Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / sctp / sm_make_chunk.c
diff --git a/kernel/net/sctp/sm_make_chunk.c b/kernel/net/sctp/sm_make_chunk.c
new file mode 100644 (file)
index 0000000..06320c8
--- /dev/null
@@ -0,0 +1,3501 @@
+/* SCTP kernel implementation
+ * (C) Copyright IBM Corp. 2001, 2004
+ * Copyright (c) 1999-2000 Cisco, Inc.
+ * Copyright (c) 1999-2001 Motorola, Inc.
+ * Copyright (c) 2001-2002 Intel Corp.
+ *
+ * This file is part of the SCTP kernel implementation
+ *
+ * These functions work with the state functions in sctp_sm_statefuns.c
+ * to implement the state operations.  These functions implement the
+ * steps which require modifying existing data structures.
+ *
+ * This SCTP implementation is free software;
+ * you can redistribute it and/or modify it under the terms of
+ * the GNU General Public License as published by
+ * the Free Software Foundation; either version 2, or (at your option)
+ * any later version.
+ *
+ * This SCTP implementation is distributed in the hope that it
+ * will be useful, but WITHOUT ANY WARRANTY; without even the implied
+ *                 ************************
+ * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ * See the GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with GNU CC; see the file COPYING.  If not, see
+ * <http://www.gnu.org/licenses/>.
+ *
+ * Please send any bug reports or fixes you make to the
+ * email address(es):
+ *    lksctp developers <linux-sctp@vger.kernel.org>
+ *
+ * Written or modified by:
+ *    La Monte H.P. Yarroll <piggy@acm.org>
+ *    Karl Knutson          <karl@athena.chicago.il.us>
+ *    C. Robin              <chris@hundredacre.ac.uk>
+ *    Jon Grimm             <jgrimm@us.ibm.com>
+ *    Xingang Guo           <xingang.guo@intel.com>
+ *    Dajiang Zhang        <dajiang.zhang@nokia.com>
+ *    Sridhar Samudrala            <sri@us.ibm.com>
+ *    Daisy Chang          <daisyc@us.ibm.com>
+ *    Ardelle Fan          <ardelle.fan@intel.com>
+ *    Kevin Gao             <kevin.gao@intel.com>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/net.h>
+#include <linux/inet.h>
+#include <linux/scatterlist.h>
+#include <linux/crypto.h>
+#include <linux/slab.h>
+#include <net/sock.h>
+
+#include <linux/skbuff.h>
+#include <linux/random.h>      /* for get_random_bytes */
+#include <net/sctp/sctp.h>
+#include <net/sctp/sm.h>
+
+static struct sctp_chunk *sctp_make_control(const struct sctp_association *asoc,
+                                           __u8 type, __u8 flags, int paylen);
+static struct sctp_chunk *sctp_make_data(const struct sctp_association *asoc,
+                                        __u8 flags, int paylen);
+static struct sctp_chunk *_sctp_make_chunk(const struct sctp_association *asoc,
+                                          __u8 type, __u8 flags, int paylen);
+static sctp_cookie_param_t *sctp_pack_cookie(const struct sctp_endpoint *ep,
+                                       const struct sctp_association *asoc,
+                                       const struct sctp_chunk *init_chunk,
+                                       int *cookie_len,
+                                       const __u8 *raw_addrs, int addrs_len);
+static int sctp_process_param(struct sctp_association *asoc,
+                             union sctp_params param,
+                             const union sctp_addr *peer_addr,
+                             gfp_t gfp);
+static void *sctp_addto_param(struct sctp_chunk *chunk, int len,
+                             const void *data);
+static void  *sctp_addto_chunk_fixed(struct sctp_chunk *, int len,
+                                    const void *data);
+
+/* Control chunk destructor */
+static void sctp_control_release_owner(struct sk_buff *skb)
+{
+       /*TODO: do memory release */
+}
+
+static void sctp_control_set_owner_w(struct sctp_chunk *chunk)
+{
+       struct sctp_association *asoc = chunk->asoc;
+       struct sk_buff *skb = chunk->skb;
+
+       /* TODO: properly account for control chunks.
+        * To do it right we'll need:
+        *  1) endpoint if association isn't known.
+        *  2) proper memory accounting.
+        *
+        *  For now don't do anything for now.
+        */
+       skb->sk = asoc ? asoc->base.sk : NULL;
+       skb->destructor = sctp_control_release_owner;
+}
+
+/* What was the inbound interface for this chunk? */
+int sctp_chunk_iif(const struct sctp_chunk *chunk)
+{
+       struct sctp_af *af;
+       int iif = 0;
+
+       af = sctp_get_af_specific(ipver2af(ip_hdr(chunk->skb)->version));
+       if (af)
+               iif = af->skb_iif(chunk->skb);
+
+       return iif;
+}
+
+/* RFC 2960 3.3.2 Initiation (INIT) (1)
+ *
+ * Note 2: The ECN capable field is reserved for future use of
+ * Explicit Congestion Notification.
+ */
+static const struct sctp_paramhdr ecap_param = {
+       SCTP_PARAM_ECN_CAPABLE,
+       cpu_to_be16(sizeof(struct sctp_paramhdr)),
+};
+static const struct sctp_paramhdr prsctp_param = {
+       SCTP_PARAM_FWD_TSN_SUPPORT,
+       cpu_to_be16(sizeof(struct sctp_paramhdr)),
+};
+
+/* A helper to initialize an op error inside a
+ * provided chunk, as most cause codes will be embedded inside an
+ * abort chunk.
+ */
+void  sctp_init_cause(struct sctp_chunk *chunk, __be16 cause_code,
+                     size_t paylen)
+{
+       sctp_errhdr_t err;
+       __u16 len;
+
+       /* Cause code constants are now defined in network order.  */
+       err.cause = cause_code;
+       len = sizeof(sctp_errhdr_t) + paylen;
+       err.length  = htons(len);
+       chunk->subh.err_hdr = sctp_addto_chunk(chunk, sizeof(sctp_errhdr_t), &err);
+}
+
+/* A helper to initialize an op error inside a
+ * provided chunk, as most cause codes will be embedded inside an
+ * abort chunk.  Differs from sctp_init_cause in that it won't oops
+ * if there isn't enough space in the op error chunk
+ */
+static int sctp_init_cause_fixed(struct sctp_chunk *chunk, __be16 cause_code,
+                     size_t paylen)
+{
+       sctp_errhdr_t err;
+       __u16 len;
+
+       /* Cause code constants are now defined in network order.  */
+       err.cause = cause_code;
+       len = sizeof(sctp_errhdr_t) + paylen;
+       err.length  = htons(len);
+
+       if (skb_tailroom(chunk->skb) < len)
+               return -ENOSPC;
+       chunk->subh.err_hdr = sctp_addto_chunk_fixed(chunk,
+                                                    sizeof(sctp_errhdr_t),
+                                                    &err);
+       return 0;
+}
+/* 3.3.2 Initiation (INIT) (1)
+ *
+ * This chunk is used to initiate a SCTP association between two
+ * endpoints. The format of the INIT chunk is shown below:
+ *
+ *     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
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    |   Type = 1    |  Chunk Flags  |      Chunk Length             |
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    |                         Initiate Tag                          |
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    |           Advertised Receiver Window Credit (a_rwnd)          |
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    |  Number of Outbound Streams   |  Number of Inbound Streams    |
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    |                          Initial TSN                          |
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    \                                                               \
+ *    /              Optional/Variable-Length Parameters              /
+ *    \                                                               \
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ *
+ * The INIT chunk contains the following parameters. Unless otherwise
+ * noted, each parameter MUST only be included once in the INIT chunk.
+ *
+ * Fixed Parameters                     Status
+ * ----------------------------------------------
+ * Initiate Tag                        Mandatory
+ * Advertised Receiver Window Credit   Mandatory
+ * Number of Outbound Streams          Mandatory
+ * Number of Inbound Streams           Mandatory
+ * Initial TSN                         Mandatory
+ *
+ * Variable Parameters                  Status     Type Value
+ * -------------------------------------------------------------
+ * IPv4 Address (Note 1)               Optional    5
+ * IPv6 Address (Note 1)               Optional    6
+ * Cookie Preservative                 Optional    9
+ * Reserved for ECN Capable (Note 2)   Optional    32768 (0x8000)
+ * Host Name Address (Note 3)          Optional    11
+ * Supported Address Types (Note 4)    Optional    12
+ */
+struct sctp_chunk *sctp_make_init(const struct sctp_association *asoc,
+                            const struct sctp_bind_addr *bp,
+                            gfp_t gfp, int vparam_len)
+{
+       struct net *net = sock_net(asoc->base.sk);
+       struct sctp_endpoint *ep = asoc->ep;
+       sctp_inithdr_t init;
+       union sctp_params addrs;
+       size_t chunksize;
+       struct sctp_chunk *retval = NULL;
+       int num_types, addrs_len = 0;
+       struct sctp_sock *sp;
+       sctp_supported_addrs_param_t sat;
+       __be16 types[2];
+       sctp_adaptation_ind_param_t aiparam;
+       sctp_supported_ext_param_t ext_param;
+       int num_ext = 0;
+       __u8 extensions[3];
+       sctp_paramhdr_t *auth_chunks = NULL,
+                       *auth_hmacs = NULL;
+
+       /* RFC 2960 3.3.2 Initiation (INIT) (1)
+        *
+        * Note 1: The INIT chunks can contain multiple addresses that
+        * can be IPv4 and/or IPv6 in any combination.
+        */
+       retval = NULL;
+
+       /* Convert the provided bind address list to raw format. */
+       addrs = sctp_bind_addrs_to_raw(bp, &addrs_len, gfp);
+
+       init.init_tag              = htonl(asoc->c.my_vtag);
+       init.a_rwnd                = htonl(asoc->rwnd);
+       init.num_outbound_streams  = htons(asoc->c.sinit_num_ostreams);
+       init.num_inbound_streams   = htons(asoc->c.sinit_max_instreams);
+       init.initial_tsn           = htonl(asoc->c.initial_tsn);
+
+       /* How many address types are needed? */
+       sp = sctp_sk(asoc->base.sk);
+       num_types = sp->pf->supported_addrs(sp, types);
+
+       chunksize = sizeof(init) + addrs_len;
+       chunksize += WORD_ROUND(SCTP_SAT_LEN(num_types));
+       chunksize += sizeof(ecap_param);
+
+       if (net->sctp.prsctp_enable)
+               chunksize += sizeof(prsctp_param);
+
+       /* ADDIP: Section 4.2.7:
+        *  An implementation supporting this extension [ADDIP] MUST list
+        *  the ASCONF,the ASCONF-ACK, and the AUTH  chunks in its INIT and
+        *  INIT-ACK parameters.
+        */
+       if (net->sctp.addip_enable) {
+               extensions[num_ext] = SCTP_CID_ASCONF;
+               extensions[num_ext+1] = SCTP_CID_ASCONF_ACK;
+               num_ext += 2;
+       }
+
+       if (sp->adaptation_ind)
+               chunksize += sizeof(aiparam);
+
+       chunksize += vparam_len;
+
+       /* Account for AUTH related parameters */
+       if (ep->auth_enable) {
+               /* Add random parameter length*/
+               chunksize += sizeof(asoc->c.auth_random);
+
+               /* Add HMACS parameter length if any were defined */
+               auth_hmacs = (sctp_paramhdr_t *)asoc->c.auth_hmacs;
+               if (auth_hmacs->length)
+                       chunksize += WORD_ROUND(ntohs(auth_hmacs->length));
+               else
+                       auth_hmacs = NULL;
+
+               /* Add CHUNKS parameter length */
+               auth_chunks = (sctp_paramhdr_t *)asoc->c.auth_chunks;
+               if (auth_chunks->length)
+                       chunksize += WORD_ROUND(ntohs(auth_chunks->length));
+               else
+                       auth_chunks = NULL;
+
+               extensions[num_ext] = SCTP_CID_AUTH;
+               num_ext += 1;
+       }
+
+       /* If we have any extensions to report, account for that */
+       if (num_ext)
+               chunksize += WORD_ROUND(sizeof(sctp_supported_ext_param_t) +
+                                       num_ext);
+
+       /* RFC 2960 3.3.2 Initiation (INIT) (1)
+        *
+        * Note 3: An INIT chunk MUST NOT contain more than one Host
+        * Name address parameter. Moreover, the sender of the INIT
+        * MUST NOT combine any other address types with the Host Name
+        * address in the INIT. The receiver of INIT MUST ignore any
+        * other address types if the Host Name address parameter is
+        * present in the received INIT chunk.
+        *
+        * PLEASE DO NOT FIXME [This version does not support Host Name.]
+        */
+
+       retval = sctp_make_control(asoc, SCTP_CID_INIT, 0, chunksize);
+       if (!retval)
+               goto nodata;
+
+       retval->subh.init_hdr =
+               sctp_addto_chunk(retval, sizeof(init), &init);
+       retval->param_hdr.v =
+               sctp_addto_chunk(retval, addrs_len, addrs.v);
+
+       /* RFC 2960 3.3.2 Initiation (INIT) (1)
+        *
+        * Note 4: This parameter, when present, specifies all the
+        * address types the sending endpoint can support. The absence
+        * of this parameter indicates that the sending endpoint can
+        * support any address type.
+        */
+       sat.param_hdr.type = SCTP_PARAM_SUPPORTED_ADDRESS_TYPES;
+       sat.param_hdr.length = htons(SCTP_SAT_LEN(num_types));
+       sctp_addto_chunk(retval, sizeof(sat), &sat);
+       sctp_addto_chunk(retval, num_types * sizeof(__u16), &types);
+
+       sctp_addto_chunk(retval, sizeof(ecap_param), &ecap_param);
+
+       /* Add the supported extensions parameter.  Be nice and add this
+        * fist before addiding the parameters for the extensions themselves
+        */
+       if (num_ext) {
+               ext_param.param_hdr.type = SCTP_PARAM_SUPPORTED_EXT;
+               ext_param.param_hdr.length =
+                           htons(sizeof(sctp_supported_ext_param_t) + num_ext);
+               sctp_addto_chunk(retval, sizeof(sctp_supported_ext_param_t),
+                               &ext_param);
+               sctp_addto_param(retval, num_ext, extensions);
+       }
+
+       if (net->sctp.prsctp_enable)
+               sctp_addto_chunk(retval, sizeof(prsctp_param), &prsctp_param);
+
+       if (sp->adaptation_ind) {
+               aiparam.param_hdr.type = SCTP_PARAM_ADAPTATION_LAYER_IND;
+               aiparam.param_hdr.length = htons(sizeof(aiparam));
+               aiparam.adaptation_ind = htonl(sp->adaptation_ind);
+               sctp_addto_chunk(retval, sizeof(aiparam), &aiparam);
+       }
+
+       /* Add SCTP-AUTH chunks to the parameter list */
+       if (ep->auth_enable) {
+               sctp_addto_chunk(retval, sizeof(asoc->c.auth_random),
+                                asoc->c.auth_random);
+               if (auth_hmacs)
+                       sctp_addto_chunk(retval, ntohs(auth_hmacs->length),
+                                       auth_hmacs);
+               if (auth_chunks)
+                       sctp_addto_chunk(retval, ntohs(auth_chunks->length),
+                                       auth_chunks);
+       }
+nodata:
+       kfree(addrs.v);
+       return retval;
+}
+
+struct sctp_chunk *sctp_make_init_ack(const struct sctp_association *asoc,
+                                const struct sctp_chunk *chunk,
+                                gfp_t gfp, int unkparam_len)
+{
+       sctp_inithdr_t initack;
+       struct sctp_chunk *retval;
+       union sctp_params addrs;
+       struct sctp_sock *sp;
+       int addrs_len;
+       sctp_cookie_param_t *cookie;
+       int cookie_len;
+       size_t chunksize;
+       sctp_adaptation_ind_param_t aiparam;
+       sctp_supported_ext_param_t ext_param;
+       int num_ext = 0;
+       __u8 extensions[3];
+       sctp_paramhdr_t *auth_chunks = NULL,
+                       *auth_hmacs = NULL,
+                       *auth_random = NULL;
+
+       retval = NULL;
+
+       /* Note: there may be no addresses to embed. */
+       addrs = sctp_bind_addrs_to_raw(&asoc->base.bind_addr, &addrs_len, gfp);
+
+       initack.init_tag                = htonl(asoc->c.my_vtag);
+       initack.a_rwnd                  = htonl(asoc->rwnd);
+       initack.num_outbound_streams    = htons(asoc->c.sinit_num_ostreams);
+       initack.num_inbound_streams     = htons(asoc->c.sinit_max_instreams);
+       initack.initial_tsn             = htonl(asoc->c.initial_tsn);
+
+       /* FIXME:  We really ought to build the cookie right
+        * into the packet instead of allocating more fresh memory.
+        */
+       cookie = sctp_pack_cookie(asoc->ep, asoc, chunk, &cookie_len,
+                                 addrs.v, addrs_len);
+       if (!cookie)
+               goto nomem_cookie;
+
+       /* Calculate the total size of allocation, include the reserved
+        * space for reporting unknown parameters if it is specified.
+        */
+       sp = sctp_sk(asoc->base.sk);
+       chunksize = sizeof(initack) + addrs_len + cookie_len + unkparam_len;
+
+       /* Tell peer that we'll do ECN only if peer advertised such cap.  */
+       if (asoc->peer.ecn_capable)
+               chunksize += sizeof(ecap_param);
+
+       if (asoc->peer.prsctp_capable)
+               chunksize += sizeof(prsctp_param);
+
+       if (asoc->peer.asconf_capable) {
+               extensions[num_ext] = SCTP_CID_ASCONF;
+               extensions[num_ext+1] = SCTP_CID_ASCONF_ACK;
+               num_ext += 2;
+       }
+
+       if (sp->adaptation_ind)
+               chunksize += sizeof(aiparam);
+
+       if (asoc->peer.auth_capable) {
+               auth_random = (sctp_paramhdr_t *)asoc->c.auth_random;
+               chunksize += ntohs(auth_random->length);
+
+               auth_hmacs = (sctp_paramhdr_t *)asoc->c.auth_hmacs;
+               if (auth_hmacs->length)
+                       chunksize += WORD_ROUND(ntohs(auth_hmacs->length));
+               else
+                       auth_hmacs = NULL;
+
+               auth_chunks = (sctp_paramhdr_t *)asoc->c.auth_chunks;
+               if (auth_chunks->length)
+                       chunksize += WORD_ROUND(ntohs(auth_chunks->length));
+               else
+                       auth_chunks = NULL;
+
+               extensions[num_ext] = SCTP_CID_AUTH;
+               num_ext += 1;
+       }
+
+       if (num_ext)
+               chunksize += WORD_ROUND(sizeof(sctp_supported_ext_param_t) +
+                                       num_ext);
+
+       /* Now allocate and fill out the chunk.  */
+       retval = sctp_make_control(asoc, SCTP_CID_INIT_ACK, 0, chunksize);
+       if (!retval)
+               goto nomem_chunk;
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [INIT ACK back to where the INIT came from.]
+        */
+       retval->transport = chunk->transport;
+
+       retval->subh.init_hdr =
+               sctp_addto_chunk(retval, sizeof(initack), &initack);
+       retval->param_hdr.v = sctp_addto_chunk(retval, addrs_len, addrs.v);
+       sctp_addto_chunk(retval, cookie_len, cookie);
+       if (asoc->peer.ecn_capable)
+               sctp_addto_chunk(retval, sizeof(ecap_param), &ecap_param);
+       if (num_ext) {
+               ext_param.param_hdr.type = SCTP_PARAM_SUPPORTED_EXT;
+               ext_param.param_hdr.length =
+                           htons(sizeof(sctp_supported_ext_param_t) + num_ext);
+               sctp_addto_chunk(retval, sizeof(sctp_supported_ext_param_t),
+                                &ext_param);
+               sctp_addto_param(retval, num_ext, extensions);
+       }
+       if (asoc->peer.prsctp_capable)
+               sctp_addto_chunk(retval, sizeof(prsctp_param), &prsctp_param);
+
+       if (sp->adaptation_ind) {
+               aiparam.param_hdr.type = SCTP_PARAM_ADAPTATION_LAYER_IND;
+               aiparam.param_hdr.length = htons(sizeof(aiparam));
+               aiparam.adaptation_ind = htonl(sp->adaptation_ind);
+               sctp_addto_chunk(retval, sizeof(aiparam), &aiparam);
+       }
+
+       if (asoc->peer.auth_capable) {
+               sctp_addto_chunk(retval, ntohs(auth_random->length),
+                                auth_random);
+               if (auth_hmacs)
+                       sctp_addto_chunk(retval, ntohs(auth_hmacs->length),
+                                       auth_hmacs);
+               if (auth_chunks)
+                       sctp_addto_chunk(retval, ntohs(auth_chunks->length),
+                                       auth_chunks);
+       }
+
+       /* We need to remove the const qualifier at this point.  */
+       retval->asoc = (struct sctp_association *) asoc;
+
+nomem_chunk:
+       kfree(cookie);
+nomem_cookie:
+       kfree(addrs.v);
+       return retval;
+}
+
+/* 3.3.11 Cookie Echo (COOKIE ECHO) (10):
+ *
+ * This chunk is used only during the initialization of an association.
+ * It is sent by the initiator of an association to its peer to complete
+ * the initialization process. This chunk MUST precede any DATA chunk
+ * sent within the association, but MAY be bundled with one or more DATA
+ * chunks in the same packet.
+ *
+ *      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
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |   Type = 10   |Chunk  Flags   |         Length                |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     /                     Cookie                                    /
+ *     \                                                               \
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * Chunk Flags: 8 bit
+ *
+ *   Set to zero on transmit and ignored on receipt.
+ *
+ * Length: 16 bits (unsigned integer)
+ *
+ *   Set to the size of the chunk in bytes, including the 4 bytes of
+ *   the chunk header and the size of the Cookie.
+ *
+ * Cookie: variable size
+ *
+ *   This field must contain the exact cookie received in the
+ *   State Cookie parameter from the previous INIT ACK.
+ *
+ *   An implementation SHOULD make the cookie as small as possible
+ *   to insure interoperability.
+ */
+struct sctp_chunk *sctp_make_cookie_echo(const struct sctp_association *asoc,
+                                   const struct sctp_chunk *chunk)
+{
+       struct sctp_chunk *retval;
+       void *cookie;
+       int cookie_len;
+
+       cookie = asoc->peer.cookie;
+       cookie_len = asoc->peer.cookie_len;
+
+       /* Build a cookie echo chunk.  */
+       retval = sctp_make_control(asoc, SCTP_CID_COOKIE_ECHO, 0, cookie_len);
+       if (!retval)
+               goto nodata;
+       retval->subh.cookie_hdr =
+               sctp_addto_chunk(retval, cookie_len, cookie);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [COOKIE ECHO back to where the INIT ACK came from.]
+        */
+       if (chunk)
+               retval->transport = chunk->transport;
+
+nodata:
+       return retval;
+}
+
+/* 3.3.12 Cookie Acknowledgement (COOKIE ACK) (11):
+ *
+ * This chunk is used only during the initialization of an
+ * association.  It is used to acknowledge the receipt of a COOKIE
+ * ECHO chunk.  This chunk MUST precede any DATA or SACK chunk sent
+ * within the association, but MAY be bundled with one or more DATA
+ * chunks or SACK chunk in the same SCTP packet.
+ *
+ *      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
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |   Type = 11   |Chunk  Flags   |     Length = 4                |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * Chunk Flags: 8 bits
+ *
+ *   Set to zero on transmit and ignored on receipt.
+ */
+struct sctp_chunk *sctp_make_cookie_ack(const struct sctp_association *asoc,
+                                  const struct sctp_chunk *chunk)
+{
+       struct sctp_chunk *retval;
+
+       retval = sctp_make_control(asoc, SCTP_CID_COOKIE_ACK, 0, 0);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [COOKIE ACK back to where the COOKIE ECHO came from.]
+        */
+       if (retval && chunk)
+               retval->transport = chunk->transport;
+
+       return retval;
+}
+
+/*
+ *  Appendix A: Explicit Congestion Notification:
+ *  CWR:
+ *
+ *  RFC 2481 details a specific bit for a sender to send in the header of
+ *  its next outbound TCP segment to indicate to its peer that it has
+ *  reduced its congestion window.  This is termed the CWR bit.  For
+ *  SCTP the same indication is made by including the CWR chunk.
+ *  This chunk contains one data element, i.e. the TSN number that
+ *  was sent in the ECNE chunk.  This element represents the lowest
+ *  TSN number in the datagram that was originally marked with the
+ *  CE bit.
+ *
+ *     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
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    | Chunk Type=13 | Flags=00000000|    Chunk Length = 8           |
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *    |                      Lowest TSN Number                        |
+ *    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ *     Note: The CWR is considered a Control chunk.
+ */
+struct sctp_chunk *sctp_make_cwr(const struct sctp_association *asoc,
+                           const __u32 lowest_tsn,
+                           const struct sctp_chunk *chunk)
+{
+       struct sctp_chunk *retval;
+       sctp_cwrhdr_t cwr;
+
+       cwr.lowest_tsn = htonl(lowest_tsn);
+       retval = sctp_make_control(asoc, SCTP_CID_ECN_CWR, 0,
+                                  sizeof(sctp_cwrhdr_t));
+
+       if (!retval)
+               goto nodata;
+
+       retval->subh.ecn_cwr_hdr =
+               sctp_addto_chunk(retval, sizeof(cwr), &cwr);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [Report a reduced congestion window back to where the ECNE
+        * came from.]
+        */
+       if (chunk)
+               retval->transport = chunk->transport;
+
+nodata:
+       return retval;
+}
+
+/* Make an ECNE chunk.  This is a congestion experienced report.  */
+struct sctp_chunk *sctp_make_ecne(const struct sctp_association *asoc,
+                            const __u32 lowest_tsn)
+{
+       struct sctp_chunk *retval;
+       sctp_ecnehdr_t ecne;
+
+       ecne.lowest_tsn = htonl(lowest_tsn);
+       retval = sctp_make_control(asoc, SCTP_CID_ECN_ECNE, 0,
+                                  sizeof(sctp_ecnehdr_t));
+       if (!retval)
+               goto nodata;
+       retval->subh.ecne_hdr =
+               sctp_addto_chunk(retval, sizeof(ecne), &ecne);
+
+nodata:
+       return retval;
+}
+
+/* Make a DATA chunk for the given association from the provided
+ * parameters.  However, do not populate the data payload.
+ */
+struct sctp_chunk *sctp_make_datafrag_empty(struct sctp_association *asoc,
+                                      const struct sctp_sndrcvinfo *sinfo,
+                                      int data_len, __u8 flags, __u16 ssn)
+{
+       struct sctp_chunk *retval;
+       struct sctp_datahdr dp;
+       int chunk_len;
+
+       /* We assign the TSN as LATE as possible, not here when
+        * creating the chunk.
+        */
+       dp.tsn = 0;
+       dp.stream = htons(sinfo->sinfo_stream);
+       dp.ppid   = sinfo->sinfo_ppid;
+
+       /* Set the flags for an unordered send.  */
+       if (sinfo->sinfo_flags & SCTP_UNORDERED) {
+               flags |= SCTP_DATA_UNORDERED;
+               dp.ssn = 0;
+       } else
+               dp.ssn = htons(ssn);
+
+       chunk_len = sizeof(dp) + data_len;
+       retval = sctp_make_data(asoc, flags, chunk_len);
+       if (!retval)
+               goto nodata;
+
+       retval->subh.data_hdr = sctp_addto_chunk(retval, sizeof(dp), &dp);
+       memcpy(&retval->sinfo, sinfo, sizeof(struct sctp_sndrcvinfo));
+
+nodata:
+       return retval;
+}
+
+/* Create a selective ackowledgement (SACK) for the given
+ * association.  This reports on which TSN's we've seen to date,
+ * including duplicates and gaps.
+ */
+struct sctp_chunk *sctp_make_sack(const struct sctp_association *asoc)
+{
+       struct sctp_chunk *retval;
+       struct sctp_sackhdr sack;
+       int len;
+       __u32 ctsn;
+       __u16 num_gabs, num_dup_tsns;
+       struct sctp_association *aptr = (struct sctp_association *)asoc;
+       struct sctp_tsnmap *map = (struct sctp_tsnmap *)&asoc->peer.tsn_map;
+       struct sctp_gap_ack_block gabs[SCTP_MAX_GABS];
+       struct sctp_transport *trans;
+
+       memset(gabs, 0, sizeof(gabs));
+       ctsn = sctp_tsnmap_get_ctsn(map);
+
+       pr_debug("%s: sackCTSNAck sent:0x%x\n", __func__, ctsn);
+
+       /* How much room is needed in the chunk? */
+       num_gabs = sctp_tsnmap_num_gabs(map, gabs);
+       num_dup_tsns = sctp_tsnmap_num_dups(map);
+
+       /* Initialize the SACK header.  */
+       sack.cum_tsn_ack            = htonl(ctsn);
+       sack.a_rwnd                 = htonl(asoc->a_rwnd);
+       sack.num_gap_ack_blocks     = htons(num_gabs);
+       sack.num_dup_tsns           = htons(num_dup_tsns);
+
+       len = sizeof(sack)
+               + sizeof(struct sctp_gap_ack_block) * num_gabs
+               + sizeof(__u32) * num_dup_tsns;
+
+       /* Create the chunk.  */
+       retval = sctp_make_control(asoc, SCTP_CID_SACK, 0, len);
+       if (!retval)
+               goto nodata;
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, etc.) to the same destination transport
+        * address from which it received the DATA or control chunk to
+        * which it is replying.  This rule should also be followed if
+        * the endpoint is bundling DATA chunks together with the
+        * reply chunk.
+        *
+        * However, when acknowledging multiple DATA chunks received
+        * in packets from different source addresses in a single
+        * SACK, the SACK chunk may be transmitted to one of the
+        * destination transport addresses from which the DATA or
+        * control chunks being acknowledged were received.
+        *
+        * [BUG:  We do not implement the following paragraph.
+        * Perhaps we should remember the last transport we used for a
+        * SACK and avoid that (if possible) if we have seen any
+        * duplicates. --piggy]
+        *
+        * When a receiver of a duplicate DATA chunk sends a SACK to a
+        * multi- homed endpoint it MAY be beneficial to vary the
+        * destination address and not use the source address of the
+        * DATA chunk.  The reason being that receiving a duplicate
+        * from a multi-homed endpoint might indicate that the return
+        * path (as specified in the source address of the DATA chunk)
+        * for the SACK is broken.
+        *
+        * [Send to the address from which we last received a DATA chunk.]
+        */
+       retval->transport = asoc->peer.last_data_from;
+
+       retval->subh.sack_hdr =
+               sctp_addto_chunk(retval, sizeof(sack), &sack);
+
+       /* Add the gap ack block information.   */
+       if (num_gabs)
+               sctp_addto_chunk(retval, sizeof(__u32) * num_gabs,
+                                gabs);
+
+       /* Add the duplicate TSN information.  */
+       if (num_dup_tsns) {
+               aptr->stats.idupchunks += num_dup_tsns;
+               sctp_addto_chunk(retval, sizeof(__u32) * num_dup_tsns,
+                                sctp_tsnmap_get_dups(map));
+       }
+       /* Once we have a sack generated, check to see what our sack
+        * generation is, if its 0, reset the transports to 0, and reset
+        * the association generation to 1
+        *
+        * The idea is that zero is never used as a valid generation for the
+        * association so no transport will match after a wrap event like this,
+        * Until the next sack
+        */
+       if (++aptr->peer.sack_generation == 0) {
+               list_for_each_entry(trans, &asoc->peer.transport_addr_list,
+                                   transports)
+                       trans->sack_generation = 0;
+               aptr->peer.sack_generation = 1;
+       }
+nodata:
+       return retval;
+}
+
+/* Make a SHUTDOWN chunk. */
+struct sctp_chunk *sctp_make_shutdown(const struct sctp_association *asoc,
+                                     const struct sctp_chunk *chunk)
+{
+       struct sctp_chunk *retval;
+       sctp_shutdownhdr_t shut;
+       __u32 ctsn;
+
+       ctsn = sctp_tsnmap_get_ctsn(&asoc->peer.tsn_map);
+       shut.cum_tsn_ack = htonl(ctsn);
+
+       retval = sctp_make_control(asoc, SCTP_CID_SHUTDOWN, 0,
+                                  sizeof(sctp_shutdownhdr_t));
+       if (!retval)
+               goto nodata;
+
+       retval->subh.shutdown_hdr =
+               sctp_addto_chunk(retval, sizeof(shut), &shut);
+
+       if (chunk)
+               retval->transport = chunk->transport;
+nodata:
+       return retval;
+}
+
+struct sctp_chunk *sctp_make_shutdown_ack(const struct sctp_association *asoc,
+                                    const struct sctp_chunk *chunk)
+{
+       struct sctp_chunk *retval;
+
+       retval = sctp_make_control(asoc, SCTP_CID_SHUTDOWN_ACK, 0, 0);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [ACK back to where the SHUTDOWN came from.]
+        */
+       if (retval && chunk)
+               retval->transport = chunk->transport;
+
+       return retval;
+}
+
+struct sctp_chunk *sctp_make_shutdown_complete(
+       const struct sctp_association *asoc,
+       const struct sctp_chunk *chunk)
+{
+       struct sctp_chunk *retval;
+       __u8 flags = 0;
+
+       /* Set the T-bit if we have no association (vtag will be
+        * reflected)
+        */
+       flags |= asoc ? 0 : SCTP_CHUNK_FLAG_T;
+
+       retval = sctp_make_control(asoc, SCTP_CID_SHUTDOWN_COMPLETE, flags, 0);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [Report SHUTDOWN COMPLETE back to where the SHUTDOWN ACK
+        * came from.]
+        */
+       if (retval && chunk)
+               retval->transport = chunk->transport;
+
+       return retval;
+}
+
+/* Create an ABORT.  Note that we set the T bit if we have no
+ * association, except when responding to an INIT (sctpimpguide 2.41).
+ */
+struct sctp_chunk *sctp_make_abort(const struct sctp_association *asoc,
+                             const struct sctp_chunk *chunk,
+                             const size_t hint)
+{
+       struct sctp_chunk *retval;
+       __u8 flags = 0;
+
+       /* Set the T-bit if we have no association and 'chunk' is not
+        * an INIT (vtag will be reflected).
+        */
+       if (!asoc) {
+               if (chunk && chunk->chunk_hdr &&
+                   chunk->chunk_hdr->type == SCTP_CID_INIT)
+                       flags = 0;
+               else
+                       flags = SCTP_CHUNK_FLAG_T;
+       }
+
+       retval = sctp_make_control(asoc, SCTP_CID_ABORT, flags, hint);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [ABORT back to where the offender came from.]
+        */
+       if (retval && chunk)
+               retval->transport = chunk->transport;
+
+       return retval;
+}
+
+/* Helper to create ABORT with a NO_USER_DATA error.  */
+struct sctp_chunk *sctp_make_abort_no_data(
+       const struct sctp_association *asoc,
+       const struct sctp_chunk *chunk, __u32 tsn)
+{
+       struct sctp_chunk *retval;
+       __be32 payload;
+
+       retval = sctp_make_abort(asoc, chunk, sizeof(sctp_errhdr_t)
+                                + sizeof(tsn));
+
+       if (!retval)
+               goto no_mem;
+
+       /* Put the tsn back into network byte order.  */
+       payload = htonl(tsn);
+       sctp_init_cause(retval, SCTP_ERROR_NO_DATA, sizeof(payload));
+       sctp_addto_chunk(retval, sizeof(payload), (const void *)&payload);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [ABORT back to where the offender came from.]
+        */
+       if (chunk)
+               retval->transport = chunk->transport;
+
+no_mem:
+       return retval;
+}
+
+/* Helper to create ABORT with a SCTP_ERROR_USER_ABORT error.  */
+struct sctp_chunk *sctp_make_abort_user(const struct sctp_association *asoc,
+                                       struct msghdr *msg,
+                                       size_t paylen)
+{
+       struct sctp_chunk *retval;
+       void *payload = NULL;
+       int err;
+
+       retval = sctp_make_abort(asoc, NULL, sizeof(sctp_errhdr_t) + paylen);
+       if (!retval)
+               goto err_chunk;
+
+       if (paylen) {
+               /* Put the msg_iov together into payload.  */
+               payload = kmalloc(paylen, GFP_KERNEL);
+               if (!payload)
+                       goto err_payload;
+
+               err = memcpy_from_msg(payload, msg, paylen);
+               if (err < 0)
+                       goto err_copy;
+       }
+
+       sctp_init_cause(retval, SCTP_ERROR_USER_ABORT, paylen);
+       sctp_addto_chunk(retval, paylen, payload);
+
+       if (paylen)
+               kfree(payload);
+
+       return retval;
+
+err_copy:
+       kfree(payload);
+err_payload:
+       sctp_chunk_free(retval);
+       retval = NULL;
+err_chunk:
+       return retval;
+}
+
+/* Append bytes to the end of a parameter.  Will panic if chunk is not big
+ * enough.
+ */
+static void *sctp_addto_param(struct sctp_chunk *chunk, int len,
+                             const void *data)
+{
+       void *target;
+       int chunklen = ntohs(chunk->chunk_hdr->length);
+
+       target = skb_put(chunk->skb, len);
+
+       if (data)
+               memcpy(target, data, len);
+       else
+               memset(target, 0, len);
+
+       /* Adjust the chunk length field.  */
+       chunk->chunk_hdr->length = htons(chunklen + len);
+       chunk->chunk_end = skb_tail_pointer(chunk->skb);
+
+       return target;
+}
+
+/* Make an ABORT chunk with a PROTOCOL VIOLATION cause code. */
+struct sctp_chunk *sctp_make_abort_violation(
+       const struct sctp_association *asoc,
+       const struct sctp_chunk *chunk,
+       const __u8   *payload,
+       const size_t paylen)
+{
+       struct sctp_chunk  *retval;
+       struct sctp_paramhdr phdr;
+
+       retval = sctp_make_abort(asoc, chunk, sizeof(sctp_errhdr_t) + paylen
+                                       + sizeof(sctp_paramhdr_t));
+       if (!retval)
+               goto end;
+
+       sctp_init_cause(retval, SCTP_ERROR_PROTO_VIOLATION, paylen
+                                       + sizeof(sctp_paramhdr_t));
+
+       phdr.type = htons(chunk->chunk_hdr->type);
+       phdr.length = chunk->chunk_hdr->length;
+       sctp_addto_chunk(retval, paylen, payload);
+       sctp_addto_param(retval, sizeof(sctp_paramhdr_t), &phdr);
+
+end:
+       return retval;
+}
+
+struct sctp_chunk *sctp_make_violation_paramlen(
+       const struct sctp_association *asoc,
+       const struct sctp_chunk *chunk,
+       struct sctp_paramhdr *param)
+{
+       struct sctp_chunk *retval;
+       static const char error[] = "The following parameter had invalid length:";
+       size_t payload_len = sizeof(error) + sizeof(sctp_errhdr_t) +
+                               sizeof(sctp_paramhdr_t);
+
+       retval = sctp_make_abort(asoc, chunk, payload_len);
+       if (!retval)
+               goto nodata;
+
+       sctp_init_cause(retval, SCTP_ERROR_PROTO_VIOLATION,
+                       sizeof(error) + sizeof(sctp_paramhdr_t));
+       sctp_addto_chunk(retval, sizeof(error), error);
+       sctp_addto_param(retval, sizeof(sctp_paramhdr_t), param);
+
+nodata:
+       return retval;
+}
+
+struct sctp_chunk *sctp_make_violation_max_retrans(
+       const struct sctp_association *asoc,
+       const struct sctp_chunk *chunk)
+{
+       struct sctp_chunk *retval;
+       static const char error[] = "Association exceeded its max_retans count";
+       size_t payload_len = sizeof(error) + sizeof(sctp_errhdr_t);
+
+       retval = sctp_make_abort(asoc, chunk, payload_len);
+       if (!retval)
+               goto nodata;
+
+       sctp_init_cause(retval, SCTP_ERROR_PROTO_VIOLATION, sizeof(error));
+       sctp_addto_chunk(retval, sizeof(error), error);
+
+nodata:
+       return retval;
+}
+
+/* Make a HEARTBEAT chunk.  */
+struct sctp_chunk *sctp_make_heartbeat(const struct sctp_association *asoc,
+                                 const struct sctp_transport *transport)
+{
+       struct sctp_chunk *retval;
+       sctp_sender_hb_info_t hbinfo;
+
+       retval = sctp_make_control(asoc, SCTP_CID_HEARTBEAT, 0, sizeof(hbinfo));
+
+       if (!retval)
+               goto nodata;
+
+       hbinfo.param_hdr.type = SCTP_PARAM_HEARTBEAT_INFO;
+       hbinfo.param_hdr.length = htons(sizeof(sctp_sender_hb_info_t));
+       hbinfo.daddr = transport->ipaddr;
+       hbinfo.sent_at = jiffies;
+       hbinfo.hb_nonce = transport->hb_nonce;
+
+       /* Cast away the 'const', as this is just telling the chunk
+        * what transport it belongs to.
+        */
+       retval->transport = (struct sctp_transport *) transport;
+       retval->subh.hbs_hdr = sctp_addto_chunk(retval, sizeof(hbinfo),
+                                               &hbinfo);
+
+nodata:
+       return retval;
+}
+
+struct sctp_chunk *sctp_make_heartbeat_ack(const struct sctp_association *asoc,
+                                     const struct sctp_chunk *chunk,
+                                     const void *payload, const size_t paylen)
+{
+       struct sctp_chunk *retval;
+
+       retval  = sctp_make_control(asoc, SCTP_CID_HEARTBEAT_ACK, 0, paylen);
+       if (!retval)
+               goto nodata;
+
+       retval->subh.hbs_hdr = sctp_addto_chunk(retval, paylen, payload);
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, * etc.) to the same destination transport
+        * address from which it * received the DATA or control chunk
+        * to which it is replying.
+        *
+        * [HBACK back to where the HEARTBEAT came from.]
+        */
+       if (chunk)
+               retval->transport = chunk->transport;
+
+nodata:
+       return retval;
+}
+
+/* Create an Operation Error chunk with the specified space reserved.
+ * This routine can be used for containing multiple causes in the chunk.
+ */
+static struct sctp_chunk *sctp_make_op_error_space(
+       const struct sctp_association *asoc,
+       const struct sctp_chunk *chunk,
+       size_t size)
+{
+       struct sctp_chunk *retval;
+
+       retval = sctp_make_control(asoc, SCTP_CID_ERROR, 0,
+                                  sizeof(sctp_errhdr_t) + size);
+       if (!retval)
+               goto nodata;
+
+       /* RFC 2960 6.4 Multi-homed SCTP Endpoints
+        *
+        * An endpoint SHOULD transmit reply chunks (e.g., SACK,
+        * HEARTBEAT ACK, etc.) to the same destination transport
+        * address from which it received the DATA or control chunk
+        * to which it is replying.
+        *
+        */
+       if (chunk)
+               retval->transport = chunk->transport;
+
+nodata:
+       return retval;
+}
+
+/* Create an Operation Error chunk of a fixed size,
+ * specifically, max(asoc->pathmtu, SCTP_DEFAULT_MAXSEGMENT)
+ * This is a helper function to allocate an error chunk for
+ * for those invalid parameter codes in which we may not want
+ * to report all the errors, if the incoming chunk is large
+ */
+static inline struct sctp_chunk *sctp_make_op_error_fixed(
+       const struct sctp_association *asoc,
+       const struct sctp_chunk *chunk)
+{
+       size_t size = asoc ? asoc->pathmtu : 0;
+
+       if (!size)
+               size = SCTP_DEFAULT_MAXSEGMENT;
+
+       return sctp_make_op_error_space(asoc, chunk, size);
+}
+
+/* Create an Operation Error chunk.  */
+struct sctp_chunk *sctp_make_op_error(const struct sctp_association *asoc,
+                                const struct sctp_chunk *chunk,
+                                __be16 cause_code, const void *payload,
+                                size_t paylen, size_t reserve_tail)
+{
+       struct sctp_chunk *retval;
+
+       retval = sctp_make_op_error_space(asoc, chunk, paylen + reserve_tail);
+       if (!retval)
+               goto nodata;
+
+       sctp_init_cause(retval, cause_code, paylen + reserve_tail);
+       sctp_addto_chunk(retval, paylen, payload);
+       if (reserve_tail)
+               sctp_addto_param(retval, reserve_tail, NULL);
+
+nodata:
+       return retval;
+}
+
+struct sctp_chunk *sctp_make_auth(const struct sctp_association *asoc)
+{
+       struct sctp_chunk *retval;
+       struct sctp_hmac *hmac_desc;
+       struct sctp_authhdr auth_hdr;
+       __u8 *hmac;
+
+       /* Get the first hmac that the peer told us to use */
+       hmac_desc = sctp_auth_asoc_get_hmac(asoc);
+       if (unlikely(!hmac_desc))
+               return NULL;
+
+       retval = sctp_make_control(asoc, SCTP_CID_AUTH, 0,
+                       hmac_desc->hmac_len + sizeof(sctp_authhdr_t));
+       if (!retval)
+               return NULL;
+
+       auth_hdr.hmac_id = htons(hmac_desc->hmac_id);
+       auth_hdr.shkey_id = htons(asoc->active_key_id);
+
+       retval->subh.auth_hdr = sctp_addto_chunk(retval, sizeof(sctp_authhdr_t),
+                                               &auth_hdr);
+
+       hmac = skb_put(retval->skb, hmac_desc->hmac_len);
+       memset(hmac, 0, hmac_desc->hmac_len);
+
+       /* Adjust the chunk header to include the empty MAC */
+       retval->chunk_hdr->length =
+               htons(ntohs(retval->chunk_hdr->length) + hmac_desc->hmac_len);
+       retval->chunk_end = skb_tail_pointer(retval->skb);
+
+       return retval;
+}
+
+
+/********************************************************************
+ * 2nd Level Abstractions
+ ********************************************************************/
+
+/* Turn an skb into a chunk.
+ * FIXME: Eventually move the structure directly inside the skb->cb[].
+ *
+ * sctpimpguide-05.txt Section 2.8.2
+ * M1) Each time a new DATA chunk is transmitted
+ * set the 'TSN.Missing.Report' count for that TSN to 0. The
+ * 'TSN.Missing.Report' count will be used to determine missing chunks
+ * and when to fast retransmit.
+ *
+ */
+struct sctp_chunk *sctp_chunkify(struct sk_buff *skb,
+                           const struct sctp_association *asoc,
+                           struct sock *sk)
+{
+       struct sctp_chunk *retval;
+
+       retval = kmem_cache_zalloc(sctp_chunk_cachep, GFP_ATOMIC);
+
+       if (!retval)
+               goto nodata;
+       if (!sk)
+               pr_debug("%s: chunkifying skb:%p w/o an sk\n", __func__, skb);
+
+       INIT_LIST_HEAD(&retval->list);
+       retval->skb             = skb;
+       retval->asoc            = (struct sctp_association *)asoc;
+       retval->singleton       = 1;
+
+       retval->fast_retransmit = SCTP_CAN_FRTX;
+
+       /* Polish the bead hole.  */
+       INIT_LIST_HEAD(&retval->transmitted_list);
+       INIT_LIST_HEAD(&retval->frag_list);
+       SCTP_DBG_OBJCNT_INC(chunk);
+       atomic_set(&retval->refcnt, 1);
+
+nodata:
+       return retval;
+}
+
+/* Set chunk->source and dest based on the IP header in chunk->skb.  */
+void sctp_init_addrs(struct sctp_chunk *chunk, union sctp_addr *src,
+                    union sctp_addr *dest)
+{
+       memcpy(&chunk->source, src, sizeof(union sctp_addr));
+       memcpy(&chunk->dest, dest, sizeof(union sctp_addr));
+}
+
+/* Extract the source address from a chunk.  */
+const union sctp_addr *sctp_source(const struct sctp_chunk *chunk)
+{
+       /* If we have a known transport, use that.  */
+       if (chunk->transport) {
+               return &chunk->transport->ipaddr;
+       } else {
+               /* Otherwise, extract it from the IP header.  */
+               return &chunk->source;
+       }
+}
+
+/* Create a new chunk, setting the type and flags headers from the
+ * arguments, reserving enough space for a 'paylen' byte payload.
+ */
+static struct sctp_chunk *_sctp_make_chunk(const struct sctp_association *asoc,
+                                           __u8 type, __u8 flags, int paylen)
+{
+       struct sctp_chunk *retval;
+       sctp_chunkhdr_t *chunk_hdr;
+       struct sk_buff *skb;
+       struct sock *sk;
+
+       /* No need to allocate LL here, as this is only a chunk. */
+       skb = alloc_skb(WORD_ROUND(sizeof(sctp_chunkhdr_t) + paylen),
+                       GFP_ATOMIC);
+       if (!skb)
+               goto nodata;
+
+       /* Make room for the chunk header.  */
+       chunk_hdr = (sctp_chunkhdr_t *)skb_put(skb, sizeof(sctp_chunkhdr_t));
+       chunk_hdr->type   = type;
+       chunk_hdr->flags  = flags;
+       chunk_hdr->length = htons(sizeof(sctp_chunkhdr_t));
+
+       sk = asoc ? asoc->base.sk : NULL;
+       retval = sctp_chunkify(skb, asoc, sk);
+       if (!retval) {
+               kfree_skb(skb);
+               goto nodata;
+       }
+
+       retval->chunk_hdr = chunk_hdr;
+       retval->chunk_end = ((__u8 *)chunk_hdr) + sizeof(struct sctp_chunkhdr);
+
+       /* Determine if the chunk needs to be authenticated */
+       if (sctp_auth_send_cid(type, asoc))
+               retval->auth = 1;
+
+       return retval;
+nodata:
+       return NULL;
+}
+
+static struct sctp_chunk *sctp_make_data(const struct sctp_association *asoc,
+                                        __u8 flags, int paylen)
+{
+       return _sctp_make_chunk(asoc, SCTP_CID_DATA, flags, paylen);
+}
+
+static struct sctp_chunk *sctp_make_control(const struct sctp_association *asoc,
+                                           __u8 type, __u8 flags, int paylen)
+{
+       struct sctp_chunk *chunk = _sctp_make_chunk(asoc, type, flags, paylen);
+
+       if (chunk)
+               sctp_control_set_owner_w(chunk);
+
+       return chunk;
+}
+
+/* Release the memory occupied by a chunk.  */
+static void sctp_chunk_destroy(struct sctp_chunk *chunk)
+{
+       BUG_ON(!list_empty(&chunk->list));
+       list_del_init(&chunk->transmitted_list);
+
+       consume_skb(chunk->skb);
+       consume_skb(chunk->auth_chunk);
+
+       SCTP_DBG_OBJCNT_DEC(chunk);
+       kmem_cache_free(sctp_chunk_cachep, chunk);
+}
+
+/* Possibly, free the chunk.  */
+void sctp_chunk_free(struct sctp_chunk *chunk)
+{
+       /* Release our reference on the message tracker. */
+       if (chunk->msg)
+               sctp_datamsg_put(chunk->msg);
+
+       sctp_chunk_put(chunk);
+}
+
+/* Grab a reference to the chunk. */
+void sctp_chunk_hold(struct sctp_chunk *ch)
+{
+       atomic_inc(&ch->refcnt);
+}
+
+/* Release a reference to the chunk. */
+void sctp_chunk_put(struct sctp_chunk *ch)
+{
+       if (atomic_dec_and_test(&ch->refcnt))
+               sctp_chunk_destroy(ch);
+}
+
+/* Append bytes to the end of a chunk.  Will panic if chunk is not big
+ * enough.
+ */
+void *sctp_addto_chunk(struct sctp_chunk *chunk, int len, const void *data)
+{
+       void *target;
+       void *padding;
+       int chunklen = ntohs(chunk->chunk_hdr->length);
+       int padlen = WORD_ROUND(chunklen) - chunklen;
+
+       padding = skb_put(chunk->skb, padlen);
+       target = skb_put(chunk->skb, len);
+
+       memset(padding, 0, padlen);
+       memcpy(target, data, len);
+
+       /* Adjust the chunk length field.  */
+       chunk->chunk_hdr->length = htons(chunklen + padlen + len);
+       chunk->chunk_end = skb_tail_pointer(chunk->skb);
+
+       return target;
+}
+
+/* Append bytes to the end of a chunk. Returns NULL if there isn't sufficient
+ * space in the chunk
+ */
+static void *sctp_addto_chunk_fixed(struct sctp_chunk *chunk,
+                                   int len, const void *data)
+{
+       if (skb_tailroom(chunk->skb) >= len)
+               return sctp_addto_chunk(chunk, len, data);
+       else
+               return NULL;
+}
+
+/* Append bytes from user space to the end of a chunk.  Will panic if
+ * chunk is not big enough.
+ * Returns a kernel err value.
+ */
+int sctp_user_addto_chunk(struct sctp_chunk *chunk, int len,
+                         struct iov_iter *from)
+{
+       void *target;
+       ssize_t copied;
+
+       /* Make room in chunk for data.  */
+       target = skb_put(chunk->skb, len);
+
+       /* Copy data (whole iovec) into chunk */
+       copied = copy_from_iter(target, len, from);
+       if (copied != len)
+               return -EFAULT;
+
+       /* Adjust the chunk length field.  */
+       chunk->chunk_hdr->length =
+               htons(ntohs(chunk->chunk_hdr->length) + len);
+       chunk->chunk_end = skb_tail_pointer(chunk->skb);
+
+       return 0;
+}
+
+/* Helper function to assign a TSN if needed.  This assumes that both
+ * the data_hdr and association have already been assigned.
+ */
+void sctp_chunk_assign_ssn(struct sctp_chunk *chunk)
+{
+       struct sctp_datamsg *msg;
+       struct sctp_chunk *lchunk;
+       struct sctp_stream *stream;
+       __u16 ssn;
+       __u16 sid;
+
+       if (chunk->has_ssn)
+               return;
+
+       /* All fragments will be on the same stream */
+       sid = ntohs(chunk->subh.data_hdr->stream);
+       stream = &chunk->asoc->ssnmap->out;
+
+       /* Now assign the sequence number to the entire message.
+        * All fragments must have the same stream sequence number.
+        */
+       msg = chunk->msg;
+       list_for_each_entry(lchunk, &msg->chunks, frag_list) {
+               if (lchunk->chunk_hdr->flags & SCTP_DATA_UNORDERED) {
+                       ssn = 0;
+               } else {
+                       if (lchunk->chunk_hdr->flags & SCTP_DATA_LAST_FRAG)
+                               ssn = sctp_ssn_next(stream, sid);
+                       else
+                               ssn = sctp_ssn_peek(stream, sid);
+               }
+
+               lchunk->subh.data_hdr->ssn = htons(ssn);
+               lchunk->has_ssn = 1;
+       }
+}
+
+/* Helper function to assign a TSN if needed.  This assumes that both
+ * the data_hdr and association have already been assigned.
+ */
+void sctp_chunk_assign_tsn(struct sctp_chunk *chunk)
+{
+       if (!chunk->has_tsn) {
+               /* This is the last possible instant to
+                * assign a TSN.
+                */
+               chunk->subh.data_hdr->tsn =
+                       htonl(sctp_association_get_next_tsn(chunk->asoc));
+               chunk->has_tsn = 1;
+       }
+}
+
+/* Create a CLOSED association to use with an incoming packet.  */
+struct sctp_association *sctp_make_temp_asoc(const struct sctp_endpoint *ep,
+                                       struct sctp_chunk *chunk,
+                                       gfp_t gfp)
+{
+       struct sctp_association *asoc;
+       struct sk_buff *skb;
+       sctp_scope_t scope;
+       struct sctp_af *af;
+
+       /* Create the bare association.  */
+       scope = sctp_scope(sctp_source(chunk));
+       asoc = sctp_association_new(ep, ep->base.sk, scope, gfp);
+       if (!asoc)
+               goto nodata;
+       asoc->temp = 1;
+       skb = chunk->skb;
+       /* Create an entry for the source address of the packet.  */
+       af = sctp_get_af_specific(ipver2af(ip_hdr(skb)->version));
+       if (unlikely(!af))
+               goto fail;
+       af->from_skb(&asoc->c.peer_addr, skb, 1);
+nodata:
+       return asoc;
+
+fail:
+       sctp_association_free(asoc);
+       return NULL;
+}
+
+/* Build a cookie representing asoc.
+ * This INCLUDES the param header needed to put the cookie in the INIT ACK.
+ */
+static sctp_cookie_param_t *sctp_pack_cookie(const struct sctp_endpoint *ep,
+                                     const struct sctp_association *asoc,
+                                     const struct sctp_chunk *init_chunk,
+                                     int *cookie_len,
+                                     const __u8 *raw_addrs, int addrs_len)
+{
+       sctp_cookie_param_t *retval;
+       struct sctp_signed_cookie *cookie;
+       struct scatterlist sg;
+       int headersize, bodysize;
+
+       /* Header size is static data prior to the actual cookie, including
+        * any padding.
+        */
+       headersize = sizeof(sctp_paramhdr_t) +
+                    (sizeof(struct sctp_signed_cookie) -
+                     sizeof(struct sctp_cookie));
+       bodysize = sizeof(struct sctp_cookie)
+               + ntohs(init_chunk->chunk_hdr->length) + addrs_len;
+
+       /* Pad out the cookie to a multiple to make the signature
+        * functions simpler to write.
+        */
+       if (bodysize % SCTP_COOKIE_MULTIPLE)
+               bodysize += SCTP_COOKIE_MULTIPLE
+                       - (bodysize % SCTP_COOKIE_MULTIPLE);
+       *cookie_len = headersize + bodysize;
+
+       /* Clear this memory since we are sending this data structure
+        * out on the network.
+        */
+       retval = kzalloc(*cookie_len, GFP_ATOMIC);
+       if (!retval)
+               goto nodata;
+
+       cookie = (struct sctp_signed_cookie *) retval->body;
+
+       /* Set up the parameter header.  */
+       retval->p.type = SCTP_PARAM_STATE_COOKIE;
+       retval->p.length = htons(*cookie_len);
+
+       /* Copy the cookie part of the association itself.  */
+       cookie->c = asoc->c;
+       /* Save the raw address list length in the cookie. */
+       cookie->c.raw_addr_list_len = addrs_len;
+
+       /* Remember PR-SCTP capability. */
+       cookie->c.prsctp_capable = asoc->peer.prsctp_capable;
+
+       /* Save adaptation indication in the cookie. */
+       cookie->c.adaptation_ind = asoc->peer.adaptation_ind;
+
+       /* Set an expiration time for the cookie.  */
+       cookie->c.expiration = ktime_add(asoc->cookie_life,
+                                        ktime_get());
+
+       /* Copy the peer's init packet.  */
+       memcpy(&cookie->c.peer_init[0], init_chunk->chunk_hdr,
+              ntohs(init_chunk->chunk_hdr->length));
+
+       /* Copy the raw local address list of the association. */
+       memcpy((__u8 *)&cookie->c.peer_init[0] +
+              ntohs(init_chunk->chunk_hdr->length), raw_addrs, addrs_len);
+
+       if (sctp_sk(ep->base.sk)->hmac) {
+               struct hash_desc desc;
+
+               /* Sign the message.  */
+               sg_init_one(&sg, &cookie->c, bodysize);
+               desc.tfm = sctp_sk(ep->base.sk)->hmac;
+               desc.flags = 0;
+
+               if (crypto_hash_setkey(desc.tfm, ep->secret_key,
+                                      sizeof(ep->secret_key)) ||
+                   crypto_hash_digest(&desc, &sg, bodysize, cookie->signature))
+                       goto free_cookie;
+       }
+
+       return retval;
+
+free_cookie:
+       kfree(retval);
+nodata:
+       *cookie_len = 0;
+       return NULL;
+}
+
+/* Unpack the cookie from COOKIE ECHO chunk, recreating the association.  */
+struct sctp_association *sctp_unpack_cookie(
+       const struct sctp_endpoint *ep,
+       const struct sctp_association *asoc,
+       struct sctp_chunk *chunk, gfp_t gfp,
+       int *error, struct sctp_chunk **errp)
+{
+       struct sctp_association *retval = NULL;
+       struct sctp_signed_cookie *cookie;
+       struct sctp_cookie *bear_cookie;
+       int headersize, bodysize, fixed_size;
+       __u8 *digest = ep->digest;
+       struct scatterlist sg;
+       unsigned int len;
+       sctp_scope_t scope;
+       struct sk_buff *skb = chunk->skb;
+       ktime_t kt;
+       struct hash_desc desc;
+
+       /* Header size is static data prior to the actual cookie, including
+        * any padding.
+        */
+       headersize = sizeof(sctp_chunkhdr_t) +
+                    (sizeof(struct sctp_signed_cookie) -
+                     sizeof(struct sctp_cookie));
+       bodysize = ntohs(chunk->chunk_hdr->length) - headersize;
+       fixed_size = headersize + sizeof(struct sctp_cookie);
+
+       /* Verify that the chunk looks like it even has a cookie.
+        * There must be enough room for our cookie and our peer's
+        * INIT chunk.
+        */
+       len = ntohs(chunk->chunk_hdr->length);
+       if (len < fixed_size + sizeof(struct sctp_chunkhdr))
+               goto malformed;
+
+       /* Verify that the cookie has been padded out. */
+       if (bodysize % SCTP_COOKIE_MULTIPLE)
+               goto malformed;
+
+       /* Process the cookie.  */
+       cookie = chunk->subh.cookie_hdr;
+       bear_cookie = &cookie->c;
+
+       if (!sctp_sk(ep->base.sk)->hmac)
+               goto no_hmac;
+
+       /* Check the signature.  */
+       sg_init_one(&sg, bear_cookie, bodysize);
+       desc.tfm = sctp_sk(ep->base.sk)->hmac;
+       desc.flags = 0;
+
+       memset(digest, 0x00, SCTP_SIGNATURE_SIZE);
+       if (crypto_hash_setkey(desc.tfm, ep->secret_key,
+                              sizeof(ep->secret_key)) ||
+           crypto_hash_digest(&desc, &sg, bodysize, digest)) {
+               *error = -SCTP_IERROR_NOMEM;
+               goto fail;
+       }
+
+       if (memcmp(digest, cookie->signature, SCTP_SIGNATURE_SIZE)) {
+               *error = -SCTP_IERROR_BAD_SIG;
+               goto fail;
+       }
+
+no_hmac:
+       /* IG Section 2.35.2:
+        *  3) Compare the port numbers and the verification tag contained
+        *     within the COOKIE ECHO chunk to the actual port numbers and the
+        *     verification tag within the SCTP common header of the received
+        *     packet. If these values do not match the packet MUST be silently
+        *     discarded,
+        */
+       if (ntohl(chunk->sctp_hdr->vtag) != bear_cookie->my_vtag) {
+               *error = -SCTP_IERROR_BAD_TAG;
+               goto fail;
+       }
+
+       if (chunk->sctp_hdr->source != bear_cookie->peer_addr.v4.sin_port ||
+           ntohs(chunk->sctp_hdr->dest) != bear_cookie->my_port) {
+               *error = -SCTP_IERROR_BAD_PORTS;
+               goto fail;
+       }
+
+       /* Check to see if the cookie is stale.  If there is already
+        * an association, there is no need to check cookie's expiration
+        * for init collision case of lost COOKIE ACK.
+        * If skb has been timestamped, then use the stamp, otherwise
+        * use current time.  This introduces a small possibility that
+        * that a cookie may be considered expired, but his would only slow
+        * down the new association establishment instead of every packet.
+        */
+       if (sock_flag(ep->base.sk, SOCK_TIMESTAMP))
+               kt = skb_get_ktime(skb);
+       else
+               kt = ktime_get();
+
+       if (!asoc && ktime_before(bear_cookie->expiration, kt)) {
+               /*
+                * Section 3.3.10.3 Stale Cookie Error (3)
+                *
+                * Cause of error
+                * ---------------
+                * Stale Cookie Error:  Indicates the receipt of a valid State
+                * Cookie that has expired.
+                */
+               len = ntohs(chunk->chunk_hdr->length);
+               *errp = sctp_make_op_error_space(asoc, chunk, len);
+               if (*errp) {
+                       suseconds_t usecs = ktime_to_us(ktime_sub(kt, bear_cookie->expiration));
+                       __be32 n = htonl(usecs);
+
+                       sctp_init_cause(*errp, SCTP_ERROR_STALE_COOKIE,
+                                       sizeof(n));
+                       sctp_addto_chunk(*errp, sizeof(n), &n);
+                       *error = -SCTP_IERROR_STALE_COOKIE;
+               } else
+                       *error = -SCTP_IERROR_NOMEM;
+
+               goto fail;
+       }
+
+       /* Make a new base association.  */
+       scope = sctp_scope(sctp_source(chunk));
+       retval = sctp_association_new(ep, ep->base.sk, scope, gfp);
+       if (!retval) {
+               *error = -SCTP_IERROR_NOMEM;
+               goto fail;
+       }
+
+       /* Set up our peer's port number.  */
+       retval->peer.port = ntohs(chunk->sctp_hdr->source);
+
+       /* Populate the association from the cookie.  */
+       memcpy(&retval->c, bear_cookie, sizeof(*bear_cookie));
+
+       if (sctp_assoc_set_bind_addr_from_cookie(retval, bear_cookie,
+                                                GFP_ATOMIC) < 0) {
+               *error = -SCTP_IERROR_NOMEM;
+               goto fail;
+       }
+
+       /* Also, add the destination address. */
+       if (list_empty(&retval->base.bind_addr.address_list)) {
+               sctp_add_bind_addr(&retval->base.bind_addr, &chunk->dest,
+                               SCTP_ADDR_SRC, GFP_ATOMIC);
+       }
+
+       retval->next_tsn = retval->c.initial_tsn;
+       retval->ctsn_ack_point = retval->next_tsn - 1;
+       retval->addip_serial = retval->c.initial_tsn;
+       retval->adv_peer_ack_point = retval->ctsn_ack_point;
+       retval->peer.prsctp_capable = retval->c.prsctp_capable;
+       retval->peer.adaptation_ind = retval->c.adaptation_ind;
+
+       /* The INIT stuff will be done by the side effects.  */
+       return retval;
+
+fail:
+       if (retval)
+               sctp_association_free(retval);
+
+       return NULL;
+
+malformed:
+       /* Yikes!  The packet is either corrupt or deliberately
+        * malformed.
+        */
+       *error = -SCTP_IERROR_MALFORMED;
+       goto fail;
+}
+
+/********************************************************************
+ * 3rd Level Abstractions
+ ********************************************************************/
+
+struct __sctp_missing {
+       __be32 num_missing;
+       __be16 type;
+}  __packed;
+
+/*
+ * Report a missing mandatory parameter.
+ */
+static int sctp_process_missing_param(const struct sctp_association *asoc,
+                                     sctp_param_t paramtype,
+                                     struct sctp_chunk *chunk,
+                                     struct sctp_chunk **errp)
+{
+       struct __sctp_missing report;
+       __u16 len;
+
+       len = WORD_ROUND(sizeof(report));
+
+       /* Make an ERROR chunk, preparing enough room for
+        * returning multiple unknown parameters.
+        */
+       if (!*errp)
+               *errp = sctp_make_op_error_space(asoc, chunk, len);
+
+       if (*errp) {
+               report.num_missing = htonl(1);
+               report.type = paramtype;
+               sctp_init_cause(*errp, SCTP_ERROR_MISS_PARAM,
+                               sizeof(report));
+               sctp_addto_chunk(*errp, sizeof(report), &report);
+       }
+
+       /* Stop processing this chunk. */
+       return 0;
+}
+
+/* Report an Invalid Mandatory Parameter.  */
+static int sctp_process_inv_mandatory(const struct sctp_association *asoc,
+                                     struct sctp_chunk *chunk,
+                                     struct sctp_chunk **errp)
+{
+       /* Invalid Mandatory Parameter Error has no payload. */
+
+       if (!*errp)
+               *errp = sctp_make_op_error_space(asoc, chunk, 0);
+
+       if (*errp)
+               sctp_init_cause(*errp, SCTP_ERROR_INV_PARAM, 0);
+
+       /* Stop processing this chunk. */
+       return 0;
+}
+
+static int sctp_process_inv_paramlength(const struct sctp_association *asoc,
+                                       struct sctp_paramhdr *param,
+                                       const struct sctp_chunk *chunk,
+                                       struct sctp_chunk **errp)
+{
+       /* This is a fatal error.  Any accumulated non-fatal errors are
+        * not reported.
+        */
+       if (*errp)
+               sctp_chunk_free(*errp);
+
+       /* Create an error chunk and fill it in with our payload. */
+       *errp = sctp_make_violation_paramlen(asoc, chunk, param);
+
+       return 0;
+}
+
+
+/* Do not attempt to handle the HOST_NAME parm.  However, do
+ * send back an indicator to the peer.
+ */
+static int sctp_process_hn_param(const struct sctp_association *asoc,
+                                union sctp_params param,
+                                struct sctp_chunk *chunk,
+                                struct sctp_chunk **errp)
+{
+       __u16 len = ntohs(param.p->length);
+
+       /* Processing of the HOST_NAME parameter will generate an
+        * ABORT.  If we've accumulated any non-fatal errors, they
+        * would be unrecognized parameters and we should not include
+        * them in the ABORT.
+        */
+       if (*errp)
+               sctp_chunk_free(*errp);
+
+       *errp = sctp_make_op_error_space(asoc, chunk, len);
+
+       if (*errp) {
+               sctp_init_cause(*errp, SCTP_ERROR_DNS_FAILED, len);
+               sctp_addto_chunk(*errp, len, param.v);
+       }
+
+       /* Stop processing this chunk. */
+       return 0;
+}
+
+static int sctp_verify_ext_param(struct net *net, union sctp_params param)
+{
+       __u16 num_ext = ntohs(param.p->length) - sizeof(sctp_paramhdr_t);
+       int have_auth = 0;
+       int have_asconf = 0;
+       int i;
+
+       for (i = 0; i < num_ext; i++) {
+               switch (param.ext->chunks[i]) {
+               case SCTP_CID_AUTH:
+                       have_auth = 1;
+                       break;
+               case SCTP_CID_ASCONF:
+               case SCTP_CID_ASCONF_ACK:
+                       have_asconf = 1;
+                       break;
+               }
+       }
+
+       /* ADD-IP Security: The draft requires us to ABORT or ignore the
+        * INIT/INIT-ACK if ADD-IP is listed, but AUTH is not.  Do this
+        * only if ADD-IP is turned on and we are not backward-compatible
+        * mode.
+        */
+       if (net->sctp.addip_noauth)
+               return 1;
+
+       if (net->sctp.addip_enable && !have_auth && have_asconf)
+               return 0;
+
+       return 1;
+}
+
+static void sctp_process_ext_param(struct sctp_association *asoc,
+                                   union sctp_params param)
+{
+       struct net *net = sock_net(asoc->base.sk);
+       __u16 num_ext = ntohs(param.p->length) - sizeof(sctp_paramhdr_t);
+       int i;
+
+       for (i = 0; i < num_ext; i++) {
+               switch (param.ext->chunks[i]) {
+               case SCTP_CID_FWD_TSN:
+                       if (net->sctp.prsctp_enable && !asoc->peer.prsctp_capable)
+                                   asoc->peer.prsctp_capable = 1;
+                       break;
+               case SCTP_CID_AUTH:
+                       /* if the peer reports AUTH, assume that he
+                        * supports AUTH.
+                        */
+                       if (asoc->ep->auth_enable)
+                               asoc->peer.auth_capable = 1;
+                       break;
+               case SCTP_CID_ASCONF:
+               case SCTP_CID_ASCONF_ACK:
+                       if (net->sctp.addip_enable)
+                               asoc->peer.asconf_capable = 1;
+                       break;
+               default:
+                       break;
+               }
+       }
+}
+
+/* RFC 3.2.1 & the Implementers Guide 2.2.
+ *
+ * The Parameter Types are encoded such that the
+ * highest-order two bits specify the action that must be
+ * taken if the processing endpoint does not recognize the
+ * Parameter Type.
+ *
+ * 00 - Stop processing this parameter; do not process any further
+ *     parameters within this chunk
+ *
+ * 01 - Stop processing this parameter, do not process any further
+ *     parameters within this chunk, and report the unrecognized
+ *     parameter in an 'Unrecognized Parameter' ERROR chunk.
+ *
+ * 10 - Skip this parameter and continue processing.
+ *
+ * 11 - Skip this parameter and continue processing but
+ *     report the unrecognized parameter in an
+ *     'Unrecognized Parameter' ERROR chunk.
+ *
+ * Return value:
+ *     SCTP_IERROR_NO_ERROR - continue with the chunk
+ *     SCTP_IERROR_ERROR    - stop and report an error.
+ *     SCTP_IERROR_NOMEME   - out of memory.
+ */
+static sctp_ierror_t sctp_process_unk_param(const struct sctp_association *asoc,
+                                           union sctp_params param,
+                                           struct sctp_chunk *chunk,
+                                           struct sctp_chunk **errp)
+{
+       int retval = SCTP_IERROR_NO_ERROR;
+
+       switch (param.p->type & SCTP_PARAM_ACTION_MASK) {
+       case SCTP_PARAM_ACTION_DISCARD:
+               retval =  SCTP_IERROR_ERROR;
+               break;
+       case SCTP_PARAM_ACTION_SKIP:
+               break;
+       case SCTP_PARAM_ACTION_DISCARD_ERR:
+               retval =  SCTP_IERROR_ERROR;
+               /* Fall through */
+       case SCTP_PARAM_ACTION_SKIP_ERR:
+               /* Make an ERROR chunk, preparing enough room for
+                * returning multiple unknown parameters.
+                */
+               if (NULL == *errp)
+                       *errp = sctp_make_op_error_fixed(asoc, chunk);
+
+               if (*errp) {
+                       if (!sctp_init_cause_fixed(*errp, SCTP_ERROR_UNKNOWN_PARAM,
+                                       WORD_ROUND(ntohs(param.p->length))))
+                               sctp_addto_chunk_fixed(*errp,
+                                               WORD_ROUND(ntohs(param.p->length)),
+                                               param.v);
+               } else {
+                       /* If there is no memory for generating the ERROR
+                        * report as specified, an ABORT will be triggered
+                        * to the peer and the association won't be
+                        * established.
+                        */
+                       retval = SCTP_IERROR_NOMEM;
+               }
+               break;
+       default:
+               break;
+       }
+
+       return retval;
+}
+
+/* Verify variable length parameters
+ * Return values:
+ *     SCTP_IERROR_ABORT - trigger an ABORT
+ *     SCTP_IERROR_NOMEM - out of memory (abort)
+ *     SCTP_IERROR_ERROR - stop processing, trigger an ERROR
+ *     SCTP_IERROR_NO_ERROR - continue with the chunk
+ */
+static sctp_ierror_t sctp_verify_param(struct net *net,
+                                       const struct sctp_endpoint *ep,
+                                       const struct sctp_association *asoc,
+                                       union sctp_params param,
+                                       sctp_cid_t cid,
+                                       struct sctp_chunk *chunk,
+                                       struct sctp_chunk **err_chunk)
+{
+       struct sctp_hmac_algo_param *hmacs;
+       int retval = SCTP_IERROR_NO_ERROR;
+       __u16 n_elt, id = 0;
+       int i;
+
+       /* FIXME - This routine is not looking at each parameter per the
+        * chunk type, i.e., unrecognized parameters should be further
+        * identified based on the chunk id.
+        */
+
+       switch (param.p->type) {
+       case SCTP_PARAM_IPV4_ADDRESS:
+       case SCTP_PARAM_IPV6_ADDRESS:
+       case SCTP_PARAM_COOKIE_PRESERVATIVE:
+       case SCTP_PARAM_SUPPORTED_ADDRESS_TYPES:
+       case SCTP_PARAM_STATE_COOKIE:
+       case SCTP_PARAM_HEARTBEAT_INFO:
+       case SCTP_PARAM_UNRECOGNIZED_PARAMETERS:
+       case SCTP_PARAM_ECN_CAPABLE:
+       case SCTP_PARAM_ADAPTATION_LAYER_IND:
+               break;
+
+       case SCTP_PARAM_SUPPORTED_EXT:
+               if (!sctp_verify_ext_param(net, param))
+                       return SCTP_IERROR_ABORT;
+               break;
+
+       case SCTP_PARAM_SET_PRIMARY:
+               if (net->sctp.addip_enable)
+                       break;
+               goto fallthrough;
+
+       case SCTP_PARAM_HOST_NAME_ADDRESS:
+               /* Tell the peer, we won't support this param.  */
+               sctp_process_hn_param(asoc, param, chunk, err_chunk);
+               retval = SCTP_IERROR_ABORT;
+               break;
+
+       case SCTP_PARAM_FWD_TSN_SUPPORT:
+               if (net->sctp.prsctp_enable)
+                       break;
+               goto fallthrough;
+
+       case SCTP_PARAM_RANDOM:
+               if (!ep->auth_enable)
+                       goto fallthrough;
+
+               /* SCTP-AUTH: Secion 6.1
+                * If the random number is not 32 byte long the association
+                * MUST be aborted.  The ABORT chunk SHOULD contain the error
+                * cause 'Protocol Violation'.
+                */
+               if (SCTP_AUTH_RANDOM_LENGTH !=
+                       ntohs(param.p->length) - sizeof(sctp_paramhdr_t)) {
+                       sctp_process_inv_paramlength(asoc, param.p,
+                                                       chunk, err_chunk);
+                       retval = SCTP_IERROR_ABORT;
+               }
+               break;
+
+       case SCTP_PARAM_CHUNKS:
+               if (!ep->auth_enable)
+                       goto fallthrough;
+
+               /* SCTP-AUTH: Section 3.2
+                * The CHUNKS parameter MUST be included once in the INIT or
+                *  INIT-ACK chunk if the sender wants to receive authenticated
+                *  chunks.  Its maximum length is 260 bytes.
+                */
+               if (260 < ntohs(param.p->length)) {
+                       sctp_process_inv_paramlength(asoc, param.p,
+                                                    chunk, err_chunk);
+                       retval = SCTP_IERROR_ABORT;
+               }
+               break;
+
+       case SCTP_PARAM_HMAC_ALGO:
+               if (!ep->auth_enable)
+                       goto fallthrough;
+
+               hmacs = (struct sctp_hmac_algo_param *)param.p;
+               n_elt = (ntohs(param.p->length) - sizeof(sctp_paramhdr_t)) >> 1;
+
+               /* SCTP-AUTH: Section 6.1
+                * The HMAC algorithm based on SHA-1 MUST be supported and
+                * included in the HMAC-ALGO parameter.
+                */
+               for (i = 0; i < n_elt; i++) {
+                       id = ntohs(hmacs->hmac_ids[i]);
+
+                       if (id == SCTP_AUTH_HMAC_ID_SHA1)
+                               break;
+               }
+
+               if (id != SCTP_AUTH_HMAC_ID_SHA1) {
+                       sctp_process_inv_paramlength(asoc, param.p, chunk,
+                                                    err_chunk);
+                       retval = SCTP_IERROR_ABORT;
+               }
+               break;
+fallthrough:
+       default:
+               pr_debug("%s: unrecognized param:%d for chunk:%d\n",
+                        __func__, ntohs(param.p->type), cid);
+
+               retval = sctp_process_unk_param(asoc, param, chunk, err_chunk);
+               break;
+       }
+       return retval;
+}
+
+/* Verify the INIT packet before we process it.  */
+int sctp_verify_init(struct net *net, const struct sctp_endpoint *ep,
+                    const struct sctp_association *asoc, sctp_cid_t cid,
+                    sctp_init_chunk_t *peer_init, struct sctp_chunk *chunk,
+                    struct sctp_chunk **errp)
+{
+       union sctp_params param;
+       bool has_cookie = false;
+       int result;
+
+       /* Check for missing mandatory parameters. Note: Initial TSN is
+        * also mandatory, but is not checked here since the valid range
+        * is 0..2**32-1. RFC4960, section 3.3.3.
+        */
+       if (peer_init->init_hdr.num_outbound_streams == 0 ||
+           peer_init->init_hdr.num_inbound_streams == 0 ||
+           peer_init->init_hdr.init_tag == 0 ||
+           ntohl(peer_init->init_hdr.a_rwnd) < SCTP_DEFAULT_MINWINDOW)
+               return sctp_process_inv_mandatory(asoc, chunk, errp);
+
+       sctp_walk_params(param, peer_init, init_hdr.params) {
+               if (param.p->type == SCTP_PARAM_STATE_COOKIE)
+                       has_cookie = true;
+       }
+
+       /* There is a possibility that a parameter length was bad and
+        * in that case we would have stoped walking the parameters.
+        * The current param.p would point at the bad one.
+        * Current consensus on the mailing list is to generate a PROTOCOL
+        * VIOLATION error.  We build the ERROR chunk here and let the normal
+        * error handling code build and send the packet.
+        */
+       if (param.v != (void *)chunk->chunk_end)
+               return sctp_process_inv_paramlength(asoc, param.p, chunk, errp);
+
+       /* The only missing mandatory param possible today is
+        * the state cookie for an INIT-ACK chunk.
+        */
+       if ((SCTP_CID_INIT_ACK == cid) && !has_cookie)
+               return sctp_process_missing_param(asoc, SCTP_PARAM_STATE_COOKIE,
+                                                 chunk, errp);
+
+       /* Verify all the variable length parameters */
+       sctp_walk_params(param, peer_init, init_hdr.params) {
+               result = sctp_verify_param(net, ep, asoc, param, cid,
+                                          chunk, errp);
+               switch (result) {
+               case SCTP_IERROR_ABORT:
+               case SCTP_IERROR_NOMEM:
+                       return 0;
+               case SCTP_IERROR_ERROR:
+                       return 1;
+               case SCTP_IERROR_NO_ERROR:
+               default:
+                       break;
+               }
+
+       } /* for (loop through all parameters) */
+
+       return 1;
+}
+
+/* Unpack the parameters in an INIT packet into an association.
+ * Returns 0 on failure, else success.
+ * FIXME:  This is an association method.
+ */
+int sctp_process_init(struct sctp_association *asoc, struct sctp_chunk *chunk,
+                     const union sctp_addr *peer_addr,
+                     sctp_init_chunk_t *peer_init, gfp_t gfp)
+{
+       struct net *net = sock_net(asoc->base.sk);
+       union sctp_params param;
+       struct sctp_transport *transport;
+       struct list_head *pos, *temp;
+       struct sctp_af *af;
+       union sctp_addr addr;
+       char *cookie;
+       int src_match = 0;
+
+       /* We must include the address that the INIT packet came from.
+        * This is the only address that matters for an INIT packet.
+        * When processing a COOKIE ECHO, we retrieve the from address
+        * of the INIT from the cookie.
+        */
+
+       /* This implementation defaults to making the first transport
+        * added as the primary transport.  The source address seems to
+        * be a a better choice than any of the embedded addresses.
+        */
+       if (!sctp_assoc_add_peer(asoc, peer_addr, gfp, SCTP_ACTIVE))
+               goto nomem;
+
+       if (sctp_cmp_addr_exact(sctp_source(chunk), peer_addr))
+               src_match = 1;
+
+       /* Process the initialization parameters.  */
+       sctp_walk_params(param, peer_init, init_hdr.params) {
+               if (!src_match && (param.p->type == SCTP_PARAM_IPV4_ADDRESS ||
+                   param.p->type == SCTP_PARAM_IPV6_ADDRESS)) {
+                       af = sctp_get_af_specific(param_type2af(param.p->type));
+                       af->from_addr_param(&addr, param.addr,
+                                           chunk->sctp_hdr->source, 0);
+                       if (sctp_cmp_addr_exact(sctp_source(chunk), &addr))
+                               src_match = 1;
+               }
+
+               if (!sctp_process_param(asoc, param, peer_addr, gfp))
+                       goto clean_up;
+       }
+
+       /* source address of chunk may not match any valid address */
+       if (!src_match)
+               goto clean_up;
+
+       /* AUTH: After processing the parameters, make sure that we
+        * have all the required info to potentially do authentications.
+        */
+       if (asoc->peer.auth_capable && (!asoc->peer.peer_random ||
+                                       !asoc->peer.peer_hmacs))
+               asoc->peer.auth_capable = 0;
+
+       /* In a non-backward compatible mode, if the peer claims
+        * support for ADD-IP but not AUTH,  the ADD-IP spec states
+        * that we MUST ABORT the association. Section 6.  The section
+        * also give us an option to silently ignore the packet, which
+        * is what we'll do here.
+        */
+       if (!net->sctp.addip_noauth &&
+            (asoc->peer.asconf_capable && !asoc->peer.auth_capable)) {
+               asoc->peer.addip_disabled_mask |= (SCTP_PARAM_ADD_IP |
+                                                 SCTP_PARAM_DEL_IP |
+                                                 SCTP_PARAM_SET_PRIMARY);
+               asoc->peer.asconf_capable = 0;
+               goto clean_up;
+       }
+
+       /* Walk list of transports, removing transports in the UNKNOWN state. */
+       list_for_each_safe(pos, temp, &asoc->peer.transport_addr_list) {
+               transport = list_entry(pos, struct sctp_transport, transports);
+               if (transport->state == SCTP_UNKNOWN) {
+                       sctp_assoc_rm_peer(asoc, transport);
+               }
+       }
+
+       /* The fixed INIT headers are always in network byte
+        * order.
+        */
+       asoc->peer.i.init_tag =
+               ntohl(peer_init->init_hdr.init_tag);
+       asoc->peer.i.a_rwnd =
+               ntohl(peer_init->init_hdr.a_rwnd);
+       asoc->peer.i.num_outbound_streams =
+               ntohs(peer_init->init_hdr.num_outbound_streams);
+       asoc->peer.i.num_inbound_streams =
+               ntohs(peer_init->init_hdr.num_inbound_streams);
+       asoc->peer.i.initial_tsn =
+               ntohl(peer_init->init_hdr.initial_tsn);
+
+       /* Apply the upper bounds for output streams based on peer's
+        * number of inbound streams.
+        */
+       if (asoc->c.sinit_num_ostreams  >
+           ntohs(peer_init->init_hdr.num_inbound_streams)) {
+               asoc->c.sinit_num_ostreams =
+                       ntohs(peer_init->init_hdr.num_inbound_streams);
+       }
+
+       if (asoc->c.sinit_max_instreams >
+           ntohs(peer_init->init_hdr.num_outbound_streams)) {
+               asoc->c.sinit_max_instreams =
+                       ntohs(peer_init->init_hdr.num_outbound_streams);
+       }
+
+       /* Copy Initiation tag from INIT to VT_peer in cookie.   */
+       asoc->c.peer_vtag = asoc->peer.i.init_tag;
+
+       /* Peer Rwnd   : Current calculated value of the peer's rwnd.  */
+       asoc->peer.rwnd = asoc->peer.i.a_rwnd;
+
+       /* Copy cookie in case we need to resend COOKIE-ECHO. */
+       cookie = asoc->peer.cookie;
+       if (cookie) {
+               asoc->peer.cookie = kmemdup(cookie, asoc->peer.cookie_len, gfp);
+               if (!asoc->peer.cookie)
+                       goto clean_up;
+       }
+
+       /* RFC 2960 7.2.1 The initial value of ssthresh MAY be arbitrarily
+        * high (for example, implementations MAY use the size of the receiver
+        * advertised window).
+        */
+       list_for_each_entry(transport, &asoc->peer.transport_addr_list,
+                       transports) {
+               transport->ssthresh = asoc->peer.i.a_rwnd;
+       }
+
+       /* Set up the TSN tracking pieces.  */
+       if (!sctp_tsnmap_init(&asoc->peer.tsn_map, SCTP_TSN_MAP_INITIAL,
+                               asoc->peer.i.initial_tsn, gfp))
+               goto clean_up;
+
+       /* RFC 2960 6.5 Stream Identifier and Stream Sequence Number
+        *
+        * The stream sequence number in all the streams shall start
+        * from 0 when the association is established.  Also, when the
+        * stream sequence number reaches the value 65535 the next
+        * stream sequence number shall be set to 0.
+        */
+
+       /* Allocate storage for the negotiated streams if it is not a temporary
+        * association.
+        */
+       if (!asoc->temp) {
+               int error;
+
+               asoc->ssnmap = sctp_ssnmap_new(asoc->c.sinit_max_instreams,
+                                              asoc->c.sinit_num_ostreams, gfp);
+               if (!asoc->ssnmap)
+                       goto clean_up;
+
+               error = sctp_assoc_set_id(asoc, gfp);
+               if (error)
+                       goto clean_up;
+       }
+
+       /* ADDIP Section 4.1 ASCONF Chunk Procedures
+        *
+        * When an endpoint has an ASCONF signaled change to be sent to the
+        * remote endpoint it should do the following:
+        * ...
+        * A2) A serial number should be assigned to the Chunk. The serial
+        * number should be a monotonically increasing number. All serial
+        * numbers are defined to be initialized at the start of the
+        * association to the same value as the Initial TSN.
+        */
+       asoc->peer.addip_serial = asoc->peer.i.initial_tsn - 1;
+       return 1;
+
+clean_up:
+       /* Release the transport structures. */
+       list_for_each_safe(pos, temp, &asoc->peer.transport_addr_list) {
+               transport = list_entry(pos, struct sctp_transport, transports);
+               if (transport->state != SCTP_ACTIVE)
+                       sctp_assoc_rm_peer(asoc, transport);
+       }
+
+nomem:
+       return 0;
+}
+
+
+/* Update asoc with the option described in param.
+ *
+ * RFC2960 3.3.2.1 Optional/Variable Length Parameters in INIT
+ *
+ * asoc is the association to update.
+ * param is the variable length parameter to use for update.
+ * cid tells us if this is an INIT, INIT ACK or COOKIE ECHO.
+ * If the current packet is an INIT we want to minimize the amount of
+ * work we do.  In particular, we should not build transport
+ * structures for the addresses.
+ */
+static int sctp_process_param(struct sctp_association *asoc,
+                             union sctp_params param,
+                             const union sctp_addr *peer_addr,
+                             gfp_t gfp)
+{
+       struct net *net = sock_net(asoc->base.sk);
+       union sctp_addr addr;
+       int i;
+       __u16 sat;
+       int retval = 1;
+       sctp_scope_t scope;
+       time_t stale;
+       struct sctp_af *af;
+       union sctp_addr_param *addr_param;
+       struct sctp_transport *t;
+       struct sctp_endpoint *ep = asoc->ep;
+
+       /* We maintain all INIT parameters in network byte order all the
+        * time.  This allows us to not worry about whether the parameters
+        * came from a fresh INIT, and INIT ACK, or were stored in a cookie.
+        */
+       switch (param.p->type) {
+       case SCTP_PARAM_IPV6_ADDRESS:
+               if (PF_INET6 != asoc->base.sk->sk_family)
+                       break;
+               goto do_addr_param;
+
+       case SCTP_PARAM_IPV4_ADDRESS:
+               /* v4 addresses are not allowed on v6-only socket */
+               if (ipv6_only_sock(asoc->base.sk))
+                       break;
+do_addr_param:
+               af = sctp_get_af_specific(param_type2af(param.p->type));
+               af->from_addr_param(&addr, param.addr, htons(asoc->peer.port), 0);
+               scope = sctp_scope(peer_addr);
+               if (sctp_in_scope(net, &addr, scope))
+                       if (!sctp_assoc_add_peer(asoc, &addr, gfp, SCTP_UNCONFIRMED))
+                               return 0;
+               break;
+
+       case SCTP_PARAM_COOKIE_PRESERVATIVE:
+               if (!net->sctp.cookie_preserve_enable)
+                       break;
+
+               stale = ntohl(param.life->lifespan_increment);
+
+               /* Suggested Cookie Life span increment's unit is msec,
+                * (1/1000sec).
+                */
+               asoc->cookie_life = ktime_add_ms(asoc->cookie_life, stale);
+               break;
+
+       case SCTP_PARAM_HOST_NAME_ADDRESS:
+               pr_debug("%s: unimplemented SCTP_HOST_NAME_ADDRESS\n", __func__);
+               break;
+
+       case SCTP_PARAM_SUPPORTED_ADDRESS_TYPES:
+               /* Turn off the default values first so we'll know which
+                * ones are really set by the peer.
+                */
+               asoc->peer.ipv4_address = 0;
+               asoc->peer.ipv6_address = 0;
+
+               /* Assume that peer supports the address family
+                * by which it sends a packet.
+                */
+               if (peer_addr->sa.sa_family == AF_INET6)
+                       asoc->peer.ipv6_address = 1;
+               else if (peer_addr->sa.sa_family == AF_INET)
+                       asoc->peer.ipv4_address = 1;
+
+               /* Cycle through address types; avoid divide by 0. */
+               sat = ntohs(param.p->length) - sizeof(sctp_paramhdr_t);
+               if (sat)
+                       sat /= sizeof(__u16);
+
+               for (i = 0; i < sat; ++i) {
+                       switch (param.sat->types[i]) {
+                       case SCTP_PARAM_IPV4_ADDRESS:
+                               asoc->peer.ipv4_address = 1;
+                               break;
+
+                       case SCTP_PARAM_IPV6_ADDRESS:
+                               if (PF_INET6 == asoc->base.sk->sk_family)
+                                       asoc->peer.ipv6_address = 1;
+                               break;
+
+                       case SCTP_PARAM_HOST_NAME_ADDRESS:
+                               asoc->peer.hostname_address = 1;
+                               break;
+
+                       default: /* Just ignore anything else.  */
+                               break;
+                       }
+               }
+               break;
+
+       case SCTP_PARAM_STATE_COOKIE:
+               asoc->peer.cookie_len =
+                       ntohs(param.p->length) - sizeof(sctp_paramhdr_t);
+               asoc->peer.cookie = param.cookie->body;
+               break;
+
+       case SCTP_PARAM_HEARTBEAT_INFO:
+               /* Would be odd to receive, but it causes no problems. */
+               break;
+
+       case SCTP_PARAM_UNRECOGNIZED_PARAMETERS:
+               /* Rejected during verify stage. */
+               break;
+
+       case SCTP_PARAM_ECN_CAPABLE:
+               asoc->peer.ecn_capable = 1;
+               break;
+
+       case SCTP_PARAM_ADAPTATION_LAYER_IND:
+               asoc->peer.adaptation_ind = ntohl(param.aind->adaptation_ind);
+               break;
+
+       case SCTP_PARAM_SET_PRIMARY:
+               if (!net->sctp.addip_enable)
+                       goto fall_through;
+
+               addr_param = param.v + sizeof(sctp_addip_param_t);
+
+               af = sctp_get_af_specific(param_type2af(addr_param->p.type));
+               if (af == NULL)
+                       break;
+
+               af->from_addr_param(&addr, addr_param,
+                                   htons(asoc->peer.port), 0);
+
+               /* if the address is invalid, we can't process it.
+                * XXX: see spec for what to do.
+                */
+               if (!af->addr_valid(&addr, NULL, NULL))
+                       break;
+
+               t = sctp_assoc_lookup_paddr(asoc, &addr);
+               if (!t)
+                       break;
+
+               sctp_assoc_set_primary(asoc, t);
+               break;
+
+       case SCTP_PARAM_SUPPORTED_EXT:
+               sctp_process_ext_param(asoc, param);
+               break;
+
+       case SCTP_PARAM_FWD_TSN_SUPPORT:
+               if (net->sctp.prsctp_enable) {
+                       asoc->peer.prsctp_capable = 1;
+                       break;
+               }
+               /* Fall Through */
+               goto fall_through;
+
+       case SCTP_PARAM_RANDOM:
+               if (!ep->auth_enable)
+                       goto fall_through;
+
+               /* Save peer's random parameter */
+               asoc->peer.peer_random = kmemdup(param.p,
+                                           ntohs(param.p->length), gfp);
+               if (!asoc->peer.peer_random) {
+                       retval = 0;
+                       break;
+               }
+               break;
+
+       case SCTP_PARAM_HMAC_ALGO:
+               if (!ep->auth_enable)
+                       goto fall_through;
+
+               /* Save peer's HMAC list */
+               asoc->peer.peer_hmacs = kmemdup(param.p,
+                                           ntohs(param.p->length), gfp);
+               if (!asoc->peer.peer_hmacs) {
+                       retval = 0;
+                       break;
+               }
+
+               /* Set the default HMAC the peer requested*/
+               sctp_auth_asoc_set_default_hmac(asoc, param.hmac_algo);
+               break;
+
+       case SCTP_PARAM_CHUNKS:
+               if (!ep->auth_enable)
+                       goto fall_through;
+
+               asoc->peer.peer_chunks = kmemdup(param.p,
+                                           ntohs(param.p->length), gfp);
+               if (!asoc->peer.peer_chunks)
+                       retval = 0;
+               break;
+fall_through:
+       default:
+               /* Any unrecognized parameters should have been caught
+                * and handled by sctp_verify_param() which should be
+                * called prior to this routine.  Simply log the error
+                * here.
+                */
+               pr_debug("%s: ignoring param:%d for association:%p.\n",
+                        __func__, ntohs(param.p->type), asoc);
+               break;
+       }
+
+       return retval;
+}
+
+/* Select a new verification tag.  */
+__u32 sctp_generate_tag(const struct sctp_endpoint *ep)
+{
+       /* I believe that this random number generator complies with RFC1750.
+        * A tag of 0 is reserved for special cases (e.g. INIT).
+        */
+       __u32 x;
+
+       do {
+               get_random_bytes(&x, sizeof(__u32));
+       } while (x == 0);
+
+       return x;
+}
+
+/* Select an initial TSN to send during startup.  */
+__u32 sctp_generate_tsn(const struct sctp_endpoint *ep)
+{
+       __u32 retval;
+
+       get_random_bytes(&retval, sizeof(__u32));
+       return retval;
+}
+
+/*
+ * ADDIP 3.1.1 Address Configuration Change Chunk (ASCONF)
+ *      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
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     | Type = 0xC1   |  Chunk Flags  |      Chunk Length             |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                       Serial Number                           |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                    Address Parameter                          |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                     ASCONF Parameter #1                       |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     \                                                               \
+ *     /                             ....                              /
+ *     \                                                               \
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                     ASCONF Parameter #N                       |
+ *      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * Address Parameter and other parameter will not be wrapped in this function
+ */
+static struct sctp_chunk *sctp_make_asconf(struct sctp_association *asoc,
+                                          union sctp_addr *addr,
+                                          int vparam_len)
+{
+       sctp_addiphdr_t asconf;
+       struct sctp_chunk *retval;
+       int length = sizeof(asconf) + vparam_len;
+       union sctp_addr_param addrparam;
+       int addrlen;
+       struct sctp_af *af = sctp_get_af_specific(addr->v4.sin_family);
+
+       addrlen = af->to_addr_param(addr, &addrparam);
+       if (!addrlen)
+               return NULL;
+       length += addrlen;
+
+       /* Create the chunk.  */
+       retval = sctp_make_control(asoc, SCTP_CID_ASCONF, 0, length);
+       if (!retval)
+               return NULL;
+
+       asconf.serial = htonl(asoc->addip_serial++);
+
+       retval->subh.addip_hdr =
+               sctp_addto_chunk(retval, sizeof(asconf), &asconf);
+       retval->param_hdr.v =
+               sctp_addto_chunk(retval, addrlen, &addrparam);
+
+       return retval;
+}
+
+/* ADDIP
+ * 3.2.1 Add IP Address
+ *     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
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |        Type = 0xC001          |    Length = Variable          |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |               ASCONF-Request Correlation ID                   |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                       Address Parameter                       |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * 3.2.2 Delete IP Address
+ *     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
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |        Type = 0xC002          |    Length = Variable          |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |               ASCONF-Request Correlation ID                   |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                       Address Parameter                       |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ */
+struct sctp_chunk *sctp_make_asconf_update_ip(struct sctp_association *asoc,
+                                             union sctp_addr         *laddr,
+                                             struct sockaddr         *addrs,
+                                             int                     addrcnt,
+                                             __be16                  flags)
+{
+       sctp_addip_param_t      param;
+       struct sctp_chunk       *retval;
+       union sctp_addr_param   addr_param;
+       union sctp_addr         *addr;
+       void                    *addr_buf;
+       struct sctp_af          *af;
+       int                     paramlen = sizeof(param);
+       int                     addr_param_len = 0;
+       int                     totallen = 0;
+       int                     i;
+       int                     del_pickup = 0;
+
+       /* Get total length of all the address parameters. */
+       addr_buf = addrs;
+       for (i = 0; i < addrcnt; i++) {
+               addr = addr_buf;
+               af = sctp_get_af_specific(addr->v4.sin_family);
+               addr_param_len = af->to_addr_param(addr, &addr_param);
+
+               totallen += paramlen;
+               totallen += addr_param_len;
+
+               addr_buf += af->sockaddr_len;
+               if (asoc->asconf_addr_del_pending && !del_pickup) {
+                       /* reuse the parameter length from the same scope one */
+                       totallen += paramlen;
+                       totallen += addr_param_len;
+                       del_pickup = 1;
+
+                       pr_debug("%s: picked same-scope del_pending addr, "
+                                "totallen for all addresses is %d\n",
+                                __func__, totallen);
+               }
+       }
+
+       /* Create an asconf chunk with the required length. */
+       retval = sctp_make_asconf(asoc, laddr, totallen);
+       if (!retval)
+               return NULL;
+
+       /* Add the address parameters to the asconf chunk. */
+       addr_buf = addrs;
+       for (i = 0; i < addrcnt; i++) {
+               addr = addr_buf;
+               af = sctp_get_af_specific(addr->v4.sin_family);
+               addr_param_len = af->to_addr_param(addr, &addr_param);
+               param.param_hdr.type = flags;
+               param.param_hdr.length = htons(paramlen + addr_param_len);
+               param.crr_id = i;
+
+               sctp_addto_chunk(retval, paramlen, &param);
+               sctp_addto_chunk(retval, addr_param_len, &addr_param);
+
+               addr_buf += af->sockaddr_len;
+       }
+       if (flags == SCTP_PARAM_ADD_IP && del_pickup) {
+               addr = asoc->asconf_addr_del_pending;
+               af = sctp_get_af_specific(addr->v4.sin_family);
+               addr_param_len = af->to_addr_param(addr, &addr_param);
+               param.param_hdr.type = SCTP_PARAM_DEL_IP;
+               param.param_hdr.length = htons(paramlen + addr_param_len);
+               param.crr_id = i;
+
+               sctp_addto_chunk(retval, paramlen, &param);
+               sctp_addto_chunk(retval, addr_param_len, &addr_param);
+       }
+       return retval;
+}
+
+/* ADDIP
+ * 3.2.4 Set Primary IP Address
+ *     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
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |        Type =0xC004           |    Length = Variable          |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |               ASCONF-Request Correlation ID                   |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                       Address Parameter                       |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * Create an ASCONF chunk with Set Primary IP address parameter.
+ */
+struct sctp_chunk *sctp_make_asconf_set_prim(struct sctp_association *asoc,
+                                            union sctp_addr *addr)
+{
+       sctp_addip_param_t      param;
+       struct sctp_chunk       *retval;
+       int                     len = sizeof(param);
+       union sctp_addr_param   addrparam;
+       int                     addrlen;
+       struct sctp_af          *af = sctp_get_af_specific(addr->v4.sin_family);
+
+       addrlen = af->to_addr_param(addr, &addrparam);
+       if (!addrlen)
+               return NULL;
+       len += addrlen;
+
+       /* Create the chunk and make asconf header. */
+       retval = sctp_make_asconf(asoc, addr, len);
+       if (!retval)
+               return NULL;
+
+       param.param_hdr.type = SCTP_PARAM_SET_PRIMARY;
+       param.param_hdr.length = htons(len);
+       param.crr_id = 0;
+
+       sctp_addto_chunk(retval, sizeof(param), &param);
+       sctp_addto_chunk(retval, addrlen, &addrparam);
+
+       return retval;
+}
+
+/* ADDIP 3.1.2 Address Configuration Acknowledgement Chunk (ASCONF-ACK)
+ *      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
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     | Type = 0x80   |  Chunk Flags  |      Chunk Length             |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                       Serial Number                           |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                 ASCONF Parameter Response#1                   |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     \                                                               \
+ *     /                             ....                              /
+ *     \                                                               \
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *     |                 ASCONF Parameter Response#N                   |
+ *     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * Create an ASCONF_ACK chunk with enough space for the parameter responses.
+ */
+static struct sctp_chunk *sctp_make_asconf_ack(const struct sctp_association *asoc,
+                                              __u32 serial, int vparam_len)
+{
+       sctp_addiphdr_t         asconf;
+       struct sctp_chunk       *retval;
+       int                     length = sizeof(asconf) + vparam_len;
+
+       /* Create the chunk.  */
+       retval = sctp_make_control(asoc, SCTP_CID_ASCONF_ACK, 0, length);
+       if (!retval)
+               return NULL;
+
+       asconf.serial = htonl(serial);
+
+       retval->subh.addip_hdr =
+               sctp_addto_chunk(retval, sizeof(asconf), &asconf);
+
+       return retval;
+}
+
+/* Add response parameters to an ASCONF_ACK chunk. */
+static void sctp_add_asconf_response(struct sctp_chunk *chunk, __be32 crr_id,
+                             __be16 err_code, sctp_addip_param_t *asconf_param)
+{
+       sctp_addip_param_t      ack_param;
+       sctp_errhdr_t           err_param;
+       int                     asconf_param_len = 0;
+       int                     err_param_len = 0;
+       __be16                  response_type;
+
+       if (SCTP_ERROR_NO_ERROR == err_code) {
+               response_type = SCTP_PARAM_SUCCESS_REPORT;
+       } else {
+               response_type = SCTP_PARAM_ERR_CAUSE;
+               err_param_len = sizeof(err_param);
+               if (asconf_param)
+                       asconf_param_len =
+                                ntohs(asconf_param->param_hdr.length);
+       }
+
+       /* Add Success Indication or Error Cause Indication parameter. */
+       ack_param.param_hdr.type = response_type;
+       ack_param.param_hdr.length = htons(sizeof(ack_param) +
+                                          err_param_len +
+                                          asconf_param_len);
+       ack_param.crr_id = crr_id;
+       sctp_addto_chunk(chunk, sizeof(ack_param), &ack_param);
+
+       if (SCTP_ERROR_NO_ERROR == err_code)
+               return;
+
+       /* Add Error Cause parameter. */
+       err_param.cause = err_code;
+       err_param.length = htons(err_param_len + asconf_param_len);
+       sctp_addto_chunk(chunk, err_param_len, &err_param);
+
+       /* Add the failed TLV copied from ASCONF chunk. */
+       if (asconf_param)
+               sctp_addto_chunk(chunk, asconf_param_len, asconf_param);
+}
+
+/* Process a asconf parameter. */
+static __be16 sctp_process_asconf_param(struct sctp_association *asoc,
+                                      struct sctp_chunk *asconf,
+                                      sctp_addip_param_t *asconf_param)
+{
+       struct sctp_transport *peer;
+       struct sctp_af *af;
+       union sctp_addr addr;
+       union sctp_addr_param *addr_param;
+
+       addr_param = (void *)asconf_param + sizeof(sctp_addip_param_t);
+
+       if (asconf_param->param_hdr.type != SCTP_PARAM_ADD_IP &&
+           asconf_param->param_hdr.type != SCTP_PARAM_DEL_IP &&
+           asconf_param->param_hdr.type != SCTP_PARAM_SET_PRIMARY)
+               return SCTP_ERROR_UNKNOWN_PARAM;
+
+       switch (addr_param->p.type) {
+       case SCTP_PARAM_IPV6_ADDRESS:
+               if (!asoc->peer.ipv6_address)
+                       return SCTP_ERROR_DNS_FAILED;
+               break;
+       case SCTP_PARAM_IPV4_ADDRESS:
+               if (!asoc->peer.ipv4_address)
+                       return SCTP_ERROR_DNS_FAILED;
+               break;
+       default:
+               return SCTP_ERROR_DNS_FAILED;
+       }
+
+       af = sctp_get_af_specific(param_type2af(addr_param->p.type));
+       if (unlikely(!af))
+               return SCTP_ERROR_DNS_FAILED;
+
+       af->from_addr_param(&addr, addr_param, htons(asoc->peer.port), 0);
+
+       /* ADDIP 4.2.1  This parameter MUST NOT contain a broadcast
+        * or multicast address.
+        * (note: wildcard is permitted and requires special handling so
+        *  make sure we check for that)
+        */
+       if (!af->is_any(&addr) && !af->addr_valid(&addr, NULL, asconf->skb))
+               return SCTP_ERROR_DNS_FAILED;
+
+       switch (asconf_param->param_hdr.type) {
+       case SCTP_PARAM_ADD_IP:
+               /* Section 4.2.1:
+                * If the address 0.0.0.0 or ::0 is provided, the source
+                * address of the packet MUST be added.
+                */
+               if (af->is_any(&addr))
+                       memcpy(&addr, &asconf->source, sizeof(addr));
+
+               /* ADDIP 4.3 D9) If an endpoint receives an ADD IP address
+                * request and does not have the local resources to add this
+                * new address to the association, it MUST return an Error
+                * Cause TLV set to the new error code 'Operation Refused
+                * Due to Resource Shortage'.
+                */
+
+               peer = sctp_assoc_add_peer(asoc, &addr, GFP_ATOMIC, SCTP_UNCONFIRMED);
+               if (!peer)
+                       return SCTP_ERROR_RSRC_LOW;
+
+               /* Start the heartbeat timer. */
+               if (!mod_timer(&peer->hb_timer, sctp_transport_timeout(peer)))
+                       sctp_transport_hold(peer);
+               asoc->new_transport = peer;
+               break;
+       case SCTP_PARAM_DEL_IP:
+               /* ADDIP 4.3 D7) If a request is received to delete the
+                * last remaining IP address of a peer endpoint, the receiver
+                * MUST send an Error Cause TLV with the error cause set to the
+                * new error code 'Request to Delete Last Remaining IP Address'.
+                */
+               if (asoc->peer.transport_count == 1)
+                       return SCTP_ERROR_DEL_LAST_IP;
+
+               /* ADDIP 4.3 D8) If a request is received to delete an IP
+                * address which is also the source address of the IP packet
+                * which contained the ASCONF chunk, the receiver MUST reject
+                * this request. To reject the request the receiver MUST send
+                * an Error Cause TLV set to the new error code 'Request to
+                * Delete Source IP Address'
+                */
+               if (sctp_cmp_addr_exact(&asconf->source, &addr))
+                       return SCTP_ERROR_DEL_SRC_IP;
+
+               /* Section 4.2.2
+                * If the address 0.0.0.0 or ::0 is provided, all
+                * addresses of the peer except the source address of the
+                * packet MUST be deleted.
+                */
+               if (af->is_any(&addr)) {
+                       sctp_assoc_set_primary(asoc, asconf->transport);
+                       sctp_assoc_del_nonprimary_peers(asoc,
+                                                       asconf->transport);
+               } else
+                       sctp_assoc_del_peer(asoc, &addr);
+               break;
+       case SCTP_PARAM_SET_PRIMARY:
+               /* ADDIP Section 4.2.4
+                * If the address 0.0.0.0 or ::0 is provided, the receiver
+                * MAY mark the source address of the packet as its
+                * primary.
+                */
+               if (af->is_any(&addr))
+                       memcpy(&addr.v4, sctp_source(asconf), sizeof(addr));
+
+               peer = sctp_assoc_lookup_paddr(asoc, &addr);
+               if (!peer)
+                       return SCTP_ERROR_DNS_FAILED;
+
+               sctp_assoc_set_primary(asoc, peer);
+               break;
+       }
+
+       return SCTP_ERROR_NO_ERROR;
+}
+
+/* Verify the ASCONF packet before we process it. */
+bool sctp_verify_asconf(const struct sctp_association *asoc,
+                       struct sctp_chunk *chunk, bool addr_param_needed,
+                       struct sctp_paramhdr **errp)
+{
+       sctp_addip_chunk_t *addip = (sctp_addip_chunk_t *) chunk->chunk_hdr;
+       union sctp_params param;
+       bool addr_param_seen = false;
+
+       sctp_walk_params(param, addip, addip_hdr.params) {
+               size_t length = ntohs(param.p->length);
+
+               *errp = param.p;
+               switch (param.p->type) {
+               case SCTP_PARAM_ERR_CAUSE:
+                       break;
+               case SCTP_PARAM_IPV4_ADDRESS:
+                       if (length != sizeof(sctp_ipv4addr_param_t))
+                               return false;
+                       addr_param_seen = true;
+                       break;
+               case SCTP_PARAM_IPV6_ADDRESS:
+                       if (length != sizeof(sctp_ipv6addr_param_t))
+                               return false;
+                       addr_param_seen = true;
+                       break;
+               case SCTP_PARAM_ADD_IP:
+               case SCTP_PARAM_DEL_IP:
+               case SCTP_PARAM_SET_PRIMARY:
+                       /* In ASCONF chunks, these need to be first. */
+                       if (addr_param_needed && !addr_param_seen)
+                               return false;
+                       length = ntohs(param.addip->param_hdr.length);
+                       if (length < sizeof(sctp_addip_param_t) +
+                                    sizeof(sctp_paramhdr_t))
+                               return false;
+                       break;
+               case SCTP_PARAM_SUCCESS_REPORT:
+               case SCTP_PARAM_ADAPTATION_LAYER_IND:
+                       if (length != sizeof(sctp_addip_param_t))
+                               return false;
+                       break;
+               default:
+                       /* This is unkown to us, reject! */
+                       return false;
+               }
+       }
+
+       /* Remaining sanity checks. */
+       if (addr_param_needed && !addr_param_seen)
+               return false;
+       if (!addr_param_needed && addr_param_seen)
+               return false;
+       if (param.v != chunk->chunk_end)
+               return false;
+
+       return true;
+}
+
+/* Process an incoming ASCONF chunk with the next expected serial no. and
+ * return an ASCONF_ACK chunk to be sent in response.
+ */
+struct sctp_chunk *sctp_process_asconf(struct sctp_association *asoc,
+                                      struct sctp_chunk *asconf)
+{
+       sctp_addip_chunk_t *addip = (sctp_addip_chunk_t *) asconf->chunk_hdr;
+       bool all_param_pass = true;
+       union sctp_params param;
+       sctp_addiphdr_t         *hdr;
+       union sctp_addr_param   *addr_param;
+       sctp_addip_param_t      *asconf_param;
+       struct sctp_chunk       *asconf_ack;
+       __be16  err_code;
+       int     length = 0;
+       int     chunk_len;
+       __u32   serial;
+
+       chunk_len = ntohs(asconf->chunk_hdr->length) - sizeof(sctp_chunkhdr_t);
+       hdr = (sctp_addiphdr_t *)asconf->skb->data;
+       serial = ntohl(hdr->serial);
+
+       /* Skip the addiphdr and store a pointer to address parameter.  */
+       length = sizeof(sctp_addiphdr_t);
+       addr_param = (union sctp_addr_param *)(asconf->skb->data + length);
+       chunk_len -= length;
+
+       /* Skip the address parameter and store a pointer to the first
+        * asconf parameter.
+        */
+       length = ntohs(addr_param->p.length);
+       asconf_param = (void *)addr_param + length;
+       chunk_len -= length;
+
+       /* create an ASCONF_ACK chunk.
+        * Based on the definitions of parameters, we know that the size of
+        * ASCONF_ACK parameters are less than or equal to the fourfold of ASCONF
+        * parameters.
+        */
+       asconf_ack = sctp_make_asconf_ack(asoc, serial, chunk_len * 4);
+       if (!asconf_ack)
+               goto done;
+
+       /* Process the TLVs contained within the ASCONF chunk. */
+       sctp_walk_params(param, addip, addip_hdr.params) {
+               /* Skip preceeding address parameters. */
+               if (param.p->type == SCTP_PARAM_IPV4_ADDRESS ||
+                   param.p->type == SCTP_PARAM_IPV6_ADDRESS)
+                       continue;
+
+               err_code = sctp_process_asconf_param(asoc, asconf,
+                                                    param.addip);
+               /* ADDIP 4.1 A7)
+                * If an error response is received for a TLV parameter,
+                * all TLVs with no response before the failed TLV are
+                * considered successful if not reported.  All TLVs after
+                * the failed response are considered unsuccessful unless
+                * a specific success indication is present for the parameter.
+                */
+               if (err_code != SCTP_ERROR_NO_ERROR)
+                       all_param_pass = false;
+               if (!all_param_pass)
+                       sctp_add_asconf_response(asconf_ack, param.addip->crr_id,
+                                                err_code, param.addip);
+
+               /* ADDIP 4.3 D11) When an endpoint receiving an ASCONF to add
+                * an IP address sends an 'Out of Resource' in its response, it
+                * MUST also fail any subsequent add or delete requests bundled
+                * in the ASCONF.
+                */
+               if (err_code == SCTP_ERROR_RSRC_LOW)
+                       goto done;
+       }
+done:
+       asoc->peer.addip_serial++;
+
+       /* If we are sending a new ASCONF_ACK hold a reference to it in assoc
+        * after freeing the reference to old asconf ack if any.
+        */
+       if (asconf_ack) {
+               sctp_chunk_hold(asconf_ack);
+               list_add_tail(&asconf_ack->transmitted_list,
+                             &asoc->asconf_ack_list);
+       }
+
+       return asconf_ack;
+}
+
+/* Process a asconf parameter that is successfully acked. */
+static void sctp_asconf_param_success(struct sctp_association *asoc,
+                                    sctp_addip_param_t *asconf_param)
+{
+       struct sctp_af *af;
+       union sctp_addr addr;
+       struct sctp_bind_addr *bp = &asoc->base.bind_addr;
+       union sctp_addr_param *addr_param;
+       struct sctp_transport *transport;
+       struct sctp_sockaddr_entry *saddr;
+
+       addr_param = (void *)asconf_param + sizeof(sctp_addip_param_t);
+
+       /* We have checked the packet before, so we do not check again. */
+       af = sctp_get_af_specific(param_type2af(addr_param->p.type));
+       af->from_addr_param(&addr, addr_param, htons(bp->port), 0);
+
+       switch (asconf_param->param_hdr.type) {
+       case SCTP_PARAM_ADD_IP:
+               /* This is always done in BH context with a socket lock
+                * held, so the list can not change.
+                */
+               local_bh_disable();
+               list_for_each_entry(saddr, &bp->address_list, list) {
+                       if (sctp_cmp_addr_exact(&saddr->a, &addr))
+                               saddr->state = SCTP_ADDR_SRC;
+               }
+               local_bh_enable();
+               list_for_each_entry(transport, &asoc->peer.transport_addr_list,
+                               transports) {
+                       dst_release(transport->dst);
+                       transport->dst = NULL;
+               }
+               break;
+       case SCTP_PARAM_DEL_IP:
+               local_bh_disable();
+               sctp_del_bind_addr(bp, &addr);
+               if (asoc->asconf_addr_del_pending != NULL &&
+                   sctp_cmp_addr_exact(asoc->asconf_addr_del_pending, &addr)) {
+                       kfree(asoc->asconf_addr_del_pending);
+                       asoc->asconf_addr_del_pending = NULL;
+               }
+               local_bh_enable();
+               list_for_each_entry(transport, &asoc->peer.transport_addr_list,
+                               transports) {
+                       dst_release(transport->dst);
+                       transport->dst = NULL;
+               }
+               break;
+       default:
+               break;
+       }
+}
+
+/* Get the corresponding ASCONF response error code from the ASCONF_ACK chunk
+ * for the given asconf parameter.  If there is no response for this parameter,
+ * return the error code based on the third argument 'no_err'.
+ * ADDIP 4.1
+ * A7) If an error response is received for a TLV parameter, all TLVs with no
+ * response before the failed TLV are considered successful if not reported.
+ * All TLVs after the failed response are considered unsuccessful unless a
+ * specific success indication is present for the parameter.
+ */
+static __be16 sctp_get_asconf_response(struct sctp_chunk *asconf_ack,
+                                     sctp_addip_param_t *asconf_param,
+                                     int no_err)
+{
+       sctp_addip_param_t      *asconf_ack_param;
+       sctp_errhdr_t           *err_param;
+       int                     length;
+       int                     asconf_ack_len;
+       __be16                  err_code;
+
+       if (no_err)
+               err_code = SCTP_ERROR_NO_ERROR;
+       else
+               err_code = SCTP_ERROR_REQ_REFUSED;
+
+       asconf_ack_len = ntohs(asconf_ack->chunk_hdr->length) -
+                            sizeof(sctp_chunkhdr_t);
+
+       /* Skip the addiphdr from the asconf_ack chunk and store a pointer to
+        * the first asconf_ack parameter.
+        */
+       length = sizeof(sctp_addiphdr_t);
+       asconf_ack_param = (sctp_addip_param_t *)(asconf_ack->skb->data +
+                                                 length);
+       asconf_ack_len -= length;
+
+       while (asconf_ack_len > 0) {
+               if (asconf_ack_param->crr_id == asconf_param->crr_id) {
+                       switch (asconf_ack_param->param_hdr.type) {
+                       case SCTP_PARAM_SUCCESS_REPORT:
+                               return SCTP_ERROR_NO_ERROR;
+                       case SCTP_PARAM_ERR_CAUSE:
+                               length = sizeof(sctp_addip_param_t);
+                               err_param = (void *)asconf_ack_param + length;
+                               asconf_ack_len -= length;
+                               if (asconf_ack_len > 0)
+                                       return err_param->cause;
+                               else
+                                       return SCTP_ERROR_INV_PARAM;
+                               break;
+                       default:
+                               return SCTP_ERROR_INV_PARAM;
+                       }
+               }
+
+               length = ntohs(asconf_ack_param->param_hdr.length);
+               asconf_ack_param = (void *)asconf_ack_param + length;
+               asconf_ack_len -= length;
+       }
+
+       return err_code;
+}
+
+/* Process an incoming ASCONF_ACK chunk against the cached last ASCONF chunk. */
+int sctp_process_asconf_ack(struct sctp_association *asoc,
+                           struct sctp_chunk *asconf_ack)
+{
+       struct sctp_chunk       *asconf = asoc->addip_last_asconf;
+       union sctp_addr_param   *addr_param;
+       sctp_addip_param_t      *asconf_param;
+       int     length = 0;
+       int     asconf_len = asconf->skb->len;
+       int     all_param_pass = 0;
+       int     no_err = 1;
+       int     retval = 0;
+       __be16  err_code = SCTP_ERROR_NO_ERROR;
+
+       /* Skip the chunkhdr and addiphdr from the last asconf sent and store
+        * a pointer to address parameter.
+        */
+       length = sizeof(sctp_addip_chunk_t);
+       addr_param = (union sctp_addr_param *)(asconf->skb->data + length);
+       asconf_len -= length;
+
+       /* Skip the address parameter in the last asconf sent and store a
+        * pointer to the first asconf parameter.
+        */
+       length = ntohs(addr_param->p.length);
+       asconf_param = (void *)addr_param + length;
+       asconf_len -= length;
+
+       /* ADDIP 4.1
+        * A8) If there is no response(s) to specific TLV parameter(s), and no
+        * failures are indicated, then all request(s) are considered
+        * successful.
+        */
+       if (asconf_ack->skb->len == sizeof(sctp_addiphdr_t))
+               all_param_pass = 1;
+
+       /* Process the TLVs contained in the last sent ASCONF chunk. */
+       while (asconf_len > 0) {
+               if (all_param_pass)
+                       err_code = SCTP_ERROR_NO_ERROR;
+               else {
+                       err_code = sctp_get_asconf_response(asconf_ack,
+                                                           asconf_param,
+                                                           no_err);
+                       if (no_err && (SCTP_ERROR_NO_ERROR != err_code))
+                               no_err = 0;
+               }
+
+               switch (err_code) {
+               case SCTP_ERROR_NO_ERROR:
+                       sctp_asconf_param_success(asoc, asconf_param);
+                       break;
+
+               case SCTP_ERROR_RSRC_LOW:
+                       retval = 1;
+                       break;
+
+               case SCTP_ERROR_UNKNOWN_PARAM:
+                       /* Disable sending this type of asconf parameter in
+                        * future.
+                        */
+                       asoc->peer.addip_disabled_mask |=
+                               asconf_param->param_hdr.type;
+                       break;
+
+               case SCTP_ERROR_REQ_REFUSED:
+               case SCTP_ERROR_DEL_LAST_IP:
+               case SCTP_ERROR_DEL_SRC_IP:
+               default:
+                        break;
+               }
+
+               /* Skip the processed asconf parameter and move to the next
+                * one.
+                */
+               length = ntohs(asconf_param->param_hdr.length);
+               asconf_param = (void *)asconf_param + length;
+               asconf_len -= length;
+       }
+
+       if (no_err && asoc->src_out_of_asoc_ok) {
+               asoc->src_out_of_asoc_ok = 0;
+               sctp_transport_immediate_rtx(asoc->peer.primary_path);
+       }
+
+       /* Free the cached last sent asconf chunk. */
+       list_del_init(&asconf->transmitted_list);
+       sctp_chunk_free(asconf);
+       asoc->addip_last_asconf = NULL;
+
+       return retval;
+}
+
+/* Make a FWD TSN chunk. */
+struct sctp_chunk *sctp_make_fwdtsn(const struct sctp_association *asoc,
+                                   __u32 new_cum_tsn, size_t nstreams,
+                                   struct sctp_fwdtsn_skip *skiplist)
+{
+       struct sctp_chunk *retval = NULL;
+       struct sctp_fwdtsn_hdr ftsn_hdr;
+       struct sctp_fwdtsn_skip skip;
+       size_t hint;
+       int i;
+
+       hint = (nstreams + 1) * sizeof(__u32);
+
+       retval = sctp_make_control(asoc, SCTP_CID_FWD_TSN, 0, hint);
+
+       if (!retval)
+               return NULL;
+
+       ftsn_hdr.new_cum_tsn = htonl(new_cum_tsn);
+       retval->subh.fwdtsn_hdr =
+               sctp_addto_chunk(retval, sizeof(ftsn_hdr), &ftsn_hdr);
+
+       for (i = 0; i < nstreams; i++) {
+               skip.stream = skiplist[i].stream;
+               skip.ssn = skiplist[i].ssn;
+               sctp_addto_chunk(retval, sizeof(skip), &skip);
+       }
+
+       return retval;
+}