Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / sctp / output.c
diff --git a/kernel/net/sctp/output.c b/kernel/net/sctp/output.c
new file mode 100644 (file)
index 0000000..abe7c2d
--- /dev/null
@@ -0,0 +1,788 @@
+/* SCTP kernel implementation
+ * (C) Copyright IBM Corp. 2001, 2004
+ * Copyright (c) 1999-2000 Cisco, Inc.
+ * Copyright (c) 1999-2001 Motorola, Inc.
+ *
+ * This file is part of the SCTP kernel implementation
+ *
+ * These functions handle output processing.
+ *
+ * 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>
+ *    Jon Grimm             <jgrimm@austin.ibm.com>
+ *    Sridhar Samudrala     <sri@us.ibm.com>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/wait.h>
+#include <linux/time.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <net/inet_ecn.h>
+#include <net/ip.h>
+#include <net/icmp.h>
+#include <net/net_namespace.h>
+
+#include <linux/socket.h> /* for sa_family_t */
+#include <net/sock.h>
+
+#include <net/sctp/sctp.h>
+#include <net/sctp/sm.h>
+#include <net/sctp/checksum.h>
+
+/* Forward declarations for private helpers. */
+static sctp_xmit_t __sctp_packet_append_chunk(struct sctp_packet *packet,
+                                             struct sctp_chunk *chunk);
+static sctp_xmit_t sctp_packet_can_append_data(struct sctp_packet *packet,
+                                          struct sctp_chunk *chunk);
+static void sctp_packet_append_data(struct sctp_packet *packet,
+                                          struct sctp_chunk *chunk);
+static sctp_xmit_t sctp_packet_will_fit(struct sctp_packet *packet,
+                                       struct sctp_chunk *chunk,
+                                       u16 chunk_len);
+
+static void sctp_packet_reset(struct sctp_packet *packet)
+{
+       packet->size = packet->overhead;
+       packet->has_cookie_echo = 0;
+       packet->has_sack = 0;
+       packet->has_data = 0;
+       packet->has_auth = 0;
+       packet->ipfragok = 0;
+       packet->auth = NULL;
+}
+
+/* Config a packet.
+ * This appears to be a followup set of initializations.
+ */
+struct sctp_packet *sctp_packet_config(struct sctp_packet *packet,
+                                      __u32 vtag, int ecn_capable)
+{
+       struct sctp_chunk *chunk = NULL;
+
+       pr_debug("%s: packet:%p vtag:0x%x\n", __func__, packet, vtag);
+
+       packet->vtag = vtag;
+
+       if (ecn_capable && sctp_packet_empty(packet)) {
+               chunk = sctp_get_ecne_prepend(packet->transport->asoc);
+
+               /* If there a is a prepend chunk stick it on the list before
+                * any other chunks get appended.
+                */
+               if (chunk)
+                       sctp_packet_append_chunk(packet, chunk);
+       }
+
+       return packet;
+}
+
+/* Initialize the packet structure. */
+struct sctp_packet *sctp_packet_init(struct sctp_packet *packet,
+                                    struct sctp_transport *transport,
+                                    __u16 sport, __u16 dport)
+{
+       struct sctp_association *asoc = transport->asoc;
+       size_t overhead;
+
+       pr_debug("%s: packet:%p transport:%p\n", __func__, packet, transport);
+
+       packet->transport = transport;
+       packet->source_port = sport;
+       packet->destination_port = dport;
+       INIT_LIST_HEAD(&packet->chunk_list);
+       if (asoc) {
+               struct sctp_sock *sp = sctp_sk(asoc->base.sk);
+               overhead = sp->pf->af->net_header_len;
+       } else {
+               overhead = sizeof(struct ipv6hdr);
+       }
+       overhead += sizeof(struct sctphdr);
+       packet->overhead = overhead;
+       sctp_packet_reset(packet);
+       packet->vtag = 0;
+
+       return packet;
+}
+
+/* Free a packet.  */
+void sctp_packet_free(struct sctp_packet *packet)
+{
+       struct sctp_chunk *chunk, *tmp;
+
+       pr_debug("%s: packet:%p\n", __func__, packet);
+
+       list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) {
+               list_del_init(&chunk->list);
+               sctp_chunk_free(chunk);
+       }
+}
+
+/* This routine tries to append the chunk to the offered packet. If adding
+ * the chunk causes the packet to exceed the path MTU and COOKIE_ECHO chunk
+ * is not present in the packet, it transmits the input packet.
+ * Data can be bundled with a packet containing a COOKIE_ECHO chunk as long
+ * as it can fit in the packet, but any more data that does not fit in this
+ * packet can be sent only after receiving the COOKIE_ACK.
+ */
+sctp_xmit_t sctp_packet_transmit_chunk(struct sctp_packet *packet,
+                                      struct sctp_chunk *chunk,
+                                      int one_packet)
+{
+       sctp_xmit_t retval;
+       int error = 0;
+
+       pr_debug("%s: packet:%p chunk:%p\n", __func__, packet, chunk);
+
+       switch ((retval = (sctp_packet_append_chunk(packet, chunk)))) {
+       case SCTP_XMIT_PMTU_FULL:
+               if (!packet->has_cookie_echo) {
+                       error = sctp_packet_transmit(packet);
+                       if (error < 0)
+                               chunk->skb->sk->sk_err = -error;
+
+                       /* If we have an empty packet, then we can NOT ever
+                        * return PMTU_FULL.
+                        */
+                       if (!one_packet)
+                               retval = sctp_packet_append_chunk(packet,
+                                                                 chunk);
+               }
+               break;
+
+       case SCTP_XMIT_RWND_FULL:
+       case SCTP_XMIT_OK:
+       case SCTP_XMIT_DELAY:
+               break;
+       }
+
+       return retval;
+}
+
+/* Try to bundle an auth chunk into the packet. */
+static sctp_xmit_t sctp_packet_bundle_auth(struct sctp_packet *pkt,
+                                          struct sctp_chunk *chunk)
+{
+       struct sctp_association *asoc = pkt->transport->asoc;
+       struct sctp_chunk *auth;
+       sctp_xmit_t retval = SCTP_XMIT_OK;
+
+       /* if we don't have an association, we can't do authentication */
+       if (!asoc)
+               return retval;
+
+       /* See if this is an auth chunk we are bundling or if
+        * auth is already bundled.
+        */
+       if (chunk->chunk_hdr->type == SCTP_CID_AUTH || pkt->has_auth)
+               return retval;
+
+       /* if the peer did not request this chunk to be authenticated,
+        * don't do it
+        */
+       if (!chunk->auth)
+               return retval;
+
+       auth = sctp_make_auth(asoc);
+       if (!auth)
+               return retval;
+
+       retval = __sctp_packet_append_chunk(pkt, auth);
+
+       if (retval != SCTP_XMIT_OK)
+               sctp_chunk_free(auth);
+
+       return retval;
+}
+
+/* Try to bundle a SACK with the packet. */
+static sctp_xmit_t sctp_packet_bundle_sack(struct sctp_packet *pkt,
+                                          struct sctp_chunk *chunk)
+{
+       sctp_xmit_t retval = SCTP_XMIT_OK;
+
+       /* If sending DATA and haven't aleady bundled a SACK, try to
+        * bundle one in to the packet.
+        */
+       if (sctp_chunk_is_data(chunk) && !pkt->has_sack &&
+           !pkt->has_cookie_echo) {
+               struct sctp_association *asoc;
+               struct timer_list *timer;
+               asoc = pkt->transport->asoc;
+               timer = &asoc->timers[SCTP_EVENT_TIMEOUT_SACK];
+
+               /* If the SACK timer is running, we have a pending SACK */
+               if (timer_pending(timer)) {
+                       struct sctp_chunk *sack;
+
+                       if (pkt->transport->sack_generation !=
+                           pkt->transport->asoc->peer.sack_generation)
+                               return retval;
+
+                       asoc->a_rwnd = asoc->rwnd;
+                       sack = sctp_make_sack(asoc);
+                       if (sack) {
+                               retval = __sctp_packet_append_chunk(pkt, sack);
+                               if (retval != SCTP_XMIT_OK) {
+                                       sctp_chunk_free(sack);
+                                       goto out;
+                               }
+                               asoc->peer.sack_needed = 0;
+                               if (del_timer(timer))
+                                       sctp_association_put(asoc);
+                       }
+               }
+       }
+out:
+       return retval;
+}
+
+
+/* Append a chunk to the offered packet reporting back any inability to do
+ * so.
+ */
+static sctp_xmit_t __sctp_packet_append_chunk(struct sctp_packet *packet,
+                                             struct sctp_chunk *chunk)
+{
+       sctp_xmit_t retval = SCTP_XMIT_OK;
+       __u16 chunk_len = WORD_ROUND(ntohs(chunk->chunk_hdr->length));
+
+       /* Check to see if this chunk will fit into the packet */
+       retval = sctp_packet_will_fit(packet, chunk, chunk_len);
+       if (retval != SCTP_XMIT_OK)
+               goto finish;
+
+       /* We believe that this chunk is OK to add to the packet */
+       switch (chunk->chunk_hdr->type) {
+       case SCTP_CID_DATA:
+               /* Account for the data being in the packet */
+               sctp_packet_append_data(packet, chunk);
+               /* Disallow SACK bundling after DATA. */
+               packet->has_sack = 1;
+               /* Disallow AUTH bundling after DATA */
+               packet->has_auth = 1;
+               /* Let it be knows that packet has DATA in it */
+               packet->has_data = 1;
+               /* timestamp the chunk for rtx purposes */
+               chunk->sent_at = jiffies;
+               break;
+       case SCTP_CID_COOKIE_ECHO:
+               packet->has_cookie_echo = 1;
+               break;
+
+       case SCTP_CID_SACK:
+               packet->has_sack = 1;
+               if (chunk->asoc)
+                       chunk->asoc->stats.osacks++;
+               break;
+
+       case SCTP_CID_AUTH:
+               packet->has_auth = 1;
+               packet->auth = chunk;
+               break;
+       }
+
+       /* It is OK to send this chunk.  */
+       list_add_tail(&chunk->list, &packet->chunk_list);
+       packet->size += chunk_len;
+       chunk->transport = packet->transport;
+finish:
+       return retval;
+}
+
+/* Append a chunk to the offered packet reporting back any inability to do
+ * so.
+ */
+sctp_xmit_t sctp_packet_append_chunk(struct sctp_packet *packet,
+                                    struct sctp_chunk *chunk)
+{
+       sctp_xmit_t retval = SCTP_XMIT_OK;
+
+       pr_debug("%s: packet:%p chunk:%p\n", __func__, packet, chunk);
+
+       /* Data chunks are special.  Before seeing what else we can
+        * bundle into this packet, check to see if we are allowed to
+        * send this DATA.
+        */
+       if (sctp_chunk_is_data(chunk)) {
+               retval = sctp_packet_can_append_data(packet, chunk);
+               if (retval != SCTP_XMIT_OK)
+                       goto finish;
+       }
+
+       /* Try to bundle AUTH chunk */
+       retval = sctp_packet_bundle_auth(packet, chunk);
+       if (retval != SCTP_XMIT_OK)
+               goto finish;
+
+       /* Try to bundle SACK chunk */
+       retval = sctp_packet_bundle_sack(packet, chunk);
+       if (retval != SCTP_XMIT_OK)
+               goto finish;
+
+       retval = __sctp_packet_append_chunk(packet, chunk);
+
+finish:
+       return retval;
+}
+
+static void sctp_packet_release_owner(struct sk_buff *skb)
+{
+       sk_free(skb->sk);
+}
+
+static void sctp_packet_set_owner_w(struct sk_buff *skb, struct sock *sk)
+{
+       skb_orphan(skb);
+       skb->sk = sk;
+       skb->destructor = sctp_packet_release_owner;
+
+       /*
+        * The data chunks have already been accounted for in sctp_sendmsg(),
+        * therefore only reserve a single byte to keep socket around until
+        * the packet has been transmitted.
+        */
+       atomic_inc(&sk->sk_wmem_alloc);
+}
+
+/* All packets are sent to the network through this function from
+ * sctp_outq_tail().
+ *
+ * The return value is a normal kernel error return value.
+ */
+int sctp_packet_transmit(struct sctp_packet *packet)
+{
+       struct sctp_transport *tp = packet->transport;
+       struct sctp_association *asoc = tp->asoc;
+       struct sctphdr *sh;
+       struct sk_buff *nskb;
+       struct sctp_chunk *chunk, *tmp;
+       struct sock *sk;
+       int err = 0;
+       int padding;            /* How much padding do we need?  */
+       __u8 has_data = 0;
+       struct dst_entry *dst;
+       unsigned char *auth = NULL;     /* pointer to auth in skb data */
+
+       pr_debug("%s: packet:%p\n", __func__, packet);
+
+       /* Do NOT generate a chunkless packet. */
+       if (list_empty(&packet->chunk_list))
+               return err;
+
+       /* Set up convenience variables... */
+       chunk = list_entry(packet->chunk_list.next, struct sctp_chunk, list);
+       sk = chunk->skb->sk;
+
+       /* Allocate the new skb.  */
+       nskb = alloc_skb(packet->size + MAX_HEADER, GFP_ATOMIC);
+       if (!nskb)
+               goto nomem;
+
+       /* Make sure the outbound skb has enough header room reserved. */
+       skb_reserve(nskb, packet->overhead + MAX_HEADER);
+
+       /* Set the owning socket so that we know where to get the
+        * destination IP address.
+        */
+       sctp_packet_set_owner_w(nskb, sk);
+
+       if (!sctp_transport_dst_check(tp)) {
+               sctp_transport_route(tp, NULL, sctp_sk(sk));
+               if (asoc && (asoc->param_flags & SPP_PMTUD_ENABLE)) {
+                       sctp_assoc_sync_pmtu(sk, asoc);
+               }
+       }
+       dst = dst_clone(tp->dst);
+       if (!dst)
+               goto no_route;
+       skb_dst_set(nskb, dst);
+
+       /* Build the SCTP header.  */
+       sh = (struct sctphdr *)skb_push(nskb, sizeof(struct sctphdr));
+       skb_reset_transport_header(nskb);
+       sh->source = htons(packet->source_port);
+       sh->dest   = htons(packet->destination_port);
+
+       /* From 6.8 Adler-32 Checksum Calculation:
+        * After the packet is constructed (containing the SCTP common
+        * header and one or more control or DATA chunks), the
+        * transmitter shall:
+        *
+        * 1) Fill in the proper Verification Tag in the SCTP common
+        *    header and initialize the checksum field to 0's.
+        */
+       sh->vtag     = htonl(packet->vtag);
+       sh->checksum = 0;
+
+       /**
+        * 6.10 Bundling
+        *
+        *    An endpoint bundles chunks by simply including multiple
+        *    chunks in one outbound SCTP packet.  ...
+        */
+
+       /**
+        * 3.2  Chunk Field Descriptions
+        *
+        * The total length of a chunk (including Type, Length and
+        * Value fields) MUST be a multiple of 4 bytes.  If the length
+        * of the chunk is not a multiple of 4 bytes, the sender MUST
+        * pad the chunk with all zero bytes and this padding is not
+        * included in the chunk length field.  The sender should
+        * never pad with more than 3 bytes.
+        *
+        * [This whole comment explains WORD_ROUND() below.]
+        */
+
+       pr_debug("***sctp_transmit_packet***\n");
+
+       list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) {
+               list_del_init(&chunk->list);
+               if (sctp_chunk_is_data(chunk)) {
+                       /* 6.3.1 C4) When data is in flight and when allowed
+                        * by rule C5, a new RTT measurement MUST be made each
+                        * round trip.  Furthermore, new RTT measurements
+                        * SHOULD be made no more than once per round-trip
+                        * for a given destination transport address.
+                        */
+
+                       if (!chunk->resent && !tp->rto_pending) {
+                               chunk->rtt_in_progress = 1;
+                               tp->rto_pending = 1;
+                       }
+
+                       has_data = 1;
+               }
+
+               padding = WORD_ROUND(chunk->skb->len) - chunk->skb->len;
+               if (padding)
+                       memset(skb_put(chunk->skb, padding), 0, padding);
+
+               /* if this is the auth chunk that we are adding,
+                * store pointer where it will be added and put
+                * the auth into the packet.
+                */
+               if (chunk == packet->auth)
+                       auth = skb_tail_pointer(nskb);
+
+               memcpy(skb_put(nskb, chunk->skb->len),
+                              chunk->skb->data, chunk->skb->len);
+
+               pr_debug("*** Chunk:%p[%s] %s 0x%x, length:%d, chunk->skb->len:%d, "
+                        "rtt_in_progress:%d\n", chunk,
+                        sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)),
+                        chunk->has_tsn ? "TSN" : "No TSN",
+                        chunk->has_tsn ? ntohl(chunk->subh.data_hdr->tsn) : 0,
+                        ntohs(chunk->chunk_hdr->length), chunk->skb->len,
+                        chunk->rtt_in_progress);
+
+               /*
+                * If this is a control chunk, this is our last
+                * reference. Free data chunks after they've been
+                * acknowledged or have failed.
+                */
+               if (!sctp_chunk_is_data(chunk))
+                       sctp_chunk_free(chunk);
+       }
+
+       /* SCTP-AUTH, Section 6.2
+        *    The sender MUST calculate the MAC as described in RFC2104 [2]
+        *    using the hash function H as described by the MAC Identifier and
+        *    the shared association key K based on the endpoint pair shared key
+        *    described by the shared key identifier.  The 'data' used for the
+        *    computation of the AUTH-chunk is given by the AUTH chunk with its
+        *    HMAC field set to zero (as shown in Figure 6) followed by all
+        *    chunks that are placed after the AUTH chunk in the SCTP packet.
+        */
+       if (auth)
+               sctp_auth_calculate_hmac(asoc, nskb,
+                                       (struct sctp_auth_chunk *)auth,
+                                       GFP_ATOMIC);
+
+       /* 2) Calculate the Adler-32 checksum of the whole packet,
+        *    including the SCTP common header and all the
+        *    chunks.
+        *
+        * Note: Adler-32 is no longer applicable, as has been replaced
+        * by CRC32-C as described in <draft-ietf-tsvwg-sctpcsum-02.txt>.
+        */
+       if (!sctp_checksum_disable) {
+               if (!(dst->dev->features & NETIF_F_SCTP_CSUM) ||
+                   (dst_xfrm(dst) != NULL) || packet->ipfragok) {
+                       sh->checksum = sctp_compute_cksum(nskb, 0);
+               } else {
+                       /* no need to seed pseudo checksum for SCTP */
+                       nskb->ip_summed = CHECKSUM_PARTIAL;
+                       nskb->csum_start = skb_transport_header(nskb) - nskb->head;
+                       nskb->csum_offset = offsetof(struct sctphdr, checksum);
+               }
+       }
+
+       /* IP layer ECN support
+        * From RFC 2481
+        *  "The ECN-Capable Transport (ECT) bit would be set by the
+        *   data sender to indicate that the end-points of the
+        *   transport protocol are ECN-capable."
+        *
+        * Now setting the ECT bit all the time, as it should not cause
+        * any problems protocol-wise even if our peer ignores it.
+        *
+        * Note: The works for IPv6 layer checks this bit too later
+        * in transmission.  See IP6_ECN_flow_xmit().
+        */
+       tp->af_specific->ecn_capable(nskb->sk);
+
+       /* Set up the IP options.  */
+       /* BUG: not implemented
+        * For v4 this all lives somewhere in sk->sk_opt...
+        */
+
+       /* Dump that on IP!  */
+       if (asoc) {
+               asoc->stats.opackets++;
+               if (asoc->peer.last_sent_to != tp)
+                       /* Considering the multiple CPU scenario, this is a
+                        * "correcter" place for last_sent_to.  --xguo
+                        */
+                       asoc->peer.last_sent_to = tp;
+       }
+
+       if (has_data) {
+               struct timer_list *timer;
+               unsigned long timeout;
+
+               /* Restart the AUTOCLOSE timer when sending data. */
+               if (sctp_state(asoc, ESTABLISHED) &&
+                   asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]) {
+                       timer = &asoc->timers[SCTP_EVENT_TIMEOUT_AUTOCLOSE];
+                       timeout = asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE];
+
+                       if (!mod_timer(timer, jiffies + timeout))
+                               sctp_association_hold(asoc);
+               }
+       }
+
+       pr_debug("***sctp_transmit_packet*** skb->len:%d\n", nskb->len);
+
+       nskb->ignore_df = packet->ipfragok;
+       tp->af_specific->sctp_xmit(nskb, tp);
+
+out:
+       sctp_packet_reset(packet);
+       return err;
+no_route:
+       kfree_skb(nskb);
+
+       if (asoc)
+               IP_INC_STATS(sock_net(asoc->base.sk), IPSTATS_MIB_OUTNOROUTES);
+
+       /* FIXME: Returning the 'err' will effect all the associations
+        * associated with a socket, although only one of the paths of the
+        * association is unreachable.
+        * The real failure of a transport or association can be passed on
+        * to the user via notifications. So setting this error may not be
+        * required.
+        */
+        /* err = -EHOSTUNREACH; */
+err:
+       /* Control chunks are unreliable so just drop them.  DATA chunks
+        * will get resent or dropped later.
+        */
+
+       list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) {
+               list_del_init(&chunk->list);
+               if (!sctp_chunk_is_data(chunk))
+                       sctp_chunk_free(chunk);
+       }
+       goto out;
+nomem:
+       err = -ENOMEM;
+       goto err;
+}
+
+/********************************************************************
+ * 2nd Level Abstractions
+ ********************************************************************/
+
+/* This private function check to see if a chunk can be added */
+static sctp_xmit_t sctp_packet_can_append_data(struct sctp_packet *packet,
+                                          struct sctp_chunk *chunk)
+{
+       size_t datasize, rwnd, inflight, flight_size;
+       struct sctp_transport *transport = packet->transport;
+       struct sctp_association *asoc = transport->asoc;
+       struct sctp_outq *q = &asoc->outqueue;
+
+       /* RFC 2960 6.1  Transmission of DATA Chunks
+        *
+        * A) At any given time, the data sender MUST NOT transmit new data to
+        * any destination transport address if its peer's rwnd indicates
+        * that the peer has no buffer space (i.e. rwnd is 0, see Section
+        * 6.2.1).  However, regardless of the value of rwnd (including if it
+        * is 0), the data sender can always have one DATA chunk in flight to
+        * the receiver if allowed by cwnd (see rule B below).  This rule
+        * allows the sender to probe for a change in rwnd that the sender
+        * missed due to the SACK having been lost in transit from the data
+        * receiver to the data sender.
+        */
+
+       rwnd = asoc->peer.rwnd;
+       inflight = q->outstanding_bytes;
+       flight_size = transport->flight_size;
+
+       datasize = sctp_data_size(chunk);
+
+       if (datasize > rwnd && inflight > 0)
+               /* We have (at least) one data chunk in flight,
+                * so we can't fall back to rule 6.1 B).
+                */
+               return SCTP_XMIT_RWND_FULL;
+
+       /* RFC 2960 6.1  Transmission of DATA Chunks
+        *
+        * B) At any given time, the sender MUST NOT transmit new data
+        * to a given transport address if it has cwnd or more bytes
+        * of data outstanding to that transport address.
+        */
+       /* RFC 7.2.4 & the Implementers Guide 2.8.
+        *
+        * 3) ...
+        *    When a Fast Retransmit is being performed the sender SHOULD
+        *    ignore the value of cwnd and SHOULD NOT delay retransmission.
+        */
+       if (chunk->fast_retransmit != SCTP_NEED_FRTX &&
+           flight_size >= transport->cwnd)
+               return SCTP_XMIT_RWND_FULL;
+
+       /* Nagle's algorithm to solve small-packet problem:
+        * Inhibit the sending of new chunks when new outgoing data arrives
+        * if any previously transmitted data on the connection remains
+        * unacknowledged.
+        */
+
+       if (sctp_sk(asoc->base.sk)->nodelay)
+               /* Nagle disabled */
+               return SCTP_XMIT_OK;
+
+       if (!sctp_packet_empty(packet))
+               /* Append to packet */
+               return SCTP_XMIT_OK;
+
+       if (inflight == 0)
+               /* Nothing unacked */
+               return SCTP_XMIT_OK;
+
+       if (!sctp_state(asoc, ESTABLISHED))
+               return SCTP_XMIT_OK;
+
+       /* Check whether this chunk and all the rest of pending data will fit
+        * or delay in hopes of bundling a full sized packet.
+        */
+       if (chunk->skb->len + q->out_qlen >= transport->pathmtu - packet->overhead)
+               /* Enough data queued to fill a packet */
+               return SCTP_XMIT_OK;
+
+       /* Don't delay large message writes that may have been fragmented */
+       if (!chunk->msg->can_delay)
+               return SCTP_XMIT_OK;
+
+       /* Defer until all data acked or packet full */
+       return SCTP_XMIT_DELAY;
+}
+
+/* This private function does management things when adding DATA chunk */
+static void sctp_packet_append_data(struct sctp_packet *packet,
+                               struct sctp_chunk *chunk)
+{
+       struct sctp_transport *transport = packet->transport;
+       size_t datasize = sctp_data_size(chunk);
+       struct sctp_association *asoc = transport->asoc;
+       u32 rwnd = asoc->peer.rwnd;
+
+       /* Keep track of how many bytes are in flight over this transport. */
+       transport->flight_size += datasize;
+
+       /* Keep track of how many bytes are in flight to the receiver. */
+       asoc->outqueue.outstanding_bytes += datasize;
+
+       /* Update our view of the receiver's rwnd. */
+       if (datasize < rwnd)
+               rwnd -= datasize;
+       else
+               rwnd = 0;
+
+       asoc->peer.rwnd = rwnd;
+       /* Has been accepted for transmission. */
+       if (!asoc->peer.prsctp_capable)
+               chunk->msg->can_abandon = 0;
+       sctp_chunk_assign_tsn(chunk);
+       sctp_chunk_assign_ssn(chunk);
+}
+
+static sctp_xmit_t sctp_packet_will_fit(struct sctp_packet *packet,
+                                       struct sctp_chunk *chunk,
+                                       u16 chunk_len)
+{
+       size_t psize;
+       size_t pmtu;
+       int too_big;
+       sctp_xmit_t retval = SCTP_XMIT_OK;
+
+       psize = packet->size;
+       pmtu  = ((packet->transport->asoc) ?
+               (packet->transport->asoc->pathmtu) :
+               (packet->transport->pathmtu));
+
+       too_big = (psize + chunk_len > pmtu);
+
+       /* Decide if we need to fragment or resubmit later. */
+       if (too_big) {
+               /* It's OK to fragmet at IP level if any one of the following
+                * is true:
+                *      1. The packet is empty (meaning this chunk is greater
+                *         the MTU)
+                *      2. The chunk we are adding is a control chunk
+                *      3. The packet doesn't have any data in it yet and data
+                *      requires authentication.
+                */
+               if (sctp_packet_empty(packet) || !sctp_chunk_is_data(chunk) ||
+                   (!packet->has_data && chunk->auth)) {
+                       /* We no longer do re-fragmentation.
+                        * Just fragment at the IP layer, if we
+                        * actually hit this condition
+                        */
+                       packet->ipfragok = 1;
+               } else {
+                       retval = SCTP_XMIT_PMTU_FULL;
+               }
+       }
+
+       return retval;
+}