Merge branch 'sii-m15w' into upstream
[pandora-kernel.git] / net / sctp / sm_statefuns.c
index 2b9a832..ead3f1b 100644 (file)
@@ -93,7 +93,7 @@ static sctp_disposition_t sctp_sf_shut_8_4_5(const struct sctp_endpoint *ep,
 static struct sctp_sackhdr *sctp_sm_pull_sack(struct sctp_chunk *chunk);
 
 static sctp_disposition_t sctp_stop_t1_and_abort(sctp_cmd_seq_t *commands,
-                                          __u16 error,
+                                          __u16 error, int sk_err,
                                           const struct sctp_association *asoc,
                                           struct sctp_transport *transport);
 
@@ -448,7 +448,7 @@ sctp_disposition_t sctp_sf_do_5_1C_ack(const struct sctp_endpoint *ep,
        __u32 init_tag;
        struct sctp_chunk *err_chunk;
        struct sctp_packet *packet;
-       sctp_disposition_t ret;
+       __u16 error;
 
        if (!sctp_vtag_verify(chunk, asoc))
                return sctp_sf_pdiscard(ep, asoc, type, arg, commands);
@@ -480,11 +480,9 @@ sctp_disposition_t sctp_sf_do_5_1C_ack(const struct sctp_endpoint *ep,
                        goto nomem;
 
                sctp_add_cmd_sf(commands, SCTP_CMD_REPLY, SCTP_CHUNK(reply));
-               sctp_add_cmd_sf(commands, SCTP_CMD_NEW_STATE,
-                               SCTP_STATE(SCTP_STATE_CLOSED));
-               SCTP_INC_STATS(SCTP_MIB_ABORTEDS);
-               sctp_add_cmd_sf(commands, SCTP_CMD_DELETE_TCB, SCTP_NULL());
-               return SCTP_DISPOSITION_DELETE_TCB;
+               return sctp_stop_t1_and_abort(commands, SCTP_ERROR_INV_PARAM,
+                                             ECONNREFUSED, asoc,
+                                             chunk->transport);
        }
 
        /* Verify the INIT chunk before processing it. */
@@ -511,27 +509,16 @@ sctp_disposition_t sctp_sf_do_5_1C_ack(const struct sctp_endpoint *ep,
                                sctp_add_cmd_sf(commands, SCTP_CMD_SEND_PKT,
                                                SCTP_PACKET(packet));
                                SCTP_INC_STATS(SCTP_MIB_OUTCTRLCHUNKS);
-                               sctp_add_cmd_sf(commands, SCTP_CMD_NEW_STATE,
-                                               SCTP_STATE(SCTP_STATE_CLOSED));
-                               sctp_add_cmd_sf(commands, SCTP_CMD_DELETE_TCB,
-                                               SCTP_NULL());
-                               return SCTP_DISPOSITION_CONSUME;
+                               error = SCTP_ERROR_INV_PARAM;
                        } else {
-                               sctp_add_cmd_sf(commands, SCTP_CMD_NEW_STATE,
-                                               SCTP_STATE(SCTP_STATE_CLOSED));
-                               sctp_add_cmd_sf(commands, SCTP_CMD_DELETE_TCB,
-                                               SCTP_NULL());
-                               return SCTP_DISPOSITION_NOMEM;
+                               error = SCTP_ERROR_NO_RESOURCE;
                        }
                } else {
-                       ret = sctp_sf_tabort_8_4_8(ep, asoc, type, arg,
-                                                  commands);
-                       sctp_add_cmd_sf(commands, SCTP_CMD_NEW_STATE,
-                                       SCTP_STATE(SCTP_STATE_CLOSED));
-                       sctp_add_cmd_sf(commands, SCTP_CMD_DELETE_TCB,
-                                       SCTP_NULL());
-                       return ret;
+                       sctp_sf_tabort_8_4_8(ep, asoc, type, arg, commands);
+                       error = SCTP_ERROR_INV_PARAM;
                }
+               return sctp_stop_t1_and_abort(commands, error, ECONNREFUSED,
+                                               asoc, chunk->transport);
        }
 
        /* Tag the variable length parameters.  Note that we never
@@ -636,8 +623,9 @@ sctp_disposition_t sctp_sf_do_5_1D_ce(const struct sctp_endpoint *ep,
         */
         chunk->subh.cookie_hdr =
                (struct sctp_signed_cookie *)chunk->skb->data;
-       skb_pull(chunk->skb,
-                ntohs(chunk->chunk_hdr->length) - sizeof(sctp_chunkhdr_t));
+       if (!pskb_pull(chunk->skb, ntohs(chunk->chunk_hdr->length) -
+                                        sizeof(sctp_chunkhdr_t)))
+               goto nomem;
 
        /* 5.1 D) Upon reception of the COOKIE ECHO chunk, Endpoint
         * "Z" will reply with a COOKIE ACK chunk after building a TCB
@@ -858,6 +846,7 @@ static sctp_disposition_t sctp_sf_heartbeat(const struct sctp_endpoint *ep,
        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;
 
        /* Send a heartbeat to our peer.  */
        paylen = sizeof(sctp_sender_hb_info_t);
@@ -885,6 +874,8 @@ sctp_disposition_t sctp_sf_sendbeat_8_3(const struct sctp_endpoint *ep,
        struct sctp_transport *transport = (struct sctp_transport *) arg;
 
        if (asoc->overall_error_count >= asoc->max_retrans) {
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ETIMEDOUT));
                /* CMD_ASSOC_FAILED calls CMD_DELETE_TCB. */
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_NO_ERROR));
@@ -965,7 +956,8 @@ sctp_disposition_t sctp_sf_beat_8_3(const struct sctp_endpoint *ep,
         */
        chunk->subh.hb_hdr = (sctp_heartbeathdr_t *) chunk->skb->data;
        paylen = ntohs(chunk->chunk_hdr->length) - sizeof(sctp_chunkhdr_t);
-       skb_pull(chunk->skb, paylen);
+       if (!pskb_pull(chunk->skb, paylen))
+               goto nomem;
 
        reply = sctp_make_heartbeat_ack(asoc, chunk,
                                        chunk->subh.hb_hdr, paylen);
@@ -1028,6 +1020,12 @@ sctp_disposition_t sctp_sf_backbeat_8_3(const struct sctp_endpoint *ep,
                                                  commands);
 
        hbinfo = (sctp_sender_hb_info_t *) chunk->skb->data;
+       /* Make sure that the length of the parameter is what we expect */
+       if (ntohs(hbinfo->param_hdr.length) !=
+                                   sizeof(sctp_sender_hb_info_t)) {
+               return SCTP_DISPOSITION_DISCARD;
+       }
+
        from_addr = hbinfo->daddr;
        link = sctp_assoc_lookup_paddr(asoc, &from_addr);
 
@@ -1051,6 +1049,10 @@ sctp_disposition_t sctp_sf_backbeat_8_3(const struct sctp_endpoint *ep,
                return SCTP_DISPOSITION_DISCARD;
        }
 
+       /* Validate the 64-bit random nonce. */
+       if (hbinfo->hb_nonce != link->hb_nonce)
+               return SCTP_DISPOSITION_DISCARD;
+
        max_interval = link->hbinterval + link->rto;
 
        /* Check if the timestamp looks valid.  */
@@ -1860,8 +1862,9 @@ sctp_disposition_t sctp_sf_do_5_2_4_dupcook(const struct sctp_endpoint *ep,
         * are in good shape.
         */
         chunk->subh.cookie_hdr = (struct sctp_signed_cookie *)chunk->skb->data;
-       skb_pull(chunk->skb, ntohs(chunk->chunk_hdr->length) -
-                sizeof(sctp_chunkhdr_t));
+       if (!pskb_pull(chunk->skb, ntohs(chunk->chunk_hdr->length) -
+                                       sizeof(sctp_chunkhdr_t)))
+               goto nomem;
 
        /* In RFC 2960 5.2.4 3, if both Verification Tags in the State Cookie
         * of a duplicate COOKIE ECHO match the Verification Tags of the
@@ -2123,6 +2126,8 @@ static sctp_disposition_t sctp_sf_do_5_2_6_stale(const struct sctp_endpoint *ep,
        int attempts = asoc->init_err_counter + 1;
 
        if (attempts > asoc->max_init_attempts) {
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ETIMEDOUT));
                sctp_add_cmd_sf(commands, SCTP_CMD_INIT_FAILED,
                                SCTP_U32(SCTP_ERROR_STALE_COOKIE));
                return SCTP_DISPOSITION_DELETE_TCB;
@@ -2259,6 +2264,7 @@ sctp_disposition_t sctp_sf_do_9_1_abort(const struct sctp_endpoint *ep,
        if (len >= sizeof(struct sctp_chunkhdr) + sizeof(struct sctp_errhdr))
                error = ((sctp_errhdr_t *)chunk->skb->data)->cause;
 
+       sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR, SCTP_ERROR(ECONNRESET));
        /* ASSOC_FAILED will DELETE_TCB. */
        sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED, SCTP_U32(error));
        SCTP_INC_STATS(SCTP_MIB_ABORTEDS);
@@ -2303,7 +2309,8 @@ sctp_disposition_t sctp_sf_cookie_wait_abort(const struct sctp_endpoint *ep,
        if (len >= sizeof(struct sctp_chunkhdr) + sizeof(struct sctp_errhdr))
                error = ((sctp_errhdr_t *)chunk->skb->data)->cause;
 
-       return sctp_stop_t1_and_abort(commands, error, asoc, chunk->transport);
+       return sctp_stop_t1_and_abort(commands, error, ECONNREFUSED, asoc,
+                                     chunk->transport);
 }
 
 /*
@@ -2315,7 +2322,8 @@ sctp_disposition_t sctp_sf_cookie_wait_icmp_abort(const struct sctp_endpoint *ep
                                        void *arg,
                                        sctp_cmd_seq_t *commands)
 {
-       return sctp_stop_t1_and_abort(commands, SCTP_ERROR_NO_ERROR, asoc,
+       return sctp_stop_t1_and_abort(commands, SCTP_ERROR_NO_ERROR,
+                                     ENOPROTOOPT, asoc,
                                      (struct sctp_transport *)arg);
 }
 
@@ -2340,7 +2348,7 @@ sctp_disposition_t sctp_sf_cookie_echoed_abort(const struct sctp_endpoint *ep,
  * This is common code called by several sctp_sf_*_abort() functions above.
  */
 static sctp_disposition_t sctp_stop_t1_and_abort(sctp_cmd_seq_t *commands,
-                                          __u16 error,
+                                          __u16 error, int sk_err,
                                           const struct sctp_association *asoc,
                                           struct sctp_transport *transport)
 {
@@ -2350,6 +2358,7 @@ static sctp_disposition_t sctp_stop_t1_and_abort(sctp_cmd_seq_t *commands,
        SCTP_INC_STATS(SCTP_MIB_ABORTEDS);
        sctp_add_cmd_sf(commands, SCTP_CMD_TIMER_STOP,
                        SCTP_TO(SCTP_EVENT_TIMEOUT_T1_INIT));
+       sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR, SCTP_ERROR(sk_err));
        /* CMD_INIT_FAILED will DELETE_TCB. */
        sctp_add_cmd_sf(commands, SCTP_CMD_INIT_FAILED,
                        SCTP_U32(error));
@@ -3333,6 +3342,8 @@ sctp_disposition_t sctp_sf_do_asconf_ack(const struct sctp_endpoint *ep,
                sctp_add_cmd_sf(commands, SCTP_CMD_TIMER_STOP,
                                SCTP_TO(SCTP_EVENT_TIMEOUT_T4_RTO));
                sctp_add_cmd_sf(commands, SCTP_CMD_DISCARD_PACKET,SCTP_NULL());
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ECONNABORTED));
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_ASCONF_ACK));
                SCTP_INC_STATS(SCTP_MIB_ABORTEDS);
@@ -3359,6 +3370,8 @@ sctp_disposition_t sctp_sf_do_asconf_ack(const struct sctp_endpoint *ep,
                 * processing the rest of the chunks in the packet.
                 */
                sctp_add_cmd_sf(commands, SCTP_CMD_DISCARD_PACKET,SCTP_NULL());
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ECONNABORTED));
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_ASCONF_ACK));
                SCTP_INC_STATS(SCTP_MIB_ABORTEDS);
@@ -3711,9 +3724,13 @@ static sctp_disposition_t sctp_sf_violation_chunklen(
        if (asoc->state <= SCTP_STATE_COOKIE_ECHOED) {
                sctp_add_cmd_sf(commands, SCTP_CMD_TIMER_STOP,
                                SCTP_TO(SCTP_EVENT_TIMEOUT_T1_INIT));
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ECONNREFUSED));
                sctp_add_cmd_sf(commands, SCTP_CMD_INIT_FAILED,
                                SCTP_U32(SCTP_ERROR_PROTO_VIOLATION));
        } else {
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ECONNABORTED));
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_PROTO_VIOLATION));
                SCTP_DEC_STATS(SCTP_MIB_CURRESTAB);
@@ -4031,6 +4048,8 @@ sctp_disposition_t sctp_sf_do_9_1_prm_abort(
         * TCB.  This is a departure from our typical NOMEM handling.
         */
 
+       sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                       SCTP_ERROR(ECONNABORTED));
        /* Delete the established association. */
        sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                        SCTP_U32(SCTP_ERROR_USER_ABORT));
@@ -4172,6 +4191,8 @@ sctp_disposition_t sctp_sf_cookie_wait_prm_abort(
         * TCB.  This is a departure from our typical NOMEM handling.
         */
 
+       sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                       SCTP_ERROR(ECONNREFUSED));
        /* Delete the established association. */
        sctp_add_cmd_sf(commands, SCTP_CMD_INIT_FAILED,
                        SCTP_U32(SCTP_ERROR_USER_ABORT));
@@ -4540,6 +4561,8 @@ sctp_disposition_t sctp_sf_do_6_3_3_rtx(const struct sctp_endpoint *ep,
        struct sctp_transport *transport = arg;
 
        if (asoc->overall_error_count >= asoc->max_retrans) {
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ETIMEDOUT));
                /* CMD_ASSOC_FAILED calls CMD_DELETE_TCB. */
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_NO_ERROR));
@@ -4659,6 +4682,8 @@ sctp_disposition_t sctp_sf_t1_init_timer_expire(const struct sctp_endpoint *ep,
                SCTP_DEBUG_PRINTK("Giving up on INIT, attempts: %d"
                                  " max_init_attempts: %d\n",
                                  attempts, asoc->max_init_attempts);
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ETIMEDOUT));
                sctp_add_cmd_sf(commands, SCTP_CMD_INIT_FAILED,
                                SCTP_U32(SCTP_ERROR_NO_ERROR));
                return SCTP_DISPOSITION_DELETE_TCB;
@@ -4708,6 +4733,8 @@ sctp_disposition_t sctp_sf_t1_cookie_timer_expire(const struct sctp_endpoint *ep
 
                sctp_add_cmd_sf(commands, SCTP_CMD_REPLY, SCTP_CHUNK(repl));
        } else {
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ETIMEDOUT));
                sctp_add_cmd_sf(commands, SCTP_CMD_INIT_FAILED,
                                SCTP_U32(SCTP_ERROR_NO_ERROR));
                return SCTP_DISPOSITION_DELETE_TCB;
@@ -4739,6 +4766,8 @@ sctp_disposition_t sctp_sf_t2_timer_expire(const struct sctp_endpoint *ep,
 
        SCTP_DEBUG_PRINTK("Timer T2 expired.\n");
        if (asoc->overall_error_count >= asoc->max_retrans) {
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ETIMEDOUT));
                /* Note:  CMD_ASSOC_FAILED calls CMD_DELETE_TCB. */
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_NO_ERROR));
@@ -4814,6 +4843,8 @@ sctp_disposition_t sctp_sf_t4_timer_expire(
        if (asoc->overall_error_count >= asoc->max_retrans) {
                sctp_add_cmd_sf(commands, SCTP_CMD_TIMER_STOP,
                                SCTP_TO(SCTP_EVENT_TIMEOUT_T4_RTO));
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ETIMEDOUT));
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_NO_ERROR));
                SCTP_INC_STATS(SCTP_MIB_ABORTEDS);
@@ -4867,6 +4898,8 @@ sctp_disposition_t sctp_sf_t5_timer_expire(const struct sctp_endpoint *ep,
                goto nomem;
 
        sctp_add_cmd_sf(commands, SCTP_CMD_REPLY, SCTP_CHUNK(reply));
+       sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                       SCTP_ERROR(ETIMEDOUT));
        sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                        SCTP_U32(SCTP_ERROR_NO_ERROR));
 
@@ -5151,7 +5184,9 @@ static int sctp_eat_data(const struct sctp_association *asoc,
        int tmp;
        __u32 tsn;
        int account_value;
+       struct sctp_tsnmap *map = (struct sctp_tsnmap *)&asoc->peer.tsn_map;
        struct sock *sk = asoc->base.sk;
+       int rcvbuf_over = 0;
 
        data_hdr = chunk->subh.data_hdr = (sctp_datahdr_t *)chunk->skb->data;
        skb_pull(chunk->skb, sizeof(sctp_datahdr_t));
@@ -5162,10 +5197,16 @@ static int sctp_eat_data(const struct sctp_association *asoc,
        /* ASSERT:  Now skb->data is really the user data.  */
 
        /*
-        * if we are established, and we have used up our receive
-        * buffer memory, drop the frame
-        */
-       if (asoc->state == SCTP_STATE_ESTABLISHED) {
+        * If we are established, and we have used up our receive buffer
+        * memory, think about droping the frame.
+        * Note that we have an opportunity to improve performance here.
+        * If we accept one chunk from an skbuff, we have to keep all the
+        * memory of that skbuff around until the chunk is read into user
+        * space. Therefore, once we accept 1 chunk we may as well accept all
+        * remaining chunks in the skbuff. The data_accepted flag helps us do
+        * that.
+        */
+       if ((asoc->state == SCTP_STATE_ESTABLISHED) && (!chunk->data_accepted)) {
                /*
                 * If the receive buffer policy is 1, then each
                 * association can allocate up to sk_rcvbuf bytes
@@ -5176,9 +5217,25 @@ static int sctp_eat_data(const struct sctp_association *asoc,
                        account_value = atomic_read(&asoc->rmem_alloc);
                else
                        account_value = atomic_read(&sk->sk_rmem_alloc);
-
-               if (account_value > sk->sk_rcvbuf)
-                       return SCTP_IERROR_IGNORE_TSN;
+               if (account_value > sk->sk_rcvbuf) {
+                       /*
+                        * We need to make forward progress, even when we are
+                        * under memory pressure, so we always allow the
+                        * next tsn after the ctsn ack point to be accepted.
+                        * This lets us avoid deadlocks in which we have to
+                        * drop frames that would otherwise let us drain the
+                        * receive queue.
+                        */
+                       if ((sctp_tsnmap_get_ctsn(map) + 1) != tsn)
+                               return SCTP_IERROR_IGNORE_TSN;
+
+                       /*
+                        * We're going to accept the frame but we should renege
+                        * to make space for it. This will send us down that
+                        * path later in this function.
+                        */
+                       rcvbuf_over = 1;
+               }
        }
 
        /* Process ECN based congestion.
@@ -5240,9 +5297,18 @@ static int sctp_eat_data(const struct sctp_association *asoc,
         * seems a bit troublesome in that frag_point varies based on
         * PMTU.  In cases, such as loopback, this might be a rather
         * large spill over.
+        * NOTE: If we have a full receive buffer here, we only renege if
+        * our receiver can still make progress without the tsn being
+        * received. We do this because in the event that the associations
+        * receive queue is empty we are filling a leading gap, and since
+        * reneging moves the gap to the end of the tsn stream, we are likely
+        * to stall again very shortly. Avoiding the renege when we fill a
+        * leading gap is a good heuristic for avoiding such steady state
+        * stalls.
         */
        if (!asoc->rwnd || asoc->rwnd_over ||
-           (datalen > asoc->rwnd + asoc->frag_point)) {
+           (datalen > asoc->rwnd + asoc->frag_point) ||
+           (rcvbuf_over && (!skb_queue_len(&sk->sk_receive_queue)))) {
 
                /* If this is the next TSN, consider reneging to make
                 * room.   Note: Playing nice with a confused sender.  A
@@ -5250,8 +5316,8 @@ static int sctp_eat_data(const struct sctp_association *asoc,
                 * space and in the future we may want to detect and
                 * do more drastic reneging.
                 */
-               if (sctp_tsnmap_has_gap(&asoc->peer.tsn_map) &&
-                   (sctp_tsnmap_get_ctsn(&asoc->peer.tsn_map) + 1) == tsn) {
+               if (sctp_tsnmap_has_gap(map) &&
+                   (sctp_tsnmap_get_ctsn(map) + 1) == tsn) {
                        SCTP_DEBUG_PRINTK("Reneging for tsn:%u\n", tsn);
                        deliver = SCTP_CMD_RENEGE;
                } else {
@@ -5280,6 +5346,8 @@ static int sctp_eat_data(const struct sctp_association *asoc,
                 * processing the rest of the chunks in the packet.
                 */
                sctp_add_cmd_sf(commands, SCTP_CMD_DISCARD_PACKET,SCTP_NULL());
+               sctp_add_cmd_sf(commands, SCTP_CMD_SET_SK_ERR,
+                               SCTP_ERROR(ECONNABORTED));
                sctp_add_cmd_sf(commands, SCTP_CMD_ASSOC_FAILED,
                                SCTP_U32(SCTP_ERROR_NO_DATA));
                SCTP_INC_STATS(SCTP_MIB_ABORTEDS);
@@ -5293,6 +5361,8 @@ static int sctp_eat_data(const struct sctp_association *asoc,
        if (SCTP_CMD_CHUNK_ULP == deliver)
                sctp_add_cmd_sf(commands, SCTP_CMD_REPORT_TSN, SCTP_U32(tsn));
 
+       chunk->data_accepted = 1;
+
        /* Note: Some chunks may get overcounted (if we drop) or overcounted
         * if we renege and the chunk arrives again.
         */