aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--include/net/sctp/structs.h14
-rw-r--r--net/sctp/associola.c82
-rw-r--r--net/sctp/sm_statefuns.c2
-rw-r--r--net/sctp/socket.c6
-rw-r--r--net/sctp/ulpevent.c8
5 files changed, 25 insertions, 87 deletions
diff --git a/include/net/sctp/structs.h b/include/net/sctp/structs.h
index d992ca3145fe..6ee76c804893 100644
--- a/include/net/sctp/structs.h
+++ b/include/net/sctp/structs.h
@@ -1653,17 +1653,6 @@ struct sctp_association {
1653 /* This is the last advertised value of rwnd over a SACK chunk. */ 1653 /* This is the last advertised value of rwnd over a SACK chunk. */
1654 __u32 a_rwnd; 1654 __u32 a_rwnd;
1655 1655
1656 /* Number of bytes by which the rwnd has slopped. The rwnd is allowed
1657 * to slop over a maximum of the association's frag_point.
1658 */
1659 __u32 rwnd_over;
1660
1661 /* Keeps treack of rwnd pressure. This happens when we have
1662 * a window, but not recevie buffer (i.e small packets). This one
1663 * is releases slowly (1 PMTU at a time ).
1664 */
1665 __u32 rwnd_press;
1666
1667 /* This is the sndbuf size in use for the association. 1656 /* This is the sndbuf size in use for the association.
1668 * This corresponds to the sndbuf size for the association, 1657 * This corresponds to the sndbuf size for the association,
1669 * as specified in the sk->sndbuf. 1658 * as specified in the sk->sndbuf.
@@ -1892,8 +1881,7 @@ void sctp_assoc_update(struct sctp_association *old,
1892__u32 sctp_association_get_next_tsn(struct sctp_association *); 1881__u32 sctp_association_get_next_tsn(struct sctp_association *);
1893 1882
1894void sctp_assoc_sync_pmtu(struct sock *, struct sctp_association *); 1883void sctp_assoc_sync_pmtu(struct sock *, struct sctp_association *);
1895void sctp_assoc_rwnd_increase(struct sctp_association *, unsigned int); 1884void sctp_assoc_rwnd_update(struct sctp_association *, bool);
1896void sctp_assoc_rwnd_decrease(struct sctp_association *, unsigned int);
1897void sctp_assoc_set_primary(struct sctp_association *, 1885void sctp_assoc_set_primary(struct sctp_association *,
1898 struct sctp_transport *); 1886 struct sctp_transport *);
1899void sctp_assoc_del_nonprimary_peers(struct sctp_association *, 1887void sctp_assoc_del_nonprimary_peers(struct sctp_association *,
diff --git a/net/sctp/associola.c b/net/sctp/associola.c
index 5ae609200674..f558433537b8 100644
--- a/net/sctp/associola.c
+++ b/net/sctp/associola.c
@@ -1367,44 +1367,35 @@ static inline bool sctp_peer_needs_update(struct sctp_association *asoc)
1367 return false; 1367 return false;
1368} 1368}
1369 1369
1370/* Increase asoc's rwnd by len and send any window update SACK if needed. */ 1370/* Update asoc's rwnd for the approximated state in the buffer,
1371void sctp_assoc_rwnd_increase(struct sctp_association *asoc, unsigned int len) 1371 * and check whether SACK needs to be sent.
1372 */
1373void sctp_assoc_rwnd_update(struct sctp_association *asoc, bool update_peer)
1372{ 1374{
1375 int rx_count;
1373 struct sctp_chunk *sack; 1376 struct sctp_chunk *sack;
1374 struct timer_list *timer; 1377 struct timer_list *timer;
1375 1378
1376 if (asoc->rwnd_over) { 1379 if (asoc->ep->rcvbuf_policy)
1377 if (asoc->rwnd_over >= len) { 1380 rx_count = atomic_read(&asoc->rmem_alloc);
1378 asoc->rwnd_over -= len; 1381 else
1379 } else { 1382 rx_count = atomic_read(&asoc->base.sk->sk_rmem_alloc);
1380 asoc->rwnd += (len - asoc->rwnd_over);
1381 asoc->rwnd_over = 0;
1382 }
1383 } else {
1384 asoc->rwnd += len;
1385 }
1386 1383
1387 /* If we had window pressure, start recovering it 1384 if ((asoc->base.sk->sk_rcvbuf - rx_count) > 0)
1388 * once our rwnd had reached the accumulated pressure 1385 asoc->rwnd = (asoc->base.sk->sk_rcvbuf - rx_count) >> 1;
1389 * threshold. The idea is to recover slowly, but up 1386 else
1390 * to the initial advertised window. 1387 asoc->rwnd = 0;
1391 */
1392 if (asoc->rwnd_press && asoc->rwnd >= asoc->rwnd_press) {
1393 int change = min(asoc->pathmtu, asoc->rwnd_press);
1394 asoc->rwnd += change;
1395 asoc->rwnd_press -= change;
1396 }
1397 1388
1398 pr_debug("%s: asoc:%p rwnd increased by %d to (%u, %u) - %u\n", 1389 pr_debug("%s: asoc:%p rwnd=%u, rx_count=%d, sk_rcvbuf=%d\n",
1399 __func__, asoc, len, asoc->rwnd, asoc->rwnd_over, 1390 __func__, asoc, asoc->rwnd, rx_count,
1400 asoc->a_rwnd); 1391 asoc->base.sk->sk_rcvbuf);
1401 1392
1402 /* Send a window update SACK if the rwnd has increased by at least the 1393 /* Send a window update SACK if the rwnd has increased by at least the
1403 * minimum of the association's PMTU and half of the receive buffer. 1394 * minimum of the association's PMTU and half of the receive buffer.
1404 * The algorithm used is similar to the one described in 1395 * The algorithm used is similar to the one described in
1405 * Section 4.2.3.3 of RFC 1122. 1396 * Section 4.2.3.3 of RFC 1122.
1406 */ 1397 */
1407 if (sctp_peer_needs_update(asoc)) { 1398 if (update_peer && sctp_peer_needs_update(asoc)) {
1408 asoc->a_rwnd = asoc->rwnd; 1399 asoc->a_rwnd = asoc->rwnd;
1409 1400
1410 pr_debug("%s: sending window update SACK- asoc:%p rwnd:%u " 1401 pr_debug("%s: sending window update SACK- asoc:%p rwnd:%u "
@@ -1426,45 +1417,6 @@ void sctp_assoc_rwnd_increase(struct sctp_association *asoc, unsigned int len)
1426 } 1417 }
1427} 1418}
1428 1419
1429/* Decrease asoc's rwnd by len. */
1430void sctp_assoc_rwnd_decrease(struct sctp_association *asoc, unsigned int len)
1431{
1432 int rx_count;
1433 int over = 0;
1434
1435 if (unlikely(!asoc->rwnd || asoc->rwnd_over))
1436 pr_debug("%s: association:%p has asoc->rwnd:%u, "
1437 "asoc->rwnd_over:%u!\n", __func__, asoc,
1438 asoc->rwnd, asoc->rwnd_over);
1439
1440 if (asoc->ep->rcvbuf_policy)
1441 rx_count = atomic_read(&asoc->rmem_alloc);
1442 else
1443 rx_count = atomic_read(&asoc->base.sk->sk_rmem_alloc);
1444
1445 /* If we've reached or overflowed our receive buffer, announce
1446 * a 0 rwnd if rwnd would still be positive. Store the
1447 * the potential pressure overflow so that the window can be restored
1448 * back to original value.
1449 */
1450 if (rx_count >= asoc->base.sk->sk_rcvbuf)
1451 over = 1;
1452
1453 if (asoc->rwnd >= len) {
1454 asoc->rwnd -= len;
1455 if (over) {
1456 asoc->rwnd_press += asoc->rwnd;
1457 asoc->rwnd = 0;
1458 }
1459 } else {
1460 asoc->rwnd_over = len - asoc->rwnd;
1461 asoc->rwnd = 0;
1462 }
1463
1464 pr_debug("%s: asoc:%p rwnd decreased by %d to (%u, %u, %u)\n",
1465 __func__, asoc, len, asoc->rwnd, asoc->rwnd_over,
1466 asoc->rwnd_press);
1467}
1468 1420
1469/* Build the bind address list for the association based on info from the 1421/* Build the bind address list for the association based on info from the
1470 * local endpoint and the remote peer. 1422 * local endpoint and the remote peer.
diff --git a/net/sctp/sm_statefuns.c b/net/sctp/sm_statefuns.c
index 483dcd71b3c5..591b44d3b7de 100644
--- a/net/sctp/sm_statefuns.c
+++ b/net/sctp/sm_statefuns.c
@@ -6176,7 +6176,7 @@ static int sctp_eat_data(const struct sctp_association *asoc,
6176 * PMTU. In cases, such as loopback, this might be a rather 6176 * PMTU. In cases, such as loopback, this might be a rather
6177 * large spill over. 6177 * large spill over.
6178 */ 6178 */
6179 if ((!chunk->data_accepted) && (!asoc->rwnd || asoc->rwnd_over || 6179 if ((!chunk->data_accepted) && (!asoc->rwnd ||
6180 (datalen > asoc->rwnd + asoc->frag_point))) { 6180 (datalen > asoc->rwnd + asoc->frag_point))) {
6181 6181
6182 /* If this is the next TSN, consider reneging to make 6182 /* If this is the next TSN, consider reneging to make
diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index 9e91d6e5df63..7075ac847fde 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -2092,12 +2092,6 @@ static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2092 sctp_skb_pull(skb, copied); 2092 sctp_skb_pull(skb, copied);
2093 skb_queue_head(&sk->sk_receive_queue, skb); 2093 skb_queue_head(&sk->sk_receive_queue, skb);
2094 2094
2095 /* When only partial message is copied to the user, increase
2096 * rwnd by that amount. If all the data in the skb is read,
2097 * rwnd is updated when the event is freed.
2098 */
2099 if (!sctp_ulpevent_is_notification(event))
2100 sctp_assoc_rwnd_increase(event->asoc, copied);
2101 goto out; 2095 goto out;
2102 } else if ((event->msg_flags & MSG_NOTIFICATION) || 2096 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2103 (event->msg_flags & MSG_EOR)) 2097 (event->msg_flags & MSG_EOR))
diff --git a/net/sctp/ulpevent.c b/net/sctp/ulpevent.c
index 85c64658bd0b..8d198ae03606 100644
--- a/net/sctp/ulpevent.c
+++ b/net/sctp/ulpevent.c
@@ -989,7 +989,7 @@ static void sctp_ulpevent_receive_data(struct sctp_ulpevent *event,
989 skb = sctp_event2skb(event); 989 skb = sctp_event2skb(event);
990 /* Set the owner and charge rwnd for bytes received. */ 990 /* Set the owner and charge rwnd for bytes received. */
991 sctp_ulpevent_set_owner(event, asoc); 991 sctp_ulpevent_set_owner(event, asoc);
992 sctp_assoc_rwnd_decrease(asoc, skb_headlen(skb)); 992 sctp_assoc_rwnd_update(asoc, false);
993 993
994 if (!skb->data_len) 994 if (!skb->data_len)
995 return; 995 return;
@@ -1011,6 +1011,7 @@ static void sctp_ulpevent_release_data(struct sctp_ulpevent *event)
1011{ 1011{
1012 struct sk_buff *skb, *frag; 1012 struct sk_buff *skb, *frag;
1013 unsigned int len; 1013 unsigned int len;
1014 struct sctp_association *asoc;
1014 1015
1015 /* Current stack structures assume that the rcv buffer is 1016 /* Current stack structures assume that the rcv buffer is
1016 * per socket. For UDP style sockets this is not true as 1017 * per socket. For UDP style sockets this is not true as
@@ -1035,8 +1036,11 @@ static void sctp_ulpevent_release_data(struct sctp_ulpevent *event)
1035 } 1036 }
1036 1037
1037done: 1038done:
1038 sctp_assoc_rwnd_increase(event->asoc, len); 1039 asoc = event->asoc;
1040 sctp_association_hold(asoc);
1039 sctp_ulpevent_release_owner(event); 1041 sctp_ulpevent_release_owner(event);
1042 sctp_assoc_rwnd_update(asoc, true);
1043 sctp_association_put(asoc);
1040} 1044}
1041 1045
1042static void sctp_ulpevent_release_frag_data(struct sctp_ulpevent *event) 1046static void sctp_ulpevent_release_frag_data(struct sctp_ulpevent *event)