2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
37 #include "opt_tcpdebug.h"
39 #include <sys/param.h>
40 #include <sys/kernel.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
44 #include <sys/socket.h>
45 #include <sys/socketvar.h>
46 #include <sys/sysctl.h>
47 #include <sys/syslog.h>
48 #include <sys/systm.h>
53 #include <net/if_var.h>
54 #include <net/route.h>
57 #include <netinet/in.h>
58 #include <netinet/in_pcb.h>
59 #include <netinet/in_systm.h>
60 #include <netinet/in_var.h>
61 #include <netinet/ip.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_options.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/in6_pcb.h>
66 #include <netinet6/ip6_var.h>
67 #include <netinet6/nd6.h>
68 #include <netinet/tcp.h>
69 #include <netinet/tcp_fsm.h>
70 #include <netinet/tcp_seq.h>
71 #include <netinet/tcp_timer.h>
72 #include <netinet/tcp_var.h>
73 #include <netinet6/tcp6_var.h>
74 #include <netinet/tcpip.h>
76 #include <netinet/tcp_debug.h>
79 static SYSCTL_NODE(_net_inet_tcp
, OID_AUTO
, reass
, CTLFLAG_RW
, 0,
80 "TCP Segment Reassembly Queue");
82 static int tcp_reass_maxseg
= 0;
83 SYSCTL_INT(_net_inet_tcp_reass
, OID_AUTO
, maxsegments
, CTLFLAG_RDTUN
,
85 "Global maximum number of TCP Segments in Reassembly Queue");
87 static uma_zone_t tcp_reass_zone
;
88 SYSCTL_UMA_CUR(_net_inet_tcp_reass
, OID_AUTO
, cursegments
, 0,
90 "Global number of TCP Segments currently in Reassembly Queue");
92 /* Initialize TCP reassembly queue */
94 tcp_reass_zone_change(void *tag
)
97 /* Set the zone limit and read back the effective value. */
98 tcp_reass_maxseg
= nmbclusters
/ 16;
99 tcp_reass_maxseg
= uma_zone_set_max(tcp_reass_zone
,
104 tcp_reass_global_init(void)
107 tcp_reass_maxseg
= nmbclusters
/ 16;
108 TUNABLE_INT_FETCH("net.inet.tcp.reass.maxsegments",
110 tcp_reass_zone
= uma_zcreate("tcpreass", sizeof (struct tseg_qent
),
111 NULL
, NULL
, NULL
, NULL
, UMA_ALIGN_PTR
, UMA_ZONE_NOFREE
);
112 /* Set the zone limit and read back the effective value. */
113 tcp_reass_maxseg
= uma_zone_set_max(tcp_reass_zone
,
115 EVENTHANDLER_REGISTER(nmbclusters_change
,
116 tcp_reass_zone_change
, NULL
, EVENTHANDLER_PRI_ANY
);
120 tcp_reass_flush(struct tcpcb
*tp
)
122 struct tseg_qent
*qe
;
124 INP_WLOCK_ASSERT(tp
->t_inpcb
);
126 while ((qe
= LIST_FIRST(&tp
->t_segq
)) != NULL
) {
127 LIST_REMOVE(qe
, tqe_q
);
129 uma_zfree(tcp_reass_zone
, qe
);
133 KASSERT((tp
->t_segqlen
== 0),
134 ("TCP reass queue %p segment count is %d instead of 0 after flush.",
139 tcp_reass(struct tcpcb
*tp
, struct tcphdr
*th
, int *tlenp
, struct mbuf
*m
)
142 struct tseg_qent
*p
= NULL
;
143 struct tseg_qent
*nq
;
144 struct tseg_qent
*te
= NULL
;
145 struct socket
*so
= tp
->t_inpcb
->inp_socket
;
148 struct tseg_qent tqs
;
150 INP_WLOCK_ASSERT(tp
->t_inpcb
);
153 * XXX: tcp_reass() is rather inefficient with its data structures
154 * and should be rewritten (see NetBSD for optimizations).
158 * Call with th==NULL after become established to
159 * force pre-ESTABLISHED data up to user socket.
165 * Limit the number of segments that can be queued to reduce the
166 * potential for mbuf exhaustion. For best performance, we want to be
167 * able to queue a full window's worth of segments. The size of the
168 * socket receive buffer determines our advertised window and grows
169 * automatically when socket buffer autotuning is enabled. Use it as the
170 * basis for our queue limit.
171 * Always let the missing segment through which caused this queue.
172 * NB: Access to the socket buffer is left intentionally unlocked as we
173 * can tolerate stale information here.
175 * XXXLAS: Using sbspace(so->so_rcv) instead of so->so_rcv.sb_hiwat
176 * should work but causes packets to be dropped when they shouldn't.
177 * Investigate why and re-evaluate the below limit after the behaviour
180 if ((th
->th_seq
!= tp
->rcv_nxt
|| !TCPS_HAVEESTABLISHED(tp
->t_state
)) &&
181 tp
->t_segqlen
>= (so
->so_rcv
.sb_hiwat
/ tp
->t_maxseg
) + 1) {
182 TCPSTAT_INC(tcps_rcvreassfull
);
184 if ((s
= tcp_log_addrs(&tp
->t_inpcb
->inp_inc
, th
, NULL
, NULL
))) {
185 log(LOG_DEBUG
, "%s; %s: queue limit reached, "
186 "segment dropped\n", s
, __func__
);
194 * Allocate a new queue entry. If we can't, or hit the zone limit
197 * Use a temporary structure on the stack for the missing segment
198 * when the zone is exhausted. Otherwise we may get stuck.
200 te
= uma_zalloc(tcp_reass_zone
, M_NOWAIT
);
202 if (th
->th_seq
!= tp
->rcv_nxt
|| !TCPS_HAVEESTABLISHED(tp
->t_state
)) {
203 TCPSTAT_INC(tcps_rcvmemdrop
);
206 if ((s
= tcp_log_addrs(&tp
->t_inpcb
->inp_inc
, th
, NULL
,
208 log(LOG_DEBUG
, "%s; %s: global zone limit "
209 "reached, segment dropped\n", s
, __func__
);
214 bzero(&tqs
, sizeof(struct tseg_qent
));
216 if ((s
= tcp_log_addrs(&tp
->t_inpcb
->inp_inc
, th
, NULL
,
219 "%s; %s: global zone limit reached, using "
220 "stack for missing segment\n", s
, __func__
);
228 * Find a segment which begins after this one does.
230 LIST_FOREACH(q
, &tp
->t_segq
, tqe_q
) {
231 if (SEQ_GT(q
->tqe_th
->th_seq
, th
->th_seq
))
237 * If there is a preceding segment, it may provide some of
238 * our data already. If so, drop the data from the incoming
239 * segment. If it provides all of our data, drop us.
243 /* conversion to int (in i) handles seq wraparound */
244 i
= p
->tqe_th
->th_seq
+ p
->tqe_len
- th
->th_seq
;
247 TCPSTAT_INC(tcps_rcvduppack
);
248 TCPSTAT_ADD(tcps_rcvdupbyte
, *tlenp
);
251 uma_zfree(tcp_reass_zone
, te
);
254 * Try to present any queued data
255 * at the left window edge to the user.
256 * This is needed after the 3-WHS
259 goto present
; /* ??? */
267 TCPSTAT_INC(tcps_rcvoopack
);
268 TCPSTAT_ADD(tcps_rcvoobyte
, *tlenp
);
271 * While we overlap succeeding segments trim them or,
272 * if they are completely covered, dequeue them.
275 int i
= (th
->th_seq
+ *tlenp
) - q
->tqe_th
->th_seq
;
278 if (i
< q
->tqe_len
) {
279 q
->tqe_th
->th_seq
+= i
;
285 nq
= LIST_NEXT(q
, tqe_q
);
286 LIST_REMOVE(q
, tqe_q
);
288 uma_zfree(tcp_reass_zone
, q
);
293 /* Insert the new segment queue entry into place. */
296 te
->tqe_len
= *tlenp
;
299 LIST_INSERT_HEAD(&tp
->t_segq
, te
, tqe_q
);
301 KASSERT(te
!= &tqs
, ("%s: temporary stack based entry not "
302 "first element in queue", __func__
));
303 LIST_INSERT_AFTER(p
, te
, tqe_q
);
308 * Present data to user, advancing rcv_nxt through
309 * completed sequence space.
311 if (!TCPS_HAVEESTABLISHED(tp
->t_state
))
313 q
= LIST_FIRST(&tp
->t_segq
);
314 if (!q
|| q
->tqe_th
->th_seq
!= tp
->rcv_nxt
)
316 SOCKBUF_LOCK(&so
->so_rcv
);
318 tp
->rcv_nxt
+= q
->tqe_len
;
319 flags
= q
->tqe_th
->th_flags
& TH_FIN
;
320 nq
= LIST_NEXT(q
, tqe_q
);
321 LIST_REMOVE(q
, tqe_q
);
322 if (so
->so_rcv
.sb_state
& SBS_CANTRCVMORE
)
325 sbappendstream_locked(&so
->so_rcv
, q
->tqe_m
, 0);
327 uma_zfree(tcp_reass_zone
, q
);
330 } while (q
&& q
->tqe_th
->th_seq
== tp
->rcv_nxt
);
331 sorwakeup_locked(so
);