multi-homing and wildcards
[unix-history] / usr / src / sys / netinet / tcp_usrreq.c
CommitLineData
1aa87517 1/* tcp_usrreq.c 1.52 82/02/27 */
72f24d7d 2
4eb5d593 3#include "../h/param.h"
72f24d7d 4#include "../h/systm.h"
dad64fdf
BJ
5#include "../h/mbuf.h"
6#include "../h/socket.h"
eee3ab16
BJ
7#include "../h/socketvar.h"
8#include "../h/protosw.h"
0974b45c
BJ
9#include "../net/in.h"
10#include "../net/in_pcb.h"
11#include "../net/in_systm.h"
4ad99bae 12#include "../net/if.h"
eee3ab16 13#include "../net/ip.h"
eb44bfb2 14#include "../net/ip_var.h"
eee3ab16 15#include "../net/tcp.h"
eee3ab16 16#include "../net/tcp_fsm.h"
0974b45c
BJ
17#include "../net/tcp_seq.h"
18#include "../net/tcp_timer.h"
eee3ab16 19#include "../net/tcp_var.h"
0974b45c 20#include "../net/tcpip.h"
17b82ed4 21#include "../net/tcp_debug.h"
f1b2fa5b 22#include "../errno.h"
eee3ab16 23
290e0b0a
BJ
24/*
25 * TCP protocol interface to socket abstraction.
26 */
27extern char *tcpstates[];
4ad99bae 28struct tcpcb *tcp_newtcpcb();
290e0b0a 29
9c5022e3 30/*
290e0b0a 31 * Process a TCP user request for TCP tb. If this is a send request
9c5022e3
BJ
32 * then m is the mbuf chain of send data. If this is a timer expiration
33 * (called from the software clock routine), then timertype tells which timer.
34 */
eee3ab16
BJ
35tcp_usrreq(so, req, m, addr)
36 struct socket *so;
37 int req;
9c5022e3 38 struct mbuf *m;
eee3ab16 39 caddr_t addr;
4eb5d593 40{
53a5409e 41 register struct inpcb *inp = sotoinpcb(so);
cdad2eb1 42 register struct tcpcb *tp;
72f24d7d 43 int s = splnet();
eee3ab16 44 int error = 0;
17b82ed4 45 int ostate;
72f24d7d
BJ
46COUNT(TCP_USRREQ);
47
53a5409e 48 /*
290e0b0a
BJ
49 * When a TCP is attached to a socket, then there will be
50 * a (struct inpcb) pointed at by the socket, and this
51 * structure will point at a subsidary (struct tcpcb).
52 * The normal sequence of events is:
53 * PRU_ATTACH creating these structures
54 * PRU_CONNECT connecting to a remote peer
55 * (PRU_SEND|PRU_RCVD)* exchanging data
56 * PRU_DISCONNECT disconnecting from remote peer
57 * PRU_DETACH deleting the structures
58 * With the operations from PRU_CONNECT through PRU_DISCONNECT
59 * possible repeated several times.
60 *
61 * MULTIPLE CONNECTS ARE NOT YET IMPLEMENTED.
53a5409e 62 */
0974b45c 63 if (inp == 0 && req != PRU_ATTACH) {
a6503abf 64 splx(s);
290e0b0a 65 return (EINVAL); /* XXX */
a6503abf
BJ
66 }
67 if (inp) {
cdad2eb1 68 tp = intotcpcb(inp);
9c5022e3 69#ifdef KPROF
a6503abf 70 tcp_acounts[tp->t_state][req]++;
9c5022e3 71#endif
17b82ed4 72 ostate = tp->t_state;
cdad2eb1 73 }
eee3ab16 74 switch (req) {
4eb5d593 75
290e0b0a
BJ
76 /*
77 * TCP attaches to socket via PRU_ATTACH, reserving space,
78 * and internet and TCP control blocks.
79 * If the socket is to receive connections,
80 * then the LISTEN state is entered.
81 */
eee3ab16 82 case PRU_ATTACH:
4ad99bae 83 if (inp) {
eee3ab16 84 error = EISCONN;
cdad2eb1 85 break;
53a5409e 86 }
290e0b0a 87 error = tcp_attach(so, (struct sockaddr *)addr);
a6503abf 88 if (error)
4ad99bae 89 break;
8e65fd66
BJ
90 if ((so->so_options & SO_DONTLINGER) == 0)
91 so->so_linger = TCP_LINGERTIME;
290e0b0a 92 tp = sototcpcb(so);
72f24d7d 93 break;
4eb5d593 94
290e0b0a
BJ
95 /*
96 * PRU_DETACH detaches the TCP protocol from the socket.
97 * If the protocol state is non-embryonic, then can't
98 * do this directly: have to initiate a PRU_DISCONNECT,
99 * which may finish later; embryonic TCB's can just
100 * be discarded here.
101 */
eee3ab16 102 case PRU_DETACH:
290e0b0a
BJ
103 if (tp->t_state > TCPS_LISTEN)
104 tcp_disconnect(tp);
105 else {
106 tcp_close(tp);
107 tp = 0;
108 }
eee3ab16
BJ
109 break;
110
290e0b0a
BJ
111 /*
112 * Initiate connection to peer.
113 * Create a template for use in transmissions on this connection.
114 * Enter SYN_SENT state, and mark socket as connecting.
115 * Start keep-alive timer, and seed output sequence space.
116 * Send initial segment on connection.
117 */
eee3ab16 118 case PRU_CONNECT:
405c9168 119 error = in_pcbconnect(inp, (struct sockaddr_in *)addr);
4ad99bae 120 if (error)
53a5409e 121 break;
b454c3ea 122 tp->t_template = tcp_template(tp);
290e0b0a
BJ
123 if (tp->t_template == 0) {
124 in_pcbdisconnect(inp);
125 error = ENOBUFS;
126 break;
127 }
53a5409e 128 soisconnecting(so);
a6503abf 129 tp->t_state = TCPS_SYN_SENT;
4aed14e3
BJ
130 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
131 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
132 tcp_sendseqinit(tp);
f1b2fa5b 133 (void) tcp_output(tp);
72f24d7d 134 break;
4eb5d593 135
290e0b0a
BJ
136 /*
137 * Initiate disconnect from peer.
138 * If connection never passed embryonic stage, just drop;
139 * else if don't need to let data drain, then can just drop anyways,
140 * else have to begin TCP shutdown process: mark socket disconnecting,
141 * drain unread data, state switch to reflect user close, and
142 * send segment (e.g. FIN) to peer. Socket will be really disconnected
143 * when peer sends FIN and acks ours.
144 *
145 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
146 */
147 case PRU_DISCONNECT:
148 tcp_disconnect(tp);
4aed14e3
BJ
149 break;
150
290e0b0a
BJ
151 /*
152 * Accept a connection. Essentially all the work is
153 * done at higher levels; just return the address
154 * of the peer, storing through addr.
155 */
2b4b57cd 156 case PRU_ACCEPT:
eee3ab16
BJ
157 break;
158
290e0b0a
BJ
159 /*
160 * Mark the connection as being incapable of further output.
161 */
eee3ab16 162 case PRU_SHUTDOWN:
0974b45c 163 socantsendmore(so);
4aed14e3
BJ
164 tcp_usrclosed(tp);
165 (void) tcp_output(tp);
72f24d7d
BJ
166 break;
167
290e0b0a
BJ
168 /*
169 * After a receive, possibly send window update to peer.
170 */
eee3ab16 171 case PRU_RCVD:
f1b2fa5b 172 (void) tcp_output(tp);
72f24d7d
BJ
173 break;
174
290e0b0a
BJ
175 /*
176 * Do a send by putting data in output queue and updating urgent
177 * marker if URG set. Possibly send more data.
178 */
eee3ab16 179 case PRU_SEND:
a6503abf 180 sbappend(&so->so_snd, m);
0974b45c
BJ
181/*
182 if (tp->t_flags & TF_PUSH)
a6503abf 183 tp->snd_end = tp->snd_una + so->so_snd.sb_cc;
0974b45c 184 */
f1b2fa5b 185 (void) tcp_output(tp);
72f24d7d
BJ
186 break;
187
290e0b0a
BJ
188 /*
189 * Abort the TCP.
190 */
eee3ab16 191 case PRU_ABORT:
a6503abf 192 tcp_drop(tp, ECONNABORTED);
72f24d7d
BJ
193 break;
194
290e0b0a 195/* SOME AS YET UNIMPLEMENTED HOOKS */
eee3ab16 196 case PRU_CONTROL:
53a5409e 197 error = EOPNOTSUPP;
eee3ab16
BJ
198 break;
199
f1b2fa5b
BJ
200 case PRU_SENSE:
201 error = EOPNOTSUPP;
202 break;
0244dbc7 203/* END UNIMPLEMENTED HOOKS */
f1b2fa5b
BJ
204
205 case PRU_RCVOOB:
8b5a83bb
BJ
206 if (so->so_oobmark == 0 &&
207 (so->so_state & SS_RCVATMARK) == 0) {
0244dbc7
BJ
208 error = EINVAL;
209 break;
210 }
b2db9217 211 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
8b5a83bb 212 error = EWOULDBLOCK;
b2db9217 213 break;
8b5a83bb 214 }
b2db9217 215 *mtod(m, caddr_t) = tp->t_iobc;
f1b2fa5b
BJ
216 break;
217
218 case PRU_SENDOOB:
b2db9217
BJ
219#ifdef TCPTRUEOOB
220 if (tp->t_flags & TF_DOOOB) {
221 tp->t_oobseq++;
222 tp->t_oobc = *mtod(m, caddr_t);
223 tp->t_oobmark = tp->snd_una + so->so_snd.sb_cc;
224printf("sendoob seq now %x oobc %x\n", tp->t_oobseq, tp->t_oobc);
225 tp->t_oobflags |= TCPOOB_NEEDACK;
226 (void) tcp_output(tp);
227 }
228#endif
8b5a83bb
BJ
229 if (sbspace(&so->so_snd) < -512) {
230 error = ENOBUFS;
231 break;
232 }
0244dbc7
BJ
233 tp->snd_up = tp->snd_una + so->so_snd.sb_cc + 1;
234 sbappend(&so->so_snd, m);
235/*
236 if (tp->t_flags & TF_PUSH)
237 tp->snd_end = tp->snd_una + so->so_snd.sb_cc;
238 */
b2db9217
BJ
239 tp->t_force = 1;
240 (void) tcp_output(tp);
241 tp->t_force = 0;
f1b2fa5b
BJ
242 break;
243
290e0b0a
BJ
244 /*
245 * TCP slow timer went off; going through this
246 * routine for tracing's sake.
247 */
eee3ab16 248 case PRU_SLOWTIMO:
a6503abf 249 tcp_timers(tp, (int)addr);
17b82ed4 250 req |= (int)addr << 8; /* for debug's sake */
eee3ab16
BJ
251 break;
252
9c5022e3
BJ
253 default:
254 panic("tcp_usrreq");
72f24d7d 255 }
17b82ed4
BJ
256 if (tp && (so->so_options & SO_DEBUG))
257 tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req);
72f24d7d 258 splx(s);
53a5409e 259 return (error);
4eb5d593 260}
4aed14e3 261
306f91c9
BJ
262int tcp_sendspace = 1024*2;
263int tcp_recvspace = 1024*3;
290e0b0a
BJ
264/*
265 * Attach TCP protocol to socket, allocating
266 * internet protocol control block, tcp control block,
267 * bufer space, and entering LISTEN state if to accept connections.
268 */
269tcp_attach(so, sa)
270 struct socket *so;
271 struct sockaddr *sa;
272{
273 register struct tcpcb *tp;
274 struct inpcb *inp;
275 int error;
276
306f91c9
BJ
277 error = in_pcbattach(so, &tcb,
278 tcp_sendspace, tcp_recvspace, (struct sockaddr_in *)sa);
290e0b0a
BJ
279 if (error)
280 return (error);
281 inp = (struct inpcb *)so->so_pcb;
282 tp = tcp_newtcpcb(inp);
283 if (so->so_options & SO_ACCEPTCONN) {
284 if (tp == 0) {
285 in_pcbdetach(inp);
286 return (ENOBUFS);
287 }
288 tp->t_state = TCPS_LISTEN;
289 } else
290 tp->t_state = TCPS_CLOSED;
291 return (0);
292}
293
294/*
295 * Initiate (or continue) disconnect.
296 * If embryonic state, just send reset (once).
297 * If not in ``let data drain'' option, just drop.
298 * Otherwise (hard), mark socket disconnecting and drop
299 * current input data; switch states based on user close, and
300 * send segment to peer (with FIN).
301 */
302tcp_disconnect(tp)
303 struct tcpcb *tp;
304{
305 struct socket *so = tp->t_inpcb->inp_socket;
306
307 if (tp->t_state < TCPS_ESTABLISHED)
308 tcp_close(tp);
8e65fd66 309 else if (so->so_linger == 0)
290e0b0a
BJ
310 tcp_drop(tp, 0);
311 else {
312 soisdisconnecting(so);
313 sbflush(&so->so_rcv);
314 tcp_usrclosed(tp);
315 (void) tcp_output(tp);
316 }
317}
318
319/*
320 * User issued close, and wish to trail through shutdown states:
321 * if never received SYN, just forget it. If got a SYN from peer,
322 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
323 * If already got a FIN from peer, then almost done; go to LAST_ACK
324 * state. In all other cases, have already sent FIN to peer (e.g.
325 * after PRU_SHUTDOWN), and just have to play tedious game waiting
326 * for peer to send FIN or not respond to keep-alives, etc.
085a0b90 327 * We can let the user exit from the close as soon as the FIN is acked.
290e0b0a 328 */
4aed14e3
BJ
329tcp_usrclosed(tp)
330 struct tcpcb *tp;
331{
332
4aed14e3
BJ
333 switch (tp->t_state) {
334
335 case TCPS_LISTEN:
336 case TCPS_SYN_SENT:
337 tp->t_state = TCPS_CLOSED;
338 tcp_close(tp);
339 break;
340
341 case TCPS_SYN_RECEIVED:
342 case TCPS_ESTABLISHED:
343 tp->t_state = TCPS_FIN_WAIT_1;
344 break;
345
346 case TCPS_CLOSE_WAIT:
347 tp->t_state = TCPS_LAST_ACK;
348 break;
349 }
085a0b90
BJ
350 if (tp->t_state >= TCPS_FIN_WAIT_2)
351 soisdisconnected(tp->t_inpcb->inp_socket);
4aed14e3 352}