ip_freef now void
[unix-history] / usr / src / sys / netinet / tcp_usrreq.c
CommitLineData
0e3936fa 1/* tcp_usrreq.c 1.73 83/01/17 */
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"
6e7edb25
BJ
9
10#include "../net/if.h"
c124e997 11#include "../net/route.h"
6e7edb25 12#include "../netinet/in.h"
839fe741
BJ
13#include "../netinet/in_pcb.h"
14#include "../netinet/in_systm.h"
839fe741
BJ
15#include "../netinet/ip.h"
16#include "../netinet/ip_var.h"
17#include "../netinet/tcp.h"
18#include "../netinet/tcp_fsm.h"
19#include "../netinet/tcp_seq.h"
20#include "../netinet/tcp_timer.h"
21#include "../netinet/tcp_var.h"
22#include "../netinet/tcpip.h"
23#include "../netinet/tcp_debug.h"
8a2f82db 24#include <errno.h>
eee3ab16 25
290e0b0a
BJ
26/*
27 * TCP protocol interface to socket abstraction.
28 */
29extern char *tcpstates[];
4ad99bae 30struct tcpcb *tcp_newtcpcb();
290e0b0a 31
9c5022e3 32/*
290e0b0a 33 * Process a TCP user request for TCP tb. If this is a send request
9c5022e3
BJ
34 * then m is the mbuf chain of send data. If this is a timer expiration
35 * (called from the software clock routine), then timertype tells which timer.
36 */
a8d3bf7f 37/*ARGSUSED*/
bc031e39 38tcp_usrreq(so, req, m, nam)
eee3ab16
BJ
39 struct socket *so;
40 int req;
8075bb0e 41 struct mbuf *m, *nam;
4eb5d593 42{
53a5409e 43 register struct inpcb *inp = sotoinpcb(so);
cdad2eb1 44 register struct tcpcb *tp;
72f24d7d 45 int s = splnet();
eee3ab16 46 int error = 0;
17b82ed4 47 int ostate;
72f24d7d 48
53a5409e 49 /*
290e0b0a
BJ
50 * When a TCP is attached to a socket, then there will be
51 * a (struct inpcb) pointed at by the socket, and this
52 * structure will point at a subsidary (struct tcpcb).
53a5409e 53 */
0974b45c 54 if (inp == 0 && req != PRU_ATTACH) {
a6503abf 55 splx(s);
290e0b0a 56 return (EINVAL); /* XXX */
a6503abf
BJ
57 }
58 if (inp) {
cdad2eb1 59 tp = intotcpcb(inp);
8075bb0e 60 /* WHAT IF TP IS 0? */
9c5022e3 61#ifdef KPROF
a6503abf 62 tcp_acounts[tp->t_state][req]++;
9c5022e3 63#endif
17b82ed4 64 ostate = tp->t_state;
ebf42a75
BJ
65 } else
66 ostate = 0;
eee3ab16 67 switch (req) {
4eb5d593 68
290e0b0a
BJ
69 /*
70 * TCP attaches to socket via PRU_ATTACH, reserving space,
8075bb0e 71 * and an internet control block.
290e0b0a 72 */
eee3ab16 73 case PRU_ATTACH:
4ad99bae 74 if (inp) {
eee3ab16 75 error = EISCONN;
cdad2eb1 76 break;
53a5409e 77 }
a1edc12b 78 error = tcp_attach(so);
a6503abf 79 if (error)
4ad99bae 80 break;
0e3936fa 81 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
8e65fd66 82 so->so_linger = TCP_LINGERTIME;
290e0b0a 83 tp = sototcpcb(so);
72f24d7d 84 break;
4eb5d593 85
290e0b0a
BJ
86 /*
87 * PRU_DETACH detaches the TCP protocol from the socket.
88 * If the protocol state is non-embryonic, then can't
89 * do this directly: have to initiate a PRU_DISCONNECT,
90 * which may finish later; embryonic TCB's can just
91 * be discarded here.
92 */
eee3ab16 93 case PRU_DETACH:
290e0b0a 94 if (tp->t_state > TCPS_LISTEN)
0e3936fa
SL
95 tp = tcp_disconnect(tp);
96 else
97 tp = tcp_close(tp);
eee3ab16
BJ
98 break;
99
8075bb0e
BJ
100 /*
101 * Give the socket an address.
102 */
103 case PRU_BIND:
104 error = in_pcbbind(inp, nam);
105 if (error)
106 break;
107 break;
108
109 /*
110 * Prepare to accept connections.
111 */
112 case PRU_LISTEN:
113 if (inp->inp_lport == 0)
114 error = in_pcbbind(inp, (struct mbuf *)0);
115 if (error == 0)
116 tp->t_state = TCPS_LISTEN;
117 break;
118
290e0b0a
BJ
119 /*
120 * Initiate connection to peer.
121 * Create a template for use in transmissions on this connection.
122 * Enter SYN_SENT state, and mark socket as connecting.
123 * Start keep-alive timer, and seed output sequence space.
124 * Send initial segment on connection.
125 */
eee3ab16 126 case PRU_CONNECT:
8075bb0e
BJ
127 if (inp->inp_lport == 0) {
128 error = in_pcbbind(inp, (struct mbuf *)0);
129 if (error)
130 break;
131 }
132 error = in_pcbconnect(inp, nam);
4ad99bae 133 if (error)
53a5409e 134 break;
b454c3ea 135 tp->t_template = tcp_template(tp);
290e0b0a
BJ
136 if (tp->t_template == 0) {
137 in_pcbdisconnect(inp);
138 error = ENOBUFS;
139 break;
140 }
53a5409e 141 soisconnecting(so);
a6503abf 142 tp->t_state = TCPS_SYN_SENT;
4aed14e3
BJ
143 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
144 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
145 tcp_sendseqinit(tp);
8a2f82db 146 error = tcp_output(tp);
72f24d7d 147 break;
4eb5d593 148
290e0b0a
BJ
149 /*
150 * Initiate disconnect from peer.
151 * If connection never passed embryonic stage, just drop;
152 * else if don't need to let data drain, then can just drop anyways,
153 * else have to begin TCP shutdown process: mark socket disconnecting,
154 * drain unread data, state switch to reflect user close, and
155 * send segment (e.g. FIN) to peer. Socket will be really disconnected
156 * when peer sends FIN and acks ours.
157 *
158 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
159 */
160 case PRU_DISCONNECT:
0e3936fa 161 tp = tcp_disconnect(tp);
4aed14e3
BJ
162 break;
163
290e0b0a
BJ
164 /*
165 * Accept a connection. Essentially all the work is
166 * done at higher levels; just return the address
167 * of the peer, storing through addr.
168 */
1acff8ec 169 case PRU_ACCEPT: {
8075bb0e 170 struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
1acff8ec 171
8075bb0e
BJ
172 nam->m_len = sizeof (struct sockaddr_in);
173 sin->sin_family = AF_INET;
174 sin->sin_port = inp->inp_fport;
175 sin->sin_addr = inp->inp_faddr;
eee3ab16 176 break;
8075bb0e 177 }
eee3ab16 178
290e0b0a
BJ
179 /*
180 * Mark the connection as being incapable of further output.
181 */
eee3ab16 182 case PRU_SHUTDOWN:
0974b45c 183 socantsendmore(so);
0e3936fa
SL
184 tp = tcp_usrclosed(tp);
185 if (tp)
186 error = tcp_output(tp);
72f24d7d
BJ
187 break;
188
290e0b0a
BJ
189 /*
190 * After a receive, possibly send window update to peer.
191 */
eee3ab16 192 case PRU_RCVD:
f1b2fa5b 193 (void) tcp_output(tp);
72f24d7d
BJ
194 break;
195
290e0b0a
BJ
196 /*
197 * Do a send by putting data in output queue and updating urgent
198 * marker if URG set. Possibly send more data.
199 */
eee3ab16 200 case PRU_SEND:
a6503abf 201 sbappend(&so->so_snd, m);
8a2f82db 202#ifdef notdef
0974b45c 203 if (tp->t_flags & TF_PUSH)
a6503abf 204 tp->snd_end = tp->snd_una + so->so_snd.sb_cc;
8a2f82db
SL
205#endif
206 error = tcp_output(tp);
72f24d7d
BJ
207 break;
208
290e0b0a
BJ
209 /*
210 * Abort the TCP.
211 */
eee3ab16 212 case PRU_ABORT:
0e3936fa 213 tp = tcp_drop(tp, ECONNABORTED);
72f24d7d
BJ
214 break;
215
290e0b0a 216/* SOME AS YET UNIMPLEMENTED HOOKS */
eee3ab16 217 case PRU_CONTROL:
53a5409e 218 error = EOPNOTSUPP;
eee3ab16
BJ
219 break;
220
f1b2fa5b
BJ
221 case PRU_SENSE:
222 error = EOPNOTSUPP;
223 break;
0244dbc7 224/* END UNIMPLEMENTED HOOKS */
f1b2fa5b
BJ
225
226 case PRU_RCVOOB:
8b5a83bb
BJ
227 if (so->so_oobmark == 0 &&
228 (so->so_state & SS_RCVATMARK) == 0) {
0244dbc7
BJ
229 error = EINVAL;
230 break;
231 }
b2db9217 232 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
8b5a83bb 233 error = EWOULDBLOCK;
b2db9217 234 break;
8b5a83bb 235 }
283ea225 236 m->m_len = 1;
b2db9217 237 *mtod(m, caddr_t) = tp->t_iobc;
f1b2fa5b
BJ
238 break;
239
240 case PRU_SENDOOB:
8b5a83bb
BJ
241 if (sbspace(&so->so_snd) < -512) {
242 error = ENOBUFS;
243 break;
244 }
0244dbc7
BJ
245 tp->snd_up = tp->snd_una + so->so_snd.sb_cc + 1;
246 sbappend(&so->so_snd, m);
b2db9217 247 tp->t_force = 1;
8a2f82db 248 error = tcp_output(tp);
b2db9217 249 tp->t_force = 0;
f1b2fa5b
BJ
250 break;
251
126472ab 252 case PRU_SOCKADDR:
8075bb0e 253 in_setsockaddr(inp, nam);
126472ab
SL
254 break;
255
290e0b0a
BJ
256 /*
257 * TCP slow timer went off; going through this
258 * routine for tracing's sake.
259 */
eee3ab16 260 case PRU_SLOWTIMO:
0e3936fa 261 tp = tcp_timers(tp, (int)nam);
8075bb0e 262 req |= (int)nam << 8; /* for debug's sake */
eee3ab16
BJ
263 break;
264
9c5022e3
BJ
265 default:
266 panic("tcp_usrreq");
72f24d7d 267 }
17b82ed4
BJ
268 if (tp && (so->so_options & SO_DEBUG))
269 tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req);
72f24d7d 270 splx(s);
53a5409e 271 return (error);
4eb5d593 272}
4aed14e3 273
306f91c9 274int tcp_sendspace = 1024*2;
1bc20e72 275int tcp_recvspace = 1024*2;
290e0b0a
BJ
276/*
277 * Attach TCP protocol to socket, allocating
278 * internet protocol control block, tcp control block,
279 * bufer space, and entering LISTEN state if to accept connections.
280 */
8075bb0e 281tcp_attach(so)
290e0b0a 282 struct socket *so;
290e0b0a
BJ
283{
284 register struct tcpcb *tp;
285 struct inpcb *inp;
286 int error;
287
59965020 288 error = soreserve(so, tcp_sendspace, tcp_recvspace);
ebf42a75
BJ
289 if (error)
290 goto bad;
291 error = in_pcballoc(so, &tcb);
290e0b0a 292 if (error)
8075bb0e
BJ
293 goto bad;
294 inp = sotoinpcb(so);
290e0b0a 295 tp = tcp_newtcpcb(inp);
ebf42a75
BJ
296 if (tp == 0) {
297 error = ENOBUFS;
298 goto bad2;
299 }
8075bb0e 300 tp->t_state = TCPS_CLOSED;
290e0b0a 301 return (0);
ebf42a75
BJ
302bad2:
303 in_pcbdetach(inp);
304bad:
305 return (error);
290e0b0a
BJ
306}
307
308/*
309 * Initiate (or continue) disconnect.
310 * If embryonic state, just send reset (once).
311 * If not in ``let data drain'' option, just drop.
312 * Otherwise (hard), mark socket disconnecting and drop
313 * current input data; switch states based on user close, and
314 * send segment to peer (with FIN).
315 */
0e3936fa 316struct tcpcb *
290e0b0a 317tcp_disconnect(tp)
0e3936fa 318 register struct tcpcb *tp;
290e0b0a
BJ
319{
320 struct socket *so = tp->t_inpcb->inp_socket;
321
322 if (tp->t_state < TCPS_ESTABLISHED)
0e3936fa 323 tp = tcp_close(tp);
8e65fd66 324 else if (so->so_linger == 0)
0e3936fa 325 tp = tcp_drop(tp, 0);
290e0b0a
BJ
326 else {
327 soisdisconnecting(so);
328 sbflush(&so->so_rcv);
0e3936fa
SL
329 tp = tcp_usrclosed(tp);
330 if (tp)
331 (void) tcp_output(tp);
290e0b0a 332 }
0e3936fa 333 return (tp);
290e0b0a
BJ
334}
335
336/*
337 * User issued close, and wish to trail through shutdown states:
338 * if never received SYN, just forget it. If got a SYN from peer,
339 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
340 * If already got a FIN from peer, then almost done; go to LAST_ACK
341 * state. In all other cases, have already sent FIN to peer (e.g.
342 * after PRU_SHUTDOWN), and just have to play tedious game waiting
343 * for peer to send FIN or not respond to keep-alives, etc.
085a0b90 344 * We can let the user exit from the close as soon as the FIN is acked.
290e0b0a 345 */
0e3936fa 346struct tcpcb *
4aed14e3 347tcp_usrclosed(tp)
0e3936fa 348 register struct tcpcb *tp;
4aed14e3
BJ
349{
350
4aed14e3
BJ
351 switch (tp->t_state) {
352
353 case TCPS_LISTEN:
354 case TCPS_SYN_SENT:
355 tp->t_state = TCPS_CLOSED;
0e3936fa 356 tp = tcp_close(tp);
4aed14e3
BJ
357 break;
358
359 case TCPS_SYN_RECEIVED:
360 case TCPS_ESTABLISHED:
361 tp->t_state = TCPS_FIN_WAIT_1;
362 break;
363
364 case TCPS_CLOSE_WAIT:
365 tp->t_state = TCPS_LAST_ACK;
366 break;
367 }
0e3936fa 368 if (tp && tp->t_state >= TCPS_FIN_WAIT_2)
085a0b90 369 soisdisconnected(tp->t_inpcb->inp_socket);
0e3936fa 370 return (tp);
4aed14e3 371}