Commit | Line | Data |
---|---|---|
ab303321 | 1 | /* uipc_socket2.c 4.36 83/05/01 */ |
681ebb17 BJ |
2 | |
3 | #include "../h/param.h" | |
4 | #include "../h/systm.h" | |
5 | #include "../h/dir.h" | |
6 | #include "../h/user.h" | |
7 | #include "../h/proc.h" | |
8 | #include "../h/file.h" | |
9 | #include "../h/inode.h" | |
10 | #include "../h/buf.h" | |
11 | #include "../h/mbuf.h" | |
681ebb17 BJ |
12 | #include "../h/protosw.h" |
13 | #include "../h/socket.h" | |
14 | #include "../h/socketvar.h" | |
681ebb17 BJ |
15 | |
16 | /* | |
17 | * Primitive routines for operating on sockets and socket buffers | |
18 | */ | |
19 | ||
20 | /* | |
21 | * Procedures to manipulate state flags of socket | |
2deddea9 BJ |
22 | * and do appropriate wakeups. Normal sequence from the |
23 | * active (originating) side is that soisconnecting() is | |
24 | * called during processing of connect() call, | |
4c078bb2 BJ |
25 | * resulting in an eventual call to soisconnected() if/when the |
26 | * connection is established. When the connection is torn down | |
27 | * soisdisconnecting() is called during processing of disconnect() call, | |
28 | * and soisdisconnected() is called when the connection to the peer | |
29 | * is totally severed. The semantics of these routines are such that | |
30 | * connectionless protocols can call soisconnected() and soisdisconnected() | |
31 | * only, bypassing the in-progress calls when setting up a ``connection'' | |
32 | * takes no time. | |
33 | * | |
2deddea9 BJ |
34 | * From the passive side, a socket is created with SO_ACCEPTCONN |
35 | * creating two queues of sockets: so_q0 for connections in progress | |
36 | * and so_q for connections already made and awaiting user acceptance. | |
37 | * As a protocol is preparing incoming connections, it creates a socket | |
38 | * structure queued on so_q0 by calling sonewconn(). When the connection | |
39 | * is established, soisconnected() is called, and transfers the | |
40 | * socket structure to so_q, making it available to accept(). | |
41 | * | |
42 | * If a SO_ACCEPTCONN socket is closed with sockets on either | |
43 | * so_q0 or so_q, these sockets are dropped. | |
44 | * | |
45 | * If and when higher level protocols are implemented in | |
4c078bb2 BJ |
46 | * the kernel, the wakeups done here will sometimes |
47 | * be implemented as software-interrupt process scheduling. | |
681ebb17 | 48 | */ |
4c078bb2 | 49 | |
681ebb17 BJ |
50 | soisconnecting(so) |
51 | struct socket *so; | |
52 | { | |
53 | ||
54 | so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); | |
55 | so->so_state |= SS_ISCONNECTING; | |
56 | wakeup((caddr_t)&so->so_timeo); | |
57 | } | |
58 | ||
59 | soisconnected(so) | |
60 | struct socket *so; | |
61 | { | |
2deddea9 | 62 | register struct socket *head = so->so_head; |
681ebb17 | 63 | |
2deddea9 BJ |
64 | if (head) { |
65 | if (soqremque(so, 0) == 0) | |
66 | panic("soisconnected"); | |
67 | soqinsque(head, so, 1); | |
68 | wakeup((caddr_t)&head->so_timeo); | |
ab303321 | 69 | sorwakeup(head); |
2deddea9 | 70 | } |
681ebb17 BJ |
71 | so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING); |
72 | so->so_state |= SS_ISCONNECTED; | |
73 | wakeup((caddr_t)&so->so_timeo); | |
f957a49a BJ |
74 | sorwakeup(so); |
75 | sowwakeup(so); | |
681ebb17 BJ |
76 | } |
77 | ||
78 | soisdisconnecting(so) | |
79 | struct socket *so; | |
80 | { | |
81 | ||
72857acf | 82 | so->so_state &= ~SS_ISCONNECTING; |
681ebb17 BJ |
83 | so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); |
84 | wakeup((caddr_t)&so->so_timeo); | |
4c078bb2 | 85 | sowwakeup(so); |
b454c3ea | 86 | sorwakeup(so); |
681ebb17 BJ |
87 | } |
88 | ||
89 | soisdisconnected(so) | |
90 | struct socket *so; | |
91 | { | |
92 | ||
93 | so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); | |
94 | so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE); | |
95 | wakeup((caddr_t)&so->so_timeo); | |
96 | sowwakeup(so); | |
97 | sorwakeup(so); | |
98 | } | |
99 | ||
2deddea9 BJ |
100 | /* |
101 | * When an attempt at a new connection is noted on a socket | |
102 | * which accepts connections, sonewconn is called. If the | |
103 | * connection is possible (subject to space constraints, etc.) | |
104 | * then we allocate a new structure, propoerly linked into the | |
105 | * data structure of the original socket, and return this. | |
106 | */ | |
107 | struct socket * | |
108 | sonewconn(head) | |
109 | register struct socket *head; | |
110 | { | |
111 | register struct socket *so; | |
112 | struct mbuf *m; | |
113 | ||
114 | if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2) | |
115 | goto bad; | |
cce93e4b | 116 | m = m_getclr(M_DONTWAIT, MT_SOCKET); |
5fe6f9d1 | 117 | if (m == NULL) |
2deddea9 BJ |
118 | goto bad; |
119 | so = mtod(m, struct socket *); | |
120 | so->so_type = head->so_type; | |
121 | so->so_options = head->so_options &~ SO_ACCEPTCONN; | |
122 | so->so_linger = head->so_linger; | |
f7428e88 | 123 | so->so_state = head->so_state | SS_NOFDREF; |
2deddea9 BJ |
124 | so->so_proto = head->so_proto; |
125 | so->so_timeo = head->so_timeo; | |
126 | so->so_pgrp = head->so_pgrp; | |
127 | soqinsque(head, so, 0); | |
5fe6f9d1 | 128 | if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0, |
c20f7455 | 129 | (struct mbuf *)0)) { |
2deddea9 | 130 | (void) soqremque(so, 0); |
30c36259 | 131 | (void) m_free(m); |
2deddea9 BJ |
132 | goto bad; |
133 | } | |
134 | return (so); | |
135 | bad: | |
136 | return ((struct socket *)0); | |
137 | } | |
138 | ||
139 | soqinsque(head, so, q) | |
140 | register struct socket *head, *so; | |
141 | int q; | |
142 | { | |
143 | ||
144 | so->so_head = head; | |
145 | if (q == 0) { | |
146 | head->so_q0len++; | |
147 | so->so_q0 = head->so_q0; | |
148 | head->so_q0 = so; | |
149 | } else { | |
150 | head->so_qlen++; | |
151 | so->so_q = head->so_q; | |
152 | head->so_q = so; | |
153 | } | |
154 | } | |
155 | ||
156 | soqremque(so, q) | |
157 | register struct socket *so; | |
158 | int q; | |
159 | { | |
160 | register struct socket *head, *prev, *next; | |
161 | ||
162 | head = so->so_head; | |
163 | prev = head; | |
164 | for (;;) { | |
165 | next = q ? prev->so_q : prev->so_q0; | |
166 | if (next == so) | |
167 | break; | |
168 | if (next == head) | |
169 | return (0); | |
170 | prev = next; | |
171 | } | |
172 | if (q == 0) { | |
173 | prev->so_q0 = next->so_q0; | |
174 | head->so_q0len--; | |
175 | } else { | |
176 | prev->so_q = next->so_q; | |
177 | head->so_qlen--; | |
178 | } | |
179 | next->so_q0 = next->so_q = 0; | |
180 | next->so_head = 0; | |
181 | return (1); | |
182 | } | |
183 | ||
4c078bb2 BJ |
184 | /* |
185 | * Socantsendmore indicates that no more data will be sent on the | |
186 | * socket; it would normally be applied to a socket when the user | |
187 | * informs the system that no more data is to be sent, by the protocol | |
188 | * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data | |
189 | * will be received, and will normally be applied to the socket by a | |
190 | * protocol when it detects that the peer will send no more data. | |
191 | * Data queued for reading in the socket may yet be read. | |
192 | */ | |
193 | ||
ae921915 BJ |
194 | socantsendmore(so) |
195 | struct socket *so; | |
196 | { | |
197 | ||
198 | so->so_state |= SS_CANTSENDMORE; | |
199 | sowwakeup(so); | |
200 | } | |
201 | ||
202 | socantrcvmore(so) | |
203 | struct socket *so; | |
204 | { | |
205 | ||
206 | so->so_state |= SS_CANTRCVMORE; | |
207 | sorwakeup(so); | |
208 | } | |
209 | ||
681ebb17 | 210 | /* |
4c078bb2 BJ |
211 | * Socket select/wakeup routines. |
212 | */ | |
213 | ||
214 | /* | |
215 | * Interface routine to select() system | |
216 | * call for sockets. | |
681ebb17 | 217 | */ |
477b2112 | 218 | soselect(so, rw) |
681ebb17 | 219 | register struct socket *so; |
477b2112 | 220 | int rw; |
681ebb17 | 221 | { |
f957a49a | 222 | int s = splnet(); |
681ebb17 | 223 | |
477b2112 BJ |
224 | switch (rw) { |
225 | ||
226 | case FREAD: | |
f957a49a BJ |
227 | if (soreadable(so)) { |
228 | splx(s); | |
681ebb17 | 229 | return (1); |
f957a49a | 230 | } |
681ebb17 | 231 | sbselqueue(&so->so_rcv); |
477b2112 BJ |
232 | break; |
233 | ||
234 | case FWRITE: | |
f957a49a BJ |
235 | if (sowriteable(so)) { |
236 | splx(s); | |
681ebb17 | 237 | return (1); |
f957a49a | 238 | } |
681ebb17 | 239 | sbselqueue(&so->so_snd); |
477b2112 | 240 | break; |
681ebb17 | 241 | } |
f957a49a | 242 | splx(s); |
681ebb17 BJ |
243 | return (0); |
244 | } | |
245 | ||
246 | /* | |
247 | * Queue a process for a select on a socket buffer. | |
248 | */ | |
249 | sbselqueue(sb) | |
250 | struct sockbuf *sb; | |
251 | { | |
252 | register struct proc *p; | |
253 | ||
ae921915 | 254 | if ((p = sb->sb_sel) && p->p_wchan == (caddr_t)&selwait) |
681ebb17 BJ |
255 | sb->sb_flags |= SB_COLL; |
256 | else | |
257 | sb->sb_sel = u.u_procp; | |
258 | } | |
259 | ||
ae921915 BJ |
260 | /* |
261 | * Wait for data to arrive at/drain from a socket buffer. | |
262 | */ | |
263 | sbwait(sb) | |
264 | struct sockbuf *sb; | |
265 | { | |
266 | ||
267 | sb->sb_flags |= SB_WAIT; | |
268 | sleep((caddr_t)&sb->sb_cc, PZERO+1); | |
269 | } | |
270 | ||
681ebb17 BJ |
271 | /* |
272 | * Wakeup processes waiting on a socket buffer. | |
273 | */ | |
274 | sbwakeup(sb) | |
275 | struct sockbuf *sb; | |
276 | { | |
277 | ||
278 | if (sb->sb_sel) { | |
279 | selwakeup(sb->sb_sel, sb->sb_flags & SB_COLL); | |
280 | sb->sb_sel = 0; | |
281 | sb->sb_flags &= ~SB_COLL; | |
282 | } | |
283 | if (sb->sb_flags & SB_WAIT) { | |
284 | sb->sb_flags &= ~SB_WAIT; | |
388ca8bd | 285 | wakeup((caddr_t)&sb->sb_cc); |
681ebb17 BJ |
286 | } |
287 | } | |
288 | ||
4c078bb2 BJ |
289 | /* |
290 | * Socket buffer (struct sockbuf) utility routines. | |
291 | * | |
292 | * Each socket contains two socket buffers: one for sending data and | |
293 | * one for receiving data. Each buffer contains a queue of mbufs, | |
294 | * information about the number of mbufs and amount of data in the | |
295 | * queue, and other fields allowing select() statements and notification | |
296 | * on data availability to be implemented. | |
297 | * | |
298 | * Before using a new socket structure it is first necessary to reserve | |
299 | * buffer space to the socket, by calling sbreserve. This commits | |
300 | * some of the available buffer space in the system buffer pool for the | |
301 | * socket. The space should be released by calling sbrelease when the | |
302 | * socket is destroyed. | |
303 | * | |
304 | * The routine sbappend() is normally called to append new mbufs | |
305 | * to a socket buffer, after checking that adequate space is available | |
306 | * comparing the function spspace() with the amount of data to be added. | |
307 | * Data is normally removed from a socket buffer in a protocol by | |
308 | * first calling m_copy on the socket buffer mbuf chain and sending this | |
309 | * to a peer, and then removing the data from the socket buffer with | |
310 | * sbdrop when the data is acknowledged by the peer (or immediately | |
b454c3ea | 311 | * in the case of unreliable protocols.) |
4c078bb2 BJ |
312 | * |
313 | * Protocols which do not require connections place both source address | |
314 | * and data information in socket buffer queues. The source addresses | |
315 | * are stored in single mbufs after each data item, and are easily found | |
316 | * as the data items are all marked with end of record markers. The | |
317 | * sbappendaddr() routine stores a datum and associated address in | |
318 | * a socket buffer. Note that, unlike sbappend(), this routine checks | |
319 | * for the caller that there will be enough space to store the data. | |
320 | * It fails if there is not enough space, or if it cannot find | |
321 | * a mbuf to store the address in. | |
322 | * | |
323 | * The higher-level routines sosend and soreceive (in socket.c) | |
b454c3ea | 324 | * also add data to, and remove data from socket buffers repectively. |
4c078bb2 BJ |
325 | */ |
326 | ||
0e18ec4a BJ |
327 | soreserve(so, sndcc, rcvcc) |
328 | struct socket *so; | |
329 | int sndcc, rcvcc; | |
330 | { | |
331 | ||
332 | if (sbreserve(&so->so_snd, sndcc) == 0) | |
333 | goto bad; | |
334 | if (sbreserve(&so->so_rcv, rcvcc) == 0) | |
335 | goto bad2; | |
336 | return (0); | |
337 | bad2: | |
338 | sbrelease(&so->so_snd); | |
339 | bad: | |
340 | return (ENOBUFS); | |
341 | } | |
342 | ||
681ebb17 BJ |
343 | /* |
344 | * Allot mbufs to a sockbuf. | |
345 | */ | |
346 | sbreserve(sb, cc) | |
347 | struct sockbuf *sb; | |
348 | { | |
349 | ||
de48daf3 | 350 | /* someday maybe this routine will fail... */ |
d028a086 | 351 | sb->sb_hiwat = cc; |
5fe6f9d1 | 352 | /* the 2 implies names can be no more than 1 mbuf each */ |
76a6e254 | 353 | sb->sb_mbmax = cc*2; |
ae921915 | 354 | return (1); |
681ebb17 BJ |
355 | } |
356 | ||
357 | /* | |
358 | * Free mbufs held by a socket, and reserved mbuf space. | |
359 | */ | |
360 | sbrelease(sb) | |
361 | struct sockbuf *sb; | |
362 | { | |
363 | ||
364 | sbflush(sb); | |
d028a086 | 365 | sb->sb_hiwat = sb->sb_mbmax = 0; |
681ebb17 BJ |
366 | } |
367 | ||
368 | /* | |
369 | * Routines to add (at the end) and remove (from the beginning) | |
370 | * data from a mbuf queue. | |
371 | */ | |
372 | ||
373 | /* | |
374 | * Append mbuf queue m to sockbuf sb. | |
375 | */ | |
376 | sbappend(sb, m) | |
377 | register struct mbuf *m; | |
378 | register struct sockbuf *sb; | |
379 | { | |
e495e1cc | 380 | register struct mbuf *n; |
681ebb17 | 381 | |
e495e1cc BJ |
382 | n = sb->sb_mb; |
383 | if (n) | |
384 | while (n->m_next) | |
385 | n = n->m_next; | |
681ebb17 | 386 | while (m) { |
a73ab5ae | 387 | if (m->m_len == 0 && (int)m->m_act == 0) { |
c64d826c | 388 | m = m_free(m); |
a73ab5ae BJ |
389 | continue; |
390 | } | |
681ebb17 BJ |
391 | if (n && n->m_off <= MMAXOFF && m->m_off <= MMAXOFF && |
392 | (int)n->m_act == 0 && (int)m->m_act == 0 && | |
76a6e254 BJ |
393 | (n->m_off + n->m_len + m->m_len) <= MMAXOFF) { |
394 | bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len, | |
ae921915 | 395 | (unsigned)m->m_len); |
681ebb17 BJ |
396 | n->m_len += m->m_len; |
397 | sb->sb_cc += m->m_len; | |
398 | m = m_free(m); | |
399 | continue; | |
400 | } | |
401 | sballoc(sb, m); | |
e495e1cc BJ |
402 | if (n == 0) |
403 | sb->sb_mb = m; | |
404 | else | |
405 | n->m_next = m; | |
681ebb17 | 406 | n = m; |
681ebb17 | 407 | m = m->m_next; |
e495e1cc | 408 | n->m_next = 0; |
681ebb17 BJ |
409 | } |
410 | } | |
411 | ||
4c078bb2 BJ |
412 | /* |
413 | * Append data and address. | |
414 | * Return 0 if no space in sockbuf or if | |
415 | * can't get mbuf to stuff address in. | |
416 | */ | |
2b4b57cd BJ |
417 | sbappendaddr(sb, asa, m0) |
418 | struct sockbuf *sb; | |
419 | struct sockaddr *asa; | |
420 | struct mbuf *m0; | |
421 | { | |
422 | struct sockaddr *msa; | |
423 | register struct mbuf *m; | |
424 | register int len = sizeof (struct sockaddr); | |
425 | ||
76a6e254 BJ |
426 | m = m0; |
427 | if (m == 0) | |
428 | panic("sbappendaddr"); | |
429 | for (;;) { | |
2b4b57cd | 430 | len += m->m_len; |
76a6e254 BJ |
431 | if (m->m_next == 0) { |
432 | m->m_act = (struct mbuf *)1; | |
433 | break; | |
434 | } | |
435 | m = m->m_next; | |
436 | } | |
509e40dd | 437 | if (len > sbspace(sb)) |
2b4b57cd | 438 | return (0); |
cce93e4b | 439 | m = m_get(M_DONTWAIT, MT_SONAME); |
509e40dd | 440 | if (m == 0) |
2b4b57cd | 441 | return (0); |
2b4b57cd BJ |
442 | m->m_len = sizeof (struct sockaddr); |
443 | msa = mtod(m, struct sockaddr *); | |
444 | *msa = *asa; | |
445 | m->m_act = (struct mbuf *)1; | |
446 | sbappend(sb, m); | |
2b4b57cd BJ |
447 | sbappend(sb, m0); |
448 | return (1); | |
449 | } | |
450 | ||
5fe6f9d1 | 451 | #ifdef notdef |
e435773e BJ |
452 | SBCHECK(sb, str) |
453 | struct sockbuf *sb; | |
454 | char *str; | |
455 | { | |
456 | register int cnt = sb->sb_cc; | |
457 | register int mbcnt = sb->sb_mbcnt; | |
458 | register struct mbuf *m; | |
459 | ||
460 | for (m = sb->sb_mb; m; m = m->m_next) { | |
461 | cnt -= m->m_len; | |
462 | mbcnt -= MSIZE; | |
463 | if (m->m_off > MMAXOFF) | |
464 | mbcnt -= CLBYTES; | |
465 | } | |
466 | if (cnt || mbcnt) { | |
467 | printf("cnt %d mbcnt %d\n", cnt, mbcnt); | |
468 | panic(str); | |
469 | } | |
470 | } | |
5fe6f9d1 | 471 | #endif |
e435773e | 472 | |
681ebb17 BJ |
473 | /* |
474 | * Free all mbufs on a sockbuf mbuf chain. | |
475 | * Check that resource allocations return to 0. | |
476 | */ | |
477 | sbflush(sb) | |
478 | struct sockbuf *sb; | |
479 | { | |
480 | ||
481 | if (sb->sb_flags & SB_LOCK) | |
482 | panic("sbflush"); | |
a73ab5ae BJ |
483 | if (sb->sb_cc) |
484 | sbdrop(sb, sb->sb_cc); | |
681ebb17 BJ |
485 | if (sb->sb_cc || sb->sb_mbcnt || sb->sb_mb) |
486 | panic("sbflush 2"); | |
487 | } | |
488 | ||
489 | /* | |
490 | * Drop data from (the front of) a sockbuf chain. | |
491 | */ | |
492 | sbdrop(sb, len) | |
493 | register struct sockbuf *sb; | |
494 | register int len; | |
495 | { | |
496 | register struct mbuf *m = sb->sb_mb, *mn; | |
497 | ||
498 | while (len > 0) { | |
499 | if (m == 0) | |
500 | panic("sbdrop"); | |
b9f0d37f | 501 | if (m->m_len > len) { |
681ebb17 BJ |
502 | m->m_len -= len; |
503 | m->m_off += len; | |
504 | sb->sb_cc -= len; | |
505 | break; | |
506 | } | |
b9f0d37f BJ |
507 | len -= m->m_len; |
508 | sbfree(sb, m); | |
509 | MFREE(m, mn); | |
510 | m = mn; | |
681ebb17 BJ |
511 | } |
512 | sb->sb_mb = m; | |
513 | } |