socket: Replicate soreceive() to sorecvtcp() for cleanup and optimization
[dragonfly.git] / sys / kern / uipc_socket.c
CommitLineData
984263bc 1/*
6ea1e9b9 2 * Copyright (c) 2004 Jeffrey M. Hsu. All rights reserved.
66d6c637
JH
3 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
4 *
5 * This code is derived from software contributed to The DragonFly Project
6 * by Jeffrey M. Hsu.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33
34/*
984263bc
MD
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94
7405c902 67 * $FreeBSD: src/sys/kern/uipc_socket.c,v 1.68.2.24 2003/11/11 17:18:18 silby Exp $
984263bc
MD
68 */
69
70#include "opt_inet.h"
78812139 71#include "opt_sctp.h"
984263bc
MD
72
73#include <sys/param.h>
74#include <sys/systm.h>
75#include <sys/fcntl.h>
76#include <sys/malloc.h>
77#include <sys/mbuf.h>
78#include <sys/domain.h>
79#include <sys/file.h> /* for struct knote */
80#include <sys/kernel.h>
984263bc 81#include <sys/event.h>
984263bc
MD
82#include <sys/proc.h>
83#include <sys/protosw.h>
84#include <sys/socket.h>
85#include <sys/socketvar.h>
6b6e0885 86#include <sys/socketops.h>
984263bc
MD
87#include <sys/resourcevar.h>
88#include <sys/signalvar.h>
89#include <sys/sysctl.h>
90#include <sys/uio.h>
91#include <sys/jail.h>
92#include <vm/vm_zone.h>
e71a125f 93#include <vm/pmap.h>
acd31a69 94#include <net/netmsg2.h>
984263bc 95
e43a034f 96#include <sys/thread2.h>
d6cb521d 97#include <sys/socketvar2.h>
e43a034f 98
984263bc
MD
99#include <machine/limits.h>
100
0df7608b
SZ
101extern int tcp_sosend_agglim;
102extern int tcp_sosend_async;
103extern int udp_sosend_async;
c16aca65 104
984263bc
MD
105#ifdef INET
106static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
107#endif /* INET */
108
109static void filt_sordetach(struct knote *kn);
110static int filt_soread(struct knote *kn, long hint);
111static void filt_sowdetach(struct knote *kn);
112static int filt_sowrite(struct knote *kn, long hint);
113static int filt_solisten(struct knote *kn, long hint);
114
acd31a69
SZ
115static void sodiscard(struct socket *so);
116static int soclose_sync(struct socket *so, int fflag);
117static void soclose_fast(struct socket *so);
118
984263bc 119static struct filterops solisten_filtops =
8d1b9f93 120 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, filt_sordetach, filt_solisten };
984263bc 121static struct filterops soread_filtops =
8d1b9f93 122 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, filt_sordetach, filt_soread };
984263bc 123static struct filterops sowrite_filtops =
8d1b9f93 124 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, filt_sowdetach, filt_sowrite };
73c344d3 125static struct filterops soexcept_filtops =
8d1b9f93 126 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, filt_sordetach, filt_soread };
984263bc 127
69ea5b8d 128MALLOC_DEFINE(M_SOCKET, "socket", "socket struct");
984263bc
MD
129MALLOC_DEFINE(M_SONAME, "soname", "socket name");
130MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
131
984263bc
MD
132
133static int somaxconn = SOMAXCONN;
134SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
135 &somaxconn, 0, "Maximum pending socket connection queue size");
136
acd31a69
SZ
137static int use_soclose_fast = 1;
138SYSCTL_INT(_kern_ipc, OID_AUTO, soclose_fast, CTLFLAG_RW,
139 &use_soclose_fast, 0, "Fast socket close");
140
5e4b3994
SZ
141int use_soaccept_pred_fast = 1;
142SYSCTL_INT(_kern_ipc, OID_AUTO, soaccept_pred_fast, CTLFLAG_RW,
143 &use_soaccept_pred_fast, 0, "Fast socket accept predication");
144
f786a6ac 145int use_sendfile_async = 1;
828cf263
SZ
146SYSCTL_INT(_kern_ipc, OID_AUTO, sendfile_async, CTLFLAG_RW,
147 &use_sendfile_async, 0, "sendfile uses asynchronized pru_send");
148
984263bc
MD
149/*
150 * Socket operation routines.
151 * These routines are called by the routines in
152 * sys_socket.c or from a system process, and
153 * implement the semantics of socket operations by
154 * switching out to the protocol specific routines.
155 */
156
157/*
69ea5b8d 158 * Get a socket structure, and initialize it.
984263bc
MD
159 * Note that it would probably be better to allocate socket
160 * and PCB at the same time, but I'm not convinced that all
161 * the protocols can be easily modified to do this.
162 */
163struct socket *
c972a82f 164soalloc(int waitok)
984263bc
MD
165{
166 struct socket *so;
69ea5b8d 167 unsigned waitmask;
984263bc 168
69ea5b8d
NT
169 waitmask = waitok ? M_WAITOK : M_NOWAIT;
170 so = kmalloc(sizeof(struct socket), M_SOCKET, M_ZERO|waitmask);
984263bc
MD
171 if (so) {
172 /* XXX race condition for reentrant kernel */
984263bc 173 TAILQ_INIT(&so->so_aiojobq);
5b22f1a7
SG
174 TAILQ_INIT(&so->so_rcv.ssb_kq.ki_mlist);
175 TAILQ_INIT(&so->so_snd.ssb_kq.ki_mlist);
a3c18566
MD
176 lwkt_token_init(&so->so_rcv.ssb_token, "rcvtok");
177 lwkt_token_init(&so->so_snd.ssb_token, "sndtok");
6cef7136
MD
178 so->so_state = SS_NOFDREF;
179 so->so_refs = 1;
984263bc
MD
180 }
181 return so;
182}
183
184int
dadab5e9
MD
185socreate(int dom, struct socket **aso, int type,
186 int proto, struct thread *td)
984263bc 187{
dadab5e9
MD
188 struct proc *p = td->td_proc;
189 struct protosw *prp;
190 struct socket *so;
e4700d00 191 struct pru_attach_info ai;
dadab5e9 192 int error;
984263bc
MD
193
194 if (proto)
195 prp = pffindproto(dom, proto, type);
196 else
197 prp = pffindtype(dom, type);
198
4090d6ff 199 if (prp == NULL || prp->pr_usrreqs->pru_attach == 0)
984263bc
MD
200 return (EPROTONOSUPPORT);
201
41c20dac 202 if (p->p_ucred->cr_prison && jail_socket_unixiproute_only &&
984263bc
MD
203 prp->pr_domain->dom_family != PF_LOCAL &&
204 prp->pr_domain->dom_family != PF_INET &&
3e4150ef 205 prp->pr_domain->dom_family != PF_INET6 &&
984263bc
MD
206 prp->pr_domain->dom_family != PF_ROUTE) {
207 return (EPROTONOSUPPORT);
208 }
209
210 if (prp->pr_type != type)
211 return (EPROTOTYPE);
4090d6ff 212 so = soalloc(p != NULL);
6cef7136 213 if (so == NULL)
984263bc
MD
214 return (ENOBUFS);
215
48e7b118 216 /*
6cef7136
MD
217 * Callers of socreate() presumably will connect up a descriptor
218 * and call soclose() if they cannot. This represents our so_refs
219 * (which should be 1) from soalloc().
220 */
221 soclrstate(so, SS_NOFDREF);
222
223 /*
48e7b118
MD
224 * Set a default port for protocol processing. No action will occur
225 * on the socket on this port until an inpcb is attached to it and
226 * is able to match incoming packets, or until the socket becomes
227 * available to userland.
002c1265
MD
228 *
229 * We normally default the socket to the protocol thread on cpu 0.
230 * If PR_SYNC_PORT is set (unix domain sockets) there is no protocol
231 * thread and all pr_*()/pru_*() calls are executed synchronously.
48e7b118 232 */
002c1265
MD
233 if (prp->pr_flags & PR_SYNC_PORT)
234 so->so_port = &netisr_sync_port;
235 else
236 so->so_port = cpu_portfn(0);
48e7b118 237
984263bc
MD
238 TAILQ_INIT(&so->so_incomp);
239 TAILQ_INIT(&so->so_comp);
240 so->so_type = type;
e9a372eb 241 so->so_cred = crhold(p->p_ucred);
984263bc 242 so->so_proto = prp;
e4700d00
JH
243 ai.sb_rlimit = &p->p_rlimit[RLIMIT_SBSIZE];
244 ai.p_ucred = p->p_ucred;
245 ai.fd_rdir = p->p_fd->fd_rdir;
48e7b118 246
5b0b9fa5
PA
247 /*
248 * Auto-sizing of socket buffers is managed by the protocols and
249 * the appropriate flags must be set in the pru_attach function.
250 */
e4700d00 251 error = so_pru_attach(so, proto, &ai);
984263bc 252 if (error) {
6cef7136
MD
253 sosetstate(so, SS_NOFDREF);
254 sofree(so); /* from soalloc */
255 return error;
984263bc 256 }
48e7b118 257
6cef7136
MD
258 /*
259 * NOTE: Returns referenced socket.
260 */
984263bc
MD
261 *aso = so;
262 return (0);
263}
264
265int
dadab5e9 266sobind(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc 267{
984263bc
MD
268 int error;
269
6b6e0885 270 error = so_pru_bind(so, nam, td);
984263bc
MD
271 return (error);
272}
273
6cef7136 274static void
dadab5e9 275sodealloc(struct socket *so)
984263bc 276{
6d49aa6f 277 if (so->so_rcv.ssb_hiwat)
984263bc 278 (void)chgsbsize(so->so_cred->cr_uidinfo,
6d49aa6f
MD
279 &so->so_rcv.ssb_hiwat, 0, RLIM_INFINITY);
280 if (so->so_snd.ssb_hiwat)
984263bc 281 (void)chgsbsize(so->so_cred->cr_uidinfo,
6d49aa6f 282 &so->so_snd.ssb_hiwat, 0, RLIM_INFINITY);
984263bc 283#ifdef INET
81d59d3d
HP
284 /* remove accept filter if present */
285 if (so->so_accf != NULL)
286 do_setopt_accept_filter(so, NULL);
984263bc
MD
287#endif /* INET */
288 crfree(so->so_cred);
88da6203
SZ
289 if (so->so_faddr != NULL)
290 kfree(so->so_faddr, M_SONAME);
69ea5b8d 291 kfree(so, M_SOCKET);
984263bc
MD
292}
293
294int
dadab5e9 295solisten(struct socket *so, int backlog, struct thread *td)
984263bc 296{
e43a034f 297 int error;
78812139
EN
298#ifdef SCTP
299 short oldopt, oldqlimit;
300#endif /* SCTP */
984263bc 301
6cef7136 302 if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING))
78812139 303 return (EINVAL);
78812139
EN
304
305#ifdef SCTP
306 oldopt = so->so_options;
307 oldqlimit = so->so_qlimit;
308#endif /* SCTP */
309
6cef7136 310 lwkt_gettoken(&so->so_rcv.ssb_token);
984263bc
MD
311 if (TAILQ_EMPTY(&so->so_comp))
312 so->so_options |= SO_ACCEPTCONN;
6cef7136 313 lwkt_reltoken(&so->so_rcv.ssb_token);
984263bc
MD
314 if (backlog < 0 || backlog > somaxconn)
315 backlog = somaxconn;
316 so->so_qlimit = backlog;
78812139
EN
317 /* SCTP needs to look at tweak both the inbound backlog parameter AND
318 * the so_options (UDP model both connect's and gets inbound
319 * connections .. implicitly).
320 */
321 error = so_pru_listen(so, td);
322 if (error) {
323#ifdef SCTP
324 /* Restore the params */
325 so->so_options = oldopt;
326 so->so_qlimit = oldqlimit;
327#endif /* SCTP */
78812139
EN
328 return (error);
329 }
984263bc
MD
330 return (0);
331}
332
4402d8a2
MD
333/*
334 * Destroy a disconnected socket. This routine is a NOP if entities
335 * still have a reference on the socket:
336 *
337 * so_pcb - The protocol stack still has a reference
338 * SS_NOFDREF - There is no longer a file pointer reference
4402d8a2 339 */
984263bc 340void
dadab5e9 341sofree(struct socket *so)
984263bc 342{
5217bcbc
MD
343 struct socket *head;
344
345 /*
346 * This is a bit hackish at the moment. We need to interlock
347 * any accept queue we are on before we potentially lose the
348 * last reference to avoid races against a re-reference from
349 * someone operating on the queue.
350 */
351 while ((head = so->so_head) != NULL) {
352 lwkt_getpooltoken(head);
353 if (so->so_head == head)
354 break;
355 lwkt_relpooltoken(head);
356 }
984263bc 357
6cef7136
MD
358 /*
359 * Arbitrage the last free.
360 */
361 KKASSERT(so->so_refs > 0);
5217bcbc
MD
362 if (atomic_fetchadd_int(&so->so_refs, -1) != 1) {
363 if (head)
364 lwkt_relpooltoken(head);
4402d8a2 365 return;
5217bcbc 366 }
6cef7136
MD
367
368 KKASSERT(so->so_pcb == NULL && (so->so_state & SS_NOFDREF));
e28d8186 369 KKASSERT((so->so_state & SS_ASSERTINPROG) == 0);
6cef7136
MD
370
371 /*
5217bcbc
MD
372 * We're done, remove ourselves from the accept queue we are
373 * on, if we are on one.
6cef7136 374 */
984263bc
MD
375 if (head != NULL) {
376 if (so->so_state & SS_INCOMP) {
377 TAILQ_REMOVE(&head->so_incomp, so, so_list);
378 head->so_incqlen--;
379 } else if (so->so_state & SS_COMP) {
380 /*
381 * We must not decommission a socket that's
382 * on the accept(2) queue. If we do, then
383 * accept(2) may hang after select(2) indicated
384 * that the listening socket was ready.
385 */
5217bcbc 386 lwkt_relpooltoken(head);
984263bc
MD
387 return;
388 } else {
389 panic("sofree: not queued");
390 }
6cef7136 391 soclrstate(so, SS_INCOMP);
984263bc 392 so->so_head = NULL;
5217bcbc 393 lwkt_relpooltoken(head);
984263bc 394 }
6d49aa6f 395 ssb_release(&so->so_snd, so);
984263bc
MD
396 sorflush(so);
397 sodealloc(so);
398}
399
400/*
401 * Close a socket on last file table reference removal.
402 * Initiate disconnect if connected.
403 * Free socket when disconnect complete.
404 */
405int
9ba76b73 406soclose(struct socket *so, int fflag)
984263bc 407{
acd31a69 408 int error;
984263bc 409
58c2553a 410 funsetown(&so->so_sigio);
acd31a69
SZ
411 if (!use_soclose_fast ||
412 (so->so_proto->pr_flags & PR_SYNC_PORT) ||
413 (so->so_options & SO_LINGER)) {
414 error = soclose_sync(so, fflag);
415 } else {
416 soclose_fast(so);
417 error = 0;
418 }
419 return error;
420}
421
422static void
423sodiscard(struct socket *so)
424{
425 lwkt_getpooltoken(so);
426 if (so->so_options & SO_ACCEPTCONN) {
427 struct socket *sp;
428
429 while ((sp = TAILQ_FIRST(&so->so_incomp)) != NULL) {
430 TAILQ_REMOVE(&so->so_incomp, sp, so_list);
431 soclrstate(sp, SS_INCOMP);
432 sp->so_head = NULL;
433 so->so_incqlen--;
434 soaborta(sp);
435 }
436 while ((sp = TAILQ_FIRST(&so->so_comp)) != NULL) {
437 TAILQ_REMOVE(&so->so_comp, sp, so_list);
438 soclrstate(sp, SS_COMP);
439 sp->so_head = NULL;
440 so->so_qlen--;
441 soaborta(sp);
442 }
443 }
444 lwkt_relpooltoken(so);
445
446 if (so->so_state & SS_NOFDREF)
447 panic("soclose: NOFDREF");
448 sosetstate(so, SS_NOFDREF); /* take ref */
449}
450
451static int
452soclose_sync(struct socket *so, int fflag)
453{
454 int error = 0;
455
19be7d32 456 if (so->so_pcb == NULL)
984263bc
MD
457 goto discard;
458 if (so->so_state & SS_ISCONNECTED) {
459 if ((so->so_state & SS_ISDISCONNECTING) == 0) {
460 error = sodisconnect(so);
461 if (error)
462 goto drop;
463 }
464 if (so->so_options & SO_LINGER) {
465 if ((so->so_state & SS_ISDISCONNECTING) &&
9ba76b73 466 (fflag & FNONBLOCK))
984263bc
MD
467 goto drop;
468 while (so->so_state & SS_ISCONNECTED) {
6cef7136
MD
469 error = tsleep(&so->so_timeo, PCATCH,
470 "soclos", so->so_linger * hz);
984263bc
MD
471 if (error)
472 break;
473 }
474 }
475 }
476drop:
477 if (so->so_pcb) {
6b6e0885
JH
478 int error2;
479
480 error2 = so_pru_detach(so);
984263bc
MD
481 if (error == 0)
482 error = error2;
483 }
484discard:
acd31a69
SZ
485 sodiscard(so);
486 so_pru_sync(so); /* unpend async sending */
487 sofree(so); /* dispose of ref */
19be7d32 488
acd31a69
SZ
489 return (error);
490}
491
492static void
493soclose_sofree_async_handler(netmsg_t msg)
494{
495 sofree(msg->base.nm_so);
496}
497
498static void
499soclose_sofree_async(struct socket *so)
500{
501 struct netmsg_base *base = &so->so_clomsg;
502
503 netmsg_init(base, so, &netisr_apanic_rport, 0,
504 soclose_sofree_async_handler);
505 lwkt_sendmsg(so->so_port, &base->lmsg);
506}
507
508static void
509soclose_disconn_async_handler(netmsg_t msg)
510{
511 struct socket *so = msg->base.nm_so;
512
513 if ((so->so_state & SS_ISCONNECTED) &&
514 (so->so_state & SS_ISDISCONNECTING) == 0)
515 so_pru_disconnect_direct(so);
516
517 if (so->so_pcb)
518 so_pru_detach_direct(so);
519
520 sodiscard(so);
521 sofree(so);
522}
523
524static void
525soclose_disconn_async(struct socket *so)
526{
527 struct netmsg_base *base = &so->so_clomsg;
528
529 netmsg_init(base, so, &netisr_apanic_rport, 0,
530 soclose_disconn_async_handler);
531 lwkt_sendmsg(so->so_port, &base->lmsg);
532}
533
534static void
535soclose_detach_async_handler(netmsg_t msg)
536{
537 struct socket *so = msg->base.nm_so;
538
539 if (so->so_pcb)
540 so_pru_detach_direct(so);
541
542 sodiscard(so);
543 sofree(so);
544}
545
546static void
547soclose_detach_async(struct socket *so)
548{
549 struct netmsg_base *base = &so->so_clomsg;
550
551 netmsg_init(base, so, &netisr_apanic_rport, 0,
552 soclose_detach_async_handler);
553 lwkt_sendmsg(so->so_port, &base->lmsg);
554}
555
556static void
557soclose_fast(struct socket *so)
558{
559 if (so->so_pcb == NULL)
560 goto discard;
561
562 if ((so->so_state & SS_ISCONNECTED) &&
563 (so->so_state & SS_ISDISCONNECTING) == 0) {
564 soclose_disconn_async(so);
565 return;
19be7d32 566 }
939c81e4 567
acd31a69
SZ
568 if (so->so_pcb) {
569 soclose_detach_async(so);
570 return;
571 }
572
573discard:
574 sodiscard(so);
575 soclose_sofree_async(so);
984263bc
MD
576}
577
578/*
9116be8e
MD
579 * Abort and destroy a socket. Only one abort can be in progress
580 * at any given moment.
984263bc 581 */
4402d8a2 582void
c972a82f 583soabort(struct socket *so)
984263bc 584{
6cef7136
MD
585 soreference(so);
586 so_pru_abort(so);
4402d8a2 587}
984263bc 588
4402d8a2
MD
589void
590soaborta(struct socket *so)
591{
6cef7136
MD
592 soreference(so);
593 so_pru_aborta(so);
984263bc
MD
594}
595
fd86a41c
SZ
596void
597soabort_oncpu(struct socket *so)
598{
6cef7136
MD
599 soreference(so);
600 so_pru_abort_oncpu(so);
fd86a41c
SZ
601}
602
c19fdb0e
MD
603/*
604 * so is passed in ref'd, which becomes owned by
605 * the cleared SS_NOFDREF flag.
606 */
88da6203
SZ
607void
608soaccept_generic(struct socket *so)
609{
610 if ((so->so_state & SS_NOFDREF) == 0)
611 panic("soaccept: !NOFDREF");
612 soclrstate(so, SS_NOFDREF); /* owned by lack of SS_NOFDREF */
613}
614
984263bc 615int
dadab5e9 616soaccept(struct socket *so, struct sockaddr **nam)
984263bc 617{
984263bc
MD
618 int error;
619
88da6203 620 soaccept_generic(so);
f328adf5 621 error = so_pru_accept(so, nam);
984263bc
MD
622 return (error);
623}
624
625int
dadab5e9 626soconnect(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc 627{
984263bc
MD
628 int error;
629
630 if (so->so_options & SO_ACCEPTCONN)
631 return (EOPNOTSUPP);
984263bc
MD
632 /*
633 * If protocol is connection-based, can only connect once.
634 * Otherwise, if connected, try to disconnect first.
635 * This allows user to disconnect by connecting to, e.g.,
636 * a null address.
637 */
638 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
639 ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
59429d28 640 (error = sodisconnect(so)))) {
984263bc 641 error = EISCONN;
59429d28
MD
642 } else {
643 /*
644 * Prevent accumulated error from previous connection
645 * from biting us.
646 */
647 so->so_error = 0;
6b6e0885 648 error = so_pru_connect(so, nam, td);
59429d28 649 }
984263bc
MD
650 return (error);
651}
652
653int
dadab5e9 654soconnect2(struct socket *so1, struct socket *so2)
984263bc 655{
984263bc
MD
656 int error;
657
6b6e0885 658 error = so_pru_connect2(so1, so2);
984263bc
MD
659 return (error);
660}
661
662int
dadab5e9 663sodisconnect(struct socket *so)
984263bc 664{
984263bc
MD
665 int error;
666
667 if ((so->so_state & SS_ISCONNECTED) == 0) {
668 error = ENOTCONN;
669 goto bad;
670 }
671 if (so->so_state & SS_ISDISCONNECTING) {
672 error = EALREADY;
673 goto bad;
674 }
6b6e0885 675 error = so_pru_disconnect(so);
984263bc 676bad:
984263bc
MD
677 return (error);
678}
679
680#define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
681/*
682 * Send on a socket.
683 * If send must go all at once and message is larger than
684 * send buffering, then hard error.
685 * Lock against other senders.
686 * If must go all at once and not enough room now, then
687 * inform user that this would block and do nothing.
688 * Otherwise, if nonblocking, send as much as possible.
689 * The data to be sent is described by "uio" if nonzero,
690 * otherwise by the mbuf chain "top" (which must be null
691 * if uio is not). Data provided in mbuf chain must be small
692 * enough to send all at once.
693 *
694 * Returns nonzero on error, timeout or signal; callers
695 * must check for short counts if EINTR/ERESTART are returned.
696 * Data and control buffers are freed on return.
697 */
698int
dadab5e9
MD
699sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
700 struct mbuf *top, struct mbuf *control, int flags,
701 struct thread *td)
984263bc
MD
702{
703 struct mbuf **mp;
dadab5e9 704 struct mbuf *m;
e54488bb
MD
705 size_t resid;
706 int space, len;
e43a034f 707 int clen = 0, error, dontroute, mlen;
984263bc 708 int atomic = sosendallatonce(so) || top;
6b6e0885 709 int pru_flags;
984263bc 710
5bd48c1d 711 if (uio) {
984263bc 712 resid = uio->uio_resid;
5bd48c1d 713 } else {
e54488bb 714 resid = (size_t)top->m_pkthdr.len;
5bd48c1d
MD
715#ifdef INVARIANTS
716 len = 0;
717 for (m = top; m; m = m->m_next)
718 len += m->m_len;
719 KKASSERT(top->m_pkthdr.len == len);
720#endif
721 }
48e7b118 722
984263bc 723 /*
e54488bb
MD
724 * WARNING! resid is unsigned, space and len are signed. space
725 * can wind up negative if the sockbuf is overcommitted.
984263bc
MD
726 *
727 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
728 * type sockets since that's an error.
729 */
e54488bb 730 if (so->so_type == SOCK_STREAM && (flags & MSG_EOR)) {
984263bc
MD
731 error = EINVAL;
732 goto out;
733 }
734
735 dontroute =
736 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
737 (so->so_proto->pr_flags & PR_ATOMIC);
fde7ac71
SS
738 if (td->td_lwp != NULL)
739 td->td_lwp->lwp_ru.ru_msgsnd++;
984263bc
MD
740 if (control)
741 clen = control->m_len;
6cef7136 742#define gotoerr(errcode) { error = errcode; goto release; }
984263bc
MD
743
744restart:
6d49aa6f 745 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags));
984263bc
MD
746 if (error)
747 goto out;
48e7b118 748
984263bc 749 do {
984263bc 750 if (so->so_state & SS_CANTSENDMORE)
6ea1e9b9 751 gotoerr(EPIPE);
984263bc
MD
752 if (so->so_error) {
753 error = so->so_error;
754 so->so_error = 0;
984263bc
MD
755 goto release;
756 }
757 if ((so->so_state & SS_ISCONNECTED) == 0) {
758 /*
759 * `sendto' and `sendmsg' is allowed on a connection-
760 * based socket if it supports implied connect.
761 * Return ENOTCONN if not connected and no address is
762 * supplied.
763 */
764 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
765 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
766 if ((so->so_state & SS_ISCONFIRMING) == 0 &&
767 !(resid == 0 && clen != 0))
6ea1e9b9 768 gotoerr(ENOTCONN);
4090d6ff 769 } else if (addr == NULL)
6ea1e9b9 770 gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
984263bc
MD
771 ENOTCONN : EDESTADDRREQ);
772 }
3a6117bb
MD
773 if ((atomic && resid > so->so_snd.ssb_hiwat) ||
774 clen > so->so_snd.ssb_hiwat) {
775 gotoerr(EMSGSIZE);
776 }
6d49aa6f 777 space = ssb_space(&so->so_snd);
984263bc
MD
778 if (flags & MSG_OOB)
779 space += 1024;
e54488bb 780 if ((space < 0 || (size_t)space < resid + clen) && uio &&
6d49aa6f 781 (atomic || space < so->so_snd.ssb_lowat || space < clen)) {
9ba76b73 782 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT))
6ea1e9b9 783 gotoerr(EWOULDBLOCK);
6d49aa6f
MD
784 ssb_unlock(&so->so_snd);
785 error = ssb_wait(&so->so_snd);
984263bc
MD
786 if (error)
787 goto out;
788 goto restart;
789 }
984263bc
MD
790 mp = &top;
791 space -= clen;
792 do {
793 if (uio == NULL) {
794 /*
795 * Data is prepackaged in "top".
796 */
797 resid = 0;
798 if (flags & MSG_EOR)
799 top->m_flags |= M_EOR;
800 } else do {
e54488bb
MD
801 if (resid > INT_MAX)
802 resid = INT_MAX;
803 m = m_getl((int)resid, MB_WAIT, MT_DATA,
50503f0f
JH
804 top == NULL ? M_PKTHDR : 0, &mlen);
805 if (top == NULL) {
984263bc 806 m->m_pkthdr.len = 0;
60233e58 807 m->m_pkthdr.rcvif = NULL;
984263bc 808 }
e54488bb 809 len = imin((int)szmin(mlen, resid), space);
50503f0f 810 if (resid < MINCLSIZE) {
984263bc
MD
811 /*
812 * For datagram protocols, leave room
813 * for protocol headers in first mbuf.
814 */
4090d6ff 815 if (atomic && top == NULL && len < mlen)
984263bc
MD
816 MH_ALIGN(m, len);
817 }
818 space -= len;
e54488bb 819 error = uiomove(mtod(m, caddr_t), (size_t)len, uio);
984263bc
MD
820 resid = uio->uio_resid;
821 m->m_len = len;
822 *mp = m;
823 top->m_pkthdr.len += len;
824 if (error)
825 goto release;
826 mp = &m->m_next;
e54488bb 827 if (resid == 0) {
984263bc
MD
828 if (flags & MSG_EOR)
829 top->m_flags |= M_EOR;
830 break;
831 }
832 } while (space > 0 && atomic);
833 if (dontroute)
834 so->so_options |= SO_DONTROUTE;
6b6e0885
JH
835 if (flags & MSG_OOB) {
836 pru_flags = PRUS_OOB;
837 } else if ((flags & MSG_EOF) &&
838 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
e54488bb 839 (resid == 0)) {
6b6e0885
JH
840 /*
841 * If the user set MSG_EOF, the protocol
842 * understands this flag and nothing left to
843 * send then use PRU_SEND_EOF instead of PRU_SEND.
844 */
845 pru_flags = PRUS_EOF;
846 } else if (resid > 0 && space > 0) {
847 /* If there is more to send, set PRUS_MORETOCOME */
848 pru_flags = PRUS_MORETOCOME;
849 } else {
850 pru_flags = 0;
851 }
984263bc
MD
852 /*
853 * XXX all the SS_CANTSENDMORE checks previously
854 * done could be out of date. We could have recieved
855 * a reset packet in an interrupt or maybe we slept
856 * while doing page faults in uiomove() etc. We could
857 * probably recheck again inside the splnet() protection
858 * here, but there are probably other places that this
859 * also happens. We must rethink this.
860 */
6b6e0885 861 error = so_pru_send(so, pru_flags, top, addr, control, td);
984263bc
MD
862 if (dontroute)
863 so->so_options &= ~SO_DONTROUTE;
864 clen = 0;
4090d6ff 865 control = NULL;
e28d8186 866 top = NULL;
984263bc
MD
867 mp = &top;
868 if (error)
6b6e0885 869 goto release;
984263bc
MD
870 } while (resid && space > 0);
871 } while (resid);
872
873release:
6d49aa6f 874 ssb_unlock(&so->so_snd);
984263bc
MD
875out:
876 if (top)
877 m_freem(top);
878 if (control)
879 m_freem(control);
880 return (error);
881}
882
883/*
6ea1e9b9
JH
884 * A specialization of sosend() for UDP based on protocol-specific knowledge:
885 * so->so_proto->pr_flags has the PR_ATOMIC field set. This means that
886 * sosendallatonce() returns true,
887 * the "atomic" variable is true,
888 * and sosendudp() blocks until space is available for the entire send.
889 * so->so_proto->pr_flags does not have the PR_CONNREQUIRED or
890 * PR_IMPLOPCL flags set.
891 * UDP has no out-of-band data.
892 * UDP has no control data.
893 * UDP does not support MSG_EOR.
894 */
895int
896sosendudp(struct socket *so, struct sockaddr *addr, struct uio *uio,
897 struct mbuf *top, struct mbuf *control, int flags, struct thread *td)
898{
e54488bb 899 size_t resid;
47d61296 900 int error, pru_flags = 0;
e54488bb 901 int space;
6ea1e9b9 902
fde7ac71
SS
903 if (td->td_lwp != NULL)
904 td->td_lwp->lwp_ru.ru_msgsnd++;
6ea1e9b9
JH
905 if (control)
906 m_freem(control);
907
908 KASSERT((uio && !top) || (top && !uio), ("bad arguments to sosendudp"));
e54488bb 909 resid = uio ? uio->uio_resid : (size_t)top->m_pkthdr.len;
6ea1e9b9
JH
910
911restart:
6d49aa6f 912 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags));
6ea1e9b9
JH
913 if (error)
914 goto out;
915
6ea1e9b9
JH
916 if (so->so_state & SS_CANTSENDMORE)
917 gotoerr(EPIPE);
918 if (so->so_error) {
919 error = so->so_error;
920 so->so_error = 0;
6ea1e9b9
JH
921 goto release;
922 }
923 if (!(so->so_state & SS_ISCONNECTED) && addr == NULL)
924 gotoerr(EDESTADDRREQ);
6d49aa6f 925 if (resid > so->so_snd.ssb_hiwat)
6ea1e9b9 926 gotoerr(EMSGSIZE);
e54488bb
MD
927 space = ssb_space(&so->so_snd);
928 if (uio && (space < 0 || (size_t)space < resid)) {
9ba76b73 929 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT))
6ea1e9b9 930 gotoerr(EWOULDBLOCK);
6d49aa6f
MD
931 ssb_unlock(&so->so_snd);
932 error = ssb_wait(&so->so_snd);
6ea1e9b9
JH
933 if (error)
934 goto out;
935 goto restart;
936 }
6ea1e9b9
JH
937
938 if (uio) {
e12241e1 939 top = m_uiomove(uio);
6ea1e9b9
JH
940 if (top == NULL)
941 goto release;
942 }
943
47d61296
SZ
944 if (flags & MSG_DONTROUTE)
945 pru_flags |= PRUS_DONTROUTE;
6ea1e9b9 946
0df7608b 947 if (udp_sosend_async && (flags & MSG_SYNC) == 0) {
6ce03112
SZ
948 so_pru_send_async(so, pru_flags, top, addr, NULL, td);
949 error = 0;
950 } else {
951 error = so_pru_send(so, pru_flags, top, addr, NULL, td);
952 }
6ea1e9b9
JH
953 top = NULL; /* sent or freed in lower layer */
954
6ea1e9b9 955release:
6d49aa6f 956 ssb_unlock(&so->so_snd);
6ea1e9b9
JH
957out:
958 if (top)
959 m_freem(top);
960 return (error);
961}
962
5bc42dd1
SZ
963int
964sosendtcp(struct socket *so, struct sockaddr *addr, struct uio *uio,
965 struct mbuf *top, struct mbuf *control, int flags,
966 struct thread *td)
967{
968 struct mbuf **mp;
969 struct mbuf *m;
970 size_t resid;
971 int space, len;
972 int error, mlen;
973 int allatonce;
974 int pru_flags;
975
976 if (uio) {
977 KKASSERT(top == NULL);
978 allatonce = 0;
979 resid = uio->uio_resid;
980 } else {
981 allatonce = 1;
982 resid = (size_t)top->m_pkthdr.len;
983#ifdef INVARIANTS
984 len = 0;
985 for (m = top; m; m = m->m_next)
986 len += m->m_len;
987 KKASSERT(top->m_pkthdr.len == len);
988#endif
989 }
990
991 /*
992 * WARNING! resid is unsigned, space and len are signed. space
993 * can wind up negative if the sockbuf is overcommitted.
994 *
995 * Also check to make sure that MSG_EOR isn't used on TCP
996 */
997 if (flags & MSG_EOR) {
998 error = EINVAL;
999 goto out;
1000 }
1001
1002 if (control) {
1003 /* TCP doesn't do control messages (rights, creds, etc) */
1004 if (control->m_len) {
1005 error = EINVAL;
1006 goto out;
1007 }
1008 m_freem(control); /* empty control, just free it */
1009 control = NULL;
1010 }
1011
1012 if (td->td_lwp != NULL)
1013 td->td_lwp->lwp_ru.ru_msgsnd++;
1014
1015#define gotoerr(errcode) { error = errcode; goto release; }
1016
1017restart:
1018 error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags));
1019 if (error)
1020 goto out;
1021
1022 do {
1023 if (so->so_state & SS_CANTSENDMORE)
1024 gotoerr(EPIPE);
1025 if (so->so_error) {
1026 error = so->so_error;
1027 so->so_error = 0;
1028 goto release;
1029 }
1030 if ((so->so_state & SS_ISCONNECTED) == 0 &&
1031 (so->so_state & SS_ISCONFIRMING) == 0)
1032 gotoerr(ENOTCONN);
1033 if (allatonce && resid > so->so_snd.ssb_hiwat)
1034 gotoerr(EMSGSIZE);
1035
f786a6ac 1036 space = ssb_space_prealloc(&so->so_snd);
5bc42dd1
SZ
1037 if (flags & MSG_OOB)
1038 space += 1024;
1039 if ((space < 0 || (size_t)space < resid) && !allatonce &&
1040 space < so->so_snd.ssb_lowat) {
1041 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT))
1042 gotoerr(EWOULDBLOCK);
1043 ssb_unlock(&so->so_snd);
1044 error = ssb_wait(&so->so_snd);
1045 if (error)
1046 goto out;
1047 goto restart;
1048 }
1049 mp = &top;
1050 do {
f2a3782e 1051 int cnt = 0, async = 0;
c16aca65 1052
5bc42dd1
SZ
1053 if (uio == NULL) {
1054 /*
1055 * Data is prepackaged in "top".
1056 */
1057 resid = 0;
1058 } else do {
1059 if (resid > INT_MAX)
1060 resid = INT_MAX;
1061 m = m_getl((int)resid, MB_WAIT, MT_DATA,
1062 top == NULL ? M_PKTHDR : 0, &mlen);
1063 if (top == NULL) {
1064 m->m_pkthdr.len = 0;
1065 m->m_pkthdr.rcvif = NULL;
1066 }
1067 len = imin((int)szmin(mlen, resid), space);
1068 space -= len;
1069 error = uiomove(mtod(m, caddr_t), (size_t)len, uio);
1070 resid = uio->uio_resid;
1071 m->m_len = len;
1072 *mp = m;
1073 top->m_pkthdr.len += len;
1074 if (error)
1075 goto release;
1076 mp = &m->m_next;
1077 if (resid == 0)
1078 break;
c16aca65 1079 ++cnt;
0df7608b 1080 } while (space > 0 && cnt < tcp_sosend_agglim);
5bc42dd1 1081
0df7608b 1082 if (tcp_sosend_async)
6d618102
SZ
1083 async = 1;
1084
5bc42dd1
SZ
1085 if (flags & MSG_OOB) {
1086 pru_flags = PRUS_OOB;
6d618102
SZ
1087 async = 0;
1088 } else if ((flags & MSG_EOF) && resid == 0) {
1089 pru_flags = PRUS_EOF;
5bc42dd1
SZ
1090 } else if (resid > 0 && space > 0) {
1091 /* If there is more to send, set PRUS_MORETOCOME */
1092 pru_flags = PRUS_MORETOCOME;
f2a3782e 1093 async = 1;
5bc42dd1
SZ
1094 } else {
1095 pru_flags = 0;
1096 }
1097
1af30d61 1098 if (flags & MSG_SYNC)
6d618102 1099 async = 0;
1af30d61 1100
5bc42dd1
SZ
1101 /*
1102 * XXX all the SS_CANTSENDMORE checks previously
1103 * done could be out of date. We could have recieved
1104 * a reset packet in an interrupt or maybe we slept
1105 * while doing page faults in uiomove() etc. We could
1106 * probably recheck again inside the splnet() protection
1107 * here, but there are probably other places that this
1108 * also happens. We must rethink this.
1109 */
f786a6ac
SZ
1110 for (m = top; m; m = m->m_next)
1111 ssb_preallocstream(&so->so_snd, m);
f2a3782e 1112 if (!async) {
0ad8e15e
SZ
1113 error = so_pru_send(so, pru_flags, top,
1114 NULL, NULL, td);
1115 } else {
1116 so_pru_send_async(so, pru_flags, top,
1117 NULL, NULL, td);
1118 error = 0;
1119 }
5bc42dd1
SZ
1120
1121 top = NULL;
1122 mp = &top;
1123 if (error)
1124 goto release;
1125 } while (resid && space > 0);
1126 } while (resid);
1127
1128release:
1129 ssb_unlock(&so->so_snd);
1130out:
1131 if (top)
1132 m_freem(top);
1133 if (control)
1134 m_freem(control);
1135 return (error);
1136}
1137
6ea1e9b9 1138/*
984263bc 1139 * Implement receive operations on a socket.
6cef7136 1140 *
6d49aa6f 1141 * We depend on the way that records are added to the signalsockbuf
984263bc
MD
1142 * by sbappend*. In particular, each record (mbufs linked through m_next)
1143 * must begin with an address if the protocol so specifies,
1144 * followed by an optional mbuf or mbufs containing ancillary data,
1145 * and then zero or more mbufs of data.
6cef7136
MD
1146 *
1147 * Although the signalsockbuf is locked, new data may still be appended.
1148 * A token inside the ssb_lock deals with MP issues and still allows
1149 * the network to access the socket if we block in a uio.
984263bc
MD
1150 *
1151 * The caller may receive the data as a single mbuf chain by supplying
1152 * an mbuf **mp0 for use in returning the chain. The uio is then used
1153 * only for the count in uio_resid.
1154 */
1155int
c972a82f 1156soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio,
6d49aa6f 1157 struct sockbuf *sio, struct mbuf **controlp, int *flagsp)
984263bc 1158{
d8a9a23b 1159 struct mbuf *m, *n;
857caa4a 1160 struct mbuf *free_chain = NULL;
e43a034f 1161 int flags, len, error, offset;
984263bc 1162 struct protosw *pr = so->so_proto;
984263bc 1163 int moff, type = 0;
e54488bb 1164 size_t resid, orig_resid;
d8a9a23b
MD
1165
1166 if (uio)
1167 resid = uio->uio_resid;
1168 else
e54488bb 1169 resid = (size_t)(sio->sb_climit - sio->sb_cc);
d8a9a23b 1170 orig_resid = resid;
984263bc 1171
984263bc 1172 if (psa)
857caa4a 1173 *psa = NULL;
984263bc 1174 if (controlp)
857caa4a 1175 *controlp = NULL;
984263bc
MD
1176 if (flagsp)
1177 flags = *flagsp &~ MSG_EOR;
1178 else
1179 flags = 0;
1180 if (flags & MSG_OOB) {
74f1caca 1181 m = m_get(MB_WAIT, MT_DATA);
984263bc
MD
1182 if (m == NULL)
1183 return (ENOBUFS);
6b6e0885 1184 error = so_pru_rcvoob(so, m, flags & MSG_PEEK);
984263bc
MD
1185 if (error)
1186 goto bad;
d8a9a23b 1187 if (sio) {
9510c423
SZ
1188 do {
1189 sbappend(sio, m);
1190 KKASSERT(resid >= (size_t)m->m_len);
1191 resid -= (size_t)m->m_len;
1192 } while (resid > 0 && m);
1193 } else {
1194 do {
1195 uio->uio_resid = resid;
1196 error = uiomove(mtod(m, caddr_t),
1197 (int)szmin(resid, m->m_len),
1198 uio);
1199 resid = uio->uio_resid;
1200 m = m_free(m);
1201 } while (uio->uio_resid && error == 0 && m);
1202 }
1203bad:
1204 if (m)
1205 m_freem(m);
1206 return (error);
1207 }
1208 if ((so->so_state & SS_ISCONFIRMING) && resid)
1209 so_pru_rcvd(so, 0);
1210
1211 /*
1212 * The token interlocks against the protocol thread while
1213 * ssb_lock is a blocking lock against other userland entities.
1214 */
1215 lwkt_gettoken(&so->so_rcv.ssb_token);
1216restart:
1217 error = ssb_lock(&so->so_rcv, SBLOCKWAIT(flags));
1218 if (error)
1219 goto done;
1220
1221 m = so->so_rcv.ssb_mb;
1222 /*
1223 * If we have less data than requested, block awaiting more
1224 * (subject to any timeout) if:
1225 * 1. the current count is less than the low water mark, or
1226 * 2. MSG_WAITALL is set, and it is possible to do the entire
1227 * receive operation at once if we block (resid <= hiwat).
1228 * 3. MSG_DONTWAIT is not set
1229 * If MSG_WAITALL is set but resid is larger than the receive buffer,
1230 * we have to do the receive in sections, and thus risk returning
1231 * a short count if a timeout or signal occurs after we start.
1232 */
1233 if (m == NULL || (((flags & MSG_DONTWAIT) == 0 &&
1234 (size_t)so->so_rcv.ssb_cc < resid) &&
1235 (so->so_rcv.ssb_cc < so->so_rcv.ssb_lowat ||
1236 ((flags & MSG_WAITALL) && resid <= (size_t)so->so_rcv.ssb_hiwat)) &&
1237 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
1238 KASSERT(m != NULL || !so->so_rcv.ssb_cc, ("receive 1"));
1239 if (so->so_error) {
1240 if (m)
1241 goto dontblock;
1242 error = so->so_error;
1243 if ((flags & MSG_PEEK) == 0)
1244 so->so_error = 0;
1245 goto release;
1246 }
1247 if (so->so_state & SS_CANTRCVMORE) {
1248 if (m)
1249 goto dontblock;
1250 else
1251 goto release;
1252 }
1253 for (; m; m = m->m_next) {
1254 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
1255 m = so->so_rcv.ssb_mb;
1256 goto dontblock;
1257 }
1258 }
1259 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
1260 (pr->pr_flags & PR_CONNREQUIRED)) {
1261 error = ENOTCONN;
1262 goto release;
1263 }
1264 if (resid == 0)
1265 goto release;
1266 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) {
1267 error = EWOULDBLOCK;
1268 goto release;
1269 }
1270 ssb_unlock(&so->so_rcv);
1271 error = ssb_wait(&so->so_rcv);
1272 if (error)
1273 goto done;
1274 goto restart;
1275 }
1276dontblock:
1277 if (uio && uio->uio_td && uio->uio_td->td_proc)
1278 uio->uio_td->td_lwp->lwp_ru.ru_msgrcv++;
1279
1280 /*
1281 * note: m should be == sb_mb here. Cache the next record while
1282 * cleaning up. Note that calling m_free*() will break out critical
1283 * section.
1284 */
1285 KKASSERT(m == so->so_rcv.ssb_mb);
1286
1287 /*
1288 * Skip any address mbufs prepending the record.
1289 */
1290 if (pr->pr_flags & PR_ADDR) {
1291 KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
1292 orig_resid = 0;
1293 if (psa)
1294 *psa = dup_sockaddr(mtod(m, struct sockaddr *));
1295 if (flags & MSG_PEEK)
1296 m = m->m_next;
1297 else
1298 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
1299 }
1300
1301 /*
1302 * Skip any control mbufs prepending the record.
1303 */
1304#ifdef SCTP
1305 if (pr->pr_flags & PR_ADDR_OPT) {
1306 /*
1307 * For SCTP we may be getting a
1308 * whole message OR a partial delivery.
1309 */
1310 if (m && m->m_type == MT_SONAME) {
1311 orig_resid = 0;
1312 if (psa)
1313 *psa = dup_sockaddr(mtod(m, struct sockaddr *));
1314 if (flags & MSG_PEEK)
1315 m = m->m_next;
1316 else
1317 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
1318 }
1319 }
1320#endif /* SCTP */
1321 while (m && m->m_type == MT_CONTROL && error == 0) {
1322 if (flags & MSG_PEEK) {
1323 if (controlp)
1324 *controlp = m_copy(m, 0, m->m_len);
1325 m = m->m_next; /* XXX race */
1326 } else {
1327 if (controlp) {
1328 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL);
1329 if (pr->pr_domain->dom_externalize &&
1330 mtod(m, struct cmsghdr *)->cmsg_type ==
1331 SCM_RIGHTS)
1332 error = (*pr->pr_domain->dom_externalize)(m);
1333 *controlp = m;
1334 m = n;
1335 } else {
1336 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
1337 }
1338 }
1339 if (controlp && *controlp) {
1340 orig_resid = 0;
1341 controlp = &(*controlp)->m_next;
1342 }
1343 }
1344
1345 /*
1346 * flag OOB data.
1347 */
1348 if (m) {
1349 type = m->m_type;
1350 if (type == MT_OOBDATA)
1351 flags |= MSG_OOB;
1352 }
1353
1354 /*
1355 * Copy to the UIO or mbuf return chain (*mp).
1356 */
1357 moff = 0;
1358 offset = 0;
1359 while (m && resid > 0 && error == 0) {
1360 if (m->m_type == MT_OOBDATA) {
1361 if (type != MT_OOBDATA)
1362 break;
1363 } else if (type == MT_OOBDATA)
1364 break;
1365 else
1366 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
1367 ("receive 3"));
1368 soclrstate(so, SS_RCVATMARK);
1369 len = (resid > INT_MAX) ? INT_MAX : resid;
1370 if (so->so_oobmark && len > so->so_oobmark - offset)
1371 len = so->so_oobmark - offset;
1372 if (len > m->m_len - moff)
1373 len = m->m_len - moff;
1374
1375 /*
1376 * Copy out to the UIO or pass the mbufs back to the SIO.
1377 * The SIO is dealt with when we eat the mbuf, but deal
1378 * with the resid here either way.
1379 */
1380 if (uio) {
1381 uio->uio_resid = resid;
1382 error = uiomove(mtod(m, caddr_t) + moff, len, uio);
1383 resid = uio->uio_resid;
1384 if (error)
1385 goto release;
1386 } else {
1387 resid -= (size_t)len;
1388 }
1389
1390 /*
1391 * Eat the entire mbuf or just a piece of it
1392 */
1393 if (len == m->m_len - moff) {
1394 if (m->m_flags & M_EOR)
1395 flags |= MSG_EOR;
1396#ifdef SCTP
1397 if (m->m_flags & M_NOTIFICATION)
1398 flags |= MSG_NOTIFICATION;
1399#endif /* SCTP */
1400 if (flags & MSG_PEEK) {
1401 m = m->m_next;
1402 moff = 0;
1403 } else {
1404 if (sio) {
1405 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL);
1406 sbappend(sio, m);
1407 m = n;
1408 } else {
1409 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
1410 }
1411 }
1412 } else {
1413 if (flags & MSG_PEEK) {
1414 moff += len;
1415 } else {
1416 if (sio) {
1417 n = m_copym(m, 0, len, MB_WAIT);
1418 if (n)
1419 sbappend(sio, n);
1420 }
1421 m->m_data += len;
1422 m->m_len -= len;
1423 so->so_rcv.ssb_cc -= len;
1424 }
1425 }
1426 if (so->so_oobmark) {
1427 if ((flags & MSG_PEEK) == 0) {
1428 so->so_oobmark -= len;
1429 if (so->so_oobmark == 0) {
1430 sosetstate(so, SS_RCVATMARK);
1431 break;
1432 }
1433 } else {
1434 offset += len;
1435 if (offset == so->so_oobmark)
1436 break;
1437 }
1438 }
1439 if (flags & MSG_EOR)
1440 break;
1441 /*
1442 * If the MSG_WAITALL flag is set (for non-atomic socket),
1443 * we must not quit until resid == 0 or an error
1444 * termination. If a signal/timeout occurs, return
1445 * with a short count but without error.
1446 * Keep signalsockbuf locked against other readers.
1447 */
1448 while ((flags & MSG_WAITALL) && m == NULL &&
1449 resid > 0 && !sosendallatonce(so) &&
1450 so->so_rcv.ssb_mb == NULL) {
1451 if (so->so_error || so->so_state & SS_CANTRCVMORE)
1452 break;
1453 /*
1454 * The window might have closed to zero, make
1455 * sure we send an ack now that we've drained
1456 * the buffer or we might end up blocking until
1457 * the idle takes over (5 seconds).
1458 */
1459 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1460 so_pru_rcvd(so, flags);
1461 error = ssb_wait(&so->so_rcv);
1462 if (error) {
1463 ssb_unlock(&so->so_rcv);
1464 error = 0;
1465 goto done;
1466 }
1467 m = so->so_rcv.ssb_mb;
1468 }
1469 }
1470
1471 /*
1472 * If an atomic read was requested but unread data still remains
1473 * in the record, set MSG_TRUNC.
1474 */
1475 if (m && pr->pr_flags & PR_ATOMIC)
1476 flags |= MSG_TRUNC;
1477
1478 /*
1479 * Cleanup. If an atomic read was requested drop any unread data.
1480 */
1481 if ((flags & MSG_PEEK) == 0) {
1482 if (m && (pr->pr_flags & PR_ATOMIC))
1483 sbdroprecord(&so->so_rcv.sb);
1484 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1485 so_pru_rcvd(so, flags);
1486 }
1487
1488 if (orig_resid == resid && orig_resid &&
1489 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1490 ssb_unlock(&so->so_rcv);
1491 goto restart;
1492 }
1493
1494 if (flagsp)
1495 *flagsp |= flags;
1496release:
1497 ssb_unlock(&so->so_rcv);
1498done:
1499 lwkt_reltoken(&so->so_rcv.ssb_token);
1500 if (free_chain)
1501 m_freem(free_chain);
1502 return (error);
1503}
1504
1505int
1506sorecvtcp(struct socket *so, struct sockaddr **psa, struct uio *uio,
1507 struct sockbuf *sio, struct mbuf **controlp, int *flagsp)
1508{
1509 struct mbuf *m, *n;
1510 struct mbuf *free_chain = NULL;
1511 int flags, len, error, offset;
1512 struct protosw *pr = so->so_proto;
1513 int moff, type = 0;
1514 size_t resid, orig_resid;
1515
1516 if (uio)
1517 resid = uio->uio_resid;
1518 else
1519 resid = (size_t)(sio->sb_climit - sio->sb_cc);
1520 orig_resid = resid;
1521
1522 if (psa)
1523 *psa = NULL;
1524 if (controlp)
1525 *controlp = NULL;
1526 if (flagsp)
1527 flags = *flagsp &~ MSG_EOR;
1528 else
1529 flags = 0;
1530 if (flags & MSG_OOB) {
1531 m = m_get(MB_WAIT, MT_DATA);
1532 if (m == NULL)
1533 return (ENOBUFS);
1534 error = so_pru_rcvoob(so, m, flags & MSG_PEEK);
1535 if (error)
1536 goto bad;
1537 if (sio) {
d8a9a23b 1538 do {
6d49aa6f 1539 sbappend(sio, m);
e54488bb
MD
1540 KKASSERT(resid >= (size_t)m->m_len);
1541 resid -= (size_t)m->m_len;
d8a9a23b
MD
1542 } while (resid > 0 && m);
1543 } else {
1544 do {
1545 uio->uio_resid = resid;
1546 error = uiomove(mtod(m, caddr_t),
e54488bb
MD
1547 (int)szmin(resid, m->m_len),
1548 uio);
d8a9a23b
MD
1549 resid = uio->uio_resid;
1550 m = m_free(m);
1551 } while (uio->uio_resid && error == 0 && m);
1552 }
984263bc
MD
1553bad:
1554 if (m)
1555 m_freem(m);
1556 return (error);
1557 }
e54488bb 1558 if ((so->so_state & SS_ISCONFIRMING) && resid)
6b6e0885 1559 so_pru_rcvd(so, 0);
984263bc 1560
20156c7a
MD
1561 /*
1562 * The token interlocks against the protocol thread while
1563 * ssb_lock is a blocking lock against other userland entities.
1564 */
1565 lwkt_gettoken(&so->so_rcv.ssb_token);
984263bc 1566restart:
6d49aa6f 1567 error = ssb_lock(&so->so_rcv, SBLOCKWAIT(flags));
984263bc 1568 if (error)
857caa4a 1569 goto done;
984263bc 1570
6d49aa6f 1571 m = so->so_rcv.ssb_mb;
984263bc
MD
1572 /*
1573 * If we have less data than requested, block awaiting more
1574 * (subject to any timeout) if:
1575 * 1. the current count is less than the low water mark, or
1576 * 2. MSG_WAITALL is set, and it is possible to do the entire
1577 * receive operation at once if we block (resid <= hiwat).
1578 * 3. MSG_DONTWAIT is not set
1579 * If MSG_WAITALL is set but resid is larger than the receive buffer,
1580 * we have to do the receive in sections, and thus risk returning
1581 * a short count if a timeout or signal occurs after we start.
1582 */
857caa4a 1583 if (m == NULL || (((flags & MSG_DONTWAIT) == 0 &&
e54488bb 1584 (size_t)so->so_rcv.ssb_cc < resid) &&
6d49aa6f 1585 (so->so_rcv.ssb_cc < so->so_rcv.ssb_lowat ||
e54488bb 1586 ((flags & MSG_WAITALL) && resid <= (size_t)so->so_rcv.ssb_hiwat)) &&
984263bc 1587 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
6d49aa6f 1588 KASSERT(m != NULL || !so->so_rcv.ssb_cc, ("receive 1"));
984263bc
MD
1589 if (so->so_error) {
1590 if (m)
1591 goto dontblock;
1592 error = so->so_error;
1593 if ((flags & MSG_PEEK) == 0)
1594 so->so_error = 0;
1595 goto release;
1596 }
1597 if (so->so_state & SS_CANTRCVMORE) {
1598 if (m)
1599 goto dontblock;
1600 else
1601 goto release;
1602 }
857caa4a 1603 for (; m; m = m->m_next) {
984263bc 1604 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
6d49aa6f 1605 m = so->so_rcv.ssb_mb;
984263bc
MD
1606 goto dontblock;
1607 }
857caa4a 1608 }
984263bc 1609 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
6b6e0885 1610 (pr->pr_flags & PR_CONNREQUIRED)) {
984263bc
MD
1611 error = ENOTCONN;
1612 goto release;
1613 }
d8a9a23b 1614 if (resid == 0)
984263bc 1615 goto release;
9ba76b73 1616 if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) {
984263bc
MD
1617 error = EWOULDBLOCK;
1618 goto release;
1619 }
6d49aa6f
MD
1620 ssb_unlock(&so->so_rcv);
1621 error = ssb_wait(&so->so_rcv);
984263bc 1622 if (error)
857caa4a 1623 goto done;
984263bc
MD
1624 goto restart;
1625 }
1626dontblock:
d8a9a23b 1627 if (uio && uio->uio_td && uio->uio_td->td_proc)
fde7ac71 1628 uio->uio_td->td_lwp->lwp_ru.ru_msgrcv++;
857caa4a
MD
1629
1630 /*
1631 * note: m should be == sb_mb here. Cache the next record while
1632 * cleaning up. Note that calling m_free*() will break out critical
1633 * section.
1634 */
6d49aa6f 1635 KKASSERT(m == so->so_rcv.ssb_mb);
857caa4a
MD
1636
1637 /*
1638 * Skip any address mbufs prepending the record.
1639 */
984263bc
MD
1640 if (pr->pr_flags & PR_ADDR) {
1641 KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
1642 orig_resid = 0;
1643 if (psa)
cfa2ba21 1644 *psa = dup_sockaddr(mtod(m, struct sockaddr *));
857caa4a 1645 if (flags & MSG_PEEK)
984263bc 1646 m = m->m_next;
857caa4a 1647 else
6d49aa6f 1648 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
984263bc 1649 }
857caa4a
MD
1650
1651 /*
1652 * Skip any control mbufs prepending the record.
1653 */
78812139
EN
1654#ifdef SCTP
1655 if (pr->pr_flags & PR_ADDR_OPT) {
1656 /*
1657 * For SCTP we may be getting a
1658 * whole message OR a partial delivery.
1659 */
857caa4a 1660 if (m && m->m_type == MT_SONAME) {
78812139
EN
1661 orig_resid = 0;
1662 if (psa)
1663 *psa = dup_sockaddr(mtod(m, struct sockaddr *));
857caa4a 1664 if (flags & MSG_PEEK)
78812139 1665 m = m->m_next;
857caa4a 1666 else
6d49aa6f 1667 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
78812139
EN
1668 }
1669 }
1670#endif /* SCTP */
984263bc
MD
1671 while (m && m->m_type == MT_CONTROL && error == 0) {
1672 if (flags & MSG_PEEK) {
1673 if (controlp)
1674 *controlp = m_copy(m, 0, m->m_len);
857caa4a 1675 m = m->m_next; /* XXX race */
984263bc 1676 } else {
984263bc 1677 if (controlp) {
6d49aa6f 1678 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL);
984263bc
MD
1679 if (pr->pr_domain->dom_externalize &&
1680 mtod(m, struct cmsghdr *)->cmsg_type ==
1681 SCM_RIGHTS)
1682 error = (*pr->pr_domain->dom_externalize)(m);
1683 *controlp = m;
857caa4a 1684 m = n;
984263bc 1685 } else {
6d49aa6f 1686 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
984263bc
MD
1687 }
1688 }
857caa4a 1689 if (controlp && *controlp) {
984263bc
MD
1690 orig_resid = 0;
1691 controlp = &(*controlp)->m_next;
1692 }
1693 }
857caa4a
MD
1694
1695 /*
1696 * flag OOB data.
1697 */
984263bc 1698 if (m) {
984263bc
MD
1699 type = m->m_type;
1700 if (type == MT_OOBDATA)
1701 flags |= MSG_OOB;
1702 }
857caa4a
MD
1703
1704 /*
1705 * Copy to the UIO or mbuf return chain (*mp).
1706 */
984263bc
MD
1707 moff = 0;
1708 offset = 0;
d8a9a23b 1709 while (m && resid > 0 && error == 0) {
984263bc
MD
1710 if (m->m_type == MT_OOBDATA) {
1711 if (type != MT_OOBDATA)
1712 break;
1713 } else if (type == MT_OOBDATA)
1714 break;
1715 else
1716 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
1717 ("receive 3"));
6cef7136 1718 soclrstate(so, SS_RCVATMARK);
e54488bb 1719 len = (resid > INT_MAX) ? INT_MAX : resid;
984263bc
MD
1720 if (so->so_oobmark && len > so->so_oobmark - offset)
1721 len = so->so_oobmark - offset;
1722 if (len > m->m_len - moff)
1723 len = m->m_len - moff;
d8a9a23b 1724
984263bc 1725 /*
d8a9a23b
MD
1726 * Copy out to the UIO or pass the mbufs back to the SIO.
1727 * The SIO is dealt with when we eat the mbuf, but deal
1728 * with the resid here either way.
984263bc 1729 */
d8a9a23b 1730 if (uio) {
d8a9a23b
MD
1731 uio->uio_resid = resid;
1732 error = uiomove(mtod(m, caddr_t) + moff, len, uio);
1733 resid = uio->uio_resid;
984263bc
MD
1734 if (error)
1735 goto release;
857caa4a 1736 } else {
e54488bb 1737 resid -= (size_t)len;
857caa4a
MD
1738 }
1739
1740 /*
1741 * Eat the entire mbuf or just a piece of it
1742 */
984263bc
MD
1743 if (len == m->m_len - moff) {
1744 if (m->m_flags & M_EOR)
1745 flags |= MSG_EOR;
78812139
EN
1746#ifdef SCTP
1747 if (m->m_flags & M_NOTIFICATION)
1748 flags |= MSG_NOTIFICATION;
1749#endif /* SCTP */
984263bc
MD
1750 if (flags & MSG_PEEK) {
1751 m = m->m_next;
1752 moff = 0;
1753 } else {
d8a9a23b 1754 if (sio) {
6d49aa6f
MD
1755 n = sbunlinkmbuf(&so->so_rcv.sb, m, NULL);
1756 sbappend(sio, m);
857caa4a 1757 m = n;
984263bc 1758 } else {
6d49aa6f 1759 m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
984263bc 1760 }
984263bc
MD
1761 }
1762 } else {
857caa4a 1763 if (flags & MSG_PEEK) {
984263bc 1764 moff += len;
857caa4a 1765 } else {
d8a9a23b 1766 if (sio) {
6d49aa6f
MD
1767 n = m_copym(m, 0, len, MB_WAIT);
1768 if (n)
1769 sbappend(sio, n);
d8a9a23b 1770 }
984263bc
MD
1771 m->m_data += len;
1772 m->m_len -= len;
6d49aa6f 1773 so->so_rcv.ssb_cc -= len;
984263bc
MD
1774 }
1775 }
1776 if (so->so_oobmark) {
1777 if ((flags & MSG_PEEK) == 0) {
1778 so->so_oobmark -= len;
1779 if (so->so_oobmark == 0) {
6cef7136 1780 sosetstate(so, SS_RCVATMARK);
984263bc
MD
1781 break;
1782 }
1783 } else {
1784 offset += len;
1785 if (offset == so->so_oobmark)
1786 break;
1787 }
1788 }
1789 if (flags & MSG_EOR)
1790 break;
1791 /*
1792 * If the MSG_WAITALL flag is set (for non-atomic socket),
d8a9a23b 1793 * we must not quit until resid == 0 or an error
984263bc
MD
1794 * termination. If a signal/timeout occurs, return
1795 * with a short count but without error.
6d49aa6f 1796 * Keep signalsockbuf locked against other readers.
984263bc 1797 */
d8a9a23b
MD
1798 while ((flags & MSG_WAITALL) && m == NULL &&
1799 resid > 0 && !sosendallatonce(so) &&
6d49aa6f 1800 so->so_rcv.ssb_mb == NULL) {
984263bc
MD
1801 if (so->so_error || so->so_state & SS_CANTRCVMORE)
1802 break;
1803 /*
1804 * The window might have closed to zero, make
1805 * sure we send an ack now that we've drained
1806 * the buffer or we might end up blocking until
1807 * the idle takes over (5 seconds).
1808 */
1809 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
6b6e0885 1810 so_pru_rcvd(so, flags);
6d49aa6f 1811 error = ssb_wait(&so->so_rcv);
984263bc 1812 if (error) {
6d49aa6f 1813 ssb_unlock(&so->so_rcv);
857caa4a
MD
1814 error = 0;
1815 goto done;
984263bc 1816 }
6d49aa6f 1817 m = so->so_rcv.ssb_mb;
984263bc
MD
1818 }
1819 }
1820
857caa4a
MD
1821 /*
1822 * If an atomic read was requested but unread data still remains
1823 * in the record, set MSG_TRUNC.
1824 */
bf8a9a6f 1825 if (m && pr->pr_flags & PR_ATOMIC)
984263bc 1826 flags |= MSG_TRUNC;
857caa4a
MD
1827
1828 /*
1829 * Cleanup. If an atomic read was requested drop any unread data.
1830 */
1831 if ((flags & MSG_PEEK) == 0) {
1832 if (m && (pr->pr_flags & PR_ATOMIC))
6d49aa6f 1833 sbdroprecord(&so->so_rcv.sb);
857caa4a 1834 if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
6b6e0885 1835 so_pru_rcvd(so, flags);
984263bc 1836 }
bf8a9a6f 1837
d8a9a23b 1838 if (orig_resid == resid && orig_resid &&
984263bc 1839 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
6d49aa6f 1840 ssb_unlock(&so->so_rcv);
984263bc
MD
1841 goto restart;
1842 }
1843
1844 if (flagsp)
1845 *flagsp |= flags;
1846release:
6d49aa6f 1847 ssb_unlock(&so->so_rcv);
857caa4a 1848done:
20156c7a 1849 lwkt_reltoken(&so->so_rcv.ssb_token);
857caa4a
MD
1850 if (free_chain)
1851 m_freem(free_chain);
984263bc
MD
1852 return (error);
1853}
1854
edf5c732
MD
1855/*
1856 * Shut a socket down. Note that we do not get a frontend lock as we
1857 * want to be able to shut the socket down even if another thread is
1858 * blocked in a read(), thus waking it up.
1859 */
984263bc 1860int
c972a82f 1861soshutdown(struct socket *so, int how)
984263bc 1862{
984263bc
MD
1863 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1864 return (EINVAL);
1865
ff518922 1866 if (how != SHUT_WR) {
edf5c732 1867 /*ssb_lock(&so->so_rcv, M_WAITOK);*/
984263bc 1868 sorflush(so);
edf5c732 1869 /*ssb_unlock(&so->so_rcv);*/
ff518922 1870 }
0cd6e642 1871 if (how != SHUT_RD)
6b6e0885 1872 return (so_pru_shutdown(so));
984263bc
MD
1873 return (0);
1874}
1875
1876void
c972a82f 1877sorflush(struct socket *so)
984263bc 1878{
6d49aa6f 1879 struct signalsockbuf *ssb = &so->so_rcv;
1fd87d54 1880 struct protosw *pr = so->so_proto;
6d49aa6f 1881 struct signalsockbuf asb;
984263bc 1882
14343ad3 1883 atomic_set_int(&ssb->ssb_flags, SSB_NOINTR);
e43a034f 1884
ff518922 1885 lwkt_gettoken(&ssb->ssb_token);
984263bc 1886 socantrcvmore(so);
6d49aa6f 1887 asb = *ssb;
14343ad3
MD
1888
1889 /*
1890 * Can't just blow up the ssb structure here
1891 */
ff518922 1892 bzero(&ssb->sb, sizeof(ssb->sb));
14343ad3 1893 ssb->ssb_timeo = 0;
14343ad3
MD
1894 ssb->ssb_lowat = 0;
1895 ssb->ssb_hiwat = 0;
1896 ssb->ssb_mbmax = 0;
1897 atomic_clear_int(&ssb->ssb_flags, SSB_CLEAR_MASK);
1898
edf5c732 1899 if ((pr->pr_flags & PR_RIGHTS) && pr->pr_domain->dom_dispose)
6d49aa6f
MD
1900 (*pr->pr_domain->dom_dispose)(asb.ssb_mb);
1901 ssb_release(&asb, so);
edf5c732
MD
1902
1903 lwkt_reltoken(&ssb->ssb_token);
984263bc
MD
1904}
1905
1906#ifdef INET
1907static int
c972a82f 1908do_setopt_accept_filter(struct socket *so, struct sockopt *sopt)
984263bc
MD
1909{
1910 struct accept_filter_arg *afap = NULL;
1911 struct accept_filter *afp;
1912 struct so_accf *af = so->so_accf;
1913 int error = 0;
1914
1915 /* do not set/remove accept filters on non listen sockets */
1916 if ((so->so_options & SO_ACCEPTCONN) == 0) {
1917 error = EINVAL;
1918 goto out;
1919 }
1920
1921 /* removing the filter */
1922 if (sopt == NULL) {
1923 if (af != NULL) {
1924 if (af->so_accept_filter != NULL &&
1925 af->so_accept_filter->accf_destroy != NULL) {
1926 af->so_accept_filter->accf_destroy(so);
1927 }
1928 if (af->so_accept_filter_str != NULL) {
884717e1 1929 kfree(af->so_accept_filter_str, M_ACCF);
984263bc 1930 }
884717e1 1931 kfree(af, M_ACCF);
984263bc
MD
1932 so->so_accf = NULL;
1933 }
1934 so->so_options &= ~SO_ACCEPTFILTER;
1935 return (0);
1936 }
1937 /* adding a filter */
1938 /* must remove previous filter first */
1939 if (af != NULL) {
1940 error = EINVAL;
1941 goto out;
1942 }
1943 /* don't put large objects on the kernel stack */
884717e1 1944 afap = kmalloc(sizeof(*afap), M_TEMP, M_WAITOK);
984263bc
MD
1945 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1946 afap->af_name[sizeof(afap->af_name)-1] = '\0';
1947 afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1948 if (error)
1949 goto out;
1950 afp = accept_filt_get(afap->af_name);
1951 if (afp == NULL) {
1952 error = ENOENT;
1953 goto out;
1954 }
884717e1 1955 af = kmalloc(sizeof(*af), M_ACCF, M_WAITOK | M_ZERO);
984263bc
MD
1956 if (afp->accf_create != NULL) {
1957 if (afap->af_name[0] != '\0') {
1958 int len = strlen(afap->af_name) + 1;
1959
884717e1
SW
1960 af->so_accept_filter_str = kmalloc(len, M_ACCF,
1961 M_WAITOK);
984263bc
MD
1962 strcpy(af->so_accept_filter_str, afap->af_name);
1963 }
1964 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1965 if (af->so_accept_filter_arg == NULL) {
884717e1
SW
1966 kfree(af->so_accept_filter_str, M_ACCF);
1967 kfree(af, M_ACCF);
984263bc
MD
1968 so->so_accf = NULL;
1969 error = EINVAL;
1970 goto out;
1971 }
1972 }
1973 af->so_accept_filter = afp;
1974 so->so_accf = af;
1975 so->so_options |= SO_ACCEPTFILTER;
1976out:
1977 if (afap != NULL)
884717e1 1978 kfree(afap, M_TEMP);
984263bc
MD
1979 return (error);
1980}
1981#endif /* INET */
1982
1983/*
1984 * Perhaps this routine, and sooptcopyout(), below, ought to come in
1985 * an additional variant to handle the case where the option value needs
1986 * to be some kind of integer, but not a specific size.
1987 * In addition to their use here, these functions are also called by the
1988 * protocol-level pr_ctloutput() routines.
1989 */
1990int
c972a82f 1991sooptcopyin(struct sockopt *sopt, void *buf, size_t len, size_t minlen)
984263bc 1992{
de0003fe
AE
1993 return soopt_to_kbuf(sopt, buf, len, minlen);
1994}
1995
1996int
1997soopt_to_kbuf(struct sockopt *sopt, void *buf, size_t len, size_t minlen)
1998{
984263bc
MD
1999 size_t valsize;
2000
792239df 2001 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val));
de0003fe
AE
2002 KKASSERT(kva_p(buf));
2003
984263bc
MD
2004 /*
2005 * If the user gives us more than we wanted, we ignore it,
2006 * but if we don't get the minimum length the caller
2007 * wants, we return EINVAL. On success, sopt->sopt_valsize
2008 * is set to however much we actually retrieved.
2009 */
2010 if ((valsize = sopt->sopt_valsize) < minlen)
2011 return EINVAL;
2012 if (valsize > len)
2013 sopt->sopt_valsize = valsize = len;
2014
984263bc
MD
2015 bcopy(sopt->sopt_val, buf, valsize);
2016 return 0;
2017}
2018
e71a125f
AE
2019
2020int
c972a82f 2021sosetopt(struct socket *so, struct sockopt *sopt)
984263bc
MD
2022{
2023 int error, optval;
2024 struct linger l;
2025 struct timeval tv;
2026 u_long val;
14343ad3 2027 struct signalsockbuf *sotmp;
984263bc
MD
2028
2029 error = 0;
e79d388f 2030 sopt->sopt_dir = SOPT_SET;
984263bc 2031 if (sopt->sopt_level != SOL_SOCKET) {
6b6e0885 2032 if (so->so_proto && so->so_proto->pr_ctloutput) {
002c1265 2033 return (so_pr_ctloutput(so, sopt));
6b6e0885 2034 }
984263bc
MD
2035 error = ENOPROTOOPT;
2036 } else {
2037 switch (sopt->sopt_name) {
2038#ifdef INET
2039 case SO_ACCEPTFILTER:
2040 error = do_setopt_accept_filter(so, sopt);
2041 if (error)
2042 goto bad;
2043 break;
2044#endif /* INET */
2045 case SO_LINGER:
2046 error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
2047 if (error)
2048 goto bad;
2049
2050 so->so_linger = l.l_linger;
2051 if (l.l_onoff)
2052 so->so_options |= SO_LINGER;
2053 else
2054 so->so_options &= ~SO_LINGER;
2055 break;
2056
2057 case SO_DEBUG:
2058 case SO_KEEPALIVE:
2059 case SO_DONTROUTE:
2060 case SO_USELOOPBACK:
2061 case SO_BROADCAST:
2062 case SO_REUSEADDR:
2063 case SO_REUSEPORT:
2064 case SO_OOBINLINE:
2065 case SO_TIMESTAMP:
2066 error = sooptcopyin(sopt, &optval, sizeof optval,
2067 sizeof optval);
2068 if (error)
2069 goto bad;
2070 if (optval)
2071 so->so_options |= sopt->sopt_name;
2072 else
2073 so->so_options &= ~sopt->sopt_name;
2074 break;
2075
2076 case SO_SNDBUF:
2077 case SO_RCVBUF:
2078 case SO_SNDLOWAT:
2079 case SO_RCVLOWAT:
2080 error = sooptcopyin(sopt, &optval, sizeof optval,
2081 sizeof optval);
2082 if (error)
2083 goto bad;
2084
2085 /*
2086 * Values < 1 make no sense for any of these
2087 * options, so disallow them.
2088 */
2089 if (optval < 1) {
2090 error = EINVAL;
2091 goto bad;
2092 }
2093
2094 switch (sopt->sopt_name) {
2095 case SO_SNDBUF:
2096 case SO_RCVBUF:
6d49aa6f 2097 if (ssb_reserve(sopt->sopt_name == SO_SNDBUF ?
984263bc 2098 &so->so_snd : &so->so_rcv, (u_long)optval,
e4700d00
JH
2099 so,
2100 &curproc->p_rlimit[RLIMIT_SBSIZE]) == 0) {
984263bc
MD
2101 error = ENOBUFS;
2102 goto bad;
2103 }
14343ad3
MD
2104 sotmp = (sopt->sopt_name == SO_SNDBUF) ?
2105 &so->so_snd : &so->so_rcv;
2106 atomic_clear_int(&sotmp->ssb_flags,
2107 SSB_AUTOSIZE);
984263bc
MD
2108 break;
2109
2110 /*
2111 * Make sure the low-water is never greater than
2112 * the high-water.
2113 */
2114 case SO_SNDLOWAT:
6d49aa6f
MD
2115 so->so_snd.ssb_lowat =
2116 (optval > so->so_snd.ssb_hiwat) ?
2117 so->so_snd.ssb_hiwat : optval;
14343ad3
MD
2118 atomic_clear_int(&so->so_snd.ssb_flags,
2119 SSB_AUTOLOWAT);
984263bc
MD
2120 break;
2121 case SO_RCVLOWAT:
6d49aa6f
MD
2122 so->so_rcv.ssb_lowat =
2123 (optval > so->so_rcv.ssb_hiwat) ?
2124 so->so_rcv.ssb_hiwat : optval;
14343ad3
MD
2125 atomic_clear_int(&so->so_rcv.ssb_flags,
2126 SSB_AUTOLOWAT);
984263bc
MD
2127 break;
2128 }
2129 break;
2130
2131 case SO_SNDTIMEO:
2132 case SO_RCVTIMEO:
2133 error = sooptcopyin(sopt, &tv, sizeof tv,
2134 sizeof tv);
2135 if (error)
2136 goto bad;
2137
2138 /* assert(hz > 0); */
45546849 2139 if (tv.tv_sec < 0 || tv.tv_sec > INT_MAX / hz ||
984263bc
MD
2140 tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
2141 error = EDOM;
2142 goto bad;
2143 }
2144 /* assert(tick > 0); */
45546849 2145 /* assert(ULONG_MAX - INT_MAX >= 1000000); */
a591f597 2146 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / ustick;
45546849 2147 if (val > INT_MAX) {
984263bc
MD
2148 error = EDOM;
2149 goto bad;
2150 }
2151 if (val == 0 && tv.tv_usec != 0)
2152 val = 1;
2153
2154 switch (sopt->sopt_name) {
2155 case SO_SNDTIMEO:
6d49aa6f 2156 so->so_snd.ssb_timeo = val;
984263bc
MD
2157 break;
2158 case SO_RCVTIMEO:
6d49aa6f 2159 so->so_rcv.ssb_timeo = val;
984263bc
MD
2160 break;
2161 }
2162 break;
2163 default:
2164 error = ENOPROTOOPT;
2165 break;
2166 }
2167 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
002c1265 2168 (void) so_pr_ctloutput(so, sopt);
984263bc
MD
2169 }
2170 }
2171bad:
2172 return (error);
2173}
2174
2175/* Helper routine for getsockopt */
2176int
f1f552f6 2177sooptcopyout(struct sockopt *sopt, const void *buf, size_t len)
984263bc 2178{
de0003fe
AE
2179 soopt_from_kbuf(sopt, buf, len);
2180 return 0;
2181}
2182
2183void
2184soopt_from_kbuf(struct sockopt *sopt, const void *buf, size_t len)
2185{
984263bc
MD
2186 size_t valsize;
2187
565d9f6f
SZ
2188 if (len == 0) {
2189 sopt->sopt_valsize = 0;
2190 return;
2191 }
2192
792239df 2193 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val));
de0003fe 2194 KKASSERT(kva_p(buf));
984263bc
MD
2195
2196 /*
2197 * Documented get behavior is that we always return a value,
2198 * possibly truncated to fit in the user's buffer.
2199 * Traditional behavior is that we always tell the user
2200 * precisely how much we copied, rather than something useful
2201 * like the total amount we had available for her.
2202 * Note that this interface is not idempotent; the entire answer must
2203 * generated ahead of time.
2204 */
231d276b 2205 valsize = szmin(len, sopt->sopt_valsize);
984263bc
MD
2206 sopt->sopt_valsize = valsize;
2207 if (sopt->sopt_val != 0) {
de0003fe 2208 bcopy(buf, sopt->sopt_val, valsize);
984263bc 2209 }
e71a125f
AE
2210}
2211
984263bc 2212int
c972a82f 2213sogetopt(struct socket *so, struct sockopt *sopt)
984263bc
MD
2214{
2215 int error, optval;
755c519c 2216 long optval_l;
984263bc
MD
2217 struct linger l;
2218 struct timeval tv;
51f4ca92 2219#ifdef INET
984263bc 2220 struct accept_filter_arg *afap;
51f4ca92 2221#endif
984263bc
MD
2222
2223 error = 0;
e79d388f 2224 sopt->sopt_dir = SOPT_GET;
984263bc
MD
2225 if (sopt->sopt_level != SOL_SOCKET) {
2226 if (so->so_proto && so->so_proto->pr_ctloutput) {
002c1265 2227 return (so_pr_ctloutput(so, sopt));
984263bc
MD
2228 } else
2229 return (ENOPROTOOPT);
2230 } else {
2231 switch (sopt->sopt_name) {
2232#ifdef INET
2233 case SO_ACCEPTFILTER:
2234 if ((so->so_options & SO_ACCEPTCONN) == 0)
2235 return (EINVAL);
884717e1
SW
2236 afap = kmalloc(sizeof(*afap), M_TEMP,
2237 M_WAITOK | M_ZERO);
984263bc
MD
2238 if ((so->so_options & SO_ACCEPTFILTER) != 0) {
2239 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
2240 if (so->so_accf->so_accept_filter_str != NULL)
2241 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
2242 }
2243 error = sooptcopyout(sopt, afap, sizeof(*afap));
884717e1 2244 kfree(afap, M_TEMP);
984263bc
MD
2245 break;
2246#endif /* INET */
2247
2248 case SO_LINGER:
2249 l.l_onoff = so->so_options & SO_LINGER;
2250 l.l_linger = so->so_linger;
2251 error = sooptcopyout(sopt, &l, sizeof l);
2252 break;
2253
2254 case SO_USELOOPBACK:
2255 case SO_DONTROUTE:
2256 case SO_DEBUG:
2257 case SO_KEEPALIVE:
2258 case SO_REUSEADDR:
2259 case SO_REUSEPORT:
2260 case SO_BROADCAST:
2261 case SO_OOBINLINE:
2262 case SO_TIMESTAMP:
2263 optval = so->so_options & sopt->sopt_name;
2264integer:
2265 error = sooptcopyout(sopt, &optval, sizeof optval);
2266 break;
2267
2268 case SO_TYPE:
2269 optval = so->so_type;
2270 goto integer;
2271
2272 case SO_ERROR:
2273 optval = so->so_error;
2274 so->so_error = 0;
2275 goto integer;
2276
2277 case SO_SNDBUF:
6d49aa6f 2278 optval = so->so_snd.ssb_hiwat;
984263bc
MD
2279 goto integer;
2280
2281 case SO_RCVBUF:
6d49aa6f 2282 optval = so->so_rcv.ssb_hiwat;
984263bc
MD
2283 goto integer;
2284
2285 case SO_SNDLOWAT:
6d49aa6f 2286 optval = so->so_snd.ssb_lowat;
984263bc
MD
2287 goto integer;
2288
2289 case SO_RCVLOWAT:
6d49aa6f 2290 optval = so->so_rcv.ssb_lowat;
984263bc
MD
2291 goto integer;
2292
2293 case SO_SNDTIMEO:
2294 case SO_RCVTIMEO:
2295 optval = (sopt->sopt_name == SO_SNDTIMEO ?
6d49aa6f 2296 so->so_snd.ssb_timeo : so->so_rcv.ssb_timeo);
984263bc
MD
2297
2298 tv.tv_sec = optval / hz;
a591f597 2299 tv.tv_usec = (optval % hz) * ustick;
984263bc
MD
2300 error = sooptcopyout(sopt, &tv, sizeof tv);
2301 break;
2302
755c519c
SZ
2303 case SO_SNDSPACE:
2304 optval_l = ssb_space(&so->so_snd);
2305 error = sooptcopyout(sopt, &optval_l, sizeof(optval_l));
2306 break;
2307
984263bc
MD
2308 default:
2309 error = ENOPROTOOPT;
2310 break;
2311 }
2312 return (error);
2313 }
2314}
2315
2316/* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
2317int
2318soopt_getm(struct sockopt *sopt, struct mbuf **mp)
2319{
2320 struct mbuf *m, *m_prev;
bf6ac9fa
JH
2321 int sopt_size = sopt->sopt_valsize, msize;
2322
2323 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_DATA,
2324 0, &msize);
2325 if (m == NULL)
2326 return (ENOBUFS);
2327 m->m_len = min(msize, sopt_size);
984263bc
MD
2328 sopt_size -= m->m_len;
2329 *mp = m;
2330 m_prev = m;
2331
bf6ac9fa
JH
2332 while (sopt_size > 0) {
2333 m = m_getl(sopt_size, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT,
2334 MT_DATA, 0, &msize);
2335 if (m == NULL) {
984263bc 2336 m_freem(*mp);
bf6ac9fa 2337 return (ENOBUFS);
984263bc 2338 }
bf6ac9fa 2339 m->m_len = min(msize, sopt_size);
984263bc
MD
2340 sopt_size -= m->m_len;
2341 m_prev->m_next = m;
2342 m_prev = m;
2343 }
bf6ac9fa 2344 return (0);
984263bc
MD
2345}
2346
2347/* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
2348int
2349soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
2350{
de0003fe
AE
2351 soopt_to_mbuf(sopt, m);
2352 return 0;
2353}
2354
2355void
2356soopt_to_mbuf(struct sockopt *sopt, struct mbuf *m)
2357{
c3e742f9
NT
2358 size_t valsize;
2359 void *val;
984263bc 2360
792239df 2361 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val));
de0003fe 2362 KKASSERT(kva_p(m));
984263bc 2363 if (sopt->sopt_val == NULL)
792239df 2364 return;
c3e742f9
NT
2365 val = sopt->sopt_val;
2366 valsize = sopt->sopt_valsize;
2367 while (m != NULL && valsize >= m->m_len) {
de0003fe 2368 bcopy(val, mtod(m, char *), m->m_len);
c3e742f9
NT
2369 valsize -= m->m_len;
2370 val = (caddr_t)val + m->m_len;
984263bc
MD
2371 m = m->m_next;
2372 }
2373 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
2374 panic("ip6_sooptmcopyin");
984263bc
MD
2375}
2376
de0003fe
AE
2377/* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
2378int
2379soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
e71a125f 2380{
de0003fe 2381 return soopt_from_mbuf(sopt, m);
e71a125f
AE
2382}
2383
984263bc 2384int
de0003fe 2385soopt_from_mbuf(struct sockopt *sopt, struct mbuf *m)
984263bc
MD
2386{
2387 struct mbuf *m0 = m;
2388 size_t valsize = 0;
c3e742f9
NT
2389 size_t maxsize;
2390 void *val;
984263bc 2391
792239df 2392 KKASSERT(!sopt->sopt_val || kva_p(sopt->sopt_val));
de0003fe 2393 KKASSERT(kva_p(m));
984263bc
MD
2394 if (sopt->sopt_val == NULL)
2395 return 0;
c3e742f9
NT
2396 val = sopt->sopt_val;
2397 maxsize = sopt->sopt_valsize;
2398 while (m != NULL && maxsize >= m->m_len) {
de0003fe 2399 bcopy(mtod(m, char *), val, m->m_len);
c3e742f9
NT
2400 maxsize -= m->m_len;
2401 val = (caddr_t)val + m->m_len;
984263bc
MD
2402 valsize += m->m_len;
2403 m = m->m_next;
2404 }
2405 if (m != NULL) {
2406 /* enough soopt buffer should be given from user-land */
2407 m_freem(m0);
bf6ac9fa 2408 return (EINVAL);
984263bc
MD
2409 }
2410 sopt->sopt_valsize = valsize;
2411 return 0;
2412}
2413
2414void
c972a82f 2415sohasoutofband(struct socket *so)
984263bc
MD
2416{
2417 if (so->so_sigio != NULL)
2418 pgsigio(so->so_sigio, SIGURG, 0);
5b22f1a7 2419 KNOTE(&so->so_rcv.ssb_kq.ki_note, NOTE_OOB);
984263bc
MD
2420}
2421
2422int
984263bc
MD
2423sokqfilter(struct file *fp, struct knote *kn)
2424{
2425 struct socket *so = (struct socket *)kn->kn_fp->f_data;
6d49aa6f 2426 struct signalsockbuf *ssb;
984263bc
MD
2427
2428 switch (kn->kn_filter) {
2429 case EVFILT_READ:
2430 if (so->so_options & SO_ACCEPTCONN)
2431 kn->kn_fop = &solisten_filtops;
2432 else
2433 kn->kn_fop = &soread_filtops;
6d49aa6f 2434 ssb = &so->so_rcv;
984263bc
MD
2435 break;
2436 case EVFILT_WRITE:
2437 kn->kn_fop = &sowrite_filtops;
6d49aa6f 2438 ssb = &so->so_snd;
984263bc 2439 break;
73c344d3
SG
2440 case EVFILT_EXCEPT:
2441 kn->kn_fop = &soexcept_filtops;
2442 ssb = &so->so_rcv;
2443 break;
984263bc 2444 default:
b287d649 2445 return (EOPNOTSUPP);
984263bc
MD
2446 }
2447
5b22f1a7 2448 knote_insert(&ssb->ssb_kq.ki_note, kn);
14343ad3 2449 atomic_set_int(&ssb->ssb_flags, SSB_KNOTE);
984263bc
MD
2450 return (0);
2451}
2452
2453static void
2454filt_sordetach(struct knote *kn)
2455{
2456 struct socket *so = (struct socket *)kn->kn_fp->f_data;
984263bc 2457
5b22f1a7
SG
2458 knote_remove(&so->so_rcv.ssb_kq.ki_note, kn);
2459 if (SLIST_EMPTY(&so->so_rcv.ssb_kq.ki_note))
14343ad3 2460 atomic_clear_int(&so->so_rcv.ssb_flags, SSB_KNOTE);
984263bc
MD
2461}
2462
2463/*ARGSUSED*/
2464static int
2465filt_soread(struct knote *kn, long hint)
2466{
2467 struct socket *so = (struct socket *)kn->kn_fp->f_data;
2468
73c344d3
SG
2469 if (kn->kn_sfflags & NOTE_OOB) {
2470 if ((so->so_oobmark || (so->so_state & SS_RCVATMARK))) {
2471 kn->kn_fflags |= NOTE_OOB;
2472 return (1);
2473 }
2474 return (0);
70a4a30f 2475 }
6d49aa6f 2476 kn->kn_data = so->so_rcv.ssb_cc;
8c4ed426 2477
3bcb6e5e
SZ
2478 if (so->so_state & SS_CANTRCVMORE) {
2479 /*
2480 * Only set NODATA if all data has been exhausted.
2481 */
2482 if (kn->kn_data == 0)
2483 kn->kn_flags |= EV_NODATA;
984263bc
MD
2484 kn->kn_flags |= EV_EOF;
2485 kn->kn_fflags = so->so_error;
2486 return (1);
2487 }
2488 if (so->so_error) /* temporary udp error */
2489 return (1);
2490 if (kn->kn_sfflags & NOTE_LOWAT)
2491 return (kn->kn_data >= kn->kn_sdata);
e5857bf7 2492 return ((kn->kn_data >= so->so_rcv.ssb_lowat) ||
6cef7136 2493 !TAILQ_EMPTY(&so->so_comp));
984263bc
MD
2494}
2495
2496static void
2497filt_sowdetach(struct knote *kn)
2498{
2499 struct socket *so = (struct socket *)kn->kn_fp->f_data;
984263bc 2500
5b22f1a7
SG
2501 knote_remove(&so->so_snd.ssb_kq.ki_note, kn);
2502 if (SLIST_EMPTY(&so->so_snd.ssb_kq.ki_note))
14343ad3 2503 atomic_clear_int(&so->so_snd.ssb_flags, SSB_KNOTE);
984263bc
MD
2504}
2505
2506/*ARGSUSED*/
2507static int
2508filt_sowrite(struct knote *kn, long hint)
2509{
2510 struct socket *so = (struct socket *)kn->kn_fp->f_data;
2511
6d49aa6f 2512 kn->kn_data = ssb_space(&so->so_snd);
984263bc 2513 if (so->so_state & SS_CANTSENDMORE) {
3bcb6e5e 2514 kn->kn_flags |= (EV_EOF | EV_NODATA);
984263bc
MD
2515 kn->kn_fflags = so->so_error;
2516 return (1);
2517 }
2518 if (so->so_error) /* temporary udp error */
2519 return (1);
2520 if (((so->so_state & SS_ISCONNECTED) == 0) &&
2521 (so->so_proto->pr_flags & PR_CONNREQUIRED))
2522 return (0);
2523 if (kn->kn_sfflags & NOTE_LOWAT)
2524 return (kn->kn_data >= kn->kn_sdata);
6d49aa6f 2525 return (kn->kn_data >= so->so_snd.ssb_lowat);
984263bc
MD
2526}
2527
2528/*ARGSUSED*/
2529static int
2530filt_solisten(struct knote *kn, long hint)
2531{
2532 struct socket *so = (struct socket *)kn->kn_fp->f_data;
2533
2534 kn->kn_data = so->so_qlen;
2535 return (! TAILQ_EMPTY(&so->so_comp));
2536}