kttcp.c revision 1.41 1 1.41 riastrad /* $NetBSD: kttcp.c,v 1.41 2018/09/03 16:29:30 riastradh Exp $ */
2 1.1 thorpej
3 1.1 thorpej /*
4 1.1 thorpej * Copyright (c) 2002 Wasabi Systems, Inc.
5 1.1 thorpej * All rights reserved.
6 1.1 thorpej *
7 1.1 thorpej * Written by Frank van der Linden and Jason R. Thorpe for
8 1.1 thorpej * Wasabi Systems, Inc.
9 1.1 thorpej *
10 1.1 thorpej * Redistribution and use in source and binary forms, with or without
11 1.1 thorpej * modification, are permitted provided that the following conditions
12 1.1 thorpej * are met:
13 1.1 thorpej * 1. Redistributions of source code must retain the above copyright
14 1.1 thorpej * notice, this list of conditions and the following disclaimer.
15 1.1 thorpej * 2. Redistributions in binary form must reproduce the above copyright
16 1.1 thorpej * notice, this list of conditions and the following disclaimer in the
17 1.1 thorpej * documentation and/or other materials provided with the distribution.
18 1.1 thorpej * 3. All advertising materials mentioning features or use of this software
19 1.1 thorpej * must display the following acknowledgement:
20 1.1 thorpej * This product includes software developed for the NetBSD Project by
21 1.1 thorpej * Wasabi Systems, Inc.
22 1.1 thorpej * 4. The name of Wasabi Systems, Inc. may not be used to endorse
23 1.1 thorpej * or promote products derived from this software without specific prior
24 1.1 thorpej * written permission.
25 1.1 thorpej *
26 1.1 thorpej * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
27 1.1 thorpej * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28 1.1 thorpej * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29 1.1 thorpej * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC
30 1.1 thorpej * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 1.1 thorpej * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 1.1 thorpej * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 1.1 thorpej * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 1.1 thorpej * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 1.1 thorpej * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 1.1 thorpej * POSSIBILITY OF SUCH DAMAGE.
37 1.1 thorpej */
38 1.1 thorpej
39 1.1 thorpej /*
40 1.22 hubertf * kttcp.c -- provides kernel support for testing network testing,
41 1.22 hubertf * see kttcp(4)
42 1.1 thorpej */
43 1.14 lukem
44 1.14 lukem #include <sys/cdefs.h>
45 1.41 riastrad __KERNEL_RCSID(0, "$NetBSD: kttcp.c,v 1.41 2018/09/03 16:29:30 riastradh Exp $");
46 1.1 thorpej
47 1.1 thorpej #include <sys/param.h>
48 1.1 thorpej #include <sys/types.h>
49 1.1 thorpej #include <sys/ioctl.h>
50 1.1 thorpej #include <sys/file.h>
51 1.1 thorpej #include <sys/filedesc.h>
52 1.1 thorpej #include <sys/conf.h>
53 1.1 thorpej #include <sys/systm.h>
54 1.1 thorpej #include <sys/protosw.h>
55 1.1 thorpej #include <sys/proc.h>
56 1.1 thorpej #include <sys/resourcevar.h>
57 1.1 thorpej #include <sys/signal.h>
58 1.1 thorpej #include <sys/socketvar.h>
59 1.1 thorpej #include <sys/socket.h>
60 1.1 thorpej #include <sys/mbuf.h>
61 1.1 thorpej #include <sys/mount.h>
62 1.1 thorpej #include <sys/syscallargs.h>
63 1.1 thorpej
64 1.1 thorpej #include <dev/kttcpio.h>
65 1.1 thorpej
66 1.38 christos #include "ioconf.h"
67 1.38 christos
68 1.17 christos static int kttcp_send(struct lwp *l, struct kttcp_io_args *);
69 1.17 christos static int kttcp_recv(struct lwp *l, struct kttcp_io_args *);
70 1.1 thorpej static int kttcp_sosend(struct socket *, unsigned long long,
71 1.17 christos unsigned long long *, struct lwp *, int);
72 1.1 thorpej static int kttcp_soreceive(struct socket *, unsigned long long,
73 1.17 christos unsigned long long *, struct lwp *, int *);
74 1.1 thorpej
75 1.5 gehenna dev_type_ioctl(kttcpioctl);
76 1.5 gehenna
77 1.5 gehenna const struct cdevsw kttcp_cdevsw = {
78 1.32 dholland .d_open = nullopen,
79 1.32 dholland .d_close = nullclose,
80 1.32 dholland .d_read = noread,
81 1.32 dholland .d_write = nowrite,
82 1.32 dholland .d_ioctl = kttcpioctl,
83 1.32 dholland .d_stop = nostop,
84 1.32 dholland .d_tty = notty,
85 1.32 dholland .d_poll = nopoll,
86 1.32 dholland .d_mmap = nommap,
87 1.32 dholland .d_kqfilter = nokqfilter,
88 1.35 dholland .d_discard = nodiscard,
89 1.32 dholland .d_flag = D_OTHER
90 1.5 gehenna };
91 1.1 thorpej
92 1.1 thorpej void
93 1.21 christos kttcpattach(int count)
94 1.1 thorpej {
95 1.1 thorpej /* Do nothing. */
96 1.1 thorpej }
97 1.1 thorpej
98 1.1 thorpej int
99 1.24 christos kttcpioctl(dev_t dev, u_long cmd, void *data, int flag,
100 1.20 christos struct lwp *l)
101 1.1 thorpej {
102 1.1 thorpej int error;
103 1.1 thorpej
104 1.1 thorpej if ((flag & FWRITE) == 0)
105 1.1 thorpej return EPERM;
106 1.1 thorpej
107 1.1 thorpej switch (cmd) {
108 1.1 thorpej case KTTCP_IO_SEND:
109 1.17 christos error = kttcp_send(l, (struct kttcp_io_args *) data);
110 1.1 thorpej break;
111 1.1 thorpej
112 1.1 thorpej case KTTCP_IO_RECV:
113 1.17 christos error = kttcp_recv(l, (struct kttcp_io_args *) data);
114 1.1 thorpej break;
115 1.1 thorpej
116 1.1 thorpej default:
117 1.1 thorpej return EINVAL;
118 1.1 thorpej }
119 1.1 thorpej
120 1.1 thorpej return error;
121 1.1 thorpej }
122 1.1 thorpej
123 1.1 thorpej static int
124 1.17 christos kttcp_send(struct lwp *l, struct kttcp_io_args *kio)
125 1.1 thorpej {
126 1.26 ad struct socket *so;
127 1.1 thorpej int error;
128 1.1 thorpej struct timeval t0, t1;
129 1.1 thorpej unsigned long long len, done;
130 1.1 thorpej
131 1.1 thorpej if (kio->kio_totalsize >= KTTCP_MAX_XMIT)
132 1.1 thorpej return EINVAL;
133 1.1 thorpej
134 1.26 ad if ((error = fd_getsock(kio->kio_socket, &so)) != 0)
135 1.26 ad return error;
136 1.1 thorpej
137 1.1 thorpej len = kio->kio_totalsize;
138 1.1 thorpej microtime(&t0);
139 1.1 thorpej do {
140 1.26 ad error = kttcp_sosend(so, len, &done, l, 0);
141 1.1 thorpej len -= done;
142 1.1 thorpej } while (error == 0 && len > 0);
143 1.8 briggs
144 1.26 ad fd_putfile(kio->kio_socket);
145 1.8 briggs
146 1.1 thorpej microtime(&t1);
147 1.1 thorpej if (error != 0)
148 1.1 thorpej return error;
149 1.1 thorpej timersub(&t1, &t0, &kio->kio_elapsed);
150 1.1 thorpej
151 1.1 thorpej kio->kio_bytesdone = kio->kio_totalsize - len;
152 1.1 thorpej
153 1.1 thorpej return 0;
154 1.1 thorpej }
155 1.1 thorpej
156 1.1 thorpej static int
157 1.17 christos kttcp_recv(struct lwp *l, struct kttcp_io_args *kio)
158 1.1 thorpej {
159 1.26 ad struct socket *so;
160 1.1 thorpej int error;
161 1.1 thorpej struct timeval t0, t1;
162 1.1 thorpej unsigned long long len, done;
163 1.1 thorpej
164 1.18 mrg done = 0; /* XXX gcc */
165 1.18 mrg
166 1.1 thorpej if (kio->kio_totalsize > KTTCP_MAX_XMIT)
167 1.1 thorpej return EINVAL;
168 1.1 thorpej
169 1.26 ad if ((error = fd_getsock(kio->kio_socket, &so)) != 0)
170 1.26 ad return error;
171 1.1 thorpej len = kio->kio_totalsize;
172 1.1 thorpej microtime(&t0);
173 1.1 thorpej do {
174 1.26 ad error = kttcp_soreceive(so, len, &done, l, NULL);
175 1.1 thorpej len -= done;
176 1.1 thorpej } while (error == 0 && len > 0 && done > 0);
177 1.8 briggs
178 1.26 ad fd_putfile(kio->kio_socket);
179 1.8 briggs
180 1.1 thorpej microtime(&t1);
181 1.1 thorpej if (error == EPIPE)
182 1.1 thorpej error = 0;
183 1.1 thorpej if (error != 0)
184 1.1 thorpej return error;
185 1.1 thorpej timersub(&t1, &t0, &kio->kio_elapsed);
186 1.1 thorpej
187 1.1 thorpej kio->kio_bytesdone = kio->kio_totalsize - len;
188 1.1 thorpej
189 1.1 thorpej return 0;
190 1.1 thorpej }
191 1.1 thorpej
192 1.1 thorpej #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
193 1.1 thorpej
194 1.1 thorpej /*
195 1.1 thorpej * Slightly changed version of sosend()
196 1.1 thorpej */
197 1.1 thorpej static int
198 1.1 thorpej kttcp_sosend(struct socket *so, unsigned long long slen,
199 1.17 christos unsigned long long *done, struct lwp *l, int flags)
200 1.1 thorpej {
201 1.1 thorpej struct mbuf **mp, *m, *top;
202 1.1 thorpej long space, len, mlen;
203 1.28 ad int error, dontroute, atomic;
204 1.1 thorpej long long resid;
205 1.1 thorpej
206 1.1 thorpej atomic = sosendallatonce(so);
207 1.1 thorpej resid = slen;
208 1.1 thorpej top = NULL;
209 1.1 thorpej /*
210 1.1 thorpej * In theory resid should be unsigned.
211 1.1 thorpej * However, space must be signed, as it might be less than 0
212 1.1 thorpej * if we over-committed, and we must use a signed comparison
213 1.1 thorpej * of space and resid. On the other hand, a negative resid
214 1.1 thorpej * causes us to loop sending 0-length segments to the protocol.
215 1.1 thorpej */
216 1.1 thorpej if (resid < 0) {
217 1.1 thorpej error = EINVAL;
218 1.1 thorpej goto out;
219 1.1 thorpej }
220 1.1 thorpej dontroute =
221 1.1 thorpej (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
222 1.1 thorpej (so->so_proto->pr_flags & PR_ATOMIC);
223 1.27 ad l->l_ru.ru_msgsnd++;
224 1.28 ad #define snderr(errno) { error = errno; goto release; }
225 1.28 ad solock(so);
226 1.1 thorpej restart:
227 1.1 thorpej if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
228 1.1 thorpej goto out;
229 1.1 thorpej do {
230 1.1 thorpej if (so->so_state & SS_CANTSENDMORE)
231 1.1 thorpej snderr(EPIPE);
232 1.1 thorpej if (so->so_error) {
233 1.1 thorpej error = so->so_error;
234 1.1 thorpej so->so_error = 0;
235 1.1 thorpej goto release;
236 1.1 thorpej }
237 1.1 thorpej if ((so->so_state & SS_ISCONNECTED) == 0) {
238 1.1 thorpej if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
239 1.31 rmind snderr(ENOTCONN);
240 1.31 rmind } else {
241 1.1 thorpej snderr(EDESTADDRREQ);
242 1.31 rmind }
243 1.1 thorpej }
244 1.1 thorpej space = sbspace(&so->so_snd);
245 1.1 thorpej if (flags & MSG_OOB)
246 1.1 thorpej space += 1024;
247 1.1 thorpej if ((atomic && resid > so->so_snd.sb_hiwat))
248 1.1 thorpej snderr(EMSGSIZE);
249 1.1 thorpej if (space < resid && (atomic || space < so->so_snd.sb_lowat)) {
250 1.29 christos if (so->so_state & SS_NBIO)
251 1.1 thorpej snderr(EWOULDBLOCK);
252 1.2 thorpej SBLASTRECORDCHK(&so->so_rcv,
253 1.2 thorpej "kttcp_soreceive sbwait 1");
254 1.2 thorpej SBLASTMBUFCHK(&so->so_rcv,
255 1.2 thorpej "kttcp_soreceive sbwait 1");
256 1.1 thorpej sbunlock(&so->so_snd);
257 1.1 thorpej error = sbwait(&so->so_snd);
258 1.1 thorpej if (error)
259 1.1 thorpej goto out;
260 1.1 thorpej goto restart;
261 1.1 thorpej }
262 1.1 thorpej mp = ⊤
263 1.1 thorpej do {
264 1.28 ad sounlock(so);
265 1.1 thorpej do {
266 1.1 thorpej if (top == 0) {
267 1.9 matt m = m_gethdr(M_WAIT, MT_DATA);
268 1.1 thorpej mlen = MHLEN;
269 1.1 thorpej m->m_pkthdr.len = 0;
270 1.39 ozaki m_reset_rcvif(m);
271 1.1 thorpej } else {
272 1.9 matt m = m_get(M_WAIT, MT_DATA);
273 1.1 thorpej mlen = MLEN;
274 1.1 thorpej }
275 1.1 thorpej if (resid >= MINCLSIZE && space >= MCLBYTES) {
276 1.9 matt m_clget(m, M_WAIT);
277 1.1 thorpej if ((m->m_flags & M_EXT) == 0)
278 1.1 thorpej goto nopages;
279 1.1 thorpej mlen = MCLBYTES;
280 1.1 thorpej #ifdef MAPPED_MBUFS
281 1.1 thorpej len = lmin(MCLBYTES, resid);
282 1.1 thorpej #else
283 1.1 thorpej if (atomic && top == 0) {
284 1.1 thorpej len = lmin(MCLBYTES - max_hdr,
285 1.1 thorpej resid);
286 1.1 thorpej m->m_data += max_hdr;
287 1.1 thorpej } else
288 1.1 thorpej len = lmin(MCLBYTES, resid);
289 1.1 thorpej #endif
290 1.1 thorpej space -= len;
291 1.1 thorpej } else {
292 1.1 thorpej nopages:
293 1.1 thorpej len = lmin(lmin(mlen, resid), space);
294 1.1 thorpej space -= len;
295 1.1 thorpej /*
296 1.1 thorpej * For datagram protocols, leave room
297 1.1 thorpej * for protocol headers in first mbuf.
298 1.1 thorpej */
299 1.1 thorpej if (atomic && top == 0 && len < mlen)
300 1.1 thorpej MH_ALIGN(m, len);
301 1.1 thorpej }
302 1.1 thorpej resid -= len;
303 1.1 thorpej m->m_len = len;
304 1.1 thorpej *mp = m;
305 1.1 thorpej top->m_pkthdr.len += len;
306 1.1 thorpej if (error)
307 1.1 thorpej goto release;
308 1.1 thorpej mp = &m->m_next;
309 1.1 thorpej if (resid <= 0) {
310 1.1 thorpej if (flags & MSG_EOR)
311 1.1 thorpej top->m_flags |= M_EOR;
312 1.1 thorpej break;
313 1.1 thorpej }
314 1.1 thorpej } while (space > 0 && atomic);
315 1.28 ad solock(so);
316 1.1 thorpej
317 1.1 thorpej if (so->so_state & SS_CANTSENDMORE)
318 1.1 thorpej snderr(EPIPE);
319 1.1 thorpej if (dontroute)
320 1.1 thorpej so->so_options |= SO_DONTROUTE;
321 1.1 thorpej if (resid > 0)
322 1.1 thorpej so->so_state |= SS_MORETOCOME;
323 1.34 rtr if (flags & MSG_OOB)
324 1.34 rtr error = (*so->so_proto->pr_usrreqs->pr_sendoob)(so,
325 1.34 rtr top, NULL);
326 1.34 rtr else
327 1.36 rtr error = (*so->so_proto->pr_usrreqs->pr_send)(so,
328 1.36 rtr top, NULL, NULL, l);
329 1.1 thorpej if (dontroute)
330 1.1 thorpej so->so_options &= ~SO_DONTROUTE;
331 1.1 thorpej if (resid > 0)
332 1.1 thorpej so->so_state &= ~SS_MORETOCOME;
333 1.1 thorpej top = 0;
334 1.1 thorpej mp = ⊤
335 1.1 thorpej if (error)
336 1.1 thorpej goto release;
337 1.1 thorpej } while (resid && space > 0);
338 1.1 thorpej } while (resid);
339 1.1 thorpej
340 1.1 thorpej release:
341 1.1 thorpej sbunlock(&so->so_snd);
342 1.1 thorpej out:
343 1.28 ad sounlock(so);
344 1.1 thorpej if (top)
345 1.1 thorpej m_freem(top);
346 1.1 thorpej *done = slen - resid;
347 1.1 thorpej #if 0
348 1.1 thorpej printf("sosend: error %d slen %llu resid %lld\n", error, slen, resid);
349 1.1 thorpej #endif
350 1.1 thorpej return (error);
351 1.1 thorpej }
352 1.1 thorpej
353 1.1 thorpej static int
354 1.1 thorpej kttcp_soreceive(struct socket *so, unsigned long long slen,
355 1.21 christos unsigned long long *done, struct lwp *l, int *flagsp)
356 1.1 thorpej {
357 1.1 thorpej struct mbuf *m, **mp;
358 1.28 ad int flags, len, error, offset, moff, type;
359 1.1 thorpej long long orig_resid, resid;
360 1.15 matt const struct protosw *pr;
361 1.1 thorpej struct mbuf *nextrecord;
362 1.1 thorpej
363 1.1 thorpej pr = so->so_proto;
364 1.1 thorpej mp = NULL;
365 1.1 thorpej type = 0;
366 1.1 thorpej resid = orig_resid = slen;
367 1.1 thorpej if (flagsp)
368 1.1 thorpej flags = *flagsp &~ MSG_EOR;
369 1.1 thorpej else
370 1.1 thorpej flags = 0;
371 1.1 thorpej if (flags & MSG_OOB) {
372 1.1 thorpej m = m_get(M_WAIT, MT_DATA);
373 1.28 ad solock(so);
374 1.34 rtr error = (*pr->pr_usrreqs->pr_recvoob)(so, m, flags & MSG_PEEK);
375 1.28 ad sounlock(so);
376 1.1 thorpej if (error)
377 1.1 thorpej goto bad;
378 1.1 thorpej do {
379 1.41 riastrad resid -= uimin(resid, m->m_len);
380 1.1 thorpej m = m_free(m);
381 1.1 thorpej } while (resid && error == 0 && m);
382 1.1 thorpej bad:
383 1.1 thorpej if (m)
384 1.1 thorpej m_freem(m);
385 1.1 thorpej return (error);
386 1.1 thorpej }
387 1.1 thorpej if (mp)
388 1.12 simonb *mp = NULL;
389 1.28 ad solock(so);
390 1.1 thorpej restart:
391 1.1 thorpej if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
392 1.1 thorpej return (error);
393 1.1 thorpej m = so->so_rcv.sb_mb;
394 1.1 thorpej /*
395 1.1 thorpej * If we have less data than requested, block awaiting more
396 1.1 thorpej * (subject to any timeout) if:
397 1.1 thorpej * 1. the current count is less than the low water mark,
398 1.1 thorpej * 2. MSG_WAITALL is set, and it is possible to do the entire
399 1.1 thorpej * receive operation at once if we block (resid <= hiwat), or
400 1.1 thorpej * 3. MSG_DONTWAIT is not set.
401 1.1 thorpej * If MSG_WAITALL is set but resid is larger than the receive buffer,
402 1.1 thorpej * we have to do the receive in sections, and thus risk returning
403 1.1 thorpej * a short count if a timeout or signal occurs after we start.
404 1.1 thorpej */
405 1.13 fvdl if (m == NULL || (((flags & MSG_DONTWAIT) == 0 &&
406 1.1 thorpej so->so_rcv.sb_cc < resid) &&
407 1.1 thorpej (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
408 1.1 thorpej ((flags & MSG_WAITALL) && resid <= so->so_rcv.sb_hiwat)) &&
409 1.13 fvdl m->m_nextpkt == NULL && (pr->pr_flags & PR_ATOMIC) == 0)) {
410 1.1 thorpej #ifdef DIAGNOSTIC
411 1.13 fvdl if (m == NULL && so->so_rcv.sb_cc)
412 1.1 thorpej panic("receive 1");
413 1.1 thorpej #endif
414 1.1 thorpej if (so->so_error) {
415 1.1 thorpej if (m)
416 1.1 thorpej goto dontblock;
417 1.1 thorpej error = so->so_error;
418 1.1 thorpej if ((flags & MSG_PEEK) == 0)
419 1.1 thorpej so->so_error = 0;
420 1.1 thorpej goto release;
421 1.1 thorpej }
422 1.1 thorpej if (so->so_state & SS_CANTRCVMORE) {
423 1.1 thorpej if (m)
424 1.1 thorpej goto dontblock;
425 1.1 thorpej else
426 1.1 thorpej goto release;
427 1.1 thorpej }
428 1.1 thorpej for (; m; m = m->m_next)
429 1.1 thorpej if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
430 1.1 thorpej m = so->so_rcv.sb_mb;
431 1.1 thorpej goto dontblock;
432 1.1 thorpej }
433 1.1 thorpej if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
434 1.1 thorpej (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
435 1.1 thorpej error = ENOTCONN;
436 1.1 thorpej goto release;
437 1.1 thorpej }
438 1.1 thorpej if (resid == 0)
439 1.1 thorpej goto release;
440 1.30 jakllsch if ((so->so_state & SS_NBIO) ||
441 1.30 jakllsch (flags & (MSG_DONTWAIT|MSG_NBIO))) {
442 1.1 thorpej error = EWOULDBLOCK;
443 1.1 thorpej goto release;
444 1.1 thorpej }
445 1.1 thorpej sbunlock(&so->so_rcv);
446 1.1 thorpej error = sbwait(&so->so_rcv);
447 1.28 ad if (error) {
448 1.28 ad sounlock(so);
449 1.1 thorpej return (error);
450 1.28 ad }
451 1.1 thorpej goto restart;
452 1.1 thorpej }
453 1.1 thorpej dontblock:
454 1.2 thorpej /*
455 1.2 thorpej * On entry here, m points to the first record of the socket buffer.
456 1.2 thorpej * While we process the initial mbufs containing address and control
457 1.2 thorpej * info, we save a copy of m->m_nextpkt into nextrecord.
458 1.2 thorpej */
459 1.1 thorpej #ifdef notyet /* XXXX */
460 1.17 christos if (uio->uio_lwp)
461 1.27 ad uio->uio_lwp->l_ru.ru_msgrcv++;
462 1.1 thorpej #endif
463 1.2 thorpej KASSERT(m == so->so_rcv.sb_mb);
464 1.2 thorpej SBLASTRECORDCHK(&so->so_rcv, "kttcp_soreceive 1");
465 1.2 thorpej SBLASTMBUFCHK(&so->so_rcv, "kttcp_soreceive 1");
466 1.1 thorpej nextrecord = m->m_nextpkt;
467 1.1 thorpej if (pr->pr_flags & PR_ADDR) {
468 1.1 thorpej #ifdef DIAGNOSTIC
469 1.1 thorpej if (m->m_type != MT_SONAME)
470 1.1 thorpej panic("receive 1a");
471 1.1 thorpej #endif
472 1.1 thorpej orig_resid = 0;
473 1.1 thorpej if (flags & MSG_PEEK) {
474 1.1 thorpej m = m->m_next;
475 1.1 thorpej } else {
476 1.1 thorpej sbfree(&so->so_rcv, m);
477 1.40 christos m = so->so_rcv.sb_mb = m_free(m);
478 1.1 thorpej }
479 1.1 thorpej }
480 1.1 thorpej while (m && m->m_type == MT_CONTROL && error == 0) {
481 1.1 thorpej if (flags & MSG_PEEK) {
482 1.1 thorpej m = m->m_next;
483 1.1 thorpej } else {
484 1.1 thorpej sbfree(&so->so_rcv, m);
485 1.40 christos m = so->so_rcv.sb_mb = m_free(m);
486 1.1 thorpej }
487 1.1 thorpej }
488 1.2 thorpej
489 1.2 thorpej /*
490 1.2 thorpej * If m is non-NULL, we have some data to read. From now on,
491 1.2 thorpej * make sure to keep sb_lastrecord consistent when working on
492 1.2 thorpej * the last packet on the chain (nextrecord == NULL) and we
493 1.2 thorpej * change m->m_nextpkt.
494 1.2 thorpej */
495 1.1 thorpej if (m) {
496 1.2 thorpej if ((flags & MSG_PEEK) == 0) {
497 1.1 thorpej m->m_nextpkt = nextrecord;
498 1.2 thorpej /*
499 1.2 thorpej * If nextrecord == NULL (this is a single chain),
500 1.2 thorpej * then sb_lastrecord may not be valid here if m
501 1.2 thorpej * was changed earlier.
502 1.2 thorpej */
503 1.2 thorpej if (nextrecord == NULL) {
504 1.2 thorpej KASSERT(so->so_rcv.sb_mb == m);
505 1.2 thorpej so->so_rcv.sb_lastrecord = m;
506 1.2 thorpej }
507 1.2 thorpej }
508 1.1 thorpej type = m->m_type;
509 1.1 thorpej if (type == MT_OOBDATA)
510 1.1 thorpej flags |= MSG_OOB;
511 1.2 thorpej } else {
512 1.2 thorpej if ((flags & MSG_PEEK) == 0) {
513 1.2 thorpej KASSERT(so->so_rcv.sb_mb == m);
514 1.2 thorpej so->so_rcv.sb_mb = nextrecord;
515 1.4 thorpej SB_EMPTY_FIXUP(&so->so_rcv);
516 1.2 thorpej }
517 1.1 thorpej }
518 1.2 thorpej SBLASTRECORDCHK(&so->so_rcv, "kttcp_soreceive 2");
519 1.2 thorpej SBLASTMBUFCHK(&so->so_rcv, "kttcp_soreceive 2");
520 1.2 thorpej
521 1.1 thorpej moff = 0;
522 1.1 thorpej offset = 0;
523 1.1 thorpej while (m && resid > 0 && error == 0) {
524 1.1 thorpej if (m->m_type == MT_OOBDATA) {
525 1.1 thorpej if (type != MT_OOBDATA)
526 1.1 thorpej break;
527 1.1 thorpej } else if (type == MT_OOBDATA)
528 1.1 thorpej break;
529 1.1 thorpej #ifdef DIAGNOSTIC
530 1.1 thorpej else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
531 1.1 thorpej panic("receive 3");
532 1.1 thorpej #endif
533 1.1 thorpej so->so_state &= ~SS_RCVATMARK;
534 1.1 thorpej len = resid;
535 1.1 thorpej if (so->so_oobmark && len > so->so_oobmark - offset)
536 1.1 thorpej len = so->so_oobmark - offset;
537 1.1 thorpej if (len > m->m_len - moff)
538 1.1 thorpej len = m->m_len - moff;
539 1.1 thorpej /*
540 1.1 thorpej * If mp is set, just pass back the mbufs.
541 1.1 thorpej * Otherwise copy them out via the uio, then free.
542 1.1 thorpej * Sockbuf must be consistent here (points to current mbuf,
543 1.1 thorpej * it points to next record) when we drop priority;
544 1.1 thorpej * we must note any additions to the sockbuf when we
545 1.1 thorpej * block interrupts again.
546 1.1 thorpej */
547 1.1 thorpej resid -= len;
548 1.1 thorpej if (len == m->m_len - moff) {
549 1.1 thorpej if (m->m_flags & M_EOR)
550 1.1 thorpej flags |= MSG_EOR;
551 1.1 thorpej if (flags & MSG_PEEK) {
552 1.1 thorpej m = m->m_next;
553 1.1 thorpej moff = 0;
554 1.1 thorpej } else {
555 1.1 thorpej nextrecord = m->m_nextpkt;
556 1.1 thorpej sbfree(&so->so_rcv, m);
557 1.1 thorpej if (mp) {
558 1.1 thorpej *mp = m;
559 1.1 thorpej mp = &m->m_next;
560 1.1 thorpej so->so_rcv.sb_mb = m = m->m_next;
561 1.12 simonb *mp = NULL;
562 1.1 thorpej } else {
563 1.40 christos m = so->so_rcv.sb_mb = m_free(m);
564 1.1 thorpej }
565 1.2 thorpej /*
566 1.2 thorpej * If m != NULL, we also know that
567 1.2 thorpej * so->so_rcv.sb_mb != NULL.
568 1.2 thorpej */
569 1.2 thorpej KASSERT(so->so_rcv.sb_mb == m);
570 1.2 thorpej if (m) {
571 1.1 thorpej m->m_nextpkt = nextrecord;
572 1.2 thorpej if (nextrecord == NULL)
573 1.2 thorpej so->so_rcv.sb_lastrecord = m;
574 1.2 thorpej } else {
575 1.2 thorpej so->so_rcv.sb_mb = nextrecord;
576 1.4 thorpej SB_EMPTY_FIXUP(&so->so_rcv);
577 1.2 thorpej }
578 1.2 thorpej SBLASTRECORDCHK(&so->so_rcv,
579 1.2 thorpej "kttcp_soreceive 3");
580 1.2 thorpej SBLASTMBUFCHK(&so->so_rcv,
581 1.2 thorpej "kttcp_soreceive 3");
582 1.1 thorpej }
583 1.1 thorpej } else {
584 1.1 thorpej if (flags & MSG_PEEK)
585 1.1 thorpej moff += len;
586 1.1 thorpej else {
587 1.28 ad if (mp) {
588 1.28 ad sounlock(so);
589 1.1 thorpej *mp = m_copym(m, 0, len, M_WAIT);
590 1.28 ad solock(so);
591 1.28 ad }
592 1.1 thorpej m->m_data += len;
593 1.1 thorpej m->m_len -= len;
594 1.1 thorpej so->so_rcv.sb_cc -= len;
595 1.1 thorpej }
596 1.1 thorpej }
597 1.1 thorpej if (so->so_oobmark) {
598 1.1 thorpej if ((flags & MSG_PEEK) == 0) {
599 1.1 thorpej so->so_oobmark -= len;
600 1.1 thorpej if (so->so_oobmark == 0) {
601 1.1 thorpej so->so_state |= SS_RCVATMARK;
602 1.1 thorpej break;
603 1.1 thorpej }
604 1.1 thorpej } else {
605 1.1 thorpej offset += len;
606 1.1 thorpej if (offset == so->so_oobmark)
607 1.1 thorpej break;
608 1.1 thorpej }
609 1.1 thorpej }
610 1.1 thorpej if (flags & MSG_EOR)
611 1.1 thorpej break;
612 1.1 thorpej /*
613 1.1 thorpej * If the MSG_WAITALL flag is set (for non-atomic socket),
614 1.1 thorpej * we must not quit until "uio->uio_resid == 0" or an error
615 1.1 thorpej * termination. If a signal/timeout occurs, return
616 1.1 thorpej * with a short count but without error.
617 1.1 thorpej * Keep sockbuf locked against other readers.
618 1.1 thorpej */
619 1.13 fvdl while (flags & MSG_WAITALL && m == NULL && resid > 0 &&
620 1.1 thorpej !sosendallatonce(so) && !nextrecord) {
621 1.1 thorpej if (so->so_error || so->so_state & SS_CANTRCVMORE)
622 1.1 thorpej break;
623 1.3 thorpej /*
624 1.3 thorpej * If we are peeking and the socket receive buffer is
625 1.3 thorpej * full, stop since we can't get more data to peek at.
626 1.3 thorpej */
627 1.3 thorpej if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
628 1.3 thorpej break;
629 1.3 thorpej /*
630 1.3 thorpej * If we've drained the socket buffer, tell the
631 1.3 thorpej * protocol in case it needs to do something to
632 1.3 thorpej * get it filled again.
633 1.3 thorpej */
634 1.33 rmind if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) {
635 1.37 rtr (*pr->pr_usrreqs->pr_rcvd)(so, flags, l);
636 1.33 rmind }
637 1.2 thorpej SBLASTRECORDCHK(&so->so_rcv,
638 1.2 thorpej "kttcp_soreceive sbwait 2");
639 1.2 thorpej SBLASTMBUFCHK(&so->so_rcv,
640 1.2 thorpej "kttcp_soreceive sbwait 2");
641 1.1 thorpej error = sbwait(&so->so_rcv);
642 1.1 thorpej if (error) {
643 1.1 thorpej sbunlock(&so->so_rcv);
644 1.28 ad sounlock(so);
645 1.1 thorpej return (0);
646 1.1 thorpej }
647 1.1 thorpej if ((m = so->so_rcv.sb_mb) != NULL)
648 1.1 thorpej nextrecord = m->m_nextpkt;
649 1.1 thorpej }
650 1.1 thorpej }
651 1.1 thorpej
652 1.1 thorpej if (m && pr->pr_flags & PR_ATOMIC) {
653 1.1 thorpej flags |= MSG_TRUNC;
654 1.1 thorpej if ((flags & MSG_PEEK) == 0)
655 1.1 thorpej (void) sbdroprecord(&so->so_rcv);
656 1.1 thorpej }
657 1.1 thorpej if ((flags & MSG_PEEK) == 0) {
658 1.13 fvdl if (m == NULL) {
659 1.2 thorpej /*
660 1.4 thorpej * First part is an SB_EMPTY_FIXUP(). Second part
661 1.2 thorpej * makes sure sb_lastrecord is up-to-date if
662 1.2 thorpej * there is still data in the socket buffer.
663 1.2 thorpej */
664 1.1 thorpej so->so_rcv.sb_mb = nextrecord;
665 1.2 thorpej if (so->so_rcv.sb_mb == NULL) {
666 1.2 thorpej so->so_rcv.sb_mbtail = NULL;
667 1.2 thorpej so->so_rcv.sb_lastrecord = NULL;
668 1.2 thorpej } else if (nextrecord->m_nextpkt == NULL)
669 1.2 thorpej so->so_rcv.sb_lastrecord = nextrecord;
670 1.2 thorpej }
671 1.2 thorpej SBLASTRECORDCHK(&so->so_rcv, "kttcp_soreceive 4");
672 1.2 thorpej SBLASTMBUFCHK(&so->so_rcv, "kttcp_soreceive 4");
673 1.33 rmind if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) {
674 1.37 rtr (*pr->pr_usrreqs->pr_rcvd)(so, flags, l);
675 1.33 rmind }
676 1.1 thorpej }
677 1.1 thorpej if (orig_resid == resid && orig_resid &&
678 1.1 thorpej (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
679 1.1 thorpej sbunlock(&so->so_rcv);
680 1.1 thorpej goto restart;
681 1.1 thorpej }
682 1.16 perry
683 1.1 thorpej if (flagsp)
684 1.1 thorpej *flagsp |= flags;
685 1.1 thorpej release:
686 1.1 thorpej sbunlock(&so->so_rcv);
687 1.28 ad sounlock(so);
688 1.1 thorpej *done = slen - resid;
689 1.1 thorpej #if 0
690 1.1 thorpej printf("soreceive: error %d slen %llu resid %lld\n", error, slen, resid);
691 1.1 thorpej #endif
692 1.1 thorpej return (error);
693 1.1 thorpej }
694