1 /*-
2 * Copyright (c) 1995 Søren Schmidt
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31
32 /* XXX we use functions that might not exist. */
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35
36 #include <sys/param.h>
37 #include <sys/proc.h>
38 #include <sys/systm.h>
39 #include <sys/sysproto.h>
40 #include <sys/capsicum.h>
41 #include <sys/fcntl.h>
42 #include <sys/file.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mutex.h>
47 #include <sys/mbuf.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/syscallsubr.h>
51 #include <sys/uio.h>
52 #include <sys/syslog.h>
53 #include <sys/un.h>
54
55 #include <net/if.h>
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/ip.h>
59 #include <netinet/tcp.h>
60 #ifdef INET6
61 #include <netinet/ip6.h>
62 #include <netinet6/ip6_var.h>
63 #include <netinet6/in6_var.h>
64 #endif
65
66 #ifdef COMPAT_LINUX32
67 #include <machine/../linux32/linux.h>
68 #include <machine/../linux32/linux32_proto.h>
69 #else
70 #include <machine/../linux/linux.h>
71 #include <machine/../linux/linux_proto.h>
72 #endif
73 #include <compat/linux/linux_file.h>
74 #include <compat/linux/linux_socket.h>
75 #include <compat/linux/linux_timer.h>
76 #include <compat/linux/linux_util.h>
77
78 static int linux_to_bsd_domain(int);
79 static int linux_sendmsg_common(struct thread *, l_int, struct l_msghdr *,
80 l_uint);
81 static int linux_recvmsg_common(struct thread *, l_int, struct l_msghdr *,
82 l_uint, struct msghdr *);
83 static int linux_set_socket_flags(int, int *);
84
85 /*
86 * Reads a linux sockaddr and does any necessary translation.
87 * Linux sockaddrs don't have a length field, only a family.
88 * Copy the osockaddr structure pointed to by osa to kernel, adjust
89 * family and convert to sockaddr.
90 */
91 static int
92 linux_getsockaddr(struct sockaddr **sap, const struct osockaddr *osa, int salen)
93 {
94 struct sockaddr *sa;
95 struct osockaddr *kosa;
96 #ifdef INET6
97 struct sockaddr_in6 *sin6;
98 int oldv6size;
99 #endif
100 char *name;
101 int bdom, error, hdrlen, namelen;
102
103 if (salen < 2 || salen > UCHAR_MAX || !osa)
104 return (EINVAL);
105
106 #ifdef INET6
107 oldv6size = 0;
108 /*
109 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
110 * if it's a v4-mapped address, so reserve the proper space
111 * for it.
112 */
113 if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
114 salen += sizeof(uint32_t);
115 oldv6size = 1;
116 }
117 #endif
118
119 kosa = malloc(salen, M_SONAME, M_WAITOK);
120
121 if ((error = copyin(osa, kosa, salen)))
122 goto out;
123
124 bdom = linux_to_bsd_domain(kosa->sa_family);
125 if (bdom == -1) {
126 error = EAFNOSUPPORT;
127 goto out;
128 }
129
130 #ifdef INET6
131 /*
132 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
133 * which lacks the scope id compared with RFC2553 one. If we detect
134 * the situation, reject the address and write a message to system log.
135 *
136 * Still accept addresses for which the scope id is not used.
137 */
138 if (oldv6size) {
139 if (bdom == AF_INET6) {
140 sin6 = (struct sockaddr_in6 *)kosa;
141 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
142 (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
143 !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
144 !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
145 !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
146 !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
147 sin6->sin6_scope_id = 0;
148 } else {
149 log(LOG_DEBUG,
150 "obsolete pre-RFC2553 sockaddr_in6 rejected\n");
151 error = EINVAL;
152 goto out;
153 }
154 } else
155 salen -= sizeof(uint32_t);
156 }
157 #endif
158 if (bdom == AF_INET) {
159 if (salen < sizeof(struct sockaddr_in)) {
160 error = EINVAL;
161 goto out;
162 }
163 salen = sizeof(struct sockaddr_in);
164 }
165
166 if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
167 hdrlen = offsetof(struct sockaddr_un, sun_path);
168 name = ((struct sockaddr_un *)kosa)->sun_path;
169 if (*name == '\0') {
170 /*
171 * Linux abstract namespace starts with a NULL byte.
172 * XXX We do not support abstract namespace yet.
173 */
174 namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
175 } else
176 namelen = strnlen(name, salen - hdrlen);
177 salen = hdrlen + namelen;
178 if (salen > sizeof(struct sockaddr_un)) {
179 error = ENAMETOOLONG;
180 goto out;
181 }
182 }
183
184 sa = (struct sockaddr *)kosa;
185 sa->sa_family = bdom;
186 sa->sa_len = salen;
187
188 *sap = sa;
189 return (0);
190
191 out:
192 free(kosa, M_SONAME);
193 return (error);
194 }
195
196 static int
197 linux_to_bsd_domain(int domain)
198 {
199
200 switch (domain) {
201 case LINUX_AF_UNSPEC:
202 return (AF_UNSPEC);
203 case LINUX_AF_UNIX:
204 return (AF_LOCAL);
205 case LINUX_AF_INET:
206 return (AF_INET);
207 case LINUX_AF_INET6:
208 return (AF_INET6);
209 case LINUX_AF_AX25:
210 return (AF_CCITT);
211 case LINUX_AF_IPX:
212 return (AF_IPX);
213 case LINUX_AF_APPLETALK:
214 return (AF_APPLETALK);
215 }
216 return (-1);
217 }
218
219 static int
220 bsd_to_linux_domain(int domain)
221 {
222
223 switch (domain) {
224 case AF_UNSPEC:
225 return (LINUX_AF_UNSPEC);
226 case AF_LOCAL:
227 return (LINUX_AF_UNIX);
228 case AF_INET:
229 return (LINUX_AF_INET);
230 case AF_INET6:
231 return (LINUX_AF_INET6);
232 case AF_CCITT:
233 return (LINUX_AF_AX25);
234 case AF_IPX:
235 return (LINUX_AF_IPX);
236 case AF_APPLETALK:
237 return (LINUX_AF_APPLETALK);
238 }
239 return (-1);
240 }
241
242 static int
243 linux_to_bsd_sockopt_level(int level)
244 {
245
246 switch (level) {
247 case LINUX_SOL_SOCKET:
248 return (SOL_SOCKET);
249 }
250 return (level);
251 }
252
253 static int
254 bsd_to_linux_sockopt_level(int level)
255 {
256
257 switch (level) {
258 case SOL_SOCKET:
259 return (LINUX_SOL_SOCKET);
260 }
261 return (level);
262 }
263
264 static int
265 linux_to_bsd_ip_sockopt(int opt)
266 {
267
268 switch (opt) {
269 case LINUX_IP_TOS:
270 return (IP_TOS);
271 case LINUX_IP_TTL:
272 return (IP_TTL);
273 case LINUX_IP_OPTIONS:
274 return (IP_OPTIONS);
275 case LINUX_IP_MULTICAST_IF:
276 return (IP_MULTICAST_IF);
277 case LINUX_IP_MULTICAST_TTL:
278 return (IP_MULTICAST_TTL);
279 case LINUX_IP_MULTICAST_LOOP:
280 return (IP_MULTICAST_LOOP);
281 case LINUX_IP_ADD_MEMBERSHIP:
282 return (IP_ADD_MEMBERSHIP);
283 case LINUX_IP_DROP_MEMBERSHIP:
284 return (IP_DROP_MEMBERSHIP);
285 case LINUX_IP_HDRINCL:
286 return (IP_HDRINCL);
287 }
288 return (-1);
289 }
290
291 static int
292 linux_to_bsd_ip6_sockopt(int opt)
293 {
294
295 switch (opt) {
296 case LINUX_IPV6_NEXTHOP:
297 return (IPV6_NEXTHOP);
298 case LINUX_IPV6_UNICAST_HOPS:
299 return (IPV6_UNICAST_HOPS);
300 case LINUX_IPV6_MULTICAST_IF:
301 return (IPV6_MULTICAST_IF);
302 case LINUX_IPV6_MULTICAST_HOPS:
303 return (IPV6_MULTICAST_HOPS);
304 case LINUX_IPV6_MULTICAST_LOOP:
305 return (IPV6_MULTICAST_LOOP);
306 case LINUX_IPV6_ADD_MEMBERSHIP:
307 return (IPV6_JOIN_GROUP);
308 case LINUX_IPV6_DROP_MEMBERSHIP:
309 return (IPV6_LEAVE_GROUP);
310 case LINUX_IPV6_V6ONLY:
311 return (IPV6_V6ONLY);
312 case LINUX_IPV6_DONTFRAG:
313 return (IPV6_DONTFRAG);
314 #if 0
315 case LINUX_IPV6_CHECKSUM:
316 return (IPV6_CHECKSUM);
317 case LINUX_IPV6_RECVPKTINFO:
318 return (IPV6_RECVPKTINFO);
319 case LINUX_IPV6_PKTINFO:
320 return (IPV6_PKTINFO);
321 case LINUX_IPV6_RECVHOPLIMIT:
322 return (IPV6_RECVHOPLIMIT);
323 case LINUX_IPV6_HOPLIMIT:
324 return (IPV6_HOPLIMIT);
325 case LINUX_IPV6_RECVHOPOPTS:
326 return (IPV6_RECVHOPOPTS);
327 case LINUX_IPV6_HOPOPTS:
328 return (IPV6_HOPOPTS);
329 case LINUX_IPV6_RTHDRDSTOPTS:
330 return (IPV6_RTHDRDSTOPTS);
331 case LINUX_IPV6_RECVRTHDR:
332 return (IPV6_RECVRTHDR);
333 case LINUX_IPV6_RTHDR:
334 return (IPV6_RTHDR);
335 case LINUX_IPV6_RECVDSTOPTS:
336 return (IPV6_RECVDSTOPTS);
337 case LINUX_IPV6_DSTOPTS:
338 return (IPV6_DSTOPTS);
339 case LINUX_IPV6_RECVPATHMTU:
340 return (IPV6_RECVPATHMTU);
341 case LINUX_IPV6_PATHMTU:
342 return (IPV6_PATHMTU);
343 #endif
344 }
345 return (-1);
346 }
347
348 static int
349 linux_to_bsd_so_sockopt(int opt)
350 {
351
352 switch (opt) {
353 case LINUX_SO_DEBUG:
354 return (SO_DEBUG);
355 case LINUX_SO_REUSEADDR:
356 return (SO_REUSEADDR);
357 case LINUX_SO_TYPE:
358 return (SO_TYPE);
359 case LINUX_SO_ERROR:
360 return (SO_ERROR);
361 case LINUX_SO_DONTROUTE:
362 return (SO_DONTROUTE);
363 case LINUX_SO_BROADCAST:
364 return (SO_BROADCAST);
365 case LINUX_SO_SNDBUF:
366 return (SO_SNDBUF);
367 case LINUX_SO_RCVBUF:
368 return (SO_RCVBUF);
369 case LINUX_SO_KEEPALIVE:
370 return (SO_KEEPALIVE);
371 case LINUX_SO_OOBINLINE:
372 return (SO_OOBINLINE);
373 case LINUX_SO_LINGER:
374 return (SO_LINGER);
375 case LINUX_SO_PEERCRED:
376 return (LOCAL_PEERCRED);
377 case LINUX_SO_RCVLOWAT:
378 return (SO_RCVLOWAT);
379 case LINUX_SO_SNDLOWAT:
380 return (SO_SNDLOWAT);
381 case LINUX_SO_RCVTIMEO:
382 return (SO_RCVTIMEO);
383 case LINUX_SO_SNDTIMEO:
384 return (SO_SNDTIMEO);
385 case LINUX_SO_TIMESTAMP:
386 return (SO_TIMESTAMP);
387 case LINUX_SO_ACCEPTCONN:
388 return (SO_ACCEPTCONN);
389 }
390 return (-1);
391 }
392
393 static int
394 linux_to_bsd_tcp_sockopt(int opt)
395 {
396
397 switch (opt) {
398 case LINUX_TCP_NODELAY:
399 return (TCP_NODELAY);
400 case LINUX_TCP_MAXSEG:
401 return (TCP_MAXSEG);
402 case LINUX_TCP_KEEPIDLE:
403 return (TCP_KEEPIDLE);
404 case LINUX_TCP_KEEPINTVL:
405 return (TCP_KEEPINTVL);
406 case LINUX_TCP_KEEPCNT:
407 return (TCP_KEEPCNT);
408 case LINUX_TCP_MD5SIG:
409 return (TCP_MD5SIG);
410 }
411 return (-1);
412 }
413
414 static int
415 linux_to_bsd_msg_flags(int flags)
416 {
417 int ret_flags = 0;
418
419 if (flags & LINUX_MSG_OOB)
420 ret_flags |= MSG_OOB;
421 if (flags & LINUX_MSG_PEEK)
422 ret_flags |= MSG_PEEK;
423 if (flags & LINUX_MSG_DONTROUTE)
424 ret_flags |= MSG_DONTROUTE;
425 if (flags & LINUX_MSG_CTRUNC)
426 ret_flags |= MSG_CTRUNC;
427 if (flags & LINUX_MSG_TRUNC)
428 ret_flags |= MSG_TRUNC;
429 if (flags & LINUX_MSG_DONTWAIT)
430 ret_flags |= MSG_DONTWAIT;
431 if (flags & LINUX_MSG_EOR)
432 ret_flags |= MSG_EOR;
433 if (flags & LINUX_MSG_WAITALL)
434 ret_flags |= MSG_WAITALL;
435 if (flags & LINUX_MSG_NOSIGNAL)
436 ret_flags |= MSG_NOSIGNAL;
437 #if 0 /* not handled */
438 if (flags & LINUX_MSG_PROXY)
439 ;
440 if (flags & LINUX_MSG_FIN)
441 ;
442 if (flags & LINUX_MSG_SYN)
443 ;
444 if (flags & LINUX_MSG_CONFIRM)
445 ;
446 if (flags & LINUX_MSG_RST)
447 ;
448 if (flags & LINUX_MSG_ERRQUEUE)
449 ;
450 #endif
451 return (ret_flags);
452 }
453
454 /*
455 * If bsd_to_linux_sockaddr() or linux_to_bsd_sockaddr() faults, then the
456 * native syscall will fault. Thus, we don't really need to check the
457 * return values for these functions.
458 */
459
460 static int
461 bsd_to_linux_sockaddr(struct sockaddr *arg)
462 {
463 struct sockaddr sa;
464 size_t sa_len = sizeof(struct sockaddr);
465 int error, bdom;
466
467 if ((error = copyin(arg, &sa, sa_len)))
468 return (error);
469
470 bdom = bsd_to_linux_domain(sa.sa_family);
471 if (bdom == -1)
472 return (EAFNOSUPPORT);
473
474 *(u_short *)&sa = bdom;
475 return (copyout(&sa, arg, sa_len));
476 }
477
478 static int
479 linux_to_bsd_sockaddr(struct sockaddr *arg, int len)
480 {
481 struct sockaddr sa;
482 size_t sa_len = sizeof(struct sockaddr);
483 int error, bdom;
484
485 if ((error = copyin(arg, &sa, sa_len)))
486 return (error);
487
488 bdom = linux_to_bsd_domain(*(sa_family_t *)&sa);
489 if (bdom == -1)
490 return (EAFNOSUPPORT);
491
492 sa.sa_family = bdom;
493 sa.sa_len = len;
494 return (copyout(&sa, arg, sa_len));
495 }
496
497 static int
498 linux_sa_put(struct osockaddr *osa)
499 {
500 struct osockaddr sa;
501 int error, bdom;
502
503 /*
504 * Only read/write the osockaddr family part, the rest is
505 * not changed.
506 */
507 error = copyin(osa, &sa, sizeof(sa.sa_family));
508 if (error)
509 return (error);
510
511 bdom = bsd_to_linux_domain(sa.sa_family);
512 if (bdom == -1)
513 return (EINVAL);
514
515 sa.sa_family = bdom;
516 return (copyout(&sa, osa, sizeof(sa.sa_family)));
517 }
518
519 static int
520 linux_to_bsd_cmsg_type(int cmsg_type)
521 {
522
523 switch (cmsg_type) {
524 case LINUX_SCM_RIGHTS:
525 return (SCM_RIGHTS);
526 case LINUX_SCM_CREDENTIALS:
527 return (SCM_CREDS);
528 }
529 return (-1);
530 }
531
532 static int
533 bsd_to_linux_cmsg_type(int cmsg_type)
534 {
535
536 switch (cmsg_type) {
537 case SCM_RIGHTS:
538 return (LINUX_SCM_RIGHTS);
539 case SCM_CREDS:
540 return (LINUX_SCM_CREDENTIALS);
541 case SCM_TIMESTAMP:
542 return (LINUX_SCM_TIMESTAMP);
543 }
544 return (-1);
545 }
546
547 static int
548 linux_to_bsd_msghdr(struct msghdr *bhdr, const struct l_msghdr *lhdr)
549 {
550 if (lhdr->msg_controllen > INT_MAX)
551 return (ENOBUFS);
552
553 bhdr->msg_name = PTRIN(lhdr->msg_name);
554 bhdr->msg_namelen = lhdr->msg_namelen;
555 bhdr->msg_iov = PTRIN(lhdr->msg_iov);
556 bhdr->msg_iovlen = lhdr->msg_iovlen;
557 bhdr->msg_control = PTRIN(lhdr->msg_control);
558
559 /*
560 * msg_controllen is skipped since BSD and LINUX control messages
561 * are potentially different sizes (e.g. the cred structure used
562 * by SCM_CREDS is different between the two operating system).
563 *
564 * The caller can set it (if necessary) after converting all the
565 * control messages.
566 */
567
568 bhdr->msg_flags = linux_to_bsd_msg_flags(lhdr->msg_flags);
569 return (0);
570 }
571
572 static int
573 bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
574 {
575 lhdr->msg_name = PTROUT(bhdr->msg_name);
576 lhdr->msg_namelen = bhdr->msg_namelen;
577 lhdr->msg_iov = PTROUT(bhdr->msg_iov);
578 lhdr->msg_iovlen = bhdr->msg_iovlen;
579 lhdr->msg_control = PTROUT(bhdr->msg_control);
580
581 /*
582 * msg_controllen is skipped since BSD and LINUX control messages
583 * are potentially different sizes (e.g. the cred structure used
584 * by SCM_CREDS is different between the two operating system).
585 *
586 * The caller can set it (if necessary) after converting all the
587 * control messages.
588 */
589
590 /* msg_flags skipped */
591 return (0);
592 }
593
594 static int
595 linux_set_socket_flags(int lflags, int *flags)
596 {
597
598 if (lflags & ~(LINUX_SOCK_CLOEXEC | LINUX_SOCK_NONBLOCK))
599 return (EINVAL);
600 if (lflags & LINUX_SOCK_NONBLOCK)
601 *flags |= SOCK_NONBLOCK;
602 if (lflags & LINUX_SOCK_CLOEXEC)
603 *flags |= SOCK_CLOEXEC;
604 return (0);
605 }
606
607 static int
608 linux_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
609 struct mbuf *control, enum uio_seg segflg)
610 {
611 struct sockaddr *to;
612 int error;
613
614 if (mp->msg_name != NULL) {
615 error = linux_getsockaddr(&to, mp->msg_name, mp->msg_namelen);
616 if (error)
617 return (error);
618 mp->msg_name = to;
619 } else
620 to = NULL;
621
622 error = kern_sendit(td, s, mp, linux_to_bsd_msg_flags(flags), control,
623 segflg);
624
625 if (to)
626 free(to, M_SONAME);
627 return (error);
628 }
629
630 /* Return 0 if IP_HDRINCL is set for the given socket. */
631 static int
632 linux_check_hdrincl(struct thread *td, int s)
633 {
634 int error, optval;
635 socklen_t size_val;
636
637 size_val = sizeof(optval);
638 error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
639 &optval, UIO_SYSSPACE, &size_val);
640 if (error)
641 return (error);
642
643 return (optval == 0);
644 }
645
646 /*
647 * Updated sendto() when IP_HDRINCL is set:
648 * tweak endian-dependent fields in the IP packet.
649 */
650 static int
651 linux_sendto_hdrincl(struct thread *td, struct linux_sendto_args *linux_args)
652 {
653 /*
654 * linux_ip_copysize defines how many bytes we should copy
655 * from the beginning of the IP packet before we customize it for BSD.
656 * It should include all the fields we modify (ip_len and ip_off).
657 */
658 #define linux_ip_copysize 8
659
660 struct ip *packet;
661 struct msghdr msg;
662 struct iovec aiov[1];
663 int error;
664
665 /* Check that the packet isn't too big or too small. */
666 if (linux_args->len < linux_ip_copysize ||
667 linux_args->len > IP_MAXPACKET)
668 return (EINVAL);
669
670 packet = (struct ip *)malloc(linux_args->len, M_LINUX, M_WAITOK);
671
672 /* Make kernel copy of the packet to be sent */
673 if ((error = copyin(PTRIN(linux_args->msg), packet,
674 linux_args->len)))
675 goto goout;
676
677 /* Convert fields from Linux to BSD raw IP socket format */
678 packet->ip_len = linux_args->len;
679 packet->ip_off = ntohs(packet->ip_off);
680
681 /* Prepare the msghdr and iovec structures describing the new packet */
682 msg.msg_name = PTRIN(linux_args->to);
683 msg.msg_namelen = linux_args->tolen;
684 msg.msg_iov = aiov;
685 msg.msg_iovlen = 1;
686 msg.msg_control = NULL;
687 msg.msg_flags = 0;
688 aiov[0].iov_base = (char *)packet;
689 aiov[0].iov_len = linux_args->len;
690 error = linux_sendit(td, linux_args->s, &msg, linux_args->flags,
691 NULL, UIO_SYSSPACE);
692 goout:
693 free(packet, M_LINUX);
694 return (error);
695 }
696
697 int
698 linux_socket(struct thread *td, struct linux_socket_args *args)
699 {
700 struct socket_args /* {
701 int domain;
702 int type;
703 int protocol;
704 } */ bsd_args;
705 int retval_socket;
706
707 bsd_args.protocol = args->protocol;
708 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
709 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
710 return (EINVAL);
711 retval_socket = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
712 &bsd_args.type);
713 if (retval_socket != 0)
714 return (retval_socket);
715 bsd_args.domain = linux_to_bsd_domain(args->domain);
716 if (bsd_args.domain == -1)
717 return (EAFNOSUPPORT);
718
719 retval_socket = sys_socket(td, &bsd_args);
720 if (retval_socket)
721 return (retval_socket);
722
723 if (bsd_args.type == SOCK_RAW
724 && (bsd_args.protocol == IPPROTO_RAW || bsd_args.protocol == 0)
725 && bsd_args.domain == PF_INET) {
726 /* It's a raw IP socket: set the IP_HDRINCL option. */
727 int hdrincl;
728
729 hdrincl = 1;
730 /* We ignore any error returned by kern_setsockopt() */
731 kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
732 &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
733 }
734 #ifdef INET6
735 /*
736 * Linux AF_INET6 socket has IPV6_V6ONLY setsockopt set to 0 by default
737 * and some apps depend on this. So, set V6ONLY to 0 for Linux apps.
738 * For simplicity we do this unconditionally of the net.inet6.ip6.v6only
739 * sysctl value.
740 */
741 if (bsd_args.domain == PF_INET6) {
742 int v6only;
743
744 v6only = 0;
745 /* We ignore any error returned by setsockopt() */
746 kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
747 &v6only, UIO_SYSSPACE, sizeof(v6only));
748 }
749 #endif
750
751 return (retval_socket);
752 }
753
754 int
755 linux_bind(struct thread *td, struct linux_bind_args *args)
756 {
757 struct sockaddr *sa;
758 int error;
759
760 error = linux_getsockaddr(&sa, PTRIN(args->name),
761 args->namelen);
762 if (error)
763 return (error);
764
765 error = kern_bind(td, args->s, sa);
766 free(sa, M_SONAME);
767 if (error == EADDRNOTAVAIL && args->namelen != sizeof(struct sockaddr_in))
768 return (EINVAL);
769 return (error);
770 }
771
772 int
773 linux_connect(struct thread *td, struct linux_connect_args *args)
774 {
775 cap_rights_t rights;
776 struct socket *so;
777 struct sockaddr *sa;
778 u_int fflag;
779 int error;
780
781 error = linux_getsockaddr(&sa, (struct osockaddr *)PTRIN(args->name),
782 args->namelen);
783 if (error)
784 return (error);
785
786 error = kern_connect(td, args->s, sa);
787 free(sa, M_SONAME);
788 if (error != EISCONN)
789 return (error);
790
791 /*
792 * Linux doesn't return EISCONN the first time it occurs,
793 * when on a non-blocking socket. Instead it returns the
794 * error getsockopt(SOL_SOCKET, SO_ERROR) would return on BSD.
795 *
796 * XXXRW: Instead of using fgetsock(), check that it is a
797 * socket and use the file descriptor reference instead of
798 * creating a new one.
799 */
800 error = fgetsock(td, args->s, cap_rights_init(&rights, CAP_CONNECT),
801 &so, &fflag);
802 if (error == 0) {
803 error = EISCONN;
804 if (fflag & FNONBLOCK) {
805 SOCK_LOCK(so);
806 if (so->so_emuldata == 0)
807 error = so->so_error;
808 so->so_emuldata = (void *)1;
809 SOCK_UNLOCK(so);
810 }
811 fputsock(so);
812 }
813 return (error);
814 }
815
816 int
817 linux_listen(struct thread *td, struct linux_listen_args *args)
818 {
819 struct listen_args /* {
820 int s;
821 int backlog;
822 } */ bsd_args;
823
824 bsd_args.s = args->s;
825 bsd_args.backlog = args->backlog;
826 return (sys_listen(td, &bsd_args));
827 }
828
829 static int
830 linux_accept_common(struct thread *td, int s, l_uintptr_t addr,
831 l_uintptr_t namelen, int flags)
832 {
833 struct accept4_args /* {
834 int s;
835 struct sockaddr * __restrict name;
836 socklen_t * __restrict anamelen;
837 int flags;
838 } */ bsd_args;
839 cap_rights_t rights;
840 struct socket *so;
841 struct file *fp;
842 int error, error1;
843
844 bsd_args.s = s;
845 /* XXX: */
846 bsd_args.name = (struct sockaddr * __restrict)PTRIN(addr);
847 bsd_args.anamelen = PTRIN(namelen);/* XXX */
848 bsd_args.flags = 0;
849 error = linux_set_socket_flags(flags, &bsd_args.flags);
850 if (error != 0)
851 return (error);
852 error = sys_accept4(td, &bsd_args);
853 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.name);
854 if (error) {
855 if (error == EFAULT && namelen != sizeof(struct sockaddr_in))
856 return (EINVAL);
857 if (error == EINVAL) {
858 error1 = getsock_cap(td, s,
859 cap_rights_init(&rights, CAP_ACCEPT), &fp, NULL);
860 if (error1 != 0)
861 return (error1);
862 so = fp->f_data;
863 if (so->so_type == SOCK_DGRAM) {
864 fdrop(fp, td);
865 return (EOPNOTSUPP);
866 }
867 fdrop(fp, td);
868 }
869 return (error);
870 }
871 if (addr)
872 error = linux_sa_put(PTRIN(addr));
873 if (error) {
874 (void)kern_close(td, td->td_retval[0]);
875 td->td_retval[0] = 0;
876 }
877 return (error);
878 }
879
880 int
881 linux_accept(struct thread *td, struct linux_accept_args *args)
882 {
883
884 return (linux_accept_common(td, args->s, args->addr,
885 args->namelen, 0));
886 }
887
888 int
889 linux_accept4(struct thread *td, struct linux_accept4_args *args)
890 {
891
892 return (linux_accept_common(td, args->s, args->addr,
893 args->namelen, args->flags));
894 }
895
896 int
897 linux_getsockname(struct thread *td, struct linux_getsockname_args *args)
898 {
899 struct getsockname_args /* {
900 int fdes;
901 struct sockaddr * __restrict asa;
902 socklen_t * __restrict alen;
903 } */ bsd_args;
904 int error;
905
906 bsd_args.fdes = args->s;
907 /* XXX: */
908 bsd_args.asa = (struct sockaddr * __restrict)PTRIN(args->addr);
909 bsd_args.alen = PTRIN(args->namelen); /* XXX */
910 error = sys_getsockname(td, &bsd_args);
911 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
912 if (error)
913 return (error);
914 return (linux_sa_put(PTRIN(args->addr)));
915 }
916
917 int
918 linux_getpeername(struct thread *td, struct linux_getpeername_args *args)
919 {
920 struct getpeername_args /* {
921 int fdes;
922 caddr_t asa;
923 int *alen;
924 } */ bsd_args;
925 int error;
926
927 bsd_args.fdes = args->s;
928 bsd_args.asa = (struct sockaddr *)PTRIN(args->addr);
929 bsd_args.alen = (socklen_t *)PTRIN(args->namelen);
930 error = sys_getpeername(td, &bsd_args);
931 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.asa);
932 if (error)
933 return (error);
934 return (linux_sa_put(PTRIN(args->addr)));
935 }
936
937 int
938 linux_socketpair(struct thread *td, struct linux_socketpair_args *args)
939 {
940 struct socketpair_args /* {
941 int domain;
942 int type;
943 int protocol;
944 int *rsv;
945 } */ bsd_args;
946 int error;
947
948 bsd_args.domain = linux_to_bsd_domain(args->domain);
949 if (bsd_args.domain != PF_LOCAL)
950 return (EAFNOSUPPORT);
951 bsd_args.type = args->type & LINUX_SOCK_TYPE_MASK;
952 if (bsd_args.type < 0 || bsd_args.type > LINUX_SOCK_MAX)
953 return (EINVAL);
954 error = linux_set_socket_flags(args->type & ~LINUX_SOCK_TYPE_MASK,
955 &bsd_args.type);
956 if (error != 0)
957 return (error);
958 if (args->protocol != 0 && args->protocol != PF_UNIX)
959
960 /*
961 * Use of PF_UNIX as protocol argument is not right,
962 * but Linux does it.
963 * Do not map PF_UNIX as its Linux value is identical
964 * to FreeBSD one.
965 */
966 return (EPROTONOSUPPORT);
967 else
968 bsd_args.protocol = 0;
969 bsd_args.rsv = (int *)PTRIN(args->rsv);
970 return (sys_socketpair(td, &bsd_args));
971 }
972
973 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
974 struct linux_send_args {
975 int s;
976 l_uintptr_t msg;
977 int len;
978 int flags;
979 };
980
981 static int
982 linux_send(struct thread *td, struct linux_send_args *args)
983 {
984 struct sendto_args /* {
985 int s;
986 caddr_t buf;
987 int len;
988 int flags;
989 caddr_t to;
990 int tolen;
991 } */ bsd_args;
992
993 bsd_args.s = args->s;
994 bsd_args.buf = (caddr_t)PTRIN(args->msg);
995 bsd_args.len = args->len;
996 bsd_args.flags = args->flags;
997 bsd_args.to = NULL;
998 bsd_args.tolen = 0;
999 return (sys_sendto(td, &bsd_args));
1000 }
1001
1002 struct linux_recv_args {
1003 int s;
1004 l_uintptr_t msg;
1005 int len;
1006 int flags;
1007 };
1008
1009 static int
1010 linux_recv(struct thread *td, struct linux_recv_args *args)
1011 {
1012 struct recvfrom_args /* {
1013 int s;
1014 caddr_t buf;
1015 int len;
1016 int flags;
1017 struct sockaddr *from;
1018 socklen_t fromlenaddr;
1019 } */ bsd_args;
1020
1021 bsd_args.s = args->s;
1022 bsd_args.buf = (caddr_t)PTRIN(args->msg);
1023 bsd_args.len = args->len;
1024 bsd_args.flags = linux_to_bsd_msg_flags(args->flags);
1025 bsd_args.from = NULL;
1026 bsd_args.fromlenaddr = 0;
1027 return (sys_recvfrom(td, &bsd_args));
1028 }
1029 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1030
1031 int
1032 linux_sendto(struct thread *td, struct linux_sendto_args *args)
1033 {
1034 struct msghdr msg;
1035 struct iovec aiov;
1036
1037 if (linux_check_hdrincl(td, args->s) == 0)
1038 /* IP_HDRINCL set, tweak the packet before sending */
1039 return (linux_sendto_hdrincl(td, args));
1040
1041 msg.msg_name = PTRIN(args->to);
1042 msg.msg_namelen = args->tolen;
1043 msg.msg_iov = &aiov;
1044 msg.msg_iovlen = 1;
1045 msg.msg_control = NULL;
1046 msg.msg_flags = 0;
1047 aiov.iov_base = PTRIN(args->msg);
1048 aiov.iov_len = args->len;
1049 return (linux_sendit(td, args->s, &msg, args->flags, NULL,
1050 UIO_USERSPACE));
1051 }
1052
1053 int
1054 linux_recvfrom(struct thread *td, struct linux_recvfrom_args *args)
1055 {
1056 struct msghdr msg;
1057 struct iovec aiov;
1058 int error, fromlen;
1059
1060 if (PTRIN(args->fromlen) != NULL) {
1061 error = copyin(PTRIN(args->fromlen), &fromlen,
1062 sizeof(fromlen));
1063 if (error != 0)
1064 return (error);
1065 if (fromlen < 0)
1066 return (EINVAL);
1067 msg.msg_namelen = fromlen;
1068 } else
1069 msg.msg_namelen = 0;
1070
1071 msg.msg_name = (struct sockaddr * __restrict)PTRIN(args->from);
1072 msg.msg_iov = &aiov;
1073 msg.msg_iovlen = 1;
1074 aiov.iov_base = PTRIN(args->buf);
1075 aiov.iov_len = args->len;
1076 msg.msg_control = 0;
1077 msg.msg_flags = linux_to_bsd_msg_flags(args->flags);
1078
1079 error = kern_recvit(td, args->s, &msg, UIO_USERSPACE, NULL);
1080 if (error != 0)
1081 return (error);
1082
1083 if (PTRIN(args->from) != NULL) {
1084 error = bsd_to_linux_sockaddr((struct sockaddr *)
1085 PTRIN(args->from));
1086 if (error != 0)
1087 return (error);
1088
1089 error = linux_sa_put((struct osockaddr *)
1090 PTRIN(args->from));
1091 }
1092
1093 if (PTRIN(args->fromlen) != NULL)
1094 error = copyout(&msg.msg_namelen, PTRIN(args->fromlen),
1095 sizeof(msg.msg_namelen));
1096
1097 return (error);
1098 }
1099
1100 static int
1101 linux_sendmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1102 l_uint flags)
1103 {
1104 struct cmsghdr *cmsg;
1105 struct cmsgcred cmcred;
1106 struct mbuf *control;
1107 struct msghdr msg;
1108 struct l_cmsghdr linux_cmsg;
1109 struct l_cmsghdr *ptr_cmsg;
1110 struct l_msghdr linux_msg;
1111 struct iovec *iov;
1112 socklen_t datalen;
1113 struct sockaddr *sa;
1114 sa_family_t sa_family;
1115 void *data;
1116 int error;
1117
1118 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1119 if (error != 0)
1120 return (error);
1121
1122 /*
1123 * Some Linux applications (ping) define a non-NULL control data
1124 * pointer, but a msg_controllen of 0, which is not allowed in the
1125 * FreeBSD system call interface. NULL the msg_control pointer in
1126 * order to handle this case. This should be checked, but allows the
1127 * Linux ping to work.
1128 */
1129 if (PTRIN(linux_msg.msg_control) != NULL && linux_msg.msg_controllen == 0)
1130 linux_msg.msg_control = PTROUT(NULL);
1131
1132 error = linux_to_bsd_msghdr(&msg, &linux_msg);
1133 if (error != 0)
1134 return (error);
1135
1136 #ifdef COMPAT_LINUX32
1137 error = linux32_copyiniov(PTRIN(msg.msg_iov), msg.msg_iovlen,
1138 &iov, EMSGSIZE);
1139 #else
1140 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1141 #endif
1142 if (error != 0)
1143 return (error);
1144
1145 control = NULL;
1146 cmsg = NULL;
1147
1148 if ((ptr_cmsg = LINUX_CMSG_FIRSTHDR(&linux_msg)) != NULL) {
1149 error = kern_getsockname(td, s, &sa, &datalen);
1150 if (error != 0)
1151 goto bad;
1152 sa_family = sa->sa_family;
1153 free(sa, M_SONAME);
1154
1155 error = ENOBUFS;
1156 cmsg = malloc(CMSG_HDRSZ, M_LINUX, M_WAITOK|M_ZERO);
1157 control = m_get(M_WAITOK, MT_CONTROL);
1158
1159 do {
1160 error = copyin(ptr_cmsg, &linux_cmsg,
1161 sizeof(struct l_cmsghdr));
1162 if (error != 0)
1163 goto bad;
1164
1165 error = EINVAL;
1166 if (linux_cmsg.cmsg_len < sizeof(struct l_cmsghdr))
1167 goto bad;
1168
1169 /*
1170 * Now we support only SCM_RIGHTS and SCM_CRED,
1171 * so return EINVAL in any other cmsg_type
1172 */
1173 cmsg->cmsg_type =
1174 linux_to_bsd_cmsg_type(linux_cmsg.cmsg_type);
1175 cmsg->cmsg_level =
1176 linux_to_bsd_sockopt_level(linux_cmsg.cmsg_level);
1177 if (cmsg->cmsg_type == -1
1178 || cmsg->cmsg_level != SOL_SOCKET)
1179 goto bad;
1180
1181 /*
1182 * Some applications (e.g. pulseaudio) attempt to
1183 * send ancillary data even if the underlying protocol
1184 * doesn't support it which is not allowed in the
1185 * FreeBSD system call interface.
1186 */
1187 if (sa_family != AF_UNIX)
1188 continue;
1189
1190 data = LINUX_CMSG_DATA(ptr_cmsg);
1191 datalen = linux_cmsg.cmsg_len - L_CMSG_HDRSZ;
1192
1193 switch (cmsg->cmsg_type)
1194 {
1195 case SCM_RIGHTS:
1196 break;
1197
1198 case SCM_CREDS:
1199 data = &cmcred;
1200 datalen = sizeof(cmcred);
1201
1202 /*
1203 * The lower levels will fill in the structure
1204 */
1205 bzero(data, datalen);
1206 break;
1207 }
1208
1209 cmsg->cmsg_len = CMSG_LEN(datalen);
1210
1211 error = ENOBUFS;
1212 if (!m_append(control, CMSG_HDRSZ, (c_caddr_t)cmsg))
1213 goto bad;
1214 if (!m_append(control, datalen, (c_caddr_t)data))
1215 goto bad;
1216 } while ((ptr_cmsg = LINUX_CMSG_NXTHDR(&linux_msg, ptr_cmsg)));
1217
1218 if (m_length(control, NULL) == 0) {
1219 m_freem(control);
1220 control = NULL;
1221 }
1222 }
1223
1224 msg.msg_iov = iov;
1225 msg.msg_flags = 0;
1226 error = linux_sendit(td, s, &msg, flags, control, UIO_USERSPACE);
1227 control = NULL;
1228
1229 bad:
1230 m_freem(control);
1231 free(iov, M_IOV);
1232 if (cmsg)
1233 free(cmsg, M_LINUX);
1234 return (error);
1235 }
1236
1237 int
1238 linux_sendmsg(struct thread *td, struct linux_sendmsg_args *args)
1239 {
1240
1241 return (linux_sendmsg_common(td, args->s, PTRIN(args->msg),
1242 args->flags));
1243 }
1244
1245 int
1246 linux_sendmmsg(struct thread *td, struct linux_sendmmsg_args *args)
1247 {
1248 struct l_mmsghdr *msg;
1249 l_uint retval;
1250 int error, datagrams;
1251
1252 if (args->vlen > UIO_MAXIOV)
1253 args->vlen = UIO_MAXIOV;
1254
1255 msg = PTRIN(args->msg);
1256 datagrams = 0;
1257 while (datagrams < args->vlen) {
1258 error = linux_sendmsg_common(td, args->s, &msg->msg_hdr,
1259 args->flags);
1260 if (error != 0)
1261 break;
1262
1263 retval = td->td_retval[0];
1264 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1265 if (error != 0)
1266 break;
1267 ++msg;
1268 ++datagrams;
1269 }
1270 if (error == 0)
1271 td->td_retval[0] = datagrams;
1272 return (error);
1273 }
1274
1275 static int
1276 linux_recvmsg_common(struct thread *td, l_int s, struct l_msghdr *msghdr,
1277 l_uint flags, struct msghdr *msg)
1278 {
1279 struct cmsghdr *cm;
1280 struct cmsgcred *cmcred;
1281 struct l_cmsghdr *linux_cmsg = NULL;
1282 struct l_ucred linux_ucred;
1283 socklen_t datalen, outlen;
1284 struct l_msghdr linux_msg;
1285 struct iovec *iov, *uiov;
1286 struct mbuf *control = NULL;
1287 struct mbuf **controlp;
1288 struct timeval *ftmvl;
1289 l_timeval ltmvl;
1290 caddr_t outbuf;
1291 void *data;
1292 int error, i, fd, fds, *fdp;
1293
1294 error = copyin(msghdr, &linux_msg, sizeof(linux_msg));
1295 if (error != 0)
1296 return (error);
1297
1298 error = linux_to_bsd_msghdr(msg, &linux_msg);
1299 if (error != 0)
1300 return (error);
1301
1302 #ifdef COMPAT_LINUX32
1303 error = linux32_copyiniov(PTRIN(msg->msg_iov), msg->msg_iovlen,
1304 &iov, EMSGSIZE);
1305 #else
1306 error = copyiniov(msg->msg_iov, msg->msg_iovlen, &iov, EMSGSIZE);
1307 #endif
1308 if (error != 0)
1309 return (error);
1310
1311 if (msg->msg_name) {
1312 error = linux_to_bsd_sockaddr((struct sockaddr *)msg->msg_name,
1313 msg->msg_namelen);
1314 if (error != 0)
1315 goto bad;
1316 }
1317
1318 uiov = msg->msg_iov;
1319 msg->msg_iov = iov;
1320 controlp = (msg->msg_control != NULL) ? &control : NULL;
1321 error = kern_recvit(td, s, msg, UIO_USERSPACE, controlp);
1322 msg->msg_iov = uiov;
1323 if (error != 0)
1324 goto bad;
1325
1326 error = bsd_to_linux_msghdr(msg, &linux_msg);
1327 if (error != 0)
1328 goto bad;
1329
1330 if (linux_msg.msg_name) {
1331 error = bsd_to_linux_sockaddr((struct sockaddr *)
1332 PTRIN(linux_msg.msg_name));
1333 if (error != 0)
1334 goto bad;
1335 }
1336 if (linux_msg.msg_name && linux_msg.msg_namelen > 2) {
1337 error = linux_sa_put(PTRIN(linux_msg.msg_name));
1338 if (error != 0)
1339 goto bad;
1340 }
1341
1342 outbuf = PTRIN(linux_msg.msg_control);
1343 outlen = 0;
1344
1345 if (control) {
1346 linux_cmsg = malloc(L_CMSG_HDRSZ, M_LINUX, M_WAITOK | M_ZERO);
1347
1348 msg->msg_control = mtod(control, struct cmsghdr *);
1349 msg->msg_controllen = control->m_len;
1350
1351 cm = CMSG_FIRSTHDR(msg);
1352
1353 while (cm != NULL) {
1354 linux_cmsg->cmsg_type =
1355 bsd_to_linux_cmsg_type(cm->cmsg_type);
1356 linux_cmsg->cmsg_level =
1357 bsd_to_linux_sockopt_level(cm->cmsg_level);
1358 if (linux_cmsg->cmsg_type == -1
1359 || cm->cmsg_level != SOL_SOCKET)
1360 {
1361 error = EINVAL;
1362 goto bad;
1363 }
1364
1365 data = CMSG_DATA(cm);
1366 datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data;
1367
1368 switch (cm->cmsg_type)
1369 {
1370 case SCM_RIGHTS:
1371 if (flags & LINUX_MSG_CMSG_CLOEXEC) {
1372 fds = datalen / sizeof(int);
1373 fdp = data;
1374 for (i = 0; i < fds; i++) {
1375 fd = *fdp++;
1376 (void)kern_fcntl(td, fd,
1377 F_SETFD, FD_CLOEXEC);
1378 }
1379 }
1380 break;
1381
1382 case SCM_CREDS:
1383 /*
1384 * Currently LOCAL_CREDS is never in
1385 * effect for Linux so no need to worry
1386 * about sockcred
1387 */
1388 if (datalen != sizeof(*cmcred)) {
1389 error = EMSGSIZE;
1390 goto bad;
1391 }
1392 cmcred = (struct cmsgcred *)data;
1393 bzero(&linux_ucred, sizeof(linux_ucred));
1394 linux_ucred.pid = cmcred->cmcred_pid;
1395 linux_ucred.uid = cmcred->cmcred_uid;
1396 linux_ucred.gid = cmcred->cmcred_gid;
1397 data = &linux_ucred;
1398 datalen = sizeof(linux_ucred);
1399 break;
1400
1401 case SCM_TIMESTAMP:
1402 if (datalen != sizeof(struct timeval)) {
1403 error = EMSGSIZE;
1404 goto bad;
1405 }
1406 ftmvl = (struct timeval *)data;
1407 ltmvl.tv_sec = ftmvl->tv_sec;
1408 ltmvl.tv_usec = ftmvl->tv_usec;
1409 data = <mvl;
1410 datalen = sizeof(ltmvl);
1411 break;
1412 }
1413
1414 if (outlen + LINUX_CMSG_LEN(datalen) >
1415 linux_msg.msg_controllen) {
1416 if (outlen == 0) {
1417 error = EMSGSIZE;
1418 goto bad;
1419 } else {
1420 linux_msg.msg_flags |=
1421 LINUX_MSG_CTRUNC;
1422 goto out;
1423 }
1424 }
1425
1426 linux_cmsg->cmsg_len = LINUX_CMSG_LEN(datalen);
1427
1428 error = copyout(linux_cmsg, outbuf, L_CMSG_HDRSZ);
1429 if (error)
1430 goto bad;
1431 outbuf += L_CMSG_HDRSZ;
1432
1433 error = copyout(data, outbuf, datalen);
1434 if (error)
1435 goto bad;
1436
1437 outbuf += LINUX_CMSG_ALIGN(datalen);
1438 outlen += LINUX_CMSG_LEN(datalen);
1439
1440 cm = CMSG_NXTHDR(msg, cm);
1441 }
1442 }
1443
1444 out:
1445 linux_msg.msg_controllen = outlen;
1446 error = copyout(&linux_msg, msghdr, sizeof(linux_msg));
1447
1448 bad:
1449 free(iov, M_IOV);
1450 m_freem(control);
1451 free(linux_cmsg, M_LINUX);
1452
1453 return (error);
1454 }
1455
1456 int
1457 linux_recvmsg(struct thread *td, struct linux_recvmsg_args *args)
1458 {
1459 struct msghdr bsd_msg;
1460
1461 return (linux_recvmsg_common(td, args->s, PTRIN(args->msg),
1462 args->flags, &bsd_msg));
1463 }
1464
1465 int
1466 linux_recvmmsg(struct thread *td, struct linux_recvmmsg_args *args)
1467 {
1468 struct l_mmsghdr *msg;
1469 struct msghdr bsd_msg;
1470 struct l_timespec lts;
1471 struct timespec ts, tts;
1472 l_uint retval;
1473 int error, datagrams;
1474
1475 if (args->timeout) {
1476 error = copyin(args->timeout, <s, sizeof(struct l_timespec));
1477 if (error != 0)
1478 return (error);
1479 error = linux_to_native_timespec(&ts, <s);
1480 if (error != 0)
1481 return (error);
1482 getnanotime(&tts);
1483 timespecadd(&tts, &ts);
1484 }
1485
1486 msg = PTRIN(args->msg);
1487 datagrams = 0;
1488 while (datagrams < args->vlen) {
1489 error = linux_recvmsg_common(td, args->s, &msg->msg_hdr,
1490 args->flags & ~LINUX_MSG_WAITFORONE, &bsd_msg);
1491 if (error != 0)
1492 break;
1493
1494 retval = td->td_retval[0];
1495 error = copyout(&retval, &msg->msg_len, sizeof(msg->msg_len));
1496 if (error != 0)
1497 break;
1498 ++msg;
1499 ++datagrams;
1500
1501 /*
1502 * MSG_WAITFORONE turns on MSG_DONTWAIT after one packet.
1503 */
1504 if (args->flags & LINUX_MSG_WAITFORONE)
1505 args->flags |= LINUX_MSG_DONTWAIT;
1506
1507 /*
1508 * See BUGS section of recvmmsg(2).
1509 */
1510 if (args->timeout) {
1511 getnanotime(&ts);
1512 timespecsub(&ts, &tts);
1513 if (!timespecisset(&ts) || ts.tv_sec > 0)
1514 break;
1515 }
1516 /* Out of band data, return right away. */
1517 if (bsd_msg.msg_flags & MSG_OOB)
1518 break;
1519 }
1520 if (error == 0)
1521 td->td_retval[0] = datagrams;
1522 return (error);
1523 }
1524
1525 int
1526 linux_shutdown(struct thread *td, struct linux_shutdown_args *args)
1527 {
1528 struct shutdown_args /* {
1529 int s;
1530 int how;
1531 } */ bsd_args;
1532
1533 bsd_args.s = args->s;
1534 bsd_args.how = args->how;
1535 return (sys_shutdown(td, &bsd_args));
1536 }
1537
1538 int
1539 linux_setsockopt(struct thread *td, struct linux_setsockopt_args *args)
1540 {
1541 struct setsockopt_args /* {
1542 int s;
1543 int level;
1544 int name;
1545 caddr_t val;
1546 int valsize;
1547 } */ bsd_args;
1548 l_timeval linux_tv;
1549 struct timeval tv;
1550 int error, name;
1551
1552 bsd_args.s = args->s;
1553 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1554 switch (bsd_args.level) {
1555 case SOL_SOCKET:
1556 name = linux_to_bsd_so_sockopt(args->optname);
1557 switch (name) {
1558 case SO_RCVTIMEO:
1559 /* FALLTHROUGH */
1560 case SO_SNDTIMEO:
1561 error = copyin(PTRIN(args->optval), &linux_tv,
1562 sizeof(linux_tv));
1563 if (error)
1564 return (error);
1565 tv.tv_sec = linux_tv.tv_sec;
1566 tv.tv_usec = linux_tv.tv_usec;
1567 return (kern_setsockopt(td, args->s, bsd_args.level,
1568 name, &tv, UIO_SYSSPACE, sizeof(tv)));
1569 /* NOTREACHED */
1570 break;
1571 default:
1572 break;
1573 }
1574 break;
1575 case IPPROTO_IP:
1576 name = linux_to_bsd_ip_sockopt(args->optname);
1577 break;
1578 case IPPROTO_IPV6:
1579 name = linux_to_bsd_ip6_sockopt(args->optname);
1580 break;
1581 case IPPROTO_TCP:
1582 name = linux_to_bsd_tcp_sockopt(args->optname);
1583 break;
1584 default:
1585 name = -1;
1586 break;
1587 }
1588 if (name == -1)
1589 return (ENOPROTOOPT);
1590
1591 bsd_args.name = name;
1592 bsd_args.val = PTRIN(args->optval);
1593 bsd_args.valsize = args->optlen;
1594
1595 if (name == IPV6_NEXTHOP) {
1596 linux_to_bsd_sockaddr((struct sockaddr *)bsd_args.val,
1597 bsd_args.valsize);
1598 error = sys_setsockopt(td, &bsd_args);
1599 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1600 } else
1601 error = sys_setsockopt(td, &bsd_args);
1602
1603 return (error);
1604 }
1605
1606 int
1607 linux_getsockopt(struct thread *td, struct linux_getsockopt_args *args)
1608 {
1609 struct getsockopt_args /* {
1610 int s;
1611 int level;
1612 int name;
1613 caddr_t val;
1614 int *avalsize;
1615 } */ bsd_args;
1616 l_timeval linux_tv;
1617 struct timeval tv;
1618 socklen_t tv_len, xulen, len;
1619 struct xucred xu;
1620 struct l_ucred lxu;
1621 int error, name, newval;
1622
1623 bsd_args.s = args->s;
1624 bsd_args.level = linux_to_bsd_sockopt_level(args->level);
1625 switch (bsd_args.level) {
1626 case SOL_SOCKET:
1627 name = linux_to_bsd_so_sockopt(args->optname);
1628 switch (name) {
1629 case SO_RCVTIMEO:
1630 /* FALLTHROUGH */
1631 case SO_SNDTIMEO:
1632 tv_len = sizeof(tv);
1633 error = kern_getsockopt(td, args->s, bsd_args.level,
1634 name, &tv, UIO_SYSSPACE, &tv_len);
1635 if (error)
1636 return (error);
1637 linux_tv.tv_sec = tv.tv_sec;
1638 linux_tv.tv_usec = tv.tv_usec;
1639 return (copyout(&linux_tv, PTRIN(args->optval),
1640 sizeof(linux_tv)));
1641 /* NOTREACHED */
1642 break;
1643 case LOCAL_PEERCRED:
1644 if (args->optlen != sizeof(lxu))
1645 return (EINVAL);
1646 xulen = sizeof(xu);
1647 error = kern_getsockopt(td, args->s, bsd_args.level,
1648 name, &xu, UIO_SYSSPACE, &xulen);
1649 if (error)
1650 return (error);
1651 /*
1652 * XXX Use 0 for pid as the FreeBSD does not cache peer pid.
1653 */
1654 lxu.pid = 0;
1655 lxu.uid = xu.cr_uid;
1656 lxu.gid = xu.cr_gid;
1657 return (copyout(&lxu, PTRIN(args->optval), sizeof(lxu)));
1658 /* NOTREACHED */
1659 break;
1660 case SO_ERROR:
1661 len = sizeof(newval);
1662 error = kern_getsockopt(td, args->s, bsd_args.level,
1663 name, &newval, UIO_SYSSPACE, &len);
1664 if (error)
1665 return (error);
1666 newval = -SV_ABI_ERRNO(td->td_proc, newval);
1667 return (copyout(&newval, PTRIN(args->optval), len));
1668 /* NOTREACHED */
1669 default:
1670 break;
1671 }
1672 break;
1673 case IPPROTO_IP:
1674 name = linux_to_bsd_ip_sockopt(args->optname);
1675 break;
1676 case IPPROTO_IPV6:
1677 name = linux_to_bsd_ip6_sockopt(args->optname);
1678 break;
1679 case IPPROTO_TCP:
1680 name = linux_to_bsd_tcp_sockopt(args->optname);
1681 break;
1682 default:
1683 name = -1;
1684 break;
1685 }
1686 if (name == -1)
1687 return (EINVAL);
1688
1689 bsd_args.name = name;
1690 bsd_args.val = PTRIN(args->optval);
1691 bsd_args.avalsize = PTRIN(args->optlen);
1692
1693 if (name == IPV6_NEXTHOP) {
1694 error = sys_getsockopt(td, &bsd_args);
1695 bsd_to_linux_sockaddr((struct sockaddr *)bsd_args.val);
1696 } else
1697 error = sys_getsockopt(td, &bsd_args);
1698
1699 return (error);
1700 }
1701
1702 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1703
1704 /* Argument list sizes for linux_socketcall */
1705
1706 #define LINUX_AL(x) ((x) * sizeof(l_ulong))
1707
1708 static const unsigned char lxs_args[] = {
1709 LINUX_AL(0) /* unused*/, LINUX_AL(3) /* socket */,
1710 LINUX_AL(3) /* bind */, LINUX_AL(3) /* connect */,
1711 LINUX_AL(2) /* listen */, LINUX_AL(3) /* accept */,
1712 LINUX_AL(3) /* getsockname */, LINUX_AL(3) /* getpeername */,
1713 LINUX_AL(4) /* socketpair */, LINUX_AL(4) /* send */,
1714 LINUX_AL(4) /* recv */, LINUX_AL(6) /* sendto */,
1715 LINUX_AL(6) /* recvfrom */, LINUX_AL(2) /* shutdown */,
1716 LINUX_AL(5) /* setsockopt */, LINUX_AL(5) /* getsockopt */,
1717 LINUX_AL(3) /* sendmsg */, LINUX_AL(3) /* recvmsg */,
1718 LINUX_AL(4) /* accept4 */, LINUX_AL(5) /* recvmmsg */,
1719 LINUX_AL(4) /* sendmmsg */
1720 };
1721
1722 #define LINUX_AL_SIZE sizeof(lxs_args) / sizeof(lxs_args[0]) - 1
1723
1724 int
1725 linux_socketcall(struct thread *td, struct linux_socketcall_args *args)
1726 {
1727 l_ulong a[6];
1728 void *arg;
1729 int error;
1730
1731 if (args->what < LINUX_SOCKET || args->what > LINUX_AL_SIZE)
1732 return (EINVAL);
1733 error = copyin(PTRIN(args->args), a, lxs_args[args->what]);
1734 if (error)
1735 return (error);
1736
1737 arg = a;
1738 switch (args->what) {
1739 case LINUX_SOCKET:
1740 return (linux_socket(td, arg));
1741 case LINUX_BIND:
1742 return (linux_bind(td, arg));
1743 case LINUX_CONNECT:
1744 return (linux_connect(td, arg));
1745 case LINUX_LISTEN:
1746 return (linux_listen(td, arg));
1747 case LINUX_ACCEPT:
1748 return (linux_accept(td, arg));
1749 case LINUX_GETSOCKNAME:
1750 return (linux_getsockname(td, arg));
1751 case LINUX_GETPEERNAME:
1752 return (linux_getpeername(td, arg));
1753 case LINUX_SOCKETPAIR:
1754 return (linux_socketpair(td, arg));
1755 case LINUX_SEND:
1756 return (linux_send(td, arg));
1757 case LINUX_RECV:
1758 return (linux_recv(td, arg));
1759 case LINUX_SENDTO:
1760 return (linux_sendto(td, arg));
1761 case LINUX_RECVFROM:
1762 return (linux_recvfrom(td, arg));
1763 case LINUX_SHUTDOWN:
1764 return (linux_shutdown(td, arg));
1765 case LINUX_SETSOCKOPT:
1766 return (linux_setsockopt(td, arg));
1767 case LINUX_GETSOCKOPT:
1768 return (linux_getsockopt(td, arg));
1769 case LINUX_SENDMSG:
1770 return (linux_sendmsg(td, arg));
1771 case LINUX_RECVMSG:
1772 return (linux_recvmsg(td, arg));
1773 case LINUX_ACCEPT4:
1774 return (linux_accept4(td, arg));
1775 case LINUX_RECVMMSG:
1776 return (linux_recvmmsg(td, arg));
1777 case LINUX_SENDMMSG:
1778 return (linux_sendmmsg(td, arg));
1779 }
1780
1781 uprintf("LINUX: 'socket' typ=%d not implemented\n", args->what);
1782 return (ENOSYS);
1783 }
1784 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
Cache object: efd6bd67376cecdd65787349ab5cfa60
|