xref: /freebsd/sys/compat/linux/linux.c (revision af23369a6deaaeb612ab266eb88b8bb8d560c322)
1 /*-
2  * Copyright (c) 2015 Dmitry Chagin <dchagin@FreeBSD.org>
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  * 1. Redistributions of source code must retain the above copyright
8  *    notice, this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright
10  *    notice, this list of conditions and the following disclaimer in the
11  *    documentation and/or other materials provided with the distribution.
12  *
13  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23  * SUCH DAMAGE.
24  */
25 
26 #include <sys/cdefs.h>
27 __FBSDID("$FreeBSD$");
28 
29 #include "opt_inet6.h"
30 
31 #include <sys/param.h>
32 #include <sys/conf.h>
33 #include <sys/ctype.h>
34 #include <sys/file.h>
35 #include <sys/filedesc.h>
36 #include <sys/jail.h>
37 #include <sys/lock.h>
38 #include <sys/malloc.h>
39 #include <sys/poll.h>
40 #include <sys/proc.h>
41 #include <sys/signalvar.h>
42 #include <sys/socket.h>
43 #include <sys/socketvar.h>
44 
45 #include <net/if.h>
46 #include <net/if_var.h>
47 #include <net/if_dl.h>
48 #include <net/if_types.h>
49 #include <netlink/netlink.h>
50 
51 #include <sys/un.h>
52 #include <netinet/in.h>
53 
54 #include <compat/linux/linux.h>
55 #include <compat/linux/linux_common.h>
56 #include <compat/linux/linux_mib.h>
57 #include <compat/linux/linux_util.h>
58 
59 CTASSERT(LINUX_IFNAMSIZ == IFNAMSIZ);
60 
61 static bool use_real_ifnames = false;
62 SYSCTL_BOOL(_compat_linux, OID_AUTO, use_real_ifnames, CTLFLAG_RWTUN,
63     &use_real_ifnames, 0,
64     "Use FreeBSD interface names instead of generating ethN aliases");
65 
66 static int bsd_to_linux_sigtbl[LINUX_SIGTBLSZ] = {
67 	LINUX_SIGHUP,	/* SIGHUP */
68 	LINUX_SIGINT,	/* SIGINT */
69 	LINUX_SIGQUIT,	/* SIGQUIT */
70 	LINUX_SIGILL,	/* SIGILL */
71 	LINUX_SIGTRAP,	/* SIGTRAP */
72 	LINUX_SIGABRT,	/* SIGABRT */
73 	0,		/* SIGEMT */
74 	LINUX_SIGFPE,	/* SIGFPE */
75 	LINUX_SIGKILL,	/* SIGKILL */
76 	LINUX_SIGBUS,	/* SIGBUS */
77 	LINUX_SIGSEGV,	/* SIGSEGV */
78 	LINUX_SIGSYS,	/* SIGSYS */
79 	LINUX_SIGPIPE,	/* SIGPIPE */
80 	LINUX_SIGALRM,	/* SIGALRM */
81 	LINUX_SIGTERM,	/* SIGTERM */
82 	LINUX_SIGURG,	/* SIGURG */
83 	LINUX_SIGSTOP,	/* SIGSTOP */
84 	LINUX_SIGTSTP,	/* SIGTSTP */
85 	LINUX_SIGCONT,	/* SIGCONT */
86 	LINUX_SIGCHLD,	/* SIGCHLD */
87 	LINUX_SIGTTIN,	/* SIGTTIN */
88 	LINUX_SIGTTOU,	/* SIGTTOU */
89 	LINUX_SIGIO,	/* SIGIO */
90 	LINUX_SIGXCPU,	/* SIGXCPU */
91 	LINUX_SIGXFSZ,	/* SIGXFSZ */
92 	LINUX_SIGVTALRM,/* SIGVTALRM */
93 	LINUX_SIGPROF,	/* SIGPROF */
94 	LINUX_SIGWINCH,	/* SIGWINCH */
95 	0,		/* SIGINFO */
96 	LINUX_SIGUSR1,	/* SIGUSR1 */
97 	LINUX_SIGUSR2	/* SIGUSR2 */
98 };
99 
100 #define	LINUX_SIGPWREMU	(SIGRTMIN + (LINUX_SIGRTMAX - LINUX_SIGRTMIN) + 1)
101 
102 static int linux_to_bsd_sigtbl[LINUX_SIGTBLSZ] = {
103 	SIGHUP,		/* LINUX_SIGHUP */
104 	SIGINT,		/* LINUX_SIGINT */
105 	SIGQUIT,	/* LINUX_SIGQUIT */
106 	SIGILL,		/* LINUX_SIGILL */
107 	SIGTRAP,	/* LINUX_SIGTRAP */
108 	SIGABRT,	/* LINUX_SIGABRT */
109 	SIGBUS,		/* LINUX_SIGBUS */
110 	SIGFPE,		/* LINUX_SIGFPE */
111 	SIGKILL,	/* LINUX_SIGKILL */
112 	SIGUSR1,	/* LINUX_SIGUSR1 */
113 	SIGSEGV,	/* LINUX_SIGSEGV */
114 	SIGUSR2,	/* LINUX_SIGUSR2 */
115 	SIGPIPE,	/* LINUX_SIGPIPE */
116 	SIGALRM,	/* LINUX_SIGALRM */
117 	SIGTERM,	/* LINUX_SIGTERM */
118 	SIGBUS,		/* LINUX_SIGSTKFLT */
119 	SIGCHLD,	/* LINUX_SIGCHLD */
120 	SIGCONT,	/* LINUX_SIGCONT */
121 	SIGSTOP,	/* LINUX_SIGSTOP */
122 	SIGTSTP,	/* LINUX_SIGTSTP */
123 	SIGTTIN,	/* LINUX_SIGTTIN */
124 	SIGTTOU,	/* LINUX_SIGTTOU */
125 	SIGURG,		/* LINUX_SIGURG */
126 	SIGXCPU,	/* LINUX_SIGXCPU */
127 	SIGXFSZ,	/* LINUX_SIGXFSZ */
128 	SIGVTALRM,	/* LINUX_SIGVTALARM */
129 	SIGPROF,	/* LINUX_SIGPROF */
130 	SIGWINCH,	/* LINUX_SIGWINCH */
131 	SIGIO,		/* LINUX_SIGIO */
132 	/*
133 	 * FreeBSD does not have SIGPWR signal, map Linux SIGPWR signal
134 	 * to the first unused FreeBSD signal number. Since Linux supports
135 	 * signals from 1 to 64 we are ok here as our SIGRTMIN = 65.
136 	 */
137 	LINUX_SIGPWREMU,/* LINUX_SIGPWR */
138 	SIGSYS		/* LINUX_SIGSYS */
139 };
140 
141 static struct cdev *dev_shm_cdev;
142 static struct cdevsw dev_shm_cdevsw = {
143      .d_version = D_VERSION,
144      .d_name    = "dev_shm",
145 };
146 
147 /*
148  * Map Linux RT signals to the FreeBSD RT signals.
149  */
150 static inline int
151 linux_to_bsd_rt_signal(int sig)
152 {
153 
154 	return (SIGRTMIN + sig - LINUX_SIGRTMIN);
155 }
156 
157 static inline int
158 bsd_to_linux_rt_signal(int sig)
159 {
160 
161 	return (sig - SIGRTMIN + LINUX_SIGRTMIN);
162 }
163 
164 int
165 linux_to_bsd_signal(int sig)
166 {
167 
168 	KASSERT(sig > 0 && sig <= LINUX_SIGRTMAX, ("invalid Linux signal %d\n", sig));
169 
170 	if (sig < LINUX_SIGRTMIN)
171 		return (linux_to_bsd_sigtbl[_SIG_IDX(sig)]);
172 
173 	return (linux_to_bsd_rt_signal(sig));
174 }
175 
176 int
177 bsd_to_linux_signal(int sig)
178 {
179 
180 	if (sig <= LINUX_SIGTBLSZ)
181 		return (bsd_to_linux_sigtbl[_SIG_IDX(sig)]);
182 	if (sig == LINUX_SIGPWREMU)
183 		return (LINUX_SIGPWR);
184 
185 	return (bsd_to_linux_rt_signal(sig));
186 }
187 
188 int
189 linux_to_bsd_sigaltstack(int lsa)
190 {
191 	int bsa = 0;
192 
193 	if (lsa & LINUX_SS_DISABLE)
194 		bsa |= SS_DISABLE;
195 	/*
196 	 * Linux ignores SS_ONSTACK flag for ss
197 	 * parameter while FreeBSD prohibits it.
198 	 */
199 	return (bsa);
200 }
201 
202 int
203 bsd_to_linux_sigaltstack(int bsa)
204 {
205 	int lsa = 0;
206 
207 	if (bsa & SS_DISABLE)
208 		lsa |= LINUX_SS_DISABLE;
209 	if (bsa & SS_ONSTACK)
210 		lsa |= LINUX_SS_ONSTACK;
211 	return (lsa);
212 }
213 
214 void
215 linux_to_bsd_sigset(l_sigset_t *lss, sigset_t *bss)
216 {
217 	int b, l;
218 
219 	SIGEMPTYSET(*bss);
220 	for (l = 1; l <= LINUX_SIGRTMAX; l++) {
221 		if (LINUX_SIGISMEMBER(*lss, l)) {
222 			b = linux_to_bsd_signal(l);
223 			if (b)
224 				SIGADDSET(*bss, b);
225 		}
226 	}
227 }
228 
229 void
230 bsd_to_linux_sigset(sigset_t *bss, l_sigset_t *lss)
231 {
232 	int b, l;
233 
234 	LINUX_SIGEMPTYSET(*lss);
235 	for (b = 1; b <= SIGRTMAX; b++) {
236 		if (SIGISMEMBER(*bss, b)) {
237 			l = bsd_to_linux_signal(b);
238 			if (l)
239 				LINUX_SIGADDSET(*lss, l);
240 		}
241 	}
242 }
243 
244 /*
245  * Translate a FreeBSD interface name to a Linux interface name
246  * by interface name, and return the number of bytes copied to lxname.
247  */
248 int
249 ifname_bsd_to_linux_name(const char *bsdname, char *lxname, size_t len)
250 {
251 	struct epoch_tracker et;
252 	struct ifnet *ifp;
253 	int ret;
254 
255 	ret = 0;
256 	CURVNET_SET(TD_TO_VNET(curthread));
257 	NET_EPOCH_ENTER(et);
258 	ifp = ifunit(bsdname);
259 	if (ifp != NULL)
260 		ret = ifname_bsd_to_linux_ifp(ifp, lxname, len);
261 	NET_EPOCH_EXIT(et);
262 	CURVNET_RESTORE();
263 	return (ret);
264 }
265 
266 /*
267  * Translate a FreeBSD interface name to a Linux interface name
268  * by interface index, and return the number of bytes copied to lxname.
269  */
270 int
271 ifname_bsd_to_linux_idx(u_int idx, char *lxname, size_t len)
272 {
273 	struct epoch_tracker et;
274 	struct ifnet *ifp;
275 	int ret;
276 
277 	ret = 0;
278 	CURVNET_SET(TD_TO_VNET(curthread));
279 	NET_EPOCH_ENTER(et);
280 	ifp = ifnet_byindex(idx);
281 	if (ifp != NULL)
282 		ret = ifname_bsd_to_linux_ifp(ifp, lxname, len);
283 	NET_EPOCH_EXIT(et);
284 	CURVNET_RESTORE();
285 	return (ret);
286 }
287 
288 /*
289  * Translate a FreeBSD interface name to a Linux interface name,
290  * and return the number of bytes copied to lxname.
291  */
292 int
293 ifname_bsd_to_linux_ifp(struct ifnet *ifp, char *lxname, size_t len)
294 {
295 	struct ifnet *ifscan;
296 	int unit;
297 
298 	NET_EPOCH_ASSERT();
299 
300 	/*
301 	 * Linux loopback interface name is lo (not lo0),
302 	 * we translate lo to lo0, loX to loX.
303 	 */
304 	if (IFP_IS_LOOP(ifp) && strncmp(ifp->if_xname, "lo0", IFNAMSIZ) == 0)
305 		return (strlcpy(lxname, "lo", len));
306 
307 	/* Short-circuit non ethernet interfaces. */
308 	if (!IFP_IS_ETH(ifp) || linux_use_real_ifname(ifp))
309 		return (strlcpy(lxname, ifp->if_xname, len));
310 
311 	/* Determine the (relative) unit number for ethernet interfaces. */
312 	unit = 0;
313 	CK_STAILQ_FOREACH(ifscan, &V_ifnet, if_link) {
314 		if (ifscan == ifp)
315 			return (snprintf(lxname, len, "eth%d", unit));
316 		if (IFP_IS_ETH(ifscan))
317 			unit++;
318 	}
319 	return (0);
320 }
321 
322 /*
323  * Translate a Linux interface name to a FreeBSD interface name,
324  * and return the associated ifnet structure
325  * bsdname and lxname need to be least IFNAMSIZ bytes long, but
326  * can point to the same buffer.
327  */
328 struct ifnet *
329 ifname_linux_to_bsd(struct thread *td, const char *lxname, char *bsdname)
330 {
331 	struct ifnet *ifp;
332 	int len, unit;
333 	char *ep;
334 	int index;
335 	bool is_eth, is_lo;
336 
337 	for (len = 0; len < LINUX_IFNAMSIZ; ++len)
338 		if (!isalpha(lxname[len]) || lxname[len] == '\0')
339 			break;
340 	if (len == 0 || len == LINUX_IFNAMSIZ)
341 		return (NULL);
342 	/*
343 	 * Linux loopback interface name is lo (not lo0),
344 	 * we translate lo to lo0, loX to loX.
345 	 */
346 	is_lo = (len == 2 && strncmp(lxname, "lo", LINUX_IFNAMSIZ) == 0);
347 	unit = (int)strtoul(lxname + len, &ep, 10);
348 	if ((ep == NULL || ep == lxname + len || ep >= lxname + LINUX_IFNAMSIZ) &&
349 	    is_lo == 0)
350 		return (NULL);
351 	index = 0;
352 	is_eth = (len == 3 && strncmp(lxname, "eth", len) == 0);
353 
354 	CURVNET_SET(TD_TO_VNET(td));
355 	IFNET_RLOCK();
356 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
357 		/*
358 		 * Allow Linux programs to use FreeBSD names. Don't presume
359 		 * we never have an interface named "eth", so don't make
360 		 * the test optional based on is_eth.
361 		 */
362 		if (strncmp(ifp->if_xname, lxname, LINUX_IFNAMSIZ) == 0)
363 			break;
364 		if (is_eth && IFP_IS_ETH(ifp) && unit == index++)
365 			break;
366 		if (is_lo && IFP_IS_LOOP(ifp))
367 			break;
368 	}
369 	IFNET_RUNLOCK();
370 	CURVNET_RESTORE();
371 	if (ifp != NULL && bsdname != NULL)
372 		strlcpy(bsdname, ifp->if_xname, IFNAMSIZ);
373 	return (ifp);
374 }
375 
376 void
377 linux_ifflags(struct ifnet *ifp, short *flags)
378 {
379 	unsigned short fl;
380 
381 	fl = (ifp->if_flags | ifp->if_drv_flags) & 0xffff;
382 	*flags = 0;
383 	if (fl & IFF_UP)
384 		*flags |= LINUX_IFF_UP;
385 	if (fl & IFF_BROADCAST)
386 		*flags |= LINUX_IFF_BROADCAST;
387 	if (fl & IFF_DEBUG)
388 		*flags |= LINUX_IFF_DEBUG;
389 	if (fl & IFF_LOOPBACK)
390 		*flags |= LINUX_IFF_LOOPBACK;
391 	if (fl & IFF_POINTOPOINT)
392 		*flags |= LINUX_IFF_POINTOPOINT;
393 	if (fl & IFF_DRV_RUNNING)
394 		*flags |= LINUX_IFF_RUNNING;
395 	if (fl & IFF_NOARP)
396 		*flags |= LINUX_IFF_NOARP;
397 	if (fl & IFF_PROMISC)
398 		*flags |= LINUX_IFF_PROMISC;
399 	if (fl & IFF_ALLMULTI)
400 		*flags |= LINUX_IFF_ALLMULTI;
401 	if (fl & IFF_MULTICAST)
402 		*flags |= LINUX_IFF_MULTICAST;
403 }
404 
405 int
406 linux_ifhwaddr(struct ifnet *ifp, struct l_sockaddr *lsa)
407 {
408 	struct ifaddr *ifa;
409 	struct sockaddr_dl *sdl;
410 
411 	if (IFP_IS_LOOP(ifp)) {
412 		bzero(lsa, sizeof(*lsa));
413 		lsa->sa_family = LINUX_ARPHRD_LOOPBACK;
414 		return (0);
415 	}
416 
417 	if (!IFP_IS_ETH(ifp))
418 		return (ENOENT);
419 
420 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
421 		sdl = (struct sockaddr_dl*)ifa->ifa_addr;
422 		if (sdl != NULL && (sdl->sdl_family == AF_LINK) &&
423 		    (sdl->sdl_type == IFT_ETHER)) {
424 			bzero(lsa, sizeof(*lsa));
425 			lsa->sa_family = LINUX_ARPHRD_ETHER;
426 			bcopy(LLADDR(sdl), lsa->sa_data, LINUX_IFHWADDRLEN);
427 			return (0);
428 		}
429 	}
430 
431 	return (ENOENT);
432 }
433 
434 int
435 linux_to_bsd_domain(int domain)
436 {
437 
438 	switch (domain) {
439 	case LINUX_AF_UNSPEC:
440 		return (AF_UNSPEC);
441 	case LINUX_AF_UNIX:
442 		return (AF_LOCAL);
443 	case LINUX_AF_INET:
444 		return (AF_INET);
445 	case LINUX_AF_INET6:
446 		return (AF_INET6);
447 	case LINUX_AF_AX25:
448 		return (AF_CCITT);
449 	case LINUX_AF_IPX:
450 		return (AF_IPX);
451 	case LINUX_AF_APPLETALK:
452 		return (AF_APPLETALK);
453 	case LINUX_AF_NETLINK:
454 		return (AF_NETLINK);
455 	}
456 	return (-1);
457 }
458 
459 int
460 bsd_to_linux_domain(int domain)
461 {
462 
463 	switch (domain) {
464 	case AF_UNSPEC:
465 		return (LINUX_AF_UNSPEC);
466 	case AF_LOCAL:
467 		return (LINUX_AF_UNIX);
468 	case AF_INET:
469 		return (LINUX_AF_INET);
470 	case AF_INET6:
471 		return (LINUX_AF_INET6);
472 	case AF_CCITT:
473 		return (LINUX_AF_AX25);
474 	case AF_IPX:
475 		return (LINUX_AF_IPX);
476 	case AF_APPLETALK:
477 		return (LINUX_AF_APPLETALK);
478 	case AF_NETLINK:
479 		return (LINUX_AF_NETLINK);
480 	}
481 	return (-1);
482 }
483 
484 /*
485  * Based on the fact that:
486  * 1. Native and Linux storage of struct sockaddr
487  * and struct sockaddr_in6 are equal.
488  * 2. On Linux sa_family is the first member of all struct sockaddr.
489  */
490 int
491 bsd_to_linux_sockaddr(const struct sockaddr *sa, struct l_sockaddr **lsa,
492     socklen_t len)
493 {
494 	struct l_sockaddr *kosa;
495 	int bdom;
496 
497 	*lsa = NULL;
498 	if (len < 2 || len > UCHAR_MAX)
499 		return (EINVAL);
500 	bdom = bsd_to_linux_domain(sa->sa_family);
501 	if (bdom == -1)
502 		return (EAFNOSUPPORT);
503 
504 	kosa = malloc(len, M_LINUX, M_WAITOK);
505 	bcopy(sa, kosa, len);
506 	kosa->sa_family = bdom;
507 	*lsa = kosa;
508 	return (0);
509 }
510 
511 int
512 linux_to_bsd_sockaddr(const struct l_sockaddr *osa, struct sockaddr **sap,
513     socklen_t *len)
514 {
515 	struct sockaddr *sa;
516 	struct l_sockaddr *kosa;
517 #ifdef INET6
518 	struct sockaddr_in6 *sin6;
519 	bool  oldv6size;
520 #endif
521 	char *name;
522 	int salen, bdom, error, hdrlen, namelen;
523 
524 	if (*len < 2 || *len > UCHAR_MAX)
525 		return (EINVAL);
526 
527 	salen = *len;
528 
529 #ifdef INET6
530 	oldv6size = false;
531 	/*
532 	 * Check for old (pre-RFC2553) sockaddr_in6. We may accept it
533 	 * if it's a v4-mapped address, so reserve the proper space
534 	 * for it.
535 	 */
536 	if (salen == sizeof(struct sockaddr_in6) - sizeof(uint32_t)) {
537 		salen += sizeof(uint32_t);
538 		oldv6size = true;
539 	}
540 #endif
541 
542 	kosa = malloc(salen, M_SONAME, M_WAITOK);
543 
544 	if ((error = copyin(osa, kosa, *len)))
545 		goto out;
546 
547 	bdom = linux_to_bsd_domain(kosa->sa_family);
548 	if (bdom == -1) {
549 		error = EAFNOSUPPORT;
550 		goto out;
551 	}
552 
553 #ifdef INET6
554 	/*
555 	 * Older Linux IPv6 code uses obsolete RFC2133 struct sockaddr_in6,
556 	 * which lacks the scope id compared with RFC2553 one. If we detect
557 	 * the situation, reject the address and write a message to system log.
558 	 *
559 	 * Still accept addresses for which the scope id is not used.
560 	 */
561 	if (oldv6size) {
562 		if (bdom == AF_INET6) {
563 			sin6 = (struct sockaddr_in6 *)kosa;
564 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ||
565 			    (!IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
566 			     !IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
567 			     !IN6_IS_ADDR_V4COMPAT(&sin6->sin6_addr) &&
568 			     !IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) &&
569 			     !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))) {
570 				sin6->sin6_scope_id = 0;
571 			} else {
572 				linux_msg(curthread,
573 				    "obsolete pre-RFC2553 sockaddr_in6 rejected");
574 				error = EINVAL;
575 				goto out;
576 			}
577 		} else
578 			salen -= sizeof(uint32_t);
579 	}
580 #endif
581 	if (bdom == AF_INET) {
582 		if (salen < sizeof(struct sockaddr_in)) {
583 			error = EINVAL;
584 			goto out;
585 		}
586 		salen = sizeof(struct sockaddr_in);
587 	}
588 
589 	if (bdom == AF_LOCAL && salen > sizeof(struct sockaddr_un)) {
590 		hdrlen = offsetof(struct sockaddr_un, sun_path);
591 		name = ((struct sockaddr_un *)kosa)->sun_path;
592 		if (*name == '\0') {
593 			/*
594 			 * Linux abstract namespace starts with a NULL byte.
595 			 * XXX We do not support abstract namespace yet.
596 			 */
597 			namelen = strnlen(name + 1, salen - hdrlen - 1) + 1;
598 		} else
599 			namelen = strnlen(name, salen - hdrlen);
600 		salen = hdrlen + namelen;
601 		if (salen > sizeof(struct sockaddr_un)) {
602 			error = ENAMETOOLONG;
603 			goto out;
604 		}
605 	}
606 
607 	if (bdom == AF_NETLINK) {
608 		if (salen < sizeof(struct sockaddr_nl)) {
609 			error = EINVAL;
610 			goto out;
611 		}
612 		salen = sizeof(struct sockaddr_nl);
613 	}
614 
615 	sa = (struct sockaddr *)kosa;
616 	sa->sa_family = bdom;
617 	sa->sa_len = salen;
618 
619 	*sap = sa;
620 	*len = salen;
621 	return (0);
622 
623 out:
624 	free(kosa, M_SONAME);
625 	return (error);
626 }
627 
628 void
629 linux_dev_shm_create(void)
630 {
631 	int error;
632 
633 	error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &dev_shm_cdev,
634 	    &dev_shm_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0, "shm/.mountpoint");
635 	if (error != 0) {
636 		printf("%s: failed to create device node, error %d\n",
637 		    __func__, error);
638 	}
639 }
640 
641 void
642 linux_dev_shm_destroy(void)
643 {
644 
645 	destroy_dev(dev_shm_cdev);
646 }
647 
648 int
649 bsd_to_linux_bits_(int value, struct bsd_to_linux_bitmap *bitmap,
650     size_t mapcnt, int no_value)
651 {
652 	int bsd_mask, bsd_value, linux_mask, linux_value;
653 	int linux_ret;
654 	size_t i;
655 	bool applied;
656 
657 	applied = false;
658 	linux_ret = 0;
659 	for (i = 0; i < mapcnt; ++i) {
660 		bsd_mask = bitmap[i].bsd_mask;
661 		bsd_value = bitmap[i].bsd_value;
662 		if (bsd_mask == 0)
663 			bsd_mask = bsd_value;
664 
665 		linux_mask = bitmap[i].linux_mask;
666 		linux_value = bitmap[i].linux_value;
667 		if (linux_mask == 0)
668 			linux_mask = linux_value;
669 
670 		/*
671 		 * If a mask larger than just the value is set, we explicitly
672 		 * want to make sure that only this bit we mapped within that
673 		 * mask is set.
674 		 */
675 		if ((value & bsd_mask) == bsd_value) {
676 			linux_ret = (linux_ret & ~linux_mask) | linux_value;
677 			applied = true;
678 		}
679 	}
680 
681 	if (!applied)
682 		return (no_value);
683 	return (linux_ret);
684 }
685 
686 int
687 linux_to_bsd_bits_(int value, struct bsd_to_linux_bitmap *bitmap,
688     size_t mapcnt, int no_value)
689 {
690 	int bsd_mask, bsd_value, linux_mask, linux_value;
691 	int bsd_ret;
692 	size_t i;
693 	bool applied;
694 
695 	applied = false;
696 	bsd_ret = 0;
697 	for (i = 0; i < mapcnt; ++i) {
698 		bsd_mask = bitmap[i].bsd_mask;
699 		bsd_value = bitmap[i].bsd_value;
700 		if (bsd_mask == 0)
701 			bsd_mask = bsd_value;
702 
703 		linux_mask = bitmap[i].linux_mask;
704 		linux_value = bitmap[i].linux_value;
705 		if (linux_mask == 0)
706 			linux_mask = linux_value;
707 
708 		/*
709 		 * If a mask larger than just the value is set, we explicitly
710 		 * want to make sure that only this bit we mapped within that
711 		 * mask is set.
712 		 */
713 		if ((value & linux_mask) == linux_value) {
714 			bsd_ret = (bsd_ret & ~bsd_mask) | bsd_value;
715 			applied = true;
716 		}
717 	}
718 
719 	if (!applied)
720 		return (no_value);
721 	return (bsd_ret);
722 }
723 
724 void
725 linux_to_bsd_poll_events(struct thread *td, int fd, short lev,
726     short *bev)
727 {
728 	struct file *fp;
729 	int error;
730 	short bits = 0;
731 
732 	if (lev & LINUX_POLLIN)
733 		bits |= POLLIN;
734 	if (lev & LINUX_POLLPRI)
735 		bits |=	POLLPRI;
736 	if (lev & LINUX_POLLOUT)
737 		bits |= POLLOUT;
738 	if (lev & LINUX_POLLERR)
739 		bits |= POLLERR;
740 	if (lev & LINUX_POLLHUP)
741 		bits |= POLLHUP;
742 	if (lev & LINUX_POLLNVAL)
743 		bits |= POLLNVAL;
744 	if (lev & LINUX_POLLRDNORM)
745 		bits |= POLLRDNORM;
746 	if (lev & LINUX_POLLRDBAND)
747 		bits |= POLLRDBAND;
748 	if (lev & LINUX_POLLWRBAND)
749 		bits |= POLLWRBAND;
750 	if (lev & LINUX_POLLWRNORM)
751 		bits |= POLLWRNORM;
752 
753 	if (lev & LINUX_POLLRDHUP) {
754 		/*
755 		 * It seems that the Linux silencly ignores POLLRDHUP
756 		 * on non-socket file descriptors unlike FreeBSD, where
757 		 * events bits is more strictly checked (POLLSTANDARD).
758 		 */
759 		error = fget_unlocked(td, fd, &cap_no_rights, &fp);
760 		if (error == 0) {
761 			/*
762 			 * XXX. On FreeBSD POLLRDHUP applies only to
763 			 * stream sockets.
764 			 */
765 			if (fp->f_type == DTYPE_SOCKET)
766 				bits |= POLLRDHUP;
767 			fdrop(fp, td);
768 		}
769 	}
770 
771 	if (lev & LINUX_POLLMSG)
772 		LINUX_RATELIMIT_MSG_OPT1("unsupported POLLMSG, events(%d)", lev);
773 	if (lev & LINUX_POLLREMOVE)
774 		LINUX_RATELIMIT_MSG_OPT1("unsupported POLLREMOVE, events(%d)", lev);
775 
776 	*bev = bits;
777 }
778 
779 void
780 bsd_to_linux_poll_events(short bev, short *lev)
781 {
782 	short bits = 0;
783 
784 	if (bev & POLLIN)
785 		bits |= LINUX_POLLIN;
786 	if (bev & POLLPRI)
787 		bits |=	LINUX_POLLPRI;
788 	if (bev & (POLLOUT | POLLWRNORM))
789 		/*
790 		 * POLLWRNORM is equal to POLLOUT on FreeBSD,
791 		 * but not on Linux
792 		 */
793 		bits |= LINUX_POLLOUT;
794 	if (bev & POLLERR)
795 		bits |= LINUX_POLLERR;
796 	if (bev & POLLHUP)
797 		bits |= LINUX_POLLHUP;
798 	if (bev & POLLNVAL)
799 		bits |= LINUX_POLLNVAL;
800 	if (bev & POLLRDNORM)
801 		bits |= LINUX_POLLRDNORM;
802 	if (bev & POLLRDBAND)
803 		bits |= LINUX_POLLRDBAND;
804 	if (bev & POLLWRBAND)
805 		bits |= LINUX_POLLWRBAND;
806 	if (bev & POLLRDHUP)
807 		bits |= LINUX_POLLRDHUP;
808 
809 	*lev = bits;
810 }
811 
812 bool
813 linux_use_real_ifname(const struct ifnet *ifp)
814 {
815 
816 	return (use_real_ifnames);
817 }
818