xref: /illumos-gate/usr/src/lib/libipadm/common/libipadm.c (revision 20a7641f9918de8574b8b3b47dbe35c4bfc78df1)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
24  * Copyright (c) 2016, Chris Fraire <cfraire@me.com>.
25  * Copyright 2021 Tintri by DDN, Inc. All rights reserved.
26  */
27 
28 #include <stdio.h>
29 #include <stdlib.h>
30 #include <string.h>
31 #include <errno.h>
32 #include <fcntl.h>
33 #include <unistd.h>
34 #include <stropts.h>
35 #include <sys/sockio.h>
36 #include <sys/types.h>
37 #include <sys/stat.h>
38 #include <sys/socket.h>
39 #include <net/route.h>
40 #include <netinet/in.h>
41 #include <inet/ip.h>
42 #include <arpa/inet.h>
43 #include <libintl.h>
44 #include <libdlpi.h>
45 #include <libinetutil.h>
46 #include <libdladm.h>
47 #include <libdllink.h>
48 #include <libdliptun.h>
49 #include <strings.h>
50 #include <zone.h>
51 #include <ctype.h>
52 #include <limits.h>
53 #include <assert.h>
54 #include <netdb.h>
55 #include <pwd.h>
56 #include <auth_attr.h>
57 #include <secdb.h>
58 #include <nss_dbdefs.h>
59 #include "libipadm_impl.h"
60 
61 /* error codes and text description */
62 static struct ipadm_error_info {
63 	ipadm_status_t	error_code;
64 	const char	*error_desc;
65 } ipadm_errors[] = {
66 	{ IPADM_SUCCESS,	"Operation succeeded" },
67 	{ IPADM_FAILURE,	"Operation failed" },
68 	{ IPADM_EAUTH,		"Insufficient user authorizations" },
69 	{ IPADM_EPERM,		"Permission denied" },
70 	{ IPADM_NO_BUFS,	"No buffer space available" },
71 	{ IPADM_NO_MEMORY,	"Insufficient memory" },
72 	{ IPADM_BAD_ADDR,	"Invalid address" },
73 	{ IPADM_BAD_PROTOCOL,	"Incorrect protocol family for operation" },
74 	{ IPADM_DAD_FOUND,	"Duplicate address detected" },
75 	{ IPADM_EXISTS,		"Already exists" },
76 	{ IPADM_IF_EXISTS,	"Interface already exists" },
77 	{ IPADM_ADDROBJ_EXISTS, "Address object already exists" },
78 	{ IPADM_ADDRCONF_EXISTS, "Addrconf already in progress" },
79 	{ IPADM_ENXIO,		"Interface does not exist" },
80 	{ IPADM_GRP_NOTEMPTY,	"IPMP group is not empty" },
81 	{ IPADM_INVALID_ARG,	"Invalid argument provided" },
82 	{ IPADM_INVALID_NAME,	"Invalid name" },
83 	{ IPADM_DLPI_FAILURE,	"Could not open DLPI link" },
84 	{ IPADM_DLADM_FAILURE,	"Datalink does not exist" },
85 	{ IPADM_PROP_UNKNOWN,   "Unknown property" },
86 	{ IPADM_ERANGE,		"Value is outside the allowed range" },
87 	{ IPADM_ESRCH,		"Value does not exist" },
88 	{ IPADM_EOVERFLOW,	"Number of values exceeds the allowed limit" },
89 	{ IPADM_NOTFOUND,	"Object not found" },
90 	{ IPADM_IF_INUSE,	"Interface already in use" },
91 	{ IPADM_ADDR_INUSE,	"Address already in use" },
92 	{ IPADM_BAD_HOSTNAME,	"Hostname maps to multiple IP addresses" },
93 	{ IPADM_ADDR_NOTAVAIL,	"Can't assign requested address" },
94 	{ IPADM_ALL_ADDRS_NOT_ENABLED, "All addresses could not be enabled" },
95 	{ IPADM_NDPD_NOT_RUNNING, "IPv6 autoconf daemon in.ndpd not running" },
96 	{ IPADM_DHCP_START_ERROR, "Could not start dhcpagent" },
97 	{ IPADM_DHCP_IPC_ERROR,	"Could not communicate with dhcpagent" },
98 	{ IPADM_DHCP_IPC_TIMEOUT, "Communication with dhcpagent timed out" },
99 	{ IPADM_TEMPORARY_OBJ,	"Persistent operation on temporary object" },
100 	{ IPADM_IPC_ERROR,	"Could not communicate with ipmgmtd" },
101 	{ IPADM_NOTSUP,		"Operation not supported" },
102 	{ IPADM_OP_DISABLE_OBJ, "Operation not supported on disabled object" },
103 	{ IPADM_EBADE,		"Invalid data exchange with daemon" },
104 	{ IPADM_GZ_PERM,	"Operation not permitted on from-gz interface"}
105 };
106 
107 #define	IPADM_NUM_ERRORS	(sizeof (ipadm_errors) / sizeof (*ipadm_errors))
108 
109 ipadm_status_t
110 ipadm_errno2status(int error)
111 {
112 	switch (error) {
113 	case 0:
114 		return (IPADM_SUCCESS);
115 	case ENXIO:
116 		return (IPADM_ENXIO);
117 	case ENOMEM:
118 		return (IPADM_NO_MEMORY);
119 	case ENOBUFS:
120 		return (IPADM_NO_BUFS);
121 	case EINVAL:
122 		return (IPADM_INVALID_ARG);
123 	case EBUSY:
124 		return (IPADM_IF_INUSE);
125 	case EEXIST:
126 		return (IPADM_EXISTS);
127 	case EADDRNOTAVAIL:
128 		return (IPADM_ADDR_NOTAVAIL);
129 	case EADDRINUSE:
130 		return (IPADM_ADDR_INUSE);
131 	case ENOENT:
132 		return (IPADM_NOTFOUND);
133 	case ERANGE:
134 		return (IPADM_ERANGE);
135 	case EPERM:
136 		return (IPADM_EPERM);
137 	case ENOTSUP:
138 	case EOPNOTSUPP:
139 		return (IPADM_NOTSUP);
140 	case EBADF:
141 		return (IPADM_IPC_ERROR);
142 	case EBADE:
143 		return (IPADM_EBADE);
144 	case ESRCH:
145 		return (IPADM_ESRCH);
146 	case EOVERFLOW:
147 		return (IPADM_EOVERFLOW);
148 	default:
149 		return (IPADM_FAILURE);
150 	}
151 }
152 
153 /*
154  * Returns a message string for the given libipadm error status.
155  */
156 const char *
157 ipadm_status2str(ipadm_status_t status)
158 {
159 	int	i;
160 
161 	for (i = 0; i < IPADM_NUM_ERRORS; i++) {
162 		if (status == ipadm_errors[i].error_code)
163 			return (dgettext(TEXT_DOMAIN,
164 			    ipadm_errors[i].error_desc));
165 	}
166 
167 	return (dgettext(TEXT_DOMAIN, "<unknown error>"));
168 }
169 
170 /*
171  * Opens a handle to libipadm.
172  * Possible values for flags:
173  *  IPH_VRRP:	Used by VRRP daemon to set the socket option SO_VRRP.
174  *  IPH_LEGACY:	This is used whenever an application needs to provide a
175  *		logical interface name while creating or deleting
176  *		interfaces and static addresses.
177  *  IPH_INIT:   Used by ipadm_init_prop(), to initialize protocol properties
178  *		on reboot.
179  */
180 ipadm_status_t
181 ipadm_open(ipadm_handle_t *handle, uint32_t flags)
182 {
183 	ipadm_handle_t	iph;
184 	ipadm_status_t	status = IPADM_SUCCESS;
185 	zoneid_t	zoneid;
186 	ushort_t	zflags;
187 	int		on = B_TRUE;
188 
189 	if (handle == NULL)
190 		return (IPADM_INVALID_ARG);
191 	*handle = NULL;
192 
193 	if (flags & ~(IPH_VRRP|IPH_LEGACY|IPH_INIT|IPH_IPMGMTD))
194 		return (IPADM_INVALID_ARG);
195 
196 	if ((iph = calloc(1, sizeof (struct ipadm_handle))) == NULL)
197 		return (IPADM_NO_MEMORY);
198 	iph->iph_sock = -1;
199 	iph->iph_sock6 = -1;
200 	iph->iph_door_fd = -1;
201 	iph->iph_rtsock = -1;
202 	iph->iph_flags = flags;
203 	(void) pthread_mutex_init(&iph->iph_lock, NULL);
204 
205 	if ((iph->iph_sock = socket(AF_INET, SOCK_DGRAM, 0)) < 0 ||
206 	    (iph->iph_sock6 = socket(AF_INET6, SOCK_DGRAM, 0)) < 0) {
207 		goto errnofail;
208 	}
209 
210 	/*
211 	 * We open a handle to libdladm here, to facilitate some daemons (like
212 	 * nwamd) which opens handle to libipadm before devfsadmd installs the
213 	 * right device permissions into the kernel and requires "all"
214 	 * privileges to open DLD_CONTROL_DEV.
215 	 *
216 	 * In a non-global shared-ip zone there will be no DLD_CONTROL_DEV node
217 	 * and dladm_open() will fail. So, we avoid this by not calling
218 	 * dladm_open() for such zones.
219 	 */
220 	zoneid = getzoneid();
221 	iph->iph_zoneid = zoneid;
222 	if (zoneid != GLOBAL_ZONEID) {
223 		if (zone_getattr(zoneid, ZONE_ATTR_FLAGS, &zflags,
224 		    sizeof (zflags)) < 0) {
225 			goto errnofail;
226 		}
227 	}
228 	if ((zoneid == GLOBAL_ZONEID) || (zflags & ZF_NET_EXCL)) {
229 		if (dladm_open(&iph->iph_dlh) != DLADM_STATUS_OK) {
230 			ipadm_close(iph);
231 			return (IPADM_DLADM_FAILURE);
232 		}
233 		if (zoneid != GLOBAL_ZONEID) {
234 			iph->iph_rtsock = socket(PF_ROUTE, SOCK_RAW, 0);
235 			/*
236 			 * Failure to open rtsock is ignored as this is
237 			 * only used in non-global zones to initialize
238 			 * routing socket information.
239 			 */
240 		}
241 	} else {
242 		assert(zoneid != GLOBAL_ZONEID);
243 		iph->iph_dlh = NULL;
244 	}
245 	if (flags & IPH_VRRP) {
246 		if (setsockopt(iph->iph_sock6, SOL_SOCKET, SO_VRRP, &on,
247 		    sizeof (on)) < 0 || setsockopt(iph->iph_sock, SOL_SOCKET,
248 		    SO_VRRP, &on, sizeof (on)) < 0) {
249 			goto errnofail;
250 		}
251 	}
252 	*handle = iph;
253 	return (status);
254 
255 errnofail:
256 	status = ipadm_errno2status(errno);
257 	ipadm_close(iph);
258 	return (status);
259 }
260 
261 /*
262  * Closes and frees the libipadm handle.
263  */
264 void
265 ipadm_close(ipadm_handle_t iph)
266 {
267 	if (iph == NULL)
268 		return;
269 	if (iph->iph_sock != -1)
270 		(void) close(iph->iph_sock);
271 	if (iph->iph_sock6 != -1)
272 		(void) close(iph->iph_sock6);
273 	if (iph->iph_rtsock != -1)
274 		(void) close(iph->iph_rtsock);
275 	if (iph->iph_door_fd != -1)
276 		(void) close(iph->iph_door_fd);
277 	dladm_close(iph->iph_dlh);
278 	(void) pthread_mutex_destroy(&iph->iph_lock);
279 	free(iph);
280 }
281 
282 /*
283  * Checks if the caller has the authorization to configure network
284  * interfaces.
285  */
286 boolean_t
287 ipadm_check_auth(void)
288 {
289 	struct passwd	pwd;
290 	char		buf[NSS_BUFLEN_PASSWD];
291 
292 	/* get the password entry for the given user ID */
293 	if (getpwuid_r(getuid(), &pwd, buf, sizeof (buf)) == NULL)
294 		return (B_FALSE);
295 
296 	/* check for presence of given authorization */
297 	return (chkauthattr(NETWORK_INTERFACE_CONFIG_AUTH, pwd.pw_name) != 0);
298 }
299 
300 /*
301  * Stores the index value of the interface in `ifname' for the address
302  * family `af' into the buffer pointed to by `index'.
303  */
304 static ipadm_status_t
305 i_ipadm_get_index(ipadm_handle_t iph, const char *ifname, sa_family_t af,
306     int *index)
307 {
308 	struct lifreq	lifr;
309 	int		sock;
310 
311 	bzero(&lifr, sizeof (lifr));
312 	(void) strlcpy(lifr.lifr_name, ifname, sizeof (lifr.lifr_name));
313 	if (af == AF_INET)
314 		sock = iph->iph_sock;
315 	else
316 		sock = iph->iph_sock6;
317 
318 	if (ioctl(sock, SIOCGLIFINDEX, (caddr_t)&lifr) < 0)
319 		return (ipadm_errno2status(errno));
320 	*index = lifr.lifr_index;
321 
322 	return (IPADM_SUCCESS);
323 }
324 
325 /*
326  * Maximum amount of time (in milliseconds) to wait for Duplicate Address
327  * Detection to complete in the kernel.
328  */
329 #define	DAD_WAIT_TIME		1000
330 
331 /*
332  * Any time that flags are changed on an interface where either the new or the
333  * existing flags have IFF_UP set, we'll get a RTM_NEWADDR message to
334  * announce the new address added and its flag status.
335  * We wait here for that message and look for IFF_UP.
336  * If something's amiss with the kernel, though, we don't wait forever.
337  * (Note that IFF_DUPLICATE is a high-order bit, and we cannot see
338  * it in the routing socket messages.)
339  */
340 static ipadm_status_t
341 i_ipadm_dad_wait(ipadm_handle_t handle, const char *lifname, sa_family_t af,
342     int rtsock)
343 {
344 	struct pollfd	fds[1];
345 	union {
346 		struct if_msghdr ifm;
347 		char buf[1024];
348 	} msg;
349 	int		index;
350 	ipadm_status_t	retv;
351 	uint64_t	flags;
352 	hrtime_t	starttime, now;
353 
354 	fds[0].fd = rtsock;
355 	fds[0].events = POLLIN;
356 	fds[0].revents = 0;
357 
358 	retv = i_ipadm_get_index(handle, lifname, af, &index);
359 	if (retv != IPADM_SUCCESS)
360 		return (retv);
361 
362 	starttime = gethrtime();
363 	for (;;) {
364 		now = gethrtime();
365 		now = (now - starttime) / 1000000;
366 		if (now >= DAD_WAIT_TIME)
367 			break;
368 		if (poll(fds, 1, DAD_WAIT_TIME - (int)now) <= 0)
369 			break;
370 		if (read(rtsock, &msg, sizeof (msg)) <= 0)
371 			break;
372 		if (msg.ifm.ifm_type != RTM_NEWADDR)
373 			continue;
374 		/* Note that ifm_index is just 16 bits */
375 		if (index == msg.ifm.ifm_index && (msg.ifm.ifm_flags & IFF_UP))
376 			return (IPADM_SUCCESS);
377 	}
378 
379 	retv = i_ipadm_get_flags(handle, lifname, af, &flags);
380 	if (retv != IPADM_SUCCESS)
381 		return (retv);
382 	if (flags & IFF_DUPLICATE)
383 		return (IPADM_DAD_FOUND);
384 
385 	return (IPADM_SUCCESS);
386 }
387 
388 /*
389  * Sets the flags `on_flags' and resets the flags `off_flags' for the logical
390  * interface in `lifname'.
391  *
392  * If the new flags value will transition the interface from "down" to "up"
393  * then duplicate address detection is performed by the kernel.  This routine
394  * waits to get the outcome of that test.
395  */
396 ipadm_status_t
397 i_ipadm_set_flags(ipadm_handle_t iph, const char *lifname, sa_family_t af,
398     uint64_t on_flags, uint64_t off_flags)
399 {
400 	struct lifreq	lifr;
401 	uint64_t	oflags;
402 	ipadm_status_t	ret;
403 	int		rtsock = -1;
404 	int		sock, err;
405 
406 	ret = i_ipadm_get_flags(iph, lifname, af, &oflags);
407 	if (ret != IPADM_SUCCESS)
408 		return (ret);
409 
410 	sock = (af == AF_INET ? iph->iph_sock : iph->iph_sock6);
411 
412 	/*
413 	 * Any time flags are changed on an interface that has IFF_UP set,
414 	 * we get a routing socket message.  We care about the status,
415 	 * though, only when the new flags are marked "up."
416 	 */
417 	if (!(oflags & IFF_UP) && (on_flags & IFF_UP))
418 		rtsock = socket(PF_ROUTE, SOCK_RAW, af);
419 
420 	oflags |= on_flags;
421 	oflags &= ~off_flags;
422 	bzero(&lifr, sizeof (lifr));
423 	(void) strlcpy(lifr.lifr_name, lifname, sizeof (lifr.lifr_name));
424 	lifr.lifr_flags = oflags;
425 	if (ioctl(sock, SIOCSLIFFLAGS, (caddr_t)&lifr) < 0) {
426 		err = errno;
427 		if (rtsock != -1)
428 			(void) close(rtsock);
429 		return (ipadm_errno2status(err));
430 	}
431 	if (rtsock == -1) {
432 		return (IPADM_SUCCESS);
433 	} else {
434 		/* Wait for DAD to complete. */
435 		ret = i_ipadm_dad_wait(iph, lifname, af, rtsock);
436 		(void) close(rtsock);
437 		return (ret);
438 	}
439 }
440 
441 /*
442  * Returns the flags value for the logical interface in `lifname'
443  * in the buffer pointed to by `flags'.
444  */
445 ipadm_status_t
446 i_ipadm_get_flags(ipadm_handle_t iph, const char *lifname, sa_family_t af,
447     uint64_t *flags)
448 {
449 	struct lifreq	lifr;
450 	int		sock;
451 
452 	bzero(&lifr, sizeof (lifr));
453 	(void) strlcpy(lifr.lifr_name, lifname, sizeof (lifr.lifr_name));
454 	if (af == AF_INET)
455 		sock = iph->iph_sock;
456 	else
457 		sock = iph->iph_sock6;
458 
459 	if (ioctl(sock, SIOCGLIFFLAGS, (caddr_t)&lifr) < 0) {
460 		return (ipadm_errno2status(errno));
461 	}
462 	*flags = lifr.lifr_flags;
463 
464 	return (IPADM_SUCCESS);
465 }
466 
467 /*
468  * Determines whether or not an interface name represents a loopback
469  * interface, before the interface has been plumbed.
470  * It is assumed that the interface name in `ifname' is of correct format
471  * as verified by ifparse_ifspec().
472  *
473  * Returns: B_TRUE if loopback, B_FALSE if not.
474  */
475 boolean_t
476 i_ipadm_is_loopback(const char *ifname)
477 {
478 	int len = strlen(LOOPBACK_IF);
479 
480 	return (strncmp(ifname, LOOPBACK_IF, len) == 0 &&
481 	    (ifname[len] == '\0' || ifname[len] == IPADM_LOGICAL_SEP));
482 }
483 
484 /*
485  * Determines whether or not an interface name represents a vni
486  * interface, before the interface has been plumbed.
487  * It is assumed that the interface name in `ifname' is of correct format
488  * as verified by ifparse_ifspec().
489  *
490  * Returns: B_TRUE if vni, B_FALSE if not.
491  */
492 boolean_t
493 i_ipadm_is_vni(const char *ifname)
494 {
495 	ifspec_t	ifsp;
496 
497 	return (ifparse_ifspec(ifname, &ifsp) &&
498 	    strcmp(ifsp.ifsp_devnm, "vni") == 0);
499 }
500 
501 /*
502  * Returns B_TRUE if `ifname' is an IP interface on a 6to4 tunnel.
503  */
504 boolean_t
505 i_ipadm_is_6to4(ipadm_handle_t iph, char *ifname)
506 {
507 	dladm_status_t		dlstatus;
508 	datalink_class_t	class;
509 	iptun_params_t		params;
510 	datalink_id_t		linkid;
511 
512 	if (iph->iph_dlh == NULL) {
513 		assert(iph->iph_zoneid != GLOBAL_ZONEID);
514 		return (B_FALSE);
515 	}
516 	dlstatus = dladm_name2info(iph->iph_dlh, ifname, &linkid, NULL,
517 	    &class, NULL);
518 	if (dlstatus == DLADM_STATUS_OK && class == DATALINK_CLASS_IPTUN) {
519 		params.iptun_param_linkid = linkid;
520 		dlstatus = dladm_iptun_getparams(iph->iph_dlh, &params,
521 		    DLADM_OPT_ACTIVE);
522 		if (dlstatus == DLADM_STATUS_OK &&
523 		    params.iptun_param_type == IPTUN_TYPE_6TO4) {
524 			return (B_TRUE);
525 		}
526 	}
527 	return (B_FALSE);
528 }
529 
530 /*
531  * For a given interface name, ipadm_if_enabled() checks if v4
532  * or v6 or both IP interfaces exist in the active configuration.
533  */
534 boolean_t
535 ipadm_if_enabled(ipadm_handle_t iph, const char *ifname, sa_family_t af)
536 {
537 	struct lifreq	lifr;
538 	int		s4 = iph->iph_sock;
539 	int		s6 = iph->iph_sock6;
540 
541 	bzero(&lifr, sizeof (lifr));
542 	(void) strlcpy(lifr.lifr_name, ifname, sizeof (lifr.lifr_name));
543 	switch (af) {
544 	case AF_INET:
545 		if (ioctl(s4, SIOCGLIFFLAGS, (caddr_t)&lifr) == 0)
546 			return (B_TRUE);
547 		break;
548 	case AF_INET6:
549 		if (ioctl(s6, SIOCGLIFFLAGS, (caddr_t)&lifr) == 0)
550 			return (B_TRUE);
551 		break;
552 	case AF_UNSPEC:
553 		if (ioctl(s4, SIOCGLIFFLAGS, (caddr_t)&lifr) == 0 ||
554 		    ioctl(s6, SIOCGLIFFLAGS, (caddr_t)&lifr) == 0) {
555 			return (B_TRUE);
556 		}
557 	}
558 	return (B_FALSE);
559 }
560 
561 /*
562  * Apply the interface property by retrieving information from nvl.
563  */
564 static ipadm_status_t
565 i_ipadm_init_ifprop(ipadm_handle_t iph, nvlist_t *nvl)
566 {
567 	nvpair_t	*nvp;
568 	char		*name, *pname = NULL;
569 	char		*protostr = NULL, *ifname = NULL, *pval = NULL;
570 	uint_t		proto;
571 	int		err = 0;
572 
573 	for (nvp = nvlist_next_nvpair(nvl, NULL); nvp != NULL;
574 	    nvp = nvlist_next_nvpair(nvl, nvp)) {
575 		name = nvpair_name(nvp);
576 		if (strcmp(name, IPADM_NVP_IFNAME) == 0) {
577 			if ((err = nvpair_value_string(nvp, &ifname)) != 0)
578 				break;
579 		} else if (strcmp(name, IPADM_NVP_PROTONAME) == 0) {
580 			if ((err = nvpair_value_string(nvp, &protostr)) != 0)
581 				break;
582 		} else {
583 			assert(!IPADM_PRIV_NVP(name));
584 			pname = name;
585 			if ((err = nvpair_value_string(nvp, &pval)) != 0)
586 				break;
587 		}
588 	}
589 	if (err != 0)
590 		return (ipadm_errno2status(err));
591 	proto = ipadm_str2proto(protostr);
592 	return (ipadm_set_ifprop(iph, ifname, pname, pval, proto,
593 	    IPADM_OPT_ACTIVE));
594 }
595 
596 /*
597  * Instantiate the address object or set the address object property by
598  * retrieving the configuration from the nvlist `nvl'.
599  */
600 ipadm_status_t
601 i_ipadm_init_addrobj(ipadm_handle_t iph, nvlist_t *nvl)
602 {
603 	nvpair_t	*nvp;
604 	char		*name;
605 	char		*aobjname = NULL, *pval = NULL, *ifname = NULL;
606 	sa_family_t	af = AF_UNSPEC;
607 	ipadm_addr_type_t atype = IPADM_ADDR_NONE;
608 	int		err = 0;
609 	ipadm_status_t	status = IPADM_SUCCESS;
610 
611 	for (nvp = nvlist_next_nvpair(nvl, NULL); nvp != NULL;
612 	    nvp = nvlist_next_nvpair(nvl, nvp)) {
613 		name = nvpair_name(nvp);
614 		if (strcmp(name, IPADM_NVP_IFNAME) == 0) {
615 			if ((err = nvpair_value_string(nvp, &ifname)) != 0)
616 				break;
617 		} else if (strcmp(name, IPADM_NVP_AOBJNAME) == 0) {
618 			if ((err = nvpair_value_string(nvp, &aobjname)) != 0)
619 				break;
620 		} else if (i_ipadm_name2atype(name, &af, &atype)) {
621 			break;
622 		} else {
623 			assert(!IPADM_PRIV_NVP(name));
624 			err = nvpair_value_string(nvp, &pval);
625 			break;
626 		}
627 	}
628 	if (err != 0)
629 		return (ipadm_errno2status(err));
630 
631 	switch (atype) {
632 	case IPADM_ADDR_STATIC:
633 		status = i_ipadm_enable_static(iph, ifname, nvl, af);
634 		break;
635 	case IPADM_ADDR_DHCP:
636 		status = i_ipadm_enable_dhcp(iph, ifname, nvl);
637 		if (status == IPADM_DHCP_IPC_TIMEOUT)
638 			status = IPADM_SUCCESS;
639 		break;
640 	case IPADM_ADDR_IPV6_ADDRCONF:
641 		status = i_ipadm_enable_addrconf(iph, ifname, nvl);
642 		break;
643 	case IPADM_ADDR_NONE:
644 		status = ipadm_set_addrprop(iph, name, pval, aobjname,
645 		    IPADM_OPT_ACTIVE);
646 		break;
647 	}
648 
649 	return (status);
650 }
651 
652 /*
653  * Instantiate the interface object by retrieving the configuration from
654  * `ifnvl'. The nvlist `ifnvl' contains all the persistent configuration
655  * (interface properties and address objects on that interface) for the
656  * given `ifname'.
657  */
658 ipadm_status_t
659 i_ipadm_init_ifobj(ipadm_handle_t iph, const char *ifname, nvlist_t *ifnvl)
660 {
661 	nvlist_t	*nvl = NULL;
662 	nvpair_t	*nvp;
663 	ipadm_status_t	status = IPADM_ENXIO;
664 	ipadm_status_t	ret_status = IPADM_SUCCESS;
665 	char		newifname[LIFNAMSIZ];
666 	char		*aobjstr;
667 	uint16_t	*afs;
668 	char		*gifname;
669 	uint_t		nelem = 0;
670 	boolean_t	init_from_gz = B_FALSE;
671 	boolean_t	move_to_group = B_FALSE;
672 
673 	(void) strlcpy(newifname, ifname, sizeof (newifname));
674 
675 	/*
676 	 * First go through the ifnvl nvlist looking for nested nvlist
677 	 * containing interface class and address families.
678 	 */
679 	for (nvp = nvlist_next_nvpair(ifnvl, NULL); nvp != NULL;
680 	    nvp = nvlist_next_nvpair(ifnvl, nvp)) {
681 		char *icstr;
682 		char **mifnames;
683 		uint32_t ipadm_flags = IPADM_OPT_ACTIVE;
684 
685 		if (nvpair_value_nvlist(nvp, &nvl) != 0 ||
686 		    nvlist_lookup_uint16_array(nvl, IPADM_NVP_FAMILIES,
687 		    &afs, &nelem) != 0)
688 			continue;
689 
690 		/* Check if this is IPMP group interface */
691 		if (nvlist_lookup_string(nvl, IPADM_NVP_IFCLASS,
692 		    &icstr) == 0 && atoi(icstr) == IPADM_IF_CLASS_IPMP)
693 			ipadm_flags |= IPADM_OPT_IPMP;
694 
695 		/* Create interfaces for address families specified */
696 		while (nelem-- > 0) {
697 			uint16_t af = afs[nelem];
698 
699 			assert(af == AF_INET || af == AF_INET6);
700 
701 			status = i_ipadm_plumb_if(iph, newifname, af,
702 			    ipadm_flags);
703 			if (status == IPADM_IF_EXISTS)
704 				status = IPADM_SUCCESS;
705 			if (status != IPADM_SUCCESS)
706 				return (status);
707 		}
708 		if (nvlist_lookup_string(nvl, IPADM_NVP_GIFNAME,
709 		    &gifname) == 0) {
710 			/*
711 			 * IPMP underlying interface. Move to the
712 			 * specified IPMP group.
713 			 */
714 			move_to_group = B_TRUE;
715 		} else if ((ipadm_flags & IPADM_OPT_IPMP) &&
716 		    nvlist_lookup_string_array(nvl, IPADM_NVP_MIFNAMES,
717 		    &mifnames, &nelem) == 0) {
718 			/* Non-empty IPMP group interface */
719 			while (nelem-- > 0) {
720 				(void) ipadm_add_ipmp_member(iph, newifname,
721 				    mifnames[nelem], IPADM_OPT_ACTIVE);
722 			}
723 		}
724 		if (iph->iph_zoneid != GLOBAL_ZONEID)
725 			init_from_gz = B_TRUE;
726 	}
727 
728 	if (status != IPADM_SUCCESS)
729 		return (status);
730 
731 	/*
732 	 * Go through the ifnvl nvlist again, applying persistent configuration.
733 	 */
734 	for (nvp = nvlist_next_nvpair(ifnvl, NULL); nvp != NULL;
735 	    nvp = nvlist_next_nvpair(ifnvl, nvp)) {
736 		if (nvpair_value_nvlist(nvp, &nvl) != 0)
737 			continue;
738 		if (nvlist_lookup_string(nvl, IPADM_NVP_AOBJNAME,
739 		    &aobjstr) == 0) {
740 			/*
741 			 * For addresses, we need to relocate addrprops from the
742 			 * nvlist `ifnvl'.
743 			 */
744 			if (nvlist_exists(nvl, IPADM_NVP_IPV4ADDR) ||
745 			    nvlist_exists(nvl, IPADM_NVP_IPV6ADDR) ||
746 			    nvlist_exists(nvl, IPADM_NVP_DHCP)) {
747 				status = i_ipadm_merge_addrprops_from_nvl(ifnvl,
748 				    nvl, aobjstr);
749 
750 				if (status != IPADM_SUCCESS)
751 					continue;
752 			}
753 			status = i_ipadm_init_addrobj(iph, nvl);
754 
755 			/*
756 			 * If this address is in use on some other interface,
757 			 * we want to record an error to be returned as
758 			 * a soft error and continue processing the rest of
759 			 * the addresses.
760 			 */
761 			if (status == IPADM_ADDR_NOTAVAIL) {
762 				ret_status = IPADM_ALL_ADDRS_NOT_ENABLED;
763 				status = IPADM_SUCCESS;
764 			}
765 		} else if (nvlist_exists(nvl, IPADM_NVP_PROTONAME) == B_TRUE) {
766 			status = i_ipadm_init_ifprop(iph, nvl);
767 		}
768 		if (status != IPADM_SUCCESS)
769 			return (status);
770 	}
771 	if (move_to_group) {
772 		(void) ipadm_add_ipmp_member(iph, gifname, newifname,
773 		    IPADM_OPT_ACTIVE);
774 	}
775 	if (init_from_gz)
776 		ret_status = ipadm_init_net_from_gz(iph, newifname, NULL);
777 	return (ret_status);
778 }
779 
780 /*
781  * Retrieves the persistent configuration for the given interface(s) in `ifs'
782  * by contacting the daemon and dumps the information in `allifs'.
783  */
784 ipadm_status_t
785 i_ipadm_init_ifs(ipadm_handle_t iph, const char *ifs, nvlist_t **allifs)
786 {
787 	nvlist_t		*nvl = NULL;
788 	size_t			nvlsize, bufsize;
789 	ipmgmt_initif_arg_t	*iargp;
790 	char			*buf = NULL, *nvlbuf = NULL;
791 	ipmgmt_get_rval_t	*rvalp = NULL;
792 	int			err;
793 	ipadm_status_t		status = IPADM_SUCCESS;
794 
795 	status = ipadm_str2nvlist(ifs, &nvl, IPADM_NORVAL);
796 	if (status != IPADM_SUCCESS)
797 		return (status);
798 
799 	err = nvlist_pack(nvl, &nvlbuf, &nvlsize, NV_ENCODE_NATIVE, 0);
800 	if (err != 0) {
801 		status = ipadm_errno2status(err);
802 		goto done;
803 	}
804 	bufsize = sizeof (*iargp) + nvlsize;
805 	if ((buf = malloc(bufsize)) == NULL) {
806 		status = ipadm_errno2status(errno);
807 		goto done;
808 	}
809 
810 	/* populate the door_call argument structure */
811 	iargp = (void *)buf;
812 	iargp->ia_cmd = IPMGMT_CMD_INITIF;
813 	iargp->ia_flags = 0;
814 	iargp->ia_family = AF_UNSPEC;
815 	iargp->ia_nvlsize = nvlsize;
816 	(void) bcopy(nvlbuf, buf + sizeof (*iargp), nvlsize);
817 
818 	if ((rvalp = malloc(sizeof (ipmgmt_get_rval_t))) == NULL) {
819 		status = ipadm_errno2status(errno);
820 		goto done;
821 	}
822 	if ((err = ipadm_door_call(iph, iargp, bufsize, (void **)&rvalp,
823 	    sizeof (*rvalp), B_TRUE)) != 0) {
824 		status = ipadm_errno2status(err);
825 		goto done;
826 	}
827 
828 	/*
829 	 * Daemon reply pointed to by rvalp contains ipmgmt_get_rval_t structure
830 	 * followed by a list of packed nvlists, each of which represents
831 	 * configuration information for the given interface(s).
832 	 */
833 	err = nvlist_unpack((char *)rvalp + sizeof (ipmgmt_get_rval_t),
834 	    rvalp->ir_nvlsize, allifs, 0);
835 	if (err != 0)
836 		status = ipadm_errno2status(err);
837 done:
838 	nvlist_free(nvl);
839 	free(buf);
840 	free(nvlbuf);
841 	free(rvalp);
842 	return (status);
843 }
844 
845 /*
846  * Returns B_FALSE if
847  * (1) `ifname' is NULL or has no string or has a string of invalid length
848  * (2) ifname is a logical interface and IPH_LEGACY is not set, or
849  */
850 boolean_t
851 i_ipadm_validate_ifname(ipadm_handle_t iph, const char *ifname)
852 {
853 	ifspec_t ifsp;
854 
855 	if (ifname == NULL || ifname[0] == '\0' ||
856 	    !ifparse_ifspec(ifname, &ifsp))
857 		return (B_FALSE);
858 	if (ifsp.ifsp_lunvalid)
859 		return (ifsp.ifsp_lun > 0 && (iph->iph_flags & IPH_LEGACY));
860 	return (B_TRUE);
861 }
862 
863 /*
864  * Wrapper for sending a non-transparent I_STR ioctl().
865  * Returns: Result from ioctl().
866  */
867 int
868 i_ipadm_strioctl(int s, int cmd, char *buf, int buflen)
869 {
870 	struct strioctl ioc;
871 
872 	(void) memset(&ioc, 0, sizeof (ioc));
873 	ioc.ic_cmd = cmd;
874 	ioc.ic_timout = 0;
875 	ioc.ic_len = buflen;
876 	ioc.ic_dp = buf;
877 
878 	return (ioctl(s, I_STR, (char *)&ioc));
879 }
880 
881 /*
882  * Make a door call to the server and checks if the door call succeeded or not.
883  * `is_varsize' specifies that the data returned by ipmgmtd daemon is of
884  * variable size and door will allocate buffer using mmap(). In such cases
885  * we re-allocate the required memory,n assign it to `rbufp', copy the data to
886  * `rbufp' and then call munmap() (see below).
887  *
888  * It also checks to see if the server side procedure ran successfully by
889  * checking for ir_err. Therefore, for some callers who just care about the
890  * return status can set `rbufp' to NULL and set `rsize' to 0.
891  */
892 int
893 ipadm_door_call(ipadm_handle_t iph, void *arg, size_t asize, void **rbufp,
894     size_t rsize, boolean_t is_varsize)
895 {
896 	door_arg_t	darg;
897 	int		err;
898 	ipmgmt_retval_t	rval, *rvalp;
899 	boolean_t	reopen = B_FALSE;
900 
901 	if (rbufp == NULL) {
902 		rvalp = &rval;
903 		rbufp = (void **)&rvalp;
904 		rsize = sizeof (rval);
905 	}
906 
907 	darg.data_ptr = arg;
908 	darg.data_size = asize;
909 	darg.desc_ptr = NULL;
910 	darg.desc_num = 0;
911 	darg.rbuf = *rbufp;
912 	darg.rsize = rsize;
913 
914 reopen:
915 	(void) pthread_mutex_lock(&iph->iph_lock);
916 	/* The door descriptor is opened if it isn't already */
917 	if (iph->iph_door_fd == -1) {
918 		if ((iph->iph_door_fd = open(IPMGMT_DOOR, O_RDONLY)) < 0) {
919 			err = errno;
920 			(void) pthread_mutex_unlock(&iph->iph_lock);
921 			return (err);
922 		}
923 	}
924 	(void) pthread_mutex_unlock(&iph->iph_lock);
925 
926 	if (door_call(iph->iph_door_fd, &darg) == -1) {
927 		/*
928 		 * Stale door descriptor is possible if ipmgmtd was restarted
929 		 * since last iph_door_fd was opened, so try re-opening door
930 		 * descriptor.
931 		 */
932 		if (!reopen && errno == EBADF) {
933 			(void) close(iph->iph_door_fd);
934 			iph->iph_door_fd = -1;
935 			reopen = B_TRUE;
936 			goto reopen;
937 		}
938 		return (errno);
939 	}
940 	err = ((ipmgmt_retval_t *)(void *)(darg.rbuf))->ir_err;
941 	if (darg.rbuf != *rbufp) {
942 		/*
943 		 * if the caller is expecting the result to fit in specified
944 		 * buffer then return failure.
945 		 */
946 		if (!is_varsize)
947 			err = EBADE;
948 		/*
949 		 * The size of the buffer `*rbufp' was not big enough
950 		 * and the door itself allocated buffer, for us. We will
951 		 * hit this, on several occasion as for some cases
952 		 * we cannot predict the size of the return structure.
953 		 * Reallocate the buffer `*rbufp' and memcpy() the contents
954 		 * to new buffer.
955 		 */
956 		if (err == 0) {
957 			void *newp;
958 
959 			/* allocated memory will be freed by the caller */
960 			if ((newp = realloc(*rbufp, darg.rsize)) == NULL) {
961 				err = ENOMEM;
962 			} else {
963 				*rbufp = newp;
964 				(void) memcpy(*rbufp, darg.rbuf, darg.rsize);
965 			}
966 		}
967 		/* munmap() the door buffer */
968 		(void) munmap(darg.rbuf, darg.rsize);
969 	} else {
970 		if (darg.rsize != rsize)
971 			err = EBADE;
972 	}
973 	return (err);
974 }
975 
976 /*
977  * A helper that is used by i_ipadm_get_db_addr and i_ipadm_get_db_if
978  * to do a door_call to ipmgmtd, that should return persistent information
979  * about interfaces or/and addresses from ipadm DB
980  */
981 ipadm_status_t
982 i_ipadm_call_ipmgmtd(ipadm_handle_t iph, void *garg, size_t garg_size,
983     nvlist_t **onvl)
984 {
985 	ipmgmt_get_rval_t	*rvalp;
986 	int			err;
987 	size_t			nvlsize;
988 	char			*nvlbuf;
989 
990 	rvalp = malloc(sizeof (ipmgmt_get_rval_t));
991 	if (rvalp == NULL)
992 		return (IPADM_NO_MEMORY);
993 
994 	err = ipadm_door_call(iph, garg, garg_size, (void **)&rvalp,
995 	    sizeof (*rvalp), B_TRUE);
996 	if (err == 0) {
997 		nvlsize = rvalp->ir_nvlsize;
998 		nvlbuf = (char *)rvalp + sizeof (ipmgmt_get_rval_t);
999 		err = nvlist_unpack(nvlbuf, nvlsize, onvl, 0);
1000 	}
1001 	free(rvalp);
1002 
1003 	return (ipadm_errno2status(err));
1004 }
1005 
1006 /*
1007  * ipadm_is_nil_hostname() : Determine if the `hostname' is nil: i.e.,
1008  *			NULL, empty, or a single space (e.g., as returned by
1009  *			domainname(8)/sysinfo).
1010  *
1011  *   input: const char *: the hostname to inspect;
1012  *  output: boolean_t: B_TRUE if `hostname' is not NULL satisfies the
1013  *			criteria above; otherwise, B_FALSE;
1014  */
1015 
1016 boolean_t
1017 ipadm_is_nil_hostname(const char *hostname)
1018 {
1019 	return (hostname == NULL || *hostname == '\0' ||
1020 	    (*hostname == ' ' && hostname[1] == '\0'));
1021 }
1022 
1023 /*
1024  * ipadm_is_valid_hostname(): check whether a string is a valid hostname
1025  *
1026  *   input: const char *: the string to verify as a hostname
1027  *  output: boolean_t: B_TRUE if the string is a valid hostname
1028  *
1029  * Note that we accept host names beginning with a digit, which is not
1030  * strictly legal according to the RFCs but is in common practice, so we
1031  * endeavour to not break what customers are using.
1032  *
1033  * RFC 1035 limits a wire-format domain name to 255 octets. For a printable
1034  * `hostname' as we have, the limit is therefore 253 characters (excluding
1035  * the terminating '\0'--or 254 characters if the last character of
1036  * `hostname' is a '.'.
1037  *
1038  * Excerpt from section 2.3.1., Preferred name syntax:
1039  *
1040  * <domain> ::= <subdomain> | " "
1041  * <subdomain> ::= <label> | <subdomain> "." <label>
1042  * <label> ::= <letter> [ [ <ldh-str> ] <let-dig> ]
1043  * <ldh-str> ::= <let-dig-hyp> | <let-dig-hyp> <ldh-str>
1044  * <let-dig-hyp> ::= <let-dig> | "-"
1045  * <let-dig> ::= <letter> | <digit>
1046  */
1047 boolean_t
1048 ipadm_is_valid_hostname(const char *hostname)
1049 {
1050 	const size_t MAX_READABLE_NAME_LEN = 253;
1051 	char last_char;
1052 	size_t has_last_dot, namelen, i;
1053 
1054 	if (hostname == NULL)
1055 		return (B_FALSE);
1056 
1057 	namelen = strlen(hostname);
1058 	if (namelen < 1)
1059 		return (B_FALSE);
1060 
1061 	last_char = hostname[namelen - 1];
1062 	has_last_dot = last_char == '.';
1063 
1064 	if (namelen > MAX_READABLE_NAME_LEN + has_last_dot ||
1065 	    last_char == '-')
1066 		return (B_FALSE);
1067 
1068 	for (i = 0; hostname[i] != '\0'; i++) {
1069 		/*
1070 		 * As noted above, this deviates from RFC 1035 in that it
1071 		 * allows a leading digit.
1072 		 */
1073 		if (isalpha(hostname[i]) || isdigit(hostname[i]) ||
1074 		    (((hostname[i] == '-') || (hostname[i] == '.')) && (i > 0)))
1075 			continue;
1076 
1077 		return (B_FALSE);
1078 	}
1079 
1080 	return (B_TRUE);
1081 }
1082