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.
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.
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]
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2013, Joyent Inc. All rights reserved.
25 * Copyright (c) 2015 by Delphix. All rights reserved.
29 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
33 * This module contains functions used to bring up and tear down the
34 * Virtual Platform: [un]mounting file-systems, [un]plumbing network
35 * interfaces, [un]configuring devices, establishing resource controls,
36 * and creating/destroying the zone in the kernel. These actions, on
37 * the way up, ready the zone; on the way down, they halt the zone.
38 * See the much longer block comment at the beginning of zoneadmd.c
39 * for a bigger picture of how the whole program functions.
41 * This module also has primary responsibility for the layout of "scratch
42 * zones." These are mounted, but inactive, zones that are used during
43 * operating system upgrade and potentially other administrative action. The
44 * scratch zone environment is similar to the miniroot environment. The zone's
45 * actual root is mounted read-write on /a, and the standard paths (/usr,
46 * /sbin, /lib) all lead to read-only copies of the running system's binaries.
47 * This allows the administrative tools to manipulate the zone using "-R /a"
48 * without relying on any binaries in the zone itself.
50 * If the scratch zone is on an alternate root (Live Upgrade [LU] boot
51 * environment), then we must resolve the lofs mounts used there to uncover
52 * writable (unshared) resources. Shared resources, though, are always
53 * read-only. In addition, if the "same" zone with a different root path is
54 * currently running, then "/b" inside the zone points to the running zone's
55 * root. This allows LU to synchronize configuration files during the upgrade
58 * To construct this environment, this module creates a tmpfs mount on
59 * $ZONEPATH/lu. Inside this scratch area, the miniroot-like environment as
60 * described above is constructed on the fly. The zone is then created using
61 * $ZONEPATH/lu as the root.
63 * Note that scratch zones are inactive. The zone's bits are not running and
64 * likely cannot be run correctly until upgrade is done. Init is not running
65 * there, nor is SMF. Because of this, the "mounted" state of a scratch zone
66 * is not a part of the usual halt/ready/boot state machine.
69 #include <sys/param.h>
70 #include <sys/mount.h>
71 #include <sys/mntent.h>
72 #include <sys/socket.h>
73 #include <sys/utsname.h>
74 #include <sys/types.h>
76 #include <sys/sockio.h>
77 #include <sys/stropts.h>
79 #include <sys/systeminfo.h>
80 #include <sys/secflags.h>
83 #include <libdllink.h>
84 #include <libdlvlan.h>
87 #include <arpa/inet.h>
88 #include <netinet/in.h>
89 #include <net/route.h>
103 #include <libdevinfo.h>
106 #include <libcontract.h>
107 #include <libcontract_priv.h>
108 #include <uuid/uuid.h>
110 #include <sys/mntio.h>
111 #include <sys/mnttab.h>
112 #include <sys/fs/autofs.h> /* for _autofssys() */
113 #include <sys/fs/lofs_info.h>
114 #include <sys/fs/zfs.h>
117 #include <sys/pool.h>
118 #include <sys/priocntl.h>
120 #include <libbrand.h>
121 #include <sys/brand.h>
122 #include <libzonecfg.h>
125 #include "zoneadmd.h"
126 #include <sys/priv.h>
127 #include <libinetutil.h>
129 #define V4_ADDR_LEN 32
130 #define V6_ADDR_LEN 128
132 #define RESOURCE_DEFAULT_OPTS \
133 MNTOPT_RO "," MNTOPT_LOFS_NOSUB "," MNTOPT_NODEVICES
135 #define DFSTYPES "/etc/dfs/fstypes"
136 #define MAXTNZLEN 2048
138 #define ALT_MOUNT(mount_cmd) ((mount_cmd) != Z_MNT_BOOT)
140 /* a reasonable estimate for the number of lwps per process */
141 #define LWPS_PER_PROCESS 10
143 /* for routing socket */
144 static int rts_seqno
= 0;
146 /* mangled zone name when mounting in an alternate root environment */
147 static char kernzone
[ZONENAME_MAX
];
149 /* array of cached mount entries for resolve_lofs */
150 static struct mnttab
*resolve_lofs_mnts
, *resolve_lofs_mnt_max
;
152 static const char *DFLT_FS_ALLOWED
= "hsfs,smbfs,nfs,nfs3,nfs4,nfsdyn";
154 /* from libsocket, not in any header file */
155 extern int getnetmaskbyaddr(struct in_addr
, struct in_addr
*);
158 extern char query_hook
[];
161 * For each "net" resource configured in zonecfg, we track a zone_addr_list_t
162 * node in a linked list that is sorted by linkid. The list is constructed as
163 * the xml configuration file is parsed, and the information
164 * contained in each node is added to the kernel before the zone is
165 * booted, to be retrieved and applied from within the exclusive-IP NGZ
168 typedef struct zone_addr_list
{
169 struct zone_addr_list
*za_next
;
170 datalink_id_t za_linkid
; /* datalink_id_t of interface */
171 struct zone_nwiftab za_nwiftab
; /* address, defrouter properties */
175 * An optimization for build_mnttable: reallocate (and potentially copy the
176 * data) only once every N times through the loop.
178 #define MNTTAB_HUNK 32
180 /* some handy macros */
181 #define SIN(s) ((struct sockaddr_in *)s)
182 #define SIN6(s) ((struct sockaddr_in6 *)s)
185 * Private autofs system call
187 extern int _autofssys(int, void *);
190 autofs_cleanup(zoneid_t zoneid
)
193 * Ask autofs to unmount all trigger nodes in the given zone.
195 return (_autofssys(AUTOFS_UNMOUNTALL
, (void *)zoneid
));
199 free_mnttable(struct mnttab
*mnt_array
, uint_t nelem
)
203 if (mnt_array
== NULL
)
205 for (i
= 0; i
< nelem
; i
++) {
206 free(mnt_array
[i
].mnt_mountp
);
207 free(mnt_array
[i
].mnt_fstype
);
208 free(mnt_array
[i
].mnt_special
);
209 free(mnt_array
[i
].mnt_mntopts
);
210 assert(mnt_array
[i
].mnt_time
== NULL
);
216 * Build the mount table for the zone rooted at "zroot", storing the resulting
217 * array of struct mnttabs in "mnt_arrayp" and the number of elements in the
221 build_mnttable(zlog_t
*zlogp
, const char *zroot
, size_t zrootlen
, FILE *mnttab
,
222 struct mnttab
**mnt_arrayp
, uint_t
*nelemp
)
233 while (getmntent(mnttab
, &mnt
) == 0) {
234 struct mnttab
*tmp_array
;
236 if (strncmp(mnt
.mnt_mountp
, zroot
, zrootlen
) != 0)
238 if (nmnt
% MNTTAB_HUNK
== 0) {
239 tmp_array
= realloc(mnts
,
240 (nmnt
+ MNTTAB_HUNK
) * sizeof (*mnts
));
241 if (tmp_array
== NULL
) {
242 free_mnttable(mnts
, nmnt
);
250 * Zero out any fields we're not using.
252 (void) memset(mnp
, 0, sizeof (*mnp
));
254 if (mnt
.mnt_special
!= NULL
)
255 mnp
->mnt_special
= strdup(mnt
.mnt_special
);
256 if (mnt
.mnt_mntopts
!= NULL
)
257 mnp
->mnt_mntopts
= strdup(mnt
.mnt_mntopts
);
258 mnp
->mnt_mountp
= strdup(mnt
.mnt_mountp
);
259 mnp
->mnt_fstype
= strdup(mnt
.mnt_fstype
);
260 if ((mnt
.mnt_special
!= NULL
&& mnp
->mnt_special
== NULL
) ||
261 (mnt
.mnt_mntopts
!= NULL
&& mnp
->mnt_mntopts
== NULL
) ||
262 mnp
->mnt_mountp
== NULL
|| mnp
->mnt_fstype
== NULL
) {
263 zerror(zlogp
, B_TRUE
, "memory allocation failed");
264 free_mnttable(mnts
, nmnt
);
274 * This is an optimization. The resolve_lofs function is used quite frequently
275 * to manipulate file paths, and on a machine with a large number of zones,
276 * there will be a huge number of mounted file systems. Thus, we trigger a
277 * reread of the list of mount points
280 lofs_discard_mnttab(void)
282 free_mnttable(resolve_lofs_mnts
,
283 resolve_lofs_mnt_max
- resolve_lofs_mnts
);
284 resolve_lofs_mnts
= resolve_lofs_mnt_max
= NULL
;
288 lofs_read_mnttab(zlog_t
*zlogp
)
293 if ((mnttab
= fopen(MNTTAB
, "r")) == NULL
)
295 if (build_mnttable(zlogp
, "", 0, mnttab
, &resolve_lofs_mnts
,
297 (void) fclose(mnttab
);
300 (void) fclose(mnttab
);
301 resolve_lofs_mnt_max
= resolve_lofs_mnts
+ nmnts
;
306 * This function loops over potential loopback mounts and symlinks in a given
307 * path and resolves them all down to an absolute path.
310 resolve_lofs(zlog_t
*zlogp
, char *path
, size_t pathlen
)
314 char tmppath
[MAXPATHLEN
];
315 boolean_t outside_altroot
;
317 if ((len
= resolvepath(path
, tmppath
, sizeof (tmppath
))) == -1)
320 (void) strlcpy(path
, tmppath
, sizeof (tmppath
));
322 /* This happens once per zoneadmd operation. */
323 if (resolve_lofs_mnts
== NULL
&& lofs_read_mnttab(zlogp
) == -1)
326 altroot
= zonecfg_get_root();
327 arlen
= strlen(altroot
);
328 outside_altroot
= B_FALSE
;
332 /* Search in reverse order to find longest match */
333 for (mnp
= resolve_lofs_mnt_max
- 1; mnp
>= resolve_lofs_mnts
;
335 if (mnp
->mnt_fstype
== NULL
||
336 mnp
->mnt_mountp
== NULL
||
337 mnp
->mnt_special
== NULL
)
339 len
= strlen(mnp
->mnt_mountp
);
340 if (strncmp(mnp
->mnt_mountp
, path
, len
) == 0 &&
341 (path
[len
] == '/' || path
[len
] == '\0'))
344 if (mnp
< resolve_lofs_mnts
)
346 /* If it's not a lofs then we're done */
347 if (strcmp(mnp
->mnt_fstype
, MNTTYPE_LOFS
) != 0)
349 if (outside_altroot
) {
351 int olen
= sizeof (MNTOPT_RO
) - 1;
354 * If we run into a read-only mount outside of the
355 * alternate root environment, then the user doesn't
356 * want this path to be made read-write.
358 if (mnp
->mnt_mntopts
!= NULL
&&
359 (cp
= strstr(mnp
->mnt_mntopts
, MNTOPT_RO
)) !=
361 (cp
== mnp
->mnt_mntopts
|| cp
[-1] == ',') &&
362 (cp
[olen
] == '\0' || cp
[olen
] == ',')) {
365 } else if (arlen
> 0 &&
366 (strncmp(mnp
->mnt_special
, altroot
, arlen
) != 0 ||
367 (mnp
->mnt_special
[arlen
] != '\0' &&
368 mnp
->mnt_special
[arlen
] != '/'))) {
369 outside_altroot
= B_TRUE
;
371 /* use temporary buffer because new path might be longer */
372 (void) snprintf(tmppath
, sizeof (tmppath
), "%s%s",
373 mnp
->mnt_special
, path
+ len
);
374 if ((len
= resolvepath(tmppath
, path
, pathlen
)) == -1)
381 * For a regular mount, check if a replacement lofs mount is needed because the
382 * referenced device is already mounted somewhere.
385 check_lofs_needed(zlog_t
*zlogp
, struct zone_fstab
*fsptr
)
388 zone_fsopt_t
*optptr
, *onext
;
390 /* This happens once per zoneadmd operation. */
391 if (resolve_lofs_mnts
== NULL
&& lofs_read_mnttab(zlogp
) == -1)
395 * If this special node isn't already in use, then it's ours alone;
396 * no need to worry about conflicting mounts.
398 for (mnp
= resolve_lofs_mnts
; mnp
< resolve_lofs_mnt_max
;
400 if (strcmp(mnp
->mnt_special
, fsptr
->zone_fs_special
) == 0)
403 if (mnp
>= resolve_lofs_mnt_max
)
407 * Convert this duplicate mount into a lofs mount.
409 (void) strlcpy(fsptr
->zone_fs_special
, mnp
->mnt_mountp
,
410 sizeof (fsptr
->zone_fs_special
));
411 (void) strlcpy(fsptr
->zone_fs_type
, MNTTYPE_LOFS
,
412 sizeof (fsptr
->zone_fs_type
));
413 fsptr
->zone_fs_raw
[0] = '\0';
416 * Discard all but one of the original options and set that to our
417 * default set of options used for resources.
419 optptr
= fsptr
->zone_fs_options
;
420 if (optptr
== NULL
) {
421 optptr
= malloc(sizeof (*optptr
));
422 if (optptr
== NULL
) {
423 zerror(zlogp
, B_TRUE
, "cannot mount %s",
428 while ((onext
= optptr
->zone_fsopt_next
) != NULL
) {
429 optptr
->zone_fsopt_next
= onext
->zone_fsopt_next
;
433 (void) strcpy(optptr
->zone_fsopt_opt
, RESOURCE_DEFAULT_OPTS
);
434 optptr
->zone_fsopt_next
= NULL
;
435 fsptr
->zone_fs_options
= optptr
;
440 make_one_dir(zlog_t
*zlogp
, const char *prefix
, const char *subdir
, mode_t mode
,
441 uid_t userid
, gid_t groupid
)
443 char path
[MAXPATHLEN
];
446 if (snprintf(path
, sizeof (path
), "%s%s", prefix
, subdir
) >
448 zerror(zlogp
, B_FALSE
, "pathname %s%s is too long", prefix
,
453 if (lstat(path
, &st
) == 0) {
455 * We don't check the file mode since presumably the zone
456 * administrator may have had good reason to change the mode,
457 * and we don't need to second guess him.
459 if (!S_ISDIR(st
.st_mode
)) {
460 zerror(zlogp
, B_FALSE
, "%s is not a directory", path
);
466 if (mkdirp(path
, mode
) != 0) {
468 zerror(zlogp
, B_FALSE
, "Could not mkdir %s.\nIt is on "
469 "a read-only file system in this local zone.\nMake "
470 "sure %s exists in the global zone.", path
, subdir
);
472 zerror(zlogp
, B_TRUE
, "mkdirp of %s failed", path
);
476 (void) chown(path
, userid
, groupid
);
481 free_remote_fstypes(char **types
)
487 for (i
= 0; types
[i
] != NULL
; i
++)
493 get_remote_fstypes(zlog_t
*zlogp
)
497 char buf
[MAXPATHLEN
];
498 char fstype
[MAXPATHLEN
];
502 if ((fp
= fopen(DFSTYPES
, "r")) == NULL
) {
503 zerror(zlogp
, B_TRUE
, "failed to open %s", DFSTYPES
);
507 * Count the number of lines
509 while (fgets(buf
, sizeof (buf
), fp
) != NULL
)
511 if (lines
== 0) /* didn't read anything; empty file */
515 * Allocate enough space for a NULL-terminated array.
517 types
= calloc(lines
+ 1, sizeof (char *));
519 zerror(zlogp
, B_TRUE
, "memory allocation failed");
523 while (fgets(buf
, sizeof (buf
), fp
) != NULL
) {
524 /* LINTED - fstype is big enough to hold buf */
525 if (sscanf(buf
, "%s", fstype
) == 0) {
526 zerror(zlogp
, B_FALSE
, "unable to parse %s", DFSTYPES
);
527 free_remote_fstypes(types
);
531 types
[i
] = strdup(fstype
);
532 if (types
[i
] == NULL
) {
533 zerror(zlogp
, B_TRUE
, "memory allocation failed");
534 free_remote_fstypes(types
);
546 is_remote_fstype(const char *fstype
, char *const *remote_fstypes
)
550 if (remote_fstypes
== NULL
)
552 for (i
= 0; remote_fstypes
[i
] != NULL
; i
++) {
553 if (strcmp(remote_fstypes
[i
], fstype
) == 0)
560 * This converts a zone root path (normally of the form .../root) to a Live
561 * Upgrade scratch zone root (of the form .../lu).
564 root_to_lu(zlog_t
*zlogp
, char *zroot
, size_t zrootlen
, boolean_t isresolved
)
566 if (!isresolved
&& zonecfg_in_alt_root())
567 resolve_lofs(zlogp
, zroot
, zrootlen
);
568 (void) strcpy(strrchr(zroot
, '/') + 1, "lu");
572 * The general strategy for unmounting filesystems is as follows:
574 * - Remote filesystems may be dead, and attempting to contact them as
575 * part of a regular unmount may hang forever; we want to always try to
576 * forcibly unmount such filesystems and only fall back to regular
577 * unmounts if the filesystem doesn't support forced unmounts.
579 * - We don't want to unnecessarily corrupt metadata on local
580 * filesystems (ie UFS), so we want to start off with graceful unmounts,
581 * and only escalate to doing forced unmounts if we get stuck.
583 * We start off walking backwards through the mount table. This doesn't
584 * give us strict ordering but ensures that we try to unmount submounts
585 * first. We thus limit the number of failed umount2(2) calls.
587 * The mechanism for determining if we're stuck is to count the number
588 * of failed unmounts each iteration through the mount table. This
589 * gives us an upper bound on the number of filesystems which remain
590 * mounted (autofs trigger nodes are dealt with separately). If at the
591 * end of one unmount+autofs_cleanup cycle we still have the same number
592 * of mounts that we started out with, we're stuck and try a forced
593 * unmount. If that fails (filesystem doesn't support forced unmounts)
594 * then we bail and are unable to teardown the zone. If it succeeds,
595 * we're no longer stuck so we continue with our policy of trying
596 * graceful mounts first.
598 * Zone must be down (ie, no processes or threads active).
601 unmount_filesystems(zlog_t
*zlogp
, zoneid_t zoneid
, boolean_t unmount_cmd
)
607 char zroot
[MAXPATHLEN
+ 1];
609 uint_t oldcount
= UINT_MAX
;
610 boolean_t stuck
= B_FALSE
;
611 char **remote_fstypes
= NULL
;
613 if (zone_get_rootpath(zone_name
, zroot
, sizeof (zroot
)) != Z_OK
) {
614 zerror(zlogp
, B_FALSE
, "unable to determine zone root");
618 root_to_lu(zlogp
, zroot
, sizeof (zroot
), B_FALSE
);
620 (void) strcat(zroot
, "/");
621 zrootlen
= strlen(zroot
);
623 if ((mnttab
= fopen(MNTTAB
, "r")) == NULL
) {
624 zerror(zlogp
, B_TRUE
, "failed to open %s", MNTTAB
);
628 * Use our hacky mntfs ioctl so we see everything, even mounts with
631 if (ioctl(fileno(mnttab
), MNTIOC_SHOWHIDDEN
, NULL
) < 0) {
632 zerror(zlogp
, B_TRUE
, "unable to configure %s", MNTTAB
);
638 * Build the list of remote fstypes so we know which ones we
639 * should forcibly unmount.
641 remote_fstypes
= get_remote_fstypes(zlogp
);
642 for (; /* ever */; ) {
652 * MNTTAB gives us a way to walk through mounted
653 * filesystems; we need to be able to walk them in
654 * reverse order, so we build a list of all mounted
657 if (build_mnttable(zlogp
, zroot
, zrootlen
, mnttab
, &mnts
,
662 for (i
= 0; i
< nmnt
; i
++) {
663 mnp
= &mnts
[nmnt
- i
- 1]; /* access in reverse order */
664 path
= mnp
->mnt_mountp
;
667 * Try forced unmount first for remote filesystems.
669 * Not all remote filesystems support forced unmounts,
670 * so if this fails (ENOTSUP) we'll continue on
671 * and try a regular unmount.
673 if (is_remote_fstype(mnp
->mnt_fstype
, remote_fstypes
)) {
674 if (umount2(path
, MS_FORCE
) == 0)
678 * Try forced unmount if we're stuck.
681 if (umount2(path
, MS_FORCE
) == 0) {
686 * The first failure indicates a
687 * mount we won't be able to get
688 * rid of automatically, so we
692 zerror(zlogp
, B_FALSE
,
693 "unable to unmount '%s'", path
);
694 free_mnttable(mnts
, nmnt
);
699 * Try regular unmounts for everything else.
701 if (!unmounted
&& umount2(path
, 0) != 0)
704 free_mnttable(mnts
, nmnt
);
708 if (newcount
>= oldcount
) {
710 * Last round didn't unmount anything; we're stuck and
711 * should start trying forced unmounts.
718 * Autofs doesn't let you unmount its trigger nodes from
719 * userland so we have to tell the kernel to cleanup for us.
721 if (autofs_cleanup(zoneid
) != 0) {
722 zerror(zlogp
, B_TRUE
, "unable to remove autofs nodes");
729 free_remote_fstypes(remote_fstypes
);
730 (void) fclose(mnttab
);
731 return (error
? -1 : 0);
735 fs_compare(const void *m1
, const void *m2
)
737 struct zone_fstab
*i
= (struct zone_fstab
*)m1
;
738 struct zone_fstab
*j
= (struct zone_fstab
*)m2
;
740 return (strcmp(i
->zone_fs_dir
, j
->zone_fs_dir
));
744 * Fork and exec (and wait for) the mentioned binary with the provided
745 * arguments. Returns (-1) if something went wrong with fork(2) or exec(2),
746 * returns the exit status otherwise.
748 * If we were unable to exec the provided pathname (for whatever
749 * reason), we return the special token ZEXIT_EXEC. The current value
750 * of ZEXIT_EXEC doesn't conflict with legitimate exit codes of the
751 * consumers of this function; any future consumers must make sure this
755 forkexec(zlog_t
*zlogp
, const char *path
, char *const argv
[])
758 int child_status
= 0;
761 * Do not let another thread localize a message while we are forking.
763 (void) mutex_lock(&msglock
);
765 (void) mutex_unlock(&msglock
);
766 if (child_pid
== -1) {
767 zerror(zlogp
, B_TRUE
, "could not fork for %s", argv
[0]);
769 } else if (child_pid
== 0) {
771 /* redirect stdin, stdout & stderr to /dev/null */
772 (void) open("/dev/null", O_RDONLY
); /* stdin */
773 (void) open("/dev/null", O_WRONLY
); /* stdout */
774 (void) open("/dev/null", O_WRONLY
); /* stderr */
775 (void) execv(path
, argv
);
777 * Since we are in the child, there is no point calling zerror()
778 * since there is nobody waiting to consume it. So exit with a
779 * special code that the parent will recognize and call zerror()
785 (void) waitpid(child_pid
, &child_status
, 0);
788 if (WIFSIGNALED(child_status
)) {
789 zerror(zlogp
, B_FALSE
, "%s unexpectedly terminated due to "
790 "signal %d", path
, WTERMSIG(child_status
));
793 assert(WIFEXITED(child_status
));
794 if (WEXITSTATUS(child_status
) == ZEXIT_EXEC
) {
795 zerror(zlogp
, B_FALSE
, "failed to exec %s", path
);
798 return (WEXITSTATUS(child_status
));
802 isregfile(const char *path
)
806 if (stat64(path
, &st
) == -1)
809 return (S_ISREG(st
.st_mode
));
813 dofsck(zlog_t
*zlogp
, const char *fstype
, const char *rawdev
)
815 char cmdbuf
[MAXPATHLEN
];
820 * We could alternatively have called /usr/sbin/fsck -F <fstype>, but
821 * that would cost us an extra fork/exec without buying us anything.
823 if (snprintf(cmdbuf
, sizeof (cmdbuf
), "/usr/lib/fs/%s/fsck", fstype
)
824 >= sizeof (cmdbuf
)) {
825 zerror(zlogp
, B_FALSE
, "file-system type %s too long", fstype
);
830 * If it doesn't exist, that's OK: we verified this previously
833 if (isregfile(cmdbuf
) == -1)
839 argv
[3] = (char *)rawdev
;
842 status
= forkexec(zlogp
, cmdbuf
, argv
);
843 if (status
== 0 || status
== -1)
845 zerror(zlogp
, B_FALSE
, "fsck of '%s' failed with exit status %d; "
846 "run fsck manually", rawdev
, status
);
851 domount(zlog_t
*zlogp
, const char *fstype
, const char *opts
,
852 const char *special
, const char *directory
)
854 char cmdbuf
[MAXPATHLEN
];
859 * We could alternatively have called /usr/sbin/mount -F <fstype>, but
860 * that would cost us an extra fork/exec without buying us anything.
862 if (snprintf(cmdbuf
, sizeof (cmdbuf
), "/usr/lib/fs/%s/mount", fstype
)
863 >= sizeof (cmdbuf
)) {
864 zerror(zlogp
, B_FALSE
, "file-system type %s too long", fstype
);
868 if (opts
[0] == '\0') {
869 argv
[1] = (char *)special
;
870 argv
[2] = (char *)directory
;
874 argv
[2] = (char *)opts
;
875 argv
[3] = (char *)special
;
876 argv
[4] = (char *)directory
;
880 status
= forkexec(zlogp
, cmdbuf
, argv
);
881 if (status
== 0 || status
== -1)
884 zerror(zlogp
, B_FALSE
, "\"%s %s %s\" "
885 "failed with exit code %d",
886 cmdbuf
, special
, directory
, status
);
888 zerror(zlogp
, B_FALSE
, "\"%s -o %s %s %s\" "
889 "failed with exit code %d",
890 cmdbuf
, opts
, special
, directory
, status
);
895 * Check if a given mount point path exists.
896 * If it does, make sure it doesn't contain any symlinks.
897 * Note that if "leaf" is false we're checking an intermediate
898 * component of the mount point path, so it must be a directory.
899 * If "leaf" is true, then we're checking the entire mount point
900 * path, so the mount point itself can be anything aside from a
903 * If the path is invalid then a negative value is returned. If the
904 * path exists and is a valid mount point path then 0 is returned.
905 * If the path doesn't exist return a positive value.
908 valid_mount_point(zlog_t
*zlogp
, const char *path
, const boolean_t leaf
)
911 char respath
[MAXPATHLEN
];
914 if (lstat(path
, &statbuf
) != 0) {
917 zerror(zlogp
, B_TRUE
, "can't stat %s", path
);
920 if (S_ISLNK(statbuf
.st_mode
)) {
921 zerror(zlogp
, B_FALSE
, "%s is a symlink", path
);
924 if (!leaf
&& !S_ISDIR(statbuf
.st_mode
)) {
925 zerror(zlogp
, B_FALSE
, "%s is not a directory", path
);
928 if ((res
= resolvepath(path
, respath
, sizeof (respath
))) == -1) {
929 zerror(zlogp
, B_TRUE
, "unable to resolve path %s", path
);
933 if (strcmp(path
, respath
) != 0) {
935 * We don't like ".."s, "."s, or "//"s throwing us off
937 zerror(zlogp
, B_FALSE
, "%s is not a canonical path", path
);
944 * Validate a mount point path. A valid mount point path is an
945 * absolute path that either doesn't exist, or, if it does exists it
946 * must be an absolute canonical path that doesn't have any symbolic
947 * links in it. The target of a mount point path can be any filesystem
948 * object. (Different filesystems can support different mount points,
949 * for example "lofs" and "mntfs" both support files and directories
950 * while "ufs" just supports directories.)
952 * If the path is invalid then a negative value is returned. If the
953 * path exists and is a valid mount point path then 0 is returned.
954 * If the path doesn't exist return a positive value.
957 valid_mount_path(zlog_t
*zlogp
, const char *rootpath
, const char *spec
,
958 const char *dir
, const char *fstype
)
960 char abspath
[MAXPATHLEN
], *slashp
, *slashp_next
;
964 * Sanity check the target mount point path.
965 * It must be a non-null string that starts with a '/'.
968 /* Something went wrong. */
969 zerror(zlogp
, B_FALSE
, "invalid mount directory, "
970 "type: \"%s\", special: \"%s\", dir: \"%s\"",
976 * Join rootpath and dir. Make sure abspath ends with '/', this
977 * is added to all paths (even non-directory paths) to allow us
978 * to detect the end of paths below. If the path already ends
979 * in a '/', then that's ok too (although we'll fail the
980 * cannonical path check in valid_mount_point()).
982 if (snprintf(abspath
, sizeof (abspath
),
983 "%s%s/", rootpath
, dir
) >= sizeof (abspath
)) {
984 zerror(zlogp
, B_FALSE
, "pathname %s%s is too long",
990 * Starting with rootpath, verify the mount path one component
991 * at a time. Continue until we've evaluated all of abspath.
993 slashp
= &abspath
[strlen(rootpath
)];
994 assert(*slashp
== '/');
996 slashp_next
= strchr(slashp
+ 1, '/');
998 if (slashp_next
!= NULL
) {
999 /* This is an intermediary mount path component. */
1000 rv
= valid_mount_point(zlogp
, abspath
, B_FALSE
);
1002 /* This is the last component of the mount path. */
1003 rv
= valid_mount_point(zlogp
, abspath
, B_TRUE
);
1008 } while ((slashp
= slashp_next
) != NULL
);
1013 mount_one_dev_device_cb(void *arg
, const char *match
, const char *name
)
1015 di_prof_t prof
= arg
;
1018 return (di_prof_add_dev(prof
, match
));
1019 return (di_prof_add_map(prof
, match
, name
));
1023 mount_one_dev_symlink_cb(void *arg
, const char *source
, const char *target
)
1025 di_prof_t prof
= arg
;
1027 return (di_prof_add_symlink(prof
, source
, target
));
1031 vplat_get_iptype(zlog_t
*zlogp
, zone_iptype_t
*iptypep
)
1033 zone_dochandle_t handle
;
1035 if ((handle
= zonecfg_init_handle()) == NULL
) {
1036 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
1039 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
) {
1040 zerror(zlogp
, B_FALSE
, "invalid configuration");
1041 zonecfg_fini_handle(handle
);
1044 if (zonecfg_get_iptype(handle
, iptypep
) != Z_OK
) {
1045 zerror(zlogp
, B_FALSE
, "invalid ip-type configuration");
1046 zonecfg_fini_handle(handle
);
1049 zonecfg_fini_handle(handle
);
1054 * Apply the standard lists of devices/symlinks/mappings and the user-specified
1055 * list of devices (via zonecfg) to the /dev filesystem. The filesystem will
1056 * use these as a profile/filter to determine what exists in /dev.
1059 mount_one_dev(zlog_t
*zlogp
, char *devpath
, zone_mnt_t mount_cmd
)
1061 char brand
[MAXNAMELEN
];
1062 zone_dochandle_t handle
= NULL
;
1063 brand_handle_t bh
= NULL
;
1064 struct zone_devtab ztab
;
1065 di_prof_t prof
= NULL
;
1068 zone_iptype_t iptype
;
1069 const char *curr_iptype
;
1071 if (di_prof_init(devpath
, &prof
)) {
1072 zerror(zlogp
, B_TRUE
, "failed to initialize profile");
1077 * Get a handle to the brand info for this zone.
1078 * If we are mounting the zone, then we must always use the default
1079 * brand device mounts.
1081 if (ALT_MOUNT(mount_cmd
)) {
1082 (void) strlcpy(brand
, default_brand
, sizeof (brand
));
1084 (void) strlcpy(brand
, brand_name
, sizeof (brand
));
1087 if ((bh
= brand_open(brand
)) == NULL
) {
1088 zerror(zlogp
, B_FALSE
, "unable to determine zone brand");
1092 if (vplat_get_iptype(zlogp
, &iptype
) < 0) {
1093 zerror(zlogp
, B_TRUE
, "unable to determine ip-type");
1098 curr_iptype
= "shared";
1101 curr_iptype
= "exclusive";
1105 if (brand_platform_iter_devices(bh
, zone_name
,
1106 mount_one_dev_device_cb
, prof
, curr_iptype
) != 0) {
1107 zerror(zlogp
, B_TRUE
, "failed to add standard device");
1111 if (brand_platform_iter_link(bh
,
1112 mount_one_dev_symlink_cb
, prof
) != 0) {
1113 zerror(zlogp
, B_TRUE
, "failed to add standard symlink");
1117 /* Add user-specified devices and directories */
1118 if ((handle
= zonecfg_init_handle()) == NULL
) {
1119 zerror(zlogp
, B_FALSE
, "can't initialize zone handle");
1122 if (err
= zonecfg_get_handle(zone_name
, handle
)) {
1123 zerror(zlogp
, B_FALSE
, "can't get handle for zone "
1124 "%s: %s", zone_name
, zonecfg_strerror(err
));
1127 if (err
= zonecfg_setdevent(handle
)) {
1128 zerror(zlogp
, B_FALSE
, "%s: %s", zone_name
,
1129 zonecfg_strerror(err
));
1132 while (zonecfg_getdevent(handle
, &ztab
) == Z_OK
) {
1133 if (di_prof_add_dev(prof
, ztab
.zone_dev_match
)) {
1134 zerror(zlogp
, B_TRUE
, "failed to add "
1135 "user-specified device");
1139 (void) zonecfg_enddevent(handle
);
1141 /* Send profile to kernel */
1142 if (di_prof_commit(prof
)) {
1143 zerror(zlogp
, B_TRUE
, "failed to commit profile");
1153 zonecfg_fini_handle(handle
);
1160 mount_one(zlog_t
*zlogp
, struct zone_fstab
*fsptr
, const char *rootpath
,
1161 zone_mnt_t mount_cmd
)
1163 char path
[MAXPATHLEN
];
1164 char optstr
[MAX_MNTOPT_STR
];
1165 zone_fsopt_t
*optptr
;
1168 if ((rv
= valid_mount_path(zlogp
, rootpath
, fsptr
->zone_fs_special
,
1169 fsptr
->zone_fs_dir
, fsptr
->zone_fs_type
)) < 0) {
1170 zerror(zlogp
, B_FALSE
, "%s%s is not a valid mount point",
1171 rootpath
, fsptr
->zone_fs_dir
);
1173 } else if (rv
> 0) {
1174 /* The mount point path doesn't exist, create it now. */
1175 if (make_one_dir(zlogp
, rootpath
, fsptr
->zone_fs_dir
,
1176 DEFAULT_DIR_MODE
, DEFAULT_DIR_USER
,
1177 DEFAULT_DIR_GROUP
) != 0) {
1178 zerror(zlogp
, B_FALSE
, "failed to create mount point");
1183 * Now this might seem weird, but we need to invoke
1184 * valid_mount_path() again. Why? Because it checks
1185 * to make sure that the mount point path is canonical,
1186 * which it can only do if the path exists, so now that
1187 * we've created the path we have to verify it again.
1189 if ((rv
= valid_mount_path(zlogp
, rootpath
,
1190 fsptr
->zone_fs_special
, fsptr
->zone_fs_dir
,
1191 fsptr
->zone_fs_type
)) < 0) {
1192 zerror(zlogp
, B_FALSE
,
1193 "%s%s is not a valid mount point",
1194 rootpath
, fsptr
->zone_fs_dir
);
1199 (void) snprintf(path
, sizeof (path
), "%s%s", rootpath
,
1200 fsptr
->zone_fs_dir
);
1203 * In general the strategy here is to do just as much verification as
1204 * necessary to avoid crashing or otherwise doing something bad; if the
1205 * administrator initiated the operation via zoneadm(1m), he'll get
1206 * auto-verification which will let him know what's wrong. If he
1207 * modifies the zone configuration of a running zone and doesn't attempt
1208 * to verify that it's OK we won't crash but won't bother trying to be
1209 * too helpful either. zoneadm verify is only a couple keystrokes away.
1211 if (!zonecfg_valid_fs_type(fsptr
->zone_fs_type
)) {
1212 zerror(zlogp
, B_FALSE
, "cannot mount %s on %s: "
1213 "invalid file-system type %s", fsptr
->zone_fs_special
,
1214 fsptr
->zone_fs_dir
, fsptr
->zone_fs_type
);
1219 * If we're looking at an alternate root environment, then construct
1220 * read-only loopback mounts as necessary. Note that any special
1221 * paths for lofs zone mounts in an alternate root must have
1222 * already been pre-pended with any alternate root path by the
1225 if (zonecfg_in_alt_root()) {
1228 if (stat64(fsptr
->zone_fs_special
, &st
) != -1 &&
1229 S_ISBLK(st
.st_mode
)) {
1231 * If we're going to mount a block device we need
1232 * to check if that device is already mounted
1233 * somewhere else, and if so, do a lofs mount
1234 * of the device instead of a direct mount
1236 if (check_lofs_needed(zlogp
, fsptr
) == -1)
1238 } else if (strcmp(fsptr
->zone_fs_type
, MNTTYPE_LOFS
) == 0) {
1240 * For lofs mounts, the special node is inside the
1241 * alternate root. We need lofs resolution for
1242 * this case in order to get at the underlying
1245 resolve_lofs(zlogp
, fsptr
->zone_fs_special
,
1246 sizeof (fsptr
->zone_fs_special
));
1251 * Run 'fsck -m' if there's a device to fsck.
1253 if (fsptr
->zone_fs_raw
[0] != '\0' &&
1254 dofsck(zlogp
, fsptr
->zone_fs_type
, fsptr
->zone_fs_raw
) != 0) {
1256 } else if (isregfile(fsptr
->zone_fs_special
) == 1 &&
1257 dofsck(zlogp
, fsptr
->zone_fs_type
, fsptr
->zone_fs_special
) != 0) {
1262 * Build up mount option string.
1265 if (fsptr
->zone_fs_options
!= NULL
) {
1266 (void) strlcpy(optstr
, fsptr
->zone_fs_options
->zone_fsopt_opt
,
1268 for (optptr
= fsptr
->zone_fs_options
->zone_fsopt_next
;
1269 optptr
!= NULL
; optptr
= optptr
->zone_fsopt_next
) {
1270 (void) strlcat(optstr
, ",", sizeof (optstr
));
1271 (void) strlcat(optstr
, optptr
->zone_fsopt_opt
,
1276 if ((rv
= domount(zlogp
, fsptr
->zone_fs_type
, optstr
,
1277 fsptr
->zone_fs_special
, path
)) != 0)
1281 * The mount succeeded. If this was not a mount of /dev then
1284 if (strcmp(fsptr
->zone_fs_type
, MNTTYPE_DEV
) != 0)
1288 * We just mounted an instance of a /dev filesystem, so now we
1289 * need to configure it.
1291 return (mount_one_dev(zlogp
, path
, mount_cmd
));
1295 free_fs_data(struct zone_fstab
*fsarray
, uint_t nelem
)
1299 if (fsarray
== NULL
)
1301 for (i
= 0; i
< nelem
; i
++)
1302 zonecfg_free_fs_option_list(fsarray
[i
].zone_fs_options
);
1307 * This function initiates the creation of a small Solaris Environment for
1308 * scratch zone. The Environment creation process is split up into two
1309 * functions(build_mounted_pre_var() and build_mounted_post_var()). It
1310 * is done this way because:
1311 * We need to have both /etc and /var in the root of the scratchzone.
1312 * We loopback mount zone's own /etc and /var into the root of the
1313 * scratch zone. Unlike /etc, /var can be a seperate filesystem. So we
1314 * need to delay the mount of /var till the zone's root gets populated.
1315 * So mounting of localdirs[](/etc and /var) have been moved to the
1316 * build_mounted_post_var() which gets called only after the zone
1317 * specific filesystems are mounted.
1319 * Note that the scratch zone we set up for updating the zone (Z_MNT_UPDATE)
1320 * does not loopback mount the zone's own /etc and /var into the root of the
1324 build_mounted_pre_var(zlog_t
*zlogp
, char *rootpath
,
1325 size_t rootlen
, const char *zonepath
, char *luroot
, size_t lurootlen
)
1327 char tmp
[MAXPATHLEN
], fromdir
[MAXPATHLEN
];
1329 static const char *mkdirs
[] = {
1330 "/system", "/system/contract", "/system/object", "/proc",
1331 "/dev", "/tmp", "/a", NULL
1337 resolve_lofs(zlogp
, rootpath
, rootlen
);
1338 (void) snprintf(luroot
, lurootlen
, "%s/lu", zonepath
);
1339 resolve_lofs(zlogp
, luroot
, lurootlen
);
1340 (void) snprintf(tmp
, sizeof (tmp
), "%s/bin", luroot
);
1341 (void) symlink("./usr/bin", tmp
);
1344 * These are mostly special mount points; not handled here. (See
1345 * zone_mount_early.)
1347 for (cpp
= mkdirs
; *cpp
!= NULL
; cpp
++) {
1348 (void) snprintf(tmp
, sizeof (tmp
), "%s%s", luroot
, *cpp
);
1349 if (mkdir(tmp
, 0755) != 0) {
1350 zerror(zlogp
, B_TRUE
, "cannot create %s", tmp
);
1355 * This is here to support lucopy. If there's an instance of this same
1356 * zone on the current running system, then we mount its root up as
1357 * read-only inside the scratch zone.
1359 (void) zonecfg_get_uuid(zone_name
, uuid
);
1360 altstr
= strdup(zonecfg_get_root());
1361 if (altstr
== NULL
) {
1362 zerror(zlogp
, B_TRUE
, "memory allocation failed");
1365 zonecfg_set_root("");
1366 (void) strlcpy(tmp
, zone_name
, sizeof (tmp
));
1367 (void) zonecfg_get_name_by_uuid(uuid
, tmp
, sizeof (tmp
));
1368 if (zone_get_rootpath(tmp
, fromdir
, sizeof (fromdir
)) == Z_OK
&&
1369 strcmp(fromdir
, rootpath
) != 0) {
1370 (void) snprintf(tmp
, sizeof (tmp
), "%s/b", luroot
);
1371 if (mkdir(tmp
, 0755) != 0) {
1372 zerror(zlogp
, B_TRUE
, "cannot create %s", tmp
);
1375 if (domount(zlogp
, MNTTYPE_LOFS
, RESOURCE_DEFAULT_OPTS
, fromdir
,
1377 zerror(zlogp
, B_TRUE
, "cannot mount %s on %s", tmp
,
1382 zonecfg_set_root(altstr
);
1385 if ((fp
= zonecfg_open_scratch(luroot
, B_TRUE
)) == NULL
) {
1386 zerror(zlogp
, B_TRUE
, "cannot open zone mapfile");
1389 (void) ftruncate(fileno(fp
), 0);
1390 if (zonecfg_add_scratch(fp
, zone_name
, kernzone
, "/") == -1) {
1391 zerror(zlogp
, B_TRUE
, "cannot add zone mapfile entry");
1393 zonecfg_close_scratch(fp
);
1394 (void) snprintf(tmp
, sizeof (tmp
), "%s/a", luroot
);
1395 if (domount(zlogp
, MNTTYPE_LOFS
, "", rootpath
, tmp
) != 0)
1397 (void) strlcpy(rootpath
, tmp
, rootlen
);
1403 build_mounted_post_var(zlog_t
*zlogp
, zone_mnt_t mount_cmd
, char *rootpath
,
1406 char tmp
[MAXPATHLEN
], fromdir
[MAXPATHLEN
];
1408 const char **loopdirs
;
1409 const char **tmpdirs
;
1410 static const char *localdirs
[] = {
1411 "/etc", "/var", NULL
1413 static const char *scr_loopdirs
[] = {
1414 "/etc/lib", "/etc/fs", "/lib", "/sbin", "/platform",
1417 static const char *upd_loopdirs
[] = {
1418 "/etc", "/kernel", "/lib", "/opt", "/platform", "/sbin",
1419 "/usr", "/var", NULL
1421 static const char *scr_tmpdirs
[] = {
1422 "/tmp", "/var/run", NULL
1424 static const char *upd_tmpdirs
[] = {
1425 "/tmp", "/var/run", "/var/tmp", NULL
1429 if (mount_cmd
== Z_MNT_SCRATCH
) {
1431 * These are mounted read-write from the zone undergoing
1432 * upgrade. We must be careful not to 'leak' things from the
1433 * main system into the zone, and this accomplishes that goal.
1435 for (cpp
= localdirs
; *cpp
!= NULL
; cpp
++) {
1436 (void) snprintf(tmp
, sizeof (tmp
), "%s%s", luroot
,
1438 (void) snprintf(fromdir
, sizeof (fromdir
), "%s%s",
1440 if (mkdir(tmp
, 0755) != 0) {
1441 zerror(zlogp
, B_TRUE
, "cannot create %s", tmp
);
1444 if (domount(zlogp
, MNTTYPE_LOFS
, "", fromdir
, tmp
)
1446 zerror(zlogp
, B_TRUE
, "cannot mount %s on %s",
1453 if (mount_cmd
== Z_MNT_UPDATE
)
1454 loopdirs
= upd_loopdirs
;
1456 loopdirs
= scr_loopdirs
;
1459 * These are things mounted read-only from the running system because
1460 * they contain binaries that must match system.
1462 for (cpp
= loopdirs
; *cpp
!= NULL
; cpp
++) {
1463 (void) snprintf(tmp
, sizeof (tmp
), "%s%s", luroot
, *cpp
);
1464 if (mkdir(tmp
, 0755) != 0) {
1465 if (errno
!= EEXIST
) {
1466 zerror(zlogp
, B_TRUE
, "cannot create %s", tmp
);
1469 if (lstat(tmp
, &st
) != 0) {
1470 zerror(zlogp
, B_TRUE
, "cannot stat %s", tmp
);
1474 * Ignore any non-directories encountered. These are
1475 * things that have been converted into symlinks
1476 * (/etc/fs and /etc/lib) and no longer need a lofs
1479 if (!S_ISDIR(st
.st_mode
))
1482 if (domount(zlogp
, MNTTYPE_LOFS
, RESOURCE_DEFAULT_OPTS
, *cpp
,
1484 zerror(zlogp
, B_TRUE
, "cannot mount %s on %s", tmp
,
1490 if (mount_cmd
== Z_MNT_UPDATE
)
1491 tmpdirs
= upd_tmpdirs
;
1493 tmpdirs
= scr_tmpdirs
;
1496 * These are things with tmpfs mounted inside.
1498 for (cpp
= tmpdirs
; *cpp
!= NULL
; cpp
++) {
1499 (void) snprintf(tmp
, sizeof (tmp
), "%s%s", luroot
, *cpp
);
1500 if (mount_cmd
== Z_MNT_SCRATCH
&& mkdir(tmp
, 0755) != 0 &&
1502 zerror(zlogp
, B_TRUE
, "cannot create %s", tmp
);
1507 * We could set the mode for /tmp when we do the mkdir but
1508 * since that can be modified by the umask we will just set
1509 * the correct mode for /tmp now.
1511 if (strcmp(*cpp
, "/tmp") == 0 && chmod(tmp
, 01777) != 0) {
1512 zerror(zlogp
, B_TRUE
, "cannot chmod %s", tmp
);
1516 if (domount(zlogp
, MNTTYPE_TMPFS
, "", "swap", tmp
) != 0) {
1517 zerror(zlogp
, B_TRUE
, "cannot mount swap on %s", *cpp
);
1524 typedef struct plat_gmount_cb_data
{
1526 struct zone_fstab
**pgcd_fs_tab
;
1528 } plat_gmount_cb_data_t
;
1531 * plat_gmount_cb() is a callback function invoked by libbrand to iterate
1532 * through all global brand platform mounts.
1535 plat_gmount_cb(void *data
, const char *spec
, const char *dir
,
1536 const char *fstype
, const char *opt
)
1538 plat_gmount_cb_data_t
*cp
= data
;
1539 zlog_t
*zlogp
= cp
->pgcd_zlogp
;
1540 struct zone_fstab
*fs_ptr
= *cp
->pgcd_fs_tab
;
1541 int num_fs
= *cp
->pgcd_num_fs
;
1542 struct zone_fstab
*fsp
, *tmp_ptr
;
1545 if ((tmp_ptr
= realloc(fs_ptr
, num_fs
* sizeof (*tmp_ptr
))) == NULL
) {
1546 zerror(zlogp
, B_TRUE
, "memory allocation failed");
1551 fsp
= &fs_ptr
[num_fs
- 1];
1553 /* update the callback struct passed in */
1554 *cp
->pgcd_fs_tab
= fs_ptr
;
1555 *cp
->pgcd_num_fs
= num_fs
;
1557 fsp
->zone_fs_raw
[0] = '\0';
1558 (void) strlcpy(fsp
->zone_fs_special
, spec
,
1559 sizeof (fsp
->zone_fs_special
));
1560 (void) strlcpy(fsp
->zone_fs_dir
, dir
, sizeof (fsp
->zone_fs_dir
));
1561 (void) strlcpy(fsp
->zone_fs_type
, fstype
, sizeof (fsp
->zone_fs_type
));
1562 fsp
->zone_fs_options
= NULL
;
1563 if ((opt
!= NULL
) &&
1564 (zonecfg_add_fs_option(fsp
, (char *)opt
) != Z_OK
)) {
1565 zerror(zlogp
, B_FALSE
, "error adding property");
1573 mount_filesystems_fsent(zone_dochandle_t handle
, zlog_t
*zlogp
,
1574 struct zone_fstab
**fs_tabp
, int *num_fsp
, zone_mnt_t mount_cmd
)
1576 struct zone_fstab
*tmp_ptr
, *fs_ptr
, *fsp
, fstab
;
1582 if (zonecfg_setfsent(handle
) != Z_OK
) {
1583 zerror(zlogp
, B_FALSE
, "invalid configuration");
1586 while (zonecfg_getfsent(handle
, &fstab
) == Z_OK
) {
1588 * ZFS filesystems will not be accessible under an alternate
1589 * root, since the pool will not be known. Ignore them in this
1592 if (ALT_MOUNT(mount_cmd
) &&
1593 strcmp(fstab
.zone_fs_type
, MNTTYPE_ZFS
) == 0)
1597 if ((tmp_ptr
= realloc(fs_ptr
,
1598 num_fs
* sizeof (*tmp_ptr
))) == NULL
) {
1599 zerror(zlogp
, B_TRUE
, "memory allocation failed");
1600 (void) zonecfg_endfsent(handle
);
1603 /* update the pointers passed in */
1608 fsp
= &fs_ptr
[num_fs
- 1];
1609 (void) strlcpy(fsp
->zone_fs_dir
,
1610 fstab
.zone_fs_dir
, sizeof (fsp
->zone_fs_dir
));
1611 (void) strlcpy(fsp
->zone_fs_raw
, fstab
.zone_fs_raw
,
1612 sizeof (fsp
->zone_fs_raw
));
1613 (void) strlcpy(fsp
->zone_fs_type
, fstab
.zone_fs_type
,
1614 sizeof (fsp
->zone_fs_type
));
1615 fsp
->zone_fs_options
= fstab
.zone_fs_options
;
1618 * For all lofs mounts, make sure that the 'special'
1619 * entry points inside the alternate root. The
1620 * source path for a lofs mount in a given zone needs
1621 * to be relative to the root of the boot environment
1622 * that contains the zone. Note that we don't do this
1623 * for non-lofs mounts since they will have a device
1624 * as a backing store and device paths must always be
1625 * specified relative to the current boot environment.
1627 fsp
->zone_fs_special
[0] = '\0';
1628 if (strcmp(fsp
->zone_fs_type
, MNTTYPE_LOFS
) == 0) {
1629 (void) strlcat(fsp
->zone_fs_special
, zonecfg_get_root(),
1630 sizeof (fsp
->zone_fs_special
));
1632 (void) strlcat(fsp
->zone_fs_special
, fstab
.zone_fs_special
,
1633 sizeof (fsp
->zone_fs_special
));
1635 (void) zonecfg_endfsent(handle
);
1640 mount_filesystems(zlog_t
*zlogp
, zone_mnt_t mount_cmd
)
1642 char rootpath
[MAXPATHLEN
];
1643 char zonepath
[MAXPATHLEN
];
1644 char brand
[MAXNAMELEN
];
1645 char luroot
[MAXPATHLEN
];
1647 struct zone_fstab
*fs_ptr
= NULL
;
1648 zone_dochandle_t handle
= NULL
;
1649 zone_state_t zstate
;
1651 plat_gmount_cb_data_t cb
;
1653 if (zone_get_state(zone_name
, &zstate
) != Z_OK
||
1654 (zstate
!= ZONE_STATE_READY
&& zstate
!= ZONE_STATE_MOUNTED
)) {
1655 zerror(zlogp
, B_FALSE
,
1656 "zone must be in '%s' or '%s' state to mount file-systems",
1657 zone_state_str(ZONE_STATE_READY
),
1658 zone_state_str(ZONE_STATE_MOUNTED
));
1662 if (zone_get_zonepath(zone_name
, zonepath
, sizeof (zonepath
)) != Z_OK
) {
1663 zerror(zlogp
, B_TRUE
, "unable to determine zone path");
1667 if (zone_get_rootpath(zone_name
, rootpath
, sizeof (rootpath
)) != Z_OK
) {
1668 zerror(zlogp
, B_TRUE
, "unable to determine zone root");
1672 if ((handle
= zonecfg_init_handle()) == NULL
) {
1673 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
1676 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
||
1677 zonecfg_setfsent(handle
) != Z_OK
) {
1678 zerror(zlogp
, B_FALSE
, "invalid configuration");
1683 * If we are mounting the zone, then we must always use the default
1684 * brand global mounts.
1686 if (ALT_MOUNT(mount_cmd
)) {
1687 (void) strlcpy(brand
, default_brand
, sizeof (brand
));
1689 (void) strlcpy(brand
, brand_name
, sizeof (brand
));
1692 /* Get a handle to the brand info for this zone */
1693 if ((bh
= brand_open(brand
)) == NULL
) {
1694 zerror(zlogp
, B_FALSE
, "unable to determine zone brand");
1695 zonecfg_fini_handle(handle
);
1700 * Get the list of global filesystems to mount from the brand
1703 cb
.pgcd_zlogp
= zlogp
;
1704 cb
.pgcd_fs_tab
= &fs_ptr
;
1705 cb
.pgcd_num_fs
= &num_fs
;
1706 if (brand_platform_iter_gmounts(bh
, zone_name
, zonepath
,
1707 plat_gmount_cb
, &cb
) != 0) {
1708 zerror(zlogp
, B_FALSE
, "unable to mount filesystems");
1710 zonecfg_fini_handle(handle
);
1716 * Iterate through the rest of the filesystems. Sort them all,
1717 * then mount them in sorted order. This is to make sure the
1718 * higher level directories (e.g., /usr) get mounted before
1719 * any beneath them (e.g., /usr/local).
1721 if (mount_filesystems_fsent(handle
, zlogp
, &fs_ptr
, &num_fs
,
1725 zonecfg_fini_handle(handle
);
1729 * Normally when we mount a zone all the zone filesystems
1730 * get mounted relative to rootpath, which is usually
1731 * <zonepath>/root. But when mounting a zone for administration
1732 * purposes via the zone "mount" state, build_mounted_pre_var()
1733 * updates rootpath to be <zonepath>/lu/a so we'll mount all
1734 * the zones filesystems there instead.
1736 * build_mounted_pre_var() and build_mounted_post_var() will
1737 * also do some extra work to create directories and lofs mount
1738 * a bunch of global zone file system paths into <zonepath>/lu.
1740 * This allows us to be able to enter the zone (now rooted at
1741 * <zonepath>/lu) and run the upgrade/patch tools that are in the
1742 * global zone and have them upgrade the to-be-modified zone's
1743 * files mounted on /a. (Which mirrors the existing standard
1744 * upgrade environment.)
1746 * There is of course one catch. When doing the upgrade
1747 * we need <zoneroot>/lu/dev to be the /dev filesystem
1748 * for the zone and we don't want to have any /dev filesystem
1749 * mounted at <zoneroot>/lu/a/dev. Since /dev is specified
1750 * as a normal zone filesystem by default we'll try to mount
1751 * it at <zoneroot>/lu/a/dev, so we have to detect this
1752 * case and instead mount it at <zoneroot>/lu/dev.
1754 * All this work is done in three phases:
1755 * 1) Create and populate lu directory (build_mounted_pre_var()).
1756 * 2) Mount the required filesystems as per the zone configuration.
1757 * 3) Set up the rest of the scratch zone environment
1758 * (build_mounted_post_var()).
1760 if (ALT_MOUNT(mount_cmd
) && !build_mounted_pre_var(zlogp
,
1761 rootpath
, sizeof (rootpath
), zonepath
, luroot
, sizeof (luroot
)))
1764 qsort(fs_ptr
, num_fs
, sizeof (*fs_ptr
), fs_compare
);
1766 for (i
= 0; i
< num_fs
; i
++) {
1767 if (ALT_MOUNT(mount_cmd
) &&
1768 strcmp(fs_ptr
[i
].zone_fs_dir
, "/dev") == 0) {
1769 size_t slen
= strlen(rootpath
) - 2;
1772 * By default we'll try to mount /dev as /a/dev
1773 * but /dev is special and always goes at the top
1774 * so strip the trailing '/a' from the rootpath.
1776 assert(strcmp(&rootpath
[slen
], "/a") == 0);
1777 rootpath
[slen
] = '\0';
1778 if (mount_one(zlogp
, &fs_ptr
[i
], rootpath
, mount_cmd
)
1781 rootpath
[slen
] = '/';
1784 if (mount_one(zlogp
, &fs_ptr
[i
], rootpath
, mount_cmd
) != 0)
1787 if (ALT_MOUNT(mount_cmd
) &&
1788 !build_mounted_post_var(zlogp
, mount_cmd
, rootpath
, luroot
))
1791 free_fs_data(fs_ptr
, num_fs
);
1794 * Everything looks fine.
1800 zonecfg_fini_handle(handle
);
1801 free_fs_data(fs_ptr
, num_fs
);
1805 /* caller makes sure neither parameter is NULL */
1807 addr2netmask(char *prefixstr
, int maxprefixlen
, uchar_t
*maskstr
)
1811 prefixlen
= atoi(prefixstr
);
1812 if (prefixlen
< 0 || prefixlen
> maxprefixlen
)
1814 while (prefixlen
> 0) {
1815 if (prefixlen
>= 8) {
1820 *maskstr
|= 1 << (8 - prefixlen
);
1827 * Tear down all interfaces belonging to the given zone. This should
1828 * be called with the zone in a state other than "running", so that
1829 * interfaces can't be assigned to the zone after this returns.
1831 * If anything goes wrong, log an error message and return an error.
1834 unconfigure_shared_network_interfaces(zlog_t
*zlogp
, zoneid_t zone_id
)
1837 struct lifconf lifc
;
1838 struct lifreq
*lifrp
, lifrl
;
1839 int64_t lifc_flags
= LIFC_NOXMIT
| LIFC_ALLZONES
;
1840 int num_ifs
, s
, i
, ret_code
= 0;
1844 if ((s
= socket(AF_INET
, SOCK_DGRAM
, 0)) < 0) {
1845 zerror(zlogp
, B_TRUE
, "could not get socket");
1849 lifn
.lifn_family
= AF_UNSPEC
;
1850 lifn
.lifn_flags
= (int)lifc_flags
;
1851 if (ioctl(s
, SIOCGLIFNUM
, (char *)&lifn
) < 0) {
1852 zerror(zlogp
, B_TRUE
,
1853 "could not determine number of network interfaces");
1857 num_ifs
= lifn
.lifn_count
;
1858 bufsize
= num_ifs
* sizeof (struct lifreq
);
1859 if ((buf
= malloc(bufsize
)) == NULL
) {
1860 zerror(zlogp
, B_TRUE
, "memory allocation failed");
1864 lifc
.lifc_family
= AF_UNSPEC
;
1865 lifc
.lifc_flags
= (int)lifc_flags
;
1866 lifc
.lifc_len
= bufsize
;
1867 lifc
.lifc_buf
= buf
;
1868 if (ioctl(s
, SIOCGLIFCONF
, (char *)&lifc
) < 0) {
1869 zerror(zlogp
, B_TRUE
, "could not get configured network "
1874 lifrp
= lifc
.lifc_req
;
1875 for (i
= lifc
.lifc_len
/ sizeof (struct lifreq
); i
> 0; i
--, lifrp
++) {
1877 if ((s
= socket(lifrp
->lifr_addr
.ss_family
, SOCK_DGRAM
, 0)) <
1879 zerror(zlogp
, B_TRUE
, "%s: could not get socket",
1884 (void) memset(&lifrl
, 0, sizeof (lifrl
));
1885 (void) strncpy(lifrl
.lifr_name
, lifrp
->lifr_name
,
1886 sizeof (lifrl
.lifr_name
));
1887 if (ioctl(s
, SIOCGLIFZONE
, (caddr_t
)&lifrl
) < 0) {
1890 * Interface may have been removed by admin or
1891 * another zone halting.
1894 zerror(zlogp
, B_TRUE
,
1895 "%s: could not determine the zone to which this "
1896 "network interface is bound", lifrl
.lifr_name
);
1900 if (lifrl
.lifr_zoneid
== zone_id
) {
1901 if (ioctl(s
, SIOCLIFREMOVEIF
, (caddr_t
)&lifrl
) < 0) {
1902 zerror(zlogp
, B_TRUE
,
1903 "%s: could not remove network interface",
1917 static union sockunion
{
1919 struct sockaddr_in sin
;
1920 struct sockaddr_dl sdl
;
1921 struct sockaddr_in6 sin6
;
1925 struct rt_msghdr hdr
;
1930 salen(struct sockaddr
*sa
)
1932 switch (sa
->sa_family
) {
1934 return (sizeof (struct sockaddr_in
));
1936 return (sizeof (struct sockaddr_dl
));
1938 return (sizeof (struct sockaddr_in6
));
1940 return (sizeof (struct sockaddr
));
1944 #define ROUNDUP_LONG(a) \
1945 ((a) > 0 ? (1 + (((a) - 1) | (sizeof (long) - 1))) : sizeof (long))
1948 * Look up which zone is using a given IP address. The address in question
1949 * is expected to have been stuffed into the structure to which lifr points
1950 * via a previous SIOCGLIFADDR ioctl().
1952 * This is done using black router socket magic.
1954 * Return the name of the zone on success or NULL on failure.
1956 * This is a lot of code for a simple task; a new ioctl request to take care
1957 * of this might be a useful RFE.
1961 who_is_using(zlog_t
*zlogp
, struct lifreq
*lifr
)
1963 static char answer
[ZONENAME_MAX
];
1966 char *cp
= rtmsg
.space
;
1967 struct sockaddr_dl
*ifp
= NULL
;
1968 struct sockaddr
*sa
;
1969 char save_if_name
[LIFNAMSIZ
];
1974 if ((s
= socket(PF_ROUTE
, SOCK_RAW
, 0)) < 0) {
1975 zerror(zlogp
, B_TRUE
, "could not get routing socket");
1979 if (lifr
->lifr_addr
.ss_family
== AF_INET
) {
1980 struct sockaddr_in
*sin4
;
1982 so_dst
.sa
.sa_family
= AF_INET
;
1983 sin4
= (struct sockaddr_in
*)&lifr
->lifr_addr
;
1984 so_dst
.sin
.sin_addr
= sin4
->sin_addr
;
1986 struct sockaddr_in6
*sin6
;
1988 so_dst
.sa
.sa_family
= AF_INET6
;
1989 sin6
= (struct sockaddr_in6
*)&lifr
->lifr_addr
;
1990 so_dst
.sin6
.sin6_addr
= sin6
->sin6_addr
;
1993 so_ifp
.sa
.sa_family
= AF_LINK
;
1995 (void) memset(&rtmsg
, 0, sizeof (rtmsg
));
1996 rtmsg
.hdr
.rtm_type
= RTM_GET
;
1997 rtmsg
.hdr
.rtm_flags
= RTF_UP
| RTF_HOST
;
1998 rtmsg
.hdr
.rtm_version
= RTM_VERSION
;
1999 rtmsg
.hdr
.rtm_seq
= ++rts_seqno
;
2000 rtmsg
.hdr
.rtm_addrs
= RTA_IFP
| RTA_DST
;
2002 l
= ROUNDUP_LONG(salen(&so_dst
.sa
));
2003 (void) memmove(cp
, &(so_dst
), l
);
2005 l
= ROUNDUP_LONG(salen(&so_ifp
.sa
));
2006 (void) memmove(cp
, &(so_ifp
), l
);
2009 rtmsg
.hdr
.rtm_msglen
= l
= cp
- (char *)&rtmsg
;
2011 if ((rlen
= write(s
, &rtmsg
, l
)) < 0) {
2012 zerror(zlogp
, B_TRUE
, "writing to routing socket");
2014 } else if (rlen
< (int)rtmsg
.hdr
.rtm_msglen
) {
2015 zerror(zlogp
, B_TRUE
,
2016 "write to routing socket got only %d for len\n", rlen
);
2020 l
= read(s
, &rtmsg
, sizeof (rtmsg
));
2021 } while (l
> 0 && (rtmsg
.hdr
.rtm_seq
!= rts_seqno
||
2022 rtmsg
.hdr
.rtm_pid
!= pid
));
2024 zerror(zlogp
, B_TRUE
, "reading from routing socket");
2028 if (rtmsg
.hdr
.rtm_version
!= RTM_VERSION
) {
2029 zerror(zlogp
, B_FALSE
,
2030 "routing message version %d not understood",
2031 rtmsg
.hdr
.rtm_version
);
2034 if (rtmsg
.hdr
.rtm_msglen
!= (ushort_t
)l
) {
2035 zerror(zlogp
, B_FALSE
, "message length mismatch, "
2036 "expected %d bytes, returned %d bytes",
2037 rtmsg
.hdr
.rtm_msglen
, l
);
2040 if (rtmsg
.hdr
.rtm_errno
!= 0) {
2041 errno
= rtmsg
.hdr
.rtm_errno
;
2042 zerror(zlogp
, B_TRUE
, "RTM_GET routing socket message");
2045 if ((rtmsg
.hdr
.rtm_addrs
& RTA_IFP
) == 0) {
2046 zerror(zlogp
, B_FALSE
, "network interface not found");
2049 cp
= ((char *)(&rtmsg
.hdr
+ 1));
2050 for (i
= 1; i
!= 0; i
<<= 1) {
2051 /* LINTED E_BAD_PTR_CAST_ALIGN */
2052 sa
= (struct sockaddr
*)cp
;
2054 if ((i
& rtmsg
.hdr
.rtm_addrs
) != 0)
2055 cp
+= ROUNDUP_LONG(salen(sa
));
2058 if (sa
->sa_family
== AF_LINK
&&
2059 ((struct sockaddr_dl
*)sa
)->sdl_nlen
!= 0)
2060 ifp
= (struct sockaddr_dl
*)sa
;
2064 zerror(zlogp
, B_FALSE
, "network interface could not be "
2070 * We need to set the I/F name to what we got above, then do the
2071 * appropriate ioctl to get its zone name. But lifr->lifr_name is
2072 * used by the calling function to do a REMOVEIF, so if we leave the
2073 * "good" zone's I/F name in place, *that* I/F will be removed instead
2074 * of the bad one. So we save the old (bad) I/F name before over-
2075 * writing it and doing the ioctl, then restore it after the ioctl.
2077 (void) strlcpy(save_if_name
, lifr
->lifr_name
, sizeof (save_if_name
));
2078 (void) strncpy(lifr
->lifr_name
, ifp
->sdl_data
, ifp
->sdl_nlen
);
2079 lifr
->lifr_name
[ifp
->sdl_nlen
] = '\0';
2080 i
= ioctl(s
, SIOCGLIFZONE
, lifr
);
2081 (void) strlcpy(lifr
->lifr_name
, save_if_name
, sizeof (save_if_name
));
2083 zerror(zlogp
, B_TRUE
,
2084 "%s: could not determine the zone network interface "
2085 "belongs to", lifr
->lifr_name
);
2088 if (getzonenamebyid(lifr
->lifr_zoneid
, answer
, sizeof (answer
)) < 0)
2089 (void) snprintf(answer
, sizeof (answer
), "%d",
2092 if (strlen(answer
) > 0)
2098 * Configures a single interface: a new virtual interface is added, based on
2099 * the physical interface nwiftabptr->zone_nwif_physical, with the address
2100 * specified in nwiftabptr->zone_nwif_address, for zone zone_id. Note that
2101 * the "address" can be an IPv6 address (with a /prefixlength required), an
2102 * IPv4 address (with a /prefixlength optional), or a name; for the latter,
2103 * an IPv4 name-to-address resolution will be attempted.
2105 * If anything goes wrong, we log an detailed error message, attempt to tear
2106 * down whatever we set up and return an error.
2109 configure_one_interface(zlog_t
*zlogp
, zoneid_t zone_id
,
2110 struct zone_nwiftab
*nwiftabptr
)
2113 struct sockaddr_in netmask4
;
2114 struct sockaddr_in6 netmask6
;
2115 struct sockaddr_storage laddr
;
2118 char *slashp
= strchr(nwiftabptr
->zone_nwif_address
, '/');
2120 boolean_t got_netmask
= B_FALSE
;
2121 boolean_t is_loopback
= B_FALSE
;
2122 char addrstr4
[INET_ADDRSTRLEN
];
2125 res
= zonecfg_valid_net_address(nwiftabptr
->zone_nwif_address
, &lifr
);
2127 zerror(zlogp
, B_FALSE
, "%s: %s", zonecfg_strerror(res
),
2128 nwiftabptr
->zone_nwif_address
);
2131 af
= lifr
.lifr_addr
.ss_family
;
2133 in4
= ((struct sockaddr_in
*)(&lifr
.lifr_addr
))->sin_addr
;
2134 if ((s
= socket(af
, SOCK_DGRAM
, 0)) < 0) {
2135 zerror(zlogp
, B_TRUE
, "could not get socket");
2140 * This is a similar kind of "hack" like in addif() to get around
2141 * the problem of SIOCLIFADDIF. The problem is that this ioctl
2142 * does not include the netmask when adding a logical interface.
2143 * To get around this problem, we first add the logical interface
2144 * with a 0 address. After that, we set the netmask if provided.
2145 * Finally we set the interface address.
2147 laddr
= lifr
.lifr_addr
;
2148 (void) strlcpy(lifr
.lifr_name
, nwiftabptr
->zone_nwif_physical
,
2149 sizeof (lifr
.lifr_name
));
2150 (void) memset(&lifr
.lifr_addr
, 0, sizeof (lifr
.lifr_addr
));
2152 if (ioctl(s
, SIOCLIFADDIF
, (caddr_t
)&lifr
) < 0) {
2154 * Here, we know that the interface can't be brought up.
2155 * A similar warning message was already printed out to
2156 * the console by zoneadm(1M) so instead we log the
2157 * message to syslog and continue.
2159 zerror(&logsys
, B_TRUE
, "WARNING: skipping network interface "
2160 "'%s' which may not be present/plumbed in the "
2161 "global zone.", lifr
.lifr_name
);
2166 /* Preserve literal IPv4 address for later potential printing. */
2168 (void) inet_ntop(AF_INET
, &in4
, addrstr4
, INET_ADDRSTRLEN
);
2170 lifr
.lifr_zoneid
= zone_id
;
2171 if (ioctl(s
, SIOCSLIFZONE
, (caddr_t
)&lifr
) < 0) {
2172 zerror(zlogp
, B_TRUE
, "%s: could not place network interface "
2173 "into zone", lifr
.lifr_name
);
2178 * Loopback interface will use the default netmask assigned, if no
2181 if (strcmp(nwiftabptr
->zone_nwif_physical
, "lo0") == 0) {
2182 is_loopback
= B_TRUE
;
2184 if (af
== AF_INET
) {
2186 * The IPv4 netmask can be determined either
2187 * directly if a prefix length was supplied with
2188 * the address or via the netmasks database. Not
2189 * being able to determine it is a common failure,
2190 * but it often is not fatal to operation of the
2191 * interface. In that case, a warning will be
2192 * printed after the rest of the interface's
2193 * parameters have been configured.
2195 (void) memset(&netmask4
, 0, sizeof (netmask4
));
2196 if (slashp
!= NULL
) {
2197 if (addr2netmask(slashp
+ 1, V4_ADDR_LEN
,
2198 (uchar_t
*)&netmask4
.sin_addr
) != 0) {
2200 zerror(zlogp
, B_FALSE
,
2201 "%s: invalid prefix length in %s",
2203 nwiftabptr
->zone_nwif_address
);
2206 got_netmask
= B_TRUE
;
2207 } else if (getnetmaskbyaddr(in4
,
2208 &netmask4
.sin_addr
) == 0) {
2209 got_netmask
= B_TRUE
;
2212 netmask4
.sin_family
= af
;
2213 (void) memcpy(&lifr
.lifr_addr
, &netmask4
,
2217 (void) memset(&netmask6
, 0, sizeof (netmask6
));
2218 if (addr2netmask(slashp
+ 1, V6_ADDR_LEN
,
2219 (uchar_t
*)&netmask6
.sin6_addr
) != 0) {
2221 zerror(zlogp
, B_FALSE
,
2222 "%s: invalid prefix length in %s",
2224 nwiftabptr
->zone_nwif_address
);
2227 got_netmask
= B_TRUE
;
2228 netmask6
.sin6_family
= af
;
2229 (void) memcpy(&lifr
.lifr_addr
, &netmask6
,
2233 ioctl(s
, SIOCSLIFNETMASK
, (caddr_t
)&lifr
) < 0) {
2234 zerror(zlogp
, B_TRUE
, "%s: could not set netmask",
2239 /* Set the interface address */
2240 lifr
.lifr_addr
= laddr
;
2241 if (ioctl(s
, SIOCSLIFADDR
, (caddr_t
)&lifr
) < 0) {
2242 zerror(zlogp
, B_TRUE
,
2243 "%s: could not set IP address to %s",
2244 lifr
.lifr_name
, nwiftabptr
->zone_nwif_address
);
2248 if (ioctl(s
, SIOCGLIFFLAGS
, (caddr_t
)&lifr
) < 0) {
2249 zerror(zlogp
, B_TRUE
, "%s: could not get flags",
2253 lifr
.lifr_flags
|= IFF_UP
;
2254 if (ioctl(s
, SIOCSLIFFLAGS
, (caddr_t
)&lifr
) < 0) {
2255 int save_errno
= errno
;
2259 * If we failed with something other than EADDRNOTAVAIL,
2260 * then skip to the end. Otherwise, look up our address,
2261 * then call a function to determine which zone is already
2262 * using that address.
2264 if (errno
!= EADDRNOTAVAIL
) {
2265 zerror(zlogp
, B_TRUE
,
2266 "%s: could not bring network interface up",
2270 if (ioctl(s
, SIOCGLIFADDR
, (caddr_t
)&lifr
) < 0) {
2271 zerror(zlogp
, B_TRUE
, "%s: could not get address",
2275 zone_using
= who_is_using(zlogp
, &lifr
);
2277 if (zone_using
== NULL
)
2278 zerror(zlogp
, B_TRUE
,
2279 "%s: could not bring network interface up",
2282 zerror(zlogp
, B_TRUE
, "%s: could not bring network "
2283 "interface up: address in use by zone '%s'",
2284 lifr
.lifr_name
, zone_using
);
2288 if (!got_netmask
&& !is_loopback
) {
2290 * A common, but often non-fatal problem, is that the system
2291 * cannot find the netmask for an interface address. This is
2292 * often caused by it being only in /etc/inet/netmasks, but
2293 * /etc/nsswitch.conf says to use NIS or NIS+ and it's not
2294 * in that. This doesn't show up at boot because the netmask
2295 * is obtained from /etc/inet/netmasks when no network
2296 * interfaces are up, but isn't consulted when NIS/NIS+ is
2297 * available. We warn the user here that something like this
2298 * has happened and we're just running with a default and
2299 * possible incorrect netmask.
2301 char buffer
[INET6_ADDRSTRLEN
];
2303 const char *nomatch
= "no matching subnet found in netmasks(4)";
2306 addr
= &((struct sockaddr_in
*)
2307 (&lifr
.lifr_addr
))->sin_addr
;
2309 addr
= &((struct sockaddr_in6
*)
2310 (&lifr
.lifr_addr
))->sin6_addr
;
2313 * Find out what netmask the interface is going to be using.
2314 * If we just brought up an IPMP data address on an underlying
2315 * interface above, the address will have already migrated, so
2316 * the SIOCGLIFNETMASK won't be able to find it (but we need
2317 * to bring the address up to get the actual netmask). Just
2318 * omit printing the actual netmask in this corner-case.
2320 if (ioctl(s
, SIOCGLIFNETMASK
, (caddr_t
)&lifr
) < 0 ||
2321 inet_ntop(af
, addr
, buffer
, sizeof (buffer
)) == NULL
) {
2322 zerror(zlogp
, B_FALSE
, "WARNING: %s; using default.",
2325 zerror(zlogp
, B_FALSE
,
2326 "WARNING: %s: %s: %s; using default of %s.",
2327 lifr
.lifr_name
, nomatch
, addrstr4
, buffer
);
2332 * If a default router was specified for this interface
2333 * set the route now. Ignore if already set.
2335 if (strlen(nwiftabptr
->zone_nwif_defrouter
) > 0) {
2342 argv
[3] = nwiftabptr
->zone_nwif_physical
;
2343 argv
[4] = "default";
2344 argv
[5] = nwiftabptr
->zone_nwif_defrouter
;
2347 status
= forkexec(zlogp
, "/usr/sbin/route", argv
);
2348 if (status
!= 0 && status
!= EEXIST
)
2349 zerror(zlogp
, B_FALSE
, "Unable to set route for "
2350 "interface %s to %s\n",
2351 nwiftabptr
->zone_nwif_physical
,
2352 nwiftabptr
->zone_nwif_defrouter
);
2358 (void) ioctl(s
, SIOCLIFREMOVEIF
, (caddr_t
)&lifr
);
2364 * Sets up network interfaces based on information from the zone configuration.
2365 * IPv4 and IPv6 loopback interfaces are set up "for free", modeling the global
2368 * If anything goes wrong, we log a general error message, attempt to tear down
2369 * whatever we set up, and return an error.
2372 configure_shared_network_interfaces(zlog_t
*zlogp
)
2374 zone_dochandle_t handle
;
2375 struct zone_nwiftab nwiftab
, loopback_iftab
;
2378 if ((zoneid
= getzoneidbyname(zone_name
)) == ZONE_ID_UNDEFINED
) {
2379 zerror(zlogp
, B_TRUE
, "unable to get zoneid");
2383 if ((handle
= zonecfg_init_handle()) == NULL
) {
2384 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
2387 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
) {
2388 zerror(zlogp
, B_FALSE
, "invalid configuration");
2389 zonecfg_fini_handle(handle
);
2392 if (zonecfg_setnwifent(handle
) == Z_OK
) {
2394 if (zonecfg_getnwifent(handle
, &nwiftab
) != Z_OK
)
2396 if (configure_one_interface(zlogp
, zoneid
, &nwiftab
) !=
2398 (void) zonecfg_endnwifent(handle
);
2399 zonecfg_fini_handle(handle
);
2403 (void) zonecfg_endnwifent(handle
);
2405 zonecfg_fini_handle(handle
);
2406 (void) strlcpy(loopback_iftab
.zone_nwif_physical
, "lo0",
2407 sizeof (loopback_iftab
.zone_nwif_physical
));
2408 (void) strlcpy(loopback_iftab
.zone_nwif_address
, "127.0.0.1",
2409 sizeof (loopback_iftab
.zone_nwif_address
));
2410 loopback_iftab
.zone_nwif_defrouter
[0] = '\0';
2411 if (configure_one_interface(zlogp
, zoneid
, &loopback_iftab
) != Z_OK
)
2414 /* Always plumb up the IPv6 loopback interface. */
2415 (void) strlcpy(loopback_iftab
.zone_nwif_address
, "::1/128",
2416 sizeof (loopback_iftab
.zone_nwif_address
));
2417 if (configure_one_interface(zlogp
, zoneid
, &loopback_iftab
) != Z_OK
)
2423 zdlerror(zlog_t
*zlogp
, dladm_status_t err
, const char *dlname
, const char *str
)
2425 char errmsg
[DLADM_STRSIZE
];
2427 (void) dladm_status2str(err
, errmsg
);
2428 zerror(zlogp
, B_FALSE
, "%s '%s': %s", str
, dlname
, errmsg
);
2432 add_datalink(zlog_t
*zlogp
, char *zone_name
, datalink_id_t linkid
, char *dlname
)
2435 boolean_t cpuset
, poolset
;
2438 /* First check if it's in use by global zone. */
2439 if (zonecfg_ifname_exists(AF_INET
, dlname
) ||
2440 zonecfg_ifname_exists(AF_INET6
, dlname
)) {
2441 zerror(zlogp
, B_FALSE
, "WARNING: skipping network interface "
2442 "'%s' which is used in the global zone", dlname
);
2446 /* Set zoneid of this link. */
2447 err
= dladm_set_linkprop(dld_handle
, linkid
, "zone", &zone_name
, 1,
2449 if (err
!= DLADM_STATUS_OK
) {
2450 zdlerror(zlogp
, err
, dlname
,
2451 "WARNING: unable to add network interface");
2456 * Set the pool of this link if the zone has a pool and
2457 * neither the cpus nor the pool datalink property is
2460 err
= dladm_linkprop_is_set(dld_handle
, linkid
, DLADM_PROP_VAL_CURRENT
,
2462 if (err
!= DLADM_STATUS_OK
) {
2463 zdlerror(zlogp
, err
, dlname
,
2464 "WARNING: unable to check if cpus link property is set");
2466 err
= dladm_linkprop_is_set(dld_handle
, linkid
, DLADM_PROP_VAL_CURRENT
,
2468 if (err
!= DLADM_STATUS_OK
) {
2469 zdlerror(zlogp
, err
, dlname
,
2470 "WARNING: unable to check if pool link property is set");
2473 if ((strlen(pool_name
) != 0) && !cpuset
&& !poolset
) {
2475 err
= dladm_set_linkprop(dld_handle
, linkid
, "pool",
2476 &poolp
, 1, DLADM_OPT_ACTIVE
);
2477 if (err
!= DLADM_STATUS_OK
) {
2478 zerror(zlogp
, B_FALSE
, "WARNING: unable to set "
2479 "pool %s to datalink %s", pool_name
, dlname
);
2480 bzero(pool_name
, sizeof (pool_name
));
2483 bzero(pool_name
, sizeof (pool_name
));
2489 sockaddr_to_str(sa_family_t af
, const struct sockaddr
*sockaddr
,
2490 char *straddr
, size_t len
)
2492 struct sockaddr_in
*sin
;
2493 struct sockaddr_in6
*sin6
;
2494 const char *str
= NULL
;
2496 if (af
== AF_INET
) {
2497 /* LINTED E_BAD_PTR_CAST_ALIGN */
2498 sin
= SIN(sockaddr
);
2499 str
= inet_ntop(AF_INET
, (void *)&sin
->sin_addr
, straddr
, len
);
2500 } else if (af
== AF_INET6
) {
2501 /* LINTED E_BAD_PTR_CAST_ALIGN */
2502 sin6
= SIN6(sockaddr
);
2503 str
= inet_ntop(AF_INET6
, (void *)&sin6
->sin6_addr
, straddr
,
2507 return (str
!= NULL
);
2511 ipv4_prefixlen(struct sockaddr_in
*sin
)
2513 struct sockaddr_in
*m
;
2514 struct sockaddr_storage mask
;
2517 m
->sin_family
= AF_INET
;
2518 if (getnetmaskbyaddr(sin
->sin_addr
, &m
->sin_addr
) == 0) {
2519 return (mask2plen((struct sockaddr
*)&mask
));
2520 } else if (IN_CLASSA(htonl(sin
->sin_addr
.s_addr
))) {
2522 } else if (IN_CLASSB(ntohl(sin
->sin_addr
.s_addr
))) {
2524 } else if (IN_CLASSC(ntohl(sin
->sin_addr
.s_addr
))) {
2531 zone_setattr_network(int type
, zoneid_t zoneid
, datalink_id_t linkid
,
2532 void *buf
, size_t bufsize
)
2534 zone_net_data_t
*zndata
;
2538 znsize
= sizeof (*zndata
) + bufsize
;
2539 zndata
= calloc(1, znsize
);
2542 zndata
->zn_type
= type
;
2543 zndata
->zn_len
= bufsize
;
2544 zndata
->zn_linkid
= linkid
;
2545 bcopy(buf
, zndata
->zn_val
, zndata
->zn_len
);
2546 err
= zone_setattr(zoneid
, ZONE_ATTR_NETWORK
, zndata
, znsize
);
2552 add_net_for_linkid(zlog_t
*zlogp
, zoneid_t zoneid
, zone_addr_list_t
*start
)
2555 char **astr
, *address
;
2556 dladm_status_t dlstatus
;
2557 char *ip_nospoof
= "ip-nospoof";
2558 int nnet
, naddr
, err
= 0, j
;
2559 size_t zlen
, cpleft
;
2560 zone_addr_list_t
*ptr
, *end
;
2561 char tmp
[INET6_ADDRSTRLEN
], *maskstr
;
2563 struct in6_addr
*routes
= NULL
;
2565 datalink_id_t linkid
;
2567 assert(start
!= NULL
);
2568 naddr
= 0; /* number of addresses */
2569 nnet
= 0; /* number of net resources */
2570 linkid
= start
->za_linkid
;
2571 for (ptr
= start
; ptr
!= NULL
&& ptr
->za_linkid
== linkid
;
2572 ptr
= ptr
->za_next
) {
2576 zlen
= nnet
* (INET6_ADDRSTRLEN
+ 1);
2577 astr
= calloc(1, nnet
* sizeof (uintptr_t));
2578 zaddr
= calloc(1, zlen
);
2579 if (astr
== NULL
|| zaddr
== NULL
) {
2586 for (ptr
= start
; ptr
!= end
; ptr
= ptr
->za_next
) {
2587 address
= ptr
->za_nwiftab
.zone_nwif_allowed_address
;
2588 if (address
[0] == '\0')
2590 (void) snprintf(tmp
, sizeof (tmp
), "%s", address
);
2592 * Validate the data. zonecfg_valid_net_address() clobbers
2593 * the /<mask> in the address string.
2595 if (zonecfg_valid_net_address(address
, &lifr
) != Z_OK
) {
2596 zerror(zlogp
, B_FALSE
, "invalid address [%s]\n",
2602 * convert any hostnames to numeric address strings.
2604 if (!sockaddr_to_str(lifr
.lifr_addr
.ss_family
,
2605 (const struct sockaddr
*)&lifr
.lifr_addr
, cp
, cpleft
)) {
2610 * make a copy of the numeric string for the data needed
2611 * by the "allowed-ips" datalink property.
2613 astr
[j
] = strdup(cp
);
2614 if (astr
[j
] == NULL
) {
2620 * compute the default netmask from the address, if necessary
2622 if ((maskstr
= strchr(tmp
, '/')) == NULL
) {
2625 if (lifr
.lifr_addr
.ss_family
== AF_INET
) {
2626 prefixlen
= ipv4_prefixlen(
2627 SIN(&lifr
.lifr_addr
));
2629 struct sockaddr_in6
*sin6
;
2631 sin6
= SIN6(&lifr
.lifr_addr
);
2632 if (IN6_IS_ADDR_LINKLOCAL(&sin6
->sin6_addr
))
2637 (void) snprintf(tmp
, sizeof (tmp
), "%d", prefixlen
);
2642 /* append the "/<netmask>" */
2643 (void) strlcat(cp
, "/", cpleft
);
2644 (void) strlcat(cp
, maskstr
, cpleft
);
2645 (void) strlcat(cp
, ",", cpleft
);
2646 cp
+= strnlen(cp
, zlen
);
2647 cpleft
= &zaddr
[INET6_ADDRSTRLEN
] - cp
;
2649 naddr
= j
; /* the actual number of addresses in the net resource */
2650 assert(naddr
<= nnet
);
2653 * zonecfg has already verified that the defrouter property can only
2654 * be set if there is at least one address defined for the net resource.
2655 * If j is 0, there are no addresses defined, and therefore no routers
2656 * to configure, and we are done at that point.
2661 /* over-write last ',' with '\0' */
2662 zaddr
[strnlen(zaddr
, zlen
) + 1] = '\0';
2665 * First make sure L3 protection is not already set on the link.
2667 dlstatus
= dladm_linkprop_is_set(dld_handle
, linkid
, DLADM_OPT_ACTIVE
,
2668 "protection", &is_set
);
2669 if (dlstatus
!= DLADM_STATUS_OK
) {
2671 zerror(zlogp
, B_FALSE
, "unable to check if protection is set");
2676 zerror(zlogp
, B_FALSE
, "Protection is already set");
2679 dlstatus
= dladm_linkprop_is_set(dld_handle
, linkid
, DLADM_OPT_ACTIVE
,
2680 "allowed-ips", &is_set
);
2681 if (dlstatus
!= DLADM_STATUS_OK
) {
2683 zerror(zlogp
, B_FALSE
, "unable to check if allowed-ips is set");
2687 zerror(zlogp
, B_FALSE
, "allowed-ips is already set");
2693 * Enable ip-nospoof for the link, and add address to the allowed-ips
2696 dlstatus
= dladm_set_linkprop(dld_handle
, linkid
, "protection",
2697 &ip_nospoof
, 1, DLADM_OPT_ACTIVE
);
2698 if (dlstatus
!= DLADM_STATUS_OK
) {
2699 zerror(zlogp
, B_FALSE
, "could not set protection\n");
2703 dlstatus
= dladm_set_linkprop(dld_handle
, linkid
, "allowed-ips",
2704 astr
, naddr
, DLADM_OPT_ACTIVE
);
2705 if (dlstatus
!= DLADM_STATUS_OK
) {
2706 zerror(zlogp
, B_FALSE
, "could not set allowed-ips\n");
2711 /* now set the address in the data-store */
2712 err
= zone_setattr_network(ZONE_NETWORK_ADDRESS
, zoneid
, linkid
,
2713 zaddr
, strnlen(zaddr
, zlen
) + 1);
2718 * add the defaultrouters
2720 routes
= calloc(1, nnet
* sizeof (*routes
));
2722 for (ptr
= start
; ptr
!= end
; ptr
= ptr
->za_next
) {
2723 address
= ptr
->za_nwiftab
.zone_nwif_defrouter
;
2724 if (address
[0] == '\0')
2726 if (strchr(address
, '/') == NULL
&& strchr(address
, ':') != 0) {
2728 * zonecfg_valid_net_address() expects numeric IPv6
2729 * addresses to have a CIDR format netmask.
2731 (void) snprintf(tmp
, sizeof (tmp
), "/%d", V6_ADDR_LEN
);
2732 (void) strlcat(address
, tmp
, INET6_ADDRSTRLEN
);
2734 if (zonecfg_valid_net_address(address
, &lifr
) != Z_OK
) {
2735 zerror(zlogp
, B_FALSE
,
2736 "invalid router [%s]\n", address
);
2740 if (lifr
.lifr_addr
.ss_family
== AF_INET6
) {
2741 routes
[j
] = SIN6(&lifr
.lifr_addr
)->sin6_addr
;
2743 IN6_INADDR_TO_V4MAPPED(&SIN(&lifr
.lifr_addr
)->sin_addr
,
2750 err
= zone_setattr_network(ZONE_NETWORK_DEFROUTER
, zoneid
,
2751 linkid
, routes
, j
* sizeof (*routes
));
2755 for (j
= 0; j
< naddr
; j
++)
2764 add_net(zlog_t
*zlogp
, zoneid_t zoneid
, zone_addr_list_t
*zalist
)
2766 zone_addr_list_t
*ptr
;
2767 datalink_id_t linkid
;
2773 linkid
= zalist
->za_linkid
;
2775 err
= add_net_for_linkid(zlogp
, zoneid
, zalist
);
2779 for (ptr
= zalist
; ptr
!= NULL
; ptr
= ptr
->za_next
) {
2780 if (ptr
->za_linkid
== linkid
)
2782 linkid
= ptr
->za_linkid
;
2783 err
= add_net_for_linkid(zlogp
, zoneid
, ptr
);
2791 * Add "new" to the list of network interfaces to be configured by
2792 * add_net on zone boot in "old". The list of interfaces in "old" is
2793 * sorted by datalink_id_t, with interfaces sorted FIFO for a given
2796 * Returns the merged list of IP interfaces containing "old" and "new"
2798 static zone_addr_list_t
*
2799 add_ip_interface(zone_addr_list_t
*old
, zone_addr_list_t
*new)
2801 zone_addr_list_t
*ptr
, *next
;
2802 datalink_id_t linkid
= new->za_linkid
;
2808 for (ptr
= old
; ptr
!= NULL
; ptr
= ptr
->za_next
) {
2809 if (ptr
->za_linkid
== linkid
)
2813 /* linkid does not already exist, add to the beginning */
2818 * adding to the middle of the list; ptr points at the first
2819 * occurrence of linkid. Find the last occurrence.
2821 while ((next
= ptr
->za_next
) != NULL
) {
2822 if (next
->za_linkid
!= linkid
)
2826 /* insert new after ptr */
2827 new->za_next
= next
;
2833 free_ip_interface(zone_addr_list_t
*zalist
)
2835 zone_addr_list_t
*ptr
, *new;
2837 for (ptr
= zalist
; ptr
!= NULL
; ) {
2845 * Add the kernel access control information for the interface names.
2846 * If anything goes wrong, we log a general error message, attempt to tear down
2847 * whatever we set up, and return an error.
2850 configure_exclusive_network_interfaces(zlog_t
*zlogp
, zoneid_t zoneid
)
2852 zone_dochandle_t handle
;
2853 struct zone_nwiftab nwiftab
;
2854 char rootpath
[MAXPATHLEN
];
2855 char path
[MAXPATHLEN
];
2856 datalink_id_t linkid
;
2857 di_prof_t prof
= NULL
;
2858 boolean_t added
= B_FALSE
;
2859 zone_addr_list_t
*zalist
= NULL
, *new;
2861 if ((handle
= zonecfg_init_handle()) == NULL
) {
2862 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
2865 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
) {
2866 zerror(zlogp
, B_FALSE
, "invalid configuration");
2867 zonecfg_fini_handle(handle
);
2871 if (zonecfg_setnwifent(handle
) != Z_OK
) {
2872 zonecfg_fini_handle(handle
);
2877 if (zonecfg_getnwifent(handle
, &nwiftab
) != Z_OK
)
2881 if (zone_get_devroot(zone_name
, rootpath
,
2882 sizeof (rootpath
)) != Z_OK
) {
2883 (void) zonecfg_endnwifent(handle
);
2884 zonecfg_fini_handle(handle
);
2885 zerror(zlogp
, B_TRUE
,
2886 "unable to determine dev root");
2889 (void) snprintf(path
, sizeof (path
), "%s%s", rootpath
,
2891 if (di_prof_init(path
, &prof
) != 0) {
2892 (void) zonecfg_endnwifent(handle
);
2893 zonecfg_fini_handle(handle
);
2894 zerror(zlogp
, B_TRUE
,
2895 "failed to initialize profile");
2901 * Create the /dev entry for backward compatibility.
2902 * Only create the /dev entry if it's not in use.
2903 * Note that the zone still boots when the assigned
2904 * interface is inaccessible, used by others, etc.
2905 * Also, when vanity naming is used, some interface do
2906 * do not have corresponding /dev node names (for example,
2907 * vanity named aggregations). The /dev entry is not
2908 * created in that case. The /dev/net entry is always
2911 if (dladm_name2info(dld_handle
, nwiftab
.zone_nwif_physical
,
2912 &linkid
, NULL
, NULL
, NULL
) == DLADM_STATUS_OK
&&
2913 add_datalink(zlogp
, zone_name
, linkid
,
2914 nwiftab
.zone_nwif_physical
) == 0) {
2917 (void) zonecfg_endnwifent(handle
);
2918 zonecfg_fini_handle(handle
);
2919 zerror(zlogp
, B_TRUE
, "failed to add network device");
2922 /* set up the new IP interface, and add them all later */
2923 new = malloc(sizeof (*new));
2925 zerror(zlogp
, B_TRUE
, "no memory for %s",
2926 nwiftab
.zone_nwif_physical
);
2927 zonecfg_fini_handle(handle
);
2928 free_ip_interface(zalist
);
2930 bzero(new, sizeof (*new));
2931 new->za_nwiftab
= nwiftab
;
2932 new->za_linkid
= linkid
;
2933 zalist
= add_ip_interface(zalist
, new);
2935 if (zalist
!= NULL
) {
2936 if ((errno
= add_net(zlogp
, zoneid
, zalist
)) != 0) {
2937 (void) zonecfg_endnwifent(handle
);
2938 zonecfg_fini_handle(handle
);
2939 zerror(zlogp
, B_TRUE
, "failed to add address");
2940 free_ip_interface(zalist
);
2943 free_ip_interface(zalist
);
2945 (void) zonecfg_endnwifent(handle
);
2946 zonecfg_fini_handle(handle
);
2948 if (prof
!= NULL
&& added
) {
2949 if (di_prof_commit(prof
) != 0) {
2950 zerror(zlogp
, B_TRUE
, "failed to commit profile");
2961 remove_datalink_pool(zlog_t
*zlogp
, zoneid_t zoneid
)
2964 zone_iptype_t iptype
;
2966 datalink_id_t
*dllink
, *dllinks
= NULL
;
2969 if (strlen(pool_name
) == 0)
2972 if (zone_getattr(zoneid
, ZONE_ATTR_FLAGS
, &flags
,
2973 sizeof (flags
)) < 0) {
2974 if (vplat_get_iptype(zlogp
, &iptype
) < 0) {
2975 zerror(zlogp
, B_FALSE
, "unable to determine ip-type");
2979 if (flags
& ZF_NET_EXCL
)
2980 iptype
= ZS_EXCLUSIVE
;
2985 if (iptype
== ZS_EXCLUSIVE
) {
2987 * Get the datalink count and for each datalink,
2988 * attempt to clear the pool property and clear
2991 if (zone_list_datalink(zoneid
, &dlnum
, NULL
) != 0) {
2992 zerror(zlogp
, B_TRUE
, "unable to count network "
3000 if ((dllinks
= malloc(dlnum
* sizeof (datalink_id_t
)))
3002 zerror(zlogp
, B_TRUE
, "memory allocation failed");
3005 if (zone_list_datalink(zoneid
, &dlnum
, dllinks
) != 0) {
3006 zerror(zlogp
, B_TRUE
, "unable to list network "
3011 bzero(pool_name
, sizeof (pool_name
));
3012 for (i
= 0, dllink
= dllinks
; i
< dlnum
; i
++, dllink
++) {
3013 err
= dladm_set_linkprop(dld_handle
, *dllink
, "pool",
3014 NULL
, 0, DLADM_OPT_ACTIVE
);
3015 if (err
!= DLADM_STATUS_OK
) {
3016 zerror(zlogp
, B_TRUE
,
3017 "WARNING: unable to clear pool");
3026 remove_datalink_protect(zlog_t
*zlogp
, zoneid_t zoneid
)
3029 zone_iptype_t iptype
;
3031 dladm_status_t dlstatus
;
3032 datalink_id_t
*dllink
, *dllinks
= NULL
;
3034 if (zone_getattr(zoneid
, ZONE_ATTR_FLAGS
, &flags
,
3035 sizeof (flags
)) < 0) {
3036 if (vplat_get_iptype(zlogp
, &iptype
) < 0) {
3037 zerror(zlogp
, B_FALSE
, "unable to determine ip-type");
3041 if (flags
& ZF_NET_EXCL
)
3042 iptype
= ZS_EXCLUSIVE
;
3047 if (iptype
!= ZS_EXCLUSIVE
)
3051 * Get the datalink count and for each datalink,
3052 * attempt to clear the pool property and clear
3055 if (zone_list_datalink(zoneid
, &dlnum
, NULL
) != 0) {
3056 zerror(zlogp
, B_TRUE
, "unable to count network interfaces");
3063 if ((dllinks
= malloc(dlnum
* sizeof (datalink_id_t
))) == NULL
) {
3064 zerror(zlogp
, B_TRUE
, "memory allocation failed");
3067 if (zone_list_datalink(zoneid
, &dlnum
, dllinks
) != 0) {
3068 zerror(zlogp
, B_TRUE
, "unable to list network interfaces");
3073 for (i
= 0, dllink
= dllinks
; i
< dlnum
; i
++, dllink
++) {
3074 char dlerr
[DLADM_STRSIZE
];
3076 dlstatus
= dladm_set_linkprop(dld_handle
, *dllink
,
3077 "protection", NULL
, 0, DLADM_OPT_ACTIVE
);
3078 if (dlstatus
== DLADM_STATUS_NOTFOUND
) {
3079 /* datalink does not belong to the GZ */
3082 if (dlstatus
!= DLADM_STATUS_OK
) {
3083 zerror(zlogp
, B_FALSE
,
3084 dladm_status2str(dlstatus
, dlerr
));
3088 dlstatus
= dladm_set_linkprop(dld_handle
, *dllink
,
3089 "allowed-ips", NULL
, 0, DLADM_OPT_ACTIVE
);
3090 if (dlstatus
!= DLADM_STATUS_OK
) {
3091 zerror(zlogp
, B_FALSE
,
3092 dladm_status2str(dlstatus
, dlerr
));
3102 unconfigure_exclusive_network_interfaces(zlog_t
*zlogp
, zoneid_t zoneid
)
3107 * The kernel shutdown callback for the dls module should have removed
3108 * all datalinks from this zone. If any remain, then there's a
3111 if (zone_list_datalink(zoneid
, &dlnum
, NULL
) != 0) {
3112 zerror(zlogp
, B_TRUE
, "unable to list network interfaces");
3116 zerror(zlogp
, B_FALSE
,
3117 "datalinks remain in zone after shutdown");
3124 tcp_abort_conn(zlog_t
*zlogp
, zoneid_t zoneid
,
3125 const struct sockaddr_storage
*local
, const struct sockaddr_storage
*remote
)
3128 struct strioctl ioc
;
3129 tcp_ioc_abort_conn_t conn
;
3132 conn
.ac_local
= *local
;
3133 conn
.ac_remote
= *remote
;
3134 conn
.ac_start
= TCPS_SYN_SENT
;
3135 conn
.ac_end
= TCPS_TIME_WAIT
;
3136 conn
.ac_zoneid
= zoneid
;
3138 ioc
.ic_cmd
= TCP_IOC_ABORT_CONN
;
3139 ioc
.ic_timout
= -1; /* infinite timeout */
3140 ioc
.ic_len
= sizeof (conn
);
3141 ioc
.ic_dp
= (char *)&conn
;
3143 if ((fd
= open("/dev/tcp", O_RDONLY
)) < 0) {
3144 zerror(zlogp
, B_TRUE
, "unable to open %s", "/dev/tcp");
3148 error
= ioctl(fd
, I_STR
, &ioc
);
3150 if (error
== 0 || errno
== ENOENT
) /* ENOENT is not an error */
3156 tcp_abort_connections(zlog_t
*zlogp
, zoneid_t zoneid
)
3158 struct sockaddr_storage l
, r
;
3159 struct sockaddr_in
*local
, *remote
;
3160 struct sockaddr_in6
*local6
, *remote6
;
3164 * Abort IPv4 connections.
3166 bzero(&l
, sizeof (*local
));
3167 local
= (struct sockaddr_in
*)&l
;
3168 local
->sin_family
= AF_INET
;
3169 local
->sin_addr
.s_addr
= INADDR_ANY
;
3170 local
->sin_port
= 0;
3172 bzero(&r
, sizeof (*remote
));
3173 remote
= (struct sockaddr_in
*)&r
;
3174 remote
->sin_family
= AF_INET
;
3175 remote
->sin_addr
.s_addr
= INADDR_ANY
;
3176 remote
->sin_port
= 0;
3178 if ((error
= tcp_abort_conn(zlogp
, zoneid
, &l
, &r
)) != 0)
3182 * Abort IPv6 connections.
3184 bzero(&l
, sizeof (*local6
));
3185 local6
= (struct sockaddr_in6
*)&l
;
3186 local6
->sin6_family
= AF_INET6
;
3187 local6
->sin6_port
= 0;
3188 local6
->sin6_addr
= in6addr_any
;
3190 bzero(&r
, sizeof (*remote6
));
3191 remote6
= (struct sockaddr_in6
*)&r
;
3192 remote6
->sin6_family
= AF_INET6
;
3193 remote6
->sin6_port
= 0;
3194 remote6
->sin6_addr
= in6addr_any
;
3196 if ((error
= tcp_abort_conn(zlogp
, zoneid
, &l
, &r
)) != 0)
3202 get_privset(zlog_t
*zlogp
, priv_set_t
*privs
, zone_mnt_t mount_cmd
)
3205 zone_dochandle_t handle
;
3206 char *privname
= NULL
;
3208 if ((handle
= zonecfg_init_handle()) == NULL
) {
3209 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
3212 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
) {
3213 zerror(zlogp
, B_FALSE
, "invalid configuration");
3214 zonecfg_fini_handle(handle
);
3218 if (ALT_MOUNT(mount_cmd
)) {
3219 zone_iptype_t iptype
;
3220 const char *curr_iptype
;
3222 if (zonecfg_get_iptype(handle
, &iptype
) != Z_OK
) {
3223 zerror(zlogp
, B_TRUE
, "unable to determine ip-type");
3224 zonecfg_fini_handle(handle
);
3230 curr_iptype
= "shared";
3233 curr_iptype
= "exclusive";
3237 if (zonecfg_default_privset(privs
, curr_iptype
) == Z_OK
) {
3238 zonecfg_fini_handle(handle
);
3241 zerror(zlogp
, B_FALSE
,
3242 "failed to determine the zone's default privilege set");
3243 zonecfg_fini_handle(handle
);
3247 switch (zonecfg_get_privset(handle
, privs
, &privname
)) {
3251 case Z_PRIV_PROHIBITED
:
3252 zerror(zlogp
, B_FALSE
, "privilege \"%s\" is not permitted "
3253 "within the zone's privilege set", privname
);
3255 case Z_PRIV_REQUIRED
:
3256 zerror(zlogp
, B_FALSE
, "required privilege \"%s\" is missing "
3257 "from the zone's privilege set", privname
);
3259 case Z_PRIV_UNKNOWN
:
3260 zerror(zlogp
, B_FALSE
, "unknown privilege \"%s\" specified "
3261 "in the zone's privilege set", privname
);
3264 zerror(zlogp
, B_FALSE
, "failed to determine the zone's "
3270 zonecfg_fini_handle(handle
);
3275 get_rctls(zlog_t
*zlogp
, char **bufp
, size_t *bufsizep
)
3277 nvlist_t
*nvl
= NULL
;
3278 char *nvl_packed
= NULL
;
3279 size_t nvl_size
= 0;
3280 nvlist_t
**nvlv
= NULL
;
3283 zone_dochandle_t handle
;
3284 struct zone_rctltab rctltab
;
3285 rctlblk_t
*rctlblk
= NULL
;
3292 if ((handle
= zonecfg_init_handle()) == NULL
) {
3293 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
3296 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
) {
3297 zerror(zlogp
, B_FALSE
, "invalid configuration");
3298 zonecfg_fini_handle(handle
);
3302 rctltab
.zone_rctl_valptr
= NULL
;
3303 if (nvlist_alloc(&nvl
, NV_UNIQUE_NAME
, 0) != 0) {
3304 zerror(zlogp
, B_TRUE
, "%s failed", "nvlist_alloc");
3309 * Allow the administrator to control both the maximum number of
3310 * process table slots and the maximum number of lwps with just the
3311 * max-processes property. If only the max-processes property is set,
3312 * we add a max-lwps property with a limit derived from max-processes.
3314 if (zonecfg_get_aliased_rctl(handle
, ALIAS_MAXPROCS
, &maxprocs
)
3316 zonecfg_get_aliased_rctl(handle
, ALIAS_MAXLWPS
, &maxlwps
)
3318 if (zonecfg_set_aliased_rctl(handle
, ALIAS_MAXLWPS
,
3319 maxprocs
* LWPS_PER_PROCESS
) != Z_OK
) {
3320 zerror(zlogp
, B_FALSE
, "unable to set max-lwps alias");
3325 if (zonecfg_setrctlent(handle
) != Z_OK
) {
3326 zerror(zlogp
, B_FALSE
, "%s failed", "zonecfg_setrctlent");
3330 if ((rctlblk
= malloc(rctlblk_size())) == NULL
) {
3331 zerror(zlogp
, B_TRUE
, "memory allocation failed");
3334 while (zonecfg_getrctlent(handle
, &rctltab
) == Z_OK
) {
3335 struct zone_rctlvaltab
*rctlval
;
3337 const char *name
= rctltab
.zone_rctl_name
;
3339 /* zoneadm should have already warned about unknown rctls. */
3340 if (!zonecfg_is_rctl(name
)) {
3341 zonecfg_free_rctl_value_list(rctltab
.zone_rctl_valptr
);
3342 rctltab
.zone_rctl_valptr
= NULL
;
3346 for (rctlval
= rctltab
.zone_rctl_valptr
; rctlval
!= NULL
;
3347 rctlval
= rctlval
->zone_rctlval_next
) {
3350 if (count
== 0) { /* ignore */
3351 continue; /* Nothing to free */
3353 if ((nvlv
= malloc(sizeof (*nvlv
) * count
)) == NULL
)
3356 for (rctlval
= rctltab
.zone_rctl_valptr
; rctlval
!= NULL
;
3357 rctlval
= rctlval
->zone_rctlval_next
, i
++) {
3358 if (nvlist_alloc(&nvlv
[i
], NV_UNIQUE_NAME
, 0) != 0) {
3359 zerror(zlogp
, B_TRUE
, "%s failed",
3363 if (zonecfg_construct_rctlblk(rctlval
, rctlblk
)
3365 zerror(zlogp
, B_FALSE
, "invalid rctl value: "
3366 "(priv=%s,limit=%s,action=%s)",
3367 rctlval
->zone_rctlval_priv
,
3368 rctlval
->zone_rctlval_limit
,
3369 rctlval
->zone_rctlval_action
);
3372 if (!zonecfg_valid_rctl(name
, rctlblk
)) {
3373 zerror(zlogp
, B_FALSE
,
3374 "(priv=%s,limit=%s,action=%s) is not a "
3375 "valid value for rctl '%s'",
3376 rctlval
->zone_rctlval_priv
,
3377 rctlval
->zone_rctlval_limit
,
3378 rctlval
->zone_rctlval_action
,
3382 if (nvlist_add_uint64(nvlv
[i
], "privilege",
3383 rctlblk_get_privilege(rctlblk
)) != 0) {
3384 zerror(zlogp
, B_FALSE
, "%s failed",
3385 "nvlist_add_uint64");
3388 if (nvlist_add_uint64(nvlv
[i
], "limit",
3389 rctlblk_get_value(rctlblk
)) != 0) {
3390 zerror(zlogp
, B_FALSE
, "%s failed",
3391 "nvlist_add_uint64");
3394 if (nvlist_add_uint64(nvlv
[i
], "action",
3395 (uint_t
)rctlblk_get_local_action(rctlblk
, NULL
))
3397 zerror(zlogp
, B_FALSE
, "%s failed",
3398 "nvlist_add_uint64");
3402 zonecfg_free_rctl_value_list(rctltab
.zone_rctl_valptr
);
3403 rctltab
.zone_rctl_valptr
= NULL
;
3404 if (nvlist_add_nvlist_array(nvl
, (char *)name
, nvlv
, count
)
3406 zerror(zlogp
, B_FALSE
, "%s failed",
3407 "nvlist_add_nvlist_array");
3410 for (i
= 0; i
< count
; i
++)
3411 nvlist_free(nvlv
[i
]);
3416 (void) zonecfg_endrctlent(handle
);
3418 if (rctlcount
== 0) {
3422 if (nvlist_pack(nvl
, &nvl_packed
, &nvl_size
, NV_ENCODE_NATIVE
, 0)
3424 zerror(zlogp
, B_FALSE
, "%s failed", "nvlist_pack");
3430 *bufsizep
= nvl_size
;
3434 zonecfg_free_rctl_value_list(rctltab
.zone_rctl_valptr
);
3440 zonecfg_fini_handle(handle
);
3445 get_implicit_datasets(zlog_t
*zlogp
, char **retstr
)
3447 char cmdbuf
[2 * MAXPATHLEN
];
3449 if (query_hook
[0] == '\0')
3452 if (snprintf(cmdbuf
, sizeof (cmdbuf
), "%s datasets", query_hook
)
3456 if (do_subproc(zlogp
, cmdbuf
, retstr
) != 0)
3463 get_datasets(zlog_t
*zlogp
, char **bufp
, size_t *bufsizep
)
3465 zone_dochandle_t handle
;
3466 struct zone_dstab dstab
;
3467 size_t total
, offset
, len
;
3470 char *implicit_datasets
= NULL
;
3471 int implicit_len
= 0;
3476 if ((handle
= zonecfg_init_handle()) == NULL
) {
3477 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
3480 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
) {
3481 zerror(zlogp
, B_FALSE
, "invalid configuration");
3482 zonecfg_fini_handle(handle
);
3486 if (get_implicit_datasets(zlogp
, &implicit_datasets
) != 0) {
3487 zerror(zlogp
, B_FALSE
, "getting implicit datasets failed");
3491 if (zonecfg_setdsent(handle
) != Z_OK
) {
3492 zerror(zlogp
, B_FALSE
, "%s failed", "zonecfg_setdsent");
3497 while (zonecfg_getdsent(handle
, &dstab
) == Z_OK
)
3498 total
+= strlen(dstab
.zone_dataset_name
) + 1;
3499 (void) zonecfg_enddsent(handle
);
3501 if (implicit_datasets
!= NULL
)
3502 implicit_len
= strlen(implicit_datasets
);
3503 if (implicit_len
> 0)
3504 total
+= implicit_len
+ 1;
3511 if ((str
= malloc(total
)) == NULL
) {
3512 zerror(zlogp
, B_TRUE
, "memory allocation failed");
3516 if (zonecfg_setdsent(handle
) != Z_OK
) {
3517 zerror(zlogp
, B_FALSE
, "%s failed", "zonecfg_setdsent");
3521 while (zonecfg_getdsent(handle
, &dstab
) == Z_OK
) {
3522 len
= strlen(dstab
.zone_dataset_name
);
3523 (void) strlcpy(str
+ offset
, dstab
.zone_dataset_name
,
3526 if (offset
< total
- 1)
3527 str
[offset
++] = ',';
3529 (void) zonecfg_enddsent(handle
);
3531 if (implicit_len
> 0)
3532 (void) strlcpy(str
+ offset
, implicit_datasets
, total
- offset
);
3542 zonecfg_fini_handle(handle
);
3543 free(implicit_datasets
);
3549 validate_datasets(zlog_t
*zlogp
)
3551 zone_dochandle_t handle
;
3552 struct zone_dstab dstab
;
3554 libzfs_handle_t
*hdl
;
3556 if ((handle
= zonecfg_init_handle()) == NULL
) {
3557 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
3560 if (zonecfg_get_snapshot_handle(zone_name
, handle
) != Z_OK
) {
3561 zerror(zlogp
, B_FALSE
, "invalid configuration");
3562 zonecfg_fini_handle(handle
);
3566 if (zonecfg_setdsent(handle
) != Z_OK
) {
3567 zerror(zlogp
, B_FALSE
, "invalid configuration");
3568 zonecfg_fini_handle(handle
);
3572 if ((hdl
= libzfs_init()) == NULL
) {
3573 zerror(zlogp
, B_FALSE
, "opening ZFS library");
3574 zonecfg_fini_handle(handle
);
3578 while (zonecfg_getdsent(handle
, &dstab
) == Z_OK
) {
3580 if ((zhp
= zfs_open(hdl
, dstab
.zone_dataset_name
,
3581 ZFS_TYPE_FILESYSTEM
)) == NULL
) {
3582 zerror(zlogp
, B_FALSE
, "cannot open ZFS dataset '%s'",
3583 dstab
.zone_dataset_name
);
3584 zonecfg_fini_handle(handle
);
3590 * Automatically set the 'zoned' property. We check the value
3591 * first because we'll get EPERM if it is already set.
3593 if (!zfs_prop_get_int(zhp
, ZFS_PROP_ZONED
) &&
3594 zfs_prop_set(zhp
, zfs_prop_to_name(ZFS_PROP_ZONED
),
3596 zerror(zlogp
, B_FALSE
, "cannot set 'zoned' "
3597 "property for ZFS dataset '%s'\n",
3598 dstab
.zone_dataset_name
);
3599 zonecfg_fini_handle(handle
);
3607 (void) zonecfg_enddsent(handle
);
3609 zonecfg_fini_handle(handle
);
3616 * Return true if the path is its own zfs file system. We determine this
3617 * by stat-ing the path to see if it is zfs and stat-ing the parent to see
3618 * if it is a different fs.
3621 is_zonepath_zfs(char *zonepath
)
3626 struct statvfs64 buf1
, buf2
;
3628 if (statvfs64(zonepath
, &buf1
) != 0)
3631 if (strcmp(buf1
.f_basetype
, "zfs") != 0)
3634 if ((path
= strdup(zonepath
)) == NULL
)
3637 parent
= dirname(path
);
3638 res
= statvfs64(parent
, &buf2
);
3644 if (buf1
.f_fsid
== buf2
.f_fsid
)
3651 prtmount(const struct mnttab
*fs
, void *x
)
3653 zerror((zlog_t
*)x
, B_FALSE
, " %s", fs
->mnt_mountp
);
3658 * Look for zones running on the main system that are using this root (or any
3659 * subdirectory of it). Return B_TRUE and print an error if a conflicting zone
3660 * is found or if we can't tell.
3663 duplicate_zone_root(zlog_t
*zlogp
, const char *rootpath
)
3665 zoneid_t
*zids
= NULL
;
3669 char zroot
[MAXPATHLEN
];
3670 char zonename
[ZONENAME_MAX
];
3674 zids
= malloc(nzids
* sizeof (*zids
));
3676 zerror(zlogp
, B_TRUE
, "memory allocation failed");
3679 if (zone_list(zids
, &nzids
) == 0)
3684 rlen
= strlen(rootpath
);
3687 * Ignore errors; they just mean that the zone has disappeared
3688 * while we were busy.
3690 if (zone_getattr(zids
[--nzids
], ZONE_ATTR_ROOT
, zroot
,
3691 sizeof (zroot
)) == -1)
3693 zlen
= strlen(zroot
);
3696 if (strncmp(rootpath
, zroot
, zlen
) == 0 &&
3697 (zroot
[zlen
] == '\0' || zroot
[zlen
] == '/') &&
3698 (rootpath
[zlen
] == '\0' || rootpath
[zlen
] == '/')) {
3699 if (getzonenamebyid(zids
[nzids
], zonename
,
3700 sizeof (zonename
)) == -1)
3701 (void) snprintf(zonename
, sizeof (zonename
),
3702 "id %d", (int)zids
[nzids
]);
3703 zerror(zlogp
, B_FALSE
,
3704 "zone root %s already in use by zone %s",
3705 rootpath
, zonename
);
3715 * Search for loopback mounts that use this same source node (same device and
3716 * inode). Return B_TRUE if there is one or if we can't tell.
3719 duplicate_reachable_path(zlog_t
*zlogp
, const char *rootpath
)
3721 struct stat64 rst
, zst
;
3724 if (stat64(rootpath
, &rst
) == -1) {
3725 zerror(zlogp
, B_TRUE
, "can't stat %s", rootpath
);
3728 if (resolve_lofs_mnts
== NULL
&& lofs_read_mnttab(zlogp
) == -1)
3730 for (mnp
= resolve_lofs_mnts
; mnp
< resolve_lofs_mnt_max
; mnp
++) {
3731 if (mnp
->mnt_fstype
== NULL
||
3732 strcmp(MNTTYPE_LOFS
, mnp
->mnt_fstype
) != 0)
3734 /* We're looking at a loopback mount. Stat it. */
3735 if (mnp
->mnt_special
!= NULL
&&
3736 stat64(mnp
->mnt_special
, &zst
) != -1 &&
3737 rst
.st_dev
== zst
.st_dev
&& rst
.st_ino
== zst
.st_ino
) {
3738 zerror(zlogp
, B_FALSE
,
3739 "zone root %s is reachable through %s",
3740 rootpath
, mnp
->mnt_mountp
);
3748 * Set memory cap and pool info for the zone's resource management
3752 setup_zone_rm(zlog_t
*zlogp
, char *zone_name
, zoneid_t zoneid
)
3756 struct zone_mcaptab mcap
;
3757 char sched
[MAXNAMELEN
];
3758 zone_dochandle_t handle
= NULL
;
3761 if ((handle
= zonecfg_init_handle()) == NULL
) {
3762 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
3763 return (Z_BAD_HANDLE
);
3766 if ((res
= zonecfg_get_snapshot_handle(zone_name
, handle
)) != Z_OK
) {
3767 zerror(zlogp
, B_FALSE
, "invalid configuration");
3768 zonecfg_fini_handle(handle
);
3773 * If a memory cap is configured, set the cap in the kernel using
3774 * zone_setattr() and make sure the rcapd SMF service is enabled.
3776 if (zonecfg_getmcapent(handle
, &mcap
) == Z_OK
) {
3780 num
= (uint64_t)strtoull(mcap
.zone_physmem_cap
, NULL
, 10);
3781 if (zone_setattr(zoneid
, ZONE_ATTR_PHYS_MCAP
, &num
, 0) == -1) {
3782 zerror(zlogp
, B_TRUE
, "could not set zone memory cap");
3783 zonecfg_fini_handle(handle
);
3787 if (zonecfg_enable_rcapd(smf_err
, sizeof (smf_err
)) != Z_OK
) {
3788 zerror(zlogp
, B_FALSE
, "enabling system/rcap service "
3789 "failed: %s", smf_err
);
3790 zonecfg_fini_handle(handle
);
3795 /* Get the scheduling class set in the zone configuration. */
3796 if (zonecfg_get_sched_class(handle
, sched
, sizeof (sched
)) == Z_OK
&&
3797 strlen(sched
) > 0) {
3798 if (zone_setattr(zoneid
, ZONE_ATTR_SCHED_CLASS
, sched
,
3799 strlen(sched
)) == -1)
3800 zerror(zlogp
, B_TRUE
, "WARNING: unable to set the "
3801 "default scheduling class");
3803 } else if (zonecfg_get_aliased_rctl(handle
, ALIAS_SHARES
, &tmp
)
3806 * If the zone has the zone.cpu-shares rctl set then we want to
3807 * use the Fair Share Scheduler (FSS) for processes in the
3808 * zone. Check what scheduling class the zone would be running
3809 * in by default so we can print a warning and modify the class
3810 * if we wouldn't be using FSS.
3812 char class_name
[PC_CLNMSZ
];
3814 if (zonecfg_get_dflt_sched_class(handle
, class_name
,
3815 sizeof (class_name
)) != Z_OK
) {
3816 zerror(zlogp
, B_FALSE
, "WARNING: unable to determine "
3817 "the zone's scheduling class");
3819 } else if (strcmp("FSS", class_name
) != 0) {
3820 zerror(zlogp
, B_FALSE
, "WARNING: The zone.cpu-shares "
3821 "rctl is set but\nFSS is not the default "
3822 "scheduling class for\nthis zone. FSS will be "
3823 "used for processes\nin the zone but to get the "
3824 "full benefit of FSS,\nit should be the default "
3825 "scheduling class.\nSee dispadmin(1M) for more "
3828 if (zone_setattr(zoneid
, ZONE_ATTR_SCHED_CLASS
, "FSS",
3829 strlen("FSS")) == -1)
3830 zerror(zlogp
, B_TRUE
, "WARNING: unable to set "
3831 "zone scheduling class to FSS");
3836 * The next few blocks of code attempt to set up temporary pools as
3837 * well as persistent pools. In all cases we call the functions
3838 * unconditionally. Within each funtion the code will check if the
3839 * zone is actually configured for a temporary pool or persistent pool
3840 * and just return if there is nothing to do.
3842 * If we are rebooting we want to attempt to reuse any temporary pool
3843 * that was previously set up. zonecfg_bind_tmp_pool() will do the
3844 * right thing in all cases (reuse or create) based on the current
3847 if ((res
= zonecfg_bind_tmp_pool(handle
, zoneid
, pool_err
,
3848 sizeof (pool_err
))) != Z_OK
) {
3849 if (res
== Z_POOL
|| res
== Z_POOL_CREATE
|| res
== Z_POOL_BIND
)
3850 zerror(zlogp
, B_FALSE
, "%s: %s\ndedicated-cpu setting "
3851 "cannot be instantiated", zonecfg_strerror(res
),
3854 zerror(zlogp
, B_FALSE
, "could not bind zone to "
3855 "temporary pool: %s", zonecfg_strerror(res
));
3856 zonecfg_fini_handle(handle
);
3857 return (Z_POOL_BIND
);
3861 * Check if we need to warn about poold not being enabled.
3863 if (zonecfg_warn_poold(handle
)) {
3864 zerror(zlogp
, B_FALSE
, "WARNING: A range of dedicated-cpus has "
3865 "been specified\nbut the dynamic pool service is not "
3866 "enabled.\nThe system will not dynamically adjust the\n"
3867 "processor allocation within the specified range\n"
3868 "until svc:/system/pools/dynamic is enabled.\n"
3872 /* The following is a warning, not an error. */
3873 if ((res
= zonecfg_bind_pool(handle
, zoneid
, pool_err
,
3874 sizeof (pool_err
))) != Z_OK
) {
3875 if (res
== Z_POOL_BIND
)
3876 zerror(zlogp
, B_FALSE
, "WARNING: unable to bind to "
3877 "pool '%s'; using default pool.", pool_err
);
3878 else if (res
== Z_POOL
)
3879 zerror(zlogp
, B_FALSE
, "WARNING: %s: %s",
3880 zonecfg_strerror(res
), pool_err
);
3882 zerror(zlogp
, B_FALSE
, "WARNING: %s",
3883 zonecfg_strerror(res
));
3886 /* Update saved pool name in case it has changed */
3887 (void) zonecfg_get_poolname(handle
, zone_name
, pool_name
,
3888 sizeof (pool_name
));
3890 zonecfg_fini_handle(handle
);
3895 report_prop_err(zlog_t
*zlogp
, const char *name
, const char *value
, int res
)
3899 zerror(zlogp
, B_FALSE
, "%s property value is too large.", name
);
3902 case Z_INVALID_PROPERTY
:
3903 zerror(zlogp
, B_FALSE
, "%s property value \"%s\" is not valid",
3908 zerror(zlogp
, B_TRUE
, "fetching property %s: %d", name
, res
);
3914 * Sets the hostid of the new zone based on its configured value. The zone's
3915 * zone_t structure must already exist in kernel memory. 'zlogp' refers to the
3916 * log used to report errors and warnings and must be non-NULL. 'zone_namep'
3917 * is the name of the new zone and must be non-NULL. 'zoneid' is the numeric
3918 * ID of the new zone.
3920 * This function returns zero on success and a nonzero error code on failure.
3923 setup_zone_hostid(zone_dochandle_t handle
, zlog_t
*zlogp
, zoneid_t zoneid
)
3926 char hostidp
[HW_HOSTID_LEN
];
3927 unsigned int hostid
;
3929 res
= zonecfg_get_hostid(handle
, hostidp
, sizeof (hostidp
));
3931 if (res
== Z_BAD_PROPERTY
) {
3933 } else if (res
!= Z_OK
) {
3934 report_prop_err(zlogp
, "hostid", hostidp
, res
);
3938 hostid
= (unsigned int)strtoul(hostidp
, NULL
, 16);
3939 if ((res
= zone_setattr(zoneid
, ZONE_ATTR_HOSTID
, &hostid
,
3940 sizeof (hostid
))) != 0) {
3941 zerror(zlogp
, B_TRUE
,
3942 "zone hostid is not valid: %s: %d", hostidp
, res
);
3950 setup_zone_secflags(zone_dochandle_t handle
, zlog_t
*zlogp
, zoneid_t zoneid
)
3952 psecflags_t secflags
;
3953 struct zone_secflagstab tab
= {0};
3954 secflagdelta_t delt
;
3957 res
= zonecfg_lookup_secflags(handle
, &tab
);
3959 if ((res
!= Z_OK
) &&
3960 /* The general defaulting code will handle this */
3961 (res
!= Z_NO_ENTRY
) && (res
!= Z_BAD_PROPERTY
)) {
3962 zerror(zlogp
, B_FALSE
, "security-flags property is "
3963 "invalid: %d", res
);
3967 if (strlen(tab
.zone_secflags_lower
) == 0)
3968 (void) strlcpy(tab
.zone_secflags_lower
, "none",
3969 sizeof (tab
.zone_secflags_lower
));
3970 if (strlen(tab
.zone_secflags_default
) == 0)
3971 (void) strlcpy(tab
.zone_secflags_default
,
3972 tab
.zone_secflags_lower
,
3973 sizeof (tab
.zone_secflags_default
));
3974 if (strlen(tab
.zone_secflags_upper
) == 0)
3975 (void) strlcpy(tab
.zone_secflags_upper
, "all",
3976 sizeof (tab
.zone_secflags_upper
));
3978 if (secflags_parse(NULL
, tab
.zone_secflags_default
,
3980 zerror(zlogp
, B_FALSE
, "default security-flags: '%s'"
3981 "are invalid", tab
.zone_secflags_default
);
3982 return (Z_BAD_PROPERTY
);
3983 } else if (delt
.psd_ass_active
!= B_TRUE
) {
3984 zerror(zlogp
, B_FALSE
, "relative security-flags are not "
3985 "allowed in zone configuration (default "
3986 "security-flags: '%s')",
3987 tab
.zone_secflags_default
);
3988 return (Z_BAD_PROPERTY
);
3990 secflags_copy(&secflags
.psf_inherit
, &delt
.psd_assign
);
3991 secflags_copy(&secflags
.psf_effective
, &delt
.psd_assign
);
3994 if (secflags_parse(NULL
, tab
.zone_secflags_lower
,
3996 zerror(zlogp
, B_FALSE
, "lower security-flags: '%s'"
3997 "are invalid", tab
.zone_secflags_lower
);
3998 return (Z_BAD_PROPERTY
);
3999 } else if (delt
.psd_ass_active
!= B_TRUE
) {
4000 zerror(zlogp
, B_FALSE
, "relative security-flags are not "
4001 "allowed in zone configuration (lower "
4002 "security-flags: '%s')",
4003 tab
.zone_secflags_lower
);
4004 return (Z_BAD_PROPERTY
);
4006 secflags_copy(&secflags
.psf_lower
, &delt
.psd_assign
);
4009 if (secflags_parse(NULL
, tab
.zone_secflags_upper
,
4011 zerror(zlogp
, B_FALSE
, "upper security-flags: '%s'"
4012 "are invalid", tab
.zone_secflags_upper
);
4013 return (Z_BAD_PROPERTY
);
4014 } else if (delt
.psd_ass_active
!= B_TRUE
) {
4015 zerror(zlogp
, B_FALSE
, "relative security-flags are not "
4016 "allowed in zone configuration (upper "
4017 "security-flags: '%s')",
4018 tab
.zone_secflags_upper
);
4019 return (Z_BAD_PROPERTY
);
4021 secflags_copy(&secflags
.psf_upper
, &delt
.psd_assign
);
4024 if (!psecflags_validate(&secflags
)) {
4025 zerror(zlogp
, B_TRUE
, "security-flags violate invariants");
4026 return (Z_BAD_PROPERTY
);
4029 if ((res
= zone_setattr(zoneid
, ZONE_ATTR_SECFLAGS
, &secflags
,
4030 sizeof (secflags
))) != 0) {
4031 zerror(zlogp
, B_TRUE
,
4032 "security-flags couldn't be set: %d", res
);
4040 setup_zone_fs_allowed(zone_dochandle_t handle
, zlog_t
*zlogp
, zoneid_t zoneid
)
4042 char fsallowed
[ZONE_FS_ALLOWED_MAX
];
4043 char *fsallowedp
= fsallowed
;
4044 int len
= sizeof (fsallowed
);
4047 res
= zonecfg_get_fs_allowed(handle
, fsallowed
, len
);
4049 if (res
== Z_BAD_PROPERTY
) {
4050 /* No value, set the defaults */
4051 (void) strlcpy(fsallowed
, DFLT_FS_ALLOWED
, len
);
4052 } else if (res
!= Z_OK
) {
4053 report_prop_err(zlogp
, "fs-allowed", fsallowed
, res
);
4055 } else if (fsallowed
[0] == '-') {
4056 /* dropping default filesystems - use remaining list */
4057 if (fsallowed
[1] != ',')
4062 /* Has a value, append the defaults */
4063 if (strlcat(fsallowed
, ",", len
) >= len
||
4064 strlcat(fsallowed
, DFLT_FS_ALLOWED
, len
) >= len
) {
4065 report_prop_err(zlogp
, "fs-allowed", fsallowed
,
4071 if (zone_setattr(zoneid
, ZONE_ATTR_FS_ALLOWED
, fsallowedp
, len
) != 0) {
4072 zerror(zlogp
, B_TRUE
,
4073 "fs-allowed couldn't be set: %s: %d", fsallowedp
, res
);
4081 setup_zone_attrs(zlog_t
*zlogp
, char *zone_namep
, zoneid_t zoneid
)
4083 zone_dochandle_t handle
;
4086 if ((handle
= zonecfg_init_handle()) == NULL
) {
4087 zerror(zlogp
, B_TRUE
, "getting zone configuration handle");
4088 return (Z_BAD_HANDLE
);
4090 if ((res
= zonecfg_get_snapshot_handle(zone_namep
, handle
)) != Z_OK
) {
4091 zerror(zlogp
, B_FALSE
, "invalid configuration");
4095 if ((res
= setup_zone_hostid(handle
, zlogp
, zoneid
)) != Z_OK
)
4098 if ((res
= setup_zone_fs_allowed(handle
, zlogp
, zoneid
)) != Z_OK
)
4101 if ((res
= setup_zone_secflags(handle
, zlogp
, zoneid
)) != Z_OK
)
4105 zonecfg_fini_handle(handle
);
4110 vplat_create(zlog_t
*zlogp
, zone_mnt_t mount_cmd
)
4114 char rootpath
[MAXPATHLEN
];
4115 char *rctlbuf
= NULL
;
4116 size_t rctlbufsz
= 0;
4117 char *zfsbuf
= NULL
;
4118 size_t zfsbufsz
= 0;
4119 zoneid_t zoneid
= -1;
4124 zone_iptype_t iptype
;
4126 if (zone_get_rootpath(zone_name
, rootpath
, sizeof (rootpath
)) != Z_OK
) {
4127 zerror(zlogp
, B_TRUE
, "unable to determine zone root");
4130 if (zonecfg_in_alt_root())
4131 resolve_lofs(zlogp
, rootpath
, sizeof (rootpath
));
4133 if (vplat_get_iptype(zlogp
, &iptype
) < 0) {
4134 zerror(zlogp
, B_TRUE
, "unable to determine ip-type");
4142 flags
= ZCF_NET_EXCL
;
4146 if ((privs
= priv_allocset()) == NULL
) {
4147 zerror(zlogp
, B_TRUE
, "%s failed", "priv_allocset");
4150 priv_emptyset(privs
);
4151 if (get_privset(zlogp
, privs
, mount_cmd
) != 0)
4154 if (mount_cmd
== Z_MNT_BOOT
&&
4155 get_rctls(zlogp
, &rctlbuf
, &rctlbufsz
) != 0) {
4156 zerror(zlogp
, B_FALSE
, "Unable to get list of rctls");
4160 if (get_datasets(zlogp
, &zfsbuf
, &zfsbufsz
) != 0) {
4161 zerror(zlogp
, B_FALSE
, "Unable to get list of ZFS datasets");
4168 * We must do this scan twice. First, we look for zones running on the
4169 * main system that are using this root (or any subdirectory of it).
4170 * Next, we reduce to the shortest path and search for loopback mounts
4171 * that use this same source node (same device and inode).
4173 if (duplicate_zone_root(zlogp
, rootpath
))
4175 if (duplicate_reachable_path(zlogp
, rootpath
))
4178 if (ALT_MOUNT(mount_cmd
)) {
4179 root_to_lu(zlogp
, rootpath
, sizeof (rootpath
), B_TRUE
);
4182 * Forge up a special root for this zone. When a zone is
4183 * mounted, we can't let the zone have its own root because the
4184 * tools that will be used in this "scratch zone" need access
4185 * to both the zone's resources and the running machine's
4188 * Note that the mkdir here also catches read-only filesystems.
4190 if (mkdir(rootpath
, 0755) != 0 && errno
!= EEXIST
) {
4191 zerror(zlogp
, B_TRUE
, "cannot create %s", rootpath
);
4194 if (domount(zlogp
, "tmpfs", "", "swap", rootpath
) != 0)
4198 if (zonecfg_in_alt_root()) {
4200 * If we are mounting up a zone in an alternate root partition,
4201 * then we have some additional work to do before starting the
4202 * zone. First, resolve the root path down so that we're not
4203 * fooled by duplicates. Then forge up an internal name for
4206 if ((fp
= zonecfg_open_scratch("", B_TRUE
)) == NULL
) {
4207 zerror(zlogp
, B_TRUE
, "cannot open mapfile");
4210 if (zonecfg_lock_scratch(fp
) != 0) {
4211 zerror(zlogp
, B_TRUE
, "cannot lock mapfile");
4214 if (zonecfg_find_scratch(fp
, zone_name
, zonecfg_get_root(),
4216 zerror(zlogp
, B_FALSE
, "scratch zone already running");
4219 /* This is the preferred name */
4220 (void) snprintf(kernzone
, sizeof (kernzone
), "SUNWlu-%s",
4223 while (zonecfg_reverse_scratch(fp
, kernzone
, NULL
, 0, NULL
,
4225 /* This is just an arbitrary name; note "." usage */
4226 (void) snprintf(kernzone
, sizeof (kernzone
),
4227 "SUNWlu.%08lX%08lX", random(), random());
4233 if ((zoneid
= zone_create(kzone
, rootpath
, privs
, rctlbuf
,
4234 rctlbufsz
, zfsbuf
, zfsbufsz
, &xerr
, flags
)) == -1) {
4235 if (xerr
== ZE_AREMOUNTS
) {
4236 if (zonecfg_find_mounts(rootpath
, NULL
, NULL
) < 1) {
4237 zerror(zlogp
, B_FALSE
,
4238 "An unknown file-system is mounted on "
4239 "a subdirectory of %s", rootpath
);
4242 zerror(zlogp
, B_FALSE
,
4243 "These file-systems are mounted on "
4244 "subdirectories of %s:", rootpath
);
4245 (void) zonecfg_find_mounts(rootpath
,
4248 } else if (xerr
== ZE_CHROOTED
) {
4249 zerror(zlogp
, B_FALSE
, "%s: "
4250 "cannot create a zone from a chrooted "
4251 "environment", "zone_create");
4253 zerror(zlogp
, B_TRUE
, "%s failed", "zone_create");
4258 if (zonecfg_in_alt_root() &&
4259 zonecfg_add_scratch(fp
, zone_name
, kernzone
,
4260 zonecfg_get_root()) == -1) {
4261 zerror(zlogp
, B_TRUE
, "cannot add mapfile entry");
4266 * The following actions are not performed when merely mounting a zone
4267 * for administrative use.
4269 if (mount_cmd
== Z_MNT_BOOT
) {
4271 struct brand_attr attr
;
4272 char modname
[MAXPATHLEN
];
4274 if (setup_zone_attrs(zlogp
, zone_name
, zoneid
) != Z_OK
)
4277 if ((bh
= brand_open(brand_name
)) == NULL
) {
4278 zerror(zlogp
, B_FALSE
,
4279 "unable to determine brand name");
4284 * If this brand requires any kernel support, now is the time to
4285 * get it loaded and initialized.
4287 if (brand_get_modname(bh
, modname
, MAXPATHLEN
) < 0) {
4289 zerror(zlogp
, B_FALSE
,
4290 "unable to determine brand kernel module");
4295 if (strlen(modname
) > 0) {
4296 (void) strlcpy(attr
.ba_brandname
, brand_name
,
4297 sizeof (attr
.ba_brandname
));
4298 (void) strlcpy(attr
.ba_modname
, modname
,
4299 sizeof (attr
.ba_modname
));
4300 if (zone_setattr(zoneid
, ZONE_ATTR_BRAND
, &attr
,
4301 sizeof (attr
) != 0)) {
4302 zerror(zlogp
, B_TRUE
,
4303 "could not set zone brand attribute.");
4308 if (setup_zone_rm(zlogp
, zone_name
, zoneid
) != Z_OK
)
4317 (void) zone_shutdown(zoneid
);
4318 (void) zone_destroy(zoneid
);
4321 priv_freeset(privs
);
4323 zonecfg_close_scratch(fp
);
4324 lofs_discard_mnttab();
4329 * Enter the zone and write a /etc/zones/index file there. This allows
4330 * libzonecfg (and thus zoneadm) to report the UUID and potentially other zone
4331 * details from inside the zone.
4334 write_index_file(zoneid_t zoneid
)
4338 struct zoneent
*zep
;
4343 char uuidstr
[UUID_PRINTABLE_STRING_LENGTH
];
4345 /* Locate the zone entry in the global zone's index file */
4346 if ((zef
= setzoneent()) == NULL
)
4348 while ((zep
= getzoneent_private(zef
)) != NULL
) {
4349 if (strcmp(zep
->zone_name
, zone_name
) == 0)
4357 if ((tmpl_fd
= init_template()) == -1) {
4362 if ((child
= fork()) == -1) {
4363 (void) ct_tmpl_clear(tmpl_fd
);
4364 (void) close(tmpl_fd
);
4369 /* parent waits for child to finish */
4372 if (contract_latest(&ct
) == -1)
4374 (void) ct_tmpl_clear(tmpl_fd
);
4375 (void) close(tmpl_fd
);
4376 (void) waitpid(child
, NULL
, 0);
4377 (void) contract_abandon_id(ct
);
4381 /* child enters zone and sets up index file */
4382 (void) ct_tmpl_clear(tmpl_fd
);
4383 if (zone_enter(zoneid
) != -1) {
4384 (void) mkdir(ZONE_CONFIG_ROOT
, ZONE_CONFIG_MODE
);
4385 (void) chown(ZONE_CONFIG_ROOT
, ZONE_CONFIG_UID
,
4387 fd
= open(ZONE_INDEX_FILE
, O_WRONLY
|O_CREAT
|O_TRUNC
,
4389 if (fd
!= -1 && (zet
= fdopen(fd
, "w")) != NULL
) {
4390 (void) fchown(fd
, ZONE_INDEX_UID
, ZONE_INDEX_GID
);
4391 if (uuid_is_null(zep
->zone_uuid
))
4394 uuid_unparse(zep
->zone_uuid
, uuidstr
);
4395 (void) fprintf(zet
, "%s:%s:/:%s\n", zep
->zone_name
,
4396 zone_state_str(zep
->zone_state
),
4405 vplat_bringup(zlog_t
*zlogp
, zone_mnt_t mount_cmd
, zoneid_t zoneid
)
4407 char zonepath
[MAXPATHLEN
];
4409 if (mount_cmd
== Z_MNT_BOOT
&& validate_datasets(zlogp
) != 0) {
4410 lofs_discard_mnttab();
4415 * Before we try to mount filesystems we need to create the
4416 * attribute backing store for /dev
4418 if (zone_get_zonepath(zone_name
, zonepath
, sizeof (zonepath
)) != Z_OK
) {
4419 lofs_discard_mnttab();
4422 resolve_lofs(zlogp
, zonepath
, sizeof (zonepath
));
4424 /* Make /dev directory owned by root, grouped sys */
4425 if (make_one_dir(zlogp
, zonepath
, "/dev", DEFAULT_DIR_MODE
,
4427 lofs_discard_mnttab();
4431 if (mount_filesystems(zlogp
, mount_cmd
) != 0) {
4432 lofs_discard_mnttab();
4436 if (mount_cmd
== Z_MNT_BOOT
) {
4437 zone_iptype_t iptype
;
4439 if (vplat_get_iptype(zlogp
, &iptype
) < 0) {
4440 zerror(zlogp
, B_TRUE
, "unable to determine ip-type");
4441 lofs_discard_mnttab();
4447 /* Always do this to make lo0 get configured */
4448 if (configure_shared_network_interfaces(zlogp
) != 0) {
4449 lofs_discard_mnttab();
4454 if (configure_exclusive_network_interfaces(zlogp
,
4457 lofs_discard_mnttab();
4464 write_index_file(zoneid
);
4466 lofs_discard_mnttab();
4471 lu_root_teardown(zlog_t
*zlogp
)
4473 char zroot
[MAXPATHLEN
];
4475 if (zone_get_rootpath(zone_name
, zroot
, sizeof (zroot
)) != Z_OK
) {
4476 zerror(zlogp
, B_FALSE
, "unable to determine zone root");
4479 root_to_lu(zlogp
, zroot
, sizeof (zroot
), B_FALSE
);
4482 * At this point, the processes are gone, the filesystems (save the
4483 * root) are unmounted, and the zone is on death row. But there may
4484 * still be creds floating about in the system that reference the
4485 * zone_t, and which pin down zone_rootvp causing this call to fail
4486 * with EBUSY. Thus, we try for a little while before just giving up.
4487 * (How I wish this were not true, and umount2 just did the right
4488 * thing, or tmpfs supported MS_FORCE This is a gross hack.)
4490 if (umount2(zroot
, MS_FORCE
) != 0) {
4491 if (errno
== ENOTSUP
&& umount2(zroot
, 0) == 0)
4493 if (errno
== EBUSY
) {
4496 while (--tries
>= 0) {
4498 if (umount2(zroot
, 0) == 0)
4504 zerror(zlogp
, B_TRUE
, "unable to unmount '%s'", zroot
);
4510 * Only zones in an alternate root environment have scratch zone
4513 if (zonecfg_in_alt_root()) {
4517 if ((fp
= zonecfg_open_scratch("", B_FALSE
)) == NULL
) {
4518 zerror(zlogp
, B_TRUE
, "cannot open mapfile");
4522 if (zonecfg_lock_scratch(fp
) != 0)
4523 zerror(zlogp
, B_TRUE
, "cannot lock mapfile");
4524 else if (zonecfg_delete_scratch(fp
, kernzone
) != 0)
4525 zerror(zlogp
, B_TRUE
, "cannot delete map entry");
4528 zonecfg_close_scratch(fp
);
4536 vplat_teardown(zlog_t
*zlogp
, boolean_t unmount_cmd
, boolean_t rebooting
)
4542 char zpath
[MAXPATHLEN
];
4543 char cmdbuf
[MAXPATHLEN
];
4544 brand_handle_t bh
= NULL
;
4545 dladm_status_t status
;
4546 char errmsg
[DLADM_STRSIZE
];
4550 if (zonecfg_in_alt_root()) {
4553 if ((fp
= zonecfg_open_scratch("", B_FALSE
)) == NULL
) {
4554 zerror(zlogp
, B_TRUE
, "unable to open map file");
4557 if (zonecfg_find_scratch(fp
, zone_name
, zonecfg_get_root(),
4558 kernzone
, sizeof (kernzone
)) != 0) {
4559 zerror(zlogp
, B_FALSE
, "unable to find scratch zone");
4560 zonecfg_close_scratch(fp
);
4563 zonecfg_close_scratch(fp
);
4567 if ((zoneid
= getzoneidbyname(kzone
)) == ZONE_ID_UNDEFINED
) {
4568 if (!bringup_failure_recovery
)
4569 zerror(zlogp
, B_TRUE
, "unable to get zoneid");
4571 (void) lu_root_teardown(zlogp
);
4575 if (remove_datalink_pool(zlogp
, zoneid
) != 0) {
4576 zerror(zlogp
, B_FALSE
, "unable clear datalink pool property");
4580 if (remove_datalink_protect(zlogp
, zoneid
) != 0) {
4581 zerror(zlogp
, B_FALSE
,
4582 "unable clear datalink protect property");
4587 * The datalinks assigned to the zone will be removed from the NGZ as
4588 * part of zone_shutdown() so that we need to remove protect/pool etc.
4589 * before zone_shutdown(). Even if the shutdown itself fails, the zone
4590 * will not be able to violate any constraints applied because the
4591 * datalinks are no longer available to the zone.
4593 if (zone_shutdown(zoneid
) != 0) {
4594 zerror(zlogp
, B_TRUE
, "unable to shutdown zone");
4598 /* Get the zonepath of this zone */
4599 if (zone_get_zonepath(zone_name
, zpath
, sizeof (zpath
)) != Z_OK
) {
4600 zerror(zlogp
, B_FALSE
, "unable to determine zone path");
4604 /* Get a handle to the brand info for this zone */
4605 if ((bh
= brand_open(brand_name
)) == NULL
) {
4606 zerror(zlogp
, B_FALSE
, "unable to determine zone brand");
4610 * If there is a brand 'halt' callback, execute it now to give the
4611 * brand a chance to cleanup any custom configuration.
4613 (void) strcpy(cmdbuf
, EXEC_PREFIX
);
4614 if (brand_get_halt(bh
, zone_name
, zpath
, cmdbuf
+ EXEC_LEN
,
4615 sizeof (cmdbuf
) - EXEC_LEN
) < 0) {
4617 zerror(zlogp
, B_FALSE
, "unable to determine branded zone's "
4623 if ((strlen(cmdbuf
) > EXEC_LEN
) &&
4624 (do_subproc(zlogp
, cmdbuf
, NULL
) != Z_OK
)) {
4625 zerror(zlogp
, B_FALSE
, "%s failed", cmdbuf
);
4630 zone_iptype_t iptype
;
4632 if (zone_getattr(zoneid
, ZONE_ATTR_FLAGS
, &flags
,
4633 sizeof (flags
)) < 0) {
4634 if (vplat_get_iptype(zlogp
, &iptype
) < 0) {
4635 zerror(zlogp
, B_TRUE
, "unable to determine "
4640 if (flags
& ZF_NET_EXCL
)
4641 iptype
= ZS_EXCLUSIVE
;
4648 if (unconfigure_shared_network_interfaces(zlogp
,
4650 zerror(zlogp
, B_FALSE
, "unable to unconfigure "
4651 "network interfaces in zone");
4656 if (unconfigure_exclusive_network_interfaces(zlogp
,
4658 zerror(zlogp
, B_FALSE
, "unable to unconfigure "
4659 "network interfaces in zone");
4662 status
= dladm_zone_halt(dld_handle
, zoneid
);
4663 if (status
!= DLADM_STATUS_OK
) {
4664 zerror(zlogp
, B_FALSE
, "unable to notify "
4665 "dlmgmtd of zone halt: %s",
4666 dladm_status2str(status
, errmsg
));
4672 if (!unmount_cmd
&& tcp_abort_connections(zlogp
, zoneid
) != 0) {
4673 zerror(zlogp
, B_TRUE
, "unable to abort TCP connections");
4677 if (unmount_filesystems(zlogp
, zoneid
, unmount_cmd
) != 0) {
4678 zerror(zlogp
, B_FALSE
,
4679 "unable to unmount file systems in zone");
4684 * If we are rebooting then we normally don't want to destroy an
4685 * existing temporary pool at this point so that we can just reuse it
4686 * when the zone boots back up. However, it is also possible we were
4687 * running with a temporary pool and the zone configuration has been
4688 * modified to no longer use a temporary pool. In that case we need
4689 * to destroy the temporary pool now. This case looks like the case
4690 * where we never had a temporary pool configured but
4691 * zonecfg_destroy_tmp_pool will do the right thing either way.
4694 boolean_t destroy_tmp_pool
= B_TRUE
;
4697 struct zone_psettab pset_tab
;
4698 zone_dochandle_t handle
;
4700 if ((handle
= zonecfg_init_handle()) != NULL
&&
4701 zonecfg_get_handle(zone_name
, handle
) == Z_OK
&&
4702 zonecfg_lookup_pset(handle
, &pset_tab
) == Z_OK
)
4703 destroy_tmp_pool
= B_FALSE
;
4705 zonecfg_fini_handle(handle
);
4708 if (destroy_tmp_pool
) {
4709 if ((res
= zonecfg_destroy_tmp_pool(zone_name
, pool_err
,
4710 sizeof (pool_err
))) != Z_OK
) {
4712 zerror(zlogp
, B_FALSE
, pool_err
);
4717 if (zone_destroy(zoneid
) != 0) {
4718 zerror(zlogp
, B_TRUE
, "unable to destroy zone");
4723 * Special teardown for alternate boot environments: remove the tmpfs
4724 * root for the zone and then remove it from the map file.
4726 if (unmount_cmd
&& lu_root_teardown(zlogp
) != 0)
4729 lofs_discard_mnttab();
4733 lofs_discard_mnttab();