]> git.proxmox.com Git - mirror_zfs.git/blame - module/zfs/zfs_ctldir.c
Rebase master to b117
[mirror_zfs.git] / module / zfs / zfs_ctldir.c
CommitLineData
34dc7c2f
BB
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
d164b209 22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
34dc7c2f
BB
23 * Use is subject to license terms.
24 */
25
34dc7c2f
BB
26/*
27 * ZFS control directory (a.k.a. ".zfs")
28 *
29 * This directory provides a common location for all ZFS meta-objects.
30 * Currently, this is only the 'snapshot' directory, but this may expand in the
31 * future. The elements are built using the GFS primitives, as the hierarchy
32 * does not actually exist on disk.
33 *
34 * For 'snapshot', we don't want to have all snapshots always mounted, because
35 * this would take up a huge amount of space in /etc/mnttab. We have three
36 * types of objects:
37 *
38 * ctldir ------> snapshotdir -------> snapshot
39 * |
40 * |
41 * V
42 * mounted fs
43 *
44 * The 'snapshot' node contains just enough information to lookup '..' and act
45 * as a mountpoint for the snapshot. Whenever we lookup a specific snapshot, we
46 * perform an automount of the underlying filesystem and return the
47 * corresponding vnode.
48 *
49 * All mounts are handled automatically by the kernel, but unmounts are
50 * (currently) handled from user land. The main reason is that there is no
51 * reliable way to auto-unmount the filesystem when it's "no longer in use".
52 * When the user unmounts a filesystem, we call zfsctl_unmount(), which
53 * unmounts any snapshots within the snapshot directory.
54 *
55 * The '.zfs', '.zfs/snapshot', and all directories created under
56 * '.zfs/snapshot' (ie: '.zfs/snapshot/<snapname>') are all GFS nodes and
57 * share the same vfs_t as the head filesystem (what '.zfs' lives under).
58 *
59 * File systems mounted ontop of the GFS nodes '.zfs/snapshot/<snapname>'
60 * (ie: snapshots) are ZFS nodes and have their own unique vfs_t.
61 * However, vnodes within these mounted on file systems have their v_vfsp
62 * fields set to the head filesystem to make NFS happy (see
63 * zfsctl_snapdir_lookup()). We VFS_HOLD the head filesystem's vfs_t
64 * so that it cannot be freed until all snapshots have been unmounted.
65 */
66
67#include <fs/fs_subr.h>
68#include <sys/zfs_ctldir.h>
69#include <sys/zfs_ioctl.h>
70#include <sys/zfs_vfsops.h>
71#include <sys/vfs_opreg.h>
72#include <sys/gfs.h>
73#include <sys/stat.h>
74#include <sys/dmu.h>
75#include <sys/dsl_deleg.h>
76#include <sys/mount.h>
77#include <sys/sunddi.h>
78
b128c09f
BB
79#include "zfs_namecheck.h"
80
34dc7c2f
BB
81typedef struct zfsctl_node {
82 gfs_dir_t zc_gfs_private;
83 uint64_t zc_id;
84 timestruc_t zc_cmtime; /* ctime and mtime, always the same */
85} zfsctl_node_t;
86
87typedef struct zfsctl_snapdir {
88 zfsctl_node_t sd_node;
89 kmutex_t sd_lock;
90 avl_tree_t sd_snaps;
91} zfsctl_snapdir_t;
92
93typedef struct {
94 char *se_name;
95 vnode_t *se_root;
96 avl_node_t se_node;
97} zfs_snapentry_t;
98
99static int
100snapentry_compare(const void *a, const void *b)
101{
102 const zfs_snapentry_t *sa = a;
103 const zfs_snapentry_t *sb = b;
104 int ret = strcmp(sa->se_name, sb->se_name);
105
106 if (ret < 0)
107 return (-1);
108 else if (ret > 0)
109 return (1);
110 else
111 return (0);
112}
113
114vnodeops_t *zfsctl_ops_root;
115vnodeops_t *zfsctl_ops_snapdir;
116vnodeops_t *zfsctl_ops_snapshot;
9babb374
BB
117vnodeops_t *zfsctl_ops_shares;
118vnodeops_t *zfsctl_ops_shares_dir;
34dc7c2f
BB
119
120static const fs_operation_def_t zfsctl_tops_root[];
121static const fs_operation_def_t zfsctl_tops_snapdir[];
122static const fs_operation_def_t zfsctl_tops_snapshot[];
9babb374 123static const fs_operation_def_t zfsctl_tops_shares[];
34dc7c2f
BB
124
125static vnode_t *zfsctl_mknode_snapdir(vnode_t *);
9babb374 126static vnode_t *zfsctl_mknode_shares(vnode_t *);
34dc7c2f
BB
127static vnode_t *zfsctl_snapshot_mknode(vnode_t *, uint64_t objset);
128static int zfsctl_unmount_snap(zfs_snapentry_t *, int, cred_t *);
129
130static gfs_opsvec_t zfsctl_opsvec[] = {
131 { ".zfs", zfsctl_tops_root, &zfsctl_ops_root },
132 { ".zfs/snapshot", zfsctl_tops_snapdir, &zfsctl_ops_snapdir },
133 { ".zfs/snapshot/vnode", zfsctl_tops_snapshot, &zfsctl_ops_snapshot },
9babb374
BB
134 { ".zfs/shares", zfsctl_tops_shares, &zfsctl_ops_shares_dir },
135 { ".zfs/shares/vnode", zfsctl_tops_shares, &zfsctl_ops_shares },
34dc7c2f
BB
136 { NULL }
137};
138
139/*
9babb374
BB
140 * Root directory elements. We only have two entries
141 * snapshot and shares.
34dc7c2f
BB
142 */
143static gfs_dirent_t zfsctl_root_entries[] = {
144 { "snapshot", zfsctl_mknode_snapdir, GFS_CACHE_VNODE },
9babb374 145 { "shares", zfsctl_mknode_shares, GFS_CACHE_VNODE },
34dc7c2f
BB
146 { NULL }
147};
148
149/* include . and .. in the calculation */
150#define NROOT_ENTRIES ((sizeof (zfsctl_root_entries) / \
151 sizeof (gfs_dirent_t)) + 1)
152
153
154/*
155 * Initialize the various GFS pieces we'll need to create and manipulate .zfs
156 * directories. This is called from the ZFS init routine, and initializes the
157 * vnode ops vectors that we'll be using.
158 */
159void
160zfsctl_init(void)
161{
162 VERIFY(gfs_make_opsvec(zfsctl_opsvec) == 0);
163}
164
165void
166zfsctl_fini(void)
167{
168 /*
169 * Remove vfsctl vnode ops
170 */
171 if (zfsctl_ops_root)
172 vn_freevnodeops(zfsctl_ops_root);
173 if (zfsctl_ops_snapdir)
174 vn_freevnodeops(zfsctl_ops_snapdir);
175 if (zfsctl_ops_snapshot)
176 vn_freevnodeops(zfsctl_ops_snapshot);
9babb374
BB
177 if (zfsctl_ops_shares)
178 vn_freevnodeops(zfsctl_ops_shares);
179 if (zfsctl_ops_shares_dir)
180 vn_freevnodeops(zfsctl_ops_shares_dir);
34dc7c2f
BB
181
182 zfsctl_ops_root = NULL;
183 zfsctl_ops_snapdir = NULL;
184 zfsctl_ops_snapshot = NULL;
9babb374
BB
185 zfsctl_ops_shares = NULL;
186 zfsctl_ops_shares_dir = NULL;
34dc7c2f
BB
187}
188
189/*
9babb374
BB
190 * Return the inode number associated with the 'snapshot' or
191 * 'shares' directory.
34dc7c2f
BB
192 */
193/* ARGSUSED */
194static ino64_t
195zfsctl_root_inode_cb(vnode_t *vp, int index)
196{
9babb374
BB
197 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
198
199 ASSERT(index <= 2);
200
201 if (index == 0)
202 return (ZFSCTL_INO_SNAPDIR);
203
204 return (zfsvfs->z_shares_dir);
34dc7c2f
BB
205}
206
207/*
208 * Create the '.zfs' directory. This directory is cached as part of the VFS
209 * structure. This results in a hold on the vfs_t. The code in zfs_umount()
210 * therefore checks against a vfs_count of 2 instead of 1. This reference
211 * is removed when the ctldir is destroyed in the unmount.
212 */
213void
214zfsctl_create(zfsvfs_t *zfsvfs)
215{
216 vnode_t *vp, *rvp;
217 zfsctl_node_t *zcp;
218
219 ASSERT(zfsvfs->z_ctldir == NULL);
220
221 vp = gfs_root_create(sizeof (zfsctl_node_t), zfsvfs->z_vfs,
222 zfsctl_ops_root, ZFSCTL_INO_ROOT, zfsctl_root_entries,
223 zfsctl_root_inode_cb, MAXNAMELEN, NULL, NULL);
224 zcp = vp->v_data;
225 zcp->zc_id = ZFSCTL_INO_ROOT;
226
227 VERIFY(VFS_ROOT(zfsvfs->z_vfs, &rvp) == 0);
228 ZFS_TIME_DECODE(&zcp->zc_cmtime, VTOZ(rvp)->z_phys->zp_crtime);
229 VN_RELE(rvp);
230
231 /*
232 * We're only faking the fact that we have a root of a filesystem for
233 * the sake of the GFS interfaces. Undo the flag manipulation it did
234 * for us.
235 */
236 vp->v_flag &= ~(VROOT | VNOCACHE | VNOMAP | VNOSWAP | VNOMOUNT);
237
238 zfsvfs->z_ctldir = vp;
239}
240
241/*
242 * Destroy the '.zfs' directory. Only called when the filesystem is unmounted.
243 * There might still be more references if we were force unmounted, but only
244 * new zfs_inactive() calls can occur and they don't reference .zfs
245 */
246void
247zfsctl_destroy(zfsvfs_t *zfsvfs)
248{
249 VN_RELE(zfsvfs->z_ctldir);
250 zfsvfs->z_ctldir = NULL;
251}
252
253/*
254 * Given a root znode, retrieve the associated .zfs directory.
255 * Add a hold to the vnode and return it.
256 */
257vnode_t *
258zfsctl_root(znode_t *zp)
259{
260 ASSERT(zfs_has_ctldir(zp));
261 VN_HOLD(zp->z_zfsvfs->z_ctldir);
262 return (zp->z_zfsvfs->z_ctldir);
263}
264
265/*
266 * Common open routine. Disallow any write access.
267 */
268/* ARGSUSED */
269static int
270zfsctl_common_open(vnode_t **vpp, int flags, cred_t *cr, caller_context_t *ct)
271{
272 if (flags & FWRITE)
273 return (EACCES);
274
275 return (0);
276}
277
278/*
279 * Common close routine. Nothing to do here.
280 */
281/* ARGSUSED */
282static int
283zfsctl_common_close(vnode_t *vpp, int flags, int count, offset_t off,
284 cred_t *cr, caller_context_t *ct)
285{
286 return (0);
287}
288
289/*
290 * Common access routine. Disallow writes.
291 */
292/* ARGSUSED */
293static int
294zfsctl_common_access(vnode_t *vp, int mode, int flags, cred_t *cr,
295 caller_context_t *ct)
296{
d164b209
BB
297 if (flags & V_ACE_MASK) {
298 if (mode & ACE_ALL_WRITE_PERMS)
299 return (EACCES);
300 } else {
301 if (mode & VWRITE)
302 return (EACCES);
303 }
34dc7c2f
BB
304
305 return (0);
306}
307
308/*
309 * Common getattr function. Fill in basic information.
310 */
311static void
312zfsctl_common_getattr(vnode_t *vp, vattr_t *vap)
313{
314 zfsctl_node_t *zcp = vp->v_data;
315 timestruc_t now;
316
317 vap->va_uid = 0;
318 vap->va_gid = 0;
319 vap->va_rdev = 0;
320 /*
321 * We are a purly virtual object, so we have no
322 * blocksize or allocated blocks.
323 */
324 vap->va_blksize = 0;
325 vap->va_nblocks = 0;
326 vap->va_seq = 0;
327 vap->va_fsid = vp->v_vfsp->vfs_dev;
328 vap->va_mode = S_IRUSR | S_IXUSR | S_IRGRP | S_IXGRP |
329 S_IROTH | S_IXOTH;
330 vap->va_type = VDIR;
331 /*
332 * We live in the now (for atime).
333 */
334 gethrestime(&now);
335 vap->va_atime = now;
336 vap->va_mtime = vap->va_ctime = zcp->zc_cmtime;
337}
338
339/*ARGSUSED*/
340static int
341zfsctl_common_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
342{
343 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
344 zfsctl_node_t *zcp = vp->v_data;
345 uint64_t object = zcp->zc_id;
346 zfid_short_t *zfid;
347 int i;
348
349 ZFS_ENTER(zfsvfs);
350
351 if (fidp->fid_len < SHORT_FID_LEN) {
352 fidp->fid_len = SHORT_FID_LEN;
353 ZFS_EXIT(zfsvfs);
354 return (ENOSPC);
355 }
356
357 zfid = (zfid_short_t *)fidp;
358
359 zfid->zf_len = SHORT_FID_LEN;
360
361 for (i = 0; i < sizeof (zfid->zf_object); i++)
362 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
363
364 /* .zfs znodes always have a generation number of 0 */
365 for (i = 0; i < sizeof (zfid->zf_gen); i++)
366 zfid->zf_gen[i] = 0;
367
368 ZFS_EXIT(zfsvfs);
369 return (0);
370}
371
9babb374
BB
372
373/*ARGSUSED*/
374static int
375zfsctl_shares_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
376{
377 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
378 znode_t *dzp;
379 int error;
380
381 ZFS_ENTER(zfsvfs);
382
383 if (zfsvfs->z_shares_dir == 0) {
384 ZFS_EXIT(zfsvfs);
385 return (ENOTSUP);
386 }
387
388 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
389 error = VOP_FID(ZTOV(dzp), fidp, ct);
390 VN_RELE(ZTOV(dzp));
391 }
392
393 ZFS_EXIT(zfsvfs);
394 return (error);
395}
34dc7c2f
BB
396/*
397 * .zfs inode namespace
398 *
399 * We need to generate unique inode numbers for all files and directories
400 * within the .zfs pseudo-filesystem. We use the following scheme:
401 *
402 * ENTRY ZFSCTL_INODE
403 * .zfs 1
404 * .zfs/snapshot 2
405 * .zfs/snapshot/<snap> objectid(snap)
406 */
407
408#define ZFSCTL_INO_SNAP(id) (id)
409
410/*
411 * Get root directory attributes.
412 */
413/* ARGSUSED */
414static int
415zfsctl_root_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
416 caller_context_t *ct)
417{
418 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
419
420 ZFS_ENTER(zfsvfs);
421 vap->va_nodeid = ZFSCTL_INO_ROOT;
422 vap->va_nlink = vap->va_size = NROOT_ENTRIES;
423
424 zfsctl_common_getattr(vp, vap);
425 ZFS_EXIT(zfsvfs);
426
427 return (0);
428}
429
430/*
431 * Special case the handling of "..".
432 */
433/* ARGSUSED */
434int
435zfsctl_root_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
436 int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
437 int *direntflags, pathname_t *realpnp)
438{
439 zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
440 int err;
441
442 /*
443 * No extended attributes allowed under .zfs
444 */
445 if (flags & LOOKUP_XATTR)
446 return (EINVAL);
447
448 ZFS_ENTER(zfsvfs);
449
450 if (strcmp(nm, "..") == 0) {
451 err = VFS_ROOT(dvp->v_vfsp, vpp);
452 } else {
453 err = gfs_vop_lookup(dvp, nm, vpp, pnp, flags, rdir,
454 cr, ct, direntflags, realpnp);
455 }
456
457 ZFS_EXIT(zfsvfs);
458
459 return (err);
460}
461
d164b209
BB
462static int
463zfsctl_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
464 caller_context_t *ct)
465{
466 /*
467 * We only care about ACL_ENABLED so that libsec can
468 * display ACL correctly and not default to POSIX draft.
469 */
470 if (cmd == _PC_ACL_ENABLED) {
471 *valp = _ACL_ACE_ENABLED;
472 return (0);
473 }
474
475 return (fs_pathconf(vp, cmd, valp, cr, ct));
476}
477
34dc7c2f
BB
478static const fs_operation_def_t zfsctl_tops_root[] = {
479 { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
480 { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
481 { VOPNAME_IOCTL, { .error = fs_inval } },
482 { VOPNAME_GETATTR, { .vop_getattr = zfsctl_root_getattr } },
483 { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
484 { VOPNAME_READDIR, { .vop_readdir = gfs_vop_readdir } },
485 { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_root_lookup } },
486 { VOPNAME_SEEK, { .vop_seek = fs_seek } },
487 { VOPNAME_INACTIVE, { .vop_inactive = gfs_vop_inactive } },
d164b209 488 { VOPNAME_PATHCONF, { .vop_pathconf = zfsctl_pathconf } },
34dc7c2f
BB
489 { VOPNAME_FID, { .vop_fid = zfsctl_common_fid } },
490 { NULL }
491};
492
493static int
494zfsctl_snapshot_zname(vnode_t *vp, const char *name, int len, char *zname)
495{
496 objset_t *os = ((zfsvfs_t *)((vp)->v_vfsp->vfs_data))->z_os;
497
b128c09f
BB
498 if (snapshot_namecheck(name, NULL, NULL) != 0)
499 return (EILSEQ);
34dc7c2f
BB
500 dmu_objset_name(os, zname);
501 if (strlen(zname) + 1 + strlen(name) >= len)
502 return (ENAMETOOLONG);
503 (void) strcat(zname, "@");
504 (void) strcat(zname, name);
505 return (0);
506}
507
508static int
509zfsctl_unmount_snap(zfs_snapentry_t *sep, int fflags, cred_t *cr)
510{
511 vnode_t *svp = sep->se_root;
512 int error;
513
514 ASSERT(vn_ismntpt(svp));
515
516 /* this will be dropped by dounmount() */
517 if ((error = vn_vfswlock(svp)) != 0)
518 return (error);
519
520 VN_HOLD(svp);
521 error = dounmount(vn_mountedvfs(svp), fflags, cr);
522 if (error) {
523 VN_RELE(svp);
524 return (error);
525 }
9babb374 526
34dc7c2f
BB
527 /*
528 * We can't use VN_RELE(), as that will try to invoke
529 * zfsctl_snapdir_inactive(), which would cause us to destroy
530 * the sd_lock mutex held by our caller.
531 */
532 ASSERT(svp->v_count == 1);
533 gfs_vop_inactive(svp, cr, NULL);
534
535 kmem_free(sep->se_name, strlen(sep->se_name) + 1);
536 kmem_free(sep, sizeof (zfs_snapentry_t));
537
538 return (0);
539}
540
541static void
542zfsctl_rename_snap(zfsctl_snapdir_t *sdp, zfs_snapentry_t *sep, const char *nm)
543{
544 avl_index_t where;
545 vfs_t *vfsp;
546 refstr_t *pathref;
547 char newpath[MAXNAMELEN];
548 char *tail;
549
550 ASSERT(MUTEX_HELD(&sdp->sd_lock));
551 ASSERT(sep != NULL);
552
553 vfsp = vn_mountedvfs(sep->se_root);
554 ASSERT(vfsp != NULL);
555
556 vfs_lock_wait(vfsp);
557
558 /*
559 * Change the name in the AVL tree.
560 */
561 avl_remove(&sdp->sd_snaps, sep);
562 kmem_free(sep->se_name, strlen(sep->se_name) + 1);
563 sep->se_name = kmem_alloc(strlen(nm) + 1, KM_SLEEP);
564 (void) strcpy(sep->se_name, nm);
565 VERIFY(avl_find(&sdp->sd_snaps, sep, &where) == NULL);
566 avl_insert(&sdp->sd_snaps, sep, where);
567
568 /*
569 * Change the current mountpoint info:
570 * - update the tail of the mntpoint path
571 * - update the tail of the resource path
572 */
573 pathref = vfs_getmntpoint(vfsp);
574 (void) strncpy(newpath, refstr_value(pathref), sizeof (newpath));
575 VERIFY((tail = strrchr(newpath, '/')) != NULL);
576 *(tail+1) = '\0';
577 ASSERT3U(strlen(newpath) + strlen(nm), <, sizeof (newpath));
578 (void) strcat(newpath, nm);
579 refstr_rele(pathref);
580 vfs_setmntpoint(vfsp, newpath);
581
582 pathref = vfs_getresource(vfsp);
583 (void) strncpy(newpath, refstr_value(pathref), sizeof (newpath));
584 VERIFY((tail = strrchr(newpath, '@')) != NULL);
585 *(tail+1) = '\0';
586 ASSERT3U(strlen(newpath) + strlen(nm), <, sizeof (newpath));
587 (void) strcat(newpath, nm);
588 refstr_rele(pathref);
589 vfs_setresource(vfsp, newpath);
590
591 vfs_unlock(vfsp);
592}
593
594/*ARGSUSED*/
595static int
596zfsctl_snapdir_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm,
597 cred_t *cr, caller_context_t *ct, int flags)
598{
599 zfsctl_snapdir_t *sdp = sdvp->v_data;
600 zfs_snapentry_t search, *sep;
601 zfsvfs_t *zfsvfs;
602 avl_index_t where;
603 char from[MAXNAMELEN], to[MAXNAMELEN];
604 char real[MAXNAMELEN];
605 int err;
606
607 zfsvfs = sdvp->v_vfsp->vfs_data;
608 ZFS_ENTER(zfsvfs);
609
610 if ((flags & FIGNORECASE) || zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
611 err = dmu_snapshot_realname(zfsvfs->z_os, snm, real,
612 MAXNAMELEN, NULL);
613 if (err == 0) {
614 snm = real;
615 } else if (err != ENOTSUP) {
616 ZFS_EXIT(zfsvfs);
617 return (err);
618 }
619 }
620
621 ZFS_EXIT(zfsvfs);
622
623 err = zfsctl_snapshot_zname(sdvp, snm, MAXNAMELEN, from);
624 if (!err)
625 err = zfsctl_snapshot_zname(tdvp, tnm, MAXNAMELEN, to);
626 if (!err)
627 err = zfs_secpolicy_rename_perms(from, to, cr);
628 if (err)
629 return (err);
630
631 /*
632 * Cannot move snapshots out of the snapdir.
633 */
634 if (sdvp != tdvp)
635 return (EINVAL);
636
637 if (strcmp(snm, tnm) == 0)
638 return (0);
639
640 mutex_enter(&sdp->sd_lock);
641
642 search.se_name = (char *)snm;
643 if ((sep = avl_find(&sdp->sd_snaps, &search, &where)) == NULL) {
644 mutex_exit(&sdp->sd_lock);
645 return (ENOENT);
646 }
647
648 err = dmu_objset_rename(from, to, B_FALSE);
649 if (err == 0)
650 zfsctl_rename_snap(sdp, sep, tnm);
651
652 mutex_exit(&sdp->sd_lock);
653
654 return (err);
655}
656
657/* ARGSUSED */
658static int
659zfsctl_snapdir_remove(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
660 caller_context_t *ct, int flags)
661{
662 zfsctl_snapdir_t *sdp = dvp->v_data;
663 zfs_snapentry_t *sep;
664 zfs_snapentry_t search;
665 zfsvfs_t *zfsvfs;
666 char snapname[MAXNAMELEN];
667 char real[MAXNAMELEN];
668 int err;
669
670 zfsvfs = dvp->v_vfsp->vfs_data;
671 ZFS_ENTER(zfsvfs);
672
673 if ((flags & FIGNORECASE) || zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
674
675 err = dmu_snapshot_realname(zfsvfs->z_os, name, real,
676 MAXNAMELEN, NULL);
677 if (err == 0) {
678 name = real;
679 } else if (err != ENOTSUP) {
680 ZFS_EXIT(zfsvfs);
681 return (err);
682 }
683 }
684
685 ZFS_EXIT(zfsvfs);
686
687 err = zfsctl_snapshot_zname(dvp, name, MAXNAMELEN, snapname);
688 if (!err)
689 err = zfs_secpolicy_destroy_perms(snapname, cr);
690 if (err)
691 return (err);
692
693 mutex_enter(&sdp->sd_lock);
694
695 search.se_name = name;
696 sep = avl_find(&sdp->sd_snaps, &search, NULL);
697 if (sep) {
698 avl_remove(&sdp->sd_snaps, sep);
699 err = zfsctl_unmount_snap(sep, MS_FORCE, cr);
700 if (err)
701 avl_add(&sdp->sd_snaps, sep);
702 else
703 err = dmu_objset_destroy(snapname);
704 } else {
705 err = ENOENT;
706 }
707
708 mutex_exit(&sdp->sd_lock);
709
710 return (err);
711}
712
713/*
714 * This creates a snapshot under '.zfs/snapshot'.
715 */
716/* ARGSUSED */
717static int
718zfsctl_snapdir_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp,
719 cred_t *cr, caller_context_t *cc, int flags, vsecattr_t *vsecp)
720{
721 zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
722 char name[MAXNAMELEN];
723 int err;
724 static enum symfollow follow = NO_FOLLOW;
725 static enum uio_seg seg = UIO_SYSSPACE;
726
b128c09f
BB
727 if (snapshot_namecheck(dirname, NULL, NULL) != 0)
728 return (EILSEQ);
729
34dc7c2f
BB
730 dmu_objset_name(zfsvfs->z_os, name);
731
732 *vpp = NULL;
733
734 err = zfs_secpolicy_snapshot_perms(name, cr);
735 if (err)
736 return (err);
737
738 if (err == 0) {
9babb374 739 err = dmu_objset_snapshot(name, dirname, NULL, B_FALSE);
34dc7c2f
BB
740 if (err)
741 return (err);
742 err = lookupnameat(dirname, seg, follow, NULL, vpp, dvp);
743 }
744
745 return (err);
746}
747
748/*
749 * Lookup entry point for the 'snapshot' directory. Try to open the
750 * snapshot if it exist, creating the pseudo filesystem vnode as necessary.
751 * Perform a mount of the associated dataset on top of the vnode.
752 */
753/* ARGSUSED */
754static int
755zfsctl_snapdir_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
756 int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
757 int *direntflags, pathname_t *realpnp)
758{
759 zfsctl_snapdir_t *sdp = dvp->v_data;
760 objset_t *snap;
761 char snapname[MAXNAMELEN];
762 char real[MAXNAMELEN];
763 char *mountpoint;
764 zfs_snapentry_t *sep, search;
765 struct mounta margs;
766 vfs_t *vfsp;
767 size_t mountpoint_len;
768 avl_index_t where;
769 zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
770 int err;
771
772 /*
773 * No extended attributes allowed under .zfs
774 */
775 if (flags & LOOKUP_XATTR)
776 return (EINVAL);
777
778 ASSERT(dvp->v_type == VDIR);
779
34dc7c2f
BB
780 /*
781 * If we get a recursive call, that means we got called
782 * from the domount() code while it was trying to look up the
783 * spec (which looks like a local path for zfs). We need to
784 * add some flag to domount() to tell it not to do this lookup.
785 */
786 if (MUTEX_HELD(&sdp->sd_lock))
787 return (ENOENT);
788
789 ZFS_ENTER(zfsvfs);
790
9babb374
BB
791 if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
792 ZFS_EXIT(zfsvfs);
793 return (0);
794 }
795
34dc7c2f
BB
796 if (flags & FIGNORECASE) {
797 boolean_t conflict = B_FALSE;
798
799 err = dmu_snapshot_realname(zfsvfs->z_os, nm, real,
800 MAXNAMELEN, &conflict);
801 if (err == 0) {
802 nm = real;
803 } else if (err != ENOTSUP) {
804 ZFS_EXIT(zfsvfs);
805 return (err);
806 }
807 if (realpnp)
808 (void) strlcpy(realpnp->pn_buf, nm,
809 realpnp->pn_bufsize);
810 if (conflict && direntflags)
811 *direntflags = ED_CASE_CONFLICT;
812 }
813
814 mutex_enter(&sdp->sd_lock);
815 search.se_name = (char *)nm;
816 if ((sep = avl_find(&sdp->sd_snaps, &search, &where)) != NULL) {
817 *vpp = sep->se_root;
818 VN_HOLD(*vpp);
819 err = traverse(vpp);
820 if (err) {
821 VN_RELE(*vpp);
822 *vpp = NULL;
823 } else if (*vpp == sep->se_root) {
824 /*
825 * The snapshot was unmounted behind our backs,
826 * try to remount it.
827 */
828 goto domount;
829 } else {
830 /*
831 * VROOT was set during the traverse call. We need
832 * to clear it since we're pretending to be part
833 * of our parent's vfs.
834 */
835 (*vpp)->v_flag &= ~VROOT;
836 }
837 mutex_exit(&sdp->sd_lock);
838 ZFS_EXIT(zfsvfs);
839 return (err);
840 }
841
842 /*
843 * The requested snapshot is not currently mounted, look it up.
844 */
845 err = zfsctl_snapshot_zname(dvp, nm, MAXNAMELEN, snapname);
846 if (err) {
847 mutex_exit(&sdp->sd_lock);
848 ZFS_EXIT(zfsvfs);
b128c09f
BB
849 /*
850 * handle "ls *" or "?" in a graceful manner,
851 * forcing EILSEQ to ENOENT.
852 * Since shell ultimately passes "*" or "?" as name to lookup
853 */
854 return (err == EILSEQ ? ENOENT : err);
34dc7c2f
BB
855 }
856 if (dmu_objset_open(snapname, DMU_OST_ZFS,
b128c09f 857 DS_MODE_USER | DS_MODE_READONLY, &snap) != 0) {
34dc7c2f
BB
858 mutex_exit(&sdp->sd_lock);
859 ZFS_EXIT(zfsvfs);
860 return (ENOENT);
861 }
862
863 sep = kmem_alloc(sizeof (zfs_snapentry_t), KM_SLEEP);
864 sep->se_name = kmem_alloc(strlen(nm) + 1, KM_SLEEP);
865 (void) strcpy(sep->se_name, nm);
866 *vpp = sep->se_root = zfsctl_snapshot_mknode(dvp, dmu_objset_id(snap));
867 avl_insert(&sdp->sd_snaps, sep, where);
868
869 dmu_objset_close(snap);
870domount:
871 mountpoint_len = strlen(refstr_value(dvp->v_vfsp->vfs_mntpt)) +
872 strlen("/.zfs/snapshot/") + strlen(nm) + 1;
873 mountpoint = kmem_alloc(mountpoint_len, KM_SLEEP);
874 (void) snprintf(mountpoint, mountpoint_len, "%s/.zfs/snapshot/%s",
875 refstr_value(dvp->v_vfsp->vfs_mntpt), nm);
876
877 margs.spec = snapname;
878 margs.dir = mountpoint;
879 margs.flags = MS_SYSSPACE | MS_NOMNTTAB;
880 margs.fstype = "zfs";
881 margs.dataptr = NULL;
882 margs.datalen = 0;
883 margs.optptr = NULL;
884 margs.optlen = 0;
885
886 err = domount("zfs", &margs, *vpp, kcred, &vfsp);
887 kmem_free(mountpoint, mountpoint_len);
888
889 if (err == 0) {
890 /*
891 * Return the mounted root rather than the covered mount point.
892 * Takes the GFS vnode at .zfs/snapshot/<snapname> and returns
893 * the ZFS vnode mounted on top of the GFS node. This ZFS
9babb374 894 * vnode is the root of the newly created vfsp.
34dc7c2f
BB
895 */
896 VFS_RELE(vfsp);
897 err = traverse(vpp);
898 }
899
900 if (err == 0) {
901 /*
902 * Fix up the root vnode mounted on .zfs/snapshot/<snapname>.
903 *
904 * This is where we lie about our v_vfsp in order to
905 * make .zfs/snapshot/<snapname> accessible over NFS
906 * without requiring manual mounts of <snapname>.
907 */
908 ASSERT(VTOZ(*vpp)->z_zfsvfs != zfsvfs);
909 VTOZ(*vpp)->z_zfsvfs->z_parent = zfsvfs;
910 (*vpp)->v_vfsp = zfsvfs->z_vfs;
911 (*vpp)->v_flag &= ~VROOT;
912 }
913 mutex_exit(&sdp->sd_lock);
914 ZFS_EXIT(zfsvfs);
915
916 /*
917 * If we had an error, drop our hold on the vnode and
918 * zfsctl_snapshot_inactive() will clean up.
919 */
920 if (err) {
921 VN_RELE(*vpp);
922 *vpp = NULL;
923 }
924 return (err);
925}
926
9babb374
BB
927/* ARGSUSED */
928static int
929zfsctl_shares_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
930 int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
931 int *direntflags, pathname_t *realpnp)
932{
933 zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
934 znode_t *dzp;
935 int error;
936
937 ZFS_ENTER(zfsvfs);
938
939 if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
940 ZFS_EXIT(zfsvfs);
941 return (0);
942 }
943
944 if (zfsvfs->z_shares_dir == 0) {
945 ZFS_EXIT(zfsvfs);
946 return (ENOTSUP);
947 }
948 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0)
949 error = VOP_LOOKUP(ZTOV(dzp), nm, vpp, pnp,
950 flags, rdir, cr, ct, direntflags, realpnp);
951
952 VN_RELE(ZTOV(dzp));
953 ZFS_EXIT(zfsvfs);
954
955 return (error);
956}
957
34dc7c2f
BB
958/* ARGSUSED */
959static int
960zfsctl_snapdir_readdir_cb(vnode_t *vp, void *dp, int *eofp,
961 offset_t *offp, offset_t *nextp, void *data, int flags)
962{
963 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
964 char snapname[MAXNAMELEN];
965 uint64_t id, cookie;
966 boolean_t case_conflict;
967 int error;
968
969 ZFS_ENTER(zfsvfs);
970
971 cookie = *offp;
972 error = dmu_snapshot_list_next(zfsvfs->z_os, MAXNAMELEN, snapname, &id,
973 &cookie, &case_conflict);
974 if (error) {
975 ZFS_EXIT(zfsvfs);
976 if (error == ENOENT) {
977 *eofp = 1;
978 return (0);
979 }
980 return (error);
981 }
982
983 if (flags & V_RDDIR_ENTFLAGS) {
984 edirent_t *eodp = dp;
985
986 (void) strcpy(eodp->ed_name, snapname);
987 eodp->ed_ino = ZFSCTL_INO_SNAP(id);
988 eodp->ed_eflags = case_conflict ? ED_CASE_CONFLICT : 0;
989 } else {
990 struct dirent64 *odp = dp;
991
992 (void) strcpy(odp->d_name, snapname);
993 odp->d_ino = ZFSCTL_INO_SNAP(id);
994 }
995 *nextp = cookie;
996
997 ZFS_EXIT(zfsvfs);
998
999 return (0);
1000}
1001
9babb374
BB
1002/* ARGSUSED */
1003static int
1004zfsctl_shares_readdir(vnode_t *vp, uio_t *uiop, cred_t *cr, int *eofp,
1005 caller_context_t *ct, int flags)
1006{
1007 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
1008 znode_t *dzp;
1009 int error;
1010
1011 ZFS_ENTER(zfsvfs);
1012
1013 if (zfsvfs->z_shares_dir == 0) {
1014 ZFS_EXIT(zfsvfs);
1015 return (ENOTSUP);
1016 }
1017 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
1018 error = VOP_READDIR(ZTOV(dzp), uiop, cr, eofp, ct, flags);
1019 VN_RELE(ZTOV(dzp));
1020 } else {
1021 *eofp = 1;
1022 error = ENOENT;
1023 }
1024
1025 ZFS_EXIT(zfsvfs);
1026 return (error);
1027}
1028
34dc7c2f
BB
1029/*
1030 * pvp is the '.zfs' directory (zfsctl_node_t).
1031 * Creates vp, which is '.zfs/snapshot' (zfsctl_snapdir_t).
1032 *
1033 * This function is the callback to create a GFS vnode for '.zfs/snapshot'
1034 * when a lookup is performed on .zfs for "snapshot".
1035 */
1036vnode_t *
1037zfsctl_mknode_snapdir(vnode_t *pvp)
1038{
1039 vnode_t *vp;
1040 zfsctl_snapdir_t *sdp;
1041
1042 vp = gfs_dir_create(sizeof (zfsctl_snapdir_t), pvp,
1043 zfsctl_ops_snapdir, NULL, NULL, MAXNAMELEN,
1044 zfsctl_snapdir_readdir_cb, NULL);
1045 sdp = vp->v_data;
1046 sdp->sd_node.zc_id = ZFSCTL_INO_SNAPDIR;
1047 sdp->sd_node.zc_cmtime = ((zfsctl_node_t *)pvp->v_data)->zc_cmtime;
1048 mutex_init(&sdp->sd_lock, NULL, MUTEX_DEFAULT, NULL);
1049 avl_create(&sdp->sd_snaps, snapentry_compare,
1050 sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node));
1051 return (vp);
1052}
1053
9babb374
BB
1054vnode_t *
1055zfsctl_mknode_shares(vnode_t *pvp)
1056{
1057 vnode_t *vp;
1058 zfsctl_node_t *sdp;
1059
1060 vp = gfs_dir_create(sizeof (zfsctl_node_t), pvp,
1061 zfsctl_ops_shares, NULL, NULL, MAXNAMELEN,
1062 NULL, NULL);
1063 sdp = vp->v_data;
1064 sdp->zc_cmtime = ((zfsctl_node_t *)pvp->v_data)->zc_cmtime;
1065 return (vp);
1066
1067}
1068
1069/* ARGSUSED */
1070static int
1071zfsctl_shares_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
1072 caller_context_t *ct)
1073{
1074 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
1075 znode_t *dzp;
1076 int error;
1077
1078 ZFS_ENTER(zfsvfs);
1079 if (zfsvfs->z_shares_dir == 0) {
1080 ZFS_EXIT(zfsvfs);
1081 return (ENOTSUP);
1082 }
1083 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
1084 error = VOP_GETATTR(ZTOV(dzp), vap, flags, cr, ct);
1085 VN_RELE(ZTOV(dzp));
1086 }
1087 ZFS_EXIT(zfsvfs);
1088 return (error);
1089
1090
1091}
1092
34dc7c2f
BB
1093/* ARGSUSED */
1094static int
1095zfsctl_snapdir_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
1096 caller_context_t *ct)
1097{
1098 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
1099 zfsctl_snapdir_t *sdp = vp->v_data;
1100
1101 ZFS_ENTER(zfsvfs);
1102 zfsctl_common_getattr(vp, vap);
1103 vap->va_nodeid = gfs_file_inode(vp);
1104 vap->va_nlink = vap->va_size = avl_numnodes(&sdp->sd_snaps) + 2;
1105 ZFS_EXIT(zfsvfs);
1106
1107 return (0);
1108}
1109
1110/* ARGSUSED */
1111static void
1112zfsctl_snapdir_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
1113{
1114 zfsctl_snapdir_t *sdp = vp->v_data;
1115 void *private;
1116
1117 private = gfs_dir_inactive(vp);
1118 if (private != NULL) {
1119 ASSERT(avl_numnodes(&sdp->sd_snaps) == 0);
1120 mutex_destroy(&sdp->sd_lock);
1121 avl_destroy(&sdp->sd_snaps);
1122 kmem_free(private, sizeof (zfsctl_snapdir_t));
1123 }
1124}
1125
1126static const fs_operation_def_t zfsctl_tops_snapdir[] = {
1127 { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
1128 { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
1129 { VOPNAME_IOCTL, { .error = fs_inval } },
1130 { VOPNAME_GETATTR, { .vop_getattr = zfsctl_snapdir_getattr } },
1131 { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
1132 { VOPNAME_RENAME, { .vop_rename = zfsctl_snapdir_rename } },
1133 { VOPNAME_RMDIR, { .vop_rmdir = zfsctl_snapdir_remove } },
1134 { VOPNAME_MKDIR, { .vop_mkdir = zfsctl_snapdir_mkdir } },
1135 { VOPNAME_READDIR, { .vop_readdir = gfs_vop_readdir } },
1136 { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_snapdir_lookup } },
1137 { VOPNAME_SEEK, { .vop_seek = fs_seek } },
1138 { VOPNAME_INACTIVE, { .vop_inactive = zfsctl_snapdir_inactive } },
1139 { VOPNAME_FID, { .vop_fid = zfsctl_common_fid } },
1140 { NULL }
1141};
1142
9babb374
BB
1143static const fs_operation_def_t zfsctl_tops_shares[] = {
1144 { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
1145 { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
1146 { VOPNAME_IOCTL, { .error = fs_inval } },
1147 { VOPNAME_GETATTR, { .vop_getattr = zfsctl_shares_getattr } },
1148 { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
1149 { VOPNAME_READDIR, { .vop_readdir = zfsctl_shares_readdir } },
1150 { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_shares_lookup } },
1151 { VOPNAME_SEEK, { .vop_seek = fs_seek } },
1152 { VOPNAME_INACTIVE, { .vop_inactive = gfs_vop_inactive } },
1153 { VOPNAME_FID, { .vop_fid = zfsctl_shares_fid } },
1154 { NULL }
1155};
1156
34dc7c2f
BB
1157/*
1158 * pvp is the GFS vnode '.zfs/snapshot'.
1159 *
1160 * This creates a GFS node under '.zfs/snapshot' representing each
1161 * snapshot. This newly created GFS node is what we mount snapshot
1162 * vfs_t's ontop of.
1163 */
1164static vnode_t *
1165zfsctl_snapshot_mknode(vnode_t *pvp, uint64_t objset)
1166{
1167 vnode_t *vp;
1168 zfsctl_node_t *zcp;
1169
1170 vp = gfs_dir_create(sizeof (zfsctl_node_t), pvp,
1171 zfsctl_ops_snapshot, NULL, NULL, MAXNAMELEN, NULL, NULL);
1172 zcp = vp->v_data;
1173 zcp->zc_id = objset;
34dc7c2f
BB
1174
1175 return (vp);
1176}
1177
1178static void
1179zfsctl_snapshot_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
1180{
1181 zfsctl_snapdir_t *sdp;
1182 zfs_snapentry_t *sep, *next;
1183 vnode_t *dvp;
1184
1185 VERIFY(gfs_dir_lookup(vp, "..", &dvp, cr, 0, NULL, NULL) == 0);
1186 sdp = dvp->v_data;
1187
1188 mutex_enter(&sdp->sd_lock);
1189
1190 if (vp->v_count > 1) {
1191 mutex_exit(&sdp->sd_lock);
1192 return;
1193 }
1194 ASSERT(!vn_ismntpt(vp));
1195
1196 sep = avl_first(&sdp->sd_snaps);
1197 while (sep != NULL) {
1198 next = AVL_NEXT(&sdp->sd_snaps, sep);
1199
1200 if (sep->se_root == vp) {
1201 avl_remove(&sdp->sd_snaps, sep);
1202 kmem_free(sep->se_name, strlen(sep->se_name) + 1);
1203 kmem_free(sep, sizeof (zfs_snapentry_t));
1204 break;
1205 }
1206 sep = next;
1207 }
1208 ASSERT(sep != NULL);
1209
1210 mutex_exit(&sdp->sd_lock);
1211 VN_RELE(dvp);
34dc7c2f
BB
1212
1213 /*
1214 * Dispose of the vnode for the snapshot mount point.
1215 * This is safe to do because once this entry has been removed
1216 * from the AVL tree, it can't be found again, so cannot become
1217 * "active". If we lookup the same name again we will end up
1218 * creating a new vnode.
1219 */
1220 gfs_vop_inactive(vp, cr, ct);
1221}
1222
1223
1224/*
1225 * These VP's should never see the light of day. They should always
1226 * be covered.
1227 */
1228static const fs_operation_def_t zfsctl_tops_snapshot[] = {
1229 VOPNAME_INACTIVE, { .vop_inactive = zfsctl_snapshot_inactive },
1230 NULL, NULL
1231};
1232
1233int
1234zfsctl_lookup_objset(vfs_t *vfsp, uint64_t objsetid, zfsvfs_t **zfsvfsp)
1235{
1236 zfsvfs_t *zfsvfs = vfsp->vfs_data;
1237 vnode_t *dvp, *vp;
1238 zfsctl_snapdir_t *sdp;
1239 zfsctl_node_t *zcp;
1240 zfs_snapentry_t *sep;
1241 int error;
1242
1243 ASSERT(zfsvfs->z_ctldir != NULL);
1244 error = zfsctl_root_lookup(zfsvfs->z_ctldir, "snapshot", &dvp,
1245 NULL, 0, NULL, kcred, NULL, NULL, NULL);
1246 if (error != 0)
1247 return (error);
1248 sdp = dvp->v_data;
1249
1250 mutex_enter(&sdp->sd_lock);
1251 sep = avl_first(&sdp->sd_snaps);
1252 while (sep != NULL) {
1253 vp = sep->se_root;
1254 zcp = vp->v_data;
1255 if (zcp->zc_id == objsetid)
1256 break;
1257
1258 sep = AVL_NEXT(&sdp->sd_snaps, sep);
1259 }
1260
1261 if (sep != NULL) {
1262 VN_HOLD(vp);
1263 /*
1264 * Return the mounted root rather than the covered mount point.
1265 * Takes the GFS vnode at .zfs/snapshot/<snapshot objsetid>
1266 * and returns the ZFS vnode mounted on top of the GFS node.
1267 * This ZFS vnode is the root of the vfs for objset 'objsetid'.
1268 */
1269 error = traverse(&vp);
1270 if (error == 0) {
1271 if (vp == sep->se_root)
1272 error = EINVAL;
1273 else
1274 *zfsvfsp = VTOZ(vp)->z_zfsvfs;
1275 }
1276 mutex_exit(&sdp->sd_lock);
1277 VN_RELE(vp);
1278 } else {
1279 error = EINVAL;
1280 mutex_exit(&sdp->sd_lock);
1281 }
1282
1283 VN_RELE(dvp);
1284
1285 return (error);
1286}
1287
1288/*
1289 * Unmount any snapshots for the given filesystem. This is called from
1290 * zfs_umount() - if we have a ctldir, then go through and unmount all the
1291 * snapshots.
1292 */
1293int
1294zfsctl_umount_snapshots(vfs_t *vfsp, int fflags, cred_t *cr)
1295{
1296 zfsvfs_t *zfsvfs = vfsp->vfs_data;
1297 vnode_t *dvp;
1298 zfsctl_snapdir_t *sdp;
1299 zfs_snapentry_t *sep, *next;
1300 int error;
1301
1302 ASSERT(zfsvfs->z_ctldir != NULL);
1303 error = zfsctl_root_lookup(zfsvfs->z_ctldir, "snapshot", &dvp,
1304 NULL, 0, NULL, cr, NULL, NULL, NULL);
1305 if (error != 0)
1306 return (error);
1307 sdp = dvp->v_data;
1308
1309 mutex_enter(&sdp->sd_lock);
1310
1311 sep = avl_first(&sdp->sd_snaps);
1312 while (sep != NULL) {
1313 next = AVL_NEXT(&sdp->sd_snaps, sep);
1314
1315 /*
1316 * If this snapshot is not mounted, then it must
1317 * have just been unmounted by somebody else, and
1318 * will be cleaned up by zfsctl_snapdir_inactive().
1319 */
1320 if (vn_ismntpt(sep->se_root)) {
1321 avl_remove(&sdp->sd_snaps, sep);
1322 error = zfsctl_unmount_snap(sep, fflags, cr);
1323 if (error) {
1324 avl_add(&sdp->sd_snaps, sep);
1325 break;
1326 }
1327 }
1328 sep = next;
1329 }
1330
1331 mutex_exit(&sdp->sd_lock);
1332 VN_RELE(dvp);
1333
1334 return (error);
1335}