vfs_mount.c revision 1.81 1 /* $NetBSD: vfs_mount.c,v 1.81 2020/04/21 21:42:47 ad Exp $ */
2
3 /*-
4 * Copyright (c) 1997-2020 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9 * NASA Ames Research Center, by Charles M. Hannum, and by Andrew Doran.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30 * POSSIBILITY OF SUCH DAMAGE.
31 */
32
33 /*
34 * Copyright (c) 1989, 1993
35 * The Regents of the University of California. All rights reserved.
36 * (c) UNIX System Laboratories, Inc.
37 * All or some portions of this file are derived from material licensed
38 * to the University of California by American Telephone and Telegraph
39 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
40 * the permission of UNIX System Laboratories, Inc.
41 *
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
44 * are met:
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94
67 */
68
69 #include <sys/cdefs.h>
70 __KERNEL_RCSID(0, "$NetBSD: vfs_mount.c,v 1.81 2020/04/21 21:42:47 ad Exp $");
71
72 #include <sys/param.h>
73 #include <sys/kernel.h>
74
75 #include <sys/atomic.h>
76 #include <sys/buf.h>
77 #include <sys/conf.h>
78 #include <sys/fcntl.h>
79 #include <sys/filedesc.h>
80 #include <sys/device.h>
81 #include <sys/kauth.h>
82 #include <sys/kmem.h>
83 #include <sys/module.h>
84 #include <sys/mount.h>
85 #include <sys/fstrans.h>
86 #include <sys/namei.h>
87 #include <sys/extattr.h>
88 #include <sys/syscallargs.h>
89 #include <sys/sysctl.h>
90 #include <sys/systm.h>
91 #include <sys/vfs_syscalls.h>
92 #include <sys/vnode_impl.h>
93
94 #include <miscfs/genfs/genfs.h>
95 #include <miscfs/specfs/specdev.h>
96
97 enum mountlist_type {
98 ME_MOUNT,
99 ME_MARKER
100 };
101 struct mountlist_entry {
102 TAILQ_ENTRY(mountlist_entry) me_list; /* Mount list. */
103 struct mount *me_mount; /* Actual mount if ME_MOUNT,
104 current mount else. */
105 enum mountlist_type me_type; /* Mount or marker. */
106 };
107 struct mount_iterator {
108 struct mountlist_entry mi_entry;
109 };
110
111 static struct vnode *vfs_vnode_iterator_next1(struct vnode_iterator *,
112 bool (*)(void *, struct vnode *), void *, bool);
113
114 /* Root filesystem. */
115 vnode_t * rootvnode;
116
117 extern struct mount *dead_rootmount;
118
119 /* Mounted filesystem list. */
120 static TAILQ_HEAD(mountlist, mountlist_entry) mountlist;
121 static kmutex_t mountlist_lock __cacheline_aligned;
122 int vnode_offset_next_by_lru /* XXX: ugly hack for pstat.c */
123 = offsetof(vnode_impl_t, vi_lrulist.tqe_next);
124
125 kmutex_t vfs_list_lock __cacheline_aligned;
126
127 static specificdata_domain_t mount_specificdata_domain;
128 static kmutex_t mntid_lock;
129
130 static kmutex_t mountgen_lock __cacheline_aligned;
131 static uint64_t mountgen;
132
133 void
134 vfs_mount_sysinit(void)
135 {
136
137 TAILQ_INIT(&mountlist);
138 mutex_init(&mountlist_lock, MUTEX_DEFAULT, IPL_NONE);
139 mutex_init(&vfs_list_lock, MUTEX_DEFAULT, IPL_NONE);
140
141 mount_specificdata_domain = specificdata_domain_create();
142 mutex_init(&mntid_lock, MUTEX_DEFAULT, IPL_NONE);
143 mutex_init(&mountgen_lock, MUTEX_DEFAULT, IPL_NONE);
144 mountgen = 0;
145 }
146
147 struct mount *
148 vfs_mountalloc(struct vfsops *vfsops, vnode_t *vp)
149 {
150 struct mount *mp;
151 int error __diagused;
152
153 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP);
154 mp->mnt_op = vfsops;
155 mp->mnt_refcnt = 1;
156 TAILQ_INIT(&mp->mnt_vnodelist);
157 mp->mnt_renamelock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
158 mp->mnt_vnodelock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
159 mp->mnt_updating = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
160 mp->mnt_vnodecovered = vp;
161 mount_initspecific(mp);
162
163 error = fstrans_mount(mp);
164 KASSERT(error == 0);
165
166 mutex_enter(&mountgen_lock);
167 mp->mnt_gen = mountgen++;
168 mutex_exit(&mountgen_lock);
169
170 return mp;
171 }
172
173 /*
174 * vfs_rootmountalloc: lookup a filesystem type, and if found allocate and
175 * initialize a mount structure for it.
176 *
177 * Devname is usually updated by mount(8) after booting.
178 */
179 int
180 vfs_rootmountalloc(const char *fstypename, const char *devname,
181 struct mount **mpp)
182 {
183 struct vfsops *vfsp = NULL;
184 struct mount *mp;
185 int error __diagused;
186
187 mutex_enter(&vfs_list_lock);
188 LIST_FOREACH(vfsp, &vfs_list, vfs_list)
189 if (!strncmp(vfsp->vfs_name, fstypename,
190 sizeof(mp->mnt_stat.f_fstypename)))
191 break;
192 if (vfsp == NULL) {
193 mutex_exit(&vfs_list_lock);
194 return (ENODEV);
195 }
196 vfsp->vfs_refcount++;
197 mutex_exit(&vfs_list_lock);
198
199 if ((mp = vfs_mountalloc(vfsp, NULL)) == NULL)
200 return ENOMEM;
201 error = vfs_busy(mp);
202 KASSERT(error == 0);
203 mp->mnt_flag = MNT_RDONLY;
204 (void)strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfs_name,
205 sizeof(mp->mnt_stat.f_fstypename));
206 mp->mnt_stat.f_mntonname[0] = '/';
207 mp->mnt_stat.f_mntonname[1] = '\0';
208 mp->mnt_stat.f_mntfromname[sizeof(mp->mnt_stat.f_mntfromname) - 1] =
209 '\0';
210 (void)copystr(devname, mp->mnt_stat.f_mntfromname,
211 sizeof(mp->mnt_stat.f_mntfromname) - 1, 0);
212 *mpp = mp;
213 return 0;
214 }
215
216 /*
217 * vfs_getnewfsid: get a new unique fsid.
218 */
219 void
220 vfs_getnewfsid(struct mount *mp)
221 {
222 static u_short xxxfs_mntid;
223 fsid_t tfsid;
224 int mtype;
225
226 mutex_enter(&mntid_lock);
227 mtype = makefstype(mp->mnt_op->vfs_name);
228 mp->mnt_stat.f_fsidx.__fsid_val[0] = makedev(mtype, 0);
229 mp->mnt_stat.f_fsidx.__fsid_val[1] = mtype;
230 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
231 if (xxxfs_mntid == 0)
232 ++xxxfs_mntid;
233 tfsid.__fsid_val[0] = makedev(mtype & 0xff, xxxfs_mntid);
234 tfsid.__fsid_val[1] = mtype;
235 while (vfs_getvfs(&tfsid)) {
236 tfsid.__fsid_val[0]++;
237 xxxfs_mntid++;
238 }
239 mp->mnt_stat.f_fsidx.__fsid_val[0] = tfsid.__fsid_val[0];
240 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
241 mutex_exit(&mntid_lock);
242 }
243
244 /*
245 * Lookup a mount point by filesystem identifier.
246 *
247 * XXX Needs to add a reference to the mount point.
248 */
249 struct mount *
250 vfs_getvfs(fsid_t *fsid)
251 {
252 mount_iterator_t *iter;
253 struct mount *mp;
254
255 mountlist_iterator_init(&iter);
256 while ((mp = mountlist_iterator_next(iter)) != NULL) {
257 if (mp->mnt_stat.f_fsidx.__fsid_val[0] == fsid->__fsid_val[0] &&
258 mp->mnt_stat.f_fsidx.__fsid_val[1] == fsid->__fsid_val[1]) {
259 mountlist_iterator_destroy(iter);
260 return mp;
261 }
262 }
263 mountlist_iterator_destroy(iter);
264 return NULL;
265 }
266
267 /*
268 * Take a reference to a mount structure.
269 */
270 void
271 vfs_ref(struct mount *mp)
272 {
273
274 KASSERT(mp->mnt_refcnt > 0 || mutex_owned(&mountlist_lock));
275
276 atomic_inc_uint(&mp->mnt_refcnt);
277 }
278
279 /*
280 * Drop a reference to a mount structure, freeing if the last reference.
281 */
282 void
283 vfs_rele(struct mount *mp)
284 {
285
286 if (__predict_true((int)atomic_dec_uint_nv(&mp->mnt_refcnt) > 0)) {
287 return;
288 }
289
290 /*
291 * Nothing else has visibility of the mount: we can now
292 * free the data structures.
293 */
294 KASSERT(mp->mnt_refcnt == 0);
295 specificdata_fini(mount_specificdata_domain, &mp->mnt_specdataref);
296 mutex_obj_free(mp->mnt_updating);
297 mutex_obj_free(mp->mnt_renamelock);
298 mutex_obj_free(mp->mnt_vnodelock);
299 if (mp->mnt_op != NULL) {
300 vfs_delref(mp->mnt_op);
301 }
302 fstrans_unmount(mp);
303 /*
304 * Final free of mp gets done from fstrans_mount_dtor().
305 *
306 * Prevents this memory to be reused as a mount before
307 * fstrans releases all references to it.
308 */
309 }
310
311 /*
312 * Mark a mount point as busy, and gain a new reference to it. Used to
313 * prevent the file system from being unmounted during critical sections.
314 *
315 * vfs_busy can be called multiple times and by multiple threads
316 * and must be accompanied by the same number of vfs_unbusy calls.
317 *
318 * => The caller must hold a pre-existing reference to the mount.
319 * => Will fail if the file system is being unmounted, or is unmounted.
320 */
321 static inline int
322 _vfs_busy(struct mount *mp, bool wait)
323 {
324
325 KASSERT(mp->mnt_refcnt > 0);
326
327 if (wait) {
328 fstrans_start(mp);
329 } else {
330 if (fstrans_start_nowait(mp))
331 return EBUSY;
332 }
333 if (__predict_false((mp->mnt_iflag & IMNT_GONE) != 0)) {
334 fstrans_done(mp);
335 return ENOENT;
336 }
337 vfs_ref(mp);
338 return 0;
339 }
340
341 int
342 vfs_busy(struct mount *mp)
343 {
344
345 return _vfs_busy(mp, true);
346 }
347
348 int
349 vfs_trybusy(struct mount *mp)
350 {
351
352 return _vfs_busy(mp, false);
353 }
354
355 /*
356 * Unbusy a busy filesystem.
357 *
358 * Every successful vfs_busy() call must be undone by a vfs_unbusy() call.
359 */
360 void
361 vfs_unbusy(struct mount *mp)
362 {
363
364 KASSERT(mp->mnt_refcnt > 0);
365
366 fstrans_done(mp);
367 vfs_rele(mp);
368 }
369
370 struct vnode_iterator {
371 vnode_impl_t vi_vnode;
372 };
373
374 void
375 vfs_vnode_iterator_init(struct mount *mp, struct vnode_iterator **vnip)
376 {
377 vnode_t *vp;
378 vnode_impl_t *vip;
379
380 vp = vnalloc_marker(mp);
381 vip = VNODE_TO_VIMPL(vp);
382
383 mutex_enter(mp->mnt_vnodelock);
384 TAILQ_INSERT_HEAD(&mp->mnt_vnodelist, vip, vi_mntvnodes);
385 vp->v_usecount = 1;
386 mutex_exit(mp->mnt_vnodelock);
387
388 *vnip = (struct vnode_iterator *)vip;
389 }
390
391 void
392 vfs_vnode_iterator_destroy(struct vnode_iterator *vni)
393 {
394 vnode_impl_t *mvip = &vni->vi_vnode;
395 vnode_t *mvp = VIMPL_TO_VNODE(mvip);
396 kmutex_t *lock;
397
398 KASSERT(vnis_marker(mvp));
399 if (vrefcnt(mvp) != 0) {
400 lock = mvp->v_mount->mnt_vnodelock;
401 mutex_enter(lock);
402 TAILQ_REMOVE(&mvp->v_mount->mnt_vnodelist, mvip, vi_mntvnodes);
403 mvp->v_usecount = 0;
404 mutex_exit(lock);
405 }
406 vnfree_marker(mvp);
407 }
408
409 static struct vnode *
410 vfs_vnode_iterator_next1(struct vnode_iterator *vni,
411 bool (*f)(void *, struct vnode *), void *cl, bool do_wait)
412 {
413 vnode_impl_t *mvip = &vni->vi_vnode;
414 struct mount *mp = VIMPL_TO_VNODE(mvip)->v_mount;
415 vnode_t *vp;
416 vnode_impl_t *vip;
417 kmutex_t *lock;
418 int error;
419
420 KASSERT(vnis_marker(VIMPL_TO_VNODE(mvip)));
421
422 lock = mp->mnt_vnodelock;
423 do {
424 mutex_enter(lock);
425 vip = TAILQ_NEXT(mvip, vi_mntvnodes);
426 TAILQ_REMOVE(&mp->mnt_vnodelist, mvip, vi_mntvnodes);
427 VIMPL_TO_VNODE(mvip)->v_usecount = 0;
428 again:
429 if (vip == NULL) {
430 mutex_exit(lock);
431 return NULL;
432 }
433 vp = VIMPL_TO_VNODE(vip);
434 KASSERT(vp != NULL);
435 mutex_enter(vp->v_interlock);
436 if (vnis_marker(vp) ||
437 vdead_check(vp, (do_wait ? 0 : VDEAD_NOWAIT)) ||
438 (f && !(*f)(cl, vp))) {
439 mutex_exit(vp->v_interlock);
440 vip = TAILQ_NEXT(vip, vi_mntvnodes);
441 goto again;
442 }
443
444 TAILQ_INSERT_AFTER(&mp->mnt_vnodelist, vip, mvip, vi_mntvnodes);
445 VIMPL_TO_VNODE(mvip)->v_usecount = 1;
446 mutex_exit(lock);
447 error = vcache_vget(vp);
448 KASSERT(error == 0 || error == ENOENT);
449 } while (error != 0);
450
451 return vp;
452 }
453
454 struct vnode *
455 vfs_vnode_iterator_next(struct vnode_iterator *vni,
456 bool (*f)(void *, struct vnode *), void *cl)
457 {
458
459 return vfs_vnode_iterator_next1(vni, f, cl, false);
460 }
461
462 /*
463 * Move a vnode from one mount queue to another.
464 */
465 void
466 vfs_insmntque(vnode_t *vp, struct mount *mp)
467 {
468 vnode_impl_t *vip = VNODE_TO_VIMPL(vp);
469 struct mount *omp;
470 kmutex_t *lock;
471
472 KASSERT(mp == NULL || (mp->mnt_iflag & IMNT_UNMOUNT) == 0 ||
473 vp->v_tag == VT_VFS);
474
475 /*
476 * Delete from old mount point vnode list, if on one.
477 */
478 if ((omp = vp->v_mount) != NULL) {
479 lock = omp->mnt_vnodelock;
480 mutex_enter(lock);
481 TAILQ_REMOVE(&vp->v_mount->mnt_vnodelist, vip, vi_mntvnodes);
482 mutex_exit(lock);
483 }
484
485 /*
486 * Insert into list of vnodes for the new mount point, if
487 * available. The caller must take a reference on the mount
488 * structure and donate to the vnode.
489 */
490 if ((vp->v_mount = mp) != NULL) {
491 lock = mp->mnt_vnodelock;
492 mutex_enter(lock);
493 TAILQ_INSERT_TAIL(&mp->mnt_vnodelist, vip, vi_mntvnodes);
494 mutex_exit(lock);
495 }
496
497 if (omp != NULL) {
498 /* Release reference to old mount. */
499 vfs_rele(omp);
500 }
501 }
502
503 /*
504 * Remove any vnodes in the vnode table belonging to mount point mp.
505 *
506 * If FORCECLOSE is not specified, there should not be any active ones,
507 * return error if any are found (nb: this is a user error, not a
508 * system error). If FORCECLOSE is specified, detach any active vnodes
509 * that are found.
510 *
511 * If WRITECLOSE is set, only flush out regular file vnodes open for
512 * writing.
513 *
514 * SKIPSYSTEM causes any vnodes marked VV_SYSTEM to be skipped.
515 */
516 #ifdef DEBUG
517 int busyprt = 0; /* print out busy vnodes */
518 struct ctldebug debug1 = { "busyprt", &busyprt };
519 #endif
520
521 static vnode_t *
522 vflushnext(struct vnode_iterator *marker, int *when)
523 {
524 if (getticks() > *when) {
525 yield();
526 *when = getticks() + hz / 10;
527 }
528 return vfs_vnode_iterator_next1(marker, NULL, NULL, true);
529 }
530
531 /*
532 * Flush one vnode. Referenced on entry, unreferenced on return.
533 */
534 static int
535 vflush_one(vnode_t *vp, vnode_t *skipvp, int flags)
536 {
537 int error;
538 struct vattr vattr;
539
540 if (vp == skipvp ||
541 ((flags & SKIPSYSTEM) && (vp->v_vflag & VV_SYSTEM))) {
542 vrele(vp);
543 return 0;
544 }
545 /*
546 * If WRITECLOSE is set, only flush out regular file
547 * vnodes open for writing or open and unlinked.
548 */
549 if ((flags & WRITECLOSE)) {
550 if (vp->v_type != VREG) {
551 vrele(vp);
552 return 0;
553 }
554 error = vn_lock(vp, LK_EXCLUSIVE);
555 if (error) {
556 KASSERT(error == ENOENT);
557 vrele(vp);
558 return 0;
559 }
560 error = VOP_FSYNC(vp, curlwp->l_cred, FSYNC_WAIT, 0, 0);
561 if (error == 0)
562 error = VOP_GETATTR(vp, &vattr, curlwp->l_cred);
563 VOP_UNLOCK(vp);
564 if (error) {
565 vrele(vp);
566 return error;
567 }
568 if (vp->v_writecount == 0 && vattr.va_nlink > 0) {
569 vrele(vp);
570 return 0;
571 }
572 }
573 /*
574 * First try to recycle the vnode.
575 */
576 if (vrecycle(vp))
577 return 0;
578 /*
579 * If FORCECLOSE is set, forcibly close the vnode.
580 * For block or character devices, revert to an
581 * anonymous device. For all other files, just
582 * kill them.
583 */
584 if (flags & FORCECLOSE) {
585 if (vrefcnt(vp) > 1 &&
586 (vp->v_type == VBLK || vp->v_type == VCHR))
587 vcache_make_anon(vp);
588 else
589 vgone(vp);
590 return 0;
591 }
592 vrele(vp);
593 return EBUSY;
594 }
595
596 int
597 vflush(struct mount *mp, vnode_t *skipvp, int flags)
598 {
599 vnode_t *vp;
600 struct vnode_iterator *marker;
601 int busy, error, when, retries = 2;
602
603 do {
604 busy = error = when = 0;
605
606 /*
607 * First, flush out any vnode references from the
608 * deferred vrele list.
609 */
610 vrele_flush(mp);
611
612 vfs_vnode_iterator_init(mp, &marker);
613
614 while ((vp = vflushnext(marker, &when)) != NULL) {
615 error = vflush_one(vp, skipvp, flags);
616 if (error == EBUSY) {
617 error = 0;
618 busy++;
619 #ifdef DEBUG
620 if (busyprt && retries == 0)
621 vprint("vflush: busy vnode", vp);
622 #endif
623 } else if (error != 0) {
624 break;
625 }
626 }
627
628 vfs_vnode_iterator_destroy(marker);
629 } while (error == 0 && busy > 0 && retries-- > 0);
630
631 if (error)
632 return error;
633 if (busy)
634 return EBUSY;
635 return 0;
636 }
637
638 /*
639 * Mount a file system.
640 */
641
642 /*
643 * Scan all active processes to see if any of them have a current or root
644 * directory onto which the new filesystem has just been mounted. If so,
645 * replace them with the new mount point.
646 */
647 static void
648 mount_checkdirs(vnode_t *olddp)
649 {
650 vnode_t *newdp, *rele1, *rele2;
651 struct cwdinfo *cwdi;
652 struct proc *p;
653 bool retry;
654
655 if (vrefcnt(olddp) == 1) {
656 return;
657 }
658 if (VFS_ROOT(olddp->v_mountedhere, LK_EXCLUSIVE, &newdp))
659 panic("mount: lost mount");
660
661 do {
662 retry = false;
663 mutex_enter(proc_lock);
664 PROCLIST_FOREACH(p, &allproc) {
665 if ((cwdi = p->p_cwdi) == NULL)
666 continue;
667 /*
668 * Cannot change to the old directory any more,
669 * so even if we see a stale value it is not a
670 * problem.
671 */
672 if (cwdi->cwdi_cdir != olddp &&
673 cwdi->cwdi_rdir != olddp)
674 continue;
675 retry = true;
676 rele1 = NULL;
677 rele2 = NULL;
678 atomic_inc_uint(&cwdi->cwdi_refcnt);
679 mutex_exit(proc_lock);
680 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
681 if (cwdi->cwdi_cdir == olddp) {
682 rele1 = cwdi->cwdi_cdir;
683 vref(newdp);
684 cwdi->cwdi_cdir = newdp;
685 }
686 if (cwdi->cwdi_rdir == olddp) {
687 rele2 = cwdi->cwdi_rdir;
688 vref(newdp);
689 cwdi->cwdi_rdir = newdp;
690 }
691 rw_exit(&cwdi->cwdi_lock);
692 cwdfree(cwdi);
693 if (rele1 != NULL)
694 vrele(rele1);
695 if (rele2 != NULL)
696 vrele(rele2);
697 mutex_enter(proc_lock);
698 break;
699 }
700 mutex_exit(proc_lock);
701 } while (retry);
702
703 if (rootvnode == olddp) {
704 vrele(rootvnode);
705 vref(newdp);
706 rootvnode = newdp;
707 }
708 vput(newdp);
709 }
710
711 /*
712 * Start extended attributes
713 */
714 static int
715 start_extattr(struct mount *mp)
716 {
717 int error;
718
719 error = VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, NULL, 0, NULL);
720 if (error)
721 printf("%s: failed to start extattr: error = %d\n",
722 mp->mnt_stat.f_mntonname, error);
723
724 return error;
725 }
726
727 int
728 mount_domount(struct lwp *l, vnode_t **vpp, struct vfsops *vfsops,
729 const char *path, int flags, void *data, size_t *data_len)
730 {
731 vnode_t *vp = *vpp;
732 struct mount *mp;
733 struct pathbuf *pb;
734 struct nameidata nd;
735 int error;
736
737 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
738 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
739 if (error) {
740 vfs_delref(vfsops);
741 return error;
742 }
743
744 /* Cannot make a non-dir a mount-point (from here anyway). */
745 if (vp->v_type != VDIR) {
746 vfs_delref(vfsops);
747 return ENOTDIR;
748 }
749
750 if (flags & MNT_EXPORTED) {
751 vfs_delref(vfsops);
752 return EINVAL;
753 }
754
755 if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) {
756 vfs_delref(vfsops);
757 return ENOMEM;
758 }
759
760 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
761
762 /*
763 * The underlying file system may refuse the mount for
764 * various reasons. Allow the user to force it to happen.
765 *
766 * Set the mount level flags.
767 */
768 mp->mnt_flag = flags & (MNT_BASIC_FLAGS | MNT_FORCE | MNT_IGNORE);
769
770 mutex_enter(mp->mnt_updating);
771 error = VFS_MOUNT(mp, path, data, data_len);
772 mp->mnt_flag &= ~MNT_OP_FLAGS;
773
774 if (error != 0)
775 goto err_unmounted;
776
777 /*
778 * Validate and prepare the mount point.
779 */
780 error = pathbuf_copyin(path, &pb);
781 if (error != 0) {
782 goto err_mounted;
783 }
784 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
785 error = namei(&nd);
786 pathbuf_destroy(pb);
787 if (error != 0) {
788 goto err_mounted;
789 }
790 if (nd.ni_vp != vp) {
791 vput(nd.ni_vp);
792 error = EINVAL;
793 goto err_mounted;
794 }
795 if (vp->v_mountedhere != NULL) {
796 vput(nd.ni_vp);
797 error = EBUSY;
798 goto err_mounted;
799 }
800 error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0);
801 if (error != 0) {
802 vput(nd.ni_vp);
803 goto err_mounted;
804 }
805
806 /*
807 * Put the new filesystem on the mount list after root.
808 */
809 cache_purge(vp);
810 mp->mnt_iflag &= ~IMNT_WANTRDWR;
811
812 mountlist_append(mp);
813 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
814 vfs_syncer_add_to_worklist(mp);
815 vp->v_mountedhere = mp;
816 vput(nd.ni_vp);
817
818 mount_checkdirs(vp);
819 mutex_exit(mp->mnt_updating);
820
821 /* Hold an additional reference to the mount across VFS_START(). */
822 vfs_ref(mp);
823 (void) VFS_STATVFS(mp, &mp->mnt_stat);
824 error = VFS_START(mp, 0);
825 if (error) {
826 vrele(vp);
827 } else if (flags & MNT_EXTATTR) {
828 if (start_extattr(mp) != 0)
829 mp->mnt_flag &= ~MNT_EXTATTR;
830 }
831 /* Drop reference held for VFS_START(). */
832 vfs_rele(mp);
833 *vpp = NULL;
834 return error;
835
836 err_mounted:
837 if (VFS_UNMOUNT(mp, MNT_FORCE) != 0)
838 panic("Unmounting fresh file system failed");
839
840 err_unmounted:
841 vp->v_mountedhere = NULL;
842 mutex_exit(mp->mnt_updating);
843 vfs_rele(mp);
844
845 return error;
846 }
847
848 /*
849 * Do the actual file system unmount. File system is assumed to have
850 * been locked by the caller.
851 *
852 * => Caller hold reference to the mount, explicitly for dounmount().
853 */
854 int
855 dounmount(struct mount *mp, int flags, struct lwp *l)
856 {
857 vnode_t *coveredvp;
858 int error, async, used_syncer, used_extattr;
859 const bool was_suspended = fstrans_is_owner(mp);
860
861 #if NVERIEXEC > 0
862 error = veriexec_unmountchk(mp);
863 if (error)
864 return (error);
865 #endif /* NVERIEXEC > 0 */
866
867 if (!was_suspended) {
868 error = vfs_suspend(mp, 0);
869 if (error) {
870 return error;
871 }
872 }
873
874 KASSERT((mp->mnt_iflag & IMNT_GONE) == 0);
875
876 used_syncer = (mp->mnt_iflag & IMNT_ONWORKLIST) != 0;
877 used_extattr = mp->mnt_flag & MNT_EXTATTR;
878
879 mp->mnt_iflag |= IMNT_UNMOUNT;
880 mutex_enter(mp->mnt_updating);
881 async = mp->mnt_flag & MNT_ASYNC;
882 mp->mnt_flag &= ~MNT_ASYNC;
883 cache_purgevfs(mp); /* remove cache entries for this file sys */
884 if (used_syncer)
885 vfs_syncer_remove_from_worklist(mp);
886 error = 0;
887 if (((mp->mnt_flag & MNT_RDONLY) == 0) && ((flags & MNT_FORCE) == 0)) {
888 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
889 }
890 if (error == 0 || (flags & MNT_FORCE)) {
891 error = VFS_UNMOUNT(mp, flags);
892 }
893 if (error) {
894 mp->mnt_iflag &= ~IMNT_UNMOUNT;
895 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
896 vfs_syncer_add_to_worklist(mp);
897 mp->mnt_flag |= async;
898 mutex_exit(mp->mnt_updating);
899 if (!was_suspended)
900 vfs_resume(mp);
901 if (used_extattr) {
902 if (start_extattr(mp) != 0)
903 mp->mnt_flag &= ~MNT_EXTATTR;
904 else
905 mp->mnt_flag |= MNT_EXTATTR;
906 }
907 return (error);
908 }
909 mutex_exit(mp->mnt_updating);
910
911 /*
912 * mark filesystem as gone to prevent further umounts
913 * after mnt_umounting lock is gone, this also prevents
914 * vfs_busy() from succeeding.
915 */
916 mp->mnt_iflag |= IMNT_GONE;
917 if (!was_suspended)
918 vfs_resume(mp);
919
920 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
921 vn_lock(coveredvp, LK_EXCLUSIVE | LK_RETRY);
922 coveredvp->v_mountedhere = NULL;
923 VOP_UNLOCK(coveredvp);
924 }
925 mountlist_remove(mp);
926 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
927 panic("unmount: dangling vnode");
928 vfs_hooks_unmount(mp);
929
930 vfs_rele(mp); /* reference from mount() */
931 if (coveredvp != NULLVP) {
932 vrele(coveredvp);
933 }
934 return (0);
935 }
936
937 /*
938 * Unmount all file systems.
939 * We traverse the list in reverse order under the assumption that doing so
940 * will avoid needing to worry about dependencies.
941 */
942 bool
943 vfs_unmountall(struct lwp *l)
944 {
945
946 printf("unmounting file systems...\n");
947 return vfs_unmountall1(l, true, true);
948 }
949
950 static void
951 vfs_unmount_print(struct mount *mp, const char *pfx)
952 {
953
954 aprint_verbose("%sunmounted %s on %s type %s\n", pfx,
955 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname,
956 mp->mnt_stat.f_fstypename);
957 }
958
959 /*
960 * Return the mount with the highest generation less than "gen".
961 */
962 static struct mount *
963 vfs_unmount_next(uint64_t gen)
964 {
965 mount_iterator_t *iter;
966 struct mount *mp, *nmp;
967
968 nmp = NULL;
969
970 mountlist_iterator_init(&iter);
971 while ((mp = mountlist_iterator_next(iter)) != NULL) {
972 if ((nmp == NULL || mp->mnt_gen > nmp->mnt_gen) &&
973 mp->mnt_gen < gen) {
974 if (nmp != NULL)
975 vfs_rele(nmp);
976 nmp = mp;
977 vfs_ref(nmp);
978 }
979 }
980 mountlist_iterator_destroy(iter);
981
982 return nmp;
983 }
984
985 bool
986 vfs_unmount_forceone(struct lwp *l)
987 {
988 struct mount *mp;
989 int error;
990
991 mp = vfs_unmount_next(mountgen);
992 if (mp == NULL) {
993 return false;
994 }
995
996 #ifdef DEBUG
997 printf("forcefully unmounting %s (%s)...\n",
998 mp->mnt_stat.f_mntonname, mp->mnt_stat.f_mntfromname);
999 #endif
1000 if ((error = dounmount(mp, MNT_FORCE, l)) == 0) {
1001 vfs_unmount_print(mp, "forcefully ");
1002 return true;
1003 } else {
1004 vfs_rele(mp);
1005 }
1006
1007 #ifdef DEBUG
1008 printf("forceful unmount of %s failed with error %d\n",
1009 mp->mnt_stat.f_mntonname, error);
1010 #endif
1011
1012 return false;
1013 }
1014
1015 bool
1016 vfs_unmountall1(struct lwp *l, bool force, bool verbose)
1017 {
1018 struct mount *mp;
1019 mount_iterator_t *iter;
1020 bool any_error = false, progress = false;
1021 uint64_t gen;
1022 int error;
1023
1024 gen = mountgen;
1025 for (;;) {
1026 mp = vfs_unmount_next(gen);
1027 if (mp == NULL)
1028 break;
1029 gen = mp->mnt_gen;
1030
1031 #ifdef DEBUG
1032 printf("unmounting %p %s (%s)...\n",
1033 (void *)mp, mp->mnt_stat.f_mntonname,
1034 mp->mnt_stat.f_mntfromname);
1035 #endif
1036 if ((error = dounmount(mp, force ? MNT_FORCE : 0, l)) == 0) {
1037 vfs_unmount_print(mp, "");
1038 progress = true;
1039 } else {
1040 vfs_rele(mp);
1041 if (verbose) {
1042 printf("unmount of %s failed with error %d\n",
1043 mp->mnt_stat.f_mntonname, error);
1044 }
1045 any_error = true;
1046 }
1047 }
1048 if (verbose) {
1049 printf("unmounting done\n");
1050 }
1051 if (any_error && verbose) {
1052 printf("WARNING: some file systems would not unmount\n");
1053 }
1054
1055 /* If the mountlist is empty destroy anonymous device vnodes. */
1056 mountlist_iterator_init(&iter);
1057 if (mountlist_iterator_next(iter) == NULL) {
1058 struct vnode_iterator *marker;
1059 vnode_t *vp;
1060
1061 vfs_vnode_iterator_init(dead_rootmount, &marker);
1062 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
1063 if (vp->v_type == VCHR || vp->v_type == VBLK)
1064 vgone(vp);
1065 else
1066 vrele(vp);
1067 }
1068 vfs_vnode_iterator_destroy(marker);
1069 }
1070 mountlist_iterator_destroy(iter);
1071
1072 return progress;
1073 }
1074
1075 void
1076 vfs_sync_all(struct lwp *l)
1077 {
1078 printf("syncing disks... ");
1079
1080 /* remove user processes from run queue */
1081 suspendsched();
1082 (void)spl0();
1083
1084 /* avoid coming back this way again if we panic. */
1085 doing_shutdown = 1;
1086
1087 do_sys_sync(l);
1088
1089 /* Wait for sync to finish. */
1090 if (vfs_syncwait() != 0) {
1091 #if defined(DDB) && defined(DEBUG_HALT_BUSY)
1092 Debugger();
1093 #endif
1094 printf("giving up\n");
1095 return;
1096 } else
1097 printf("done\n");
1098 }
1099
1100 /*
1101 * Sync and unmount file systems before shutting down.
1102 */
1103 void
1104 vfs_shutdown(void)
1105 {
1106 lwp_t *l = curlwp;
1107
1108 vfs_sync_all(l);
1109
1110 /*
1111 * If we have paniced - do not make the situation potentially
1112 * worse by unmounting the file systems.
1113 */
1114 if (panicstr != NULL) {
1115 return;
1116 }
1117
1118 /* Unmount file systems. */
1119 vfs_unmountall(l);
1120 }
1121
1122 /*
1123 * Print a list of supported file system types (used by vfs_mountroot)
1124 */
1125 static void
1126 vfs_print_fstypes(void)
1127 {
1128 struct vfsops *v;
1129 int cnt = 0;
1130
1131 mutex_enter(&vfs_list_lock);
1132 LIST_FOREACH(v, &vfs_list, vfs_list)
1133 ++cnt;
1134 mutex_exit(&vfs_list_lock);
1135
1136 if (cnt == 0) {
1137 printf("WARNING: No file system modules have been loaded.\n");
1138 return;
1139 }
1140
1141 printf("Supported file systems:");
1142 mutex_enter(&vfs_list_lock);
1143 LIST_FOREACH(v, &vfs_list, vfs_list) {
1144 printf(" %s", v->vfs_name);
1145 }
1146 mutex_exit(&vfs_list_lock);
1147 printf("\n");
1148 }
1149
1150 /*
1151 * Mount the root file system. If the operator didn't specify a
1152 * file system to use, try all possible file systems until one
1153 * succeeds.
1154 */
1155 int
1156 vfs_mountroot(void)
1157 {
1158 struct vfsops *v;
1159 int error = ENODEV;
1160
1161 if (root_device == NULL)
1162 panic("vfs_mountroot: root device unknown");
1163
1164 switch (device_class(root_device)) {
1165 case DV_IFNET:
1166 if (rootdev != NODEV)
1167 panic("vfs_mountroot: rootdev set for DV_IFNET "
1168 "(0x%llx -> %llu,%llu)",
1169 (unsigned long long)rootdev,
1170 (unsigned long long)major(rootdev),
1171 (unsigned long long)minor(rootdev));
1172 break;
1173
1174 case DV_DISK:
1175 if (rootdev == NODEV)
1176 panic("vfs_mountroot: rootdev not set for DV_DISK");
1177 if (bdevvp(rootdev, &rootvp))
1178 panic("vfs_mountroot: can't get vnode for rootdev");
1179 error = VOP_OPEN(rootvp, FREAD, FSCRED);
1180 if (error) {
1181 printf("vfs_mountroot: can't open root device\n");
1182 return (error);
1183 }
1184 break;
1185
1186 case DV_VIRTUAL:
1187 break;
1188
1189 default:
1190 printf("%s: inappropriate for root file system\n",
1191 device_xname(root_device));
1192 return (ENODEV);
1193 }
1194
1195 /*
1196 * If user specified a root fs type, use it. Make sure the
1197 * specified type exists and has a mount_root()
1198 */
1199 if (strcmp(rootfstype, ROOT_FSTYPE_ANY) != 0) {
1200 v = vfs_getopsbyname(rootfstype);
1201 error = EFTYPE;
1202 if (v != NULL) {
1203 if (v->vfs_mountroot != NULL) {
1204 error = (v->vfs_mountroot)();
1205 }
1206 v->vfs_refcount--;
1207 }
1208 goto done;
1209 }
1210
1211 /*
1212 * Try each file system currently configured into the kernel.
1213 */
1214 mutex_enter(&vfs_list_lock);
1215 LIST_FOREACH(v, &vfs_list, vfs_list) {
1216 if (v->vfs_mountroot == NULL)
1217 continue;
1218 #ifdef DEBUG
1219 aprint_normal("mountroot: trying %s...\n", v->vfs_name);
1220 #endif
1221 v->vfs_refcount++;
1222 mutex_exit(&vfs_list_lock);
1223 error = (*v->vfs_mountroot)();
1224 mutex_enter(&vfs_list_lock);
1225 v->vfs_refcount--;
1226 if (!error) {
1227 aprint_normal("root file system type: %s\n",
1228 v->vfs_name);
1229 break;
1230 }
1231 }
1232 mutex_exit(&vfs_list_lock);
1233
1234 if (v == NULL) {
1235 vfs_print_fstypes();
1236 printf("no file system for %s", device_xname(root_device));
1237 if (device_class(root_device) == DV_DISK)
1238 printf(" (dev 0x%llx)", (unsigned long long)rootdev);
1239 printf("\n");
1240 error = EFTYPE;
1241 }
1242
1243 done:
1244 if (error && device_class(root_device) == DV_DISK) {
1245 VOP_CLOSE(rootvp, FREAD, FSCRED);
1246 vrele(rootvp);
1247 }
1248 if (error == 0) {
1249 mount_iterator_t *iter;
1250 struct mount *mp;
1251 extern struct cwdinfo cwdi0;
1252
1253 mountlist_iterator_init(&iter);
1254 mp = mountlist_iterator_next(iter);
1255 KASSERT(mp != NULL);
1256 mountlist_iterator_destroy(iter);
1257
1258 mp->mnt_flag |= MNT_ROOTFS;
1259 mp->mnt_op->vfs_refcount++;
1260
1261 /*
1262 * Get the vnode for '/'. Set cwdi0.cwdi_cdir to
1263 * reference it, and donate it the reference grabbed
1264 * with VFS_ROOT().
1265 */
1266 error = VFS_ROOT(mp, LK_NONE, &rootvnode);
1267 if (error)
1268 panic("cannot find root vnode, error=%d", error);
1269 cwdi0.cwdi_cdir = rootvnode;
1270 cwdi0.cwdi_rdir = NULL;
1271
1272 /*
1273 * Now that root is mounted, we can fixup initproc's CWD
1274 * info. All other processes are kthreads, which merely
1275 * share proc0's CWD info.
1276 */
1277 initproc->p_cwdi->cwdi_cdir = rootvnode;
1278 vref(initproc->p_cwdi->cwdi_cdir);
1279 initproc->p_cwdi->cwdi_rdir = NULL;
1280 /*
1281 * Enable loading of modules from the filesystem
1282 */
1283 module_load_vfs_init();
1284
1285 }
1286 return (error);
1287 }
1288
1289 /*
1290 * mount_specific_key_create --
1291 * Create a key for subsystem mount-specific data.
1292 */
1293 int
1294 mount_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1295 {
1296
1297 return specificdata_key_create(mount_specificdata_domain, keyp, dtor);
1298 }
1299
1300 /*
1301 * mount_specific_key_delete --
1302 * Delete a key for subsystem mount-specific data.
1303 */
1304 void
1305 mount_specific_key_delete(specificdata_key_t key)
1306 {
1307
1308 specificdata_key_delete(mount_specificdata_domain, key);
1309 }
1310
1311 /*
1312 * mount_initspecific --
1313 * Initialize a mount's specificdata container.
1314 */
1315 void
1316 mount_initspecific(struct mount *mp)
1317 {
1318 int error __diagused;
1319
1320 error = specificdata_init(mount_specificdata_domain,
1321 &mp->mnt_specdataref);
1322 KASSERT(error == 0);
1323 }
1324
1325 /*
1326 * mount_finispecific --
1327 * Finalize a mount's specificdata container.
1328 */
1329 void
1330 mount_finispecific(struct mount *mp)
1331 {
1332
1333 specificdata_fini(mount_specificdata_domain, &mp->mnt_specdataref);
1334 }
1335
1336 /*
1337 * mount_getspecific --
1338 * Return mount-specific data corresponding to the specified key.
1339 */
1340 void *
1341 mount_getspecific(struct mount *mp, specificdata_key_t key)
1342 {
1343
1344 return specificdata_getspecific(mount_specificdata_domain,
1345 &mp->mnt_specdataref, key);
1346 }
1347
1348 /*
1349 * mount_setspecific --
1350 * Set mount-specific data corresponding to the specified key.
1351 */
1352 void
1353 mount_setspecific(struct mount *mp, specificdata_key_t key, void *data)
1354 {
1355
1356 specificdata_setspecific(mount_specificdata_domain,
1357 &mp->mnt_specdataref, key, data);
1358 }
1359
1360 /*
1361 * Check to see if a filesystem is mounted on a block device.
1362 */
1363 int
1364 vfs_mountedon(vnode_t *vp)
1365 {
1366 vnode_t *vq;
1367 int error = 0;
1368
1369 if (vp->v_type != VBLK)
1370 return ENOTBLK;
1371 if (spec_node_getmountedfs(vp) != NULL)
1372 return EBUSY;
1373 if (spec_node_lookup_by_dev(vp->v_type, vp->v_rdev, &vq) == 0) {
1374 if (spec_node_getmountedfs(vq) != NULL)
1375 error = EBUSY;
1376 vrele(vq);
1377 }
1378
1379 return error;
1380 }
1381
1382 /*
1383 * Check if a device pointed to by vp is mounted.
1384 *
1385 * Returns:
1386 * EINVAL if it's not a disk
1387 * EBUSY if it's a disk and mounted
1388 * 0 if it's a disk and not mounted
1389 */
1390 int
1391 rawdev_mounted(vnode_t *vp, vnode_t **bvpp)
1392 {
1393 vnode_t *bvp;
1394 dev_t dev;
1395 int d_type;
1396
1397 bvp = NULL;
1398 d_type = D_OTHER;
1399
1400 if (iskmemvp(vp))
1401 return EINVAL;
1402
1403 switch (vp->v_type) {
1404 case VCHR: {
1405 const struct cdevsw *cdev;
1406
1407 dev = vp->v_rdev;
1408 cdev = cdevsw_lookup(dev);
1409 if (cdev != NULL) {
1410 dev_t blkdev;
1411
1412 blkdev = devsw_chr2blk(dev);
1413 if (blkdev != NODEV) {
1414 if (vfinddev(blkdev, VBLK, &bvp) != 0) {
1415 d_type = (cdev->d_flag & D_TYPEMASK);
1416 /* XXX: what if bvp disappears? */
1417 vrele(bvp);
1418 }
1419 }
1420 }
1421
1422 break;
1423 }
1424
1425 case VBLK: {
1426 const struct bdevsw *bdev;
1427
1428 dev = vp->v_rdev;
1429 bdev = bdevsw_lookup(dev);
1430 if (bdev != NULL)
1431 d_type = (bdev->d_flag & D_TYPEMASK);
1432
1433 bvp = vp;
1434
1435 break;
1436 }
1437
1438 default:
1439 break;
1440 }
1441
1442 if (d_type != D_DISK)
1443 return EINVAL;
1444
1445 if (bvpp != NULL)
1446 *bvpp = bvp;
1447
1448 /*
1449 * XXX: This is bogus. We should be failing the request
1450 * XXX: not only if this specific slice is mounted, but
1451 * XXX: if it's on a disk with any other mounted slice.
1452 */
1453 if (vfs_mountedon(bvp))
1454 return EBUSY;
1455
1456 return 0;
1457 }
1458
1459 /*
1460 * Make a 'unique' number from a mount type name.
1461 */
1462 long
1463 makefstype(const char *type)
1464 {
1465 long rv;
1466
1467 for (rv = 0; *type; type++) {
1468 rv <<= 2;
1469 rv ^= *type;
1470 }
1471 return rv;
1472 }
1473
1474 static struct mountlist_entry *
1475 mountlist_alloc(enum mountlist_type type, struct mount *mp)
1476 {
1477 struct mountlist_entry *me;
1478
1479 me = kmem_zalloc(sizeof(*me), KM_SLEEP);
1480 me->me_mount = mp;
1481 me->me_type = type;
1482
1483 return me;
1484 }
1485
1486 static void
1487 mountlist_free(struct mountlist_entry *me)
1488 {
1489
1490 kmem_free(me, sizeof(*me));
1491 }
1492
1493 void
1494 mountlist_iterator_init(mount_iterator_t **mip)
1495 {
1496 struct mountlist_entry *me;
1497
1498 me = mountlist_alloc(ME_MARKER, NULL);
1499 mutex_enter(&mountlist_lock);
1500 TAILQ_INSERT_HEAD(&mountlist, me, me_list);
1501 mutex_exit(&mountlist_lock);
1502 *mip = (mount_iterator_t *)me;
1503 }
1504
1505 void
1506 mountlist_iterator_destroy(mount_iterator_t *mi)
1507 {
1508 struct mountlist_entry *marker = &mi->mi_entry;
1509
1510 if (marker->me_mount != NULL)
1511 vfs_unbusy(marker->me_mount);
1512
1513 mutex_enter(&mountlist_lock);
1514 TAILQ_REMOVE(&mountlist, marker, me_list);
1515 mutex_exit(&mountlist_lock);
1516
1517 mountlist_free(marker);
1518
1519 }
1520
1521 /*
1522 * Return the next mount or NULL for this iterator.
1523 * Mark it busy on success.
1524 */
1525 static inline struct mount *
1526 _mountlist_iterator_next(mount_iterator_t *mi, bool wait)
1527 {
1528 struct mountlist_entry *me, *marker = &mi->mi_entry;
1529 struct mount *mp;
1530 int error;
1531
1532 if (marker->me_mount != NULL) {
1533 vfs_unbusy(marker->me_mount);
1534 marker->me_mount = NULL;
1535 }
1536
1537 mutex_enter(&mountlist_lock);
1538 for (;;) {
1539 KASSERT(marker->me_type == ME_MARKER);
1540
1541 me = TAILQ_NEXT(marker, me_list);
1542 if (me == NULL) {
1543 /* End of list: keep marker and return. */
1544 mutex_exit(&mountlist_lock);
1545 return NULL;
1546 }
1547 TAILQ_REMOVE(&mountlist, marker, me_list);
1548 TAILQ_INSERT_AFTER(&mountlist, me, marker, me_list);
1549
1550 /* Skip other markers. */
1551 if (me->me_type != ME_MOUNT)
1552 continue;
1553
1554 /* Take an initial reference for vfs_busy() below. */
1555 mp = me->me_mount;
1556 KASSERT(mp != NULL);
1557 vfs_ref(mp);
1558 mutex_exit(&mountlist_lock);
1559
1560 /* Try to mark this mount busy and return on success. */
1561 if (wait)
1562 error = vfs_busy(mp);
1563 else
1564 error = vfs_trybusy(mp);
1565 if (error == 0) {
1566 vfs_rele(mp);
1567 marker->me_mount = mp;
1568 return mp;
1569 }
1570 vfs_rele(mp);
1571 mutex_enter(&mountlist_lock);
1572 }
1573 }
1574
1575 struct mount *
1576 mountlist_iterator_next(mount_iterator_t *mi)
1577 {
1578
1579 return _mountlist_iterator_next(mi, true);
1580 }
1581
1582 struct mount *
1583 mountlist_iterator_trynext(mount_iterator_t *mi)
1584 {
1585
1586 return _mountlist_iterator_next(mi, false);
1587 }
1588
1589 /*
1590 * Attach new mount to the end of the mount list.
1591 */
1592 void
1593 mountlist_append(struct mount *mp)
1594 {
1595 struct mountlist_entry *me;
1596
1597 me = mountlist_alloc(ME_MOUNT, mp);
1598 mutex_enter(&mountlist_lock);
1599 TAILQ_INSERT_TAIL(&mountlist, me, me_list);
1600 mutex_exit(&mountlist_lock);
1601 }
1602
1603 /*
1604 * Remove mount from mount list.
1605 */void
1606 mountlist_remove(struct mount *mp)
1607 {
1608 struct mountlist_entry *me;
1609
1610 mutex_enter(&mountlist_lock);
1611 TAILQ_FOREACH(me, &mountlist, me_list)
1612 if (me->me_type == ME_MOUNT && me->me_mount == mp)
1613 break;
1614 KASSERT(me != NULL);
1615 TAILQ_REMOVE(&mountlist, me, me_list);
1616 mutex_exit(&mountlist_lock);
1617 mountlist_free(me);
1618 }
1619
1620 /*
1621 * Unlocked variant to traverse the mountlist.
1622 * To be used from DDB only.
1623 */
1624 struct mount *
1625 _mountlist_next(struct mount *mp)
1626 {
1627 struct mountlist_entry *me;
1628
1629 if (mp == NULL) {
1630 me = TAILQ_FIRST(&mountlist);
1631 } else {
1632 TAILQ_FOREACH(me, &mountlist, me_list)
1633 if (me->me_type == ME_MOUNT && me->me_mount == mp)
1634 break;
1635 if (me != NULL)
1636 me = TAILQ_NEXT(me, me_list);
1637 }
1638
1639 while (me != NULL && me->me_type != ME_MOUNT)
1640 me = TAILQ_NEXT(me, me_list);
1641
1642 return (me ? me->me_mount : NULL);
1643 }
1644