2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2000,2004
5 * Poul-Henning Kamp. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Neither the name of the University nor the names of its contributors
13 * may be used to endorse or promote products derived from this software
14 * without specific prior written permission.
16 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * From: FreeBSD: src/sys/miscfs/kernfs/kernfs_vfsops.c 1.36
33 #include <sys/param.h>
34 #include <sys/systm.h>
36 #include <sys/dirent.h>
37 #include <sys/kernel.h>
38 #include <sys/limits.h>
40 #include <sys/malloc.h>
43 #include <sys/sysctl.h>
44 #include <sys/vnode.h>
48 #include <fs/devfs/devfs.h>
49 #include <fs/devfs/devfs_int.h>
51 #include <security/mac/mac_framework.h>
54 * The one true (but secret) list of active devices in the system.
55 * Locked by dev_lock()/devmtx
57 struct cdev_priv_list cdevp_list = TAILQ_HEAD_INITIALIZER(cdevp_list);
59 struct unrhdr *devfs_inos;
62 static MALLOC_DEFINE(M_DEVFS2, "DEVFS2", "DEVFS data 2");
63 static MALLOC_DEFINE(M_DEVFS3, "DEVFS3", "DEVFS data 3");
64 static MALLOC_DEFINE(M_CDEVP, "DEVFS1", "DEVFS cdev_priv storage");
66 SYSCTL_NODE(_vfs, OID_AUTO, devfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
69 static unsigned devfs_generation;
70 SYSCTL_UINT(_vfs_devfs, OID_AUTO, generation, CTLFLAG_RD,
71 &devfs_generation, 0, "DEVFS generation number");
73 unsigned devfs_rule_depth = 1;
74 SYSCTL_UINT(_vfs_devfs, OID_AUTO, rule_depth, CTLFLAG_RW,
75 &devfs_rule_depth, 0, "Max depth of ruleset include");
78 * Helper sysctl for devname(3). We're given a dev_t and return the
79 * name, if any, registered by the device driver.
82 sysctl_devname(SYSCTL_HANDLER_ARGS)
86 #ifdef COMPAT_FREEBSD11
89 struct cdev_priv *cdp;
92 #ifdef COMPAT_FREEBSD11
93 if (req->newlen == sizeof(ud_compat)) {
94 error = SYSCTL_IN(req, &ud_compat, sizeof(ud_compat));
96 ud = ud_compat == (uint32_t)NODEV ? NODEV : ud_compat;
99 error = SYSCTL_IN(req, &ud, sizeof (ud));
106 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list)
107 if (cdp->cdp_inode == ud) {
115 error = SYSCTL_OUT(req, dev->si_name, strlen(dev->si_name) + 1);
120 SYSCTL_PROC(_kern, OID_AUTO, devname,
121 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
122 NULL, 0, sysctl_devname, "", "devname(3) handler");
124 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev, CTLFLAG_RD,
125 SYSCTL_NULL_INT_PTR, sizeof(struct cdev), "sizeof(struct cdev)");
127 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev_priv, CTLFLAG_RD,
128 SYSCTL_NULL_INT_PTR, sizeof(struct cdev_priv), "sizeof(struct cdev_priv)");
131 devfs_alloc(int flags)
133 struct cdev_priv *cdp;
137 cdp = malloc(sizeof *cdp, M_CDEVP, M_ZERO |
138 ((flags & MAKEDEV_NOWAIT) ? M_NOWAIT : M_WAITOK));
142 mtx_init(&cdp->cdp_threadlock, "devthrd", NULL, MTX_DEF);
144 cdp->cdp_dirents = &cdp->cdp_dirent0;
147 LIST_INIT(&cdev->si_children);
149 cdev->si_atime = cdev->si_mtime = cdev->si_ctime = ts;
155 devfs_dev_exists(const char *name)
157 struct cdev_priv *cdp;
159 mtx_assert(&devmtx, MA_OWNED);
161 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
162 if ((cdp->cdp_flags & CDP_ACTIVE) == 0)
164 if (devfs_pathpath(cdp->cdp_c.si_name, name) != 0)
166 if (devfs_pathpath(name, cdp->cdp_c.si_name) != 0)
169 if (devfs_dir_find(name) != 0)
176 devfs_free(struct cdev *cdev)
178 struct cdev_priv *cdp;
180 cdp = cdev2priv(cdev);
181 if (cdev->si_cred != NULL)
182 crfree(cdev->si_cred);
183 devfs_free_cdp_inode(cdp->cdp_inode);
184 if (cdp->cdp_maxdirent > 0)
185 free(cdp->cdp_dirents, M_DEVFS2);
186 mtx_destroy(&cdp->cdp_threadlock);
190 struct devfs_dirent *
191 devfs_find(struct devfs_dirent *dd, const char *name, int namelen, int type)
193 struct devfs_dirent *de;
195 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
196 if (namelen != de->de_dirent->d_namlen)
198 if (type != 0 && type != de->de_dirent->d_type)
202 * The race with finding non-active name is not
203 * completely closed by the check, but it is similar
204 * to the devfs_allocv() in making it unlikely enough.
206 if (de->de_dirent->d_type == DT_CHR &&
207 (de->de_cdp->cdp_flags & CDP_ACTIVE) == 0)
210 if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
214 KASSERT(de == NULL || (de->de_flags & DE_DOOMED) == 0,
215 ("devfs_find: returning a doomed entry"));
219 struct devfs_dirent *
220 devfs_newdirent(char *name, int namelen)
223 struct devfs_dirent *de;
226 d.d_namlen = namelen;
227 i = sizeof(*de) + GENERIC_DIRSIZ(&d);
228 de = malloc(i, M_DEVFS3, M_WAITOK | M_ZERO);
229 de->de_dirent = (struct dirent *)(de + 1);
230 de->de_dirent->d_namlen = namelen;
231 de->de_dirent->d_reclen = GENERIC_DIRSIZ(&d);
232 bcopy(name, de->de_dirent->d_name, namelen);
233 dirent_terminate(de->de_dirent);
234 vfs_timestamp(&de->de_ctime);
235 de->de_mtime = de->de_atime = de->de_ctime;
244 struct devfs_dirent *
245 devfs_parent_dirent(struct devfs_dirent *de)
248 if (de->de_dirent->d_type != DT_DIR)
251 if (de->de_flags & (DE_DOT | DE_DOTDOT))
254 de = TAILQ_FIRST(&de->de_dlist); /* "." */
257 de = TAILQ_NEXT(de, de_list); /* ".." */
264 struct devfs_dirent *
265 devfs_vmkdir(struct devfs_mount *dmp, char *name, int namelen,
266 struct devfs_dirent *dotdot, u_int inode)
268 struct devfs_dirent *dd;
269 struct devfs_dirent *de;
271 /* Create the new directory */
272 dd = devfs_newdirent(name, namelen);
273 TAILQ_INIT(&dd->de_dlist);
274 dd->de_dirent->d_type = DT_DIR;
279 dd->de_inode = inode;
281 dd->de_inode = alloc_unr(devfs_inos);
284 * "." and ".." are always the two first entries in the
287 * Create the "." entry in the new directory.
289 de = devfs_newdirent(".", 1);
290 de->de_dirent->d_type = DT_DIR;
291 de->de_flags |= DE_DOT;
292 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
295 /* Create the ".." entry in the new directory. */
296 de = devfs_newdirent("..", 2);
297 de->de_dirent->d_type = DT_DIR;
298 de->de_flags |= DE_DOTDOT;
299 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
300 if (dotdot == NULL) {
304 sx_assert(&dmp->dm_lock, SX_XLOCKED);
305 TAILQ_INSERT_TAIL(&dotdot->de_dlist, dd, de_list);
307 devfs_rules_apply(dmp, dd);
311 mac_devfs_create_directory(dmp->dm_mount, name, namelen, dd);
317 devfs_dirent_free(struct devfs_dirent *de)
322 mtx_lock(&devfs_de_interlock);
323 if (vp != NULL && vp->v_data == de)
325 mtx_unlock(&devfs_de_interlock);
330 * Removes a directory if it is empty. Also empty parent directories are
331 * removed recursively.
334 devfs_rmdir_empty(struct devfs_mount *dm, struct devfs_dirent *de)
336 struct devfs_dirent *dd, *de_dot, *de_dotdot;
338 sx_assert(&dm->dm_lock, SX_XLOCKED);
341 KASSERT(de->de_dirent->d_type == DT_DIR,
342 ("devfs_rmdir_empty: de is not a directory"));
344 if ((de->de_flags & DE_DOOMED) != 0 || de == dm->dm_rootdir)
347 de_dot = TAILQ_FIRST(&de->de_dlist);
348 KASSERT(de_dot != NULL, ("devfs_rmdir_empty: . missing"));
349 de_dotdot = TAILQ_NEXT(de_dot, de_list);
350 KASSERT(de_dotdot != NULL, ("devfs_rmdir_empty: .. missing"));
351 /* Return if the directory is not empty. */
352 if (TAILQ_NEXT(de_dotdot, de_list) != NULL)
355 dd = devfs_parent_dirent(de);
356 KASSERT(dd != NULL, ("devfs_rmdir_empty: NULL dd"));
357 TAILQ_REMOVE(&de->de_dlist, de_dot, de_list);
358 TAILQ_REMOVE(&de->de_dlist, de_dotdot, de_list);
359 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
361 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
362 devfs_delete(dm, de_dot, DEVFS_DEL_NORECURSE);
363 devfs_delete(dm, de_dotdot, DEVFS_DEL_NORECURSE);
364 if (DEVFS_DE_DROP(dd)) {
365 devfs_dirent_free(dd);
374 * The caller needs to hold the dm for the duration of the call since
375 * dm->dm_lock may be temporary dropped.
378 devfs_delete(struct devfs_mount *dm, struct devfs_dirent *de, int flags)
380 struct devfs_dirent *dd;
383 KASSERT((de->de_flags & DE_DOOMED) == 0,
384 ("devfs_delete doomed dirent"));
385 de->de_flags |= DE_DOOMED;
387 if ((flags & DEVFS_DEL_NORECURSE) == 0) {
388 dd = devfs_parent_dirent(de);
391 if (de->de_flags & DE_USER) {
392 KASSERT(dd != NULL, ("devfs_delete: NULL dd"));
393 devfs_dir_unref_de(dm, dd);
398 mtx_lock(&devfs_de_interlock);
402 mtx_unlock(&devfs_de_interlock);
404 sx_unlock(&dm->dm_lock);
405 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
406 vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
410 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
413 sx_xlock(&dm->dm_lock);
415 mtx_unlock(&devfs_de_interlock);
416 if (de->de_symlink) {
417 free(de->de_symlink, M_DEVFS);
418 de->de_symlink = NULL;
421 mac_devfs_destroy(de);
423 if (de->de_inode > DEVFS_ROOTINO) {
424 devfs_free_cdp_inode(de->de_inode);
427 if (DEVFS_DE_DROP(de))
428 devfs_dirent_free(de);
431 if (DEVFS_DE_DROP(dd))
432 devfs_dirent_free(dd);
434 devfs_rmdir_empty(dm, dd);
440 * Recursively removes the entire tree.
441 * The caller needs to hold the dm for the duration of the call.
445 devfs_purge(struct devfs_mount *dm, struct devfs_dirent *dd)
447 struct devfs_dirent *de;
449 sx_assert(&dm->dm_lock, SX_XLOCKED);
454 * Use TAILQ_LAST() to remove "." and ".." last.
455 * We might need ".." to resolve a path in
456 * devfs_dir_unref_de().
458 de = TAILQ_LAST(&dd->de_dlist, devfs_dlist_head);
461 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
462 if (de->de_flags & DE_USER)
463 devfs_dir_unref_de(dm, dd);
464 if (de->de_flags & (DE_DOT | DE_DOTDOT))
465 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
466 else if (de->de_dirent->d_type == DT_DIR)
469 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
471 if (DEVFS_DE_DROP(dd))
472 devfs_dirent_free(dd);
473 else if ((dd->de_flags & DE_DOOMED) == 0)
474 devfs_delete(dm, dd, DEVFS_DEL_NORECURSE);
478 * Each cdev_priv has an array of pointers to devfs_dirent which is indexed
479 * by the mount points dm_idx.
480 * This function extends the array when necessary, taking into account that
481 * the default array is 1 element and not malloc'ed.
484 devfs_metoo(struct cdev_priv *cdp, struct devfs_mount *dm)
486 struct devfs_dirent **dep;
489 siz = (dm->dm_idx + 1) * sizeof *dep;
490 dep = malloc(siz, M_DEVFS2, M_WAITOK | M_ZERO);
492 if (dm->dm_idx <= cdp->cdp_maxdirent) {
498 memcpy(dep, cdp->cdp_dirents, (cdp->cdp_maxdirent + 1) * sizeof *dep);
499 if (cdp->cdp_maxdirent > 0)
500 free(cdp->cdp_dirents, M_DEVFS2);
501 cdp->cdp_dirents = dep;
503 * XXX: if malloc told us how much we actually got this could
506 cdp->cdp_maxdirent = dm->dm_idx;
511 * The caller needs to hold the dm for the duration of the call.
514 devfs_populate_loop(struct devfs_mount *dm, int cleanup)
516 struct cdev_priv *cdp;
517 struct devfs_dirent *de;
518 struct devfs_dirent *dd, *dt;
520 int de_flags, depth, j;
523 sx_assert(&dm->dm_lock, SX_XLOCKED);
525 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
527 KASSERT(cdp->cdp_dirents != NULL, ("NULL cdp_dirents"));
530 * If we are unmounting, or the device has been destroyed,
531 * clean up our dirent.
533 if ((cleanup || !(cdp->cdp_flags & CDP_ACTIVE)) &&
534 dm->dm_idx <= cdp->cdp_maxdirent &&
535 cdp->cdp_dirents[dm->dm_idx] != NULL) {
536 de = cdp->cdp_dirents[dm->dm_idx];
537 cdp->cdp_dirents[dm->dm_idx] = NULL;
538 KASSERT(cdp == de->de_cdp,
539 ("%s %d %s %p %p", __func__, __LINE__,
540 cdp->cdp_c.si_name, cdp, de->de_cdp));
541 KASSERT(de->de_dir != NULL, ("Null de->de_dir"));
544 TAILQ_REMOVE(&de->de_dir->de_dlist, de, de_list);
547 devfs_delete(dm, de, 0);
554 * GC any lingering devices
556 if (!(cdp->cdp_flags & CDP_ACTIVE)) {
557 if (cdp->cdp_inuse > 0)
559 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
561 dev_rel(&cdp->cdp_c);
565 * Don't create any new dirents if we are unmounting
569 KASSERT((cdp->cdp_flags & CDP_ACTIVE), ("Bogons, I tell ya'!"));
571 if (dm->dm_idx <= cdp->cdp_maxdirent &&
572 cdp->cdp_dirents[dm->dm_idx] != NULL) {
573 de = cdp->cdp_dirents[dm->dm_idx];
574 KASSERT(cdp == de->de_cdp, ("inconsistent cdp"));
582 if (dm->dm_idx > cdp->cdp_maxdirent)
583 devfs_metoo(cdp, dm);
586 s = cdp->cdp_c.si_name;
588 for (q = s; *q != '/' && *q != '\0'; q++)
592 de = devfs_find(dd, s, q - s, 0);
594 de = devfs_vmkdir(dm, s, q - s, dd, 0);
595 else if (de->de_dirent->d_type == DT_LNK) {
596 de = devfs_find(dd, s, q - s, DT_DIR);
598 de = devfs_vmkdir(dm, s, q - s, dd, 0);
599 de->de_flags |= DE_COVERED;
603 KASSERT(dd->de_dirent->d_type == DT_DIR &&
604 (dd->de_flags & (DE_DOT | DE_DOTDOT)) == 0,
605 ("%s: invalid directory (si_name=%s)",
606 __func__, cdp->cdp_c.si_name));
610 de = devfs_find(dd, s, q - s, DT_LNK);
612 de_flags |= DE_COVERED;
614 de = devfs_newdirent(s, q - s);
615 if (cdp->cdp_c.si_flags & SI_ALIAS) {
619 de->de_dirent->d_type = DT_LNK;
620 pdev = cdp->cdp_c.si_parent;
623 while (dt != dm->dm_rootdir &&
624 (dt = devfs_parent_dirent(dt)) != NULL)
626 j = depth * 3 + strlen(pdev->si_name) + 1;
627 de->de_symlink = malloc(j, M_DEVFS, M_WAITOK);
628 de->de_symlink[0] = 0;
630 strcat(de->de_symlink, "../");
631 strcat(de->de_symlink, pdev->si_name);
633 de->de_uid = cdp->cdp_c.si_uid;
634 de->de_gid = cdp->cdp_c.si_gid;
635 de->de_mode = cdp->cdp_c.si_mode;
636 de->de_dirent->d_type = DT_CHR;
638 de->de_flags |= de_flags;
639 de->de_inode = cdp->cdp_inode;
642 mac_devfs_create_device(cdp->cdp_c.si_cred, dm->dm_mount,
646 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
647 devfs_rules_apply(dm, de);
649 /* XXX: could check that cdp is still active here */
650 KASSERT(cdp->cdp_dirents[dm->dm_idx] == NULL,
651 ("%s %d\n", __func__, __LINE__));
652 cdp->cdp_dirents[dm->dm_idx] = de;
653 KASSERT(de->de_cdp != (void *)0xdeadc0de,
654 ("%s %d\n", __func__, __LINE__));
663 * The caller needs to hold the dm for the duration of the call.
666 devfs_populate(struct devfs_mount *dm)
670 sx_assert(&dm->dm_lock, SX_XLOCKED);
671 gen = devfs_generation;
672 if (dm->dm_generation == gen)
674 while (devfs_populate_loop(dm, 0))
676 dm->dm_generation = gen;
680 * The caller needs to hold the dm for the duration of the call.
683 devfs_cleanup(struct devfs_mount *dm)
686 sx_assert(&dm->dm_lock, SX_XLOCKED);
687 while (devfs_populate_loop(dm, 1))
689 devfs_purge(dm, dm->dm_rootdir);
693 * devfs_create() and devfs_destroy() are called from kern_conf.c and
694 * in both cases the devlock() mutex is held, so no further locking
695 * is necessary and no sleeping allowed.
699 devfs_create(struct cdev *dev)
701 struct cdev_priv *cdp;
703 mtx_assert(&devmtx, MA_OWNED);
704 cdp = cdev2priv(dev);
705 cdp->cdp_flags |= CDP_ACTIVE;
706 cdp->cdp_inode = alloc_unrl(devfs_inos);
708 TAILQ_INSERT_TAIL(&cdevp_list, cdp, cdp_list);
713 devfs_destroy(struct cdev *dev)
715 struct cdev_priv *cdp;
717 mtx_assert(&devmtx, MA_OWNED);
718 cdp = cdev2priv(dev);
719 cdp->cdp_flags &= ~CDP_ACTIVE;
724 devfs_alloc_cdp_inode(void)
727 return (alloc_unr(devfs_inos));
731 devfs_free_cdp_inode(ino_t ino)
735 free_unr(devfs_inos, ino);
739 devfs_devs_init(void *junk __unused)
742 devfs_inos = new_unrhdr(DEVFS_ROOTINO + 1, INT_MAX, &devmtx);
745 SYSINIT(devfs_devs, SI_SUB_DEVFS, SI_ORDER_FIRST, devfs_devs_init, NULL);