2 * Copyright (c) 2000,2004
3 * Poul-Henning Kamp. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Neither the name of the University nor the names of its contributors
11 * may be used to endorse or promote products derived from this software
12 * without specific prior written permission.
14 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * From: FreeBSD: src/sys/miscfs/kernfs/kernfs_vfsops.c 1.36
31 #include <sys/param.h>
32 #include <sys/systm.h>
34 #include <sys/dirent.h>
35 #include <sys/kernel.h>
36 #include <sys/limits.h>
38 #include <sys/malloc.h>
41 #include <sys/sysctl.h>
42 #include <sys/vnode.h>
46 #include <fs/devfs/devfs.h>
47 #include <fs/devfs/devfs_int.h>
49 #include <security/mac/mac_framework.h>
52 * The one true (but secret) list of active devices in the system.
53 * Locked by dev_lock()/devmtx
55 struct cdev_priv_list cdevp_list = TAILQ_HEAD_INITIALIZER(cdevp_list);
57 struct unrhdr *devfs_inos;
60 static MALLOC_DEFINE(M_DEVFS2, "DEVFS2", "DEVFS data 2");
61 static MALLOC_DEFINE(M_DEVFS3, "DEVFS3", "DEVFS data 3");
62 static MALLOC_DEFINE(M_CDEVP, "DEVFS1", "DEVFS cdev_priv storage");
64 static SYSCTL_NODE(_vfs, OID_AUTO, devfs, CTLFLAG_RW, 0, "DEVFS filesystem");
66 static unsigned devfs_generation;
67 SYSCTL_UINT(_vfs_devfs, OID_AUTO, generation, CTLFLAG_RD,
68 &devfs_generation, 0, "DEVFS generation number");
70 unsigned devfs_rule_depth = 1;
71 SYSCTL_UINT(_vfs_devfs, OID_AUTO, rule_depth, CTLFLAG_RW,
72 &devfs_rule_depth, 0, "Max depth of ruleset include");
75 * Helper sysctl for devname(3). We're given a dev_t and return the
76 * name, if any, registered by the device driver.
79 sysctl_devname(SYSCTL_HANDLER_ARGS)
83 struct cdev_priv *cdp;
86 error = SYSCTL_IN(req, &ud, sizeof (ud));
93 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list)
94 if (cdp->cdp_inode == ud) {
102 error = SYSCTL_OUT(req, dev->si_name, strlen(dev->si_name) + 1);
107 SYSCTL_PROC(_kern, OID_AUTO, devname,
108 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
109 NULL, 0, sysctl_devname, "", "devname(3) handler");
111 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev, CTLFLAG_RD,
112 0, sizeof(struct cdev), "sizeof(struct cdev)");
114 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev_priv, CTLFLAG_RD,
115 0, sizeof(struct cdev_priv), "sizeof(struct cdev_priv)");
118 devfs_alloc(int flags)
120 struct cdev_priv *cdp;
124 cdp = malloc(sizeof *cdp, M_CDEVP, M_ZERO |
125 ((flags & MAKEDEV_NOWAIT) ? M_NOWAIT : M_WAITOK));
129 cdp->cdp_dirents = &cdp->cdp_dirent0;
130 cdp->cdp_dirent0 = NULL;
131 cdp->cdp_maxdirent = 0;
136 LIST_INIT(&cdev->si_children);
138 cdev->si_atime = cdev->si_mtime = cdev->si_ctime = ts;
139 cdev->si_cred = NULL;
145 devfs_dev_exists(const char *name)
147 struct cdev_priv *cdp;
149 mtx_assert(&devmtx, MA_OWNED);
151 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
152 if ((cdp->cdp_flags & CDP_ACTIVE) == 0)
154 if (devfs_pathpath(cdp->cdp_c.si_name, name) != 0)
156 if (devfs_pathpath(name, cdp->cdp_c.si_name) != 0)
159 if (devfs_dir_find(name) != 0)
166 devfs_free(struct cdev *cdev)
168 struct cdev_priv *cdp;
170 cdp = cdev2priv(cdev);
171 if (cdev->si_cred != NULL)
172 crfree(cdev->si_cred);
173 devfs_free_cdp_inode(cdp->cdp_inode);
174 if (cdp->cdp_maxdirent > 0)
175 free(cdp->cdp_dirents, M_DEVFS2);
179 struct devfs_dirent *
180 devfs_find(struct devfs_dirent *dd, const char *name, int namelen, int type)
182 struct devfs_dirent *de;
184 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
185 if (namelen != de->de_dirent->d_namlen)
187 if (type != 0 && type != de->de_dirent->d_type)
189 if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
193 KASSERT(de == NULL || (de->de_flags & DE_DOOMED) == 0,
194 ("devfs_find: returning a doomed entry"));
198 struct devfs_dirent *
199 devfs_newdirent(char *name, int namelen)
202 struct devfs_dirent *de;
205 d.d_namlen = namelen;
206 i = sizeof (*de) + GENERIC_DIRSIZ(&d);
207 de = malloc(i, M_DEVFS3, M_WAITOK | M_ZERO);
208 de->de_dirent = (struct dirent *)(de + 1);
209 de->de_dirent->d_namlen = namelen;
210 de->de_dirent->d_reclen = GENERIC_DIRSIZ(&d);
211 bcopy(name, de->de_dirent->d_name, namelen);
212 de->de_dirent->d_name[namelen] = '\0';
213 vfs_timestamp(&de->de_ctime);
214 de->de_mtime = de->de_atime = de->de_ctime;
223 struct devfs_dirent *
224 devfs_parent_dirent(struct devfs_dirent *de)
227 if (de->de_dirent->d_type != DT_DIR)
230 if (de->de_flags & (DE_DOT | DE_DOTDOT))
233 de = TAILQ_FIRST(&de->de_dlist); /* "." */
236 de = TAILQ_NEXT(de, de_list); /* ".." */
243 struct devfs_dirent *
244 devfs_vmkdir(struct devfs_mount *dmp, char *name, int namelen, struct devfs_dirent *dotdot, u_int inode)
246 struct devfs_dirent *dd;
247 struct devfs_dirent *de;
249 /* Create the new directory */
250 dd = devfs_newdirent(name, namelen);
251 TAILQ_INIT(&dd->de_dlist);
252 dd->de_dirent->d_type = DT_DIR;
257 dd->de_inode = inode;
259 dd->de_inode = alloc_unr(devfs_inos);
262 * "." and ".." are always the two first entries in the
265 * Create the "." entry in the new directory.
267 de = devfs_newdirent(".", 1);
268 de->de_dirent->d_type = DT_DIR;
269 de->de_flags |= DE_DOT;
270 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
273 /* Create the ".." entry in the new directory. */
274 de = devfs_newdirent("..", 2);
275 de->de_dirent->d_type = DT_DIR;
276 de->de_flags |= DE_DOTDOT;
277 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
278 if (dotdot == NULL) {
282 sx_assert(&dmp->dm_lock, SX_XLOCKED);
283 TAILQ_INSERT_TAIL(&dotdot->de_dlist, dd, de_list);
285 devfs_rules_apply(dmp, dd);
289 mac_devfs_create_directory(dmp->dm_mount, name, namelen, dd);
295 devfs_dirent_free(struct devfs_dirent *de)
301 * Removes a directory if it is empty. Also empty parent directories are
302 * removed recursively.
305 devfs_rmdir_empty(struct devfs_mount *dm, struct devfs_dirent *de)
307 struct devfs_dirent *dd, *de_dot, *de_dotdot;
309 sx_assert(&dm->dm_lock, SX_XLOCKED);
312 KASSERT(de->de_dirent->d_type == DT_DIR,
313 ("devfs_rmdir_empty: de is not a directory"));
315 if ((de->de_flags & DE_DOOMED) != 0 || de == dm->dm_rootdir)
318 de_dot = TAILQ_FIRST(&de->de_dlist);
319 KASSERT(de_dot != NULL, ("devfs_rmdir_empty: . missing"));
320 de_dotdot = TAILQ_NEXT(de_dot, de_list);
321 KASSERT(de_dotdot != NULL, ("devfs_rmdir_empty: .. missing"));
322 /* Return if the directory is not empty. */
323 if (TAILQ_NEXT(de_dotdot, de_list) != NULL)
326 dd = devfs_parent_dirent(de);
327 KASSERT(dd != NULL, ("devfs_rmdir_empty: NULL dd"));
328 TAILQ_REMOVE(&de->de_dlist, de_dot, de_list);
329 TAILQ_REMOVE(&de->de_dlist, de_dotdot, de_list);
330 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
332 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
333 devfs_delete(dm, de_dot, DEVFS_DEL_NORECURSE);
334 devfs_delete(dm, de_dotdot, DEVFS_DEL_NORECURSE);
335 if (DEVFS_DE_DROP(dd)) {
336 devfs_dirent_free(dd);
345 * The caller needs to hold the dm for the duration of the call since
346 * dm->dm_lock may be temporary dropped.
349 devfs_delete(struct devfs_mount *dm, struct devfs_dirent *de, int flags)
351 struct devfs_dirent *dd;
354 KASSERT((de->de_flags & DE_DOOMED) == 0,
355 ("devfs_delete doomed dirent"));
356 de->de_flags |= DE_DOOMED;
358 if ((flags & DEVFS_DEL_NORECURSE) == 0) {
359 dd = devfs_parent_dirent(de);
362 if (de->de_flags & DE_USER) {
363 KASSERT(dd != NULL, ("devfs_delete: NULL dd"));
364 devfs_dir_unref_de(dm, dd);
369 mtx_lock(&devfs_de_interlock);
373 mtx_unlock(&devfs_de_interlock);
375 sx_unlock(&dm->dm_lock);
376 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
377 vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
381 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
384 sx_xlock(&dm->dm_lock);
386 mtx_unlock(&devfs_de_interlock);
387 if (de->de_symlink) {
388 free(de->de_symlink, M_DEVFS);
389 de->de_symlink = NULL;
392 mac_devfs_destroy(de);
394 if (de->de_inode > DEVFS_ROOTINO) {
395 devfs_free_cdp_inode(de->de_inode);
398 if (DEVFS_DE_DROP(de))
399 devfs_dirent_free(de);
402 if (DEVFS_DE_DROP(dd))
403 devfs_dirent_free(dd);
405 devfs_rmdir_empty(dm, dd);
411 * Recursively removes the entire tree.
412 * The caller needs to hold the dm for the duration of the call.
416 devfs_purge(struct devfs_mount *dm, struct devfs_dirent *dd)
418 struct devfs_dirent *de;
420 sx_assert(&dm->dm_lock, SX_XLOCKED);
425 * Use TAILQ_LAST() to remove "." and ".." last.
426 * We might need ".." to resolve a path in
427 * devfs_dir_unref_de().
429 de = TAILQ_LAST(&dd->de_dlist, devfs_dlist_head);
432 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
433 if (de->de_flags & DE_USER)
434 devfs_dir_unref_de(dm, dd);
435 if (de->de_flags & (DE_DOT | DE_DOTDOT))
436 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
437 else if (de->de_dirent->d_type == DT_DIR)
440 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
442 if (DEVFS_DE_DROP(dd))
443 devfs_dirent_free(dd);
444 else if ((dd->de_flags & DE_DOOMED) == 0)
445 devfs_delete(dm, dd, DEVFS_DEL_NORECURSE);
449 * Each cdev_priv has an array of pointers to devfs_dirent which is indexed
450 * by the mount points dm_idx.
451 * This function extends the array when necessary, taking into account that
452 * the default array is 1 element and not malloc'ed.
455 devfs_metoo(struct cdev_priv *cdp, struct devfs_mount *dm)
457 struct devfs_dirent **dep;
460 siz = (dm->dm_idx + 1) * sizeof *dep;
461 dep = malloc(siz, M_DEVFS2, M_WAITOK | M_ZERO);
463 if (dm->dm_idx <= cdp->cdp_maxdirent) {
469 memcpy(dep, cdp->cdp_dirents, (cdp->cdp_maxdirent + 1) * sizeof *dep);
470 if (cdp->cdp_maxdirent > 0)
471 free(cdp->cdp_dirents, M_DEVFS2);
472 cdp->cdp_dirents = dep;
474 * XXX: if malloc told us how much we actually got this could
477 cdp->cdp_maxdirent = dm->dm_idx;
482 * The caller needs to hold the dm for the duration of the call.
485 devfs_populate_loop(struct devfs_mount *dm, int cleanup)
487 struct cdev_priv *cdp;
488 struct devfs_dirent *de;
489 struct devfs_dirent *dd;
494 sx_assert(&dm->dm_lock, SX_XLOCKED);
496 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
498 KASSERT(cdp->cdp_dirents != NULL, ("NULL cdp_dirents"));
501 * If we are unmounting, or the device has been destroyed,
502 * clean up our dirent.
504 if ((cleanup || !(cdp->cdp_flags & CDP_ACTIVE)) &&
505 dm->dm_idx <= cdp->cdp_maxdirent &&
506 cdp->cdp_dirents[dm->dm_idx] != NULL) {
507 de = cdp->cdp_dirents[dm->dm_idx];
508 cdp->cdp_dirents[dm->dm_idx] = NULL;
509 KASSERT(cdp == de->de_cdp,
510 ("%s %d %s %p %p", __func__, __LINE__,
511 cdp->cdp_c.si_name, cdp, de->de_cdp));
512 KASSERT(de->de_dir != NULL, ("Null de->de_dir"));
515 TAILQ_REMOVE(&de->de_dir->de_dlist, de, de_list);
518 devfs_delete(dm, de, 0);
525 * GC any lingering devices
527 if (!(cdp->cdp_flags & CDP_ACTIVE)) {
528 if (cdp->cdp_inuse > 0)
530 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
532 dev_rel(&cdp->cdp_c);
536 * Don't create any new dirents if we are unmounting
540 KASSERT((cdp->cdp_flags & CDP_ACTIVE), ("Bogons, I tell ya'!"));
542 if (dm->dm_idx <= cdp->cdp_maxdirent &&
543 cdp->cdp_dirents[dm->dm_idx] != NULL) {
544 de = cdp->cdp_dirents[dm->dm_idx];
545 KASSERT(cdp == de->de_cdp, ("inconsistent cdp"));
553 if (dm->dm_idx > cdp->cdp_maxdirent)
554 devfs_metoo(cdp, dm);
557 s = cdp->cdp_c.si_name;
559 for (q = s; *q != '/' && *q != '\0'; q++)
563 de = devfs_find(dd, s, q - s, 0);
565 de = devfs_vmkdir(dm, s, q - s, dd, 0);
566 else if (de->de_dirent->d_type == DT_LNK) {
567 de = devfs_find(dd, s, q - s, DT_DIR);
569 de = devfs_vmkdir(dm, s, q - s, dd, 0);
570 de->de_flags |= DE_COVERED;
574 KASSERT(dd->de_dirent->d_type == DT_DIR &&
575 (dd->de_flags & (DE_DOT | DE_DOTDOT)) == 0,
576 ("%s: invalid directory (si_name=%s)",
577 __func__, cdp->cdp_c.si_name));
581 de = devfs_find(dd, s, q - s, DT_LNK);
583 de_flags |= DE_COVERED;
585 de = devfs_newdirent(s, q - s);
586 if (cdp->cdp_c.si_flags & SI_ALIAS) {
590 de->de_dirent->d_type = DT_LNK;
591 pdev = cdp->cdp_c.si_parent;
592 j = strlen(pdev->si_name) + 1;
593 de->de_symlink = malloc(j, M_DEVFS, M_WAITOK);
594 bcopy(pdev->si_name, de->de_symlink, j);
596 de->de_uid = cdp->cdp_c.si_uid;
597 de->de_gid = cdp->cdp_c.si_gid;
598 de->de_mode = cdp->cdp_c.si_mode;
599 de->de_dirent->d_type = DT_CHR;
601 de->de_flags |= de_flags;
602 de->de_inode = cdp->cdp_inode;
605 mac_devfs_create_device(cdp->cdp_c.si_cred, dm->dm_mount,
609 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
610 devfs_rules_apply(dm, de);
612 /* XXX: could check that cdp is still active here */
613 KASSERT(cdp->cdp_dirents[dm->dm_idx] == NULL,
614 ("%s %d\n", __func__, __LINE__));
615 cdp->cdp_dirents[dm->dm_idx] = de;
616 KASSERT(de->de_cdp != (void *)0xdeadc0de,
617 ("%s %d\n", __func__, __LINE__));
626 * The caller needs to hold the dm for the duration of the call.
629 devfs_populate(struct devfs_mount *dm)
633 sx_assert(&dm->dm_lock, SX_XLOCKED);
634 gen = devfs_generation;
635 if (dm->dm_generation == gen)
637 while (devfs_populate_loop(dm, 0))
639 dm->dm_generation = gen;
643 * The caller needs to hold the dm for the duration of the call.
646 devfs_cleanup(struct devfs_mount *dm)
649 sx_assert(&dm->dm_lock, SX_XLOCKED);
650 while (devfs_populate_loop(dm, 1))
652 devfs_purge(dm, dm->dm_rootdir);
656 * devfs_create() and devfs_destroy() are called from kern_conf.c and
657 * in both cases the devlock() mutex is held, so no further locking
658 * is necessary and no sleeping allowed.
662 devfs_create(struct cdev *dev)
664 struct cdev_priv *cdp;
666 mtx_assert(&devmtx, MA_OWNED);
667 cdp = cdev2priv(dev);
668 cdp->cdp_flags |= CDP_ACTIVE;
669 cdp->cdp_inode = alloc_unrl(devfs_inos);
671 TAILQ_INSERT_TAIL(&cdevp_list, cdp, cdp_list);
676 devfs_destroy(struct cdev *dev)
678 struct cdev_priv *cdp;
680 mtx_assert(&devmtx, MA_OWNED);
681 cdp = cdev2priv(dev);
682 cdp->cdp_flags &= ~CDP_ACTIVE;
687 devfs_alloc_cdp_inode(void)
690 return (alloc_unr(devfs_inos));
694 devfs_free_cdp_inode(ino_t ino)
698 free_unr(devfs_inos, ino);
702 devfs_devs_init(void *junk __unused)
705 devfs_inos = new_unrhdr(DEVFS_ROOTINO + 1, INT_MAX, &devmtx);
708 SYSINIT(devfs_devs, SI_SUB_DEVFS, SI_ORDER_FIRST, devfs_devs_init, NULL);