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 SYSCTL_NULL_INT_PTR, sizeof(struct cdev), "sizeof(struct cdev)");
114 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev_priv, CTLFLAG_RD,
115 SYSCTL_NULL_INT_PTR, 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;
141 * Avoid race with dev_rel() by setting the initial
142 * reference count to 1. This last reference is taken
143 * by the destroy_dev() function.
145 cdev->si_refcount = 1;
151 devfs_dev_exists(const char *name)
153 struct cdev_priv *cdp;
155 mtx_assert(&devmtx, MA_OWNED);
157 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
158 if ((cdp->cdp_flags & CDP_ACTIVE) == 0)
160 if (devfs_pathpath(cdp->cdp_c.si_name, name) != 0)
162 if (devfs_pathpath(name, cdp->cdp_c.si_name) != 0)
165 if (devfs_dir_find(name) != 0)
172 devfs_free(struct cdev *cdev)
174 struct cdev_priv *cdp;
176 cdp = cdev2priv(cdev);
177 if (cdev->si_cred != NULL)
178 crfree(cdev->si_cred);
179 devfs_free_cdp_inode(cdp->cdp_inode);
180 if (cdp->cdp_maxdirent > 0)
181 free(cdp->cdp_dirents, M_DEVFS2);
185 struct devfs_dirent *
186 devfs_find(struct devfs_dirent *dd, const char *name, int namelen, int type)
188 struct devfs_dirent *de;
190 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
191 if (namelen != de->de_dirent->d_namlen)
193 if (type != 0 && type != de->de_dirent->d_type)
195 if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
199 KASSERT(de == NULL || (de->de_flags & DE_DOOMED) == 0,
200 ("devfs_find: returning a doomed entry"));
204 struct devfs_dirent *
205 devfs_newdirent(char *name, int namelen)
208 struct devfs_dirent *de;
211 d.d_namlen = namelen;
212 i = sizeof (*de) + GENERIC_DIRSIZ(&d);
213 de = malloc(i, M_DEVFS3, M_WAITOK | M_ZERO);
214 de->de_dirent = (struct dirent *)(de + 1);
215 de->de_dirent->d_namlen = namelen;
216 de->de_dirent->d_reclen = GENERIC_DIRSIZ(&d);
217 bcopy(name, de->de_dirent->d_name, namelen);
218 de->de_dirent->d_name[namelen] = '\0';
219 vfs_timestamp(&de->de_ctime);
220 de->de_mtime = de->de_atime = de->de_ctime;
229 struct devfs_dirent *
230 devfs_parent_dirent(struct devfs_dirent *de)
233 if (de->de_dirent->d_type != DT_DIR)
236 if (de->de_flags & (DE_DOT | DE_DOTDOT))
239 de = TAILQ_FIRST(&de->de_dlist); /* "." */
242 de = TAILQ_NEXT(de, de_list); /* ".." */
249 struct devfs_dirent *
250 devfs_vmkdir(struct devfs_mount *dmp, char *name, int namelen, struct devfs_dirent *dotdot, u_int inode)
252 struct devfs_dirent *dd;
253 struct devfs_dirent *de;
255 /* Create the new directory */
256 dd = devfs_newdirent(name, namelen);
257 TAILQ_INIT(&dd->de_dlist);
258 dd->de_dirent->d_type = DT_DIR;
263 dd->de_inode = inode;
265 dd->de_inode = alloc_unr(devfs_inos);
268 * "." and ".." are always the two first entries in the
271 * Create the "." entry in the new directory.
273 de = devfs_newdirent(".", 1);
274 de->de_dirent->d_type = DT_DIR;
275 de->de_flags |= DE_DOT;
276 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
279 /* Create the ".." entry in the new directory. */
280 de = devfs_newdirent("..", 2);
281 de->de_dirent->d_type = DT_DIR;
282 de->de_flags |= DE_DOTDOT;
283 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
284 if (dotdot == NULL) {
288 sx_assert(&dmp->dm_lock, SX_XLOCKED);
289 TAILQ_INSERT_TAIL(&dotdot->de_dlist, dd, de_list);
291 devfs_rules_apply(dmp, dd);
295 mac_devfs_create_directory(dmp->dm_mount, name, namelen, dd);
301 devfs_dirent_free(struct devfs_dirent *de)
307 * Removes a directory if it is empty. Also empty parent directories are
308 * removed recursively.
311 devfs_rmdir_empty(struct devfs_mount *dm, struct devfs_dirent *de)
313 struct devfs_dirent *dd, *de_dot, *de_dotdot;
315 sx_assert(&dm->dm_lock, SX_XLOCKED);
318 KASSERT(de->de_dirent->d_type == DT_DIR,
319 ("devfs_rmdir_empty: de is not a directory"));
321 if ((de->de_flags & DE_DOOMED) != 0 || de == dm->dm_rootdir)
324 de_dot = TAILQ_FIRST(&de->de_dlist);
325 KASSERT(de_dot != NULL, ("devfs_rmdir_empty: . missing"));
326 de_dotdot = TAILQ_NEXT(de_dot, de_list);
327 KASSERT(de_dotdot != NULL, ("devfs_rmdir_empty: .. missing"));
328 /* Return if the directory is not empty. */
329 if (TAILQ_NEXT(de_dotdot, de_list) != NULL)
332 dd = devfs_parent_dirent(de);
333 KASSERT(dd != NULL, ("devfs_rmdir_empty: NULL dd"));
334 TAILQ_REMOVE(&de->de_dlist, de_dot, de_list);
335 TAILQ_REMOVE(&de->de_dlist, de_dotdot, de_list);
336 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
338 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
339 devfs_delete(dm, de_dot, DEVFS_DEL_NORECURSE);
340 devfs_delete(dm, de_dotdot, DEVFS_DEL_NORECURSE);
341 if (DEVFS_DE_DROP(dd)) {
342 devfs_dirent_free(dd);
351 * The caller needs to hold the dm for the duration of the call since
352 * dm->dm_lock may be temporary dropped.
355 devfs_delete(struct devfs_mount *dm, struct devfs_dirent *de, int flags)
357 struct devfs_dirent *dd;
360 KASSERT((de->de_flags & DE_DOOMED) == 0,
361 ("devfs_delete doomed dirent"));
362 de->de_flags |= DE_DOOMED;
364 if ((flags & DEVFS_DEL_NORECURSE) == 0) {
365 dd = devfs_parent_dirent(de);
368 if (de->de_flags & DE_USER) {
369 KASSERT(dd != NULL, ("devfs_delete: NULL dd"));
370 devfs_dir_unref_de(dm, dd);
375 mtx_lock(&devfs_de_interlock);
379 mtx_unlock(&devfs_de_interlock);
381 sx_unlock(&dm->dm_lock);
382 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
383 vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
387 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
390 sx_xlock(&dm->dm_lock);
392 mtx_unlock(&devfs_de_interlock);
393 if (de->de_symlink) {
394 free(de->de_symlink, M_DEVFS);
395 de->de_symlink = NULL;
398 mac_devfs_destroy(de);
400 if (de->de_inode > DEVFS_ROOTINO) {
401 devfs_free_cdp_inode(de->de_inode);
404 if (DEVFS_DE_DROP(de))
405 devfs_dirent_free(de);
408 if (DEVFS_DE_DROP(dd))
409 devfs_dirent_free(dd);
411 devfs_rmdir_empty(dm, dd);
417 * Recursively removes the entire tree.
418 * The caller needs to hold the dm for the duration of the call.
422 devfs_purge(struct devfs_mount *dm, struct devfs_dirent *dd)
424 struct devfs_dirent *de;
426 sx_assert(&dm->dm_lock, SX_XLOCKED);
431 * Use TAILQ_LAST() to remove "." and ".." last.
432 * We might need ".." to resolve a path in
433 * devfs_dir_unref_de().
435 de = TAILQ_LAST(&dd->de_dlist, devfs_dlist_head);
438 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
439 if (de->de_flags & DE_USER)
440 devfs_dir_unref_de(dm, dd);
441 if (de->de_flags & (DE_DOT | DE_DOTDOT))
442 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
443 else if (de->de_dirent->d_type == DT_DIR)
446 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
448 if (DEVFS_DE_DROP(dd))
449 devfs_dirent_free(dd);
450 else if ((dd->de_flags & DE_DOOMED) == 0)
451 devfs_delete(dm, dd, DEVFS_DEL_NORECURSE);
455 * Each cdev_priv has an array of pointers to devfs_dirent which is indexed
456 * by the mount points dm_idx.
457 * This function extends the array when necessary, taking into account that
458 * the default array is 1 element and not malloc'ed.
461 devfs_metoo(struct cdev_priv *cdp, struct devfs_mount *dm)
463 struct devfs_dirent **dep;
466 siz = (dm->dm_idx + 1) * sizeof *dep;
467 dep = malloc(siz, M_DEVFS2, M_WAITOK | M_ZERO);
469 if (dm->dm_idx <= cdp->cdp_maxdirent) {
475 memcpy(dep, cdp->cdp_dirents, (cdp->cdp_maxdirent + 1) * sizeof *dep);
476 if (cdp->cdp_maxdirent > 0)
477 free(cdp->cdp_dirents, M_DEVFS2);
478 cdp->cdp_dirents = dep;
480 * XXX: if malloc told us how much we actually got this could
483 cdp->cdp_maxdirent = dm->dm_idx;
488 * The caller needs to hold the dm for the duration of the call.
491 devfs_populate_loop(struct devfs_mount *dm, int cleanup)
493 struct cdev_priv *cdp;
494 struct devfs_dirent *de;
495 struct devfs_dirent *dd, *dt;
497 int de_flags, depth, j;
500 sx_assert(&dm->dm_lock, SX_XLOCKED);
502 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
504 KASSERT(cdp->cdp_dirents != NULL, ("NULL cdp_dirents"));
507 * If we are unmounting, or the device has been destroyed,
508 * clean up our dirent.
510 if ((cleanup || !(cdp->cdp_flags & CDP_ACTIVE)) &&
511 dm->dm_idx <= cdp->cdp_maxdirent &&
512 cdp->cdp_dirents[dm->dm_idx] != NULL) {
513 de = cdp->cdp_dirents[dm->dm_idx];
514 cdp->cdp_dirents[dm->dm_idx] = NULL;
515 KASSERT(cdp == de->de_cdp,
516 ("%s %d %s %p %p", __func__, __LINE__,
517 cdp->cdp_c.si_name, cdp, de->de_cdp));
518 KASSERT(de->de_dir != NULL, ("Null de->de_dir"));
521 TAILQ_REMOVE(&de->de_dir->de_dlist, de, de_list);
524 devfs_delete(dm, de, 0);
531 * GC any lingering devices
533 if (!(cdp->cdp_flags & CDP_ACTIVE)) {
534 if (cdp->cdp_inuse > 0)
536 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
538 dev_rel(&cdp->cdp_c);
542 * Don't create any new dirents if we are unmounting
546 KASSERT((cdp->cdp_flags & CDP_ACTIVE), ("Bogons, I tell ya'!"));
548 if (dm->dm_idx <= cdp->cdp_maxdirent &&
549 cdp->cdp_dirents[dm->dm_idx] != NULL) {
550 de = cdp->cdp_dirents[dm->dm_idx];
551 KASSERT(cdp == de->de_cdp, ("inconsistent cdp"));
559 if (dm->dm_idx > cdp->cdp_maxdirent)
560 devfs_metoo(cdp, dm);
563 s = cdp->cdp_c.si_name;
565 for (q = s; *q != '/' && *q != '\0'; q++)
569 de = devfs_find(dd, s, q - s, 0);
571 de = devfs_vmkdir(dm, s, q - s, dd, 0);
572 else if (de->de_dirent->d_type == DT_LNK) {
573 de = devfs_find(dd, s, q - s, DT_DIR);
575 de = devfs_vmkdir(dm, s, q - s, dd, 0);
576 de->de_flags |= DE_COVERED;
580 KASSERT(dd->de_dirent->d_type == DT_DIR &&
581 (dd->de_flags & (DE_DOT | DE_DOTDOT)) == 0,
582 ("%s: invalid directory (si_name=%s)",
583 __func__, cdp->cdp_c.si_name));
587 de = devfs_find(dd, s, q - s, DT_LNK);
589 de_flags |= DE_COVERED;
591 de = devfs_newdirent(s, q - s);
592 if (cdp->cdp_c.si_flags & SI_ALIAS) {
596 de->de_dirent->d_type = DT_LNK;
597 pdev = cdp->cdp_c.si_parent;
600 while (dt != dm->dm_rootdir &&
601 (dt = devfs_parent_dirent(dt)) != NULL)
603 j = depth * 3 + strlen(pdev->si_name) + 1;
604 de->de_symlink = malloc(j, M_DEVFS, M_WAITOK);
605 de->de_symlink[0] = 0;
607 strcat(de->de_symlink, "../");
608 strcat(de->de_symlink, pdev->si_name);
610 de->de_uid = cdp->cdp_c.si_uid;
611 de->de_gid = cdp->cdp_c.si_gid;
612 de->de_mode = cdp->cdp_c.si_mode;
613 de->de_dirent->d_type = DT_CHR;
615 de->de_flags |= de_flags;
616 de->de_inode = cdp->cdp_inode;
619 mac_devfs_create_device(cdp->cdp_c.si_cred, dm->dm_mount,
623 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
624 devfs_rules_apply(dm, de);
626 /* XXX: could check that cdp is still active here */
627 KASSERT(cdp->cdp_dirents[dm->dm_idx] == NULL,
628 ("%s %d\n", __func__, __LINE__));
629 cdp->cdp_dirents[dm->dm_idx] = de;
630 KASSERT(de->de_cdp != (void *)0xdeadc0de,
631 ("%s %d\n", __func__, __LINE__));
640 * The caller needs to hold the dm for the duration of the call.
643 devfs_populate(struct devfs_mount *dm)
647 sx_assert(&dm->dm_lock, SX_XLOCKED);
648 gen = devfs_generation;
649 if (dm->dm_generation == gen)
651 while (devfs_populate_loop(dm, 0))
653 dm->dm_generation = gen;
657 * The caller needs to hold the dm for the duration of the call.
660 devfs_cleanup(struct devfs_mount *dm)
663 sx_assert(&dm->dm_lock, SX_XLOCKED);
664 while (devfs_populate_loop(dm, 1))
666 devfs_purge(dm, dm->dm_rootdir);
670 * devfs_create() and devfs_destroy() are called from kern_conf.c and
671 * in both cases the devlock() mutex is held, so no further locking
672 * is necessary and no sleeping allowed.
676 devfs_create(struct cdev *dev)
678 struct cdev_priv *cdp;
680 mtx_assert(&devmtx, MA_OWNED);
681 cdp = cdev2priv(dev);
682 cdp->cdp_flags |= CDP_ACTIVE;
683 cdp->cdp_inode = alloc_unrl(devfs_inos);
685 TAILQ_INSERT_TAIL(&cdevp_list, cdp, cdp_list);
690 devfs_destroy(struct cdev *dev)
692 struct cdev_priv *cdp;
694 mtx_assert(&devmtx, MA_OWNED);
695 cdp = cdev2priv(dev);
696 cdp->cdp_flags &= ~CDP_ACTIVE;
701 devfs_alloc_cdp_inode(void)
704 return (alloc_unr(devfs_inos));
708 devfs_free_cdp_inode(ino_t ino)
712 free_unr(devfs_inos, ino);
716 devfs_devs_init(void *junk __unused)
719 devfs_inos = new_unrhdr(DEVFS_ROOTINO + 1, INT_MAX, &devmtx);
722 SYSINIT(devfs_devs, SI_SUB_DEVFS, SI_ORDER_FIRST, devfs_devs_init, NULL);