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_USE_RESERVE | 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 cdev->si_name = cdev->__si_namebuf;
137 LIST_INIT(&cdev->si_children);
139 cdev->si_atime = cdev->si_mtime = cdev->si_ctime = ts;
140 cdev->si_cred = NULL;
146 devfs_dev_exists(const char *name)
148 struct cdev_priv *cdp;
150 mtx_assert(&devmtx, MA_OWNED);
152 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
153 if ((cdp->cdp_flags & CDP_ACTIVE) == 0)
155 if (devfs_pathpath(cdp->cdp_c.si_name, name) != 0)
157 if (devfs_pathpath(name, cdp->cdp_c.si_name) != 0)
160 if (devfs_dir_find(name) != 0)
167 devfs_free(struct cdev *cdev)
169 struct cdev_priv *cdp;
171 cdp = cdev2priv(cdev);
172 if (cdev->si_cred != NULL)
173 crfree(cdev->si_cred);
174 if (cdp->cdp_inode > 0)
175 free_unr(devfs_inos, cdp->cdp_inode);
176 if (cdp->cdp_maxdirent > 0)
177 free(cdp->cdp_dirents, M_DEVFS2);
181 struct devfs_dirent *
182 devfs_find(struct devfs_dirent *dd, const char *name, int namelen, int type)
184 struct devfs_dirent *de;
186 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
187 if (namelen != de->de_dirent->d_namlen)
189 if (type != 0 && type != de->de_dirent->d_type)
191 if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
195 KASSERT(de == NULL || (de->de_flags & DE_DOOMED) == 0,
196 ("devfs_find: returning a doomed entry"));
200 struct devfs_dirent *
201 devfs_newdirent(char *name, int namelen)
204 struct devfs_dirent *de;
207 d.d_namlen = namelen;
208 i = sizeof (*de) + GENERIC_DIRSIZ(&d);
209 de = malloc(i, M_DEVFS3, M_WAITOK | M_ZERO);
210 de->de_dirent = (struct dirent *)(de + 1);
211 de->de_dirent->d_namlen = namelen;
212 de->de_dirent->d_reclen = GENERIC_DIRSIZ(&d);
213 bcopy(name, de->de_dirent->d_name, namelen);
214 de->de_dirent->d_name[namelen] = '\0';
215 vfs_timestamp(&de->de_ctime);
216 de->de_mtime = de->de_atime = de->de_ctime;
225 struct devfs_dirent *
226 devfs_parent_dirent(struct devfs_dirent *de)
229 if (de->de_dirent->d_type != DT_DIR)
232 if (de->de_flags & (DE_DOT | DE_DOTDOT))
235 de = TAILQ_FIRST(&de->de_dlist); /* "." */
238 de = TAILQ_NEXT(de, de_list); /* ".." */
245 struct devfs_dirent *
246 devfs_vmkdir(struct devfs_mount *dmp, char *name, int namelen, struct devfs_dirent *dotdot, u_int inode)
248 struct devfs_dirent *dd;
249 struct devfs_dirent *de;
251 /* Create the new directory */
252 dd = devfs_newdirent(name, namelen);
253 TAILQ_INIT(&dd->de_dlist);
254 dd->de_dirent->d_type = DT_DIR;
259 dd->de_inode = inode;
261 dd->de_inode = alloc_unr(devfs_inos);
264 * "." and ".." are always the two first entries in the
267 * Create the "." entry in the new directory.
269 de = devfs_newdirent(".", 1);
270 de->de_dirent->d_type = DT_DIR;
271 de->de_flags |= DE_DOT;
272 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
275 /* Create the ".." entry in the new directory. */
276 de = devfs_newdirent("..", 2);
277 de->de_dirent->d_type = DT_DIR;
278 de->de_flags |= DE_DOTDOT;
279 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
280 if (dotdot == NULL) {
284 sx_assert(&dmp->dm_lock, SX_XLOCKED);
285 TAILQ_INSERT_TAIL(&dotdot->de_dlist, dd, de_list);
287 devfs_rules_apply(dmp, dd);
291 mac_devfs_create_directory(dmp->dm_mount, name, namelen, dd);
297 devfs_dirent_free(struct devfs_dirent *de)
303 * Removes a directory if it is empty. Also empty parent directories are
304 * removed recursively.
307 devfs_rmdir_empty(struct devfs_mount *dm, struct devfs_dirent *de)
309 struct devfs_dirent *dd, *de_dot, *de_dotdot;
311 sx_assert(&dm->dm_lock, SX_XLOCKED);
314 KASSERT(de->de_dirent->d_type == DT_DIR,
315 ("devfs_rmdir_empty: de is not a directory"));
317 if ((de->de_flags & DE_DOOMED) != 0 || de == dm->dm_rootdir)
320 de_dot = TAILQ_FIRST(&de->de_dlist);
321 KASSERT(de_dot != NULL, ("devfs_rmdir_empty: . missing"));
322 de_dotdot = TAILQ_NEXT(de_dot, de_list);
323 KASSERT(de_dotdot != NULL, ("devfs_rmdir_empty: .. missing"));
324 /* Return if the directory is not empty. */
325 if (TAILQ_NEXT(de_dotdot, de_list) != NULL)
328 dd = devfs_parent_dirent(de);
329 KASSERT(dd != NULL, ("devfs_rmdir_empty: NULL dd"));
330 TAILQ_REMOVE(&de->de_dlist, de_dot, de_list);
331 TAILQ_REMOVE(&de->de_dlist, de_dotdot, de_list);
332 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
334 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
335 devfs_delete(dm, de_dot, DEVFS_DEL_NORECURSE);
336 devfs_delete(dm, de_dotdot, DEVFS_DEL_NORECURSE);
337 if (DEVFS_DE_DROP(dd)) {
338 devfs_dirent_free(dd);
347 * The caller needs to hold the dm for the duration of the call since
348 * dm->dm_lock may be temporary dropped.
351 devfs_delete(struct devfs_mount *dm, struct devfs_dirent *de, int flags)
353 struct devfs_dirent *dd;
356 KASSERT((de->de_flags & DE_DOOMED) == 0,
357 ("devfs_delete doomed dirent"));
358 de->de_flags |= DE_DOOMED;
360 if ((flags & DEVFS_DEL_NORECURSE) == 0) {
361 dd = devfs_parent_dirent(de);
364 if (de->de_flags & DE_USER) {
365 KASSERT(dd != NULL, ("devfs_delete: NULL dd"));
366 devfs_dir_unref_de(dm, dd);
371 mtx_lock(&devfs_de_interlock);
375 mtx_unlock(&devfs_de_interlock);
377 sx_unlock(&dm->dm_lock);
378 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
379 vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
383 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
386 sx_xlock(&dm->dm_lock);
388 mtx_unlock(&devfs_de_interlock);
389 if (de->de_symlink) {
390 free(de->de_symlink, M_DEVFS);
391 de->de_symlink = NULL;
394 mac_devfs_destroy(de);
396 if (de->de_inode > DEVFS_ROOTINO) {
397 free_unr(devfs_inos, de->de_inode);
400 if (DEVFS_DE_DROP(de))
401 devfs_dirent_free(de);
404 if (DEVFS_DE_DROP(dd))
405 devfs_dirent_free(dd);
407 devfs_rmdir_empty(dm, dd);
413 * Recursively removes the entire tree.
414 * The caller needs to hold the dm for the duration of the call.
418 devfs_purge(struct devfs_mount *dm, struct devfs_dirent *dd)
420 struct devfs_dirent *de;
422 sx_assert(&dm->dm_lock, SX_XLOCKED);
427 * Use TAILQ_LAST() to remove "." and ".." last.
428 * We might need ".." to resolve a path in
429 * devfs_dir_unref_de().
431 de = TAILQ_LAST(&dd->de_dlist, devfs_dlist_head);
434 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
435 if (de->de_flags & DE_USER)
436 devfs_dir_unref_de(dm, dd);
437 if (de->de_flags & (DE_DOT | DE_DOTDOT))
438 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
439 else if (de->de_dirent->d_type == DT_DIR)
442 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
444 if (DEVFS_DE_DROP(dd))
445 devfs_dirent_free(dd);
446 else if ((dd->de_flags & DE_DOOMED) == 0)
447 devfs_delete(dm, dd, DEVFS_DEL_NORECURSE);
451 * Each cdev_priv has an array of pointers to devfs_dirent which is indexed
452 * by the mount points dm_idx.
453 * This function extends the array when necessary, taking into account that
454 * the default array is 1 element and not malloc'ed.
457 devfs_metoo(struct cdev_priv *cdp, struct devfs_mount *dm)
459 struct devfs_dirent **dep;
462 siz = (dm->dm_idx + 1) * sizeof *dep;
463 dep = malloc(siz, M_DEVFS2, M_WAITOK | M_ZERO);
465 if (dm->dm_idx <= cdp->cdp_maxdirent) {
471 memcpy(dep, cdp->cdp_dirents, (cdp->cdp_maxdirent + 1) * sizeof *dep);
472 if (cdp->cdp_maxdirent > 0)
473 free(cdp->cdp_dirents, M_DEVFS2);
474 cdp->cdp_dirents = dep;
476 * XXX: if malloc told us how much we actually got this could
479 cdp->cdp_maxdirent = dm->dm_idx;
484 * The caller needs to hold the dm for the duration of the call.
487 devfs_populate_loop(struct devfs_mount *dm, int cleanup)
489 struct cdev_priv *cdp;
490 struct devfs_dirent *de;
491 struct devfs_dirent *dd;
496 sx_assert(&dm->dm_lock, SX_XLOCKED);
498 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
500 KASSERT(cdp->cdp_dirents != NULL, ("NULL cdp_dirents"));
503 * If we are unmounting, or the device has been destroyed,
504 * clean up our dirent.
506 if ((cleanup || !(cdp->cdp_flags & CDP_ACTIVE)) &&
507 dm->dm_idx <= cdp->cdp_maxdirent &&
508 cdp->cdp_dirents[dm->dm_idx] != NULL) {
509 de = cdp->cdp_dirents[dm->dm_idx];
510 cdp->cdp_dirents[dm->dm_idx] = NULL;
511 KASSERT(cdp == de->de_cdp,
512 ("%s %d %s %p %p", __func__, __LINE__,
513 cdp->cdp_c.si_name, cdp, de->de_cdp));
514 KASSERT(de->de_dir != NULL, ("Null de->de_dir"));
517 TAILQ_REMOVE(&de->de_dir->de_dlist, de, de_list);
520 devfs_delete(dm, de, 0);
527 * GC any lingering devices
529 if (!(cdp->cdp_flags & CDP_ACTIVE)) {
530 if (cdp->cdp_inuse > 0)
532 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
534 dev_rel(&cdp->cdp_c);
538 * Don't create any new dirents if we are unmounting
542 KASSERT((cdp->cdp_flags & CDP_ACTIVE), ("Bogons, I tell ya'!"));
544 if (dm->dm_idx <= cdp->cdp_maxdirent &&
545 cdp->cdp_dirents[dm->dm_idx] != NULL) {
546 de = cdp->cdp_dirents[dm->dm_idx];
547 KASSERT(cdp == de->de_cdp, ("inconsistent cdp"));
555 if (dm->dm_idx > cdp->cdp_maxdirent)
556 devfs_metoo(cdp, dm);
559 s = cdp->cdp_c.si_name;
561 for (q = s; *q != '/' && *q != '\0'; q++)
565 de = devfs_find(dd, s, q - s, 0);
567 de = devfs_vmkdir(dm, s, q - s, dd, 0);
568 else if (de->de_dirent->d_type == DT_LNK) {
569 de = devfs_find(dd, s, q - s, DT_DIR);
571 de = devfs_vmkdir(dm, s, q - s, dd, 0);
572 de->de_flags |= DE_COVERED;
576 KASSERT(dd->de_dirent->d_type == DT_DIR &&
577 (dd->de_flags & (DE_DOT | DE_DOTDOT)) == 0,
578 ("%s: invalid directory (si_name=%s)",
579 __func__, cdp->cdp_c.si_name));
583 de = devfs_find(dd, s, q - s, DT_LNK);
585 de_flags |= DE_COVERED;
587 de = devfs_newdirent(s, q - s);
588 if (cdp->cdp_c.si_flags & SI_ALIAS) {
592 de->de_dirent->d_type = DT_LNK;
593 pdev = cdp->cdp_c.si_parent;
594 j = strlen(pdev->si_name) + 1;
595 de->de_symlink = malloc(j, M_DEVFS, M_WAITOK);
596 bcopy(pdev->si_name, de->de_symlink, j);
598 de->de_uid = cdp->cdp_c.si_uid;
599 de->de_gid = cdp->cdp_c.si_gid;
600 de->de_mode = cdp->cdp_c.si_mode;
601 de->de_dirent->d_type = DT_CHR;
603 de->de_flags |= de_flags;
604 de->de_inode = cdp->cdp_inode;
607 mac_devfs_create_device(cdp->cdp_c.si_cred, dm->dm_mount,
611 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
612 devfs_rules_apply(dm, de);
614 /* XXX: could check that cdp is still active here */
615 KASSERT(cdp->cdp_dirents[dm->dm_idx] == NULL,
616 ("%s %d\n", __func__, __LINE__));
617 cdp->cdp_dirents[dm->dm_idx] = de;
618 KASSERT(de->de_cdp != (void *)0xdeadc0de,
619 ("%s %d\n", __func__, __LINE__));
628 * The caller needs to hold the dm for the duration of the call.
631 devfs_populate(struct devfs_mount *dm)
635 sx_assert(&dm->dm_lock, SX_XLOCKED);
636 gen = devfs_generation;
637 if (dm->dm_generation == gen)
639 while (devfs_populate_loop(dm, 0))
641 dm->dm_generation = gen;
645 * The caller needs to hold the dm for the duration of the call.
648 devfs_cleanup(struct devfs_mount *dm)
651 sx_assert(&dm->dm_lock, SX_XLOCKED);
652 while (devfs_populate_loop(dm, 1))
654 devfs_purge(dm, dm->dm_rootdir);
658 * devfs_create() and devfs_destroy() are called from kern_conf.c and
659 * in both cases the devlock() mutex is held, so no further locking
660 * is necesary and no sleeping allowed.
664 devfs_create(struct cdev *dev)
666 struct cdev_priv *cdp;
668 mtx_assert(&devmtx, MA_OWNED);
669 cdp = cdev2priv(dev);
670 cdp->cdp_flags |= CDP_ACTIVE;
671 cdp->cdp_inode = alloc_unrl(devfs_inos);
673 TAILQ_INSERT_TAIL(&cdevp_list, cdp, cdp_list);
678 devfs_destroy(struct cdev *dev)
680 struct cdev_priv *cdp;
682 mtx_assert(&devmtx, MA_OWNED);
683 cdp = cdev2priv(dev);
684 cdp->cdp_flags &= ~CDP_ACTIVE;
689 devfs_devs_init(void *junk __unused)
692 devfs_inos = new_unrhdr(DEVFS_ROOTINO + 1, INT_MAX, &devmtx);
695 SYSINIT(devfs_devs, SI_SUB_DEVFS, SI_ORDER_FIRST, devfs_devs_init, NULL);