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 "opt_compat.h"
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, 0, "DEVFS filesystem");
68 static unsigned devfs_generation;
69 SYSCTL_UINT(_vfs_devfs, OID_AUTO, generation, CTLFLAG_RD,
70 &devfs_generation, 0, "DEVFS generation number");
72 unsigned devfs_rule_depth = 1;
73 SYSCTL_UINT(_vfs_devfs, OID_AUTO, rule_depth, CTLFLAG_RW,
74 &devfs_rule_depth, 0, "Max depth of ruleset include");
77 * Helper sysctl for devname(3). We're given a dev_t and return the
78 * name, if any, registered by the device driver.
81 sysctl_devname(SYSCTL_HANDLER_ARGS)
85 #ifdef COMPAT_FREEBSD11
88 struct cdev_priv *cdp;
91 #ifdef COMPAT_FREEBSD11
92 if (req->newlen == sizeof(ud_compat)) {
93 error = SYSCTL_IN(req, &ud_compat, sizeof(ud_compat));
95 ud = ud_compat == (uint32_t)NODEV ? NODEV : ud_compat;
98 error = SYSCTL_IN(req, &ud, sizeof (ud));
105 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list)
106 if (cdp->cdp_inode == ud) {
114 error = SYSCTL_OUT(req, dev->si_name, strlen(dev->si_name) + 1);
119 SYSCTL_PROC(_kern, OID_AUTO, devname,
120 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
121 NULL, 0, sysctl_devname, "", "devname(3) handler");
123 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev, CTLFLAG_RD,
124 SYSCTL_NULL_INT_PTR, sizeof(struct cdev), "sizeof(struct cdev)");
126 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev_priv, CTLFLAG_RD,
127 SYSCTL_NULL_INT_PTR, sizeof(struct cdev_priv), "sizeof(struct cdev_priv)");
130 devfs_alloc(int flags)
132 struct cdev_priv *cdp;
136 cdp = malloc(sizeof *cdp, M_CDEVP, M_ZERO |
137 ((flags & MAKEDEV_NOWAIT) ? M_NOWAIT : M_WAITOK));
141 cdp->cdp_dirents = &cdp->cdp_dirent0;
144 LIST_INIT(&cdev->si_children);
146 cdev->si_atime = cdev->si_mtime = cdev->si_ctime = ts;
152 devfs_dev_exists(const char *name)
154 struct cdev_priv *cdp;
156 mtx_assert(&devmtx, MA_OWNED);
158 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
159 if ((cdp->cdp_flags & CDP_ACTIVE) == 0)
161 if (devfs_pathpath(cdp->cdp_c.si_name, name) != 0)
163 if (devfs_pathpath(name, cdp->cdp_c.si_name) != 0)
166 if (devfs_dir_find(name) != 0)
173 devfs_free(struct cdev *cdev)
175 struct cdev_priv *cdp;
177 cdp = cdev2priv(cdev);
178 if (cdev->si_cred != NULL)
179 crfree(cdev->si_cred);
180 devfs_free_cdp_inode(cdp->cdp_inode);
181 if (cdp->cdp_maxdirent > 0)
182 free(cdp->cdp_dirents, M_DEVFS2);
186 struct devfs_dirent *
187 devfs_find(struct devfs_dirent *dd, const char *name, int namelen, int type)
189 struct devfs_dirent *de;
191 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
192 if (namelen != de->de_dirent->d_namlen)
194 if (type != 0 && type != de->de_dirent->d_type)
198 * The race with finding non-active name is not
199 * completely closed by the check, but it is similar
200 * to the devfs_allocv() in making it unlikely enough.
202 if (de->de_dirent->d_type == DT_CHR &&
203 (de->de_cdp->cdp_flags & CDP_ACTIVE) == 0)
206 if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
210 KASSERT(de == NULL || (de->de_flags & DE_DOOMED) == 0,
211 ("devfs_find: returning a doomed entry"));
215 struct devfs_dirent *
216 devfs_newdirent(char *name, int namelen)
219 struct devfs_dirent *de;
222 d.d_namlen = namelen;
223 i = sizeof(*de) + GENERIC_DIRSIZ(&d);
224 de = malloc(i, M_DEVFS3, M_WAITOK | M_ZERO);
225 de->de_dirent = (struct dirent *)(de + 1);
226 de->de_dirent->d_namlen = namelen;
227 de->de_dirent->d_reclen = GENERIC_DIRSIZ(&d);
228 bcopy(name, de->de_dirent->d_name, namelen);
229 de->de_dirent->d_name[namelen] = '\0';
230 vfs_timestamp(&de->de_ctime);
231 de->de_mtime = de->de_atime = de->de_ctime;
240 struct devfs_dirent *
241 devfs_parent_dirent(struct devfs_dirent *de)
244 if (de->de_dirent->d_type != DT_DIR)
247 if (de->de_flags & (DE_DOT | DE_DOTDOT))
250 de = TAILQ_FIRST(&de->de_dlist); /* "." */
253 de = TAILQ_NEXT(de, de_list); /* ".." */
260 struct devfs_dirent *
261 devfs_vmkdir(struct devfs_mount *dmp, char *name, int namelen,
262 struct devfs_dirent *dotdot, u_int inode)
264 struct devfs_dirent *dd;
265 struct devfs_dirent *de;
267 /* Create the new directory */
268 dd = devfs_newdirent(name, namelen);
269 TAILQ_INIT(&dd->de_dlist);
270 dd->de_dirent->d_type = DT_DIR;
275 dd->de_inode = inode;
277 dd->de_inode = alloc_unr(devfs_inos);
280 * "." and ".." are always the two first entries in the
283 * Create the "." entry in the new directory.
285 de = devfs_newdirent(".", 1);
286 de->de_dirent->d_type = DT_DIR;
287 de->de_flags |= DE_DOT;
288 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
291 /* Create the ".." entry in the new directory. */
292 de = devfs_newdirent("..", 2);
293 de->de_dirent->d_type = DT_DIR;
294 de->de_flags |= DE_DOTDOT;
295 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
296 if (dotdot == NULL) {
300 sx_assert(&dmp->dm_lock, SX_XLOCKED);
301 TAILQ_INSERT_TAIL(&dotdot->de_dlist, dd, de_list);
303 devfs_rules_apply(dmp, dd);
307 mac_devfs_create_directory(dmp->dm_mount, name, namelen, dd);
313 devfs_dirent_free(struct devfs_dirent *de)
318 mtx_lock(&devfs_de_interlock);
319 if (vp != NULL && vp->v_data == de)
321 mtx_unlock(&devfs_de_interlock);
326 * Removes a directory if it is empty. Also empty parent directories are
327 * removed recursively.
330 devfs_rmdir_empty(struct devfs_mount *dm, struct devfs_dirent *de)
332 struct devfs_dirent *dd, *de_dot, *de_dotdot;
334 sx_assert(&dm->dm_lock, SX_XLOCKED);
337 KASSERT(de->de_dirent->d_type == DT_DIR,
338 ("devfs_rmdir_empty: de is not a directory"));
340 if ((de->de_flags & DE_DOOMED) != 0 || de == dm->dm_rootdir)
343 de_dot = TAILQ_FIRST(&de->de_dlist);
344 KASSERT(de_dot != NULL, ("devfs_rmdir_empty: . missing"));
345 de_dotdot = TAILQ_NEXT(de_dot, de_list);
346 KASSERT(de_dotdot != NULL, ("devfs_rmdir_empty: .. missing"));
347 /* Return if the directory is not empty. */
348 if (TAILQ_NEXT(de_dotdot, de_list) != NULL)
351 dd = devfs_parent_dirent(de);
352 KASSERT(dd != NULL, ("devfs_rmdir_empty: NULL dd"));
353 TAILQ_REMOVE(&de->de_dlist, de_dot, de_list);
354 TAILQ_REMOVE(&de->de_dlist, de_dotdot, de_list);
355 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
357 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
358 devfs_delete(dm, de_dot, DEVFS_DEL_NORECURSE);
359 devfs_delete(dm, de_dotdot, DEVFS_DEL_NORECURSE);
360 if (DEVFS_DE_DROP(dd)) {
361 devfs_dirent_free(dd);
370 * The caller needs to hold the dm for the duration of the call since
371 * dm->dm_lock may be temporary dropped.
374 devfs_delete(struct devfs_mount *dm, struct devfs_dirent *de, int flags)
376 struct devfs_dirent *dd;
379 KASSERT((de->de_flags & DE_DOOMED) == 0,
380 ("devfs_delete doomed dirent"));
381 de->de_flags |= DE_DOOMED;
383 if ((flags & DEVFS_DEL_NORECURSE) == 0) {
384 dd = devfs_parent_dirent(de);
387 if (de->de_flags & DE_USER) {
388 KASSERT(dd != NULL, ("devfs_delete: NULL dd"));
389 devfs_dir_unref_de(dm, dd);
394 mtx_lock(&devfs_de_interlock);
398 mtx_unlock(&devfs_de_interlock);
400 sx_unlock(&dm->dm_lock);
401 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
402 vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
406 if ((flags & DEVFS_DEL_VNLOCKED) == 0)
409 sx_xlock(&dm->dm_lock);
411 mtx_unlock(&devfs_de_interlock);
412 if (de->de_symlink) {
413 free(de->de_symlink, M_DEVFS);
414 de->de_symlink = NULL;
417 mac_devfs_destroy(de);
419 if (de->de_inode > DEVFS_ROOTINO) {
420 devfs_free_cdp_inode(de->de_inode);
423 if (DEVFS_DE_DROP(de))
424 devfs_dirent_free(de);
427 if (DEVFS_DE_DROP(dd))
428 devfs_dirent_free(dd);
430 devfs_rmdir_empty(dm, dd);
436 * Recursively removes the entire tree.
437 * The caller needs to hold the dm for the duration of the call.
441 devfs_purge(struct devfs_mount *dm, struct devfs_dirent *dd)
443 struct devfs_dirent *de;
445 sx_assert(&dm->dm_lock, SX_XLOCKED);
450 * Use TAILQ_LAST() to remove "." and ".." last.
451 * We might need ".." to resolve a path in
452 * devfs_dir_unref_de().
454 de = TAILQ_LAST(&dd->de_dlist, devfs_dlist_head);
457 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
458 if (de->de_flags & DE_USER)
459 devfs_dir_unref_de(dm, dd);
460 if (de->de_flags & (DE_DOT | DE_DOTDOT))
461 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
462 else if (de->de_dirent->d_type == DT_DIR)
465 devfs_delete(dm, de, DEVFS_DEL_NORECURSE);
467 if (DEVFS_DE_DROP(dd))
468 devfs_dirent_free(dd);
469 else if ((dd->de_flags & DE_DOOMED) == 0)
470 devfs_delete(dm, dd, DEVFS_DEL_NORECURSE);
474 * Each cdev_priv has an array of pointers to devfs_dirent which is indexed
475 * by the mount points dm_idx.
476 * This function extends the array when necessary, taking into account that
477 * the default array is 1 element and not malloc'ed.
480 devfs_metoo(struct cdev_priv *cdp, struct devfs_mount *dm)
482 struct devfs_dirent **dep;
485 siz = (dm->dm_idx + 1) * sizeof *dep;
486 dep = malloc(siz, M_DEVFS2, M_WAITOK | M_ZERO);
488 if (dm->dm_idx <= cdp->cdp_maxdirent) {
494 memcpy(dep, cdp->cdp_dirents, (cdp->cdp_maxdirent + 1) * sizeof *dep);
495 if (cdp->cdp_maxdirent > 0)
496 free(cdp->cdp_dirents, M_DEVFS2);
497 cdp->cdp_dirents = dep;
499 * XXX: if malloc told us how much we actually got this could
502 cdp->cdp_maxdirent = dm->dm_idx;
507 * The caller needs to hold the dm for the duration of the call.
510 devfs_populate_loop(struct devfs_mount *dm, int cleanup)
512 struct cdev_priv *cdp;
513 struct devfs_dirent *de;
514 struct devfs_dirent *dd, *dt;
516 int de_flags, depth, j;
519 sx_assert(&dm->dm_lock, SX_XLOCKED);
521 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
523 KASSERT(cdp->cdp_dirents != NULL, ("NULL cdp_dirents"));
526 * If we are unmounting, or the device has been destroyed,
527 * clean up our dirent.
529 if ((cleanup || !(cdp->cdp_flags & CDP_ACTIVE)) &&
530 dm->dm_idx <= cdp->cdp_maxdirent &&
531 cdp->cdp_dirents[dm->dm_idx] != NULL) {
532 de = cdp->cdp_dirents[dm->dm_idx];
533 cdp->cdp_dirents[dm->dm_idx] = NULL;
534 KASSERT(cdp == de->de_cdp,
535 ("%s %d %s %p %p", __func__, __LINE__,
536 cdp->cdp_c.si_name, cdp, de->de_cdp));
537 KASSERT(de->de_dir != NULL, ("Null de->de_dir"));
540 TAILQ_REMOVE(&de->de_dir->de_dlist, de, de_list);
543 devfs_delete(dm, de, 0);
550 * GC any lingering devices
552 if (!(cdp->cdp_flags & CDP_ACTIVE)) {
553 if (cdp->cdp_inuse > 0)
555 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
557 dev_rel(&cdp->cdp_c);
561 * Don't create any new dirents if we are unmounting
565 KASSERT((cdp->cdp_flags & CDP_ACTIVE), ("Bogons, I tell ya'!"));
567 if (dm->dm_idx <= cdp->cdp_maxdirent &&
568 cdp->cdp_dirents[dm->dm_idx] != NULL) {
569 de = cdp->cdp_dirents[dm->dm_idx];
570 KASSERT(cdp == de->de_cdp, ("inconsistent cdp"));
578 if (dm->dm_idx > cdp->cdp_maxdirent)
579 devfs_metoo(cdp, dm);
582 s = cdp->cdp_c.si_name;
584 for (q = s; *q != '/' && *q != '\0'; q++)
588 de = devfs_find(dd, s, q - s, 0);
590 de = devfs_vmkdir(dm, s, q - s, dd, 0);
591 else if (de->de_dirent->d_type == DT_LNK) {
592 de = devfs_find(dd, s, q - s, DT_DIR);
594 de = devfs_vmkdir(dm, s, q - s, dd, 0);
595 de->de_flags |= DE_COVERED;
599 KASSERT(dd->de_dirent->d_type == DT_DIR &&
600 (dd->de_flags & (DE_DOT | DE_DOTDOT)) == 0,
601 ("%s: invalid directory (si_name=%s)",
602 __func__, cdp->cdp_c.si_name));
606 de = devfs_find(dd, s, q - s, DT_LNK);
608 de_flags |= DE_COVERED;
610 de = devfs_newdirent(s, q - s);
611 if (cdp->cdp_c.si_flags & SI_ALIAS) {
615 de->de_dirent->d_type = DT_LNK;
616 pdev = cdp->cdp_c.si_parent;
619 while (dt != dm->dm_rootdir &&
620 (dt = devfs_parent_dirent(dt)) != NULL)
622 j = depth * 3 + strlen(pdev->si_name) + 1;
623 de->de_symlink = malloc(j, M_DEVFS, M_WAITOK);
624 de->de_symlink[0] = 0;
626 strcat(de->de_symlink, "../");
627 strcat(de->de_symlink, pdev->si_name);
629 de->de_uid = cdp->cdp_c.si_uid;
630 de->de_gid = cdp->cdp_c.si_gid;
631 de->de_mode = cdp->cdp_c.si_mode;
632 de->de_dirent->d_type = DT_CHR;
634 de->de_flags |= de_flags;
635 de->de_inode = cdp->cdp_inode;
638 mac_devfs_create_device(cdp->cdp_c.si_cred, dm->dm_mount,
642 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
643 devfs_rules_apply(dm, de);
645 /* XXX: could check that cdp is still active here */
646 KASSERT(cdp->cdp_dirents[dm->dm_idx] == NULL,
647 ("%s %d\n", __func__, __LINE__));
648 cdp->cdp_dirents[dm->dm_idx] = de;
649 KASSERT(de->de_cdp != (void *)0xdeadc0de,
650 ("%s %d\n", __func__, __LINE__));
659 * The caller needs to hold the dm for the duration of the call.
662 devfs_populate(struct devfs_mount *dm)
666 sx_assert(&dm->dm_lock, SX_XLOCKED);
667 gen = devfs_generation;
668 if (dm->dm_generation == gen)
670 while (devfs_populate_loop(dm, 0))
672 dm->dm_generation = gen;
676 * The caller needs to hold the dm for the duration of the call.
679 devfs_cleanup(struct devfs_mount *dm)
682 sx_assert(&dm->dm_lock, SX_XLOCKED);
683 while (devfs_populate_loop(dm, 1))
685 devfs_purge(dm, dm->dm_rootdir);
689 * devfs_create() and devfs_destroy() are called from kern_conf.c and
690 * in both cases the devlock() mutex is held, so no further locking
691 * is necessary and no sleeping allowed.
695 devfs_create(struct cdev *dev)
697 struct cdev_priv *cdp;
699 mtx_assert(&devmtx, MA_OWNED);
700 cdp = cdev2priv(dev);
701 cdp->cdp_flags |= CDP_ACTIVE;
702 cdp->cdp_inode = alloc_unrl(devfs_inos);
704 TAILQ_INSERT_TAIL(&cdevp_list, cdp, cdp_list);
709 devfs_destroy(struct cdev *dev)
711 struct cdev_priv *cdp;
713 mtx_assert(&devmtx, MA_OWNED);
714 cdp = cdev2priv(dev);
715 cdp->cdp_flags &= ~CDP_ACTIVE;
720 devfs_alloc_cdp_inode(void)
723 return (alloc_unr(devfs_inos));
727 devfs_free_cdp_inode(ino_t ino)
731 free_unr(devfs_inos, ino);
735 devfs_devs_init(void *junk __unused)
738 devfs_inos = new_unrhdr(DEVFS_ROOTINO + 1, INT_MAX, &devmtx);
741 SYSINIT(devfs_devs, SI_SUB_DEVFS, SI_ORDER_FIRST, devfs_devs_init, NULL);