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 struct cdev * and return
76 * the name, if any, registered by the device driver.
79 sysctl_devname(SYSCTL_HANDLER_ARGS)
83 struct cdev_priv *cdp;
85 error = SYSCTL_IN(req, &ud, sizeof (ud));
94 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list)
95 if (cdp->cdp_inode == ud)
100 return(SYSCTL_OUT(req, cdp->cdp_c.si_name, strlen(cdp->cdp_c.si_name) + 1));
104 SYSCTL_PROC(_kern, OID_AUTO, devname,
105 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
106 NULL, 0, sysctl_devname, "", "devname(3) handler");
108 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev, CTLFLAG_RD,
109 0, sizeof(struct cdev), "sizeof(struct cdev)");
111 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev_priv, CTLFLAG_RD,
112 0, sizeof(struct cdev_priv), "sizeof(struct cdev_priv)");
115 devfs_alloc(int flags)
117 struct cdev_priv *cdp;
121 cdp = malloc(sizeof *cdp, M_CDEVP, M_USE_RESERVE | M_ZERO |
122 ((flags & MAKEDEV_NOWAIT) ? M_NOWAIT : M_WAITOK));
126 cdp->cdp_dirents = &cdp->cdp_dirent0;
127 cdp->cdp_dirent0 = NULL;
128 cdp->cdp_maxdirent = 0;
133 cdev->si_name = cdev->__si_namebuf;
134 LIST_INIT(&cdev->si_children);
136 cdev->si_atime = cdev->si_mtime = cdev->si_ctime = ts;
137 cdev->si_cred = NULL;
143 devfs_free(struct cdev *cdev)
145 struct cdev_priv *cdp;
147 cdp = cdev2priv(cdev);
148 if (cdev->si_cred != NULL)
149 crfree(cdev->si_cred);
150 if (cdp->cdp_inode > 0)
151 free_unr(devfs_inos, cdp->cdp_inode);
152 if (cdp->cdp_maxdirent > 0)
153 free(cdp->cdp_dirents, M_DEVFS2);
157 struct devfs_dirent *
158 devfs_find(struct devfs_dirent *dd, const char *name, int namelen)
160 struct devfs_dirent *de;
162 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
163 if (namelen != de->de_dirent->d_namlen)
165 if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
172 struct devfs_dirent *
173 devfs_newdirent(char *name, int namelen)
176 struct devfs_dirent *de;
179 d.d_namlen = namelen;
180 i = sizeof (*de) + GENERIC_DIRSIZ(&d);
181 de = malloc(i, M_DEVFS3, M_WAITOK | M_ZERO);
182 de->de_dirent = (struct dirent *)(de + 1);
183 de->de_dirent->d_namlen = namelen;
184 de->de_dirent->d_reclen = GENERIC_DIRSIZ(&d);
185 bcopy(name, de->de_dirent->d_name, namelen);
186 de->de_dirent->d_name[namelen] = '\0';
187 vfs_timestamp(&de->de_ctime);
188 de->de_mtime = de->de_atime = de->de_ctime;
197 struct devfs_dirent *
198 devfs_parent_dirent(struct devfs_dirent *de)
201 if (de->de_dirent->d_type != DT_DIR)
204 if (de->de_flags & (DE_DOT | DE_DOTDOT))
207 de = TAILQ_FIRST(&de->de_dlist); /* "." */
210 de = TAILQ_NEXT(de, de_list); /* ".." */
217 struct devfs_dirent *
218 devfs_vmkdir(struct devfs_mount *dmp, char *name, int namelen, struct devfs_dirent *dotdot, u_int inode)
220 struct devfs_dirent *dd;
221 struct devfs_dirent *de;
223 /* Create the new directory */
224 dd = devfs_newdirent(name, namelen);
225 TAILQ_INIT(&dd->de_dlist);
226 dd->de_dirent->d_type = DT_DIR;
231 dd->de_inode = inode;
233 dd->de_inode = alloc_unr(devfs_inos);
235 /* Create the "." entry in the new directory */
236 de = devfs_newdirent(".", 1);
237 de->de_dirent->d_type = DT_DIR;
238 de->de_flags |= DE_DOT;
239 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
242 /* Create the ".." entry in the new directory */
243 de = devfs_newdirent("..", 2);
244 de->de_dirent->d_type = DT_DIR;
245 de->de_flags |= DE_DOTDOT;
246 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
247 if (dotdot == NULL) {
251 TAILQ_INSERT_TAIL(&dotdot->de_dlist, dd, de_list);
256 mac_devfs_create_directory(dmp->dm_mount, name, namelen, dd);
262 devfs_dirent_free(struct devfs_dirent *de)
268 * The caller needs to hold the dm for the duration of the call since
269 * dm->dm_lock may be temporary dropped.
272 devfs_delete(struct devfs_mount *dm, struct devfs_dirent *de, int vp_locked)
276 KASSERT((de->de_flags & DE_DOOMED) == 0,
277 ("devfs_delete doomed dirent"));
278 de->de_flags |= DE_DOOMED;
279 mtx_lock(&devfs_de_interlock);
283 mtx_unlock(&devfs_de_interlock);
285 sx_unlock(&dm->dm_lock);
287 vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
294 sx_xlock(&dm->dm_lock);
296 mtx_unlock(&devfs_de_interlock);
297 if (de->de_symlink) {
298 free(de->de_symlink, M_DEVFS);
299 de->de_symlink = NULL;
302 mac_devfs_destroy(de);
304 if (de->de_inode > DEVFS_ROOTINO) {
305 free_unr(devfs_inos, de->de_inode);
308 if (DEVFS_DE_DROP(de))
309 devfs_dirent_free(de);
314 * Recursively removes the entire tree.
315 * The caller needs to hold the dm for the duration of the call.
319 devfs_purge(struct devfs_mount *dm, struct devfs_dirent *dd)
321 struct devfs_dirent *de;
323 sx_assert(&dm->dm_lock, SX_XLOCKED);
325 de = TAILQ_FIRST(&dd->de_dlist);
328 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
329 if (de->de_flags & (DE_DOT|DE_DOTDOT))
330 devfs_delete(dm, de, 0);
331 else if (de->de_dirent->d_type == DT_DIR)
334 devfs_delete(dm, de, 0);
336 devfs_delete(dm, dd, 0);
340 * Each cdev_priv has an array of pointers to devfs_dirent which is indexed
341 * by the mount points dm_idx.
342 * This function extends the array when necessary, taking into account that
343 * the default array is 1 element and not malloc'ed.
346 devfs_metoo(struct cdev_priv *cdp, struct devfs_mount *dm)
348 struct devfs_dirent **dep;
351 siz = (dm->dm_idx + 1) * sizeof *dep;
352 dep = malloc(siz, M_DEVFS2, M_WAITOK | M_ZERO);
354 if (dm->dm_idx <= cdp->cdp_maxdirent) {
360 memcpy(dep, cdp->cdp_dirents, (cdp->cdp_maxdirent + 1) * sizeof *dep);
361 if (cdp->cdp_maxdirent > 0)
362 free(cdp->cdp_dirents, M_DEVFS2);
363 cdp->cdp_dirents = dep;
365 * XXX: if malloc told us how much we actually got this could
368 cdp->cdp_maxdirent = dm->dm_idx;
373 * The caller needs to hold the dm for the duration of the call.
376 devfs_populate_loop(struct devfs_mount *dm, int cleanup)
378 struct cdev_priv *cdp;
379 struct devfs_dirent *de;
380 struct devfs_dirent *dd;
385 sx_assert(&dm->dm_lock, SX_XLOCKED);
387 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
389 KASSERT(cdp->cdp_dirents != NULL, ("NULL cdp_dirents"));
392 * If we are unmounting, or the device has been destroyed,
393 * clean up our dirent.
395 if ((cleanup || !(cdp->cdp_flags & CDP_ACTIVE)) &&
396 dm->dm_idx <= cdp->cdp_maxdirent &&
397 cdp->cdp_dirents[dm->dm_idx] != NULL) {
398 de = cdp->cdp_dirents[dm->dm_idx];
399 cdp->cdp_dirents[dm->dm_idx] = NULL;
400 KASSERT(cdp == de->de_cdp,
401 ("%s %d %s %p %p", __func__, __LINE__,
402 cdp->cdp_c.si_name, cdp, de->de_cdp));
403 KASSERT(de->de_dir != NULL, ("Null de->de_dir"));
406 TAILQ_REMOVE(&de->de_dir->de_dlist, de, de_list);
409 devfs_delete(dm, de, 0);
416 * GC any lingering devices
418 if (!(cdp->cdp_flags & CDP_ACTIVE)) {
419 if (cdp->cdp_inuse > 0)
421 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
423 dev_rel(&cdp->cdp_c);
427 * Don't create any new dirents if we are unmounting
431 KASSERT((cdp->cdp_flags & CDP_ACTIVE), ("Bogons, I tell ya'!"));
433 if (dm->dm_idx <= cdp->cdp_maxdirent &&
434 cdp->cdp_dirents[dm->dm_idx] != NULL) {
435 de = cdp->cdp_dirents[dm->dm_idx];
436 KASSERT(cdp == de->de_cdp, ("inconsistent cdp"));
444 if (dm->dm_idx > cdp->cdp_maxdirent)
445 devfs_metoo(cdp, dm);
448 s = cdp->cdp_c.si_name;
450 for (q = s; *q != '/' && *q != '\0'; q++)
454 de = devfs_find(dd, s, q - s);
456 de = devfs_vmkdir(dm, s, q - s, dd, 0);
461 de = devfs_newdirent(s, q - s);
462 if (cdp->cdp_c.si_flags & SI_ALIAS) {
466 de->de_dirent->d_type = DT_LNK;
467 pdev = cdp->cdp_c.si_parent;
468 j = strlen(pdev->si_name) + 1;
469 de->de_symlink = malloc(j, M_DEVFS, M_WAITOK);
470 bcopy(pdev->si_name, de->de_symlink, j);
472 de->de_uid = cdp->cdp_c.si_uid;
473 de->de_gid = cdp->cdp_c.si_gid;
474 de->de_mode = cdp->cdp_c.si_mode;
475 de->de_dirent->d_type = DT_CHR;
477 de->de_inode = cdp->cdp_inode;
480 mac_devfs_create_device(cdp->cdp_c.si_cred, dm->dm_mount,
484 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
485 devfs_rules_apply(dm, de);
487 /* XXX: could check that cdp is still active here */
488 KASSERT(cdp->cdp_dirents[dm->dm_idx] == NULL,
489 ("%s %d\n", __func__, __LINE__));
490 cdp->cdp_dirents[dm->dm_idx] = de;
491 KASSERT(de->de_cdp != (void *)0xdeadc0de,
492 ("%s %d\n", __func__, __LINE__));
501 * The caller needs to hold the dm for the duration of the call.
504 devfs_populate(struct devfs_mount *dm)
507 sx_assert(&dm->dm_lock, SX_XLOCKED);
508 if (dm->dm_generation == devfs_generation)
510 while (devfs_populate_loop(dm, 0))
512 dm->dm_generation = devfs_generation;
516 * The caller needs to hold the dm for the duration of the call.
519 devfs_cleanup(struct devfs_mount *dm)
522 sx_assert(&dm->dm_lock, SX_XLOCKED);
523 while (devfs_populate_loop(dm, 1))
525 devfs_purge(dm, dm->dm_rootdir);
529 * devfs_create() and devfs_destroy() are called from kern_conf.c and
530 * in both cases the devlock() mutex is held, so no further locking
531 * is necesary and no sleeping allowed.
535 devfs_create(struct cdev *dev)
537 struct cdev_priv *cdp;
539 mtx_assert(&devmtx, MA_OWNED);
540 cdp = cdev2priv(dev);
541 cdp->cdp_flags |= CDP_ACTIVE;
542 cdp->cdp_inode = alloc_unrl(devfs_inos);
544 TAILQ_INSERT_TAIL(&cdevp_list, cdp, cdp_list);
549 devfs_destroy(struct cdev *dev)
551 struct cdev_priv *cdp;
553 mtx_assert(&devmtx, MA_OWNED);
554 cdp = cdev2priv(dev);
555 cdp->cdp_flags &= ~CDP_ACTIVE;
560 devfs_devs_init(void *junk __unused)
563 devfs_inos = new_unrhdr(DEVFS_ROOTINO + 1, INT_MAX, &devmtx);
566 SYSINIT(devfs_devs, SI_SUB_DEVFS, SI_ORDER_FIRST, devfs_devs_init, NULL);