1 /*-
2 * Copyright (c) 2000,2004
3 * Poul-Henning Kamp. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
13 *
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
24 * SUCH DAMAGE.
25 *
26 * From: FreeBSD: src/sys/miscfs/kernfs/kernfs_vfsops.c 1.36
27 *
28 * $FreeBSD: releng/8.4/sys/fs/devfs/devfs_devs.c 224904 2011-08-16 08:41:09Z kib $
29 */
30
31 #include <sys/param.h>
32 #include <sys/systm.h>
33 #include <sys/conf.h>
34 #include <sys/dirent.h>
35 #include <sys/kernel.h>
36 #include <sys/limits.h>
37 #include <sys/lock.h>
38 #include <sys/malloc.h>
39 #include <sys/proc.h>
40 #include <sys/sx.h>
41 #include <sys/sysctl.h>
42 #include <sys/vnode.h>
43
44 #include <sys/kdb.h>
45
46 #include <fs/devfs/devfs.h>
47 #include <fs/devfs/devfs_int.h>
48
49 #include <security/mac/mac_framework.h>
50
51 /*
52 * The one true (but secret) list of active devices in the system.
53 * Locked by dev_lock()/devmtx
54 */
55 struct cdev_priv_list cdevp_list = TAILQ_HEAD_INITIALIZER(cdevp_list);
56
57 struct unrhdr *devfs_inos;
58
59
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");
63
64 static SYSCTL_NODE(_vfs, OID_AUTO, devfs, CTLFLAG_RW, 0, "DEVFS filesystem");
65
66 static unsigned devfs_generation;
67 SYSCTL_UINT(_vfs_devfs, OID_AUTO, generation, CTLFLAG_RD,
68 &devfs_generation, 0, "DEVFS generation number");
69
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");
73
74 /*
75 * Helper sysctl for devname(3). We're given a struct cdev * and return
76 * the name, if any, registered by the device driver.
77 */
78 static int
79 sysctl_devname(SYSCTL_HANDLER_ARGS)
80 {
81 int error;
82 dev_t ud;
83 struct cdev_priv *cdp;
84
85 error = SYSCTL_IN(req, &ud, sizeof (ud));
86 if (error)
87 return (error);
88 if (ud == NODEV)
89 return(EINVAL);
90 /*
91 ud ^ devfs_random();
92 */
93 dev_lock();
94 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list)
95 if (cdp->cdp_inode == ud)
96 break;
97 dev_unlock();
98 if (cdp == NULL)
99 return(ENOENT);
100 return(SYSCTL_OUT(req, cdp->cdp_c.si_name, strlen(cdp->cdp_c.si_name) + 1));
101 return (error);
102 }
103
104 SYSCTL_PROC(_kern, OID_AUTO, devname,
105 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
106 NULL, 0, sysctl_devname, "", "devname(3) handler");
107
108 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev, CTLFLAG_RD,
109 0, sizeof(struct cdev), "sizeof(struct cdev)");
110
111 SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev_priv, CTLFLAG_RD,
112 0, sizeof(struct cdev_priv), "sizeof(struct cdev_priv)");
113
114 struct cdev *
115 devfs_alloc(int flags)
116 {
117 struct cdev_priv *cdp;
118 struct cdev *cdev;
119 struct timespec ts;
120
121 cdp = malloc(sizeof *cdp, M_CDEVP, M_USE_RESERVE | M_ZERO |
122 ((flags & MAKEDEV_NOWAIT) ? M_NOWAIT : M_WAITOK));
123 if (cdp == NULL)
124 return (NULL);
125
126 cdp->cdp_dirents = &cdp->cdp_dirent0;
127 cdp->cdp_dirent0 = NULL;
128 cdp->cdp_maxdirent = 0;
129 cdp->cdp_inode = 0;
130
131 cdev = &cdp->cdp_c;
132
133 cdev->si_name = cdev->__si_namebuf;
134 LIST_INIT(&cdev->si_children);
135 vfs_timestamp(&ts);
136 cdev->si_atime = cdev->si_mtime = cdev->si_ctime = ts;
137 cdev->si_cred = NULL;
138
139 return (cdev);
140 }
141
142 void
143 devfs_free(struct cdev *cdev)
144 {
145 struct cdev_priv *cdp;
146
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);
154 free(cdp, M_CDEVP);
155 }
156
157 struct devfs_dirent *
158 devfs_find(struct devfs_dirent *dd, const char *name, int namelen)
159 {
160 struct devfs_dirent *de;
161
162 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
163 if (namelen != de->de_dirent->d_namlen)
164 continue;
165 if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
166 continue;
167 break;
168 }
169 return (de);
170 }
171
172 struct devfs_dirent *
173 devfs_newdirent(char *name, int namelen)
174 {
175 int i;
176 struct devfs_dirent *de;
177 struct dirent d;
178
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;
189 de->de_links = 1;
190 de->de_holdcnt = 1;
191 #ifdef MAC
192 mac_devfs_init(de);
193 #endif
194 return (de);
195 }
196
197 struct devfs_dirent *
198 devfs_parent_dirent(struct devfs_dirent *de)
199 {
200
201 if (de->de_dirent->d_type != DT_DIR)
202 return (de->de_dir);
203
204 if (de->de_flags & (DE_DOT | DE_DOTDOT))
205 return (NULL);
206
207 de = TAILQ_FIRST(&de->de_dlist); /* "." */
208 if (de == NULL)
209 return (NULL);
210 de = TAILQ_NEXT(de, de_list); /* ".." */
211 if (de == NULL)
212 return (NULL);
213
214 return (de->de_dir);
215 }
216
217 struct devfs_dirent *
218 devfs_vmkdir(struct devfs_mount *dmp, char *name, int namelen, struct devfs_dirent *dotdot, u_int inode)
219 {
220 struct devfs_dirent *dd;
221 struct devfs_dirent *de;
222
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;
227 dd->de_mode = 0555;
228 dd->de_links = 2;
229 dd->de_dir = dd;
230 if (inode != 0)
231 dd->de_inode = inode;
232 else
233 dd->de_inode = alloc_unr(devfs_inos);
234
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);
240 de->de_dir = dd;
241
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) {
248 de->de_dir = dd;
249 } else {
250 de->de_dir = dotdot;
251 sx_assert(&dmp->dm_lock, SX_XLOCKED);
252 TAILQ_INSERT_TAIL(&dotdot->de_dlist, dd, de_list);
253 dotdot->de_links++;
254 devfs_rules_apply(dmp, dd);
255 }
256
257 #ifdef MAC
258 mac_devfs_create_directory(dmp->dm_mount, name, namelen, dd);
259 #endif
260 return (dd);
261 }
262
263 void
264 devfs_dirent_free(struct devfs_dirent *de)
265 {
266 free(de, M_DEVFS3);
267 }
268
269 /*
270 * The caller needs to hold the dm for the duration of the call since
271 * dm->dm_lock may be temporary dropped.
272 */
273 void
274 devfs_delete(struct devfs_mount *dm, struct devfs_dirent *de, int vp_locked)
275 {
276 struct vnode *vp;
277
278 KASSERT((de->de_flags & DE_DOOMED) == 0,
279 ("devfs_delete doomed dirent"));
280 de->de_flags |= DE_DOOMED;
281 mtx_lock(&devfs_de_interlock);
282 vp = de->de_vnode;
283 if (vp != NULL) {
284 VI_LOCK(vp);
285 mtx_unlock(&devfs_de_interlock);
286 vholdl(vp);
287 sx_unlock(&dm->dm_lock);
288 if (!vp_locked)
289 vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
290 else
291 VI_UNLOCK(vp);
292 vgone(vp);
293 if (!vp_locked)
294 VOP_UNLOCK(vp, 0);
295 vdrop(vp);
296 sx_xlock(&dm->dm_lock);
297 } else
298 mtx_unlock(&devfs_de_interlock);
299 if (de->de_symlink) {
300 free(de->de_symlink, M_DEVFS);
301 de->de_symlink = NULL;
302 }
303 #ifdef MAC
304 mac_devfs_destroy(de);
305 #endif
306 if (de->de_inode > DEVFS_ROOTINO) {
307 free_unr(devfs_inos, de->de_inode);
308 de->de_inode = 0;
309 }
310 if (DEVFS_DE_DROP(de))
311 devfs_dirent_free(de);
312 }
313
314 /*
315 * Called on unmount.
316 * Recursively removes the entire tree.
317 * The caller needs to hold the dm for the duration of the call.
318 */
319
320 static void
321 devfs_purge(struct devfs_mount *dm, struct devfs_dirent *dd)
322 {
323 struct devfs_dirent *de;
324
325 sx_assert(&dm->dm_lock, SX_XLOCKED);
326 for (;;) {
327 de = TAILQ_FIRST(&dd->de_dlist);
328 if (de == NULL)
329 break;
330 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
331 if (de->de_flags & (DE_DOT|DE_DOTDOT))
332 devfs_delete(dm, de, 0);
333 else if (de->de_dirent->d_type == DT_DIR)
334 devfs_purge(dm, de);
335 else
336 devfs_delete(dm, de, 0);
337 }
338 devfs_delete(dm, dd, 0);
339 }
340
341 /*
342 * Each cdev_priv has an array of pointers to devfs_dirent which is indexed
343 * by the mount points dm_idx.
344 * This function extends the array when necessary, taking into account that
345 * the default array is 1 element and not malloc'ed.
346 */
347 static void
348 devfs_metoo(struct cdev_priv *cdp, struct devfs_mount *dm)
349 {
350 struct devfs_dirent **dep;
351 int siz;
352
353 siz = (dm->dm_idx + 1) * sizeof *dep;
354 dep = malloc(siz, M_DEVFS2, M_WAITOK | M_ZERO);
355 dev_lock();
356 if (dm->dm_idx <= cdp->cdp_maxdirent) {
357 /* We got raced */
358 dev_unlock();
359 free(dep, M_DEVFS2);
360 return;
361 }
362 memcpy(dep, cdp->cdp_dirents, (cdp->cdp_maxdirent + 1) * sizeof *dep);
363 if (cdp->cdp_maxdirent > 0)
364 free(cdp->cdp_dirents, M_DEVFS2);
365 cdp->cdp_dirents = dep;
366 /*
367 * XXX: if malloc told us how much we actually got this could
368 * XXX: be optimized.
369 */
370 cdp->cdp_maxdirent = dm->dm_idx;
371 dev_unlock();
372 }
373
374 /*
375 * The caller needs to hold the dm for the duration of the call.
376 */
377 static int
378 devfs_populate_loop(struct devfs_mount *dm, int cleanup)
379 {
380 struct cdev_priv *cdp;
381 struct devfs_dirent *de;
382 struct devfs_dirent *dd;
383 struct cdev *pdev;
384 int j;
385 char *q, *s;
386
387 sx_assert(&dm->dm_lock, SX_XLOCKED);
388 dev_lock();
389 TAILQ_FOREACH(cdp, &cdevp_list, cdp_list) {
390
391 KASSERT(cdp->cdp_dirents != NULL, ("NULL cdp_dirents"));
392
393 /*
394 * If we are unmounting, or the device has been destroyed,
395 * clean up our dirent.
396 */
397 if ((cleanup || !(cdp->cdp_flags & CDP_ACTIVE)) &&
398 dm->dm_idx <= cdp->cdp_maxdirent &&
399 cdp->cdp_dirents[dm->dm_idx] != NULL) {
400 de = cdp->cdp_dirents[dm->dm_idx];
401 cdp->cdp_dirents[dm->dm_idx] = NULL;
402 KASSERT(cdp == de->de_cdp,
403 ("%s %d %s %p %p", __func__, __LINE__,
404 cdp->cdp_c.si_name, cdp, de->de_cdp));
405 KASSERT(de->de_dir != NULL, ("Null de->de_dir"));
406 dev_unlock();
407
408 TAILQ_REMOVE(&de->de_dir->de_dlist, de, de_list);
409 de->de_cdp = NULL;
410 de->de_inode = 0;
411 devfs_delete(dm, de, 0);
412 dev_lock();
413 cdp->cdp_inuse--;
414 dev_unlock();
415 return (1);
416 }
417 /*
418 * GC any lingering devices
419 */
420 if (!(cdp->cdp_flags & CDP_ACTIVE)) {
421 if (cdp->cdp_inuse > 0)
422 continue;
423 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
424 dev_unlock();
425 dev_rel(&cdp->cdp_c);
426 return (1);
427 }
428 /*
429 * Don't create any new dirents if we are unmounting
430 */
431 if (cleanup)
432 continue;
433 KASSERT((cdp->cdp_flags & CDP_ACTIVE), ("Bogons, I tell ya'!"));
434
435 if (dm->dm_idx <= cdp->cdp_maxdirent &&
436 cdp->cdp_dirents[dm->dm_idx] != NULL) {
437 de = cdp->cdp_dirents[dm->dm_idx];
438 KASSERT(cdp == de->de_cdp, ("inconsistent cdp"));
439 continue;
440 }
441
442
443 cdp->cdp_inuse++;
444 dev_unlock();
445
446 if (dm->dm_idx > cdp->cdp_maxdirent)
447 devfs_metoo(cdp, dm);
448
449 dd = dm->dm_rootdir;
450 s = cdp->cdp_c.si_name;
451 for (;;) {
452 for (q = s; *q != '/' && *q != '\0'; q++)
453 continue;
454 if (*q != '/')
455 break;
456 de = devfs_find(dd, s, q - s);
457 if (de == NULL)
458 de = devfs_vmkdir(dm, s, q - s, dd, 0);
459 s = q + 1;
460 dd = de;
461 }
462
463 de = devfs_newdirent(s, q - s);
464 if (cdp->cdp_c.si_flags & SI_ALIAS) {
465 de->de_uid = 0;
466 de->de_gid = 0;
467 de->de_mode = 0755;
468 de->de_dirent->d_type = DT_LNK;
469 pdev = cdp->cdp_c.si_parent;
470 j = strlen(pdev->si_name) + 1;
471 de->de_symlink = malloc(j, M_DEVFS, M_WAITOK);
472 bcopy(pdev->si_name, de->de_symlink, j);
473 } else {
474 de->de_uid = cdp->cdp_c.si_uid;
475 de->de_gid = cdp->cdp_c.si_gid;
476 de->de_mode = cdp->cdp_c.si_mode;
477 de->de_dirent->d_type = DT_CHR;
478 }
479 de->de_inode = cdp->cdp_inode;
480 de->de_cdp = cdp;
481 #ifdef MAC
482 mac_devfs_create_device(cdp->cdp_c.si_cred, dm->dm_mount,
483 &cdp->cdp_c, de);
484 #endif
485 de->de_dir = dd;
486 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
487 devfs_rules_apply(dm, de);
488 dev_lock();
489 /* XXX: could check that cdp is still active here */
490 KASSERT(cdp->cdp_dirents[dm->dm_idx] == NULL,
491 ("%s %d\n", __func__, __LINE__));
492 cdp->cdp_dirents[dm->dm_idx] = de;
493 KASSERT(de->de_cdp != (void *)0xdeadc0de,
494 ("%s %d\n", __func__, __LINE__));
495 dev_unlock();
496 return (1);
497 }
498 dev_unlock();
499 return (0);
500 }
501
502 /*
503 * The caller needs to hold the dm for the duration of the call.
504 */
505 void
506 devfs_populate(struct devfs_mount *dm)
507 {
508 unsigned gen;
509
510 sx_assert(&dm->dm_lock, SX_XLOCKED);
511 gen = devfs_generation;
512 if (dm->dm_generation == gen)
513 return;
514 while (devfs_populate_loop(dm, 0))
515 continue;
516 dm->dm_generation = gen;
517 }
518
519 /*
520 * The caller needs to hold the dm for the duration of the call.
521 */
522 void
523 devfs_cleanup(struct devfs_mount *dm)
524 {
525
526 sx_assert(&dm->dm_lock, SX_XLOCKED);
527 while (devfs_populate_loop(dm, 1))
528 continue;
529 devfs_purge(dm, dm->dm_rootdir);
530 }
531
532 /*
533 * devfs_create() and devfs_destroy() are called from kern_conf.c and
534 * in both cases the devlock() mutex is held, so no further locking
535 * is necesary and no sleeping allowed.
536 */
537
538 void
539 devfs_create(struct cdev *dev)
540 {
541 struct cdev_priv *cdp;
542
543 mtx_assert(&devmtx, MA_OWNED);
544 cdp = cdev2priv(dev);
545 cdp->cdp_flags |= CDP_ACTIVE;
546 cdp->cdp_inode = alloc_unrl(devfs_inos);
547 dev_refl(dev);
548 TAILQ_INSERT_TAIL(&cdevp_list, cdp, cdp_list);
549 devfs_generation++;
550 }
551
552 void
553 devfs_destroy(struct cdev *dev)
554 {
555 struct cdev_priv *cdp;
556
557 mtx_assert(&devmtx, MA_OWNED);
558 cdp = cdev2priv(dev);
559 cdp->cdp_flags &= ~CDP_ACTIVE;
560 devfs_generation++;
561 }
562
563 static void
564 devfs_devs_init(void *junk __unused)
565 {
566
567 devfs_inos = new_unrhdr(DEVFS_ROOTINO + 1, INT_MAX, &devmtx);
568 }
569
570 SYSINIT(devfs_devs, SI_SUB_DEVFS, SI_ORDER_FIRST, devfs_devs_init, NULL);
Cache object: 17f209742413e900867a9b36f66a06ab
|