devfs_devs.c revision 150147 
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: head/sys/fs/devfs/devfs_devs.c 150147 2005-09-15 06:57:28Z phk $
29 */
30
31#include "opt_devfs.h"
32#include "opt_mac.h"
33
34#include <sys/param.h>
35#include <sys/systm.h>
36#include <sys/conf.h>
37#include <sys/dirent.h>
38#include <sys/kernel.h>
39#include <sys/lock.h>
40#include <sys/mac.h>
41#include <sys/malloc.h>
42#include <sys/proc.h>
43#include <sys/sysctl.h>
44#include <sys/vnode.h>
45
46#include <machine/atomic.h>
47
48#include <fs/devfs/devfs.h>
49#include <fs/devfs/devfs_int.h>
50
51static struct cdev *devfs_inot[NDEVFSINO];
52static struct cdev **devfs_overflow;
53static int devfs_ref[NDEVFSINO];
54static int *devfs_refoverflow;
55static int devfs_nextino = 3;
56static int devfs_numino;
57static int devfs_topino;
58static int devfs_noverflowwant = NDEVFSOVERFLOW;
59static int devfs_noverflow;
60static unsigned devfs_generation;
61
62static struct devfs_dirent *devfs_find (struct devfs_dirent *dd, const char *name, int namelen);
63
64static SYSCTL_NODE(_vfs, OID_AUTO, devfs, CTLFLAG_RW, 0, "DEVFS filesystem");
65SYSCTL_UINT(_vfs_devfs, OID_AUTO, noverflow, CTLFLAG_RW,
66	&devfs_noverflowwant, 0, "Size of DEVFS overflow table");
67SYSCTL_UINT(_vfs_devfs, OID_AUTO, generation, CTLFLAG_RD,
68	&devfs_generation, 0, "DEVFS generation number");
69SYSCTL_UINT(_vfs_devfs, OID_AUTO, inodes, CTLFLAG_RD,
70	&devfs_numino, 0, "DEVFS inodes");
71SYSCTL_UINT(_vfs_devfs, OID_AUTO, topinode, CTLFLAG_RD,
72	&devfs_topino, 0, "DEVFS highest inode#");
73
74unsigned devfs_rule_depth = 1;
75SYSCTL_UINT(_vfs_devfs, OID_AUTO, rule_depth, CTLFLAG_RW,
76    &devfs_rule_depth, 0, "Max depth of ruleset include");
77
78/*
79 * Helper sysctl for devname(3).  We're given a struct cdev * and return
80 * the name, if any, registered by the device driver.
81 */
82static int
83sysctl_devname(SYSCTL_HANDLER_ARGS)
84{
85	int error;
86	dev_t ud;
87	struct cdev *dev, **dp;
88
89	error = SYSCTL_IN(req, &ud, sizeof (ud));
90	if (error)
91		return (error);
92	if (ud == NODEV)
93		return(EINVAL);
94	dp = devfs_itod(ud);
95	if (dp == NULL)
96		return(ENOENT);
97	dev = *dp;
98	if (dev == NULL)
99		return(ENOENT);
100	return(SYSCTL_OUT(req, dev->si_name, strlen(dev->si_name) + 1));
101	return (error);
102}
103
104SYSCTL_PROC(_kern, OID_AUTO, devname, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_ANYBODY,
105	NULL, 0, sysctl_devname, "", "devname(3) handler");
106
107SYSCTL_INT(_debug_sizeof, OID_AUTO, cdev, CTLFLAG_RD,
108    0, sizeof(struct cdev), "sizeof(struct cdev)");
109
110static int *
111devfs_itor(int inode)
112{
113	if (inode < NDEVFSINO)
114		return (&devfs_ref[inode]);
115	else if (inode < NDEVFSINO + devfs_noverflow)
116		return (&devfs_refoverflow[inode - NDEVFSINO]);
117	else
118		panic ("YRK!");
119}
120
121static void
122devfs_dropref(int inode)
123{
124	int *ip;
125
126	ip = devfs_itor(inode);
127	atomic_add_int(ip, -1);
128}
129
130static int
131devfs_getref(int inode)
132{
133	int *ip, i, j;
134	struct cdev **dp;
135
136	ip = devfs_itor(inode);
137	dp = devfs_itod(inode);
138	for (;;) {
139		i = *ip;
140		j = i + 1;
141		if (!atomic_cmpset_int(ip, i, j))
142			continue;
143		if (*dp != NULL)
144			return (1);
145		atomic_add_int(ip, -1);
146		return(0);
147	}
148}
149
150struct devfs_dirent **
151devfs_itode (struct devfs_mount *dm, int inode)
152{
153
154	if (inode < 0)
155		return (NULL);
156	if (inode < NDEVFSINO)
157		return (&dm->dm_dirent[inode]);
158	if (devfs_overflow == NULL)
159		return (NULL);
160	if (inode < NDEVFSINO + devfs_noverflow)
161		return (&dm->dm_overflow[inode - NDEVFSINO]);
162	return (NULL);
163}
164
165struct cdev **
166devfs_itod (int inode)
167{
168
169	if (inode < 0)
170		return (NULL);
171	if (inode < NDEVFSINO)
172		return (&devfs_inot[inode]);
173	if (devfs_overflow == NULL)
174		return (NULL);
175	if (inode < NDEVFSINO + devfs_noverflow)
176		return (&devfs_overflow[inode - NDEVFSINO]);
177	return (NULL);
178}
179
180static struct devfs_dirent *
181devfs_find(struct devfs_dirent *dd, const char *name, int namelen)
182{
183	struct devfs_dirent *de;
184
185	TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
186		if (namelen != de->de_dirent->d_namlen)
187			continue;
188		if (bcmp(name, de->de_dirent->d_name, namelen) != 0)
189			continue;
190		break;
191	}
192	return (de);
193}
194
195struct devfs_dirent *
196devfs_newdirent(char *name, int namelen)
197{
198	int i;
199	struct devfs_dirent *de;
200	struct dirent d;
201
202	d.d_namlen = namelen;
203	i = sizeof (*de) + GENERIC_DIRSIZ(&d);
204	MALLOC(de, struct devfs_dirent *, i, M_DEVFS, M_WAITOK | M_ZERO);
205	de->de_dirent = (struct dirent *)(de + 1);
206	de->de_dirent->d_namlen = namelen;
207	de->de_dirent->d_reclen = GENERIC_DIRSIZ(&d);
208	bcopy(name, de->de_dirent->d_name, namelen);
209	de->de_dirent->d_name[namelen] = '\0';
210	vfs_timestamp(&de->de_ctime);
211	de->de_mtime = de->de_atime = de->de_ctime;
212	de->de_links = 1;
213#ifdef MAC
214	mac_init_devfsdirent(de);
215#endif
216	return (de);
217}
218
219struct devfs_dirent *
220devfs_vmkdir(char *name, int namelen, struct devfs_dirent *dotdot)
221{
222	struct devfs_dirent *dd;
223	struct devfs_dirent *de;
224
225	dd = devfs_newdirent(name, namelen);
226
227	TAILQ_INIT(&dd->de_dlist);
228
229	dd->de_dirent->d_type = DT_DIR;
230	dd->de_mode = 0555;
231	dd->de_links = 2;
232	dd->de_dir = dd;
233
234	de = devfs_newdirent(".", 1);
235	de->de_dirent->d_type = DT_DIR;
236	de->de_dir = dd;
237	de->de_flags |= DE_DOT;
238	TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
239
240	de = devfs_newdirent("..", 2);
241	de->de_dirent->d_type = DT_DIR;
242	if (dotdot == NULL)
243		de->de_dir = dd;
244	else
245		de->de_dir = dotdot;
246	de->de_flags |= DE_DOTDOT;
247	TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
248
249	return (dd);
250}
251
252static void
253devfs_delete(struct devfs_dirent *dd, struct devfs_dirent *de)
254{
255
256	if (de->de_symlink) {
257		FREE(de->de_symlink, M_DEVFS);
258		de->de_symlink = NULL;
259	}
260	if (de->de_vnode)
261		de->de_vnode->v_data = NULL;
262	TAILQ_REMOVE(&dd->de_dlist, de, de_list);
263#ifdef MAC
264	mac_destroy_devfsdirent(de);
265#endif
266	FREE(de, M_DEVFS);
267}
268
269void
270devfs_purge(struct devfs_dirent *dd)
271{
272	struct devfs_dirent *de;
273
274	for (;;) {
275		de = TAILQ_FIRST(&dd->de_dlist);
276		if (de == NULL)
277			break;
278		devfs_delete(dd, de);
279	}
280	FREE(dd, M_DEVFS);
281}
282
283
284int
285devfs_populate(struct devfs_mount *dm)
286{
287	int i, j;
288	struct cdev *dev, *pdev;
289	struct devfs_dirent *dd;
290	struct devfs_dirent *de, **dep;
291	char *q, *s;
292
293	if (dm->dm_generation == devfs_generation)
294		return (0);
295	lockmgr(&dm->dm_lock, LK_UPGRADE, 0, curthread);
296	if (devfs_noverflow && dm->dm_overflow == NULL) {
297		i = devfs_noverflow * sizeof (struct devfs_dirent *);
298		MALLOC(dm->dm_overflow, struct devfs_dirent **, i,
299			M_DEVFS, M_WAITOK | M_ZERO);
300	}
301	while (dm->dm_generation != devfs_generation) {
302		dm->dm_generation = devfs_generation;
303		for (i = 0; i <= devfs_topino; i++) {
304			dev = *devfs_itod(i);
305			dep = devfs_itode(dm, i);
306			de = *dep;
307			if (dev == NULL && de == DE_DELETED) {
308				*dep = NULL;
309				continue;
310			}
311			if (dev == NULL && de != NULL) {
312				dd = de->de_dir;
313				*dep = NULL;
314				devfs_delete(dd, de);
315				devfs_dropref(i);
316				continue;
317			}
318			if (dev == NULL)
319				continue;
320			if (de != NULL)
321				continue;
322			if (!devfs_getref(i))
323				continue;
324			dd = dm->dm_rootdir;
325			s = dev->si_name;
326			for (;;) {
327				for (q = s; *q != '/' && *q != '\0'; q++)
328					continue;
329				if (*q != '/')
330					break;
331				de = devfs_find(dd, s, q - s);
332				if (de == NULL) {
333					de = devfs_vmkdir(s, q - s, dd);
334#ifdef MAC
335					mac_create_devfs_directory(
336					    dm->dm_mount, s, q - s, de);
337#endif
338					de->de_inode = dm->dm_inode++;
339					TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
340					dd->de_links++;
341				}
342				s = q + 1;
343				dd = de;
344			}
345			de = devfs_newdirent(s, q - s);
346			if (dev->si_flags & SI_ALIAS) {
347				de->de_inode = dm->dm_inode++;
348				de->de_uid = 0;
349				de->de_gid = 0;
350				de->de_mode = 0755;
351				de->de_dirent->d_type = DT_LNK;
352				pdev = dev->si_parent;
353				j = strlen(pdev->si_name) + 1;
354				MALLOC(de->de_symlink, char *, j, M_DEVFS, M_WAITOK);
355				bcopy(pdev->si_name, de->de_symlink, j);
356			} else {
357				de->de_inode = i;
358				de->de_uid = dev->si_uid;
359				de->de_gid = dev->si_gid;
360				de->de_mode = dev->si_mode;
361				de->de_dirent->d_type = DT_CHR;
362			}
363#ifdef MAC
364			mac_create_devfs_device(dev->si_cred, dm->dm_mount,
365			    dev, de);
366#endif
367			*dep = de;
368			de->de_dir = dd;
369			devfs_rules_apply(dm, de);
370			TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
371		}
372	}
373	lockmgr(&dm->dm_lock, LK_DOWNGRADE, 0, curthread);
374	return (0);
375}
376
377/*
378 * devfs_create() and devfs_destroy() are called from kern_conf.c and
379 * in both cases the devlock() mutex is held, so no further locking
380 * is necesary and no sleeping allowed.
381 */
382
383void
384devfs_create(struct cdev *dev)
385{
386	int ino, i, *ip;
387	struct cdev **dp;
388	struct cdev **ot;
389	int *or;
390	int n;
391
392	for (;;) {
393		/* Grab the next inode number */
394		ino = devfs_nextino;
395		i = ino + 1;
396		/* wrap around when we reach the end */
397		if (i >= NDEVFSINO + devfs_noverflow)
398			i = 3;
399		devfs_nextino = i;
400
401		/* see if it was occupied */
402		dp = devfs_itod(ino);
403		KASSERT(dp != NULL, ("DEVFS: No devptr inode %d", ino));
404		if (*dp != NULL)
405			continue;
406		ip = devfs_itor(ino);
407		KASSERT(ip != NULL, ("DEVFS: No iptr inode %d", ino));
408		if (*ip != 0)
409			continue;
410		break;
411	}
412
413	*dp = dev;
414	dev->si_inode = ino;
415	if (i > devfs_topino)
416		devfs_topino = i;
417
418	devfs_numino++;
419	devfs_generation++;
420
421	if (devfs_overflow != NULL || devfs_numino + 100 < NDEVFSINO)
422		return;
423
424	/*
425	 * Try to allocate overflow table
426	 * XXX: we can probably be less panicy these days and a linked
427	 * XXX: list of PAGESIZE/PTRSIZE entries might be a better idea.
428	 *
429	 * XXX: we may be into witness unlove here.
430	 */
431	n = devfs_noverflowwant;
432	ot = malloc(sizeof(*ot) * n, M_DEVFS, M_NOWAIT | M_ZERO);
433	if (ot == NULL)
434		return;
435	or = malloc(sizeof(*or) * n, M_DEVFS, M_NOWAIT | M_ZERO);
436	if (or == NULL) {
437		free(ot, M_DEVFS);
438		return;
439	}
440	devfs_overflow = ot;
441	devfs_refoverflow = or;
442	devfs_noverflow = n;
443	printf("DEVFS Overflow table with %d entries allocated\n", n);
444	return;
445}
446
447void
448devfs_destroy(struct cdev *dev)
449{
450	int ino;
451	struct cdev **dp;
452
453	ino = dev->si_inode;
454	dev->si_inode = 0;
455	if (ino == 0)
456		return;
457	dp = devfs_itod(ino);
458	KASSERT(*dp == dev,
459	    ("DEVFS: destroying wrong cdev ino %d", ino));
460	*dp = NULL;
461	devfs_numino--;
462	devfs_generation++;
463	if (ino < devfs_nextino)
464		devfs_nextino = ino;
465}
466