tmpfs_vfsops.c revision 1.42
1/*	$NetBSD: tmpfs_vfsops.c,v 1.42 2008/07/28 12:42:12 pooka Exp $	*/
2
3/*
4 * Copyright (c) 2005, 2006, 2007 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Julio M. Merino Vidal, developed as part of Google's Summer of Code
9 * 2005 program.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 *    notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 *    notice, this list of conditions and the following disclaimer in the
18 *    documentation and/or other materials provided with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30 * POSSIBILITY OF SUCH DAMAGE.
31 */
32
33/*
34 * Efficient memory file system.
35 *
36 * tmpfs is a file system that uses NetBSD's virtual memory sub-system
37 * (the well-known UVM) to store file data and metadata in an efficient
38 * way.  This means that it does not follow the structure of an on-disk
39 * file system because it simply does not need to.  Instead, it uses
40 * memory-specific data structures and algorithms to automatically
41 * allocate and release resources.
42 */
43
44#include <sys/cdefs.h>
45__KERNEL_RCSID(0, "$NetBSD: tmpfs_vfsops.c,v 1.42 2008/07/28 12:42:12 pooka Exp $");
46
47#include <sys/param.h>
48#include <sys/types.h>
49#include <sys/kmem.h>
50#include <sys/mount.h>
51#include <sys/stat.h>
52#include <sys/systm.h>
53#include <sys/vnode.h>
54#include <sys/proc.h>
55#include <sys/module.h>
56
57#include <miscfs/genfs/genfs.h>
58#include <fs/tmpfs/tmpfs.h>
59#include <fs/tmpfs/tmpfs_args.h>
60
61MODULE(MODULE_CLASS_VFS, tmpfs, NULL);
62
63/* --------------------------------------------------------------------- */
64
65static int	tmpfs_mount(struct mount *, const char *, void *, size_t *);
66static int	tmpfs_start(struct mount *, int);
67static int	tmpfs_unmount(struct mount *, int);
68static int	tmpfs_root(struct mount *, struct vnode **);
69static int	tmpfs_vget(struct mount *, ino_t, struct vnode **);
70static int	tmpfs_fhtovp(struct mount *, struct fid *, struct vnode **);
71static int	tmpfs_vptofh(struct vnode *, struct fid *, size_t *);
72static int	tmpfs_statvfs(struct mount *, struct statvfs *);
73static int	tmpfs_sync(struct mount *, int, kauth_cred_t);
74static void	tmpfs_init(void);
75static void	tmpfs_done(void);
76static int	tmpfs_snapshot(struct mount *, struct vnode *,
77		    struct timespec *);
78
79/* --------------------------------------------------------------------- */
80
81static int
82tmpfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
83{
84	struct lwp *l = curlwp;
85	int error;
86	ino_t nodes;
87	size_t pages;
88	struct tmpfs_mount *tmp;
89	struct tmpfs_node *root;
90	struct tmpfs_args *args = data;
91
92	if (*data_len < sizeof *args)
93		return EINVAL;
94
95	/* Handle retrieval of mount point arguments. */
96	if (mp->mnt_flag & MNT_GETARGS) {
97		if (mp->mnt_data == NULL)
98			return EIO;
99		tmp = VFS_TO_TMPFS(mp);
100
101		args->ta_version = TMPFS_ARGS_VERSION;
102		args->ta_nodes_max = tmp->tm_nodes_max;
103		args->ta_size_max = tmp->tm_pages_max * PAGE_SIZE;
104
105		root = tmp->tm_root;
106		args->ta_root_uid = root->tn_uid;
107		args->ta_root_gid = root->tn_gid;
108		args->ta_root_mode = root->tn_mode;
109
110		*data_len = sizeof *args;
111		return 0;
112	}
113
114	if (mp->mnt_flag & MNT_UPDATE) {
115		/* XXX: There is no support yet to update file system
116		 * settings.  Should be added. */
117
118		return EOPNOTSUPP;
119	}
120
121	if (args->ta_version != TMPFS_ARGS_VERSION)
122		return EINVAL;
123
124	/* Do not allow mounts if we do not have enough memory to preserve
125	 * the minimum reserved pages. */
126	if (tmpfs_mem_info(true) < TMPFS_PAGES_RESERVED)
127		return EINVAL;
128
129	/* Get the maximum number of memory pages this file system is
130	 * allowed to use, based on the maximum size the user passed in
131	 * the mount structure.  A value of zero is treated as if the
132	 * maximum available space was requested. */
133	if (args->ta_size_max < PAGE_SIZE || args->ta_size_max >= SIZE_MAX)
134		pages = SIZE_MAX;
135	else
136		pages = args->ta_size_max / PAGE_SIZE +
137		    (args->ta_size_max % PAGE_SIZE == 0 ? 0 : 1);
138	if (pages > INT_MAX)
139		pages = INT_MAX;
140	KASSERT(pages > 0);
141
142	if (args->ta_nodes_max <= 3)
143		nodes = 3 + pages * PAGE_SIZE / 1024;
144	else
145		nodes = args->ta_nodes_max;
146	if (nodes > INT_MAX)
147		nodes = INT_MAX;
148	KASSERT(nodes >= 3);
149
150	/* Allocate the tmpfs mount structure and fill it. */
151	tmp = kmem_alloc(sizeof(struct tmpfs_mount), KM_SLEEP);
152	if (tmp == NULL)
153		return ENOMEM;
154
155	tmp->tm_nodes_max = nodes;
156	tmp->tm_nodes_cnt = 0;
157	LIST_INIT(&tmp->tm_nodes);
158
159	mutex_init(&tmp->tm_lock, MUTEX_DEFAULT, IPL_NONE);
160
161	tmp->tm_pages_max = pages;
162	tmp->tm_pages_used = 0;
163	tmpfs_pool_init(&tmp->tm_dirent_pool, sizeof(struct tmpfs_dirent),
164	    "dirent", tmp);
165	tmpfs_pool_init(&tmp->tm_node_pool, sizeof(struct tmpfs_node),
166	    "node", tmp);
167	tmpfs_str_pool_init(&tmp->tm_str_pool, tmp);
168
169	/* Allocate the root node. */
170	error = tmpfs_alloc_node(tmp, VDIR, args->ta_root_uid,
171	    args->ta_root_gid, args->ta_root_mode & ALLPERMS, NULL, NULL,
172	    VNOVAL, &root);
173	KASSERT(error == 0 && root != NULL);
174	root->tn_links++;
175	tmp->tm_root = root;
176
177	mp->mnt_data = tmp;
178	mp->mnt_flag |= MNT_LOCAL;
179	mp->mnt_stat.f_namemax = MAXNAMLEN;
180	mp->mnt_fs_bshift = PAGE_SHIFT;
181	mp->mnt_dev_bshift = DEV_BSHIFT;
182	mp->mnt_iflag |= IMNT_MPSAFE;
183	vfs_getnewfsid(mp);
184
185	return set_statvfs_info(path, UIO_USERSPACE, "tmpfs", UIO_SYSSPACE,
186	    mp->mnt_op->vfs_name, mp, l);
187}
188
189/* --------------------------------------------------------------------- */
190
191static int
192tmpfs_start(struct mount *mp, int flags)
193{
194
195	return 0;
196}
197
198/* --------------------------------------------------------------------- */
199
200/* ARGSUSED2 */
201static int
202tmpfs_unmount(struct mount *mp, int mntflags)
203{
204	int error;
205	int flags = 0;
206	struct tmpfs_mount *tmp;
207	struct tmpfs_node *node;
208
209	/* Handle forced unmounts. */
210	if (mntflags & MNT_FORCE)
211		flags |= FORCECLOSE;
212
213	/* Finalize all pending I/O. */
214	error = vflush(mp, NULL, flags);
215	if (error != 0)
216		return error;
217
218	tmp = VFS_TO_TMPFS(mp);
219
220	/* Free all associated data.  The loop iterates over the linked list
221	 * we have containing all used nodes.  For each of them that is
222	 * a directory, we free all its directory entries.  Note that after
223	 * freeing a node, it will automatically go to the available list,
224	 * so we will later have to iterate over it to release its items. */
225	node = LIST_FIRST(&tmp->tm_nodes);
226	while (node != NULL) {
227		struct tmpfs_node *next;
228
229		if (node->tn_type == VDIR) {
230			struct tmpfs_dirent *de;
231
232			de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir);
233			while (de != NULL) {
234				struct tmpfs_dirent *nde;
235
236				nde = TAILQ_NEXT(de, td_entries);
237				tmpfs_free_dirent(tmp, de, false);
238				de = nde;
239				node->tn_size -= sizeof(struct tmpfs_dirent);
240			}
241		}
242
243		next = LIST_NEXT(node, tn_entries);
244		tmpfs_free_node(tmp, node);
245		node = next;
246	}
247
248	tmpfs_pool_destroy(&tmp->tm_dirent_pool);
249	tmpfs_pool_destroy(&tmp->tm_node_pool);
250	tmpfs_str_pool_destroy(&tmp->tm_str_pool);
251
252	KASSERT(tmp->tm_pages_used == 0);
253
254	/* Throw away the tmpfs_mount structure. */
255	mutex_destroy(&tmp->tm_lock);
256	kmem_free(tmp, sizeof(*tmp));
257	mp->mnt_data = NULL;
258
259	return 0;
260}
261
262/* --------------------------------------------------------------------- */
263
264static int
265tmpfs_root(struct mount *mp, struct vnode **vpp)
266{
267
268	return tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, vpp);
269}
270
271/* --------------------------------------------------------------------- */
272
273static int
274tmpfs_vget(struct mount *mp, ino_t ino,
275    struct vnode **vpp)
276{
277
278	printf("tmpfs_vget called; need for it unknown yet\n");
279	return EOPNOTSUPP;
280}
281
282/* --------------------------------------------------------------------- */
283
284static int
285tmpfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
286{
287	bool found;
288	struct tmpfs_fid tfh;
289	struct tmpfs_mount *tmp;
290	struct tmpfs_node *node;
291
292	tmp = VFS_TO_TMPFS(mp);
293
294	if (fhp->fid_len != sizeof(struct tmpfs_fid))
295		return EINVAL;
296
297	memcpy(&tfh, fhp, sizeof(struct tmpfs_fid));
298
299	if (tfh.tf_id >= tmp->tm_nodes_max)
300		return EINVAL;
301
302	found = false;
303	mutex_enter(&tmp->tm_lock);
304	LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
305		if (node->tn_id == tfh.tf_id &&
306		    node->tn_gen == tfh.tf_gen) {
307			found = true;
308			break;
309		}
310	}
311	mutex_exit(&tmp->tm_lock);
312
313	/* XXXAD nothing to prevent 'node' from being removed. */
314	return found ? tmpfs_alloc_vp(mp, node, vpp) : EINVAL;
315}
316
317/* --------------------------------------------------------------------- */
318
319static int
320tmpfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
321{
322	struct tmpfs_fid tfh;
323	struct tmpfs_node *node;
324
325	if (*fh_size < sizeof(struct tmpfs_fid)) {
326		*fh_size = sizeof(struct tmpfs_fid);
327		return E2BIG;
328	}
329	*fh_size = sizeof(struct tmpfs_fid);
330	node = VP_TO_TMPFS_NODE(vp);
331
332	memset(&tfh, 0, sizeof(tfh));
333	tfh.tf_len = sizeof(struct tmpfs_fid);
334	tfh.tf_gen = node->tn_gen;
335	tfh.tf_id = node->tn_id;
336	memcpy(fhp, &tfh, sizeof(tfh));
337
338	return 0;
339}
340
341/* --------------------------------------------------------------------- */
342
343/* ARGSUSED2 */
344static int
345tmpfs_statvfs(struct mount *mp, struct statvfs *sbp)
346{
347	fsfilcnt_t freenodes;
348	struct tmpfs_mount *tmp;
349
350	tmp = VFS_TO_TMPFS(mp);
351
352	sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE;
353
354	sbp->f_blocks = TMPFS_PAGES_MAX(tmp);
355	sbp->f_bavail = sbp->f_bfree = TMPFS_PAGES_AVAIL(tmp);
356	sbp->f_bresvd = 0;
357
358	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt,
359	    TMPFS_PAGES_AVAIL(tmp) * PAGE_SIZE / sizeof(struct tmpfs_node));
360
361	sbp->f_files = tmp->tm_nodes_cnt + freenodes;
362	sbp->f_favail = sbp->f_ffree = freenodes;
363	sbp->f_fresvd = 0;
364
365	copy_statvfs_info(sbp, mp);
366
367	return 0;
368}
369
370/* --------------------------------------------------------------------- */
371
372/* ARGSUSED0 */
373static int
374tmpfs_sync(struct mount *mp, int waitfor,
375    kauth_cred_t uc)
376{
377
378	return 0;
379}
380
381/* --------------------------------------------------------------------- */
382
383static void
384tmpfs_init(void)
385{
386
387}
388
389/* --------------------------------------------------------------------- */
390
391static void
392tmpfs_done(void)
393{
394
395}
396
397/* --------------------------------------------------------------------- */
398
399static int
400tmpfs_snapshot(struct mount *mp, struct vnode *vp,
401    struct timespec *ctime)
402{
403
404	return EOPNOTSUPP;
405}
406
407/* --------------------------------------------------------------------- */
408
409/*
410 * tmpfs vfs operations.
411 */
412
413extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc;
414extern const struct vnodeopv_desc tmpfs_specop_opv_desc;
415extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc;
416
417const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = {
418	&tmpfs_fifoop_opv_desc,
419	&tmpfs_specop_opv_desc,
420	&tmpfs_vnodeop_opv_desc,
421	NULL,
422};
423
424struct vfsops tmpfs_vfsops = {
425	MOUNT_TMPFS,			/* vfs_name */
426	sizeof (struct tmpfs_args),
427	tmpfs_mount,			/* vfs_mount */
428	tmpfs_start,			/* vfs_start */
429	tmpfs_unmount,			/* vfs_unmount */
430	tmpfs_root,			/* vfs_root */
431	(void *)eopnotsupp,		/* vfs_quotactl */
432	tmpfs_statvfs,			/* vfs_statvfs */
433	tmpfs_sync,			/* vfs_sync */
434	tmpfs_vget,			/* vfs_vget */
435	tmpfs_fhtovp,			/* vfs_fhtovp */
436	tmpfs_vptofh,			/* vfs_vptofh */
437	tmpfs_init,			/* vfs_init */
438	NULL,				/* vfs_reinit */
439	tmpfs_done,			/* vfs_done */
440	NULL,				/* vfs_mountroot */
441	tmpfs_snapshot,			/* vfs_snapshot */
442	vfs_stdextattrctl,		/* vfs_extattrctl */
443	(void *)eopnotsupp,		/* vfs_suspendctl */
444	genfs_renamelock_enter,
445	genfs_renamelock_exit,
446	(void *)eopnotsupp,
447	tmpfs_vnodeopv_descs,
448	0,				/* vfs_refcount */
449	{ NULL, NULL },
450};
451
452static int
453tmpfs_modcmd(modcmd_t cmd, void *arg)
454{
455
456	switch (cmd) {
457	case MODULE_CMD_INIT:
458		return vfs_attach(&tmpfs_vfsops);
459	case MODULE_CMD_FINI:
460		return vfs_detach(&tmpfs_vfsops);
461	default:
462		return ENOTTY;
463	}
464}
465