xref: /netbsd-src/sys/fs/tmpfs/tmpfs_vfsops.c (revision b1bb3099bf4d47bbe8c7be5b78240a535263771f)
1 /*	$NetBSD: tmpfs_vfsops.c,v 1.61 2014/04/30 01:59:30 christos 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 1.61 2014/04/30 01:59:30 christos 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/module.h>
55 
56 #include <miscfs/genfs/genfs.h>
57 #include <fs/tmpfs/tmpfs.h>
58 #include <fs/tmpfs/tmpfs_args.h>
59 
60 MODULE(MODULE_CLASS_VFS, tmpfs, NULL);
61 
62 struct pool	tmpfs_dirent_pool;
63 struct pool	tmpfs_node_pool;
64 
65 static int	tmpfs_mount(struct mount *, const char *, void *, size_t *);
66 static int	tmpfs_start(struct mount *, int);
67 static int	tmpfs_unmount(struct mount *, int);
68 static int	tmpfs_root(struct mount *, vnode_t **);
69 static int	tmpfs_vget(struct mount *, ino_t, vnode_t **);
70 static int	tmpfs_fhtovp(struct mount *, struct fid *, vnode_t **);
71 static int	tmpfs_vptofh(struct vnode *, struct fid *, size_t *);
72 static int	tmpfs_statvfs(struct mount *, struct statvfs *);
73 static int	tmpfs_sync(struct mount *, int, kauth_cred_t);
74 static void	tmpfs_init(void);
75 static void	tmpfs_done(void);
76 static int	tmpfs_snapshot(struct mount *, vnode_t *, struct timespec *);
77 
78 static void
79 tmpfs_init(void)
80 {
81 
82 	pool_init(&tmpfs_dirent_pool, sizeof(tmpfs_dirent_t), 0, 0, 0,
83 	    "tmpfs_dirent", &pool_allocator_nointr, IPL_NONE);
84 	pool_init(&tmpfs_node_pool, sizeof(tmpfs_node_t), 0, 0, 0,
85 	    "tmpfs_node", &pool_allocator_nointr, IPL_NONE);
86 }
87 
88 static void
89 tmpfs_done(void)
90 {
91 
92 	pool_destroy(&tmpfs_dirent_pool);
93 	pool_destroy(&tmpfs_node_pool);
94 }
95 
96 static int
97 tmpfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
98 {
99 	struct tmpfs_args *args = data;
100 	tmpfs_mount_t *tmp;
101 	tmpfs_node_t *root;
102 	uint64_t memlimit;
103 	ino_t nodes;
104 	int error;
105 
106 	if (args == NULL)
107 		return EINVAL;
108 
109 	/* Validate the version. */
110 	if (*data_len < sizeof(*args) ||
111 	    args->ta_version != TMPFS_ARGS_VERSION)
112 		return EINVAL;
113 
114 	/* Handle retrieval of mount point arguments. */
115 	if (mp->mnt_flag & MNT_GETARGS) {
116 		if (mp->mnt_data == NULL)
117 			return EIO;
118 		tmp = VFS_TO_TMPFS(mp);
119 
120 		args->ta_version = TMPFS_ARGS_VERSION;
121 		args->ta_nodes_max = tmp->tm_nodes_max;
122 		args->ta_size_max = tmp->tm_mem_limit;
123 
124 		root = tmp->tm_root;
125 		args->ta_root_uid = root->tn_uid;
126 		args->ta_root_gid = root->tn_gid;
127 		args->ta_root_mode = root->tn_mode;
128 
129 		*data_len = sizeof(*args);
130 		return 0;
131 	}
132 
133 
134 	/* Prohibit mounts if there is not enough memory. */
135 	if (tmpfs_mem_info(true) < TMPFS_PAGES_RESERVED)
136 		return EINVAL;
137 
138 	/* Get the memory usage limit for this file-system. */
139 	if (args->ta_size_max < PAGE_SIZE) {
140 		memlimit = UINT64_MAX;
141 	} else {
142 		memlimit = args->ta_size_max;
143 	}
144 	KASSERT(memlimit > 0);
145 
146 	if (args->ta_nodes_max <= 3) {
147 		nodes = 3 + (memlimit / 1024);
148 	} else {
149 		nodes = args->ta_nodes_max;
150 	}
151 	nodes = MIN(nodes, INT_MAX);
152 	KASSERT(nodes >= 3);
153 
154 	if (mp->mnt_flag & MNT_UPDATE) {
155 		tmp = VFS_TO_TMPFS(mp);
156 		if (nodes < tmp->tm_nodes_cnt)
157 			return EBUSY;
158 		if ((error = tmpfs_mntmem_set(tmp, memlimit)) != 0)
159 			return error;
160 		tmp->tm_nodes_max = nodes;
161 		root = tmp->tm_root;
162 		root->tn_uid = args->ta_root_uid;
163 		root->tn_gid = args->ta_root_gid;
164 		root->tn_mode = args->ta_root_mode;
165 		return 0;
166 	}
167 
168 	/* Allocate the tmpfs mount structure and fill it. */
169 	tmp = kmem_zalloc(sizeof(tmpfs_mount_t), KM_SLEEP);
170 	if (tmp == NULL)
171 		return ENOMEM;
172 
173 	tmp->tm_nodes_max = nodes;
174 	tmp->tm_nodes_cnt = 0;
175 	LIST_INIT(&tmp->tm_nodes);
176 
177 	mutex_init(&tmp->tm_lock, MUTEX_DEFAULT, IPL_NONE);
178 	tmpfs_mntmem_init(tmp, memlimit);
179 
180 	/* Allocate the root node. */
181 	error = tmpfs_alloc_node(tmp, VDIR, args->ta_root_uid,
182 	    args->ta_root_gid, args->ta_root_mode & ALLPERMS, NULL,
183 	    VNOVAL, &root);
184 	KASSERT(error == 0 && root != NULL);
185 
186 	/*
187 	 * Parent of the root inode is itself.  Also, root inode has no
188 	 * directory entry (i.e. is never attached), thus hold an extra
189 	 * reference (link) for it.
190 	 */
191 	root->tn_links++;
192 	root->tn_spec.tn_dir.tn_parent = root;
193 	tmp->tm_root = root;
194 
195 	mp->mnt_data = tmp;
196 	mp->mnt_flag |= MNT_LOCAL;
197 	mp->mnt_stat.f_namemax = TMPFS_MAXNAMLEN;
198 	mp->mnt_fs_bshift = PAGE_SHIFT;
199 	mp->mnt_dev_bshift = DEV_BSHIFT;
200 	mp->mnt_iflag |= IMNT_MPSAFE;
201 	vfs_getnewfsid(mp);
202 
203 	error = set_statvfs_info(path, UIO_USERSPACE, "tmpfs", UIO_SYSSPACE,
204 	    mp->mnt_op->vfs_name, mp, curlwp);
205 	if (error) {
206 		(void)tmpfs_unmount(mp, MNT_FORCE);
207 	}
208 	return error;
209 }
210 
211 static int
212 tmpfs_start(struct mount *mp, int flags)
213 {
214 
215 	return 0;
216 }
217 
218 static int
219 tmpfs_unmount(struct mount *mp, int mntflags)
220 {
221 	tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp);
222 	tmpfs_node_t *node, *cnode;
223 	int error, flags = 0;
224 
225 	/* Handle forced unmounts. */
226 	if (mntflags & MNT_FORCE)
227 		flags |= FORCECLOSE;
228 
229 	/* Finalize all pending I/O. */
230 	error = vflush(mp, NULL, flags);
231 	if (error != 0)
232 		return error;
233 
234 	/*
235 	 * First round, detach and destroy all directory entries.
236 	 * Also, clear the pointers to the vnodes - they are gone.
237 	 */
238 	LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
239 		tmpfs_dirent_t *de;
240 
241 		node->tn_vnode = NULL;
242 		if (node->tn_type != VDIR) {
243 			continue;
244 		}
245 		while ((de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir)) != NULL) {
246 			cnode = de->td_node;
247 			if (cnode && cnode != TMPFS_NODE_WHITEOUT) {
248 				cnode->tn_vnode = NULL;
249 			}
250 			tmpfs_dir_detach(node, de);
251 			tmpfs_free_dirent(tmp, de);
252 		}
253 		/* Extra virtual entry (itself for the root). */
254 		node->tn_links--;
255 	}
256 
257 	/* Release the reference on root (diagnostic). */
258 	node = tmp->tm_root;
259 	node->tn_links--;
260 
261 	/* Second round, destroy all inodes. */
262 	while ((node = LIST_FIRST(&tmp->tm_nodes)) != NULL) {
263 		tmpfs_free_node(tmp, node);
264 	}
265 
266 	/* Throw away the tmpfs_mount structure. */
267 	tmpfs_mntmem_destroy(tmp);
268 	mutex_destroy(&tmp->tm_lock);
269 	kmem_free(tmp, sizeof(*tmp));
270 	mp->mnt_data = NULL;
271 
272 	return 0;
273 }
274 
275 static int
276 tmpfs_root(struct mount *mp, vnode_t **vpp)
277 {
278 	tmpfs_node_t *node = VFS_TO_TMPFS(mp)->tm_root;
279 
280 	mutex_enter(&node->tn_vlock);
281 	return tmpfs_vnode_get(mp, node, vpp);
282 }
283 
284 static int
285 tmpfs_vget(struct mount *mp, ino_t ino, vnode_t **vpp)
286 {
287 
288 	return EOPNOTSUPP;
289 }
290 
291 static int
292 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, vnode_t **vpp)
293 {
294 	tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp);
295 	tmpfs_node_t *node;
296 	tmpfs_fid_t tfh;
297 	int error;
298 
299 	if (fhp->fid_len != sizeof(tmpfs_fid_t)) {
300 		return EINVAL;
301 	}
302 	memcpy(&tfh, fhp, sizeof(tmpfs_fid_t));
303 
304 	mutex_enter(&tmp->tm_lock);
305 	LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
306 		if (node->tn_id == tfh.tf_id) {
307 			mutex_enter(&node->tn_vlock);
308 			break;
309 		}
310 	}
311 	mutex_exit(&tmp->tm_lock);
312 
313 	if (node == NULL)
314 		return ESTALE;
315 	/* Will release the tn_vlock. */
316 	if ((error = tmpfs_vnode_get(mp, node, vpp)) != 0)
317 		return error;
318 	if (TMPFS_NODE_GEN(node) != tfh.tf_gen) {
319 		vput(*vpp);
320 		*vpp = NULL;
321 		return ESTALE;
322 	}
323 
324 	return 0;
325 }
326 
327 static int
328 tmpfs_vptofh(vnode_t *vp, struct fid *fhp, size_t *fh_size)
329 {
330 	tmpfs_fid_t tfh;
331 	tmpfs_node_t *node;
332 
333 	if (*fh_size < sizeof(tmpfs_fid_t)) {
334 		*fh_size = sizeof(tmpfs_fid_t);
335 		return E2BIG;
336 	}
337 	*fh_size = sizeof(tmpfs_fid_t);
338 	node = VP_TO_TMPFS_NODE(vp);
339 
340 	memset(&tfh, 0, sizeof(tfh));
341 	tfh.tf_len = sizeof(tmpfs_fid_t);
342 	tfh.tf_gen = TMPFS_NODE_GEN(node);
343 	tfh.tf_id = node->tn_id;
344 	memcpy(fhp, &tfh, sizeof(tfh));
345 
346 	return 0;
347 }
348 
349 static int
350 tmpfs_statvfs(struct mount *mp, struct statvfs *sbp)
351 {
352 	tmpfs_mount_t *tmp;
353 	fsfilcnt_t freenodes;
354 	size_t avail;
355 
356 	tmp = VFS_TO_TMPFS(mp);
357 
358 	sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE;
359 
360 	mutex_enter(&tmp->tm_acc_lock);
361 	avail =  tmpfs_pages_avail(tmp);
362 	sbp->f_blocks = (tmpfs_bytes_max(tmp) >> PAGE_SHIFT);
363 	sbp->f_bavail = sbp->f_bfree = avail;
364 	sbp->f_bresvd = 0;
365 
366 	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt,
367 	    avail * PAGE_SIZE / sizeof(tmpfs_node_t));
368 
369 	sbp->f_files = tmp->tm_nodes_cnt + freenodes;
370 	sbp->f_favail = sbp->f_ffree = freenodes;
371 	sbp->f_fresvd = 0;
372 	mutex_exit(&tmp->tm_acc_lock);
373 
374 	copy_statvfs_info(sbp, mp);
375 
376 	return 0;
377 }
378 
379 static int
380 tmpfs_sync(struct mount *mp, int waitfor, kauth_cred_t uc)
381 {
382 
383 	return 0;
384 }
385 
386 static int
387 tmpfs_snapshot(struct mount *mp, vnode_t *vp, struct timespec *ctime)
388 {
389 
390 	return EOPNOTSUPP;
391 }
392 
393 /*
394  * tmpfs vfs operations.
395  */
396 
397 extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc;
398 extern const struct vnodeopv_desc tmpfs_specop_opv_desc;
399 extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc;
400 
401 const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = {
402 	&tmpfs_fifoop_opv_desc,
403 	&tmpfs_specop_opv_desc,
404 	&tmpfs_vnodeop_opv_desc,
405 	NULL,
406 };
407 
408 struct vfsops tmpfs_vfsops = {
409 	.vfs_name = MOUNT_TMPFS,
410 	.vfs_min_mount_data = sizeof (struct tmpfs_args),
411 	.vfs_mount = tmpfs_mount,
412 	.vfs_start = tmpfs_start,
413 	.vfs_unmount = tmpfs_unmount,
414 	.vfs_root = tmpfs_root,
415 	.vfs_quotactl = (void *)eopnotsupp,
416 	.vfs_statvfs = tmpfs_statvfs,
417 	.vfs_sync = tmpfs_sync,
418 	.vfs_vget = tmpfs_vget,
419 	.vfs_fhtovp = tmpfs_fhtovp,
420 	.vfs_vptofh = tmpfs_vptofh,
421 	.vfs_init = tmpfs_init,
422 	.vfs_done = tmpfs_done,
423 	.vfs_snapshot = tmpfs_snapshot,
424 	.vfs_extattrctl = vfs_stdextattrctl,
425 	.vfs_suspendctl = (void *)eopnotsupp,
426 	.vfs_renamelock_enter = genfs_renamelock_enter,
427 	.vfs_renamelock_exit = genfs_renamelock_exit,
428 	.vfs_fsync = (void *)eopnotsupp,
429 	.vfs_opv_descs = tmpfs_vnodeopv_descs
430 };
431 
432 static int
433 tmpfs_modcmd(modcmd_t cmd, void *arg)
434 {
435 
436 	switch (cmd) {
437 	case MODULE_CMD_INIT:
438 		return vfs_attach(&tmpfs_vfsops);
439 	case MODULE_CMD_FINI:
440 		return vfs_detach(&tmpfs_vfsops);
441 	default:
442 		return ENOTTY;
443 	}
444 }
445