xref: /netbsd-src/sys/fs/tmpfs/tmpfs_vfsops.c (revision aad9773e38ed2370a628a6416e098f9008fc10a7)
1 /*	$NetBSD: tmpfs_vfsops.c,v 1.63 2014/06/10 16:10:59 martin 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 1.63 2014/06/10 16:10:59 martin 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) < uvmexp.freetarg)
136 		return EINVAL;
137 
138 	/* Check for invalid uid and gid arguments */
139 	if (args->ta_root_uid == VNOVAL || args->ta_root_gid == VNOVAL)
140 		return EINVAL;
141 
142 	/* This can never happen? */
143 	if ((args->ta_root_mode & ALLPERMS) == VNOVAL)
144 		return EINVAL;
145 
146 	/* Get the memory usage limit for this file-system. */
147 	if (args->ta_size_max < PAGE_SIZE) {
148 		memlimit = UINT64_MAX;
149 	} else {
150 		memlimit = args->ta_size_max;
151 	}
152 	KASSERT(memlimit > 0);
153 
154 	if (args->ta_nodes_max <= 3) {
155 		nodes = 3 + (memlimit / 1024);
156 	} else {
157 		nodes = args->ta_nodes_max;
158 	}
159 	nodes = MIN(nodes, INT_MAX);
160 	KASSERT(nodes >= 3);
161 
162 	if (mp->mnt_flag & MNT_UPDATE) {
163 		tmp = VFS_TO_TMPFS(mp);
164 		if (nodes < tmp->tm_nodes_cnt)
165 			return EBUSY;
166 		if ((error = tmpfs_mntmem_set(tmp, memlimit)) != 0)
167 			return error;
168 		tmp->tm_nodes_max = nodes;
169 		root = tmp->tm_root;
170 		root->tn_uid = args->ta_root_uid;
171 		root->tn_gid = args->ta_root_gid;
172 		root->tn_mode = args->ta_root_mode;
173 		return 0;
174 	}
175 
176 	/* Allocate the tmpfs mount structure and fill it. */
177 	tmp = kmem_zalloc(sizeof(tmpfs_mount_t), KM_SLEEP);
178 	if (tmp == NULL)
179 		return ENOMEM;
180 
181 	tmp->tm_nodes_max = nodes;
182 	tmp->tm_nodes_cnt = 0;
183 	LIST_INIT(&tmp->tm_nodes);
184 
185 	mutex_init(&tmp->tm_lock, MUTEX_DEFAULT, IPL_NONE);
186 	tmpfs_mntmem_init(tmp, memlimit);
187 
188 	/* Allocate the root node. */
189 	error = tmpfs_alloc_node(tmp, VDIR, args->ta_root_uid,
190 	    args->ta_root_gid, args->ta_root_mode & ALLPERMS, NULL,
191 	    VNOVAL, &root);
192 	KASSERT(error == 0 && root != NULL);
193 
194 	/*
195 	 * Parent of the root inode is itself.  Also, root inode has no
196 	 * directory entry (i.e. is never attached), thus hold an extra
197 	 * reference (link) for it.
198 	 */
199 	root->tn_links++;
200 	root->tn_spec.tn_dir.tn_parent = root;
201 	tmp->tm_root = root;
202 
203 	mp->mnt_data = tmp;
204 	mp->mnt_flag |= MNT_LOCAL;
205 	mp->mnt_stat.f_namemax = TMPFS_MAXNAMLEN;
206 	mp->mnt_fs_bshift = PAGE_SHIFT;
207 	mp->mnt_dev_bshift = DEV_BSHIFT;
208 	mp->mnt_iflag |= IMNT_MPSAFE;
209 	vfs_getnewfsid(mp);
210 
211 	error = set_statvfs_info(path, UIO_USERSPACE, "tmpfs", UIO_SYSSPACE,
212 	    mp->mnt_op->vfs_name, mp, curlwp);
213 	if (error) {
214 		(void)tmpfs_unmount(mp, MNT_FORCE);
215 	}
216 	return error;
217 }
218 
219 static int
220 tmpfs_start(struct mount *mp, int flags)
221 {
222 
223 	return 0;
224 }
225 
226 static int
227 tmpfs_unmount(struct mount *mp, int mntflags)
228 {
229 	tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp);
230 	tmpfs_node_t *node, *cnode;
231 	int error, flags = 0;
232 
233 	/* Handle forced unmounts. */
234 	if (mntflags & MNT_FORCE)
235 		flags |= FORCECLOSE;
236 
237 	/* Finalize all pending I/O. */
238 	error = vflush(mp, NULL, flags);
239 	if (error != 0)
240 		return error;
241 
242 	/*
243 	 * First round, detach and destroy all directory entries.
244 	 * Also, clear the pointers to the vnodes - they are gone.
245 	 */
246 	LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
247 		tmpfs_dirent_t *de;
248 
249 		node->tn_vnode = NULL;
250 		if (node->tn_type != VDIR) {
251 			continue;
252 		}
253 		while ((de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir)) != NULL) {
254 			cnode = de->td_node;
255 			if (cnode && cnode != TMPFS_NODE_WHITEOUT) {
256 				cnode->tn_vnode = NULL;
257 			}
258 			tmpfs_dir_detach(node, de);
259 			tmpfs_free_dirent(tmp, de);
260 		}
261 		/* Extra virtual entry (itself for the root). */
262 		node->tn_links--;
263 	}
264 
265 	/* Release the reference on root (diagnostic). */
266 	node = tmp->tm_root;
267 	node->tn_links--;
268 
269 	/* Second round, destroy all inodes. */
270 	while ((node = LIST_FIRST(&tmp->tm_nodes)) != NULL) {
271 		tmpfs_free_node(tmp, node);
272 	}
273 
274 	/* Throw away the tmpfs_mount structure. */
275 	tmpfs_mntmem_destroy(tmp);
276 	mutex_destroy(&tmp->tm_lock);
277 	kmem_free(tmp, sizeof(*tmp));
278 	mp->mnt_data = NULL;
279 
280 	return 0;
281 }
282 
283 static int
284 tmpfs_root(struct mount *mp, vnode_t **vpp)
285 {
286 	tmpfs_node_t *node = VFS_TO_TMPFS(mp)->tm_root;
287 
288 	mutex_enter(&node->tn_vlock);
289 	return tmpfs_vnode_get(mp, node, vpp);
290 }
291 
292 static int
293 tmpfs_vget(struct mount *mp, ino_t ino, vnode_t **vpp)
294 {
295 
296 	return EOPNOTSUPP;
297 }
298 
299 static int
300 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, vnode_t **vpp)
301 {
302 	tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp);
303 	tmpfs_node_t *node;
304 	tmpfs_fid_t tfh;
305 	int error;
306 
307 	if (fhp->fid_len != sizeof(tmpfs_fid_t)) {
308 		return EINVAL;
309 	}
310 	memcpy(&tfh, fhp, sizeof(tmpfs_fid_t));
311 
312 	mutex_enter(&tmp->tm_lock);
313 	LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
314 		if (node->tn_id == tfh.tf_id) {
315 			mutex_enter(&node->tn_vlock);
316 			break;
317 		}
318 	}
319 	mutex_exit(&tmp->tm_lock);
320 
321 	if (node == NULL)
322 		return ESTALE;
323 	/* Will release the tn_vlock. */
324 	if ((error = tmpfs_vnode_get(mp, node, vpp)) != 0)
325 		return error;
326 	if (TMPFS_NODE_GEN(node) != tfh.tf_gen) {
327 		vput(*vpp);
328 		*vpp = NULL;
329 		return ESTALE;
330 	}
331 
332 	return 0;
333 }
334 
335 static int
336 tmpfs_vptofh(vnode_t *vp, struct fid *fhp, size_t *fh_size)
337 {
338 	tmpfs_fid_t tfh;
339 	tmpfs_node_t *node;
340 
341 	if (*fh_size < sizeof(tmpfs_fid_t)) {
342 		*fh_size = sizeof(tmpfs_fid_t);
343 		return E2BIG;
344 	}
345 	*fh_size = sizeof(tmpfs_fid_t);
346 	node = VP_TO_TMPFS_NODE(vp);
347 
348 	memset(&tfh, 0, sizeof(tfh));
349 	tfh.tf_len = sizeof(tmpfs_fid_t);
350 	tfh.tf_gen = TMPFS_NODE_GEN(node);
351 	tfh.tf_id = node->tn_id;
352 	memcpy(fhp, &tfh, sizeof(tfh));
353 
354 	return 0;
355 }
356 
357 static int
358 tmpfs_statvfs(struct mount *mp, struct statvfs *sbp)
359 {
360 	tmpfs_mount_t *tmp;
361 	fsfilcnt_t freenodes;
362 	size_t avail;
363 
364 	tmp = VFS_TO_TMPFS(mp);
365 
366 	sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE;
367 
368 	mutex_enter(&tmp->tm_acc_lock);
369 	avail =  tmpfs_pages_avail(tmp);
370 	sbp->f_blocks = (tmpfs_bytes_max(tmp) >> PAGE_SHIFT);
371 	sbp->f_bavail = sbp->f_bfree = avail;
372 	sbp->f_bresvd = 0;
373 
374 	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt,
375 	    avail * PAGE_SIZE / sizeof(tmpfs_node_t));
376 
377 	sbp->f_files = tmp->tm_nodes_cnt + freenodes;
378 	sbp->f_favail = sbp->f_ffree = freenodes;
379 	sbp->f_fresvd = 0;
380 	mutex_exit(&tmp->tm_acc_lock);
381 
382 	copy_statvfs_info(sbp, mp);
383 
384 	return 0;
385 }
386 
387 static int
388 tmpfs_sync(struct mount *mp, int waitfor, kauth_cred_t uc)
389 {
390 
391 	return 0;
392 }
393 
394 static int
395 tmpfs_snapshot(struct mount *mp, vnode_t *vp, struct timespec *ctime)
396 {
397 
398 	return EOPNOTSUPP;
399 }
400 
401 /*
402  * tmpfs vfs operations.
403  */
404 
405 extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc;
406 extern const struct vnodeopv_desc tmpfs_specop_opv_desc;
407 extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc;
408 
409 const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = {
410 	&tmpfs_fifoop_opv_desc,
411 	&tmpfs_specop_opv_desc,
412 	&tmpfs_vnodeop_opv_desc,
413 	NULL,
414 };
415 
416 struct vfsops tmpfs_vfsops = {
417 	.vfs_name = MOUNT_TMPFS,
418 	.vfs_min_mount_data = sizeof (struct tmpfs_args),
419 	.vfs_mount = tmpfs_mount,
420 	.vfs_start = tmpfs_start,
421 	.vfs_unmount = tmpfs_unmount,
422 	.vfs_root = tmpfs_root,
423 	.vfs_quotactl = (void *)eopnotsupp,
424 	.vfs_statvfs = tmpfs_statvfs,
425 	.vfs_sync = tmpfs_sync,
426 	.vfs_vget = tmpfs_vget,
427 	.vfs_fhtovp = tmpfs_fhtovp,
428 	.vfs_vptofh = tmpfs_vptofh,
429 	.vfs_init = tmpfs_init,
430 	.vfs_done = tmpfs_done,
431 	.vfs_snapshot = tmpfs_snapshot,
432 	.vfs_extattrctl = vfs_stdextattrctl,
433 	.vfs_suspendctl = (void *)eopnotsupp,
434 	.vfs_renamelock_enter = genfs_renamelock_enter,
435 	.vfs_renamelock_exit = genfs_renamelock_exit,
436 	.vfs_fsync = (void *)eopnotsupp,
437 	.vfs_opv_descs = tmpfs_vnodeopv_descs
438 };
439 
440 static int
441 tmpfs_modcmd(modcmd_t cmd, void *arg)
442 {
443 
444 	switch (cmd) {
445 	case MODULE_CMD_INIT:
446 		return vfs_attach(&tmpfs_vfsops);
447 	case MODULE_CMD_FINI:
448 		return vfs_detach(&tmpfs_vfsops);
449 	default:
450 		return ENOTTY;
451 	}
452 }
453