xref: /netbsd-src/sys/fs/tmpfs/tmpfs_vfsops.c (revision da39824b722dbd84beb9a1ab7e8de6710cc44d4b)
1 /*	$NetBSD: tmpfs_vfsops.c,v 1.57 2014/02/06 16:18:38 hannken 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 1.57 2014/02/06 16:18:38 hannken 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 	/* Validate the version. */
107 	if (*data_len < sizeof(*args) ||
108 	    args->ta_version != TMPFS_ARGS_VERSION)
109 		return EINVAL;
110 
111 	/* Handle retrieval of mount point arguments. */
112 	if (mp->mnt_flag & MNT_GETARGS) {
113 		if (mp->mnt_data == NULL)
114 			return EIO;
115 		tmp = VFS_TO_TMPFS(mp);
116 
117 		args->ta_version = TMPFS_ARGS_VERSION;
118 		args->ta_nodes_max = tmp->tm_nodes_max;
119 		args->ta_size_max = tmp->tm_mem_limit;
120 
121 		root = tmp->tm_root;
122 		args->ta_root_uid = root->tn_uid;
123 		args->ta_root_gid = root->tn_gid;
124 		args->ta_root_mode = root->tn_mode;
125 
126 		*data_len = sizeof(*args);
127 		return 0;
128 	}
129 
130 	if (mp->mnt_flag & MNT_UPDATE) {
131 		/* TODO */
132 		return EOPNOTSUPP;
133 	}
134 
135 	/* Prohibit mounts if there is not enough memory. */
136 	if (tmpfs_mem_info(true) < TMPFS_PAGES_RESERVED)
137 		return EINVAL;
138 
139 	/* Get the memory usage limit for this file-system. */
140 	if (args->ta_size_max < PAGE_SIZE) {
141 		memlimit = UINT64_MAX;
142 	} else {
143 		memlimit = args->ta_size_max;
144 	}
145 	KASSERT(memlimit > 0);
146 
147 	if (args->ta_nodes_max <= 3) {
148 		nodes = 3 + (memlimit / 1024);
149 	} else {
150 		nodes = args->ta_nodes_max;
151 	}
152 	nodes = MIN(nodes, INT_MAX);
153 	KASSERT(nodes >= 3);
154 
155 	/* Allocate the tmpfs mount structure and fill it. */
156 	tmp = kmem_zalloc(sizeof(tmpfs_mount_t), KM_SLEEP);
157 	if (tmp == NULL)
158 		return ENOMEM;
159 
160 	tmp->tm_nodes_max = nodes;
161 	tmp->tm_nodes_cnt = 0;
162 	LIST_INIT(&tmp->tm_nodes);
163 
164 	mutex_init(&tmp->tm_lock, MUTEX_DEFAULT, IPL_NONE);
165 	tmpfs_mntmem_init(tmp, memlimit);
166 
167 	/* Allocate the root node. */
168 	error = tmpfs_alloc_node(tmp, VDIR, args->ta_root_uid,
169 	    args->ta_root_gid, args->ta_root_mode & ALLPERMS, NULL,
170 	    VNOVAL, &root);
171 	KASSERT(error == 0 && root != NULL);
172 
173 	/*
174 	 * Parent of the root inode is itself.  Also, root inode has no
175 	 * directory entry (i.e. is never attached), thus hold an extra
176 	 * reference (link) for it.
177 	 */
178 	root->tn_links++;
179 	root->tn_spec.tn_dir.tn_parent = root;
180 	tmp->tm_root = root;
181 
182 	mp->mnt_data = tmp;
183 	mp->mnt_flag |= MNT_LOCAL;
184 	mp->mnt_stat.f_namemax = TMPFS_MAXNAMLEN;
185 	mp->mnt_fs_bshift = PAGE_SHIFT;
186 	mp->mnt_dev_bshift = DEV_BSHIFT;
187 	mp->mnt_iflag |= IMNT_MPSAFE;
188 	vfs_getnewfsid(mp);
189 
190 	error = set_statvfs_info(path, UIO_USERSPACE, "tmpfs", UIO_SYSSPACE,
191 	    mp->mnt_op->vfs_name, mp, curlwp);
192 	if (error) {
193 		(void)tmpfs_unmount(mp, MNT_FORCE);
194 	}
195 	return error;
196 }
197 
198 static int
199 tmpfs_start(struct mount *mp, int flags)
200 {
201 
202 	return 0;
203 }
204 
205 static int
206 tmpfs_unmount(struct mount *mp, int mntflags)
207 {
208 	tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp);
209 	tmpfs_node_t *node, *cnode;
210 	int error, flags = 0;
211 
212 	/* Handle forced unmounts. */
213 	if (mntflags & MNT_FORCE)
214 		flags |= FORCECLOSE;
215 
216 	/* Finalize all pending I/O. */
217 	error = vflush(mp, NULL, flags);
218 	if (error != 0)
219 		return error;
220 
221 	/*
222 	 * First round, detach and destroy all directory entries.
223 	 * Also, clear the pointers to the vnodes - they are gone.
224 	 */
225 	LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
226 		tmpfs_dirent_t *de;
227 
228 		node->tn_vnode = NULL;
229 		if (node->tn_type != VDIR) {
230 			continue;
231 		}
232 		while ((de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir)) != NULL) {
233 			cnode = de->td_node;
234 			if (cnode && cnode != TMPFS_NODE_WHITEOUT) {
235 				cnode->tn_vnode = NULL;
236 			}
237 			tmpfs_dir_detach(node, de);
238 			tmpfs_free_dirent(tmp, de);
239 		}
240 		/* Extra virtual entry (itself for the root). */
241 		node->tn_links--;
242 	}
243 
244 	/* Release the reference on root (diagnostic). */
245 	node = tmp->tm_root;
246 	node->tn_links--;
247 
248 	/* Second round, destroy all inodes. */
249 	while ((node = LIST_FIRST(&tmp->tm_nodes)) != NULL) {
250 		tmpfs_free_node(tmp, node);
251 	}
252 
253 	/* Throw away the tmpfs_mount structure. */
254 	tmpfs_mntmem_destroy(tmp);
255 	mutex_destroy(&tmp->tm_lock);
256 	kmem_free(tmp, sizeof(*tmp));
257 	mp->mnt_data = NULL;
258 
259 	return 0;
260 }
261 
262 static int
263 tmpfs_root(struct mount *mp, vnode_t **vpp)
264 {
265 	tmpfs_node_t *node = VFS_TO_TMPFS(mp)->tm_root;
266 
267 	mutex_enter(&node->tn_vlock);
268 	return tmpfs_vnode_get(mp, node, vpp);
269 }
270 
271 static int
272 tmpfs_vget(struct mount *mp, ino_t ino, vnode_t **vpp)
273 {
274 
275 	return EOPNOTSUPP;
276 }
277 
278 static int
279 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, vnode_t **vpp)
280 {
281 	tmpfs_mount_t *tmp = VFS_TO_TMPFS(mp);
282 	tmpfs_node_t *node;
283 	tmpfs_fid_t tfh;
284 	int error;
285 
286 	if (fhp->fid_len != sizeof(tmpfs_fid_t)) {
287 		return EINVAL;
288 	}
289 	memcpy(&tfh, fhp, sizeof(tmpfs_fid_t));
290 
291 	mutex_enter(&tmp->tm_lock);
292 	LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
293 		if (node->tn_id == tfh.tf_id) {
294 			mutex_enter(&node->tn_vlock);
295 			break;
296 		}
297 	}
298 	mutex_exit(&tmp->tm_lock);
299 
300 	if (node == NULL)
301 		return ESTALE;
302 	/* Will release the tn_vlock. */
303 	if ((error = tmpfs_vnode_get(mp, node, vpp)) != 0)
304 		return error;
305 	if (TMPFS_NODE_GEN(node) != tfh.tf_gen) {
306 		vput(*vpp);
307 		*vpp = NULL;
308 		return ESTALE;
309 	}
310 
311 	return 0;
312 }
313 
314 static int
315 tmpfs_vptofh(vnode_t *vp, struct fid *fhp, size_t *fh_size)
316 {
317 	tmpfs_fid_t tfh;
318 	tmpfs_node_t *node;
319 
320 	if (*fh_size < sizeof(tmpfs_fid_t)) {
321 		*fh_size = sizeof(tmpfs_fid_t);
322 		return E2BIG;
323 	}
324 	*fh_size = sizeof(tmpfs_fid_t);
325 	node = VP_TO_TMPFS_NODE(vp);
326 
327 	memset(&tfh, 0, sizeof(tfh));
328 	tfh.tf_len = sizeof(tmpfs_fid_t);
329 	tfh.tf_gen = TMPFS_NODE_GEN(node);
330 	tfh.tf_id = node->tn_id;
331 	memcpy(fhp, &tfh, sizeof(tfh));
332 
333 	return 0;
334 }
335 
336 static int
337 tmpfs_statvfs(struct mount *mp, struct statvfs *sbp)
338 {
339 	tmpfs_mount_t *tmp;
340 	fsfilcnt_t freenodes;
341 	size_t avail;
342 
343 	tmp = VFS_TO_TMPFS(mp);
344 
345 	sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE;
346 
347 	mutex_enter(&tmp->tm_acc_lock);
348 	avail =  tmpfs_pages_avail(tmp);
349 	sbp->f_blocks = (tmpfs_bytes_max(tmp) >> PAGE_SHIFT);
350 	sbp->f_bavail = sbp->f_bfree = avail;
351 	sbp->f_bresvd = 0;
352 
353 	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt,
354 	    avail * PAGE_SIZE / sizeof(tmpfs_node_t));
355 
356 	sbp->f_files = tmp->tm_nodes_cnt + freenodes;
357 	sbp->f_favail = sbp->f_ffree = freenodes;
358 	sbp->f_fresvd = 0;
359 	mutex_exit(&tmp->tm_acc_lock);
360 
361 	copy_statvfs_info(sbp, mp);
362 
363 	return 0;
364 }
365 
366 static int
367 tmpfs_sync(struct mount *mp, int waitfor, kauth_cred_t uc)
368 {
369 
370 	return 0;
371 }
372 
373 static int
374 tmpfs_snapshot(struct mount *mp, vnode_t *vp, struct timespec *ctime)
375 {
376 
377 	return EOPNOTSUPP;
378 }
379 
380 /*
381  * tmpfs vfs operations.
382  */
383 
384 extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc;
385 extern const struct vnodeopv_desc tmpfs_specop_opv_desc;
386 extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc;
387 
388 const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = {
389 	&tmpfs_fifoop_opv_desc,
390 	&tmpfs_specop_opv_desc,
391 	&tmpfs_vnodeop_opv_desc,
392 	NULL,
393 };
394 
395 struct vfsops tmpfs_vfsops = {
396 	MOUNT_TMPFS,			/* vfs_name */
397 	sizeof (struct tmpfs_args),
398 	tmpfs_mount,			/* vfs_mount */
399 	tmpfs_start,			/* vfs_start */
400 	tmpfs_unmount,			/* vfs_unmount */
401 	tmpfs_root,			/* vfs_root */
402 	(void *)eopnotsupp,		/* vfs_quotactl */
403 	tmpfs_statvfs,			/* vfs_statvfs */
404 	tmpfs_sync,			/* vfs_sync */
405 	tmpfs_vget,			/* vfs_vget */
406 	tmpfs_fhtovp,			/* vfs_fhtovp */
407 	tmpfs_vptofh,			/* vfs_vptofh */
408 	tmpfs_init,			/* vfs_init */
409 	NULL,				/* vfs_reinit */
410 	tmpfs_done,			/* vfs_done */
411 	NULL,				/* vfs_mountroot */
412 	tmpfs_snapshot,			/* vfs_snapshot */
413 	vfs_stdextattrctl,		/* vfs_extattrctl */
414 	(void *)eopnotsupp,		/* vfs_suspendctl */
415 	genfs_renamelock_enter,
416 	genfs_renamelock_exit,
417 	(void *)eopnotsupp,
418 	tmpfs_vnodeopv_descs,
419 	0,				/* vfs_refcount */
420 	{ NULL, NULL },
421 };
422 
423 static int
424 tmpfs_modcmd(modcmd_t cmd, void *arg)
425 {
426 
427 	switch (cmd) {
428 	case MODULE_CMD_INIT:
429 		return vfs_attach(&tmpfs_vfsops);
430 	case MODULE_CMD_FINI:
431 		return vfs_detach(&tmpfs_vfsops);
432 	default:
433 		return ENOTTY;
434 	}
435 }
436