tmpfs_vfsops.c revision 1.50 1 /* $NetBSD: tmpfs_vfsops.c,v 1.50 2011/05/24 20:17:49 rmind 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 1.50 2011/05/24 20:17:49 rmind 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, NULL,
170 VNOVAL, &root);
171 KASSERT(error == 0 && root != NULL);
172 root->tn_links++;
173 tmp->tm_root = root;
174
175 mp->mnt_data = tmp;
176 mp->mnt_flag |= MNT_LOCAL;
177 mp->mnt_stat.f_namemax = MAXNAMLEN;
178 mp->mnt_fs_bshift = PAGE_SHIFT;
179 mp->mnt_dev_bshift = DEV_BSHIFT;
180 mp->mnt_iflag |= IMNT_MPSAFE;
181 vfs_getnewfsid(mp);
182
183 error = set_statvfs_info(path, UIO_USERSPACE, "tmpfs", UIO_SYSSPACE,
184 mp->mnt_op->vfs_name, mp, curlwp);
185 if (error) {
186 (void)tmpfs_unmount(mp, MNT_FORCE);
187 }
188 return error;
189 }
190
191 static int
192 tmpfs_start(struct mount *mp, int flags)
193 {
194
195 return 0;
196 }
197
198 static int
199 tmpfs_unmount(struct mount *mp, int mntflags)
200 {
201 tmpfs_mount_t *tmp;
202 tmpfs_node_t *node;
203 int error, flags = 0;
204
205 /* Handle forced unmounts. */
206 if (mntflags & MNT_FORCE)
207 flags |= FORCECLOSE;
208
209 /* Finalize all pending I/O. */
210 error = vflush(mp, NULL, flags);
211 if (error != 0)
212 return error;
213
214 tmp = VFS_TO_TMPFS(mp);
215
216 /* Destroy any existing inodes. */
217 while ((node = LIST_FIRST(&tmp->tm_nodes)) != NULL) {
218 if (node->tn_type == VDIR) {
219 tmpfs_dirent_t *de;
220
221 /* Destroy any directory entries. */
222 de = TAILQ_FIRST(&node->tn_spec.tn_dir.tn_dir);
223 while (de != NULL) {
224 tmpfs_dirent_t *nde;
225
226 nde = TAILQ_NEXT(de, td_entries);
227 tmpfs_free_dirent(tmp, de, false);
228 node->tn_size -= sizeof(tmpfs_dirent_t);
229 de = nde;
230 }
231 }
232 /* Removes inode from the list. */
233 tmpfs_free_node(tmp, node);
234 }
235
236 /* Throw away the tmpfs_mount structure. */
237 tmpfs_mntmem_destroy(tmp);
238 mutex_destroy(&tmp->tm_lock);
239 kmem_free(tmp, sizeof(*tmp));
240 mp->mnt_data = NULL;
241
242 return 0;
243 }
244
245 static int
246 tmpfs_root(struct mount *mp, vnode_t **vpp)
247 {
248
249 return tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, vpp);
250 }
251
252 static int
253 tmpfs_vget(struct mount *mp, ino_t ino, vnode_t **vpp)
254 {
255
256 printf("tmpfs_vget called; need for it unknown yet\n");
257 return EOPNOTSUPP;
258 }
259
260 static int
261 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, vnode_t **vpp)
262 {
263 tmpfs_mount_t *tmp;
264 tmpfs_node_t *node;
265 tmpfs_fid_t tfh;
266 bool found;
267
268 tmp = VFS_TO_TMPFS(mp);
269
270 if (fhp->fid_len != sizeof(tmpfs_fid_t))
271 return EINVAL;
272
273 memcpy(&tfh, fhp, sizeof(tmpfs_fid_t));
274
275 found = false;
276 mutex_enter(&tmp->tm_lock);
277 LIST_FOREACH(node, &tmp->tm_nodes, tn_entries) {
278 if (node->tn_id == tfh.tf_id &&
279 node->tn_gen == tfh.tf_gen) {
280 found = true;
281 break;
282 }
283 }
284 mutex_exit(&tmp->tm_lock);
285
286 /* XXXAD nothing to prevent 'node' from being removed. */
287 return found ? tmpfs_alloc_vp(mp, node, vpp) : ESTALE;
288 }
289
290 static int
291 tmpfs_vptofh(vnode_t *vp, struct fid *fhp, size_t *fh_size)
292 {
293 tmpfs_fid_t tfh;
294 tmpfs_node_t *node;
295
296 if (*fh_size < sizeof(tmpfs_fid_t)) {
297 *fh_size = sizeof(tmpfs_fid_t);
298 return E2BIG;
299 }
300 *fh_size = sizeof(tmpfs_fid_t);
301 node = VP_TO_TMPFS_NODE(vp);
302
303 memset(&tfh, 0, sizeof(tfh));
304 tfh.tf_len = sizeof(tmpfs_fid_t);
305 tfh.tf_gen = node->tn_gen;
306 tfh.tf_id = node->tn_id;
307 memcpy(fhp, &tfh, sizeof(tfh));
308
309 return 0;
310 }
311
312 static int
313 tmpfs_statvfs(struct mount *mp, struct statvfs *sbp)
314 {
315 tmpfs_mount_t *tmp;
316 fsfilcnt_t freenodes;
317 size_t avail;
318
319 tmp = VFS_TO_TMPFS(mp);
320
321 sbp->f_iosize = sbp->f_frsize = sbp->f_bsize = PAGE_SIZE;
322
323 mutex_enter(&tmp->tm_acc_lock);
324 avail = tmpfs_pages_avail(tmp);
325 sbp->f_blocks = (tmpfs_bytes_max(tmp) >> PAGE_SHIFT);
326 sbp->f_bavail = sbp->f_bfree = avail;
327 sbp->f_bresvd = 0;
328
329 freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_cnt,
330 avail * PAGE_SIZE / sizeof(tmpfs_node_t));
331
332 sbp->f_files = tmp->tm_nodes_cnt + freenodes;
333 sbp->f_favail = sbp->f_ffree = freenodes;
334 sbp->f_fresvd = 0;
335 mutex_exit(&tmp->tm_acc_lock);
336
337 copy_statvfs_info(sbp, mp);
338
339 return 0;
340 }
341
342 static int
343 tmpfs_sync(struct mount *mp, int waitfor, kauth_cred_t uc)
344 {
345
346 return 0;
347 }
348
349 static int
350 tmpfs_snapshot(struct mount *mp, vnode_t *vp, struct timespec *ctime)
351 {
352
353 return EOPNOTSUPP;
354 }
355
356 /*
357 * tmpfs vfs operations.
358 */
359
360 extern const struct vnodeopv_desc tmpfs_fifoop_opv_desc;
361 extern const struct vnodeopv_desc tmpfs_specop_opv_desc;
362 extern const struct vnodeopv_desc tmpfs_vnodeop_opv_desc;
363
364 const struct vnodeopv_desc * const tmpfs_vnodeopv_descs[] = {
365 &tmpfs_fifoop_opv_desc,
366 &tmpfs_specop_opv_desc,
367 &tmpfs_vnodeop_opv_desc,
368 NULL,
369 };
370
371 struct vfsops tmpfs_vfsops = {
372 MOUNT_TMPFS, /* vfs_name */
373 sizeof (struct tmpfs_args),
374 tmpfs_mount, /* vfs_mount */
375 tmpfs_start, /* vfs_start */
376 tmpfs_unmount, /* vfs_unmount */
377 tmpfs_root, /* vfs_root */
378 (void *)eopnotsupp, /* vfs_quotactl */
379 tmpfs_statvfs, /* vfs_statvfs */
380 tmpfs_sync, /* vfs_sync */
381 tmpfs_vget, /* vfs_vget */
382 tmpfs_fhtovp, /* vfs_fhtovp */
383 tmpfs_vptofh, /* vfs_vptofh */
384 tmpfs_init, /* vfs_init */
385 NULL, /* vfs_reinit */
386 tmpfs_done, /* vfs_done */
387 NULL, /* vfs_mountroot */
388 tmpfs_snapshot, /* vfs_snapshot */
389 vfs_stdextattrctl, /* vfs_extattrctl */
390 (void *)eopnotsupp, /* vfs_suspendctl */
391 genfs_renamelock_enter,
392 genfs_renamelock_exit,
393 (void *)eopnotsupp,
394 tmpfs_vnodeopv_descs,
395 0, /* vfs_refcount */
396 { NULL, NULL },
397 };
398
399 static int
400 tmpfs_modcmd(modcmd_t cmd, void *arg)
401 {
402
403 switch (cmd) {
404 case MODULE_CMD_INIT:
405 return vfs_attach(&tmpfs_vfsops);
406 case MODULE_CMD_FINI:
407 return vfs_detach(&tmpfs_vfsops);
408 default:
409 return ENOTTY;
410 }
411 }
412