1 /* $NetBSD: tmpfs_vfsops.c,v 1.10 2005/12/11 12:24:29 christos Exp $ */
2
3 /*-
4 * Copyright (c) 2005 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 #include <sys/cdefs.h>
44 __FBSDID("$FreeBSD$");
45
46 #include <sys/param.h>
47 #include <sys/limits.h>
48 #include <sys/lock.h>
49 #include <sys/mutex.h>
50 #include <sys/kernel.h>
51 #include <sys/stat.h>
52 #include <sys/systm.h>
53 #include <sys/sysctl.h>
54
55 #include <vm/vm.h>
56 #include <vm/vm_object.h>
57 #include <vm/vm_param.h>
58
59 #include <fs/tmpfs/tmpfs.h>
60
61 /*
62 * Default permission for root node
63 */
64 #define TMPFS_DEFAULT_ROOT_MODE (S_IRWXU|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH)
65
66 MALLOC_DEFINE(M_TMPFSMNT, "tmpfs mount", "tmpfs mount structures");
67 MALLOC_DEFINE(M_TMPFSNAME, "tmpfs name", "tmpfs file names");
68
69 /* --------------------------------------------------------------------- */
70
71 static int tmpfs_mount(struct mount *);
72 static int tmpfs_unmount(struct mount *, int);
73 static int tmpfs_root(struct mount *, int flags, struct vnode **);
74 static int tmpfs_fhtovp(struct mount *, struct fid *, int,
75 struct vnode **);
76 static int tmpfs_statfs(struct mount *, struct statfs *);
77
78 /* --------------------------------------------------------------------- */
79
80 static const char *tmpfs_opts[] = {
81 "from", "size", "maxfilesize", "inodes", "uid", "gid", "mode", "export",
82 NULL
83 };
84
85 static const char *tmpfs_updateopts[] = {
86 "from", "export", NULL
87 };
88
89 /* --------------------------------------------------------------------- */
90
91 static int
92 tmpfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
93 {
94 char *opt_value, *vtp;
95 quad_t iv;
96 int error, opt_len;
97
98 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
99 if (error != 0)
100 return (error);
101 if (opt_len == 0 || opt_value == NULL)
102 return (EINVAL);
103 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
104 return (EINVAL);
105 iv = strtoq(opt_value, &vtp, 0);
106 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
107 return (EINVAL);
108 if (iv < 0)
109 return (EINVAL);
110 switch (vtp[0]) {
111 case 't':
112 case 'T':
113 iv *= 1024;
114 case 'g':
115 case 'G':
116 iv *= 1024;
117 case 'm':
118 case 'M':
119 iv *= 1024;
120 case 'k':
121 case 'K':
122 iv *= 1024;
123 case '\0':
124 break;
125 default:
126 return (EINVAL);
127 }
128 *value = iv;
129
130 return (0);
131 }
132
133
134 static int
135 tmpfs_node_ctor(void *mem, int size, void *arg, int flags)
136 {
137 struct tmpfs_node *node = (struct tmpfs_node *)mem;
138
139 node->tn_gen++;
140 node->tn_size = 0;
141 node->tn_status = 0;
142 node->tn_flags = 0;
143 node->tn_links = 0;
144 node->tn_vnode = NULL;
145 node->tn_vpstate = 0;
146
147 return (0);
148 }
149
150 static void
151 tmpfs_node_dtor(void *mem, int size, void *arg)
152 {
153 struct tmpfs_node *node = (struct tmpfs_node *)mem;
154 node->tn_type = VNON;
155 }
156
157 static int
158 tmpfs_node_init(void *mem, int size, int flags)
159 {
160 struct tmpfs_node *node = (struct tmpfs_node *)mem;
161 node->tn_id = 0;
162
163 mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF);
164 node->tn_gen = arc4random();
165
166 return (0);
167 }
168
169 static void
170 tmpfs_node_fini(void *mem, int size)
171 {
172 struct tmpfs_node *node = (struct tmpfs_node *)mem;
173
174 mtx_destroy(&node->tn_interlock);
175 }
176
177 static int
178 tmpfs_mount(struct mount *mp)
179 {
180 const size_t nodes_per_page = howmany(PAGE_SIZE,
181 sizeof(struct tmpfs_dirent) + sizeof(struct tmpfs_node));
182 struct tmpfs_mount *tmp;
183 struct tmpfs_node *root;
184 int error;
185 /* Size counters. */
186 u_quad_t pages;
187 off_t nodes_max, size_max, maxfilesize;
188
189 /* Root node attributes. */
190 uid_t root_uid;
191 gid_t root_gid;
192 mode_t root_mode;
193
194 struct vattr va;
195
196 if (vfs_filteropt(mp->mnt_optnew, tmpfs_opts))
197 return (EINVAL);
198
199 if (mp->mnt_flag & MNT_UPDATE) {
200 /* Only support update mounts for certain options. */
201 if (vfs_filteropt(mp->mnt_optnew, tmpfs_updateopts) != 0)
202 return (EOPNOTSUPP);
203 if (vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0) !=
204 ((struct tmpfs_mount *)mp->mnt_data)->tm_ronly)
205 return (EOPNOTSUPP);
206 return (0);
207 }
208
209 vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY);
210 error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred);
211 VOP_UNLOCK(mp->mnt_vnodecovered, 0);
212 if (error)
213 return (error);
214
215 if (mp->mnt_cred->cr_ruid != 0 ||
216 vfs_scanopt(mp->mnt_optnew, "gid", "%d", &root_gid) != 1)
217 root_gid = va.va_gid;
218 if (mp->mnt_cred->cr_ruid != 0 ||
219 vfs_scanopt(mp->mnt_optnew, "uid", "%d", &root_uid) != 1)
220 root_uid = va.va_uid;
221 if (mp->mnt_cred->cr_ruid != 0 ||
222 vfs_scanopt(mp->mnt_optnew, "mode", "%ho", &root_mode) != 1)
223 root_mode = va.va_mode;
224 if (tmpfs_getopt_size(mp->mnt_optnew, "inodes", &nodes_max) != 0)
225 nodes_max = 0;
226 if (tmpfs_getopt_size(mp->mnt_optnew, "size", &size_max) != 0)
227 size_max = 0;
228 if (tmpfs_getopt_size(mp->mnt_optnew, "maxfilesize", &maxfilesize) != 0)
229 maxfilesize = 0;
230
231 /* Do not allow mounts if we do not have enough memory to preserve
232 * the minimum reserved pages. */
233 if (tmpfs_mem_avail() < TMPFS_PAGES_MINRESERVED)
234 return ENOSPC;
235
236 /* Get the maximum number of memory pages this file system is
237 * allowed to use, based on the maximum size the user passed in
238 * the mount structure. A value of zero is treated as if the
239 * maximum available space was requested. */
240 if (size_max == 0 || size_max > OFF_MAX - PAGE_SIZE ||
241 (SIZE_MAX < OFF_MAX && size_max / PAGE_SIZE >= SIZE_MAX))
242 pages = SIZE_MAX;
243 else {
244 size_max = roundup(size_max, PAGE_SIZE);
245 pages = howmany(size_max, PAGE_SIZE);
246 }
247 MPASS(pages > 0);
248
249 if (nodes_max <= 3) {
250 if (pages < INT_MAX / nodes_per_page)
251 nodes_max = pages * nodes_per_page;
252 else
253 nodes_max = INT_MAX;
254 }
255 if (nodes_max > INT_MAX)
256 nodes_max = INT_MAX;
257 MPASS(nodes_max >= 3);
258
259 /* Allocate the tmpfs mount structure and fill it. */
260 tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount),
261 M_TMPFSMNT, M_WAITOK | M_ZERO);
262
263 mtx_init(&tmp->allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF);
264 tmp->tm_nodes_max = nodes_max;
265 tmp->tm_nodes_inuse = 0;
266 tmp->tm_maxfilesize = maxfilesize > 0 ? maxfilesize : OFF_MAX;
267 LIST_INIT(&tmp->tm_nodes_used);
268
269 tmp->tm_pages_max = pages;
270 tmp->tm_pages_used = 0;
271 tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->allnode_lock);
272 tmp->tm_dirent_pool = uma_zcreate("TMPFS dirent",
273 sizeof(struct tmpfs_dirent),
274 NULL, NULL, NULL, NULL,
275 UMA_ALIGN_PTR, 0);
276 tmp->tm_node_pool = uma_zcreate("TMPFS node",
277 sizeof(struct tmpfs_node),
278 tmpfs_node_ctor, tmpfs_node_dtor,
279 tmpfs_node_init, tmpfs_node_fini,
280 UMA_ALIGN_PTR, 0);
281 tmp->tm_ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
282
283 /* Allocate the root node. */
284 error = tmpfs_alloc_node(tmp, VDIR, root_uid,
285 root_gid, root_mode & ALLPERMS, NULL, NULL,
286 VNOVAL, &root);
287
288 if (error != 0 || root == NULL) {
289 uma_zdestroy(tmp->tm_node_pool);
290 uma_zdestroy(tmp->tm_dirent_pool);
291 delete_unrhdr(tmp->tm_ino_unr);
292 free(tmp, M_TMPFSMNT);
293 return error;
294 }
295 KASSERT(root->tn_id == 2, ("tmpfs root with invalid ino: %d", root->tn_id));
296 tmp->tm_root = root;
297
298 MNT_ILOCK(mp);
299 mp->mnt_flag |= MNT_LOCAL;
300 mp->mnt_kern_flag |= MNTK_MPSAFE;
301 MNT_IUNLOCK(mp);
302
303 mp->mnt_data = tmp;
304 mp->mnt_stat.f_namemax = MAXNAMLEN;
305 vfs_getnewfsid(mp);
306 vfs_mountedfrom(mp, "tmpfs");
307
308 return 0;
309 }
310
311 /* --------------------------------------------------------------------- */
312
313 /* ARGSUSED2 */
314 static int
315 tmpfs_unmount(struct mount *mp, int mntflags)
316 {
317 int error;
318 int flags = 0;
319 struct tmpfs_mount *tmp;
320 struct tmpfs_node *node;
321
322 /* Handle forced unmounts. */
323 if (mntflags & MNT_FORCE)
324 flags |= FORCECLOSE;
325
326 /* Finalize all pending I/O. */
327 error = vflush(mp, 0, flags, curthread);
328 if (error != 0)
329 return error;
330
331 tmp = VFS_TO_TMPFS(mp);
332
333 /* Free all associated data. The loop iterates over the linked list
334 * we have containing all used nodes. For each of them that is
335 * a directory, we free all its directory entries. Note that after
336 * freeing a node, it will automatically go to the available list,
337 * so we will later have to iterate over it to release its items. */
338 node = LIST_FIRST(&tmp->tm_nodes_used);
339 while (node != NULL) {
340 struct tmpfs_node *next;
341
342 if (node->tn_type == VDIR) {
343 struct tmpfs_dirent *de;
344
345 de = TAILQ_FIRST(&node->tn_dir.tn_dirhead);
346 while (de != NULL) {
347 struct tmpfs_dirent *nde;
348
349 nde = TAILQ_NEXT(de, td_entries);
350 tmpfs_free_dirent(tmp, de, FALSE);
351 de = nde;
352 node->tn_size -= sizeof(struct tmpfs_dirent);
353 }
354 }
355
356 next = LIST_NEXT(node, tn_entries);
357 tmpfs_free_node(tmp, node);
358 node = next;
359 }
360
361 uma_zdestroy(tmp->tm_dirent_pool);
362 uma_zdestroy(tmp->tm_node_pool);
363 delete_unrhdr(tmp->tm_ino_unr);
364
365 mtx_destroy(&tmp->allnode_lock);
366 MPASS(tmp->tm_pages_used == 0);
367 MPASS(tmp->tm_nodes_inuse == 0);
368
369 /* Throw away the tmpfs_mount structure. */
370 free(mp->mnt_data, M_TMPFSMNT);
371 mp->mnt_data = NULL;
372
373 MNT_ILOCK(mp);
374 mp->mnt_flag &= ~MNT_LOCAL;
375 MNT_IUNLOCK(mp);
376 return 0;
377 }
378
379 /* --------------------------------------------------------------------- */
380
381 static int
382 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp)
383 {
384 int error;
385 error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp);
386
387 if (!error)
388 (*vpp)->v_vflag |= VV_ROOT;
389
390 return error;
391 }
392
393 /* --------------------------------------------------------------------- */
394
395 static int
396 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, int flags,
397 struct vnode **vpp)
398 {
399 boolean_t found;
400 struct tmpfs_fid *tfhp;
401 struct tmpfs_mount *tmp;
402 struct tmpfs_node *node;
403
404 tmp = VFS_TO_TMPFS(mp);
405
406 tfhp = (struct tmpfs_fid *)fhp;
407 if (tfhp->tf_len != sizeof(struct tmpfs_fid))
408 return EINVAL;
409
410 if (tfhp->tf_id >= tmp->tm_nodes_max)
411 return EINVAL;
412
413 found = FALSE;
414
415 TMPFS_LOCK(tmp);
416 LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) {
417 if (node->tn_id == tfhp->tf_id &&
418 node->tn_gen == tfhp->tf_gen) {
419 found = TRUE;
420 break;
421 }
422 }
423 TMPFS_UNLOCK(tmp);
424
425 if (found)
426 return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp));
427
428 return (EINVAL);
429 }
430
431 /* --------------------------------------------------------------------- */
432
433 /* ARGSUSED2 */
434 static int
435 tmpfs_statfs(struct mount *mp, struct statfs *sbp)
436 {
437 struct tmpfs_mount *tmp;
438 size_t used;
439
440 tmp = VFS_TO_TMPFS(mp);
441
442 sbp->f_iosize = PAGE_SIZE;
443 sbp->f_bsize = PAGE_SIZE;
444
445 used = tmpfs_pages_used(tmp);
446 if (tmp->tm_pages_max != SIZE_MAX)
447 sbp->f_blocks = tmp->tm_pages_max;
448 else
449 sbp->f_blocks = used + tmpfs_mem_avail();
450 if (sbp->f_blocks <= used)
451 sbp->f_bavail = 0;
452 else
453 sbp->f_bavail = sbp->f_blocks - used;
454 sbp->f_bfree = sbp->f_bavail;
455 used = tmp->tm_nodes_inuse;
456 sbp->f_files = tmp->tm_nodes_max;
457 if (sbp->f_files <= used)
458 sbp->f_ffree = 0;
459 else
460 sbp->f_ffree = sbp->f_files - used;
461 /* sbp->f_owner = tmp->tn_uid; */
462
463 return 0;
464 }
465
466 /* --------------------------------------------------------------------- */
467
468 /*
469 * tmpfs vfs operations.
470 */
471
472 struct vfsops tmpfs_vfsops = {
473 .vfs_mount = tmpfs_mount,
474 .vfs_unmount = tmpfs_unmount,
475 .vfs_root = tmpfs_root,
476 .vfs_statfs = tmpfs_statfs,
477 .vfs_fhtovp = tmpfs_fhtovp,
478 };
479 VFS_SET(tmpfs_vfsops, tmpfs, 0);
Cache object: 1ce1117a0f6842c90bfd937eb63fdfed
|