1 /* $NetBSD: tmpfs_vfsops.c,v 1.10 2005/12/11 12:24:29 christos Exp $ */
4 * Copyright (c) 2005 The NetBSD Foundation, Inc.
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
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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.
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.
34 * Efficient memory file system.
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.
43 #include <sys/cdefs.h>
44 __FBSDID("$FreeBSD$");
46 #include <sys/param.h>
47 #include <sys/limits.h>
49 #include <sys/mutex.h>
50 #include <sys/kernel.h>
52 #include <sys/systm.h>
53 #include <sys/sysctl.h>
56 #include <vm/vm_object.h>
57 #include <vm/vm_param.h>
59 #include <fs/tmpfs/tmpfs.h>
62 * Default permission for root node
64 #define TMPFS_DEFAULT_ROOT_MODE (S_IRWXU|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH)
66 MALLOC_DEFINE(M_TMPFSMNT, "tmpfs mount", "tmpfs mount structures");
67 MALLOC_DEFINE(M_TMPFSNAME, "tmpfs name", "tmpfs file names");
69 /* --------------------------------------------------------------------- */
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,
76 static int tmpfs_statfs(struct mount *, struct statfs *);
78 /* --------------------------------------------------------------------- */
80 static const char *tmpfs_opts[] = {
81 "from", "size", "maxfilesize", "inodes", "uid", "gid", "mode", "export",
85 static const char *tmpfs_updateopts[] = {
86 "from", "export", NULL
89 /* --------------------------------------------------------------------- */
92 tmpfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
94 char *opt_value, *vtp;
98 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
101 if (opt_len == 0 || opt_value == NULL)
103 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
105 iv = strtoq(opt_value, &vtp, 0);
106 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
135 tmpfs_node_ctor(void *mem, int size, void *arg, int flags)
137 struct tmpfs_node *node = (struct tmpfs_node *)mem;
144 node->tn_vnode = NULL;
145 node->tn_vpstate = 0;
151 tmpfs_node_dtor(void *mem, int size, void *arg)
153 struct tmpfs_node *node = (struct tmpfs_node *)mem;
154 node->tn_type = VNON;
158 tmpfs_node_init(void *mem, int size, int flags)
160 struct tmpfs_node *node = (struct tmpfs_node *)mem;
163 mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF);
164 node->tn_gen = arc4random();
170 tmpfs_node_fini(void *mem, int size)
172 struct tmpfs_node *node = (struct tmpfs_node *)mem;
174 mtx_destroy(&node->tn_interlock);
178 tmpfs_mount(struct mount *mp)
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;
187 off_t nodes_max, size_max, maxfilesize;
189 /* Root node attributes. */
196 if (vfs_filteropt(mp->mnt_optnew, tmpfs_opts))
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)
203 if (vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0) !=
204 ((struct tmpfs_mount *)mp->mnt_data)->tm_ronly)
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);
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)
226 if (tmpfs_getopt_size(mp->mnt_optnew, "size", &size_max) != 0)
228 if (tmpfs_getopt_size(mp->mnt_optnew, "maxfilesize", &maxfilesize) != 0)
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)
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 < PAGE_SIZE || size_max > OFF_MAX - PAGE_SIZE ||
241 (SIZE_MAX < OFF_MAX && size_max / PAGE_SIZE >= SIZE_MAX))
244 pages = howmany(size_max, PAGE_SIZE);
247 if (nodes_max <= 3) {
248 if (pages < INT_MAX / nodes_per_page)
249 nodes_max = pages * nodes_per_page;
253 if (nodes_max > INT_MAX)
255 MPASS(nodes_max >= 3);
257 /* Allocate the tmpfs mount structure and fill it. */
258 tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount),
259 M_TMPFSMNT, M_WAITOK | M_ZERO);
261 mtx_init(&tmp->allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF);
262 tmp->tm_nodes_max = nodes_max;
263 tmp->tm_nodes_inuse = 0;
264 tmp->tm_maxfilesize = maxfilesize > 0 ? maxfilesize : OFF_MAX;
265 LIST_INIT(&tmp->tm_nodes_used);
267 tmp->tm_pages_max = pages;
268 tmp->tm_pages_used = 0;
269 tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->allnode_lock);
270 tmp->tm_dirent_pool = uma_zcreate("TMPFS dirent",
271 sizeof(struct tmpfs_dirent),
272 NULL, NULL, NULL, NULL,
274 tmp->tm_node_pool = uma_zcreate("TMPFS node",
275 sizeof(struct tmpfs_node),
276 tmpfs_node_ctor, tmpfs_node_dtor,
277 tmpfs_node_init, tmpfs_node_fini,
279 tmp->tm_ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
281 /* Allocate the root node. */
282 error = tmpfs_alloc_node(tmp, VDIR, root_uid,
283 root_gid, root_mode & ALLPERMS, NULL, NULL,
286 if (error != 0 || root == NULL) {
287 uma_zdestroy(tmp->tm_node_pool);
288 uma_zdestroy(tmp->tm_dirent_pool);
289 delete_unrhdr(tmp->tm_ino_unr);
290 free(tmp, M_TMPFSMNT);
293 KASSERT(root->tn_id == 2, ("tmpfs root with invalid ino: %d", root->tn_id));
297 mp->mnt_flag |= MNT_LOCAL;
298 mp->mnt_kern_flag |= MNTK_MPSAFE;
302 mp->mnt_stat.f_namemax = MAXNAMLEN;
304 vfs_mountedfrom(mp, "tmpfs");
309 /* --------------------------------------------------------------------- */
313 tmpfs_unmount(struct mount *mp, int mntflags)
317 struct tmpfs_mount *tmp;
318 struct tmpfs_node *node;
320 /* Handle forced unmounts. */
321 if (mntflags & MNT_FORCE)
324 /* Finalize all pending I/O. */
325 error = vflush(mp, 0, flags, curthread);
329 tmp = VFS_TO_TMPFS(mp);
331 /* Free all associated data. The loop iterates over the linked list
332 * we have containing all used nodes. For each of them that is
333 * a directory, we free all its directory entries. Note that after
334 * freeing a node, it will automatically go to the available list,
335 * so we will later have to iterate over it to release its items. */
336 node = LIST_FIRST(&tmp->tm_nodes_used);
337 while (node != NULL) {
338 struct tmpfs_node *next;
340 if (node->tn_type == VDIR) {
341 struct tmpfs_dirent *de;
343 de = TAILQ_FIRST(&node->tn_dir.tn_dirhead);
345 struct tmpfs_dirent *nde;
347 nde = TAILQ_NEXT(de, td_entries);
348 tmpfs_free_dirent(tmp, de, FALSE);
350 node->tn_size -= sizeof(struct tmpfs_dirent);
354 next = LIST_NEXT(node, tn_entries);
355 tmpfs_free_node(tmp, node);
359 uma_zdestroy(tmp->tm_dirent_pool);
360 uma_zdestroy(tmp->tm_node_pool);
361 delete_unrhdr(tmp->tm_ino_unr);
363 mtx_destroy(&tmp->allnode_lock);
364 MPASS(tmp->tm_pages_used == 0);
365 MPASS(tmp->tm_nodes_inuse == 0);
367 /* Throw away the tmpfs_mount structure. */
368 free(mp->mnt_data, M_TMPFSMNT);
372 mp->mnt_flag &= ~MNT_LOCAL;
377 /* --------------------------------------------------------------------- */
380 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp)
383 error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp);
386 (*vpp)->v_vflag |= VV_ROOT;
391 /* --------------------------------------------------------------------- */
394 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, int flags,
398 struct tmpfs_fid *tfhp;
399 struct tmpfs_mount *tmp;
400 struct tmpfs_node *node;
402 tmp = VFS_TO_TMPFS(mp);
404 tfhp = (struct tmpfs_fid *)fhp;
405 if (tfhp->tf_len != sizeof(struct tmpfs_fid))
408 if (tfhp->tf_id >= tmp->tm_nodes_max)
414 LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) {
415 if (node->tn_id == tfhp->tf_id &&
416 node->tn_gen == tfhp->tf_gen) {
424 return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp));
429 /* --------------------------------------------------------------------- */
433 tmpfs_statfs(struct mount *mp, struct statfs *sbp)
435 struct tmpfs_mount *tmp;
438 tmp = VFS_TO_TMPFS(mp);
440 sbp->f_iosize = PAGE_SIZE;
441 sbp->f_bsize = PAGE_SIZE;
443 used = tmpfs_pages_used(tmp);
444 if (tmp->tm_pages_max != SIZE_MAX)
445 sbp->f_blocks = tmp->tm_pages_max;
447 sbp->f_blocks = used + tmpfs_mem_avail();
448 if (sbp->f_blocks <= used)
451 sbp->f_bavail = sbp->f_blocks - used;
452 sbp->f_bfree = sbp->f_bavail;
453 used = tmp->tm_nodes_inuse;
454 sbp->f_files = tmp->tm_nodes_max;
455 if (sbp->f_files <= used)
458 sbp->f_ffree = sbp->f_files - used;
459 /* sbp->f_owner = tmp->tn_uid; */
464 /* --------------------------------------------------------------------- */
467 * tmpfs vfs operations.
470 struct vfsops tmpfs_vfsops = {
471 .vfs_mount = tmpfs_mount,
472 .vfs_unmount = tmpfs_unmount,
473 .vfs_root = tmpfs_root,
474 .vfs_statfs = tmpfs_statfs,
475 .vfs_fhtovp = tmpfs_fhtovp,
477 VFS_SET(tmpfs_vfsops, tmpfs, 0);