xref: /freebsd/sys/fs/tmpfs/tmpfs_vfsops.c (revision f0720ed5f83d956846b999aa395d4697503d21cd)
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 /* --------------------------------------------------------------------- */
86 
87 static int
88 tmpfs_node_ctor(void *mem, int size, void *arg, int flags)
89 {
90 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
91 
92 	node->tn_gen++;
93 	node->tn_size = 0;
94 	node->tn_status = 0;
95 	node->tn_flags = 0;
96 	node->tn_links = 0;
97 	node->tn_vnode = NULL;
98 	node->tn_vpstate = 0;
99 
100 	return (0);
101 }
102 
103 static void
104 tmpfs_node_dtor(void *mem, int size, void *arg)
105 {
106 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
107 	node->tn_type = VNON;
108 }
109 
110 static int
111 tmpfs_node_init(void *mem, int size, int flags)
112 {
113 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
114 	node->tn_id = 0;
115 
116 	mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF);
117 	node->tn_gen = arc4random();
118 
119 	return (0);
120 }
121 
122 static void
123 tmpfs_node_fini(void *mem, int size)
124 {
125 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
126 
127 	mtx_destroy(&node->tn_interlock);
128 }
129 
130 static int
131 tmpfs_mount(struct mount *mp)
132 {
133 	struct tmpfs_mount *tmp;
134 	struct tmpfs_node *root;
135 	size_t pages;
136 	uint32_t nodes;
137 	int error;
138 	/* Size counters. */
139 	u_int nodes_max;
140 	u_quad_t size_max, maxfilesize;
141 
142 	/* Root node attributes. */
143 	uid_t root_uid;
144 	gid_t root_gid;
145 	mode_t root_mode;
146 
147 	struct vattr va;
148 
149 	if (vfs_filteropt(mp->mnt_optnew, tmpfs_opts))
150 		return (EINVAL);
151 
152 	if (mp->mnt_flag & MNT_UPDATE) {
153 		/* XXX: There is no support yet to update file system
154 		 * settings.  Should be added. */
155 
156 		return EOPNOTSUPP;
157 	}
158 
159 	printf("WARNING: TMPFS is considered to be a highly experimental "
160 	    "feature in FreeBSD.\n");
161 
162 	vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY);
163 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred);
164 	VOP_UNLOCK(mp->mnt_vnodecovered, 0);
165 	if (error)
166 		return (error);
167 
168 	if (mp->mnt_cred->cr_ruid != 0 ||
169 	    vfs_scanopt(mp->mnt_optnew, "gid", "%d", &root_gid) != 1)
170 		root_gid = va.va_gid;
171 	if (mp->mnt_cred->cr_ruid != 0 ||
172 	    vfs_scanopt(mp->mnt_optnew, "uid", "%d", &root_uid) != 1)
173 		root_uid = va.va_uid;
174 	if (mp->mnt_cred->cr_ruid != 0 ||
175 	    vfs_scanopt(mp->mnt_optnew, "mode", "%ho", &root_mode) != 1)
176 		root_mode = va.va_mode;
177 	if (vfs_scanopt(mp->mnt_optnew, "inodes", "%u", &nodes_max) != 1)
178 		nodes_max = 0;
179 	if (vfs_scanopt(mp->mnt_optnew, "size", "%qu", &size_max) != 1)
180 		size_max = 0;
181 	if (vfs_scanopt(mp->mnt_optnew, "maxfilesize", "%qu",
182 	    &maxfilesize) != 1)
183 		maxfilesize = 0;
184 
185 	/* Do not allow mounts if we do not have enough memory to preserve
186 	 * the minimum reserved pages. */
187 	if (tmpfs_mem_info() < TMPFS_PAGES_RESERVED)
188 		return ENOSPC;
189 
190 	/* Get the maximum number of memory pages this file system is
191 	 * allowed to use, based on the maximum size the user passed in
192 	 * the mount structure.  A value of zero is treated as if the
193 	 * maximum available space was requested. */
194 	if (size_max < PAGE_SIZE || size_max > SIZE_MAX - PAGE_SIZE)
195 		pages = SIZE_MAX;
196 	else
197 		pages = howmany(size_max, PAGE_SIZE);
198 	MPASS(pages > 0);
199 
200 	if (nodes_max <= 3) {
201 		if (pages > UINT32_MAX - 3)
202 			nodes = UINT32_MAX;
203 		else
204 			nodes = pages + 3;
205 	} else
206 		nodes = nodes_max;
207 	MPASS(nodes >= 3);
208 
209 	/* Allocate the tmpfs mount structure and fill it. */
210 	tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount),
211 	    M_TMPFSMNT, M_WAITOK | M_ZERO);
212 
213 	mtx_init(&tmp->allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF);
214 	tmp->tm_nodes_max = nodes;
215 	tmp->tm_nodes_inuse = 0;
216 	tmp->tm_maxfilesize = maxfilesize > 0 ? maxfilesize : UINT64_MAX;
217 	LIST_INIT(&tmp->tm_nodes_used);
218 
219 	tmp->tm_pages_max = pages;
220 	tmp->tm_pages_used = 0;
221 	tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->allnode_lock);
222 	tmp->tm_dirent_pool = uma_zcreate("TMPFS dirent",
223 	    sizeof(struct tmpfs_dirent),
224 	    NULL, NULL, NULL, NULL,
225 	    UMA_ALIGN_PTR, 0);
226 	tmp->tm_node_pool = uma_zcreate("TMPFS node",
227 	    sizeof(struct tmpfs_node),
228 	    tmpfs_node_ctor, tmpfs_node_dtor,
229 	    tmpfs_node_init, tmpfs_node_fini,
230 	    UMA_ALIGN_PTR, 0);
231 
232 	/* Allocate the root node. */
233 	error = tmpfs_alloc_node(tmp, VDIR, root_uid,
234 	    root_gid, root_mode & ALLPERMS, NULL, NULL,
235 	    VNOVAL, &root);
236 
237 	if (error != 0 || root == NULL) {
238 	    uma_zdestroy(tmp->tm_node_pool);
239 	    uma_zdestroy(tmp->tm_dirent_pool);
240 	    delete_unrhdr(tmp->tm_ino_unr);
241 	    free(tmp, M_TMPFSMNT);
242 	    return error;
243 	}
244 	KASSERT(root->tn_id == 2, ("tmpfs root with invalid ino: %d", root->tn_id));
245 	tmp->tm_root = root;
246 
247 	MNT_ILOCK(mp);
248 	mp->mnt_flag |= MNT_LOCAL;
249 	mp->mnt_kern_flag |= MNTK_MPSAFE;
250 	MNT_IUNLOCK(mp);
251 
252 	mp->mnt_data = tmp;
253 	mp->mnt_stat.f_namemax = MAXNAMLEN;
254 	vfs_getnewfsid(mp);
255 	vfs_mountedfrom(mp, "tmpfs");
256 
257 	return 0;
258 }
259 
260 /* --------------------------------------------------------------------- */
261 
262 /* ARGSUSED2 */
263 static int
264 tmpfs_unmount(struct mount *mp, int mntflags)
265 {
266 	int error;
267 	int flags = 0;
268 	struct tmpfs_mount *tmp;
269 	struct tmpfs_node *node;
270 
271 	/* Handle forced unmounts. */
272 	if (mntflags & MNT_FORCE)
273 		flags |= FORCECLOSE;
274 
275 	/* Finalize all pending I/O. */
276 	error = vflush(mp, 0, flags, curthread);
277 	if (error != 0)
278 		return error;
279 
280 	tmp = VFS_TO_TMPFS(mp);
281 
282 	/* Free all associated data.  The loop iterates over the linked list
283 	 * we have containing all used nodes.  For each of them that is
284 	 * a directory, we free all its directory entries.  Note that after
285 	 * freeing a node, it will automatically go to the available list,
286 	 * so we will later have to iterate over it to release its items. */
287 	node = LIST_FIRST(&tmp->tm_nodes_used);
288 	while (node != NULL) {
289 		struct tmpfs_node *next;
290 
291 		if (node->tn_type == VDIR) {
292 			struct tmpfs_dirent *de;
293 
294 			de = TAILQ_FIRST(&node->tn_dir.tn_dirhead);
295 			while (de != NULL) {
296 				struct tmpfs_dirent *nde;
297 
298 				nde = TAILQ_NEXT(de, td_entries);
299 				tmpfs_free_dirent(tmp, de, FALSE);
300 				de = nde;
301 				node->tn_size -= sizeof(struct tmpfs_dirent);
302 			}
303 		}
304 
305 		next = LIST_NEXT(node, tn_entries);
306 		tmpfs_free_node(tmp, node);
307 		node = next;
308 	}
309 
310 	uma_zdestroy(tmp->tm_dirent_pool);
311 	uma_zdestroy(tmp->tm_node_pool);
312 	delete_unrhdr(tmp->tm_ino_unr);
313 
314 	mtx_destroy(&tmp->allnode_lock);
315 	MPASS(tmp->tm_pages_used == 0);
316 	MPASS(tmp->tm_nodes_inuse == 0);
317 
318 	/* Throw away the tmpfs_mount structure. */
319 	free(mp->mnt_data, M_TMPFSMNT);
320 	mp->mnt_data = NULL;
321 
322 	MNT_ILOCK(mp);
323 	mp->mnt_flag &= ~MNT_LOCAL;
324 	MNT_IUNLOCK(mp);
325 	return 0;
326 }
327 
328 /* --------------------------------------------------------------------- */
329 
330 static int
331 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp)
332 {
333 	int error;
334 	error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp);
335 
336 	if (!error)
337 		(*vpp)->v_vflag |= VV_ROOT;
338 
339 	return error;
340 }
341 
342 /* --------------------------------------------------------------------- */
343 
344 static int
345 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, int flags,
346     struct vnode **vpp)
347 {
348 	boolean_t found;
349 	struct tmpfs_fid *tfhp;
350 	struct tmpfs_mount *tmp;
351 	struct tmpfs_node *node;
352 
353 	tmp = VFS_TO_TMPFS(mp);
354 
355 	tfhp = (struct tmpfs_fid *)fhp;
356 	if (tfhp->tf_len != sizeof(struct tmpfs_fid))
357 		return EINVAL;
358 
359 	if (tfhp->tf_id >= tmp->tm_nodes_max)
360 		return EINVAL;
361 
362 	found = FALSE;
363 
364 	TMPFS_LOCK(tmp);
365 	LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) {
366 		if (node->tn_id == tfhp->tf_id &&
367 		    node->tn_gen == tfhp->tf_gen) {
368 			found = TRUE;
369 			break;
370 		}
371 	}
372 	TMPFS_UNLOCK(tmp);
373 
374 	if (found)
375 		return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp));
376 
377 	return (EINVAL);
378 }
379 
380 /* --------------------------------------------------------------------- */
381 
382 /* ARGSUSED2 */
383 static int
384 tmpfs_statfs(struct mount *mp, struct statfs *sbp)
385 {
386 	fsfilcnt_t freenodes;
387 	struct tmpfs_mount *tmp;
388 
389 	tmp = VFS_TO_TMPFS(mp);
390 
391 	sbp->f_iosize = PAGE_SIZE;
392 	sbp->f_bsize = PAGE_SIZE;
393 
394 	sbp->f_blocks = TMPFS_PAGES_MAX(tmp);
395 	sbp->f_bavail = sbp->f_bfree = TMPFS_PAGES_AVAIL(tmp);
396 
397 	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_inuse,
398 	    TMPFS_PAGES_AVAIL(tmp) * PAGE_SIZE / sizeof(struct tmpfs_node));
399 
400 	sbp->f_files = freenodes + tmp->tm_nodes_inuse;
401 	sbp->f_ffree = freenodes;
402 	/* sbp->f_owner = tmp->tn_uid; */
403 
404 	return 0;
405 }
406 
407 /* --------------------------------------------------------------------- */
408 
409 /*
410  * tmpfs vfs operations.
411  */
412 
413 struct vfsops tmpfs_vfsops = {
414 	.vfs_mount =			tmpfs_mount,
415 	.vfs_unmount =			tmpfs_unmount,
416 	.vfs_root =			tmpfs_root,
417 	.vfs_statfs =			tmpfs_statfs,
418 	.vfs_fhtovp =			tmpfs_fhtovp,
419 };
420 VFS_SET(tmpfs_vfsops, tmpfs, 0);
421