xref: /freebsd/sys/fs/tmpfs/tmpfs_vfsops.c (revision 61af1d13936ec56808f62d13dd8698f73b440dc1)
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 	vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY);
160 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred);
161 	VOP_UNLOCK(mp->mnt_vnodecovered, 0);
162 	if (error)
163 		return (error);
164 
165 	if (mp->mnt_cred->cr_ruid != 0 ||
166 	    vfs_scanopt(mp->mnt_optnew, "gid", "%d", &root_gid) != 1)
167 		root_gid = va.va_gid;
168 	if (mp->mnt_cred->cr_ruid != 0 ||
169 	    vfs_scanopt(mp->mnt_optnew, "uid", "%d", &root_uid) != 1)
170 		root_uid = va.va_uid;
171 	if (mp->mnt_cred->cr_ruid != 0 ||
172 	    vfs_scanopt(mp->mnt_optnew, "mode", "%ho", &root_mode) != 1)
173 		root_mode = va.va_mode;
174 	if (vfs_scanopt(mp->mnt_optnew, "inodes", "%u", &nodes_max) != 1)
175 		nodes_max = 0;
176 	if (vfs_scanopt(mp->mnt_optnew, "size", "%qu", &size_max) != 1)
177 		size_max = 0;
178 	if (vfs_scanopt(mp->mnt_optnew, "maxfilesize", "%qu",
179 	    &maxfilesize) != 1)
180 		maxfilesize = 0;
181 
182 	/* Do not allow mounts if we do not have enough memory to preserve
183 	 * the minimum reserved pages. */
184 	if (tmpfs_mem_info() < TMPFS_PAGES_RESERVED)
185 		return ENOSPC;
186 
187 	/* Get the maximum number of memory pages this file system is
188 	 * allowed to use, based on the maximum size the user passed in
189 	 * the mount structure.  A value of zero is treated as if the
190 	 * maximum available space was requested. */
191 	if (size_max < PAGE_SIZE || size_max > SIZE_MAX - PAGE_SIZE)
192 		pages = SIZE_MAX;
193 	else
194 		pages = howmany(size_max, PAGE_SIZE);
195 	MPASS(pages > 0);
196 
197 	if (nodes_max <= 3) {
198 		if (pages > UINT32_MAX - 3)
199 			nodes = UINT32_MAX;
200 		else
201 			nodes = pages + 3;
202 	} else
203 		nodes = nodes_max;
204 	MPASS(nodes >= 3);
205 
206 	/* Allocate the tmpfs mount structure and fill it. */
207 	tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount),
208 	    M_TMPFSMNT, M_WAITOK | M_ZERO);
209 
210 	mtx_init(&tmp->allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF);
211 	tmp->tm_nodes_max = nodes;
212 	tmp->tm_nodes_inuse = 0;
213 	tmp->tm_maxfilesize = maxfilesize > 0 ? maxfilesize : UINT64_MAX;
214 	LIST_INIT(&tmp->tm_nodes_used);
215 
216 	tmp->tm_pages_max = pages;
217 	tmp->tm_pages_used = 0;
218 	tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->allnode_lock);
219 	tmp->tm_dirent_pool = uma_zcreate("TMPFS dirent",
220 	    sizeof(struct tmpfs_dirent),
221 	    NULL, NULL, NULL, NULL,
222 	    UMA_ALIGN_PTR, 0);
223 	tmp->tm_node_pool = uma_zcreate("TMPFS node",
224 	    sizeof(struct tmpfs_node),
225 	    tmpfs_node_ctor, tmpfs_node_dtor,
226 	    tmpfs_node_init, tmpfs_node_fini,
227 	    UMA_ALIGN_PTR, 0);
228 
229 	/* Allocate the root node. */
230 	error = tmpfs_alloc_node(tmp, VDIR, root_uid,
231 	    root_gid, root_mode & ALLPERMS, NULL, NULL,
232 	    VNOVAL, &root);
233 
234 	if (error != 0 || root == NULL) {
235 	    uma_zdestroy(tmp->tm_node_pool);
236 	    uma_zdestroy(tmp->tm_dirent_pool);
237 	    delete_unrhdr(tmp->tm_ino_unr);
238 	    free(tmp, M_TMPFSMNT);
239 	    return error;
240 	}
241 	KASSERT(root->tn_id == 2, ("tmpfs root with invalid ino: %d", root->tn_id));
242 	tmp->tm_root = root;
243 
244 	MNT_ILOCK(mp);
245 	mp->mnt_flag |= MNT_LOCAL;
246 	mp->mnt_kern_flag |= MNTK_MPSAFE;
247 	MNT_IUNLOCK(mp);
248 
249 	mp->mnt_data = tmp;
250 	mp->mnt_stat.f_namemax = MAXNAMLEN;
251 	vfs_getnewfsid(mp);
252 	vfs_mountedfrom(mp, "tmpfs");
253 
254 	return 0;
255 }
256 
257 /* --------------------------------------------------------------------- */
258 
259 /* ARGSUSED2 */
260 static int
261 tmpfs_unmount(struct mount *mp, int mntflags)
262 {
263 	int error;
264 	int flags = 0;
265 	struct tmpfs_mount *tmp;
266 	struct tmpfs_node *node;
267 
268 	/* Handle forced unmounts. */
269 	if (mntflags & MNT_FORCE)
270 		flags |= FORCECLOSE;
271 
272 	/* Finalize all pending I/O. */
273 	error = vflush(mp, 0, flags, curthread);
274 	if (error != 0)
275 		return error;
276 
277 	tmp = VFS_TO_TMPFS(mp);
278 
279 	/* Free all associated data.  The loop iterates over the linked list
280 	 * we have containing all used nodes.  For each of them that is
281 	 * a directory, we free all its directory entries.  Note that after
282 	 * freeing a node, it will automatically go to the available list,
283 	 * so we will later have to iterate over it to release its items. */
284 	node = LIST_FIRST(&tmp->tm_nodes_used);
285 	while (node != NULL) {
286 		struct tmpfs_node *next;
287 
288 		if (node->tn_type == VDIR) {
289 			struct tmpfs_dirent *de;
290 
291 			de = TAILQ_FIRST(&node->tn_dir.tn_dirhead);
292 			while (de != NULL) {
293 				struct tmpfs_dirent *nde;
294 
295 				nde = TAILQ_NEXT(de, td_entries);
296 				tmpfs_free_dirent(tmp, de, FALSE);
297 				de = nde;
298 				node->tn_size -= sizeof(struct tmpfs_dirent);
299 			}
300 		}
301 
302 		next = LIST_NEXT(node, tn_entries);
303 		tmpfs_free_node(tmp, node);
304 		node = next;
305 	}
306 
307 	uma_zdestroy(tmp->tm_dirent_pool);
308 	uma_zdestroy(tmp->tm_node_pool);
309 	delete_unrhdr(tmp->tm_ino_unr);
310 
311 	mtx_destroy(&tmp->allnode_lock);
312 	MPASS(tmp->tm_pages_used == 0);
313 	MPASS(tmp->tm_nodes_inuse == 0);
314 
315 	/* Throw away the tmpfs_mount structure. */
316 	free(mp->mnt_data, M_TMPFSMNT);
317 	mp->mnt_data = NULL;
318 
319 	MNT_ILOCK(mp);
320 	mp->mnt_flag &= ~MNT_LOCAL;
321 	MNT_IUNLOCK(mp);
322 	return 0;
323 }
324 
325 /* --------------------------------------------------------------------- */
326 
327 static int
328 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp)
329 {
330 	int error;
331 	error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp);
332 
333 	if (!error)
334 		(*vpp)->v_vflag |= VV_ROOT;
335 
336 	return error;
337 }
338 
339 /* --------------------------------------------------------------------- */
340 
341 static int
342 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, int flags,
343     struct vnode **vpp)
344 {
345 	boolean_t found;
346 	struct tmpfs_fid *tfhp;
347 	struct tmpfs_mount *tmp;
348 	struct tmpfs_node *node;
349 
350 	tmp = VFS_TO_TMPFS(mp);
351 
352 	tfhp = (struct tmpfs_fid *)fhp;
353 	if (tfhp->tf_len != sizeof(struct tmpfs_fid))
354 		return EINVAL;
355 
356 	if (tfhp->tf_id >= tmp->tm_nodes_max)
357 		return EINVAL;
358 
359 	found = FALSE;
360 
361 	TMPFS_LOCK(tmp);
362 	LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) {
363 		if (node->tn_id == tfhp->tf_id &&
364 		    node->tn_gen == tfhp->tf_gen) {
365 			found = TRUE;
366 			break;
367 		}
368 	}
369 	TMPFS_UNLOCK(tmp);
370 
371 	if (found)
372 		return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp));
373 
374 	return (EINVAL);
375 }
376 
377 /* --------------------------------------------------------------------- */
378 
379 /* ARGSUSED2 */
380 static int
381 tmpfs_statfs(struct mount *mp, struct statfs *sbp)
382 {
383 	fsfilcnt_t freenodes;
384 	struct tmpfs_mount *tmp;
385 
386 	tmp = VFS_TO_TMPFS(mp);
387 
388 	sbp->f_iosize = PAGE_SIZE;
389 	sbp->f_bsize = PAGE_SIZE;
390 
391 	sbp->f_blocks = TMPFS_PAGES_MAX(tmp);
392 	sbp->f_bavail = sbp->f_bfree = TMPFS_PAGES_AVAIL(tmp);
393 
394 	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_inuse,
395 	    TMPFS_PAGES_AVAIL(tmp) * PAGE_SIZE / sizeof(struct tmpfs_node));
396 
397 	sbp->f_files = freenodes + tmp->tm_nodes_inuse;
398 	sbp->f_ffree = freenodes;
399 	/* sbp->f_owner = tmp->tn_uid; */
400 
401 	return 0;
402 }
403 
404 /* --------------------------------------------------------------------- */
405 
406 /*
407  * tmpfs vfs operations.
408  */
409 
410 struct vfsops tmpfs_vfsops = {
411 	.vfs_mount =			tmpfs_mount,
412 	.vfs_unmount =			tmpfs_unmount,
413 	.vfs_root =			tmpfs_root,
414 	.vfs_statfs =			tmpfs_statfs,
415 	.vfs_fhtovp =			tmpfs_fhtovp,
416 };
417 VFS_SET(tmpfs_vfsops, tmpfs, 0);
418