xref: /freebsd/sys/fs/tmpfs/tmpfs_vfsops.c (revision b28624fde638caadd4a89f50c9b7e7da0f98c4d2)
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  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *        This product includes software developed by the NetBSD
22  *        Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Efficient memory file system.
42  *
43  * tmpfs is a file system that uses NetBSD's virtual memory sub-system
44  * (the well-known UVM) to store file data and metadata in an efficient
45  * way.  This means that it does not follow the structure of an on-disk
46  * file system because it simply does not need to.  Instead, it uses
47  * memory-specific data structures and algorithms to automatically
48  * allocate and release resources.
49  */
50 #include <sys/cdefs.h>
51 __FBSDID("$FreeBSD$");
52 
53 #include <sys/param.h>
54 #include <sys/limits.h>
55 #include <sys/lock.h>
56 #include <sys/mutex.h>
57 #include <sys/kernel.h>
58 #include <sys/stat.h>
59 #include <sys/systm.h>
60 #include <sys/sysctl.h>
61 
62 #include <vm/vm.h>
63 #include <vm/vm_object.h>
64 #include <vm/vm_param.h>
65 
66 #include <fs/tmpfs/tmpfs.h>
67 
68 /*
69  * Default permission for root node
70  */
71 #define TMPFS_DEFAULT_ROOT_MODE	(S_IRWXU|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH)
72 
73 MALLOC_DEFINE(M_TMPFSMNT, "tmpfs mount", "tmpfs mount structures");
74 MALLOC_DEFINE(M_TMPFSNAME, "tmpfs name", "tmpfs file names");
75 
76 /* --------------------------------------------------------------------- */
77 
78 static int	tmpfs_mount(struct mount *, struct thread *);
79 static int	tmpfs_unmount(struct mount *, int, struct thread *);
80 static int	tmpfs_root(struct mount *, int flags, struct vnode **,
81 		    struct thread *);
82 static int	tmpfs_fhtovp(struct mount *, struct fid *, struct vnode **);
83 static int	tmpfs_statfs(struct mount *, struct statfs *, struct thread *);
84 
85 /* --------------------------------------------------------------------- */
86 
87 static const char *tmpfs_opts[] = {
88 	"from", "size", "inodes", "uid", "gid", "mode",
89 	NULL
90 };
91 
92 /* --------------------------------------------------------------------- */
93 
94 #define SWI_MAXMIB	3
95 
96 static u_int
97 get_swpgtotal(void)
98 {
99 	struct xswdev xsd;
100 	char *sname = "vm.swap_info";
101 	int soid[SWI_MAXMIB], oid[2];
102 	u_int unswdev, total, dmmax, nswapdev;
103 	size_t mibi, len;
104 
105 	total = 0;
106 
107 	len = sizeof(dmmax);
108 	if (kernel_sysctlbyname(curthread, "vm.dmmax", &dmmax, &len,
109 				NULL, 0, NULL, 0) != 0)
110 		return total;
111 
112 	len = sizeof(nswapdev);
113 	if (kernel_sysctlbyname(curthread, "vm.nswapdev",
114 				&nswapdev, &len,
115 				NULL, 0, NULL, 0) != 0)
116 		return total;
117 
118 	mibi = (SWI_MAXMIB - 1) * sizeof(int);
119 	oid[0] = 0;
120 	oid[1] = 3;
121 
122 	if (kernel_sysctl(curthread, oid, 2,
123 			soid, &mibi, (void *)sname, strlen(sname),
124 			NULL, 0) != 0)
125 		return total;
126 
127 	mibi = (SWI_MAXMIB - 1);
128 	for (unswdev = 0; unswdev < nswapdev; ++unswdev) {
129 		soid[mibi] = unswdev;
130 		len = sizeof(struct xswdev);
131 		if (kernel_sysctl(curthread,
132 				soid, mibi + 1, &xsd, &len, NULL, 0,
133 				NULL, 0) != 0)
134 			return total;
135 		if (len == sizeof(struct xswdev))
136 			total += (xsd.xsw_nblks - dmmax);
137 	}
138 
139 	/* Not Reached */
140 	return total;
141 }
142 
143 /* --------------------------------------------------------------------- */
144 static int
145 tmpfs_node_ctor(void *mem, int size, void *arg, int flags)
146 {
147 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
148 
149 	node->tn_gen++;
150 	node->tn_size = 0;
151 	node->tn_status = 0;
152 	node->tn_flags = 0;
153 	node->tn_links = 0;
154 	node->tn_lockf = NULL;
155 	node->tn_vnode = NULL;
156 	node->tn_vpstate = 0;
157 	node->tn_lookup_dirent = NULL;
158 
159 	return (0);
160 }
161 
162 static void
163 tmpfs_node_dtor(void *mem, int size, void *arg)
164 {
165 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
166 	node->tn_type = VNON;
167 }
168 
169 static int
170 tmpfs_node_init(void *mem, int size, int flags)
171 {
172 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
173 	node->tn_id = 0;
174 
175 	mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF);
176 	node->tn_gen = arc4random();
177 
178 	return (0);
179 }
180 
181 static void
182 tmpfs_node_fini(void *mem, int size)
183 {
184 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
185 
186 	mtx_destroy(&node->tn_interlock);
187 }
188 
189 static int
190 tmpfs_mount(struct mount *mp, struct thread *td)
191 {
192 	struct tmpfs_mount *tmp;
193 	struct tmpfs_node *root;
194 	size_t pages, mem_size;
195 	ino_t nodes;
196 	int error;
197 	/* Size counters. */
198 	ino_t	nodes_max;
199 	off_t	size_max;
200 
201 	/* Root node attributes. */
202 	uid_t	root_uid;
203 	gid_t	root_gid;
204 	mode_t	root_mode;
205 
206 	struct vattr	va;
207 
208 	if (vfs_filteropt(mp->mnt_optnew, tmpfs_opts))
209 		return (EINVAL);
210 
211 	if (mp->mnt_flag & MNT_UPDATE) {
212 		/* XXX: There is no support yet to update file system
213 		 * settings.  Should be added. */
214 
215 		return EOPNOTSUPP;
216 	}
217 
218 	vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY, td);
219 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred, td);
220 	VOP_UNLOCK(mp->mnt_vnodecovered, 0, td);
221 	if (error)
222 		return (error);
223 
224 	if (mp->mnt_cred->cr_ruid != 0 ||
225 	    vfs_scanopt(mp->mnt_optnew, "gid", "%d", &root_gid) != 1)
226 		root_gid = va.va_gid;
227 	if (mp->mnt_cred->cr_ruid != 0 ||
228 	    vfs_scanopt(mp->mnt_optnew, "uid", "%d", &root_uid) != 1)
229 		root_uid = va.va_uid;
230 	if (mp->mnt_cred->cr_ruid != 0 ||
231 	    vfs_scanopt(mp->mnt_optnew, "mode", "%o", &root_mode) != 1)
232 		root_mode = va.va_mode;
233 	if(vfs_scanopt(mp->mnt_optnew, "inodes", "%d", &nodes_max) != 1)
234 		nodes_max = 0;
235 
236 	if(vfs_scanopt(mp->mnt_optnew,
237 			"size",
238 			"%qu", &size_max) != 1)
239 		size_max = 0;
240 
241 	/* Do not allow mounts if we do not have enough memory to preserve
242 	 * the minimum reserved pages. */
243 	mem_size = cnt.v_free_count + cnt.v_inactive_count + get_swpgtotal();
244 	mem_size -= mem_size > cnt.v_wire_count ? cnt.v_wire_count : mem_size;
245 	if (mem_size < TMPFS_PAGES_RESERVED)
246 		return ENOSPC;
247 
248 	/* Get the maximum number of memory pages this file system is
249 	 * allowed to use, based on the maximum size the user passed in
250 	 * the mount structure.  A value of zero is treated as if the
251 	 * maximum available space was requested. */
252 	if (size_max < PAGE_SIZE || size_max >= SIZE_MAX)
253 		pages = SIZE_MAX;
254 	else
255 		pages = howmany(size_max, PAGE_SIZE);
256 	MPASS(pages > 0);
257 
258 	if (nodes_max <= 3)
259 		nodes = 3 + pages * PAGE_SIZE / 1024;
260 	else
261 		nodes = nodes_max;
262 	MPASS(nodes >= 3);
263 
264 	/* Allocate the tmpfs mount structure and fill it. */
265 	tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount),
266 	    M_TMPFSMNT, M_WAITOK | M_ZERO);
267 
268 	mtx_init(&tmp->allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF);
269 	tmp->tm_nodes_max = nodes;
270 	tmp->tm_nodes_inuse = 0;
271 	tmp->tm_maxfilesize = (u_int64_t)(cnt.v_page_count + get_swpgtotal()) * PAGE_SIZE;
272 	LIST_INIT(&tmp->tm_nodes_used);
273 
274 	tmp->tm_pages_max = pages;
275 	tmp->tm_pages_used = 0;
276 	tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->allnode_lock);
277 	tmp->tm_dirent_pool = uma_zcreate(
278 					"TMPFS dirent",
279 					sizeof(struct tmpfs_dirent),
280 					NULL, NULL, NULL, NULL,
281 					UMA_ALIGN_PTR,
282 					0);
283 	tmp->tm_node_pool = uma_zcreate(
284 					"TMPFS node",
285 					sizeof(struct tmpfs_node),
286 					tmpfs_node_ctor, tmpfs_node_dtor,
287 					tmpfs_node_init, tmpfs_node_fini,
288 					UMA_ALIGN_PTR,
289 					0);
290 
291 	/* Allocate the root node. */
292 	error = tmpfs_alloc_node(tmp, VDIR, root_uid,
293 	    root_gid, root_mode & ALLPERMS, NULL, NULL,
294 	    VNOVAL, td, &root);
295 
296 	if (error != 0 || root == NULL) {
297 	    uma_zdestroy(tmp->tm_node_pool);
298 	    uma_zdestroy(tmp->tm_dirent_pool);
299 	    delete_unrhdr(tmp->tm_ino_unr);
300 	    free(tmp, M_TMPFSMNT);
301 	    return error;
302 	}
303 	KASSERT(root->tn_id == 2, ("tmpfs root with invalid ino: %d", root->tn_id));
304 	tmp->tm_root = root;
305 
306 	MNT_ILOCK(mp);
307 	mp->mnt_flag |= MNT_LOCAL;
308 	mp->mnt_kern_flag |= MNTK_MPSAFE;
309 	MNT_IUNLOCK(mp);
310 
311 	mp->mnt_data = tmp;
312 	mp->mnt_stat.f_namemax = MAXNAMLEN;
313 	vfs_getnewfsid(mp);
314 	vfs_mountedfrom(mp, "tmpfs");
315 
316 	return 0;
317 }
318 
319 /* --------------------------------------------------------------------- */
320 
321 /* ARGSUSED2 */
322 static int
323 tmpfs_unmount(struct mount *mp, int mntflags, struct thread *l)
324 {
325 	int error;
326 	int flags = 0;
327 	struct tmpfs_mount *tmp;
328 	struct tmpfs_node *node;
329 
330 	/* Handle forced unmounts. */
331 	if (mntflags & MNT_FORCE)
332 		flags |= FORCECLOSE;
333 
334 	/* Finalize all pending I/O. */
335 	error = vflush(mp, 0, flags, l);
336 	if (error != 0)
337 		return error;
338 
339 	tmp = VFS_TO_TMPFS(mp);
340 
341 	/* Free all associated data.  The loop iterates over the linked list
342 	 * we have containing all used nodes.  For each of them that is
343 	 * a directory, we free all its directory entries.  Note that after
344 	 * freeing a node, it will automatically go to the available list,
345 	 * so we will later have to iterate over it to release its items. */
346 	node = LIST_FIRST(&tmp->tm_nodes_used);
347 	while (node != NULL) {
348 		struct tmpfs_node *next;
349 
350 		if (node->tn_type == VDIR) {
351 			struct tmpfs_dirent *de;
352 
353 			de = TAILQ_FIRST(&node->tn_dir.tn_dirhead);
354 			while (de != NULL) {
355 				struct tmpfs_dirent *nde;
356 
357 				nde = TAILQ_NEXT(de, td_entries);
358 				tmpfs_free_dirent(tmp, de, FALSE);
359 				de = nde;
360 				node->tn_size -= sizeof(struct tmpfs_dirent);
361 			}
362 		}
363 
364 		next = LIST_NEXT(node, tn_entries);
365 		tmpfs_free_node(tmp, node);
366 		node = next;
367 	}
368 
369 	uma_zdestroy(tmp->tm_dirent_pool);
370 	uma_zdestroy(tmp->tm_node_pool);
371 	delete_unrhdr(tmp->tm_ino_unr);
372 
373 	mtx_destroy(&tmp->allnode_lock);
374 	MPASS(tmp->tm_pages_used == 0);
375 	MPASS(tmp->tm_nodes_inuse == 0);
376 
377 	/* Throw away the tmpfs_mount structure. */
378 	free(mp->mnt_data, M_TMPFSMNT);
379 	mp->mnt_data = NULL;
380 
381 	MNT_ILOCK(mp);
382 	mp->mnt_flag &= ~MNT_LOCAL;
383 	MNT_IUNLOCK(mp);
384 	return 0;
385 }
386 
387 /* --------------------------------------------------------------------- */
388 
389 static int
390 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp, struct thread *td)
391 {
392 	int error;
393 	error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp, td);
394 
395 	if (!error)
396 		(*vpp)->v_vflag |= VV_ROOT;
397 
398 	return error;
399 }
400 
401 /* --------------------------------------------------------------------- */
402 
403 static int
404 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
405 {
406 	boolean_t found;
407 	struct tmpfs_fid *tfhp;
408 	struct tmpfs_mount *tmp;
409 	struct tmpfs_node *node;
410 
411 	tmp = VFS_TO_TMPFS(mp);
412 
413 	tfhp = (struct tmpfs_fid *)fhp;
414 	if (tfhp->tf_len != sizeof(struct tmpfs_fid))
415 		return EINVAL;
416 
417 	if (tfhp->tf_id >= tmp->tm_nodes_max)
418 		return EINVAL;
419 
420 	found = FALSE;
421 
422 	TMPFS_LOCK(tmp);
423 	LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) {
424 		if (node->tn_id == tfhp->tf_id &&
425 		    node->tn_gen == tfhp->tf_gen) {
426 			found = TRUE;
427 			break;
428 		}
429 	}
430 	TMPFS_UNLOCK(tmp);
431 
432 	if (found)
433 		return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp, curthread));
434 
435 	return (EINVAL);
436 }
437 
438 /* --------------------------------------------------------------------- */
439 
440 /* ARGSUSED2 */
441 static int
442 tmpfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *l)
443 {
444 	fsfilcnt_t freenodes;
445 	struct tmpfs_mount *tmp;
446 
447 	tmp = VFS_TO_TMPFS(mp);
448 
449 	sbp->f_iosize = PAGE_SIZE;
450 	sbp->f_bsize = PAGE_SIZE;
451 
452 	sbp->f_blocks = TMPFS_PAGES_MAX(tmp);
453 	sbp->f_bavail = sbp->f_bfree = TMPFS_PAGES_AVAIL(tmp);
454 
455 	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_inuse,
456 	    TMPFS_PAGES_AVAIL(tmp) * PAGE_SIZE / sizeof(struct tmpfs_node));
457 
458 	sbp->f_files = freenodes + tmp->tm_nodes_inuse;
459 	sbp->f_ffree = freenodes;
460 	/* sbp->f_owner = tmp->tn_uid; */
461 
462 	return 0;
463 }
464 
465 /* --------------------------------------------------------------------- */
466 
467 /*
468  * tmpfs vfs operations.
469  */
470 
471 struct vfsops tmpfs_vfsops = {
472 	.vfs_mount =			tmpfs_mount,
473 	.vfs_unmount =			tmpfs_unmount,
474 	.vfs_root =			tmpfs_root,
475 	.vfs_statfs =			tmpfs_statfs,
476 	.vfs_fhtovp =			tmpfs_fhtovp,
477 };
478 VFS_SET(tmpfs_vfsops, tmpfs, 0);
479