xref: /freebsd/sys/fs/tmpfs/tmpfs_vfsops.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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 *, struct vnode **);
75 static int	tmpfs_statfs(struct mount *, struct statfs *);
76 
77 /* --------------------------------------------------------------------- */
78 
79 static const char *tmpfs_opts[] = {
80 	"from", "size", "inodes", "uid", "gid", "mode", "export",
81 	NULL
82 };
83 
84 /* --------------------------------------------------------------------- */
85 
86 #define SWI_MAXMIB	3
87 
88 static u_int
89 get_swpgtotal(void)
90 {
91 	struct xswdev xsd;
92 	char *sname = "vm.swap_info";
93 	int soid[SWI_MAXMIB], oid[2];
94 	u_int unswdev, total, dmmax, nswapdev;
95 	size_t mibi, len;
96 
97 	total = 0;
98 
99 	len = sizeof(dmmax);
100 	if (kernel_sysctlbyname(curthread, "vm.dmmax", &dmmax, &len,
101 				NULL, 0, NULL, 0) != 0)
102 		return total;
103 
104 	len = sizeof(nswapdev);
105 	if (kernel_sysctlbyname(curthread, "vm.nswapdev",
106 				&nswapdev, &len,
107 				NULL, 0, NULL, 0) != 0)
108 		return total;
109 
110 	mibi = (SWI_MAXMIB - 1) * sizeof(int);
111 	oid[0] = 0;
112 	oid[1] = 3;
113 
114 	if (kernel_sysctl(curthread, oid, 2,
115 			soid, &mibi, (void *)sname, strlen(sname),
116 			NULL, 0) != 0)
117 		return total;
118 
119 	mibi = (SWI_MAXMIB - 1);
120 	for (unswdev = 0; unswdev < nswapdev; ++unswdev) {
121 		soid[mibi] = unswdev;
122 		len = sizeof(struct xswdev);
123 		if (kernel_sysctl(curthread,
124 				soid, mibi + 1, &xsd, &len, NULL, 0,
125 				NULL, 0) != 0)
126 			return total;
127 		if (len == sizeof(struct xswdev))
128 			total += (xsd.xsw_nblks - dmmax);
129 	}
130 
131 	/* Not Reached */
132 	return total;
133 }
134 
135 /* --------------------------------------------------------------------- */
136 static int
137 tmpfs_node_ctor(void *mem, int size, void *arg, int flags)
138 {
139 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
140 
141 	node->tn_gen++;
142 	node->tn_size = 0;
143 	node->tn_status = 0;
144 	node->tn_flags = 0;
145 	node->tn_links = 0;
146 	node->tn_vnode = NULL;
147 	node->tn_vpstate = 0;
148 
149 	return (0);
150 }
151 
152 static void
153 tmpfs_node_dtor(void *mem, int size, void *arg)
154 {
155 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
156 	node->tn_type = VNON;
157 }
158 
159 static int
160 tmpfs_node_init(void *mem, int size, int flags)
161 {
162 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
163 	node->tn_id = 0;
164 
165 	mtx_init(&node->tn_interlock, "tmpfs node interlock", NULL, MTX_DEF);
166 	node->tn_gen = arc4random();
167 
168 	return (0);
169 }
170 
171 static void
172 tmpfs_node_fini(void *mem, int size)
173 {
174 	struct tmpfs_node *node = (struct tmpfs_node *)mem;
175 
176 	mtx_destroy(&node->tn_interlock);
177 }
178 
179 static int
180 tmpfs_mount(struct mount *mp)
181 {
182 	struct tmpfs_mount *tmp;
183 	struct tmpfs_node *root;
184 	size_t pages, mem_size;
185 	ino_t nodes;
186 	int error;
187 	/* Size counters. */
188 	ino_t	nodes_max;
189 	size_t	size_max;
190 
191 	/* Root node attributes. */
192 	uid_t	root_uid;
193 	gid_t	root_gid;
194 	mode_t	root_mode;
195 
196 	struct vattr	va;
197 
198 	if (vfs_filteropt(mp->mnt_optnew, tmpfs_opts))
199 		return (EINVAL);
200 
201 	if (mp->mnt_flag & MNT_UPDATE) {
202 		/* XXX: There is no support yet to update file system
203 		 * settings.  Should be added. */
204 
205 		return EOPNOTSUPP;
206 	}
207 
208 	printf("WARNING: TMPFS is considered to be a highly experimental "
209 	    "feature in FreeBSD.\n");
210 
211 	vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY);
212 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred);
213 	VOP_UNLOCK(mp->mnt_vnodecovered, 0);
214 	if (error)
215 		return (error);
216 
217 	if (mp->mnt_cred->cr_ruid != 0 ||
218 	    vfs_scanopt(mp->mnt_optnew, "gid", "%d", &root_gid) != 1)
219 		root_gid = va.va_gid;
220 	if (mp->mnt_cred->cr_ruid != 0 ||
221 	    vfs_scanopt(mp->mnt_optnew, "uid", "%d", &root_uid) != 1)
222 		root_uid = va.va_uid;
223 	if (mp->mnt_cred->cr_ruid != 0 ||
224 	    vfs_scanopt(mp->mnt_optnew, "mode", "%ho", &root_mode) != 1)
225 		root_mode = va.va_mode;
226 	if (vfs_scanopt(mp->mnt_optnew, "inodes", "%d", &nodes_max) != 1)
227 		nodes_max = 0;
228 	if (vfs_scanopt(mp->mnt_optnew, "size", "%qu", &size_max) != 1)
229 		size_max = 0;
230 
231 	/* Do not allow mounts if we do not have enough memory to preserve
232 	 * the minimum reserved pages. */
233 	mem_size = cnt.v_free_count + cnt.v_inactive_count + get_swpgtotal();
234 	mem_size -= mem_size > cnt.v_wire_count ? cnt.v_wire_count : mem_size;
235 	if (mem_size < TMPFS_PAGES_RESERVED)
236 		return ENOSPC;
237 
238 	/* Get the maximum number of memory pages this file system is
239 	 * allowed to use, based on the maximum size the user passed in
240 	 * the mount structure.  A value of zero is treated as if the
241 	 * maximum available space was requested. */
242 	if (size_max < PAGE_SIZE || size_max >= SIZE_MAX)
243 		pages = SIZE_MAX;
244 	else
245 		pages = howmany(size_max, PAGE_SIZE);
246 	MPASS(pages > 0);
247 
248 	if (nodes_max <= 3)
249 		nodes = 3 + pages * PAGE_SIZE / 1024;
250 	else
251 		nodes = nodes_max;
252 	MPASS(nodes >= 3);
253 
254 	/* Allocate the tmpfs mount structure and fill it. */
255 	tmp = (struct tmpfs_mount *)malloc(sizeof(struct tmpfs_mount),
256 	    M_TMPFSMNT, M_WAITOK | M_ZERO);
257 
258 	mtx_init(&tmp->allnode_lock, "tmpfs allnode lock", NULL, MTX_DEF);
259 	tmp->tm_nodes_max = nodes;
260 	tmp->tm_nodes_inuse = 0;
261 	tmp->tm_maxfilesize = (u_int64_t)(cnt.v_page_count + get_swpgtotal()) * PAGE_SIZE;
262 	LIST_INIT(&tmp->tm_nodes_used);
263 
264 	tmp->tm_pages_max = pages;
265 	tmp->tm_pages_used = 0;
266 	tmp->tm_ino_unr = new_unrhdr(2, INT_MAX, &tmp->allnode_lock);
267 	tmp->tm_dirent_pool = uma_zcreate("TMPFS dirent",
268 	    sizeof(struct tmpfs_dirent),
269 	    NULL, NULL, NULL, NULL,
270 	    UMA_ALIGN_PTR, 0);
271 	tmp->tm_node_pool = uma_zcreate("TMPFS node",
272 	    sizeof(struct tmpfs_node),
273 	    tmpfs_node_ctor, tmpfs_node_dtor,
274 	    tmpfs_node_init, tmpfs_node_fini,
275 	    UMA_ALIGN_PTR, 0);
276 
277 	/* Allocate the root node. */
278 	error = tmpfs_alloc_node(tmp, VDIR, root_uid,
279 	    root_gid, root_mode & ALLPERMS, NULL, NULL,
280 	    VNOVAL, &root);
281 
282 	if (error != 0 || root == NULL) {
283 	    uma_zdestroy(tmp->tm_node_pool);
284 	    uma_zdestroy(tmp->tm_dirent_pool);
285 	    delete_unrhdr(tmp->tm_ino_unr);
286 	    free(tmp, M_TMPFSMNT);
287 	    return error;
288 	}
289 	KASSERT(root->tn_id == 2, ("tmpfs root with invalid ino: %d", root->tn_id));
290 	tmp->tm_root = root;
291 
292 	MNT_ILOCK(mp);
293 	mp->mnt_flag |= MNT_LOCAL;
294 	mp->mnt_kern_flag |= MNTK_MPSAFE;
295 	MNT_IUNLOCK(mp);
296 
297 	mp->mnt_data = tmp;
298 	mp->mnt_stat.f_namemax = MAXNAMLEN;
299 	vfs_getnewfsid(mp);
300 	vfs_mountedfrom(mp, "tmpfs");
301 
302 	return 0;
303 }
304 
305 /* --------------------------------------------------------------------- */
306 
307 /* ARGSUSED2 */
308 static int
309 tmpfs_unmount(struct mount *mp, int mntflags)
310 {
311 	int error;
312 	int flags = 0;
313 	struct tmpfs_mount *tmp;
314 	struct tmpfs_node *node;
315 
316 	/* Handle forced unmounts. */
317 	if (mntflags & MNT_FORCE)
318 		flags |= FORCECLOSE;
319 
320 	/* Finalize all pending I/O. */
321 	error = vflush(mp, 0, flags, curthread);
322 	if (error != 0)
323 		return error;
324 
325 	tmp = VFS_TO_TMPFS(mp);
326 
327 	/* Free all associated data.  The loop iterates over the linked list
328 	 * we have containing all used nodes.  For each of them that is
329 	 * a directory, we free all its directory entries.  Note that after
330 	 * freeing a node, it will automatically go to the available list,
331 	 * so we will later have to iterate over it to release its items. */
332 	node = LIST_FIRST(&tmp->tm_nodes_used);
333 	while (node != NULL) {
334 		struct tmpfs_node *next;
335 
336 		if (node->tn_type == VDIR) {
337 			struct tmpfs_dirent *de;
338 
339 			de = TAILQ_FIRST(&node->tn_dir.tn_dirhead);
340 			while (de != NULL) {
341 				struct tmpfs_dirent *nde;
342 
343 				nde = TAILQ_NEXT(de, td_entries);
344 				tmpfs_free_dirent(tmp, de, FALSE);
345 				de = nde;
346 				node->tn_size -= sizeof(struct tmpfs_dirent);
347 			}
348 		}
349 
350 		next = LIST_NEXT(node, tn_entries);
351 		tmpfs_free_node(tmp, node);
352 		node = next;
353 	}
354 
355 	uma_zdestroy(tmp->tm_dirent_pool);
356 	uma_zdestroy(tmp->tm_node_pool);
357 	delete_unrhdr(tmp->tm_ino_unr);
358 
359 	mtx_destroy(&tmp->allnode_lock);
360 	MPASS(tmp->tm_pages_used == 0);
361 	MPASS(tmp->tm_nodes_inuse == 0);
362 
363 	/* Throw away the tmpfs_mount structure. */
364 	free(mp->mnt_data, M_TMPFSMNT);
365 	mp->mnt_data = NULL;
366 
367 	MNT_ILOCK(mp);
368 	mp->mnt_flag &= ~MNT_LOCAL;
369 	MNT_IUNLOCK(mp);
370 	return 0;
371 }
372 
373 /* --------------------------------------------------------------------- */
374 
375 static int
376 tmpfs_root(struct mount *mp, int flags, struct vnode **vpp)
377 {
378 	int error;
379 	error = tmpfs_alloc_vp(mp, VFS_TO_TMPFS(mp)->tm_root, flags, vpp);
380 
381 	if (!error)
382 		(*vpp)->v_vflag |= VV_ROOT;
383 
384 	return error;
385 }
386 
387 /* --------------------------------------------------------------------- */
388 
389 static int
390 tmpfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
391 {
392 	boolean_t found;
393 	struct tmpfs_fid *tfhp;
394 	struct tmpfs_mount *tmp;
395 	struct tmpfs_node *node;
396 
397 	tmp = VFS_TO_TMPFS(mp);
398 
399 	tfhp = (struct tmpfs_fid *)fhp;
400 	if (tfhp->tf_len != sizeof(struct tmpfs_fid))
401 		return EINVAL;
402 
403 	if (tfhp->tf_id >= tmp->tm_nodes_max)
404 		return EINVAL;
405 
406 	found = FALSE;
407 
408 	TMPFS_LOCK(tmp);
409 	LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) {
410 		if (node->tn_id == tfhp->tf_id &&
411 		    node->tn_gen == tfhp->tf_gen) {
412 			found = TRUE;
413 			break;
414 		}
415 	}
416 	TMPFS_UNLOCK(tmp);
417 
418 	if (found)
419 		return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp));
420 
421 	return (EINVAL);
422 }
423 
424 /* --------------------------------------------------------------------- */
425 
426 /* ARGSUSED2 */
427 static int
428 tmpfs_statfs(struct mount *mp, struct statfs *sbp)
429 {
430 	fsfilcnt_t freenodes;
431 	struct tmpfs_mount *tmp;
432 
433 	tmp = VFS_TO_TMPFS(mp);
434 
435 	sbp->f_iosize = PAGE_SIZE;
436 	sbp->f_bsize = PAGE_SIZE;
437 
438 	sbp->f_blocks = TMPFS_PAGES_MAX(tmp);
439 	sbp->f_bavail = sbp->f_bfree = TMPFS_PAGES_AVAIL(tmp);
440 
441 	freenodes = MIN(tmp->tm_nodes_max - tmp->tm_nodes_inuse,
442 	    TMPFS_PAGES_AVAIL(tmp) * PAGE_SIZE / sizeof(struct tmpfs_node));
443 
444 	sbp->f_files = freenodes + tmp->tm_nodes_inuse;
445 	sbp->f_ffree = freenodes;
446 	/* sbp->f_owner = tmp->tn_uid; */
447 
448 	return 0;
449 }
450 
451 /* --------------------------------------------------------------------- */
452 
453 /*
454  * tmpfs vfs operations.
455  */
456 
457 struct vfsops tmpfs_vfsops = {
458 	.vfs_mount =			tmpfs_mount,
459 	.vfs_unmount =			tmpfs_unmount,
460 	.vfs_root =			tmpfs_root,
461 	.vfs_statfs =			tmpfs_statfs,
462 	.vfs_fhtovp =			tmpfs_fhtovp,
463 };
464 VFS_SET(tmpfs_vfsops, tmpfs, 0);
465