xref: /netbsd-src/sys/fs/tmpfs/tmpfs.h (revision 2980e352a13e8f0b545a366830c411e7a542ada8)
1 /*	$NetBSD: tmpfs.h,v 1.36 2008/07/28 18:00:20 pooka Exp $	*/
2 
3 /*
4  * Copyright (c) 2005, 2006, 2007 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 #ifndef _FS_TMPFS_TMPFS_H_
34 #define _FS_TMPFS_TMPFS_H_
35 
36 #include <sys/dirent.h>
37 #include <sys/mount.h>
38 #include <sys/queue.h>
39 #include <sys/vnode.h>
40 
41 #include <fs/tmpfs/tmpfs_pool.h>
42 #include <fs/tmpfs/tmpfs_args.h>
43 
44 /* --------------------------------------------------------------------- */
45 
46 /* Each entry in a directory has a cookie that identifies it.  Cookies
47  * supersede offsets within directories because, given how tmpfs stores
48  * directories in memory, there is no such thing as an offset.  (Emulating
49  * a real offset could be very difficult.)
50  *
51  * The '.', '..' and the end of directory markers have fixed cookies which
52  * cannot collide with the cookies generated by other entries.  The cookies
53  * fot the other entries are generated based on the memory address on which
54  * stores their information is stored.
55  *
56  * Ideally, using the entry's memory pointer as the cookie would be enough
57  * to represent it and it wouldn't cause collisions in any system.
58  * Unfortunately, this results in "offsets" with very large values which
59  * later raise problems in the Linux compatibility layer (and maybe in other
60  * places) as described in PR kern/32034.  Hence we need to workaround this
61  * with a rather ugly hack.
62  *
63  * Linux 32-bit binaries, unless built with _FILE_OFFSET_BITS=64, have off_t
64  * set to 'long', which is a 32-bit *signed* long integer.  Regardless of
65  * the macro value, GLIBC (2.3 at least) always uses the getdents64
66  * system call (when calling readdir) which internally returns off64_t
67  * offsets.  In order to make 32-bit binaries work, *GLIBC* converts the
68  * 64-bit values returned by the kernel to 32-bit ones and aborts with
69  * EOVERFLOW if the conversion results in values that won't fit in 32-bit
70  * integers (which it assumes is because the directory is extremely large).
71  * This wouldn't cause problems if we were dealing with unsigned integers,
72  * but as we have signed integers, this check fails due to sign expansion.
73  *
74  * For example, consider that the kernel returns the 0xc1234567 cookie to
75  * userspace in a off64_t integer.  Later on, GLIBC casts this value to
76  * off_t (remember, signed) with code similar to:
77  *     system call returns the offset in kernel_value;
78  *     off_t casted_value = kernel_value;
79  *     if (sizeof(off_t) != sizeof(off64_t) &&
80  *         kernel_value != casted_value)
81  *             error!
82  * In this case, casted_value still has 0xc1234567, but when it is compared
83  * for equality against kernel_value, it is promoted to a 64-bit integer and
84  * becomes 0xffffffffc1234567, which is different than 0x00000000c1234567.
85  * Then, GLIBC assumes this is because the directory is very large.
86  *
87  * Given that all the above happens in user-space, we have no control over
88  * it; therefore we must workaround the issue here.  We do this by
89  * truncating the pointer value to a 32-bit integer and hope that there
90  * won't be collisions.  In fact, this will not cause any problems in
91  * 32-bit platforms but some might arise in 64-bit machines (I'm not sure
92  * if they can happen at all in practice).
93  *
94  * XXX A nicer solution shall be attempted. */
95 #define	TMPFS_DIRCOOKIE_DOT	0
96 #define	TMPFS_DIRCOOKIE_DOTDOT	1
97 #define	TMPFS_DIRCOOKIE_EOF	2
98 static __inline
99 off_t
100 tmpfs_dircookie(struct tmpfs_dirent *de)
101 {
102 	off_t cookie;
103 
104 	cookie = ((off_t)(uintptr_t)de >> 1) & 0x7FFFFFFF;
105 	KASSERT(cookie != TMPFS_DIRCOOKIE_DOT);
106 	KASSERT(cookie != TMPFS_DIRCOOKIE_DOTDOT);
107 	KASSERT(cookie != TMPFS_DIRCOOKIE_EOF);
108 
109 	return cookie;
110 }
111 
112 /* --------------------------------------------------------------------- */
113 
114 LIST_HEAD(tmpfs_node_list, tmpfs_node);
115 
116 /* --------------------------------------------------------------------- */
117 
118 /*
119  * Internal representation of a tmpfs mount point.
120  */
121 struct tmpfs_mount {
122 	/* Maximum number of memory pages available for use by the file
123 	 * system, set during mount time.  This variable must never be
124 	 * used directly as it may be bigger than the current amount of
125 	 * free memory; in the extreme case, it will hold the SIZE_MAX
126 	 * value.  Instead, use the TMPFS_PAGES_MAX macro. */
127 	unsigned int		tm_pages_max;
128 
129 	/* Number of pages in use by the file system.  Cannot be bigger
130 	 * than the value returned by TMPFS_PAGES_MAX in any case. */
131 	unsigned int		tm_pages_used;
132 
133 	/* Pointer to the node representing the root directory of this
134 	 * file system. */
135 	struct tmpfs_node *	tm_root;
136 
137 	/* Maximum number of possible nodes for this file system; set
138 	 * during mount time.  We need a hard limit on the maximum number
139 	 * of nodes to avoid allocating too much of them; their objects
140 	 * cannot be released until the file system is unmounted.
141 	 * Otherwise, we could easily run out of memory by creating lots
142 	 * of empty files and then simply removing them. */
143 	unsigned int		tm_nodes_max;
144 
145 	/* Number of nodes currently allocated.  This number only grows.
146 	 * When it reaches tm_nodes_max, no more new nodes can be allocated.
147 	 * Of course, the old, unused ones can be reused. */
148 	unsigned int		tm_nodes_cnt;
149 
150 	/* Node list. */
151 	kmutex_t		tm_lock;
152 	struct tmpfs_node_list	tm_nodes;
153 
154 	/* Pools used to store file system meta data.  These are not shared
155 	 * across several instances of tmpfs for the reasons described in
156 	 * tmpfs_pool.c. */
157 	struct tmpfs_pool	tm_dirent_pool;
158 	struct tmpfs_pool	tm_node_pool;
159 	struct tmpfs_str_pool	tm_str_pool;
160 };
161 
162 /* --------------------------------------------------------------------- */
163 
164 /*
165  * This structure maps a file identifier to a tmpfs node.  Used by the
166  * NFS code.
167  */
168 struct tmpfs_fid {
169 	uint16_t		tf_len;
170 	uint16_t		tf_pad;
171 	uint32_t		tf_gen;
172 	ino_t			tf_id;
173 };
174 
175 /* --------------------------------------------------------------------- */
176 
177 /*
178  * Prototypes for tmpfs_subr.c.
179  */
180 
181 int	tmpfs_alloc_node(struct tmpfs_mount *, enum vtype,
182 	    uid_t uid, gid_t gid, mode_t mode, struct tmpfs_node *,
183 	    char *, dev_t, struct tmpfs_node **);
184 void	tmpfs_free_node(struct tmpfs_mount *, struct tmpfs_node *);
185 int	tmpfs_alloc_dirent(struct tmpfs_mount *, struct tmpfs_node *,
186 	    const char *, uint16_t, struct tmpfs_dirent **);
187 void	tmpfs_free_dirent(struct tmpfs_mount *, struct tmpfs_dirent *,
188 	    bool);
189 int	tmpfs_alloc_vp(struct mount *, struct tmpfs_node *, struct vnode **);
190 void	tmpfs_free_vp(struct vnode *);
191 int	tmpfs_alloc_file(struct vnode *, struct vnode **, struct vattr *,
192 	    struct componentname *, char *);
193 void	tmpfs_dir_attach(struct vnode *, struct tmpfs_dirent *);
194 void	tmpfs_dir_detach(struct vnode *, struct tmpfs_dirent *);
195 struct tmpfs_dirent *	tmpfs_dir_lookup(struct tmpfs_node *node,
196 			    struct componentname *cnp);
197 int	tmpfs_dir_getdotdent(struct tmpfs_node *, struct uio *);
198 int	tmpfs_dir_getdotdotdent(struct tmpfs_node *, struct uio *);
199 struct tmpfs_dirent *	tmpfs_dir_lookupbycookie(struct tmpfs_node *, off_t);
200 int	tmpfs_dir_getdents(struct tmpfs_node *, struct uio *, off_t *);
201 int	tmpfs_reg_resize(struct vnode *, off_t);
202 size_t	tmpfs_mem_info(bool);
203 int	tmpfs_chflags(struct vnode *, int, kauth_cred_t, struct lwp *);
204 int	tmpfs_chmod(struct vnode *, mode_t, kauth_cred_t, struct lwp *);
205 int	tmpfs_chown(struct vnode *, uid_t, gid_t, kauth_cred_t, struct lwp *);
206 int	tmpfs_chsize(struct vnode *, u_quad_t, kauth_cred_t, struct lwp *);
207 int	tmpfs_chtimes(struct vnode *, const struct timespec *,
208     const struct timespec *, const struct timespec *, int, kauth_cred_t,
209     struct lwp *);
210 void	tmpfs_itimes(struct vnode *, const struct timespec *,
211 	    const struct timespec *, const struct timespec *);
212 
213 void	tmpfs_update(struct vnode *, const struct timespec *,
214 	    const struct timespec *, const struct timespec *, int);
215 int	tmpfs_truncate(struct vnode *, off_t);
216 
217 /* --------------------------------------------------------------------- */
218 
219 /*
220  * Convenience macros to simplify some logical expressions.
221  */
222 #define IMPLIES(a, b) (!(a) || (b))
223 #define IFF(a, b) (IMPLIES(a, b) && IMPLIES(b, a))
224 
225 /* --------------------------------------------------------------------- */
226 
227 /*
228  * Checks that the directory entry pointed by 'de' matches the name 'name'
229  * with a length of 'len'.
230  */
231 #define TMPFS_DIRENT_MATCHES(de, name, len) \
232     (de->td_namelen == (uint16_t)len && \
233     memcmp((de)->td_name, (name), (de)->td_namelen) == 0)
234 
235 /* --------------------------------------------------------------------- */
236 
237 /*
238  * Ensures that the node pointed by 'node' is a directory and that its
239  * contents are consistent with respect to directories.
240  */
241 #define TMPFS_VALIDATE_DIR(node) \
242     KASSERT((node)->tn_type == VDIR); \
243     KASSERT((node)->tn_size % sizeof(struct tmpfs_dirent) == 0); \
244     KASSERT((node)->tn_spec.tn_dir.tn_readdir_lastp == NULL || \
245         tmpfs_dircookie((node)->tn_spec.tn_dir.tn_readdir_lastp) == \
246         (node)->tn_spec.tn_dir.tn_readdir_lastn);
247 
248 /* --------------------------------------------------------------------- */
249 
250 /*
251  * Memory management stuff.
252  */
253 
254 /* Amount of memory pages to reserve for the system (e.g., to not use by
255  * tmpfs).
256  * XXX: Should this be tunable through sysctl, for instance? */
257 #define TMPFS_PAGES_RESERVED (4 * 1024 * 1024 / PAGE_SIZE)
258 
259 /* Returns the maximum size allowed for a tmpfs file system.  This macro
260  * must be used instead of directly retrieving the value from tm_pages_max.
261  * The reason is that the size of a tmpfs file system is dynamic: it lets
262  * the user store files as long as there is enough free memory (including
263  * physical memory and swap space).  Therefore, the amount of memory to be
264  * used is either the limit imposed by the user during mount time or the
265  * amount of available memory, whichever is lower.  To avoid consuming all
266  * the memory for a given mount point, the system will always reserve a
267  * minimum of TMPFS_PAGES_RESERVED pages, which is also taken into account
268  * by this macro (see above). */
269 static __inline size_t
270 TMPFS_PAGES_MAX(struct tmpfs_mount *tmp)
271 {
272 	size_t freepages;
273 
274 	freepages = tmpfs_mem_info(false);
275 	if (freepages < TMPFS_PAGES_RESERVED)
276 		freepages = 0;
277 	else
278 		freepages -= TMPFS_PAGES_RESERVED;
279 
280 	return MIN(tmp->tm_pages_max, freepages + tmp->tm_pages_used);
281 }
282 
283 /* Returns the available space for the given file system. */
284 #define TMPFS_PAGES_AVAIL(tmp)		\
285     ((ssize_t)(TMPFS_PAGES_MAX(tmp) - (tmp)->tm_pages_used))
286 
287 /* --------------------------------------------------------------------- */
288 
289 /*
290  * Macros/functions to convert from generic data structures to tmpfs
291  * specific ones.
292  */
293 
294 static __inline
295 struct tmpfs_mount *
296 VFS_TO_TMPFS(struct mount *mp)
297 {
298 	struct tmpfs_mount *tmp;
299 
300 #ifdef KASSERT
301 	KASSERT((mp) != NULL && (mp)->mnt_data != NULL);
302 #endif
303 	tmp = (struct tmpfs_mount *)(mp)->mnt_data;
304 	return tmp;
305 }
306 
307 static __inline
308 struct tmpfs_node *
309 VP_TO_TMPFS_DIR(struct vnode *vp)
310 {
311 	struct tmpfs_node *node;
312 
313 	node = VP_TO_TMPFS_NODE(vp);
314 #ifdef KASSERT
315 	TMPFS_VALIDATE_DIR(node);
316 #endif
317 	return node;
318 }
319 #endif /* _FS_TMPFS_TMPFS_H_ */
320