xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/zfs_znode.h (revision cdf9f8c9206117516a14f3c70f86510326c3d5fa)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef	_SYS_FS_ZFS_ZNODE_H
27 #define	_SYS_FS_ZFS_ZNODE_H
28 
29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
30 
31 #ifdef _KERNEL
32 #include <sys/isa_defs.h>
33 #include <sys/types32.h>
34 #include <sys/list.h>
35 #include <sys/dmu.h>
36 #include <sys/zfs_vfsops.h>
37 #endif
38 #include <sys/zfs_acl.h>
39 #include <sys/zil.h>
40 
41 #ifdef	__cplusplus
42 extern "C" {
43 #endif
44 
45 /*
46  * Define special zfs pflags
47  */
48 #define	ZFS_XATTR	0x1		/* is an extended attribute */
49 #define	ZFS_INHERIT_ACE	0x2		/* ace has inheritable ACEs */
50 #define	ZFS_ACL_TRIVIAL 0x4		/* files ACL is trivial */
51 
52 #define	MASTER_NODE_OBJ	1
53 
54 /*
55  * special attributes for master node.
56  */
57 #define	ZFS_FSID		"FSID"
58 #define	ZFS_UNLINKED_SET	"DELETE_QUEUE"
59 #define	ZFS_ROOT_OBJ		"ROOT"
60 #define	ZPL_VERSION_STR		"VERSION"
61 
62 
63 #define	ZFS_MAX_BLOCKSIZE	(SPA_MAXBLOCKSIZE)
64 
65 /* Path component length */
66 /*
67  * The generic fs code uses MAXNAMELEN to represent
68  * what the largest component length is.  Unfortunately,
69  * this length includes the terminating NULL.  ZFS needs
70  * to tell the users via pathconf() and statvfs() what the
71  * true maximum length of a component is, excluding the NULL.
72  */
73 #define	ZFS_MAXNAMELEN	(MAXNAMELEN - 1)
74 
75 /*
76  * Convert mode bits (zp_mode) to BSD-style DT_* values for storing in
77  * the directory entries.
78  */
79 #define	IFTODT(mode) (((mode) & S_IFMT) >> 12)
80 
81 /*
82  * The directory entry has the type (currently unused on Solaris) in the
83  * top 4 bits, and the object number in the low 48 bits.  The "middle"
84  * 12 bits are unused.
85  */
86 #define	ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
87 #define	ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
88 
89 /*
90  * This is the persistent portion of the znode.  It is stored
91  * in the "bonus buffer" of the file.  Short symbolic links
92  * are also stored in the bonus buffer.
93  */
94 typedef struct znode_phys {
95 	uint64_t zp_atime[2];		/*  0 - last file access time */
96 	uint64_t zp_mtime[2];		/* 16 - last file modification time */
97 	uint64_t zp_ctime[2];		/* 32 - last file change time */
98 	uint64_t zp_crtime[2];		/* 48 - creation time */
99 	uint64_t zp_gen;		/* 64 - generation (txg of creation) */
100 	uint64_t zp_mode;		/* 72 - file mode bits */
101 	uint64_t zp_size;		/* 80 - size of file */
102 	uint64_t zp_parent;		/* 88 - directory parent (`..') */
103 	uint64_t zp_links;		/* 96 - number of links to file */
104 	uint64_t zp_xattr;		/* 104 - DMU object for xattrs */
105 	uint64_t zp_rdev;		/* 112 - dev_t for VBLK & VCHR files */
106 	uint64_t zp_flags;		/* 120 - persistent flags */
107 	uint64_t zp_uid;		/* 128 - file owner */
108 	uint64_t zp_gid;		/* 136 - owning group */
109 	uint64_t zp_pad[4];		/* 144 - future */
110 	zfs_znode_acl_t zp_acl;		/* 176 - 263 ACL */
111 	/*
112 	 * Data may pad out any remaining bytes in the znode buffer, eg:
113 	 *
114 	 * |<---------------------- dnode_phys (512) ------------------------>|
115 	 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
116 	 *			|<---- znode (264) ---->|<---- data (56) ---->|
117 	 *
118 	 * At present, we only use this space to store symbolic links.
119 	 */
120 } znode_phys_t;
121 
122 /*
123  * Directory entry locks control access to directory entries.
124  * They are used to protect creates, deletes, and renames.
125  * Each directory znode has a mutex and a list of locked names.
126  */
127 #ifdef _KERNEL
128 typedef struct zfs_dirlock {
129 	char		*dl_name;	/* directory entry being locked */
130 	uint32_t	dl_sharecnt;	/* 0 if exclusive, > 0 if shared */
131 	uint16_t	dl_namesize;	/* set if dl_name was allocated */
132 	kcondvar_t	dl_cv;		/* wait for entry to be unlocked */
133 	struct znode	*dl_dzp;	/* directory znode */
134 	struct zfs_dirlock *dl_next;	/* next in z_dirlocks list */
135 } zfs_dirlock_t;
136 
137 typedef struct znode {
138 	struct zfsvfs	*z_zfsvfs;
139 	vnode_t		*z_vnode;
140 	uint64_t	z_id;		/* object ID for this znode */
141 	kmutex_t	z_lock;		/* znode modification lock */
142 	krwlock_t	z_map_lock;	/* page map lock */
143 	krwlock_t	z_parent_lock;	/* parent lock for directories */
144 	krwlock_t	z_name_lock;	/* "master" lock for dirent locks */
145 	zfs_dirlock_t	*z_dirlocks;	/* directory entry lock list */
146 	kmutex_t	z_range_lock;	/* protects changes to z_range_avl */
147 	avl_tree_t	z_range_avl;	/* avl tree of file range locks */
148 	uint8_t		z_unlinked;	/* file has been unlinked */
149 	uint8_t		z_atime_dirty;	/* atime needs to be synced */
150 	uint8_t		z_dbuf_held;	/* Is z_dbuf already held? */
151 	uint8_t		z_zn_prefetch;	/* Prefetch znodes? */
152 	uint_t		z_blksz;	/* block size in bytes */
153 	uint_t		z_seq;		/* modification sequence number */
154 	uint64_t	z_mapcnt;	/* number of pages mapped to file */
155 	uint64_t	z_last_itx;	/* last ZIL itx on this znode */
156 	uint32_t	z_sync_cnt;	/* synchronous open count */
157 	kmutex_t	z_acl_lock;	/* acl data lock */
158 	list_node_t	z_link_node;	/* all znodes in fs link */
159 	/*
160 	 * These are dmu managed fields.
161 	 */
162 	znode_phys_t	*z_phys;	/* pointer to persistent znode */
163 	dmu_buf_t	*z_dbuf;	/* buffer containing the z_phys */
164 } znode_t;
165 
166 
167 /*
168  * Range locking rules
169  * --------------------
170  * 1. When truncating a file (zfs_create, zfs_setattr, zfs_space) the whole
171  *    file range needs to be locked as RL_WRITER. Only then can the pages be
172  *    freed etc and zp_size reset. zp_size must be set within range lock.
173  * 2. For writes and punching holes (zfs_write & zfs_space) just the range
174  *    being written or freed needs to be locked as RL_WRITER.
175  *    Multiple writes at the end of the file must coordinate zp_size updates
176  *    to ensure data isn't lost. A compare and swap loop is currently used
177  *    to ensure the file size is at least the offset last written.
178  * 3. For reads (zfs_read, zfs_get_data & zfs_putapage) just the range being
179  *    read needs to be locked as RL_READER. A check against zp_size can then
180  *    be made for reading beyond end of file.
181  */
182 
183 /*
184  * Convert between znode pointers and vnode pointers
185  */
186 #define	ZTOV(ZP)	((ZP)->z_vnode)
187 #define	VTOZ(VP)	((znode_t *)(VP)->v_data)
188 
189 /*
190  * ZFS_ENTER() is called on entry to each ZFS vnode and vfs operation.
191  * ZFS_EXIT() must be called before exitting the vop.
192  */
193 #define	ZFS_ENTER(zfsvfs) \
194 	{ \
195 		if (rw_tryenter(&(zfsvfs)->z_unmount_lock, RW_READER) == 0) \
196 			return (EIO); \
197 		if ((zfsvfs)->z_unmounted) { \
198 			ZFS_EXIT(zfsvfs); \
199 			return (EIO); \
200 		} \
201 	}
202 
203 #define	ZFS_EXIT(zfsvfs) rw_exit(&(zfsvfs)->z_unmount_lock)
204 
205 /*
206  * Macros for dealing with dmu_buf_hold
207  */
208 #define	ZFS_OBJ_HASH(obj_num)	(obj_num & (ZFS_OBJ_MTX_SZ - 1))
209 #define	ZFS_OBJ_MUTEX(zp)	\
210 	(&zp->z_zfsvfs->z_hold_mtx[ZFS_OBJ_HASH(zp->z_id)])
211 #define	ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num) \
212 	mutex_enter(&zfsvfs->z_hold_mtx[ZFS_OBJ_HASH(obj_num)]);
213 
214 #define	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num) \
215 	mutex_exit(&zfsvfs->z_hold_mtx[ZFS_OBJ_HASH(obj_num)])
216 
217 /*
218  * Macros to encode/decode ZFS stored time values from/to struct timespec
219  */
220 #define	ZFS_TIME_ENCODE(tp, stmp)		\
221 {						\
222 	stmp[0] = (uint64_t)(tp)->tv_sec; 	\
223 	stmp[1] = (uint64_t)(tp)->tv_nsec;	\
224 }
225 
226 #define	ZFS_TIME_DECODE(tp, stmp)		\
227 {						\
228 	(tp)->tv_sec = (time_t)stmp[0];		\
229 	(tp)->tv_nsec = (long)stmp[1];		\
230 }
231 
232 /*
233  * Timestamp defines
234  */
235 #define	ACCESSED		(AT_ATIME)
236 #define	STATE_CHANGED		(AT_CTIME)
237 #define	CONTENT_MODIFIED	(AT_MTIME | AT_CTIME)
238 
239 #define	ZFS_ACCESSTIME_STAMP(zfsvfs, zp) \
240 	if ((zfsvfs)->z_atime && !((zfsvfs)->z_vfs->vfs_flag & VFS_RDONLY)) \
241 		zfs_time_stamper(zp, ACCESSED, NULL)
242 
243 extern int	zfs_init_fs(zfsvfs_t *, znode_t **, cred_t *);
244 extern void	zfs_set_dataprop(objset_t *);
245 extern void	zfs_create_fs(objset_t *os, cred_t *cr, uint64_t, dmu_tx_t *tx);
246 extern void	zfs_time_stamper(znode_t *, uint_t, dmu_tx_t *);
247 extern void	zfs_time_stamper_locked(znode_t *, uint_t, dmu_tx_t *);
248 extern void	zfs_grow_blocksize(znode_t *, uint64_t, dmu_tx_t *);
249 extern int	zfs_freesp(znode_t *, uint64_t, uint64_t, int, boolean_t);
250 extern void	zfs_znode_init(void);
251 extern void	zfs_znode_fini(void);
252 extern int	zfs_zget(zfsvfs_t *, uint64_t, znode_t **);
253 extern void	zfs_zinactive(znode_t *);
254 extern void	zfs_znode_delete(znode_t *, dmu_tx_t *);
255 extern void	zfs_znode_free(znode_t *);
256 extern void	zfs_remove_op_tables();
257 extern int	zfs_create_op_tables();
258 extern int	zfs_sync(vfs_t *vfsp, short flag, cred_t *cr);
259 extern dev_t	zfs_cmpldev(uint64_t);
260 extern int	zfs_get_version(objset_t *os, uint64_t *version);
261 extern int	zfs_set_version(const char *name, uint64_t newvers);
262 
263 extern void zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, int txtype,
264     znode_t *dzp, znode_t *zp, char *name);
265 extern void zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, int txtype,
266     znode_t *dzp, char *name);
267 extern void zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, int txtype,
268     znode_t *dzp, znode_t *zp, char *name);
269 extern void zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, int txtype,
270     znode_t *dzp, znode_t *zp, char *name, char *link);
271 extern void zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, int txtype,
272     znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp);
273 extern void zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
274     znode_t *zp, offset_t off, ssize_t len, int ioflag);
275 extern void zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
276     znode_t *zp, uint64_t off, uint64_t len);
277 extern void zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
278     znode_t *zp, vattr_t *vap, uint_t mask_applied);
279 extern void zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, int txtype,
280     znode_t *zp, int aclcnt, ace_t *z_ace);
281 
282 extern zil_get_data_t zfs_get_data;
283 extern zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE];
284 extern int zfsfstype;
285 
286 #endif /* _KERNEL */
287 
288 extern int zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len);
289 
290 #ifdef	__cplusplus
291 }
292 #endif
293 
294 #endif	/* _SYS_FS_ZFS_ZNODE_H */
295