xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/zfs_znode.h (revision bdb9230ac765cb7af3fc1f4119caf2c5720dceb3)
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 2009 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 #ifdef _KERNEL
30 #include <sys/isa_defs.h>
31 #include <sys/types32.h>
32 #include <sys/attr.h>
33 #include <sys/list.h>
34 #include <sys/dmu.h>
35 #include <sys/zfs_vfsops.h>
36 #include <sys/rrwlock.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  * Additional file level attributes, that are stored
47  * in the upper half of zp_flags
48  */
49 #define	ZFS_READONLY		0x0000000100000000
50 #define	ZFS_HIDDEN		0x0000000200000000
51 #define	ZFS_SYSTEM		0x0000000400000000
52 #define	ZFS_ARCHIVE		0x0000000800000000
53 #define	ZFS_IMMUTABLE		0x0000001000000000
54 #define	ZFS_NOUNLINK		0x0000002000000000
55 #define	ZFS_APPENDONLY		0x0000004000000000
56 #define	ZFS_NODUMP		0x0000008000000000
57 #define	ZFS_OPAQUE		0x0000010000000000
58 #define	ZFS_AV_QUARANTINED 	0x0000020000000000
59 #define	ZFS_AV_MODIFIED 	0x0000040000000000
60 
61 #define	ZFS_ATTR_SET(zp, attr, value)	\
62 { \
63 	if (value) \
64 		zp->z_phys->zp_flags |= attr; \
65 	else \
66 		zp->z_phys->zp_flags &= ~attr; \
67 }
68 
69 /*
70  * Define special zfs pflags
71  */
72 #define	ZFS_XATTR		0x1		/* is an extended attribute */
73 #define	ZFS_INHERIT_ACE		0x2		/* ace has inheritable ACEs */
74 #define	ZFS_ACL_TRIVIAL 	0x4		/* files ACL is trivial */
75 #define	ZFS_ACL_OBJ_ACE 	0x8		/* ACL has CMPLX Object ACE */
76 #define	ZFS_ACL_PROTECTED	0x10		/* ACL protected */
77 #define	ZFS_ACL_DEFAULTED	0x20		/* ACL should be defaulted */
78 #define	ZFS_ACL_AUTO_INHERIT	0x40		/* ACL should be inherited */
79 #define	ZFS_BONUS_SCANSTAMP	0x80		/* Scanstamp in bonus area */
80 
81 /*
82  * Is ID ephemeral?
83  */
84 #define	IS_EPHEMERAL(x)		(x > MAXUID)
85 
86 /*
87  * Should we use FUIDs?
88  */
89 #define	USE_FUIDS(version, os)	(version >= ZPL_VERSION_FUID &&\
90     spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
91 
92 #define	MASTER_NODE_OBJ	1
93 
94 /*
95  * Special attributes for master node.
96  * "userquota@" and "groupquota@" are also valid (from
97  * zfs_userquota_prop_prefixes[]).
98  */
99 #define	ZFS_FSID		"FSID"
100 #define	ZFS_UNLINKED_SET	"DELETE_QUEUE"
101 #define	ZFS_ROOT_OBJ		"ROOT"
102 #define	ZPL_VERSION_STR		"VERSION"
103 #define	ZFS_FUID_TABLES		"FUID"
104 #define	ZFS_SHARES_DIR		"SHARES"
105 
106 #define	ZFS_MAX_BLOCKSIZE	(SPA_MAXBLOCKSIZE)
107 
108 /* Path component length */
109 /*
110  * The generic fs code uses MAXNAMELEN to represent
111  * what the largest component length is.  Unfortunately,
112  * this length includes the terminating NULL.  ZFS needs
113  * to tell the users via pathconf() and statvfs() what the
114  * true maximum length of a component is, excluding the NULL.
115  */
116 #define	ZFS_MAXNAMELEN	(MAXNAMELEN - 1)
117 
118 /*
119  * Convert mode bits (zp_mode) to BSD-style DT_* values for storing in
120  * the directory entries.
121  */
122 #define	IFTODT(mode) (((mode) & S_IFMT) >> 12)
123 
124 /*
125  * The directory entry has the type (currently unused on Solaris) in the
126  * top 4 bits, and the object number in the low 48 bits.  The "middle"
127  * 12 bits are unused.
128  */
129 #define	ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
130 #define	ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
131 
132 /*
133  * This is the persistent portion of the znode.  It is stored
134  * in the "bonus buffer" of the file.  Short symbolic links
135  * are also stored in the bonus buffer.
136  */
137 typedef struct znode_phys {
138 	uint64_t zp_atime[2];		/*  0 - last file access time */
139 	uint64_t zp_mtime[2];		/* 16 - last file modification time */
140 	uint64_t zp_ctime[2];		/* 32 - last file change time */
141 	uint64_t zp_crtime[2];		/* 48 - creation time */
142 	uint64_t zp_gen;		/* 64 - generation (txg of creation) */
143 	uint64_t zp_mode;		/* 72 - file mode bits */
144 	uint64_t zp_size;		/* 80 - size of file */
145 	uint64_t zp_parent;		/* 88 - directory parent (`..') */
146 	uint64_t zp_links;		/* 96 - number of links to file */
147 	uint64_t zp_xattr;		/* 104 - DMU object for xattrs */
148 	uint64_t zp_rdev;		/* 112 - dev_t for VBLK & VCHR files */
149 	uint64_t zp_flags;		/* 120 - persistent flags */
150 	uint64_t zp_uid;		/* 128 - file owner */
151 	uint64_t zp_gid;		/* 136 - owning group */
152 	uint64_t zp_zap;		/* 144 - extra attributes */
153 	uint64_t zp_pad[3];		/* 152 - future */
154 	zfs_acl_phys_t zp_acl;		/* 176 - 263 ACL */
155 	/*
156 	 * Data may pad out any remaining bytes in the znode buffer, eg:
157 	 *
158 	 * |<---------------------- dnode_phys (512) ------------------------>|
159 	 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
160 	 *			|<---- znode (264) ---->|<---- data (56) ---->|
161 	 *
162 	 * At present, we use this space for the following:
163 	 *  - symbolic links
164 	 *  - 32-byte anti-virus scanstamp (regular files only)
165 	 */
166 } znode_phys_t;
167 
168 /*
169  * Directory entry locks control access to directory entries.
170  * They are used to protect creates, deletes, and renames.
171  * Each directory znode has a mutex and a list of locked names.
172  */
173 #ifdef _KERNEL
174 typedef struct zfs_dirlock {
175 	char		*dl_name;	/* directory entry being locked */
176 	uint32_t	dl_sharecnt;	/* 0 if exclusive, > 0 if shared */
177 	uint16_t	dl_namesize;	/* set if dl_name was allocated */
178 	kcondvar_t	dl_cv;		/* wait for entry to be unlocked */
179 	struct znode	*dl_dzp;	/* directory znode */
180 	struct zfs_dirlock *dl_next;	/* next in z_dirlocks list */
181 } zfs_dirlock_t;
182 
183 typedef struct znode {
184 	struct zfsvfs	*z_zfsvfs;
185 	vnode_t		*z_vnode;
186 	uint64_t	z_id;		/* object ID for this znode */
187 	kmutex_t	z_lock;		/* znode modification lock */
188 	krwlock_t	z_parent_lock;	/* parent lock for directories */
189 	krwlock_t	z_name_lock;	/* "master" lock for dirent locks */
190 	zfs_dirlock_t	*z_dirlocks;	/* directory entry lock list */
191 	kmutex_t	z_range_lock;	/* protects changes to z_range_avl */
192 	avl_tree_t	z_range_avl;	/* avl tree of file range locks */
193 	uint8_t		z_unlinked;	/* file has been unlinked */
194 	uint8_t		z_atime_dirty;	/* atime needs to be synced */
195 	uint8_t		z_zn_prefetch;	/* Prefetch znodes? */
196 	uint_t		z_blksz;	/* block size in bytes */
197 	uint_t		z_seq;		/* modification sequence number */
198 	uint64_t	z_mapcnt;	/* number of pages mapped to file */
199 	uint64_t	z_last_itx;	/* last ZIL itx on this znode */
200 	uint64_t	z_gen;		/* generation (same as zp_gen) */
201 	uint32_t	z_sync_cnt;	/* synchronous open count */
202 	kmutex_t	z_acl_lock;	/* acl data lock */
203 	list_node_t	z_link_node;	/* all znodes in fs link */
204 	/*
205 	 * These are dmu managed fields.
206 	 */
207 	znode_phys_t	*z_phys;	/* pointer to persistent znode */
208 	dmu_buf_t	*z_dbuf;	/* buffer containing the z_phys */
209 } znode_t;
210 
211 
212 /*
213  * Range locking rules
214  * --------------------
215  * 1. When truncating a file (zfs_create, zfs_setattr, zfs_space) the whole
216  *    file range needs to be locked as RL_WRITER. Only then can the pages be
217  *    freed etc and zp_size reset. zp_size must be set within range lock.
218  * 2. For writes and punching holes (zfs_write & zfs_space) just the range
219  *    being written or freed needs to be locked as RL_WRITER.
220  *    Multiple writes at the end of the file must coordinate zp_size updates
221  *    to ensure data isn't lost. A compare and swap loop is currently used
222  *    to ensure the file size is at least the offset last written.
223  * 3. For reads (zfs_read, zfs_get_data & zfs_putapage) just the range being
224  *    read needs to be locked as RL_READER. A check against zp_size can then
225  *    be made for reading beyond end of file.
226  */
227 
228 /*
229  * Convert between znode pointers and vnode pointers
230  */
231 #define	ZTOV(ZP)	((ZP)->z_vnode)
232 #define	VTOZ(VP)	((znode_t *)(VP)->v_data)
233 
234 /*
235  * ZFS_ENTER() is called on entry to each ZFS vnode and vfs operation.
236  * ZFS_EXIT() must be called before exitting the vop.
237  * ZFS_VERIFY_ZP() verifies the znode is valid.
238  */
239 #define	ZFS_ENTER(zfsvfs) \
240 	{ \
241 		rrw_enter(&(zfsvfs)->z_teardown_lock, RW_READER, FTAG); \
242 		if ((zfsvfs)->z_unmounted) { \
243 			ZFS_EXIT(zfsvfs); \
244 			return (EIO); \
245 		} \
246 	}
247 
248 #define	ZFS_EXIT(zfsvfs) rrw_exit(&(zfsvfs)->z_teardown_lock, FTAG)
249 
250 #define	ZFS_VERIFY_ZP(zp) \
251 	if ((zp)->z_dbuf == NULL) { \
252 		ZFS_EXIT((zp)->z_zfsvfs); \
253 		return (EIO); \
254 	} \
255 
256 /*
257  * Macros for dealing with dmu_buf_hold
258  */
259 #define	ZFS_OBJ_HASH(obj_num)	((obj_num) & (ZFS_OBJ_MTX_SZ - 1))
260 #define	ZFS_OBJ_MUTEX(zfsvfs, obj_num)	\
261 	(&(zfsvfs)->z_hold_mtx[ZFS_OBJ_HASH(obj_num)])
262 #define	ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num) \
263 	mutex_enter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
264 #define	ZFS_OBJ_HOLD_TRYENTER(zfsvfs, obj_num) \
265 	mutex_tryenter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
266 #define	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num) \
267 	mutex_exit(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
268 
269 /*
270  * Macros to encode/decode ZFS stored time values from/to struct timespec
271  */
272 #define	ZFS_TIME_ENCODE(tp, stmp)		\
273 {						\
274 	(stmp)[0] = (uint64_t)(tp)->tv_sec;	\
275 	(stmp)[1] = (uint64_t)(tp)->tv_nsec;	\
276 }
277 
278 #define	ZFS_TIME_DECODE(tp, stmp)		\
279 {						\
280 	(tp)->tv_sec = (time_t)(stmp)[0];		\
281 	(tp)->tv_nsec = (long)(stmp)[1];		\
282 }
283 
284 /*
285  * Timestamp defines
286  */
287 #define	ACCESSED		(AT_ATIME)
288 #define	STATE_CHANGED		(AT_CTIME)
289 #define	CONTENT_MODIFIED	(AT_MTIME | AT_CTIME)
290 
291 #define	ZFS_ACCESSTIME_STAMP(zfsvfs, zp) \
292 	if ((zfsvfs)->z_atime && !((zfsvfs)->z_vfs->vfs_flag & VFS_RDONLY)) \
293 		zfs_time_stamper(zp, ACCESSED, NULL)
294 
295 extern int	zfs_init_fs(zfsvfs_t *, znode_t **);
296 extern void	zfs_set_dataprop(objset_t *);
297 extern void	zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *,
298     dmu_tx_t *tx);
299 extern void	zfs_time_stamper(znode_t *, uint_t, dmu_tx_t *);
300 extern void	zfs_time_stamper_locked(znode_t *, uint_t, dmu_tx_t *);
301 extern void	zfs_grow_blocksize(znode_t *, uint64_t, dmu_tx_t *);
302 extern int	zfs_freesp(znode_t *, uint64_t, uint64_t, int, boolean_t);
303 extern void	zfs_znode_init(void);
304 extern void	zfs_znode_fini(void);
305 extern int	zfs_zget(zfsvfs_t *, uint64_t, znode_t **);
306 extern int	zfs_rezget(znode_t *);
307 extern void	zfs_zinactive(znode_t *);
308 extern void	zfs_znode_delete(znode_t *, dmu_tx_t *);
309 extern void	zfs_znode_free(znode_t *);
310 extern void	zfs_remove_op_tables();
311 extern int	zfs_create_op_tables();
312 extern int	zfs_sync(vfs_t *vfsp, short flag, cred_t *cr);
313 extern dev_t	zfs_cmpldev(uint64_t);
314 extern int	zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value);
315 extern int	zfs_get_stats(objset_t *os, nvlist_t *nv);
316 extern void	zfs_znode_dmu_fini(znode_t *);
317 
318 extern void zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
319     znode_t *dzp, znode_t *zp, char *name, vsecattr_t *, zfs_fuid_info_t *,
320     vattr_t *vap);
321 extern int zfs_log_create_txtype(zil_create_t, vsecattr_t *vsecp,
322     vattr_t *vap);
323 extern void zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
324     znode_t *dzp, char *name);
325 extern void zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
326     znode_t *dzp, znode_t *zp, char *name);
327 extern void zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
328     znode_t *dzp, znode_t *zp, char *name, char *link);
329 extern void zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
330     znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp);
331 extern void zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
332     znode_t *zp, offset_t off, ssize_t len, int ioflag);
333 extern void zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
334     znode_t *zp, uint64_t off, uint64_t len);
335 extern void zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
336     znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp);
337 extern void zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
338     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp);
339 extern void zfs_xvattr_set(znode_t *zp, xvattr_t *xvap);
340 extern void zfs_upgrade(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
341 extern int zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
342 
343 extern caddr_t zfs_map_page(page_t *, enum seg_rw);
344 extern void zfs_unmap_page(page_t *, caddr_t);
345 
346 extern zil_get_data_t zfs_get_data;
347 extern zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE];
348 extern int zfsfstype;
349 
350 #endif /* _KERNEL */
351 
352 extern int zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len);
353 
354 #ifdef	__cplusplus
355 }
356 #endif
357 
358 #endif	/* _SYS_FS_ZFS_ZNODE_H */
359