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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23  * Copyright (c) 2013, 2018 by Delphix. All rights reserved.
24  */
25 
26 #ifndef	_SYS_FS_ZFS_VFSOPS_H
27 #define	_SYS_FS_ZFS_VFSOPS_H
28 
29 #include <sys/dataset_kstats.h>
30 #include <sys/isa_defs.h>
31 #include <sys/types32.h>
32 #include <sys/list.h>
33 #include <sys/vfs.h>
34 #include <sys/zil.h>
35 #include <sys/sa.h>
36 #include <sys/rrwlock.h>
37 #include <sys/dsl_dataset.h>
38 #include <sys/zfs_ioctl.h>
39 #include <sys/objlist.h>
40 
41 #ifdef	__cplusplus
42 extern "C" {
43 #endif
44 
45 typedef struct zfsvfs zfsvfs_t;
46 struct znode;
47 
48 /*
49  * This structure emulates the vfs_t from other platforms.  It's purpose
50  * is to facilitate the handling of mount options and minimize structural
51  * differences between the platforms.
52  */
53 typedef struct vfs {
54 	struct zfsvfs	*vfs_data;
55 	char		*vfs_mntpoint;	/* Primary mount point */
56 	uint64_t	vfs_xattr;
57 	boolean_t	vfs_readonly;
58 	boolean_t	vfs_do_readonly;
59 	boolean_t	vfs_setuid;
60 	boolean_t	vfs_do_setuid;
61 	boolean_t	vfs_exec;
62 	boolean_t	vfs_do_exec;
63 	boolean_t	vfs_devices;
64 	boolean_t	vfs_do_devices;
65 	boolean_t	vfs_do_xattr;
66 	boolean_t	vfs_atime;
67 	boolean_t	vfs_do_atime;
68 	boolean_t	vfs_relatime;
69 	boolean_t	vfs_do_relatime;
70 	boolean_t	vfs_nbmand;
71 	boolean_t	vfs_do_nbmand;
72 } vfs_t;
73 
74 typedef struct zfs_mnt {
75 	const char	*mnt_osname;	/* Objset name */
76 	char		*mnt_data;	/* Raw mount options */
77 } zfs_mnt_t;
78 
79 struct zfsvfs {
80 	vfs_t		*z_vfs;		/* generic fs struct */
81 	struct super_block *z_sb;	/* generic super_block */
82 	struct zfsvfs	*z_parent;	/* parent fs */
83 	objset_t	*z_os;		/* objset reference */
84 	uint64_t	z_flags;	/* super_block flags */
85 	uint64_t	z_root;		/* id of root znode */
86 	uint64_t	z_unlinkedobj;	/* id of unlinked zapobj */
87 	uint64_t	z_max_blksz;	/* maximum block size for files */
88 	uint64_t	z_fuid_obj;	/* fuid table object number */
89 	uint64_t	z_fuid_size;	/* fuid table size */
90 	avl_tree_t	z_fuid_idx;	/* fuid tree keyed by index */
91 	avl_tree_t	z_fuid_domain;	/* fuid tree keyed by domain */
92 	krwlock_t	z_fuid_lock;	/* fuid lock */
93 	boolean_t	z_fuid_loaded;	/* fuid tables are loaded */
94 	boolean_t	z_fuid_dirty;   /* need to sync fuid table ? */
95 	struct zfs_fuid_info	*z_fuid_replay; /* fuid info for replay */
96 	zilog_t		*z_log;		/* intent log pointer */
97 	uint_t		z_acl_mode;	/* acl chmod/mode behavior */
98 	uint_t		z_acl_inherit;	/* acl inheritance behavior */
99 	uint_t		z_acl_type;	/* type of ACL usable on this FS */
100 	zfs_case_t	z_case;		/* case-sense */
101 	boolean_t	z_utf8;		/* utf8-only */
102 	int		z_norm;		/* normalization flags */
103 	boolean_t	z_relatime;	/* enable relatime mount option */
104 	boolean_t	z_unmounted;	/* unmounted */
105 	rrmlock_t	z_teardown_lock;
106 	krwlock_t	z_teardown_inactive_lock;
107 	list_t		z_all_znodes;	/* all znodes in the fs */
108 	uint64_t	z_nr_znodes;	/* number of znodes in the fs */
109 	unsigned long	z_rollback_time; /* last online rollback time */
110 	unsigned long	z_snap_defer_time; /* last snapshot unmount deferral */
111 	kmutex_t	z_znodes_lock;	/* lock for z_all_znodes */
112 	arc_prune_t	*z_arc_prune;	/* called by ARC to prune caches */
113 	struct inode	*z_ctldir;	/* .zfs directory inode */
114 	boolean_t	z_show_ctldir;	/* expose .zfs in the root dir */
115 	boolean_t	z_issnap;	/* true if this is a snapshot */
116 	boolean_t	z_use_fuids;	/* version allows fuids */
117 	boolean_t	z_replay;	/* set during ZIL replay */
118 	boolean_t	z_use_sa;	/* version allow system attributes */
119 	boolean_t	z_xattr_sa;	/* allow xattrs to be stores as SA */
120 	boolean_t	z_draining;	/* is true when drain is active */
121 	boolean_t	z_drain_cancel; /* signal the unlinked drain to stop */
122 	uint64_t	z_version;	/* ZPL version */
123 	uint64_t	z_shares_dir;	/* hidden shares dir */
124 	dataset_kstats_t	z_kstat;	/* fs kstats */
125 	kmutex_t	z_lock;
126 	uint64_t	z_userquota_obj;
127 	uint64_t	z_groupquota_obj;
128 	uint64_t	z_userobjquota_obj;
129 	uint64_t	z_groupobjquota_obj;
130 	uint64_t	z_projectquota_obj;
131 	uint64_t	z_projectobjquota_obj;
132 	uint64_t	z_replay_eof;	/* New end of file - replay only */
133 	sa_attr_type_t	*z_attr_table;	/* SA attr mapping->id */
134 	uint64_t	z_hold_size;	/* znode hold array size */
135 	avl_tree_t	*z_hold_trees;	/* znode hold trees */
136 	kmutex_t	*z_hold_locks;	/* znode hold locks */
137 	taskqid_t	z_drain_task;	/* task id for the unlink drain task */
138 };
139 
140 #define	ZFS_TEARDOWN_INIT(zfsvfs)		\
141 	rrm_init(&(zfsvfs)->z_teardown_lock, B_FALSE)
142 
143 #define	ZFS_TEARDOWN_DESTROY(zfsvfs)		\
144 	rrm_destroy(&(zfsvfs)->z_teardown_lock)
145 
146 #define	ZFS_TEARDOWN_TRY_ENTER_READ(zfsvfs)	\
147 	rw_tryenter(&(zfsvfs)->z_teardown_lock, RW_READER)
148 
149 #define	ZFS_TEARDOWN_ENTER_READ(zfsvfs, tag)	\
150 	rrm_enter_read(&(zfsvfs)->z_teardown_lock, tag);
151 
152 #define	ZFS_TEARDOWN_EXIT_READ(zfsvfs, tag)	\
153 	rrm_exit(&(zfsvfs)->z_teardown_lock, tag)
154 
155 #define	ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, tag)	\
156 	rrm_enter(&(zfsvfs)->z_teardown_lock, RW_WRITER, tag)
157 
158 #define	ZFS_TEARDOWN_EXIT_WRITE(zfsvfs)		\
159 	rrm_exit(&(zfsvfs)->z_teardown_lock, tag)
160 
161 #define	ZFS_TEARDOWN_EXIT(zfsvfs, tag)		\
162 	rrm_exit(&(zfsvfs)->z_teardown_lock, tag)
163 
164 #define	ZFS_TEARDOWN_READ_HELD(zfsvfs)		\
165 	RRM_READ_HELD(&(zfsvfs)->z_teardown_lock)
166 
167 #define	ZFS_TEARDOWN_WRITE_HELD(zfsvfs)		\
168 	RRM_WRITE_HELD(&(zfsvfs)->z_teardown_lock)
169 
170 #define	ZFS_TEARDOWN_HELD(zfsvfs)		\
171 	RRM_LOCK_HELD(&(zfsvfs)->z_teardown_lock)
172 
173 #define	ZSB_XATTR	0x0001		/* Enable user xattrs */
174 
175 /*
176  * Allow a maximum number of links.  While ZFS does not internally limit
177  * this the inode->i_nlink member is defined as an unsigned int.  To be
178  * safe we use 2^31-1 as the limit.
179  */
180 #define	ZFS_LINK_MAX		((1U << 31) - 1U)
181 
182 /*
183  * Normal filesystems (those not under .zfs/snapshot) have a total
184  * file ID size limited to 12 bytes (including the length field) due to
185  * NFSv2 protocol's limitation of 32 bytes for a filehandle.  For historical
186  * reasons, this same limit is being imposed by the Solaris NFSv3 implementation
187  * (although the NFSv3 protocol actually permits a maximum of 64 bytes).  It
188  * is not possible to expand beyond 12 bytes without abandoning support
189  * of NFSv2.
190  *
191  * For normal filesystems, we partition up the available space as follows:
192  *	2 bytes		fid length (required)
193  *	6 bytes		object number (48 bits)
194  *	4 bytes		generation number (32 bits)
195  *
196  * We reserve only 48 bits for the object number, as this is the limit
197  * currently defined and imposed by the DMU.
198  */
199 typedef struct zfid_short {
200 	uint16_t	zf_len;
201 	uint8_t		zf_object[6];		/* obj[i] = obj >> (8 * i) */
202 	uint8_t		zf_gen[4];		/* gen[i] = gen >> (8 * i) */
203 } zfid_short_t;
204 
205 /*
206  * Filesystems under .zfs/snapshot have a total file ID size of 22 bytes
207  * (including the length field).  This makes files under .zfs/snapshot
208  * accessible by NFSv3 and NFSv4, but not NFSv2.
209  *
210  * For files under .zfs/snapshot, we partition up the available space
211  * as follows:
212  *	2 bytes		fid length (required)
213  *	6 bytes		object number (48 bits)
214  *	4 bytes		generation number (32 bits)
215  *	6 bytes		objset id (48 bits)
216  *	4 bytes		currently just zero (32 bits)
217  *
218  * We reserve only 48 bits for the object number and objset id, as these are
219  * the limits currently defined and imposed by the DMU.
220  */
221 typedef struct zfid_long {
222 	zfid_short_t	z_fid;
223 	uint8_t		zf_setid[6];		/* obj[i] = obj >> (8 * i) */
224 	uint8_t		zf_setgen[4];		/* gen[i] = gen >> (8 * i) */
225 } zfid_long_t;
226 
227 #define	SHORT_FID_LEN	(sizeof (zfid_short_t) - sizeof (uint16_t))
228 #define	LONG_FID_LEN	(sizeof (zfid_long_t) - sizeof (uint16_t))
229 
230 extern void zfs_init(void);
231 extern void zfs_fini(void);
232 
233 extern int zfs_suspend_fs(zfsvfs_t *zfsvfs);
234 extern int zfs_resume_fs(zfsvfs_t *zfsvfs, struct dsl_dataset *ds);
235 extern int zfs_end_fs(zfsvfs_t *zfsvfs, struct dsl_dataset *ds);
236 extern void zfs_exit_fs(zfsvfs_t *zfsvfs);
237 extern int zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers);
238 extern int zfsvfs_create(const char *name, boolean_t readony, zfsvfs_t **zfvp);
239 extern int zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os);
240 extern void zfsvfs_free(zfsvfs_t *zfsvfs);
241 extern int zfs_check_global_label(const char *dsname, const char *hexsl);
242 
243 extern boolean_t zfs_is_readonly(zfsvfs_t *zfsvfs);
244 extern int zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent);
245 extern void zfs_preumount(struct super_block *sb);
246 extern int zfs_umount(struct super_block *sb);
247 extern int zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm);
248 extern int zfs_statvfs(struct inode *ip, struct kstatfs *statp);
249 extern int zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp);
250 extern int zfs_prune(struct super_block *sb, unsigned long nr_to_scan,
251     int *objects);
252 extern int zfs_get_temporary_prop(dsl_dataset_t *ds, zfs_prop_t zfs_prop,
253     uint64_t *val, char *setpoint);
254 
255 #ifdef	__cplusplus
256 }
257 #endif
258 
259 #endif	/* _SYS_FS_ZFS_VFSOPS_H */
260