xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/vdev_impl.h (revision 3d7072f8bd27709dba14f6fe336f149d25d9e207)
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_VDEV_IMPL_H
27 #define	_SYS_VDEV_IMPL_H
28 
29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
30 
31 #include <sys/avl.h>
32 #include <sys/dmu.h>
33 #include <sys/metaslab.h>
34 #include <sys/nvpair.h>
35 #include <sys/space_map.h>
36 #include <sys/vdev.h>
37 #include <sys/dkio.h>
38 #include <sys/uberblock_impl.h>
39 
40 #ifdef	__cplusplus
41 extern "C" {
42 #endif
43 
44 /*
45  * Virtual device descriptors.
46  *
47  * All storage pool operations go through the virtual device framework,
48  * which provides data replication and I/O scheduling.
49  */
50 
51 /*
52  * Forward declarations that lots of things need.
53  */
54 typedef struct vdev_queue vdev_queue_t;
55 typedef struct vdev_cache vdev_cache_t;
56 typedef struct vdev_cache_entry vdev_cache_entry_t;
57 
58 /*
59  * Virtual device operations
60  */
61 typedef int	vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *ashift);
62 typedef void	vdev_close_func_t(vdev_t *vd);
63 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
64 typedef void	vdev_io_start_func_t(zio_t *zio);
65 typedef void	vdev_io_done_func_t(zio_t *zio);
66 typedef void	vdev_state_change_func_t(vdev_t *vd, int, int);
67 
68 typedef struct vdev_ops {
69 	vdev_open_func_t		*vdev_op_open;
70 	vdev_close_func_t		*vdev_op_close;
71 	vdev_asize_func_t		*vdev_op_asize;
72 	vdev_io_start_func_t		*vdev_op_io_start;
73 	vdev_io_done_func_t		*vdev_op_io_done;
74 	vdev_state_change_func_t	*vdev_op_state_change;
75 	char				vdev_op_type[16];
76 	boolean_t			vdev_op_leaf;
77 } vdev_ops_t;
78 
79 /*
80  * Virtual device properties
81  */
82 struct vdev_cache_entry {
83 	char		*ve_data;
84 	uint64_t	ve_offset;
85 	uint64_t	ve_lastused;
86 	avl_node_t	ve_offset_node;
87 	avl_node_t	ve_lastused_node;
88 	uint32_t	ve_hits;
89 	uint16_t	ve_missed_update;
90 	zio_t		*ve_fill_io;
91 };
92 
93 struct vdev_cache {
94 	avl_tree_t	vc_offset_tree;
95 	avl_tree_t	vc_lastused_tree;
96 	kmutex_t	vc_lock;
97 };
98 
99 struct vdev_queue {
100 	avl_tree_t	vq_deadline_tree;
101 	avl_tree_t	vq_read_tree;
102 	avl_tree_t	vq_write_tree;
103 	avl_tree_t	vq_pending_tree;
104 	kmutex_t	vq_lock;
105 };
106 
107 /*
108  * Virtual device descriptor
109  */
110 struct vdev {
111 	/*
112 	 * Common to all vdev types.
113 	 */
114 	uint64_t	vdev_id;	/* child number in vdev parent	*/
115 	uint64_t	vdev_guid;	/* unique ID for this vdev	*/
116 	uint64_t	vdev_guid_sum;	/* self guid + all child guids	*/
117 	uint64_t	vdev_asize;	/* allocatable device capacity	*/
118 	uint64_t	vdev_ashift;	/* block alignment shift	*/
119 	uint64_t	vdev_state;	/* see VDEV_STATE_* #defines	*/
120 	uint64_t	vdev_prevstate;	/* used when reopening a vdev	*/
121 	vdev_ops_t	*vdev_ops;	/* vdev operations		*/
122 	spa_t		*vdev_spa;	/* spa for this vdev		*/
123 	void		*vdev_tsd;	/* type-specific data		*/
124 	vdev_t		*vdev_top;	/* top-level vdev		*/
125 	vdev_t		*vdev_parent;	/* parent vdev			*/
126 	vdev_t		**vdev_child;	/* array of children		*/
127 	uint64_t	vdev_children;	/* number of children		*/
128 	space_map_t	vdev_dtl_map;	/* dirty time log in-core state	*/
129 	space_map_t	vdev_dtl_scrub;	/* DTL for scrub repair writes	*/
130 	vdev_stat_t	vdev_stat;	/* virtual device statistics	*/
131 
132 	/*
133 	 * Top-level vdev state.
134 	 */
135 	uint64_t	vdev_ms_array;	/* metaslab array object	*/
136 	uint64_t	vdev_ms_shift;	/* metaslab size shift		*/
137 	uint64_t	vdev_ms_count;	/* number of metaslabs		*/
138 	metaslab_group_t *vdev_mg;	/* metaslab group		*/
139 	metaslab_t	**vdev_ms;	/* metaslab array		*/
140 	txg_list_t	vdev_ms_list;	/* per-txg dirty metaslab lists	*/
141 	txg_list_t	vdev_dtl_list;	/* per-txg dirty DTL lists	*/
142 	txg_node_t	vdev_txg_node;	/* per-txg dirty vdev linkage	*/
143 	boolean_t	vdev_remove_wanted; /* async remove wanted?	*/
144 	list_node_t	vdev_dirty_node; /* config dirty list		*/
145 	uint64_t	vdev_deflate_ratio; /* deflation ratio (x512)	*/
146 
147 	/*
148 	 * Leaf vdev state.
149 	 */
150 	uint64_t	vdev_psize;	/* physical device capacity	*/
151 	space_map_obj_t	vdev_dtl;	/* dirty time log on-disk state	*/
152 	txg_node_t	vdev_dtl_node;	/* per-txg dirty DTL linkage	*/
153 	uint64_t	vdev_wholedisk;	/* true if this is a whole disk */
154 	uint64_t	vdev_offline;	/* persistent offline state	*/
155 	uint64_t	vdev_faulted;	/* persistent faulted state	*/
156 	uint64_t	vdev_degraded;	/* persistent degraded state	*/
157 	uint64_t	vdev_removed;	/* persistent removed state	*/
158 	uint64_t	vdev_nparity;	/* number of parity devices for raidz */
159 	char		*vdev_path;	/* vdev path (if any)		*/
160 	char		*vdev_devid;	/* vdev devid (if any)		*/
161 	char		*vdev_physpath;	/* vdev device path (if any)	*/
162 	uint64_t	vdev_fault_arg; /* fault injection paramater	*/
163 	int		vdev_fault_mask; /* zio types to fault		*/
164 	uint8_t		vdev_fault_mode; /* fault injection mode	*/
165 	uint8_t		vdev_tmpoffline; /* device taken offline temporarily? */
166 	uint8_t		vdev_detached;	/* device detached?		*/
167 	uint64_t	vdev_isspare;	/* was a hot spare */
168 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
169 	vdev_cache_t	vdev_cache;	/* physical block cache		*/
170 	uint64_t	vdev_not_present; /* not present during import	*/
171 	hrtime_t	vdev_last_try;	/* last reopen time		*/
172 	boolean_t	vdev_nowritecache; /* true if flushwritecache failed */
173 	uint64_t	vdev_unspare;	/* unspare when resilvering done */
174 	boolean_t	vdev_checkremove; /* temporary online test	*/
175 	boolean_t	vdev_forcefault; /* force online fault		*/
176 
177 	/*
178 	 * For DTrace to work in userland (libzpool) context, these fields must
179 	 * remain at the end of the structure.  DTrace will use the kernel's
180 	 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
181 	 * larger in userland, the offsets for the rest fields would be
182 	 * incorrect.
183 	 */
184 	kmutex_t	vdev_dtl_lock;	/* vdev_dtl_{map,resilver}	*/
185 	kmutex_t	vdev_stat_lock;	/* vdev_stat			*/
186 };
187 
188 #define	VDEV_SKIP_SIZE		(8 << 10)
189 #define	VDEV_BOOT_HEADER_SIZE	(8 << 10)
190 #define	VDEV_PHYS_SIZE		(112 << 10)
191 #define	VDEV_UBERBLOCK_RING	(128 << 10)
192 
193 #define	VDEV_UBERBLOCK_SHIFT(vd)	\
194 	MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT)
195 #define	VDEV_UBERBLOCK_COUNT(vd)	\
196 	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
197 #define	VDEV_UBERBLOCK_OFFSET(vd, n)	\
198 	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
199 #define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
200 
201 /* ZFS boot block */
202 #define	VDEV_BOOT_MAGIC		0x2f5b007b10cULL
203 #define	VDEV_BOOT_VERSION	1		/* version number	*/
204 
205 typedef struct vdev_boot_header {
206 	uint64_t	vb_magic;		/* VDEV_BOOT_MAGIC	*/
207 	uint64_t	vb_version;		/* VDEV_BOOT_VERSION	*/
208 	uint64_t	vb_offset;		/* start offset	(bytes) */
209 	uint64_t	vb_size;		/* size (bytes)		*/
210 	char		vb_pad[VDEV_BOOT_HEADER_SIZE - 4 * sizeof (uint64_t)];
211 } vdev_boot_header_t;
212 
213 typedef struct vdev_phys {
214 	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_block_tail_t)];
215 	zio_block_tail_t vp_zbt;
216 } vdev_phys_t;
217 
218 typedef struct vdev_label {
219 	char		vl_pad[VDEV_SKIP_SIZE];			/*   8K	*/
220 	vdev_boot_header_t vl_boot_header;			/*   8K	*/
221 	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
222 	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
223 } vdev_label_t;							/* 256K total */
224 
225 /*
226  * vdev_dirty() flags
227  */
228 #define	VDD_METASLAB	0x01
229 #define	VDD_DTL		0x02
230 
231 /*
232  * Size and offset of embedded boot loader region on each label.
233  * The total size of the first two labels plus the boot area is 4MB.
234  */
235 #define	VDEV_BOOT_OFFSET	(2 * sizeof (vdev_label_t))
236 #define	VDEV_BOOT_SIZE		(7ULL << 19)			/* 3.5M	*/
237 
238 /*
239  * Size of label regions at the start and end of each leaf device.
240  */
241 #define	VDEV_LABEL_START_SIZE	(2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
242 #define	VDEV_LABEL_END_SIZE	(2 * sizeof (vdev_label_t))
243 #define	VDEV_LABELS		4
244 
245 #define	VDEV_ALLOC_LOAD		0
246 #define	VDEV_ALLOC_ADD		1
247 #define	VDEV_ALLOC_SPARE	2
248 
249 /*
250  * Allocate or free a vdev
251  */
252 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
253     vdev_t *parent, uint_t id, int alloctype);
254 extern void vdev_free(vdev_t *vd);
255 
256 /*
257  * Add or remove children and parents
258  */
259 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
260 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
261 extern void vdev_compact_children(vdev_t *pvd);
262 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
263 extern void vdev_remove_parent(vdev_t *cvd);
264 
265 /*
266  * vdev sync load and sync
267  */
268 extern void vdev_load(vdev_t *vd);
269 extern void vdev_sync(vdev_t *vd, uint64_t txg);
270 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
271 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
272 
273 /*
274  * Available vdev types.
275  */
276 extern vdev_ops_t vdev_root_ops;
277 extern vdev_ops_t vdev_mirror_ops;
278 extern vdev_ops_t vdev_replacing_ops;
279 extern vdev_ops_t vdev_raidz_ops;
280 extern vdev_ops_t vdev_disk_ops;
281 extern vdev_ops_t vdev_file_ops;
282 extern vdev_ops_t vdev_missing_ops;
283 extern vdev_ops_t vdev_spare_ops;
284 
285 /*
286  * Common size functions
287  */
288 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
289 extern uint64_t vdev_get_rsize(vdev_t *vd);
290 
291 /*
292  * zdb uses this tunable, so it must be declared here to make lint happy.
293  */
294 extern int zfs_vdev_cache_size;
295 
296 #ifdef	__cplusplus
297 }
298 #endif
299 
300 #endif	/* _SYS_VDEV_IMPL_H */
301