initial commit
[freebsd-arm:freebsd-arm.git] / cddl / contrib / opensolaris / uts / common / fs / zfs / sys / vdev_impl.h
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 2008 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 #include <sys/avl.h>
30 #include <sys/dmu.h>
31 #include <sys/metaslab.h>
32 #include <sys/nvpair.h>
33 #include <sys/space_map.h>
34 #include <sys/vdev.h>
35 #include <sys/dkio.h>
36 #include <sys/uberblock_impl.h>
37
38 #ifdef  __cplusplus
39 extern "C" {
40 #endif
41
42 /*
43  * Virtual device descriptors.
44  *
45  * All storage pool operations go through the virtual device framework,
46  * which provides data replication and I/O scheduling.
47  */
48
49 /*
50  * Forward declarations that lots of things need.
51  */
52 typedef struct vdev_queue vdev_queue_t;
53 typedef struct vdev_cache vdev_cache_t;
54 typedef struct vdev_cache_entry vdev_cache_entry_t;
55
56 /*
57  * Virtual device operations
58  */
59 typedef int     vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *ashift);
60 typedef void    vdev_close_func_t(vdev_t *vd);
61 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
62 typedef int     vdev_io_start_func_t(zio_t *zio);
63 typedef void    vdev_io_done_func_t(zio_t *zio);
64 typedef void    vdev_state_change_func_t(vdev_t *vd, int, int);
65
66 typedef struct vdev_ops {
67         vdev_open_func_t                *vdev_op_open;
68         vdev_close_func_t               *vdev_op_close;
69         vdev_asize_func_t               *vdev_op_asize;
70         vdev_io_start_func_t            *vdev_op_io_start;
71         vdev_io_done_func_t             *vdev_op_io_done;
72         vdev_state_change_func_t        *vdev_op_state_change;
73         char                            vdev_op_type[16];
74         boolean_t                       vdev_op_leaf;
75 } vdev_ops_t;
76
77 /*
78  * Virtual device properties
79  */
80 struct vdev_cache_entry {
81         char            *ve_data;
82         uint64_t        ve_offset;
83         uint64_t        ve_lastused;
84         avl_node_t      ve_offset_node;
85         avl_node_t      ve_lastused_node;
86         uint32_t        ve_hits;
87         uint16_t        ve_missed_update;
88         zio_t           *ve_fill_io;
89 };
90
91 struct vdev_cache {
92         avl_tree_t      vc_offset_tree;
93         avl_tree_t      vc_lastused_tree;
94         kmutex_t        vc_lock;
95 };
96
97 struct vdev_queue {
98         avl_tree_t      vq_deadline_tree;
99         avl_tree_t      vq_read_tree;
100         avl_tree_t      vq_write_tree;
101         avl_tree_t      vq_pending_tree;
102         kmutex_t        vq_lock;
103 };
104
105 /*
106  * Virtual device descriptor
107  */
108 struct vdev {
109         /*
110          * Common to all vdev types.
111          */
112         uint64_t        vdev_id;        /* child number in vdev parent  */
113         uint64_t        vdev_guid;      /* unique ID for this vdev      */
114         uint64_t        vdev_guid_sum;  /* self guid + all child guids  */
115         uint64_t        vdev_asize;     /* allocatable device capacity  */
116         uint64_t        vdev_ashift;    /* block alignment shift        */
117         uint64_t        vdev_state;     /* see VDEV_STATE_* #defines    */
118         uint64_t        vdev_prevstate; /* used when reopening a vdev   */
119         vdev_ops_t      *vdev_ops;      /* vdev operations              */
120         spa_t           *vdev_spa;      /* spa for this vdev            */
121         void            *vdev_tsd;      /* type-specific data           */
122         vdev_t          *vdev_top;      /* top-level vdev               */
123         vdev_t          *vdev_parent;   /* parent vdev                  */
124         vdev_t          **vdev_child;   /* array of children            */
125         uint64_t        vdev_children;  /* number of children           */
126         space_map_t     vdev_dtl_map;   /* dirty time log in-core state */
127         space_map_t     vdev_dtl_scrub; /* DTL for scrub repair writes  */
128         vdev_stat_t     vdev_stat;      /* virtual device statistics    */
129
130         /*
131          * Top-level vdev state.
132          */
133         uint64_t        vdev_ms_array;  /* metaslab array object        */
134         uint64_t        vdev_ms_shift;  /* metaslab size shift          */
135         uint64_t        vdev_ms_count;  /* number of metaslabs          */
136         metaslab_group_t *vdev_mg;      /* metaslab group               */
137         metaslab_t      **vdev_ms;      /* metaslab array               */
138         txg_list_t      vdev_ms_list;   /* per-txg dirty metaslab lists */
139         txg_list_t      vdev_dtl_list;  /* per-txg dirty DTL lists      */
140         txg_node_t      vdev_txg_node;  /* per-txg dirty vdev linkage   */
141         boolean_t       vdev_remove_wanted; /* async remove wanted?     */
142         boolean_t       vdev_probe_wanted; /* async probe wanted?       */
143         list_node_t     vdev_config_dirty_node; /* config dirty list    */
144         list_node_t     vdev_state_dirty_node; /* state dirty list      */
145         uint64_t        vdev_deflate_ratio; /* deflation ratio (x512)   */
146         uint64_t        vdev_islog;     /* is an intent log device      */
147
148         /*
149          * Leaf vdev state.
150          */
151         uint64_t        vdev_psize;     /* physical device capacity     */
152         space_map_obj_t vdev_dtl;       /* dirty time log on-disk state */
153         txg_node_t      vdev_dtl_node;  /* per-txg dirty DTL linkage    */
154         uint64_t        vdev_wholedisk; /* true if this is a whole disk */
155         uint64_t        vdev_offline;   /* persistent offline state     */
156         uint64_t        vdev_faulted;   /* persistent faulted state     */
157         uint64_t        vdev_degraded;  /* persistent degraded state    */
158         uint64_t        vdev_removed;   /* persistent removed state     */
159         uint64_t        vdev_nparity;   /* number of parity devices for raidz */
160         char            *vdev_path;     /* vdev path (if any)           */
161         char            *vdev_devid;    /* vdev devid (if any)          */
162         char            *vdev_physpath; /* vdev device path (if any)    */
163         uint64_t        vdev_not_present; /* not present during import  */
164         uint64_t        vdev_unspare;   /* unspare when resilvering done */
165         hrtime_t        vdev_last_try;  /* last reopen time             */
166         boolean_t       vdev_nowritecache; /* true if flushwritecache failed */
167         boolean_t       vdev_checkremove; /* temporary online test      */
168         boolean_t       vdev_forcefault; /* force online fault          */
169         uint8_t         vdev_tmpoffline; /* device taken offline temporarily? */
170         uint8_t         vdev_detached;  /* device detached?             */
171         uint8_t         vdev_cant_read; /* vdev is failing all reads    */
172         uint8_t         vdev_cant_write; /* vdev is failing all writes  */
173         uint64_t        vdev_isspare;   /* was a hot spare              */
174         uint64_t        vdev_isl2cache; /* was a l2cache device         */
175         vdev_queue_t    vdev_queue;     /* I/O deadline schedule queue  */
176         vdev_cache_t    vdev_cache;     /* physical block cache         */
177         spa_aux_vdev_t  *vdev_aux;      /* for l2cache vdevs            */
178         zio_t           *vdev_probe_zio; /* root of current probe       */
179
180         /*
181          * For DTrace to work in userland (libzpool) context, these fields must
182          * remain at the end of the structure.  DTrace will use the kernel's
183          * CTF definition for 'struct vdev', and since the size of a kmutex_t is
184          * larger in userland, the offsets for the rest fields would be
185          * incorrect.
186          */
187         kmutex_t        vdev_dtl_lock;  /* vdev_dtl_{map,resilver}      */
188         kmutex_t        vdev_stat_lock; /* vdev_stat                    */
189         kmutex_t        vdev_probe_lock; /* protects vdev_probe_zio     */
190 };
191
192 #define VDEV_SKIP_SIZE          (8 << 10)
193 #define VDEV_BOOT_HEADER_SIZE   (8 << 10)
194 #define VDEV_PHYS_SIZE          (112 << 10)
195 #define VDEV_UBERBLOCK_RING     (128 << 10)
196
197 #define VDEV_UBERBLOCK_SHIFT(vd)        \
198         MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT)
199 #define VDEV_UBERBLOCK_COUNT(vd)        \
200         (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
201 #define VDEV_UBERBLOCK_OFFSET(vd, n)    \
202         offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
203 #define VDEV_UBERBLOCK_SIZE(vd)         (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
204
205 /* ZFS boot block */
206 #define VDEV_BOOT_MAGIC         0x2f5b007b10cULL
207 #define VDEV_BOOT_VERSION       1               /* version number       */
208
209 typedef struct vdev_boot_header {
210         uint64_t        vb_magic;               /* VDEV_BOOT_MAGIC      */
211         uint64_t        vb_version;             /* VDEV_BOOT_VERSION    */
212         uint64_t        vb_offset;              /* start offset (bytes) */
213         uint64_t        vb_size;                /* size (bytes)         */
214         char            vb_pad[VDEV_BOOT_HEADER_SIZE - 4 * sizeof (uint64_t)];
215 } vdev_boot_header_t;
216
217 typedef struct vdev_phys {
218         char            vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_block_tail_t)];
219         zio_block_tail_t vp_zbt;
220 } vdev_phys_t;
221
222 typedef struct vdev_label {
223         char            vl_pad[VDEV_SKIP_SIZE];                 /*   8K */
224         vdev_boot_header_t vl_boot_header;                      /*   8K */
225         vdev_phys_t     vl_vdev_phys;                           /* 112K */
226         char            vl_uberblock[VDEV_UBERBLOCK_RING];      /* 128K */
227 } vdev_label_t;                                                 /* 256K total */
228
229 /*
230  * vdev_dirty() flags
231  */
232 #define VDD_METASLAB    0x01
233 #define VDD_DTL         0x02
234
235 /*
236  * Size and offset of embedded boot loader region on each label.
237  * The total size of the first two labels plus the boot area is 4MB.
238  */
239 #define VDEV_BOOT_OFFSET        (2 * sizeof (vdev_label_t))
240 #define VDEV_BOOT_SIZE          (7ULL << 19)                    /* 3.5M */
241
242 /*
243  * Size of label regions at the start and end of each leaf device.
244  */
245 #define VDEV_LABEL_START_SIZE   (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
246 #define VDEV_LABEL_END_SIZE     (2 * sizeof (vdev_label_t))
247 #define VDEV_LABELS             4
248
249 #define VDEV_ALLOC_LOAD         0
250 #define VDEV_ALLOC_ADD          1
251 #define VDEV_ALLOC_SPARE        2
252 #define VDEV_ALLOC_L2CACHE      3
253
254 /*
255  * Allocate or free a vdev
256  */
257 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
258     vdev_t *parent, uint_t id, int alloctype);
259 extern void vdev_free(vdev_t *vd);
260
261 /*
262  * Add or remove children and parents
263  */
264 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
265 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
266 extern void vdev_compact_children(vdev_t *pvd);
267 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
268 extern void vdev_remove_parent(vdev_t *cvd);
269
270 /*
271  * vdev sync load and sync
272  */
273 extern void vdev_load(vdev_t *vd);
274 extern void vdev_sync(vdev_t *vd, uint64_t txg);
275 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
276 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
277
278 /*
279  * Available vdev types.
280  */
281 extern vdev_ops_t vdev_root_ops;
282 extern vdev_ops_t vdev_mirror_ops;
283 extern vdev_ops_t vdev_replacing_ops;
284 extern vdev_ops_t vdev_raidz_ops;
285 #ifdef _KERNEL
286 extern vdev_ops_t vdev_geom_ops;
287 #else
288 extern vdev_ops_t vdev_disk_ops;
289 #endif
290 extern vdev_ops_t vdev_file_ops;
291 extern vdev_ops_t vdev_missing_ops;
292 extern vdev_ops_t vdev_spare_ops;
293
294 /*
295  * Common size functions
296  */
297 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
298 extern uint64_t vdev_get_rsize(vdev_t *vd);
299
300 /*
301  * zdb uses this tunable, so it must be declared here to make lint happy.
302  */
303 extern int zfs_vdev_cache_size;
304
305 #ifdef  __cplusplus
306 }
307 #endif
308
309 #endif  /* _SYS_VDEV_IMPL_H */