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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
26 #ifndef _SYS_VDEV_IMPL_H
27 #define _SYS_VDEV_IMPL_H
31 #include <sys/metaslab.h>
32 #include <sys/nvpair.h>
33 #include <sys/space_map.h>
36 #include <sys/uberblock_impl.h>
37 #include <sys/zfs_ratelimit.h>
44 * Virtual device descriptors.
46 * All storage pool operations go through the virtual device framework,
47 * which provides data replication and I/O scheduling.
51 * Forward declarations that lots of things need.
53 typedef struct vdev_queue vdev_queue_t
;
54 typedef struct vdev_cache vdev_cache_t
;
55 typedef struct vdev_cache_entry vdev_cache_entry_t
;
58 extern int zfs_vdev_queue_depth_pct
;
59 extern uint32_t zfs_vdev_async_write_max_active
;
62 * Virtual device operations
64 typedef int vdev_open_func_t(vdev_t
*vd
, uint64_t *size
, uint64_t *max_size
,
66 typedef void vdev_close_func_t(vdev_t
*vd
);
67 typedef uint64_t vdev_asize_func_t(vdev_t
*vd
, uint64_t psize
);
68 typedef void vdev_io_start_func_t(zio_t
*zio
);
69 typedef void vdev_io_done_func_t(zio_t
*zio
);
70 typedef void vdev_state_change_func_t(vdev_t
*vd
, int, int);
71 typedef boolean_t
vdev_need_resilver_func_t(vdev_t
*vd
, uint64_t, size_t);
72 typedef void vdev_hold_func_t(vdev_t
*vd
);
73 typedef void vdev_rele_func_t(vdev_t
*vd
);
75 typedef const struct vdev_ops
{
76 vdev_open_func_t
*vdev_op_open
;
77 vdev_close_func_t
*vdev_op_close
;
78 vdev_asize_func_t
*vdev_op_asize
;
79 vdev_io_start_func_t
*vdev_op_io_start
;
80 vdev_io_done_func_t
*vdev_op_io_done
;
81 vdev_state_change_func_t
*vdev_op_state_change
;
82 vdev_need_resilver_func_t
*vdev_op_need_resilver
;
83 vdev_hold_func_t
*vdev_op_hold
;
84 vdev_rele_func_t
*vdev_op_rele
;
85 char vdev_op_type
[16];
86 boolean_t vdev_op_leaf
;
90 * Virtual device properties
92 struct vdev_cache_entry
{
96 avl_node_t ve_offset_node
;
97 avl_node_t ve_lastused_node
;
99 uint16_t ve_missed_update
;
104 avl_tree_t vc_offset_tree
;
105 avl_tree_t vc_lastused_tree
;
109 typedef struct vdev_queue_class
{
113 * Sorted by offset or timestamp, depending on if the queue is
114 * LBA-ordered vs FIFO.
116 avl_tree_t vqc_queued_tree
;
117 } vdev_queue_class_t
;
121 vdev_queue_class_t vq_class
[ZIO_PRIORITY_NUM_QUEUEABLE
];
122 avl_tree_t vq_active_tree
;
123 avl_tree_t vq_read_offset_tree
;
124 avl_tree_t vq_write_offset_tree
;
125 uint64_t vq_last_offset
;
126 hrtime_t vq_io_complete_ts
; /* time last i/o completed */
127 hrtime_t vq_io_delta_ts
;
128 zio_t vq_io_search
; /* used as local for stack reduction */
133 * Virtual device descriptor
137 * Common to all vdev types.
139 uint64_t vdev_id
; /* child number in vdev parent */
140 uint64_t vdev_guid
; /* unique ID for this vdev */
141 uint64_t vdev_guid_sum
; /* self guid + all child guids */
142 uint64_t vdev_orig_guid
; /* orig. guid prior to remove */
143 uint64_t vdev_asize
; /* allocatable device capacity */
144 uint64_t vdev_min_asize
; /* min acceptable asize */
145 uint64_t vdev_max_asize
; /* max acceptable asize */
146 uint64_t vdev_ashift
; /* block alignment shift */
147 uint64_t vdev_state
; /* see VDEV_STATE_* #defines */
148 uint64_t vdev_prevstate
; /* used when reopening a vdev */
149 vdev_ops_t
*vdev_ops
; /* vdev operations */
150 spa_t
*vdev_spa
; /* spa for this vdev */
151 void *vdev_tsd
; /* type-specific data */
152 vnode_t
*vdev_name_vp
; /* vnode for pathname */
153 vnode_t
*vdev_devid_vp
; /* vnode for devid */
154 vdev_t
*vdev_top
; /* top-level vdev */
155 vdev_t
*vdev_parent
; /* parent vdev */
156 vdev_t
**vdev_child
; /* array of children */
157 uint64_t vdev_children
; /* number of children */
158 vdev_stat_t vdev_stat
; /* virtual device statistics */
159 vdev_stat_ex_t vdev_stat_ex
; /* extended statistics */
160 boolean_t vdev_expanding
; /* expand the vdev? */
161 boolean_t vdev_reopening
; /* reopen in progress? */
162 boolean_t vdev_nonrot
; /* true if solid state */
163 int vdev_open_error
; /* error on last open */
164 kthread_t
*vdev_open_thread
; /* thread opening children */
165 uint64_t vdev_crtxg
; /* txg when top-level was added */
168 * Top-level vdev state.
170 uint64_t vdev_ms_array
; /* metaslab array object */
171 uint64_t vdev_ms_shift
; /* metaslab size shift */
172 uint64_t vdev_ms_count
; /* number of metaslabs */
173 metaslab_group_t
*vdev_mg
; /* metaslab group */
174 metaslab_t
**vdev_ms
; /* metaslab array */
175 uint64_t vdev_pending_fastwrite
; /* allocated fastwrites */
176 txg_list_t vdev_ms_list
; /* per-txg dirty metaslab lists */
177 txg_list_t vdev_dtl_list
; /* per-txg dirty DTL lists */
178 txg_node_t vdev_txg_node
; /* per-txg dirty vdev linkage */
179 boolean_t vdev_remove_wanted
; /* async remove wanted? */
180 boolean_t vdev_probe_wanted
; /* async probe wanted? */
181 list_node_t vdev_config_dirty_node
; /* config dirty list */
182 list_node_t vdev_state_dirty_node
; /* state dirty list */
183 uint64_t vdev_deflate_ratio
; /* deflation ratio (x512) */
184 uint64_t vdev_islog
; /* is an intent log device */
185 uint64_t vdev_removing
; /* device is being removed? */
186 boolean_t vdev_ishole
; /* is a hole in the namespace */
187 kmutex_t vdev_queue_lock
; /* protects vdev_queue_depth */
188 uint64_t vdev_top_zap
;
191 * The queue depth parameters determine how many async writes are
192 * still pending (i.e. allocated by net yet issued to disk) per
193 * top-level (vdev_async_write_queue_depth) and the maximum allowed
194 * (vdev_max_async_write_queue_depth). These values only apply to
197 uint64_t vdev_async_write_queue_depth
;
198 uint64_t vdev_max_async_write_queue_depth
;
203 range_tree_t
*vdev_dtl
[DTL_TYPES
]; /* dirty time logs */
204 space_map_t
*vdev_dtl_sm
; /* dirty time log space map */
205 txg_node_t vdev_dtl_node
; /* per-txg dirty DTL linkage */
206 uint64_t vdev_dtl_object
; /* DTL object */
207 uint64_t vdev_psize
; /* physical device capacity */
208 uint64_t vdev_wholedisk
; /* true if this is a whole disk */
209 uint64_t vdev_offline
; /* persistent offline state */
210 uint64_t vdev_faulted
; /* persistent faulted state */
211 uint64_t vdev_degraded
; /* persistent degraded state */
212 uint64_t vdev_removed
; /* persistent removed state */
213 uint64_t vdev_resilver_txg
; /* persistent resilvering state */
214 uint64_t vdev_nparity
; /* number of parity devices for raidz */
215 char *vdev_path
; /* vdev path (if any) */
216 char *vdev_devid
; /* vdev devid (if any) */
217 char *vdev_physpath
; /* vdev device path (if any) */
218 char *vdev_enc_sysfs_path
; /* enclosure sysfs path */
219 char *vdev_fru
; /* physical FRU location */
220 uint64_t vdev_not_present
; /* not present during import */
221 uint64_t vdev_unspare
; /* unspare when resilvering done */
222 boolean_t vdev_nowritecache
; /* true if flushwritecache failed */
223 boolean_t vdev_checkremove
; /* temporary online test */
224 boolean_t vdev_forcefault
; /* force online fault */
225 boolean_t vdev_splitting
; /* split or repair in progress */
226 boolean_t vdev_delayed_close
; /* delayed device close? */
227 boolean_t vdev_tmpoffline
; /* device taken offline temporarily? */
228 boolean_t vdev_detached
; /* device detached? */
229 boolean_t vdev_cant_read
; /* vdev is failing all reads */
230 boolean_t vdev_cant_write
; /* vdev is failing all writes */
231 boolean_t vdev_isspare
; /* was a hot spare */
232 boolean_t vdev_isl2cache
; /* was a l2cache device */
233 boolean_t vdev_copy_uberblocks
; /* post expand copy uberblocks */
234 vdev_queue_t vdev_queue
; /* I/O deadline schedule queue */
235 vdev_cache_t vdev_cache
; /* physical block cache */
236 spa_aux_vdev_t
*vdev_aux
; /* for l2cache and spares vdevs */
237 zio_t
*vdev_probe_zio
; /* root of current probe */
238 vdev_aux_t vdev_label_aux
; /* on-disk aux state */
239 uint64_t vdev_leaf_zap
;
240 hrtime_t vdev_mmp_pending
; /* 0 if write finished */
243 * For DTrace to work in userland (libzpool) context, these fields must
244 * remain at the end of the structure. DTrace will use the kernel's
245 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
246 * larger in userland, the offsets for the rest of the fields would be
249 kmutex_t vdev_dtl_lock
; /* vdev_dtl_{map,resilver} */
250 kmutex_t vdev_stat_lock
; /* vdev_stat */
251 kmutex_t vdev_probe_lock
; /* protects vdev_probe_zio */
254 * We rate limit ZIO delay and ZIO checksum events, since they
255 * can flood ZED with tons of events when a drive is acting up.
257 #define DELAYS_PER_SECOND 5
258 #define CHECKSUMS_PER_SECOND 5
259 zfs_ratelimit_t vdev_delay_rl
;
260 zfs_ratelimit_t vdev_checksum_rl
;
263 #define VDEV_RAIDZ_MAXPARITY 3
265 #define VDEV_PAD_SIZE (8 << 10)
266 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
267 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
268 #define VDEV_PHYS_SIZE (112 << 10)
269 #define VDEV_UBERBLOCK_RING (128 << 10)
272 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
273 * ring when MMP is enabled.
275 #define MMP_BLOCKS_PER_LABEL 1
277 /* The largest uberblock we support is 8k. */
278 #define MAX_UBERBLOCK_SHIFT (13)
279 #define VDEV_UBERBLOCK_SHIFT(vd) \
280 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
282 #define VDEV_UBERBLOCK_COUNT(vd) \
283 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
284 #define VDEV_UBERBLOCK_OFFSET(vd, n) \
285 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
286 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
288 typedef struct vdev_phys
{
289 char vp_nvlist
[VDEV_PHYS_SIZE
- sizeof (zio_eck_t
)];
293 typedef struct vdev_label
{
294 char vl_pad1
[VDEV_PAD_SIZE
]; /* 8K */
295 char vl_pad2
[VDEV_PAD_SIZE
]; /* 8K */
296 vdev_phys_t vl_vdev_phys
; /* 112K */
297 char vl_uberblock
[VDEV_UBERBLOCK_RING
]; /* 128K */
298 } vdev_label_t
; /* 256K total */
303 #define VDD_METASLAB 0x01
306 /* Offset of embedded boot loader region on each label */
307 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
309 * Size of embedded boot loader region on each label.
310 * The total size of the first two labels plus the boot area is 4MB.
312 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
315 * Size of label regions at the start and end of each leaf device.
317 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
318 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
319 #define VDEV_LABELS 4
320 #define VDEV_BEST_LABEL VDEV_LABELS
322 #define VDEV_ALLOC_LOAD 0
323 #define VDEV_ALLOC_ADD 1
324 #define VDEV_ALLOC_SPARE 2
325 #define VDEV_ALLOC_L2CACHE 3
326 #define VDEV_ALLOC_ROOTPOOL 4
327 #define VDEV_ALLOC_SPLIT 5
328 #define VDEV_ALLOC_ATTACH 6
331 * Allocate or free a vdev
333 extern vdev_t
*vdev_alloc_common(spa_t
*spa
, uint_t id
, uint64_t guid
,
335 extern int vdev_alloc(spa_t
*spa
, vdev_t
**vdp
, nvlist_t
*config
,
336 vdev_t
*parent
, uint_t id
, int alloctype
);
337 extern void vdev_free(vdev_t
*vd
);
340 * Add or remove children and parents
342 extern void vdev_add_child(vdev_t
*pvd
, vdev_t
*cvd
);
343 extern void vdev_remove_child(vdev_t
*pvd
, vdev_t
*cvd
);
344 extern void vdev_compact_children(vdev_t
*pvd
);
345 extern vdev_t
*vdev_add_parent(vdev_t
*cvd
, vdev_ops_t
*ops
);
346 extern void vdev_remove_parent(vdev_t
*cvd
);
349 * vdev sync load and sync
351 extern void vdev_load_log_state(vdev_t
*nvd
, vdev_t
*ovd
);
352 extern boolean_t
vdev_log_state_valid(vdev_t
*vd
);
353 extern void vdev_load(vdev_t
*vd
);
354 extern int vdev_dtl_load(vdev_t
*vd
);
355 extern void vdev_sync(vdev_t
*vd
, uint64_t txg
);
356 extern void vdev_sync_done(vdev_t
*vd
, uint64_t txg
);
357 extern void vdev_dirty(vdev_t
*vd
, int flags
, void *arg
, uint64_t txg
);
358 extern void vdev_dirty_leaves(vdev_t
*vd
, int flags
, uint64_t txg
);
361 * Available vdev types.
363 extern vdev_ops_t vdev_root_ops
;
364 extern vdev_ops_t vdev_mirror_ops
;
365 extern vdev_ops_t vdev_replacing_ops
;
366 extern vdev_ops_t vdev_raidz_ops
;
367 extern vdev_ops_t vdev_disk_ops
;
368 extern vdev_ops_t vdev_file_ops
;
369 extern vdev_ops_t vdev_missing_ops
;
370 extern vdev_ops_t vdev_hole_ops
;
371 extern vdev_ops_t vdev_spare_ops
;
374 * Common size functions
376 extern uint64_t vdev_default_asize(vdev_t
*vd
, uint64_t psize
);
377 extern uint64_t vdev_get_min_asize(vdev_t
*vd
);
378 extern void vdev_set_min_asize(vdev_t
*vd
);
383 /* zdb uses this tunable, so it must be declared here to make lint happy. */
384 extern int zfs_vdev_cache_size
;
390 #endif /* _SYS_VDEV_IMPL_H */