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) 2011, 2015 by Delphix. All rights reserved.
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
39extern "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 */
52typedef struct vdev_queue vdev_queue_t;
53typedef struct vdev_cache vdev_cache_t;
54typedef struct vdev_cache_entry vdev_cache_entry_t;
55
56extern int zfs_vdev_queue_depth_pct;
57extern uint32_t zfs_vdev_async_write_max_active;
58
59/*
60 * Virtual device operations
61 */
62typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
63 uint64_t *logical_ashift, uint64_t *physical_ashift);
64typedef void vdev_close_func_t(vdev_t *vd);
65typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
66typedef void vdev_io_start_func_t(zio_t *zio);
67typedef void vdev_io_done_func_t(zio_t *zio);
68typedef void vdev_state_change_func_t(vdev_t *vd, int, int);
69typedef void vdev_hold_func_t(vdev_t *vd);
70typedef void vdev_rele_func_t(vdev_t *vd);
71
72typedef struct vdev_ops {
73 vdev_open_func_t *vdev_op_open;
74 vdev_close_func_t *vdev_op_close;
75 vdev_asize_func_t *vdev_op_asize;
76 vdev_io_start_func_t *vdev_op_io_start;
77 vdev_io_done_func_t *vdev_op_io_done;
78 vdev_state_change_func_t *vdev_op_state_change;
79 vdev_hold_func_t *vdev_op_hold;
80 vdev_rele_func_t *vdev_op_rele;
81 char vdev_op_type[16];
82 boolean_t vdev_op_leaf;
83} vdev_ops_t;
84
85/*
86 * Virtual device properties
87 */
88struct vdev_cache_entry {
89 char *ve_data;
90 uint64_t ve_offset;
91 uint64_t ve_lastused;
92 avl_node_t ve_offset_node;
93 avl_node_t ve_lastused_node;
94 uint32_t ve_hits;
95 uint16_t ve_missed_update;
96 zio_t *ve_fill_io;
97};
98
99struct vdev_cache {
100 avl_tree_t vc_offset_tree;
101 avl_tree_t vc_lastused_tree;
102 kmutex_t vc_lock;
103};
104
105typedef struct vdev_queue_class {
106 uint32_t vqc_active;
107
108 /*
109 * Sorted by offset or timestamp, depending on if the queue is
110 * LBA-ordered vs FIFO.
111 */
112 avl_tree_t vqc_queued_tree;
113} vdev_queue_class_t;
114
115struct vdev_queue {
116 vdev_t *vq_vdev;
117 vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE];
118 avl_tree_t vq_active_tree;
119 avl_tree_t vq_read_offset_tree;
120 avl_tree_t vq_write_offset_tree;
121 uint64_t vq_last_offset;
122 hrtime_t vq_io_complete_ts; /* time last i/o completed */
123 kmutex_t vq_lock;
124 uint64_t vq_lastoffset;
125};
126
127/*
128 * Virtual device descriptor
129 */
130struct vdev {
131 /*
132 * Common to all vdev types.
133 */
134 uint64_t vdev_id; /* child number in vdev parent */
135 uint64_t vdev_guid; /* unique ID for this vdev */
136 uint64_t vdev_guid_sum; /* self guid + all child guids */
137 uint64_t vdev_orig_guid; /* orig. guid prior to remove */
138 uint64_t vdev_asize; /* allocatable device capacity */
139 uint64_t vdev_min_asize; /* min acceptable asize */
140 uint64_t vdev_max_asize; /* max acceptable asize */
141 uint64_t vdev_ashift; /* block alignment shift */
142 /*
143 * Logical block alignment shift
144 *
145 * The smallest sized/aligned I/O supported by the device.
146 */
147 uint64_t vdev_logical_ashift;
148 /*
149 * Physical block alignment shift
150 *
151 * The device supports logical I/Os with vdev_logical_ashift
152 * size/alignment, but optimum performance will be achieved by
153 * aligning/sizing requests to vdev_physical_ashift. Smaller
154 * requests may be inflated or incur device level read-modify-write
155 * operations.
156 *
157 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift).
158 */
159 uint64_t vdev_physical_ashift;
160 uint64_t vdev_state; /* see VDEV_STATE_* #defines */
161 uint64_t vdev_prevstate; /* used when reopening a vdev */
162 vdev_ops_t *vdev_ops; /* vdev operations */
163 spa_t *vdev_spa; /* spa for this vdev */
164 void *vdev_tsd; /* type-specific data */
165 vnode_t *vdev_name_vp; /* vnode for pathname */
166 vnode_t *vdev_devid_vp; /* vnode for devid */
167 vdev_t *vdev_top; /* top-level vdev */
168 vdev_t *vdev_parent; /* parent vdev */
169 vdev_t **vdev_child; /* array of children */
170 uint64_t vdev_children; /* number of children */
171 vdev_stat_t vdev_stat; /* virtual device statistics */
172 boolean_t vdev_expanding; /* expand the vdev? */
173 boolean_t vdev_reopening; /* reopen in progress? */
174 int vdev_open_error; /* error on last open */
175 kthread_t *vdev_open_thread; /* thread opening children */
176 uint64_t vdev_crtxg; /* txg when top-level was added */
177
178 /*
179 * Top-level vdev state.
180 */
181 uint64_t vdev_ms_array; /* metaslab array object */
182 uint64_t vdev_ms_shift; /* metaslab size shift */
183 uint64_t vdev_ms_count; /* number of metaslabs */
184 metaslab_group_t *vdev_mg; /* metaslab group */
185 metaslab_t **vdev_ms; /* metaslab array */
186 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */
187 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */
188 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */
189 boolean_t vdev_remove_wanted; /* async remove wanted? */
190 boolean_t vdev_probe_wanted; /* async probe wanted? */
191 list_node_t vdev_config_dirty_node; /* config dirty list */
192 list_node_t vdev_state_dirty_node; /* state dirty list */
193 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */
194 uint64_t vdev_islog; /* is an intent log device */
195 uint64_t vdev_removing; /* device is being removed? */
196 boolean_t vdev_ishole; /* is a hole in the namespace */
197 kmutex_t vdev_queue_lock; /* protects vdev_queue_depth */
198 uint64_t vdev_top_zap;
199
200 /*
201 * The queue depth parameters determine how many async writes are
202 * still pending (i.e. allocated by net yet issued to disk) per
203 * top-level (vdev_async_write_queue_depth) and the maximum allowed
204 * (vdev_max_async_write_queue_depth). These values only apply to
205 * top-level vdevs.
206 */
207 uint64_t vdev_async_write_queue_depth;
208 uint64_t vdev_max_async_write_queue_depth;
209
210 /*
211 * Leaf vdev state.
212 */
213 range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */
214 space_map_t *vdev_dtl_sm; /* dirty time log space map */
215 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */
216 uint64_t vdev_dtl_object; /* DTL object */
217 uint64_t vdev_psize; /* physical device capacity */
218 uint64_t vdev_wholedisk; /* true if this is a whole disk */
219 uint64_t vdev_offline; /* persistent offline state */
220 uint64_t vdev_faulted; /* persistent faulted state */
221 uint64_t vdev_degraded; /* persistent degraded state */
222 uint64_t vdev_removed; /* persistent removed state */
223 uint64_t vdev_resilver_txg; /* persistent resilvering state */
224 uint64_t vdev_nparity; /* number of parity devices for raidz */
225 char *vdev_path; /* vdev path (if any) */
226 char *vdev_devid; /* vdev devid (if any) */
227 char *vdev_physpath; /* vdev device path (if any) */
228 char *vdev_fru; /* physical FRU location */
229 uint64_t vdev_not_present; /* not present during import */
230 uint64_t vdev_unspare; /* unspare when resilvering done */
231 boolean_t vdev_nowritecache; /* true if flushwritecache failed */
232 boolean_t vdev_notrim; /* true if trim failed */
233 boolean_t vdev_checkremove; /* temporary online test */
234 boolean_t vdev_forcefault; /* force online fault */
235 boolean_t vdev_splitting; /* split or repair in progress */
236 boolean_t vdev_delayed_close; /* delayed device close? */
237 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */
238 boolean_t vdev_detached; /* device detached? */
239 boolean_t vdev_cant_read; /* vdev is failing all reads */
240 boolean_t vdev_cant_write; /* vdev is failing all writes */
241 boolean_t vdev_isspare; /* was a hot spare */
242 boolean_t vdev_isl2cache; /* was a l2cache device */
243 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */
244 vdev_cache_t vdev_cache; /* physical block cache */
245 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */
246 zio_t *vdev_probe_zio; /* root of current probe */
247 vdev_aux_t vdev_label_aux; /* on-disk aux state */
248 struct trim_map *vdev_trimmap; /* map on outstanding trims */
249 uint16_t vdev_rotation_rate; /* rotational rate of the media */
250#define VDEV_RATE_UNKNOWN 0
251#define VDEV_RATE_NON_ROTATING 1
252 uint64_t vdev_leaf_zap;
253
254 /*
255 * For DTrace to work in userland (libzpool) context, these fields must
256 * remain at the end of the structure. DTrace will use the kernel's
257 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
258 * larger in userland, the offsets for the rest of the fields would be
259 * incorrect.
260 */
261 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */
262 kmutex_t vdev_stat_lock; /* vdev_stat */
263 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */
264};
265
266#define VDEV_RAIDZ_MAXPARITY 3
267
268#define VDEV_PAD_SIZE (8 << 10)
269/* 2 padding areas (vl_pad1 and vl_pad2) to skip */
270#define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
271#define VDEV_PHYS_SIZE (112 << 10)
272#define VDEV_UBERBLOCK_RING (128 << 10)
273
274/* The largest uberblock we support is 8k. */
275#define MAX_UBERBLOCK_SHIFT (13)
276#define VDEV_UBERBLOCK_SHIFT(vd) \
277 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
278 MAX_UBERBLOCK_SHIFT)
279#define VDEV_UBERBLOCK_COUNT(vd) \
280 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
281#define VDEV_UBERBLOCK_OFFSET(vd, n) \
282 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
283#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
284
285typedef struct vdev_phys {
286 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
287 zio_eck_t vp_zbt;
288} vdev_phys_t;
289
290typedef struct vdev_label {
291 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
292 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */
293 vdev_phys_t vl_vdev_phys; /* 112K */
294 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
295} vdev_label_t; /* 256K total */
296
297/*
298 * vdev_dirty() flags
299 */
300#define VDD_METASLAB 0x01
301#define VDD_DTL 0x02
302
303/* Offset of embedded boot loader region on each label */
304#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
305/*
306 * Size of embedded boot loader region on each label.
307 * The total size of the first two labels plus the boot area is 4MB.
308 */
309#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
310
311/*
312 * Size of label regions at the start and end of each leaf device.
313 */
314#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
315#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
316#define VDEV_LABELS 4
317#define VDEV_BEST_LABEL VDEV_LABELS
318
319#define VDEV_ALLOC_LOAD 0
320#define VDEV_ALLOC_ADD 1
321#define VDEV_ALLOC_SPARE 2
322#define VDEV_ALLOC_L2CACHE 3
323#define VDEV_ALLOC_ROOTPOOL 4
324#define VDEV_ALLOC_SPLIT 5
325#define VDEV_ALLOC_ATTACH 6
326
327/*
328 * Allocate or free a vdev
329 */
330extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
331 vdev_ops_t *ops);
332extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
333 vdev_t *parent, uint_t id, int alloctype);
334extern void vdev_free(vdev_t *vd);
335
336/*
337 * Add or remove children and parents
338 */
339extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
340extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
341extern void vdev_compact_children(vdev_t *pvd);
342extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
343extern void vdev_remove_parent(vdev_t *cvd);
344
345/*
346 * vdev sync load and sync
347 */
348extern void vdev_load_log_state(vdev_t *nvd, vdev_t *ovd);
349extern boolean_t vdev_log_state_valid(vdev_t *vd);
350extern void vdev_load(vdev_t *vd);
351extern int vdev_dtl_load(vdev_t *vd);
352extern void vdev_sync(vdev_t *vd, uint64_t txg);
353extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
354extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
355extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
356
357/*
358 * Available vdev types.
359 */
360extern vdev_ops_t vdev_root_ops;
361extern vdev_ops_t vdev_mirror_ops;
362extern vdev_ops_t vdev_replacing_ops;
363extern vdev_ops_t vdev_raidz_ops;
364#if defined(__FreeBSD__) && defined(_KERNEL)
365extern vdev_ops_t vdev_geom_ops;
366#else
367extern vdev_ops_t vdev_disk_ops;
368#endif
369extern vdev_ops_t vdev_file_ops;
370extern vdev_ops_t vdev_missing_ops;
371extern vdev_ops_t vdev_hole_ops;
372extern vdev_ops_t vdev_spare_ops;
373
374/*
375 * Common size functions
376 */
377extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
378extern uint64_t vdev_get_min_asize(vdev_t *vd);
379extern void vdev_set_min_asize(vdev_t *vd);
380
381/*
382 * Global variables
383 */
384/* zdb uses this tunable, so it must be declared here to make lint happy. */
385extern int zfs_vdev_cache_size;
386extern uint_t zfs_geom_probe_vdev_key;
387
388#ifdef illumos
389/*
390 * The vdev_buf_t is used to translate between zio_t and buf_t, and back again.
391 */
392typedef struct vdev_buf {
393 buf_t vb_buf; /* buffer that describes the io */
394 zio_t *vb_io; /* pointer back to the original zio_t */
395} vdev_buf_t;
396#endif
397
398#ifdef __cplusplus
399}
400#endif
401
402#endif /* _SYS_VDEV_IMPL_H */
403