vdev_impl.h revision 297078
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 56/* 57 * Virtual device operations 58 */ 59typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size, 60 uint64_t *logical_ashift, uint64_t *physical_ashift); 61typedef void vdev_close_func_t(vdev_t *vd); 62typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize); 63typedef void vdev_io_start_func_t(zio_t *zio); 64typedef void vdev_io_done_func_t(zio_t *zio); 65typedef void vdev_state_change_func_t(vdev_t *vd, int, int); 66typedef void vdev_hold_func_t(vdev_t *vd); 67typedef void vdev_rele_func_t(vdev_t *vd); 68 69typedef struct vdev_ops { 70 vdev_open_func_t *vdev_op_open; 71 vdev_close_func_t *vdev_op_close; 72 vdev_asize_func_t *vdev_op_asize; 73 vdev_io_start_func_t *vdev_op_io_start; 74 vdev_io_done_func_t *vdev_op_io_done; 75 vdev_state_change_func_t *vdev_op_state_change; 76 vdev_hold_func_t *vdev_op_hold; 77 vdev_rele_func_t *vdev_op_rele; 78 char vdev_op_type[16]; 79 boolean_t vdev_op_leaf; 80} vdev_ops_t; 81 82/* 83 * Virtual device properties 84 */ 85struct vdev_cache_entry { 86 char *ve_data; 87 uint64_t ve_offset; 88 uint64_t ve_lastused; 89 avl_node_t ve_offset_node; 90 avl_node_t ve_lastused_node; 91 uint32_t ve_hits; 92 uint16_t ve_missed_update; 93 zio_t *ve_fill_io; 94}; 95 96struct vdev_cache { 97 avl_tree_t vc_offset_tree; 98 avl_tree_t vc_lastused_tree; 99 kmutex_t vc_lock; 100}; 101 102typedef struct vdev_queue_class { 103 uint32_t vqc_active; 104 105 /* 106 * Sorted by offset or timestamp, depending on if the queue is 107 * LBA-ordered vs FIFO. 108 */ 109 avl_tree_t vqc_queued_tree; 110} vdev_queue_class_t; 111 112struct vdev_queue { 113 vdev_t *vq_vdev; 114 vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE]; 115 avl_tree_t vq_active_tree; 116 avl_tree_t vq_read_offset_tree; 117 avl_tree_t vq_write_offset_tree; 118 uint64_t vq_last_offset; 119 hrtime_t vq_io_complete_ts; /* time last i/o completed */ 120 kmutex_t vq_lock; 121 uint64_t vq_lastoffset; 122}; 123 124/* 125 * Virtual device descriptor 126 */ 127struct vdev { 128 /* 129 * Common to all vdev types. 130 */ 131 uint64_t vdev_id; /* child number in vdev parent */ 132 uint64_t vdev_guid; /* unique ID for this vdev */ 133 uint64_t vdev_guid_sum; /* self guid + all child guids */ 134 uint64_t vdev_orig_guid; /* orig. guid prior to remove */ 135 uint64_t vdev_asize; /* allocatable device capacity */ 136 uint64_t vdev_min_asize; /* min acceptable asize */ 137 uint64_t vdev_max_asize; /* max acceptable asize */ 138 uint64_t vdev_ashift; /* block alignment shift */ 139 /* 140 * Logical block alignment shift 141 * 142 * The smallest sized/aligned I/O supported by the device. 143 */ 144 uint64_t vdev_logical_ashift; 145 /* 146 * Physical block alignment shift 147 * 148 * The device supports logical I/Os with vdev_logical_ashift 149 * size/alignment, but optimum performance will be achieved by 150 * aligning/sizing requests to vdev_physical_ashift. Smaller 151 * requests may be inflated or incur device level read-modify-write 152 * operations. 153 * 154 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift). 155 */ 156 uint64_t vdev_physical_ashift; 157 uint64_t vdev_state; /* see VDEV_STATE_* #defines */ 158 uint64_t vdev_prevstate; /* used when reopening a vdev */ 159 vdev_ops_t *vdev_ops; /* vdev operations */ 160 spa_t *vdev_spa; /* spa for this vdev */ 161 void *vdev_tsd; /* type-specific data */ 162 vnode_t *vdev_name_vp; /* vnode for pathname */ 163 vnode_t *vdev_devid_vp; /* vnode for devid */ 164 vdev_t *vdev_top; /* top-level vdev */ 165 vdev_t *vdev_parent; /* parent vdev */ 166 vdev_t **vdev_child; /* array of children */ 167 uint64_t vdev_children; /* number of children */ 168 vdev_stat_t vdev_stat; /* virtual device statistics */ 169 boolean_t vdev_expanding; /* expand the vdev? */ 170 boolean_t vdev_reopening; /* reopen in progress? */ 171 int vdev_open_error; /* error on last open */ 172 kthread_t *vdev_open_thread; /* thread opening children */ 173 uint64_t vdev_crtxg; /* txg when top-level was added */ 174 175 /* 176 * Top-level vdev state. 177 */ 178 uint64_t vdev_ms_array; /* metaslab array object */ 179 uint64_t vdev_ms_shift; /* metaslab size shift */ 180 uint64_t vdev_ms_count; /* number of metaslabs */ 181 metaslab_group_t *vdev_mg; /* metaslab group */ 182 metaslab_t **vdev_ms; /* metaslab array */ 183 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */ 184 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */ 185 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */ 186 boolean_t vdev_remove_wanted; /* async remove wanted? */ 187 boolean_t vdev_probe_wanted; /* async probe wanted? */ 188 list_node_t vdev_config_dirty_node; /* config dirty list */ 189 list_node_t vdev_state_dirty_node; /* state dirty list */ 190 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */ 191 uint64_t vdev_islog; /* is an intent log device */ 192 uint64_t vdev_removing; /* device is being removed? */ 193 boolean_t vdev_ishole; /* is a hole in the namespace */ 194 195 /* 196 * Leaf vdev state. 197 */ 198 range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */ 199 space_map_t *vdev_dtl_sm; /* dirty time log space map */ 200 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */ 201 uint64_t vdev_dtl_object; /* DTL object */ 202 uint64_t vdev_psize; /* physical device capacity */ 203 uint64_t vdev_wholedisk; /* true if this is a whole disk */ 204 uint64_t vdev_offline; /* persistent offline state */ 205 uint64_t vdev_faulted; /* persistent faulted state */ 206 uint64_t vdev_degraded; /* persistent degraded state */ 207 uint64_t vdev_removed; /* persistent removed state */ 208 uint64_t vdev_resilver_txg; /* persistent resilvering state */ 209 uint64_t vdev_nparity; /* number of parity devices for raidz */ 210 char *vdev_path; /* vdev path (if any) */ 211 char *vdev_devid; /* vdev devid (if any) */ 212 char *vdev_physpath; /* vdev device path (if any) */ 213 char *vdev_fru; /* physical FRU location */ 214 uint64_t vdev_not_present; /* not present during import */ 215 uint64_t vdev_unspare; /* unspare when resilvering done */ 216 boolean_t vdev_nowritecache; /* true if flushwritecache failed */ 217 boolean_t vdev_notrim; /* true if trim failed */ 218 boolean_t vdev_checkremove; /* temporary online test */ 219 boolean_t vdev_forcefault; /* force online fault */ 220 boolean_t vdev_splitting; /* split or repair in progress */ 221 boolean_t vdev_delayed_close; /* delayed device close? */ 222 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */ 223 boolean_t vdev_detached; /* device detached? */ 224 boolean_t vdev_cant_read; /* vdev is failing all reads */ 225 boolean_t vdev_cant_write; /* vdev is failing all writes */ 226 boolean_t vdev_isspare; /* was a hot spare */ 227 boolean_t vdev_isl2cache; /* was a l2cache device */ 228 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */ 229 vdev_cache_t vdev_cache; /* physical block cache */ 230 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */ 231 zio_t *vdev_probe_zio; /* root of current probe */ 232 vdev_aux_t vdev_label_aux; /* on-disk aux state */ 233 struct trim_map *vdev_trimmap; /* map on outstanding trims */ 234 uint16_t vdev_rotation_rate; /* rotational rate of the media */ 235#define VDEV_RATE_UNKNOWN 0 236#define VDEV_RATE_NON_ROTATING 1 237 238 /* 239 * For DTrace to work in userland (libzpool) context, these fields must 240 * remain at the end of the structure. DTrace will use the kernel's 241 * CTF definition for 'struct vdev', and since the size of a kmutex_t is 242 * larger in userland, the offsets for the rest of the fields would be 243 * incorrect. 244 */ 245 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */ 246 kmutex_t vdev_stat_lock; /* vdev_stat */ 247 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */ 248}; 249 250#define VDEV_RAIDZ_MAXPARITY 3 251 252#define VDEV_PAD_SIZE (8 << 10) 253/* 2 padding areas (vl_pad1 and vl_pad2) to skip */ 254#define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 255#define VDEV_PHYS_SIZE (112 << 10) 256#define VDEV_UBERBLOCK_RING (128 << 10) 257 258/* The largest uberblock we support is 8k. */ 259#define MAX_UBERBLOCK_SHIFT (13) 260#define VDEV_UBERBLOCK_SHIFT(vd) \ 261 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \ 262 MAX_UBERBLOCK_SHIFT) 263#define VDEV_UBERBLOCK_COUNT(vd) \ 264 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 265#define VDEV_UBERBLOCK_OFFSET(vd, n) \ 266 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 267#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 268 269typedef struct vdev_phys { 270 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 271 zio_eck_t vp_zbt; 272} vdev_phys_t; 273 274typedef struct vdev_label { 275 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 276 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */ 277 vdev_phys_t vl_vdev_phys; /* 112K */ 278 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 279} vdev_label_t; /* 256K total */ 280 281/* 282 * vdev_dirty() flags 283 */ 284#define VDD_METASLAB 0x01 285#define VDD_DTL 0x02 286 287/* Offset of embedded boot loader region on each label */ 288#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 289/* 290 * Size of embedded boot loader region on each label. 291 * The total size of the first two labels plus the boot area is 4MB. 292 */ 293#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 294 295/* 296 * Size of label regions at the start and end of each leaf device. 297 */ 298#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 299#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 300#define VDEV_LABELS 4 301#define VDEV_BEST_LABEL VDEV_LABELS 302 303#define VDEV_ALLOC_LOAD 0 304#define VDEV_ALLOC_ADD 1 305#define VDEV_ALLOC_SPARE 2 306#define VDEV_ALLOC_L2CACHE 3 307#define VDEV_ALLOC_ROOTPOOL 4 308#define VDEV_ALLOC_SPLIT 5 309#define VDEV_ALLOC_ATTACH 6 310 311/* 312 * Allocate or free a vdev 313 */ 314extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, 315 vdev_ops_t *ops); 316extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config, 317 vdev_t *parent, uint_t id, int alloctype); 318extern void vdev_free(vdev_t *vd); 319 320/* 321 * Add or remove children and parents 322 */ 323extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd); 324extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd); 325extern void vdev_compact_children(vdev_t *pvd); 326extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops); 327extern void vdev_remove_parent(vdev_t *cvd); 328 329/* 330 * vdev sync load and sync 331 */ 332extern void vdev_load_log_state(vdev_t *nvd, vdev_t *ovd); 333extern boolean_t vdev_log_state_valid(vdev_t *vd); 334extern void vdev_load(vdev_t *vd); 335extern int vdev_dtl_load(vdev_t *vd); 336extern void vdev_sync(vdev_t *vd, uint64_t txg); 337extern void vdev_sync_done(vdev_t *vd, uint64_t txg); 338extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg); 339extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg); 340 341/* 342 * Available vdev types. 343 */ 344extern vdev_ops_t vdev_root_ops; 345extern vdev_ops_t vdev_mirror_ops; 346extern vdev_ops_t vdev_replacing_ops; 347extern vdev_ops_t vdev_raidz_ops; 348#ifdef _KERNEL 349extern vdev_ops_t vdev_geom_ops; 350#else 351extern vdev_ops_t vdev_disk_ops; 352#endif 353extern vdev_ops_t vdev_file_ops; 354extern vdev_ops_t vdev_missing_ops; 355extern vdev_ops_t vdev_hole_ops; 356extern vdev_ops_t vdev_spare_ops; 357 358/* 359 * Common size functions 360 */ 361extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize); 362extern uint64_t vdev_get_min_asize(vdev_t *vd); 363extern void vdev_set_min_asize(vdev_t *vd); 364 365/* 366 * Global variables 367 */ 368/* zdb uses this tunable, so it must be declared here to make lint happy. */ 369extern int zfs_vdev_cache_size; 370 371#ifdef illumos 372/* 373 * The vdev_buf_t is used to translate between zio_t and buf_t, and back again. 374 */ 375typedef struct vdev_buf { 376 buf_t vb_buf; /* buffer that describes the io */ 377 zio_t *vb_io; /* pointer back to the original zio_t */ 378} vdev_buf_t; 379#endif 380 381#ifdef __cplusplus 382} 383#endif 384 385#endif /* _SYS_VDEV_IMPL_H */ 386