1 /* SPDX-License-Identifier: BSD-3-Clause 2 * Copyright (c) Intel Corporation. 3 * All rights reserved. 4 */ 5 6 #ifndef SPDK_BDEV_RAID_INTERNAL_H 7 #define SPDK_BDEV_RAID_INTERNAL_H 8 9 #include "spdk/bdev_module.h" 10 11 enum raid_level { 12 INVALID_RAID_LEVEL = -1, 13 RAID0 = 0, 14 RAID5F = 95, /* 0x5f */ 15 CONCAT = 99, 16 }; 17 18 /* 19 * Raid state describes the state of the raid. This raid bdev can be either in 20 * configured list or configuring list 21 */ 22 enum raid_bdev_state { 23 /* raid bdev is ready and is seen by upper layers */ 24 RAID_BDEV_STATE_ONLINE, 25 26 /* 27 * raid bdev is configuring, not all underlying bdevs are present. 28 * And can't be seen by upper layers. 29 */ 30 RAID_BDEV_STATE_CONFIGURING, 31 32 /* 33 * In offline state, raid bdev layer will complete all incoming commands without 34 * submitting to underlying base nvme bdevs 35 */ 36 RAID_BDEV_STATE_OFFLINE, 37 38 /* raid bdev max, new states should be added before this */ 39 RAID_BDEV_MAX 40 }; 41 42 /* 43 * raid_base_bdev_info contains information for the base bdevs which are part of some 44 * raid. This structure contains the per base bdev information. Whatever is 45 * required per base device for raid bdev will be kept here 46 */ 47 struct raid_base_bdev_info { 48 /* pointer to base spdk bdev */ 49 struct spdk_bdev *bdev; 50 51 /* pointer to base bdev descriptor opened by raid bdev */ 52 struct spdk_bdev_desc *desc; 53 54 /* 55 * When underlying base device calls the hot plug function on drive removal, 56 * this flag will be set and later after doing some processing, base device 57 * descriptor will be closed 58 */ 59 bool remove_scheduled; 60 61 /* thread where base device is opened */ 62 struct spdk_thread *thread; 63 }; 64 65 /* 66 * raid_bdev_io is the context part of bdev_io. It contains the information 67 * related to bdev_io for a raid bdev 68 */ 69 struct raid_bdev_io { 70 /* The raid bdev associated with this IO */ 71 struct raid_bdev *raid_bdev; 72 73 /* WaitQ entry, used only in waitq logic */ 74 struct spdk_bdev_io_wait_entry waitq_entry; 75 76 /* Context of the original channel for this IO */ 77 struct raid_bdev_io_channel *raid_ch; 78 79 /* Used for tracking progress on io requests sent to member disks. */ 80 uint64_t base_bdev_io_remaining; 81 uint8_t base_bdev_io_submitted; 82 uint8_t base_bdev_io_status; 83 84 /* Private data for the raid module */ 85 void *module_private; 86 }; 87 88 /* 89 * raid_bdev is the single entity structure which contains SPDK block device 90 * and the information related to any raid bdev either configured or 91 * in configuring list. io device is created on this. 92 */ 93 struct raid_bdev { 94 /* raid bdev device, this will get registered in bdev layer */ 95 struct spdk_bdev bdev; 96 97 /* link of raid bdev to link it to configured, configuring or offline list */ 98 TAILQ_ENTRY(raid_bdev) state_link; 99 100 /* link of raid bdev to link it to global raid bdev list */ 101 TAILQ_ENTRY(raid_bdev) global_link; 102 103 /* pointer to config file entry */ 104 struct raid_bdev_config *config; 105 106 /* array of base bdev info */ 107 struct raid_base_bdev_info *base_bdev_info; 108 109 /* strip size of raid bdev in blocks */ 110 uint32_t strip_size; 111 112 /* strip size of raid bdev in KB */ 113 uint32_t strip_size_kb; 114 115 /* strip size bit shift for optimized calculation */ 116 uint32_t strip_size_shift; 117 118 /* block length bit shift for optimized calculation */ 119 uint32_t blocklen_shift; 120 121 /* state of raid bdev */ 122 enum raid_bdev_state state; 123 124 /* number of base bdevs comprising raid bdev */ 125 uint8_t num_base_bdevs; 126 127 /* number of base bdevs discovered */ 128 uint8_t num_base_bdevs_discovered; 129 130 /* Raid Level of this raid bdev */ 131 enum raid_level level; 132 133 /* Set to true if destruct is called for this raid bdev */ 134 bool destruct_called; 135 136 /* Set to true if destroy of this raid bdev is started. */ 137 bool destroy_started; 138 139 /* Module for RAID-level specific operations */ 140 struct raid_bdev_module *module; 141 142 /* Private data for the raid module */ 143 void *module_private; 144 }; 145 146 #define RAID_FOR_EACH_BASE_BDEV(r, i) \ 147 for (i = r->base_bdev_info; i < r->base_bdev_info + r->num_base_bdevs; i++) 148 149 /* 150 * raid_base_bdev_config is the per base bdev data structure which contains 151 * information w.r.t to per base bdev during parsing config 152 */ 153 struct raid_base_bdev_config { 154 /* base bdev name from config file */ 155 char *name; 156 }; 157 158 /* 159 * raid_bdev_config contains the raid bdev config related information after 160 * parsing the config file 161 */ 162 struct raid_bdev_config { 163 /* base bdev config per underlying bdev */ 164 struct raid_base_bdev_config *base_bdev; 165 166 /* Points to already created raid bdev */ 167 struct raid_bdev *raid_bdev; 168 169 char *name; 170 171 /* strip size of this raid bdev in KB */ 172 uint32_t strip_size; 173 174 /* number of base bdevs */ 175 uint8_t num_base_bdevs; 176 177 /* raid level */ 178 enum raid_level level; 179 180 TAILQ_ENTRY(raid_bdev_config) link; 181 }; 182 183 /* 184 * raid_config is the top level structure representing the raid bdev config as read 185 * from config file for all raids 186 */ 187 struct raid_config { 188 /* raid bdev context from config file */ 189 TAILQ_HEAD(, raid_bdev_config) raid_bdev_config_head; 190 191 /* total raid bdev from config file */ 192 uint8_t total_raid_bdev; 193 }; 194 195 /* 196 * raid_bdev_io_channel is the context of spdk_io_channel for raid bdev device. It 197 * contains the relationship of raid bdev io channel with base bdev io channels. 198 */ 199 struct raid_bdev_io_channel { 200 /* Array of IO channels of base bdevs */ 201 struct spdk_io_channel **base_channel; 202 203 /* Number of IO channels */ 204 uint8_t num_channels; 205 206 /* Private raid module IO channel */ 207 struct spdk_io_channel *module_channel; 208 }; 209 210 /* TAIL heads for various raid bdev lists */ 211 TAILQ_HEAD(raid_configured_tailq, raid_bdev); 212 TAILQ_HEAD(raid_configuring_tailq, raid_bdev); 213 TAILQ_HEAD(raid_all_tailq, raid_bdev); 214 TAILQ_HEAD(raid_offline_tailq, raid_bdev); 215 216 extern struct raid_configured_tailq g_raid_bdev_configured_list; 217 extern struct raid_configuring_tailq g_raid_bdev_configuring_list; 218 extern struct raid_all_tailq g_raid_bdev_list; 219 extern struct raid_offline_tailq g_raid_bdev_offline_list; 220 extern struct raid_config g_raid_config; 221 222 typedef void (*raid_bdev_destruct_cb)(void *cb_ctx, int rc); 223 224 int raid_bdev_create(struct raid_bdev_config *raid_cfg); 225 int raid_bdev_add_base_devices(struct raid_bdev_config *raid_cfg); 226 void raid_bdev_remove_base_devices(struct raid_bdev_config *raid_cfg, 227 raid_bdev_destruct_cb cb_fn, void *cb_ctx); 228 int raid_bdev_config_add(const char *raid_name, uint32_t strip_size, uint8_t num_base_bdevs, 229 enum raid_level level, struct raid_bdev_config **_raid_cfg); 230 int raid_bdev_config_add_base_bdev(struct raid_bdev_config *raid_cfg, 231 const char *base_bdev_name, uint8_t slot); 232 void raid_bdev_config_cleanup(struct raid_bdev_config *raid_cfg); 233 struct raid_bdev_config *raid_bdev_config_find_by_name(const char *raid_name); 234 enum raid_level raid_bdev_parse_raid_level(const char *str); 235 const char *raid_bdev_level_to_str(enum raid_level level); 236 237 /* 238 * RAID module descriptor 239 */ 240 struct raid_bdev_module { 241 /* RAID level implemented by this module */ 242 enum raid_level level; 243 244 /* Minimum required number of base bdevs. Must be > 0. */ 245 uint8_t base_bdevs_min; 246 247 /* 248 * Maximum number of base bdevs that can be removed without failing 249 * the array. 250 */ 251 uint8_t base_bdevs_max_degraded; 252 253 /* 254 * Called when the raid is starting, right before changing the state to 255 * online and registering the bdev. Parameters of the bdev like blockcnt 256 * should be set here. 257 * 258 * Non-zero return value will abort the startup process. 259 */ 260 int (*start)(struct raid_bdev *raid_bdev); 261 262 /* 263 * Called when the raid is stopping, right before changing the state to 264 * offline and unregistering the bdev. Optional. 265 * 266 * The function should return false if it is asynchronous. Then, after 267 * the async operation has completed and the module is fully stopped 268 * raid_bdev_module_stop_done() must be called. 269 */ 270 bool (*stop)(struct raid_bdev *raid_bdev); 271 272 /* Handler for R/W requests */ 273 void (*submit_rw_request)(struct raid_bdev_io *raid_io); 274 275 /* Handler for requests without payload (flush, unmap). Optional. */ 276 void (*submit_null_payload_request)(struct raid_bdev_io *raid_io); 277 278 /* 279 * Called when the bdev's IO channel is created to get the module's private IO channel. 280 * Optional. 281 */ 282 struct spdk_io_channel *(*get_io_channel)(struct raid_bdev *raid_bdev); 283 284 TAILQ_ENTRY(raid_bdev_module) link; 285 }; 286 287 void raid_bdev_module_list_add(struct raid_bdev_module *raid_module); 288 289 #define __RAID_MODULE_REGISTER(line) __RAID_MODULE_REGISTER_(line) 290 #define __RAID_MODULE_REGISTER_(line) raid_module_register_##line 291 292 #define RAID_MODULE_REGISTER(_module) \ 293 __attribute__((constructor)) static void \ 294 __RAID_MODULE_REGISTER(__LINE__)(void) \ 295 { \ 296 raid_bdev_module_list_add(_module); \ 297 } 298 299 bool raid_bdev_io_complete_part(struct raid_bdev_io *raid_io, uint64_t completed, 300 enum spdk_bdev_io_status status); 301 void raid_bdev_queue_io_wait(struct raid_bdev_io *raid_io, struct spdk_bdev *bdev, 302 struct spdk_io_channel *ch, spdk_bdev_io_wait_cb cb_fn); 303 void raid_bdev_io_complete(struct raid_bdev_io *raid_io, enum spdk_bdev_io_status status); 304 void raid_bdev_module_stop_done(struct raid_bdev *raid_bdev); 305 306 #endif /* SPDK_BDEV_RAID_INTERNAL_H */ 307