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 RAID5 = 5, 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 85 /* 86 * raid_bdev is the single entity structure which contains SPDK block device 87 * and the information related to any raid bdev either configured or 88 * in configuring list. io device is created on this. 89 */ 90 struct raid_bdev { 91 /* raid bdev device, this will get registered in bdev layer */ 92 struct spdk_bdev bdev; 93 94 /* link of raid bdev to link it to configured, configuring or offline list */ 95 TAILQ_ENTRY(raid_bdev) state_link; 96 97 /* link of raid bdev to link it to global raid bdev list */ 98 TAILQ_ENTRY(raid_bdev) global_link; 99 100 /* pointer to config file entry */ 101 struct raid_bdev_config *config; 102 103 /* array of base bdev info */ 104 struct raid_base_bdev_info *base_bdev_info; 105 106 /* strip size of raid bdev in blocks */ 107 uint32_t strip_size; 108 109 /* strip size of raid bdev in KB */ 110 uint32_t strip_size_kb; 111 112 /* strip size bit shift for optimized calculation */ 113 uint32_t strip_size_shift; 114 115 /* block length bit shift for optimized calculation */ 116 uint32_t blocklen_shift; 117 118 /* state of raid bdev */ 119 enum raid_bdev_state state; 120 121 /* number of base bdevs comprising raid bdev */ 122 uint8_t num_base_bdevs; 123 124 /* number of base bdevs discovered */ 125 uint8_t num_base_bdevs_discovered; 126 127 /* Raid Level of this raid bdev */ 128 enum raid_level level; 129 130 /* Set to true if destruct is called for this raid bdev */ 131 bool destruct_called; 132 133 /* Set to true if destroy of this raid bdev is started. */ 134 bool destroy_started; 135 136 /* Module for RAID-level specific operations */ 137 struct raid_bdev_module *module; 138 139 /* Private data for the raid module */ 140 void *module_private; 141 }; 142 143 #define RAID_FOR_EACH_BASE_BDEV(r, i) \ 144 for (i = r->base_bdev_info; i < r->base_bdev_info + r->num_base_bdevs; i++) 145 146 /* 147 * raid_base_bdev_config is the per base bdev data structure which contains 148 * information w.r.t to per base bdev during parsing config 149 */ 150 struct raid_base_bdev_config { 151 /* base bdev name from config file */ 152 char *name; 153 }; 154 155 /* 156 * raid_bdev_config contains the raid bdev config related information after 157 * parsing the config file 158 */ 159 struct raid_bdev_config { 160 /* base bdev config per underlying bdev */ 161 struct raid_base_bdev_config *base_bdev; 162 163 /* Points to already created raid bdev */ 164 struct raid_bdev *raid_bdev; 165 166 char *name; 167 168 /* strip size of this raid bdev in KB */ 169 uint32_t strip_size; 170 171 /* number of base bdevs */ 172 uint8_t num_base_bdevs; 173 174 /* raid level */ 175 enum raid_level level; 176 177 TAILQ_ENTRY(raid_bdev_config) link; 178 }; 179 180 /* 181 * raid_config is the top level structure representing the raid bdev config as read 182 * from config file for all raids 183 */ 184 struct raid_config { 185 /* raid bdev context from config file */ 186 TAILQ_HEAD(, raid_bdev_config) raid_bdev_config_head; 187 188 /* total raid bdev from config file */ 189 uint8_t total_raid_bdev; 190 }; 191 192 /* 193 * raid_bdev_io_channel is the context of spdk_io_channel for raid bdev device. It 194 * contains the relationship of raid bdev io channel with base bdev io channels. 195 */ 196 struct raid_bdev_io_channel { 197 /* Array of IO channels of base bdevs */ 198 struct spdk_io_channel **base_channel; 199 200 /* Number of IO channels */ 201 uint8_t num_channels; 202 }; 203 204 /* TAIL heads for various raid bdev lists */ 205 TAILQ_HEAD(raid_configured_tailq, raid_bdev); 206 TAILQ_HEAD(raid_configuring_tailq, raid_bdev); 207 TAILQ_HEAD(raid_all_tailq, raid_bdev); 208 TAILQ_HEAD(raid_offline_tailq, raid_bdev); 209 210 extern struct raid_configured_tailq g_raid_bdev_configured_list; 211 extern struct raid_configuring_tailq g_raid_bdev_configuring_list; 212 extern struct raid_all_tailq g_raid_bdev_list; 213 extern struct raid_offline_tailq g_raid_bdev_offline_list; 214 extern struct raid_config g_raid_config; 215 216 typedef void (*raid_bdev_destruct_cb)(void *cb_ctx, int rc); 217 218 int raid_bdev_create(struct raid_bdev_config *raid_cfg); 219 int raid_bdev_add_base_devices(struct raid_bdev_config *raid_cfg); 220 void raid_bdev_remove_base_devices(struct raid_bdev_config *raid_cfg, 221 raid_bdev_destruct_cb cb_fn, void *cb_ctx); 222 int raid_bdev_config_add(const char *raid_name, uint32_t strip_size, uint8_t num_base_bdevs, 223 enum raid_level level, struct raid_bdev_config **_raid_cfg); 224 int raid_bdev_config_add_base_bdev(struct raid_bdev_config *raid_cfg, 225 const char *base_bdev_name, uint8_t slot); 226 void raid_bdev_config_cleanup(struct raid_bdev_config *raid_cfg); 227 struct raid_bdev_config *raid_bdev_config_find_by_name(const char *raid_name); 228 enum raid_level raid_bdev_parse_raid_level(const char *str); 229 const char *raid_bdev_level_to_str(enum raid_level level); 230 231 /* 232 * RAID module descriptor 233 */ 234 struct raid_bdev_module { 235 /* RAID level implemented by this module */ 236 enum raid_level level; 237 238 /* Minimum required number of base bdevs. Must be > 0. */ 239 uint8_t base_bdevs_min; 240 241 /* 242 * Maximum number of base bdevs that can be removed without failing 243 * the array. 244 */ 245 uint8_t base_bdevs_max_degraded; 246 247 /* 248 * Called when the raid is starting, right before changing the state to 249 * online and registering the bdev. Parameters of the bdev like blockcnt 250 * should be set here. 251 * 252 * Non-zero return value will abort the startup process. 253 */ 254 int (*start)(struct raid_bdev *raid_bdev); 255 256 /* 257 * Called when the raid is stopping, right before changing the state to 258 * offline and unregistering the bdev. Optional. 259 */ 260 void (*stop)(struct raid_bdev *raid_bdev); 261 262 /* Handler for R/W requests */ 263 void (*submit_rw_request)(struct raid_bdev_io *raid_io); 264 265 /* Handler for requests without payload (flush, unmap). Optional. */ 266 void (*submit_null_payload_request)(struct raid_bdev_io *raid_io); 267 268 TAILQ_ENTRY(raid_bdev_module) link; 269 }; 270 271 void raid_bdev_module_list_add(struct raid_bdev_module *raid_module); 272 273 #define __RAID_MODULE_REGISTER(line) __RAID_MODULE_REGISTER_(line) 274 #define __RAID_MODULE_REGISTER_(line) raid_module_register_##line 275 276 #define RAID_MODULE_REGISTER(_module) \ 277 __attribute__((constructor)) static void \ 278 __RAID_MODULE_REGISTER(__LINE__)(void) \ 279 { \ 280 raid_bdev_module_list_add(_module); \ 281 } 282 283 bool raid_bdev_io_complete_part(struct raid_bdev_io *raid_io, uint64_t completed, 284 enum spdk_bdev_io_status status); 285 void raid_bdev_queue_io_wait(struct raid_bdev_io *raid_io, struct spdk_bdev *bdev, 286 struct spdk_io_channel *ch, spdk_bdev_io_wait_cb cb_fn); 287 void raid_bdev_io_complete(struct raid_bdev_io *raid_io, enum spdk_bdev_io_status status); 288 289 #endif /* SPDK_BDEV_RAID_INTERNAL_H */ 290