| 1 | /* |
| 2 | * Internal header file _only_ for device mapper core |
| 3 | * |
| 4 | * Copyright (C) 2016 Red Hat, Inc. All rights reserved. |
| 5 | * |
| 6 | * This file is released under the LGPL. |
| 7 | */ |
| 8 | |
| 9 | #ifndef DM_CORE_INTERNAL_H |
| 10 | #define DM_CORE_INTERNAL_H |
| 11 | |
| 12 | #include <linux/kthread.h> |
| 13 | #include <linux/ktime.h> |
| 14 | #include <linux/blk-mq.h> |
| 15 | #include <linux/blk-crypto-profile.h> |
| 16 | #include <linux/jump_label.h> |
| 17 | |
| 18 | #include <trace/events/block.h> |
| 19 | |
| 20 | #include "dm.h" |
| 21 | #include "dm-ima.h" |
| 22 | |
| 23 | #define DM_RESERVED_MAX_IOS 1024 |
| 24 | |
| 25 | struct dm_io; |
| 26 | |
| 27 | struct dm_kobject_holder { |
| 28 | struct kobject kobj; |
| 29 | struct completion completion; |
| 30 | }; |
| 31 | |
| 32 | /* |
| 33 | * DM core internal structures used directly by dm.c, dm-rq.c and dm-table.c. |
| 34 | * DM targets must _not_ deference a mapped_device or dm_table to directly |
| 35 | * access their members! |
| 36 | */ |
| 37 | |
| 38 | /* |
| 39 | * For mempools pre-allocation at the table loading time. |
| 40 | */ |
| 41 | struct dm_md_mempools { |
| 42 | struct bio_set bs; |
| 43 | struct bio_set io_bs; |
| 44 | }; |
| 45 | |
| 46 | struct mapped_device { |
| 47 | struct mutex suspend_lock; |
| 48 | |
| 49 | struct mutex table_devices_lock; |
| 50 | struct list_head table_devices; |
| 51 | |
| 52 | /* |
| 53 | * The current mapping (struct dm_table *). |
| 54 | * Use dm_get_live_table{_fast} or take suspend_lock for |
| 55 | * dereference. |
| 56 | */ |
| 57 | void __rcu *map; |
| 58 | |
| 59 | unsigned long flags; |
| 60 | |
| 61 | /* Protect queue and type against concurrent access. */ |
| 62 | struct mutex type_lock; |
| 63 | enum dm_queue_mode type; |
| 64 | |
| 65 | int numa_node_id; |
| 66 | struct request_queue *queue; |
| 67 | |
| 68 | atomic_t holders; |
| 69 | atomic_t open_count; |
| 70 | |
| 71 | struct dm_target *immutable_target; |
| 72 | struct target_type *immutable_target_type; |
| 73 | |
| 74 | char name[16]; |
| 75 | struct gendisk *disk; |
| 76 | struct dax_device *dax_dev; |
| 77 | |
| 78 | wait_queue_head_t wait; |
| 79 | unsigned long __percpu *pending_io; |
| 80 | |
| 81 | /* forced geometry settings */ |
| 82 | struct hd_geometry geometry; |
| 83 | |
| 84 | /* |
| 85 | * Processing queue (flush) |
| 86 | */ |
| 87 | struct workqueue_struct *wq; |
| 88 | |
| 89 | /* |
| 90 | * A list of ios that arrived while we were suspended. |
| 91 | */ |
| 92 | struct work_struct work; |
| 93 | spinlock_t deferred_lock; |
| 94 | struct bio_list deferred; |
| 95 | |
| 96 | /* |
| 97 | * requeue work context is needed for cloning one new bio |
| 98 | * to represent the dm_io to be requeued, since each |
| 99 | * dm_io may point to the original bio from FS. |
| 100 | */ |
| 101 | struct work_struct requeue_work; |
| 102 | struct dm_io *requeue_list; |
| 103 | |
| 104 | void *interface_ptr; |
| 105 | |
| 106 | /* |
| 107 | * Event handling. |
| 108 | */ |
| 109 | wait_queue_head_t eventq; |
| 110 | atomic_t event_nr; |
| 111 | atomic_t uevent_seq; |
| 112 | struct list_head uevent_list; |
| 113 | spinlock_t uevent_lock; /* Protect access to uevent_list */ |
| 114 | |
| 115 | /* for blk-mq request-based DM support */ |
| 116 | bool init_tio_pdu:1; |
| 117 | struct blk_mq_tag_set *tag_set; |
| 118 | |
| 119 | struct dm_stats stats; |
| 120 | |
| 121 | /* the number of internal suspends */ |
| 122 | unsigned internal_suspend_count; |
| 123 | |
| 124 | int swap_bios; |
| 125 | struct semaphore swap_bios_semaphore; |
| 126 | struct mutex swap_bios_lock; |
| 127 | |
| 128 | /* |
| 129 | * io objects are allocated from here. |
| 130 | */ |
| 131 | struct dm_md_mempools *mempools; |
| 132 | |
| 133 | /* kobject and completion */ |
| 134 | struct dm_kobject_holder kobj_holder; |
| 135 | |
| 136 | struct srcu_struct io_barrier; |
| 137 | |
| 138 | #ifdef CONFIG_BLK_DEV_ZONED |
| 139 | unsigned int nr_zones; |
| 140 | unsigned int *zwp_offset; |
| 141 | #endif |
| 142 | |
| 143 | #ifdef CONFIG_IMA |
| 144 | struct dm_ima_measurements ima; |
| 145 | #endif |
| 146 | }; |
| 147 | |
| 148 | /* |
| 149 | * Bits for the flags field of struct mapped_device. |
| 150 | */ |
| 151 | #define DMF_BLOCK_IO_FOR_SUSPEND 0 |
| 152 | #define DMF_SUSPENDED 1 |
| 153 | #define DMF_FROZEN 2 |
| 154 | #define DMF_FREEING 3 |
| 155 | #define DMF_DELETING 4 |
| 156 | #define DMF_NOFLUSH_SUSPENDING 5 |
| 157 | #define DMF_DEFERRED_REMOVE 6 |
| 158 | #define DMF_SUSPENDED_INTERNALLY 7 |
| 159 | #define DMF_POST_SUSPENDING 8 |
| 160 | #define DMF_EMULATE_ZONE_APPEND 9 |
| 161 | |
| 162 | void disable_discard(struct mapped_device *md); |
| 163 | void disable_write_zeroes(struct mapped_device *md); |
| 164 | |
| 165 | static inline sector_t dm_get_size(struct mapped_device *md) |
| 166 | { |
| 167 | return get_capacity(md->disk); |
| 168 | } |
| 169 | |
| 170 | static inline struct dm_stats *dm_get_stats(struct mapped_device *md) |
| 171 | { |
| 172 | return &md->stats; |
| 173 | } |
| 174 | |
| 175 | DECLARE_STATIC_KEY_FALSE(stats_enabled); |
| 176 | DECLARE_STATIC_KEY_FALSE(swap_bios_enabled); |
| 177 | DECLARE_STATIC_KEY_FALSE(zoned_enabled); |
| 178 | |
| 179 | static inline bool dm_emulate_zone_append(struct mapped_device *md) |
| 180 | { |
| 181 | if (blk_queue_is_zoned(md->queue)) |
| 182 | return test_bit(DMF_EMULATE_ZONE_APPEND, &md->flags); |
| 183 | return false; |
| 184 | } |
| 185 | |
| 186 | #define DM_TABLE_MAX_DEPTH 16 |
| 187 | |
| 188 | struct dm_table { |
| 189 | struct mapped_device *md; |
| 190 | enum dm_queue_mode type; |
| 191 | |
| 192 | /* btree table */ |
| 193 | unsigned int depth; |
| 194 | unsigned int counts[DM_TABLE_MAX_DEPTH]; /* in nodes */ |
| 195 | sector_t *index[DM_TABLE_MAX_DEPTH]; |
| 196 | |
| 197 | unsigned int num_targets; |
| 198 | unsigned int num_allocated; |
| 199 | sector_t *highs; |
| 200 | struct dm_target *targets; |
| 201 | |
| 202 | struct target_type *immutable_target_type; |
| 203 | |
| 204 | bool integrity_supported:1; |
| 205 | bool singleton:1; |
| 206 | unsigned integrity_added:1; |
| 207 | |
| 208 | /* |
| 209 | * Indicates the rw permissions for the new logical |
| 210 | * device. This should be a combination of FMODE_READ |
| 211 | * and FMODE_WRITE. |
| 212 | */ |
| 213 | fmode_t mode; |
| 214 | |
| 215 | /* a list of devices used by this table */ |
| 216 | struct list_head devices; |
| 217 | |
| 218 | /* events get handed up using this callback */ |
| 219 | void (*event_fn)(void *); |
| 220 | void *event_context; |
| 221 | |
| 222 | struct dm_md_mempools *mempools; |
| 223 | |
| 224 | #ifdef CONFIG_BLK_INLINE_ENCRYPTION |
| 225 | struct blk_crypto_profile *crypto_profile; |
| 226 | #endif |
| 227 | }; |
| 228 | |
| 229 | /* |
| 230 | * One of these is allocated per clone bio. |
| 231 | */ |
| 232 | #define DM_TIO_MAGIC 28714 |
| 233 | struct dm_target_io { |
| 234 | unsigned short magic; |
| 235 | blk_short_t flags; |
| 236 | unsigned int target_bio_nr; |
| 237 | struct dm_io *io; |
| 238 | struct dm_target *ti; |
| 239 | unsigned int *len_ptr; |
| 240 | sector_t old_sector; |
| 241 | struct bio clone; |
| 242 | }; |
| 243 | #define DM_TARGET_IO_BIO_OFFSET (offsetof(struct dm_target_io, clone)) |
| 244 | #define DM_IO_BIO_OFFSET \ |
| 245 | (offsetof(struct dm_target_io, clone) + offsetof(struct dm_io, tio)) |
| 246 | |
| 247 | /* |
| 248 | * dm_target_io flags |
| 249 | */ |
| 250 | enum { |
| 251 | DM_TIO_INSIDE_DM_IO, |
| 252 | DM_TIO_IS_DUPLICATE_BIO |
| 253 | }; |
| 254 | |
| 255 | static inline bool dm_tio_flagged(struct dm_target_io *tio, unsigned int bit) |
| 256 | { |
| 257 | return (tio->flags & (1U << bit)) != 0; |
| 258 | } |
| 259 | |
| 260 | static inline void dm_tio_set_flag(struct dm_target_io *tio, unsigned int bit) |
| 261 | { |
| 262 | tio->flags |= (1U << bit); |
| 263 | } |
| 264 | |
| 265 | static inline bool dm_tio_is_normal(struct dm_target_io *tio) |
| 266 | { |
| 267 | return (dm_tio_flagged(tio, DM_TIO_INSIDE_DM_IO) && |
| 268 | !dm_tio_flagged(tio, DM_TIO_IS_DUPLICATE_BIO)); |
| 269 | } |
| 270 | |
| 271 | /* |
| 272 | * One of these is allocated per original bio. |
| 273 | * It contains the first clone used for that original. |
| 274 | */ |
| 275 | #define DM_IO_MAGIC 19577 |
| 276 | struct dm_io { |
| 277 | unsigned short magic; |
| 278 | blk_short_t flags; |
| 279 | spinlock_t lock; |
| 280 | unsigned long start_time; |
| 281 | void *data; |
| 282 | struct dm_io *next; |
| 283 | struct dm_stats_aux stats_aux; |
| 284 | blk_status_t status; |
| 285 | atomic_t io_count; |
| 286 | struct mapped_device *md; |
| 287 | |
| 288 | /* The three fields represent mapped part of original bio */ |
| 289 | struct bio *orig_bio; |
| 290 | unsigned int sector_offset; /* offset to end of orig_bio */ |
| 291 | unsigned int sectors; |
| 292 | |
| 293 | /* last member of dm_target_io is 'struct bio' */ |
| 294 | struct dm_target_io tio; |
| 295 | }; |
| 296 | |
| 297 | /* |
| 298 | * dm_io flags |
| 299 | */ |
| 300 | enum { |
| 301 | DM_IO_ACCOUNTED, |
| 302 | DM_IO_WAS_SPLIT |
| 303 | }; |
| 304 | |
| 305 | static inline bool dm_io_flagged(struct dm_io *io, unsigned int bit) |
| 306 | { |
| 307 | return (io->flags & (1U << bit)) != 0; |
| 308 | } |
| 309 | |
| 310 | static inline void dm_io_set_flag(struct dm_io *io, unsigned int bit) |
| 311 | { |
| 312 | io->flags |= (1U << bit); |
| 313 | } |
| 314 | |
| 315 | void dm_io_rewind(struct dm_io *io, struct bio_set *bs); |
| 316 | |
| 317 | static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) |
| 318 | { |
| 319 | return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; |
| 320 | } |
| 321 | |
| 322 | unsigned __dm_get_module_param(unsigned *module_param, unsigned def, unsigned max); |
| 323 | |
| 324 | static inline bool dm_message_test_buffer_overflow(char *result, unsigned maxlen) |
| 325 | { |
| 326 | return !maxlen || strlen(result) + 1 >= maxlen; |
| 327 | } |
| 328 | |
| 329 | extern atomic_t dm_global_event_nr; |
| 330 | extern wait_queue_head_t dm_global_eventq; |
| 331 | void dm_issue_global_event(void); |
| 332 | |
| 333 | #endif |