drivers/md/dm-bio-prison.c

   1 /*
   2  * Copyright (C) 2012 Red Hat, Inc.
   3  *
   4  * This file is released under the GPL.
   5  */
   6
   7 #include "dm.h"
   8 #include "dm-bio-prison.h"
   9
  10 #include <linux/spinlock.h>
  11 #include <linux/mempool.h>
  12 #include <linux/module.h>
  13 #include <linux/slab.h>
  14
  15 /*----------------------------------------------------------------*/
  16
  17 struct dm_bio_prison_cell {
  18         struct hlist_node list;
  19         struct dm_bio_prison *prison;
  20         struct dm_cell_key key;
  21         struct bio *holder;
  22         struct bio_list bios;
  23 };
  24
  25 struct dm_bio_prison {
  26         spinlock_t lock;
  27         mempool_t *cell_pool;
  28
  29         unsigned nr_buckets;
  30         unsigned hash_mask;
  31         struct hlist_head *cells;
  32 };
  33
  34 /*----------------------------------------------------------------*/
  35
  36 static uint32_t calc_nr_buckets(unsigned nr_cells)
  37 {
  38         uint32_t n = 128;
  39
  40         nr_cells /= 4;
  41         nr_cells = min(nr_cells, 8192u);
  42
  43         while (n < nr_cells)
  44                 n <<= 1;
  45
  46         return n;
  47 }
  48
  49 static struct kmem_cache *_cell_cache;
  50
  51 /*
  52  * @nr_cells should be the number of cells you want in use _concurrently_.
  53  * Don't confuse it with the number of distinct keys.
  54  */
  55 struct dm_bio_prison *dm_bio_prison_create(unsigned nr_cells)
  56 {
  57         unsigned i;
  58         uint32_t nr_buckets = calc_nr_buckets(nr_cells);
  59         size_t len = sizeof(struct dm_bio_prison) +
  60                 (sizeof(struct hlist_head) * nr_buckets);
  61         struct dm_bio_prison *prison = kmalloc(len, GFP_KERNEL);
  62
  63         if (!prison)
  64                 return NULL;
  65
  66         spin_lock_init(&prison->lock);
  67         prison->cell_pool = mempool_create_slab_pool(nr_cells, _cell_cache);
  68         if (!prison->cell_pool) {
  69                 kfree(prison);
  70                 return NULL;
  71         }
  72
  73         prison->nr_buckets = nr_buckets;
  74         prison->hash_mask = nr_buckets - 1;
  75         prison->cells = (struct hlist_head *) (prison + 1);
  76         for (i = 0; i < nr_buckets; i++)
  77                 INIT_HLIST_HEAD(prison->cells + i);
  78
  79         return prison;
  80 }
  81 EXPORT_SYMBOL_GPL(dm_bio_prison_create);
  82
  83 void dm_bio_prison_destroy(struct dm_bio_prison *prison)
  84 {
  85         mempool_destroy(prison->cell_pool);
  86         kfree(prison);
  87 }
  88 EXPORT_SYMBOL_GPL(dm_bio_prison_destroy);
  89
  90 static uint32_t hash_key(struct dm_bio_prison *prison, struct dm_cell_key *key)
  91 {
  92         const unsigned long BIG_PRIME = 4294967291UL;
  93         uint64_t hash = key->block * BIG_PRIME;
  94
  95         return (uint32_t) (hash & prison->hash_mask);
  96 }
  97
  98 static int keys_equal(struct dm_cell_key *lhs, struct dm_cell_key *rhs)
  99 {
 100                return (lhs->virtual == rhs->virtual) &&
 101                        (lhs->dev == rhs->dev) &&
 102                        (lhs->block == rhs->block);
 103 }
 104
 105 static struct dm_bio_prison_cell *__search_bucket(struct hlist_head *bucket,
 106                                                   struct dm_cell_key *key)
 107 {
 108         struct dm_bio_prison_cell *cell;
 109
 110         hlist_for_each_entry(cell, bucket, list)
 111                 if (keys_equal(&cell->key, key))
 112                         return cell;
 113
 114         return NULL;
 115 }
 116
 117 /*
 118  * This may block if a new cell needs allocating.  You must ensure that
 119  * cells will be unlocked even if the calling thread is blocked.
 120  *
 121  * Returns 1 if the cell was already held, 0 if @inmate is the new holder.
 122  */
 123 int dm_bio_detain(struct dm_bio_prison *prison, struct dm_cell_key *key,
 124                   struct bio *inmate, struct dm_bio_prison_cell **ref)
 125 {
 126         int r = 1;
 127         unsigned long flags;
 128         uint32_t hash = hash_key(prison, key);
 129         struct dm_bio_prison_cell *cell, *cell2;
 130
 131         BUG_ON(hash > prison->nr_buckets);
 132
 133         spin_lock_irqsave(&prison->lock, flags);
 134
 135         cell = __search_bucket(prison->cells + hash, key);
 136         if (cell) {
 137                 bio_list_add(&cell->bios, inmate);
 138                 goto out;
 139         }
 140
 141         /*
 142          * Allocate a new cell
 143          */
 144         spin_unlock_irqrestore(&prison->lock, flags);
 145         cell2 = mempool_alloc(prison->cell_pool, GFP_NOIO);
 146         spin_lock_irqsave(&prison->lock, flags);
 147
 148         /*
 149          * We've been unlocked, so we have to double check that
 150          * nobody else has inserted this cell in the meantime.
 151          */
 152         cell = __search_bucket(prison->cells + hash, key);
 153         if (cell) {
 154                 mempool_free(cell2, prison->cell_pool);
 155                 bio_list_add(&cell->bios, inmate);
 156                 goto out;
 157         }
 158
 159         /*
 160          * Use new cell.
 161          */
 162         cell = cell2;
 163
 164         cell->prison = prison;
 165         memcpy(&cell->key, key, sizeof(cell->key));
 166         cell->holder = inmate;
 167         bio_list_init(&cell->bios);
 168         hlist_add_head(&cell->list, prison->cells + hash);
 169
 170         r = 0;
 171
 172 out:
 173         spin_unlock_irqrestore(&prison->lock, flags);
 174
 175         *ref = cell;
 176
 177         return r;
 178 }
 179 EXPORT_SYMBOL_GPL(dm_bio_detain);
 180
 181 /*
 182  * @inmates must have been initialised prior to this call
 183  */
 184 static void __cell_release(struct dm_bio_prison_cell *cell, struct bio_list *inmates)
 185 {
 186         struct dm_bio_prison *prison = cell->prison;
 187
 188         hlist_del(&cell->list);
 189
 190         if (inmates) {
 191                 bio_list_add(inmates, cell->holder);
 192                 bio_list_merge(inmates, &cell->bios);
 193         }
 194
 195         mempool_free(cell, prison->cell_pool);
 196 }
 197
 198 void dm_cell_release(struct dm_bio_prison_cell *cell, struct bio_list *bios)
 199 {
 200         unsigned long flags;
 201         struct dm_bio_prison *prison = cell->prison;
 202
 203         spin_lock_irqsave(&prison->lock, flags);
 204         __cell_release(cell, bios);
 205         spin_unlock_irqrestore(&prison->lock, flags);
 206 }
 207 EXPORT_SYMBOL_GPL(dm_cell_release);
 208
 209 /*
 210  * Sometimes we don't want the holder, just the additional bios.
 211  */
 212 static void __cell_release_no_holder(struct dm_bio_prison_cell *cell, struct bio_list *inmates)
 213 {
 214         struct dm_bio_prison *prison = cell->prison;
 215
 216         hlist_del(&cell->list);
 217         bio_list_merge(inmates, &cell->bios);
 218
 219         mempool_free(cell, prison->cell_pool);
 220 }
 221
 222 void dm_cell_release_no_holder(struct dm_bio_prison_cell *cell, struct bio_list *inmates)
 223 {
 224         unsigned long flags;
 225         struct dm_bio_prison *prison = cell->prison;
 226
 227         spin_lock_irqsave(&prison->lock, flags);
 228         __cell_release_no_holder(cell, inmates);
 229         spin_unlock_irqrestore(&prison->lock, flags);
 230 }
 231 EXPORT_SYMBOL_GPL(dm_cell_release_no_holder);
 232
 233 void dm_cell_error(struct dm_bio_prison_cell *cell)
 234 {
 235         struct dm_bio_prison *prison = cell->prison;
 236         struct bio_list bios;
 237         struct bio *bio;
 238         unsigned long flags;
 239
 240         bio_list_init(&bios);
 241
 242         spin_lock_irqsave(&prison->lock, flags);
 243         __cell_release(cell, &bios);
 244         spin_unlock_irqrestore(&prison->lock, flags);
 245
 246         while ((bio = bio_list_pop(&bios)))
 247                 bio_io_error(bio);
 248 }
 249 EXPORT_SYMBOL_GPL(dm_cell_error);
 250
 251 /*----------------------------------------------------------------*/
 252
 253 #define DEFERRED_SET_SIZE 64
 254
 255 struct dm_deferred_entry {
 256         struct dm_deferred_set *ds;
 257         unsigned count;
 258         struct list_head work_items;
 259 };
 260
 261 struct dm_deferred_set {
 262         spinlock_t lock;
 263         unsigned current_entry;
 264         unsigned sweeper;
 265         struct dm_deferred_entry entries[DEFERRED_SET_SIZE];
 266 };
 267
 268 struct dm_deferred_set *dm_deferred_set_create(void)
 269 {
 270         int i;
 271         struct dm_deferred_set *ds;
 272
 273         ds = kmalloc(sizeof(*ds), GFP_KERNEL);
 274         if (!ds)
 275                 return NULL;
 276
 277         spin_lock_init(&ds->lock);
 278         ds->current_entry = 0;
 279         ds->sweeper = 0;
 280         for (i = 0; i < DEFERRED_SET_SIZE; i++) {
 281                 ds->entries[i].ds = ds;
 282                 ds->entries[i].count = 0;
 283                 INIT_LIST_HEAD(&ds->entries[i].work_items);
 284         }
 285
 286         return ds;
 287 }
 288 EXPORT_SYMBOL_GPL(dm_deferred_set_create);
 289
 290 void dm_deferred_set_destroy(struct dm_deferred_set *ds)
 291 {
 292         kfree(ds);
 293 }
 294 EXPORT_SYMBOL_GPL(dm_deferred_set_destroy);
 295
 296 struct dm_deferred_entry *dm_deferred_entry_inc(struct dm_deferred_set *ds)
 297 {
 298         unsigned long flags;
 299         struct dm_deferred_entry *entry;
 300
 301         spin_lock_irqsave(&ds->lock, flags);
 302         entry = ds->entries + ds->current_entry;
 303         entry->count++;
 304         spin_unlock_irqrestore(&ds->lock, flags);
 305
 306         return entry;
 307 }
 308 EXPORT_SYMBOL_GPL(dm_deferred_entry_inc);
 309
 310 static unsigned ds_next(unsigned index)
 311 {
 312         return (index + 1) % DEFERRED_SET_SIZE;
 313 }
 314
 315 static void __sweep(struct dm_deferred_set *ds, struct list_head *head)
 316 {
 317         while ((ds->sweeper != ds->current_entry) &&
 318                !ds->entries[ds->sweeper].count) {
 319                 list_splice_init(&ds->entries[ds->sweeper].work_items, head);
 320                 ds->sweeper = ds_next(ds->sweeper);
 321         }
 322
 323         if ((ds->sweeper == ds->current_entry) && !ds->entries[ds->sweeper].count)
 324                 list_splice_init(&ds->entries[ds->sweeper].work_items, head);
 325 }
 326
 327 void dm_deferred_entry_dec(struct dm_deferred_entry *entry, struct list_head *head)
 328 {
 329         unsigned long flags;
 330
 331         spin_lock_irqsave(&entry->ds->lock, flags);
 332         BUG_ON(!entry->count);
 333         --entry->count;
 334         __sweep(entry->ds, head);
 335         spin_unlock_irqrestore(&entry->ds->lock, flags);
 336 }
 337 EXPORT_SYMBOL_GPL(dm_deferred_entry_dec);
 338
 339 /*
 340  * Returns 1 if deferred or 0 if no pending items to delay job.
 341  */
 342 int dm_deferred_set_add_work(struct dm_deferred_set *ds, struct list_head *work)
 343 {
 344         int r = 1;
 345         unsigned long flags;
 346         unsigned next_entry;
 347
 348         spin_lock_irqsave(&ds->lock, flags);
 349         if ((ds->sweeper == ds->current_entry) &&
 350             !ds->entries[ds->current_entry].count)
 351                 r = 0;
 352         else {
 353                 list_add(work, &ds->entries[ds->current_entry].work_items);
 354                 next_entry = ds_next(ds->current_entry);
 355                 if (!ds->entries[next_entry].count)
 356                         ds->current_entry = next_entry;
 357         }
 358         spin_unlock_irqrestore(&ds->lock, flags);
 359
 360         return r;
 361 }
 362 EXPORT_SYMBOL_GPL(dm_deferred_set_add_work);
 363
 364 /*----------------------------------------------------------------*/
 365
 366 static int __init dm_bio_prison_init(void)
 367 {
 368         _cell_cache = KMEM_CACHE(dm_bio_prison_cell, 0);
 369         if (!_cell_cache)
 370                 return -ENOMEM;
 371
 372         return 0;
 373 }
 374
 375 static void __exit dm_bio_prison_exit(void)
 376 {
 377         kmem_cache_destroy(_cell_cache);
 378         _cell_cache = NULL;
 379 }
 380
 381 /*
 382  * module hooks
 383  */
 384 module_init(dm_bio_prison_init);
 385 module_exit(dm_bio_prison_exit);
 386
 387 MODULE_DESCRIPTION(DM_NAME " bio prison");
 388 MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
 389 MODULE_LICENSE("GPL");