drivers/md/bcache/btree.c

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
write_block
bch_btree_init_next
bkey_put
btree_csum_set
bch_btree_node_read_done
btree_node_read_endio
bch_btree_node_read
btree_complete_write
btree_node_write_unlock
__btree_node_write_done
btree_node_write_done
btree_node_write_endio
do_btree_node_write
__bch_btree_node_write
bch_btree_node_write
bch_btree_node_write_sync
btree_node_write_work
bch_btree_leaf_dirty
mca_data_free
mca_bucket_free
btree_order
mca_data_alloc
mca_bucket_alloc
mca_reap
bch_mca_scan
bch_mca_count
bch_btree_cache_free
bch_btree_cache_alloc
mca_hash
mca_find
mca_cannibalize_lock
mca_cannibalize
bch_cannibalize_unlock
mca_alloc
bch_btree_node_get
btree_node_prefetch
btree_node_free
__bch_btree_node_alloc
bch_btree_node_alloc
btree_node_alloc_replacement
make_btree_freeing_key
btree_check_reserve
__bch_btree_mark_key
bch_initial_mark_key
bch_update_bucket_in_use
btree_gc_mark_node
btree_gc_coalesce
btree_gc_rewrite_node
btree_gc_count_keys
btree_gc_min_nodes
btree_gc_recurse
bch_btree_gc_root
btree_gc_start
bch_btree_gc_finish
bch_btree_gc
gc_should_run
bch_gc_thread
bch_gc_thread_start
bch_btree_check_recurse
bch_btree_check
bch_initial_gc_finish
btree_insert_key
insert_u64s_remaining
bch_btree_insert_keys
btree_split
bch_btree_insert_node
bch_btree_insert_check_key
btree_insert_fn
bch_btree_insert
bch_btree_set_root
bch_btree_map_nodes_recurse
__bch_btree_map_nodes
bch_btree_map_keys_recurse
bch_btree_map_keys
keybuf_cmp
keybuf_nonoverlapping_cmp
refill_keybuf_fn
bch_refill_keybuf
__bch_keybuf_del
bch_keybuf_del
bch_keybuf_check_overlapping
bch_keybuf_next
bch_keybuf_next_rescan
bch_keybuf_init
   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2010 Kent Overstreet <kent.overstreet@gmail.com>
   4  *
   5  * Uses a block device as cache for other block devices; optimized for SSDs.
   6  * All allocation is done in buckets, which should match the erase block size
   7  * of the device.
   8  *
   9  * Buckets containing cached data are kept on a heap sorted by priority;
  10  * bucket priority is increased on cache hit, and periodically all the buckets
  11  * on the heap have their priority scaled down. This currently is just used as
  12  * an LRU but in the future should allow for more intelligent heuristics.
  13  *
  14  * Buckets have an 8 bit counter; freeing is accomplished by incrementing the
  15  * counter. Garbage collection is used to remove stale pointers.
  16  *
  17  * Indexing is done via a btree; nodes are not necessarily fully sorted, rather
  18  * as keys are inserted we only sort the pages that have not yet been written.
  19  * When garbage collection is run, we resort the entire node.
  20  *
  21  * All configuration is done via sysfs; see Documentation/admin-guide/bcache.rst.
  22  */
  23 
  24 #include "bcache.h"
  25 #include "btree.h"
  26 #include "debug.h"
  27 #include "extents.h"
  28 
  29 #include <linux/slab.h>
  30 #include <linux/bitops.h>
  31 #include <linux/hash.h>
  32 #include <linux/kthread.h>
  33 #include <linux/prefetch.h>
  34 #include <linux/random.h>
  35 #include <linux/rcupdate.h>
  36 #include <linux/sched/clock.h>
  37 #include <linux/rculist.h>
  38 #include <linux/delay.h>
  39 #include <trace/events/bcache.h>
  40 
  41 /*
  42  * Todo:
  43  * register_bcache: Return errors out to userspace correctly
  44  *
  45  * Writeback: don't undirty key until after a cache flush
  46  *
  47  * Create an iterator for key pointers
  48  *
  49  * On btree write error, mark bucket such that it won't be freed from the cache
  50  *
  51  * Journalling:
  52  *   Check for bad keys in replay
  53  *   Propagate barriers
  54  *   Refcount journal entries in journal_replay
  55  *
  56  * Garbage collection:
  57  *   Finish incremental gc
  58  *   Gc should free old UUIDs, data for invalid UUIDs
  59  *
  60  * Provide a way to list backing device UUIDs we have data cached for, and
  61  * probably how long it's been since we've seen them, and a way to invalidate
  62  * dirty data for devices that will never be attached again
  63  *
  64  * Keep 1 min/5 min/15 min statistics of how busy a block device has been, so
  65  * that based on that and how much dirty data we have we can keep writeback
  66  * from being starved
  67  *
  68  * Add a tracepoint or somesuch to watch for writeback starvation
  69  *
  70  * When btree depth > 1 and splitting an interior node, we have to make sure
  71  * alloc_bucket() cannot fail. This should be true but is not completely
  72  * obvious.
  73  *
  74  * Plugging?
  75  *
  76  * If data write is less than hard sector size of ssd, round up offset in open
  77  * bucket to the next whole sector
  78  *
  79  * Superblock needs to be fleshed out for multiple cache devices
  80  *
  81  * Add a sysfs tunable for the number of writeback IOs in flight
  82  *
  83  * Add a sysfs tunable for the number of open data buckets
  84  *
  85  * IO tracking: Can we track when one process is doing io on behalf of another?
  86  * IO tracking: Don't use just an average, weigh more recent stuff higher
  87  *
  88  * Test module load/unload
  89  */
  90 
  91 #define MAX_NEED_GC             64
  92 #define MAX_SAVE_PRIO           72
  93 #define MAX_GC_TIMES            100
  94 #define MIN_GC_NODES            100
  95 #define GC_SLEEP_MS             100
  96 
  97 #define PTR_DIRTY_BIT           (((uint64_t) 1 << 36))
  98 
  99 #define PTR_HASH(c, k)                                                  \
 100         (((k)->ptr[0] >> c->bucket_bits) | PTR_GEN(k, 0))
 101 
 102 #define insert_lock(s, b)       ((b)->level <= (s)->lock)
 103 
 104 /*
 105  * These macros are for recursing down the btree - they handle the details of
 106  * locking and looking up nodes in the cache for you. They're best treated as
 107  * mere syntax when reading code that uses them.
 108  *
 109  * op->lock determines whether we take a read or a write lock at a given depth.
 110  * If you've got a read lock and find that you need a write lock (i.e. you're
 111  * going to have to split), set op->lock and return -EINTR; btree_root() will
 112  * call you again and you'll have the correct lock.
 113  */
 114 
 115 /**
 116  * btree - recurse down the btree on a specified key
 117  * @fn:         function to call, which will be passed the child node
 118  * @key:        key to recurse on
 119  * @b:          parent btree node
 120  * @op:         pointer to struct btree_op
 121  */
 122 #define btree(fn, key, b, op, ...)                                      \
 123 ({                                                                      \
 124         int _r, l = (b)->level - 1;                                     \
 125         bool _w = l <= (op)->lock;                                      \
 126         struct btree *_child = bch_btree_node_get((b)->c, op, key, l,   \
 127                                                   _w, b);               \
 128         if (!IS_ERR(_child)) {                                          \
 129                 _r = bch_btree_ ## fn(_child, op, ##__VA_ARGS__);       \
 130                 rw_unlock(_w, _child);                                  \
 131         } else                                                          \
 132                 _r = PTR_ERR(_child);                                   \
 133         _r;                                                             \
 134 })
 135 
 136 /**
 137  * btree_root - call a function on the root of the btree
 138  * @fn:         function to call, which will be passed the child node
 139  * @c:          cache set
 140  * @op:         pointer to struct btree_op
 141  */
 142 #define btree_root(fn, c, op, ...)                                      \
 143 ({                                                                      \
 144         int _r = -EINTR;                                                \
 145         do {                                                            \
 146                 struct btree *_b = (c)->root;                           \
 147                 bool _w = insert_lock(op, _b);                          \
 148                 rw_lock(_w, _b, _b->level);                             \
 149                 if (_b == (c)->root &&                                  \
 150                     _w == insert_lock(op, _b)) {                        \
 151                         _r = bch_btree_ ## fn(_b, op, ##__VA_ARGS__);   \
 152                 }                                                       \
 153                 rw_unlock(_w, _b);                                      \
 154                 bch_cannibalize_unlock(c);                              \
 155                 if (_r == -EINTR)                                       \
 156                         schedule();                                     \
 157         } while (_r == -EINTR);                                         \
 158                                                                         \
 159         finish_wait(&(c)->btree_cache_wait, &(op)->wait);               \
 160         _r;                                                             \
 161 })
 162 
 163 static inline struct bset *write_block(struct btree *b)
 164 {
 165         return ((void *) btree_bset_first(b)) + b->written * block_bytes(b->c);
 166 }
 167 
 168 static void bch_btree_init_next(struct btree *b)
 169 {
 170         /* If not a leaf node, always sort */
 171         if (b->level && b->keys.nsets)
 172                 bch_btree_sort(&b->keys, &b->c->sort);
 173         else
 174                 bch_btree_sort_lazy(&b->keys, &b->c->sort);
 175 
 176         if (b->written < btree_blocks(b))
 177                 bch_bset_init_next(&b->keys, write_block(b),
 178                                    bset_magic(&b->c->sb));
 179 
 180 }
 181 
 182 /* Btree key manipulation */
 183 
 184 void bkey_put(struct cache_set *c, struct bkey *k)
 185 {
 186         unsigned int i;
 187 
 188         for (i = 0; i < KEY_PTRS(k); i++)
 189                 if (ptr_available(c, k, i))
 190                         atomic_dec_bug(&PTR_BUCKET(c, k, i)->pin);
 191 }
 192 
 193 /* Btree IO */
 194 
 195 static uint64_t btree_csum_set(struct btree *b, struct bset *i)
 196 {
 197         uint64_t crc = b->key.ptr[0];
 198         void *data = (void *) i + 8, *end = bset_bkey_last(i);
 199 
 200         crc = bch_crc64_update(crc, data, end - data);
 201         return crc ^ 0xffffffffffffffffULL;
 202 }
 203 
 204 void bch_btree_node_read_done(struct btree *b)
 205 {
 206         const char *err = "bad btree header";
 207         struct bset *i = btree_bset_first(b);
 208         struct btree_iter *iter;
 209 
 210         /*
 211          * c->fill_iter can allocate an iterator with more memory space
 212          * than static MAX_BSETS.
 213          * See the comment arount cache_set->fill_iter.
 214          */
 215         iter = mempool_alloc(&b->c->fill_iter, GFP_NOIO);
 216         iter->size = b->c->sb.bucket_size / b->c->sb.block_size;
 217         iter->used = 0;
 218 
 219 #ifdef CONFIG_BCACHE_DEBUG
 220         iter->b = &b->keys;
 221 #endif
 222 
 223         if (!i->seq)
 224                 goto err;
 225 
 226         for (;
 227              b->written < btree_blocks(b) && i->seq == b->keys.set[0].data->seq;
 228              i = write_block(b)) {
 229                 err = "unsupported bset version";
 230                 if (i->version > BCACHE_BSET_VERSION)
 231                         goto err;
 232 
 233                 err = "bad btree header";
 234                 if (b->written + set_blocks(i, block_bytes(b->c)) >
 235                     btree_blocks(b))
 236                         goto err;
 237 
 238                 err = "bad magic";
 239                 if (i->magic != bset_magic(&b->c->sb))
 240                         goto err;
 241 
 242                 err = "bad checksum";
 243                 switch (i->version) {
 244                 case 0:
 245                         if (i->csum != csum_set(i))
 246                                 goto err;
 247                         break;
 248                 case BCACHE_BSET_VERSION:
 249                         if (i->csum != btree_csum_set(b, i))
 250                                 goto err;
 251                         break;
 252                 }
 253 
 254                 err = "empty set";
 255                 if (i != b->keys.set[0].data && !i->keys)
 256                         goto err;
 257 
 258                 bch_btree_iter_push(iter, i->start, bset_bkey_last(i));
 259 
 260                 b->written += set_blocks(i, block_bytes(b->c));
 261         }
 262 
 263         err = "corrupted btree";
 264         for (i = write_block(b);
 265              bset_sector_offset(&b->keys, i) < KEY_SIZE(&b->key);
 266              i = ((void *) i) + block_bytes(b->c))
 267                 if (i->seq == b->keys.set[0].data->seq)
 268                         goto err;
 269 
 270         bch_btree_sort_and_fix_extents(&b->keys, iter, &b->c->sort);
 271 
 272         i = b->keys.set[0].data;
 273         err = "short btree key";
 274         if (b->keys.set[0].size &&
 275             bkey_cmp(&b->key, &b->keys.set[0].end) < 0)
 276                 goto err;
 277 
 278         if (b->written < btree_blocks(b))
 279                 bch_bset_init_next(&b->keys, write_block(b),
 280                                    bset_magic(&b->c->sb));
 281 out:
 282         mempool_free(iter, &b->c->fill_iter);
 283         return;
 284 err:
 285         set_btree_node_io_error(b);
 286         bch_cache_set_error(b->c, "%s at bucket %zu, block %u, %u keys",
 287                             err, PTR_BUCKET_NR(b->c, &b->key, 0),
 288                             bset_block_offset(b, i), i->keys);
 289         goto out;
 290 }
 291 
 292 static void btree_node_read_endio(struct bio *bio)
 293 {
 294         struct closure *cl = bio->bi_private;
 295 
 296         closure_put(cl);
 297 }
 298 
 299 static void bch_btree_node_read(struct btree *b)
 300 {
 301         uint64_t start_time = local_clock();
 302         struct closure cl;
 303         struct bio *bio;
 304 
 305         trace_bcache_btree_read(b);
 306 
 307         closure_init_stack(&cl);
 308 
 309         bio = bch_bbio_alloc(b->c);
 310         bio->bi_iter.bi_size = KEY_SIZE(&b->key) << 9;
 311         bio->bi_end_io  = btree_node_read_endio;
 312         bio->bi_private = &cl;
 313         bio->bi_opf = REQ_OP_READ | REQ_META;
 314 
 315         bch_bio_map(bio, b->keys.set[0].data);
 316 
 317         bch_submit_bbio(bio, b->c, &b->key, 0);
 318         closure_sync(&cl);
 319 
 320         if (bio->bi_status)
 321                 set_btree_node_io_error(b);
 322 
 323         bch_bbio_free(bio, b->c);
 324 
 325         if (btree_node_io_error(b))
 326                 goto err;
 327 
 328         bch_btree_node_read_done(b);
 329         bch_time_stats_update(&b->c->btree_read_time, start_time);
 330 
 331         return;
 332 err:
 333         bch_cache_set_error(b->c, "io error reading bucket %zu",
 334                             PTR_BUCKET_NR(b->c, &b->key, 0));
 335 }
 336 
 337 static void btree_complete_write(struct btree *b, struct btree_write *w)
 338 {
 339         if (w->prio_blocked &&
 340             !atomic_sub_return(w->prio_blocked, &b->c->prio_blocked))
 341                 wake_up_allocators(b->c);
 342 
 343         if (w->journal) {
 344                 atomic_dec_bug(w->journal);
 345                 __closure_wake_up(&b->c->journal.wait);
 346         }
 347 
 348         w->prio_blocked = 0;
 349         w->journal      = NULL;
 350 }
 351 
 352 static void btree_node_write_unlock(struct closure *cl)
 353 {
 354         struct btree *b = container_of(cl, struct btree, io);
 355 
 356         up(&b->io_mutex);
 357 }
 358 
 359 static void __btree_node_write_done(struct closure *cl)
 360 {
 361         struct btree *b = container_of(cl, struct btree, io);
 362         struct btree_write *w = btree_prev_write(b);
 363 
 364         bch_bbio_free(b->bio, b->c);
 365         b->bio = NULL;
 366         btree_complete_write(b, w);
 367 
 368         if (btree_node_dirty(b))
 369                 schedule_delayed_work(&b->work, 30 * HZ);
 370 
 371         closure_return_with_destructor(cl, btree_node_write_unlock);
 372 }
 373 
 374 static void btree_node_write_done(struct closure *cl)
 375 {
 376         struct btree *b = container_of(cl, struct btree, io);
 377 
 378         bio_free_pages(b->bio);
 379         __btree_node_write_done(cl);
 380 }
 381 
 382 static void btree_node_write_endio(struct bio *bio)
 383 {
 384         struct closure *cl = bio->bi_private;
 385         struct btree *b = container_of(cl, struct btree, io);
 386 
 387         if (bio->bi_status)
 388                 set_btree_node_io_error(b);
 389 
 390         bch_bbio_count_io_errors(b->c, bio, bio->bi_status, "writing btree");
 391         closure_put(cl);
 392 }
 393 
 394 static void do_btree_node_write(struct btree *b)
 395 {
 396         struct closure *cl = &b->io;
 397         struct bset *i = btree_bset_last(b);
 398         BKEY_PADDED(key) k;
 399 
 400         i->version      = BCACHE_BSET_VERSION;
 401         i->csum         = btree_csum_set(b, i);
 402 
 403         BUG_ON(b->bio);
 404         b->bio = bch_bbio_alloc(b->c);
 405 
 406         b->bio->bi_end_io       = btree_node_write_endio;
 407         b->bio->bi_private      = cl;
 408         b->bio->bi_iter.bi_size = roundup(set_bytes(i), block_bytes(b->c));
 409         b->bio->bi_opf          = REQ_OP_WRITE | REQ_META | REQ_FUA;
 410         bch_bio_map(b->bio, i);
 411 
 412         /*
 413          * If we're appending to a leaf node, we don't technically need FUA -
 414          * this write just needs to be persisted before the next journal write,
 415          * which will be marked FLUSH|FUA.
 416          *
 417          * Similarly if we're writing a new btree root - the pointer is going to
 418          * be in the next journal entry.
 419          *
 420          * But if we're writing a new btree node (that isn't a root) or
 421          * appending to a non leaf btree node, we need either FUA or a flush
 422          * when we write the parent with the new pointer. FUA is cheaper than a
 423          * flush, and writes appending to leaf nodes aren't blocking anything so
 424          * just make all btree node writes FUA to keep things sane.
 425          */
 426 
 427         bkey_copy(&k.key, &b->key);
 428         SET_PTR_OFFSET(&k.key, 0, PTR_OFFSET(&k.key, 0) +
 429                        bset_sector_offset(&b->keys, i));
 430 
 431         if (!bch_bio_alloc_pages(b->bio, __GFP_NOWARN|GFP_NOWAIT)) {
 432                 struct bio_vec *bv;
 433                 void *addr = (void *) ((unsigned long) i & ~(PAGE_SIZE - 1));
 434                 struct bvec_iter_all iter_all;
 435 
 436                 bio_for_each_segment_all(bv, b->bio, iter_all) {
 437                         memcpy(page_address(bv->bv_page), addr, PAGE_SIZE);
 438                         addr += PAGE_SIZE;
 439                 }
 440 
 441                 bch_submit_bbio(b->bio, b->c, &k.key, 0);
 442 
 443                 continue_at(cl, btree_node_write_done, NULL);
 444         } else {
 445                 /*
 446                  * No problem for multipage bvec since the bio is
 447                  * just allocated
 448                  */
 449                 b->bio->bi_vcnt = 0;
 450                 bch_bio_map(b->bio, i);
 451 
 452                 bch_submit_bbio(b->bio, b->c, &k.key, 0);
 453 
 454                 closure_sync(cl);
 455                 continue_at_nobarrier(cl, __btree_node_write_done, NULL);
 456         }
 457 }
 458 
 459 void __bch_btree_node_write(struct btree *b, struct closure *parent)
 460 {
 461         struct bset *i = btree_bset_last(b);
 462 
 463         lockdep_assert_held(&b->write_lock);
 464 
 465         trace_bcache_btree_write(b);
 466 
 467         BUG_ON(current->bio_list);
 468         BUG_ON(b->written >= btree_blocks(b));
 469         BUG_ON(b->written && !i->keys);
 470         BUG_ON(btree_bset_first(b)->seq != i->seq);
 471         bch_check_keys(&b->keys, "writing");
 472 
 473         cancel_delayed_work(&b->work);
 474 
 475         /* If caller isn't waiting for write, parent refcount is cache set */
 476         down(&b->io_mutex);
 477         closure_init(&b->io, parent ?: &b->c->cl);
 478 
 479         clear_bit(BTREE_NODE_dirty,      &b->flags);
 480         change_bit(BTREE_NODE_write_idx, &b->flags);
 481 
 482         do_btree_node_write(b);
 483 
 484         atomic_long_add(set_blocks(i, block_bytes(b->c)) * b->c->sb.block_size,
 485                         &PTR_CACHE(b->c, &b->key, 0)->btree_sectors_written);
 486 
 487         b->written += set_blocks(i, block_bytes(b->c));
 488 }
 489 
 490 void bch_btree_node_write(struct btree *b, struct closure *parent)
 491 {
 492         unsigned int nsets = b->keys.nsets;
 493 
 494         lockdep_assert_held(&b->lock);
 495 
 496         __bch_btree_node_write(b, parent);
 497 
 498         /*
 499          * do verify if there was more than one set initially (i.e. we did a
 500          * sort) and we sorted down to a single set:
 501          */
 502         if (nsets && !b->keys.nsets)
 503                 bch_btree_verify(b);
 504 
 505         bch_btree_init_next(b);
 506 }
 507 
 508 static void bch_btree_node_write_sync(struct btree *b)
 509 {
 510         struct closure cl;
 511 
 512         closure_init_stack(&cl);
 513 
 514         mutex_lock(&b->write_lock);
 515         bch_btree_node_write(b, &cl);
 516         mutex_unlock(&b->write_lock);
 517 
 518         closure_sync(&cl);
 519 }
 520 
 521 static void btree_node_write_work(struct work_struct *w)
 522 {
 523         struct btree *b = container_of(to_delayed_work(w), struct btree, work);
 524 
 525         mutex_lock(&b->write_lock);
 526         if (btree_node_dirty(b))
 527                 __bch_btree_node_write(b, NULL);
 528         mutex_unlock(&b->write_lock);
 529 }
 530 
 531 static void bch_btree_leaf_dirty(struct btree *b, atomic_t *journal_ref)
 532 {
 533         struct bset *i = btree_bset_last(b);
 534         struct btree_write *w = btree_current_write(b);
 535 
 536         lockdep_assert_held(&b->write_lock);
 537 
 538         BUG_ON(!b->written);
 539         BUG_ON(!i->keys);
 540 
 541         if (!btree_node_dirty(b))
 542                 schedule_delayed_work(&b->work, 30 * HZ);
 543 
 544         set_btree_node_dirty(b);
 545 
 546         if (journal_ref) {
 547                 if (w->journal &&
 548                     journal_pin_cmp(b->c, w->journal, journal_ref)) {
 549                         atomic_dec_bug(w->journal);
 550                         w->journal = NULL;
 551                 }
 552 
 553                 if (!w->journal) {
 554                         w->journal = journal_ref;
 555                         atomic_inc(w->journal);
 556                 }
 557         }
 558 
 559         /* Force write if set is too big */
 560         if (set_bytes(i) > PAGE_SIZE - 48 &&
 561             !current->bio_list)
 562                 bch_btree_node_write(b, NULL);
 563 }
 564 
 565 /*
 566  * Btree in memory cache - allocation/freeing
 567  * mca -> memory cache
 568  */
 569 
 570 #define mca_reserve(c)  (((c->root && c->root->level)           \
 571                           ? c->root->level : 1) * 8 + 16)
 572 #define mca_can_free(c)                                         \
 573         max_t(int, 0, c->btree_cache_used - mca_reserve(c))
 574 
 575 static void mca_data_free(struct btree *b)
 576 {
 577         BUG_ON(b->io_mutex.count != 1);
 578 
 579         bch_btree_keys_free(&b->keys);
 580 
 581         b->c->btree_cache_used--;
 582         list_move(&b->list, &b->c->btree_cache_freed);
 583 }
 584 
 585 static void mca_bucket_free(struct btree *b)
 586 {
 587         BUG_ON(btree_node_dirty(b));
 588 
 589         b->key.ptr[0] = 0;
 590         hlist_del_init_rcu(&b->hash);
 591         list_move(&b->list, &b->c->btree_cache_freeable);
 592 }
 593 
 594 static unsigned int btree_order(struct bkey *k)
 595 {
 596         return ilog2(KEY_SIZE(k) / PAGE_SECTORS ?: 1);
 597 }
 598 
 599 static void mca_data_alloc(struct btree *b, struct bkey *k, gfp_t gfp)
 600 {
 601         if (!bch_btree_keys_alloc(&b->keys,
 602                                   max_t(unsigned int,
 603                                         ilog2(b->c->btree_pages),
 604                                         btree_order(k)),
 605                                   gfp)) {
 606                 b->c->btree_cache_used++;
 607                 list_move(&b->list, &b->c->btree_cache);
 608         } else {
 609                 list_move(&b->list, &b->c->btree_cache_freed);
 610         }
 611 }
 612 
 613 static struct btree *mca_bucket_alloc(struct cache_set *c,
 614                                       struct bkey *k, gfp_t gfp)
 615 {
 616         /*
 617          * kzalloc() is necessary here for initialization,
 618          * see code comments in bch_btree_keys_init().
 619          */
 620         struct btree *b = kzalloc(sizeof(struct btree), gfp);
 621 
 622         if (!b)
 623                 return NULL;
 624 
 625         init_rwsem(&b->lock);
 626         lockdep_set_novalidate_class(&b->lock);
 627         mutex_init(&b->write_lock);
 628         lockdep_set_novalidate_class(&b->write_lock);
 629         INIT_LIST_HEAD(&b->list);
 630         INIT_DELAYED_WORK(&b->work, btree_node_write_work);
 631         b->c = c;
 632         sema_init(&b->io_mutex, 1);
 633 
 634         mca_data_alloc(b, k, gfp);
 635         return b;
 636 }
 637 
 638 static int mca_reap(struct btree *b, unsigned int min_order, bool flush)
 639 {
 640         struct closure cl;
 641 
 642         closure_init_stack(&cl);
 643         lockdep_assert_held(&b->c->bucket_lock);
 644 
 645         if (!down_write_trylock(&b->lock))
 646                 return -ENOMEM;
 647 
 648         BUG_ON(btree_node_dirty(b) && !b->keys.set[0].data);
 649 
 650         if (b->keys.page_order < min_order)
 651                 goto out_unlock;
 652 
 653         if (!flush) {
 654                 if (btree_node_dirty(b))
 655                         goto out_unlock;
 656 
 657                 if (down_trylock(&b->io_mutex))
 658                         goto out_unlock;
 659                 up(&b->io_mutex);
 660         }
 661 
 662 retry:
 663         /*
 664          * BTREE_NODE_dirty might be cleared in btree_flush_btree() by
 665          * __bch_btree_node_write(). To avoid an extra flush, acquire
 666          * b->write_lock before checking BTREE_NODE_dirty bit.
 667          */
 668         mutex_lock(&b->write_lock);
 669         /*
 670          * If this btree node is selected in btree_flush_write() by journal
 671          * code, delay and retry until the node is flushed by journal code
 672          * and BTREE_NODE_journal_flush bit cleared by btree_flush_write().
 673          */
 674         if (btree_node_journal_flush(b)) {
 675                 pr_debug("bnode %p is flushing by journal, retry", b);
 676                 mutex_unlock(&b->write_lock);
 677                 udelay(1);
 678                 goto retry;
 679         }
 680 
 681         if (btree_node_dirty(b))
 682                 __bch_btree_node_write(b, &cl);
 683         mutex_unlock(&b->write_lock);
 684 
 685         closure_sync(&cl);
 686 
 687         /* wait for any in flight btree write */
 688         down(&b->io_mutex);
 689         up(&b->io_mutex);
 690 
 691         return 0;
 692 out_unlock:
 693         rw_unlock(true, b);
 694         return -ENOMEM;
 695 }
 696 
 697 static unsigned long bch_mca_scan(struct shrinker *shrink,
 698                                   struct shrink_control *sc)
 699 {
 700         struct cache_set *c = container_of(shrink, struct cache_set, shrink);
 701         struct btree *b, *t;
 702         unsigned long i, nr = sc->nr_to_scan;
 703         unsigned long freed = 0;
 704         unsigned int btree_cache_used;
 705 
 706         if (c->shrinker_disabled)
 707                 return SHRINK_STOP;
 708 
 709         if (c->btree_cache_alloc_lock)
 710                 return SHRINK_STOP;
 711 
 712         /* Return -1 if we can't do anything right now */
 713         if (sc->gfp_mask & __GFP_IO)
 714                 mutex_lock(&c->bucket_lock);
 715         else if (!mutex_trylock(&c->bucket_lock))
 716                 return -1;
 717 
 718         /*
 719          * It's _really_ critical that we don't free too many btree nodes - we
 720          * have to always leave ourselves a reserve. The reserve is how we
 721          * guarantee that allocating memory for a new btree node can always
 722          * succeed, so that inserting keys into the btree can always succeed and
 723          * IO can always make forward progress:
 724          */
 725         nr /= c->btree_pages;
 726         if (nr == 0)
 727                 nr = 1;
 728         nr = min_t(unsigned long, nr, mca_can_free(c));
 729 
 730         i = 0;
 731         btree_cache_used = c->btree_cache_used;
 732         list_for_each_entry_safe(b, t, &c->btree_cache_freeable, list) {
 733                 if (nr <= 0)
 734                         goto out;
 735 
 736                 if (++i > 3 &&
 737                     !mca_reap(b, 0, false)) {
 738                         mca_data_free(b);
 739                         rw_unlock(true, b);
 740                         freed++;
 741                 }
 742                 nr--;
 743         }
 744 
 745         for (;  (nr--) && i < btree_cache_used; i++) {
 746                 if (list_empty(&c->btree_cache))
 747                         goto out;
 748 
 749                 b = list_first_entry(&c->btree_cache, struct btree, list);
 750                 list_rotate_left(&c->btree_cache);
 751 
 752                 if (!b->accessed &&
 753                     !mca_reap(b, 0, false)) {
 754                         mca_bucket_free(b);
 755                         mca_data_free(b);
 756                         rw_unlock(true, b);
 757                         freed++;
 758                 } else
 759                         b->accessed = 0;
 760         }
 761 out:
 762         mutex_unlock(&c->bucket_lock);
 763         return freed * c->btree_pages;
 764 }
 765 
 766 static unsigned long bch_mca_count(struct shrinker *shrink,
 767                                    struct shrink_control *sc)
 768 {
 769         struct cache_set *c = container_of(shrink, struct cache_set, shrink);
 770 
 771         if (c->shrinker_disabled)
 772                 return 0;
 773 
 774         if (c->btree_cache_alloc_lock)
 775                 return 0;
 776 
 777         return mca_can_free(c) * c->btree_pages;
 778 }
 779 
 780 void bch_btree_cache_free(struct cache_set *c)
 781 {
 782         struct btree *b;
 783         struct closure cl;
 784 
 785         closure_init_stack(&cl);
 786 
 787         if (c->shrink.list.next)
 788                 unregister_shrinker(&c->shrink);
 789 
 790         mutex_lock(&c->bucket_lock);
 791 
 792 #ifdef CONFIG_BCACHE_DEBUG
 793         if (c->verify_data)
 794                 list_move(&c->verify_data->list, &c->btree_cache);
 795 
 796         free_pages((unsigned long) c->verify_ondisk, ilog2(bucket_pages(c)));
 797 #endif
 798 
 799         list_splice(&c->btree_cache_freeable,
 800                     &c->btree_cache);
 801 
 802         while (!list_empty(&c->btree_cache)) {
 803                 b = list_first_entry(&c->btree_cache, struct btree, list);
 804 
 805                 /*
 806                  * This function is called by cache_set_free(), no I/O
 807                  * request on cache now, it is unnecessary to acquire
 808                  * b->write_lock before clearing BTREE_NODE_dirty anymore.
 809                  */
 810                 if (btree_node_dirty(b)) {
 811                         btree_complete_write(b, btree_current_write(b));
 812                         clear_bit(BTREE_NODE_dirty, &b->flags);
 813                 }
 814                 mca_data_free(b);
 815         }
 816 
 817         while (!list_empty(&c->btree_cache_freed)) {
 818                 b = list_first_entry(&c->btree_cache_freed,
 819                                      struct btree, list);
 820                 list_del(&b->list);
 821                 cancel_delayed_work_sync(&b->work);
 822                 kfree(b);
 823         }
 824 
 825         mutex_unlock(&c->bucket_lock);
 826 }
 827 
 828 int bch_btree_cache_alloc(struct cache_set *c)
 829 {
 830         unsigned int i;
 831 
 832         for (i = 0; i < mca_reserve(c); i++)
 833                 if (!mca_bucket_alloc(c, &ZERO_KEY, GFP_KERNEL))
 834                         return -ENOMEM;
 835 
 836         list_splice_init(&c->btree_cache,
 837                          &c->btree_cache_freeable);
 838 
 839 #ifdef CONFIG_BCACHE_DEBUG
 840         mutex_init(&c->verify_lock);
 841 
 842         c->verify_ondisk = (void *)
 843                 __get_free_pages(GFP_KERNEL, ilog2(bucket_pages(c)));
 844 
 845         c->verify_data = mca_bucket_alloc(c, &ZERO_KEY, GFP_KERNEL);
 846 
 847         if (c->verify_data &&
 848             c->verify_data->keys.set->data)
 849                 list_del_init(&c->verify_data->list);
 850         else
 851                 c->verify_data = NULL;
 852 #endif
 853 
 854         c->shrink.count_objects = bch_mca_count;
 855         c->shrink.scan_objects = bch_mca_scan;
 856         c->shrink.seeks = 4;
 857         c->shrink.batch = c->btree_pages * 2;
 858 
 859         if (register_shrinker(&c->shrink))
 860                 pr_warn("bcache: %s: could not register shrinker",
 861                                 __func__);
 862 
 863         return 0;
 864 }
 865 
 866 /* Btree in memory cache - hash table */
 867 
 868 static struct hlist_head *mca_hash(struct cache_set *c, struct bkey *k)
 869 {
 870         return &c->bucket_hash[hash_32(PTR_HASH(c, k), BUCKET_HASH_BITS)];
 871 }
 872 
 873 static struct btree *mca_find(struct cache_set *c, struct bkey *k)
 874 {
 875         struct btree *b;
 876 
 877         rcu_read_lock();
 878         hlist_for_each_entry_rcu(b, mca_hash(c, k), hash)
 879                 if (PTR_HASH(c, &b->key) == PTR_HASH(c, k))
 880                         goto out;
 881         b = NULL;
 882 out:
 883         rcu_read_unlock();
 884         return b;
 885 }
 886 
 887 static int mca_cannibalize_lock(struct cache_set *c, struct btree_op *op)
 888 {
 889         struct task_struct *old;
 890 
 891         old = cmpxchg(&c->btree_cache_alloc_lock, NULL, current);
 892         if (old && old != current) {
 893                 if (op)
 894                         prepare_to_wait(&c->btree_cache_wait, &op->wait,
 895                                         TASK_UNINTERRUPTIBLE);
 896                 return -EINTR;
 897         }
 898 
 899         return 0;
 900 }
 901 
 902 static struct btree *mca_cannibalize(struct cache_set *c, struct btree_op *op,
 903                                      struct bkey *k)
 904 {
 905         struct btree *b;
 906 
 907         trace_bcache_btree_cache_cannibalize(c);
 908 
 909         if (mca_cannibalize_lock(c, op))
 910                 return ERR_PTR(-EINTR);
 911 
 912         list_for_each_entry_reverse(b, &c->btree_cache, list)
 913                 if (!mca_reap(b, btree_order(k), false))
 914                         return b;
 915 
 916         list_for_each_entry_reverse(b, &c->btree_cache, list)
 917                 if (!mca_reap(b, btree_order(k), true))
 918                         return b;
 919 
 920         WARN(1, "btree cache cannibalize failed\n");
 921         return ERR_PTR(-ENOMEM);
 922 }
 923 
 924 /*
 925  * We can only have one thread cannibalizing other cached btree nodes at a time,
 926  * or we'll deadlock. We use an open coded mutex to ensure that, which a
 927  * cannibalize_bucket() will take. This means every time we unlock the root of
 928  * the btree, we need to release this lock if we have it held.
 929  */
 930 static void bch_cannibalize_unlock(struct cache_set *c)
 931 {
 932         if (c->btree_cache_alloc_lock == current) {
 933                 c->btree_cache_alloc_lock = NULL;
 934                 wake_up(&c->btree_cache_wait);
 935         }
 936 }
 937 
 938 static struct btree *mca_alloc(struct cache_set *c, struct btree_op *op,
 939                                struct bkey *k, int level)
 940 {
 941         struct btree *b;
 942 
 943         BUG_ON(current->bio_list);
 944 
 945         lockdep_assert_held(&c->bucket_lock);
 946 
 947         if (mca_find(c, k))
 948                 return NULL;
 949 
 950         /* btree_free() doesn't free memory; it sticks the node on the end of
 951          * the list. Check if there's any freed nodes there:
 952          */
 953         list_for_each_entry(b, &c->btree_cache_freeable, list)
 954                 if (!mca_reap(b, btree_order(k), false))
 955                         goto out;
 956 
 957         /* We never free struct btree itself, just the memory that holds the on
 958          * disk node. Check the freed list before allocating a new one:
 959          */
 960         list_for_each_entry(b, &c->btree_cache_freed, list)
 961                 if (!mca_reap(b, 0, false)) {
 962                         mca_data_alloc(b, k, __GFP_NOWARN|GFP_NOIO);
 963                         if (!b->keys.set[0].data)
 964                                 goto err;
 965                         else
 966                                 goto out;
 967                 }
 968 
 969         b = mca_bucket_alloc(c, k, __GFP_NOWARN|GFP_NOIO);
 970         if (!b)
 971                 goto err;
 972 
 973         BUG_ON(!down_write_trylock(&b->lock));
 974         if (!b->keys.set->data)
 975                 goto err;
 976 out:
 977         BUG_ON(b->io_mutex.count != 1);
 978 
 979         bkey_copy(&b->key, k);
 980         list_move(&b->list, &c->btree_cache);
 981         hlist_del_init_rcu(&b->hash);
 982         hlist_add_head_rcu(&b->hash, mca_hash(c, k));
 983 
 984         lock_set_subclass(&b->lock.dep_map, level + 1, _THIS_IP_);
 985         b->parent       = (void *) ~0UL;
 986         b->flags        = 0;
 987         b->written      = 0;
 988         b->level        = level;
 989 
 990         if (!b->level)
 991                 bch_btree_keys_init(&b->keys, &bch_extent_keys_ops,
 992                                     &b->c->expensive_debug_checks);
 993         else
 994                 bch_btree_keys_init(&b->keys, &bch_btree_keys_ops,
 995                                     &b->c->expensive_debug_checks);
 996 
 997         return b;
 998 err:
 999         if (b)
1000                 rw_unlock(true, b);
1001 
1002         b = mca_cannibalize(c, op, k);
1003         if (!IS_ERR(b))
1004                 goto out;
1005 
1006         return b;
1007 }
1008 
1009 /*
1010  * bch_btree_node_get - find a btree node in the cache and lock it, reading it
1011  * in from disk if necessary.
1012  *
1013  * If IO is necessary and running under generic_make_request, returns -EAGAIN.
1014  *
1015  * The btree node will have either a read or a write lock held, depending on
1016  * level and op->lock.
1017  */
1018 struct btree *bch_btree_node_get(struct cache_set *c, struct btree_op *op,
1019                                  struct bkey *k, int level, bool write,
1020                                  struct btree *parent)
1021 {
1022         int i = 0;
1023         struct btree *b;
1024 
1025         BUG_ON(level < 0);
1026 retry:
1027         b = mca_find(c, k);
1028 
1029         if (!b) {
1030                 if (current->bio_list)
1031                         return ERR_PTR(-EAGAIN);
1032 
1033                 mutex_lock(&c->bucket_lock);
1034                 b = mca_alloc(c, op, k, level);
1035                 mutex_unlock(&c->bucket_lock);
1036 
1037                 if (!b)
1038                         goto retry;
1039                 if (IS_ERR(b))
1040                         return b;
1041 
1042                 bch_btree_node_read(b);
1043 
1044                 if (!write)
1045                         downgrade_write(&b->lock);
1046         } else {
1047                 rw_lock(write, b, level);
1048                 if (PTR_HASH(c, &b->key) != PTR_HASH(c, k)) {
1049                         rw_unlock(write, b);
1050                         goto retry;
1051                 }
1052                 BUG_ON(b->level != level);
1053         }
1054 
1055         if (btree_node_io_error(b)) {
1056                 rw_unlock(write, b);
1057                 return ERR_PTR(-EIO);
1058         }
1059 
1060         BUG_ON(!b->written);
1061 
1062         b->parent = parent;
1063         b->accessed = 1;
1064 
1065         for (; i <= b->keys.nsets && b->keys.set[i].size; i++) {
1066                 prefetch(b->keys.set[i].tree);
1067                 prefetch(b->keys.set[i].data);
1068         }
1069 
1070         for (; i <= b->keys.nsets; i++)
1071                 prefetch(b->keys.set[i].data);
1072 
1073         return b;
1074 }
1075 
1076 static void btree_node_prefetch(struct btree *parent, struct bkey *k)
1077 {
1078         struct btree *b;
1079 
1080         mutex_lock(&parent->c->bucket_lock);
1081         b = mca_alloc(parent->c, NULL, k, parent->level - 1);
1082         mutex_unlock(&parent->c->bucket_lock);
1083 
1084         if (!IS_ERR_OR_NULL(b)) {
1085                 b->parent = parent;
1086                 bch_btree_node_read(b);
1087                 rw_unlock(true, b);
1088         }
1089 }
1090 
1091 /* Btree alloc */
1092 
1093 static void btree_node_free(struct btree *b)
1094 {
1095         trace_bcache_btree_node_free(b);
1096 
1097         BUG_ON(b == b->c->root);
1098 
1099 retry:
1100         mutex_lock(&b->write_lock);
1101         /*
1102          * If the btree node is selected and flushing in btree_flush_write(),
1103          * delay and retry until the BTREE_NODE_journal_flush bit cleared,
1104          * then it is safe to free the btree node here. Otherwise this btree
1105          * node will be in race condition.
1106          */
1107         if (btree_node_journal_flush(b)) {
1108                 mutex_unlock(&b->write_lock);
1109                 pr_debug("bnode %p journal_flush set, retry", b);
1110                 udelay(1);
1111                 goto retry;
1112         }
1113 
1114         if (btree_node_dirty(b)) {
1115                 btree_complete_write(b, btree_current_write(b));
1116                 clear_bit(BTREE_NODE_dirty, &b->flags);
1117         }
1118 
1119         mutex_unlock(&b->write_lock);
1120 
1121         cancel_delayed_work(&b->work);
1122 
1123         mutex_lock(&b->c->bucket_lock);
1124         bch_bucket_free(b->c, &b->key);
1125         mca_bucket_free(b);
1126         mutex_unlock(&b->c->bucket_lock);
1127 }
1128 
1129 struct btree *__bch_btree_node_alloc(struct cache_set *c, struct btree_op *op,
1130                                      int level, bool wait,
1131                                      struct btree *parent)
1132 {
1133         BKEY_PADDED(key) k;
1134         struct btree *b = ERR_PTR(-EAGAIN);
1135 
1136         mutex_lock(&c->bucket_lock);
1137 retry:
1138         if (__bch_bucket_alloc_set(c, RESERVE_BTREE, &k.key, 1, wait))
1139                 goto err;
1140 
1141         bkey_put(c, &k.key);
1142         SET_KEY_SIZE(&k.key, c->btree_pages * PAGE_SECTORS);
1143 
1144         b = mca_alloc(c, op, &k.key, level);
1145         if (IS_ERR(b))
1146                 goto err_free;
1147 
1148         if (!b) {
1149                 cache_bug(c,
1150                         "Tried to allocate bucket that was in btree cache");
1151                 goto retry;
1152         }
1153 
1154         b->accessed = 1;
1155         b->parent = parent;
1156         bch_bset_init_next(&b->keys, b->keys.set->data, bset_magic(&b->c->sb));
1157 
1158         mutex_unlock(&c->bucket_lock);
1159 
1160         trace_bcache_btree_node_alloc(b);
1161         return b;
1162 err_free:
1163         bch_bucket_free(c, &k.key);
1164 err:
1165         mutex_unlock(&c->bucket_lock);
1166 
1167         trace_bcache_btree_node_alloc_fail(c);
1168         return b;
1169 }
1170 
1171 static struct btree *bch_btree_node_alloc(struct cache_set *c,
1172                                           struct btree_op *op, int level,
1173                                           struct btree *parent)
1174 {
1175         return __bch_btree_node_alloc(c, op, level, op != NULL, parent);
1176 }
1177 
1178 static struct btree *btree_node_alloc_replacement(struct btree *b,
1179                                                   struct btree_op *op)
1180 {
1181         struct btree *n = bch_btree_node_alloc(b->c, op, b->level, b->parent);
1182 
1183         if (!IS_ERR_OR_NULL(n)) {
1184                 mutex_lock(&n->write_lock);
1185                 bch_btree_sort_into(&b->keys, &n->keys, &b->c->sort);
1186                 bkey_copy_key(&n->key, &b->key);
1187                 mutex_unlock(&n->write_lock);
1188         }
1189 
1190         return n;
1191 }
1192 
1193 static void make_btree_freeing_key(struct btree *b, struct bkey *k)
1194 {
1195         unsigned int i;
1196 
1197         mutex_lock(&b->c->bucket_lock);
1198 
1199         atomic_inc(&b->c->prio_blocked);
1200 
1201         bkey_copy(k, &b->key);
1202         bkey_copy_key(k, &ZERO_KEY);
1203 
1204         for (i = 0; i < KEY_PTRS(k); i++)
1205                 SET_PTR_GEN(k, i,
1206                             bch_inc_gen(PTR_CACHE(b->c, &b->key, i),
1207                                         PTR_BUCKET(b->c, &b->key, i)));
1208 
1209         mutex_unlock(&b->c->bucket_lock);
1210 }
1211 
1212 static int btree_check_reserve(struct btree *b, struct btree_op *op)
1213 {
1214         struct cache_set *c = b->c;
1215         struct cache *ca;
1216         unsigned int i, reserve = (c->root->level - b->level) * 2 + 1;
1217 
1218         mutex_lock(&c->bucket_lock);
1219 
1220         for_each_cache(ca, c, i)
1221                 if (fifo_used(&ca->free[RESERVE_BTREE]) < reserve) {
1222                         if (op)
1223                                 prepare_to_wait(&c->btree_cache_wait, &op->wait,
1224                                                 TASK_UNINTERRUPTIBLE);
1225                         mutex_unlock(&c->bucket_lock);
1226                         return -EINTR;
1227                 }
1228 
1229         mutex_unlock(&c->bucket_lock);
1230 
1231         return mca_cannibalize_lock(b->c, op);
1232 }
1233 
1234 /* Garbage collection */
1235 
1236 static uint8_t __bch_btree_mark_key(struct cache_set *c, int level,
1237                                     struct bkey *k)
1238 {
1239         uint8_t stale = 0;
1240         unsigned int i;
1241         struct bucket *g;
1242 
1243         /*
1244          * ptr_invalid() can't return true for the keys that mark btree nodes as
1245          * freed, but since ptr_bad() returns true we'll never actually use them
1246          * for anything and thus we don't want mark their pointers here
1247          */
1248         if (!bkey_cmp(k, &ZERO_KEY))
1249                 return stale;
1250 
1251         for (i = 0; i < KEY_PTRS(k); i++) {
1252                 if (!ptr_available(c, k, i))
1253                         continue;
1254 
1255                 g = PTR_BUCKET(c, k, i);
1256 
1257                 if (gen_after(g->last_gc, PTR_GEN(k, i)))
1258                         g->last_gc = PTR_GEN(k, i);
1259 
1260                 if (ptr_stale(c, k, i)) {
1261                         stale = max(stale, ptr_stale(c, k, i));
1262                         continue;
1263                 }
1264 
1265                 cache_bug_on(GC_MARK(g) &&
1266                              (GC_MARK(g) == GC_MARK_METADATA) != (level != 0),
1267                              c, "inconsistent ptrs: mark = %llu, level = %i",
1268                              GC_MARK(g), level);
1269 
1270                 if (level)
1271                         SET_GC_MARK(g, GC_MARK_METADATA);
1272                 else if (KEY_DIRTY(k))
1273                         SET_GC_MARK(g, GC_MARK_DIRTY);
1274                 else if (!GC_MARK(g))
1275                         SET_GC_MARK(g, GC_MARK_RECLAIMABLE);
1276 
1277                 /* guard against overflow */
1278                 SET_GC_SECTORS_USED(g, min_t(unsigned int,
1279                                              GC_SECTORS_USED(g) + KEY_SIZE(k),
1280                                              MAX_GC_SECTORS_USED));
1281 
1282                 BUG_ON(!GC_SECTORS_USED(g));
1283         }
1284 
1285         return stale;
1286 }
1287 
1288 #define btree_mark_key(b, k)    __bch_btree_mark_key(b->c, b->level, k)
1289 
1290 void bch_initial_mark_key(struct cache_set *c, int level, struct bkey *k)
1291 {
1292         unsigned int i;
1293 
1294         for (i = 0; i < KEY_PTRS(k); i++)
1295                 if (ptr_available(c, k, i) &&
1296                     !ptr_stale(c, k, i)) {
1297                         struct bucket *b = PTR_BUCKET(c, k, i);
1298 
1299                         b->gen = PTR_GEN(k, i);
1300 
1301                         if (level && bkey_cmp(k, &ZERO_KEY))
1302                                 b->prio = BTREE_PRIO;
1303                         else if (!level && b->prio == BTREE_PRIO)
1304                                 b->prio = INITIAL_PRIO;
1305                 }
1306 
1307         __bch_btree_mark_key(c, level, k);
1308 }
1309 
1310 void bch_update_bucket_in_use(struct cache_set *c, struct gc_stat *stats)
1311 {
1312         stats->in_use = (c->nbuckets - c->avail_nbuckets) * 100 / c->nbuckets;
1313 }
1314 
1315 static bool btree_gc_mark_node(struct btree *b, struct gc_stat *gc)
1316 {
1317         uint8_t stale = 0;
1318         unsigned int keys = 0, good_keys = 0;
1319         struct bkey *k;
1320         struct btree_iter iter;
1321         struct bset_tree *t;
1322 
1323         gc->nodes++;
1324 
1325         for_each_key_filter(&b->keys, k, &iter, bch_ptr_invalid) {
1326                 stale = max(stale, btree_mark_key(b, k));
1327                 keys++;
1328 
1329                 if (bch_ptr_bad(&b->keys, k))
1330                         continue;
1331 
1332                 gc->key_bytes += bkey_u64s(k);
1333                 gc->nkeys++;
1334                 good_keys++;
1335 
1336                 gc->data += KEY_SIZE(k);
1337         }
1338 
1339         for (t = b->keys.set; t <= &b->keys.set[b->keys.nsets]; t++)
1340                 btree_bug_on(t->size &&
1341                              bset_written(&b->keys, t) &&
1342                              bkey_cmp(&b->key, &t->end) < 0,
1343                              b, "found short btree key in gc");
1344 
1345         if (b->c->gc_always_rewrite)
1346                 return true;
1347 
1348         if (stale > 10)
1349                 return true;
1350 
1351         if ((keys - good_keys) * 2 > keys)
1352                 return true;
1353 
1354         return false;
1355 }
1356 
1357 #define GC_MERGE_NODES  4U
1358 
1359 struct gc_merge_info {
1360         struct btree    *b;
1361         unsigned int    keys;
1362 };
1363 
1364 static int bch_btree_insert_node(struct btree *b, struct btree_op *op,
1365                                  struct keylist *insert_keys,
1366                                  atomic_t *journal_ref,
1367                                  struct bkey *replace_key);
1368 
1369 static int btree_gc_coalesce(struct btree *b, struct btree_op *op,
1370                              struct gc_stat *gc, struct gc_merge_info *r)
1371 {
1372         unsigned int i, nodes = 0, keys = 0, blocks;
1373         struct btree *new_nodes[GC_MERGE_NODES];
1374         struct keylist keylist;
1375         struct closure cl;
1376         struct bkey *k;
1377 
1378         bch_keylist_init(&keylist);
1379 
1380         if (btree_check_reserve(b, NULL))
1381                 return 0;
1382 
1383         memset(new_nodes, 0, sizeof(new_nodes));
1384         closure_init_stack(&cl);
1385 
1386         while (nodes < GC_MERGE_NODES && !IS_ERR_OR_NULL(r[nodes].b))
1387                 keys += r[nodes++].keys;
1388 
1389         blocks = btree_default_blocks(b->c) * 2 / 3;
1390 
1391         if (nodes < 2 ||
1392             __set_blocks(b->keys.set[0].data, keys,
1393                          block_bytes(b->c)) > blocks * (nodes - 1))
1394                 return 0;
1395 
1396         for (i = 0; i < nodes; i++) {
1397                 new_nodes[i] = btree_node_alloc_replacement(r[i].b, NULL);
1398                 if (IS_ERR_OR_NULL(new_nodes[i]))
1399                         goto out_nocoalesce;
1400         }
1401 
1402         /*
1403          * We have to check the reserve here, after we've allocated our new
1404          * nodes, to make sure the insert below will succeed - we also check
1405          * before as an optimization to potentially avoid a bunch of expensive
1406          * allocs/sorts
1407          */
1408         if (btree_check_reserve(b, NULL))
1409                 goto out_nocoalesce;
1410 
1411         for (i = 0; i < nodes; i++)
1412                 mutex_lock(&new_nodes[i]->write_lock);
1413 
1414         for (i = nodes - 1; i > 0; --i) {
1415                 struct bset *n1 = btree_bset_first(new_nodes[i]);
1416                 struct bset *n2 = btree_bset_first(new_nodes[i - 1]);
1417                 struct bkey *k, *last = NULL;
1418 
1419                 keys = 0;
1420 
1421                 if (i > 1) {
1422                         for (k = n2->start;
1423                              k < bset_bkey_last(n2);
1424                              k = bkey_next(k)) {
1425                                 if (__set_blocks(n1, n1->keys + keys +
1426                                                  bkey_u64s(k),
1427                                                  block_bytes(b->c)) > blocks)
1428                                         break;
1429 
1430                                 last = k;
1431                                 keys += bkey_u64s(k);
1432                         }
1433                 } else {
1434                         /*
1435                          * Last node we're not getting rid of - we're getting
1436                          * rid of the node at r[0]. Have to try and fit all of
1437                          * the remaining keys into this node; we can't ensure
1438                          * they will always fit due to rounding and variable
1439                          * length keys (shouldn't be possible in practice,
1440                          * though)
1441                          */
1442                         if (__set_blocks(n1, n1->keys + n2->keys,
1443                                          block_bytes(b->c)) >
1444                             btree_blocks(new_nodes[i]))
1445                                 goto out_nocoalesce;
1446 
1447                         keys = n2->keys;
1448                         /* Take the key of the node we're getting rid of */
1449                         last = &r->b->key;
1450                 }
1451 
1452                 BUG_ON(__set_blocks(n1, n1->keys + keys, block_bytes(b->c)) >
1453                        btree_blocks(new_nodes[i]));
1454 
1455                 if (last)
1456                         bkey_copy_key(&new_nodes[i]->key, last);
1457 
1458                 memcpy(bset_bkey_last(n1),
1459                        n2->start,
1460                        (void *) bset_bkey_idx(n2, keys) - (void *) n2->start);
1461 
1462                 n1->keys += keys;
1463                 r[i].keys = n1->keys;
1464 
1465                 memmove(n2->start,
1466                         bset_bkey_idx(n2, keys),
1467                         (void *) bset_bkey_last(n2) -
1468                         (void *) bset_bkey_idx(n2, keys));
1469 
1470                 n2->keys -= keys;
1471 
1472                 if (__bch_keylist_realloc(&keylist,
1473                                           bkey_u64s(&new_nodes[i]->key)))
1474                         goto out_nocoalesce;
1475 
1476                 bch_btree_node_write(new_nodes[i], &cl);
1477                 bch_keylist_add(&keylist, &new_nodes[i]->key);
1478         }
1479 
1480         for (i = 0; i < nodes; i++)
1481                 mutex_unlock(&new_nodes[i]->write_lock);
1482 
1483         closure_sync(&cl);
1484 
1485         /* We emptied out this node */
1486         BUG_ON(btree_bset_first(new_nodes[0])->keys);
1487         btree_node_free(new_nodes[0]);
1488         rw_unlock(true, new_nodes[0]);
1489         new_nodes[0] = NULL;
1490 
1491         for (i = 0; i < nodes; i++) {
1492                 if (__bch_keylist_realloc(&keylist, bkey_u64s(&r[i].b->key)))
1493                         goto out_nocoalesce;
1494 
1495                 make_btree_freeing_key(r[i].b, keylist.top);
1496                 bch_keylist_push(&keylist);
1497         }
1498 
1499         bch_btree_insert_node(b, op, &keylist, NULL, NULL);
1500         BUG_ON(!bch_keylist_empty(&keylist));
1501 
1502         for (i = 0; i < nodes; i++) {
1503                 btree_node_free(r[i].b);
1504                 rw_unlock(true, r[i].b);
1505 
1506                 r[i].b = new_nodes[i];
1507         }
1508 
1509         memmove(r, r + 1, sizeof(r[0]) * (nodes - 1));
1510         r[nodes - 1].b = ERR_PTR(-EINTR);
1511 
1512         trace_bcache_btree_gc_coalesce(nodes);
1513         gc->nodes--;
1514 
1515         bch_keylist_free(&keylist);
1516 
1517         /* Invalidated our iterator */
1518         return -EINTR;
1519 
1520 out_nocoalesce:
1521         closure_sync(&cl);
1522 
1523         while ((k = bch_keylist_pop(&keylist)))
1524                 if (!bkey_cmp(k, &ZERO_KEY))
1525                         atomic_dec(&b->c->prio_blocked);
1526         bch_keylist_free(&keylist);
1527 
1528         for (i = 0; i < nodes; i++)
1529                 if (!IS_ERR_OR_NULL(new_nodes[i])) {
1530                         btree_node_free(new_nodes[i]);
1531                         rw_unlock(true, new_nodes[i]);
1532                 }
1533         return 0;
1534 }
1535 
1536 static int btree_gc_rewrite_node(struct btree *b, struct btree_op *op,
1537                                  struct btree *replace)
1538 {
1539         struct keylist keys;
1540         struct btree *n;
1541 
1542         if (btree_check_reserve(b, NULL))
1543                 return 0;
1544 
1545         n = btree_node_alloc_replacement(replace, NULL);
1546 
1547         /* recheck reserve after allocating replacement node */
1548         if (btree_check_reserve(b, NULL)) {
1549                 btree_node_free(n);
1550                 rw_unlock(true, n);
1551                 return 0;
1552         }
1553 
1554         bch_btree_node_write_sync(n);
1555 
1556         bch_keylist_init(&keys);
1557         bch_keylist_add(&keys, &n->key);
1558 
1559         make_btree_freeing_key(replace, keys.top);
1560         bch_keylist_push(&keys);
1561 
1562         bch_btree_insert_node(b, op, &keys, NULL, NULL);
1563         BUG_ON(!bch_keylist_empty(&keys));
1564 
1565         btree_node_free(replace);
1566         rw_unlock(true, n);
1567 
1568         /* Invalidated our iterator */
1569         return -EINTR;
1570 }
1571 
1572 static unsigned int btree_gc_count_keys(struct btree *b)
1573 {
1574         struct bkey *k;
1575         struct btree_iter iter;
1576         unsigned int ret = 0;
1577 
1578         for_each_key_filter(&b->keys, k, &iter, bch_ptr_bad)
1579                 ret += bkey_u64s(k);
1580 
1581         return ret;
1582 }
1583 
1584 static size_t btree_gc_min_nodes(struct cache_set *c)
1585 {
1586         size_t min_nodes;
1587 
1588         /*
1589          * Since incremental GC would stop 100ms when front
1590          * side I/O comes, so when there are many btree nodes,
1591          * if GC only processes constant (100) nodes each time,
1592          * GC would last a long time, and the front side I/Os
1593          * would run out of the buckets (since no new bucket
1594          * can be allocated during GC), and be blocked again.
1595          * So GC should not process constant nodes, but varied
1596          * nodes according to the number of btree nodes, which
1597          * realized by dividing GC into constant(100) times,
1598          * so when there are many btree nodes, GC can process
1599          * more nodes each time, otherwise, GC will process less
1600          * nodes each time (but no less than MIN_GC_NODES)
1601          */
1602         min_nodes = c->gc_stats.nodes / MAX_GC_TIMES;
1603         if (min_nodes < MIN_GC_NODES)
1604                 min_nodes = MIN_GC_NODES;
1605 
1606         return min_nodes;
1607 }
1608 
1609 
1610 static int btree_gc_recurse(struct btree *b, struct btree_op *op,
1611                             struct closure *writes, struct gc_stat *gc)
1612 {
1613         int ret = 0;
1614         bool should_rewrite;
1615         struct bkey *k;
1616         struct btree_iter iter;
1617         struct gc_merge_info r[GC_MERGE_NODES];
1618         struct gc_merge_info *i, *last = r + ARRAY_SIZE(r) - 1;
1619 
1620         bch_btree_iter_init(&b->keys, &iter, &b->c->gc_done);
1621 
1622         for (i = r; i < r + ARRAY_SIZE(r); i++)
1623                 i->b = ERR_PTR(-EINTR);
1624 
1625         while (1) {
1626                 k = bch_btree_iter_next_filter(&iter, &b->keys, bch_ptr_bad);
1627                 if (k) {
1628                         r->b = bch_btree_node_get(b->c, op, k, b->level - 1,
1629                                                   true, b);
1630                         if (IS_ERR(r->b)) {
1631                                 ret = PTR_ERR(r->b);
1632                                 break;
1633                         }
1634 
1635                         r->keys = btree_gc_count_keys(r->b);
1636 
1637                         ret = btree_gc_coalesce(b, op, gc, r);
1638                         if (ret)
1639                                 break;
1640                 }
1641 
1642                 if (!last->b)
1643                         break;
1644 
1645                 if (!IS_ERR(last->b)) {
1646                         should_rewrite = btree_gc_mark_node(last->b, gc);
1647                         if (should_rewrite) {
1648                                 ret = btree_gc_rewrite_node(b, op, last->b);
1649                                 if (ret)
1650                                         break;
1651                         }
1652 
1653                         if (last->b->level) {
1654                                 ret = btree_gc_recurse(last->b, op, writes, gc);
1655                                 if (ret)
1656                                         break;
1657                         }
1658 
1659                         bkey_copy_key(&b->c->gc_done, &last->b->key);
1660 
1661                         /*
1662                          * Must flush leaf nodes before gc ends, since replace
1663                          * operations aren't journalled
1664                          */
1665                         mutex_lock(&last->b->write_lock);
1666                         if (btree_node_dirty(last->b))
1667                                 bch_btree_node_write(last->b, writes);
1668                         mutex_unlock(&last->b->write_lock);
1669                         rw_unlock(true, last->b);
1670                 }
1671 
1672                 memmove(r + 1, r, sizeof(r[0]) * (GC_MERGE_NODES - 1));
1673                 r->b = NULL;
1674 
1675                 if (atomic_read(&b->c->search_inflight) &&
1676                     gc->nodes >= gc->nodes_pre + btree_gc_min_nodes(b->c)) {
1677                         gc->nodes_pre =  gc->nodes;
1678                         ret = -EAGAIN;
1679                         break;
1680                 }
1681 
1682                 if (need_resched()) {
1683                         ret = -EAGAIN;
1684                         break;
1685                 }
1686         }
1687 
1688         for (i = r; i < r + ARRAY_SIZE(r); i++)
1689                 if (!IS_ERR_OR_NULL(i->b)) {
1690                         mutex_lock(&i->b->write_lock);
1691                         if (btree_node_dirty(i->b))
1692                                 bch_btree_node_write(i->b, writes);
1693                         mutex_unlock(&i->b->write_lock);
1694                         rw_unlock(true, i->b);
1695                 }
1696 
1697         return ret;
1698 }
1699 
1700 static int bch_btree_gc_root(struct btree *b, struct btree_op *op,
1701                              struct closure *writes, struct gc_stat *gc)
1702 {
1703         struct btree *n = NULL;
1704         int ret = 0;
1705         bool should_rewrite;
1706 
1707         should_rewrite = btree_gc_mark_node(b, gc);
1708         if (should_rewrite) {
1709                 n = btree_node_alloc_replacement(b, NULL);
1710 
1711                 if (!IS_ERR_OR_NULL(n)) {
1712                         bch_btree_node_write_sync(n);
1713 
1714                         bch_btree_set_root(n);
1715                         btree_node_free(b);
1716                         rw_unlock(true, n);
1717 
1718                         return -EINTR;
1719                 }
1720         }
1721 
1722         __bch_btree_mark_key(b->c, b->level + 1, &b->key);
1723 
1724         if (b->level) {
1725                 ret = btree_gc_recurse(b, op, writes, gc);
1726                 if (ret)
1727                         return ret;
1728         }
1729 
1730         bkey_copy_key(&b->c->gc_done, &b->key);
1731 
1732         return ret;
1733 }
1734 
1735 static void btree_gc_start(struct cache_set *c)
1736 {
1737         struct cache *ca;
1738         struct bucket *b;
1739         unsigned int i;
1740 
1741         if (!c->gc_mark_valid)
1742                 return;
1743 
1744         mutex_lock(&c->bucket_lock);
1745 
1746         c->gc_mark_valid = 0;
1747         c->gc_done = ZERO_KEY;
1748 
1749         for_each_cache(ca, c, i)
1750                 for_each_bucket(b, ca) {
1751                         b->last_gc = b->gen;
1752                         if (!atomic_read(&b->pin)) {
1753                                 SET_GC_MARK(b, 0);
1754                                 SET_GC_SECTORS_USED(b, 0);
1755                         }
1756                 }
1757 
1758         mutex_unlock(&c->bucket_lock);
1759 }
1760 
1761 static void bch_btree_gc_finish(struct cache_set *c)
1762 {
1763         struct bucket *b;
1764         struct cache *ca;
1765         unsigned int i;
1766 
1767         mutex_lock(&c->bucket_lock);
1768 
1769         set_gc_sectors(c);
1770         c->gc_mark_valid = 1;
1771         c->need_gc      = 0;
1772 
1773         for (i = 0; i < KEY_PTRS(&c->uuid_bucket); i++)
1774                 SET_GC_MARK(PTR_BUCKET(c, &c->uuid_bucket, i),
1775                             GC_MARK_METADATA);
1776 
1777         /* don't reclaim buckets to which writeback keys point */
1778         rcu_read_lock();
1779         for (i = 0; i < c->devices_max_used; i++) {
1780                 struct bcache_device *d = c->devices[i];
1781                 struct cached_dev *dc;
1782                 struct keybuf_key *w, *n;
1783                 unsigned int j;
1784 
1785                 if (!d || UUID_FLASH_ONLY(&c->uuids[i]))
1786                         continue;
1787                 dc = container_of(d, struct cached_dev, disk);
1788 
1789                 spin_lock(&dc->writeback_keys.lock);
1790                 rbtree_postorder_for_each_entry_safe(w, n,
1791                                         &dc->writeback_keys.keys, node)
1792                         for (j = 0; j < KEY_PTRS(&w->key); j++)
1793                                 SET_GC_MARK(PTR_BUCKET(c, &w->key, j),
1794                                             GC_MARK_DIRTY);
1795                 spin_unlock(&dc->writeback_keys.lock);
1796         }
1797         rcu_read_unlock();
1798 
1799         c->avail_nbuckets = 0;
1800         for_each_cache(ca, c, i) {
1801                 uint64_t *i;
1802 
1803                 ca->invalidate_needs_gc = 0;
1804 
1805                 for (i = ca->sb.d; i < ca->sb.d + ca->sb.keys; i++)
1806                         SET_GC_MARK(ca->buckets + *i, GC_MARK_METADATA);
1807 
1808                 for (i = ca->prio_buckets;
1809                      i < ca->prio_buckets + prio_buckets(ca) * 2; i++)
1810                         SET_GC_MARK(ca->buckets + *i, GC_MARK_METADATA);
1811 
1812                 for_each_bucket(b, ca) {
1813                         c->need_gc      = max(c->need_gc, bucket_gc_gen(b));
1814 
1815                         if (atomic_read(&b->pin))
1816                                 continue;
1817 
1818                         BUG_ON(!GC_MARK(b) && GC_SECTORS_USED(b));
1819 
1820                         if (!GC_MARK(b) || GC_MARK(b) == GC_MARK_RECLAIMABLE)
1821                                 c->avail_nbuckets++;
1822                 }
1823         }
1824 
1825         mutex_unlock(&c->bucket_lock);
1826 }
1827 
1828 static void bch_btree_gc(struct cache_set *c)
1829 {
1830         int ret;
1831         struct gc_stat stats;
1832         struct closure writes;
1833         struct btree_op op;
1834         uint64_t start_time = local_clock();
1835 
1836         trace_bcache_gc_start(c);
1837 
1838         memset(&stats, 0, sizeof(struct gc_stat));
1839         closure_init_stack(&writes);
1840         bch_btree_op_init(&op, SHRT_MAX);
1841 
1842         btree_gc_start(c);
1843 
1844         /* if CACHE_SET_IO_DISABLE set, gc thread should stop too */
1845         do {
1846                 ret = btree_root(gc_root, c, &op, &writes, &stats);
1847                 closure_sync(&writes);
1848                 cond_resched();
1849 
1850                 if (ret == -EAGAIN)
1851                         schedule_timeout_interruptible(msecs_to_jiffies
1852                                                        (GC_SLEEP_MS));
1853                 else if (ret)
1854                         pr_warn("gc failed!");
1855         } while (ret && !test_bit(CACHE_SET_IO_DISABLE, &c->flags));
1856 
1857         bch_btree_gc_finish(c);
1858         wake_up_allocators(c);
1859 
1860         bch_time_stats_update(&c->btree_gc_time, start_time);
1861 
1862         stats.key_bytes *= sizeof(uint64_t);
1863         stats.data      <<= 9;
1864         bch_update_bucket_in_use(c, &stats);
1865         memcpy(&c->gc_stats, &stats, sizeof(struct gc_stat));
1866 
1867         trace_bcache_gc_end(c);
1868 
1869         bch_moving_gc(c);
1870 }
1871 
1872 static bool gc_should_run(struct cache_set *c)
1873 {
1874         struct cache *ca;
1875         unsigned int i;
1876 
1877         for_each_cache(ca, c, i)
1878                 if (ca->invalidate_needs_gc)
1879                         return true;
1880 
1881         if (atomic_read(&c->sectors_to_gc) < 0)
1882                 return true;
1883 
1884         return false;
1885 }
1886 
1887 static int bch_gc_thread(void *arg)
1888 {
1889         struct cache_set *c = arg;
1890 
1891         while (1) {
1892                 wait_event_interruptible(c->gc_wait,
1893                            kthread_should_stop() ||
1894                            test_bit(CACHE_SET_IO_DISABLE, &c->flags) ||
1895                            gc_should_run(c));
1896 
1897                 if (kthread_should_stop() ||
1898                     test_bit(CACHE_SET_IO_DISABLE, &c->flags))
1899                         break;
1900 
1901                 set_gc_sectors(c);
1902                 bch_btree_gc(c);
1903         }
1904 
1905         wait_for_kthread_stop();
1906         return 0;
1907 }
1908 
1909 int bch_gc_thread_start(struct cache_set *c)
1910 {
1911         c->gc_thread = kthread_run(bch_gc_thread, c, "bcache_gc");
1912         return PTR_ERR_OR_ZERO(c->gc_thread);
1913 }
1914 
1915 /* Initial partial gc */
1916 
1917 static int bch_btree_check_recurse(struct btree *b, struct btree_op *op)
1918 {
1919         int ret = 0;
1920         struct bkey *k, *p = NULL;
1921         struct btree_iter iter;
1922 
1923         for_each_key_filter(&b->keys, k, &iter, bch_ptr_invalid)
1924                 bch_initial_mark_key(b->c, b->level, k);
1925 
1926         bch_initial_mark_key(b->c, b->level + 1, &b->key);
1927 
1928         if (b->level) {
1929                 bch_btree_iter_init(&b->keys, &iter, NULL);
1930 
1931                 do {
1932                         k = bch_btree_iter_next_filter(&iter, &b->keys,
1933                                                        bch_ptr_bad);
1934                         if (k) {
1935                                 btree_node_prefetch(b, k);
1936                                 /*
1937                                  * initiallize c->gc_stats.nodes
1938                                  * for incremental GC
1939                                  */
1940                                 b->c->gc_stats.nodes++;
1941                         }
1942 
1943                         if (p)
1944                                 ret = btree(check_recurse, p, b, op);
1945 
1946                         p = k;
1947                 } while (p && !ret);
1948         }
1949 
1950         return ret;
1951 }
1952 
1953 int bch_btree_check(struct cache_set *c)
1954 {
1955         struct btree_op op;
1956 
1957         bch_btree_op_init(&op, SHRT_MAX);
1958 
1959         return btree_root(check_recurse, c, &op);
1960 }
1961 
1962 void bch_initial_gc_finish(struct cache_set *c)
1963 {
1964         struct cache *ca;
1965         struct bucket *b;
1966         unsigned int i;
1967 
1968         bch_btree_gc_finish(c);
1969 
1970         mutex_lock(&c->bucket_lock);
1971 
1972         /*
1973          * We need to put some unused buckets directly on the prio freelist in
1974          * order to get the allocator thread started - it needs freed buckets in
1975          * order to rewrite the prios and gens, and it needs to rewrite prios
1976          * and gens in order to free buckets.
1977          *
1978          * This is only safe for buckets that have no live data in them, which
1979          * there should always be some of.
1980          */
1981         for_each_cache(ca, c, i) {
1982                 for_each_bucket(b, ca) {
1983                         if (fifo_full(&ca->free[RESERVE_PRIO]) &&
1984                             fifo_full(&ca->free[RESERVE_BTREE]))
1985                                 break;
1986 
1987                         if (bch_can_invalidate_bucket(ca, b) &&
1988                             !GC_MARK(b)) {
1989                                 __bch_invalidate_one_bucket(ca, b);
1990                                 if (!fifo_push(&ca->free[RESERVE_PRIO],
1991                                    b - ca->buckets))
1992                                         fifo_push(&ca->free[RESERVE_BTREE],
1993                                                   b - ca->buckets);
1994                         }
1995                 }
1996         }
1997 
1998         mutex_unlock(&c->bucket_lock);
1999 }
2000 
2001 /* Btree insertion */
2002 
2003 static bool btree_insert_key(struct btree *b, struct bkey *k,
2004                              struct bkey *replace_key)
2005 {
2006         unsigned int status;
2007 
2008         BUG_ON(bkey_cmp(k, &b->key) > 0);
2009 
2010         status = bch_btree_insert_key(&b->keys, k, replace_key);
2011         if (status != BTREE_INSERT_STATUS_NO_INSERT) {
2012                 bch_check_keys(&b->keys, "%u for %s", status,
2013                                replace_key ? "replace" : "insert");
2014 
2015                 trace_bcache_btree_insert_key(b, k, replace_key != NULL,
2016                                               status);
2017                 return true;
2018         } else
2019                 return false;
2020 }
2021 
2022 static size_t insert_u64s_remaining(struct btree *b)
2023 {
2024         long ret = bch_btree_keys_u64s_remaining(&b->keys);
2025 
2026         /*
2027          * Might land in the middle of an existing extent and have to split it
2028          */
2029         if (b->keys.ops->is_extents)
2030                 ret -= KEY_MAX_U64S;
2031 
2032         return max(ret, 0L);
2033 }
2034 
2035 static bool bch_btree_insert_keys(struct btree *b, struct btree_op *op,
2036                                   struct keylist *insert_keys,
2037                                   struct bkey *replace_key)
2038 {
2039         bool ret = false;
2040         int oldsize = bch_count_data(&b->keys);
2041 
2042         while (!bch_keylist_empty(insert_keys)) {
2043                 struct bkey *k = insert_keys->keys;
2044 
2045                 if (bkey_u64s(k) > insert_u64s_remaining(b))
2046                         break;
2047 
2048                 if (bkey_cmp(k, &b->key) <= 0) {
2049                         if (!b->level)
2050                                 bkey_put(b->c, k);
2051 
2052                         ret |= btree_insert_key(b, k, replace_key);
2053                         bch_keylist_pop_front(insert_keys);
2054                 } else if (bkey_cmp(&START_KEY(k), &b->key) < 0) {
2055                         BKEY_PADDED(key) temp;
2056                         bkey_copy(&temp.key, insert_keys->keys);
2057 
2058                         bch_cut_back(&b->key, &temp.key);
2059                         bch_cut_front(&b->key, insert_keys->keys);
2060 
2061                         ret |= btree_insert_key(b, &temp.key, replace_key);
2062                         break;
2063                 } else {
2064                         break;
2065                 }
2066         }
2067 
2068         if (!ret)
2069                 op->insert_collision = true;
2070 
2071         BUG_ON(!bch_keylist_empty(insert_keys) && b->level);
2072 
2073         BUG_ON(bch_count_data(&b->keys) < oldsize);
2074         return ret;
2075 }
2076 
2077 static int btree_split(struct btree *b, struct btree_op *op,
2078                        struct keylist *insert_keys,
2079                        struct bkey *replace_key)
2080 {
2081         bool split;
2082         struct btree *n1, *n2 = NULL, *n3 = NULL;
2083         uint64_t start_time = local_clock();
2084         struct closure cl;
2085         struct keylist parent_keys;
2086 
2087         closure_init_stack(&cl);
2088         bch_keylist_init(&parent_keys);
2089 
2090         if (btree_check_reserve(b, op)) {
2091                 if (!b->level)
2092                         return -EINTR;
2093                 else
2094                         WARN(1, "insufficient reserve for split\n");
2095         }
2096 
2097         n1 = btree_node_alloc_replacement(b, op);
2098         if (IS_ERR(n1))
2099                 goto err;
2100 
2101         split = set_blocks(btree_bset_first(n1),
2102                            block_bytes(n1->c)) > (btree_blocks(b) * 4) / 5;
2103 
2104         if (split) {
2105                 unsigned int keys = 0;
2106 
2107                 trace_bcache_btree_node_split(b, btree_bset_first(n1)->keys);
2108 
2109                 n2 = bch_btree_node_alloc(b->c, op, b->level, b->parent);
2110                 if (IS_ERR(n2))
2111                         goto err_free1;
2112 
2113                 if (!b->parent) {
2114                         n3 = bch_btree_node_alloc(b->c, op, b->level + 1, NULL);
2115                         if (IS_ERR(n3))
2116                                 goto err_free2;
2117                 }
2118 
2119                 mutex_lock(&n1->write_lock);
2120                 mutex_lock(&n2->write_lock);
2121 
2122                 bch_btree_insert_keys(n1, op, insert_keys, replace_key);
2123 
2124                 /*
2125                  * Has to be a linear search because we don't have an auxiliary
2126                  * search tree yet
2127                  */
2128 
2129                 while (keys < (btree_bset_first(n1)->keys * 3) / 5)
2130                         keys += bkey_u64s(bset_bkey_idx(btree_bset_first(n1),
2131                                                         keys));
2132 
2133                 bkey_copy_key(&n1->key,
2134                               bset_bkey_idx(btree_bset_first(n1), keys));
2135                 keys += bkey_u64s(bset_bkey_idx(btree_bset_first(n1), keys));
2136 
2137                 btree_bset_first(n2)->keys = btree_bset_first(n1)->keys - keys;
2138                 btree_bset_first(n1)->keys = keys;
2139 
2140                 memcpy(btree_bset_first(n2)->start,
2141                        bset_bkey_last(btree_bset_first(n1)),
2142                        btree_bset_first(n2)->keys * sizeof(uint64_t));
2143 
2144                 bkey_copy_key(&n2->key, &b->key);
2145 
2146                 bch_keylist_add(&parent_keys, &n2->key);
2147                 bch_btree_node_write(n2, &cl);
2148                 mutex_unlock(&n2->write_lock);
2149                 rw_unlock(true, n2);
2150         } else {
2151                 trace_bcache_btree_node_compact(b, btree_bset_first(n1)->keys);
2152 
2153                 mutex_lock(&n1->write_lock);
2154                 bch_btree_insert_keys(n1, op, insert_keys, replace_key);
2155         }
2156 
2157         bch_keylist_add(&parent_keys, &n1->key);
2158         bch_btree_node_write(n1, &cl);
2159         mutex_unlock(&n1->write_lock);
2160 
2161         if (n3) {
2162                 /* Depth increases, make a new root */
2163                 mutex_lock(&n3->write_lock);
2164                 bkey_copy_key(&n3->key, &MAX_KEY);
2165                 bch_btree_insert_keys(n3, op, &parent_keys, NULL);
2166                 bch_btree_node_write(n3, &cl);
2167                 mutex_unlock(&n3->write_lock);
2168 
2169                 closure_sync(&cl);
2170                 bch_btree_set_root(n3);
2171                 rw_unlock(true, n3);
2172         } else if (!b->parent) {
2173                 /* Root filled up but didn't need to be split */
2174                 closure_sync(&cl);
2175                 bch_btree_set_root(n1);
2176         } else {
2177                 /* Split a non root node */
2178                 closure_sync(&cl);
2179                 make_btree_freeing_key(b, parent_keys.top);
2180                 bch_keylist_push(&parent_keys);
2181 
2182                 bch_btree_insert_node(b->parent, op, &parent_keys, NULL, NULL);
2183                 BUG_ON(!bch_keylist_empty(&parent_keys));
2184         }
2185 
2186         btree_node_free(b);
2187         rw_unlock(true, n1);
2188 
2189         bch_time_stats_update(&b->c->btree_split_time, start_time);
2190 
2191         return 0;
2192 err_free2:
2193         bkey_put(b->c, &n2->key);
2194         btree_node_free(n2);
2195         rw_unlock(true, n2);
2196 err_free1:
2197         bkey_put(b->c, &n1->key);
2198         btree_node_free(n1);
2199         rw_unlock(true, n1);
2200 err:
2201         WARN(1, "bcache: btree split failed (level %u)", b->level);
2202 
2203         if (n3 == ERR_PTR(-EAGAIN) ||
2204             n2 == ERR_PTR(-EAGAIN) ||
2205             n1 == ERR_PTR(-EAGAIN))
2206                 return -EAGAIN;
2207 
2208         return -ENOMEM;
2209 }
2210 
2211 static int bch_btree_insert_node(struct btree *b, struct btree_op *op,
2212                                  struct keylist *insert_keys,
2213                                  atomic_t *journal_ref,
2214                                  struct bkey *replace_key)
2215 {
2216         struct closure cl;
2217 
2218         BUG_ON(b->level && replace_key);
2219 
2220         closure_init_stack(&cl);
2221 
2222         mutex_lock(&b->write_lock);
2223 
2224         if (write_block(b) != btree_bset_last(b) &&
2225             b->keys.last_set_unwritten)
2226                 bch_btree_init_next(b); /* just wrote a set */
2227 
2228         if (bch_keylist_nkeys(insert_keys) > insert_u64s_remaining(b)) {
2229                 mutex_unlock(&b->write_lock);
2230                 goto split;
2231         }
2232 
2233         BUG_ON(write_block(b) != btree_bset_last(b));
2234 
2235         if (bch_btree_insert_keys(b, op, insert_keys, replace_key)) {
2236                 if (!b->level)
2237                         bch_btree_leaf_dirty(b, journal_ref);
2238                 else
2239                         bch_btree_node_write(b, &cl);
2240         }
2241 
2242         mutex_unlock(&b->write_lock);
2243 
2244         /* wait for btree node write if necessary, after unlock */
2245         closure_sync(&cl);
2246 
2247         return 0;
2248 split:
2249         if (current->bio_list) {
2250                 op->lock = b->c->root->level + 1;
2251                 return -EAGAIN;
2252         } else if (op->lock <= b->c->root->level) {
2253                 op->lock = b->c->root->level + 1;
2254                 return -EINTR;
2255         } else {
2256                 /* Invalidated all iterators */
2257                 int ret = btree_split(b, op, insert_keys, replace_key);
2258 
2259                 if (bch_keylist_empty(insert_keys))
2260                         return 0;
2261                 else if (!ret)
2262                         return -EINTR;
2263                 return ret;
2264         }
2265 }
2266 
2267 int bch_btree_insert_check_key(struct btree *b, struct btree_op *op,
2268                                struct bkey *check_key)
2269 {
2270         int ret = -EINTR;
2271         uint64_t btree_ptr = b->key.ptr[0];
2272         unsigned long seq = b->seq;
2273         struct keylist insert;
2274         bool upgrade = op->lock == -1;
2275 
2276         bch_keylist_init(&insert);
2277 
2278         if (upgrade) {
2279                 rw_unlock(false, b);
2280                 rw_lock(true, b, b->level);
2281 
2282                 if (b->key.ptr[0] != btree_ptr ||
2283                     b->seq != seq + 1) {
2284                         op->lock = b->level;
2285                         goto out;
2286                 }
2287         }
2288 
2289         SET_KEY_PTRS(check_key, 1);
2290         get_random_bytes(&check_key->ptr[0], sizeof(uint64_t));
2291 
2292         SET_PTR_DEV(check_key, 0, PTR_CHECK_DEV);
2293 
2294         bch_keylist_add(&insert, check_key);
2295 
2296         ret = bch_btree_insert_node(b, op, &insert, NULL, NULL);
2297 
2298         BUG_ON(!ret && !bch_keylist_empty(&insert));
2299 out:
2300         if (upgrade)
2301                 downgrade_write(&b->lock);
2302         return ret;
2303 }
2304 
2305 struct btree_insert_op {
2306         struct btree_op op;
2307         struct keylist  *keys;
2308         atomic_t        *journal_ref;
2309         struct bkey     *replace_key;
2310 };
2311 
2312 static int btree_insert_fn(struct btree_op *b_op, struct btree *b)
2313 {
2314         struct btree_insert_op *op = container_of(b_op,
2315                                         struct btree_insert_op, op);
2316 
2317         int ret = bch_btree_insert_node(b, &op->op, op->keys,
2318                                         op->journal_ref, op->replace_key);
2319         if (ret && !bch_keylist_empty(op->keys))
2320                 return ret;
2321         else
2322                 return MAP_DONE;
2323 }
2324 
2325 int bch_btree_insert(struct cache_set *c, struct keylist *keys,
2326                      atomic_t *journal_ref, struct bkey *replace_key)
2327 {
2328         struct btree_insert_op op;
2329         int ret = 0;
2330 
2331         BUG_ON(current->bio_list);
2332         BUG_ON(bch_keylist_empty(keys));
2333 
2334         bch_btree_op_init(&op.op, 0);
2335         op.keys         = keys;
2336         op.journal_ref  = journal_ref;
2337         op.replace_key  = replace_key;
2338 
2339         while (!ret && !bch_keylist_empty(keys)) {
2340                 op.op.lock = 0;
2341                 ret = bch_btree_map_leaf_nodes(&op.op, c,
2342                                                &START_KEY(keys->keys),
2343                                                btree_insert_fn);
2344         }
2345 
2346         if (ret) {
2347                 struct bkey *k;
2348 
2349                 pr_err("error %i", ret);
2350 
2351                 while ((k = bch_keylist_pop(keys)))
2352                         bkey_put(c, k);
2353         } else if (op.op.insert_collision)
2354                 ret = -ESRCH;
2355 
2356         return ret;
2357 }
2358 
2359 void bch_btree_set_root(struct btree *b)
2360 {
2361         unsigned int i;
2362         struct closure cl;
2363 
2364         closure_init_stack(&cl);
2365 
2366         trace_bcache_btree_set_root(b);
2367 
2368         BUG_ON(!b->written);
2369 
2370         for (i = 0; i < KEY_PTRS(&b->key); i++)
2371                 BUG_ON(PTR_BUCKET(b->c, &b->key, i)->prio != BTREE_PRIO);
2372 
2373         mutex_lock(&b->c->bucket_lock);
2374         list_del_init(&b->list);
2375         mutex_unlock(&b->c->bucket_lock);
2376 
2377         b->c->root = b;
2378 
2379         bch_journal_meta(b->c, &cl);
2380         closure_sync(&cl);
2381 }
2382 
2383 /* Map across nodes or keys */
2384 
2385 static int bch_btree_map_nodes_recurse(struct btree *b, struct btree_op *op,
2386                                        struct bkey *from,
2387                                        btree_map_nodes_fn *fn, int flags)
2388 {
2389         int ret = MAP_CONTINUE;
2390 
2391         if (b->level) {
2392                 struct bkey *k;
2393                 struct btree_iter iter;
2394 
2395                 bch_btree_iter_init(&b->keys, &iter, from);
2396 
2397                 while ((k = bch_btree_iter_next_filter(&iter, &b->keys,
2398                                                        bch_ptr_bad))) {
2399                         ret = btree(map_nodes_recurse, k, b,
2400                                     op, from, fn, flags);
2401                         from = NULL;
2402 
2403                         if (ret != MAP_CONTINUE)
2404                                 return ret;
2405                 }
2406         }
2407 
2408         if (!b->level || flags == MAP_ALL_NODES)
2409                 ret = fn(op, b);
2410 
2411         return ret;
2412 }
2413 
2414 int __bch_btree_map_nodes(struct btree_op *op, struct cache_set *c,
2415                           struct bkey *from, btree_map_nodes_fn *fn, int flags)
2416 {
2417         return btree_root(map_nodes_recurse, c, op, from, fn, flags);
2418 }
2419 
2420 static int bch_btree_map_keys_recurse(struct btree *b, struct btree_op *op,
2421                                       struct bkey *from, btree_map_keys_fn *fn,
2422                                       int flags)
2423 {
2424         int ret = MAP_CONTINUE;
2425         struct bkey *k;
2426         struct btree_iter iter;
2427 
2428         bch_btree_iter_init(&b->keys, &iter, from);
2429 
2430         while ((k = bch_btree_iter_next_filter(&iter, &b->keys, bch_ptr_bad))) {
2431                 ret = !b->level
2432                         ? fn(op, b, k)
2433                         : btree(map_keys_recurse, k, b, op, from, fn, flags);
2434                 from = NULL;
2435 
2436                 if (ret != MAP_CONTINUE)
2437                         return ret;
2438         }
2439 
2440         if (!b->level && (flags & MAP_END_KEY))
2441                 ret = fn(op, b, &KEY(KEY_INODE(&b->key),
2442                                      KEY_OFFSET(&b->key), 0));
2443 
2444         return ret;
2445 }
2446 
2447 int bch_btree_map_keys(struct btree_op *op, struct cache_set *c,
2448                        struct bkey *from, btree_map_keys_fn *fn, int flags)
2449 {
2450         return btree_root(map_keys_recurse, c, op, from, fn, flags);
2451 }
2452 
2453 /* Keybuf code */
2454 
2455 static inline int keybuf_cmp(struct keybuf_key *l, struct keybuf_key *r)
2456 {
2457         /* Overlapping keys compare equal */
2458         if (bkey_cmp(&l->key, &START_KEY(&r->key)) <= 0)
2459                 return -1;
2460         if (bkey_cmp(&START_KEY(&l->key), &r->key) >= 0)
2461                 return 1;
2462         return 0;
2463 }
2464 
2465 static inline int keybuf_nonoverlapping_cmp(struct keybuf_key *l,
2466                                             struct keybuf_key *r)
2467 {
2468         return clamp_t(int64_t, bkey_cmp(&l->key, &r->key), -1, 1);
2469 }
2470 
2471 struct refill {
2472         struct btree_op op;
2473         unsigned int    nr_found;
2474         struct keybuf   *buf;
2475         struct bkey     *end;
2476         keybuf_pred_fn  *pred;
2477 };
2478 
2479 static int refill_keybuf_fn(struct btree_op *op, struct btree *b,
2480                             struct bkey *k)
2481 {
2482         struct refill *refill = container_of(op, struct refill, op);
2483         struct keybuf *buf = refill->buf;
2484         int ret = MAP_CONTINUE;
2485 
2486         if (bkey_cmp(k, refill->end) > 0) {
2487                 ret = MAP_DONE;
2488                 goto out;
2489         }
2490 
2491         if (!KEY_SIZE(k)) /* end key */
2492                 goto out;
2493 
2494         if (refill->pred(buf, k)) {
2495                 struct keybuf_key *w;
2496 
2497                 spin_lock(&buf->lock);
2498 
2499                 w = array_alloc(&buf->freelist);
2500                 if (!w) {
2501                         spin_unlock(&buf->lock);
2502                         return MAP_DONE;
2503                 }
2504 
2505                 w->private = NULL;
2506                 bkey_copy(&w->key, k);
2507 
2508                 if (RB_INSERT(&buf->keys, w, node, keybuf_cmp))
2509                         array_free(&buf->freelist, w);
2510                 else
2511                         refill->nr_found++;
2512 
2513                 if (array_freelist_empty(&buf->freelist))
2514                         ret = MAP_DONE;
2515 
2516                 spin_unlock(&buf->lock);
2517         }
2518 out:
2519         buf->last_scanned = *k;
2520         return ret;
2521 }
2522 
2523 void bch_refill_keybuf(struct cache_set *c, struct keybuf *buf,
2524                        struct bkey *end, keybuf_pred_fn *pred)
2525 {
2526         struct bkey start = buf->last_scanned;
2527         struct refill refill;
2528 
2529         cond_resched();
2530 
2531         bch_btree_op_init(&refill.op, -1);
2532         refill.nr_found = 0;
2533         refill.buf      = buf;
2534         refill.end      = end;
2535         refill.pred     = pred;
2536 
2537         bch_btree_map_keys(&refill.op, c, &buf->last_scanned,
2538                            refill_keybuf_fn, MAP_END_KEY);
2539 
2540         trace_bcache_keyscan(refill.nr_found,
2541                              KEY_INODE(&start), KEY_OFFSET(&start),
2542                              KEY_INODE(&buf->last_scanned),
2543                              KEY_OFFSET(&buf->last_scanned));
2544 
2545         spin_lock(&buf->lock);
2546 
2547         if (!RB_EMPTY_ROOT(&buf->keys)) {
2548                 struct keybuf_key *w;
2549 
2550                 w = RB_FIRST(&buf->keys, struct keybuf_key, node);
2551                 buf->start      = START_KEY(&w->key);
2552 
2553                 w = RB_LAST(&buf->keys, struct keybuf_key, node);
2554                 buf->end        = w->key;
2555         } else {
2556                 buf->start      = MAX_KEY;
2557                 buf->end        = MAX_KEY;
2558         }
2559 
2560         spin_unlock(&buf->lock);
2561 }
2562 
2563 static void __bch_keybuf_del(struct keybuf *buf, struct keybuf_key *w)
2564 {
2565         rb_erase(&w->node, &buf->keys);
2566         array_free(&buf->freelist, w);
2567 }
2568 
2569 void bch_keybuf_del(struct keybuf *buf, struct keybuf_key *w)
2570 {
2571         spin_lock(&buf->lock);
2572         __bch_keybuf_del(buf, w);
2573         spin_unlock(&buf->lock);
2574 }
2575 
2576 bool bch_keybuf_check_overlapping(struct keybuf *buf, struct bkey *start,
2577                                   struct bkey *end)
2578 {
2579         bool ret = false;
2580         struct keybuf_key *p, *w, s;
2581 
2582         s.key = *start;
2583 
2584         if (bkey_cmp(end, &buf->start) <= 0 ||
2585             bkey_cmp(start, &buf->end) >= 0)
2586                 return false;
2587 
2588         spin_lock(&buf->lock);
2589         w = RB_GREATER(&buf->keys, s, node, keybuf_nonoverlapping_cmp);
2590 
2591         while (w && bkey_cmp(&START_KEY(&w->key), end) < 0) {
2592                 p = w;
2593                 w = RB_NEXT(w, node);
2594 
2595                 if (p->private)
2596                         ret = true;
2597                 else
2598                         __bch_keybuf_del(buf, p);
2599         }
2600 
2601         spin_unlock(&buf->lock);
2602         return ret;
2603 }
2604 
2605 struct keybuf_key *bch_keybuf_next(struct keybuf *buf)
2606 {
2607         struct keybuf_key *w;
2608 
2609         spin_lock(&buf->lock);
2610 
2611         w = RB_FIRST(&buf->keys, struct keybuf_key, node);
2612 
2613         while (w && w->private)
2614                 w = RB_NEXT(w, node);
2615 
2616         if (w)
2617                 w->private = ERR_PTR(-EINTR);
2618 
2619         spin_unlock(&buf->lock);
2620         return w;
2621 }
2622 
2623 struct keybuf_key *bch_keybuf_next_rescan(struct cache_set *c,
2624                                           struct keybuf *buf,
2625                                           struct bkey *end,
2626                                           keybuf_pred_fn *pred)
2627 {
2628         struct keybuf_key *ret;
2629 
2630         while (1) {
2631                 ret = bch_keybuf_next(buf);
2632                 if (ret)
2633                         break;
2634 
2635                 if (bkey_cmp(&buf->last_scanned, end) >= 0) {
2636                         pr_debug("scan finished");
2637                         break;
2638                 }
2639 
2640                 bch_refill_keybuf(c, buf, end, pred);
2641         }
2642 
2643         return ret;
2644 }
2645 
2646 void bch_keybuf_init(struct keybuf *buf)
2647 {
2648         buf->last_scanned       = MAX_KEY;
2649         buf->keys               = RB_ROOT;
2650 
2651         spin_lock_init(&buf->lock);
2652         array_allocator_init(&buf->freelist);
2653 }
/* [<][>][^][v][top][bottom][index][help] */
root/drivers/md/bcache/btree.c

DEFINITIONS
