fs/btrfs/free-space-cache.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2008 Red Hat.  All rights reserved.
   4  */
   5
   6 #include <linux/pagemap.h>
   7 #include <linux/sched.h>
   8 #include <linux/sched/signal.h>
   9 #include <linux/slab.h>
  10 #include <linux/math64.h>
  11 #include <linux/ratelimit.h>
  12 #include <linux/error-injection.h>
  13 #include <linux/sched/mm.h>
  14 #include "ctree.h"
  15 #include "fs.h"
  16 #include "messages.h"
  17 #include "misc.h"
  18 #include "free-space-cache.h"
  19 #include "transaction.h"
  20 #include "disk-io.h"
  21 #include "extent_io.h"
  22 #include "volumes.h"
  23 #include "space-info.h"
  24 #include "delalloc-space.h"
  25 #include "block-group.h"
  26 #include "discard.h"
  27 #include "subpage.h"
  28 #include "inode-item.h"
  29 #include "accessors.h"
  30 #include "file-item.h"
  31 #include "file.h"
  32 #include "super.h"
  33
  34 #define BITS_PER_BITMAP         (PAGE_SIZE * 8UL)
  35 #define MAX_CACHE_BYTES_PER_GIG SZ_64K
  36 #define FORCE_EXTENT_THRESHOLD  SZ_1M
  37
  38 static struct kmem_cache *btrfs_free_space_cachep;
  39 static struct kmem_cache *btrfs_free_space_bitmap_cachep;
  40
  41 struct btrfs_trim_range {
  42         u64 start;
  43         u64 bytes;
  44         struct list_head list;
  45 };
  46
  47 static int link_free_space(struct btrfs_free_space_ctl *ctl,
  48                            struct btrfs_free_space *info);
  49 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
  50                               struct btrfs_free_space *info, bool update_stat);
  51 static int search_bitmap(struct btrfs_free_space_ctl *ctl,
  52                          struct btrfs_free_space *bitmap_info, u64 *offset,
  53                          u64 *bytes, bool for_alloc);
  54 static void free_bitmap(struct btrfs_free_space_ctl *ctl,
  55                         struct btrfs_free_space *bitmap_info);
  56 static void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
  57                               struct btrfs_free_space *info, u64 offset,
  58                               u64 bytes, bool update_stats);
  59
  60 static void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl)
  61 {
  62         struct btrfs_free_space *info;
  63         struct rb_node *node;
  64
  65         while ((node = rb_last(&ctl->free_space_offset)) != NULL) {
  66                 info = rb_entry(node, struct btrfs_free_space, offset_index);
  67                 if (!info->bitmap) {
  68                         unlink_free_space(ctl, info, true);
  69                         kmem_cache_free(btrfs_free_space_cachep, info);
  70                 } else {
  71                         free_bitmap(ctl, info);
  72                 }
  73
  74                 cond_resched_lock(&ctl->tree_lock);
  75         }
  76 }
  77
  78 static struct inode *__lookup_free_space_inode(struct btrfs_root *root,
  79                                                struct btrfs_path *path,
  80                                                u64 offset)
  81 {
  82         struct btrfs_fs_info *fs_info = root->fs_info;
  83         struct btrfs_key key;
  84         struct btrfs_key location;
  85         struct btrfs_disk_key disk_key;
  86         struct btrfs_free_space_header *header;
  87         struct extent_buffer *leaf;
  88         struct inode *inode = NULL;
  89         unsigned nofs_flag;
  90         int ret;
  91
  92         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
  93         key.offset = offset;
  94         key.type = 0;
  95
  96         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
  97         if (ret < 0)
  98                 return ERR_PTR(ret);
  99         if (ret > 0) {
 100                 btrfs_release_path(path);
 101                 return ERR_PTR(-ENOENT);
 102         }
 103
 104         leaf = path->nodes[0];
 105         header = btrfs_item_ptr(leaf, path->slots[0],
 106                                 struct btrfs_free_space_header);
 107         btrfs_free_space_key(leaf, header, &disk_key);
 108         btrfs_disk_key_to_cpu(&location, &disk_key);
 109         btrfs_release_path(path);
 110
 111         /*
 112          * We are often under a trans handle at this point, so we need to make
 113          * sure NOFS is set to keep us from deadlocking.
 114          */
 115         nofs_flag = memalloc_nofs_save();
 116         inode = btrfs_iget_path(fs_info->sb, location.objectid, root, path);
 117         btrfs_release_path(path);
 118         memalloc_nofs_restore(nofs_flag);
 119         if (IS_ERR(inode))
 120                 return inode;
 121
 122         mapping_set_gfp_mask(inode->i_mapping,
 123                         mapping_gfp_constraint(inode->i_mapping,
 124                         ~(__GFP_FS | __GFP_HIGHMEM)));
 125
 126         return inode;
 127 }
 128
 129 struct inode *lookup_free_space_inode(struct btrfs_block_group *block_group,
 130                 struct btrfs_path *path)
 131 {
 132         struct btrfs_fs_info *fs_info = block_group->fs_info;
 133         struct inode *inode = NULL;
 134         u32 flags = BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
 135
 136         spin_lock(&block_group->lock);
 137         if (block_group->inode)
 138                 inode = igrab(block_group->inode);
 139         spin_unlock(&block_group->lock);
 140         if (inode)
 141                 return inode;
 142
 143         inode = __lookup_free_space_inode(fs_info->tree_root, path,
 144                                           block_group->start);
 145         if (IS_ERR(inode))
 146                 return inode;
 147
 148         spin_lock(&block_group->lock);
 149         if (!((BTRFS_I(inode)->flags & flags) == flags)) {
 150                 btrfs_info(fs_info, "Old style space inode found, converting.");
 151                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM |
 152                         BTRFS_INODE_NODATACOW;
 153                 block_group->disk_cache_state = BTRFS_DC_CLEAR;
 154         }
 155
 156         if (!test_and_set_bit(BLOCK_GROUP_FLAG_IREF, &block_group->runtime_flags))
 157                 block_group->inode = igrab(inode);
 158         spin_unlock(&block_group->lock);
 159
 160         return inode;
 161 }
 162
 163 static int __create_free_space_inode(struct btrfs_root *root,
 164                                      struct btrfs_trans_handle *trans,
 165                                      struct btrfs_path *path,
 166                                      u64 ino, u64 offset)
 167 {
 168         struct btrfs_key key;
 169         struct btrfs_disk_key disk_key;
 170         struct btrfs_free_space_header *header;
 171         struct btrfs_inode_item *inode_item;
 172         struct extent_buffer *leaf;
 173         /* We inline CRCs for the free disk space cache */
 174         const u64 flags = BTRFS_INODE_NOCOMPRESS | BTRFS_INODE_PREALLOC |
 175                           BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
 176         int ret;
 177
 178         ret = btrfs_insert_empty_inode(trans, root, path, ino);
 179         if (ret)
 180                 return ret;
 181
 182         leaf = path->nodes[0];
 183         inode_item = btrfs_item_ptr(leaf, path->slots[0],
 184                                     struct btrfs_inode_item);
 185         btrfs_item_key(leaf, &disk_key, path->slots[0]);
 186         memzero_extent_buffer(leaf, (unsigned long)inode_item,
 187                              sizeof(*inode_item));
 188         btrfs_set_inode_generation(leaf, inode_item, trans->transid);
 189         btrfs_set_inode_size(leaf, inode_item, 0);
 190         btrfs_set_inode_nbytes(leaf, inode_item, 0);
 191         btrfs_set_inode_uid(leaf, inode_item, 0);
 192         btrfs_set_inode_gid(leaf, inode_item, 0);
 193         btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
 194         btrfs_set_inode_flags(leaf, inode_item, flags);
 195         btrfs_set_inode_nlink(leaf, inode_item, 1);
 196         btrfs_set_inode_transid(leaf, inode_item, trans->transid);
 197         btrfs_set_inode_block_group(leaf, inode_item, offset);
 198         btrfs_mark_buffer_dirty(leaf);
 199         btrfs_release_path(path);
 200
 201         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
 202         key.offset = offset;
 203         key.type = 0;
 204         ret = btrfs_insert_empty_item(trans, root, path, &key,
 205                                       sizeof(struct btrfs_free_space_header));
 206         if (ret < 0) {
 207                 btrfs_release_path(path);
 208                 return ret;
 209         }
 210
 211         leaf = path->nodes[0];
 212         header = btrfs_item_ptr(leaf, path->slots[0],
 213                                 struct btrfs_free_space_header);
 214         memzero_extent_buffer(leaf, (unsigned long)header, sizeof(*header));
 215         btrfs_set_free_space_key(leaf, header, &disk_key);
 216         btrfs_mark_buffer_dirty(leaf);
 217         btrfs_release_path(path);
 218
 219         return 0;
 220 }
 221
 222 int create_free_space_inode(struct btrfs_trans_handle *trans,
 223                             struct btrfs_block_group *block_group,
 224                             struct btrfs_path *path)
 225 {
 226         int ret;
 227         u64 ino;
 228
 229         ret = btrfs_get_free_objectid(trans->fs_info->tree_root, &ino);
 230         if (ret < 0)
 231                 return ret;
 232
 233         return __create_free_space_inode(trans->fs_info->tree_root, trans, path,
 234                                          ino, block_group->start);
 235 }
 236
 237 /*
 238  * inode is an optional sink: if it is NULL, btrfs_remove_free_space_inode
 239  * handles lookup, otherwise it takes ownership and iputs the inode.
 240  * Don't reuse an inode pointer after passing it into this function.
 241  */
 242 int btrfs_remove_free_space_inode(struct btrfs_trans_handle *trans,
 243                                   struct inode *inode,
 244                                   struct btrfs_block_group *block_group)
 245 {
 246         struct btrfs_path *path;
 247         struct btrfs_key key;
 248         int ret = 0;
 249
 250         path = btrfs_alloc_path();
 251         if (!path)
 252                 return -ENOMEM;
 253
 254         if (!inode)
 255                 inode = lookup_free_space_inode(block_group, path);
 256         if (IS_ERR(inode)) {
 257                 if (PTR_ERR(inode) != -ENOENT)
 258                         ret = PTR_ERR(inode);
 259                 goto out;
 260         }
 261         ret = btrfs_orphan_add(trans, BTRFS_I(inode));
 262         if (ret) {
 263                 btrfs_add_delayed_iput(BTRFS_I(inode));
 264                 goto out;
 265         }
 266         clear_nlink(inode);
 267         /* One for the block groups ref */
 268         spin_lock(&block_group->lock);
 269         if (test_and_clear_bit(BLOCK_GROUP_FLAG_IREF, &block_group->runtime_flags)) {
 270                 block_group->inode = NULL;
 271                 spin_unlock(&block_group->lock);
 272                 iput(inode);
 273         } else {
 274                 spin_unlock(&block_group->lock);
 275         }
 276         /* One for the lookup ref */
 277         btrfs_add_delayed_iput(BTRFS_I(inode));
 278
 279         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
 280         key.type = 0;
 281         key.offset = block_group->start;
 282         ret = btrfs_search_slot(trans, trans->fs_info->tree_root, &key, path,
 283                                 -1, 1);
 284         if (ret) {
 285                 if (ret > 0)
 286                         ret = 0;
 287                 goto out;
 288         }
 289         ret = btrfs_del_item(trans, trans->fs_info->tree_root, path);
 290 out:
 291         btrfs_free_path(path);
 292         return ret;
 293 }
 294
 295 int btrfs_check_trunc_cache_free_space(struct btrfs_fs_info *fs_info,
 296                                        struct btrfs_block_rsv *rsv)
 297 {
 298         u64 needed_bytes;
 299         int ret;
 300
 301         /* 1 for slack space, 1 for updating the inode */
 302         needed_bytes = btrfs_calc_insert_metadata_size(fs_info, 1) +
 303                 btrfs_calc_metadata_size(fs_info, 1);
 304
 305         spin_lock(&rsv->lock);
 306         if (rsv->reserved < needed_bytes)
 307                 ret = -ENOSPC;
 308         else
 309                 ret = 0;
 310         spin_unlock(&rsv->lock);
 311         return ret;
 312 }
 313
 314 int btrfs_truncate_free_space_cache(struct btrfs_trans_handle *trans,
 315                                     struct btrfs_block_group *block_group,
 316                                     struct inode *vfs_inode)
 317 {
 318         struct btrfs_truncate_control control = {
 319                 .inode = BTRFS_I(vfs_inode),
 320                 .new_size = 0,
 321                 .ino = btrfs_ino(BTRFS_I(vfs_inode)),
 322                 .min_type = BTRFS_EXTENT_DATA_KEY,
 323                 .clear_extent_range = true,
 324         };
 325         struct btrfs_inode *inode = BTRFS_I(vfs_inode);
 326         struct btrfs_root *root = inode->root;
 327         struct extent_state *cached_state = NULL;
 328         int ret = 0;
 329         bool locked = false;
 330
 331         if (block_group) {
 332                 struct btrfs_path *path = btrfs_alloc_path();
 333
 334                 if (!path) {
 335                         ret = -ENOMEM;
 336                         goto fail;
 337                 }
 338                 locked = true;
 339                 mutex_lock(&trans->transaction->cache_write_mutex);
 340                 if (!list_empty(&block_group->io_list)) {
 341                         list_del_init(&block_group->io_list);
 342
 343                         btrfs_wait_cache_io(trans, block_group, path);
 344                         btrfs_put_block_group(block_group);
 345                 }
 346
 347                 /*
 348                  * now that we've truncated the cache away, its no longer
 349                  * setup or written
 350                  */
 351                 spin_lock(&block_group->lock);
 352                 block_group->disk_cache_state = BTRFS_DC_CLEAR;
 353                 spin_unlock(&block_group->lock);
 354                 btrfs_free_path(path);
 355         }
 356
 357         btrfs_i_size_write(inode, 0);
 358         truncate_pagecache(vfs_inode, 0);
 359
 360         lock_extent(&inode->io_tree, 0, (u64)-1, &cached_state);
 361         btrfs_drop_extent_map_range(inode, 0, (u64)-1, false);
 362
 363         /*
 364          * We skip the throttling logic for free space cache inodes, so we don't
 365          * need to check for -EAGAIN.
 366          */
 367         ret = btrfs_truncate_inode_items(trans, root, &control);
 368
 369         inode_sub_bytes(&inode->vfs_inode, control.sub_bytes);
 370         btrfs_inode_safe_disk_i_size_write(inode, control.last_size);
 371
 372         unlock_extent(&inode->io_tree, 0, (u64)-1, &cached_state);
 373         if (ret)
 374                 goto fail;
 375
 376         ret = btrfs_update_inode(trans, root, inode);
 377
 378 fail:
 379         if (locked)
 380                 mutex_unlock(&trans->transaction->cache_write_mutex);
 381         if (ret)
 382                 btrfs_abort_transaction(trans, ret);
 383
 384         return ret;
 385 }
 386
 387 static void readahead_cache(struct inode *inode)
 388 {
 389         struct file_ra_state ra;
 390         unsigned long last_index;
 391
 392         file_ra_state_init(&ra, inode->i_mapping);
 393         last_index = (i_size_read(inode) - 1) >> PAGE_SHIFT;
 394
 395         page_cache_sync_readahead(inode->i_mapping, &ra, NULL, 0, last_index);
 396 }
 397
 398 static int io_ctl_init(struct btrfs_io_ctl *io_ctl, struct inode *inode,
 399                        int write)
 400 {
 401         int num_pages;
 402
 403         num_pages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
 404
 405         /* Make sure we can fit our crcs and generation into the first page */
 406         if (write && (num_pages * sizeof(u32) + sizeof(u64)) > PAGE_SIZE)
 407                 return -ENOSPC;
 408
 409         memset(io_ctl, 0, sizeof(struct btrfs_io_ctl));
 410
 411         io_ctl->pages = kcalloc(num_pages, sizeof(struct page *), GFP_NOFS);
 412         if (!io_ctl->pages)
 413                 return -ENOMEM;
 414
 415         io_ctl->num_pages = num_pages;
 416         io_ctl->fs_info = btrfs_sb(inode->i_sb);
 417         io_ctl->inode = inode;
 418
 419         return 0;
 420 }
 421 ALLOW_ERROR_INJECTION(io_ctl_init, ERRNO);
 422
 423 static void io_ctl_free(struct btrfs_io_ctl *io_ctl)
 424 {
 425         kfree(io_ctl->pages);
 426         io_ctl->pages = NULL;
 427 }
 428
 429 static void io_ctl_unmap_page(struct btrfs_io_ctl *io_ctl)
 430 {
 431         if (io_ctl->cur) {
 432                 io_ctl->cur = NULL;
 433                 io_ctl->orig = NULL;
 434         }
 435 }
 436
 437 static void io_ctl_map_page(struct btrfs_io_ctl *io_ctl, int clear)
 438 {
 439         ASSERT(io_ctl->index < io_ctl->num_pages);
 440         io_ctl->page = io_ctl->pages[io_ctl->index++];
 441         io_ctl->cur = page_address(io_ctl->page);
 442         io_ctl->orig = io_ctl->cur;
 443         io_ctl->size = PAGE_SIZE;
 444         if (clear)
 445                 clear_page(io_ctl->cur);
 446 }
 447
 448 static void io_ctl_drop_pages(struct btrfs_io_ctl *io_ctl)
 449 {
 450         int i;
 451
 452         io_ctl_unmap_page(io_ctl);
 453
 454         for (i = 0; i < io_ctl->num_pages; i++) {
 455                 if (io_ctl->pages[i]) {
 456                         btrfs_page_clear_checked(io_ctl->fs_info,
 457                                         io_ctl->pages[i],
 458                                         page_offset(io_ctl->pages[i]),
 459                                         PAGE_SIZE);
 460                         unlock_page(io_ctl->pages[i]);
 461                         put_page(io_ctl->pages[i]);
 462                 }
 463         }
 464 }
 465
 466 static int io_ctl_prepare_pages(struct btrfs_io_ctl *io_ctl, bool uptodate)
 467 {
 468         struct page *page;
 469         struct inode *inode = io_ctl->inode;
 470         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 471         int i;
 472
 473         for (i = 0; i < io_ctl->num_pages; i++) {
 474                 int ret;
 475
 476                 page = find_or_create_page(inode->i_mapping, i, mask);
 477                 if (!page) {
 478                         io_ctl_drop_pages(io_ctl);
 479                         return -ENOMEM;
 480                 }
 481
 482                 ret = set_page_extent_mapped(page);
 483                 if (ret < 0) {
 484                         unlock_page(page);
 485                         put_page(page);
 486                         io_ctl_drop_pages(io_ctl);
 487                         return ret;
 488                 }
 489
 490                 io_ctl->pages[i] = page;
 491                 if (uptodate && !PageUptodate(page)) {
 492                         btrfs_read_folio(NULL, page_folio(page));
 493                         lock_page(page);
 494                         if (page->mapping != inode->i_mapping) {
 495                                 btrfs_err(BTRFS_I(inode)->root->fs_info,
 496                                           "free space cache page truncated");
 497                                 io_ctl_drop_pages(io_ctl);
 498                                 return -EIO;
 499                         }
 500                         if (!PageUptodate(page)) {
 501                                 btrfs_err(BTRFS_I(inode)->root->fs_info,
 502                                            "error reading free space cache");
 503                                 io_ctl_drop_pages(io_ctl);
 504                                 return -EIO;
 505                         }
 506                 }
 507         }
 508
 509         for (i = 0; i < io_ctl->num_pages; i++)
 510                 clear_page_dirty_for_io(io_ctl->pages[i]);
 511
 512         return 0;
 513 }
 514
 515 static void io_ctl_set_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
 516 {
 517         io_ctl_map_page(io_ctl, 1);
 518
 519         /*
 520          * Skip the csum areas.  If we don't check crcs then we just have a
 521          * 64bit chunk at the front of the first page.
 522          */
 523         io_ctl->cur += (sizeof(u32) * io_ctl->num_pages);
 524         io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
 525
 526         put_unaligned_le64(generation, io_ctl->cur);
 527         io_ctl->cur += sizeof(u64);
 528 }
 529
 530 static int io_ctl_check_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
 531 {
 532         u64 cache_gen;
 533
 534         /*
 535          * Skip the crc area.  If we don't check crcs then we just have a 64bit
 536          * chunk at the front of the first page.
 537          */
 538         io_ctl->cur += sizeof(u32) * io_ctl->num_pages;
 539         io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
 540
 541         cache_gen = get_unaligned_le64(io_ctl->cur);
 542         if (cache_gen != generation) {
 543                 btrfs_err_rl(io_ctl->fs_info,
 544                         "space cache generation (%llu) does not match inode (%llu)",
 545                                 cache_gen, generation);
 546                 io_ctl_unmap_page(io_ctl);
 547                 return -EIO;
 548         }
 549         io_ctl->cur += sizeof(u64);
 550         return 0;
 551 }
 552
 553 static void io_ctl_set_crc(struct btrfs_io_ctl *io_ctl, int index)
 554 {
 555         u32 *tmp;
 556         u32 crc = ~(u32)0;
 557         unsigned offset = 0;
 558
 559         if (index == 0)
 560                 offset = sizeof(u32) * io_ctl->num_pages;
 561
 562         crc = btrfs_crc32c(crc, io_ctl->orig + offset, PAGE_SIZE - offset);
 563         btrfs_crc32c_final(crc, (u8 *)&crc);
 564         io_ctl_unmap_page(io_ctl);
 565         tmp = page_address(io_ctl->pages[0]);
 566         tmp += index;
 567         *tmp = crc;
 568 }
 569
 570 static int io_ctl_check_crc(struct btrfs_io_ctl *io_ctl, int index)
 571 {
 572         u32 *tmp, val;
 573         u32 crc = ~(u32)0;
 574         unsigned offset = 0;
 575
 576         if (index == 0)
 577                 offset = sizeof(u32) * io_ctl->num_pages;
 578
 579         tmp = page_address(io_ctl->pages[0]);
 580         tmp += index;
 581         val = *tmp;
 582
 583         io_ctl_map_page(io_ctl, 0);
 584         crc = btrfs_crc32c(crc, io_ctl->orig + offset, PAGE_SIZE - offset);
 585         btrfs_crc32c_final(crc, (u8 *)&crc);
 586         if (val != crc) {
 587                 btrfs_err_rl(io_ctl->fs_info,
 588                         "csum mismatch on free space cache");
 589                 io_ctl_unmap_page(io_ctl);
 590                 return -EIO;
 591         }
 592
 593         return 0;
 594 }
 595
 596 static int io_ctl_add_entry(struct btrfs_io_ctl *io_ctl, u64 offset, u64 bytes,
 597                             void *bitmap)
 598 {
 599         struct btrfs_free_space_entry *entry;
 600
 601         if (!io_ctl->cur)
 602                 return -ENOSPC;
 603
 604         entry = io_ctl->cur;
 605         put_unaligned_le64(offset, &entry->offset);
 606         put_unaligned_le64(bytes, &entry->bytes);
 607         entry->type = (bitmap) ? BTRFS_FREE_SPACE_BITMAP :
 608                 BTRFS_FREE_SPACE_EXTENT;
 609         io_ctl->cur += sizeof(struct btrfs_free_space_entry);
 610         io_ctl->size -= sizeof(struct btrfs_free_space_entry);
 611
 612         if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
 613                 return 0;
 614
 615         io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 616
 617         /* No more pages to map */
 618         if (io_ctl->index >= io_ctl->num_pages)
 619                 return 0;
 620
 621         /* map the next page */
 622         io_ctl_map_page(io_ctl, 1);
 623         return 0;
 624 }
 625
 626 static int io_ctl_add_bitmap(struct btrfs_io_ctl *io_ctl, void *bitmap)
 627 {
 628         if (!io_ctl->cur)
 629                 return -ENOSPC;
 630
 631         /*
 632          * If we aren't at the start of the current page, unmap this one and
 633          * map the next one if there is any left.
 634          */
 635         if (io_ctl->cur != io_ctl->orig) {
 636                 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 637                 if (io_ctl->index >= io_ctl->num_pages)
 638                         return -ENOSPC;
 639                 io_ctl_map_page(io_ctl, 0);
 640         }
 641
 642         copy_page(io_ctl->cur, bitmap);
 643         io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 644         if (io_ctl->index < io_ctl->num_pages)
 645                 io_ctl_map_page(io_ctl, 0);
 646         return 0;
 647 }
 648
 649 static void io_ctl_zero_remaining_pages(struct btrfs_io_ctl *io_ctl)
 650 {
 651         /*
 652          * If we're not on the boundary we know we've modified the page and we
 653          * need to crc the page.
 654          */
 655         if (io_ctl->cur != io_ctl->orig)
 656                 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 657         else
 658                 io_ctl_unmap_page(io_ctl);
 659
 660         while (io_ctl->index < io_ctl->num_pages) {
 661                 io_ctl_map_page(io_ctl, 1);
 662                 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 663         }
 664 }
 665
 666 static int io_ctl_read_entry(struct btrfs_io_ctl *io_ctl,
 667                             struct btrfs_free_space *entry, u8 *type)
 668 {
 669         struct btrfs_free_space_entry *e;
 670         int ret;
 671
 672         if (!io_ctl->cur) {
 673                 ret = io_ctl_check_crc(io_ctl, io_ctl->index);
 674                 if (ret)
 675                         return ret;
 676         }
 677
 678         e = io_ctl->cur;
 679         entry->offset = get_unaligned_le64(&e->offset);
 680         entry->bytes = get_unaligned_le64(&e->bytes);
 681         *type = e->type;
 682         io_ctl->cur += sizeof(struct btrfs_free_space_entry);
 683         io_ctl->size -= sizeof(struct btrfs_free_space_entry);
 684
 685         if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
 686                 return 0;
 687
 688         io_ctl_unmap_page(io_ctl);
 689
 690         return 0;
 691 }
 692
 693 static int io_ctl_read_bitmap(struct btrfs_io_ctl *io_ctl,
 694                               struct btrfs_free_space *entry)
 695 {
 696         int ret;
 697
 698         ret = io_ctl_check_crc(io_ctl, io_ctl->index);
 699         if (ret)
 700                 return ret;
 701
 702         copy_page(entry->bitmap, io_ctl->cur);
 703         io_ctl_unmap_page(io_ctl);
 704
 705         return 0;
 706 }
 707
 708 static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
 709 {
 710         struct btrfs_block_group *block_group = ctl->block_group;
 711         u64 max_bytes;
 712         u64 bitmap_bytes;
 713         u64 extent_bytes;
 714         u64 size = block_group->length;
 715         u64 bytes_per_bg = BITS_PER_BITMAP * ctl->unit;
 716         u64 max_bitmaps = div64_u64(size + bytes_per_bg - 1, bytes_per_bg);
 717
 718         max_bitmaps = max_t(u64, max_bitmaps, 1);
 719
 720         if (ctl->total_bitmaps > max_bitmaps)
 721                 btrfs_err(block_group->fs_info,
 722 "invalid free space control: bg start=%llu len=%llu total_bitmaps=%u unit=%u max_bitmaps=%llu bytes_per_bg=%llu",
 723                           block_group->start, block_group->length,
 724                           ctl->total_bitmaps, ctl->unit, max_bitmaps,
 725                           bytes_per_bg);
 726         ASSERT(ctl->total_bitmaps <= max_bitmaps);
 727
 728         /*
 729          * We are trying to keep the total amount of memory used per 1GiB of
 730          * space to be MAX_CACHE_BYTES_PER_GIG.  However, with a reclamation
 731          * mechanism of pulling extents >= FORCE_EXTENT_THRESHOLD out of
 732          * bitmaps, we may end up using more memory than this.
 733          */
 734         if (size < SZ_1G)
 735                 max_bytes = MAX_CACHE_BYTES_PER_GIG;
 736         else
 737                 max_bytes = MAX_CACHE_BYTES_PER_GIG * div_u64(size, SZ_1G);
 738
 739         bitmap_bytes = ctl->total_bitmaps * ctl->unit;
 740
 741         /*
 742          * we want the extent entry threshold to always be at most 1/2 the max
 743          * bytes we can have, or whatever is less than that.
 744          */
 745         extent_bytes = max_bytes - bitmap_bytes;
 746         extent_bytes = min_t(u64, extent_bytes, max_bytes >> 1);
 747
 748         ctl->extents_thresh =
 749                 div_u64(extent_bytes, sizeof(struct btrfs_free_space));
 750 }
 751
 752 static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
 753                                    struct btrfs_free_space_ctl *ctl,
 754                                    struct btrfs_path *path, u64 offset)
 755 {
 756         struct btrfs_fs_info *fs_info = root->fs_info;
 757         struct btrfs_free_space_header *header;
 758         struct extent_buffer *leaf;
 759         struct btrfs_io_ctl io_ctl;
 760         struct btrfs_key key;
 761         struct btrfs_free_space *e, *n;
 762         LIST_HEAD(bitmaps);
 763         u64 num_entries;
 764         u64 num_bitmaps;
 765         u64 generation;
 766         u8 type;
 767         int ret = 0;
 768
 769         /* Nothing in the space cache, goodbye */
 770         if (!i_size_read(inode))
 771                 return 0;
 772
 773         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
 774         key.offset = offset;
 775         key.type = 0;
 776
 777         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 778         if (ret < 0)
 779                 return 0;
 780         else if (ret > 0) {
 781                 btrfs_release_path(path);
 782                 return 0;
 783         }
 784
 785         ret = -1;
 786
 787         leaf = path->nodes[0];
 788         header = btrfs_item_ptr(leaf, path->slots[0],
 789                                 struct btrfs_free_space_header);
 790         num_entries = btrfs_free_space_entries(leaf, header);
 791         num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
 792         generation = btrfs_free_space_generation(leaf, header);
 793         btrfs_release_path(path);
 794
 795         if (!BTRFS_I(inode)->generation) {
 796                 btrfs_info(fs_info,
 797                            "the free space cache file (%llu) is invalid, skip it",
 798                            offset);
 799                 return 0;
 800         }
 801
 802         if (BTRFS_I(inode)->generation != generation) {
 803                 btrfs_err(fs_info,
 804                           "free space inode generation (%llu) did not match free space cache generation (%llu)",
 805                           BTRFS_I(inode)->generation, generation);
 806                 return 0;
 807         }
 808
 809         if (!num_entries)
 810                 return 0;
 811
 812         ret = io_ctl_init(&io_ctl, inode, 0);
 813         if (ret)
 814                 return ret;
 815
 816         readahead_cache(inode);
 817
 818         ret = io_ctl_prepare_pages(&io_ctl, true);
 819         if (ret)
 820                 goto out;
 821
 822         ret = io_ctl_check_crc(&io_ctl, 0);
 823         if (ret)
 824                 goto free_cache;
 825
 826         ret = io_ctl_check_generation(&io_ctl, generation);
 827         if (ret)
 828                 goto free_cache;
 829
 830         while (num_entries) {
 831                 e = kmem_cache_zalloc(btrfs_free_space_cachep,
 832                                       GFP_NOFS);
 833                 if (!e) {
 834                         ret = -ENOMEM;
 835                         goto free_cache;
 836                 }
 837
 838                 ret = io_ctl_read_entry(&io_ctl, e, &type);
 839                 if (ret) {
 840                         kmem_cache_free(btrfs_free_space_cachep, e);
 841                         goto free_cache;
 842                 }
 843
 844                 if (!e->bytes) {
 845                         ret = -1;
 846                         kmem_cache_free(btrfs_free_space_cachep, e);
 847                         goto free_cache;
 848                 }
 849
 850                 if (type == BTRFS_FREE_SPACE_EXTENT) {
 851                         spin_lock(&ctl->tree_lock);
 852                         ret = link_free_space(ctl, e);
 853                         spin_unlock(&ctl->tree_lock);
 854                         if (ret) {
 855                                 btrfs_err(fs_info,
 856                                         "Duplicate entries in free space cache, dumping");
 857                                 kmem_cache_free(btrfs_free_space_cachep, e);
 858                                 goto free_cache;
 859                         }
 860                 } else {
 861                         ASSERT(num_bitmaps);
 862                         num_bitmaps--;
 863                         e->bitmap = kmem_cache_zalloc(
 864                                         btrfs_free_space_bitmap_cachep, GFP_NOFS);
 865                         if (!e->bitmap) {
 866                                 ret = -ENOMEM;
 867                                 kmem_cache_free(
 868                                         btrfs_free_space_cachep, e);
 869                                 goto free_cache;
 870                         }
 871                         spin_lock(&ctl->tree_lock);
 872                         ret = link_free_space(ctl, e);
 873                         if (ret) {
 874                                 spin_unlock(&ctl->tree_lock);
 875                                 btrfs_err(fs_info,
 876                                         "Duplicate entries in free space cache, dumping");
 877                                 kmem_cache_free(btrfs_free_space_cachep, e);
 878                                 goto free_cache;
 879                         }
 880                         ctl->total_bitmaps++;
 881                         recalculate_thresholds(ctl);
 882                         spin_unlock(&ctl->tree_lock);
 883                         list_add_tail(&e->list, &bitmaps);
 884                 }
 885
 886                 num_entries--;
 887         }
 888
 889         io_ctl_unmap_page(&io_ctl);
 890
 891         /*
 892          * We add the bitmaps at the end of the entries in order that
 893          * the bitmap entries are added to the cache.
 894          */
 895         list_for_each_entry_safe(e, n, &bitmaps, list) {
 896                 list_del_init(&e->list);
 897                 ret = io_ctl_read_bitmap(&io_ctl, e);
 898                 if (ret)
 899                         goto free_cache;
 900         }
 901
 902         io_ctl_drop_pages(&io_ctl);
 903         ret = 1;
 904 out:
 905         io_ctl_free(&io_ctl);
 906         return ret;
 907 free_cache:
 908         io_ctl_drop_pages(&io_ctl);
 909
 910         spin_lock(&ctl->tree_lock);
 911         __btrfs_remove_free_space_cache(ctl);
 912         spin_unlock(&ctl->tree_lock);
 913         goto out;
 914 }
 915
 916 static int copy_free_space_cache(struct btrfs_block_group *block_group,
 917                                  struct btrfs_free_space_ctl *ctl)
 918 {
 919         struct btrfs_free_space *info;
 920         struct rb_node *n;
 921         int ret = 0;
 922
 923         while (!ret && (n = rb_first(&ctl->free_space_offset)) != NULL) {
 924                 info = rb_entry(n, struct btrfs_free_space, offset_index);
 925                 if (!info->bitmap) {
 926                         unlink_free_space(ctl, info, true);
 927                         ret = btrfs_add_free_space(block_group, info->offset,
 928                                                    info->bytes);
 929                         kmem_cache_free(btrfs_free_space_cachep, info);
 930                 } else {
 931                         u64 offset = info->offset;
 932                         u64 bytes = ctl->unit;
 933
 934                         while (search_bitmap(ctl, info, &offset, &bytes,
 935                                              false) == 0) {
 936                                 ret = btrfs_add_free_space(block_group, offset,
 937                                                            bytes);
 938                                 if (ret)
 939                                         break;
 940                                 bitmap_clear_bits(ctl, info, offset, bytes, true);
 941                                 offset = info->offset;
 942                                 bytes = ctl->unit;
 943                         }
 944                         free_bitmap(ctl, info);
 945                 }
 946                 cond_resched();
 947         }
 948         return ret;
 949 }
 950
 951 static struct lock_class_key btrfs_free_space_inode_key;
 952
 953 int load_free_space_cache(struct btrfs_block_group *block_group)
 954 {
 955         struct btrfs_fs_info *fs_info = block_group->fs_info;
 956         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 957         struct btrfs_free_space_ctl tmp_ctl = {};
 958         struct inode *inode;
 959         struct btrfs_path *path;
 960         int ret = 0;
 961         bool matched;
 962         u64 used = block_group->used;
 963
 964         /*
 965          * Because we could potentially discard our loaded free space, we want
 966          * to load everything into a temporary structure first, and then if it's
 967          * valid copy it all into the actual free space ctl.
 968          */
 969         btrfs_init_free_space_ctl(block_group, &tmp_ctl);
 970
 971         /*
 972          * If this block group has been marked to be cleared for one reason or
 973          * another then we can't trust the on disk cache, so just return.
 974          */
 975         spin_lock(&block_group->lock);
 976         if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
 977                 spin_unlock(&block_group->lock);
 978                 return 0;
 979         }
 980         spin_unlock(&block_group->lock);
 981
 982         path = btrfs_alloc_path();
 983         if (!path)
 984                 return 0;
 985         path->search_commit_root = 1;
 986         path->skip_locking = 1;
 987
 988         /*
 989          * We must pass a path with search_commit_root set to btrfs_iget in
 990          * order to avoid a deadlock when allocating extents for the tree root.
 991          *
 992          * When we are COWing an extent buffer from the tree root, when looking
 993          * for a free extent, at extent-tree.c:find_free_extent(), we can find
 994          * block group without its free space cache loaded. When we find one
 995          * we must load its space cache which requires reading its free space
 996          * cache's inode item from the root tree. If this inode item is located
 997          * in the same leaf that we started COWing before, then we end up in
 998          * deadlock on the extent buffer (trying to read lock it when we
 999          * previously write locked it).
1000          *
1001          * It's safe to read the inode item using the commit root because
1002          * block groups, once loaded, stay in memory forever (until they are
1003          * removed) as well as their space caches once loaded. New block groups
1004          * once created get their ->cached field set to BTRFS_CACHE_FINISHED so
1005          * we will never try to read their inode item while the fs is mounted.
1006          */
1007         inode = lookup_free_space_inode(block_group, path);
1008         if (IS_ERR(inode)) {
1009                 btrfs_free_path(path);
1010                 return 0;
1011         }
1012
1013         /* We may have converted the inode and made the cache invalid. */
1014         spin_lock(&block_group->lock);
1015         if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
1016                 spin_unlock(&block_group->lock);
1017                 btrfs_free_path(path);
1018                 goto out;
1019         }
1020         spin_unlock(&block_group->lock);
1021
1022         /*
1023          * Reinitialize the class of struct inode's mapping->invalidate_lock for
1024          * free space inodes to prevent false positives related to locks for normal
1025          * inodes.
1026          */
1027         lockdep_set_class(&(&inode->i_data)->invalidate_lock,
1028                           &btrfs_free_space_inode_key);
1029
1030         ret = __load_free_space_cache(fs_info->tree_root, inode, &tmp_ctl,
1031                                       path, block_group->start);
1032         btrfs_free_path(path);
1033         if (ret <= 0)
1034                 goto out;
1035
1036         matched = (tmp_ctl.free_space == (block_group->length - used -
1037                                           block_group->bytes_super));
1038
1039         if (matched) {
1040                 ret = copy_free_space_cache(block_group, &tmp_ctl);
1041                 /*
1042                  * ret == 1 means we successfully loaded the free space cache,
1043                  * so we need to re-set it here.
1044                  */
1045                 if (ret == 0)
1046                         ret = 1;
1047         } else {
1048                 /*
1049                  * We need to call the _locked variant so we don't try to update
1050                  * the discard counters.
1051                  */
1052                 spin_lock(&tmp_ctl.tree_lock);
1053                 __btrfs_remove_free_space_cache(&tmp_ctl);
1054                 spin_unlock(&tmp_ctl.tree_lock);
1055                 btrfs_warn(fs_info,
1056                            "block group %llu has wrong amount of free space",
1057                            block_group->start);
1058                 ret = -1;
1059         }
1060 out:
1061         if (ret < 0) {
1062                 /* This cache is bogus, make sure it gets cleared */
1063                 spin_lock(&block_group->lock);
1064                 block_group->disk_cache_state = BTRFS_DC_CLEAR;
1065                 spin_unlock(&block_group->lock);
1066                 ret = 0;
1067
1068                 btrfs_warn(fs_info,
1069                            "failed to load free space cache for block group %llu, rebuilding it now",
1070                            block_group->start);
1071         }
1072
1073         spin_lock(&ctl->tree_lock);
1074         btrfs_discard_update_discardable(block_group);
1075         spin_unlock(&ctl->tree_lock);
1076         iput(inode);
1077         return ret;
1078 }
1079
1080 static noinline_for_stack
1081 int write_cache_extent_entries(struct btrfs_io_ctl *io_ctl,
1082                               struct btrfs_free_space_ctl *ctl,
1083                               struct btrfs_block_group *block_group,
1084                               int *entries, int *bitmaps,
1085                               struct list_head *bitmap_list)
1086 {
1087         int ret;
1088         struct btrfs_free_cluster *cluster = NULL;
1089         struct btrfs_free_cluster *cluster_locked = NULL;
1090         struct rb_node *node = rb_first(&ctl->free_space_offset);
1091         struct btrfs_trim_range *trim_entry;
1092
1093         /* Get the cluster for this block_group if it exists */
1094         if (block_group && !list_empty(&block_group->cluster_list)) {
1095                 cluster = list_entry(block_group->cluster_list.next,
1096                                      struct btrfs_free_cluster,
1097                                      block_group_list);
1098         }
1099
1100         if (!node && cluster) {
1101                 cluster_locked = cluster;
1102                 spin_lock(&cluster_locked->lock);
1103                 node = rb_first(&cluster->root);
1104                 cluster = NULL;
1105         }
1106
1107         /* Write out the extent entries */
1108         while (node) {
1109                 struct btrfs_free_space *e;
1110
1111                 e = rb_entry(node, struct btrfs_free_space, offset_index);
1112                 *entries += 1;
1113
1114                 ret = io_ctl_add_entry(io_ctl, e->offset, e->bytes,
1115                                        e->bitmap);
1116                 if (ret)
1117                         goto fail;
1118
1119                 if (e->bitmap) {
1120                         list_add_tail(&e->list, bitmap_list);
1121                         *bitmaps += 1;
1122                 }
1123                 node = rb_next(node);
1124                 if (!node && cluster) {
1125                         node = rb_first(&cluster->root);
1126                         cluster_locked = cluster;
1127                         spin_lock(&cluster_locked->lock);
1128                         cluster = NULL;
1129                 }
1130         }
1131         if (cluster_locked) {
1132                 spin_unlock(&cluster_locked->lock);
1133                 cluster_locked = NULL;
1134         }
1135
1136         /*
1137          * Make sure we don't miss any range that was removed from our rbtree
1138          * because trimming is running. Otherwise after a umount+mount (or crash
1139          * after committing the transaction) we would leak free space and get
1140          * an inconsistent free space cache report from fsck.
1141          */
1142         list_for_each_entry(trim_entry, &ctl->trimming_ranges, list) {
1143                 ret = io_ctl_add_entry(io_ctl, trim_entry->start,
1144                                        trim_entry->bytes, NULL);
1145                 if (ret)
1146                         goto fail;
1147                 *entries += 1;
1148         }
1149
1150         return 0;
1151 fail:
1152         if (cluster_locked)
1153                 spin_unlock(&cluster_locked->lock);
1154         return -ENOSPC;
1155 }
1156
1157 static noinline_for_stack int
1158 update_cache_item(struct btrfs_trans_handle *trans,
1159                   struct btrfs_root *root,
1160                   struct inode *inode,
1161                   struct btrfs_path *path, u64 offset,
1162                   int entries, int bitmaps)
1163 {
1164         struct btrfs_key key;
1165         struct btrfs_free_space_header *header;
1166         struct extent_buffer *leaf;
1167         int ret;
1168
1169         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
1170         key.offset = offset;
1171         key.type = 0;
1172
1173         ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
1174         if (ret < 0) {
1175                 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
1176                                  EXTENT_DELALLOC, NULL);
1177                 goto fail;
1178         }
1179         leaf = path->nodes[0];
1180         if (ret > 0) {
1181                 struct btrfs_key found_key;
1182                 ASSERT(path->slots[0]);
1183                 path->slots[0]--;
1184                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1185                 if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
1186                     found_key.offset != offset) {
1187                         clear_extent_bit(&BTRFS_I(inode)->io_tree, 0,
1188                                          inode->i_size - 1, EXTENT_DELALLOC,
1189                                          NULL);
1190                         btrfs_release_path(path);
1191                         goto fail;
1192                 }
1193         }
1194
1195         BTRFS_I(inode)->generation = trans->transid;
1196         header = btrfs_item_ptr(leaf, path->slots[0],
1197                                 struct btrfs_free_space_header);
1198         btrfs_set_free_space_entries(leaf, header, entries);
1199         btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
1200         btrfs_set_free_space_generation(leaf, header, trans->transid);
1201         btrfs_mark_buffer_dirty(leaf);
1202         btrfs_release_path(path);
1203
1204         return 0;
1205
1206 fail:
1207         return -1;
1208 }
1209
1210 static noinline_for_stack int write_pinned_extent_entries(
1211                             struct btrfs_trans_handle *trans,
1212                             struct btrfs_block_group *block_group,
1213                             struct btrfs_io_ctl *io_ctl,
1214                             int *entries)
1215 {
1216         u64 start, extent_start, extent_end, len;
1217         struct extent_io_tree *unpin = NULL;
1218         int ret;
1219
1220         if (!block_group)
1221                 return 0;
1222
1223         /*
1224          * We want to add any pinned extents to our free space cache
1225          * so we don't leak the space
1226          *
1227          * We shouldn't have switched the pinned extents yet so this is the
1228          * right one
1229          */
1230         unpin = &trans->transaction->pinned_extents;
1231
1232         start = block_group->start;
1233
1234         while (start < block_group->start + block_group->length) {
1235                 ret = find_first_extent_bit(unpin, start,
1236                                             &extent_start, &extent_end,
1237                                             EXTENT_DIRTY, NULL);
1238                 if (ret)
1239                         return 0;
1240
1241                 /* This pinned extent is out of our range */
1242                 if (extent_start >= block_group->start + block_group->length)
1243                         return 0;
1244
1245                 extent_start = max(extent_start, start);
1246                 extent_end = min(block_group->start + block_group->length,
1247                                  extent_end + 1);
1248                 len = extent_end - extent_start;
1249
1250                 *entries += 1;
1251                 ret = io_ctl_add_entry(io_ctl, extent_start, len, NULL);
1252                 if (ret)
1253                         return -ENOSPC;
1254
1255                 start = extent_end;
1256         }
1257
1258         return 0;
1259 }
1260
1261 static noinline_for_stack int
1262 write_bitmap_entries(struct btrfs_io_ctl *io_ctl, struct list_head *bitmap_list)
1263 {
1264         struct btrfs_free_space *entry, *next;
1265         int ret;
1266
1267         /* Write out the bitmaps */
1268         list_for_each_entry_safe(entry, next, bitmap_list, list) {
1269                 ret = io_ctl_add_bitmap(io_ctl, entry->bitmap);
1270                 if (ret)
1271                         return -ENOSPC;
1272                 list_del_init(&entry->list);
1273         }
1274
1275         return 0;
1276 }
1277
1278 static int flush_dirty_cache(struct inode *inode)
1279 {
1280         int ret;
1281
1282         ret = btrfs_wait_ordered_range(inode, 0, (u64)-1);
1283         if (ret)
1284                 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
1285                                  EXTENT_DELALLOC, NULL);
1286
1287         return ret;
1288 }
1289
1290 static void noinline_for_stack
1291 cleanup_bitmap_list(struct list_head *bitmap_list)
1292 {
1293         struct btrfs_free_space *entry, *next;
1294
1295         list_for_each_entry_safe(entry, next, bitmap_list, list)
1296                 list_del_init(&entry->list);
1297 }
1298
1299 static void noinline_for_stack
1300 cleanup_write_cache_enospc(struct inode *inode,
1301                            struct btrfs_io_ctl *io_ctl,
1302                            struct extent_state **cached_state)
1303 {
1304         io_ctl_drop_pages(io_ctl);
1305         unlock_extent(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
1306                       cached_state);
1307 }
1308
1309 static int __btrfs_wait_cache_io(struct btrfs_root *root,
1310                                  struct btrfs_trans_handle *trans,
1311                                  struct btrfs_block_group *block_group,
1312                                  struct btrfs_io_ctl *io_ctl,
1313                                  struct btrfs_path *path, u64 offset)
1314 {
1315         int ret;
1316         struct inode *inode = io_ctl->inode;
1317
1318         if (!inode)
1319                 return 0;
1320
1321         /* Flush the dirty pages in the cache file. */
1322         ret = flush_dirty_cache(inode);
1323         if (ret)
1324                 goto out;
1325
1326         /* Update the cache item to tell everyone this cache file is valid. */
1327         ret = update_cache_item(trans, root, inode, path, offset,
1328                                 io_ctl->entries, io_ctl->bitmaps);
1329 out:
1330         if (ret) {
1331                 invalidate_inode_pages2(inode->i_mapping);
1332                 BTRFS_I(inode)->generation = 0;
1333                 if (block_group)
1334                         btrfs_debug(root->fs_info,
1335           "failed to write free space cache for block group %llu error %d",
1336                                   block_group->start, ret);
1337         }
1338         btrfs_update_inode(trans, root, BTRFS_I(inode));
1339
1340         if (block_group) {
1341                 /* the dirty list is protected by the dirty_bgs_lock */
1342                 spin_lock(&trans->transaction->dirty_bgs_lock);
1343
1344                 /* the disk_cache_state is protected by the block group lock */
1345                 spin_lock(&block_group->lock);
1346
1347                 /*
1348                  * only mark this as written if we didn't get put back on
1349                  * the dirty list while waiting for IO.   Otherwise our
1350                  * cache state won't be right, and we won't get written again
1351                  */
1352                 if (!ret && list_empty(&block_group->dirty_list))
1353                         block_group->disk_cache_state = BTRFS_DC_WRITTEN;
1354                 else if (ret)
1355                         block_group->disk_cache_state = BTRFS_DC_ERROR;
1356
1357                 spin_unlock(&block_group->lock);
1358                 spin_unlock(&trans->transaction->dirty_bgs_lock);
1359                 io_ctl->inode = NULL;
1360                 iput(inode);
1361         }
1362
1363         return ret;
1364
1365 }
1366
1367 int btrfs_wait_cache_io(struct btrfs_trans_handle *trans,
1368                         struct btrfs_block_group *block_group,
1369                         struct btrfs_path *path)
1370 {
1371         return __btrfs_wait_cache_io(block_group->fs_info->tree_root, trans,
1372                                      block_group, &block_group->io_ctl,
1373                                      path, block_group->start);
1374 }
1375
1376 /*
1377  * Write out cached info to an inode.
1378  *
1379  * @root:        root the inode belongs to
1380  * @inode:       freespace inode we are writing out
1381  * @ctl:         free space cache we are going to write out
1382  * @block_group: block_group for this cache if it belongs to a block_group
1383  * @io_ctl:      holds context for the io
1384  * @trans:       the trans handle
1385  *
1386  * This function writes out a free space cache struct to disk for quick recovery
1387  * on mount.  This will return 0 if it was successful in writing the cache out,
1388  * or an errno if it was not.
1389  */
1390 static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
1391                                    struct btrfs_free_space_ctl *ctl,
1392                                    struct btrfs_block_group *block_group,
1393                                    struct btrfs_io_ctl *io_ctl,
1394                                    struct btrfs_trans_handle *trans)
1395 {
1396         struct extent_state *cached_state = NULL;
1397         LIST_HEAD(bitmap_list);
1398         int entries = 0;
1399         int bitmaps = 0;
1400         int ret;
1401         int must_iput = 0;
1402
1403         if (!i_size_read(inode))
1404                 return -EIO;
1405
1406         WARN_ON(io_ctl->pages);
1407         ret = io_ctl_init(io_ctl, inode, 1);
1408         if (ret)
1409                 return ret;
1410
1411         if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA)) {
1412                 down_write(&block_group->data_rwsem);
1413                 spin_lock(&block_group->lock);
1414                 if (block_group->delalloc_bytes) {
1415                         block_group->disk_cache_state = BTRFS_DC_WRITTEN;
1416                         spin_unlock(&block_group->lock);
1417                         up_write(&block_group->data_rwsem);
1418                         BTRFS_I(inode)->generation = 0;
1419                         ret = 0;
1420                         must_iput = 1;
1421                         goto out;
1422                 }
1423                 spin_unlock(&block_group->lock);
1424         }
1425
1426         /* Lock all pages first so we can lock the extent safely. */
1427         ret = io_ctl_prepare_pages(io_ctl, false);
1428         if (ret)
1429                 goto out_unlock;
1430
1431         lock_extent(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
1432                     &cached_state);
1433
1434         io_ctl_set_generation(io_ctl, trans->transid);
1435
1436         mutex_lock(&ctl->cache_writeout_mutex);
1437         /* Write out the extent entries in the free space cache */
1438         spin_lock(&ctl->tree_lock);
1439         ret = write_cache_extent_entries(io_ctl, ctl,
1440                                          block_group, &entries, &bitmaps,
1441                                          &bitmap_list);
1442         if (ret)
1443                 goto out_nospc_locked;
1444
1445         /*
1446          * Some spaces that are freed in the current transaction are pinned,
1447          * they will be added into free space cache after the transaction is
1448          * committed, we shouldn't lose them.
1449          *
1450          * If this changes while we are working we'll get added back to
1451          * the dirty list and redo it.  No locking needed
1452          */
1453         ret = write_pinned_extent_entries(trans, block_group, io_ctl, &entries);
1454         if (ret)
1455                 goto out_nospc_locked;
1456
1457         /*
1458          * At last, we write out all the bitmaps and keep cache_writeout_mutex
1459          * locked while doing it because a concurrent trim can be manipulating
1460          * or freeing the bitmap.
1461          */
1462         ret = write_bitmap_entries(io_ctl, &bitmap_list);
1463         spin_unlock(&ctl->tree_lock);
1464         mutex_unlock(&ctl->cache_writeout_mutex);
1465         if (ret)
1466                 goto out_nospc;
1467
1468         /* Zero out the rest of the pages just to make sure */
1469         io_ctl_zero_remaining_pages(io_ctl);
1470
1471         /* Everything is written out, now we dirty the pages in the file. */
1472         ret = btrfs_dirty_pages(BTRFS_I(inode), io_ctl->pages,
1473                                 io_ctl->num_pages, 0, i_size_read(inode),
1474                                 &cached_state, false);
1475         if (ret)
1476                 goto out_nospc;
1477
1478         if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
1479                 up_write(&block_group->data_rwsem);
1480         /*
1481          * Release the pages and unlock the extent, we will flush
1482          * them out later
1483          */
1484         io_ctl_drop_pages(io_ctl);
1485         io_ctl_free(io_ctl);
1486
1487         unlock_extent(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
1488                       &cached_state);
1489
1490         /*
1491          * at this point the pages are under IO and we're happy,
1492          * The caller is responsible for waiting on them and updating
1493          * the cache and the inode
1494          */
1495         io_ctl->entries = entries;
1496         io_ctl->bitmaps = bitmaps;
1497
1498         ret = btrfs_fdatawrite_range(inode, 0, (u64)-1);
1499         if (ret)
1500                 goto out;
1501
1502         return 0;
1503
1504 out_nospc_locked:
1505         cleanup_bitmap_list(&bitmap_list);
1506         spin_unlock(&ctl->tree_lock);
1507         mutex_unlock(&ctl->cache_writeout_mutex);
1508
1509 out_nospc:
1510         cleanup_write_cache_enospc(inode, io_ctl, &cached_state);
1511
1512 out_unlock:
1513         if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
1514                 up_write(&block_group->data_rwsem);
1515
1516 out:
1517         io_ctl->inode = NULL;
1518         io_ctl_free(io_ctl);
1519         if (ret) {
1520                 invalidate_inode_pages2(inode->i_mapping);
1521                 BTRFS_I(inode)->generation = 0;
1522         }
1523         btrfs_update_inode(trans, root, BTRFS_I(inode));
1524         if (must_iput)
1525                 iput(inode);
1526         return ret;
1527 }
1528
1529 int btrfs_write_out_cache(struct btrfs_trans_handle *trans,
1530                           struct btrfs_block_group *block_group,
1531                           struct btrfs_path *path)
1532 {
1533         struct btrfs_fs_info *fs_info = trans->fs_info;
1534         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1535         struct inode *inode;
1536         int ret = 0;
1537
1538         spin_lock(&block_group->lock);
1539         if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
1540                 spin_unlock(&block_group->lock);
1541                 return 0;
1542         }
1543         spin_unlock(&block_group->lock);
1544
1545         inode = lookup_free_space_inode(block_group, path);
1546         if (IS_ERR(inode))
1547                 return 0;
1548
1549         ret = __btrfs_write_out_cache(fs_info->tree_root, inode, ctl,
1550                                 block_group, &block_group->io_ctl, trans);
1551         if (ret) {
1552                 btrfs_debug(fs_info,
1553           "failed to write free space cache for block group %llu error %d",
1554                           block_group->start, ret);
1555                 spin_lock(&block_group->lock);
1556                 block_group->disk_cache_state = BTRFS_DC_ERROR;
1557                 spin_unlock(&block_group->lock);
1558
1559                 block_group->io_ctl.inode = NULL;
1560                 iput(inode);
1561         }
1562
1563         /*
1564          * if ret == 0 the caller is expected to call btrfs_wait_cache_io
1565          * to wait for IO and put the inode
1566          */
1567
1568         return ret;
1569 }
1570
1571 static inline unsigned long offset_to_bit(u64 bitmap_start, u32 unit,
1572                                           u64 offset)
1573 {
1574         ASSERT(offset >= bitmap_start);
1575         offset -= bitmap_start;
1576         return (unsigned long)(div_u64(offset, unit));
1577 }
1578
1579 static inline unsigned long bytes_to_bits(u64 bytes, u32 unit)
1580 {
1581         return (unsigned long)(div_u64(bytes, unit));
1582 }
1583
1584 static inline u64 offset_to_bitmap(struct btrfs_free_space_ctl *ctl,
1585                                    u64 offset)
1586 {
1587         u64 bitmap_start;
1588         u64 bytes_per_bitmap;
1589
1590         bytes_per_bitmap = BITS_PER_BITMAP * ctl->unit;
1591         bitmap_start = offset - ctl->start;
1592         bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
1593         bitmap_start *= bytes_per_bitmap;
1594         bitmap_start += ctl->start;
1595
1596         return bitmap_start;
1597 }
1598
1599 static int tree_insert_offset(struct rb_root *root, u64 offset,
1600                               struct rb_node *node, int bitmap)
1601 {
1602         struct rb_node **p = &root->rb_node;
1603         struct rb_node *parent = NULL;
1604         struct btrfs_free_space *info;
1605
1606         while (*p) {
1607                 parent = *p;
1608                 info = rb_entry(parent, struct btrfs_free_space, offset_index);
1609
1610                 if (offset < info->offset) {
1611                         p = &(*p)->rb_left;
1612                 } else if (offset > info->offset) {
1613                         p = &(*p)->rb_right;
1614                 } else {
1615                         /*
1616                          * we could have a bitmap entry and an extent entry
1617                          * share the same offset.  If this is the case, we want
1618                          * the extent entry to always be found first if we do a
1619                          * linear search through the tree, since we want to have
1620                          * the quickest allocation time, and allocating from an
1621                          * extent is faster than allocating from a bitmap.  So
1622                          * if we're inserting a bitmap and we find an entry at
1623                          * this offset, we want to go right, or after this entry
1624                          * logically.  If we are inserting an extent and we've
1625                          * found a bitmap, we want to go left, or before
1626                          * logically.
1627                          */
1628                         if (bitmap) {
1629                                 if (info->bitmap) {
1630                                         WARN_ON_ONCE(1);
1631                                         return -EEXIST;
1632                                 }
1633                                 p = &(*p)->rb_right;
1634                         } else {
1635                                 if (!info->bitmap) {
1636                                         WARN_ON_ONCE(1);
1637                                         return -EEXIST;
1638                                 }
1639                                 p = &(*p)->rb_left;
1640                         }
1641                 }
1642         }
1643
1644         rb_link_node(node, parent, p);
1645         rb_insert_color(node, root);
1646
1647         return 0;
1648 }
1649
1650 /*
1651  * This is a little subtle.  We *only* have ->max_extent_size set if we actually
1652  * searched through the bitmap and figured out the largest ->max_extent_size,
1653  * otherwise it's 0.  In the case that it's 0 we don't want to tell the
1654  * allocator the wrong thing, we want to use the actual real max_extent_size
1655  * we've found already if it's larger, or we want to use ->bytes.
1656  *
1657  * This matters because find_free_space() will skip entries who's ->bytes is
1658  * less than the required bytes.  So if we didn't search down this bitmap, we
1659  * may pick some previous entry that has a smaller ->max_extent_size than we
1660  * have.  For example, assume we have two entries, one that has
1661  * ->max_extent_size set to 4K and ->bytes set to 1M.  A second entry hasn't set
1662  * ->max_extent_size yet, has ->bytes set to 8K and it's contiguous.  We will
1663  *  call into find_free_space(), and return with max_extent_size == 4K, because
1664  *  that first bitmap entry had ->max_extent_size set, but the second one did
1665  *  not.  If instead we returned 8K we'd come in searching for 8K, and find the
1666  *  8K contiguous range.
1667  *
1668  *  Consider the other case, we have 2 8K chunks in that second entry and still
1669  *  don't have ->max_extent_size set.  We'll return 16K, and the next time the
1670  *  allocator comes in it'll fully search our second bitmap, and this time it'll
1671  *  get an uptodate value of 8K as the maximum chunk size.  Then we'll get the
1672  *  right allocation the next loop through.
1673  */
1674 static inline u64 get_max_extent_size(const struct btrfs_free_space *entry)
1675 {
1676         if (entry->bitmap && entry->max_extent_size)
1677                 return entry->max_extent_size;
1678         return entry->bytes;
1679 }
1680
1681 /*
1682  * We want the largest entry to be leftmost, so this is inverted from what you'd
1683  * normally expect.
1684  */
1685 static bool entry_less(struct rb_node *node, const struct rb_node *parent)
1686 {
1687         const struct btrfs_free_space *entry, *exist;
1688
1689         entry = rb_entry(node, struct btrfs_free_space, bytes_index);
1690         exist = rb_entry(parent, struct btrfs_free_space, bytes_index);
1691         return get_max_extent_size(exist) < get_max_extent_size(entry);
1692 }
1693
1694 /*
1695  * searches the tree for the given offset.
1696  *
1697  * fuzzy - If this is set, then we are trying to make an allocation, and we just
1698  * want a section that has at least bytes size and comes at or after the given
1699  * offset.
1700  */
1701 static struct btrfs_free_space *
1702 tree_search_offset(struct btrfs_free_space_ctl *ctl,
1703                    u64 offset, int bitmap_only, int fuzzy)
1704 {
1705         struct rb_node *n = ctl->free_space_offset.rb_node;
1706         struct btrfs_free_space *entry = NULL, *prev = NULL;
1707
1708         /* find entry that is closest to the 'offset' */
1709         while (n) {
1710                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1711                 prev = entry;
1712
1713                 if (offset < entry->offset)
1714                         n = n->rb_left;
1715                 else if (offset > entry->offset)
1716                         n = n->rb_right;
1717                 else
1718                         break;
1719
1720                 entry = NULL;
1721         }
1722
1723         if (bitmap_only) {
1724                 if (!entry)
1725                         return NULL;
1726                 if (entry->bitmap)
1727                         return entry;
1728
1729                 /*
1730                  * bitmap entry and extent entry may share same offset,
1731                  * in that case, bitmap entry comes after extent entry.
1732                  */
1733                 n = rb_next(n);
1734                 if (!n)
1735                         return NULL;
1736                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1737                 if (entry->offset != offset)
1738                         return NULL;
1739
1740                 WARN_ON(!entry->bitmap);
1741                 return entry;
1742         } else if (entry) {
1743                 if (entry->bitmap) {
1744                         /*
1745                          * if previous extent entry covers the offset,
1746                          * we should return it instead of the bitmap entry
1747                          */
1748                         n = rb_prev(&entry->offset_index);
1749                         if (n) {
1750                                 prev = rb_entry(n, struct btrfs_free_space,
1751                                                 offset_index);
1752                                 if (!prev->bitmap &&
1753                                     prev->offset + prev->bytes > offset)
1754                                         entry = prev;
1755                         }
1756                 }
1757                 return entry;
1758         }
1759
1760         if (!prev)
1761                 return NULL;
1762
1763         /* find last entry before the 'offset' */
1764         entry = prev;
1765         if (entry->offset > offset) {
1766                 n = rb_prev(&entry->offset_index);
1767                 if (n) {
1768                         entry = rb_entry(n, struct btrfs_free_space,
1769                                         offset_index);
1770                         ASSERT(entry->offset <= offset);
1771                 } else {
1772                         if (fuzzy)
1773                                 return entry;
1774                         else
1775                                 return NULL;
1776                 }
1777         }
1778
1779         if (entry->bitmap) {
1780                 n = rb_prev(&entry->offset_index);
1781                 if (n) {
1782                         prev = rb_entry(n, struct btrfs_free_space,
1783                                         offset_index);
1784                         if (!prev->bitmap &&
1785                             prev->offset + prev->bytes > offset)
1786                                 return prev;
1787                 }
1788                 if (entry->offset + BITS_PER_BITMAP * ctl->unit > offset)
1789                         return entry;
1790         } else if (entry->offset + entry->bytes > offset)
1791                 return entry;
1792
1793         if (!fuzzy)
1794                 return NULL;
1795
1796         while (1) {
1797                 n = rb_next(&entry->offset_index);
1798                 if (!n)
1799                         return NULL;
1800                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1801                 if (entry->bitmap) {
1802                         if (entry->offset + BITS_PER_BITMAP *
1803                             ctl->unit > offset)
1804                                 break;
1805                 } else {
1806                         if (entry->offset + entry->bytes > offset)
1807                                 break;
1808                 }
1809         }
1810         return entry;
1811 }
1812
1813 static inline void unlink_free_space(struct btrfs_free_space_ctl *ctl,
1814                                      struct btrfs_free_space *info,
1815                                      bool update_stat)
1816 {
1817         rb_erase(&info->offset_index, &ctl->free_space_offset);
1818         rb_erase_cached(&info->bytes_index, &ctl->free_space_bytes);
1819         ctl->free_extents--;
1820
1821         if (!info->bitmap && !btrfs_free_space_trimmed(info)) {
1822                 ctl->discardable_extents[BTRFS_STAT_CURR]--;
1823                 ctl->discardable_bytes[BTRFS_STAT_CURR] -= info->bytes;
1824         }
1825
1826         if (update_stat)
1827                 ctl->free_space -= info->bytes;
1828 }
1829
1830 static int link_free_space(struct btrfs_free_space_ctl *ctl,
1831                            struct btrfs_free_space *info)
1832 {
1833         int ret = 0;
1834
1835         ASSERT(info->bytes || info->bitmap);
1836         ret = tree_insert_offset(&ctl->free_space_offset, info->offset,
1837                                  &info->offset_index, (info->bitmap != NULL));
1838         if (ret)
1839                 return ret;
1840
1841         rb_add_cached(&info->bytes_index, &ctl->free_space_bytes, entry_less);
1842
1843         if (!info->bitmap && !btrfs_free_space_trimmed(info)) {
1844                 ctl->discardable_extents[BTRFS_STAT_CURR]++;
1845                 ctl->discardable_bytes[BTRFS_STAT_CURR] += info->bytes;
1846         }
1847
1848         ctl->free_space += info->bytes;
1849         ctl->free_extents++;
1850         return ret;
1851 }
1852
1853 static void relink_bitmap_entry(struct btrfs_free_space_ctl *ctl,
1854                                 struct btrfs_free_space *info)
1855 {
1856         ASSERT(info->bitmap);
1857
1858         /*
1859          * If our entry is empty it's because we're on a cluster and we don't
1860          * want to re-link it into our ctl bytes index.
1861          */
1862         if (RB_EMPTY_NODE(&info->bytes_index))
1863                 return;
1864
1865         rb_erase_cached(&info->bytes_index, &ctl->free_space_bytes);
1866         rb_add_cached(&info->bytes_index, &ctl->free_space_bytes, entry_less);
1867 }
1868
1869 static inline void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1870                                      struct btrfs_free_space *info,
1871                                      u64 offset, u64 bytes, bool update_stat)
1872 {
1873         unsigned long start, count, end;
1874         int extent_delta = -1;
1875
1876         start = offset_to_bit(info->offset, ctl->unit, offset);
1877         count = bytes_to_bits(bytes, ctl->unit);
1878         end = start + count;
1879         ASSERT(end <= BITS_PER_BITMAP);
1880
1881         bitmap_clear(info->bitmap, start, count);
1882
1883         info->bytes -= bytes;
1884         if (info->max_extent_size > ctl->unit)
1885                 info->max_extent_size = 0;
1886
1887         relink_bitmap_entry(ctl, info);
1888
1889         if (start && test_bit(start - 1, info->bitmap))
1890                 extent_delta++;
1891
1892         if (end < BITS_PER_BITMAP && test_bit(end, info->bitmap))
1893                 extent_delta++;
1894
1895         info->bitmap_extents += extent_delta;
1896         if (!btrfs_free_space_trimmed(info)) {
1897                 ctl->discardable_extents[BTRFS_STAT_CURR] += extent_delta;
1898                 ctl->discardable_bytes[BTRFS_STAT_CURR] -= bytes;
1899         }
1900
1901         if (update_stat)
1902                 ctl->free_space -= bytes;
1903 }
1904
1905 static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
1906                             struct btrfs_free_space *info, u64 offset,
1907                             u64 bytes)
1908 {
1909         unsigned long start, count, end;
1910         int extent_delta = 1;
1911
1912         start = offset_to_bit(info->offset, ctl->unit, offset);
1913         count = bytes_to_bits(bytes, ctl->unit);
1914         end = start + count;
1915         ASSERT(end <= BITS_PER_BITMAP);
1916
1917         bitmap_set(info->bitmap, start, count);
1918
1919         /*
1920          * We set some bytes, we have no idea what the max extent size is
1921          * anymore.
1922          */
1923         info->max_extent_size = 0;
1924         info->bytes += bytes;
1925         ctl->free_space += bytes;
1926
1927         relink_bitmap_entry(ctl, info);
1928
1929         if (start && test_bit(start - 1, info->bitmap))
1930                 extent_delta--;
1931
1932         if (end < BITS_PER_BITMAP && test_bit(end, info->bitmap))
1933                 extent_delta--;
1934
1935         info->bitmap_extents += extent_delta;
1936         if (!btrfs_free_space_trimmed(info)) {
1937                 ctl->discardable_extents[BTRFS_STAT_CURR] += extent_delta;
1938                 ctl->discardable_bytes[BTRFS_STAT_CURR] += bytes;
1939         }
1940 }
1941
1942 /*
1943  * If we can not find suitable extent, we will use bytes to record
1944  * the size of the max extent.
1945  */
1946 static int search_bitmap(struct btrfs_free_space_ctl *ctl,
1947                          struct btrfs_free_space *bitmap_info, u64 *offset,
1948                          u64 *bytes, bool for_alloc)
1949 {
1950         unsigned long found_bits = 0;
1951         unsigned long max_bits = 0;
1952         unsigned long bits, i;
1953         unsigned long next_zero;
1954         unsigned long extent_bits;
1955
1956         /*
1957          * Skip searching the bitmap if we don't have a contiguous section that
1958          * is large enough for this allocation.
1959          */
1960         if (for_alloc &&
1961             bitmap_info->max_extent_size &&
1962             bitmap_info->max_extent_size < *bytes) {
1963                 *bytes = bitmap_info->max_extent_size;
1964                 return -1;
1965         }
1966
1967         i = offset_to_bit(bitmap_info->offset, ctl->unit,
1968                           max_t(u64, *offset, bitmap_info->offset));
1969         bits = bytes_to_bits(*bytes, ctl->unit);
1970
1971         for_each_set_bit_from(i, bitmap_info->bitmap, BITS_PER_BITMAP) {
1972                 if (for_alloc && bits == 1) {
1973                         found_bits = 1;
1974                         break;
1975                 }
1976                 next_zero = find_next_zero_bit(bitmap_info->bitmap,
1977                                                BITS_PER_BITMAP, i);
1978                 extent_bits = next_zero - i;
1979                 if (extent_bits >= bits) {
1980                         found_bits = extent_bits;
1981                         break;
1982                 } else if (extent_bits > max_bits) {
1983                         max_bits = extent_bits;
1984                 }
1985                 i = next_zero;
1986         }
1987
1988         if (found_bits) {
1989                 *offset = (u64)(i * ctl->unit) + bitmap_info->offset;
1990                 *bytes = (u64)(found_bits) * ctl->unit;
1991                 return 0;
1992         }
1993
1994         *bytes = (u64)(max_bits) * ctl->unit;
1995         bitmap_info->max_extent_size = *bytes;
1996         relink_bitmap_entry(ctl, bitmap_info);
1997         return -1;
1998 }
1999
2000 /* Cache the size of the max extent in bytes */
2001 static struct btrfs_free_space *
2002 find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
2003                 unsigned long align, u64 *max_extent_size, bool use_bytes_index)
2004 {
2005         struct btrfs_free_space *entry;
2006         struct rb_node *node;
2007         u64 tmp;
2008         u64 align_off;
2009         int ret;
2010
2011         if (!ctl->free_space_offset.rb_node)
2012                 goto out;
2013 again:
2014         if (use_bytes_index) {
2015                 node = rb_first_cached(&ctl->free_space_bytes);
2016         } else {
2017                 entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset),
2018                                            0, 1);
2019                 if (!entry)
2020                         goto out;
2021                 node = &entry->offset_index;
2022         }
2023
2024         for (; node; node = rb_next(node)) {
2025                 if (use_bytes_index)
2026                         entry = rb_entry(node, struct btrfs_free_space,
2027                                          bytes_index);
2028                 else
2029                         entry = rb_entry(node, struct btrfs_free_space,
2030                                          offset_index);
2031
2032                 /*
2033                  * If we are using the bytes index then all subsequent entries
2034                  * in this tree are going to be < bytes, so simply set the max
2035                  * extent size and exit the loop.
2036                  *
2037                  * If we're using the offset index then we need to keep going
2038                  * through the rest of the tree.
2039                  */
2040                 if (entry->bytes < *bytes) {
2041                         *max_extent_size = max(get_max_extent_size(entry),
2042                                                *max_extent_size);
2043                         if (use_bytes_index)
2044                                 break;
2045                         continue;
2046                 }
2047
2048                 /* make sure the space returned is big enough
2049                  * to match our requested alignment
2050                  */
2051                 if (*bytes >= align) {
2052                         tmp = entry->offset - ctl->start + align - 1;
2053                         tmp = div64_u64(tmp, align);
2054                         tmp = tmp * align + ctl->start;
2055                         align_off = tmp - entry->offset;
2056                 } else {
2057                         align_off = 0;
2058                         tmp = entry->offset;
2059                 }
2060
2061                 /*
2062                  * We don't break here if we're using the bytes index because we
2063                  * may have another entry that has the correct alignment that is
2064                  * the right size, so we don't want to miss that possibility.
2065                  * At worst this adds another loop through the logic, but if we
2066                  * broke here we could prematurely ENOSPC.
2067                  */
2068                 if (entry->bytes < *bytes + align_off) {
2069                         *max_extent_size = max(get_max_extent_size(entry),
2070                                                *max_extent_size);
2071                         continue;
2072                 }
2073
2074                 if (entry->bitmap) {
2075                         struct rb_node *old_next = rb_next(node);
2076                         u64 size = *bytes;
2077
2078                         ret = search_bitmap(ctl, entry, &tmp, &size, true);
2079                         if (!ret) {
2080                                 *offset = tmp;
2081                                 *bytes = size;
2082                                 return entry;
2083                         } else {
2084                                 *max_extent_size =
2085                                         max(get_max_extent_size(entry),
2086                                             *max_extent_size);
2087                         }
2088
2089                         /*
2090                          * The bitmap may have gotten re-arranged in the space
2091                          * index here because the max_extent_size may have been
2092                          * updated.  Start from the beginning again if this
2093                          * happened.
2094                          */
2095                         if (use_bytes_index && old_next != rb_next(node))
2096                                 goto again;
2097                         continue;
2098                 }
2099
2100                 *offset = tmp;
2101                 *bytes = entry->bytes - align_off;
2102                 return entry;
2103         }
2104 out:
2105         return NULL;
2106 }
2107
2108 static void add_new_bitmap(struct btrfs_free_space_ctl *ctl,
2109                            struct btrfs_free_space *info, u64 offset)
2110 {
2111         info->offset = offset_to_bitmap(ctl, offset);
2112         info->bytes = 0;
2113         info->bitmap_extents = 0;
2114         INIT_LIST_HEAD(&info->list);
2115         link_free_space(ctl, info);
2116         ctl->total_bitmaps++;
2117         recalculate_thresholds(ctl);
2118 }
2119
2120 static void free_bitmap(struct btrfs_free_space_ctl *ctl,
2121                         struct btrfs_free_space *bitmap_info)
2122 {
2123         /*
2124          * Normally when this is called, the bitmap is completely empty. However,
2125          * if we are blowing up the free space cache for one reason or another
2126          * via __btrfs_remove_free_space_cache(), then it may not be freed and
2127          * we may leave stats on the table.
2128          */
2129         if (bitmap_info->bytes && !btrfs_free_space_trimmed(bitmap_info)) {
2130                 ctl->discardable_extents[BTRFS_STAT_CURR] -=
2131                         bitmap_info->bitmap_extents;
2132                 ctl->discardable_bytes[BTRFS_STAT_CURR] -= bitmap_info->bytes;
2133
2134         }
2135         unlink_free_space(ctl, bitmap_info, true);
2136         kmem_cache_free(btrfs_free_space_bitmap_cachep, bitmap_info->bitmap);
2137         kmem_cache_free(btrfs_free_space_cachep, bitmap_info);
2138         ctl->total_bitmaps--;
2139         recalculate_thresholds(ctl);
2140 }
2141
2142 static noinline int remove_from_bitmap(struct btrfs_free_space_ctl *ctl,
2143                               struct btrfs_free_space *bitmap_info,
2144                               u64 *offset, u64 *bytes)
2145 {
2146         u64 end;
2147         u64 search_start, search_bytes;
2148         int ret;
2149
2150 again:
2151         end = bitmap_info->offset + (u64)(BITS_PER_BITMAP * ctl->unit) - 1;
2152
2153         /*
2154          * We need to search for bits in this bitmap.  We could only cover some
2155          * of the extent in this bitmap thanks to how we add space, so we need
2156          * to search for as much as it as we can and clear that amount, and then
2157          * go searching for the next bit.
2158          */
2159         search_start = *offset;
2160         search_bytes = ctl->unit;
2161         search_bytes = min(search_bytes, end - search_start + 1);
2162         ret = search_bitmap(ctl, bitmap_info, &search_start, &search_bytes,
2163                             false);
2164         if (ret < 0 || search_start != *offset)
2165                 return -EINVAL;
2166
2167         /* We may have found more bits than what we need */
2168         search_bytes = min(search_bytes, *bytes);
2169
2170         /* Cannot clear past the end of the bitmap */
2171         search_bytes = min(search_bytes, end - search_start + 1);
2172
2173         bitmap_clear_bits(ctl, bitmap_info, search_start, search_bytes, true);
2174         *offset += search_bytes;
2175         *bytes -= search_bytes;
2176
2177         if (*bytes) {
2178                 struct rb_node *next = rb_next(&bitmap_info->offset_index);
2179                 if (!bitmap_info->bytes)
2180                         free_bitmap(ctl, bitmap_info);
2181
2182                 /*
2183                  * no entry after this bitmap, but we still have bytes to
2184                  * remove, so something has gone wrong.
2185                  */
2186                 if (!next)
2187                         return -EINVAL;
2188
2189                 bitmap_info = rb_entry(next, struct btrfs_free_space,
2190                                        offset_index);
2191
2192                 /*
2193                  * if the next entry isn't a bitmap we need to return to let the
2194                  * extent stuff do its work.
2195                  */
2196                 if (!bitmap_info->bitmap)
2197                         return -EAGAIN;
2198
2199                 /*
2200                  * Ok the next item is a bitmap, but it may not actually hold
2201                  * the information for the rest of this free space stuff, so
2202                  * look for it, and if we don't find it return so we can try
2203                  * everything over again.
2204                  */
2205                 search_start = *offset;
2206                 search_bytes = ctl->unit;
2207                 ret = search_bitmap(ctl, bitmap_info, &search_start,
2208                                     &search_bytes, false);
2209                 if (ret < 0 || search_start != *offset)
2210                         return -EAGAIN;
2211
2212                 goto again;
2213         } else if (!bitmap_info->bytes)
2214                 free_bitmap(ctl, bitmap_info);
2215
2216         return 0;
2217 }
2218
2219 static u64 add_bytes_to_bitmap(struct btrfs_free_space_ctl *ctl,
2220                                struct btrfs_free_space *info, u64 offset,
2221                                u64 bytes, enum btrfs_trim_state trim_state)
2222 {
2223         u64 bytes_to_set = 0;
2224         u64 end;
2225
2226         /*
2227          * This is a tradeoff to make bitmap trim state minimal.  We mark the
2228          * whole bitmap untrimmed if at any point we add untrimmed regions.
2229          */
2230         if (trim_state == BTRFS_TRIM_STATE_UNTRIMMED) {
2231                 if (btrfs_free_space_trimmed(info)) {
2232                         ctl->discardable_extents[BTRFS_STAT_CURR] +=
2233                                 info->bitmap_extents;
2234                         ctl->discardable_bytes[BTRFS_STAT_CURR] += info->bytes;
2235                 }
2236                 info->trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
2237         }
2238
2239         end = info->offset + (u64)(BITS_PER_BITMAP * ctl->unit);
2240
2241         bytes_to_set = min(end - offset, bytes);
2242
2243         bitmap_set_bits(ctl, info, offset, bytes_to_set);
2244
2245         return bytes_to_set;
2246
2247 }
2248
2249 static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
2250                       struct btrfs_free_space *info)
2251 {
2252         struct btrfs_block_group *block_group = ctl->block_group;
2253         struct btrfs_fs_info *fs_info = block_group->fs_info;
2254         bool forced = false;
2255
2256 #ifdef CONFIG_BTRFS_DEBUG
2257         if (btrfs_should_fragment_free_space(block_group))
2258                 forced = true;
2259 #endif
2260
2261         /* This is a way to reclaim large regions from the bitmaps. */
2262         if (!forced && info->bytes >= FORCE_EXTENT_THRESHOLD)
2263                 return false;
2264
2265         /*
2266          * If we are below the extents threshold then we can add this as an
2267          * extent, and don't have to deal with the bitmap
2268          */
2269         if (!forced && ctl->free_extents < ctl->extents_thresh) {
2270                 /*
2271                  * If this block group has some small extents we don't want to
2272                  * use up all of our free slots in the cache with them, we want
2273                  * to reserve them to larger extents, however if we have plenty
2274                  * of cache left then go ahead an dadd them, no sense in adding
2275                  * the overhead of a bitmap if we don't have to.
2276                  */
2277                 if (info->bytes <= fs_info->sectorsize * 8) {
2278                         if (ctl->free_extents * 3 <= ctl->extents_thresh)
2279                                 return false;
2280                 } else {
2281                         return false;
2282                 }
2283         }
2284
2285         /*
2286          * The original block groups from mkfs can be really small, like 8
2287          * megabytes, so don't bother with a bitmap for those entries.  However
2288          * some block groups can be smaller than what a bitmap would cover but
2289          * are still large enough that they could overflow the 32k memory limit,
2290          * so allow those block groups to still be allowed to have a bitmap
2291          * entry.
2292          */
2293         if (((BITS_PER_BITMAP * ctl->unit) >> 1) > block_group->length)
2294                 return false;
2295
2296         return true;
2297 }
2298
2299 static const struct btrfs_free_space_op free_space_op = {
2300         .use_bitmap             = use_bitmap,
2301 };
2302
2303 static int insert_into_bitmap(struct btrfs_free_space_ctl *ctl,
2304                               struct btrfs_free_space *info)
2305 {
2306         struct btrfs_free_space *bitmap_info;
2307         struct btrfs_block_group *block_group = NULL;
2308         int added = 0;
2309         u64 bytes, offset, bytes_added;
2310         enum btrfs_trim_state trim_state;
2311         int ret;
2312
2313         bytes = info->bytes;
2314         offset = info->offset;
2315         trim_state = info->trim_state;
2316
2317         if (!ctl->op->use_bitmap(ctl, info))
2318                 return 0;
2319
2320         if (ctl->op == &free_space_op)
2321                 block_group = ctl->block_group;
2322 again:
2323         /*
2324          * Since we link bitmaps right into the cluster we need to see if we
2325          * have a cluster here, and if so and it has our bitmap we need to add
2326          * the free space to that bitmap.
2327          */
2328         if (block_group && !list_empty(&block_group->cluster_list)) {
2329                 struct btrfs_free_cluster *cluster;
2330                 struct rb_node *node;
2331                 struct btrfs_free_space *entry;
2332
2333                 cluster = list_entry(block_group->cluster_list.next,
2334                                      struct btrfs_free_cluster,
2335                                      block_group_list);
2336                 spin_lock(&cluster->lock);
2337                 node = rb_first(&cluster->root);
2338                 if (!node) {
2339                         spin_unlock(&cluster->lock);
2340                         goto no_cluster_bitmap;
2341                 }
2342
2343                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2344                 if (!entry->bitmap) {
2345                         spin_unlock(&cluster->lock);
2346                         goto no_cluster_bitmap;
2347                 }
2348
2349                 if (entry->offset == offset_to_bitmap(ctl, offset)) {
2350                         bytes_added = add_bytes_to_bitmap(ctl, entry, offset,
2351                                                           bytes, trim_state);
2352                         bytes -= bytes_added;
2353                         offset += bytes_added;
2354                 }
2355                 spin_unlock(&cluster->lock);
2356                 if (!bytes) {
2357                         ret = 1;
2358                         goto out;
2359                 }
2360         }
2361
2362 no_cluster_bitmap:
2363         bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
2364                                          1, 0);
2365         if (!bitmap_info) {
2366                 ASSERT(added == 0);
2367                 goto new_bitmap;
2368         }
2369
2370         bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes,
2371                                           trim_state);
2372         bytes -= bytes_added;
2373         offset += bytes_added;
2374         added = 0;
2375
2376         if (!bytes) {
2377                 ret = 1;
2378                 goto out;
2379         } else
2380                 goto again;
2381
2382 new_bitmap:
2383         if (info && info->bitmap) {
2384                 add_new_bitmap(ctl, info, offset);
2385                 added = 1;
2386                 info = NULL;
2387                 goto again;
2388         } else {
2389                 spin_unlock(&ctl->tree_lock);
2390
2391                 /* no pre-allocated info, allocate a new one */
2392                 if (!info) {
2393                         info = kmem_cache_zalloc(btrfs_free_space_cachep,
2394                                                  GFP_NOFS);
2395                         if (!info) {
2396                                 spin_lock(&ctl->tree_lock);
2397                                 ret = -ENOMEM;
2398                                 goto out;
2399                         }
2400                 }
2401
2402                 /* allocate the bitmap */
2403                 info->bitmap = kmem_cache_zalloc(btrfs_free_space_bitmap_cachep,
2404                                                  GFP_NOFS);
2405                 info->trim_state = BTRFS_TRIM_STATE_TRIMMED;
2406                 spin_lock(&ctl->tree_lock);
2407                 if (!info->bitmap) {
2408                         ret = -ENOMEM;
2409                         goto out;
2410                 }
2411                 goto again;
2412         }
2413
2414 out:
2415         if (info) {
2416                 if (info->bitmap)
2417                         kmem_cache_free(btrfs_free_space_bitmap_cachep,
2418                                         info->bitmap);
2419                 kmem_cache_free(btrfs_free_space_cachep, info);
2420         }
2421
2422         return ret;
2423 }
2424
2425 /*
2426  * Free space merging rules:
2427  *  1) Merge trimmed areas together
2428  *  2) Let untrimmed areas coalesce with trimmed areas
2429  *  3) Always pull neighboring regions from bitmaps
2430  *
2431  * The above rules are for when we merge free space based on btrfs_trim_state.
2432  * Rules 2 and 3 are subtle because they are suboptimal, but are done for the
2433  * same reason: to promote larger extent regions which makes life easier for
2434  * find_free_extent().  Rule 2 enables coalescing based on the common path
2435  * being returning free space from btrfs_finish_extent_commit().  So when free
2436  * space is trimmed, it will prevent aggregating trimmed new region and
2437  * untrimmed regions in the rb_tree.  Rule 3 is purely to obtain larger extents
2438  * and provide find_free_extent() with the largest extents possible hoping for
2439  * the reuse path.
2440  */
2441 static bool try_merge_free_space(struct btrfs_free_space_ctl *ctl,
2442                           struct btrfs_free_space *info, bool update_stat)
2443 {
2444         struct btrfs_free_space *left_info = NULL;
2445         struct btrfs_free_space *right_info;
2446         bool merged = false;
2447         u64 offset = info->offset;
2448         u64 bytes = info->bytes;
2449         const bool is_trimmed = btrfs_free_space_trimmed(info);
2450
2451         /*
2452          * first we want to see if there is free space adjacent to the range we
2453          * are adding, if there is remove that struct and add a new one to
2454          * cover the entire range
2455          */
2456         right_info = tree_search_offset(ctl, offset + bytes, 0, 0);
2457         if (right_info && rb_prev(&right_info->offset_index))
2458                 left_info = rb_entry(rb_prev(&right_info->offset_index),
2459                                      struct btrfs_free_space, offset_index);
2460         else if (!right_info)
2461                 left_info = tree_search_offset(ctl, offset - 1, 0, 0);
2462
2463         /* See try_merge_free_space() comment. */
2464         if (right_info && !right_info->bitmap &&
2465             (!is_trimmed || btrfs_free_space_trimmed(right_info))) {
2466                 unlink_free_space(ctl, right_info, update_stat);
2467                 info->bytes += right_info->bytes;
2468                 kmem_cache_free(btrfs_free_space_cachep, right_info);
2469                 merged = true;
2470         }
2471
2472         /* See try_merge_free_space() comment. */
2473         if (left_info && !left_info->bitmap &&
2474             left_info->offset + left_info->bytes == offset &&
2475             (!is_trimmed || btrfs_free_space_trimmed(left_info))) {
2476                 unlink_free_space(ctl, left_info, update_stat);
2477                 info->offset = left_info->offset;
2478                 info->bytes += left_info->bytes;
2479                 kmem_cache_free(btrfs_free_space_cachep, left_info);
2480                 merged = true;
2481         }
2482
2483         return merged;
2484 }
2485
2486 static bool steal_from_bitmap_to_end(struct btrfs_free_space_ctl *ctl,
2487                                      struct btrfs_free_space *info,
2488                                      bool update_stat)
2489 {
2490         struct btrfs_free_space *bitmap;
2491         unsigned long i;
2492         unsigned long j;
2493         const u64 end = info->offset + info->bytes;
2494         const u64 bitmap_offset = offset_to_bitmap(ctl, end);
2495         u64 bytes;
2496
2497         bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
2498         if (!bitmap)
2499                 return false;
2500
2501         i = offset_to_bit(bitmap->offset, ctl->unit, end);
2502         j = find_next_zero_bit(bitmap->bitmap, BITS_PER_BITMAP, i);
2503         if (j == i)
2504                 return false;
2505         bytes = (j - i) * ctl->unit;
2506         info->bytes += bytes;
2507
2508         /* See try_merge_free_space() comment. */
2509         if (!btrfs_free_space_trimmed(bitmap))
2510                 info->trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
2511
2512         bitmap_clear_bits(ctl, bitmap, end, bytes, update_stat);
2513
2514         if (!bitmap->bytes)
2515                 free_bitmap(ctl, bitmap);
2516
2517         return true;
2518 }
2519
2520 static bool steal_from_bitmap_to_front(struct btrfs_free_space_ctl *ctl,
2521                                        struct btrfs_free_space *info,
2522                                        bool update_stat)
2523 {
2524         struct btrfs_free_space *bitmap;
2525         u64 bitmap_offset;
2526         unsigned long i;
2527         unsigned long j;
2528         unsigned long prev_j;
2529         u64 bytes;
2530
2531         bitmap_offset = offset_to_bitmap(ctl, info->offset);
2532         /* If we're on a boundary, try the previous logical bitmap. */
2533         if (bitmap_offset == info->offset) {
2534                 if (info->offset == 0)
2535                         return false;
2536                 bitmap_offset = offset_to_bitmap(ctl, info->offset - 1);
2537         }
2538
2539         bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
2540         if (!bitmap)
2541                 return false;
2542
2543         i = offset_to_bit(bitmap->offset, ctl->unit, info->offset) - 1;
2544         j = 0;
2545         prev_j = (unsigned long)-1;
2546         for_each_clear_bit_from(j, bitmap->bitmap, BITS_PER_BITMAP) {
2547                 if (j > i)
2548                         break;
2549                 prev_j = j;
2550         }
2551         if (prev_j == i)
2552                 return false;
2553
2554         if (prev_j == (unsigned long)-1)
2555                 bytes = (i + 1) * ctl->unit;
2556         else
2557                 bytes = (i - prev_j) * ctl->unit;
2558
2559         info->offset -= bytes;
2560         info->bytes += bytes;
2561
2562         /* See try_merge_free_space() comment. */
2563         if (!btrfs_free_space_trimmed(bitmap))
2564                 info->trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
2565
2566         bitmap_clear_bits(ctl, bitmap, info->offset, bytes, update_stat);
2567
2568         if (!bitmap->bytes)
2569                 free_bitmap(ctl, bitmap);
2570
2571         return true;
2572 }
2573
2574 /*
2575  * We prefer always to allocate from extent entries, both for clustered and
2576  * non-clustered allocation requests. So when attempting to add a new extent
2577  * entry, try to see if there's adjacent free space in bitmap entries, and if
2578  * there is, migrate that space from the bitmaps to the extent.
2579  * Like this we get better chances of satisfying space allocation requests
2580  * because we attempt to satisfy them based on a single cache entry, and never
2581  * on 2 or more entries - even if the entries represent a contiguous free space
2582  * region (e.g. 1 extent entry + 1 bitmap entry starting where the extent entry
2583  * ends).
2584  */
2585 static void steal_from_bitmap(struct btrfs_free_space_ctl *ctl,
2586                               struct btrfs_free_space *info,
2587                               bool update_stat)
2588 {
2589         /*
2590          * Only work with disconnected entries, as we can change their offset,
2591          * and must be extent entries.
2592          */
2593         ASSERT(!info->bitmap);
2594         ASSERT(RB_EMPTY_NODE(&info->offset_index));
2595
2596         if (ctl->total_bitmaps > 0) {
2597                 bool stole_end;
2598                 bool stole_front = false;
2599
2600                 stole_end = steal_from_bitmap_to_end(ctl, info, update_stat);
2601                 if (ctl->total_bitmaps > 0)
2602                         stole_front = steal_from_bitmap_to_front(ctl, info,
2603                                                                  update_stat);
2604
2605                 if (stole_end || stole_front)
2606                         try_merge_free_space(ctl, info, update_stat);
2607         }
2608 }
2609
2610 int __btrfs_add_free_space(struct btrfs_block_group *block_group,
2611                            u64 offset, u64 bytes,
2612                            enum btrfs_trim_state trim_state)
2613 {
2614         struct btrfs_fs_info *fs_info = block_group->fs_info;
2615         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2616         struct btrfs_free_space *info;
2617         int ret = 0;
2618         u64 filter_bytes = bytes;
2619
2620         ASSERT(!btrfs_is_zoned(fs_info));
2621
2622         info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
2623         if (!info)
2624                 return -ENOMEM;
2625
2626         info->offset = offset;
2627         info->bytes = bytes;
2628         info->trim_state = trim_state;
2629         RB_CLEAR_NODE(&info->offset_index);
2630         RB_CLEAR_NODE(&info->bytes_index);
2631
2632         spin_lock(&ctl->tree_lock);
2633
2634         if (try_merge_free_space(ctl, info, true))
2635                 goto link;
2636
2637         /*
2638          * There was no extent directly to the left or right of this new
2639          * extent then we know we're going to have to allocate a new extent, so
2640          * before we do that see if we need to drop this into a bitmap
2641          */
2642         ret = insert_into_bitmap(ctl, info);
2643         if (ret < 0) {
2644                 goto out;
2645         } else if (ret) {
2646                 ret = 0;
2647                 goto out;
2648         }
2649 link:
2650         /*
2651          * Only steal free space from adjacent bitmaps if we're sure we're not
2652          * going to add the new free space to existing bitmap entries - because
2653          * that would mean unnecessary work that would be reverted. Therefore
2654          * attempt to steal space from bitmaps if we're adding an extent entry.
2655          */
2656         steal_from_bitmap(ctl, info, true);
2657
2658         filter_bytes = max(filter_bytes, info->bytes);
2659
2660         ret = link_free_space(ctl, info);
2661         if (ret)
2662                 kmem_cache_free(btrfs_free_space_cachep, info);
2663 out:
2664         btrfs_discard_update_discardable(block_group);
2665         spin_unlock(&ctl->tree_lock);
2666
2667         if (ret) {
2668                 btrfs_crit(fs_info, "unable to add free space :%d", ret);
2669                 ASSERT(ret != -EEXIST);
2670         }
2671
2672         if (trim_state != BTRFS_TRIM_STATE_TRIMMED) {
2673                 btrfs_discard_check_filter(block_group, filter_bytes);
2674                 btrfs_discard_queue_work(&fs_info->discard_ctl, block_group);
2675         }
2676
2677         return ret;
2678 }
2679
2680 static int __btrfs_add_free_space_zoned(struct btrfs_block_group *block_group,
2681                                         u64 bytenr, u64 size, bool used)
2682 {
2683         struct btrfs_space_info *sinfo = block_group->space_info;
2684         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2685         u64 offset = bytenr - block_group->start;
2686         u64 to_free, to_unusable;
2687         int bg_reclaim_threshold = 0;
2688         bool initial = (size == block_group->length);
2689         u64 reclaimable_unusable;
2690
2691         WARN_ON(!initial && offset + size > block_group->zone_capacity);
2692
2693         if (!initial)
2694                 bg_reclaim_threshold = READ_ONCE(sinfo->bg_reclaim_threshold);
2695
2696         spin_lock(&ctl->tree_lock);
2697         /* Count initial region as zone_unusable until it gets activated. */
2698         if (!used)
2699                 to_free = size;
2700         else if (initial &&
2701                  test_bit(BTRFS_FS_ACTIVE_ZONE_TRACKING, &block_group->fs_info->flags) &&
2702                  (block_group->flags & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_SYSTEM)))
2703                 to_free = 0;
2704         else if (initial)
2705                 to_free = block_group->zone_capacity;
2706         else if (offset >= block_group->alloc_offset)
2707                 to_free = size;
2708         else if (offset + size <= block_group->alloc_offset)
2709                 to_free = 0;
2710         else
2711                 to_free = offset + size - block_group->alloc_offset;
2712         to_unusable = size - to_free;
2713
2714         ctl->free_space += to_free;
2715         /*
2716          * If the block group is read-only, we should account freed space into
2717          * bytes_readonly.
2718          */
2719         if (!block_group->ro)
2720                 block_group->zone_unusable += to_unusable;
2721         spin_unlock(&ctl->tree_lock);
2722         if (!used) {
2723                 spin_lock(&block_group->lock);
2724                 block_group->alloc_offset -= size;
2725                 spin_unlock(&block_group->lock);
2726         }
2727
2728         reclaimable_unusable = block_group->zone_unusable -
2729                                (block_group->length - block_group->zone_capacity);
2730         /* All the region is now unusable. Mark it as unused and reclaim */
2731         if (block_group->zone_unusable == block_group->length &&
2732             block_group->alloc_offset) {
2733                 btrfs_mark_bg_unused(block_group);
2734         } else if (bg_reclaim_threshold &&
2735                    reclaimable_unusable >=
2736                    mult_perc(block_group->zone_capacity, bg_reclaim_threshold)) {
2737                 btrfs_mark_bg_to_reclaim(block_group);
2738         }
2739
2740         return 0;
2741 }
2742
2743 int btrfs_add_free_space(struct btrfs_block_group *block_group,
2744                          u64 bytenr, u64 size)
2745 {
2746         enum btrfs_trim_state trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
2747
2748         if (btrfs_is_zoned(block_group->fs_info))
2749                 return __btrfs_add_free_space_zoned(block_group, bytenr, size,
2750                                                     true);
2751
2752         if (btrfs_test_opt(block_group->fs_info, DISCARD_SYNC))
2753                 trim_state = BTRFS_TRIM_STATE_TRIMMED;
2754
2755         return __btrfs_add_free_space(block_group, bytenr, size, trim_state);
2756 }
2757
2758 int btrfs_add_free_space_unused(struct btrfs_block_group *block_group,
2759                                 u64 bytenr, u64 size)
2760 {
2761         if (btrfs_is_zoned(block_group->fs_info))
2762                 return __btrfs_add_free_space_zoned(block_group, bytenr, size,
2763                                                     false);
2764
2765         return btrfs_add_free_space(block_group, bytenr, size);
2766 }
2767
2768 /*
2769  * This is a subtle distinction because when adding free space back in general,
2770  * we want it to be added as untrimmed for async. But in the case where we add
2771  * it on loading of a block group, we want to consider it trimmed.
2772  */
2773 int btrfs_add_free_space_async_trimmed(struct btrfs_block_group *block_group,
2774                                        u64 bytenr, u64 size)
2775 {
2776         enum btrfs_trim_state trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
2777
2778         if (btrfs_is_zoned(block_group->fs_info))
2779                 return __btrfs_add_free_space_zoned(block_group, bytenr, size,
2780                                                     true);
2781
2782         if (btrfs_test_opt(block_group->fs_info, DISCARD_SYNC) ||
2783             btrfs_test_opt(block_group->fs_info, DISCARD_ASYNC))
2784                 trim_state = BTRFS_TRIM_STATE_TRIMMED;
2785
2786         return __btrfs_add_free_space(block_group, bytenr, size, trim_state);
2787 }
2788
2789 int btrfs_remove_free_space(struct btrfs_block_group *block_group,
2790                             u64 offset, u64 bytes)
2791 {
2792         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2793         struct btrfs_free_space *info;
2794         int ret;
2795         bool re_search = false;
2796
2797         if (btrfs_is_zoned(block_group->fs_info)) {
2798                 /*
2799                  * This can happen with conventional zones when replaying log.
2800                  * Since the allocation info of tree-log nodes are not recorded
2801                  * to the extent-tree, calculate_alloc_pointer() failed to
2802                  * advance the allocation pointer after last allocated tree log
2803                  * node blocks.
2804                  *
2805                  * This function is called from
2806                  * btrfs_pin_extent_for_log_replay() when replaying the log.
2807                  * Advance the pointer not to overwrite the tree-log nodes.
2808                  */
2809                 if (block_group->start + block_group->alloc_offset <
2810                     offset + bytes) {
2811                         block_group->alloc_offset =
2812                                 offset + bytes - block_group->start;
2813                 }
2814                 return 0;
2815         }
2816
2817         spin_lock(&ctl->tree_lock);
2818
2819 again:
2820         ret = 0;
2821         if (!bytes)
2822                 goto out_lock;
2823
2824         info = tree_search_offset(ctl, offset, 0, 0);
2825         if (!info) {
2826                 /*
2827                  * oops didn't find an extent that matched the space we wanted
2828                  * to remove, look for a bitmap instead
2829                  */
2830                 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
2831                                           1, 0);
2832                 if (!info) {
2833                         /*
2834                          * If we found a partial bit of our free space in a
2835                          * bitmap but then couldn't find the other part this may
2836                          * be a problem, so WARN about it.
2837                          */
2838                         WARN_ON(re_search);
2839                         goto out_lock;
2840                 }
2841         }
2842
2843         re_search = false;
2844         if (!info->bitmap) {
2845                 unlink_free_space(ctl, info, true);
2846                 if (offset == info->offset) {
2847                         u64 to_free = min(bytes, info->bytes);
2848
2849                         info->bytes -= to_free;
2850                         info->offset += to_free;
2851                         if (info->bytes) {
2852                                 ret = link_free_space(ctl, info);
2853                                 WARN_ON(ret);
2854                         } else {
2855                                 kmem_cache_free(btrfs_free_space_cachep, info);
2856                         }
2857
2858                         offset += to_free;
2859                         bytes -= to_free;
2860                         goto again;
2861                 } else {
2862                         u64 old_end = info->bytes + info->offset;
2863
2864                         info->bytes = offset - info->offset;
2865                         ret = link_free_space(ctl, info);
2866                         WARN_ON(ret);
2867                         if (ret)
2868                                 goto out_lock;
2869
2870                         /* Not enough bytes in this entry to satisfy us */
2871                         if (old_end < offset + bytes) {
2872                                 bytes -= old_end - offset;
2873                                 offset = old_end;
2874                                 goto again;
2875                         } else if (old_end == offset + bytes) {
2876                                 /* all done */
2877                                 goto out_lock;
2878                         }
2879                         spin_unlock(&ctl->tree_lock);
2880
2881                         ret = __btrfs_add_free_space(block_group,
2882                                                      offset + bytes,
2883                                                      old_end - (offset + bytes),
2884                                                      info->trim_state);
2885                         WARN_ON(ret);
2886                         goto out;
2887                 }
2888         }
2889
2890         ret = remove_from_bitmap(ctl, info, &offset, &bytes);
2891         if (ret == -EAGAIN) {
2892                 re_search = true;
2893                 goto again;
2894         }
2895 out_lock:
2896         btrfs_discard_update_discardable(block_group);
2897         spin_unlock(&ctl->tree_lock);
2898 out:
2899         return ret;
2900 }
2901
2902 void btrfs_dump_free_space(struct btrfs_block_group *block_group,
2903                            u64 bytes)
2904 {
2905         struct btrfs_fs_info *fs_info = block_group->fs_info;
2906         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2907         struct btrfs_free_space *info;
2908         struct rb_node *n;
2909         int count = 0;
2910
2911         /*
2912          * Zoned btrfs does not use free space tree and cluster. Just print
2913          * out the free space after the allocation offset.
2914          */
2915         if (btrfs_is_zoned(fs_info)) {
2916                 btrfs_info(fs_info, "free space %llu active %d",
2917                            block_group->zone_capacity - block_group->alloc_offset,
2918                            test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
2919                                     &block_group->runtime_flags));
2920                 return;
2921         }
2922
2923         spin_lock(&ctl->tree_lock);
2924         for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
2925                 info = rb_entry(n, struct btrfs_free_space, offset_index);
2926                 if (info->bytes >= bytes && !block_group->ro)
2927                         count++;
2928                 btrfs_crit(fs_info, "entry offset %llu, bytes %llu, bitmap %s",
2929                            info->offset, info->bytes,
2930                        (info->bitmap) ? "yes" : "no");
2931         }
2932         spin_unlock(&ctl->tree_lock);
2933         btrfs_info(fs_info, "block group has cluster?: %s",
2934                list_empty(&block_group->cluster_list) ? "no" : "yes");
2935         btrfs_info(fs_info,
2936                    "%d blocks of free space at or bigger than bytes is", count);
2937 }
2938
2939 void btrfs_init_free_space_ctl(struct btrfs_block_group *block_group,
2940                                struct btrfs_free_space_ctl *ctl)
2941 {
2942         struct btrfs_fs_info *fs_info = block_group->fs_info;
2943
2944         spin_lock_init(&ctl->tree_lock);
2945         ctl->unit = fs_info->sectorsize;
2946         ctl->start = block_group->start;
2947         ctl->block_group = block_group;
2948         ctl->op = &free_space_op;
2949         ctl->free_space_bytes = RB_ROOT_CACHED;
2950         INIT_LIST_HEAD(&ctl->trimming_ranges);
2951         mutex_init(&ctl->cache_writeout_mutex);
2952
2953         /*
2954          * we only want to have 32k of ram per block group for keeping
2955          * track of free space, and if we pass 1/2 of that we want to
2956          * start converting things over to using bitmaps
2957          */
2958         ctl->extents_thresh = (SZ_32K / 2) / sizeof(struct btrfs_free_space);
2959 }
2960
2961 /*
2962  * for a given cluster, put all of its extents back into the free
2963  * space cache.  If the block group passed doesn't match the block group
2964  * pointed to by the cluster, someone else raced in and freed the
2965  * cluster already.  In that case, we just return without changing anything
2966  */
2967 static void __btrfs_return_cluster_to_free_space(
2968                              struct btrfs_block_group *block_group,
2969                              struct btrfs_free_cluster *cluster)
2970 {
2971         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2972         struct btrfs_free_space *entry;
2973         struct rb_node *node;
2974
2975         spin_lock(&cluster->lock);
2976         if (cluster->block_group != block_group) {
2977                 spin_unlock(&cluster->lock);
2978                 return;
2979         }
2980
2981         cluster->block_group = NULL;
2982         cluster->window_start = 0;
2983         list_del_init(&cluster->block_group_list);
2984
2985         node = rb_first(&cluster->root);
2986         while (node) {
2987                 bool bitmap;
2988
2989                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2990                 node = rb_next(&entry->offset_index);
2991                 rb_erase(&entry->offset_index, &cluster->root);
2992                 RB_CLEAR_NODE(&entry->offset_index);
2993
2994                 bitmap = (entry->bitmap != NULL);
2995                 if (!bitmap) {
2996                         /* Merging treats extents as if they were new */
2997                         if (!btrfs_free_space_trimmed(entry)) {
2998                                 ctl->discardable_extents[BTRFS_STAT_CURR]--;
2999                                 ctl->discardable_bytes[BTRFS_STAT_CURR] -=
3000                                         entry->bytes;
3001                         }
3002
3003                         try_merge_free_space(ctl, entry, false);
3004                         steal_from_bitmap(ctl, entry, false);
3005
3006                         /* As we insert directly, update these statistics */
3007                         if (!btrfs_free_space_trimmed(entry)) {
3008                                 ctl->discardable_extents[BTRFS_STAT_CURR]++;
3009                                 ctl->discardable_bytes[BTRFS_STAT_CURR] +=
3010                                         entry->bytes;
3011                         }
3012                 }
3013                 tree_insert_offset(&ctl->free_space_offset,
3014                                    entry->offset, &entry->offset_index, bitmap);
3015                 rb_add_cached(&entry->bytes_index, &ctl->free_space_bytes,
3016                               entry_less);
3017         }
3018         cluster->root = RB_ROOT;
3019         spin_unlock(&cluster->lock);
3020         btrfs_put_block_group(block_group);
3021 }
3022
3023 void btrfs_remove_free_space_cache(struct btrfs_block_group *block_group)
3024 {
3025         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3026         struct btrfs_free_cluster *cluster;
3027         struct list_head *head;
3028
3029         spin_lock(&ctl->tree_lock);
3030         while ((head = block_group->cluster_list.next) !=
3031                &block_group->cluster_list) {
3032                 cluster = list_entry(head, struct btrfs_free_cluster,
3033                                      block_group_list);
3034
3035                 WARN_ON(cluster->block_group != block_group);
3036                 __btrfs_return_cluster_to_free_space(block_group, cluster);
3037
3038                 cond_resched_lock(&ctl->tree_lock);
3039         }
3040         __btrfs_remove_free_space_cache(ctl);
3041         btrfs_discard_update_discardable(block_group);
3042         spin_unlock(&ctl->tree_lock);
3043
3044 }
3045
3046 /*
3047  * Walk @block_group's free space rb_tree to determine if everything is trimmed.
3048  */
3049 bool btrfs_is_free_space_trimmed(struct btrfs_block_group *block_group)
3050 {
3051         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3052         struct btrfs_free_space *info;
3053         struct rb_node *node;
3054         bool ret = true;
3055
3056         spin_lock(&ctl->tree_lock);
3057         node = rb_first(&ctl->free_space_offset);
3058
3059         while (node) {
3060                 info = rb_entry(node, struct btrfs_free_space, offset_index);
3061
3062                 if (!btrfs_free_space_trimmed(info)) {
3063                         ret = false;
3064                         break;
3065                 }
3066
3067                 node = rb_next(node);
3068         }
3069
3070         spin_unlock(&ctl->tree_lock);
3071         return ret;
3072 }
3073
3074 u64 btrfs_find_space_for_alloc(struct btrfs_block_group *block_group,
3075                                u64 offset, u64 bytes, u64 empty_size,
3076                                u64 *max_extent_size)
3077 {
3078         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3079         struct btrfs_discard_ctl *discard_ctl =
3080                                         &block_group->fs_info->discard_ctl;
3081         struct btrfs_free_space *entry = NULL;
3082         u64 bytes_search = bytes + empty_size;
3083         u64 ret = 0;
3084         u64 align_gap = 0;
3085         u64 align_gap_len = 0;
3086         enum btrfs_trim_state align_gap_trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
3087         bool use_bytes_index = (offset == block_group->start);
3088
3089         ASSERT(!btrfs_is_zoned(block_group->fs_info));
3090
3091         spin_lock(&ctl->tree_lock);
3092         entry = find_free_space(ctl, &offset, &bytes_search,
3093                                 block_group->full_stripe_len, max_extent_size,
3094                                 use_bytes_index);
3095         if (!entry)
3096                 goto out;
3097
3098         ret = offset;
3099         if (entry->bitmap) {
3100                 bitmap_clear_bits(ctl, entry, offset, bytes, true);
3101
3102                 if (!btrfs_free_space_trimmed(entry))
3103                         atomic64_add(bytes, &discard_ctl->discard_bytes_saved);
3104
3105                 if (!entry->bytes)
3106                         free_bitmap(ctl, entry);
3107         } else {
3108                 unlink_free_space(ctl, entry, true);
3109                 align_gap_len = offset - entry->offset;
3110                 align_gap = entry->offset;
3111                 align_gap_trim_state = entry->trim_state;
3112
3113                 if (!btrfs_free_space_trimmed(entry))
3114                         atomic64_add(bytes, &discard_ctl->discard_bytes_saved);
3115
3116                 entry->offset = offset + bytes;
3117                 WARN_ON(entry->bytes < bytes + align_gap_len);
3118
3119                 entry->bytes -= bytes + align_gap_len;
3120                 if (!entry->bytes)
3121                         kmem_cache_free(btrfs_free_space_cachep, entry);
3122                 else
3123                         link_free_space(ctl, entry);
3124         }
3125 out:
3126         btrfs_discard_update_discardable(block_group);
3127         spin_unlock(&ctl->tree_lock);
3128
3129         if (align_gap_len)
3130                 __btrfs_add_free_space(block_group, align_gap, align_gap_len,
3131                                        align_gap_trim_state);
3132         return ret;
3133 }
3134
3135 /*
3136  * given a cluster, put all of its extents back into the free space
3137  * cache.  If a block group is passed, this function will only free
3138  * a cluster that belongs to the passed block group.
3139  *
3140  * Otherwise, it'll get a reference on the block group pointed to by the
3141  * cluster and remove the cluster from it.
3142  */
3143 void btrfs_return_cluster_to_free_space(
3144                                struct btrfs_block_group *block_group,
3145                                struct btrfs_free_cluster *cluster)
3146 {
3147         struct btrfs_free_space_ctl *ctl;
3148
3149         /* first, get a safe pointer to the block group */
3150         spin_lock(&cluster->lock);
3151         if (!block_group) {
3152                 block_group = cluster->block_group;
3153                 if (!block_group) {
3154                         spin_unlock(&cluster->lock);
3155                         return;
3156                 }
3157         } else if (cluster->block_group != block_group) {
3158                 /* someone else has already freed it don't redo their work */
3159                 spin_unlock(&cluster->lock);
3160                 return;
3161         }
3162         btrfs_get_block_group(block_group);
3163         spin_unlock(&cluster->lock);
3164
3165         ctl = block_group->free_space_ctl;
3166
3167         /* now return any extents the cluster had on it */
3168         spin_lock(&ctl->tree_lock);
3169         __btrfs_return_cluster_to_free_space(block_group, cluster);
3170         spin_unlock(&ctl->tree_lock);
3171
3172         btrfs_discard_queue_work(&block_group->fs_info->discard_ctl, block_group);
3173
3174         /* finally drop our ref */
3175         btrfs_put_block_group(block_group);
3176 }
3177
3178 static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group *block_group,
3179                                    struct btrfs_free_cluster *cluster,
3180                                    struct btrfs_free_space *entry,
3181                                    u64 bytes, u64 min_start,
3182                                    u64 *max_extent_size)
3183 {
3184         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3185         int err;
3186         u64 search_start = cluster->window_start;
3187         u64 search_bytes = bytes;
3188         u64 ret = 0;
3189
3190         search_start = min_start;
3191         search_bytes = bytes;
3192
3193         err = search_bitmap(ctl, entry, &search_start, &search_bytes, true);
3194         if (err) {
3195                 *max_extent_size = max(get_max_extent_size(entry),
3196                                        *max_extent_size);
3197                 return 0;
3198         }
3199
3200         ret = search_start;
3201         bitmap_clear_bits(ctl, entry, ret, bytes, false);
3202
3203         return ret;
3204 }
3205
3206 /*
3207  * given a cluster, try to allocate 'bytes' from it, returns 0
3208  * if it couldn't find anything suitably large, or a logical disk offset
3209  * if things worked out
3210  */
3211 u64 btrfs_alloc_from_cluster(struct btrfs_block_group *block_group,
3212                              struct btrfs_free_cluster *cluster, u64 bytes,
3213                              u64 min_start, u64 *max_extent_size)
3214 {
3215         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3216         struct btrfs_discard_ctl *discard_ctl =
3217                                         &block_group->fs_info->discard_ctl;
3218         struct btrfs_free_space *entry = NULL;
3219         struct rb_node *node;
3220         u64 ret = 0;
3221
3222         ASSERT(!btrfs_is_zoned(block_group->fs_info));
3223
3224         spin_lock(&cluster->lock);
3225         if (bytes > cluster->max_size)
3226                 goto out;
3227
3228         if (cluster->block_group != block_group)
3229                 goto out;
3230
3231         node = rb_first(&cluster->root);
3232         if (!node)
3233                 goto out;
3234
3235         entry = rb_entry(node, struct btrfs_free_space, offset_index);
3236         while (1) {
3237                 if (entry->bytes < bytes)
3238                         *max_extent_size = max(get_max_extent_size(entry),
3239                                                *max_extent_size);
3240
3241                 if (entry->bytes < bytes ||
3242                     (!entry->bitmap && entry->offset < min_start)) {
3243                         node = rb_next(&entry->offset_index);
3244                         if (!node)
3245                                 break;
3246                         entry = rb_entry(node, struct btrfs_free_space,
3247                                          offset_index);
3248                         continue;
3249                 }
3250
3251                 if (entry->bitmap) {
3252                         ret = btrfs_alloc_from_bitmap(block_group,
3253                                                       cluster, entry, bytes,
3254                                                       cluster->window_start,
3255                                                       max_extent_size);
3256                         if (ret == 0) {
3257                                 node = rb_next(&entry->offset_index);
3258                                 if (!node)
3259                                         break;
3260                                 entry = rb_entry(node, struct btrfs_free_space,
3261                                                  offset_index);
3262                                 continue;
3263                         }
3264                         cluster->window_start += bytes;
3265                 } else {
3266                         ret = entry->offset;
3267
3268                         entry->offset += bytes;
3269                         entry->bytes -= bytes;
3270                 }
3271
3272                 break;
3273         }
3274 out:
3275         spin_unlock(&cluster->lock);
3276
3277         if (!ret)
3278                 return 0;
3279
3280         spin_lock(&ctl->tree_lock);
3281
3282         if (!btrfs_free_space_trimmed(entry))
3283                 atomic64_add(bytes, &discard_ctl->discard_bytes_saved);
3284
3285         ctl->free_space -= bytes;
3286         if (!entry->bitmap && !btrfs_free_space_trimmed(entry))
3287                 ctl->discardable_bytes[BTRFS_STAT_CURR] -= bytes;
3288
3289         spin_lock(&cluster->lock);
3290         if (entry->bytes == 0) {
3291                 rb_erase(&entry->offset_index, &cluster->root);
3292                 ctl->free_extents--;
3293                 if (entry->bitmap) {
3294                         kmem_cache_free(btrfs_free_space_bitmap_cachep,
3295                                         entry->bitmap);
3296                         ctl->total_bitmaps--;
3297                         recalculate_thresholds(ctl);
3298                 } else if (!btrfs_free_space_trimmed(entry)) {
3299                         ctl->discardable_extents[BTRFS_STAT_CURR]--;
3300                 }
3301                 kmem_cache_free(btrfs_free_space_cachep, entry);
3302         }
3303
3304         spin_unlock(&cluster->lock);
3305         spin_unlock(&ctl->tree_lock);
3306
3307         return ret;
3308 }
3309
3310 static int btrfs_bitmap_cluster(struct btrfs_block_group *block_group,
3311                                 struct btrfs_free_space *entry,
3312                                 struct btrfs_free_cluster *cluster,
3313                                 u64 offset, u64 bytes,
3314                                 u64 cont1_bytes, u64 min_bytes)
3315 {
3316         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3317         unsigned long next_zero;
3318         unsigned long i;
3319         unsigned long want_bits;
3320         unsigned long min_bits;
3321         unsigned long found_bits;
3322         unsigned long max_bits = 0;
3323         unsigned long start = 0;
3324         unsigned long total_found = 0;
3325         int ret;
3326
3327         i = offset_to_bit(entry->offset, ctl->unit,
3328                           max_t(u64, offset, entry->offset));
3329         want_bits = bytes_to_bits(bytes, ctl->unit);
3330         min_bits = bytes_to_bits(min_bytes, ctl->unit);
3331
3332         /*
3333          * Don't bother looking for a cluster in this bitmap if it's heavily
3334          * fragmented.
3335          */
3336         if (entry->max_extent_size &&
3337             entry->max_extent_size < cont1_bytes)
3338                 return -ENOSPC;
3339 again:
3340         found_bits = 0;
3341         for_each_set_bit_from(i, entry->bitmap, BITS_PER_BITMAP) {
3342                 next_zero = find_next_zero_bit(entry->bitmap,
3343                                                BITS_PER_BITMAP, i);
3344                 if (next_zero - i >= min_bits) {
3345                         found_bits = next_zero - i;
3346                         if (found_bits > max_bits)
3347                                 max_bits = found_bits;
3348                         break;
3349                 }
3350                 if (next_zero - i > max_bits)
3351                         max_bits = next_zero - i;
3352                 i = next_zero;
3353         }
3354
3355         if (!found_bits) {
3356                 entry->max_extent_size = (u64)max_bits * ctl->unit;
3357                 return -ENOSPC;
3358         }
3359
3360         if (!total_found) {
3361                 start = i;
3362                 cluster->max_size = 0;
3363         }
3364
3365         total_found += found_bits;
3366
3367         if (cluster->max_size < found_bits * ctl->unit)
3368                 cluster->max_size = found_bits * ctl->unit;
3369
3370         if (total_found < want_bits || cluster->max_size < cont1_bytes) {
3371                 i = next_zero + 1;
3372                 goto again;
3373         }
3374
3375         cluster->window_start = start * ctl->unit + entry->offset;
3376         rb_erase(&entry->offset_index, &ctl->free_space_offset);
3377         rb_erase_cached(&entry->bytes_index, &ctl->free_space_bytes);
3378
3379         /*
3380          * We need to know if we're currently on the normal space index when we
3381          * manipulate the bitmap so that we know we need to remove and re-insert
3382          * it into the space_index tree.  Clear the bytes_index node here so the
3383          * bitmap manipulation helpers know not to mess with the space_index
3384          * until this bitmap entry is added back into the normal cache.
3385          */
3386         RB_CLEAR_NODE(&entry->bytes_index);
3387
3388         ret = tree_insert_offset(&cluster->root, entry->offset,
3389                                  &entry->offset_index, 1);
3390         ASSERT(!ret); /* -EEXIST; Logic error */
3391
3392         trace_btrfs_setup_cluster(block_group, cluster,
3393                                   total_found * ctl->unit, 1);
3394         return 0;
3395 }
3396
3397 /*
3398  * This searches the block group for just extents to fill the cluster with.
3399  * Try to find a cluster with at least bytes total bytes, at least one
3400  * extent of cont1_bytes, and other clusters of at least min_bytes.
3401  */
3402 static noinline int
3403 setup_cluster_no_bitmap(struct btrfs_block_group *block_group,
3404                         struct btrfs_free_cluster *cluster,
3405                         struct list_head *bitmaps, u64 offset, u64 bytes,
3406                         u64 cont1_bytes, u64 min_bytes)
3407 {
3408         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3409         struct btrfs_free_space *first = NULL;
3410         struct btrfs_free_space *entry = NULL;
3411         struct btrfs_free_space *last;
3412         struct rb_node *node;
3413         u64 window_free;
3414         u64 max_extent;
3415         u64 total_size = 0;
3416
3417         entry = tree_search_offset(ctl, offset, 0, 1);
3418         if (!entry)
3419                 return -ENOSPC;
3420
3421         /*
3422          * We don't want bitmaps, so just move along until we find a normal
3423          * extent entry.
3424          */
3425         while (entry->bitmap || entry->bytes < min_bytes) {
3426                 if (entry->bitmap && list_empty(&entry->list))
3427                         list_add_tail(&entry->list, bitmaps);
3428                 node = rb_next(&entry->offset_index);
3429                 if (!node)
3430                         return -ENOSPC;
3431                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
3432         }
3433
3434         window_free = entry->bytes;
3435         max_extent = entry->bytes;
3436         first = entry;
3437         last = entry;
3438
3439         for (node = rb_next(&entry->offset_index); node;
3440              node = rb_next(&entry->offset_index)) {
3441                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
3442
3443                 if (entry->bitmap) {
3444                         if (list_empty(&entry->list))
3445                                 list_add_tail(&entry->list, bitmaps);
3446                         continue;
3447                 }
3448
3449                 if (entry->bytes < min_bytes)
3450                         continue;
3451
3452                 last = entry;
3453                 window_free += entry->bytes;
3454                 if (entry->bytes > max_extent)
3455                         max_extent = entry->bytes;
3456         }
3457
3458         if (window_free < bytes || max_extent < cont1_bytes)
3459                 return -ENOSPC;
3460
3461         cluster->window_start = first->offset;
3462
3463         node = &first->offset_index;
3464
3465         /*
3466          * now we've found our entries, pull them out of the free space
3467          * cache and put them into the cluster rbtree
3468          */
3469         do {
3470                 int ret;
3471
3472                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
3473                 node = rb_next(&entry->offset_index);
3474                 if (entry->bitmap || entry->bytes < min_bytes)
3475                         continue;
3476
3477                 rb_erase(&entry->offset_index, &ctl->free_space_offset);
3478                 rb_erase_cached(&entry->bytes_index, &ctl->free_space_bytes);
3479                 ret = tree_insert_offset(&cluster->root, entry->offset,
3480                                          &entry->offset_index, 0);
3481                 total_size += entry->bytes;
3482                 ASSERT(!ret); /* -EEXIST; Logic error */
3483         } while (node && entry != last);
3484
3485         cluster->max_size = max_extent;
3486         trace_btrfs_setup_cluster(block_group, cluster, total_size, 0);
3487         return 0;
3488 }
3489
3490 /*
3491  * This specifically looks for bitmaps that may work in the cluster, we assume
3492  * that we have already failed to find extents that will work.
3493  */
3494 static noinline int
3495 setup_cluster_bitmap(struct btrfs_block_group *block_group,
3496                      struct btrfs_free_cluster *cluster,
3497                      struct list_head *bitmaps, u64 offset, u64 bytes,
3498                      u64 cont1_bytes, u64 min_bytes)
3499 {
3500         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3501         struct btrfs_free_space *entry = NULL;
3502         int ret = -ENOSPC;
3503         u64 bitmap_offset = offset_to_bitmap(ctl, offset);
3504
3505         if (ctl->total_bitmaps == 0)
3506                 return -ENOSPC;
3507
3508         /*
3509          * The bitmap that covers offset won't be in the list unless offset
3510          * is just its start offset.
3511          */
3512         if (!list_empty(bitmaps))
3513                 entry = list_first_entry(bitmaps, struct btrfs_free_space, list);
3514
3515         if (!entry || entry->offset != bitmap_offset) {
3516                 entry = tree_search_offset(ctl, bitmap_offset, 1, 0);
3517                 if (entry && list_empty(&entry->list))
3518                         list_add(&entry->list, bitmaps);
3519         }
3520
3521         list_for_each_entry(entry, bitmaps, list) {
3522                 if (entry->bytes < bytes)
3523                         continue;
3524                 ret = btrfs_bitmap_cluster(block_group, entry, cluster, offset,
3525                                            bytes, cont1_bytes, min_bytes);
3526                 if (!ret)
3527                         return 0;
3528         }
3529
3530         /*
3531          * The bitmaps list has all the bitmaps that record free space
3532          * starting after offset, so no more search is required.
3533          */
3534         return -ENOSPC;
3535 }
3536
3537 /*
3538  * here we try to find a cluster of blocks in a block group.  The goal
3539  * is to find at least bytes+empty_size.
3540  * We might not find them all in one contiguous area.
3541  *
3542  * returns zero and sets up cluster if things worked out, otherwise
3543  * it returns -enospc
3544  */
3545 int btrfs_find_space_cluster(struct btrfs_block_group *block_group,
3546                              struct btrfs_free_cluster *cluster,
3547                              u64 offset, u64 bytes, u64 empty_size)
3548 {
3549         struct btrfs_fs_info *fs_info = block_group->fs_info;
3550         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3551         struct btrfs_free_space *entry, *tmp;
3552         LIST_HEAD(bitmaps);
3553         u64 min_bytes;
3554         u64 cont1_bytes;
3555         int ret;
3556
3557         /*
3558          * Choose the minimum extent size we'll require for this
3559          * cluster.  For SSD_SPREAD, don't allow any fragmentation.
3560          * For metadata, allow allocates with smaller extents.  For
3561          * data, keep it dense.
3562          */
3563         if (btrfs_test_opt(fs_info, SSD_SPREAD)) {
3564                 cont1_bytes = bytes + empty_size;
3565                 min_bytes = cont1_bytes;
3566         } else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
3567                 cont1_bytes = bytes;
3568                 min_bytes = fs_info->sectorsize;
3569         } else {
3570                 cont1_bytes = max(bytes, (bytes + empty_size) >> 2);
3571                 min_bytes = fs_info->sectorsize;
3572         }
3573
3574         spin_lock(&ctl->tree_lock);
3575
3576         /*
3577          * If we know we don't have enough space to make a cluster don't even
3578          * bother doing all the work to try and find one.
3579          */
3580         if (ctl->free_space < bytes) {
3581                 spin_unlock(&ctl->tree_lock);
3582                 return -ENOSPC;
3583         }
3584
3585         spin_lock(&cluster->lock);
3586
3587         /* someone already found a cluster, hooray */
3588         if (cluster->block_group) {
3589                 ret = 0;
3590                 goto out;
3591         }
3592
3593         trace_btrfs_find_cluster(block_group, offset, bytes, empty_size,
3594                                  min_bytes);
3595
3596         ret = setup_cluster_no_bitmap(block_group, cluster, &bitmaps, offset,
3597                                       bytes + empty_size,
3598                                       cont1_bytes, min_bytes);
3599         if (ret)
3600                 ret = setup_cluster_bitmap(block_group, cluster, &bitmaps,
3601                                            offset, bytes + empty_size,
3602                                            cont1_bytes, min_bytes);
3603
3604         /* Clear our temporary list */
3605         list_for_each_entry_safe(entry, tmp, &bitmaps, list)
3606                 list_del_init(&entry->list);
3607
3608         if (!ret) {
3609                 btrfs_get_block_group(block_group);
3610                 list_add_tail(&cluster->block_group_list,
3611                               &block_group->cluster_list);
3612                 cluster->block_group = block_group;
3613         } else {
3614                 trace_btrfs_failed_cluster_setup(block_group);
3615         }
3616 out:
3617         spin_unlock(&cluster->lock);
3618         spin_unlock(&ctl->tree_lock);
3619
3620         return ret;
3621 }
3622
3623 /*
3624  * simple code to zero out a cluster
3625  */
3626 void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
3627 {
3628         spin_lock_init(&cluster->lock);
3629         spin_lock_init(&cluster->refill_lock);
3630         cluster->root = RB_ROOT;
3631         cluster->max_size = 0;
3632         cluster->fragmented = false;
3633         INIT_LIST_HEAD(&cluster->block_group_list);
3634         cluster->block_group = NULL;
3635 }
3636
3637 static int do_trimming(struct btrfs_block_group *block_group,
3638                        u64 *total_trimmed, u64 start, u64 bytes,
3639                        u64 reserved_start, u64 reserved_bytes,
3640                        enum btrfs_trim_state reserved_trim_state,
3641                        struct btrfs_trim_range *trim_entry)
3642 {
3643         struct btrfs_space_info *space_info = block_group->space_info;
3644         struct btrfs_fs_info *fs_info = block_group->fs_info;
3645         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3646         int ret;
3647         int update = 0;
3648         const u64 end = start + bytes;
3649         const u64 reserved_end = reserved_start + reserved_bytes;
3650         enum btrfs_trim_state trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
3651         u64 trimmed = 0;
3652
3653         spin_lock(&space_info->lock);
3654         spin_lock(&block_group->lock);
3655         if (!block_group->ro) {
3656                 block_group->reserved += reserved_bytes;
3657                 space_info->bytes_reserved += reserved_bytes;
3658                 update = 1;
3659         }
3660         spin_unlock(&block_group->lock);
3661         spin_unlock(&space_info->lock);
3662
3663         ret = btrfs_discard_extent(fs_info, start, bytes, &trimmed);
3664         if (!ret) {
3665                 *total_trimmed += trimmed;
3666                 trim_state = BTRFS_TRIM_STATE_TRIMMED;
3667         }
3668
3669         mutex_lock(&ctl->cache_writeout_mutex);
3670         if (reserved_start < start)
3671                 __btrfs_add_free_space(block_group, reserved_start,
3672                                        start - reserved_start,
3673                                        reserved_trim_state);
3674         if (start + bytes < reserved_start + reserved_bytes)
3675                 __btrfs_add_free_space(block_group, end, reserved_end - end,
3676                                        reserved_trim_state);
3677         __btrfs_add_free_space(block_group, start, bytes, trim_state);
3678         list_del(&trim_entry->list);
3679         mutex_unlock(&ctl->cache_writeout_mutex);
3680
3681         if (update) {
3682                 spin_lock(&space_info->lock);
3683                 spin_lock(&block_group->lock);
3684                 if (block_group->ro)
3685                         space_info->bytes_readonly += reserved_bytes;
3686                 block_group->reserved -= reserved_bytes;
3687                 space_info->bytes_reserved -= reserved_bytes;
3688                 spin_unlock(&block_group->lock);
3689                 spin_unlock(&space_info->lock);
3690         }
3691
3692         return ret;
3693 }
3694
3695 /*
3696  * If @async is set, then we will trim 1 region and return.
3697  */
3698 static int trim_no_bitmap(struct btrfs_block_group *block_group,
3699                           u64 *total_trimmed, u64 start, u64 end, u64 minlen,
3700                           bool async)
3701 {
3702         struct btrfs_discard_ctl *discard_ctl =
3703                                         &block_group->fs_info->discard_ctl;
3704         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3705         struct btrfs_free_space *entry;
3706         struct rb_node *node;
3707         int ret = 0;
3708         u64 extent_start;
3709         u64 extent_bytes;
3710         enum btrfs_trim_state extent_trim_state;
3711         u64 bytes;
3712         const u64 max_discard_size = READ_ONCE(discard_ctl->max_discard_size);
3713
3714         while (start < end) {
3715                 struct btrfs_trim_range trim_entry;
3716
3717                 mutex_lock(&ctl->cache_writeout_mutex);
3718                 spin_lock(&ctl->tree_lock);
3719
3720                 if (ctl->free_space < minlen)
3721                         goto out_unlock;
3722
3723                 entry = tree_search_offset(ctl, start, 0, 1);
3724                 if (!entry)
3725                         goto out_unlock;
3726
3727                 /* Skip bitmaps and if async, already trimmed entries */
3728                 while (entry->bitmap ||
3729                        (async && btrfs_free_space_trimmed(entry))) {
3730                         node = rb_next(&entry->offset_index);
3731                         if (!node)
3732                                 goto out_unlock;
3733                         entry = rb_entry(node, struct btrfs_free_space,
3734                                          offset_index);
3735                 }
3736
3737                 if (entry->offset >= end)
3738                         goto out_unlock;
3739
3740                 extent_start = entry->offset;
3741                 extent_bytes = entry->bytes;
3742                 extent_trim_state = entry->trim_state;
3743                 if (async) {
3744                         start = entry->offset;
3745                         bytes = entry->bytes;
3746                         if (bytes < minlen) {
3747                                 spin_unlock(&ctl->tree_lock);
3748                                 mutex_unlock(&ctl->cache_writeout_mutex);
3749                                 goto next;
3750                         }
3751                         unlink_free_space(ctl, entry, true);
3752                         /*
3753                          * Let bytes = BTRFS_MAX_DISCARD_SIZE + X.
3754                          * If X < BTRFS_ASYNC_DISCARD_MIN_FILTER, we won't trim
3755                          * X when we come back around.  So trim it now.
3756                          */
3757                         if (max_discard_size &&
3758                             bytes >= (max_discard_size +
3759                                       BTRFS_ASYNC_DISCARD_MIN_FILTER)) {
3760                                 bytes = max_discard_size;
3761                                 extent_bytes = max_discard_size;
3762                                 entry->offset += max_discard_size;
3763                                 entry->bytes -= max_discard_size;
3764                                 link_free_space(ctl, entry);
3765                         } else {
3766                                 kmem_cache_free(btrfs_free_space_cachep, entry);
3767                         }
3768                 } else {
3769                         start = max(start, extent_start);
3770                         bytes = min(extent_start + extent_bytes, end) - start;
3771                         if (bytes < minlen) {
3772                                 spin_unlock(&ctl->tree_lock);
3773                                 mutex_unlock(&ctl->cache_writeout_mutex);
3774                                 goto next;
3775                         }
3776
3777                         unlink_free_space(ctl, entry, true);
3778                         kmem_cache_free(btrfs_free_space_cachep, entry);
3779                 }
3780
3781                 spin_unlock(&ctl->tree_lock);
3782                 trim_entry.start = extent_start;
3783                 trim_entry.bytes = extent_bytes;
3784                 list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
3785                 mutex_unlock(&ctl->cache_writeout_mutex);
3786
3787                 ret = do_trimming(block_group, total_trimmed, start, bytes,
3788                                   extent_start, extent_bytes, extent_trim_state,
3789                                   &trim_entry);
3790                 if (ret) {
3791                         block_group->discard_cursor = start + bytes;
3792                         break;
3793                 }
3794 next:
3795                 start += bytes;
3796                 block_group->discard_cursor = start;
3797                 if (async && *total_trimmed)
3798                         break;
3799
3800                 if (fatal_signal_pending(current)) {
3801                         ret = -ERESTARTSYS;
3802                         break;
3803                 }
3804
3805                 cond_resched();
3806         }
3807
3808         return ret;
3809
3810 out_unlock:
3811         block_group->discard_cursor = btrfs_block_group_end(block_group);
3812         spin_unlock(&ctl->tree_lock);
3813         mutex_unlock(&ctl->cache_writeout_mutex);
3814
3815         return ret;
3816 }
3817
3818 /*
3819  * If we break out of trimming a bitmap prematurely, we should reset the
3820  * trimming bit.  In a rather contrieved case, it's possible to race here so
3821  * reset the state to BTRFS_TRIM_STATE_UNTRIMMED.
3822  *
3823  * start = start of bitmap
3824  * end = near end of bitmap
3825  *
3826  * Thread 1:                    Thread 2:
3827  * trim_bitmaps(start)
3828  *                              trim_bitmaps(end)
3829  *                              end_trimming_bitmap()
3830  * reset_trimming_bitmap()
3831  */
3832 static void reset_trimming_bitmap(struct btrfs_free_space_ctl *ctl, u64 offset)
3833 {
3834         struct btrfs_free_space *entry;
3835
3836         spin_lock(&ctl->tree_lock);
3837         entry = tree_search_offset(ctl, offset, 1, 0);
3838         if (entry) {
3839                 if (btrfs_free_space_trimmed(entry)) {
3840                         ctl->discardable_extents[BTRFS_STAT_CURR] +=
3841                                 entry->bitmap_extents;
3842                         ctl->discardable_bytes[BTRFS_STAT_CURR] += entry->bytes;
3843                 }
3844                 entry->trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
3845         }
3846
3847         spin_unlock(&ctl->tree_lock);
3848 }
3849
3850 static void end_trimming_bitmap(struct btrfs_free_space_ctl *ctl,
3851                                 struct btrfs_free_space *entry)
3852 {
3853         if (btrfs_free_space_trimming_bitmap(entry)) {
3854                 entry->trim_state = BTRFS_TRIM_STATE_TRIMMED;
3855                 ctl->discardable_extents[BTRFS_STAT_CURR] -=
3856                         entry->bitmap_extents;
3857                 ctl->discardable_bytes[BTRFS_STAT_CURR] -= entry->bytes;
3858         }
3859 }
3860
3861 /*
3862  * If @async is set, then we will trim 1 region and return.
3863  */
3864 static int trim_bitmaps(struct btrfs_block_group *block_group,
3865                         u64 *total_trimmed, u64 start, u64 end, u64 minlen,
3866                         u64 maxlen, bool async)
3867 {
3868         struct btrfs_discard_ctl *discard_ctl =
3869                                         &block_group->fs_info->discard_ctl;
3870         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3871         struct btrfs_free_space *entry;
3872         int ret = 0;
3873         int ret2;
3874         u64 bytes;
3875         u64 offset = offset_to_bitmap(ctl, start);
3876         const u64 max_discard_size = READ_ONCE(discard_ctl->max_discard_size);
3877
3878         while (offset < end) {
3879                 bool next_bitmap = false;
3880                 struct btrfs_trim_range trim_entry;
3881
3882                 mutex_lock(&ctl->cache_writeout_mutex);
3883                 spin_lock(&ctl->tree_lock);
3884
3885                 if (ctl->free_space < minlen) {
3886                         block_group->discard_cursor =
3887                                 btrfs_block_group_end(block_group);
3888                         spin_unlock(&ctl->tree_lock);
3889                         mutex_unlock(&ctl->cache_writeout_mutex);
3890                         break;
3891                 }
3892
3893                 entry = tree_search_offset(ctl, offset, 1, 0);
3894                 /*
3895                  * Bitmaps are marked trimmed lossily now to prevent constant
3896                  * discarding of the same bitmap (the reason why we are bound
3897                  * by the filters).  So, retrim the block group bitmaps when we
3898                  * are preparing to punt to the unused_bgs list.  This uses
3899                  * @minlen to determine if we are in BTRFS_DISCARD_INDEX_UNUSED
3900                  * which is the only discard index which sets minlen to 0.
3901                  */
3902                 if (!entry || (async && minlen && start == offset &&
3903                                btrfs_free_space_trimmed(entry))) {
3904                         spin_unlock(&ctl->tree_lock);
3905                         mutex_unlock(&ctl->cache_writeout_mutex);
3906                         next_bitmap = true;
3907                         goto next;
3908                 }
3909
3910                 /*
3911                  * Async discard bitmap trimming begins at by setting the start
3912                  * to be key.objectid and the offset_to_bitmap() aligns to the
3913                  * start of the bitmap.  This lets us know we are fully
3914                  * scanning the bitmap rather than only some portion of it.
3915                  */
3916                 if (start == offset)
3917                         entry->trim_state = BTRFS_TRIM_STATE_TRIMMING;
3918
3919                 bytes = minlen;
3920                 ret2 = search_bitmap(ctl, entry, &start, &bytes, false);
3921                 if (ret2 || start >= end) {
3922                         /*
3923                          * We lossily consider a bitmap trimmed if we only skip
3924                          * over regions <= BTRFS_ASYNC_DISCARD_MIN_FILTER.
3925                          */
3926                         if (ret2 && minlen <= BTRFS_ASYNC_DISCARD_MIN_FILTER)
3927                                 end_trimming_bitmap(ctl, entry);
3928                         else
3929                                 entry->trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
3930                         spin_unlock(&ctl->tree_lock);
3931                         mutex_unlock(&ctl->cache_writeout_mutex);
3932                         next_bitmap = true;
3933                         goto next;
3934                 }
3935
3936                 /*
3937                  * We already trimmed a region, but are using the locking above
3938                  * to reset the trim_state.
3939                  */
3940                 if (async && *total_trimmed) {
3941                         spin_unlock(&ctl->tree_lock);
3942                         mutex_unlock(&ctl->cache_writeout_mutex);
3943                         goto out;
3944                 }
3945
3946                 bytes = min(bytes, end - start);
3947                 if (bytes < minlen || (async && maxlen && bytes > maxlen)) {
3948                         spin_unlock(&ctl->tree_lock);
3949                         mutex_unlock(&ctl->cache_writeout_mutex);
3950                         goto next;
3951                 }
3952
3953                 /*
3954                  * Let bytes = BTRFS_MAX_DISCARD_SIZE + X.
3955                  * If X < @minlen, we won't trim X when we come back around.
3956                  * So trim it now.  We differ here from trimming extents as we
3957                  * don't keep individual state per bit.
3958                  */
3959                 if (async &&
3960                     max_discard_size &&
3961                     bytes > (max_discard_size + minlen))
3962                         bytes = max_discard_size;
3963
3964                 bitmap_clear_bits(ctl, entry, start, bytes, true);
3965                 if (entry->bytes == 0)
3966                         free_bitmap(ctl, entry);
3967
3968                 spin_unlock(&ctl->tree_lock);
3969                 trim_entry.start = start;
3970                 trim_entry.bytes = bytes;
3971                 list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
3972                 mutex_unlock(&ctl->cache_writeout_mutex);
3973
3974                 ret = do_trimming(block_group, total_trimmed, start, bytes,
3975                                   start, bytes, 0, &trim_entry);
3976                 if (ret) {
3977                         reset_trimming_bitmap(ctl, offset);
3978                         block_group->discard_cursor =
3979                                 btrfs_block_group_end(block_group);
3980                         break;
3981                 }
3982 next:
3983                 if (next_bitmap) {
3984                         offset += BITS_PER_BITMAP * ctl->unit;
3985                         start = offset;
3986                 } else {
3987                         start += bytes;
3988                 }
3989                 block_group->discard_cursor = start;
3990
3991                 if (fatal_signal_pending(current)) {
3992                         if (start != offset)
3993                                 reset_trimming_bitmap(ctl, offset);
3994                         ret = -ERESTARTSYS;
3995                         break;
3996                 }
3997
3998                 cond_resched();
3999         }
4000
4001         if (offset >= end)
4002                 block_group->discard_cursor = end;
4003
4004 out:
4005         return ret;
4006 }
4007
4008 int btrfs_trim_block_group(struct btrfs_block_group *block_group,
4009                            u64 *trimmed, u64 start, u64 end, u64 minlen)
4010 {
4011         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
4012         int ret;
4013         u64 rem = 0;
4014
4015         ASSERT(!btrfs_is_zoned(block_group->fs_info));
4016
4017         *trimmed = 0;
4018
4019         spin_lock(&block_group->lock);
4020         if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags)) {
4021                 spin_unlock(&block_group->lock);
4022                 return 0;
4023         }
4024         btrfs_freeze_block_group(block_group);
4025         spin_unlock(&block_group->lock);
4026
4027         ret = trim_no_bitmap(block_group, trimmed, start, end, minlen, false);
4028         if (ret)
4029                 goto out;
4030
4031         ret = trim_bitmaps(block_group, trimmed, start, end, minlen, 0, false);
4032         div64_u64_rem(end, BITS_PER_BITMAP * ctl->unit, &rem);
4033         /* If we ended in the middle of a bitmap, reset the trimming flag */
4034         if (rem)
4035                 reset_trimming_bitmap(ctl, offset_to_bitmap(ctl, end));
4036 out:
4037         btrfs_unfreeze_block_group(block_group);
4038         return ret;
4039 }
4040
4041 int btrfs_trim_block_group_extents(struct btrfs_block_group *block_group,
4042                                    u64 *trimmed, u64 start, u64 end, u64 minlen,
4043                                    bool async)
4044 {
4045         int ret;
4046
4047         *trimmed = 0;
4048
4049         spin_lock(&block_group->lock);
4050         if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags)) {
4051                 spin_unlock(&block_group->lock);
4052                 return 0;
4053         }
4054         btrfs_freeze_block_group(block_group);
4055         spin_unlock(&block_group->lock);
4056
4057         ret = trim_no_bitmap(block_group, trimmed, start, end, minlen, async);
4058         btrfs_unfreeze_block_group(block_group);
4059
4060         return ret;
4061 }
4062
4063 int btrfs_trim_block_group_bitmaps(struct btrfs_block_group *block_group,
4064                                    u64 *trimmed, u64 start, u64 end, u64 minlen,
4065                                    u64 maxlen, bool async)
4066 {
4067         int ret;
4068
4069         *trimmed = 0;
4070
4071         spin_lock(&block_group->lock);
4072         if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags)) {
4073                 spin_unlock(&block_group->lock);
4074                 return 0;
4075         }
4076         btrfs_freeze_block_group(block_group);
4077         spin_unlock(&block_group->lock);
4078
4079         ret = trim_bitmaps(block_group, trimmed, start, end, minlen, maxlen,
4080                            async);
4081
4082         btrfs_unfreeze_block_group(block_group);
4083
4084         return ret;
4085 }
4086
4087 bool btrfs_free_space_cache_v1_active(struct btrfs_fs_info *fs_info)
4088 {
4089         return btrfs_super_cache_generation(fs_info->super_copy);
4090 }
4091
4092 static int cleanup_free_space_cache_v1(struct btrfs_fs_info *fs_info,
4093                                        struct btrfs_trans_handle *trans)
4094 {
4095         struct btrfs_block_group *block_group;
4096         struct rb_node *node;
4097         int ret = 0;
4098
4099         btrfs_info(fs_info, "cleaning free space cache v1");
4100
4101         node = rb_first_cached(&fs_info->block_group_cache_tree);
4102         while (node) {
4103                 block_group = rb_entry(node, struct btrfs_block_group, cache_node);
4104                 ret = btrfs_remove_free_space_inode(trans, NULL, block_group);
4105                 if (ret)
4106                         goto out;
4107                 node = rb_next(node);
4108         }
4109 out:
4110         return ret;
4111 }
4112
4113 int btrfs_set_free_space_cache_v1_active(struct btrfs_fs_info *fs_info, bool active)
4114 {
4115         struct btrfs_trans_handle *trans;
4116         int ret;
4117
4118         /*
4119          * update_super_roots will appropriately set or unset
4120          * super_copy->cache_generation based on SPACE_CACHE and
4121          * BTRFS_FS_CLEANUP_SPACE_CACHE_V1. For this reason, we need a
4122          * transaction commit whether we are enabling space cache v1 and don't
4123          * have any other work to do, or are disabling it and removing free
4124          * space inodes.
4125          */
4126         trans = btrfs_start_transaction(fs_info->tree_root, 0);
4127         if (IS_ERR(trans))
4128                 return PTR_ERR(trans);
4129
4130         if (!active) {
4131                 set_bit(BTRFS_FS_CLEANUP_SPACE_CACHE_V1, &fs_info->flags);
4132                 ret = cleanup_free_space_cache_v1(fs_info, trans);
4133                 if (ret) {
4134                         btrfs_abort_transaction(trans, ret);
4135                         btrfs_end_transaction(trans);
4136                         goto out;
4137                 }
4138         }
4139
4140         ret = btrfs_commit_transaction(trans);
4141 out:
4142         clear_bit(BTRFS_FS_CLEANUP_SPACE_CACHE_V1, &fs_info->flags);
4143
4144         return ret;
4145 }
4146
4147 int __init btrfs_free_space_init(void)
4148 {
4149         btrfs_free_space_cachep = kmem_cache_create("btrfs_free_space",
4150                         sizeof(struct btrfs_free_space), 0,
4151                         SLAB_MEM_SPREAD, NULL);
4152         if (!btrfs_free_space_cachep)
4153                 return -ENOMEM;
4154
4155         btrfs_free_space_bitmap_cachep = kmem_cache_create("btrfs_free_space_bitmap",
4156                                                         PAGE_SIZE, PAGE_SIZE,
4157                                                         SLAB_MEM_SPREAD, NULL);
4158         if (!btrfs_free_space_bitmap_cachep) {
4159                 kmem_cache_destroy(btrfs_free_space_cachep);
4160                 return -ENOMEM;
4161         }
4162
4163         return 0;
4164 }
4165
4166 void __cold btrfs_free_space_exit(void)
4167 {
4168         kmem_cache_destroy(btrfs_free_space_cachep);
4169         kmem_cache_destroy(btrfs_free_space_bitmap_cachep);
4170 }
4171
4172 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4173 /*
4174  * Use this if you need to make a bitmap or extent entry specifically, it
4175  * doesn't do any of the merging that add_free_space does, this acts a lot like
4176  * how the free space cache loading stuff works, so you can get really weird
4177  * configurations.
4178  */
4179 int test_add_free_space_entry(struct btrfs_block_group *cache,
4180                               u64 offset, u64 bytes, bool bitmap)
4181 {
4182         struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
4183         struct btrfs_free_space *info = NULL, *bitmap_info;
4184         void *map = NULL;
4185         enum btrfs_trim_state trim_state = BTRFS_TRIM_STATE_TRIMMED;
4186         u64 bytes_added;
4187         int ret;
4188
4189 again:
4190         if (!info) {
4191                 info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
4192                 if (!info)
4193                         return -ENOMEM;
4194         }
4195
4196         if (!bitmap) {
4197                 spin_lock(&ctl->tree_lock);
4198                 info->offset = offset;
4199                 info->bytes = bytes;
4200                 info->max_extent_size = 0;
4201                 ret = link_free_space(ctl, info);
4202                 spin_unlock(&ctl->tree_lock);
4203                 if (ret)
4204                         kmem_cache_free(btrfs_free_space_cachep, info);
4205                 return ret;
4206         }
4207
4208         if (!map) {
4209                 map = kmem_cache_zalloc(btrfs_free_space_bitmap_cachep, GFP_NOFS);
4210                 if (!map) {
4211                         kmem_cache_free(btrfs_free_space_cachep, info);
4212                         return -ENOMEM;
4213                 }
4214         }
4215
4216         spin_lock(&ctl->tree_lock);
4217         bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
4218                                          1, 0);
4219         if (!bitmap_info) {
4220                 info->bitmap = map;
4221                 map = NULL;
4222                 add_new_bitmap(ctl, info, offset);
4223                 bitmap_info = info;
4224                 info = NULL;
4225         }
4226
4227         bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes,
4228                                           trim_state);
4229
4230         bytes -= bytes_added;
4231         offset += bytes_added;
4232         spin_unlock(&ctl->tree_lock);
4233
4234         if (bytes)
4235                 goto again;
4236
4237         if (info)
4238                 kmem_cache_free(btrfs_free_space_cachep, info);
4239         if (map)
4240                 kmem_cache_free(btrfs_free_space_bitmap_cachep, map);
4241         return 0;
4242 }
4243
4244 /*
4245  * Checks to see if the given range is in the free space cache.  This is really
4246  * just used to check the absence of space, so if there is free space in the
4247  * range at all we will return 1.
4248  */
4249 int test_check_exists(struct btrfs_block_group *cache,
4250                       u64 offset, u64 bytes)
4251 {
4252         struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
4253         struct btrfs_free_space *info;
4254         int ret = 0;
4255
4256         spin_lock(&ctl->tree_lock);
4257         info = tree_search_offset(ctl, offset, 0, 0);
4258         if (!info) {
4259                 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
4260                                           1, 0);
4261                 if (!info)
4262                         goto out;
4263         }
4264
4265 have_info:
4266         if (info->bitmap) {
4267                 u64 bit_off, bit_bytes;
4268                 struct rb_node *n;
4269                 struct btrfs_free_space *tmp;
4270
4271                 bit_off = offset;
4272                 bit_bytes = ctl->unit;
4273                 ret = search_bitmap(ctl, info, &bit_off, &bit_bytes, false);
4274                 if (!ret) {
4275                         if (bit_off == offset) {
4276                                 ret = 1;
4277                                 goto out;
4278                         } else if (bit_off > offset &&
4279                                    offset + bytes > bit_off) {
4280                                 ret = 1;
4281                                 goto out;
4282                         }
4283                 }
4284
4285                 n = rb_prev(&info->offset_index);
4286                 while (n) {
4287                         tmp = rb_entry(n, struct btrfs_free_space,
4288                                        offset_index);
4289                         if (tmp->offset + tmp->bytes < offset)
4290                                 break;
4291                         if (offset + bytes < tmp->offset) {
4292                                 n = rb_prev(&tmp->offset_index);
4293                                 continue;
4294                         }
4295                         info = tmp;
4296                         goto have_info;
4297                 }
4298
4299                 n = rb_next(&info->offset_index);
4300                 while (n) {
4301                         tmp = rb_entry(n, struct btrfs_free_space,
4302                                        offset_index);
4303                         if (offset + bytes < tmp->offset)
4304                                 break;
4305                         if (tmp->offset + tmp->bytes < offset) {
4306                                 n = rb_next(&tmp->offset_index);
4307                                 continue;
4308                         }
4309                         info = tmp;
4310                         goto have_info;
4311                 }
4312
4313                 ret = 0;
4314                 goto out;
4315         }
4316
4317         if (info->offset == offset) {
4318                 ret = 1;
4319                 goto out;
4320         }
4321
4322         if (offset > info->offset && offset < info->offset + info->bytes)
4323                 ret = 1;
4324 out:
4325         spin_unlock(&ctl->tree_lock);
4326         return ret;
4327 }
4328 #endif /* CONFIG_BTRFS_FS_RUN_SANITY_TESTS */