fs/btrfs/qgroup.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2011 STRATO.  All rights reserved.
   4  */
   5
   6 #include <linux/sched.h>
   7 #include <linux/pagemap.h>
   8 #include <linux/writeback.h>
   9 #include <linux/blkdev.h>
  10 #include <linux/rbtree.h>
  11 #include <linux/slab.h>
  12 #include <linux/workqueue.h>
  13 #include <linux/btrfs.h>
  14 #include <linux/sizes.h>
  15
  16 #include "ctree.h"
  17 #include "transaction.h"
  18 #include "disk-io.h"
  19 #include "locking.h"
  20 #include "ulist.h"
  21 #include "backref.h"
  22 #include "extent_io.h"
  23 #include "qgroup.h"
  24
  25
  26 /* TODO XXX FIXME
  27  *  - subvol delete -> delete when ref goes to 0? delete limits also?
  28  *  - reorganize keys
  29  *  - compressed
  30  *  - sync
  31  *  - copy also limits on subvol creation
  32  *  - limit
  33  *  - caches for ulists
  34  *  - performance benchmarks
  35  *  - check all ioctl parameters
  36  */
  37
  38 /*
  39  * Helpers to access qgroup reservation
  40  *
  41  * Callers should ensure the lock context and type are valid
  42  */
  43
  44 static u64 qgroup_rsv_total(const struct btrfs_qgroup *qgroup)
  45 {
  46         u64 ret = 0;
  47         int i;
  48
  49         for (i = 0; i < BTRFS_QGROUP_RSV_LAST; i++)
  50                 ret += qgroup->rsv.values[i];
  51
  52         return ret;
  53 }
  54
  55 #ifdef CONFIG_BTRFS_DEBUG
  56 static const char *qgroup_rsv_type_str(enum btrfs_qgroup_rsv_type type)
  57 {
  58         if (type == BTRFS_QGROUP_RSV_DATA)
  59                 return "data";
  60         if (type == BTRFS_QGROUP_RSV_META_PERTRANS)
  61                 return "meta_pertrans";
  62         if (type == BTRFS_QGROUP_RSV_META_PREALLOC)
  63                 return "meta_prealloc";
  64         return NULL;
  65 }
  66 #endif
  67
  68 static void qgroup_rsv_add(struct btrfs_fs_info *fs_info,
  69                            struct btrfs_qgroup *qgroup, u64 num_bytes,
  70                            enum btrfs_qgroup_rsv_type type)
  71 {
  72         trace_qgroup_update_reserve(fs_info, qgroup, num_bytes, type);
  73         qgroup->rsv.values[type] += num_bytes;
  74 }
  75
  76 static void qgroup_rsv_release(struct btrfs_fs_info *fs_info,
  77                                struct btrfs_qgroup *qgroup, u64 num_bytes,
  78                                enum btrfs_qgroup_rsv_type type)
  79 {
  80         trace_qgroup_update_reserve(fs_info, qgroup, -(s64)num_bytes, type);
  81         if (qgroup->rsv.values[type] >= num_bytes) {
  82                 qgroup->rsv.values[type] -= num_bytes;
  83                 return;
  84         }
  85 #ifdef CONFIG_BTRFS_DEBUG
  86         WARN_RATELIMIT(1,
  87                 "qgroup %llu %s reserved space underflow, have %llu to free %llu",
  88                 qgroup->qgroupid, qgroup_rsv_type_str(type),
  89                 qgroup->rsv.values[type], num_bytes);
  90 #endif
  91         qgroup->rsv.values[type] = 0;
  92 }
  93
  94 static void qgroup_rsv_add_by_qgroup(struct btrfs_fs_info *fs_info,
  95                                      struct btrfs_qgroup *dest,
  96                                      struct btrfs_qgroup *src)
  97 {
  98         int i;
  99
 100         for (i = 0; i < BTRFS_QGROUP_RSV_LAST; i++)
 101                 qgroup_rsv_add(fs_info, dest, src->rsv.values[i], i);
 102 }
 103
 104 static void qgroup_rsv_release_by_qgroup(struct btrfs_fs_info *fs_info,
 105                                          struct btrfs_qgroup *dest,
 106                                           struct btrfs_qgroup *src)
 107 {
 108         int i;
 109
 110         for (i = 0; i < BTRFS_QGROUP_RSV_LAST; i++)
 111                 qgroup_rsv_release(fs_info, dest, src->rsv.values[i], i);
 112 }
 113
 114 static void btrfs_qgroup_update_old_refcnt(struct btrfs_qgroup *qg, u64 seq,
 115                                            int mod)
 116 {
 117         if (qg->old_refcnt < seq)
 118                 qg->old_refcnt = seq;
 119         qg->old_refcnt += mod;
 120 }
 121
 122 static void btrfs_qgroup_update_new_refcnt(struct btrfs_qgroup *qg, u64 seq,
 123                                            int mod)
 124 {
 125         if (qg->new_refcnt < seq)
 126                 qg->new_refcnt = seq;
 127         qg->new_refcnt += mod;
 128 }
 129
 130 static inline u64 btrfs_qgroup_get_old_refcnt(struct btrfs_qgroup *qg, u64 seq)
 131 {
 132         if (qg->old_refcnt < seq)
 133                 return 0;
 134         return qg->old_refcnt - seq;
 135 }
 136
 137 static inline u64 btrfs_qgroup_get_new_refcnt(struct btrfs_qgroup *qg, u64 seq)
 138 {
 139         if (qg->new_refcnt < seq)
 140                 return 0;
 141         return qg->new_refcnt - seq;
 142 }
 143
 144 /*
 145  * glue structure to represent the relations between qgroups.
 146  */
 147 struct btrfs_qgroup_list {
 148         struct list_head next_group;
 149         struct list_head next_member;
 150         struct btrfs_qgroup *group;
 151         struct btrfs_qgroup *member;
 152 };
 153
 154 static inline u64 qgroup_to_aux(struct btrfs_qgroup *qg)
 155 {
 156         return (u64)(uintptr_t)qg;
 157 }
 158
 159 static inline struct btrfs_qgroup* unode_aux_to_qgroup(struct ulist_node *n)
 160 {
 161         return (struct btrfs_qgroup *)(uintptr_t)n->aux;
 162 }
 163
 164 static int
 165 qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid,
 166                    int init_flags);
 167 static void qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info);
 168
 169 /* must be called with qgroup_ioctl_lock held */
 170 static struct btrfs_qgroup *find_qgroup_rb(struct btrfs_fs_info *fs_info,
 171                                            u64 qgroupid)
 172 {
 173         struct rb_node *n = fs_info->qgroup_tree.rb_node;
 174         struct btrfs_qgroup *qgroup;
 175
 176         while (n) {
 177                 qgroup = rb_entry(n, struct btrfs_qgroup, node);
 178                 if (qgroup->qgroupid < qgroupid)
 179                         n = n->rb_left;
 180                 else if (qgroup->qgroupid > qgroupid)
 181                         n = n->rb_right;
 182                 else
 183                         return qgroup;
 184         }
 185         return NULL;
 186 }
 187
 188 /* must be called with qgroup_lock held */
 189 static struct btrfs_qgroup *add_qgroup_rb(struct btrfs_fs_info *fs_info,
 190                                           u64 qgroupid)
 191 {
 192         struct rb_node **p = &fs_info->qgroup_tree.rb_node;
 193         struct rb_node *parent = NULL;
 194         struct btrfs_qgroup *qgroup;
 195
 196         while (*p) {
 197                 parent = *p;
 198                 qgroup = rb_entry(parent, struct btrfs_qgroup, node);
 199
 200                 if (qgroup->qgroupid < qgroupid)
 201                         p = &(*p)->rb_left;
 202                 else if (qgroup->qgroupid > qgroupid)
 203                         p = &(*p)->rb_right;
 204                 else
 205                         return qgroup;
 206         }
 207
 208         qgroup = kzalloc(sizeof(*qgroup), GFP_ATOMIC);
 209         if (!qgroup)
 210                 return ERR_PTR(-ENOMEM);
 211
 212         qgroup->qgroupid = qgroupid;
 213         INIT_LIST_HEAD(&qgroup->groups);
 214         INIT_LIST_HEAD(&qgroup->members);
 215         INIT_LIST_HEAD(&qgroup->dirty);
 216
 217         rb_link_node(&qgroup->node, parent, p);
 218         rb_insert_color(&qgroup->node, &fs_info->qgroup_tree);
 219
 220         return qgroup;
 221 }
 222
 223 static void __del_qgroup_rb(struct btrfs_qgroup *qgroup)
 224 {
 225         struct btrfs_qgroup_list *list;
 226
 227         list_del(&qgroup->dirty);
 228         while (!list_empty(&qgroup->groups)) {
 229                 list = list_first_entry(&qgroup->groups,
 230                                         struct btrfs_qgroup_list, next_group);
 231                 list_del(&list->next_group);
 232                 list_del(&list->next_member);
 233                 kfree(list);
 234         }
 235
 236         while (!list_empty(&qgroup->members)) {
 237                 list = list_first_entry(&qgroup->members,
 238                                         struct btrfs_qgroup_list, next_member);
 239                 list_del(&list->next_group);
 240                 list_del(&list->next_member);
 241                 kfree(list);
 242         }
 243         kfree(qgroup);
 244 }
 245
 246 /* must be called with qgroup_lock held */
 247 static int del_qgroup_rb(struct btrfs_fs_info *fs_info, u64 qgroupid)
 248 {
 249         struct btrfs_qgroup *qgroup = find_qgroup_rb(fs_info, qgroupid);
 250
 251         if (!qgroup)
 252                 return -ENOENT;
 253
 254         rb_erase(&qgroup->node, &fs_info->qgroup_tree);
 255         __del_qgroup_rb(qgroup);
 256         return 0;
 257 }
 258
 259 /* must be called with qgroup_lock held */
 260 static int add_relation_rb(struct btrfs_fs_info *fs_info,
 261                            u64 memberid, u64 parentid)
 262 {
 263         struct btrfs_qgroup *member;
 264         struct btrfs_qgroup *parent;
 265         struct btrfs_qgroup_list *list;
 266
 267         member = find_qgroup_rb(fs_info, memberid);
 268         parent = find_qgroup_rb(fs_info, parentid);
 269         if (!member || !parent)
 270                 return -ENOENT;
 271
 272         list = kzalloc(sizeof(*list), GFP_ATOMIC);
 273         if (!list)
 274                 return -ENOMEM;
 275
 276         list->group = parent;
 277         list->member = member;
 278         list_add_tail(&list->next_group, &member->groups);
 279         list_add_tail(&list->next_member, &parent->members);
 280
 281         return 0;
 282 }
 283
 284 /* must be called with qgroup_lock held */
 285 static int del_relation_rb(struct btrfs_fs_info *fs_info,
 286                            u64 memberid, u64 parentid)
 287 {
 288         struct btrfs_qgroup *member;
 289         struct btrfs_qgroup *parent;
 290         struct btrfs_qgroup_list *list;
 291
 292         member = find_qgroup_rb(fs_info, memberid);
 293         parent = find_qgroup_rb(fs_info, parentid);
 294         if (!member || !parent)
 295                 return -ENOENT;
 296
 297         list_for_each_entry(list, &member->groups, next_group) {
 298                 if (list->group == parent) {
 299                         list_del(&list->next_group);
 300                         list_del(&list->next_member);
 301                         kfree(list);
 302                         return 0;
 303                 }
 304         }
 305         return -ENOENT;
 306 }
 307
 308 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
 309 int btrfs_verify_qgroup_counts(struct btrfs_fs_info *fs_info, u64 qgroupid,
 310                                u64 rfer, u64 excl)
 311 {
 312         struct btrfs_qgroup *qgroup;
 313
 314         qgroup = find_qgroup_rb(fs_info, qgroupid);
 315         if (!qgroup)
 316                 return -EINVAL;
 317         if (qgroup->rfer != rfer || qgroup->excl != excl)
 318                 return -EINVAL;
 319         return 0;
 320 }
 321 #endif
 322
 323 /*
 324  * The full config is read in one go, only called from open_ctree()
 325  * It doesn't use any locking, as at this point we're still single-threaded
 326  */
 327 int btrfs_read_qgroup_config(struct btrfs_fs_info *fs_info)
 328 {
 329         struct btrfs_key key;
 330         struct btrfs_key found_key;
 331         struct btrfs_root *quota_root = fs_info->quota_root;
 332         struct btrfs_path *path = NULL;
 333         struct extent_buffer *l;
 334         int slot;
 335         int ret = 0;
 336         u64 flags = 0;
 337         u64 rescan_progress = 0;
 338
 339         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
 340                 return 0;
 341
 342         fs_info->qgroup_ulist = ulist_alloc(GFP_KERNEL);
 343         if (!fs_info->qgroup_ulist) {
 344                 ret = -ENOMEM;
 345                 goto out;
 346         }
 347
 348         path = btrfs_alloc_path();
 349         if (!path) {
 350                 ret = -ENOMEM;
 351                 goto out;
 352         }
 353
 354         /* default this to quota off, in case no status key is found */
 355         fs_info->qgroup_flags = 0;
 356
 357         /*
 358          * pass 1: read status, all qgroup infos and limits
 359          */
 360         key.objectid = 0;
 361         key.type = 0;
 362         key.offset = 0;
 363         ret = btrfs_search_slot_for_read(quota_root, &key, path, 1, 1);
 364         if (ret)
 365                 goto out;
 366
 367         while (1) {
 368                 struct btrfs_qgroup *qgroup;
 369
 370                 slot = path->slots[0];
 371                 l = path->nodes[0];
 372                 btrfs_item_key_to_cpu(l, &found_key, slot);
 373
 374                 if (found_key.type == BTRFS_QGROUP_STATUS_KEY) {
 375                         struct btrfs_qgroup_status_item *ptr;
 376
 377                         ptr = btrfs_item_ptr(l, slot,
 378                                              struct btrfs_qgroup_status_item);
 379
 380                         if (btrfs_qgroup_status_version(l, ptr) !=
 381                             BTRFS_QGROUP_STATUS_VERSION) {
 382                                 btrfs_err(fs_info,
 383                                  "old qgroup version, quota disabled");
 384                                 goto out;
 385                         }
 386                         if (btrfs_qgroup_status_generation(l, ptr) !=
 387                             fs_info->generation) {
 388                                 flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
 389                                 btrfs_err(fs_info,
 390                                         "qgroup generation mismatch, marked as inconsistent");
 391                         }
 392                         fs_info->qgroup_flags = btrfs_qgroup_status_flags(l,
 393                                                                           ptr);
 394                         rescan_progress = btrfs_qgroup_status_rescan(l, ptr);
 395                         goto next1;
 396                 }
 397
 398                 if (found_key.type != BTRFS_QGROUP_INFO_KEY &&
 399                     found_key.type != BTRFS_QGROUP_LIMIT_KEY)
 400                         goto next1;
 401
 402                 qgroup = find_qgroup_rb(fs_info, found_key.offset);
 403                 if ((qgroup && found_key.type == BTRFS_QGROUP_INFO_KEY) ||
 404                     (!qgroup && found_key.type == BTRFS_QGROUP_LIMIT_KEY)) {
 405                         btrfs_err(fs_info, "inconsistent qgroup config");
 406                         flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
 407                 }
 408                 if (!qgroup) {
 409                         qgroup = add_qgroup_rb(fs_info, found_key.offset);
 410                         if (IS_ERR(qgroup)) {
 411                                 ret = PTR_ERR(qgroup);
 412                                 goto out;
 413                         }
 414                 }
 415                 switch (found_key.type) {
 416                 case BTRFS_QGROUP_INFO_KEY: {
 417                         struct btrfs_qgroup_info_item *ptr;
 418
 419                         ptr = btrfs_item_ptr(l, slot,
 420                                              struct btrfs_qgroup_info_item);
 421                         qgroup->rfer = btrfs_qgroup_info_rfer(l, ptr);
 422                         qgroup->rfer_cmpr = btrfs_qgroup_info_rfer_cmpr(l, ptr);
 423                         qgroup->excl = btrfs_qgroup_info_excl(l, ptr);
 424                         qgroup->excl_cmpr = btrfs_qgroup_info_excl_cmpr(l, ptr);
 425                         /* generation currently unused */
 426                         break;
 427                 }
 428                 case BTRFS_QGROUP_LIMIT_KEY: {
 429                         struct btrfs_qgroup_limit_item *ptr;
 430
 431                         ptr = btrfs_item_ptr(l, slot,
 432                                              struct btrfs_qgroup_limit_item);
 433                         qgroup->lim_flags = btrfs_qgroup_limit_flags(l, ptr);
 434                         qgroup->max_rfer = btrfs_qgroup_limit_max_rfer(l, ptr);
 435                         qgroup->max_excl = btrfs_qgroup_limit_max_excl(l, ptr);
 436                         qgroup->rsv_rfer = btrfs_qgroup_limit_rsv_rfer(l, ptr);
 437                         qgroup->rsv_excl = btrfs_qgroup_limit_rsv_excl(l, ptr);
 438                         break;
 439                 }
 440                 }
 441 next1:
 442                 ret = btrfs_next_item(quota_root, path);
 443                 if (ret < 0)
 444                         goto out;
 445                 if (ret)
 446                         break;
 447         }
 448         btrfs_release_path(path);
 449
 450         /*
 451          * pass 2: read all qgroup relations
 452          */
 453         key.objectid = 0;
 454         key.type = BTRFS_QGROUP_RELATION_KEY;
 455         key.offset = 0;
 456         ret = btrfs_search_slot_for_read(quota_root, &key, path, 1, 0);
 457         if (ret)
 458                 goto out;
 459         while (1) {
 460                 slot = path->slots[0];
 461                 l = path->nodes[0];
 462                 btrfs_item_key_to_cpu(l, &found_key, slot);
 463
 464                 if (found_key.type != BTRFS_QGROUP_RELATION_KEY)
 465                         goto next2;
 466
 467                 if (found_key.objectid > found_key.offset) {
 468                         /* parent <- member, not needed to build config */
 469                         /* FIXME should we omit the key completely? */
 470                         goto next2;
 471                 }
 472
 473                 ret = add_relation_rb(fs_info, found_key.objectid,
 474                                       found_key.offset);
 475                 if (ret == -ENOENT) {
 476                         btrfs_warn(fs_info,
 477                                 "orphan qgroup relation 0x%llx->0x%llx",
 478                                 found_key.objectid, found_key.offset);
 479                         ret = 0;        /* ignore the error */
 480                 }
 481                 if (ret)
 482                         goto out;
 483 next2:
 484                 ret = btrfs_next_item(quota_root, path);
 485                 if (ret < 0)
 486                         goto out;
 487                 if (ret)
 488                         break;
 489         }
 490 out:
 491         fs_info->qgroup_flags |= flags;
 492         if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON))
 493                 clear_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
 494         else if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN &&
 495                  ret >= 0)
 496                 ret = qgroup_rescan_init(fs_info, rescan_progress, 0);
 497         btrfs_free_path(path);
 498
 499         if (ret < 0) {
 500                 ulist_free(fs_info->qgroup_ulist);
 501                 fs_info->qgroup_ulist = NULL;
 502                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
 503         }
 504
 505         return ret < 0 ? ret : 0;
 506 }
 507
 508 /*
 509  * This is called from close_ctree() or open_ctree() or btrfs_quota_disable(),
 510  * first two are in single-threaded paths.And for the third one, we have set
 511  * quota_root to be null with qgroup_lock held before, so it is safe to clean
 512  * up the in-memory structures without qgroup_lock held.
 513  */
 514 void btrfs_free_qgroup_config(struct btrfs_fs_info *fs_info)
 515 {
 516         struct rb_node *n;
 517         struct btrfs_qgroup *qgroup;
 518
 519         while ((n = rb_first(&fs_info->qgroup_tree))) {
 520                 qgroup = rb_entry(n, struct btrfs_qgroup, node);
 521                 rb_erase(n, &fs_info->qgroup_tree);
 522                 __del_qgroup_rb(qgroup);
 523         }
 524         /*
 525          * We call btrfs_free_qgroup_config() when unmounting
 526          * filesystem and disabling quota, so we set qgroup_ulist
 527          * to be null here to avoid double free.
 528          */
 529         ulist_free(fs_info->qgroup_ulist);
 530         fs_info->qgroup_ulist = NULL;
 531 }
 532
 533 static int add_qgroup_relation_item(struct btrfs_trans_handle *trans, u64 src,
 534                                     u64 dst)
 535 {
 536         int ret;
 537         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 538         struct btrfs_path *path;
 539         struct btrfs_key key;
 540
 541         path = btrfs_alloc_path();
 542         if (!path)
 543                 return -ENOMEM;
 544
 545         key.objectid = src;
 546         key.type = BTRFS_QGROUP_RELATION_KEY;
 547         key.offset = dst;
 548
 549         ret = btrfs_insert_empty_item(trans, quota_root, path, &key, 0);
 550
 551         btrfs_mark_buffer_dirty(path->nodes[0]);
 552
 553         btrfs_free_path(path);
 554         return ret;
 555 }
 556
 557 static int del_qgroup_relation_item(struct btrfs_trans_handle *trans, u64 src,
 558                                     u64 dst)
 559 {
 560         int ret;
 561         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 562         struct btrfs_path *path;
 563         struct btrfs_key key;
 564
 565         path = btrfs_alloc_path();
 566         if (!path)
 567                 return -ENOMEM;
 568
 569         key.objectid = src;
 570         key.type = BTRFS_QGROUP_RELATION_KEY;
 571         key.offset = dst;
 572
 573         ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
 574         if (ret < 0)
 575                 goto out;
 576
 577         if (ret > 0) {
 578                 ret = -ENOENT;
 579                 goto out;
 580         }
 581
 582         ret = btrfs_del_item(trans, quota_root, path);
 583 out:
 584         btrfs_free_path(path);
 585         return ret;
 586 }
 587
 588 static int add_qgroup_item(struct btrfs_trans_handle *trans,
 589                            struct btrfs_root *quota_root, u64 qgroupid)
 590 {
 591         int ret;
 592         struct btrfs_path *path;
 593         struct btrfs_qgroup_info_item *qgroup_info;
 594         struct btrfs_qgroup_limit_item *qgroup_limit;
 595         struct extent_buffer *leaf;
 596         struct btrfs_key key;
 597
 598         if (btrfs_is_testing(quota_root->fs_info))
 599                 return 0;
 600
 601         path = btrfs_alloc_path();
 602         if (!path)
 603                 return -ENOMEM;
 604
 605         key.objectid = 0;
 606         key.type = BTRFS_QGROUP_INFO_KEY;
 607         key.offset = qgroupid;
 608
 609         /*
 610          * Avoid a transaction abort by catching -EEXIST here. In that
 611          * case, we proceed by re-initializing the existing structure
 612          * on disk.
 613          */
 614
 615         ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
 616                                       sizeof(*qgroup_info));
 617         if (ret && ret != -EEXIST)
 618                 goto out;
 619
 620         leaf = path->nodes[0];
 621         qgroup_info = btrfs_item_ptr(leaf, path->slots[0],
 622                                  struct btrfs_qgroup_info_item);
 623         btrfs_set_qgroup_info_generation(leaf, qgroup_info, trans->transid);
 624         btrfs_set_qgroup_info_rfer(leaf, qgroup_info, 0);
 625         btrfs_set_qgroup_info_rfer_cmpr(leaf, qgroup_info, 0);
 626         btrfs_set_qgroup_info_excl(leaf, qgroup_info, 0);
 627         btrfs_set_qgroup_info_excl_cmpr(leaf, qgroup_info, 0);
 628
 629         btrfs_mark_buffer_dirty(leaf);
 630
 631         btrfs_release_path(path);
 632
 633         key.type = BTRFS_QGROUP_LIMIT_KEY;
 634         ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
 635                                       sizeof(*qgroup_limit));
 636         if (ret && ret != -EEXIST)
 637                 goto out;
 638
 639         leaf = path->nodes[0];
 640         qgroup_limit = btrfs_item_ptr(leaf, path->slots[0],
 641                                   struct btrfs_qgroup_limit_item);
 642         btrfs_set_qgroup_limit_flags(leaf, qgroup_limit, 0);
 643         btrfs_set_qgroup_limit_max_rfer(leaf, qgroup_limit, 0);
 644         btrfs_set_qgroup_limit_max_excl(leaf, qgroup_limit, 0);
 645         btrfs_set_qgroup_limit_rsv_rfer(leaf, qgroup_limit, 0);
 646         btrfs_set_qgroup_limit_rsv_excl(leaf, qgroup_limit, 0);
 647
 648         btrfs_mark_buffer_dirty(leaf);
 649
 650         ret = 0;
 651 out:
 652         btrfs_free_path(path);
 653         return ret;
 654 }
 655
 656 static int del_qgroup_item(struct btrfs_trans_handle *trans, u64 qgroupid)
 657 {
 658         int ret;
 659         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 660         struct btrfs_path *path;
 661         struct btrfs_key key;
 662
 663         path = btrfs_alloc_path();
 664         if (!path)
 665                 return -ENOMEM;
 666
 667         key.objectid = 0;
 668         key.type = BTRFS_QGROUP_INFO_KEY;
 669         key.offset = qgroupid;
 670         ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
 671         if (ret < 0)
 672                 goto out;
 673
 674         if (ret > 0) {
 675                 ret = -ENOENT;
 676                 goto out;
 677         }
 678
 679         ret = btrfs_del_item(trans, quota_root, path);
 680         if (ret)
 681                 goto out;
 682
 683         btrfs_release_path(path);
 684
 685         key.type = BTRFS_QGROUP_LIMIT_KEY;
 686         ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
 687         if (ret < 0)
 688                 goto out;
 689
 690         if (ret > 0) {
 691                 ret = -ENOENT;
 692                 goto out;
 693         }
 694
 695         ret = btrfs_del_item(trans, quota_root, path);
 696
 697 out:
 698         btrfs_free_path(path);
 699         return ret;
 700 }
 701
 702 static int update_qgroup_limit_item(struct btrfs_trans_handle *trans,
 703                                     struct btrfs_qgroup *qgroup)
 704 {
 705         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 706         struct btrfs_path *path;
 707         struct btrfs_key key;
 708         struct extent_buffer *l;
 709         struct btrfs_qgroup_limit_item *qgroup_limit;
 710         int ret;
 711         int slot;
 712
 713         key.objectid = 0;
 714         key.type = BTRFS_QGROUP_LIMIT_KEY;
 715         key.offset = qgroup->qgroupid;
 716
 717         path = btrfs_alloc_path();
 718         if (!path)
 719                 return -ENOMEM;
 720
 721         ret = btrfs_search_slot(trans, quota_root, &key, path, 0, 1);
 722         if (ret > 0)
 723                 ret = -ENOENT;
 724
 725         if (ret)
 726                 goto out;
 727
 728         l = path->nodes[0];
 729         slot = path->slots[0];
 730         qgroup_limit = btrfs_item_ptr(l, slot, struct btrfs_qgroup_limit_item);
 731         btrfs_set_qgroup_limit_flags(l, qgroup_limit, qgroup->lim_flags);
 732         btrfs_set_qgroup_limit_max_rfer(l, qgroup_limit, qgroup->max_rfer);
 733         btrfs_set_qgroup_limit_max_excl(l, qgroup_limit, qgroup->max_excl);
 734         btrfs_set_qgroup_limit_rsv_rfer(l, qgroup_limit, qgroup->rsv_rfer);
 735         btrfs_set_qgroup_limit_rsv_excl(l, qgroup_limit, qgroup->rsv_excl);
 736
 737         btrfs_mark_buffer_dirty(l);
 738
 739 out:
 740         btrfs_free_path(path);
 741         return ret;
 742 }
 743
 744 static int update_qgroup_info_item(struct btrfs_trans_handle *trans,
 745                                    struct btrfs_qgroup *qgroup)
 746 {
 747         struct btrfs_fs_info *fs_info = trans->fs_info;
 748         struct btrfs_root *quota_root = fs_info->quota_root;
 749         struct btrfs_path *path;
 750         struct btrfs_key key;
 751         struct extent_buffer *l;
 752         struct btrfs_qgroup_info_item *qgroup_info;
 753         int ret;
 754         int slot;
 755
 756         if (btrfs_is_testing(fs_info))
 757                 return 0;
 758
 759         key.objectid = 0;
 760         key.type = BTRFS_QGROUP_INFO_KEY;
 761         key.offset = qgroup->qgroupid;
 762
 763         path = btrfs_alloc_path();
 764         if (!path)
 765                 return -ENOMEM;
 766
 767         ret = btrfs_search_slot(trans, quota_root, &key, path, 0, 1);
 768         if (ret > 0)
 769                 ret = -ENOENT;
 770
 771         if (ret)
 772                 goto out;
 773
 774         l = path->nodes[0];
 775         slot = path->slots[0];
 776         qgroup_info = btrfs_item_ptr(l, slot, struct btrfs_qgroup_info_item);
 777         btrfs_set_qgroup_info_generation(l, qgroup_info, trans->transid);
 778         btrfs_set_qgroup_info_rfer(l, qgroup_info, qgroup->rfer);
 779         btrfs_set_qgroup_info_rfer_cmpr(l, qgroup_info, qgroup->rfer_cmpr);
 780         btrfs_set_qgroup_info_excl(l, qgroup_info, qgroup->excl);
 781         btrfs_set_qgroup_info_excl_cmpr(l, qgroup_info, qgroup->excl_cmpr);
 782
 783         btrfs_mark_buffer_dirty(l);
 784
 785 out:
 786         btrfs_free_path(path);
 787         return ret;
 788 }
 789
 790 static int update_qgroup_status_item(struct btrfs_trans_handle *trans)
 791 {
 792         struct btrfs_fs_info *fs_info = trans->fs_info;
 793         struct btrfs_root *quota_root = fs_info->quota_root;
 794         struct btrfs_path *path;
 795         struct btrfs_key key;
 796         struct extent_buffer *l;
 797         struct btrfs_qgroup_status_item *ptr;
 798         int ret;
 799         int slot;
 800
 801         key.objectid = 0;
 802         key.type = BTRFS_QGROUP_STATUS_KEY;
 803         key.offset = 0;
 804
 805         path = btrfs_alloc_path();
 806         if (!path)
 807                 return -ENOMEM;
 808
 809         ret = btrfs_search_slot(trans, quota_root, &key, path, 0, 1);
 810         if (ret > 0)
 811                 ret = -ENOENT;
 812
 813         if (ret)
 814                 goto out;
 815
 816         l = path->nodes[0];
 817         slot = path->slots[0];
 818         ptr = btrfs_item_ptr(l, slot, struct btrfs_qgroup_status_item);
 819         btrfs_set_qgroup_status_flags(l, ptr, fs_info->qgroup_flags);
 820         btrfs_set_qgroup_status_generation(l, ptr, trans->transid);
 821         btrfs_set_qgroup_status_rescan(l, ptr,
 822                                 fs_info->qgroup_rescan_progress.objectid);
 823
 824         btrfs_mark_buffer_dirty(l);
 825
 826 out:
 827         btrfs_free_path(path);
 828         return ret;
 829 }
 830
 831 /*
 832  * called with qgroup_lock held
 833  */
 834 static int btrfs_clean_quota_tree(struct btrfs_trans_handle *trans,
 835                                   struct btrfs_root *root)
 836 {
 837         struct btrfs_path *path;
 838         struct btrfs_key key;
 839         struct extent_buffer *leaf = NULL;
 840         int ret;
 841         int nr = 0;
 842
 843         path = btrfs_alloc_path();
 844         if (!path)
 845                 return -ENOMEM;
 846
 847         path->leave_spinning = 1;
 848
 849         key.objectid = 0;
 850         key.offset = 0;
 851         key.type = 0;
 852
 853         while (1) {
 854                 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 855                 if (ret < 0)
 856                         goto out;
 857                 leaf = path->nodes[0];
 858                 nr = btrfs_header_nritems(leaf);
 859                 if (!nr)
 860                         break;
 861                 /*
 862                  * delete the leaf one by one
 863                  * since the whole tree is going
 864                  * to be deleted.
 865                  */
 866                 path->slots[0] = 0;
 867                 ret = btrfs_del_items(trans, root, path, 0, nr);
 868                 if (ret)
 869                         goto out;
 870
 871                 btrfs_release_path(path);
 872         }
 873         ret = 0;
 874 out:
 875         btrfs_free_path(path);
 876         return ret;
 877 }
 878
 879 int btrfs_quota_enable(struct btrfs_fs_info *fs_info)
 880 {
 881         struct btrfs_root *quota_root;
 882         struct btrfs_root *tree_root = fs_info->tree_root;
 883         struct btrfs_path *path = NULL;
 884         struct btrfs_qgroup_status_item *ptr;
 885         struct extent_buffer *leaf;
 886         struct btrfs_key key;
 887         struct btrfs_key found_key;
 888         struct btrfs_qgroup *qgroup = NULL;
 889         struct btrfs_trans_handle *trans = NULL;
 890         int ret = 0;
 891         int slot;
 892
 893         mutex_lock(&fs_info->qgroup_ioctl_lock);
 894         if (fs_info->quota_root)
 895                 goto out;
 896
 897         fs_info->qgroup_ulist = ulist_alloc(GFP_KERNEL);
 898         if (!fs_info->qgroup_ulist) {
 899                 ret = -ENOMEM;
 900                 goto out;
 901         }
 902
 903         /*
 904          * 1 for quota root item
 905          * 1 for BTRFS_QGROUP_STATUS item
 906          *
 907          * Yet we also need 2*n items for a QGROUP_INFO/QGROUP_LIMIT items
 908          * per subvolume. However those are not currently reserved since it
 909          * would be a lot of overkill.
 910          */
 911         trans = btrfs_start_transaction(tree_root, 2);
 912         if (IS_ERR(trans)) {
 913                 ret = PTR_ERR(trans);
 914                 trans = NULL;
 915                 goto out;
 916         }
 917
 918         /*
 919          * initially create the quota tree
 920          */
 921         quota_root = btrfs_create_tree(trans, BTRFS_QUOTA_TREE_OBJECTID);
 922         if (IS_ERR(quota_root)) {
 923                 ret =  PTR_ERR(quota_root);
 924                 btrfs_abort_transaction(trans, ret);
 925                 goto out;
 926         }
 927
 928         path = btrfs_alloc_path();
 929         if (!path) {
 930                 ret = -ENOMEM;
 931                 btrfs_abort_transaction(trans, ret);
 932                 goto out_free_root;
 933         }
 934
 935         key.objectid = 0;
 936         key.type = BTRFS_QGROUP_STATUS_KEY;
 937         key.offset = 0;
 938
 939         ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
 940                                       sizeof(*ptr));
 941         if (ret) {
 942                 btrfs_abort_transaction(trans, ret);
 943                 goto out_free_path;
 944         }
 945
 946         leaf = path->nodes[0];
 947         ptr = btrfs_item_ptr(leaf, path->slots[0],
 948                                  struct btrfs_qgroup_status_item);
 949         btrfs_set_qgroup_status_generation(leaf, ptr, trans->transid);
 950         btrfs_set_qgroup_status_version(leaf, ptr, BTRFS_QGROUP_STATUS_VERSION);
 951         fs_info->qgroup_flags = BTRFS_QGROUP_STATUS_FLAG_ON |
 952                                 BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
 953         btrfs_set_qgroup_status_flags(leaf, ptr, fs_info->qgroup_flags);
 954         btrfs_set_qgroup_status_rescan(leaf, ptr, 0);
 955
 956         btrfs_mark_buffer_dirty(leaf);
 957
 958         key.objectid = 0;
 959         key.type = BTRFS_ROOT_REF_KEY;
 960         key.offset = 0;
 961
 962         btrfs_release_path(path);
 963         ret = btrfs_search_slot_for_read(tree_root, &key, path, 1, 0);
 964         if (ret > 0)
 965                 goto out_add_root;
 966         if (ret < 0) {
 967                 btrfs_abort_transaction(trans, ret);
 968                 goto out_free_path;
 969         }
 970
 971         while (1) {
 972                 slot = path->slots[0];
 973                 leaf = path->nodes[0];
 974                 btrfs_item_key_to_cpu(leaf, &found_key, slot);
 975
 976                 if (found_key.type == BTRFS_ROOT_REF_KEY) {
 977                         ret = add_qgroup_item(trans, quota_root,
 978                                               found_key.offset);
 979                         if (ret) {
 980                                 btrfs_abort_transaction(trans, ret);
 981                                 goto out_free_path;
 982                         }
 983
 984                         qgroup = add_qgroup_rb(fs_info, found_key.offset);
 985                         if (IS_ERR(qgroup)) {
 986                                 ret = PTR_ERR(qgroup);
 987                                 btrfs_abort_transaction(trans, ret);
 988                                 goto out_free_path;
 989                         }
 990                 }
 991                 ret = btrfs_next_item(tree_root, path);
 992                 if (ret < 0) {
 993                         btrfs_abort_transaction(trans, ret);
 994                         goto out_free_path;
 995                 }
 996                 if (ret)
 997                         break;
 998         }
 999
1000 out_add_root:
1001         btrfs_release_path(path);
1002         ret = add_qgroup_item(trans, quota_root, BTRFS_FS_TREE_OBJECTID);
1003         if (ret) {
1004                 btrfs_abort_transaction(trans, ret);
1005                 goto out_free_path;
1006         }
1007
1008         qgroup = add_qgroup_rb(fs_info, BTRFS_FS_TREE_OBJECTID);
1009         if (IS_ERR(qgroup)) {
1010                 ret = PTR_ERR(qgroup);
1011                 btrfs_abort_transaction(trans, ret);
1012                 goto out_free_path;
1013         }
1014
1015         ret = btrfs_commit_transaction(trans);
1016         trans = NULL;
1017         if (ret)
1018                 goto out_free_path;
1019
1020         /*
1021          * Set quota enabled flag after committing the transaction, to avoid
1022          * deadlocks on fs_info->qgroup_ioctl_lock with concurrent snapshot
1023          * creation.
1024          */
1025         spin_lock(&fs_info->qgroup_lock);
1026         fs_info->quota_root = quota_root;
1027         set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
1028         spin_unlock(&fs_info->qgroup_lock);
1029
1030         ret = qgroup_rescan_init(fs_info, 0, 1);
1031         if (!ret) {
1032                 qgroup_rescan_zero_tracking(fs_info);
1033                 btrfs_queue_work(fs_info->qgroup_rescan_workers,
1034                                  &fs_info->qgroup_rescan_work);
1035         }
1036
1037 out_free_path:
1038         btrfs_free_path(path);
1039 out_free_root:
1040         if (ret) {
1041                 free_extent_buffer(quota_root->node);
1042                 free_extent_buffer(quota_root->commit_root);
1043                 kfree(quota_root);
1044         }
1045 out:
1046         if (ret) {
1047                 ulist_free(fs_info->qgroup_ulist);
1048                 fs_info->qgroup_ulist = NULL;
1049                 if (trans)
1050                         btrfs_end_transaction(trans);
1051         }
1052         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1053         return ret;
1054 }
1055
1056 int btrfs_quota_disable(struct btrfs_fs_info *fs_info)
1057 {
1058         struct btrfs_root *quota_root;
1059         struct btrfs_trans_handle *trans = NULL;
1060         int ret = 0;
1061
1062         mutex_lock(&fs_info->qgroup_ioctl_lock);
1063         if (!fs_info->quota_root)
1064                 goto out;
1065
1066         /*
1067          * 1 For the root item
1068          *
1069          * We should also reserve enough items for the quota tree deletion in
1070          * btrfs_clean_quota_tree but this is not done.
1071          */
1072         trans = btrfs_start_transaction(fs_info->tree_root, 1);
1073         if (IS_ERR(trans)) {
1074                 ret = PTR_ERR(trans);
1075                 goto out;
1076         }
1077
1078         clear_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
1079         btrfs_qgroup_wait_for_completion(fs_info, false);
1080         spin_lock(&fs_info->qgroup_lock);
1081         quota_root = fs_info->quota_root;
1082         fs_info->quota_root = NULL;
1083         fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_ON;
1084         spin_unlock(&fs_info->qgroup_lock);
1085
1086         btrfs_free_qgroup_config(fs_info);
1087
1088         ret = btrfs_clean_quota_tree(trans, quota_root);
1089         if (ret) {
1090                 btrfs_abort_transaction(trans, ret);
1091                 goto end_trans;
1092         }
1093
1094         ret = btrfs_del_root(trans, &quota_root->root_key);
1095         if (ret) {
1096                 btrfs_abort_transaction(trans, ret);
1097                 goto end_trans;
1098         }
1099
1100         list_del(&quota_root->dirty_list);
1101
1102         btrfs_tree_lock(quota_root->node);
1103         btrfs_clean_tree_block(quota_root->node);
1104         btrfs_tree_unlock(quota_root->node);
1105         btrfs_free_tree_block(trans, quota_root, quota_root->node, 0, 1);
1106
1107         free_extent_buffer(quota_root->node);
1108         free_extent_buffer(quota_root->commit_root);
1109         kfree(quota_root);
1110
1111 end_trans:
1112         ret = btrfs_end_transaction(trans);
1113 out:
1114         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1115         return ret;
1116 }
1117
1118 static void qgroup_dirty(struct btrfs_fs_info *fs_info,
1119                          struct btrfs_qgroup *qgroup)
1120 {
1121         if (list_empty(&qgroup->dirty))
1122                 list_add(&qgroup->dirty, &fs_info->dirty_qgroups);
1123 }
1124
1125 /*
1126  * The easy accounting, we're updating qgroup relationship whose child qgroup
1127  * only has exclusive extents.
1128  *
1129  * In this case, all exclusive extents will also be exclusive for parent, so
1130  * excl/rfer just get added/removed.
1131  *
1132  * So is qgroup reservation space, which should also be added/removed to
1133  * parent.
1134  * Or when child tries to release reservation space, parent will underflow its
1135  * reservation (for relationship adding case).
1136  *
1137  * Caller should hold fs_info->qgroup_lock.
1138  */
1139 static int __qgroup_excl_accounting(struct btrfs_fs_info *fs_info,
1140                                     struct ulist *tmp, u64 ref_root,
1141                                     struct btrfs_qgroup *src, int sign)
1142 {
1143         struct btrfs_qgroup *qgroup;
1144         struct btrfs_qgroup_list *glist;
1145         struct ulist_node *unode;
1146         struct ulist_iterator uiter;
1147         u64 num_bytes = src->excl;
1148         int ret = 0;
1149
1150         qgroup = find_qgroup_rb(fs_info, ref_root);
1151         if (!qgroup)
1152                 goto out;
1153
1154         qgroup->rfer += sign * num_bytes;
1155         qgroup->rfer_cmpr += sign * num_bytes;
1156
1157         WARN_ON(sign < 0 && qgroup->excl < num_bytes);
1158         qgroup->excl += sign * num_bytes;
1159         qgroup->excl_cmpr += sign * num_bytes;
1160
1161         if (sign > 0)
1162                 qgroup_rsv_add_by_qgroup(fs_info, qgroup, src);
1163         else
1164                 qgroup_rsv_release_by_qgroup(fs_info, qgroup, src);
1165
1166         qgroup_dirty(fs_info, qgroup);
1167
1168         /* Get all of the parent groups that contain this qgroup */
1169         list_for_each_entry(glist, &qgroup->groups, next_group) {
1170                 ret = ulist_add(tmp, glist->group->qgroupid,
1171                                 qgroup_to_aux(glist->group), GFP_ATOMIC);
1172                 if (ret < 0)
1173                         goto out;
1174         }
1175
1176         /* Iterate all of the parents and adjust their reference counts */
1177         ULIST_ITER_INIT(&uiter);
1178         while ((unode = ulist_next(tmp, &uiter))) {
1179                 qgroup = unode_aux_to_qgroup(unode);
1180                 qgroup->rfer += sign * num_bytes;
1181                 qgroup->rfer_cmpr += sign * num_bytes;
1182                 WARN_ON(sign < 0 && qgroup->excl < num_bytes);
1183                 qgroup->excl += sign * num_bytes;
1184                 if (sign > 0)
1185                         qgroup_rsv_add_by_qgroup(fs_info, qgroup, src);
1186                 else
1187                         qgroup_rsv_release_by_qgroup(fs_info, qgroup, src);
1188                 qgroup->excl_cmpr += sign * num_bytes;
1189                 qgroup_dirty(fs_info, qgroup);
1190
1191                 /* Add any parents of the parents */
1192                 list_for_each_entry(glist, &qgroup->groups, next_group) {
1193                         ret = ulist_add(tmp, glist->group->qgroupid,
1194                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
1195                         if (ret < 0)
1196                                 goto out;
1197                 }
1198         }
1199         ret = 0;
1200 out:
1201         return ret;
1202 }
1203
1204
1205 /*
1206  * Quick path for updating qgroup with only excl refs.
1207  *
1208  * In that case, just update all parent will be enough.
1209  * Or we needs to do a full rescan.
1210  * Caller should also hold fs_info->qgroup_lock.
1211  *
1212  * Return 0 for quick update, return >0 for need to full rescan
1213  * and mark INCONSISTENT flag.
1214  * Return < 0 for other error.
1215  */
1216 static int quick_update_accounting(struct btrfs_fs_info *fs_info,
1217                                    struct ulist *tmp, u64 src, u64 dst,
1218                                    int sign)
1219 {
1220         struct btrfs_qgroup *qgroup;
1221         int ret = 1;
1222         int err = 0;
1223
1224         qgroup = find_qgroup_rb(fs_info, src);
1225         if (!qgroup)
1226                 goto out;
1227         if (qgroup->excl == qgroup->rfer) {
1228                 ret = 0;
1229                 err = __qgroup_excl_accounting(fs_info, tmp, dst,
1230                                                qgroup, sign);
1231                 if (err < 0) {
1232                         ret = err;
1233                         goto out;
1234                 }
1235         }
1236 out:
1237         if (ret)
1238                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
1239         return ret;
1240 }
1241
1242 int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
1243                               u64 dst)
1244 {
1245         struct btrfs_fs_info *fs_info = trans->fs_info;
1246         struct btrfs_root *quota_root;
1247         struct btrfs_qgroup *parent;
1248         struct btrfs_qgroup *member;
1249         struct btrfs_qgroup_list *list;
1250         struct ulist *tmp;
1251         int ret = 0;
1252
1253         /* Check the level of src and dst first */
1254         if (btrfs_qgroup_level(src) >= btrfs_qgroup_level(dst))
1255                 return -EINVAL;
1256
1257         tmp = ulist_alloc(GFP_KERNEL);
1258         if (!tmp)
1259                 return -ENOMEM;
1260
1261         mutex_lock(&fs_info->qgroup_ioctl_lock);
1262         quota_root = fs_info->quota_root;
1263         if (!quota_root) {
1264                 ret = -EINVAL;
1265                 goto out;
1266         }
1267         member = find_qgroup_rb(fs_info, src);
1268         parent = find_qgroup_rb(fs_info, dst);
1269         if (!member || !parent) {
1270                 ret = -EINVAL;
1271                 goto out;
1272         }
1273
1274         /* check if such qgroup relation exist firstly */
1275         list_for_each_entry(list, &member->groups, next_group) {
1276                 if (list->group == parent) {
1277                         ret = -EEXIST;
1278                         goto out;
1279                 }
1280         }
1281
1282         ret = add_qgroup_relation_item(trans, src, dst);
1283         if (ret)
1284                 goto out;
1285
1286         ret = add_qgroup_relation_item(trans, dst, src);
1287         if (ret) {
1288                 del_qgroup_relation_item(trans, src, dst);
1289                 goto out;
1290         }
1291
1292         spin_lock(&fs_info->qgroup_lock);
1293         ret = add_relation_rb(fs_info, src, dst);
1294         if (ret < 0) {
1295                 spin_unlock(&fs_info->qgroup_lock);
1296                 goto out;
1297         }
1298         ret = quick_update_accounting(fs_info, tmp, src, dst, 1);
1299         spin_unlock(&fs_info->qgroup_lock);
1300 out:
1301         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1302         ulist_free(tmp);
1303         return ret;
1304 }
1305
1306 static int __del_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
1307                                  u64 dst)
1308 {
1309         struct btrfs_fs_info *fs_info = trans->fs_info;
1310         struct btrfs_root *quota_root;
1311         struct btrfs_qgroup *parent;
1312         struct btrfs_qgroup *member;
1313         struct btrfs_qgroup_list *list;
1314         struct ulist *tmp;
1315         int ret = 0;
1316         int err;
1317
1318         tmp = ulist_alloc(GFP_KERNEL);
1319         if (!tmp)
1320                 return -ENOMEM;
1321
1322         quota_root = fs_info->quota_root;
1323         if (!quota_root) {
1324                 ret = -EINVAL;
1325                 goto out;
1326         }
1327
1328         member = find_qgroup_rb(fs_info, src);
1329         parent = find_qgroup_rb(fs_info, dst);
1330         if (!member || !parent) {
1331                 ret = -EINVAL;
1332                 goto out;
1333         }
1334
1335         /* check if such qgroup relation exist firstly */
1336         list_for_each_entry(list, &member->groups, next_group) {
1337                 if (list->group == parent)
1338                         goto exist;
1339         }
1340         ret = -ENOENT;
1341         goto out;
1342 exist:
1343         ret = del_qgroup_relation_item(trans, src, dst);
1344         err = del_qgroup_relation_item(trans, dst, src);
1345         if (err && !ret)
1346                 ret = err;
1347
1348         spin_lock(&fs_info->qgroup_lock);
1349         del_relation_rb(fs_info, src, dst);
1350         ret = quick_update_accounting(fs_info, tmp, src, dst, -1);
1351         spin_unlock(&fs_info->qgroup_lock);
1352 out:
1353         ulist_free(tmp);
1354         return ret;
1355 }
1356
1357 int btrfs_del_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
1358                               u64 dst)
1359 {
1360         struct btrfs_fs_info *fs_info = trans->fs_info;
1361         int ret = 0;
1362
1363         mutex_lock(&fs_info->qgroup_ioctl_lock);
1364         ret = __del_qgroup_relation(trans, src, dst);
1365         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1366
1367         return ret;
1368 }
1369
1370 int btrfs_create_qgroup(struct btrfs_trans_handle *trans, u64 qgroupid)
1371 {
1372         struct btrfs_fs_info *fs_info = trans->fs_info;
1373         struct btrfs_root *quota_root;
1374         struct btrfs_qgroup *qgroup;
1375         int ret = 0;
1376
1377         mutex_lock(&fs_info->qgroup_ioctl_lock);
1378         quota_root = fs_info->quota_root;
1379         if (!quota_root) {
1380                 ret = -EINVAL;
1381                 goto out;
1382         }
1383         qgroup = find_qgroup_rb(fs_info, qgroupid);
1384         if (qgroup) {
1385                 ret = -EEXIST;
1386                 goto out;
1387         }
1388
1389         ret = add_qgroup_item(trans, quota_root, qgroupid);
1390         if (ret)
1391                 goto out;
1392
1393         spin_lock(&fs_info->qgroup_lock);
1394         qgroup = add_qgroup_rb(fs_info, qgroupid);
1395         spin_unlock(&fs_info->qgroup_lock);
1396
1397         if (IS_ERR(qgroup))
1398                 ret = PTR_ERR(qgroup);
1399 out:
1400         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1401         return ret;
1402 }
1403
1404 int btrfs_remove_qgroup(struct btrfs_trans_handle *trans, u64 qgroupid)
1405 {
1406         struct btrfs_fs_info *fs_info = trans->fs_info;
1407         struct btrfs_root *quota_root;
1408         struct btrfs_qgroup *qgroup;
1409         struct btrfs_qgroup_list *list;
1410         int ret = 0;
1411
1412         mutex_lock(&fs_info->qgroup_ioctl_lock);
1413         quota_root = fs_info->quota_root;
1414         if (!quota_root) {
1415                 ret = -EINVAL;
1416                 goto out;
1417         }
1418
1419         qgroup = find_qgroup_rb(fs_info, qgroupid);
1420         if (!qgroup) {
1421                 ret = -ENOENT;
1422                 goto out;
1423         }
1424
1425         /* Check if there are no children of this qgroup */
1426         if (!list_empty(&qgroup->members)) {
1427                 ret = -EBUSY;
1428                 goto out;
1429         }
1430
1431         ret = del_qgroup_item(trans, qgroupid);
1432         if (ret && ret != -ENOENT)
1433                 goto out;
1434
1435         while (!list_empty(&qgroup->groups)) {
1436                 list = list_first_entry(&qgroup->groups,
1437                                         struct btrfs_qgroup_list, next_group);
1438                 ret = __del_qgroup_relation(trans, qgroupid,
1439                                             list->group->qgroupid);
1440                 if (ret)
1441                         goto out;
1442         }
1443
1444         spin_lock(&fs_info->qgroup_lock);
1445         del_qgroup_rb(fs_info, qgroupid);
1446         spin_unlock(&fs_info->qgroup_lock);
1447 out:
1448         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1449         return ret;
1450 }
1451
1452 int btrfs_limit_qgroup(struct btrfs_trans_handle *trans, u64 qgroupid,
1453                        struct btrfs_qgroup_limit *limit)
1454 {
1455         struct btrfs_fs_info *fs_info = trans->fs_info;
1456         struct btrfs_root *quota_root;
1457         struct btrfs_qgroup *qgroup;
1458         int ret = 0;
1459         /* Sometimes we would want to clear the limit on this qgroup.
1460          * To meet this requirement, we treat the -1 as a special value
1461          * which tell kernel to clear the limit on this qgroup.
1462          */
1463         const u64 CLEAR_VALUE = -1;
1464
1465         mutex_lock(&fs_info->qgroup_ioctl_lock);
1466         quota_root = fs_info->quota_root;
1467         if (!quota_root) {
1468                 ret = -EINVAL;
1469                 goto out;
1470         }
1471
1472         qgroup = find_qgroup_rb(fs_info, qgroupid);
1473         if (!qgroup) {
1474                 ret = -ENOENT;
1475                 goto out;
1476         }
1477
1478         spin_lock(&fs_info->qgroup_lock);
1479         if (limit->flags & BTRFS_QGROUP_LIMIT_MAX_RFER) {
1480                 if (limit->max_rfer == CLEAR_VALUE) {
1481                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_MAX_RFER;
1482                         limit->flags &= ~BTRFS_QGROUP_LIMIT_MAX_RFER;
1483                         qgroup->max_rfer = 0;
1484                 } else {
1485                         qgroup->max_rfer = limit->max_rfer;
1486                 }
1487         }
1488         if (limit->flags & BTRFS_QGROUP_LIMIT_MAX_EXCL) {
1489                 if (limit->max_excl == CLEAR_VALUE) {
1490                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_MAX_EXCL;
1491                         limit->flags &= ~BTRFS_QGROUP_LIMIT_MAX_EXCL;
1492                         qgroup->max_excl = 0;
1493                 } else {
1494                         qgroup->max_excl = limit->max_excl;
1495                 }
1496         }
1497         if (limit->flags & BTRFS_QGROUP_LIMIT_RSV_RFER) {
1498                 if (limit->rsv_rfer == CLEAR_VALUE) {
1499                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_RSV_RFER;
1500                         limit->flags &= ~BTRFS_QGROUP_LIMIT_RSV_RFER;
1501                         qgroup->rsv_rfer = 0;
1502                 } else {
1503                         qgroup->rsv_rfer = limit->rsv_rfer;
1504                 }
1505         }
1506         if (limit->flags & BTRFS_QGROUP_LIMIT_RSV_EXCL) {
1507                 if (limit->rsv_excl == CLEAR_VALUE) {
1508                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_RSV_EXCL;
1509                         limit->flags &= ~BTRFS_QGROUP_LIMIT_RSV_EXCL;
1510                         qgroup->rsv_excl = 0;
1511                 } else {
1512                         qgroup->rsv_excl = limit->rsv_excl;
1513                 }
1514         }
1515         qgroup->lim_flags |= limit->flags;
1516
1517         spin_unlock(&fs_info->qgroup_lock);
1518
1519         ret = update_qgroup_limit_item(trans, qgroup);
1520         if (ret) {
1521                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
1522                 btrfs_info(fs_info, "unable to update quota limit for %llu",
1523                        qgroupid);
1524         }
1525
1526 out:
1527         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1528         return ret;
1529 }
1530
1531 int btrfs_qgroup_trace_extent_nolock(struct btrfs_fs_info *fs_info,
1532                                 struct btrfs_delayed_ref_root *delayed_refs,
1533                                 struct btrfs_qgroup_extent_record *record)
1534 {
1535         struct rb_node **p = &delayed_refs->dirty_extent_root.rb_node;
1536         struct rb_node *parent_node = NULL;
1537         struct btrfs_qgroup_extent_record *entry;
1538         u64 bytenr = record->bytenr;
1539
1540         lockdep_assert_held(&delayed_refs->lock);
1541         trace_btrfs_qgroup_trace_extent(fs_info, record);
1542
1543         while (*p) {
1544                 parent_node = *p;
1545                 entry = rb_entry(parent_node, struct btrfs_qgroup_extent_record,
1546                                  node);
1547                 if (bytenr < entry->bytenr) {
1548                         p = &(*p)->rb_left;
1549                 } else if (bytenr > entry->bytenr) {
1550                         p = &(*p)->rb_right;
1551                 } else {
1552                         if (record->data_rsv && !entry->data_rsv) {
1553                                 entry->data_rsv = record->data_rsv;
1554                                 entry->data_rsv_refroot =
1555                                         record->data_rsv_refroot;
1556                         }
1557                         return 1;
1558                 }
1559         }
1560
1561         rb_link_node(&record->node, parent_node, p);
1562         rb_insert_color(&record->node, &delayed_refs->dirty_extent_root);
1563         return 0;
1564 }
1565
1566 int btrfs_qgroup_trace_extent_post(struct btrfs_fs_info *fs_info,
1567                                    struct btrfs_qgroup_extent_record *qrecord)
1568 {
1569         struct ulist *old_root;
1570         u64 bytenr = qrecord->bytenr;
1571         int ret;
1572
1573         ret = btrfs_find_all_roots(NULL, fs_info, bytenr, 0, &old_root, false);
1574         if (ret < 0) {
1575                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
1576                 btrfs_warn(fs_info,
1577 "error accounting new delayed refs extent (err code: %d), quota inconsistent",
1578                         ret);
1579                 return 0;
1580         }
1581
1582         /*
1583          * Here we don't need to get the lock of
1584          * trans->transaction->delayed_refs, since inserted qrecord won't
1585          * be deleted, only qrecord->node may be modified (new qrecord insert)
1586          *
1587          * So modifying qrecord->old_roots is safe here
1588          */
1589         qrecord->old_roots = old_root;
1590         return 0;
1591 }
1592
1593 int btrfs_qgroup_trace_extent(struct btrfs_trans_handle *trans, u64 bytenr,
1594                               u64 num_bytes, gfp_t gfp_flag)
1595 {
1596         struct btrfs_fs_info *fs_info = trans->fs_info;
1597         struct btrfs_qgroup_extent_record *record;
1598         struct btrfs_delayed_ref_root *delayed_refs;
1599         int ret;
1600
1601         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags)
1602             || bytenr == 0 || num_bytes == 0)
1603                 return 0;
1604         record = kzalloc(sizeof(*record), gfp_flag);
1605         if (!record)
1606                 return -ENOMEM;
1607
1608         delayed_refs = &trans->transaction->delayed_refs;
1609         record->bytenr = bytenr;
1610         record->num_bytes = num_bytes;
1611         record->old_roots = NULL;
1612
1613         spin_lock(&delayed_refs->lock);
1614         ret = btrfs_qgroup_trace_extent_nolock(fs_info, delayed_refs, record);
1615         spin_unlock(&delayed_refs->lock);
1616         if (ret > 0) {
1617                 kfree(record);
1618                 return 0;
1619         }
1620         return btrfs_qgroup_trace_extent_post(fs_info, record);
1621 }
1622
1623 int btrfs_qgroup_trace_leaf_items(struct btrfs_trans_handle *trans,
1624                                   struct extent_buffer *eb)
1625 {
1626         struct btrfs_fs_info *fs_info = trans->fs_info;
1627         int nr = btrfs_header_nritems(eb);
1628         int i, extent_type, ret;
1629         struct btrfs_key key;
1630         struct btrfs_file_extent_item *fi;
1631         u64 bytenr, num_bytes;
1632
1633         /* We can be called directly from walk_up_proc() */
1634         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
1635                 return 0;
1636
1637         for (i = 0; i < nr; i++) {
1638                 btrfs_item_key_to_cpu(eb, &key, i);
1639
1640                 if (key.type != BTRFS_EXTENT_DATA_KEY)
1641                         continue;
1642
1643                 fi = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
1644                 /* filter out non qgroup-accountable extents  */
1645                 extent_type = btrfs_file_extent_type(eb, fi);
1646
1647                 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
1648                         continue;
1649
1650                 bytenr = btrfs_file_extent_disk_bytenr(eb, fi);
1651                 if (!bytenr)
1652                         continue;
1653
1654                 num_bytes = btrfs_file_extent_disk_num_bytes(eb, fi);
1655
1656                 ret = btrfs_qgroup_trace_extent(trans, bytenr, num_bytes,
1657                                                 GFP_NOFS);
1658                 if (ret)
1659                         return ret;
1660         }
1661         cond_resched();
1662         return 0;
1663 }
1664
1665 /*
1666  * Walk up the tree from the bottom, freeing leaves and any interior
1667  * nodes which have had all slots visited. If a node (leaf or
1668  * interior) is freed, the node above it will have it's slot
1669  * incremented. The root node will never be freed.
1670  *
1671  * At the end of this function, we should have a path which has all
1672  * slots incremented to the next position for a search. If we need to
1673  * read a new node it will be NULL and the node above it will have the
1674  * correct slot selected for a later read.
1675  *
1676  * If we increment the root nodes slot counter past the number of
1677  * elements, 1 is returned to signal completion of the search.
1678  */
1679 static int adjust_slots_upwards(struct btrfs_path *path, int root_level)
1680 {
1681         int level = 0;
1682         int nr, slot;
1683         struct extent_buffer *eb;
1684
1685         if (root_level == 0)
1686                 return 1;
1687
1688         while (level <= root_level) {
1689                 eb = path->nodes[level];
1690                 nr = btrfs_header_nritems(eb);
1691                 path->slots[level]++;
1692                 slot = path->slots[level];
1693                 if (slot >= nr || level == 0) {
1694                         /*
1695                          * Don't free the root -  we will detect this
1696                          * condition after our loop and return a
1697                          * positive value for caller to stop walking the tree.
1698                          */
1699                         if (level != root_level) {
1700                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
1701                                 path->locks[level] = 0;
1702
1703                                 free_extent_buffer(eb);
1704                                 path->nodes[level] = NULL;
1705                                 path->slots[level] = 0;
1706                         }
1707                 } else {
1708                         /*
1709                          * We have a valid slot to walk back down
1710                          * from. Stop here so caller can process these
1711                          * new nodes.
1712                          */
1713                         break;
1714                 }
1715
1716                 level++;
1717         }
1718
1719         eb = path->nodes[root_level];
1720         if (path->slots[root_level] >= btrfs_header_nritems(eb))
1721                 return 1;
1722
1723         return 0;
1724 }
1725
1726 /*
1727  * Helper function to trace a subtree tree block swap.
1728  *
1729  * The swap will happen in highest tree block, but there may be a lot of
1730  * tree blocks involved.
1731  *
1732  * For example:
1733  *  OO = Old tree blocks
1734  *  NN = New tree blocks allocated during balance
1735  *
1736  *           File tree (257)                  Reloc tree for 257
1737  * L2              OO                                NN
1738  *               /    \                            /    \
1739  * L1          OO      OO (a)                    OO      NN (a)
1740  *            / \     / \                       / \     / \
1741  * L0       OO   OO OO   OO                   OO   OO NN   NN
1742  *                  (b)  (c)                          (b)  (c)
1743  *
1744  * When calling qgroup_trace_extent_swap(), we will pass:
1745  * @src_eb = OO(a)
1746  * @dst_path = [ nodes[1] = NN(a), nodes[0] = NN(c) ]
1747  * @dst_level = 0
1748  * @root_level = 1
1749  *
1750  * In that case, qgroup_trace_extent_swap() will search from OO(a) to
1751  * reach OO(c), then mark both OO(c) and NN(c) as qgroup dirty.
1752  *
1753  * The main work of qgroup_trace_extent_swap() can be split into 3 parts:
1754  *
1755  * 1) Tree search from @src_eb
1756  *    It should acts as a simplified btrfs_search_slot().
1757  *    The key for search can be extracted from @dst_path->nodes[dst_level]
1758  *    (first key).
1759  *
1760  * 2) Mark the final tree blocks in @src_path and @dst_path qgroup dirty
1761  *    NOTE: In above case, OO(a) and NN(a) won't be marked qgroup dirty.
1762  *    They should be marked during previous (@dst_level = 1) iteration.
1763  *
1764  * 3) Mark file extents in leaves dirty
1765  *    We don't have good way to pick out new file extents only.
1766  *    So we still follow the old method by scanning all file extents in
1767  *    the leave.
1768  *
1769  * This function can free us from keeping two paths, thus later we only need
1770  * to care about how to iterate all new tree blocks in reloc tree.
1771  */
1772 static int qgroup_trace_extent_swap(struct btrfs_trans_handle* trans,
1773                                     struct extent_buffer *src_eb,
1774                                     struct btrfs_path *dst_path,
1775                                     int dst_level, int root_level,
1776                                     bool trace_leaf)
1777 {
1778         struct btrfs_key key;
1779         struct btrfs_path *src_path;
1780         struct btrfs_fs_info *fs_info = trans->fs_info;
1781         u32 nodesize = fs_info->nodesize;
1782         int cur_level = root_level;
1783         int ret;
1784
1785         BUG_ON(dst_level > root_level);
1786         /* Level mismatch */
1787         if (btrfs_header_level(src_eb) != root_level)
1788                 return -EINVAL;
1789
1790         src_path = btrfs_alloc_path();
1791         if (!src_path) {
1792                 ret = -ENOMEM;
1793                 goto out;
1794         }
1795
1796         if (dst_level)
1797                 btrfs_node_key_to_cpu(dst_path->nodes[dst_level], &key, 0);
1798         else
1799                 btrfs_item_key_to_cpu(dst_path->nodes[dst_level], &key, 0);
1800
1801         /* For src_path */
1802         extent_buffer_get(src_eb);
1803         src_path->nodes[root_level] = src_eb;
1804         src_path->slots[root_level] = dst_path->slots[root_level];
1805         src_path->locks[root_level] = 0;
1806
1807         /* A simplified version of btrfs_search_slot() */
1808         while (cur_level >= dst_level) {
1809                 struct btrfs_key src_key;
1810                 struct btrfs_key dst_key;
1811
1812                 if (src_path->nodes[cur_level] == NULL) {
1813                         struct btrfs_key first_key;
1814                         struct extent_buffer *eb;
1815                         int parent_slot;
1816                         u64 child_gen;
1817                         u64 child_bytenr;
1818
1819                         eb = src_path->nodes[cur_level + 1];
1820                         parent_slot = src_path->slots[cur_level + 1];
1821                         child_bytenr = btrfs_node_blockptr(eb, parent_slot);
1822                         child_gen = btrfs_node_ptr_generation(eb, parent_slot);
1823                         btrfs_node_key_to_cpu(eb, &first_key, parent_slot);
1824
1825                         eb = read_tree_block(fs_info, child_bytenr, child_gen,
1826                                              cur_level, &first_key);
1827                         if (IS_ERR(eb)) {
1828                                 ret = PTR_ERR(eb);
1829                                 goto out;
1830                         } else if (!extent_buffer_uptodate(eb)) {
1831                                 free_extent_buffer(eb);
1832                                 ret = -EIO;
1833                                 goto out;
1834                         }
1835
1836                         src_path->nodes[cur_level] = eb;
1837
1838                         btrfs_tree_read_lock(eb);
1839                         btrfs_set_lock_blocking_read(eb);
1840                         src_path->locks[cur_level] = BTRFS_READ_LOCK_BLOCKING;
1841                 }
1842
1843                 src_path->slots[cur_level] = dst_path->slots[cur_level];
1844                 if (cur_level) {
1845                         btrfs_node_key_to_cpu(dst_path->nodes[cur_level],
1846                                         &dst_key, dst_path->slots[cur_level]);
1847                         btrfs_node_key_to_cpu(src_path->nodes[cur_level],
1848                                         &src_key, src_path->slots[cur_level]);
1849                 } else {
1850                         btrfs_item_key_to_cpu(dst_path->nodes[cur_level],
1851                                         &dst_key, dst_path->slots[cur_level]);
1852                         btrfs_item_key_to_cpu(src_path->nodes[cur_level],
1853                                         &src_key, src_path->slots[cur_level]);
1854                 }
1855                 /* Content mismatch, something went wrong */
1856                 if (btrfs_comp_cpu_keys(&dst_key, &src_key)) {
1857                         ret = -ENOENT;
1858                         goto out;
1859                 }
1860                 cur_level--;
1861         }
1862
1863         /*
1864          * Now both @dst_path and @src_path have been populated, record the tree
1865          * blocks for qgroup accounting.
1866          */
1867         ret = btrfs_qgroup_trace_extent(trans, src_path->nodes[dst_level]->start,
1868                         nodesize, GFP_NOFS);
1869         if (ret < 0)
1870                 goto out;
1871         ret = btrfs_qgroup_trace_extent(trans,
1872                         dst_path->nodes[dst_level]->start,
1873                         nodesize, GFP_NOFS);
1874         if (ret < 0)
1875                 goto out;
1876
1877         /* Record leaf file extents */
1878         if (dst_level == 0 && trace_leaf) {
1879                 ret = btrfs_qgroup_trace_leaf_items(trans, src_path->nodes[0]);
1880                 if (ret < 0)
1881                         goto out;
1882                 ret = btrfs_qgroup_trace_leaf_items(trans, dst_path->nodes[0]);
1883         }
1884 out:
1885         btrfs_free_path(src_path);
1886         return ret;
1887 }
1888
1889 /*
1890  * Helper function to do recursive generation-aware depth-first search, to
1891  * locate all new tree blocks in a subtree of reloc tree.
1892  *
1893  * E.g. (OO = Old tree blocks, NN = New tree blocks, whose gen == last_snapshot)
1894  *         reloc tree
1895  * L2         NN (a)
1896  *          /    \
1897  * L1    OO        NN (b)
1898  *      /  \      /  \
1899  * L0  OO  OO    OO  NN
1900  *               (c) (d)
1901  * If we pass:
1902  * @dst_path = [ nodes[1] = NN(b), nodes[0] = NULL ],
1903  * @cur_level = 1
1904  * @root_level = 1
1905  *
1906  * We will iterate through tree blocks NN(b), NN(d) and info qgroup to trace
1907  * above tree blocks along with their counter parts in file tree.
1908  * While during search, old tree blocks OO(c) will be skipped as tree block swap
1909  * won't affect OO(c).
1910  */
1911 static int qgroup_trace_new_subtree_blocks(struct btrfs_trans_handle* trans,
1912                                            struct extent_buffer *src_eb,
1913                                            struct btrfs_path *dst_path,
1914                                            int cur_level, int root_level,
1915                                            u64 last_snapshot, bool trace_leaf)
1916 {
1917         struct btrfs_fs_info *fs_info = trans->fs_info;
1918         struct extent_buffer *eb;
1919         bool need_cleanup = false;
1920         int ret = 0;
1921         int i;
1922
1923         /* Level sanity check */
1924         if (cur_level < 0 || cur_level >= BTRFS_MAX_LEVEL - 1 ||
1925             root_level < 0 || root_level >= BTRFS_MAX_LEVEL - 1 ||
1926             root_level < cur_level) {
1927                 btrfs_err_rl(fs_info,
1928                         "%s: bad levels, cur_level=%d root_level=%d",
1929                         __func__, cur_level, root_level);
1930                 return -EUCLEAN;
1931         }
1932
1933         /* Read the tree block if needed */
1934         if (dst_path->nodes[cur_level] == NULL) {
1935                 struct btrfs_key first_key;
1936                 int parent_slot;
1937                 u64 child_gen;
1938                 u64 child_bytenr;
1939
1940                 /*
1941                  * dst_path->nodes[root_level] must be initialized before
1942                  * calling this function.
1943                  */
1944                 if (cur_level == root_level) {
1945                         btrfs_err_rl(fs_info,
1946         "%s: dst_path->nodes[%d] not initialized, root_level=%d cur_level=%d",
1947                                 __func__, root_level, root_level, cur_level);
1948                         return -EUCLEAN;
1949                 }
1950
1951                 /*
1952                  * We need to get child blockptr/gen from parent before we can
1953                  * read it.
1954                   */
1955                 eb = dst_path->nodes[cur_level + 1];
1956                 parent_slot = dst_path->slots[cur_level + 1];
1957                 child_bytenr = btrfs_node_blockptr(eb, parent_slot);
1958                 child_gen = btrfs_node_ptr_generation(eb, parent_slot);
1959                 btrfs_node_key_to_cpu(eb, &first_key, parent_slot);
1960
1961                 /* This node is old, no need to trace */
1962                 if (child_gen < last_snapshot)
1963                         goto out;
1964
1965                 eb = read_tree_block(fs_info, child_bytenr, child_gen,
1966                                      cur_level, &first_key);
1967                 if (IS_ERR(eb)) {
1968                         ret = PTR_ERR(eb);
1969                         goto out;
1970                 } else if (!extent_buffer_uptodate(eb)) {
1971                         free_extent_buffer(eb);
1972                         ret = -EIO;
1973                         goto out;
1974                 }
1975
1976                 dst_path->nodes[cur_level] = eb;
1977                 dst_path->slots[cur_level] = 0;
1978
1979                 btrfs_tree_read_lock(eb);
1980                 btrfs_set_lock_blocking_read(eb);
1981                 dst_path->locks[cur_level] = BTRFS_READ_LOCK_BLOCKING;
1982                 need_cleanup = true;
1983         }
1984
1985         /* Now record this tree block and its counter part for qgroups */
1986         ret = qgroup_trace_extent_swap(trans, src_eb, dst_path, cur_level,
1987                                        root_level, trace_leaf);
1988         if (ret < 0)
1989                 goto cleanup;
1990
1991         eb = dst_path->nodes[cur_level];
1992
1993         if (cur_level > 0) {
1994                 /* Iterate all child tree blocks */
1995                 for (i = 0; i < btrfs_header_nritems(eb); i++) {
1996                         /* Skip old tree blocks as they won't be swapped */
1997                         if (btrfs_node_ptr_generation(eb, i) < last_snapshot)
1998                                 continue;
1999                         dst_path->slots[cur_level] = i;
2000
2001                         /* Recursive call (at most 7 times) */
2002                         ret = qgroup_trace_new_subtree_blocks(trans, src_eb,
2003                                         dst_path, cur_level - 1, root_level,
2004                                         last_snapshot, trace_leaf);
2005                         if (ret < 0)
2006                                 goto cleanup;
2007                 }
2008         }
2009
2010 cleanup:
2011         if (need_cleanup) {
2012                 /* Clean up */
2013                 btrfs_tree_unlock_rw(dst_path->nodes[cur_level],
2014                                      dst_path->locks[cur_level]);
2015                 free_extent_buffer(dst_path->nodes[cur_level]);
2016                 dst_path->nodes[cur_level] = NULL;
2017                 dst_path->slots[cur_level] = 0;
2018                 dst_path->locks[cur_level] = 0;
2019         }
2020 out:
2021         return ret;
2022 }
2023
2024 static int qgroup_trace_subtree_swap(struct btrfs_trans_handle *trans,
2025                                 struct extent_buffer *src_eb,
2026                                 struct extent_buffer *dst_eb,
2027                                 u64 last_snapshot, bool trace_leaf)
2028 {
2029         struct btrfs_fs_info *fs_info = trans->fs_info;
2030         struct btrfs_path *dst_path = NULL;
2031         int level;
2032         int ret;
2033
2034         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2035                 return 0;
2036
2037         /* Wrong parameter order */
2038         if (btrfs_header_generation(src_eb) > btrfs_header_generation(dst_eb)) {
2039                 btrfs_err_rl(fs_info,
2040                 "%s: bad parameter order, src_gen=%llu dst_gen=%llu", __func__,
2041                              btrfs_header_generation(src_eb),
2042                              btrfs_header_generation(dst_eb));
2043                 return -EUCLEAN;
2044         }
2045
2046         if (!extent_buffer_uptodate(src_eb) || !extent_buffer_uptodate(dst_eb)) {
2047                 ret = -EIO;
2048                 goto out;
2049         }
2050
2051         level = btrfs_header_level(dst_eb);
2052         dst_path = btrfs_alloc_path();
2053         if (!dst_path) {
2054                 ret = -ENOMEM;
2055                 goto out;
2056         }
2057         /* For dst_path */
2058         extent_buffer_get(dst_eb);
2059         dst_path->nodes[level] = dst_eb;
2060         dst_path->slots[level] = 0;
2061         dst_path->locks[level] = 0;
2062
2063         /* Do the generation aware breadth-first search */
2064         ret = qgroup_trace_new_subtree_blocks(trans, src_eb, dst_path, level,
2065                                               level, last_snapshot, trace_leaf);
2066         if (ret < 0)
2067                 goto out;
2068         ret = 0;
2069
2070 out:
2071         btrfs_free_path(dst_path);
2072         if (ret < 0)
2073                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2074         return ret;
2075 }
2076
2077 int btrfs_qgroup_trace_subtree(struct btrfs_trans_handle *trans,
2078                                struct extent_buffer *root_eb,
2079                                u64 root_gen, int root_level)
2080 {
2081         struct btrfs_fs_info *fs_info = trans->fs_info;
2082         int ret = 0;
2083         int level;
2084         struct extent_buffer *eb = root_eb;
2085         struct btrfs_path *path = NULL;
2086
2087         BUG_ON(root_level < 0 || root_level >= BTRFS_MAX_LEVEL);
2088         BUG_ON(root_eb == NULL);
2089
2090         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2091                 return 0;
2092
2093         if (!extent_buffer_uptodate(root_eb)) {
2094                 ret = btrfs_read_buffer(root_eb, root_gen, root_level, NULL);
2095                 if (ret)
2096                         goto out;
2097         }
2098
2099         if (root_level == 0) {
2100                 ret = btrfs_qgroup_trace_leaf_items(trans, root_eb);
2101                 goto out;
2102         }
2103
2104         path = btrfs_alloc_path();
2105         if (!path)
2106                 return -ENOMEM;
2107
2108         /*
2109          * Walk down the tree.  Missing extent blocks are filled in as
2110          * we go. Metadata is accounted every time we read a new
2111          * extent block.
2112          *
2113          * When we reach a leaf, we account for file extent items in it,
2114          * walk back up the tree (adjusting slot pointers as we go)
2115          * and restart the search process.
2116          */
2117         extent_buffer_get(root_eb); /* For path */
2118         path->nodes[root_level] = root_eb;
2119         path->slots[root_level] = 0;
2120         path->locks[root_level] = 0; /* so release_path doesn't try to unlock */
2121 walk_down:
2122         level = root_level;
2123         while (level >= 0) {
2124                 if (path->nodes[level] == NULL) {
2125                         struct btrfs_key first_key;
2126                         int parent_slot;
2127                         u64 child_gen;
2128                         u64 child_bytenr;
2129
2130                         /*
2131                          * We need to get child blockptr/gen from parent before
2132                          * we can read it.
2133                           */
2134                         eb = path->nodes[level + 1];
2135                         parent_slot = path->slots[level + 1];
2136                         child_bytenr = btrfs_node_blockptr(eb, parent_slot);
2137                         child_gen = btrfs_node_ptr_generation(eb, parent_slot);
2138                         btrfs_node_key_to_cpu(eb, &first_key, parent_slot);
2139
2140                         eb = read_tree_block(fs_info, child_bytenr, child_gen,
2141                                              level, &first_key);
2142                         if (IS_ERR(eb)) {
2143                                 ret = PTR_ERR(eb);
2144                                 goto out;
2145                         } else if (!extent_buffer_uptodate(eb)) {
2146                                 free_extent_buffer(eb);
2147                                 ret = -EIO;
2148                                 goto out;
2149                         }
2150
2151                         path->nodes[level] = eb;
2152                         path->slots[level] = 0;
2153
2154                         btrfs_tree_read_lock(eb);
2155                         btrfs_set_lock_blocking_read(eb);
2156                         path->locks[level] = BTRFS_READ_LOCK_BLOCKING;
2157
2158                         ret = btrfs_qgroup_trace_extent(trans, child_bytenr,
2159                                                         fs_info->nodesize,
2160                                                         GFP_NOFS);
2161                         if (ret)
2162                                 goto out;
2163                 }
2164
2165                 if (level == 0) {
2166                         ret = btrfs_qgroup_trace_leaf_items(trans,
2167                                                             path->nodes[level]);
2168                         if (ret)
2169                                 goto out;
2170
2171                         /* Nonzero return here means we completed our search */
2172                         ret = adjust_slots_upwards(path, root_level);
2173                         if (ret)
2174                                 break;
2175
2176                         /* Restart search with new slots */
2177                         goto walk_down;
2178                 }
2179
2180                 level--;
2181         }
2182
2183         ret = 0;
2184 out:
2185         btrfs_free_path(path);
2186
2187         return ret;
2188 }
2189
2190 #define UPDATE_NEW      0
2191 #define UPDATE_OLD      1
2192 /*
2193  * Walk all of the roots that points to the bytenr and adjust their refcnts.
2194  */
2195 static int qgroup_update_refcnt(struct btrfs_fs_info *fs_info,
2196                                 struct ulist *roots, struct ulist *tmp,
2197                                 struct ulist *qgroups, u64 seq, int update_old)
2198 {
2199         struct ulist_node *unode;
2200         struct ulist_iterator uiter;
2201         struct ulist_node *tmp_unode;
2202         struct ulist_iterator tmp_uiter;
2203         struct btrfs_qgroup *qg;
2204         int ret = 0;
2205
2206         if (!roots)
2207                 return 0;
2208         ULIST_ITER_INIT(&uiter);
2209         while ((unode = ulist_next(roots, &uiter))) {
2210                 qg = find_qgroup_rb(fs_info, unode->val);
2211                 if (!qg)
2212                         continue;
2213
2214                 ulist_reinit(tmp);
2215                 ret = ulist_add(qgroups, qg->qgroupid, qgroup_to_aux(qg),
2216                                 GFP_ATOMIC);
2217                 if (ret < 0)
2218                         return ret;
2219                 ret = ulist_add(tmp, qg->qgroupid, qgroup_to_aux(qg), GFP_ATOMIC);
2220                 if (ret < 0)
2221                         return ret;
2222                 ULIST_ITER_INIT(&tmp_uiter);
2223                 while ((tmp_unode = ulist_next(tmp, &tmp_uiter))) {
2224                         struct btrfs_qgroup_list *glist;
2225
2226                         qg = unode_aux_to_qgroup(tmp_unode);
2227                         if (update_old)
2228                                 btrfs_qgroup_update_old_refcnt(qg, seq, 1);
2229                         else
2230                                 btrfs_qgroup_update_new_refcnt(qg, seq, 1);
2231                         list_for_each_entry(glist, &qg->groups, next_group) {
2232                                 ret = ulist_add(qgroups, glist->group->qgroupid,
2233                                                 qgroup_to_aux(glist->group),
2234                                                 GFP_ATOMIC);
2235                                 if (ret < 0)
2236                                         return ret;
2237                                 ret = ulist_add(tmp, glist->group->qgroupid,
2238                                                 qgroup_to_aux(glist->group),
2239                                                 GFP_ATOMIC);
2240                                 if (ret < 0)
2241                                         return ret;
2242                         }
2243                 }
2244         }
2245         return 0;
2246 }
2247
2248 /*
2249  * Update qgroup rfer/excl counters.
2250  * Rfer update is easy, codes can explain themselves.
2251  *
2252  * Excl update is tricky, the update is split into 2 part.
2253  * Part 1: Possible exclusive <-> sharing detect:
2254  *      |       A       |       !A      |
2255  *  -------------------------------------
2256  *  B   |       *       |       -       |
2257  *  -------------------------------------
2258  *  !B  |       +       |       **      |
2259  *  -------------------------------------
2260  *
2261  * Conditions:
2262  * A:   cur_old_roots < nr_old_roots    (not exclusive before)
2263  * !A:  cur_old_roots == nr_old_roots   (possible exclusive before)
2264  * B:   cur_new_roots < nr_new_roots    (not exclusive now)
2265  * !B:  cur_new_roots == nr_new_roots   (possible exclusive now)
2266  *
2267  * Results:
2268  * +: Possible sharing -> exclusive     -: Possible exclusive -> sharing
2269  * *: Definitely not changed.           **: Possible unchanged.
2270  *
2271  * For !A and !B condition, the exception is cur_old/new_roots == 0 case.
2272  *
2273  * To make the logic clear, we first use condition A and B to split
2274  * combination into 4 results.
2275  *
2276  * Then, for result "+" and "-", check old/new_roots == 0 case, as in them
2277  * only on variant maybe 0.
2278  *
2279  * Lastly, check result **, since there are 2 variants maybe 0, split them
2280  * again(2x2).
2281  * But this time we don't need to consider other things, the codes and logic
2282  * is easy to understand now.
2283  */
2284 static int qgroup_update_counters(struct btrfs_fs_info *fs_info,
2285                                   struct ulist *qgroups,
2286                                   u64 nr_old_roots,
2287                                   u64 nr_new_roots,
2288                                   u64 num_bytes, u64 seq)
2289 {
2290         struct ulist_node *unode;
2291         struct ulist_iterator uiter;
2292         struct btrfs_qgroup *qg;
2293         u64 cur_new_count, cur_old_count;
2294
2295         ULIST_ITER_INIT(&uiter);
2296         while ((unode = ulist_next(qgroups, &uiter))) {
2297                 bool dirty = false;
2298
2299                 qg = unode_aux_to_qgroup(unode);
2300                 cur_old_count = btrfs_qgroup_get_old_refcnt(qg, seq);
2301                 cur_new_count = btrfs_qgroup_get_new_refcnt(qg, seq);
2302
2303                 trace_qgroup_update_counters(fs_info, qg, cur_old_count,
2304                                              cur_new_count);
2305
2306                 /* Rfer update part */
2307                 if (cur_old_count == 0 && cur_new_count > 0) {
2308                         qg->rfer += num_bytes;
2309                         qg->rfer_cmpr += num_bytes;
2310                         dirty = true;
2311                 }
2312                 if (cur_old_count > 0 && cur_new_count == 0) {
2313                         qg->rfer -= num_bytes;
2314                         qg->rfer_cmpr -= num_bytes;
2315                         dirty = true;
2316                 }
2317
2318                 /* Excl update part */
2319                 /* Exclusive/none -> shared case */
2320                 if (cur_old_count == nr_old_roots &&
2321                     cur_new_count < nr_new_roots) {
2322                         /* Exclusive -> shared */
2323                         if (cur_old_count != 0) {
2324                                 qg->excl -= num_bytes;
2325                                 qg->excl_cmpr -= num_bytes;
2326                                 dirty = true;
2327                         }
2328                 }
2329
2330                 /* Shared -> exclusive/none case */
2331                 if (cur_old_count < nr_old_roots &&
2332                     cur_new_count == nr_new_roots) {
2333                         /* Shared->exclusive */
2334                         if (cur_new_count != 0) {
2335                                 qg->excl += num_bytes;
2336                                 qg->excl_cmpr += num_bytes;
2337                                 dirty = true;
2338                         }
2339                 }
2340
2341                 /* Exclusive/none -> exclusive/none case */
2342                 if (cur_old_count == nr_old_roots &&
2343                     cur_new_count == nr_new_roots) {
2344                         if (cur_old_count == 0) {
2345                                 /* None -> exclusive/none */
2346
2347                                 if (cur_new_count != 0) {
2348                                         /* None -> exclusive */
2349                                         qg->excl += num_bytes;
2350                                         qg->excl_cmpr += num_bytes;
2351                                         dirty = true;
2352                                 }
2353                                 /* None -> none, nothing changed */
2354                         } else {
2355                                 /* Exclusive -> exclusive/none */
2356
2357                                 if (cur_new_count == 0) {
2358                                         /* Exclusive -> none */
2359                                         qg->excl -= num_bytes;
2360                                         qg->excl_cmpr -= num_bytes;
2361                                         dirty = true;
2362                                 }
2363                                 /* Exclusive -> exclusive, nothing changed */
2364                         }
2365                 }
2366
2367                 if (dirty)
2368                         qgroup_dirty(fs_info, qg);
2369         }
2370         return 0;
2371 }
2372
2373 /*
2374  * Check if the @roots potentially is a list of fs tree roots
2375  *
2376  * Return 0 for definitely not a fs/subvol tree roots ulist
2377  * Return 1 for possible fs/subvol tree roots in the list (considering an empty
2378  *          one as well)
2379  */
2380 static int maybe_fs_roots(struct ulist *roots)
2381 {
2382         struct ulist_node *unode;
2383         struct ulist_iterator uiter;
2384
2385         /* Empty one, still possible for fs roots */
2386         if (!roots || roots->nnodes == 0)
2387                 return 1;
2388
2389         ULIST_ITER_INIT(&uiter);
2390         unode = ulist_next(roots, &uiter);
2391         if (!unode)
2392                 return 1;
2393
2394         /*
2395          * If it contains fs tree roots, then it must belong to fs/subvol
2396          * trees.
2397          * If it contains a non-fs tree, it won't be shared with fs/subvol trees.
2398          */
2399         return is_fstree(unode->val);
2400 }
2401
2402 int btrfs_qgroup_account_extent(struct btrfs_trans_handle *trans, u64 bytenr,
2403                                 u64 num_bytes, struct ulist *old_roots,
2404                                 struct ulist *new_roots)
2405 {
2406         struct btrfs_fs_info *fs_info = trans->fs_info;
2407         struct ulist *qgroups = NULL;
2408         struct ulist *tmp = NULL;
2409         u64 seq;
2410         u64 nr_new_roots = 0;
2411         u64 nr_old_roots = 0;
2412         int ret = 0;
2413
2414         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2415                 return 0;
2416
2417         if (new_roots) {
2418                 if (!maybe_fs_roots(new_roots))
2419                         goto out_free;
2420                 nr_new_roots = new_roots->nnodes;
2421         }
2422         if (old_roots) {
2423                 if (!maybe_fs_roots(old_roots))
2424                         goto out_free;
2425                 nr_old_roots = old_roots->nnodes;
2426         }
2427
2428         /* Quick exit, either not fs tree roots, or won't affect any qgroup */
2429         if (nr_old_roots == 0 && nr_new_roots == 0)
2430                 goto out_free;
2431
2432         BUG_ON(!fs_info->quota_root);
2433
2434         trace_btrfs_qgroup_account_extent(fs_info, trans->transid, bytenr,
2435                                         num_bytes, nr_old_roots, nr_new_roots);
2436
2437         qgroups = ulist_alloc(GFP_NOFS);
2438         if (!qgroups) {
2439                 ret = -ENOMEM;
2440                 goto out_free;
2441         }
2442         tmp = ulist_alloc(GFP_NOFS);
2443         if (!tmp) {
2444                 ret = -ENOMEM;
2445                 goto out_free;
2446         }
2447
2448         mutex_lock(&fs_info->qgroup_rescan_lock);
2449         if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
2450                 if (fs_info->qgroup_rescan_progress.objectid <= bytenr) {
2451                         mutex_unlock(&fs_info->qgroup_rescan_lock);
2452                         ret = 0;
2453                         goto out_free;
2454                 }
2455         }
2456         mutex_unlock(&fs_info->qgroup_rescan_lock);
2457
2458         spin_lock(&fs_info->qgroup_lock);
2459         seq = fs_info->qgroup_seq;
2460
2461         /* Update old refcnts using old_roots */
2462         ret = qgroup_update_refcnt(fs_info, old_roots, tmp, qgroups, seq,
2463                                    UPDATE_OLD);
2464         if (ret < 0)
2465                 goto out;
2466
2467         /* Update new refcnts using new_roots */
2468         ret = qgroup_update_refcnt(fs_info, new_roots, tmp, qgroups, seq,
2469                                    UPDATE_NEW);
2470         if (ret < 0)
2471                 goto out;
2472
2473         qgroup_update_counters(fs_info, qgroups, nr_old_roots, nr_new_roots,
2474                                num_bytes, seq);
2475
2476         /*
2477          * Bump qgroup_seq to avoid seq overlap
2478          */
2479         fs_info->qgroup_seq += max(nr_old_roots, nr_new_roots) + 1;
2480 out:
2481         spin_unlock(&fs_info->qgroup_lock);
2482 out_free:
2483         ulist_free(tmp);
2484         ulist_free(qgroups);
2485         ulist_free(old_roots);
2486         ulist_free(new_roots);
2487         return ret;
2488 }
2489
2490 int btrfs_qgroup_account_extents(struct btrfs_trans_handle *trans)
2491 {
2492         struct btrfs_fs_info *fs_info = trans->fs_info;
2493         struct btrfs_qgroup_extent_record *record;
2494         struct btrfs_delayed_ref_root *delayed_refs;
2495         struct ulist *new_roots = NULL;
2496         struct rb_node *node;
2497         u64 num_dirty_extents = 0;
2498         u64 qgroup_to_skip;
2499         int ret = 0;
2500
2501         delayed_refs = &trans->transaction->delayed_refs;
2502         qgroup_to_skip = delayed_refs->qgroup_to_skip;
2503         while ((node = rb_first(&delayed_refs->dirty_extent_root))) {
2504                 record = rb_entry(node, struct btrfs_qgroup_extent_record,
2505                                   node);
2506
2507                 num_dirty_extents++;
2508                 trace_btrfs_qgroup_account_extents(fs_info, record);
2509
2510                 if (!ret) {
2511                         /*
2512                          * Old roots should be searched when inserting qgroup
2513                          * extent record
2514                          */
2515                         if (WARN_ON(!record->old_roots)) {
2516                                 /* Search commit root to find old_roots */
2517                                 ret = btrfs_find_all_roots(NULL, fs_info,
2518                                                 record->bytenr, 0,
2519                                                 &record->old_roots, false);
2520                                 if (ret < 0)
2521                                         goto cleanup;
2522                         }
2523
2524                         /* Free the reserved data space */
2525                         btrfs_qgroup_free_refroot(fs_info,
2526                                         record->data_rsv_refroot,
2527                                         record->data_rsv,
2528                                         BTRFS_QGROUP_RSV_DATA);
2529                         /*
2530                          * Use SEQ_LAST as time_seq to do special search, which
2531                          * doesn't lock tree or delayed_refs and search current
2532                          * root. It's safe inside commit_transaction().
2533                          */
2534                         ret = btrfs_find_all_roots(trans, fs_info,
2535                                 record->bytenr, SEQ_LAST, &new_roots, false);
2536                         if (ret < 0)
2537                                 goto cleanup;
2538                         if (qgroup_to_skip) {
2539                                 ulist_del(new_roots, qgroup_to_skip, 0);
2540                                 ulist_del(record->old_roots, qgroup_to_skip,
2541                                           0);
2542                         }
2543                         ret = btrfs_qgroup_account_extent(trans, record->bytenr,
2544                                                           record->num_bytes,
2545                                                           record->old_roots,
2546                                                           new_roots);
2547                         record->old_roots = NULL;
2548                         new_roots = NULL;
2549                 }
2550 cleanup:
2551                 ulist_free(record->old_roots);
2552                 ulist_free(new_roots);
2553                 new_roots = NULL;
2554                 rb_erase(node, &delayed_refs->dirty_extent_root);
2555                 kfree(record);
2556
2557         }
2558         trace_qgroup_num_dirty_extents(fs_info, trans->transid,
2559                                        num_dirty_extents);
2560         return ret;
2561 }
2562
2563 /*
2564  * called from commit_transaction. Writes all changed qgroups to disk.
2565  */
2566 int btrfs_run_qgroups(struct btrfs_trans_handle *trans)
2567 {
2568         struct btrfs_fs_info *fs_info = trans->fs_info;
2569         struct btrfs_root *quota_root = fs_info->quota_root;
2570         int ret = 0;
2571
2572         if (!quota_root)
2573                 return ret;
2574
2575         spin_lock(&fs_info->qgroup_lock);
2576         while (!list_empty(&fs_info->dirty_qgroups)) {
2577                 struct btrfs_qgroup *qgroup;
2578                 qgroup = list_first_entry(&fs_info->dirty_qgroups,
2579                                           struct btrfs_qgroup, dirty);
2580                 list_del_init(&qgroup->dirty);
2581                 spin_unlock(&fs_info->qgroup_lock);
2582                 ret = update_qgroup_info_item(trans, qgroup);
2583                 if (ret)
2584                         fs_info->qgroup_flags |=
2585                                         BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2586                 ret = update_qgroup_limit_item(trans, qgroup);
2587                 if (ret)
2588                         fs_info->qgroup_flags |=
2589                                         BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2590                 spin_lock(&fs_info->qgroup_lock);
2591         }
2592         if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2593                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_ON;
2594         else
2595                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_ON;
2596         spin_unlock(&fs_info->qgroup_lock);
2597
2598         ret = update_qgroup_status_item(trans);
2599         if (ret)
2600                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2601
2602         return ret;
2603 }
2604
2605 /*
2606  * Copy the accounting information between qgroups. This is necessary
2607  * when a snapshot or a subvolume is created. Throwing an error will
2608  * cause a transaction abort so we take extra care here to only error
2609  * when a readonly fs is a reasonable outcome.
2610  */
2611 int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
2612                          u64 objectid, struct btrfs_qgroup_inherit *inherit)
2613 {
2614         int ret = 0;
2615         int i;
2616         u64 *i_qgroups;
2617         bool committing = false;
2618         struct btrfs_fs_info *fs_info = trans->fs_info;
2619         struct btrfs_root *quota_root;
2620         struct btrfs_qgroup *srcgroup;
2621         struct btrfs_qgroup *dstgroup;
2622         u32 level_size = 0;
2623         u64 nums;
2624
2625         /*
2626          * There are only two callers of this function.
2627          *
2628          * One in create_subvol() in the ioctl context, which needs to hold
2629          * the qgroup_ioctl_lock.
2630          *
2631          * The other one in create_pending_snapshot() where no other qgroup
2632          * code can modify the fs as they all need to either start a new trans
2633          * or hold a trans handler, thus we don't need to hold
2634          * qgroup_ioctl_lock.
2635          * This would avoid long and complex lock chain and make lockdep happy.
2636          */
2637         spin_lock(&fs_info->trans_lock);
2638         if (trans->transaction->state == TRANS_STATE_COMMIT_DOING)
2639                 committing = true;
2640         spin_unlock(&fs_info->trans_lock);
2641
2642         if (!committing)
2643                 mutex_lock(&fs_info->qgroup_ioctl_lock);
2644         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2645                 goto out;
2646
2647         quota_root = fs_info->quota_root;
2648         if (!quota_root) {
2649                 ret = -EINVAL;
2650                 goto out;
2651         }
2652
2653         if (inherit) {
2654                 i_qgroups = (u64 *)(inherit + 1);
2655                 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
2656                        2 * inherit->num_excl_copies;
2657                 for (i = 0; i < nums; ++i) {
2658                         srcgroup = find_qgroup_rb(fs_info, *i_qgroups);
2659
2660                         /*
2661                          * Zero out invalid groups so we can ignore
2662                          * them later.
2663                          */
2664                         if (!srcgroup ||
2665                             ((srcgroup->qgroupid >> 48) <= (objectid >> 48)))
2666                                 *i_qgroups = 0ULL;
2667
2668                         ++i_qgroups;
2669                 }
2670         }
2671
2672         /*
2673          * create a tracking group for the subvol itself
2674          */
2675         ret = add_qgroup_item(trans, quota_root, objectid);
2676         if (ret)
2677                 goto out;
2678
2679         /*
2680          * add qgroup to all inherited groups
2681          */
2682         if (inherit) {
2683                 i_qgroups = (u64 *)(inherit + 1);
2684                 for (i = 0; i < inherit->num_qgroups; ++i, ++i_qgroups) {
2685                         if (*i_qgroups == 0)
2686                                 continue;
2687                         ret = add_qgroup_relation_item(trans, objectid,
2688                                                        *i_qgroups);
2689                         if (ret && ret != -EEXIST)
2690                                 goto out;
2691                         ret = add_qgroup_relation_item(trans, *i_qgroups,
2692                                                        objectid);
2693                         if (ret && ret != -EEXIST)
2694                                 goto out;
2695                 }
2696                 ret = 0;
2697         }
2698
2699
2700         spin_lock(&fs_info->qgroup_lock);
2701
2702         dstgroup = add_qgroup_rb(fs_info, objectid);
2703         if (IS_ERR(dstgroup)) {
2704                 ret = PTR_ERR(dstgroup);
2705                 goto unlock;
2706         }
2707
2708         if (inherit && inherit->flags & BTRFS_QGROUP_INHERIT_SET_LIMITS) {
2709                 dstgroup->lim_flags = inherit->lim.flags;
2710                 dstgroup->max_rfer = inherit->lim.max_rfer;
2711                 dstgroup->max_excl = inherit->lim.max_excl;
2712                 dstgroup->rsv_rfer = inherit->lim.rsv_rfer;
2713                 dstgroup->rsv_excl = inherit->lim.rsv_excl;
2714
2715                 ret = update_qgroup_limit_item(trans, dstgroup);
2716                 if (ret) {
2717                         fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2718                         btrfs_info(fs_info,
2719                                    "unable to update quota limit for %llu",
2720                                    dstgroup->qgroupid);
2721                         goto unlock;
2722                 }
2723         }
2724
2725         if (srcid) {
2726                 srcgroup = find_qgroup_rb(fs_info, srcid);
2727                 if (!srcgroup)
2728                         goto unlock;
2729
2730                 /*
2731                  * We call inherit after we clone the root in order to make sure
2732                  * our counts don't go crazy, so at this point the only
2733                  * difference between the two roots should be the root node.
2734                  */
2735                 level_size = fs_info->nodesize;
2736                 dstgroup->rfer = srcgroup->rfer;
2737                 dstgroup->rfer_cmpr = srcgroup->rfer_cmpr;
2738                 dstgroup->excl = level_size;
2739                 dstgroup->excl_cmpr = level_size;
2740                 srcgroup->excl = level_size;
2741                 srcgroup->excl_cmpr = level_size;
2742
2743                 /* inherit the limit info */
2744                 dstgroup->lim_flags = srcgroup->lim_flags;
2745                 dstgroup->max_rfer = srcgroup->max_rfer;
2746                 dstgroup->max_excl = srcgroup->max_excl;
2747                 dstgroup->rsv_rfer = srcgroup->rsv_rfer;
2748                 dstgroup->rsv_excl = srcgroup->rsv_excl;
2749
2750                 qgroup_dirty(fs_info, dstgroup);
2751                 qgroup_dirty(fs_info, srcgroup);
2752         }
2753
2754         if (!inherit)
2755                 goto unlock;
2756
2757         i_qgroups = (u64 *)(inherit + 1);
2758         for (i = 0; i < inherit->num_qgroups; ++i) {
2759                 if (*i_qgroups) {
2760                         ret = add_relation_rb(fs_info, objectid, *i_qgroups);
2761                         if (ret)
2762                                 goto unlock;
2763                 }
2764                 ++i_qgroups;
2765         }
2766
2767         for (i = 0; i <  inherit->num_ref_copies; ++i, i_qgroups += 2) {
2768                 struct btrfs_qgroup *src;
2769                 struct btrfs_qgroup *dst;
2770
2771                 if (!i_qgroups[0] || !i_qgroups[1])
2772                         continue;
2773
2774                 src = find_qgroup_rb(fs_info, i_qgroups[0]);
2775                 dst = find_qgroup_rb(fs_info, i_qgroups[1]);
2776
2777                 if (!src || !dst) {
2778                         ret = -EINVAL;
2779                         goto unlock;
2780                 }
2781
2782                 dst->rfer = src->rfer - level_size;
2783                 dst->rfer_cmpr = src->rfer_cmpr - level_size;
2784         }
2785         for (i = 0; i <  inherit->num_excl_copies; ++i, i_qgroups += 2) {
2786                 struct btrfs_qgroup *src;
2787                 struct btrfs_qgroup *dst;
2788
2789                 if (!i_qgroups[0] || !i_qgroups[1])
2790                         continue;
2791
2792                 src = find_qgroup_rb(fs_info, i_qgroups[0]);
2793                 dst = find_qgroup_rb(fs_info, i_qgroups[1]);
2794
2795                 if (!src || !dst) {
2796                         ret = -EINVAL;
2797                         goto unlock;
2798                 }
2799
2800                 dst->excl = src->excl + level_size;
2801                 dst->excl_cmpr = src->excl_cmpr + level_size;
2802         }
2803
2804 unlock:
2805         spin_unlock(&fs_info->qgroup_lock);
2806 out:
2807         if (!committing)
2808                 mutex_unlock(&fs_info->qgroup_ioctl_lock);
2809         return ret;
2810 }
2811
2812 /*
2813  * Two limits to commit transaction in advance.
2814  *
2815  * For RATIO, it will be 1/RATIO of the remaining limit as threshold.
2816  * For SIZE, it will be in byte unit as threshold.
2817  */
2818 #define QGROUP_FREE_RATIO               32
2819 #define QGROUP_FREE_SIZE                SZ_32M
2820 static bool qgroup_check_limits(struct btrfs_fs_info *fs_info,
2821                                 const struct btrfs_qgroup *qg, u64 num_bytes)
2822 {
2823         u64 free;
2824         u64 threshold;
2825
2826         if ((qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_RFER) &&
2827             qgroup_rsv_total(qg) + (s64)qg->rfer + num_bytes > qg->max_rfer)
2828                 return false;
2829
2830         if ((qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_EXCL) &&
2831             qgroup_rsv_total(qg) + (s64)qg->excl + num_bytes > qg->max_excl)
2832                 return false;
2833
2834         /*
2835          * Even if we passed the check, it's better to check if reservation
2836          * for meta_pertrans is pushing us near limit.
2837          * If there is too much pertrans reservation or it's near the limit,
2838          * let's try commit transaction to free some, using transaction_kthread
2839          */
2840         if ((qg->lim_flags & (BTRFS_QGROUP_LIMIT_MAX_RFER |
2841                               BTRFS_QGROUP_LIMIT_MAX_EXCL))) {
2842                 if (qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_EXCL) {
2843                         free = qg->max_excl - qgroup_rsv_total(qg) - qg->excl;
2844                         threshold = min_t(u64, qg->max_excl / QGROUP_FREE_RATIO,
2845                                           QGROUP_FREE_SIZE);
2846                 } else {
2847                         free = qg->max_rfer - qgroup_rsv_total(qg) - qg->rfer;
2848                         threshold = min_t(u64, qg->max_rfer / QGROUP_FREE_RATIO,
2849                                           QGROUP_FREE_SIZE);
2850                 }
2851
2852                 /*
2853                  * Use transaction_kthread to commit transaction, so we no
2854                  * longer need to bother nested transaction nor lock context.
2855                  */
2856                 if (free < threshold)
2857                         btrfs_commit_transaction_locksafe(fs_info);
2858         }
2859
2860         return true;
2861 }
2862
2863 static int qgroup_reserve(struct btrfs_root *root, u64 num_bytes, bool enforce,
2864                           enum btrfs_qgroup_rsv_type type)
2865 {
2866         struct btrfs_root *quota_root;
2867         struct btrfs_qgroup *qgroup;
2868         struct btrfs_fs_info *fs_info = root->fs_info;
2869         u64 ref_root = root->root_key.objectid;
2870         int ret = 0;
2871         struct ulist_node *unode;
2872         struct ulist_iterator uiter;
2873
2874         if (!is_fstree(ref_root))
2875                 return 0;
2876
2877         if (num_bytes == 0)
2878                 return 0;
2879
2880         if (test_bit(BTRFS_FS_QUOTA_OVERRIDE, &fs_info->flags) &&
2881             capable(CAP_SYS_RESOURCE))
2882                 enforce = false;
2883
2884         spin_lock(&fs_info->qgroup_lock);
2885         quota_root = fs_info->quota_root;
2886         if (!quota_root)
2887                 goto out;
2888
2889         qgroup = find_qgroup_rb(fs_info, ref_root);
2890         if (!qgroup)
2891                 goto out;
2892
2893         /*
2894          * in a first step, we check all affected qgroups if any limits would
2895          * be exceeded
2896          */
2897         ulist_reinit(fs_info->qgroup_ulist);
2898         ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
2899                         qgroup_to_aux(qgroup), GFP_ATOMIC);
2900         if (ret < 0)
2901                 goto out;
2902         ULIST_ITER_INIT(&uiter);
2903         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
2904                 struct btrfs_qgroup *qg;
2905                 struct btrfs_qgroup_list *glist;
2906
2907                 qg = unode_aux_to_qgroup(unode);
2908
2909                 if (enforce && !qgroup_check_limits(fs_info, qg, num_bytes)) {
2910                         ret = -EDQUOT;
2911                         goto out;
2912                 }
2913
2914                 list_for_each_entry(glist, &qg->groups, next_group) {
2915                         ret = ulist_add(fs_info->qgroup_ulist,
2916                                         glist->group->qgroupid,
2917                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
2918                         if (ret < 0)
2919                                 goto out;
2920                 }
2921         }
2922         ret = 0;
2923         /*
2924          * no limits exceeded, now record the reservation into all qgroups
2925          */
2926         ULIST_ITER_INIT(&uiter);
2927         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
2928                 struct btrfs_qgroup *qg;
2929
2930                 qg = unode_aux_to_qgroup(unode);
2931
2932                 qgroup_rsv_add(fs_info, qg, num_bytes, type);
2933         }
2934
2935 out:
2936         spin_unlock(&fs_info->qgroup_lock);
2937         return ret;
2938 }
2939
2940 /*
2941  * Free @num_bytes of reserved space with @type for qgroup.  (Normally level 0
2942  * qgroup).
2943  *
2944  * Will handle all higher level qgroup too.
2945  *
2946  * NOTE: If @num_bytes is (u64)-1, this means to free all bytes of this qgroup.
2947  * This special case is only used for META_PERTRANS type.
2948  */
2949 void btrfs_qgroup_free_refroot(struct btrfs_fs_info *fs_info,
2950                                u64 ref_root, u64 num_bytes,
2951                                enum btrfs_qgroup_rsv_type type)
2952 {
2953         struct btrfs_root *quota_root;
2954         struct btrfs_qgroup *qgroup;
2955         struct ulist_node *unode;
2956         struct ulist_iterator uiter;
2957         int ret = 0;
2958
2959         if (!is_fstree(ref_root))
2960                 return;
2961
2962         if (num_bytes == 0)
2963                 return;
2964
2965         if (num_bytes == (u64)-1 && type != BTRFS_QGROUP_RSV_META_PERTRANS) {
2966                 WARN(1, "%s: Invalid type to free", __func__);
2967                 return;
2968         }
2969         spin_lock(&fs_info->qgroup_lock);
2970
2971         quota_root = fs_info->quota_root;
2972         if (!quota_root)
2973                 goto out;
2974
2975         qgroup = find_qgroup_rb(fs_info, ref_root);
2976         if (!qgroup)
2977                 goto out;
2978
2979         if (num_bytes == (u64)-1)
2980                 /*
2981                  * We're freeing all pertrans rsv, get reserved value from
2982                  * level 0 qgroup as real num_bytes to free.
2983                  */
2984                 num_bytes = qgroup->rsv.values[type];
2985
2986         ulist_reinit(fs_info->qgroup_ulist);
2987         ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
2988                         qgroup_to_aux(qgroup), GFP_ATOMIC);
2989         if (ret < 0)
2990                 goto out;
2991         ULIST_ITER_INIT(&uiter);
2992         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
2993                 struct btrfs_qgroup *qg;
2994                 struct btrfs_qgroup_list *glist;
2995
2996                 qg = unode_aux_to_qgroup(unode);
2997
2998                 qgroup_rsv_release(fs_info, qg, num_bytes, type);
2999
3000                 list_for_each_entry(glist, &qg->groups, next_group) {
3001                         ret = ulist_add(fs_info->qgroup_ulist,
3002                                         glist->group->qgroupid,
3003                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
3004                         if (ret < 0)
3005                                 goto out;
3006                 }
3007         }
3008
3009 out:
3010         spin_unlock(&fs_info->qgroup_lock);
3011 }
3012
3013 /*
3014  * Check if the leaf is the last leaf. Which means all node pointers
3015  * are at their last position.
3016  */
3017 static bool is_last_leaf(struct btrfs_path *path)
3018 {
3019         int i;
3020
3021         for (i = 1; i < BTRFS_MAX_LEVEL && path->nodes[i]; i++) {
3022                 if (path->slots[i] != btrfs_header_nritems(path->nodes[i]) - 1)
3023                         return false;
3024         }
3025         return true;
3026 }
3027
3028 /*
3029  * returns < 0 on error, 0 when more leafs are to be scanned.
3030  * returns 1 when done.
3031  */
3032 static int qgroup_rescan_leaf(struct btrfs_trans_handle *trans,
3033                               struct btrfs_path *path)
3034 {
3035         struct btrfs_fs_info *fs_info = trans->fs_info;
3036         struct btrfs_key found;
3037         struct extent_buffer *scratch_leaf = NULL;
3038         struct ulist *roots = NULL;
3039         u64 num_bytes;
3040         bool done;
3041         int slot;
3042         int ret;
3043
3044         mutex_lock(&fs_info->qgroup_rescan_lock);
3045         ret = btrfs_search_slot_for_read(fs_info->extent_root,
3046                                          &fs_info->qgroup_rescan_progress,
3047                                          path, 1, 0);
3048
3049         btrfs_debug(fs_info,
3050                 "current progress key (%llu %u %llu), search_slot ret %d",
3051                 fs_info->qgroup_rescan_progress.objectid,
3052                 fs_info->qgroup_rescan_progress.type,
3053                 fs_info->qgroup_rescan_progress.offset, ret);
3054
3055         if (ret) {
3056                 /*
3057                  * The rescan is about to end, we will not be scanning any
3058                  * further blocks. We cannot unset the RESCAN flag here, because
3059                  * we want to commit the transaction if everything went well.
3060                  * To make the live accounting work in this phase, we set our
3061                  * scan progress pointer such that every real extent objectid
3062                  * will be smaller.
3063                  */
3064                 fs_info->qgroup_rescan_progress.objectid = (u64)-1;
3065                 btrfs_release_path(path);
3066                 mutex_unlock(&fs_info->qgroup_rescan_lock);
3067                 return ret;
3068         }
3069         done = is_last_leaf(path);
3070
3071         btrfs_item_key_to_cpu(path->nodes[0], &found,
3072                               btrfs_header_nritems(path->nodes[0]) - 1);
3073         fs_info->qgroup_rescan_progress.objectid = found.objectid + 1;
3074
3075         scratch_leaf = btrfs_clone_extent_buffer(path->nodes[0]);
3076         if (!scratch_leaf) {
3077                 ret = -ENOMEM;
3078                 mutex_unlock(&fs_info->qgroup_rescan_lock);
3079                 goto out;
3080         }
3081         slot = path->slots[0];
3082         btrfs_release_path(path);
3083         mutex_unlock(&fs_info->qgroup_rescan_lock);
3084
3085         for (; slot < btrfs_header_nritems(scratch_leaf); ++slot) {
3086                 btrfs_item_key_to_cpu(scratch_leaf, &found, slot);
3087                 if (found.type != BTRFS_EXTENT_ITEM_KEY &&
3088                     found.type != BTRFS_METADATA_ITEM_KEY)
3089                         continue;
3090                 if (found.type == BTRFS_METADATA_ITEM_KEY)
3091                         num_bytes = fs_info->nodesize;
3092                 else
3093                         num_bytes = found.offset;
3094
3095                 ret = btrfs_find_all_roots(NULL, fs_info, found.objectid, 0,
3096                                            &roots, false);
3097                 if (ret < 0)
3098                         goto out;
3099                 /* For rescan, just pass old_roots as NULL */
3100                 ret = btrfs_qgroup_account_extent(trans, found.objectid,
3101                                                   num_bytes, NULL, roots);
3102                 if (ret < 0)
3103                         goto out;
3104         }
3105 out:
3106         if (scratch_leaf)
3107                 free_extent_buffer(scratch_leaf);
3108
3109         if (done && !ret) {
3110                 ret = 1;
3111                 fs_info->qgroup_rescan_progress.objectid = (u64)-1;
3112         }
3113         return ret;
3114 }
3115
3116 static void btrfs_qgroup_rescan_worker(struct btrfs_work *work)
3117 {
3118         struct btrfs_fs_info *fs_info = container_of(work, struct btrfs_fs_info,
3119                                                      qgroup_rescan_work);
3120         struct btrfs_path *path;
3121         struct btrfs_trans_handle *trans = NULL;
3122         int err = -ENOMEM;
3123         int ret = 0;
3124
3125         path = btrfs_alloc_path();
3126         if (!path)
3127                 goto out;
3128         /*
3129          * Rescan should only search for commit root, and any later difference
3130          * should be recorded by qgroup
3131          */
3132         path->search_commit_root = 1;
3133         path->skip_locking = 1;
3134
3135         err = 0;
3136         while (!err && !btrfs_fs_closing(fs_info)) {
3137                 trans = btrfs_start_transaction(fs_info->fs_root, 0);
3138                 if (IS_ERR(trans)) {
3139                         err = PTR_ERR(trans);
3140                         break;
3141                 }
3142                 if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags)) {
3143                         err = -EINTR;
3144                 } else {
3145                         err = qgroup_rescan_leaf(trans, path);
3146                 }
3147                 if (err > 0)
3148                         btrfs_commit_transaction(trans);
3149                 else
3150                         btrfs_end_transaction(trans);
3151         }
3152
3153 out:
3154         btrfs_free_path(path);
3155
3156         mutex_lock(&fs_info->qgroup_rescan_lock);
3157         if (!btrfs_fs_closing(fs_info))
3158                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3159
3160         if (err > 0 &&
3161             fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT) {
3162                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
3163         } else if (err < 0) {
3164                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
3165         }
3166         mutex_unlock(&fs_info->qgroup_rescan_lock);
3167
3168         /*
3169          * only update status, since the previous part has already updated the
3170          * qgroup info.
3171          */
3172         trans = btrfs_start_transaction(fs_info->quota_root, 1);
3173         if (IS_ERR(trans)) {
3174                 err = PTR_ERR(trans);
3175                 btrfs_err(fs_info,
3176                           "fail to start transaction for status update: %d",
3177                           err);
3178                 goto done;
3179         }
3180         ret = update_qgroup_status_item(trans);
3181         if (ret < 0) {
3182                 err = ret;
3183                 btrfs_err(fs_info, "fail to update qgroup status: %d", err);
3184         }
3185         btrfs_end_transaction(trans);
3186
3187         if (btrfs_fs_closing(fs_info)) {
3188                 btrfs_info(fs_info, "qgroup scan paused");
3189         } else if (err >= 0) {
3190                 btrfs_info(fs_info, "qgroup scan completed%s",
3191                         err > 0 ? " (inconsistency flag cleared)" : "");
3192         } else {
3193                 btrfs_err(fs_info, "qgroup scan failed with %d", err);
3194         }
3195
3196 done:
3197         mutex_lock(&fs_info->qgroup_rescan_lock);
3198         fs_info->qgroup_rescan_running = false;
3199         mutex_unlock(&fs_info->qgroup_rescan_lock);
3200         complete_all(&fs_info->qgroup_rescan_completion);
3201 }
3202
3203 /*
3204  * Checks that (a) no rescan is running and (b) quota is enabled. Allocates all
3205  * memory required for the rescan context.
3206  */
3207 static int
3208 qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid,
3209                    int init_flags)
3210 {
3211         int ret = 0;
3212
3213         if (!init_flags) {
3214                 /* we're resuming qgroup rescan at mount time */
3215                 if (!(fs_info->qgroup_flags &
3216                       BTRFS_QGROUP_STATUS_FLAG_RESCAN)) {
3217                         btrfs_warn(fs_info,
3218                         "qgroup rescan init failed, qgroup is not enabled");
3219                         ret = -EINVAL;
3220                 } else if (!(fs_info->qgroup_flags &
3221                              BTRFS_QGROUP_STATUS_FLAG_ON)) {
3222                         btrfs_warn(fs_info,
3223                         "qgroup rescan init failed, qgroup rescan is not queued");
3224                         ret = -EINVAL;
3225                 }
3226
3227                 if (ret)
3228                         return ret;
3229         }
3230
3231         mutex_lock(&fs_info->qgroup_rescan_lock);
3232         spin_lock(&fs_info->qgroup_lock);
3233
3234         if (init_flags) {
3235                 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3236                         btrfs_warn(fs_info,
3237                                    "qgroup rescan is already in progress");
3238                         ret = -EINPROGRESS;
3239                 } else if (!(fs_info->qgroup_flags &
3240                              BTRFS_QGROUP_STATUS_FLAG_ON)) {
3241                         btrfs_warn(fs_info,
3242                         "qgroup rescan init failed, qgroup is not enabled");
3243                         ret = -EINVAL;
3244                 }
3245
3246                 if (ret) {
3247                         spin_unlock(&fs_info->qgroup_lock);
3248                         mutex_unlock(&fs_info->qgroup_rescan_lock);
3249                         return ret;
3250                 }
3251                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3252         }
3253
3254         memset(&fs_info->qgroup_rescan_progress, 0,
3255                 sizeof(fs_info->qgroup_rescan_progress));
3256         fs_info->qgroup_rescan_progress.objectid = progress_objectid;
3257         init_completion(&fs_info->qgroup_rescan_completion);
3258         fs_info->qgroup_rescan_running = true;
3259
3260         spin_unlock(&fs_info->qgroup_lock);
3261         mutex_unlock(&fs_info->qgroup_rescan_lock);
3262
3263         memset(&fs_info->qgroup_rescan_work, 0,
3264                sizeof(fs_info->qgroup_rescan_work));
3265         btrfs_init_work(&fs_info->qgroup_rescan_work,
3266                         btrfs_qgroup_rescan_helper,
3267                         btrfs_qgroup_rescan_worker, NULL, NULL);
3268         return 0;
3269 }
3270
3271 static void
3272 qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info)
3273 {
3274         struct rb_node *n;
3275         struct btrfs_qgroup *qgroup;
3276
3277         spin_lock(&fs_info->qgroup_lock);
3278         /* clear all current qgroup tracking information */
3279         for (n = rb_first(&fs_info->qgroup_tree); n; n = rb_next(n)) {
3280                 qgroup = rb_entry(n, struct btrfs_qgroup, node);
3281                 qgroup->rfer = 0;
3282                 qgroup->rfer_cmpr = 0;
3283                 qgroup->excl = 0;
3284                 qgroup->excl_cmpr = 0;
3285                 qgroup_dirty(fs_info, qgroup);
3286         }
3287         spin_unlock(&fs_info->qgroup_lock);
3288 }
3289
3290 int
3291 btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info)
3292 {
3293         int ret = 0;
3294         struct btrfs_trans_handle *trans;
3295
3296         ret = qgroup_rescan_init(fs_info, 0, 1);
3297         if (ret)
3298                 return ret;
3299
3300         /*
3301          * We have set the rescan_progress to 0, which means no more
3302          * delayed refs will be accounted by btrfs_qgroup_account_ref.
3303          * However, btrfs_qgroup_account_ref may be right after its call
3304          * to btrfs_find_all_roots, in which case it would still do the
3305          * accounting.
3306          * To solve this, we're committing the transaction, which will
3307          * ensure we run all delayed refs and only after that, we are
3308          * going to clear all tracking information for a clean start.
3309          */
3310
3311         trans = btrfs_join_transaction(fs_info->fs_root);
3312         if (IS_ERR(trans)) {
3313                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3314                 return PTR_ERR(trans);
3315         }
3316         ret = btrfs_commit_transaction(trans);
3317         if (ret) {
3318                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3319                 return ret;
3320         }
3321
3322         qgroup_rescan_zero_tracking(fs_info);
3323
3324         btrfs_queue_work(fs_info->qgroup_rescan_workers,
3325                          &fs_info->qgroup_rescan_work);
3326
3327         return 0;
3328 }
3329
3330 int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info,
3331                                      bool interruptible)
3332 {
3333         int running;
3334         int ret = 0;
3335
3336         mutex_lock(&fs_info->qgroup_rescan_lock);
3337         spin_lock(&fs_info->qgroup_lock);
3338         running = fs_info->qgroup_rescan_running;
3339         spin_unlock(&fs_info->qgroup_lock);
3340         mutex_unlock(&fs_info->qgroup_rescan_lock);
3341
3342         if (!running)
3343                 return 0;
3344
3345         if (interruptible)
3346                 ret = wait_for_completion_interruptible(
3347                                         &fs_info->qgroup_rescan_completion);
3348         else
3349                 wait_for_completion(&fs_info->qgroup_rescan_completion);
3350
3351         return ret;
3352 }
3353
3354 /*
3355  * this is only called from open_ctree where we're still single threaded, thus
3356  * locking is omitted here.
3357  */
3358 void
3359 btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info)
3360 {
3361         if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN)
3362                 btrfs_queue_work(fs_info->qgroup_rescan_workers,
3363                                  &fs_info->qgroup_rescan_work);
3364 }
3365
3366 /*
3367  * Reserve qgroup space for range [start, start + len).
3368  *
3369  * This function will either reserve space from related qgroups or doing
3370  * nothing if the range is already reserved.
3371  *
3372  * Return 0 for successful reserve
3373  * Return <0 for error (including -EQUOT)
3374  *
3375  * NOTE: this function may sleep for memory allocation.
3376  *       if btrfs_qgroup_reserve_data() is called multiple times with
3377  *       same @reserved, caller must ensure when error happens it's OK
3378  *       to free *ALL* reserved space.
3379  */
3380 int btrfs_qgroup_reserve_data(struct inode *inode,
3381                         struct extent_changeset **reserved_ret, u64 start,
3382                         u64 len)
3383 {
3384         struct btrfs_root *root = BTRFS_I(inode)->root;
3385         struct ulist_node *unode;
3386         struct ulist_iterator uiter;
3387         struct extent_changeset *reserved;
3388         u64 orig_reserved;
3389         u64 to_reserve;
3390         int ret;
3391
3392         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &root->fs_info->flags) ||
3393             !is_fstree(root->root_key.objectid) || len == 0)
3394                 return 0;
3395
3396         /* @reserved parameter is mandatory for qgroup */
3397         if (WARN_ON(!reserved_ret))
3398                 return -EINVAL;
3399         if (!*reserved_ret) {
3400                 *reserved_ret = extent_changeset_alloc();
3401                 if (!*reserved_ret)
3402                         return -ENOMEM;
3403         }
3404         reserved = *reserved_ret;
3405         /* Record already reserved space */
3406         orig_reserved = reserved->bytes_changed;
3407         ret = set_record_extent_bits(&BTRFS_I(inode)->io_tree, start,
3408                         start + len -1, EXTENT_QGROUP_RESERVED, reserved);
3409
3410         /* Newly reserved space */
3411         to_reserve = reserved->bytes_changed - orig_reserved;
3412         trace_btrfs_qgroup_reserve_data(inode, start, len,
3413                                         to_reserve, QGROUP_RESERVE);
3414         if (ret < 0)
3415                 goto cleanup;
3416         ret = qgroup_reserve(root, to_reserve, true, BTRFS_QGROUP_RSV_DATA);
3417         if (ret < 0)
3418                 goto cleanup;
3419
3420         return ret;
3421
3422 cleanup:
3423         /* cleanup *ALL* already reserved ranges */
3424         ULIST_ITER_INIT(&uiter);
3425         while ((unode = ulist_next(&reserved->range_changed, &uiter)))
3426                 clear_extent_bit(&BTRFS_I(inode)->io_tree, unode->val,
3427                                  unode->aux, EXTENT_QGROUP_RESERVED, 0, 0, NULL);
3428         extent_changeset_release(reserved);
3429         return ret;
3430 }
3431
3432 /* Free ranges specified by @reserved, normally in error path */
3433 static int qgroup_free_reserved_data(struct inode *inode,
3434                         struct extent_changeset *reserved, u64 start, u64 len)
3435 {
3436         struct btrfs_root *root = BTRFS_I(inode)->root;
3437         struct ulist_node *unode;
3438         struct ulist_iterator uiter;
3439         struct extent_changeset changeset;
3440         int freed = 0;
3441         int ret;
3442
3443         extent_changeset_init(&changeset);
3444         len = round_up(start + len, root->fs_info->sectorsize);
3445         start = round_down(start, root->fs_info->sectorsize);
3446
3447         ULIST_ITER_INIT(&uiter);
3448         while ((unode = ulist_next(&reserved->range_changed, &uiter))) {
3449                 u64 range_start = unode->val;
3450                 /* unode->aux is the inclusive end */
3451                 u64 range_len = unode->aux - range_start + 1;
3452                 u64 free_start;
3453                 u64 free_len;
3454
3455                 extent_changeset_release(&changeset);
3456
3457                 /* Only free range in range [start, start + len) */
3458                 if (range_start >= start + len ||
3459                     range_start + range_len <= start)
3460                         continue;
3461                 free_start = max(range_start, start);
3462                 free_len = min(start + len, range_start + range_len) -
3463                            free_start;
3464                 /*
3465                  * TODO: To also modify reserved->ranges_reserved to reflect
3466                  * the modification.
3467                  *
3468                  * However as long as we free qgroup reserved according to
3469                  * EXTENT_QGROUP_RESERVED, we won't double free.
3470                  * So not need to rush.
3471                  */
3472                 ret = clear_record_extent_bits(&BTRFS_I(inode)->io_failure_tree,
3473                                 free_start, free_start + free_len - 1,
3474                                 EXTENT_QGROUP_RESERVED, &changeset);
3475                 if (ret < 0)
3476                         goto out;
3477                 freed += changeset.bytes_changed;
3478         }
3479         btrfs_qgroup_free_refroot(root->fs_info, root->root_key.objectid, freed,
3480                                   BTRFS_QGROUP_RSV_DATA);
3481         ret = freed;
3482 out:
3483         extent_changeset_release(&changeset);
3484         return ret;
3485 }
3486
3487 static int __btrfs_qgroup_release_data(struct inode *inode,
3488                         struct extent_changeset *reserved, u64 start, u64 len,
3489                         int free)
3490 {
3491         struct extent_changeset changeset;
3492         int trace_op = QGROUP_RELEASE;
3493         int ret;
3494
3495         if (!test_bit(BTRFS_FS_QUOTA_ENABLED,
3496                       &BTRFS_I(inode)->root->fs_info->flags))
3497                 return 0;
3498
3499         /* In release case, we shouldn't have @reserved */
3500         WARN_ON(!free && reserved);
3501         if (free && reserved)
3502                 return qgroup_free_reserved_data(inode, reserved, start, len);
3503         extent_changeset_init(&changeset);
3504         ret = clear_record_extent_bits(&BTRFS_I(inode)->io_tree, start,
3505                         start + len -1, EXTENT_QGROUP_RESERVED, &changeset);
3506         if (ret < 0)
3507                 goto out;
3508
3509         if (free)
3510                 trace_op = QGROUP_FREE;
3511         trace_btrfs_qgroup_release_data(inode, start, len,
3512                                         changeset.bytes_changed, trace_op);
3513         if (free)
3514                 btrfs_qgroup_free_refroot(BTRFS_I(inode)->root->fs_info,
3515                                 BTRFS_I(inode)->root->root_key.objectid,
3516                                 changeset.bytes_changed, BTRFS_QGROUP_RSV_DATA);
3517         ret = changeset.bytes_changed;
3518 out:
3519         extent_changeset_release(&changeset);
3520         return ret;
3521 }
3522
3523 /*
3524  * Free a reserved space range from io_tree and related qgroups
3525  *
3526  * Should be called when a range of pages get invalidated before reaching disk.
3527  * Or for error cleanup case.
3528  * if @reserved is given, only reserved range in [@start, @start + @len) will
3529  * be freed.
3530  *
3531  * For data written to disk, use btrfs_qgroup_release_data().
3532  *
3533  * NOTE: This function may sleep for memory allocation.
3534  */
3535 int btrfs_qgroup_free_data(struct inode *inode,
3536                         struct extent_changeset *reserved, u64 start, u64 len)
3537 {
3538         return __btrfs_qgroup_release_data(inode, reserved, start, len, 1);
3539 }
3540
3541 /*
3542  * Release a reserved space range from io_tree only.
3543  *
3544  * Should be called when a range of pages get written to disk and corresponding
3545  * FILE_EXTENT is inserted into corresponding root.
3546  *
3547  * Since new qgroup accounting framework will only update qgroup numbers at
3548  * commit_transaction() time, its reserved space shouldn't be freed from
3549  * related qgroups.
3550  *
3551  * But we should release the range from io_tree, to allow further write to be
3552  * COWed.
3553  *
3554  * NOTE: This function may sleep for memory allocation.
3555  */
3556 int btrfs_qgroup_release_data(struct inode *inode, u64 start, u64 len)
3557 {
3558         return __btrfs_qgroup_release_data(inode, NULL, start, len, 0);
3559 }
3560
3561 static void add_root_meta_rsv(struct btrfs_root *root, int num_bytes,
3562                               enum btrfs_qgroup_rsv_type type)
3563 {
3564         if (type != BTRFS_QGROUP_RSV_META_PREALLOC &&
3565             type != BTRFS_QGROUP_RSV_META_PERTRANS)
3566                 return;
3567         if (num_bytes == 0)
3568                 return;
3569
3570         spin_lock(&root->qgroup_meta_rsv_lock);
3571         if (type == BTRFS_QGROUP_RSV_META_PREALLOC)
3572                 root->qgroup_meta_rsv_prealloc += num_bytes;
3573         else
3574                 root->qgroup_meta_rsv_pertrans += num_bytes;
3575         spin_unlock(&root->qgroup_meta_rsv_lock);
3576 }
3577
3578 static int sub_root_meta_rsv(struct btrfs_root *root, int num_bytes,
3579                              enum btrfs_qgroup_rsv_type type)
3580 {
3581         if (type != BTRFS_QGROUP_RSV_META_PREALLOC &&
3582             type != BTRFS_QGROUP_RSV_META_PERTRANS)
3583                 return 0;
3584         if (num_bytes == 0)
3585                 return 0;
3586
3587         spin_lock(&root->qgroup_meta_rsv_lock);
3588         if (type == BTRFS_QGROUP_RSV_META_PREALLOC) {
3589                 num_bytes = min_t(u64, root->qgroup_meta_rsv_prealloc,
3590                                   num_bytes);
3591                 root->qgroup_meta_rsv_prealloc -= num_bytes;
3592         } else {
3593                 num_bytes = min_t(u64, root->qgroup_meta_rsv_pertrans,
3594                                   num_bytes);
3595                 root->qgroup_meta_rsv_pertrans -= num_bytes;
3596         }
3597         spin_unlock(&root->qgroup_meta_rsv_lock);
3598         return num_bytes;
3599 }
3600
3601 int __btrfs_qgroup_reserve_meta(struct btrfs_root *root, int num_bytes,
3602                                 enum btrfs_qgroup_rsv_type type, bool enforce)
3603 {
3604         struct btrfs_fs_info *fs_info = root->fs_info;
3605         int ret;
3606
3607         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3608             !is_fstree(root->root_key.objectid) || num_bytes == 0)
3609                 return 0;
3610
3611         BUG_ON(num_bytes != round_down(num_bytes, fs_info->nodesize));
3612         trace_qgroup_meta_reserve(root, type, (s64)num_bytes);
3613         ret = qgroup_reserve(root, num_bytes, enforce, type);
3614         if (ret < 0)
3615                 return ret;
3616         /*
3617          * Record what we have reserved into root.
3618          *
3619          * To avoid quota disabled->enabled underflow.
3620          * In that case, we may try to free space we haven't reserved
3621          * (since quota was disabled), so record what we reserved into root.
3622          * And ensure later release won't underflow this number.
3623          */
3624         add_root_meta_rsv(root, num_bytes, type);
3625         return ret;
3626 }
3627
3628 void btrfs_qgroup_free_meta_all_pertrans(struct btrfs_root *root)
3629 {
3630         struct btrfs_fs_info *fs_info = root->fs_info;
3631
3632         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3633             !is_fstree(root->root_key.objectid))
3634                 return;
3635
3636         /* TODO: Update trace point to handle such free */
3637         trace_qgroup_meta_free_all_pertrans(root);
3638         /* Special value -1 means to free all reserved space */
3639         btrfs_qgroup_free_refroot(fs_info, root->root_key.objectid, (u64)-1,
3640                                   BTRFS_QGROUP_RSV_META_PERTRANS);
3641 }
3642
3643 void __btrfs_qgroup_free_meta(struct btrfs_root *root, int num_bytes,
3644                               enum btrfs_qgroup_rsv_type type)
3645 {
3646         struct btrfs_fs_info *fs_info = root->fs_info;
3647
3648         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3649             !is_fstree(root->root_key.objectid))
3650                 return;
3651
3652         /*
3653          * reservation for META_PREALLOC can happen before quota is enabled,
3654          * which can lead to underflow.
3655          * Here ensure we will only free what we really have reserved.
3656          */
3657         num_bytes = sub_root_meta_rsv(root, num_bytes, type);
3658         BUG_ON(num_bytes != round_down(num_bytes, fs_info->nodesize));
3659         trace_qgroup_meta_reserve(root, type, -(s64)num_bytes);
3660         btrfs_qgroup_free_refroot(fs_info, root->root_key.objectid,
3661                                   num_bytes, type);
3662 }
3663
3664 static void qgroup_convert_meta(struct btrfs_fs_info *fs_info, u64 ref_root,
3665                                 int num_bytes)
3666 {
3667         struct btrfs_root *quota_root = fs_info->quota_root;
3668         struct btrfs_qgroup *qgroup;
3669         struct ulist_node *unode;
3670         struct ulist_iterator uiter;
3671         int ret = 0;
3672
3673         if (num_bytes == 0)
3674                 return;
3675         if (!quota_root)
3676                 return;
3677
3678         spin_lock(&fs_info->qgroup_lock);
3679         qgroup = find_qgroup_rb(fs_info, ref_root);
3680         if (!qgroup)
3681                 goto out;
3682         ulist_reinit(fs_info->qgroup_ulist);
3683         ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
3684                        qgroup_to_aux(qgroup), GFP_ATOMIC);
3685         if (ret < 0)
3686                 goto out;
3687         ULIST_ITER_INIT(&uiter);
3688         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
3689                 struct btrfs_qgroup *qg;
3690                 struct btrfs_qgroup_list *glist;
3691
3692                 qg = unode_aux_to_qgroup(unode);
3693
3694                 qgroup_rsv_release(fs_info, qg, num_bytes,
3695                                 BTRFS_QGROUP_RSV_META_PREALLOC);
3696                 qgroup_rsv_add(fs_info, qg, num_bytes,
3697                                 BTRFS_QGROUP_RSV_META_PERTRANS);
3698                 list_for_each_entry(glist, &qg->groups, next_group) {
3699                         ret = ulist_add(fs_info->qgroup_ulist,
3700                                         glist->group->qgroupid,
3701                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
3702                         if (ret < 0)
3703                                 goto out;
3704                 }
3705         }
3706 out:
3707         spin_unlock(&fs_info->qgroup_lock);
3708 }
3709
3710 void btrfs_qgroup_convert_reserved_meta(struct btrfs_root *root, int num_bytes)
3711 {
3712         struct btrfs_fs_info *fs_info = root->fs_info;
3713
3714         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3715             !is_fstree(root->root_key.objectid))
3716                 return;
3717         /* Same as btrfs_qgroup_free_meta_prealloc() */
3718         num_bytes = sub_root_meta_rsv(root, num_bytes,
3719                                       BTRFS_QGROUP_RSV_META_PREALLOC);
3720         trace_qgroup_meta_convert(root, num_bytes);
3721         qgroup_convert_meta(fs_info, root->root_key.objectid, num_bytes);
3722 }
3723
3724 /*
3725  * Check qgroup reserved space leaking, normally at destroy inode
3726  * time
3727  */
3728 void btrfs_qgroup_check_reserved_leak(struct inode *inode)
3729 {
3730         struct extent_changeset changeset;
3731         struct ulist_node *unode;
3732         struct ulist_iterator iter;
3733         int ret;
3734
3735         extent_changeset_init(&changeset);
3736         ret = clear_record_extent_bits(&BTRFS_I(inode)->io_tree, 0, (u64)-1,
3737                         EXTENT_QGROUP_RESERVED, &changeset);
3738
3739         WARN_ON(ret < 0);
3740         if (WARN_ON(changeset.bytes_changed)) {
3741                 ULIST_ITER_INIT(&iter);
3742                 while ((unode = ulist_next(&changeset.range_changed, &iter))) {
3743                         btrfs_warn(BTRFS_I(inode)->root->fs_info,
3744                                 "leaking qgroup reserved space, ino: %lu, start: %llu, end: %llu",
3745                                 inode->i_ino, unode->val, unode->aux);
3746                 }
3747                 btrfs_qgroup_free_refroot(BTRFS_I(inode)->root->fs_info,
3748                                 BTRFS_I(inode)->root->root_key.objectid,
3749                                 changeset.bytes_changed, BTRFS_QGROUP_RSV_DATA);
3750
3751         }
3752         extent_changeset_release(&changeset);
3753 }
3754
3755 void btrfs_qgroup_init_swapped_blocks(
3756         struct btrfs_qgroup_swapped_blocks *swapped_blocks)
3757 {
3758         int i;
3759
3760         spin_lock_init(&swapped_blocks->lock);
3761         for (i = 0; i < BTRFS_MAX_LEVEL; i++)
3762                 swapped_blocks->blocks[i] = RB_ROOT;
3763         swapped_blocks->swapped = false;
3764 }
3765
3766 /*
3767  * Delete all swapped blocks record of @root.
3768  * Every record here means we skipped a full subtree scan for qgroup.
3769  *
3770  * Gets called when committing one transaction.
3771  */
3772 void btrfs_qgroup_clean_swapped_blocks(struct btrfs_root *root)
3773 {
3774         struct btrfs_qgroup_swapped_blocks *swapped_blocks;
3775         int i;
3776
3777         swapped_blocks = &root->swapped_blocks;
3778
3779         spin_lock(&swapped_blocks->lock);
3780         if (!swapped_blocks->swapped)
3781                 goto out;
3782         for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
3783                 struct rb_root *cur_root = &swapped_blocks->blocks[i];
3784                 struct btrfs_qgroup_swapped_block *entry;
3785                 struct btrfs_qgroup_swapped_block *next;
3786
3787                 rbtree_postorder_for_each_entry_safe(entry, next, cur_root,
3788                                                      node)
3789                         kfree(entry);
3790                 swapped_blocks->blocks[i] = RB_ROOT;
3791         }
3792         swapped_blocks->swapped = false;
3793 out:
3794         spin_unlock(&swapped_blocks->lock);
3795 }
3796
3797 /*
3798  * Add subtree roots record into @subvol_root.
3799  *
3800  * @subvol_root:        tree root of the subvolume tree get swapped
3801  * @bg:                 block group under balance
3802  * @subvol_parent/slot: pointer to the subtree root in subvolume tree
3803  * @reloc_parent/slot:  pointer to the subtree root in reloc tree
3804  *                      BOTH POINTERS ARE BEFORE TREE SWAP
3805  * @last_snapshot:      last snapshot generation of the subvolume tree
3806  */
3807 int btrfs_qgroup_add_swapped_blocks(struct btrfs_trans_handle *trans,
3808                 struct btrfs_root *subvol_root,
3809                 struct btrfs_block_group_cache *bg,
3810                 struct extent_buffer *subvol_parent, int subvol_slot,
3811                 struct extent_buffer *reloc_parent, int reloc_slot,
3812                 u64 last_snapshot)
3813 {
3814         struct btrfs_fs_info *fs_info = subvol_root->fs_info;
3815         struct btrfs_qgroup_swapped_blocks *blocks = &subvol_root->swapped_blocks;
3816         struct btrfs_qgroup_swapped_block *block;
3817         struct rb_node **cur;
3818         struct rb_node *parent = NULL;
3819         int level = btrfs_header_level(subvol_parent) - 1;
3820         int ret = 0;
3821
3822         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
3823                 return 0;
3824
3825         if (btrfs_node_ptr_generation(subvol_parent, subvol_slot) >
3826             btrfs_node_ptr_generation(reloc_parent, reloc_slot)) {
3827                 btrfs_err_rl(fs_info,
3828                 "%s: bad parameter order, subvol_gen=%llu reloc_gen=%llu",
3829                         __func__,
3830                         btrfs_node_ptr_generation(subvol_parent, subvol_slot),
3831                         btrfs_node_ptr_generation(reloc_parent, reloc_slot));
3832                 return -EUCLEAN;
3833         }
3834
3835         block = kmalloc(sizeof(*block), GFP_NOFS);
3836         if (!block) {
3837                 ret = -ENOMEM;
3838                 goto out;
3839         }
3840
3841         /*
3842          * @reloc_parent/slot is still before swap, while @block is going to
3843          * record the bytenr after swap, so we do the swap here.
3844          */
3845         block->subvol_bytenr = btrfs_node_blockptr(reloc_parent, reloc_slot);
3846         block->subvol_generation = btrfs_node_ptr_generation(reloc_parent,
3847                                                              reloc_slot);
3848         block->reloc_bytenr = btrfs_node_blockptr(subvol_parent, subvol_slot);
3849         block->reloc_generation = btrfs_node_ptr_generation(subvol_parent,
3850                                                             subvol_slot);
3851         block->last_snapshot = last_snapshot;
3852         block->level = level;
3853
3854         /*
3855          * If we have bg == NULL, we're called from btrfs_recover_relocation(),
3856          * no one else can modify tree blocks thus we qgroup will not change
3857          * no matter the value of trace_leaf.
3858          */
3859         if (bg && bg->flags & BTRFS_BLOCK_GROUP_DATA)
3860                 block->trace_leaf = true;
3861         else
3862                 block->trace_leaf = false;
3863         btrfs_node_key_to_cpu(reloc_parent, &block->first_key, reloc_slot);
3864
3865         /* Insert @block into @blocks */
3866         spin_lock(&blocks->lock);
3867         cur = &blocks->blocks[level].rb_node;
3868         while (*cur) {
3869                 struct btrfs_qgroup_swapped_block *entry;
3870
3871                 parent = *cur;
3872                 entry = rb_entry(parent, struct btrfs_qgroup_swapped_block,
3873                                  node);
3874
3875                 if (entry->subvol_bytenr < block->subvol_bytenr) {
3876                         cur = &(*cur)->rb_left;
3877                 } else if (entry->subvol_bytenr > block->subvol_bytenr) {
3878                         cur = &(*cur)->rb_right;
3879                 } else {
3880                         if (entry->subvol_generation !=
3881                                         block->subvol_generation ||
3882                             entry->reloc_bytenr != block->reloc_bytenr ||
3883                             entry->reloc_generation !=
3884                                         block->reloc_generation) {
3885                                 /*
3886                                  * Duplicated but mismatch entry found.
3887                                  * Shouldn't happen.
3888                                  *
3889                                  * Marking qgroup inconsistent should be enough
3890                                  * for end users.
3891                                  */
3892                                 WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
3893                                 ret = -EEXIST;
3894                         }
3895                         kfree(block);
3896                         goto out_unlock;
3897                 }
3898         }
3899         rb_link_node(&block->node, parent, cur);
3900         rb_insert_color(&block->node, &blocks->blocks[level]);
3901         blocks->swapped = true;
3902 out_unlock:
3903         spin_unlock(&blocks->lock);
3904 out:
3905         if (ret < 0)
3906                 fs_info->qgroup_flags |=
3907                         BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
3908         return ret;
3909 }
3910
3911 /*
3912  * Check if the tree block is a subtree root, and if so do the needed
3913  * delayed subtree trace for qgroup.
3914  *
3915  * This is called during btrfs_cow_block().
3916  */
3917 int btrfs_qgroup_trace_subtree_after_cow(struct btrfs_trans_handle *trans,
3918                                          struct btrfs_root *root,
3919                                          struct extent_buffer *subvol_eb)
3920 {
3921         struct btrfs_fs_info *fs_info = root->fs_info;
3922         struct btrfs_qgroup_swapped_blocks *blocks = &root->swapped_blocks;
3923         struct btrfs_qgroup_swapped_block *block;
3924         struct extent_buffer *reloc_eb = NULL;
3925         struct rb_node *node;
3926         bool found = false;
3927         bool swapped = false;
3928         int level = btrfs_header_level(subvol_eb);
3929         int ret = 0;
3930         int i;
3931
3932         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
3933                 return 0;
3934         if (!is_fstree(root->root_key.objectid) || !root->reloc_root)
3935                 return 0;
3936
3937         spin_lock(&blocks->lock);
3938         if (!blocks->swapped) {
3939                 spin_unlock(&blocks->lock);
3940                 return 0;
3941         }
3942         node = blocks->blocks[level].rb_node;
3943
3944         while (node) {
3945                 block = rb_entry(node, struct btrfs_qgroup_swapped_block, node);
3946                 if (block->subvol_bytenr < subvol_eb->start) {
3947                         node = node->rb_left;
3948                 } else if (block->subvol_bytenr > subvol_eb->start) {
3949                         node = node->rb_right;
3950                 } else {
3951                         found = true;
3952                         break;
3953                 }
3954         }
3955         if (!found) {
3956                 spin_unlock(&blocks->lock);
3957                 goto out;
3958         }
3959         /* Found one, remove it from @blocks first and update blocks->swapped */
3960         rb_erase(&block->node, &blocks->blocks[level]);
3961         for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
3962                 if (RB_EMPTY_ROOT(&blocks->blocks[i])) {
3963                         swapped = true;
3964                         break;
3965                 }
3966         }
3967         blocks->swapped = swapped;
3968         spin_unlock(&blocks->lock);
3969
3970         /* Read out reloc subtree root */
3971         reloc_eb = read_tree_block(fs_info, block->reloc_bytenr,
3972                                    block->reloc_generation, block->level,
3973                                    &block->first_key);
3974         if (IS_ERR(reloc_eb)) {
3975                 ret = PTR_ERR(reloc_eb);
3976                 reloc_eb = NULL;
3977                 goto free_out;
3978         }
3979         if (!extent_buffer_uptodate(reloc_eb)) {
3980                 ret = -EIO;
3981                 goto free_out;
3982         }
3983
3984         ret = qgroup_trace_subtree_swap(trans, reloc_eb, subvol_eb,
3985                         block->last_snapshot, block->trace_leaf);
3986 free_out:
3987         kfree(block);
3988         free_extent_buffer(reloc_eb);
3989 out:
3990         if (ret < 0) {
3991                 btrfs_err_rl(fs_info,
3992                              "failed to account subtree at bytenr %llu: %d",
3993                              subvol_eb->start, ret);
3994                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
3995         }
3996         return ret;
3997 }