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/sched/mm.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 #include "block-group.h"
  25 #include "sysfs.h"
  26 #include "tree-mod-log.h"
  27
  28 /* TODO XXX FIXME
  29  *  - subvol delete -> delete when ref goes to 0? delete limits also?
  30  *  - reorganize keys
  31  *  - compressed
  32  *  - sync
  33  *  - copy also limits on subvol creation
  34  *  - limit
  35  *  - caches for ulists
  36  *  - performance benchmarks
  37  *  - check all ioctl parameters
  38  */
  39
  40 /*
  41  * Helpers to access qgroup reservation
  42  *
  43  * Callers should ensure the lock context and type are valid
  44  */
  45
  46 static u64 qgroup_rsv_total(const struct btrfs_qgroup *qgroup)
  47 {
  48         u64 ret = 0;
  49         int i;
  50
  51         for (i = 0; i < BTRFS_QGROUP_RSV_LAST; i++)
  52                 ret += qgroup->rsv.values[i];
  53
  54         return ret;
  55 }
  56
  57 #ifdef CONFIG_BTRFS_DEBUG
  58 static const char *qgroup_rsv_type_str(enum btrfs_qgroup_rsv_type type)
  59 {
  60         if (type == BTRFS_QGROUP_RSV_DATA)
  61                 return "data";
  62         if (type == BTRFS_QGROUP_RSV_META_PERTRANS)
  63                 return "meta_pertrans";
  64         if (type == BTRFS_QGROUP_RSV_META_PREALLOC)
  65                 return "meta_prealloc";
  66         return NULL;
  67 }
  68 #endif
  69
  70 static void qgroup_rsv_add(struct btrfs_fs_info *fs_info,
  71                            struct btrfs_qgroup *qgroup, u64 num_bytes,
  72                            enum btrfs_qgroup_rsv_type type)
  73 {
  74         trace_qgroup_update_reserve(fs_info, qgroup, num_bytes, type);
  75         qgroup->rsv.values[type] += num_bytes;
  76 }
  77
  78 static void qgroup_rsv_release(struct btrfs_fs_info *fs_info,
  79                                struct btrfs_qgroup *qgroup, u64 num_bytes,
  80                                enum btrfs_qgroup_rsv_type type)
  81 {
  82         trace_qgroup_update_reserve(fs_info, qgroup, -(s64)num_bytes, type);
  83         if (qgroup->rsv.values[type] >= num_bytes) {
  84                 qgroup->rsv.values[type] -= num_bytes;
  85                 return;
  86         }
  87 #ifdef CONFIG_BTRFS_DEBUG
  88         WARN_RATELIMIT(1,
  89                 "qgroup %llu %s reserved space underflow, have %llu to free %llu",
  90                 qgroup->qgroupid, qgroup_rsv_type_str(type),
  91                 qgroup->rsv.values[type], num_bytes);
  92 #endif
  93         qgroup->rsv.values[type] = 0;
  94 }
  95
  96 static void qgroup_rsv_add_by_qgroup(struct btrfs_fs_info *fs_info,
  97                                      struct btrfs_qgroup *dest,
  98                                      struct btrfs_qgroup *src)
  99 {
 100         int i;
 101
 102         for (i = 0; i < BTRFS_QGROUP_RSV_LAST; i++)
 103                 qgroup_rsv_add(fs_info, dest, src->rsv.values[i], i);
 104 }
 105
 106 static void qgroup_rsv_release_by_qgroup(struct btrfs_fs_info *fs_info,
 107                                          struct btrfs_qgroup *dest,
 108                                           struct btrfs_qgroup *src)
 109 {
 110         int i;
 111
 112         for (i = 0; i < BTRFS_QGROUP_RSV_LAST; i++)
 113                 qgroup_rsv_release(fs_info, dest, src->rsv.values[i], i);
 114 }
 115
 116 static void btrfs_qgroup_update_old_refcnt(struct btrfs_qgroup *qg, u64 seq,
 117                                            int mod)
 118 {
 119         if (qg->old_refcnt < seq)
 120                 qg->old_refcnt = seq;
 121         qg->old_refcnt += mod;
 122 }
 123
 124 static void btrfs_qgroup_update_new_refcnt(struct btrfs_qgroup *qg, u64 seq,
 125                                            int mod)
 126 {
 127         if (qg->new_refcnt < seq)
 128                 qg->new_refcnt = seq;
 129         qg->new_refcnt += mod;
 130 }
 131
 132 static inline u64 btrfs_qgroup_get_old_refcnt(struct btrfs_qgroup *qg, u64 seq)
 133 {
 134         if (qg->old_refcnt < seq)
 135                 return 0;
 136         return qg->old_refcnt - seq;
 137 }
 138
 139 static inline u64 btrfs_qgroup_get_new_refcnt(struct btrfs_qgroup *qg, u64 seq)
 140 {
 141         if (qg->new_refcnt < seq)
 142                 return 0;
 143         return qg->new_refcnt - seq;
 144 }
 145
 146 /*
 147  * glue structure to represent the relations between qgroups.
 148  */
 149 struct btrfs_qgroup_list {
 150         struct list_head next_group;
 151         struct list_head next_member;
 152         struct btrfs_qgroup *group;
 153         struct btrfs_qgroup *member;
 154 };
 155
 156 static inline u64 qgroup_to_aux(struct btrfs_qgroup *qg)
 157 {
 158         return (u64)(uintptr_t)qg;
 159 }
 160
 161 static inline struct btrfs_qgroup* unode_aux_to_qgroup(struct ulist_node *n)
 162 {
 163         return (struct btrfs_qgroup *)(uintptr_t)n->aux;
 164 }
 165
 166 static int
 167 qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid,
 168                    int init_flags);
 169 static void qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info);
 170
 171 /* must be called with qgroup_ioctl_lock held */
 172 static struct btrfs_qgroup *find_qgroup_rb(struct btrfs_fs_info *fs_info,
 173                                            u64 qgroupid)
 174 {
 175         struct rb_node *n = fs_info->qgroup_tree.rb_node;
 176         struct btrfs_qgroup *qgroup;
 177
 178         while (n) {
 179                 qgroup = rb_entry(n, struct btrfs_qgroup, node);
 180                 if (qgroup->qgroupid < qgroupid)
 181                         n = n->rb_left;
 182                 else if (qgroup->qgroupid > qgroupid)
 183                         n = n->rb_right;
 184                 else
 185                         return qgroup;
 186         }
 187         return NULL;
 188 }
 189
 190 /* must be called with qgroup_lock held */
 191 static struct btrfs_qgroup *add_qgroup_rb(struct btrfs_fs_info *fs_info,
 192                                           u64 qgroupid)
 193 {
 194         struct rb_node **p = &fs_info->qgroup_tree.rb_node;
 195         struct rb_node *parent = NULL;
 196         struct btrfs_qgroup *qgroup;
 197
 198         while (*p) {
 199                 parent = *p;
 200                 qgroup = rb_entry(parent, struct btrfs_qgroup, node);
 201
 202                 if (qgroup->qgroupid < qgroupid)
 203                         p = &(*p)->rb_left;
 204                 else if (qgroup->qgroupid > qgroupid)
 205                         p = &(*p)->rb_right;
 206                 else
 207                         return qgroup;
 208         }
 209
 210         qgroup = kzalloc(sizeof(*qgroup), GFP_ATOMIC);
 211         if (!qgroup)
 212                 return ERR_PTR(-ENOMEM);
 213
 214         qgroup->qgroupid = qgroupid;
 215         INIT_LIST_HEAD(&qgroup->groups);
 216         INIT_LIST_HEAD(&qgroup->members);
 217         INIT_LIST_HEAD(&qgroup->dirty);
 218
 219         rb_link_node(&qgroup->node, parent, p);
 220         rb_insert_color(&qgroup->node, &fs_info->qgroup_tree);
 221
 222         return qgroup;
 223 }
 224
 225 static void __del_qgroup_rb(struct btrfs_fs_info *fs_info,
 226                             struct btrfs_qgroup *qgroup)
 227 {
 228         struct btrfs_qgroup_list *list;
 229
 230         list_del(&qgroup->dirty);
 231         while (!list_empty(&qgroup->groups)) {
 232                 list = list_first_entry(&qgroup->groups,
 233                                         struct btrfs_qgroup_list, next_group);
 234                 list_del(&list->next_group);
 235                 list_del(&list->next_member);
 236                 kfree(list);
 237         }
 238
 239         while (!list_empty(&qgroup->members)) {
 240                 list = list_first_entry(&qgroup->members,
 241                                         struct btrfs_qgroup_list, next_member);
 242                 list_del(&list->next_group);
 243                 list_del(&list->next_member);
 244                 kfree(list);
 245         }
 246 }
 247
 248 /* must be called with qgroup_lock held */
 249 static int del_qgroup_rb(struct btrfs_fs_info *fs_info, u64 qgroupid)
 250 {
 251         struct btrfs_qgroup *qgroup = find_qgroup_rb(fs_info, qgroupid);
 252
 253         if (!qgroup)
 254                 return -ENOENT;
 255
 256         rb_erase(&qgroup->node, &fs_info->qgroup_tree);
 257         __del_qgroup_rb(fs_info, qgroup);
 258         return 0;
 259 }
 260
 261 /* must be called with qgroup_lock held */
 262 static int add_relation_rb(struct btrfs_fs_info *fs_info,
 263                            u64 memberid, u64 parentid)
 264 {
 265         struct btrfs_qgroup *member;
 266         struct btrfs_qgroup *parent;
 267         struct btrfs_qgroup_list *list;
 268
 269         member = find_qgroup_rb(fs_info, memberid);
 270         parent = find_qgroup_rb(fs_info, parentid);
 271         if (!member || !parent)
 272                 return -ENOENT;
 273
 274         list = kzalloc(sizeof(*list), GFP_ATOMIC);
 275         if (!list)
 276                 return -ENOMEM;
 277
 278         list->group = parent;
 279         list->member = member;
 280         list_add_tail(&list->next_group, &member->groups);
 281         list_add_tail(&list->next_member, &parent->members);
 282
 283         return 0;
 284 }
 285
 286 /* must be called with qgroup_lock held */
 287 static int del_relation_rb(struct btrfs_fs_info *fs_info,
 288                            u64 memberid, u64 parentid)
 289 {
 290         struct btrfs_qgroup *member;
 291         struct btrfs_qgroup *parent;
 292         struct btrfs_qgroup_list *list;
 293
 294         member = find_qgroup_rb(fs_info, memberid);
 295         parent = find_qgroup_rb(fs_info, parentid);
 296         if (!member || !parent)
 297                 return -ENOENT;
 298
 299         list_for_each_entry(list, &member->groups, next_group) {
 300                 if (list->group == parent) {
 301                         list_del(&list->next_group);
 302                         list_del(&list->next_member);
 303                         kfree(list);
 304                         return 0;
 305                 }
 306         }
 307         return -ENOENT;
 308 }
 309
 310 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
 311 int btrfs_verify_qgroup_counts(struct btrfs_fs_info *fs_info, u64 qgroupid,
 312                                u64 rfer, u64 excl)
 313 {
 314         struct btrfs_qgroup *qgroup;
 315
 316         qgroup = find_qgroup_rb(fs_info, qgroupid);
 317         if (!qgroup)
 318                 return -EINVAL;
 319         if (qgroup->rfer != rfer || qgroup->excl != excl)
 320                 return -EINVAL;
 321         return 0;
 322 }
 323 #endif
 324
 325 /*
 326  * The full config is read in one go, only called from open_ctree()
 327  * It doesn't use any locking, as at this point we're still single-threaded
 328  */
 329 int btrfs_read_qgroup_config(struct btrfs_fs_info *fs_info)
 330 {
 331         struct btrfs_key key;
 332         struct btrfs_key found_key;
 333         struct btrfs_root *quota_root = fs_info->quota_root;
 334         struct btrfs_path *path = NULL;
 335         struct extent_buffer *l;
 336         int slot;
 337         int ret = 0;
 338         u64 flags = 0;
 339         u64 rescan_progress = 0;
 340
 341         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
 342                 return 0;
 343
 344         fs_info->qgroup_ulist = ulist_alloc(GFP_KERNEL);
 345         if (!fs_info->qgroup_ulist) {
 346                 ret = -ENOMEM;
 347                 goto out;
 348         }
 349
 350         path = btrfs_alloc_path();
 351         if (!path) {
 352                 ret = -ENOMEM;
 353                 goto out;
 354         }
 355
 356         ret = btrfs_sysfs_add_qgroups(fs_info);
 357         if (ret < 0)
 358                 goto out;
 359         /* default this to quota off, in case no status key is found */
 360         fs_info->qgroup_flags = 0;
 361
 362         /*
 363          * pass 1: read status, all qgroup infos and limits
 364          */
 365         key.objectid = 0;
 366         key.type = 0;
 367         key.offset = 0;
 368         ret = btrfs_search_slot_for_read(quota_root, &key, path, 1, 1);
 369         if (ret)
 370                 goto out;
 371
 372         while (1) {
 373                 struct btrfs_qgroup *qgroup;
 374
 375                 slot = path->slots[0];
 376                 l = path->nodes[0];
 377                 btrfs_item_key_to_cpu(l, &found_key, slot);
 378
 379                 if (found_key.type == BTRFS_QGROUP_STATUS_KEY) {
 380                         struct btrfs_qgroup_status_item *ptr;
 381
 382                         ptr = btrfs_item_ptr(l, slot,
 383                                              struct btrfs_qgroup_status_item);
 384
 385                         if (btrfs_qgroup_status_version(l, ptr) !=
 386                             BTRFS_QGROUP_STATUS_VERSION) {
 387                                 btrfs_err(fs_info,
 388                                  "old qgroup version, quota disabled");
 389                                 goto out;
 390                         }
 391                         if (btrfs_qgroup_status_generation(l, ptr) !=
 392                             fs_info->generation) {
 393                                 flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
 394                                 btrfs_err(fs_info,
 395                                         "qgroup generation mismatch, marked as inconsistent");
 396                         }
 397                         fs_info->qgroup_flags = btrfs_qgroup_status_flags(l,
 398                                                                           ptr);
 399                         rescan_progress = btrfs_qgroup_status_rescan(l, ptr);
 400                         goto next1;
 401                 }
 402
 403                 if (found_key.type != BTRFS_QGROUP_INFO_KEY &&
 404                     found_key.type != BTRFS_QGROUP_LIMIT_KEY)
 405                         goto next1;
 406
 407                 qgroup = find_qgroup_rb(fs_info, found_key.offset);
 408                 if ((qgroup && found_key.type == BTRFS_QGROUP_INFO_KEY) ||
 409                     (!qgroup && found_key.type == BTRFS_QGROUP_LIMIT_KEY)) {
 410                         btrfs_err(fs_info, "inconsistent qgroup config");
 411                         flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
 412                 }
 413                 if (!qgroup) {
 414                         qgroup = add_qgroup_rb(fs_info, found_key.offset);
 415                         if (IS_ERR(qgroup)) {
 416                                 ret = PTR_ERR(qgroup);
 417                                 goto out;
 418                         }
 419                 }
 420                 ret = btrfs_sysfs_add_one_qgroup(fs_info, qgroup);
 421                 if (ret < 0)
 422                         goto out;
 423
 424                 switch (found_key.type) {
 425                 case BTRFS_QGROUP_INFO_KEY: {
 426                         struct btrfs_qgroup_info_item *ptr;
 427
 428                         ptr = btrfs_item_ptr(l, slot,
 429                                              struct btrfs_qgroup_info_item);
 430                         qgroup->rfer = btrfs_qgroup_info_rfer(l, ptr);
 431                         qgroup->rfer_cmpr = btrfs_qgroup_info_rfer_cmpr(l, ptr);
 432                         qgroup->excl = btrfs_qgroup_info_excl(l, ptr);
 433                         qgroup->excl_cmpr = btrfs_qgroup_info_excl_cmpr(l, ptr);
 434                         /* generation currently unused */
 435                         break;
 436                 }
 437                 case BTRFS_QGROUP_LIMIT_KEY: {
 438                         struct btrfs_qgroup_limit_item *ptr;
 439
 440                         ptr = btrfs_item_ptr(l, slot,
 441                                              struct btrfs_qgroup_limit_item);
 442                         qgroup->lim_flags = btrfs_qgroup_limit_flags(l, ptr);
 443                         qgroup->max_rfer = btrfs_qgroup_limit_max_rfer(l, ptr);
 444                         qgroup->max_excl = btrfs_qgroup_limit_max_excl(l, ptr);
 445                         qgroup->rsv_rfer = btrfs_qgroup_limit_rsv_rfer(l, ptr);
 446                         qgroup->rsv_excl = btrfs_qgroup_limit_rsv_excl(l, ptr);
 447                         break;
 448                 }
 449                 }
 450 next1:
 451                 ret = btrfs_next_item(quota_root, path);
 452                 if (ret < 0)
 453                         goto out;
 454                 if (ret)
 455                         break;
 456         }
 457         btrfs_release_path(path);
 458
 459         /*
 460          * pass 2: read all qgroup relations
 461          */
 462         key.objectid = 0;
 463         key.type = BTRFS_QGROUP_RELATION_KEY;
 464         key.offset = 0;
 465         ret = btrfs_search_slot_for_read(quota_root, &key, path, 1, 0);
 466         if (ret)
 467                 goto out;
 468         while (1) {
 469                 slot = path->slots[0];
 470                 l = path->nodes[0];
 471                 btrfs_item_key_to_cpu(l, &found_key, slot);
 472
 473                 if (found_key.type != BTRFS_QGROUP_RELATION_KEY)
 474                         goto next2;
 475
 476                 if (found_key.objectid > found_key.offset) {
 477                         /* parent <- member, not needed to build config */
 478                         /* FIXME should we omit the key completely? */
 479                         goto next2;
 480                 }
 481
 482                 ret = add_relation_rb(fs_info, found_key.objectid,
 483                                       found_key.offset);
 484                 if (ret == -ENOENT) {
 485                         btrfs_warn(fs_info,
 486                                 "orphan qgroup relation 0x%llx->0x%llx",
 487                                 found_key.objectid, found_key.offset);
 488                         ret = 0;        /* ignore the error */
 489                 }
 490                 if (ret)
 491                         goto out;
 492 next2:
 493                 ret = btrfs_next_item(quota_root, path);
 494                 if (ret < 0)
 495                         goto out;
 496                 if (ret)
 497                         break;
 498         }
 499 out:
 500         btrfs_free_path(path);
 501         fs_info->qgroup_flags |= flags;
 502         if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON))
 503                 clear_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
 504         else if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN &&
 505                  ret >= 0)
 506                 ret = qgroup_rescan_init(fs_info, rescan_progress, 0);
 507
 508         if (ret < 0) {
 509                 ulist_free(fs_info->qgroup_ulist);
 510                 fs_info->qgroup_ulist = NULL;
 511                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
 512                 btrfs_sysfs_del_qgroups(fs_info);
 513         }
 514
 515         return ret < 0 ? ret : 0;
 516 }
 517
 518 /*
 519  * Called in close_ctree() when quota is still enabled.  This verifies we don't
 520  * leak some reserved space.
 521  *
 522  * Return false if no reserved space is left.
 523  * Return true if some reserved space is leaked.
 524  */
 525 bool btrfs_check_quota_leak(struct btrfs_fs_info *fs_info)
 526 {
 527         struct rb_node *node;
 528         bool ret = false;
 529
 530         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
 531                 return ret;
 532         /*
 533          * Since we're unmounting, there is no race and no need to grab qgroup
 534          * lock.  And here we don't go post-order to provide a more user
 535          * friendly sorted result.
 536          */
 537         for (node = rb_first(&fs_info->qgroup_tree); node; node = rb_next(node)) {
 538                 struct btrfs_qgroup *qgroup;
 539                 int i;
 540
 541                 qgroup = rb_entry(node, struct btrfs_qgroup, node);
 542                 for (i = 0; i < BTRFS_QGROUP_RSV_LAST; i++) {
 543                         if (qgroup->rsv.values[i]) {
 544                                 ret = true;
 545                                 btrfs_warn(fs_info,
 546                 "qgroup %hu/%llu has unreleased space, type %d rsv %llu",
 547                                    btrfs_qgroup_level(qgroup->qgroupid),
 548                                    btrfs_qgroup_subvolid(qgroup->qgroupid),
 549                                    i, qgroup->rsv.values[i]);
 550                         }
 551                 }
 552         }
 553         return ret;
 554 }
 555
 556 /*
 557  * This is called from close_ctree() or open_ctree() or btrfs_quota_disable(),
 558  * first two are in single-threaded paths.And for the third one, we have set
 559  * quota_root to be null with qgroup_lock held before, so it is safe to clean
 560  * up the in-memory structures without qgroup_lock held.
 561  */
 562 void btrfs_free_qgroup_config(struct btrfs_fs_info *fs_info)
 563 {
 564         struct rb_node *n;
 565         struct btrfs_qgroup *qgroup;
 566
 567         while ((n = rb_first(&fs_info->qgroup_tree))) {
 568                 qgroup = rb_entry(n, struct btrfs_qgroup, node);
 569                 rb_erase(n, &fs_info->qgroup_tree);
 570                 __del_qgroup_rb(fs_info, qgroup);
 571                 btrfs_sysfs_del_one_qgroup(fs_info, qgroup);
 572                 kfree(qgroup);
 573         }
 574         /*
 575          * We call btrfs_free_qgroup_config() when unmounting
 576          * filesystem and disabling quota, so we set qgroup_ulist
 577          * to be null here to avoid double free.
 578          */
 579         ulist_free(fs_info->qgroup_ulist);
 580         fs_info->qgroup_ulist = NULL;
 581         btrfs_sysfs_del_qgroups(fs_info);
 582 }
 583
 584 static int add_qgroup_relation_item(struct btrfs_trans_handle *trans, u64 src,
 585                                     u64 dst)
 586 {
 587         int ret;
 588         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 589         struct btrfs_path *path;
 590         struct btrfs_key key;
 591
 592         path = btrfs_alloc_path();
 593         if (!path)
 594                 return -ENOMEM;
 595
 596         key.objectid = src;
 597         key.type = BTRFS_QGROUP_RELATION_KEY;
 598         key.offset = dst;
 599
 600         ret = btrfs_insert_empty_item(trans, quota_root, path, &key, 0);
 601
 602         btrfs_mark_buffer_dirty(path->nodes[0]);
 603
 604         btrfs_free_path(path);
 605         return ret;
 606 }
 607
 608 static int del_qgroup_relation_item(struct btrfs_trans_handle *trans, u64 src,
 609                                     u64 dst)
 610 {
 611         int ret;
 612         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 613         struct btrfs_path *path;
 614         struct btrfs_key key;
 615
 616         path = btrfs_alloc_path();
 617         if (!path)
 618                 return -ENOMEM;
 619
 620         key.objectid = src;
 621         key.type = BTRFS_QGROUP_RELATION_KEY;
 622         key.offset = dst;
 623
 624         ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
 625         if (ret < 0)
 626                 goto out;
 627
 628         if (ret > 0) {
 629                 ret = -ENOENT;
 630                 goto out;
 631         }
 632
 633         ret = btrfs_del_item(trans, quota_root, path);
 634 out:
 635         btrfs_free_path(path);
 636         return ret;
 637 }
 638
 639 static int add_qgroup_item(struct btrfs_trans_handle *trans,
 640                            struct btrfs_root *quota_root, u64 qgroupid)
 641 {
 642         int ret;
 643         struct btrfs_path *path;
 644         struct btrfs_qgroup_info_item *qgroup_info;
 645         struct btrfs_qgroup_limit_item *qgroup_limit;
 646         struct extent_buffer *leaf;
 647         struct btrfs_key key;
 648
 649         if (btrfs_is_testing(quota_root->fs_info))
 650                 return 0;
 651
 652         path = btrfs_alloc_path();
 653         if (!path)
 654                 return -ENOMEM;
 655
 656         key.objectid = 0;
 657         key.type = BTRFS_QGROUP_INFO_KEY;
 658         key.offset = qgroupid;
 659
 660         /*
 661          * Avoid a transaction abort by catching -EEXIST here. In that
 662          * case, we proceed by re-initializing the existing structure
 663          * on disk.
 664          */
 665
 666         ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
 667                                       sizeof(*qgroup_info));
 668         if (ret && ret != -EEXIST)
 669                 goto out;
 670
 671         leaf = path->nodes[0];
 672         qgroup_info = btrfs_item_ptr(leaf, path->slots[0],
 673                                  struct btrfs_qgroup_info_item);
 674         btrfs_set_qgroup_info_generation(leaf, qgroup_info, trans->transid);
 675         btrfs_set_qgroup_info_rfer(leaf, qgroup_info, 0);
 676         btrfs_set_qgroup_info_rfer_cmpr(leaf, qgroup_info, 0);
 677         btrfs_set_qgroup_info_excl(leaf, qgroup_info, 0);
 678         btrfs_set_qgroup_info_excl_cmpr(leaf, qgroup_info, 0);
 679
 680         btrfs_mark_buffer_dirty(leaf);
 681
 682         btrfs_release_path(path);
 683
 684         key.type = BTRFS_QGROUP_LIMIT_KEY;
 685         ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
 686                                       sizeof(*qgroup_limit));
 687         if (ret && ret != -EEXIST)
 688                 goto out;
 689
 690         leaf = path->nodes[0];
 691         qgroup_limit = btrfs_item_ptr(leaf, path->slots[0],
 692                                   struct btrfs_qgroup_limit_item);
 693         btrfs_set_qgroup_limit_flags(leaf, qgroup_limit, 0);
 694         btrfs_set_qgroup_limit_max_rfer(leaf, qgroup_limit, 0);
 695         btrfs_set_qgroup_limit_max_excl(leaf, qgroup_limit, 0);
 696         btrfs_set_qgroup_limit_rsv_rfer(leaf, qgroup_limit, 0);
 697         btrfs_set_qgroup_limit_rsv_excl(leaf, qgroup_limit, 0);
 698
 699         btrfs_mark_buffer_dirty(leaf);
 700
 701         ret = 0;
 702 out:
 703         btrfs_free_path(path);
 704         return ret;
 705 }
 706
 707 static int del_qgroup_item(struct btrfs_trans_handle *trans, u64 qgroupid)
 708 {
 709         int ret;
 710         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 711         struct btrfs_path *path;
 712         struct btrfs_key key;
 713
 714         path = btrfs_alloc_path();
 715         if (!path)
 716                 return -ENOMEM;
 717
 718         key.objectid = 0;
 719         key.type = BTRFS_QGROUP_INFO_KEY;
 720         key.offset = qgroupid;
 721         ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
 722         if (ret < 0)
 723                 goto out;
 724
 725         if (ret > 0) {
 726                 ret = -ENOENT;
 727                 goto out;
 728         }
 729
 730         ret = btrfs_del_item(trans, quota_root, path);
 731         if (ret)
 732                 goto out;
 733
 734         btrfs_release_path(path);
 735
 736         key.type = BTRFS_QGROUP_LIMIT_KEY;
 737         ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
 738         if (ret < 0)
 739                 goto out;
 740
 741         if (ret > 0) {
 742                 ret = -ENOENT;
 743                 goto out;
 744         }
 745
 746         ret = btrfs_del_item(trans, quota_root, path);
 747
 748 out:
 749         btrfs_free_path(path);
 750         return ret;
 751 }
 752
 753 static int update_qgroup_limit_item(struct btrfs_trans_handle *trans,
 754                                     struct btrfs_qgroup *qgroup)
 755 {
 756         struct btrfs_root *quota_root = trans->fs_info->quota_root;
 757         struct btrfs_path *path;
 758         struct btrfs_key key;
 759         struct extent_buffer *l;
 760         struct btrfs_qgroup_limit_item *qgroup_limit;
 761         int ret;
 762         int slot;
 763
 764         key.objectid = 0;
 765         key.type = BTRFS_QGROUP_LIMIT_KEY;
 766         key.offset = qgroup->qgroupid;
 767
 768         path = btrfs_alloc_path();
 769         if (!path)
 770                 return -ENOMEM;
 771
 772         ret = btrfs_search_slot(trans, quota_root, &key, path, 0, 1);
 773         if (ret > 0)
 774                 ret = -ENOENT;
 775
 776         if (ret)
 777                 goto out;
 778
 779         l = path->nodes[0];
 780         slot = path->slots[0];
 781         qgroup_limit = btrfs_item_ptr(l, slot, struct btrfs_qgroup_limit_item);
 782         btrfs_set_qgroup_limit_flags(l, qgroup_limit, qgroup->lim_flags);
 783         btrfs_set_qgroup_limit_max_rfer(l, qgroup_limit, qgroup->max_rfer);
 784         btrfs_set_qgroup_limit_max_excl(l, qgroup_limit, qgroup->max_excl);
 785         btrfs_set_qgroup_limit_rsv_rfer(l, qgroup_limit, qgroup->rsv_rfer);
 786         btrfs_set_qgroup_limit_rsv_excl(l, qgroup_limit, qgroup->rsv_excl);
 787
 788         btrfs_mark_buffer_dirty(l);
 789
 790 out:
 791         btrfs_free_path(path);
 792         return ret;
 793 }
 794
 795 static int update_qgroup_info_item(struct btrfs_trans_handle *trans,
 796                                    struct btrfs_qgroup *qgroup)
 797 {
 798         struct btrfs_fs_info *fs_info = trans->fs_info;
 799         struct btrfs_root *quota_root = fs_info->quota_root;
 800         struct btrfs_path *path;
 801         struct btrfs_key key;
 802         struct extent_buffer *l;
 803         struct btrfs_qgroup_info_item *qgroup_info;
 804         int ret;
 805         int slot;
 806
 807         if (btrfs_is_testing(fs_info))
 808                 return 0;
 809
 810         key.objectid = 0;
 811         key.type = BTRFS_QGROUP_INFO_KEY;
 812         key.offset = qgroup->qgroupid;
 813
 814         path = btrfs_alloc_path();
 815         if (!path)
 816                 return -ENOMEM;
 817
 818         ret = btrfs_search_slot(trans, quota_root, &key, path, 0, 1);
 819         if (ret > 0)
 820                 ret = -ENOENT;
 821
 822         if (ret)
 823                 goto out;
 824
 825         l = path->nodes[0];
 826         slot = path->slots[0];
 827         qgroup_info = btrfs_item_ptr(l, slot, struct btrfs_qgroup_info_item);
 828         btrfs_set_qgroup_info_generation(l, qgroup_info, trans->transid);
 829         btrfs_set_qgroup_info_rfer(l, qgroup_info, qgroup->rfer);
 830         btrfs_set_qgroup_info_rfer_cmpr(l, qgroup_info, qgroup->rfer_cmpr);
 831         btrfs_set_qgroup_info_excl(l, qgroup_info, qgroup->excl);
 832         btrfs_set_qgroup_info_excl_cmpr(l, qgroup_info, qgroup->excl_cmpr);
 833
 834         btrfs_mark_buffer_dirty(l);
 835
 836 out:
 837         btrfs_free_path(path);
 838         return ret;
 839 }
 840
 841 static int update_qgroup_status_item(struct btrfs_trans_handle *trans)
 842 {
 843         struct btrfs_fs_info *fs_info = trans->fs_info;
 844         struct btrfs_root *quota_root = fs_info->quota_root;
 845         struct btrfs_path *path;
 846         struct btrfs_key key;
 847         struct extent_buffer *l;
 848         struct btrfs_qgroup_status_item *ptr;
 849         int ret;
 850         int slot;
 851
 852         key.objectid = 0;
 853         key.type = BTRFS_QGROUP_STATUS_KEY;
 854         key.offset = 0;
 855
 856         path = btrfs_alloc_path();
 857         if (!path)
 858                 return -ENOMEM;
 859
 860         ret = btrfs_search_slot(trans, quota_root, &key, path, 0, 1);
 861         if (ret > 0)
 862                 ret = -ENOENT;
 863
 864         if (ret)
 865                 goto out;
 866
 867         l = path->nodes[0];
 868         slot = path->slots[0];
 869         ptr = btrfs_item_ptr(l, slot, struct btrfs_qgroup_status_item);
 870         btrfs_set_qgroup_status_flags(l, ptr, fs_info->qgroup_flags);
 871         btrfs_set_qgroup_status_generation(l, ptr, trans->transid);
 872         btrfs_set_qgroup_status_rescan(l, ptr,
 873                                 fs_info->qgroup_rescan_progress.objectid);
 874
 875         btrfs_mark_buffer_dirty(l);
 876
 877 out:
 878         btrfs_free_path(path);
 879         return ret;
 880 }
 881
 882 /*
 883  * called with qgroup_lock held
 884  */
 885 static int btrfs_clean_quota_tree(struct btrfs_trans_handle *trans,
 886                                   struct btrfs_root *root)
 887 {
 888         struct btrfs_path *path;
 889         struct btrfs_key key;
 890         struct extent_buffer *leaf = NULL;
 891         int ret;
 892         int nr = 0;
 893
 894         path = btrfs_alloc_path();
 895         if (!path)
 896                 return -ENOMEM;
 897
 898         key.objectid = 0;
 899         key.offset = 0;
 900         key.type = 0;
 901
 902         while (1) {
 903                 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 904                 if (ret < 0)
 905                         goto out;
 906                 leaf = path->nodes[0];
 907                 nr = btrfs_header_nritems(leaf);
 908                 if (!nr)
 909                         break;
 910                 /*
 911                  * delete the leaf one by one
 912                  * since the whole tree is going
 913                  * to be deleted.
 914                  */
 915                 path->slots[0] = 0;
 916                 ret = btrfs_del_items(trans, root, path, 0, nr);
 917                 if (ret)
 918                         goto out;
 919
 920                 btrfs_release_path(path);
 921         }
 922         ret = 0;
 923 out:
 924         btrfs_free_path(path);
 925         return ret;
 926 }
 927
 928 int btrfs_quota_enable(struct btrfs_fs_info *fs_info)
 929 {
 930         struct btrfs_root *quota_root;
 931         struct btrfs_root *tree_root = fs_info->tree_root;
 932         struct btrfs_path *path = NULL;
 933         struct btrfs_qgroup_status_item *ptr;
 934         struct extent_buffer *leaf;
 935         struct btrfs_key key;
 936         struct btrfs_key found_key;
 937         struct btrfs_qgroup *qgroup = NULL;
 938         struct btrfs_trans_handle *trans = NULL;
 939         struct ulist *ulist = NULL;
 940         int ret = 0;
 941         int slot;
 942
 943         mutex_lock(&fs_info->qgroup_ioctl_lock);
 944         if (fs_info->quota_root)
 945                 goto out;
 946
 947         ulist = ulist_alloc(GFP_KERNEL);
 948         if (!ulist) {
 949                 ret = -ENOMEM;
 950                 goto out;
 951         }
 952
 953         ret = btrfs_sysfs_add_qgroups(fs_info);
 954         if (ret < 0)
 955                 goto out;
 956
 957         /*
 958          * Unlock qgroup_ioctl_lock before starting the transaction. This is to
 959          * avoid lock acquisition inversion problems (reported by lockdep) between
 960          * qgroup_ioctl_lock and the vfs freeze semaphores, acquired when we
 961          * start a transaction.
 962          * After we started the transaction lock qgroup_ioctl_lock again and
 963          * check if someone else created the quota root in the meanwhile. If so,
 964          * just return success and release the transaction handle.
 965          *
 966          * Also we don't need to worry about someone else calling
 967          * btrfs_sysfs_add_qgroups() after we unlock and getting an error because
 968          * that function returns 0 (success) when the sysfs entries already exist.
 969          */
 970         mutex_unlock(&fs_info->qgroup_ioctl_lock);
 971
 972         /*
 973          * 1 for quota root item
 974          * 1 for BTRFS_QGROUP_STATUS item
 975          *
 976          * Yet we also need 2*n items for a QGROUP_INFO/QGROUP_LIMIT items
 977          * per subvolume. However those are not currently reserved since it
 978          * would be a lot of overkill.
 979          */
 980         trans = btrfs_start_transaction(tree_root, 2);
 981
 982         mutex_lock(&fs_info->qgroup_ioctl_lock);
 983         if (IS_ERR(trans)) {
 984                 ret = PTR_ERR(trans);
 985                 trans = NULL;
 986                 goto out;
 987         }
 988
 989         if (fs_info->quota_root)
 990                 goto out;
 991
 992         fs_info->qgroup_ulist = ulist;
 993         ulist = NULL;
 994
 995         /*
 996          * initially create the quota tree
 997          */
 998         quota_root = btrfs_create_tree(trans, BTRFS_QUOTA_TREE_OBJECTID);
 999         if (IS_ERR(quota_root)) {
1000                 ret =  PTR_ERR(quota_root);
1001                 btrfs_abort_transaction(trans, ret);
1002                 goto out;
1003         }
1004
1005         path = btrfs_alloc_path();
1006         if (!path) {
1007                 ret = -ENOMEM;
1008                 btrfs_abort_transaction(trans, ret);
1009                 goto out_free_root;
1010         }
1011
1012         key.objectid = 0;
1013         key.type = BTRFS_QGROUP_STATUS_KEY;
1014         key.offset = 0;
1015
1016         ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
1017                                       sizeof(*ptr));
1018         if (ret) {
1019                 btrfs_abort_transaction(trans, ret);
1020                 goto out_free_path;
1021         }
1022
1023         leaf = path->nodes[0];
1024         ptr = btrfs_item_ptr(leaf, path->slots[0],
1025                                  struct btrfs_qgroup_status_item);
1026         btrfs_set_qgroup_status_generation(leaf, ptr, trans->transid);
1027         btrfs_set_qgroup_status_version(leaf, ptr, BTRFS_QGROUP_STATUS_VERSION);
1028         fs_info->qgroup_flags = BTRFS_QGROUP_STATUS_FLAG_ON |
1029                                 BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
1030         btrfs_set_qgroup_status_flags(leaf, ptr, fs_info->qgroup_flags);
1031         btrfs_set_qgroup_status_rescan(leaf, ptr, 0);
1032
1033         btrfs_mark_buffer_dirty(leaf);
1034
1035         key.objectid = 0;
1036         key.type = BTRFS_ROOT_REF_KEY;
1037         key.offset = 0;
1038
1039         btrfs_release_path(path);
1040         ret = btrfs_search_slot_for_read(tree_root, &key, path, 1, 0);
1041         if (ret > 0)
1042                 goto out_add_root;
1043         if (ret < 0) {
1044                 btrfs_abort_transaction(trans, ret);
1045                 goto out_free_path;
1046         }
1047
1048         while (1) {
1049                 slot = path->slots[0];
1050                 leaf = path->nodes[0];
1051                 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1052
1053                 if (found_key.type == BTRFS_ROOT_REF_KEY) {
1054
1055                         /* Release locks on tree_root before we access quota_root */
1056                         btrfs_release_path(path);
1057
1058                         ret = add_qgroup_item(trans, quota_root,
1059                                               found_key.offset);
1060                         if (ret) {
1061                                 btrfs_abort_transaction(trans, ret);
1062                                 goto out_free_path;
1063                         }
1064
1065                         qgroup = add_qgroup_rb(fs_info, found_key.offset);
1066                         if (IS_ERR(qgroup)) {
1067                                 ret = PTR_ERR(qgroup);
1068                                 btrfs_abort_transaction(trans, ret);
1069                                 goto out_free_path;
1070                         }
1071                         ret = btrfs_sysfs_add_one_qgroup(fs_info, qgroup);
1072                         if (ret < 0) {
1073                                 btrfs_abort_transaction(trans, ret);
1074                                 goto out_free_path;
1075                         }
1076                         ret = btrfs_search_slot_for_read(tree_root, &found_key,
1077                                                          path, 1, 0);
1078                         if (ret < 0) {
1079                                 btrfs_abort_transaction(trans, ret);
1080                                 goto out_free_path;
1081                         }
1082                         if (ret > 0) {
1083                                 /*
1084                                  * Shouldn't happen, but in case it does we
1085                                  * don't need to do the btrfs_next_item, just
1086                                  * continue.
1087                                  */
1088                                 continue;
1089                         }
1090                 }
1091                 ret = btrfs_next_item(tree_root, path);
1092                 if (ret < 0) {
1093                         btrfs_abort_transaction(trans, ret);
1094                         goto out_free_path;
1095                 }
1096                 if (ret)
1097                         break;
1098         }
1099
1100 out_add_root:
1101         btrfs_release_path(path);
1102         ret = add_qgroup_item(trans, quota_root, BTRFS_FS_TREE_OBJECTID);
1103         if (ret) {
1104                 btrfs_abort_transaction(trans, ret);
1105                 goto out_free_path;
1106         }
1107
1108         qgroup = add_qgroup_rb(fs_info, BTRFS_FS_TREE_OBJECTID);
1109         if (IS_ERR(qgroup)) {
1110                 ret = PTR_ERR(qgroup);
1111                 btrfs_abort_transaction(trans, ret);
1112                 goto out_free_path;
1113         }
1114         ret = btrfs_sysfs_add_one_qgroup(fs_info, qgroup);
1115         if (ret < 0) {
1116                 btrfs_abort_transaction(trans, ret);
1117                 goto out_free_path;
1118         }
1119
1120         ret = btrfs_commit_transaction(trans);
1121         trans = NULL;
1122         if (ret)
1123                 goto out_free_path;
1124
1125         /*
1126          * Set quota enabled flag after committing the transaction, to avoid
1127          * deadlocks on fs_info->qgroup_ioctl_lock with concurrent snapshot
1128          * creation.
1129          */
1130         spin_lock(&fs_info->qgroup_lock);
1131         fs_info->quota_root = quota_root;
1132         set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
1133         spin_unlock(&fs_info->qgroup_lock);
1134
1135         ret = qgroup_rescan_init(fs_info, 0, 1);
1136         if (!ret) {
1137                 qgroup_rescan_zero_tracking(fs_info);
1138                 fs_info->qgroup_rescan_running = true;
1139                 btrfs_queue_work(fs_info->qgroup_rescan_workers,
1140                                  &fs_info->qgroup_rescan_work);
1141         }
1142
1143 out_free_path:
1144         btrfs_free_path(path);
1145 out_free_root:
1146         if (ret)
1147                 btrfs_put_root(quota_root);
1148 out:
1149         if (ret) {
1150                 ulist_free(fs_info->qgroup_ulist);
1151                 fs_info->qgroup_ulist = NULL;
1152                 btrfs_sysfs_del_qgroups(fs_info);
1153         }
1154         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1155         if (ret && trans)
1156                 btrfs_end_transaction(trans);
1157         else if (trans)
1158                 ret = btrfs_end_transaction(trans);
1159         ulist_free(ulist);
1160         return ret;
1161 }
1162
1163 int btrfs_quota_disable(struct btrfs_fs_info *fs_info)
1164 {
1165         struct btrfs_root *quota_root;
1166         struct btrfs_trans_handle *trans = NULL;
1167         int ret = 0;
1168
1169         mutex_lock(&fs_info->qgroup_ioctl_lock);
1170         if (!fs_info->quota_root)
1171                 goto out;
1172         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1173
1174         /*
1175          * 1 For the root item
1176          *
1177          * We should also reserve enough items for the quota tree deletion in
1178          * btrfs_clean_quota_tree but this is not done.
1179          *
1180          * Also, we must always start a transaction without holding the mutex
1181          * qgroup_ioctl_lock, see btrfs_quota_enable().
1182          */
1183         trans = btrfs_start_transaction(fs_info->tree_root, 1);
1184
1185         mutex_lock(&fs_info->qgroup_ioctl_lock);
1186         if (IS_ERR(trans)) {
1187                 ret = PTR_ERR(trans);
1188                 trans = NULL;
1189                 goto out;
1190         }
1191
1192         if (!fs_info->quota_root)
1193                 goto out;
1194
1195         clear_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
1196         btrfs_qgroup_wait_for_completion(fs_info, false);
1197         spin_lock(&fs_info->qgroup_lock);
1198         quota_root = fs_info->quota_root;
1199         fs_info->quota_root = NULL;
1200         fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_ON;
1201         spin_unlock(&fs_info->qgroup_lock);
1202
1203         btrfs_free_qgroup_config(fs_info);
1204
1205         ret = btrfs_clean_quota_tree(trans, quota_root);
1206         if (ret) {
1207                 btrfs_abort_transaction(trans, ret);
1208                 goto out;
1209         }
1210
1211         ret = btrfs_del_root(trans, &quota_root->root_key);
1212         if (ret) {
1213                 btrfs_abort_transaction(trans, ret);
1214                 goto out;
1215         }
1216
1217         list_del(&quota_root->dirty_list);
1218
1219         btrfs_tree_lock(quota_root->node);
1220         btrfs_clean_tree_block(quota_root->node);
1221         btrfs_tree_unlock(quota_root->node);
1222         btrfs_free_tree_block(trans, quota_root, quota_root->node, 0, 1);
1223
1224         btrfs_put_root(quota_root);
1225
1226 out:
1227         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1228         if (ret && trans)
1229                 btrfs_end_transaction(trans);
1230         else if (trans)
1231                 ret = btrfs_end_transaction(trans);
1232
1233         return ret;
1234 }
1235
1236 static void qgroup_dirty(struct btrfs_fs_info *fs_info,
1237                          struct btrfs_qgroup *qgroup)
1238 {
1239         if (list_empty(&qgroup->dirty))
1240                 list_add(&qgroup->dirty, &fs_info->dirty_qgroups);
1241 }
1242
1243 /*
1244  * The easy accounting, we're updating qgroup relationship whose child qgroup
1245  * only has exclusive extents.
1246  *
1247  * In this case, all exclusive extents will also be exclusive for parent, so
1248  * excl/rfer just get added/removed.
1249  *
1250  * So is qgroup reservation space, which should also be added/removed to
1251  * parent.
1252  * Or when child tries to release reservation space, parent will underflow its
1253  * reservation (for relationship adding case).
1254  *
1255  * Caller should hold fs_info->qgroup_lock.
1256  */
1257 static int __qgroup_excl_accounting(struct btrfs_fs_info *fs_info,
1258                                     struct ulist *tmp, u64 ref_root,
1259                                     struct btrfs_qgroup *src, int sign)
1260 {
1261         struct btrfs_qgroup *qgroup;
1262         struct btrfs_qgroup_list *glist;
1263         struct ulist_node *unode;
1264         struct ulist_iterator uiter;
1265         u64 num_bytes = src->excl;
1266         int ret = 0;
1267
1268         qgroup = find_qgroup_rb(fs_info, ref_root);
1269         if (!qgroup)
1270                 goto out;
1271
1272         qgroup->rfer += sign * num_bytes;
1273         qgroup->rfer_cmpr += sign * num_bytes;
1274
1275         WARN_ON(sign < 0 && qgroup->excl < num_bytes);
1276         qgroup->excl += sign * num_bytes;
1277         qgroup->excl_cmpr += sign * num_bytes;
1278
1279         if (sign > 0)
1280                 qgroup_rsv_add_by_qgroup(fs_info, qgroup, src);
1281         else
1282                 qgroup_rsv_release_by_qgroup(fs_info, qgroup, src);
1283
1284         qgroup_dirty(fs_info, qgroup);
1285
1286         /* Get all of the parent groups that contain this qgroup */
1287         list_for_each_entry(glist, &qgroup->groups, next_group) {
1288                 ret = ulist_add(tmp, glist->group->qgroupid,
1289                                 qgroup_to_aux(glist->group), GFP_ATOMIC);
1290                 if (ret < 0)
1291                         goto out;
1292         }
1293
1294         /* Iterate all of the parents and adjust their reference counts */
1295         ULIST_ITER_INIT(&uiter);
1296         while ((unode = ulist_next(tmp, &uiter))) {
1297                 qgroup = unode_aux_to_qgroup(unode);
1298                 qgroup->rfer += sign * num_bytes;
1299                 qgroup->rfer_cmpr += sign * num_bytes;
1300                 WARN_ON(sign < 0 && qgroup->excl < num_bytes);
1301                 qgroup->excl += sign * num_bytes;
1302                 if (sign > 0)
1303                         qgroup_rsv_add_by_qgroup(fs_info, qgroup, src);
1304                 else
1305                         qgroup_rsv_release_by_qgroup(fs_info, qgroup, src);
1306                 qgroup->excl_cmpr += sign * num_bytes;
1307                 qgroup_dirty(fs_info, qgroup);
1308
1309                 /* Add any parents of the parents */
1310                 list_for_each_entry(glist, &qgroup->groups, next_group) {
1311                         ret = ulist_add(tmp, glist->group->qgroupid,
1312                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
1313                         if (ret < 0)
1314                                 goto out;
1315                 }
1316         }
1317         ret = 0;
1318 out:
1319         return ret;
1320 }
1321
1322
1323 /*
1324  * Quick path for updating qgroup with only excl refs.
1325  *
1326  * In that case, just update all parent will be enough.
1327  * Or we needs to do a full rescan.
1328  * Caller should also hold fs_info->qgroup_lock.
1329  *
1330  * Return 0 for quick update, return >0 for need to full rescan
1331  * and mark INCONSISTENT flag.
1332  * Return < 0 for other error.
1333  */
1334 static int quick_update_accounting(struct btrfs_fs_info *fs_info,
1335                                    struct ulist *tmp, u64 src, u64 dst,
1336                                    int sign)
1337 {
1338         struct btrfs_qgroup *qgroup;
1339         int ret = 1;
1340         int err = 0;
1341
1342         qgroup = find_qgroup_rb(fs_info, src);
1343         if (!qgroup)
1344                 goto out;
1345         if (qgroup->excl == qgroup->rfer) {
1346                 ret = 0;
1347                 err = __qgroup_excl_accounting(fs_info, tmp, dst,
1348                                                qgroup, sign);
1349                 if (err < 0) {
1350                         ret = err;
1351                         goto out;
1352                 }
1353         }
1354 out:
1355         if (ret)
1356                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
1357         return ret;
1358 }
1359
1360 int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
1361                               u64 dst)
1362 {
1363         struct btrfs_fs_info *fs_info = trans->fs_info;
1364         struct btrfs_qgroup *parent;
1365         struct btrfs_qgroup *member;
1366         struct btrfs_qgroup_list *list;
1367         struct ulist *tmp;
1368         unsigned int nofs_flag;
1369         int ret = 0;
1370
1371         /* Check the level of src and dst first */
1372         if (btrfs_qgroup_level(src) >= btrfs_qgroup_level(dst))
1373                 return -EINVAL;
1374
1375         /* We hold a transaction handle open, must do a NOFS allocation. */
1376         nofs_flag = memalloc_nofs_save();
1377         tmp = ulist_alloc(GFP_KERNEL);
1378         memalloc_nofs_restore(nofs_flag);
1379         if (!tmp)
1380                 return -ENOMEM;
1381
1382         mutex_lock(&fs_info->qgroup_ioctl_lock);
1383         if (!fs_info->quota_root) {
1384                 ret = -ENOTCONN;
1385                 goto out;
1386         }
1387         member = find_qgroup_rb(fs_info, src);
1388         parent = find_qgroup_rb(fs_info, dst);
1389         if (!member || !parent) {
1390                 ret = -EINVAL;
1391                 goto out;
1392         }
1393
1394         /* check if such qgroup relation exist firstly */
1395         list_for_each_entry(list, &member->groups, next_group) {
1396                 if (list->group == parent) {
1397                         ret = -EEXIST;
1398                         goto out;
1399                 }
1400         }
1401
1402         ret = add_qgroup_relation_item(trans, src, dst);
1403         if (ret)
1404                 goto out;
1405
1406         ret = add_qgroup_relation_item(trans, dst, src);
1407         if (ret) {
1408                 del_qgroup_relation_item(trans, src, dst);
1409                 goto out;
1410         }
1411
1412         spin_lock(&fs_info->qgroup_lock);
1413         ret = add_relation_rb(fs_info, src, dst);
1414         if (ret < 0) {
1415                 spin_unlock(&fs_info->qgroup_lock);
1416                 goto out;
1417         }
1418         ret = quick_update_accounting(fs_info, tmp, src, dst, 1);
1419         spin_unlock(&fs_info->qgroup_lock);
1420 out:
1421         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1422         ulist_free(tmp);
1423         return ret;
1424 }
1425
1426 static int __del_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
1427                                  u64 dst)
1428 {
1429         struct btrfs_fs_info *fs_info = trans->fs_info;
1430         struct btrfs_qgroup *parent;
1431         struct btrfs_qgroup *member;
1432         struct btrfs_qgroup_list *list;
1433         struct ulist *tmp;
1434         bool found = false;
1435         unsigned int nofs_flag;
1436         int ret = 0;
1437         int ret2;
1438
1439         /* We hold a transaction handle open, must do a NOFS allocation. */
1440         nofs_flag = memalloc_nofs_save();
1441         tmp = ulist_alloc(GFP_KERNEL);
1442         memalloc_nofs_restore(nofs_flag);
1443         if (!tmp)
1444                 return -ENOMEM;
1445
1446         if (!fs_info->quota_root) {
1447                 ret = -ENOTCONN;
1448                 goto out;
1449         }
1450
1451         member = find_qgroup_rb(fs_info, src);
1452         parent = find_qgroup_rb(fs_info, dst);
1453         /*
1454          * The parent/member pair doesn't exist, then try to delete the dead
1455          * relation items only.
1456          */
1457         if (!member || !parent)
1458                 goto delete_item;
1459
1460         /* check if such qgroup relation exist firstly */
1461         list_for_each_entry(list, &member->groups, next_group) {
1462                 if (list->group == parent) {
1463                         found = true;
1464                         break;
1465                 }
1466         }
1467
1468 delete_item:
1469         ret = del_qgroup_relation_item(trans, src, dst);
1470         if (ret < 0 && ret != -ENOENT)
1471                 goto out;
1472         ret2 = del_qgroup_relation_item(trans, dst, src);
1473         if (ret2 < 0 && ret2 != -ENOENT)
1474                 goto out;
1475
1476         /* At least one deletion succeeded, return 0 */
1477         if (!ret || !ret2)
1478                 ret = 0;
1479
1480         if (found) {
1481                 spin_lock(&fs_info->qgroup_lock);
1482                 del_relation_rb(fs_info, src, dst);
1483                 ret = quick_update_accounting(fs_info, tmp, src, dst, -1);
1484                 spin_unlock(&fs_info->qgroup_lock);
1485         }
1486 out:
1487         ulist_free(tmp);
1488         return ret;
1489 }
1490
1491 int btrfs_del_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
1492                               u64 dst)
1493 {
1494         struct btrfs_fs_info *fs_info = trans->fs_info;
1495         int ret = 0;
1496
1497         mutex_lock(&fs_info->qgroup_ioctl_lock);
1498         ret = __del_qgroup_relation(trans, src, dst);
1499         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1500
1501         return ret;
1502 }
1503
1504 int btrfs_create_qgroup(struct btrfs_trans_handle *trans, u64 qgroupid)
1505 {
1506         struct btrfs_fs_info *fs_info = trans->fs_info;
1507         struct btrfs_root *quota_root;
1508         struct btrfs_qgroup *qgroup;
1509         int ret = 0;
1510
1511         mutex_lock(&fs_info->qgroup_ioctl_lock);
1512         if (!fs_info->quota_root) {
1513                 ret = -ENOTCONN;
1514                 goto out;
1515         }
1516         quota_root = fs_info->quota_root;
1517         qgroup = find_qgroup_rb(fs_info, qgroupid);
1518         if (qgroup) {
1519                 ret = -EEXIST;
1520                 goto out;
1521         }
1522
1523         ret = add_qgroup_item(trans, quota_root, qgroupid);
1524         if (ret)
1525                 goto out;
1526
1527         spin_lock(&fs_info->qgroup_lock);
1528         qgroup = add_qgroup_rb(fs_info, qgroupid);
1529         spin_unlock(&fs_info->qgroup_lock);
1530
1531         if (IS_ERR(qgroup)) {
1532                 ret = PTR_ERR(qgroup);
1533                 goto out;
1534         }
1535         ret = btrfs_sysfs_add_one_qgroup(fs_info, qgroup);
1536 out:
1537         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1538         return ret;
1539 }
1540
1541 int btrfs_remove_qgroup(struct btrfs_trans_handle *trans, u64 qgroupid)
1542 {
1543         struct btrfs_fs_info *fs_info = trans->fs_info;
1544         struct btrfs_qgroup *qgroup;
1545         struct btrfs_qgroup_list *list;
1546         int ret = 0;
1547
1548         mutex_lock(&fs_info->qgroup_ioctl_lock);
1549         if (!fs_info->quota_root) {
1550                 ret = -ENOTCONN;
1551                 goto out;
1552         }
1553
1554         qgroup = find_qgroup_rb(fs_info, qgroupid);
1555         if (!qgroup) {
1556                 ret = -ENOENT;
1557                 goto out;
1558         }
1559
1560         /* Check if there are no children of this qgroup */
1561         if (!list_empty(&qgroup->members)) {
1562                 ret = -EBUSY;
1563                 goto out;
1564         }
1565
1566         ret = del_qgroup_item(trans, qgroupid);
1567         if (ret && ret != -ENOENT)
1568                 goto out;
1569
1570         while (!list_empty(&qgroup->groups)) {
1571                 list = list_first_entry(&qgroup->groups,
1572                                         struct btrfs_qgroup_list, next_group);
1573                 ret = __del_qgroup_relation(trans, qgroupid,
1574                                             list->group->qgroupid);
1575                 if (ret)
1576                         goto out;
1577         }
1578
1579         spin_lock(&fs_info->qgroup_lock);
1580         del_qgroup_rb(fs_info, qgroupid);
1581         spin_unlock(&fs_info->qgroup_lock);
1582
1583         /*
1584          * Remove the qgroup from sysfs now without holding the qgroup_lock
1585          * spinlock, since the sysfs_remove_group() function needs to take
1586          * the mutex kernfs_mutex through kernfs_remove_by_name_ns().
1587          */
1588         btrfs_sysfs_del_one_qgroup(fs_info, qgroup);
1589         kfree(qgroup);
1590 out:
1591         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1592         return ret;
1593 }
1594
1595 int btrfs_limit_qgroup(struct btrfs_trans_handle *trans, u64 qgroupid,
1596                        struct btrfs_qgroup_limit *limit)
1597 {
1598         struct btrfs_fs_info *fs_info = trans->fs_info;
1599         struct btrfs_qgroup *qgroup;
1600         int ret = 0;
1601         /* Sometimes we would want to clear the limit on this qgroup.
1602          * To meet this requirement, we treat the -1 as a special value
1603          * which tell kernel to clear the limit on this qgroup.
1604          */
1605         const u64 CLEAR_VALUE = -1;
1606
1607         mutex_lock(&fs_info->qgroup_ioctl_lock);
1608         if (!fs_info->quota_root) {
1609                 ret = -ENOTCONN;
1610                 goto out;
1611         }
1612
1613         qgroup = find_qgroup_rb(fs_info, qgroupid);
1614         if (!qgroup) {
1615                 ret = -ENOENT;
1616                 goto out;
1617         }
1618
1619         spin_lock(&fs_info->qgroup_lock);
1620         if (limit->flags & BTRFS_QGROUP_LIMIT_MAX_RFER) {
1621                 if (limit->max_rfer == CLEAR_VALUE) {
1622                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_MAX_RFER;
1623                         limit->flags &= ~BTRFS_QGROUP_LIMIT_MAX_RFER;
1624                         qgroup->max_rfer = 0;
1625                 } else {
1626                         qgroup->max_rfer = limit->max_rfer;
1627                 }
1628         }
1629         if (limit->flags & BTRFS_QGROUP_LIMIT_MAX_EXCL) {
1630                 if (limit->max_excl == CLEAR_VALUE) {
1631                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_MAX_EXCL;
1632                         limit->flags &= ~BTRFS_QGROUP_LIMIT_MAX_EXCL;
1633                         qgroup->max_excl = 0;
1634                 } else {
1635                         qgroup->max_excl = limit->max_excl;
1636                 }
1637         }
1638         if (limit->flags & BTRFS_QGROUP_LIMIT_RSV_RFER) {
1639                 if (limit->rsv_rfer == CLEAR_VALUE) {
1640                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_RSV_RFER;
1641                         limit->flags &= ~BTRFS_QGROUP_LIMIT_RSV_RFER;
1642                         qgroup->rsv_rfer = 0;
1643                 } else {
1644                         qgroup->rsv_rfer = limit->rsv_rfer;
1645                 }
1646         }
1647         if (limit->flags & BTRFS_QGROUP_LIMIT_RSV_EXCL) {
1648                 if (limit->rsv_excl == CLEAR_VALUE) {
1649                         qgroup->lim_flags &= ~BTRFS_QGROUP_LIMIT_RSV_EXCL;
1650                         limit->flags &= ~BTRFS_QGROUP_LIMIT_RSV_EXCL;
1651                         qgroup->rsv_excl = 0;
1652                 } else {
1653                         qgroup->rsv_excl = limit->rsv_excl;
1654                 }
1655         }
1656         qgroup->lim_flags |= limit->flags;
1657
1658         spin_unlock(&fs_info->qgroup_lock);
1659
1660         ret = update_qgroup_limit_item(trans, qgroup);
1661         if (ret) {
1662                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
1663                 btrfs_info(fs_info, "unable to update quota limit for %llu",
1664                        qgroupid);
1665         }
1666
1667 out:
1668         mutex_unlock(&fs_info->qgroup_ioctl_lock);
1669         return ret;
1670 }
1671
1672 int btrfs_qgroup_trace_extent_nolock(struct btrfs_fs_info *fs_info,
1673                                 struct btrfs_delayed_ref_root *delayed_refs,
1674                                 struct btrfs_qgroup_extent_record *record)
1675 {
1676         struct rb_node **p = &delayed_refs->dirty_extent_root.rb_node;
1677         struct rb_node *parent_node = NULL;
1678         struct btrfs_qgroup_extent_record *entry;
1679         u64 bytenr = record->bytenr;
1680
1681         lockdep_assert_held(&delayed_refs->lock);
1682         trace_btrfs_qgroup_trace_extent(fs_info, record);
1683
1684         while (*p) {
1685                 parent_node = *p;
1686                 entry = rb_entry(parent_node, struct btrfs_qgroup_extent_record,
1687                                  node);
1688                 if (bytenr < entry->bytenr) {
1689                         p = &(*p)->rb_left;
1690                 } else if (bytenr > entry->bytenr) {
1691                         p = &(*p)->rb_right;
1692                 } else {
1693                         if (record->data_rsv && !entry->data_rsv) {
1694                                 entry->data_rsv = record->data_rsv;
1695                                 entry->data_rsv_refroot =
1696                                         record->data_rsv_refroot;
1697                         }
1698                         return 1;
1699                 }
1700         }
1701
1702         rb_link_node(&record->node, parent_node, p);
1703         rb_insert_color(&record->node, &delayed_refs->dirty_extent_root);
1704         return 0;
1705 }
1706
1707 int btrfs_qgroup_trace_extent_post(struct btrfs_trans_handle *trans,
1708                                    struct btrfs_qgroup_extent_record *qrecord)
1709 {
1710         struct ulist *old_root;
1711         u64 bytenr = qrecord->bytenr;
1712         int ret;
1713
1714         /*
1715          * We are always called in a context where we are already holding a
1716          * transaction handle. Often we are called when adding a data delayed
1717          * reference from btrfs_truncate_inode_items() (truncating or unlinking),
1718          * in which case we will be holding a write lock on extent buffer from a
1719          * subvolume tree. In this case we can't allow btrfs_find_all_roots() to
1720          * acquire fs_info->commit_root_sem, because that is a higher level lock
1721          * that must be acquired before locking any extent buffers.
1722          *
1723          * So we want btrfs_find_all_roots() to not acquire the commit_root_sem
1724          * but we can't pass it a non-NULL transaction handle, because otherwise
1725          * it would not use commit roots and would lock extent buffers, causing
1726          * a deadlock if it ends up trying to read lock the same extent buffer
1727          * that was previously write locked at btrfs_truncate_inode_items().
1728          *
1729          * So pass a NULL transaction handle to btrfs_find_all_roots() and
1730          * explicitly tell it to not acquire the commit_root_sem - if we are
1731          * holding a transaction handle we don't need its protection.
1732          */
1733         ASSERT(trans != NULL);
1734
1735         ret = btrfs_find_all_roots(NULL, trans->fs_info, bytenr, 0, &old_root,
1736                                    false, true);
1737         if (ret < 0) {
1738                 trans->fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
1739                 btrfs_warn(trans->fs_info,
1740 "error accounting new delayed refs extent (err code: %d), quota inconsistent",
1741                         ret);
1742                 return 0;
1743         }
1744
1745         /*
1746          * Here we don't need to get the lock of
1747          * trans->transaction->delayed_refs, since inserted qrecord won't
1748          * be deleted, only qrecord->node may be modified (new qrecord insert)
1749          *
1750          * So modifying qrecord->old_roots is safe here
1751          */
1752         qrecord->old_roots = old_root;
1753         return 0;
1754 }
1755
1756 int btrfs_qgroup_trace_extent(struct btrfs_trans_handle *trans, u64 bytenr,
1757                               u64 num_bytes, gfp_t gfp_flag)
1758 {
1759         struct btrfs_fs_info *fs_info = trans->fs_info;
1760         struct btrfs_qgroup_extent_record *record;
1761         struct btrfs_delayed_ref_root *delayed_refs;
1762         int ret;
1763
1764         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags)
1765             || bytenr == 0 || num_bytes == 0)
1766                 return 0;
1767         record = kzalloc(sizeof(*record), gfp_flag);
1768         if (!record)
1769                 return -ENOMEM;
1770
1771         delayed_refs = &trans->transaction->delayed_refs;
1772         record->bytenr = bytenr;
1773         record->num_bytes = num_bytes;
1774         record->old_roots = NULL;
1775
1776         spin_lock(&delayed_refs->lock);
1777         ret = btrfs_qgroup_trace_extent_nolock(fs_info, delayed_refs, record);
1778         spin_unlock(&delayed_refs->lock);
1779         if (ret > 0) {
1780                 kfree(record);
1781                 return 0;
1782         }
1783         return btrfs_qgroup_trace_extent_post(trans, record);
1784 }
1785
1786 int btrfs_qgroup_trace_leaf_items(struct btrfs_trans_handle *trans,
1787                                   struct extent_buffer *eb)
1788 {
1789         struct btrfs_fs_info *fs_info = trans->fs_info;
1790         int nr = btrfs_header_nritems(eb);
1791         int i, extent_type, ret;
1792         struct btrfs_key key;
1793         struct btrfs_file_extent_item *fi;
1794         u64 bytenr, num_bytes;
1795
1796         /* We can be called directly from walk_up_proc() */
1797         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
1798                 return 0;
1799
1800         for (i = 0; i < nr; i++) {
1801                 btrfs_item_key_to_cpu(eb, &key, i);
1802
1803                 if (key.type != BTRFS_EXTENT_DATA_KEY)
1804                         continue;
1805
1806                 fi = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
1807                 /* filter out non qgroup-accountable extents  */
1808                 extent_type = btrfs_file_extent_type(eb, fi);
1809
1810                 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
1811                         continue;
1812
1813                 bytenr = btrfs_file_extent_disk_bytenr(eb, fi);
1814                 if (!bytenr)
1815                         continue;
1816
1817                 num_bytes = btrfs_file_extent_disk_num_bytes(eb, fi);
1818
1819                 ret = btrfs_qgroup_trace_extent(trans, bytenr, num_bytes,
1820                                                 GFP_NOFS);
1821                 if (ret)
1822                         return ret;
1823         }
1824         cond_resched();
1825         return 0;
1826 }
1827
1828 /*
1829  * Walk up the tree from the bottom, freeing leaves and any interior
1830  * nodes which have had all slots visited. If a node (leaf or
1831  * interior) is freed, the node above it will have it's slot
1832  * incremented. The root node will never be freed.
1833  *
1834  * At the end of this function, we should have a path which has all
1835  * slots incremented to the next position for a search. If we need to
1836  * read a new node it will be NULL and the node above it will have the
1837  * correct slot selected for a later read.
1838  *
1839  * If we increment the root nodes slot counter past the number of
1840  * elements, 1 is returned to signal completion of the search.
1841  */
1842 static int adjust_slots_upwards(struct btrfs_path *path, int root_level)
1843 {
1844         int level = 0;
1845         int nr, slot;
1846         struct extent_buffer *eb;
1847
1848         if (root_level == 0)
1849                 return 1;
1850
1851         while (level <= root_level) {
1852                 eb = path->nodes[level];
1853                 nr = btrfs_header_nritems(eb);
1854                 path->slots[level]++;
1855                 slot = path->slots[level];
1856                 if (slot >= nr || level == 0) {
1857                         /*
1858                          * Don't free the root -  we will detect this
1859                          * condition after our loop and return a
1860                          * positive value for caller to stop walking the tree.
1861                          */
1862                         if (level != root_level) {
1863                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
1864                                 path->locks[level] = 0;
1865
1866                                 free_extent_buffer(eb);
1867                                 path->nodes[level] = NULL;
1868                                 path->slots[level] = 0;
1869                         }
1870                 } else {
1871                         /*
1872                          * We have a valid slot to walk back down
1873                          * from. Stop here so caller can process these
1874                          * new nodes.
1875                          */
1876                         break;
1877                 }
1878
1879                 level++;
1880         }
1881
1882         eb = path->nodes[root_level];
1883         if (path->slots[root_level] >= btrfs_header_nritems(eb))
1884                 return 1;
1885
1886         return 0;
1887 }
1888
1889 /*
1890  * Helper function to trace a subtree tree block swap.
1891  *
1892  * The swap will happen in highest tree block, but there may be a lot of
1893  * tree blocks involved.
1894  *
1895  * For example:
1896  *  OO = Old tree blocks
1897  *  NN = New tree blocks allocated during balance
1898  *
1899  *           File tree (257)                  Reloc tree for 257
1900  * L2              OO                                NN
1901  *               /    \                            /    \
1902  * L1          OO      OO (a)                    OO      NN (a)
1903  *            / \     / \                       / \     / \
1904  * L0       OO   OO OO   OO                   OO   OO NN   NN
1905  *                  (b)  (c)                          (b)  (c)
1906  *
1907  * When calling qgroup_trace_extent_swap(), we will pass:
1908  * @src_eb = OO(a)
1909  * @dst_path = [ nodes[1] = NN(a), nodes[0] = NN(c) ]
1910  * @dst_level = 0
1911  * @root_level = 1
1912  *
1913  * In that case, qgroup_trace_extent_swap() will search from OO(a) to
1914  * reach OO(c), then mark both OO(c) and NN(c) as qgroup dirty.
1915  *
1916  * The main work of qgroup_trace_extent_swap() can be split into 3 parts:
1917  *
1918  * 1) Tree search from @src_eb
1919  *    It should acts as a simplified btrfs_search_slot().
1920  *    The key for search can be extracted from @dst_path->nodes[dst_level]
1921  *    (first key).
1922  *
1923  * 2) Mark the final tree blocks in @src_path and @dst_path qgroup dirty
1924  *    NOTE: In above case, OO(a) and NN(a) won't be marked qgroup dirty.
1925  *    They should be marked during previous (@dst_level = 1) iteration.
1926  *
1927  * 3) Mark file extents in leaves dirty
1928  *    We don't have good way to pick out new file extents only.
1929  *    So we still follow the old method by scanning all file extents in
1930  *    the leave.
1931  *
1932  * This function can free us from keeping two paths, thus later we only need
1933  * to care about how to iterate all new tree blocks in reloc tree.
1934  */
1935 static int qgroup_trace_extent_swap(struct btrfs_trans_handle* trans,
1936                                     struct extent_buffer *src_eb,
1937                                     struct btrfs_path *dst_path,
1938                                     int dst_level, int root_level,
1939                                     bool trace_leaf)
1940 {
1941         struct btrfs_key key;
1942         struct btrfs_path *src_path;
1943         struct btrfs_fs_info *fs_info = trans->fs_info;
1944         u32 nodesize = fs_info->nodesize;
1945         int cur_level = root_level;
1946         int ret;
1947
1948         BUG_ON(dst_level > root_level);
1949         /* Level mismatch */
1950         if (btrfs_header_level(src_eb) != root_level)
1951                 return -EINVAL;
1952
1953         src_path = btrfs_alloc_path();
1954         if (!src_path) {
1955                 ret = -ENOMEM;
1956                 goto out;
1957         }
1958
1959         if (dst_level)
1960                 btrfs_node_key_to_cpu(dst_path->nodes[dst_level], &key, 0);
1961         else
1962                 btrfs_item_key_to_cpu(dst_path->nodes[dst_level], &key, 0);
1963
1964         /* For src_path */
1965         atomic_inc(&src_eb->refs);
1966         src_path->nodes[root_level] = src_eb;
1967         src_path->slots[root_level] = dst_path->slots[root_level];
1968         src_path->locks[root_level] = 0;
1969
1970         /* A simplified version of btrfs_search_slot() */
1971         while (cur_level >= dst_level) {
1972                 struct btrfs_key src_key;
1973                 struct btrfs_key dst_key;
1974
1975                 if (src_path->nodes[cur_level] == NULL) {
1976                         struct extent_buffer *eb;
1977                         int parent_slot;
1978
1979                         eb = src_path->nodes[cur_level + 1];
1980                         parent_slot = src_path->slots[cur_level + 1];
1981
1982                         eb = btrfs_read_node_slot(eb, parent_slot);
1983                         if (IS_ERR(eb)) {
1984                                 ret = PTR_ERR(eb);
1985                                 goto out;
1986                         }
1987
1988                         src_path->nodes[cur_level] = eb;
1989
1990                         btrfs_tree_read_lock(eb);
1991                         src_path->locks[cur_level] = BTRFS_READ_LOCK;
1992                 }
1993
1994                 src_path->slots[cur_level] = dst_path->slots[cur_level];
1995                 if (cur_level) {
1996                         btrfs_node_key_to_cpu(dst_path->nodes[cur_level],
1997                                         &dst_key, dst_path->slots[cur_level]);
1998                         btrfs_node_key_to_cpu(src_path->nodes[cur_level],
1999                                         &src_key, src_path->slots[cur_level]);
2000                 } else {
2001                         btrfs_item_key_to_cpu(dst_path->nodes[cur_level],
2002                                         &dst_key, dst_path->slots[cur_level]);
2003                         btrfs_item_key_to_cpu(src_path->nodes[cur_level],
2004                                         &src_key, src_path->slots[cur_level]);
2005                 }
2006                 /* Content mismatch, something went wrong */
2007                 if (btrfs_comp_cpu_keys(&dst_key, &src_key)) {
2008                         ret = -ENOENT;
2009                         goto out;
2010                 }
2011                 cur_level--;
2012         }
2013
2014         /*
2015          * Now both @dst_path and @src_path have been populated, record the tree
2016          * blocks for qgroup accounting.
2017          */
2018         ret = btrfs_qgroup_trace_extent(trans, src_path->nodes[dst_level]->start,
2019                         nodesize, GFP_NOFS);
2020         if (ret < 0)
2021                 goto out;
2022         ret = btrfs_qgroup_trace_extent(trans,
2023                         dst_path->nodes[dst_level]->start,
2024                         nodesize, GFP_NOFS);
2025         if (ret < 0)
2026                 goto out;
2027
2028         /* Record leaf file extents */
2029         if (dst_level == 0 && trace_leaf) {
2030                 ret = btrfs_qgroup_trace_leaf_items(trans, src_path->nodes[0]);
2031                 if (ret < 0)
2032                         goto out;
2033                 ret = btrfs_qgroup_trace_leaf_items(trans, dst_path->nodes[0]);
2034         }
2035 out:
2036         btrfs_free_path(src_path);
2037         return ret;
2038 }
2039
2040 /*
2041  * Helper function to do recursive generation-aware depth-first search, to
2042  * locate all new tree blocks in a subtree of reloc tree.
2043  *
2044  * E.g. (OO = Old tree blocks, NN = New tree blocks, whose gen == last_snapshot)
2045  *         reloc tree
2046  * L2         NN (a)
2047  *          /    \
2048  * L1    OO        NN (b)
2049  *      /  \      /  \
2050  * L0  OO  OO    OO  NN
2051  *               (c) (d)
2052  * If we pass:
2053  * @dst_path = [ nodes[1] = NN(b), nodes[0] = NULL ],
2054  * @cur_level = 1
2055  * @root_level = 1
2056  *
2057  * We will iterate through tree blocks NN(b), NN(d) and info qgroup to trace
2058  * above tree blocks along with their counter parts in file tree.
2059  * While during search, old tree blocks OO(c) will be skipped as tree block swap
2060  * won't affect OO(c).
2061  */
2062 static int qgroup_trace_new_subtree_blocks(struct btrfs_trans_handle* trans,
2063                                            struct extent_buffer *src_eb,
2064                                            struct btrfs_path *dst_path,
2065                                            int cur_level, int root_level,
2066                                            u64 last_snapshot, bool trace_leaf)
2067 {
2068         struct btrfs_fs_info *fs_info = trans->fs_info;
2069         struct extent_buffer *eb;
2070         bool need_cleanup = false;
2071         int ret = 0;
2072         int i;
2073
2074         /* Level sanity check */
2075         if (cur_level < 0 || cur_level >= BTRFS_MAX_LEVEL - 1 ||
2076             root_level < 0 || root_level >= BTRFS_MAX_LEVEL - 1 ||
2077             root_level < cur_level) {
2078                 btrfs_err_rl(fs_info,
2079                         "%s: bad levels, cur_level=%d root_level=%d",
2080                         __func__, cur_level, root_level);
2081                 return -EUCLEAN;
2082         }
2083
2084         /* Read the tree block if needed */
2085         if (dst_path->nodes[cur_level] == NULL) {
2086                 int parent_slot;
2087                 u64 child_gen;
2088
2089                 /*
2090                  * dst_path->nodes[root_level] must be initialized before
2091                  * calling this function.
2092                  */
2093                 if (cur_level == root_level) {
2094                         btrfs_err_rl(fs_info,
2095         "%s: dst_path->nodes[%d] not initialized, root_level=%d cur_level=%d",
2096                                 __func__, root_level, root_level, cur_level);
2097                         return -EUCLEAN;
2098                 }
2099
2100                 /*
2101                  * We need to get child blockptr/gen from parent before we can
2102                  * read it.
2103                   */
2104                 eb = dst_path->nodes[cur_level + 1];
2105                 parent_slot = dst_path->slots[cur_level + 1];
2106                 child_gen = btrfs_node_ptr_generation(eb, parent_slot);
2107
2108                 /* This node is old, no need to trace */
2109                 if (child_gen < last_snapshot)
2110                         goto out;
2111
2112                 eb = btrfs_read_node_slot(eb, parent_slot);
2113                 if (IS_ERR(eb)) {
2114                         ret = PTR_ERR(eb);
2115                         goto out;
2116                 }
2117
2118                 dst_path->nodes[cur_level] = eb;
2119                 dst_path->slots[cur_level] = 0;
2120
2121                 btrfs_tree_read_lock(eb);
2122                 dst_path->locks[cur_level] = BTRFS_READ_LOCK;
2123                 need_cleanup = true;
2124         }
2125
2126         /* Now record this tree block and its counter part for qgroups */
2127         ret = qgroup_trace_extent_swap(trans, src_eb, dst_path, cur_level,
2128                                        root_level, trace_leaf);
2129         if (ret < 0)
2130                 goto cleanup;
2131
2132         eb = dst_path->nodes[cur_level];
2133
2134         if (cur_level > 0) {
2135                 /* Iterate all child tree blocks */
2136                 for (i = 0; i < btrfs_header_nritems(eb); i++) {
2137                         /* Skip old tree blocks as they won't be swapped */
2138                         if (btrfs_node_ptr_generation(eb, i) < last_snapshot)
2139                                 continue;
2140                         dst_path->slots[cur_level] = i;
2141
2142                         /* Recursive call (at most 7 times) */
2143                         ret = qgroup_trace_new_subtree_blocks(trans, src_eb,
2144                                         dst_path, cur_level - 1, root_level,
2145                                         last_snapshot, trace_leaf);
2146                         if (ret < 0)
2147                                 goto cleanup;
2148                 }
2149         }
2150
2151 cleanup:
2152         if (need_cleanup) {
2153                 /* Clean up */
2154                 btrfs_tree_unlock_rw(dst_path->nodes[cur_level],
2155                                      dst_path->locks[cur_level]);
2156                 free_extent_buffer(dst_path->nodes[cur_level]);
2157                 dst_path->nodes[cur_level] = NULL;
2158                 dst_path->slots[cur_level] = 0;
2159                 dst_path->locks[cur_level] = 0;
2160         }
2161 out:
2162         return ret;
2163 }
2164
2165 static int qgroup_trace_subtree_swap(struct btrfs_trans_handle *trans,
2166                                 struct extent_buffer *src_eb,
2167                                 struct extent_buffer *dst_eb,
2168                                 u64 last_snapshot, bool trace_leaf)
2169 {
2170         struct btrfs_fs_info *fs_info = trans->fs_info;
2171         struct btrfs_path *dst_path = NULL;
2172         int level;
2173         int ret;
2174
2175         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2176                 return 0;
2177
2178         /* Wrong parameter order */
2179         if (btrfs_header_generation(src_eb) > btrfs_header_generation(dst_eb)) {
2180                 btrfs_err_rl(fs_info,
2181                 "%s: bad parameter order, src_gen=%llu dst_gen=%llu", __func__,
2182                              btrfs_header_generation(src_eb),
2183                              btrfs_header_generation(dst_eb));
2184                 return -EUCLEAN;
2185         }
2186
2187         if (!extent_buffer_uptodate(src_eb) || !extent_buffer_uptodate(dst_eb)) {
2188                 ret = -EIO;
2189                 goto out;
2190         }
2191
2192         level = btrfs_header_level(dst_eb);
2193         dst_path = btrfs_alloc_path();
2194         if (!dst_path) {
2195                 ret = -ENOMEM;
2196                 goto out;
2197         }
2198         /* For dst_path */
2199         atomic_inc(&dst_eb->refs);
2200         dst_path->nodes[level] = dst_eb;
2201         dst_path->slots[level] = 0;
2202         dst_path->locks[level] = 0;
2203
2204         /* Do the generation aware breadth-first search */
2205         ret = qgroup_trace_new_subtree_blocks(trans, src_eb, dst_path, level,
2206                                               level, last_snapshot, trace_leaf);
2207         if (ret < 0)
2208                 goto out;
2209         ret = 0;
2210
2211 out:
2212         btrfs_free_path(dst_path);
2213         if (ret < 0)
2214                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2215         return ret;
2216 }
2217
2218 int btrfs_qgroup_trace_subtree(struct btrfs_trans_handle *trans,
2219                                struct extent_buffer *root_eb,
2220                                u64 root_gen, int root_level)
2221 {
2222         struct btrfs_fs_info *fs_info = trans->fs_info;
2223         int ret = 0;
2224         int level;
2225         struct extent_buffer *eb = root_eb;
2226         struct btrfs_path *path = NULL;
2227
2228         BUG_ON(root_level < 0 || root_level >= BTRFS_MAX_LEVEL);
2229         BUG_ON(root_eb == NULL);
2230
2231         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2232                 return 0;
2233
2234         if (!extent_buffer_uptodate(root_eb)) {
2235                 ret = btrfs_read_buffer(root_eb, root_gen, root_level, NULL);
2236                 if (ret)
2237                         goto out;
2238         }
2239
2240         if (root_level == 0) {
2241                 ret = btrfs_qgroup_trace_leaf_items(trans, root_eb);
2242                 goto out;
2243         }
2244
2245         path = btrfs_alloc_path();
2246         if (!path)
2247                 return -ENOMEM;
2248
2249         /*
2250          * Walk down the tree.  Missing extent blocks are filled in as
2251          * we go. Metadata is accounted every time we read a new
2252          * extent block.
2253          *
2254          * When we reach a leaf, we account for file extent items in it,
2255          * walk back up the tree (adjusting slot pointers as we go)
2256          * and restart the search process.
2257          */
2258         atomic_inc(&root_eb->refs);     /* For path */
2259         path->nodes[root_level] = root_eb;
2260         path->slots[root_level] = 0;
2261         path->locks[root_level] = 0; /* so release_path doesn't try to unlock */
2262 walk_down:
2263         level = root_level;
2264         while (level >= 0) {
2265                 if (path->nodes[level] == NULL) {
2266                         int parent_slot;
2267                         u64 child_bytenr;
2268
2269                         /*
2270                          * We need to get child blockptr from parent before we
2271                          * can read it.
2272                           */
2273                         eb = path->nodes[level + 1];
2274                         parent_slot = path->slots[level + 1];
2275                         child_bytenr = btrfs_node_blockptr(eb, parent_slot);
2276
2277                         eb = btrfs_read_node_slot(eb, parent_slot);
2278                         if (IS_ERR(eb)) {
2279                                 ret = PTR_ERR(eb);
2280                                 goto out;
2281                         }
2282
2283                         path->nodes[level] = eb;
2284                         path->slots[level] = 0;
2285
2286                         btrfs_tree_read_lock(eb);
2287                         path->locks[level] = BTRFS_READ_LOCK;
2288
2289                         ret = btrfs_qgroup_trace_extent(trans, child_bytenr,
2290                                                         fs_info->nodesize,
2291                                                         GFP_NOFS);
2292                         if (ret)
2293                                 goto out;
2294                 }
2295
2296                 if (level == 0) {
2297                         ret = btrfs_qgroup_trace_leaf_items(trans,
2298                                                             path->nodes[level]);
2299                         if (ret)
2300                                 goto out;
2301
2302                         /* Nonzero return here means we completed our search */
2303                         ret = adjust_slots_upwards(path, root_level);
2304                         if (ret)
2305                                 break;
2306
2307                         /* Restart search with new slots */
2308                         goto walk_down;
2309                 }
2310
2311                 level--;
2312         }
2313
2314         ret = 0;
2315 out:
2316         btrfs_free_path(path);
2317
2318         return ret;
2319 }
2320
2321 #define UPDATE_NEW      0
2322 #define UPDATE_OLD      1
2323 /*
2324  * Walk all of the roots that points to the bytenr and adjust their refcnts.
2325  */
2326 static int qgroup_update_refcnt(struct btrfs_fs_info *fs_info,
2327                                 struct ulist *roots, struct ulist *tmp,
2328                                 struct ulist *qgroups, u64 seq, int update_old)
2329 {
2330         struct ulist_node *unode;
2331         struct ulist_iterator uiter;
2332         struct ulist_node *tmp_unode;
2333         struct ulist_iterator tmp_uiter;
2334         struct btrfs_qgroup *qg;
2335         int ret = 0;
2336
2337         if (!roots)
2338                 return 0;
2339         ULIST_ITER_INIT(&uiter);
2340         while ((unode = ulist_next(roots, &uiter))) {
2341                 qg = find_qgroup_rb(fs_info, unode->val);
2342                 if (!qg)
2343                         continue;
2344
2345                 ulist_reinit(tmp);
2346                 ret = ulist_add(qgroups, qg->qgroupid, qgroup_to_aux(qg),
2347                                 GFP_ATOMIC);
2348                 if (ret < 0)
2349                         return ret;
2350                 ret = ulist_add(tmp, qg->qgroupid, qgroup_to_aux(qg), GFP_ATOMIC);
2351                 if (ret < 0)
2352                         return ret;
2353                 ULIST_ITER_INIT(&tmp_uiter);
2354                 while ((tmp_unode = ulist_next(tmp, &tmp_uiter))) {
2355                         struct btrfs_qgroup_list *glist;
2356
2357                         qg = unode_aux_to_qgroup(tmp_unode);
2358                         if (update_old)
2359                                 btrfs_qgroup_update_old_refcnt(qg, seq, 1);
2360                         else
2361                                 btrfs_qgroup_update_new_refcnt(qg, seq, 1);
2362                         list_for_each_entry(glist, &qg->groups, next_group) {
2363                                 ret = ulist_add(qgroups, glist->group->qgroupid,
2364                                                 qgroup_to_aux(glist->group),
2365                                                 GFP_ATOMIC);
2366                                 if (ret < 0)
2367                                         return ret;
2368                                 ret = ulist_add(tmp, glist->group->qgroupid,
2369                                                 qgroup_to_aux(glist->group),
2370                                                 GFP_ATOMIC);
2371                                 if (ret < 0)
2372                                         return ret;
2373                         }
2374                 }
2375         }
2376         return 0;
2377 }
2378
2379 /*
2380  * Update qgroup rfer/excl counters.
2381  * Rfer update is easy, codes can explain themselves.
2382  *
2383  * Excl update is tricky, the update is split into 2 parts.
2384  * Part 1: Possible exclusive <-> sharing detect:
2385  *      |       A       |       !A      |
2386  *  -------------------------------------
2387  *  B   |       *       |       -       |
2388  *  -------------------------------------
2389  *  !B  |       +       |       **      |
2390  *  -------------------------------------
2391  *
2392  * Conditions:
2393  * A:   cur_old_roots < nr_old_roots    (not exclusive before)
2394  * !A:  cur_old_roots == nr_old_roots   (possible exclusive before)
2395  * B:   cur_new_roots < nr_new_roots    (not exclusive now)
2396  * !B:  cur_new_roots == nr_new_roots   (possible exclusive now)
2397  *
2398  * Results:
2399  * +: Possible sharing -> exclusive     -: Possible exclusive -> sharing
2400  * *: Definitely not changed.           **: Possible unchanged.
2401  *
2402  * For !A and !B condition, the exception is cur_old/new_roots == 0 case.
2403  *
2404  * To make the logic clear, we first use condition A and B to split
2405  * combination into 4 results.
2406  *
2407  * Then, for result "+" and "-", check old/new_roots == 0 case, as in them
2408  * only on variant maybe 0.
2409  *
2410  * Lastly, check result **, since there are 2 variants maybe 0, split them
2411  * again(2x2).
2412  * But this time we don't need to consider other things, the codes and logic
2413  * is easy to understand now.
2414  */
2415 static int qgroup_update_counters(struct btrfs_fs_info *fs_info,
2416                                   struct ulist *qgroups,
2417                                   u64 nr_old_roots,
2418                                   u64 nr_new_roots,
2419                                   u64 num_bytes, u64 seq)
2420 {
2421         struct ulist_node *unode;
2422         struct ulist_iterator uiter;
2423         struct btrfs_qgroup *qg;
2424         u64 cur_new_count, cur_old_count;
2425
2426         ULIST_ITER_INIT(&uiter);
2427         while ((unode = ulist_next(qgroups, &uiter))) {
2428                 bool dirty = false;
2429
2430                 qg = unode_aux_to_qgroup(unode);
2431                 cur_old_count = btrfs_qgroup_get_old_refcnt(qg, seq);
2432                 cur_new_count = btrfs_qgroup_get_new_refcnt(qg, seq);
2433
2434                 trace_qgroup_update_counters(fs_info, qg, cur_old_count,
2435                                              cur_new_count);
2436
2437                 /* Rfer update part */
2438                 if (cur_old_count == 0 && cur_new_count > 0) {
2439                         qg->rfer += num_bytes;
2440                         qg->rfer_cmpr += num_bytes;
2441                         dirty = true;
2442                 }
2443                 if (cur_old_count > 0 && cur_new_count == 0) {
2444                         qg->rfer -= num_bytes;
2445                         qg->rfer_cmpr -= num_bytes;
2446                         dirty = true;
2447                 }
2448
2449                 /* Excl update part */
2450                 /* Exclusive/none -> shared case */
2451                 if (cur_old_count == nr_old_roots &&
2452                     cur_new_count < nr_new_roots) {
2453                         /* Exclusive -> shared */
2454                         if (cur_old_count != 0) {
2455                                 qg->excl -= num_bytes;
2456                                 qg->excl_cmpr -= num_bytes;
2457                                 dirty = true;
2458                         }
2459                 }
2460
2461                 /* Shared -> exclusive/none case */
2462                 if (cur_old_count < nr_old_roots &&
2463                     cur_new_count == nr_new_roots) {
2464                         /* Shared->exclusive */
2465                         if (cur_new_count != 0) {
2466                                 qg->excl += num_bytes;
2467                                 qg->excl_cmpr += num_bytes;
2468                                 dirty = true;
2469                         }
2470                 }
2471
2472                 /* Exclusive/none -> exclusive/none case */
2473                 if (cur_old_count == nr_old_roots &&
2474                     cur_new_count == nr_new_roots) {
2475                         if (cur_old_count == 0) {
2476                                 /* None -> exclusive/none */
2477
2478                                 if (cur_new_count != 0) {
2479                                         /* None -> exclusive */
2480                                         qg->excl += num_bytes;
2481                                         qg->excl_cmpr += num_bytes;
2482                                         dirty = true;
2483                                 }
2484                                 /* None -> none, nothing changed */
2485                         } else {
2486                                 /* Exclusive -> exclusive/none */
2487
2488                                 if (cur_new_count == 0) {
2489                                         /* Exclusive -> none */
2490                                         qg->excl -= num_bytes;
2491                                         qg->excl_cmpr -= num_bytes;
2492                                         dirty = true;
2493                                 }
2494                                 /* Exclusive -> exclusive, nothing changed */
2495                         }
2496                 }
2497
2498                 if (dirty)
2499                         qgroup_dirty(fs_info, qg);
2500         }
2501         return 0;
2502 }
2503
2504 /*
2505  * Check if the @roots potentially is a list of fs tree roots
2506  *
2507  * Return 0 for definitely not a fs/subvol tree roots ulist
2508  * Return 1 for possible fs/subvol tree roots in the list (considering an empty
2509  *          one as well)
2510  */
2511 static int maybe_fs_roots(struct ulist *roots)
2512 {
2513         struct ulist_node *unode;
2514         struct ulist_iterator uiter;
2515
2516         /* Empty one, still possible for fs roots */
2517         if (!roots || roots->nnodes == 0)
2518                 return 1;
2519
2520         ULIST_ITER_INIT(&uiter);
2521         unode = ulist_next(roots, &uiter);
2522         if (!unode)
2523                 return 1;
2524
2525         /*
2526          * If it contains fs tree roots, then it must belong to fs/subvol
2527          * trees.
2528          * If it contains a non-fs tree, it won't be shared with fs/subvol trees.
2529          */
2530         return is_fstree(unode->val);
2531 }
2532
2533 int btrfs_qgroup_account_extent(struct btrfs_trans_handle *trans, u64 bytenr,
2534                                 u64 num_bytes, struct ulist *old_roots,
2535                                 struct ulist *new_roots)
2536 {
2537         struct btrfs_fs_info *fs_info = trans->fs_info;
2538         struct ulist *qgroups = NULL;
2539         struct ulist *tmp = NULL;
2540         u64 seq;
2541         u64 nr_new_roots = 0;
2542         u64 nr_old_roots = 0;
2543         int ret = 0;
2544
2545         /*
2546          * If quotas get disabled meanwhile, the resources need to be freed and
2547          * we can't just exit here.
2548          */
2549         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2550                 goto out_free;
2551
2552         if (new_roots) {
2553                 if (!maybe_fs_roots(new_roots))
2554                         goto out_free;
2555                 nr_new_roots = new_roots->nnodes;
2556         }
2557         if (old_roots) {
2558                 if (!maybe_fs_roots(old_roots))
2559                         goto out_free;
2560                 nr_old_roots = old_roots->nnodes;
2561         }
2562
2563         /* Quick exit, either not fs tree roots, or won't affect any qgroup */
2564         if (nr_old_roots == 0 && nr_new_roots == 0)
2565                 goto out_free;
2566
2567         BUG_ON(!fs_info->quota_root);
2568
2569         trace_btrfs_qgroup_account_extent(fs_info, trans->transid, bytenr,
2570                                         num_bytes, nr_old_roots, nr_new_roots);
2571
2572         qgroups = ulist_alloc(GFP_NOFS);
2573         if (!qgroups) {
2574                 ret = -ENOMEM;
2575                 goto out_free;
2576         }
2577         tmp = ulist_alloc(GFP_NOFS);
2578         if (!tmp) {
2579                 ret = -ENOMEM;
2580                 goto out_free;
2581         }
2582
2583         mutex_lock(&fs_info->qgroup_rescan_lock);
2584         if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
2585                 if (fs_info->qgroup_rescan_progress.objectid <= bytenr) {
2586                         mutex_unlock(&fs_info->qgroup_rescan_lock);
2587                         ret = 0;
2588                         goto out_free;
2589                 }
2590         }
2591         mutex_unlock(&fs_info->qgroup_rescan_lock);
2592
2593         spin_lock(&fs_info->qgroup_lock);
2594         seq = fs_info->qgroup_seq;
2595
2596         /* Update old refcnts using old_roots */
2597         ret = qgroup_update_refcnt(fs_info, old_roots, tmp, qgroups, seq,
2598                                    UPDATE_OLD);
2599         if (ret < 0)
2600                 goto out;
2601
2602         /* Update new refcnts using new_roots */
2603         ret = qgroup_update_refcnt(fs_info, new_roots, tmp, qgroups, seq,
2604                                    UPDATE_NEW);
2605         if (ret < 0)
2606                 goto out;
2607
2608         qgroup_update_counters(fs_info, qgroups, nr_old_roots, nr_new_roots,
2609                                num_bytes, seq);
2610
2611         /*
2612          * Bump qgroup_seq to avoid seq overlap
2613          */
2614         fs_info->qgroup_seq += max(nr_old_roots, nr_new_roots) + 1;
2615 out:
2616         spin_unlock(&fs_info->qgroup_lock);
2617 out_free:
2618         ulist_free(tmp);
2619         ulist_free(qgroups);
2620         ulist_free(old_roots);
2621         ulist_free(new_roots);
2622         return ret;
2623 }
2624
2625 int btrfs_qgroup_account_extents(struct btrfs_trans_handle *trans)
2626 {
2627         struct btrfs_fs_info *fs_info = trans->fs_info;
2628         struct btrfs_qgroup_extent_record *record;
2629         struct btrfs_delayed_ref_root *delayed_refs;
2630         struct ulist *new_roots = NULL;
2631         struct rb_node *node;
2632         u64 num_dirty_extents = 0;
2633         u64 qgroup_to_skip;
2634         int ret = 0;
2635
2636         delayed_refs = &trans->transaction->delayed_refs;
2637         qgroup_to_skip = delayed_refs->qgroup_to_skip;
2638         while ((node = rb_first(&delayed_refs->dirty_extent_root))) {
2639                 record = rb_entry(node, struct btrfs_qgroup_extent_record,
2640                                   node);
2641
2642                 num_dirty_extents++;
2643                 trace_btrfs_qgroup_account_extents(fs_info, record);
2644
2645                 if (!ret) {
2646                         /*
2647                          * Old roots should be searched when inserting qgroup
2648                          * extent record
2649                          */
2650                         if (WARN_ON(!record->old_roots)) {
2651                                 /* Search commit root to find old_roots */
2652                                 ret = btrfs_find_all_roots(NULL, fs_info,
2653                                                 record->bytenr, 0,
2654                                                 &record->old_roots, false, false);
2655                                 if (ret < 0)
2656                                         goto cleanup;
2657                         }
2658
2659                         /* Free the reserved data space */
2660                         btrfs_qgroup_free_refroot(fs_info,
2661                                         record->data_rsv_refroot,
2662                                         record->data_rsv,
2663                                         BTRFS_QGROUP_RSV_DATA);
2664                         /*
2665                          * Use BTRFS_SEQ_LAST as time_seq to do special search,
2666                          * which doesn't lock tree or delayed_refs and search
2667                          * current root. It's safe inside commit_transaction().
2668                          */
2669                         ret = btrfs_find_all_roots(trans, fs_info,
2670                            record->bytenr, BTRFS_SEQ_LAST, &new_roots, false, false);
2671                         if (ret < 0)
2672                                 goto cleanup;
2673                         if (qgroup_to_skip) {
2674                                 ulist_del(new_roots, qgroup_to_skip, 0);
2675                                 ulist_del(record->old_roots, qgroup_to_skip,
2676                                           0);
2677                         }
2678                         ret = btrfs_qgroup_account_extent(trans, record->bytenr,
2679                                                           record->num_bytes,
2680                                                           record->old_roots,
2681                                                           new_roots);
2682                         record->old_roots = NULL;
2683                         new_roots = NULL;
2684                 }
2685 cleanup:
2686                 ulist_free(record->old_roots);
2687                 ulist_free(new_roots);
2688                 new_roots = NULL;
2689                 rb_erase(node, &delayed_refs->dirty_extent_root);
2690                 kfree(record);
2691
2692         }
2693         trace_qgroup_num_dirty_extents(fs_info, trans->transid,
2694                                        num_dirty_extents);
2695         return ret;
2696 }
2697
2698 /*
2699  * called from commit_transaction. Writes all changed qgroups to disk.
2700  */
2701 int btrfs_run_qgroups(struct btrfs_trans_handle *trans)
2702 {
2703         struct btrfs_fs_info *fs_info = trans->fs_info;
2704         int ret = 0;
2705
2706         if (!fs_info->quota_root)
2707                 return ret;
2708
2709         spin_lock(&fs_info->qgroup_lock);
2710         while (!list_empty(&fs_info->dirty_qgroups)) {
2711                 struct btrfs_qgroup *qgroup;
2712                 qgroup = list_first_entry(&fs_info->dirty_qgroups,
2713                                           struct btrfs_qgroup, dirty);
2714                 list_del_init(&qgroup->dirty);
2715                 spin_unlock(&fs_info->qgroup_lock);
2716                 ret = update_qgroup_info_item(trans, qgroup);
2717                 if (ret)
2718                         fs_info->qgroup_flags |=
2719                                         BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2720                 ret = update_qgroup_limit_item(trans, qgroup);
2721                 if (ret)
2722                         fs_info->qgroup_flags |=
2723                                         BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2724                 spin_lock(&fs_info->qgroup_lock);
2725         }
2726         if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2727                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_ON;
2728         else
2729                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_ON;
2730         spin_unlock(&fs_info->qgroup_lock);
2731
2732         ret = update_qgroup_status_item(trans);
2733         if (ret)
2734                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2735
2736         return ret;
2737 }
2738
2739 /*
2740  * Copy the accounting information between qgroups. This is necessary
2741  * when a snapshot or a subvolume is created. Throwing an error will
2742  * cause a transaction abort so we take extra care here to only error
2743  * when a readonly fs is a reasonable outcome.
2744  */
2745 int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
2746                          u64 objectid, struct btrfs_qgroup_inherit *inherit)
2747 {
2748         int ret = 0;
2749         int i;
2750         u64 *i_qgroups;
2751         bool committing = false;
2752         struct btrfs_fs_info *fs_info = trans->fs_info;
2753         struct btrfs_root *quota_root;
2754         struct btrfs_qgroup *srcgroup;
2755         struct btrfs_qgroup *dstgroup;
2756         bool need_rescan = false;
2757         u32 level_size = 0;
2758         u64 nums;
2759
2760         /*
2761          * There are only two callers of this function.
2762          *
2763          * One in create_subvol() in the ioctl context, which needs to hold
2764          * the qgroup_ioctl_lock.
2765          *
2766          * The other one in create_pending_snapshot() where no other qgroup
2767          * code can modify the fs as they all need to either start a new trans
2768          * or hold a trans handler, thus we don't need to hold
2769          * qgroup_ioctl_lock.
2770          * This would avoid long and complex lock chain and make lockdep happy.
2771          */
2772         spin_lock(&fs_info->trans_lock);
2773         if (trans->transaction->state == TRANS_STATE_COMMIT_DOING)
2774                 committing = true;
2775         spin_unlock(&fs_info->trans_lock);
2776
2777         if (!committing)
2778                 mutex_lock(&fs_info->qgroup_ioctl_lock);
2779         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
2780                 goto out;
2781
2782         quota_root = fs_info->quota_root;
2783         if (!quota_root) {
2784                 ret = -EINVAL;
2785                 goto out;
2786         }
2787
2788         if (inherit) {
2789                 i_qgroups = (u64 *)(inherit + 1);
2790                 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
2791                        2 * inherit->num_excl_copies;
2792                 for (i = 0; i < nums; ++i) {
2793                         srcgroup = find_qgroup_rb(fs_info, *i_qgroups);
2794
2795                         /*
2796                          * Zero out invalid groups so we can ignore
2797                          * them later.
2798                          */
2799                         if (!srcgroup ||
2800                             ((srcgroup->qgroupid >> 48) <= (objectid >> 48)))
2801                                 *i_qgroups = 0ULL;
2802
2803                         ++i_qgroups;
2804                 }
2805         }
2806
2807         /*
2808          * create a tracking group for the subvol itself
2809          */
2810         ret = add_qgroup_item(trans, quota_root, objectid);
2811         if (ret)
2812                 goto out;
2813
2814         /*
2815          * add qgroup to all inherited groups
2816          */
2817         if (inherit) {
2818                 i_qgroups = (u64 *)(inherit + 1);
2819                 for (i = 0; i < inherit->num_qgroups; ++i, ++i_qgroups) {
2820                         if (*i_qgroups == 0)
2821                                 continue;
2822                         ret = add_qgroup_relation_item(trans, objectid,
2823                                                        *i_qgroups);
2824                         if (ret && ret != -EEXIST)
2825                                 goto out;
2826                         ret = add_qgroup_relation_item(trans, *i_qgroups,
2827                                                        objectid);
2828                         if (ret && ret != -EEXIST)
2829                                 goto out;
2830                 }
2831                 ret = 0;
2832         }
2833
2834
2835         spin_lock(&fs_info->qgroup_lock);
2836
2837         dstgroup = add_qgroup_rb(fs_info, objectid);
2838         if (IS_ERR(dstgroup)) {
2839                 ret = PTR_ERR(dstgroup);
2840                 goto unlock;
2841         }
2842
2843         if (inherit && inherit->flags & BTRFS_QGROUP_INHERIT_SET_LIMITS) {
2844                 dstgroup->lim_flags = inherit->lim.flags;
2845                 dstgroup->max_rfer = inherit->lim.max_rfer;
2846                 dstgroup->max_excl = inherit->lim.max_excl;
2847                 dstgroup->rsv_rfer = inherit->lim.rsv_rfer;
2848                 dstgroup->rsv_excl = inherit->lim.rsv_excl;
2849
2850                 ret = update_qgroup_limit_item(trans, dstgroup);
2851                 if (ret) {
2852                         fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2853                         btrfs_info(fs_info,
2854                                    "unable to update quota limit for %llu",
2855                                    dstgroup->qgroupid);
2856                         goto unlock;
2857                 }
2858         }
2859
2860         if (srcid) {
2861                 srcgroup = find_qgroup_rb(fs_info, srcid);
2862                 if (!srcgroup)
2863                         goto unlock;
2864
2865                 /*
2866                  * We call inherit after we clone the root in order to make sure
2867                  * our counts don't go crazy, so at this point the only
2868                  * difference between the two roots should be the root node.
2869                  */
2870                 level_size = fs_info->nodesize;
2871                 dstgroup->rfer = srcgroup->rfer;
2872                 dstgroup->rfer_cmpr = srcgroup->rfer_cmpr;
2873                 dstgroup->excl = level_size;
2874                 dstgroup->excl_cmpr = level_size;
2875                 srcgroup->excl = level_size;
2876                 srcgroup->excl_cmpr = level_size;
2877
2878                 /* inherit the limit info */
2879                 dstgroup->lim_flags = srcgroup->lim_flags;
2880                 dstgroup->max_rfer = srcgroup->max_rfer;
2881                 dstgroup->max_excl = srcgroup->max_excl;
2882                 dstgroup->rsv_rfer = srcgroup->rsv_rfer;
2883                 dstgroup->rsv_excl = srcgroup->rsv_excl;
2884
2885                 qgroup_dirty(fs_info, dstgroup);
2886                 qgroup_dirty(fs_info, srcgroup);
2887         }
2888
2889         if (!inherit)
2890                 goto unlock;
2891
2892         i_qgroups = (u64 *)(inherit + 1);
2893         for (i = 0; i < inherit->num_qgroups; ++i) {
2894                 if (*i_qgroups) {
2895                         ret = add_relation_rb(fs_info, objectid, *i_qgroups);
2896                         if (ret)
2897                                 goto unlock;
2898                 }
2899                 ++i_qgroups;
2900
2901                 /*
2902                  * If we're doing a snapshot, and adding the snapshot to a new
2903                  * qgroup, the numbers are guaranteed to be incorrect.
2904                  */
2905                 if (srcid)
2906                         need_rescan = true;
2907         }
2908
2909         for (i = 0; i <  inherit->num_ref_copies; ++i, i_qgroups += 2) {
2910                 struct btrfs_qgroup *src;
2911                 struct btrfs_qgroup *dst;
2912
2913                 if (!i_qgroups[0] || !i_qgroups[1])
2914                         continue;
2915
2916                 src = find_qgroup_rb(fs_info, i_qgroups[0]);
2917                 dst = find_qgroup_rb(fs_info, i_qgroups[1]);
2918
2919                 if (!src || !dst) {
2920                         ret = -EINVAL;
2921                         goto unlock;
2922                 }
2923
2924                 dst->rfer = src->rfer - level_size;
2925                 dst->rfer_cmpr = src->rfer_cmpr - level_size;
2926
2927                 /* Manually tweaking numbers certainly needs a rescan */
2928                 need_rescan = true;
2929         }
2930         for (i = 0; i <  inherit->num_excl_copies; ++i, i_qgroups += 2) {
2931                 struct btrfs_qgroup *src;
2932                 struct btrfs_qgroup *dst;
2933
2934                 if (!i_qgroups[0] || !i_qgroups[1])
2935                         continue;
2936
2937                 src = find_qgroup_rb(fs_info, i_qgroups[0]);
2938                 dst = find_qgroup_rb(fs_info, i_qgroups[1]);
2939
2940                 if (!src || !dst) {
2941                         ret = -EINVAL;
2942                         goto unlock;
2943                 }
2944
2945                 dst->excl = src->excl + level_size;
2946                 dst->excl_cmpr = src->excl_cmpr + level_size;
2947                 need_rescan = true;
2948         }
2949
2950 unlock:
2951         spin_unlock(&fs_info->qgroup_lock);
2952         if (!ret)
2953                 ret = btrfs_sysfs_add_one_qgroup(fs_info, dstgroup);
2954 out:
2955         if (!committing)
2956                 mutex_unlock(&fs_info->qgroup_ioctl_lock);
2957         if (need_rescan)
2958                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
2959         return ret;
2960 }
2961
2962 static bool qgroup_check_limits(const struct btrfs_qgroup *qg, u64 num_bytes)
2963 {
2964         if ((qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_RFER) &&
2965             qgroup_rsv_total(qg) + (s64)qg->rfer + num_bytes > qg->max_rfer)
2966                 return false;
2967
2968         if ((qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_EXCL) &&
2969             qgroup_rsv_total(qg) + (s64)qg->excl + num_bytes > qg->max_excl)
2970                 return false;
2971
2972         return true;
2973 }
2974
2975 static int qgroup_reserve(struct btrfs_root *root, u64 num_bytes, bool enforce,
2976                           enum btrfs_qgroup_rsv_type type)
2977 {
2978         struct btrfs_qgroup *qgroup;
2979         struct btrfs_fs_info *fs_info = root->fs_info;
2980         u64 ref_root = root->root_key.objectid;
2981         int ret = 0;
2982         struct ulist_node *unode;
2983         struct ulist_iterator uiter;
2984
2985         if (!is_fstree(ref_root))
2986                 return 0;
2987
2988         if (num_bytes == 0)
2989                 return 0;
2990
2991         if (test_bit(BTRFS_FS_QUOTA_OVERRIDE, &fs_info->flags) &&
2992             capable(CAP_SYS_RESOURCE))
2993                 enforce = false;
2994
2995         spin_lock(&fs_info->qgroup_lock);
2996         if (!fs_info->quota_root)
2997                 goto out;
2998
2999         qgroup = find_qgroup_rb(fs_info, ref_root);
3000         if (!qgroup)
3001                 goto out;
3002
3003         /*
3004          * in a first step, we check all affected qgroups if any limits would
3005          * be exceeded
3006          */
3007         ulist_reinit(fs_info->qgroup_ulist);
3008         ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
3009                         qgroup_to_aux(qgroup), GFP_ATOMIC);
3010         if (ret < 0)
3011                 goto out;
3012         ULIST_ITER_INIT(&uiter);
3013         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
3014                 struct btrfs_qgroup *qg;
3015                 struct btrfs_qgroup_list *glist;
3016
3017                 qg = unode_aux_to_qgroup(unode);
3018
3019                 if (enforce && !qgroup_check_limits(qg, num_bytes)) {
3020                         ret = -EDQUOT;
3021                         goto out;
3022                 }
3023
3024                 list_for_each_entry(glist, &qg->groups, next_group) {
3025                         ret = ulist_add(fs_info->qgroup_ulist,
3026                                         glist->group->qgroupid,
3027                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
3028                         if (ret < 0)
3029                                 goto out;
3030                 }
3031         }
3032         ret = 0;
3033         /*
3034          * no limits exceeded, now record the reservation into all qgroups
3035          */
3036         ULIST_ITER_INIT(&uiter);
3037         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
3038                 struct btrfs_qgroup *qg;
3039
3040                 qg = unode_aux_to_qgroup(unode);
3041
3042                 qgroup_rsv_add(fs_info, qg, num_bytes, type);
3043         }
3044
3045 out:
3046         spin_unlock(&fs_info->qgroup_lock);
3047         return ret;
3048 }
3049
3050 /*
3051  * Free @num_bytes of reserved space with @type for qgroup.  (Normally level 0
3052  * qgroup).
3053  *
3054  * Will handle all higher level qgroup too.
3055  *
3056  * NOTE: If @num_bytes is (u64)-1, this means to free all bytes of this qgroup.
3057  * This special case is only used for META_PERTRANS type.
3058  */
3059 void btrfs_qgroup_free_refroot(struct btrfs_fs_info *fs_info,
3060                                u64 ref_root, u64 num_bytes,
3061                                enum btrfs_qgroup_rsv_type type)
3062 {
3063         struct btrfs_qgroup *qgroup;
3064         struct ulist_node *unode;
3065         struct ulist_iterator uiter;
3066         int ret = 0;
3067
3068         if (!is_fstree(ref_root))
3069                 return;
3070
3071         if (num_bytes == 0)
3072                 return;
3073
3074         if (num_bytes == (u64)-1 && type != BTRFS_QGROUP_RSV_META_PERTRANS) {
3075                 WARN(1, "%s: Invalid type to free", __func__);
3076                 return;
3077         }
3078         spin_lock(&fs_info->qgroup_lock);
3079
3080         if (!fs_info->quota_root)
3081                 goto out;
3082
3083         qgroup = find_qgroup_rb(fs_info, ref_root);
3084         if (!qgroup)
3085                 goto out;
3086
3087         if (num_bytes == (u64)-1)
3088                 /*
3089                  * We're freeing all pertrans rsv, get reserved value from
3090                  * level 0 qgroup as real num_bytes to free.
3091                  */
3092                 num_bytes = qgroup->rsv.values[type];
3093
3094         ulist_reinit(fs_info->qgroup_ulist);
3095         ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
3096                         qgroup_to_aux(qgroup), GFP_ATOMIC);
3097         if (ret < 0)
3098                 goto out;
3099         ULIST_ITER_INIT(&uiter);
3100         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
3101                 struct btrfs_qgroup *qg;
3102                 struct btrfs_qgroup_list *glist;
3103
3104                 qg = unode_aux_to_qgroup(unode);
3105
3106                 qgroup_rsv_release(fs_info, qg, num_bytes, type);
3107
3108                 list_for_each_entry(glist, &qg->groups, next_group) {
3109                         ret = ulist_add(fs_info->qgroup_ulist,
3110                                         glist->group->qgroupid,
3111                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
3112                         if (ret < 0)
3113                                 goto out;
3114                 }
3115         }
3116
3117 out:
3118         spin_unlock(&fs_info->qgroup_lock);
3119 }
3120
3121 /*
3122  * Check if the leaf is the last leaf. Which means all node pointers
3123  * are at their last position.
3124  */
3125 static bool is_last_leaf(struct btrfs_path *path)
3126 {
3127         int i;
3128
3129         for (i = 1; i < BTRFS_MAX_LEVEL && path->nodes[i]; i++) {
3130                 if (path->slots[i] != btrfs_header_nritems(path->nodes[i]) - 1)
3131                         return false;
3132         }
3133         return true;
3134 }
3135
3136 /*
3137  * returns < 0 on error, 0 when more leafs are to be scanned.
3138  * returns 1 when done.
3139  */
3140 static int qgroup_rescan_leaf(struct btrfs_trans_handle *trans,
3141                               struct btrfs_path *path)
3142 {
3143         struct btrfs_fs_info *fs_info = trans->fs_info;
3144         struct btrfs_key found;
3145         struct extent_buffer *scratch_leaf = NULL;
3146         struct ulist *roots = NULL;
3147         u64 num_bytes;
3148         bool done;
3149         int slot;
3150         int ret;
3151
3152         mutex_lock(&fs_info->qgroup_rescan_lock);
3153         ret = btrfs_search_slot_for_read(fs_info->extent_root,
3154                                          &fs_info->qgroup_rescan_progress,
3155                                          path, 1, 0);
3156
3157         btrfs_debug(fs_info,
3158                 "current progress key (%llu %u %llu), search_slot ret %d",
3159                 fs_info->qgroup_rescan_progress.objectid,
3160                 fs_info->qgroup_rescan_progress.type,
3161                 fs_info->qgroup_rescan_progress.offset, ret);
3162
3163         if (ret) {
3164                 /*
3165                  * The rescan is about to end, we will not be scanning any
3166                  * further blocks. We cannot unset the RESCAN flag here, because
3167                  * we want to commit the transaction if everything went well.
3168                  * To make the live accounting work in this phase, we set our
3169                  * scan progress pointer such that every real extent objectid
3170                  * will be smaller.
3171                  */
3172                 fs_info->qgroup_rescan_progress.objectid = (u64)-1;
3173                 btrfs_release_path(path);
3174                 mutex_unlock(&fs_info->qgroup_rescan_lock);
3175                 return ret;
3176         }
3177         done = is_last_leaf(path);
3178
3179         btrfs_item_key_to_cpu(path->nodes[0], &found,
3180                               btrfs_header_nritems(path->nodes[0]) - 1);
3181         fs_info->qgroup_rescan_progress.objectid = found.objectid + 1;
3182
3183         scratch_leaf = btrfs_clone_extent_buffer(path->nodes[0]);
3184         if (!scratch_leaf) {
3185                 ret = -ENOMEM;
3186                 mutex_unlock(&fs_info->qgroup_rescan_lock);
3187                 goto out;
3188         }
3189         slot = path->slots[0];
3190         btrfs_release_path(path);
3191         mutex_unlock(&fs_info->qgroup_rescan_lock);
3192
3193         for (; slot < btrfs_header_nritems(scratch_leaf); ++slot) {
3194                 btrfs_item_key_to_cpu(scratch_leaf, &found, slot);
3195                 if (found.type != BTRFS_EXTENT_ITEM_KEY &&
3196                     found.type != BTRFS_METADATA_ITEM_KEY)
3197                         continue;
3198                 if (found.type == BTRFS_METADATA_ITEM_KEY)
3199                         num_bytes = fs_info->nodesize;
3200                 else
3201                         num_bytes = found.offset;
3202
3203                 ret = btrfs_find_all_roots(NULL, fs_info, found.objectid, 0,
3204                                            &roots, false, false);
3205                 if (ret < 0)
3206                         goto out;
3207                 /* For rescan, just pass old_roots as NULL */
3208                 ret = btrfs_qgroup_account_extent(trans, found.objectid,
3209                                                   num_bytes, NULL, roots);
3210                 if (ret < 0)
3211                         goto out;
3212         }
3213 out:
3214         if (scratch_leaf)
3215                 free_extent_buffer(scratch_leaf);
3216
3217         if (done && !ret) {
3218                 ret = 1;
3219                 fs_info->qgroup_rescan_progress.objectid = (u64)-1;
3220         }
3221         return ret;
3222 }
3223
3224 static bool rescan_should_stop(struct btrfs_fs_info *fs_info)
3225 {
3226         return btrfs_fs_closing(fs_info) ||
3227                 test_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
3228 }
3229
3230 static void btrfs_qgroup_rescan_worker(struct btrfs_work *work)
3231 {
3232         struct btrfs_fs_info *fs_info = container_of(work, struct btrfs_fs_info,
3233                                                      qgroup_rescan_work);
3234         struct btrfs_path *path;
3235         struct btrfs_trans_handle *trans = NULL;
3236         int err = -ENOMEM;
3237         int ret = 0;
3238         bool stopped = false;
3239
3240         path = btrfs_alloc_path();
3241         if (!path)
3242                 goto out;
3243         /*
3244          * Rescan should only search for commit root, and any later difference
3245          * should be recorded by qgroup
3246          */
3247         path->search_commit_root = 1;
3248         path->skip_locking = 1;
3249
3250         err = 0;
3251         while (!err && !(stopped = rescan_should_stop(fs_info))) {
3252                 trans = btrfs_start_transaction(fs_info->fs_root, 0);
3253                 if (IS_ERR(trans)) {
3254                         err = PTR_ERR(trans);
3255                         break;
3256                 }
3257                 if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags)) {
3258                         err = -EINTR;
3259                 } else {
3260                         err = qgroup_rescan_leaf(trans, path);
3261                 }
3262                 if (err > 0)
3263                         btrfs_commit_transaction(trans);
3264                 else
3265                         btrfs_end_transaction(trans);
3266         }
3267
3268 out:
3269         btrfs_free_path(path);
3270
3271         mutex_lock(&fs_info->qgroup_rescan_lock);
3272         if (err > 0 &&
3273             fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT) {
3274                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
3275         } else if (err < 0) {
3276                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
3277         }
3278         mutex_unlock(&fs_info->qgroup_rescan_lock);
3279
3280         /*
3281          * only update status, since the previous part has already updated the
3282          * qgroup info.
3283          */
3284         trans = btrfs_start_transaction(fs_info->quota_root, 1);
3285         if (IS_ERR(trans)) {
3286                 err = PTR_ERR(trans);
3287                 trans = NULL;
3288                 btrfs_err(fs_info,
3289                           "fail to start transaction for status update: %d",
3290                           err);
3291         }
3292
3293         mutex_lock(&fs_info->qgroup_rescan_lock);
3294         if (!stopped)
3295                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3296         if (trans) {
3297                 ret = update_qgroup_status_item(trans);
3298                 if (ret < 0) {
3299                         err = ret;
3300                         btrfs_err(fs_info, "fail to update qgroup status: %d",
3301                                   err);
3302                 }
3303         }
3304         fs_info->qgroup_rescan_running = false;
3305         complete_all(&fs_info->qgroup_rescan_completion);
3306         mutex_unlock(&fs_info->qgroup_rescan_lock);
3307
3308         if (!trans)
3309                 return;
3310
3311         btrfs_end_transaction(trans);
3312
3313         if (stopped) {
3314                 btrfs_info(fs_info, "qgroup scan paused");
3315         } else if (err >= 0) {
3316                 btrfs_info(fs_info, "qgroup scan completed%s",
3317                         err > 0 ? " (inconsistency flag cleared)" : "");
3318         } else {
3319                 btrfs_err(fs_info, "qgroup scan failed with %d", err);
3320         }
3321 }
3322
3323 /*
3324  * Checks that (a) no rescan is running and (b) quota is enabled. Allocates all
3325  * memory required for the rescan context.
3326  */
3327 static int
3328 qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid,
3329                    int init_flags)
3330 {
3331         int ret = 0;
3332
3333         if (!init_flags) {
3334                 /* we're resuming qgroup rescan at mount time */
3335                 if (!(fs_info->qgroup_flags &
3336                       BTRFS_QGROUP_STATUS_FLAG_RESCAN)) {
3337                         btrfs_warn(fs_info,
3338                         "qgroup rescan init failed, qgroup rescan is not queued");
3339                         ret = -EINVAL;
3340                 } else if (!(fs_info->qgroup_flags &
3341                              BTRFS_QGROUP_STATUS_FLAG_ON)) {
3342                         btrfs_warn(fs_info,
3343                         "qgroup rescan init failed, qgroup is not enabled");
3344                         ret = -EINVAL;
3345                 }
3346
3347                 if (ret)
3348                         return ret;
3349         }
3350
3351         mutex_lock(&fs_info->qgroup_rescan_lock);
3352
3353         if (init_flags) {
3354                 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3355                         btrfs_warn(fs_info,
3356                                    "qgroup rescan is already in progress");
3357                         ret = -EINPROGRESS;
3358                 } else if (!(fs_info->qgroup_flags &
3359                              BTRFS_QGROUP_STATUS_FLAG_ON)) {
3360                         btrfs_warn(fs_info,
3361                         "qgroup rescan init failed, qgroup is not enabled");
3362                         ret = -EINVAL;
3363                 }
3364
3365                 if (ret) {
3366                         mutex_unlock(&fs_info->qgroup_rescan_lock);
3367                         return ret;
3368                 }
3369                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3370         }
3371
3372         memset(&fs_info->qgroup_rescan_progress, 0,
3373                 sizeof(fs_info->qgroup_rescan_progress));
3374         fs_info->qgroup_rescan_progress.objectid = progress_objectid;
3375         init_completion(&fs_info->qgroup_rescan_completion);
3376         mutex_unlock(&fs_info->qgroup_rescan_lock);
3377
3378         btrfs_init_work(&fs_info->qgroup_rescan_work,
3379                         btrfs_qgroup_rescan_worker, NULL, NULL);
3380         return 0;
3381 }
3382
3383 static void
3384 qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info)
3385 {
3386         struct rb_node *n;
3387         struct btrfs_qgroup *qgroup;
3388
3389         spin_lock(&fs_info->qgroup_lock);
3390         /* clear all current qgroup tracking information */
3391         for (n = rb_first(&fs_info->qgroup_tree); n; n = rb_next(n)) {
3392                 qgroup = rb_entry(n, struct btrfs_qgroup, node);
3393                 qgroup->rfer = 0;
3394                 qgroup->rfer_cmpr = 0;
3395                 qgroup->excl = 0;
3396                 qgroup->excl_cmpr = 0;
3397                 qgroup_dirty(fs_info, qgroup);
3398         }
3399         spin_unlock(&fs_info->qgroup_lock);
3400 }
3401
3402 int
3403 btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info)
3404 {
3405         int ret = 0;
3406         struct btrfs_trans_handle *trans;
3407
3408         ret = qgroup_rescan_init(fs_info, 0, 1);
3409         if (ret)
3410                 return ret;
3411
3412         /*
3413          * We have set the rescan_progress to 0, which means no more
3414          * delayed refs will be accounted by btrfs_qgroup_account_ref.
3415          * However, btrfs_qgroup_account_ref may be right after its call
3416          * to btrfs_find_all_roots, in which case it would still do the
3417          * accounting.
3418          * To solve this, we're committing the transaction, which will
3419          * ensure we run all delayed refs and only after that, we are
3420          * going to clear all tracking information for a clean start.
3421          */
3422
3423         trans = btrfs_join_transaction(fs_info->fs_root);
3424         if (IS_ERR(trans)) {
3425                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3426                 return PTR_ERR(trans);
3427         }
3428         ret = btrfs_commit_transaction(trans);
3429         if (ret) {
3430                 fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
3431                 return ret;
3432         }
3433
3434         qgroup_rescan_zero_tracking(fs_info);
3435
3436         mutex_lock(&fs_info->qgroup_rescan_lock);
3437         fs_info->qgroup_rescan_running = true;
3438         btrfs_queue_work(fs_info->qgroup_rescan_workers,
3439                          &fs_info->qgroup_rescan_work);
3440         mutex_unlock(&fs_info->qgroup_rescan_lock);
3441
3442         return 0;
3443 }
3444
3445 int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info,
3446                                      bool interruptible)
3447 {
3448         int running;
3449         int ret = 0;
3450
3451         mutex_lock(&fs_info->qgroup_rescan_lock);
3452         running = fs_info->qgroup_rescan_running;
3453         mutex_unlock(&fs_info->qgroup_rescan_lock);
3454
3455         if (!running)
3456                 return 0;
3457
3458         if (interruptible)
3459                 ret = wait_for_completion_interruptible(
3460                                         &fs_info->qgroup_rescan_completion);
3461         else
3462                 wait_for_completion(&fs_info->qgroup_rescan_completion);
3463
3464         return ret;
3465 }
3466
3467 /*
3468  * this is only called from open_ctree where we're still single threaded, thus
3469  * locking is omitted here.
3470  */
3471 void
3472 btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info)
3473 {
3474         if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3475                 mutex_lock(&fs_info->qgroup_rescan_lock);
3476                 fs_info->qgroup_rescan_running = true;
3477                 btrfs_queue_work(fs_info->qgroup_rescan_workers,
3478                                  &fs_info->qgroup_rescan_work);
3479                 mutex_unlock(&fs_info->qgroup_rescan_lock);
3480         }
3481 }
3482
3483 #define rbtree_iterate_from_safe(node, next, start)                             \
3484        for (node = start; node && ({ next = rb_next(node); 1;}); node = next)
3485
3486 static int qgroup_unreserve_range(struct btrfs_inode *inode,
3487                                   struct extent_changeset *reserved, u64 start,
3488                                   u64 len)
3489 {
3490         struct rb_node *node;
3491         struct rb_node *next;
3492         struct ulist_node *entry;
3493         int ret = 0;
3494
3495         node = reserved->range_changed.root.rb_node;
3496         if (!node)
3497                 return 0;
3498         while (node) {
3499                 entry = rb_entry(node, struct ulist_node, rb_node);
3500                 if (entry->val < start)
3501                         node = node->rb_right;
3502                 else
3503                         node = node->rb_left;
3504         }
3505
3506         if (entry->val > start && rb_prev(&entry->rb_node))
3507                 entry = rb_entry(rb_prev(&entry->rb_node), struct ulist_node,
3508                                  rb_node);
3509
3510         rbtree_iterate_from_safe(node, next, &entry->rb_node) {
3511                 u64 entry_start;
3512                 u64 entry_end;
3513                 u64 entry_len;
3514                 int clear_ret;
3515
3516                 entry = rb_entry(node, struct ulist_node, rb_node);
3517                 entry_start = entry->val;
3518                 entry_end = entry->aux;
3519                 entry_len = entry_end - entry_start + 1;
3520
3521                 if (entry_start >= start + len)
3522                         break;
3523                 if (entry_start + entry_len <= start)
3524                         continue;
3525                 /*
3526                  * Now the entry is in [start, start + len), revert the
3527                  * EXTENT_QGROUP_RESERVED bit.
3528                  */
3529                 clear_ret = clear_extent_bits(&inode->io_tree, entry_start,
3530                                               entry_end, EXTENT_QGROUP_RESERVED);
3531                 if (!ret && clear_ret < 0)
3532                         ret = clear_ret;
3533
3534                 ulist_del(&reserved->range_changed, entry->val, entry->aux);
3535                 if (likely(reserved->bytes_changed >= entry_len)) {
3536                         reserved->bytes_changed -= entry_len;
3537                 } else {
3538                         WARN_ON(1);
3539                         reserved->bytes_changed = 0;
3540                 }
3541         }
3542
3543         return ret;
3544 }
3545
3546 /*
3547  * Try to free some space for qgroup.
3548  *
3549  * For qgroup, there are only 3 ways to free qgroup space:
3550  * - Flush nodatacow write
3551  *   Any nodatacow write will free its reserved data space at run_delalloc_range().
3552  *   In theory, we should only flush nodatacow inodes, but it's not yet
3553  *   possible, so we need to flush the whole root.
3554  *
3555  * - Wait for ordered extents
3556  *   When ordered extents are finished, their reserved metadata is finally
3557  *   converted to per_trans status, which can be freed by later commit
3558  *   transaction.
3559  *
3560  * - Commit transaction
3561  *   This would free the meta_per_trans space.
3562  *   In theory this shouldn't provide much space, but any more qgroup space
3563  *   is needed.
3564  */
3565 static int try_flush_qgroup(struct btrfs_root *root)
3566 {
3567         struct btrfs_trans_handle *trans;
3568         int ret;
3569
3570         /* Can't hold an open transaction or we run the risk of deadlocking. */
3571         ASSERT(current->journal_info == NULL);
3572         if (WARN_ON(current->journal_info))
3573                 return 0;
3574
3575         /*
3576          * We don't want to run flush again and again, so if there is a running
3577          * one, we won't try to start a new flush, but exit directly.
3578          */
3579         if (test_and_set_bit(BTRFS_ROOT_QGROUP_FLUSHING, &root->state)) {
3580                 wait_event(root->qgroup_flush_wait,
3581                         !test_bit(BTRFS_ROOT_QGROUP_FLUSHING, &root->state));
3582                 return 0;
3583         }
3584
3585         ret = btrfs_start_delalloc_snapshot(root, true);
3586         if (ret < 0)
3587                 goto out;
3588         btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
3589
3590         trans = btrfs_join_transaction(root);
3591         if (IS_ERR(trans)) {
3592                 ret = PTR_ERR(trans);
3593                 goto out;
3594         }
3595
3596         ret = btrfs_commit_transaction(trans);
3597 out:
3598         clear_bit(BTRFS_ROOT_QGROUP_FLUSHING, &root->state);
3599         wake_up(&root->qgroup_flush_wait);
3600         return ret;
3601 }
3602
3603 static int qgroup_reserve_data(struct btrfs_inode *inode,
3604                         struct extent_changeset **reserved_ret, u64 start,
3605                         u64 len)
3606 {
3607         struct btrfs_root *root = inode->root;
3608         struct extent_changeset *reserved;
3609         bool new_reserved = false;
3610         u64 orig_reserved;
3611         u64 to_reserve;
3612         int ret;
3613
3614         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &root->fs_info->flags) ||
3615             !is_fstree(root->root_key.objectid) || len == 0)
3616                 return 0;
3617
3618         /* @reserved parameter is mandatory for qgroup */
3619         if (WARN_ON(!reserved_ret))
3620                 return -EINVAL;
3621         if (!*reserved_ret) {
3622                 new_reserved = true;
3623                 *reserved_ret = extent_changeset_alloc();
3624                 if (!*reserved_ret)
3625                         return -ENOMEM;
3626         }
3627         reserved = *reserved_ret;
3628         /* Record already reserved space */
3629         orig_reserved = reserved->bytes_changed;
3630         ret = set_record_extent_bits(&inode->io_tree, start,
3631                         start + len -1, EXTENT_QGROUP_RESERVED, reserved);
3632
3633         /* Newly reserved space */
3634         to_reserve = reserved->bytes_changed - orig_reserved;
3635         trace_btrfs_qgroup_reserve_data(&inode->vfs_inode, start, len,
3636                                         to_reserve, QGROUP_RESERVE);
3637         if (ret < 0)
3638                 goto out;
3639         ret = qgroup_reserve(root, to_reserve, true, BTRFS_QGROUP_RSV_DATA);
3640         if (ret < 0)
3641                 goto cleanup;
3642
3643         return ret;
3644
3645 cleanup:
3646         qgroup_unreserve_range(inode, reserved, start, len);
3647 out:
3648         if (new_reserved) {
3649                 extent_changeset_free(reserved);
3650                 *reserved_ret = NULL;
3651         }
3652         return ret;
3653 }
3654
3655 /*
3656  * Reserve qgroup space for range [start, start + len).
3657  *
3658  * This function will either reserve space from related qgroups or do nothing
3659  * if the range is already reserved.
3660  *
3661  * Return 0 for successful reservation
3662  * Return <0 for error (including -EQUOT)
3663  *
3664  * NOTE: This function may sleep for memory allocation, dirty page flushing and
3665  *       commit transaction. So caller should not hold any dirty page locked.
3666  */
3667 int btrfs_qgroup_reserve_data(struct btrfs_inode *inode,
3668                         struct extent_changeset **reserved_ret, u64 start,
3669                         u64 len)
3670 {
3671         int ret;
3672
3673         ret = qgroup_reserve_data(inode, reserved_ret, start, len);
3674         if (ret <= 0 && ret != -EDQUOT)
3675                 return ret;
3676
3677         ret = try_flush_qgroup(inode->root);
3678         if (ret < 0)
3679                 return ret;
3680         return qgroup_reserve_data(inode, reserved_ret, start, len);
3681 }
3682
3683 /* Free ranges specified by @reserved, normally in error path */
3684 static int qgroup_free_reserved_data(struct btrfs_inode *inode,
3685                         struct extent_changeset *reserved, u64 start, u64 len)
3686 {
3687         struct btrfs_root *root = inode->root;
3688         struct ulist_node *unode;
3689         struct ulist_iterator uiter;
3690         struct extent_changeset changeset;
3691         int freed = 0;
3692         int ret;
3693
3694         extent_changeset_init(&changeset);
3695         len = round_up(start + len, root->fs_info->sectorsize);
3696         start = round_down(start, root->fs_info->sectorsize);
3697
3698         ULIST_ITER_INIT(&uiter);
3699         while ((unode = ulist_next(&reserved->range_changed, &uiter))) {
3700                 u64 range_start = unode->val;
3701                 /* unode->aux is the inclusive end */
3702                 u64 range_len = unode->aux - range_start + 1;
3703                 u64 free_start;
3704                 u64 free_len;
3705
3706                 extent_changeset_release(&changeset);
3707
3708                 /* Only free range in range [start, start + len) */
3709                 if (range_start >= start + len ||
3710                     range_start + range_len <= start)
3711                         continue;
3712                 free_start = max(range_start, start);
3713                 free_len = min(start + len, range_start + range_len) -
3714                            free_start;
3715                 /*
3716                  * TODO: To also modify reserved->ranges_reserved to reflect
3717                  * the modification.
3718                  *
3719                  * However as long as we free qgroup reserved according to
3720                  * EXTENT_QGROUP_RESERVED, we won't double free.
3721                  * So not need to rush.
3722                  */
3723                 ret = clear_record_extent_bits(&inode->io_tree, free_start,
3724                                 free_start + free_len - 1,
3725                                 EXTENT_QGROUP_RESERVED, &changeset);
3726                 if (ret < 0)
3727                         goto out;
3728                 freed += changeset.bytes_changed;
3729         }
3730         btrfs_qgroup_free_refroot(root->fs_info, root->root_key.objectid, freed,
3731                                   BTRFS_QGROUP_RSV_DATA);
3732         ret = freed;
3733 out:
3734         extent_changeset_release(&changeset);
3735         return ret;
3736 }
3737
3738 static int __btrfs_qgroup_release_data(struct btrfs_inode *inode,
3739                         struct extent_changeset *reserved, u64 start, u64 len,
3740                         int free)
3741 {
3742         struct extent_changeset changeset;
3743         int trace_op = QGROUP_RELEASE;
3744         int ret;
3745
3746         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &inode->root->fs_info->flags))
3747                 return 0;
3748
3749         /* In release case, we shouldn't have @reserved */
3750         WARN_ON(!free && reserved);
3751         if (free && reserved)
3752                 return qgroup_free_reserved_data(inode, reserved, start, len);
3753         extent_changeset_init(&changeset);
3754         ret = clear_record_extent_bits(&inode->io_tree, start, start + len -1,
3755                                        EXTENT_QGROUP_RESERVED, &changeset);
3756         if (ret < 0)
3757                 goto out;
3758
3759         if (free)
3760                 trace_op = QGROUP_FREE;
3761         trace_btrfs_qgroup_release_data(&inode->vfs_inode, start, len,
3762                                         changeset.bytes_changed, trace_op);
3763         if (free)
3764                 btrfs_qgroup_free_refroot(inode->root->fs_info,
3765                                 inode->root->root_key.objectid,
3766                                 changeset.bytes_changed, BTRFS_QGROUP_RSV_DATA);
3767         ret = changeset.bytes_changed;
3768 out:
3769         extent_changeset_release(&changeset);
3770         return ret;
3771 }
3772
3773 /*
3774  * Free a reserved space range from io_tree and related qgroups
3775  *
3776  * Should be called when a range of pages get invalidated before reaching disk.
3777  * Or for error cleanup case.
3778  * if @reserved is given, only reserved range in [@start, @start + @len) will
3779  * be freed.
3780  *
3781  * For data written to disk, use btrfs_qgroup_release_data().
3782  *
3783  * NOTE: This function may sleep for memory allocation.
3784  */
3785 int btrfs_qgroup_free_data(struct btrfs_inode *inode,
3786                         struct extent_changeset *reserved, u64 start, u64 len)
3787 {
3788         return __btrfs_qgroup_release_data(inode, reserved, start, len, 1);
3789 }
3790
3791 /*
3792  * Release a reserved space range from io_tree only.
3793  *
3794  * Should be called when a range of pages get written to disk and corresponding
3795  * FILE_EXTENT is inserted into corresponding root.
3796  *
3797  * Since new qgroup accounting framework will only update qgroup numbers at
3798  * commit_transaction() time, its reserved space shouldn't be freed from
3799  * related qgroups.
3800  *
3801  * But we should release the range from io_tree, to allow further write to be
3802  * COWed.
3803  *
3804  * NOTE: This function may sleep for memory allocation.
3805  */
3806 int btrfs_qgroup_release_data(struct btrfs_inode *inode, u64 start, u64 len)
3807 {
3808         return __btrfs_qgroup_release_data(inode, NULL, start, len, 0);
3809 }
3810
3811 static void add_root_meta_rsv(struct btrfs_root *root, int num_bytes,
3812                               enum btrfs_qgroup_rsv_type type)
3813 {
3814         if (type != BTRFS_QGROUP_RSV_META_PREALLOC &&
3815             type != BTRFS_QGROUP_RSV_META_PERTRANS)
3816                 return;
3817         if (num_bytes == 0)
3818                 return;
3819
3820         spin_lock(&root->qgroup_meta_rsv_lock);
3821         if (type == BTRFS_QGROUP_RSV_META_PREALLOC)
3822                 root->qgroup_meta_rsv_prealloc += num_bytes;
3823         else
3824                 root->qgroup_meta_rsv_pertrans += num_bytes;
3825         spin_unlock(&root->qgroup_meta_rsv_lock);
3826 }
3827
3828 static int sub_root_meta_rsv(struct btrfs_root *root, int num_bytes,
3829                              enum btrfs_qgroup_rsv_type type)
3830 {
3831         if (type != BTRFS_QGROUP_RSV_META_PREALLOC &&
3832             type != BTRFS_QGROUP_RSV_META_PERTRANS)
3833                 return 0;
3834         if (num_bytes == 0)
3835                 return 0;
3836
3837         spin_lock(&root->qgroup_meta_rsv_lock);
3838         if (type == BTRFS_QGROUP_RSV_META_PREALLOC) {
3839                 num_bytes = min_t(u64, root->qgroup_meta_rsv_prealloc,
3840                                   num_bytes);
3841                 root->qgroup_meta_rsv_prealloc -= num_bytes;
3842         } else {
3843                 num_bytes = min_t(u64, root->qgroup_meta_rsv_pertrans,
3844                                   num_bytes);
3845                 root->qgroup_meta_rsv_pertrans -= num_bytes;
3846         }
3847         spin_unlock(&root->qgroup_meta_rsv_lock);
3848         return num_bytes;
3849 }
3850
3851 int btrfs_qgroup_reserve_meta(struct btrfs_root *root, int num_bytes,
3852                               enum btrfs_qgroup_rsv_type type, bool enforce)
3853 {
3854         struct btrfs_fs_info *fs_info = root->fs_info;
3855         int ret;
3856
3857         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3858             !is_fstree(root->root_key.objectid) || num_bytes == 0)
3859                 return 0;
3860
3861         BUG_ON(num_bytes != round_down(num_bytes, fs_info->nodesize));
3862         trace_qgroup_meta_reserve(root, (s64)num_bytes, type);
3863         ret = qgroup_reserve(root, num_bytes, enforce, type);
3864         if (ret < 0)
3865                 return ret;
3866         /*
3867          * Record what we have reserved into root.
3868          *
3869          * To avoid quota disabled->enabled underflow.
3870          * In that case, we may try to free space we haven't reserved
3871          * (since quota was disabled), so record what we reserved into root.
3872          * And ensure later release won't underflow this number.
3873          */
3874         add_root_meta_rsv(root, num_bytes, type);
3875         return ret;
3876 }
3877
3878 int __btrfs_qgroup_reserve_meta(struct btrfs_root *root, int num_bytes,
3879                                 enum btrfs_qgroup_rsv_type type, bool enforce)
3880 {
3881         int ret;
3882
3883         ret = btrfs_qgroup_reserve_meta(root, num_bytes, type, enforce);
3884         if (ret <= 0 && ret != -EDQUOT)
3885                 return ret;
3886
3887         ret = try_flush_qgroup(root);
3888         if (ret < 0)
3889                 return ret;
3890         return btrfs_qgroup_reserve_meta(root, num_bytes, type, enforce);
3891 }
3892
3893 void btrfs_qgroup_free_meta_all_pertrans(struct btrfs_root *root)
3894 {
3895         struct btrfs_fs_info *fs_info = root->fs_info;
3896
3897         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3898             !is_fstree(root->root_key.objectid))
3899                 return;
3900
3901         /* TODO: Update trace point to handle such free */
3902         trace_qgroup_meta_free_all_pertrans(root);
3903         /* Special value -1 means to free all reserved space */
3904         btrfs_qgroup_free_refroot(fs_info, root->root_key.objectid, (u64)-1,
3905                                   BTRFS_QGROUP_RSV_META_PERTRANS);
3906 }
3907
3908 void __btrfs_qgroup_free_meta(struct btrfs_root *root, int num_bytes,
3909                               enum btrfs_qgroup_rsv_type type)
3910 {
3911         struct btrfs_fs_info *fs_info = root->fs_info;
3912
3913         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3914             !is_fstree(root->root_key.objectid))
3915                 return;
3916
3917         /*
3918          * reservation for META_PREALLOC can happen before quota is enabled,
3919          * which can lead to underflow.
3920          * Here ensure we will only free what we really have reserved.
3921          */
3922         num_bytes = sub_root_meta_rsv(root, num_bytes, type);
3923         BUG_ON(num_bytes != round_down(num_bytes, fs_info->nodesize));
3924         trace_qgroup_meta_reserve(root, -(s64)num_bytes, type);
3925         btrfs_qgroup_free_refroot(fs_info, root->root_key.objectid,
3926                                   num_bytes, type);
3927 }
3928
3929 static void qgroup_convert_meta(struct btrfs_fs_info *fs_info, u64 ref_root,
3930                                 int num_bytes)
3931 {
3932         struct btrfs_qgroup *qgroup;
3933         struct ulist_node *unode;
3934         struct ulist_iterator uiter;
3935         int ret = 0;
3936
3937         if (num_bytes == 0)
3938                 return;
3939         if (!fs_info->quota_root)
3940                 return;
3941
3942         spin_lock(&fs_info->qgroup_lock);
3943         qgroup = find_qgroup_rb(fs_info, ref_root);
3944         if (!qgroup)
3945                 goto out;
3946         ulist_reinit(fs_info->qgroup_ulist);
3947         ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
3948                        qgroup_to_aux(qgroup), GFP_ATOMIC);
3949         if (ret < 0)
3950                 goto out;
3951         ULIST_ITER_INIT(&uiter);
3952         while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
3953                 struct btrfs_qgroup *qg;
3954                 struct btrfs_qgroup_list *glist;
3955
3956                 qg = unode_aux_to_qgroup(unode);
3957
3958                 qgroup_rsv_release(fs_info, qg, num_bytes,
3959                                 BTRFS_QGROUP_RSV_META_PREALLOC);
3960                 qgroup_rsv_add(fs_info, qg, num_bytes,
3961                                 BTRFS_QGROUP_RSV_META_PERTRANS);
3962                 list_for_each_entry(glist, &qg->groups, next_group) {
3963                         ret = ulist_add(fs_info->qgroup_ulist,
3964                                         glist->group->qgroupid,
3965                                         qgroup_to_aux(glist->group), GFP_ATOMIC);
3966                         if (ret < 0)
3967                                 goto out;
3968                 }
3969         }
3970 out:
3971         spin_unlock(&fs_info->qgroup_lock);
3972 }
3973
3974 void btrfs_qgroup_convert_reserved_meta(struct btrfs_root *root, int num_bytes)
3975 {
3976         struct btrfs_fs_info *fs_info = root->fs_info;
3977
3978         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) ||
3979             !is_fstree(root->root_key.objectid))
3980                 return;
3981         /* Same as btrfs_qgroup_free_meta_prealloc() */
3982         num_bytes = sub_root_meta_rsv(root, num_bytes,
3983                                       BTRFS_QGROUP_RSV_META_PREALLOC);
3984         trace_qgroup_meta_convert(root, num_bytes);
3985         qgroup_convert_meta(fs_info, root->root_key.objectid, num_bytes);
3986 }
3987
3988 /*
3989  * Check qgroup reserved space leaking, normally at destroy inode
3990  * time
3991  */
3992 void btrfs_qgroup_check_reserved_leak(struct btrfs_inode *inode)
3993 {
3994         struct extent_changeset changeset;
3995         struct ulist_node *unode;
3996         struct ulist_iterator iter;
3997         int ret;
3998
3999         extent_changeset_init(&changeset);
4000         ret = clear_record_extent_bits(&inode->io_tree, 0, (u64)-1,
4001                         EXTENT_QGROUP_RESERVED, &changeset);
4002
4003         WARN_ON(ret < 0);
4004         if (WARN_ON(changeset.bytes_changed)) {
4005                 ULIST_ITER_INIT(&iter);
4006                 while ((unode = ulist_next(&changeset.range_changed, &iter))) {
4007                         btrfs_warn(inode->root->fs_info,
4008                 "leaking qgroup reserved space, ino: %llu, start: %llu, end: %llu",
4009                                 btrfs_ino(inode), unode->val, unode->aux);
4010                 }
4011                 btrfs_qgroup_free_refroot(inode->root->fs_info,
4012                                 inode->root->root_key.objectid,
4013                                 changeset.bytes_changed, BTRFS_QGROUP_RSV_DATA);
4014
4015         }
4016         extent_changeset_release(&changeset);
4017 }
4018
4019 void btrfs_qgroup_init_swapped_blocks(
4020         struct btrfs_qgroup_swapped_blocks *swapped_blocks)
4021 {
4022         int i;
4023
4024         spin_lock_init(&swapped_blocks->lock);
4025         for (i = 0; i < BTRFS_MAX_LEVEL; i++)
4026                 swapped_blocks->blocks[i] = RB_ROOT;
4027         swapped_blocks->swapped = false;
4028 }
4029
4030 /*
4031  * Delete all swapped blocks record of @root.
4032  * Every record here means we skipped a full subtree scan for qgroup.
4033  *
4034  * Gets called when committing one transaction.
4035  */
4036 void btrfs_qgroup_clean_swapped_blocks(struct btrfs_root *root)
4037 {
4038         struct btrfs_qgroup_swapped_blocks *swapped_blocks;
4039         int i;
4040
4041         swapped_blocks = &root->swapped_blocks;
4042
4043         spin_lock(&swapped_blocks->lock);
4044         if (!swapped_blocks->swapped)
4045                 goto out;
4046         for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
4047                 struct rb_root *cur_root = &swapped_blocks->blocks[i];
4048                 struct btrfs_qgroup_swapped_block *entry;
4049                 struct btrfs_qgroup_swapped_block *next;
4050
4051                 rbtree_postorder_for_each_entry_safe(entry, next, cur_root,
4052                                                      node)
4053                         kfree(entry);
4054                 swapped_blocks->blocks[i] = RB_ROOT;
4055         }
4056         swapped_blocks->swapped = false;
4057 out:
4058         spin_unlock(&swapped_blocks->lock);
4059 }
4060
4061 /*
4062  * Add subtree roots record into @subvol_root.
4063  *
4064  * @subvol_root:        tree root of the subvolume tree get swapped
4065  * @bg:                 block group under balance
4066  * @subvol_parent/slot: pointer to the subtree root in subvolume tree
4067  * @reloc_parent/slot:  pointer to the subtree root in reloc tree
4068  *                      BOTH POINTERS ARE BEFORE TREE SWAP
4069  * @last_snapshot:      last snapshot generation of the subvolume tree
4070  */
4071 int btrfs_qgroup_add_swapped_blocks(struct btrfs_trans_handle *trans,
4072                 struct btrfs_root *subvol_root,
4073                 struct btrfs_block_group *bg,
4074                 struct extent_buffer *subvol_parent, int subvol_slot,
4075                 struct extent_buffer *reloc_parent, int reloc_slot,
4076                 u64 last_snapshot)
4077 {
4078         struct btrfs_fs_info *fs_info = subvol_root->fs_info;
4079         struct btrfs_qgroup_swapped_blocks *blocks = &subvol_root->swapped_blocks;
4080         struct btrfs_qgroup_swapped_block *block;
4081         struct rb_node **cur;
4082         struct rb_node *parent = NULL;
4083         int level = btrfs_header_level(subvol_parent) - 1;
4084         int ret = 0;
4085
4086         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
4087                 return 0;
4088
4089         if (btrfs_node_ptr_generation(subvol_parent, subvol_slot) >
4090             btrfs_node_ptr_generation(reloc_parent, reloc_slot)) {
4091                 btrfs_err_rl(fs_info,
4092                 "%s: bad parameter order, subvol_gen=%llu reloc_gen=%llu",
4093                         __func__,
4094                         btrfs_node_ptr_generation(subvol_parent, subvol_slot),
4095                         btrfs_node_ptr_generation(reloc_parent, reloc_slot));
4096                 return -EUCLEAN;
4097         }
4098
4099         block = kmalloc(sizeof(*block), GFP_NOFS);
4100         if (!block) {
4101                 ret = -ENOMEM;
4102                 goto out;
4103         }
4104
4105         /*
4106          * @reloc_parent/slot is still before swap, while @block is going to
4107          * record the bytenr after swap, so we do the swap here.
4108          */
4109         block->subvol_bytenr = btrfs_node_blockptr(reloc_parent, reloc_slot);
4110         block->subvol_generation = btrfs_node_ptr_generation(reloc_parent,
4111                                                              reloc_slot);
4112         block->reloc_bytenr = btrfs_node_blockptr(subvol_parent, subvol_slot);
4113         block->reloc_generation = btrfs_node_ptr_generation(subvol_parent,
4114                                                             subvol_slot);
4115         block->last_snapshot = last_snapshot;
4116         block->level = level;
4117
4118         /*
4119          * If we have bg == NULL, we're called from btrfs_recover_relocation(),
4120          * no one else can modify tree blocks thus we qgroup will not change
4121          * no matter the value of trace_leaf.
4122          */
4123         if (bg && bg->flags & BTRFS_BLOCK_GROUP_DATA)
4124                 block->trace_leaf = true;
4125         else
4126                 block->trace_leaf = false;
4127         btrfs_node_key_to_cpu(reloc_parent, &block->first_key, reloc_slot);
4128
4129         /* Insert @block into @blocks */
4130         spin_lock(&blocks->lock);
4131         cur = &blocks->blocks[level].rb_node;
4132         while (*cur) {
4133                 struct btrfs_qgroup_swapped_block *entry;
4134
4135                 parent = *cur;
4136                 entry = rb_entry(parent, struct btrfs_qgroup_swapped_block,
4137                                  node);
4138
4139                 if (entry->subvol_bytenr < block->subvol_bytenr) {
4140                         cur = &(*cur)->rb_left;
4141                 } else if (entry->subvol_bytenr > block->subvol_bytenr) {
4142                         cur = &(*cur)->rb_right;
4143                 } else {
4144                         if (entry->subvol_generation !=
4145                                         block->subvol_generation ||
4146                             entry->reloc_bytenr != block->reloc_bytenr ||
4147                             entry->reloc_generation !=
4148                                         block->reloc_generation) {
4149                                 /*
4150                                  * Duplicated but mismatch entry found.
4151                                  * Shouldn't happen.
4152                                  *
4153                                  * Marking qgroup inconsistent should be enough
4154                                  * for end users.
4155                                  */
4156                                 WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
4157                                 ret = -EEXIST;
4158                         }
4159                         kfree(block);
4160                         goto out_unlock;
4161                 }
4162         }
4163         rb_link_node(&block->node, parent, cur);
4164         rb_insert_color(&block->node, &blocks->blocks[level]);
4165         blocks->swapped = true;
4166 out_unlock:
4167         spin_unlock(&blocks->lock);
4168 out:
4169         if (ret < 0)
4170                 fs_info->qgroup_flags |=
4171                         BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
4172         return ret;
4173 }
4174
4175 /*
4176  * Check if the tree block is a subtree root, and if so do the needed
4177  * delayed subtree trace for qgroup.
4178  *
4179  * This is called during btrfs_cow_block().
4180  */
4181 int btrfs_qgroup_trace_subtree_after_cow(struct btrfs_trans_handle *trans,
4182                                          struct btrfs_root *root,
4183                                          struct extent_buffer *subvol_eb)
4184 {
4185         struct btrfs_fs_info *fs_info = root->fs_info;
4186         struct btrfs_qgroup_swapped_blocks *blocks = &root->swapped_blocks;
4187         struct btrfs_qgroup_swapped_block *block;
4188         struct extent_buffer *reloc_eb = NULL;
4189         struct rb_node *node;
4190         bool found = false;
4191         bool swapped = false;
4192         int level = btrfs_header_level(subvol_eb);
4193         int ret = 0;
4194         int i;
4195
4196         if (!test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags))
4197                 return 0;
4198         if (!is_fstree(root->root_key.objectid) || !root->reloc_root)
4199                 return 0;
4200
4201         spin_lock(&blocks->lock);
4202         if (!blocks->swapped) {
4203                 spin_unlock(&blocks->lock);
4204                 return 0;
4205         }
4206         node = blocks->blocks[level].rb_node;
4207
4208         while (node) {
4209                 block = rb_entry(node, struct btrfs_qgroup_swapped_block, node);
4210                 if (block->subvol_bytenr < subvol_eb->start) {
4211                         node = node->rb_left;
4212                 } else if (block->subvol_bytenr > subvol_eb->start) {
4213                         node = node->rb_right;
4214                 } else {
4215                         found = true;
4216                         break;
4217                 }
4218         }
4219         if (!found) {
4220                 spin_unlock(&blocks->lock);
4221                 goto out;
4222         }
4223         /* Found one, remove it from @blocks first and update blocks->swapped */
4224         rb_erase(&block->node, &blocks->blocks[level]);
4225         for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
4226                 if (RB_EMPTY_ROOT(&blocks->blocks[i])) {
4227                         swapped = true;
4228                         break;
4229                 }
4230         }
4231         blocks->swapped = swapped;
4232         spin_unlock(&blocks->lock);
4233
4234         /* Read out reloc subtree root */
4235         reloc_eb = read_tree_block(fs_info, block->reloc_bytenr, 0,
4236                                    block->reloc_generation, block->level,
4237                                    &block->first_key);
4238         if (IS_ERR(reloc_eb)) {
4239                 ret = PTR_ERR(reloc_eb);
4240                 reloc_eb = NULL;
4241                 goto free_out;
4242         }
4243         if (!extent_buffer_uptodate(reloc_eb)) {
4244                 ret = -EIO;
4245                 goto free_out;
4246         }
4247
4248         ret = qgroup_trace_subtree_swap(trans, reloc_eb, subvol_eb,
4249                         block->last_snapshot, block->trace_leaf);
4250 free_out:
4251         kfree(block);
4252         free_extent_buffer(reloc_eb);
4253 out:
4254         if (ret < 0) {
4255                 btrfs_err_rl(fs_info,
4256                              "failed to account subtree at bytenr %llu: %d",
4257                              subvol_eb->start, ret);
4258                 fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
4259         }
4260         return ret;
4261 }
4262
4263 void btrfs_qgroup_destroy_extent_records(struct btrfs_transaction *trans)
4264 {
4265         struct btrfs_qgroup_extent_record *entry;
4266         struct btrfs_qgroup_extent_record *next;
4267         struct rb_root *root;
4268
4269         root = &trans->delayed_refs.dirty_extent_root;
4270         rbtree_postorder_for_each_entry_safe(entry, next, root, node) {
4271                 ulist_free(entry->old_roots);
4272                 kfree(entry);
4273         }
4274 }