linux/fs/f2fs/dir.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * fs/f2fs/dir.c
   4 *
   5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6 *             http://www.samsung.com/
   7 */
   8#include <asm/unaligned.h>
   9#include <linux/fs.h>
  10#include <linux/f2fs_fs.h>
  11#include <linux/sched/signal.h>
  12#include <linux/unicode.h>
  13#include "f2fs.h"
  14#include "node.h"
  15#include "acl.h"
  16#include "xattr.h"
  17#include <trace/events/f2fs.h>
  18
  19static unsigned long dir_blocks(struct inode *inode)
  20{
  21        return ((unsigned long long) (i_size_read(inode) + PAGE_SIZE - 1))
  22                                                        >> PAGE_SHIFT;
  23}
  24
  25static unsigned int dir_buckets(unsigned int level, int dir_level)
  26{
  27        if (level + dir_level < MAX_DIR_HASH_DEPTH / 2)
  28                return 1 << (level + dir_level);
  29        else
  30                return MAX_DIR_BUCKETS;
  31}
  32
  33static unsigned int bucket_blocks(unsigned int level)
  34{
  35        if (level < MAX_DIR_HASH_DEPTH / 2)
  36                return 2;
  37        else
  38                return 4;
  39}
  40
  41static unsigned char f2fs_filetype_table[F2FS_FT_MAX] = {
  42        [F2FS_FT_UNKNOWN]       = DT_UNKNOWN,
  43        [F2FS_FT_REG_FILE]      = DT_REG,
  44        [F2FS_FT_DIR]           = DT_DIR,
  45        [F2FS_FT_CHRDEV]        = DT_CHR,
  46        [F2FS_FT_BLKDEV]        = DT_BLK,
  47        [F2FS_FT_FIFO]          = DT_FIFO,
  48        [F2FS_FT_SOCK]          = DT_SOCK,
  49        [F2FS_FT_SYMLINK]       = DT_LNK,
  50};
  51
  52static unsigned char f2fs_type_by_mode[S_IFMT >> S_SHIFT] = {
  53        [S_IFREG >> S_SHIFT]    = F2FS_FT_REG_FILE,
  54        [S_IFDIR >> S_SHIFT]    = F2FS_FT_DIR,
  55        [S_IFCHR >> S_SHIFT]    = F2FS_FT_CHRDEV,
  56        [S_IFBLK >> S_SHIFT]    = F2FS_FT_BLKDEV,
  57        [S_IFIFO >> S_SHIFT]    = F2FS_FT_FIFO,
  58        [S_IFSOCK >> S_SHIFT]   = F2FS_FT_SOCK,
  59        [S_IFLNK >> S_SHIFT]    = F2FS_FT_SYMLINK,
  60};
  61
  62static void set_de_type(struct f2fs_dir_entry *de, umode_t mode)
  63{
  64        de->file_type = f2fs_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
  65}
  66
  67unsigned char f2fs_get_de_type(struct f2fs_dir_entry *de)
  68{
  69        if (de->file_type < F2FS_FT_MAX)
  70                return f2fs_filetype_table[de->file_type];
  71        return DT_UNKNOWN;
  72}
  73
  74/* If @dir is casefolded, initialize @fname->cf_name from @fname->usr_fname. */
  75int f2fs_init_casefolded_name(const struct inode *dir,
  76                              struct f2fs_filename *fname)
  77{
  78#ifdef CONFIG_UNICODE
  79        struct super_block *sb = dir->i_sb;
  80        struct f2fs_sb_info *sbi = F2FS_SB(sb);
  81
  82        if (IS_CASEFOLDED(dir)) {
  83                fname->cf_name.name = f2fs_kmalloc(sbi, F2FS_NAME_LEN,
  84                                                   GFP_NOFS);
  85                if (!fname->cf_name.name)
  86                        return -ENOMEM;
  87                fname->cf_name.len = utf8_casefold(sb->s_encoding,
  88                                                   fname->usr_fname,
  89                                                   fname->cf_name.name,
  90                                                   F2FS_NAME_LEN);
  91                if ((int)fname->cf_name.len <= 0) {
  92                        kfree(fname->cf_name.name);
  93                        fname->cf_name.name = NULL;
  94                        if (sb_has_strict_encoding(sb))
  95                                return -EINVAL;
  96                        /* fall back to treating name as opaque byte sequence */
  97                }
  98        }
  99#endif
 100        return 0;
 101}
 102
 103static int __f2fs_setup_filename(const struct inode *dir,
 104                                 const struct fscrypt_name *crypt_name,
 105                                 struct f2fs_filename *fname)
 106{
 107        int err;
 108
 109        memset(fname, 0, sizeof(*fname));
 110
 111        fname->usr_fname = crypt_name->usr_fname;
 112        fname->disk_name = crypt_name->disk_name;
 113#ifdef CONFIG_FS_ENCRYPTION
 114        fname->crypto_buf = crypt_name->crypto_buf;
 115#endif
 116        if (crypt_name->is_nokey_name) {
 117                /* hash was decoded from the no-key name */
 118                fname->hash = cpu_to_le32(crypt_name->hash);
 119        } else {
 120                err = f2fs_init_casefolded_name(dir, fname);
 121                if (err) {
 122                        f2fs_free_filename(fname);
 123                        return err;
 124                }
 125                f2fs_hash_filename(dir, fname);
 126        }
 127        return 0;
 128}
 129
 130/*
 131 * Prepare to search for @iname in @dir.  This is similar to
 132 * fscrypt_setup_filename(), but this also handles computing the casefolded name
 133 * and the f2fs dirhash if needed, then packing all the information about this
 134 * filename up into a 'struct f2fs_filename'.
 135 */
 136int f2fs_setup_filename(struct inode *dir, const struct qstr *iname,
 137                        int lookup, struct f2fs_filename *fname)
 138{
 139        struct fscrypt_name crypt_name;
 140        int err;
 141
 142        err = fscrypt_setup_filename(dir, iname, lookup, &crypt_name);
 143        if (err)
 144                return err;
 145
 146        return __f2fs_setup_filename(dir, &crypt_name, fname);
 147}
 148
 149/*
 150 * Prepare to look up @dentry in @dir.  This is similar to
 151 * fscrypt_prepare_lookup(), but this also handles computing the casefolded name
 152 * and the f2fs dirhash if needed, then packing all the information about this
 153 * filename up into a 'struct f2fs_filename'.
 154 */
 155int f2fs_prepare_lookup(struct inode *dir, struct dentry *dentry,
 156                        struct f2fs_filename *fname)
 157{
 158        struct fscrypt_name crypt_name;
 159        int err;
 160
 161        err = fscrypt_prepare_lookup(dir, dentry, &crypt_name);
 162        if (err)
 163                return err;
 164
 165        return __f2fs_setup_filename(dir, &crypt_name, fname);
 166}
 167
 168void f2fs_free_filename(struct f2fs_filename *fname)
 169{
 170#ifdef CONFIG_FS_ENCRYPTION
 171        kfree(fname->crypto_buf.name);
 172        fname->crypto_buf.name = NULL;
 173#endif
 174#ifdef CONFIG_UNICODE
 175        kfree(fname->cf_name.name);
 176        fname->cf_name.name = NULL;
 177#endif
 178}
 179
 180static unsigned long dir_block_index(unsigned int level,
 181                                int dir_level, unsigned int idx)
 182{
 183        unsigned long i;
 184        unsigned long bidx = 0;
 185
 186        for (i = 0; i < level; i++)
 187                bidx += dir_buckets(i, dir_level) * bucket_blocks(i);
 188        bidx += idx * bucket_blocks(level);
 189        return bidx;
 190}
 191
 192static struct f2fs_dir_entry *find_in_block(struct inode *dir,
 193                                struct page *dentry_page,
 194                                const struct f2fs_filename *fname,
 195                                int *max_slots)
 196{
 197        struct f2fs_dentry_block *dentry_blk;
 198        struct f2fs_dentry_ptr d;
 199
 200        dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
 201
 202        make_dentry_ptr_block(dir, &d, dentry_blk);
 203        return f2fs_find_target_dentry(&d, fname, max_slots);
 204}
 205
 206#ifdef CONFIG_UNICODE
 207/*
 208 * Test whether a case-insensitive directory entry matches the filename
 209 * being searched for.
 210 *
 211 * Returns 1 for a match, 0 for no match, and -errno on an error.
 212 */
 213static int f2fs_match_ci_name(const struct inode *dir, const struct qstr *name,
 214                               const u8 *de_name, u32 de_name_len)
 215{
 216        const struct super_block *sb = dir->i_sb;
 217        const struct unicode_map *um = sb->s_encoding;
 218        struct fscrypt_str decrypted_name = FSTR_INIT(NULL, de_name_len);
 219        struct qstr entry = QSTR_INIT(de_name, de_name_len);
 220        int res;
 221
 222        if (IS_ENCRYPTED(dir)) {
 223                const struct fscrypt_str encrypted_name =
 224                        FSTR_INIT((u8 *)de_name, de_name_len);
 225
 226                if (WARN_ON_ONCE(!fscrypt_has_encryption_key(dir)))
 227                        return -EINVAL;
 228
 229                decrypted_name.name = kmalloc(de_name_len, GFP_KERNEL);
 230                if (!decrypted_name.name)
 231                        return -ENOMEM;
 232                res = fscrypt_fname_disk_to_usr(dir, 0, 0, &encrypted_name,
 233                                                &decrypted_name);
 234                if (res < 0)
 235                        goto out;
 236                entry.name = decrypted_name.name;
 237                entry.len = decrypted_name.len;
 238        }
 239
 240        res = utf8_strncasecmp_folded(um, name, &entry);
 241        /*
 242         * In strict mode, ignore invalid names.  In non-strict mode,
 243         * fall back to treating them as opaque byte sequences.
 244         */
 245        if (res < 0 && !sb_has_strict_encoding(sb)) {
 246                res = name->len == entry.len &&
 247                                memcmp(name->name, entry.name, name->len) == 0;
 248        } else {
 249                /* utf8_strncasecmp_folded returns 0 on match */
 250                res = (res == 0);
 251        }
 252out:
 253        kfree(decrypted_name.name);
 254        return res;
 255}
 256#endif /* CONFIG_UNICODE */
 257
 258static inline int f2fs_match_name(const struct inode *dir,
 259                                   const struct f2fs_filename *fname,
 260                                   const u8 *de_name, u32 de_name_len)
 261{
 262        struct fscrypt_name f;
 263
 264#ifdef CONFIG_UNICODE
 265        if (fname->cf_name.name) {
 266                struct qstr cf = FSTR_TO_QSTR(&fname->cf_name);
 267
 268                return f2fs_match_ci_name(dir, &cf, de_name, de_name_len);
 269        }
 270#endif
 271        f.usr_fname = fname->usr_fname;
 272        f.disk_name = fname->disk_name;
 273#ifdef CONFIG_FS_ENCRYPTION
 274        f.crypto_buf = fname->crypto_buf;
 275#endif
 276        return fscrypt_match_name(&f, de_name, de_name_len);
 277}
 278
 279struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
 280                        const struct f2fs_filename *fname, int *max_slots)
 281{
 282        struct f2fs_dir_entry *de;
 283        unsigned long bit_pos = 0;
 284        int max_len = 0;
 285        int res = 0;
 286
 287        if (max_slots)
 288                *max_slots = 0;
 289        while (bit_pos < d->max) {
 290                if (!test_bit_le(bit_pos, d->bitmap)) {
 291                        bit_pos++;
 292                        max_len++;
 293                        continue;
 294                }
 295
 296                de = &d->dentry[bit_pos];
 297
 298                if (unlikely(!de->name_len)) {
 299                        bit_pos++;
 300                        continue;
 301                }
 302
 303                if (de->hash_code == fname->hash) {
 304                        res = f2fs_match_name(d->inode, fname,
 305                                              d->filename[bit_pos],
 306                                              le16_to_cpu(de->name_len));
 307                        if (res < 0)
 308                                return ERR_PTR(res);
 309                        if (res)
 310                                goto found;
 311                }
 312
 313                if (max_slots && max_len > *max_slots)
 314                        *max_slots = max_len;
 315                max_len = 0;
 316
 317                bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
 318        }
 319
 320        de = NULL;
 321found:
 322        if (max_slots && max_len > *max_slots)
 323                *max_slots = max_len;
 324        return de;
 325}
 326
 327static struct f2fs_dir_entry *find_in_level(struct inode *dir,
 328                                        unsigned int level,
 329                                        const struct f2fs_filename *fname,
 330                                        struct page **res_page)
 331{
 332        int s = GET_DENTRY_SLOTS(fname->disk_name.len);
 333        unsigned int nbucket, nblock;
 334        unsigned int bidx, end_block;
 335        struct page *dentry_page;
 336        struct f2fs_dir_entry *de = NULL;
 337        bool room = false;
 338        int max_slots;
 339
 340        nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
 341        nblock = bucket_blocks(level);
 342
 343        bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
 344                               le32_to_cpu(fname->hash) % nbucket);
 345        end_block = bidx + nblock;
 346
 347        for (; bidx < end_block; bidx++) {
 348                /* no need to allocate new dentry pages to all the indices */
 349                dentry_page = f2fs_find_data_page(dir, bidx);
 350                if (IS_ERR(dentry_page)) {
 351                        if (PTR_ERR(dentry_page) == -ENOENT) {
 352                                room = true;
 353                                continue;
 354                        } else {
 355                                *res_page = dentry_page;
 356                                break;
 357                        }
 358                }
 359
 360                de = find_in_block(dir, dentry_page, fname, &max_slots);
 361                if (IS_ERR(de)) {
 362                        *res_page = ERR_CAST(de);
 363                        de = NULL;
 364                        break;
 365                } else if (de) {
 366                        *res_page = dentry_page;
 367                        break;
 368                }
 369
 370                if (max_slots >= s)
 371                        room = true;
 372                f2fs_put_page(dentry_page, 0);
 373        }
 374
 375        if (!de && room && F2FS_I(dir)->chash != fname->hash) {
 376                F2FS_I(dir)->chash = fname->hash;
 377                F2FS_I(dir)->clevel = level;
 378        }
 379
 380        return de;
 381}
 382
 383struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
 384                                         const struct f2fs_filename *fname,
 385                                         struct page **res_page)
 386{
 387        unsigned long npages = dir_blocks(dir);
 388        struct f2fs_dir_entry *de = NULL;
 389        unsigned int max_depth;
 390        unsigned int level;
 391
 392        *res_page = NULL;
 393
 394        if (f2fs_has_inline_dentry(dir)) {
 395                de = f2fs_find_in_inline_dir(dir, fname, res_page);
 396                goto out;
 397        }
 398
 399        if (npages == 0)
 400                goto out;
 401
 402        max_depth = F2FS_I(dir)->i_current_depth;
 403        if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
 404                f2fs_warn(F2FS_I_SB(dir), "Corrupted max_depth of %lu: %u",
 405                          dir->i_ino, max_depth);
 406                max_depth = MAX_DIR_HASH_DEPTH;
 407                f2fs_i_depth_write(dir, max_depth);
 408        }
 409
 410        for (level = 0; level < max_depth; level++) {
 411                de = find_in_level(dir, level, fname, res_page);
 412                if (de || IS_ERR(*res_page))
 413                        break;
 414        }
 415out:
 416        /* This is to increase the speed of f2fs_create */
 417        if (!de)
 418                F2FS_I(dir)->task = current;
 419        return de;
 420}
 421
 422/*
 423 * Find an entry in the specified directory with the wanted name.
 424 * It returns the page where the entry was found (as a parameter - res_page),
 425 * and the entry itself. Page is returned mapped and unlocked.
 426 * Entry is guaranteed to be valid.
 427 */
 428struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
 429                        const struct qstr *child, struct page **res_page)
 430{
 431        struct f2fs_dir_entry *de = NULL;
 432        struct f2fs_filename fname;
 433        int err;
 434
 435        err = f2fs_setup_filename(dir, child, 1, &fname);
 436        if (err) {
 437                if (err == -ENOENT)
 438                        *res_page = NULL;
 439                else
 440                        *res_page = ERR_PTR(err);
 441                return NULL;
 442        }
 443
 444        de = __f2fs_find_entry(dir, &fname, res_page);
 445
 446        f2fs_free_filename(&fname);
 447        return de;
 448}
 449
 450struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
 451{
 452        struct qstr dotdot = QSTR_INIT("..", 2);
 453
 454        return f2fs_find_entry(dir, &dotdot, p);
 455}
 456
 457ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
 458                                                        struct page **page)
 459{
 460        ino_t res = 0;
 461        struct f2fs_dir_entry *de;
 462
 463        de = f2fs_find_entry(dir, qstr, page);
 464        if (de) {
 465                res = le32_to_cpu(de->ino);
 466                f2fs_put_page(*page, 0);
 467        }
 468
 469        return res;
 470}
 471
 472void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
 473                struct page *page, struct inode *inode)
 474{
 475        enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
 476        lock_page(page);
 477        f2fs_wait_on_page_writeback(page, type, true, true);
 478        de->ino = cpu_to_le32(inode->i_ino);
 479        set_de_type(de, inode->i_mode);
 480        set_page_dirty(page);
 481
 482        dir->i_mtime = dir->i_ctime = current_time(dir);
 483        f2fs_mark_inode_dirty_sync(dir, false);
 484        f2fs_put_page(page, 1);
 485}
 486
 487static void init_dent_inode(struct inode *dir, struct inode *inode,
 488                            const struct f2fs_filename *fname,
 489                            struct page *ipage)
 490{
 491        struct f2fs_inode *ri;
 492
 493        if (!fname) /* tmpfile case? */
 494                return;
 495
 496        f2fs_wait_on_page_writeback(ipage, NODE, true, true);
 497
 498        /* copy name info. to this inode page */
 499        ri = F2FS_INODE(ipage);
 500        ri->i_namelen = cpu_to_le32(fname->disk_name.len);
 501        memcpy(ri->i_name, fname->disk_name.name, fname->disk_name.len);
 502        if (IS_ENCRYPTED(dir)) {
 503                file_set_enc_name(inode);
 504                /*
 505                 * Roll-forward recovery doesn't have encryption keys available,
 506                 * so it can't compute the dirhash for encrypted+casefolded
 507                 * filenames.  Append it to i_name if possible.  Else, disable
 508                 * roll-forward recovery of the dentry (i.e., make fsync'ing the
 509                 * file force a checkpoint) by setting LOST_PINO.
 510                 */
 511                if (IS_CASEFOLDED(dir)) {
 512                        if (fname->disk_name.len + sizeof(f2fs_hash_t) <=
 513                            F2FS_NAME_LEN)
 514                                put_unaligned(fname->hash, (f2fs_hash_t *)
 515                                        &ri->i_name[fname->disk_name.len]);
 516                        else
 517                                file_lost_pino(inode);
 518                }
 519        }
 520        set_page_dirty(ipage);
 521}
 522
 523void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
 524                                        struct f2fs_dentry_ptr *d)
 525{
 526        struct fscrypt_str dot = FSTR_INIT(".", 1);
 527        struct fscrypt_str dotdot = FSTR_INIT("..", 2);
 528
 529        /* update dirent of "." */
 530        f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
 531
 532        /* update dirent of ".." */
 533        f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
 534}
 535
 536static int make_empty_dir(struct inode *inode,
 537                struct inode *parent, struct page *page)
 538{
 539        struct page *dentry_page;
 540        struct f2fs_dentry_block *dentry_blk;
 541        struct f2fs_dentry_ptr d;
 542
 543        if (f2fs_has_inline_dentry(inode))
 544                return f2fs_make_empty_inline_dir(inode, parent, page);
 545
 546        dentry_page = f2fs_get_new_data_page(inode, page, 0, true);
 547        if (IS_ERR(dentry_page))
 548                return PTR_ERR(dentry_page);
 549
 550        dentry_blk = page_address(dentry_page);
 551
 552        make_dentry_ptr_block(NULL, &d, dentry_blk);
 553        f2fs_do_make_empty_dir(inode, parent, &d);
 554
 555        set_page_dirty(dentry_page);
 556        f2fs_put_page(dentry_page, 1);
 557        return 0;
 558}
 559
 560struct page *f2fs_init_inode_metadata(struct inode *inode, struct inode *dir,
 561                        const struct f2fs_filename *fname, struct page *dpage)
 562{
 563        struct page *page;
 564        int err;
 565
 566        if (is_inode_flag_set(inode, FI_NEW_INODE)) {
 567                page = f2fs_new_inode_page(inode);
 568                if (IS_ERR(page))
 569                        return page;
 570
 571                if (S_ISDIR(inode->i_mode)) {
 572                        /* in order to handle error case */
 573                        get_page(page);
 574                        err = make_empty_dir(inode, dir, page);
 575                        if (err) {
 576                                lock_page(page);
 577                                goto put_error;
 578                        }
 579                        put_page(page);
 580                }
 581
 582                err = f2fs_init_acl(inode, dir, page, dpage);
 583                if (err)
 584                        goto put_error;
 585
 586                err = f2fs_init_security(inode, dir,
 587                                         fname ? fname->usr_fname : NULL, page);
 588                if (err)
 589                        goto put_error;
 590
 591                if (IS_ENCRYPTED(inode)) {
 592                        err = fscrypt_set_context(inode, page);
 593                        if (err)
 594                                goto put_error;
 595                }
 596        } else {
 597                page = f2fs_get_node_page(F2FS_I_SB(dir), inode->i_ino);
 598                if (IS_ERR(page))
 599                        return page;
 600        }
 601
 602        init_dent_inode(dir, inode, fname, page);
 603
 604        /*
 605         * This file should be checkpointed during fsync.
 606         * We lost i_pino from now on.
 607         */
 608        if (is_inode_flag_set(inode, FI_INC_LINK)) {
 609                if (!S_ISDIR(inode->i_mode))
 610                        file_lost_pino(inode);
 611                /*
 612                 * If link the tmpfile to alias through linkat path,
 613                 * we should remove this inode from orphan list.
 614                 */
 615                if (inode->i_nlink == 0)
 616                        f2fs_remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
 617                f2fs_i_links_write(inode, true);
 618        }
 619        return page;
 620
 621put_error:
 622        clear_nlink(inode);
 623        f2fs_update_inode(inode, page);
 624        f2fs_put_page(page, 1);
 625        return ERR_PTR(err);
 626}
 627
 628void f2fs_update_parent_metadata(struct inode *dir, struct inode *inode,
 629                                                unsigned int current_depth)
 630{
 631        if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
 632                if (S_ISDIR(inode->i_mode))
 633                        f2fs_i_links_write(dir, true);
 634                clear_inode_flag(inode, FI_NEW_INODE);
 635        }
 636        dir->i_mtime = dir->i_ctime = current_time(dir);
 637        f2fs_mark_inode_dirty_sync(dir, false);
 638
 639        if (F2FS_I(dir)->i_current_depth != current_depth)
 640                f2fs_i_depth_write(dir, current_depth);
 641
 642        if (inode && is_inode_flag_set(inode, FI_INC_LINK))
 643                clear_inode_flag(inode, FI_INC_LINK);
 644}
 645
 646int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
 647{
 648        int bit_start = 0;
 649        int zero_start, zero_end;
 650next:
 651        zero_start = find_next_zero_bit_le(bitmap, max_slots, bit_start);
 652        if (zero_start >= max_slots)
 653                return max_slots;
 654
 655        zero_end = find_next_bit_le(bitmap, max_slots, zero_start);
 656        if (zero_end - zero_start >= slots)
 657                return zero_start;
 658
 659        bit_start = zero_end + 1;
 660
 661        if (zero_end + 1 >= max_slots)
 662                return max_slots;
 663        goto next;
 664}
 665
 666bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
 667                          const struct f2fs_filename *fname)
 668{
 669        struct f2fs_dentry_ptr d;
 670        unsigned int bit_pos;
 671        int slots = GET_DENTRY_SLOTS(fname->disk_name.len);
 672
 673        make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
 674
 675        bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
 676
 677        return bit_pos < d.max;
 678}
 679
 680void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
 681                        const struct fscrypt_str *name, f2fs_hash_t name_hash,
 682                        unsigned int bit_pos)
 683{
 684        struct f2fs_dir_entry *de;
 685        int slots = GET_DENTRY_SLOTS(name->len);
 686        int i;
 687
 688        de = &d->dentry[bit_pos];
 689        de->hash_code = name_hash;
 690        de->name_len = cpu_to_le16(name->len);
 691        memcpy(d->filename[bit_pos], name->name, name->len);
 692        de->ino = cpu_to_le32(ino);
 693        set_de_type(de, mode);
 694        for (i = 0; i < slots; i++) {
 695                __set_bit_le(bit_pos + i, (void *)d->bitmap);
 696                /* avoid wrong garbage data for readdir */
 697                if (i)
 698                        (de + i)->name_len = 0;
 699        }
 700}
 701
 702int f2fs_add_regular_entry(struct inode *dir, const struct f2fs_filename *fname,
 703                           struct inode *inode, nid_t ino, umode_t mode)
 704{
 705        unsigned int bit_pos;
 706        unsigned int level;
 707        unsigned int current_depth;
 708        unsigned long bidx, block;
 709        unsigned int nbucket, nblock;
 710        struct page *dentry_page = NULL;
 711        struct f2fs_dentry_block *dentry_blk = NULL;
 712        struct f2fs_dentry_ptr d;
 713        struct page *page = NULL;
 714        int slots, err = 0;
 715
 716        level = 0;
 717        slots = GET_DENTRY_SLOTS(fname->disk_name.len);
 718
 719        current_depth = F2FS_I(dir)->i_current_depth;
 720        if (F2FS_I(dir)->chash == fname->hash) {
 721                level = F2FS_I(dir)->clevel;
 722                F2FS_I(dir)->chash = 0;
 723        }
 724
 725start:
 726        if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH)) {
 727                f2fs_show_injection_info(F2FS_I_SB(dir), FAULT_DIR_DEPTH);
 728                return -ENOSPC;
 729        }
 730
 731        if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
 732                return -ENOSPC;
 733
 734        /* Increase the depth, if required */
 735        if (level == current_depth)
 736                ++current_depth;
 737
 738        nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
 739        nblock = bucket_blocks(level);
 740
 741        bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
 742                                (le32_to_cpu(fname->hash) % nbucket));
 743
 744        for (block = bidx; block <= (bidx + nblock - 1); block++) {
 745                dentry_page = f2fs_get_new_data_page(dir, NULL, block, true);
 746                if (IS_ERR(dentry_page))
 747                        return PTR_ERR(dentry_page);
 748
 749                dentry_blk = page_address(dentry_page);
 750                bit_pos = f2fs_room_for_filename(&dentry_blk->dentry_bitmap,
 751                                                slots, NR_DENTRY_IN_BLOCK);
 752                if (bit_pos < NR_DENTRY_IN_BLOCK)
 753                        goto add_dentry;
 754
 755                f2fs_put_page(dentry_page, 1);
 756        }
 757
 758        /* Move to next level to find the empty slot for new dentry */
 759        ++level;
 760        goto start;
 761add_dentry:
 762        f2fs_wait_on_page_writeback(dentry_page, DATA, true, true);
 763
 764        if (inode) {
 765                down_write(&F2FS_I(inode)->i_sem);
 766                page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
 767                if (IS_ERR(page)) {
 768                        err = PTR_ERR(page);
 769                        goto fail;
 770                }
 771        }
 772
 773        make_dentry_ptr_block(NULL, &d, dentry_blk);
 774        f2fs_update_dentry(ino, mode, &d, &fname->disk_name, fname->hash,
 775                           bit_pos);
 776
 777        set_page_dirty(dentry_page);
 778
 779        if (inode) {
 780                f2fs_i_pino_write(inode, dir->i_ino);
 781
 782                /* synchronize inode page's data from inode cache */
 783                if (is_inode_flag_set(inode, FI_NEW_INODE))
 784                        f2fs_update_inode(inode, page);
 785
 786                f2fs_put_page(page, 1);
 787        }
 788
 789        f2fs_update_parent_metadata(dir, inode, current_depth);
 790fail:
 791        if (inode)
 792                up_write(&F2FS_I(inode)->i_sem);
 793
 794        f2fs_put_page(dentry_page, 1);
 795
 796        return err;
 797}
 798
 799int f2fs_add_dentry(struct inode *dir, const struct f2fs_filename *fname,
 800                    struct inode *inode, nid_t ino, umode_t mode)
 801{
 802        int err = -EAGAIN;
 803
 804        if (f2fs_has_inline_dentry(dir))
 805                err = f2fs_add_inline_entry(dir, fname, inode, ino, mode);
 806        if (err == -EAGAIN)
 807                err = f2fs_add_regular_entry(dir, fname, inode, ino, mode);
 808
 809        f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
 810        return err;
 811}
 812
 813/*
 814 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
 815 * f2fs_unlock_op().
 816 */
 817int f2fs_do_add_link(struct inode *dir, const struct qstr *name,
 818                                struct inode *inode, nid_t ino, umode_t mode)
 819{
 820        struct f2fs_filename fname;
 821        struct page *page = NULL;
 822        struct f2fs_dir_entry *de = NULL;
 823        int err;
 824
 825        err = f2fs_setup_filename(dir, name, 0, &fname);
 826        if (err)
 827                return err;
 828
 829        /*
 830         * An immature stackable filesystem shows a race condition between lookup
 831         * and create. If we have same task when doing lookup and create, it's
 832         * definitely fine as expected by VFS normally. Otherwise, let's just
 833         * verify on-disk dentry one more time, which guarantees filesystem
 834         * consistency more.
 835         */
 836        if (current != F2FS_I(dir)->task) {
 837                de = __f2fs_find_entry(dir, &fname, &page);
 838                F2FS_I(dir)->task = NULL;
 839        }
 840        if (de) {
 841                f2fs_put_page(page, 0);
 842                err = -EEXIST;
 843        } else if (IS_ERR(page)) {
 844                err = PTR_ERR(page);
 845        } else {
 846                err = f2fs_add_dentry(dir, &fname, inode, ino, mode);
 847        }
 848        f2fs_free_filename(&fname);
 849        return err;
 850}
 851
 852int f2fs_do_tmpfile(struct inode *inode, struct inode *dir)
 853{
 854        struct page *page;
 855        int err = 0;
 856
 857        down_write(&F2FS_I(inode)->i_sem);
 858        page = f2fs_init_inode_metadata(inode, dir, NULL, NULL);
 859        if (IS_ERR(page)) {
 860                err = PTR_ERR(page);
 861                goto fail;
 862        }
 863        f2fs_put_page(page, 1);
 864
 865        clear_inode_flag(inode, FI_NEW_INODE);
 866        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
 867fail:
 868        up_write(&F2FS_I(inode)->i_sem);
 869        return err;
 870}
 871
 872void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
 873{
 874        struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
 875
 876        down_write(&F2FS_I(inode)->i_sem);
 877
 878        if (S_ISDIR(inode->i_mode))
 879                f2fs_i_links_write(dir, false);
 880        inode->i_ctime = current_time(inode);
 881
 882        f2fs_i_links_write(inode, false);
 883        if (S_ISDIR(inode->i_mode)) {
 884                f2fs_i_links_write(inode, false);
 885                f2fs_i_size_write(inode, 0);
 886        }
 887        up_write(&F2FS_I(inode)->i_sem);
 888
 889        if (inode->i_nlink == 0)
 890                f2fs_add_orphan_inode(inode);
 891        else
 892                f2fs_release_orphan_inode(sbi);
 893}
 894
 895/*
 896 * It only removes the dentry from the dentry page, corresponding name
 897 * entry in name page does not need to be touched during deletion.
 898 */
 899void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
 900                                        struct inode *dir, struct inode *inode)
 901{
 902        struct  f2fs_dentry_block *dentry_blk;
 903        unsigned int bit_pos;
 904        int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
 905        int i;
 906
 907        f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
 908
 909        if (F2FS_OPTION(F2FS_I_SB(dir)).fsync_mode == FSYNC_MODE_STRICT)
 910                f2fs_add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
 911
 912        if (f2fs_has_inline_dentry(dir))
 913                return f2fs_delete_inline_entry(dentry, page, dir, inode);
 914
 915        lock_page(page);
 916        f2fs_wait_on_page_writeback(page, DATA, true, true);
 917
 918        dentry_blk = page_address(page);
 919        bit_pos = dentry - dentry_blk->dentry;
 920        for (i = 0; i < slots; i++)
 921                __clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
 922
 923        /* Let's check and deallocate this dentry page */
 924        bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
 925                        NR_DENTRY_IN_BLOCK,
 926                        0);
 927        set_page_dirty(page);
 928
 929        if (bit_pos == NR_DENTRY_IN_BLOCK &&
 930                !f2fs_truncate_hole(dir, page->index, page->index + 1)) {
 931                f2fs_clear_page_cache_dirty_tag(page);
 932                clear_page_dirty_for_io(page);
 933                f2fs_clear_page_private(page);
 934                ClearPageUptodate(page);
 935                clear_cold_data(page);
 936                inode_dec_dirty_pages(dir);
 937                f2fs_remove_dirty_inode(dir);
 938        }
 939        f2fs_put_page(page, 1);
 940
 941        dir->i_ctime = dir->i_mtime = current_time(dir);
 942        f2fs_mark_inode_dirty_sync(dir, false);
 943
 944        if (inode)
 945                f2fs_drop_nlink(dir, inode);
 946}
 947
 948bool f2fs_empty_dir(struct inode *dir)
 949{
 950        unsigned long bidx;
 951        struct page *dentry_page;
 952        unsigned int bit_pos;
 953        struct f2fs_dentry_block *dentry_blk;
 954        unsigned long nblock = dir_blocks(dir);
 955
 956        if (f2fs_has_inline_dentry(dir))
 957                return f2fs_empty_inline_dir(dir);
 958
 959        for (bidx = 0; bidx < nblock; bidx++) {
 960                dentry_page = f2fs_get_lock_data_page(dir, bidx, false);
 961                if (IS_ERR(dentry_page)) {
 962                        if (PTR_ERR(dentry_page) == -ENOENT)
 963                                continue;
 964                        else
 965                                return false;
 966                }
 967
 968                dentry_blk = page_address(dentry_page);
 969                if (bidx == 0)
 970                        bit_pos = 2;
 971                else
 972                        bit_pos = 0;
 973                bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
 974                                                NR_DENTRY_IN_BLOCK,
 975                                                bit_pos);
 976
 977                f2fs_put_page(dentry_page, 1);
 978
 979                if (bit_pos < NR_DENTRY_IN_BLOCK)
 980                        return false;
 981        }
 982        return true;
 983}
 984
 985int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
 986                        unsigned int start_pos, struct fscrypt_str *fstr)
 987{
 988        unsigned char d_type = DT_UNKNOWN;
 989        unsigned int bit_pos;
 990        struct f2fs_dir_entry *de = NULL;
 991        struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
 992        struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
 993        struct blk_plug plug;
 994        bool readdir_ra = sbi->readdir_ra == 1;
 995        int err = 0;
 996
 997        bit_pos = ((unsigned long)ctx->pos % d->max);
 998
 999        if (readdir_ra)
1000                blk_start_plug(&plug);
1001
1002        while (bit_pos < d->max) {
1003                bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
1004                if (bit_pos >= d->max)
1005                        break;
1006
1007                de = &d->dentry[bit_pos];
1008                if (de->name_len == 0) {
1009                        bit_pos++;
1010                        ctx->pos = start_pos + bit_pos;
1011                        printk_ratelimited(
1012                                "%sF2FS-fs (%s): invalid namelen(0), ino:%u, run fsck to fix.",
1013                                KERN_WARNING, sbi->sb->s_id,
1014                                le32_to_cpu(de->ino));
1015                        set_sbi_flag(sbi, SBI_NEED_FSCK);
1016                        continue;
1017                }
1018
1019                d_type = f2fs_get_de_type(de);
1020
1021                de_name.name = d->filename[bit_pos];
1022                de_name.len = le16_to_cpu(de->name_len);
1023
1024                /* check memory boundary before moving forward */
1025                bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
1026                if (unlikely(bit_pos > d->max ||
1027                                le16_to_cpu(de->name_len) > F2FS_NAME_LEN)) {
1028                        f2fs_warn(sbi, "%s: corrupted namelen=%d, run fsck to fix.",
1029                                  __func__, le16_to_cpu(de->name_len));
1030                        set_sbi_flag(sbi, SBI_NEED_FSCK);
1031                        err = -EFSCORRUPTED;
1032                        goto out;
1033                }
1034
1035                if (IS_ENCRYPTED(d->inode)) {
1036                        int save_len = fstr->len;
1037
1038                        err = fscrypt_fname_disk_to_usr(d->inode,
1039                                                (u32)le32_to_cpu(de->hash_code),
1040                                                0, &de_name, fstr);
1041                        if (err)
1042                                goto out;
1043
1044                        de_name = *fstr;
1045                        fstr->len = save_len;
1046                }
1047
1048                if (!dir_emit(ctx, de_name.name, de_name.len,
1049                                        le32_to_cpu(de->ino), d_type)) {
1050                        err = 1;
1051                        goto out;
1052                }
1053
1054                if (readdir_ra)
1055                        f2fs_ra_node_page(sbi, le32_to_cpu(de->ino));
1056
1057                ctx->pos = start_pos + bit_pos;
1058        }
1059out:
1060        if (readdir_ra)
1061                blk_finish_plug(&plug);
1062        return err;
1063}
1064
1065static int f2fs_readdir(struct file *file, struct dir_context *ctx)
1066{
1067        struct inode *inode = file_inode(file);
1068        unsigned long npages = dir_blocks(inode);
1069        struct f2fs_dentry_block *dentry_blk = NULL;
1070        struct page *dentry_page = NULL;
1071        struct file_ra_state *ra = &file->f_ra;
1072        loff_t start_pos = ctx->pos;
1073        unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
1074        struct f2fs_dentry_ptr d;
1075        struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
1076        int err = 0;
1077
1078        if (IS_ENCRYPTED(inode)) {
1079                err = fscrypt_prepare_readdir(inode);
1080                if (err)
1081                        goto out;
1082
1083                err = fscrypt_fname_alloc_buffer(F2FS_NAME_LEN, &fstr);
1084                if (err < 0)
1085                        goto out;
1086        }
1087
1088        if (f2fs_has_inline_dentry(inode)) {
1089                err = f2fs_read_inline_dir(file, ctx, &fstr);
1090                goto out_free;
1091        }
1092
1093        for (; n < npages; n++, ctx->pos = n * NR_DENTRY_IN_BLOCK) {
1094
1095                /* allow readdir() to be interrupted */
1096                if (fatal_signal_pending(current)) {
1097                        err = -ERESTARTSYS;
1098                        goto out_free;
1099                }
1100                cond_resched();
1101
1102                /* readahead for multi pages of dir */
1103                if (npages - n > 1 && !ra_has_index(ra, n))
1104                        page_cache_sync_readahead(inode->i_mapping, ra, file, n,
1105                                min(npages - n, (pgoff_t)MAX_DIR_RA_PAGES));
1106
1107                dentry_page = f2fs_find_data_page(inode, n);
1108                if (IS_ERR(dentry_page)) {
1109                        err = PTR_ERR(dentry_page);
1110                        if (err == -ENOENT) {
1111                                err = 0;
1112                                continue;
1113                        } else {
1114                                goto out_free;
1115                        }
1116                }
1117
1118                dentry_blk = page_address(dentry_page);
1119
1120                make_dentry_ptr_block(inode, &d, dentry_blk);
1121
1122                err = f2fs_fill_dentries(ctx, &d,
1123                                n * NR_DENTRY_IN_BLOCK, &fstr);
1124                if (err) {
1125                        f2fs_put_page(dentry_page, 0);
1126                        break;
1127                }
1128
1129                f2fs_put_page(dentry_page, 0);
1130        }
1131out_free:
1132        fscrypt_fname_free_buffer(&fstr);
1133out:
1134        trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
1135        return err < 0 ? err : 0;
1136}
1137
1138const struct file_operations f2fs_dir_operations = {
1139        .llseek         = generic_file_llseek,
1140        .read           = generic_read_dir,
1141        .iterate_shared = f2fs_readdir,
1142        .fsync          = f2fs_sync_file,
1143        .unlocked_ioctl = f2fs_ioctl,
1144#ifdef CONFIG_COMPAT
1145        .compat_ioctl   = f2fs_compat_ioctl,
1146#endif
1147};
1148