linux/fs/f2fs/file.c
<<
>>
Prefs
   1/*
   2 * fs/f2fs/file.c
   3 *
   4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   5 *             http://www.samsung.com/
   6 *
   7 * This program is free software; you can redistribute it and/or modify
   8 * it under the terms of the GNU General Public License version 2 as
   9 * published by the Free Software Foundation.
  10 */
  11#include <linux/fs.h>
  12#include <linux/f2fs_fs.h>
  13#include <linux/stat.h>
  14#include <linux/buffer_head.h>
  15#include <linux/writeback.h>
  16#include <linux/blkdev.h>
  17#include <linux/falloc.h>
  18#include <linux/types.h>
  19#include <linux/compat.h>
  20#include <linux/uaccess.h>
  21#include <linux/mount.h>
  22#include <linux/pagevec.h>
  23#include <linux/uio.h>
  24#include <linux/uuid.h>
  25#include <linux/file.h>
  26
  27#include "f2fs.h"
  28#include "node.h"
  29#include "segment.h"
  30#include "xattr.h"
  31#include "acl.h"
  32#include "gc.h"
  33#include "trace.h"
  34#include <trace/events/f2fs.h>
  35
  36static int f2fs_filemap_fault(struct vm_fault *vmf)
  37{
  38        struct inode *inode = file_inode(vmf->vma->vm_file);
  39        int err;
  40
  41        down_read(&F2FS_I(inode)->i_mmap_sem);
  42        err = filemap_fault(vmf);
  43        up_read(&F2FS_I(inode)->i_mmap_sem);
  44
  45        return err;
  46}
  47
  48static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
  49{
  50        struct page *page = vmf->page;
  51        struct inode *inode = file_inode(vmf->vma->vm_file);
  52        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
  53        struct dnode_of_data dn;
  54        int err;
  55
  56        if (unlikely(f2fs_cp_error(sbi))) {
  57                err = -EIO;
  58                goto err;
  59        }
  60
  61        sb_start_pagefault(inode->i_sb);
  62
  63        f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
  64
  65        /* block allocation */
  66        f2fs_lock_op(sbi);
  67        set_new_dnode(&dn, inode, NULL, NULL, 0);
  68        err = f2fs_reserve_block(&dn, page->index);
  69        if (err) {
  70                f2fs_unlock_op(sbi);
  71                goto out;
  72        }
  73        f2fs_put_dnode(&dn);
  74        f2fs_unlock_op(sbi);
  75
  76        f2fs_balance_fs(sbi, dn.node_changed);
  77
  78        file_update_time(vmf->vma->vm_file);
  79        down_read(&F2FS_I(inode)->i_mmap_sem);
  80        lock_page(page);
  81        if (unlikely(page->mapping != inode->i_mapping ||
  82                        page_offset(page) > i_size_read(inode) ||
  83                        !PageUptodate(page))) {
  84                unlock_page(page);
  85                err = -EFAULT;
  86                goto out_sem;
  87        }
  88
  89        /*
  90         * check to see if the page is mapped already (no holes)
  91         */
  92        if (PageMappedToDisk(page))
  93                goto mapped;
  94
  95        /* page is wholly or partially inside EOF */
  96        if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
  97                                                i_size_read(inode)) {
  98                unsigned offset;
  99                offset = i_size_read(inode) & ~PAGE_MASK;
 100                zero_user_segment(page, offset, PAGE_SIZE);
 101        }
 102        set_page_dirty(page);
 103        if (!PageUptodate(page))
 104                SetPageUptodate(page);
 105
 106        f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
 107
 108        trace_f2fs_vm_page_mkwrite(page, DATA);
 109mapped:
 110        /* fill the page */
 111        f2fs_wait_on_page_writeback(page, DATA, false);
 112
 113        /* wait for GCed encrypted page writeback */
 114        if (f2fs_encrypted_file(inode))
 115                f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
 116
 117out_sem:
 118        up_read(&F2FS_I(inode)->i_mmap_sem);
 119out:
 120        sb_end_pagefault(inode->i_sb);
 121        f2fs_update_time(sbi, REQ_TIME);
 122err:
 123        return block_page_mkwrite_return(err);
 124}
 125
 126static const struct vm_operations_struct f2fs_file_vm_ops = {
 127        .fault          = f2fs_filemap_fault,
 128        .map_pages      = filemap_map_pages,
 129        .page_mkwrite   = f2fs_vm_page_mkwrite,
 130};
 131
 132static int get_parent_ino(struct inode *inode, nid_t *pino)
 133{
 134        struct dentry *dentry;
 135
 136        inode = igrab(inode);
 137        dentry = d_find_any_alias(inode);
 138        iput(inode);
 139        if (!dentry)
 140                return 0;
 141
 142        *pino = parent_ino(dentry);
 143        dput(dentry);
 144        return 1;
 145}
 146
 147static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
 148{
 149        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 150        enum cp_reason_type cp_reason = CP_NO_NEEDED;
 151
 152        if (!S_ISREG(inode->i_mode))
 153                cp_reason = CP_NON_REGULAR;
 154        else if (inode->i_nlink != 1)
 155                cp_reason = CP_HARDLINK;
 156        else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
 157                cp_reason = CP_SB_NEED_CP;
 158        else if (file_wrong_pino(inode))
 159                cp_reason = CP_WRONG_PINO;
 160        else if (!space_for_roll_forward(sbi))
 161                cp_reason = CP_NO_SPC_ROLL;
 162        else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
 163                cp_reason = CP_NODE_NEED_CP;
 164        else if (test_opt(sbi, FASTBOOT))
 165                cp_reason = CP_FASTBOOT_MODE;
 166        else if (F2FS_OPTION(sbi).active_logs == 2)
 167                cp_reason = CP_SPEC_LOG_NUM;
 168        else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
 169                need_dentry_mark(sbi, inode->i_ino) &&
 170                exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
 171                cp_reason = CP_RECOVER_DIR;
 172
 173        return cp_reason;
 174}
 175
 176static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
 177{
 178        struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
 179        bool ret = false;
 180        /* But we need to avoid that there are some inode updates */
 181        if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
 182                ret = true;
 183        f2fs_put_page(i, 0);
 184        return ret;
 185}
 186
 187static void try_to_fix_pino(struct inode *inode)
 188{
 189        struct f2fs_inode_info *fi = F2FS_I(inode);
 190        nid_t pino;
 191
 192        down_write(&fi->i_sem);
 193        if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
 194                        get_parent_ino(inode, &pino)) {
 195                f2fs_i_pino_write(inode, pino);
 196                file_got_pino(inode);
 197        }
 198        up_write(&fi->i_sem);
 199}
 200
 201static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
 202                                                int datasync, bool atomic)
 203{
 204        struct inode *inode = file->f_mapping->host;
 205        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 206        nid_t ino = inode->i_ino;
 207        int ret = 0;
 208        enum cp_reason_type cp_reason = 0;
 209        struct writeback_control wbc = {
 210                .sync_mode = WB_SYNC_ALL,
 211                .nr_to_write = LONG_MAX,
 212                .for_reclaim = 0,
 213        };
 214
 215        if (unlikely(f2fs_readonly(inode->i_sb)))
 216                return 0;
 217
 218        trace_f2fs_sync_file_enter(inode);
 219
 220        /* if fdatasync is triggered, let's do in-place-update */
 221        if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
 222                set_inode_flag(inode, FI_NEED_IPU);
 223        ret = file_write_and_wait_range(file, start, end);
 224        clear_inode_flag(inode, FI_NEED_IPU);
 225
 226        if (ret) {
 227                trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 228                return ret;
 229        }
 230
 231        /* if the inode is dirty, let's recover all the time */
 232        if (!f2fs_skip_inode_update(inode, datasync)) {
 233                f2fs_write_inode(inode, NULL);
 234                goto go_write;
 235        }
 236
 237        /*
 238         * if there is no written data, don't waste time to write recovery info.
 239         */
 240        if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
 241                        !exist_written_data(sbi, ino, APPEND_INO)) {
 242
 243                /* it may call write_inode just prior to fsync */
 244                if (need_inode_page_update(sbi, ino))
 245                        goto go_write;
 246
 247                if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
 248                                exist_written_data(sbi, ino, UPDATE_INO))
 249                        goto flush_out;
 250                goto out;
 251        }
 252go_write:
 253        /*
 254         * Both of fdatasync() and fsync() are able to be recovered from
 255         * sudden-power-off.
 256         */
 257        down_read(&F2FS_I(inode)->i_sem);
 258        cp_reason = need_do_checkpoint(inode);
 259        up_read(&F2FS_I(inode)->i_sem);
 260
 261        if (cp_reason) {
 262                /* all the dirty node pages should be flushed for POR */
 263                ret = f2fs_sync_fs(inode->i_sb, 1);
 264
 265                /*
 266                 * We've secured consistency through sync_fs. Following pino
 267                 * will be used only for fsynced inodes after checkpoint.
 268                 */
 269                try_to_fix_pino(inode);
 270                clear_inode_flag(inode, FI_APPEND_WRITE);
 271                clear_inode_flag(inode, FI_UPDATE_WRITE);
 272                goto out;
 273        }
 274sync_nodes:
 275        ret = fsync_node_pages(sbi, inode, &wbc, atomic);
 276        if (ret)
 277                goto out;
 278
 279        /* if cp_error was enabled, we should avoid infinite loop */
 280        if (unlikely(f2fs_cp_error(sbi))) {
 281                ret = -EIO;
 282                goto out;
 283        }
 284
 285        if (need_inode_block_update(sbi, ino)) {
 286                f2fs_mark_inode_dirty_sync(inode, true);
 287                f2fs_write_inode(inode, NULL);
 288                goto sync_nodes;
 289        }
 290
 291        /*
 292         * If it's atomic_write, it's just fine to keep write ordering. So
 293         * here we don't need to wait for node write completion, since we use
 294         * node chain which serializes node blocks. If one of node writes are
 295         * reordered, we can see simply broken chain, resulting in stopping
 296         * roll-forward recovery. It means we'll recover all or none node blocks
 297         * given fsync mark.
 298         */
 299        if (!atomic) {
 300                ret = wait_on_node_pages_writeback(sbi, ino);
 301                if (ret)
 302                        goto out;
 303        }
 304
 305        /* once recovery info is written, don't need to tack this */
 306        remove_ino_entry(sbi, ino, APPEND_INO);
 307        clear_inode_flag(inode, FI_APPEND_WRITE);
 308flush_out:
 309        if (!atomic)
 310                ret = f2fs_issue_flush(sbi, inode->i_ino);
 311        if (!ret) {
 312                remove_ino_entry(sbi, ino, UPDATE_INO);
 313                clear_inode_flag(inode, FI_UPDATE_WRITE);
 314                remove_ino_entry(sbi, ino, FLUSH_INO);
 315        }
 316        f2fs_update_time(sbi, REQ_TIME);
 317out:
 318        trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 319        f2fs_trace_ios(NULL, 1);
 320        return ret;
 321}
 322
 323int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 324{
 325        if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
 326                return -EIO;
 327        return f2fs_do_sync_file(file, start, end, datasync, false);
 328}
 329
 330static pgoff_t __get_first_dirty_index(struct address_space *mapping,
 331                                                pgoff_t pgofs, int whence)
 332{
 333        struct page *page;
 334        int nr_pages;
 335
 336        if (whence != SEEK_DATA)
 337                return 0;
 338
 339        /* find first dirty page index */
 340        nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
 341                                      1, &page);
 342        if (!nr_pages)
 343                return ULONG_MAX;
 344        pgofs = page->index;
 345        put_page(page);
 346        return pgofs;
 347}
 348
 349static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
 350                                                        int whence)
 351{
 352        switch (whence) {
 353        case SEEK_DATA:
 354                if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
 355                        (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
 356                        return true;
 357                break;
 358        case SEEK_HOLE:
 359                if (blkaddr == NULL_ADDR)
 360                        return true;
 361                break;
 362        }
 363        return false;
 364}
 365
 366static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 367{
 368        struct inode *inode = file->f_mapping->host;
 369        loff_t maxbytes = inode->i_sb->s_maxbytes;
 370        struct dnode_of_data dn;
 371        pgoff_t pgofs, end_offset, dirty;
 372        loff_t data_ofs = offset;
 373        loff_t isize;
 374        int err = 0;
 375
 376        inode_lock(inode);
 377
 378        isize = i_size_read(inode);
 379        if (offset >= isize)
 380                goto fail;
 381
 382        /* handle inline data case */
 383        if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
 384                if (whence == SEEK_HOLE)
 385                        data_ofs = isize;
 386                goto found;
 387        }
 388
 389        pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
 390
 391        dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
 392
 393        for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 394                set_new_dnode(&dn, inode, NULL, NULL, 0);
 395                err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
 396                if (err && err != -ENOENT) {
 397                        goto fail;
 398                } else if (err == -ENOENT) {
 399                        /* direct node does not exists */
 400                        if (whence == SEEK_DATA) {
 401                                pgofs = get_next_page_offset(&dn, pgofs);
 402                                continue;
 403                        } else {
 404                                goto found;
 405                        }
 406                }
 407
 408                end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 409
 410                /* find data/hole in dnode block */
 411                for (; dn.ofs_in_node < end_offset;
 412                                dn.ofs_in_node++, pgofs++,
 413                                data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 414                        block_t blkaddr;
 415                        blkaddr = datablock_addr(dn.inode,
 416                                        dn.node_page, dn.ofs_in_node);
 417
 418                        if (__found_offset(blkaddr, dirty, pgofs, whence)) {
 419                                f2fs_put_dnode(&dn);
 420                                goto found;
 421                        }
 422                }
 423                f2fs_put_dnode(&dn);
 424        }
 425
 426        if (whence == SEEK_DATA)
 427                goto fail;
 428found:
 429        if (whence == SEEK_HOLE && data_ofs > isize)
 430                data_ofs = isize;
 431        inode_unlock(inode);
 432        return vfs_setpos(file, data_ofs, maxbytes);
 433fail:
 434        inode_unlock(inode);
 435        return -ENXIO;
 436}
 437
 438static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
 439{
 440        struct inode *inode = file->f_mapping->host;
 441        loff_t maxbytes = inode->i_sb->s_maxbytes;
 442
 443        switch (whence) {
 444        case SEEK_SET:
 445        case SEEK_CUR:
 446        case SEEK_END:
 447                return generic_file_llseek_size(file, offset, whence,
 448                                                maxbytes, i_size_read(inode));
 449        case SEEK_DATA:
 450        case SEEK_HOLE:
 451                if (offset < 0)
 452                        return -ENXIO;
 453                return f2fs_seek_block(file, offset, whence);
 454        }
 455
 456        return -EINVAL;
 457}
 458
 459static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 460{
 461        struct inode *inode = file_inode(file);
 462        int err;
 463
 464        if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 465                return -EIO;
 466
 467        /* we don't need to use inline_data strictly */
 468        err = f2fs_convert_inline_inode(inode);
 469        if (err)
 470                return err;
 471
 472        file_accessed(file);
 473        vma->vm_ops = &f2fs_file_vm_ops;
 474        return 0;
 475}
 476
 477static int f2fs_file_open(struct inode *inode, struct file *filp)
 478{
 479        int err = fscrypt_file_open(inode, filp);
 480
 481        if (err)
 482                return err;
 483
 484        filp->f_mode |= FMODE_NOWAIT;
 485
 486        return dquot_file_open(inode, filp);
 487}
 488
 489void truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 490{
 491        struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 492        struct f2fs_node *raw_node;
 493        int nr_free = 0, ofs = dn->ofs_in_node, len = count;
 494        __le32 *addr;
 495        int base = 0;
 496
 497        if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
 498                base = get_extra_isize(dn->inode);
 499
 500        raw_node = F2FS_NODE(dn->node_page);
 501        addr = blkaddr_in_node(raw_node) + base + ofs;
 502
 503        for (; count > 0; count--, addr++, dn->ofs_in_node++) {
 504                block_t blkaddr = le32_to_cpu(*addr);
 505                if (blkaddr == NULL_ADDR)
 506                        continue;
 507
 508                dn->data_blkaddr = NULL_ADDR;
 509                set_data_blkaddr(dn);
 510                invalidate_blocks(sbi, blkaddr);
 511                if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 512                        clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 513                nr_free++;
 514        }
 515
 516        if (nr_free) {
 517                pgoff_t fofs;
 518                /*
 519                 * once we invalidate valid blkaddr in range [ofs, ofs + count],
 520                 * we will invalidate all blkaddr in the whole range.
 521                 */
 522                fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
 523                                                        dn->inode) + ofs;
 524                f2fs_update_extent_cache_range(dn, fofs, 0, len);
 525                dec_valid_block_count(sbi, dn->inode, nr_free);
 526        }
 527        dn->ofs_in_node = ofs;
 528
 529        f2fs_update_time(sbi, REQ_TIME);
 530        trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 531                                         dn->ofs_in_node, nr_free);
 532}
 533
 534void truncate_data_blocks(struct dnode_of_data *dn)
 535{
 536        truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
 537}
 538
 539static int truncate_partial_data_page(struct inode *inode, u64 from,
 540                                                                bool cache_only)
 541{
 542        unsigned offset = from & (PAGE_SIZE - 1);
 543        pgoff_t index = from >> PAGE_SHIFT;
 544        struct address_space *mapping = inode->i_mapping;
 545        struct page *page;
 546
 547        if (!offset && !cache_only)
 548                return 0;
 549
 550        if (cache_only) {
 551                page = find_lock_page(mapping, index);
 552                if (page && PageUptodate(page))
 553                        goto truncate_out;
 554                f2fs_put_page(page, 1);
 555                return 0;
 556        }
 557
 558        page = get_lock_data_page(inode, index, true);
 559        if (IS_ERR(page))
 560                return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 561truncate_out:
 562        f2fs_wait_on_page_writeback(page, DATA, true);
 563        zero_user(page, offset, PAGE_SIZE - offset);
 564
 565        /* An encrypted inode should have a key and truncate the last page. */
 566        f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
 567        if (!cache_only)
 568                set_page_dirty(page);
 569        f2fs_put_page(page, 1);
 570        return 0;
 571}
 572
 573int truncate_blocks(struct inode *inode, u64 from, bool lock)
 574{
 575        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 576        struct dnode_of_data dn;
 577        pgoff_t free_from;
 578        int count = 0, err = 0;
 579        struct page *ipage;
 580        bool truncate_page = false;
 581
 582        trace_f2fs_truncate_blocks_enter(inode, from);
 583
 584        free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 585
 586        if (free_from >= sbi->max_file_blocks)
 587                goto free_partial;
 588
 589        if (lock)
 590                f2fs_lock_op(sbi);
 591
 592        ipage = get_node_page(sbi, inode->i_ino);
 593        if (IS_ERR(ipage)) {
 594                err = PTR_ERR(ipage);
 595                goto out;
 596        }
 597
 598        if (f2fs_has_inline_data(inode)) {
 599                truncate_inline_inode(inode, ipage, from);
 600                f2fs_put_page(ipage, 1);
 601                truncate_page = true;
 602                goto out;
 603        }
 604
 605        set_new_dnode(&dn, inode, ipage, NULL, 0);
 606        err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 607        if (err) {
 608                if (err == -ENOENT)
 609                        goto free_next;
 610                goto out;
 611        }
 612
 613        count = ADDRS_PER_PAGE(dn.node_page, inode);
 614
 615        count -= dn.ofs_in_node;
 616        f2fs_bug_on(sbi, count < 0);
 617
 618        if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 619                truncate_data_blocks_range(&dn, count);
 620                free_from += count;
 621        }
 622
 623        f2fs_put_dnode(&dn);
 624free_next:
 625        err = truncate_inode_blocks(inode, free_from);
 626out:
 627        if (lock)
 628                f2fs_unlock_op(sbi);
 629free_partial:
 630        /* lastly zero out the first data page */
 631        if (!err)
 632                err = truncate_partial_data_page(inode, from, truncate_page);
 633
 634        trace_f2fs_truncate_blocks_exit(inode, err);
 635        return err;
 636}
 637
 638int f2fs_truncate(struct inode *inode)
 639{
 640        int err;
 641
 642        if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 643                return -EIO;
 644
 645        if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 646                                S_ISLNK(inode->i_mode)))
 647                return 0;
 648
 649        trace_f2fs_truncate(inode);
 650
 651#ifdef CONFIG_F2FS_FAULT_INJECTION
 652        if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
 653                f2fs_show_injection_info(FAULT_TRUNCATE);
 654                return -EIO;
 655        }
 656#endif
 657        /* we should check inline_data size */
 658        if (!f2fs_may_inline_data(inode)) {
 659                err = f2fs_convert_inline_inode(inode);
 660                if (err)
 661                        return err;
 662        }
 663
 664        err = truncate_blocks(inode, i_size_read(inode), true);
 665        if (err)
 666                return err;
 667
 668        inode->i_mtime = inode->i_ctime = current_time(inode);
 669        f2fs_mark_inode_dirty_sync(inode, false);
 670        return 0;
 671}
 672
 673int f2fs_getattr(const struct path *path, struct kstat *stat,
 674                 u32 request_mask, unsigned int query_flags)
 675{
 676        struct inode *inode = d_inode(path->dentry);
 677        struct f2fs_inode_info *fi = F2FS_I(inode);
 678        struct f2fs_inode *ri;
 679        unsigned int flags;
 680
 681        if (f2fs_has_extra_attr(inode) &&
 682                        f2fs_sb_has_inode_crtime(inode->i_sb) &&
 683                        F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 684                stat->result_mask |= STATX_BTIME;
 685                stat->btime.tv_sec = fi->i_crtime.tv_sec;
 686                stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 687        }
 688
 689        flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
 690        if (flags & FS_APPEND_FL)
 691                stat->attributes |= STATX_ATTR_APPEND;
 692        if (flags & FS_COMPR_FL)
 693                stat->attributes |= STATX_ATTR_COMPRESSED;
 694        if (f2fs_encrypted_inode(inode))
 695                stat->attributes |= STATX_ATTR_ENCRYPTED;
 696        if (flags & FS_IMMUTABLE_FL)
 697                stat->attributes |= STATX_ATTR_IMMUTABLE;
 698        if (flags & FS_NODUMP_FL)
 699                stat->attributes |= STATX_ATTR_NODUMP;
 700
 701        stat->attributes_mask |= (STATX_ATTR_APPEND |
 702                                  STATX_ATTR_COMPRESSED |
 703                                  STATX_ATTR_ENCRYPTED |
 704                                  STATX_ATTR_IMMUTABLE |
 705                                  STATX_ATTR_NODUMP);
 706
 707        generic_fillattr(inode, stat);
 708
 709        /* we need to show initial sectors used for inline_data/dentries */
 710        if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 711                                        f2fs_has_inline_dentry(inode))
 712                stat->blocks += (stat->size + 511) >> 9;
 713
 714        return 0;
 715}
 716
 717#ifdef CONFIG_F2FS_FS_POSIX_ACL
 718static void __setattr_copy(struct inode *inode, const struct iattr *attr)
 719{
 720        unsigned int ia_valid = attr->ia_valid;
 721
 722        if (ia_valid & ATTR_UID)
 723                inode->i_uid = attr->ia_uid;
 724        if (ia_valid & ATTR_GID)
 725                inode->i_gid = attr->ia_gid;
 726        if (ia_valid & ATTR_ATIME)
 727                inode->i_atime = timespec_trunc(attr->ia_atime,
 728                                                inode->i_sb->s_time_gran);
 729        if (ia_valid & ATTR_MTIME)
 730                inode->i_mtime = timespec_trunc(attr->ia_mtime,
 731                                                inode->i_sb->s_time_gran);
 732        if (ia_valid & ATTR_CTIME)
 733                inode->i_ctime = timespec_trunc(attr->ia_ctime,
 734                                                inode->i_sb->s_time_gran);
 735        if (ia_valid & ATTR_MODE) {
 736                umode_t mode = attr->ia_mode;
 737
 738                if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
 739                        mode &= ~S_ISGID;
 740                set_acl_inode(inode, mode);
 741        }
 742}
 743#else
 744#define __setattr_copy setattr_copy
 745#endif
 746
 747int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
 748{
 749        struct inode *inode = d_inode(dentry);
 750        int err;
 751        bool size_changed = false;
 752
 753        if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 754                return -EIO;
 755
 756        err = setattr_prepare(dentry, attr);
 757        if (err)
 758                return err;
 759
 760        err = fscrypt_prepare_setattr(dentry, attr);
 761        if (err)
 762                return err;
 763
 764        if (is_quota_modification(inode, attr)) {
 765                err = dquot_initialize(inode);
 766                if (err)
 767                        return err;
 768        }
 769        if ((attr->ia_valid & ATTR_UID &&
 770                !uid_eq(attr->ia_uid, inode->i_uid)) ||
 771                (attr->ia_valid & ATTR_GID &&
 772                !gid_eq(attr->ia_gid, inode->i_gid))) {
 773                err = dquot_transfer(inode, attr);
 774                if (err)
 775                        return err;
 776        }
 777
 778        if (attr->ia_valid & ATTR_SIZE) {
 779                if (attr->ia_size <= i_size_read(inode)) {
 780                        down_write(&F2FS_I(inode)->i_mmap_sem);
 781                        truncate_setsize(inode, attr->ia_size);
 782                        err = f2fs_truncate(inode);
 783                        up_write(&F2FS_I(inode)->i_mmap_sem);
 784                        if (err)
 785                                return err;
 786                } else {
 787                        /*
 788                         * do not trim all blocks after i_size if target size is
 789                         * larger than i_size.
 790                         */
 791                        down_write(&F2FS_I(inode)->i_mmap_sem);
 792                        truncate_setsize(inode, attr->ia_size);
 793                        up_write(&F2FS_I(inode)->i_mmap_sem);
 794
 795                        /* should convert inline inode here */
 796                        if (!f2fs_may_inline_data(inode)) {
 797                                err = f2fs_convert_inline_inode(inode);
 798                                if (err)
 799                                        return err;
 800                        }
 801                        inode->i_mtime = inode->i_ctime = current_time(inode);
 802                }
 803
 804                down_write(&F2FS_I(inode)->i_sem);
 805                F2FS_I(inode)->last_disk_size = i_size_read(inode);
 806                up_write(&F2FS_I(inode)->i_sem);
 807
 808                size_changed = true;
 809        }
 810
 811        __setattr_copy(inode, attr);
 812
 813        if (attr->ia_valid & ATTR_MODE) {
 814                err = posix_acl_chmod(inode, get_inode_mode(inode));
 815                if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
 816                        inode->i_mode = F2FS_I(inode)->i_acl_mode;
 817                        clear_inode_flag(inode, FI_ACL_MODE);
 818                }
 819        }
 820
 821        /* file size may changed here */
 822        f2fs_mark_inode_dirty_sync(inode, size_changed);
 823
 824        /* inode change will produce dirty node pages flushed by checkpoint */
 825        f2fs_balance_fs(F2FS_I_SB(inode), true);
 826
 827        return err;
 828}
 829
 830const struct inode_operations f2fs_file_inode_operations = {
 831        .getattr        = f2fs_getattr,
 832        .setattr        = f2fs_setattr,
 833        .get_acl        = f2fs_get_acl,
 834        .set_acl        = f2fs_set_acl,
 835#ifdef CONFIG_F2FS_FS_XATTR
 836        .listxattr      = f2fs_listxattr,
 837#endif
 838        .fiemap         = f2fs_fiemap,
 839};
 840
 841static int fill_zero(struct inode *inode, pgoff_t index,
 842                                        loff_t start, loff_t len)
 843{
 844        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 845        struct page *page;
 846
 847        if (!len)
 848                return 0;
 849
 850        f2fs_balance_fs(sbi, true);
 851
 852        f2fs_lock_op(sbi);
 853        page = get_new_data_page(inode, NULL, index, false);
 854        f2fs_unlock_op(sbi);
 855
 856        if (IS_ERR(page))
 857                return PTR_ERR(page);
 858
 859        f2fs_wait_on_page_writeback(page, DATA, true);
 860        zero_user(page, start, len);
 861        set_page_dirty(page);
 862        f2fs_put_page(page, 1);
 863        return 0;
 864}
 865
 866int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
 867{
 868        int err;
 869
 870        while (pg_start < pg_end) {
 871                struct dnode_of_data dn;
 872                pgoff_t end_offset, count;
 873
 874                set_new_dnode(&dn, inode, NULL, NULL, 0);
 875                err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
 876                if (err) {
 877                        if (err == -ENOENT) {
 878                                pg_start = get_next_page_offset(&dn, pg_start);
 879                                continue;
 880                        }
 881                        return err;
 882                }
 883
 884                end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 885                count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
 886
 887                f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
 888
 889                truncate_data_blocks_range(&dn, count);
 890                f2fs_put_dnode(&dn);
 891
 892                pg_start += count;
 893        }
 894        return 0;
 895}
 896
 897static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
 898{
 899        pgoff_t pg_start, pg_end;
 900        loff_t off_start, off_end;
 901        int ret;
 902
 903        ret = f2fs_convert_inline_inode(inode);
 904        if (ret)
 905                return ret;
 906
 907        pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
 908        pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
 909
 910        off_start = offset & (PAGE_SIZE - 1);
 911        off_end = (offset + len) & (PAGE_SIZE - 1);
 912
 913        if (pg_start == pg_end) {
 914                ret = fill_zero(inode, pg_start, off_start,
 915                                                off_end - off_start);
 916                if (ret)
 917                        return ret;
 918        } else {
 919                if (off_start) {
 920                        ret = fill_zero(inode, pg_start++, off_start,
 921                                                PAGE_SIZE - off_start);
 922                        if (ret)
 923                                return ret;
 924                }
 925                if (off_end) {
 926                        ret = fill_zero(inode, pg_end, 0, off_end);
 927                        if (ret)
 928                                return ret;
 929                }
 930
 931                if (pg_start < pg_end) {
 932                        struct address_space *mapping = inode->i_mapping;
 933                        loff_t blk_start, blk_end;
 934                        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 935
 936                        f2fs_balance_fs(sbi, true);
 937
 938                        blk_start = (loff_t)pg_start << PAGE_SHIFT;
 939                        blk_end = (loff_t)pg_end << PAGE_SHIFT;
 940                        down_write(&F2FS_I(inode)->i_mmap_sem);
 941                        truncate_inode_pages_range(mapping, blk_start,
 942                                        blk_end - 1);
 943
 944                        f2fs_lock_op(sbi);
 945                        ret = truncate_hole(inode, pg_start, pg_end);
 946                        f2fs_unlock_op(sbi);
 947                        up_write(&F2FS_I(inode)->i_mmap_sem);
 948                }
 949        }
 950
 951        return ret;
 952}
 953
 954static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
 955                                int *do_replace, pgoff_t off, pgoff_t len)
 956{
 957        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 958        struct dnode_of_data dn;
 959        int ret, done, i;
 960
 961next_dnode:
 962        set_new_dnode(&dn, inode, NULL, NULL, 0);
 963        ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
 964        if (ret && ret != -ENOENT) {
 965                return ret;
 966        } else if (ret == -ENOENT) {
 967                if (dn.max_level == 0)
 968                        return -ENOENT;
 969                done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
 970                blkaddr += done;
 971                do_replace += done;
 972                goto next;
 973        }
 974
 975        done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
 976                                                        dn.ofs_in_node, len);
 977        for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
 978                *blkaddr = datablock_addr(dn.inode,
 979                                        dn.node_page, dn.ofs_in_node);
 980                if (!is_checkpointed_data(sbi, *blkaddr)) {
 981
 982                        if (test_opt(sbi, LFS)) {
 983                                f2fs_put_dnode(&dn);
 984                                return -ENOTSUPP;
 985                        }
 986
 987                        /* do not invalidate this block address */
 988                        f2fs_update_data_blkaddr(&dn, NULL_ADDR);
 989                        *do_replace = 1;
 990                }
 991        }
 992        f2fs_put_dnode(&dn);
 993next:
 994        len -= done;
 995        off += done;
 996        if (len)
 997                goto next_dnode;
 998        return 0;
 999}
1000
1001static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1002                                int *do_replace, pgoff_t off, int len)
1003{
1004        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1005        struct dnode_of_data dn;
1006        int ret, i;
1007
1008        for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1009                if (*do_replace == 0)
1010                        continue;
1011
1012                set_new_dnode(&dn, inode, NULL, NULL, 0);
1013                ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1014                if (ret) {
1015                        dec_valid_block_count(sbi, inode, 1);
1016                        invalidate_blocks(sbi, *blkaddr);
1017                } else {
1018                        f2fs_update_data_blkaddr(&dn, *blkaddr);
1019                }
1020                f2fs_put_dnode(&dn);
1021        }
1022        return 0;
1023}
1024
1025static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1026                        block_t *blkaddr, int *do_replace,
1027                        pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1028{
1029        struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1030        pgoff_t i = 0;
1031        int ret;
1032
1033        while (i < len) {
1034                if (blkaddr[i] == NULL_ADDR && !full) {
1035                        i++;
1036                        continue;
1037                }
1038
1039                if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1040                        struct dnode_of_data dn;
1041                        struct node_info ni;
1042                        size_t new_size;
1043                        pgoff_t ilen;
1044
1045                        set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1046                        ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1047                        if (ret)
1048                                return ret;
1049
1050                        get_node_info(sbi, dn.nid, &ni);
1051                        ilen = min((pgoff_t)
1052                                ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1053                                                dn.ofs_in_node, len - i);
1054                        do {
1055                                dn.data_blkaddr = datablock_addr(dn.inode,
1056                                                dn.node_page, dn.ofs_in_node);
1057                                truncate_data_blocks_range(&dn, 1);
1058
1059                                if (do_replace[i]) {
1060                                        f2fs_i_blocks_write(src_inode,
1061                                                        1, false, false);
1062                                        f2fs_i_blocks_write(dst_inode,
1063                                                        1, true, false);
1064                                        f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1065                                        blkaddr[i], ni.version, true, false);
1066
1067                                        do_replace[i] = 0;
1068                                }
1069                                dn.ofs_in_node++;
1070                                i++;
1071                                new_size = (dst + i) << PAGE_SHIFT;
1072                                if (dst_inode->i_size < new_size)
1073                                        f2fs_i_size_write(dst_inode, new_size);
1074                        } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1075
1076                        f2fs_put_dnode(&dn);
1077                } else {
1078                        struct page *psrc, *pdst;
1079
1080                        psrc = get_lock_data_page(src_inode, src + i, true);
1081                        if (IS_ERR(psrc))
1082                                return PTR_ERR(psrc);
1083                        pdst = get_new_data_page(dst_inode, NULL, dst + i,
1084                                                                true);
1085                        if (IS_ERR(pdst)) {
1086                                f2fs_put_page(psrc, 1);
1087                                return PTR_ERR(pdst);
1088                        }
1089                        f2fs_copy_page(psrc, pdst);
1090                        set_page_dirty(pdst);
1091                        f2fs_put_page(pdst, 1);
1092                        f2fs_put_page(psrc, 1);
1093
1094                        ret = truncate_hole(src_inode, src + i, src + i + 1);
1095                        if (ret)
1096                                return ret;
1097                        i++;
1098                }
1099        }
1100        return 0;
1101}
1102
1103static int __exchange_data_block(struct inode *src_inode,
1104                        struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1105                        pgoff_t len, bool full)
1106{
1107        block_t *src_blkaddr;
1108        int *do_replace;
1109        pgoff_t olen;
1110        int ret;
1111
1112        while (len) {
1113                olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1114
1115                src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1116                                        sizeof(block_t) * olen, GFP_KERNEL);
1117                if (!src_blkaddr)
1118                        return -ENOMEM;
1119
1120                do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1121                                        sizeof(int) * olen, GFP_KERNEL);
1122                if (!do_replace) {
1123                        kvfree(src_blkaddr);
1124                        return -ENOMEM;
1125                }
1126
1127                ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1128                                        do_replace, src, olen);
1129                if (ret)
1130                        goto roll_back;
1131
1132                ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1133                                        do_replace, src, dst, olen, full);
1134                if (ret)
1135                        goto roll_back;
1136
1137                src += olen;
1138                dst += olen;
1139                len -= olen;
1140
1141                kvfree(src_blkaddr);
1142                kvfree(do_replace);
1143        }
1144        return 0;
1145
1146roll_back:
1147        __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1148        kvfree(src_blkaddr);
1149        kvfree(do_replace);
1150        return ret;
1151}
1152
1153static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1154{
1155        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1156        pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1157        int ret;
1158
1159        f2fs_balance_fs(sbi, true);
1160        f2fs_lock_op(sbi);
1161
1162        f2fs_drop_extent_tree(inode);
1163
1164        ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1165        f2fs_unlock_op(sbi);
1166        return ret;
1167}
1168
1169static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1170{
1171        pgoff_t pg_start, pg_end;
1172        loff_t new_size;
1173        int ret;
1174
1175        if (offset + len >= i_size_read(inode))
1176                return -EINVAL;
1177
1178        /* collapse range should be aligned to block size of f2fs. */
1179        if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1180                return -EINVAL;
1181
1182        ret = f2fs_convert_inline_inode(inode);
1183        if (ret)
1184                return ret;
1185
1186        pg_start = offset >> PAGE_SHIFT;
1187        pg_end = (offset + len) >> PAGE_SHIFT;
1188
1189        /* avoid gc operation during block exchange */
1190        down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1191
1192        down_write(&F2FS_I(inode)->i_mmap_sem);
1193        /* write out all dirty pages from offset */
1194        ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1195        if (ret)
1196                goto out_unlock;
1197
1198        truncate_pagecache(inode, offset);
1199
1200        ret = f2fs_do_collapse(inode, pg_start, pg_end);
1201        if (ret)
1202                goto out_unlock;
1203
1204        /* write out all moved pages, if possible */
1205        filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1206        truncate_pagecache(inode, offset);
1207
1208        new_size = i_size_read(inode) - len;
1209        truncate_pagecache(inode, new_size);
1210
1211        ret = truncate_blocks(inode, new_size, true);
1212        if (!ret)
1213                f2fs_i_size_write(inode, new_size);
1214out_unlock:
1215        up_write(&F2FS_I(inode)->i_mmap_sem);
1216        up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1217        return ret;
1218}
1219
1220static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1221                                                                pgoff_t end)
1222{
1223        struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1224        pgoff_t index = start;
1225        unsigned int ofs_in_node = dn->ofs_in_node;
1226        blkcnt_t count = 0;
1227        int ret;
1228
1229        for (; index < end; index++, dn->ofs_in_node++) {
1230                if (datablock_addr(dn->inode, dn->node_page,
1231                                        dn->ofs_in_node) == NULL_ADDR)
1232                        count++;
1233        }
1234
1235        dn->ofs_in_node = ofs_in_node;
1236        ret = reserve_new_blocks(dn, count);
1237        if (ret)
1238                return ret;
1239
1240        dn->ofs_in_node = ofs_in_node;
1241        for (index = start; index < end; index++, dn->ofs_in_node++) {
1242                dn->data_blkaddr = datablock_addr(dn->inode,
1243                                        dn->node_page, dn->ofs_in_node);
1244                /*
1245                 * reserve_new_blocks will not guarantee entire block
1246                 * allocation.
1247                 */
1248                if (dn->data_blkaddr == NULL_ADDR) {
1249                        ret = -ENOSPC;
1250                        break;
1251                }
1252                if (dn->data_blkaddr != NEW_ADDR) {
1253                        invalidate_blocks(sbi, dn->data_blkaddr);
1254                        dn->data_blkaddr = NEW_ADDR;
1255                        set_data_blkaddr(dn);
1256                }
1257        }
1258
1259        f2fs_update_extent_cache_range(dn, start, 0, index - start);
1260
1261        return ret;
1262}
1263
1264static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1265                                                                int mode)
1266{
1267        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1268        struct address_space *mapping = inode->i_mapping;
1269        pgoff_t index, pg_start, pg_end;
1270        loff_t new_size = i_size_read(inode);
1271        loff_t off_start, off_end;
1272        int ret = 0;
1273
1274        ret = inode_newsize_ok(inode, (len + offset));
1275        if (ret)
1276                return ret;
1277
1278        ret = f2fs_convert_inline_inode(inode);
1279        if (ret)
1280                return ret;
1281
1282        down_write(&F2FS_I(inode)->i_mmap_sem);
1283        ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1284        if (ret)
1285                goto out_sem;
1286
1287        truncate_pagecache_range(inode, offset, offset + len - 1);
1288
1289        pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1290        pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1291
1292        off_start = offset & (PAGE_SIZE - 1);
1293        off_end = (offset + len) & (PAGE_SIZE - 1);
1294
1295        if (pg_start == pg_end) {
1296                ret = fill_zero(inode, pg_start, off_start,
1297                                                off_end - off_start);
1298                if (ret)
1299                        goto out_sem;
1300
1301                new_size = max_t(loff_t, new_size, offset + len);
1302        } else {
1303                if (off_start) {
1304                        ret = fill_zero(inode, pg_start++, off_start,
1305                                                PAGE_SIZE - off_start);
1306                        if (ret)
1307                                goto out_sem;
1308
1309                        new_size = max_t(loff_t, new_size,
1310                                        (loff_t)pg_start << PAGE_SHIFT);
1311                }
1312
1313                for (index = pg_start; index < pg_end;) {
1314                        struct dnode_of_data dn;
1315                        unsigned int end_offset;
1316                        pgoff_t end;
1317
1318                        f2fs_lock_op(sbi);
1319
1320                        set_new_dnode(&dn, inode, NULL, NULL, 0);
1321                        ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1322                        if (ret) {
1323                                f2fs_unlock_op(sbi);
1324                                goto out;
1325                        }
1326
1327                        end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1328                        end = min(pg_end, end_offset - dn.ofs_in_node + index);
1329
1330                        ret = f2fs_do_zero_range(&dn, index, end);
1331                        f2fs_put_dnode(&dn);
1332                        f2fs_unlock_op(sbi);
1333
1334                        f2fs_balance_fs(sbi, dn.node_changed);
1335
1336                        if (ret)
1337                                goto out;
1338
1339                        index = end;
1340                        new_size = max_t(loff_t, new_size,
1341                                        (loff_t)index << PAGE_SHIFT);
1342                }
1343
1344                if (off_end) {
1345                        ret = fill_zero(inode, pg_end, 0, off_end);
1346                        if (ret)
1347                                goto out;
1348
1349                        new_size = max_t(loff_t, new_size, offset + len);
1350                }
1351        }
1352
1353out:
1354        if (new_size > i_size_read(inode)) {
1355                if (mode & FALLOC_FL_KEEP_SIZE)
1356                        file_set_keep_isize(inode);
1357                else
1358                        f2fs_i_size_write(inode, new_size);
1359        }
1360out_sem:
1361        up_write(&F2FS_I(inode)->i_mmap_sem);
1362
1363        return ret;
1364}
1365
1366static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1367{
1368        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1369        pgoff_t nr, pg_start, pg_end, delta, idx;
1370        loff_t new_size;
1371        int ret = 0;
1372
1373        new_size = i_size_read(inode) + len;
1374        ret = inode_newsize_ok(inode, new_size);
1375        if (ret)
1376                return ret;
1377
1378        if (offset >= i_size_read(inode))
1379                return -EINVAL;
1380
1381        /* insert range should be aligned to block size of f2fs. */
1382        if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1383                return -EINVAL;
1384
1385        ret = f2fs_convert_inline_inode(inode);
1386        if (ret)
1387                return ret;
1388
1389        f2fs_balance_fs(sbi, true);
1390
1391        /* avoid gc operation during block exchange */
1392        down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1393
1394        down_write(&F2FS_I(inode)->i_mmap_sem);
1395        ret = truncate_blocks(inode, i_size_read(inode), true);
1396        if (ret)
1397                goto out;
1398
1399        /* write out all dirty pages from offset */
1400        ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1401        if (ret)
1402                goto out;
1403
1404        truncate_pagecache(inode, offset);
1405
1406        pg_start = offset >> PAGE_SHIFT;
1407        pg_end = (offset + len) >> PAGE_SHIFT;
1408        delta = pg_end - pg_start;
1409        idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1410
1411        while (!ret && idx > pg_start) {
1412                nr = idx - pg_start;
1413                if (nr > delta)
1414                        nr = delta;
1415                idx -= nr;
1416
1417                f2fs_lock_op(sbi);
1418                f2fs_drop_extent_tree(inode);
1419
1420                ret = __exchange_data_block(inode, inode, idx,
1421                                        idx + delta, nr, false);
1422                f2fs_unlock_op(sbi);
1423        }
1424
1425        /* write out all moved pages, if possible */
1426        filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1427        truncate_pagecache(inode, offset);
1428
1429        if (!ret)
1430                f2fs_i_size_write(inode, new_size);
1431out:
1432        up_write(&F2FS_I(inode)->i_mmap_sem);
1433        up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1434        return ret;
1435}
1436
1437static int expand_inode_data(struct inode *inode, loff_t offset,
1438                                        loff_t len, int mode)
1439{
1440        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1441        struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1442                        .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1443        pgoff_t pg_end;
1444        loff_t new_size = i_size_read(inode);
1445        loff_t off_end;
1446        int err;
1447
1448        err = inode_newsize_ok(inode, (len + offset));
1449        if (err)
1450                return err;
1451
1452        err = f2fs_convert_inline_inode(inode);
1453        if (err)
1454                return err;
1455
1456        f2fs_balance_fs(sbi, true);
1457
1458        pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1459        off_end = (offset + len) & (PAGE_SIZE - 1);
1460
1461        map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1462        map.m_len = pg_end - map.m_lblk;
1463        if (off_end)
1464                map.m_len++;
1465
1466        err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1467        if (err) {
1468                pgoff_t last_off;
1469
1470                if (!map.m_len)
1471                        return err;
1472
1473                last_off = map.m_lblk + map.m_len - 1;
1474
1475                /* update new size to the failed position */
1476                new_size = (last_off == pg_end) ? offset + len:
1477                                        (loff_t)(last_off + 1) << PAGE_SHIFT;
1478        } else {
1479                new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1480        }
1481
1482        if (new_size > i_size_read(inode)) {
1483                if (mode & FALLOC_FL_KEEP_SIZE)
1484                        file_set_keep_isize(inode);
1485                else
1486                        f2fs_i_size_write(inode, new_size);
1487        }
1488
1489        return err;
1490}
1491
1492static long f2fs_fallocate(struct file *file, int mode,
1493                                loff_t offset, loff_t len)
1494{
1495        struct inode *inode = file_inode(file);
1496        long ret = 0;
1497
1498        if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1499                return -EIO;
1500
1501        /* f2fs only support ->fallocate for regular file */
1502        if (!S_ISREG(inode->i_mode))
1503                return -EINVAL;
1504
1505        if (f2fs_encrypted_inode(inode) &&
1506                (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1507                return -EOPNOTSUPP;
1508
1509        if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1510                        FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1511                        FALLOC_FL_INSERT_RANGE))
1512                return -EOPNOTSUPP;
1513
1514        inode_lock(inode);
1515
1516        if (mode & FALLOC_FL_PUNCH_HOLE) {
1517                if (offset >= inode->i_size)
1518                        goto out;
1519
1520                ret = punch_hole(inode, offset, len);
1521        } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1522                ret = f2fs_collapse_range(inode, offset, len);
1523        } else if (mode & FALLOC_FL_ZERO_RANGE) {
1524                ret = f2fs_zero_range(inode, offset, len, mode);
1525        } else if (mode & FALLOC_FL_INSERT_RANGE) {
1526                ret = f2fs_insert_range(inode, offset, len);
1527        } else {
1528                ret = expand_inode_data(inode, offset, len, mode);
1529        }
1530
1531        if (!ret) {
1532                inode->i_mtime = inode->i_ctime = current_time(inode);
1533                f2fs_mark_inode_dirty_sync(inode, false);
1534                f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1535        }
1536
1537out:
1538        inode_unlock(inode);
1539
1540        trace_f2fs_fallocate(inode, mode, offset, len, ret);
1541        return ret;
1542}
1543
1544static int f2fs_release_file(struct inode *inode, struct file *filp)
1545{
1546        /*
1547         * f2fs_relase_file is called at every close calls. So we should
1548         * not drop any inmemory pages by close called by other process.
1549         */
1550        if (!(filp->f_mode & FMODE_WRITE) ||
1551                        atomic_read(&inode->i_writecount) != 1)
1552                return 0;
1553
1554        /* some remained atomic pages should discarded */
1555        if (f2fs_is_atomic_file(inode))
1556                drop_inmem_pages(inode);
1557        if (f2fs_is_volatile_file(inode)) {
1558                clear_inode_flag(inode, FI_VOLATILE_FILE);
1559                stat_dec_volatile_write(inode);
1560                set_inode_flag(inode, FI_DROP_CACHE);
1561                filemap_fdatawrite(inode->i_mapping);
1562                clear_inode_flag(inode, FI_DROP_CACHE);
1563        }
1564        return 0;
1565}
1566
1567static int f2fs_file_flush(struct file *file, fl_owner_t id)
1568{
1569        struct inode *inode = file_inode(file);
1570
1571        /*
1572         * If the process doing a transaction is crashed, we should do
1573         * roll-back. Otherwise, other reader/write can see corrupted database
1574         * until all the writers close its file. Since this should be done
1575         * before dropping file lock, it needs to do in ->flush.
1576         */
1577        if (f2fs_is_atomic_file(inode) &&
1578                        F2FS_I(inode)->inmem_task == current)
1579                drop_inmem_pages(inode);
1580        return 0;
1581}
1582
1583static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1584{
1585        struct inode *inode = file_inode(filp);
1586        struct f2fs_inode_info *fi = F2FS_I(inode);
1587        unsigned int flags = fi->i_flags &
1588                        (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1589        return put_user(flags, (int __user *)arg);
1590}
1591
1592static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1593{
1594        struct f2fs_inode_info *fi = F2FS_I(inode);
1595        unsigned int oldflags;
1596
1597        /* Is it quota file? Do not allow user to mess with it */
1598        if (IS_NOQUOTA(inode))
1599                return -EPERM;
1600
1601        flags = f2fs_mask_flags(inode->i_mode, flags);
1602
1603        oldflags = fi->i_flags;
1604
1605        if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1606                if (!capable(CAP_LINUX_IMMUTABLE))
1607                        return -EPERM;
1608
1609        flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1610        flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1611        fi->i_flags = flags;
1612
1613        if (fi->i_flags & FS_PROJINHERIT_FL)
1614                set_inode_flag(inode, FI_PROJ_INHERIT);
1615        else
1616                clear_inode_flag(inode, FI_PROJ_INHERIT);
1617
1618        inode->i_ctime = current_time(inode);
1619        f2fs_set_inode_flags(inode);
1620        f2fs_mark_inode_dirty_sync(inode, false);
1621        return 0;
1622}
1623
1624static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1625{
1626        struct inode *inode = file_inode(filp);
1627        unsigned int flags;
1628        int ret;
1629
1630        if (!inode_owner_or_capable(inode))
1631                return -EACCES;
1632
1633        if (get_user(flags, (int __user *)arg))
1634                return -EFAULT;
1635
1636        ret = mnt_want_write_file(filp);
1637        if (ret)
1638                return ret;
1639
1640        inode_lock(inode);
1641
1642        ret = __f2fs_ioc_setflags(inode, flags);
1643
1644        inode_unlock(inode);
1645        mnt_drop_write_file(filp);
1646        return ret;
1647}
1648
1649static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1650{
1651        struct inode *inode = file_inode(filp);
1652
1653        return put_user(inode->i_generation, (int __user *)arg);
1654}
1655
1656static int f2fs_ioc_start_atomic_write(struct file *filp)
1657{
1658        struct inode *inode = file_inode(filp);
1659        int ret;
1660
1661        if (!inode_owner_or_capable(inode))
1662                return -EACCES;
1663
1664        if (!S_ISREG(inode->i_mode))
1665                return -EINVAL;
1666
1667        ret = mnt_want_write_file(filp);
1668        if (ret)
1669                return ret;
1670
1671        inode_lock(inode);
1672
1673        if (f2fs_is_atomic_file(inode))
1674                goto out;
1675
1676        ret = f2fs_convert_inline_inode(inode);
1677        if (ret)
1678                goto out;
1679
1680        set_inode_flag(inode, FI_ATOMIC_FILE);
1681        set_inode_flag(inode, FI_HOT_DATA);
1682        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1683
1684        if (!get_dirty_pages(inode))
1685                goto inc_stat;
1686
1687        f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1688                "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1689                                        inode->i_ino, get_dirty_pages(inode));
1690        ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1691        if (ret) {
1692                clear_inode_flag(inode, FI_ATOMIC_FILE);
1693                clear_inode_flag(inode, FI_HOT_DATA);
1694                goto out;
1695        }
1696
1697inc_stat:
1698        F2FS_I(inode)->inmem_task = current;
1699        stat_inc_atomic_write(inode);
1700        stat_update_max_atomic_write(inode);
1701out:
1702        inode_unlock(inode);
1703        mnt_drop_write_file(filp);
1704        return ret;
1705}
1706
1707static int f2fs_ioc_commit_atomic_write(struct file *filp)
1708{
1709        struct inode *inode = file_inode(filp);
1710        int ret;
1711
1712        if (!inode_owner_or_capable(inode))
1713                return -EACCES;
1714
1715        ret = mnt_want_write_file(filp);
1716        if (ret)
1717                return ret;
1718
1719        inode_lock(inode);
1720
1721        down_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1722
1723        if (f2fs_is_volatile_file(inode))
1724                goto err_out;
1725
1726        if (f2fs_is_atomic_file(inode)) {
1727                ret = commit_inmem_pages(inode);
1728                if (ret)
1729                        goto err_out;
1730
1731                ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1732                if (!ret) {
1733                        clear_inode_flag(inode, FI_ATOMIC_FILE);
1734                        clear_inode_flag(inode, FI_HOT_DATA);
1735                        stat_dec_atomic_write(inode);
1736                }
1737        } else {
1738                ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1739        }
1740err_out:
1741        up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
1742        inode_unlock(inode);
1743        mnt_drop_write_file(filp);
1744        return ret;
1745}
1746
1747static int f2fs_ioc_start_volatile_write(struct file *filp)
1748{
1749        struct inode *inode = file_inode(filp);
1750        int ret;
1751
1752        if (!inode_owner_or_capable(inode))
1753                return -EACCES;
1754
1755        if (!S_ISREG(inode->i_mode))
1756                return -EINVAL;
1757
1758        ret = mnt_want_write_file(filp);
1759        if (ret)
1760                return ret;
1761
1762        inode_lock(inode);
1763
1764        if (f2fs_is_volatile_file(inode))
1765                goto out;
1766
1767        ret = f2fs_convert_inline_inode(inode);
1768        if (ret)
1769                goto out;
1770
1771        stat_inc_volatile_write(inode);
1772        stat_update_max_volatile_write(inode);
1773
1774        set_inode_flag(inode, FI_VOLATILE_FILE);
1775        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1776out:
1777        inode_unlock(inode);
1778        mnt_drop_write_file(filp);
1779        return ret;
1780}
1781
1782static int f2fs_ioc_release_volatile_write(struct file *filp)
1783{
1784        struct inode *inode = file_inode(filp);
1785        int ret;
1786
1787        if (!inode_owner_or_capable(inode))
1788                return -EACCES;
1789
1790        ret = mnt_want_write_file(filp);
1791        if (ret)
1792                return ret;
1793
1794        inode_lock(inode);
1795
1796        if (!f2fs_is_volatile_file(inode))
1797                goto out;
1798
1799        if (!f2fs_is_first_block_written(inode)) {
1800                ret = truncate_partial_data_page(inode, 0, true);
1801                goto out;
1802        }
1803
1804        ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1805out:
1806        inode_unlock(inode);
1807        mnt_drop_write_file(filp);
1808        return ret;
1809}
1810
1811static int f2fs_ioc_abort_volatile_write(struct file *filp)
1812{
1813        struct inode *inode = file_inode(filp);
1814        int ret;
1815
1816        if (!inode_owner_or_capable(inode))
1817                return -EACCES;
1818
1819        ret = mnt_want_write_file(filp);
1820        if (ret)
1821                return ret;
1822
1823        inode_lock(inode);
1824
1825        if (f2fs_is_atomic_file(inode))
1826                drop_inmem_pages(inode);
1827        if (f2fs_is_volatile_file(inode)) {
1828                clear_inode_flag(inode, FI_VOLATILE_FILE);
1829                stat_dec_volatile_write(inode);
1830                ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1831        }
1832
1833        inode_unlock(inode);
1834
1835        mnt_drop_write_file(filp);
1836        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1837        return ret;
1838}
1839
1840static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1841{
1842        struct inode *inode = file_inode(filp);
1843        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1844        struct super_block *sb = sbi->sb;
1845        __u32 in;
1846        int ret;
1847
1848        if (!capable(CAP_SYS_ADMIN))
1849                return -EPERM;
1850
1851        if (get_user(in, (__u32 __user *)arg))
1852                return -EFAULT;
1853
1854        ret = mnt_want_write_file(filp);
1855        if (ret)
1856                return ret;
1857
1858        switch (in) {
1859        case F2FS_GOING_DOWN_FULLSYNC:
1860                sb = freeze_bdev(sb->s_bdev);
1861                if (IS_ERR(sb)) {
1862                        ret = PTR_ERR(sb);
1863                        goto out;
1864                }
1865                if (sb) {
1866                        f2fs_stop_checkpoint(sbi, false);
1867                        thaw_bdev(sb->s_bdev, sb);
1868                }
1869                break;
1870        case F2FS_GOING_DOWN_METASYNC:
1871                /* do checkpoint only */
1872                ret = f2fs_sync_fs(sb, 1);
1873                if (ret)
1874                        goto out;
1875                f2fs_stop_checkpoint(sbi, false);
1876                break;
1877        case F2FS_GOING_DOWN_NOSYNC:
1878                f2fs_stop_checkpoint(sbi, false);
1879                break;
1880        case F2FS_GOING_DOWN_METAFLUSH:
1881                sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1882                f2fs_stop_checkpoint(sbi, false);
1883                break;
1884        default:
1885                ret = -EINVAL;
1886                goto out;
1887        }
1888
1889        stop_gc_thread(sbi);
1890        stop_discard_thread(sbi);
1891
1892        drop_discard_cmd(sbi);
1893        clear_opt(sbi, DISCARD);
1894
1895        f2fs_update_time(sbi, REQ_TIME);
1896out:
1897        mnt_drop_write_file(filp);
1898        return ret;
1899}
1900
1901static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1902{
1903        struct inode *inode = file_inode(filp);
1904        struct super_block *sb = inode->i_sb;
1905        struct request_queue *q = bdev_get_queue(sb->s_bdev);
1906        struct fstrim_range range;
1907        int ret;
1908
1909        if (!capable(CAP_SYS_ADMIN))
1910                return -EPERM;
1911
1912        if (!blk_queue_discard(q))
1913                return -EOPNOTSUPP;
1914
1915        if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1916                                sizeof(range)))
1917                return -EFAULT;
1918
1919        ret = mnt_want_write_file(filp);
1920        if (ret)
1921                return ret;
1922
1923        range.minlen = max((unsigned int)range.minlen,
1924                                q->limits.discard_granularity);
1925        ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1926        mnt_drop_write_file(filp);
1927        if (ret < 0)
1928                return ret;
1929
1930        if (copy_to_user((struct fstrim_range __user *)arg, &range,
1931                                sizeof(range)))
1932                return -EFAULT;
1933        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1934        return 0;
1935}
1936
1937static bool uuid_is_nonzero(__u8 u[16])
1938{
1939        int i;
1940
1941        for (i = 0; i < 16; i++)
1942                if (u[i])
1943                        return true;
1944        return false;
1945}
1946
1947static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1948{
1949        struct inode *inode = file_inode(filp);
1950
1951        if (!f2fs_sb_has_encrypt(inode->i_sb))
1952                return -EOPNOTSUPP;
1953
1954        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1955
1956        return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1957}
1958
1959static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1960{
1961        if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
1962                return -EOPNOTSUPP;
1963        return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1964}
1965
1966static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1967{
1968        struct inode *inode = file_inode(filp);
1969        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1970        int err;
1971
1972        if (!f2fs_sb_has_encrypt(inode->i_sb))
1973                return -EOPNOTSUPP;
1974
1975        err = mnt_want_write_file(filp);
1976        if (err)
1977                return err;
1978
1979        down_write(&sbi->sb_lock);
1980
1981        if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1982                goto got_it;
1983
1984        /* update superblock with uuid */
1985        generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1986
1987        err = f2fs_commit_super(sbi, false);
1988        if (err) {
1989                /* undo new data */
1990                memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1991                goto out_err;
1992        }
1993got_it:
1994        if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1995                                                                        16))
1996                err = -EFAULT;
1997out_err:
1998        up_write(&sbi->sb_lock);
1999        mnt_drop_write_file(filp);
2000        return err;
2001}
2002
2003static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2004{
2005        struct inode *inode = file_inode(filp);
2006        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2007        __u32 sync;
2008        int ret;
2009
2010        if (!capable(CAP_SYS_ADMIN))
2011                return -EPERM;
2012
2013        if (get_user(sync, (__u32 __user *)arg))
2014                return -EFAULT;
2015
2016        if (f2fs_readonly(sbi->sb))
2017                return -EROFS;
2018
2019        ret = mnt_want_write_file(filp);
2020        if (ret)
2021                return ret;
2022
2023        if (!sync) {
2024                if (!mutex_trylock(&sbi->gc_mutex)) {
2025                        ret = -EBUSY;
2026                        goto out;
2027                }
2028        } else {
2029                mutex_lock(&sbi->gc_mutex);
2030        }
2031
2032        ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2033out:
2034        mnt_drop_write_file(filp);
2035        return ret;
2036}
2037
2038static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2039{
2040        struct inode *inode = file_inode(filp);
2041        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2042        struct f2fs_gc_range range;
2043        u64 end;
2044        int ret;
2045
2046        if (!capable(CAP_SYS_ADMIN))
2047                return -EPERM;
2048
2049        if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2050                                                        sizeof(range)))
2051                return -EFAULT;
2052
2053        if (f2fs_readonly(sbi->sb))
2054                return -EROFS;
2055
2056        ret = mnt_want_write_file(filp);
2057        if (ret)
2058                return ret;
2059
2060        end = range.start + range.len;
2061        if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2062                ret = -EINVAL;
2063                goto out;
2064        }
2065do_more:
2066        if (!range.sync) {
2067                if (!mutex_trylock(&sbi->gc_mutex)) {
2068                        ret = -EBUSY;
2069                        goto out;
2070                }
2071        } else {
2072                mutex_lock(&sbi->gc_mutex);
2073        }
2074
2075        ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2076        range.start += sbi->blocks_per_seg;
2077        if (range.start <= end)
2078                goto do_more;
2079out:
2080        mnt_drop_write_file(filp);
2081        return ret;
2082}
2083
2084static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2085{
2086        struct inode *inode = file_inode(filp);
2087        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2088        int ret;
2089
2090        if (!capable(CAP_SYS_ADMIN))
2091                return -EPERM;
2092
2093        if (f2fs_readonly(sbi->sb))
2094                return -EROFS;
2095
2096        ret = mnt_want_write_file(filp);
2097        if (ret)
2098                return ret;
2099
2100        ret = f2fs_sync_fs(sbi->sb, 1);
2101
2102        mnt_drop_write_file(filp);
2103        return ret;
2104}
2105
2106static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2107                                        struct file *filp,
2108                                        struct f2fs_defragment *range)
2109{
2110        struct inode *inode = file_inode(filp);
2111        struct f2fs_map_blocks map = { .m_next_extent = NULL,
2112                                        .m_seg_type = NO_CHECK_TYPE };
2113        struct extent_info ei = {0,0,0};
2114        pgoff_t pg_start, pg_end, next_pgofs;
2115        unsigned int blk_per_seg = sbi->blocks_per_seg;
2116        unsigned int total = 0, sec_num;
2117        block_t blk_end = 0;
2118        bool fragmented = false;
2119        int err;
2120
2121        /* if in-place-update policy is enabled, don't waste time here */
2122        if (should_update_inplace(inode, NULL))
2123                return -EINVAL;
2124
2125        pg_start = range->start >> PAGE_SHIFT;
2126        pg_end = (range->start + range->len) >> PAGE_SHIFT;
2127
2128        f2fs_balance_fs(sbi, true);
2129
2130        inode_lock(inode);
2131
2132        /* writeback all dirty pages in the range */
2133        err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2134                                                range->start + range->len - 1);
2135        if (err)
2136                goto out;
2137
2138        /*
2139         * lookup mapping info in extent cache, skip defragmenting if physical
2140         * block addresses are continuous.
2141         */
2142        if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2143                if (ei.fofs + ei.len >= pg_end)
2144                        goto out;
2145        }
2146
2147        map.m_lblk = pg_start;
2148        map.m_next_pgofs = &next_pgofs;
2149
2150        /*
2151         * lookup mapping info in dnode page cache, skip defragmenting if all
2152         * physical block addresses are continuous even if there are hole(s)
2153         * in logical blocks.
2154         */
2155        while (map.m_lblk < pg_end) {
2156                map.m_len = pg_end - map.m_lblk;
2157                err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2158                if (err)
2159                        goto out;
2160
2161                if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2162                        map.m_lblk = next_pgofs;
2163                        continue;
2164                }
2165
2166                if (blk_end && blk_end != map.m_pblk)
2167                        fragmented = true;
2168
2169                /* record total count of block that we're going to move */
2170                total += map.m_len;
2171
2172                blk_end = map.m_pblk + map.m_len;
2173
2174                map.m_lblk += map.m_len;
2175        }
2176
2177        if (!fragmented)
2178                goto out;
2179
2180        sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2181
2182        /*
2183         * make sure there are enough free section for LFS allocation, this can
2184         * avoid defragment running in SSR mode when free section are allocated
2185         * intensively
2186         */
2187        if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2188                err = -EAGAIN;
2189                goto out;
2190        }
2191
2192        map.m_lblk = pg_start;
2193        map.m_len = pg_end - pg_start;
2194        total = 0;
2195
2196        while (map.m_lblk < pg_end) {
2197                pgoff_t idx;
2198                int cnt = 0;
2199
2200do_map:
2201                map.m_len = pg_end - map.m_lblk;
2202                err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2203                if (err)
2204                        goto clear_out;
2205
2206                if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2207                        map.m_lblk = next_pgofs;
2208                        continue;
2209                }
2210
2211                set_inode_flag(inode, FI_DO_DEFRAG);
2212
2213                idx = map.m_lblk;
2214                while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2215                        struct page *page;
2216
2217                        page = get_lock_data_page(inode, idx, true);
2218                        if (IS_ERR(page)) {
2219                                err = PTR_ERR(page);
2220                                goto clear_out;
2221                        }
2222
2223                        set_page_dirty(page);
2224                        f2fs_put_page(page, 1);
2225
2226                        idx++;
2227                        cnt++;
2228                        total++;
2229                }
2230
2231                map.m_lblk = idx;
2232
2233                if (idx < pg_end && cnt < blk_per_seg)
2234                        goto do_map;
2235
2236                clear_inode_flag(inode, FI_DO_DEFRAG);
2237
2238                err = filemap_fdatawrite(inode->i_mapping);
2239                if (err)
2240                        goto out;
2241        }
2242clear_out:
2243        clear_inode_flag(inode, FI_DO_DEFRAG);
2244out:
2245        inode_unlock(inode);
2246        if (!err)
2247                range->len = (u64)total << PAGE_SHIFT;
2248        return err;
2249}
2250
2251static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2252{
2253        struct inode *inode = file_inode(filp);
2254        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2255        struct f2fs_defragment range;
2256        int err;
2257
2258        if (!capable(CAP_SYS_ADMIN))
2259                return -EPERM;
2260
2261        if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2262                return -EINVAL;
2263
2264        if (f2fs_readonly(sbi->sb))
2265                return -EROFS;
2266
2267        if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2268                                                        sizeof(range)))
2269                return -EFAULT;
2270
2271        /* verify alignment of offset & size */
2272        if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2273                return -EINVAL;
2274
2275        if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2276                                        sbi->max_file_blocks))
2277                return -EINVAL;
2278
2279        err = mnt_want_write_file(filp);
2280        if (err)
2281                return err;
2282
2283        err = f2fs_defragment_range(sbi, filp, &range);
2284        mnt_drop_write_file(filp);
2285
2286        f2fs_update_time(sbi, REQ_TIME);
2287        if (err < 0)
2288                return err;
2289
2290        if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2291                                                        sizeof(range)))
2292                return -EFAULT;
2293
2294        return 0;
2295}
2296
2297static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2298                        struct file *file_out, loff_t pos_out, size_t len)
2299{
2300        struct inode *src = file_inode(file_in);
2301        struct inode *dst = file_inode(file_out);
2302        struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2303        size_t olen = len, dst_max_i_size = 0;
2304        size_t dst_osize;
2305        int ret;
2306
2307        if (file_in->f_path.mnt != file_out->f_path.mnt ||
2308                                src->i_sb != dst->i_sb)
2309                return -EXDEV;
2310
2311        if (unlikely(f2fs_readonly(src->i_sb)))
2312                return -EROFS;
2313
2314        if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2315                return -EINVAL;
2316
2317        if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2318                return -EOPNOTSUPP;
2319
2320        if (src == dst) {
2321                if (pos_in == pos_out)
2322                        return 0;
2323                if (pos_out > pos_in && pos_out < pos_in + len)
2324                        return -EINVAL;
2325        }
2326
2327        inode_lock(src);
2328        down_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2329        if (src != dst) {
2330                ret = -EBUSY;
2331                if (!inode_trylock(dst))
2332                        goto out;
2333                if (!down_write_trylock(&F2FS_I(dst)->dio_rwsem[WRITE])) {
2334                        inode_unlock(dst);
2335                        goto out;
2336                }
2337        }
2338
2339        ret = -EINVAL;
2340        if (pos_in + len > src->i_size || pos_in + len < pos_in)
2341                goto out_unlock;
2342        if (len == 0)
2343                olen = len = src->i_size - pos_in;
2344        if (pos_in + len == src->i_size)
2345                len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2346        if (len == 0) {
2347                ret = 0;
2348                goto out_unlock;
2349        }
2350
2351        dst_osize = dst->i_size;
2352        if (pos_out + olen > dst->i_size)
2353                dst_max_i_size = pos_out + olen;
2354
2355        /* verify the end result is block aligned */
2356        if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2357                        !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2358                        !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2359                goto out_unlock;
2360
2361        ret = f2fs_convert_inline_inode(src);
2362        if (ret)
2363                goto out_unlock;
2364
2365        ret = f2fs_convert_inline_inode(dst);
2366        if (ret)
2367                goto out_unlock;
2368
2369        /* write out all dirty pages from offset */
2370        ret = filemap_write_and_wait_range(src->i_mapping,
2371                                        pos_in, pos_in + len);
2372        if (ret)
2373                goto out_unlock;
2374
2375        ret = filemap_write_and_wait_range(dst->i_mapping,
2376                                        pos_out, pos_out + len);
2377        if (ret)
2378                goto out_unlock;
2379
2380        f2fs_balance_fs(sbi, true);
2381        f2fs_lock_op(sbi);
2382        ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2383                                pos_out >> F2FS_BLKSIZE_BITS,
2384                                len >> F2FS_BLKSIZE_BITS, false);
2385
2386        if (!ret) {
2387                if (dst_max_i_size)
2388                        f2fs_i_size_write(dst, dst_max_i_size);
2389                else if (dst_osize != dst->i_size)
2390                        f2fs_i_size_write(dst, dst_osize);
2391        }
2392        f2fs_unlock_op(sbi);
2393out_unlock:
2394        if (src != dst) {
2395                up_write(&F2FS_I(dst)->dio_rwsem[WRITE]);
2396                inode_unlock(dst);
2397        }
2398out:
2399        up_write(&F2FS_I(src)->dio_rwsem[WRITE]);
2400        inode_unlock(src);
2401        return ret;
2402}
2403
2404static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2405{
2406        struct f2fs_move_range range;
2407        struct fd dst;
2408        int err;
2409
2410        if (!(filp->f_mode & FMODE_READ) ||
2411                        !(filp->f_mode & FMODE_WRITE))
2412                return -EBADF;
2413
2414        if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2415                                                        sizeof(range)))
2416                return -EFAULT;
2417
2418        dst = fdget(range.dst_fd);
2419        if (!dst.file)
2420                return -EBADF;
2421
2422        if (!(dst.file->f_mode & FMODE_WRITE)) {
2423                err = -EBADF;
2424                goto err_out;
2425        }
2426
2427        err = mnt_want_write_file(filp);
2428        if (err)
2429                goto err_out;
2430
2431        err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2432                                        range.pos_out, range.len);
2433
2434        mnt_drop_write_file(filp);
2435        if (err)
2436                goto err_out;
2437
2438        if (copy_to_user((struct f2fs_move_range __user *)arg,
2439                                                &range, sizeof(range)))
2440                err = -EFAULT;
2441err_out:
2442        fdput(dst);
2443        return err;
2444}
2445
2446static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2447{
2448        struct inode *inode = file_inode(filp);
2449        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2450        struct sit_info *sm = SIT_I(sbi);
2451        unsigned int start_segno = 0, end_segno = 0;
2452        unsigned int dev_start_segno = 0, dev_end_segno = 0;
2453        struct f2fs_flush_device range;
2454        int ret;
2455
2456        if (!capable(CAP_SYS_ADMIN))
2457                return -EPERM;
2458
2459        if (f2fs_readonly(sbi->sb))
2460                return -EROFS;
2461
2462        if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2463                                                        sizeof(range)))
2464                return -EFAULT;
2465
2466        if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2467                        sbi->segs_per_sec != 1) {
2468                f2fs_msg(sbi->sb, KERN_WARNING,
2469                        "Can't flush %u in %d for segs_per_sec %u != 1\n",
2470                                range.dev_num, sbi->s_ndevs,
2471                                sbi->segs_per_sec);
2472                return -EINVAL;
2473        }
2474
2475        ret = mnt_want_write_file(filp);
2476        if (ret)
2477                return ret;
2478
2479        if (range.dev_num != 0)
2480                dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2481        dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2482
2483        start_segno = sm->last_victim[FLUSH_DEVICE];
2484        if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2485                start_segno = dev_start_segno;
2486        end_segno = min(start_segno + range.segments, dev_end_segno);
2487
2488        while (start_segno < end_segno) {
2489                if (!mutex_trylock(&sbi->gc_mutex)) {
2490                        ret = -EBUSY;
2491                        goto out;
2492                }
2493                sm->last_victim[GC_CB] = end_segno + 1;
2494                sm->last_victim[GC_GREEDY] = end_segno + 1;
2495                sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2496                ret = f2fs_gc(sbi, true, true, start_segno);
2497                if (ret == -EAGAIN)
2498                        ret = 0;
2499                else if (ret < 0)
2500                        break;
2501                start_segno++;
2502        }
2503out:
2504        mnt_drop_write_file(filp);
2505        return ret;
2506}
2507
2508static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2509{
2510        struct inode *inode = file_inode(filp);
2511        u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2512
2513        /* Must validate to set it with SQLite behavior in Android. */
2514        sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2515
2516        return put_user(sb_feature, (u32 __user *)arg);
2517}
2518
2519#ifdef CONFIG_QUOTA
2520static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2521{
2522        struct inode *inode = file_inode(filp);
2523        struct f2fs_inode_info *fi = F2FS_I(inode);
2524        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2525        struct super_block *sb = sbi->sb;
2526        struct dquot *transfer_to[MAXQUOTAS] = {};
2527        struct page *ipage;
2528        kprojid_t kprojid;
2529        int err;
2530
2531        if (!f2fs_sb_has_project_quota(sb)) {
2532                if (projid != F2FS_DEF_PROJID)
2533                        return -EOPNOTSUPP;
2534                else
2535                        return 0;
2536        }
2537
2538        if (!f2fs_has_extra_attr(inode))
2539                return -EOPNOTSUPP;
2540
2541        kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2542
2543        if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2544                return 0;
2545
2546        err = mnt_want_write_file(filp);
2547        if (err)
2548                return err;
2549
2550        err = -EPERM;
2551        inode_lock(inode);
2552
2553        /* Is it quota file? Do not allow user to mess with it */
2554        if (IS_NOQUOTA(inode))
2555                goto out_unlock;
2556
2557        ipage = get_node_page(sbi, inode->i_ino);
2558        if (IS_ERR(ipage)) {
2559                err = PTR_ERR(ipage);
2560                goto out_unlock;
2561        }
2562
2563        if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2564                                                                i_projid)) {
2565                err = -EOVERFLOW;
2566                f2fs_put_page(ipage, 1);
2567                goto out_unlock;
2568        }
2569        f2fs_put_page(ipage, 1);
2570
2571        dquot_initialize(inode);
2572
2573        transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2574        if (!IS_ERR(transfer_to[PRJQUOTA])) {
2575                err = __dquot_transfer(inode, transfer_to);
2576                dqput(transfer_to[PRJQUOTA]);
2577                if (err)
2578                        goto out_dirty;
2579        }
2580
2581        F2FS_I(inode)->i_projid = kprojid;
2582        inode->i_ctime = current_time(inode);
2583out_dirty:
2584        f2fs_mark_inode_dirty_sync(inode, true);
2585out_unlock:
2586        inode_unlock(inode);
2587        mnt_drop_write_file(filp);
2588        return err;
2589}
2590#else
2591static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2592{
2593        if (projid != F2FS_DEF_PROJID)
2594                return -EOPNOTSUPP;
2595        return 0;
2596}
2597#endif
2598
2599/* Transfer internal flags to xflags */
2600static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2601{
2602        __u32 xflags = 0;
2603
2604        if (iflags & FS_SYNC_FL)
2605                xflags |= FS_XFLAG_SYNC;
2606        if (iflags & FS_IMMUTABLE_FL)
2607                xflags |= FS_XFLAG_IMMUTABLE;
2608        if (iflags & FS_APPEND_FL)
2609                xflags |= FS_XFLAG_APPEND;
2610        if (iflags & FS_NODUMP_FL)
2611                xflags |= FS_XFLAG_NODUMP;
2612        if (iflags & FS_NOATIME_FL)
2613                xflags |= FS_XFLAG_NOATIME;
2614        if (iflags & FS_PROJINHERIT_FL)
2615                xflags |= FS_XFLAG_PROJINHERIT;
2616        return xflags;
2617}
2618
2619#define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2620                                  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2621                                  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2622
2623/* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2624#define F2FS_FL_XFLAG_VISIBLE           (FS_SYNC_FL | \
2625                                         FS_IMMUTABLE_FL | \
2626                                         FS_APPEND_FL | \
2627                                         FS_NODUMP_FL | \
2628                                         FS_NOATIME_FL | \
2629                                         FS_PROJINHERIT_FL)
2630
2631/* Transfer xflags flags to internal */
2632static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2633{
2634        unsigned long iflags = 0;
2635
2636        if (xflags & FS_XFLAG_SYNC)
2637                iflags |= FS_SYNC_FL;
2638        if (xflags & FS_XFLAG_IMMUTABLE)
2639                iflags |= FS_IMMUTABLE_FL;
2640        if (xflags & FS_XFLAG_APPEND)
2641                iflags |= FS_APPEND_FL;
2642        if (xflags & FS_XFLAG_NODUMP)
2643                iflags |= FS_NODUMP_FL;
2644        if (xflags & FS_XFLAG_NOATIME)
2645                iflags |= FS_NOATIME_FL;
2646        if (xflags & FS_XFLAG_PROJINHERIT)
2647                iflags |= FS_PROJINHERIT_FL;
2648
2649        return iflags;
2650}
2651
2652static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2653{
2654        struct inode *inode = file_inode(filp);
2655        struct f2fs_inode_info *fi = F2FS_I(inode);
2656        struct fsxattr fa;
2657
2658        memset(&fa, 0, sizeof(struct fsxattr));
2659        fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2660                                (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2661
2662        if (f2fs_sb_has_project_quota(inode->i_sb))
2663                fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2664                                                        fi->i_projid);
2665
2666        if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2667                return -EFAULT;
2668        return 0;
2669}
2670
2671static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2672{
2673        struct inode *inode = file_inode(filp);
2674        struct f2fs_inode_info *fi = F2FS_I(inode);
2675        struct fsxattr fa;
2676        unsigned int flags;
2677        int err;
2678
2679        if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2680                return -EFAULT;
2681
2682        /* Make sure caller has proper permission */
2683        if (!inode_owner_or_capable(inode))
2684                return -EACCES;
2685
2686        if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2687                return -EOPNOTSUPP;
2688
2689        flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2690        if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2691                return -EOPNOTSUPP;
2692
2693        err = mnt_want_write_file(filp);
2694        if (err)
2695                return err;
2696
2697        inode_lock(inode);
2698        flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2699                                (flags & F2FS_FL_XFLAG_VISIBLE);
2700        err = __f2fs_ioc_setflags(inode, flags);
2701        inode_unlock(inode);
2702        mnt_drop_write_file(filp);
2703        if (err)
2704                return err;
2705
2706        err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2707        if (err)
2708                return err;
2709
2710        return 0;
2711}
2712
2713int f2fs_pin_file_control(struct inode *inode, bool inc)
2714{
2715        struct f2fs_inode_info *fi = F2FS_I(inode);
2716        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2717
2718        /* Use i_gc_failures for normal file as a risk signal. */
2719        if (inc)
2720                f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
2721
2722        if (fi->i_gc_failures > sbi->gc_pin_file_threshold) {
2723                f2fs_msg(sbi->sb, KERN_WARNING,
2724                        "%s: Enable GC = ino %lx after %x GC trials\n",
2725                        __func__, inode->i_ino, fi->i_gc_failures);
2726                clear_inode_flag(inode, FI_PIN_FILE);
2727                return -EAGAIN;
2728        }
2729        return 0;
2730}
2731
2732static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2733{
2734        struct inode *inode = file_inode(filp);
2735        __u32 pin;
2736        int ret = 0;
2737
2738        if (!inode_owner_or_capable(inode))
2739                return -EACCES;
2740
2741        if (get_user(pin, (__u32 __user *)arg))
2742                return -EFAULT;
2743
2744        if (!S_ISREG(inode->i_mode))
2745                return -EINVAL;
2746
2747        if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2748                return -EROFS;
2749
2750        ret = mnt_want_write_file(filp);
2751        if (ret)
2752                return ret;
2753
2754        inode_lock(inode);
2755
2756        if (should_update_outplace(inode, NULL)) {
2757                ret = -EINVAL;
2758                goto out;
2759        }
2760
2761        if (!pin) {
2762                clear_inode_flag(inode, FI_PIN_FILE);
2763                F2FS_I(inode)->i_gc_failures = 1;
2764                goto done;
2765        }
2766
2767        if (f2fs_pin_file_control(inode, false)) {
2768                ret = -EAGAIN;
2769                goto out;
2770        }
2771        ret = f2fs_convert_inline_inode(inode);
2772        if (ret)
2773                goto out;
2774
2775        set_inode_flag(inode, FI_PIN_FILE);
2776        ret = F2FS_I(inode)->i_gc_failures;
2777done:
2778        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2779out:
2780        inode_unlock(inode);
2781        mnt_drop_write_file(filp);
2782        return ret;
2783}
2784
2785static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2786{
2787        struct inode *inode = file_inode(filp);
2788        __u32 pin = 0;
2789
2790        if (is_inode_flag_set(inode, FI_PIN_FILE))
2791                pin = F2FS_I(inode)->i_gc_failures;
2792        return put_user(pin, (u32 __user *)arg);
2793}
2794
2795int f2fs_precache_extents(struct inode *inode)
2796{
2797        struct f2fs_inode_info *fi = F2FS_I(inode);
2798        struct f2fs_map_blocks map;
2799        pgoff_t m_next_extent;
2800        loff_t end;
2801        int err;
2802
2803        if (is_inode_flag_set(inode, FI_NO_EXTENT))
2804                return -EOPNOTSUPP;
2805
2806        map.m_lblk = 0;
2807        map.m_next_pgofs = NULL;
2808        map.m_next_extent = &m_next_extent;
2809        map.m_seg_type = NO_CHECK_TYPE;
2810        end = F2FS_I_SB(inode)->max_file_blocks;
2811
2812        while (map.m_lblk < end) {
2813                map.m_len = end - map.m_lblk;
2814
2815                down_write(&fi->dio_rwsem[WRITE]);
2816                err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2817                up_write(&fi->dio_rwsem[WRITE]);
2818                if (err)
2819                        return err;
2820
2821                map.m_lblk = m_next_extent;
2822        }
2823
2824        return err;
2825}
2826
2827static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2828{
2829        return f2fs_precache_extents(file_inode(filp));
2830}
2831
2832long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2833{
2834        if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2835                return -EIO;
2836
2837        switch (cmd) {
2838        case F2FS_IOC_GETFLAGS:
2839                return f2fs_ioc_getflags(filp, arg);
2840        case F2FS_IOC_SETFLAGS:
2841                return f2fs_ioc_setflags(filp, arg);
2842        case F2FS_IOC_GETVERSION:
2843                return f2fs_ioc_getversion(filp, arg);
2844        case F2FS_IOC_START_ATOMIC_WRITE:
2845                return f2fs_ioc_start_atomic_write(filp);
2846        case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2847                return f2fs_ioc_commit_atomic_write(filp);
2848        case F2FS_IOC_START_VOLATILE_WRITE:
2849                return f2fs_ioc_start_volatile_write(filp);
2850        case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2851                return f2fs_ioc_release_volatile_write(filp);
2852        case F2FS_IOC_ABORT_VOLATILE_WRITE:
2853                return f2fs_ioc_abort_volatile_write(filp);
2854        case F2FS_IOC_SHUTDOWN:
2855                return f2fs_ioc_shutdown(filp, arg);
2856        case FITRIM:
2857                return f2fs_ioc_fitrim(filp, arg);
2858        case F2FS_IOC_SET_ENCRYPTION_POLICY:
2859                return f2fs_ioc_set_encryption_policy(filp, arg);
2860        case F2FS_IOC_GET_ENCRYPTION_POLICY:
2861                return f2fs_ioc_get_encryption_policy(filp, arg);
2862        case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2863                return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2864        case F2FS_IOC_GARBAGE_COLLECT:
2865                return f2fs_ioc_gc(filp, arg);
2866        case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2867                return f2fs_ioc_gc_range(filp, arg);
2868        case F2FS_IOC_WRITE_CHECKPOINT:
2869                return f2fs_ioc_write_checkpoint(filp, arg);
2870        case F2FS_IOC_DEFRAGMENT:
2871                return f2fs_ioc_defragment(filp, arg);
2872        case F2FS_IOC_MOVE_RANGE:
2873                return f2fs_ioc_move_range(filp, arg);
2874        case F2FS_IOC_FLUSH_DEVICE:
2875                return f2fs_ioc_flush_device(filp, arg);
2876        case F2FS_IOC_GET_FEATURES:
2877                return f2fs_ioc_get_features(filp, arg);
2878        case F2FS_IOC_FSGETXATTR:
2879                return f2fs_ioc_fsgetxattr(filp, arg);
2880        case F2FS_IOC_FSSETXATTR:
2881                return f2fs_ioc_fssetxattr(filp, arg);
2882        case F2FS_IOC_GET_PIN_FILE:
2883                return f2fs_ioc_get_pin_file(filp, arg);
2884        case F2FS_IOC_SET_PIN_FILE:
2885                return f2fs_ioc_set_pin_file(filp, arg);
2886        case F2FS_IOC_PRECACHE_EXTENTS:
2887                return f2fs_ioc_precache_extents(filp, arg);
2888        default:
2889                return -ENOTTY;
2890        }
2891}
2892
2893static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2894{
2895        struct file *file = iocb->ki_filp;
2896        struct inode *inode = file_inode(file);
2897        struct blk_plug plug;
2898        ssize_t ret;
2899
2900        if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2901                return -EIO;
2902
2903        if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2904                return -EINVAL;
2905
2906        if (!inode_trylock(inode)) {
2907                if (iocb->ki_flags & IOCB_NOWAIT)
2908                        return -EAGAIN;
2909                inode_lock(inode);
2910        }
2911
2912        ret = generic_write_checks(iocb, from);
2913        if (ret > 0) {
2914                bool preallocated = false;
2915                size_t target_size = 0;
2916                int err;
2917
2918                if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2919                        set_inode_flag(inode, FI_NO_PREALLOC);
2920
2921                if ((iocb->ki_flags & IOCB_NOWAIT) &&
2922                        (iocb->ki_flags & IOCB_DIRECT)) {
2923                                if (!f2fs_overwrite_io(inode, iocb->ki_pos,
2924                                                iov_iter_count(from)) ||
2925                                        f2fs_has_inline_data(inode) ||
2926                                        f2fs_force_buffered_io(inode, WRITE)) {
2927                                                inode_unlock(inode);
2928                                                return -EAGAIN;
2929                                }
2930
2931                } else {
2932                        preallocated = true;
2933                        target_size = iocb->ki_pos + iov_iter_count(from);
2934
2935                        err = f2fs_preallocate_blocks(iocb, from);
2936                        if (err) {
2937                                clear_inode_flag(inode, FI_NO_PREALLOC);
2938                                inode_unlock(inode);
2939                                return err;
2940                        }
2941                }
2942                blk_start_plug(&plug);
2943                ret = __generic_file_write_iter(iocb, from);
2944                blk_finish_plug(&plug);
2945                clear_inode_flag(inode, FI_NO_PREALLOC);
2946
2947                /* if we couldn't write data, we should deallocate blocks. */
2948                if (preallocated && i_size_read(inode) < target_size)
2949                        f2fs_truncate(inode);
2950
2951                if (ret > 0)
2952                        f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2953        }
2954        inode_unlock(inode);
2955
2956        if (ret > 0)
2957                ret = generic_write_sync(iocb, ret);
2958        return ret;
2959}
2960
2961#ifdef CONFIG_COMPAT
2962long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2963{
2964        switch (cmd) {
2965        case F2FS_IOC32_GETFLAGS:
2966                cmd = F2FS_IOC_GETFLAGS;
2967                break;
2968        case F2FS_IOC32_SETFLAGS:
2969                cmd = F2FS_IOC_SETFLAGS;
2970                break;
2971        case F2FS_IOC32_GETVERSION:
2972                cmd = F2FS_IOC_GETVERSION;
2973                break;
2974        case F2FS_IOC_START_ATOMIC_WRITE:
2975        case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2976        case F2FS_IOC_START_VOLATILE_WRITE:
2977        case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2978        case F2FS_IOC_ABORT_VOLATILE_WRITE:
2979        case F2FS_IOC_SHUTDOWN:
2980        case F2FS_IOC_SET_ENCRYPTION_POLICY:
2981        case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2982        case F2FS_IOC_GET_ENCRYPTION_POLICY:
2983        case F2FS_IOC_GARBAGE_COLLECT:
2984        case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2985        case F2FS_IOC_WRITE_CHECKPOINT:
2986        case F2FS_IOC_DEFRAGMENT:
2987        case F2FS_IOC_MOVE_RANGE:
2988        case F2FS_IOC_FLUSH_DEVICE:
2989        case F2FS_IOC_GET_FEATURES:
2990        case F2FS_IOC_FSGETXATTR:
2991        case F2FS_IOC_FSSETXATTR:
2992        case F2FS_IOC_GET_PIN_FILE:
2993        case F2FS_IOC_SET_PIN_FILE:
2994        case F2FS_IOC_PRECACHE_EXTENTS:
2995                break;
2996        default:
2997                return -ENOIOCTLCMD;
2998        }
2999        return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3000}
3001#endif
3002
3003const struct file_operations f2fs_file_operations = {
3004        .llseek         = f2fs_llseek,
3005        .read_iter      = generic_file_read_iter,
3006        .write_iter     = f2fs_file_write_iter,
3007        .open           = f2fs_file_open,
3008        .release        = f2fs_release_file,
3009        .mmap           = f2fs_file_mmap,
3010        .flush          = f2fs_file_flush,
3011        .fsync          = f2fs_sync_file,
3012        .fallocate      = f2fs_fallocate,
3013        .unlocked_ioctl = f2fs_ioctl,
3014#ifdef CONFIG_COMPAT
3015        .compat_ioctl   = f2fs_compat_ioctl,
3016#endif
3017        .splice_read    = generic_file_splice_read,
3018        .splice_write   = iter_file_splice_write,
3019};
3020