fs/f2fs/file.c

   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/uuid.h>
  24 #include <linux/file.h>
  25
  26 #include "f2fs.h"
  27 #include "node.h"
  28 #include "segment.h"
  29 #include "xattr.h"
  30 #include "acl.h"
  31 #include "gc.h"
  32 #include "trace.h"
  33 #include <trace/events/f2fs.h>
  34
  35 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
  36                                                 struct vm_fault *vmf)
  37 {
  38         struct page *page = vmf->page;
  39         struct inode *inode = file_inode(vma->vm_file);
  40         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
  41         struct dnode_of_data dn;
  42         int err;
  43
  44         sb_start_pagefault(inode->i_sb);
  45
  46         f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
  47
  48         /* block allocation */
  49         f2fs_lock_op(sbi);
  50         set_new_dnode(&dn, inode, NULL, NULL, 0);
  51         err = f2fs_reserve_block(&dn, page->index);
  52         if (err) {
  53                 f2fs_unlock_op(sbi);
  54                 goto out;
  55         }
  56         f2fs_put_dnode(&dn);
  57         f2fs_unlock_op(sbi);
  58
  59         f2fs_balance_fs(sbi, dn.node_changed);
  60
  61         file_update_time(vma->vm_file);
  62         lock_page(page);
  63         if (unlikely(page->mapping != inode->i_mapping ||
  64                         page_offset(page) > i_size_read(inode) ||
  65                         !PageUptodate(page))) {
  66                 unlock_page(page);
  67                 err = -EFAULT;
  68                 goto out;
  69         }
  70
  71         /*
  72          * check to see if the page is mapped already (no holes)
  73          */
  74         if (PageMappedToDisk(page))
  75                 goto mapped;
  76
  77         /* page is wholly or partially inside EOF */
  78         if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
  79                                                 i_size_read(inode)) {
  80                 unsigned offset;
  81                 offset = i_size_read(inode) & ~PAGE_MASK;
  82                 zero_user_segment(page, offset, PAGE_SIZE);
  83         }
  84         set_page_dirty(page);
  85         if (!PageUptodate(page))
  86                 SetPageUptodate(page);
  87
  88         trace_f2fs_vm_page_mkwrite(page, DATA);
  89 mapped:
  90         /* fill the page */
  91         f2fs_wait_on_page_writeback(page, DATA, false);
  92
  93         /* wait for GCed encrypted page writeback */
  94         if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
  95                 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
  96
  97 out:
  98         sb_end_pagefault(inode->i_sb);
  99         f2fs_update_time(sbi, REQ_TIME);
 100         return block_page_mkwrite_return(err);
 101 }
 102
 103 static const struct vm_operations_struct f2fs_file_vm_ops = {
 104         .fault          = filemap_fault,
 105         .map_pages      = filemap_map_pages,
 106         .page_mkwrite   = f2fs_vm_page_mkwrite,
 107 };
 108
 109 static int get_parent_ino(struct inode *inode, nid_t *pino)
 110 {
 111         struct dentry *dentry;
 112
 113         inode = igrab(inode);
 114         dentry = d_find_any_alias(inode);
 115         iput(inode);
 116         if (!dentry)
 117                 return 0;
 118
 119         if (update_dent_inode(inode, inode, &dentry->d_name)) {
 120                 dput(dentry);
 121                 return 0;
 122         }
 123
 124         *pino = parent_ino(dentry);
 125         dput(dentry);
 126         return 1;
 127 }
 128
 129 static inline bool need_do_checkpoint(struct inode *inode)
 130 {
 131         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 132         bool need_cp = false;
 133
 134         if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
 135                 need_cp = true;
 136         else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
 137                 need_cp = true;
 138         else if (file_wrong_pino(inode))
 139                 need_cp = true;
 140         else if (!space_for_roll_forward(sbi))
 141                 need_cp = true;
 142         else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
 143                 need_cp = true;
 144         else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
 145                 need_cp = true;
 146         else if (test_opt(sbi, FASTBOOT))
 147                 need_cp = true;
 148         else if (sbi->active_logs == 2)
 149                 need_cp = true;
 150
 151         return need_cp;
 152 }
 153
 154 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
 155 {
 156         struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
 157         bool ret = false;
 158         /* But we need to avoid that there are some inode updates */
 159         if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
 160                 ret = true;
 161         f2fs_put_page(i, 0);
 162         return ret;
 163 }
 164
 165 static void try_to_fix_pino(struct inode *inode)
 166 {
 167         struct f2fs_inode_info *fi = F2FS_I(inode);
 168         nid_t pino;
 169
 170         down_write(&fi->i_sem);
 171         fi->xattr_ver = 0;
 172         if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
 173                         get_parent_ino(inode, &pino)) {
 174                 f2fs_i_pino_write(inode, pino);
 175                 file_got_pino(inode);
 176         }
 177         up_write(&fi->i_sem);
 178 }
 179
 180 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
 181                                                 int datasync, bool atomic)
 182 {
 183         struct inode *inode = file->f_mapping->host;
 184         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 185         nid_t ino = inode->i_ino;
 186         int ret = 0;
 187         bool need_cp = false;
 188         struct writeback_control wbc = {
 189                 .sync_mode = WB_SYNC_ALL,
 190                 .nr_to_write = LONG_MAX,
 191                 .for_reclaim = 0,
 192         };
 193
 194         if (unlikely(f2fs_readonly(inode->i_sb)))
 195                 return 0;
 196
 197         trace_f2fs_sync_file_enter(inode);
 198
 199         /* if fdatasync is triggered, let's do in-place-update */
 200         if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
 201                 set_inode_flag(inode, FI_NEED_IPU);
 202         ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
 203         clear_inode_flag(inode, FI_NEED_IPU);
 204
 205         if (ret) {
 206                 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
 207                 return ret;
 208         }
 209
 210         /* if the inode is dirty, let's recover all the time */
 211         if (!f2fs_skip_inode_update(inode, datasync)) {
 212                 f2fs_write_inode(inode, NULL);
 213                 goto go_write;
 214         }
 215
 216         /*
 217          * if there is no written data, don't waste time to write recovery info.
 218          */
 219         if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
 220                         !exist_written_data(sbi, ino, APPEND_INO)) {
 221
 222                 /* it may call write_inode just prior to fsync */
 223                 if (need_inode_page_update(sbi, ino))
 224                         goto go_write;
 225
 226                 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
 227                                 exist_written_data(sbi, ino, UPDATE_INO))
 228                         goto flush_out;
 229                 goto out;
 230         }
 231 go_write:
 232         /*
 233          * Both of fdatasync() and fsync() are able to be recovered from
 234          * sudden-power-off.
 235          */
 236         down_read(&F2FS_I(inode)->i_sem);
 237         need_cp = need_do_checkpoint(inode);
 238         up_read(&F2FS_I(inode)->i_sem);
 239
 240         if (need_cp) {
 241                 /* all the dirty node pages should be flushed for POR */
 242                 ret = f2fs_sync_fs(inode->i_sb, 1);
 243
 244                 /*
 245                  * We've secured consistency through sync_fs. Following pino
 246                  * will be used only for fsynced inodes after checkpoint.
 247                  */
 248                 try_to_fix_pino(inode);
 249                 clear_inode_flag(inode, FI_APPEND_WRITE);
 250                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 251                 goto out;
 252         }
 253 sync_nodes:
 254         ret = fsync_node_pages(sbi, inode, &wbc, atomic);
 255         if (ret)
 256                 goto out;
 257
 258         /* if cp_error was enabled, we should avoid infinite loop */
 259         if (unlikely(f2fs_cp_error(sbi))) {
 260                 ret = -EIO;
 261                 goto out;
 262         }
 263
 264         if (need_inode_block_update(sbi, ino)) {
 265                 f2fs_mark_inode_dirty_sync(inode, true);
 266                 f2fs_write_inode(inode, NULL);
 267                 goto sync_nodes;
 268         }
 269
 270         ret = wait_on_node_pages_writeback(sbi, ino);
 271         if (ret)
 272                 goto out;
 273
 274         /* once recovery info is written, don't need to tack this */
 275         remove_ino_entry(sbi, ino, APPEND_INO);
 276         clear_inode_flag(inode, FI_APPEND_WRITE);
 277 flush_out:
 278         remove_ino_entry(sbi, ino, UPDATE_INO);
 279         clear_inode_flag(inode, FI_UPDATE_WRITE);
 280         ret = f2fs_issue_flush(sbi);
 281         f2fs_update_time(sbi, REQ_TIME);
 282 out:
 283         trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
 284         f2fs_trace_ios(NULL, 1);
 285         return ret;
 286 }
 287
 288 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 289 {
 290         return f2fs_do_sync_file(file, start, end, datasync, false);
 291 }
 292
 293 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
 294                                                 pgoff_t pgofs, int whence)
 295 {
 296         struct pagevec pvec;
 297         int nr_pages;
 298
 299         if (whence != SEEK_DATA)
 300                 return 0;
 301
 302         /* find first dirty page index */
 303         pagevec_init(&pvec, 0);
 304         nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
 305                                         PAGECACHE_TAG_DIRTY, 1);
 306         pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
 307         pagevec_release(&pvec);
 308         return pgofs;
 309 }
 310
 311 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
 312                                                         int whence)
 313 {
 314         switch (whence) {
 315         case SEEK_DATA:
 316                 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
 317                         (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
 318                         return true;
 319                 break;
 320         case SEEK_HOLE:
 321                 if (blkaddr == NULL_ADDR)
 322                         return true;
 323                 break;
 324         }
 325         return false;
 326 }
 327
 328 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 329 {
 330         struct inode *inode = file->f_mapping->host;
 331         loff_t maxbytes = inode->i_sb->s_maxbytes;
 332         struct dnode_of_data dn;
 333         pgoff_t pgofs, end_offset, dirty;
 334         loff_t data_ofs = offset;
 335         loff_t isize;
 336         int err = 0;
 337
 338         inode_lock(inode);
 339
 340         isize = i_size_read(inode);
 341         if (offset >= isize)
 342                 goto fail;
 343
 344         /* handle inline data case */
 345         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
 346                 if (whence == SEEK_HOLE)
 347                         data_ofs = isize;
 348                 goto found;
 349         }
 350
 351         pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
 352
 353         dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
 354
 355         for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 356                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 357                 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
 358                 if (err && err != -ENOENT) {
 359                         goto fail;
 360                 } else if (err == -ENOENT) {
 361                         /* direct node does not exists */
 362                         if (whence == SEEK_DATA) {
 363                                 pgofs = get_next_page_offset(&dn, pgofs);
 364                                 continue;
 365                         } else {
 366                                 goto found;
 367                         }
 368                 }
 369
 370                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 371
 372                 /* find data/hole in dnode block */
 373                 for (; dn.ofs_in_node < end_offset;
 374                                 dn.ofs_in_node++, pgofs++,
 375                                 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 376                         block_t blkaddr;
 377                         blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
 378
 379                         if (__found_offset(blkaddr, dirty, pgofs, whence)) {
 380                                 f2fs_put_dnode(&dn);
 381                                 goto found;
 382                         }
 383                 }
 384                 f2fs_put_dnode(&dn);
 385         }
 386
 387         if (whence == SEEK_DATA)
 388                 goto fail;
 389 found:
 390         if (whence == SEEK_HOLE && data_ofs > isize)
 391                 data_ofs = isize;
 392         inode_unlock(inode);
 393         return vfs_setpos(file, data_ofs, maxbytes);
 394 fail:
 395         inode_unlock(inode);
 396         return -ENXIO;
 397 }
 398
 399 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
 400 {
 401         struct inode *inode = file->f_mapping->host;
 402         loff_t maxbytes = inode->i_sb->s_maxbytes;
 403
 404         switch (whence) {
 405         case SEEK_SET:
 406         case SEEK_CUR:
 407         case SEEK_END:
 408                 return generic_file_llseek_size(file, offset, whence,
 409                                                 maxbytes, i_size_read(inode));
 410         case SEEK_DATA:
 411         case SEEK_HOLE:
 412                 if (offset < 0)
 413                         return -ENXIO;
 414                 return f2fs_seek_block(file, offset, whence);
 415         }
 416
 417         return -EINVAL;
 418 }
 419
 420 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 421 {
 422         struct inode *inode = file_inode(file);
 423         int err;
 424
 425         if (f2fs_encrypted_inode(inode)) {
 426                 err = fscrypt_get_encryption_info(inode);
 427                 if (err)
 428                         return 0;
 429                 if (!f2fs_encrypted_inode(inode))
 430                         return -ENOKEY;
 431         }
 432
 433         /* we don't need to use inline_data strictly */
 434         err = f2fs_convert_inline_inode(inode);
 435         if (err)
 436                 return err;
 437
 438         file_accessed(file);
 439         vma->vm_ops = &f2fs_file_vm_ops;
 440         return 0;
 441 }
 442
 443 static int f2fs_file_open(struct inode *inode, struct file *filp)
 444 {
 445         int ret = generic_file_open(inode, filp);
 446         struct dentry *dir;
 447
 448         if (!ret && f2fs_encrypted_inode(inode)) {
 449                 ret = fscrypt_get_encryption_info(inode);
 450                 if (ret)
 451                         return -EACCES;
 452                 if (!fscrypt_has_encryption_key(inode))
 453                         return -ENOKEY;
 454         }
 455         dir = dget_parent(file_dentry(filp));
 456         if (f2fs_encrypted_inode(d_inode(dir)) &&
 457                         !fscrypt_has_permitted_context(d_inode(dir), inode)) {
 458                 dput(dir);
 459                 return -EPERM;
 460         }
 461         dput(dir);
 462         return ret;
 463 }
 464
 465 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 466 {
 467         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 468         struct f2fs_node *raw_node;
 469         int nr_free = 0, ofs = dn->ofs_in_node, len = count;
 470         __le32 *addr;
 471
 472         raw_node = F2FS_NODE(dn->node_page);
 473         addr = blkaddr_in_node(raw_node) + ofs;
 474
 475         for (; count > 0; count--, addr++, dn->ofs_in_node++) {
 476                 block_t blkaddr = le32_to_cpu(*addr);
 477                 if (blkaddr == NULL_ADDR)
 478                         continue;
 479
 480                 dn->data_blkaddr = NULL_ADDR;
 481                 set_data_blkaddr(dn);
 482                 invalidate_blocks(sbi, blkaddr);
 483                 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 484                         clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 485                 nr_free++;
 486         }
 487
 488         if (nr_free) {
 489                 pgoff_t fofs;
 490                 /*
 491                  * once we invalidate valid blkaddr in range [ofs, ofs + count],
 492                  * we will invalidate all blkaddr in the whole range.
 493                  */
 494                 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
 495                                                         dn->inode) + ofs;
 496                 f2fs_update_extent_cache_range(dn, fofs, 0, len);
 497                 dec_valid_block_count(sbi, dn->inode, nr_free);
 498         }
 499         dn->ofs_in_node = ofs;
 500
 501         f2fs_update_time(sbi, REQ_TIME);
 502         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 503                                          dn->ofs_in_node, nr_free);
 504         return nr_free;
 505 }
 506
 507 void truncate_data_blocks(struct dnode_of_data *dn)
 508 {
 509         truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
 510 }
 511
 512 static int truncate_partial_data_page(struct inode *inode, u64 from,
 513                                                                 bool cache_only)
 514 {
 515         unsigned offset = from & (PAGE_SIZE - 1);
 516         pgoff_t index = from >> PAGE_SHIFT;
 517         struct address_space *mapping = inode->i_mapping;
 518         struct page *page;
 519
 520         if (!offset && !cache_only)
 521                 return 0;
 522
 523         if (cache_only) {
 524                 page = find_lock_page(mapping, index);
 525                 if (page && PageUptodate(page))
 526                         goto truncate_out;
 527                 f2fs_put_page(page, 1);
 528                 return 0;
 529         }
 530
 531         page = get_lock_data_page(inode, index, true);
 532         if (IS_ERR(page))
 533                 return 0;
 534 truncate_out:
 535         f2fs_wait_on_page_writeback(page, DATA, true);
 536         zero_user(page, offset, PAGE_SIZE - offset);
 537         if (!cache_only || !f2fs_encrypted_inode(inode) ||
 538                                         !S_ISREG(inode->i_mode))
 539                 set_page_dirty(page);
 540         f2fs_put_page(page, 1);
 541         return 0;
 542 }
 543
 544 int truncate_blocks(struct inode *inode, u64 from, bool lock)
 545 {
 546         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 547         unsigned int blocksize = inode->i_sb->s_blocksize;
 548         struct dnode_of_data dn;
 549         pgoff_t free_from;
 550         int count = 0, err = 0;
 551         struct page *ipage;
 552         bool truncate_page = false;
 553
 554         trace_f2fs_truncate_blocks_enter(inode, from);
 555
 556         free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
 557
 558         if (free_from >= sbi->max_file_blocks)
 559                 goto free_partial;
 560
 561         if (lock)
 562                 f2fs_lock_op(sbi);
 563
 564         ipage = get_node_page(sbi, inode->i_ino);
 565         if (IS_ERR(ipage)) {
 566                 err = PTR_ERR(ipage);
 567                 goto out;
 568         }
 569
 570         if (f2fs_has_inline_data(inode)) {
 571                 if (truncate_inline_inode(ipage, from))
 572                         set_page_dirty(ipage);
 573                 f2fs_put_page(ipage, 1);
 574                 truncate_page = true;
 575                 goto out;
 576         }
 577
 578         set_new_dnode(&dn, inode, ipage, NULL, 0);
 579         err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 580         if (err) {
 581                 if (err == -ENOENT)
 582                         goto free_next;
 583                 goto out;
 584         }
 585
 586         count = ADDRS_PER_PAGE(dn.node_page, inode);
 587
 588         count -= dn.ofs_in_node;
 589         f2fs_bug_on(sbi, count < 0);
 590
 591         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 592                 truncate_data_blocks_range(&dn, count);
 593                 free_from += count;
 594         }
 595
 596         f2fs_put_dnode(&dn);
 597 free_next:
 598         err = truncate_inode_blocks(inode, free_from);
 599 out:
 600         if (lock)
 601                 f2fs_unlock_op(sbi);
 602 free_partial:
 603         /* lastly zero out the first data page */
 604         if (!err)
 605                 err = truncate_partial_data_page(inode, from, truncate_page);
 606
 607         trace_f2fs_truncate_blocks_exit(inode, err);
 608         return err;
 609 }
 610
 611 int f2fs_truncate(struct inode *inode)
 612 {
 613         int err;
 614
 615         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 616                                 S_ISLNK(inode->i_mode)))
 617                 return 0;
 618
 619         trace_f2fs_truncate(inode);
 620
 621         /* we should check inline_data size */
 622         if (!f2fs_may_inline_data(inode)) {
 623                 err = f2fs_convert_inline_inode(inode);
 624                 if (err)
 625                         return err;
 626         }
 627
 628         err = truncate_blocks(inode, i_size_read(inode), true);
 629         if (err)
 630                 return err;
 631
 632         inode->i_mtime = inode->i_ctime = current_time(inode);
 633         f2fs_mark_inode_dirty_sync(inode, false);
 634         return 0;
 635 }
 636
 637 int f2fs_getattr(struct vfsmount *mnt,
 638                          struct dentry *dentry, struct kstat *stat)
 639 {
 640         struct inode *inode = d_inode(dentry);
 641         generic_fillattr(inode, stat);
 642         stat->blocks <<= 3;
 643         return 0;
 644 }
 645
 646 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 647 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
 648 {
 649         unsigned int ia_valid = attr->ia_valid;
 650
 651         if (ia_valid & ATTR_UID)
 652                 inode->i_uid = attr->ia_uid;
 653         if (ia_valid & ATTR_GID)
 654                 inode->i_gid = attr->ia_gid;
 655         if (ia_valid & ATTR_ATIME)
 656                 inode->i_atime = timespec_trunc(attr->ia_atime,
 657                                                 inode->i_sb->s_time_gran);
 658         if (ia_valid & ATTR_MTIME)
 659                 inode->i_mtime = timespec_trunc(attr->ia_mtime,
 660                                                 inode->i_sb->s_time_gran);
 661         if (ia_valid & ATTR_CTIME)
 662                 inode->i_ctime = timespec_trunc(attr->ia_ctime,
 663                                                 inode->i_sb->s_time_gran);
 664         if (ia_valid & ATTR_MODE) {
 665                 umode_t mode = attr->ia_mode;
 666
 667                 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
 668                         mode &= ~S_ISGID;
 669                 set_acl_inode(inode, mode);
 670         }
 671 }
 672 #else
 673 #define __setattr_copy setattr_copy
 674 #endif
 675
 676 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
 677 {
 678         struct inode *inode = d_inode(dentry);
 679         int err;
 680         bool size_changed = false;
 681
 682         err = setattr_prepare(dentry, attr);
 683         if (err)
 684                 return err;
 685
 686         if (attr->ia_valid & ATTR_SIZE) {
 687                 if (f2fs_encrypted_inode(inode) &&
 688                                 fscrypt_get_encryption_info(inode))
 689                         return -EACCES;
 690
 691                 if (attr->ia_size <= i_size_read(inode)) {
 692                         truncate_setsize(inode, attr->ia_size);
 693                         err = f2fs_truncate(inode);
 694                         if (err)
 695                                 return err;
 696                 } else {
 697                         /*
 698                          * do not trim all blocks after i_size if target size is
 699                          * larger than i_size.
 700                          */
 701                         truncate_setsize(inode, attr->ia_size);
 702
 703                         /* should convert inline inode here */
 704                         if (!f2fs_may_inline_data(inode)) {
 705                                 err = f2fs_convert_inline_inode(inode);
 706                                 if (err)
 707                                         return err;
 708                         }
 709                         inode->i_mtime = inode->i_ctime = current_time(inode);
 710                 }
 711
 712                 size_changed = true;
 713         }
 714
 715         __setattr_copy(inode, attr);
 716
 717         if (attr->ia_valid & ATTR_MODE) {
 718                 err = posix_acl_chmod(inode, get_inode_mode(inode));
 719                 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
 720                         inode->i_mode = F2FS_I(inode)->i_acl_mode;
 721                         clear_inode_flag(inode, FI_ACL_MODE);
 722                 }
 723         }
 724
 725         /* file size may changed here */
 726         f2fs_mark_inode_dirty_sync(inode, size_changed);
 727
 728         /* inode change will produce dirty node pages flushed by checkpoint */
 729         f2fs_balance_fs(F2FS_I_SB(inode), true);
 730
 731         return err;
 732 }
 733
 734 const struct inode_operations f2fs_file_inode_operations = {
 735         .getattr        = f2fs_getattr,
 736         .setattr        = f2fs_setattr,
 737         .get_acl        = f2fs_get_acl,
 738         .set_acl        = f2fs_set_acl,
 739 #ifdef CONFIG_F2FS_FS_XATTR
 740         .listxattr      = f2fs_listxattr,
 741 #endif
 742         .fiemap         = f2fs_fiemap,
 743 };
 744
 745 static int fill_zero(struct inode *inode, pgoff_t index,
 746                                         loff_t start, loff_t len)
 747 {
 748         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 749         struct page *page;
 750
 751         if (!len)
 752                 return 0;
 753
 754         f2fs_balance_fs(sbi, true);
 755
 756         f2fs_lock_op(sbi);
 757         page = get_new_data_page(inode, NULL, index, false);
 758         f2fs_unlock_op(sbi);
 759
 760         if (IS_ERR(page))
 761                 return PTR_ERR(page);
 762
 763         f2fs_wait_on_page_writeback(page, DATA, true);
 764         zero_user(page, start, len);
 765         set_page_dirty(page);
 766         f2fs_put_page(page, 1);
 767         return 0;
 768 }
 769
 770 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
 771 {
 772         int err;
 773
 774         while (pg_start < pg_end) {
 775                 struct dnode_of_data dn;
 776                 pgoff_t end_offset, count;
 777
 778                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 779                 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
 780                 if (err) {
 781                         if (err == -ENOENT) {
 782                                 pg_start++;
 783                                 continue;
 784                         }
 785                         return err;
 786                 }
 787
 788                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 789                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
 790
 791                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
 792
 793                 truncate_data_blocks_range(&dn, count);
 794                 f2fs_put_dnode(&dn);
 795
 796                 pg_start += count;
 797         }
 798         return 0;
 799 }
 800
 801 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
 802 {
 803         pgoff_t pg_start, pg_end;
 804         loff_t off_start, off_end;
 805         int ret;
 806
 807         ret = f2fs_convert_inline_inode(inode);
 808         if (ret)
 809                 return ret;
 810
 811         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
 812         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
 813
 814         off_start = offset & (PAGE_SIZE - 1);
 815         off_end = (offset + len) & (PAGE_SIZE - 1);
 816
 817         if (pg_start == pg_end) {
 818                 ret = fill_zero(inode, pg_start, off_start,
 819                                                 off_end - off_start);
 820                 if (ret)
 821                         return ret;
 822         } else {
 823                 if (off_start) {
 824                         ret = fill_zero(inode, pg_start++, off_start,
 825                                                 PAGE_SIZE - off_start);
 826                         if (ret)
 827                                 return ret;
 828                 }
 829                 if (off_end) {
 830                         ret = fill_zero(inode, pg_end, 0, off_end);
 831                         if (ret)
 832                                 return ret;
 833                 }
 834
 835                 if (pg_start < pg_end) {
 836                         struct address_space *mapping = inode->i_mapping;
 837                         loff_t blk_start, blk_end;
 838                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 839
 840                         f2fs_balance_fs(sbi, true);
 841
 842                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
 843                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
 844                         truncate_inode_pages_range(mapping, blk_start,
 845                                         blk_end - 1);
 846
 847                         f2fs_lock_op(sbi);
 848                         ret = truncate_hole(inode, pg_start, pg_end);
 849                         f2fs_unlock_op(sbi);
 850                 }
 851         }
 852
 853         return ret;
 854 }
 855
 856 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
 857                                 int *do_replace, pgoff_t off, pgoff_t len)
 858 {
 859         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 860         struct dnode_of_data dn;
 861         int ret, done, i;
 862
 863 next_dnode:
 864         set_new_dnode(&dn, inode, NULL, NULL, 0);
 865         ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
 866         if (ret && ret != -ENOENT) {
 867                 return ret;
 868         } else if (ret == -ENOENT) {
 869                 if (dn.max_level == 0)
 870                         return -ENOENT;
 871                 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
 872                 blkaddr += done;
 873                 do_replace += done;
 874                 goto next;
 875         }
 876
 877         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
 878                                                         dn.ofs_in_node, len);
 879         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
 880                 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
 881                 if (!is_checkpointed_data(sbi, *blkaddr)) {
 882
 883                         if (test_opt(sbi, LFS)) {
 884                                 f2fs_put_dnode(&dn);
 885                                 return -ENOTSUPP;
 886                         }
 887
 888                         /* do not invalidate this block address */
 889                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
 890                         *do_replace = 1;
 891                 }
 892         }
 893         f2fs_put_dnode(&dn);
 894 next:
 895         len -= done;
 896         off += done;
 897         if (len)
 898                 goto next_dnode;
 899         return 0;
 900 }
 901
 902 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
 903                                 int *do_replace, pgoff_t off, int len)
 904 {
 905         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 906         struct dnode_of_data dn;
 907         int ret, i;
 908
 909         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
 910                 if (*do_replace == 0)
 911                         continue;
 912
 913                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 914                 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
 915                 if (ret) {
 916                         dec_valid_block_count(sbi, inode, 1);
 917                         invalidate_blocks(sbi, *blkaddr);
 918                 } else {
 919                         f2fs_update_data_blkaddr(&dn, *blkaddr);
 920                 }
 921                 f2fs_put_dnode(&dn);
 922         }
 923         return 0;
 924 }
 925
 926 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
 927                         block_t *blkaddr, int *do_replace,
 928                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
 929 {
 930         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
 931         pgoff_t i = 0;
 932         int ret;
 933
 934         while (i < len) {
 935                 if (blkaddr[i] == NULL_ADDR && !full) {
 936                         i++;
 937                         continue;
 938                 }
 939
 940                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
 941                         struct dnode_of_data dn;
 942                         struct node_info ni;
 943                         size_t new_size;
 944                         pgoff_t ilen;
 945
 946                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
 947                         ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
 948                         if (ret)
 949                                 return ret;
 950
 951                         get_node_info(sbi, dn.nid, &ni);
 952                         ilen = min((pgoff_t)
 953                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
 954                                                 dn.ofs_in_node, len - i);
 955                         do {
 956                                 dn.data_blkaddr = datablock_addr(dn.node_page,
 957                                                                 dn.ofs_in_node);
 958                                 truncate_data_blocks_range(&dn, 1);
 959
 960                                 if (do_replace[i]) {
 961                                         f2fs_i_blocks_write(src_inode,
 962                                                                 1, false);
 963                                         f2fs_i_blocks_write(dst_inode,
 964                                                                 1, true);
 965                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
 966                                         blkaddr[i], ni.version, true, false);
 967
 968                                         do_replace[i] = 0;
 969                                 }
 970                                 dn.ofs_in_node++;
 971                                 i++;
 972                                 new_size = (dst + i) << PAGE_SHIFT;
 973                                 if (dst_inode->i_size < new_size)
 974                                         f2fs_i_size_write(dst_inode, new_size);
 975                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
 976
 977                         f2fs_put_dnode(&dn);
 978                 } else {
 979                         struct page *psrc, *pdst;
 980
 981                         psrc = get_lock_data_page(src_inode, src + i, true);
 982                         if (IS_ERR(psrc))
 983                                 return PTR_ERR(psrc);
 984                         pdst = get_new_data_page(dst_inode, NULL, dst + i,
 985                                                                 true);
 986                         if (IS_ERR(pdst)) {
 987                                 f2fs_put_page(psrc, 1);
 988                                 return PTR_ERR(pdst);
 989                         }
 990                         f2fs_copy_page(psrc, pdst);
 991                         set_page_dirty(pdst);
 992                         f2fs_put_page(pdst, 1);
 993                         f2fs_put_page(psrc, 1);
 994
 995                         ret = truncate_hole(src_inode, src + i, src + i + 1);
 996                         if (ret)
 997                                 return ret;
 998                         i++;
 999                 }
1000         }
1001         return 0;
1002 }
1003
1004 static int __exchange_data_block(struct inode *src_inode,
1005                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1006                         pgoff_t len, bool full)
1007 {
1008         block_t *src_blkaddr;
1009         int *do_replace;
1010         pgoff_t olen;
1011         int ret;
1012
1013         while (len) {
1014                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1015
1016                 src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1017                 if (!src_blkaddr)
1018                         return -ENOMEM;
1019
1020                 do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1021                 if (!do_replace) {
1022                         kvfree(src_blkaddr);
1023                         return -ENOMEM;
1024                 }
1025
1026                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1027                                         do_replace, src, olen);
1028                 if (ret)
1029                         goto roll_back;
1030
1031                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1032                                         do_replace, src, dst, olen, full);
1033                 if (ret)
1034                         goto roll_back;
1035
1036                 src += olen;
1037                 dst += olen;
1038                 len -= olen;
1039
1040                 kvfree(src_blkaddr);
1041                 kvfree(do_replace);
1042         }
1043         return 0;
1044
1045 roll_back:
1046         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1047         kvfree(src_blkaddr);
1048         kvfree(do_replace);
1049         return ret;
1050 }
1051
1052 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1053 {
1054         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1055         pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1056         int ret;
1057
1058         f2fs_balance_fs(sbi, true);
1059         f2fs_lock_op(sbi);
1060
1061         f2fs_drop_extent_tree(inode);
1062
1063         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1064         f2fs_unlock_op(sbi);
1065         return ret;
1066 }
1067
1068 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1069 {
1070         pgoff_t pg_start, pg_end;
1071         loff_t new_size;
1072         int ret;
1073
1074         if (offset + len >= i_size_read(inode))
1075                 return -EINVAL;
1076
1077         /* collapse range should be aligned to block size of f2fs. */
1078         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1079                 return -EINVAL;
1080
1081         ret = f2fs_convert_inline_inode(inode);
1082         if (ret)
1083                 return ret;
1084
1085         pg_start = offset >> PAGE_SHIFT;
1086         pg_end = (offset + len) >> PAGE_SHIFT;
1087
1088         /* write out all dirty pages from offset */
1089         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1090         if (ret)
1091                 return ret;
1092
1093         truncate_pagecache(inode, offset);
1094
1095         ret = f2fs_do_collapse(inode, pg_start, pg_end);
1096         if (ret)
1097                 return ret;
1098
1099         /* write out all moved pages, if possible */
1100         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1101         truncate_pagecache(inode, offset);
1102
1103         new_size = i_size_read(inode) - len;
1104         truncate_pagecache(inode, new_size);
1105
1106         ret = truncate_blocks(inode, new_size, true);
1107         if (!ret)
1108                 f2fs_i_size_write(inode, new_size);
1109
1110         return ret;
1111 }
1112
1113 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1114                                                                 pgoff_t end)
1115 {
1116         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1117         pgoff_t index = start;
1118         unsigned int ofs_in_node = dn->ofs_in_node;
1119         blkcnt_t count = 0;
1120         int ret;
1121
1122         for (; index < end; index++, dn->ofs_in_node++) {
1123                 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR)
1124                         count++;
1125         }
1126
1127         dn->ofs_in_node = ofs_in_node;
1128         ret = reserve_new_blocks(dn, count);
1129         if (ret)
1130                 return ret;
1131
1132         dn->ofs_in_node = ofs_in_node;
1133         for (index = start; index < end; index++, dn->ofs_in_node++) {
1134                 dn->data_blkaddr =
1135                                 datablock_addr(dn->node_page, dn->ofs_in_node);
1136                 /*
1137                  * reserve_new_blocks will not guarantee entire block
1138                  * allocation.
1139                  */
1140                 if (dn->data_blkaddr == NULL_ADDR) {
1141                         ret = -ENOSPC;
1142                         break;
1143                 }
1144                 if (dn->data_blkaddr != NEW_ADDR) {
1145                         invalidate_blocks(sbi, dn->data_blkaddr);
1146                         dn->data_blkaddr = NEW_ADDR;
1147                         set_data_blkaddr(dn);
1148                 }
1149         }
1150
1151         f2fs_update_extent_cache_range(dn, start, 0, index - start);
1152
1153         return ret;
1154 }
1155
1156 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1157                                                                 int mode)
1158 {
1159         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1160         struct address_space *mapping = inode->i_mapping;
1161         pgoff_t index, pg_start, pg_end;
1162         loff_t new_size = i_size_read(inode);
1163         loff_t off_start, off_end;
1164         int ret = 0;
1165
1166         ret = inode_newsize_ok(inode, (len + offset));
1167         if (ret)
1168                 return ret;
1169
1170         ret = f2fs_convert_inline_inode(inode);
1171         if (ret)
1172                 return ret;
1173
1174         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1175         if (ret)
1176                 return ret;
1177
1178         truncate_pagecache_range(inode, offset, offset + len - 1);
1179
1180         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1181         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1182
1183         off_start = offset & (PAGE_SIZE - 1);
1184         off_end = (offset + len) & (PAGE_SIZE - 1);
1185
1186         if (pg_start == pg_end) {
1187                 ret = fill_zero(inode, pg_start, off_start,
1188                                                 off_end - off_start);
1189                 if (ret)
1190                         return ret;
1191
1192                 if (offset + len > new_size)
1193                         new_size = offset + len;
1194                 new_size = max_t(loff_t, new_size, offset + len);
1195         } else {
1196                 if (off_start) {
1197                         ret = fill_zero(inode, pg_start++, off_start,
1198                                                 PAGE_SIZE - off_start);
1199                         if (ret)
1200                                 return ret;
1201
1202                         new_size = max_t(loff_t, new_size,
1203                                         (loff_t)pg_start << PAGE_SHIFT);
1204                 }
1205
1206                 for (index = pg_start; index < pg_end;) {
1207                         struct dnode_of_data dn;
1208                         unsigned int end_offset;
1209                         pgoff_t end;
1210
1211                         f2fs_lock_op(sbi);
1212
1213                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1214                         ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1215                         if (ret) {
1216                                 f2fs_unlock_op(sbi);
1217                                 goto out;
1218                         }
1219
1220                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1221                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1222
1223                         ret = f2fs_do_zero_range(&dn, index, end);
1224                         f2fs_put_dnode(&dn);
1225                         f2fs_unlock_op(sbi);
1226
1227                         f2fs_balance_fs(sbi, dn.node_changed);
1228
1229                         if (ret)
1230                                 goto out;
1231
1232                         index = end;
1233                         new_size = max_t(loff_t, new_size,
1234                                         (loff_t)index << PAGE_SHIFT);
1235                 }
1236
1237                 if (off_end) {
1238                         ret = fill_zero(inode, pg_end, 0, off_end);
1239                         if (ret)
1240                                 goto out;
1241
1242                         new_size = max_t(loff_t, new_size, offset + len);
1243                 }
1244         }
1245
1246 out:
1247         if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1248                 f2fs_i_size_write(inode, new_size);
1249
1250         return ret;
1251 }
1252
1253 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1254 {
1255         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1256         pgoff_t nr, pg_start, pg_end, delta, idx;
1257         loff_t new_size;
1258         int ret = 0;
1259
1260         new_size = i_size_read(inode) + len;
1261         if (new_size > inode->i_sb->s_maxbytes)
1262                 return -EFBIG;
1263
1264         if (offset >= i_size_read(inode))
1265                 return -EINVAL;
1266
1267         /* insert range should be aligned to block size of f2fs. */
1268         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1269                 return -EINVAL;
1270
1271         ret = f2fs_convert_inline_inode(inode);
1272         if (ret)
1273                 return ret;
1274
1275         f2fs_balance_fs(sbi, true);
1276
1277         ret = truncate_blocks(inode, i_size_read(inode), true);
1278         if (ret)
1279                 return ret;
1280
1281         /* write out all dirty pages from offset */
1282         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1283         if (ret)
1284                 return ret;
1285
1286         truncate_pagecache(inode, offset);
1287
1288         pg_start = offset >> PAGE_SHIFT;
1289         pg_end = (offset + len) >> PAGE_SHIFT;
1290         delta = pg_end - pg_start;
1291         idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1292
1293         while (!ret && idx > pg_start) {
1294                 nr = idx - pg_start;
1295                 if (nr > delta)
1296                         nr = delta;
1297                 idx -= nr;
1298
1299                 f2fs_lock_op(sbi);
1300                 f2fs_drop_extent_tree(inode);
1301
1302                 ret = __exchange_data_block(inode, inode, idx,
1303                                         idx + delta, nr, false);
1304                 f2fs_unlock_op(sbi);
1305         }
1306
1307         /* write out all moved pages, if possible */
1308         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1309         truncate_pagecache(inode, offset);
1310
1311         if (!ret)
1312                 f2fs_i_size_write(inode, new_size);
1313         return ret;
1314 }
1315
1316 static int expand_inode_data(struct inode *inode, loff_t offset,
1317                                         loff_t len, int mode)
1318 {
1319         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1320         struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1321         pgoff_t pg_end;
1322         loff_t new_size = i_size_read(inode);
1323         loff_t off_end;
1324         int err;
1325
1326         err = inode_newsize_ok(inode, (len + offset));
1327         if (err)
1328                 return err;
1329
1330         err = f2fs_convert_inline_inode(inode);
1331         if (err)
1332                 return err;
1333
1334         f2fs_balance_fs(sbi, true);
1335
1336         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1337         off_end = (offset + len) & (PAGE_SIZE - 1);
1338
1339         map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1340         map.m_len = pg_end - map.m_lblk;
1341         if (off_end)
1342                 map.m_len++;
1343
1344         err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1345         if (err) {
1346                 pgoff_t last_off;
1347
1348                 if (!map.m_len)
1349                         return err;
1350
1351                 last_off = map.m_lblk + map.m_len - 1;
1352
1353                 /* update new size to the failed position */
1354                 new_size = (last_off == pg_end) ? offset + len:
1355                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1356         } else {
1357                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1358         }
1359
1360         if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1361                 f2fs_i_size_write(inode, new_size);
1362
1363         return err;
1364 }
1365
1366 static long f2fs_fallocate(struct file *file, int mode,
1367                                 loff_t offset, loff_t len)
1368 {
1369         struct inode *inode = file_inode(file);
1370         long ret = 0;
1371
1372         /* f2fs only support ->fallocate for regular file */
1373         if (!S_ISREG(inode->i_mode))
1374                 return -EINVAL;
1375
1376         if (f2fs_encrypted_inode(inode) &&
1377                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1378                 return -EOPNOTSUPP;
1379
1380         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1381                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1382                         FALLOC_FL_INSERT_RANGE))
1383                 return -EOPNOTSUPP;
1384
1385         inode_lock(inode);
1386
1387         if (mode & FALLOC_FL_PUNCH_HOLE) {
1388                 if (offset >= inode->i_size)
1389                         goto out;
1390
1391                 ret = punch_hole(inode, offset, len);
1392         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1393                 ret = f2fs_collapse_range(inode, offset, len);
1394         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1395                 ret = f2fs_zero_range(inode, offset, len, mode);
1396         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1397                 ret = f2fs_insert_range(inode, offset, len);
1398         } else {
1399                 ret = expand_inode_data(inode, offset, len, mode);
1400         }
1401
1402         if (!ret) {
1403                 inode->i_mtime = inode->i_ctime = current_time(inode);
1404                 f2fs_mark_inode_dirty_sync(inode, false);
1405                 if (mode & FALLOC_FL_KEEP_SIZE)
1406                         file_set_keep_isize(inode);
1407                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1408         }
1409
1410 out:
1411         inode_unlock(inode);
1412
1413         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1414         return ret;
1415 }
1416
1417 static int f2fs_release_file(struct inode *inode, struct file *filp)
1418 {
1419         /*
1420          * f2fs_relase_file is called at every close calls. So we should
1421          * not drop any inmemory pages by close called by other process.
1422          */
1423         if (!(filp->f_mode & FMODE_WRITE) ||
1424                         atomic_read(&inode->i_writecount) != 1)
1425                 return 0;
1426
1427         /* some remained atomic pages should discarded */
1428         if (f2fs_is_atomic_file(inode))
1429                 drop_inmem_pages(inode);
1430         if (f2fs_is_volatile_file(inode)) {
1431                 clear_inode_flag(inode, FI_VOLATILE_FILE);
1432                 set_inode_flag(inode, FI_DROP_CACHE);
1433                 filemap_fdatawrite(inode->i_mapping);
1434                 clear_inode_flag(inode, FI_DROP_CACHE);
1435         }
1436         return 0;
1437 }
1438
1439 #define F2FS_REG_FLMASK         (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
1440 #define F2FS_OTHER_FLMASK       (FS_NODUMP_FL | FS_NOATIME_FL)
1441
1442 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
1443 {
1444         if (S_ISDIR(mode))
1445                 return flags;
1446         else if (S_ISREG(mode))
1447                 return flags & F2FS_REG_FLMASK;
1448         else
1449                 return flags & F2FS_OTHER_FLMASK;
1450 }
1451
1452 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1453 {
1454         struct inode *inode = file_inode(filp);
1455         struct f2fs_inode_info *fi = F2FS_I(inode);
1456         unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE;
1457         return put_user(flags, (int __user *)arg);
1458 }
1459
1460 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1461 {
1462         struct inode *inode = file_inode(filp);
1463         struct f2fs_inode_info *fi = F2FS_I(inode);
1464         unsigned int flags;
1465         unsigned int oldflags;
1466         int ret;
1467
1468         if (!inode_owner_or_capable(inode))
1469                 return -EACCES;
1470
1471         if (get_user(flags, (int __user *)arg))
1472                 return -EFAULT;
1473
1474         ret = mnt_want_write_file(filp);
1475         if (ret)
1476                 return ret;
1477
1478         flags = f2fs_mask_flags(inode->i_mode, flags);
1479
1480         inode_lock(inode);
1481
1482         oldflags = fi->i_flags;
1483
1484         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
1485                 if (!capable(CAP_LINUX_IMMUTABLE)) {
1486                         inode_unlock(inode);
1487                         ret = -EPERM;
1488                         goto out;
1489                 }
1490         }
1491
1492         flags = flags & FS_FL_USER_MODIFIABLE;
1493         flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
1494         fi->i_flags = flags;
1495         inode_unlock(inode);
1496
1497         inode->i_ctime = current_time(inode);
1498         f2fs_set_inode_flags(inode);
1499 out:
1500         mnt_drop_write_file(filp);
1501         return ret;
1502 }
1503
1504 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1505 {
1506         struct inode *inode = file_inode(filp);
1507
1508         return put_user(inode->i_generation, (int __user *)arg);
1509 }
1510
1511 static int f2fs_ioc_start_atomic_write(struct file *filp)
1512 {
1513         struct inode *inode = file_inode(filp);
1514         int ret;
1515
1516         if (!inode_owner_or_capable(inode))
1517                 return -EACCES;
1518
1519         ret = mnt_want_write_file(filp);
1520         if (ret)
1521                 return ret;
1522
1523         inode_lock(inode);
1524
1525         if (f2fs_is_atomic_file(inode))
1526                 goto out;
1527
1528         ret = f2fs_convert_inline_inode(inode);
1529         if (ret)
1530                 goto out;
1531
1532         set_inode_flag(inode, FI_ATOMIC_FILE);
1533         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1534
1535         if (!get_dirty_pages(inode))
1536                 goto out;
1537
1538         f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1539                 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1540                                         inode->i_ino, get_dirty_pages(inode));
1541         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1542         if (ret)
1543                 clear_inode_flag(inode, FI_ATOMIC_FILE);
1544 out:
1545         inode_unlock(inode);
1546         mnt_drop_write_file(filp);
1547         return ret;
1548 }
1549
1550 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1551 {
1552         struct inode *inode = file_inode(filp);
1553         int ret;
1554
1555         if (!inode_owner_or_capable(inode))
1556                 return -EACCES;
1557
1558         ret = mnt_want_write_file(filp);
1559         if (ret)
1560                 return ret;
1561
1562         inode_lock(inode);
1563
1564         if (f2fs_is_volatile_file(inode))
1565                 goto err_out;
1566
1567         if (f2fs_is_atomic_file(inode)) {
1568                 clear_inode_flag(inode, FI_ATOMIC_FILE);
1569                 ret = commit_inmem_pages(inode);
1570                 if (ret) {
1571                         set_inode_flag(inode, FI_ATOMIC_FILE);
1572                         goto err_out;
1573                 }
1574         }
1575
1576         ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1577 err_out:
1578         inode_unlock(inode);
1579         mnt_drop_write_file(filp);
1580         return ret;
1581 }
1582
1583 static int f2fs_ioc_start_volatile_write(struct file *filp)
1584 {
1585         struct inode *inode = file_inode(filp);
1586         int ret;
1587
1588         if (!inode_owner_or_capable(inode))
1589                 return -EACCES;
1590
1591         ret = mnt_want_write_file(filp);
1592         if (ret)
1593                 return ret;
1594
1595         inode_lock(inode);
1596
1597         if (f2fs_is_volatile_file(inode))
1598                 goto out;
1599
1600         ret = f2fs_convert_inline_inode(inode);
1601         if (ret)
1602                 goto out;
1603
1604         set_inode_flag(inode, FI_VOLATILE_FILE);
1605         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1606 out:
1607         inode_unlock(inode);
1608         mnt_drop_write_file(filp);
1609         return ret;
1610 }
1611
1612 static int f2fs_ioc_release_volatile_write(struct file *filp)
1613 {
1614         struct inode *inode = file_inode(filp);
1615         int ret;
1616
1617         if (!inode_owner_or_capable(inode))
1618                 return -EACCES;
1619
1620         ret = mnt_want_write_file(filp);
1621         if (ret)
1622                 return ret;
1623
1624         inode_lock(inode);
1625
1626         if (!f2fs_is_volatile_file(inode))
1627                 goto out;
1628
1629         if (!f2fs_is_first_block_written(inode)) {
1630                 ret = truncate_partial_data_page(inode, 0, true);
1631                 goto out;
1632         }
1633
1634         ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1635 out:
1636         inode_unlock(inode);
1637         mnt_drop_write_file(filp);
1638         return ret;
1639 }
1640
1641 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1642 {
1643         struct inode *inode = file_inode(filp);
1644         int ret;
1645
1646         if (!inode_owner_or_capable(inode))
1647                 return -EACCES;
1648
1649         ret = mnt_want_write_file(filp);
1650         if (ret)
1651                 return ret;
1652
1653         inode_lock(inode);
1654
1655         if (f2fs_is_atomic_file(inode))
1656                 drop_inmem_pages(inode);
1657         if (f2fs_is_volatile_file(inode)) {
1658                 clear_inode_flag(inode, FI_VOLATILE_FILE);
1659                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1660         }
1661
1662         inode_unlock(inode);
1663
1664         mnt_drop_write_file(filp);
1665         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1666         return ret;
1667 }
1668
1669 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1670 {
1671         struct inode *inode = file_inode(filp);
1672         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1673         struct super_block *sb = sbi->sb;
1674         __u32 in;
1675         int ret;
1676
1677         if (!capable(CAP_SYS_ADMIN))
1678                 return -EPERM;
1679
1680         if (get_user(in, (__u32 __user *)arg))
1681                 return -EFAULT;
1682
1683         ret = mnt_want_write_file(filp);
1684         if (ret)
1685                 return ret;
1686
1687         switch (in) {
1688         case F2FS_GOING_DOWN_FULLSYNC:
1689                 sb = freeze_bdev(sb->s_bdev);
1690                 if (sb && !IS_ERR(sb)) {
1691                         f2fs_stop_checkpoint(sbi, false);
1692                         thaw_bdev(sb->s_bdev, sb);
1693                 }
1694                 break;
1695         case F2FS_GOING_DOWN_METASYNC:
1696                 /* do checkpoint only */
1697                 f2fs_sync_fs(sb, 1);
1698                 f2fs_stop_checkpoint(sbi, false);
1699                 break;
1700         case F2FS_GOING_DOWN_NOSYNC:
1701                 f2fs_stop_checkpoint(sbi, false);
1702                 break;
1703         case F2FS_GOING_DOWN_METAFLUSH:
1704                 sync_meta_pages(sbi, META, LONG_MAX);
1705                 f2fs_stop_checkpoint(sbi, false);
1706                 break;
1707         default:
1708                 ret = -EINVAL;
1709                 goto out;
1710         }
1711         f2fs_update_time(sbi, REQ_TIME);
1712 out:
1713         mnt_drop_write_file(filp);
1714         return ret;
1715 }
1716
1717 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1718 {
1719         struct inode *inode = file_inode(filp);
1720         struct super_block *sb = inode->i_sb;
1721         struct request_queue *q = bdev_get_queue(sb->s_bdev);
1722         struct fstrim_range range;
1723         int ret;
1724
1725         if (!capable(CAP_SYS_ADMIN))
1726                 return -EPERM;
1727
1728         if (!blk_queue_discard(q))
1729                 return -EOPNOTSUPP;
1730
1731         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1732                                 sizeof(range)))
1733                 return -EFAULT;
1734
1735         ret = mnt_want_write_file(filp);
1736         if (ret)
1737                 return ret;
1738
1739         range.minlen = max((unsigned int)range.minlen,
1740                                 q->limits.discard_granularity);
1741         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1742         mnt_drop_write_file(filp);
1743         if (ret < 0)
1744                 return ret;
1745
1746         if (copy_to_user((struct fstrim_range __user *)arg, &range,
1747                                 sizeof(range)))
1748                 return -EFAULT;
1749         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1750         return 0;
1751 }
1752
1753 static bool uuid_is_nonzero(__u8 u[16])
1754 {
1755         int i;
1756
1757         for (i = 0; i < 16; i++)
1758                 if (u[i])
1759                         return true;
1760         return false;
1761 }
1762
1763 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1764 {
1765         struct inode *inode = file_inode(filp);
1766
1767         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1768
1769         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1770 }
1771
1772 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1773 {
1774         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1775 }
1776
1777 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1778 {
1779         struct inode *inode = file_inode(filp);
1780         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1781         int err;
1782
1783         if (!f2fs_sb_has_crypto(inode->i_sb))
1784                 return -EOPNOTSUPP;
1785
1786         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1787                 goto got_it;
1788
1789         err = mnt_want_write_file(filp);
1790         if (err)
1791                 return err;
1792
1793         /* update superblock with uuid */
1794         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1795
1796         err = f2fs_commit_super(sbi, false);
1797         if (err) {
1798                 /* undo new data */
1799                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1800                 mnt_drop_write_file(filp);
1801                 return err;
1802         }
1803         mnt_drop_write_file(filp);
1804 got_it:
1805         if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1806                                                                         16))
1807                 return -EFAULT;
1808         return 0;
1809 }
1810
1811 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1812 {
1813         struct inode *inode = file_inode(filp);
1814         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1815         __u32 sync;
1816         int ret;
1817
1818         if (!capable(CAP_SYS_ADMIN))
1819                 return -EPERM;
1820
1821         if (get_user(sync, (__u32 __user *)arg))
1822                 return -EFAULT;
1823
1824         if (f2fs_readonly(sbi->sb))
1825                 return -EROFS;
1826
1827         ret = mnt_want_write_file(filp);
1828         if (ret)
1829                 return ret;
1830
1831         if (!sync) {
1832                 if (!mutex_trylock(&sbi->gc_mutex)) {
1833                         ret = -EBUSY;
1834                         goto out;
1835                 }
1836         } else {
1837                 mutex_lock(&sbi->gc_mutex);
1838         }
1839
1840         ret = f2fs_gc(sbi, sync, true);
1841 out:
1842         mnt_drop_write_file(filp);
1843         return ret;
1844 }
1845
1846 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
1847 {
1848         struct inode *inode = file_inode(filp);
1849         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1850         int ret;
1851
1852         if (!capable(CAP_SYS_ADMIN))
1853                 return -EPERM;
1854
1855         if (f2fs_readonly(sbi->sb))
1856                 return -EROFS;
1857
1858         ret = mnt_want_write_file(filp);
1859         if (ret)
1860                 return ret;
1861
1862         ret = f2fs_sync_fs(sbi->sb, 1);
1863
1864         mnt_drop_write_file(filp);
1865         return ret;
1866 }
1867
1868 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
1869                                         struct file *filp,
1870                                         struct f2fs_defragment *range)
1871 {
1872         struct inode *inode = file_inode(filp);
1873         struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1874         struct extent_info ei;
1875         pgoff_t pg_start, pg_end;
1876         unsigned int blk_per_seg = sbi->blocks_per_seg;
1877         unsigned int total = 0, sec_num;
1878         unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg;
1879         block_t blk_end = 0;
1880         bool fragmented = false;
1881         int err;
1882
1883         /* if in-place-update policy is enabled, don't waste time here */
1884         if (need_inplace_update(inode))
1885                 return -EINVAL;
1886
1887         pg_start = range->start >> PAGE_SHIFT;
1888         pg_end = (range->start + range->len) >> PAGE_SHIFT;
1889
1890         f2fs_balance_fs(sbi, true);
1891
1892         inode_lock(inode);
1893
1894         /* writeback all dirty pages in the range */
1895         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
1896                                                 range->start + range->len - 1);
1897         if (err)
1898                 goto out;
1899
1900         /*
1901          * lookup mapping info in extent cache, skip defragmenting if physical
1902          * block addresses are continuous.
1903          */
1904         if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
1905                 if (ei.fofs + ei.len >= pg_end)
1906                         goto out;
1907         }
1908
1909         map.m_lblk = pg_start;
1910
1911         /*
1912          * lookup mapping info in dnode page cache, skip defragmenting if all
1913          * physical block addresses are continuous even if there are hole(s)
1914          * in logical blocks.
1915          */
1916         while (map.m_lblk < pg_end) {
1917                 map.m_len = pg_end - map.m_lblk;
1918                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1919                 if (err)
1920                         goto out;
1921
1922                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1923                         map.m_lblk++;
1924                         continue;
1925                 }
1926
1927                 if (blk_end && blk_end != map.m_pblk) {
1928                         fragmented = true;
1929                         break;
1930                 }
1931                 blk_end = map.m_pblk + map.m_len;
1932
1933                 map.m_lblk += map.m_len;
1934         }
1935
1936         if (!fragmented)
1937                 goto out;
1938
1939         map.m_lblk = pg_start;
1940         map.m_len = pg_end - pg_start;
1941
1942         sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec;
1943
1944         /*
1945          * make sure there are enough free section for LFS allocation, this can
1946          * avoid defragment running in SSR mode when free section are allocated
1947          * intensively
1948          */
1949         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
1950                 err = -EAGAIN;
1951                 goto out;
1952         }
1953
1954         while (map.m_lblk < pg_end) {
1955                 pgoff_t idx;
1956                 int cnt = 0;
1957
1958 do_map:
1959                 map.m_len = pg_end - map.m_lblk;
1960                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ);
1961                 if (err)
1962                         goto clear_out;
1963
1964                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
1965                         map.m_lblk++;
1966                         continue;
1967                 }
1968
1969                 set_inode_flag(inode, FI_DO_DEFRAG);
1970
1971                 idx = map.m_lblk;
1972                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
1973                         struct page *page;
1974
1975                         page = get_lock_data_page(inode, idx, true);
1976                         if (IS_ERR(page)) {
1977                                 err = PTR_ERR(page);
1978                                 goto clear_out;
1979                         }
1980
1981                         set_page_dirty(page);
1982                         f2fs_put_page(page, 1);
1983
1984                         idx++;
1985                         cnt++;
1986                         total++;
1987                 }
1988
1989                 map.m_lblk = idx;
1990
1991                 if (idx < pg_end && cnt < blk_per_seg)
1992                         goto do_map;
1993
1994                 clear_inode_flag(inode, FI_DO_DEFRAG);
1995
1996                 err = filemap_fdatawrite(inode->i_mapping);
1997                 if (err)
1998                         goto out;
1999         }
2000 clear_out:
2001         clear_inode_flag(inode, FI_DO_DEFRAG);
2002 out:
2003         inode_unlock(inode);
2004         if (!err)
2005                 range->len = (u64)total << PAGE_SHIFT;
2006         return err;
2007 }
2008
2009 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2010 {
2011         struct inode *inode = file_inode(filp);
2012         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2013         struct f2fs_defragment range;
2014         int err;
2015
2016         if (!capable(CAP_SYS_ADMIN))
2017                 return -EPERM;
2018
2019         if (!S_ISREG(inode->i_mode))
2020                 return -EINVAL;
2021
2022         err = mnt_want_write_file(filp);
2023         if (err)
2024                 return err;
2025
2026         if (f2fs_readonly(sbi->sb)) {
2027                 err = -EROFS;
2028                 goto out;
2029         }
2030
2031         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2032                                                         sizeof(range))) {
2033                 err = -EFAULT;
2034                 goto out;
2035         }
2036
2037         /* verify alignment of offset & size */
2038         if (range.start & (F2FS_BLKSIZE - 1) ||
2039                 range.len & (F2FS_BLKSIZE - 1)) {
2040                 err = -EINVAL;
2041                 goto out;
2042         }
2043
2044         err = f2fs_defragment_range(sbi, filp, &range);
2045         f2fs_update_time(sbi, REQ_TIME);
2046         if (err < 0)
2047                 goto out;
2048
2049         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2050                                                         sizeof(range)))
2051                 err = -EFAULT;
2052 out:
2053         mnt_drop_write_file(filp);
2054         return err;
2055 }
2056
2057 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2058                         struct file *file_out, loff_t pos_out, size_t len)
2059 {
2060         struct inode *src = file_inode(file_in);
2061         struct inode *dst = file_inode(file_out);
2062         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2063         size_t olen = len, dst_max_i_size = 0;
2064         size_t dst_osize;
2065         int ret;
2066
2067         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2068                                 src->i_sb != dst->i_sb)
2069                 return -EXDEV;
2070
2071         if (unlikely(f2fs_readonly(src->i_sb)))
2072                 return -EROFS;
2073
2074         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2075                 return -EINVAL;
2076
2077         if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2078                 return -EOPNOTSUPP;
2079
2080         if (src == dst) {
2081                 if (pos_in == pos_out)
2082                         return 0;
2083                 if (pos_out > pos_in && pos_out < pos_in + len)
2084                         return -EINVAL;
2085         }
2086
2087         inode_lock(src);
2088         if (src != dst) {
2089                 if (!inode_trylock(dst)) {
2090                         ret = -EBUSY;
2091                         goto out;
2092                 }
2093         }
2094
2095         ret = -EINVAL;
2096         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2097                 goto out_unlock;
2098         if (len == 0)
2099                 olen = len = src->i_size - pos_in;
2100         if (pos_in + len == src->i_size)
2101                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2102         if (len == 0) {
2103                 ret = 0;
2104                 goto out_unlock;
2105         }
2106
2107         dst_osize = dst->i_size;
2108         if (pos_out + olen > dst->i_size)
2109                 dst_max_i_size = pos_out + olen;
2110
2111         /* verify the end result is block aligned */
2112         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2113                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2114                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2115                 goto out_unlock;
2116
2117         ret = f2fs_convert_inline_inode(src);
2118         if (ret)
2119                 goto out_unlock;
2120
2121         ret = f2fs_convert_inline_inode(dst);
2122         if (ret)
2123                 goto out_unlock;
2124
2125         /* write out all dirty pages from offset */
2126         ret = filemap_write_and_wait_range(src->i_mapping,
2127                                         pos_in, pos_in + len);
2128         if (ret)
2129                 goto out_unlock;
2130
2131         ret = filemap_write_and_wait_range(dst->i_mapping,
2132                                         pos_out, pos_out + len);
2133         if (ret)
2134                 goto out_unlock;
2135
2136         f2fs_balance_fs(sbi, true);
2137         f2fs_lock_op(sbi);
2138         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2139                                 pos_out >> F2FS_BLKSIZE_BITS,
2140                                 len >> F2FS_BLKSIZE_BITS, false);
2141
2142         if (!ret) {
2143                 if (dst_max_i_size)
2144                         f2fs_i_size_write(dst, dst_max_i_size);
2145                 else if (dst_osize != dst->i_size)
2146                         f2fs_i_size_write(dst, dst_osize);
2147         }
2148         f2fs_unlock_op(sbi);
2149 out_unlock:
2150         if (src != dst)
2151                 inode_unlock(dst);
2152 out:
2153         inode_unlock(src);
2154         return ret;
2155 }
2156
2157 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2158 {
2159         struct f2fs_move_range range;
2160         struct fd dst;
2161         int err;
2162
2163         if (!(filp->f_mode & FMODE_READ) ||
2164                         !(filp->f_mode & FMODE_WRITE))
2165                 return -EBADF;
2166
2167         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2168                                                         sizeof(range)))
2169                 return -EFAULT;
2170
2171         dst = fdget(range.dst_fd);
2172         if (!dst.file)
2173                 return -EBADF;
2174
2175         if (!(dst.file->f_mode & FMODE_WRITE)) {
2176                 err = -EBADF;
2177                 goto err_out;
2178         }
2179
2180         err = mnt_want_write_file(filp);
2181         if (err)
2182                 goto err_out;
2183
2184         err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2185                                         range.pos_out, range.len);
2186
2187         mnt_drop_write_file(filp);
2188
2189         if (copy_to_user((struct f2fs_move_range __user *)arg,
2190                                                 &range, sizeof(range)))
2191                 err = -EFAULT;
2192 err_out:
2193         fdput(dst);
2194         return err;
2195 }
2196
2197 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2198 {
2199         switch (cmd) {
2200         case F2FS_IOC_GETFLAGS:
2201                 return f2fs_ioc_getflags(filp, arg);
2202         case F2FS_IOC_SETFLAGS:
2203                 return f2fs_ioc_setflags(filp, arg);
2204         case F2FS_IOC_GETVERSION:
2205                 return f2fs_ioc_getversion(filp, arg);
2206         case F2FS_IOC_START_ATOMIC_WRITE:
2207                 return f2fs_ioc_start_atomic_write(filp);
2208         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2209                 return f2fs_ioc_commit_atomic_write(filp);
2210         case F2FS_IOC_START_VOLATILE_WRITE:
2211                 return f2fs_ioc_start_volatile_write(filp);
2212         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2213                 return f2fs_ioc_release_volatile_write(filp);
2214         case F2FS_IOC_ABORT_VOLATILE_WRITE:
2215                 return f2fs_ioc_abort_volatile_write(filp);
2216         case F2FS_IOC_SHUTDOWN:
2217                 return f2fs_ioc_shutdown(filp, arg);
2218         case FITRIM:
2219                 return f2fs_ioc_fitrim(filp, arg);
2220         case F2FS_IOC_SET_ENCRYPTION_POLICY:
2221                 return f2fs_ioc_set_encryption_policy(filp, arg);
2222         case F2FS_IOC_GET_ENCRYPTION_POLICY:
2223                 return f2fs_ioc_get_encryption_policy(filp, arg);
2224         case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2225                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2226         case F2FS_IOC_GARBAGE_COLLECT:
2227                 return f2fs_ioc_gc(filp, arg);
2228         case F2FS_IOC_WRITE_CHECKPOINT:
2229                 return f2fs_ioc_write_checkpoint(filp, arg);
2230         case F2FS_IOC_DEFRAGMENT:
2231                 return f2fs_ioc_defragment(filp, arg);
2232         case F2FS_IOC_MOVE_RANGE:
2233                 return f2fs_ioc_move_range(filp, arg);
2234         default:
2235                 return -ENOTTY;
2236         }
2237 }
2238
2239 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2240 {
2241         struct file *file = iocb->ki_filp;
2242         struct inode *inode = file_inode(file);
2243         struct blk_plug plug;
2244         ssize_t ret;
2245
2246         if (f2fs_encrypted_inode(inode) &&
2247                                 !fscrypt_has_encryption_key(inode) &&
2248                                 fscrypt_get_encryption_info(inode))
2249                 return -EACCES;
2250
2251         inode_lock(inode);
2252         ret = generic_write_checks(iocb, from);
2253         if (ret > 0) {
2254                 int err = f2fs_preallocate_blocks(iocb, from);
2255
2256                 if (err) {
2257                         inode_unlock(inode);
2258                         return err;
2259                 }
2260                 blk_start_plug(&plug);
2261                 ret = __generic_file_write_iter(iocb, from);
2262                 blk_finish_plug(&plug);
2263         }
2264         inode_unlock(inode);
2265
2266         if (ret > 0)
2267                 ret = generic_write_sync(iocb, ret);
2268         return ret;
2269 }
2270
2271 #ifdef CONFIG_COMPAT
2272 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2273 {
2274         switch (cmd) {
2275         case F2FS_IOC32_GETFLAGS:
2276                 cmd = F2FS_IOC_GETFLAGS;
2277                 break;
2278         case F2FS_IOC32_SETFLAGS:
2279                 cmd = F2FS_IOC_SETFLAGS;
2280                 break;
2281         case F2FS_IOC32_GETVERSION:
2282                 cmd = F2FS_IOC_GETVERSION;
2283                 break;
2284         case F2FS_IOC_START_ATOMIC_WRITE:
2285         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2286         case F2FS_IOC_START_VOLATILE_WRITE:
2287         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2288         case F2FS_IOC_ABORT_VOLATILE_WRITE:
2289         case F2FS_IOC_SHUTDOWN:
2290         case F2FS_IOC_SET_ENCRYPTION_POLICY:
2291         case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2292         case F2FS_IOC_GET_ENCRYPTION_POLICY:
2293         case F2FS_IOC_GARBAGE_COLLECT:
2294         case F2FS_IOC_WRITE_CHECKPOINT:
2295         case F2FS_IOC_DEFRAGMENT:
2296                 break;
2297         case F2FS_IOC_MOVE_RANGE:
2298                 break;
2299         default:
2300                 return -ENOIOCTLCMD;
2301         }
2302         return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2303 }
2304 #endif
2305
2306 const struct file_operations f2fs_file_operations = {
2307         .llseek         = f2fs_llseek,
2308         .read_iter      = generic_file_read_iter,
2309         .write_iter     = f2fs_file_write_iter,
2310         .open           = f2fs_file_open,
2311         .release        = f2fs_release_file,
2312         .mmap           = f2fs_file_mmap,
2313         .fsync          = f2fs_sync_file,
2314         .fallocate      = f2fs_fallocate,
2315         .unlocked_ioctl = f2fs_ioctl,
2316 #ifdef CONFIG_COMPAT
2317         .compat_ioctl   = f2fs_compat_ioctl,
2318 #endif
2319         .splice_read    = generic_file_splice_read,
2320         .splice_write   = iter_file_splice_write,
2321 };