fs/f2fs/file.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * fs/f2fs/file.c
   4  *
   5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6  *             http://www.samsung.com/
   7  */
   8 #include <linux/fs.h>
   9 #include <linux/f2fs_fs.h>
  10 #include <linux/stat.h>
  11 #include <linux/buffer_head.h>
  12 #include <linux/writeback.h>
  13 #include <linux/blkdev.h>
  14 #include <linux/falloc.h>
  15 #include <linux/types.h>
  16 #include <linux/compat.h>
  17 #include <linux/uaccess.h>
  18 #include <linux/mount.h>
  19 #include <linux/pagevec.h>
  20 #include <linux/uio.h>
  21 #include <linux/uuid.h>
  22 #include <linux/file.h>
  23 #include <linux/nls.h>
  24 #include <linux/sched/signal.h>
  25 #include <linux/fileattr.h>
  26 #include <linux/fadvise.h>
  27 #include <linux/iomap.h>
  28
  29 #include "f2fs.h"
  30 #include "node.h"
  31 #include "segment.h"
  32 #include "xattr.h"
  33 #include "acl.h"
  34 #include "gc.h"
  35 #include "iostat.h"
  36 #include <trace/events/f2fs.h>
  37 #include <uapi/linux/f2fs.h>
  38
  39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
  40 {
  41         struct inode *inode = file_inode(vmf->vma->vm_file);
  42         vm_fault_t ret;
  43
  44         ret = filemap_fault(vmf);
  45         if (!ret)
  46                 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
  47                                                         F2FS_BLKSIZE);
  48
  49         trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
  50
  51         return ret;
  52 }
  53
  54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
  55 {
  56         struct page *page = vmf->page;
  57         struct inode *inode = file_inode(vmf->vma->vm_file);
  58         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
  59         struct dnode_of_data dn;
  60         bool need_alloc = true;
  61         int err = 0;
  62
  63         if (unlikely(IS_IMMUTABLE(inode)))
  64                 return VM_FAULT_SIGBUS;
  65
  66         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
  67                 return VM_FAULT_SIGBUS;
  68
  69         if (unlikely(f2fs_cp_error(sbi))) {
  70                 err = -EIO;
  71                 goto err;
  72         }
  73
  74         if (!f2fs_is_checkpoint_ready(sbi)) {
  75                 err = -ENOSPC;
  76                 goto err;
  77         }
  78
  79         err = f2fs_convert_inline_inode(inode);
  80         if (err)
  81                 goto err;
  82
  83 #ifdef CONFIG_F2FS_FS_COMPRESSION
  84         if (f2fs_compressed_file(inode)) {
  85                 int ret = f2fs_is_compressed_cluster(inode, page->index);
  86
  87                 if (ret < 0) {
  88                         err = ret;
  89                         goto err;
  90                 } else if (ret) {
  91                         need_alloc = false;
  92                 }
  93         }
  94 #endif
  95         /* should do out of any locked page */
  96         if (need_alloc)
  97                 f2fs_balance_fs(sbi, true);
  98
  99         sb_start_pagefault(inode->i_sb);
 100
 101         f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
 102
 103         file_update_time(vmf->vma->vm_file);
 104         filemap_invalidate_lock_shared(inode->i_mapping);
 105         lock_page(page);
 106         if (unlikely(page->mapping != inode->i_mapping ||
 107                         page_offset(page) > i_size_read(inode) ||
 108                         !PageUptodate(page))) {
 109                 unlock_page(page);
 110                 err = -EFAULT;
 111                 goto out_sem;
 112         }
 113
 114         if (need_alloc) {
 115                 /* block allocation */
 116                 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
 117                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 118                 err = f2fs_get_block(&dn, page->index);
 119                 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
 120         }
 121
 122 #ifdef CONFIG_F2FS_FS_COMPRESSION
 123         if (!need_alloc) {
 124                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 125                 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
 126                 f2fs_put_dnode(&dn);
 127         }
 128 #endif
 129         if (err) {
 130                 unlock_page(page);
 131                 goto out_sem;
 132         }
 133
 134         f2fs_wait_on_page_writeback(page, DATA, false, true);
 135
 136         /* wait for GCed page writeback via META_MAPPING */
 137         f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
 138
 139         /*
 140          * check to see if the page is mapped already (no holes)
 141          */
 142         if (PageMappedToDisk(page))
 143                 goto out_sem;
 144
 145         /* page is wholly or partially inside EOF */
 146         if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
 147                                                 i_size_read(inode)) {
 148                 loff_t offset;
 149
 150                 offset = i_size_read(inode) & ~PAGE_MASK;
 151                 zero_user_segment(page, offset, PAGE_SIZE);
 152         }
 153         set_page_dirty(page);
 154         if (!PageUptodate(page))
 155                 SetPageUptodate(page);
 156
 157         f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
 158         f2fs_update_time(sbi, REQ_TIME);
 159
 160         trace_f2fs_vm_page_mkwrite(page, DATA);
 161 out_sem:
 162         filemap_invalidate_unlock_shared(inode->i_mapping);
 163
 164         sb_end_pagefault(inode->i_sb);
 165 err:
 166         return block_page_mkwrite_return(err);
 167 }
 168
 169 static const struct vm_operations_struct f2fs_file_vm_ops = {
 170         .fault          = f2fs_filemap_fault,
 171         .map_pages      = filemap_map_pages,
 172         .page_mkwrite   = f2fs_vm_page_mkwrite,
 173 };
 174
 175 static int get_parent_ino(struct inode *inode, nid_t *pino)
 176 {
 177         struct dentry *dentry;
 178
 179         /*
 180          * Make sure to get the non-deleted alias.  The alias associated with
 181          * the open file descriptor being fsync()'ed may be deleted already.
 182          */
 183         dentry = d_find_alias(inode);
 184         if (!dentry)
 185                 return 0;
 186
 187         *pino = parent_ino(dentry);
 188         dput(dentry);
 189         return 1;
 190 }
 191
 192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
 193 {
 194         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 195         enum cp_reason_type cp_reason = CP_NO_NEEDED;
 196
 197         if (!S_ISREG(inode->i_mode))
 198                 cp_reason = CP_NON_REGULAR;
 199         else if (f2fs_compressed_file(inode))
 200                 cp_reason = CP_COMPRESSED;
 201         else if (inode->i_nlink != 1)
 202                 cp_reason = CP_HARDLINK;
 203         else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
 204                 cp_reason = CP_SB_NEED_CP;
 205         else if (file_wrong_pino(inode))
 206                 cp_reason = CP_WRONG_PINO;
 207         else if (!f2fs_space_for_roll_forward(sbi))
 208                 cp_reason = CP_NO_SPC_ROLL;
 209         else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
 210                 cp_reason = CP_NODE_NEED_CP;
 211         else if (test_opt(sbi, FASTBOOT))
 212                 cp_reason = CP_FASTBOOT_MODE;
 213         else if (F2FS_OPTION(sbi).active_logs == 2)
 214                 cp_reason = CP_SPEC_LOG_NUM;
 215         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
 216                 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
 217                 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
 218                                                         TRANS_DIR_INO))
 219                 cp_reason = CP_RECOVER_DIR;
 220
 221         return cp_reason;
 222 }
 223
 224 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
 225 {
 226         struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
 227         bool ret = false;
 228         /* But we need to avoid that there are some inode updates */
 229         if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
 230                 ret = true;
 231         f2fs_put_page(i, 0);
 232         return ret;
 233 }
 234
 235 static void try_to_fix_pino(struct inode *inode)
 236 {
 237         struct f2fs_inode_info *fi = F2FS_I(inode);
 238         nid_t pino;
 239
 240         down_write(&fi->i_sem);
 241         if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
 242                         get_parent_ino(inode, &pino)) {
 243                 f2fs_i_pino_write(inode, pino);
 244                 file_got_pino(inode);
 245         }
 246         up_write(&fi->i_sem);
 247 }
 248
 249 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
 250                                                 int datasync, bool atomic)
 251 {
 252         struct inode *inode = file->f_mapping->host;
 253         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 254         nid_t ino = inode->i_ino;
 255         int ret = 0;
 256         enum cp_reason_type cp_reason = 0;
 257         struct writeback_control wbc = {
 258                 .sync_mode = WB_SYNC_ALL,
 259                 .nr_to_write = LONG_MAX,
 260                 .for_reclaim = 0,
 261         };
 262         unsigned int seq_id = 0;
 263
 264         if (unlikely(f2fs_readonly(inode->i_sb)))
 265                 return 0;
 266
 267         trace_f2fs_sync_file_enter(inode);
 268
 269         if (S_ISDIR(inode->i_mode))
 270                 goto go_write;
 271
 272         /* if fdatasync is triggered, let's do in-place-update */
 273         if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
 274                 set_inode_flag(inode, FI_NEED_IPU);
 275         ret = file_write_and_wait_range(file, start, end);
 276         clear_inode_flag(inode, FI_NEED_IPU);
 277
 278         if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
 279                 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 280                 return ret;
 281         }
 282
 283         /* if the inode is dirty, let's recover all the time */
 284         if (!f2fs_skip_inode_update(inode, datasync)) {
 285                 f2fs_write_inode(inode, NULL);
 286                 goto go_write;
 287         }
 288
 289         /*
 290          * if there is no written data, don't waste time to write recovery info.
 291          */
 292         if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
 293                         !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
 294
 295                 /* it may call write_inode just prior to fsync */
 296                 if (need_inode_page_update(sbi, ino))
 297                         goto go_write;
 298
 299                 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
 300                                 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
 301                         goto flush_out;
 302                 goto out;
 303         } else {
 304                 /*
 305                  * for OPU case, during fsync(), node can be persisted before
 306                  * data when lower device doesn't support write barrier, result
 307                  * in data corruption after SPO.
 308                  * So for strict fsync mode, force to use atomic write sematics
 309                  * to keep write order in between data/node and last node to
 310                  * avoid potential data corruption.
 311                  */
 312                 if (F2FS_OPTION(sbi).fsync_mode ==
 313                                 FSYNC_MODE_STRICT && !atomic)
 314                         atomic = true;
 315         }
 316 go_write:
 317         /*
 318          * Both of fdatasync() and fsync() are able to be recovered from
 319          * sudden-power-off.
 320          */
 321         down_read(&F2FS_I(inode)->i_sem);
 322         cp_reason = need_do_checkpoint(inode);
 323         up_read(&F2FS_I(inode)->i_sem);
 324
 325         if (cp_reason) {
 326                 /* all the dirty node pages should be flushed for POR */
 327                 ret = f2fs_sync_fs(inode->i_sb, 1);
 328
 329                 /*
 330                  * We've secured consistency through sync_fs. Following pino
 331                  * will be used only for fsynced inodes after checkpoint.
 332                  */
 333                 try_to_fix_pino(inode);
 334                 clear_inode_flag(inode, FI_APPEND_WRITE);
 335                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 336                 goto out;
 337         }
 338 sync_nodes:
 339         atomic_inc(&sbi->wb_sync_req[NODE]);
 340         ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
 341         atomic_dec(&sbi->wb_sync_req[NODE]);
 342         if (ret)
 343                 goto out;
 344
 345         /* if cp_error was enabled, we should avoid infinite loop */
 346         if (unlikely(f2fs_cp_error(sbi))) {
 347                 ret = -EIO;
 348                 goto out;
 349         }
 350
 351         if (f2fs_need_inode_block_update(sbi, ino)) {
 352                 f2fs_mark_inode_dirty_sync(inode, true);
 353                 f2fs_write_inode(inode, NULL);
 354                 goto sync_nodes;
 355         }
 356
 357         /*
 358          * If it's atomic_write, it's just fine to keep write ordering. So
 359          * here we don't need to wait for node write completion, since we use
 360          * node chain which serializes node blocks. If one of node writes are
 361          * reordered, we can see simply broken chain, resulting in stopping
 362          * roll-forward recovery. It means we'll recover all or none node blocks
 363          * given fsync mark.
 364          */
 365         if (!atomic) {
 366                 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
 367                 if (ret)
 368                         goto out;
 369         }
 370
 371         /* once recovery info is written, don't need to tack this */
 372         f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
 373         clear_inode_flag(inode, FI_APPEND_WRITE);
 374 flush_out:
 375         if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
 376                 ret = f2fs_issue_flush(sbi, inode->i_ino);
 377         if (!ret) {
 378                 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
 379                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 380                 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
 381         }
 382         f2fs_update_time(sbi, REQ_TIME);
 383 out:
 384         trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 385         return ret;
 386 }
 387
 388 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 389 {
 390         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
 391                 return -EIO;
 392         return f2fs_do_sync_file(file, start, end, datasync, false);
 393 }
 394
 395 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
 396                                 pgoff_t index, int whence)
 397 {
 398         switch (whence) {
 399         case SEEK_DATA:
 400                 if (__is_valid_data_blkaddr(blkaddr))
 401                         return true;
 402                 if (blkaddr == NEW_ADDR &&
 403                     xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
 404                         return true;
 405                 break;
 406         case SEEK_HOLE:
 407                 if (blkaddr == NULL_ADDR)
 408                         return true;
 409                 break;
 410         }
 411         return false;
 412 }
 413
 414 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 415 {
 416         struct inode *inode = file->f_mapping->host;
 417         loff_t maxbytes = inode->i_sb->s_maxbytes;
 418         struct dnode_of_data dn;
 419         pgoff_t pgofs, end_offset;
 420         loff_t data_ofs = offset;
 421         loff_t isize;
 422         int err = 0;
 423
 424         inode_lock(inode);
 425
 426         isize = i_size_read(inode);
 427         if (offset >= isize)
 428                 goto fail;
 429
 430         /* handle inline data case */
 431         if (f2fs_has_inline_data(inode)) {
 432                 if (whence == SEEK_HOLE) {
 433                         data_ofs = isize;
 434                         goto found;
 435                 } else if (whence == SEEK_DATA) {
 436                         data_ofs = offset;
 437                         goto found;
 438                 }
 439         }
 440
 441         pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
 442
 443         for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 444                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 445                 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
 446                 if (err && err != -ENOENT) {
 447                         goto fail;
 448                 } else if (err == -ENOENT) {
 449                         /* direct node does not exists */
 450                         if (whence == SEEK_DATA) {
 451                                 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
 452                                 continue;
 453                         } else {
 454                                 goto found;
 455                         }
 456                 }
 457
 458                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 459
 460                 /* find data/hole in dnode block */
 461                 for (; dn.ofs_in_node < end_offset;
 462                                 dn.ofs_in_node++, pgofs++,
 463                                 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 464                         block_t blkaddr;
 465
 466                         blkaddr = f2fs_data_blkaddr(&dn);
 467
 468                         if (__is_valid_data_blkaddr(blkaddr) &&
 469                                 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
 470                                         blkaddr, DATA_GENERIC_ENHANCE)) {
 471                                 f2fs_put_dnode(&dn);
 472                                 goto fail;
 473                         }
 474
 475                         if (__found_offset(file->f_mapping, blkaddr,
 476                                                         pgofs, whence)) {
 477                                 f2fs_put_dnode(&dn);
 478                                 goto found;
 479                         }
 480                 }
 481                 f2fs_put_dnode(&dn);
 482         }
 483
 484         if (whence == SEEK_DATA)
 485                 goto fail;
 486 found:
 487         if (whence == SEEK_HOLE && data_ofs > isize)
 488                 data_ofs = isize;
 489         inode_unlock(inode);
 490         return vfs_setpos(file, data_ofs, maxbytes);
 491 fail:
 492         inode_unlock(inode);
 493         return -ENXIO;
 494 }
 495
 496 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
 497 {
 498         struct inode *inode = file->f_mapping->host;
 499         loff_t maxbytes = inode->i_sb->s_maxbytes;
 500
 501         if (f2fs_compressed_file(inode))
 502                 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
 503
 504         switch (whence) {
 505         case SEEK_SET:
 506         case SEEK_CUR:
 507         case SEEK_END:
 508                 return generic_file_llseek_size(file, offset, whence,
 509                                                 maxbytes, i_size_read(inode));
 510         case SEEK_DATA:
 511         case SEEK_HOLE:
 512                 if (offset < 0)
 513                         return -ENXIO;
 514                 return f2fs_seek_block(file, offset, whence);
 515         }
 516
 517         return -EINVAL;
 518 }
 519
 520 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 521 {
 522         struct inode *inode = file_inode(file);
 523
 524         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 525                 return -EIO;
 526
 527         if (!f2fs_is_compress_backend_ready(inode))
 528                 return -EOPNOTSUPP;
 529
 530         file_accessed(file);
 531         vma->vm_ops = &f2fs_file_vm_ops;
 532         set_inode_flag(inode, FI_MMAP_FILE);
 533         return 0;
 534 }
 535
 536 static int f2fs_file_open(struct inode *inode, struct file *filp)
 537 {
 538         int err = fscrypt_file_open(inode, filp);
 539
 540         if (err)
 541                 return err;
 542
 543         if (!f2fs_is_compress_backend_ready(inode))
 544                 return -EOPNOTSUPP;
 545
 546         err = fsverity_file_open(inode, filp);
 547         if (err)
 548                 return err;
 549
 550         filp->f_mode |= FMODE_NOWAIT;
 551
 552         return dquot_file_open(inode, filp);
 553 }
 554
 555 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 556 {
 557         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 558         struct f2fs_node *raw_node;
 559         int nr_free = 0, ofs = dn->ofs_in_node, len = count;
 560         __le32 *addr;
 561         int base = 0;
 562         bool compressed_cluster = false;
 563         int cluster_index = 0, valid_blocks = 0;
 564         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
 565         bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
 566
 567         if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
 568                 base = get_extra_isize(dn->inode);
 569
 570         raw_node = F2FS_NODE(dn->node_page);
 571         addr = blkaddr_in_node(raw_node) + base + ofs;
 572
 573         /* Assumption: truncateion starts with cluster */
 574         for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
 575                 block_t blkaddr = le32_to_cpu(*addr);
 576
 577                 if (f2fs_compressed_file(dn->inode) &&
 578                                         !(cluster_index & (cluster_size - 1))) {
 579                         if (compressed_cluster)
 580                                 f2fs_i_compr_blocks_update(dn->inode,
 581                                                         valid_blocks, false);
 582                         compressed_cluster = (blkaddr == COMPRESS_ADDR);
 583                         valid_blocks = 0;
 584                 }
 585
 586                 if (blkaddr == NULL_ADDR)
 587                         continue;
 588
 589                 dn->data_blkaddr = NULL_ADDR;
 590                 f2fs_set_data_blkaddr(dn);
 591
 592                 if (__is_valid_data_blkaddr(blkaddr)) {
 593                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
 594                                         DATA_GENERIC_ENHANCE))
 595                                 continue;
 596                         if (compressed_cluster)
 597                                 valid_blocks++;
 598                 }
 599
 600                 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 601                         clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 602
 603                 f2fs_invalidate_blocks(sbi, blkaddr);
 604
 605                 if (!released || blkaddr != COMPRESS_ADDR)
 606                         nr_free++;
 607         }
 608
 609         if (compressed_cluster)
 610                 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
 611
 612         if (nr_free) {
 613                 pgoff_t fofs;
 614                 /*
 615                  * once we invalidate valid blkaddr in range [ofs, ofs + count],
 616                  * we will invalidate all blkaddr in the whole range.
 617                  */
 618                 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
 619                                                         dn->inode) + ofs;
 620                 f2fs_update_extent_cache_range(dn, fofs, 0, len);
 621                 dec_valid_block_count(sbi, dn->inode, nr_free);
 622         }
 623         dn->ofs_in_node = ofs;
 624
 625         f2fs_update_time(sbi, REQ_TIME);
 626         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 627                                          dn->ofs_in_node, nr_free);
 628 }
 629
 630 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
 631 {
 632         f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
 633 }
 634
 635 static int truncate_partial_data_page(struct inode *inode, u64 from,
 636                                                                 bool cache_only)
 637 {
 638         loff_t offset = from & (PAGE_SIZE - 1);
 639         pgoff_t index = from >> PAGE_SHIFT;
 640         struct address_space *mapping = inode->i_mapping;
 641         struct page *page;
 642
 643         if (!offset && !cache_only)
 644                 return 0;
 645
 646         if (cache_only) {
 647                 page = find_lock_page(mapping, index);
 648                 if (page && PageUptodate(page))
 649                         goto truncate_out;
 650                 f2fs_put_page(page, 1);
 651                 return 0;
 652         }
 653
 654         page = f2fs_get_lock_data_page(inode, index, true);
 655         if (IS_ERR(page))
 656                 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 657 truncate_out:
 658         f2fs_wait_on_page_writeback(page, DATA, true, true);
 659         zero_user(page, offset, PAGE_SIZE - offset);
 660
 661         /* An encrypted inode should have a key and truncate the last page. */
 662         f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
 663         if (!cache_only)
 664                 set_page_dirty(page);
 665         f2fs_put_page(page, 1);
 666         return 0;
 667 }
 668
 669 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
 670 {
 671         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 672         struct dnode_of_data dn;
 673         pgoff_t free_from;
 674         int count = 0, err = 0;
 675         struct page *ipage;
 676         bool truncate_page = false;
 677
 678         trace_f2fs_truncate_blocks_enter(inode, from);
 679
 680         free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 681
 682         if (free_from >= max_file_blocks(inode))
 683                 goto free_partial;
 684
 685         if (lock)
 686                 f2fs_lock_op(sbi);
 687
 688         ipage = f2fs_get_node_page(sbi, inode->i_ino);
 689         if (IS_ERR(ipage)) {
 690                 err = PTR_ERR(ipage);
 691                 goto out;
 692         }
 693
 694         if (f2fs_has_inline_data(inode)) {
 695                 f2fs_truncate_inline_inode(inode, ipage, from);
 696                 f2fs_put_page(ipage, 1);
 697                 truncate_page = true;
 698                 goto out;
 699         }
 700
 701         set_new_dnode(&dn, inode, ipage, NULL, 0);
 702         err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 703         if (err) {
 704                 if (err == -ENOENT)
 705                         goto free_next;
 706                 goto out;
 707         }
 708
 709         count = ADDRS_PER_PAGE(dn.node_page, inode);
 710
 711         count -= dn.ofs_in_node;
 712         f2fs_bug_on(sbi, count < 0);
 713
 714         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 715                 f2fs_truncate_data_blocks_range(&dn, count);
 716                 free_from += count;
 717         }
 718
 719         f2fs_put_dnode(&dn);
 720 free_next:
 721         err = f2fs_truncate_inode_blocks(inode, free_from);
 722 out:
 723         if (lock)
 724                 f2fs_unlock_op(sbi);
 725 free_partial:
 726         /* lastly zero out the first data page */
 727         if (!err)
 728                 err = truncate_partial_data_page(inode, from, truncate_page);
 729
 730         trace_f2fs_truncate_blocks_exit(inode, err);
 731         return err;
 732 }
 733
 734 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
 735 {
 736         u64 free_from = from;
 737         int err;
 738
 739 #ifdef CONFIG_F2FS_FS_COMPRESSION
 740         /*
 741          * for compressed file, only support cluster size
 742          * aligned truncation.
 743          */
 744         if (f2fs_compressed_file(inode))
 745                 free_from = round_up(from,
 746                                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
 747 #endif
 748
 749         err = f2fs_do_truncate_blocks(inode, free_from, lock);
 750         if (err)
 751                 return err;
 752
 753 #ifdef CONFIG_F2FS_FS_COMPRESSION
 754         /*
 755          * For compressed file, after release compress blocks, don't allow write
 756          * direct, but we should allow write direct after truncate to zero.
 757          */
 758         if (f2fs_compressed_file(inode) && !free_from
 759                         && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
 760                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 761
 762         if (from != free_from) {
 763                 err = f2fs_truncate_partial_cluster(inode, from, lock);
 764                 if (err)
 765                         return err;
 766         }
 767 #endif
 768
 769         return 0;
 770 }
 771
 772 int f2fs_truncate(struct inode *inode)
 773 {
 774         int err;
 775
 776         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 777                 return -EIO;
 778
 779         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 780                                 S_ISLNK(inode->i_mode)))
 781                 return 0;
 782
 783         trace_f2fs_truncate(inode);
 784
 785         if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
 786                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
 787                 return -EIO;
 788         }
 789
 790         err = f2fs_dquot_initialize(inode);
 791         if (err)
 792                 return err;
 793
 794         /* we should check inline_data size */
 795         if (!f2fs_may_inline_data(inode)) {
 796                 err = f2fs_convert_inline_inode(inode);
 797                 if (err)
 798                         return err;
 799         }
 800
 801         err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
 802         if (err)
 803                 return err;
 804
 805         inode->i_mtime = inode->i_ctime = current_time(inode);
 806         f2fs_mark_inode_dirty_sync(inode, false);
 807         return 0;
 808 }
 809
 810 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 811                  struct kstat *stat, u32 request_mask, unsigned int query_flags)
 812 {
 813         struct inode *inode = d_inode(path->dentry);
 814         struct f2fs_inode_info *fi = F2FS_I(inode);
 815         struct f2fs_inode *ri;
 816         unsigned int flags;
 817
 818         if (f2fs_has_extra_attr(inode) &&
 819                         f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
 820                         F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 821                 stat->result_mask |= STATX_BTIME;
 822                 stat->btime.tv_sec = fi->i_crtime.tv_sec;
 823                 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 824         }
 825
 826         flags = fi->i_flags;
 827         if (flags & F2FS_COMPR_FL)
 828                 stat->attributes |= STATX_ATTR_COMPRESSED;
 829         if (flags & F2FS_APPEND_FL)
 830                 stat->attributes |= STATX_ATTR_APPEND;
 831         if (IS_ENCRYPTED(inode))
 832                 stat->attributes |= STATX_ATTR_ENCRYPTED;
 833         if (flags & F2FS_IMMUTABLE_FL)
 834                 stat->attributes |= STATX_ATTR_IMMUTABLE;
 835         if (flags & F2FS_NODUMP_FL)
 836                 stat->attributes |= STATX_ATTR_NODUMP;
 837         if (IS_VERITY(inode))
 838                 stat->attributes |= STATX_ATTR_VERITY;
 839
 840         stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 841                                   STATX_ATTR_APPEND |
 842                                   STATX_ATTR_ENCRYPTED |
 843                                   STATX_ATTR_IMMUTABLE |
 844                                   STATX_ATTR_NODUMP |
 845                                   STATX_ATTR_VERITY);
 846
 847         generic_fillattr(&init_user_ns, inode, stat);
 848
 849         /* we need to show initial sectors used for inline_data/dentries */
 850         if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 851                                         f2fs_has_inline_dentry(inode))
 852                 stat->blocks += (stat->size + 511) >> 9;
 853
 854         return 0;
 855 }
 856
 857 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 858 static void __setattr_copy(struct user_namespace *mnt_userns,
 859                            struct inode *inode, const struct iattr *attr)
 860 {
 861         unsigned int ia_valid = attr->ia_valid;
 862
 863         if (ia_valid & ATTR_UID)
 864                 inode->i_uid = attr->ia_uid;
 865         if (ia_valid & ATTR_GID)
 866                 inode->i_gid = attr->ia_gid;
 867         if (ia_valid & ATTR_ATIME)
 868                 inode->i_atime = attr->ia_atime;
 869         if (ia_valid & ATTR_MTIME)
 870                 inode->i_mtime = attr->ia_mtime;
 871         if (ia_valid & ATTR_CTIME)
 872                 inode->i_ctime = attr->ia_ctime;
 873         if (ia_valid & ATTR_MODE) {
 874                 umode_t mode = attr->ia_mode;
 875                 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
 876
 877                 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
 878                         mode &= ~S_ISGID;
 879                 set_acl_inode(inode, mode);
 880         }
 881 }
 882 #else
 883 #define __setattr_copy setattr_copy
 884 #endif
 885
 886 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 887                  struct iattr *attr)
 888 {
 889         struct inode *inode = d_inode(dentry);
 890         int err;
 891
 892         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 893                 return -EIO;
 894
 895         if (unlikely(IS_IMMUTABLE(inode)))
 896                 return -EPERM;
 897
 898         if (unlikely(IS_APPEND(inode) &&
 899                         (attr->ia_valid & (ATTR_MODE | ATTR_UID |
 900                                   ATTR_GID | ATTR_TIMES_SET))))
 901                 return -EPERM;
 902
 903         if ((attr->ia_valid & ATTR_SIZE) &&
 904                 !f2fs_is_compress_backend_ready(inode))
 905                 return -EOPNOTSUPP;
 906
 907         err = setattr_prepare(&init_user_ns, dentry, attr);
 908         if (err)
 909                 return err;
 910
 911         err = fscrypt_prepare_setattr(dentry, attr);
 912         if (err)
 913                 return err;
 914
 915         err = fsverity_prepare_setattr(dentry, attr);
 916         if (err)
 917                 return err;
 918
 919         if (is_quota_modification(inode, attr)) {
 920                 err = f2fs_dquot_initialize(inode);
 921                 if (err)
 922                         return err;
 923         }
 924         if ((attr->ia_valid & ATTR_UID &&
 925                 !uid_eq(attr->ia_uid, inode->i_uid)) ||
 926                 (attr->ia_valid & ATTR_GID &&
 927                 !gid_eq(attr->ia_gid, inode->i_gid))) {
 928                 f2fs_lock_op(F2FS_I_SB(inode));
 929                 err = dquot_transfer(inode, attr);
 930                 if (err) {
 931                         set_sbi_flag(F2FS_I_SB(inode),
 932                                         SBI_QUOTA_NEED_REPAIR);
 933                         f2fs_unlock_op(F2FS_I_SB(inode));
 934                         return err;
 935                 }
 936                 /*
 937                  * update uid/gid under lock_op(), so that dquot and inode can
 938                  * be updated atomically.
 939                  */
 940                 if (attr->ia_valid & ATTR_UID)
 941                         inode->i_uid = attr->ia_uid;
 942                 if (attr->ia_valid & ATTR_GID)
 943                         inode->i_gid = attr->ia_gid;
 944                 f2fs_mark_inode_dirty_sync(inode, true);
 945                 f2fs_unlock_op(F2FS_I_SB(inode));
 946         }
 947
 948         if (attr->ia_valid & ATTR_SIZE) {
 949                 loff_t old_size = i_size_read(inode);
 950
 951                 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
 952                         /*
 953                          * should convert inline inode before i_size_write to
 954                          * keep smaller than inline_data size with inline flag.
 955                          */
 956                         err = f2fs_convert_inline_inode(inode);
 957                         if (err)
 958                                 return err;
 959                 }
 960
 961                 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 962                 filemap_invalidate_lock(inode->i_mapping);
 963
 964                 truncate_setsize(inode, attr->ia_size);
 965
 966                 if (attr->ia_size <= old_size)
 967                         err = f2fs_truncate(inode);
 968                 /*
 969                  * do not trim all blocks after i_size if target size is
 970                  * larger than i_size.
 971                  */
 972                 filemap_invalidate_unlock(inode->i_mapping);
 973                 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 974                 if (err)
 975                         return err;
 976
 977                 spin_lock(&F2FS_I(inode)->i_size_lock);
 978                 inode->i_mtime = inode->i_ctime = current_time(inode);
 979                 F2FS_I(inode)->last_disk_size = i_size_read(inode);
 980                 spin_unlock(&F2FS_I(inode)->i_size_lock);
 981         }
 982
 983         __setattr_copy(&init_user_ns, inode, attr);
 984
 985         if (attr->ia_valid & ATTR_MODE) {
 986                 err = posix_acl_chmod(&init_user_ns, inode, f2fs_get_inode_mode(inode));
 987
 988                 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
 989                         if (!err)
 990                                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
 991                         clear_inode_flag(inode, FI_ACL_MODE);
 992                 }
 993         }
 994
 995         /* file size may changed here */
 996         f2fs_mark_inode_dirty_sync(inode, true);
 997
 998         /* inode change will produce dirty node pages flushed by checkpoint */
 999         f2fs_balance_fs(F2FS_I_SB(inode), true);
1000
1001         return err;
1002 }
1003
1004 const struct inode_operations f2fs_file_inode_operations = {
1005         .getattr        = f2fs_getattr,
1006         .setattr        = f2fs_setattr,
1007         .get_acl        = f2fs_get_acl,
1008         .set_acl        = f2fs_set_acl,
1009         .listxattr      = f2fs_listxattr,
1010         .fiemap         = f2fs_fiemap,
1011         .fileattr_get   = f2fs_fileattr_get,
1012         .fileattr_set   = f2fs_fileattr_set,
1013 };
1014
1015 static int fill_zero(struct inode *inode, pgoff_t index,
1016                                         loff_t start, loff_t len)
1017 {
1018         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1019         struct page *page;
1020
1021         if (!len)
1022                 return 0;
1023
1024         f2fs_balance_fs(sbi, true);
1025
1026         f2fs_lock_op(sbi);
1027         page = f2fs_get_new_data_page(inode, NULL, index, false);
1028         f2fs_unlock_op(sbi);
1029
1030         if (IS_ERR(page))
1031                 return PTR_ERR(page);
1032
1033         f2fs_wait_on_page_writeback(page, DATA, true, true);
1034         zero_user(page, start, len);
1035         set_page_dirty(page);
1036         f2fs_put_page(page, 1);
1037         return 0;
1038 }
1039
1040 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1041 {
1042         int err;
1043
1044         while (pg_start < pg_end) {
1045                 struct dnode_of_data dn;
1046                 pgoff_t end_offset, count;
1047
1048                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1049                 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1050                 if (err) {
1051                         if (err == -ENOENT) {
1052                                 pg_start = f2fs_get_next_page_offset(&dn,
1053                                                                 pg_start);
1054                                 continue;
1055                         }
1056                         return err;
1057                 }
1058
1059                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1060                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1061
1062                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1063
1064                 f2fs_truncate_data_blocks_range(&dn, count);
1065                 f2fs_put_dnode(&dn);
1066
1067                 pg_start += count;
1068         }
1069         return 0;
1070 }
1071
1072 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1073 {
1074         pgoff_t pg_start, pg_end;
1075         loff_t off_start, off_end;
1076         int ret;
1077
1078         ret = f2fs_convert_inline_inode(inode);
1079         if (ret)
1080                 return ret;
1081
1082         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1083         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1084
1085         off_start = offset & (PAGE_SIZE - 1);
1086         off_end = (offset + len) & (PAGE_SIZE - 1);
1087
1088         if (pg_start == pg_end) {
1089                 ret = fill_zero(inode, pg_start, off_start,
1090                                                 off_end - off_start);
1091                 if (ret)
1092                         return ret;
1093         } else {
1094                 if (off_start) {
1095                         ret = fill_zero(inode, pg_start++, off_start,
1096                                                 PAGE_SIZE - off_start);
1097                         if (ret)
1098                                 return ret;
1099                 }
1100                 if (off_end) {
1101                         ret = fill_zero(inode, pg_end, 0, off_end);
1102                         if (ret)
1103                                 return ret;
1104                 }
1105
1106                 if (pg_start < pg_end) {
1107                         loff_t blk_start, blk_end;
1108                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1109
1110                         f2fs_balance_fs(sbi, true);
1111
1112                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
1113                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
1114
1115                         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1116                         filemap_invalidate_lock(inode->i_mapping);
1117
1118                         truncate_pagecache_range(inode, blk_start, blk_end - 1);
1119
1120                         f2fs_lock_op(sbi);
1121                         ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1122                         f2fs_unlock_op(sbi);
1123
1124                         filemap_invalidate_unlock(inode->i_mapping);
1125                         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1126                 }
1127         }
1128
1129         return ret;
1130 }
1131
1132 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1133                                 int *do_replace, pgoff_t off, pgoff_t len)
1134 {
1135         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1136         struct dnode_of_data dn;
1137         int ret, done, i;
1138
1139 next_dnode:
1140         set_new_dnode(&dn, inode, NULL, NULL, 0);
1141         ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1142         if (ret && ret != -ENOENT) {
1143                 return ret;
1144         } else if (ret == -ENOENT) {
1145                 if (dn.max_level == 0)
1146                         return -ENOENT;
1147                 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1148                                                 dn.ofs_in_node, len);
1149                 blkaddr += done;
1150                 do_replace += done;
1151                 goto next;
1152         }
1153
1154         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1155                                                         dn.ofs_in_node, len);
1156         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1157                 *blkaddr = f2fs_data_blkaddr(&dn);
1158
1159                 if (__is_valid_data_blkaddr(*blkaddr) &&
1160                         !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1161                                         DATA_GENERIC_ENHANCE)) {
1162                         f2fs_put_dnode(&dn);
1163                         return -EFSCORRUPTED;
1164                 }
1165
1166                 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1167
1168                         if (f2fs_lfs_mode(sbi)) {
1169                                 f2fs_put_dnode(&dn);
1170                                 return -EOPNOTSUPP;
1171                         }
1172
1173                         /* do not invalidate this block address */
1174                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1175                         *do_replace = 1;
1176                 }
1177         }
1178         f2fs_put_dnode(&dn);
1179 next:
1180         len -= done;
1181         off += done;
1182         if (len)
1183                 goto next_dnode;
1184         return 0;
1185 }
1186
1187 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1188                                 int *do_replace, pgoff_t off, int len)
1189 {
1190         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1191         struct dnode_of_data dn;
1192         int ret, i;
1193
1194         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1195                 if (*do_replace == 0)
1196                         continue;
1197
1198                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1199                 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1200                 if (ret) {
1201                         dec_valid_block_count(sbi, inode, 1);
1202                         f2fs_invalidate_blocks(sbi, *blkaddr);
1203                 } else {
1204                         f2fs_update_data_blkaddr(&dn, *blkaddr);
1205                 }
1206                 f2fs_put_dnode(&dn);
1207         }
1208         return 0;
1209 }
1210
1211 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1212                         block_t *blkaddr, int *do_replace,
1213                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1214 {
1215         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1216         pgoff_t i = 0;
1217         int ret;
1218
1219         while (i < len) {
1220                 if (blkaddr[i] == NULL_ADDR && !full) {
1221                         i++;
1222                         continue;
1223                 }
1224
1225                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1226                         struct dnode_of_data dn;
1227                         struct node_info ni;
1228                         size_t new_size;
1229                         pgoff_t ilen;
1230
1231                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1232                         ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1233                         if (ret)
1234                                 return ret;
1235
1236                         ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1237                         if (ret) {
1238                                 f2fs_put_dnode(&dn);
1239                                 return ret;
1240                         }
1241
1242                         ilen = min((pgoff_t)
1243                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1244                                                 dn.ofs_in_node, len - i);
1245                         do {
1246                                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1247                                 f2fs_truncate_data_blocks_range(&dn, 1);
1248
1249                                 if (do_replace[i]) {
1250                                         f2fs_i_blocks_write(src_inode,
1251                                                         1, false, false);
1252                                         f2fs_i_blocks_write(dst_inode,
1253                                                         1, true, false);
1254                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1255                                         blkaddr[i], ni.version, true, false);
1256
1257                                         do_replace[i] = 0;
1258                                 }
1259                                 dn.ofs_in_node++;
1260                                 i++;
1261                                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1262                                 if (dst_inode->i_size < new_size)
1263                                         f2fs_i_size_write(dst_inode, new_size);
1264                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1265
1266                         f2fs_put_dnode(&dn);
1267                 } else {
1268                         struct page *psrc, *pdst;
1269
1270                         psrc = f2fs_get_lock_data_page(src_inode,
1271                                                         src + i, true);
1272                         if (IS_ERR(psrc))
1273                                 return PTR_ERR(psrc);
1274                         pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1275                                                                 true);
1276                         if (IS_ERR(pdst)) {
1277                                 f2fs_put_page(psrc, 1);
1278                                 return PTR_ERR(pdst);
1279                         }
1280                         f2fs_copy_page(psrc, pdst);
1281                         set_page_dirty(pdst);
1282                         f2fs_put_page(pdst, 1);
1283                         f2fs_put_page(psrc, 1);
1284
1285                         ret = f2fs_truncate_hole(src_inode,
1286                                                 src + i, src + i + 1);
1287                         if (ret)
1288                                 return ret;
1289                         i++;
1290                 }
1291         }
1292         return 0;
1293 }
1294
1295 static int __exchange_data_block(struct inode *src_inode,
1296                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1297                         pgoff_t len, bool full)
1298 {
1299         block_t *src_blkaddr;
1300         int *do_replace;
1301         pgoff_t olen;
1302         int ret;
1303
1304         while (len) {
1305                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1306
1307                 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1308                                         array_size(olen, sizeof(block_t)),
1309                                         GFP_NOFS);
1310                 if (!src_blkaddr)
1311                         return -ENOMEM;
1312
1313                 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1314                                         array_size(olen, sizeof(int)),
1315                                         GFP_NOFS);
1316                 if (!do_replace) {
1317                         kvfree(src_blkaddr);
1318                         return -ENOMEM;
1319                 }
1320
1321                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1322                                         do_replace, src, olen);
1323                 if (ret)
1324                         goto roll_back;
1325
1326                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1327                                         do_replace, src, dst, olen, full);
1328                 if (ret)
1329                         goto roll_back;
1330
1331                 src += olen;
1332                 dst += olen;
1333                 len -= olen;
1334
1335                 kvfree(src_blkaddr);
1336                 kvfree(do_replace);
1337         }
1338         return 0;
1339
1340 roll_back:
1341         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1342         kvfree(src_blkaddr);
1343         kvfree(do_replace);
1344         return ret;
1345 }
1346
1347 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1348 {
1349         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1350         pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1351         pgoff_t start = offset >> PAGE_SHIFT;
1352         pgoff_t end = (offset + len) >> PAGE_SHIFT;
1353         int ret;
1354
1355         f2fs_balance_fs(sbi, true);
1356
1357         /* avoid gc operation during block exchange */
1358         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1359         filemap_invalidate_lock(inode->i_mapping);
1360
1361         f2fs_lock_op(sbi);
1362         f2fs_drop_extent_tree(inode);
1363         truncate_pagecache(inode, offset);
1364         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1365         f2fs_unlock_op(sbi);
1366
1367         filemap_invalidate_unlock(inode->i_mapping);
1368         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1369         return ret;
1370 }
1371
1372 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1373 {
1374         loff_t new_size;
1375         int ret;
1376
1377         if (offset + len >= i_size_read(inode))
1378                 return -EINVAL;
1379
1380         /* collapse range should be aligned to block size of f2fs. */
1381         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1382                 return -EINVAL;
1383
1384         ret = f2fs_convert_inline_inode(inode);
1385         if (ret)
1386                 return ret;
1387
1388         /* write out all dirty pages from offset */
1389         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1390         if (ret)
1391                 return ret;
1392
1393         ret = f2fs_do_collapse(inode, offset, len);
1394         if (ret)
1395                 return ret;
1396
1397         /* write out all moved pages, if possible */
1398         filemap_invalidate_lock(inode->i_mapping);
1399         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1400         truncate_pagecache(inode, offset);
1401
1402         new_size = i_size_read(inode) - len;
1403         ret = f2fs_truncate_blocks(inode, new_size, true);
1404         filemap_invalidate_unlock(inode->i_mapping);
1405         if (!ret)
1406                 f2fs_i_size_write(inode, new_size);
1407         return ret;
1408 }
1409
1410 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1411                                                                 pgoff_t end)
1412 {
1413         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1414         pgoff_t index = start;
1415         unsigned int ofs_in_node = dn->ofs_in_node;
1416         blkcnt_t count = 0;
1417         int ret;
1418
1419         for (; index < end; index++, dn->ofs_in_node++) {
1420                 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1421                         count++;
1422         }
1423
1424         dn->ofs_in_node = ofs_in_node;
1425         ret = f2fs_reserve_new_blocks(dn, count);
1426         if (ret)
1427                 return ret;
1428
1429         dn->ofs_in_node = ofs_in_node;
1430         for (index = start; index < end; index++, dn->ofs_in_node++) {
1431                 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1432                 /*
1433                  * f2fs_reserve_new_blocks will not guarantee entire block
1434                  * allocation.
1435                  */
1436                 if (dn->data_blkaddr == NULL_ADDR) {
1437                         ret = -ENOSPC;
1438                         break;
1439                 }
1440                 if (dn->data_blkaddr != NEW_ADDR) {
1441                         f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1442                         dn->data_blkaddr = NEW_ADDR;
1443                         f2fs_set_data_blkaddr(dn);
1444                 }
1445         }
1446
1447         f2fs_update_extent_cache_range(dn, start, 0, index - start);
1448
1449         return ret;
1450 }
1451
1452 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1453                                                                 int mode)
1454 {
1455         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1456         struct address_space *mapping = inode->i_mapping;
1457         pgoff_t index, pg_start, pg_end;
1458         loff_t new_size = i_size_read(inode);
1459         loff_t off_start, off_end;
1460         int ret = 0;
1461
1462         ret = inode_newsize_ok(inode, (len + offset));
1463         if (ret)
1464                 return ret;
1465
1466         ret = f2fs_convert_inline_inode(inode);
1467         if (ret)
1468                 return ret;
1469
1470         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1471         if (ret)
1472                 return ret;
1473
1474         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1475         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1476
1477         off_start = offset & (PAGE_SIZE - 1);
1478         off_end = (offset + len) & (PAGE_SIZE - 1);
1479
1480         if (pg_start == pg_end) {
1481                 ret = fill_zero(inode, pg_start, off_start,
1482                                                 off_end - off_start);
1483                 if (ret)
1484                         return ret;
1485
1486                 new_size = max_t(loff_t, new_size, offset + len);
1487         } else {
1488                 if (off_start) {
1489                         ret = fill_zero(inode, pg_start++, off_start,
1490                                                 PAGE_SIZE - off_start);
1491                         if (ret)
1492                                 return ret;
1493
1494                         new_size = max_t(loff_t, new_size,
1495                                         (loff_t)pg_start << PAGE_SHIFT);
1496                 }
1497
1498                 for (index = pg_start; index < pg_end;) {
1499                         struct dnode_of_data dn;
1500                         unsigned int end_offset;
1501                         pgoff_t end;
1502
1503                         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1504                         filemap_invalidate_lock(mapping);
1505
1506                         truncate_pagecache_range(inode,
1507                                 (loff_t)index << PAGE_SHIFT,
1508                                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1509
1510                         f2fs_lock_op(sbi);
1511
1512                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1513                         ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1514                         if (ret) {
1515                                 f2fs_unlock_op(sbi);
1516                                 filemap_invalidate_unlock(mapping);
1517                                 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1518                                 goto out;
1519                         }
1520
1521                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1522                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1523
1524                         ret = f2fs_do_zero_range(&dn, index, end);
1525                         f2fs_put_dnode(&dn);
1526
1527                         f2fs_unlock_op(sbi);
1528                         filemap_invalidate_unlock(mapping);
1529                         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1530
1531                         f2fs_balance_fs(sbi, dn.node_changed);
1532
1533                         if (ret)
1534                                 goto out;
1535
1536                         index = end;
1537                         new_size = max_t(loff_t, new_size,
1538                                         (loff_t)index << PAGE_SHIFT);
1539                 }
1540
1541                 if (off_end) {
1542                         ret = fill_zero(inode, pg_end, 0, off_end);
1543                         if (ret)
1544                                 goto out;
1545
1546                         new_size = max_t(loff_t, new_size, offset + len);
1547                 }
1548         }
1549
1550 out:
1551         if (new_size > i_size_read(inode)) {
1552                 if (mode & FALLOC_FL_KEEP_SIZE)
1553                         file_set_keep_isize(inode);
1554                 else
1555                         f2fs_i_size_write(inode, new_size);
1556         }
1557         return ret;
1558 }
1559
1560 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1561 {
1562         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1563         struct address_space *mapping = inode->i_mapping;
1564         pgoff_t nr, pg_start, pg_end, delta, idx;
1565         loff_t new_size;
1566         int ret = 0;
1567
1568         new_size = i_size_read(inode) + len;
1569         ret = inode_newsize_ok(inode, new_size);
1570         if (ret)
1571                 return ret;
1572
1573         if (offset >= i_size_read(inode))
1574                 return -EINVAL;
1575
1576         /* insert range should be aligned to block size of f2fs. */
1577         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1578                 return -EINVAL;
1579
1580         ret = f2fs_convert_inline_inode(inode);
1581         if (ret)
1582                 return ret;
1583
1584         f2fs_balance_fs(sbi, true);
1585
1586         filemap_invalidate_lock(mapping);
1587         ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1588         filemap_invalidate_unlock(mapping);
1589         if (ret)
1590                 return ret;
1591
1592         /* write out all dirty pages from offset */
1593         ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1594         if (ret)
1595                 return ret;
1596
1597         pg_start = offset >> PAGE_SHIFT;
1598         pg_end = (offset + len) >> PAGE_SHIFT;
1599         delta = pg_end - pg_start;
1600         idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1601
1602         /* avoid gc operation during block exchange */
1603         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1604         filemap_invalidate_lock(mapping);
1605         truncate_pagecache(inode, offset);
1606
1607         while (!ret && idx > pg_start) {
1608                 nr = idx - pg_start;
1609                 if (nr > delta)
1610                         nr = delta;
1611                 idx -= nr;
1612
1613                 f2fs_lock_op(sbi);
1614                 f2fs_drop_extent_tree(inode);
1615
1616                 ret = __exchange_data_block(inode, inode, idx,
1617                                         idx + delta, nr, false);
1618                 f2fs_unlock_op(sbi);
1619         }
1620         filemap_invalidate_unlock(mapping);
1621         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1622
1623         /* write out all moved pages, if possible */
1624         filemap_invalidate_lock(mapping);
1625         filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1626         truncate_pagecache(inode, offset);
1627         filemap_invalidate_unlock(mapping);
1628
1629         if (!ret)
1630                 f2fs_i_size_write(inode, new_size);
1631         return ret;
1632 }
1633
1634 static int expand_inode_data(struct inode *inode, loff_t offset,
1635                                         loff_t len, int mode)
1636 {
1637         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1638         struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1639                         .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1640                         .m_may_create = true };
1641         pgoff_t pg_start, pg_end;
1642         loff_t new_size = i_size_read(inode);
1643         loff_t off_end;
1644         block_t expanded = 0;
1645         int err;
1646
1647         err = inode_newsize_ok(inode, (len + offset));
1648         if (err)
1649                 return err;
1650
1651         err = f2fs_convert_inline_inode(inode);
1652         if (err)
1653                 return err;
1654
1655         f2fs_balance_fs(sbi, true);
1656
1657         pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1658         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1659         off_end = (offset + len) & (PAGE_SIZE - 1);
1660
1661         map.m_lblk = pg_start;
1662         map.m_len = pg_end - pg_start;
1663         if (off_end)
1664                 map.m_len++;
1665
1666         if (!map.m_len)
1667                 return 0;
1668
1669         if (f2fs_is_pinned_file(inode)) {
1670                 block_t sec_blks = BLKS_PER_SEC(sbi);
1671                 block_t sec_len = roundup(map.m_len, sec_blks);
1672
1673                 map.m_len = sec_blks;
1674 next_alloc:
1675                 if (has_not_enough_free_secs(sbi, 0,
1676                         GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1677                         down_write(&sbi->gc_lock);
1678                         err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1679                         if (err && err != -ENODATA && err != -EAGAIN)
1680                                 goto out_err;
1681                 }
1682
1683                 down_write(&sbi->pin_sem);
1684
1685                 f2fs_lock_op(sbi);
1686                 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1687                 f2fs_unlock_op(sbi);
1688
1689                 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1690                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1691                 file_dont_truncate(inode);
1692
1693                 up_write(&sbi->pin_sem);
1694
1695                 expanded += map.m_len;
1696                 sec_len -= map.m_len;
1697                 map.m_lblk += map.m_len;
1698                 if (!err && sec_len)
1699                         goto next_alloc;
1700
1701                 map.m_len = expanded;
1702         } else {
1703                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1704                 expanded = map.m_len;
1705         }
1706 out_err:
1707         if (err) {
1708                 pgoff_t last_off;
1709
1710                 if (!expanded)
1711                         return err;
1712
1713                 last_off = pg_start + expanded - 1;
1714
1715                 /* update new size to the failed position */
1716                 new_size = (last_off == pg_end) ? offset + len :
1717                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1718         } else {
1719                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1720         }
1721
1722         if (new_size > i_size_read(inode)) {
1723                 if (mode & FALLOC_FL_KEEP_SIZE)
1724                         file_set_keep_isize(inode);
1725                 else
1726                         f2fs_i_size_write(inode, new_size);
1727         }
1728
1729         return err;
1730 }
1731
1732 static long f2fs_fallocate(struct file *file, int mode,
1733                                 loff_t offset, loff_t len)
1734 {
1735         struct inode *inode = file_inode(file);
1736         long ret = 0;
1737
1738         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1739                 return -EIO;
1740         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1741                 return -ENOSPC;
1742         if (!f2fs_is_compress_backend_ready(inode))
1743                 return -EOPNOTSUPP;
1744
1745         /* f2fs only support ->fallocate for regular file */
1746         if (!S_ISREG(inode->i_mode))
1747                 return -EINVAL;
1748
1749         if (IS_ENCRYPTED(inode) &&
1750                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1751                 return -EOPNOTSUPP;
1752
1753         /*
1754          * Pinned file should not support partial trucation since the block
1755          * can be used by applications.
1756          */
1757         if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1758                 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1759                         FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1760                 return -EOPNOTSUPP;
1761
1762         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1763                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1764                         FALLOC_FL_INSERT_RANGE))
1765                 return -EOPNOTSUPP;
1766
1767         inode_lock(inode);
1768
1769         if (mode & FALLOC_FL_PUNCH_HOLE) {
1770                 if (offset >= inode->i_size)
1771                         goto out;
1772
1773                 ret = punch_hole(inode, offset, len);
1774         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1775                 ret = f2fs_collapse_range(inode, offset, len);
1776         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1777                 ret = f2fs_zero_range(inode, offset, len, mode);
1778         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1779                 ret = f2fs_insert_range(inode, offset, len);
1780         } else {
1781                 ret = expand_inode_data(inode, offset, len, mode);
1782         }
1783
1784         if (!ret) {
1785                 inode->i_mtime = inode->i_ctime = current_time(inode);
1786                 f2fs_mark_inode_dirty_sync(inode, false);
1787                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1788         }
1789
1790 out:
1791         inode_unlock(inode);
1792
1793         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1794         return ret;
1795 }
1796
1797 static int f2fs_release_file(struct inode *inode, struct file *filp)
1798 {
1799         /*
1800          * f2fs_relase_file is called at every close calls. So we should
1801          * not drop any inmemory pages by close called by other process.
1802          */
1803         if (!(filp->f_mode & FMODE_WRITE) ||
1804                         atomic_read(&inode->i_writecount) != 1)
1805                 return 0;
1806
1807         /* some remained atomic pages should discarded */
1808         if (f2fs_is_atomic_file(inode))
1809                 f2fs_drop_inmem_pages(inode);
1810         if (f2fs_is_volatile_file(inode)) {
1811                 set_inode_flag(inode, FI_DROP_CACHE);
1812                 filemap_fdatawrite(inode->i_mapping);
1813                 clear_inode_flag(inode, FI_DROP_CACHE);
1814                 clear_inode_flag(inode, FI_VOLATILE_FILE);
1815                 stat_dec_volatile_write(inode);
1816         }
1817         return 0;
1818 }
1819
1820 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1821 {
1822         struct inode *inode = file_inode(file);
1823
1824         /*
1825          * If the process doing a transaction is crashed, we should do
1826          * roll-back. Otherwise, other reader/write can see corrupted database
1827          * until all the writers close its file. Since this should be done
1828          * before dropping file lock, it needs to do in ->flush.
1829          */
1830         if (f2fs_is_atomic_file(inode) &&
1831                         F2FS_I(inode)->inmem_task == current)
1832                 f2fs_drop_inmem_pages(inode);
1833         return 0;
1834 }
1835
1836 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1837 {
1838         struct f2fs_inode_info *fi = F2FS_I(inode);
1839         u32 masked_flags = fi->i_flags & mask;
1840
1841         /* mask can be shrunk by flags_valid selector */
1842         iflags &= mask;
1843
1844         /* Is it quota file? Do not allow user to mess with it */
1845         if (IS_NOQUOTA(inode))
1846                 return -EPERM;
1847
1848         if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1849                 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1850                         return -EOPNOTSUPP;
1851                 if (!f2fs_empty_dir(inode))
1852                         return -ENOTEMPTY;
1853         }
1854
1855         if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1856                 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1857                         return -EOPNOTSUPP;
1858                 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1859                         return -EINVAL;
1860         }
1861
1862         if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1863                 if (masked_flags & F2FS_COMPR_FL) {
1864                         if (!f2fs_disable_compressed_file(inode))
1865                                 return -EINVAL;
1866                 }
1867                 if (iflags & F2FS_NOCOMP_FL)
1868                         return -EINVAL;
1869                 if (iflags & F2FS_COMPR_FL) {
1870                         if (!f2fs_may_compress(inode))
1871                                 return -EINVAL;
1872                         if (S_ISREG(inode->i_mode) && inode->i_size)
1873                                 return -EINVAL;
1874
1875                         set_compress_context(inode);
1876                 }
1877         }
1878         if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1879                 if (masked_flags & F2FS_COMPR_FL)
1880                         return -EINVAL;
1881         }
1882
1883         fi->i_flags = iflags | (fi->i_flags & ~mask);
1884         f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1885                                         (fi->i_flags & F2FS_NOCOMP_FL));
1886
1887         if (fi->i_flags & F2FS_PROJINHERIT_FL)
1888                 set_inode_flag(inode, FI_PROJ_INHERIT);
1889         else
1890                 clear_inode_flag(inode, FI_PROJ_INHERIT);
1891
1892         inode->i_ctime = current_time(inode);
1893         f2fs_set_inode_flags(inode);
1894         f2fs_mark_inode_dirty_sync(inode, true);
1895         return 0;
1896 }
1897
1898 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1899
1900 /*
1901  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1902  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1903  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1904  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1905  *
1906  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1907  * FS_IOC_FSSETXATTR is done by the VFS.
1908  */
1909
1910 static const struct {
1911         u32 iflag;
1912         u32 fsflag;
1913 } f2fs_fsflags_map[] = {
1914         { F2FS_COMPR_FL,        FS_COMPR_FL },
1915         { F2FS_SYNC_FL,         FS_SYNC_FL },
1916         { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1917         { F2FS_APPEND_FL,       FS_APPEND_FL },
1918         { F2FS_NODUMP_FL,       FS_NODUMP_FL },
1919         { F2FS_NOATIME_FL,      FS_NOATIME_FL },
1920         { F2FS_NOCOMP_FL,       FS_NOCOMP_FL },
1921         { F2FS_INDEX_FL,        FS_INDEX_FL },
1922         { F2FS_DIRSYNC_FL,      FS_DIRSYNC_FL },
1923         { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1924         { F2FS_CASEFOLD_FL,     FS_CASEFOLD_FL },
1925 };
1926
1927 #define F2FS_GETTABLE_FS_FL (           \
1928                 FS_COMPR_FL |           \
1929                 FS_SYNC_FL |            \
1930                 FS_IMMUTABLE_FL |       \
1931                 FS_APPEND_FL |          \
1932                 FS_NODUMP_FL |          \
1933                 FS_NOATIME_FL |         \
1934                 FS_NOCOMP_FL |          \
1935                 FS_INDEX_FL |           \
1936                 FS_DIRSYNC_FL |         \
1937                 FS_PROJINHERIT_FL |     \
1938                 FS_ENCRYPT_FL |         \
1939                 FS_INLINE_DATA_FL |     \
1940                 FS_NOCOW_FL |           \
1941                 FS_VERITY_FL |          \
1942                 FS_CASEFOLD_FL)
1943
1944 #define F2FS_SETTABLE_FS_FL (           \
1945                 FS_COMPR_FL |           \
1946                 FS_SYNC_FL |            \
1947                 FS_IMMUTABLE_FL |       \
1948                 FS_APPEND_FL |          \
1949                 FS_NODUMP_FL |          \
1950                 FS_NOATIME_FL |         \
1951                 FS_NOCOMP_FL |          \
1952                 FS_DIRSYNC_FL |         \
1953                 FS_PROJINHERIT_FL |     \
1954                 FS_CASEFOLD_FL)
1955
1956 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1957 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1958 {
1959         u32 fsflags = 0;
1960         int i;
1961
1962         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1963                 if (iflags & f2fs_fsflags_map[i].iflag)
1964                         fsflags |= f2fs_fsflags_map[i].fsflag;
1965
1966         return fsflags;
1967 }
1968
1969 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1970 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1971 {
1972         u32 iflags = 0;
1973         int i;
1974
1975         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1976                 if (fsflags & f2fs_fsflags_map[i].fsflag)
1977                         iflags |= f2fs_fsflags_map[i].iflag;
1978
1979         return iflags;
1980 }
1981
1982 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1983 {
1984         struct inode *inode = file_inode(filp);
1985
1986         return put_user(inode->i_generation, (int __user *)arg);
1987 }
1988
1989 static int f2fs_ioc_start_atomic_write(struct file *filp)
1990 {
1991         struct inode *inode = file_inode(filp);
1992         struct f2fs_inode_info *fi = F2FS_I(inode);
1993         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1994         int ret;
1995
1996         if (!inode_owner_or_capable(&init_user_ns, inode))
1997                 return -EACCES;
1998
1999         if (!S_ISREG(inode->i_mode))
2000                 return -EINVAL;
2001
2002         if (filp->f_flags & O_DIRECT)
2003                 return -EINVAL;
2004
2005         ret = mnt_want_write_file(filp);
2006         if (ret)
2007                 return ret;
2008
2009         inode_lock(inode);
2010
2011         f2fs_disable_compressed_file(inode);
2012
2013         if (f2fs_is_atomic_file(inode)) {
2014                 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2015                         ret = -EINVAL;
2016                 goto out;
2017         }
2018
2019         ret = f2fs_convert_inline_inode(inode);
2020         if (ret)
2021                 goto out;
2022
2023         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2024
2025         /*
2026          * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2027          * f2fs_is_atomic_file.
2028          */
2029         if (get_dirty_pages(inode))
2030                 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2031                           inode->i_ino, get_dirty_pages(inode));
2032         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2033         if (ret) {
2034                 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2035                 goto out;
2036         }
2037
2038         spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2039         if (list_empty(&fi->inmem_ilist))
2040                 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2041         sbi->atomic_files++;
2042         spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2043
2044         /* add inode in inmem_list first and set atomic_file */
2045         set_inode_flag(inode, FI_ATOMIC_FILE);
2046         clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2047         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2048
2049         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2050         F2FS_I(inode)->inmem_task = current;
2051         stat_update_max_atomic_write(inode);
2052 out:
2053         inode_unlock(inode);
2054         mnt_drop_write_file(filp);
2055         return ret;
2056 }
2057
2058 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2059 {
2060         struct inode *inode = file_inode(filp);
2061         int ret;
2062
2063         if (!inode_owner_or_capable(&init_user_ns, inode))
2064                 return -EACCES;
2065
2066         ret = mnt_want_write_file(filp);
2067         if (ret)
2068                 return ret;
2069
2070         f2fs_balance_fs(F2FS_I_SB(inode), true);
2071
2072         inode_lock(inode);
2073
2074         if (f2fs_is_volatile_file(inode)) {
2075                 ret = -EINVAL;
2076                 goto err_out;
2077         }
2078
2079         if (f2fs_is_atomic_file(inode)) {
2080                 ret = f2fs_commit_inmem_pages(inode);
2081                 if (ret)
2082                         goto err_out;
2083
2084                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2085                 if (!ret)
2086                         f2fs_drop_inmem_pages(inode);
2087         } else {
2088                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2089         }
2090 err_out:
2091         if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2092                 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2093                 ret = -EINVAL;
2094         }
2095         inode_unlock(inode);
2096         mnt_drop_write_file(filp);
2097         return ret;
2098 }
2099
2100 static int f2fs_ioc_start_volatile_write(struct file *filp)
2101 {
2102         struct inode *inode = file_inode(filp);
2103         int ret;
2104
2105         if (!inode_owner_or_capable(&init_user_ns, inode))
2106                 return -EACCES;
2107
2108         if (!S_ISREG(inode->i_mode))
2109                 return -EINVAL;
2110
2111         ret = mnt_want_write_file(filp);
2112         if (ret)
2113                 return ret;
2114
2115         inode_lock(inode);
2116
2117         if (f2fs_is_volatile_file(inode))
2118                 goto out;
2119
2120         ret = f2fs_convert_inline_inode(inode);
2121         if (ret)
2122                 goto out;
2123
2124         stat_inc_volatile_write(inode);
2125         stat_update_max_volatile_write(inode);
2126
2127         set_inode_flag(inode, FI_VOLATILE_FILE);
2128         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2129 out:
2130         inode_unlock(inode);
2131         mnt_drop_write_file(filp);
2132         return ret;
2133 }
2134
2135 static int f2fs_ioc_release_volatile_write(struct file *filp)
2136 {
2137         struct inode *inode = file_inode(filp);
2138         int ret;
2139
2140         if (!inode_owner_or_capable(&init_user_ns, inode))
2141                 return -EACCES;
2142
2143         ret = mnt_want_write_file(filp);
2144         if (ret)
2145                 return ret;
2146
2147         inode_lock(inode);
2148
2149         if (!f2fs_is_volatile_file(inode))
2150                 goto out;
2151
2152         if (!f2fs_is_first_block_written(inode)) {
2153                 ret = truncate_partial_data_page(inode, 0, true);
2154                 goto out;
2155         }
2156
2157         ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2158 out:
2159         inode_unlock(inode);
2160         mnt_drop_write_file(filp);
2161         return ret;
2162 }
2163
2164 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2165 {
2166         struct inode *inode = file_inode(filp);
2167         int ret;
2168
2169         if (!inode_owner_or_capable(&init_user_ns, inode))
2170                 return -EACCES;
2171
2172         ret = mnt_want_write_file(filp);
2173         if (ret)
2174                 return ret;
2175
2176         inode_lock(inode);
2177
2178         if (f2fs_is_atomic_file(inode))
2179                 f2fs_drop_inmem_pages(inode);
2180         if (f2fs_is_volatile_file(inode)) {
2181                 clear_inode_flag(inode, FI_VOLATILE_FILE);
2182                 stat_dec_volatile_write(inode);
2183                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2184         }
2185
2186         clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2187
2188         inode_unlock(inode);
2189
2190         mnt_drop_write_file(filp);
2191         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2192         return ret;
2193 }
2194
2195 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2196 {
2197         struct inode *inode = file_inode(filp);
2198         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2199         struct super_block *sb = sbi->sb;
2200         __u32 in;
2201         int ret = 0;
2202
2203         if (!capable(CAP_SYS_ADMIN))
2204                 return -EPERM;
2205
2206         if (get_user(in, (__u32 __user *)arg))
2207                 return -EFAULT;
2208
2209         if (in != F2FS_GOING_DOWN_FULLSYNC) {
2210                 ret = mnt_want_write_file(filp);
2211                 if (ret) {
2212                         if (ret == -EROFS) {
2213                                 ret = 0;
2214                                 f2fs_stop_checkpoint(sbi, false);
2215                                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2216                                 trace_f2fs_shutdown(sbi, in, ret);
2217                         }
2218                         return ret;
2219                 }
2220         }
2221
2222         switch (in) {
2223         case F2FS_GOING_DOWN_FULLSYNC:
2224                 ret = freeze_bdev(sb->s_bdev);
2225                 if (ret)
2226                         goto out;
2227                 f2fs_stop_checkpoint(sbi, false);
2228                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2229                 thaw_bdev(sb->s_bdev);
2230                 break;
2231         case F2FS_GOING_DOWN_METASYNC:
2232                 /* do checkpoint only */
2233                 ret = f2fs_sync_fs(sb, 1);
2234                 if (ret)
2235                         goto out;
2236                 f2fs_stop_checkpoint(sbi, false);
2237                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2238                 break;
2239         case F2FS_GOING_DOWN_NOSYNC:
2240                 f2fs_stop_checkpoint(sbi, false);
2241                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2242                 break;
2243         case F2FS_GOING_DOWN_METAFLUSH:
2244                 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2245                 f2fs_stop_checkpoint(sbi, false);
2246                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2247                 break;
2248         case F2FS_GOING_DOWN_NEED_FSCK:
2249                 set_sbi_flag(sbi, SBI_NEED_FSCK);
2250                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2251                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2252                 /* do checkpoint only */
2253                 ret = f2fs_sync_fs(sb, 1);
2254                 goto out;
2255         default:
2256                 ret = -EINVAL;
2257                 goto out;
2258         }
2259
2260         f2fs_stop_gc_thread(sbi);
2261         f2fs_stop_discard_thread(sbi);
2262
2263         f2fs_drop_discard_cmd(sbi);
2264         clear_opt(sbi, DISCARD);
2265
2266         f2fs_update_time(sbi, REQ_TIME);
2267 out:
2268         if (in != F2FS_GOING_DOWN_FULLSYNC)
2269                 mnt_drop_write_file(filp);
2270
2271         trace_f2fs_shutdown(sbi, in, ret);
2272
2273         return ret;
2274 }
2275
2276 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2277 {
2278         struct inode *inode = file_inode(filp);
2279         struct super_block *sb = inode->i_sb;
2280         struct request_queue *q = bdev_get_queue(sb->s_bdev);
2281         struct fstrim_range range;
2282         int ret;
2283
2284         if (!capable(CAP_SYS_ADMIN))
2285                 return -EPERM;
2286
2287         if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2288                 return -EOPNOTSUPP;
2289
2290         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2291                                 sizeof(range)))
2292                 return -EFAULT;
2293
2294         ret = mnt_want_write_file(filp);
2295         if (ret)
2296                 return ret;
2297
2298         range.minlen = max((unsigned int)range.minlen,
2299                                 q->limits.discard_granularity);
2300         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2301         mnt_drop_write_file(filp);
2302         if (ret < 0)
2303                 return ret;
2304
2305         if (copy_to_user((struct fstrim_range __user *)arg, &range,
2306                                 sizeof(range)))
2307                 return -EFAULT;
2308         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2309         return 0;
2310 }
2311
2312 static bool uuid_is_nonzero(__u8 u[16])
2313 {
2314         int i;
2315
2316         for (i = 0; i < 16; i++)
2317                 if (u[i])
2318                         return true;
2319         return false;
2320 }
2321
2322 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2323 {
2324         struct inode *inode = file_inode(filp);
2325
2326         if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2327                 return -EOPNOTSUPP;
2328
2329         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2330
2331         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2332 }
2333
2334 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2335 {
2336         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2337                 return -EOPNOTSUPP;
2338         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2339 }
2340
2341 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2342 {
2343         struct inode *inode = file_inode(filp);
2344         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2345         int err;
2346
2347         if (!f2fs_sb_has_encrypt(sbi))
2348                 return -EOPNOTSUPP;
2349
2350         err = mnt_want_write_file(filp);
2351         if (err)
2352                 return err;
2353
2354         down_write(&sbi->sb_lock);
2355
2356         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2357                 goto got_it;
2358
2359         /* update superblock with uuid */
2360         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2361
2362         err = f2fs_commit_super(sbi, false);
2363         if (err) {
2364                 /* undo new data */
2365                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2366                 goto out_err;
2367         }
2368 got_it:
2369         if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2370                                                                         16))
2371                 err = -EFAULT;
2372 out_err:
2373         up_write(&sbi->sb_lock);
2374         mnt_drop_write_file(filp);
2375         return err;
2376 }
2377
2378 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2379                                              unsigned long arg)
2380 {
2381         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2382                 return -EOPNOTSUPP;
2383
2384         return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2385 }
2386
2387 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2388 {
2389         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2390                 return -EOPNOTSUPP;
2391
2392         return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2393 }
2394
2395 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2396 {
2397         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2398                 return -EOPNOTSUPP;
2399
2400         return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2401 }
2402
2403 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2404                                                     unsigned long arg)
2405 {
2406         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2407                 return -EOPNOTSUPP;
2408
2409         return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2410 }
2411
2412 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2413                                               unsigned long arg)
2414 {
2415         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2416                 return -EOPNOTSUPP;
2417
2418         return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2419 }
2420
2421 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2422 {
2423         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2424                 return -EOPNOTSUPP;
2425
2426         return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2427 }
2428
2429 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2430 {
2431         struct inode *inode = file_inode(filp);
2432         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2433         __u32 sync;
2434         int ret;
2435
2436         if (!capable(CAP_SYS_ADMIN))
2437                 return -EPERM;
2438
2439         if (get_user(sync, (__u32 __user *)arg))
2440                 return -EFAULT;
2441
2442         if (f2fs_readonly(sbi->sb))
2443                 return -EROFS;
2444
2445         ret = mnt_want_write_file(filp);
2446         if (ret)
2447                 return ret;
2448
2449         if (!sync) {
2450                 if (!down_write_trylock(&sbi->gc_lock)) {
2451                         ret = -EBUSY;
2452                         goto out;
2453                 }
2454         } else {
2455                 down_write(&sbi->gc_lock);
2456         }
2457
2458         ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO);
2459 out:
2460         mnt_drop_write_file(filp);
2461         return ret;
2462 }
2463
2464 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2465 {
2466         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2467         u64 end;
2468         int ret;
2469
2470         if (!capable(CAP_SYS_ADMIN))
2471                 return -EPERM;
2472         if (f2fs_readonly(sbi->sb))
2473                 return -EROFS;
2474
2475         end = range->start + range->len;
2476         if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2477                                         end >= MAX_BLKADDR(sbi))
2478                 return -EINVAL;
2479
2480         ret = mnt_want_write_file(filp);
2481         if (ret)
2482                 return ret;
2483
2484 do_more:
2485         if (!range->sync) {
2486                 if (!down_write_trylock(&sbi->gc_lock)) {
2487                         ret = -EBUSY;
2488                         goto out;
2489                 }
2490         } else {
2491                 down_write(&sbi->gc_lock);
2492         }
2493
2494         ret = f2fs_gc(sbi, range->sync, true, false,
2495                                 GET_SEGNO(sbi, range->start));
2496         if (ret) {
2497                 if (ret == -EBUSY)
2498                         ret = -EAGAIN;
2499                 goto out;
2500         }
2501         range->start += BLKS_PER_SEC(sbi);
2502         if (range->start <= end)
2503                 goto do_more;
2504 out:
2505         mnt_drop_write_file(filp);
2506         return ret;
2507 }
2508
2509 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2510 {
2511         struct f2fs_gc_range range;
2512
2513         if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2514                                                         sizeof(range)))
2515                 return -EFAULT;
2516         return __f2fs_ioc_gc_range(filp, &range);
2517 }
2518
2519 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2520 {
2521         struct inode *inode = file_inode(filp);
2522         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2523         int ret;
2524
2525         if (!capable(CAP_SYS_ADMIN))
2526                 return -EPERM;
2527
2528         if (f2fs_readonly(sbi->sb))
2529                 return -EROFS;
2530
2531         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2532                 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2533                 return -EINVAL;
2534         }
2535
2536         ret = mnt_want_write_file(filp);
2537         if (ret)
2538                 return ret;
2539
2540         ret = f2fs_sync_fs(sbi->sb, 1);
2541
2542         mnt_drop_write_file(filp);
2543         return ret;
2544 }
2545
2546 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2547                                         struct file *filp,
2548                                         struct f2fs_defragment *range)
2549 {
2550         struct inode *inode = file_inode(filp);
2551         struct f2fs_map_blocks map = { .m_next_extent = NULL,
2552                                         .m_seg_type = NO_CHECK_TYPE,
2553                                         .m_may_create = false };
2554         struct extent_info ei = {0, 0, 0};
2555         pgoff_t pg_start, pg_end, next_pgofs;
2556         unsigned int blk_per_seg = sbi->blocks_per_seg;
2557         unsigned int total = 0, sec_num;
2558         block_t blk_end = 0;
2559         bool fragmented = false;
2560         int err;
2561
2562         /* if in-place-update policy is enabled, don't waste time here */
2563         if (f2fs_should_update_inplace(inode, NULL))
2564                 return -EINVAL;
2565
2566         pg_start = range->start >> PAGE_SHIFT;
2567         pg_end = (range->start + range->len) >> PAGE_SHIFT;
2568
2569         f2fs_balance_fs(sbi, true);
2570
2571         inode_lock(inode);
2572
2573         /* writeback all dirty pages in the range */
2574         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2575                                                 range->start + range->len - 1);
2576         if (err)
2577                 goto out;
2578
2579         /*
2580          * lookup mapping info in extent cache, skip defragmenting if physical
2581          * block addresses are continuous.
2582          */
2583         if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2584                 if (ei.fofs + ei.len >= pg_end)
2585                         goto out;
2586         }
2587
2588         map.m_lblk = pg_start;
2589         map.m_next_pgofs = &next_pgofs;
2590
2591         /*
2592          * lookup mapping info in dnode page cache, skip defragmenting if all
2593          * physical block addresses are continuous even if there are hole(s)
2594          * in logical blocks.
2595          */
2596         while (map.m_lblk < pg_end) {
2597                 map.m_len = pg_end - map.m_lblk;
2598                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2599                 if (err)
2600                         goto out;
2601
2602                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2603                         map.m_lblk = next_pgofs;
2604                         continue;
2605                 }
2606
2607                 if (blk_end && blk_end != map.m_pblk)
2608                         fragmented = true;
2609
2610                 /* record total count of block that we're going to move */
2611                 total += map.m_len;
2612
2613                 blk_end = map.m_pblk + map.m_len;
2614
2615                 map.m_lblk += map.m_len;
2616         }
2617
2618         if (!fragmented) {
2619                 total = 0;
2620                 goto out;
2621         }
2622
2623         sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2624
2625         /*
2626          * make sure there are enough free section for LFS allocation, this can
2627          * avoid defragment running in SSR mode when free section are allocated
2628          * intensively
2629          */
2630         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2631                 err = -EAGAIN;
2632                 goto out;
2633         }
2634
2635         map.m_lblk = pg_start;
2636         map.m_len = pg_end - pg_start;
2637         total = 0;
2638
2639         while (map.m_lblk < pg_end) {
2640                 pgoff_t idx;
2641                 int cnt = 0;
2642
2643 do_map:
2644                 map.m_len = pg_end - map.m_lblk;
2645                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2646                 if (err)
2647                         goto clear_out;
2648
2649                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2650                         map.m_lblk = next_pgofs;
2651                         goto check;
2652                 }
2653
2654                 set_inode_flag(inode, FI_DO_DEFRAG);
2655
2656                 idx = map.m_lblk;
2657                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2658                         struct page *page;
2659
2660                         page = f2fs_get_lock_data_page(inode, idx, true);
2661                         if (IS_ERR(page)) {
2662                                 err = PTR_ERR(page);
2663                                 goto clear_out;
2664                         }
2665
2666                         set_page_dirty(page);
2667                         f2fs_put_page(page, 1);
2668
2669                         idx++;
2670                         cnt++;
2671                         total++;
2672                 }
2673
2674                 map.m_lblk = idx;
2675 check:
2676                 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2677                         goto do_map;
2678
2679                 clear_inode_flag(inode, FI_DO_DEFRAG);
2680
2681                 err = filemap_fdatawrite(inode->i_mapping);
2682                 if (err)
2683                         goto out;
2684         }
2685 clear_out:
2686         clear_inode_flag(inode, FI_DO_DEFRAG);
2687 out:
2688         inode_unlock(inode);
2689         if (!err)
2690                 range->len = (u64)total << PAGE_SHIFT;
2691         return err;
2692 }
2693
2694 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2695 {
2696         struct inode *inode = file_inode(filp);
2697         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2698         struct f2fs_defragment range;
2699         int err;
2700
2701         if (!capable(CAP_SYS_ADMIN))
2702                 return -EPERM;
2703
2704         if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2705                 return -EINVAL;
2706
2707         if (f2fs_readonly(sbi->sb))
2708                 return -EROFS;
2709
2710         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2711                                                         sizeof(range)))
2712                 return -EFAULT;
2713
2714         /* verify alignment of offset & size */
2715         if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2716                 return -EINVAL;
2717
2718         if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2719                                         max_file_blocks(inode)))
2720                 return -EINVAL;
2721
2722         err = mnt_want_write_file(filp);
2723         if (err)
2724                 return err;
2725
2726         err = f2fs_defragment_range(sbi, filp, &range);
2727         mnt_drop_write_file(filp);
2728
2729         f2fs_update_time(sbi, REQ_TIME);
2730         if (err < 0)
2731                 return err;
2732
2733         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2734                                                         sizeof(range)))
2735                 return -EFAULT;
2736
2737         return 0;
2738 }
2739
2740 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2741                         struct file *file_out, loff_t pos_out, size_t len)
2742 {
2743         struct inode *src = file_inode(file_in);
2744         struct inode *dst = file_inode(file_out);
2745         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2746         size_t olen = len, dst_max_i_size = 0;
2747         size_t dst_osize;
2748         int ret;
2749
2750         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2751                                 src->i_sb != dst->i_sb)
2752                 return -EXDEV;
2753
2754         if (unlikely(f2fs_readonly(src->i_sb)))
2755                 return -EROFS;
2756
2757         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2758                 return -EINVAL;
2759
2760         if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2761                 return -EOPNOTSUPP;
2762
2763         if (pos_out < 0 || pos_in < 0)
2764                 return -EINVAL;
2765
2766         if (src == dst) {
2767                 if (pos_in == pos_out)
2768                         return 0;
2769                 if (pos_out > pos_in && pos_out < pos_in + len)
2770                         return -EINVAL;
2771         }
2772
2773         inode_lock(src);
2774         if (src != dst) {
2775                 ret = -EBUSY;
2776                 if (!inode_trylock(dst))
2777                         goto out;
2778         }
2779
2780         ret = -EINVAL;
2781         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2782                 goto out_unlock;
2783         if (len == 0)
2784                 olen = len = src->i_size - pos_in;
2785         if (pos_in + len == src->i_size)
2786                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2787         if (len == 0) {
2788                 ret = 0;
2789                 goto out_unlock;
2790         }
2791
2792         dst_osize = dst->i_size;
2793         if (pos_out + olen > dst->i_size)
2794                 dst_max_i_size = pos_out + olen;
2795
2796         /* verify the end result is block aligned */
2797         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2798                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2799                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2800                 goto out_unlock;
2801
2802         ret = f2fs_convert_inline_inode(src);
2803         if (ret)
2804                 goto out_unlock;
2805
2806         ret = f2fs_convert_inline_inode(dst);
2807         if (ret)
2808                 goto out_unlock;
2809
2810         /* write out all dirty pages from offset */
2811         ret = filemap_write_and_wait_range(src->i_mapping,
2812                                         pos_in, pos_in + len);
2813         if (ret)
2814                 goto out_unlock;
2815
2816         ret = filemap_write_and_wait_range(dst->i_mapping,
2817                                         pos_out, pos_out + len);
2818         if (ret)
2819                 goto out_unlock;
2820
2821         f2fs_balance_fs(sbi, true);
2822
2823         down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2824         if (src != dst) {
2825                 ret = -EBUSY;
2826                 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2827                         goto out_src;
2828         }
2829
2830         f2fs_lock_op(sbi);
2831         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2832                                 pos_out >> F2FS_BLKSIZE_BITS,
2833                                 len >> F2FS_BLKSIZE_BITS, false);
2834
2835         if (!ret) {
2836                 if (dst_max_i_size)
2837                         f2fs_i_size_write(dst, dst_max_i_size);
2838                 else if (dst_osize != dst->i_size)
2839                         f2fs_i_size_write(dst, dst_osize);
2840         }
2841         f2fs_unlock_op(sbi);
2842
2843         if (src != dst)
2844                 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2845 out_src:
2846         up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2847 out_unlock:
2848         if (src != dst)
2849                 inode_unlock(dst);
2850 out:
2851         inode_unlock(src);
2852         return ret;
2853 }
2854
2855 static int __f2fs_ioc_move_range(struct file *filp,
2856                                 struct f2fs_move_range *range)
2857 {
2858         struct fd dst;
2859         int err;
2860
2861         if (!(filp->f_mode & FMODE_READ) ||
2862                         !(filp->f_mode & FMODE_WRITE))
2863                 return -EBADF;
2864
2865         dst = fdget(range->dst_fd);
2866         if (!dst.file)
2867                 return -EBADF;
2868
2869         if (!(dst.file->f_mode & FMODE_WRITE)) {
2870                 err = -EBADF;
2871                 goto err_out;
2872         }
2873
2874         err = mnt_want_write_file(filp);
2875         if (err)
2876                 goto err_out;
2877
2878         err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2879                                         range->pos_out, range->len);
2880
2881         mnt_drop_write_file(filp);
2882 err_out:
2883         fdput(dst);
2884         return err;
2885 }
2886
2887 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2888 {
2889         struct f2fs_move_range range;
2890
2891         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2892                                                         sizeof(range)))
2893                 return -EFAULT;
2894         return __f2fs_ioc_move_range(filp, &range);
2895 }
2896
2897 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2898 {
2899         struct inode *inode = file_inode(filp);
2900         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2901         struct sit_info *sm = SIT_I(sbi);
2902         unsigned int start_segno = 0, end_segno = 0;
2903         unsigned int dev_start_segno = 0, dev_end_segno = 0;
2904         struct f2fs_flush_device range;
2905         int ret;
2906
2907         if (!capable(CAP_SYS_ADMIN))
2908                 return -EPERM;
2909
2910         if (f2fs_readonly(sbi->sb))
2911                 return -EROFS;
2912
2913         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2914                 return -EINVAL;
2915
2916         if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2917                                                         sizeof(range)))
2918                 return -EFAULT;
2919
2920         if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2921                         __is_large_section(sbi)) {
2922                 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2923                           range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2924                 return -EINVAL;
2925         }
2926
2927         ret = mnt_want_write_file(filp);
2928         if (ret)
2929                 return ret;
2930
2931         if (range.dev_num != 0)
2932                 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2933         dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2934
2935         start_segno = sm->last_victim[FLUSH_DEVICE];
2936         if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2937                 start_segno = dev_start_segno;
2938         end_segno = min(start_segno + range.segments, dev_end_segno);
2939
2940         while (start_segno < end_segno) {
2941                 if (!down_write_trylock(&sbi->gc_lock)) {
2942                         ret = -EBUSY;
2943                         goto out;
2944                 }
2945                 sm->last_victim[GC_CB] = end_segno + 1;
2946                 sm->last_victim[GC_GREEDY] = end_segno + 1;
2947                 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2948                 ret = f2fs_gc(sbi, true, true, true, start_segno);
2949                 if (ret == -EAGAIN)
2950                         ret = 0;
2951                 else if (ret < 0)
2952                         break;
2953                 start_segno++;
2954         }
2955 out:
2956         mnt_drop_write_file(filp);
2957         return ret;
2958 }
2959
2960 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2961 {
2962         struct inode *inode = file_inode(filp);
2963         u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2964
2965         /* Must validate to set it with SQLite behavior in Android. */
2966         sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2967
2968         return put_user(sb_feature, (u32 __user *)arg);
2969 }
2970
2971 #ifdef CONFIG_QUOTA
2972 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2973 {
2974         struct dquot *transfer_to[MAXQUOTAS] = {};
2975         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2976         struct super_block *sb = sbi->sb;
2977         int err = 0;
2978
2979         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2980         if (!IS_ERR(transfer_to[PRJQUOTA])) {
2981                 err = __dquot_transfer(inode, transfer_to);
2982                 if (err)
2983                         set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2984                 dqput(transfer_to[PRJQUOTA]);
2985         }
2986         return err;
2987 }
2988
2989 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2990 {
2991         struct f2fs_inode_info *fi = F2FS_I(inode);
2992         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2993         struct page *ipage;
2994         kprojid_t kprojid;
2995         int err;
2996
2997         if (!f2fs_sb_has_project_quota(sbi)) {
2998                 if (projid != F2FS_DEF_PROJID)
2999                         return -EOPNOTSUPP;
3000                 else
3001                         return 0;
3002         }
3003
3004         if (!f2fs_has_extra_attr(inode))
3005                 return -EOPNOTSUPP;
3006
3007         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3008
3009         if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
3010                 return 0;
3011
3012         err = -EPERM;
3013         /* Is it quota file? Do not allow user to mess with it */
3014         if (IS_NOQUOTA(inode))
3015                 return err;
3016
3017         ipage = f2fs_get_node_page(sbi, inode->i_ino);
3018         if (IS_ERR(ipage))
3019                 return PTR_ERR(ipage);
3020
3021         if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
3022                                                                 i_projid)) {
3023                 err = -EOVERFLOW;
3024                 f2fs_put_page(ipage, 1);
3025                 return err;
3026         }
3027         f2fs_put_page(ipage, 1);
3028
3029         err = f2fs_dquot_initialize(inode);
3030         if (err)
3031                 return err;
3032
3033         f2fs_lock_op(sbi);
3034         err = f2fs_transfer_project_quota(inode, kprojid);
3035         if (err)
3036                 goto out_unlock;
3037
3038         F2FS_I(inode)->i_projid = kprojid;
3039         inode->i_ctime = current_time(inode);
3040         f2fs_mark_inode_dirty_sync(inode, true);
3041 out_unlock:
3042         f2fs_unlock_op(sbi);
3043         return err;
3044 }
3045 #else
3046 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3047 {
3048         return 0;
3049 }
3050
3051 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3052 {
3053         if (projid != F2FS_DEF_PROJID)
3054                 return -EOPNOTSUPP;
3055         return 0;
3056 }
3057 #endif
3058
3059 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3060 {
3061         struct inode *inode = d_inode(dentry);
3062         struct f2fs_inode_info *fi = F2FS_I(inode);
3063         u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3064
3065         if (IS_ENCRYPTED(inode))
3066                 fsflags |= FS_ENCRYPT_FL;
3067         if (IS_VERITY(inode))
3068                 fsflags |= FS_VERITY_FL;
3069         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3070                 fsflags |= FS_INLINE_DATA_FL;
3071         if (is_inode_flag_set(inode, FI_PIN_FILE))
3072                 fsflags |= FS_NOCOW_FL;
3073
3074         fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3075
3076         if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3077                 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3078
3079         return 0;
3080 }
3081
3082 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3083                       struct dentry *dentry, struct fileattr *fa)
3084 {
3085         struct inode *inode = d_inode(dentry);
3086         u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3087         u32 iflags;
3088         int err;
3089
3090         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3091                 return -EIO;
3092         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3093                 return -ENOSPC;
3094         if (fsflags & ~F2FS_GETTABLE_FS_FL)
3095                 return -EOPNOTSUPP;
3096         fsflags &= F2FS_SETTABLE_FS_FL;
3097         if (!fa->flags_valid)
3098                 mask &= FS_COMMON_FL;
3099
3100         iflags = f2fs_fsflags_to_iflags(fsflags);
3101         if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3102                 return -EOPNOTSUPP;
3103
3104         err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3105         if (!err)
3106                 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3107
3108         return err;
3109 }
3110
3111 int f2fs_pin_file_control(struct inode *inode, bool inc)
3112 {
3113         struct f2fs_inode_info *fi = F2FS_I(inode);
3114         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3115
3116         /* Use i_gc_failures for normal file as a risk signal. */
3117         if (inc)
3118                 f2fs_i_gc_failures_write(inode,
3119                                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3120
3121         if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3122                 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3123                           __func__, inode->i_ino,
3124                           fi->i_gc_failures[GC_FAILURE_PIN]);
3125                 clear_inode_flag(inode, FI_PIN_FILE);
3126                 return -EAGAIN;
3127         }
3128         return 0;
3129 }
3130
3131 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3132 {
3133         struct inode *inode = file_inode(filp);
3134         __u32 pin;
3135         int ret = 0;
3136
3137         if (get_user(pin, (__u32 __user *)arg))
3138                 return -EFAULT;
3139
3140         if (!S_ISREG(inode->i_mode))
3141                 return -EINVAL;
3142
3143         if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3144                 return -EROFS;
3145
3146         ret = mnt_want_write_file(filp);
3147         if (ret)
3148                 return ret;
3149
3150         inode_lock(inode);
3151
3152         if (!pin) {
3153                 clear_inode_flag(inode, FI_PIN_FILE);
3154                 f2fs_i_gc_failures_write(inode, 0);
3155                 goto done;
3156         }
3157
3158         if (f2fs_should_update_outplace(inode, NULL)) {
3159                 ret = -EINVAL;
3160                 goto out;
3161         }
3162
3163         if (f2fs_pin_file_control(inode, false)) {
3164                 ret = -EAGAIN;
3165                 goto out;
3166         }
3167
3168         ret = f2fs_convert_inline_inode(inode);
3169         if (ret)
3170                 goto out;
3171
3172         if (!f2fs_disable_compressed_file(inode)) {
3173                 ret = -EOPNOTSUPP;
3174                 goto out;
3175         }
3176
3177         set_inode_flag(inode, FI_PIN_FILE);
3178         ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3179 done:
3180         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3181 out:
3182         inode_unlock(inode);
3183         mnt_drop_write_file(filp);
3184         return ret;
3185 }
3186
3187 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3188 {
3189         struct inode *inode = file_inode(filp);
3190         __u32 pin = 0;
3191
3192         if (is_inode_flag_set(inode, FI_PIN_FILE))
3193                 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3194         return put_user(pin, (u32 __user *)arg);
3195 }
3196
3197 int f2fs_precache_extents(struct inode *inode)
3198 {
3199         struct f2fs_inode_info *fi = F2FS_I(inode);
3200         struct f2fs_map_blocks map;
3201         pgoff_t m_next_extent;
3202         loff_t end;
3203         int err;
3204
3205         if (is_inode_flag_set(inode, FI_NO_EXTENT))
3206                 return -EOPNOTSUPP;
3207
3208         map.m_lblk = 0;
3209         map.m_next_pgofs = NULL;
3210         map.m_next_extent = &m_next_extent;
3211         map.m_seg_type = NO_CHECK_TYPE;
3212         map.m_may_create = false;
3213         end = max_file_blocks(inode);
3214
3215         while (map.m_lblk < end) {
3216                 map.m_len = end - map.m_lblk;
3217
3218                 down_write(&fi->i_gc_rwsem[WRITE]);
3219                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3220                 up_write(&fi->i_gc_rwsem[WRITE]);
3221                 if (err)
3222                         return err;
3223
3224                 map.m_lblk = m_next_extent;
3225         }
3226
3227         return 0;
3228 }
3229
3230 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3231 {
3232         return f2fs_precache_extents(file_inode(filp));
3233 }
3234
3235 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3236 {
3237         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3238         __u64 block_count;
3239
3240         if (!capable(CAP_SYS_ADMIN))
3241                 return -EPERM;
3242
3243         if (f2fs_readonly(sbi->sb))
3244                 return -EROFS;
3245
3246         if (copy_from_user(&block_count, (void __user *)arg,
3247                            sizeof(block_count)))
3248                 return -EFAULT;
3249
3250         return f2fs_resize_fs(sbi, block_count);
3251 }
3252
3253 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3254 {
3255         struct inode *inode = file_inode(filp);
3256
3257         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3258
3259         if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3260                 f2fs_warn(F2FS_I_SB(inode),
3261                           "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3262                           inode->i_ino);
3263                 return -EOPNOTSUPP;
3264         }
3265
3266         return fsverity_ioctl_enable(filp, (const void __user *)arg);
3267 }
3268
3269 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3270 {
3271         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3272                 return -EOPNOTSUPP;
3273
3274         return fsverity_ioctl_measure(filp, (void __user *)arg);
3275 }
3276
3277 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3278 {
3279         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3280                 return -EOPNOTSUPP;
3281
3282         return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3283 }
3284
3285 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3286 {
3287         struct inode *inode = file_inode(filp);
3288         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3289         char *vbuf;
3290         int count;
3291         int err = 0;
3292
3293         vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3294         if (!vbuf)
3295                 return -ENOMEM;
3296
3297         down_read(&sbi->sb_lock);
3298         count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3299                         ARRAY_SIZE(sbi->raw_super->volume_name),
3300                         UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3301         up_read(&sbi->sb_lock);
3302
3303         if (copy_to_user((char __user *)arg, vbuf,
3304                                 min(FSLABEL_MAX, count)))
3305                 err = -EFAULT;
3306
3307         kfree(vbuf);
3308         return err;
3309 }
3310
3311 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3312 {
3313         struct inode *inode = file_inode(filp);
3314         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3315         char *vbuf;
3316         int err = 0;
3317
3318         if (!capable(CAP_SYS_ADMIN))
3319                 return -EPERM;
3320
3321         vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3322         if (IS_ERR(vbuf))
3323                 return PTR_ERR(vbuf);
3324
3325         err = mnt_want_write_file(filp);
3326         if (err)
3327                 goto out;
3328
3329         down_write(&sbi->sb_lock);
3330
3331         memset(sbi->raw_super->volume_name, 0,
3332                         sizeof(sbi->raw_super->volume_name));
3333         utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3334                         sbi->raw_super->volume_name,
3335                         ARRAY_SIZE(sbi->raw_super->volume_name));
3336
3337         err = f2fs_commit_super(sbi, false);
3338
3339         up_write(&sbi->sb_lock);
3340
3341         mnt_drop_write_file(filp);
3342 out:
3343         kfree(vbuf);
3344         return err;
3345 }
3346
3347 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3348 {
3349         struct inode *inode = file_inode(filp);
3350         __u64 blocks;
3351
3352         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3353                 return -EOPNOTSUPP;
3354
3355         if (!f2fs_compressed_file(inode))
3356                 return -EINVAL;
3357
3358         blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3359         return put_user(blocks, (u64 __user *)arg);
3360 }
3361
3362 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3363 {
3364         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3365         unsigned int released_blocks = 0;
3366         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3367         block_t blkaddr;
3368         int i;
3369
3370         for (i = 0; i < count; i++) {
3371                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3372                                                 dn->ofs_in_node + i);
3373
3374                 if (!__is_valid_data_blkaddr(blkaddr))
3375                         continue;
3376                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3377                                         DATA_GENERIC_ENHANCE)))
3378                         return -EFSCORRUPTED;
3379         }
3380
3381         while (count) {
3382                 int compr_blocks = 0;
3383
3384                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3385                         blkaddr = f2fs_data_blkaddr(dn);
3386
3387                         if (i == 0) {
3388                                 if (blkaddr == COMPRESS_ADDR)
3389                                         continue;
3390                                 dn->ofs_in_node += cluster_size;
3391                                 goto next;
3392                         }
3393
3394                         if (__is_valid_data_blkaddr(blkaddr))
3395                                 compr_blocks++;
3396
3397                         if (blkaddr != NEW_ADDR)
3398                                 continue;
3399
3400                         dn->data_blkaddr = NULL_ADDR;
3401                         f2fs_set_data_blkaddr(dn);
3402                 }
3403
3404                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3405                 dec_valid_block_count(sbi, dn->inode,
3406                                         cluster_size - compr_blocks);
3407
3408                 released_blocks += cluster_size - compr_blocks;
3409 next:
3410                 count -= cluster_size;
3411         }
3412
3413         return released_blocks;
3414 }
3415
3416 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3417 {
3418         struct inode *inode = file_inode(filp);
3419         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3420         pgoff_t page_idx = 0, last_idx;
3421         unsigned int released_blocks = 0;
3422         int ret;
3423         int writecount;
3424
3425         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3426                 return -EOPNOTSUPP;
3427
3428         if (!f2fs_compressed_file(inode))
3429                 return -EINVAL;
3430
3431         if (f2fs_readonly(sbi->sb))
3432                 return -EROFS;
3433
3434         ret = mnt_want_write_file(filp);
3435         if (ret)
3436                 return ret;
3437
3438         f2fs_balance_fs(F2FS_I_SB(inode), true);
3439
3440         inode_lock(inode);
3441
3442         writecount = atomic_read(&inode->i_writecount);
3443         if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3444                         (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3445                 ret = -EBUSY;
3446                 goto out;
3447         }
3448
3449         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3450                 ret = -EINVAL;
3451                 goto out;
3452         }
3453
3454         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3455         if (ret)
3456                 goto out;
3457
3458         set_inode_flag(inode, FI_COMPRESS_RELEASED);
3459         inode->i_ctime = current_time(inode);
3460         f2fs_mark_inode_dirty_sync(inode, true);
3461
3462         if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3463                 goto out;
3464
3465         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3466         filemap_invalidate_lock(inode->i_mapping);
3467
3468         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3469
3470         while (page_idx < last_idx) {
3471                 struct dnode_of_data dn;
3472                 pgoff_t end_offset, count;
3473
3474                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3475                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3476                 if (ret) {
3477                         if (ret == -ENOENT) {
3478                                 page_idx = f2fs_get_next_page_offset(&dn,
3479                                                                 page_idx);
3480                                 ret = 0;
3481                                 continue;
3482                         }
3483                         break;
3484                 }
3485
3486                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3487                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3488                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3489
3490                 ret = release_compress_blocks(&dn, count);
3491
3492                 f2fs_put_dnode(&dn);
3493
3494                 if (ret < 0)
3495                         break;
3496
3497                 page_idx += count;
3498                 released_blocks += ret;
3499         }
3500
3501         filemap_invalidate_unlock(inode->i_mapping);
3502         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3503 out:
3504         inode_unlock(inode);
3505
3506         mnt_drop_write_file(filp);
3507
3508         if (ret >= 0) {
3509                 ret = put_user(released_blocks, (u64 __user *)arg);
3510         } else if (released_blocks &&
3511                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3512                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3513                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3514                         "iblocks=%llu, released=%u, compr_blocks=%u, "
3515                         "run fsck to fix.",
3516                         __func__, inode->i_ino, inode->i_blocks,
3517                         released_blocks,
3518                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3519         }
3520
3521         return ret;
3522 }
3523
3524 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3525 {
3526         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3527         unsigned int reserved_blocks = 0;
3528         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3529         block_t blkaddr;
3530         int i;
3531
3532         for (i = 0; i < count; i++) {
3533                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3534                                                 dn->ofs_in_node + i);
3535
3536                 if (!__is_valid_data_blkaddr(blkaddr))
3537                         continue;
3538                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3539                                         DATA_GENERIC_ENHANCE)))
3540                         return -EFSCORRUPTED;
3541         }
3542
3543         while (count) {
3544                 int compr_blocks = 0;
3545                 blkcnt_t reserved;
3546                 int ret;
3547
3548                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3549                         blkaddr = f2fs_data_blkaddr(dn);
3550
3551                         if (i == 0) {
3552                                 if (blkaddr == COMPRESS_ADDR)
3553                                         continue;
3554                                 dn->ofs_in_node += cluster_size;
3555                                 goto next;
3556                         }
3557
3558                         if (__is_valid_data_blkaddr(blkaddr)) {
3559                                 compr_blocks++;
3560                                 continue;
3561                         }
3562
3563                         dn->data_blkaddr = NEW_ADDR;
3564                         f2fs_set_data_blkaddr(dn);
3565                 }
3566
3567                 reserved = cluster_size - compr_blocks;
3568                 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3569                 if (ret)
3570                         return ret;
3571
3572                 if (reserved != cluster_size - compr_blocks)
3573                         return -ENOSPC;
3574
3575                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3576
3577                 reserved_blocks += reserved;
3578 next:
3579                 count -= cluster_size;
3580         }
3581
3582         return reserved_blocks;
3583 }
3584
3585 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3586 {
3587         struct inode *inode = file_inode(filp);
3588         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3589         pgoff_t page_idx = 0, last_idx;
3590         unsigned int reserved_blocks = 0;
3591         int ret;
3592
3593         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3594                 return -EOPNOTSUPP;
3595
3596         if (!f2fs_compressed_file(inode))
3597                 return -EINVAL;
3598
3599         if (f2fs_readonly(sbi->sb))
3600                 return -EROFS;
3601
3602         ret = mnt_want_write_file(filp);
3603         if (ret)
3604                 return ret;
3605
3606         if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3607                 goto out;
3608
3609         f2fs_balance_fs(F2FS_I_SB(inode), true);
3610
3611         inode_lock(inode);
3612
3613         if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3614                 ret = -EINVAL;
3615                 goto unlock_inode;
3616         }
3617
3618         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3619         filemap_invalidate_lock(inode->i_mapping);
3620
3621         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3622
3623         while (page_idx < last_idx) {
3624                 struct dnode_of_data dn;
3625                 pgoff_t end_offset, count;
3626
3627                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3628                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3629                 if (ret) {
3630                         if (ret == -ENOENT) {
3631                                 page_idx = f2fs_get_next_page_offset(&dn,
3632                                                                 page_idx);
3633                                 ret = 0;
3634                                 continue;
3635                         }
3636                         break;
3637                 }
3638
3639                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3640                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3641                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3642
3643                 ret = reserve_compress_blocks(&dn, count);
3644
3645                 f2fs_put_dnode(&dn);
3646
3647                 if (ret < 0)
3648                         break;
3649
3650                 page_idx += count;
3651                 reserved_blocks += ret;
3652         }
3653
3654         filemap_invalidate_unlock(inode->i_mapping);
3655         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3656
3657         if (ret >= 0) {
3658                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3659                 inode->i_ctime = current_time(inode);
3660                 f2fs_mark_inode_dirty_sync(inode, true);
3661         }
3662 unlock_inode:
3663         inode_unlock(inode);
3664 out:
3665         mnt_drop_write_file(filp);
3666
3667         if (ret >= 0) {
3668                 ret = put_user(reserved_blocks, (u64 __user *)arg);
3669         } else if (reserved_blocks &&
3670                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3671                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3672                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3673                         "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3674                         "run fsck to fix.",
3675                         __func__, inode->i_ino, inode->i_blocks,
3676                         reserved_blocks,
3677                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3678         }
3679
3680         return ret;
3681 }
3682
3683 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3684                 pgoff_t off, block_t block, block_t len, u32 flags)
3685 {
3686         struct request_queue *q = bdev_get_queue(bdev);
3687         sector_t sector = SECTOR_FROM_BLOCK(block);
3688         sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3689         int ret = 0;
3690
3691         if (!q)
3692                 return -ENXIO;
3693
3694         if (flags & F2FS_TRIM_FILE_DISCARD)
3695                 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3696                                                 blk_queue_secure_erase(q) ?
3697                                                 BLKDEV_DISCARD_SECURE : 0);
3698
3699         if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3700                 if (IS_ENCRYPTED(inode))
3701                         ret = fscrypt_zeroout_range(inode, off, block, len);
3702                 else
3703                         ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3704                                         GFP_NOFS, 0);
3705         }
3706
3707         return ret;
3708 }
3709
3710 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3711 {
3712         struct inode *inode = file_inode(filp);
3713         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3714         struct address_space *mapping = inode->i_mapping;
3715         struct block_device *prev_bdev = NULL;
3716         struct f2fs_sectrim_range range;
3717         pgoff_t index, pg_end, prev_index = 0;
3718         block_t prev_block = 0, len = 0;
3719         loff_t end_addr;
3720         bool to_end = false;
3721         int ret = 0;
3722
3723         if (!(filp->f_mode & FMODE_WRITE))
3724                 return -EBADF;
3725
3726         if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3727                                 sizeof(range)))
3728                 return -EFAULT;
3729
3730         if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3731                         !S_ISREG(inode->i_mode))
3732                 return -EINVAL;
3733
3734         if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3735                         !f2fs_hw_support_discard(sbi)) ||
3736                         ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3737                          IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3738                 return -EOPNOTSUPP;
3739
3740         file_start_write(filp);
3741         inode_lock(inode);
3742
3743         if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3744                         range.start >= inode->i_size) {
3745                 ret = -EINVAL;
3746                 goto err;
3747         }
3748
3749         if (range.len == 0)
3750                 goto err;
3751
3752         if (inode->i_size - range.start > range.len) {
3753                 end_addr = range.start + range.len;
3754         } else {
3755                 end_addr = range.len == (u64)-1 ?
3756                         sbi->sb->s_maxbytes : inode->i_size;
3757                 to_end = true;
3758         }
3759
3760         if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3761                         (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3762                 ret = -EINVAL;
3763                 goto err;
3764         }
3765
3766         index = F2FS_BYTES_TO_BLK(range.start);
3767         pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3768
3769         ret = f2fs_convert_inline_inode(inode);
3770         if (ret)
3771                 goto err;
3772
3773         down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3774         filemap_invalidate_lock(mapping);
3775
3776         ret = filemap_write_and_wait_range(mapping, range.start,
3777                         to_end ? LLONG_MAX : end_addr - 1);
3778         if (ret)
3779                 goto out;
3780
3781         truncate_inode_pages_range(mapping, range.start,
3782                         to_end ? -1 : end_addr - 1);
3783
3784         while (index < pg_end) {
3785                 struct dnode_of_data dn;
3786                 pgoff_t end_offset, count;
3787                 int i;
3788
3789                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3790                 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3791                 if (ret) {
3792                         if (ret == -ENOENT) {
3793                                 index = f2fs_get_next_page_offset(&dn, index);
3794                                 continue;
3795                         }
3796                         goto out;
3797                 }
3798
3799                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3800                 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3801                 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3802                         struct block_device *cur_bdev;
3803                         block_t blkaddr = f2fs_data_blkaddr(&dn);
3804
3805                         if (!__is_valid_data_blkaddr(blkaddr))
3806                                 continue;
3807
3808                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3809                                                 DATA_GENERIC_ENHANCE)) {
3810                                 ret = -EFSCORRUPTED;
3811                                 f2fs_put_dnode(&dn);
3812                                 goto out;
3813                         }
3814
3815                         cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3816                         if (f2fs_is_multi_device(sbi)) {
3817                                 int di = f2fs_target_device_index(sbi, blkaddr);
3818
3819                                 blkaddr -= FDEV(di).start_blk;
3820                         }
3821
3822                         if (len) {
3823                                 if (prev_bdev == cur_bdev &&
3824                                                 index == prev_index + len &&
3825                                                 blkaddr == prev_block + len) {
3826                                         len++;
3827                                 } else {
3828                                         ret = f2fs_secure_erase(prev_bdev,
3829                                                 inode, prev_index, prev_block,
3830                                                 len, range.flags);
3831                                         if (ret) {
3832                                                 f2fs_put_dnode(&dn);
3833                                                 goto out;
3834                                         }
3835
3836                                         len = 0;
3837                                 }
3838                         }
3839
3840                         if (!len) {
3841                                 prev_bdev = cur_bdev;
3842                                 prev_index = index;
3843                                 prev_block = blkaddr;
3844                                 len = 1;
3845                         }
3846                 }
3847
3848                 f2fs_put_dnode(&dn);
3849
3850                 if (fatal_signal_pending(current)) {
3851                         ret = -EINTR;
3852                         goto out;
3853                 }
3854                 cond_resched();
3855         }
3856
3857         if (len)
3858                 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3859                                 prev_block, len, range.flags);
3860 out:
3861         filemap_invalidate_unlock(mapping);
3862         up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3863 err:
3864         inode_unlock(inode);
3865         file_end_write(filp);
3866
3867         return ret;
3868 }
3869
3870 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3871 {
3872         struct inode *inode = file_inode(filp);
3873         struct f2fs_comp_option option;
3874
3875         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3876                 return -EOPNOTSUPP;
3877
3878         inode_lock_shared(inode);
3879
3880         if (!f2fs_compressed_file(inode)) {
3881                 inode_unlock_shared(inode);
3882                 return -ENODATA;
3883         }
3884
3885         option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3886         option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3887
3888         inode_unlock_shared(inode);
3889
3890         if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3891                                 sizeof(option)))
3892                 return -EFAULT;
3893
3894         return 0;
3895 }
3896
3897 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3898 {
3899         struct inode *inode = file_inode(filp);
3900         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3901         struct f2fs_comp_option option;
3902         int ret = 0;
3903
3904         if (!f2fs_sb_has_compression(sbi))
3905                 return -EOPNOTSUPP;
3906
3907         if (!(filp->f_mode & FMODE_WRITE))
3908                 return -EBADF;
3909
3910         if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3911                                 sizeof(option)))
3912                 return -EFAULT;
3913
3914         if (!f2fs_compressed_file(inode) ||
3915                         option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3916                         option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3917                         option.algorithm >= COMPRESS_MAX)
3918                 return -EINVAL;
3919
3920         file_start_write(filp);
3921         inode_lock(inode);
3922
3923         if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3924                 ret = -EBUSY;
3925                 goto out;
3926         }
3927
3928         if (inode->i_size != 0) {
3929                 ret = -EFBIG;
3930                 goto out;
3931         }
3932
3933         F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3934         F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3935         F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3936         f2fs_mark_inode_dirty_sync(inode, true);
3937
3938         if (!f2fs_is_compress_backend_ready(inode))
3939                 f2fs_warn(sbi, "compression algorithm is successfully set, "
3940                         "but current kernel doesn't support this algorithm.");
3941 out:
3942         inode_unlock(inode);
3943         file_end_write(filp);
3944
3945         return ret;
3946 }
3947
3948 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3949 {
3950         DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3951         struct address_space *mapping = inode->i_mapping;
3952         struct page *page;
3953         pgoff_t redirty_idx = page_idx;
3954         int i, page_len = 0, ret = 0;
3955
3956         page_cache_ra_unbounded(&ractl, len, 0);
3957
3958         for (i = 0; i < len; i++, page_idx++) {
3959                 page = read_cache_page(mapping, page_idx, NULL, NULL);
3960                 if (IS_ERR(page)) {
3961                         ret = PTR_ERR(page);
3962                         break;
3963                 }
3964                 page_len++;
3965         }
3966
3967         for (i = 0; i < page_len; i++, redirty_idx++) {
3968                 page = find_lock_page(mapping, redirty_idx);
3969                 if (!page) {
3970                         ret = -ENOMEM;
3971                         break;
3972                 }
3973                 set_page_dirty(page);
3974                 f2fs_put_page(page, 1);
3975                 f2fs_put_page(page, 0);
3976         }
3977
3978         return ret;
3979 }
3980
3981 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3982 {
3983         struct inode *inode = file_inode(filp);
3984         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3985         struct f2fs_inode_info *fi = F2FS_I(inode);
3986         pgoff_t page_idx = 0, last_idx;
3987         unsigned int blk_per_seg = sbi->blocks_per_seg;
3988         int cluster_size = F2FS_I(inode)->i_cluster_size;
3989         int count, ret;
3990
3991         if (!f2fs_sb_has_compression(sbi) ||
3992                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3993                 return -EOPNOTSUPP;
3994
3995         if (!(filp->f_mode & FMODE_WRITE))
3996                 return -EBADF;
3997
3998         if (!f2fs_compressed_file(inode))
3999                 return -EINVAL;
4000
4001         f2fs_balance_fs(F2FS_I_SB(inode), true);
4002
4003         file_start_write(filp);
4004         inode_lock(inode);
4005
4006         if (!f2fs_is_compress_backend_ready(inode)) {
4007                 ret = -EOPNOTSUPP;
4008                 goto out;
4009         }
4010
4011         if (f2fs_is_mmap_file(inode)) {
4012                 ret = -EBUSY;
4013                 goto out;
4014         }
4015
4016         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4017         if (ret)
4018                 goto out;
4019
4020         if (!atomic_read(&fi->i_compr_blocks))
4021                 goto out;
4022
4023         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4024
4025         count = last_idx - page_idx;
4026         while (count) {
4027                 int len = min(cluster_size, count);
4028
4029                 ret = redirty_blocks(inode, page_idx, len);
4030                 if (ret < 0)
4031                         break;
4032
4033                 if (get_dirty_pages(inode) >= blk_per_seg)
4034                         filemap_fdatawrite(inode->i_mapping);
4035
4036                 count -= len;
4037                 page_idx += len;
4038         }
4039
4040         if (!ret)
4041                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4042                                                         LLONG_MAX);
4043
4044         if (ret)
4045                 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4046                           __func__, ret);
4047 out:
4048         inode_unlock(inode);
4049         file_end_write(filp);
4050
4051         return ret;
4052 }
4053
4054 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4055 {
4056         struct inode *inode = file_inode(filp);
4057         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4058         pgoff_t page_idx = 0, last_idx;
4059         unsigned int blk_per_seg = sbi->blocks_per_seg;
4060         int cluster_size = F2FS_I(inode)->i_cluster_size;
4061         int count, ret;
4062
4063         if (!f2fs_sb_has_compression(sbi) ||
4064                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4065                 return -EOPNOTSUPP;
4066
4067         if (!(filp->f_mode & FMODE_WRITE))
4068                 return -EBADF;
4069
4070         if (!f2fs_compressed_file(inode))
4071                 return -EINVAL;
4072
4073         f2fs_balance_fs(F2FS_I_SB(inode), true);
4074
4075         file_start_write(filp);
4076         inode_lock(inode);
4077
4078         if (!f2fs_is_compress_backend_ready(inode)) {
4079                 ret = -EOPNOTSUPP;
4080                 goto out;
4081         }
4082
4083         if (f2fs_is_mmap_file(inode)) {
4084                 ret = -EBUSY;
4085                 goto out;
4086         }
4087
4088         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4089         if (ret)
4090                 goto out;
4091
4092         set_inode_flag(inode, FI_ENABLE_COMPRESS);
4093
4094         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4095
4096         count = last_idx - page_idx;
4097         while (count) {
4098                 int len = min(cluster_size, count);
4099
4100                 ret = redirty_blocks(inode, page_idx, len);
4101                 if (ret < 0)
4102                         break;
4103
4104                 if (get_dirty_pages(inode) >= blk_per_seg)
4105                         filemap_fdatawrite(inode->i_mapping);
4106
4107                 count -= len;
4108                 page_idx += len;
4109         }
4110
4111         if (!ret)
4112                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4113                                                         LLONG_MAX);
4114
4115         clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4116
4117         if (ret)
4118                 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4119                           __func__, ret);
4120 out:
4121         inode_unlock(inode);
4122         file_end_write(filp);
4123
4124         return ret;
4125 }
4126
4127 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4128 {
4129         switch (cmd) {
4130         case FS_IOC_GETVERSION:
4131                 return f2fs_ioc_getversion(filp, arg);
4132         case F2FS_IOC_START_ATOMIC_WRITE:
4133                 return f2fs_ioc_start_atomic_write(filp);
4134         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4135                 return f2fs_ioc_commit_atomic_write(filp);
4136         case F2FS_IOC_START_VOLATILE_WRITE:
4137                 return f2fs_ioc_start_volatile_write(filp);
4138         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4139                 return f2fs_ioc_release_volatile_write(filp);
4140         case F2FS_IOC_ABORT_VOLATILE_WRITE:
4141                 return f2fs_ioc_abort_volatile_write(filp);
4142         case F2FS_IOC_SHUTDOWN:
4143                 return f2fs_ioc_shutdown(filp, arg);
4144         case FITRIM:
4145                 return f2fs_ioc_fitrim(filp, arg);
4146         case FS_IOC_SET_ENCRYPTION_POLICY:
4147                 return f2fs_ioc_set_encryption_policy(filp, arg);
4148         case FS_IOC_GET_ENCRYPTION_POLICY:
4149                 return f2fs_ioc_get_encryption_policy(filp, arg);
4150         case FS_IOC_GET_ENCRYPTION_PWSALT:
4151                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4152         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4153                 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4154         case FS_IOC_ADD_ENCRYPTION_KEY:
4155                 return f2fs_ioc_add_encryption_key(filp, arg);
4156         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4157                 return f2fs_ioc_remove_encryption_key(filp, arg);
4158         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4159                 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4160         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4161                 return f2fs_ioc_get_encryption_key_status(filp, arg);
4162         case FS_IOC_GET_ENCRYPTION_NONCE:
4163                 return f2fs_ioc_get_encryption_nonce(filp, arg);
4164         case F2FS_IOC_GARBAGE_COLLECT:
4165                 return f2fs_ioc_gc(filp, arg);
4166         case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4167                 return f2fs_ioc_gc_range(filp, arg);
4168         case F2FS_IOC_WRITE_CHECKPOINT:
4169                 return f2fs_ioc_write_checkpoint(filp, arg);
4170         case F2FS_IOC_DEFRAGMENT:
4171                 return f2fs_ioc_defragment(filp, arg);
4172         case F2FS_IOC_MOVE_RANGE:
4173                 return f2fs_ioc_move_range(filp, arg);
4174         case F2FS_IOC_FLUSH_DEVICE:
4175                 return f2fs_ioc_flush_device(filp, arg);
4176         case F2FS_IOC_GET_FEATURES:
4177                 return f2fs_ioc_get_features(filp, arg);
4178         case F2FS_IOC_GET_PIN_FILE:
4179                 return f2fs_ioc_get_pin_file(filp, arg);
4180         case F2FS_IOC_SET_PIN_FILE:
4181                 return f2fs_ioc_set_pin_file(filp, arg);
4182         case F2FS_IOC_PRECACHE_EXTENTS:
4183                 return f2fs_ioc_precache_extents(filp, arg);
4184         case F2FS_IOC_RESIZE_FS:
4185                 return f2fs_ioc_resize_fs(filp, arg);
4186         case FS_IOC_ENABLE_VERITY:
4187                 return f2fs_ioc_enable_verity(filp, arg);
4188         case FS_IOC_MEASURE_VERITY:
4189                 return f2fs_ioc_measure_verity(filp, arg);
4190         case FS_IOC_READ_VERITY_METADATA:
4191                 return f2fs_ioc_read_verity_metadata(filp, arg);
4192         case FS_IOC_GETFSLABEL:
4193                 return f2fs_ioc_getfslabel(filp, arg);
4194         case FS_IOC_SETFSLABEL:
4195                 return f2fs_ioc_setfslabel(filp, arg);
4196         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4197                 return f2fs_get_compress_blocks(filp, arg);
4198         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4199                 return f2fs_release_compress_blocks(filp, arg);
4200         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4201                 return f2fs_reserve_compress_blocks(filp, arg);
4202         case F2FS_IOC_SEC_TRIM_FILE:
4203                 return f2fs_sec_trim_file(filp, arg);
4204         case F2FS_IOC_GET_COMPRESS_OPTION:
4205                 return f2fs_ioc_get_compress_option(filp, arg);
4206         case F2FS_IOC_SET_COMPRESS_OPTION:
4207                 return f2fs_ioc_set_compress_option(filp, arg);
4208         case F2FS_IOC_DECOMPRESS_FILE:
4209                 return f2fs_ioc_decompress_file(filp, arg);
4210         case F2FS_IOC_COMPRESS_FILE:
4211                 return f2fs_ioc_compress_file(filp, arg);
4212         default:
4213                 return -ENOTTY;
4214         }
4215 }
4216
4217 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4218 {
4219         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4220                 return -EIO;
4221         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4222                 return -ENOSPC;
4223
4224         return __f2fs_ioctl(filp, cmd, arg);
4225 }
4226
4227 /*
4228  * Return %true if the given read or write request should use direct I/O, or
4229  * %false if it should use buffered I/O.
4230  */
4231 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4232                                 struct iov_iter *iter)
4233 {
4234         unsigned int align;
4235
4236         if (!(iocb->ki_flags & IOCB_DIRECT))
4237                 return false;
4238
4239         if (f2fs_force_buffered_io(inode, iocb, iter))
4240                 return false;
4241
4242         /*
4243          * Direct I/O not aligned to the disk's logical_block_size will be
4244          * attempted, but will fail with -EINVAL.
4245          *
4246          * f2fs additionally requires that direct I/O be aligned to the
4247          * filesystem block size, which is often a stricter requirement.
4248          * However, f2fs traditionally falls back to buffered I/O on requests
4249          * that are logical_block_size-aligned but not fs-block aligned.
4250          *
4251          * The below logic implements this behavior.
4252          */
4253         align = iocb->ki_pos | iov_iter_alignment(iter);
4254         if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4255             IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4256                 return false;
4257
4258         return true;
4259 }
4260
4261 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4262                                 unsigned int flags)
4263 {
4264         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4265
4266         dec_page_count(sbi, F2FS_DIO_READ);
4267         if (error)
4268                 return error;
4269         f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4270         return 0;
4271 }
4272
4273 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4274         .end_io = f2fs_dio_read_end_io,
4275 };
4276
4277 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4278 {
4279         struct file *file = iocb->ki_filp;
4280         struct inode *inode = file_inode(file);
4281         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4282         struct f2fs_inode_info *fi = F2FS_I(inode);
4283         const loff_t pos = iocb->ki_pos;
4284         const size_t count = iov_iter_count(to);
4285         struct iomap_dio *dio;
4286         ssize_t ret;
4287
4288         if (count == 0)
4289                 return 0; /* skip atime update */
4290
4291         trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4292
4293         if (iocb->ki_flags & IOCB_NOWAIT) {
4294                 if (!down_read_trylock(&fi->i_gc_rwsem[READ])) {
4295                         ret = -EAGAIN;
4296                         goto out;
4297                 }
4298         } else {
4299                 down_read(&fi->i_gc_rwsem[READ]);
4300         }
4301
4302         /*
4303          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4304          * the higher-level function iomap_dio_rw() in order to ensure that the
4305          * F2FS_DIO_READ counter will be decremented correctly in all cases.
4306          */
4307         inc_page_count(sbi, F2FS_DIO_READ);
4308         dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4309                              &f2fs_iomap_dio_read_ops, 0, 0);
4310         if (IS_ERR_OR_NULL(dio)) {
4311                 ret = PTR_ERR_OR_ZERO(dio);
4312                 if (ret != -EIOCBQUEUED)
4313                         dec_page_count(sbi, F2FS_DIO_READ);
4314         } else {
4315                 ret = iomap_dio_complete(dio);
4316         }
4317
4318         up_read(&fi->i_gc_rwsem[READ]);
4319
4320         file_accessed(file);
4321 out:
4322         trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4323         return ret;
4324 }
4325
4326 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4327 {
4328         struct inode *inode = file_inode(iocb->ki_filp);
4329         ssize_t ret;
4330
4331         if (!f2fs_is_compress_backend_ready(inode))
4332                 return -EOPNOTSUPP;
4333
4334         if (f2fs_should_use_dio(inode, iocb, to))
4335                 return f2fs_dio_read_iter(iocb, to);
4336
4337         ret = filemap_read(iocb, to, 0);
4338         if (ret > 0)
4339                 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4340         return ret;
4341 }
4342
4343 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4344 {
4345         struct file *file = iocb->ki_filp;
4346         struct inode *inode = file_inode(file);
4347         ssize_t count;
4348         int err;
4349
4350         if (IS_IMMUTABLE(inode))
4351                 return -EPERM;
4352
4353         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4354                 return -EPERM;
4355
4356         count = generic_write_checks(iocb, from);
4357         if (count <= 0)
4358                 return count;
4359
4360         err = file_modified(file);
4361         if (err)
4362                 return err;
4363         return count;
4364 }
4365
4366 /*
4367  * Preallocate blocks for a write request, if it is possible and helpful to do
4368  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4369  * blocks were preallocated, or a negative errno value if something went
4370  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4371  * requested blocks (not just some of them) have been allocated.
4372  */
4373 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4374                                    bool dio)
4375 {
4376         struct inode *inode = file_inode(iocb->ki_filp);
4377         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4378         const loff_t pos = iocb->ki_pos;
4379         const size_t count = iov_iter_count(iter);
4380         struct f2fs_map_blocks map = {};
4381         int flag;
4382         int ret;
4383
4384         /* If it will be an out-of-place direct write, don't bother. */
4385         if (dio && f2fs_lfs_mode(sbi))
4386                 return 0;
4387         /*
4388          * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4389          * buffered IO, if DIO meets any holes.
4390          */
4391         if (dio && i_size_read(inode) &&
4392                 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4393                 return 0;
4394
4395         /* No-wait I/O can't allocate blocks. */
4396         if (iocb->ki_flags & IOCB_NOWAIT)
4397                 return 0;
4398
4399         /* If it will be a short write, don't bother. */
4400         if (fault_in_iov_iter_readable(iter, count))
4401                 return 0;
4402
4403         if (f2fs_has_inline_data(inode)) {
4404                 /* If the data will fit inline, don't bother. */
4405                 if (pos + count <= MAX_INLINE_DATA(inode))
4406                         return 0;
4407                 ret = f2fs_convert_inline_inode(inode);
4408                 if (ret)
4409                         return ret;
4410         }
4411
4412         /* Do not preallocate blocks that will be written partially in 4KB. */
4413         map.m_lblk = F2FS_BLK_ALIGN(pos);
4414         map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4415         if (map.m_len > map.m_lblk)
4416                 map.m_len -= map.m_lblk;
4417         else
4418                 map.m_len = 0;
4419         map.m_may_create = true;
4420         if (dio) {
4421                 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4422                 flag = F2FS_GET_BLOCK_PRE_DIO;
4423         } else {
4424                 map.m_seg_type = NO_CHECK_TYPE;
4425                 flag = F2FS_GET_BLOCK_PRE_AIO;
4426         }
4427
4428         ret = f2fs_map_blocks(inode, &map, 1, flag);
4429         /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4430         if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4431                 return ret;
4432         if (ret == 0)
4433                 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4434         return map.m_len;
4435 }
4436
4437 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4438                                         struct iov_iter *from)
4439 {
4440         struct file *file = iocb->ki_filp;
4441         struct inode *inode = file_inode(file);
4442         ssize_t ret;
4443
4444         if (iocb->ki_flags & IOCB_NOWAIT)
4445                 return -EOPNOTSUPP;
4446
4447         current->backing_dev_info = inode_to_bdi(inode);
4448         ret = generic_perform_write(file, from, iocb->ki_pos);
4449         current->backing_dev_info = NULL;
4450
4451         if (ret > 0) {
4452                 iocb->ki_pos += ret;
4453                 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4454         }
4455         return ret;
4456 }
4457
4458 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4459                                  unsigned int flags)
4460 {
4461         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4462
4463         dec_page_count(sbi, F2FS_DIO_WRITE);
4464         if (error)
4465                 return error;
4466         f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4467         return 0;
4468 }
4469
4470 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4471         .end_io = f2fs_dio_write_end_io,
4472 };
4473
4474 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4475                                    bool *may_need_sync)
4476 {
4477         struct file *file = iocb->ki_filp;
4478         struct inode *inode = file_inode(file);
4479         struct f2fs_inode_info *fi = F2FS_I(inode);
4480         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4481         const bool do_opu = f2fs_lfs_mode(sbi);
4482         const int whint_mode = F2FS_OPTION(sbi).whint_mode;
4483         const loff_t pos = iocb->ki_pos;
4484         const ssize_t count = iov_iter_count(from);
4485         const enum rw_hint hint = iocb->ki_hint;
4486         unsigned int dio_flags;
4487         struct iomap_dio *dio;
4488         ssize_t ret;
4489
4490         trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4491
4492         if (iocb->ki_flags & IOCB_NOWAIT) {
4493                 /* f2fs_convert_inline_inode() and block allocation can block */
4494                 if (f2fs_has_inline_data(inode) ||
4495                     !f2fs_overwrite_io(inode, pos, count)) {
4496                         ret = -EAGAIN;
4497                         goto out;
4498                 }
4499
4500                 if (!down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4501                         ret = -EAGAIN;
4502                         goto out;
4503                 }
4504                 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) {
4505                         up_read(&fi->i_gc_rwsem[WRITE]);
4506                         ret = -EAGAIN;
4507                         goto out;
4508                 }
4509         } else {
4510                 ret = f2fs_convert_inline_inode(inode);
4511                 if (ret)
4512                         goto out;
4513
4514                 down_read(&fi->i_gc_rwsem[WRITE]);
4515                 if (do_opu)
4516                         down_read(&fi->i_gc_rwsem[READ]);
4517         }
4518         if (whint_mode == WHINT_MODE_OFF)
4519                 iocb->ki_hint = WRITE_LIFE_NOT_SET;
4520
4521         /*
4522          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4523          * the higher-level function iomap_dio_rw() in order to ensure that the
4524          * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4525          */
4526         inc_page_count(sbi, F2FS_DIO_WRITE);
4527         dio_flags = 0;
4528         if (pos + count > inode->i_size)
4529                 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4530         dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4531                              &f2fs_iomap_dio_write_ops, dio_flags, 0);
4532         if (IS_ERR_OR_NULL(dio)) {
4533                 ret = PTR_ERR_OR_ZERO(dio);
4534                 if (ret == -ENOTBLK)
4535                         ret = 0;
4536                 if (ret != -EIOCBQUEUED)
4537                         dec_page_count(sbi, F2FS_DIO_WRITE);
4538         } else {
4539                 ret = iomap_dio_complete(dio);
4540         }
4541
4542         if (whint_mode == WHINT_MODE_OFF)
4543                 iocb->ki_hint = hint;
4544         if (do_opu)
4545                 up_read(&fi->i_gc_rwsem[READ]);
4546         up_read(&fi->i_gc_rwsem[WRITE]);
4547
4548         if (ret < 0)
4549                 goto out;
4550         if (pos + ret > inode->i_size)
4551                 f2fs_i_size_write(inode, pos + ret);
4552         if (!do_opu)
4553                 set_inode_flag(inode, FI_UPDATE_WRITE);
4554
4555         if (iov_iter_count(from)) {
4556                 ssize_t ret2;
4557                 loff_t bufio_start_pos = iocb->ki_pos;
4558
4559                 /*
4560                  * The direct write was partial, so we need to fall back to a
4561                  * buffered write for the remainder.
4562                  */
4563
4564                 ret2 = f2fs_buffered_write_iter(iocb, from);
4565                 if (iov_iter_count(from))
4566                         f2fs_write_failed(inode, iocb->ki_pos);
4567                 if (ret2 < 0)
4568                         goto out;
4569
4570                 /*
4571                  * Ensure that the pagecache pages are written to disk and
4572                  * invalidated to preserve the expected O_DIRECT semantics.
4573                  */
4574                 if (ret2 > 0) {
4575                         loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4576
4577                         ret += ret2;
4578
4579                         ret2 = filemap_write_and_wait_range(file->f_mapping,
4580                                                             bufio_start_pos,
4581                                                             bufio_end_pos);
4582                         if (ret2 < 0)
4583                                 goto out;
4584                         invalidate_mapping_pages(file->f_mapping,
4585                                                  bufio_start_pos >> PAGE_SHIFT,
4586                                                  bufio_end_pos >> PAGE_SHIFT);
4587                 }
4588         } else {
4589                 /* iomap_dio_rw() already handled the generic_write_sync(). */
4590                 *may_need_sync = false;
4591         }
4592 out:
4593         trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4594         return ret;
4595 }
4596
4597 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4598 {
4599         struct inode *inode = file_inode(iocb->ki_filp);
4600         const loff_t orig_pos = iocb->ki_pos;
4601         const size_t orig_count = iov_iter_count(from);
4602         loff_t target_size;
4603         bool dio;
4604         bool may_need_sync = true;
4605         int preallocated;
4606         ssize_t ret;
4607
4608         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4609                 ret = -EIO;
4610                 goto out;
4611         }
4612
4613         if (!f2fs_is_compress_backend_ready(inode)) {
4614                 ret = -EOPNOTSUPP;
4615                 goto out;
4616         }
4617
4618         if (iocb->ki_flags & IOCB_NOWAIT) {
4619                 if (!inode_trylock(inode)) {
4620                         ret = -EAGAIN;
4621                         goto out;
4622                 }
4623         } else {
4624                 inode_lock(inode);
4625         }
4626
4627         ret = f2fs_write_checks(iocb, from);
4628         if (ret <= 0)
4629                 goto out_unlock;
4630
4631         /* Determine whether we will do a direct write or a buffered write. */
4632         dio = f2fs_should_use_dio(inode, iocb, from);
4633
4634         /* Possibly preallocate the blocks for the write. */
4635         target_size = iocb->ki_pos + iov_iter_count(from);
4636         preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4637         if (preallocated < 0)
4638                 ret = preallocated;
4639         else
4640                 /* Do the actual write. */
4641                 ret = dio ?
4642                         f2fs_dio_write_iter(iocb, from, &may_need_sync):
4643                         f2fs_buffered_write_iter(iocb, from);
4644
4645         /* Don't leave any preallocated blocks around past i_size. */
4646         if (preallocated && i_size_read(inode) < target_size) {
4647                 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4648                 filemap_invalidate_lock(inode->i_mapping);
4649                 if (!f2fs_truncate(inode))
4650                         file_dont_truncate(inode);
4651                 filemap_invalidate_unlock(inode->i_mapping);
4652                 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4653         } else {
4654                 file_dont_truncate(inode);
4655         }
4656
4657         clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4658 out_unlock:
4659         inode_unlock(inode);
4660 out:
4661         trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4662         if (ret > 0 && may_need_sync)
4663                 ret = generic_write_sync(iocb, ret);
4664         return ret;
4665 }
4666
4667 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4668                 int advice)
4669 {
4670         struct address_space *mapping;
4671         struct backing_dev_info *bdi;
4672         struct inode *inode = file_inode(filp);
4673         int err;
4674
4675         if (advice == POSIX_FADV_SEQUENTIAL) {
4676                 if (S_ISFIFO(inode->i_mode))
4677                         return -ESPIPE;
4678
4679                 mapping = filp->f_mapping;
4680                 if (!mapping || len < 0)
4681                         return -EINVAL;
4682
4683                 bdi = inode_to_bdi(mapping->host);
4684                 filp->f_ra.ra_pages = bdi->ra_pages *
4685                         F2FS_I_SB(inode)->seq_file_ra_mul;
4686                 spin_lock(&filp->f_lock);
4687                 filp->f_mode &= ~FMODE_RANDOM;
4688                 spin_unlock(&filp->f_lock);
4689                 return 0;
4690         }
4691
4692         err = generic_fadvise(filp, offset, len, advice);
4693         if (!err && advice == POSIX_FADV_DONTNEED &&
4694                 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4695                 f2fs_compressed_file(inode))
4696                 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4697
4698         return err;
4699 }
4700
4701 #ifdef CONFIG_COMPAT
4702 struct compat_f2fs_gc_range {
4703         u32 sync;
4704         compat_u64 start;
4705         compat_u64 len;
4706 };
4707 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE        _IOW(F2FS_IOCTL_MAGIC, 11,\
4708                                                 struct compat_f2fs_gc_range)
4709
4710 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4711 {
4712         struct compat_f2fs_gc_range __user *urange;
4713         struct f2fs_gc_range range;
4714         int err;
4715
4716         urange = compat_ptr(arg);
4717         err = get_user(range.sync, &urange->sync);
4718         err |= get_user(range.start, &urange->start);
4719         err |= get_user(range.len, &urange->len);
4720         if (err)
4721                 return -EFAULT;
4722
4723         return __f2fs_ioc_gc_range(file, &range);
4724 }
4725
4726 struct compat_f2fs_move_range {
4727         u32 dst_fd;
4728         compat_u64 pos_in;
4729         compat_u64 pos_out;
4730         compat_u64 len;
4731 };
4732 #define F2FS_IOC32_MOVE_RANGE           _IOWR(F2FS_IOCTL_MAGIC, 9,      \
4733                                         struct compat_f2fs_move_range)
4734
4735 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4736 {
4737         struct compat_f2fs_move_range __user *urange;
4738         struct f2fs_move_range range;
4739         int err;
4740
4741         urange = compat_ptr(arg);
4742         err = get_user(range.dst_fd, &urange->dst_fd);
4743         err |= get_user(range.pos_in, &urange->pos_in);
4744         err |= get_user(range.pos_out, &urange->pos_out);
4745         err |= get_user(range.len, &urange->len);
4746         if (err)
4747                 return -EFAULT;
4748
4749         return __f2fs_ioc_move_range(file, &range);
4750 }
4751
4752 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4753 {
4754         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4755                 return -EIO;
4756         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4757                 return -ENOSPC;
4758
4759         switch (cmd) {
4760         case FS_IOC32_GETVERSION:
4761                 cmd = FS_IOC_GETVERSION;
4762                 break;
4763         case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4764                 return f2fs_compat_ioc_gc_range(file, arg);
4765         case F2FS_IOC32_MOVE_RANGE:
4766                 return f2fs_compat_ioc_move_range(file, arg);
4767         case F2FS_IOC_START_ATOMIC_WRITE:
4768         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4769         case F2FS_IOC_START_VOLATILE_WRITE:
4770         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4771         case F2FS_IOC_ABORT_VOLATILE_WRITE:
4772         case F2FS_IOC_SHUTDOWN:
4773         case FITRIM:
4774         case FS_IOC_SET_ENCRYPTION_POLICY:
4775         case FS_IOC_GET_ENCRYPTION_PWSALT:
4776         case FS_IOC_GET_ENCRYPTION_POLICY:
4777         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4778         case FS_IOC_ADD_ENCRYPTION_KEY:
4779         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4780         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4781         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4782         case FS_IOC_GET_ENCRYPTION_NONCE:
4783         case F2FS_IOC_GARBAGE_COLLECT:
4784         case F2FS_IOC_WRITE_CHECKPOINT:
4785         case F2FS_IOC_DEFRAGMENT:
4786         case F2FS_IOC_FLUSH_DEVICE:
4787         case F2FS_IOC_GET_FEATURES:
4788         case F2FS_IOC_GET_PIN_FILE:
4789         case F2FS_IOC_SET_PIN_FILE:
4790         case F2FS_IOC_PRECACHE_EXTENTS:
4791         case F2FS_IOC_RESIZE_FS:
4792         case FS_IOC_ENABLE_VERITY:
4793         case FS_IOC_MEASURE_VERITY:
4794         case FS_IOC_READ_VERITY_METADATA:
4795         case FS_IOC_GETFSLABEL:
4796         case FS_IOC_SETFSLABEL:
4797         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4798         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4799         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4800         case F2FS_IOC_SEC_TRIM_FILE:
4801         case F2FS_IOC_GET_COMPRESS_OPTION:
4802         case F2FS_IOC_SET_COMPRESS_OPTION:
4803         case F2FS_IOC_DECOMPRESS_FILE:
4804         case F2FS_IOC_COMPRESS_FILE:
4805                 break;
4806         default:
4807                 return -ENOIOCTLCMD;
4808         }
4809         return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4810 }
4811 #endif
4812
4813 const struct file_operations f2fs_file_operations = {
4814         .llseek         = f2fs_llseek,
4815         .read_iter      = f2fs_file_read_iter,
4816         .write_iter     = f2fs_file_write_iter,
4817         .open           = f2fs_file_open,
4818         .release        = f2fs_release_file,
4819         .mmap           = f2fs_file_mmap,
4820         .flush          = f2fs_file_flush,
4821         .fsync          = f2fs_sync_file,
4822         .fallocate      = f2fs_fallocate,
4823         .unlocked_ioctl = f2fs_ioctl,
4824 #ifdef CONFIG_COMPAT
4825         .compat_ioctl   = f2fs_compat_ioctl,
4826 #endif
4827         .splice_read    = generic_file_splice_read,
4828         .splice_write   = iter_file_splice_write,
4829         .fadvise        = f2fs_file_fadvise,
4830 };