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         f2fs_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         f2fs_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         f2fs_down_read(&F2FS_I(inode)->i_sem);
 322         cp_reason = need_do_checkpoint(inode);
 323         f2fs_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             (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
 377                 ret = f2fs_issue_flush(sbi, inode->i_ino);
 378         if (!ret) {
 379                 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
 380                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 381                 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
 382         }
 383         f2fs_update_time(sbi, REQ_TIME);
 384 out:
 385         trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 386         return ret;
 387 }
 388
 389 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 390 {
 391         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
 392                 return -EIO;
 393         return f2fs_do_sync_file(file, start, end, datasync, false);
 394 }
 395
 396 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
 397                                 pgoff_t index, int whence)
 398 {
 399         switch (whence) {
 400         case SEEK_DATA:
 401                 if (__is_valid_data_blkaddr(blkaddr))
 402                         return true;
 403                 if (blkaddr == NEW_ADDR &&
 404                     xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
 405                         return true;
 406                 break;
 407         case SEEK_HOLE:
 408                 if (blkaddr == NULL_ADDR)
 409                         return true;
 410                 break;
 411         }
 412         return false;
 413 }
 414
 415 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 416 {
 417         struct inode *inode = file->f_mapping->host;
 418         loff_t maxbytes = inode->i_sb->s_maxbytes;
 419         struct dnode_of_data dn;
 420         pgoff_t pgofs, end_offset;
 421         loff_t data_ofs = offset;
 422         loff_t isize;
 423         int err = 0;
 424
 425         inode_lock(inode);
 426
 427         isize = i_size_read(inode);
 428         if (offset >= isize)
 429                 goto fail;
 430
 431         /* handle inline data case */
 432         if (f2fs_has_inline_data(inode)) {
 433                 if (whence == SEEK_HOLE) {
 434                         data_ofs = isize;
 435                         goto found;
 436                 } else if (whence == SEEK_DATA) {
 437                         data_ofs = offset;
 438                         goto found;
 439                 }
 440         }
 441
 442         pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
 443
 444         for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 445                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 446                 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
 447                 if (err && err != -ENOENT) {
 448                         goto fail;
 449                 } else if (err == -ENOENT) {
 450                         /* direct node does not exists */
 451                         if (whence == SEEK_DATA) {
 452                                 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
 453                                 continue;
 454                         } else {
 455                                 goto found;
 456                         }
 457                 }
 458
 459                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 460
 461                 /* find data/hole in dnode block */
 462                 for (; dn.ofs_in_node < end_offset;
 463                                 dn.ofs_in_node++, pgofs++,
 464                                 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 465                         block_t blkaddr;
 466
 467                         blkaddr = f2fs_data_blkaddr(&dn);
 468
 469                         if (__is_valid_data_blkaddr(blkaddr) &&
 470                                 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
 471                                         blkaddr, DATA_GENERIC_ENHANCE)) {
 472                                 f2fs_put_dnode(&dn);
 473                                 goto fail;
 474                         }
 475
 476                         if (__found_offset(file->f_mapping, blkaddr,
 477                                                         pgofs, whence)) {
 478                                 f2fs_put_dnode(&dn);
 479                                 goto found;
 480                         }
 481                 }
 482                 f2fs_put_dnode(&dn);
 483         }
 484
 485         if (whence == SEEK_DATA)
 486                 goto fail;
 487 found:
 488         if (whence == SEEK_HOLE && data_ofs > isize)
 489                 data_ofs = isize;
 490         inode_unlock(inode);
 491         return vfs_setpos(file, data_ofs, maxbytes);
 492 fail:
 493         inode_unlock(inode);
 494         return -ENXIO;
 495 }
 496
 497 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
 498 {
 499         struct inode *inode = file->f_mapping->host;
 500         loff_t maxbytes = inode->i_sb->s_maxbytes;
 501
 502         if (f2fs_compressed_file(inode))
 503                 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
 504
 505         switch (whence) {
 506         case SEEK_SET:
 507         case SEEK_CUR:
 508         case SEEK_END:
 509                 return generic_file_llseek_size(file, offset, whence,
 510                                                 maxbytes, i_size_read(inode));
 511         case SEEK_DATA:
 512         case SEEK_HOLE:
 513                 if (offset < 0)
 514                         return -ENXIO;
 515                 return f2fs_seek_block(file, offset, whence);
 516         }
 517
 518         return -EINVAL;
 519 }
 520
 521 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 522 {
 523         struct inode *inode = file_inode(file);
 524
 525         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 526                 return -EIO;
 527
 528         if (!f2fs_is_compress_backend_ready(inode))
 529                 return -EOPNOTSUPP;
 530
 531         file_accessed(file);
 532         vma->vm_ops = &f2fs_file_vm_ops;
 533         set_inode_flag(inode, FI_MMAP_FILE);
 534         return 0;
 535 }
 536
 537 static int f2fs_file_open(struct inode *inode, struct file *filp)
 538 {
 539         int err = fscrypt_file_open(inode, filp);
 540
 541         if (err)
 542                 return err;
 543
 544         if (!f2fs_is_compress_backend_ready(inode))
 545                 return -EOPNOTSUPP;
 546
 547         err = fsverity_file_open(inode, filp);
 548         if (err)
 549                 return err;
 550
 551         filp->f_mode |= FMODE_NOWAIT;
 552
 553         return dquot_file_open(inode, filp);
 554 }
 555
 556 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 557 {
 558         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 559         struct f2fs_node *raw_node;
 560         int nr_free = 0, ofs = dn->ofs_in_node, len = count;
 561         __le32 *addr;
 562         int base = 0;
 563         bool compressed_cluster = false;
 564         int cluster_index = 0, valid_blocks = 0;
 565         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
 566         bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
 567
 568         if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
 569                 base = get_extra_isize(dn->inode);
 570
 571         raw_node = F2FS_NODE(dn->node_page);
 572         addr = blkaddr_in_node(raw_node) + base + ofs;
 573
 574         /* Assumption: truncateion starts with cluster */
 575         for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
 576                 block_t blkaddr = le32_to_cpu(*addr);
 577
 578                 if (f2fs_compressed_file(dn->inode) &&
 579                                         !(cluster_index & (cluster_size - 1))) {
 580                         if (compressed_cluster)
 581                                 f2fs_i_compr_blocks_update(dn->inode,
 582                                                         valid_blocks, false);
 583                         compressed_cluster = (blkaddr == COMPRESS_ADDR);
 584                         valid_blocks = 0;
 585                 }
 586
 587                 if (blkaddr == NULL_ADDR)
 588                         continue;
 589
 590                 dn->data_blkaddr = NULL_ADDR;
 591                 f2fs_set_data_blkaddr(dn);
 592
 593                 if (__is_valid_data_blkaddr(blkaddr)) {
 594                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
 595                                         DATA_GENERIC_ENHANCE))
 596                                 continue;
 597                         if (compressed_cluster)
 598                                 valid_blocks++;
 599                 }
 600
 601                 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 602                         clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 603
 604                 f2fs_invalidate_blocks(sbi, blkaddr);
 605
 606                 if (!released || blkaddr != COMPRESS_ADDR)
 607                         nr_free++;
 608         }
 609
 610         if (compressed_cluster)
 611                 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
 612
 613         if (nr_free) {
 614                 pgoff_t fofs;
 615                 /*
 616                  * once we invalidate valid blkaddr in range [ofs, ofs + count],
 617                  * we will invalidate all blkaddr in the whole range.
 618                  */
 619                 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
 620                                                         dn->inode) + ofs;
 621                 f2fs_update_extent_cache_range(dn, fofs, 0, len);
 622                 dec_valid_block_count(sbi, dn->inode, nr_free);
 623         }
 624         dn->ofs_in_node = ofs;
 625
 626         f2fs_update_time(sbi, REQ_TIME);
 627         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 628                                          dn->ofs_in_node, nr_free);
 629 }
 630
 631 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
 632 {
 633         f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
 634 }
 635
 636 static int truncate_partial_data_page(struct inode *inode, u64 from,
 637                                                                 bool cache_only)
 638 {
 639         loff_t offset = from & (PAGE_SIZE - 1);
 640         pgoff_t index = from >> PAGE_SHIFT;
 641         struct address_space *mapping = inode->i_mapping;
 642         struct page *page;
 643
 644         if (!offset && !cache_only)
 645                 return 0;
 646
 647         if (cache_only) {
 648                 page = find_lock_page(mapping, index);
 649                 if (page && PageUptodate(page))
 650                         goto truncate_out;
 651                 f2fs_put_page(page, 1);
 652                 return 0;
 653         }
 654
 655         page = f2fs_get_lock_data_page(inode, index, true);
 656         if (IS_ERR(page))
 657                 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 658 truncate_out:
 659         f2fs_wait_on_page_writeback(page, DATA, true, true);
 660         zero_user(page, offset, PAGE_SIZE - offset);
 661
 662         /* An encrypted inode should have a key and truncate the last page. */
 663         f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
 664         if (!cache_only)
 665                 set_page_dirty(page);
 666         f2fs_put_page(page, 1);
 667         return 0;
 668 }
 669
 670 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
 671 {
 672         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 673         struct dnode_of_data dn;
 674         pgoff_t free_from;
 675         int count = 0, err = 0;
 676         struct page *ipage;
 677         bool truncate_page = false;
 678
 679         trace_f2fs_truncate_blocks_enter(inode, from);
 680
 681         free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 682
 683         if (free_from >= max_file_blocks(inode))
 684                 goto free_partial;
 685
 686         if (lock)
 687                 f2fs_lock_op(sbi);
 688
 689         ipage = f2fs_get_node_page(sbi, inode->i_ino);
 690         if (IS_ERR(ipage)) {
 691                 err = PTR_ERR(ipage);
 692                 goto out;
 693         }
 694
 695         if (f2fs_has_inline_data(inode)) {
 696                 f2fs_truncate_inline_inode(inode, ipage, from);
 697                 f2fs_put_page(ipage, 1);
 698                 truncate_page = true;
 699                 goto out;
 700         }
 701
 702         set_new_dnode(&dn, inode, ipage, NULL, 0);
 703         err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 704         if (err) {
 705                 if (err == -ENOENT)
 706                         goto free_next;
 707                 goto out;
 708         }
 709
 710         count = ADDRS_PER_PAGE(dn.node_page, inode);
 711
 712         count -= dn.ofs_in_node;
 713         f2fs_bug_on(sbi, count < 0);
 714
 715         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 716                 f2fs_truncate_data_blocks_range(&dn, count);
 717                 free_from += count;
 718         }
 719
 720         f2fs_put_dnode(&dn);
 721 free_next:
 722         err = f2fs_truncate_inode_blocks(inode, free_from);
 723 out:
 724         if (lock)
 725                 f2fs_unlock_op(sbi);
 726 free_partial:
 727         /* lastly zero out the first data page */
 728         if (!err)
 729                 err = truncate_partial_data_page(inode, from, truncate_page);
 730
 731         trace_f2fs_truncate_blocks_exit(inode, err);
 732         return err;
 733 }
 734
 735 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
 736 {
 737         u64 free_from = from;
 738         int err;
 739
 740 #ifdef CONFIG_F2FS_FS_COMPRESSION
 741         /*
 742          * for compressed file, only support cluster size
 743          * aligned truncation.
 744          */
 745         if (f2fs_compressed_file(inode))
 746                 free_from = round_up(from,
 747                                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
 748 #endif
 749
 750         err = f2fs_do_truncate_blocks(inode, free_from, lock);
 751         if (err)
 752                 return err;
 753
 754 #ifdef CONFIG_F2FS_FS_COMPRESSION
 755         /*
 756          * For compressed file, after release compress blocks, don't allow write
 757          * direct, but we should allow write direct after truncate to zero.
 758          */
 759         if (f2fs_compressed_file(inode) && !free_from
 760                         && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
 761                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 762
 763         if (from != free_from) {
 764                 err = f2fs_truncate_partial_cluster(inode, from, lock);
 765                 if (err)
 766                         return err;
 767         }
 768 #endif
 769
 770         return 0;
 771 }
 772
 773 int f2fs_truncate(struct inode *inode)
 774 {
 775         int err;
 776
 777         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 778                 return -EIO;
 779
 780         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 781                                 S_ISLNK(inode->i_mode)))
 782                 return 0;
 783
 784         trace_f2fs_truncate(inode);
 785
 786         if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
 787                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
 788                 return -EIO;
 789         }
 790
 791         err = f2fs_dquot_initialize(inode);
 792         if (err)
 793                 return err;
 794
 795         /* we should check inline_data size */
 796         if (!f2fs_may_inline_data(inode)) {
 797                 err = f2fs_convert_inline_inode(inode);
 798                 if (err)
 799                         return err;
 800         }
 801
 802         err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
 803         if (err)
 804                 return err;
 805
 806         inode->i_mtime = inode->i_ctime = current_time(inode);
 807         f2fs_mark_inode_dirty_sync(inode, false);
 808         return 0;
 809 }
 810
 811 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 812                  struct kstat *stat, u32 request_mask, unsigned int query_flags)
 813 {
 814         struct inode *inode = d_inode(path->dentry);
 815         struct f2fs_inode_info *fi = F2FS_I(inode);
 816         struct f2fs_inode *ri = NULL;
 817         unsigned int flags;
 818
 819         if (f2fs_has_extra_attr(inode) &&
 820                         f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
 821                         F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 822                 stat->result_mask |= STATX_BTIME;
 823                 stat->btime.tv_sec = fi->i_crtime.tv_sec;
 824                 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 825         }
 826
 827         flags = fi->i_flags;
 828         if (flags & F2FS_COMPR_FL)
 829                 stat->attributes |= STATX_ATTR_COMPRESSED;
 830         if (flags & F2FS_APPEND_FL)
 831                 stat->attributes |= STATX_ATTR_APPEND;
 832         if (IS_ENCRYPTED(inode))
 833                 stat->attributes |= STATX_ATTR_ENCRYPTED;
 834         if (flags & F2FS_IMMUTABLE_FL)
 835                 stat->attributes |= STATX_ATTR_IMMUTABLE;
 836         if (flags & F2FS_NODUMP_FL)
 837                 stat->attributes |= STATX_ATTR_NODUMP;
 838         if (IS_VERITY(inode))
 839                 stat->attributes |= STATX_ATTR_VERITY;
 840
 841         stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 842                                   STATX_ATTR_APPEND |
 843                                   STATX_ATTR_ENCRYPTED |
 844                                   STATX_ATTR_IMMUTABLE |
 845                                   STATX_ATTR_NODUMP |
 846                                   STATX_ATTR_VERITY);
 847
 848         generic_fillattr(mnt_userns, inode, stat);
 849
 850         /* we need to show initial sectors used for inline_data/dentries */
 851         if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 852                                         f2fs_has_inline_dentry(inode))
 853                 stat->blocks += (stat->size + 511) >> 9;
 854
 855         return 0;
 856 }
 857
 858 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 859 static void __setattr_copy(struct user_namespace *mnt_userns,
 860                            struct inode *inode, const struct iattr *attr)
 861 {
 862         unsigned int ia_valid = attr->ia_valid;
 863
 864         if (ia_valid & ATTR_UID)
 865                 inode->i_uid = attr->ia_uid;
 866         if (ia_valid & ATTR_GID)
 867                 inode->i_gid = attr->ia_gid;
 868         if (ia_valid & ATTR_ATIME)
 869                 inode->i_atime = attr->ia_atime;
 870         if (ia_valid & ATTR_MTIME)
 871                 inode->i_mtime = attr->ia_mtime;
 872         if (ia_valid & ATTR_CTIME)
 873                 inode->i_ctime = attr->ia_ctime;
 874         if (ia_valid & ATTR_MODE) {
 875                 umode_t mode = attr->ia_mode;
 876                 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
 877
 878                 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
 879                         mode &= ~S_ISGID;
 880                 set_acl_inode(inode, mode);
 881         }
 882 }
 883 #else
 884 #define __setattr_copy setattr_copy
 885 #endif
 886
 887 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 888                  struct iattr *attr)
 889 {
 890         struct inode *inode = d_inode(dentry);
 891         int err;
 892
 893         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 894                 return -EIO;
 895
 896         if (unlikely(IS_IMMUTABLE(inode)))
 897                 return -EPERM;
 898
 899         if (unlikely(IS_APPEND(inode) &&
 900                         (attr->ia_valid & (ATTR_MODE | ATTR_UID |
 901                                   ATTR_GID | ATTR_TIMES_SET))))
 902                 return -EPERM;
 903
 904         if ((attr->ia_valid & ATTR_SIZE) &&
 905                 !f2fs_is_compress_backend_ready(inode))
 906                 return -EOPNOTSUPP;
 907
 908         err = setattr_prepare(mnt_userns, dentry, attr);
 909         if (err)
 910                 return err;
 911
 912         err = fscrypt_prepare_setattr(dentry, attr);
 913         if (err)
 914                 return err;
 915
 916         err = fsverity_prepare_setattr(dentry, attr);
 917         if (err)
 918                 return err;
 919
 920         if (is_quota_modification(inode, attr)) {
 921                 err = f2fs_dquot_initialize(inode);
 922                 if (err)
 923                         return err;
 924         }
 925         if ((attr->ia_valid & ATTR_UID &&
 926                 !uid_eq(attr->ia_uid, inode->i_uid)) ||
 927                 (attr->ia_valid & ATTR_GID &&
 928                 !gid_eq(attr->ia_gid, inode->i_gid))) {
 929                 f2fs_lock_op(F2FS_I_SB(inode));
 930                 err = dquot_transfer(inode, attr);
 931                 if (err) {
 932                         set_sbi_flag(F2FS_I_SB(inode),
 933                                         SBI_QUOTA_NEED_REPAIR);
 934                         f2fs_unlock_op(F2FS_I_SB(inode));
 935                         return err;
 936                 }
 937                 /*
 938                  * update uid/gid under lock_op(), so that dquot and inode can
 939                  * be updated atomically.
 940                  */
 941                 if (attr->ia_valid & ATTR_UID)
 942                         inode->i_uid = attr->ia_uid;
 943                 if (attr->ia_valid & ATTR_GID)
 944                         inode->i_gid = attr->ia_gid;
 945                 f2fs_mark_inode_dirty_sync(inode, true);
 946                 f2fs_unlock_op(F2FS_I_SB(inode));
 947         }
 948
 949         if (attr->ia_valid & ATTR_SIZE) {
 950                 loff_t old_size = i_size_read(inode);
 951
 952                 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
 953                         /*
 954                          * should convert inline inode before i_size_write to
 955                          * keep smaller than inline_data size with inline flag.
 956                          */
 957                         err = f2fs_convert_inline_inode(inode);
 958                         if (err)
 959                                 return err;
 960                 }
 961
 962                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 963                 filemap_invalidate_lock(inode->i_mapping);
 964
 965                 truncate_setsize(inode, attr->ia_size);
 966
 967                 if (attr->ia_size <= old_size)
 968                         err = f2fs_truncate(inode);
 969                 /*
 970                  * do not trim all blocks after i_size if target size is
 971                  * larger than i_size.
 972                  */
 973                 filemap_invalidate_unlock(inode->i_mapping);
 974                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 975                 if (err)
 976                         return err;
 977
 978                 spin_lock(&F2FS_I(inode)->i_size_lock);
 979                 inode->i_mtime = inode->i_ctime = current_time(inode);
 980                 F2FS_I(inode)->last_disk_size = i_size_read(inode);
 981                 spin_unlock(&F2FS_I(inode)->i_size_lock);
 982         }
 983
 984         __setattr_copy(mnt_userns, inode, attr);
 985
 986         if (attr->ia_valid & ATTR_MODE) {
 987                 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
 988
 989                 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
 990                         if (!err)
 991                                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
 992                         clear_inode_flag(inode, FI_ACL_MODE);
 993                 }
 994         }
 995
 996         /* file size may changed here */
 997         f2fs_mark_inode_dirty_sync(inode, true);
 998
 999         /* inode change will produce dirty node pages flushed by checkpoint */
1000         f2fs_balance_fs(F2FS_I_SB(inode), true);
1001
1002         return err;
1003 }
1004
1005 const struct inode_operations f2fs_file_inode_operations = {
1006         .getattr        = f2fs_getattr,
1007         .setattr        = f2fs_setattr,
1008         .get_acl        = f2fs_get_acl,
1009         .set_acl        = f2fs_set_acl,
1010         .listxattr      = f2fs_listxattr,
1011         .fiemap         = f2fs_fiemap,
1012         .fileattr_get   = f2fs_fileattr_get,
1013         .fileattr_set   = f2fs_fileattr_set,
1014 };
1015
1016 static int fill_zero(struct inode *inode, pgoff_t index,
1017                                         loff_t start, loff_t len)
1018 {
1019         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1020         struct page *page;
1021
1022         if (!len)
1023                 return 0;
1024
1025         f2fs_balance_fs(sbi, true);
1026
1027         f2fs_lock_op(sbi);
1028         page = f2fs_get_new_data_page(inode, NULL, index, false);
1029         f2fs_unlock_op(sbi);
1030
1031         if (IS_ERR(page))
1032                 return PTR_ERR(page);
1033
1034         f2fs_wait_on_page_writeback(page, DATA, true, true);
1035         zero_user(page, start, len);
1036         set_page_dirty(page);
1037         f2fs_put_page(page, 1);
1038         return 0;
1039 }
1040
1041 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1042 {
1043         int err;
1044
1045         while (pg_start < pg_end) {
1046                 struct dnode_of_data dn;
1047                 pgoff_t end_offset, count;
1048
1049                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1050                 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1051                 if (err) {
1052                         if (err == -ENOENT) {
1053                                 pg_start = f2fs_get_next_page_offset(&dn,
1054                                                                 pg_start);
1055                                 continue;
1056                         }
1057                         return err;
1058                 }
1059
1060                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1061                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1062
1063                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1064
1065                 f2fs_truncate_data_blocks_range(&dn, count);
1066                 f2fs_put_dnode(&dn);
1067
1068                 pg_start += count;
1069         }
1070         return 0;
1071 }
1072
1073 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1074 {
1075         pgoff_t pg_start, pg_end;
1076         loff_t off_start, off_end;
1077         int ret;
1078
1079         ret = f2fs_convert_inline_inode(inode);
1080         if (ret)
1081                 return ret;
1082
1083         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1084         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1085
1086         off_start = offset & (PAGE_SIZE - 1);
1087         off_end = (offset + len) & (PAGE_SIZE - 1);
1088
1089         if (pg_start == pg_end) {
1090                 ret = fill_zero(inode, pg_start, off_start,
1091                                                 off_end - off_start);
1092                 if (ret)
1093                         return ret;
1094         } else {
1095                 if (off_start) {
1096                         ret = fill_zero(inode, pg_start++, off_start,
1097                                                 PAGE_SIZE - off_start);
1098                         if (ret)
1099                                 return ret;
1100                 }
1101                 if (off_end) {
1102                         ret = fill_zero(inode, pg_end, 0, off_end);
1103                         if (ret)
1104                                 return ret;
1105                 }
1106
1107                 if (pg_start < pg_end) {
1108                         loff_t blk_start, blk_end;
1109                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1110
1111                         f2fs_balance_fs(sbi, true);
1112
1113                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
1114                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
1115
1116                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1117                         filemap_invalidate_lock(inode->i_mapping);
1118
1119                         truncate_pagecache_range(inode, blk_start, blk_end - 1);
1120
1121                         f2fs_lock_op(sbi);
1122                         ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1123                         f2fs_unlock_op(sbi);
1124
1125                         filemap_invalidate_unlock(inode->i_mapping);
1126                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1127                 }
1128         }
1129
1130         return ret;
1131 }
1132
1133 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1134                                 int *do_replace, pgoff_t off, pgoff_t len)
1135 {
1136         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1137         struct dnode_of_data dn;
1138         int ret, done, i;
1139
1140 next_dnode:
1141         set_new_dnode(&dn, inode, NULL, NULL, 0);
1142         ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1143         if (ret && ret != -ENOENT) {
1144                 return ret;
1145         } else if (ret == -ENOENT) {
1146                 if (dn.max_level == 0)
1147                         return -ENOENT;
1148                 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1149                                                 dn.ofs_in_node, len);
1150                 blkaddr += done;
1151                 do_replace += done;
1152                 goto next;
1153         }
1154
1155         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1156                                                         dn.ofs_in_node, len);
1157         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1158                 *blkaddr = f2fs_data_blkaddr(&dn);
1159
1160                 if (__is_valid_data_blkaddr(*blkaddr) &&
1161                         !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1162                                         DATA_GENERIC_ENHANCE)) {
1163                         f2fs_put_dnode(&dn);
1164                         return -EFSCORRUPTED;
1165                 }
1166
1167                 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1168
1169                         if (f2fs_lfs_mode(sbi)) {
1170                                 f2fs_put_dnode(&dn);
1171                                 return -EOPNOTSUPP;
1172                         }
1173
1174                         /* do not invalidate this block address */
1175                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1176                         *do_replace = 1;
1177                 }
1178         }
1179         f2fs_put_dnode(&dn);
1180 next:
1181         len -= done;
1182         off += done;
1183         if (len)
1184                 goto next_dnode;
1185         return 0;
1186 }
1187
1188 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1189                                 int *do_replace, pgoff_t off, int len)
1190 {
1191         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1192         struct dnode_of_data dn;
1193         int ret, i;
1194
1195         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1196                 if (*do_replace == 0)
1197                         continue;
1198
1199                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1200                 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1201                 if (ret) {
1202                         dec_valid_block_count(sbi, inode, 1);
1203                         f2fs_invalidate_blocks(sbi, *blkaddr);
1204                 } else {
1205                         f2fs_update_data_blkaddr(&dn, *blkaddr);
1206                 }
1207                 f2fs_put_dnode(&dn);
1208         }
1209         return 0;
1210 }
1211
1212 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1213                         block_t *blkaddr, int *do_replace,
1214                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1215 {
1216         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1217         pgoff_t i = 0;
1218         int ret;
1219
1220         while (i < len) {
1221                 if (blkaddr[i] == NULL_ADDR && !full) {
1222                         i++;
1223                         continue;
1224                 }
1225
1226                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1227                         struct dnode_of_data dn;
1228                         struct node_info ni;
1229                         size_t new_size;
1230                         pgoff_t ilen;
1231
1232                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1233                         ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1234                         if (ret)
1235                                 return ret;
1236
1237                         ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1238                         if (ret) {
1239                                 f2fs_put_dnode(&dn);
1240                                 return ret;
1241                         }
1242
1243                         ilen = min((pgoff_t)
1244                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1245                                                 dn.ofs_in_node, len - i);
1246                         do {
1247                                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1248                                 f2fs_truncate_data_blocks_range(&dn, 1);
1249
1250                                 if (do_replace[i]) {
1251                                         f2fs_i_blocks_write(src_inode,
1252                                                         1, false, false);
1253                                         f2fs_i_blocks_write(dst_inode,
1254                                                         1, true, false);
1255                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1256                                         blkaddr[i], ni.version, true, false);
1257
1258                                         do_replace[i] = 0;
1259                                 }
1260                                 dn.ofs_in_node++;
1261                                 i++;
1262                                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1263                                 if (dst_inode->i_size < new_size)
1264                                         f2fs_i_size_write(dst_inode, new_size);
1265                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1266
1267                         f2fs_put_dnode(&dn);
1268                 } else {
1269                         struct page *psrc, *pdst;
1270
1271                         psrc = f2fs_get_lock_data_page(src_inode,
1272                                                         src + i, true);
1273                         if (IS_ERR(psrc))
1274                                 return PTR_ERR(psrc);
1275                         pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1276                                                                 true);
1277                         if (IS_ERR(pdst)) {
1278                                 f2fs_put_page(psrc, 1);
1279                                 return PTR_ERR(pdst);
1280                         }
1281                         f2fs_copy_page(psrc, pdst);
1282                         set_page_dirty(pdst);
1283                         f2fs_put_page(pdst, 1);
1284                         f2fs_put_page(psrc, 1);
1285
1286                         ret = f2fs_truncate_hole(src_inode,
1287                                                 src + i, src + i + 1);
1288                         if (ret)
1289                                 return ret;
1290                         i++;
1291                 }
1292         }
1293         return 0;
1294 }
1295
1296 static int __exchange_data_block(struct inode *src_inode,
1297                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1298                         pgoff_t len, bool full)
1299 {
1300         block_t *src_blkaddr;
1301         int *do_replace;
1302         pgoff_t olen;
1303         int ret;
1304
1305         while (len) {
1306                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1307
1308                 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1309                                         array_size(olen, sizeof(block_t)),
1310                                         GFP_NOFS);
1311                 if (!src_blkaddr)
1312                         return -ENOMEM;
1313
1314                 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1315                                         array_size(olen, sizeof(int)),
1316                                         GFP_NOFS);
1317                 if (!do_replace) {
1318                         kvfree(src_blkaddr);
1319                         return -ENOMEM;
1320                 }
1321
1322                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1323                                         do_replace, src, olen);
1324                 if (ret)
1325                         goto roll_back;
1326
1327                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1328                                         do_replace, src, dst, olen, full);
1329                 if (ret)
1330                         goto roll_back;
1331
1332                 src += olen;
1333                 dst += olen;
1334                 len -= olen;
1335
1336                 kvfree(src_blkaddr);
1337                 kvfree(do_replace);
1338         }
1339         return 0;
1340
1341 roll_back:
1342         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1343         kvfree(src_blkaddr);
1344         kvfree(do_replace);
1345         return ret;
1346 }
1347
1348 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1349 {
1350         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1351         pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1352         pgoff_t start = offset >> PAGE_SHIFT;
1353         pgoff_t end = (offset + len) >> PAGE_SHIFT;
1354         int ret;
1355
1356         f2fs_balance_fs(sbi, true);
1357
1358         /* avoid gc operation during block exchange */
1359         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1360         filemap_invalidate_lock(inode->i_mapping);
1361
1362         f2fs_lock_op(sbi);
1363         f2fs_drop_extent_tree(inode);
1364         truncate_pagecache(inode, offset);
1365         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1366         f2fs_unlock_op(sbi);
1367
1368         filemap_invalidate_unlock(inode->i_mapping);
1369         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1370         return ret;
1371 }
1372
1373 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1374 {
1375         loff_t new_size;
1376         int ret;
1377
1378         if (offset + len >= i_size_read(inode))
1379                 return -EINVAL;
1380
1381         /* collapse range should be aligned to block size of f2fs. */
1382         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1383                 return -EINVAL;
1384
1385         ret = f2fs_convert_inline_inode(inode);
1386         if (ret)
1387                 return ret;
1388
1389         /* write out all dirty pages from offset */
1390         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1391         if (ret)
1392                 return ret;
1393
1394         ret = f2fs_do_collapse(inode, offset, len);
1395         if (ret)
1396                 return ret;
1397
1398         /* write out all moved pages, if possible */
1399         filemap_invalidate_lock(inode->i_mapping);
1400         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1401         truncate_pagecache(inode, offset);
1402
1403         new_size = i_size_read(inode) - len;
1404         ret = f2fs_truncate_blocks(inode, new_size, true);
1405         filemap_invalidate_unlock(inode->i_mapping);
1406         if (!ret)
1407                 f2fs_i_size_write(inode, new_size);
1408         return ret;
1409 }
1410
1411 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1412                                                                 pgoff_t end)
1413 {
1414         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1415         pgoff_t index = start;
1416         unsigned int ofs_in_node = dn->ofs_in_node;
1417         blkcnt_t count = 0;
1418         int ret;
1419
1420         for (; index < end; index++, dn->ofs_in_node++) {
1421                 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1422                         count++;
1423         }
1424
1425         dn->ofs_in_node = ofs_in_node;
1426         ret = f2fs_reserve_new_blocks(dn, count);
1427         if (ret)
1428                 return ret;
1429
1430         dn->ofs_in_node = ofs_in_node;
1431         for (index = start; index < end; index++, dn->ofs_in_node++) {
1432                 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1433                 /*
1434                  * f2fs_reserve_new_blocks will not guarantee entire block
1435                  * allocation.
1436                  */
1437                 if (dn->data_blkaddr == NULL_ADDR) {
1438                         ret = -ENOSPC;
1439                         break;
1440                 }
1441
1442                 if (dn->data_blkaddr == NEW_ADDR)
1443                         continue;
1444
1445                 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1446                                         DATA_GENERIC_ENHANCE)) {
1447                         ret = -EFSCORRUPTED;
1448                         break;
1449                 }
1450
1451                 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1452                 dn->data_blkaddr = NEW_ADDR;
1453                 f2fs_set_data_blkaddr(dn);
1454         }
1455
1456         f2fs_update_extent_cache_range(dn, start, 0, index - start);
1457
1458         return ret;
1459 }
1460
1461 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1462                                                                 int mode)
1463 {
1464         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1465         struct address_space *mapping = inode->i_mapping;
1466         pgoff_t index, pg_start, pg_end;
1467         loff_t new_size = i_size_read(inode);
1468         loff_t off_start, off_end;
1469         int ret = 0;
1470
1471         ret = inode_newsize_ok(inode, (len + offset));
1472         if (ret)
1473                 return ret;
1474
1475         ret = f2fs_convert_inline_inode(inode);
1476         if (ret)
1477                 return ret;
1478
1479         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1480         if (ret)
1481                 return ret;
1482
1483         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1484         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1485
1486         off_start = offset & (PAGE_SIZE - 1);
1487         off_end = (offset + len) & (PAGE_SIZE - 1);
1488
1489         if (pg_start == pg_end) {
1490                 ret = fill_zero(inode, pg_start, off_start,
1491                                                 off_end - off_start);
1492                 if (ret)
1493                         return ret;
1494
1495                 new_size = max_t(loff_t, new_size, offset + len);
1496         } else {
1497                 if (off_start) {
1498                         ret = fill_zero(inode, pg_start++, off_start,
1499                                                 PAGE_SIZE - off_start);
1500                         if (ret)
1501                                 return ret;
1502
1503                         new_size = max_t(loff_t, new_size,
1504                                         (loff_t)pg_start << PAGE_SHIFT);
1505                 }
1506
1507                 for (index = pg_start; index < pg_end;) {
1508                         struct dnode_of_data dn;
1509                         unsigned int end_offset;
1510                         pgoff_t end;
1511
1512                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1513                         filemap_invalidate_lock(mapping);
1514
1515                         truncate_pagecache_range(inode,
1516                                 (loff_t)index << PAGE_SHIFT,
1517                                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1518
1519                         f2fs_lock_op(sbi);
1520
1521                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1522                         ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1523                         if (ret) {
1524                                 f2fs_unlock_op(sbi);
1525                                 filemap_invalidate_unlock(mapping);
1526                                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1527                                 goto out;
1528                         }
1529
1530                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1531                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1532
1533                         ret = f2fs_do_zero_range(&dn, index, end);
1534                         f2fs_put_dnode(&dn);
1535
1536                         f2fs_unlock_op(sbi);
1537                         filemap_invalidate_unlock(mapping);
1538                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1539
1540                         f2fs_balance_fs(sbi, dn.node_changed);
1541
1542                         if (ret)
1543                                 goto out;
1544
1545                         index = end;
1546                         new_size = max_t(loff_t, new_size,
1547                                         (loff_t)index << PAGE_SHIFT);
1548                 }
1549
1550                 if (off_end) {
1551                         ret = fill_zero(inode, pg_end, 0, off_end);
1552                         if (ret)
1553                                 goto out;
1554
1555                         new_size = max_t(loff_t, new_size, offset + len);
1556                 }
1557         }
1558
1559 out:
1560         if (new_size > i_size_read(inode)) {
1561                 if (mode & FALLOC_FL_KEEP_SIZE)
1562                         file_set_keep_isize(inode);
1563                 else
1564                         f2fs_i_size_write(inode, new_size);
1565         }
1566         return ret;
1567 }
1568
1569 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1570 {
1571         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1572         struct address_space *mapping = inode->i_mapping;
1573         pgoff_t nr, pg_start, pg_end, delta, idx;
1574         loff_t new_size;
1575         int ret = 0;
1576
1577         new_size = i_size_read(inode) + len;
1578         ret = inode_newsize_ok(inode, new_size);
1579         if (ret)
1580                 return ret;
1581
1582         if (offset >= i_size_read(inode))
1583                 return -EINVAL;
1584
1585         /* insert range should be aligned to block size of f2fs. */
1586         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1587                 return -EINVAL;
1588
1589         ret = f2fs_convert_inline_inode(inode);
1590         if (ret)
1591                 return ret;
1592
1593         f2fs_balance_fs(sbi, true);
1594
1595         filemap_invalidate_lock(mapping);
1596         ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1597         filemap_invalidate_unlock(mapping);
1598         if (ret)
1599                 return ret;
1600
1601         /* write out all dirty pages from offset */
1602         ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1603         if (ret)
1604                 return ret;
1605
1606         pg_start = offset >> PAGE_SHIFT;
1607         pg_end = (offset + len) >> PAGE_SHIFT;
1608         delta = pg_end - pg_start;
1609         idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1610
1611         /* avoid gc operation during block exchange */
1612         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1613         filemap_invalidate_lock(mapping);
1614         truncate_pagecache(inode, offset);
1615
1616         while (!ret && idx > pg_start) {
1617                 nr = idx - pg_start;
1618                 if (nr > delta)
1619                         nr = delta;
1620                 idx -= nr;
1621
1622                 f2fs_lock_op(sbi);
1623                 f2fs_drop_extent_tree(inode);
1624
1625                 ret = __exchange_data_block(inode, inode, idx,
1626                                         idx + delta, nr, false);
1627                 f2fs_unlock_op(sbi);
1628         }
1629         filemap_invalidate_unlock(mapping);
1630         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1631
1632         /* write out all moved pages, if possible */
1633         filemap_invalidate_lock(mapping);
1634         filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1635         truncate_pagecache(inode, offset);
1636         filemap_invalidate_unlock(mapping);
1637
1638         if (!ret)
1639                 f2fs_i_size_write(inode, new_size);
1640         return ret;
1641 }
1642
1643 static int expand_inode_data(struct inode *inode, loff_t offset,
1644                                         loff_t len, int mode)
1645 {
1646         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1647         struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1648                         .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1649                         .m_may_create = true };
1650         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1651                         .init_gc_type = FG_GC,
1652                         .should_migrate_blocks = false,
1653                         .err_gc_skipped = true,
1654                         .nr_free_secs = 0 };
1655         pgoff_t pg_start, pg_end;
1656         loff_t new_size = i_size_read(inode);
1657         loff_t off_end;
1658         block_t expanded = 0;
1659         int err;
1660
1661         err = inode_newsize_ok(inode, (len + offset));
1662         if (err)
1663                 return err;
1664
1665         err = f2fs_convert_inline_inode(inode);
1666         if (err)
1667                 return err;
1668
1669         f2fs_balance_fs(sbi, true);
1670
1671         pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1672         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1673         off_end = (offset + len) & (PAGE_SIZE - 1);
1674
1675         map.m_lblk = pg_start;
1676         map.m_len = pg_end - pg_start;
1677         if (off_end)
1678                 map.m_len++;
1679
1680         if (!map.m_len)
1681                 return 0;
1682
1683         if (f2fs_is_pinned_file(inode)) {
1684                 block_t sec_blks = BLKS_PER_SEC(sbi);
1685                 block_t sec_len = roundup(map.m_len, sec_blks);
1686
1687                 map.m_len = sec_blks;
1688 next_alloc:
1689                 if (has_not_enough_free_secs(sbi, 0,
1690                         GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1691                         f2fs_down_write(&sbi->gc_lock);
1692                         err = f2fs_gc(sbi, &gc_control);
1693                         if (err && err != -ENODATA)
1694                                 goto out_err;
1695                 }
1696
1697                 f2fs_down_write(&sbi->pin_sem);
1698
1699                 f2fs_lock_op(sbi);
1700                 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1701                 f2fs_unlock_op(sbi);
1702
1703                 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1704                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1705                 file_dont_truncate(inode);
1706
1707                 f2fs_up_write(&sbi->pin_sem);
1708
1709                 expanded += map.m_len;
1710                 sec_len -= map.m_len;
1711                 map.m_lblk += map.m_len;
1712                 if (!err && sec_len)
1713                         goto next_alloc;
1714
1715                 map.m_len = expanded;
1716         } else {
1717                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1718                 expanded = map.m_len;
1719         }
1720 out_err:
1721         if (err) {
1722                 pgoff_t last_off;
1723
1724                 if (!expanded)
1725                         return err;
1726
1727                 last_off = pg_start + expanded - 1;
1728
1729                 /* update new size to the failed position */
1730                 new_size = (last_off == pg_end) ? offset + len :
1731                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1732         } else {
1733                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1734         }
1735
1736         if (new_size > i_size_read(inode)) {
1737                 if (mode & FALLOC_FL_KEEP_SIZE)
1738                         file_set_keep_isize(inode);
1739                 else
1740                         f2fs_i_size_write(inode, new_size);
1741         }
1742
1743         return err;
1744 }
1745
1746 static long f2fs_fallocate(struct file *file, int mode,
1747                                 loff_t offset, loff_t len)
1748 {
1749         struct inode *inode = file_inode(file);
1750         long ret = 0;
1751
1752         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1753                 return -EIO;
1754         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1755                 return -ENOSPC;
1756         if (!f2fs_is_compress_backend_ready(inode))
1757                 return -EOPNOTSUPP;
1758
1759         /* f2fs only support ->fallocate for regular file */
1760         if (!S_ISREG(inode->i_mode))
1761                 return -EINVAL;
1762
1763         if (IS_ENCRYPTED(inode) &&
1764                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1765                 return -EOPNOTSUPP;
1766
1767         /*
1768          * Pinned file should not support partial trucation since the block
1769          * can be used by applications.
1770          */
1771         if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1772                 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1773                         FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1774                 return -EOPNOTSUPP;
1775
1776         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1777                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1778                         FALLOC_FL_INSERT_RANGE))
1779                 return -EOPNOTSUPP;
1780
1781         inode_lock(inode);
1782
1783         ret = file_modified(file);
1784         if (ret)
1785                 goto out;
1786
1787         if (mode & FALLOC_FL_PUNCH_HOLE) {
1788                 if (offset >= inode->i_size)
1789                         goto out;
1790
1791                 ret = punch_hole(inode, offset, len);
1792         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1793                 ret = f2fs_collapse_range(inode, offset, len);
1794         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1795                 ret = f2fs_zero_range(inode, offset, len, mode);
1796         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1797                 ret = f2fs_insert_range(inode, offset, len);
1798         } else {
1799                 ret = expand_inode_data(inode, offset, len, mode);
1800         }
1801
1802         if (!ret) {
1803                 inode->i_mtime = inode->i_ctime = current_time(inode);
1804                 f2fs_mark_inode_dirty_sync(inode, false);
1805                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1806         }
1807
1808 out:
1809         inode_unlock(inode);
1810
1811         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1812         return ret;
1813 }
1814
1815 static int f2fs_release_file(struct inode *inode, struct file *filp)
1816 {
1817         /*
1818          * f2fs_relase_file is called at every close calls. So we should
1819          * not drop any inmemory pages by close called by other process.
1820          */
1821         if (!(filp->f_mode & FMODE_WRITE) ||
1822                         atomic_read(&inode->i_writecount) != 1)
1823                 return 0;
1824
1825         if (f2fs_is_atomic_file(inode))
1826                 f2fs_abort_atomic_write(inode, true);
1827         return 0;
1828 }
1829
1830 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1831 {
1832         struct inode *inode = file_inode(file);
1833
1834         /*
1835          * If the process doing a transaction is crashed, we should do
1836          * roll-back. Otherwise, other reader/write can see corrupted database
1837          * until all the writers close its file. Since this should be done
1838          * before dropping file lock, it needs to do in ->flush.
1839          */
1840         if (f2fs_is_atomic_file(inode) &&
1841                         F2FS_I(inode)->atomic_write_task == current)
1842                 f2fs_abort_atomic_write(inode, true);
1843         return 0;
1844 }
1845
1846 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1847 {
1848         struct f2fs_inode_info *fi = F2FS_I(inode);
1849         u32 masked_flags = fi->i_flags & mask;
1850
1851         /* mask can be shrunk by flags_valid selector */
1852         iflags &= mask;
1853
1854         /* Is it quota file? Do not allow user to mess with it */
1855         if (IS_NOQUOTA(inode))
1856                 return -EPERM;
1857
1858         if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1859                 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1860                         return -EOPNOTSUPP;
1861                 if (!f2fs_empty_dir(inode))
1862                         return -ENOTEMPTY;
1863         }
1864
1865         if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1866                 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1867                         return -EOPNOTSUPP;
1868                 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1869                         return -EINVAL;
1870         }
1871
1872         if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1873                 if (masked_flags & F2FS_COMPR_FL) {
1874                         if (!f2fs_disable_compressed_file(inode))
1875                                 return -EINVAL;
1876                 }
1877                 if (iflags & F2FS_NOCOMP_FL)
1878                         return -EINVAL;
1879                 if (iflags & F2FS_COMPR_FL) {
1880                         if (!f2fs_may_compress(inode))
1881                                 return -EINVAL;
1882                         if (S_ISREG(inode->i_mode) && inode->i_size)
1883                                 return -EINVAL;
1884
1885                         set_compress_context(inode);
1886                 }
1887         }
1888         if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1889                 if (masked_flags & F2FS_COMPR_FL)
1890                         return -EINVAL;
1891         }
1892
1893         fi->i_flags = iflags | (fi->i_flags & ~mask);
1894         f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1895                                         (fi->i_flags & F2FS_NOCOMP_FL));
1896
1897         if (fi->i_flags & F2FS_PROJINHERIT_FL)
1898                 set_inode_flag(inode, FI_PROJ_INHERIT);
1899         else
1900                 clear_inode_flag(inode, FI_PROJ_INHERIT);
1901
1902         inode->i_ctime = current_time(inode);
1903         f2fs_set_inode_flags(inode);
1904         f2fs_mark_inode_dirty_sync(inode, true);
1905         return 0;
1906 }
1907
1908 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1909
1910 /*
1911  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1912  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1913  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1914  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1915  *
1916  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1917  * FS_IOC_FSSETXATTR is done by the VFS.
1918  */
1919
1920 static const struct {
1921         u32 iflag;
1922         u32 fsflag;
1923 } f2fs_fsflags_map[] = {
1924         { F2FS_COMPR_FL,        FS_COMPR_FL },
1925         { F2FS_SYNC_FL,         FS_SYNC_FL },
1926         { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1927         { F2FS_APPEND_FL,       FS_APPEND_FL },
1928         { F2FS_NODUMP_FL,       FS_NODUMP_FL },
1929         { F2FS_NOATIME_FL,      FS_NOATIME_FL },
1930         { F2FS_NOCOMP_FL,       FS_NOCOMP_FL },
1931         { F2FS_INDEX_FL,        FS_INDEX_FL },
1932         { F2FS_DIRSYNC_FL,      FS_DIRSYNC_FL },
1933         { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1934         { F2FS_CASEFOLD_FL,     FS_CASEFOLD_FL },
1935 };
1936
1937 #define F2FS_GETTABLE_FS_FL (           \
1938                 FS_COMPR_FL |           \
1939                 FS_SYNC_FL |            \
1940                 FS_IMMUTABLE_FL |       \
1941                 FS_APPEND_FL |          \
1942                 FS_NODUMP_FL |          \
1943                 FS_NOATIME_FL |         \
1944                 FS_NOCOMP_FL |          \
1945                 FS_INDEX_FL |           \
1946                 FS_DIRSYNC_FL |         \
1947                 FS_PROJINHERIT_FL |     \
1948                 FS_ENCRYPT_FL |         \
1949                 FS_INLINE_DATA_FL |     \
1950                 FS_NOCOW_FL |           \
1951                 FS_VERITY_FL |          \
1952                 FS_CASEFOLD_FL)
1953
1954 #define F2FS_SETTABLE_FS_FL (           \
1955                 FS_COMPR_FL |           \
1956                 FS_SYNC_FL |            \
1957                 FS_IMMUTABLE_FL |       \
1958                 FS_APPEND_FL |          \
1959                 FS_NODUMP_FL |          \
1960                 FS_NOATIME_FL |         \
1961                 FS_NOCOMP_FL |          \
1962                 FS_DIRSYNC_FL |         \
1963                 FS_PROJINHERIT_FL |     \
1964                 FS_CASEFOLD_FL)
1965
1966 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1967 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1968 {
1969         u32 fsflags = 0;
1970         int i;
1971
1972         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1973                 if (iflags & f2fs_fsflags_map[i].iflag)
1974                         fsflags |= f2fs_fsflags_map[i].fsflag;
1975
1976         return fsflags;
1977 }
1978
1979 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1980 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1981 {
1982         u32 iflags = 0;
1983         int i;
1984
1985         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1986                 if (fsflags & f2fs_fsflags_map[i].fsflag)
1987                         iflags |= f2fs_fsflags_map[i].iflag;
1988
1989         return iflags;
1990 }
1991
1992 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1993 {
1994         struct inode *inode = file_inode(filp);
1995
1996         return put_user(inode->i_generation, (int __user *)arg);
1997 }
1998
1999 static int f2fs_ioc_start_atomic_write(struct file *filp)
2000 {
2001         struct inode *inode = file_inode(filp);
2002         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2003         struct f2fs_inode_info *fi = F2FS_I(inode);
2004         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2005         struct inode *pinode;
2006         int ret;
2007
2008         if (!inode_owner_or_capable(mnt_userns, inode))
2009                 return -EACCES;
2010
2011         if (!S_ISREG(inode->i_mode))
2012                 return -EINVAL;
2013
2014         if (filp->f_flags & O_DIRECT)
2015                 return -EINVAL;
2016
2017         ret = mnt_want_write_file(filp);
2018         if (ret)
2019                 return ret;
2020
2021         inode_lock(inode);
2022
2023         if (!f2fs_disable_compressed_file(inode)) {
2024                 ret = -EINVAL;
2025                 goto out;
2026         }
2027
2028         if (f2fs_is_atomic_file(inode))
2029                 goto out;
2030
2031         ret = f2fs_convert_inline_inode(inode);
2032         if (ret)
2033                 goto out;
2034
2035         f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2036
2037         /*
2038          * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2039          * f2fs_is_atomic_file.
2040          */
2041         if (get_dirty_pages(inode))
2042                 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2043                           inode->i_ino, get_dirty_pages(inode));
2044         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2045         if (ret) {
2046                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2047                 goto out;
2048         }
2049
2050         /* Create a COW inode for atomic write */
2051         pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2052         if (IS_ERR(pinode)) {
2053                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2054                 ret = PTR_ERR(pinode);
2055                 goto out;
2056         }
2057
2058         ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2059         iput(pinode);
2060         if (ret) {
2061                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2062                 goto out;
2063         }
2064         f2fs_i_size_write(fi->cow_inode, i_size_read(inode));
2065
2066         spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2067         sbi->atomic_files++;
2068         spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2069
2070         set_inode_flag(inode, FI_ATOMIC_FILE);
2071         set_inode_flag(fi->cow_inode, FI_ATOMIC_FILE);
2072         clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2073         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2074
2075         f2fs_update_time(sbi, REQ_TIME);
2076         fi->atomic_write_task = current;
2077         stat_update_max_atomic_write(inode);
2078 out:
2079         inode_unlock(inode);
2080         mnt_drop_write_file(filp);
2081         return ret;
2082 }
2083
2084 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2085 {
2086         struct inode *inode = file_inode(filp);
2087         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2088         int ret;
2089
2090         if (!inode_owner_or_capable(mnt_userns, inode))
2091                 return -EACCES;
2092
2093         ret = mnt_want_write_file(filp);
2094         if (ret)
2095                 return ret;
2096
2097         f2fs_balance_fs(F2FS_I_SB(inode), true);
2098
2099         inode_lock(inode);
2100
2101         if (f2fs_is_atomic_file(inode)) {
2102                 ret = f2fs_commit_atomic_write(inode);
2103                 if (ret)
2104                         goto unlock_out;
2105
2106                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2107                 if (!ret)
2108                         f2fs_abort_atomic_write(inode, false);
2109         } else {
2110                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2111         }
2112 unlock_out:
2113         inode_unlock(inode);
2114         mnt_drop_write_file(filp);
2115         return ret;
2116 }
2117
2118 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2119 {
2120         struct inode *inode = file_inode(filp);
2121         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2122         struct super_block *sb = sbi->sb;
2123         __u32 in;
2124         int ret = 0;
2125
2126         if (!capable(CAP_SYS_ADMIN))
2127                 return -EPERM;
2128
2129         if (get_user(in, (__u32 __user *)arg))
2130                 return -EFAULT;
2131
2132         if (in != F2FS_GOING_DOWN_FULLSYNC) {
2133                 ret = mnt_want_write_file(filp);
2134                 if (ret) {
2135                         if (ret == -EROFS) {
2136                                 ret = 0;
2137                                 f2fs_stop_checkpoint(sbi, false);
2138                                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2139                                 trace_f2fs_shutdown(sbi, in, ret);
2140                         }
2141                         return ret;
2142                 }
2143         }
2144
2145         switch (in) {
2146         case F2FS_GOING_DOWN_FULLSYNC:
2147                 ret = freeze_bdev(sb->s_bdev);
2148                 if (ret)
2149                         goto out;
2150                 f2fs_stop_checkpoint(sbi, false);
2151                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2152                 thaw_bdev(sb->s_bdev);
2153                 break;
2154         case F2FS_GOING_DOWN_METASYNC:
2155                 /* do checkpoint only */
2156                 ret = f2fs_sync_fs(sb, 1);
2157                 if (ret)
2158                         goto out;
2159                 f2fs_stop_checkpoint(sbi, false);
2160                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2161                 break;
2162         case F2FS_GOING_DOWN_NOSYNC:
2163                 f2fs_stop_checkpoint(sbi, false);
2164                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2165                 break;
2166         case F2FS_GOING_DOWN_METAFLUSH:
2167                 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2168                 f2fs_stop_checkpoint(sbi, false);
2169                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2170                 break;
2171         case F2FS_GOING_DOWN_NEED_FSCK:
2172                 set_sbi_flag(sbi, SBI_NEED_FSCK);
2173                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2174                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2175                 /* do checkpoint only */
2176                 ret = f2fs_sync_fs(sb, 1);
2177                 goto out;
2178         default:
2179                 ret = -EINVAL;
2180                 goto out;
2181         }
2182
2183         f2fs_stop_gc_thread(sbi);
2184         f2fs_stop_discard_thread(sbi);
2185
2186         f2fs_drop_discard_cmd(sbi);
2187         clear_opt(sbi, DISCARD);
2188
2189         f2fs_update_time(sbi, REQ_TIME);
2190 out:
2191         if (in != F2FS_GOING_DOWN_FULLSYNC)
2192                 mnt_drop_write_file(filp);
2193
2194         trace_f2fs_shutdown(sbi, in, ret);
2195
2196         return ret;
2197 }
2198
2199 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2200 {
2201         struct inode *inode = file_inode(filp);
2202         struct super_block *sb = inode->i_sb;
2203         struct fstrim_range range;
2204         int ret;
2205
2206         if (!capable(CAP_SYS_ADMIN))
2207                 return -EPERM;
2208
2209         if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2210                 return -EOPNOTSUPP;
2211
2212         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2213                                 sizeof(range)))
2214                 return -EFAULT;
2215
2216         ret = mnt_want_write_file(filp);
2217         if (ret)
2218                 return ret;
2219
2220         range.minlen = max((unsigned int)range.minlen,
2221                            bdev_discard_granularity(sb->s_bdev));
2222         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2223         mnt_drop_write_file(filp);
2224         if (ret < 0)
2225                 return ret;
2226
2227         if (copy_to_user((struct fstrim_range __user *)arg, &range,
2228                                 sizeof(range)))
2229                 return -EFAULT;
2230         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2231         return 0;
2232 }
2233
2234 static bool uuid_is_nonzero(__u8 u[16])
2235 {
2236         int i;
2237
2238         for (i = 0; i < 16; i++)
2239                 if (u[i])
2240                         return true;
2241         return false;
2242 }
2243
2244 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2245 {
2246         struct inode *inode = file_inode(filp);
2247
2248         if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2249                 return -EOPNOTSUPP;
2250
2251         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2252
2253         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2254 }
2255
2256 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2257 {
2258         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2259                 return -EOPNOTSUPP;
2260         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2261 }
2262
2263 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2264 {
2265         struct inode *inode = file_inode(filp);
2266         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2267         int err;
2268
2269         if (!f2fs_sb_has_encrypt(sbi))
2270                 return -EOPNOTSUPP;
2271
2272         err = mnt_want_write_file(filp);
2273         if (err)
2274                 return err;
2275
2276         f2fs_down_write(&sbi->sb_lock);
2277
2278         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2279                 goto got_it;
2280
2281         /* update superblock with uuid */
2282         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2283
2284         err = f2fs_commit_super(sbi, false);
2285         if (err) {
2286                 /* undo new data */
2287                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2288                 goto out_err;
2289         }
2290 got_it:
2291         if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2292                                                                         16))
2293                 err = -EFAULT;
2294 out_err:
2295         f2fs_up_write(&sbi->sb_lock);
2296         mnt_drop_write_file(filp);
2297         return err;
2298 }
2299
2300 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2301                                              unsigned long arg)
2302 {
2303         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2304                 return -EOPNOTSUPP;
2305
2306         return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2307 }
2308
2309 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2310 {
2311         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2312                 return -EOPNOTSUPP;
2313
2314         return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2315 }
2316
2317 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2318 {
2319         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2320                 return -EOPNOTSUPP;
2321
2322         return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2323 }
2324
2325 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2326                                                     unsigned long arg)
2327 {
2328         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2329                 return -EOPNOTSUPP;
2330
2331         return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2332 }
2333
2334 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2335                                               unsigned long arg)
2336 {
2337         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2338                 return -EOPNOTSUPP;
2339
2340         return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2341 }
2342
2343 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2344 {
2345         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2346                 return -EOPNOTSUPP;
2347
2348         return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2349 }
2350
2351 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2352 {
2353         struct inode *inode = file_inode(filp);
2354         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2355         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2356                         .no_bg_gc = false,
2357                         .should_migrate_blocks = false,
2358                         .nr_free_secs = 0 };
2359         __u32 sync;
2360         int ret;
2361
2362         if (!capable(CAP_SYS_ADMIN))
2363                 return -EPERM;
2364
2365         if (get_user(sync, (__u32 __user *)arg))
2366                 return -EFAULT;
2367
2368         if (f2fs_readonly(sbi->sb))
2369                 return -EROFS;
2370
2371         ret = mnt_want_write_file(filp);
2372         if (ret)
2373                 return ret;
2374
2375         if (!sync) {
2376                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2377                         ret = -EBUSY;
2378                         goto out;
2379                 }
2380         } else {
2381                 f2fs_down_write(&sbi->gc_lock);
2382         }
2383
2384         gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2385         gc_control.err_gc_skipped = sync;
2386         ret = f2fs_gc(sbi, &gc_control);
2387 out:
2388         mnt_drop_write_file(filp);
2389         return ret;
2390 }
2391
2392 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2393 {
2394         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2395         struct f2fs_gc_control gc_control = {
2396                         .init_gc_type = range->sync ? FG_GC : BG_GC,
2397                         .no_bg_gc = false,
2398                         .should_migrate_blocks = false,
2399                         .err_gc_skipped = range->sync,
2400                         .nr_free_secs = 0 };
2401         u64 end;
2402         int ret;
2403
2404         if (!capable(CAP_SYS_ADMIN))
2405                 return -EPERM;
2406         if (f2fs_readonly(sbi->sb))
2407                 return -EROFS;
2408
2409         end = range->start + range->len;
2410         if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2411                                         end >= MAX_BLKADDR(sbi))
2412                 return -EINVAL;
2413
2414         ret = mnt_want_write_file(filp);
2415         if (ret)
2416                 return ret;
2417
2418 do_more:
2419         if (!range->sync) {
2420                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2421                         ret = -EBUSY;
2422                         goto out;
2423                 }
2424         } else {
2425                 f2fs_down_write(&sbi->gc_lock);
2426         }
2427
2428         gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2429         ret = f2fs_gc(sbi, &gc_control);
2430         if (ret) {
2431                 if (ret == -EBUSY)
2432                         ret = -EAGAIN;
2433                 goto out;
2434         }
2435         range->start += BLKS_PER_SEC(sbi);
2436         if (range->start <= end)
2437                 goto do_more;
2438 out:
2439         mnt_drop_write_file(filp);
2440         return ret;
2441 }
2442
2443 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2444 {
2445         struct f2fs_gc_range range;
2446
2447         if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2448                                                         sizeof(range)))
2449                 return -EFAULT;
2450         return __f2fs_ioc_gc_range(filp, &range);
2451 }
2452
2453 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2454 {
2455         struct inode *inode = file_inode(filp);
2456         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2457         int ret;
2458
2459         if (!capable(CAP_SYS_ADMIN))
2460                 return -EPERM;
2461
2462         if (f2fs_readonly(sbi->sb))
2463                 return -EROFS;
2464
2465         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2466                 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2467                 return -EINVAL;
2468         }
2469
2470         ret = mnt_want_write_file(filp);
2471         if (ret)
2472                 return ret;
2473
2474         ret = f2fs_sync_fs(sbi->sb, 1);
2475
2476         mnt_drop_write_file(filp);
2477         return ret;
2478 }
2479
2480 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2481                                         struct file *filp,
2482                                         struct f2fs_defragment *range)
2483 {
2484         struct inode *inode = file_inode(filp);
2485         struct f2fs_map_blocks map = { .m_next_extent = NULL,
2486                                         .m_seg_type = NO_CHECK_TYPE,
2487                                         .m_may_create = false };
2488         struct extent_info ei = {0, 0, 0};
2489         pgoff_t pg_start, pg_end, next_pgofs;
2490         unsigned int blk_per_seg = sbi->blocks_per_seg;
2491         unsigned int total = 0, sec_num;
2492         block_t blk_end = 0;
2493         bool fragmented = false;
2494         int err;
2495
2496         pg_start = range->start >> PAGE_SHIFT;
2497         pg_end = (range->start + range->len) >> PAGE_SHIFT;
2498
2499         f2fs_balance_fs(sbi, true);
2500
2501         inode_lock(inode);
2502
2503         /* if in-place-update policy is enabled, don't waste time here */
2504         set_inode_flag(inode, FI_OPU_WRITE);
2505         if (f2fs_should_update_inplace(inode, NULL)) {
2506                 err = -EINVAL;
2507                 goto out;
2508         }
2509
2510         /* writeback all dirty pages in the range */
2511         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2512                                                 range->start + range->len - 1);
2513         if (err)
2514                 goto out;
2515
2516         /*
2517          * lookup mapping info in extent cache, skip defragmenting if physical
2518          * block addresses are continuous.
2519          */
2520         if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2521                 if (ei.fofs + ei.len >= pg_end)
2522                         goto out;
2523         }
2524
2525         map.m_lblk = pg_start;
2526         map.m_next_pgofs = &next_pgofs;
2527
2528         /*
2529          * lookup mapping info in dnode page cache, skip defragmenting if all
2530          * physical block addresses are continuous even if there are hole(s)
2531          * in logical blocks.
2532          */
2533         while (map.m_lblk < pg_end) {
2534                 map.m_len = pg_end - map.m_lblk;
2535                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2536                 if (err)
2537                         goto out;
2538
2539                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2540                         map.m_lblk = next_pgofs;
2541                         continue;
2542                 }
2543
2544                 if (blk_end && blk_end != map.m_pblk)
2545                         fragmented = true;
2546
2547                 /* record total count of block that we're going to move */
2548                 total += map.m_len;
2549
2550                 blk_end = map.m_pblk + map.m_len;
2551
2552                 map.m_lblk += map.m_len;
2553         }
2554
2555         if (!fragmented) {
2556                 total = 0;
2557                 goto out;
2558         }
2559
2560         sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2561
2562         /*
2563          * make sure there are enough free section for LFS allocation, this can
2564          * avoid defragment running in SSR mode when free section are allocated
2565          * intensively
2566          */
2567         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2568                 err = -EAGAIN;
2569                 goto out;
2570         }
2571
2572         map.m_lblk = pg_start;
2573         map.m_len = pg_end - pg_start;
2574         total = 0;
2575
2576         while (map.m_lblk < pg_end) {
2577                 pgoff_t idx;
2578                 int cnt = 0;
2579
2580 do_map:
2581                 map.m_len = pg_end - map.m_lblk;
2582                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2583                 if (err)
2584                         goto clear_out;
2585
2586                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2587                         map.m_lblk = next_pgofs;
2588                         goto check;
2589                 }
2590
2591                 set_inode_flag(inode, FI_SKIP_WRITES);
2592
2593                 idx = map.m_lblk;
2594                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2595                         struct page *page;
2596
2597                         page = f2fs_get_lock_data_page(inode, idx, true);
2598                         if (IS_ERR(page)) {
2599                                 err = PTR_ERR(page);
2600                                 goto clear_out;
2601                         }
2602
2603                         set_page_dirty(page);
2604                         set_page_private_gcing(page);
2605                         f2fs_put_page(page, 1);
2606
2607                         idx++;
2608                         cnt++;
2609                         total++;
2610                 }
2611
2612                 map.m_lblk = idx;
2613 check:
2614                 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2615                         goto do_map;
2616
2617                 clear_inode_flag(inode, FI_SKIP_WRITES);
2618
2619                 err = filemap_fdatawrite(inode->i_mapping);
2620                 if (err)
2621                         goto out;
2622         }
2623 clear_out:
2624         clear_inode_flag(inode, FI_SKIP_WRITES);
2625 out:
2626         clear_inode_flag(inode, FI_OPU_WRITE);
2627         inode_unlock(inode);
2628         if (!err)
2629                 range->len = (u64)total << PAGE_SHIFT;
2630         return err;
2631 }
2632
2633 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2634 {
2635         struct inode *inode = file_inode(filp);
2636         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2637         struct f2fs_defragment range;
2638         int err;
2639
2640         if (!capable(CAP_SYS_ADMIN))
2641                 return -EPERM;
2642
2643         if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2644                 return -EINVAL;
2645
2646         if (f2fs_readonly(sbi->sb))
2647                 return -EROFS;
2648
2649         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2650                                                         sizeof(range)))
2651                 return -EFAULT;
2652
2653         /* verify alignment of offset & size */
2654         if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2655                 return -EINVAL;
2656
2657         if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2658                                         max_file_blocks(inode)))
2659                 return -EINVAL;
2660
2661         err = mnt_want_write_file(filp);
2662         if (err)
2663                 return err;
2664
2665         err = f2fs_defragment_range(sbi, filp, &range);
2666         mnt_drop_write_file(filp);
2667
2668         f2fs_update_time(sbi, REQ_TIME);
2669         if (err < 0)
2670                 return err;
2671
2672         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2673                                                         sizeof(range)))
2674                 return -EFAULT;
2675
2676         return 0;
2677 }
2678
2679 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2680                         struct file *file_out, loff_t pos_out, size_t len)
2681 {
2682         struct inode *src = file_inode(file_in);
2683         struct inode *dst = file_inode(file_out);
2684         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2685         size_t olen = len, dst_max_i_size = 0;
2686         size_t dst_osize;
2687         int ret;
2688
2689         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2690                                 src->i_sb != dst->i_sb)
2691                 return -EXDEV;
2692
2693         if (unlikely(f2fs_readonly(src->i_sb)))
2694                 return -EROFS;
2695
2696         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2697                 return -EINVAL;
2698
2699         if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2700                 return -EOPNOTSUPP;
2701
2702         if (pos_out < 0 || pos_in < 0)
2703                 return -EINVAL;
2704
2705         if (src == dst) {
2706                 if (pos_in == pos_out)
2707                         return 0;
2708                 if (pos_out > pos_in && pos_out < pos_in + len)
2709                         return -EINVAL;
2710         }
2711
2712         inode_lock(src);
2713         if (src != dst) {
2714                 ret = -EBUSY;
2715                 if (!inode_trylock(dst))
2716                         goto out;
2717         }
2718
2719         ret = -EINVAL;
2720         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2721                 goto out_unlock;
2722         if (len == 0)
2723                 olen = len = src->i_size - pos_in;
2724         if (pos_in + len == src->i_size)
2725                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2726         if (len == 0) {
2727                 ret = 0;
2728                 goto out_unlock;
2729         }
2730
2731         dst_osize = dst->i_size;
2732         if (pos_out + olen > dst->i_size)
2733                 dst_max_i_size = pos_out + olen;
2734
2735         /* verify the end result is block aligned */
2736         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2737                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2738                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2739                 goto out_unlock;
2740
2741         ret = f2fs_convert_inline_inode(src);
2742         if (ret)
2743                 goto out_unlock;
2744
2745         ret = f2fs_convert_inline_inode(dst);
2746         if (ret)
2747                 goto out_unlock;
2748
2749         /* write out all dirty pages from offset */
2750         ret = filemap_write_and_wait_range(src->i_mapping,
2751                                         pos_in, pos_in + len);
2752         if (ret)
2753                 goto out_unlock;
2754
2755         ret = filemap_write_and_wait_range(dst->i_mapping,
2756                                         pos_out, pos_out + len);
2757         if (ret)
2758                 goto out_unlock;
2759
2760         f2fs_balance_fs(sbi, true);
2761
2762         f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2763         if (src != dst) {
2764                 ret = -EBUSY;
2765                 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2766                         goto out_src;
2767         }
2768
2769         f2fs_lock_op(sbi);
2770         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2771                                 pos_out >> F2FS_BLKSIZE_BITS,
2772                                 len >> F2FS_BLKSIZE_BITS, false);
2773
2774         if (!ret) {
2775                 if (dst_max_i_size)
2776                         f2fs_i_size_write(dst, dst_max_i_size);
2777                 else if (dst_osize != dst->i_size)
2778                         f2fs_i_size_write(dst, dst_osize);
2779         }
2780         f2fs_unlock_op(sbi);
2781
2782         if (src != dst)
2783                 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2784 out_src:
2785         f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2786 out_unlock:
2787         if (src != dst)
2788                 inode_unlock(dst);
2789 out:
2790         inode_unlock(src);
2791         return ret;
2792 }
2793
2794 static int __f2fs_ioc_move_range(struct file *filp,
2795                                 struct f2fs_move_range *range)
2796 {
2797         struct fd dst;
2798         int err;
2799
2800         if (!(filp->f_mode & FMODE_READ) ||
2801                         !(filp->f_mode & FMODE_WRITE))
2802                 return -EBADF;
2803
2804         dst = fdget(range->dst_fd);
2805         if (!dst.file)
2806                 return -EBADF;
2807
2808         if (!(dst.file->f_mode & FMODE_WRITE)) {
2809                 err = -EBADF;
2810                 goto err_out;
2811         }
2812
2813         err = mnt_want_write_file(filp);
2814         if (err)
2815                 goto err_out;
2816
2817         err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2818                                         range->pos_out, range->len);
2819
2820         mnt_drop_write_file(filp);
2821 err_out:
2822         fdput(dst);
2823         return err;
2824 }
2825
2826 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2827 {
2828         struct f2fs_move_range range;
2829
2830         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2831                                                         sizeof(range)))
2832                 return -EFAULT;
2833         return __f2fs_ioc_move_range(filp, &range);
2834 }
2835
2836 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2837 {
2838         struct inode *inode = file_inode(filp);
2839         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2840         struct sit_info *sm = SIT_I(sbi);
2841         unsigned int start_segno = 0, end_segno = 0;
2842         unsigned int dev_start_segno = 0, dev_end_segno = 0;
2843         struct f2fs_flush_device range;
2844         struct f2fs_gc_control gc_control = {
2845                         .init_gc_type = FG_GC,
2846                         .should_migrate_blocks = true,
2847                         .err_gc_skipped = true,
2848                         .nr_free_secs = 0 };
2849         int ret;
2850
2851         if (!capable(CAP_SYS_ADMIN))
2852                 return -EPERM;
2853
2854         if (f2fs_readonly(sbi->sb))
2855                 return -EROFS;
2856
2857         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2858                 return -EINVAL;
2859
2860         if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2861                                                         sizeof(range)))
2862                 return -EFAULT;
2863
2864         if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2865                         __is_large_section(sbi)) {
2866                 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2867                           range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2868                 return -EINVAL;
2869         }
2870
2871         ret = mnt_want_write_file(filp);
2872         if (ret)
2873                 return ret;
2874
2875         if (range.dev_num != 0)
2876                 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2877         dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2878
2879         start_segno = sm->last_victim[FLUSH_DEVICE];
2880         if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2881                 start_segno = dev_start_segno;
2882         end_segno = min(start_segno + range.segments, dev_end_segno);
2883
2884         while (start_segno < end_segno) {
2885                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2886                         ret = -EBUSY;
2887                         goto out;
2888                 }
2889                 sm->last_victim[GC_CB] = end_segno + 1;
2890                 sm->last_victim[GC_GREEDY] = end_segno + 1;
2891                 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2892
2893                 gc_control.victim_segno = start_segno;
2894                 ret = f2fs_gc(sbi, &gc_control);
2895                 if (ret == -EAGAIN)
2896                         ret = 0;
2897                 else if (ret < 0)
2898                         break;
2899                 start_segno++;
2900         }
2901 out:
2902         mnt_drop_write_file(filp);
2903         return ret;
2904 }
2905
2906 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2907 {
2908         struct inode *inode = file_inode(filp);
2909         u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2910
2911         /* Must validate to set it with SQLite behavior in Android. */
2912         sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2913
2914         return put_user(sb_feature, (u32 __user *)arg);
2915 }
2916
2917 #ifdef CONFIG_QUOTA
2918 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2919 {
2920         struct dquot *transfer_to[MAXQUOTAS] = {};
2921         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2922         struct super_block *sb = sbi->sb;
2923         int err = 0;
2924
2925         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2926         if (!IS_ERR(transfer_to[PRJQUOTA])) {
2927                 err = __dquot_transfer(inode, transfer_to);
2928                 if (err)
2929                         set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2930                 dqput(transfer_to[PRJQUOTA]);
2931         }
2932         return err;
2933 }
2934
2935 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2936 {
2937         struct f2fs_inode_info *fi = F2FS_I(inode);
2938         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2939         struct f2fs_inode *ri = NULL;
2940         kprojid_t kprojid;
2941         int err;
2942
2943         if (!f2fs_sb_has_project_quota(sbi)) {
2944                 if (projid != F2FS_DEF_PROJID)
2945                         return -EOPNOTSUPP;
2946                 else
2947                         return 0;
2948         }
2949
2950         if (!f2fs_has_extra_attr(inode))
2951                 return -EOPNOTSUPP;
2952
2953         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2954
2955         if (projid_eq(kprojid, fi->i_projid))
2956                 return 0;
2957
2958         err = -EPERM;
2959         /* Is it quota file? Do not allow user to mess with it */
2960         if (IS_NOQUOTA(inode))
2961                 return err;
2962
2963         if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
2964                 return -EOVERFLOW;
2965
2966         err = f2fs_dquot_initialize(inode);
2967         if (err)
2968                 return err;
2969
2970         f2fs_lock_op(sbi);
2971         err = f2fs_transfer_project_quota(inode, kprojid);
2972         if (err)
2973                 goto out_unlock;
2974
2975         fi->i_projid = kprojid;
2976         inode->i_ctime = current_time(inode);
2977         f2fs_mark_inode_dirty_sync(inode, true);
2978 out_unlock:
2979         f2fs_unlock_op(sbi);
2980         return err;
2981 }
2982 #else
2983 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2984 {
2985         return 0;
2986 }
2987
2988 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2989 {
2990         if (projid != F2FS_DEF_PROJID)
2991                 return -EOPNOTSUPP;
2992         return 0;
2993 }
2994 #endif
2995
2996 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
2997 {
2998         struct inode *inode = d_inode(dentry);
2999         struct f2fs_inode_info *fi = F2FS_I(inode);
3000         u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3001
3002         if (IS_ENCRYPTED(inode))
3003                 fsflags |= FS_ENCRYPT_FL;
3004         if (IS_VERITY(inode))
3005                 fsflags |= FS_VERITY_FL;
3006         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3007                 fsflags |= FS_INLINE_DATA_FL;
3008         if (is_inode_flag_set(inode, FI_PIN_FILE))
3009                 fsflags |= FS_NOCOW_FL;
3010
3011         fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3012
3013         if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3014                 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3015
3016         return 0;
3017 }
3018
3019 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3020                       struct dentry *dentry, struct fileattr *fa)
3021 {
3022         struct inode *inode = d_inode(dentry);
3023         u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3024         u32 iflags;
3025         int err;
3026
3027         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3028                 return -EIO;
3029         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3030                 return -ENOSPC;
3031         if (fsflags & ~F2FS_GETTABLE_FS_FL)
3032                 return -EOPNOTSUPP;
3033         fsflags &= F2FS_SETTABLE_FS_FL;
3034         if (!fa->flags_valid)
3035                 mask &= FS_COMMON_FL;
3036
3037         iflags = f2fs_fsflags_to_iflags(fsflags);
3038         if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3039                 return -EOPNOTSUPP;
3040
3041         err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3042         if (!err)
3043                 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3044
3045         return err;
3046 }
3047
3048 int f2fs_pin_file_control(struct inode *inode, bool inc)
3049 {
3050         struct f2fs_inode_info *fi = F2FS_I(inode);
3051         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3052
3053         /* Use i_gc_failures for normal file as a risk signal. */
3054         if (inc)
3055                 f2fs_i_gc_failures_write(inode,
3056                                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3057
3058         if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3059                 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3060                           __func__, inode->i_ino,
3061                           fi->i_gc_failures[GC_FAILURE_PIN]);
3062                 clear_inode_flag(inode, FI_PIN_FILE);
3063                 return -EAGAIN;
3064         }
3065         return 0;
3066 }
3067
3068 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3069 {
3070         struct inode *inode = file_inode(filp);
3071         __u32 pin;
3072         int ret = 0;
3073
3074         if (get_user(pin, (__u32 __user *)arg))
3075                 return -EFAULT;
3076
3077         if (!S_ISREG(inode->i_mode))
3078                 return -EINVAL;
3079
3080         if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3081                 return -EROFS;
3082
3083         ret = mnt_want_write_file(filp);
3084         if (ret)
3085                 return ret;
3086
3087         inode_lock(inode);
3088
3089         if (!pin) {
3090                 clear_inode_flag(inode, FI_PIN_FILE);
3091                 f2fs_i_gc_failures_write(inode, 0);
3092                 goto done;
3093         }
3094
3095         if (f2fs_should_update_outplace(inode, NULL)) {
3096                 ret = -EINVAL;
3097                 goto out;
3098         }
3099
3100         if (f2fs_pin_file_control(inode, false)) {
3101                 ret = -EAGAIN;
3102                 goto out;
3103         }
3104
3105         ret = f2fs_convert_inline_inode(inode);
3106         if (ret)
3107                 goto out;
3108
3109         if (!f2fs_disable_compressed_file(inode)) {
3110                 ret = -EOPNOTSUPP;
3111                 goto out;
3112         }
3113
3114         set_inode_flag(inode, FI_PIN_FILE);
3115         ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3116 done:
3117         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3118 out:
3119         inode_unlock(inode);
3120         mnt_drop_write_file(filp);
3121         return ret;
3122 }
3123
3124 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3125 {
3126         struct inode *inode = file_inode(filp);
3127         __u32 pin = 0;
3128
3129         if (is_inode_flag_set(inode, FI_PIN_FILE))
3130                 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3131         return put_user(pin, (u32 __user *)arg);
3132 }
3133
3134 int f2fs_precache_extents(struct inode *inode)
3135 {
3136         struct f2fs_inode_info *fi = F2FS_I(inode);
3137         struct f2fs_map_blocks map;
3138         pgoff_t m_next_extent;
3139         loff_t end;
3140         int err;
3141
3142         if (is_inode_flag_set(inode, FI_NO_EXTENT))
3143                 return -EOPNOTSUPP;
3144
3145         map.m_lblk = 0;
3146         map.m_next_pgofs = NULL;
3147         map.m_next_extent = &m_next_extent;
3148         map.m_seg_type = NO_CHECK_TYPE;
3149         map.m_may_create = false;
3150         end = max_file_blocks(inode);
3151
3152         while (map.m_lblk < end) {
3153                 map.m_len = end - map.m_lblk;
3154
3155                 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3156                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3157                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3158                 if (err)
3159                         return err;
3160
3161                 map.m_lblk = m_next_extent;
3162         }
3163
3164         return 0;
3165 }
3166
3167 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3168 {
3169         return f2fs_precache_extents(file_inode(filp));
3170 }
3171
3172 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3173 {
3174         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3175         __u64 block_count;
3176
3177         if (!capable(CAP_SYS_ADMIN))
3178                 return -EPERM;
3179
3180         if (f2fs_readonly(sbi->sb))
3181                 return -EROFS;
3182
3183         if (copy_from_user(&block_count, (void __user *)arg,
3184                            sizeof(block_count)))
3185                 return -EFAULT;
3186
3187         return f2fs_resize_fs(sbi, block_count);
3188 }
3189
3190 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3191 {
3192         struct inode *inode = file_inode(filp);
3193
3194         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3195
3196         if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3197                 f2fs_warn(F2FS_I_SB(inode),
3198                           "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3199                           inode->i_ino);
3200                 return -EOPNOTSUPP;
3201         }
3202
3203         return fsverity_ioctl_enable(filp, (const void __user *)arg);
3204 }
3205
3206 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3207 {
3208         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3209                 return -EOPNOTSUPP;
3210
3211         return fsverity_ioctl_measure(filp, (void __user *)arg);
3212 }
3213
3214 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3215 {
3216         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3217                 return -EOPNOTSUPP;
3218
3219         return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3220 }
3221
3222 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3223 {
3224         struct inode *inode = file_inode(filp);
3225         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3226         char *vbuf;
3227         int count;
3228         int err = 0;
3229
3230         vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3231         if (!vbuf)
3232                 return -ENOMEM;
3233
3234         f2fs_down_read(&sbi->sb_lock);
3235         count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3236                         ARRAY_SIZE(sbi->raw_super->volume_name),
3237                         UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3238         f2fs_up_read(&sbi->sb_lock);
3239
3240         if (copy_to_user((char __user *)arg, vbuf,
3241                                 min(FSLABEL_MAX, count)))
3242                 err = -EFAULT;
3243
3244         kfree(vbuf);
3245         return err;
3246 }
3247
3248 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3249 {
3250         struct inode *inode = file_inode(filp);
3251         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3252         char *vbuf;
3253         int err = 0;
3254
3255         if (!capable(CAP_SYS_ADMIN))
3256                 return -EPERM;
3257
3258         vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3259         if (IS_ERR(vbuf))
3260                 return PTR_ERR(vbuf);
3261
3262         err = mnt_want_write_file(filp);
3263         if (err)
3264                 goto out;
3265
3266         f2fs_down_write(&sbi->sb_lock);
3267
3268         memset(sbi->raw_super->volume_name, 0,
3269                         sizeof(sbi->raw_super->volume_name));
3270         utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3271                         sbi->raw_super->volume_name,
3272                         ARRAY_SIZE(sbi->raw_super->volume_name));
3273
3274         err = f2fs_commit_super(sbi, false);
3275
3276         f2fs_up_write(&sbi->sb_lock);
3277
3278         mnt_drop_write_file(filp);
3279 out:
3280         kfree(vbuf);
3281         return err;
3282 }
3283
3284 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3285 {
3286         struct inode *inode = file_inode(filp);
3287         __u64 blocks;
3288
3289         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3290                 return -EOPNOTSUPP;
3291
3292         if (!f2fs_compressed_file(inode))
3293                 return -EINVAL;
3294
3295         blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3296         return put_user(blocks, (u64 __user *)arg);
3297 }
3298
3299 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3300 {
3301         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3302         unsigned int released_blocks = 0;
3303         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3304         block_t blkaddr;
3305         int i;
3306
3307         for (i = 0; i < count; i++) {
3308                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3309                                                 dn->ofs_in_node + i);
3310
3311                 if (!__is_valid_data_blkaddr(blkaddr))
3312                         continue;
3313                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3314                                         DATA_GENERIC_ENHANCE)))
3315                         return -EFSCORRUPTED;
3316         }
3317
3318         while (count) {
3319                 int compr_blocks = 0;
3320
3321                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3322                         blkaddr = f2fs_data_blkaddr(dn);
3323
3324                         if (i == 0) {
3325                                 if (blkaddr == COMPRESS_ADDR)
3326                                         continue;
3327                                 dn->ofs_in_node += cluster_size;
3328                                 goto next;
3329                         }
3330
3331                         if (__is_valid_data_blkaddr(blkaddr))
3332                                 compr_blocks++;
3333
3334                         if (blkaddr != NEW_ADDR)
3335                                 continue;
3336
3337                         dn->data_blkaddr = NULL_ADDR;
3338                         f2fs_set_data_blkaddr(dn);
3339                 }
3340
3341                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3342                 dec_valid_block_count(sbi, dn->inode,
3343                                         cluster_size - compr_blocks);
3344
3345                 released_blocks += cluster_size - compr_blocks;
3346 next:
3347                 count -= cluster_size;
3348         }
3349
3350         return released_blocks;
3351 }
3352
3353 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3354 {
3355         struct inode *inode = file_inode(filp);
3356         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3357         pgoff_t page_idx = 0, last_idx;
3358         unsigned int released_blocks = 0;
3359         int ret;
3360         int writecount;
3361
3362         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3363                 return -EOPNOTSUPP;
3364
3365         if (!f2fs_compressed_file(inode))
3366                 return -EINVAL;
3367
3368         if (f2fs_readonly(sbi->sb))
3369                 return -EROFS;
3370
3371         ret = mnt_want_write_file(filp);
3372         if (ret)
3373                 return ret;
3374
3375         f2fs_balance_fs(F2FS_I_SB(inode), true);
3376
3377         inode_lock(inode);
3378
3379         writecount = atomic_read(&inode->i_writecount);
3380         if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3381                         (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3382                 ret = -EBUSY;
3383                 goto out;
3384         }
3385
3386         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3387                 ret = -EINVAL;
3388                 goto out;
3389         }
3390
3391         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3392         if (ret)
3393                 goto out;
3394
3395         set_inode_flag(inode, FI_COMPRESS_RELEASED);
3396         inode->i_ctime = current_time(inode);
3397         f2fs_mark_inode_dirty_sync(inode, true);
3398
3399         if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3400                 goto out;
3401
3402         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3403         filemap_invalidate_lock(inode->i_mapping);
3404
3405         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3406
3407         while (page_idx < last_idx) {
3408                 struct dnode_of_data dn;
3409                 pgoff_t end_offset, count;
3410
3411                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3412                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3413                 if (ret) {
3414                         if (ret == -ENOENT) {
3415                                 page_idx = f2fs_get_next_page_offset(&dn,
3416                                                                 page_idx);
3417                                 ret = 0;
3418                                 continue;
3419                         }
3420                         break;
3421                 }
3422
3423                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3424                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3425                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3426
3427                 ret = release_compress_blocks(&dn, count);
3428
3429                 f2fs_put_dnode(&dn);
3430
3431                 if (ret < 0)
3432                         break;
3433
3434                 page_idx += count;
3435                 released_blocks += ret;
3436         }
3437
3438         filemap_invalidate_unlock(inode->i_mapping);
3439         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3440 out:
3441         inode_unlock(inode);
3442
3443         mnt_drop_write_file(filp);
3444
3445         if (ret >= 0) {
3446                 ret = put_user(released_blocks, (u64 __user *)arg);
3447         } else if (released_blocks &&
3448                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3449                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3450                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3451                         "iblocks=%llu, released=%u, compr_blocks=%u, "
3452                         "run fsck to fix.",
3453                         __func__, inode->i_ino, inode->i_blocks,
3454                         released_blocks,
3455                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3456         }
3457
3458         return ret;
3459 }
3460
3461 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3462 {
3463         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3464         unsigned int reserved_blocks = 0;
3465         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3466         block_t blkaddr;
3467         int i;
3468
3469         for (i = 0; i < count; i++) {
3470                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3471                                                 dn->ofs_in_node + i);
3472
3473                 if (!__is_valid_data_blkaddr(blkaddr))
3474                         continue;
3475                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3476                                         DATA_GENERIC_ENHANCE)))
3477                         return -EFSCORRUPTED;
3478         }
3479
3480         while (count) {
3481                 int compr_blocks = 0;
3482                 blkcnt_t reserved;
3483                 int ret;
3484
3485                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3486                         blkaddr = f2fs_data_blkaddr(dn);
3487
3488                         if (i == 0) {
3489                                 if (blkaddr == COMPRESS_ADDR)
3490                                         continue;
3491                                 dn->ofs_in_node += cluster_size;
3492                                 goto next;
3493                         }
3494
3495                         if (__is_valid_data_blkaddr(blkaddr)) {
3496                                 compr_blocks++;
3497                                 continue;
3498                         }
3499
3500                         dn->data_blkaddr = NEW_ADDR;
3501                         f2fs_set_data_blkaddr(dn);
3502                 }
3503
3504                 reserved = cluster_size - compr_blocks;
3505                 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3506                 if (ret)
3507                         return ret;
3508
3509                 if (reserved != cluster_size - compr_blocks)
3510                         return -ENOSPC;
3511
3512                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3513
3514                 reserved_blocks += reserved;
3515 next:
3516                 count -= cluster_size;
3517         }
3518
3519         return reserved_blocks;
3520 }
3521
3522 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3523 {
3524         struct inode *inode = file_inode(filp);
3525         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3526         pgoff_t page_idx = 0, last_idx;
3527         unsigned int reserved_blocks = 0;
3528         int ret;
3529
3530         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3531                 return -EOPNOTSUPP;
3532
3533         if (!f2fs_compressed_file(inode))
3534                 return -EINVAL;
3535
3536         if (f2fs_readonly(sbi->sb))
3537                 return -EROFS;
3538
3539         ret = mnt_want_write_file(filp);
3540         if (ret)
3541                 return ret;
3542
3543         if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3544                 goto out;
3545
3546         f2fs_balance_fs(F2FS_I_SB(inode), true);
3547
3548         inode_lock(inode);
3549
3550         if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3551                 ret = -EINVAL;
3552                 goto unlock_inode;
3553         }
3554
3555         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3556         filemap_invalidate_lock(inode->i_mapping);
3557
3558         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3559
3560         while (page_idx < last_idx) {
3561                 struct dnode_of_data dn;
3562                 pgoff_t end_offset, count;
3563
3564                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3565                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3566                 if (ret) {
3567                         if (ret == -ENOENT) {
3568                                 page_idx = f2fs_get_next_page_offset(&dn,
3569                                                                 page_idx);
3570                                 ret = 0;
3571                                 continue;
3572                         }
3573                         break;
3574                 }
3575
3576                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3577                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3578                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3579
3580                 ret = reserve_compress_blocks(&dn, count);
3581
3582                 f2fs_put_dnode(&dn);
3583
3584                 if (ret < 0)
3585                         break;
3586
3587                 page_idx += count;
3588                 reserved_blocks += ret;
3589         }
3590
3591         filemap_invalidate_unlock(inode->i_mapping);
3592         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3593
3594         if (ret >= 0) {
3595                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3596                 inode->i_ctime = current_time(inode);
3597                 f2fs_mark_inode_dirty_sync(inode, true);
3598         }
3599 unlock_inode:
3600         inode_unlock(inode);
3601 out:
3602         mnt_drop_write_file(filp);
3603
3604         if (ret >= 0) {
3605                 ret = put_user(reserved_blocks, (u64 __user *)arg);
3606         } else if (reserved_blocks &&
3607                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3608                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3609                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3610                         "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3611                         "run fsck to fix.",
3612                         __func__, inode->i_ino, inode->i_blocks,
3613                         reserved_blocks,
3614                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3615         }
3616
3617         return ret;
3618 }
3619
3620 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3621                 pgoff_t off, block_t block, block_t len, u32 flags)
3622 {
3623         sector_t sector = SECTOR_FROM_BLOCK(block);
3624         sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3625         int ret = 0;
3626
3627         if (flags & F2FS_TRIM_FILE_DISCARD) {
3628                 if (bdev_max_secure_erase_sectors(bdev))
3629                         ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3630                                         GFP_NOFS);
3631                 else
3632                         ret = blkdev_issue_discard(bdev, sector, nr_sects,
3633                                         GFP_NOFS);
3634         }
3635
3636         if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3637                 if (IS_ENCRYPTED(inode))
3638                         ret = fscrypt_zeroout_range(inode, off, block, len);
3639                 else
3640                         ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3641                                         GFP_NOFS, 0);
3642         }
3643
3644         return ret;
3645 }
3646
3647 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3648 {
3649         struct inode *inode = file_inode(filp);
3650         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3651         struct address_space *mapping = inode->i_mapping;
3652         struct block_device *prev_bdev = NULL;
3653         struct f2fs_sectrim_range range;
3654         pgoff_t index, pg_end, prev_index = 0;
3655         block_t prev_block = 0, len = 0;
3656         loff_t end_addr;
3657         bool to_end = false;
3658         int ret = 0;
3659
3660         if (!(filp->f_mode & FMODE_WRITE))
3661                 return -EBADF;
3662
3663         if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3664                                 sizeof(range)))
3665                 return -EFAULT;
3666
3667         if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3668                         !S_ISREG(inode->i_mode))
3669                 return -EINVAL;
3670
3671         if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3672                         !f2fs_hw_support_discard(sbi)) ||
3673                         ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3674                          IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3675                 return -EOPNOTSUPP;
3676
3677         file_start_write(filp);
3678         inode_lock(inode);
3679
3680         if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3681                         range.start >= inode->i_size) {
3682                 ret = -EINVAL;
3683                 goto err;
3684         }
3685
3686         if (range.len == 0)
3687                 goto err;
3688
3689         if (inode->i_size - range.start > range.len) {
3690                 end_addr = range.start + range.len;
3691         } else {
3692                 end_addr = range.len == (u64)-1 ?
3693                         sbi->sb->s_maxbytes : inode->i_size;
3694                 to_end = true;
3695         }
3696
3697         if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3698                         (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3699                 ret = -EINVAL;
3700                 goto err;
3701         }
3702
3703         index = F2FS_BYTES_TO_BLK(range.start);
3704         pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3705
3706         ret = f2fs_convert_inline_inode(inode);
3707         if (ret)
3708                 goto err;
3709
3710         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3711         filemap_invalidate_lock(mapping);
3712
3713         ret = filemap_write_and_wait_range(mapping, range.start,
3714                         to_end ? LLONG_MAX : end_addr - 1);
3715         if (ret)
3716                 goto out;
3717
3718         truncate_inode_pages_range(mapping, range.start,
3719                         to_end ? -1 : end_addr - 1);
3720
3721         while (index < pg_end) {
3722                 struct dnode_of_data dn;
3723                 pgoff_t end_offset, count;
3724                 int i;
3725
3726                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3727                 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3728                 if (ret) {
3729                         if (ret == -ENOENT) {
3730                                 index = f2fs_get_next_page_offset(&dn, index);
3731                                 continue;
3732                         }
3733                         goto out;
3734                 }
3735
3736                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3737                 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3738                 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3739                         struct block_device *cur_bdev;
3740                         block_t blkaddr = f2fs_data_blkaddr(&dn);
3741
3742                         if (!__is_valid_data_blkaddr(blkaddr))
3743                                 continue;
3744
3745                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3746                                                 DATA_GENERIC_ENHANCE)) {
3747                                 ret = -EFSCORRUPTED;
3748                                 f2fs_put_dnode(&dn);
3749                                 goto out;
3750                         }
3751
3752                         cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3753                         if (f2fs_is_multi_device(sbi)) {
3754                                 int di = f2fs_target_device_index(sbi, blkaddr);
3755
3756                                 blkaddr -= FDEV(di).start_blk;
3757                         }
3758
3759                         if (len) {
3760                                 if (prev_bdev == cur_bdev &&
3761                                                 index == prev_index + len &&
3762                                                 blkaddr == prev_block + len) {
3763                                         len++;
3764                                 } else {
3765                                         ret = f2fs_secure_erase(prev_bdev,
3766                                                 inode, prev_index, prev_block,
3767                                                 len, range.flags);
3768                                         if (ret) {
3769                                                 f2fs_put_dnode(&dn);
3770                                                 goto out;
3771                                         }
3772
3773                                         len = 0;
3774                                 }
3775                         }
3776
3777                         if (!len) {
3778                                 prev_bdev = cur_bdev;
3779                                 prev_index = index;
3780                                 prev_block = blkaddr;
3781                                 len = 1;
3782                         }
3783                 }
3784
3785                 f2fs_put_dnode(&dn);
3786
3787                 if (fatal_signal_pending(current)) {
3788                         ret = -EINTR;
3789                         goto out;
3790                 }
3791                 cond_resched();
3792         }
3793
3794         if (len)
3795                 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3796                                 prev_block, len, range.flags);
3797 out:
3798         filemap_invalidate_unlock(mapping);
3799         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3800 err:
3801         inode_unlock(inode);
3802         file_end_write(filp);
3803
3804         return ret;
3805 }
3806
3807 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3808 {
3809         struct inode *inode = file_inode(filp);
3810         struct f2fs_comp_option option;
3811
3812         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3813                 return -EOPNOTSUPP;
3814
3815         inode_lock_shared(inode);
3816
3817         if (!f2fs_compressed_file(inode)) {
3818                 inode_unlock_shared(inode);
3819                 return -ENODATA;
3820         }
3821
3822         option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3823         option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3824
3825         inode_unlock_shared(inode);
3826
3827         if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3828                                 sizeof(option)))
3829                 return -EFAULT;
3830
3831         return 0;
3832 }
3833
3834 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3835 {
3836         struct inode *inode = file_inode(filp);
3837         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3838         struct f2fs_comp_option option;
3839         int ret = 0;
3840
3841         if (!f2fs_sb_has_compression(sbi))
3842                 return -EOPNOTSUPP;
3843
3844         if (!(filp->f_mode & FMODE_WRITE))
3845                 return -EBADF;
3846
3847         if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3848                                 sizeof(option)))
3849                 return -EFAULT;
3850
3851         if (!f2fs_compressed_file(inode) ||
3852                         option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3853                         option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3854                         option.algorithm >= COMPRESS_MAX)
3855                 return -EINVAL;
3856
3857         file_start_write(filp);
3858         inode_lock(inode);
3859
3860         if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3861                 ret = -EBUSY;
3862                 goto out;
3863         }
3864
3865         if (inode->i_size != 0) {
3866                 ret = -EFBIG;
3867                 goto out;
3868         }
3869
3870         F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3871         F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3872         F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3873         f2fs_mark_inode_dirty_sync(inode, true);
3874
3875         if (!f2fs_is_compress_backend_ready(inode))
3876                 f2fs_warn(sbi, "compression algorithm is successfully set, "
3877                         "but current kernel doesn't support this algorithm.");
3878 out:
3879         inode_unlock(inode);
3880         file_end_write(filp);
3881
3882         return ret;
3883 }
3884
3885 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3886 {
3887         DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3888         struct address_space *mapping = inode->i_mapping;
3889         struct page *page;
3890         pgoff_t redirty_idx = page_idx;
3891         int i, page_len = 0, ret = 0;
3892
3893         page_cache_ra_unbounded(&ractl, len, 0);
3894
3895         for (i = 0; i < len; i++, page_idx++) {
3896                 page = read_cache_page(mapping, page_idx, NULL, NULL);
3897                 if (IS_ERR(page)) {
3898                         ret = PTR_ERR(page);
3899                         break;
3900                 }
3901                 page_len++;
3902         }
3903
3904         for (i = 0; i < page_len; i++, redirty_idx++) {
3905                 page = find_lock_page(mapping, redirty_idx);
3906                 if (!page) {
3907                         ret = -ENOMEM;
3908                         break;
3909                 }
3910                 set_page_dirty(page);
3911                 f2fs_put_page(page, 1);
3912                 f2fs_put_page(page, 0);
3913         }
3914
3915         return ret;
3916 }
3917
3918 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3919 {
3920         struct inode *inode = file_inode(filp);
3921         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3922         struct f2fs_inode_info *fi = F2FS_I(inode);
3923         pgoff_t page_idx = 0, last_idx;
3924         unsigned int blk_per_seg = sbi->blocks_per_seg;
3925         int cluster_size = fi->i_cluster_size;
3926         int count, ret;
3927
3928         if (!f2fs_sb_has_compression(sbi) ||
3929                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3930                 return -EOPNOTSUPP;
3931
3932         if (!(filp->f_mode & FMODE_WRITE))
3933                 return -EBADF;
3934
3935         if (!f2fs_compressed_file(inode))
3936                 return -EINVAL;
3937
3938         f2fs_balance_fs(F2FS_I_SB(inode), true);
3939
3940         file_start_write(filp);
3941         inode_lock(inode);
3942
3943         if (!f2fs_is_compress_backend_ready(inode)) {
3944                 ret = -EOPNOTSUPP;
3945                 goto out;
3946         }
3947
3948         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3949         if (ret)
3950                 goto out;
3951
3952         if (!atomic_read(&fi->i_compr_blocks))
3953                 goto out;
3954
3955         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3956
3957         count = last_idx - page_idx;
3958         while (count) {
3959                 int len = min(cluster_size, count);
3960
3961                 ret = redirty_blocks(inode, page_idx, len);
3962                 if (ret < 0)
3963                         break;
3964
3965                 if (get_dirty_pages(inode) >= blk_per_seg)
3966                         filemap_fdatawrite(inode->i_mapping);
3967
3968                 count -= len;
3969                 page_idx += len;
3970         }
3971
3972         if (!ret)
3973                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
3974                                                         LLONG_MAX);
3975
3976         if (ret)
3977                 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
3978                           __func__, ret);
3979 out:
3980         inode_unlock(inode);
3981         file_end_write(filp);
3982
3983         return ret;
3984 }
3985
3986 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
3987 {
3988         struct inode *inode = file_inode(filp);
3989         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3990         pgoff_t page_idx = 0, last_idx;
3991         unsigned int blk_per_seg = sbi->blocks_per_seg;
3992         int cluster_size = F2FS_I(inode)->i_cluster_size;
3993         int count, ret;
3994
3995         if (!f2fs_sb_has_compression(sbi) ||
3996                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3997                 return -EOPNOTSUPP;
3998
3999         if (!(filp->f_mode & FMODE_WRITE))
4000                 return -EBADF;
4001
4002         if (!f2fs_compressed_file(inode))
4003                 return -EINVAL;
4004
4005         f2fs_balance_fs(F2FS_I_SB(inode), true);
4006
4007         file_start_write(filp);
4008         inode_lock(inode);
4009
4010         if (!f2fs_is_compress_backend_ready(inode)) {
4011                 ret = -EOPNOTSUPP;
4012                 goto out;
4013         }
4014
4015         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4016         if (ret)
4017                 goto out;
4018
4019         set_inode_flag(inode, FI_ENABLE_COMPRESS);
4020
4021         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4022
4023         count = last_idx - page_idx;
4024         while (count) {
4025                 int len = min(cluster_size, count);
4026
4027                 ret = redirty_blocks(inode, page_idx, len);
4028                 if (ret < 0)
4029                         break;
4030
4031                 if (get_dirty_pages(inode) >= blk_per_seg)
4032                         filemap_fdatawrite(inode->i_mapping);
4033
4034                 count -= len;
4035                 page_idx += len;
4036         }
4037
4038         if (!ret)
4039                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4040                                                         LLONG_MAX);
4041
4042         clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4043
4044         if (ret)
4045                 f2fs_warn(sbi, "%s: The file might be partially compressed (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 long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4055 {
4056         switch (cmd) {
4057         case FS_IOC_GETVERSION:
4058                 return f2fs_ioc_getversion(filp, arg);
4059         case F2FS_IOC_START_ATOMIC_WRITE:
4060                 return f2fs_ioc_start_atomic_write(filp);
4061         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4062                 return f2fs_ioc_commit_atomic_write(filp);
4063         case F2FS_IOC_START_VOLATILE_WRITE:
4064         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4065         case F2FS_IOC_ABORT_VOLATILE_WRITE:
4066                 return -EOPNOTSUPP;
4067         case F2FS_IOC_SHUTDOWN:
4068                 return f2fs_ioc_shutdown(filp, arg);
4069         case FITRIM:
4070                 return f2fs_ioc_fitrim(filp, arg);
4071         case FS_IOC_SET_ENCRYPTION_POLICY:
4072                 return f2fs_ioc_set_encryption_policy(filp, arg);
4073         case FS_IOC_GET_ENCRYPTION_POLICY:
4074                 return f2fs_ioc_get_encryption_policy(filp, arg);
4075         case FS_IOC_GET_ENCRYPTION_PWSALT:
4076                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4077         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4078                 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4079         case FS_IOC_ADD_ENCRYPTION_KEY:
4080                 return f2fs_ioc_add_encryption_key(filp, arg);
4081         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4082                 return f2fs_ioc_remove_encryption_key(filp, arg);
4083         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4084                 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4085         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4086                 return f2fs_ioc_get_encryption_key_status(filp, arg);
4087         case FS_IOC_GET_ENCRYPTION_NONCE:
4088                 return f2fs_ioc_get_encryption_nonce(filp, arg);
4089         case F2FS_IOC_GARBAGE_COLLECT:
4090                 return f2fs_ioc_gc(filp, arg);
4091         case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4092                 return f2fs_ioc_gc_range(filp, arg);
4093         case F2FS_IOC_WRITE_CHECKPOINT:
4094                 return f2fs_ioc_write_checkpoint(filp, arg);
4095         case F2FS_IOC_DEFRAGMENT:
4096                 return f2fs_ioc_defragment(filp, arg);
4097         case F2FS_IOC_MOVE_RANGE:
4098                 return f2fs_ioc_move_range(filp, arg);
4099         case F2FS_IOC_FLUSH_DEVICE:
4100                 return f2fs_ioc_flush_device(filp, arg);
4101         case F2FS_IOC_GET_FEATURES:
4102                 return f2fs_ioc_get_features(filp, arg);
4103         case F2FS_IOC_GET_PIN_FILE:
4104                 return f2fs_ioc_get_pin_file(filp, arg);
4105         case F2FS_IOC_SET_PIN_FILE:
4106                 return f2fs_ioc_set_pin_file(filp, arg);
4107         case F2FS_IOC_PRECACHE_EXTENTS:
4108                 return f2fs_ioc_precache_extents(filp, arg);
4109         case F2FS_IOC_RESIZE_FS:
4110                 return f2fs_ioc_resize_fs(filp, arg);
4111         case FS_IOC_ENABLE_VERITY:
4112                 return f2fs_ioc_enable_verity(filp, arg);
4113         case FS_IOC_MEASURE_VERITY:
4114                 return f2fs_ioc_measure_verity(filp, arg);
4115         case FS_IOC_READ_VERITY_METADATA:
4116                 return f2fs_ioc_read_verity_metadata(filp, arg);
4117         case FS_IOC_GETFSLABEL:
4118                 return f2fs_ioc_getfslabel(filp, arg);
4119         case FS_IOC_SETFSLABEL:
4120                 return f2fs_ioc_setfslabel(filp, arg);
4121         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4122                 return f2fs_get_compress_blocks(filp, arg);
4123         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4124                 return f2fs_release_compress_blocks(filp, arg);
4125         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4126                 return f2fs_reserve_compress_blocks(filp, arg);
4127         case F2FS_IOC_SEC_TRIM_FILE:
4128                 return f2fs_sec_trim_file(filp, arg);
4129         case F2FS_IOC_GET_COMPRESS_OPTION:
4130                 return f2fs_ioc_get_compress_option(filp, arg);
4131         case F2FS_IOC_SET_COMPRESS_OPTION:
4132                 return f2fs_ioc_set_compress_option(filp, arg);
4133         case F2FS_IOC_DECOMPRESS_FILE:
4134                 return f2fs_ioc_decompress_file(filp, arg);
4135         case F2FS_IOC_COMPRESS_FILE:
4136                 return f2fs_ioc_compress_file(filp, arg);
4137         default:
4138                 return -ENOTTY;
4139         }
4140 }
4141
4142 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4143 {
4144         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4145                 return -EIO;
4146         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4147                 return -ENOSPC;
4148
4149         return __f2fs_ioctl(filp, cmd, arg);
4150 }
4151
4152 /*
4153  * Return %true if the given read or write request should use direct I/O, or
4154  * %false if it should use buffered I/O.
4155  */
4156 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4157                                 struct iov_iter *iter)
4158 {
4159         unsigned int align;
4160
4161         if (!(iocb->ki_flags & IOCB_DIRECT))
4162                 return false;
4163
4164         if (f2fs_force_buffered_io(inode, iocb, iter))
4165                 return false;
4166
4167         /*
4168          * Direct I/O not aligned to the disk's logical_block_size will be
4169          * attempted, but will fail with -EINVAL.
4170          *
4171          * f2fs additionally requires that direct I/O be aligned to the
4172          * filesystem block size, which is often a stricter requirement.
4173          * However, f2fs traditionally falls back to buffered I/O on requests
4174          * that are logical_block_size-aligned but not fs-block aligned.
4175          *
4176          * The below logic implements this behavior.
4177          */
4178         align = iocb->ki_pos | iov_iter_alignment(iter);
4179         if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4180             IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4181                 return false;
4182
4183         return true;
4184 }
4185
4186 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4187                                 unsigned int flags)
4188 {
4189         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4190
4191         dec_page_count(sbi, F2FS_DIO_READ);
4192         if (error)
4193                 return error;
4194         f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4195         return 0;
4196 }
4197
4198 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4199         .end_io = f2fs_dio_read_end_io,
4200 };
4201
4202 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4203 {
4204         struct file *file = iocb->ki_filp;
4205         struct inode *inode = file_inode(file);
4206         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4207         struct f2fs_inode_info *fi = F2FS_I(inode);
4208         const loff_t pos = iocb->ki_pos;
4209         const size_t count = iov_iter_count(to);
4210         struct iomap_dio *dio;
4211         ssize_t ret;
4212
4213         if (count == 0)
4214                 return 0; /* skip atime update */
4215
4216         trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4217
4218         if (iocb->ki_flags & IOCB_NOWAIT) {
4219                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4220                         ret = -EAGAIN;
4221                         goto out;
4222                 }
4223         } else {
4224                 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4225         }
4226
4227         /*
4228          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4229          * the higher-level function iomap_dio_rw() in order to ensure that the
4230          * F2FS_DIO_READ counter will be decremented correctly in all cases.
4231          */
4232         inc_page_count(sbi, F2FS_DIO_READ);
4233         dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4234                              &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4235         if (IS_ERR_OR_NULL(dio)) {
4236                 ret = PTR_ERR_OR_ZERO(dio);
4237                 if (ret != -EIOCBQUEUED)
4238                         dec_page_count(sbi, F2FS_DIO_READ);
4239         } else {
4240                 ret = iomap_dio_complete(dio);
4241         }
4242
4243         f2fs_up_read(&fi->i_gc_rwsem[READ]);
4244
4245         file_accessed(file);
4246 out:
4247         trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4248         return ret;
4249 }
4250
4251 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4252 {
4253         struct inode *inode = file_inode(iocb->ki_filp);
4254         const loff_t pos = iocb->ki_pos;
4255         ssize_t ret;
4256
4257         if (!f2fs_is_compress_backend_ready(inode))
4258                 return -EOPNOTSUPP;
4259
4260         if (trace_f2fs_dataread_start_enabled()) {
4261                 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4262                 char *path;
4263
4264                 if (!p)
4265                         goto skip_read_trace;
4266
4267                 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4268                 if (IS_ERR(path)) {
4269                         kfree(p);
4270                         goto skip_read_trace;
4271                 }
4272
4273                 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4274                                         current->pid, path, current->comm);
4275                 kfree(p);
4276         }
4277 skip_read_trace:
4278         if (f2fs_should_use_dio(inode, iocb, to)) {
4279                 ret = f2fs_dio_read_iter(iocb, to);
4280         } else {
4281                 ret = filemap_read(iocb, to, 0);
4282                 if (ret > 0)
4283                         f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4284         }
4285         if (trace_f2fs_dataread_end_enabled())
4286                 trace_f2fs_dataread_end(inode, pos, ret);
4287         return ret;
4288 }
4289
4290 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4291 {
4292         struct file *file = iocb->ki_filp;
4293         struct inode *inode = file_inode(file);
4294         ssize_t count;
4295         int err;
4296
4297         if (IS_IMMUTABLE(inode))
4298                 return -EPERM;
4299
4300         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4301                 return -EPERM;
4302
4303         count = generic_write_checks(iocb, from);
4304         if (count <= 0)
4305                 return count;
4306
4307         err = file_modified(file);
4308         if (err)
4309                 return err;
4310         return count;
4311 }
4312
4313 /*
4314  * Preallocate blocks for a write request, if it is possible and helpful to do
4315  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4316  * blocks were preallocated, or a negative errno value if something went
4317  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4318  * requested blocks (not just some of them) have been allocated.
4319  */
4320 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4321                                    bool dio)
4322 {
4323         struct inode *inode = file_inode(iocb->ki_filp);
4324         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4325         const loff_t pos = iocb->ki_pos;
4326         const size_t count = iov_iter_count(iter);
4327         struct f2fs_map_blocks map = {};
4328         int flag;
4329         int ret;
4330
4331         /* If it will be an out-of-place direct write, don't bother. */
4332         if (dio && f2fs_lfs_mode(sbi))
4333                 return 0;
4334         /*
4335          * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4336          * buffered IO, if DIO meets any holes.
4337          */
4338         if (dio && i_size_read(inode) &&
4339                 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4340                 return 0;
4341
4342         /* No-wait I/O can't allocate blocks. */
4343         if (iocb->ki_flags & IOCB_NOWAIT)
4344                 return 0;
4345
4346         /* If it will be a short write, don't bother. */
4347         if (fault_in_iov_iter_readable(iter, count))
4348                 return 0;
4349
4350         if (f2fs_has_inline_data(inode)) {
4351                 /* If the data will fit inline, don't bother. */
4352                 if (pos + count <= MAX_INLINE_DATA(inode))
4353                         return 0;
4354                 ret = f2fs_convert_inline_inode(inode);
4355                 if (ret)
4356                         return ret;
4357         }
4358
4359         /* Do not preallocate blocks that will be written partially in 4KB. */
4360         map.m_lblk = F2FS_BLK_ALIGN(pos);
4361         map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4362         if (map.m_len > map.m_lblk)
4363                 map.m_len -= map.m_lblk;
4364         else
4365                 map.m_len = 0;
4366         map.m_may_create = true;
4367         if (dio) {
4368                 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4369                 flag = F2FS_GET_BLOCK_PRE_DIO;
4370         } else {
4371                 map.m_seg_type = NO_CHECK_TYPE;
4372                 flag = F2FS_GET_BLOCK_PRE_AIO;
4373         }
4374
4375         ret = f2fs_map_blocks(inode, &map, 1, flag);
4376         /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4377         if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4378                 return ret;
4379         if (ret == 0)
4380                 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4381         return map.m_len;
4382 }
4383
4384 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4385                                         struct iov_iter *from)
4386 {
4387         struct file *file = iocb->ki_filp;
4388         struct inode *inode = file_inode(file);
4389         ssize_t ret;
4390
4391         if (iocb->ki_flags & IOCB_NOWAIT)
4392                 return -EOPNOTSUPP;
4393
4394         current->backing_dev_info = inode_to_bdi(inode);
4395         ret = generic_perform_write(iocb, from);
4396         current->backing_dev_info = NULL;
4397
4398         if (ret > 0) {
4399                 iocb->ki_pos += ret;
4400                 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4401         }
4402         return ret;
4403 }
4404
4405 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4406                                  unsigned int flags)
4407 {
4408         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4409
4410         dec_page_count(sbi, F2FS_DIO_WRITE);
4411         if (error)
4412                 return error;
4413         f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4414         return 0;
4415 }
4416
4417 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4418         .end_io = f2fs_dio_write_end_io,
4419 };
4420
4421 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4422                                    bool *may_need_sync)
4423 {
4424         struct file *file = iocb->ki_filp;
4425         struct inode *inode = file_inode(file);
4426         struct f2fs_inode_info *fi = F2FS_I(inode);
4427         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4428         const bool do_opu = f2fs_lfs_mode(sbi);
4429         const loff_t pos = iocb->ki_pos;
4430         const ssize_t count = iov_iter_count(from);
4431         unsigned int dio_flags;
4432         struct iomap_dio *dio;
4433         ssize_t ret;
4434
4435         trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4436
4437         if (iocb->ki_flags & IOCB_NOWAIT) {
4438                 /* f2fs_convert_inline_inode() and block allocation can block */
4439                 if (f2fs_has_inline_data(inode) ||
4440                     !f2fs_overwrite_io(inode, pos, count)) {
4441                         ret = -EAGAIN;
4442                         goto out;
4443                 }
4444
4445                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4446                         ret = -EAGAIN;
4447                         goto out;
4448                 }
4449                 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4450                         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4451                         ret = -EAGAIN;
4452                         goto out;
4453                 }
4454         } else {
4455                 ret = f2fs_convert_inline_inode(inode);
4456                 if (ret)
4457                         goto out;
4458
4459                 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4460                 if (do_opu)
4461                         f2fs_down_read(&fi->i_gc_rwsem[READ]);
4462         }
4463
4464         /*
4465          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4466          * the higher-level function iomap_dio_rw() in order to ensure that the
4467          * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4468          */
4469         inc_page_count(sbi, F2FS_DIO_WRITE);
4470         dio_flags = 0;
4471         if (pos + count > inode->i_size)
4472                 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4473         dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4474                              &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4475         if (IS_ERR_OR_NULL(dio)) {
4476                 ret = PTR_ERR_OR_ZERO(dio);
4477                 if (ret == -ENOTBLK)
4478                         ret = 0;
4479                 if (ret != -EIOCBQUEUED)
4480                         dec_page_count(sbi, F2FS_DIO_WRITE);
4481         } else {
4482                 ret = iomap_dio_complete(dio);
4483         }
4484
4485         if (do_opu)
4486                 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4487         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4488
4489         if (ret < 0)
4490                 goto out;
4491         if (pos + ret > inode->i_size)
4492                 f2fs_i_size_write(inode, pos + ret);
4493         if (!do_opu)
4494                 set_inode_flag(inode, FI_UPDATE_WRITE);
4495
4496         if (iov_iter_count(from)) {
4497                 ssize_t ret2;
4498                 loff_t bufio_start_pos = iocb->ki_pos;
4499
4500                 /*
4501                  * The direct write was partial, so we need to fall back to a
4502                  * buffered write for the remainder.
4503                  */
4504
4505                 ret2 = f2fs_buffered_write_iter(iocb, from);
4506                 if (iov_iter_count(from))
4507                         f2fs_write_failed(inode, iocb->ki_pos);
4508                 if (ret2 < 0)
4509                         goto out;
4510
4511                 /*
4512                  * Ensure that the pagecache pages are written to disk and
4513                  * invalidated to preserve the expected O_DIRECT semantics.
4514                  */
4515                 if (ret2 > 0) {
4516                         loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4517
4518                         ret += ret2;
4519
4520                         ret2 = filemap_write_and_wait_range(file->f_mapping,
4521                                                             bufio_start_pos,
4522                                                             bufio_end_pos);
4523                         if (ret2 < 0)
4524                                 goto out;
4525                         invalidate_mapping_pages(file->f_mapping,
4526                                                  bufio_start_pos >> PAGE_SHIFT,
4527                                                  bufio_end_pos >> PAGE_SHIFT);
4528                 }
4529         } else {
4530                 /* iomap_dio_rw() already handled the generic_write_sync(). */
4531                 *may_need_sync = false;
4532         }
4533 out:
4534         trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4535         return ret;
4536 }
4537
4538 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4539 {
4540         struct inode *inode = file_inode(iocb->ki_filp);
4541         const loff_t orig_pos = iocb->ki_pos;
4542         const size_t orig_count = iov_iter_count(from);
4543         loff_t target_size;
4544         bool dio;
4545         bool may_need_sync = true;
4546         int preallocated;
4547         ssize_t ret;
4548
4549         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4550                 ret = -EIO;
4551                 goto out;
4552         }
4553
4554         if (!f2fs_is_compress_backend_ready(inode)) {
4555                 ret = -EOPNOTSUPP;
4556                 goto out;
4557         }
4558
4559         if (iocb->ki_flags & IOCB_NOWAIT) {
4560                 if (!inode_trylock(inode)) {
4561                         ret = -EAGAIN;
4562                         goto out;
4563                 }
4564         } else {
4565                 inode_lock(inode);
4566         }
4567
4568         ret = f2fs_write_checks(iocb, from);
4569         if (ret <= 0)
4570                 goto out_unlock;
4571
4572         /* Determine whether we will do a direct write or a buffered write. */
4573         dio = f2fs_should_use_dio(inode, iocb, from);
4574
4575         /* Possibly preallocate the blocks for the write. */
4576         target_size = iocb->ki_pos + iov_iter_count(from);
4577         preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4578         if (preallocated < 0) {
4579                 ret = preallocated;
4580         } else {
4581                 if (trace_f2fs_datawrite_start_enabled()) {
4582                         char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4583                                                 PATH_MAX, GFP_KERNEL);
4584                         char *path;
4585
4586                         if (!p)
4587                                 goto skip_write_trace;
4588                         path = dentry_path_raw(file_dentry(iocb->ki_filp),
4589                                                                 p, PATH_MAX);
4590                         if (IS_ERR(path)) {
4591                                 kfree(p);
4592                                 goto skip_write_trace;
4593                         }
4594                         trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4595                                         current->pid, path, current->comm);
4596                         kfree(p);
4597                 }
4598 skip_write_trace:
4599                 /* Do the actual write. */
4600                 ret = dio ?
4601                         f2fs_dio_write_iter(iocb, from, &may_need_sync):
4602                         f2fs_buffered_write_iter(iocb, from);
4603
4604                 if (trace_f2fs_datawrite_end_enabled())
4605                         trace_f2fs_datawrite_end(inode, orig_pos, ret);
4606         }
4607
4608         /* Don't leave any preallocated blocks around past i_size. */
4609         if (preallocated && i_size_read(inode) < target_size) {
4610                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4611                 filemap_invalidate_lock(inode->i_mapping);
4612                 if (!f2fs_truncate(inode))
4613                         file_dont_truncate(inode);
4614                 filemap_invalidate_unlock(inode->i_mapping);
4615                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4616         } else {
4617                 file_dont_truncate(inode);
4618         }
4619
4620         clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4621 out_unlock:
4622         inode_unlock(inode);
4623 out:
4624         trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4625         if (ret > 0 && may_need_sync)
4626                 ret = generic_write_sync(iocb, ret);
4627         return ret;
4628 }
4629
4630 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4631                 int advice)
4632 {
4633         struct address_space *mapping;
4634         struct backing_dev_info *bdi;
4635         struct inode *inode = file_inode(filp);
4636         int err;
4637
4638         if (advice == POSIX_FADV_SEQUENTIAL) {
4639                 if (S_ISFIFO(inode->i_mode))
4640                         return -ESPIPE;
4641
4642                 mapping = filp->f_mapping;
4643                 if (!mapping || len < 0)
4644                         return -EINVAL;
4645
4646                 bdi = inode_to_bdi(mapping->host);
4647                 filp->f_ra.ra_pages = bdi->ra_pages *
4648                         F2FS_I_SB(inode)->seq_file_ra_mul;
4649                 spin_lock(&filp->f_lock);
4650                 filp->f_mode &= ~FMODE_RANDOM;
4651                 spin_unlock(&filp->f_lock);
4652                 return 0;
4653         }
4654
4655         err = generic_fadvise(filp, offset, len, advice);
4656         if (!err && advice == POSIX_FADV_DONTNEED &&
4657                 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4658                 f2fs_compressed_file(inode))
4659                 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4660
4661         return err;
4662 }
4663
4664 #ifdef CONFIG_COMPAT
4665 struct compat_f2fs_gc_range {
4666         u32 sync;
4667         compat_u64 start;
4668         compat_u64 len;
4669 };
4670 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE        _IOW(F2FS_IOCTL_MAGIC, 11,\
4671                                                 struct compat_f2fs_gc_range)
4672
4673 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4674 {
4675         struct compat_f2fs_gc_range __user *urange;
4676         struct f2fs_gc_range range;
4677         int err;
4678
4679         urange = compat_ptr(arg);
4680         err = get_user(range.sync, &urange->sync);
4681         err |= get_user(range.start, &urange->start);
4682         err |= get_user(range.len, &urange->len);
4683         if (err)
4684                 return -EFAULT;
4685
4686         return __f2fs_ioc_gc_range(file, &range);
4687 }
4688
4689 struct compat_f2fs_move_range {
4690         u32 dst_fd;
4691         compat_u64 pos_in;
4692         compat_u64 pos_out;
4693         compat_u64 len;
4694 };
4695 #define F2FS_IOC32_MOVE_RANGE           _IOWR(F2FS_IOCTL_MAGIC, 9,      \
4696                                         struct compat_f2fs_move_range)
4697
4698 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4699 {
4700         struct compat_f2fs_move_range __user *urange;
4701         struct f2fs_move_range range;
4702         int err;
4703
4704         urange = compat_ptr(arg);
4705         err = get_user(range.dst_fd, &urange->dst_fd);
4706         err |= get_user(range.pos_in, &urange->pos_in);
4707         err |= get_user(range.pos_out, &urange->pos_out);
4708         err |= get_user(range.len, &urange->len);
4709         if (err)
4710                 return -EFAULT;
4711
4712         return __f2fs_ioc_move_range(file, &range);
4713 }
4714
4715 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4716 {
4717         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4718                 return -EIO;
4719         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4720                 return -ENOSPC;
4721
4722         switch (cmd) {
4723         case FS_IOC32_GETVERSION:
4724                 cmd = FS_IOC_GETVERSION;
4725                 break;
4726         case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4727                 return f2fs_compat_ioc_gc_range(file, arg);
4728         case F2FS_IOC32_MOVE_RANGE:
4729                 return f2fs_compat_ioc_move_range(file, arg);
4730         case F2FS_IOC_START_ATOMIC_WRITE:
4731         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4732         case F2FS_IOC_START_VOLATILE_WRITE:
4733         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4734         case F2FS_IOC_ABORT_VOLATILE_WRITE:
4735         case F2FS_IOC_SHUTDOWN:
4736         case FITRIM:
4737         case FS_IOC_SET_ENCRYPTION_POLICY:
4738         case FS_IOC_GET_ENCRYPTION_PWSALT:
4739         case FS_IOC_GET_ENCRYPTION_POLICY:
4740         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4741         case FS_IOC_ADD_ENCRYPTION_KEY:
4742         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4743         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4744         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4745         case FS_IOC_GET_ENCRYPTION_NONCE:
4746         case F2FS_IOC_GARBAGE_COLLECT:
4747         case F2FS_IOC_WRITE_CHECKPOINT:
4748         case F2FS_IOC_DEFRAGMENT:
4749         case F2FS_IOC_FLUSH_DEVICE:
4750         case F2FS_IOC_GET_FEATURES:
4751         case F2FS_IOC_GET_PIN_FILE:
4752         case F2FS_IOC_SET_PIN_FILE:
4753         case F2FS_IOC_PRECACHE_EXTENTS:
4754         case F2FS_IOC_RESIZE_FS:
4755         case FS_IOC_ENABLE_VERITY:
4756         case FS_IOC_MEASURE_VERITY:
4757         case FS_IOC_READ_VERITY_METADATA:
4758         case FS_IOC_GETFSLABEL:
4759         case FS_IOC_SETFSLABEL:
4760         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4761         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4762         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4763         case F2FS_IOC_SEC_TRIM_FILE:
4764         case F2FS_IOC_GET_COMPRESS_OPTION:
4765         case F2FS_IOC_SET_COMPRESS_OPTION:
4766         case F2FS_IOC_DECOMPRESS_FILE:
4767         case F2FS_IOC_COMPRESS_FILE:
4768                 break;
4769         default:
4770                 return -ENOIOCTLCMD;
4771         }
4772         return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4773 }
4774 #endif
4775
4776 const struct file_operations f2fs_file_operations = {
4777         .llseek         = f2fs_llseek,
4778         .read_iter      = f2fs_file_read_iter,
4779         .write_iter     = f2fs_file_write_iter,
4780         .open           = f2fs_file_open,
4781         .release        = f2fs_release_file,
4782         .mmap           = f2fs_file_mmap,
4783         .flush          = f2fs_file_flush,
4784         .fsync          = f2fs_sync_file,
4785         .fallocate      = f2fs_fallocate,
4786         .unlocked_ioctl = f2fs_ioctl,
4787 #ifdef CONFIG_COMPAT
4788         .compat_ioctl   = f2fs_compat_ioctl,
4789 #endif
4790         .splice_read    = generic_file_splice_read,
4791         .splice_write   = iter_file_splice_write,
4792         .fadvise        = f2fs_file_fadvise,
4793 };