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), inode,
  47                                         APP_MAPPED_READ_IO, 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, inode, 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: truncation 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_read_extent_cache_range(dn, fofs, 0, len);
 622                 f2fs_update_age_extent_cache_range(dn, fofs, nr_free);
 623                 dec_valid_block_count(sbi, dn->inode, nr_free);
 624         }
 625         dn->ofs_in_node = ofs;
 626
 627         f2fs_update_time(sbi, REQ_TIME);
 628         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 629                                          dn->ofs_in_node, nr_free);
 630 }
 631
 632 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
 633 {
 634         f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
 635 }
 636
 637 static int truncate_partial_data_page(struct inode *inode, u64 from,
 638                                                                 bool cache_only)
 639 {
 640         loff_t offset = from & (PAGE_SIZE - 1);
 641         pgoff_t index = from >> PAGE_SHIFT;
 642         struct address_space *mapping = inode->i_mapping;
 643         struct page *page;
 644
 645         if (!offset && !cache_only)
 646                 return 0;
 647
 648         if (cache_only) {
 649                 page = find_lock_page(mapping, index);
 650                 if (page && PageUptodate(page))
 651                         goto truncate_out;
 652                 f2fs_put_page(page, 1);
 653                 return 0;
 654         }
 655
 656         page = f2fs_get_lock_data_page(inode, index, true);
 657         if (IS_ERR(page))
 658                 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 659 truncate_out:
 660         f2fs_wait_on_page_writeback(page, DATA, true, true);
 661         zero_user(page, offset, PAGE_SIZE - offset);
 662
 663         /* An encrypted inode should have a key and truncate the last page. */
 664         f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
 665         if (!cache_only)
 666                 set_page_dirty(page);
 667         f2fs_put_page(page, 1);
 668         return 0;
 669 }
 670
 671 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
 672 {
 673         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 674         struct dnode_of_data dn;
 675         pgoff_t free_from;
 676         int count = 0, err = 0;
 677         struct page *ipage;
 678         bool truncate_page = false;
 679
 680         trace_f2fs_truncate_blocks_enter(inode, from);
 681
 682         free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 683
 684         if (free_from >= max_file_blocks(inode))
 685                 goto free_partial;
 686
 687         if (lock)
 688                 f2fs_lock_op(sbi);
 689
 690         ipage = f2fs_get_node_page(sbi, inode->i_ino);
 691         if (IS_ERR(ipage)) {
 692                 err = PTR_ERR(ipage);
 693                 goto out;
 694         }
 695
 696         if (f2fs_has_inline_data(inode)) {
 697                 f2fs_truncate_inline_inode(inode, ipage, from);
 698                 f2fs_put_page(ipage, 1);
 699                 truncate_page = true;
 700                 goto out;
 701         }
 702
 703         set_new_dnode(&dn, inode, ipage, NULL, 0);
 704         err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 705         if (err) {
 706                 if (err == -ENOENT)
 707                         goto free_next;
 708                 goto out;
 709         }
 710
 711         count = ADDRS_PER_PAGE(dn.node_page, inode);
 712
 713         count -= dn.ofs_in_node;
 714         f2fs_bug_on(sbi, count < 0);
 715
 716         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 717                 f2fs_truncate_data_blocks_range(&dn, count);
 718                 free_from += count;
 719         }
 720
 721         f2fs_put_dnode(&dn);
 722 free_next:
 723         err = f2fs_truncate_inode_blocks(inode, free_from);
 724 out:
 725         if (lock)
 726                 f2fs_unlock_op(sbi);
 727 free_partial:
 728         /* lastly zero out the first data page */
 729         if (!err)
 730                 err = truncate_partial_data_page(inode, from, truncate_page);
 731
 732         trace_f2fs_truncate_blocks_exit(inode, err);
 733         return err;
 734 }
 735
 736 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
 737 {
 738         u64 free_from = from;
 739         int err;
 740
 741 #ifdef CONFIG_F2FS_FS_COMPRESSION
 742         /*
 743          * for compressed file, only support cluster size
 744          * aligned truncation.
 745          */
 746         if (f2fs_compressed_file(inode))
 747                 free_from = round_up(from,
 748                                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
 749 #endif
 750
 751         err = f2fs_do_truncate_blocks(inode, free_from, lock);
 752         if (err)
 753                 return err;
 754
 755 #ifdef CONFIG_F2FS_FS_COMPRESSION
 756         /*
 757          * For compressed file, after release compress blocks, don't allow write
 758          * direct, but we should allow write direct after truncate to zero.
 759          */
 760         if (f2fs_compressed_file(inode) && !free_from
 761                         && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
 762                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 763
 764         if (from != free_from) {
 765                 err = f2fs_truncate_partial_cluster(inode, from, lock);
 766                 if (err)
 767                         return err;
 768         }
 769 #endif
 770
 771         return 0;
 772 }
 773
 774 int f2fs_truncate(struct inode *inode)
 775 {
 776         int err;
 777
 778         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 779                 return -EIO;
 780
 781         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 782                                 S_ISLNK(inode->i_mode)))
 783                 return 0;
 784
 785         trace_f2fs_truncate(inode);
 786
 787         if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
 788                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
 789                 return -EIO;
 790         }
 791
 792         err = f2fs_dquot_initialize(inode);
 793         if (err)
 794                 return err;
 795
 796         /* we should check inline_data size */
 797         if (!f2fs_may_inline_data(inode)) {
 798                 err = f2fs_convert_inline_inode(inode);
 799                 if (err)
 800                         return err;
 801         }
 802
 803         err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
 804         if (err)
 805                 return err;
 806
 807         inode->i_mtime = inode->i_ctime = current_time(inode);
 808         f2fs_mark_inode_dirty_sync(inode, false);
 809         return 0;
 810 }
 811
 812 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
 813 {
 814         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 815
 816         if (!fscrypt_dio_supported(inode))
 817                 return true;
 818         if (fsverity_active(inode))
 819                 return true;
 820         if (f2fs_compressed_file(inode))
 821                 return true;
 822
 823         /* disallow direct IO if any of devices has unaligned blksize */
 824         if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
 825                 return true;
 826         /*
 827          * for blkzoned device, fallback direct IO to buffered IO, so
 828          * all IOs can be serialized by log-structured write.
 829          */
 830         if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
 831                 return true;
 832         if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi))
 833                 return true;
 834         if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
 835                 return true;
 836
 837         return false;
 838 }
 839
 840 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
 841                  struct kstat *stat, u32 request_mask, unsigned int query_flags)
 842 {
 843         struct inode *inode = d_inode(path->dentry);
 844         struct f2fs_inode_info *fi = F2FS_I(inode);
 845         struct f2fs_inode *ri = NULL;
 846         unsigned int flags;
 847
 848         if (f2fs_has_extra_attr(inode) &&
 849                         f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
 850                         F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 851                 stat->result_mask |= STATX_BTIME;
 852                 stat->btime.tv_sec = fi->i_crtime.tv_sec;
 853                 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 854         }
 855
 856         /*
 857          * Return the DIO alignment restrictions if requested.  We only return
 858          * this information when requested, since on encrypted files it might
 859          * take a fair bit of work to get if the file wasn't opened recently.
 860          *
 861          * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
 862          * cannot represent that, so in that case we report no DIO support.
 863          */
 864         if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
 865                 unsigned int bsize = i_blocksize(inode);
 866
 867                 stat->result_mask |= STATX_DIOALIGN;
 868                 if (!f2fs_force_buffered_io(inode, WRITE)) {
 869                         stat->dio_mem_align = bsize;
 870                         stat->dio_offset_align = bsize;
 871                 }
 872         }
 873
 874         flags = fi->i_flags;
 875         if (flags & F2FS_COMPR_FL)
 876                 stat->attributes |= STATX_ATTR_COMPRESSED;
 877         if (flags & F2FS_APPEND_FL)
 878                 stat->attributes |= STATX_ATTR_APPEND;
 879         if (IS_ENCRYPTED(inode))
 880                 stat->attributes |= STATX_ATTR_ENCRYPTED;
 881         if (flags & F2FS_IMMUTABLE_FL)
 882                 stat->attributes |= STATX_ATTR_IMMUTABLE;
 883         if (flags & F2FS_NODUMP_FL)
 884                 stat->attributes |= STATX_ATTR_NODUMP;
 885         if (IS_VERITY(inode))
 886                 stat->attributes |= STATX_ATTR_VERITY;
 887
 888         stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 889                                   STATX_ATTR_APPEND |
 890                                   STATX_ATTR_ENCRYPTED |
 891                                   STATX_ATTR_IMMUTABLE |
 892                                   STATX_ATTR_NODUMP |
 893                                   STATX_ATTR_VERITY);
 894
 895         generic_fillattr(idmap, inode, stat);
 896
 897         /* we need to show initial sectors used for inline_data/dentries */
 898         if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 899                                         f2fs_has_inline_dentry(inode))
 900                 stat->blocks += (stat->size + 511) >> 9;
 901
 902         return 0;
 903 }
 904
 905 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 906 static void __setattr_copy(struct mnt_idmap *idmap,
 907                            struct inode *inode, const struct iattr *attr)
 908 {
 909         unsigned int ia_valid = attr->ia_valid;
 910         struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
 911
 912         i_uid_update(mnt_userns, attr, inode);
 913         i_gid_update(mnt_userns, attr, inode);
 914         if (ia_valid & ATTR_ATIME)
 915                 inode->i_atime = attr->ia_atime;
 916         if (ia_valid & ATTR_MTIME)
 917                 inode->i_mtime = attr->ia_mtime;
 918         if (ia_valid & ATTR_CTIME)
 919                 inode->i_ctime = attr->ia_ctime;
 920         if (ia_valid & ATTR_MODE) {
 921                 umode_t mode = attr->ia_mode;
 922                 vfsgid_t vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
 923
 924                 if (!vfsgid_in_group_p(vfsgid) &&
 925                     !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
 926                         mode &= ~S_ISGID;
 927                 set_acl_inode(inode, mode);
 928         }
 929 }
 930 #else
 931 #define __setattr_copy setattr_copy
 932 #endif
 933
 934 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
 935                  struct iattr *attr)
 936 {
 937         struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
 938         struct inode *inode = d_inode(dentry);
 939         int err;
 940
 941         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 942                 return -EIO;
 943
 944         if (unlikely(IS_IMMUTABLE(inode)))
 945                 return -EPERM;
 946
 947         if (unlikely(IS_APPEND(inode) &&
 948                         (attr->ia_valid & (ATTR_MODE | ATTR_UID |
 949                                   ATTR_GID | ATTR_TIMES_SET))))
 950                 return -EPERM;
 951
 952         if ((attr->ia_valid & ATTR_SIZE) &&
 953                 !f2fs_is_compress_backend_ready(inode))
 954                 return -EOPNOTSUPP;
 955
 956         err = setattr_prepare(idmap, dentry, attr);
 957         if (err)
 958                 return err;
 959
 960         err = fscrypt_prepare_setattr(dentry, attr);
 961         if (err)
 962                 return err;
 963
 964         err = fsverity_prepare_setattr(dentry, attr);
 965         if (err)
 966                 return err;
 967
 968         if (is_quota_modification(mnt_userns, inode, attr)) {
 969                 err = f2fs_dquot_initialize(inode);
 970                 if (err)
 971                         return err;
 972         }
 973         if (i_uid_needs_update(mnt_userns, attr, inode) ||
 974             i_gid_needs_update(mnt_userns, attr, inode)) {
 975                 f2fs_lock_op(F2FS_I_SB(inode));
 976                 err = dquot_transfer(mnt_userns, inode, attr);
 977                 if (err) {
 978                         set_sbi_flag(F2FS_I_SB(inode),
 979                                         SBI_QUOTA_NEED_REPAIR);
 980                         f2fs_unlock_op(F2FS_I_SB(inode));
 981                         return err;
 982                 }
 983                 /*
 984                  * update uid/gid under lock_op(), so that dquot and inode can
 985                  * be updated atomically.
 986                  */
 987                 i_uid_update(mnt_userns, attr, inode);
 988                 i_gid_update(mnt_userns, attr, inode);
 989                 f2fs_mark_inode_dirty_sync(inode, true);
 990                 f2fs_unlock_op(F2FS_I_SB(inode));
 991         }
 992
 993         if (attr->ia_valid & ATTR_SIZE) {
 994                 loff_t old_size = i_size_read(inode);
 995
 996                 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
 997                         /*
 998                          * should convert inline inode before i_size_write to
 999                          * keep smaller than inline_data size with inline flag.
1000                          */
1001                         err = f2fs_convert_inline_inode(inode);
1002                         if (err)
1003                                 return err;
1004                 }
1005
1006                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1007                 filemap_invalidate_lock(inode->i_mapping);
1008
1009                 truncate_setsize(inode, attr->ia_size);
1010
1011                 if (attr->ia_size <= old_size)
1012                         err = f2fs_truncate(inode);
1013                 /*
1014                  * do not trim all blocks after i_size if target size is
1015                  * larger than i_size.
1016                  */
1017                 filemap_invalidate_unlock(inode->i_mapping);
1018                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1019                 if (err)
1020                         return err;
1021
1022                 spin_lock(&F2FS_I(inode)->i_size_lock);
1023                 inode->i_mtime = inode->i_ctime = current_time(inode);
1024                 F2FS_I(inode)->last_disk_size = i_size_read(inode);
1025                 spin_unlock(&F2FS_I(inode)->i_size_lock);
1026         }
1027
1028         __setattr_copy(idmap, inode, attr);
1029
1030         if (attr->ia_valid & ATTR_MODE) {
1031                 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1032
1033                 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1034                         if (!err)
1035                                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
1036                         clear_inode_flag(inode, FI_ACL_MODE);
1037                 }
1038         }
1039
1040         /* file size may changed here */
1041         f2fs_mark_inode_dirty_sync(inode, true);
1042
1043         /* inode change will produce dirty node pages flushed by checkpoint */
1044         f2fs_balance_fs(F2FS_I_SB(inode), true);
1045
1046         return err;
1047 }
1048
1049 const struct inode_operations f2fs_file_inode_operations = {
1050         .getattr        = f2fs_getattr,
1051         .setattr        = f2fs_setattr,
1052         .get_inode_acl  = f2fs_get_acl,
1053         .set_acl        = f2fs_set_acl,
1054         .listxattr      = f2fs_listxattr,
1055         .fiemap         = f2fs_fiemap,
1056         .fileattr_get   = f2fs_fileattr_get,
1057         .fileattr_set   = f2fs_fileattr_set,
1058 };
1059
1060 static int fill_zero(struct inode *inode, pgoff_t index,
1061                                         loff_t start, loff_t len)
1062 {
1063         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1064         struct page *page;
1065
1066         if (!len)
1067                 return 0;
1068
1069         f2fs_balance_fs(sbi, true);
1070
1071         f2fs_lock_op(sbi);
1072         page = f2fs_get_new_data_page(inode, NULL, index, false);
1073         f2fs_unlock_op(sbi);
1074
1075         if (IS_ERR(page))
1076                 return PTR_ERR(page);
1077
1078         f2fs_wait_on_page_writeback(page, DATA, true, true);
1079         zero_user(page, start, len);
1080         set_page_dirty(page);
1081         f2fs_put_page(page, 1);
1082         return 0;
1083 }
1084
1085 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1086 {
1087         int err;
1088
1089         while (pg_start < pg_end) {
1090                 struct dnode_of_data dn;
1091                 pgoff_t end_offset, count;
1092
1093                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1094                 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1095                 if (err) {
1096                         if (err == -ENOENT) {
1097                                 pg_start = f2fs_get_next_page_offset(&dn,
1098                                                                 pg_start);
1099                                 continue;
1100                         }
1101                         return err;
1102                 }
1103
1104                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1105                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1106
1107                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1108
1109                 f2fs_truncate_data_blocks_range(&dn, count);
1110                 f2fs_put_dnode(&dn);
1111
1112                 pg_start += count;
1113         }
1114         return 0;
1115 }
1116
1117 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1118 {
1119         pgoff_t pg_start, pg_end;
1120         loff_t off_start, off_end;
1121         int ret;
1122
1123         ret = f2fs_convert_inline_inode(inode);
1124         if (ret)
1125                 return ret;
1126
1127         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1128         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1129
1130         off_start = offset & (PAGE_SIZE - 1);
1131         off_end = (offset + len) & (PAGE_SIZE - 1);
1132
1133         if (pg_start == pg_end) {
1134                 ret = fill_zero(inode, pg_start, off_start,
1135                                                 off_end - off_start);
1136                 if (ret)
1137                         return ret;
1138         } else {
1139                 if (off_start) {
1140                         ret = fill_zero(inode, pg_start++, off_start,
1141                                                 PAGE_SIZE - off_start);
1142                         if (ret)
1143                                 return ret;
1144                 }
1145                 if (off_end) {
1146                         ret = fill_zero(inode, pg_end, 0, off_end);
1147                         if (ret)
1148                                 return ret;
1149                 }
1150
1151                 if (pg_start < pg_end) {
1152                         loff_t blk_start, blk_end;
1153                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1154
1155                         f2fs_balance_fs(sbi, true);
1156
1157                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
1158                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
1159
1160                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1161                         filemap_invalidate_lock(inode->i_mapping);
1162
1163                         truncate_pagecache_range(inode, blk_start, blk_end - 1);
1164
1165                         f2fs_lock_op(sbi);
1166                         ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1167                         f2fs_unlock_op(sbi);
1168
1169                         filemap_invalidate_unlock(inode->i_mapping);
1170                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1171                 }
1172         }
1173
1174         return ret;
1175 }
1176
1177 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1178                                 int *do_replace, pgoff_t off, pgoff_t len)
1179 {
1180         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1181         struct dnode_of_data dn;
1182         int ret, done, i;
1183
1184 next_dnode:
1185         set_new_dnode(&dn, inode, NULL, NULL, 0);
1186         ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1187         if (ret && ret != -ENOENT) {
1188                 return ret;
1189         } else if (ret == -ENOENT) {
1190                 if (dn.max_level == 0)
1191                         return -ENOENT;
1192                 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1193                                                 dn.ofs_in_node, len);
1194                 blkaddr += done;
1195                 do_replace += done;
1196                 goto next;
1197         }
1198
1199         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1200                                                         dn.ofs_in_node, len);
1201         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1202                 *blkaddr = f2fs_data_blkaddr(&dn);
1203
1204                 if (__is_valid_data_blkaddr(*blkaddr) &&
1205                         !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1206                                         DATA_GENERIC_ENHANCE)) {
1207                         f2fs_put_dnode(&dn);
1208                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1209                         return -EFSCORRUPTED;
1210                 }
1211
1212                 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1213
1214                         if (f2fs_lfs_mode(sbi)) {
1215                                 f2fs_put_dnode(&dn);
1216                                 return -EOPNOTSUPP;
1217                         }
1218
1219                         /* do not invalidate this block address */
1220                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1221                         *do_replace = 1;
1222                 }
1223         }
1224         f2fs_put_dnode(&dn);
1225 next:
1226         len -= done;
1227         off += done;
1228         if (len)
1229                 goto next_dnode;
1230         return 0;
1231 }
1232
1233 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1234                                 int *do_replace, pgoff_t off, int len)
1235 {
1236         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1237         struct dnode_of_data dn;
1238         int ret, i;
1239
1240         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1241                 if (*do_replace == 0)
1242                         continue;
1243
1244                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1245                 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1246                 if (ret) {
1247                         dec_valid_block_count(sbi, inode, 1);
1248                         f2fs_invalidate_blocks(sbi, *blkaddr);
1249                 } else {
1250                         f2fs_update_data_blkaddr(&dn, *blkaddr);
1251                 }
1252                 f2fs_put_dnode(&dn);
1253         }
1254         return 0;
1255 }
1256
1257 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1258                         block_t *blkaddr, int *do_replace,
1259                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1260 {
1261         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1262         pgoff_t i = 0;
1263         int ret;
1264
1265         while (i < len) {
1266                 if (blkaddr[i] == NULL_ADDR && !full) {
1267                         i++;
1268                         continue;
1269                 }
1270
1271                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1272                         struct dnode_of_data dn;
1273                         struct node_info ni;
1274                         size_t new_size;
1275                         pgoff_t ilen;
1276
1277                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1278                         ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1279                         if (ret)
1280                                 return ret;
1281
1282                         ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1283                         if (ret) {
1284                                 f2fs_put_dnode(&dn);
1285                                 return ret;
1286                         }
1287
1288                         ilen = min((pgoff_t)
1289                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1290                                                 dn.ofs_in_node, len - i);
1291                         do {
1292                                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1293                                 f2fs_truncate_data_blocks_range(&dn, 1);
1294
1295                                 if (do_replace[i]) {
1296                                         f2fs_i_blocks_write(src_inode,
1297                                                         1, false, false);
1298                                         f2fs_i_blocks_write(dst_inode,
1299                                                         1, true, false);
1300                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1301                                         blkaddr[i], ni.version, true, false);
1302
1303                                         do_replace[i] = 0;
1304                                 }
1305                                 dn.ofs_in_node++;
1306                                 i++;
1307                                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1308                                 if (dst_inode->i_size < new_size)
1309                                         f2fs_i_size_write(dst_inode, new_size);
1310                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1311
1312                         f2fs_put_dnode(&dn);
1313                 } else {
1314                         struct page *psrc, *pdst;
1315
1316                         psrc = f2fs_get_lock_data_page(src_inode,
1317                                                         src + i, true);
1318                         if (IS_ERR(psrc))
1319                                 return PTR_ERR(psrc);
1320                         pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1321                                                                 true);
1322                         if (IS_ERR(pdst)) {
1323                                 f2fs_put_page(psrc, 1);
1324                                 return PTR_ERR(pdst);
1325                         }
1326                         memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1327                         set_page_dirty(pdst);
1328                         f2fs_put_page(pdst, 1);
1329                         f2fs_put_page(psrc, 1);
1330
1331                         ret = f2fs_truncate_hole(src_inode,
1332                                                 src + i, src + i + 1);
1333                         if (ret)
1334                                 return ret;
1335                         i++;
1336                 }
1337         }
1338         return 0;
1339 }
1340
1341 static int __exchange_data_block(struct inode *src_inode,
1342                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1343                         pgoff_t len, bool full)
1344 {
1345         block_t *src_blkaddr;
1346         int *do_replace;
1347         pgoff_t olen;
1348         int ret;
1349
1350         while (len) {
1351                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1352
1353                 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1354                                         array_size(olen, sizeof(block_t)),
1355                                         GFP_NOFS);
1356                 if (!src_blkaddr)
1357                         return -ENOMEM;
1358
1359                 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1360                                         array_size(olen, sizeof(int)),
1361                                         GFP_NOFS);
1362                 if (!do_replace) {
1363                         kvfree(src_blkaddr);
1364                         return -ENOMEM;
1365                 }
1366
1367                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1368                                         do_replace, src, olen);
1369                 if (ret)
1370                         goto roll_back;
1371
1372                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1373                                         do_replace, src, dst, olen, full);
1374                 if (ret)
1375                         goto roll_back;
1376
1377                 src += olen;
1378                 dst += olen;
1379                 len -= olen;
1380
1381                 kvfree(src_blkaddr);
1382                 kvfree(do_replace);
1383         }
1384         return 0;
1385
1386 roll_back:
1387         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1388         kvfree(src_blkaddr);
1389         kvfree(do_replace);
1390         return ret;
1391 }
1392
1393 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1394 {
1395         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1396         pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1397         pgoff_t start = offset >> PAGE_SHIFT;
1398         pgoff_t end = (offset + len) >> PAGE_SHIFT;
1399         int ret;
1400
1401         f2fs_balance_fs(sbi, true);
1402
1403         /* avoid gc operation during block exchange */
1404         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1405         filemap_invalidate_lock(inode->i_mapping);
1406
1407         f2fs_lock_op(sbi);
1408         f2fs_drop_extent_tree(inode);
1409         truncate_pagecache(inode, offset);
1410         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1411         f2fs_unlock_op(sbi);
1412
1413         filemap_invalidate_unlock(inode->i_mapping);
1414         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1415         return ret;
1416 }
1417
1418 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1419 {
1420         loff_t new_size;
1421         int ret;
1422
1423         if (offset + len >= i_size_read(inode))
1424                 return -EINVAL;
1425
1426         /* collapse range should be aligned to block size of f2fs. */
1427         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1428                 return -EINVAL;
1429
1430         ret = f2fs_convert_inline_inode(inode);
1431         if (ret)
1432                 return ret;
1433
1434         /* write out all dirty pages from offset */
1435         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1436         if (ret)
1437                 return ret;
1438
1439         ret = f2fs_do_collapse(inode, offset, len);
1440         if (ret)
1441                 return ret;
1442
1443         /* write out all moved pages, if possible */
1444         filemap_invalidate_lock(inode->i_mapping);
1445         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1446         truncate_pagecache(inode, offset);
1447
1448         new_size = i_size_read(inode) - len;
1449         ret = f2fs_truncate_blocks(inode, new_size, true);
1450         filemap_invalidate_unlock(inode->i_mapping);
1451         if (!ret)
1452                 f2fs_i_size_write(inode, new_size);
1453         return ret;
1454 }
1455
1456 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1457                                                                 pgoff_t end)
1458 {
1459         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1460         pgoff_t index = start;
1461         unsigned int ofs_in_node = dn->ofs_in_node;
1462         blkcnt_t count = 0;
1463         int ret;
1464
1465         for (; index < end; index++, dn->ofs_in_node++) {
1466                 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1467                         count++;
1468         }
1469
1470         dn->ofs_in_node = ofs_in_node;
1471         ret = f2fs_reserve_new_blocks(dn, count);
1472         if (ret)
1473                 return ret;
1474
1475         dn->ofs_in_node = ofs_in_node;
1476         for (index = start; index < end; index++, dn->ofs_in_node++) {
1477                 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1478                 /*
1479                  * f2fs_reserve_new_blocks will not guarantee entire block
1480                  * allocation.
1481                  */
1482                 if (dn->data_blkaddr == NULL_ADDR) {
1483                         ret = -ENOSPC;
1484                         break;
1485                 }
1486
1487                 if (dn->data_blkaddr == NEW_ADDR)
1488                         continue;
1489
1490                 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1491                                         DATA_GENERIC_ENHANCE)) {
1492                         ret = -EFSCORRUPTED;
1493                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1494                         break;
1495                 }
1496
1497                 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1498                 dn->data_blkaddr = NEW_ADDR;
1499                 f2fs_set_data_blkaddr(dn);
1500         }
1501
1502         f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1503
1504         return ret;
1505 }
1506
1507 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1508                                                                 int mode)
1509 {
1510         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1511         struct address_space *mapping = inode->i_mapping;
1512         pgoff_t index, pg_start, pg_end;
1513         loff_t new_size = i_size_read(inode);
1514         loff_t off_start, off_end;
1515         int ret = 0;
1516
1517         ret = inode_newsize_ok(inode, (len + offset));
1518         if (ret)
1519                 return ret;
1520
1521         ret = f2fs_convert_inline_inode(inode);
1522         if (ret)
1523                 return ret;
1524
1525         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1526         if (ret)
1527                 return ret;
1528
1529         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1530         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1531
1532         off_start = offset & (PAGE_SIZE - 1);
1533         off_end = (offset + len) & (PAGE_SIZE - 1);
1534
1535         if (pg_start == pg_end) {
1536                 ret = fill_zero(inode, pg_start, off_start,
1537                                                 off_end - off_start);
1538                 if (ret)
1539                         return ret;
1540
1541                 new_size = max_t(loff_t, new_size, offset + len);
1542         } else {
1543                 if (off_start) {
1544                         ret = fill_zero(inode, pg_start++, off_start,
1545                                                 PAGE_SIZE - off_start);
1546                         if (ret)
1547                                 return ret;
1548
1549                         new_size = max_t(loff_t, new_size,
1550                                         (loff_t)pg_start << PAGE_SHIFT);
1551                 }
1552
1553                 for (index = pg_start; index < pg_end;) {
1554                         struct dnode_of_data dn;
1555                         unsigned int end_offset;
1556                         pgoff_t end;
1557
1558                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1559                         filemap_invalidate_lock(mapping);
1560
1561                         truncate_pagecache_range(inode,
1562                                 (loff_t)index << PAGE_SHIFT,
1563                                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1564
1565                         f2fs_lock_op(sbi);
1566
1567                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1568                         ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1569                         if (ret) {
1570                                 f2fs_unlock_op(sbi);
1571                                 filemap_invalidate_unlock(mapping);
1572                                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1573                                 goto out;
1574                         }
1575
1576                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1577                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1578
1579                         ret = f2fs_do_zero_range(&dn, index, end);
1580                         f2fs_put_dnode(&dn);
1581
1582                         f2fs_unlock_op(sbi);
1583                         filemap_invalidate_unlock(mapping);
1584                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1585
1586                         f2fs_balance_fs(sbi, dn.node_changed);
1587
1588                         if (ret)
1589                                 goto out;
1590
1591                         index = end;
1592                         new_size = max_t(loff_t, new_size,
1593                                         (loff_t)index << PAGE_SHIFT);
1594                 }
1595
1596                 if (off_end) {
1597                         ret = fill_zero(inode, pg_end, 0, off_end);
1598                         if (ret)
1599                                 goto out;
1600
1601                         new_size = max_t(loff_t, new_size, offset + len);
1602                 }
1603         }
1604
1605 out:
1606         if (new_size > i_size_read(inode)) {
1607                 if (mode & FALLOC_FL_KEEP_SIZE)
1608                         file_set_keep_isize(inode);
1609                 else
1610                         f2fs_i_size_write(inode, new_size);
1611         }
1612         return ret;
1613 }
1614
1615 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1616 {
1617         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1618         struct address_space *mapping = inode->i_mapping;
1619         pgoff_t nr, pg_start, pg_end, delta, idx;
1620         loff_t new_size;
1621         int ret = 0;
1622
1623         new_size = i_size_read(inode) + len;
1624         ret = inode_newsize_ok(inode, new_size);
1625         if (ret)
1626                 return ret;
1627
1628         if (offset >= i_size_read(inode))
1629                 return -EINVAL;
1630
1631         /* insert range should be aligned to block size of f2fs. */
1632         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1633                 return -EINVAL;
1634
1635         ret = f2fs_convert_inline_inode(inode);
1636         if (ret)
1637                 return ret;
1638
1639         f2fs_balance_fs(sbi, true);
1640
1641         filemap_invalidate_lock(mapping);
1642         ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1643         filemap_invalidate_unlock(mapping);
1644         if (ret)
1645                 return ret;
1646
1647         /* write out all dirty pages from offset */
1648         ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1649         if (ret)
1650                 return ret;
1651
1652         pg_start = offset >> PAGE_SHIFT;
1653         pg_end = (offset + len) >> PAGE_SHIFT;
1654         delta = pg_end - pg_start;
1655         idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1656
1657         /* avoid gc operation during block exchange */
1658         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1659         filemap_invalidate_lock(mapping);
1660         truncate_pagecache(inode, offset);
1661
1662         while (!ret && idx > pg_start) {
1663                 nr = idx - pg_start;
1664                 if (nr > delta)
1665                         nr = delta;
1666                 idx -= nr;
1667
1668                 f2fs_lock_op(sbi);
1669                 f2fs_drop_extent_tree(inode);
1670
1671                 ret = __exchange_data_block(inode, inode, idx,
1672                                         idx + delta, nr, false);
1673                 f2fs_unlock_op(sbi);
1674         }
1675         filemap_invalidate_unlock(mapping);
1676         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1677
1678         /* write out all moved pages, if possible */
1679         filemap_invalidate_lock(mapping);
1680         filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1681         truncate_pagecache(inode, offset);
1682         filemap_invalidate_unlock(mapping);
1683
1684         if (!ret)
1685                 f2fs_i_size_write(inode, new_size);
1686         return ret;
1687 }
1688
1689 static int expand_inode_data(struct inode *inode, loff_t offset,
1690                                         loff_t len, int mode)
1691 {
1692         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1693         struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1694                         .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1695                         .m_may_create = true };
1696         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1697                         .init_gc_type = FG_GC,
1698                         .should_migrate_blocks = false,
1699                         .err_gc_skipped = true,
1700                         .nr_free_secs = 0 };
1701         pgoff_t pg_start, pg_end;
1702         loff_t new_size = i_size_read(inode);
1703         loff_t off_end;
1704         block_t expanded = 0;
1705         int err;
1706
1707         err = inode_newsize_ok(inode, (len + offset));
1708         if (err)
1709                 return err;
1710
1711         err = f2fs_convert_inline_inode(inode);
1712         if (err)
1713                 return err;
1714
1715         f2fs_balance_fs(sbi, true);
1716
1717         pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1718         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1719         off_end = (offset + len) & (PAGE_SIZE - 1);
1720
1721         map.m_lblk = pg_start;
1722         map.m_len = pg_end - pg_start;
1723         if (off_end)
1724                 map.m_len++;
1725
1726         if (!map.m_len)
1727                 return 0;
1728
1729         if (f2fs_is_pinned_file(inode)) {
1730                 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1731                 block_t sec_len = roundup(map.m_len, sec_blks);
1732
1733                 map.m_len = sec_blks;
1734 next_alloc:
1735                 if (has_not_enough_free_secs(sbi, 0,
1736                         GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1737                         f2fs_down_write(&sbi->gc_lock);
1738                         err = f2fs_gc(sbi, &gc_control);
1739                         if (err && err != -ENODATA)
1740                                 goto out_err;
1741                 }
1742
1743                 f2fs_down_write(&sbi->pin_sem);
1744
1745                 f2fs_lock_op(sbi);
1746                 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1747                 f2fs_unlock_op(sbi);
1748
1749                 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1750                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1751                 file_dont_truncate(inode);
1752
1753                 f2fs_up_write(&sbi->pin_sem);
1754
1755                 expanded += map.m_len;
1756                 sec_len -= map.m_len;
1757                 map.m_lblk += map.m_len;
1758                 if (!err && sec_len)
1759                         goto next_alloc;
1760
1761                 map.m_len = expanded;
1762         } else {
1763                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1764                 expanded = map.m_len;
1765         }
1766 out_err:
1767         if (err) {
1768                 pgoff_t last_off;
1769
1770                 if (!expanded)
1771                         return err;
1772
1773                 last_off = pg_start + expanded - 1;
1774
1775                 /* update new size to the failed position */
1776                 new_size = (last_off == pg_end) ? offset + len :
1777                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1778         } else {
1779                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1780         }
1781
1782         if (new_size > i_size_read(inode)) {
1783                 if (mode & FALLOC_FL_KEEP_SIZE)
1784                         file_set_keep_isize(inode);
1785                 else
1786                         f2fs_i_size_write(inode, new_size);
1787         }
1788
1789         return err;
1790 }
1791
1792 static long f2fs_fallocate(struct file *file, int mode,
1793                                 loff_t offset, loff_t len)
1794 {
1795         struct inode *inode = file_inode(file);
1796         long ret = 0;
1797
1798         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1799                 return -EIO;
1800         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1801                 return -ENOSPC;
1802         if (!f2fs_is_compress_backend_ready(inode))
1803                 return -EOPNOTSUPP;
1804
1805         /* f2fs only support ->fallocate for regular file */
1806         if (!S_ISREG(inode->i_mode))
1807                 return -EINVAL;
1808
1809         if (IS_ENCRYPTED(inode) &&
1810                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1811                 return -EOPNOTSUPP;
1812
1813         /*
1814          * Pinned file should not support partial trucation since the block
1815          * can be used by applications.
1816          */
1817         if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1818                 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1819                         FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1820                 return -EOPNOTSUPP;
1821
1822         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1823                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1824                         FALLOC_FL_INSERT_RANGE))
1825                 return -EOPNOTSUPP;
1826
1827         inode_lock(inode);
1828
1829         ret = file_modified(file);
1830         if (ret)
1831                 goto out;
1832
1833         if (mode & FALLOC_FL_PUNCH_HOLE) {
1834                 if (offset >= inode->i_size)
1835                         goto out;
1836
1837                 ret = punch_hole(inode, offset, len);
1838         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1839                 ret = f2fs_collapse_range(inode, offset, len);
1840         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1841                 ret = f2fs_zero_range(inode, offset, len, mode);
1842         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1843                 ret = f2fs_insert_range(inode, offset, len);
1844         } else {
1845                 ret = expand_inode_data(inode, offset, len, mode);
1846         }
1847
1848         if (!ret) {
1849                 inode->i_mtime = inode->i_ctime = current_time(inode);
1850                 f2fs_mark_inode_dirty_sync(inode, false);
1851                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1852         }
1853
1854 out:
1855         inode_unlock(inode);
1856
1857         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1858         return ret;
1859 }
1860
1861 static int f2fs_release_file(struct inode *inode, struct file *filp)
1862 {
1863         /*
1864          * f2fs_relase_file is called at every close calls. So we should
1865          * not drop any inmemory pages by close called by other process.
1866          */
1867         if (!(filp->f_mode & FMODE_WRITE) ||
1868                         atomic_read(&inode->i_writecount) != 1)
1869                 return 0;
1870
1871         f2fs_abort_atomic_write(inode, true);
1872         return 0;
1873 }
1874
1875 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1876 {
1877         struct inode *inode = file_inode(file);
1878
1879         /*
1880          * If the process doing a transaction is crashed, we should do
1881          * roll-back. Otherwise, other reader/write can see corrupted database
1882          * until all the writers close its file. Since this should be done
1883          * before dropping file lock, it needs to do in ->flush.
1884          */
1885         if (F2FS_I(inode)->atomic_write_task == current)
1886                 f2fs_abort_atomic_write(inode, true);
1887         return 0;
1888 }
1889
1890 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1891 {
1892         struct f2fs_inode_info *fi = F2FS_I(inode);
1893         u32 masked_flags = fi->i_flags & mask;
1894
1895         /* mask can be shrunk by flags_valid selector */
1896         iflags &= mask;
1897
1898         /* Is it quota file? Do not allow user to mess with it */
1899         if (IS_NOQUOTA(inode))
1900                 return -EPERM;
1901
1902         if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1903                 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1904                         return -EOPNOTSUPP;
1905                 if (!f2fs_empty_dir(inode))
1906                         return -ENOTEMPTY;
1907         }
1908
1909         if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1910                 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1911                         return -EOPNOTSUPP;
1912                 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1913                         return -EINVAL;
1914         }
1915
1916         if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1917                 if (masked_flags & F2FS_COMPR_FL) {
1918                         if (!f2fs_disable_compressed_file(inode))
1919                                 return -EINVAL;
1920                 } else {
1921                         /* try to convert inline_data to support compression */
1922                         int err = f2fs_convert_inline_inode(inode);
1923                         if (err)
1924                                 return err;
1925                         if (!f2fs_may_compress(inode))
1926                                 return -EINVAL;
1927                         if (S_ISREG(inode->i_mode) && F2FS_HAS_BLOCKS(inode))
1928                                 return -EINVAL;
1929                         if (set_compress_context(inode))
1930                                 return -EOPNOTSUPP;
1931                 }
1932         }
1933
1934         fi->i_flags = iflags | (fi->i_flags & ~mask);
1935         f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1936                                         (fi->i_flags & F2FS_NOCOMP_FL));
1937
1938         if (fi->i_flags & F2FS_PROJINHERIT_FL)
1939                 set_inode_flag(inode, FI_PROJ_INHERIT);
1940         else
1941                 clear_inode_flag(inode, FI_PROJ_INHERIT);
1942
1943         inode->i_ctime = current_time(inode);
1944         f2fs_set_inode_flags(inode);
1945         f2fs_mark_inode_dirty_sync(inode, true);
1946         return 0;
1947 }
1948
1949 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1950
1951 /*
1952  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1953  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1954  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1955  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1956  *
1957  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1958  * FS_IOC_FSSETXATTR is done by the VFS.
1959  */
1960
1961 static const struct {
1962         u32 iflag;
1963         u32 fsflag;
1964 } f2fs_fsflags_map[] = {
1965         { F2FS_COMPR_FL,        FS_COMPR_FL },
1966         { F2FS_SYNC_FL,         FS_SYNC_FL },
1967         { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1968         { F2FS_APPEND_FL,       FS_APPEND_FL },
1969         { F2FS_NODUMP_FL,       FS_NODUMP_FL },
1970         { F2FS_NOATIME_FL,      FS_NOATIME_FL },
1971         { F2FS_NOCOMP_FL,       FS_NOCOMP_FL },
1972         { F2FS_INDEX_FL,        FS_INDEX_FL },
1973         { F2FS_DIRSYNC_FL,      FS_DIRSYNC_FL },
1974         { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1975         { F2FS_CASEFOLD_FL,     FS_CASEFOLD_FL },
1976 };
1977
1978 #define F2FS_GETTABLE_FS_FL (           \
1979                 FS_COMPR_FL |           \
1980                 FS_SYNC_FL |            \
1981                 FS_IMMUTABLE_FL |       \
1982                 FS_APPEND_FL |          \
1983                 FS_NODUMP_FL |          \
1984                 FS_NOATIME_FL |         \
1985                 FS_NOCOMP_FL |          \
1986                 FS_INDEX_FL |           \
1987                 FS_DIRSYNC_FL |         \
1988                 FS_PROJINHERIT_FL |     \
1989                 FS_ENCRYPT_FL |         \
1990                 FS_INLINE_DATA_FL |     \
1991                 FS_NOCOW_FL |           \
1992                 FS_VERITY_FL |          \
1993                 FS_CASEFOLD_FL)
1994
1995 #define F2FS_SETTABLE_FS_FL (           \
1996                 FS_COMPR_FL |           \
1997                 FS_SYNC_FL |            \
1998                 FS_IMMUTABLE_FL |       \
1999                 FS_APPEND_FL |          \
2000                 FS_NODUMP_FL |          \
2001                 FS_NOATIME_FL |         \
2002                 FS_NOCOMP_FL |          \
2003                 FS_DIRSYNC_FL |         \
2004                 FS_PROJINHERIT_FL |     \
2005                 FS_CASEFOLD_FL)
2006
2007 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2008 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2009 {
2010         u32 fsflags = 0;
2011         int i;
2012
2013         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2014                 if (iflags & f2fs_fsflags_map[i].iflag)
2015                         fsflags |= f2fs_fsflags_map[i].fsflag;
2016
2017         return fsflags;
2018 }
2019
2020 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2021 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2022 {
2023         u32 iflags = 0;
2024         int i;
2025
2026         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2027                 if (fsflags & f2fs_fsflags_map[i].fsflag)
2028                         iflags |= f2fs_fsflags_map[i].iflag;
2029
2030         return iflags;
2031 }
2032
2033 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2034 {
2035         struct inode *inode = file_inode(filp);
2036
2037         return put_user(inode->i_generation, (int __user *)arg);
2038 }
2039
2040 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2041 {
2042         struct inode *inode = file_inode(filp);
2043         struct mnt_idmap *idmap = file_mnt_idmap(filp);
2044         struct f2fs_inode_info *fi = F2FS_I(inode);
2045         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2046         struct inode *pinode;
2047         loff_t isize;
2048         int ret;
2049
2050         if (!inode_owner_or_capable(idmap, inode))
2051                 return -EACCES;
2052
2053         if (!S_ISREG(inode->i_mode))
2054                 return -EINVAL;
2055
2056         if (filp->f_flags & O_DIRECT)
2057                 return -EINVAL;
2058
2059         ret = mnt_want_write_file(filp);
2060         if (ret)
2061                 return ret;
2062
2063         inode_lock(inode);
2064
2065         if (!f2fs_disable_compressed_file(inode)) {
2066                 ret = -EINVAL;
2067                 goto out;
2068         }
2069
2070         if (f2fs_is_atomic_file(inode))
2071                 goto out;
2072
2073         ret = f2fs_convert_inline_inode(inode);
2074         if (ret)
2075                 goto out;
2076
2077         f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2078
2079         /*
2080          * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2081          * f2fs_is_atomic_file.
2082          */
2083         if (get_dirty_pages(inode))
2084                 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2085                           inode->i_ino, get_dirty_pages(inode));
2086         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2087         if (ret) {
2088                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2089                 goto out;
2090         }
2091
2092         /* Create a COW inode for atomic write */
2093         pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2094         if (IS_ERR(pinode)) {
2095                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2096                 ret = PTR_ERR(pinode);
2097                 goto out;
2098         }
2099
2100         ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode);
2101         iput(pinode);
2102         if (ret) {
2103                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2104                 goto out;
2105         }
2106
2107         f2fs_write_inode(inode, NULL);
2108
2109         stat_inc_atomic_inode(inode);
2110
2111         set_inode_flag(inode, FI_ATOMIC_FILE);
2112         set_inode_flag(fi->cow_inode, FI_COW_FILE);
2113         clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2114
2115         isize = i_size_read(inode);
2116         fi->original_i_size = isize;
2117         if (truncate) {
2118                 set_inode_flag(inode, FI_ATOMIC_REPLACE);
2119                 truncate_inode_pages_final(inode->i_mapping);
2120                 f2fs_i_size_write(inode, 0);
2121                 isize = 0;
2122         }
2123         f2fs_i_size_write(fi->cow_inode, isize);
2124
2125         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2126
2127         f2fs_update_time(sbi, REQ_TIME);
2128         fi->atomic_write_task = current;
2129         stat_update_max_atomic_write(inode);
2130         fi->atomic_write_cnt = 0;
2131 out:
2132         inode_unlock(inode);
2133         mnt_drop_write_file(filp);
2134         return ret;
2135 }
2136
2137 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2138 {
2139         struct inode *inode = file_inode(filp);
2140         struct mnt_idmap *idmap = file_mnt_idmap(filp);
2141         int ret;
2142
2143         if (!inode_owner_or_capable(idmap, inode))
2144                 return -EACCES;
2145
2146         ret = mnt_want_write_file(filp);
2147         if (ret)
2148                 return ret;
2149
2150         f2fs_balance_fs(F2FS_I_SB(inode), true);
2151
2152         inode_lock(inode);
2153
2154         if (f2fs_is_atomic_file(inode)) {
2155                 ret = f2fs_commit_atomic_write(inode);
2156                 if (!ret)
2157                         ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2158
2159                 f2fs_abort_atomic_write(inode, ret);
2160         } else {
2161                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2162         }
2163
2164         inode_unlock(inode);
2165         mnt_drop_write_file(filp);
2166         return ret;
2167 }
2168
2169 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2170 {
2171         struct inode *inode = file_inode(filp);
2172         struct mnt_idmap *idmap = file_mnt_idmap(filp);
2173         int ret;
2174
2175         if (!inode_owner_or_capable(idmap, inode))
2176                 return -EACCES;
2177
2178         ret = mnt_want_write_file(filp);
2179         if (ret)
2180                 return ret;
2181
2182         inode_lock(inode);
2183
2184         f2fs_abort_atomic_write(inode, true);
2185
2186         inode_unlock(inode);
2187
2188         mnt_drop_write_file(filp);
2189         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2190         return ret;
2191 }
2192
2193 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2194 {
2195         struct inode *inode = file_inode(filp);
2196         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2197         struct super_block *sb = sbi->sb;
2198         __u32 in;
2199         int ret = 0;
2200
2201         if (!capable(CAP_SYS_ADMIN))
2202                 return -EPERM;
2203
2204         if (get_user(in, (__u32 __user *)arg))
2205                 return -EFAULT;
2206
2207         if (in != F2FS_GOING_DOWN_FULLSYNC) {
2208                 ret = mnt_want_write_file(filp);
2209                 if (ret) {
2210                         if (ret == -EROFS) {
2211                                 ret = 0;
2212                                 f2fs_stop_checkpoint(sbi, false,
2213                                                 STOP_CP_REASON_SHUTDOWN);
2214                                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2215                                 trace_f2fs_shutdown(sbi, in, ret);
2216                         }
2217                         return ret;
2218                 }
2219         }
2220
2221         switch (in) {
2222         case F2FS_GOING_DOWN_FULLSYNC:
2223                 ret = freeze_bdev(sb->s_bdev);
2224                 if (ret)
2225                         goto out;
2226                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2227                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2228                 thaw_bdev(sb->s_bdev);
2229                 break;
2230         case F2FS_GOING_DOWN_METASYNC:
2231                 /* do checkpoint only */
2232                 ret = f2fs_sync_fs(sb, 1);
2233                 if (ret)
2234                         goto out;
2235                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2236                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2237                 break;
2238         case F2FS_GOING_DOWN_NOSYNC:
2239                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2240                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2241                 break;
2242         case F2FS_GOING_DOWN_METAFLUSH:
2243                 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2244                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2245                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2246                 break;
2247         case F2FS_GOING_DOWN_NEED_FSCK:
2248                 set_sbi_flag(sbi, SBI_NEED_FSCK);
2249                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2250                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2251                 /* do checkpoint only */
2252                 ret = f2fs_sync_fs(sb, 1);
2253                 goto out;
2254         default:
2255                 ret = -EINVAL;
2256                 goto out;
2257         }
2258
2259         f2fs_stop_gc_thread(sbi);
2260         f2fs_stop_discard_thread(sbi);
2261
2262         f2fs_drop_discard_cmd(sbi);
2263         clear_opt(sbi, DISCARD);
2264
2265         f2fs_update_time(sbi, REQ_TIME);
2266 out:
2267         if (in != F2FS_GOING_DOWN_FULLSYNC)
2268                 mnt_drop_write_file(filp);
2269
2270         trace_f2fs_shutdown(sbi, in, ret);
2271
2272         return ret;
2273 }
2274
2275 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2276 {
2277         struct inode *inode = file_inode(filp);
2278         struct super_block *sb = inode->i_sb;
2279         struct fstrim_range range;
2280         int ret;
2281
2282         if (!capable(CAP_SYS_ADMIN))
2283                 return -EPERM;
2284
2285         if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2286                 return -EOPNOTSUPP;
2287
2288         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2289                                 sizeof(range)))
2290                 return -EFAULT;
2291
2292         ret = mnt_want_write_file(filp);
2293         if (ret)
2294                 return ret;
2295
2296         range.minlen = max((unsigned int)range.minlen,
2297                            bdev_discard_granularity(sb->s_bdev));
2298         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2299         mnt_drop_write_file(filp);
2300         if (ret < 0)
2301                 return ret;
2302
2303         if (copy_to_user((struct fstrim_range __user *)arg, &range,
2304                                 sizeof(range)))
2305                 return -EFAULT;
2306         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2307         return 0;
2308 }
2309
2310 static bool uuid_is_nonzero(__u8 u[16])
2311 {
2312         int i;
2313
2314         for (i = 0; i < 16; i++)
2315                 if (u[i])
2316                         return true;
2317         return false;
2318 }
2319
2320 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2321 {
2322         struct inode *inode = file_inode(filp);
2323
2324         if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2325                 return -EOPNOTSUPP;
2326
2327         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2328
2329         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2330 }
2331
2332 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2333 {
2334         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2335                 return -EOPNOTSUPP;
2336         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2337 }
2338
2339 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2340 {
2341         struct inode *inode = file_inode(filp);
2342         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2343         int err;
2344
2345         if (!f2fs_sb_has_encrypt(sbi))
2346                 return -EOPNOTSUPP;
2347
2348         err = mnt_want_write_file(filp);
2349         if (err)
2350                 return err;
2351
2352         f2fs_down_write(&sbi->sb_lock);
2353
2354         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2355                 goto got_it;
2356
2357         /* update superblock with uuid */
2358         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2359
2360         err = f2fs_commit_super(sbi, false);
2361         if (err) {
2362                 /* undo new data */
2363                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2364                 goto out_err;
2365         }
2366 got_it:
2367         if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2368                                                                         16))
2369                 err = -EFAULT;
2370 out_err:
2371         f2fs_up_write(&sbi->sb_lock);
2372         mnt_drop_write_file(filp);
2373         return err;
2374 }
2375
2376 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2377                                              unsigned long arg)
2378 {
2379         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2380                 return -EOPNOTSUPP;
2381
2382         return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2383 }
2384
2385 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2386 {
2387         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2388                 return -EOPNOTSUPP;
2389
2390         return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2391 }
2392
2393 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2394 {
2395         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2396                 return -EOPNOTSUPP;
2397
2398         return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2399 }
2400
2401 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2402                                                     unsigned long arg)
2403 {
2404         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2405                 return -EOPNOTSUPP;
2406
2407         return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2408 }
2409
2410 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2411                                               unsigned long arg)
2412 {
2413         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2414                 return -EOPNOTSUPP;
2415
2416         return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2417 }
2418
2419 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2420 {
2421         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2422                 return -EOPNOTSUPP;
2423
2424         return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2425 }
2426
2427 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2428 {
2429         struct inode *inode = file_inode(filp);
2430         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2431         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2432                         .no_bg_gc = false,
2433                         .should_migrate_blocks = false,
2434                         .nr_free_secs = 0 };
2435         __u32 sync;
2436         int ret;
2437
2438         if (!capable(CAP_SYS_ADMIN))
2439                 return -EPERM;
2440
2441         if (get_user(sync, (__u32 __user *)arg))
2442                 return -EFAULT;
2443
2444         if (f2fs_readonly(sbi->sb))
2445                 return -EROFS;
2446
2447         ret = mnt_want_write_file(filp);
2448         if (ret)
2449                 return ret;
2450
2451         if (!sync) {
2452                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2453                         ret = -EBUSY;
2454                         goto out;
2455                 }
2456         } else {
2457                 f2fs_down_write(&sbi->gc_lock);
2458         }
2459
2460         gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2461         gc_control.err_gc_skipped = sync;
2462         ret = f2fs_gc(sbi, &gc_control);
2463 out:
2464         mnt_drop_write_file(filp);
2465         return ret;
2466 }
2467
2468 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2469 {
2470         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2471         struct f2fs_gc_control gc_control = {
2472                         .init_gc_type = range->sync ? FG_GC : BG_GC,
2473                         .no_bg_gc = false,
2474                         .should_migrate_blocks = false,
2475                         .err_gc_skipped = range->sync,
2476                         .nr_free_secs = 0 };
2477         u64 end;
2478         int ret;
2479
2480         if (!capable(CAP_SYS_ADMIN))
2481                 return -EPERM;
2482         if (f2fs_readonly(sbi->sb))
2483                 return -EROFS;
2484
2485         end = range->start + range->len;
2486         if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2487                                         end >= MAX_BLKADDR(sbi))
2488                 return -EINVAL;
2489
2490         ret = mnt_want_write_file(filp);
2491         if (ret)
2492                 return ret;
2493
2494 do_more:
2495         if (!range->sync) {
2496                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2497                         ret = -EBUSY;
2498                         goto out;
2499                 }
2500         } else {
2501                 f2fs_down_write(&sbi->gc_lock);
2502         }
2503
2504         gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2505         ret = f2fs_gc(sbi, &gc_control);
2506         if (ret) {
2507                 if (ret == -EBUSY)
2508                         ret = -EAGAIN;
2509                 goto out;
2510         }
2511         range->start += CAP_BLKS_PER_SEC(sbi);
2512         if (range->start <= end)
2513                 goto do_more;
2514 out:
2515         mnt_drop_write_file(filp);
2516         return ret;
2517 }
2518
2519 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2520 {
2521         struct f2fs_gc_range range;
2522
2523         if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2524                                                         sizeof(range)))
2525                 return -EFAULT;
2526         return __f2fs_ioc_gc_range(filp, &range);
2527 }
2528
2529 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2530 {
2531         struct inode *inode = file_inode(filp);
2532         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2533         int ret;
2534
2535         if (!capable(CAP_SYS_ADMIN))
2536                 return -EPERM;
2537
2538         if (f2fs_readonly(sbi->sb))
2539                 return -EROFS;
2540
2541         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2542                 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2543                 return -EINVAL;
2544         }
2545
2546         ret = mnt_want_write_file(filp);
2547         if (ret)
2548                 return ret;
2549
2550         ret = f2fs_sync_fs(sbi->sb, 1);
2551
2552         mnt_drop_write_file(filp);
2553         return ret;
2554 }
2555
2556 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2557                                         struct file *filp,
2558                                         struct f2fs_defragment *range)
2559 {
2560         struct inode *inode = file_inode(filp);
2561         struct f2fs_map_blocks map = { .m_next_extent = NULL,
2562                                         .m_seg_type = NO_CHECK_TYPE,
2563                                         .m_may_create = false };
2564         struct extent_info ei = {0, };
2565         pgoff_t pg_start, pg_end, next_pgofs;
2566         unsigned int blk_per_seg = sbi->blocks_per_seg;
2567         unsigned int total = 0, sec_num;
2568         block_t blk_end = 0;
2569         bool fragmented = false;
2570         int err;
2571
2572         pg_start = range->start >> PAGE_SHIFT;
2573         pg_end = (range->start + range->len) >> PAGE_SHIFT;
2574
2575         f2fs_balance_fs(sbi, true);
2576
2577         inode_lock(inode);
2578
2579         /* if in-place-update policy is enabled, don't waste time here */
2580         set_inode_flag(inode, FI_OPU_WRITE);
2581         if (f2fs_should_update_inplace(inode, NULL)) {
2582                 err = -EINVAL;
2583                 goto out;
2584         }
2585
2586         /* writeback all dirty pages in the range */
2587         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2588                                                 range->start + range->len - 1);
2589         if (err)
2590                 goto out;
2591
2592         /*
2593          * lookup mapping info in extent cache, skip defragmenting if physical
2594          * block addresses are continuous.
2595          */
2596         if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2597                 if (ei.fofs + ei.len >= pg_end)
2598                         goto out;
2599         }
2600
2601         map.m_lblk = pg_start;
2602         map.m_next_pgofs = &next_pgofs;
2603
2604         /*
2605          * lookup mapping info in dnode page cache, skip defragmenting if all
2606          * physical block addresses are continuous even if there are hole(s)
2607          * in logical blocks.
2608          */
2609         while (map.m_lblk < pg_end) {
2610                 map.m_len = pg_end - map.m_lblk;
2611                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2612                 if (err)
2613                         goto out;
2614
2615                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2616                         map.m_lblk = next_pgofs;
2617                         continue;
2618                 }
2619
2620                 if (blk_end && blk_end != map.m_pblk)
2621                         fragmented = true;
2622
2623                 /* record total count of block that we're going to move */
2624                 total += map.m_len;
2625
2626                 blk_end = map.m_pblk + map.m_len;
2627
2628                 map.m_lblk += map.m_len;
2629         }
2630
2631         if (!fragmented) {
2632                 total = 0;
2633                 goto out;
2634         }
2635
2636         sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2637
2638         /*
2639          * make sure there are enough free section for LFS allocation, this can
2640          * avoid defragment running in SSR mode when free section are allocated
2641          * intensively
2642          */
2643         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2644                 err = -EAGAIN;
2645                 goto out;
2646         }
2647
2648         map.m_lblk = pg_start;
2649         map.m_len = pg_end - pg_start;
2650         total = 0;
2651
2652         while (map.m_lblk < pg_end) {
2653                 pgoff_t idx;
2654                 int cnt = 0;
2655
2656 do_map:
2657                 map.m_len = pg_end - map.m_lblk;
2658                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2659                 if (err)
2660                         goto clear_out;
2661
2662                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2663                         map.m_lblk = next_pgofs;
2664                         goto check;
2665                 }
2666
2667                 set_inode_flag(inode, FI_SKIP_WRITES);
2668
2669                 idx = map.m_lblk;
2670                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2671                         struct page *page;
2672
2673                         page = f2fs_get_lock_data_page(inode, idx, true);
2674                         if (IS_ERR(page)) {
2675                                 err = PTR_ERR(page);
2676                                 goto clear_out;
2677                         }
2678
2679                         set_page_dirty(page);
2680                         set_page_private_gcing(page);
2681                         f2fs_put_page(page, 1);
2682
2683                         idx++;
2684                         cnt++;
2685                         total++;
2686                 }
2687
2688                 map.m_lblk = idx;
2689 check:
2690                 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2691                         goto do_map;
2692
2693                 clear_inode_flag(inode, FI_SKIP_WRITES);
2694
2695                 err = filemap_fdatawrite(inode->i_mapping);
2696                 if (err)
2697                         goto out;
2698         }
2699 clear_out:
2700         clear_inode_flag(inode, FI_SKIP_WRITES);
2701 out:
2702         clear_inode_flag(inode, FI_OPU_WRITE);
2703         inode_unlock(inode);
2704         if (!err)
2705                 range->len = (u64)total << PAGE_SHIFT;
2706         return err;
2707 }
2708
2709 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2710 {
2711         struct inode *inode = file_inode(filp);
2712         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2713         struct f2fs_defragment range;
2714         int err;
2715
2716         if (!capable(CAP_SYS_ADMIN))
2717                 return -EPERM;
2718
2719         if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2720                 return -EINVAL;
2721
2722         if (f2fs_readonly(sbi->sb))
2723                 return -EROFS;
2724
2725         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2726                                                         sizeof(range)))
2727                 return -EFAULT;
2728
2729         /* verify alignment of offset & size */
2730         if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2731                 return -EINVAL;
2732
2733         if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2734                                         max_file_blocks(inode)))
2735                 return -EINVAL;
2736
2737         err = mnt_want_write_file(filp);
2738         if (err)
2739                 return err;
2740
2741         err = f2fs_defragment_range(sbi, filp, &range);
2742         mnt_drop_write_file(filp);
2743
2744         f2fs_update_time(sbi, REQ_TIME);
2745         if (err < 0)
2746                 return err;
2747
2748         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2749                                                         sizeof(range)))
2750                 return -EFAULT;
2751
2752         return 0;
2753 }
2754
2755 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2756                         struct file *file_out, loff_t pos_out, size_t len)
2757 {
2758         struct inode *src = file_inode(file_in);
2759         struct inode *dst = file_inode(file_out);
2760         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2761         size_t olen = len, dst_max_i_size = 0;
2762         size_t dst_osize;
2763         int ret;
2764
2765         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2766                                 src->i_sb != dst->i_sb)
2767                 return -EXDEV;
2768
2769         if (unlikely(f2fs_readonly(src->i_sb)))
2770                 return -EROFS;
2771
2772         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2773                 return -EINVAL;
2774
2775         if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2776                 return -EOPNOTSUPP;
2777
2778         if (pos_out < 0 || pos_in < 0)
2779                 return -EINVAL;
2780
2781         if (src == dst) {
2782                 if (pos_in == pos_out)
2783                         return 0;
2784                 if (pos_out > pos_in && pos_out < pos_in + len)
2785                         return -EINVAL;
2786         }
2787
2788         inode_lock(src);
2789         if (src != dst) {
2790                 ret = -EBUSY;
2791                 if (!inode_trylock(dst))
2792                         goto out;
2793         }
2794
2795         ret = -EINVAL;
2796         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2797                 goto out_unlock;
2798         if (len == 0)
2799                 olen = len = src->i_size - pos_in;
2800         if (pos_in + len == src->i_size)
2801                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2802         if (len == 0) {
2803                 ret = 0;
2804                 goto out_unlock;
2805         }
2806
2807         dst_osize = dst->i_size;
2808         if (pos_out + olen > dst->i_size)
2809                 dst_max_i_size = pos_out + olen;
2810
2811         /* verify the end result is block aligned */
2812         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2813                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2814                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2815                 goto out_unlock;
2816
2817         ret = f2fs_convert_inline_inode(src);
2818         if (ret)
2819                 goto out_unlock;
2820
2821         ret = f2fs_convert_inline_inode(dst);
2822         if (ret)
2823                 goto out_unlock;
2824
2825         /* write out all dirty pages from offset */
2826         ret = filemap_write_and_wait_range(src->i_mapping,
2827                                         pos_in, pos_in + len);
2828         if (ret)
2829                 goto out_unlock;
2830
2831         ret = filemap_write_and_wait_range(dst->i_mapping,
2832                                         pos_out, pos_out + len);
2833         if (ret)
2834                 goto out_unlock;
2835
2836         f2fs_balance_fs(sbi, true);
2837
2838         f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2839         if (src != dst) {
2840                 ret = -EBUSY;
2841                 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2842                         goto out_src;
2843         }
2844
2845         f2fs_lock_op(sbi);
2846         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2847                                 pos_out >> F2FS_BLKSIZE_BITS,
2848                                 len >> F2FS_BLKSIZE_BITS, false);
2849
2850         if (!ret) {
2851                 if (dst_max_i_size)
2852                         f2fs_i_size_write(dst, dst_max_i_size);
2853                 else if (dst_osize != dst->i_size)
2854                         f2fs_i_size_write(dst, dst_osize);
2855         }
2856         f2fs_unlock_op(sbi);
2857
2858         if (src != dst)
2859                 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2860 out_src:
2861         f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2862 out_unlock:
2863         if (src != dst)
2864                 inode_unlock(dst);
2865 out:
2866         inode_unlock(src);
2867         return ret;
2868 }
2869
2870 static int __f2fs_ioc_move_range(struct file *filp,
2871                                 struct f2fs_move_range *range)
2872 {
2873         struct fd dst;
2874         int err;
2875
2876         if (!(filp->f_mode & FMODE_READ) ||
2877                         !(filp->f_mode & FMODE_WRITE))
2878                 return -EBADF;
2879
2880         dst = fdget(range->dst_fd);
2881         if (!dst.file)
2882                 return -EBADF;
2883
2884         if (!(dst.file->f_mode & FMODE_WRITE)) {
2885                 err = -EBADF;
2886                 goto err_out;
2887         }
2888
2889         err = mnt_want_write_file(filp);
2890         if (err)
2891                 goto err_out;
2892
2893         err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2894                                         range->pos_out, range->len);
2895
2896         mnt_drop_write_file(filp);
2897 err_out:
2898         fdput(dst);
2899         return err;
2900 }
2901
2902 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2903 {
2904         struct f2fs_move_range range;
2905
2906         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2907                                                         sizeof(range)))
2908                 return -EFAULT;
2909         return __f2fs_ioc_move_range(filp, &range);
2910 }
2911
2912 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2913 {
2914         struct inode *inode = file_inode(filp);
2915         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2916         struct sit_info *sm = SIT_I(sbi);
2917         unsigned int start_segno = 0, end_segno = 0;
2918         unsigned int dev_start_segno = 0, dev_end_segno = 0;
2919         struct f2fs_flush_device range;
2920         struct f2fs_gc_control gc_control = {
2921                         .init_gc_type = FG_GC,
2922                         .should_migrate_blocks = true,
2923                         .err_gc_skipped = true,
2924                         .nr_free_secs = 0 };
2925         int ret;
2926
2927         if (!capable(CAP_SYS_ADMIN))
2928                 return -EPERM;
2929
2930         if (f2fs_readonly(sbi->sb))
2931                 return -EROFS;
2932
2933         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2934                 return -EINVAL;
2935
2936         if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2937                                                         sizeof(range)))
2938                 return -EFAULT;
2939
2940         if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2941                         __is_large_section(sbi)) {
2942                 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2943                           range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2944                 return -EINVAL;
2945         }
2946
2947         ret = mnt_want_write_file(filp);
2948         if (ret)
2949                 return ret;
2950
2951         if (range.dev_num != 0)
2952                 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2953         dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2954
2955         start_segno = sm->last_victim[FLUSH_DEVICE];
2956         if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2957                 start_segno = dev_start_segno;
2958         end_segno = min(start_segno + range.segments, dev_end_segno);
2959
2960         while (start_segno < end_segno) {
2961                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2962                         ret = -EBUSY;
2963                         goto out;
2964                 }
2965                 sm->last_victim[GC_CB] = end_segno + 1;
2966                 sm->last_victim[GC_GREEDY] = end_segno + 1;
2967                 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2968
2969                 gc_control.victim_segno = start_segno;
2970                 ret = f2fs_gc(sbi, &gc_control);
2971                 if (ret == -EAGAIN)
2972                         ret = 0;
2973                 else if (ret < 0)
2974                         break;
2975                 start_segno++;
2976         }
2977 out:
2978         mnt_drop_write_file(filp);
2979         return ret;
2980 }
2981
2982 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2983 {
2984         struct inode *inode = file_inode(filp);
2985         u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2986
2987         /* Must validate to set it with SQLite behavior in Android. */
2988         sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2989
2990         return put_user(sb_feature, (u32 __user *)arg);
2991 }
2992
2993 #ifdef CONFIG_QUOTA
2994 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2995 {
2996         struct dquot *transfer_to[MAXQUOTAS] = {};
2997         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2998         struct super_block *sb = sbi->sb;
2999         int err = 0;
3000
3001         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3002         if (!IS_ERR(transfer_to[PRJQUOTA])) {
3003                 err = __dquot_transfer(inode, transfer_to);
3004                 if (err)
3005                         set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3006                 dqput(transfer_to[PRJQUOTA]);
3007         }
3008         return err;
3009 }
3010
3011 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3012 {
3013         struct f2fs_inode_info *fi = F2FS_I(inode);
3014         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3015         struct f2fs_inode *ri = NULL;
3016         kprojid_t kprojid;
3017         int err;
3018
3019         if (!f2fs_sb_has_project_quota(sbi)) {
3020                 if (projid != F2FS_DEF_PROJID)
3021                         return -EOPNOTSUPP;
3022                 else
3023                         return 0;
3024         }
3025
3026         if (!f2fs_has_extra_attr(inode))
3027                 return -EOPNOTSUPP;
3028
3029         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3030
3031         if (projid_eq(kprojid, fi->i_projid))
3032                 return 0;
3033
3034         err = -EPERM;
3035         /* Is it quota file? Do not allow user to mess with it */
3036         if (IS_NOQUOTA(inode))
3037                 return err;
3038
3039         if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3040                 return -EOVERFLOW;
3041
3042         err = f2fs_dquot_initialize(inode);
3043         if (err)
3044                 return err;
3045
3046         f2fs_lock_op(sbi);
3047         err = f2fs_transfer_project_quota(inode, kprojid);
3048         if (err)
3049                 goto out_unlock;
3050
3051         fi->i_projid = kprojid;
3052         inode->i_ctime = current_time(inode);
3053         f2fs_mark_inode_dirty_sync(inode, true);
3054 out_unlock:
3055         f2fs_unlock_op(sbi);
3056         return err;
3057 }
3058 #else
3059 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3060 {
3061         return 0;
3062 }
3063
3064 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3065 {
3066         if (projid != F2FS_DEF_PROJID)
3067                 return -EOPNOTSUPP;
3068         return 0;
3069 }
3070 #endif
3071
3072 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3073 {
3074         struct inode *inode = d_inode(dentry);
3075         struct f2fs_inode_info *fi = F2FS_I(inode);
3076         u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3077
3078         if (IS_ENCRYPTED(inode))
3079                 fsflags |= FS_ENCRYPT_FL;
3080         if (IS_VERITY(inode))
3081                 fsflags |= FS_VERITY_FL;
3082         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3083                 fsflags |= FS_INLINE_DATA_FL;
3084         if (is_inode_flag_set(inode, FI_PIN_FILE))
3085                 fsflags |= FS_NOCOW_FL;
3086
3087         fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3088
3089         if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3090                 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3091
3092         return 0;
3093 }
3094
3095 int f2fs_fileattr_set(struct mnt_idmap *idmap,
3096                       struct dentry *dentry, struct fileattr *fa)
3097 {
3098         struct inode *inode = d_inode(dentry);
3099         u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3100         u32 iflags;
3101         int err;
3102
3103         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3104                 return -EIO;
3105         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3106                 return -ENOSPC;
3107         if (fsflags & ~F2FS_GETTABLE_FS_FL)
3108                 return -EOPNOTSUPP;
3109         fsflags &= F2FS_SETTABLE_FS_FL;
3110         if (!fa->flags_valid)
3111                 mask &= FS_COMMON_FL;
3112
3113         iflags = f2fs_fsflags_to_iflags(fsflags);
3114         if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3115                 return -EOPNOTSUPP;
3116
3117         err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3118         if (!err)
3119                 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3120
3121         return err;
3122 }
3123
3124 int f2fs_pin_file_control(struct inode *inode, bool inc)
3125 {
3126         struct f2fs_inode_info *fi = F2FS_I(inode);
3127         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3128
3129         /* Use i_gc_failures for normal file as a risk signal. */
3130         if (inc)
3131                 f2fs_i_gc_failures_write(inode,
3132                                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3133
3134         if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3135                 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3136                           __func__, inode->i_ino,
3137                           fi->i_gc_failures[GC_FAILURE_PIN]);
3138                 clear_inode_flag(inode, FI_PIN_FILE);
3139                 return -EAGAIN;
3140         }
3141         return 0;
3142 }
3143
3144 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3145 {
3146         struct inode *inode = file_inode(filp);
3147         __u32 pin;
3148         int ret = 0;
3149
3150         if (get_user(pin, (__u32 __user *)arg))
3151                 return -EFAULT;
3152
3153         if (!S_ISREG(inode->i_mode))
3154                 return -EINVAL;
3155
3156         if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3157                 return -EROFS;
3158
3159         ret = mnt_want_write_file(filp);
3160         if (ret)
3161                 return ret;
3162
3163         inode_lock(inode);
3164
3165         if (!pin) {
3166                 clear_inode_flag(inode, FI_PIN_FILE);
3167                 f2fs_i_gc_failures_write(inode, 0);
3168                 goto done;
3169         }
3170
3171         if (f2fs_should_update_outplace(inode, NULL)) {
3172                 ret = -EINVAL;
3173                 goto out;
3174         }
3175
3176         if (f2fs_pin_file_control(inode, false)) {
3177                 ret = -EAGAIN;
3178                 goto out;
3179         }
3180
3181         ret = f2fs_convert_inline_inode(inode);
3182         if (ret)
3183                 goto out;
3184
3185         if (!f2fs_disable_compressed_file(inode)) {
3186                 ret = -EOPNOTSUPP;
3187                 goto out;
3188         }
3189
3190         set_inode_flag(inode, FI_PIN_FILE);
3191         ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3192 done:
3193         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3194 out:
3195         inode_unlock(inode);
3196         mnt_drop_write_file(filp);
3197         return ret;
3198 }
3199
3200 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3201 {
3202         struct inode *inode = file_inode(filp);
3203         __u32 pin = 0;
3204
3205         if (is_inode_flag_set(inode, FI_PIN_FILE))
3206                 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3207         return put_user(pin, (u32 __user *)arg);
3208 }
3209
3210 int f2fs_precache_extents(struct inode *inode)
3211 {
3212         struct f2fs_inode_info *fi = F2FS_I(inode);
3213         struct f2fs_map_blocks map;
3214         pgoff_t m_next_extent;
3215         loff_t end;
3216         int err;
3217
3218         if (is_inode_flag_set(inode, FI_NO_EXTENT))
3219                 return -EOPNOTSUPP;
3220
3221         map.m_lblk = 0;
3222         map.m_next_pgofs = NULL;
3223         map.m_next_extent = &m_next_extent;
3224         map.m_seg_type = NO_CHECK_TYPE;
3225         map.m_may_create = false;
3226         end = max_file_blocks(inode);
3227
3228         while (map.m_lblk < end) {
3229                 map.m_len = end - map.m_lblk;
3230
3231                 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3232                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3233                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3234                 if (err)
3235                         return err;
3236
3237                 map.m_lblk = m_next_extent;
3238         }
3239
3240         return 0;
3241 }
3242
3243 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3244 {
3245         return f2fs_precache_extents(file_inode(filp));
3246 }
3247
3248 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3249 {
3250         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3251         __u64 block_count;
3252
3253         if (!capable(CAP_SYS_ADMIN))
3254                 return -EPERM;
3255
3256         if (f2fs_readonly(sbi->sb))
3257                 return -EROFS;
3258
3259         if (copy_from_user(&block_count, (void __user *)arg,
3260                            sizeof(block_count)))
3261                 return -EFAULT;
3262
3263         return f2fs_resize_fs(sbi, block_count);
3264 }
3265
3266 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3267 {
3268         struct inode *inode = file_inode(filp);
3269
3270         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3271
3272         if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3273                 f2fs_warn(F2FS_I_SB(inode),
3274                           "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3275                           inode->i_ino);
3276                 return -EOPNOTSUPP;
3277         }
3278
3279         return fsverity_ioctl_enable(filp, (const void __user *)arg);
3280 }
3281
3282 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3283 {
3284         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3285                 return -EOPNOTSUPP;
3286
3287         return fsverity_ioctl_measure(filp, (void __user *)arg);
3288 }
3289
3290 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3291 {
3292         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3293                 return -EOPNOTSUPP;
3294
3295         return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3296 }
3297
3298 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3299 {
3300         struct inode *inode = file_inode(filp);
3301         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3302         char *vbuf;
3303         int count;
3304         int err = 0;
3305
3306         vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3307         if (!vbuf)
3308                 return -ENOMEM;
3309
3310         f2fs_down_read(&sbi->sb_lock);
3311         count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3312                         ARRAY_SIZE(sbi->raw_super->volume_name),
3313                         UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3314         f2fs_up_read(&sbi->sb_lock);
3315
3316         if (copy_to_user((char __user *)arg, vbuf,
3317                                 min(FSLABEL_MAX, count)))
3318                 err = -EFAULT;
3319
3320         kfree(vbuf);
3321         return err;
3322 }
3323
3324 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3325 {
3326         struct inode *inode = file_inode(filp);
3327         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3328         char *vbuf;
3329         int err = 0;
3330
3331         if (!capable(CAP_SYS_ADMIN))
3332                 return -EPERM;
3333
3334         vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3335         if (IS_ERR(vbuf))
3336                 return PTR_ERR(vbuf);
3337
3338         err = mnt_want_write_file(filp);
3339         if (err)
3340                 goto out;
3341
3342         f2fs_down_write(&sbi->sb_lock);
3343
3344         memset(sbi->raw_super->volume_name, 0,
3345                         sizeof(sbi->raw_super->volume_name));
3346         utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3347                         sbi->raw_super->volume_name,
3348                         ARRAY_SIZE(sbi->raw_super->volume_name));
3349
3350         err = f2fs_commit_super(sbi, false);
3351
3352         f2fs_up_write(&sbi->sb_lock);
3353
3354         mnt_drop_write_file(filp);
3355 out:
3356         kfree(vbuf);
3357         return err;
3358 }
3359
3360 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3361 {
3362         struct inode *inode = file_inode(filp);
3363         __u64 blocks;
3364
3365         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3366                 return -EOPNOTSUPP;
3367
3368         if (!f2fs_compressed_file(inode))
3369                 return -EINVAL;
3370
3371         blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3372         return put_user(blocks, (u64 __user *)arg);
3373 }
3374
3375 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3376 {
3377         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3378         unsigned int released_blocks = 0;
3379         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3380         block_t blkaddr;
3381         int i;
3382
3383         for (i = 0; i < count; i++) {
3384                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3385                                                 dn->ofs_in_node + i);
3386
3387                 if (!__is_valid_data_blkaddr(blkaddr))
3388                         continue;
3389                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3390                                         DATA_GENERIC_ENHANCE))) {
3391                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3392                         return -EFSCORRUPTED;
3393                 }
3394         }
3395
3396         while (count) {
3397                 int compr_blocks = 0;
3398
3399                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3400                         blkaddr = f2fs_data_blkaddr(dn);
3401
3402                         if (i == 0) {
3403                                 if (blkaddr == COMPRESS_ADDR)
3404                                         continue;
3405                                 dn->ofs_in_node += cluster_size;
3406                                 goto next;
3407                         }
3408
3409                         if (__is_valid_data_blkaddr(blkaddr))
3410                                 compr_blocks++;
3411
3412                         if (blkaddr != NEW_ADDR)
3413                                 continue;
3414
3415                         dn->data_blkaddr = NULL_ADDR;
3416                         f2fs_set_data_blkaddr(dn);
3417                 }
3418
3419                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3420                 dec_valid_block_count(sbi, dn->inode,
3421                                         cluster_size - compr_blocks);
3422
3423                 released_blocks += cluster_size - compr_blocks;
3424 next:
3425                 count -= cluster_size;
3426         }
3427
3428         return released_blocks;
3429 }
3430
3431 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3432 {
3433         struct inode *inode = file_inode(filp);
3434         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3435         pgoff_t page_idx = 0, last_idx;
3436         unsigned int released_blocks = 0;
3437         int ret;
3438         int writecount;
3439
3440         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3441                 return -EOPNOTSUPP;
3442
3443         if (!f2fs_compressed_file(inode))
3444                 return -EINVAL;
3445
3446         if (f2fs_readonly(sbi->sb))
3447                 return -EROFS;
3448
3449         ret = mnt_want_write_file(filp);
3450         if (ret)
3451                 return ret;
3452
3453         f2fs_balance_fs(F2FS_I_SB(inode), true);
3454
3455         inode_lock(inode);
3456
3457         writecount = atomic_read(&inode->i_writecount);
3458         if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3459                         (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3460                 ret = -EBUSY;
3461                 goto out;
3462         }
3463
3464         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3465                 ret = -EINVAL;
3466                 goto out;
3467         }
3468
3469         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3470         if (ret)
3471                 goto out;
3472
3473         set_inode_flag(inode, FI_COMPRESS_RELEASED);
3474         inode->i_ctime = current_time(inode);
3475         f2fs_mark_inode_dirty_sync(inode, true);
3476
3477         if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3478                 goto out;
3479
3480         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3481         filemap_invalidate_lock(inode->i_mapping);
3482
3483         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3484
3485         while (page_idx < last_idx) {
3486                 struct dnode_of_data dn;
3487                 pgoff_t end_offset, count;
3488
3489                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3490                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3491                 if (ret) {
3492                         if (ret == -ENOENT) {
3493                                 page_idx = f2fs_get_next_page_offset(&dn,
3494                                                                 page_idx);
3495                                 ret = 0;
3496                                 continue;
3497                         }
3498                         break;
3499                 }
3500
3501                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3502                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3503                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3504
3505                 ret = release_compress_blocks(&dn, count);
3506
3507                 f2fs_put_dnode(&dn);
3508
3509                 if (ret < 0)
3510                         break;
3511
3512                 page_idx += count;
3513                 released_blocks += ret;
3514         }
3515
3516         filemap_invalidate_unlock(inode->i_mapping);
3517         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3518 out:
3519         inode_unlock(inode);
3520
3521         mnt_drop_write_file(filp);
3522
3523         if (ret >= 0) {
3524                 ret = put_user(released_blocks, (u64 __user *)arg);
3525         } else if (released_blocks &&
3526                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3527                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3528                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3529                         "iblocks=%llu, released=%u, compr_blocks=%u, "
3530                         "run fsck to fix.",
3531                         __func__, inode->i_ino, inode->i_blocks,
3532                         released_blocks,
3533                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3534         }
3535
3536         return ret;
3537 }
3538
3539 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3540 {
3541         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3542         unsigned int reserved_blocks = 0;
3543         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3544         block_t blkaddr;
3545         int i;
3546
3547         for (i = 0; i < count; i++) {
3548                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3549                                                 dn->ofs_in_node + i);
3550
3551                 if (!__is_valid_data_blkaddr(blkaddr))
3552                         continue;
3553                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3554                                         DATA_GENERIC_ENHANCE))) {
3555                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3556                         return -EFSCORRUPTED;
3557                 }
3558         }
3559
3560         while (count) {
3561                 int compr_blocks = 0;
3562                 blkcnt_t reserved;
3563                 int ret;
3564
3565                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3566                         blkaddr = f2fs_data_blkaddr(dn);
3567
3568                         if (i == 0) {
3569                                 if (blkaddr == COMPRESS_ADDR)
3570                                         continue;
3571                                 dn->ofs_in_node += cluster_size;
3572                                 goto next;
3573                         }
3574
3575                         if (__is_valid_data_blkaddr(blkaddr)) {
3576                                 compr_blocks++;
3577                                 continue;
3578                         }
3579
3580                         dn->data_blkaddr = NEW_ADDR;
3581                         f2fs_set_data_blkaddr(dn);
3582                 }
3583
3584                 reserved = cluster_size - compr_blocks;
3585                 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3586                 if (ret)
3587                         return ret;
3588
3589                 if (reserved != cluster_size - compr_blocks)
3590                         return -ENOSPC;
3591
3592                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3593
3594                 reserved_blocks += reserved;
3595 next:
3596                 count -= cluster_size;
3597         }
3598
3599         return reserved_blocks;
3600 }
3601
3602 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3603 {
3604         struct inode *inode = file_inode(filp);
3605         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3606         pgoff_t page_idx = 0, last_idx;
3607         unsigned int reserved_blocks = 0;
3608         int ret;
3609
3610         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3611                 return -EOPNOTSUPP;
3612
3613         if (!f2fs_compressed_file(inode))
3614                 return -EINVAL;
3615
3616         if (f2fs_readonly(sbi->sb))
3617                 return -EROFS;
3618
3619         ret = mnt_want_write_file(filp);
3620         if (ret)
3621                 return ret;
3622
3623         if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3624                 goto out;
3625
3626         f2fs_balance_fs(F2FS_I_SB(inode), true);
3627
3628         inode_lock(inode);
3629
3630         if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3631                 ret = -EINVAL;
3632                 goto unlock_inode;
3633         }
3634
3635         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3636         filemap_invalidate_lock(inode->i_mapping);
3637
3638         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3639
3640         while (page_idx < last_idx) {
3641                 struct dnode_of_data dn;
3642                 pgoff_t end_offset, count;
3643
3644                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3645                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3646                 if (ret) {
3647                         if (ret == -ENOENT) {
3648                                 page_idx = f2fs_get_next_page_offset(&dn,
3649                                                                 page_idx);
3650                                 ret = 0;
3651                                 continue;
3652                         }
3653                         break;
3654                 }
3655
3656                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3657                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3658                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3659
3660                 ret = reserve_compress_blocks(&dn, count);
3661
3662                 f2fs_put_dnode(&dn);
3663
3664                 if (ret < 0)
3665                         break;
3666
3667                 page_idx += count;
3668                 reserved_blocks += ret;
3669         }
3670
3671         filemap_invalidate_unlock(inode->i_mapping);
3672         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3673
3674         if (ret >= 0) {
3675                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3676                 inode->i_ctime = current_time(inode);
3677                 f2fs_mark_inode_dirty_sync(inode, true);
3678         }
3679 unlock_inode:
3680         inode_unlock(inode);
3681 out:
3682         mnt_drop_write_file(filp);
3683
3684         if (ret >= 0) {
3685                 ret = put_user(reserved_blocks, (u64 __user *)arg);
3686         } else if (reserved_blocks &&
3687                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3688                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3689                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3690                         "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3691                         "run fsck to fix.",
3692                         __func__, inode->i_ino, inode->i_blocks,
3693                         reserved_blocks,
3694                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3695         }
3696
3697         return ret;
3698 }
3699
3700 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3701                 pgoff_t off, block_t block, block_t len, u32 flags)
3702 {
3703         sector_t sector = SECTOR_FROM_BLOCK(block);
3704         sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3705         int ret = 0;
3706
3707         if (flags & F2FS_TRIM_FILE_DISCARD) {
3708                 if (bdev_max_secure_erase_sectors(bdev))
3709                         ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3710                                         GFP_NOFS);
3711                 else
3712                         ret = blkdev_issue_discard(bdev, sector, nr_sects,
3713                                         GFP_NOFS);
3714         }
3715
3716         if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3717                 if (IS_ENCRYPTED(inode))
3718                         ret = fscrypt_zeroout_range(inode, off, block, len);
3719                 else
3720                         ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3721                                         GFP_NOFS, 0);
3722         }
3723
3724         return ret;
3725 }
3726
3727 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3728 {
3729         struct inode *inode = file_inode(filp);
3730         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3731         struct address_space *mapping = inode->i_mapping;
3732         struct block_device *prev_bdev = NULL;
3733         struct f2fs_sectrim_range range;
3734         pgoff_t index, pg_end, prev_index = 0;
3735         block_t prev_block = 0, len = 0;
3736         loff_t end_addr;
3737         bool to_end = false;
3738         int ret = 0;
3739
3740         if (!(filp->f_mode & FMODE_WRITE))
3741                 return -EBADF;
3742
3743         if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3744                                 sizeof(range)))
3745                 return -EFAULT;
3746
3747         if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3748                         !S_ISREG(inode->i_mode))
3749                 return -EINVAL;
3750
3751         if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3752                         !f2fs_hw_support_discard(sbi)) ||
3753                         ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3754                          IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3755                 return -EOPNOTSUPP;
3756
3757         file_start_write(filp);
3758         inode_lock(inode);
3759
3760         if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3761                         range.start >= inode->i_size) {
3762                 ret = -EINVAL;
3763                 goto err;
3764         }
3765
3766         if (range.len == 0)
3767                 goto err;
3768
3769         if (inode->i_size - range.start > range.len) {
3770                 end_addr = range.start + range.len;
3771         } else {
3772                 end_addr = range.len == (u64)-1 ?
3773                         sbi->sb->s_maxbytes : inode->i_size;
3774                 to_end = true;
3775         }
3776
3777         if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3778                         (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3779                 ret = -EINVAL;
3780                 goto err;
3781         }
3782
3783         index = F2FS_BYTES_TO_BLK(range.start);
3784         pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3785
3786         ret = f2fs_convert_inline_inode(inode);
3787         if (ret)
3788                 goto err;
3789
3790         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3791         filemap_invalidate_lock(mapping);
3792
3793         ret = filemap_write_and_wait_range(mapping, range.start,
3794                         to_end ? LLONG_MAX : end_addr - 1);
3795         if (ret)
3796                 goto out;
3797
3798         truncate_inode_pages_range(mapping, range.start,
3799                         to_end ? -1 : end_addr - 1);
3800
3801         while (index < pg_end) {
3802                 struct dnode_of_data dn;
3803                 pgoff_t end_offset, count;
3804                 int i;
3805
3806                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3807                 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3808                 if (ret) {
3809                         if (ret == -ENOENT) {
3810                                 index = f2fs_get_next_page_offset(&dn, index);
3811                                 continue;
3812                         }
3813                         goto out;
3814                 }
3815
3816                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3817                 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3818                 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3819                         struct block_device *cur_bdev;
3820                         block_t blkaddr = f2fs_data_blkaddr(&dn);
3821
3822                         if (!__is_valid_data_blkaddr(blkaddr))
3823                                 continue;
3824
3825                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3826                                                 DATA_GENERIC_ENHANCE)) {
3827                                 ret = -EFSCORRUPTED;
3828                                 f2fs_put_dnode(&dn);
3829                                 f2fs_handle_error(sbi,
3830                                                 ERROR_INVALID_BLKADDR);
3831                                 goto out;
3832                         }
3833
3834                         cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3835                         if (f2fs_is_multi_device(sbi)) {
3836                                 int di = f2fs_target_device_index(sbi, blkaddr);
3837
3838                                 blkaddr -= FDEV(di).start_blk;
3839                         }
3840
3841                         if (len) {
3842                                 if (prev_bdev == cur_bdev &&
3843                                                 index == prev_index + len &&
3844                                                 blkaddr == prev_block + len) {
3845                                         len++;
3846                                 } else {
3847                                         ret = f2fs_secure_erase(prev_bdev,
3848                                                 inode, prev_index, prev_block,
3849                                                 len, range.flags);
3850                                         if (ret) {
3851                                                 f2fs_put_dnode(&dn);
3852                                                 goto out;
3853                                         }
3854
3855                                         len = 0;
3856                                 }
3857                         }
3858
3859                         if (!len) {
3860                                 prev_bdev = cur_bdev;
3861                                 prev_index = index;
3862                                 prev_block = blkaddr;
3863                                 len = 1;
3864                         }
3865                 }
3866
3867                 f2fs_put_dnode(&dn);
3868
3869                 if (fatal_signal_pending(current)) {
3870                         ret = -EINTR;
3871                         goto out;
3872                 }
3873                 cond_resched();
3874         }
3875
3876         if (len)
3877                 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3878                                 prev_block, len, range.flags);
3879 out:
3880         filemap_invalidate_unlock(mapping);
3881         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3882 err:
3883         inode_unlock(inode);
3884         file_end_write(filp);
3885
3886         return ret;
3887 }
3888
3889 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3890 {
3891         struct inode *inode = file_inode(filp);
3892         struct f2fs_comp_option option;
3893
3894         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3895                 return -EOPNOTSUPP;
3896
3897         inode_lock_shared(inode);
3898
3899         if (!f2fs_compressed_file(inode)) {
3900                 inode_unlock_shared(inode);
3901                 return -ENODATA;
3902         }
3903
3904         option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3905         option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3906
3907         inode_unlock_shared(inode);
3908
3909         if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3910                                 sizeof(option)))
3911                 return -EFAULT;
3912
3913         return 0;
3914 }
3915
3916 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3917 {
3918         struct inode *inode = file_inode(filp);
3919         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3920         struct f2fs_comp_option option;
3921         int ret = 0;
3922
3923         if (!f2fs_sb_has_compression(sbi))
3924                 return -EOPNOTSUPP;
3925
3926         if (!(filp->f_mode & FMODE_WRITE))
3927                 return -EBADF;
3928
3929         if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3930                                 sizeof(option)))
3931                 return -EFAULT;
3932
3933         if (!f2fs_compressed_file(inode) ||
3934                         option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3935                         option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3936                         option.algorithm >= COMPRESS_MAX)
3937                 return -EINVAL;
3938
3939         file_start_write(filp);
3940         inode_lock(inode);
3941
3942         if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3943                 ret = -EBUSY;
3944                 goto out;
3945         }
3946
3947         if (inode->i_size != 0) {
3948                 ret = -EFBIG;
3949                 goto out;
3950         }
3951
3952         F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3953         F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3954         F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3955         f2fs_mark_inode_dirty_sync(inode, true);
3956
3957         if (!f2fs_is_compress_backend_ready(inode))
3958                 f2fs_warn(sbi, "compression algorithm is successfully set, "
3959                         "but current kernel doesn't support this algorithm.");
3960 out:
3961         inode_unlock(inode);
3962         file_end_write(filp);
3963
3964         return ret;
3965 }
3966
3967 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3968 {
3969         DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3970         struct address_space *mapping = inode->i_mapping;
3971         struct page *page;
3972         pgoff_t redirty_idx = page_idx;
3973         int i, page_len = 0, ret = 0;
3974
3975         page_cache_ra_unbounded(&ractl, len, 0);
3976
3977         for (i = 0; i < len; i++, page_idx++) {
3978                 page = read_cache_page(mapping, page_idx, NULL, NULL);
3979                 if (IS_ERR(page)) {
3980                         ret = PTR_ERR(page);
3981                         break;
3982                 }
3983                 page_len++;
3984         }
3985
3986         for (i = 0; i < page_len; i++, redirty_idx++) {
3987                 page = find_lock_page(mapping, redirty_idx);
3988
3989                 /* It will never fail, when page has pinned above */
3990                 f2fs_bug_on(F2FS_I_SB(inode), !page);
3991
3992                 set_page_dirty(page);
3993                 f2fs_put_page(page, 1);
3994                 f2fs_put_page(page, 0);
3995         }
3996
3997         return ret;
3998 }
3999
4000 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
4001 {
4002         struct inode *inode = file_inode(filp);
4003         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4004         struct f2fs_inode_info *fi = F2FS_I(inode);
4005         pgoff_t page_idx = 0, last_idx;
4006         unsigned int blk_per_seg = sbi->blocks_per_seg;
4007         int cluster_size = fi->i_cluster_size;
4008         int count, ret;
4009
4010         if (!f2fs_sb_has_compression(sbi) ||
4011                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4012                 return -EOPNOTSUPP;
4013
4014         if (!(filp->f_mode & FMODE_WRITE))
4015                 return -EBADF;
4016
4017         if (!f2fs_compressed_file(inode))
4018                 return -EINVAL;
4019
4020         f2fs_balance_fs(F2FS_I_SB(inode), true);
4021
4022         file_start_write(filp);
4023         inode_lock(inode);
4024
4025         if (!f2fs_is_compress_backend_ready(inode)) {
4026                 ret = -EOPNOTSUPP;
4027                 goto out;
4028         }
4029
4030         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4031                 ret = -EINVAL;
4032                 goto out;
4033         }
4034
4035         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4036         if (ret)
4037                 goto out;
4038
4039         if (!atomic_read(&fi->i_compr_blocks))
4040                 goto out;
4041
4042         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4043
4044         count = last_idx - page_idx;
4045         while (count) {
4046                 int len = min(cluster_size, count);
4047
4048                 ret = redirty_blocks(inode, page_idx, len);
4049                 if (ret < 0)
4050                         break;
4051
4052                 if (get_dirty_pages(inode) >= blk_per_seg)
4053                         filemap_fdatawrite(inode->i_mapping);
4054
4055                 count -= len;
4056                 page_idx += len;
4057         }
4058
4059         if (!ret)
4060                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4061                                                         LLONG_MAX);
4062
4063         if (ret)
4064                 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4065                           __func__, ret);
4066 out:
4067         inode_unlock(inode);
4068         file_end_write(filp);
4069
4070         return ret;
4071 }
4072
4073 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4074 {
4075         struct inode *inode = file_inode(filp);
4076         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4077         pgoff_t page_idx = 0, last_idx;
4078         unsigned int blk_per_seg = sbi->blocks_per_seg;
4079         int cluster_size = F2FS_I(inode)->i_cluster_size;
4080         int count, ret;
4081
4082         if (!f2fs_sb_has_compression(sbi) ||
4083                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4084                 return -EOPNOTSUPP;
4085
4086         if (!(filp->f_mode & FMODE_WRITE))
4087                 return -EBADF;
4088
4089         if (!f2fs_compressed_file(inode))
4090                 return -EINVAL;
4091
4092         f2fs_balance_fs(F2FS_I_SB(inode), true);
4093
4094         file_start_write(filp);
4095         inode_lock(inode);
4096
4097         if (!f2fs_is_compress_backend_ready(inode)) {
4098                 ret = -EOPNOTSUPP;
4099                 goto out;
4100         }
4101
4102         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4103                 ret = -EINVAL;
4104                 goto out;
4105         }
4106
4107         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4108         if (ret)
4109                 goto out;
4110
4111         set_inode_flag(inode, FI_ENABLE_COMPRESS);
4112
4113         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4114
4115         count = last_idx - page_idx;
4116         while (count) {
4117                 int len = min(cluster_size, count);
4118
4119                 ret = redirty_blocks(inode, page_idx, len);
4120                 if (ret < 0)
4121                         break;
4122
4123                 if (get_dirty_pages(inode) >= blk_per_seg)
4124                         filemap_fdatawrite(inode->i_mapping);
4125
4126                 count -= len;
4127                 page_idx += len;
4128         }
4129
4130         if (!ret)
4131                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4132                                                         LLONG_MAX);
4133
4134         clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4135
4136         if (ret)
4137                 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4138                           __func__, ret);
4139 out:
4140         inode_unlock(inode);
4141         file_end_write(filp);
4142
4143         return ret;
4144 }
4145
4146 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4147 {
4148         switch (cmd) {
4149         case FS_IOC_GETVERSION:
4150                 return f2fs_ioc_getversion(filp, arg);
4151         case F2FS_IOC_START_ATOMIC_WRITE:
4152                 return f2fs_ioc_start_atomic_write(filp, false);
4153         case F2FS_IOC_START_ATOMIC_REPLACE:
4154                 return f2fs_ioc_start_atomic_write(filp, true);
4155         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4156                 return f2fs_ioc_commit_atomic_write(filp);
4157         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4158                 return f2fs_ioc_abort_atomic_write(filp);
4159         case F2FS_IOC_START_VOLATILE_WRITE:
4160         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4161                 return -EOPNOTSUPP;
4162         case F2FS_IOC_SHUTDOWN:
4163                 return f2fs_ioc_shutdown(filp, arg);
4164         case FITRIM:
4165                 return f2fs_ioc_fitrim(filp, arg);
4166         case FS_IOC_SET_ENCRYPTION_POLICY:
4167                 return f2fs_ioc_set_encryption_policy(filp, arg);
4168         case FS_IOC_GET_ENCRYPTION_POLICY:
4169                 return f2fs_ioc_get_encryption_policy(filp, arg);
4170         case FS_IOC_GET_ENCRYPTION_PWSALT:
4171                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4172         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4173                 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4174         case FS_IOC_ADD_ENCRYPTION_KEY:
4175                 return f2fs_ioc_add_encryption_key(filp, arg);
4176         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4177                 return f2fs_ioc_remove_encryption_key(filp, arg);
4178         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4179                 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4180         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4181                 return f2fs_ioc_get_encryption_key_status(filp, arg);
4182         case FS_IOC_GET_ENCRYPTION_NONCE:
4183                 return f2fs_ioc_get_encryption_nonce(filp, arg);
4184         case F2FS_IOC_GARBAGE_COLLECT:
4185                 return f2fs_ioc_gc(filp, arg);
4186         case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4187                 return f2fs_ioc_gc_range(filp, arg);
4188         case F2FS_IOC_WRITE_CHECKPOINT:
4189                 return f2fs_ioc_write_checkpoint(filp, arg);
4190         case F2FS_IOC_DEFRAGMENT:
4191                 return f2fs_ioc_defragment(filp, arg);
4192         case F2FS_IOC_MOVE_RANGE:
4193                 return f2fs_ioc_move_range(filp, arg);
4194         case F2FS_IOC_FLUSH_DEVICE:
4195                 return f2fs_ioc_flush_device(filp, arg);
4196         case F2FS_IOC_GET_FEATURES:
4197                 return f2fs_ioc_get_features(filp, arg);
4198         case F2FS_IOC_GET_PIN_FILE:
4199                 return f2fs_ioc_get_pin_file(filp, arg);
4200         case F2FS_IOC_SET_PIN_FILE:
4201                 return f2fs_ioc_set_pin_file(filp, arg);
4202         case F2FS_IOC_PRECACHE_EXTENTS:
4203                 return f2fs_ioc_precache_extents(filp, arg);
4204         case F2FS_IOC_RESIZE_FS:
4205                 return f2fs_ioc_resize_fs(filp, arg);
4206         case FS_IOC_ENABLE_VERITY:
4207                 return f2fs_ioc_enable_verity(filp, arg);
4208         case FS_IOC_MEASURE_VERITY:
4209                 return f2fs_ioc_measure_verity(filp, arg);
4210         case FS_IOC_READ_VERITY_METADATA:
4211                 return f2fs_ioc_read_verity_metadata(filp, arg);
4212         case FS_IOC_GETFSLABEL:
4213                 return f2fs_ioc_getfslabel(filp, arg);
4214         case FS_IOC_SETFSLABEL:
4215                 return f2fs_ioc_setfslabel(filp, arg);
4216         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4217                 return f2fs_get_compress_blocks(filp, arg);
4218         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4219                 return f2fs_release_compress_blocks(filp, arg);
4220         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4221                 return f2fs_reserve_compress_blocks(filp, arg);
4222         case F2FS_IOC_SEC_TRIM_FILE:
4223                 return f2fs_sec_trim_file(filp, arg);
4224         case F2FS_IOC_GET_COMPRESS_OPTION:
4225                 return f2fs_ioc_get_compress_option(filp, arg);
4226         case F2FS_IOC_SET_COMPRESS_OPTION:
4227                 return f2fs_ioc_set_compress_option(filp, arg);
4228         case F2FS_IOC_DECOMPRESS_FILE:
4229                 return f2fs_ioc_decompress_file(filp, arg);
4230         case F2FS_IOC_COMPRESS_FILE:
4231                 return f2fs_ioc_compress_file(filp, arg);
4232         default:
4233                 return -ENOTTY;
4234         }
4235 }
4236
4237 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4238 {
4239         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4240                 return -EIO;
4241         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4242                 return -ENOSPC;
4243
4244         return __f2fs_ioctl(filp, cmd, arg);
4245 }
4246
4247 /*
4248  * Return %true if the given read or write request should use direct I/O, or
4249  * %false if it should use buffered I/O.
4250  */
4251 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4252                                 struct iov_iter *iter)
4253 {
4254         unsigned int align;
4255
4256         if (!(iocb->ki_flags & IOCB_DIRECT))
4257                 return false;
4258
4259         if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4260                 return false;
4261
4262         /*
4263          * Direct I/O not aligned to the disk's logical_block_size will be
4264          * attempted, but will fail with -EINVAL.
4265          *
4266          * f2fs additionally requires that direct I/O be aligned to the
4267          * filesystem block size, which is often a stricter requirement.
4268          * However, f2fs traditionally falls back to buffered I/O on requests
4269          * that are logical_block_size-aligned but not fs-block aligned.
4270          *
4271          * The below logic implements this behavior.
4272          */
4273         align = iocb->ki_pos | iov_iter_alignment(iter);
4274         if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4275             IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4276                 return false;
4277
4278         return true;
4279 }
4280
4281 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4282                                 unsigned int flags)
4283 {
4284         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4285
4286         dec_page_count(sbi, F2FS_DIO_READ);
4287         if (error)
4288                 return error;
4289         f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4290         return 0;
4291 }
4292
4293 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4294         .end_io = f2fs_dio_read_end_io,
4295 };
4296
4297 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4298 {
4299         struct file *file = iocb->ki_filp;
4300         struct inode *inode = file_inode(file);
4301         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4302         struct f2fs_inode_info *fi = F2FS_I(inode);
4303         const loff_t pos = iocb->ki_pos;
4304         const size_t count = iov_iter_count(to);
4305         struct iomap_dio *dio;
4306         ssize_t ret;
4307
4308         if (count == 0)
4309                 return 0; /* skip atime update */
4310
4311         trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4312
4313         if (iocb->ki_flags & IOCB_NOWAIT) {
4314                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4315                         ret = -EAGAIN;
4316                         goto out;
4317                 }
4318         } else {
4319                 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4320         }
4321
4322         /*
4323          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4324          * the higher-level function iomap_dio_rw() in order to ensure that the
4325          * F2FS_DIO_READ counter will be decremented correctly in all cases.
4326          */
4327         inc_page_count(sbi, F2FS_DIO_READ);
4328         dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4329                              &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4330         if (IS_ERR_OR_NULL(dio)) {
4331                 ret = PTR_ERR_OR_ZERO(dio);
4332                 if (ret != -EIOCBQUEUED)
4333                         dec_page_count(sbi, F2FS_DIO_READ);
4334         } else {
4335                 ret = iomap_dio_complete(dio);
4336         }
4337
4338         f2fs_up_read(&fi->i_gc_rwsem[READ]);
4339
4340         file_accessed(file);
4341 out:
4342         trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4343         return ret;
4344 }
4345
4346 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4347 {
4348         struct inode *inode = file_inode(iocb->ki_filp);
4349         const loff_t pos = iocb->ki_pos;
4350         ssize_t ret;
4351
4352         if (!f2fs_is_compress_backend_ready(inode))
4353                 return -EOPNOTSUPP;
4354
4355         if (trace_f2fs_dataread_start_enabled()) {
4356                 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4357                 char *path;
4358
4359                 if (!p)
4360                         goto skip_read_trace;
4361
4362                 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4363                 if (IS_ERR(path)) {
4364                         kfree(p);
4365                         goto skip_read_trace;
4366                 }
4367
4368                 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4369                                         current->pid, path, current->comm);
4370                 kfree(p);
4371         }
4372 skip_read_trace:
4373         if (f2fs_should_use_dio(inode, iocb, to)) {
4374                 ret = f2fs_dio_read_iter(iocb, to);
4375         } else {
4376                 ret = filemap_read(iocb, to, 0);
4377                 if (ret > 0)
4378                         f2fs_update_iostat(F2FS_I_SB(inode), inode,
4379                                                 APP_BUFFERED_READ_IO, ret);
4380         }
4381         if (trace_f2fs_dataread_end_enabled())
4382                 trace_f2fs_dataread_end(inode, pos, ret);
4383         return ret;
4384 }
4385
4386 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4387 {
4388         struct file *file = iocb->ki_filp;
4389         struct inode *inode = file_inode(file);
4390         ssize_t count;
4391         int err;
4392
4393         if (IS_IMMUTABLE(inode))
4394                 return -EPERM;
4395
4396         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4397                 return -EPERM;
4398
4399         count = generic_write_checks(iocb, from);
4400         if (count <= 0)
4401                 return count;
4402
4403         err = file_modified(file);
4404         if (err)
4405                 return err;
4406         return count;
4407 }
4408
4409 /*
4410  * Preallocate blocks for a write request, if it is possible and helpful to do
4411  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4412  * blocks were preallocated, or a negative errno value if something went
4413  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4414  * requested blocks (not just some of them) have been allocated.
4415  */
4416 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4417                                    bool dio)
4418 {
4419         struct inode *inode = file_inode(iocb->ki_filp);
4420         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4421         const loff_t pos = iocb->ki_pos;
4422         const size_t count = iov_iter_count(iter);
4423         struct f2fs_map_blocks map = {};
4424         int flag;
4425         int ret;
4426
4427         /* If it will be an out-of-place direct write, don't bother. */
4428         if (dio && f2fs_lfs_mode(sbi))
4429                 return 0;
4430         /*
4431          * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4432          * buffered IO, if DIO meets any holes.
4433          */
4434         if (dio && i_size_read(inode) &&
4435                 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4436                 return 0;
4437
4438         /* No-wait I/O can't allocate blocks. */
4439         if (iocb->ki_flags & IOCB_NOWAIT)
4440                 return 0;
4441
4442         /* If it will be a short write, don't bother. */
4443         if (fault_in_iov_iter_readable(iter, count))
4444                 return 0;
4445
4446         if (f2fs_has_inline_data(inode)) {
4447                 /* If the data will fit inline, don't bother. */
4448                 if (pos + count <= MAX_INLINE_DATA(inode))
4449                         return 0;
4450                 ret = f2fs_convert_inline_inode(inode);
4451                 if (ret)
4452                         return ret;
4453         }
4454
4455         /* Do not preallocate blocks that will be written partially in 4KB. */
4456         map.m_lblk = F2FS_BLK_ALIGN(pos);
4457         map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4458         if (map.m_len > map.m_lblk)
4459                 map.m_len -= map.m_lblk;
4460         else
4461                 map.m_len = 0;
4462         map.m_may_create = true;
4463         if (dio) {
4464                 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4465                 flag = F2FS_GET_BLOCK_PRE_DIO;
4466         } else {
4467                 map.m_seg_type = NO_CHECK_TYPE;
4468                 flag = F2FS_GET_BLOCK_PRE_AIO;
4469         }
4470
4471         ret = f2fs_map_blocks(inode, &map, 1, flag);
4472         /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4473         if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4474                 return ret;
4475         if (ret == 0)
4476                 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4477         return map.m_len;
4478 }
4479
4480 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4481                                         struct iov_iter *from)
4482 {
4483         struct file *file = iocb->ki_filp;
4484         struct inode *inode = file_inode(file);
4485         ssize_t ret;
4486
4487         if (iocb->ki_flags & IOCB_NOWAIT)
4488                 return -EOPNOTSUPP;
4489
4490         current->backing_dev_info = inode_to_bdi(inode);
4491         ret = generic_perform_write(iocb, from);
4492         current->backing_dev_info = NULL;
4493
4494         if (ret > 0) {
4495                 iocb->ki_pos += ret;
4496                 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4497                                                 APP_BUFFERED_IO, ret);
4498         }
4499         return ret;
4500 }
4501
4502 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4503                                  unsigned int flags)
4504 {
4505         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4506
4507         dec_page_count(sbi, F2FS_DIO_WRITE);
4508         if (error)
4509                 return error;
4510         f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4511         return 0;
4512 }
4513
4514 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4515         .end_io = f2fs_dio_write_end_io,
4516 };
4517
4518 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4519                                    bool *may_need_sync)
4520 {
4521         struct file *file = iocb->ki_filp;
4522         struct inode *inode = file_inode(file);
4523         struct f2fs_inode_info *fi = F2FS_I(inode);
4524         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4525         const bool do_opu = f2fs_lfs_mode(sbi);
4526         const loff_t pos = iocb->ki_pos;
4527         const ssize_t count = iov_iter_count(from);
4528         unsigned int dio_flags;
4529         struct iomap_dio *dio;
4530         ssize_t ret;
4531
4532         trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4533
4534         if (iocb->ki_flags & IOCB_NOWAIT) {
4535                 /* f2fs_convert_inline_inode() and block allocation can block */
4536                 if (f2fs_has_inline_data(inode) ||
4537                     !f2fs_overwrite_io(inode, pos, count)) {
4538                         ret = -EAGAIN;
4539                         goto out;
4540                 }
4541
4542                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4543                         ret = -EAGAIN;
4544                         goto out;
4545                 }
4546                 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4547                         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4548                         ret = -EAGAIN;
4549                         goto out;
4550                 }
4551         } else {
4552                 ret = f2fs_convert_inline_inode(inode);
4553                 if (ret)
4554                         goto out;
4555
4556                 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4557                 if (do_opu)
4558                         f2fs_down_read(&fi->i_gc_rwsem[READ]);
4559         }
4560
4561         /*
4562          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4563          * the higher-level function iomap_dio_rw() in order to ensure that the
4564          * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4565          */
4566         inc_page_count(sbi, F2FS_DIO_WRITE);
4567         dio_flags = 0;
4568         if (pos + count > inode->i_size)
4569                 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4570         dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4571                              &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4572         if (IS_ERR_OR_NULL(dio)) {
4573                 ret = PTR_ERR_OR_ZERO(dio);
4574                 if (ret == -ENOTBLK)
4575                         ret = 0;
4576                 if (ret != -EIOCBQUEUED)
4577                         dec_page_count(sbi, F2FS_DIO_WRITE);
4578         } else {
4579                 ret = iomap_dio_complete(dio);
4580         }
4581
4582         if (do_opu)
4583                 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4584         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4585
4586         if (ret < 0)
4587                 goto out;
4588         if (pos + ret > inode->i_size)
4589                 f2fs_i_size_write(inode, pos + ret);
4590         if (!do_opu)
4591                 set_inode_flag(inode, FI_UPDATE_WRITE);
4592
4593         if (iov_iter_count(from)) {
4594                 ssize_t ret2;
4595                 loff_t bufio_start_pos = iocb->ki_pos;
4596
4597                 /*
4598                  * The direct write was partial, so we need to fall back to a
4599                  * buffered write for the remainder.
4600                  */
4601
4602                 ret2 = f2fs_buffered_write_iter(iocb, from);
4603                 if (iov_iter_count(from))
4604                         f2fs_write_failed(inode, iocb->ki_pos);
4605                 if (ret2 < 0)
4606                         goto out;
4607
4608                 /*
4609                  * Ensure that the pagecache pages are written to disk and
4610                  * invalidated to preserve the expected O_DIRECT semantics.
4611                  */
4612                 if (ret2 > 0) {
4613                         loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4614
4615                         ret += ret2;
4616
4617                         ret2 = filemap_write_and_wait_range(file->f_mapping,
4618                                                             bufio_start_pos,
4619                                                             bufio_end_pos);
4620                         if (ret2 < 0)
4621                                 goto out;
4622                         invalidate_mapping_pages(file->f_mapping,
4623                                                  bufio_start_pos >> PAGE_SHIFT,
4624                                                  bufio_end_pos >> PAGE_SHIFT);
4625                 }
4626         } else {
4627                 /* iomap_dio_rw() already handled the generic_write_sync(). */
4628                 *may_need_sync = false;
4629         }
4630 out:
4631         trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4632         return ret;
4633 }
4634
4635 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4636 {
4637         struct inode *inode = file_inode(iocb->ki_filp);
4638         const loff_t orig_pos = iocb->ki_pos;
4639         const size_t orig_count = iov_iter_count(from);
4640         loff_t target_size;
4641         bool dio;
4642         bool may_need_sync = true;
4643         int preallocated;
4644         ssize_t ret;
4645
4646         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4647                 ret = -EIO;
4648                 goto out;
4649         }
4650
4651         if (!f2fs_is_compress_backend_ready(inode)) {
4652                 ret = -EOPNOTSUPP;
4653                 goto out;
4654         }
4655
4656         if (iocb->ki_flags & IOCB_NOWAIT) {
4657                 if (!inode_trylock(inode)) {
4658                         ret = -EAGAIN;
4659                         goto out;
4660                 }
4661         } else {
4662                 inode_lock(inode);
4663         }
4664
4665         ret = f2fs_write_checks(iocb, from);
4666         if (ret <= 0)
4667                 goto out_unlock;
4668
4669         /* Determine whether we will do a direct write or a buffered write. */
4670         dio = f2fs_should_use_dio(inode, iocb, from);
4671
4672         /* Possibly preallocate the blocks for the write. */
4673         target_size = iocb->ki_pos + iov_iter_count(from);
4674         preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4675         if (preallocated < 0) {
4676                 ret = preallocated;
4677         } else {
4678                 if (trace_f2fs_datawrite_start_enabled()) {
4679                         char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4680                                                 PATH_MAX, GFP_KERNEL);
4681                         char *path;
4682
4683                         if (!p)
4684                                 goto skip_write_trace;
4685                         path = dentry_path_raw(file_dentry(iocb->ki_filp),
4686                                                                 p, PATH_MAX);
4687                         if (IS_ERR(path)) {
4688                                 kfree(p);
4689                                 goto skip_write_trace;
4690                         }
4691                         trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4692                                         current->pid, path, current->comm);
4693                         kfree(p);
4694                 }
4695 skip_write_trace:
4696                 /* Do the actual write. */
4697                 ret = dio ?
4698                         f2fs_dio_write_iter(iocb, from, &may_need_sync) :
4699                         f2fs_buffered_write_iter(iocb, from);
4700
4701                 if (trace_f2fs_datawrite_end_enabled())
4702                         trace_f2fs_datawrite_end(inode, orig_pos, ret);
4703         }
4704
4705         /* Don't leave any preallocated blocks around past i_size. */
4706         if (preallocated && i_size_read(inode) < target_size) {
4707                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4708                 filemap_invalidate_lock(inode->i_mapping);
4709                 if (!f2fs_truncate(inode))
4710                         file_dont_truncate(inode);
4711                 filemap_invalidate_unlock(inode->i_mapping);
4712                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4713         } else {
4714                 file_dont_truncate(inode);
4715         }
4716
4717         clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4718 out_unlock:
4719         inode_unlock(inode);
4720 out:
4721         trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4722         if (ret > 0 && may_need_sync)
4723                 ret = generic_write_sync(iocb, ret);
4724         return ret;
4725 }
4726
4727 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4728                 int advice)
4729 {
4730         struct address_space *mapping;
4731         struct backing_dev_info *bdi;
4732         struct inode *inode = file_inode(filp);
4733         int err;
4734
4735         if (advice == POSIX_FADV_SEQUENTIAL) {
4736                 if (S_ISFIFO(inode->i_mode))
4737                         return -ESPIPE;
4738
4739                 mapping = filp->f_mapping;
4740                 if (!mapping || len < 0)
4741                         return -EINVAL;
4742
4743                 bdi = inode_to_bdi(mapping->host);
4744                 filp->f_ra.ra_pages = bdi->ra_pages *
4745                         F2FS_I_SB(inode)->seq_file_ra_mul;
4746                 spin_lock(&filp->f_lock);
4747                 filp->f_mode &= ~FMODE_RANDOM;
4748                 spin_unlock(&filp->f_lock);
4749                 return 0;
4750         }
4751
4752         err = generic_fadvise(filp, offset, len, advice);
4753         if (!err && advice == POSIX_FADV_DONTNEED &&
4754                 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4755                 f2fs_compressed_file(inode))
4756                 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4757
4758         return err;
4759 }
4760
4761 #ifdef CONFIG_COMPAT
4762 struct compat_f2fs_gc_range {
4763         u32 sync;
4764         compat_u64 start;
4765         compat_u64 len;
4766 };
4767 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE        _IOW(F2FS_IOCTL_MAGIC, 11,\
4768                                                 struct compat_f2fs_gc_range)
4769
4770 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4771 {
4772         struct compat_f2fs_gc_range __user *urange;
4773         struct f2fs_gc_range range;
4774         int err;
4775
4776         urange = compat_ptr(arg);
4777         err = get_user(range.sync, &urange->sync);
4778         err |= get_user(range.start, &urange->start);
4779         err |= get_user(range.len, &urange->len);
4780         if (err)
4781                 return -EFAULT;
4782
4783         return __f2fs_ioc_gc_range(file, &range);
4784 }
4785
4786 struct compat_f2fs_move_range {
4787         u32 dst_fd;
4788         compat_u64 pos_in;
4789         compat_u64 pos_out;
4790         compat_u64 len;
4791 };
4792 #define F2FS_IOC32_MOVE_RANGE           _IOWR(F2FS_IOCTL_MAGIC, 9,      \
4793                                         struct compat_f2fs_move_range)
4794
4795 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4796 {
4797         struct compat_f2fs_move_range __user *urange;
4798         struct f2fs_move_range range;
4799         int err;
4800
4801         urange = compat_ptr(arg);
4802         err = get_user(range.dst_fd, &urange->dst_fd);
4803         err |= get_user(range.pos_in, &urange->pos_in);
4804         err |= get_user(range.pos_out, &urange->pos_out);
4805         err |= get_user(range.len, &urange->len);
4806         if (err)
4807                 return -EFAULT;
4808
4809         return __f2fs_ioc_move_range(file, &range);
4810 }
4811
4812 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4813 {
4814         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4815                 return -EIO;
4816         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4817                 return -ENOSPC;
4818
4819         switch (cmd) {
4820         case FS_IOC32_GETVERSION:
4821                 cmd = FS_IOC_GETVERSION;
4822                 break;
4823         case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4824                 return f2fs_compat_ioc_gc_range(file, arg);
4825         case F2FS_IOC32_MOVE_RANGE:
4826                 return f2fs_compat_ioc_move_range(file, arg);
4827         case F2FS_IOC_START_ATOMIC_WRITE:
4828         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4829         case F2FS_IOC_START_VOLATILE_WRITE:
4830         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4831         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4832         case F2FS_IOC_SHUTDOWN:
4833         case FITRIM:
4834         case FS_IOC_SET_ENCRYPTION_POLICY:
4835         case FS_IOC_GET_ENCRYPTION_PWSALT:
4836         case FS_IOC_GET_ENCRYPTION_POLICY:
4837         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4838         case FS_IOC_ADD_ENCRYPTION_KEY:
4839         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4840         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4841         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4842         case FS_IOC_GET_ENCRYPTION_NONCE:
4843         case F2FS_IOC_GARBAGE_COLLECT:
4844         case F2FS_IOC_WRITE_CHECKPOINT:
4845         case F2FS_IOC_DEFRAGMENT:
4846         case F2FS_IOC_FLUSH_DEVICE:
4847         case F2FS_IOC_GET_FEATURES:
4848         case F2FS_IOC_GET_PIN_FILE:
4849         case F2FS_IOC_SET_PIN_FILE:
4850         case F2FS_IOC_PRECACHE_EXTENTS:
4851         case F2FS_IOC_RESIZE_FS:
4852         case FS_IOC_ENABLE_VERITY:
4853         case FS_IOC_MEASURE_VERITY:
4854         case FS_IOC_READ_VERITY_METADATA:
4855         case FS_IOC_GETFSLABEL:
4856         case FS_IOC_SETFSLABEL:
4857         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4858         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4859         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4860         case F2FS_IOC_SEC_TRIM_FILE:
4861         case F2FS_IOC_GET_COMPRESS_OPTION:
4862         case F2FS_IOC_SET_COMPRESS_OPTION:
4863         case F2FS_IOC_DECOMPRESS_FILE:
4864         case F2FS_IOC_COMPRESS_FILE:
4865                 break;
4866         default:
4867                 return -ENOIOCTLCMD;
4868         }
4869         return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4870 }
4871 #endif
4872
4873 const struct file_operations f2fs_file_operations = {
4874         .llseek         = f2fs_llseek,
4875         .read_iter      = f2fs_file_read_iter,
4876         .write_iter     = f2fs_file_write_iter,
4877         .open           = f2fs_file_open,
4878         .release        = f2fs_release_file,
4879         .mmap           = f2fs_file_mmap,
4880         .flush          = f2fs_file_flush,
4881         .fsync          = f2fs_sync_file,
4882         .fallocate      = f2fs_fallocate,
4883         .unlocked_ioctl = f2fs_ioctl,
4884 #ifdef CONFIG_COMPAT
4885         .compat_ioctl   = f2fs_compat_ioctl,
4886 #endif
4887         .splice_read    = generic_file_splice_read,
4888         .splice_write   = iter_file_splice_write,
4889         .fadvise        = f2fs_file_fadvise,
4890 };