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: truncateion starts with cluster */
 575         for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
 576                 block_t blkaddr = le32_to_cpu(*addr);
 577
 578                 if (f2fs_compressed_file(dn->inode) &&
 579                                         !(cluster_index & (cluster_size - 1))) {
 580                         if (compressed_cluster)
 581                                 f2fs_i_compr_blocks_update(dn->inode,
 582                                                         valid_blocks, false);
 583                         compressed_cluster = (blkaddr == COMPRESS_ADDR);
 584                         valid_blocks = 0;
 585                 }
 586
 587                 if (blkaddr == NULL_ADDR)
 588                         continue;
 589
 590                 dn->data_blkaddr = NULL_ADDR;
 591                 f2fs_set_data_blkaddr(dn);
 592
 593                 if (__is_valid_data_blkaddr(blkaddr)) {
 594                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
 595                                         DATA_GENERIC_ENHANCE))
 596                                 continue;
 597                         if (compressed_cluster)
 598                                 valid_blocks++;
 599                 }
 600
 601                 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 602                         clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 603
 604                 f2fs_invalidate_blocks(sbi, blkaddr);
 605
 606                 if (!released || blkaddr != COMPRESS_ADDR)
 607                         nr_free++;
 608         }
 609
 610         if (compressed_cluster)
 611                 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
 612
 613         if (nr_free) {
 614                 pgoff_t fofs;
 615                 /*
 616                  * once we invalidate valid blkaddr in range [ofs, ofs + count],
 617                  * we will invalidate all blkaddr in the whole range.
 618                  */
 619                 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
 620                                                         dn->inode) + ofs;
 621                 f2fs_update_extent_cache_range(dn, fofs, 0, len);
 622                 dec_valid_block_count(sbi, dn->inode, nr_free);
 623         }
 624         dn->ofs_in_node = ofs;
 625
 626         f2fs_update_time(sbi, REQ_TIME);
 627         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 628                                          dn->ofs_in_node, nr_free);
 629 }
 630
 631 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
 632 {
 633         f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
 634 }
 635
 636 static int truncate_partial_data_page(struct inode *inode, u64 from,
 637                                                                 bool cache_only)
 638 {
 639         loff_t offset = from & (PAGE_SIZE - 1);
 640         pgoff_t index = from >> PAGE_SHIFT;
 641         struct address_space *mapping = inode->i_mapping;
 642         struct page *page;
 643
 644         if (!offset && !cache_only)
 645                 return 0;
 646
 647         if (cache_only) {
 648                 page = find_lock_page(mapping, index);
 649                 if (page && PageUptodate(page))
 650                         goto truncate_out;
 651                 f2fs_put_page(page, 1);
 652                 return 0;
 653         }
 654
 655         page = f2fs_get_lock_data_page(inode, index, true);
 656         if (IS_ERR(page))
 657                 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 658 truncate_out:
 659         f2fs_wait_on_page_writeback(page, DATA, true, true);
 660         zero_user(page, offset, PAGE_SIZE - offset);
 661
 662         /* An encrypted inode should have a key and truncate the last page. */
 663         f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
 664         if (!cache_only)
 665                 set_page_dirty(page);
 666         f2fs_put_page(page, 1);
 667         return 0;
 668 }
 669
 670 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
 671 {
 672         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 673         struct dnode_of_data dn;
 674         pgoff_t free_from;
 675         int count = 0, err = 0;
 676         struct page *ipage;
 677         bool truncate_page = false;
 678
 679         trace_f2fs_truncate_blocks_enter(inode, from);
 680
 681         free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 682
 683         if (free_from >= max_file_blocks(inode))
 684                 goto free_partial;
 685
 686         if (lock)
 687                 f2fs_lock_op(sbi);
 688
 689         ipage = f2fs_get_node_page(sbi, inode->i_ino);
 690         if (IS_ERR(ipage)) {
 691                 err = PTR_ERR(ipage);
 692                 goto out;
 693         }
 694
 695         if (f2fs_has_inline_data(inode)) {
 696                 f2fs_truncate_inline_inode(inode, ipage, from);
 697                 f2fs_put_page(ipage, 1);
 698                 truncate_page = true;
 699                 goto out;
 700         }
 701
 702         set_new_dnode(&dn, inode, ipage, NULL, 0);
 703         err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 704         if (err) {
 705                 if (err == -ENOENT)
 706                         goto free_next;
 707                 goto out;
 708         }
 709
 710         count = ADDRS_PER_PAGE(dn.node_page, inode);
 711
 712         count -= dn.ofs_in_node;
 713         f2fs_bug_on(sbi, count < 0);
 714
 715         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 716                 f2fs_truncate_data_blocks_range(&dn, count);
 717                 free_from += count;
 718         }
 719
 720         f2fs_put_dnode(&dn);
 721 free_next:
 722         err = f2fs_truncate_inode_blocks(inode, free_from);
 723 out:
 724         if (lock)
 725                 f2fs_unlock_op(sbi);
 726 free_partial:
 727         /* lastly zero out the first data page */
 728         if (!err)
 729                 err = truncate_partial_data_page(inode, from, truncate_page);
 730
 731         trace_f2fs_truncate_blocks_exit(inode, err);
 732         return err;
 733 }
 734
 735 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
 736 {
 737         u64 free_from = from;
 738         int err;
 739
 740 #ifdef CONFIG_F2FS_FS_COMPRESSION
 741         /*
 742          * for compressed file, only support cluster size
 743          * aligned truncation.
 744          */
 745         if (f2fs_compressed_file(inode))
 746                 free_from = round_up(from,
 747                                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
 748 #endif
 749
 750         err = f2fs_do_truncate_blocks(inode, free_from, lock);
 751         if (err)
 752                 return err;
 753
 754 #ifdef CONFIG_F2FS_FS_COMPRESSION
 755         /*
 756          * For compressed file, after release compress blocks, don't allow write
 757          * direct, but we should allow write direct after truncate to zero.
 758          */
 759         if (f2fs_compressed_file(inode) && !free_from
 760                         && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
 761                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 762
 763         if (from != free_from) {
 764                 err = f2fs_truncate_partial_cluster(inode, from, lock);
 765                 if (err)
 766                         return err;
 767         }
 768 #endif
 769
 770         return 0;
 771 }
 772
 773 int f2fs_truncate(struct inode *inode)
 774 {
 775         int err;
 776
 777         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 778                 return -EIO;
 779
 780         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 781                                 S_ISLNK(inode->i_mode)))
 782                 return 0;
 783
 784         trace_f2fs_truncate(inode);
 785
 786         if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
 787                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
 788                 return -EIO;
 789         }
 790
 791         err = f2fs_dquot_initialize(inode);
 792         if (err)
 793                 return err;
 794
 795         /* we should check inline_data size */
 796         if (!f2fs_may_inline_data(inode)) {
 797                 err = f2fs_convert_inline_inode(inode);
 798                 if (err)
 799                         return err;
 800         }
 801
 802         err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
 803         if (err)
 804                 return err;
 805
 806         inode->i_mtime = inode->i_ctime = current_time(inode);
 807         f2fs_mark_inode_dirty_sync(inode, false);
 808         return 0;
 809 }
 810
 811 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
 812 {
 813         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 814
 815         if (!fscrypt_dio_supported(inode))
 816                 return true;
 817         if (fsverity_active(inode))
 818                 return true;
 819         if (f2fs_compressed_file(inode))
 820                 return true;
 821
 822         /* disallow direct IO if any of devices has unaligned blksize */
 823         if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
 824                 return true;
 825         /*
 826          * for blkzoned device, fallback direct IO to buffered IO, so
 827          * all IOs can be serialized by log-structured write.
 828          */
 829         if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
 830                 return true;
 831         if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi))
 832                 return true;
 833         if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
 834                 return true;
 835
 836         return false;
 837 }
 838
 839 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 840                  struct kstat *stat, u32 request_mask, unsigned int query_flags)
 841 {
 842         struct inode *inode = d_inode(path->dentry);
 843         struct f2fs_inode_info *fi = F2FS_I(inode);
 844         struct f2fs_inode *ri = NULL;
 845         unsigned int flags;
 846
 847         if (f2fs_has_extra_attr(inode) &&
 848                         f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
 849                         F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 850                 stat->result_mask |= STATX_BTIME;
 851                 stat->btime.tv_sec = fi->i_crtime.tv_sec;
 852                 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 853         }
 854
 855         /*
 856          * Return the DIO alignment restrictions if requested.  We only return
 857          * this information when requested, since on encrypted files it might
 858          * take a fair bit of work to get if the file wasn't opened recently.
 859          *
 860          * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
 861          * cannot represent that, so in that case we report no DIO support.
 862          */
 863         if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
 864                 unsigned int bsize = i_blocksize(inode);
 865
 866                 stat->result_mask |= STATX_DIOALIGN;
 867                 if (!f2fs_force_buffered_io(inode, WRITE)) {
 868                         stat->dio_mem_align = bsize;
 869                         stat->dio_offset_align = bsize;
 870                 }
 871         }
 872
 873         flags = fi->i_flags;
 874         if (flags & F2FS_COMPR_FL)
 875                 stat->attributes |= STATX_ATTR_COMPRESSED;
 876         if (flags & F2FS_APPEND_FL)
 877                 stat->attributes |= STATX_ATTR_APPEND;
 878         if (IS_ENCRYPTED(inode))
 879                 stat->attributes |= STATX_ATTR_ENCRYPTED;
 880         if (flags & F2FS_IMMUTABLE_FL)
 881                 stat->attributes |= STATX_ATTR_IMMUTABLE;
 882         if (flags & F2FS_NODUMP_FL)
 883                 stat->attributes |= STATX_ATTR_NODUMP;
 884         if (IS_VERITY(inode))
 885                 stat->attributes |= STATX_ATTR_VERITY;
 886
 887         stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 888                                   STATX_ATTR_APPEND |
 889                                   STATX_ATTR_ENCRYPTED |
 890                                   STATX_ATTR_IMMUTABLE |
 891                                   STATX_ATTR_NODUMP |
 892                                   STATX_ATTR_VERITY);
 893
 894         generic_fillattr(mnt_userns, inode, stat);
 895
 896         /* we need to show initial sectors used for inline_data/dentries */
 897         if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 898                                         f2fs_has_inline_dentry(inode))
 899                 stat->blocks += (stat->size + 511) >> 9;
 900
 901         return 0;
 902 }
 903
 904 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 905 static void __setattr_copy(struct user_namespace *mnt_userns,
 906                            struct inode *inode, const struct iattr *attr)
 907 {
 908         unsigned int ia_valid = attr->ia_valid;
 909
 910         i_uid_update(mnt_userns, attr, inode);
 911         i_gid_update(mnt_userns, attr, inode);
 912         if (ia_valid & ATTR_ATIME)
 913                 inode->i_atime = attr->ia_atime;
 914         if (ia_valid & ATTR_MTIME)
 915                 inode->i_mtime = attr->ia_mtime;
 916         if (ia_valid & ATTR_CTIME)
 917                 inode->i_ctime = attr->ia_ctime;
 918         if (ia_valid & ATTR_MODE) {
 919                 umode_t mode = attr->ia_mode;
 920                 vfsgid_t vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
 921
 922                 if (!vfsgid_in_group_p(vfsgid) &&
 923                     !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
 924                         mode &= ~S_ISGID;
 925                 set_acl_inode(inode, mode);
 926         }
 927 }
 928 #else
 929 #define __setattr_copy setattr_copy
 930 #endif
 931
 932 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 933                  struct iattr *attr)
 934 {
 935         struct inode *inode = d_inode(dentry);
 936         int err;
 937
 938         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 939                 return -EIO;
 940
 941         if (unlikely(IS_IMMUTABLE(inode)))
 942                 return -EPERM;
 943
 944         if (unlikely(IS_APPEND(inode) &&
 945                         (attr->ia_valid & (ATTR_MODE | ATTR_UID |
 946                                   ATTR_GID | ATTR_TIMES_SET))))
 947                 return -EPERM;
 948
 949         if ((attr->ia_valid & ATTR_SIZE) &&
 950                 !f2fs_is_compress_backend_ready(inode))
 951                 return -EOPNOTSUPP;
 952
 953         err = setattr_prepare(mnt_userns, dentry, attr);
 954         if (err)
 955                 return err;
 956
 957         err = fscrypt_prepare_setattr(dentry, attr);
 958         if (err)
 959                 return err;
 960
 961         err = fsverity_prepare_setattr(dentry, attr);
 962         if (err)
 963                 return err;
 964
 965         if (is_quota_modification(mnt_userns, inode, attr)) {
 966                 err = f2fs_dquot_initialize(inode);
 967                 if (err)
 968                         return err;
 969         }
 970         if (i_uid_needs_update(mnt_userns, attr, inode) ||
 971             i_gid_needs_update(mnt_userns, attr, inode)) {
 972                 f2fs_lock_op(F2FS_I_SB(inode));
 973                 err = dquot_transfer(mnt_userns, inode, attr);
 974                 if (err) {
 975                         set_sbi_flag(F2FS_I_SB(inode),
 976                                         SBI_QUOTA_NEED_REPAIR);
 977                         f2fs_unlock_op(F2FS_I_SB(inode));
 978                         return err;
 979                 }
 980                 /*
 981                  * update uid/gid under lock_op(), so that dquot and inode can
 982                  * be updated atomically.
 983                  */
 984                 i_uid_update(mnt_userns, attr, inode);
 985                 i_gid_update(mnt_userns, attr, inode);
 986                 f2fs_mark_inode_dirty_sync(inode, true);
 987                 f2fs_unlock_op(F2FS_I_SB(inode));
 988         }
 989
 990         if (attr->ia_valid & ATTR_SIZE) {
 991                 loff_t old_size = i_size_read(inode);
 992
 993                 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
 994                         /*
 995                          * should convert inline inode before i_size_write to
 996                          * keep smaller than inline_data size with inline flag.
 997                          */
 998                         err = f2fs_convert_inline_inode(inode);
 999                         if (err)
1000                                 return err;
1001                 }
1002
1003                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1004                 filemap_invalidate_lock(inode->i_mapping);
1005
1006                 truncate_setsize(inode, attr->ia_size);
1007
1008                 if (attr->ia_size <= old_size)
1009                         err = f2fs_truncate(inode);
1010                 /*
1011                  * do not trim all blocks after i_size if target size is
1012                  * larger than i_size.
1013                  */
1014                 filemap_invalidate_unlock(inode->i_mapping);
1015                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1016                 if (err)
1017                         return err;
1018
1019                 spin_lock(&F2FS_I(inode)->i_size_lock);
1020                 inode->i_mtime = inode->i_ctime = current_time(inode);
1021                 F2FS_I(inode)->last_disk_size = i_size_read(inode);
1022                 spin_unlock(&F2FS_I(inode)->i_size_lock);
1023         }
1024
1025         __setattr_copy(mnt_userns, inode, attr);
1026
1027         if (attr->ia_valid & ATTR_MODE) {
1028                 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
1029
1030                 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1031                         if (!err)
1032                                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
1033                         clear_inode_flag(inode, FI_ACL_MODE);
1034                 }
1035         }
1036
1037         /* file size may changed here */
1038         f2fs_mark_inode_dirty_sync(inode, true);
1039
1040         /* inode change will produce dirty node pages flushed by checkpoint */
1041         f2fs_balance_fs(F2FS_I_SB(inode), true);
1042
1043         return err;
1044 }
1045
1046 const struct inode_operations f2fs_file_inode_operations = {
1047         .getattr        = f2fs_getattr,
1048         .setattr        = f2fs_setattr,
1049         .get_acl        = f2fs_get_acl,
1050         .set_acl        = f2fs_set_acl,
1051         .listxattr      = f2fs_listxattr,
1052         .fiemap         = f2fs_fiemap,
1053         .fileattr_get   = f2fs_fileattr_get,
1054         .fileattr_set   = f2fs_fileattr_set,
1055 };
1056
1057 static int fill_zero(struct inode *inode, pgoff_t index,
1058                                         loff_t start, loff_t len)
1059 {
1060         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1061         struct page *page;
1062
1063         if (!len)
1064                 return 0;
1065
1066         f2fs_balance_fs(sbi, true);
1067
1068         f2fs_lock_op(sbi);
1069         page = f2fs_get_new_data_page(inode, NULL, index, false);
1070         f2fs_unlock_op(sbi);
1071
1072         if (IS_ERR(page))
1073                 return PTR_ERR(page);
1074
1075         f2fs_wait_on_page_writeback(page, DATA, true, true);
1076         zero_user(page, start, len);
1077         set_page_dirty(page);
1078         f2fs_put_page(page, 1);
1079         return 0;
1080 }
1081
1082 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1083 {
1084         int err;
1085
1086         while (pg_start < pg_end) {
1087                 struct dnode_of_data dn;
1088                 pgoff_t end_offset, count;
1089
1090                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1091                 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1092                 if (err) {
1093                         if (err == -ENOENT) {
1094                                 pg_start = f2fs_get_next_page_offset(&dn,
1095                                                                 pg_start);
1096                                 continue;
1097                         }
1098                         return err;
1099                 }
1100
1101                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1102                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1103
1104                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1105
1106                 f2fs_truncate_data_blocks_range(&dn, count);
1107                 f2fs_put_dnode(&dn);
1108
1109                 pg_start += count;
1110         }
1111         return 0;
1112 }
1113
1114 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1115 {
1116         pgoff_t pg_start, pg_end;
1117         loff_t off_start, off_end;
1118         int ret;
1119
1120         ret = f2fs_convert_inline_inode(inode);
1121         if (ret)
1122                 return ret;
1123
1124         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1125         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1126
1127         off_start = offset & (PAGE_SIZE - 1);
1128         off_end = (offset + len) & (PAGE_SIZE - 1);
1129
1130         if (pg_start == pg_end) {
1131                 ret = fill_zero(inode, pg_start, off_start,
1132                                                 off_end - off_start);
1133                 if (ret)
1134                         return ret;
1135         } else {
1136                 if (off_start) {
1137                         ret = fill_zero(inode, pg_start++, off_start,
1138                                                 PAGE_SIZE - off_start);
1139                         if (ret)
1140                                 return ret;
1141                 }
1142                 if (off_end) {
1143                         ret = fill_zero(inode, pg_end, 0, off_end);
1144                         if (ret)
1145                                 return ret;
1146                 }
1147
1148                 if (pg_start < pg_end) {
1149                         loff_t blk_start, blk_end;
1150                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1151
1152                         f2fs_balance_fs(sbi, true);
1153
1154                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
1155                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
1156
1157                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1158                         filemap_invalidate_lock(inode->i_mapping);
1159
1160                         truncate_pagecache_range(inode, blk_start, blk_end - 1);
1161
1162                         f2fs_lock_op(sbi);
1163                         ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1164                         f2fs_unlock_op(sbi);
1165
1166                         filemap_invalidate_unlock(inode->i_mapping);
1167                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1168                 }
1169         }
1170
1171         return ret;
1172 }
1173
1174 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1175                                 int *do_replace, pgoff_t off, pgoff_t len)
1176 {
1177         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1178         struct dnode_of_data dn;
1179         int ret, done, i;
1180
1181 next_dnode:
1182         set_new_dnode(&dn, inode, NULL, NULL, 0);
1183         ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1184         if (ret && ret != -ENOENT) {
1185                 return ret;
1186         } else if (ret == -ENOENT) {
1187                 if (dn.max_level == 0)
1188                         return -ENOENT;
1189                 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1190                                                 dn.ofs_in_node, len);
1191                 blkaddr += done;
1192                 do_replace += done;
1193                 goto next;
1194         }
1195
1196         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1197                                                         dn.ofs_in_node, len);
1198         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1199                 *blkaddr = f2fs_data_blkaddr(&dn);
1200
1201                 if (__is_valid_data_blkaddr(*blkaddr) &&
1202                         !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1203                                         DATA_GENERIC_ENHANCE)) {
1204                         f2fs_put_dnode(&dn);
1205                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1206                         return -EFSCORRUPTED;
1207                 }
1208
1209                 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1210
1211                         if (f2fs_lfs_mode(sbi)) {
1212                                 f2fs_put_dnode(&dn);
1213                                 return -EOPNOTSUPP;
1214                         }
1215
1216                         /* do not invalidate this block address */
1217                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1218                         *do_replace = 1;
1219                 }
1220         }
1221         f2fs_put_dnode(&dn);
1222 next:
1223         len -= done;
1224         off += done;
1225         if (len)
1226                 goto next_dnode;
1227         return 0;
1228 }
1229
1230 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1231                                 int *do_replace, pgoff_t off, int len)
1232 {
1233         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1234         struct dnode_of_data dn;
1235         int ret, i;
1236
1237         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1238                 if (*do_replace == 0)
1239                         continue;
1240
1241                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1242                 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1243                 if (ret) {
1244                         dec_valid_block_count(sbi, inode, 1);
1245                         f2fs_invalidate_blocks(sbi, *blkaddr);
1246                 } else {
1247                         f2fs_update_data_blkaddr(&dn, *blkaddr);
1248                 }
1249                 f2fs_put_dnode(&dn);
1250         }
1251         return 0;
1252 }
1253
1254 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1255                         block_t *blkaddr, int *do_replace,
1256                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1257 {
1258         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1259         pgoff_t i = 0;
1260         int ret;
1261
1262         while (i < len) {
1263                 if (blkaddr[i] == NULL_ADDR && !full) {
1264                         i++;
1265                         continue;
1266                 }
1267
1268                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1269                         struct dnode_of_data dn;
1270                         struct node_info ni;
1271                         size_t new_size;
1272                         pgoff_t ilen;
1273
1274                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1275                         ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1276                         if (ret)
1277                                 return ret;
1278
1279                         ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1280                         if (ret) {
1281                                 f2fs_put_dnode(&dn);
1282                                 return ret;
1283                         }
1284
1285                         ilen = min((pgoff_t)
1286                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1287                                                 dn.ofs_in_node, len - i);
1288                         do {
1289                                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1290                                 f2fs_truncate_data_blocks_range(&dn, 1);
1291
1292                                 if (do_replace[i]) {
1293                                         f2fs_i_blocks_write(src_inode,
1294                                                         1, false, false);
1295                                         f2fs_i_blocks_write(dst_inode,
1296                                                         1, true, false);
1297                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1298                                         blkaddr[i], ni.version, true, false);
1299
1300                                         do_replace[i] = 0;
1301                                 }
1302                                 dn.ofs_in_node++;
1303                                 i++;
1304                                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1305                                 if (dst_inode->i_size < new_size)
1306                                         f2fs_i_size_write(dst_inode, new_size);
1307                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1308
1309                         f2fs_put_dnode(&dn);
1310                 } else {
1311                         struct page *psrc, *pdst;
1312
1313                         psrc = f2fs_get_lock_data_page(src_inode,
1314                                                         src + i, true);
1315                         if (IS_ERR(psrc))
1316                                 return PTR_ERR(psrc);
1317                         pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1318                                                                 true);
1319                         if (IS_ERR(pdst)) {
1320                                 f2fs_put_page(psrc, 1);
1321                                 return PTR_ERR(pdst);
1322                         }
1323                         memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1324                         set_page_dirty(pdst);
1325                         f2fs_put_page(pdst, 1);
1326                         f2fs_put_page(psrc, 1);
1327
1328                         ret = f2fs_truncate_hole(src_inode,
1329                                                 src + i, src + i + 1);
1330                         if (ret)
1331                                 return ret;
1332                         i++;
1333                 }
1334         }
1335         return 0;
1336 }
1337
1338 static int __exchange_data_block(struct inode *src_inode,
1339                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1340                         pgoff_t len, bool full)
1341 {
1342         block_t *src_blkaddr;
1343         int *do_replace;
1344         pgoff_t olen;
1345         int ret;
1346
1347         while (len) {
1348                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1349
1350                 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1351                                         array_size(olen, sizeof(block_t)),
1352                                         GFP_NOFS);
1353                 if (!src_blkaddr)
1354                         return -ENOMEM;
1355
1356                 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1357                                         array_size(olen, sizeof(int)),
1358                                         GFP_NOFS);
1359                 if (!do_replace) {
1360                         kvfree(src_blkaddr);
1361                         return -ENOMEM;
1362                 }
1363
1364                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1365                                         do_replace, src, olen);
1366                 if (ret)
1367                         goto roll_back;
1368
1369                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1370                                         do_replace, src, dst, olen, full);
1371                 if (ret)
1372                         goto roll_back;
1373
1374                 src += olen;
1375                 dst += olen;
1376                 len -= olen;
1377
1378                 kvfree(src_blkaddr);
1379                 kvfree(do_replace);
1380         }
1381         return 0;
1382
1383 roll_back:
1384         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1385         kvfree(src_blkaddr);
1386         kvfree(do_replace);
1387         return ret;
1388 }
1389
1390 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1391 {
1392         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1393         pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1394         pgoff_t start = offset >> PAGE_SHIFT;
1395         pgoff_t end = (offset + len) >> PAGE_SHIFT;
1396         int ret;
1397
1398         f2fs_balance_fs(sbi, true);
1399
1400         /* avoid gc operation during block exchange */
1401         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1402         filemap_invalidate_lock(inode->i_mapping);
1403
1404         f2fs_lock_op(sbi);
1405         f2fs_drop_extent_tree(inode);
1406         truncate_pagecache(inode, offset);
1407         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1408         f2fs_unlock_op(sbi);
1409
1410         filemap_invalidate_unlock(inode->i_mapping);
1411         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1412         return ret;
1413 }
1414
1415 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1416 {
1417         loff_t new_size;
1418         int ret;
1419
1420         if (offset + len >= i_size_read(inode))
1421                 return -EINVAL;
1422
1423         /* collapse range should be aligned to block size of f2fs. */
1424         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1425                 return -EINVAL;
1426
1427         ret = f2fs_convert_inline_inode(inode);
1428         if (ret)
1429                 return ret;
1430
1431         /* write out all dirty pages from offset */
1432         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1433         if (ret)
1434                 return ret;
1435
1436         ret = f2fs_do_collapse(inode, offset, len);
1437         if (ret)
1438                 return ret;
1439
1440         /* write out all moved pages, if possible */
1441         filemap_invalidate_lock(inode->i_mapping);
1442         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1443         truncate_pagecache(inode, offset);
1444
1445         new_size = i_size_read(inode) - len;
1446         ret = f2fs_truncate_blocks(inode, new_size, true);
1447         filemap_invalidate_unlock(inode->i_mapping);
1448         if (!ret)
1449                 f2fs_i_size_write(inode, new_size);
1450         return ret;
1451 }
1452
1453 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1454                                                                 pgoff_t end)
1455 {
1456         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1457         pgoff_t index = start;
1458         unsigned int ofs_in_node = dn->ofs_in_node;
1459         blkcnt_t count = 0;
1460         int ret;
1461
1462         for (; index < end; index++, dn->ofs_in_node++) {
1463                 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1464                         count++;
1465         }
1466
1467         dn->ofs_in_node = ofs_in_node;
1468         ret = f2fs_reserve_new_blocks(dn, count);
1469         if (ret)
1470                 return ret;
1471
1472         dn->ofs_in_node = ofs_in_node;
1473         for (index = start; index < end; index++, dn->ofs_in_node++) {
1474                 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1475                 /*
1476                  * f2fs_reserve_new_blocks will not guarantee entire block
1477                  * allocation.
1478                  */
1479                 if (dn->data_blkaddr == NULL_ADDR) {
1480                         ret = -ENOSPC;
1481                         break;
1482                 }
1483
1484                 if (dn->data_blkaddr == NEW_ADDR)
1485                         continue;
1486
1487                 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1488                                         DATA_GENERIC_ENHANCE)) {
1489                         ret = -EFSCORRUPTED;
1490                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1491                         break;
1492                 }
1493
1494                 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1495                 dn->data_blkaddr = NEW_ADDR;
1496                 f2fs_set_data_blkaddr(dn);
1497         }
1498
1499         f2fs_update_extent_cache_range(dn, start, 0, index - start);
1500
1501         return ret;
1502 }
1503
1504 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1505                                                                 int mode)
1506 {
1507         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1508         struct address_space *mapping = inode->i_mapping;
1509         pgoff_t index, pg_start, pg_end;
1510         loff_t new_size = i_size_read(inode);
1511         loff_t off_start, off_end;
1512         int ret = 0;
1513
1514         ret = inode_newsize_ok(inode, (len + offset));
1515         if (ret)
1516                 return ret;
1517
1518         ret = f2fs_convert_inline_inode(inode);
1519         if (ret)
1520                 return ret;
1521
1522         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1523         if (ret)
1524                 return ret;
1525
1526         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1527         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1528
1529         off_start = offset & (PAGE_SIZE - 1);
1530         off_end = (offset + len) & (PAGE_SIZE - 1);
1531
1532         if (pg_start == pg_end) {
1533                 ret = fill_zero(inode, pg_start, off_start,
1534                                                 off_end - off_start);
1535                 if (ret)
1536                         return ret;
1537
1538                 new_size = max_t(loff_t, new_size, offset + len);
1539         } else {
1540                 if (off_start) {
1541                         ret = fill_zero(inode, pg_start++, off_start,
1542                                                 PAGE_SIZE - off_start);
1543                         if (ret)
1544                                 return ret;
1545
1546                         new_size = max_t(loff_t, new_size,
1547                                         (loff_t)pg_start << PAGE_SHIFT);
1548                 }
1549
1550                 for (index = pg_start; index < pg_end;) {
1551                         struct dnode_of_data dn;
1552                         unsigned int end_offset;
1553                         pgoff_t end;
1554
1555                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1556                         filemap_invalidate_lock(mapping);
1557
1558                         truncate_pagecache_range(inode,
1559                                 (loff_t)index << PAGE_SHIFT,
1560                                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1561
1562                         f2fs_lock_op(sbi);
1563
1564                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1565                         ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1566                         if (ret) {
1567                                 f2fs_unlock_op(sbi);
1568                                 filemap_invalidate_unlock(mapping);
1569                                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1570                                 goto out;
1571                         }
1572
1573                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1574                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1575
1576                         ret = f2fs_do_zero_range(&dn, index, end);
1577                         f2fs_put_dnode(&dn);
1578
1579                         f2fs_unlock_op(sbi);
1580                         filemap_invalidate_unlock(mapping);
1581                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1582
1583                         f2fs_balance_fs(sbi, dn.node_changed);
1584
1585                         if (ret)
1586                                 goto out;
1587
1588                         index = end;
1589                         new_size = max_t(loff_t, new_size,
1590                                         (loff_t)index << PAGE_SHIFT);
1591                 }
1592
1593                 if (off_end) {
1594                         ret = fill_zero(inode, pg_end, 0, off_end);
1595                         if (ret)
1596                                 goto out;
1597
1598                         new_size = max_t(loff_t, new_size, offset + len);
1599                 }
1600         }
1601
1602 out:
1603         if (new_size > i_size_read(inode)) {
1604                 if (mode & FALLOC_FL_KEEP_SIZE)
1605                         file_set_keep_isize(inode);
1606                 else
1607                         f2fs_i_size_write(inode, new_size);
1608         }
1609         return ret;
1610 }
1611
1612 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1613 {
1614         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1615         struct address_space *mapping = inode->i_mapping;
1616         pgoff_t nr, pg_start, pg_end, delta, idx;
1617         loff_t new_size;
1618         int ret = 0;
1619
1620         new_size = i_size_read(inode) + len;
1621         ret = inode_newsize_ok(inode, new_size);
1622         if (ret)
1623                 return ret;
1624
1625         if (offset >= i_size_read(inode))
1626                 return -EINVAL;
1627
1628         /* insert range should be aligned to block size of f2fs. */
1629         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1630                 return -EINVAL;
1631
1632         ret = f2fs_convert_inline_inode(inode);
1633         if (ret)
1634                 return ret;
1635
1636         f2fs_balance_fs(sbi, true);
1637
1638         filemap_invalidate_lock(mapping);
1639         ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1640         filemap_invalidate_unlock(mapping);
1641         if (ret)
1642                 return ret;
1643
1644         /* write out all dirty pages from offset */
1645         ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1646         if (ret)
1647                 return ret;
1648
1649         pg_start = offset >> PAGE_SHIFT;
1650         pg_end = (offset + len) >> PAGE_SHIFT;
1651         delta = pg_end - pg_start;
1652         idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1653
1654         /* avoid gc operation during block exchange */
1655         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1656         filemap_invalidate_lock(mapping);
1657         truncate_pagecache(inode, offset);
1658
1659         while (!ret && idx > pg_start) {
1660                 nr = idx - pg_start;
1661                 if (nr > delta)
1662                         nr = delta;
1663                 idx -= nr;
1664
1665                 f2fs_lock_op(sbi);
1666                 f2fs_drop_extent_tree(inode);
1667
1668                 ret = __exchange_data_block(inode, inode, idx,
1669                                         idx + delta, nr, false);
1670                 f2fs_unlock_op(sbi);
1671         }
1672         filemap_invalidate_unlock(mapping);
1673         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1674
1675         /* write out all moved pages, if possible */
1676         filemap_invalidate_lock(mapping);
1677         filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1678         truncate_pagecache(inode, offset);
1679         filemap_invalidate_unlock(mapping);
1680
1681         if (!ret)
1682                 f2fs_i_size_write(inode, new_size);
1683         return ret;
1684 }
1685
1686 static int expand_inode_data(struct inode *inode, loff_t offset,
1687                                         loff_t len, int mode)
1688 {
1689         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1690         struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1691                         .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1692                         .m_may_create = true };
1693         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1694                         .init_gc_type = FG_GC,
1695                         .should_migrate_blocks = false,
1696                         .err_gc_skipped = true,
1697                         .nr_free_secs = 0 };
1698         pgoff_t pg_start, pg_end;
1699         loff_t new_size = i_size_read(inode);
1700         loff_t off_end;
1701         block_t expanded = 0;
1702         int err;
1703
1704         err = inode_newsize_ok(inode, (len + offset));
1705         if (err)
1706                 return err;
1707
1708         err = f2fs_convert_inline_inode(inode);
1709         if (err)
1710                 return err;
1711
1712         f2fs_balance_fs(sbi, true);
1713
1714         pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1715         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1716         off_end = (offset + len) & (PAGE_SIZE - 1);
1717
1718         map.m_lblk = pg_start;
1719         map.m_len = pg_end - pg_start;
1720         if (off_end)
1721                 map.m_len++;
1722
1723         if (!map.m_len)
1724                 return 0;
1725
1726         if (f2fs_is_pinned_file(inode)) {
1727                 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1728                 block_t sec_len = roundup(map.m_len, sec_blks);
1729
1730                 map.m_len = sec_blks;
1731 next_alloc:
1732                 if (has_not_enough_free_secs(sbi, 0,
1733                         GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1734                         f2fs_down_write(&sbi->gc_lock);
1735                         err = f2fs_gc(sbi, &gc_control);
1736                         if (err && err != -ENODATA)
1737                                 goto out_err;
1738                 }
1739
1740                 f2fs_down_write(&sbi->pin_sem);
1741
1742                 f2fs_lock_op(sbi);
1743                 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1744                 f2fs_unlock_op(sbi);
1745
1746                 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1747                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1748                 file_dont_truncate(inode);
1749
1750                 f2fs_up_write(&sbi->pin_sem);
1751
1752                 expanded += map.m_len;
1753                 sec_len -= map.m_len;
1754                 map.m_lblk += map.m_len;
1755                 if (!err && sec_len)
1756                         goto next_alloc;
1757
1758                 map.m_len = expanded;
1759         } else {
1760                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1761                 expanded = map.m_len;
1762         }
1763 out_err:
1764         if (err) {
1765                 pgoff_t last_off;
1766
1767                 if (!expanded)
1768                         return err;
1769
1770                 last_off = pg_start + expanded - 1;
1771
1772                 /* update new size to the failed position */
1773                 new_size = (last_off == pg_end) ? offset + len :
1774                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1775         } else {
1776                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1777         }
1778
1779         if (new_size > i_size_read(inode)) {
1780                 if (mode & FALLOC_FL_KEEP_SIZE)
1781                         file_set_keep_isize(inode);
1782                 else
1783                         f2fs_i_size_write(inode, new_size);
1784         }
1785
1786         return err;
1787 }
1788
1789 static long f2fs_fallocate(struct file *file, int mode,
1790                                 loff_t offset, loff_t len)
1791 {
1792         struct inode *inode = file_inode(file);
1793         long ret = 0;
1794
1795         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1796                 return -EIO;
1797         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1798                 return -ENOSPC;
1799         if (!f2fs_is_compress_backend_ready(inode))
1800                 return -EOPNOTSUPP;
1801
1802         /* f2fs only support ->fallocate for regular file */
1803         if (!S_ISREG(inode->i_mode))
1804                 return -EINVAL;
1805
1806         if (IS_ENCRYPTED(inode) &&
1807                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1808                 return -EOPNOTSUPP;
1809
1810         /*
1811          * Pinned file should not support partial trucation since the block
1812          * can be used by applications.
1813          */
1814         if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1815                 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1816                         FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1817                 return -EOPNOTSUPP;
1818
1819         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1820                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1821                         FALLOC_FL_INSERT_RANGE))
1822                 return -EOPNOTSUPP;
1823
1824         inode_lock(inode);
1825
1826         ret = file_modified(file);
1827         if (ret)
1828                 goto out;
1829
1830         if (mode & FALLOC_FL_PUNCH_HOLE) {
1831                 if (offset >= inode->i_size)
1832                         goto out;
1833
1834                 ret = punch_hole(inode, offset, len);
1835         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1836                 ret = f2fs_collapse_range(inode, offset, len);
1837         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1838                 ret = f2fs_zero_range(inode, offset, len, mode);
1839         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1840                 ret = f2fs_insert_range(inode, offset, len);
1841         } else {
1842                 ret = expand_inode_data(inode, offset, len, mode);
1843         }
1844
1845         if (!ret) {
1846                 inode->i_mtime = inode->i_ctime = current_time(inode);
1847                 f2fs_mark_inode_dirty_sync(inode, false);
1848                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1849         }
1850
1851 out:
1852         inode_unlock(inode);
1853
1854         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1855         return ret;
1856 }
1857
1858 static int f2fs_release_file(struct inode *inode, struct file *filp)
1859 {
1860         /*
1861          * f2fs_relase_file is called at every close calls. So we should
1862          * not drop any inmemory pages by close called by other process.
1863          */
1864         if (!(filp->f_mode & FMODE_WRITE) ||
1865                         atomic_read(&inode->i_writecount) != 1)
1866                 return 0;
1867
1868         inode_lock(inode);
1869         f2fs_abort_atomic_write(inode, true);
1870         inode_unlock(inode);
1871
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                                 (current->flags & PF_EXITING)) {
1887                 inode_lock(inode);
1888                 f2fs_abort_atomic_write(inode, true);
1889                 inode_unlock(inode);
1890         }
1891
1892         return 0;
1893 }
1894
1895 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1896 {
1897         struct f2fs_inode_info *fi = F2FS_I(inode);
1898         u32 masked_flags = fi->i_flags & mask;
1899
1900         /* mask can be shrunk by flags_valid selector */
1901         iflags &= mask;
1902
1903         /* Is it quota file? Do not allow user to mess with it */
1904         if (IS_NOQUOTA(inode))
1905                 return -EPERM;
1906
1907         if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1908                 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1909                         return -EOPNOTSUPP;
1910                 if (!f2fs_empty_dir(inode))
1911                         return -ENOTEMPTY;
1912         }
1913
1914         if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1915                 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1916                         return -EOPNOTSUPP;
1917                 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1918                         return -EINVAL;
1919         }
1920
1921         if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1922                 if (masked_flags & F2FS_COMPR_FL) {
1923                         if (!f2fs_disable_compressed_file(inode))
1924                                 return -EINVAL;
1925                 } else {
1926                         /* try to convert inline_data to support compression */
1927                         int err = f2fs_convert_inline_inode(inode);
1928                         if (err)
1929                                 return err;
1930                         if (!f2fs_may_compress(inode))
1931                                 return -EINVAL;
1932                         if (S_ISREG(inode->i_mode) && F2FS_HAS_BLOCKS(inode))
1933                                 return -EINVAL;
1934                         if (set_compress_context(inode))
1935                                 return -EOPNOTSUPP;
1936                 }
1937         }
1938
1939         fi->i_flags = iflags | (fi->i_flags & ~mask);
1940         f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1941                                         (fi->i_flags & F2FS_NOCOMP_FL));
1942
1943         if (fi->i_flags & F2FS_PROJINHERIT_FL)
1944                 set_inode_flag(inode, FI_PROJ_INHERIT);
1945         else
1946                 clear_inode_flag(inode, FI_PROJ_INHERIT);
1947
1948         inode->i_ctime = current_time(inode);
1949         f2fs_set_inode_flags(inode);
1950         f2fs_mark_inode_dirty_sync(inode, true);
1951         return 0;
1952 }
1953
1954 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1955
1956 /*
1957  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1958  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1959  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1960  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1961  *
1962  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1963  * FS_IOC_FSSETXATTR is done by the VFS.
1964  */
1965
1966 static const struct {
1967         u32 iflag;
1968         u32 fsflag;
1969 } f2fs_fsflags_map[] = {
1970         { F2FS_COMPR_FL,        FS_COMPR_FL },
1971         { F2FS_SYNC_FL,         FS_SYNC_FL },
1972         { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1973         { F2FS_APPEND_FL,       FS_APPEND_FL },
1974         { F2FS_NODUMP_FL,       FS_NODUMP_FL },
1975         { F2FS_NOATIME_FL,      FS_NOATIME_FL },
1976         { F2FS_NOCOMP_FL,       FS_NOCOMP_FL },
1977         { F2FS_INDEX_FL,        FS_INDEX_FL },
1978         { F2FS_DIRSYNC_FL,      FS_DIRSYNC_FL },
1979         { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1980         { F2FS_CASEFOLD_FL,     FS_CASEFOLD_FL },
1981 };
1982
1983 #define F2FS_GETTABLE_FS_FL (           \
1984                 FS_COMPR_FL |           \
1985                 FS_SYNC_FL |            \
1986                 FS_IMMUTABLE_FL |       \
1987                 FS_APPEND_FL |          \
1988                 FS_NODUMP_FL |          \
1989                 FS_NOATIME_FL |         \
1990                 FS_NOCOMP_FL |          \
1991                 FS_INDEX_FL |           \
1992                 FS_DIRSYNC_FL |         \
1993                 FS_PROJINHERIT_FL |     \
1994                 FS_ENCRYPT_FL |         \
1995                 FS_INLINE_DATA_FL |     \
1996                 FS_NOCOW_FL |           \
1997                 FS_VERITY_FL |          \
1998                 FS_CASEFOLD_FL)
1999
2000 #define F2FS_SETTABLE_FS_FL (           \
2001                 FS_COMPR_FL |           \
2002                 FS_SYNC_FL |            \
2003                 FS_IMMUTABLE_FL |       \
2004                 FS_APPEND_FL |          \
2005                 FS_NODUMP_FL |          \
2006                 FS_NOATIME_FL |         \
2007                 FS_NOCOMP_FL |          \
2008                 FS_DIRSYNC_FL |         \
2009                 FS_PROJINHERIT_FL |     \
2010                 FS_CASEFOLD_FL)
2011
2012 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2013 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2014 {
2015         u32 fsflags = 0;
2016         int i;
2017
2018         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2019                 if (iflags & f2fs_fsflags_map[i].iflag)
2020                         fsflags |= f2fs_fsflags_map[i].fsflag;
2021
2022         return fsflags;
2023 }
2024
2025 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2026 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2027 {
2028         u32 iflags = 0;
2029         int i;
2030
2031         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2032                 if (fsflags & f2fs_fsflags_map[i].fsflag)
2033                         iflags |= f2fs_fsflags_map[i].iflag;
2034
2035         return iflags;
2036 }
2037
2038 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2039 {
2040         struct inode *inode = file_inode(filp);
2041
2042         return put_user(inode->i_generation, (int __user *)arg);
2043 }
2044
2045 static int f2fs_ioc_start_atomic_write(struct file *filp)
2046 {
2047         struct inode *inode = file_inode(filp);
2048         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2049         struct f2fs_inode_info *fi = F2FS_I(inode);
2050         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2051         struct inode *pinode;
2052         loff_t isize;
2053         int ret;
2054
2055         if (!inode_owner_or_capable(mnt_userns, inode))
2056                 return -EACCES;
2057
2058         if (!S_ISREG(inode->i_mode))
2059                 return -EINVAL;
2060
2061         if (filp->f_flags & O_DIRECT)
2062                 return -EINVAL;
2063
2064         ret = mnt_want_write_file(filp);
2065         if (ret)
2066                 return ret;
2067
2068         inode_lock(inode);
2069
2070         if (!f2fs_disable_compressed_file(inode)) {
2071                 ret = -EINVAL;
2072                 goto out;
2073         }
2074
2075         if (f2fs_is_atomic_file(inode))
2076                 goto out;
2077
2078         ret = f2fs_convert_inline_inode(inode);
2079         if (ret)
2080                 goto out;
2081
2082         f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2083
2084         /*
2085          * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2086          * f2fs_is_atomic_file.
2087          */
2088         if (get_dirty_pages(inode))
2089                 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2090                           inode->i_ino, get_dirty_pages(inode));
2091         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2092         if (ret) {
2093                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2094                 goto out;
2095         }
2096
2097         /* Check if the inode already has a COW inode */
2098         if (fi->cow_inode == NULL) {
2099                 /* Create a COW inode for atomic write */
2100                 pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2101                 if (IS_ERR(pinode)) {
2102                         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2103                         ret = PTR_ERR(pinode);
2104                         goto out;
2105                 }
2106
2107                 ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2108                 iput(pinode);
2109                 if (ret) {
2110                         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2111                         goto out;
2112                 }
2113
2114                 set_inode_flag(fi->cow_inode, FI_COW_FILE);
2115                 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2116         } else {
2117                 /* Reuse the already created COW inode */
2118                 f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2119         }
2120
2121         f2fs_write_inode(inode, NULL);
2122
2123         isize = i_size_read(inode);
2124         fi->original_i_size = isize;
2125         f2fs_i_size_write(fi->cow_inode, isize);
2126
2127         stat_inc_atomic_inode(inode);
2128
2129         set_inode_flag(inode, FI_ATOMIC_FILE);
2130         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2131
2132         f2fs_update_time(sbi, REQ_TIME);
2133         fi->atomic_write_task = current;
2134         stat_update_max_atomic_write(inode);
2135         fi->atomic_write_cnt = 0;
2136 out:
2137         inode_unlock(inode);
2138         mnt_drop_write_file(filp);
2139         return ret;
2140 }
2141
2142 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2143 {
2144         struct inode *inode = file_inode(filp);
2145         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2146         int ret;
2147
2148         if (!inode_owner_or_capable(mnt_userns, inode))
2149                 return -EACCES;
2150
2151         ret = mnt_want_write_file(filp);
2152         if (ret)
2153                 return ret;
2154
2155         f2fs_balance_fs(F2FS_I_SB(inode), true);
2156
2157         inode_lock(inode);
2158
2159         if (f2fs_is_atomic_file(inode)) {
2160                 ret = f2fs_commit_atomic_write(inode);
2161                 if (!ret)
2162                         ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2163
2164                 f2fs_abort_atomic_write(inode, ret);
2165         } else {
2166                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2167         }
2168
2169         inode_unlock(inode);
2170         mnt_drop_write_file(filp);
2171         return ret;
2172 }
2173
2174 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2175 {
2176         struct inode *inode = file_inode(filp);
2177         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2178         int ret;
2179
2180         if (!inode_owner_or_capable(mnt_userns, inode))
2181                 return -EACCES;
2182
2183         ret = mnt_want_write_file(filp);
2184         if (ret)
2185                 return ret;
2186
2187         inode_lock(inode);
2188
2189         f2fs_abort_atomic_write(inode, true);
2190
2191         inode_unlock(inode);
2192
2193         mnt_drop_write_file(filp);
2194         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2195         return ret;
2196 }
2197
2198 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2199 {
2200         struct inode *inode = file_inode(filp);
2201         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2202         struct super_block *sb = sbi->sb;
2203         __u32 in;
2204         int ret = 0;
2205
2206         if (!capable(CAP_SYS_ADMIN))
2207                 return -EPERM;
2208
2209         if (get_user(in, (__u32 __user *)arg))
2210                 return -EFAULT;
2211
2212         if (in != F2FS_GOING_DOWN_FULLSYNC) {
2213                 ret = mnt_want_write_file(filp);
2214                 if (ret) {
2215                         if (ret == -EROFS) {
2216                                 ret = 0;
2217                                 f2fs_stop_checkpoint(sbi, false,
2218                                                 STOP_CP_REASON_SHUTDOWN);
2219                                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2220                                 trace_f2fs_shutdown(sbi, in, ret);
2221                         }
2222                         return ret;
2223                 }
2224         }
2225
2226         switch (in) {
2227         case F2FS_GOING_DOWN_FULLSYNC:
2228                 ret = freeze_bdev(sb->s_bdev);
2229                 if (ret)
2230                         goto out;
2231                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2232                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2233                 thaw_bdev(sb->s_bdev);
2234                 break;
2235         case F2FS_GOING_DOWN_METASYNC:
2236                 /* do checkpoint only */
2237                 ret = f2fs_sync_fs(sb, 1);
2238                 if (ret)
2239                         goto out;
2240                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2241                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2242                 break;
2243         case F2FS_GOING_DOWN_NOSYNC:
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_METAFLUSH:
2248                 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2249                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2250                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2251                 break;
2252         case F2FS_GOING_DOWN_NEED_FSCK:
2253                 set_sbi_flag(sbi, SBI_NEED_FSCK);
2254                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2255                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2256                 /* do checkpoint only */
2257                 ret = f2fs_sync_fs(sb, 1);
2258                 goto out;
2259         default:
2260                 ret = -EINVAL;
2261                 goto out;
2262         }
2263
2264         f2fs_stop_gc_thread(sbi);
2265         f2fs_stop_discard_thread(sbi);
2266
2267         f2fs_drop_discard_cmd(sbi);
2268         clear_opt(sbi, DISCARD);
2269
2270         f2fs_update_time(sbi, REQ_TIME);
2271 out:
2272         if (in != F2FS_GOING_DOWN_FULLSYNC)
2273                 mnt_drop_write_file(filp);
2274
2275         trace_f2fs_shutdown(sbi, in, ret);
2276
2277         return ret;
2278 }
2279
2280 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2281 {
2282         struct inode *inode = file_inode(filp);
2283         struct super_block *sb = inode->i_sb;
2284         struct fstrim_range range;
2285         int ret;
2286
2287         if (!capable(CAP_SYS_ADMIN))
2288                 return -EPERM;
2289
2290         if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2291                 return -EOPNOTSUPP;
2292
2293         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2294                                 sizeof(range)))
2295                 return -EFAULT;
2296
2297         ret = mnt_want_write_file(filp);
2298         if (ret)
2299                 return ret;
2300
2301         range.minlen = max((unsigned int)range.minlen,
2302                            bdev_discard_granularity(sb->s_bdev));
2303         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2304         mnt_drop_write_file(filp);
2305         if (ret < 0)
2306                 return ret;
2307
2308         if (copy_to_user((struct fstrim_range __user *)arg, &range,
2309                                 sizeof(range)))
2310                 return -EFAULT;
2311         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2312         return 0;
2313 }
2314
2315 static bool uuid_is_nonzero(__u8 u[16])
2316 {
2317         int i;
2318
2319         for (i = 0; i < 16; i++)
2320                 if (u[i])
2321                         return true;
2322         return false;
2323 }
2324
2325 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2326 {
2327         struct inode *inode = file_inode(filp);
2328
2329         if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2330                 return -EOPNOTSUPP;
2331
2332         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2333
2334         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2335 }
2336
2337 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2338 {
2339         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2340                 return -EOPNOTSUPP;
2341         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2342 }
2343
2344 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2345 {
2346         struct inode *inode = file_inode(filp);
2347         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2348         u8 encrypt_pw_salt[16];
2349         int err;
2350
2351         if (!f2fs_sb_has_encrypt(sbi))
2352                 return -EOPNOTSUPP;
2353
2354         err = mnt_want_write_file(filp);
2355         if (err)
2356                 return err;
2357
2358         f2fs_down_write(&sbi->sb_lock);
2359
2360         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2361                 goto got_it;
2362
2363         /* update superblock with uuid */
2364         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2365
2366         err = f2fs_commit_super(sbi, false);
2367         if (err) {
2368                 /* undo new data */
2369                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2370                 goto out_err;
2371         }
2372 got_it:
2373         memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2374 out_err:
2375         f2fs_up_write(&sbi->sb_lock);
2376         mnt_drop_write_file(filp);
2377
2378         if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2379                 err = -EFAULT;
2380
2381         return err;
2382 }
2383
2384 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2385                                              unsigned long arg)
2386 {
2387         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2388                 return -EOPNOTSUPP;
2389
2390         return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2391 }
2392
2393 static int f2fs_ioc_add_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_add_key(filp, (void __user *)arg);
2399 }
2400
2401 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2402 {
2403         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2404                 return -EOPNOTSUPP;
2405
2406         return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2407 }
2408
2409 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2410                                                     unsigned long arg)
2411 {
2412         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2413                 return -EOPNOTSUPP;
2414
2415         return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2416 }
2417
2418 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2419                                               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_key_status(filp, (void __user *)arg);
2425 }
2426
2427 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2428 {
2429         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2430                 return -EOPNOTSUPP;
2431
2432         return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2433 }
2434
2435 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2436 {
2437         struct inode *inode = file_inode(filp);
2438         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2439         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2440                         .no_bg_gc = false,
2441                         .should_migrate_blocks = false,
2442                         .nr_free_secs = 0 };
2443         __u32 sync;
2444         int ret;
2445
2446         if (!capable(CAP_SYS_ADMIN))
2447                 return -EPERM;
2448
2449         if (get_user(sync, (__u32 __user *)arg))
2450                 return -EFAULT;
2451
2452         if (f2fs_readonly(sbi->sb))
2453                 return -EROFS;
2454
2455         ret = mnt_want_write_file(filp);
2456         if (ret)
2457                 return ret;
2458
2459         if (!sync) {
2460                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2461                         ret = -EBUSY;
2462                         goto out;
2463                 }
2464         } else {
2465                 f2fs_down_write(&sbi->gc_lock);
2466         }
2467
2468         gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2469         gc_control.err_gc_skipped = sync;
2470         ret = f2fs_gc(sbi, &gc_control);
2471 out:
2472         mnt_drop_write_file(filp);
2473         return ret;
2474 }
2475
2476 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2477 {
2478         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2479         struct f2fs_gc_control gc_control = {
2480                         .init_gc_type = range->sync ? FG_GC : BG_GC,
2481                         .no_bg_gc = false,
2482                         .should_migrate_blocks = false,
2483                         .err_gc_skipped = range->sync,
2484                         .nr_free_secs = 0 };
2485         u64 end;
2486         int ret;
2487
2488         if (!capable(CAP_SYS_ADMIN))
2489                 return -EPERM;
2490         if (f2fs_readonly(sbi->sb))
2491                 return -EROFS;
2492
2493         end = range->start + range->len;
2494         if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2495                                         end >= MAX_BLKADDR(sbi))
2496                 return -EINVAL;
2497
2498         ret = mnt_want_write_file(filp);
2499         if (ret)
2500                 return ret;
2501
2502 do_more:
2503         if (!range->sync) {
2504                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2505                         ret = -EBUSY;
2506                         goto out;
2507                 }
2508         } else {
2509                 f2fs_down_write(&sbi->gc_lock);
2510         }
2511
2512         gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2513         ret = f2fs_gc(sbi, &gc_control);
2514         if (ret) {
2515                 if (ret == -EBUSY)
2516                         ret = -EAGAIN;
2517                 goto out;
2518         }
2519         range->start += CAP_BLKS_PER_SEC(sbi);
2520         if (range->start <= end)
2521                 goto do_more;
2522 out:
2523         mnt_drop_write_file(filp);
2524         return ret;
2525 }
2526
2527 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2528 {
2529         struct f2fs_gc_range range;
2530
2531         if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2532                                                         sizeof(range)))
2533                 return -EFAULT;
2534         return __f2fs_ioc_gc_range(filp, &range);
2535 }
2536
2537 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2538 {
2539         struct inode *inode = file_inode(filp);
2540         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2541         int ret;
2542
2543         if (!capable(CAP_SYS_ADMIN))
2544                 return -EPERM;
2545
2546         if (f2fs_readonly(sbi->sb))
2547                 return -EROFS;
2548
2549         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2550                 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2551                 return -EINVAL;
2552         }
2553
2554         ret = mnt_want_write_file(filp);
2555         if (ret)
2556                 return ret;
2557
2558         ret = f2fs_sync_fs(sbi->sb, 1);
2559
2560         mnt_drop_write_file(filp);
2561         return ret;
2562 }
2563
2564 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2565                                         struct file *filp,
2566                                         struct f2fs_defragment *range)
2567 {
2568         struct inode *inode = file_inode(filp);
2569         struct f2fs_map_blocks map = { .m_next_extent = NULL,
2570                                         .m_seg_type = NO_CHECK_TYPE,
2571                                         .m_may_create = false };
2572         struct extent_info ei = {0, 0, 0};
2573         pgoff_t pg_start, pg_end, next_pgofs;
2574         unsigned int blk_per_seg = sbi->blocks_per_seg;
2575         unsigned int total = 0, sec_num;
2576         block_t blk_end = 0;
2577         bool fragmented = false;
2578         int err;
2579
2580         pg_start = range->start >> PAGE_SHIFT;
2581         pg_end = (range->start + range->len) >> PAGE_SHIFT;
2582
2583         f2fs_balance_fs(sbi, true);
2584
2585         inode_lock(inode);
2586
2587         /* if in-place-update policy is enabled, don't waste time here */
2588         set_inode_flag(inode, FI_OPU_WRITE);
2589         if (f2fs_should_update_inplace(inode, NULL)) {
2590                 err = -EINVAL;
2591                 goto out;
2592         }
2593
2594         /* writeback all dirty pages in the range */
2595         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2596                                                 range->start + range->len - 1);
2597         if (err)
2598                 goto out;
2599
2600         /*
2601          * lookup mapping info in extent cache, skip defragmenting if physical
2602          * block addresses are continuous.
2603          */
2604         if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2605                 if (ei.fofs + ei.len >= pg_end)
2606                         goto out;
2607         }
2608
2609         map.m_lblk = pg_start;
2610         map.m_next_pgofs = &next_pgofs;
2611
2612         /*
2613          * lookup mapping info in dnode page cache, skip defragmenting if all
2614          * physical block addresses are continuous even if there are hole(s)
2615          * in logical blocks.
2616          */
2617         while (map.m_lblk < pg_end) {
2618                 map.m_len = pg_end - map.m_lblk;
2619                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2620                 if (err)
2621                         goto out;
2622
2623                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2624                         map.m_lblk = next_pgofs;
2625                         continue;
2626                 }
2627
2628                 if (blk_end && blk_end != map.m_pblk)
2629                         fragmented = true;
2630
2631                 /* record total count of block that we're going to move */
2632                 total += map.m_len;
2633
2634                 blk_end = map.m_pblk + map.m_len;
2635
2636                 map.m_lblk += map.m_len;
2637         }
2638
2639         if (!fragmented) {
2640                 total = 0;
2641                 goto out;
2642         }
2643
2644         sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2645
2646         /*
2647          * make sure there are enough free section for LFS allocation, this can
2648          * avoid defragment running in SSR mode when free section are allocated
2649          * intensively
2650          */
2651         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2652                 err = -EAGAIN;
2653                 goto out;
2654         }
2655
2656         map.m_lblk = pg_start;
2657         map.m_len = pg_end - pg_start;
2658         total = 0;
2659
2660         while (map.m_lblk < pg_end) {
2661                 pgoff_t idx;
2662                 int cnt = 0;
2663
2664 do_map:
2665                 map.m_len = pg_end - map.m_lblk;
2666                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2667                 if (err)
2668                         goto clear_out;
2669
2670                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2671                         map.m_lblk = next_pgofs;
2672                         goto check;
2673                 }
2674
2675                 set_inode_flag(inode, FI_SKIP_WRITES);
2676
2677                 idx = map.m_lblk;
2678                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2679                         struct page *page;
2680
2681                         page = f2fs_get_lock_data_page(inode, idx, true);
2682                         if (IS_ERR(page)) {
2683                                 err = PTR_ERR(page);
2684                                 goto clear_out;
2685                         }
2686
2687                         set_page_dirty(page);
2688                         set_page_private_gcing(page);
2689                         f2fs_put_page(page, 1);
2690
2691                         idx++;
2692                         cnt++;
2693                         total++;
2694                 }
2695
2696                 map.m_lblk = idx;
2697 check:
2698                 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2699                         goto do_map;
2700
2701                 clear_inode_flag(inode, FI_SKIP_WRITES);
2702
2703                 err = filemap_fdatawrite(inode->i_mapping);
2704                 if (err)
2705                         goto out;
2706         }
2707 clear_out:
2708         clear_inode_flag(inode, FI_SKIP_WRITES);
2709 out:
2710         clear_inode_flag(inode, FI_OPU_WRITE);
2711         inode_unlock(inode);
2712         if (!err)
2713                 range->len = (u64)total << PAGE_SHIFT;
2714         return err;
2715 }
2716
2717 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2718 {
2719         struct inode *inode = file_inode(filp);
2720         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2721         struct f2fs_defragment range;
2722         int err;
2723
2724         if (!capable(CAP_SYS_ADMIN))
2725                 return -EPERM;
2726
2727         if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2728                 return -EINVAL;
2729
2730         if (f2fs_readonly(sbi->sb))
2731                 return -EROFS;
2732
2733         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2734                                                         sizeof(range)))
2735                 return -EFAULT;
2736
2737         /* verify alignment of offset & size */
2738         if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2739                 return -EINVAL;
2740
2741         if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2742                                         max_file_blocks(inode)))
2743                 return -EINVAL;
2744
2745         err = mnt_want_write_file(filp);
2746         if (err)
2747                 return err;
2748
2749         err = f2fs_defragment_range(sbi, filp, &range);
2750         mnt_drop_write_file(filp);
2751
2752         f2fs_update_time(sbi, REQ_TIME);
2753         if (err < 0)
2754                 return err;
2755
2756         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2757                                                         sizeof(range)))
2758                 return -EFAULT;
2759
2760         return 0;
2761 }
2762
2763 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2764                         struct file *file_out, loff_t pos_out, size_t len)
2765 {
2766         struct inode *src = file_inode(file_in);
2767         struct inode *dst = file_inode(file_out);
2768         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2769         size_t olen = len, dst_max_i_size = 0;
2770         size_t dst_osize;
2771         int ret;
2772
2773         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2774                                 src->i_sb != dst->i_sb)
2775                 return -EXDEV;
2776
2777         if (unlikely(f2fs_readonly(src->i_sb)))
2778                 return -EROFS;
2779
2780         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2781                 return -EINVAL;
2782
2783         if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2784                 return -EOPNOTSUPP;
2785
2786         if (pos_out < 0 || pos_in < 0)
2787                 return -EINVAL;
2788
2789         if (src == dst) {
2790                 if (pos_in == pos_out)
2791                         return 0;
2792                 if (pos_out > pos_in && pos_out < pos_in + len)
2793                         return -EINVAL;
2794         }
2795
2796         inode_lock(src);
2797         if (src != dst) {
2798                 ret = -EBUSY;
2799                 if (!inode_trylock(dst))
2800                         goto out;
2801         }
2802
2803         ret = -EINVAL;
2804         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2805                 goto out_unlock;
2806         if (len == 0)
2807                 olen = len = src->i_size - pos_in;
2808         if (pos_in + len == src->i_size)
2809                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2810         if (len == 0) {
2811                 ret = 0;
2812                 goto out_unlock;
2813         }
2814
2815         dst_osize = dst->i_size;
2816         if (pos_out + olen > dst->i_size)
2817                 dst_max_i_size = pos_out + olen;
2818
2819         /* verify the end result is block aligned */
2820         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2821                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2822                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2823                 goto out_unlock;
2824
2825         ret = f2fs_convert_inline_inode(src);
2826         if (ret)
2827                 goto out_unlock;
2828
2829         ret = f2fs_convert_inline_inode(dst);
2830         if (ret)
2831                 goto out_unlock;
2832
2833         /* write out all dirty pages from offset */
2834         ret = filemap_write_and_wait_range(src->i_mapping,
2835                                         pos_in, pos_in + len);
2836         if (ret)
2837                 goto out_unlock;
2838
2839         ret = filemap_write_and_wait_range(dst->i_mapping,
2840                                         pos_out, pos_out + len);
2841         if (ret)
2842                 goto out_unlock;
2843
2844         f2fs_balance_fs(sbi, true);
2845
2846         f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2847         if (src != dst) {
2848                 ret = -EBUSY;
2849                 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2850                         goto out_src;
2851         }
2852
2853         f2fs_lock_op(sbi);
2854         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2855                                 pos_out >> F2FS_BLKSIZE_BITS,
2856                                 len >> F2FS_BLKSIZE_BITS, false);
2857
2858         if (!ret) {
2859                 if (dst_max_i_size)
2860                         f2fs_i_size_write(dst, dst_max_i_size);
2861                 else if (dst_osize != dst->i_size)
2862                         f2fs_i_size_write(dst, dst_osize);
2863         }
2864         f2fs_unlock_op(sbi);
2865
2866         if (src != dst)
2867                 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2868 out_src:
2869         f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2870 out_unlock:
2871         if (src != dst)
2872                 inode_unlock(dst);
2873 out:
2874         inode_unlock(src);
2875         return ret;
2876 }
2877
2878 static int __f2fs_ioc_move_range(struct file *filp,
2879                                 struct f2fs_move_range *range)
2880 {
2881         struct fd dst;
2882         int err;
2883
2884         if (!(filp->f_mode & FMODE_READ) ||
2885                         !(filp->f_mode & FMODE_WRITE))
2886                 return -EBADF;
2887
2888         dst = fdget(range->dst_fd);
2889         if (!dst.file)
2890                 return -EBADF;
2891
2892         if (!(dst.file->f_mode & FMODE_WRITE)) {
2893                 err = -EBADF;
2894                 goto err_out;
2895         }
2896
2897         err = mnt_want_write_file(filp);
2898         if (err)
2899                 goto err_out;
2900
2901         err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2902                                         range->pos_out, range->len);
2903
2904         mnt_drop_write_file(filp);
2905 err_out:
2906         fdput(dst);
2907         return err;
2908 }
2909
2910 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2911 {
2912         struct f2fs_move_range range;
2913
2914         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2915                                                         sizeof(range)))
2916                 return -EFAULT;
2917         return __f2fs_ioc_move_range(filp, &range);
2918 }
2919
2920 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2921 {
2922         struct inode *inode = file_inode(filp);
2923         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2924         struct sit_info *sm = SIT_I(sbi);
2925         unsigned int start_segno = 0, end_segno = 0;
2926         unsigned int dev_start_segno = 0, dev_end_segno = 0;
2927         struct f2fs_flush_device range;
2928         struct f2fs_gc_control gc_control = {
2929                         .init_gc_type = FG_GC,
2930                         .should_migrate_blocks = true,
2931                         .err_gc_skipped = true,
2932                         .nr_free_secs = 0 };
2933         int ret;
2934
2935         if (!capable(CAP_SYS_ADMIN))
2936                 return -EPERM;
2937
2938         if (f2fs_readonly(sbi->sb))
2939                 return -EROFS;
2940
2941         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2942                 return -EINVAL;
2943
2944         if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2945                                                         sizeof(range)))
2946                 return -EFAULT;
2947
2948         if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2949                         __is_large_section(sbi)) {
2950                 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2951                           range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2952                 return -EINVAL;
2953         }
2954
2955         ret = mnt_want_write_file(filp);
2956         if (ret)
2957                 return ret;
2958
2959         if (range.dev_num != 0)
2960                 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2961         dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2962
2963         start_segno = sm->last_victim[FLUSH_DEVICE];
2964         if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2965                 start_segno = dev_start_segno;
2966         end_segno = min(start_segno + range.segments, dev_end_segno);
2967
2968         while (start_segno < end_segno) {
2969                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2970                         ret = -EBUSY;
2971                         goto out;
2972                 }
2973                 sm->last_victim[GC_CB] = end_segno + 1;
2974                 sm->last_victim[GC_GREEDY] = end_segno + 1;
2975                 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2976
2977                 gc_control.victim_segno = start_segno;
2978                 ret = f2fs_gc(sbi, &gc_control);
2979                 if (ret == -EAGAIN)
2980                         ret = 0;
2981                 else if (ret < 0)
2982                         break;
2983                 start_segno++;
2984         }
2985 out:
2986         mnt_drop_write_file(filp);
2987         return ret;
2988 }
2989
2990 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2991 {
2992         struct inode *inode = file_inode(filp);
2993         u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2994
2995         /* Must validate to set it with SQLite behavior in Android. */
2996         sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2997
2998         return put_user(sb_feature, (u32 __user *)arg);
2999 }
3000
3001 #ifdef CONFIG_QUOTA
3002 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3003 {
3004         struct dquot *transfer_to[MAXQUOTAS] = {};
3005         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3006         struct super_block *sb = sbi->sb;
3007         int err;
3008
3009         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3010         if (IS_ERR(transfer_to[PRJQUOTA]))
3011                 return PTR_ERR(transfer_to[PRJQUOTA]);
3012
3013         err = __dquot_transfer(inode, transfer_to);
3014         if (err)
3015                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3016         dqput(transfer_to[PRJQUOTA]);
3017         return err;
3018 }
3019
3020 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3021 {
3022         struct f2fs_inode_info *fi = F2FS_I(inode);
3023         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3024         struct f2fs_inode *ri = NULL;
3025         kprojid_t kprojid;
3026         int err;
3027
3028         if (!f2fs_sb_has_project_quota(sbi)) {
3029                 if (projid != F2FS_DEF_PROJID)
3030                         return -EOPNOTSUPP;
3031                 else
3032                         return 0;
3033         }
3034
3035         if (!f2fs_has_extra_attr(inode))
3036                 return -EOPNOTSUPP;
3037
3038         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3039
3040         if (projid_eq(kprojid, fi->i_projid))
3041                 return 0;
3042
3043         err = -EPERM;
3044         /* Is it quota file? Do not allow user to mess with it */
3045         if (IS_NOQUOTA(inode))
3046                 return err;
3047
3048         if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3049                 return -EOVERFLOW;
3050
3051         err = f2fs_dquot_initialize(inode);
3052         if (err)
3053                 return err;
3054
3055         f2fs_lock_op(sbi);
3056         err = f2fs_transfer_project_quota(inode, kprojid);
3057         if (err)
3058                 goto out_unlock;
3059
3060         fi->i_projid = kprojid;
3061         inode->i_ctime = current_time(inode);
3062         f2fs_mark_inode_dirty_sync(inode, true);
3063 out_unlock:
3064         f2fs_unlock_op(sbi);
3065         return err;
3066 }
3067 #else
3068 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3069 {
3070         return 0;
3071 }
3072
3073 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3074 {
3075         if (projid != F2FS_DEF_PROJID)
3076                 return -EOPNOTSUPP;
3077         return 0;
3078 }
3079 #endif
3080
3081 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3082 {
3083         struct inode *inode = d_inode(dentry);
3084         struct f2fs_inode_info *fi = F2FS_I(inode);
3085         u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3086
3087         if (IS_ENCRYPTED(inode))
3088                 fsflags |= FS_ENCRYPT_FL;
3089         if (IS_VERITY(inode))
3090                 fsflags |= FS_VERITY_FL;
3091         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3092                 fsflags |= FS_INLINE_DATA_FL;
3093         if (is_inode_flag_set(inode, FI_PIN_FILE))
3094                 fsflags |= FS_NOCOW_FL;
3095
3096         fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3097
3098         if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3099                 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3100
3101         return 0;
3102 }
3103
3104 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3105                       struct dentry *dentry, struct fileattr *fa)
3106 {
3107         struct inode *inode = d_inode(dentry);
3108         u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3109         u32 iflags;
3110         int err;
3111
3112         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3113                 return -EIO;
3114         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3115                 return -ENOSPC;
3116         if (fsflags & ~F2FS_GETTABLE_FS_FL)
3117                 return -EOPNOTSUPP;
3118         fsflags &= F2FS_SETTABLE_FS_FL;
3119         if (!fa->flags_valid)
3120                 mask &= FS_COMMON_FL;
3121
3122         iflags = f2fs_fsflags_to_iflags(fsflags);
3123         if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3124                 return -EOPNOTSUPP;
3125
3126         err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3127         if (!err)
3128                 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3129
3130         return err;
3131 }
3132
3133 int f2fs_pin_file_control(struct inode *inode, bool inc)
3134 {
3135         struct f2fs_inode_info *fi = F2FS_I(inode);
3136         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3137
3138         /* Use i_gc_failures for normal file as a risk signal. */
3139         if (inc)
3140                 f2fs_i_gc_failures_write(inode,
3141                                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3142
3143         if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3144                 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3145                           __func__, inode->i_ino,
3146                           fi->i_gc_failures[GC_FAILURE_PIN]);
3147                 clear_inode_flag(inode, FI_PIN_FILE);
3148                 return -EAGAIN;
3149         }
3150         return 0;
3151 }
3152
3153 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3154 {
3155         struct inode *inode = file_inode(filp);
3156         __u32 pin;
3157         int ret = 0;
3158
3159         if (get_user(pin, (__u32 __user *)arg))
3160                 return -EFAULT;
3161
3162         if (!S_ISREG(inode->i_mode))
3163                 return -EINVAL;
3164
3165         if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3166                 return -EROFS;
3167
3168         ret = mnt_want_write_file(filp);
3169         if (ret)
3170                 return ret;
3171
3172         inode_lock(inode);
3173
3174         if (!pin) {
3175                 clear_inode_flag(inode, FI_PIN_FILE);
3176                 f2fs_i_gc_failures_write(inode, 0);
3177                 goto done;
3178         }
3179
3180         if (f2fs_should_update_outplace(inode, NULL)) {
3181                 ret = -EINVAL;
3182                 goto out;
3183         }
3184
3185         if (f2fs_pin_file_control(inode, false)) {
3186                 ret = -EAGAIN;
3187                 goto out;
3188         }
3189
3190         ret = f2fs_convert_inline_inode(inode);
3191         if (ret)
3192                 goto out;
3193
3194         if (!f2fs_disable_compressed_file(inode)) {
3195                 ret = -EOPNOTSUPP;
3196                 goto out;
3197         }
3198
3199         set_inode_flag(inode, FI_PIN_FILE);
3200         ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3201 done:
3202         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3203 out:
3204         inode_unlock(inode);
3205         mnt_drop_write_file(filp);
3206         return ret;
3207 }
3208
3209 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3210 {
3211         struct inode *inode = file_inode(filp);
3212         __u32 pin = 0;
3213
3214         if (is_inode_flag_set(inode, FI_PIN_FILE))
3215                 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3216         return put_user(pin, (u32 __user *)arg);
3217 }
3218
3219 int f2fs_precache_extents(struct inode *inode)
3220 {
3221         struct f2fs_inode_info *fi = F2FS_I(inode);
3222         struct f2fs_map_blocks map;
3223         pgoff_t m_next_extent;
3224         loff_t end;
3225         int err;
3226
3227         if (is_inode_flag_set(inode, FI_NO_EXTENT))
3228                 return -EOPNOTSUPP;
3229
3230         map.m_lblk = 0;
3231         map.m_next_pgofs = NULL;
3232         map.m_next_extent = &m_next_extent;
3233         map.m_seg_type = NO_CHECK_TYPE;
3234         map.m_may_create = false;
3235         end = max_file_blocks(inode);
3236
3237         while (map.m_lblk < end) {
3238                 map.m_len = end - map.m_lblk;
3239
3240                 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3241                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3242                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3243                 if (err)
3244                         return err;
3245
3246                 map.m_lblk = m_next_extent;
3247         }
3248
3249         return 0;
3250 }
3251
3252 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3253 {
3254         return f2fs_precache_extents(file_inode(filp));
3255 }
3256
3257 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3258 {
3259         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3260         __u64 block_count;
3261
3262         if (!capable(CAP_SYS_ADMIN))
3263                 return -EPERM;
3264
3265         if (f2fs_readonly(sbi->sb))
3266                 return -EROFS;
3267
3268         if (copy_from_user(&block_count, (void __user *)arg,
3269                            sizeof(block_count)))
3270                 return -EFAULT;
3271
3272         return f2fs_resize_fs(sbi, block_count);
3273 }
3274
3275 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3276 {
3277         struct inode *inode = file_inode(filp);
3278
3279         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3280
3281         if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3282                 f2fs_warn(F2FS_I_SB(inode),
3283                           "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3284                           inode->i_ino);
3285                 return -EOPNOTSUPP;
3286         }
3287
3288         return fsverity_ioctl_enable(filp, (const void __user *)arg);
3289 }
3290
3291 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3292 {
3293         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3294                 return -EOPNOTSUPP;
3295
3296         return fsverity_ioctl_measure(filp, (void __user *)arg);
3297 }
3298
3299 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3300 {
3301         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3302                 return -EOPNOTSUPP;
3303
3304         return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3305 }
3306
3307 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3308 {
3309         struct inode *inode = file_inode(filp);
3310         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3311         char *vbuf;
3312         int count;
3313         int err = 0;
3314
3315         vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3316         if (!vbuf)
3317                 return -ENOMEM;
3318
3319         f2fs_down_read(&sbi->sb_lock);
3320         count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3321                         ARRAY_SIZE(sbi->raw_super->volume_name),
3322                         UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3323         f2fs_up_read(&sbi->sb_lock);
3324
3325         if (copy_to_user((char __user *)arg, vbuf,
3326                                 min(FSLABEL_MAX, count)))
3327                 err = -EFAULT;
3328
3329         kfree(vbuf);
3330         return err;
3331 }
3332
3333 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3334 {
3335         struct inode *inode = file_inode(filp);
3336         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3337         char *vbuf;
3338         int err = 0;
3339
3340         if (!capable(CAP_SYS_ADMIN))
3341                 return -EPERM;
3342
3343         vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3344         if (IS_ERR(vbuf))
3345                 return PTR_ERR(vbuf);
3346
3347         err = mnt_want_write_file(filp);
3348         if (err)
3349                 goto out;
3350
3351         f2fs_down_write(&sbi->sb_lock);
3352
3353         memset(sbi->raw_super->volume_name, 0,
3354                         sizeof(sbi->raw_super->volume_name));
3355         utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3356                         sbi->raw_super->volume_name,
3357                         ARRAY_SIZE(sbi->raw_super->volume_name));
3358
3359         err = f2fs_commit_super(sbi, false);
3360
3361         f2fs_up_write(&sbi->sb_lock);
3362
3363         mnt_drop_write_file(filp);
3364 out:
3365         kfree(vbuf);
3366         return err;
3367 }
3368
3369 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3370 {
3371         struct inode *inode = file_inode(filp);
3372         __u64 blocks;
3373
3374         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3375                 return -EOPNOTSUPP;
3376
3377         if (!f2fs_compressed_file(inode))
3378                 return -EINVAL;
3379
3380         blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3381         return put_user(blocks, (u64 __user *)arg);
3382 }
3383
3384 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3385 {
3386         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3387         unsigned int released_blocks = 0;
3388         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3389         block_t blkaddr;
3390         int i;
3391
3392         for (i = 0; i < count; i++) {
3393                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3394                                                 dn->ofs_in_node + i);
3395
3396                 if (!__is_valid_data_blkaddr(blkaddr))
3397                         continue;
3398                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3399                                         DATA_GENERIC_ENHANCE))) {
3400                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3401                         return -EFSCORRUPTED;
3402                 }
3403         }
3404
3405         while (count) {
3406                 int compr_blocks = 0;
3407
3408                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3409                         blkaddr = f2fs_data_blkaddr(dn);
3410
3411                         if (i == 0) {
3412                                 if (blkaddr == COMPRESS_ADDR)
3413                                         continue;
3414                                 dn->ofs_in_node += cluster_size;
3415                                 goto next;
3416                         }
3417
3418                         if (__is_valid_data_blkaddr(blkaddr))
3419                                 compr_blocks++;
3420
3421                         if (blkaddr != NEW_ADDR)
3422                                 continue;
3423
3424                         dn->data_blkaddr = NULL_ADDR;
3425                         f2fs_set_data_blkaddr(dn);
3426                 }
3427
3428                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3429                 dec_valid_block_count(sbi, dn->inode,
3430                                         cluster_size - compr_blocks);
3431
3432                 released_blocks += cluster_size - compr_blocks;
3433 next:
3434                 count -= cluster_size;
3435         }
3436
3437         return released_blocks;
3438 }
3439
3440 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3441 {
3442         struct inode *inode = file_inode(filp);
3443         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3444         pgoff_t page_idx = 0, last_idx;
3445         unsigned int released_blocks = 0;
3446         int ret;
3447         int writecount;
3448
3449         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3450                 return -EOPNOTSUPP;
3451
3452         if (!f2fs_compressed_file(inode))
3453                 return -EINVAL;
3454
3455         if (f2fs_readonly(sbi->sb))
3456                 return -EROFS;
3457
3458         ret = mnt_want_write_file(filp);
3459         if (ret)
3460                 return ret;
3461
3462         f2fs_balance_fs(F2FS_I_SB(inode), true);
3463
3464         inode_lock(inode);
3465
3466         writecount = atomic_read(&inode->i_writecount);
3467         if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3468                         (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3469                 ret = -EBUSY;
3470                 goto out;
3471         }
3472
3473         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3474                 ret = -EINVAL;
3475                 goto out;
3476         }
3477
3478         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3479         if (ret)
3480                 goto out;
3481
3482         set_inode_flag(inode, FI_COMPRESS_RELEASED);
3483         inode->i_ctime = current_time(inode);
3484         f2fs_mark_inode_dirty_sync(inode, true);
3485
3486         if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3487                 goto out;
3488
3489         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3490         filemap_invalidate_lock(inode->i_mapping);
3491
3492         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3493
3494         while (page_idx < last_idx) {
3495                 struct dnode_of_data dn;
3496                 pgoff_t end_offset, count;
3497
3498                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3499                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3500                 if (ret) {
3501                         if (ret == -ENOENT) {
3502                                 page_idx = f2fs_get_next_page_offset(&dn,
3503                                                                 page_idx);
3504                                 ret = 0;
3505                                 continue;
3506                         }
3507                         break;
3508                 }
3509
3510                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3511                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3512                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3513
3514                 ret = release_compress_blocks(&dn, count);
3515
3516                 f2fs_put_dnode(&dn);
3517
3518                 if (ret < 0)
3519                         break;
3520
3521                 page_idx += count;
3522                 released_blocks += ret;
3523         }
3524
3525         filemap_invalidate_unlock(inode->i_mapping);
3526         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3527 out:
3528         inode_unlock(inode);
3529
3530         mnt_drop_write_file(filp);
3531
3532         if (ret >= 0) {
3533                 ret = put_user(released_blocks, (u64 __user *)arg);
3534         } else if (released_blocks &&
3535                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3536                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3537                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3538                         "iblocks=%llu, released=%u, compr_blocks=%u, "
3539                         "run fsck to fix.",
3540                         __func__, inode->i_ino, inode->i_blocks,
3541                         released_blocks,
3542                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3543         }
3544
3545         return ret;
3546 }
3547
3548 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3549 {
3550         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3551         unsigned int reserved_blocks = 0;
3552         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3553         block_t blkaddr;
3554         int i;
3555
3556         for (i = 0; i < count; i++) {
3557                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3558                                                 dn->ofs_in_node + i);
3559
3560                 if (!__is_valid_data_blkaddr(blkaddr))
3561                         continue;
3562                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3563                                         DATA_GENERIC_ENHANCE))) {
3564                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3565                         return -EFSCORRUPTED;
3566                 }
3567         }
3568
3569         while (count) {
3570                 int compr_blocks = 0;
3571                 blkcnt_t reserved;
3572                 int ret;
3573
3574                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3575                         blkaddr = f2fs_data_blkaddr(dn);
3576
3577                         if (i == 0) {
3578                                 if (blkaddr == COMPRESS_ADDR)
3579                                         continue;
3580                                 dn->ofs_in_node += cluster_size;
3581                                 goto next;
3582                         }
3583
3584                         if (__is_valid_data_blkaddr(blkaddr)) {
3585                                 compr_blocks++;
3586                                 continue;
3587                         }
3588
3589                         dn->data_blkaddr = NEW_ADDR;
3590                         f2fs_set_data_blkaddr(dn);
3591                 }
3592
3593                 reserved = cluster_size - compr_blocks;
3594                 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3595                 if (ret)
3596                         return ret;
3597
3598                 if (reserved != cluster_size - compr_blocks)
3599                         return -ENOSPC;
3600
3601                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3602
3603                 reserved_blocks += reserved;
3604 next:
3605                 count -= cluster_size;
3606         }
3607
3608         return reserved_blocks;
3609 }
3610
3611 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3612 {
3613         struct inode *inode = file_inode(filp);
3614         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3615         pgoff_t page_idx = 0, last_idx;
3616         unsigned int reserved_blocks = 0;
3617         int ret;
3618
3619         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3620                 return -EOPNOTSUPP;
3621
3622         if (!f2fs_compressed_file(inode))
3623                 return -EINVAL;
3624
3625         if (f2fs_readonly(sbi->sb))
3626                 return -EROFS;
3627
3628         ret = mnt_want_write_file(filp);
3629         if (ret)
3630                 return ret;
3631
3632         if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3633                 goto out;
3634
3635         f2fs_balance_fs(F2FS_I_SB(inode), true);
3636
3637         inode_lock(inode);
3638
3639         if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3640                 ret = -EINVAL;
3641                 goto unlock_inode;
3642         }
3643
3644         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3645         filemap_invalidate_lock(inode->i_mapping);
3646
3647         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3648
3649         while (page_idx < last_idx) {
3650                 struct dnode_of_data dn;
3651                 pgoff_t end_offset, count;
3652
3653                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3654                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3655                 if (ret) {
3656                         if (ret == -ENOENT) {
3657                                 page_idx = f2fs_get_next_page_offset(&dn,
3658                                                                 page_idx);
3659                                 ret = 0;
3660                                 continue;
3661                         }
3662                         break;
3663                 }
3664
3665                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3666                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3667                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3668
3669                 ret = reserve_compress_blocks(&dn, count);
3670
3671                 f2fs_put_dnode(&dn);
3672
3673                 if (ret < 0)
3674                         break;
3675
3676                 page_idx += count;
3677                 reserved_blocks += ret;
3678         }
3679
3680         filemap_invalidate_unlock(inode->i_mapping);
3681         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3682
3683         if (ret >= 0) {
3684                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3685                 inode->i_ctime = current_time(inode);
3686                 f2fs_mark_inode_dirty_sync(inode, true);
3687         }
3688 unlock_inode:
3689         inode_unlock(inode);
3690 out:
3691         mnt_drop_write_file(filp);
3692
3693         if (ret >= 0) {
3694                 ret = put_user(reserved_blocks, (u64 __user *)arg);
3695         } else if (reserved_blocks &&
3696                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3697                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3698                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3699                         "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3700                         "run fsck to fix.",
3701                         __func__, inode->i_ino, inode->i_blocks,
3702                         reserved_blocks,
3703                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3704         }
3705
3706         return ret;
3707 }
3708
3709 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3710                 pgoff_t off, block_t block, block_t len, u32 flags)
3711 {
3712         sector_t sector = SECTOR_FROM_BLOCK(block);
3713         sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3714         int ret = 0;
3715
3716         if (flags & F2FS_TRIM_FILE_DISCARD) {
3717                 if (bdev_max_secure_erase_sectors(bdev))
3718                         ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3719                                         GFP_NOFS);
3720                 else
3721                         ret = blkdev_issue_discard(bdev, sector, nr_sects,
3722                                         GFP_NOFS);
3723         }
3724
3725         if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3726                 if (IS_ENCRYPTED(inode))
3727                         ret = fscrypt_zeroout_range(inode, off, block, len);
3728                 else
3729                         ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3730                                         GFP_NOFS, 0);
3731         }
3732
3733         return ret;
3734 }
3735
3736 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3737 {
3738         struct inode *inode = file_inode(filp);
3739         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3740         struct address_space *mapping = inode->i_mapping;
3741         struct block_device *prev_bdev = NULL;
3742         struct f2fs_sectrim_range range;
3743         pgoff_t index, pg_end, prev_index = 0;
3744         block_t prev_block = 0, len = 0;
3745         loff_t end_addr;
3746         bool to_end = false;
3747         int ret = 0;
3748
3749         if (!(filp->f_mode & FMODE_WRITE))
3750                 return -EBADF;
3751
3752         if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3753                                 sizeof(range)))
3754                 return -EFAULT;
3755
3756         if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3757                         !S_ISREG(inode->i_mode))
3758                 return -EINVAL;
3759
3760         if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3761                         !f2fs_hw_support_discard(sbi)) ||
3762                         ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3763                          IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3764                 return -EOPNOTSUPP;
3765
3766         file_start_write(filp);
3767         inode_lock(inode);
3768
3769         if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3770                         range.start >= inode->i_size) {
3771                 ret = -EINVAL;
3772                 goto err;
3773         }
3774
3775         if (range.len == 0)
3776                 goto err;
3777
3778         if (inode->i_size - range.start > range.len) {
3779                 end_addr = range.start + range.len;
3780         } else {
3781                 end_addr = range.len == (u64)-1 ?
3782                         sbi->sb->s_maxbytes : inode->i_size;
3783                 to_end = true;
3784         }
3785
3786         if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3787                         (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3788                 ret = -EINVAL;
3789                 goto err;
3790         }
3791
3792         index = F2FS_BYTES_TO_BLK(range.start);
3793         pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3794
3795         ret = f2fs_convert_inline_inode(inode);
3796         if (ret)
3797                 goto err;
3798
3799         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3800         filemap_invalidate_lock(mapping);
3801
3802         ret = filemap_write_and_wait_range(mapping, range.start,
3803                         to_end ? LLONG_MAX : end_addr - 1);
3804         if (ret)
3805                 goto out;
3806
3807         truncate_inode_pages_range(mapping, range.start,
3808                         to_end ? -1 : end_addr - 1);
3809
3810         while (index < pg_end) {
3811                 struct dnode_of_data dn;
3812                 pgoff_t end_offset, count;
3813                 int i;
3814
3815                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3816                 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3817                 if (ret) {
3818                         if (ret == -ENOENT) {
3819                                 index = f2fs_get_next_page_offset(&dn, index);
3820                                 continue;
3821                         }
3822                         goto out;
3823                 }
3824
3825                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3826                 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3827                 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3828                         struct block_device *cur_bdev;
3829                         block_t blkaddr = f2fs_data_blkaddr(&dn);
3830
3831                         if (!__is_valid_data_blkaddr(blkaddr))
3832                                 continue;
3833
3834                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3835                                                 DATA_GENERIC_ENHANCE)) {
3836                                 ret = -EFSCORRUPTED;
3837                                 f2fs_put_dnode(&dn);
3838                                 f2fs_handle_error(sbi,
3839                                                 ERROR_INVALID_BLKADDR);
3840                                 goto out;
3841                         }
3842
3843                         cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3844                         if (f2fs_is_multi_device(sbi)) {
3845                                 int di = f2fs_target_device_index(sbi, blkaddr);
3846
3847                                 blkaddr -= FDEV(di).start_blk;
3848                         }
3849
3850                         if (len) {
3851                                 if (prev_bdev == cur_bdev &&
3852                                                 index == prev_index + len &&
3853                                                 blkaddr == prev_block + len) {
3854                                         len++;
3855                                 } else {
3856                                         ret = f2fs_secure_erase(prev_bdev,
3857                                                 inode, prev_index, prev_block,
3858                                                 len, range.flags);
3859                                         if (ret) {
3860                                                 f2fs_put_dnode(&dn);
3861                                                 goto out;
3862                                         }
3863
3864                                         len = 0;
3865                                 }
3866                         }
3867
3868                         if (!len) {
3869                                 prev_bdev = cur_bdev;
3870                                 prev_index = index;
3871                                 prev_block = blkaddr;
3872                                 len = 1;
3873                         }
3874                 }
3875
3876                 f2fs_put_dnode(&dn);
3877
3878                 if (fatal_signal_pending(current)) {
3879                         ret = -EINTR;
3880                         goto out;
3881                 }
3882                 cond_resched();
3883         }
3884
3885         if (len)
3886                 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3887                                 prev_block, len, range.flags);
3888 out:
3889         filemap_invalidate_unlock(mapping);
3890         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3891 err:
3892         inode_unlock(inode);
3893         file_end_write(filp);
3894
3895         return ret;
3896 }
3897
3898 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3899 {
3900         struct inode *inode = file_inode(filp);
3901         struct f2fs_comp_option option;
3902
3903         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3904                 return -EOPNOTSUPP;
3905
3906         inode_lock_shared(inode);
3907
3908         if (!f2fs_compressed_file(inode)) {
3909                 inode_unlock_shared(inode);
3910                 return -ENODATA;
3911         }
3912
3913         option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3914         option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3915
3916         inode_unlock_shared(inode);
3917
3918         if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3919                                 sizeof(option)))
3920                 return -EFAULT;
3921
3922         return 0;
3923 }
3924
3925 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3926 {
3927         struct inode *inode = file_inode(filp);
3928         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3929         struct f2fs_comp_option option;
3930         int ret = 0;
3931
3932         if (!f2fs_sb_has_compression(sbi))
3933                 return -EOPNOTSUPP;
3934
3935         if (!(filp->f_mode & FMODE_WRITE))
3936                 return -EBADF;
3937
3938         if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3939                                 sizeof(option)))
3940                 return -EFAULT;
3941
3942         if (!f2fs_compressed_file(inode) ||
3943                         option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3944                         option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3945                         option.algorithm >= COMPRESS_MAX)
3946                 return -EINVAL;
3947
3948         file_start_write(filp);
3949         inode_lock(inode);
3950
3951         if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3952                 ret = -EBUSY;
3953                 goto out;
3954         }
3955
3956         if (F2FS_HAS_BLOCKS(inode)) {
3957                 ret = -EFBIG;
3958                 goto out;
3959         }
3960
3961         F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3962         F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3963         F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3964         f2fs_mark_inode_dirty_sync(inode, true);
3965
3966         if (!f2fs_is_compress_backend_ready(inode))
3967                 f2fs_warn(sbi, "compression algorithm is successfully set, "
3968                         "but current kernel doesn't support this algorithm.");
3969 out:
3970         inode_unlock(inode);
3971         file_end_write(filp);
3972
3973         return ret;
3974 }
3975
3976 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3977 {
3978         DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3979         struct address_space *mapping = inode->i_mapping;
3980         struct page *page;
3981         pgoff_t redirty_idx = page_idx;
3982         int i, page_len = 0, ret = 0;
3983
3984         page_cache_ra_unbounded(&ractl, len, 0);
3985
3986         for (i = 0; i < len; i++, page_idx++) {
3987                 page = read_cache_page(mapping, page_idx, NULL, NULL);
3988                 if (IS_ERR(page)) {
3989                         ret = PTR_ERR(page);
3990                         break;
3991                 }
3992                 page_len++;
3993         }
3994
3995         for (i = 0; i < page_len; i++, redirty_idx++) {
3996                 page = find_lock_page(mapping, redirty_idx);
3997
3998                 /* It will never fail, when page has pinned above */
3999                 f2fs_bug_on(F2FS_I_SB(inode), !page);
4000
4001                 set_page_dirty(page);
4002                 f2fs_put_page(page, 1);
4003                 f2fs_put_page(page, 0);
4004         }
4005
4006         return ret;
4007 }
4008
4009 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
4010 {
4011         struct inode *inode = file_inode(filp);
4012         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4013         struct f2fs_inode_info *fi = F2FS_I(inode);
4014         pgoff_t page_idx = 0, last_idx;
4015         unsigned int blk_per_seg = sbi->blocks_per_seg;
4016         int cluster_size = fi->i_cluster_size;
4017         int count, ret;
4018
4019         if (!f2fs_sb_has_compression(sbi) ||
4020                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4021                 return -EOPNOTSUPP;
4022
4023         if (!(filp->f_mode & FMODE_WRITE))
4024                 return -EBADF;
4025
4026         if (!f2fs_compressed_file(inode))
4027                 return -EINVAL;
4028
4029         f2fs_balance_fs(F2FS_I_SB(inode), true);
4030
4031         file_start_write(filp);
4032         inode_lock(inode);
4033
4034         if (!f2fs_is_compress_backend_ready(inode)) {
4035                 ret = -EOPNOTSUPP;
4036                 goto out;
4037         }
4038
4039         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4040                 ret = -EINVAL;
4041                 goto out;
4042         }
4043
4044         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4045         if (ret)
4046                 goto out;
4047
4048         if (!atomic_read(&fi->i_compr_blocks))
4049                 goto out;
4050
4051         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4052
4053         count = last_idx - page_idx;
4054         while (count) {
4055                 int len = min(cluster_size, count);
4056
4057                 ret = redirty_blocks(inode, page_idx, len);
4058                 if (ret < 0)
4059                         break;
4060
4061                 if (get_dirty_pages(inode) >= blk_per_seg)
4062                         filemap_fdatawrite(inode->i_mapping);
4063
4064                 count -= len;
4065                 page_idx += len;
4066         }
4067
4068         if (!ret)
4069                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4070                                                         LLONG_MAX);
4071
4072         if (ret)
4073                 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4074                           __func__, ret);
4075 out:
4076         inode_unlock(inode);
4077         file_end_write(filp);
4078
4079         return ret;
4080 }
4081
4082 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4083 {
4084         struct inode *inode = file_inode(filp);
4085         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4086         pgoff_t page_idx = 0, last_idx;
4087         unsigned int blk_per_seg = sbi->blocks_per_seg;
4088         int cluster_size = F2FS_I(inode)->i_cluster_size;
4089         int count, ret;
4090
4091         if (!f2fs_sb_has_compression(sbi) ||
4092                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4093                 return -EOPNOTSUPP;
4094
4095         if (!(filp->f_mode & FMODE_WRITE))
4096                 return -EBADF;
4097
4098         if (!f2fs_compressed_file(inode))
4099                 return -EINVAL;
4100
4101         f2fs_balance_fs(F2FS_I_SB(inode), true);
4102
4103         file_start_write(filp);
4104         inode_lock(inode);
4105
4106         if (!f2fs_is_compress_backend_ready(inode)) {
4107                 ret = -EOPNOTSUPP;
4108                 goto out;
4109         }
4110
4111         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4112                 ret = -EINVAL;
4113                 goto out;
4114         }
4115
4116         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4117         if (ret)
4118                 goto out;
4119
4120         set_inode_flag(inode, FI_ENABLE_COMPRESS);
4121
4122         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4123
4124         count = last_idx - page_idx;
4125         while (count) {
4126                 int len = min(cluster_size, count);
4127
4128                 ret = redirty_blocks(inode, page_idx, len);
4129                 if (ret < 0)
4130                         break;
4131
4132                 if (get_dirty_pages(inode) >= blk_per_seg)
4133                         filemap_fdatawrite(inode->i_mapping);
4134
4135                 count -= len;
4136                 page_idx += len;
4137         }
4138
4139         if (!ret)
4140                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4141                                                         LLONG_MAX);
4142
4143         clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4144
4145         if (ret)
4146                 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4147                           __func__, ret);
4148 out:
4149         inode_unlock(inode);
4150         file_end_write(filp);
4151
4152         return ret;
4153 }
4154
4155 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4156 {
4157         switch (cmd) {
4158         case FS_IOC_GETVERSION:
4159                 return f2fs_ioc_getversion(filp, arg);
4160         case F2FS_IOC_START_ATOMIC_WRITE:
4161                 return f2fs_ioc_start_atomic_write(filp);
4162         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4163                 return f2fs_ioc_commit_atomic_write(filp);
4164         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4165                 return f2fs_ioc_abort_atomic_write(filp);
4166         case F2FS_IOC_START_VOLATILE_WRITE:
4167         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4168                 return -EOPNOTSUPP;
4169         case F2FS_IOC_SHUTDOWN:
4170                 return f2fs_ioc_shutdown(filp, arg);
4171         case FITRIM:
4172                 return f2fs_ioc_fitrim(filp, arg);
4173         case FS_IOC_SET_ENCRYPTION_POLICY:
4174                 return f2fs_ioc_set_encryption_policy(filp, arg);
4175         case FS_IOC_GET_ENCRYPTION_POLICY:
4176                 return f2fs_ioc_get_encryption_policy(filp, arg);
4177         case FS_IOC_GET_ENCRYPTION_PWSALT:
4178                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4179         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4180                 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4181         case FS_IOC_ADD_ENCRYPTION_KEY:
4182                 return f2fs_ioc_add_encryption_key(filp, arg);
4183         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4184                 return f2fs_ioc_remove_encryption_key(filp, arg);
4185         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4186                 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4187         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4188                 return f2fs_ioc_get_encryption_key_status(filp, arg);
4189         case FS_IOC_GET_ENCRYPTION_NONCE:
4190                 return f2fs_ioc_get_encryption_nonce(filp, arg);
4191         case F2FS_IOC_GARBAGE_COLLECT:
4192                 return f2fs_ioc_gc(filp, arg);
4193         case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4194                 return f2fs_ioc_gc_range(filp, arg);
4195         case F2FS_IOC_WRITE_CHECKPOINT:
4196                 return f2fs_ioc_write_checkpoint(filp, arg);
4197         case F2FS_IOC_DEFRAGMENT:
4198                 return f2fs_ioc_defragment(filp, arg);
4199         case F2FS_IOC_MOVE_RANGE:
4200                 return f2fs_ioc_move_range(filp, arg);
4201         case F2FS_IOC_FLUSH_DEVICE:
4202                 return f2fs_ioc_flush_device(filp, arg);
4203         case F2FS_IOC_GET_FEATURES:
4204                 return f2fs_ioc_get_features(filp, arg);
4205         case F2FS_IOC_GET_PIN_FILE:
4206                 return f2fs_ioc_get_pin_file(filp, arg);
4207         case F2FS_IOC_SET_PIN_FILE:
4208                 return f2fs_ioc_set_pin_file(filp, arg);
4209         case F2FS_IOC_PRECACHE_EXTENTS:
4210                 return f2fs_ioc_precache_extents(filp, arg);
4211         case F2FS_IOC_RESIZE_FS:
4212                 return f2fs_ioc_resize_fs(filp, arg);
4213         case FS_IOC_ENABLE_VERITY:
4214                 return f2fs_ioc_enable_verity(filp, arg);
4215         case FS_IOC_MEASURE_VERITY:
4216                 return f2fs_ioc_measure_verity(filp, arg);
4217         case FS_IOC_READ_VERITY_METADATA:
4218                 return f2fs_ioc_read_verity_metadata(filp, arg);
4219         case FS_IOC_GETFSLABEL:
4220                 return f2fs_ioc_getfslabel(filp, arg);
4221         case FS_IOC_SETFSLABEL:
4222                 return f2fs_ioc_setfslabel(filp, arg);
4223         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4224                 return f2fs_get_compress_blocks(filp, arg);
4225         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4226                 return f2fs_release_compress_blocks(filp, arg);
4227         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4228                 return f2fs_reserve_compress_blocks(filp, arg);
4229         case F2FS_IOC_SEC_TRIM_FILE:
4230                 return f2fs_sec_trim_file(filp, arg);
4231         case F2FS_IOC_GET_COMPRESS_OPTION:
4232                 return f2fs_ioc_get_compress_option(filp, arg);
4233         case F2FS_IOC_SET_COMPRESS_OPTION:
4234                 return f2fs_ioc_set_compress_option(filp, arg);
4235         case F2FS_IOC_DECOMPRESS_FILE:
4236                 return f2fs_ioc_decompress_file(filp, arg);
4237         case F2FS_IOC_COMPRESS_FILE:
4238                 return f2fs_ioc_compress_file(filp, arg);
4239         default:
4240                 return -ENOTTY;
4241         }
4242 }
4243
4244 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4245 {
4246         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4247                 return -EIO;
4248         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4249                 return -ENOSPC;
4250
4251         return __f2fs_ioctl(filp, cmd, arg);
4252 }
4253
4254 /*
4255  * Return %true if the given read or write request should use direct I/O, or
4256  * %false if it should use buffered I/O.
4257  */
4258 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4259                                 struct iov_iter *iter)
4260 {
4261         unsigned int align;
4262
4263         if (!(iocb->ki_flags & IOCB_DIRECT))
4264                 return false;
4265
4266         if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4267                 return false;
4268
4269         /*
4270          * Direct I/O not aligned to the disk's logical_block_size will be
4271          * attempted, but will fail with -EINVAL.
4272          *
4273          * f2fs additionally requires that direct I/O be aligned to the
4274          * filesystem block size, which is often a stricter requirement.
4275          * However, f2fs traditionally falls back to buffered I/O on requests
4276          * that are logical_block_size-aligned but not fs-block aligned.
4277          *
4278          * The below logic implements this behavior.
4279          */
4280         align = iocb->ki_pos | iov_iter_alignment(iter);
4281         if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4282             IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4283                 return false;
4284
4285         return true;
4286 }
4287
4288 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4289                                 unsigned int flags)
4290 {
4291         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4292
4293         dec_page_count(sbi, F2FS_DIO_READ);
4294         if (error)
4295                 return error;
4296         f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4297         return 0;
4298 }
4299
4300 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4301         .end_io = f2fs_dio_read_end_io,
4302 };
4303
4304 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4305 {
4306         struct file *file = iocb->ki_filp;
4307         struct inode *inode = file_inode(file);
4308         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4309         struct f2fs_inode_info *fi = F2FS_I(inode);
4310         const loff_t pos = iocb->ki_pos;
4311         const size_t count = iov_iter_count(to);
4312         struct iomap_dio *dio;
4313         ssize_t ret;
4314
4315         if (count == 0)
4316                 return 0; /* skip atime update */
4317
4318         trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4319
4320         if (iocb->ki_flags & IOCB_NOWAIT) {
4321                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4322                         ret = -EAGAIN;
4323                         goto out;
4324                 }
4325         } else {
4326                 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4327         }
4328
4329         /*
4330          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4331          * the higher-level function iomap_dio_rw() in order to ensure that the
4332          * F2FS_DIO_READ counter will be decremented correctly in all cases.
4333          */
4334         inc_page_count(sbi, F2FS_DIO_READ);
4335         dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4336                              &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4337         if (IS_ERR_OR_NULL(dio)) {
4338                 ret = PTR_ERR_OR_ZERO(dio);
4339                 if (ret != -EIOCBQUEUED)
4340                         dec_page_count(sbi, F2FS_DIO_READ);
4341         } else {
4342                 ret = iomap_dio_complete(dio);
4343         }
4344
4345         f2fs_up_read(&fi->i_gc_rwsem[READ]);
4346
4347         file_accessed(file);
4348 out:
4349         trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4350         return ret;
4351 }
4352
4353 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4354 {
4355         struct inode *inode = file_inode(iocb->ki_filp);
4356         const loff_t pos = iocb->ki_pos;
4357         ssize_t ret;
4358
4359         if (!f2fs_is_compress_backend_ready(inode))
4360                 return -EOPNOTSUPP;
4361
4362         if (trace_f2fs_dataread_start_enabled()) {
4363                 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4364                 char *path;
4365
4366                 if (!p)
4367                         goto skip_read_trace;
4368
4369                 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4370                 if (IS_ERR(path)) {
4371                         kfree(p);
4372                         goto skip_read_trace;
4373                 }
4374
4375                 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4376                                         current->pid, path, current->comm);
4377                 kfree(p);
4378         }
4379 skip_read_trace:
4380         if (f2fs_should_use_dio(inode, iocb, to)) {
4381                 ret = f2fs_dio_read_iter(iocb, to);
4382         } else {
4383                 ret = filemap_read(iocb, to, 0);
4384                 if (ret > 0)
4385                         f2fs_update_iostat(F2FS_I_SB(inode), inode,
4386                                                 APP_BUFFERED_READ_IO, ret);
4387         }
4388         if (trace_f2fs_dataread_end_enabled())
4389                 trace_f2fs_dataread_end(inode, pos, ret);
4390         return ret;
4391 }
4392
4393 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4394 {
4395         struct file *file = iocb->ki_filp;
4396         struct inode *inode = file_inode(file);
4397         ssize_t count;
4398         int err;
4399
4400         if (IS_IMMUTABLE(inode))
4401                 return -EPERM;
4402
4403         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4404                 return -EPERM;
4405
4406         count = generic_write_checks(iocb, from);
4407         if (count <= 0)
4408                 return count;
4409
4410         err = file_modified(file);
4411         if (err)
4412                 return err;
4413         return count;
4414 }
4415
4416 /*
4417  * Preallocate blocks for a write request, if it is possible and helpful to do
4418  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4419  * blocks were preallocated, or a negative errno value if something went
4420  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4421  * requested blocks (not just some of them) have been allocated.
4422  */
4423 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4424                                    bool dio)
4425 {
4426         struct inode *inode = file_inode(iocb->ki_filp);
4427         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4428         const loff_t pos = iocb->ki_pos;
4429         const size_t count = iov_iter_count(iter);
4430         struct f2fs_map_blocks map = {};
4431         int flag;
4432         int ret;
4433
4434         /* If it will be an out-of-place direct write, don't bother. */
4435         if (dio && f2fs_lfs_mode(sbi))
4436                 return 0;
4437         /*
4438          * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4439          * buffered IO, if DIO meets any holes.
4440          */
4441         if (dio && i_size_read(inode) &&
4442                 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4443                 return 0;
4444
4445         /* No-wait I/O can't allocate blocks. */
4446         if (iocb->ki_flags & IOCB_NOWAIT)
4447                 return 0;
4448
4449         /* If it will be a short write, don't bother. */
4450         if (fault_in_iov_iter_readable(iter, count))
4451                 return 0;
4452
4453         if (f2fs_has_inline_data(inode)) {
4454                 /* If the data will fit inline, don't bother. */
4455                 if (pos + count <= MAX_INLINE_DATA(inode))
4456                         return 0;
4457                 ret = f2fs_convert_inline_inode(inode);
4458                 if (ret)
4459                         return ret;
4460         }
4461
4462         /* Do not preallocate blocks that will be written partially in 4KB. */
4463         map.m_lblk = F2FS_BLK_ALIGN(pos);
4464         map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4465         if (map.m_len > map.m_lblk)
4466                 map.m_len -= map.m_lblk;
4467         else
4468                 map.m_len = 0;
4469         map.m_may_create = true;
4470         if (dio) {
4471                 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4472                 flag = F2FS_GET_BLOCK_PRE_DIO;
4473         } else {
4474                 map.m_seg_type = NO_CHECK_TYPE;
4475                 flag = F2FS_GET_BLOCK_PRE_AIO;
4476         }
4477
4478         ret = f2fs_map_blocks(inode, &map, 1, flag);
4479         /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4480         if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4481                 return ret;
4482         if (ret == 0)
4483                 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4484         return map.m_len;
4485 }
4486
4487 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4488                                         struct iov_iter *from)
4489 {
4490         struct file *file = iocb->ki_filp;
4491         struct inode *inode = file_inode(file);
4492         ssize_t ret;
4493
4494         if (iocb->ki_flags & IOCB_NOWAIT)
4495                 return -EOPNOTSUPP;
4496
4497         current->backing_dev_info = inode_to_bdi(inode);
4498         ret = generic_perform_write(iocb, from);
4499         current->backing_dev_info = NULL;
4500
4501         if (ret > 0) {
4502                 iocb->ki_pos += ret;
4503                 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4504                                                 APP_BUFFERED_IO, ret);
4505         }
4506         return ret;
4507 }
4508
4509 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4510                                  unsigned int flags)
4511 {
4512         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4513
4514         dec_page_count(sbi, F2FS_DIO_WRITE);
4515         if (error)
4516                 return error;
4517         f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4518         return 0;
4519 }
4520
4521 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4522         .end_io = f2fs_dio_write_end_io,
4523 };
4524
4525 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4526                                    bool *may_need_sync)
4527 {
4528         struct file *file = iocb->ki_filp;
4529         struct inode *inode = file_inode(file);
4530         struct f2fs_inode_info *fi = F2FS_I(inode);
4531         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4532         const bool do_opu = f2fs_lfs_mode(sbi);
4533         const loff_t pos = iocb->ki_pos;
4534         const ssize_t count = iov_iter_count(from);
4535         unsigned int dio_flags;
4536         struct iomap_dio *dio;
4537         ssize_t ret;
4538
4539         trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4540
4541         if (iocb->ki_flags & IOCB_NOWAIT) {
4542                 /* f2fs_convert_inline_inode() and block allocation can block */
4543                 if (f2fs_has_inline_data(inode) ||
4544                     !f2fs_overwrite_io(inode, pos, count)) {
4545                         ret = -EAGAIN;
4546                         goto out;
4547                 }
4548
4549                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4550                         ret = -EAGAIN;
4551                         goto out;
4552                 }
4553                 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4554                         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4555                         ret = -EAGAIN;
4556                         goto out;
4557                 }
4558         } else {
4559                 ret = f2fs_convert_inline_inode(inode);
4560                 if (ret)
4561                         goto out;
4562
4563                 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4564                 if (do_opu)
4565                         f2fs_down_read(&fi->i_gc_rwsem[READ]);
4566         }
4567
4568         /*
4569          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4570          * the higher-level function iomap_dio_rw() in order to ensure that the
4571          * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4572          */
4573         inc_page_count(sbi, F2FS_DIO_WRITE);
4574         dio_flags = 0;
4575         if (pos + count > inode->i_size)
4576                 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4577         dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4578                              &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4579         if (IS_ERR_OR_NULL(dio)) {
4580                 ret = PTR_ERR_OR_ZERO(dio);
4581                 if (ret == -ENOTBLK)
4582                         ret = 0;
4583                 if (ret != -EIOCBQUEUED)
4584                         dec_page_count(sbi, F2FS_DIO_WRITE);
4585         } else {
4586                 ret = iomap_dio_complete(dio);
4587         }
4588
4589         if (do_opu)
4590                 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4591         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4592
4593         if (ret < 0)
4594                 goto out;
4595         if (pos + ret > inode->i_size)
4596                 f2fs_i_size_write(inode, pos + ret);
4597         if (!do_opu)
4598                 set_inode_flag(inode, FI_UPDATE_WRITE);
4599
4600         if (iov_iter_count(from)) {
4601                 ssize_t ret2;
4602                 loff_t bufio_start_pos = iocb->ki_pos;
4603
4604                 /*
4605                  * The direct write was partial, so we need to fall back to a
4606                  * buffered write for the remainder.
4607                  */
4608
4609                 ret2 = f2fs_buffered_write_iter(iocb, from);
4610                 if (iov_iter_count(from))
4611                         f2fs_write_failed(inode, iocb->ki_pos);
4612                 if (ret2 < 0)
4613                         goto out;
4614
4615                 /*
4616                  * Ensure that the pagecache pages are written to disk and
4617                  * invalidated to preserve the expected O_DIRECT semantics.
4618                  */
4619                 if (ret2 > 0) {
4620                         loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4621
4622                         ret += ret2;
4623
4624                         ret2 = filemap_write_and_wait_range(file->f_mapping,
4625                                                             bufio_start_pos,
4626                                                             bufio_end_pos);
4627                         if (ret2 < 0)
4628                                 goto out;
4629                         invalidate_mapping_pages(file->f_mapping,
4630                                                  bufio_start_pos >> PAGE_SHIFT,
4631                                                  bufio_end_pos >> PAGE_SHIFT);
4632                 }
4633         } else {
4634                 /* iomap_dio_rw() already handled the generic_write_sync(). */
4635                 *may_need_sync = false;
4636         }
4637 out:
4638         trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4639         return ret;
4640 }
4641
4642 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4643 {
4644         struct inode *inode = file_inode(iocb->ki_filp);
4645         const loff_t orig_pos = iocb->ki_pos;
4646         const size_t orig_count = iov_iter_count(from);
4647         loff_t target_size;
4648         bool dio;
4649         bool may_need_sync = true;
4650         int preallocated;
4651         ssize_t ret;
4652
4653         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4654                 ret = -EIO;
4655                 goto out;
4656         }
4657
4658         if (!f2fs_is_compress_backend_ready(inode)) {
4659                 ret = -EOPNOTSUPP;
4660                 goto out;
4661         }
4662
4663         if (iocb->ki_flags & IOCB_NOWAIT) {
4664                 if (!inode_trylock(inode)) {
4665                         ret = -EAGAIN;
4666                         goto out;
4667                 }
4668         } else {
4669                 inode_lock(inode);
4670         }
4671
4672         ret = f2fs_write_checks(iocb, from);
4673         if (ret <= 0)
4674                 goto out_unlock;
4675
4676         /* Determine whether we will do a direct write or a buffered write. */
4677         dio = f2fs_should_use_dio(inode, iocb, from);
4678
4679         /* Possibly preallocate the blocks for the write. */
4680         target_size = iocb->ki_pos + iov_iter_count(from);
4681         preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4682         if (preallocated < 0) {
4683                 ret = preallocated;
4684         } else {
4685                 if (trace_f2fs_datawrite_start_enabled()) {
4686                         char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4687                                                 PATH_MAX, GFP_KERNEL);
4688                         char *path;
4689
4690                         if (!p)
4691                                 goto skip_write_trace;
4692                         path = dentry_path_raw(file_dentry(iocb->ki_filp),
4693                                                                 p, PATH_MAX);
4694                         if (IS_ERR(path)) {
4695                                 kfree(p);
4696                                 goto skip_write_trace;
4697                         }
4698                         trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4699                                         current->pid, path, current->comm);
4700                         kfree(p);
4701                 }
4702 skip_write_trace:
4703                 /* Do the actual write. */
4704                 ret = dio ?
4705                         f2fs_dio_write_iter(iocb, from, &may_need_sync) :
4706                         f2fs_buffered_write_iter(iocb, from);
4707
4708                 if (trace_f2fs_datawrite_end_enabled())
4709                         trace_f2fs_datawrite_end(inode, orig_pos, ret);
4710         }
4711
4712         /* Don't leave any preallocated blocks around past i_size. */
4713         if (preallocated && i_size_read(inode) < target_size) {
4714                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4715                 filemap_invalidate_lock(inode->i_mapping);
4716                 if (!f2fs_truncate(inode))
4717                         file_dont_truncate(inode);
4718                 filemap_invalidate_unlock(inode->i_mapping);
4719                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4720         } else {
4721                 file_dont_truncate(inode);
4722         }
4723
4724         clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4725 out_unlock:
4726         inode_unlock(inode);
4727 out:
4728         trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4729         if (ret > 0 && may_need_sync)
4730                 ret = generic_write_sync(iocb, ret);
4731         return ret;
4732 }
4733
4734 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4735                 int advice)
4736 {
4737         struct address_space *mapping;
4738         struct backing_dev_info *bdi;
4739         struct inode *inode = file_inode(filp);
4740         int err;
4741
4742         if (advice == POSIX_FADV_SEQUENTIAL) {
4743                 if (S_ISFIFO(inode->i_mode))
4744                         return -ESPIPE;
4745
4746                 mapping = filp->f_mapping;
4747                 if (!mapping || len < 0)
4748                         return -EINVAL;
4749
4750                 bdi = inode_to_bdi(mapping->host);
4751                 filp->f_ra.ra_pages = bdi->ra_pages *
4752                         F2FS_I_SB(inode)->seq_file_ra_mul;
4753                 spin_lock(&filp->f_lock);
4754                 filp->f_mode &= ~FMODE_RANDOM;
4755                 spin_unlock(&filp->f_lock);
4756                 return 0;
4757         }
4758
4759         err = generic_fadvise(filp, offset, len, advice);
4760         if (!err && advice == POSIX_FADV_DONTNEED &&
4761                 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4762                 f2fs_compressed_file(inode))
4763                 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4764
4765         return err;
4766 }
4767
4768 #ifdef CONFIG_COMPAT
4769 struct compat_f2fs_gc_range {
4770         u32 sync;
4771         compat_u64 start;
4772         compat_u64 len;
4773 };
4774 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE        _IOW(F2FS_IOCTL_MAGIC, 11,\
4775                                                 struct compat_f2fs_gc_range)
4776
4777 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4778 {
4779         struct compat_f2fs_gc_range __user *urange;
4780         struct f2fs_gc_range range;
4781         int err;
4782
4783         urange = compat_ptr(arg);
4784         err = get_user(range.sync, &urange->sync);
4785         err |= get_user(range.start, &urange->start);
4786         err |= get_user(range.len, &urange->len);
4787         if (err)
4788                 return -EFAULT;
4789
4790         return __f2fs_ioc_gc_range(file, &range);
4791 }
4792
4793 struct compat_f2fs_move_range {
4794         u32 dst_fd;
4795         compat_u64 pos_in;
4796         compat_u64 pos_out;
4797         compat_u64 len;
4798 };
4799 #define F2FS_IOC32_MOVE_RANGE           _IOWR(F2FS_IOCTL_MAGIC, 9,      \
4800                                         struct compat_f2fs_move_range)
4801
4802 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4803 {
4804         struct compat_f2fs_move_range __user *urange;
4805         struct f2fs_move_range range;
4806         int err;
4807
4808         urange = compat_ptr(arg);
4809         err = get_user(range.dst_fd, &urange->dst_fd);
4810         err |= get_user(range.pos_in, &urange->pos_in);
4811         err |= get_user(range.pos_out, &urange->pos_out);
4812         err |= get_user(range.len, &urange->len);
4813         if (err)
4814                 return -EFAULT;
4815
4816         return __f2fs_ioc_move_range(file, &range);
4817 }
4818
4819 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4820 {
4821         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4822                 return -EIO;
4823         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4824                 return -ENOSPC;
4825
4826         switch (cmd) {
4827         case FS_IOC32_GETVERSION:
4828                 cmd = FS_IOC_GETVERSION;
4829                 break;
4830         case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4831                 return f2fs_compat_ioc_gc_range(file, arg);
4832         case F2FS_IOC32_MOVE_RANGE:
4833                 return f2fs_compat_ioc_move_range(file, arg);
4834         case F2FS_IOC_START_ATOMIC_WRITE:
4835         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4836         case F2FS_IOC_START_VOLATILE_WRITE:
4837         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4838         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4839         case F2FS_IOC_SHUTDOWN:
4840         case FITRIM:
4841         case FS_IOC_SET_ENCRYPTION_POLICY:
4842         case FS_IOC_GET_ENCRYPTION_PWSALT:
4843         case FS_IOC_GET_ENCRYPTION_POLICY:
4844         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4845         case FS_IOC_ADD_ENCRYPTION_KEY:
4846         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4847         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4848         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4849         case FS_IOC_GET_ENCRYPTION_NONCE:
4850         case F2FS_IOC_GARBAGE_COLLECT:
4851         case F2FS_IOC_WRITE_CHECKPOINT:
4852         case F2FS_IOC_DEFRAGMENT:
4853         case F2FS_IOC_FLUSH_DEVICE:
4854         case F2FS_IOC_GET_FEATURES:
4855         case F2FS_IOC_GET_PIN_FILE:
4856         case F2FS_IOC_SET_PIN_FILE:
4857         case F2FS_IOC_PRECACHE_EXTENTS:
4858         case F2FS_IOC_RESIZE_FS:
4859         case FS_IOC_ENABLE_VERITY:
4860         case FS_IOC_MEASURE_VERITY:
4861         case FS_IOC_READ_VERITY_METADATA:
4862         case FS_IOC_GETFSLABEL:
4863         case FS_IOC_SETFSLABEL:
4864         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4865         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4866         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4867         case F2FS_IOC_SEC_TRIM_FILE:
4868         case F2FS_IOC_GET_COMPRESS_OPTION:
4869         case F2FS_IOC_SET_COMPRESS_OPTION:
4870         case F2FS_IOC_DECOMPRESS_FILE:
4871         case F2FS_IOC_COMPRESS_FILE:
4872                 break;
4873         default:
4874                 return -ENOIOCTLCMD;
4875         }
4876         return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4877 }
4878 #endif
4879
4880 const struct file_operations f2fs_file_operations = {
4881         .llseek         = f2fs_llseek,
4882         .read_iter      = f2fs_file_read_iter,
4883         .write_iter     = f2fs_file_write_iter,
4884         .open           = f2fs_file_open,
4885         .release        = f2fs_release_file,
4886         .mmap           = f2fs_file_mmap,
4887         .flush          = f2fs_file_flush,
4888         .fsync          = f2fs_sync_file,
4889         .fallocate      = f2fs_fallocate,
4890         .unlocked_ioctl = f2fs_ioctl,
4891 #ifdef CONFIG_COMPAT
4892         .compat_ioctl   = f2fs_compat_ioctl,
4893 #endif
4894         .splice_read    = generic_file_splice_read,
4895         .splice_write   = iter_file_splice_write,
4896         .fadvise        = f2fs_file_fadvise,
4897 };