fs/f2fs/file.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * fs/f2fs/file.c
   4  *
   5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6  *             http://www.samsung.com/
   7  */
   8 #include <linux/fs.h>
   9 #include <linux/f2fs_fs.h>
  10 #include <linux/stat.h>
  11 #include <linux/buffer_head.h>
  12 #include <linux/writeback.h>
  13 #include <linux/blkdev.h>
  14 #include <linux/falloc.h>
  15 #include <linux/types.h>
  16 #include <linux/compat.h>
  17 #include <linux/uaccess.h>
  18 #include <linux/mount.h>
  19 #include <linux/pagevec.h>
  20 #include <linux/uio.h>
  21 #include <linux/uuid.h>
  22 #include <linux/file.h>
  23 #include <linux/nls.h>
  24 #include <linux/sched/signal.h>
  25 #include <linux/fileattr.h>
  26 #include <linux/fadvise.h>
  27 #include <linux/iomap.h>
  28
  29 #include "f2fs.h"
  30 #include "node.h"
  31 #include "segment.h"
  32 #include "xattr.h"
  33 #include "acl.h"
  34 #include "gc.h"
  35 #include "iostat.h"
  36 #include <trace/events/f2fs.h>
  37 #include <uapi/linux/f2fs.h>
  38
  39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
  40 {
  41         struct inode *inode = file_inode(vmf->vma->vm_file);
  42         vm_fault_t ret;
  43
  44         ret = filemap_fault(vmf);
  45         if (!ret)
  46                 f2fs_update_iostat(F2FS_I_SB(inode), inode,
  47                                         APP_MAPPED_READ_IO, F2FS_BLKSIZE);
  48
  49         trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
  50
  51         return ret;
  52 }
  53
  54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
  55 {
  56         struct page *page = vmf->page;
  57         struct inode *inode = file_inode(vmf->vma->vm_file);
  58         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
  59         struct dnode_of_data dn;
  60         bool need_alloc = true;
  61         int err = 0;
  62
  63         if (unlikely(IS_IMMUTABLE(inode)))
  64                 return VM_FAULT_SIGBUS;
  65
  66         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
  67                 return VM_FAULT_SIGBUS;
  68
  69         if (unlikely(f2fs_cp_error(sbi))) {
  70                 err = -EIO;
  71                 goto err;
  72         }
  73
  74         if (!f2fs_is_checkpoint_ready(sbi)) {
  75                 err = -ENOSPC;
  76                 goto err;
  77         }
  78
  79         err = f2fs_convert_inline_inode(inode);
  80         if (err)
  81                 goto err;
  82
  83 #ifdef CONFIG_F2FS_FS_COMPRESSION
  84         if (f2fs_compressed_file(inode)) {
  85                 int ret = f2fs_is_compressed_cluster(inode, page->index);
  86
  87                 if (ret < 0) {
  88                         err = ret;
  89                         goto err;
  90                 } else if (ret) {
  91                         need_alloc = false;
  92                 }
  93         }
  94 #endif
  95         /* should do out of any locked page */
  96         if (need_alloc)
  97                 f2fs_balance_fs(sbi, true);
  98
  99         sb_start_pagefault(inode->i_sb);
 100
 101         f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
 102
 103         file_update_time(vmf->vma->vm_file);
 104         filemap_invalidate_lock_shared(inode->i_mapping);
 105         lock_page(page);
 106         if (unlikely(page->mapping != inode->i_mapping ||
 107                         page_offset(page) > i_size_read(inode) ||
 108                         !PageUptodate(page))) {
 109                 unlock_page(page);
 110                 err = -EFAULT;
 111                 goto out_sem;
 112         }
 113
 114         if (need_alloc) {
 115                 /* block allocation */
 116                 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
 117                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 118                 err = f2fs_get_block(&dn, page->index);
 119                 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
 120         }
 121
 122 #ifdef CONFIG_F2FS_FS_COMPRESSION
 123         if (!need_alloc) {
 124                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 125                 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
 126                 f2fs_put_dnode(&dn);
 127         }
 128 #endif
 129         if (err) {
 130                 unlock_page(page);
 131                 goto out_sem;
 132         }
 133
 134         f2fs_wait_on_page_writeback(page, DATA, false, true);
 135
 136         /* wait for GCed page writeback via META_MAPPING */
 137         f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
 138
 139         /*
 140          * check to see if the page is mapped already (no holes)
 141          */
 142         if (PageMappedToDisk(page))
 143                 goto out_sem;
 144
 145         /* page is wholly or partially inside EOF */
 146         if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
 147                                                 i_size_read(inode)) {
 148                 loff_t offset;
 149
 150                 offset = i_size_read(inode) & ~PAGE_MASK;
 151                 zero_user_segment(page, offset, PAGE_SIZE);
 152         }
 153         set_page_dirty(page);
 154         if (!PageUptodate(page))
 155                 SetPageUptodate(page);
 156
 157         f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
 158         f2fs_update_time(sbi, REQ_TIME);
 159
 160         trace_f2fs_vm_page_mkwrite(page, DATA);
 161 out_sem:
 162         filemap_invalidate_unlock_shared(inode->i_mapping);
 163
 164         sb_end_pagefault(inode->i_sb);
 165 err:
 166         return block_page_mkwrite_return(err);
 167 }
 168
 169 static const struct vm_operations_struct f2fs_file_vm_ops = {
 170         .fault          = f2fs_filemap_fault,
 171         .map_pages      = filemap_map_pages,
 172         .page_mkwrite   = f2fs_vm_page_mkwrite,
 173 };
 174
 175 static int get_parent_ino(struct inode *inode, nid_t *pino)
 176 {
 177         struct dentry *dentry;
 178
 179         /*
 180          * Make sure to get the non-deleted alias.  The alias associated with
 181          * the open file descriptor being fsync()'ed may be deleted already.
 182          */
 183         dentry = d_find_alias(inode);
 184         if (!dentry)
 185                 return 0;
 186
 187         *pino = parent_ino(dentry);
 188         dput(dentry);
 189         return 1;
 190 }
 191
 192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
 193 {
 194         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 195         enum cp_reason_type cp_reason = CP_NO_NEEDED;
 196
 197         if (!S_ISREG(inode->i_mode))
 198                 cp_reason = CP_NON_REGULAR;
 199         else if (f2fs_compressed_file(inode))
 200                 cp_reason = CP_COMPRESSED;
 201         else if (inode->i_nlink != 1)
 202                 cp_reason = CP_HARDLINK;
 203         else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
 204                 cp_reason = CP_SB_NEED_CP;
 205         else if (file_wrong_pino(inode))
 206                 cp_reason = CP_WRONG_PINO;
 207         else if (!f2fs_space_for_roll_forward(sbi))
 208                 cp_reason = CP_NO_SPC_ROLL;
 209         else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
 210                 cp_reason = CP_NODE_NEED_CP;
 211         else if (test_opt(sbi, FASTBOOT))
 212                 cp_reason = CP_FASTBOOT_MODE;
 213         else if (F2FS_OPTION(sbi).active_logs == 2)
 214                 cp_reason = CP_SPEC_LOG_NUM;
 215         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
 216                 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
 217                 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
 218                                                         TRANS_DIR_INO))
 219                 cp_reason = CP_RECOVER_DIR;
 220
 221         return cp_reason;
 222 }
 223
 224 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
 225 {
 226         struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
 227         bool ret = false;
 228         /* But we need to avoid that there are some inode updates */
 229         if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
 230                 ret = true;
 231         f2fs_put_page(i, 0);
 232         return ret;
 233 }
 234
 235 static void try_to_fix_pino(struct inode *inode)
 236 {
 237         struct f2fs_inode_info *fi = F2FS_I(inode);
 238         nid_t pino;
 239
 240         f2fs_down_write(&fi->i_sem);
 241         if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
 242                         get_parent_ino(inode, &pino)) {
 243                 f2fs_i_pino_write(inode, pino);
 244                 file_got_pino(inode);
 245         }
 246         f2fs_up_write(&fi->i_sem);
 247 }
 248
 249 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
 250                                                 int datasync, bool atomic)
 251 {
 252         struct inode *inode = file->f_mapping->host;
 253         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 254         nid_t ino = inode->i_ino;
 255         int ret = 0;
 256         enum cp_reason_type cp_reason = 0;
 257         struct writeback_control wbc = {
 258                 .sync_mode = WB_SYNC_ALL,
 259                 .nr_to_write = LONG_MAX,
 260                 .for_reclaim = 0,
 261         };
 262         unsigned int seq_id = 0;
 263
 264         if (unlikely(f2fs_readonly(inode->i_sb)))
 265                 return 0;
 266
 267         trace_f2fs_sync_file_enter(inode);
 268
 269         if (S_ISDIR(inode->i_mode))
 270                 goto go_write;
 271
 272         /* if fdatasync is triggered, let's do in-place-update */
 273         if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
 274                 set_inode_flag(inode, FI_NEED_IPU);
 275         ret = file_write_and_wait_range(file, start, end);
 276         clear_inode_flag(inode, FI_NEED_IPU);
 277
 278         if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
 279                 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 280                 return ret;
 281         }
 282
 283         /* if the inode is dirty, let's recover all the time */
 284         if (!f2fs_skip_inode_update(inode, datasync)) {
 285                 f2fs_write_inode(inode, NULL);
 286                 goto go_write;
 287         }
 288
 289         /*
 290          * if there is no written data, don't waste time to write recovery info.
 291          */
 292         if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
 293                         !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
 294
 295                 /* it may call write_inode just prior to fsync */
 296                 if (need_inode_page_update(sbi, ino))
 297                         goto go_write;
 298
 299                 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
 300                                 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
 301                         goto flush_out;
 302                 goto out;
 303         } else {
 304                 /*
 305                  * for OPU case, during fsync(), node can be persisted before
 306                  * data when lower device doesn't support write barrier, result
 307                  * in data corruption after SPO.
 308                  * So for strict fsync mode, force to use atomic write sematics
 309                  * to keep write order in between data/node and last node to
 310                  * avoid potential data corruption.
 311                  */
 312                 if (F2FS_OPTION(sbi).fsync_mode ==
 313                                 FSYNC_MODE_STRICT && !atomic)
 314                         atomic = true;
 315         }
 316 go_write:
 317         /*
 318          * Both of fdatasync() and fsync() are able to be recovered from
 319          * sudden-power-off.
 320          */
 321         f2fs_down_read(&F2FS_I(inode)->i_sem);
 322         cp_reason = need_do_checkpoint(inode);
 323         f2fs_up_read(&F2FS_I(inode)->i_sem);
 324
 325         if (cp_reason) {
 326                 /* all the dirty node pages should be flushed for POR */
 327                 ret = f2fs_sync_fs(inode->i_sb, 1);
 328
 329                 /*
 330                  * We've secured consistency through sync_fs. Following pino
 331                  * will be used only for fsynced inodes after checkpoint.
 332                  */
 333                 try_to_fix_pino(inode);
 334                 clear_inode_flag(inode, FI_APPEND_WRITE);
 335                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 336                 goto out;
 337         }
 338 sync_nodes:
 339         atomic_inc(&sbi->wb_sync_req[NODE]);
 340         ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
 341         atomic_dec(&sbi->wb_sync_req[NODE]);
 342         if (ret)
 343                 goto out;
 344
 345         /* if cp_error was enabled, we should avoid infinite loop */
 346         if (unlikely(f2fs_cp_error(sbi))) {
 347                 ret = -EIO;
 348                 goto out;
 349         }
 350
 351         if (f2fs_need_inode_block_update(sbi, ino)) {
 352                 f2fs_mark_inode_dirty_sync(inode, true);
 353                 f2fs_write_inode(inode, NULL);
 354                 goto sync_nodes;
 355         }
 356
 357         /*
 358          * If it's atomic_write, it's just fine to keep write ordering. So
 359          * here we don't need to wait for node write completion, since we use
 360          * node chain which serializes node blocks. If one of node writes are
 361          * reordered, we can see simply broken chain, resulting in stopping
 362          * roll-forward recovery. It means we'll recover all or none node blocks
 363          * given fsync mark.
 364          */
 365         if (!atomic) {
 366                 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
 367                 if (ret)
 368                         goto out;
 369         }
 370
 371         /* once recovery info is written, don't need to tack this */
 372         f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
 373         clear_inode_flag(inode, FI_APPEND_WRITE);
 374 flush_out:
 375         if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
 376             (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
 377                 ret = f2fs_issue_flush(sbi, inode->i_ino);
 378         if (!ret) {
 379                 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
 380                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 381                 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
 382         }
 383         f2fs_update_time(sbi, REQ_TIME);
 384 out:
 385         trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 386         return ret;
 387 }
 388
 389 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 390 {
 391         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
 392                 return -EIO;
 393         return f2fs_do_sync_file(file, start, end, datasync, false);
 394 }
 395
 396 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
 397                                 pgoff_t index, int whence)
 398 {
 399         switch (whence) {
 400         case SEEK_DATA:
 401                 if (__is_valid_data_blkaddr(blkaddr))
 402                         return true;
 403                 if (blkaddr == NEW_ADDR &&
 404                     xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
 405                         return true;
 406                 break;
 407         case SEEK_HOLE:
 408                 if (blkaddr == NULL_ADDR)
 409                         return true;
 410                 break;
 411         }
 412         return false;
 413 }
 414
 415 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 416 {
 417         struct inode *inode = file->f_mapping->host;
 418         loff_t maxbytes = inode->i_sb->s_maxbytes;
 419         struct dnode_of_data dn;
 420         pgoff_t pgofs, end_offset;
 421         loff_t data_ofs = offset;
 422         loff_t isize;
 423         int err = 0;
 424
 425         inode_lock(inode);
 426
 427         isize = i_size_read(inode);
 428         if (offset >= isize)
 429                 goto fail;
 430
 431         /* handle inline data case */
 432         if (f2fs_has_inline_data(inode)) {
 433                 if (whence == SEEK_HOLE) {
 434                         data_ofs = isize;
 435                         goto found;
 436                 } else if (whence == SEEK_DATA) {
 437                         data_ofs = offset;
 438                         goto found;
 439                 }
 440         }
 441
 442         pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
 443
 444         for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 445                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 446                 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
 447                 if (err && err != -ENOENT) {
 448                         goto fail;
 449                 } else if (err == -ENOENT) {
 450                         /* direct node does not exists */
 451                         if (whence == SEEK_DATA) {
 452                                 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
 453                                 continue;
 454                         } else {
 455                                 goto found;
 456                         }
 457                 }
 458
 459                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 460
 461                 /* find data/hole in dnode block */
 462                 for (; dn.ofs_in_node < end_offset;
 463                                 dn.ofs_in_node++, pgofs++,
 464                                 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 465                         block_t blkaddr;
 466
 467                         blkaddr = f2fs_data_blkaddr(&dn);
 468
 469                         if (__is_valid_data_blkaddr(blkaddr) &&
 470                                 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
 471                                         blkaddr, DATA_GENERIC_ENHANCE)) {
 472                                 f2fs_put_dnode(&dn);
 473                                 goto fail;
 474                         }
 475
 476                         if (__found_offset(file->f_mapping, blkaddr,
 477                                                         pgofs, whence)) {
 478                                 f2fs_put_dnode(&dn);
 479                                 goto found;
 480                         }
 481                 }
 482                 f2fs_put_dnode(&dn);
 483         }
 484
 485         if (whence == SEEK_DATA)
 486                 goto fail;
 487 found:
 488         if (whence == SEEK_HOLE && data_ofs > isize)
 489                 data_ofs = isize;
 490         inode_unlock(inode);
 491         return vfs_setpos(file, data_ofs, maxbytes);
 492 fail:
 493         inode_unlock(inode);
 494         return -ENXIO;
 495 }
 496
 497 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
 498 {
 499         struct inode *inode = file->f_mapping->host;
 500         loff_t maxbytes = inode->i_sb->s_maxbytes;
 501
 502         if (f2fs_compressed_file(inode))
 503                 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
 504
 505         switch (whence) {
 506         case SEEK_SET:
 507         case SEEK_CUR:
 508         case SEEK_END:
 509                 return generic_file_llseek_size(file, offset, whence,
 510                                                 maxbytes, i_size_read(inode));
 511         case SEEK_DATA:
 512         case SEEK_HOLE:
 513                 if (offset < 0)
 514                         return -ENXIO;
 515                 return f2fs_seek_block(file, offset, whence);
 516         }
 517
 518         return -EINVAL;
 519 }
 520
 521 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 522 {
 523         struct inode *inode = file_inode(file);
 524
 525         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 526                 return -EIO;
 527
 528         if (!f2fs_is_compress_backend_ready(inode))
 529                 return -EOPNOTSUPP;
 530
 531         file_accessed(file);
 532         vma->vm_ops = &f2fs_file_vm_ops;
 533         set_inode_flag(inode, FI_MMAP_FILE);
 534         return 0;
 535 }
 536
 537 static int f2fs_file_open(struct inode *inode, struct file *filp)
 538 {
 539         int err = fscrypt_file_open(inode, filp);
 540
 541         if (err)
 542                 return err;
 543
 544         if (!f2fs_is_compress_backend_ready(inode))
 545                 return -EOPNOTSUPP;
 546
 547         err = fsverity_file_open(inode, filp);
 548         if (err)
 549                 return err;
 550
 551         filp->f_mode |= FMODE_NOWAIT;
 552
 553         return dquot_file_open(inode, filp);
 554 }
 555
 556 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 557 {
 558         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 559         struct f2fs_node *raw_node;
 560         int nr_free = 0, ofs = dn->ofs_in_node, len = count;
 561         __le32 *addr;
 562         int base = 0;
 563         bool compressed_cluster = false;
 564         int cluster_index = 0, valid_blocks = 0;
 565         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
 566         bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
 567
 568         if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
 569                 base = get_extra_isize(dn->inode);
 570
 571         raw_node = F2FS_NODE(dn->node_page);
 572         addr = blkaddr_in_node(raw_node) + base + ofs;
 573
 574         /* Assumption: truncation starts with cluster */
 575         for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
 576                 block_t blkaddr = le32_to_cpu(*addr);
 577
 578                 if (f2fs_compressed_file(dn->inode) &&
 579                                         !(cluster_index & (cluster_size - 1))) {
 580                         if (compressed_cluster)
 581                                 f2fs_i_compr_blocks_update(dn->inode,
 582                                                         valid_blocks, false);
 583                         compressed_cluster = (blkaddr == COMPRESS_ADDR);
 584                         valid_blocks = 0;
 585                 }
 586
 587                 if (blkaddr == NULL_ADDR)
 588                         continue;
 589
 590                 dn->data_blkaddr = NULL_ADDR;
 591                 f2fs_set_data_blkaddr(dn);
 592
 593                 if (__is_valid_data_blkaddr(blkaddr)) {
 594                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
 595                                         DATA_GENERIC_ENHANCE))
 596                                 continue;
 597                         if (compressed_cluster)
 598                                 valid_blocks++;
 599                 }
 600
 601                 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 602                         clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 603
 604                 f2fs_invalidate_blocks(sbi, blkaddr);
 605
 606                 if (!released || blkaddr != COMPRESS_ADDR)
 607                         nr_free++;
 608         }
 609
 610         if (compressed_cluster)
 611                 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
 612
 613         if (nr_free) {
 614                 pgoff_t fofs;
 615                 /*
 616                  * once we invalidate valid blkaddr in range [ofs, ofs + count],
 617                  * we will invalidate all blkaddr in the whole range.
 618                  */
 619                 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
 620                                                         dn->inode) + ofs;
 621                 f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
 622                 f2fs_update_age_extent_cache_range(dn, fofs, nr_free);
 623                 dec_valid_block_count(sbi, dn->inode, nr_free);
 624         }
 625         dn->ofs_in_node = ofs;
 626
 627         f2fs_update_time(sbi, REQ_TIME);
 628         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 629                                          dn->ofs_in_node, nr_free);
 630 }
 631
 632 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
 633 {
 634         f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
 635 }
 636
 637 static int truncate_partial_data_page(struct inode *inode, u64 from,
 638                                                                 bool cache_only)
 639 {
 640         loff_t offset = from & (PAGE_SIZE - 1);
 641         pgoff_t index = from >> PAGE_SHIFT;
 642         struct address_space *mapping = inode->i_mapping;
 643         struct page *page;
 644
 645         if (!offset && !cache_only)
 646                 return 0;
 647
 648         if (cache_only) {
 649                 page = find_lock_page(mapping, index);
 650                 if (page && PageUptodate(page))
 651                         goto truncate_out;
 652                 f2fs_put_page(page, 1);
 653                 return 0;
 654         }
 655
 656         page = f2fs_get_lock_data_page(inode, index, true);
 657         if (IS_ERR(page))
 658                 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 659 truncate_out:
 660         f2fs_wait_on_page_writeback(page, DATA, true, true);
 661         zero_user(page, offset, PAGE_SIZE - offset);
 662
 663         /* An encrypted inode should have a key and truncate the last page. */
 664         f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
 665         if (!cache_only)
 666                 set_page_dirty(page);
 667         f2fs_put_page(page, 1);
 668         return 0;
 669 }
 670
 671 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
 672 {
 673         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 674         struct dnode_of_data dn;
 675         pgoff_t free_from;
 676         int count = 0, err = 0;
 677         struct page *ipage;
 678         bool truncate_page = false;
 679
 680         trace_f2fs_truncate_blocks_enter(inode, from);
 681
 682         free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 683
 684         if (free_from >= max_file_blocks(inode))
 685                 goto free_partial;
 686
 687         if (lock)
 688                 f2fs_lock_op(sbi);
 689
 690         ipage = f2fs_get_node_page(sbi, inode->i_ino);
 691         if (IS_ERR(ipage)) {
 692                 err = PTR_ERR(ipage);
 693                 goto out;
 694         }
 695
 696         if (f2fs_has_inline_data(inode)) {
 697                 f2fs_truncate_inline_inode(inode, ipage, from);
 698                 f2fs_put_page(ipage, 1);
 699                 truncate_page = true;
 700                 goto out;
 701         }
 702
 703         set_new_dnode(&dn, inode, ipage, NULL, 0);
 704         err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 705         if (err) {
 706                 if (err == -ENOENT)
 707                         goto free_next;
 708                 goto out;
 709         }
 710
 711         count = ADDRS_PER_PAGE(dn.node_page, inode);
 712
 713         count -= dn.ofs_in_node;
 714         f2fs_bug_on(sbi, count < 0);
 715
 716         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 717                 f2fs_truncate_data_blocks_range(&dn, count);
 718                 free_from += count;
 719         }
 720
 721         f2fs_put_dnode(&dn);
 722 free_next:
 723         err = f2fs_truncate_inode_blocks(inode, free_from);
 724 out:
 725         if (lock)
 726                 f2fs_unlock_op(sbi);
 727 free_partial:
 728         /* lastly zero out the first data page */
 729         if (!err)
 730                 err = truncate_partial_data_page(inode, from, truncate_page);
 731
 732         trace_f2fs_truncate_blocks_exit(inode, err);
 733         return err;
 734 }
 735
 736 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
 737 {
 738         u64 free_from = from;
 739         int err;
 740
 741 #ifdef CONFIG_F2FS_FS_COMPRESSION
 742         /*
 743          * for compressed file, only support cluster size
 744          * aligned truncation.
 745          */
 746         if (f2fs_compressed_file(inode))
 747                 free_from = round_up(from,
 748                                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
 749 #endif
 750
 751         err = f2fs_do_truncate_blocks(inode, free_from, lock);
 752         if (err)
 753                 return err;
 754
 755 #ifdef CONFIG_F2FS_FS_COMPRESSION
 756         /*
 757          * For compressed file, after release compress blocks, don't allow write
 758          * direct, but we should allow write direct after truncate to zero.
 759          */
 760         if (f2fs_compressed_file(inode) && !free_from
 761                         && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
 762                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 763
 764         if (from != free_from) {
 765                 err = f2fs_truncate_partial_cluster(inode, from, lock);
 766                 if (err)
 767                         return err;
 768         }
 769 #endif
 770
 771         return 0;
 772 }
 773
 774 int f2fs_truncate(struct inode *inode)
 775 {
 776         int err;
 777
 778         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 779                 return -EIO;
 780
 781         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 782                                 S_ISLNK(inode->i_mode)))
 783                 return 0;
 784
 785         trace_f2fs_truncate(inode);
 786
 787         if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
 788                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
 789                 return -EIO;
 790         }
 791
 792         err = f2fs_dquot_initialize(inode);
 793         if (err)
 794                 return err;
 795
 796         /* we should check inline_data size */
 797         if (!f2fs_may_inline_data(inode)) {
 798                 err = f2fs_convert_inline_inode(inode);
 799                 if (err)
 800                         return err;
 801         }
 802
 803         err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
 804         if (err)
 805                 return err;
 806
 807         inode->i_mtime = inode->i_ctime = current_time(inode);
 808         f2fs_mark_inode_dirty_sync(inode, false);
 809         return 0;
 810 }
 811
 812 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
 813 {
 814         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 815
 816         if (!fscrypt_dio_supported(inode))
 817                 return true;
 818         if (fsverity_active(inode))
 819                 return true;
 820         if (f2fs_compressed_file(inode))
 821                 return true;
 822
 823         /* disallow direct IO if any of devices has unaligned blksize */
 824         if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
 825                 return true;
 826         /*
 827          * for blkzoned device, fallback direct IO to buffered IO, so
 828          * all IOs can be serialized by log-structured write.
 829          */
 830         if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
 831                 return true;
 832         if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi))
 833                 return true;
 834         if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
 835                 return true;
 836
 837         return false;
 838 }
 839
 840 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 841                  struct kstat *stat, u32 request_mask, unsigned int query_flags)
 842 {
 843         struct inode *inode = d_inode(path->dentry);
 844         struct f2fs_inode_info *fi = F2FS_I(inode);
 845         struct f2fs_inode *ri = NULL;
 846         unsigned int flags;
 847
 848         if (f2fs_has_extra_attr(inode) &&
 849                         f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
 850                         F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 851                 stat->result_mask |= STATX_BTIME;
 852                 stat->btime.tv_sec = fi->i_crtime.tv_sec;
 853                 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 854         }
 855
 856         /*
 857          * Return the DIO alignment restrictions if requested.  We only return
 858          * this information when requested, since on encrypted files it might
 859          * take a fair bit of work to get if the file wasn't opened recently.
 860          *
 861          * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
 862          * cannot represent that, so in that case we report no DIO support.
 863          */
 864         if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
 865                 unsigned int bsize = i_blocksize(inode);
 866
 867                 stat->result_mask |= STATX_DIOALIGN;
 868                 if (!f2fs_force_buffered_io(inode, WRITE)) {
 869                         stat->dio_mem_align = bsize;
 870                         stat->dio_offset_align = bsize;
 871                 }
 872         }
 873
 874         flags = fi->i_flags;
 875         if (flags & F2FS_COMPR_FL)
 876                 stat->attributes |= STATX_ATTR_COMPRESSED;
 877         if (flags & F2FS_APPEND_FL)
 878                 stat->attributes |= STATX_ATTR_APPEND;
 879         if (IS_ENCRYPTED(inode))
 880                 stat->attributes |= STATX_ATTR_ENCRYPTED;
 881         if (flags & F2FS_IMMUTABLE_FL)
 882                 stat->attributes |= STATX_ATTR_IMMUTABLE;
 883         if (flags & F2FS_NODUMP_FL)
 884                 stat->attributes |= STATX_ATTR_NODUMP;
 885         if (IS_VERITY(inode))
 886                 stat->attributes |= STATX_ATTR_VERITY;
 887
 888         stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 889                                   STATX_ATTR_APPEND |
 890                                   STATX_ATTR_ENCRYPTED |
 891                                   STATX_ATTR_IMMUTABLE |
 892                                   STATX_ATTR_NODUMP |
 893                                   STATX_ATTR_VERITY);
 894
 895         generic_fillattr(mnt_userns, inode, stat);
 896
 897         /* we need to show initial sectors used for inline_data/dentries */
 898         if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 899                                         f2fs_has_inline_dentry(inode))
 900                 stat->blocks += (stat->size + 511) >> 9;
 901
 902         return 0;
 903 }
 904
 905 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 906 static void __setattr_copy(struct user_namespace *mnt_userns,
 907                            struct inode *inode, const struct iattr *attr)
 908 {
 909         unsigned int ia_valid = attr->ia_valid;
 910
 911         i_uid_update(mnt_userns, attr, inode);
 912         i_gid_update(mnt_userns, attr, inode);
 913         if (ia_valid & ATTR_ATIME)
 914                 inode->i_atime = attr->ia_atime;
 915         if (ia_valid & ATTR_MTIME)
 916                 inode->i_mtime = attr->ia_mtime;
 917         if (ia_valid & ATTR_CTIME)
 918                 inode->i_ctime = attr->ia_ctime;
 919         if (ia_valid & ATTR_MODE) {
 920                 umode_t mode = attr->ia_mode;
 921                 vfsgid_t vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
 922
 923                 if (!vfsgid_in_group_p(vfsgid) &&
 924                     !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
 925                         mode &= ~S_ISGID;
 926                 set_acl_inode(inode, mode);
 927         }
 928 }
 929 #else
 930 #define __setattr_copy setattr_copy
 931 #endif
 932
 933 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 934                  struct iattr *attr)
 935 {
 936         struct inode *inode = d_inode(dentry);
 937         int err;
 938
 939         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 940                 return -EIO;
 941
 942         if (unlikely(IS_IMMUTABLE(inode)))
 943                 return -EPERM;
 944
 945         if (unlikely(IS_APPEND(inode) &&
 946                         (attr->ia_valid & (ATTR_MODE | ATTR_UID |
 947                                   ATTR_GID | ATTR_TIMES_SET))))
 948                 return -EPERM;
 949
 950         if ((attr->ia_valid & ATTR_SIZE) &&
 951                 !f2fs_is_compress_backend_ready(inode))
 952                 return -EOPNOTSUPP;
 953
 954         err = setattr_prepare(mnt_userns, dentry, attr);
 955         if (err)
 956                 return err;
 957
 958         err = fscrypt_prepare_setattr(dentry, attr);
 959         if (err)
 960                 return err;
 961
 962         err = fsverity_prepare_setattr(dentry, attr);
 963         if (err)
 964                 return err;
 965
 966         if (is_quota_modification(mnt_userns, inode, attr)) {
 967                 err = f2fs_dquot_initialize(inode);
 968                 if (err)
 969                         return err;
 970         }
 971         if (i_uid_needs_update(mnt_userns, attr, inode) ||
 972             i_gid_needs_update(mnt_userns, attr, inode)) {
 973                 f2fs_lock_op(F2FS_I_SB(inode));
 974                 err = dquot_transfer(mnt_userns, inode, attr);
 975                 if (err) {
 976                         set_sbi_flag(F2FS_I_SB(inode),
 977                                         SBI_QUOTA_NEED_REPAIR);
 978                         f2fs_unlock_op(F2FS_I_SB(inode));
 979                         return err;
 980                 }
 981                 /*
 982                  * update uid/gid under lock_op(), so that dquot and inode can
 983                  * be updated atomically.
 984                  */
 985                 i_uid_update(mnt_userns, attr, inode);
 986                 i_gid_update(mnt_userns, attr, inode);
 987                 f2fs_mark_inode_dirty_sync(inode, true);
 988                 f2fs_unlock_op(F2FS_I_SB(inode));
 989         }
 990
 991         if (attr->ia_valid & ATTR_SIZE) {
 992                 loff_t old_size = i_size_read(inode);
 993
 994                 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
 995                         /*
 996                          * should convert inline inode before i_size_write to
 997                          * keep smaller than inline_data size with inline flag.
 998                          */
 999                         err = f2fs_convert_inline_inode(inode);
1000                         if (err)
1001                                 return err;
1002                 }
1003
1004                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1005                 filemap_invalidate_lock(inode->i_mapping);
1006
1007                 truncate_setsize(inode, attr->ia_size);
1008
1009                 if (attr->ia_size <= old_size)
1010                         err = f2fs_truncate(inode);
1011                 /*
1012                  * do not trim all blocks after i_size if target size is
1013                  * larger than i_size.
1014                  */
1015                 filemap_invalidate_unlock(inode->i_mapping);
1016                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1017                 if (err)
1018                         return err;
1019
1020                 spin_lock(&F2FS_I(inode)->i_size_lock);
1021                 inode->i_mtime = inode->i_ctime = current_time(inode);
1022                 F2FS_I(inode)->last_disk_size = i_size_read(inode);
1023                 spin_unlock(&F2FS_I(inode)->i_size_lock);
1024         }
1025
1026         __setattr_copy(mnt_userns, inode, attr);
1027
1028         if (attr->ia_valid & ATTR_MODE) {
1029                 err = posix_acl_chmod(mnt_userns, dentry, f2fs_get_inode_mode(inode));
1030
1031                 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1032                         if (!err)
1033                                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
1034                         clear_inode_flag(inode, FI_ACL_MODE);
1035                 }
1036         }
1037
1038         /* file size may changed here */
1039         f2fs_mark_inode_dirty_sync(inode, true);
1040
1041         /* inode change will produce dirty node pages flushed by checkpoint */
1042         f2fs_balance_fs(F2FS_I_SB(inode), true);
1043
1044         return err;
1045 }
1046
1047 const struct inode_operations f2fs_file_inode_operations = {
1048         .getattr        = f2fs_getattr,
1049         .setattr        = f2fs_setattr,
1050         .get_inode_acl  = f2fs_get_acl,
1051         .set_acl        = f2fs_set_acl,
1052         .listxattr      = f2fs_listxattr,
1053         .fiemap         = f2fs_fiemap,
1054         .fileattr_get   = f2fs_fileattr_get,
1055         .fileattr_set   = f2fs_fileattr_set,
1056 };
1057
1058 static int fill_zero(struct inode *inode, pgoff_t index,
1059                                         loff_t start, loff_t len)
1060 {
1061         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1062         struct page *page;
1063
1064         if (!len)
1065                 return 0;
1066
1067         f2fs_balance_fs(sbi, true);
1068
1069         f2fs_lock_op(sbi);
1070         page = f2fs_get_new_data_page(inode, NULL, index, false);
1071         f2fs_unlock_op(sbi);
1072
1073         if (IS_ERR(page))
1074                 return PTR_ERR(page);
1075
1076         f2fs_wait_on_page_writeback(page, DATA, true, true);
1077         zero_user(page, start, len);
1078         set_page_dirty(page);
1079         f2fs_put_page(page, 1);
1080         return 0;
1081 }
1082
1083 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1084 {
1085         int err;
1086
1087         while (pg_start < pg_end) {
1088                 struct dnode_of_data dn;
1089                 pgoff_t end_offset, count;
1090
1091                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1092                 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1093                 if (err) {
1094                         if (err == -ENOENT) {
1095                                 pg_start = f2fs_get_next_page_offset(&dn,
1096                                                                 pg_start);
1097                                 continue;
1098                         }
1099                         return err;
1100                 }
1101
1102                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1103                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1104
1105                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1106
1107                 f2fs_truncate_data_blocks_range(&dn, count);
1108                 f2fs_put_dnode(&dn);
1109
1110                 pg_start += count;
1111         }
1112         return 0;
1113 }
1114
1115 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1116 {
1117         pgoff_t pg_start, pg_end;
1118         loff_t off_start, off_end;
1119         int ret;
1120
1121         ret = f2fs_convert_inline_inode(inode);
1122         if (ret)
1123                 return ret;
1124
1125         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1126         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1127
1128         off_start = offset & (PAGE_SIZE - 1);
1129         off_end = (offset + len) & (PAGE_SIZE - 1);
1130
1131         if (pg_start == pg_end) {
1132                 ret = fill_zero(inode, pg_start, off_start,
1133                                                 off_end - off_start);
1134                 if (ret)
1135                         return ret;
1136         } else {
1137                 if (off_start) {
1138                         ret = fill_zero(inode, pg_start++, off_start,
1139                                                 PAGE_SIZE - off_start);
1140                         if (ret)
1141                                 return ret;
1142                 }
1143                 if (off_end) {
1144                         ret = fill_zero(inode, pg_end, 0, off_end);
1145                         if (ret)
1146                                 return ret;
1147                 }
1148
1149                 if (pg_start < pg_end) {
1150                         loff_t blk_start, blk_end;
1151                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1152
1153                         f2fs_balance_fs(sbi, true);
1154
1155                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
1156                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
1157
1158                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1159                         filemap_invalidate_lock(inode->i_mapping);
1160
1161                         truncate_pagecache_range(inode, blk_start, blk_end - 1);
1162
1163                         f2fs_lock_op(sbi);
1164                         ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1165                         f2fs_unlock_op(sbi);
1166
1167                         filemap_invalidate_unlock(inode->i_mapping);
1168                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1169                 }
1170         }
1171
1172         return ret;
1173 }
1174
1175 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1176                                 int *do_replace, pgoff_t off, pgoff_t len)
1177 {
1178         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1179         struct dnode_of_data dn;
1180         int ret, done, i;
1181
1182 next_dnode:
1183         set_new_dnode(&dn, inode, NULL, NULL, 0);
1184         ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1185         if (ret && ret != -ENOENT) {
1186                 return ret;
1187         } else if (ret == -ENOENT) {
1188                 if (dn.max_level == 0)
1189                         return -ENOENT;
1190                 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1191                                                 dn.ofs_in_node, len);
1192                 blkaddr += done;
1193                 do_replace += done;
1194                 goto next;
1195         }
1196
1197         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1198                                                         dn.ofs_in_node, len);
1199         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1200                 *blkaddr = f2fs_data_blkaddr(&dn);
1201
1202                 if (__is_valid_data_blkaddr(*blkaddr) &&
1203                         !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1204                                         DATA_GENERIC_ENHANCE)) {
1205                         f2fs_put_dnode(&dn);
1206                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1207                         return -EFSCORRUPTED;
1208                 }
1209
1210                 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1211
1212                         if (f2fs_lfs_mode(sbi)) {
1213                                 f2fs_put_dnode(&dn);
1214                                 return -EOPNOTSUPP;
1215                         }
1216
1217                         /* do not invalidate this block address */
1218                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1219                         *do_replace = 1;
1220                 }
1221         }
1222         f2fs_put_dnode(&dn);
1223 next:
1224         len -= done;
1225         off += done;
1226         if (len)
1227                 goto next_dnode;
1228         return 0;
1229 }
1230
1231 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1232                                 int *do_replace, pgoff_t off, int len)
1233 {
1234         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1235         struct dnode_of_data dn;
1236         int ret, i;
1237
1238         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1239                 if (*do_replace == 0)
1240                         continue;
1241
1242                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1243                 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1244                 if (ret) {
1245                         dec_valid_block_count(sbi, inode, 1);
1246                         f2fs_invalidate_blocks(sbi, *blkaddr);
1247                 } else {
1248                         f2fs_update_data_blkaddr(&dn, *blkaddr);
1249                 }
1250                 f2fs_put_dnode(&dn);
1251         }
1252         return 0;
1253 }
1254
1255 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1256                         block_t *blkaddr, int *do_replace,
1257                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1258 {
1259         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1260         pgoff_t i = 0;
1261         int ret;
1262
1263         while (i < len) {
1264                 if (blkaddr[i] == NULL_ADDR && !full) {
1265                         i++;
1266                         continue;
1267                 }
1268
1269                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1270                         struct dnode_of_data dn;
1271                         struct node_info ni;
1272                         size_t new_size;
1273                         pgoff_t ilen;
1274
1275                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1276                         ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1277                         if (ret)
1278                                 return ret;
1279
1280                         ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1281                         if (ret) {
1282                                 f2fs_put_dnode(&dn);
1283                                 return ret;
1284                         }
1285
1286                         ilen = min((pgoff_t)
1287                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1288                                                 dn.ofs_in_node, len - i);
1289                         do {
1290                                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1291                                 f2fs_truncate_data_blocks_range(&dn, 1);
1292
1293                                 if (do_replace[i]) {
1294                                         f2fs_i_blocks_write(src_inode,
1295                                                         1, false, false);
1296                                         f2fs_i_blocks_write(dst_inode,
1297                                                         1, true, false);
1298                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1299                                         blkaddr[i], ni.version, true, false);
1300
1301                                         do_replace[i] = 0;
1302                                 }
1303                                 dn.ofs_in_node++;
1304                                 i++;
1305                                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1306                                 if (dst_inode->i_size < new_size)
1307                                         f2fs_i_size_write(dst_inode, new_size);
1308                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1309
1310                         f2fs_put_dnode(&dn);
1311                 } else {
1312                         struct page *psrc, *pdst;
1313
1314                         psrc = f2fs_get_lock_data_page(src_inode,
1315                                                         src + i, true);
1316                         if (IS_ERR(psrc))
1317                                 return PTR_ERR(psrc);
1318                         pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1319                                                                 true);
1320                         if (IS_ERR(pdst)) {
1321                                 f2fs_put_page(psrc, 1);
1322                                 return PTR_ERR(pdst);
1323                         }
1324                         memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1325                         set_page_dirty(pdst);
1326                         f2fs_put_page(pdst, 1);
1327                         f2fs_put_page(psrc, 1);
1328
1329                         ret = f2fs_truncate_hole(src_inode,
1330                                                 src + i, src + i + 1);
1331                         if (ret)
1332                                 return ret;
1333                         i++;
1334                 }
1335         }
1336         return 0;
1337 }
1338
1339 static int __exchange_data_block(struct inode *src_inode,
1340                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1341                         pgoff_t len, bool full)
1342 {
1343         block_t *src_blkaddr;
1344         int *do_replace;
1345         pgoff_t olen;
1346         int ret;
1347
1348         while (len) {
1349                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1350
1351                 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1352                                         array_size(olen, sizeof(block_t)),
1353                                         GFP_NOFS);
1354                 if (!src_blkaddr)
1355                         return -ENOMEM;
1356
1357                 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1358                                         array_size(olen, sizeof(int)),
1359                                         GFP_NOFS);
1360                 if (!do_replace) {
1361                         kvfree(src_blkaddr);
1362                         return -ENOMEM;
1363                 }
1364
1365                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1366                                         do_replace, src, olen);
1367                 if (ret)
1368                         goto roll_back;
1369
1370                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1371                                         do_replace, src, dst, olen, full);
1372                 if (ret)
1373                         goto roll_back;
1374
1375                 src += olen;
1376                 dst += olen;
1377                 len -= olen;
1378
1379                 kvfree(src_blkaddr);
1380                 kvfree(do_replace);
1381         }
1382         return 0;
1383
1384 roll_back:
1385         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1386         kvfree(src_blkaddr);
1387         kvfree(do_replace);
1388         return ret;
1389 }
1390
1391 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1392 {
1393         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1394         pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1395         pgoff_t start = offset >> PAGE_SHIFT;
1396         pgoff_t end = (offset + len) >> PAGE_SHIFT;
1397         int ret;
1398
1399         f2fs_balance_fs(sbi, true);
1400
1401         /* avoid gc operation during block exchange */
1402         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1403         filemap_invalidate_lock(inode->i_mapping);
1404
1405         f2fs_lock_op(sbi);
1406         f2fs_drop_extent_tree(inode);
1407         truncate_pagecache(inode, offset);
1408         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1409         f2fs_unlock_op(sbi);
1410
1411         filemap_invalidate_unlock(inode->i_mapping);
1412         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1413         return ret;
1414 }
1415
1416 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1417 {
1418         loff_t new_size;
1419         int ret;
1420
1421         if (offset + len >= i_size_read(inode))
1422                 return -EINVAL;
1423
1424         /* collapse range should be aligned to block size of f2fs. */
1425         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1426                 return -EINVAL;
1427
1428         ret = f2fs_convert_inline_inode(inode);
1429         if (ret)
1430                 return ret;
1431
1432         /* write out all dirty pages from offset */
1433         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1434         if (ret)
1435                 return ret;
1436
1437         ret = f2fs_do_collapse(inode, offset, len);
1438         if (ret)
1439                 return ret;
1440
1441         /* write out all moved pages, if possible */
1442         filemap_invalidate_lock(inode->i_mapping);
1443         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1444         truncate_pagecache(inode, offset);
1445
1446         new_size = i_size_read(inode) - len;
1447         ret = f2fs_truncate_blocks(inode, new_size, true);
1448         filemap_invalidate_unlock(inode->i_mapping);
1449         if (!ret)
1450                 f2fs_i_size_write(inode, new_size);
1451         return ret;
1452 }
1453
1454 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1455                                                                 pgoff_t end)
1456 {
1457         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1458         pgoff_t index = start;
1459         unsigned int ofs_in_node = dn->ofs_in_node;
1460         blkcnt_t count = 0;
1461         int ret;
1462
1463         for (; index < end; index++, dn->ofs_in_node++) {
1464                 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1465                         count++;
1466         }
1467
1468         dn->ofs_in_node = ofs_in_node;
1469         ret = f2fs_reserve_new_blocks(dn, count);
1470         if (ret)
1471                 return ret;
1472
1473         dn->ofs_in_node = ofs_in_node;
1474         for (index = start; index < end; index++, dn->ofs_in_node++) {
1475                 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1476                 /*
1477                  * f2fs_reserve_new_blocks will not guarantee entire block
1478                  * allocation.
1479                  */
1480                 if (dn->data_blkaddr == NULL_ADDR) {
1481                         ret = -ENOSPC;
1482                         break;
1483                 }
1484
1485                 if (dn->data_blkaddr == NEW_ADDR)
1486                         continue;
1487
1488                 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1489                                         DATA_GENERIC_ENHANCE)) {
1490                         ret = -EFSCORRUPTED;
1491                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1492                         break;
1493                 }
1494
1495                 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1496                 dn->data_blkaddr = NEW_ADDR;
1497                 f2fs_set_data_blkaddr(dn);
1498         }
1499
1500         f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1501
1502         return ret;
1503 }
1504
1505 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1506                                                                 int mode)
1507 {
1508         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1509         struct address_space *mapping = inode->i_mapping;
1510         pgoff_t index, pg_start, pg_end;
1511         loff_t new_size = i_size_read(inode);
1512         loff_t off_start, off_end;
1513         int ret = 0;
1514
1515         ret = inode_newsize_ok(inode, (len + offset));
1516         if (ret)
1517                 return ret;
1518
1519         ret = f2fs_convert_inline_inode(inode);
1520         if (ret)
1521                 return ret;
1522
1523         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1524         if (ret)
1525                 return ret;
1526
1527         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1528         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1529
1530         off_start = offset & (PAGE_SIZE - 1);
1531         off_end = (offset + len) & (PAGE_SIZE - 1);
1532
1533         if (pg_start == pg_end) {
1534                 ret = fill_zero(inode, pg_start, off_start,
1535                                                 off_end - off_start);
1536                 if (ret)
1537                         return ret;
1538
1539                 new_size = max_t(loff_t, new_size, offset + len);
1540         } else {
1541                 if (off_start) {
1542                         ret = fill_zero(inode, pg_start++, off_start,
1543                                                 PAGE_SIZE - off_start);
1544                         if (ret)
1545                                 return ret;
1546
1547                         new_size = max_t(loff_t, new_size,
1548                                         (loff_t)pg_start << PAGE_SHIFT);
1549                 }
1550
1551                 for (index = pg_start; index < pg_end;) {
1552                         struct dnode_of_data dn;
1553                         unsigned int end_offset;
1554                         pgoff_t end;
1555
1556                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1557                         filemap_invalidate_lock(mapping);
1558
1559                         truncate_pagecache_range(inode,
1560                                 (loff_t)index << PAGE_SHIFT,
1561                                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1562
1563                         f2fs_lock_op(sbi);
1564
1565                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1566                         ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1567                         if (ret) {
1568                                 f2fs_unlock_op(sbi);
1569                                 filemap_invalidate_unlock(mapping);
1570                                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1571                                 goto out;
1572                         }
1573
1574                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1575                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1576
1577                         ret = f2fs_do_zero_range(&dn, index, end);
1578                         f2fs_put_dnode(&dn);
1579
1580                         f2fs_unlock_op(sbi);
1581                         filemap_invalidate_unlock(mapping);
1582                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1583
1584                         f2fs_balance_fs(sbi, dn.node_changed);
1585
1586                         if (ret)
1587                                 goto out;
1588
1589                         index = end;
1590                         new_size = max_t(loff_t, new_size,
1591                                         (loff_t)index << PAGE_SHIFT);
1592                 }
1593
1594                 if (off_end) {
1595                         ret = fill_zero(inode, pg_end, 0, off_end);
1596                         if (ret)
1597                                 goto out;
1598
1599                         new_size = max_t(loff_t, new_size, offset + len);
1600                 }
1601         }
1602
1603 out:
1604         if (new_size > i_size_read(inode)) {
1605                 if (mode & FALLOC_FL_KEEP_SIZE)
1606                         file_set_keep_isize(inode);
1607                 else
1608                         f2fs_i_size_write(inode, new_size);
1609         }
1610         return ret;
1611 }
1612
1613 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1614 {
1615         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1616         struct address_space *mapping = inode->i_mapping;
1617         pgoff_t nr, pg_start, pg_end, delta, idx;
1618         loff_t new_size;
1619         int ret = 0;
1620
1621         new_size = i_size_read(inode) + len;
1622         ret = inode_newsize_ok(inode, new_size);
1623         if (ret)
1624                 return ret;
1625
1626         if (offset >= i_size_read(inode))
1627                 return -EINVAL;
1628
1629         /* insert range should be aligned to block size of f2fs. */
1630         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1631                 return -EINVAL;
1632
1633         ret = f2fs_convert_inline_inode(inode);
1634         if (ret)
1635                 return ret;
1636
1637         f2fs_balance_fs(sbi, true);
1638
1639         filemap_invalidate_lock(mapping);
1640         ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1641         filemap_invalidate_unlock(mapping);
1642         if (ret)
1643                 return ret;
1644
1645         /* write out all dirty pages from offset */
1646         ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1647         if (ret)
1648                 return ret;
1649
1650         pg_start = offset >> PAGE_SHIFT;
1651         pg_end = (offset + len) >> PAGE_SHIFT;
1652         delta = pg_end - pg_start;
1653         idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1654
1655         /* avoid gc operation during block exchange */
1656         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1657         filemap_invalidate_lock(mapping);
1658         truncate_pagecache(inode, offset);
1659
1660         while (!ret && idx > pg_start) {
1661                 nr = idx - pg_start;
1662                 if (nr > delta)
1663                         nr = delta;
1664                 idx -= nr;
1665
1666                 f2fs_lock_op(sbi);
1667                 f2fs_drop_extent_tree(inode);
1668
1669                 ret = __exchange_data_block(inode, inode, idx,
1670                                         idx + delta, nr, false);
1671                 f2fs_unlock_op(sbi);
1672         }
1673         filemap_invalidate_unlock(mapping);
1674         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1675
1676         /* write out all moved pages, if possible */
1677         filemap_invalidate_lock(mapping);
1678         filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1679         truncate_pagecache(inode, offset);
1680         filemap_invalidate_unlock(mapping);
1681
1682         if (!ret)
1683                 f2fs_i_size_write(inode, new_size);
1684         return ret;
1685 }
1686
1687 static int expand_inode_data(struct inode *inode, loff_t offset,
1688                                         loff_t len, int mode)
1689 {
1690         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1691         struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1692                         .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1693                         .m_may_create = true };
1694         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1695                         .init_gc_type = FG_GC,
1696                         .should_migrate_blocks = false,
1697                         .err_gc_skipped = true,
1698                         .nr_free_secs = 0 };
1699         pgoff_t pg_start, pg_end;
1700         loff_t new_size = i_size_read(inode);
1701         loff_t off_end;
1702         block_t expanded = 0;
1703         int err;
1704
1705         err = inode_newsize_ok(inode, (len + offset));
1706         if (err)
1707                 return err;
1708
1709         err = f2fs_convert_inline_inode(inode);
1710         if (err)
1711                 return err;
1712
1713         f2fs_balance_fs(sbi, true);
1714
1715         pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1716         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1717         off_end = (offset + len) & (PAGE_SIZE - 1);
1718
1719         map.m_lblk = pg_start;
1720         map.m_len = pg_end - pg_start;
1721         if (off_end)
1722                 map.m_len++;
1723
1724         if (!map.m_len)
1725                 return 0;
1726
1727         if (f2fs_is_pinned_file(inode)) {
1728                 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1729                 block_t sec_len = roundup(map.m_len, sec_blks);
1730
1731                 map.m_len = sec_blks;
1732 next_alloc:
1733                 if (has_not_enough_free_secs(sbi, 0,
1734                         GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1735                         f2fs_down_write(&sbi->gc_lock);
1736                         err = f2fs_gc(sbi, &gc_control);
1737                         if (err && err != -ENODATA)
1738                                 goto out_err;
1739                 }
1740
1741                 f2fs_down_write(&sbi->pin_sem);
1742
1743                 f2fs_lock_op(sbi);
1744                 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1745                 f2fs_unlock_op(sbi);
1746
1747                 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1748                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1749                 file_dont_truncate(inode);
1750
1751                 f2fs_up_write(&sbi->pin_sem);
1752
1753                 expanded += map.m_len;
1754                 sec_len -= map.m_len;
1755                 map.m_lblk += map.m_len;
1756                 if (!err && sec_len)
1757                         goto next_alloc;
1758
1759                 map.m_len = expanded;
1760         } else {
1761                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1762                 expanded = map.m_len;
1763         }
1764 out_err:
1765         if (err) {
1766                 pgoff_t last_off;
1767
1768                 if (!expanded)
1769                         return err;
1770
1771                 last_off = pg_start + expanded - 1;
1772
1773                 /* update new size to the failed position */
1774                 new_size = (last_off == pg_end) ? offset + len :
1775                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1776         } else {
1777                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1778         }
1779
1780         if (new_size > i_size_read(inode)) {
1781                 if (mode & FALLOC_FL_KEEP_SIZE)
1782                         file_set_keep_isize(inode);
1783                 else
1784                         f2fs_i_size_write(inode, new_size);
1785         }
1786
1787         return err;
1788 }
1789
1790 static long f2fs_fallocate(struct file *file, int mode,
1791                                 loff_t offset, loff_t len)
1792 {
1793         struct inode *inode = file_inode(file);
1794         long ret = 0;
1795
1796         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1797                 return -EIO;
1798         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1799                 return -ENOSPC;
1800         if (!f2fs_is_compress_backend_ready(inode))
1801                 return -EOPNOTSUPP;
1802
1803         /* f2fs only support ->fallocate for regular file */
1804         if (!S_ISREG(inode->i_mode))
1805                 return -EINVAL;
1806
1807         if (IS_ENCRYPTED(inode) &&
1808                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1809                 return -EOPNOTSUPP;
1810
1811         /*
1812          * Pinned file should not support partial trucation since the block
1813          * can be used by applications.
1814          */
1815         if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1816                 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1817                         FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1818                 return -EOPNOTSUPP;
1819
1820         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1821                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1822                         FALLOC_FL_INSERT_RANGE))
1823                 return -EOPNOTSUPP;
1824
1825         inode_lock(inode);
1826
1827         ret = file_modified(file);
1828         if (ret)
1829                 goto out;
1830
1831         if (mode & FALLOC_FL_PUNCH_HOLE) {
1832                 if (offset >= inode->i_size)
1833                         goto out;
1834
1835                 ret = punch_hole(inode, offset, len);
1836         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1837                 ret = f2fs_collapse_range(inode, offset, len);
1838         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1839                 ret = f2fs_zero_range(inode, offset, len, mode);
1840         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1841                 ret = f2fs_insert_range(inode, offset, len);
1842         } else {
1843                 ret = expand_inode_data(inode, offset, len, mode);
1844         }
1845
1846         if (!ret) {
1847                 inode->i_mtime = inode->i_ctime = current_time(inode);
1848                 f2fs_mark_inode_dirty_sync(inode, false);
1849                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1850         }
1851
1852 out:
1853         inode_unlock(inode);
1854
1855         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1856         return ret;
1857 }
1858
1859 static int f2fs_release_file(struct inode *inode, struct file *filp)
1860 {
1861         /*
1862          * f2fs_relase_file is called at every close calls. So we should
1863          * not drop any inmemory pages by close called by other process.
1864          */
1865         if (!(filp->f_mode & FMODE_WRITE) ||
1866                         atomic_read(&inode->i_writecount) != 1)
1867                 return 0;
1868
1869         f2fs_abort_atomic_write(inode, true);
1870         return 0;
1871 }
1872
1873 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1874 {
1875         struct inode *inode = file_inode(file);
1876
1877         /*
1878          * If the process doing a transaction is crashed, we should do
1879          * roll-back. Otherwise, other reader/write can see corrupted database
1880          * until all the writers close its file. Since this should be done
1881          * before dropping file lock, it needs to do in ->flush.
1882          */
1883         if (F2FS_I(inode)->atomic_write_task == current)
1884                 f2fs_abort_atomic_write(inode, true);
1885         return 0;
1886 }
1887
1888 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1889 {
1890         struct f2fs_inode_info *fi = F2FS_I(inode);
1891         u32 masked_flags = fi->i_flags & mask;
1892
1893         /* mask can be shrunk by flags_valid selector */
1894         iflags &= mask;
1895
1896         /* Is it quota file? Do not allow user to mess with it */
1897         if (IS_NOQUOTA(inode))
1898                 return -EPERM;
1899
1900         if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1901                 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1902                         return -EOPNOTSUPP;
1903                 if (!f2fs_empty_dir(inode))
1904                         return -ENOTEMPTY;
1905         }
1906
1907         if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1908                 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1909                         return -EOPNOTSUPP;
1910                 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1911                         return -EINVAL;
1912         }
1913
1914         if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1915                 if (masked_flags & F2FS_COMPR_FL) {
1916                         if (!f2fs_disable_compressed_file(inode))
1917                                 return -EINVAL;
1918                 } else {
1919                         /* try to convert inline_data to support compression */
1920                         int err = f2fs_convert_inline_inode(inode);
1921                         if (err)
1922                                 return err;
1923                         if (!f2fs_may_compress(inode))
1924                                 return -EINVAL;
1925                         if (S_ISREG(inode->i_mode) && F2FS_HAS_BLOCKS(inode))
1926                                 return -EINVAL;
1927                         if (set_compress_context(inode))
1928                                 return -EOPNOTSUPP;
1929                 }
1930         }
1931
1932         fi->i_flags = iflags | (fi->i_flags & ~mask);
1933         f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1934                                         (fi->i_flags & F2FS_NOCOMP_FL));
1935
1936         if (fi->i_flags & F2FS_PROJINHERIT_FL)
1937                 set_inode_flag(inode, FI_PROJ_INHERIT);
1938         else
1939                 clear_inode_flag(inode, FI_PROJ_INHERIT);
1940
1941         inode->i_ctime = current_time(inode);
1942         f2fs_set_inode_flags(inode);
1943         f2fs_mark_inode_dirty_sync(inode, true);
1944         return 0;
1945 }
1946
1947 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1948
1949 /*
1950  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1951  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1952  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1953  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1954  *
1955  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1956  * FS_IOC_FSSETXATTR is done by the VFS.
1957  */
1958
1959 static const struct {
1960         u32 iflag;
1961         u32 fsflag;
1962 } f2fs_fsflags_map[] = {
1963         { F2FS_COMPR_FL,        FS_COMPR_FL },
1964         { F2FS_SYNC_FL,         FS_SYNC_FL },
1965         { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1966         { F2FS_APPEND_FL,       FS_APPEND_FL },
1967         { F2FS_NODUMP_FL,       FS_NODUMP_FL },
1968         { F2FS_NOATIME_FL,      FS_NOATIME_FL },
1969         { F2FS_NOCOMP_FL,       FS_NOCOMP_FL },
1970         { F2FS_INDEX_FL,        FS_INDEX_FL },
1971         { F2FS_DIRSYNC_FL,      FS_DIRSYNC_FL },
1972         { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1973         { F2FS_CASEFOLD_FL,     FS_CASEFOLD_FL },
1974 };
1975
1976 #define F2FS_GETTABLE_FS_FL (           \
1977                 FS_COMPR_FL |           \
1978                 FS_SYNC_FL |            \
1979                 FS_IMMUTABLE_FL |       \
1980                 FS_APPEND_FL |          \
1981                 FS_NODUMP_FL |          \
1982                 FS_NOATIME_FL |         \
1983                 FS_NOCOMP_FL |          \
1984                 FS_INDEX_FL |           \
1985                 FS_DIRSYNC_FL |         \
1986                 FS_PROJINHERIT_FL |     \
1987                 FS_ENCRYPT_FL |         \
1988                 FS_INLINE_DATA_FL |     \
1989                 FS_NOCOW_FL |           \
1990                 FS_VERITY_FL |          \
1991                 FS_CASEFOLD_FL)
1992
1993 #define F2FS_SETTABLE_FS_FL (           \
1994                 FS_COMPR_FL |           \
1995                 FS_SYNC_FL |            \
1996                 FS_IMMUTABLE_FL |       \
1997                 FS_APPEND_FL |          \
1998                 FS_NODUMP_FL |          \
1999                 FS_NOATIME_FL |         \
2000                 FS_NOCOMP_FL |          \
2001                 FS_DIRSYNC_FL |         \
2002                 FS_PROJINHERIT_FL |     \
2003                 FS_CASEFOLD_FL)
2004
2005 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2006 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2007 {
2008         u32 fsflags = 0;
2009         int i;
2010
2011         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2012                 if (iflags & f2fs_fsflags_map[i].iflag)
2013                         fsflags |= f2fs_fsflags_map[i].fsflag;
2014
2015         return fsflags;
2016 }
2017
2018 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2019 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2020 {
2021         u32 iflags = 0;
2022         int i;
2023
2024         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2025                 if (fsflags & f2fs_fsflags_map[i].fsflag)
2026                         iflags |= f2fs_fsflags_map[i].iflag;
2027
2028         return iflags;
2029 }
2030
2031 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2032 {
2033         struct inode *inode = file_inode(filp);
2034
2035         return put_user(inode->i_generation, (int __user *)arg);
2036 }
2037
2038 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2039 {
2040         struct inode *inode = file_inode(filp);
2041         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2042         struct f2fs_inode_info *fi = F2FS_I(inode);
2043         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2044         struct inode *pinode;
2045         loff_t isize;
2046         int ret;
2047
2048         if (!inode_owner_or_capable(mnt_userns, inode))
2049                 return -EACCES;
2050
2051         if (!S_ISREG(inode->i_mode))
2052                 return -EINVAL;
2053
2054         if (filp->f_flags & O_DIRECT)
2055                 return -EINVAL;
2056
2057         ret = mnt_want_write_file(filp);
2058         if (ret)
2059                 return ret;
2060
2061         inode_lock(inode);
2062
2063         if (!f2fs_disable_compressed_file(inode)) {
2064                 ret = -EINVAL;
2065                 goto out;
2066         }
2067
2068         if (f2fs_is_atomic_file(inode))
2069                 goto out;
2070
2071         ret = f2fs_convert_inline_inode(inode);
2072         if (ret)
2073                 goto out;
2074
2075         f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2076
2077         /*
2078          * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2079          * f2fs_is_atomic_file.
2080          */
2081         if (get_dirty_pages(inode))
2082                 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2083                           inode->i_ino, get_dirty_pages(inode));
2084         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2085         if (ret) {
2086                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2087                 goto out;
2088         }
2089
2090         /* Create a COW inode for atomic write */
2091         pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2092         if (IS_ERR(pinode)) {
2093                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2094                 ret = PTR_ERR(pinode);
2095                 goto out;
2096         }
2097
2098         ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2099         iput(pinode);
2100         if (ret) {
2101                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2102                 goto out;
2103         }
2104
2105         f2fs_write_inode(inode, NULL);
2106
2107         stat_inc_atomic_inode(inode);
2108
2109         set_inode_flag(inode, FI_ATOMIC_FILE);
2110         set_inode_flag(fi->cow_inode, FI_COW_FILE);
2111         clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2112
2113         isize = i_size_read(inode);
2114         fi->original_i_size = isize;
2115         if (truncate) {
2116                 set_inode_flag(inode, FI_ATOMIC_REPLACE);
2117                 truncate_inode_pages_final(inode->i_mapping);
2118                 f2fs_i_size_write(inode, 0);
2119                 isize = 0;
2120         }
2121         f2fs_i_size_write(fi->cow_inode, isize);
2122
2123         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2124
2125         f2fs_update_time(sbi, REQ_TIME);
2126         fi->atomic_write_task = current;
2127         stat_update_max_atomic_write(inode);
2128         fi->atomic_write_cnt = 0;
2129 out:
2130         inode_unlock(inode);
2131         mnt_drop_write_file(filp);
2132         return ret;
2133 }
2134
2135 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2136 {
2137         struct inode *inode = file_inode(filp);
2138         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2139         int ret;
2140
2141         if (!inode_owner_or_capable(mnt_userns, inode))
2142                 return -EACCES;
2143
2144         ret = mnt_want_write_file(filp);
2145         if (ret)
2146                 return ret;
2147
2148         f2fs_balance_fs(F2FS_I_SB(inode), true);
2149
2150         inode_lock(inode);
2151
2152         if (f2fs_is_atomic_file(inode)) {
2153                 ret = f2fs_commit_atomic_write(inode);
2154                 if (!ret)
2155                         ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2156
2157                 f2fs_abort_atomic_write(inode, ret);
2158         } else {
2159                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2160         }
2161
2162         inode_unlock(inode);
2163         mnt_drop_write_file(filp);
2164         return ret;
2165 }
2166
2167 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2168 {
2169         struct inode *inode = file_inode(filp);
2170         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2171         int ret;
2172
2173         if (!inode_owner_or_capable(mnt_userns, inode))
2174                 return -EACCES;
2175
2176         ret = mnt_want_write_file(filp);
2177         if (ret)
2178                 return ret;
2179
2180         inode_lock(inode);
2181
2182         f2fs_abort_atomic_write(inode, true);
2183
2184         inode_unlock(inode);
2185
2186         mnt_drop_write_file(filp);
2187         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2188         return ret;
2189 }
2190
2191 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2192 {
2193         struct inode *inode = file_inode(filp);
2194         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2195         struct super_block *sb = sbi->sb;
2196         __u32 in;
2197         int ret = 0;
2198
2199         if (!capable(CAP_SYS_ADMIN))
2200                 return -EPERM;
2201
2202         if (get_user(in, (__u32 __user *)arg))
2203                 return -EFAULT;
2204
2205         if (in != F2FS_GOING_DOWN_FULLSYNC) {
2206                 ret = mnt_want_write_file(filp);
2207                 if (ret) {
2208                         if (ret == -EROFS) {
2209                                 ret = 0;
2210                                 f2fs_stop_checkpoint(sbi, false,
2211                                                 STOP_CP_REASON_SHUTDOWN);
2212                                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2213                                 trace_f2fs_shutdown(sbi, in, ret);
2214                         }
2215                         return ret;
2216                 }
2217         }
2218
2219         switch (in) {
2220         case F2FS_GOING_DOWN_FULLSYNC:
2221                 ret = freeze_bdev(sb->s_bdev);
2222                 if (ret)
2223                         goto out;
2224                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2225                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2226                 thaw_bdev(sb->s_bdev);
2227                 break;
2228         case F2FS_GOING_DOWN_METASYNC:
2229                 /* do checkpoint only */
2230                 ret = f2fs_sync_fs(sb, 1);
2231                 if (ret)
2232                         goto out;
2233                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2234                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2235                 break;
2236         case F2FS_GOING_DOWN_NOSYNC:
2237                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2238                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2239                 break;
2240         case F2FS_GOING_DOWN_METAFLUSH:
2241                 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2242                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2243                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2244                 break;
2245         case F2FS_GOING_DOWN_NEED_FSCK:
2246                 set_sbi_flag(sbi, SBI_NEED_FSCK);
2247                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2248                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2249                 /* do checkpoint only */
2250                 ret = f2fs_sync_fs(sb, 1);
2251                 goto out;
2252         default:
2253                 ret = -EINVAL;
2254                 goto out;
2255         }
2256
2257         f2fs_stop_gc_thread(sbi);
2258         f2fs_stop_discard_thread(sbi);
2259
2260         f2fs_drop_discard_cmd(sbi);
2261         clear_opt(sbi, DISCARD);
2262
2263         f2fs_update_time(sbi, REQ_TIME);
2264 out:
2265         if (in != F2FS_GOING_DOWN_FULLSYNC)
2266                 mnt_drop_write_file(filp);
2267
2268         trace_f2fs_shutdown(sbi, in, ret);
2269
2270         return ret;
2271 }
2272
2273 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2274 {
2275         struct inode *inode = file_inode(filp);
2276         struct super_block *sb = inode->i_sb;
2277         struct fstrim_range range;
2278         int ret;
2279
2280         if (!capable(CAP_SYS_ADMIN))
2281                 return -EPERM;
2282
2283         if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2284                 return -EOPNOTSUPP;
2285
2286         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2287                                 sizeof(range)))
2288                 return -EFAULT;
2289
2290         ret = mnt_want_write_file(filp);
2291         if (ret)
2292                 return ret;
2293
2294         range.minlen = max((unsigned int)range.minlen,
2295                            bdev_discard_granularity(sb->s_bdev));
2296         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2297         mnt_drop_write_file(filp);
2298         if (ret < 0)
2299                 return ret;
2300
2301         if (copy_to_user((struct fstrim_range __user *)arg, &range,
2302                                 sizeof(range)))
2303                 return -EFAULT;
2304         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2305         return 0;
2306 }
2307
2308 static bool uuid_is_nonzero(__u8 u[16])
2309 {
2310         int i;
2311
2312         for (i = 0; i < 16; i++)
2313                 if (u[i])
2314                         return true;
2315         return false;
2316 }
2317
2318 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2319 {
2320         struct inode *inode = file_inode(filp);
2321
2322         if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2323                 return -EOPNOTSUPP;
2324
2325         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2326
2327         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2328 }
2329
2330 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2331 {
2332         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2333                 return -EOPNOTSUPP;
2334         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2335 }
2336
2337 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2338 {
2339         struct inode *inode = file_inode(filp);
2340         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2341         int err;
2342
2343         if (!f2fs_sb_has_encrypt(sbi))
2344                 return -EOPNOTSUPP;
2345
2346         err = mnt_want_write_file(filp);
2347         if (err)
2348                 return err;
2349
2350         f2fs_down_write(&sbi->sb_lock);
2351
2352         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2353                 goto got_it;
2354
2355         /* update superblock with uuid */
2356         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2357
2358         err = f2fs_commit_super(sbi, false);
2359         if (err) {
2360                 /* undo new data */
2361                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2362                 goto out_err;
2363         }
2364 got_it:
2365         if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2366                                                                         16))
2367                 err = -EFAULT;
2368 out_err:
2369         f2fs_up_write(&sbi->sb_lock);
2370         mnt_drop_write_file(filp);
2371         return err;
2372 }
2373
2374 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2375                                              unsigned long arg)
2376 {
2377         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2378                 return -EOPNOTSUPP;
2379
2380         return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2381 }
2382
2383 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2384 {
2385         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2386                 return -EOPNOTSUPP;
2387
2388         return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2389 }
2390
2391 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2392 {
2393         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2394                 return -EOPNOTSUPP;
2395
2396         return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2397 }
2398
2399 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2400                                                     unsigned long arg)
2401 {
2402         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2403                 return -EOPNOTSUPP;
2404
2405         return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2406 }
2407
2408 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2409                                               unsigned long arg)
2410 {
2411         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2412                 return -EOPNOTSUPP;
2413
2414         return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2415 }
2416
2417 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2418 {
2419         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2420                 return -EOPNOTSUPP;
2421
2422         return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2423 }
2424
2425 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2426 {
2427         struct inode *inode = file_inode(filp);
2428         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2429         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2430                         .no_bg_gc = false,
2431                         .should_migrate_blocks = false,
2432                         .nr_free_secs = 0 };
2433         __u32 sync;
2434         int ret;
2435
2436         if (!capable(CAP_SYS_ADMIN))
2437                 return -EPERM;
2438
2439         if (get_user(sync, (__u32 __user *)arg))
2440                 return -EFAULT;
2441
2442         if (f2fs_readonly(sbi->sb))
2443                 return -EROFS;
2444
2445         ret = mnt_want_write_file(filp);
2446         if (ret)
2447                 return ret;
2448
2449         if (!sync) {
2450                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2451                         ret = -EBUSY;
2452                         goto out;
2453                 }
2454         } else {
2455                 f2fs_down_write(&sbi->gc_lock);
2456         }
2457
2458         gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2459         gc_control.err_gc_skipped = sync;
2460         ret = f2fs_gc(sbi, &gc_control);
2461 out:
2462         mnt_drop_write_file(filp);
2463         return ret;
2464 }
2465
2466 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2467 {
2468         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2469         struct f2fs_gc_control gc_control = {
2470                         .init_gc_type = range->sync ? FG_GC : BG_GC,
2471                         .no_bg_gc = false,
2472                         .should_migrate_blocks = false,
2473                         .err_gc_skipped = range->sync,
2474                         .nr_free_secs = 0 };
2475         u64 end;
2476         int ret;
2477
2478         if (!capable(CAP_SYS_ADMIN))
2479                 return -EPERM;
2480         if (f2fs_readonly(sbi->sb))
2481                 return -EROFS;
2482
2483         end = range->start + range->len;
2484         if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2485                                         end >= MAX_BLKADDR(sbi))
2486                 return -EINVAL;
2487
2488         ret = mnt_want_write_file(filp);
2489         if (ret)
2490                 return ret;
2491
2492 do_more:
2493         if (!range->sync) {
2494                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2495                         ret = -EBUSY;
2496                         goto out;
2497                 }
2498         } else {
2499                 f2fs_down_write(&sbi->gc_lock);
2500         }
2501
2502         gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2503         ret = f2fs_gc(sbi, &gc_control);
2504         if (ret) {
2505                 if (ret == -EBUSY)
2506                         ret = -EAGAIN;
2507                 goto out;
2508         }
2509         range->start += CAP_BLKS_PER_SEC(sbi);
2510         if (range->start <= end)
2511                 goto do_more;
2512 out:
2513         mnt_drop_write_file(filp);
2514         return ret;
2515 }
2516
2517 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2518 {
2519         struct f2fs_gc_range range;
2520
2521         if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2522                                                         sizeof(range)))
2523                 return -EFAULT;
2524         return __f2fs_ioc_gc_range(filp, &range);
2525 }
2526
2527 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2528 {
2529         struct inode *inode = file_inode(filp);
2530         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2531         int ret;
2532
2533         if (!capable(CAP_SYS_ADMIN))
2534                 return -EPERM;
2535
2536         if (f2fs_readonly(sbi->sb))
2537                 return -EROFS;
2538
2539         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2540                 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2541                 return -EINVAL;
2542         }
2543
2544         ret = mnt_want_write_file(filp);
2545         if (ret)
2546                 return ret;
2547
2548         ret = f2fs_sync_fs(sbi->sb, 1);
2549
2550         mnt_drop_write_file(filp);
2551         return ret;
2552 }
2553
2554 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2555                                         struct file *filp,
2556                                         struct f2fs_defragment *range)
2557 {
2558         struct inode *inode = file_inode(filp);
2559         struct f2fs_map_blocks map = { .m_next_extent = NULL,
2560                                         .m_seg_type = NO_CHECK_TYPE,
2561                                         .m_may_create = false };
2562         struct extent_info ei = {};
2563         pgoff_t pg_start, pg_end, next_pgofs;
2564         unsigned int blk_per_seg = sbi->blocks_per_seg;
2565         unsigned int total = 0, sec_num;
2566         block_t blk_end = 0;
2567         bool fragmented = false;
2568         int err;
2569
2570         pg_start = range->start >> PAGE_SHIFT;
2571         pg_end = (range->start + range->len) >> PAGE_SHIFT;
2572
2573         f2fs_balance_fs(sbi, true);
2574
2575         inode_lock(inode);
2576
2577         /* if in-place-update policy is enabled, don't waste time here */
2578         set_inode_flag(inode, FI_OPU_WRITE);
2579         if (f2fs_should_update_inplace(inode, NULL)) {
2580                 err = -EINVAL;
2581                 goto out;
2582         }
2583
2584         /* writeback all dirty pages in the range */
2585         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2586                                                 range->start + range->len - 1);
2587         if (err)
2588                 goto out;
2589
2590         /*
2591          * lookup mapping info in extent cache, skip defragmenting if physical
2592          * block addresses are continuous.
2593          */
2594         if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2595                 if (ei.fofs + ei.len >= pg_end)
2596                         goto out;
2597         }
2598
2599         map.m_lblk = pg_start;
2600         map.m_next_pgofs = &next_pgofs;
2601
2602         /*
2603          * lookup mapping info in dnode page cache, skip defragmenting if all
2604          * physical block addresses are continuous even if there are hole(s)
2605          * in logical blocks.
2606          */
2607         while (map.m_lblk < pg_end) {
2608                 map.m_len = pg_end - map.m_lblk;
2609                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2610                 if (err)
2611                         goto out;
2612
2613                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2614                         map.m_lblk = next_pgofs;
2615                         continue;
2616                 }
2617
2618                 if (blk_end && blk_end != map.m_pblk)
2619                         fragmented = true;
2620
2621                 /* record total count of block that we're going to move */
2622                 total += map.m_len;
2623
2624                 blk_end = map.m_pblk + map.m_len;
2625
2626                 map.m_lblk += map.m_len;
2627         }
2628
2629         if (!fragmented) {
2630                 total = 0;
2631                 goto out;
2632         }
2633
2634         sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2635
2636         /*
2637          * make sure there are enough free section for LFS allocation, this can
2638          * avoid defragment running in SSR mode when free section are allocated
2639          * intensively
2640          */
2641         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2642                 err = -EAGAIN;
2643                 goto out;
2644         }
2645
2646         map.m_lblk = pg_start;
2647         map.m_len = pg_end - pg_start;
2648         total = 0;
2649
2650         while (map.m_lblk < pg_end) {
2651                 pgoff_t idx;
2652                 int cnt = 0;
2653
2654 do_map:
2655                 map.m_len = pg_end - map.m_lblk;
2656                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2657                 if (err)
2658                         goto clear_out;
2659
2660                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2661                         map.m_lblk = next_pgofs;
2662                         goto check;
2663                 }
2664
2665                 set_inode_flag(inode, FI_SKIP_WRITES);
2666
2667                 idx = map.m_lblk;
2668                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2669                         struct page *page;
2670
2671                         page = f2fs_get_lock_data_page(inode, idx, true);
2672                         if (IS_ERR(page)) {
2673                                 err = PTR_ERR(page);
2674                                 goto clear_out;
2675                         }
2676
2677                         set_page_dirty(page);
2678                         set_page_private_gcing(page);
2679                         f2fs_put_page(page, 1);
2680
2681                         idx++;
2682                         cnt++;
2683                         total++;
2684                 }
2685
2686                 map.m_lblk = idx;
2687 check:
2688                 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2689                         goto do_map;
2690
2691                 clear_inode_flag(inode, FI_SKIP_WRITES);
2692
2693                 err = filemap_fdatawrite(inode->i_mapping);
2694                 if (err)
2695                         goto out;
2696         }
2697 clear_out:
2698         clear_inode_flag(inode, FI_SKIP_WRITES);
2699 out:
2700         clear_inode_flag(inode, FI_OPU_WRITE);
2701         inode_unlock(inode);
2702         if (!err)
2703                 range->len = (u64)total << PAGE_SHIFT;
2704         return err;
2705 }
2706
2707 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2708 {
2709         struct inode *inode = file_inode(filp);
2710         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2711         struct f2fs_defragment range;
2712         int err;
2713
2714         if (!capable(CAP_SYS_ADMIN))
2715                 return -EPERM;
2716
2717         if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2718                 return -EINVAL;
2719
2720         if (f2fs_readonly(sbi->sb))
2721                 return -EROFS;
2722
2723         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2724                                                         sizeof(range)))
2725                 return -EFAULT;
2726
2727         /* verify alignment of offset & size */
2728         if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2729                 return -EINVAL;
2730
2731         if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2732                                         max_file_blocks(inode)))
2733                 return -EINVAL;
2734
2735         err = mnt_want_write_file(filp);
2736         if (err)
2737                 return err;
2738
2739         err = f2fs_defragment_range(sbi, filp, &range);
2740         mnt_drop_write_file(filp);
2741
2742         f2fs_update_time(sbi, REQ_TIME);
2743         if (err < 0)
2744                 return err;
2745
2746         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2747                                                         sizeof(range)))
2748                 return -EFAULT;
2749
2750         return 0;
2751 }
2752
2753 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2754                         struct file *file_out, loff_t pos_out, size_t len)
2755 {
2756         struct inode *src = file_inode(file_in);
2757         struct inode *dst = file_inode(file_out);
2758         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2759         size_t olen = len, dst_max_i_size = 0;
2760         size_t dst_osize;
2761         int ret;
2762
2763         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2764                                 src->i_sb != dst->i_sb)
2765                 return -EXDEV;
2766
2767         if (unlikely(f2fs_readonly(src->i_sb)))
2768                 return -EROFS;
2769
2770         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2771                 return -EINVAL;
2772
2773         if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2774                 return -EOPNOTSUPP;
2775
2776         if (pos_out < 0 || pos_in < 0)
2777                 return -EINVAL;
2778
2779         if (src == dst) {
2780                 if (pos_in == pos_out)
2781                         return 0;
2782                 if (pos_out > pos_in && pos_out < pos_in + len)
2783                         return -EINVAL;
2784         }
2785
2786         inode_lock(src);
2787         if (src != dst) {
2788                 ret = -EBUSY;
2789                 if (!inode_trylock(dst))
2790                         goto out;
2791         }
2792
2793         ret = -EINVAL;
2794         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2795                 goto out_unlock;
2796         if (len == 0)
2797                 olen = len = src->i_size - pos_in;
2798         if (pos_in + len == src->i_size)
2799                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2800         if (len == 0) {
2801                 ret = 0;
2802                 goto out_unlock;
2803         }
2804
2805         dst_osize = dst->i_size;
2806         if (pos_out + olen > dst->i_size)
2807                 dst_max_i_size = pos_out + olen;
2808
2809         /* verify the end result is block aligned */
2810         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2811                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2812                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2813                 goto out_unlock;
2814
2815         ret = f2fs_convert_inline_inode(src);
2816         if (ret)
2817                 goto out_unlock;
2818
2819         ret = f2fs_convert_inline_inode(dst);
2820         if (ret)
2821                 goto out_unlock;
2822
2823         /* write out all dirty pages from offset */
2824         ret = filemap_write_and_wait_range(src->i_mapping,
2825                                         pos_in, pos_in + len);
2826         if (ret)
2827                 goto out_unlock;
2828
2829         ret = filemap_write_and_wait_range(dst->i_mapping,
2830                                         pos_out, pos_out + len);
2831         if (ret)
2832                 goto out_unlock;
2833
2834         f2fs_balance_fs(sbi, true);
2835
2836         f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2837         if (src != dst) {
2838                 ret = -EBUSY;
2839                 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2840                         goto out_src;
2841         }
2842
2843         f2fs_lock_op(sbi);
2844         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2845                                 pos_out >> F2FS_BLKSIZE_BITS,
2846                                 len >> F2FS_BLKSIZE_BITS, false);
2847
2848         if (!ret) {
2849                 if (dst_max_i_size)
2850                         f2fs_i_size_write(dst, dst_max_i_size);
2851                 else if (dst_osize != dst->i_size)
2852                         f2fs_i_size_write(dst, dst_osize);
2853         }
2854         f2fs_unlock_op(sbi);
2855
2856         if (src != dst)
2857                 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2858 out_src:
2859         f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2860 out_unlock:
2861         if (src != dst)
2862                 inode_unlock(dst);
2863 out:
2864         inode_unlock(src);
2865         return ret;
2866 }
2867
2868 static int __f2fs_ioc_move_range(struct file *filp,
2869                                 struct f2fs_move_range *range)
2870 {
2871         struct fd dst;
2872         int err;
2873
2874         if (!(filp->f_mode & FMODE_READ) ||
2875                         !(filp->f_mode & FMODE_WRITE))
2876                 return -EBADF;
2877
2878         dst = fdget(range->dst_fd);
2879         if (!dst.file)
2880                 return -EBADF;
2881
2882         if (!(dst.file->f_mode & FMODE_WRITE)) {
2883                 err = -EBADF;
2884                 goto err_out;
2885         }
2886
2887         err = mnt_want_write_file(filp);
2888         if (err)
2889                 goto err_out;
2890
2891         err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2892                                         range->pos_out, range->len);
2893
2894         mnt_drop_write_file(filp);
2895 err_out:
2896         fdput(dst);
2897         return err;
2898 }
2899
2900 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2901 {
2902         struct f2fs_move_range range;
2903
2904         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2905                                                         sizeof(range)))
2906                 return -EFAULT;
2907         return __f2fs_ioc_move_range(filp, &range);
2908 }
2909
2910 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2911 {
2912         struct inode *inode = file_inode(filp);
2913         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2914         struct sit_info *sm = SIT_I(sbi);
2915         unsigned int start_segno = 0, end_segno = 0;
2916         unsigned int dev_start_segno = 0, dev_end_segno = 0;
2917         struct f2fs_flush_device range;
2918         struct f2fs_gc_control gc_control = {
2919                         .init_gc_type = FG_GC,
2920                         .should_migrate_blocks = true,
2921                         .err_gc_skipped = true,
2922                         .nr_free_secs = 0 };
2923         int ret;
2924
2925         if (!capable(CAP_SYS_ADMIN))
2926                 return -EPERM;
2927
2928         if (f2fs_readonly(sbi->sb))
2929                 return -EROFS;
2930
2931         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2932                 return -EINVAL;
2933
2934         if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2935                                                         sizeof(range)))
2936                 return -EFAULT;
2937
2938         if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2939                         __is_large_section(sbi)) {
2940                 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2941                           range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2942                 return -EINVAL;
2943         }
2944
2945         ret = mnt_want_write_file(filp);
2946         if (ret)
2947                 return ret;
2948
2949         if (range.dev_num != 0)
2950                 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2951         dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2952
2953         start_segno = sm->last_victim[FLUSH_DEVICE];
2954         if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2955                 start_segno = dev_start_segno;
2956         end_segno = min(start_segno + range.segments, dev_end_segno);
2957
2958         while (start_segno < end_segno) {
2959                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2960                         ret = -EBUSY;
2961                         goto out;
2962                 }
2963                 sm->last_victim[GC_CB] = end_segno + 1;
2964                 sm->last_victim[GC_GREEDY] = end_segno + 1;
2965                 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2966
2967                 gc_control.victim_segno = start_segno;
2968                 ret = f2fs_gc(sbi, &gc_control);
2969                 if (ret == -EAGAIN)
2970                         ret = 0;
2971                 else if (ret < 0)
2972                         break;
2973                 start_segno++;
2974         }
2975 out:
2976         mnt_drop_write_file(filp);
2977         return ret;
2978 }
2979
2980 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2981 {
2982         struct inode *inode = file_inode(filp);
2983         u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2984
2985         /* Must validate to set it with SQLite behavior in Android. */
2986         sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2987
2988         return put_user(sb_feature, (u32 __user *)arg);
2989 }
2990
2991 #ifdef CONFIG_QUOTA
2992 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2993 {
2994         struct dquot *transfer_to[MAXQUOTAS] = {};
2995         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2996         struct super_block *sb = sbi->sb;
2997         int err = 0;
2998
2999         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3000         if (!IS_ERR(transfer_to[PRJQUOTA])) {
3001                 err = __dquot_transfer(inode, transfer_to);
3002                 if (err)
3003                         set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3004                 dqput(transfer_to[PRJQUOTA]);
3005         }
3006         return err;
3007 }
3008
3009 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3010 {
3011         struct f2fs_inode_info *fi = F2FS_I(inode);
3012         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3013         struct f2fs_inode *ri = NULL;
3014         kprojid_t kprojid;
3015         int err;
3016
3017         if (!f2fs_sb_has_project_quota(sbi)) {
3018                 if (projid != F2FS_DEF_PROJID)
3019                         return -EOPNOTSUPP;
3020                 else
3021                         return 0;
3022         }
3023
3024         if (!f2fs_has_extra_attr(inode))
3025                 return -EOPNOTSUPP;
3026
3027         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3028
3029         if (projid_eq(kprojid, fi->i_projid))
3030                 return 0;
3031
3032         err = -EPERM;
3033         /* Is it quota file? Do not allow user to mess with it */
3034         if (IS_NOQUOTA(inode))
3035                 return err;
3036
3037         if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3038                 return -EOVERFLOW;
3039
3040         err = f2fs_dquot_initialize(inode);
3041         if (err)
3042                 return err;
3043
3044         f2fs_lock_op(sbi);
3045         err = f2fs_transfer_project_quota(inode, kprojid);
3046         if (err)
3047                 goto out_unlock;
3048
3049         fi->i_projid = kprojid;
3050         inode->i_ctime = current_time(inode);
3051         f2fs_mark_inode_dirty_sync(inode, true);
3052 out_unlock:
3053         f2fs_unlock_op(sbi);
3054         return err;
3055 }
3056 #else
3057 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3058 {
3059         return 0;
3060 }
3061
3062 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3063 {
3064         if (projid != F2FS_DEF_PROJID)
3065                 return -EOPNOTSUPP;
3066         return 0;
3067 }
3068 #endif
3069
3070 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3071 {
3072         struct inode *inode = d_inode(dentry);
3073         struct f2fs_inode_info *fi = F2FS_I(inode);
3074         u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3075
3076         if (IS_ENCRYPTED(inode))
3077                 fsflags |= FS_ENCRYPT_FL;
3078         if (IS_VERITY(inode))
3079                 fsflags |= FS_VERITY_FL;
3080         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3081                 fsflags |= FS_INLINE_DATA_FL;
3082         if (is_inode_flag_set(inode, FI_PIN_FILE))
3083                 fsflags |= FS_NOCOW_FL;
3084
3085         fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3086
3087         if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3088                 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3089
3090         return 0;
3091 }
3092
3093 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3094                       struct dentry *dentry, struct fileattr *fa)
3095 {
3096         struct inode *inode = d_inode(dentry);
3097         u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3098         u32 iflags;
3099         int err;
3100
3101         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3102                 return -EIO;
3103         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3104                 return -ENOSPC;
3105         if (fsflags & ~F2FS_GETTABLE_FS_FL)
3106                 return -EOPNOTSUPP;
3107         fsflags &= F2FS_SETTABLE_FS_FL;
3108         if (!fa->flags_valid)
3109                 mask &= FS_COMMON_FL;
3110
3111         iflags = f2fs_fsflags_to_iflags(fsflags);
3112         if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3113                 return -EOPNOTSUPP;
3114
3115         err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3116         if (!err)
3117                 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3118
3119         return err;
3120 }
3121
3122 int f2fs_pin_file_control(struct inode *inode, bool inc)
3123 {
3124         struct f2fs_inode_info *fi = F2FS_I(inode);
3125         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3126
3127         /* Use i_gc_failures for normal file as a risk signal. */
3128         if (inc)
3129                 f2fs_i_gc_failures_write(inode,
3130                                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3131
3132         if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3133                 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3134                           __func__, inode->i_ino,
3135                           fi->i_gc_failures[GC_FAILURE_PIN]);
3136                 clear_inode_flag(inode, FI_PIN_FILE);
3137                 return -EAGAIN;
3138         }
3139         return 0;
3140 }
3141
3142 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3143 {
3144         struct inode *inode = file_inode(filp);
3145         __u32 pin;
3146         int ret = 0;
3147
3148         if (get_user(pin, (__u32 __user *)arg))
3149                 return -EFAULT;
3150
3151         if (!S_ISREG(inode->i_mode))
3152                 return -EINVAL;
3153
3154         if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3155                 return -EROFS;
3156
3157         ret = mnt_want_write_file(filp);
3158         if (ret)
3159                 return ret;
3160
3161         inode_lock(inode);
3162
3163         if (!pin) {
3164                 clear_inode_flag(inode, FI_PIN_FILE);
3165                 f2fs_i_gc_failures_write(inode, 0);
3166                 goto done;
3167         }
3168
3169         if (f2fs_should_update_outplace(inode, NULL)) {
3170                 ret = -EINVAL;
3171                 goto out;
3172         }
3173
3174         if (f2fs_pin_file_control(inode, false)) {
3175                 ret = -EAGAIN;
3176                 goto out;
3177         }
3178
3179         ret = f2fs_convert_inline_inode(inode);
3180         if (ret)
3181                 goto out;
3182
3183         if (!f2fs_disable_compressed_file(inode)) {
3184                 ret = -EOPNOTSUPP;
3185                 goto out;
3186         }
3187
3188         set_inode_flag(inode, FI_PIN_FILE);
3189         ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3190 done:
3191         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3192 out:
3193         inode_unlock(inode);
3194         mnt_drop_write_file(filp);
3195         return ret;
3196 }
3197
3198 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3199 {
3200         struct inode *inode = file_inode(filp);
3201         __u32 pin = 0;
3202
3203         if (is_inode_flag_set(inode, FI_PIN_FILE))
3204                 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3205         return put_user(pin, (u32 __user *)arg);
3206 }
3207
3208 int f2fs_precache_extents(struct inode *inode)
3209 {
3210         struct f2fs_inode_info *fi = F2FS_I(inode);
3211         struct f2fs_map_blocks map;
3212         pgoff_t m_next_extent;
3213         loff_t end;
3214         int err;
3215
3216         if (is_inode_flag_set(inode, FI_NO_EXTENT))
3217                 return -EOPNOTSUPP;
3218
3219         map.m_lblk = 0;
3220         map.m_next_pgofs = NULL;
3221         map.m_next_extent = &m_next_extent;
3222         map.m_seg_type = NO_CHECK_TYPE;
3223         map.m_may_create = false;
3224         end = max_file_blocks(inode);
3225
3226         while (map.m_lblk < end) {
3227                 map.m_len = end - map.m_lblk;
3228
3229                 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3230                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3231                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3232                 if (err)
3233                         return err;
3234
3235                 map.m_lblk = m_next_extent;
3236         }
3237
3238         return 0;
3239 }
3240
3241 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3242 {
3243         return f2fs_precache_extents(file_inode(filp));
3244 }
3245
3246 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3247 {
3248         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3249         __u64 block_count;
3250
3251         if (!capable(CAP_SYS_ADMIN))
3252                 return -EPERM;
3253
3254         if (f2fs_readonly(sbi->sb))
3255                 return -EROFS;
3256
3257         if (copy_from_user(&block_count, (void __user *)arg,
3258                            sizeof(block_count)))
3259                 return -EFAULT;
3260
3261         return f2fs_resize_fs(sbi, block_count);
3262 }
3263
3264 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3265 {
3266         struct inode *inode = file_inode(filp);
3267
3268         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3269
3270         if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3271                 f2fs_warn(F2FS_I_SB(inode),
3272                           "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3273                           inode->i_ino);
3274                 return -EOPNOTSUPP;
3275         }
3276
3277         return fsverity_ioctl_enable(filp, (const void __user *)arg);
3278 }
3279
3280 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3281 {
3282         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3283                 return -EOPNOTSUPP;
3284
3285         return fsverity_ioctl_measure(filp, (void __user *)arg);
3286 }
3287
3288 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3289 {
3290         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3291                 return -EOPNOTSUPP;
3292
3293         return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3294 }
3295
3296 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3297 {
3298         struct inode *inode = file_inode(filp);
3299         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3300         char *vbuf;
3301         int count;
3302         int err = 0;
3303
3304         vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3305         if (!vbuf)
3306                 return -ENOMEM;
3307
3308         f2fs_down_read(&sbi->sb_lock);
3309         count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3310                         ARRAY_SIZE(sbi->raw_super->volume_name),
3311                         UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3312         f2fs_up_read(&sbi->sb_lock);
3313
3314         if (copy_to_user((char __user *)arg, vbuf,
3315                                 min(FSLABEL_MAX, count)))
3316                 err = -EFAULT;
3317
3318         kfree(vbuf);
3319         return err;
3320 }
3321
3322 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3323 {
3324         struct inode *inode = file_inode(filp);
3325         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3326         char *vbuf;
3327         int err = 0;
3328
3329         if (!capable(CAP_SYS_ADMIN))
3330                 return -EPERM;
3331
3332         vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3333         if (IS_ERR(vbuf))
3334                 return PTR_ERR(vbuf);
3335
3336         err = mnt_want_write_file(filp);
3337         if (err)
3338                 goto out;
3339
3340         f2fs_down_write(&sbi->sb_lock);
3341
3342         memset(sbi->raw_super->volume_name, 0,
3343                         sizeof(sbi->raw_super->volume_name));
3344         utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3345                         sbi->raw_super->volume_name,
3346                         ARRAY_SIZE(sbi->raw_super->volume_name));
3347
3348         err = f2fs_commit_super(sbi, false);
3349
3350         f2fs_up_write(&sbi->sb_lock);
3351
3352         mnt_drop_write_file(filp);
3353 out:
3354         kfree(vbuf);
3355         return err;
3356 }
3357
3358 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3359 {
3360         struct inode *inode = file_inode(filp);
3361         __u64 blocks;
3362
3363         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3364                 return -EOPNOTSUPP;
3365
3366         if (!f2fs_compressed_file(inode))
3367                 return -EINVAL;
3368
3369         blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3370         return put_user(blocks, (u64 __user *)arg);
3371 }
3372
3373 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3374 {
3375         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3376         unsigned int released_blocks = 0;
3377         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3378         block_t blkaddr;
3379         int i;
3380
3381         for (i = 0; i < count; i++) {
3382                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3383                                                 dn->ofs_in_node + i);
3384
3385                 if (!__is_valid_data_blkaddr(blkaddr))
3386                         continue;
3387                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3388                                         DATA_GENERIC_ENHANCE))) {
3389                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3390                         return -EFSCORRUPTED;
3391                 }
3392         }
3393
3394         while (count) {
3395                 int compr_blocks = 0;
3396
3397                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3398                         blkaddr = f2fs_data_blkaddr(dn);
3399
3400                         if (i == 0) {
3401                                 if (blkaddr == COMPRESS_ADDR)
3402                                         continue;
3403                                 dn->ofs_in_node += cluster_size;
3404                                 goto next;
3405                         }
3406
3407                         if (__is_valid_data_blkaddr(blkaddr))
3408                                 compr_blocks++;
3409
3410                         if (blkaddr != NEW_ADDR)
3411                                 continue;
3412
3413                         dn->data_blkaddr = NULL_ADDR;
3414                         f2fs_set_data_blkaddr(dn);
3415                 }
3416
3417                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3418                 dec_valid_block_count(sbi, dn->inode,
3419                                         cluster_size - compr_blocks);
3420
3421                 released_blocks += cluster_size - compr_blocks;
3422 next:
3423                 count -= cluster_size;
3424         }
3425
3426         return released_blocks;
3427 }
3428
3429 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3430 {
3431         struct inode *inode = file_inode(filp);
3432         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3433         pgoff_t page_idx = 0, last_idx;
3434         unsigned int released_blocks = 0;
3435         int ret;
3436         int writecount;
3437
3438         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3439                 return -EOPNOTSUPP;
3440
3441         if (!f2fs_compressed_file(inode))
3442                 return -EINVAL;
3443
3444         if (f2fs_readonly(sbi->sb))
3445                 return -EROFS;
3446
3447         ret = mnt_want_write_file(filp);
3448         if (ret)
3449                 return ret;
3450
3451         f2fs_balance_fs(F2FS_I_SB(inode), true);
3452
3453         inode_lock(inode);
3454
3455         writecount = atomic_read(&inode->i_writecount);
3456         if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3457                         (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3458                 ret = -EBUSY;
3459                 goto out;
3460         }
3461
3462         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3463                 ret = -EINVAL;
3464                 goto out;
3465         }
3466
3467         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3468         if (ret)
3469                 goto out;
3470
3471         set_inode_flag(inode, FI_COMPRESS_RELEASED);
3472         inode->i_ctime = current_time(inode);
3473         f2fs_mark_inode_dirty_sync(inode, true);
3474
3475         if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3476                 goto out;
3477
3478         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3479         filemap_invalidate_lock(inode->i_mapping);
3480
3481         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3482
3483         while (page_idx < last_idx) {
3484                 struct dnode_of_data dn;
3485                 pgoff_t end_offset, count;
3486
3487                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3488                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3489                 if (ret) {
3490                         if (ret == -ENOENT) {
3491                                 page_idx = f2fs_get_next_page_offset(&dn,
3492                                                                 page_idx);
3493                                 ret = 0;
3494                                 continue;
3495                         }
3496                         break;
3497                 }
3498
3499                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3500                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3501                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3502
3503                 ret = release_compress_blocks(&dn, count);
3504
3505                 f2fs_put_dnode(&dn);
3506
3507                 if (ret < 0)
3508                         break;
3509
3510                 page_idx += count;
3511                 released_blocks += ret;
3512         }
3513
3514         filemap_invalidate_unlock(inode->i_mapping);
3515         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3516 out:
3517         inode_unlock(inode);
3518
3519         mnt_drop_write_file(filp);
3520
3521         if (ret >= 0) {
3522                 ret = put_user(released_blocks, (u64 __user *)arg);
3523         } else if (released_blocks &&
3524                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3525                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3526                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3527                         "iblocks=%llu, released=%u, compr_blocks=%u, "
3528                         "run fsck to fix.",
3529                         __func__, inode->i_ino, inode->i_blocks,
3530                         released_blocks,
3531                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3532         }
3533
3534         return ret;
3535 }
3536
3537 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3538 {
3539         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3540         unsigned int reserved_blocks = 0;
3541         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3542         block_t blkaddr;
3543         int i;
3544
3545         for (i = 0; i < count; i++) {
3546                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3547                                                 dn->ofs_in_node + i);
3548
3549                 if (!__is_valid_data_blkaddr(blkaddr))
3550                         continue;
3551                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3552                                         DATA_GENERIC_ENHANCE))) {
3553                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3554                         return -EFSCORRUPTED;
3555                 }
3556         }
3557
3558         while (count) {
3559                 int compr_blocks = 0;
3560                 blkcnt_t reserved;
3561                 int ret;
3562
3563                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3564                         blkaddr = f2fs_data_blkaddr(dn);
3565
3566                         if (i == 0) {
3567                                 if (blkaddr == COMPRESS_ADDR)
3568                                         continue;
3569                                 dn->ofs_in_node += cluster_size;
3570                                 goto next;
3571                         }
3572
3573                         if (__is_valid_data_blkaddr(blkaddr)) {
3574                                 compr_blocks++;
3575                                 continue;
3576                         }
3577
3578                         dn->data_blkaddr = NEW_ADDR;
3579                         f2fs_set_data_blkaddr(dn);
3580                 }
3581
3582                 reserved = cluster_size - compr_blocks;
3583                 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3584                 if (ret)
3585                         return ret;
3586
3587                 if (reserved != cluster_size - compr_blocks)
3588                         return -ENOSPC;
3589
3590                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3591
3592                 reserved_blocks += reserved;
3593 next:
3594                 count -= cluster_size;
3595         }
3596
3597         return reserved_blocks;
3598 }
3599
3600 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3601 {
3602         struct inode *inode = file_inode(filp);
3603         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3604         pgoff_t page_idx = 0, last_idx;
3605         unsigned int reserved_blocks = 0;
3606         int ret;
3607
3608         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3609                 return -EOPNOTSUPP;
3610
3611         if (!f2fs_compressed_file(inode))
3612                 return -EINVAL;
3613
3614         if (f2fs_readonly(sbi->sb))
3615                 return -EROFS;
3616
3617         ret = mnt_want_write_file(filp);
3618         if (ret)
3619                 return ret;
3620
3621         if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3622                 goto out;
3623
3624         f2fs_balance_fs(F2FS_I_SB(inode), true);
3625
3626         inode_lock(inode);
3627
3628         if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3629                 ret = -EINVAL;
3630                 goto unlock_inode;
3631         }
3632
3633         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3634         filemap_invalidate_lock(inode->i_mapping);
3635
3636         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3637
3638         while (page_idx < last_idx) {
3639                 struct dnode_of_data dn;
3640                 pgoff_t end_offset, count;
3641
3642                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3643                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3644                 if (ret) {
3645                         if (ret == -ENOENT) {
3646                                 page_idx = f2fs_get_next_page_offset(&dn,
3647                                                                 page_idx);
3648                                 ret = 0;
3649                                 continue;
3650                         }
3651                         break;
3652                 }
3653
3654                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3655                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3656                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3657
3658                 ret = reserve_compress_blocks(&dn, count);
3659
3660                 f2fs_put_dnode(&dn);
3661
3662                 if (ret < 0)
3663                         break;
3664
3665                 page_idx += count;
3666                 reserved_blocks += ret;
3667         }
3668
3669         filemap_invalidate_unlock(inode->i_mapping);
3670         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3671
3672         if (ret >= 0) {
3673                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3674                 inode->i_ctime = current_time(inode);
3675                 f2fs_mark_inode_dirty_sync(inode, true);
3676         }
3677 unlock_inode:
3678         inode_unlock(inode);
3679 out:
3680         mnt_drop_write_file(filp);
3681
3682         if (ret >= 0) {
3683                 ret = put_user(reserved_blocks, (u64 __user *)arg);
3684         } else if (reserved_blocks &&
3685                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3686                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3687                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3688                         "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3689                         "run fsck to fix.",
3690                         __func__, inode->i_ino, inode->i_blocks,
3691                         reserved_blocks,
3692                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3693         }
3694
3695         return ret;
3696 }
3697
3698 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3699                 pgoff_t off, block_t block, block_t len, u32 flags)
3700 {
3701         sector_t sector = SECTOR_FROM_BLOCK(block);
3702         sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3703         int ret = 0;
3704
3705         if (flags & F2FS_TRIM_FILE_DISCARD) {
3706                 if (bdev_max_secure_erase_sectors(bdev))
3707                         ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3708                                         GFP_NOFS);
3709                 else
3710                         ret = blkdev_issue_discard(bdev, sector, nr_sects,
3711                                         GFP_NOFS);
3712         }
3713
3714         if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3715                 if (IS_ENCRYPTED(inode))
3716                         ret = fscrypt_zeroout_range(inode, off, block, len);
3717                 else
3718                         ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3719                                         GFP_NOFS, 0);
3720         }
3721
3722         return ret;
3723 }
3724
3725 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3726 {
3727         struct inode *inode = file_inode(filp);
3728         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3729         struct address_space *mapping = inode->i_mapping;
3730         struct block_device *prev_bdev = NULL;
3731         struct f2fs_sectrim_range range;
3732         pgoff_t index, pg_end, prev_index = 0;
3733         block_t prev_block = 0, len = 0;
3734         loff_t end_addr;
3735         bool to_end = false;
3736         int ret = 0;
3737
3738         if (!(filp->f_mode & FMODE_WRITE))
3739                 return -EBADF;
3740
3741         if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3742                                 sizeof(range)))
3743                 return -EFAULT;
3744
3745         if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3746                         !S_ISREG(inode->i_mode))
3747                 return -EINVAL;
3748
3749         if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3750                         !f2fs_hw_support_discard(sbi)) ||
3751                         ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3752                          IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3753                 return -EOPNOTSUPP;
3754
3755         file_start_write(filp);
3756         inode_lock(inode);
3757
3758         if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3759                         range.start >= inode->i_size) {
3760                 ret = -EINVAL;
3761                 goto err;
3762         }
3763
3764         if (range.len == 0)
3765                 goto err;
3766
3767         if (inode->i_size - range.start > range.len) {
3768                 end_addr = range.start + range.len;
3769         } else {
3770                 end_addr = range.len == (u64)-1 ?
3771                         sbi->sb->s_maxbytes : inode->i_size;
3772                 to_end = true;
3773         }
3774
3775         if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3776                         (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3777                 ret = -EINVAL;
3778                 goto err;
3779         }
3780
3781         index = F2FS_BYTES_TO_BLK(range.start);
3782         pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3783
3784         ret = f2fs_convert_inline_inode(inode);
3785         if (ret)
3786                 goto err;
3787
3788         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3789         filemap_invalidate_lock(mapping);
3790
3791         ret = filemap_write_and_wait_range(mapping, range.start,
3792                         to_end ? LLONG_MAX : end_addr - 1);
3793         if (ret)
3794                 goto out;
3795
3796         truncate_inode_pages_range(mapping, range.start,
3797                         to_end ? -1 : end_addr - 1);
3798
3799         while (index < pg_end) {
3800                 struct dnode_of_data dn;
3801                 pgoff_t end_offset, count;
3802                 int i;
3803
3804                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3805                 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3806                 if (ret) {
3807                         if (ret == -ENOENT) {
3808                                 index = f2fs_get_next_page_offset(&dn, index);
3809                                 continue;
3810                         }
3811                         goto out;
3812                 }
3813
3814                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3815                 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3816                 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3817                         struct block_device *cur_bdev;
3818                         block_t blkaddr = f2fs_data_blkaddr(&dn);
3819
3820                         if (!__is_valid_data_blkaddr(blkaddr))
3821                                 continue;
3822
3823                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3824                                                 DATA_GENERIC_ENHANCE)) {
3825                                 ret = -EFSCORRUPTED;
3826                                 f2fs_put_dnode(&dn);
3827                                 f2fs_handle_error(sbi,
3828                                                 ERROR_INVALID_BLKADDR);
3829                                 goto out;
3830                         }
3831
3832                         cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3833                         if (f2fs_is_multi_device(sbi)) {
3834                                 int di = f2fs_target_device_index(sbi, blkaddr);
3835
3836                                 blkaddr -= FDEV(di).start_blk;
3837                         }
3838
3839                         if (len) {
3840                                 if (prev_bdev == cur_bdev &&
3841                                                 index == prev_index + len &&
3842                                                 blkaddr == prev_block + len) {
3843                                         len++;
3844                                 } else {
3845                                         ret = f2fs_secure_erase(prev_bdev,
3846                                                 inode, prev_index, prev_block,
3847                                                 len, range.flags);
3848                                         if (ret) {
3849                                                 f2fs_put_dnode(&dn);
3850                                                 goto out;
3851                                         }
3852
3853                                         len = 0;
3854                                 }
3855                         }
3856
3857                         if (!len) {
3858                                 prev_bdev = cur_bdev;
3859                                 prev_index = index;
3860                                 prev_block = blkaddr;
3861                                 len = 1;
3862                         }
3863                 }
3864
3865                 f2fs_put_dnode(&dn);
3866
3867                 if (fatal_signal_pending(current)) {
3868                         ret = -EINTR;
3869                         goto out;
3870                 }
3871                 cond_resched();
3872         }
3873
3874         if (len)
3875                 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3876                                 prev_block, len, range.flags);
3877 out:
3878         filemap_invalidate_unlock(mapping);
3879         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3880 err:
3881         inode_unlock(inode);
3882         file_end_write(filp);
3883
3884         return ret;
3885 }
3886
3887 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3888 {
3889         struct inode *inode = file_inode(filp);
3890         struct f2fs_comp_option option;
3891
3892         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3893                 return -EOPNOTSUPP;
3894
3895         inode_lock_shared(inode);
3896
3897         if (!f2fs_compressed_file(inode)) {
3898                 inode_unlock_shared(inode);
3899                 return -ENODATA;
3900         }
3901
3902         option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3903         option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3904
3905         inode_unlock_shared(inode);
3906
3907         if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3908                                 sizeof(option)))
3909                 return -EFAULT;
3910
3911         return 0;
3912 }
3913
3914 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3915 {
3916         struct inode *inode = file_inode(filp);
3917         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3918         struct f2fs_comp_option option;
3919         int ret = 0;
3920
3921         if (!f2fs_sb_has_compression(sbi))
3922                 return -EOPNOTSUPP;
3923
3924         if (!(filp->f_mode & FMODE_WRITE))
3925                 return -EBADF;
3926
3927         if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3928                                 sizeof(option)))
3929                 return -EFAULT;
3930
3931         if (!f2fs_compressed_file(inode) ||
3932                         option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3933                         option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3934                         option.algorithm >= COMPRESS_MAX)
3935                 return -EINVAL;
3936
3937         file_start_write(filp);
3938         inode_lock(inode);
3939
3940         if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3941                 ret = -EBUSY;
3942                 goto out;
3943         }
3944
3945         if (inode->i_size != 0) {
3946                 ret = -EFBIG;
3947                 goto out;
3948         }
3949
3950         F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3951         F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3952         F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3953         f2fs_mark_inode_dirty_sync(inode, true);
3954
3955         if (!f2fs_is_compress_backend_ready(inode))
3956                 f2fs_warn(sbi, "compression algorithm is successfully set, "
3957                         "but current kernel doesn't support this algorithm.");
3958 out:
3959         inode_unlock(inode);
3960         file_end_write(filp);
3961
3962         return ret;
3963 }
3964
3965 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3966 {
3967         DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3968         struct address_space *mapping = inode->i_mapping;
3969         struct page *page;
3970         pgoff_t redirty_idx = page_idx;
3971         int i, page_len = 0, ret = 0;
3972
3973         page_cache_ra_unbounded(&ractl, len, 0);
3974
3975         for (i = 0; i < len; i++, page_idx++) {
3976                 page = read_cache_page(mapping, page_idx, NULL, NULL);
3977                 if (IS_ERR(page)) {
3978                         ret = PTR_ERR(page);
3979                         break;
3980                 }
3981                 page_len++;
3982         }
3983
3984         for (i = 0; i < page_len; i++, redirty_idx++) {
3985                 page = find_lock_page(mapping, redirty_idx);
3986
3987                 /* It will never fail, when page has pinned above */
3988                 f2fs_bug_on(F2FS_I_SB(inode), !page);
3989
3990                 set_page_dirty(page);
3991                 f2fs_put_page(page, 1);
3992                 f2fs_put_page(page, 0);
3993         }
3994
3995         return ret;
3996 }
3997
3998 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3999 {
4000         struct inode *inode = file_inode(filp);
4001         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4002         struct f2fs_inode_info *fi = F2FS_I(inode);
4003         pgoff_t page_idx = 0, last_idx;
4004         unsigned int blk_per_seg = sbi->blocks_per_seg;
4005         int cluster_size = fi->i_cluster_size;
4006         int count, ret;
4007
4008         if (!f2fs_sb_has_compression(sbi) ||
4009                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4010                 return -EOPNOTSUPP;
4011
4012         if (!(filp->f_mode & FMODE_WRITE))
4013                 return -EBADF;
4014
4015         if (!f2fs_compressed_file(inode))
4016                 return -EINVAL;
4017
4018         f2fs_balance_fs(F2FS_I_SB(inode), true);
4019
4020         file_start_write(filp);
4021         inode_lock(inode);
4022
4023         if (!f2fs_is_compress_backend_ready(inode)) {
4024                 ret = -EOPNOTSUPP;
4025                 goto out;
4026         }
4027
4028         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4029                 ret = -EINVAL;
4030                 goto out;
4031         }
4032
4033         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4034         if (ret)
4035                 goto out;
4036
4037         if (!atomic_read(&fi->i_compr_blocks))
4038                 goto out;
4039
4040         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4041
4042         count = last_idx - page_idx;
4043         while (count) {
4044                 int len = min(cluster_size, count);
4045
4046                 ret = redirty_blocks(inode, page_idx, len);
4047                 if (ret < 0)
4048                         break;
4049
4050                 if (get_dirty_pages(inode) >= blk_per_seg)
4051                         filemap_fdatawrite(inode->i_mapping);
4052
4053                 count -= len;
4054                 page_idx += len;
4055         }
4056
4057         if (!ret)
4058                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4059                                                         LLONG_MAX);
4060
4061         if (ret)
4062                 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4063                           __func__, ret);
4064 out:
4065         inode_unlock(inode);
4066         file_end_write(filp);
4067
4068         return ret;
4069 }
4070
4071 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4072 {
4073         struct inode *inode = file_inode(filp);
4074         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4075         pgoff_t page_idx = 0, last_idx;
4076         unsigned int blk_per_seg = sbi->blocks_per_seg;
4077         int cluster_size = F2FS_I(inode)->i_cluster_size;
4078         int count, ret;
4079
4080         if (!f2fs_sb_has_compression(sbi) ||
4081                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4082                 return -EOPNOTSUPP;
4083
4084         if (!(filp->f_mode & FMODE_WRITE))
4085                 return -EBADF;
4086
4087         if (!f2fs_compressed_file(inode))
4088                 return -EINVAL;
4089
4090         f2fs_balance_fs(F2FS_I_SB(inode), true);
4091
4092         file_start_write(filp);
4093         inode_lock(inode);
4094
4095         if (!f2fs_is_compress_backend_ready(inode)) {
4096                 ret = -EOPNOTSUPP;
4097                 goto out;
4098         }
4099
4100         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4101                 ret = -EINVAL;
4102                 goto out;
4103         }
4104
4105         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4106         if (ret)
4107                 goto out;
4108
4109         set_inode_flag(inode, FI_ENABLE_COMPRESS);
4110
4111         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4112
4113         count = last_idx - page_idx;
4114         while (count) {
4115                 int len = min(cluster_size, count);
4116
4117                 ret = redirty_blocks(inode, page_idx, len);
4118                 if (ret < 0)
4119                         break;
4120
4121                 if (get_dirty_pages(inode) >= blk_per_seg)
4122                         filemap_fdatawrite(inode->i_mapping);
4123
4124                 count -= len;
4125                 page_idx += len;
4126         }
4127
4128         if (!ret)
4129                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4130                                                         LLONG_MAX);
4131
4132         clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4133
4134         if (ret)
4135                 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4136                           __func__, ret);
4137 out:
4138         inode_unlock(inode);
4139         file_end_write(filp);
4140
4141         return ret;
4142 }
4143
4144 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4145 {
4146         switch (cmd) {
4147         case FS_IOC_GETVERSION:
4148                 return f2fs_ioc_getversion(filp, arg);
4149         case F2FS_IOC_START_ATOMIC_WRITE:
4150                 return f2fs_ioc_start_atomic_write(filp, false);
4151         case F2FS_IOC_START_ATOMIC_REPLACE:
4152                 return f2fs_ioc_start_atomic_write(filp, true);
4153         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4154                 return f2fs_ioc_commit_atomic_write(filp);
4155         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4156                 return f2fs_ioc_abort_atomic_write(filp);
4157         case F2FS_IOC_START_VOLATILE_WRITE:
4158         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4159                 return -EOPNOTSUPP;
4160         case F2FS_IOC_SHUTDOWN:
4161                 return f2fs_ioc_shutdown(filp, arg);
4162         case FITRIM:
4163                 return f2fs_ioc_fitrim(filp, arg);
4164         case FS_IOC_SET_ENCRYPTION_POLICY:
4165                 return f2fs_ioc_set_encryption_policy(filp, arg);
4166         case FS_IOC_GET_ENCRYPTION_POLICY:
4167                 return f2fs_ioc_get_encryption_policy(filp, arg);
4168         case FS_IOC_GET_ENCRYPTION_PWSALT:
4169                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4170         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4171                 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4172         case FS_IOC_ADD_ENCRYPTION_KEY:
4173                 return f2fs_ioc_add_encryption_key(filp, arg);
4174         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4175                 return f2fs_ioc_remove_encryption_key(filp, arg);
4176         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4177                 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4178         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4179                 return f2fs_ioc_get_encryption_key_status(filp, arg);
4180         case FS_IOC_GET_ENCRYPTION_NONCE:
4181                 return f2fs_ioc_get_encryption_nonce(filp, arg);
4182         case F2FS_IOC_GARBAGE_COLLECT:
4183                 return f2fs_ioc_gc(filp, arg);
4184         case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4185                 return f2fs_ioc_gc_range(filp, arg);
4186         case F2FS_IOC_WRITE_CHECKPOINT:
4187                 return f2fs_ioc_write_checkpoint(filp, arg);
4188         case F2FS_IOC_DEFRAGMENT:
4189                 return f2fs_ioc_defragment(filp, arg);
4190         case F2FS_IOC_MOVE_RANGE:
4191                 return f2fs_ioc_move_range(filp, arg);
4192         case F2FS_IOC_FLUSH_DEVICE:
4193                 return f2fs_ioc_flush_device(filp, arg);
4194         case F2FS_IOC_GET_FEATURES:
4195                 return f2fs_ioc_get_features(filp, arg);
4196         case F2FS_IOC_GET_PIN_FILE:
4197                 return f2fs_ioc_get_pin_file(filp, arg);
4198         case F2FS_IOC_SET_PIN_FILE:
4199                 return f2fs_ioc_set_pin_file(filp, arg);
4200         case F2FS_IOC_PRECACHE_EXTENTS:
4201                 return f2fs_ioc_precache_extents(filp, arg);
4202         case F2FS_IOC_RESIZE_FS:
4203                 return f2fs_ioc_resize_fs(filp, arg);
4204         case FS_IOC_ENABLE_VERITY:
4205                 return f2fs_ioc_enable_verity(filp, arg);
4206         case FS_IOC_MEASURE_VERITY:
4207                 return f2fs_ioc_measure_verity(filp, arg);
4208         case FS_IOC_READ_VERITY_METADATA:
4209                 return f2fs_ioc_read_verity_metadata(filp, arg);
4210         case FS_IOC_GETFSLABEL:
4211                 return f2fs_ioc_getfslabel(filp, arg);
4212         case FS_IOC_SETFSLABEL:
4213                 return f2fs_ioc_setfslabel(filp, arg);
4214         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4215                 return f2fs_get_compress_blocks(filp, arg);
4216         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4217                 return f2fs_release_compress_blocks(filp, arg);
4218         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4219                 return f2fs_reserve_compress_blocks(filp, arg);
4220         case F2FS_IOC_SEC_TRIM_FILE:
4221                 return f2fs_sec_trim_file(filp, arg);
4222         case F2FS_IOC_GET_COMPRESS_OPTION:
4223                 return f2fs_ioc_get_compress_option(filp, arg);
4224         case F2FS_IOC_SET_COMPRESS_OPTION:
4225                 return f2fs_ioc_set_compress_option(filp, arg);
4226         case F2FS_IOC_DECOMPRESS_FILE:
4227                 return f2fs_ioc_decompress_file(filp, arg);
4228         case F2FS_IOC_COMPRESS_FILE:
4229                 return f2fs_ioc_compress_file(filp, arg);
4230         default:
4231                 return -ENOTTY;
4232         }
4233 }
4234
4235 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4236 {
4237         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4238                 return -EIO;
4239         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4240                 return -ENOSPC;
4241
4242         return __f2fs_ioctl(filp, cmd, arg);
4243 }
4244
4245 /*
4246  * Return %true if the given read or write request should use direct I/O, or
4247  * %false if it should use buffered I/O.
4248  */
4249 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4250                                 struct iov_iter *iter)
4251 {
4252         unsigned int align;
4253
4254         if (!(iocb->ki_flags & IOCB_DIRECT))
4255                 return false;
4256
4257         if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4258                 return false;
4259
4260         /*
4261          * Direct I/O not aligned to the disk's logical_block_size will be
4262          * attempted, but will fail with -EINVAL.
4263          *
4264          * f2fs additionally requires that direct I/O be aligned to the
4265          * filesystem block size, which is often a stricter requirement.
4266          * However, f2fs traditionally falls back to buffered I/O on requests
4267          * that are logical_block_size-aligned but not fs-block aligned.
4268          *
4269          * The below logic implements this behavior.
4270          */
4271         align = iocb->ki_pos | iov_iter_alignment(iter);
4272         if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4273             IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4274                 return false;
4275
4276         return true;
4277 }
4278
4279 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4280                                 unsigned int flags)
4281 {
4282         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4283
4284         dec_page_count(sbi, F2FS_DIO_READ);
4285         if (error)
4286                 return error;
4287         f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4288         return 0;
4289 }
4290
4291 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4292         .end_io = f2fs_dio_read_end_io,
4293 };
4294
4295 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4296 {
4297         struct file *file = iocb->ki_filp;
4298         struct inode *inode = file_inode(file);
4299         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4300         struct f2fs_inode_info *fi = F2FS_I(inode);
4301         const loff_t pos = iocb->ki_pos;
4302         const size_t count = iov_iter_count(to);
4303         struct iomap_dio *dio;
4304         ssize_t ret;
4305
4306         if (count == 0)
4307                 return 0; /* skip atime update */
4308
4309         trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4310
4311         if (iocb->ki_flags & IOCB_NOWAIT) {
4312                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4313                         ret = -EAGAIN;
4314                         goto out;
4315                 }
4316         } else {
4317                 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4318         }
4319
4320         /*
4321          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4322          * the higher-level function iomap_dio_rw() in order to ensure that the
4323          * F2FS_DIO_READ counter will be decremented correctly in all cases.
4324          */
4325         inc_page_count(sbi, F2FS_DIO_READ);
4326         dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4327                              &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4328         if (IS_ERR_OR_NULL(dio)) {
4329                 ret = PTR_ERR_OR_ZERO(dio);
4330                 if (ret != -EIOCBQUEUED)
4331                         dec_page_count(sbi, F2FS_DIO_READ);
4332         } else {
4333                 ret = iomap_dio_complete(dio);
4334         }
4335
4336         f2fs_up_read(&fi->i_gc_rwsem[READ]);
4337
4338         file_accessed(file);
4339 out:
4340         trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4341         return ret;
4342 }
4343
4344 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4345 {
4346         struct inode *inode = file_inode(iocb->ki_filp);
4347         const loff_t pos = iocb->ki_pos;
4348         ssize_t ret;
4349
4350         if (!f2fs_is_compress_backend_ready(inode))
4351                 return -EOPNOTSUPP;
4352
4353         if (trace_f2fs_dataread_start_enabled()) {
4354                 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4355                 char *path;
4356
4357                 if (!p)
4358                         goto skip_read_trace;
4359
4360                 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4361                 if (IS_ERR(path)) {
4362                         kfree(p);
4363                         goto skip_read_trace;
4364                 }
4365
4366                 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4367                                         current->pid, path, current->comm);
4368                 kfree(p);
4369         }
4370 skip_read_trace:
4371         if (f2fs_should_use_dio(inode, iocb, to)) {
4372                 ret = f2fs_dio_read_iter(iocb, to);
4373         } else {
4374                 ret = filemap_read(iocb, to, 0);
4375                 if (ret > 0)
4376                         f2fs_update_iostat(F2FS_I_SB(inode), inode,
4377                                                 APP_BUFFERED_READ_IO, ret);
4378         }
4379         if (trace_f2fs_dataread_end_enabled())
4380                 trace_f2fs_dataread_end(inode, pos, ret);
4381         return ret;
4382 }
4383
4384 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4385 {
4386         struct file *file = iocb->ki_filp;
4387         struct inode *inode = file_inode(file);
4388         ssize_t count;
4389         int err;
4390
4391         if (IS_IMMUTABLE(inode))
4392                 return -EPERM;
4393
4394         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4395                 return -EPERM;
4396
4397         count = generic_write_checks(iocb, from);
4398         if (count <= 0)
4399                 return count;
4400
4401         err = file_modified(file);
4402         if (err)
4403                 return err;
4404         return count;
4405 }
4406
4407 /*
4408  * Preallocate blocks for a write request, if it is possible and helpful to do
4409  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4410  * blocks were preallocated, or a negative errno value if something went
4411  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4412  * requested blocks (not just some of them) have been allocated.
4413  */
4414 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4415                                    bool dio)
4416 {
4417         struct inode *inode = file_inode(iocb->ki_filp);
4418         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4419         const loff_t pos = iocb->ki_pos;
4420         const size_t count = iov_iter_count(iter);
4421         struct f2fs_map_blocks map = {};
4422         int flag;
4423         int ret;
4424
4425         /* If it will be an out-of-place direct write, don't bother. */
4426         if (dio && f2fs_lfs_mode(sbi))
4427                 return 0;
4428         /*
4429          * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4430          * buffered IO, if DIO meets any holes.
4431          */
4432         if (dio && i_size_read(inode) &&
4433                 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4434                 return 0;
4435
4436         /* No-wait I/O can't allocate blocks. */
4437         if (iocb->ki_flags & IOCB_NOWAIT)
4438                 return 0;
4439
4440         /* If it will be a short write, don't bother. */
4441         if (fault_in_iov_iter_readable(iter, count))
4442                 return 0;
4443
4444         if (f2fs_has_inline_data(inode)) {
4445                 /* If the data will fit inline, don't bother. */
4446                 if (pos + count <= MAX_INLINE_DATA(inode))
4447                         return 0;
4448                 ret = f2fs_convert_inline_inode(inode);
4449                 if (ret)
4450                         return ret;
4451         }
4452
4453         /* Do not preallocate blocks that will be written partially in 4KB. */
4454         map.m_lblk = F2FS_BLK_ALIGN(pos);
4455         map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4456         if (map.m_len > map.m_lblk)
4457                 map.m_len -= map.m_lblk;
4458         else
4459                 map.m_len = 0;
4460         map.m_may_create = true;
4461         if (dio) {
4462                 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4463                 flag = F2FS_GET_BLOCK_PRE_DIO;
4464         } else {
4465                 map.m_seg_type = NO_CHECK_TYPE;
4466                 flag = F2FS_GET_BLOCK_PRE_AIO;
4467         }
4468
4469         ret = f2fs_map_blocks(inode, &map, 1, flag);
4470         /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4471         if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4472                 return ret;
4473         if (ret == 0)
4474                 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4475         return map.m_len;
4476 }
4477
4478 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4479                                         struct iov_iter *from)
4480 {
4481         struct file *file = iocb->ki_filp;
4482         struct inode *inode = file_inode(file);
4483         ssize_t ret;
4484
4485         if (iocb->ki_flags & IOCB_NOWAIT)
4486                 return -EOPNOTSUPP;
4487
4488         current->backing_dev_info = inode_to_bdi(inode);
4489         ret = generic_perform_write(iocb, from);
4490         current->backing_dev_info = NULL;
4491
4492         if (ret > 0) {
4493                 iocb->ki_pos += ret;
4494                 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4495                                                 APP_BUFFERED_IO, ret);
4496         }
4497         return ret;
4498 }
4499
4500 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4501                                  unsigned int flags)
4502 {
4503         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4504
4505         dec_page_count(sbi, F2FS_DIO_WRITE);
4506         if (error)
4507                 return error;
4508         f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4509         return 0;
4510 }
4511
4512 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4513         .end_io = f2fs_dio_write_end_io,
4514 };
4515
4516 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4517                                    bool *may_need_sync)
4518 {
4519         struct file *file = iocb->ki_filp;
4520         struct inode *inode = file_inode(file);
4521         struct f2fs_inode_info *fi = F2FS_I(inode);
4522         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4523         const bool do_opu = f2fs_lfs_mode(sbi);
4524         const loff_t pos = iocb->ki_pos;
4525         const ssize_t count = iov_iter_count(from);
4526         unsigned int dio_flags;
4527         struct iomap_dio *dio;
4528         ssize_t ret;
4529
4530         trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4531
4532         if (iocb->ki_flags & IOCB_NOWAIT) {
4533                 /* f2fs_convert_inline_inode() and block allocation can block */
4534                 if (f2fs_has_inline_data(inode) ||
4535                     !f2fs_overwrite_io(inode, pos, count)) {
4536                         ret = -EAGAIN;
4537                         goto out;
4538                 }
4539
4540                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4541                         ret = -EAGAIN;
4542                         goto out;
4543                 }
4544                 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4545                         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4546                         ret = -EAGAIN;
4547                         goto out;
4548                 }
4549         } else {
4550                 ret = f2fs_convert_inline_inode(inode);
4551                 if (ret)
4552                         goto out;
4553
4554                 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4555                 if (do_opu)
4556                         f2fs_down_read(&fi->i_gc_rwsem[READ]);
4557         }
4558
4559         /*
4560          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4561          * the higher-level function iomap_dio_rw() in order to ensure that the
4562          * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4563          */
4564         inc_page_count(sbi, F2FS_DIO_WRITE);
4565         dio_flags = 0;
4566         if (pos + count > inode->i_size)
4567                 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4568         dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4569                              &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4570         if (IS_ERR_OR_NULL(dio)) {
4571                 ret = PTR_ERR_OR_ZERO(dio);
4572                 if (ret == -ENOTBLK)
4573                         ret = 0;
4574                 if (ret != -EIOCBQUEUED)
4575                         dec_page_count(sbi, F2FS_DIO_WRITE);
4576         } else {
4577                 ret = iomap_dio_complete(dio);
4578         }
4579
4580         if (do_opu)
4581                 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4582         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4583
4584         if (ret < 0)
4585                 goto out;
4586         if (pos + ret > inode->i_size)
4587                 f2fs_i_size_write(inode, pos + ret);
4588         if (!do_opu)
4589                 set_inode_flag(inode, FI_UPDATE_WRITE);
4590
4591         if (iov_iter_count(from)) {
4592                 ssize_t ret2;
4593                 loff_t bufio_start_pos = iocb->ki_pos;
4594
4595                 /*
4596                  * The direct write was partial, so we need to fall back to a
4597                  * buffered write for the remainder.
4598                  */
4599
4600                 ret2 = f2fs_buffered_write_iter(iocb, from);
4601                 if (iov_iter_count(from))
4602                         f2fs_write_failed(inode, iocb->ki_pos);
4603                 if (ret2 < 0)
4604                         goto out;
4605
4606                 /*
4607                  * Ensure that the pagecache pages are written to disk and
4608                  * invalidated to preserve the expected O_DIRECT semantics.
4609                  */
4610                 if (ret2 > 0) {
4611                         loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4612
4613                         ret += ret2;
4614
4615                         ret2 = filemap_write_and_wait_range(file->f_mapping,
4616                                                             bufio_start_pos,
4617                                                             bufio_end_pos);
4618                         if (ret2 < 0)
4619                                 goto out;
4620                         invalidate_mapping_pages(file->f_mapping,
4621                                                  bufio_start_pos >> PAGE_SHIFT,
4622                                                  bufio_end_pos >> PAGE_SHIFT);
4623                 }
4624         } else {
4625                 /* iomap_dio_rw() already handled the generic_write_sync(). */
4626                 *may_need_sync = false;
4627         }
4628 out:
4629         trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4630         return ret;
4631 }
4632
4633 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4634 {
4635         struct inode *inode = file_inode(iocb->ki_filp);
4636         const loff_t orig_pos = iocb->ki_pos;
4637         const size_t orig_count = iov_iter_count(from);
4638         loff_t target_size;
4639         bool dio;
4640         bool may_need_sync = true;
4641         int preallocated;
4642         ssize_t ret;
4643
4644         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4645                 ret = -EIO;
4646                 goto out;
4647         }
4648
4649         if (!f2fs_is_compress_backend_ready(inode)) {
4650                 ret = -EOPNOTSUPP;
4651                 goto out;
4652         }
4653
4654         if (iocb->ki_flags & IOCB_NOWAIT) {
4655                 if (!inode_trylock(inode)) {
4656                         ret = -EAGAIN;
4657                         goto out;
4658                 }
4659         } else {
4660                 inode_lock(inode);
4661         }
4662
4663         ret = f2fs_write_checks(iocb, from);
4664         if (ret <= 0)
4665                 goto out_unlock;
4666
4667         /* Determine whether we will do a direct write or a buffered write. */
4668         dio = f2fs_should_use_dio(inode, iocb, from);
4669
4670         /* Possibly preallocate the blocks for the write. */
4671         target_size = iocb->ki_pos + iov_iter_count(from);
4672         preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4673         if (preallocated < 0) {
4674                 ret = preallocated;
4675         } else {
4676                 if (trace_f2fs_datawrite_start_enabled()) {
4677                         char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4678                                                 PATH_MAX, GFP_KERNEL);
4679                         char *path;
4680
4681                         if (!p)
4682                                 goto skip_write_trace;
4683                         path = dentry_path_raw(file_dentry(iocb->ki_filp),
4684                                                                 p, PATH_MAX);
4685                         if (IS_ERR(path)) {
4686                                 kfree(p);
4687                                 goto skip_write_trace;
4688                         }
4689                         trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4690                                         current->pid, path, current->comm);
4691                         kfree(p);
4692                 }
4693 skip_write_trace:
4694                 /* Do the actual write. */
4695                 ret = dio ?
4696                         f2fs_dio_write_iter(iocb, from, &may_need_sync) :
4697                         f2fs_buffered_write_iter(iocb, from);
4698
4699                 if (trace_f2fs_datawrite_end_enabled())
4700                         trace_f2fs_datawrite_end(inode, orig_pos, ret);
4701         }
4702
4703         /* Don't leave any preallocated blocks around past i_size. */
4704         if (preallocated && i_size_read(inode) < target_size) {
4705                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4706                 filemap_invalidate_lock(inode->i_mapping);
4707                 if (!f2fs_truncate(inode))
4708                         file_dont_truncate(inode);
4709                 filemap_invalidate_unlock(inode->i_mapping);
4710                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4711         } else {
4712                 file_dont_truncate(inode);
4713         }
4714
4715         clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4716 out_unlock:
4717         inode_unlock(inode);
4718 out:
4719         trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4720         if (ret > 0 && may_need_sync)
4721                 ret = generic_write_sync(iocb, ret);
4722         return ret;
4723 }
4724
4725 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4726                 int advice)
4727 {
4728         struct address_space *mapping;
4729         struct backing_dev_info *bdi;
4730         struct inode *inode = file_inode(filp);
4731         int err;
4732
4733         if (advice == POSIX_FADV_SEQUENTIAL) {
4734                 if (S_ISFIFO(inode->i_mode))
4735                         return -ESPIPE;
4736
4737                 mapping = filp->f_mapping;
4738                 if (!mapping || len < 0)
4739                         return -EINVAL;
4740
4741                 bdi = inode_to_bdi(mapping->host);
4742                 filp->f_ra.ra_pages = bdi->ra_pages *
4743                         F2FS_I_SB(inode)->seq_file_ra_mul;
4744                 spin_lock(&filp->f_lock);
4745                 filp->f_mode &= ~FMODE_RANDOM;
4746                 spin_unlock(&filp->f_lock);
4747                 return 0;
4748         }
4749
4750         err = generic_fadvise(filp, offset, len, advice);
4751         if (!err && advice == POSIX_FADV_DONTNEED &&
4752                 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4753                 f2fs_compressed_file(inode))
4754                 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4755
4756         return err;
4757 }
4758
4759 #ifdef CONFIG_COMPAT
4760 struct compat_f2fs_gc_range {
4761         u32 sync;
4762         compat_u64 start;
4763         compat_u64 len;
4764 };
4765 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE        _IOW(F2FS_IOCTL_MAGIC, 11,\
4766                                                 struct compat_f2fs_gc_range)
4767
4768 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4769 {
4770         struct compat_f2fs_gc_range __user *urange;
4771         struct f2fs_gc_range range;
4772         int err;
4773
4774         urange = compat_ptr(arg);
4775         err = get_user(range.sync, &urange->sync);
4776         err |= get_user(range.start, &urange->start);
4777         err |= get_user(range.len, &urange->len);
4778         if (err)
4779                 return -EFAULT;
4780
4781         return __f2fs_ioc_gc_range(file, &range);
4782 }
4783
4784 struct compat_f2fs_move_range {
4785         u32 dst_fd;
4786         compat_u64 pos_in;
4787         compat_u64 pos_out;
4788         compat_u64 len;
4789 };
4790 #define F2FS_IOC32_MOVE_RANGE           _IOWR(F2FS_IOCTL_MAGIC, 9,      \
4791                                         struct compat_f2fs_move_range)
4792
4793 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4794 {
4795         struct compat_f2fs_move_range __user *urange;
4796         struct f2fs_move_range range;
4797         int err;
4798
4799         urange = compat_ptr(arg);
4800         err = get_user(range.dst_fd, &urange->dst_fd);
4801         err |= get_user(range.pos_in, &urange->pos_in);
4802         err |= get_user(range.pos_out, &urange->pos_out);
4803         err |= get_user(range.len, &urange->len);
4804         if (err)
4805                 return -EFAULT;
4806
4807         return __f2fs_ioc_move_range(file, &range);
4808 }
4809
4810 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4811 {
4812         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4813                 return -EIO;
4814         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4815                 return -ENOSPC;
4816
4817         switch (cmd) {
4818         case FS_IOC32_GETVERSION:
4819                 cmd = FS_IOC_GETVERSION;
4820                 break;
4821         case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4822                 return f2fs_compat_ioc_gc_range(file, arg);
4823         case F2FS_IOC32_MOVE_RANGE:
4824                 return f2fs_compat_ioc_move_range(file, arg);
4825         case F2FS_IOC_START_ATOMIC_WRITE:
4826         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4827         case F2FS_IOC_START_VOLATILE_WRITE:
4828         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4829         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4830         case F2FS_IOC_SHUTDOWN:
4831         case FITRIM:
4832         case FS_IOC_SET_ENCRYPTION_POLICY:
4833         case FS_IOC_GET_ENCRYPTION_PWSALT:
4834         case FS_IOC_GET_ENCRYPTION_POLICY:
4835         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4836         case FS_IOC_ADD_ENCRYPTION_KEY:
4837         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4838         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4839         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4840         case FS_IOC_GET_ENCRYPTION_NONCE:
4841         case F2FS_IOC_GARBAGE_COLLECT:
4842         case F2FS_IOC_WRITE_CHECKPOINT:
4843         case F2FS_IOC_DEFRAGMENT:
4844         case F2FS_IOC_FLUSH_DEVICE:
4845         case F2FS_IOC_GET_FEATURES:
4846         case F2FS_IOC_GET_PIN_FILE:
4847         case F2FS_IOC_SET_PIN_FILE:
4848         case F2FS_IOC_PRECACHE_EXTENTS:
4849         case F2FS_IOC_RESIZE_FS:
4850         case FS_IOC_ENABLE_VERITY:
4851         case FS_IOC_MEASURE_VERITY:
4852         case FS_IOC_READ_VERITY_METADATA:
4853         case FS_IOC_GETFSLABEL:
4854         case FS_IOC_SETFSLABEL:
4855         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4856         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4857         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4858         case F2FS_IOC_SEC_TRIM_FILE:
4859         case F2FS_IOC_GET_COMPRESS_OPTION:
4860         case F2FS_IOC_SET_COMPRESS_OPTION:
4861         case F2FS_IOC_DECOMPRESS_FILE:
4862         case F2FS_IOC_COMPRESS_FILE:
4863                 break;
4864         default:
4865                 return -ENOIOCTLCMD;
4866         }
4867         return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4868 }
4869 #endif
4870
4871 const struct file_operations f2fs_file_operations = {
4872         .llseek         = f2fs_llseek,
4873         .read_iter      = f2fs_file_read_iter,
4874         .write_iter     = f2fs_file_write_iter,
4875         .open           = f2fs_file_open,
4876         .release        = f2fs_release_file,
4877         .mmap           = f2fs_file_mmap,
4878         .flush          = f2fs_file_flush,
4879         .fsync          = f2fs_sync_file,
4880         .fallocate      = f2fs_fallocate,
4881         .unlocked_ioctl = f2fs_ioctl,
4882 #ifdef CONFIG_COMPAT
4883         .compat_ioctl   = f2fs_compat_ioctl,
4884 #endif
4885         .splice_read    = generic_file_splice_read,
4886         .splice_write   = iter_file_splice_write,
4887         .fadvise        = f2fs_file_fadvise,
4888 };