fs/f2fs/file.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * fs/f2fs/file.c
   4  *
   5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6  *             http://www.samsung.com/
   7  */
   8 #include <linux/fs.h>
   9 #include <linux/f2fs_fs.h>
  10 #include <linux/stat.h>
  11 #include <linux/buffer_head.h>
  12 #include <linux/writeback.h>
  13 #include <linux/blkdev.h>
  14 #include <linux/falloc.h>
  15 #include <linux/types.h>
  16 #include <linux/compat.h>
  17 #include <linux/uaccess.h>
  18 #include <linux/mount.h>
  19 #include <linux/pagevec.h>
  20 #include <linux/uio.h>
  21 #include <linux/uuid.h>
  22 #include <linux/file.h>
  23 #include <linux/nls.h>
  24 #include <linux/sched/signal.h>
  25 #include <linux/fileattr.h>
  26 #include <linux/fadvise.h>
  27 #include <linux/iomap.h>
  28
  29 #include "f2fs.h"
  30 #include "node.h"
  31 #include "segment.h"
  32 #include "xattr.h"
  33 #include "acl.h"
  34 #include "gc.h"
  35 #include "iostat.h"
  36 #include <trace/events/f2fs.h>
  37 #include <uapi/linux/f2fs.h>
  38
  39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
  40 {
  41         struct inode *inode = file_inode(vmf->vma->vm_file);
  42         vm_fault_t ret;
  43
  44         ret = filemap_fault(vmf);
  45         if (!ret)
  46                 f2fs_update_iostat(F2FS_I_SB(inode), inode,
  47                                         APP_MAPPED_READ_IO, F2FS_BLKSIZE);
  48
  49         trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
  50
  51         return ret;
  52 }
  53
  54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
  55 {
  56         struct page *page = vmf->page;
  57         struct inode *inode = file_inode(vmf->vma->vm_file);
  58         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
  59         struct dnode_of_data dn;
  60         bool need_alloc = true;
  61         int err = 0;
  62
  63         if (unlikely(IS_IMMUTABLE(inode)))
  64                 return VM_FAULT_SIGBUS;
  65
  66         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
  67                 return VM_FAULT_SIGBUS;
  68
  69         if (unlikely(f2fs_cp_error(sbi))) {
  70                 err = -EIO;
  71                 goto err;
  72         }
  73
  74         if (!f2fs_is_checkpoint_ready(sbi)) {
  75                 err = -ENOSPC;
  76                 goto err;
  77         }
  78
  79         err = f2fs_convert_inline_inode(inode);
  80         if (err)
  81                 goto err;
  82
  83 #ifdef CONFIG_F2FS_FS_COMPRESSION
  84         if (f2fs_compressed_file(inode)) {
  85                 int ret = f2fs_is_compressed_cluster(inode, page->index);
  86
  87                 if (ret < 0) {
  88                         err = ret;
  89                         goto err;
  90                 } else if (ret) {
  91                         need_alloc = false;
  92                 }
  93         }
  94 #endif
  95         /* should do out of any locked page */
  96         if (need_alloc)
  97                 f2fs_balance_fs(sbi, true);
  98
  99         sb_start_pagefault(inode->i_sb);
 100
 101         f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
 102
 103         file_update_time(vmf->vma->vm_file);
 104         filemap_invalidate_lock_shared(inode->i_mapping);
 105         lock_page(page);
 106         if (unlikely(page->mapping != inode->i_mapping ||
 107                         page_offset(page) > i_size_read(inode) ||
 108                         !PageUptodate(page))) {
 109                 unlock_page(page);
 110                 err = -EFAULT;
 111                 goto out_sem;
 112         }
 113
 114         if (need_alloc) {
 115                 /* block allocation */
 116                 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
 117                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 118                 err = f2fs_get_block(&dn, page->index);
 119                 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
 120         }
 121
 122 #ifdef CONFIG_F2FS_FS_COMPRESSION
 123         if (!need_alloc) {
 124                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 125                 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
 126                 f2fs_put_dnode(&dn);
 127         }
 128 #endif
 129         if (err) {
 130                 unlock_page(page);
 131                 goto out_sem;
 132         }
 133
 134         f2fs_wait_on_page_writeback(page, DATA, false, true);
 135
 136         /* wait for GCed page writeback via META_MAPPING */
 137         f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
 138
 139         /*
 140          * check to see if the page is mapped already (no holes)
 141          */
 142         if (PageMappedToDisk(page))
 143                 goto out_sem;
 144
 145         /* page is wholly or partially inside EOF */
 146         if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
 147                                                 i_size_read(inode)) {
 148                 loff_t offset;
 149
 150                 offset = i_size_read(inode) & ~PAGE_MASK;
 151                 zero_user_segment(page, offset, PAGE_SIZE);
 152         }
 153         set_page_dirty(page);
 154         if (!PageUptodate(page))
 155                 SetPageUptodate(page);
 156
 157         f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
 158         f2fs_update_time(sbi, REQ_TIME);
 159
 160         trace_f2fs_vm_page_mkwrite(page, DATA);
 161 out_sem:
 162         filemap_invalidate_unlock_shared(inode->i_mapping);
 163
 164         sb_end_pagefault(inode->i_sb);
 165 err:
 166         return block_page_mkwrite_return(err);
 167 }
 168
 169 static const struct vm_operations_struct f2fs_file_vm_ops = {
 170         .fault          = f2fs_filemap_fault,
 171         .map_pages      = filemap_map_pages,
 172         .page_mkwrite   = f2fs_vm_page_mkwrite,
 173 };
 174
 175 static int get_parent_ino(struct inode *inode, nid_t *pino)
 176 {
 177         struct dentry *dentry;
 178
 179         /*
 180          * Make sure to get the non-deleted alias.  The alias associated with
 181          * the open file descriptor being fsync()'ed may be deleted already.
 182          */
 183         dentry = d_find_alias(inode);
 184         if (!dentry)
 185                 return 0;
 186
 187         *pino = parent_ino(dentry);
 188         dput(dentry);
 189         return 1;
 190 }
 191
 192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
 193 {
 194         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 195         enum cp_reason_type cp_reason = CP_NO_NEEDED;
 196
 197         if (!S_ISREG(inode->i_mode))
 198                 cp_reason = CP_NON_REGULAR;
 199         else if (f2fs_compressed_file(inode))
 200                 cp_reason = CP_COMPRESSED;
 201         else if (inode->i_nlink != 1)
 202                 cp_reason = CP_HARDLINK;
 203         else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
 204                 cp_reason = CP_SB_NEED_CP;
 205         else if (file_wrong_pino(inode))
 206                 cp_reason = CP_WRONG_PINO;
 207         else if (!f2fs_space_for_roll_forward(sbi))
 208                 cp_reason = CP_NO_SPC_ROLL;
 209         else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
 210                 cp_reason = CP_NODE_NEED_CP;
 211         else if (test_opt(sbi, FASTBOOT))
 212                 cp_reason = CP_FASTBOOT_MODE;
 213         else if (F2FS_OPTION(sbi).active_logs == 2)
 214                 cp_reason = CP_SPEC_LOG_NUM;
 215         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
 216                 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
 217                 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
 218                                                         TRANS_DIR_INO))
 219                 cp_reason = CP_RECOVER_DIR;
 220
 221         return cp_reason;
 222 }
 223
 224 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
 225 {
 226         struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
 227         bool ret = false;
 228         /* But we need to avoid that there are some inode updates */
 229         if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
 230                 ret = true;
 231         f2fs_put_page(i, 0);
 232         return ret;
 233 }
 234
 235 static void try_to_fix_pino(struct inode *inode)
 236 {
 237         struct f2fs_inode_info *fi = F2FS_I(inode);
 238         nid_t pino;
 239
 240         f2fs_down_write(&fi->i_sem);
 241         if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
 242                         get_parent_ino(inode, &pino)) {
 243                 f2fs_i_pino_write(inode, pino);
 244                 file_got_pino(inode);
 245         }
 246         f2fs_up_write(&fi->i_sem);
 247 }
 248
 249 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
 250                                                 int datasync, bool atomic)
 251 {
 252         struct inode *inode = file->f_mapping->host;
 253         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 254         nid_t ino = inode->i_ino;
 255         int ret = 0;
 256         enum cp_reason_type cp_reason = 0;
 257         struct writeback_control wbc = {
 258                 .sync_mode = WB_SYNC_ALL,
 259                 .nr_to_write = LONG_MAX,
 260                 .for_reclaim = 0,
 261         };
 262         unsigned int seq_id = 0;
 263
 264         if (unlikely(f2fs_readonly(inode->i_sb)))
 265                 return 0;
 266
 267         trace_f2fs_sync_file_enter(inode);
 268
 269         if (S_ISDIR(inode->i_mode))
 270                 goto go_write;
 271
 272         /* if fdatasync is triggered, let's do in-place-update */
 273         if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
 274                 set_inode_flag(inode, FI_NEED_IPU);
 275         ret = file_write_and_wait_range(file, start, end);
 276         clear_inode_flag(inode, FI_NEED_IPU);
 277
 278         if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
 279                 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 280                 return ret;
 281         }
 282
 283         /* if the inode is dirty, let's recover all the time */
 284         if (!f2fs_skip_inode_update(inode, datasync)) {
 285                 f2fs_write_inode(inode, NULL);
 286                 goto go_write;
 287         }
 288
 289         /*
 290          * if there is no written data, don't waste time to write recovery info.
 291          */
 292         if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
 293                         !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
 294
 295                 /* it may call write_inode just prior to fsync */
 296                 if (need_inode_page_update(sbi, ino))
 297                         goto go_write;
 298
 299                 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
 300                                 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
 301                         goto flush_out;
 302                 goto out;
 303         } else {
 304                 /*
 305                  * for OPU case, during fsync(), node can be persisted before
 306                  * data when lower device doesn't support write barrier, result
 307                  * in data corruption after SPO.
 308                  * So for strict fsync mode, force to use atomic write sematics
 309                  * to keep write order in between data/node and last node to
 310                  * avoid potential data corruption.
 311                  */
 312                 if (F2FS_OPTION(sbi).fsync_mode ==
 313                                 FSYNC_MODE_STRICT && !atomic)
 314                         atomic = true;
 315         }
 316 go_write:
 317         /*
 318          * Both of fdatasync() and fsync() are able to be recovered from
 319          * sudden-power-off.
 320          */
 321         f2fs_down_read(&F2FS_I(inode)->i_sem);
 322         cp_reason = need_do_checkpoint(inode);
 323         f2fs_up_read(&F2FS_I(inode)->i_sem);
 324
 325         if (cp_reason) {
 326                 /* all the dirty node pages should be flushed for POR */
 327                 ret = f2fs_sync_fs(inode->i_sb, 1);
 328
 329                 /*
 330                  * We've secured consistency through sync_fs. Following pino
 331                  * will be used only for fsynced inodes after checkpoint.
 332                  */
 333                 try_to_fix_pino(inode);
 334                 clear_inode_flag(inode, FI_APPEND_WRITE);
 335                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 336                 goto out;
 337         }
 338 sync_nodes:
 339         atomic_inc(&sbi->wb_sync_req[NODE]);
 340         ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
 341         atomic_dec(&sbi->wb_sync_req[NODE]);
 342         if (ret)
 343                 goto out;
 344
 345         /* if cp_error was enabled, we should avoid infinite loop */
 346         if (unlikely(f2fs_cp_error(sbi))) {
 347                 ret = -EIO;
 348                 goto out;
 349         }
 350
 351         if (f2fs_need_inode_block_update(sbi, ino)) {
 352                 f2fs_mark_inode_dirty_sync(inode, true);
 353                 f2fs_write_inode(inode, NULL);
 354                 goto sync_nodes;
 355         }
 356
 357         /*
 358          * If it's atomic_write, it's just fine to keep write ordering. So
 359          * here we don't need to wait for node write completion, since we use
 360          * node chain which serializes node blocks. If one of node writes are
 361          * reordered, we can see simply broken chain, resulting in stopping
 362          * roll-forward recovery. It means we'll recover all or none node blocks
 363          * given fsync mark.
 364          */
 365         if (!atomic) {
 366                 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
 367                 if (ret)
 368                         goto out;
 369         }
 370
 371         /* once recovery info is written, don't need to tack this */
 372         f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
 373         clear_inode_flag(inode, FI_APPEND_WRITE);
 374 flush_out:
 375         if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
 376             (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
 377                 ret = f2fs_issue_flush(sbi, inode->i_ino);
 378         if (!ret) {
 379                 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
 380                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 381                 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
 382         }
 383         f2fs_update_time(sbi, REQ_TIME);
 384 out:
 385         trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 386         return ret;
 387 }
 388
 389 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 390 {
 391         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
 392                 return -EIO;
 393         return f2fs_do_sync_file(file, start, end, datasync, false);
 394 }
 395
 396 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
 397                                 pgoff_t index, int whence)
 398 {
 399         switch (whence) {
 400         case SEEK_DATA:
 401                 if (__is_valid_data_blkaddr(blkaddr))
 402                         return true;
 403                 if (blkaddr == NEW_ADDR &&
 404                     xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
 405                         return true;
 406                 break;
 407         case SEEK_HOLE:
 408                 if (blkaddr == NULL_ADDR)
 409                         return true;
 410                 break;
 411         }
 412         return false;
 413 }
 414
 415 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 416 {
 417         struct inode *inode = file->f_mapping->host;
 418         loff_t maxbytes = inode->i_sb->s_maxbytes;
 419         struct dnode_of_data dn;
 420         pgoff_t pgofs, end_offset;
 421         loff_t data_ofs = offset;
 422         loff_t isize;
 423         int err = 0;
 424
 425         inode_lock(inode);
 426
 427         isize = i_size_read(inode);
 428         if (offset >= isize)
 429                 goto fail;
 430
 431         /* handle inline data case */
 432         if (f2fs_has_inline_data(inode)) {
 433                 if (whence == SEEK_HOLE) {
 434                         data_ofs = isize;
 435                         goto found;
 436                 } else if (whence == SEEK_DATA) {
 437                         data_ofs = offset;
 438                         goto found;
 439                 }
 440         }
 441
 442         pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
 443
 444         for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 445                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 446                 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
 447                 if (err && err != -ENOENT) {
 448                         goto fail;
 449                 } else if (err == -ENOENT) {
 450                         /* direct node does not exists */
 451                         if (whence == SEEK_DATA) {
 452                                 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
 453                                 continue;
 454                         } else {
 455                                 goto found;
 456                         }
 457                 }
 458
 459                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 460
 461                 /* find data/hole in dnode block */
 462                 for (; dn.ofs_in_node < end_offset;
 463                                 dn.ofs_in_node++, pgofs++,
 464                                 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 465                         block_t blkaddr;
 466
 467                         blkaddr = f2fs_data_blkaddr(&dn);
 468
 469                         if (__is_valid_data_blkaddr(blkaddr) &&
 470                                 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
 471                                         blkaddr, DATA_GENERIC_ENHANCE)) {
 472                                 f2fs_put_dnode(&dn);
 473                                 goto fail;
 474                         }
 475
 476                         if (__found_offset(file->f_mapping, blkaddr,
 477                                                         pgofs, whence)) {
 478                                 f2fs_put_dnode(&dn);
 479                                 goto found;
 480                         }
 481                 }
 482                 f2fs_put_dnode(&dn);
 483         }
 484
 485         if (whence == SEEK_DATA)
 486                 goto fail;
 487 found:
 488         if (whence == SEEK_HOLE && data_ofs > isize)
 489                 data_ofs = isize;
 490         inode_unlock(inode);
 491         return vfs_setpos(file, data_ofs, maxbytes);
 492 fail:
 493         inode_unlock(inode);
 494         return -ENXIO;
 495 }
 496
 497 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
 498 {
 499         struct inode *inode = file->f_mapping->host;
 500         loff_t maxbytes = inode->i_sb->s_maxbytes;
 501
 502         if (f2fs_compressed_file(inode))
 503                 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
 504
 505         switch (whence) {
 506         case SEEK_SET:
 507         case SEEK_CUR:
 508         case SEEK_END:
 509                 return generic_file_llseek_size(file, offset, whence,
 510                                                 maxbytes, i_size_read(inode));
 511         case SEEK_DATA:
 512         case SEEK_HOLE:
 513                 if (offset < 0)
 514                         return -ENXIO;
 515                 return f2fs_seek_block(file, offset, whence);
 516         }
 517
 518         return -EINVAL;
 519 }
 520
 521 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 522 {
 523         struct inode *inode = file_inode(file);
 524
 525         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 526                 return -EIO;
 527
 528         if (!f2fs_is_compress_backend_ready(inode))
 529                 return -EOPNOTSUPP;
 530
 531         file_accessed(file);
 532         vma->vm_ops = &f2fs_file_vm_ops;
 533         set_inode_flag(inode, FI_MMAP_FILE);
 534         return 0;
 535 }
 536
 537 static int f2fs_file_open(struct inode *inode, struct file *filp)
 538 {
 539         int err = fscrypt_file_open(inode, filp);
 540
 541         if (err)
 542                 return err;
 543
 544         if (!f2fs_is_compress_backend_ready(inode))
 545                 return -EOPNOTSUPP;
 546
 547         err = fsverity_file_open(inode, filp);
 548         if (err)
 549                 return err;
 550
 551         filp->f_mode |= FMODE_NOWAIT;
 552
 553         return dquot_file_open(inode, filp);
 554 }
 555
 556 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 557 {
 558         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 559         struct f2fs_node *raw_node;
 560         int nr_free = 0, ofs = dn->ofs_in_node, len = count;
 561         __le32 *addr;
 562         int base = 0;
 563         bool compressed_cluster = false;
 564         int cluster_index = 0, valid_blocks = 0;
 565         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
 566         bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
 567
 568         if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
 569                 base = get_extra_isize(dn->inode);
 570
 571         raw_node = F2FS_NODE(dn->node_page);
 572         addr = blkaddr_in_node(raw_node) + base + ofs;
 573
 574         /* Assumption: truncateion starts with cluster */
 575         for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
 576                 block_t blkaddr = le32_to_cpu(*addr);
 577
 578                 if (f2fs_compressed_file(dn->inode) &&
 579                                         !(cluster_index & (cluster_size - 1))) {
 580                         if (compressed_cluster)
 581                                 f2fs_i_compr_blocks_update(dn->inode,
 582                                                         valid_blocks, false);
 583                         compressed_cluster = (blkaddr == COMPRESS_ADDR);
 584                         valid_blocks = 0;
 585                 }
 586
 587                 if (blkaddr == NULL_ADDR)
 588                         continue;
 589
 590                 dn->data_blkaddr = NULL_ADDR;
 591                 f2fs_set_data_blkaddr(dn);
 592
 593                 if (__is_valid_data_blkaddr(blkaddr)) {
 594                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
 595                                         DATA_GENERIC_ENHANCE))
 596                                 continue;
 597                         if (compressed_cluster)
 598                                 valid_blocks++;
 599                 }
 600
 601                 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 602                         clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 603
 604                 f2fs_invalidate_blocks(sbi, blkaddr);
 605
 606                 if (!released || blkaddr != COMPRESS_ADDR)
 607                         nr_free++;
 608         }
 609
 610         if (compressed_cluster)
 611                 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
 612
 613         if (nr_free) {
 614                 pgoff_t fofs;
 615                 /*
 616                  * once we invalidate valid blkaddr in range [ofs, ofs + count],
 617                  * we will invalidate all blkaddr in the whole range.
 618                  */
 619                 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
 620                                                         dn->inode) + ofs;
 621                 f2fs_update_extent_cache_range(dn, fofs, 0, len);
 622                 dec_valid_block_count(sbi, dn->inode, nr_free);
 623         }
 624         dn->ofs_in_node = ofs;
 625
 626         f2fs_update_time(sbi, REQ_TIME);
 627         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 628                                          dn->ofs_in_node, nr_free);
 629 }
 630
 631 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
 632 {
 633         f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
 634 }
 635
 636 static int truncate_partial_data_page(struct inode *inode, u64 from,
 637                                                                 bool cache_only)
 638 {
 639         loff_t offset = from & (PAGE_SIZE - 1);
 640         pgoff_t index = from >> PAGE_SHIFT;
 641         struct address_space *mapping = inode->i_mapping;
 642         struct page *page;
 643
 644         if (!offset && !cache_only)
 645                 return 0;
 646
 647         if (cache_only) {
 648                 page = find_lock_page(mapping, index);
 649                 if (page && PageUptodate(page))
 650                         goto truncate_out;
 651                 f2fs_put_page(page, 1);
 652                 return 0;
 653         }
 654
 655         page = f2fs_get_lock_data_page(inode, index, true);
 656         if (IS_ERR(page))
 657                 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 658 truncate_out:
 659         f2fs_wait_on_page_writeback(page, DATA, true, true);
 660         zero_user(page, offset, PAGE_SIZE - offset);
 661
 662         /* An encrypted inode should have a key and truncate the last page. */
 663         f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
 664         if (!cache_only)
 665                 set_page_dirty(page);
 666         f2fs_put_page(page, 1);
 667         return 0;
 668 }
 669
 670 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
 671 {
 672         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 673         struct dnode_of_data dn;
 674         pgoff_t free_from;
 675         int count = 0, err = 0;
 676         struct page *ipage;
 677         bool truncate_page = false;
 678
 679         trace_f2fs_truncate_blocks_enter(inode, from);
 680
 681         free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 682
 683         if (free_from >= max_file_blocks(inode))
 684                 goto free_partial;
 685
 686         if (lock)
 687                 f2fs_lock_op(sbi);
 688
 689         ipage = f2fs_get_node_page(sbi, inode->i_ino);
 690         if (IS_ERR(ipage)) {
 691                 err = PTR_ERR(ipage);
 692                 goto out;
 693         }
 694
 695         if (f2fs_has_inline_data(inode)) {
 696                 f2fs_truncate_inline_inode(inode, ipage, from);
 697                 f2fs_put_page(ipage, 1);
 698                 truncate_page = true;
 699                 goto out;
 700         }
 701
 702         set_new_dnode(&dn, inode, ipage, NULL, 0);
 703         err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 704         if (err) {
 705                 if (err == -ENOENT)
 706                         goto free_next;
 707                 goto out;
 708         }
 709
 710         count = ADDRS_PER_PAGE(dn.node_page, inode);
 711
 712         count -= dn.ofs_in_node;
 713         f2fs_bug_on(sbi, count < 0);
 714
 715         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 716                 f2fs_truncate_data_blocks_range(&dn, count);
 717                 free_from += count;
 718         }
 719
 720         f2fs_put_dnode(&dn);
 721 free_next:
 722         err = f2fs_truncate_inode_blocks(inode, free_from);
 723 out:
 724         if (lock)
 725                 f2fs_unlock_op(sbi);
 726 free_partial:
 727         /* lastly zero out the first data page */
 728         if (!err)
 729                 err = truncate_partial_data_page(inode, from, truncate_page);
 730
 731         trace_f2fs_truncate_blocks_exit(inode, err);
 732         return err;
 733 }
 734
 735 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
 736 {
 737         u64 free_from = from;
 738         int err;
 739
 740 #ifdef CONFIG_F2FS_FS_COMPRESSION
 741         /*
 742          * for compressed file, only support cluster size
 743          * aligned truncation.
 744          */
 745         if (f2fs_compressed_file(inode))
 746                 free_from = round_up(from,
 747                                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
 748 #endif
 749
 750         err = f2fs_do_truncate_blocks(inode, free_from, lock);
 751         if (err)
 752                 return err;
 753
 754 #ifdef CONFIG_F2FS_FS_COMPRESSION
 755         /*
 756          * For compressed file, after release compress blocks, don't allow write
 757          * direct, but we should allow write direct after truncate to zero.
 758          */
 759         if (f2fs_compressed_file(inode) && !free_from
 760                         && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
 761                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 762
 763         if (from != free_from) {
 764                 err = f2fs_truncate_partial_cluster(inode, from, lock);
 765                 if (err)
 766                         return err;
 767         }
 768 #endif
 769
 770         return 0;
 771 }
 772
 773 int f2fs_truncate(struct inode *inode)
 774 {
 775         int err;
 776
 777         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 778                 return -EIO;
 779
 780         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 781                                 S_ISLNK(inode->i_mode)))
 782                 return 0;
 783
 784         trace_f2fs_truncate(inode);
 785
 786         if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
 787                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
 788                 return -EIO;
 789         }
 790
 791         err = f2fs_dquot_initialize(inode);
 792         if (err)
 793                 return err;
 794
 795         /* we should check inline_data size */
 796         if (!f2fs_may_inline_data(inode)) {
 797                 err = f2fs_convert_inline_inode(inode);
 798                 if (err)
 799                         return err;
 800         }
 801
 802         err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
 803         if (err)
 804                 return err;
 805
 806         inode->i_mtime = inode->i_ctime = current_time(inode);
 807         f2fs_mark_inode_dirty_sync(inode, false);
 808         return 0;
 809 }
 810
 811 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
 812 {
 813         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 814
 815         if (!fscrypt_dio_supported(inode))
 816                 return true;
 817         if (fsverity_active(inode))
 818                 return true;
 819         if (f2fs_compressed_file(inode))
 820                 return true;
 821
 822         /* disallow direct IO if any of devices has unaligned blksize */
 823         if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
 824                 return true;
 825         /*
 826          * for blkzoned device, fallback direct IO to buffered IO, so
 827          * all IOs can be serialized by log-structured write.
 828          */
 829         if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
 830                 return true;
 831         if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi))
 832                 return true;
 833         if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
 834                 return true;
 835
 836         return false;
 837 }
 838
 839 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 840                  struct kstat *stat, u32 request_mask, unsigned int query_flags)
 841 {
 842         struct inode *inode = d_inode(path->dentry);
 843         struct f2fs_inode_info *fi = F2FS_I(inode);
 844         struct f2fs_inode *ri = NULL;
 845         unsigned int flags;
 846
 847         if (f2fs_has_extra_attr(inode) &&
 848                         f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
 849                         F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 850                 stat->result_mask |= STATX_BTIME;
 851                 stat->btime.tv_sec = fi->i_crtime.tv_sec;
 852                 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 853         }
 854
 855         /*
 856          * Return the DIO alignment restrictions if requested.  We only return
 857          * this information when requested, since on encrypted files it might
 858          * take a fair bit of work to get if the file wasn't opened recently.
 859          *
 860          * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
 861          * cannot represent that, so in that case we report no DIO support.
 862          */
 863         if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
 864                 unsigned int bsize = i_blocksize(inode);
 865
 866                 stat->result_mask |= STATX_DIOALIGN;
 867                 if (!f2fs_force_buffered_io(inode, WRITE)) {
 868                         stat->dio_mem_align = bsize;
 869                         stat->dio_offset_align = bsize;
 870                 }
 871         }
 872
 873         flags = fi->i_flags;
 874         if (flags & F2FS_COMPR_FL)
 875                 stat->attributes |= STATX_ATTR_COMPRESSED;
 876         if (flags & F2FS_APPEND_FL)
 877                 stat->attributes |= STATX_ATTR_APPEND;
 878         if (IS_ENCRYPTED(inode))
 879                 stat->attributes |= STATX_ATTR_ENCRYPTED;
 880         if (flags & F2FS_IMMUTABLE_FL)
 881                 stat->attributes |= STATX_ATTR_IMMUTABLE;
 882         if (flags & F2FS_NODUMP_FL)
 883                 stat->attributes |= STATX_ATTR_NODUMP;
 884         if (IS_VERITY(inode))
 885                 stat->attributes |= STATX_ATTR_VERITY;
 886
 887         stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 888                                   STATX_ATTR_APPEND |
 889                                   STATX_ATTR_ENCRYPTED |
 890                                   STATX_ATTR_IMMUTABLE |
 891                                   STATX_ATTR_NODUMP |
 892                                   STATX_ATTR_VERITY);
 893
 894         generic_fillattr(mnt_userns, inode, stat);
 895
 896         /* we need to show initial sectors used for inline_data/dentries */
 897         if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 898                                         f2fs_has_inline_dentry(inode))
 899                 stat->blocks += (stat->size + 511) >> 9;
 900
 901         return 0;
 902 }
 903
 904 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 905 static void __setattr_copy(struct user_namespace *mnt_userns,
 906                            struct inode *inode, const struct iattr *attr)
 907 {
 908         unsigned int ia_valid = attr->ia_valid;
 909
 910         i_uid_update(mnt_userns, attr, inode);
 911         i_gid_update(mnt_userns, attr, inode);
 912         if (ia_valid & ATTR_ATIME)
 913                 inode->i_atime = attr->ia_atime;
 914         if (ia_valid & ATTR_MTIME)
 915                 inode->i_mtime = attr->ia_mtime;
 916         if (ia_valid & ATTR_CTIME)
 917                 inode->i_ctime = attr->ia_ctime;
 918         if (ia_valid & ATTR_MODE) {
 919                 umode_t mode = attr->ia_mode;
 920                 vfsgid_t vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
 921
 922                 if (!vfsgid_in_group_p(vfsgid) &&
 923                     !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
 924                         mode &= ~S_ISGID;
 925                 set_acl_inode(inode, mode);
 926         }
 927 }
 928 #else
 929 #define __setattr_copy setattr_copy
 930 #endif
 931
 932 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 933                  struct iattr *attr)
 934 {
 935         struct inode *inode = d_inode(dentry);
 936         int err;
 937
 938         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 939                 return -EIO;
 940
 941         if (unlikely(IS_IMMUTABLE(inode)))
 942                 return -EPERM;
 943
 944         if (unlikely(IS_APPEND(inode) &&
 945                         (attr->ia_valid & (ATTR_MODE | ATTR_UID |
 946                                   ATTR_GID | ATTR_TIMES_SET))))
 947                 return -EPERM;
 948
 949         if ((attr->ia_valid & ATTR_SIZE) &&
 950                 !f2fs_is_compress_backend_ready(inode))
 951                 return -EOPNOTSUPP;
 952
 953         err = setattr_prepare(mnt_userns, dentry, attr);
 954         if (err)
 955                 return err;
 956
 957         err = fscrypt_prepare_setattr(dentry, attr);
 958         if (err)
 959                 return err;
 960
 961         err = fsverity_prepare_setattr(dentry, attr);
 962         if (err)
 963                 return err;
 964
 965         if (is_quota_modification(mnt_userns, inode, attr)) {
 966                 err = f2fs_dquot_initialize(inode);
 967                 if (err)
 968                         return err;
 969         }
 970         if (i_uid_needs_update(mnt_userns, attr, inode) ||
 971             i_gid_needs_update(mnt_userns, attr, inode)) {
 972                 f2fs_lock_op(F2FS_I_SB(inode));
 973                 err = dquot_transfer(mnt_userns, inode, attr);
 974                 if (err) {
 975                         set_sbi_flag(F2FS_I_SB(inode),
 976                                         SBI_QUOTA_NEED_REPAIR);
 977                         f2fs_unlock_op(F2FS_I_SB(inode));
 978                         return err;
 979                 }
 980                 /*
 981                  * update uid/gid under lock_op(), so that dquot and inode can
 982                  * be updated atomically.
 983                  */
 984                 i_uid_update(mnt_userns, attr, inode);
 985                 i_gid_update(mnt_userns, attr, inode);
 986                 f2fs_mark_inode_dirty_sync(inode, true);
 987                 f2fs_unlock_op(F2FS_I_SB(inode));
 988         }
 989
 990         if (attr->ia_valid & ATTR_SIZE) {
 991                 loff_t old_size = i_size_read(inode);
 992
 993                 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
 994                         /*
 995                          * should convert inline inode before i_size_write to
 996                          * keep smaller than inline_data size with inline flag.
 997                          */
 998                         err = f2fs_convert_inline_inode(inode);
 999                         if (err)
1000                                 return err;
1001                 }
1002
1003                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1004                 filemap_invalidate_lock(inode->i_mapping);
1005
1006                 truncate_setsize(inode, attr->ia_size);
1007
1008                 if (attr->ia_size <= old_size)
1009                         err = f2fs_truncate(inode);
1010                 /*
1011                  * do not trim all blocks after i_size if target size is
1012                  * larger than i_size.
1013                  */
1014                 filemap_invalidate_unlock(inode->i_mapping);
1015                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1016                 if (err)
1017                         return err;
1018
1019                 spin_lock(&F2FS_I(inode)->i_size_lock);
1020                 inode->i_mtime = inode->i_ctime = current_time(inode);
1021                 F2FS_I(inode)->last_disk_size = i_size_read(inode);
1022                 spin_unlock(&F2FS_I(inode)->i_size_lock);
1023         }
1024
1025         __setattr_copy(mnt_userns, inode, attr);
1026
1027         if (attr->ia_valid & ATTR_MODE) {
1028                 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
1029
1030                 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1031                         if (!err)
1032                                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
1033                         clear_inode_flag(inode, FI_ACL_MODE);
1034                 }
1035         }
1036
1037         /* file size may changed here */
1038         f2fs_mark_inode_dirty_sync(inode, true);
1039
1040         /* inode change will produce dirty node pages flushed by checkpoint */
1041         f2fs_balance_fs(F2FS_I_SB(inode), true);
1042
1043         return err;
1044 }
1045
1046 const struct inode_operations f2fs_file_inode_operations = {
1047         .getattr        = f2fs_getattr,
1048         .setattr        = f2fs_setattr,
1049         .get_acl        = f2fs_get_acl,
1050         .set_acl        = f2fs_set_acl,
1051         .listxattr      = f2fs_listxattr,
1052         .fiemap         = f2fs_fiemap,
1053         .fileattr_get   = f2fs_fileattr_get,
1054         .fileattr_set   = f2fs_fileattr_set,
1055 };
1056
1057 static int fill_zero(struct inode *inode, pgoff_t index,
1058                                         loff_t start, loff_t len)
1059 {
1060         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1061         struct page *page;
1062
1063         if (!len)
1064                 return 0;
1065
1066         f2fs_balance_fs(sbi, true);
1067
1068         f2fs_lock_op(sbi);
1069         page = f2fs_get_new_data_page(inode, NULL, index, false);
1070         f2fs_unlock_op(sbi);
1071
1072         if (IS_ERR(page))
1073                 return PTR_ERR(page);
1074
1075         f2fs_wait_on_page_writeback(page, DATA, true, true);
1076         zero_user(page, start, len);
1077         set_page_dirty(page);
1078         f2fs_put_page(page, 1);
1079         return 0;
1080 }
1081
1082 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1083 {
1084         int err;
1085
1086         while (pg_start < pg_end) {
1087                 struct dnode_of_data dn;
1088                 pgoff_t end_offset, count;
1089
1090                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1091                 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1092                 if (err) {
1093                         if (err == -ENOENT) {
1094                                 pg_start = f2fs_get_next_page_offset(&dn,
1095                                                                 pg_start);
1096                                 continue;
1097                         }
1098                         return err;
1099                 }
1100
1101                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1102                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1103
1104                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1105
1106                 f2fs_truncate_data_blocks_range(&dn, count);
1107                 f2fs_put_dnode(&dn);
1108
1109                 pg_start += count;
1110         }
1111         return 0;
1112 }
1113
1114 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1115 {
1116         pgoff_t pg_start, pg_end;
1117         loff_t off_start, off_end;
1118         int ret;
1119
1120         ret = f2fs_convert_inline_inode(inode);
1121         if (ret)
1122                 return ret;
1123
1124         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1125         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1126
1127         off_start = offset & (PAGE_SIZE - 1);
1128         off_end = (offset + len) & (PAGE_SIZE - 1);
1129
1130         if (pg_start == pg_end) {
1131                 ret = fill_zero(inode, pg_start, off_start,
1132                                                 off_end - off_start);
1133                 if (ret)
1134                         return ret;
1135         } else {
1136                 if (off_start) {
1137                         ret = fill_zero(inode, pg_start++, off_start,
1138                                                 PAGE_SIZE - off_start);
1139                         if (ret)
1140                                 return ret;
1141                 }
1142                 if (off_end) {
1143                         ret = fill_zero(inode, pg_end, 0, off_end);
1144                         if (ret)
1145                                 return ret;
1146                 }
1147
1148                 if (pg_start < pg_end) {
1149                         loff_t blk_start, blk_end;
1150                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1151
1152                         f2fs_balance_fs(sbi, true);
1153
1154                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
1155                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
1156
1157                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1158                         filemap_invalidate_lock(inode->i_mapping);
1159
1160                         truncate_pagecache_range(inode, blk_start, blk_end - 1);
1161
1162                         f2fs_lock_op(sbi);
1163                         ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1164                         f2fs_unlock_op(sbi);
1165
1166                         filemap_invalidate_unlock(inode->i_mapping);
1167                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1168                 }
1169         }
1170
1171         return ret;
1172 }
1173
1174 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1175                                 int *do_replace, pgoff_t off, pgoff_t len)
1176 {
1177         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1178         struct dnode_of_data dn;
1179         int ret, done, i;
1180
1181 next_dnode:
1182         set_new_dnode(&dn, inode, NULL, NULL, 0);
1183         ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1184         if (ret && ret != -ENOENT) {
1185                 return ret;
1186         } else if (ret == -ENOENT) {
1187                 if (dn.max_level == 0)
1188                         return -ENOENT;
1189                 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1190                                                 dn.ofs_in_node, len);
1191                 blkaddr += done;
1192                 do_replace += done;
1193                 goto next;
1194         }
1195
1196         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1197                                                         dn.ofs_in_node, len);
1198         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1199                 *blkaddr = f2fs_data_blkaddr(&dn);
1200
1201                 if (__is_valid_data_blkaddr(*blkaddr) &&
1202                         !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1203                                         DATA_GENERIC_ENHANCE)) {
1204                         f2fs_put_dnode(&dn);
1205                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1206                         return -EFSCORRUPTED;
1207                 }
1208
1209                 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1210
1211                         if (f2fs_lfs_mode(sbi)) {
1212                                 f2fs_put_dnode(&dn);
1213                                 return -EOPNOTSUPP;
1214                         }
1215
1216                         /* do not invalidate this block address */
1217                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1218                         *do_replace = 1;
1219                 }
1220         }
1221         f2fs_put_dnode(&dn);
1222 next:
1223         len -= done;
1224         off += done;
1225         if (len)
1226                 goto next_dnode;
1227         return 0;
1228 }
1229
1230 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1231                                 int *do_replace, pgoff_t off, int len)
1232 {
1233         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1234         struct dnode_of_data dn;
1235         int ret, i;
1236
1237         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1238                 if (*do_replace == 0)
1239                         continue;
1240
1241                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1242                 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1243                 if (ret) {
1244                         dec_valid_block_count(sbi, inode, 1);
1245                         f2fs_invalidate_blocks(sbi, *blkaddr);
1246                 } else {
1247                         f2fs_update_data_blkaddr(&dn, *blkaddr);
1248                 }
1249                 f2fs_put_dnode(&dn);
1250         }
1251         return 0;
1252 }
1253
1254 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1255                         block_t *blkaddr, int *do_replace,
1256                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1257 {
1258         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1259         pgoff_t i = 0;
1260         int ret;
1261
1262         while (i < len) {
1263                 if (blkaddr[i] == NULL_ADDR && !full) {
1264                         i++;
1265                         continue;
1266                 }
1267
1268                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1269                         struct dnode_of_data dn;
1270                         struct node_info ni;
1271                         size_t new_size;
1272                         pgoff_t ilen;
1273
1274                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1275                         ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1276                         if (ret)
1277                                 return ret;
1278
1279                         ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1280                         if (ret) {
1281                                 f2fs_put_dnode(&dn);
1282                                 return ret;
1283                         }
1284
1285                         ilen = min((pgoff_t)
1286                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1287                                                 dn.ofs_in_node, len - i);
1288                         do {
1289                                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1290                                 f2fs_truncate_data_blocks_range(&dn, 1);
1291
1292                                 if (do_replace[i]) {
1293                                         f2fs_i_blocks_write(src_inode,
1294                                                         1, false, false);
1295                                         f2fs_i_blocks_write(dst_inode,
1296                                                         1, true, false);
1297                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1298                                         blkaddr[i], ni.version, true, false);
1299
1300                                         do_replace[i] = 0;
1301                                 }
1302                                 dn.ofs_in_node++;
1303                                 i++;
1304                                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1305                                 if (dst_inode->i_size < new_size)
1306                                         f2fs_i_size_write(dst_inode, new_size);
1307                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1308
1309                         f2fs_put_dnode(&dn);
1310                 } else {
1311                         struct page *psrc, *pdst;
1312
1313                         psrc = f2fs_get_lock_data_page(src_inode,
1314                                                         src + i, true);
1315                         if (IS_ERR(psrc))
1316                                 return PTR_ERR(psrc);
1317                         pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1318                                                                 true);
1319                         if (IS_ERR(pdst)) {
1320                                 f2fs_put_page(psrc, 1);
1321                                 return PTR_ERR(pdst);
1322                         }
1323                         memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1324                         set_page_dirty(pdst);
1325                         f2fs_put_page(pdst, 1);
1326                         f2fs_put_page(psrc, 1);
1327
1328                         ret = f2fs_truncate_hole(src_inode,
1329                                                 src + i, src + i + 1);
1330                         if (ret)
1331                                 return ret;
1332                         i++;
1333                 }
1334         }
1335         return 0;
1336 }
1337
1338 static int __exchange_data_block(struct inode *src_inode,
1339                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1340                         pgoff_t len, bool full)
1341 {
1342         block_t *src_blkaddr;
1343         int *do_replace;
1344         pgoff_t olen;
1345         int ret;
1346
1347         while (len) {
1348                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1349
1350                 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1351                                         array_size(olen, sizeof(block_t)),
1352                                         GFP_NOFS);
1353                 if (!src_blkaddr)
1354                         return -ENOMEM;
1355
1356                 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1357                                         array_size(olen, sizeof(int)),
1358                                         GFP_NOFS);
1359                 if (!do_replace) {
1360                         kvfree(src_blkaddr);
1361                         return -ENOMEM;
1362                 }
1363
1364                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1365                                         do_replace, src, olen);
1366                 if (ret)
1367                         goto roll_back;
1368
1369                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1370                                         do_replace, src, dst, olen, full);
1371                 if (ret)
1372                         goto roll_back;
1373
1374                 src += olen;
1375                 dst += olen;
1376                 len -= olen;
1377
1378                 kvfree(src_blkaddr);
1379                 kvfree(do_replace);
1380         }
1381         return 0;
1382
1383 roll_back:
1384         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1385         kvfree(src_blkaddr);
1386         kvfree(do_replace);
1387         return ret;
1388 }
1389
1390 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1391 {
1392         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1393         pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1394         pgoff_t start = offset >> PAGE_SHIFT;
1395         pgoff_t end = (offset + len) >> PAGE_SHIFT;
1396         int ret;
1397
1398         f2fs_balance_fs(sbi, true);
1399
1400         /* avoid gc operation during block exchange */
1401         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1402         filemap_invalidate_lock(inode->i_mapping);
1403
1404         f2fs_lock_op(sbi);
1405         f2fs_drop_extent_tree(inode);
1406         truncate_pagecache(inode, offset);
1407         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1408         f2fs_unlock_op(sbi);
1409
1410         filemap_invalidate_unlock(inode->i_mapping);
1411         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1412         return ret;
1413 }
1414
1415 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1416 {
1417         loff_t new_size;
1418         int ret;
1419
1420         if (offset + len >= i_size_read(inode))
1421                 return -EINVAL;
1422
1423         /* collapse range should be aligned to block size of f2fs. */
1424         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1425                 return -EINVAL;
1426
1427         ret = f2fs_convert_inline_inode(inode);
1428         if (ret)
1429                 return ret;
1430
1431         /* write out all dirty pages from offset */
1432         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1433         if (ret)
1434                 return ret;
1435
1436         ret = f2fs_do_collapse(inode, offset, len);
1437         if (ret)
1438                 return ret;
1439
1440         /* write out all moved pages, if possible */
1441         filemap_invalidate_lock(inode->i_mapping);
1442         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1443         truncate_pagecache(inode, offset);
1444
1445         new_size = i_size_read(inode) - len;
1446         ret = f2fs_truncate_blocks(inode, new_size, true);
1447         filemap_invalidate_unlock(inode->i_mapping);
1448         if (!ret)
1449                 f2fs_i_size_write(inode, new_size);
1450         return ret;
1451 }
1452
1453 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1454                                                                 pgoff_t end)
1455 {
1456         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1457         pgoff_t index = start;
1458         unsigned int ofs_in_node = dn->ofs_in_node;
1459         blkcnt_t count = 0;
1460         int ret;
1461
1462         for (; index < end; index++, dn->ofs_in_node++) {
1463                 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1464                         count++;
1465         }
1466
1467         dn->ofs_in_node = ofs_in_node;
1468         ret = f2fs_reserve_new_blocks(dn, count);
1469         if (ret)
1470                 return ret;
1471
1472         dn->ofs_in_node = ofs_in_node;
1473         for (index = start; index < end; index++, dn->ofs_in_node++) {
1474                 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1475                 /*
1476                  * f2fs_reserve_new_blocks will not guarantee entire block
1477                  * allocation.
1478                  */
1479                 if (dn->data_blkaddr == NULL_ADDR) {
1480                         ret = -ENOSPC;
1481                         break;
1482                 }
1483
1484                 if (dn->data_blkaddr == NEW_ADDR)
1485                         continue;
1486
1487                 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1488                                         DATA_GENERIC_ENHANCE)) {
1489                         ret = -EFSCORRUPTED;
1490                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1491                         break;
1492                 }
1493
1494                 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1495                 dn->data_blkaddr = NEW_ADDR;
1496                 f2fs_set_data_blkaddr(dn);
1497         }
1498
1499         f2fs_update_extent_cache_range(dn, start, 0, index - start);
1500
1501         return ret;
1502 }
1503
1504 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1505                                                                 int mode)
1506 {
1507         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1508         struct address_space *mapping = inode->i_mapping;
1509         pgoff_t index, pg_start, pg_end;
1510         loff_t new_size = i_size_read(inode);
1511         loff_t off_start, off_end;
1512         int ret = 0;
1513
1514         ret = inode_newsize_ok(inode, (len + offset));
1515         if (ret)
1516                 return ret;
1517
1518         ret = f2fs_convert_inline_inode(inode);
1519         if (ret)
1520                 return ret;
1521
1522         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1523         if (ret)
1524                 return ret;
1525
1526         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1527         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1528
1529         off_start = offset & (PAGE_SIZE - 1);
1530         off_end = (offset + len) & (PAGE_SIZE - 1);
1531
1532         if (pg_start == pg_end) {
1533                 ret = fill_zero(inode, pg_start, off_start,
1534                                                 off_end - off_start);
1535                 if (ret)
1536                         return ret;
1537
1538                 new_size = max_t(loff_t, new_size, offset + len);
1539         } else {
1540                 if (off_start) {
1541                         ret = fill_zero(inode, pg_start++, off_start,
1542                                                 PAGE_SIZE - off_start);
1543                         if (ret)
1544                                 return ret;
1545
1546                         new_size = max_t(loff_t, new_size,
1547                                         (loff_t)pg_start << PAGE_SHIFT);
1548                 }
1549
1550                 for (index = pg_start; index < pg_end;) {
1551                         struct dnode_of_data dn;
1552                         unsigned int end_offset;
1553                         pgoff_t end;
1554
1555                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1556                         filemap_invalidate_lock(mapping);
1557
1558                         truncate_pagecache_range(inode,
1559                                 (loff_t)index << PAGE_SHIFT,
1560                                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1561
1562                         f2fs_lock_op(sbi);
1563
1564                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1565                         ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1566                         if (ret) {
1567                                 f2fs_unlock_op(sbi);
1568                                 filemap_invalidate_unlock(mapping);
1569                                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1570                                 goto out;
1571                         }
1572
1573                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1574                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1575
1576                         ret = f2fs_do_zero_range(&dn, index, end);
1577                         f2fs_put_dnode(&dn);
1578
1579                         f2fs_unlock_op(sbi);
1580                         filemap_invalidate_unlock(mapping);
1581                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1582
1583                         f2fs_balance_fs(sbi, dn.node_changed);
1584
1585                         if (ret)
1586                                 goto out;
1587
1588                         index = end;
1589                         new_size = max_t(loff_t, new_size,
1590                                         (loff_t)index << PAGE_SHIFT);
1591                 }
1592
1593                 if (off_end) {
1594                         ret = fill_zero(inode, pg_end, 0, off_end);
1595                         if (ret)
1596                                 goto out;
1597
1598                         new_size = max_t(loff_t, new_size, offset + len);
1599                 }
1600         }
1601
1602 out:
1603         if (new_size > i_size_read(inode)) {
1604                 if (mode & FALLOC_FL_KEEP_SIZE)
1605                         file_set_keep_isize(inode);
1606                 else
1607                         f2fs_i_size_write(inode, new_size);
1608         }
1609         return ret;
1610 }
1611
1612 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1613 {
1614         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1615         struct address_space *mapping = inode->i_mapping;
1616         pgoff_t nr, pg_start, pg_end, delta, idx;
1617         loff_t new_size;
1618         int ret = 0;
1619
1620         new_size = i_size_read(inode) + len;
1621         ret = inode_newsize_ok(inode, new_size);
1622         if (ret)
1623                 return ret;
1624
1625         if (offset >= i_size_read(inode))
1626                 return -EINVAL;
1627
1628         /* insert range should be aligned to block size of f2fs. */
1629         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1630                 return -EINVAL;
1631
1632         ret = f2fs_convert_inline_inode(inode);
1633         if (ret)
1634                 return ret;
1635
1636         f2fs_balance_fs(sbi, true);
1637
1638         filemap_invalidate_lock(mapping);
1639         ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1640         filemap_invalidate_unlock(mapping);
1641         if (ret)
1642                 return ret;
1643
1644         /* write out all dirty pages from offset */
1645         ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1646         if (ret)
1647                 return ret;
1648
1649         pg_start = offset >> PAGE_SHIFT;
1650         pg_end = (offset + len) >> PAGE_SHIFT;
1651         delta = pg_end - pg_start;
1652         idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1653
1654         /* avoid gc operation during block exchange */
1655         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1656         filemap_invalidate_lock(mapping);
1657         truncate_pagecache(inode, offset);
1658
1659         while (!ret && idx > pg_start) {
1660                 nr = idx - pg_start;
1661                 if (nr > delta)
1662                         nr = delta;
1663                 idx -= nr;
1664
1665                 f2fs_lock_op(sbi);
1666                 f2fs_drop_extent_tree(inode);
1667
1668                 ret = __exchange_data_block(inode, inode, idx,
1669                                         idx + delta, nr, false);
1670                 f2fs_unlock_op(sbi);
1671         }
1672         filemap_invalidate_unlock(mapping);
1673         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1674
1675         /* write out all moved pages, if possible */
1676         filemap_invalidate_lock(mapping);
1677         filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1678         truncate_pagecache(inode, offset);
1679         filemap_invalidate_unlock(mapping);
1680
1681         if (!ret)
1682                 f2fs_i_size_write(inode, new_size);
1683         return ret;
1684 }
1685
1686 static int expand_inode_data(struct inode *inode, loff_t offset,
1687                                         loff_t len, int mode)
1688 {
1689         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1690         struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1691                         .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1692                         .m_may_create = true };
1693         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1694                         .init_gc_type = FG_GC,
1695                         .should_migrate_blocks = false,
1696                         .err_gc_skipped = true,
1697                         .nr_free_secs = 0 };
1698         pgoff_t pg_start, pg_end;
1699         loff_t new_size = i_size_read(inode);
1700         loff_t off_end;
1701         block_t expanded = 0;
1702         int err;
1703
1704         err = inode_newsize_ok(inode, (len + offset));
1705         if (err)
1706                 return err;
1707
1708         err = f2fs_convert_inline_inode(inode);
1709         if (err)
1710                 return err;
1711
1712         f2fs_balance_fs(sbi, true);
1713
1714         pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1715         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1716         off_end = (offset + len) & (PAGE_SIZE - 1);
1717
1718         map.m_lblk = pg_start;
1719         map.m_len = pg_end - pg_start;
1720         if (off_end)
1721                 map.m_len++;
1722
1723         if (!map.m_len)
1724                 return 0;
1725
1726         if (f2fs_is_pinned_file(inode)) {
1727                 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1728                 block_t sec_len = roundup(map.m_len, sec_blks);
1729
1730                 map.m_len = sec_blks;
1731 next_alloc:
1732                 if (has_not_enough_free_secs(sbi, 0,
1733                         GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1734                         f2fs_down_write(&sbi->gc_lock);
1735                         err = f2fs_gc(sbi, &gc_control);
1736                         if (err && err != -ENODATA)
1737                                 goto out_err;
1738                 }
1739
1740                 f2fs_down_write(&sbi->pin_sem);
1741
1742                 f2fs_lock_op(sbi);
1743                 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1744                 f2fs_unlock_op(sbi);
1745
1746                 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1747                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1748                 file_dont_truncate(inode);
1749
1750                 f2fs_up_write(&sbi->pin_sem);
1751
1752                 expanded += map.m_len;
1753                 sec_len -= map.m_len;
1754                 map.m_lblk += map.m_len;
1755                 if (!err && sec_len)
1756                         goto next_alloc;
1757
1758                 map.m_len = expanded;
1759         } else {
1760                 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1761                 expanded = map.m_len;
1762         }
1763 out_err:
1764         if (err) {
1765                 pgoff_t last_off;
1766
1767                 if (!expanded)
1768                         return err;
1769
1770                 last_off = pg_start + expanded - 1;
1771
1772                 /* update new size to the failed position */
1773                 new_size = (last_off == pg_end) ? offset + len :
1774                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1775         } else {
1776                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1777         }
1778
1779         if (new_size > i_size_read(inode)) {
1780                 if (mode & FALLOC_FL_KEEP_SIZE)
1781                         file_set_keep_isize(inode);
1782                 else
1783                         f2fs_i_size_write(inode, new_size);
1784         }
1785
1786         return err;
1787 }
1788
1789 static long f2fs_fallocate(struct file *file, int mode,
1790                                 loff_t offset, loff_t len)
1791 {
1792         struct inode *inode = file_inode(file);
1793         long ret = 0;
1794
1795         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1796                 return -EIO;
1797         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1798                 return -ENOSPC;
1799         if (!f2fs_is_compress_backend_ready(inode))
1800                 return -EOPNOTSUPP;
1801
1802         /* f2fs only support ->fallocate for regular file */
1803         if (!S_ISREG(inode->i_mode))
1804                 return -EINVAL;
1805
1806         if (IS_ENCRYPTED(inode) &&
1807                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1808                 return -EOPNOTSUPP;
1809
1810         /*
1811          * Pinned file should not support partial trucation since the block
1812          * can be used by applications.
1813          */
1814         if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1815                 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1816                         FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1817                 return -EOPNOTSUPP;
1818
1819         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1820                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1821                         FALLOC_FL_INSERT_RANGE))
1822                 return -EOPNOTSUPP;
1823
1824         inode_lock(inode);
1825
1826         ret = file_modified(file);
1827         if (ret)
1828                 goto out;
1829
1830         if (mode & FALLOC_FL_PUNCH_HOLE) {
1831                 if (offset >= inode->i_size)
1832                         goto out;
1833
1834                 ret = punch_hole(inode, offset, len);
1835         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1836                 ret = f2fs_collapse_range(inode, offset, len);
1837         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1838                 ret = f2fs_zero_range(inode, offset, len, mode);
1839         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1840                 ret = f2fs_insert_range(inode, offset, len);
1841         } else {
1842                 ret = expand_inode_data(inode, offset, len, mode);
1843         }
1844
1845         if (!ret) {
1846                 inode->i_mtime = inode->i_ctime = current_time(inode);
1847                 f2fs_mark_inode_dirty_sync(inode, false);
1848                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1849         }
1850
1851 out:
1852         inode_unlock(inode);
1853
1854         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1855         return ret;
1856 }
1857
1858 static int f2fs_release_file(struct inode *inode, struct file *filp)
1859 {
1860         /*
1861          * f2fs_relase_file is called at every close calls. So we should
1862          * not drop any inmemory pages by close called by other process.
1863          */
1864         if (!(filp->f_mode & FMODE_WRITE) ||
1865                         atomic_read(&inode->i_writecount) != 1)
1866                 return 0;
1867
1868         f2fs_abort_atomic_write(inode, true);
1869         return 0;
1870 }
1871
1872 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1873 {
1874         struct inode *inode = file_inode(file);
1875
1876         /*
1877          * If the process doing a transaction is crashed, we should do
1878          * roll-back. Otherwise, other reader/write can see corrupted database
1879          * until all the writers close its file. Since this should be done
1880          * before dropping file lock, it needs to do in ->flush.
1881          */
1882         if (F2FS_I(inode)->atomic_write_task == current)
1883                 f2fs_abort_atomic_write(inode, true);
1884         return 0;
1885 }
1886
1887 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1888 {
1889         struct f2fs_inode_info *fi = F2FS_I(inode);
1890         u32 masked_flags = fi->i_flags & mask;
1891
1892         /* mask can be shrunk by flags_valid selector */
1893         iflags &= mask;
1894
1895         /* Is it quota file? Do not allow user to mess with it */
1896         if (IS_NOQUOTA(inode))
1897                 return -EPERM;
1898
1899         if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1900                 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1901                         return -EOPNOTSUPP;
1902                 if (!f2fs_empty_dir(inode))
1903                         return -ENOTEMPTY;
1904         }
1905
1906         if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1907                 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1908                         return -EOPNOTSUPP;
1909                 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1910                         return -EINVAL;
1911         }
1912
1913         if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1914                 if (masked_flags & F2FS_COMPR_FL) {
1915                         if (!f2fs_disable_compressed_file(inode))
1916                                 return -EINVAL;
1917                 } else {
1918                         if (!f2fs_may_compress(inode))
1919                                 return -EINVAL;
1920                         if (S_ISREG(inode->i_mode) && F2FS_HAS_BLOCKS(inode))
1921                                 return -EINVAL;
1922                         if (set_compress_context(inode))
1923                                 return -EOPNOTSUPP;
1924                 }
1925         }
1926
1927         fi->i_flags = iflags | (fi->i_flags & ~mask);
1928         f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1929                                         (fi->i_flags & F2FS_NOCOMP_FL));
1930
1931         if (fi->i_flags & F2FS_PROJINHERIT_FL)
1932                 set_inode_flag(inode, FI_PROJ_INHERIT);
1933         else
1934                 clear_inode_flag(inode, FI_PROJ_INHERIT);
1935
1936         inode->i_ctime = current_time(inode);
1937         f2fs_set_inode_flags(inode);
1938         f2fs_mark_inode_dirty_sync(inode, true);
1939         return 0;
1940 }
1941
1942 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1943
1944 /*
1945  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1946  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1947  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1948  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1949  *
1950  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1951  * FS_IOC_FSSETXATTR is done by the VFS.
1952  */
1953
1954 static const struct {
1955         u32 iflag;
1956         u32 fsflag;
1957 } f2fs_fsflags_map[] = {
1958         { F2FS_COMPR_FL,        FS_COMPR_FL },
1959         { F2FS_SYNC_FL,         FS_SYNC_FL },
1960         { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1961         { F2FS_APPEND_FL,       FS_APPEND_FL },
1962         { F2FS_NODUMP_FL,       FS_NODUMP_FL },
1963         { F2FS_NOATIME_FL,      FS_NOATIME_FL },
1964         { F2FS_NOCOMP_FL,       FS_NOCOMP_FL },
1965         { F2FS_INDEX_FL,        FS_INDEX_FL },
1966         { F2FS_DIRSYNC_FL,      FS_DIRSYNC_FL },
1967         { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1968         { F2FS_CASEFOLD_FL,     FS_CASEFOLD_FL },
1969 };
1970
1971 #define F2FS_GETTABLE_FS_FL (           \
1972                 FS_COMPR_FL |           \
1973                 FS_SYNC_FL |            \
1974                 FS_IMMUTABLE_FL |       \
1975                 FS_APPEND_FL |          \
1976                 FS_NODUMP_FL |          \
1977                 FS_NOATIME_FL |         \
1978                 FS_NOCOMP_FL |          \
1979                 FS_INDEX_FL |           \
1980                 FS_DIRSYNC_FL |         \
1981                 FS_PROJINHERIT_FL |     \
1982                 FS_ENCRYPT_FL |         \
1983                 FS_INLINE_DATA_FL |     \
1984                 FS_NOCOW_FL |           \
1985                 FS_VERITY_FL |          \
1986                 FS_CASEFOLD_FL)
1987
1988 #define F2FS_SETTABLE_FS_FL (           \
1989                 FS_COMPR_FL |           \
1990                 FS_SYNC_FL |            \
1991                 FS_IMMUTABLE_FL |       \
1992                 FS_APPEND_FL |          \
1993                 FS_NODUMP_FL |          \
1994                 FS_NOATIME_FL |         \
1995                 FS_NOCOMP_FL |          \
1996                 FS_DIRSYNC_FL |         \
1997                 FS_PROJINHERIT_FL |     \
1998                 FS_CASEFOLD_FL)
1999
2000 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2001 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2002 {
2003         u32 fsflags = 0;
2004         int i;
2005
2006         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2007                 if (iflags & f2fs_fsflags_map[i].iflag)
2008                         fsflags |= f2fs_fsflags_map[i].fsflag;
2009
2010         return fsflags;
2011 }
2012
2013 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2014 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2015 {
2016         u32 iflags = 0;
2017         int i;
2018
2019         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2020                 if (fsflags & f2fs_fsflags_map[i].fsflag)
2021                         iflags |= f2fs_fsflags_map[i].iflag;
2022
2023         return iflags;
2024 }
2025
2026 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2027 {
2028         struct inode *inode = file_inode(filp);
2029
2030         return put_user(inode->i_generation, (int __user *)arg);
2031 }
2032
2033 static int f2fs_ioc_start_atomic_write(struct file *filp)
2034 {
2035         struct inode *inode = file_inode(filp);
2036         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2037         struct f2fs_inode_info *fi = F2FS_I(inode);
2038         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2039         struct inode *pinode;
2040         int ret;
2041
2042         if (!inode_owner_or_capable(mnt_userns, inode))
2043                 return -EACCES;
2044
2045         if (!S_ISREG(inode->i_mode))
2046                 return -EINVAL;
2047
2048         if (filp->f_flags & O_DIRECT)
2049                 return -EINVAL;
2050
2051         ret = mnt_want_write_file(filp);
2052         if (ret)
2053                 return ret;
2054
2055         inode_lock(inode);
2056
2057         if (!f2fs_disable_compressed_file(inode)) {
2058                 ret = -EINVAL;
2059                 goto out;
2060         }
2061
2062         if (f2fs_is_atomic_file(inode))
2063                 goto out;
2064
2065         ret = f2fs_convert_inline_inode(inode);
2066         if (ret)
2067                 goto out;
2068
2069         f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2070
2071         /*
2072          * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2073          * f2fs_is_atomic_file.
2074          */
2075         if (get_dirty_pages(inode))
2076                 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2077                           inode->i_ino, get_dirty_pages(inode));
2078         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2079         if (ret) {
2080                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2081                 goto out;
2082         }
2083
2084         /* Create a COW inode for atomic write */
2085         pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2086         if (IS_ERR(pinode)) {
2087                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2088                 ret = PTR_ERR(pinode);
2089                 goto out;
2090         }
2091
2092         ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2093         iput(pinode);
2094         if (ret) {
2095                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2096                 goto out;
2097         }
2098         f2fs_i_size_write(fi->cow_inode, i_size_read(inode));
2099
2100         stat_inc_atomic_inode(inode);
2101
2102         set_inode_flag(inode, FI_ATOMIC_FILE);
2103         set_inode_flag(fi->cow_inode, FI_COW_FILE);
2104         clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2105         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2106
2107         f2fs_update_time(sbi, REQ_TIME);
2108         fi->atomic_write_task = current;
2109         stat_update_max_atomic_write(inode);
2110         fi->atomic_write_cnt = 0;
2111 out:
2112         inode_unlock(inode);
2113         mnt_drop_write_file(filp);
2114         return ret;
2115 }
2116
2117 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2118 {
2119         struct inode *inode = file_inode(filp);
2120         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2121         int ret;
2122
2123         if (!inode_owner_or_capable(mnt_userns, inode))
2124                 return -EACCES;
2125
2126         ret = mnt_want_write_file(filp);
2127         if (ret)
2128                 return ret;
2129
2130         f2fs_balance_fs(F2FS_I_SB(inode), true);
2131
2132         inode_lock(inode);
2133
2134         if (f2fs_is_atomic_file(inode)) {
2135                 ret = f2fs_commit_atomic_write(inode);
2136                 if (ret)
2137                         goto unlock_out;
2138
2139                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2140                 if (!ret)
2141                         f2fs_abort_atomic_write(inode, false);
2142         } else {
2143                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2144         }
2145 unlock_out:
2146         inode_unlock(inode);
2147         mnt_drop_write_file(filp);
2148         return ret;
2149 }
2150
2151 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2152 {
2153         struct inode *inode = file_inode(filp);
2154         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2155         int ret;
2156
2157         if (!inode_owner_or_capable(mnt_userns, inode))
2158                 return -EACCES;
2159
2160         ret = mnt_want_write_file(filp);
2161         if (ret)
2162                 return ret;
2163
2164         inode_lock(inode);
2165
2166         f2fs_abort_atomic_write(inode, true);
2167
2168         inode_unlock(inode);
2169
2170         mnt_drop_write_file(filp);
2171         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2172         return ret;
2173 }
2174
2175 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2176 {
2177         struct inode *inode = file_inode(filp);
2178         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2179         struct super_block *sb = sbi->sb;
2180         __u32 in;
2181         int ret = 0;
2182
2183         if (!capable(CAP_SYS_ADMIN))
2184                 return -EPERM;
2185
2186         if (get_user(in, (__u32 __user *)arg))
2187                 return -EFAULT;
2188
2189         if (in != F2FS_GOING_DOWN_FULLSYNC) {
2190                 ret = mnt_want_write_file(filp);
2191                 if (ret) {
2192                         if (ret == -EROFS) {
2193                                 ret = 0;
2194                                 f2fs_stop_checkpoint(sbi, false,
2195                                                 STOP_CP_REASON_SHUTDOWN);
2196                                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2197                                 trace_f2fs_shutdown(sbi, in, ret);
2198                         }
2199                         return ret;
2200                 }
2201         }
2202
2203         switch (in) {
2204         case F2FS_GOING_DOWN_FULLSYNC:
2205                 ret = freeze_bdev(sb->s_bdev);
2206                 if (ret)
2207                         goto out;
2208                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2209                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2210                 thaw_bdev(sb->s_bdev);
2211                 break;
2212         case F2FS_GOING_DOWN_METASYNC:
2213                 /* do checkpoint only */
2214                 ret = f2fs_sync_fs(sb, 1);
2215                 if (ret)
2216                         goto out;
2217                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2218                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2219                 break;
2220         case F2FS_GOING_DOWN_NOSYNC:
2221                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2222                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2223                 break;
2224         case F2FS_GOING_DOWN_METAFLUSH:
2225                 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2226                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2227                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2228                 break;
2229         case F2FS_GOING_DOWN_NEED_FSCK:
2230                 set_sbi_flag(sbi, SBI_NEED_FSCK);
2231                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2232                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2233                 /* do checkpoint only */
2234                 ret = f2fs_sync_fs(sb, 1);
2235                 goto out;
2236         default:
2237                 ret = -EINVAL;
2238                 goto out;
2239         }
2240
2241         f2fs_stop_gc_thread(sbi);
2242         f2fs_stop_discard_thread(sbi);
2243
2244         f2fs_drop_discard_cmd(sbi);
2245         clear_opt(sbi, DISCARD);
2246
2247         f2fs_update_time(sbi, REQ_TIME);
2248 out:
2249         if (in != F2FS_GOING_DOWN_FULLSYNC)
2250                 mnt_drop_write_file(filp);
2251
2252         trace_f2fs_shutdown(sbi, in, ret);
2253
2254         return ret;
2255 }
2256
2257 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2258 {
2259         struct inode *inode = file_inode(filp);
2260         struct super_block *sb = inode->i_sb;
2261         struct fstrim_range range;
2262         int ret;
2263
2264         if (!capable(CAP_SYS_ADMIN))
2265                 return -EPERM;
2266
2267         if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2268                 return -EOPNOTSUPP;
2269
2270         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2271                                 sizeof(range)))
2272                 return -EFAULT;
2273
2274         ret = mnt_want_write_file(filp);
2275         if (ret)
2276                 return ret;
2277
2278         range.minlen = max((unsigned int)range.minlen,
2279                            bdev_discard_granularity(sb->s_bdev));
2280         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2281         mnt_drop_write_file(filp);
2282         if (ret < 0)
2283                 return ret;
2284
2285         if (copy_to_user((struct fstrim_range __user *)arg, &range,
2286                                 sizeof(range)))
2287                 return -EFAULT;
2288         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2289         return 0;
2290 }
2291
2292 static bool uuid_is_nonzero(__u8 u[16])
2293 {
2294         int i;
2295
2296         for (i = 0; i < 16; i++)
2297                 if (u[i])
2298                         return true;
2299         return false;
2300 }
2301
2302 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2303 {
2304         struct inode *inode = file_inode(filp);
2305
2306         if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2307                 return -EOPNOTSUPP;
2308
2309         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2310
2311         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2312 }
2313
2314 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2315 {
2316         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2317                 return -EOPNOTSUPP;
2318         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2319 }
2320
2321 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2322 {
2323         struct inode *inode = file_inode(filp);
2324         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2325         int err;
2326
2327         if (!f2fs_sb_has_encrypt(sbi))
2328                 return -EOPNOTSUPP;
2329
2330         err = mnt_want_write_file(filp);
2331         if (err)
2332                 return err;
2333
2334         f2fs_down_write(&sbi->sb_lock);
2335
2336         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2337                 goto got_it;
2338
2339         /* update superblock with uuid */
2340         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2341
2342         err = f2fs_commit_super(sbi, false);
2343         if (err) {
2344                 /* undo new data */
2345                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2346                 goto out_err;
2347         }
2348 got_it:
2349         if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2350                                                                         16))
2351                 err = -EFAULT;
2352 out_err:
2353         f2fs_up_write(&sbi->sb_lock);
2354         mnt_drop_write_file(filp);
2355         return err;
2356 }
2357
2358 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2359                                              unsigned long arg)
2360 {
2361         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2362                 return -EOPNOTSUPP;
2363
2364         return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2365 }
2366
2367 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2368 {
2369         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2370                 return -EOPNOTSUPP;
2371
2372         return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2373 }
2374
2375 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2376 {
2377         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2378                 return -EOPNOTSUPP;
2379
2380         return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2381 }
2382
2383 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2384                                                     unsigned long arg)
2385 {
2386         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2387                 return -EOPNOTSUPP;
2388
2389         return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2390 }
2391
2392 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2393                                               unsigned long arg)
2394 {
2395         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2396                 return -EOPNOTSUPP;
2397
2398         return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2399 }
2400
2401 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2402 {
2403         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2404                 return -EOPNOTSUPP;
2405
2406         return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2407 }
2408
2409 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2410 {
2411         struct inode *inode = file_inode(filp);
2412         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2413         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2414                         .no_bg_gc = false,
2415                         .should_migrate_blocks = false,
2416                         .nr_free_secs = 0 };
2417         __u32 sync;
2418         int ret;
2419
2420         if (!capable(CAP_SYS_ADMIN))
2421                 return -EPERM;
2422
2423         if (get_user(sync, (__u32 __user *)arg))
2424                 return -EFAULT;
2425
2426         if (f2fs_readonly(sbi->sb))
2427                 return -EROFS;
2428
2429         ret = mnt_want_write_file(filp);
2430         if (ret)
2431                 return ret;
2432
2433         if (!sync) {
2434                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2435                         ret = -EBUSY;
2436                         goto out;
2437                 }
2438         } else {
2439                 f2fs_down_write(&sbi->gc_lock);
2440         }
2441
2442         gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2443         gc_control.err_gc_skipped = sync;
2444         ret = f2fs_gc(sbi, &gc_control);
2445 out:
2446         mnt_drop_write_file(filp);
2447         return ret;
2448 }
2449
2450 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2451 {
2452         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2453         struct f2fs_gc_control gc_control = {
2454                         .init_gc_type = range->sync ? FG_GC : BG_GC,
2455                         .no_bg_gc = false,
2456                         .should_migrate_blocks = false,
2457                         .err_gc_skipped = range->sync,
2458                         .nr_free_secs = 0 };
2459         u64 end;
2460         int ret;
2461
2462         if (!capable(CAP_SYS_ADMIN))
2463                 return -EPERM;
2464         if (f2fs_readonly(sbi->sb))
2465                 return -EROFS;
2466
2467         end = range->start + range->len;
2468         if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2469                                         end >= MAX_BLKADDR(sbi))
2470                 return -EINVAL;
2471
2472         ret = mnt_want_write_file(filp);
2473         if (ret)
2474                 return ret;
2475
2476 do_more:
2477         if (!range->sync) {
2478                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2479                         ret = -EBUSY;
2480                         goto out;
2481                 }
2482         } else {
2483                 f2fs_down_write(&sbi->gc_lock);
2484         }
2485
2486         gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2487         ret = f2fs_gc(sbi, &gc_control);
2488         if (ret) {
2489                 if (ret == -EBUSY)
2490                         ret = -EAGAIN;
2491                 goto out;
2492         }
2493         range->start += CAP_BLKS_PER_SEC(sbi);
2494         if (range->start <= end)
2495                 goto do_more;
2496 out:
2497         mnt_drop_write_file(filp);
2498         return ret;
2499 }
2500
2501 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2502 {
2503         struct f2fs_gc_range range;
2504
2505         if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2506                                                         sizeof(range)))
2507                 return -EFAULT;
2508         return __f2fs_ioc_gc_range(filp, &range);
2509 }
2510
2511 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2512 {
2513         struct inode *inode = file_inode(filp);
2514         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2515         int ret;
2516
2517         if (!capable(CAP_SYS_ADMIN))
2518                 return -EPERM;
2519
2520         if (f2fs_readonly(sbi->sb))
2521                 return -EROFS;
2522
2523         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2524                 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2525                 return -EINVAL;
2526         }
2527
2528         ret = mnt_want_write_file(filp);
2529         if (ret)
2530                 return ret;
2531
2532         ret = f2fs_sync_fs(sbi->sb, 1);
2533
2534         mnt_drop_write_file(filp);
2535         return ret;
2536 }
2537
2538 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2539                                         struct file *filp,
2540                                         struct f2fs_defragment *range)
2541 {
2542         struct inode *inode = file_inode(filp);
2543         struct f2fs_map_blocks map = { .m_next_extent = NULL,
2544                                         .m_seg_type = NO_CHECK_TYPE,
2545                                         .m_may_create = false };
2546         struct extent_info ei = {0, 0, 0};
2547         pgoff_t pg_start, pg_end, next_pgofs;
2548         unsigned int blk_per_seg = sbi->blocks_per_seg;
2549         unsigned int total = 0, sec_num;
2550         block_t blk_end = 0;
2551         bool fragmented = false;
2552         int err;
2553
2554         pg_start = range->start >> PAGE_SHIFT;
2555         pg_end = (range->start + range->len) >> PAGE_SHIFT;
2556
2557         f2fs_balance_fs(sbi, true);
2558
2559         inode_lock(inode);
2560
2561         /* if in-place-update policy is enabled, don't waste time here */
2562         set_inode_flag(inode, FI_OPU_WRITE);
2563         if (f2fs_should_update_inplace(inode, NULL)) {
2564                 err = -EINVAL;
2565                 goto out;
2566         }
2567
2568         /* writeback all dirty pages in the range */
2569         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2570                                                 range->start + range->len - 1);
2571         if (err)
2572                 goto out;
2573
2574         /*
2575          * lookup mapping info in extent cache, skip defragmenting if physical
2576          * block addresses are continuous.
2577          */
2578         if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2579                 if (ei.fofs + ei.len >= pg_end)
2580                         goto out;
2581         }
2582
2583         map.m_lblk = pg_start;
2584         map.m_next_pgofs = &next_pgofs;
2585
2586         /*
2587          * lookup mapping info in dnode page cache, skip defragmenting if all
2588          * physical block addresses are continuous even if there are hole(s)
2589          * in logical blocks.
2590          */
2591         while (map.m_lblk < pg_end) {
2592                 map.m_len = pg_end - map.m_lblk;
2593                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2594                 if (err)
2595                         goto out;
2596
2597                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2598                         map.m_lblk = next_pgofs;
2599                         continue;
2600                 }
2601
2602                 if (blk_end && blk_end != map.m_pblk)
2603                         fragmented = true;
2604
2605                 /* record total count of block that we're going to move */
2606                 total += map.m_len;
2607
2608                 blk_end = map.m_pblk + map.m_len;
2609
2610                 map.m_lblk += map.m_len;
2611         }
2612
2613         if (!fragmented) {
2614                 total = 0;
2615                 goto out;
2616         }
2617
2618         sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2619
2620         /*
2621          * make sure there are enough free section for LFS allocation, this can
2622          * avoid defragment running in SSR mode when free section are allocated
2623          * intensively
2624          */
2625         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2626                 err = -EAGAIN;
2627                 goto out;
2628         }
2629
2630         map.m_lblk = pg_start;
2631         map.m_len = pg_end - pg_start;
2632         total = 0;
2633
2634         while (map.m_lblk < pg_end) {
2635                 pgoff_t idx;
2636                 int cnt = 0;
2637
2638 do_map:
2639                 map.m_len = pg_end - map.m_lblk;
2640                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2641                 if (err)
2642                         goto clear_out;
2643
2644                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2645                         map.m_lblk = next_pgofs;
2646                         goto check;
2647                 }
2648
2649                 set_inode_flag(inode, FI_SKIP_WRITES);
2650
2651                 idx = map.m_lblk;
2652                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2653                         struct page *page;
2654
2655                         page = f2fs_get_lock_data_page(inode, idx, true);
2656                         if (IS_ERR(page)) {
2657                                 err = PTR_ERR(page);
2658                                 goto clear_out;
2659                         }
2660
2661                         set_page_dirty(page);
2662                         set_page_private_gcing(page);
2663                         f2fs_put_page(page, 1);
2664
2665                         idx++;
2666                         cnt++;
2667                         total++;
2668                 }
2669
2670                 map.m_lblk = idx;
2671 check:
2672                 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2673                         goto do_map;
2674
2675                 clear_inode_flag(inode, FI_SKIP_WRITES);
2676
2677                 err = filemap_fdatawrite(inode->i_mapping);
2678                 if (err)
2679                         goto out;
2680         }
2681 clear_out:
2682         clear_inode_flag(inode, FI_SKIP_WRITES);
2683 out:
2684         clear_inode_flag(inode, FI_OPU_WRITE);
2685         inode_unlock(inode);
2686         if (!err)
2687                 range->len = (u64)total << PAGE_SHIFT;
2688         return err;
2689 }
2690
2691 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2692 {
2693         struct inode *inode = file_inode(filp);
2694         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2695         struct f2fs_defragment range;
2696         int err;
2697
2698         if (!capable(CAP_SYS_ADMIN))
2699                 return -EPERM;
2700
2701         if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2702                 return -EINVAL;
2703
2704         if (f2fs_readonly(sbi->sb))
2705                 return -EROFS;
2706
2707         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2708                                                         sizeof(range)))
2709                 return -EFAULT;
2710
2711         /* verify alignment of offset & size */
2712         if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2713                 return -EINVAL;
2714
2715         if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2716                                         max_file_blocks(inode)))
2717                 return -EINVAL;
2718
2719         err = mnt_want_write_file(filp);
2720         if (err)
2721                 return err;
2722
2723         err = f2fs_defragment_range(sbi, filp, &range);
2724         mnt_drop_write_file(filp);
2725
2726         f2fs_update_time(sbi, REQ_TIME);
2727         if (err < 0)
2728                 return err;
2729
2730         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2731                                                         sizeof(range)))
2732                 return -EFAULT;
2733
2734         return 0;
2735 }
2736
2737 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2738                         struct file *file_out, loff_t pos_out, size_t len)
2739 {
2740         struct inode *src = file_inode(file_in);
2741         struct inode *dst = file_inode(file_out);
2742         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2743         size_t olen = len, dst_max_i_size = 0;
2744         size_t dst_osize;
2745         int ret;
2746
2747         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2748                                 src->i_sb != dst->i_sb)
2749                 return -EXDEV;
2750
2751         if (unlikely(f2fs_readonly(src->i_sb)))
2752                 return -EROFS;
2753
2754         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2755                 return -EINVAL;
2756
2757         if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2758                 return -EOPNOTSUPP;
2759
2760         if (pos_out < 0 || pos_in < 0)
2761                 return -EINVAL;
2762
2763         if (src == dst) {
2764                 if (pos_in == pos_out)
2765                         return 0;
2766                 if (pos_out > pos_in && pos_out < pos_in + len)
2767                         return -EINVAL;
2768         }
2769
2770         inode_lock(src);
2771         if (src != dst) {
2772                 ret = -EBUSY;
2773                 if (!inode_trylock(dst))
2774                         goto out;
2775         }
2776
2777         ret = -EINVAL;
2778         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2779                 goto out_unlock;
2780         if (len == 0)
2781                 olen = len = src->i_size - pos_in;
2782         if (pos_in + len == src->i_size)
2783                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2784         if (len == 0) {
2785                 ret = 0;
2786                 goto out_unlock;
2787         }
2788
2789         dst_osize = dst->i_size;
2790         if (pos_out + olen > dst->i_size)
2791                 dst_max_i_size = pos_out + olen;
2792
2793         /* verify the end result is block aligned */
2794         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2795                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2796                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2797                 goto out_unlock;
2798
2799         ret = f2fs_convert_inline_inode(src);
2800         if (ret)
2801                 goto out_unlock;
2802
2803         ret = f2fs_convert_inline_inode(dst);
2804         if (ret)
2805                 goto out_unlock;
2806
2807         /* write out all dirty pages from offset */
2808         ret = filemap_write_and_wait_range(src->i_mapping,
2809                                         pos_in, pos_in + len);
2810         if (ret)
2811                 goto out_unlock;
2812
2813         ret = filemap_write_and_wait_range(dst->i_mapping,
2814                                         pos_out, pos_out + len);
2815         if (ret)
2816                 goto out_unlock;
2817
2818         f2fs_balance_fs(sbi, true);
2819
2820         f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2821         if (src != dst) {
2822                 ret = -EBUSY;
2823                 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2824                         goto out_src;
2825         }
2826
2827         f2fs_lock_op(sbi);
2828         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2829                                 pos_out >> F2FS_BLKSIZE_BITS,
2830                                 len >> F2FS_BLKSIZE_BITS, false);
2831
2832         if (!ret) {
2833                 if (dst_max_i_size)
2834                         f2fs_i_size_write(dst, dst_max_i_size);
2835                 else if (dst_osize != dst->i_size)
2836                         f2fs_i_size_write(dst, dst_osize);
2837         }
2838         f2fs_unlock_op(sbi);
2839
2840         if (src != dst)
2841                 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2842 out_src:
2843         f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2844 out_unlock:
2845         if (src != dst)
2846                 inode_unlock(dst);
2847 out:
2848         inode_unlock(src);
2849         return ret;
2850 }
2851
2852 static int __f2fs_ioc_move_range(struct file *filp,
2853                                 struct f2fs_move_range *range)
2854 {
2855         struct fd dst;
2856         int err;
2857
2858         if (!(filp->f_mode & FMODE_READ) ||
2859                         !(filp->f_mode & FMODE_WRITE))
2860                 return -EBADF;
2861
2862         dst = fdget(range->dst_fd);
2863         if (!dst.file)
2864                 return -EBADF;
2865
2866         if (!(dst.file->f_mode & FMODE_WRITE)) {
2867                 err = -EBADF;
2868                 goto err_out;
2869         }
2870
2871         err = mnt_want_write_file(filp);
2872         if (err)
2873                 goto err_out;
2874
2875         err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2876                                         range->pos_out, range->len);
2877
2878         mnt_drop_write_file(filp);
2879 err_out:
2880         fdput(dst);
2881         return err;
2882 }
2883
2884 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2885 {
2886         struct f2fs_move_range range;
2887
2888         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2889                                                         sizeof(range)))
2890                 return -EFAULT;
2891         return __f2fs_ioc_move_range(filp, &range);
2892 }
2893
2894 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2895 {
2896         struct inode *inode = file_inode(filp);
2897         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2898         struct sit_info *sm = SIT_I(sbi);
2899         unsigned int start_segno = 0, end_segno = 0;
2900         unsigned int dev_start_segno = 0, dev_end_segno = 0;
2901         struct f2fs_flush_device range;
2902         struct f2fs_gc_control gc_control = {
2903                         .init_gc_type = FG_GC,
2904                         .should_migrate_blocks = true,
2905                         .err_gc_skipped = true,
2906                         .nr_free_secs = 0 };
2907         int ret;
2908
2909         if (!capable(CAP_SYS_ADMIN))
2910                 return -EPERM;
2911
2912         if (f2fs_readonly(sbi->sb))
2913                 return -EROFS;
2914
2915         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2916                 return -EINVAL;
2917
2918         if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2919                                                         sizeof(range)))
2920                 return -EFAULT;
2921
2922         if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2923                         __is_large_section(sbi)) {
2924                 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2925                           range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2926                 return -EINVAL;
2927         }
2928
2929         ret = mnt_want_write_file(filp);
2930         if (ret)
2931                 return ret;
2932
2933         if (range.dev_num != 0)
2934                 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2935         dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2936
2937         start_segno = sm->last_victim[FLUSH_DEVICE];
2938         if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2939                 start_segno = dev_start_segno;
2940         end_segno = min(start_segno + range.segments, dev_end_segno);
2941
2942         while (start_segno < end_segno) {
2943                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2944                         ret = -EBUSY;
2945                         goto out;
2946                 }
2947                 sm->last_victim[GC_CB] = end_segno + 1;
2948                 sm->last_victim[GC_GREEDY] = end_segno + 1;
2949                 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2950
2951                 gc_control.victim_segno = start_segno;
2952                 ret = f2fs_gc(sbi, &gc_control);
2953                 if (ret == -EAGAIN)
2954                         ret = 0;
2955                 else if (ret < 0)
2956                         break;
2957                 start_segno++;
2958         }
2959 out:
2960         mnt_drop_write_file(filp);
2961         return ret;
2962 }
2963
2964 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2965 {
2966         struct inode *inode = file_inode(filp);
2967         u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2968
2969         /* Must validate to set it with SQLite behavior in Android. */
2970         sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2971
2972         return put_user(sb_feature, (u32 __user *)arg);
2973 }
2974
2975 #ifdef CONFIG_QUOTA
2976 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2977 {
2978         struct dquot *transfer_to[MAXQUOTAS] = {};
2979         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2980         struct super_block *sb = sbi->sb;
2981         int err = 0;
2982
2983         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2984         if (!IS_ERR(transfer_to[PRJQUOTA])) {
2985                 err = __dquot_transfer(inode, transfer_to);
2986                 if (err)
2987                         set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2988                 dqput(transfer_to[PRJQUOTA]);
2989         }
2990         return err;
2991 }
2992
2993 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2994 {
2995         struct f2fs_inode_info *fi = F2FS_I(inode);
2996         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2997         struct f2fs_inode *ri = NULL;
2998         kprojid_t kprojid;
2999         int err;
3000
3001         if (!f2fs_sb_has_project_quota(sbi)) {
3002                 if (projid != F2FS_DEF_PROJID)
3003                         return -EOPNOTSUPP;
3004                 else
3005                         return 0;
3006         }
3007
3008         if (!f2fs_has_extra_attr(inode))
3009                 return -EOPNOTSUPP;
3010
3011         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3012
3013         if (projid_eq(kprojid, fi->i_projid))
3014                 return 0;
3015
3016         err = -EPERM;
3017         /* Is it quota file? Do not allow user to mess with it */
3018         if (IS_NOQUOTA(inode))
3019                 return err;
3020
3021         if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3022                 return -EOVERFLOW;
3023
3024         err = f2fs_dquot_initialize(inode);
3025         if (err)
3026                 return err;
3027
3028         f2fs_lock_op(sbi);
3029         err = f2fs_transfer_project_quota(inode, kprojid);
3030         if (err)
3031                 goto out_unlock;
3032
3033         fi->i_projid = kprojid;
3034         inode->i_ctime = current_time(inode);
3035         f2fs_mark_inode_dirty_sync(inode, true);
3036 out_unlock:
3037         f2fs_unlock_op(sbi);
3038         return err;
3039 }
3040 #else
3041 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3042 {
3043         return 0;
3044 }
3045
3046 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3047 {
3048         if (projid != F2FS_DEF_PROJID)
3049                 return -EOPNOTSUPP;
3050         return 0;
3051 }
3052 #endif
3053
3054 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3055 {
3056         struct inode *inode = d_inode(dentry);
3057         struct f2fs_inode_info *fi = F2FS_I(inode);
3058         u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3059
3060         if (IS_ENCRYPTED(inode))
3061                 fsflags |= FS_ENCRYPT_FL;
3062         if (IS_VERITY(inode))
3063                 fsflags |= FS_VERITY_FL;
3064         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3065                 fsflags |= FS_INLINE_DATA_FL;
3066         if (is_inode_flag_set(inode, FI_PIN_FILE))
3067                 fsflags |= FS_NOCOW_FL;
3068
3069         fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3070
3071         if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3072                 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3073
3074         return 0;
3075 }
3076
3077 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3078                       struct dentry *dentry, struct fileattr *fa)
3079 {
3080         struct inode *inode = d_inode(dentry);
3081         u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3082         u32 iflags;
3083         int err;
3084
3085         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3086                 return -EIO;
3087         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3088                 return -ENOSPC;
3089         if (fsflags & ~F2FS_GETTABLE_FS_FL)
3090                 return -EOPNOTSUPP;
3091         fsflags &= F2FS_SETTABLE_FS_FL;
3092         if (!fa->flags_valid)
3093                 mask &= FS_COMMON_FL;
3094
3095         iflags = f2fs_fsflags_to_iflags(fsflags);
3096         if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3097                 return -EOPNOTSUPP;
3098
3099         err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3100         if (!err)
3101                 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3102
3103         return err;
3104 }
3105
3106 int f2fs_pin_file_control(struct inode *inode, bool inc)
3107 {
3108         struct f2fs_inode_info *fi = F2FS_I(inode);
3109         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3110
3111         /* Use i_gc_failures for normal file as a risk signal. */
3112         if (inc)
3113                 f2fs_i_gc_failures_write(inode,
3114                                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3115
3116         if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3117                 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3118                           __func__, inode->i_ino,
3119                           fi->i_gc_failures[GC_FAILURE_PIN]);
3120                 clear_inode_flag(inode, FI_PIN_FILE);
3121                 return -EAGAIN;
3122         }
3123         return 0;
3124 }
3125
3126 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3127 {
3128         struct inode *inode = file_inode(filp);
3129         __u32 pin;
3130         int ret = 0;
3131
3132         if (get_user(pin, (__u32 __user *)arg))
3133                 return -EFAULT;
3134
3135         if (!S_ISREG(inode->i_mode))
3136                 return -EINVAL;
3137
3138         if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3139                 return -EROFS;
3140
3141         ret = mnt_want_write_file(filp);
3142         if (ret)
3143                 return ret;
3144
3145         inode_lock(inode);
3146
3147         if (!pin) {
3148                 clear_inode_flag(inode, FI_PIN_FILE);
3149                 f2fs_i_gc_failures_write(inode, 0);
3150                 goto done;
3151         }
3152
3153         if (f2fs_should_update_outplace(inode, NULL)) {
3154                 ret = -EINVAL;
3155                 goto out;
3156         }
3157
3158         if (f2fs_pin_file_control(inode, false)) {
3159                 ret = -EAGAIN;
3160                 goto out;
3161         }
3162
3163         ret = f2fs_convert_inline_inode(inode);
3164         if (ret)
3165                 goto out;
3166
3167         if (!f2fs_disable_compressed_file(inode)) {
3168                 ret = -EOPNOTSUPP;
3169                 goto out;
3170         }
3171
3172         set_inode_flag(inode, FI_PIN_FILE);
3173         ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3174 done:
3175         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3176 out:
3177         inode_unlock(inode);
3178         mnt_drop_write_file(filp);
3179         return ret;
3180 }
3181
3182 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3183 {
3184         struct inode *inode = file_inode(filp);
3185         __u32 pin = 0;
3186
3187         if (is_inode_flag_set(inode, FI_PIN_FILE))
3188                 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3189         return put_user(pin, (u32 __user *)arg);
3190 }
3191
3192 int f2fs_precache_extents(struct inode *inode)
3193 {
3194         struct f2fs_inode_info *fi = F2FS_I(inode);
3195         struct f2fs_map_blocks map;
3196         pgoff_t m_next_extent;
3197         loff_t end;
3198         int err;
3199
3200         if (is_inode_flag_set(inode, FI_NO_EXTENT))
3201                 return -EOPNOTSUPP;
3202
3203         map.m_lblk = 0;
3204         map.m_next_pgofs = NULL;
3205         map.m_next_extent = &m_next_extent;
3206         map.m_seg_type = NO_CHECK_TYPE;
3207         map.m_may_create = false;
3208         end = max_file_blocks(inode);
3209
3210         while (map.m_lblk < end) {
3211                 map.m_len = end - map.m_lblk;
3212
3213                 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3214                 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3215                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3216                 if (err)
3217                         return err;
3218
3219                 map.m_lblk = m_next_extent;
3220         }
3221
3222         return 0;
3223 }
3224
3225 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3226 {
3227         return f2fs_precache_extents(file_inode(filp));
3228 }
3229
3230 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3231 {
3232         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3233         __u64 block_count;
3234
3235         if (!capable(CAP_SYS_ADMIN))
3236                 return -EPERM;
3237
3238         if (f2fs_readonly(sbi->sb))
3239                 return -EROFS;
3240
3241         if (copy_from_user(&block_count, (void __user *)arg,
3242                            sizeof(block_count)))
3243                 return -EFAULT;
3244
3245         return f2fs_resize_fs(sbi, block_count);
3246 }
3247
3248 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3249 {
3250         struct inode *inode = file_inode(filp);
3251
3252         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3253
3254         if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3255                 f2fs_warn(F2FS_I_SB(inode),
3256                           "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3257                           inode->i_ino);
3258                 return -EOPNOTSUPP;
3259         }
3260
3261         return fsverity_ioctl_enable(filp, (const void __user *)arg);
3262 }
3263
3264 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3265 {
3266         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3267                 return -EOPNOTSUPP;
3268
3269         return fsverity_ioctl_measure(filp, (void __user *)arg);
3270 }
3271
3272 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3273 {
3274         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3275                 return -EOPNOTSUPP;
3276
3277         return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3278 }
3279
3280 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3281 {
3282         struct inode *inode = file_inode(filp);
3283         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3284         char *vbuf;
3285         int count;
3286         int err = 0;
3287
3288         vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3289         if (!vbuf)
3290                 return -ENOMEM;
3291
3292         f2fs_down_read(&sbi->sb_lock);
3293         count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3294                         ARRAY_SIZE(sbi->raw_super->volume_name),
3295                         UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3296         f2fs_up_read(&sbi->sb_lock);
3297
3298         if (copy_to_user((char __user *)arg, vbuf,
3299                                 min(FSLABEL_MAX, count)))
3300                 err = -EFAULT;
3301
3302         kfree(vbuf);
3303         return err;
3304 }
3305
3306 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3307 {
3308         struct inode *inode = file_inode(filp);
3309         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3310         char *vbuf;
3311         int err = 0;
3312
3313         if (!capable(CAP_SYS_ADMIN))
3314                 return -EPERM;
3315
3316         vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3317         if (IS_ERR(vbuf))
3318                 return PTR_ERR(vbuf);
3319
3320         err = mnt_want_write_file(filp);
3321         if (err)
3322                 goto out;
3323
3324         f2fs_down_write(&sbi->sb_lock);
3325
3326         memset(sbi->raw_super->volume_name, 0,
3327                         sizeof(sbi->raw_super->volume_name));
3328         utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3329                         sbi->raw_super->volume_name,
3330                         ARRAY_SIZE(sbi->raw_super->volume_name));
3331
3332         err = f2fs_commit_super(sbi, false);
3333
3334         f2fs_up_write(&sbi->sb_lock);
3335
3336         mnt_drop_write_file(filp);
3337 out:
3338         kfree(vbuf);
3339         return err;
3340 }
3341
3342 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3343 {
3344         struct inode *inode = file_inode(filp);
3345         __u64 blocks;
3346
3347         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3348                 return -EOPNOTSUPP;
3349
3350         if (!f2fs_compressed_file(inode))
3351                 return -EINVAL;
3352
3353         blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3354         return put_user(blocks, (u64 __user *)arg);
3355 }
3356
3357 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3358 {
3359         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3360         unsigned int released_blocks = 0;
3361         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3362         block_t blkaddr;
3363         int i;
3364
3365         for (i = 0; i < count; i++) {
3366                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3367                                                 dn->ofs_in_node + i);
3368
3369                 if (!__is_valid_data_blkaddr(blkaddr))
3370                         continue;
3371                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3372                                         DATA_GENERIC_ENHANCE))) {
3373                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3374                         return -EFSCORRUPTED;
3375                 }
3376         }
3377
3378         while (count) {
3379                 int compr_blocks = 0;
3380
3381                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3382                         blkaddr = f2fs_data_blkaddr(dn);
3383
3384                         if (i == 0) {
3385                                 if (blkaddr == COMPRESS_ADDR)
3386                                         continue;
3387                                 dn->ofs_in_node += cluster_size;
3388                                 goto next;
3389                         }
3390
3391                         if (__is_valid_data_blkaddr(blkaddr))
3392                                 compr_blocks++;
3393
3394                         if (blkaddr != NEW_ADDR)
3395                                 continue;
3396
3397                         dn->data_blkaddr = NULL_ADDR;
3398                         f2fs_set_data_blkaddr(dn);
3399                 }
3400
3401                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3402                 dec_valid_block_count(sbi, dn->inode,
3403                                         cluster_size - compr_blocks);
3404
3405                 released_blocks += cluster_size - compr_blocks;
3406 next:
3407                 count -= cluster_size;
3408         }
3409
3410         return released_blocks;
3411 }
3412
3413 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3414 {
3415         struct inode *inode = file_inode(filp);
3416         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3417         pgoff_t page_idx = 0, last_idx;
3418         unsigned int released_blocks = 0;
3419         int ret;
3420         int writecount;
3421
3422         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3423                 return -EOPNOTSUPP;
3424
3425         if (!f2fs_compressed_file(inode))
3426                 return -EINVAL;
3427
3428         if (f2fs_readonly(sbi->sb))
3429                 return -EROFS;
3430
3431         ret = mnt_want_write_file(filp);
3432         if (ret)
3433                 return ret;
3434
3435         f2fs_balance_fs(F2FS_I_SB(inode), true);
3436
3437         inode_lock(inode);
3438
3439         writecount = atomic_read(&inode->i_writecount);
3440         if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3441                         (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3442                 ret = -EBUSY;
3443                 goto out;
3444         }
3445
3446         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3447                 ret = -EINVAL;
3448                 goto out;
3449         }
3450
3451         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3452         if (ret)
3453                 goto out;
3454
3455         set_inode_flag(inode, FI_COMPRESS_RELEASED);
3456         inode->i_ctime = current_time(inode);
3457         f2fs_mark_inode_dirty_sync(inode, true);
3458
3459         if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3460                 goto out;
3461
3462         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3463         filemap_invalidate_lock(inode->i_mapping);
3464
3465         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3466
3467         while (page_idx < last_idx) {
3468                 struct dnode_of_data dn;
3469                 pgoff_t end_offset, count;
3470
3471                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3472                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3473                 if (ret) {
3474                         if (ret == -ENOENT) {
3475                                 page_idx = f2fs_get_next_page_offset(&dn,
3476                                                                 page_idx);
3477                                 ret = 0;
3478                                 continue;
3479                         }
3480                         break;
3481                 }
3482
3483                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3484                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3485                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3486
3487                 ret = release_compress_blocks(&dn, count);
3488
3489                 f2fs_put_dnode(&dn);
3490
3491                 if (ret < 0)
3492                         break;
3493
3494                 page_idx += count;
3495                 released_blocks += ret;
3496         }
3497
3498         filemap_invalidate_unlock(inode->i_mapping);
3499         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3500 out:
3501         inode_unlock(inode);
3502
3503         mnt_drop_write_file(filp);
3504
3505         if (ret >= 0) {
3506                 ret = put_user(released_blocks, (u64 __user *)arg);
3507         } else if (released_blocks &&
3508                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3509                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3510                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3511                         "iblocks=%llu, released=%u, compr_blocks=%u, "
3512                         "run fsck to fix.",
3513                         __func__, inode->i_ino, inode->i_blocks,
3514                         released_blocks,
3515                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3516         }
3517
3518         return ret;
3519 }
3520
3521 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3522 {
3523         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3524         unsigned int reserved_blocks = 0;
3525         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3526         block_t blkaddr;
3527         int i;
3528
3529         for (i = 0; i < count; i++) {
3530                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3531                                                 dn->ofs_in_node + i);
3532
3533                 if (!__is_valid_data_blkaddr(blkaddr))
3534                         continue;
3535                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3536                                         DATA_GENERIC_ENHANCE))) {
3537                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3538                         return -EFSCORRUPTED;
3539                 }
3540         }
3541
3542         while (count) {
3543                 int compr_blocks = 0;
3544                 blkcnt_t reserved;
3545                 int ret;
3546
3547                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3548                         blkaddr = f2fs_data_blkaddr(dn);
3549
3550                         if (i == 0) {
3551                                 if (blkaddr == COMPRESS_ADDR)
3552                                         continue;
3553                                 dn->ofs_in_node += cluster_size;
3554                                 goto next;
3555                         }
3556
3557                         if (__is_valid_data_blkaddr(blkaddr)) {
3558                                 compr_blocks++;
3559                                 continue;
3560                         }
3561
3562                         dn->data_blkaddr = NEW_ADDR;
3563                         f2fs_set_data_blkaddr(dn);
3564                 }
3565
3566                 reserved = cluster_size - compr_blocks;
3567                 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3568                 if (ret)
3569                         return ret;
3570
3571                 if (reserved != cluster_size - compr_blocks)
3572                         return -ENOSPC;
3573
3574                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3575
3576                 reserved_blocks += reserved;
3577 next:
3578                 count -= cluster_size;
3579         }
3580
3581         return reserved_blocks;
3582 }
3583
3584 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3585 {
3586         struct inode *inode = file_inode(filp);
3587         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3588         pgoff_t page_idx = 0, last_idx;
3589         unsigned int reserved_blocks = 0;
3590         int ret;
3591
3592         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3593                 return -EOPNOTSUPP;
3594
3595         if (!f2fs_compressed_file(inode))
3596                 return -EINVAL;
3597
3598         if (f2fs_readonly(sbi->sb))
3599                 return -EROFS;
3600
3601         ret = mnt_want_write_file(filp);
3602         if (ret)
3603                 return ret;
3604
3605         if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3606                 goto out;
3607
3608         f2fs_balance_fs(F2FS_I_SB(inode), true);
3609
3610         inode_lock(inode);
3611
3612         if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3613                 ret = -EINVAL;
3614                 goto unlock_inode;
3615         }
3616
3617         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3618         filemap_invalidate_lock(inode->i_mapping);
3619
3620         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3621
3622         while (page_idx < last_idx) {
3623                 struct dnode_of_data dn;
3624                 pgoff_t end_offset, count;
3625
3626                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3627                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3628                 if (ret) {
3629                         if (ret == -ENOENT) {
3630                                 page_idx = f2fs_get_next_page_offset(&dn,
3631                                                                 page_idx);
3632                                 ret = 0;
3633                                 continue;
3634                         }
3635                         break;
3636                 }
3637
3638                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3639                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3640                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3641
3642                 ret = reserve_compress_blocks(&dn, count);
3643
3644                 f2fs_put_dnode(&dn);
3645
3646                 if (ret < 0)
3647                         break;
3648
3649                 page_idx += count;
3650                 reserved_blocks += ret;
3651         }
3652
3653         filemap_invalidate_unlock(inode->i_mapping);
3654         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3655
3656         if (ret >= 0) {
3657                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3658                 inode->i_ctime = current_time(inode);
3659                 f2fs_mark_inode_dirty_sync(inode, true);
3660         }
3661 unlock_inode:
3662         inode_unlock(inode);
3663 out:
3664         mnt_drop_write_file(filp);
3665
3666         if (ret >= 0) {
3667                 ret = put_user(reserved_blocks, (u64 __user *)arg);
3668         } else if (reserved_blocks &&
3669                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3670                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3671                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3672                         "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3673                         "run fsck to fix.",
3674                         __func__, inode->i_ino, inode->i_blocks,
3675                         reserved_blocks,
3676                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3677         }
3678
3679         return ret;
3680 }
3681
3682 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3683                 pgoff_t off, block_t block, block_t len, u32 flags)
3684 {
3685         sector_t sector = SECTOR_FROM_BLOCK(block);
3686         sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3687         int ret = 0;
3688
3689         if (flags & F2FS_TRIM_FILE_DISCARD) {
3690                 if (bdev_max_secure_erase_sectors(bdev))
3691                         ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3692                                         GFP_NOFS);
3693                 else
3694                         ret = blkdev_issue_discard(bdev, sector, nr_sects,
3695                                         GFP_NOFS);
3696         }
3697
3698         if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3699                 if (IS_ENCRYPTED(inode))
3700                         ret = fscrypt_zeroout_range(inode, off, block, len);
3701                 else
3702                         ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3703                                         GFP_NOFS, 0);
3704         }
3705
3706         return ret;
3707 }
3708
3709 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3710 {
3711         struct inode *inode = file_inode(filp);
3712         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3713         struct address_space *mapping = inode->i_mapping;
3714         struct block_device *prev_bdev = NULL;
3715         struct f2fs_sectrim_range range;
3716         pgoff_t index, pg_end, prev_index = 0;
3717         block_t prev_block = 0, len = 0;
3718         loff_t end_addr;
3719         bool to_end = false;
3720         int ret = 0;
3721
3722         if (!(filp->f_mode & FMODE_WRITE))
3723                 return -EBADF;
3724
3725         if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3726                                 sizeof(range)))
3727                 return -EFAULT;
3728
3729         if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3730                         !S_ISREG(inode->i_mode))
3731                 return -EINVAL;
3732
3733         if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3734                         !f2fs_hw_support_discard(sbi)) ||
3735                         ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3736                          IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3737                 return -EOPNOTSUPP;
3738
3739         file_start_write(filp);
3740         inode_lock(inode);
3741
3742         if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3743                         range.start >= inode->i_size) {
3744                 ret = -EINVAL;
3745                 goto err;
3746         }
3747
3748         if (range.len == 0)
3749                 goto err;
3750
3751         if (inode->i_size - range.start > range.len) {
3752                 end_addr = range.start + range.len;
3753         } else {
3754                 end_addr = range.len == (u64)-1 ?
3755                         sbi->sb->s_maxbytes : inode->i_size;
3756                 to_end = true;
3757         }
3758
3759         if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3760                         (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3761                 ret = -EINVAL;
3762                 goto err;
3763         }
3764
3765         index = F2FS_BYTES_TO_BLK(range.start);
3766         pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3767
3768         ret = f2fs_convert_inline_inode(inode);
3769         if (ret)
3770                 goto err;
3771
3772         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3773         filemap_invalidate_lock(mapping);
3774
3775         ret = filemap_write_and_wait_range(mapping, range.start,
3776                         to_end ? LLONG_MAX : end_addr - 1);
3777         if (ret)
3778                 goto out;
3779
3780         truncate_inode_pages_range(mapping, range.start,
3781                         to_end ? -1 : end_addr - 1);
3782
3783         while (index < pg_end) {
3784                 struct dnode_of_data dn;
3785                 pgoff_t end_offset, count;
3786                 int i;
3787
3788                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3789                 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3790                 if (ret) {
3791                         if (ret == -ENOENT) {
3792                                 index = f2fs_get_next_page_offset(&dn, index);
3793                                 continue;
3794                         }
3795                         goto out;
3796                 }
3797
3798                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3799                 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3800                 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3801                         struct block_device *cur_bdev;
3802                         block_t blkaddr = f2fs_data_blkaddr(&dn);
3803
3804                         if (!__is_valid_data_blkaddr(blkaddr))
3805                                 continue;
3806
3807                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3808                                                 DATA_GENERIC_ENHANCE)) {
3809                                 ret = -EFSCORRUPTED;
3810                                 f2fs_put_dnode(&dn);
3811                                 f2fs_handle_error(sbi,
3812                                                 ERROR_INVALID_BLKADDR);
3813                                 goto out;
3814                         }
3815
3816                         cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3817                         if (f2fs_is_multi_device(sbi)) {
3818                                 int di = f2fs_target_device_index(sbi, blkaddr);
3819
3820                                 blkaddr -= FDEV(di).start_blk;
3821                         }
3822
3823                         if (len) {
3824                                 if (prev_bdev == cur_bdev &&
3825                                                 index == prev_index + len &&
3826                                                 blkaddr == prev_block + len) {
3827                                         len++;
3828                                 } else {
3829                                         ret = f2fs_secure_erase(prev_bdev,
3830                                                 inode, prev_index, prev_block,
3831                                                 len, range.flags);
3832                                         if (ret) {
3833                                                 f2fs_put_dnode(&dn);
3834                                                 goto out;
3835                                         }
3836
3837                                         len = 0;
3838                                 }
3839                         }
3840
3841                         if (!len) {
3842                                 prev_bdev = cur_bdev;
3843                                 prev_index = index;
3844                                 prev_block = blkaddr;
3845                                 len = 1;
3846                         }
3847                 }
3848
3849                 f2fs_put_dnode(&dn);
3850
3851                 if (fatal_signal_pending(current)) {
3852                         ret = -EINTR;
3853                         goto out;
3854                 }
3855                 cond_resched();
3856         }
3857
3858         if (len)
3859                 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3860                                 prev_block, len, range.flags);
3861 out:
3862         filemap_invalidate_unlock(mapping);
3863         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3864 err:
3865         inode_unlock(inode);
3866         file_end_write(filp);
3867
3868         return ret;
3869 }
3870
3871 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3872 {
3873         struct inode *inode = file_inode(filp);
3874         struct f2fs_comp_option option;
3875
3876         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3877                 return -EOPNOTSUPP;
3878
3879         inode_lock_shared(inode);
3880
3881         if (!f2fs_compressed_file(inode)) {
3882                 inode_unlock_shared(inode);
3883                 return -ENODATA;
3884         }
3885
3886         option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3887         option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3888
3889         inode_unlock_shared(inode);
3890
3891         if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3892                                 sizeof(option)))
3893                 return -EFAULT;
3894
3895         return 0;
3896 }
3897
3898 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3899 {
3900         struct inode *inode = file_inode(filp);
3901         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3902         struct f2fs_comp_option option;
3903         int ret = 0;
3904
3905         if (!f2fs_sb_has_compression(sbi))
3906                 return -EOPNOTSUPP;
3907
3908         if (!(filp->f_mode & FMODE_WRITE))
3909                 return -EBADF;
3910
3911         if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3912                                 sizeof(option)))
3913                 return -EFAULT;
3914
3915         if (!f2fs_compressed_file(inode) ||
3916                         option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3917                         option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3918                         option.algorithm >= COMPRESS_MAX)
3919                 return -EINVAL;
3920
3921         file_start_write(filp);
3922         inode_lock(inode);
3923
3924         if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3925                 ret = -EBUSY;
3926                 goto out;
3927         }
3928
3929         if (inode->i_size != 0) {
3930                 ret = -EFBIG;
3931                 goto out;
3932         }
3933
3934         F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3935         F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3936         F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3937         f2fs_mark_inode_dirty_sync(inode, true);
3938
3939         if (!f2fs_is_compress_backend_ready(inode))
3940                 f2fs_warn(sbi, "compression algorithm is successfully set, "
3941                         "but current kernel doesn't support this algorithm.");
3942 out:
3943         inode_unlock(inode);
3944         file_end_write(filp);
3945
3946         return ret;
3947 }
3948
3949 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3950 {
3951         DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3952         struct address_space *mapping = inode->i_mapping;
3953         struct page *page;
3954         pgoff_t redirty_idx = page_idx;
3955         int i, page_len = 0, ret = 0;
3956
3957         page_cache_ra_unbounded(&ractl, len, 0);
3958
3959         for (i = 0; i < len; i++, page_idx++) {
3960                 page = read_cache_page(mapping, page_idx, NULL, NULL);
3961                 if (IS_ERR(page)) {
3962                         ret = PTR_ERR(page);
3963                         break;
3964                 }
3965                 page_len++;
3966         }
3967
3968         for (i = 0; i < page_len; i++, redirty_idx++) {
3969                 page = find_lock_page(mapping, redirty_idx);
3970
3971                 /* It will never fail, when page has pinned above */
3972                 f2fs_bug_on(F2FS_I_SB(inode), !page);
3973
3974                 set_page_dirty(page);
3975                 f2fs_put_page(page, 1);
3976                 f2fs_put_page(page, 0);
3977         }
3978
3979         return ret;
3980 }
3981
3982 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3983 {
3984         struct inode *inode = file_inode(filp);
3985         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3986         struct f2fs_inode_info *fi = F2FS_I(inode);
3987         pgoff_t page_idx = 0, last_idx;
3988         unsigned int blk_per_seg = sbi->blocks_per_seg;
3989         int cluster_size = fi->i_cluster_size;
3990         int count, ret;
3991
3992         if (!f2fs_sb_has_compression(sbi) ||
3993                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3994                 return -EOPNOTSUPP;
3995
3996         if (!(filp->f_mode & FMODE_WRITE))
3997                 return -EBADF;
3998
3999         if (!f2fs_compressed_file(inode))
4000                 return -EINVAL;
4001
4002         f2fs_balance_fs(F2FS_I_SB(inode), true);
4003
4004         file_start_write(filp);
4005         inode_lock(inode);
4006
4007         if (!f2fs_is_compress_backend_ready(inode)) {
4008                 ret = -EOPNOTSUPP;
4009                 goto out;
4010         }
4011
4012         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4013                 ret = -EINVAL;
4014                 goto out;
4015         }
4016
4017         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4018         if (ret)
4019                 goto out;
4020
4021         if (!atomic_read(&fi->i_compr_blocks))
4022                 goto out;
4023
4024         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4025
4026         count = last_idx - page_idx;
4027         while (count) {
4028                 int len = min(cluster_size, count);
4029
4030                 ret = redirty_blocks(inode, page_idx, len);
4031                 if (ret < 0)
4032                         break;
4033
4034                 if (get_dirty_pages(inode) >= blk_per_seg)
4035                         filemap_fdatawrite(inode->i_mapping);
4036
4037                 count -= len;
4038                 page_idx += len;
4039         }
4040
4041         if (!ret)
4042                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4043                                                         LLONG_MAX);
4044
4045         if (ret)
4046                 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4047                           __func__, ret);
4048 out:
4049         inode_unlock(inode);
4050         file_end_write(filp);
4051
4052         return ret;
4053 }
4054
4055 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4056 {
4057         struct inode *inode = file_inode(filp);
4058         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4059         pgoff_t page_idx = 0, last_idx;
4060         unsigned int blk_per_seg = sbi->blocks_per_seg;
4061         int cluster_size = F2FS_I(inode)->i_cluster_size;
4062         int count, ret;
4063
4064         if (!f2fs_sb_has_compression(sbi) ||
4065                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4066                 return -EOPNOTSUPP;
4067
4068         if (!(filp->f_mode & FMODE_WRITE))
4069                 return -EBADF;
4070
4071         if (!f2fs_compressed_file(inode))
4072                 return -EINVAL;
4073
4074         f2fs_balance_fs(F2FS_I_SB(inode), true);
4075
4076         file_start_write(filp);
4077         inode_lock(inode);
4078
4079         if (!f2fs_is_compress_backend_ready(inode)) {
4080                 ret = -EOPNOTSUPP;
4081                 goto out;
4082         }
4083
4084         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4085                 ret = -EINVAL;
4086                 goto out;
4087         }
4088
4089         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4090         if (ret)
4091                 goto out;
4092
4093         set_inode_flag(inode, FI_ENABLE_COMPRESS);
4094
4095         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4096
4097         count = last_idx - page_idx;
4098         while (count) {
4099                 int len = min(cluster_size, count);
4100
4101                 ret = redirty_blocks(inode, page_idx, len);
4102                 if (ret < 0)
4103                         break;
4104
4105                 if (get_dirty_pages(inode) >= blk_per_seg)
4106                         filemap_fdatawrite(inode->i_mapping);
4107
4108                 count -= len;
4109                 page_idx += len;
4110         }
4111
4112         if (!ret)
4113                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4114                                                         LLONG_MAX);
4115
4116         clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4117
4118         if (ret)
4119                 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4120                           __func__, ret);
4121 out:
4122         inode_unlock(inode);
4123         file_end_write(filp);
4124
4125         return ret;
4126 }
4127
4128 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4129 {
4130         switch (cmd) {
4131         case FS_IOC_GETVERSION:
4132                 return f2fs_ioc_getversion(filp, arg);
4133         case F2FS_IOC_START_ATOMIC_WRITE:
4134                 return f2fs_ioc_start_atomic_write(filp);
4135         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4136                 return f2fs_ioc_commit_atomic_write(filp);
4137         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4138                 return f2fs_ioc_abort_atomic_write(filp);
4139         case F2FS_IOC_START_VOLATILE_WRITE:
4140         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4141                 return -EOPNOTSUPP;
4142         case F2FS_IOC_SHUTDOWN:
4143                 return f2fs_ioc_shutdown(filp, arg);
4144         case FITRIM:
4145                 return f2fs_ioc_fitrim(filp, arg);
4146         case FS_IOC_SET_ENCRYPTION_POLICY:
4147                 return f2fs_ioc_set_encryption_policy(filp, arg);
4148         case FS_IOC_GET_ENCRYPTION_POLICY:
4149                 return f2fs_ioc_get_encryption_policy(filp, arg);
4150         case FS_IOC_GET_ENCRYPTION_PWSALT:
4151                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4152         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4153                 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4154         case FS_IOC_ADD_ENCRYPTION_KEY:
4155                 return f2fs_ioc_add_encryption_key(filp, arg);
4156         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4157                 return f2fs_ioc_remove_encryption_key(filp, arg);
4158         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4159                 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4160         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4161                 return f2fs_ioc_get_encryption_key_status(filp, arg);
4162         case FS_IOC_GET_ENCRYPTION_NONCE:
4163                 return f2fs_ioc_get_encryption_nonce(filp, arg);
4164         case F2FS_IOC_GARBAGE_COLLECT:
4165                 return f2fs_ioc_gc(filp, arg);
4166         case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4167                 return f2fs_ioc_gc_range(filp, arg);
4168         case F2FS_IOC_WRITE_CHECKPOINT:
4169                 return f2fs_ioc_write_checkpoint(filp, arg);
4170         case F2FS_IOC_DEFRAGMENT:
4171                 return f2fs_ioc_defragment(filp, arg);
4172         case F2FS_IOC_MOVE_RANGE:
4173                 return f2fs_ioc_move_range(filp, arg);
4174         case F2FS_IOC_FLUSH_DEVICE:
4175                 return f2fs_ioc_flush_device(filp, arg);
4176         case F2FS_IOC_GET_FEATURES:
4177                 return f2fs_ioc_get_features(filp, arg);
4178         case F2FS_IOC_GET_PIN_FILE:
4179                 return f2fs_ioc_get_pin_file(filp, arg);
4180         case F2FS_IOC_SET_PIN_FILE:
4181                 return f2fs_ioc_set_pin_file(filp, arg);
4182         case F2FS_IOC_PRECACHE_EXTENTS:
4183                 return f2fs_ioc_precache_extents(filp, arg);
4184         case F2FS_IOC_RESIZE_FS:
4185                 return f2fs_ioc_resize_fs(filp, arg);
4186         case FS_IOC_ENABLE_VERITY:
4187                 return f2fs_ioc_enable_verity(filp, arg);
4188         case FS_IOC_MEASURE_VERITY:
4189                 return f2fs_ioc_measure_verity(filp, arg);
4190         case FS_IOC_READ_VERITY_METADATA:
4191                 return f2fs_ioc_read_verity_metadata(filp, arg);
4192         case FS_IOC_GETFSLABEL:
4193                 return f2fs_ioc_getfslabel(filp, arg);
4194         case FS_IOC_SETFSLABEL:
4195                 return f2fs_ioc_setfslabel(filp, arg);
4196         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4197                 return f2fs_get_compress_blocks(filp, arg);
4198         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4199                 return f2fs_release_compress_blocks(filp, arg);
4200         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4201                 return f2fs_reserve_compress_blocks(filp, arg);
4202         case F2FS_IOC_SEC_TRIM_FILE:
4203                 return f2fs_sec_trim_file(filp, arg);
4204         case F2FS_IOC_GET_COMPRESS_OPTION:
4205                 return f2fs_ioc_get_compress_option(filp, arg);
4206         case F2FS_IOC_SET_COMPRESS_OPTION:
4207                 return f2fs_ioc_set_compress_option(filp, arg);
4208         case F2FS_IOC_DECOMPRESS_FILE:
4209                 return f2fs_ioc_decompress_file(filp, arg);
4210         case F2FS_IOC_COMPRESS_FILE:
4211                 return f2fs_ioc_compress_file(filp, arg);
4212         default:
4213                 return -ENOTTY;
4214         }
4215 }
4216
4217 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4218 {
4219         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4220                 return -EIO;
4221         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4222                 return -ENOSPC;
4223
4224         return __f2fs_ioctl(filp, cmd, arg);
4225 }
4226
4227 /*
4228  * Return %true if the given read or write request should use direct I/O, or
4229  * %false if it should use buffered I/O.
4230  */
4231 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4232                                 struct iov_iter *iter)
4233 {
4234         unsigned int align;
4235
4236         if (!(iocb->ki_flags & IOCB_DIRECT))
4237                 return false;
4238
4239         if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4240                 return false;
4241
4242         /*
4243          * Direct I/O not aligned to the disk's logical_block_size will be
4244          * attempted, but will fail with -EINVAL.
4245          *
4246          * f2fs additionally requires that direct I/O be aligned to the
4247          * filesystem block size, which is often a stricter requirement.
4248          * However, f2fs traditionally falls back to buffered I/O on requests
4249          * that are logical_block_size-aligned but not fs-block aligned.
4250          *
4251          * The below logic implements this behavior.
4252          */
4253         align = iocb->ki_pos | iov_iter_alignment(iter);
4254         if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4255             IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4256                 return false;
4257
4258         return true;
4259 }
4260
4261 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4262                                 unsigned int flags)
4263 {
4264         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4265
4266         dec_page_count(sbi, F2FS_DIO_READ);
4267         if (error)
4268                 return error;
4269         f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4270         return 0;
4271 }
4272
4273 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4274         .end_io = f2fs_dio_read_end_io,
4275 };
4276
4277 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4278 {
4279         struct file *file = iocb->ki_filp;
4280         struct inode *inode = file_inode(file);
4281         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4282         struct f2fs_inode_info *fi = F2FS_I(inode);
4283         const loff_t pos = iocb->ki_pos;
4284         const size_t count = iov_iter_count(to);
4285         struct iomap_dio *dio;
4286         ssize_t ret;
4287
4288         if (count == 0)
4289                 return 0; /* skip atime update */
4290
4291         trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4292
4293         if (iocb->ki_flags & IOCB_NOWAIT) {
4294                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4295                         ret = -EAGAIN;
4296                         goto out;
4297                 }
4298         } else {
4299                 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4300         }
4301
4302         /*
4303          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4304          * the higher-level function iomap_dio_rw() in order to ensure that the
4305          * F2FS_DIO_READ counter will be decremented correctly in all cases.
4306          */
4307         inc_page_count(sbi, F2FS_DIO_READ);
4308         dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4309                              &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4310         if (IS_ERR_OR_NULL(dio)) {
4311                 ret = PTR_ERR_OR_ZERO(dio);
4312                 if (ret != -EIOCBQUEUED)
4313                         dec_page_count(sbi, F2FS_DIO_READ);
4314         } else {
4315                 ret = iomap_dio_complete(dio);
4316         }
4317
4318         f2fs_up_read(&fi->i_gc_rwsem[READ]);
4319
4320         file_accessed(file);
4321 out:
4322         trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4323         return ret;
4324 }
4325
4326 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4327 {
4328         struct inode *inode = file_inode(iocb->ki_filp);
4329         const loff_t pos = iocb->ki_pos;
4330         ssize_t ret;
4331
4332         if (!f2fs_is_compress_backend_ready(inode))
4333                 return -EOPNOTSUPP;
4334
4335         if (trace_f2fs_dataread_start_enabled()) {
4336                 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4337                 char *path;
4338
4339                 if (!p)
4340                         goto skip_read_trace;
4341
4342                 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4343                 if (IS_ERR(path)) {
4344                         kfree(p);
4345                         goto skip_read_trace;
4346                 }
4347
4348                 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4349                                         current->pid, path, current->comm);
4350                 kfree(p);
4351         }
4352 skip_read_trace:
4353         if (f2fs_should_use_dio(inode, iocb, to)) {
4354                 ret = f2fs_dio_read_iter(iocb, to);
4355         } else {
4356                 ret = filemap_read(iocb, to, 0);
4357                 if (ret > 0)
4358                         f2fs_update_iostat(F2FS_I_SB(inode), inode,
4359                                                 APP_BUFFERED_READ_IO, ret);
4360         }
4361         if (trace_f2fs_dataread_end_enabled())
4362                 trace_f2fs_dataread_end(inode, pos, ret);
4363         return ret;
4364 }
4365
4366 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4367 {
4368         struct file *file = iocb->ki_filp;
4369         struct inode *inode = file_inode(file);
4370         ssize_t count;
4371         int err;
4372
4373         if (IS_IMMUTABLE(inode))
4374                 return -EPERM;
4375
4376         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4377                 return -EPERM;
4378
4379         count = generic_write_checks(iocb, from);
4380         if (count <= 0)
4381                 return count;
4382
4383         err = file_modified(file);
4384         if (err)
4385                 return err;
4386         return count;
4387 }
4388
4389 /*
4390  * Preallocate blocks for a write request, if it is possible and helpful to do
4391  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4392  * blocks were preallocated, or a negative errno value if something went
4393  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4394  * requested blocks (not just some of them) have been allocated.
4395  */
4396 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4397                                    bool dio)
4398 {
4399         struct inode *inode = file_inode(iocb->ki_filp);
4400         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4401         const loff_t pos = iocb->ki_pos;
4402         const size_t count = iov_iter_count(iter);
4403         struct f2fs_map_blocks map = {};
4404         int flag;
4405         int ret;
4406
4407         /* If it will be an out-of-place direct write, don't bother. */
4408         if (dio && f2fs_lfs_mode(sbi))
4409                 return 0;
4410         /*
4411          * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4412          * buffered IO, if DIO meets any holes.
4413          */
4414         if (dio && i_size_read(inode) &&
4415                 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4416                 return 0;
4417
4418         /* No-wait I/O can't allocate blocks. */
4419         if (iocb->ki_flags & IOCB_NOWAIT)
4420                 return 0;
4421
4422         /* If it will be a short write, don't bother. */
4423         if (fault_in_iov_iter_readable(iter, count))
4424                 return 0;
4425
4426         if (f2fs_has_inline_data(inode)) {
4427                 /* If the data will fit inline, don't bother. */
4428                 if (pos + count <= MAX_INLINE_DATA(inode))
4429                         return 0;
4430                 ret = f2fs_convert_inline_inode(inode);
4431                 if (ret)
4432                         return ret;
4433         }
4434
4435         /* Do not preallocate blocks that will be written partially in 4KB. */
4436         map.m_lblk = F2FS_BLK_ALIGN(pos);
4437         map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4438         if (map.m_len > map.m_lblk)
4439                 map.m_len -= map.m_lblk;
4440         else
4441                 map.m_len = 0;
4442         map.m_may_create = true;
4443         if (dio) {
4444                 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4445                 flag = F2FS_GET_BLOCK_PRE_DIO;
4446         } else {
4447                 map.m_seg_type = NO_CHECK_TYPE;
4448                 flag = F2FS_GET_BLOCK_PRE_AIO;
4449         }
4450
4451         ret = f2fs_map_blocks(inode, &map, 1, flag);
4452         /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4453         if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4454                 return ret;
4455         if (ret == 0)
4456                 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4457         return map.m_len;
4458 }
4459
4460 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4461                                         struct iov_iter *from)
4462 {
4463         struct file *file = iocb->ki_filp;
4464         struct inode *inode = file_inode(file);
4465         ssize_t ret;
4466
4467         if (iocb->ki_flags & IOCB_NOWAIT)
4468                 return -EOPNOTSUPP;
4469
4470         current->backing_dev_info = inode_to_bdi(inode);
4471         ret = generic_perform_write(iocb, from);
4472         current->backing_dev_info = NULL;
4473
4474         if (ret > 0) {
4475                 iocb->ki_pos += ret;
4476                 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4477                                                 APP_BUFFERED_IO, ret);
4478         }
4479         return ret;
4480 }
4481
4482 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4483                                  unsigned int flags)
4484 {
4485         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4486
4487         dec_page_count(sbi, F2FS_DIO_WRITE);
4488         if (error)
4489                 return error;
4490         f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4491         return 0;
4492 }
4493
4494 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4495         .end_io = f2fs_dio_write_end_io,
4496 };
4497
4498 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4499                                    bool *may_need_sync)
4500 {
4501         struct file *file = iocb->ki_filp;
4502         struct inode *inode = file_inode(file);
4503         struct f2fs_inode_info *fi = F2FS_I(inode);
4504         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4505         const bool do_opu = f2fs_lfs_mode(sbi);
4506         const loff_t pos = iocb->ki_pos;
4507         const ssize_t count = iov_iter_count(from);
4508         unsigned int dio_flags;
4509         struct iomap_dio *dio;
4510         ssize_t ret;
4511
4512         trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4513
4514         if (iocb->ki_flags & IOCB_NOWAIT) {
4515                 /* f2fs_convert_inline_inode() and block allocation can block */
4516                 if (f2fs_has_inline_data(inode) ||
4517                     !f2fs_overwrite_io(inode, pos, count)) {
4518                         ret = -EAGAIN;
4519                         goto out;
4520                 }
4521
4522                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4523                         ret = -EAGAIN;
4524                         goto out;
4525                 }
4526                 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4527                         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4528                         ret = -EAGAIN;
4529                         goto out;
4530                 }
4531         } else {
4532                 ret = f2fs_convert_inline_inode(inode);
4533                 if (ret)
4534                         goto out;
4535
4536                 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4537                 if (do_opu)
4538                         f2fs_down_read(&fi->i_gc_rwsem[READ]);
4539         }
4540
4541         /*
4542          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4543          * the higher-level function iomap_dio_rw() in order to ensure that the
4544          * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4545          */
4546         inc_page_count(sbi, F2FS_DIO_WRITE);
4547         dio_flags = 0;
4548         if (pos + count > inode->i_size)
4549                 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4550         dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4551                              &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4552         if (IS_ERR_OR_NULL(dio)) {
4553                 ret = PTR_ERR_OR_ZERO(dio);
4554                 if (ret == -ENOTBLK)
4555                         ret = 0;
4556                 if (ret != -EIOCBQUEUED)
4557                         dec_page_count(sbi, F2FS_DIO_WRITE);
4558         } else {
4559                 ret = iomap_dio_complete(dio);
4560         }
4561
4562         if (do_opu)
4563                 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4564         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4565
4566         if (ret < 0)
4567                 goto out;
4568         if (pos + ret > inode->i_size)
4569                 f2fs_i_size_write(inode, pos + ret);
4570         if (!do_opu)
4571                 set_inode_flag(inode, FI_UPDATE_WRITE);
4572
4573         if (iov_iter_count(from)) {
4574                 ssize_t ret2;
4575                 loff_t bufio_start_pos = iocb->ki_pos;
4576
4577                 /*
4578                  * The direct write was partial, so we need to fall back to a
4579                  * buffered write for the remainder.
4580                  */
4581
4582                 ret2 = f2fs_buffered_write_iter(iocb, from);
4583                 if (iov_iter_count(from))
4584                         f2fs_write_failed(inode, iocb->ki_pos);
4585                 if (ret2 < 0)
4586                         goto out;
4587
4588                 /*
4589                  * Ensure that the pagecache pages are written to disk and
4590                  * invalidated to preserve the expected O_DIRECT semantics.
4591                  */
4592                 if (ret2 > 0) {
4593                         loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4594
4595                         ret += ret2;
4596
4597                         ret2 = filemap_write_and_wait_range(file->f_mapping,
4598                                                             bufio_start_pos,
4599                                                             bufio_end_pos);
4600                         if (ret2 < 0)
4601                                 goto out;
4602                         invalidate_mapping_pages(file->f_mapping,
4603                                                  bufio_start_pos >> PAGE_SHIFT,
4604                                                  bufio_end_pos >> PAGE_SHIFT);
4605                 }
4606         } else {
4607                 /* iomap_dio_rw() already handled the generic_write_sync(). */
4608                 *may_need_sync = false;
4609         }
4610 out:
4611         trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4612         return ret;
4613 }
4614
4615 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4616 {
4617         struct inode *inode = file_inode(iocb->ki_filp);
4618         const loff_t orig_pos = iocb->ki_pos;
4619         const size_t orig_count = iov_iter_count(from);
4620         loff_t target_size;
4621         bool dio;
4622         bool may_need_sync = true;
4623         int preallocated;
4624         ssize_t ret;
4625
4626         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4627                 ret = -EIO;
4628                 goto out;
4629         }
4630
4631         if (!f2fs_is_compress_backend_ready(inode)) {
4632                 ret = -EOPNOTSUPP;
4633                 goto out;
4634         }
4635
4636         if (iocb->ki_flags & IOCB_NOWAIT) {
4637                 if (!inode_trylock(inode)) {
4638                         ret = -EAGAIN;
4639                         goto out;
4640                 }
4641         } else {
4642                 inode_lock(inode);
4643         }
4644
4645         ret = f2fs_write_checks(iocb, from);
4646         if (ret <= 0)
4647                 goto out_unlock;
4648
4649         /* Determine whether we will do a direct write or a buffered write. */
4650         dio = f2fs_should_use_dio(inode, iocb, from);
4651
4652         /* Possibly preallocate the blocks for the write. */
4653         target_size = iocb->ki_pos + iov_iter_count(from);
4654         preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4655         if (preallocated < 0) {
4656                 ret = preallocated;
4657         } else {
4658                 if (trace_f2fs_datawrite_start_enabled()) {
4659                         char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4660                                                 PATH_MAX, GFP_KERNEL);
4661                         char *path;
4662
4663                         if (!p)
4664                                 goto skip_write_trace;
4665                         path = dentry_path_raw(file_dentry(iocb->ki_filp),
4666                                                                 p, PATH_MAX);
4667                         if (IS_ERR(path)) {
4668                                 kfree(p);
4669                                 goto skip_write_trace;
4670                         }
4671                         trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4672                                         current->pid, path, current->comm);
4673                         kfree(p);
4674                 }
4675 skip_write_trace:
4676                 /* Do the actual write. */
4677                 ret = dio ?
4678                         f2fs_dio_write_iter(iocb, from, &may_need_sync) :
4679                         f2fs_buffered_write_iter(iocb, from);
4680
4681                 if (trace_f2fs_datawrite_end_enabled())
4682                         trace_f2fs_datawrite_end(inode, orig_pos, ret);
4683         }
4684
4685         /* Don't leave any preallocated blocks around past i_size. */
4686         if (preallocated && i_size_read(inode) < target_size) {
4687                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4688                 filemap_invalidate_lock(inode->i_mapping);
4689                 if (!f2fs_truncate(inode))
4690                         file_dont_truncate(inode);
4691                 filemap_invalidate_unlock(inode->i_mapping);
4692                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4693         } else {
4694                 file_dont_truncate(inode);
4695         }
4696
4697         clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4698 out_unlock:
4699         inode_unlock(inode);
4700 out:
4701         trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4702         if (ret > 0 && may_need_sync)
4703                 ret = generic_write_sync(iocb, ret);
4704         return ret;
4705 }
4706
4707 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4708                 int advice)
4709 {
4710         struct address_space *mapping;
4711         struct backing_dev_info *bdi;
4712         struct inode *inode = file_inode(filp);
4713         int err;
4714
4715         if (advice == POSIX_FADV_SEQUENTIAL) {
4716                 if (S_ISFIFO(inode->i_mode))
4717                         return -ESPIPE;
4718
4719                 mapping = filp->f_mapping;
4720                 if (!mapping || len < 0)
4721                         return -EINVAL;
4722
4723                 bdi = inode_to_bdi(mapping->host);
4724                 filp->f_ra.ra_pages = bdi->ra_pages *
4725                         F2FS_I_SB(inode)->seq_file_ra_mul;
4726                 spin_lock(&filp->f_lock);
4727                 filp->f_mode &= ~FMODE_RANDOM;
4728                 spin_unlock(&filp->f_lock);
4729                 return 0;
4730         }
4731
4732         err = generic_fadvise(filp, offset, len, advice);
4733         if (!err && advice == POSIX_FADV_DONTNEED &&
4734                 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4735                 f2fs_compressed_file(inode))
4736                 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4737
4738         return err;
4739 }
4740
4741 #ifdef CONFIG_COMPAT
4742 struct compat_f2fs_gc_range {
4743         u32 sync;
4744         compat_u64 start;
4745         compat_u64 len;
4746 };
4747 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE        _IOW(F2FS_IOCTL_MAGIC, 11,\
4748                                                 struct compat_f2fs_gc_range)
4749
4750 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4751 {
4752         struct compat_f2fs_gc_range __user *urange;
4753         struct f2fs_gc_range range;
4754         int err;
4755
4756         urange = compat_ptr(arg);
4757         err = get_user(range.sync, &urange->sync);
4758         err |= get_user(range.start, &urange->start);
4759         err |= get_user(range.len, &urange->len);
4760         if (err)
4761                 return -EFAULT;
4762
4763         return __f2fs_ioc_gc_range(file, &range);
4764 }
4765
4766 struct compat_f2fs_move_range {
4767         u32 dst_fd;
4768         compat_u64 pos_in;
4769         compat_u64 pos_out;
4770         compat_u64 len;
4771 };
4772 #define F2FS_IOC32_MOVE_RANGE           _IOWR(F2FS_IOCTL_MAGIC, 9,      \
4773                                         struct compat_f2fs_move_range)
4774
4775 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4776 {
4777         struct compat_f2fs_move_range __user *urange;
4778         struct f2fs_move_range range;
4779         int err;
4780
4781         urange = compat_ptr(arg);
4782         err = get_user(range.dst_fd, &urange->dst_fd);
4783         err |= get_user(range.pos_in, &urange->pos_in);
4784         err |= get_user(range.pos_out, &urange->pos_out);
4785         err |= get_user(range.len, &urange->len);
4786         if (err)
4787                 return -EFAULT;
4788
4789         return __f2fs_ioc_move_range(file, &range);
4790 }
4791
4792 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4793 {
4794         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4795                 return -EIO;
4796         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4797                 return -ENOSPC;
4798
4799         switch (cmd) {
4800         case FS_IOC32_GETVERSION:
4801                 cmd = FS_IOC_GETVERSION;
4802                 break;
4803         case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4804                 return f2fs_compat_ioc_gc_range(file, arg);
4805         case F2FS_IOC32_MOVE_RANGE:
4806                 return f2fs_compat_ioc_move_range(file, arg);
4807         case F2FS_IOC_START_ATOMIC_WRITE:
4808         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4809         case F2FS_IOC_START_VOLATILE_WRITE:
4810         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4811         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4812         case F2FS_IOC_SHUTDOWN:
4813         case FITRIM:
4814         case FS_IOC_SET_ENCRYPTION_POLICY:
4815         case FS_IOC_GET_ENCRYPTION_PWSALT:
4816         case FS_IOC_GET_ENCRYPTION_POLICY:
4817         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4818         case FS_IOC_ADD_ENCRYPTION_KEY:
4819         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4820         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4821         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4822         case FS_IOC_GET_ENCRYPTION_NONCE:
4823         case F2FS_IOC_GARBAGE_COLLECT:
4824         case F2FS_IOC_WRITE_CHECKPOINT:
4825         case F2FS_IOC_DEFRAGMENT:
4826         case F2FS_IOC_FLUSH_DEVICE:
4827         case F2FS_IOC_GET_FEATURES:
4828         case F2FS_IOC_GET_PIN_FILE:
4829         case F2FS_IOC_SET_PIN_FILE:
4830         case F2FS_IOC_PRECACHE_EXTENTS:
4831         case F2FS_IOC_RESIZE_FS:
4832         case FS_IOC_ENABLE_VERITY:
4833         case FS_IOC_MEASURE_VERITY:
4834         case FS_IOC_READ_VERITY_METADATA:
4835         case FS_IOC_GETFSLABEL:
4836         case FS_IOC_SETFSLABEL:
4837         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4838         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4839         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4840         case F2FS_IOC_SEC_TRIM_FILE:
4841         case F2FS_IOC_GET_COMPRESS_OPTION:
4842         case F2FS_IOC_SET_COMPRESS_OPTION:
4843         case F2FS_IOC_DECOMPRESS_FILE:
4844         case F2FS_IOC_COMPRESS_FILE:
4845                 break;
4846         default:
4847                 return -ENOIOCTLCMD;
4848         }
4849         return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4850 }
4851 #endif
4852
4853 const struct file_operations f2fs_file_operations = {
4854         .llseek         = f2fs_llseek,
4855         .read_iter      = f2fs_file_read_iter,
4856         .write_iter     = f2fs_file_write_iter,
4857         .open           = f2fs_file_open,
4858         .release        = f2fs_release_file,
4859         .mmap           = f2fs_file_mmap,
4860         .flush          = f2fs_file_flush,
4861         .fsync          = f2fs_sync_file,
4862         .fallocate      = f2fs_fallocate,
4863         .unlocked_ioctl = f2fs_ioctl,
4864 #ifdef CONFIG_COMPAT
4865         .compat_ioctl   = f2fs_compat_ioctl,
4866 #endif
4867         .splice_read    = generic_file_splice_read,
4868         .splice_write   = iter_file_splice_write,
4869         .fadvise        = f2fs_file_fadvise,
4870 };