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                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 117                 err = f2fs_get_block_locked(&dn, page->index);
 118         }
 119
 120 #ifdef CONFIG_F2FS_FS_COMPRESSION
 121         if (!need_alloc) {
 122                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 123                 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
 124                 f2fs_put_dnode(&dn);
 125         }
 126 #endif
 127         if (err) {
 128                 unlock_page(page);
 129                 goto out_sem;
 130         }
 131
 132         f2fs_wait_on_page_writeback(page, DATA, false, true);
 133
 134         /* wait for GCed page writeback via META_MAPPING */
 135         f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
 136
 137         /*
 138          * check to see if the page is mapped already (no holes)
 139          */
 140         if (PageMappedToDisk(page))
 141                 goto out_sem;
 142
 143         /* page is wholly or partially inside EOF */
 144         if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
 145                                                 i_size_read(inode)) {
 146                 loff_t offset;
 147
 148                 offset = i_size_read(inode) & ~PAGE_MASK;
 149                 zero_user_segment(page, offset, PAGE_SIZE);
 150         }
 151         set_page_dirty(page);
 152         if (!PageUptodate(page))
 153                 SetPageUptodate(page);
 154
 155         f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
 156         f2fs_update_time(sbi, REQ_TIME);
 157
 158         trace_f2fs_vm_page_mkwrite(page, DATA);
 159 out_sem:
 160         filemap_invalidate_unlock_shared(inode->i_mapping);
 161
 162         sb_end_pagefault(inode->i_sb);
 163 err:
 164         return block_page_mkwrite_return(err);
 165 }
 166
 167 static const struct vm_operations_struct f2fs_file_vm_ops = {
 168         .fault          = f2fs_filemap_fault,
 169         .map_pages      = filemap_map_pages,
 170         .page_mkwrite   = f2fs_vm_page_mkwrite,
 171 };
 172
 173 static int get_parent_ino(struct inode *inode, nid_t *pino)
 174 {
 175         struct dentry *dentry;
 176
 177         /*
 178          * Make sure to get the non-deleted alias.  The alias associated with
 179          * the open file descriptor being fsync()'ed may be deleted already.
 180          */
 181         dentry = d_find_alias(inode);
 182         if (!dentry)
 183                 return 0;
 184
 185         *pino = parent_ino(dentry);
 186         dput(dentry);
 187         return 1;
 188 }
 189
 190 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
 191 {
 192         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 193         enum cp_reason_type cp_reason = CP_NO_NEEDED;
 194
 195         if (!S_ISREG(inode->i_mode))
 196                 cp_reason = CP_NON_REGULAR;
 197         else if (f2fs_compressed_file(inode))
 198                 cp_reason = CP_COMPRESSED;
 199         else if (inode->i_nlink != 1)
 200                 cp_reason = CP_HARDLINK;
 201         else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
 202                 cp_reason = CP_SB_NEED_CP;
 203         else if (file_wrong_pino(inode))
 204                 cp_reason = CP_WRONG_PINO;
 205         else if (!f2fs_space_for_roll_forward(sbi))
 206                 cp_reason = CP_NO_SPC_ROLL;
 207         else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
 208                 cp_reason = CP_NODE_NEED_CP;
 209         else if (test_opt(sbi, FASTBOOT))
 210                 cp_reason = CP_FASTBOOT_MODE;
 211         else if (F2FS_OPTION(sbi).active_logs == 2)
 212                 cp_reason = CP_SPEC_LOG_NUM;
 213         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
 214                 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
 215                 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
 216                                                         TRANS_DIR_INO))
 217                 cp_reason = CP_RECOVER_DIR;
 218
 219         return cp_reason;
 220 }
 221
 222 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
 223 {
 224         struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
 225         bool ret = false;
 226         /* But we need to avoid that there are some inode updates */
 227         if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
 228                 ret = true;
 229         f2fs_put_page(i, 0);
 230         return ret;
 231 }
 232
 233 static void try_to_fix_pino(struct inode *inode)
 234 {
 235         struct f2fs_inode_info *fi = F2FS_I(inode);
 236         nid_t pino;
 237
 238         f2fs_down_write(&fi->i_sem);
 239         if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
 240                         get_parent_ino(inode, &pino)) {
 241                 f2fs_i_pino_write(inode, pino);
 242                 file_got_pino(inode);
 243         }
 244         f2fs_up_write(&fi->i_sem);
 245 }
 246
 247 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
 248                                                 int datasync, bool atomic)
 249 {
 250         struct inode *inode = file->f_mapping->host;
 251         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 252         nid_t ino = inode->i_ino;
 253         int ret = 0;
 254         enum cp_reason_type cp_reason = 0;
 255         struct writeback_control wbc = {
 256                 .sync_mode = WB_SYNC_ALL,
 257                 .nr_to_write = LONG_MAX,
 258                 .for_reclaim = 0,
 259         };
 260         unsigned int seq_id = 0;
 261
 262         if (unlikely(f2fs_readonly(inode->i_sb)))
 263                 return 0;
 264
 265         trace_f2fs_sync_file_enter(inode);
 266
 267         if (S_ISDIR(inode->i_mode))
 268                 goto go_write;
 269
 270         /* if fdatasync is triggered, let's do in-place-update */
 271         if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
 272                 set_inode_flag(inode, FI_NEED_IPU);
 273         ret = file_write_and_wait_range(file, start, end);
 274         clear_inode_flag(inode, FI_NEED_IPU);
 275
 276         if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
 277                 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 278                 return ret;
 279         }
 280
 281         /* if the inode is dirty, let's recover all the time */
 282         if (!f2fs_skip_inode_update(inode, datasync)) {
 283                 f2fs_write_inode(inode, NULL);
 284                 goto go_write;
 285         }
 286
 287         /*
 288          * if there is no written data, don't waste time to write recovery info.
 289          */
 290         if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
 291                         !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
 292
 293                 /* it may call write_inode just prior to fsync */
 294                 if (need_inode_page_update(sbi, ino))
 295                         goto go_write;
 296
 297                 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
 298                                 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
 299                         goto flush_out;
 300                 goto out;
 301         } else {
 302                 /*
 303                  * for OPU case, during fsync(), node can be persisted before
 304                  * data when lower device doesn't support write barrier, result
 305                  * in data corruption after SPO.
 306                  * So for strict fsync mode, force to use atomic write semantics
 307                  * to keep write order in between data/node and last node to
 308                  * avoid potential data corruption.
 309                  */
 310                 if (F2FS_OPTION(sbi).fsync_mode ==
 311                                 FSYNC_MODE_STRICT && !atomic)
 312                         atomic = true;
 313         }
 314 go_write:
 315         /*
 316          * Both of fdatasync() and fsync() are able to be recovered from
 317          * sudden-power-off.
 318          */
 319         f2fs_down_read(&F2FS_I(inode)->i_sem);
 320         cp_reason = need_do_checkpoint(inode);
 321         f2fs_up_read(&F2FS_I(inode)->i_sem);
 322
 323         if (cp_reason) {
 324                 /* all the dirty node pages should be flushed for POR */
 325                 ret = f2fs_sync_fs(inode->i_sb, 1);
 326
 327                 /*
 328                  * We've secured consistency through sync_fs. Following pino
 329                  * will be used only for fsynced inodes after checkpoint.
 330                  */
 331                 try_to_fix_pino(inode);
 332                 clear_inode_flag(inode, FI_APPEND_WRITE);
 333                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 334                 goto out;
 335         }
 336 sync_nodes:
 337         atomic_inc(&sbi->wb_sync_req[NODE]);
 338         ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
 339         atomic_dec(&sbi->wb_sync_req[NODE]);
 340         if (ret)
 341                 goto out;
 342
 343         /* if cp_error was enabled, we should avoid infinite loop */
 344         if (unlikely(f2fs_cp_error(sbi))) {
 345                 ret = -EIO;
 346                 goto out;
 347         }
 348
 349         if (f2fs_need_inode_block_update(sbi, ino)) {
 350                 f2fs_mark_inode_dirty_sync(inode, true);
 351                 f2fs_write_inode(inode, NULL);
 352                 goto sync_nodes;
 353         }
 354
 355         /*
 356          * If it's atomic_write, it's just fine to keep write ordering. So
 357          * here we don't need to wait for node write completion, since we use
 358          * node chain which serializes node blocks. If one of node writes are
 359          * reordered, we can see simply broken chain, resulting in stopping
 360          * roll-forward recovery. It means we'll recover all or none node blocks
 361          * given fsync mark.
 362          */
 363         if (!atomic) {
 364                 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
 365                 if (ret)
 366                         goto out;
 367         }
 368
 369         /* once recovery info is written, don't need to tack this */
 370         f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
 371         clear_inode_flag(inode, FI_APPEND_WRITE);
 372 flush_out:
 373         if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
 374             (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
 375                 ret = f2fs_issue_flush(sbi, inode->i_ino);
 376         if (!ret) {
 377                 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
 378                 clear_inode_flag(inode, FI_UPDATE_WRITE);
 379                 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
 380         }
 381         f2fs_update_time(sbi, REQ_TIME);
 382 out:
 383         trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
 384         return ret;
 385 }
 386
 387 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 388 {
 389         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
 390                 return -EIO;
 391         return f2fs_do_sync_file(file, start, end, datasync, false);
 392 }
 393
 394 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
 395                                 pgoff_t index, int whence)
 396 {
 397         switch (whence) {
 398         case SEEK_DATA:
 399                 if (__is_valid_data_blkaddr(blkaddr))
 400                         return true;
 401                 if (blkaddr == NEW_ADDR &&
 402                     xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
 403                         return true;
 404                 break;
 405         case SEEK_HOLE:
 406                 if (blkaddr == NULL_ADDR)
 407                         return true;
 408                 break;
 409         }
 410         return false;
 411 }
 412
 413 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 414 {
 415         struct inode *inode = file->f_mapping->host;
 416         loff_t maxbytes = inode->i_sb->s_maxbytes;
 417         struct dnode_of_data dn;
 418         pgoff_t pgofs, end_offset;
 419         loff_t data_ofs = offset;
 420         loff_t isize;
 421         int err = 0;
 422
 423         inode_lock(inode);
 424
 425         isize = i_size_read(inode);
 426         if (offset >= isize)
 427                 goto fail;
 428
 429         /* handle inline data case */
 430         if (f2fs_has_inline_data(inode)) {
 431                 if (whence == SEEK_HOLE) {
 432                         data_ofs = isize;
 433                         goto found;
 434                 } else if (whence == SEEK_DATA) {
 435                         data_ofs = offset;
 436                         goto found;
 437                 }
 438         }
 439
 440         pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
 441
 442         for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 443                 set_new_dnode(&dn, inode, NULL, NULL, 0);
 444                 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
 445                 if (err && err != -ENOENT) {
 446                         goto fail;
 447                 } else if (err == -ENOENT) {
 448                         /* direct node does not exists */
 449                         if (whence == SEEK_DATA) {
 450                                 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
 451                                 continue;
 452                         } else {
 453                                 goto found;
 454                         }
 455                 }
 456
 457                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
 458
 459                 /* find data/hole in dnode block */
 460                 for (; dn.ofs_in_node < end_offset;
 461                                 dn.ofs_in_node++, pgofs++,
 462                                 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
 463                         block_t blkaddr;
 464
 465                         blkaddr = f2fs_data_blkaddr(&dn);
 466
 467                         if (__is_valid_data_blkaddr(blkaddr) &&
 468                                 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
 469                                         blkaddr, DATA_GENERIC_ENHANCE)) {
 470                                 f2fs_put_dnode(&dn);
 471                                 goto fail;
 472                         }
 473
 474                         if (__found_offset(file->f_mapping, blkaddr,
 475                                                         pgofs, whence)) {
 476                                 f2fs_put_dnode(&dn);
 477                                 goto found;
 478                         }
 479                 }
 480                 f2fs_put_dnode(&dn);
 481         }
 482
 483         if (whence == SEEK_DATA)
 484                 goto fail;
 485 found:
 486         if (whence == SEEK_HOLE && data_ofs > isize)
 487                 data_ofs = isize;
 488         inode_unlock(inode);
 489         return vfs_setpos(file, data_ofs, maxbytes);
 490 fail:
 491         inode_unlock(inode);
 492         return -ENXIO;
 493 }
 494
 495 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
 496 {
 497         struct inode *inode = file->f_mapping->host;
 498         loff_t maxbytes = inode->i_sb->s_maxbytes;
 499
 500         if (f2fs_compressed_file(inode))
 501                 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
 502
 503         switch (whence) {
 504         case SEEK_SET:
 505         case SEEK_CUR:
 506         case SEEK_END:
 507                 return generic_file_llseek_size(file, offset, whence,
 508                                                 maxbytes, i_size_read(inode));
 509         case SEEK_DATA:
 510         case SEEK_HOLE:
 511                 if (offset < 0)
 512                         return -ENXIO;
 513                 return f2fs_seek_block(file, offset, whence);
 514         }
 515
 516         return -EINVAL;
 517 }
 518
 519 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 520 {
 521         struct inode *inode = file_inode(file);
 522
 523         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 524                 return -EIO;
 525
 526         if (!f2fs_is_compress_backend_ready(inode))
 527                 return -EOPNOTSUPP;
 528
 529         file_accessed(file);
 530         vma->vm_ops = &f2fs_file_vm_ops;
 531         set_inode_flag(inode, FI_MMAP_FILE);
 532         return 0;
 533 }
 534
 535 static int f2fs_file_open(struct inode *inode, struct file *filp)
 536 {
 537         int err = fscrypt_file_open(inode, filp);
 538
 539         if (err)
 540                 return err;
 541
 542         if (!f2fs_is_compress_backend_ready(inode))
 543                 return -EOPNOTSUPP;
 544
 545         err = fsverity_file_open(inode, filp);
 546         if (err)
 547                 return err;
 548
 549         filp->f_mode |= FMODE_NOWAIT;
 550
 551         return dquot_file_open(inode, filp);
 552 }
 553
 554 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 555 {
 556         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 557         struct f2fs_node *raw_node;
 558         int nr_free = 0, ofs = dn->ofs_in_node, len = count;
 559         __le32 *addr;
 560         int base = 0;
 561         bool compressed_cluster = false;
 562         int cluster_index = 0, valid_blocks = 0;
 563         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
 564         bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
 565
 566         if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
 567                 base = get_extra_isize(dn->inode);
 568
 569         raw_node = F2FS_NODE(dn->node_page);
 570         addr = blkaddr_in_node(raw_node) + base + ofs;
 571
 572         /* Assumption: truncation starts with cluster */
 573         for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
 574                 block_t blkaddr = le32_to_cpu(*addr);
 575
 576                 if (f2fs_compressed_file(dn->inode) &&
 577                                         !(cluster_index & (cluster_size - 1))) {
 578                         if (compressed_cluster)
 579                                 f2fs_i_compr_blocks_update(dn->inode,
 580                                                         valid_blocks, false);
 581                         compressed_cluster = (blkaddr == COMPRESS_ADDR);
 582                         valid_blocks = 0;
 583                 }
 584
 585                 if (blkaddr == NULL_ADDR)
 586                         continue;
 587
 588                 dn->data_blkaddr = NULL_ADDR;
 589                 f2fs_set_data_blkaddr(dn);
 590
 591                 if (__is_valid_data_blkaddr(blkaddr)) {
 592                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
 593                                         DATA_GENERIC_ENHANCE))
 594                                 continue;
 595                         if (compressed_cluster)
 596                                 valid_blocks++;
 597                 }
 598
 599                 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
 600                         clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
 601
 602                 f2fs_invalidate_blocks(sbi, blkaddr);
 603
 604                 if (!released || blkaddr != COMPRESS_ADDR)
 605                         nr_free++;
 606         }
 607
 608         if (compressed_cluster)
 609                 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
 610
 611         if (nr_free) {
 612                 pgoff_t fofs;
 613                 /*
 614                  * once we invalidate valid blkaddr in range [ofs, ofs + count],
 615                  * we will invalidate all blkaddr in the whole range.
 616                  */
 617                 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
 618                                                         dn->inode) + ofs;
 619                 f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
 620                 f2fs_update_age_extent_cache_range(dn, fofs, len);
 621                 dec_valid_block_count(sbi, dn->inode, nr_free);
 622         }
 623         dn->ofs_in_node = ofs;
 624
 625         f2fs_update_time(sbi, REQ_TIME);
 626         trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
 627                                          dn->ofs_in_node, nr_free);
 628 }
 629
 630 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
 631 {
 632         f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
 633 }
 634
 635 static int truncate_partial_data_page(struct inode *inode, u64 from,
 636                                                                 bool cache_only)
 637 {
 638         loff_t offset = from & (PAGE_SIZE - 1);
 639         pgoff_t index = from >> PAGE_SHIFT;
 640         struct address_space *mapping = inode->i_mapping;
 641         struct page *page;
 642
 643         if (!offset && !cache_only)
 644                 return 0;
 645
 646         if (cache_only) {
 647                 page = find_lock_page(mapping, index);
 648                 if (page && PageUptodate(page))
 649                         goto truncate_out;
 650                 f2fs_put_page(page, 1);
 651                 return 0;
 652         }
 653
 654         page = f2fs_get_lock_data_page(inode, index, true);
 655         if (IS_ERR(page))
 656                 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
 657 truncate_out:
 658         f2fs_wait_on_page_writeback(page, DATA, true, true);
 659         zero_user(page, offset, PAGE_SIZE - offset);
 660
 661         /* An encrypted inode should have a key and truncate the last page. */
 662         f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
 663         if (!cache_only)
 664                 set_page_dirty(page);
 665         f2fs_put_page(page, 1);
 666         return 0;
 667 }
 668
 669 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
 670 {
 671         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 672         struct dnode_of_data dn;
 673         pgoff_t free_from;
 674         int count = 0, err = 0;
 675         struct page *ipage;
 676         bool truncate_page = false;
 677
 678         trace_f2fs_truncate_blocks_enter(inode, from);
 679
 680         free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
 681
 682         if (free_from >= max_file_blocks(inode))
 683                 goto free_partial;
 684
 685         if (lock)
 686                 f2fs_lock_op(sbi);
 687
 688         ipage = f2fs_get_node_page(sbi, inode->i_ino);
 689         if (IS_ERR(ipage)) {
 690                 err = PTR_ERR(ipage);
 691                 goto out;
 692         }
 693
 694         if (f2fs_has_inline_data(inode)) {
 695                 f2fs_truncate_inline_inode(inode, ipage, from);
 696                 f2fs_put_page(ipage, 1);
 697                 truncate_page = true;
 698                 goto out;
 699         }
 700
 701         set_new_dnode(&dn, inode, ipage, NULL, 0);
 702         err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
 703         if (err) {
 704                 if (err == -ENOENT)
 705                         goto free_next;
 706                 goto out;
 707         }
 708
 709         count = ADDRS_PER_PAGE(dn.node_page, inode);
 710
 711         count -= dn.ofs_in_node;
 712         f2fs_bug_on(sbi, count < 0);
 713
 714         if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
 715                 f2fs_truncate_data_blocks_range(&dn, count);
 716                 free_from += count;
 717         }
 718
 719         f2fs_put_dnode(&dn);
 720 free_next:
 721         err = f2fs_truncate_inode_blocks(inode, free_from);
 722 out:
 723         if (lock)
 724                 f2fs_unlock_op(sbi);
 725 free_partial:
 726         /* lastly zero out the first data page */
 727         if (!err)
 728                 err = truncate_partial_data_page(inode, from, truncate_page);
 729
 730         trace_f2fs_truncate_blocks_exit(inode, err);
 731         return err;
 732 }
 733
 734 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
 735 {
 736         u64 free_from = from;
 737         int err;
 738
 739 #ifdef CONFIG_F2FS_FS_COMPRESSION
 740         /*
 741          * for compressed file, only support cluster size
 742          * aligned truncation.
 743          */
 744         if (f2fs_compressed_file(inode))
 745                 free_from = round_up(from,
 746                                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
 747 #endif
 748
 749         err = f2fs_do_truncate_blocks(inode, free_from, lock);
 750         if (err)
 751                 return err;
 752
 753 #ifdef CONFIG_F2FS_FS_COMPRESSION
 754         /*
 755          * For compressed file, after release compress blocks, don't allow write
 756          * direct, but we should allow write direct after truncate to zero.
 757          */
 758         if (f2fs_compressed_file(inode) && !free_from
 759                         && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
 760                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 761
 762         if (from != free_from) {
 763                 err = f2fs_truncate_partial_cluster(inode, from, lock);
 764                 if (err)
 765                         return err;
 766         }
 767 #endif
 768
 769         return 0;
 770 }
 771
 772 int f2fs_truncate(struct inode *inode)
 773 {
 774         int err;
 775
 776         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 777                 return -EIO;
 778
 779         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 780                                 S_ISLNK(inode->i_mode)))
 781                 return 0;
 782
 783         trace_f2fs_truncate(inode);
 784
 785         if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
 786                 return -EIO;
 787
 788         err = f2fs_dquot_initialize(inode);
 789         if (err)
 790                 return err;
 791
 792         /* we should check inline_data size */
 793         if (!f2fs_may_inline_data(inode)) {
 794                 err = f2fs_convert_inline_inode(inode);
 795                 if (err)
 796                         return err;
 797         }
 798
 799         err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
 800         if (err)
 801                 return err;
 802
 803         inode->i_mtime = inode->i_ctime = current_time(inode);
 804         f2fs_mark_inode_dirty_sync(inode, false);
 805         return 0;
 806 }
 807
 808 static bool f2fs_force_buffered_io(struct inode *inode, int rw)
 809 {
 810         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 811
 812         if (!fscrypt_dio_supported(inode))
 813                 return true;
 814         if (fsverity_active(inode))
 815                 return true;
 816         if (f2fs_compressed_file(inode))
 817                 return true;
 818
 819         /* disallow direct IO if any of devices has unaligned blksize */
 820         if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
 821                 return true;
 822         /*
 823          * for blkzoned device, fallback direct IO to buffered IO, so
 824          * all IOs can be serialized by log-structured write.
 825          */
 826         if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
 827                 return true;
 828         if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi))
 829                 return true;
 830         if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
 831                 return true;
 832
 833         return false;
 834 }
 835
 836 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
 837                  struct kstat *stat, u32 request_mask, unsigned int query_flags)
 838 {
 839         struct inode *inode = d_inode(path->dentry);
 840         struct f2fs_inode_info *fi = F2FS_I(inode);
 841         struct f2fs_inode *ri = NULL;
 842         unsigned int flags;
 843
 844         if (f2fs_has_extra_attr(inode) &&
 845                         f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
 846                         F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
 847                 stat->result_mask |= STATX_BTIME;
 848                 stat->btime.tv_sec = fi->i_crtime.tv_sec;
 849                 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
 850         }
 851
 852         /*
 853          * Return the DIO alignment restrictions if requested.  We only return
 854          * this information when requested, since on encrypted files it might
 855          * take a fair bit of work to get if the file wasn't opened recently.
 856          *
 857          * f2fs sometimes supports DIO reads but not DIO writes.  STATX_DIOALIGN
 858          * cannot represent that, so in that case we report no DIO support.
 859          */
 860         if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
 861                 unsigned int bsize = i_blocksize(inode);
 862
 863                 stat->result_mask |= STATX_DIOALIGN;
 864                 if (!f2fs_force_buffered_io(inode, WRITE)) {
 865                         stat->dio_mem_align = bsize;
 866                         stat->dio_offset_align = bsize;
 867                 }
 868         }
 869
 870         flags = fi->i_flags;
 871         if (flags & F2FS_COMPR_FL)
 872                 stat->attributes |= STATX_ATTR_COMPRESSED;
 873         if (flags & F2FS_APPEND_FL)
 874                 stat->attributes |= STATX_ATTR_APPEND;
 875         if (IS_ENCRYPTED(inode))
 876                 stat->attributes |= STATX_ATTR_ENCRYPTED;
 877         if (flags & F2FS_IMMUTABLE_FL)
 878                 stat->attributes |= STATX_ATTR_IMMUTABLE;
 879         if (flags & F2FS_NODUMP_FL)
 880                 stat->attributes |= STATX_ATTR_NODUMP;
 881         if (IS_VERITY(inode))
 882                 stat->attributes |= STATX_ATTR_VERITY;
 883
 884         stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 885                                   STATX_ATTR_APPEND |
 886                                   STATX_ATTR_ENCRYPTED |
 887                                   STATX_ATTR_IMMUTABLE |
 888                                   STATX_ATTR_NODUMP |
 889                                   STATX_ATTR_VERITY);
 890
 891         generic_fillattr(idmap, inode, stat);
 892
 893         /* we need to show initial sectors used for inline_data/dentries */
 894         if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
 895                                         f2fs_has_inline_dentry(inode))
 896                 stat->blocks += (stat->size + 511) >> 9;
 897
 898         return 0;
 899 }
 900
 901 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 902 static void __setattr_copy(struct mnt_idmap *idmap,
 903                            struct inode *inode, const struct iattr *attr)
 904 {
 905         unsigned int ia_valid = attr->ia_valid;
 906
 907         i_uid_update(idmap, attr, inode);
 908         i_gid_update(idmap, attr, inode);
 909         if (ia_valid & ATTR_ATIME)
 910                 inode->i_atime = attr->ia_atime;
 911         if (ia_valid & ATTR_MTIME)
 912                 inode->i_mtime = attr->ia_mtime;
 913         if (ia_valid & ATTR_CTIME)
 914                 inode->i_ctime = attr->ia_ctime;
 915         if (ia_valid & ATTR_MODE) {
 916                 umode_t mode = attr->ia_mode;
 917                 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode);
 918
 919                 if (!vfsgid_in_group_p(vfsgid) &&
 920                     !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID))
 921                         mode &= ~S_ISGID;
 922                 set_acl_inode(inode, mode);
 923         }
 924 }
 925 #else
 926 #define __setattr_copy setattr_copy
 927 #endif
 928
 929 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
 930                  struct iattr *attr)
 931 {
 932         struct inode *inode = d_inode(dentry);
 933         int err;
 934
 935         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
 936                 return -EIO;
 937
 938         if (unlikely(IS_IMMUTABLE(inode)))
 939                 return -EPERM;
 940
 941         if (unlikely(IS_APPEND(inode) &&
 942                         (attr->ia_valid & (ATTR_MODE | ATTR_UID |
 943                                   ATTR_GID | ATTR_TIMES_SET))))
 944                 return -EPERM;
 945
 946         if ((attr->ia_valid & ATTR_SIZE) &&
 947                 !f2fs_is_compress_backend_ready(inode))
 948                 return -EOPNOTSUPP;
 949
 950         err = setattr_prepare(idmap, dentry, attr);
 951         if (err)
 952                 return err;
 953
 954         err = fscrypt_prepare_setattr(dentry, attr);
 955         if (err)
 956                 return err;
 957
 958         err = fsverity_prepare_setattr(dentry, attr);
 959         if (err)
 960                 return err;
 961
 962         if (is_quota_modification(idmap, inode, attr)) {
 963                 err = f2fs_dquot_initialize(inode);
 964                 if (err)
 965                         return err;
 966         }
 967         if (i_uid_needs_update(idmap, attr, inode) ||
 968             i_gid_needs_update(idmap, attr, inode)) {
 969                 f2fs_lock_op(F2FS_I_SB(inode));
 970                 err = dquot_transfer(idmap, inode, attr);
 971                 if (err) {
 972                         set_sbi_flag(F2FS_I_SB(inode),
 973                                         SBI_QUOTA_NEED_REPAIR);
 974                         f2fs_unlock_op(F2FS_I_SB(inode));
 975                         return err;
 976                 }
 977                 /*
 978                  * update uid/gid under lock_op(), so that dquot and inode can
 979                  * be updated atomically.
 980                  */
 981                 i_uid_update(idmap, attr, inode);
 982                 i_gid_update(idmap, attr, inode);
 983                 f2fs_mark_inode_dirty_sync(inode, true);
 984                 f2fs_unlock_op(F2FS_I_SB(inode));
 985         }
 986
 987         if (attr->ia_valid & ATTR_SIZE) {
 988                 loff_t old_size = i_size_read(inode);
 989
 990                 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
 991                         /*
 992                          * should convert inline inode before i_size_write to
 993                          * keep smaller than inline_data size with inline flag.
 994                          */
 995                         err = f2fs_convert_inline_inode(inode);
 996                         if (err)
 997                                 return err;
 998                 }
 999
1000                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1001                 filemap_invalidate_lock(inode->i_mapping);
1002
1003                 truncate_setsize(inode, attr->ia_size);
1004
1005                 if (attr->ia_size <= old_size)
1006                         err = f2fs_truncate(inode);
1007                 /*
1008                  * do not trim all blocks after i_size if target size is
1009                  * larger than i_size.
1010                  */
1011                 filemap_invalidate_unlock(inode->i_mapping);
1012                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1013                 if (err)
1014                         return err;
1015
1016                 spin_lock(&F2FS_I(inode)->i_size_lock);
1017                 inode->i_mtime = inode->i_ctime = current_time(inode);
1018                 F2FS_I(inode)->last_disk_size = i_size_read(inode);
1019                 spin_unlock(&F2FS_I(inode)->i_size_lock);
1020         }
1021
1022         __setattr_copy(idmap, inode, attr);
1023
1024         if (attr->ia_valid & ATTR_MODE) {
1025                 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
1026
1027                 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1028                         if (!err)
1029                                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
1030                         clear_inode_flag(inode, FI_ACL_MODE);
1031                 }
1032         }
1033
1034         /* file size may changed here */
1035         f2fs_mark_inode_dirty_sync(inode, true);
1036
1037         /* inode change will produce dirty node pages flushed by checkpoint */
1038         f2fs_balance_fs(F2FS_I_SB(inode), true);
1039
1040         return err;
1041 }
1042
1043 const struct inode_operations f2fs_file_inode_operations = {
1044         .getattr        = f2fs_getattr,
1045         .setattr        = f2fs_setattr,
1046         .get_inode_acl  = f2fs_get_acl,
1047         .set_acl        = f2fs_set_acl,
1048         .listxattr      = f2fs_listxattr,
1049         .fiemap         = f2fs_fiemap,
1050         .fileattr_get   = f2fs_fileattr_get,
1051         .fileattr_set   = f2fs_fileattr_set,
1052 };
1053
1054 static int fill_zero(struct inode *inode, pgoff_t index,
1055                                         loff_t start, loff_t len)
1056 {
1057         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1058         struct page *page;
1059
1060         if (!len)
1061                 return 0;
1062
1063         f2fs_balance_fs(sbi, true);
1064
1065         f2fs_lock_op(sbi);
1066         page = f2fs_get_new_data_page(inode, NULL, index, false);
1067         f2fs_unlock_op(sbi);
1068
1069         if (IS_ERR(page))
1070                 return PTR_ERR(page);
1071
1072         f2fs_wait_on_page_writeback(page, DATA, true, true);
1073         zero_user(page, start, len);
1074         set_page_dirty(page);
1075         f2fs_put_page(page, 1);
1076         return 0;
1077 }
1078
1079 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1080 {
1081         int err;
1082
1083         while (pg_start < pg_end) {
1084                 struct dnode_of_data dn;
1085                 pgoff_t end_offset, count;
1086
1087                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1088                 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1089                 if (err) {
1090                         if (err == -ENOENT) {
1091                                 pg_start = f2fs_get_next_page_offset(&dn,
1092                                                                 pg_start);
1093                                 continue;
1094                         }
1095                         return err;
1096                 }
1097
1098                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1099                 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1100
1101                 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1102
1103                 f2fs_truncate_data_blocks_range(&dn, count);
1104                 f2fs_put_dnode(&dn);
1105
1106                 pg_start += count;
1107         }
1108         return 0;
1109 }
1110
1111 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
1112 {
1113         pgoff_t pg_start, pg_end;
1114         loff_t off_start, off_end;
1115         int ret;
1116
1117         ret = f2fs_convert_inline_inode(inode);
1118         if (ret)
1119                 return ret;
1120
1121         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1122         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1123
1124         off_start = offset & (PAGE_SIZE - 1);
1125         off_end = (offset + len) & (PAGE_SIZE - 1);
1126
1127         if (pg_start == pg_end) {
1128                 ret = fill_zero(inode, pg_start, off_start,
1129                                                 off_end - off_start);
1130                 if (ret)
1131                         return ret;
1132         } else {
1133                 if (off_start) {
1134                         ret = fill_zero(inode, pg_start++, off_start,
1135                                                 PAGE_SIZE - off_start);
1136                         if (ret)
1137                                 return ret;
1138                 }
1139                 if (off_end) {
1140                         ret = fill_zero(inode, pg_end, 0, off_end);
1141                         if (ret)
1142                                 return ret;
1143                 }
1144
1145                 if (pg_start < pg_end) {
1146                         loff_t blk_start, blk_end;
1147                         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1148
1149                         f2fs_balance_fs(sbi, true);
1150
1151                         blk_start = (loff_t)pg_start << PAGE_SHIFT;
1152                         blk_end = (loff_t)pg_end << PAGE_SHIFT;
1153
1154                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1155                         filemap_invalidate_lock(inode->i_mapping);
1156
1157                         truncate_pagecache_range(inode, blk_start, blk_end - 1);
1158
1159                         f2fs_lock_op(sbi);
1160                         ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1161                         f2fs_unlock_op(sbi);
1162
1163                         filemap_invalidate_unlock(inode->i_mapping);
1164                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1165                 }
1166         }
1167
1168         return ret;
1169 }
1170
1171 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1172                                 int *do_replace, pgoff_t off, pgoff_t len)
1173 {
1174         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1175         struct dnode_of_data dn;
1176         int ret, done, i;
1177
1178 next_dnode:
1179         set_new_dnode(&dn, inode, NULL, NULL, 0);
1180         ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1181         if (ret && ret != -ENOENT) {
1182                 return ret;
1183         } else if (ret == -ENOENT) {
1184                 if (dn.max_level == 0)
1185                         return -ENOENT;
1186                 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1187                                                 dn.ofs_in_node, len);
1188                 blkaddr += done;
1189                 do_replace += done;
1190                 goto next;
1191         }
1192
1193         done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1194                                                         dn.ofs_in_node, len);
1195         for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1196                 *blkaddr = f2fs_data_blkaddr(&dn);
1197
1198                 if (__is_valid_data_blkaddr(*blkaddr) &&
1199                         !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1200                                         DATA_GENERIC_ENHANCE)) {
1201                         f2fs_put_dnode(&dn);
1202                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1203                         return -EFSCORRUPTED;
1204                 }
1205
1206                 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1207
1208                         if (f2fs_lfs_mode(sbi)) {
1209                                 f2fs_put_dnode(&dn);
1210                                 return -EOPNOTSUPP;
1211                         }
1212
1213                         /* do not invalidate this block address */
1214                         f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1215                         *do_replace = 1;
1216                 }
1217         }
1218         f2fs_put_dnode(&dn);
1219 next:
1220         len -= done;
1221         off += done;
1222         if (len)
1223                 goto next_dnode;
1224         return 0;
1225 }
1226
1227 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1228                                 int *do_replace, pgoff_t off, int len)
1229 {
1230         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1231         struct dnode_of_data dn;
1232         int ret, i;
1233
1234         for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1235                 if (*do_replace == 0)
1236                         continue;
1237
1238                 set_new_dnode(&dn, inode, NULL, NULL, 0);
1239                 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1240                 if (ret) {
1241                         dec_valid_block_count(sbi, inode, 1);
1242                         f2fs_invalidate_blocks(sbi, *blkaddr);
1243                 } else {
1244                         f2fs_update_data_blkaddr(&dn, *blkaddr);
1245                 }
1246                 f2fs_put_dnode(&dn);
1247         }
1248         return 0;
1249 }
1250
1251 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1252                         block_t *blkaddr, int *do_replace,
1253                         pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1254 {
1255         struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1256         pgoff_t i = 0;
1257         int ret;
1258
1259         while (i < len) {
1260                 if (blkaddr[i] == NULL_ADDR && !full) {
1261                         i++;
1262                         continue;
1263                 }
1264
1265                 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1266                         struct dnode_of_data dn;
1267                         struct node_info ni;
1268                         size_t new_size;
1269                         pgoff_t ilen;
1270
1271                         set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1272                         ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1273                         if (ret)
1274                                 return ret;
1275
1276                         ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1277                         if (ret) {
1278                                 f2fs_put_dnode(&dn);
1279                                 return ret;
1280                         }
1281
1282                         ilen = min((pgoff_t)
1283                                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1284                                                 dn.ofs_in_node, len - i);
1285                         do {
1286                                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1287                                 f2fs_truncate_data_blocks_range(&dn, 1);
1288
1289                                 if (do_replace[i]) {
1290                                         f2fs_i_blocks_write(src_inode,
1291                                                         1, false, false);
1292                                         f2fs_i_blocks_write(dst_inode,
1293                                                         1, true, false);
1294                                         f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1295                                         blkaddr[i], ni.version, true, false);
1296
1297                                         do_replace[i] = 0;
1298                                 }
1299                                 dn.ofs_in_node++;
1300                                 i++;
1301                                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1302                                 if (dst_inode->i_size < new_size)
1303                                         f2fs_i_size_write(dst_inode, new_size);
1304                         } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1305
1306                         f2fs_put_dnode(&dn);
1307                 } else {
1308                         struct page *psrc, *pdst;
1309
1310                         psrc = f2fs_get_lock_data_page(src_inode,
1311                                                         src + i, true);
1312                         if (IS_ERR(psrc))
1313                                 return PTR_ERR(psrc);
1314                         pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1315                                                                 true);
1316                         if (IS_ERR(pdst)) {
1317                                 f2fs_put_page(psrc, 1);
1318                                 return PTR_ERR(pdst);
1319                         }
1320                         memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1321                         set_page_dirty(pdst);
1322                         f2fs_put_page(pdst, 1);
1323                         f2fs_put_page(psrc, 1);
1324
1325                         ret = f2fs_truncate_hole(src_inode,
1326                                                 src + i, src + i + 1);
1327                         if (ret)
1328                                 return ret;
1329                         i++;
1330                 }
1331         }
1332         return 0;
1333 }
1334
1335 static int __exchange_data_block(struct inode *src_inode,
1336                         struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1337                         pgoff_t len, bool full)
1338 {
1339         block_t *src_blkaddr;
1340         int *do_replace;
1341         pgoff_t olen;
1342         int ret;
1343
1344         while (len) {
1345                 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1346
1347                 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1348                                         array_size(olen, sizeof(block_t)),
1349                                         GFP_NOFS);
1350                 if (!src_blkaddr)
1351                         return -ENOMEM;
1352
1353                 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1354                                         array_size(olen, sizeof(int)),
1355                                         GFP_NOFS);
1356                 if (!do_replace) {
1357                         kvfree(src_blkaddr);
1358                         return -ENOMEM;
1359                 }
1360
1361                 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1362                                         do_replace, src, olen);
1363                 if (ret)
1364                         goto roll_back;
1365
1366                 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1367                                         do_replace, src, dst, olen, full);
1368                 if (ret)
1369                         goto roll_back;
1370
1371                 src += olen;
1372                 dst += olen;
1373                 len -= olen;
1374
1375                 kvfree(src_blkaddr);
1376                 kvfree(do_replace);
1377         }
1378         return 0;
1379
1380 roll_back:
1381         __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1382         kvfree(src_blkaddr);
1383         kvfree(do_replace);
1384         return ret;
1385 }
1386
1387 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1388 {
1389         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1390         pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1391         pgoff_t start = offset >> PAGE_SHIFT;
1392         pgoff_t end = (offset + len) >> PAGE_SHIFT;
1393         int ret;
1394
1395         f2fs_balance_fs(sbi, true);
1396
1397         /* avoid gc operation during block exchange */
1398         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1399         filemap_invalidate_lock(inode->i_mapping);
1400
1401         f2fs_lock_op(sbi);
1402         f2fs_drop_extent_tree(inode);
1403         truncate_pagecache(inode, offset);
1404         ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1405         f2fs_unlock_op(sbi);
1406
1407         filemap_invalidate_unlock(inode->i_mapping);
1408         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1409         return ret;
1410 }
1411
1412 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1413 {
1414         loff_t new_size;
1415         int ret;
1416
1417         if (offset + len >= i_size_read(inode))
1418                 return -EINVAL;
1419
1420         /* collapse range should be aligned to block size of f2fs. */
1421         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1422                 return -EINVAL;
1423
1424         ret = f2fs_convert_inline_inode(inode);
1425         if (ret)
1426                 return ret;
1427
1428         /* write out all dirty pages from offset */
1429         ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1430         if (ret)
1431                 return ret;
1432
1433         ret = f2fs_do_collapse(inode, offset, len);
1434         if (ret)
1435                 return ret;
1436
1437         /* write out all moved pages, if possible */
1438         filemap_invalidate_lock(inode->i_mapping);
1439         filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1440         truncate_pagecache(inode, offset);
1441
1442         new_size = i_size_read(inode) - len;
1443         ret = f2fs_truncate_blocks(inode, new_size, true);
1444         filemap_invalidate_unlock(inode->i_mapping);
1445         if (!ret)
1446                 f2fs_i_size_write(inode, new_size);
1447         return ret;
1448 }
1449
1450 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1451                                                                 pgoff_t end)
1452 {
1453         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1454         pgoff_t index = start;
1455         unsigned int ofs_in_node = dn->ofs_in_node;
1456         blkcnt_t count = 0;
1457         int ret;
1458
1459         for (; index < end; index++, dn->ofs_in_node++) {
1460                 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1461                         count++;
1462         }
1463
1464         dn->ofs_in_node = ofs_in_node;
1465         ret = f2fs_reserve_new_blocks(dn, count);
1466         if (ret)
1467                 return ret;
1468
1469         dn->ofs_in_node = ofs_in_node;
1470         for (index = start; index < end; index++, dn->ofs_in_node++) {
1471                 dn->data_blkaddr = f2fs_data_blkaddr(dn);
1472                 /*
1473                  * f2fs_reserve_new_blocks will not guarantee entire block
1474                  * allocation.
1475                  */
1476                 if (dn->data_blkaddr == NULL_ADDR) {
1477                         ret = -ENOSPC;
1478                         break;
1479                 }
1480
1481                 if (dn->data_blkaddr == NEW_ADDR)
1482                         continue;
1483
1484                 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1485                                         DATA_GENERIC_ENHANCE)) {
1486                         ret = -EFSCORRUPTED;
1487                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
1488                         break;
1489                 }
1490
1491                 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1492                 dn->data_blkaddr = NEW_ADDR;
1493                 f2fs_set_data_blkaddr(dn);
1494         }
1495
1496         f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
1497         f2fs_update_age_extent_cache_range(dn, start, index - start);
1498
1499         return ret;
1500 }
1501
1502 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1503                                                                 int mode)
1504 {
1505         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1506         struct address_space *mapping = inode->i_mapping;
1507         pgoff_t index, pg_start, pg_end;
1508         loff_t new_size = i_size_read(inode);
1509         loff_t off_start, off_end;
1510         int ret = 0;
1511
1512         ret = inode_newsize_ok(inode, (len + offset));
1513         if (ret)
1514                 return ret;
1515
1516         ret = f2fs_convert_inline_inode(inode);
1517         if (ret)
1518                 return ret;
1519
1520         ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1521         if (ret)
1522                 return ret;
1523
1524         pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1525         pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1526
1527         off_start = offset & (PAGE_SIZE - 1);
1528         off_end = (offset + len) & (PAGE_SIZE - 1);
1529
1530         if (pg_start == pg_end) {
1531                 ret = fill_zero(inode, pg_start, off_start,
1532                                                 off_end - off_start);
1533                 if (ret)
1534                         return ret;
1535
1536                 new_size = max_t(loff_t, new_size, offset + len);
1537         } else {
1538                 if (off_start) {
1539                         ret = fill_zero(inode, pg_start++, off_start,
1540                                                 PAGE_SIZE - off_start);
1541                         if (ret)
1542                                 return ret;
1543
1544                         new_size = max_t(loff_t, new_size,
1545                                         (loff_t)pg_start << PAGE_SHIFT);
1546                 }
1547
1548                 for (index = pg_start; index < pg_end;) {
1549                         struct dnode_of_data dn;
1550                         unsigned int end_offset;
1551                         pgoff_t end;
1552
1553                         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1554                         filemap_invalidate_lock(mapping);
1555
1556                         truncate_pagecache_range(inode,
1557                                 (loff_t)index << PAGE_SHIFT,
1558                                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1559
1560                         f2fs_lock_op(sbi);
1561
1562                         set_new_dnode(&dn, inode, NULL, NULL, 0);
1563                         ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1564                         if (ret) {
1565                                 f2fs_unlock_op(sbi);
1566                                 filemap_invalidate_unlock(mapping);
1567                                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1568                                 goto out;
1569                         }
1570
1571                         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1572                         end = min(pg_end, end_offset - dn.ofs_in_node + index);
1573
1574                         ret = f2fs_do_zero_range(&dn, index, end);
1575                         f2fs_put_dnode(&dn);
1576
1577                         f2fs_unlock_op(sbi);
1578                         filemap_invalidate_unlock(mapping);
1579                         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1580
1581                         f2fs_balance_fs(sbi, dn.node_changed);
1582
1583                         if (ret)
1584                                 goto out;
1585
1586                         index = end;
1587                         new_size = max_t(loff_t, new_size,
1588                                         (loff_t)index << PAGE_SHIFT);
1589                 }
1590
1591                 if (off_end) {
1592                         ret = fill_zero(inode, pg_end, 0, off_end);
1593                         if (ret)
1594                                 goto out;
1595
1596                         new_size = max_t(loff_t, new_size, offset + len);
1597                 }
1598         }
1599
1600 out:
1601         if (new_size > i_size_read(inode)) {
1602                 if (mode & FALLOC_FL_KEEP_SIZE)
1603                         file_set_keep_isize(inode);
1604                 else
1605                         f2fs_i_size_write(inode, new_size);
1606         }
1607         return ret;
1608 }
1609
1610 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1611 {
1612         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1613         struct address_space *mapping = inode->i_mapping;
1614         pgoff_t nr, pg_start, pg_end, delta, idx;
1615         loff_t new_size;
1616         int ret = 0;
1617
1618         new_size = i_size_read(inode) + len;
1619         ret = inode_newsize_ok(inode, new_size);
1620         if (ret)
1621                 return ret;
1622
1623         if (offset >= i_size_read(inode))
1624                 return -EINVAL;
1625
1626         /* insert range should be aligned to block size of f2fs. */
1627         if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1628                 return -EINVAL;
1629
1630         ret = f2fs_convert_inline_inode(inode);
1631         if (ret)
1632                 return ret;
1633
1634         f2fs_balance_fs(sbi, true);
1635
1636         filemap_invalidate_lock(mapping);
1637         ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1638         filemap_invalidate_unlock(mapping);
1639         if (ret)
1640                 return ret;
1641
1642         /* write out all dirty pages from offset */
1643         ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1644         if (ret)
1645                 return ret;
1646
1647         pg_start = offset >> PAGE_SHIFT;
1648         pg_end = (offset + len) >> PAGE_SHIFT;
1649         delta = pg_end - pg_start;
1650         idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1651
1652         /* avoid gc operation during block exchange */
1653         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1654         filemap_invalidate_lock(mapping);
1655         truncate_pagecache(inode, offset);
1656
1657         while (!ret && idx > pg_start) {
1658                 nr = idx - pg_start;
1659                 if (nr > delta)
1660                         nr = delta;
1661                 idx -= nr;
1662
1663                 f2fs_lock_op(sbi);
1664                 f2fs_drop_extent_tree(inode);
1665
1666                 ret = __exchange_data_block(inode, inode, idx,
1667                                         idx + delta, nr, false);
1668                 f2fs_unlock_op(sbi);
1669         }
1670         filemap_invalidate_unlock(mapping);
1671         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1672
1673         /* write out all moved pages, if possible */
1674         filemap_invalidate_lock(mapping);
1675         filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1676         truncate_pagecache(inode, offset);
1677         filemap_invalidate_unlock(mapping);
1678
1679         if (!ret)
1680                 f2fs_i_size_write(inode, new_size);
1681         return ret;
1682 }
1683
1684 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
1685                                         loff_t len, int mode)
1686 {
1687         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1688         struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1689                         .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1690                         .m_may_create = true };
1691         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1692                         .init_gc_type = FG_GC,
1693                         .should_migrate_blocks = false,
1694                         .err_gc_skipped = true,
1695                         .nr_free_secs = 0 };
1696         pgoff_t pg_start, pg_end;
1697         loff_t new_size;
1698         loff_t off_end;
1699         block_t expanded = 0;
1700         int err;
1701
1702         err = inode_newsize_ok(inode, (len + offset));
1703         if (err)
1704                 return err;
1705
1706         err = f2fs_convert_inline_inode(inode);
1707         if (err)
1708                 return err;
1709
1710         f2fs_balance_fs(sbi, true);
1711
1712         pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1713         pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1714         off_end = (offset + len) & (PAGE_SIZE - 1);
1715
1716         map.m_lblk = pg_start;
1717         map.m_len = pg_end - pg_start;
1718         if (off_end)
1719                 map.m_len++;
1720
1721         if (!map.m_len)
1722                 return 0;
1723
1724         if (f2fs_is_pinned_file(inode)) {
1725                 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1726                 block_t sec_len = roundup(map.m_len, sec_blks);
1727
1728                 map.m_len = sec_blks;
1729 next_alloc:
1730                 if (has_not_enough_free_secs(sbi, 0,
1731                         GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1732                         f2fs_down_write(&sbi->gc_lock);
1733                         err = f2fs_gc(sbi, &gc_control);
1734                         if (err && err != -ENODATA)
1735                                 goto out_err;
1736                 }
1737
1738                 f2fs_down_write(&sbi->pin_sem);
1739
1740                 f2fs_lock_op(sbi);
1741                 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1742                 f2fs_unlock_op(sbi);
1743
1744                 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1745                 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
1746                 file_dont_truncate(inode);
1747
1748                 f2fs_up_write(&sbi->pin_sem);
1749
1750                 expanded += map.m_len;
1751                 sec_len -= map.m_len;
1752                 map.m_lblk += map.m_len;
1753                 if (!err && sec_len)
1754                         goto next_alloc;
1755
1756                 map.m_len = expanded;
1757         } else {
1758                 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
1759                 expanded = map.m_len;
1760         }
1761 out_err:
1762         if (err) {
1763                 pgoff_t last_off;
1764
1765                 if (!expanded)
1766                         return err;
1767
1768                 last_off = pg_start + expanded - 1;
1769
1770                 /* update new size to the failed position */
1771                 new_size = (last_off == pg_end) ? offset + len :
1772                                         (loff_t)(last_off + 1) << PAGE_SHIFT;
1773         } else {
1774                 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1775         }
1776
1777         if (new_size > i_size_read(inode)) {
1778                 if (mode & FALLOC_FL_KEEP_SIZE)
1779                         file_set_keep_isize(inode);
1780                 else
1781                         f2fs_i_size_write(inode, new_size);
1782         }
1783
1784         return err;
1785 }
1786
1787 static long f2fs_fallocate(struct file *file, int mode,
1788                                 loff_t offset, loff_t len)
1789 {
1790         struct inode *inode = file_inode(file);
1791         long ret = 0;
1792
1793         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1794                 return -EIO;
1795         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1796                 return -ENOSPC;
1797         if (!f2fs_is_compress_backend_ready(inode))
1798                 return -EOPNOTSUPP;
1799
1800         /* f2fs only support ->fallocate for regular file */
1801         if (!S_ISREG(inode->i_mode))
1802                 return -EINVAL;
1803
1804         if (IS_ENCRYPTED(inode) &&
1805                 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1806                 return -EOPNOTSUPP;
1807
1808         /*
1809          * Pinned file should not support partial truncation since the block
1810          * can be used by applications.
1811          */
1812         if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1813                 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1814                         FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1815                 return -EOPNOTSUPP;
1816
1817         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1818                         FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1819                         FALLOC_FL_INSERT_RANGE))
1820                 return -EOPNOTSUPP;
1821
1822         inode_lock(inode);
1823
1824         ret = file_modified(file);
1825         if (ret)
1826                 goto out;
1827
1828         if (mode & FALLOC_FL_PUNCH_HOLE) {
1829                 if (offset >= inode->i_size)
1830                         goto out;
1831
1832                 ret = f2fs_punch_hole(inode, offset, len);
1833         } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1834                 ret = f2fs_collapse_range(inode, offset, len);
1835         } else if (mode & FALLOC_FL_ZERO_RANGE) {
1836                 ret = f2fs_zero_range(inode, offset, len, mode);
1837         } else if (mode & FALLOC_FL_INSERT_RANGE) {
1838                 ret = f2fs_insert_range(inode, offset, len);
1839         } else {
1840                 ret = f2fs_expand_inode_data(inode, offset, len, mode);
1841         }
1842
1843         if (!ret) {
1844                 inode->i_mtime = inode->i_ctime = current_time(inode);
1845                 f2fs_mark_inode_dirty_sync(inode, false);
1846                 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1847         }
1848
1849 out:
1850         inode_unlock(inode);
1851
1852         trace_f2fs_fallocate(inode, mode, offset, len, ret);
1853         return ret;
1854 }
1855
1856 static int f2fs_release_file(struct inode *inode, struct file *filp)
1857 {
1858         /*
1859          * f2fs_release_file is called at every close calls. So we should
1860          * not drop any inmemory pages by close called by other process.
1861          */
1862         if (!(filp->f_mode & FMODE_WRITE) ||
1863                         atomic_read(&inode->i_writecount) != 1)
1864                 return 0;
1865
1866         inode_lock(inode);
1867         f2fs_abort_atomic_write(inode, true);
1868         inode_unlock(inode);
1869
1870         return 0;
1871 }
1872
1873 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1874 {
1875         struct inode *inode = file_inode(file);
1876
1877         /*
1878          * If the process doing a transaction is crashed, we should do
1879          * roll-back. Otherwise, other reader/write can see corrupted database
1880          * until all the writers close its file. Since this should be done
1881          * before dropping file lock, it needs to do in ->flush.
1882          */
1883         if (F2FS_I(inode)->atomic_write_task == current &&
1884                                 (current->flags & PF_EXITING)) {
1885                 inode_lock(inode);
1886                 f2fs_abort_atomic_write(inode, true);
1887                 inode_unlock(inode);
1888         }
1889
1890         return 0;
1891 }
1892
1893 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1894 {
1895         struct f2fs_inode_info *fi = F2FS_I(inode);
1896         u32 masked_flags = fi->i_flags & mask;
1897
1898         /* mask can be shrunk by flags_valid selector */
1899         iflags &= mask;
1900
1901         /* Is it quota file? Do not allow user to mess with it */
1902         if (IS_NOQUOTA(inode))
1903                 return -EPERM;
1904
1905         if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1906                 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1907                         return -EOPNOTSUPP;
1908                 if (!f2fs_empty_dir(inode))
1909                         return -ENOTEMPTY;
1910         }
1911
1912         if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1913                 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1914                         return -EOPNOTSUPP;
1915                 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1916                         return -EINVAL;
1917         }
1918
1919         if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1920                 if (masked_flags & F2FS_COMPR_FL) {
1921                         if (!f2fs_disable_compressed_file(inode))
1922                                 return -EINVAL;
1923                 } else {
1924                         /* try to convert inline_data to support compression */
1925                         int err = f2fs_convert_inline_inode(inode);
1926                         if (err)
1927                                 return err;
1928                         if (!f2fs_may_compress(inode))
1929                                 return -EINVAL;
1930                         if (S_ISREG(inode->i_mode) && F2FS_HAS_BLOCKS(inode))
1931                                 return -EINVAL;
1932                         if (set_compress_context(inode))
1933                                 return -EOPNOTSUPP;
1934                 }
1935         }
1936
1937         fi->i_flags = iflags | (fi->i_flags & ~mask);
1938         f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1939                                         (fi->i_flags & F2FS_NOCOMP_FL));
1940
1941         if (fi->i_flags & F2FS_PROJINHERIT_FL)
1942                 set_inode_flag(inode, FI_PROJ_INHERIT);
1943         else
1944                 clear_inode_flag(inode, FI_PROJ_INHERIT);
1945
1946         inode->i_ctime = current_time(inode);
1947         f2fs_set_inode_flags(inode);
1948         f2fs_mark_inode_dirty_sync(inode, true);
1949         return 0;
1950 }
1951
1952 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1953
1954 /*
1955  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1956  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1957  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1958  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1959  *
1960  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1961  * FS_IOC_FSSETXATTR is done by the VFS.
1962  */
1963
1964 static const struct {
1965         u32 iflag;
1966         u32 fsflag;
1967 } f2fs_fsflags_map[] = {
1968         { F2FS_COMPR_FL,        FS_COMPR_FL },
1969         { F2FS_SYNC_FL,         FS_SYNC_FL },
1970         { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1971         { F2FS_APPEND_FL,       FS_APPEND_FL },
1972         { F2FS_NODUMP_FL,       FS_NODUMP_FL },
1973         { F2FS_NOATIME_FL,      FS_NOATIME_FL },
1974         { F2FS_NOCOMP_FL,       FS_NOCOMP_FL },
1975         { F2FS_INDEX_FL,        FS_INDEX_FL },
1976         { F2FS_DIRSYNC_FL,      FS_DIRSYNC_FL },
1977         { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1978         { F2FS_CASEFOLD_FL,     FS_CASEFOLD_FL },
1979 };
1980
1981 #define F2FS_GETTABLE_FS_FL (           \
1982                 FS_COMPR_FL |           \
1983                 FS_SYNC_FL |            \
1984                 FS_IMMUTABLE_FL |       \
1985                 FS_APPEND_FL |          \
1986                 FS_NODUMP_FL |          \
1987                 FS_NOATIME_FL |         \
1988                 FS_NOCOMP_FL |          \
1989                 FS_INDEX_FL |           \
1990                 FS_DIRSYNC_FL |         \
1991                 FS_PROJINHERIT_FL |     \
1992                 FS_ENCRYPT_FL |         \
1993                 FS_INLINE_DATA_FL |     \
1994                 FS_NOCOW_FL |           \
1995                 FS_VERITY_FL |          \
1996                 FS_CASEFOLD_FL)
1997
1998 #define F2FS_SETTABLE_FS_FL (           \
1999                 FS_COMPR_FL |           \
2000                 FS_SYNC_FL |            \
2001                 FS_IMMUTABLE_FL |       \
2002                 FS_APPEND_FL |          \
2003                 FS_NODUMP_FL |          \
2004                 FS_NOATIME_FL |         \
2005                 FS_NOCOMP_FL |          \
2006                 FS_DIRSYNC_FL |         \
2007                 FS_PROJINHERIT_FL |     \
2008                 FS_CASEFOLD_FL)
2009
2010 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2011 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2012 {
2013         u32 fsflags = 0;
2014         int i;
2015
2016         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2017                 if (iflags & f2fs_fsflags_map[i].iflag)
2018                         fsflags |= f2fs_fsflags_map[i].fsflag;
2019
2020         return fsflags;
2021 }
2022
2023 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2024 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2025 {
2026         u32 iflags = 0;
2027         int i;
2028
2029         for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2030                 if (fsflags & f2fs_fsflags_map[i].fsflag)
2031                         iflags |= f2fs_fsflags_map[i].iflag;
2032
2033         return iflags;
2034 }
2035
2036 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2037 {
2038         struct inode *inode = file_inode(filp);
2039
2040         return put_user(inode->i_generation, (int __user *)arg);
2041 }
2042
2043 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
2044 {
2045         struct inode *inode = file_inode(filp);
2046         struct mnt_idmap *idmap = file_mnt_idmap(filp);
2047         struct f2fs_inode_info *fi = F2FS_I(inode);
2048         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2049         struct inode *pinode;
2050         loff_t isize;
2051         int ret;
2052
2053         if (!inode_owner_or_capable(idmap, inode))
2054                 return -EACCES;
2055
2056         if (!S_ISREG(inode->i_mode))
2057                 return -EINVAL;
2058
2059         if (filp->f_flags & O_DIRECT)
2060                 return -EINVAL;
2061
2062         ret = mnt_want_write_file(filp);
2063         if (ret)
2064                 return ret;
2065
2066         inode_lock(inode);
2067
2068         if (!f2fs_disable_compressed_file(inode)) {
2069                 ret = -EINVAL;
2070                 goto out;
2071         }
2072
2073         if (f2fs_is_atomic_file(inode))
2074                 goto out;
2075
2076         ret = f2fs_convert_inline_inode(inode);
2077         if (ret)
2078                 goto out;
2079
2080         f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2081
2082         /*
2083          * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2084          * f2fs_is_atomic_file.
2085          */
2086         if (get_dirty_pages(inode))
2087                 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2088                           inode->i_ino, get_dirty_pages(inode));
2089         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2090         if (ret) {
2091                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2092                 goto out;
2093         }
2094
2095         /* Check if the inode already has a COW inode */
2096         if (fi->cow_inode == NULL) {
2097                 /* Create a COW inode for atomic write */
2098                 pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2099                 if (IS_ERR(pinode)) {
2100                         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2101                         ret = PTR_ERR(pinode);
2102                         goto out;
2103                 }
2104
2105                 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode);
2106                 iput(pinode);
2107                 if (ret) {
2108                         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2109                         goto out;
2110                 }
2111
2112                 set_inode_flag(fi->cow_inode, FI_COW_FILE);
2113                 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2114         } else {
2115                 /* Reuse the already created COW inode */
2116                 f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
2117         }
2118
2119         f2fs_write_inode(inode, NULL);
2120
2121         stat_inc_atomic_inode(inode);
2122
2123         set_inode_flag(inode, FI_ATOMIC_FILE);
2124
2125         isize = i_size_read(inode);
2126         fi->original_i_size = isize;
2127         if (truncate) {
2128                 set_inode_flag(inode, FI_ATOMIC_REPLACE);
2129                 truncate_inode_pages_final(inode->i_mapping);
2130                 f2fs_i_size_write(inode, 0);
2131                 isize = 0;
2132         }
2133         f2fs_i_size_write(fi->cow_inode, isize);
2134
2135         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2136
2137         f2fs_update_time(sbi, REQ_TIME);
2138         fi->atomic_write_task = current;
2139         stat_update_max_atomic_write(inode);
2140         fi->atomic_write_cnt = 0;
2141 out:
2142         inode_unlock(inode);
2143         mnt_drop_write_file(filp);
2144         return ret;
2145 }
2146
2147 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2148 {
2149         struct inode *inode = file_inode(filp);
2150         struct mnt_idmap *idmap = file_mnt_idmap(filp);
2151         int ret;
2152
2153         if (!inode_owner_or_capable(idmap, inode))
2154                 return -EACCES;
2155
2156         ret = mnt_want_write_file(filp);
2157         if (ret)
2158                 return ret;
2159
2160         f2fs_balance_fs(F2FS_I_SB(inode), true);
2161
2162         inode_lock(inode);
2163
2164         if (f2fs_is_atomic_file(inode)) {
2165                 ret = f2fs_commit_atomic_write(inode);
2166                 if (!ret)
2167                         ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2168
2169                 f2fs_abort_atomic_write(inode, ret);
2170         } else {
2171                 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2172         }
2173
2174         inode_unlock(inode);
2175         mnt_drop_write_file(filp);
2176         return ret;
2177 }
2178
2179 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2180 {
2181         struct inode *inode = file_inode(filp);
2182         struct mnt_idmap *idmap = file_mnt_idmap(filp);
2183         int ret;
2184
2185         if (!inode_owner_or_capable(idmap, inode))
2186                 return -EACCES;
2187
2188         ret = mnt_want_write_file(filp);
2189         if (ret)
2190                 return ret;
2191
2192         inode_lock(inode);
2193
2194         f2fs_abort_atomic_write(inode, true);
2195
2196         inode_unlock(inode);
2197
2198         mnt_drop_write_file(filp);
2199         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2200         return ret;
2201 }
2202
2203 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2204 {
2205         struct inode *inode = file_inode(filp);
2206         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2207         struct super_block *sb = sbi->sb;
2208         __u32 in;
2209         int ret = 0;
2210
2211         if (!capable(CAP_SYS_ADMIN))
2212                 return -EPERM;
2213
2214         if (get_user(in, (__u32 __user *)arg))
2215                 return -EFAULT;
2216
2217         if (in != F2FS_GOING_DOWN_FULLSYNC) {
2218                 ret = mnt_want_write_file(filp);
2219                 if (ret) {
2220                         if (ret == -EROFS) {
2221                                 ret = 0;
2222                                 f2fs_stop_checkpoint(sbi, false,
2223                                                 STOP_CP_REASON_SHUTDOWN);
2224                                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2225                                 trace_f2fs_shutdown(sbi, in, ret);
2226                         }
2227                         return ret;
2228                 }
2229         }
2230
2231         switch (in) {
2232         case F2FS_GOING_DOWN_FULLSYNC:
2233                 ret = freeze_bdev(sb->s_bdev);
2234                 if (ret)
2235                         goto out;
2236                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2237                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2238                 thaw_bdev(sb->s_bdev);
2239                 break;
2240         case F2FS_GOING_DOWN_METASYNC:
2241                 /* do checkpoint only */
2242                 ret = f2fs_sync_fs(sb, 1);
2243                 if (ret)
2244                         goto out;
2245                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2246                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2247                 break;
2248         case F2FS_GOING_DOWN_NOSYNC:
2249                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2250                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2251                 break;
2252         case F2FS_GOING_DOWN_METAFLUSH:
2253                 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2254                 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
2255                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2256                 break;
2257         case F2FS_GOING_DOWN_NEED_FSCK:
2258                 set_sbi_flag(sbi, SBI_NEED_FSCK);
2259                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2260                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2261                 /* do checkpoint only */
2262                 ret = f2fs_sync_fs(sb, 1);
2263                 goto out;
2264         default:
2265                 ret = -EINVAL;
2266                 goto out;
2267         }
2268
2269         f2fs_stop_gc_thread(sbi);
2270         f2fs_stop_discard_thread(sbi);
2271
2272         f2fs_drop_discard_cmd(sbi);
2273         clear_opt(sbi, DISCARD);
2274
2275         f2fs_update_time(sbi, REQ_TIME);
2276 out:
2277         if (in != F2FS_GOING_DOWN_FULLSYNC)
2278                 mnt_drop_write_file(filp);
2279
2280         trace_f2fs_shutdown(sbi, in, ret);
2281
2282         return ret;
2283 }
2284
2285 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2286 {
2287         struct inode *inode = file_inode(filp);
2288         struct super_block *sb = inode->i_sb;
2289         struct fstrim_range range;
2290         int ret;
2291
2292         if (!capable(CAP_SYS_ADMIN))
2293                 return -EPERM;
2294
2295         if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2296                 return -EOPNOTSUPP;
2297
2298         if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2299                                 sizeof(range)))
2300                 return -EFAULT;
2301
2302         ret = mnt_want_write_file(filp);
2303         if (ret)
2304                 return ret;
2305
2306         range.minlen = max((unsigned int)range.minlen,
2307                            bdev_discard_granularity(sb->s_bdev));
2308         ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2309         mnt_drop_write_file(filp);
2310         if (ret < 0)
2311                 return ret;
2312
2313         if (copy_to_user((struct fstrim_range __user *)arg, &range,
2314                                 sizeof(range)))
2315                 return -EFAULT;
2316         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2317         return 0;
2318 }
2319
2320 static bool uuid_is_nonzero(__u8 u[16])
2321 {
2322         int i;
2323
2324         for (i = 0; i < 16; i++)
2325                 if (u[i])
2326                         return true;
2327         return false;
2328 }
2329
2330 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2331 {
2332         struct inode *inode = file_inode(filp);
2333
2334         if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2335                 return -EOPNOTSUPP;
2336
2337         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2338
2339         return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2340 }
2341
2342 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2343 {
2344         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2345                 return -EOPNOTSUPP;
2346         return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2347 }
2348
2349 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2350 {
2351         struct inode *inode = file_inode(filp);
2352         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2353         u8 encrypt_pw_salt[16];
2354         int err;
2355
2356         if (!f2fs_sb_has_encrypt(sbi))
2357                 return -EOPNOTSUPP;
2358
2359         err = mnt_want_write_file(filp);
2360         if (err)
2361                 return err;
2362
2363         f2fs_down_write(&sbi->sb_lock);
2364
2365         if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2366                 goto got_it;
2367
2368         /* update superblock with uuid */
2369         generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2370
2371         err = f2fs_commit_super(sbi, false);
2372         if (err) {
2373                 /* undo new data */
2374                 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2375                 goto out_err;
2376         }
2377 got_it:
2378         memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
2379 out_err:
2380         f2fs_up_write(&sbi->sb_lock);
2381         mnt_drop_write_file(filp);
2382
2383         if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2384                 err = -EFAULT;
2385
2386         return err;
2387 }
2388
2389 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2390                                              unsigned long arg)
2391 {
2392         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2393                 return -EOPNOTSUPP;
2394
2395         return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2396 }
2397
2398 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2399 {
2400         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2401                 return -EOPNOTSUPP;
2402
2403         return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2404 }
2405
2406 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2407 {
2408         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2409                 return -EOPNOTSUPP;
2410
2411         return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2412 }
2413
2414 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2415                                                     unsigned long arg)
2416 {
2417         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2418                 return -EOPNOTSUPP;
2419
2420         return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2421 }
2422
2423 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2424                                               unsigned long arg)
2425 {
2426         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2427                 return -EOPNOTSUPP;
2428
2429         return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2430 }
2431
2432 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2433 {
2434         if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2435                 return -EOPNOTSUPP;
2436
2437         return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2438 }
2439
2440 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2441 {
2442         struct inode *inode = file_inode(filp);
2443         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2444         struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2445                         .no_bg_gc = false,
2446                         .should_migrate_blocks = false,
2447                         .nr_free_secs = 0 };
2448         __u32 sync;
2449         int ret;
2450
2451         if (!capable(CAP_SYS_ADMIN))
2452                 return -EPERM;
2453
2454         if (get_user(sync, (__u32 __user *)arg))
2455                 return -EFAULT;
2456
2457         if (f2fs_readonly(sbi->sb))
2458                 return -EROFS;
2459
2460         ret = mnt_want_write_file(filp);
2461         if (ret)
2462                 return ret;
2463
2464         if (!sync) {
2465                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2466                         ret = -EBUSY;
2467                         goto out;
2468                 }
2469         } else {
2470                 f2fs_down_write(&sbi->gc_lock);
2471         }
2472
2473         gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2474         gc_control.err_gc_skipped = sync;
2475         ret = f2fs_gc(sbi, &gc_control);
2476 out:
2477         mnt_drop_write_file(filp);
2478         return ret;
2479 }
2480
2481 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2482 {
2483         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2484         struct f2fs_gc_control gc_control = {
2485                         .init_gc_type = range->sync ? FG_GC : BG_GC,
2486                         .no_bg_gc = false,
2487                         .should_migrate_blocks = false,
2488                         .err_gc_skipped = range->sync,
2489                         .nr_free_secs = 0 };
2490         u64 end;
2491         int ret;
2492
2493         if (!capable(CAP_SYS_ADMIN))
2494                 return -EPERM;
2495         if (f2fs_readonly(sbi->sb))
2496                 return -EROFS;
2497
2498         end = range->start + range->len;
2499         if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2500                                         end >= MAX_BLKADDR(sbi))
2501                 return -EINVAL;
2502
2503         ret = mnt_want_write_file(filp);
2504         if (ret)
2505                 return ret;
2506
2507 do_more:
2508         if (!range->sync) {
2509                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2510                         ret = -EBUSY;
2511                         goto out;
2512                 }
2513         } else {
2514                 f2fs_down_write(&sbi->gc_lock);
2515         }
2516
2517         gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2518         ret = f2fs_gc(sbi, &gc_control);
2519         if (ret) {
2520                 if (ret == -EBUSY)
2521                         ret = -EAGAIN;
2522                 goto out;
2523         }
2524         range->start += CAP_BLKS_PER_SEC(sbi);
2525         if (range->start <= end)
2526                 goto do_more;
2527 out:
2528         mnt_drop_write_file(filp);
2529         return ret;
2530 }
2531
2532 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2533 {
2534         struct f2fs_gc_range range;
2535
2536         if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2537                                                         sizeof(range)))
2538                 return -EFAULT;
2539         return __f2fs_ioc_gc_range(filp, &range);
2540 }
2541
2542 static int f2fs_ioc_write_checkpoint(struct file *filp)
2543 {
2544         struct inode *inode = file_inode(filp);
2545         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2546         int ret;
2547
2548         if (!capable(CAP_SYS_ADMIN))
2549                 return -EPERM;
2550
2551         if (f2fs_readonly(sbi->sb))
2552                 return -EROFS;
2553
2554         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2555                 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2556                 return -EINVAL;
2557         }
2558
2559         ret = mnt_want_write_file(filp);
2560         if (ret)
2561                 return ret;
2562
2563         ret = f2fs_sync_fs(sbi->sb, 1);
2564
2565         mnt_drop_write_file(filp);
2566         return ret;
2567 }
2568
2569 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2570                                         struct file *filp,
2571                                         struct f2fs_defragment *range)
2572 {
2573         struct inode *inode = file_inode(filp);
2574         struct f2fs_map_blocks map = { .m_next_extent = NULL,
2575                                         .m_seg_type = NO_CHECK_TYPE,
2576                                         .m_may_create = false };
2577         struct extent_info ei = {};
2578         pgoff_t pg_start, pg_end, next_pgofs;
2579         unsigned int blk_per_seg = sbi->blocks_per_seg;
2580         unsigned int total = 0, sec_num;
2581         block_t blk_end = 0;
2582         bool fragmented = false;
2583         int err;
2584
2585         pg_start = range->start >> PAGE_SHIFT;
2586         pg_end = (range->start + range->len) >> PAGE_SHIFT;
2587
2588         f2fs_balance_fs(sbi, true);
2589
2590         inode_lock(inode);
2591
2592         /* if in-place-update policy is enabled, don't waste time here */
2593         set_inode_flag(inode, FI_OPU_WRITE);
2594         if (f2fs_should_update_inplace(inode, NULL)) {
2595                 err = -EINVAL;
2596                 goto out;
2597         }
2598
2599         /* writeback all dirty pages in the range */
2600         err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2601                                                 range->start + range->len - 1);
2602         if (err)
2603                 goto out;
2604
2605         /*
2606          * lookup mapping info in extent cache, skip defragmenting if physical
2607          * block addresses are continuous.
2608          */
2609         if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
2610                 if (ei.fofs + ei.len >= pg_end)
2611                         goto out;
2612         }
2613
2614         map.m_lblk = pg_start;
2615         map.m_next_pgofs = &next_pgofs;
2616
2617         /*
2618          * lookup mapping info in dnode page cache, skip defragmenting if all
2619          * physical block addresses are continuous even if there are hole(s)
2620          * in logical blocks.
2621          */
2622         while (map.m_lblk < pg_end) {
2623                 map.m_len = pg_end - map.m_lblk;
2624                 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2625                 if (err)
2626                         goto out;
2627
2628                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2629                         map.m_lblk = next_pgofs;
2630                         continue;
2631                 }
2632
2633                 if (blk_end && blk_end != map.m_pblk)
2634                         fragmented = true;
2635
2636                 /* record total count of block that we're going to move */
2637                 total += map.m_len;
2638
2639                 blk_end = map.m_pblk + map.m_len;
2640
2641                 map.m_lblk += map.m_len;
2642         }
2643
2644         if (!fragmented) {
2645                 total = 0;
2646                 goto out;
2647         }
2648
2649         sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2650
2651         /*
2652          * make sure there are enough free section for LFS allocation, this can
2653          * avoid defragment running in SSR mode when free section are allocated
2654          * intensively
2655          */
2656         if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2657                 err = -EAGAIN;
2658                 goto out;
2659         }
2660
2661         map.m_lblk = pg_start;
2662         map.m_len = pg_end - pg_start;
2663         total = 0;
2664
2665         while (map.m_lblk < pg_end) {
2666                 pgoff_t idx;
2667                 int cnt = 0;
2668
2669 do_map:
2670                 map.m_len = pg_end - map.m_lblk;
2671                 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
2672                 if (err)
2673                         goto clear_out;
2674
2675                 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2676                         map.m_lblk = next_pgofs;
2677                         goto check;
2678                 }
2679
2680                 set_inode_flag(inode, FI_SKIP_WRITES);
2681
2682                 idx = map.m_lblk;
2683                 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2684                         struct page *page;
2685
2686                         page = f2fs_get_lock_data_page(inode, idx, true);
2687                         if (IS_ERR(page)) {
2688                                 err = PTR_ERR(page);
2689                                 goto clear_out;
2690                         }
2691
2692                         set_page_dirty(page);
2693                         set_page_private_gcing(page);
2694                         f2fs_put_page(page, 1);
2695
2696                         idx++;
2697                         cnt++;
2698                         total++;
2699                 }
2700
2701                 map.m_lblk = idx;
2702 check:
2703                 if (map.m_lblk < pg_end && cnt < blk_per_seg)
2704                         goto do_map;
2705
2706                 clear_inode_flag(inode, FI_SKIP_WRITES);
2707
2708                 err = filemap_fdatawrite(inode->i_mapping);
2709                 if (err)
2710                         goto out;
2711         }
2712 clear_out:
2713         clear_inode_flag(inode, FI_SKIP_WRITES);
2714 out:
2715         clear_inode_flag(inode, FI_OPU_WRITE);
2716         inode_unlock(inode);
2717         if (!err)
2718                 range->len = (u64)total << PAGE_SHIFT;
2719         return err;
2720 }
2721
2722 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2723 {
2724         struct inode *inode = file_inode(filp);
2725         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2726         struct f2fs_defragment range;
2727         int err;
2728
2729         if (!capable(CAP_SYS_ADMIN))
2730                 return -EPERM;
2731
2732         if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2733                 return -EINVAL;
2734
2735         if (f2fs_readonly(sbi->sb))
2736                 return -EROFS;
2737
2738         if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2739                                                         sizeof(range)))
2740                 return -EFAULT;
2741
2742         /* verify alignment of offset & size */
2743         if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2744                 return -EINVAL;
2745
2746         if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2747                                         max_file_blocks(inode)))
2748                 return -EINVAL;
2749
2750         err = mnt_want_write_file(filp);
2751         if (err)
2752                 return err;
2753
2754         err = f2fs_defragment_range(sbi, filp, &range);
2755         mnt_drop_write_file(filp);
2756
2757         f2fs_update_time(sbi, REQ_TIME);
2758         if (err < 0)
2759                 return err;
2760
2761         if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2762                                                         sizeof(range)))
2763                 return -EFAULT;
2764
2765         return 0;
2766 }
2767
2768 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2769                         struct file *file_out, loff_t pos_out, size_t len)
2770 {
2771         struct inode *src = file_inode(file_in);
2772         struct inode *dst = file_inode(file_out);
2773         struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2774         size_t olen = len, dst_max_i_size = 0;
2775         size_t dst_osize;
2776         int ret;
2777
2778         if (file_in->f_path.mnt != file_out->f_path.mnt ||
2779                                 src->i_sb != dst->i_sb)
2780                 return -EXDEV;
2781
2782         if (unlikely(f2fs_readonly(src->i_sb)))
2783                 return -EROFS;
2784
2785         if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2786                 return -EINVAL;
2787
2788         if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2789                 return -EOPNOTSUPP;
2790
2791         if (pos_out < 0 || pos_in < 0)
2792                 return -EINVAL;
2793
2794         if (src == dst) {
2795                 if (pos_in == pos_out)
2796                         return 0;
2797                 if (pos_out > pos_in && pos_out < pos_in + len)
2798                         return -EINVAL;
2799         }
2800
2801         inode_lock(src);
2802         if (src != dst) {
2803                 ret = -EBUSY;
2804                 if (!inode_trylock(dst))
2805                         goto out;
2806         }
2807
2808         ret = -EINVAL;
2809         if (pos_in + len > src->i_size || pos_in + len < pos_in)
2810                 goto out_unlock;
2811         if (len == 0)
2812                 olen = len = src->i_size - pos_in;
2813         if (pos_in + len == src->i_size)
2814                 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2815         if (len == 0) {
2816                 ret = 0;
2817                 goto out_unlock;
2818         }
2819
2820         dst_osize = dst->i_size;
2821         if (pos_out + olen > dst->i_size)
2822                 dst_max_i_size = pos_out + olen;
2823
2824         /* verify the end result is block aligned */
2825         if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2826                         !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2827                         !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2828                 goto out_unlock;
2829
2830         ret = f2fs_convert_inline_inode(src);
2831         if (ret)
2832                 goto out_unlock;
2833
2834         ret = f2fs_convert_inline_inode(dst);
2835         if (ret)
2836                 goto out_unlock;
2837
2838         /* write out all dirty pages from offset */
2839         ret = filemap_write_and_wait_range(src->i_mapping,
2840                                         pos_in, pos_in + len);
2841         if (ret)
2842                 goto out_unlock;
2843
2844         ret = filemap_write_and_wait_range(dst->i_mapping,
2845                                         pos_out, pos_out + len);
2846         if (ret)
2847                 goto out_unlock;
2848
2849         f2fs_balance_fs(sbi, true);
2850
2851         f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2852         if (src != dst) {
2853                 ret = -EBUSY;
2854                 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2855                         goto out_src;
2856         }
2857
2858         f2fs_lock_op(sbi);
2859         ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2860                                 pos_out >> F2FS_BLKSIZE_BITS,
2861                                 len >> F2FS_BLKSIZE_BITS, false);
2862
2863         if (!ret) {
2864                 if (dst_max_i_size)
2865                         f2fs_i_size_write(dst, dst_max_i_size);
2866                 else if (dst_osize != dst->i_size)
2867                         f2fs_i_size_write(dst, dst_osize);
2868         }
2869         f2fs_unlock_op(sbi);
2870
2871         if (src != dst)
2872                 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2873 out_src:
2874         f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2875 out_unlock:
2876         if (src != dst)
2877                 inode_unlock(dst);
2878 out:
2879         inode_unlock(src);
2880         return ret;
2881 }
2882
2883 static int __f2fs_ioc_move_range(struct file *filp,
2884                                 struct f2fs_move_range *range)
2885 {
2886         struct fd dst;
2887         int err;
2888
2889         if (!(filp->f_mode & FMODE_READ) ||
2890                         !(filp->f_mode & FMODE_WRITE))
2891                 return -EBADF;
2892
2893         dst = fdget(range->dst_fd);
2894         if (!dst.file)
2895                 return -EBADF;
2896
2897         if (!(dst.file->f_mode & FMODE_WRITE)) {
2898                 err = -EBADF;
2899                 goto err_out;
2900         }
2901
2902         err = mnt_want_write_file(filp);
2903         if (err)
2904                 goto err_out;
2905
2906         err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2907                                         range->pos_out, range->len);
2908
2909         mnt_drop_write_file(filp);
2910 err_out:
2911         fdput(dst);
2912         return err;
2913 }
2914
2915 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2916 {
2917         struct f2fs_move_range range;
2918
2919         if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2920                                                         sizeof(range)))
2921                 return -EFAULT;
2922         return __f2fs_ioc_move_range(filp, &range);
2923 }
2924
2925 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2926 {
2927         struct inode *inode = file_inode(filp);
2928         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2929         struct sit_info *sm = SIT_I(sbi);
2930         unsigned int start_segno = 0, end_segno = 0;
2931         unsigned int dev_start_segno = 0, dev_end_segno = 0;
2932         struct f2fs_flush_device range;
2933         struct f2fs_gc_control gc_control = {
2934                         .init_gc_type = FG_GC,
2935                         .should_migrate_blocks = true,
2936                         .err_gc_skipped = true,
2937                         .nr_free_secs = 0 };
2938         int ret;
2939
2940         if (!capable(CAP_SYS_ADMIN))
2941                 return -EPERM;
2942
2943         if (f2fs_readonly(sbi->sb))
2944                 return -EROFS;
2945
2946         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2947                 return -EINVAL;
2948
2949         if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2950                                                         sizeof(range)))
2951                 return -EFAULT;
2952
2953         if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2954                         __is_large_section(sbi)) {
2955                 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2956                           range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2957                 return -EINVAL;
2958         }
2959
2960         ret = mnt_want_write_file(filp);
2961         if (ret)
2962                 return ret;
2963
2964         if (range.dev_num != 0)
2965                 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2966         dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2967
2968         start_segno = sm->last_victim[FLUSH_DEVICE];
2969         if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2970                 start_segno = dev_start_segno;
2971         end_segno = min(start_segno + range.segments, dev_end_segno);
2972
2973         while (start_segno < end_segno) {
2974                 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2975                         ret = -EBUSY;
2976                         goto out;
2977                 }
2978                 sm->last_victim[GC_CB] = end_segno + 1;
2979                 sm->last_victim[GC_GREEDY] = end_segno + 1;
2980                 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2981
2982                 gc_control.victim_segno = start_segno;
2983                 ret = f2fs_gc(sbi, &gc_control);
2984                 if (ret == -EAGAIN)
2985                         ret = 0;
2986                 else if (ret < 0)
2987                         break;
2988                 start_segno++;
2989         }
2990 out:
2991         mnt_drop_write_file(filp);
2992         return ret;
2993 }
2994
2995 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2996 {
2997         struct inode *inode = file_inode(filp);
2998         u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2999
3000         /* Must validate to set it with SQLite behavior in Android. */
3001         sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3002
3003         return put_user(sb_feature, (u32 __user *)arg);
3004 }
3005
3006 #ifdef CONFIG_QUOTA
3007 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3008 {
3009         struct dquot *transfer_to[MAXQUOTAS] = {};
3010         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3011         struct super_block *sb = sbi->sb;
3012         int err = 0;
3013
3014         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
3015         if (!IS_ERR(transfer_to[PRJQUOTA])) {
3016                 err = __dquot_transfer(inode, transfer_to);
3017                 if (err)
3018                         set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3019                 dqput(transfer_to[PRJQUOTA]);
3020         }
3021         return err;
3022 }
3023
3024 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3025 {
3026         struct f2fs_inode_info *fi = F2FS_I(inode);
3027         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3028         struct f2fs_inode *ri = NULL;
3029         kprojid_t kprojid;
3030         int err;
3031
3032         if (!f2fs_sb_has_project_quota(sbi)) {
3033                 if (projid != F2FS_DEF_PROJID)
3034                         return -EOPNOTSUPP;
3035                 else
3036                         return 0;
3037         }
3038
3039         if (!f2fs_has_extra_attr(inode))
3040                 return -EOPNOTSUPP;
3041
3042         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3043
3044         if (projid_eq(kprojid, fi->i_projid))
3045                 return 0;
3046
3047         err = -EPERM;
3048         /* Is it quota file? Do not allow user to mess with it */
3049         if (IS_NOQUOTA(inode))
3050                 return err;
3051
3052         if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3053                 return -EOVERFLOW;
3054
3055         err = f2fs_dquot_initialize(inode);
3056         if (err)
3057                 return err;
3058
3059         f2fs_lock_op(sbi);
3060         err = f2fs_transfer_project_quota(inode, kprojid);
3061         if (err)
3062                 goto out_unlock;
3063
3064         fi->i_projid = kprojid;
3065         inode->i_ctime = current_time(inode);
3066         f2fs_mark_inode_dirty_sync(inode, true);
3067 out_unlock:
3068         f2fs_unlock_op(sbi);
3069         return err;
3070 }
3071 #else
3072 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3073 {
3074         return 0;
3075 }
3076
3077 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
3078 {
3079         if (projid != F2FS_DEF_PROJID)
3080                 return -EOPNOTSUPP;
3081         return 0;
3082 }
3083 #endif
3084
3085 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3086 {
3087         struct inode *inode = d_inode(dentry);
3088         struct f2fs_inode_info *fi = F2FS_I(inode);
3089         u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3090
3091         if (IS_ENCRYPTED(inode))
3092                 fsflags |= FS_ENCRYPT_FL;
3093         if (IS_VERITY(inode))
3094                 fsflags |= FS_VERITY_FL;
3095         if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3096                 fsflags |= FS_INLINE_DATA_FL;
3097         if (is_inode_flag_set(inode, FI_PIN_FILE))
3098                 fsflags |= FS_NOCOW_FL;
3099
3100         fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3101
3102         if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3103                 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3104
3105         return 0;
3106 }
3107
3108 int f2fs_fileattr_set(struct mnt_idmap *idmap,
3109                       struct dentry *dentry, struct fileattr *fa)
3110 {
3111         struct inode *inode = d_inode(dentry);
3112         u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3113         u32 iflags;
3114         int err;
3115
3116         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3117                 return -EIO;
3118         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3119                 return -ENOSPC;
3120         if (fsflags & ~F2FS_GETTABLE_FS_FL)
3121                 return -EOPNOTSUPP;
3122         fsflags &= F2FS_SETTABLE_FS_FL;
3123         if (!fa->flags_valid)
3124                 mask &= FS_COMMON_FL;
3125
3126         iflags = f2fs_fsflags_to_iflags(fsflags);
3127         if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3128                 return -EOPNOTSUPP;
3129
3130         err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3131         if (!err)
3132                 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3133
3134         return err;
3135 }
3136
3137 int f2fs_pin_file_control(struct inode *inode, bool inc)
3138 {
3139         struct f2fs_inode_info *fi = F2FS_I(inode);
3140         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3141
3142         /* Use i_gc_failures for normal file as a risk signal. */
3143         if (inc)
3144                 f2fs_i_gc_failures_write(inode,
3145                                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3146
3147         if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3148                 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3149                           __func__, inode->i_ino,
3150                           fi->i_gc_failures[GC_FAILURE_PIN]);
3151                 clear_inode_flag(inode, FI_PIN_FILE);
3152                 return -EAGAIN;
3153         }
3154         return 0;
3155 }
3156
3157 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3158 {
3159         struct inode *inode = file_inode(filp);
3160         __u32 pin;
3161         int ret = 0;
3162
3163         if (get_user(pin, (__u32 __user *)arg))
3164                 return -EFAULT;
3165
3166         if (!S_ISREG(inode->i_mode))
3167                 return -EINVAL;
3168
3169         if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3170                 return -EROFS;
3171
3172         ret = mnt_want_write_file(filp);
3173         if (ret)
3174                 return ret;
3175
3176         inode_lock(inode);
3177
3178         if (!pin) {
3179                 clear_inode_flag(inode, FI_PIN_FILE);
3180                 f2fs_i_gc_failures_write(inode, 0);
3181                 goto done;
3182         }
3183
3184         if (f2fs_should_update_outplace(inode, NULL)) {
3185                 ret = -EINVAL;
3186                 goto out;
3187         }
3188
3189         if (f2fs_pin_file_control(inode, false)) {
3190                 ret = -EAGAIN;
3191                 goto out;
3192         }
3193
3194         ret = f2fs_convert_inline_inode(inode);
3195         if (ret)
3196                 goto out;
3197
3198         if (!f2fs_disable_compressed_file(inode)) {
3199                 ret = -EOPNOTSUPP;
3200                 goto out;
3201         }
3202
3203         set_inode_flag(inode, FI_PIN_FILE);
3204         ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3205 done:
3206         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3207 out:
3208         inode_unlock(inode);
3209         mnt_drop_write_file(filp);
3210         return ret;
3211 }
3212
3213 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3214 {
3215         struct inode *inode = file_inode(filp);
3216         __u32 pin = 0;
3217
3218         if (is_inode_flag_set(inode, FI_PIN_FILE))
3219                 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3220         return put_user(pin, (u32 __user *)arg);
3221 }
3222
3223 int f2fs_precache_extents(struct inode *inode)
3224 {
3225         struct f2fs_inode_info *fi = F2FS_I(inode);
3226         struct f2fs_map_blocks map;
3227         pgoff_t m_next_extent;
3228         loff_t end;
3229         int err;
3230
3231         if (is_inode_flag_set(inode, FI_NO_EXTENT))
3232                 return -EOPNOTSUPP;
3233
3234         map.m_lblk = 0;
3235         map.m_next_pgofs = NULL;
3236         map.m_next_extent = &m_next_extent;
3237         map.m_seg_type = NO_CHECK_TYPE;
3238         map.m_may_create = false;
3239         end = max_file_blocks(inode);
3240
3241         while (map.m_lblk < end) {
3242                 map.m_len = end - map.m_lblk;
3243
3244                 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3245                 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
3246                 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3247                 if (err)
3248                         return err;
3249
3250                 map.m_lblk = m_next_extent;
3251         }
3252
3253         return 0;
3254 }
3255
3256 static int f2fs_ioc_precache_extents(struct file *filp)
3257 {
3258         return f2fs_precache_extents(file_inode(filp));
3259 }
3260
3261 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3262 {
3263         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3264         __u64 block_count;
3265
3266         if (!capable(CAP_SYS_ADMIN))
3267                 return -EPERM;
3268
3269         if (f2fs_readonly(sbi->sb))
3270                 return -EROFS;
3271
3272         if (copy_from_user(&block_count, (void __user *)arg,
3273                            sizeof(block_count)))
3274                 return -EFAULT;
3275
3276         return f2fs_resize_fs(sbi, block_count);
3277 }
3278
3279 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3280 {
3281         struct inode *inode = file_inode(filp);
3282
3283         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3284
3285         if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3286                 f2fs_warn(F2FS_I_SB(inode),
3287                           "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3288                           inode->i_ino);
3289                 return -EOPNOTSUPP;
3290         }
3291
3292         return fsverity_ioctl_enable(filp, (const void __user *)arg);
3293 }
3294
3295 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3296 {
3297         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3298                 return -EOPNOTSUPP;
3299
3300         return fsverity_ioctl_measure(filp, (void __user *)arg);
3301 }
3302
3303 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3304 {
3305         if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3306                 return -EOPNOTSUPP;
3307
3308         return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3309 }
3310
3311 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3312 {
3313         struct inode *inode = file_inode(filp);
3314         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3315         char *vbuf;
3316         int count;
3317         int err = 0;
3318
3319         vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3320         if (!vbuf)
3321                 return -ENOMEM;
3322
3323         f2fs_down_read(&sbi->sb_lock);
3324         count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3325                         ARRAY_SIZE(sbi->raw_super->volume_name),
3326                         UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3327         f2fs_up_read(&sbi->sb_lock);
3328
3329         if (copy_to_user((char __user *)arg, vbuf,
3330                                 min(FSLABEL_MAX, count)))
3331                 err = -EFAULT;
3332
3333         kfree(vbuf);
3334         return err;
3335 }
3336
3337 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3338 {
3339         struct inode *inode = file_inode(filp);
3340         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3341         char *vbuf;
3342         int err = 0;
3343
3344         if (!capable(CAP_SYS_ADMIN))
3345                 return -EPERM;
3346
3347         vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3348         if (IS_ERR(vbuf))
3349                 return PTR_ERR(vbuf);
3350
3351         err = mnt_want_write_file(filp);
3352         if (err)
3353                 goto out;
3354
3355         f2fs_down_write(&sbi->sb_lock);
3356
3357         memset(sbi->raw_super->volume_name, 0,
3358                         sizeof(sbi->raw_super->volume_name));
3359         utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3360                         sbi->raw_super->volume_name,
3361                         ARRAY_SIZE(sbi->raw_super->volume_name));
3362
3363         err = f2fs_commit_super(sbi, false);
3364
3365         f2fs_up_write(&sbi->sb_lock);
3366
3367         mnt_drop_write_file(filp);
3368 out:
3369         kfree(vbuf);
3370         return err;
3371 }
3372
3373 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3374 {
3375         struct inode *inode = file_inode(filp);
3376         __u64 blocks;
3377
3378         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3379                 return -EOPNOTSUPP;
3380
3381         if (!f2fs_compressed_file(inode))
3382                 return -EINVAL;
3383
3384         blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3385         return put_user(blocks, (u64 __user *)arg);
3386 }
3387
3388 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3389 {
3390         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3391         unsigned int released_blocks = 0;
3392         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3393         block_t blkaddr;
3394         int i;
3395
3396         for (i = 0; i < count; i++) {
3397                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3398                                                 dn->ofs_in_node + i);
3399
3400                 if (!__is_valid_data_blkaddr(blkaddr))
3401                         continue;
3402                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3403                                         DATA_GENERIC_ENHANCE))) {
3404                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3405                         return -EFSCORRUPTED;
3406                 }
3407         }
3408
3409         while (count) {
3410                 int compr_blocks = 0;
3411
3412                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3413                         blkaddr = f2fs_data_blkaddr(dn);
3414
3415                         if (i == 0) {
3416                                 if (blkaddr == COMPRESS_ADDR)
3417                                         continue;
3418                                 dn->ofs_in_node += cluster_size;
3419                                 goto next;
3420                         }
3421
3422                         if (__is_valid_data_blkaddr(blkaddr))
3423                                 compr_blocks++;
3424
3425                         if (blkaddr != NEW_ADDR)
3426                                 continue;
3427
3428                         dn->data_blkaddr = NULL_ADDR;
3429                         f2fs_set_data_blkaddr(dn);
3430                 }
3431
3432                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3433                 dec_valid_block_count(sbi, dn->inode,
3434                                         cluster_size - compr_blocks);
3435
3436                 released_blocks += cluster_size - compr_blocks;
3437 next:
3438                 count -= cluster_size;
3439         }
3440
3441         return released_blocks;
3442 }
3443
3444 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3445 {
3446         struct inode *inode = file_inode(filp);
3447         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3448         pgoff_t page_idx = 0, last_idx;
3449         unsigned int released_blocks = 0;
3450         int ret;
3451         int writecount;
3452
3453         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3454                 return -EOPNOTSUPP;
3455
3456         if (!f2fs_compressed_file(inode))
3457                 return -EINVAL;
3458
3459         if (f2fs_readonly(sbi->sb))
3460                 return -EROFS;
3461
3462         ret = mnt_want_write_file(filp);
3463         if (ret)
3464                 return ret;
3465
3466         f2fs_balance_fs(F2FS_I_SB(inode), true);
3467
3468         inode_lock(inode);
3469
3470         writecount = atomic_read(&inode->i_writecount);
3471         if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3472                         (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3473                 ret = -EBUSY;
3474                 goto out;
3475         }
3476
3477         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3478                 ret = -EINVAL;
3479                 goto out;
3480         }
3481
3482         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3483         if (ret)
3484                 goto out;
3485
3486         set_inode_flag(inode, FI_COMPRESS_RELEASED);
3487         inode->i_ctime = current_time(inode);
3488         f2fs_mark_inode_dirty_sync(inode, true);
3489
3490         if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3491                 goto out;
3492
3493         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3494         filemap_invalidate_lock(inode->i_mapping);
3495
3496         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3497
3498         while (page_idx < last_idx) {
3499                 struct dnode_of_data dn;
3500                 pgoff_t end_offset, count;
3501
3502                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3503                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3504                 if (ret) {
3505                         if (ret == -ENOENT) {
3506                                 page_idx = f2fs_get_next_page_offset(&dn,
3507                                                                 page_idx);
3508                                 ret = 0;
3509                                 continue;
3510                         }
3511                         break;
3512                 }
3513
3514                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3515                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3516                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3517
3518                 ret = release_compress_blocks(&dn, count);
3519
3520                 f2fs_put_dnode(&dn);
3521
3522                 if (ret < 0)
3523                         break;
3524
3525                 page_idx += count;
3526                 released_blocks += ret;
3527         }
3528
3529         filemap_invalidate_unlock(inode->i_mapping);
3530         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3531 out:
3532         inode_unlock(inode);
3533
3534         mnt_drop_write_file(filp);
3535
3536         if (ret >= 0) {
3537                 ret = put_user(released_blocks, (u64 __user *)arg);
3538         } else if (released_blocks &&
3539                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3540                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3541                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3542                         "iblocks=%llu, released=%u, compr_blocks=%u, "
3543                         "run fsck to fix.",
3544                         __func__, inode->i_ino, inode->i_blocks,
3545                         released_blocks,
3546                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3547         }
3548
3549         return ret;
3550 }
3551
3552 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3553 {
3554         struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3555         unsigned int reserved_blocks = 0;
3556         int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3557         block_t blkaddr;
3558         int i;
3559
3560         for (i = 0; i < count; i++) {
3561                 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3562                                                 dn->ofs_in_node + i);
3563
3564                 if (!__is_valid_data_blkaddr(blkaddr))
3565                         continue;
3566                 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3567                                         DATA_GENERIC_ENHANCE))) {
3568                         f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
3569                         return -EFSCORRUPTED;
3570                 }
3571         }
3572
3573         while (count) {
3574                 int compr_blocks = 0;
3575                 blkcnt_t reserved;
3576                 int ret;
3577
3578                 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3579                         blkaddr = f2fs_data_blkaddr(dn);
3580
3581                         if (i == 0) {
3582                                 if (blkaddr == COMPRESS_ADDR)
3583                                         continue;
3584                                 dn->ofs_in_node += cluster_size;
3585                                 goto next;
3586                         }
3587
3588                         if (__is_valid_data_blkaddr(blkaddr)) {
3589                                 compr_blocks++;
3590                                 continue;
3591                         }
3592
3593                         dn->data_blkaddr = NEW_ADDR;
3594                         f2fs_set_data_blkaddr(dn);
3595                 }
3596
3597                 reserved = cluster_size - compr_blocks;
3598                 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3599                 if (ret)
3600                         return ret;
3601
3602                 if (reserved != cluster_size - compr_blocks)
3603                         return -ENOSPC;
3604
3605                 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3606
3607                 reserved_blocks += reserved;
3608 next:
3609                 count -= cluster_size;
3610         }
3611
3612         return reserved_blocks;
3613 }
3614
3615 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3616 {
3617         struct inode *inode = file_inode(filp);
3618         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3619         pgoff_t page_idx = 0, last_idx;
3620         unsigned int reserved_blocks = 0;
3621         int ret;
3622
3623         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3624                 return -EOPNOTSUPP;
3625
3626         if (!f2fs_compressed_file(inode))
3627                 return -EINVAL;
3628
3629         if (f2fs_readonly(sbi->sb))
3630                 return -EROFS;
3631
3632         ret = mnt_want_write_file(filp);
3633         if (ret)
3634                 return ret;
3635
3636         if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3637                 goto out;
3638
3639         f2fs_balance_fs(F2FS_I_SB(inode), true);
3640
3641         inode_lock(inode);
3642
3643         if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3644                 ret = -EINVAL;
3645                 goto unlock_inode;
3646         }
3647
3648         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3649         filemap_invalidate_lock(inode->i_mapping);
3650
3651         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3652
3653         while (page_idx < last_idx) {
3654                 struct dnode_of_data dn;
3655                 pgoff_t end_offset, count;
3656
3657                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3658                 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3659                 if (ret) {
3660                         if (ret == -ENOENT) {
3661                                 page_idx = f2fs_get_next_page_offset(&dn,
3662                                                                 page_idx);
3663                                 ret = 0;
3664                                 continue;
3665                         }
3666                         break;
3667                 }
3668
3669                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3670                 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3671                 count = round_up(count, F2FS_I(inode)->i_cluster_size);
3672
3673                 ret = reserve_compress_blocks(&dn, count);
3674
3675                 f2fs_put_dnode(&dn);
3676
3677                 if (ret < 0)
3678                         break;
3679
3680                 page_idx += count;
3681                 reserved_blocks += ret;
3682         }
3683
3684         filemap_invalidate_unlock(inode->i_mapping);
3685         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3686
3687         if (ret >= 0) {
3688                 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3689                 inode->i_ctime = current_time(inode);
3690                 f2fs_mark_inode_dirty_sync(inode, true);
3691         }
3692 unlock_inode:
3693         inode_unlock(inode);
3694 out:
3695         mnt_drop_write_file(filp);
3696
3697         if (ret >= 0) {
3698                 ret = put_user(reserved_blocks, (u64 __user *)arg);
3699         } else if (reserved_blocks &&
3700                         atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3701                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3702                 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3703                         "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3704                         "run fsck to fix.",
3705                         __func__, inode->i_ino, inode->i_blocks,
3706                         reserved_blocks,
3707                         atomic_read(&F2FS_I(inode)->i_compr_blocks));
3708         }
3709
3710         return ret;
3711 }
3712
3713 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3714                 pgoff_t off, block_t block, block_t len, u32 flags)
3715 {
3716         sector_t sector = SECTOR_FROM_BLOCK(block);
3717         sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3718         int ret = 0;
3719
3720         if (flags & F2FS_TRIM_FILE_DISCARD) {
3721                 if (bdev_max_secure_erase_sectors(bdev))
3722                         ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3723                                         GFP_NOFS);
3724                 else
3725                         ret = blkdev_issue_discard(bdev, sector, nr_sects,
3726                                         GFP_NOFS);
3727         }
3728
3729         if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3730                 if (IS_ENCRYPTED(inode))
3731                         ret = fscrypt_zeroout_range(inode, off, block, len);
3732                 else
3733                         ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3734                                         GFP_NOFS, 0);
3735         }
3736
3737         return ret;
3738 }
3739
3740 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3741 {
3742         struct inode *inode = file_inode(filp);
3743         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3744         struct address_space *mapping = inode->i_mapping;
3745         struct block_device *prev_bdev = NULL;
3746         struct f2fs_sectrim_range range;
3747         pgoff_t index, pg_end, prev_index = 0;
3748         block_t prev_block = 0, len = 0;
3749         loff_t end_addr;
3750         bool to_end = false;
3751         int ret = 0;
3752
3753         if (!(filp->f_mode & FMODE_WRITE))
3754                 return -EBADF;
3755
3756         if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3757                                 sizeof(range)))
3758                 return -EFAULT;
3759
3760         if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3761                         !S_ISREG(inode->i_mode))
3762                 return -EINVAL;
3763
3764         if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3765                         !f2fs_hw_support_discard(sbi)) ||
3766                         ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3767                          IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3768                 return -EOPNOTSUPP;
3769
3770         file_start_write(filp);
3771         inode_lock(inode);
3772
3773         if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3774                         range.start >= inode->i_size) {
3775                 ret = -EINVAL;
3776                 goto err;
3777         }
3778
3779         if (range.len == 0)
3780                 goto err;
3781
3782         if (inode->i_size - range.start > range.len) {
3783                 end_addr = range.start + range.len;
3784         } else {
3785                 end_addr = range.len == (u64)-1 ?
3786                         sbi->sb->s_maxbytes : inode->i_size;
3787                 to_end = true;
3788         }
3789
3790         if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3791                         (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3792                 ret = -EINVAL;
3793                 goto err;
3794         }
3795
3796         index = F2FS_BYTES_TO_BLK(range.start);
3797         pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3798
3799         ret = f2fs_convert_inline_inode(inode);
3800         if (ret)
3801                 goto err;
3802
3803         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3804         filemap_invalidate_lock(mapping);
3805
3806         ret = filemap_write_and_wait_range(mapping, range.start,
3807                         to_end ? LLONG_MAX : end_addr - 1);
3808         if (ret)
3809                 goto out;
3810
3811         truncate_inode_pages_range(mapping, range.start,
3812                         to_end ? -1 : end_addr - 1);
3813
3814         while (index < pg_end) {
3815                 struct dnode_of_data dn;
3816                 pgoff_t end_offset, count;
3817                 int i;
3818
3819                 set_new_dnode(&dn, inode, NULL, NULL, 0);
3820                 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3821                 if (ret) {
3822                         if (ret == -ENOENT) {
3823                                 index = f2fs_get_next_page_offset(&dn, index);
3824                                 continue;
3825                         }
3826                         goto out;
3827                 }
3828
3829                 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3830                 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3831                 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3832                         struct block_device *cur_bdev;
3833                         block_t blkaddr = f2fs_data_blkaddr(&dn);
3834
3835                         if (!__is_valid_data_blkaddr(blkaddr))
3836                                 continue;
3837
3838                         if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3839                                                 DATA_GENERIC_ENHANCE)) {
3840                                 ret = -EFSCORRUPTED;
3841                                 f2fs_put_dnode(&dn);
3842                                 f2fs_handle_error(sbi,
3843                                                 ERROR_INVALID_BLKADDR);
3844                                 goto out;
3845                         }
3846
3847                         cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3848                         if (f2fs_is_multi_device(sbi)) {
3849                                 int di = f2fs_target_device_index(sbi, blkaddr);
3850
3851                                 blkaddr -= FDEV(di).start_blk;
3852                         }
3853
3854                         if (len) {
3855                                 if (prev_bdev == cur_bdev &&
3856                                                 index == prev_index + len &&
3857                                                 blkaddr == prev_block + len) {
3858                                         len++;
3859                                 } else {
3860                                         ret = f2fs_secure_erase(prev_bdev,
3861                                                 inode, prev_index, prev_block,
3862                                                 len, range.flags);
3863                                         if (ret) {
3864                                                 f2fs_put_dnode(&dn);
3865                                                 goto out;
3866                                         }
3867
3868                                         len = 0;
3869                                 }
3870                         }
3871
3872                         if (!len) {
3873                                 prev_bdev = cur_bdev;
3874                                 prev_index = index;
3875                                 prev_block = blkaddr;
3876                                 len = 1;
3877                         }
3878                 }
3879
3880                 f2fs_put_dnode(&dn);
3881
3882                 if (fatal_signal_pending(current)) {
3883                         ret = -EINTR;
3884                         goto out;
3885                 }
3886                 cond_resched();
3887         }
3888
3889         if (len)
3890                 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3891                                 prev_block, len, range.flags);
3892 out:
3893         filemap_invalidate_unlock(mapping);
3894         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3895 err:
3896         inode_unlock(inode);
3897         file_end_write(filp);
3898
3899         return ret;
3900 }
3901
3902 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3903 {
3904         struct inode *inode = file_inode(filp);
3905         struct f2fs_comp_option option;
3906
3907         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3908                 return -EOPNOTSUPP;
3909
3910         inode_lock_shared(inode);
3911
3912         if (!f2fs_compressed_file(inode)) {
3913                 inode_unlock_shared(inode);
3914                 return -ENODATA;
3915         }
3916
3917         option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3918         option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3919
3920         inode_unlock_shared(inode);
3921
3922         if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3923                                 sizeof(option)))
3924                 return -EFAULT;
3925
3926         return 0;
3927 }
3928
3929 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3930 {
3931         struct inode *inode = file_inode(filp);
3932         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3933         struct f2fs_comp_option option;
3934         int ret = 0;
3935
3936         if (!f2fs_sb_has_compression(sbi))
3937                 return -EOPNOTSUPP;
3938
3939         if (!(filp->f_mode & FMODE_WRITE))
3940                 return -EBADF;
3941
3942         if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3943                                 sizeof(option)))
3944                 return -EFAULT;
3945
3946         if (!f2fs_compressed_file(inode) ||
3947                         option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3948                         option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3949                         option.algorithm >= COMPRESS_MAX)
3950                 return -EINVAL;
3951
3952         file_start_write(filp);
3953         inode_lock(inode);
3954
3955         if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3956                 ret = -EBUSY;
3957                 goto out;
3958         }
3959
3960         if (F2FS_HAS_BLOCKS(inode)) {
3961                 ret = -EFBIG;
3962                 goto out;
3963         }
3964
3965         F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3966         F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3967         F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3968         f2fs_mark_inode_dirty_sync(inode, true);
3969
3970         if (!f2fs_is_compress_backend_ready(inode))
3971                 f2fs_warn(sbi, "compression algorithm is successfully set, "
3972                         "but current kernel doesn't support this algorithm.");
3973 out:
3974         inode_unlock(inode);
3975         file_end_write(filp);
3976
3977         return ret;
3978 }
3979
3980 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3981 {
3982         DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3983         struct address_space *mapping = inode->i_mapping;
3984         struct page *page;
3985         pgoff_t redirty_idx = page_idx;
3986         int i, page_len = 0, ret = 0;
3987
3988         page_cache_ra_unbounded(&ractl, len, 0);
3989
3990         for (i = 0; i < len; i++, page_idx++) {
3991                 page = read_cache_page(mapping, page_idx, NULL, NULL);
3992                 if (IS_ERR(page)) {
3993                         ret = PTR_ERR(page);
3994                         break;
3995                 }
3996                 page_len++;
3997         }
3998
3999         for (i = 0; i < page_len; i++, redirty_idx++) {
4000                 page = find_lock_page(mapping, redirty_idx);
4001
4002                 /* It will never fail, when page has pinned above */
4003                 f2fs_bug_on(F2FS_I_SB(inode), !page);
4004
4005                 set_page_dirty(page);
4006                 f2fs_put_page(page, 1);
4007                 f2fs_put_page(page, 0);
4008         }
4009
4010         return ret;
4011 }
4012
4013 static int f2fs_ioc_decompress_file(struct file *filp)
4014 {
4015         struct inode *inode = file_inode(filp);
4016         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4017         struct f2fs_inode_info *fi = F2FS_I(inode);
4018         pgoff_t page_idx = 0, last_idx;
4019         unsigned int blk_per_seg = sbi->blocks_per_seg;
4020         int cluster_size = fi->i_cluster_size;
4021         int count, ret;
4022
4023         if (!f2fs_sb_has_compression(sbi) ||
4024                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4025                 return -EOPNOTSUPP;
4026
4027         if (!(filp->f_mode & FMODE_WRITE))
4028                 return -EBADF;
4029
4030         if (!f2fs_compressed_file(inode))
4031                 return -EINVAL;
4032
4033         f2fs_balance_fs(F2FS_I_SB(inode), true);
4034
4035         file_start_write(filp);
4036         inode_lock(inode);
4037
4038         if (!f2fs_is_compress_backend_ready(inode)) {
4039                 ret = -EOPNOTSUPP;
4040                 goto out;
4041         }
4042
4043         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4044                 ret = -EINVAL;
4045                 goto out;
4046         }
4047
4048         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4049         if (ret)
4050                 goto out;
4051
4052         if (!atomic_read(&fi->i_compr_blocks))
4053                 goto out;
4054
4055         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4056
4057         count = last_idx - page_idx;
4058         while (count) {
4059                 int len = min(cluster_size, count);
4060
4061                 ret = redirty_blocks(inode, page_idx, len);
4062                 if (ret < 0)
4063                         break;
4064
4065                 if (get_dirty_pages(inode) >= blk_per_seg)
4066                         filemap_fdatawrite(inode->i_mapping);
4067
4068                 count -= len;
4069                 page_idx += len;
4070         }
4071
4072         if (!ret)
4073                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4074                                                         LLONG_MAX);
4075
4076         if (ret)
4077                 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4078                           __func__, ret);
4079 out:
4080         inode_unlock(inode);
4081         file_end_write(filp);
4082
4083         return ret;
4084 }
4085
4086 static int f2fs_ioc_compress_file(struct file *filp)
4087 {
4088         struct inode *inode = file_inode(filp);
4089         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4090         pgoff_t page_idx = 0, last_idx;
4091         unsigned int blk_per_seg = sbi->blocks_per_seg;
4092         int cluster_size = F2FS_I(inode)->i_cluster_size;
4093         int count, ret;
4094
4095         if (!f2fs_sb_has_compression(sbi) ||
4096                         F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4097                 return -EOPNOTSUPP;
4098
4099         if (!(filp->f_mode & FMODE_WRITE))
4100                 return -EBADF;
4101
4102         if (!f2fs_compressed_file(inode))
4103                 return -EINVAL;
4104
4105         f2fs_balance_fs(F2FS_I_SB(inode), true);
4106
4107         file_start_write(filp);
4108         inode_lock(inode);
4109
4110         if (!f2fs_is_compress_backend_ready(inode)) {
4111                 ret = -EOPNOTSUPP;
4112                 goto out;
4113         }
4114
4115         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4116                 ret = -EINVAL;
4117                 goto out;
4118         }
4119
4120         ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4121         if (ret)
4122                 goto out;
4123
4124         set_inode_flag(inode, FI_ENABLE_COMPRESS);
4125
4126         last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4127
4128         count = last_idx - page_idx;
4129         while (count) {
4130                 int len = min(cluster_size, count);
4131
4132                 ret = redirty_blocks(inode, page_idx, len);
4133                 if (ret < 0)
4134                         break;
4135
4136                 if (get_dirty_pages(inode) >= blk_per_seg)
4137                         filemap_fdatawrite(inode->i_mapping);
4138
4139                 count -= len;
4140                 page_idx += len;
4141         }
4142
4143         if (!ret)
4144                 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4145                                                         LLONG_MAX);
4146
4147         clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4148
4149         if (ret)
4150                 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4151                           __func__, ret);
4152 out:
4153         inode_unlock(inode);
4154         file_end_write(filp);
4155
4156         return ret;
4157 }
4158
4159 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4160 {
4161         switch (cmd) {
4162         case FS_IOC_GETVERSION:
4163                 return f2fs_ioc_getversion(filp, arg);
4164         case F2FS_IOC_START_ATOMIC_WRITE:
4165                 return f2fs_ioc_start_atomic_write(filp, false);
4166         case F2FS_IOC_START_ATOMIC_REPLACE:
4167                 return f2fs_ioc_start_atomic_write(filp, true);
4168         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4169                 return f2fs_ioc_commit_atomic_write(filp);
4170         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4171                 return f2fs_ioc_abort_atomic_write(filp);
4172         case F2FS_IOC_START_VOLATILE_WRITE:
4173         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4174                 return -EOPNOTSUPP;
4175         case F2FS_IOC_SHUTDOWN:
4176                 return f2fs_ioc_shutdown(filp, arg);
4177         case FITRIM:
4178                 return f2fs_ioc_fitrim(filp, arg);
4179         case FS_IOC_SET_ENCRYPTION_POLICY:
4180                 return f2fs_ioc_set_encryption_policy(filp, arg);
4181         case FS_IOC_GET_ENCRYPTION_POLICY:
4182                 return f2fs_ioc_get_encryption_policy(filp, arg);
4183         case FS_IOC_GET_ENCRYPTION_PWSALT:
4184                 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4185         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4186                 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4187         case FS_IOC_ADD_ENCRYPTION_KEY:
4188                 return f2fs_ioc_add_encryption_key(filp, arg);
4189         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4190                 return f2fs_ioc_remove_encryption_key(filp, arg);
4191         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4192                 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4193         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4194                 return f2fs_ioc_get_encryption_key_status(filp, arg);
4195         case FS_IOC_GET_ENCRYPTION_NONCE:
4196                 return f2fs_ioc_get_encryption_nonce(filp, arg);
4197         case F2FS_IOC_GARBAGE_COLLECT:
4198                 return f2fs_ioc_gc(filp, arg);
4199         case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4200                 return f2fs_ioc_gc_range(filp, arg);
4201         case F2FS_IOC_WRITE_CHECKPOINT:
4202                 return f2fs_ioc_write_checkpoint(filp);
4203         case F2FS_IOC_DEFRAGMENT:
4204                 return f2fs_ioc_defragment(filp, arg);
4205         case F2FS_IOC_MOVE_RANGE:
4206                 return f2fs_ioc_move_range(filp, arg);
4207         case F2FS_IOC_FLUSH_DEVICE:
4208                 return f2fs_ioc_flush_device(filp, arg);
4209         case F2FS_IOC_GET_FEATURES:
4210                 return f2fs_ioc_get_features(filp, arg);
4211         case F2FS_IOC_GET_PIN_FILE:
4212                 return f2fs_ioc_get_pin_file(filp, arg);
4213         case F2FS_IOC_SET_PIN_FILE:
4214                 return f2fs_ioc_set_pin_file(filp, arg);
4215         case F2FS_IOC_PRECACHE_EXTENTS:
4216                 return f2fs_ioc_precache_extents(filp);
4217         case F2FS_IOC_RESIZE_FS:
4218                 return f2fs_ioc_resize_fs(filp, arg);
4219         case FS_IOC_ENABLE_VERITY:
4220                 return f2fs_ioc_enable_verity(filp, arg);
4221         case FS_IOC_MEASURE_VERITY:
4222                 return f2fs_ioc_measure_verity(filp, arg);
4223         case FS_IOC_READ_VERITY_METADATA:
4224                 return f2fs_ioc_read_verity_metadata(filp, arg);
4225         case FS_IOC_GETFSLABEL:
4226                 return f2fs_ioc_getfslabel(filp, arg);
4227         case FS_IOC_SETFSLABEL:
4228                 return f2fs_ioc_setfslabel(filp, arg);
4229         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4230                 return f2fs_get_compress_blocks(filp, arg);
4231         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4232                 return f2fs_release_compress_blocks(filp, arg);
4233         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4234                 return f2fs_reserve_compress_blocks(filp, arg);
4235         case F2FS_IOC_SEC_TRIM_FILE:
4236                 return f2fs_sec_trim_file(filp, arg);
4237         case F2FS_IOC_GET_COMPRESS_OPTION:
4238                 return f2fs_ioc_get_compress_option(filp, arg);
4239         case F2FS_IOC_SET_COMPRESS_OPTION:
4240                 return f2fs_ioc_set_compress_option(filp, arg);
4241         case F2FS_IOC_DECOMPRESS_FILE:
4242                 return f2fs_ioc_decompress_file(filp);
4243         case F2FS_IOC_COMPRESS_FILE:
4244                 return f2fs_ioc_compress_file(filp);
4245         default:
4246                 return -ENOTTY;
4247         }
4248 }
4249
4250 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4251 {
4252         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4253                 return -EIO;
4254         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4255                 return -ENOSPC;
4256
4257         return __f2fs_ioctl(filp, cmd, arg);
4258 }
4259
4260 /*
4261  * Return %true if the given read or write request should use direct I/O, or
4262  * %false if it should use buffered I/O.
4263  */
4264 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4265                                 struct iov_iter *iter)
4266 {
4267         unsigned int align;
4268
4269         if (!(iocb->ki_flags & IOCB_DIRECT))
4270                 return false;
4271
4272         if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
4273                 return false;
4274
4275         /*
4276          * Direct I/O not aligned to the disk's logical_block_size will be
4277          * attempted, but will fail with -EINVAL.
4278          *
4279          * f2fs additionally requires that direct I/O be aligned to the
4280          * filesystem block size, which is often a stricter requirement.
4281          * However, f2fs traditionally falls back to buffered I/O on requests
4282          * that are logical_block_size-aligned but not fs-block aligned.
4283          *
4284          * The below logic implements this behavior.
4285          */
4286         align = iocb->ki_pos | iov_iter_alignment(iter);
4287         if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4288             IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4289                 return false;
4290
4291         return true;
4292 }
4293
4294 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4295                                 unsigned int flags)
4296 {
4297         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4298
4299         dec_page_count(sbi, F2FS_DIO_READ);
4300         if (error)
4301                 return error;
4302         f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
4303         return 0;
4304 }
4305
4306 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4307         .end_io = f2fs_dio_read_end_io,
4308 };
4309
4310 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4311 {
4312         struct file *file = iocb->ki_filp;
4313         struct inode *inode = file_inode(file);
4314         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4315         struct f2fs_inode_info *fi = F2FS_I(inode);
4316         const loff_t pos = iocb->ki_pos;
4317         const size_t count = iov_iter_count(to);
4318         struct iomap_dio *dio;
4319         ssize_t ret;
4320
4321         if (count == 0)
4322                 return 0; /* skip atime update */
4323
4324         trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4325
4326         if (iocb->ki_flags & IOCB_NOWAIT) {
4327                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4328                         ret = -EAGAIN;
4329                         goto out;
4330                 }
4331         } else {
4332                 f2fs_down_read(&fi->i_gc_rwsem[READ]);
4333         }
4334
4335         /*
4336          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4337          * the higher-level function iomap_dio_rw() in order to ensure that the
4338          * F2FS_DIO_READ counter will be decremented correctly in all cases.
4339          */
4340         inc_page_count(sbi, F2FS_DIO_READ);
4341         dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4342                              &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4343         if (IS_ERR_OR_NULL(dio)) {
4344                 ret = PTR_ERR_OR_ZERO(dio);
4345                 if (ret != -EIOCBQUEUED)
4346                         dec_page_count(sbi, F2FS_DIO_READ);
4347         } else {
4348                 ret = iomap_dio_complete(dio);
4349         }
4350
4351         f2fs_up_read(&fi->i_gc_rwsem[READ]);
4352
4353         file_accessed(file);
4354 out:
4355         trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4356         return ret;
4357 }
4358
4359 static void f2fs_trace_rw_file_path(struct kiocb *iocb, size_t count, int rw)
4360 {
4361         struct inode *inode = file_inode(iocb->ki_filp);
4362         char *buf, *path;
4363
4364         buf = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4365         if (!buf)
4366                 return;
4367         path = dentry_path_raw(file_dentry(iocb->ki_filp), buf, PATH_MAX);
4368         if (IS_ERR(path))
4369                 goto free_buf;
4370         if (rw == WRITE)
4371                 trace_f2fs_datawrite_start(inode, iocb->ki_pos, count,
4372                                 current->pid, path, current->comm);
4373         else
4374                 trace_f2fs_dataread_start(inode, iocb->ki_pos, count,
4375                                 current->pid, path, current->comm);
4376 free_buf:
4377         kfree(buf);
4378 }
4379
4380 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4381 {
4382         struct inode *inode = file_inode(iocb->ki_filp);
4383         const loff_t pos = iocb->ki_pos;
4384         ssize_t ret;
4385
4386         if (!f2fs_is_compress_backend_ready(inode))
4387                 return -EOPNOTSUPP;
4388
4389         if (trace_f2fs_dataread_start_enabled())
4390                 f2fs_trace_rw_file_path(iocb, iov_iter_count(to), READ);
4391
4392         if (f2fs_should_use_dio(inode, iocb, to)) {
4393                 ret = f2fs_dio_read_iter(iocb, to);
4394         } else {
4395                 ret = filemap_read(iocb, to, 0);
4396                 if (ret > 0)
4397                         f2fs_update_iostat(F2FS_I_SB(inode), inode,
4398                                                 APP_BUFFERED_READ_IO, ret);
4399         }
4400         if (trace_f2fs_dataread_end_enabled())
4401                 trace_f2fs_dataread_end(inode, pos, ret);
4402         return ret;
4403 }
4404
4405 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4406 {
4407         struct file *file = iocb->ki_filp;
4408         struct inode *inode = file_inode(file);
4409         ssize_t count;
4410         int err;
4411
4412         if (IS_IMMUTABLE(inode))
4413                 return -EPERM;
4414
4415         if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4416                 return -EPERM;
4417
4418         count = generic_write_checks(iocb, from);
4419         if (count <= 0)
4420                 return count;
4421
4422         err = file_modified(file);
4423         if (err)
4424                 return err;
4425         return count;
4426 }
4427
4428 /*
4429  * Preallocate blocks for a write request, if it is possible and helpful to do
4430  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4431  * blocks were preallocated, or a negative errno value if something went
4432  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4433  * requested blocks (not just some of them) have been allocated.
4434  */
4435 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4436                                    bool dio)
4437 {
4438         struct inode *inode = file_inode(iocb->ki_filp);
4439         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4440         const loff_t pos = iocb->ki_pos;
4441         const size_t count = iov_iter_count(iter);
4442         struct f2fs_map_blocks map = {};
4443         int flag;
4444         int ret;
4445
4446         /* If it will be an out-of-place direct write, don't bother. */
4447         if (dio && f2fs_lfs_mode(sbi))
4448                 return 0;
4449         /*
4450          * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4451          * buffered IO, if DIO meets any holes.
4452          */
4453         if (dio && i_size_read(inode) &&
4454                 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4455                 return 0;
4456
4457         /* No-wait I/O can't allocate blocks. */
4458         if (iocb->ki_flags & IOCB_NOWAIT)
4459                 return 0;
4460
4461         /* If it will be a short write, don't bother. */
4462         if (fault_in_iov_iter_readable(iter, count))
4463                 return 0;
4464
4465         if (f2fs_has_inline_data(inode)) {
4466                 /* If the data will fit inline, don't bother. */
4467                 if (pos + count <= MAX_INLINE_DATA(inode))
4468                         return 0;
4469                 ret = f2fs_convert_inline_inode(inode);
4470                 if (ret)
4471                         return ret;
4472         }
4473
4474         /* Do not preallocate blocks that will be written partially in 4KB. */
4475         map.m_lblk = F2FS_BLK_ALIGN(pos);
4476         map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4477         if (map.m_len > map.m_lblk)
4478                 map.m_len -= map.m_lblk;
4479         else
4480                 map.m_len = 0;
4481         map.m_may_create = true;
4482         if (dio) {
4483                 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4484                 flag = F2FS_GET_BLOCK_PRE_DIO;
4485         } else {
4486                 map.m_seg_type = NO_CHECK_TYPE;
4487                 flag = F2FS_GET_BLOCK_PRE_AIO;
4488         }
4489
4490         ret = f2fs_map_blocks(inode, &map, flag);
4491         /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4492         if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4493                 return ret;
4494         if (ret == 0)
4495                 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4496         return map.m_len;
4497 }
4498
4499 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4500                                         struct iov_iter *from)
4501 {
4502         struct file *file = iocb->ki_filp;
4503         struct inode *inode = file_inode(file);
4504         ssize_t ret;
4505
4506         if (iocb->ki_flags & IOCB_NOWAIT)
4507                 return -EOPNOTSUPP;
4508
4509         current->backing_dev_info = inode_to_bdi(inode);
4510         ret = generic_perform_write(iocb, from);
4511         current->backing_dev_info = NULL;
4512
4513         if (ret > 0) {
4514                 iocb->ki_pos += ret;
4515                 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4516                                                 APP_BUFFERED_IO, ret);
4517         }
4518         return ret;
4519 }
4520
4521 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4522                                  unsigned int flags)
4523 {
4524         struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4525
4526         dec_page_count(sbi, F2FS_DIO_WRITE);
4527         if (error)
4528                 return error;
4529         f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
4530         return 0;
4531 }
4532
4533 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4534         .end_io = f2fs_dio_write_end_io,
4535 };
4536
4537 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4538                                    bool *may_need_sync)
4539 {
4540         struct file *file = iocb->ki_filp;
4541         struct inode *inode = file_inode(file);
4542         struct f2fs_inode_info *fi = F2FS_I(inode);
4543         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4544         const bool do_opu = f2fs_lfs_mode(sbi);
4545         const loff_t pos = iocb->ki_pos;
4546         const ssize_t count = iov_iter_count(from);
4547         unsigned int dio_flags;
4548         struct iomap_dio *dio;
4549         ssize_t ret;
4550
4551         trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4552
4553         if (iocb->ki_flags & IOCB_NOWAIT) {
4554                 /* f2fs_convert_inline_inode() and block allocation can block */
4555                 if (f2fs_has_inline_data(inode) ||
4556                     !f2fs_overwrite_io(inode, pos, count)) {
4557                         ret = -EAGAIN;
4558                         goto out;
4559                 }
4560
4561                 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4562                         ret = -EAGAIN;
4563                         goto out;
4564                 }
4565                 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4566                         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4567                         ret = -EAGAIN;
4568                         goto out;
4569                 }
4570         } else {
4571                 ret = f2fs_convert_inline_inode(inode);
4572                 if (ret)
4573                         goto out;
4574
4575                 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4576                 if (do_opu)
4577                         f2fs_down_read(&fi->i_gc_rwsem[READ]);
4578         }
4579
4580         /*
4581          * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4582          * the higher-level function iomap_dio_rw() in order to ensure that the
4583          * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4584          */
4585         inc_page_count(sbi, F2FS_DIO_WRITE);
4586         dio_flags = 0;
4587         if (pos + count > inode->i_size)
4588                 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4589         dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4590                              &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4591         if (IS_ERR_OR_NULL(dio)) {
4592                 ret = PTR_ERR_OR_ZERO(dio);
4593                 if (ret == -ENOTBLK)
4594                         ret = 0;
4595                 if (ret != -EIOCBQUEUED)
4596                         dec_page_count(sbi, F2FS_DIO_WRITE);
4597         } else {
4598                 ret = iomap_dio_complete(dio);
4599         }
4600
4601         if (do_opu)
4602                 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4603         f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4604
4605         if (ret < 0)
4606                 goto out;
4607         if (pos + ret > inode->i_size)
4608                 f2fs_i_size_write(inode, pos + ret);
4609         if (!do_opu)
4610                 set_inode_flag(inode, FI_UPDATE_WRITE);
4611
4612         if (iov_iter_count(from)) {
4613                 ssize_t ret2;
4614                 loff_t bufio_start_pos = iocb->ki_pos;
4615
4616                 /*
4617                  * The direct write was partial, so we need to fall back to a
4618                  * buffered write for the remainder.
4619                  */
4620
4621                 ret2 = f2fs_buffered_write_iter(iocb, from);
4622                 if (iov_iter_count(from))
4623                         f2fs_write_failed(inode, iocb->ki_pos);
4624                 if (ret2 < 0)
4625                         goto out;
4626
4627                 /*
4628                  * Ensure that the pagecache pages are written to disk and
4629                  * invalidated to preserve the expected O_DIRECT semantics.
4630                  */
4631                 if (ret2 > 0) {
4632                         loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4633
4634                         ret += ret2;
4635
4636                         ret2 = filemap_write_and_wait_range(file->f_mapping,
4637                                                             bufio_start_pos,
4638                                                             bufio_end_pos);
4639                         if (ret2 < 0)
4640                                 goto out;
4641                         invalidate_mapping_pages(file->f_mapping,
4642                                                  bufio_start_pos >> PAGE_SHIFT,
4643                                                  bufio_end_pos >> PAGE_SHIFT);
4644                 }
4645         } else {
4646                 /* iomap_dio_rw() already handled the generic_write_sync(). */
4647                 *may_need_sync = false;
4648         }
4649 out:
4650         trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4651         return ret;
4652 }
4653
4654 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4655 {
4656         struct inode *inode = file_inode(iocb->ki_filp);
4657         const loff_t orig_pos = iocb->ki_pos;
4658         const size_t orig_count = iov_iter_count(from);
4659         loff_t target_size;
4660         bool dio;
4661         bool may_need_sync = true;
4662         int preallocated;
4663         ssize_t ret;
4664
4665         if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4666                 ret = -EIO;
4667                 goto out;
4668         }
4669
4670         if (!f2fs_is_compress_backend_ready(inode)) {
4671                 ret = -EOPNOTSUPP;
4672                 goto out;
4673         }
4674
4675         if (iocb->ki_flags & IOCB_NOWAIT) {
4676                 if (!inode_trylock(inode)) {
4677                         ret = -EAGAIN;
4678                         goto out;
4679                 }
4680         } else {
4681                 inode_lock(inode);
4682         }
4683
4684         ret = f2fs_write_checks(iocb, from);
4685         if (ret <= 0)
4686                 goto out_unlock;
4687
4688         /* Determine whether we will do a direct write or a buffered write. */
4689         dio = f2fs_should_use_dio(inode, iocb, from);
4690
4691         /* Possibly preallocate the blocks for the write. */
4692         target_size = iocb->ki_pos + iov_iter_count(from);
4693         preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4694         if (preallocated < 0) {
4695                 ret = preallocated;
4696         } else {
4697                 if (trace_f2fs_datawrite_start_enabled())
4698                         f2fs_trace_rw_file_path(iocb, orig_count, WRITE);
4699
4700                 /* Do the actual write. */
4701                 ret = dio ?
4702                         f2fs_dio_write_iter(iocb, from, &may_need_sync) :
4703                         f2fs_buffered_write_iter(iocb, from);
4704
4705                 if (trace_f2fs_datawrite_end_enabled())
4706                         trace_f2fs_datawrite_end(inode, orig_pos, ret);
4707         }
4708
4709         /* Don't leave any preallocated blocks around past i_size. */
4710         if (preallocated && i_size_read(inode) < target_size) {
4711                 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4712                 filemap_invalidate_lock(inode->i_mapping);
4713                 if (!f2fs_truncate(inode))
4714                         file_dont_truncate(inode);
4715                 filemap_invalidate_unlock(inode->i_mapping);
4716                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4717         } else {
4718                 file_dont_truncate(inode);
4719         }
4720
4721         clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4722 out_unlock:
4723         inode_unlock(inode);
4724 out:
4725         trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4726         if (ret > 0 && may_need_sync)
4727                 ret = generic_write_sync(iocb, ret);
4728         return ret;
4729 }
4730
4731 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4732                 int advice)
4733 {
4734         struct address_space *mapping;
4735         struct backing_dev_info *bdi;
4736         struct inode *inode = file_inode(filp);
4737         int err;
4738
4739         if (advice == POSIX_FADV_SEQUENTIAL) {
4740                 if (S_ISFIFO(inode->i_mode))
4741                         return -ESPIPE;
4742
4743                 mapping = filp->f_mapping;
4744                 if (!mapping || len < 0)
4745                         return -EINVAL;
4746
4747                 bdi = inode_to_bdi(mapping->host);
4748                 filp->f_ra.ra_pages = bdi->ra_pages *
4749                         F2FS_I_SB(inode)->seq_file_ra_mul;
4750                 spin_lock(&filp->f_lock);
4751                 filp->f_mode &= ~FMODE_RANDOM;
4752                 spin_unlock(&filp->f_lock);
4753                 return 0;
4754         }
4755
4756         err = generic_fadvise(filp, offset, len, advice);
4757         if (!err && advice == POSIX_FADV_DONTNEED &&
4758                 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4759                 f2fs_compressed_file(inode))
4760                 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4761
4762         return err;
4763 }
4764
4765 #ifdef CONFIG_COMPAT
4766 struct compat_f2fs_gc_range {
4767         u32 sync;
4768         compat_u64 start;
4769         compat_u64 len;
4770 };
4771 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE        _IOW(F2FS_IOCTL_MAGIC, 11,\
4772                                                 struct compat_f2fs_gc_range)
4773
4774 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4775 {
4776         struct compat_f2fs_gc_range __user *urange;
4777         struct f2fs_gc_range range;
4778         int err;
4779
4780         urange = compat_ptr(arg);
4781         err = get_user(range.sync, &urange->sync);
4782         err |= get_user(range.start, &urange->start);
4783         err |= get_user(range.len, &urange->len);
4784         if (err)
4785                 return -EFAULT;
4786
4787         return __f2fs_ioc_gc_range(file, &range);
4788 }
4789
4790 struct compat_f2fs_move_range {
4791         u32 dst_fd;
4792         compat_u64 pos_in;
4793         compat_u64 pos_out;
4794         compat_u64 len;
4795 };
4796 #define F2FS_IOC32_MOVE_RANGE           _IOWR(F2FS_IOCTL_MAGIC, 9,      \
4797                                         struct compat_f2fs_move_range)
4798
4799 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4800 {
4801         struct compat_f2fs_move_range __user *urange;
4802         struct f2fs_move_range range;
4803         int err;
4804
4805         urange = compat_ptr(arg);
4806         err = get_user(range.dst_fd, &urange->dst_fd);
4807         err |= get_user(range.pos_in, &urange->pos_in);
4808         err |= get_user(range.pos_out, &urange->pos_out);
4809         err |= get_user(range.len, &urange->len);
4810         if (err)
4811                 return -EFAULT;
4812
4813         return __f2fs_ioc_move_range(file, &range);
4814 }
4815
4816 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4817 {
4818         if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4819                 return -EIO;
4820         if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4821                 return -ENOSPC;
4822
4823         switch (cmd) {
4824         case FS_IOC32_GETVERSION:
4825                 cmd = FS_IOC_GETVERSION;
4826                 break;
4827         case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4828                 return f2fs_compat_ioc_gc_range(file, arg);
4829         case F2FS_IOC32_MOVE_RANGE:
4830                 return f2fs_compat_ioc_move_range(file, arg);
4831         case F2FS_IOC_START_ATOMIC_WRITE:
4832         case F2FS_IOC_START_ATOMIC_REPLACE:
4833         case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4834         case F2FS_IOC_START_VOLATILE_WRITE:
4835         case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4836         case F2FS_IOC_ABORT_ATOMIC_WRITE:
4837         case F2FS_IOC_SHUTDOWN:
4838         case FITRIM:
4839         case FS_IOC_SET_ENCRYPTION_POLICY:
4840         case FS_IOC_GET_ENCRYPTION_PWSALT:
4841         case FS_IOC_GET_ENCRYPTION_POLICY:
4842         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4843         case FS_IOC_ADD_ENCRYPTION_KEY:
4844         case FS_IOC_REMOVE_ENCRYPTION_KEY:
4845         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4846         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4847         case FS_IOC_GET_ENCRYPTION_NONCE:
4848         case F2FS_IOC_GARBAGE_COLLECT:
4849         case F2FS_IOC_WRITE_CHECKPOINT:
4850         case F2FS_IOC_DEFRAGMENT:
4851         case F2FS_IOC_FLUSH_DEVICE:
4852         case F2FS_IOC_GET_FEATURES:
4853         case F2FS_IOC_GET_PIN_FILE:
4854         case F2FS_IOC_SET_PIN_FILE:
4855         case F2FS_IOC_PRECACHE_EXTENTS:
4856         case F2FS_IOC_RESIZE_FS:
4857         case FS_IOC_ENABLE_VERITY:
4858         case FS_IOC_MEASURE_VERITY:
4859         case FS_IOC_READ_VERITY_METADATA:
4860         case FS_IOC_GETFSLABEL:
4861         case FS_IOC_SETFSLABEL:
4862         case F2FS_IOC_GET_COMPRESS_BLOCKS:
4863         case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4864         case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4865         case F2FS_IOC_SEC_TRIM_FILE:
4866         case F2FS_IOC_GET_COMPRESS_OPTION:
4867         case F2FS_IOC_SET_COMPRESS_OPTION:
4868         case F2FS_IOC_DECOMPRESS_FILE:
4869         case F2FS_IOC_COMPRESS_FILE:
4870                 break;
4871         default:
4872                 return -ENOIOCTLCMD;
4873         }
4874         return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4875 }
4876 #endif
4877
4878 const struct file_operations f2fs_file_operations = {
4879         .llseek         = f2fs_llseek,
4880         .read_iter      = f2fs_file_read_iter,
4881         .write_iter     = f2fs_file_write_iter,
4882         .open           = f2fs_file_open,
4883         .release        = f2fs_release_file,
4884         .mmap           = f2fs_file_mmap,
4885         .flush          = f2fs_file_flush,
4886         .fsync          = f2fs_sync_file,
4887         .fallocate      = f2fs_fallocate,
4888         .unlocked_ioctl = f2fs_ioctl,
4889 #ifdef CONFIG_COMPAT
4890         .compat_ioctl   = f2fs_compat_ioctl,
4891 #endif
4892         .splice_read    = generic_file_splice_read,
4893         .splice_write   = iter_file_splice_write,
4894         .fadvise        = f2fs_file_fadvise,
4895 };