1 /* 2 * fs/f2fs/file.c 3 * 4 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 5 * http://www.samsung.com/ 6 * 7 * This program is free software; you can redistribute it and/or modify 8 * it under the terms of the GNU General Public License version 2 as 9 * published by the Free Software Foundation. 10 */ 11 #include <linux/fs.h> 12 #include <linux/f2fs_fs.h> 13 #include <linux/stat.h> 14 #include <linux/buffer_head.h> 15 #include <linux/writeback.h> 16 #include <linux/blkdev.h> 17 #include <linux/falloc.h> 18 #include <linux/types.h> 19 #include <linux/compat.h> 20 #include <linux/uaccess.h> 21 #include <linux/mount.h> 22 #include <linux/pagevec.h> 23 #include <linux/uuid.h> 24 #include <linux/file.h> 25 26 #include "f2fs.h" 27 #include "node.h" 28 #include "segment.h" 29 #include "xattr.h" 30 #include "acl.h" 31 #include "gc.h" 32 #include "trace.h" 33 #include <trace/events/f2fs.h> 34 35 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma, 36 struct vm_fault *vmf) 37 { 38 struct page *page = vmf->page; 39 struct inode *inode = file_inode(vma->vm_file); 40 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 41 struct dnode_of_data dn; 42 int err; 43 44 sb_start_pagefault(inode->i_sb); 45 46 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 47 48 /* block allocation */ 49 f2fs_lock_op(sbi); 50 set_new_dnode(&dn, inode, NULL, NULL, 0); 51 err = f2fs_reserve_block(&dn, page->index); 52 if (err) { 53 f2fs_unlock_op(sbi); 54 goto out; 55 } 56 f2fs_put_dnode(&dn); 57 f2fs_unlock_op(sbi); 58 59 f2fs_balance_fs(sbi, dn.node_changed); 60 61 file_update_time(vma->vm_file); 62 lock_page(page); 63 if (unlikely(page->mapping != inode->i_mapping || 64 page_offset(page) > i_size_read(inode) || 65 !PageUptodate(page))) { 66 unlock_page(page); 67 err = -EFAULT; 68 goto out; 69 } 70 71 /* 72 * check to see if the page is mapped already (no holes) 73 */ 74 if (PageMappedToDisk(page)) 75 goto mapped; 76 77 /* page is wholly or partially inside EOF */ 78 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 79 i_size_read(inode)) { 80 unsigned offset; 81 offset = i_size_read(inode) & ~PAGE_MASK; 82 zero_user_segment(page, offset, PAGE_SIZE); 83 } 84 set_page_dirty(page); 85 if (!PageUptodate(page)) 86 SetPageUptodate(page); 87 88 trace_f2fs_vm_page_mkwrite(page, DATA); 89 mapped: 90 /* fill the page */ 91 f2fs_wait_on_page_writeback(page, DATA, false); 92 93 /* wait for GCed encrypted page writeback */ 94 if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) 95 f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr); 96 97 /* if gced page is attached, don't write to cold segment */ 98 clear_cold_data(page); 99 out: 100 sb_end_pagefault(inode->i_sb); 101 f2fs_update_time(sbi, REQ_TIME); 102 return block_page_mkwrite_return(err); 103 } 104 105 static const struct vm_operations_struct f2fs_file_vm_ops = { 106 .fault = filemap_fault, 107 .map_pages = filemap_map_pages, 108 .page_mkwrite = f2fs_vm_page_mkwrite, 109 }; 110 111 static int get_parent_ino(struct inode *inode, nid_t *pino) 112 { 113 struct dentry *dentry; 114 115 inode = igrab(inode); 116 dentry = d_find_any_alias(inode); 117 iput(inode); 118 if (!dentry) 119 return 0; 120 121 if (update_dent_inode(inode, inode, &dentry->d_name)) { 122 dput(dentry); 123 return 0; 124 } 125 126 *pino = parent_ino(dentry); 127 dput(dentry); 128 return 1; 129 } 130 131 static inline bool need_do_checkpoint(struct inode *inode) 132 { 133 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 134 bool need_cp = false; 135 136 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1) 137 need_cp = true; 138 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 139 need_cp = true; 140 else if (file_wrong_pino(inode)) 141 need_cp = true; 142 else if (!space_for_roll_forward(sbi)) 143 need_cp = true; 144 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 145 need_cp = true; 146 else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi))) 147 need_cp = true; 148 else if (test_opt(sbi, FASTBOOT)) 149 need_cp = true; 150 else if (sbi->active_logs == 2) 151 need_cp = true; 152 153 return need_cp; 154 } 155 156 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 157 { 158 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 159 bool ret = false; 160 /* But we need to avoid that there are some inode updates */ 161 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino)) 162 ret = true; 163 f2fs_put_page(i, 0); 164 return ret; 165 } 166 167 static void try_to_fix_pino(struct inode *inode) 168 { 169 struct f2fs_inode_info *fi = F2FS_I(inode); 170 nid_t pino; 171 172 down_write(&fi->i_sem); 173 fi->xattr_ver = 0; 174 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 175 get_parent_ino(inode, &pino)) { 176 f2fs_i_pino_write(inode, pino); 177 file_got_pino(inode); 178 } 179 up_write(&fi->i_sem); 180 } 181 182 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 183 int datasync, bool atomic) 184 { 185 struct inode *inode = file->f_mapping->host; 186 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 187 nid_t ino = inode->i_ino; 188 int ret = 0; 189 bool need_cp = false; 190 struct writeback_control wbc = { 191 .sync_mode = WB_SYNC_ALL, 192 .nr_to_write = LONG_MAX, 193 .for_reclaim = 0, 194 }; 195 196 if (unlikely(f2fs_readonly(inode->i_sb))) 197 return 0; 198 199 trace_f2fs_sync_file_enter(inode); 200 201 /* if fdatasync is triggered, let's do in-place-update */ 202 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 203 set_inode_flag(inode, FI_NEED_IPU); 204 ret = filemap_write_and_wait_range(inode->i_mapping, start, end); 205 clear_inode_flag(inode, FI_NEED_IPU); 206 207 if (ret) { 208 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret); 209 return ret; 210 } 211 212 /* if the inode is dirty, let's recover all the time */ 213 if (!datasync && !f2fs_skip_inode_update(inode)) { 214 f2fs_write_inode(inode, NULL); 215 goto go_write; 216 } 217 218 /* 219 * if there is no written data, don't waste time to write recovery info. 220 */ 221 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 222 !exist_written_data(sbi, ino, APPEND_INO)) { 223 224 /* it may call write_inode just prior to fsync */ 225 if (need_inode_page_update(sbi, ino)) 226 goto go_write; 227 228 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 229 exist_written_data(sbi, ino, UPDATE_INO)) 230 goto flush_out; 231 goto out; 232 } 233 go_write: 234 /* 235 * Both of fdatasync() and fsync() are able to be recovered from 236 * sudden-power-off. 237 */ 238 down_read(&F2FS_I(inode)->i_sem); 239 need_cp = need_do_checkpoint(inode); 240 up_read(&F2FS_I(inode)->i_sem); 241 242 if (need_cp) { 243 /* all the dirty node pages should be flushed for POR */ 244 ret = f2fs_sync_fs(inode->i_sb, 1); 245 246 /* 247 * We've secured consistency through sync_fs. Following pino 248 * will be used only for fsynced inodes after checkpoint. 249 */ 250 try_to_fix_pino(inode); 251 clear_inode_flag(inode, FI_APPEND_WRITE); 252 clear_inode_flag(inode, FI_UPDATE_WRITE); 253 goto out; 254 } 255 sync_nodes: 256 ret = fsync_node_pages(sbi, inode, &wbc, atomic); 257 if (ret) 258 goto out; 259 260 /* if cp_error was enabled, we should avoid infinite loop */ 261 if (unlikely(f2fs_cp_error(sbi))) { 262 ret = -EIO; 263 goto out; 264 } 265 266 if (need_inode_block_update(sbi, ino)) { 267 f2fs_mark_inode_dirty_sync(inode, true); 268 f2fs_write_inode(inode, NULL); 269 goto sync_nodes; 270 } 271 272 ret = wait_on_node_pages_writeback(sbi, ino); 273 if (ret) 274 goto out; 275 276 /* once recovery info is written, don't need to tack this */ 277 remove_ino_entry(sbi, ino, APPEND_INO); 278 clear_inode_flag(inode, FI_APPEND_WRITE); 279 flush_out: 280 remove_ino_entry(sbi, ino, UPDATE_INO); 281 clear_inode_flag(inode, FI_UPDATE_WRITE); 282 ret = f2fs_issue_flush(sbi); 283 f2fs_update_time(sbi, REQ_TIME); 284 out: 285 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret); 286 f2fs_trace_ios(NULL, 1); 287 return ret; 288 } 289 290 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 291 { 292 return f2fs_do_sync_file(file, start, end, datasync, false); 293 } 294 295 static pgoff_t __get_first_dirty_index(struct address_space *mapping, 296 pgoff_t pgofs, int whence) 297 { 298 struct pagevec pvec; 299 int nr_pages; 300 301 if (whence != SEEK_DATA) 302 return 0; 303 304 /* find first dirty page index */ 305 pagevec_init(&pvec, 0); 306 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs, 307 PAGECACHE_TAG_DIRTY, 1); 308 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX; 309 pagevec_release(&pvec); 310 return pgofs; 311 } 312 313 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs, 314 int whence) 315 { 316 switch (whence) { 317 case SEEK_DATA: 318 if ((blkaddr == NEW_ADDR && dirty == pgofs) || 319 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR)) 320 return true; 321 break; 322 case SEEK_HOLE: 323 if (blkaddr == NULL_ADDR) 324 return true; 325 break; 326 } 327 return false; 328 } 329 330 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 331 { 332 struct inode *inode = file->f_mapping->host; 333 loff_t maxbytes = inode->i_sb->s_maxbytes; 334 struct dnode_of_data dn; 335 pgoff_t pgofs, end_offset, dirty; 336 loff_t data_ofs = offset; 337 loff_t isize; 338 int err = 0; 339 340 inode_lock(inode); 341 342 isize = i_size_read(inode); 343 if (offset >= isize) 344 goto fail; 345 346 /* handle inline data case */ 347 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) { 348 if (whence == SEEK_HOLE) 349 data_ofs = isize; 350 goto found; 351 } 352 353 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 354 355 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence); 356 357 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 358 set_new_dnode(&dn, inode, NULL, NULL, 0); 359 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 360 if (err && err != -ENOENT) { 361 goto fail; 362 } else if (err == -ENOENT) { 363 /* direct node does not exists */ 364 if (whence == SEEK_DATA) { 365 pgofs = get_next_page_offset(&dn, pgofs); 366 continue; 367 } else { 368 goto found; 369 } 370 } 371 372 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 373 374 /* find data/hole in dnode block */ 375 for (; dn.ofs_in_node < end_offset; 376 dn.ofs_in_node++, pgofs++, 377 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 378 block_t blkaddr; 379 blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); 380 381 if (__found_offset(blkaddr, dirty, pgofs, whence)) { 382 f2fs_put_dnode(&dn); 383 goto found; 384 } 385 } 386 f2fs_put_dnode(&dn); 387 } 388 389 if (whence == SEEK_DATA) 390 goto fail; 391 found: 392 if (whence == SEEK_HOLE && data_ofs > isize) 393 data_ofs = isize; 394 inode_unlock(inode); 395 return vfs_setpos(file, data_ofs, maxbytes); 396 fail: 397 inode_unlock(inode); 398 return -ENXIO; 399 } 400 401 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 402 { 403 struct inode *inode = file->f_mapping->host; 404 loff_t maxbytes = inode->i_sb->s_maxbytes; 405 406 switch (whence) { 407 case SEEK_SET: 408 case SEEK_CUR: 409 case SEEK_END: 410 return generic_file_llseek_size(file, offset, whence, 411 maxbytes, i_size_read(inode)); 412 case SEEK_DATA: 413 case SEEK_HOLE: 414 if (offset < 0) 415 return -ENXIO; 416 return f2fs_seek_block(file, offset, whence); 417 } 418 419 return -EINVAL; 420 } 421 422 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 423 { 424 struct inode *inode = file_inode(file); 425 int err; 426 427 if (f2fs_encrypted_inode(inode)) { 428 err = fscrypt_get_encryption_info(inode); 429 if (err) 430 return 0; 431 if (!f2fs_encrypted_inode(inode)) 432 return -ENOKEY; 433 } 434 435 /* we don't need to use inline_data strictly */ 436 err = f2fs_convert_inline_inode(inode); 437 if (err) 438 return err; 439 440 file_accessed(file); 441 vma->vm_ops = &f2fs_file_vm_ops; 442 return 0; 443 } 444 445 static int f2fs_file_open(struct inode *inode, struct file *filp) 446 { 447 int ret = generic_file_open(inode, filp); 448 struct dentry *dir; 449 450 if (!ret && f2fs_encrypted_inode(inode)) { 451 ret = fscrypt_get_encryption_info(inode); 452 if (ret) 453 return -EACCES; 454 if (!fscrypt_has_encryption_key(inode)) 455 return -ENOKEY; 456 } 457 dir = dget_parent(file_dentry(filp)); 458 if (f2fs_encrypted_inode(d_inode(dir)) && 459 !fscrypt_has_permitted_context(d_inode(dir), inode)) { 460 dput(dir); 461 return -EPERM; 462 } 463 dput(dir); 464 return ret; 465 } 466 467 int truncate_data_blocks_range(struct dnode_of_data *dn, int count) 468 { 469 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 470 struct f2fs_node *raw_node; 471 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 472 __le32 *addr; 473 474 raw_node = F2FS_NODE(dn->node_page); 475 addr = blkaddr_in_node(raw_node) + ofs; 476 477 for (; count > 0; count--, addr++, dn->ofs_in_node++) { 478 block_t blkaddr = le32_to_cpu(*addr); 479 if (blkaddr == NULL_ADDR) 480 continue; 481 482 dn->data_blkaddr = NULL_ADDR; 483 set_data_blkaddr(dn); 484 invalidate_blocks(sbi, blkaddr); 485 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 486 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 487 nr_free++; 488 } 489 490 if (nr_free) { 491 pgoff_t fofs; 492 /* 493 * once we invalidate valid blkaddr in range [ofs, ofs + count], 494 * we will invalidate all blkaddr in the whole range. 495 */ 496 fofs = start_bidx_of_node(ofs_of_node(dn->node_page), 497 dn->inode) + ofs; 498 f2fs_update_extent_cache_range(dn, fofs, 0, len); 499 dec_valid_block_count(sbi, dn->inode, nr_free); 500 } 501 dn->ofs_in_node = ofs; 502 503 f2fs_update_time(sbi, REQ_TIME); 504 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 505 dn->ofs_in_node, nr_free); 506 return nr_free; 507 } 508 509 void truncate_data_blocks(struct dnode_of_data *dn) 510 { 511 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK); 512 } 513 514 static int truncate_partial_data_page(struct inode *inode, u64 from, 515 bool cache_only) 516 { 517 unsigned offset = from & (PAGE_SIZE - 1); 518 pgoff_t index = from >> PAGE_SHIFT; 519 struct address_space *mapping = inode->i_mapping; 520 struct page *page; 521 522 if (!offset && !cache_only) 523 return 0; 524 525 if (cache_only) { 526 page = find_lock_page(mapping, index); 527 if (page && PageUptodate(page)) 528 goto truncate_out; 529 f2fs_put_page(page, 1); 530 return 0; 531 } 532 533 page = get_lock_data_page(inode, index, true); 534 if (IS_ERR(page)) 535 return 0; 536 truncate_out: 537 f2fs_wait_on_page_writeback(page, DATA, true); 538 zero_user(page, offset, PAGE_SIZE - offset); 539 if (!cache_only || !f2fs_encrypted_inode(inode) || 540 !S_ISREG(inode->i_mode)) 541 set_page_dirty(page); 542 f2fs_put_page(page, 1); 543 return 0; 544 } 545 546 int truncate_blocks(struct inode *inode, u64 from, bool lock) 547 { 548 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 549 unsigned int blocksize = inode->i_sb->s_blocksize; 550 struct dnode_of_data dn; 551 pgoff_t free_from; 552 int count = 0, err = 0; 553 struct page *ipage; 554 bool truncate_page = false; 555 556 trace_f2fs_truncate_blocks_enter(inode, from); 557 558 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1); 559 560 if (free_from >= sbi->max_file_blocks) 561 goto free_partial; 562 563 if (lock) 564 f2fs_lock_op(sbi); 565 566 ipage = get_node_page(sbi, inode->i_ino); 567 if (IS_ERR(ipage)) { 568 err = PTR_ERR(ipage); 569 goto out; 570 } 571 572 if (f2fs_has_inline_data(inode)) { 573 if (truncate_inline_inode(ipage, from)) 574 set_page_dirty(ipage); 575 f2fs_put_page(ipage, 1); 576 truncate_page = true; 577 goto out; 578 } 579 580 set_new_dnode(&dn, inode, ipage, NULL, 0); 581 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 582 if (err) { 583 if (err == -ENOENT) 584 goto free_next; 585 goto out; 586 } 587 588 count = ADDRS_PER_PAGE(dn.node_page, inode); 589 590 count -= dn.ofs_in_node; 591 f2fs_bug_on(sbi, count < 0); 592 593 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 594 truncate_data_blocks_range(&dn, count); 595 free_from += count; 596 } 597 598 f2fs_put_dnode(&dn); 599 free_next: 600 err = truncate_inode_blocks(inode, free_from); 601 out: 602 if (lock) 603 f2fs_unlock_op(sbi); 604 free_partial: 605 /* lastly zero out the first data page */ 606 if (!err) 607 err = truncate_partial_data_page(inode, from, truncate_page); 608 609 trace_f2fs_truncate_blocks_exit(inode, err); 610 return err; 611 } 612 613 int f2fs_truncate(struct inode *inode) 614 { 615 int err; 616 617 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 618 S_ISLNK(inode->i_mode))) 619 return 0; 620 621 trace_f2fs_truncate(inode); 622 623 /* we should check inline_data size */ 624 if (!f2fs_may_inline_data(inode)) { 625 err = f2fs_convert_inline_inode(inode); 626 if (err) 627 return err; 628 } 629 630 err = truncate_blocks(inode, i_size_read(inode), true); 631 if (err) 632 return err; 633 634 inode->i_mtime = inode->i_ctime = current_time(inode); 635 f2fs_mark_inode_dirty_sync(inode, false); 636 return 0; 637 } 638 639 int f2fs_getattr(struct vfsmount *mnt, 640 struct dentry *dentry, struct kstat *stat) 641 { 642 struct inode *inode = d_inode(dentry); 643 generic_fillattr(inode, stat); 644 stat->blocks <<= 3; 645 return 0; 646 } 647 648 #ifdef CONFIG_F2FS_FS_POSIX_ACL 649 static void __setattr_copy(struct inode *inode, const struct iattr *attr) 650 { 651 unsigned int ia_valid = attr->ia_valid; 652 653 if (ia_valid & ATTR_UID) 654 inode->i_uid = attr->ia_uid; 655 if (ia_valid & ATTR_GID) 656 inode->i_gid = attr->ia_gid; 657 if (ia_valid & ATTR_ATIME) 658 inode->i_atime = timespec_trunc(attr->ia_atime, 659 inode->i_sb->s_time_gran); 660 if (ia_valid & ATTR_MTIME) 661 inode->i_mtime = timespec_trunc(attr->ia_mtime, 662 inode->i_sb->s_time_gran); 663 if (ia_valid & ATTR_CTIME) 664 inode->i_ctime = timespec_trunc(attr->ia_ctime, 665 inode->i_sb->s_time_gran); 666 if (ia_valid & ATTR_MODE) { 667 umode_t mode = attr->ia_mode; 668 669 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) 670 mode &= ~S_ISGID; 671 set_acl_inode(inode, mode); 672 } 673 } 674 #else 675 #define __setattr_copy setattr_copy 676 #endif 677 678 int f2fs_setattr(struct dentry *dentry, struct iattr *attr) 679 { 680 struct inode *inode = d_inode(dentry); 681 int err; 682 683 err = setattr_prepare(dentry, attr); 684 if (err) 685 return err; 686 687 if (attr->ia_valid & ATTR_SIZE) { 688 if (f2fs_encrypted_inode(inode) && 689 fscrypt_get_encryption_info(inode)) 690 return -EACCES; 691 692 if (attr->ia_size <= i_size_read(inode)) { 693 truncate_setsize(inode, attr->ia_size); 694 err = f2fs_truncate(inode); 695 if (err) 696 return err; 697 } else { 698 /* 699 * do not trim all blocks after i_size if target size is 700 * larger than i_size. 701 */ 702 truncate_setsize(inode, attr->ia_size); 703 704 /* should convert inline inode here */ 705 if (!f2fs_may_inline_data(inode)) { 706 err = f2fs_convert_inline_inode(inode); 707 if (err) 708 return err; 709 } 710 inode->i_mtime = inode->i_ctime = current_time(inode); 711 } 712 } 713 714 __setattr_copy(inode, attr); 715 716 if (attr->ia_valid & ATTR_MODE) { 717 err = posix_acl_chmod(inode, get_inode_mode(inode)); 718 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) { 719 inode->i_mode = F2FS_I(inode)->i_acl_mode; 720 clear_inode_flag(inode, FI_ACL_MODE); 721 } 722 } 723 724 /* update attributes only */ 725 f2fs_mark_inode_dirty_sync(inode, false); 726 727 /* inode change will produce dirty node pages flushed by checkpoint */ 728 f2fs_balance_fs(F2FS_I_SB(inode), true); 729 730 return err; 731 } 732 733 const struct inode_operations f2fs_file_inode_operations = { 734 .getattr = f2fs_getattr, 735 .setattr = f2fs_setattr, 736 .get_acl = f2fs_get_acl, 737 .set_acl = f2fs_set_acl, 738 #ifdef CONFIG_F2FS_FS_XATTR 739 .listxattr = f2fs_listxattr, 740 #endif 741 .fiemap = f2fs_fiemap, 742 }; 743 744 static int fill_zero(struct inode *inode, pgoff_t index, 745 loff_t start, loff_t len) 746 { 747 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 748 struct page *page; 749 750 if (!len) 751 return 0; 752 753 f2fs_balance_fs(sbi, true); 754 755 f2fs_lock_op(sbi); 756 page = get_new_data_page(inode, NULL, index, false); 757 f2fs_unlock_op(sbi); 758 759 if (IS_ERR(page)) 760 return PTR_ERR(page); 761 762 f2fs_wait_on_page_writeback(page, DATA, true); 763 zero_user(page, start, len); 764 set_page_dirty(page); 765 f2fs_put_page(page, 1); 766 return 0; 767 } 768 769 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 770 { 771 int err; 772 773 while (pg_start < pg_end) { 774 struct dnode_of_data dn; 775 pgoff_t end_offset, count; 776 777 set_new_dnode(&dn, inode, NULL, NULL, 0); 778 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 779 if (err) { 780 if (err == -ENOENT) { 781 pg_start++; 782 continue; 783 } 784 return err; 785 } 786 787 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 788 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 789 790 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 791 792 truncate_data_blocks_range(&dn, count); 793 f2fs_put_dnode(&dn); 794 795 pg_start += count; 796 } 797 return 0; 798 } 799 800 static int punch_hole(struct inode *inode, loff_t offset, loff_t len) 801 { 802 pgoff_t pg_start, pg_end; 803 loff_t off_start, off_end; 804 int ret; 805 806 ret = f2fs_convert_inline_inode(inode); 807 if (ret) 808 return ret; 809 810 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 811 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 812 813 off_start = offset & (PAGE_SIZE - 1); 814 off_end = (offset + len) & (PAGE_SIZE - 1); 815 816 if (pg_start == pg_end) { 817 ret = fill_zero(inode, pg_start, off_start, 818 off_end - off_start); 819 if (ret) 820 return ret; 821 } else { 822 if (off_start) { 823 ret = fill_zero(inode, pg_start++, off_start, 824 PAGE_SIZE - off_start); 825 if (ret) 826 return ret; 827 } 828 if (off_end) { 829 ret = fill_zero(inode, pg_end, 0, off_end); 830 if (ret) 831 return ret; 832 } 833 834 if (pg_start < pg_end) { 835 struct address_space *mapping = inode->i_mapping; 836 loff_t blk_start, blk_end; 837 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 838 839 f2fs_balance_fs(sbi, true); 840 841 blk_start = (loff_t)pg_start << PAGE_SHIFT; 842 blk_end = (loff_t)pg_end << PAGE_SHIFT; 843 truncate_inode_pages_range(mapping, blk_start, 844 blk_end - 1); 845 846 f2fs_lock_op(sbi); 847 ret = truncate_hole(inode, pg_start, pg_end); 848 f2fs_unlock_op(sbi); 849 } 850 } 851 852 return ret; 853 } 854 855 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 856 int *do_replace, pgoff_t off, pgoff_t len) 857 { 858 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 859 struct dnode_of_data dn; 860 int ret, done, i; 861 862 next_dnode: 863 set_new_dnode(&dn, inode, NULL, NULL, 0); 864 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 865 if (ret && ret != -ENOENT) { 866 return ret; 867 } else if (ret == -ENOENT) { 868 if (dn.max_level == 0) 869 return -ENOENT; 870 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len); 871 blkaddr += done; 872 do_replace += done; 873 goto next; 874 } 875 876 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 877 dn.ofs_in_node, len); 878 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 879 *blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); 880 if (!is_checkpointed_data(sbi, *blkaddr)) { 881 882 if (test_opt(sbi, LFS)) { 883 f2fs_put_dnode(&dn); 884 return -ENOTSUPP; 885 } 886 887 /* do not invalidate this block address */ 888 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 889 *do_replace = 1; 890 } 891 } 892 f2fs_put_dnode(&dn); 893 next: 894 len -= done; 895 off += done; 896 if (len) 897 goto next_dnode; 898 return 0; 899 } 900 901 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 902 int *do_replace, pgoff_t off, int len) 903 { 904 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 905 struct dnode_of_data dn; 906 int ret, i; 907 908 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 909 if (*do_replace == 0) 910 continue; 911 912 set_new_dnode(&dn, inode, NULL, NULL, 0); 913 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 914 if (ret) { 915 dec_valid_block_count(sbi, inode, 1); 916 invalidate_blocks(sbi, *blkaddr); 917 } else { 918 f2fs_update_data_blkaddr(&dn, *blkaddr); 919 } 920 f2fs_put_dnode(&dn); 921 } 922 return 0; 923 } 924 925 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 926 block_t *blkaddr, int *do_replace, 927 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 928 { 929 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 930 pgoff_t i = 0; 931 int ret; 932 933 while (i < len) { 934 if (blkaddr[i] == NULL_ADDR && !full) { 935 i++; 936 continue; 937 } 938 939 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 940 struct dnode_of_data dn; 941 struct node_info ni; 942 size_t new_size; 943 pgoff_t ilen; 944 945 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 946 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 947 if (ret) 948 return ret; 949 950 get_node_info(sbi, dn.nid, &ni); 951 ilen = min((pgoff_t) 952 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 953 dn.ofs_in_node, len - i); 954 do { 955 dn.data_blkaddr = datablock_addr(dn.node_page, 956 dn.ofs_in_node); 957 truncate_data_blocks_range(&dn, 1); 958 959 if (do_replace[i]) { 960 f2fs_i_blocks_write(src_inode, 961 1, false); 962 f2fs_i_blocks_write(dst_inode, 963 1, true); 964 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 965 blkaddr[i], ni.version, true, false); 966 967 do_replace[i] = 0; 968 } 969 dn.ofs_in_node++; 970 i++; 971 new_size = (dst + i) << PAGE_SHIFT; 972 if (dst_inode->i_size < new_size) 973 f2fs_i_size_write(dst_inode, new_size); 974 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 975 976 f2fs_put_dnode(&dn); 977 } else { 978 struct page *psrc, *pdst; 979 980 psrc = get_lock_data_page(src_inode, src + i, true); 981 if (IS_ERR(psrc)) 982 return PTR_ERR(psrc); 983 pdst = get_new_data_page(dst_inode, NULL, dst + i, 984 true); 985 if (IS_ERR(pdst)) { 986 f2fs_put_page(psrc, 1); 987 return PTR_ERR(pdst); 988 } 989 f2fs_copy_page(psrc, pdst); 990 set_page_dirty(pdst); 991 f2fs_put_page(pdst, 1); 992 f2fs_put_page(psrc, 1); 993 994 ret = truncate_hole(src_inode, src + i, src + i + 1); 995 if (ret) 996 return ret; 997 i++; 998 } 999 } 1000 return 0; 1001 } 1002 1003 static int __exchange_data_block(struct inode *src_inode, 1004 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1005 pgoff_t len, bool full) 1006 { 1007 block_t *src_blkaddr; 1008 int *do_replace; 1009 pgoff_t olen; 1010 int ret; 1011 1012 while (len) { 1013 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len); 1014 1015 src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL); 1016 if (!src_blkaddr) 1017 return -ENOMEM; 1018 1019 do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL); 1020 if (!do_replace) { 1021 kvfree(src_blkaddr); 1022 return -ENOMEM; 1023 } 1024 1025 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1026 do_replace, src, olen); 1027 if (ret) 1028 goto roll_back; 1029 1030 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1031 do_replace, src, dst, olen, full); 1032 if (ret) 1033 goto roll_back; 1034 1035 src += olen; 1036 dst += olen; 1037 len -= olen; 1038 1039 kvfree(src_blkaddr); 1040 kvfree(do_replace); 1041 } 1042 return 0; 1043 1044 roll_back: 1045 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len); 1046 kvfree(src_blkaddr); 1047 kvfree(do_replace); 1048 return ret; 1049 } 1050 1051 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end) 1052 { 1053 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1054 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1055 int ret; 1056 1057 f2fs_balance_fs(sbi, true); 1058 f2fs_lock_op(sbi); 1059 1060 f2fs_drop_extent_tree(inode); 1061 1062 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1063 f2fs_unlock_op(sbi); 1064 return ret; 1065 } 1066 1067 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1068 { 1069 pgoff_t pg_start, pg_end; 1070 loff_t new_size; 1071 int ret; 1072 1073 if (offset + len >= i_size_read(inode)) 1074 return -EINVAL; 1075 1076 /* collapse range should be aligned to block size of f2fs. */ 1077 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1078 return -EINVAL; 1079 1080 ret = f2fs_convert_inline_inode(inode); 1081 if (ret) 1082 return ret; 1083 1084 pg_start = offset >> PAGE_SHIFT; 1085 pg_end = (offset + len) >> PAGE_SHIFT; 1086 1087 /* write out all dirty pages from offset */ 1088 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1089 if (ret) 1090 return ret; 1091 1092 truncate_pagecache(inode, offset); 1093 1094 ret = f2fs_do_collapse(inode, pg_start, pg_end); 1095 if (ret) 1096 return ret; 1097 1098 /* write out all moved pages, if possible */ 1099 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1100 truncate_pagecache(inode, offset); 1101 1102 new_size = i_size_read(inode) - len; 1103 truncate_pagecache(inode, new_size); 1104 1105 ret = truncate_blocks(inode, new_size, true); 1106 if (!ret) 1107 f2fs_i_size_write(inode, new_size); 1108 1109 return ret; 1110 } 1111 1112 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1113 pgoff_t end) 1114 { 1115 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1116 pgoff_t index = start; 1117 unsigned int ofs_in_node = dn->ofs_in_node; 1118 blkcnt_t count = 0; 1119 int ret; 1120 1121 for (; index < end; index++, dn->ofs_in_node++) { 1122 if (datablock_addr(dn->node_page, dn->ofs_in_node) == NULL_ADDR) 1123 count++; 1124 } 1125 1126 dn->ofs_in_node = ofs_in_node; 1127 ret = reserve_new_blocks(dn, count); 1128 if (ret) 1129 return ret; 1130 1131 dn->ofs_in_node = ofs_in_node; 1132 for (index = start; index < end; index++, dn->ofs_in_node++) { 1133 dn->data_blkaddr = 1134 datablock_addr(dn->node_page, dn->ofs_in_node); 1135 /* 1136 * reserve_new_blocks will not guarantee entire block 1137 * allocation. 1138 */ 1139 if (dn->data_blkaddr == NULL_ADDR) { 1140 ret = -ENOSPC; 1141 break; 1142 } 1143 if (dn->data_blkaddr != NEW_ADDR) { 1144 invalidate_blocks(sbi, dn->data_blkaddr); 1145 dn->data_blkaddr = NEW_ADDR; 1146 set_data_blkaddr(dn); 1147 } 1148 } 1149 1150 f2fs_update_extent_cache_range(dn, start, 0, index - start); 1151 1152 return ret; 1153 } 1154 1155 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1156 int mode) 1157 { 1158 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1159 struct address_space *mapping = inode->i_mapping; 1160 pgoff_t index, pg_start, pg_end; 1161 loff_t new_size = i_size_read(inode); 1162 loff_t off_start, off_end; 1163 int ret = 0; 1164 1165 ret = inode_newsize_ok(inode, (len + offset)); 1166 if (ret) 1167 return ret; 1168 1169 ret = f2fs_convert_inline_inode(inode); 1170 if (ret) 1171 return ret; 1172 1173 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1174 if (ret) 1175 return ret; 1176 1177 truncate_pagecache_range(inode, offset, offset + len - 1); 1178 1179 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1180 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1181 1182 off_start = offset & (PAGE_SIZE - 1); 1183 off_end = (offset + len) & (PAGE_SIZE - 1); 1184 1185 if (pg_start == pg_end) { 1186 ret = fill_zero(inode, pg_start, off_start, 1187 off_end - off_start); 1188 if (ret) 1189 return ret; 1190 1191 if (offset + len > new_size) 1192 new_size = offset + len; 1193 new_size = max_t(loff_t, new_size, offset + len); 1194 } else { 1195 if (off_start) { 1196 ret = fill_zero(inode, pg_start++, off_start, 1197 PAGE_SIZE - off_start); 1198 if (ret) 1199 return ret; 1200 1201 new_size = max_t(loff_t, new_size, 1202 (loff_t)pg_start << PAGE_SHIFT); 1203 } 1204 1205 for (index = pg_start; index < pg_end;) { 1206 struct dnode_of_data dn; 1207 unsigned int end_offset; 1208 pgoff_t end; 1209 1210 f2fs_lock_op(sbi); 1211 1212 set_new_dnode(&dn, inode, NULL, NULL, 0); 1213 ret = get_dnode_of_data(&dn, index, ALLOC_NODE); 1214 if (ret) { 1215 f2fs_unlock_op(sbi); 1216 goto out; 1217 } 1218 1219 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1220 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1221 1222 ret = f2fs_do_zero_range(&dn, index, end); 1223 f2fs_put_dnode(&dn); 1224 f2fs_unlock_op(sbi); 1225 1226 f2fs_balance_fs(sbi, dn.node_changed); 1227 1228 if (ret) 1229 goto out; 1230 1231 index = end; 1232 new_size = max_t(loff_t, new_size, 1233 (loff_t)index << PAGE_SHIFT); 1234 } 1235 1236 if (off_end) { 1237 ret = fill_zero(inode, pg_end, 0, off_end); 1238 if (ret) 1239 goto out; 1240 1241 new_size = max_t(loff_t, new_size, offset + len); 1242 } 1243 } 1244 1245 out: 1246 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) 1247 f2fs_i_size_write(inode, new_size); 1248 1249 return ret; 1250 } 1251 1252 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1253 { 1254 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1255 pgoff_t nr, pg_start, pg_end, delta, idx; 1256 loff_t new_size; 1257 int ret = 0; 1258 1259 new_size = i_size_read(inode) + len; 1260 if (new_size > inode->i_sb->s_maxbytes) 1261 return -EFBIG; 1262 1263 if (offset >= i_size_read(inode)) 1264 return -EINVAL; 1265 1266 /* insert range should be aligned to block size of f2fs. */ 1267 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1268 return -EINVAL; 1269 1270 ret = f2fs_convert_inline_inode(inode); 1271 if (ret) 1272 return ret; 1273 1274 f2fs_balance_fs(sbi, true); 1275 1276 ret = truncate_blocks(inode, i_size_read(inode), true); 1277 if (ret) 1278 return ret; 1279 1280 /* write out all dirty pages from offset */ 1281 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1282 if (ret) 1283 return ret; 1284 1285 truncate_pagecache(inode, offset); 1286 1287 pg_start = offset >> PAGE_SHIFT; 1288 pg_end = (offset + len) >> PAGE_SHIFT; 1289 delta = pg_end - pg_start; 1290 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1291 1292 while (!ret && idx > pg_start) { 1293 nr = idx - pg_start; 1294 if (nr > delta) 1295 nr = delta; 1296 idx -= nr; 1297 1298 f2fs_lock_op(sbi); 1299 f2fs_drop_extent_tree(inode); 1300 1301 ret = __exchange_data_block(inode, inode, idx, 1302 idx + delta, nr, false); 1303 f2fs_unlock_op(sbi); 1304 } 1305 1306 /* write out all moved pages, if possible */ 1307 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1308 truncate_pagecache(inode, offset); 1309 1310 if (!ret) 1311 f2fs_i_size_write(inode, new_size); 1312 return ret; 1313 } 1314 1315 static int expand_inode_data(struct inode *inode, loff_t offset, 1316 loff_t len, int mode) 1317 { 1318 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1319 struct f2fs_map_blocks map = { .m_next_pgofs = NULL }; 1320 pgoff_t pg_end; 1321 loff_t new_size = i_size_read(inode); 1322 loff_t off_end; 1323 int ret; 1324 1325 ret = inode_newsize_ok(inode, (len + offset)); 1326 if (ret) 1327 return ret; 1328 1329 ret = f2fs_convert_inline_inode(inode); 1330 if (ret) 1331 return ret; 1332 1333 f2fs_balance_fs(sbi, true); 1334 1335 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1336 off_end = (offset + len) & (PAGE_SIZE - 1); 1337 1338 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT; 1339 map.m_len = pg_end - map.m_lblk; 1340 if (off_end) 1341 map.m_len++; 1342 1343 ret = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); 1344 if (ret) { 1345 pgoff_t last_off; 1346 1347 if (!map.m_len) 1348 return ret; 1349 1350 last_off = map.m_lblk + map.m_len - 1; 1351 1352 /* update new size to the failed position */ 1353 new_size = (last_off == pg_end) ? offset + len: 1354 (loff_t)(last_off + 1) << PAGE_SHIFT; 1355 } else { 1356 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1357 } 1358 1359 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) 1360 f2fs_i_size_write(inode, new_size); 1361 1362 return ret; 1363 } 1364 1365 static long f2fs_fallocate(struct file *file, int mode, 1366 loff_t offset, loff_t len) 1367 { 1368 struct inode *inode = file_inode(file); 1369 long ret = 0; 1370 1371 /* f2fs only support ->fallocate for regular file */ 1372 if (!S_ISREG(inode->i_mode)) 1373 return -EINVAL; 1374 1375 if (f2fs_encrypted_inode(inode) && 1376 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1377 return -EOPNOTSUPP; 1378 1379 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1380 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1381 FALLOC_FL_INSERT_RANGE)) 1382 return -EOPNOTSUPP; 1383 1384 inode_lock(inode); 1385 1386 if (mode & FALLOC_FL_PUNCH_HOLE) { 1387 if (offset >= inode->i_size) 1388 goto out; 1389 1390 ret = punch_hole(inode, offset, len); 1391 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1392 ret = f2fs_collapse_range(inode, offset, len); 1393 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1394 ret = f2fs_zero_range(inode, offset, len, mode); 1395 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1396 ret = f2fs_insert_range(inode, offset, len); 1397 } else { 1398 ret = expand_inode_data(inode, offset, len, mode); 1399 } 1400 1401 if (!ret) { 1402 inode->i_mtime = inode->i_ctime = current_time(inode); 1403 f2fs_mark_inode_dirty_sync(inode, false); 1404 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1405 } 1406 1407 out: 1408 inode_unlock(inode); 1409 1410 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1411 return ret; 1412 } 1413 1414 static int f2fs_release_file(struct inode *inode, struct file *filp) 1415 { 1416 /* 1417 * f2fs_relase_file is called at every close calls. So we should 1418 * not drop any inmemory pages by close called by other process. 1419 */ 1420 if (!(filp->f_mode & FMODE_WRITE) || 1421 atomic_read(&inode->i_writecount) != 1) 1422 return 0; 1423 1424 /* some remained atomic pages should discarded */ 1425 if (f2fs_is_atomic_file(inode)) 1426 drop_inmem_pages(inode); 1427 if (f2fs_is_volatile_file(inode)) { 1428 clear_inode_flag(inode, FI_VOLATILE_FILE); 1429 set_inode_flag(inode, FI_DROP_CACHE); 1430 filemap_fdatawrite(inode->i_mapping); 1431 clear_inode_flag(inode, FI_DROP_CACHE); 1432 } 1433 return 0; 1434 } 1435 1436 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL)) 1437 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL) 1438 1439 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags) 1440 { 1441 if (S_ISDIR(mode)) 1442 return flags; 1443 else if (S_ISREG(mode)) 1444 return flags & F2FS_REG_FLMASK; 1445 else 1446 return flags & F2FS_OTHER_FLMASK; 1447 } 1448 1449 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg) 1450 { 1451 struct inode *inode = file_inode(filp); 1452 struct f2fs_inode_info *fi = F2FS_I(inode); 1453 unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE; 1454 return put_user(flags, (int __user *)arg); 1455 } 1456 1457 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg) 1458 { 1459 struct inode *inode = file_inode(filp); 1460 struct f2fs_inode_info *fi = F2FS_I(inode); 1461 unsigned int flags; 1462 unsigned int oldflags; 1463 int ret; 1464 1465 if (!inode_owner_or_capable(inode)) 1466 return -EACCES; 1467 1468 if (get_user(flags, (int __user *)arg)) 1469 return -EFAULT; 1470 1471 ret = mnt_want_write_file(filp); 1472 if (ret) 1473 return ret; 1474 1475 flags = f2fs_mask_flags(inode->i_mode, flags); 1476 1477 inode_lock(inode); 1478 1479 oldflags = fi->i_flags; 1480 1481 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 1482 if (!capable(CAP_LINUX_IMMUTABLE)) { 1483 inode_unlock(inode); 1484 ret = -EPERM; 1485 goto out; 1486 } 1487 } 1488 1489 flags = flags & FS_FL_USER_MODIFIABLE; 1490 flags |= oldflags & ~FS_FL_USER_MODIFIABLE; 1491 fi->i_flags = flags; 1492 inode_unlock(inode); 1493 1494 inode->i_ctime = current_time(inode); 1495 f2fs_set_inode_flags(inode); 1496 out: 1497 mnt_drop_write_file(filp); 1498 return ret; 1499 } 1500 1501 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1502 { 1503 struct inode *inode = file_inode(filp); 1504 1505 return put_user(inode->i_generation, (int __user *)arg); 1506 } 1507 1508 static int f2fs_ioc_start_atomic_write(struct file *filp) 1509 { 1510 struct inode *inode = file_inode(filp); 1511 int ret; 1512 1513 if (!inode_owner_or_capable(inode)) 1514 return -EACCES; 1515 1516 ret = mnt_want_write_file(filp); 1517 if (ret) 1518 return ret; 1519 1520 inode_lock(inode); 1521 1522 if (f2fs_is_atomic_file(inode)) 1523 goto out; 1524 1525 ret = f2fs_convert_inline_inode(inode); 1526 if (ret) 1527 goto out; 1528 1529 set_inode_flag(inode, FI_ATOMIC_FILE); 1530 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1531 1532 if (!get_dirty_pages(inode)) 1533 goto out; 1534 1535 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING, 1536 "Unexpected flush for atomic writes: ino=%lu, npages=%lld", 1537 inode->i_ino, get_dirty_pages(inode)); 1538 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 1539 if (ret) 1540 clear_inode_flag(inode, FI_ATOMIC_FILE); 1541 out: 1542 inode_unlock(inode); 1543 mnt_drop_write_file(filp); 1544 return ret; 1545 } 1546 1547 static int f2fs_ioc_commit_atomic_write(struct file *filp) 1548 { 1549 struct inode *inode = file_inode(filp); 1550 int ret; 1551 1552 if (!inode_owner_or_capable(inode)) 1553 return -EACCES; 1554 1555 ret = mnt_want_write_file(filp); 1556 if (ret) 1557 return ret; 1558 1559 inode_lock(inode); 1560 1561 if (f2fs_is_volatile_file(inode)) 1562 goto err_out; 1563 1564 if (f2fs_is_atomic_file(inode)) { 1565 clear_inode_flag(inode, FI_ATOMIC_FILE); 1566 ret = commit_inmem_pages(inode); 1567 if (ret) { 1568 set_inode_flag(inode, FI_ATOMIC_FILE); 1569 goto err_out; 1570 } 1571 } 1572 1573 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1574 err_out: 1575 inode_unlock(inode); 1576 mnt_drop_write_file(filp); 1577 return ret; 1578 } 1579 1580 static int f2fs_ioc_start_volatile_write(struct file *filp) 1581 { 1582 struct inode *inode = file_inode(filp); 1583 int ret; 1584 1585 if (!inode_owner_or_capable(inode)) 1586 return -EACCES; 1587 1588 ret = mnt_want_write_file(filp); 1589 if (ret) 1590 return ret; 1591 1592 inode_lock(inode); 1593 1594 if (f2fs_is_volatile_file(inode)) 1595 goto out; 1596 1597 ret = f2fs_convert_inline_inode(inode); 1598 if (ret) 1599 goto out; 1600 1601 set_inode_flag(inode, FI_VOLATILE_FILE); 1602 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1603 out: 1604 inode_unlock(inode); 1605 mnt_drop_write_file(filp); 1606 return ret; 1607 } 1608 1609 static int f2fs_ioc_release_volatile_write(struct file *filp) 1610 { 1611 struct inode *inode = file_inode(filp); 1612 int ret; 1613 1614 if (!inode_owner_or_capable(inode)) 1615 return -EACCES; 1616 1617 ret = mnt_want_write_file(filp); 1618 if (ret) 1619 return ret; 1620 1621 inode_lock(inode); 1622 1623 if (!f2fs_is_volatile_file(inode)) 1624 goto out; 1625 1626 if (!f2fs_is_first_block_written(inode)) { 1627 ret = truncate_partial_data_page(inode, 0, true); 1628 goto out; 1629 } 1630 1631 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 1632 out: 1633 inode_unlock(inode); 1634 mnt_drop_write_file(filp); 1635 return ret; 1636 } 1637 1638 static int f2fs_ioc_abort_volatile_write(struct file *filp) 1639 { 1640 struct inode *inode = file_inode(filp); 1641 int ret; 1642 1643 if (!inode_owner_or_capable(inode)) 1644 return -EACCES; 1645 1646 ret = mnt_want_write_file(filp); 1647 if (ret) 1648 return ret; 1649 1650 inode_lock(inode); 1651 1652 if (f2fs_is_atomic_file(inode)) 1653 drop_inmem_pages(inode); 1654 if (f2fs_is_volatile_file(inode)) { 1655 clear_inode_flag(inode, FI_VOLATILE_FILE); 1656 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1657 } 1658 1659 inode_unlock(inode); 1660 1661 mnt_drop_write_file(filp); 1662 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1663 return ret; 1664 } 1665 1666 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 1667 { 1668 struct inode *inode = file_inode(filp); 1669 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1670 struct super_block *sb = sbi->sb; 1671 __u32 in; 1672 int ret; 1673 1674 if (!capable(CAP_SYS_ADMIN)) 1675 return -EPERM; 1676 1677 if (get_user(in, (__u32 __user *)arg)) 1678 return -EFAULT; 1679 1680 ret = mnt_want_write_file(filp); 1681 if (ret) 1682 return ret; 1683 1684 switch (in) { 1685 case F2FS_GOING_DOWN_FULLSYNC: 1686 sb = freeze_bdev(sb->s_bdev); 1687 if (sb && !IS_ERR(sb)) { 1688 f2fs_stop_checkpoint(sbi, false); 1689 thaw_bdev(sb->s_bdev, sb); 1690 } 1691 break; 1692 case F2FS_GOING_DOWN_METASYNC: 1693 /* do checkpoint only */ 1694 f2fs_sync_fs(sb, 1); 1695 f2fs_stop_checkpoint(sbi, false); 1696 break; 1697 case F2FS_GOING_DOWN_NOSYNC: 1698 f2fs_stop_checkpoint(sbi, false); 1699 break; 1700 case F2FS_GOING_DOWN_METAFLUSH: 1701 sync_meta_pages(sbi, META, LONG_MAX); 1702 f2fs_stop_checkpoint(sbi, false); 1703 break; 1704 default: 1705 ret = -EINVAL; 1706 goto out; 1707 } 1708 f2fs_update_time(sbi, REQ_TIME); 1709 out: 1710 mnt_drop_write_file(filp); 1711 return ret; 1712 } 1713 1714 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 1715 { 1716 struct inode *inode = file_inode(filp); 1717 struct super_block *sb = inode->i_sb; 1718 struct request_queue *q = bdev_get_queue(sb->s_bdev); 1719 struct fstrim_range range; 1720 int ret; 1721 1722 if (!capable(CAP_SYS_ADMIN)) 1723 return -EPERM; 1724 1725 if (!blk_queue_discard(q)) 1726 return -EOPNOTSUPP; 1727 1728 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 1729 sizeof(range))) 1730 return -EFAULT; 1731 1732 ret = mnt_want_write_file(filp); 1733 if (ret) 1734 return ret; 1735 1736 range.minlen = max((unsigned int)range.minlen, 1737 q->limits.discard_granularity); 1738 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 1739 mnt_drop_write_file(filp); 1740 if (ret < 0) 1741 return ret; 1742 1743 if (copy_to_user((struct fstrim_range __user *)arg, &range, 1744 sizeof(range))) 1745 return -EFAULT; 1746 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1747 return 0; 1748 } 1749 1750 static bool uuid_is_nonzero(__u8 u[16]) 1751 { 1752 int i; 1753 1754 for (i = 0; i < 16; i++) 1755 if (u[i]) 1756 return true; 1757 return false; 1758 } 1759 1760 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 1761 { 1762 struct fscrypt_policy policy; 1763 struct inode *inode = file_inode(filp); 1764 1765 if (copy_from_user(&policy, (struct fscrypt_policy __user *)arg, 1766 sizeof(policy))) 1767 return -EFAULT; 1768 1769 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1770 1771 return fscrypt_process_policy(filp, &policy); 1772 } 1773 1774 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 1775 { 1776 struct fscrypt_policy policy; 1777 struct inode *inode = file_inode(filp); 1778 int err; 1779 1780 err = fscrypt_get_policy(inode, &policy); 1781 if (err) 1782 return err; 1783 1784 if (copy_to_user((struct fscrypt_policy __user *)arg, &policy, sizeof(policy))) 1785 return -EFAULT; 1786 return 0; 1787 } 1788 1789 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 1790 { 1791 struct inode *inode = file_inode(filp); 1792 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1793 int err; 1794 1795 if (!f2fs_sb_has_crypto(inode->i_sb)) 1796 return -EOPNOTSUPP; 1797 1798 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 1799 goto got_it; 1800 1801 err = mnt_want_write_file(filp); 1802 if (err) 1803 return err; 1804 1805 /* update superblock with uuid */ 1806 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 1807 1808 err = f2fs_commit_super(sbi, false); 1809 if (err) { 1810 /* undo new data */ 1811 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 1812 mnt_drop_write_file(filp); 1813 return err; 1814 } 1815 mnt_drop_write_file(filp); 1816 got_it: 1817 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 1818 16)) 1819 return -EFAULT; 1820 return 0; 1821 } 1822 1823 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 1824 { 1825 struct inode *inode = file_inode(filp); 1826 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1827 __u32 sync; 1828 int ret; 1829 1830 if (!capable(CAP_SYS_ADMIN)) 1831 return -EPERM; 1832 1833 if (get_user(sync, (__u32 __user *)arg)) 1834 return -EFAULT; 1835 1836 if (f2fs_readonly(sbi->sb)) 1837 return -EROFS; 1838 1839 ret = mnt_want_write_file(filp); 1840 if (ret) 1841 return ret; 1842 1843 if (!sync) { 1844 if (!mutex_trylock(&sbi->gc_mutex)) { 1845 ret = -EBUSY; 1846 goto out; 1847 } 1848 } else { 1849 mutex_lock(&sbi->gc_mutex); 1850 } 1851 1852 ret = f2fs_gc(sbi, sync); 1853 out: 1854 mnt_drop_write_file(filp); 1855 return ret; 1856 } 1857 1858 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg) 1859 { 1860 struct inode *inode = file_inode(filp); 1861 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1862 int ret; 1863 1864 if (!capable(CAP_SYS_ADMIN)) 1865 return -EPERM; 1866 1867 if (f2fs_readonly(sbi->sb)) 1868 return -EROFS; 1869 1870 ret = mnt_want_write_file(filp); 1871 if (ret) 1872 return ret; 1873 1874 ret = f2fs_sync_fs(sbi->sb, 1); 1875 1876 mnt_drop_write_file(filp); 1877 return ret; 1878 } 1879 1880 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 1881 struct file *filp, 1882 struct f2fs_defragment *range) 1883 { 1884 struct inode *inode = file_inode(filp); 1885 struct f2fs_map_blocks map = { .m_next_pgofs = NULL }; 1886 struct extent_info ei; 1887 pgoff_t pg_start, pg_end; 1888 unsigned int blk_per_seg = sbi->blocks_per_seg; 1889 unsigned int total = 0, sec_num; 1890 unsigned int pages_per_sec = sbi->segs_per_sec * blk_per_seg; 1891 block_t blk_end = 0; 1892 bool fragmented = false; 1893 int err; 1894 1895 /* if in-place-update policy is enabled, don't waste time here */ 1896 if (need_inplace_update(inode)) 1897 return -EINVAL; 1898 1899 pg_start = range->start >> PAGE_SHIFT; 1900 pg_end = (range->start + range->len) >> PAGE_SHIFT; 1901 1902 f2fs_balance_fs(sbi, true); 1903 1904 inode_lock(inode); 1905 1906 /* writeback all dirty pages in the range */ 1907 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 1908 range->start + range->len - 1); 1909 if (err) 1910 goto out; 1911 1912 /* 1913 * lookup mapping info in extent cache, skip defragmenting if physical 1914 * block addresses are continuous. 1915 */ 1916 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) { 1917 if (ei.fofs + ei.len >= pg_end) 1918 goto out; 1919 } 1920 1921 map.m_lblk = pg_start; 1922 1923 /* 1924 * lookup mapping info in dnode page cache, skip defragmenting if all 1925 * physical block addresses are continuous even if there are hole(s) 1926 * in logical blocks. 1927 */ 1928 while (map.m_lblk < pg_end) { 1929 map.m_len = pg_end - map.m_lblk; 1930 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ); 1931 if (err) 1932 goto out; 1933 1934 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 1935 map.m_lblk++; 1936 continue; 1937 } 1938 1939 if (blk_end && blk_end != map.m_pblk) { 1940 fragmented = true; 1941 break; 1942 } 1943 blk_end = map.m_pblk + map.m_len; 1944 1945 map.m_lblk += map.m_len; 1946 } 1947 1948 if (!fragmented) 1949 goto out; 1950 1951 map.m_lblk = pg_start; 1952 map.m_len = pg_end - pg_start; 1953 1954 sec_num = (map.m_len + pages_per_sec - 1) / pages_per_sec; 1955 1956 /* 1957 * make sure there are enough free section for LFS allocation, this can 1958 * avoid defragment running in SSR mode when free section are allocated 1959 * intensively 1960 */ 1961 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 1962 err = -EAGAIN; 1963 goto out; 1964 } 1965 1966 while (map.m_lblk < pg_end) { 1967 pgoff_t idx; 1968 int cnt = 0; 1969 1970 do_map: 1971 map.m_len = pg_end - map.m_lblk; 1972 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_READ); 1973 if (err) 1974 goto clear_out; 1975 1976 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 1977 map.m_lblk++; 1978 continue; 1979 } 1980 1981 set_inode_flag(inode, FI_DO_DEFRAG); 1982 1983 idx = map.m_lblk; 1984 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 1985 struct page *page; 1986 1987 page = get_lock_data_page(inode, idx, true); 1988 if (IS_ERR(page)) { 1989 err = PTR_ERR(page); 1990 goto clear_out; 1991 } 1992 1993 set_page_dirty(page); 1994 f2fs_put_page(page, 1); 1995 1996 idx++; 1997 cnt++; 1998 total++; 1999 } 2000 2001 map.m_lblk = idx; 2002 2003 if (idx < pg_end && cnt < blk_per_seg) 2004 goto do_map; 2005 2006 clear_inode_flag(inode, FI_DO_DEFRAG); 2007 2008 err = filemap_fdatawrite(inode->i_mapping); 2009 if (err) 2010 goto out; 2011 } 2012 clear_out: 2013 clear_inode_flag(inode, FI_DO_DEFRAG); 2014 out: 2015 inode_unlock(inode); 2016 if (!err) 2017 range->len = (u64)total << PAGE_SHIFT; 2018 return err; 2019 } 2020 2021 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2022 { 2023 struct inode *inode = file_inode(filp); 2024 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2025 struct f2fs_defragment range; 2026 int err; 2027 2028 if (!capable(CAP_SYS_ADMIN)) 2029 return -EPERM; 2030 2031 if (!S_ISREG(inode->i_mode)) 2032 return -EINVAL; 2033 2034 err = mnt_want_write_file(filp); 2035 if (err) 2036 return err; 2037 2038 if (f2fs_readonly(sbi->sb)) { 2039 err = -EROFS; 2040 goto out; 2041 } 2042 2043 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2044 sizeof(range))) { 2045 err = -EFAULT; 2046 goto out; 2047 } 2048 2049 /* verify alignment of offset & size */ 2050 if (range.start & (F2FS_BLKSIZE - 1) || 2051 range.len & (F2FS_BLKSIZE - 1)) { 2052 err = -EINVAL; 2053 goto out; 2054 } 2055 2056 err = f2fs_defragment_range(sbi, filp, &range); 2057 f2fs_update_time(sbi, REQ_TIME); 2058 if (err < 0) 2059 goto out; 2060 2061 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2062 sizeof(range))) 2063 err = -EFAULT; 2064 out: 2065 mnt_drop_write_file(filp); 2066 return err; 2067 } 2068 2069 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2070 struct file *file_out, loff_t pos_out, size_t len) 2071 { 2072 struct inode *src = file_inode(file_in); 2073 struct inode *dst = file_inode(file_out); 2074 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2075 size_t olen = len, dst_max_i_size = 0; 2076 size_t dst_osize; 2077 int ret; 2078 2079 if (file_in->f_path.mnt != file_out->f_path.mnt || 2080 src->i_sb != dst->i_sb) 2081 return -EXDEV; 2082 2083 if (unlikely(f2fs_readonly(src->i_sb))) 2084 return -EROFS; 2085 2086 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2087 return -EINVAL; 2088 2089 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst)) 2090 return -EOPNOTSUPP; 2091 2092 if (src == dst) { 2093 if (pos_in == pos_out) 2094 return 0; 2095 if (pos_out > pos_in && pos_out < pos_in + len) 2096 return -EINVAL; 2097 } 2098 2099 inode_lock(src); 2100 if (src != dst) { 2101 if (!inode_trylock(dst)) { 2102 ret = -EBUSY; 2103 goto out; 2104 } 2105 } 2106 2107 ret = -EINVAL; 2108 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2109 goto out_unlock; 2110 if (len == 0) 2111 olen = len = src->i_size - pos_in; 2112 if (pos_in + len == src->i_size) 2113 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2114 if (len == 0) { 2115 ret = 0; 2116 goto out_unlock; 2117 } 2118 2119 dst_osize = dst->i_size; 2120 if (pos_out + olen > dst->i_size) 2121 dst_max_i_size = pos_out + olen; 2122 2123 /* verify the end result is block aligned */ 2124 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2125 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2126 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2127 goto out_unlock; 2128 2129 ret = f2fs_convert_inline_inode(src); 2130 if (ret) 2131 goto out_unlock; 2132 2133 ret = f2fs_convert_inline_inode(dst); 2134 if (ret) 2135 goto out_unlock; 2136 2137 /* write out all dirty pages from offset */ 2138 ret = filemap_write_and_wait_range(src->i_mapping, 2139 pos_in, pos_in + len); 2140 if (ret) 2141 goto out_unlock; 2142 2143 ret = filemap_write_and_wait_range(dst->i_mapping, 2144 pos_out, pos_out + len); 2145 if (ret) 2146 goto out_unlock; 2147 2148 f2fs_balance_fs(sbi, true); 2149 f2fs_lock_op(sbi); 2150 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2151 pos_out >> F2FS_BLKSIZE_BITS, 2152 len >> F2FS_BLKSIZE_BITS, false); 2153 2154 if (!ret) { 2155 if (dst_max_i_size) 2156 f2fs_i_size_write(dst, dst_max_i_size); 2157 else if (dst_osize != dst->i_size) 2158 f2fs_i_size_write(dst, dst_osize); 2159 } 2160 f2fs_unlock_op(sbi); 2161 out_unlock: 2162 if (src != dst) 2163 inode_unlock(dst); 2164 out: 2165 inode_unlock(src); 2166 return ret; 2167 } 2168 2169 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2170 { 2171 struct f2fs_move_range range; 2172 struct fd dst; 2173 int err; 2174 2175 if (!(filp->f_mode & FMODE_READ) || 2176 !(filp->f_mode & FMODE_WRITE)) 2177 return -EBADF; 2178 2179 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2180 sizeof(range))) 2181 return -EFAULT; 2182 2183 dst = fdget(range.dst_fd); 2184 if (!dst.file) 2185 return -EBADF; 2186 2187 if (!(dst.file->f_mode & FMODE_WRITE)) { 2188 err = -EBADF; 2189 goto err_out; 2190 } 2191 2192 err = mnt_want_write_file(filp); 2193 if (err) 2194 goto err_out; 2195 2196 err = f2fs_move_file_range(filp, range.pos_in, dst.file, 2197 range.pos_out, range.len); 2198 2199 mnt_drop_write_file(filp); 2200 2201 if (copy_to_user((struct f2fs_move_range __user *)arg, 2202 &range, sizeof(range))) 2203 err = -EFAULT; 2204 err_out: 2205 fdput(dst); 2206 return err; 2207 } 2208 2209 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 2210 { 2211 switch (cmd) { 2212 case F2FS_IOC_GETFLAGS: 2213 return f2fs_ioc_getflags(filp, arg); 2214 case F2FS_IOC_SETFLAGS: 2215 return f2fs_ioc_setflags(filp, arg); 2216 case F2FS_IOC_GETVERSION: 2217 return f2fs_ioc_getversion(filp, arg); 2218 case F2FS_IOC_START_ATOMIC_WRITE: 2219 return f2fs_ioc_start_atomic_write(filp); 2220 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 2221 return f2fs_ioc_commit_atomic_write(filp); 2222 case F2FS_IOC_START_VOLATILE_WRITE: 2223 return f2fs_ioc_start_volatile_write(filp); 2224 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 2225 return f2fs_ioc_release_volatile_write(filp); 2226 case F2FS_IOC_ABORT_VOLATILE_WRITE: 2227 return f2fs_ioc_abort_volatile_write(filp); 2228 case F2FS_IOC_SHUTDOWN: 2229 return f2fs_ioc_shutdown(filp, arg); 2230 case FITRIM: 2231 return f2fs_ioc_fitrim(filp, arg); 2232 case F2FS_IOC_SET_ENCRYPTION_POLICY: 2233 return f2fs_ioc_set_encryption_policy(filp, arg); 2234 case F2FS_IOC_GET_ENCRYPTION_POLICY: 2235 return f2fs_ioc_get_encryption_policy(filp, arg); 2236 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 2237 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 2238 case F2FS_IOC_GARBAGE_COLLECT: 2239 return f2fs_ioc_gc(filp, arg); 2240 case F2FS_IOC_WRITE_CHECKPOINT: 2241 return f2fs_ioc_write_checkpoint(filp, arg); 2242 case F2FS_IOC_DEFRAGMENT: 2243 return f2fs_ioc_defragment(filp, arg); 2244 case F2FS_IOC_MOVE_RANGE: 2245 return f2fs_ioc_move_range(filp, arg); 2246 default: 2247 return -ENOTTY; 2248 } 2249 } 2250 2251 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 2252 { 2253 struct file *file = iocb->ki_filp; 2254 struct inode *inode = file_inode(file); 2255 struct blk_plug plug; 2256 ssize_t ret; 2257 2258 if (f2fs_encrypted_inode(inode) && 2259 !fscrypt_has_encryption_key(inode) && 2260 fscrypt_get_encryption_info(inode)) 2261 return -EACCES; 2262 2263 inode_lock(inode); 2264 ret = generic_write_checks(iocb, from); 2265 if (ret > 0) { 2266 ret = f2fs_preallocate_blocks(iocb, from); 2267 if (!ret) { 2268 blk_start_plug(&plug); 2269 ret = __generic_file_write_iter(iocb, from); 2270 blk_finish_plug(&plug); 2271 } 2272 } 2273 inode_unlock(inode); 2274 2275 if (ret > 0) 2276 ret = generic_write_sync(iocb, ret); 2277 return ret; 2278 } 2279 2280 #ifdef CONFIG_COMPAT 2281 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 2282 { 2283 switch (cmd) { 2284 case F2FS_IOC32_GETFLAGS: 2285 cmd = F2FS_IOC_GETFLAGS; 2286 break; 2287 case F2FS_IOC32_SETFLAGS: 2288 cmd = F2FS_IOC_SETFLAGS; 2289 break; 2290 case F2FS_IOC32_GETVERSION: 2291 cmd = F2FS_IOC_GETVERSION; 2292 break; 2293 case F2FS_IOC_START_ATOMIC_WRITE: 2294 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 2295 case F2FS_IOC_START_VOLATILE_WRITE: 2296 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 2297 case F2FS_IOC_ABORT_VOLATILE_WRITE: 2298 case F2FS_IOC_SHUTDOWN: 2299 case F2FS_IOC_SET_ENCRYPTION_POLICY: 2300 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 2301 case F2FS_IOC_GET_ENCRYPTION_POLICY: 2302 case F2FS_IOC_GARBAGE_COLLECT: 2303 case F2FS_IOC_WRITE_CHECKPOINT: 2304 case F2FS_IOC_DEFRAGMENT: 2305 break; 2306 case F2FS_IOC_MOVE_RANGE: 2307 break; 2308 default: 2309 return -ENOIOCTLCMD; 2310 } 2311 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 2312 } 2313 #endif 2314 2315 const struct file_operations f2fs_file_operations = { 2316 .llseek = f2fs_llseek, 2317 .read_iter = generic_file_read_iter, 2318 .write_iter = f2fs_file_write_iter, 2319 .open = f2fs_file_open, 2320 .release = f2fs_release_file, 2321 .mmap = f2fs_file_mmap, 2322 .fsync = f2fs_sync_file, 2323 .fallocate = f2fs_fallocate, 2324 .unlocked_ioctl = f2fs_ioctl, 2325 #ifdef CONFIG_COMPAT 2326 .compat_ioctl = f2fs_compat_ioctl, 2327 #endif 2328 .splice_read = generic_file_splice_read, 2329 .splice_write = iter_file_splice_write, 2330 }; 2331