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/writeback.h> 12 #include <linux/blkdev.h> 13 #include <linux/falloc.h> 14 #include <linux/filelock.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 void f2fs_zero_post_eof_page(struct inode *inode, 40 loff_t new_size, bool lock) 41 { 42 loff_t old_size = i_size_read(inode); 43 44 if (old_size >= new_size) 45 return; 46 47 if (mapping_empty(inode->i_mapping)) 48 return; 49 50 if (lock) 51 filemap_invalidate_lock(inode->i_mapping); 52 /* zero or drop pages only in range of [old_size, new_size] */ 53 truncate_inode_pages_range(inode->i_mapping, old_size, new_size); 54 if (lock) 55 filemap_invalidate_unlock(inode->i_mapping); 56 } 57 58 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 59 { 60 struct inode *inode = file_inode(vmf->vma->vm_file); 61 vm_flags_t flags = vmf->vma->vm_flags; 62 vm_fault_t ret; 63 64 ret = filemap_fault(vmf); 65 if (ret & VM_FAULT_LOCKED) 66 f2fs_update_iostat(F2FS_I_SB(inode), inode, 67 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 68 69 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret); 70 71 return ret; 72 } 73 74 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 75 { 76 struct folio *folio = page_folio(vmf->page); 77 struct inode *inode = file_inode(vmf->vma->vm_file); 78 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 79 struct dnode_of_data dn; 80 bool need_alloc = !f2fs_is_pinned_file(inode); 81 int err = 0; 82 vm_fault_t ret; 83 84 if (unlikely(IS_IMMUTABLE(inode))) 85 return VM_FAULT_SIGBUS; 86 87 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 88 err = -EIO; 89 goto out; 90 } 91 92 if (unlikely(f2fs_cp_error(sbi))) { 93 err = -EIO; 94 goto out; 95 } 96 97 if (!f2fs_is_checkpoint_ready(sbi)) { 98 err = -ENOSPC; 99 goto out; 100 } 101 102 err = f2fs_convert_inline_inode(inode); 103 if (err) 104 goto out; 105 106 #ifdef CONFIG_F2FS_FS_COMPRESSION 107 if (f2fs_compressed_file(inode)) { 108 int ret = f2fs_is_compressed_cluster(inode, folio->index); 109 110 if (ret < 0) { 111 err = ret; 112 goto out; 113 } else if (ret) { 114 need_alloc = false; 115 } 116 } 117 #endif 118 /* should do out of any locked page */ 119 if (need_alloc) 120 f2fs_balance_fs(sbi, true); 121 122 sb_start_pagefault(inode->i_sb); 123 124 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 125 126 f2fs_zero_post_eof_page(inode, (folio->index + 1) << PAGE_SHIFT, true); 127 128 file_update_time(vmf->vma->vm_file); 129 filemap_invalidate_lock_shared(inode->i_mapping); 130 131 folio_lock(folio); 132 if (unlikely(folio->mapping != inode->i_mapping || 133 folio_pos(folio) > i_size_read(inode) || 134 !folio_test_uptodate(folio))) { 135 folio_unlock(folio); 136 err = -EFAULT; 137 goto out_sem; 138 } 139 140 set_new_dnode(&dn, inode, NULL, NULL, 0); 141 if (need_alloc) { 142 /* block allocation */ 143 err = f2fs_get_block_locked(&dn, folio->index); 144 } else { 145 err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE); 146 f2fs_put_dnode(&dn); 147 if (f2fs_is_pinned_file(inode) && 148 !__is_valid_data_blkaddr(dn.data_blkaddr)) 149 err = -EIO; 150 } 151 152 if (err) { 153 folio_unlock(folio); 154 goto out_sem; 155 } 156 157 f2fs_folio_wait_writeback(folio, DATA, false, true); 158 159 /* wait for GCed page writeback via META_MAPPING */ 160 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 161 162 /* 163 * check to see if the page is mapped already (no holes) 164 */ 165 if (folio_test_mappedtodisk(folio)) 166 goto out_sem; 167 168 /* page is wholly or partially inside EOF */ 169 if (((loff_t)(folio->index + 1) << PAGE_SHIFT) > 170 i_size_read(inode)) { 171 loff_t offset; 172 173 offset = i_size_read(inode) & ~PAGE_MASK; 174 folio_zero_segment(folio, offset, folio_size(folio)); 175 } 176 folio_mark_dirty(folio); 177 178 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 179 f2fs_update_time(sbi, REQ_TIME); 180 181 out_sem: 182 filemap_invalidate_unlock_shared(inode->i_mapping); 183 184 sb_end_pagefault(inode->i_sb); 185 out: 186 ret = vmf_fs_error(err); 187 188 trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret); 189 return ret; 190 } 191 192 static const struct vm_operations_struct f2fs_file_vm_ops = { 193 .fault = f2fs_filemap_fault, 194 .map_pages = filemap_map_pages, 195 .page_mkwrite = f2fs_vm_page_mkwrite, 196 }; 197 198 static int get_parent_ino(struct inode *inode, nid_t *pino) 199 { 200 struct dentry *dentry; 201 202 /* 203 * Make sure to get the non-deleted alias. The alias associated with 204 * the open file descriptor being fsync()'ed may be deleted already. 205 */ 206 dentry = d_find_alias(inode); 207 if (!dentry) 208 return 0; 209 210 *pino = d_parent_ino(dentry); 211 dput(dentry); 212 return 1; 213 } 214 215 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 216 { 217 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 218 enum cp_reason_type cp_reason = CP_NO_NEEDED; 219 220 if (!S_ISREG(inode->i_mode)) 221 cp_reason = CP_NON_REGULAR; 222 else if (f2fs_compressed_file(inode)) 223 cp_reason = CP_COMPRESSED; 224 else if (inode->i_nlink != 1) 225 cp_reason = CP_HARDLINK; 226 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 227 cp_reason = CP_SB_NEED_CP; 228 else if (file_wrong_pino(inode)) 229 cp_reason = CP_WRONG_PINO; 230 else if (!f2fs_space_for_roll_forward(sbi)) 231 cp_reason = CP_NO_SPC_ROLL; 232 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 233 cp_reason = CP_NODE_NEED_CP; 234 else if (test_opt(sbi, FASTBOOT)) 235 cp_reason = CP_FASTBOOT_MODE; 236 else if (F2FS_OPTION(sbi).active_logs == 2) 237 cp_reason = CP_SPEC_LOG_NUM; 238 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 239 f2fs_need_dentry_mark(sbi, inode->i_ino) && 240 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 241 TRANS_DIR_INO)) 242 cp_reason = CP_RECOVER_DIR; 243 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 244 XATTR_DIR_INO)) 245 cp_reason = CP_XATTR_DIR; 246 247 return cp_reason; 248 } 249 250 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 251 { 252 struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino); 253 bool ret = false; 254 /* But we need to avoid that there are some inode updates */ 255 if ((!IS_ERR(i) && folio_test_dirty(i)) || 256 f2fs_need_inode_block_update(sbi, ino)) 257 ret = true; 258 f2fs_folio_put(i, false); 259 return ret; 260 } 261 262 static void try_to_fix_pino(struct inode *inode) 263 { 264 struct f2fs_inode_info *fi = F2FS_I(inode); 265 nid_t pino; 266 267 f2fs_down_write(&fi->i_sem); 268 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 269 get_parent_ino(inode, &pino)) { 270 f2fs_i_pino_write(inode, pino); 271 file_got_pino(inode); 272 } 273 f2fs_up_write(&fi->i_sem); 274 } 275 276 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 277 int datasync, bool atomic) 278 { 279 struct inode *inode = file->f_mapping->host; 280 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 281 nid_t ino = inode->i_ino; 282 int ret = 0; 283 enum cp_reason_type cp_reason = 0; 284 struct writeback_control wbc = { 285 .sync_mode = WB_SYNC_ALL, 286 .nr_to_write = LONG_MAX, 287 }; 288 unsigned int seq_id = 0; 289 290 if (unlikely(f2fs_readonly(inode->i_sb))) 291 return 0; 292 293 trace_f2fs_sync_file_enter(inode); 294 295 if (S_ISDIR(inode->i_mode)) 296 goto go_write; 297 298 /* if fdatasync is triggered, let's do in-place-update */ 299 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 300 set_inode_flag(inode, FI_NEED_IPU); 301 ret = file_write_and_wait_range(file, start, end); 302 clear_inode_flag(inode, FI_NEED_IPU); 303 304 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 305 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 306 return ret; 307 } 308 309 /* if the inode is dirty, let's recover all the time */ 310 if (!f2fs_skip_inode_update(inode, datasync)) { 311 f2fs_write_inode(inode, NULL); 312 goto go_write; 313 } 314 315 /* 316 * if there is no written data, don't waste time to write recovery info. 317 */ 318 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 319 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 320 321 /* it may call write_inode just prior to fsync */ 322 if (need_inode_page_update(sbi, ino)) 323 goto go_write; 324 325 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 326 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 327 goto flush_out; 328 goto out; 329 } else { 330 /* 331 * for OPU case, during fsync(), node can be persisted before 332 * data when lower device doesn't support write barrier, result 333 * in data corruption after SPO. 334 * So for strict fsync mode, force to use atomic write semantics 335 * to keep write order in between data/node and last node to 336 * avoid potential data corruption. 337 */ 338 if (F2FS_OPTION(sbi).fsync_mode == 339 FSYNC_MODE_STRICT && !atomic) 340 atomic = true; 341 } 342 go_write: 343 /* 344 * Both of fdatasync() and fsync() are able to be recovered from 345 * sudden-power-off. 346 */ 347 f2fs_down_read(&F2FS_I(inode)->i_sem); 348 cp_reason = need_do_checkpoint(inode); 349 f2fs_up_read(&F2FS_I(inode)->i_sem); 350 351 if (cp_reason) { 352 /* all the dirty node pages should be flushed for POR */ 353 ret = f2fs_sync_fs(inode->i_sb, 1); 354 355 /* 356 * We've secured consistency through sync_fs. Following pino 357 * will be used only for fsynced inodes after checkpoint. 358 */ 359 try_to_fix_pino(inode); 360 clear_inode_flag(inode, FI_APPEND_WRITE); 361 clear_inode_flag(inode, FI_UPDATE_WRITE); 362 goto out; 363 } 364 sync_nodes: 365 atomic_inc(&sbi->wb_sync_req[NODE]); 366 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 367 atomic_dec(&sbi->wb_sync_req[NODE]); 368 if (ret) 369 goto out; 370 371 /* if cp_error was enabled, we should avoid infinite loop */ 372 if (unlikely(f2fs_cp_error(sbi))) { 373 ret = -EIO; 374 goto out; 375 } 376 377 if (f2fs_need_inode_block_update(sbi, ino)) { 378 f2fs_mark_inode_dirty_sync(inode, true); 379 f2fs_write_inode(inode, NULL); 380 goto sync_nodes; 381 } 382 383 /* 384 * If it's atomic_write, it's just fine to keep write ordering. So 385 * here we don't need to wait for node write completion, since we use 386 * node chain which serializes node blocks. If one of node writes are 387 * reordered, we can see simply broken chain, resulting in stopping 388 * roll-forward recovery. It means we'll recover all or none node blocks 389 * given fsync mark. 390 */ 391 if (!atomic) { 392 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 393 if (ret) 394 goto out; 395 } 396 397 /* once recovery info is written, don't need to tack this */ 398 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 399 clear_inode_flag(inode, FI_APPEND_WRITE); 400 flush_out: 401 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 402 ret = f2fs_issue_flush(sbi, inode->i_ino); 403 if (!ret) { 404 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 405 clear_inode_flag(inode, FI_UPDATE_WRITE); 406 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 407 } 408 f2fs_update_time(sbi, REQ_TIME); 409 out: 410 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 411 return ret; 412 } 413 414 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 415 { 416 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 417 return -EIO; 418 return f2fs_do_sync_file(file, start, end, datasync, false); 419 } 420 421 static bool __found_offset(struct address_space *mapping, 422 struct dnode_of_data *dn, pgoff_t index, int whence) 423 { 424 block_t blkaddr = f2fs_data_blkaddr(dn); 425 struct inode *inode = mapping->host; 426 bool compressed_cluster = false; 427 428 if (f2fs_compressed_file(inode)) { 429 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio, 430 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size)); 431 432 compressed_cluster = first_blkaddr == COMPRESS_ADDR; 433 } 434 435 switch (whence) { 436 case SEEK_DATA: 437 if (__is_valid_data_blkaddr(blkaddr)) 438 return true; 439 if (blkaddr == NEW_ADDR && 440 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 441 return true; 442 if (compressed_cluster) 443 return true; 444 break; 445 case SEEK_HOLE: 446 if (compressed_cluster) 447 return false; 448 if (blkaddr == NULL_ADDR) 449 return true; 450 break; 451 } 452 return false; 453 } 454 455 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 456 { 457 struct inode *inode = file->f_mapping->host; 458 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 459 struct dnode_of_data dn; 460 pgoff_t pgofs, end_offset; 461 loff_t data_ofs = offset; 462 loff_t isize; 463 int err = 0; 464 465 inode_lock_shared(inode); 466 467 isize = i_size_read(inode); 468 if (offset >= isize) 469 goto fail; 470 471 /* handle inline data case */ 472 if (f2fs_has_inline_data(inode)) { 473 if (whence == SEEK_HOLE) { 474 data_ofs = isize; 475 goto found; 476 } else if (whence == SEEK_DATA) { 477 data_ofs = offset; 478 goto found; 479 } 480 } 481 482 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 483 484 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 485 set_new_dnode(&dn, inode, NULL, NULL, 0); 486 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 487 if (err && err != -ENOENT) { 488 goto fail; 489 } else if (err == -ENOENT) { 490 /* direct node does not exists */ 491 if (whence == SEEK_DATA) { 492 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 493 continue; 494 } else { 495 goto found; 496 } 497 } 498 499 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 500 501 /* find data/hole in dnode block */ 502 for (; dn.ofs_in_node < end_offset; 503 dn.ofs_in_node++, pgofs++, 504 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 505 block_t blkaddr; 506 507 blkaddr = f2fs_data_blkaddr(&dn); 508 509 if (__is_valid_data_blkaddr(blkaddr) && 510 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 511 blkaddr, DATA_GENERIC_ENHANCE)) { 512 f2fs_put_dnode(&dn); 513 goto fail; 514 } 515 516 if (__found_offset(file->f_mapping, &dn, 517 pgofs, whence)) { 518 f2fs_put_dnode(&dn); 519 goto found; 520 } 521 } 522 f2fs_put_dnode(&dn); 523 } 524 525 if (whence == SEEK_DATA) 526 goto fail; 527 found: 528 if (whence == SEEK_HOLE && data_ofs > isize) 529 data_ofs = isize; 530 inode_unlock_shared(inode); 531 return vfs_setpos(file, data_ofs, maxbytes); 532 fail: 533 inode_unlock_shared(inode); 534 return -ENXIO; 535 } 536 537 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 538 { 539 struct inode *inode = file->f_mapping->host; 540 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 541 542 switch (whence) { 543 case SEEK_SET: 544 case SEEK_CUR: 545 case SEEK_END: 546 return generic_file_llseek_size(file, offset, whence, 547 maxbytes, i_size_read(inode)); 548 case SEEK_DATA: 549 case SEEK_HOLE: 550 if (offset < 0) 551 return -ENXIO; 552 return f2fs_seek_block(file, offset, whence); 553 } 554 555 return -EINVAL; 556 } 557 558 static int f2fs_file_mmap_prepare(struct vm_area_desc *desc) 559 { 560 struct file *file = desc->file; 561 struct inode *inode = file_inode(file); 562 563 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 564 return -EIO; 565 566 if (!f2fs_is_compress_backend_ready(inode)) 567 return -EOPNOTSUPP; 568 569 file_accessed(file); 570 desc->vm_ops = &f2fs_file_vm_ops; 571 572 f2fs_down_read(&F2FS_I(inode)->i_sem); 573 set_inode_flag(inode, FI_MMAP_FILE); 574 f2fs_up_read(&F2FS_I(inode)->i_sem); 575 576 return 0; 577 } 578 579 static int finish_preallocate_blocks(struct inode *inode) 580 { 581 int ret = 0; 582 bool opened; 583 584 f2fs_down_read(&F2FS_I(inode)->i_sem); 585 opened = is_inode_flag_set(inode, FI_OPENED_FILE); 586 f2fs_up_read(&F2FS_I(inode)->i_sem); 587 if (opened) 588 return 0; 589 590 inode_lock(inode); 591 if (is_inode_flag_set(inode, FI_OPENED_FILE)) 592 goto out_unlock; 593 594 if (!file_should_truncate(inode)) 595 goto out_update; 596 597 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 598 filemap_invalidate_lock(inode->i_mapping); 599 600 truncate_setsize(inode, i_size_read(inode)); 601 ret = f2fs_truncate(inode); 602 603 filemap_invalidate_unlock(inode->i_mapping); 604 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 605 if (ret) 606 goto out_unlock; 607 608 file_dont_truncate(inode); 609 out_update: 610 f2fs_down_write(&F2FS_I(inode)->i_sem); 611 set_inode_flag(inode, FI_OPENED_FILE); 612 f2fs_up_write(&F2FS_I(inode)->i_sem); 613 out_unlock: 614 inode_unlock(inode); 615 return ret; 616 } 617 618 static int f2fs_file_open(struct inode *inode, struct file *filp) 619 { 620 int err = fscrypt_file_open(inode, filp); 621 622 if (err) 623 return err; 624 625 if (!f2fs_is_compress_backend_ready(inode)) 626 return -EOPNOTSUPP; 627 628 err = fsverity_file_open(inode, filp); 629 if (err) 630 return err; 631 632 filp->f_mode |= FMODE_NOWAIT; 633 filp->f_mode |= FMODE_CAN_ODIRECT; 634 635 err = dquot_file_open(inode, filp); 636 if (err) 637 return err; 638 639 err = finish_preallocate_blocks(inode); 640 if (!err) 641 atomic_inc(&F2FS_I(inode)->open_count); 642 return err; 643 } 644 645 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 646 { 647 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 648 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 649 __le32 *addr; 650 bool compressed_cluster = false; 651 int cluster_index = 0, valid_blocks = 0; 652 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 653 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 654 block_t blkstart; 655 int blklen = 0; 656 657 addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs; 658 blkstart = le32_to_cpu(*addr); 659 660 /* Assumption: truncation starts with cluster */ 661 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 662 block_t blkaddr = le32_to_cpu(*addr); 663 664 if (f2fs_compressed_file(dn->inode) && 665 !(cluster_index & (cluster_size - 1))) { 666 if (compressed_cluster) 667 f2fs_i_compr_blocks_update(dn->inode, 668 valid_blocks, false); 669 compressed_cluster = (blkaddr == COMPRESS_ADDR); 670 valid_blocks = 0; 671 } 672 673 if (blkaddr == NULL_ADDR) 674 goto next; 675 676 f2fs_set_data_blkaddr(dn, NULL_ADDR); 677 678 if (__is_valid_data_blkaddr(blkaddr)) { 679 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE)) 680 goto next; 681 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr, 682 DATA_GENERIC_ENHANCE)) 683 goto next; 684 if (compressed_cluster) 685 valid_blocks++; 686 } 687 688 if (blkstart + blklen == blkaddr) { 689 blklen++; 690 } else { 691 f2fs_invalidate_blocks(sbi, blkstart, blklen); 692 blkstart = blkaddr; 693 blklen = 1; 694 } 695 696 if (!released || blkaddr != COMPRESS_ADDR) 697 nr_free++; 698 699 continue; 700 701 next: 702 if (blklen) 703 f2fs_invalidate_blocks(sbi, blkstart, blklen); 704 705 blkstart = le32_to_cpu(*(addr + 1)); 706 blklen = 0; 707 } 708 709 if (blklen) 710 f2fs_invalidate_blocks(sbi, blkstart, blklen); 711 712 if (compressed_cluster) 713 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 714 715 if (nr_free) { 716 pgoff_t fofs; 717 /* 718 * once we invalidate valid blkaddr in range [ofs, ofs + count], 719 * we will invalidate all blkaddr in the whole range. 720 */ 721 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_folio), 722 dn->inode) + ofs; 723 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 724 f2fs_update_age_extent_cache_range(dn, fofs, len); 725 dec_valid_block_count(sbi, dn->inode, nr_free); 726 } 727 dn->ofs_in_node = ofs; 728 729 f2fs_update_time(sbi, REQ_TIME); 730 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 731 dn->ofs_in_node, nr_free); 732 } 733 734 static int truncate_partial_data_page(struct inode *inode, u64 from, 735 bool cache_only) 736 { 737 loff_t offset = from & (PAGE_SIZE - 1); 738 pgoff_t index = from >> PAGE_SHIFT; 739 struct address_space *mapping = inode->i_mapping; 740 struct folio *folio; 741 742 if (!offset && !cache_only) 743 return 0; 744 745 if (cache_only) { 746 folio = filemap_lock_folio(mapping, index); 747 if (IS_ERR(folio)) 748 return 0; 749 if (folio_test_uptodate(folio)) 750 goto truncate_out; 751 f2fs_folio_put(folio, true); 752 return 0; 753 } 754 755 folio = f2fs_get_lock_data_folio(inode, index, true); 756 if (IS_ERR(folio)) 757 return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio); 758 truncate_out: 759 f2fs_folio_wait_writeback(folio, DATA, true, true); 760 folio_zero_segment(folio, offset, folio_size(folio)); 761 762 /* An encrypted inode should have a key and truncate the last page. */ 763 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 764 if (!cache_only) 765 folio_mark_dirty(folio); 766 f2fs_folio_put(folio, true); 767 return 0; 768 } 769 770 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 771 { 772 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 773 struct dnode_of_data dn; 774 pgoff_t free_from; 775 int count = 0, err = 0; 776 struct folio *ifolio; 777 bool truncate_page = false; 778 779 trace_f2fs_truncate_blocks_enter(inode, from); 780 781 if (IS_DEVICE_ALIASING(inode) && from) { 782 err = -EINVAL; 783 goto out_err; 784 } 785 786 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 787 788 if (free_from >= max_file_blocks(inode)) 789 goto free_partial; 790 791 if (lock) 792 f2fs_lock_op(sbi); 793 794 ifolio = f2fs_get_inode_folio(sbi, inode->i_ino); 795 if (IS_ERR(ifolio)) { 796 err = PTR_ERR(ifolio); 797 goto out; 798 } 799 800 if (IS_DEVICE_ALIASING(inode)) { 801 struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ]; 802 struct extent_info ei = et->largest; 803 804 f2fs_invalidate_blocks(sbi, ei.blk, ei.len); 805 806 dec_valid_block_count(sbi, inode, ei.len); 807 f2fs_update_time(sbi, REQ_TIME); 808 809 f2fs_folio_put(ifolio, true); 810 goto out; 811 } 812 813 if (f2fs_has_inline_data(inode)) { 814 f2fs_truncate_inline_inode(inode, ifolio, from); 815 f2fs_folio_put(ifolio, true); 816 truncate_page = true; 817 goto out; 818 } 819 820 set_new_dnode(&dn, inode, ifolio, NULL, 0); 821 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 822 if (err) { 823 if (err == -ENOENT) 824 goto free_next; 825 goto out; 826 } 827 828 count = ADDRS_PER_PAGE(dn.node_folio, inode); 829 830 count -= dn.ofs_in_node; 831 f2fs_bug_on(sbi, count < 0); 832 833 if (dn.ofs_in_node || IS_INODE(dn.node_folio)) { 834 f2fs_truncate_data_blocks_range(&dn, count); 835 free_from += count; 836 } 837 838 f2fs_put_dnode(&dn); 839 free_next: 840 err = f2fs_truncate_inode_blocks(inode, free_from); 841 out: 842 if (lock) 843 f2fs_unlock_op(sbi); 844 free_partial: 845 /* lastly zero out the first data page */ 846 if (!err) 847 err = truncate_partial_data_page(inode, from, truncate_page); 848 out_err: 849 trace_f2fs_truncate_blocks_exit(inode, err); 850 return err; 851 } 852 853 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 854 { 855 u64 free_from = from; 856 int err; 857 858 #ifdef CONFIG_F2FS_FS_COMPRESSION 859 /* 860 * for compressed file, only support cluster size 861 * aligned truncation. 862 */ 863 if (f2fs_compressed_file(inode)) 864 free_from = round_up(from, 865 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 866 #endif 867 868 err = f2fs_do_truncate_blocks(inode, free_from, lock); 869 if (err) 870 return err; 871 872 #ifdef CONFIG_F2FS_FS_COMPRESSION 873 /* 874 * For compressed file, after release compress blocks, don't allow write 875 * direct, but we should allow write direct after truncate to zero. 876 */ 877 if (f2fs_compressed_file(inode) && !free_from 878 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 879 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 880 881 if (from != free_from) { 882 err = f2fs_truncate_partial_cluster(inode, from, lock); 883 if (err) 884 return err; 885 } 886 #endif 887 888 return 0; 889 } 890 891 int f2fs_truncate(struct inode *inode) 892 { 893 int err; 894 895 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 896 return -EIO; 897 898 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 899 S_ISLNK(inode->i_mode))) 900 return 0; 901 902 trace_f2fs_truncate(inode); 903 904 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 905 return -EIO; 906 907 err = f2fs_dquot_initialize(inode); 908 if (err) 909 return err; 910 911 /* we should check inline_data size */ 912 if (!f2fs_may_inline_data(inode)) { 913 err = f2fs_convert_inline_inode(inode); 914 if (err) { 915 /* 916 * Always truncate page #0 to avoid page cache 917 * leak in evict() path. 918 */ 919 truncate_inode_pages_range(inode->i_mapping, 920 F2FS_BLK_TO_BYTES(0), 921 F2FS_BLK_END_BYTES(0)); 922 return err; 923 } 924 } 925 926 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 927 if (err) 928 return err; 929 930 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 931 f2fs_mark_inode_dirty_sync(inode, false); 932 return 0; 933 } 934 935 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 936 { 937 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 938 939 if (!fscrypt_dio_supported(inode)) 940 return true; 941 if (fsverity_active(inode)) 942 return true; 943 if (f2fs_compressed_file(inode)) 944 return true; 945 /* 946 * only force direct read to use buffered IO, for direct write, 947 * it expects inline data conversion before committing IO. 948 */ 949 if (f2fs_has_inline_data(inode) && rw == READ) 950 return true; 951 952 /* disallow direct IO if any of devices has unaligned blksize */ 953 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 954 return true; 955 /* 956 * for blkzoned device, fallback direct IO to buffered IO, so 957 * all IOs can be serialized by log-structured write. 958 */ 959 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) && 960 !f2fs_is_pinned_file(inode)) 961 return true; 962 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 963 return true; 964 965 return false; 966 } 967 968 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 969 struct kstat *stat, u32 request_mask, unsigned int query_flags) 970 { 971 struct inode *inode = d_inode(path->dentry); 972 struct f2fs_inode_info *fi = F2FS_I(inode); 973 struct f2fs_inode *ri = NULL; 974 unsigned int flags; 975 976 if (f2fs_has_extra_attr(inode) && 977 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 978 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 979 stat->result_mask |= STATX_BTIME; 980 stat->btime.tv_sec = fi->i_crtime.tv_sec; 981 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 982 } 983 984 /* 985 * Return the DIO alignment restrictions if requested. We only return 986 * this information when requested, since on encrypted files it might 987 * take a fair bit of work to get if the file wasn't opened recently. 988 * 989 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 990 * cannot represent that, so in that case we report no DIO support. 991 */ 992 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 993 unsigned int bsize = i_blocksize(inode); 994 995 stat->result_mask |= STATX_DIOALIGN; 996 if (!f2fs_force_buffered_io(inode, WRITE)) { 997 stat->dio_mem_align = bsize; 998 stat->dio_offset_align = bsize; 999 } 1000 } 1001 1002 flags = fi->i_flags; 1003 if (flags & F2FS_COMPR_FL) 1004 stat->attributes |= STATX_ATTR_COMPRESSED; 1005 if (flags & F2FS_APPEND_FL) 1006 stat->attributes |= STATX_ATTR_APPEND; 1007 if (IS_ENCRYPTED(inode)) 1008 stat->attributes |= STATX_ATTR_ENCRYPTED; 1009 if (flags & F2FS_IMMUTABLE_FL) 1010 stat->attributes |= STATX_ATTR_IMMUTABLE; 1011 if (flags & F2FS_NODUMP_FL) 1012 stat->attributes |= STATX_ATTR_NODUMP; 1013 if (IS_VERITY(inode)) 1014 stat->attributes |= STATX_ATTR_VERITY; 1015 1016 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 1017 STATX_ATTR_APPEND | 1018 STATX_ATTR_ENCRYPTED | 1019 STATX_ATTR_IMMUTABLE | 1020 STATX_ATTR_NODUMP | 1021 STATX_ATTR_VERITY); 1022 1023 generic_fillattr(idmap, request_mask, inode, stat); 1024 1025 /* we need to show initial sectors used for inline_data/dentries */ 1026 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 1027 f2fs_has_inline_dentry(inode)) 1028 stat->blocks += (stat->size + 511) >> 9; 1029 1030 return 0; 1031 } 1032 1033 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1034 static void __setattr_copy(struct mnt_idmap *idmap, 1035 struct inode *inode, const struct iattr *attr) 1036 { 1037 unsigned int ia_valid = attr->ia_valid; 1038 1039 i_uid_update(idmap, attr, inode); 1040 i_gid_update(idmap, attr, inode); 1041 if (ia_valid & ATTR_ATIME) 1042 inode_set_atime_to_ts(inode, attr->ia_atime); 1043 if (ia_valid & ATTR_MTIME) 1044 inode_set_mtime_to_ts(inode, attr->ia_mtime); 1045 if (ia_valid & ATTR_CTIME) 1046 inode_set_ctime_to_ts(inode, attr->ia_ctime); 1047 if (ia_valid & ATTR_MODE) { 1048 umode_t mode = attr->ia_mode; 1049 1050 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode))) 1051 mode &= ~S_ISGID; 1052 set_acl_inode(inode, mode); 1053 } 1054 } 1055 #else 1056 #define __setattr_copy setattr_copy 1057 #endif 1058 1059 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 1060 struct iattr *attr) 1061 { 1062 struct inode *inode = d_inode(dentry); 1063 struct f2fs_inode_info *fi = F2FS_I(inode); 1064 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1065 int err; 1066 1067 if (unlikely(f2fs_cp_error(sbi))) 1068 return -EIO; 1069 1070 err = setattr_prepare(idmap, dentry, attr); 1071 if (err) 1072 return err; 1073 1074 err = fscrypt_prepare_setattr(dentry, attr); 1075 if (err) 1076 return err; 1077 1078 err = fsverity_prepare_setattr(dentry, attr); 1079 if (err) 1080 return err; 1081 1082 if (unlikely(IS_IMMUTABLE(inode))) 1083 return -EPERM; 1084 1085 if (unlikely(IS_APPEND(inode) && 1086 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 1087 ATTR_GID | ATTR_TIMES_SET)))) 1088 return -EPERM; 1089 1090 if ((attr->ia_valid & ATTR_SIZE)) { 1091 if (!f2fs_is_compress_backend_ready(inode) || 1092 IS_DEVICE_ALIASING(inode)) 1093 return -EOPNOTSUPP; 1094 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 1095 !IS_ALIGNED(attr->ia_size, 1096 F2FS_BLK_TO_BYTES(fi->i_cluster_size))) 1097 return -EINVAL; 1098 /* 1099 * To prevent scattered pin block generation, we don't allow 1100 * smaller/equal size unaligned truncation for pinned file. 1101 * We only support overwrite IO to pinned file, so don't 1102 * care about larger size truncation. 1103 */ 1104 if (f2fs_is_pinned_file(inode) && 1105 attr->ia_size <= i_size_read(inode) && 1106 !IS_ALIGNED(attr->ia_size, 1107 F2FS_BLK_TO_BYTES(CAP_BLKS_PER_SEC(sbi)))) 1108 return -EINVAL; 1109 } 1110 1111 if (is_quota_modification(idmap, inode, attr)) { 1112 err = f2fs_dquot_initialize(inode); 1113 if (err) 1114 return err; 1115 } 1116 if (i_uid_needs_update(idmap, attr, inode) || 1117 i_gid_needs_update(idmap, attr, inode)) { 1118 f2fs_lock_op(sbi); 1119 err = dquot_transfer(idmap, inode, attr); 1120 if (err) { 1121 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 1122 f2fs_unlock_op(sbi); 1123 return err; 1124 } 1125 /* 1126 * update uid/gid under lock_op(), so that dquot and inode can 1127 * be updated atomically. 1128 */ 1129 i_uid_update(idmap, attr, inode); 1130 i_gid_update(idmap, attr, inode); 1131 f2fs_mark_inode_dirty_sync(inode, true); 1132 f2fs_unlock_op(sbi); 1133 } 1134 1135 if (attr->ia_valid & ATTR_SIZE) { 1136 loff_t old_size = i_size_read(inode); 1137 1138 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1139 /* 1140 * should convert inline inode before i_size_write to 1141 * keep smaller than inline_data size with inline flag. 1142 */ 1143 err = f2fs_convert_inline_inode(inode); 1144 if (err) 1145 return err; 1146 } 1147 1148 /* 1149 * wait for inflight dio, blocks should be removed after 1150 * IO completion. 1151 */ 1152 if (attr->ia_size < old_size) 1153 inode_dio_wait(inode); 1154 1155 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 1156 filemap_invalidate_lock(inode->i_mapping); 1157 1158 if (attr->ia_size > old_size) 1159 f2fs_zero_post_eof_page(inode, attr->ia_size, false); 1160 truncate_setsize(inode, attr->ia_size); 1161 1162 if (attr->ia_size <= old_size) 1163 err = f2fs_truncate(inode); 1164 /* 1165 * do not trim all blocks after i_size if target size is 1166 * larger than i_size. 1167 */ 1168 filemap_invalidate_unlock(inode->i_mapping); 1169 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 1170 if (err) 1171 return err; 1172 1173 spin_lock(&fi->i_size_lock); 1174 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1175 fi->last_disk_size = i_size_read(inode); 1176 spin_unlock(&fi->i_size_lock); 1177 } 1178 1179 __setattr_copy(idmap, inode, attr); 1180 1181 if (attr->ia_valid & ATTR_MODE) { 1182 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1183 1184 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1185 if (!err) 1186 inode->i_mode = fi->i_acl_mode; 1187 clear_inode_flag(inode, FI_ACL_MODE); 1188 } 1189 } 1190 1191 /* file size may changed here */ 1192 f2fs_mark_inode_dirty_sync(inode, true); 1193 1194 /* inode change will produce dirty node pages flushed by checkpoint */ 1195 f2fs_balance_fs(sbi, true); 1196 1197 return err; 1198 } 1199 1200 const struct inode_operations f2fs_file_inode_operations = { 1201 .getattr = f2fs_getattr, 1202 .setattr = f2fs_setattr, 1203 .get_inode_acl = f2fs_get_acl, 1204 .set_acl = f2fs_set_acl, 1205 .listxattr = f2fs_listxattr, 1206 .fiemap = f2fs_fiemap, 1207 .fileattr_get = f2fs_fileattr_get, 1208 .fileattr_set = f2fs_fileattr_set, 1209 }; 1210 1211 static int fill_zero(struct inode *inode, pgoff_t index, 1212 loff_t start, loff_t len) 1213 { 1214 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1215 struct folio *folio; 1216 1217 if (!len) 1218 return 0; 1219 1220 f2fs_balance_fs(sbi, true); 1221 1222 f2fs_lock_op(sbi); 1223 folio = f2fs_get_new_data_folio(inode, NULL, index, false); 1224 f2fs_unlock_op(sbi); 1225 1226 if (IS_ERR(folio)) 1227 return PTR_ERR(folio); 1228 1229 f2fs_folio_wait_writeback(folio, DATA, true, true); 1230 folio_zero_range(folio, start, len); 1231 folio_mark_dirty(folio); 1232 f2fs_folio_put(folio, true); 1233 return 0; 1234 } 1235 1236 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1237 { 1238 int err; 1239 1240 while (pg_start < pg_end) { 1241 struct dnode_of_data dn; 1242 pgoff_t end_offset, count; 1243 1244 set_new_dnode(&dn, inode, NULL, NULL, 0); 1245 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1246 if (err) { 1247 if (err == -ENOENT) { 1248 pg_start = f2fs_get_next_page_offset(&dn, 1249 pg_start); 1250 continue; 1251 } 1252 return err; 1253 } 1254 1255 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 1256 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1257 1258 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1259 1260 f2fs_truncate_data_blocks_range(&dn, count); 1261 f2fs_put_dnode(&dn); 1262 1263 pg_start += count; 1264 } 1265 return 0; 1266 } 1267 1268 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1269 { 1270 pgoff_t pg_start, pg_end; 1271 loff_t off_start, off_end; 1272 int ret; 1273 1274 ret = f2fs_convert_inline_inode(inode); 1275 if (ret) 1276 return ret; 1277 1278 f2fs_zero_post_eof_page(inode, offset + len, true); 1279 1280 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1281 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1282 1283 off_start = offset & (PAGE_SIZE - 1); 1284 off_end = (offset + len) & (PAGE_SIZE - 1); 1285 1286 if (pg_start == pg_end) { 1287 ret = fill_zero(inode, pg_start, off_start, 1288 off_end - off_start); 1289 if (ret) 1290 return ret; 1291 } else { 1292 if (off_start) { 1293 ret = fill_zero(inode, pg_start++, off_start, 1294 PAGE_SIZE - off_start); 1295 if (ret) 1296 return ret; 1297 } 1298 if (off_end) { 1299 ret = fill_zero(inode, pg_end, 0, off_end); 1300 if (ret) 1301 return ret; 1302 } 1303 1304 if (pg_start < pg_end) { 1305 loff_t blk_start, blk_end; 1306 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1307 1308 f2fs_balance_fs(sbi, true); 1309 1310 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1311 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1312 1313 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1314 filemap_invalidate_lock(inode->i_mapping); 1315 1316 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1317 1318 f2fs_lock_op(sbi); 1319 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1320 f2fs_unlock_op(sbi); 1321 1322 filemap_invalidate_unlock(inode->i_mapping); 1323 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1324 } 1325 } 1326 1327 return ret; 1328 } 1329 1330 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1331 int *do_replace, pgoff_t off, pgoff_t len) 1332 { 1333 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1334 struct dnode_of_data dn; 1335 int ret, done, i; 1336 1337 next_dnode: 1338 set_new_dnode(&dn, inode, NULL, NULL, 0); 1339 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1340 if (ret && ret != -ENOENT) { 1341 return ret; 1342 } else if (ret == -ENOENT) { 1343 if (dn.max_level == 0) 1344 return -ENOENT; 1345 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1346 dn.ofs_in_node, len); 1347 blkaddr += done; 1348 do_replace += done; 1349 goto next; 1350 } 1351 1352 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_folio, inode) - 1353 dn.ofs_in_node, len); 1354 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1355 *blkaddr = f2fs_data_blkaddr(&dn); 1356 1357 if (__is_valid_data_blkaddr(*blkaddr) && 1358 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1359 DATA_GENERIC_ENHANCE)) { 1360 f2fs_put_dnode(&dn); 1361 return -EFSCORRUPTED; 1362 } 1363 1364 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1365 1366 if (f2fs_lfs_mode(sbi)) { 1367 f2fs_put_dnode(&dn); 1368 return -EOPNOTSUPP; 1369 } 1370 1371 /* do not invalidate this block address */ 1372 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1373 *do_replace = 1; 1374 } 1375 } 1376 f2fs_put_dnode(&dn); 1377 next: 1378 len -= done; 1379 off += done; 1380 if (len) 1381 goto next_dnode; 1382 return 0; 1383 } 1384 1385 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1386 int *do_replace, pgoff_t off, int len) 1387 { 1388 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1389 struct dnode_of_data dn; 1390 int ret, i; 1391 1392 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1393 if (*do_replace == 0) 1394 continue; 1395 1396 set_new_dnode(&dn, inode, NULL, NULL, 0); 1397 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1398 if (ret) { 1399 dec_valid_block_count(sbi, inode, 1); 1400 f2fs_invalidate_blocks(sbi, *blkaddr, 1); 1401 } else { 1402 f2fs_update_data_blkaddr(&dn, *blkaddr); 1403 } 1404 f2fs_put_dnode(&dn); 1405 } 1406 return 0; 1407 } 1408 1409 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1410 block_t *blkaddr, int *do_replace, 1411 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1412 { 1413 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1414 pgoff_t i = 0; 1415 int ret; 1416 1417 while (i < len) { 1418 if (blkaddr[i] == NULL_ADDR && !full) { 1419 i++; 1420 continue; 1421 } 1422 1423 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1424 struct dnode_of_data dn; 1425 struct node_info ni; 1426 size_t new_size; 1427 pgoff_t ilen; 1428 1429 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1430 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1431 if (ret) 1432 return ret; 1433 1434 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1435 if (ret) { 1436 f2fs_put_dnode(&dn); 1437 return ret; 1438 } 1439 1440 ilen = min((pgoff_t) 1441 ADDRS_PER_PAGE(dn.node_folio, dst_inode) - 1442 dn.ofs_in_node, len - i); 1443 do { 1444 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1445 f2fs_truncate_data_blocks_range(&dn, 1); 1446 1447 if (do_replace[i]) { 1448 f2fs_i_blocks_write(src_inode, 1449 1, false, false); 1450 f2fs_i_blocks_write(dst_inode, 1451 1, true, false); 1452 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1453 blkaddr[i], ni.version, true, false); 1454 1455 do_replace[i] = 0; 1456 } 1457 dn.ofs_in_node++; 1458 i++; 1459 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1460 if (dst_inode->i_size < new_size) 1461 f2fs_i_size_write(dst_inode, new_size); 1462 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1463 1464 f2fs_put_dnode(&dn); 1465 } else { 1466 struct folio *fsrc, *fdst; 1467 1468 fsrc = f2fs_get_lock_data_folio(src_inode, 1469 src + i, true); 1470 if (IS_ERR(fsrc)) 1471 return PTR_ERR(fsrc); 1472 fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i, 1473 true); 1474 if (IS_ERR(fdst)) { 1475 f2fs_folio_put(fsrc, true); 1476 return PTR_ERR(fdst); 1477 } 1478 1479 f2fs_folio_wait_writeback(fdst, DATA, true, true); 1480 1481 memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE); 1482 folio_mark_dirty(fdst); 1483 folio_set_f2fs_gcing(fdst); 1484 f2fs_folio_put(fdst, true); 1485 f2fs_folio_put(fsrc, true); 1486 1487 ret = f2fs_truncate_hole(src_inode, 1488 src + i, src + i + 1); 1489 if (ret) 1490 return ret; 1491 i++; 1492 } 1493 } 1494 return 0; 1495 } 1496 1497 static int __exchange_data_block(struct inode *src_inode, 1498 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1499 pgoff_t len, bool full) 1500 { 1501 block_t *src_blkaddr; 1502 int *do_replace; 1503 pgoff_t olen; 1504 int ret; 1505 1506 while (len) { 1507 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1508 1509 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1510 array_size(olen, sizeof(block_t)), 1511 GFP_NOFS); 1512 if (!src_blkaddr) 1513 return -ENOMEM; 1514 1515 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1516 array_size(olen, sizeof(int)), 1517 GFP_NOFS); 1518 if (!do_replace) { 1519 kvfree(src_blkaddr); 1520 return -ENOMEM; 1521 } 1522 1523 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1524 do_replace, src, olen); 1525 if (ret) 1526 goto roll_back; 1527 1528 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1529 do_replace, src, dst, olen, full); 1530 if (ret) 1531 goto roll_back; 1532 1533 src += olen; 1534 dst += olen; 1535 len -= olen; 1536 1537 kvfree(src_blkaddr); 1538 kvfree(do_replace); 1539 } 1540 return 0; 1541 1542 roll_back: 1543 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1544 kvfree(src_blkaddr); 1545 kvfree(do_replace); 1546 return ret; 1547 } 1548 1549 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1550 { 1551 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1552 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1553 pgoff_t start = offset >> PAGE_SHIFT; 1554 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1555 int ret; 1556 1557 f2fs_balance_fs(sbi, true); 1558 1559 /* avoid gc operation during block exchange */ 1560 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1561 filemap_invalidate_lock(inode->i_mapping); 1562 1563 f2fs_zero_post_eof_page(inode, offset + len, false); 1564 1565 f2fs_lock_op(sbi); 1566 f2fs_drop_extent_tree(inode); 1567 truncate_pagecache(inode, offset); 1568 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1569 f2fs_unlock_op(sbi); 1570 1571 filemap_invalidate_unlock(inode->i_mapping); 1572 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1573 return ret; 1574 } 1575 1576 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1577 { 1578 loff_t new_size; 1579 int ret; 1580 1581 if (offset + len >= i_size_read(inode)) 1582 return -EINVAL; 1583 1584 /* collapse range should be aligned to block size of f2fs. */ 1585 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1586 return -EINVAL; 1587 1588 ret = f2fs_convert_inline_inode(inode); 1589 if (ret) 1590 return ret; 1591 1592 /* write out all dirty pages from offset */ 1593 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1594 if (ret) 1595 return ret; 1596 1597 ret = f2fs_do_collapse(inode, offset, len); 1598 if (ret) 1599 return ret; 1600 1601 /* write out all moved pages, if possible */ 1602 filemap_invalidate_lock(inode->i_mapping); 1603 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1604 truncate_pagecache(inode, offset); 1605 1606 new_size = i_size_read(inode) - len; 1607 ret = f2fs_truncate_blocks(inode, new_size, true); 1608 filemap_invalidate_unlock(inode->i_mapping); 1609 if (!ret) 1610 f2fs_i_size_write(inode, new_size); 1611 return ret; 1612 } 1613 1614 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1615 pgoff_t end) 1616 { 1617 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1618 pgoff_t index = start; 1619 unsigned int ofs_in_node = dn->ofs_in_node; 1620 blkcnt_t count = 0; 1621 int ret; 1622 1623 for (; index < end; index++, dn->ofs_in_node++) { 1624 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1625 count++; 1626 } 1627 1628 dn->ofs_in_node = ofs_in_node; 1629 ret = f2fs_reserve_new_blocks(dn, count); 1630 if (ret) 1631 return ret; 1632 1633 dn->ofs_in_node = ofs_in_node; 1634 for (index = start; index < end; index++, dn->ofs_in_node++) { 1635 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1636 /* 1637 * f2fs_reserve_new_blocks will not guarantee entire block 1638 * allocation. 1639 */ 1640 if (dn->data_blkaddr == NULL_ADDR) { 1641 ret = -ENOSPC; 1642 break; 1643 } 1644 1645 if (dn->data_blkaddr == NEW_ADDR) 1646 continue; 1647 1648 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1649 DATA_GENERIC_ENHANCE)) { 1650 ret = -EFSCORRUPTED; 1651 break; 1652 } 1653 1654 f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1); 1655 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1656 } 1657 1658 if (index > start) { 1659 f2fs_update_read_extent_cache_range(dn, start, 0, 1660 index - start); 1661 f2fs_update_age_extent_cache_range(dn, start, index - start); 1662 } 1663 1664 return ret; 1665 } 1666 1667 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1668 int mode) 1669 { 1670 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1671 struct address_space *mapping = inode->i_mapping; 1672 pgoff_t index, pg_start, pg_end; 1673 loff_t new_size = i_size_read(inode); 1674 loff_t off_start, off_end; 1675 int ret = 0; 1676 1677 ret = inode_newsize_ok(inode, (len + offset)); 1678 if (ret) 1679 return ret; 1680 1681 ret = f2fs_convert_inline_inode(inode); 1682 if (ret) 1683 return ret; 1684 1685 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1686 if (ret) 1687 return ret; 1688 1689 f2fs_zero_post_eof_page(inode, offset + len, true); 1690 1691 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1692 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1693 1694 off_start = offset & (PAGE_SIZE - 1); 1695 off_end = (offset + len) & (PAGE_SIZE - 1); 1696 1697 if (pg_start == pg_end) { 1698 ret = fill_zero(inode, pg_start, off_start, 1699 off_end - off_start); 1700 if (ret) 1701 return ret; 1702 1703 new_size = max_t(loff_t, new_size, offset + len); 1704 } else { 1705 if (off_start) { 1706 ret = fill_zero(inode, pg_start++, off_start, 1707 PAGE_SIZE - off_start); 1708 if (ret) 1709 return ret; 1710 1711 new_size = max_t(loff_t, new_size, 1712 (loff_t)pg_start << PAGE_SHIFT); 1713 } 1714 1715 for (index = pg_start; index < pg_end;) { 1716 struct dnode_of_data dn; 1717 unsigned int end_offset; 1718 pgoff_t end; 1719 1720 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1721 filemap_invalidate_lock(mapping); 1722 1723 truncate_pagecache_range(inode, 1724 (loff_t)index << PAGE_SHIFT, 1725 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1726 1727 f2fs_lock_op(sbi); 1728 1729 set_new_dnode(&dn, inode, NULL, NULL, 0); 1730 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1731 if (ret) { 1732 f2fs_unlock_op(sbi); 1733 filemap_invalidate_unlock(mapping); 1734 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1735 goto out; 1736 } 1737 1738 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 1739 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1740 1741 ret = f2fs_do_zero_range(&dn, index, end); 1742 f2fs_put_dnode(&dn); 1743 1744 f2fs_unlock_op(sbi); 1745 filemap_invalidate_unlock(mapping); 1746 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1747 1748 f2fs_balance_fs(sbi, dn.node_changed); 1749 1750 if (ret) 1751 goto out; 1752 1753 index = end; 1754 new_size = max_t(loff_t, new_size, 1755 (loff_t)index << PAGE_SHIFT); 1756 } 1757 1758 if (off_end) { 1759 ret = fill_zero(inode, pg_end, 0, off_end); 1760 if (ret) 1761 goto out; 1762 1763 new_size = max_t(loff_t, new_size, offset + len); 1764 } 1765 } 1766 1767 out: 1768 if (new_size > i_size_read(inode)) { 1769 if (mode & FALLOC_FL_KEEP_SIZE) 1770 file_set_keep_isize(inode); 1771 else 1772 f2fs_i_size_write(inode, new_size); 1773 } 1774 return ret; 1775 } 1776 1777 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1778 { 1779 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1780 struct address_space *mapping = inode->i_mapping; 1781 pgoff_t nr, pg_start, pg_end, delta, idx; 1782 loff_t new_size; 1783 int ret = 0; 1784 1785 new_size = i_size_read(inode) + len; 1786 ret = inode_newsize_ok(inode, new_size); 1787 if (ret) 1788 return ret; 1789 1790 if (offset >= i_size_read(inode)) 1791 return -EINVAL; 1792 1793 /* insert range should be aligned to block size of f2fs. */ 1794 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1795 return -EINVAL; 1796 1797 ret = f2fs_convert_inline_inode(inode); 1798 if (ret) 1799 return ret; 1800 1801 f2fs_balance_fs(sbi, true); 1802 1803 filemap_invalidate_lock(mapping); 1804 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1805 filemap_invalidate_unlock(mapping); 1806 if (ret) 1807 return ret; 1808 1809 /* write out all dirty pages from offset */ 1810 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1811 if (ret) 1812 return ret; 1813 1814 pg_start = offset >> PAGE_SHIFT; 1815 pg_end = (offset + len) >> PAGE_SHIFT; 1816 delta = pg_end - pg_start; 1817 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1818 1819 /* avoid gc operation during block exchange */ 1820 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1821 filemap_invalidate_lock(mapping); 1822 1823 f2fs_zero_post_eof_page(inode, offset + len, false); 1824 truncate_pagecache(inode, offset); 1825 1826 while (!ret && idx > pg_start) { 1827 nr = idx - pg_start; 1828 if (nr > delta) 1829 nr = delta; 1830 idx -= nr; 1831 1832 f2fs_lock_op(sbi); 1833 f2fs_drop_extent_tree(inode); 1834 1835 ret = __exchange_data_block(inode, inode, idx, 1836 idx + delta, nr, false); 1837 f2fs_unlock_op(sbi); 1838 } 1839 filemap_invalidate_unlock(mapping); 1840 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1841 if (ret) 1842 return ret; 1843 1844 /* write out all moved pages, if possible */ 1845 filemap_invalidate_lock(mapping); 1846 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1847 truncate_pagecache(inode, offset); 1848 filemap_invalidate_unlock(mapping); 1849 1850 if (!ret) 1851 f2fs_i_size_write(inode, new_size); 1852 return ret; 1853 } 1854 1855 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1856 loff_t len, int mode) 1857 { 1858 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1859 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1860 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1861 .m_may_create = true }; 1862 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1863 .init_gc_type = FG_GC, 1864 .should_migrate_blocks = false, 1865 .err_gc_skipped = true, 1866 .nr_free_secs = 0 }; 1867 pgoff_t pg_start, pg_end; 1868 loff_t new_size; 1869 loff_t off_end; 1870 block_t expanded = 0; 1871 int err; 1872 1873 err = inode_newsize_ok(inode, (len + offset)); 1874 if (err) 1875 return err; 1876 1877 err = f2fs_convert_inline_inode(inode); 1878 if (err) 1879 return err; 1880 1881 f2fs_zero_post_eof_page(inode, offset + len, true); 1882 1883 f2fs_balance_fs(sbi, true); 1884 1885 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1886 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1887 off_end = (offset + len) & (PAGE_SIZE - 1); 1888 1889 map.m_lblk = pg_start; 1890 map.m_len = pg_end - pg_start; 1891 if (off_end) 1892 map.m_len++; 1893 1894 if (!map.m_len) 1895 return 0; 1896 1897 if (f2fs_is_pinned_file(inode)) { 1898 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1899 block_t sec_len = roundup(map.m_len, sec_blks); 1900 1901 map.m_len = sec_blks; 1902 next_alloc: 1903 f2fs_down_write(&sbi->pin_sem); 1904 1905 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1906 if (has_not_enough_free_secs(sbi, 0, 0)) { 1907 f2fs_up_write(&sbi->pin_sem); 1908 err = -ENOSPC; 1909 f2fs_warn_ratelimited(sbi, 1910 "ino:%lu, start:%lu, end:%lu, need to trigger GC to " 1911 "reclaim enough free segment when checkpoint is enabled", 1912 inode->i_ino, pg_start, pg_end); 1913 goto out_err; 1914 } 1915 } 1916 1917 if (has_not_enough_free_secs(sbi, 0, 1918 sbi->reserved_pin_section)) { 1919 f2fs_down_write(&sbi->gc_lock); 1920 stat_inc_gc_call_count(sbi, FOREGROUND); 1921 err = f2fs_gc(sbi, &gc_control); 1922 if (err && err != -ENODATA) { 1923 f2fs_up_write(&sbi->pin_sem); 1924 goto out_err; 1925 } 1926 } 1927 1928 err = f2fs_allocate_pinning_section(sbi); 1929 if (err) { 1930 f2fs_up_write(&sbi->pin_sem); 1931 goto out_err; 1932 } 1933 1934 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1935 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1936 file_dont_truncate(inode); 1937 1938 f2fs_up_write(&sbi->pin_sem); 1939 1940 expanded += map.m_len; 1941 sec_len -= map.m_len; 1942 map.m_lblk += map.m_len; 1943 if (!err && sec_len) 1944 goto next_alloc; 1945 1946 map.m_len = expanded; 1947 } else { 1948 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1949 expanded = map.m_len; 1950 } 1951 out_err: 1952 if (err) { 1953 pgoff_t last_off; 1954 1955 if (!expanded) 1956 return err; 1957 1958 last_off = pg_start + expanded - 1; 1959 1960 /* update new size to the failed position */ 1961 new_size = (last_off == pg_end) ? offset + len : 1962 (loff_t)(last_off + 1) << PAGE_SHIFT; 1963 } else { 1964 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1965 } 1966 1967 if (new_size > i_size_read(inode)) { 1968 if (mode & FALLOC_FL_KEEP_SIZE) 1969 file_set_keep_isize(inode); 1970 else 1971 f2fs_i_size_write(inode, new_size); 1972 } 1973 1974 return err; 1975 } 1976 1977 static long f2fs_fallocate(struct file *file, int mode, 1978 loff_t offset, loff_t len) 1979 { 1980 struct inode *inode = file_inode(file); 1981 long ret = 0; 1982 1983 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1984 return -EIO; 1985 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1986 return -ENOSPC; 1987 if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode)) 1988 return -EOPNOTSUPP; 1989 1990 /* f2fs only support ->fallocate for regular file */ 1991 if (!S_ISREG(inode->i_mode)) 1992 return -EINVAL; 1993 1994 if (IS_ENCRYPTED(inode) && 1995 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1996 return -EOPNOTSUPP; 1997 1998 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1999 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 2000 FALLOC_FL_INSERT_RANGE)) 2001 return -EOPNOTSUPP; 2002 2003 inode_lock(inode); 2004 2005 /* 2006 * Pinned file should not support partial truncation since the block 2007 * can be used by applications. 2008 */ 2009 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 2010 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 2011 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 2012 ret = -EOPNOTSUPP; 2013 goto out; 2014 } 2015 2016 ret = file_modified(file); 2017 if (ret) 2018 goto out; 2019 2020 /* 2021 * wait for inflight dio, blocks should be removed after IO 2022 * completion. 2023 */ 2024 inode_dio_wait(inode); 2025 2026 if (mode & FALLOC_FL_PUNCH_HOLE) { 2027 if (offset >= inode->i_size) 2028 goto out; 2029 2030 ret = f2fs_punch_hole(inode, offset, len); 2031 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 2032 ret = f2fs_collapse_range(inode, offset, len); 2033 } else if (mode & FALLOC_FL_ZERO_RANGE) { 2034 ret = f2fs_zero_range(inode, offset, len, mode); 2035 } else if (mode & FALLOC_FL_INSERT_RANGE) { 2036 ret = f2fs_insert_range(inode, offset, len); 2037 } else { 2038 ret = f2fs_expand_inode_data(inode, offset, len, mode); 2039 } 2040 2041 if (!ret) { 2042 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 2043 f2fs_mark_inode_dirty_sync(inode, false); 2044 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2045 } 2046 2047 out: 2048 inode_unlock(inode); 2049 2050 trace_f2fs_fallocate(inode, mode, offset, len, ret); 2051 return ret; 2052 } 2053 2054 static int f2fs_release_file(struct inode *inode, struct file *filp) 2055 { 2056 if (atomic_dec_and_test(&F2FS_I(inode)->open_count)) 2057 f2fs_remove_donate_inode(inode); 2058 2059 /* 2060 * f2fs_release_file is called at every close calls. So we should 2061 * not drop any inmemory pages by close called by other process. 2062 */ 2063 if (!(filp->f_mode & FMODE_WRITE) || 2064 atomic_read(&inode->i_writecount) != 1) 2065 return 0; 2066 2067 inode_lock(inode); 2068 f2fs_abort_atomic_write(inode, true); 2069 inode_unlock(inode); 2070 2071 return 0; 2072 } 2073 2074 static int f2fs_file_flush(struct file *file, fl_owner_t id) 2075 { 2076 struct inode *inode = file_inode(file); 2077 2078 /* 2079 * If the process doing a transaction is crashed, we should do 2080 * roll-back. Otherwise, other reader/write can see corrupted database 2081 * until all the writers close its file. Since this should be done 2082 * before dropping file lock, it needs to do in ->flush. 2083 */ 2084 if (F2FS_I(inode)->atomic_write_task == current && 2085 (current->flags & PF_EXITING)) { 2086 inode_lock(inode); 2087 f2fs_abort_atomic_write(inode, true); 2088 inode_unlock(inode); 2089 } 2090 2091 return 0; 2092 } 2093 2094 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 2095 { 2096 struct f2fs_inode_info *fi = F2FS_I(inode); 2097 u32 masked_flags = fi->i_flags & mask; 2098 2099 /* mask can be shrunk by flags_valid selector */ 2100 iflags &= mask; 2101 2102 /* Is it quota file? Do not allow user to mess with it */ 2103 if (IS_NOQUOTA(inode)) 2104 return -EPERM; 2105 2106 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 2107 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 2108 return -EOPNOTSUPP; 2109 if (!f2fs_empty_dir(inode)) 2110 return -ENOTEMPTY; 2111 } 2112 2113 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 2114 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 2115 return -EOPNOTSUPP; 2116 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 2117 return -EINVAL; 2118 } 2119 2120 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 2121 if (masked_flags & F2FS_COMPR_FL) { 2122 if (!f2fs_disable_compressed_file(inode)) 2123 return -EINVAL; 2124 } else { 2125 /* try to convert inline_data to support compression */ 2126 int err = f2fs_convert_inline_inode(inode); 2127 if (err) 2128 return err; 2129 2130 f2fs_down_write(&fi->i_sem); 2131 if (!f2fs_may_compress(inode) || 2132 atomic_read(&fi->writeback) || 2133 (S_ISREG(inode->i_mode) && 2134 F2FS_HAS_BLOCKS(inode))) { 2135 f2fs_up_write(&fi->i_sem); 2136 return -EINVAL; 2137 } 2138 err = set_compress_context(inode); 2139 f2fs_up_write(&fi->i_sem); 2140 2141 if (err) 2142 return err; 2143 } 2144 } 2145 2146 fi->i_flags = iflags | (fi->i_flags & ~mask); 2147 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 2148 (fi->i_flags & F2FS_NOCOMP_FL)); 2149 2150 if (fi->i_flags & F2FS_PROJINHERIT_FL) 2151 set_inode_flag(inode, FI_PROJ_INHERIT); 2152 else 2153 clear_inode_flag(inode, FI_PROJ_INHERIT); 2154 2155 inode_set_ctime_current(inode); 2156 f2fs_set_inode_flags(inode); 2157 f2fs_mark_inode_dirty_sync(inode, true); 2158 return 0; 2159 } 2160 2161 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 2162 2163 /* 2164 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 2165 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 2166 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 2167 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 2168 * 2169 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 2170 * FS_IOC_FSSETXATTR is done by the VFS. 2171 */ 2172 2173 static const struct { 2174 u32 iflag; 2175 u32 fsflag; 2176 } f2fs_fsflags_map[] = { 2177 { F2FS_COMPR_FL, FS_COMPR_FL }, 2178 { F2FS_SYNC_FL, FS_SYNC_FL }, 2179 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 2180 { F2FS_APPEND_FL, FS_APPEND_FL }, 2181 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 2182 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 2183 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 2184 { F2FS_INDEX_FL, FS_INDEX_FL }, 2185 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 2186 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2187 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2188 }; 2189 2190 #define F2FS_GETTABLE_FS_FL ( \ 2191 FS_COMPR_FL | \ 2192 FS_SYNC_FL | \ 2193 FS_IMMUTABLE_FL | \ 2194 FS_APPEND_FL | \ 2195 FS_NODUMP_FL | \ 2196 FS_NOATIME_FL | \ 2197 FS_NOCOMP_FL | \ 2198 FS_INDEX_FL | \ 2199 FS_DIRSYNC_FL | \ 2200 FS_PROJINHERIT_FL | \ 2201 FS_ENCRYPT_FL | \ 2202 FS_INLINE_DATA_FL | \ 2203 FS_NOCOW_FL | \ 2204 FS_VERITY_FL | \ 2205 FS_CASEFOLD_FL) 2206 2207 #define F2FS_SETTABLE_FS_FL ( \ 2208 FS_COMPR_FL | \ 2209 FS_SYNC_FL | \ 2210 FS_IMMUTABLE_FL | \ 2211 FS_APPEND_FL | \ 2212 FS_NODUMP_FL | \ 2213 FS_NOATIME_FL | \ 2214 FS_NOCOMP_FL | \ 2215 FS_DIRSYNC_FL | \ 2216 FS_PROJINHERIT_FL | \ 2217 FS_CASEFOLD_FL) 2218 2219 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2220 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2221 { 2222 u32 fsflags = 0; 2223 int i; 2224 2225 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2226 if (iflags & f2fs_fsflags_map[i].iflag) 2227 fsflags |= f2fs_fsflags_map[i].fsflag; 2228 2229 return fsflags; 2230 } 2231 2232 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2233 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2234 { 2235 u32 iflags = 0; 2236 int i; 2237 2238 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2239 if (fsflags & f2fs_fsflags_map[i].fsflag) 2240 iflags |= f2fs_fsflags_map[i].iflag; 2241 2242 return iflags; 2243 } 2244 2245 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2246 { 2247 struct inode *inode = file_inode(filp); 2248 2249 return put_user(inode->i_generation, (int __user *)arg); 2250 } 2251 2252 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2253 { 2254 struct inode *inode = file_inode(filp); 2255 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2256 struct f2fs_inode_info *fi = F2FS_I(inode); 2257 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2258 loff_t isize; 2259 int ret; 2260 2261 if (!(filp->f_mode & FMODE_WRITE)) 2262 return -EBADF; 2263 2264 if (!inode_owner_or_capable(idmap, inode)) 2265 return -EACCES; 2266 2267 if (!S_ISREG(inode->i_mode)) 2268 return -EINVAL; 2269 2270 if (filp->f_flags & O_DIRECT) 2271 return -EINVAL; 2272 2273 ret = mnt_want_write_file(filp); 2274 if (ret) 2275 return ret; 2276 2277 inode_lock(inode); 2278 2279 if (!f2fs_disable_compressed_file(inode) || 2280 f2fs_is_pinned_file(inode)) { 2281 ret = -EINVAL; 2282 goto out; 2283 } 2284 2285 if (f2fs_is_atomic_file(inode)) 2286 goto out; 2287 2288 ret = f2fs_convert_inline_inode(inode); 2289 if (ret) 2290 goto out; 2291 2292 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2293 f2fs_down_write(&fi->i_gc_rwsem[READ]); 2294 2295 /* 2296 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2297 * f2fs_is_atomic_file. 2298 */ 2299 if (get_dirty_pages(inode)) 2300 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2301 inode->i_ino, get_dirty_pages(inode)); 2302 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2303 if (ret) 2304 goto out_unlock; 2305 2306 /* Check if the inode already has a COW inode */ 2307 if (fi->cow_inode == NULL) { 2308 /* Create a COW inode for atomic write */ 2309 struct dentry *dentry = file_dentry(filp); 2310 struct inode *dir = d_inode(dentry->d_parent); 2311 2312 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode); 2313 if (ret) 2314 goto out_unlock; 2315 2316 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2317 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2318 2319 /* Set the COW inode's atomic_inode to the atomic inode */ 2320 F2FS_I(fi->cow_inode)->atomic_inode = inode; 2321 } else { 2322 /* Reuse the already created COW inode */ 2323 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode)); 2324 2325 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1); 2326 2327 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2328 if (ret) 2329 goto out_unlock; 2330 } 2331 2332 f2fs_write_inode(inode, NULL); 2333 2334 stat_inc_atomic_inode(inode); 2335 2336 set_inode_flag(inode, FI_ATOMIC_FILE); 2337 2338 isize = i_size_read(inode); 2339 fi->original_i_size = isize; 2340 if (truncate) { 2341 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2342 truncate_inode_pages_final(inode->i_mapping); 2343 f2fs_i_size_write(inode, 0); 2344 isize = 0; 2345 } 2346 f2fs_i_size_write(fi->cow_inode, isize); 2347 2348 out_unlock: 2349 f2fs_up_write(&fi->i_gc_rwsem[READ]); 2350 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2351 if (ret) 2352 goto out; 2353 2354 f2fs_update_time(sbi, REQ_TIME); 2355 fi->atomic_write_task = current; 2356 stat_update_max_atomic_write(inode); 2357 fi->atomic_write_cnt = 0; 2358 out: 2359 inode_unlock(inode); 2360 mnt_drop_write_file(filp); 2361 return ret; 2362 } 2363 2364 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2365 { 2366 struct inode *inode = file_inode(filp); 2367 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2368 int ret; 2369 2370 if (!(filp->f_mode & FMODE_WRITE)) 2371 return -EBADF; 2372 2373 if (!inode_owner_or_capable(idmap, inode)) 2374 return -EACCES; 2375 2376 ret = mnt_want_write_file(filp); 2377 if (ret) 2378 return ret; 2379 2380 f2fs_balance_fs(F2FS_I_SB(inode), true); 2381 2382 inode_lock(inode); 2383 2384 if (f2fs_is_atomic_file(inode)) { 2385 ret = f2fs_commit_atomic_write(inode); 2386 if (!ret) 2387 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2388 2389 f2fs_abort_atomic_write(inode, ret); 2390 } else { 2391 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2392 } 2393 2394 inode_unlock(inode); 2395 mnt_drop_write_file(filp); 2396 return ret; 2397 } 2398 2399 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2400 { 2401 struct inode *inode = file_inode(filp); 2402 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2403 int ret; 2404 2405 if (!(filp->f_mode & FMODE_WRITE)) 2406 return -EBADF; 2407 2408 if (!inode_owner_or_capable(idmap, inode)) 2409 return -EACCES; 2410 2411 ret = mnt_want_write_file(filp); 2412 if (ret) 2413 return ret; 2414 2415 inode_lock(inode); 2416 2417 f2fs_abort_atomic_write(inode, true); 2418 2419 inode_unlock(inode); 2420 2421 mnt_drop_write_file(filp); 2422 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2423 return ret; 2424 } 2425 2426 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag, 2427 bool readonly, bool need_lock) 2428 { 2429 struct super_block *sb = sbi->sb; 2430 int ret = 0; 2431 2432 switch (flag) { 2433 case F2FS_GOING_DOWN_FULLSYNC: 2434 ret = bdev_freeze(sb->s_bdev); 2435 if (ret) 2436 goto out; 2437 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2438 bdev_thaw(sb->s_bdev); 2439 break; 2440 case F2FS_GOING_DOWN_METASYNC: 2441 /* do checkpoint only */ 2442 ret = f2fs_sync_fs(sb, 1); 2443 if (ret) { 2444 if (ret == -EIO) 2445 ret = 0; 2446 goto out; 2447 } 2448 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2449 break; 2450 case F2FS_GOING_DOWN_NOSYNC: 2451 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2452 break; 2453 case F2FS_GOING_DOWN_METAFLUSH: 2454 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2455 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2456 break; 2457 case F2FS_GOING_DOWN_NEED_FSCK: 2458 set_sbi_flag(sbi, SBI_NEED_FSCK); 2459 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2460 set_sbi_flag(sbi, SBI_IS_DIRTY); 2461 /* do checkpoint only */ 2462 ret = f2fs_sync_fs(sb, 1); 2463 if (ret == -EIO) 2464 ret = 0; 2465 goto out; 2466 default: 2467 ret = -EINVAL; 2468 goto out; 2469 } 2470 2471 if (readonly) 2472 goto out; 2473 2474 /* 2475 * grab sb->s_umount to avoid racing w/ remount() and other shutdown 2476 * paths. 2477 */ 2478 if (need_lock) 2479 down_write(&sbi->sb->s_umount); 2480 2481 f2fs_stop_gc_thread(sbi); 2482 f2fs_stop_discard_thread(sbi); 2483 2484 f2fs_drop_discard_cmd(sbi); 2485 clear_opt(sbi, DISCARD); 2486 2487 if (need_lock) 2488 up_write(&sbi->sb->s_umount); 2489 2490 f2fs_update_time(sbi, REQ_TIME); 2491 out: 2492 2493 trace_f2fs_shutdown(sbi, flag, ret); 2494 2495 return ret; 2496 } 2497 2498 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2499 { 2500 struct inode *inode = file_inode(filp); 2501 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2502 __u32 in; 2503 int ret; 2504 bool need_drop = false, readonly = false; 2505 2506 if (!capable(CAP_SYS_ADMIN)) 2507 return -EPERM; 2508 2509 if (get_user(in, (__u32 __user *)arg)) 2510 return -EFAULT; 2511 2512 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2513 ret = mnt_want_write_file(filp); 2514 if (ret) { 2515 if (ret != -EROFS) 2516 return ret; 2517 2518 /* fallback to nosync shutdown for readonly fs */ 2519 in = F2FS_GOING_DOWN_NOSYNC; 2520 readonly = true; 2521 } else { 2522 need_drop = true; 2523 } 2524 } 2525 2526 ret = f2fs_do_shutdown(sbi, in, readonly, true); 2527 2528 if (need_drop) 2529 mnt_drop_write_file(filp); 2530 2531 return ret; 2532 } 2533 2534 static int f2fs_keep_noreuse_range(struct inode *inode, 2535 loff_t offset, loff_t len) 2536 { 2537 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2538 u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 2539 u64 start, end; 2540 int ret = 0; 2541 2542 if (!S_ISREG(inode->i_mode)) 2543 return 0; 2544 2545 if (offset >= max_bytes || len > max_bytes || 2546 (offset + len) > max_bytes) 2547 return 0; 2548 2549 start = offset >> PAGE_SHIFT; 2550 end = DIV_ROUND_UP(offset + len, PAGE_SIZE); 2551 2552 inode_lock(inode); 2553 if (f2fs_is_atomic_file(inode)) { 2554 inode_unlock(inode); 2555 return 0; 2556 } 2557 2558 spin_lock(&sbi->inode_lock[DONATE_INODE]); 2559 /* let's remove the range, if len = 0 */ 2560 if (!len) { 2561 if (!list_empty(&F2FS_I(inode)->gdonate_list)) { 2562 list_del_init(&F2FS_I(inode)->gdonate_list); 2563 sbi->donate_files--; 2564 if (is_inode_flag_set(inode, FI_DONATE_FINISHED)) 2565 ret = -EALREADY; 2566 else 2567 set_inode_flag(inode, FI_DONATE_FINISHED); 2568 } else 2569 ret = -ENOENT; 2570 } else { 2571 if (list_empty(&F2FS_I(inode)->gdonate_list)) { 2572 list_add_tail(&F2FS_I(inode)->gdonate_list, 2573 &sbi->inode_list[DONATE_INODE]); 2574 sbi->donate_files++; 2575 } else { 2576 list_move_tail(&F2FS_I(inode)->gdonate_list, 2577 &sbi->inode_list[DONATE_INODE]); 2578 } 2579 F2FS_I(inode)->donate_start = start; 2580 F2FS_I(inode)->donate_end = end - 1; 2581 clear_inode_flag(inode, FI_DONATE_FINISHED); 2582 } 2583 spin_unlock(&sbi->inode_lock[DONATE_INODE]); 2584 inode_unlock(inode); 2585 2586 return ret; 2587 } 2588 2589 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2590 { 2591 struct inode *inode = file_inode(filp); 2592 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2593 struct fstrim_range range; 2594 int ret; 2595 2596 if (!capable(CAP_SYS_ADMIN)) 2597 return -EPERM; 2598 2599 if (!f2fs_hw_support_discard(sbi)) 2600 return -EOPNOTSUPP; 2601 2602 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2603 sizeof(range))) 2604 return -EFAULT; 2605 2606 ret = mnt_want_write_file(filp); 2607 if (ret) 2608 return ret; 2609 2610 range.minlen = max_t(unsigned int, range.minlen, 2611 f2fs_hw_discard_granularity(sbi)); 2612 ret = f2fs_trim_fs(sbi, &range); 2613 mnt_drop_write_file(filp); 2614 if (ret < 0) 2615 return ret; 2616 2617 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2618 sizeof(range))) 2619 return -EFAULT; 2620 f2fs_update_time(sbi, REQ_TIME); 2621 return 0; 2622 } 2623 2624 static bool uuid_is_nonzero(__u8 u[16]) 2625 { 2626 int i; 2627 2628 for (i = 0; i < 16; i++) 2629 if (u[i]) 2630 return true; 2631 return false; 2632 } 2633 2634 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2635 { 2636 struct inode *inode = file_inode(filp); 2637 int ret; 2638 2639 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2640 return -EOPNOTSUPP; 2641 2642 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2643 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2644 return ret; 2645 } 2646 2647 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2648 { 2649 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2650 return -EOPNOTSUPP; 2651 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2652 } 2653 2654 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2655 { 2656 struct inode *inode = file_inode(filp); 2657 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2658 u8 encrypt_pw_salt[16]; 2659 int err; 2660 2661 if (!f2fs_sb_has_encrypt(sbi)) 2662 return -EOPNOTSUPP; 2663 2664 err = mnt_want_write_file(filp); 2665 if (err) 2666 return err; 2667 2668 f2fs_down_write(&sbi->sb_lock); 2669 2670 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2671 goto got_it; 2672 2673 /* update superblock with uuid */ 2674 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2675 2676 err = f2fs_commit_super(sbi, false); 2677 if (err) { 2678 /* undo new data */ 2679 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2680 goto out_err; 2681 } 2682 got_it: 2683 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2684 out_err: 2685 f2fs_up_write(&sbi->sb_lock); 2686 mnt_drop_write_file(filp); 2687 2688 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2689 err = -EFAULT; 2690 2691 return err; 2692 } 2693 2694 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2695 unsigned long arg) 2696 { 2697 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2698 return -EOPNOTSUPP; 2699 2700 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2701 } 2702 2703 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2704 { 2705 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2706 return -EOPNOTSUPP; 2707 2708 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2709 } 2710 2711 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2712 { 2713 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2714 return -EOPNOTSUPP; 2715 2716 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2717 } 2718 2719 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2720 unsigned long arg) 2721 { 2722 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2723 return -EOPNOTSUPP; 2724 2725 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2726 } 2727 2728 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2729 unsigned long arg) 2730 { 2731 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2732 return -EOPNOTSUPP; 2733 2734 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2735 } 2736 2737 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2738 { 2739 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2740 return -EOPNOTSUPP; 2741 2742 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2743 } 2744 2745 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2746 { 2747 struct inode *inode = file_inode(filp); 2748 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2749 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2750 .no_bg_gc = false, 2751 .should_migrate_blocks = false, 2752 .nr_free_secs = 0 }; 2753 __u32 sync; 2754 int ret; 2755 2756 if (!capable(CAP_SYS_ADMIN)) 2757 return -EPERM; 2758 2759 if (get_user(sync, (__u32 __user *)arg)) 2760 return -EFAULT; 2761 2762 if (f2fs_readonly(sbi->sb)) 2763 return -EROFS; 2764 2765 ret = mnt_want_write_file(filp); 2766 if (ret) 2767 return ret; 2768 2769 if (!sync) { 2770 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2771 ret = -EBUSY; 2772 goto out; 2773 } 2774 } else { 2775 f2fs_down_write(&sbi->gc_lock); 2776 } 2777 2778 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2779 gc_control.err_gc_skipped = sync; 2780 stat_inc_gc_call_count(sbi, FOREGROUND); 2781 ret = f2fs_gc(sbi, &gc_control); 2782 out: 2783 mnt_drop_write_file(filp); 2784 return ret; 2785 } 2786 2787 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2788 { 2789 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2790 struct f2fs_gc_control gc_control = { 2791 .init_gc_type = range->sync ? FG_GC : BG_GC, 2792 .no_bg_gc = false, 2793 .should_migrate_blocks = false, 2794 .err_gc_skipped = range->sync, 2795 .nr_free_secs = 0 }; 2796 u64 end; 2797 int ret; 2798 2799 if (!capable(CAP_SYS_ADMIN)) 2800 return -EPERM; 2801 if (f2fs_readonly(sbi->sb)) 2802 return -EROFS; 2803 2804 end = range->start + range->len; 2805 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2806 end >= MAX_BLKADDR(sbi)) 2807 return -EINVAL; 2808 2809 ret = mnt_want_write_file(filp); 2810 if (ret) 2811 return ret; 2812 2813 do_more: 2814 if (!range->sync) { 2815 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2816 ret = -EBUSY; 2817 goto out; 2818 } 2819 } else { 2820 f2fs_down_write(&sbi->gc_lock); 2821 } 2822 2823 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2824 stat_inc_gc_call_count(sbi, FOREGROUND); 2825 ret = f2fs_gc(sbi, &gc_control); 2826 if (ret) { 2827 if (ret == -EBUSY) 2828 ret = -EAGAIN; 2829 goto out; 2830 } 2831 range->start += CAP_BLKS_PER_SEC(sbi); 2832 if (range->start <= end) 2833 goto do_more; 2834 out: 2835 mnt_drop_write_file(filp); 2836 return ret; 2837 } 2838 2839 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2840 { 2841 struct f2fs_gc_range range; 2842 2843 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2844 sizeof(range))) 2845 return -EFAULT; 2846 return __f2fs_ioc_gc_range(filp, &range); 2847 } 2848 2849 static int f2fs_ioc_write_checkpoint(struct file *filp) 2850 { 2851 struct inode *inode = file_inode(filp); 2852 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2853 int ret; 2854 2855 if (!capable(CAP_SYS_ADMIN)) 2856 return -EPERM; 2857 2858 if (f2fs_readonly(sbi->sb)) 2859 return -EROFS; 2860 2861 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2862 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2863 return -EINVAL; 2864 } 2865 2866 ret = mnt_want_write_file(filp); 2867 if (ret) 2868 return ret; 2869 2870 ret = f2fs_sync_fs(sbi->sb, 1); 2871 2872 mnt_drop_write_file(filp); 2873 return ret; 2874 } 2875 2876 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2877 struct file *filp, 2878 struct f2fs_defragment *range) 2879 { 2880 struct inode *inode = file_inode(filp); 2881 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2882 .m_seg_type = NO_CHECK_TYPE, 2883 .m_may_create = false }; 2884 struct extent_info ei = {}; 2885 pgoff_t pg_start, pg_end, next_pgofs; 2886 unsigned int total = 0, sec_num; 2887 block_t blk_end = 0; 2888 bool fragmented = false; 2889 int err; 2890 2891 f2fs_balance_fs(sbi, true); 2892 2893 inode_lock(inode); 2894 pg_start = range->start >> PAGE_SHIFT; 2895 pg_end = min_t(pgoff_t, 2896 (range->start + range->len) >> PAGE_SHIFT, 2897 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE)); 2898 2899 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) || 2900 f2fs_is_atomic_file(inode)) { 2901 err = -EINVAL; 2902 goto unlock_out; 2903 } 2904 2905 /* if in-place-update policy is enabled, don't waste time here */ 2906 set_inode_flag(inode, FI_OPU_WRITE); 2907 if (f2fs_should_update_inplace(inode, NULL)) { 2908 err = -EINVAL; 2909 goto out; 2910 } 2911 2912 /* writeback all dirty pages in the range */ 2913 err = filemap_write_and_wait_range(inode->i_mapping, 2914 pg_start << PAGE_SHIFT, 2915 (pg_end << PAGE_SHIFT) - 1); 2916 if (err) 2917 goto out; 2918 2919 /* 2920 * lookup mapping info in extent cache, skip defragmenting if physical 2921 * block addresses are continuous. 2922 */ 2923 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2924 if ((pgoff_t)ei.fofs + ei.len >= pg_end) 2925 goto out; 2926 } 2927 2928 map.m_lblk = pg_start; 2929 map.m_next_pgofs = &next_pgofs; 2930 2931 /* 2932 * lookup mapping info in dnode page cache, skip defragmenting if all 2933 * physical block addresses are continuous even if there are hole(s) 2934 * in logical blocks. 2935 */ 2936 while (map.m_lblk < pg_end) { 2937 map.m_len = pg_end - map.m_lblk; 2938 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2939 if (err) 2940 goto out; 2941 2942 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2943 map.m_lblk = next_pgofs; 2944 continue; 2945 } 2946 2947 if (blk_end && blk_end != map.m_pblk) 2948 fragmented = true; 2949 2950 /* record total count of block that we're going to move */ 2951 total += map.m_len; 2952 2953 blk_end = map.m_pblk + map.m_len; 2954 2955 map.m_lblk += map.m_len; 2956 } 2957 2958 if (!fragmented) { 2959 total = 0; 2960 goto out; 2961 } 2962 2963 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2964 2965 /* 2966 * make sure there are enough free section for LFS allocation, this can 2967 * avoid defragment running in SSR mode when free section are allocated 2968 * intensively 2969 */ 2970 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2971 err = -EAGAIN; 2972 goto out; 2973 } 2974 2975 map.m_lblk = pg_start; 2976 map.m_len = pg_end - pg_start; 2977 total = 0; 2978 2979 while (map.m_lblk < pg_end) { 2980 pgoff_t idx; 2981 int cnt = 0; 2982 2983 do_map: 2984 map.m_len = pg_end - map.m_lblk; 2985 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2986 if (err) 2987 goto clear_out; 2988 2989 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2990 map.m_lblk = next_pgofs; 2991 goto check; 2992 } 2993 2994 set_inode_flag(inode, FI_SKIP_WRITES); 2995 2996 idx = map.m_lblk; 2997 while (idx < map.m_lblk + map.m_len && 2998 cnt < BLKS_PER_SEG(sbi)) { 2999 struct folio *folio; 3000 3001 folio = f2fs_get_lock_data_folio(inode, idx, true); 3002 if (IS_ERR(folio)) { 3003 err = PTR_ERR(folio); 3004 goto clear_out; 3005 } 3006 3007 f2fs_folio_wait_writeback(folio, DATA, true, true); 3008 3009 folio_mark_dirty(folio); 3010 folio_set_f2fs_gcing(folio); 3011 f2fs_folio_put(folio, true); 3012 3013 idx++; 3014 cnt++; 3015 total++; 3016 } 3017 3018 map.m_lblk = idx; 3019 check: 3020 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 3021 goto do_map; 3022 3023 clear_inode_flag(inode, FI_SKIP_WRITES); 3024 3025 err = filemap_fdatawrite(inode->i_mapping); 3026 if (err) 3027 goto out; 3028 } 3029 clear_out: 3030 clear_inode_flag(inode, FI_SKIP_WRITES); 3031 out: 3032 clear_inode_flag(inode, FI_OPU_WRITE); 3033 unlock_out: 3034 inode_unlock(inode); 3035 if (!err) 3036 range->len = (u64)total << PAGE_SHIFT; 3037 return err; 3038 } 3039 3040 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 3041 { 3042 struct inode *inode = file_inode(filp); 3043 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3044 struct f2fs_defragment range; 3045 int err; 3046 3047 if (!capable(CAP_SYS_ADMIN)) 3048 return -EPERM; 3049 3050 if (!S_ISREG(inode->i_mode)) 3051 return -EINVAL; 3052 3053 if (f2fs_readonly(sbi->sb)) 3054 return -EROFS; 3055 3056 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 3057 sizeof(range))) 3058 return -EFAULT; 3059 3060 /* verify alignment of offset & size */ 3061 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 3062 return -EINVAL; 3063 3064 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 3065 max_file_blocks(inode))) 3066 return -EINVAL; 3067 3068 err = mnt_want_write_file(filp); 3069 if (err) 3070 return err; 3071 3072 err = f2fs_defragment_range(sbi, filp, &range); 3073 mnt_drop_write_file(filp); 3074 3075 if (range.len) 3076 f2fs_update_time(sbi, REQ_TIME); 3077 if (err < 0) 3078 return err; 3079 3080 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 3081 sizeof(range))) 3082 return -EFAULT; 3083 3084 return 0; 3085 } 3086 3087 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 3088 struct file *file_out, loff_t pos_out, size_t len) 3089 { 3090 struct inode *src = file_inode(file_in); 3091 struct inode *dst = file_inode(file_out); 3092 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 3093 size_t olen = len, dst_max_i_size = 0; 3094 size_t dst_osize; 3095 int ret; 3096 3097 if (file_in->f_path.mnt != file_out->f_path.mnt || 3098 src->i_sb != dst->i_sb) 3099 return -EXDEV; 3100 3101 if (unlikely(f2fs_readonly(src->i_sb))) 3102 return -EROFS; 3103 3104 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 3105 return -EINVAL; 3106 3107 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 3108 return -EOPNOTSUPP; 3109 3110 if (pos_out < 0 || pos_in < 0) 3111 return -EINVAL; 3112 3113 if (src == dst) { 3114 if (pos_in == pos_out) 3115 return 0; 3116 if (pos_out > pos_in && pos_out < pos_in + len) 3117 return -EINVAL; 3118 } 3119 3120 inode_lock(src); 3121 if (src != dst) { 3122 ret = -EBUSY; 3123 if (!inode_trylock(dst)) 3124 goto out; 3125 } 3126 3127 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) || 3128 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 3129 ret = -EOPNOTSUPP; 3130 goto out_unlock; 3131 } 3132 3133 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) { 3134 ret = -EINVAL; 3135 goto out_unlock; 3136 } 3137 3138 ret = -EINVAL; 3139 if (pos_in + len > src->i_size || pos_in + len < pos_in) 3140 goto out_unlock; 3141 if (len == 0) 3142 olen = len = src->i_size - pos_in; 3143 if (pos_in + len == src->i_size) 3144 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 3145 if (len == 0) { 3146 ret = 0; 3147 goto out_unlock; 3148 } 3149 3150 dst_osize = dst->i_size; 3151 if (pos_out + olen > dst->i_size) 3152 dst_max_i_size = pos_out + olen; 3153 3154 /* verify the end result is block aligned */ 3155 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 3156 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 3157 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 3158 goto out_unlock; 3159 3160 ret = f2fs_convert_inline_inode(src); 3161 if (ret) 3162 goto out_unlock; 3163 3164 ret = f2fs_convert_inline_inode(dst); 3165 if (ret) 3166 goto out_unlock; 3167 3168 /* write out all dirty pages from offset */ 3169 ret = filemap_write_and_wait_range(src->i_mapping, 3170 pos_in, pos_in + len); 3171 if (ret) 3172 goto out_unlock; 3173 3174 ret = filemap_write_and_wait_range(dst->i_mapping, 3175 pos_out, pos_out + len); 3176 if (ret) 3177 goto out_unlock; 3178 3179 f2fs_balance_fs(sbi, true); 3180 3181 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3182 if (src != dst) { 3183 ret = -EBUSY; 3184 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 3185 goto out_src; 3186 } 3187 3188 f2fs_lock_op(sbi); 3189 ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in), 3190 F2FS_BYTES_TO_BLK(pos_out), 3191 F2FS_BYTES_TO_BLK(len), false); 3192 3193 if (!ret) { 3194 if (dst_max_i_size) 3195 f2fs_i_size_write(dst, dst_max_i_size); 3196 else if (dst_osize != dst->i_size) 3197 f2fs_i_size_write(dst, dst_osize); 3198 } 3199 f2fs_unlock_op(sbi); 3200 3201 if (src != dst) 3202 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 3203 out_src: 3204 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3205 if (ret) 3206 goto out_unlock; 3207 3208 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 3209 f2fs_mark_inode_dirty_sync(src, false); 3210 if (src != dst) { 3211 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 3212 f2fs_mark_inode_dirty_sync(dst, false); 3213 } 3214 f2fs_update_time(sbi, REQ_TIME); 3215 3216 out_unlock: 3217 if (src != dst) 3218 inode_unlock(dst); 3219 out: 3220 inode_unlock(src); 3221 return ret; 3222 } 3223 3224 static int __f2fs_ioc_move_range(struct file *filp, 3225 struct f2fs_move_range *range) 3226 { 3227 int err; 3228 3229 if (!(filp->f_mode & FMODE_READ) || 3230 !(filp->f_mode & FMODE_WRITE)) 3231 return -EBADF; 3232 3233 CLASS(fd, dst)(range->dst_fd); 3234 if (fd_empty(dst)) 3235 return -EBADF; 3236 3237 if (!(fd_file(dst)->f_mode & FMODE_WRITE)) 3238 return -EBADF; 3239 3240 err = mnt_want_write_file(filp); 3241 if (err) 3242 return err; 3243 3244 err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst), 3245 range->pos_out, range->len); 3246 3247 mnt_drop_write_file(filp); 3248 return err; 3249 } 3250 3251 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 3252 { 3253 struct f2fs_move_range range; 3254 3255 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 3256 sizeof(range))) 3257 return -EFAULT; 3258 return __f2fs_ioc_move_range(filp, &range); 3259 } 3260 3261 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 3262 { 3263 struct inode *inode = file_inode(filp); 3264 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3265 struct sit_info *sm = SIT_I(sbi); 3266 unsigned int start_segno = 0, end_segno = 0; 3267 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3268 struct f2fs_flush_device range; 3269 struct f2fs_gc_control gc_control = { 3270 .init_gc_type = FG_GC, 3271 .should_migrate_blocks = true, 3272 .err_gc_skipped = true, 3273 .nr_free_secs = 0 }; 3274 int ret; 3275 3276 if (!capable(CAP_SYS_ADMIN)) 3277 return -EPERM; 3278 3279 if (f2fs_readonly(sbi->sb)) 3280 return -EROFS; 3281 3282 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3283 return -EINVAL; 3284 3285 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3286 sizeof(range))) 3287 return -EFAULT; 3288 3289 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3290 __is_large_section(sbi)) { 3291 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3292 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3293 return -EINVAL; 3294 } 3295 3296 ret = mnt_want_write_file(filp); 3297 if (ret) 3298 return ret; 3299 3300 if (range.dev_num != 0) 3301 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3302 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3303 3304 start_segno = sm->last_victim[FLUSH_DEVICE]; 3305 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3306 start_segno = dev_start_segno; 3307 end_segno = min(start_segno + range.segments, dev_end_segno); 3308 3309 while (start_segno < end_segno) { 3310 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3311 ret = -EBUSY; 3312 goto out; 3313 } 3314 sm->last_victim[GC_CB] = end_segno + 1; 3315 sm->last_victim[GC_GREEDY] = end_segno + 1; 3316 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3317 3318 gc_control.victim_segno = start_segno; 3319 stat_inc_gc_call_count(sbi, FOREGROUND); 3320 ret = f2fs_gc(sbi, &gc_control); 3321 if (ret == -EAGAIN) 3322 ret = 0; 3323 else if (ret < 0) 3324 break; 3325 start_segno++; 3326 } 3327 out: 3328 mnt_drop_write_file(filp); 3329 return ret; 3330 } 3331 3332 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3333 { 3334 struct inode *inode = file_inode(filp); 3335 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3336 3337 /* Must validate to set it with SQLite behavior in Android. */ 3338 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3339 3340 return put_user(sb_feature, (u32 __user *)arg); 3341 } 3342 3343 #ifdef CONFIG_QUOTA 3344 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3345 { 3346 struct dquot *transfer_to[MAXQUOTAS] = {}; 3347 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3348 struct super_block *sb = sbi->sb; 3349 int err; 3350 3351 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3352 if (IS_ERR(transfer_to[PRJQUOTA])) 3353 return PTR_ERR(transfer_to[PRJQUOTA]); 3354 3355 err = __dquot_transfer(inode, transfer_to); 3356 if (err) 3357 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3358 dqput(transfer_to[PRJQUOTA]); 3359 return err; 3360 } 3361 3362 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3363 { 3364 struct f2fs_inode_info *fi = F2FS_I(inode); 3365 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3366 struct f2fs_inode *ri = NULL; 3367 kprojid_t kprojid; 3368 int err; 3369 3370 if (!f2fs_sb_has_project_quota(sbi)) { 3371 if (projid != F2FS_DEF_PROJID) 3372 return -EOPNOTSUPP; 3373 else 3374 return 0; 3375 } 3376 3377 if (!f2fs_has_extra_attr(inode)) 3378 return -EOPNOTSUPP; 3379 3380 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3381 3382 if (projid_eq(kprojid, fi->i_projid)) 3383 return 0; 3384 3385 err = -EPERM; 3386 /* Is it quota file? Do not allow user to mess with it */ 3387 if (IS_NOQUOTA(inode)) 3388 return err; 3389 3390 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3391 return -EOVERFLOW; 3392 3393 err = f2fs_dquot_initialize(inode); 3394 if (err) 3395 return err; 3396 3397 f2fs_lock_op(sbi); 3398 err = f2fs_transfer_project_quota(inode, kprojid); 3399 if (err) 3400 goto out_unlock; 3401 3402 fi->i_projid = kprojid; 3403 inode_set_ctime_current(inode); 3404 f2fs_mark_inode_dirty_sync(inode, true); 3405 out_unlock: 3406 f2fs_unlock_op(sbi); 3407 return err; 3408 } 3409 #else 3410 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3411 { 3412 return 0; 3413 } 3414 3415 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3416 { 3417 if (projid != F2FS_DEF_PROJID) 3418 return -EOPNOTSUPP; 3419 return 0; 3420 } 3421 #endif 3422 3423 int f2fs_fileattr_get(struct dentry *dentry, struct file_kattr *fa) 3424 { 3425 struct inode *inode = d_inode(dentry); 3426 struct f2fs_inode_info *fi = F2FS_I(inode); 3427 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3428 3429 if (IS_ENCRYPTED(inode)) 3430 fsflags |= FS_ENCRYPT_FL; 3431 if (IS_VERITY(inode)) 3432 fsflags |= FS_VERITY_FL; 3433 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3434 fsflags |= FS_INLINE_DATA_FL; 3435 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3436 fsflags |= FS_NOCOW_FL; 3437 3438 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3439 3440 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3441 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3442 3443 return 0; 3444 } 3445 3446 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3447 struct dentry *dentry, struct file_kattr *fa) 3448 { 3449 struct inode *inode = d_inode(dentry); 3450 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3451 u32 iflags; 3452 int err; 3453 3454 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3455 return -EIO; 3456 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3457 return -ENOSPC; 3458 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3459 return -EOPNOTSUPP; 3460 fsflags &= F2FS_SETTABLE_FS_FL; 3461 if (!fa->flags_valid) 3462 mask &= FS_COMMON_FL; 3463 3464 iflags = f2fs_fsflags_to_iflags(fsflags); 3465 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3466 return -EOPNOTSUPP; 3467 3468 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3469 if (!err) 3470 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3471 3472 return err; 3473 } 3474 3475 int f2fs_pin_file_control(struct inode *inode, bool inc) 3476 { 3477 struct f2fs_inode_info *fi = F2FS_I(inode); 3478 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3479 3480 if (IS_DEVICE_ALIASING(inode)) 3481 return -EINVAL; 3482 3483 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) { 3484 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3485 __func__, inode->i_ino, fi->i_gc_failures); 3486 clear_inode_flag(inode, FI_PIN_FILE); 3487 return -EAGAIN; 3488 } 3489 3490 /* Use i_gc_failures for normal file as a risk signal. */ 3491 if (inc) 3492 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1); 3493 3494 return 0; 3495 } 3496 3497 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3498 { 3499 struct inode *inode = file_inode(filp); 3500 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3501 __u32 pin; 3502 int ret = 0; 3503 3504 if (get_user(pin, (__u32 __user *)arg)) 3505 return -EFAULT; 3506 3507 if (!S_ISREG(inode->i_mode)) 3508 return -EINVAL; 3509 3510 if (f2fs_readonly(sbi->sb)) 3511 return -EROFS; 3512 3513 if (!pin && IS_DEVICE_ALIASING(inode)) 3514 return -EOPNOTSUPP; 3515 3516 ret = mnt_want_write_file(filp); 3517 if (ret) 3518 return ret; 3519 3520 inode_lock(inode); 3521 3522 if (f2fs_is_atomic_file(inode)) { 3523 ret = -EINVAL; 3524 goto out; 3525 } 3526 3527 if (!pin) { 3528 clear_inode_flag(inode, FI_PIN_FILE); 3529 f2fs_i_gc_failures_write(inode, 0); 3530 goto done; 3531 } else if (f2fs_is_pinned_file(inode)) { 3532 goto done; 3533 } 3534 3535 if (F2FS_HAS_BLOCKS(inode)) { 3536 ret = -EFBIG; 3537 goto out; 3538 } 3539 3540 /* Let's allow file pinning on zoned device. */ 3541 if (!f2fs_sb_has_blkzoned(sbi) && 3542 f2fs_should_update_outplace(inode, NULL)) { 3543 ret = -EINVAL; 3544 goto out; 3545 } 3546 3547 if (f2fs_pin_file_control(inode, false)) { 3548 ret = -EAGAIN; 3549 goto out; 3550 } 3551 3552 ret = f2fs_convert_inline_inode(inode); 3553 if (ret) 3554 goto out; 3555 3556 if (!f2fs_disable_compressed_file(inode)) { 3557 ret = -EOPNOTSUPP; 3558 goto out; 3559 } 3560 3561 set_inode_flag(inode, FI_PIN_FILE); 3562 ret = F2FS_I(inode)->i_gc_failures; 3563 done: 3564 f2fs_update_time(sbi, REQ_TIME); 3565 out: 3566 inode_unlock(inode); 3567 mnt_drop_write_file(filp); 3568 return ret; 3569 } 3570 3571 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3572 { 3573 struct inode *inode = file_inode(filp); 3574 __u32 pin = 0; 3575 3576 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3577 pin = F2FS_I(inode)->i_gc_failures; 3578 return put_user(pin, (u32 __user *)arg); 3579 } 3580 3581 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg) 3582 { 3583 return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0, 3584 (u32 __user *)arg); 3585 } 3586 3587 static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg) 3588 { 3589 struct inode *inode = file_inode(filp); 3590 __u32 level; 3591 3592 if (get_user(level, (__u32 __user *)arg)) 3593 return -EFAULT; 3594 3595 if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX) 3596 return -EINVAL; 3597 3598 inode_lock(inode); 3599 F2FS_I(inode)->ioprio_hint = level; 3600 inode_unlock(inode); 3601 return 0; 3602 } 3603 3604 int f2fs_precache_extents(struct inode *inode) 3605 { 3606 struct f2fs_inode_info *fi = F2FS_I(inode); 3607 struct f2fs_map_blocks map; 3608 pgoff_t m_next_extent; 3609 loff_t end; 3610 int err; 3611 3612 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3613 return -EOPNOTSUPP; 3614 3615 map.m_lblk = 0; 3616 map.m_pblk = 0; 3617 map.m_next_pgofs = NULL; 3618 map.m_next_extent = &m_next_extent; 3619 map.m_seg_type = NO_CHECK_TYPE; 3620 map.m_may_create = false; 3621 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3622 3623 while (map.m_lblk < end) { 3624 map.m_len = end - map.m_lblk; 3625 3626 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3627 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3628 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3629 if (err || !map.m_len) 3630 return err; 3631 3632 map.m_lblk = m_next_extent; 3633 } 3634 3635 return 0; 3636 } 3637 3638 static int f2fs_ioc_precache_extents(struct file *filp) 3639 { 3640 return f2fs_precache_extents(file_inode(filp)); 3641 } 3642 3643 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3644 { 3645 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3646 __u64 block_count; 3647 3648 if (!capable(CAP_SYS_ADMIN)) 3649 return -EPERM; 3650 3651 if (f2fs_readonly(sbi->sb)) 3652 return -EROFS; 3653 3654 if (copy_from_user(&block_count, (void __user *)arg, 3655 sizeof(block_count))) 3656 return -EFAULT; 3657 3658 return f2fs_resize_fs(filp, block_count); 3659 } 3660 3661 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3662 { 3663 struct inode *inode = file_inode(filp); 3664 3665 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3666 3667 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3668 f2fs_warn(F2FS_I_SB(inode), 3669 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3670 inode->i_ino); 3671 return -EOPNOTSUPP; 3672 } 3673 3674 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3675 } 3676 3677 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3678 { 3679 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3680 return -EOPNOTSUPP; 3681 3682 return fsverity_ioctl_measure(filp, (void __user *)arg); 3683 } 3684 3685 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3686 { 3687 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3688 return -EOPNOTSUPP; 3689 3690 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3691 } 3692 3693 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3694 { 3695 struct inode *inode = file_inode(filp); 3696 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3697 char *vbuf; 3698 int count; 3699 int err = 0; 3700 3701 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3702 if (!vbuf) 3703 return -ENOMEM; 3704 3705 f2fs_down_read(&sbi->sb_lock); 3706 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3707 ARRAY_SIZE(sbi->raw_super->volume_name), 3708 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3709 f2fs_up_read(&sbi->sb_lock); 3710 3711 if (copy_to_user((char __user *)arg, vbuf, 3712 min(FSLABEL_MAX, count))) 3713 err = -EFAULT; 3714 3715 kfree(vbuf); 3716 return err; 3717 } 3718 3719 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3720 { 3721 struct inode *inode = file_inode(filp); 3722 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3723 char *vbuf; 3724 int err = 0; 3725 3726 if (!capable(CAP_SYS_ADMIN)) 3727 return -EPERM; 3728 3729 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3730 if (IS_ERR(vbuf)) 3731 return PTR_ERR(vbuf); 3732 3733 err = mnt_want_write_file(filp); 3734 if (err) 3735 goto out; 3736 3737 f2fs_down_write(&sbi->sb_lock); 3738 3739 memset(sbi->raw_super->volume_name, 0, 3740 sizeof(sbi->raw_super->volume_name)); 3741 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3742 sbi->raw_super->volume_name, 3743 ARRAY_SIZE(sbi->raw_super->volume_name)); 3744 3745 err = f2fs_commit_super(sbi, false); 3746 3747 f2fs_up_write(&sbi->sb_lock); 3748 3749 mnt_drop_write_file(filp); 3750 out: 3751 kfree(vbuf); 3752 return err; 3753 } 3754 3755 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3756 { 3757 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3758 return -EOPNOTSUPP; 3759 3760 if (!f2fs_compressed_file(inode)) 3761 return -EINVAL; 3762 3763 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3764 3765 return 0; 3766 } 3767 3768 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3769 { 3770 struct inode *inode = file_inode(filp); 3771 __u64 blocks; 3772 int ret; 3773 3774 ret = f2fs_get_compress_blocks(inode, &blocks); 3775 if (ret < 0) 3776 return ret; 3777 3778 return put_user(blocks, (u64 __user *)arg); 3779 } 3780 3781 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3782 { 3783 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3784 unsigned int released_blocks = 0; 3785 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3786 block_t blkaddr; 3787 int i; 3788 3789 for (i = 0; i < count; i++) { 3790 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3791 dn->ofs_in_node + i); 3792 3793 if (!__is_valid_data_blkaddr(blkaddr)) 3794 continue; 3795 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3796 DATA_GENERIC_ENHANCE))) 3797 return -EFSCORRUPTED; 3798 } 3799 3800 while (count) { 3801 int compr_blocks = 0; 3802 3803 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3804 blkaddr = f2fs_data_blkaddr(dn); 3805 3806 if (i == 0) { 3807 if (blkaddr == COMPRESS_ADDR) 3808 continue; 3809 dn->ofs_in_node += cluster_size; 3810 goto next; 3811 } 3812 3813 if (__is_valid_data_blkaddr(blkaddr)) 3814 compr_blocks++; 3815 3816 if (blkaddr != NEW_ADDR) 3817 continue; 3818 3819 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3820 } 3821 3822 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3823 dec_valid_block_count(sbi, dn->inode, 3824 cluster_size - compr_blocks); 3825 3826 released_blocks += cluster_size - compr_blocks; 3827 next: 3828 count -= cluster_size; 3829 } 3830 3831 return released_blocks; 3832 } 3833 3834 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3835 { 3836 struct inode *inode = file_inode(filp); 3837 struct f2fs_inode_info *fi = F2FS_I(inode); 3838 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3839 pgoff_t page_idx = 0, last_idx; 3840 unsigned int released_blocks = 0; 3841 int ret; 3842 int writecount; 3843 3844 if (!f2fs_sb_has_compression(sbi)) 3845 return -EOPNOTSUPP; 3846 3847 if (f2fs_readonly(sbi->sb)) 3848 return -EROFS; 3849 3850 ret = mnt_want_write_file(filp); 3851 if (ret) 3852 return ret; 3853 3854 f2fs_balance_fs(sbi, true); 3855 3856 inode_lock(inode); 3857 3858 writecount = atomic_read(&inode->i_writecount); 3859 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3860 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3861 ret = -EBUSY; 3862 goto out; 3863 } 3864 3865 if (!f2fs_compressed_file(inode) || 3866 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3867 ret = -EINVAL; 3868 goto out; 3869 } 3870 3871 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3872 if (ret) 3873 goto out; 3874 3875 if (!atomic_read(&fi->i_compr_blocks)) { 3876 ret = -EPERM; 3877 goto out; 3878 } 3879 3880 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3881 inode_set_ctime_current(inode); 3882 f2fs_mark_inode_dirty_sync(inode, true); 3883 3884 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3885 filemap_invalidate_lock(inode->i_mapping); 3886 3887 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3888 3889 while (page_idx < last_idx) { 3890 struct dnode_of_data dn; 3891 pgoff_t end_offset, count; 3892 3893 f2fs_lock_op(sbi); 3894 3895 set_new_dnode(&dn, inode, NULL, NULL, 0); 3896 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3897 if (ret) { 3898 f2fs_unlock_op(sbi); 3899 if (ret == -ENOENT) { 3900 page_idx = f2fs_get_next_page_offset(&dn, 3901 page_idx); 3902 ret = 0; 3903 continue; 3904 } 3905 break; 3906 } 3907 3908 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 3909 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3910 count = round_up(count, fi->i_cluster_size); 3911 3912 ret = release_compress_blocks(&dn, count); 3913 3914 f2fs_put_dnode(&dn); 3915 3916 f2fs_unlock_op(sbi); 3917 3918 if (ret < 0) 3919 break; 3920 3921 page_idx += count; 3922 released_blocks += ret; 3923 } 3924 3925 filemap_invalidate_unlock(inode->i_mapping); 3926 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3927 out: 3928 if (released_blocks) 3929 f2fs_update_time(sbi, REQ_TIME); 3930 inode_unlock(inode); 3931 3932 mnt_drop_write_file(filp); 3933 3934 if (ret >= 0) { 3935 ret = put_user(released_blocks, (u64 __user *)arg); 3936 } else if (released_blocks && 3937 atomic_read(&fi->i_compr_blocks)) { 3938 set_sbi_flag(sbi, SBI_NEED_FSCK); 3939 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3940 "iblocks=%llu, released=%u, compr_blocks=%u, " 3941 "run fsck to fix.", 3942 __func__, inode->i_ino, inode->i_blocks, 3943 released_blocks, 3944 atomic_read(&fi->i_compr_blocks)); 3945 } 3946 3947 return ret; 3948 } 3949 3950 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3951 unsigned int *reserved_blocks) 3952 { 3953 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3954 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3955 block_t blkaddr; 3956 int i; 3957 3958 for (i = 0; i < count; i++) { 3959 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3960 dn->ofs_in_node + i); 3961 3962 if (!__is_valid_data_blkaddr(blkaddr)) 3963 continue; 3964 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3965 DATA_GENERIC_ENHANCE))) 3966 return -EFSCORRUPTED; 3967 } 3968 3969 while (count) { 3970 int compr_blocks = 0; 3971 blkcnt_t reserved = 0; 3972 blkcnt_t to_reserved; 3973 int ret; 3974 3975 for (i = 0; i < cluster_size; i++) { 3976 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3977 dn->ofs_in_node + i); 3978 3979 if (i == 0) { 3980 if (blkaddr != COMPRESS_ADDR) { 3981 dn->ofs_in_node += cluster_size; 3982 goto next; 3983 } 3984 continue; 3985 } 3986 3987 /* 3988 * compressed cluster was not released due to it 3989 * fails in release_compress_blocks(), so NEW_ADDR 3990 * is a possible case. 3991 */ 3992 if (blkaddr == NEW_ADDR) { 3993 reserved++; 3994 continue; 3995 } 3996 if (__is_valid_data_blkaddr(blkaddr)) { 3997 compr_blocks++; 3998 continue; 3999 } 4000 } 4001 4002 to_reserved = cluster_size - compr_blocks - reserved; 4003 4004 /* for the case all blocks in cluster were reserved */ 4005 if (reserved && to_reserved == 1) { 4006 dn->ofs_in_node += cluster_size; 4007 goto next; 4008 } 4009 4010 ret = inc_valid_block_count(sbi, dn->inode, 4011 &to_reserved, false); 4012 if (unlikely(ret)) 4013 return ret; 4014 4015 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 4016 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 4017 f2fs_set_data_blkaddr(dn, NEW_ADDR); 4018 } 4019 4020 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 4021 4022 *reserved_blocks += to_reserved; 4023 next: 4024 count -= cluster_size; 4025 } 4026 4027 return 0; 4028 } 4029 4030 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 4031 { 4032 struct inode *inode = file_inode(filp); 4033 struct f2fs_inode_info *fi = F2FS_I(inode); 4034 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4035 pgoff_t page_idx = 0, last_idx; 4036 unsigned int reserved_blocks = 0; 4037 int ret; 4038 4039 if (!f2fs_sb_has_compression(sbi)) 4040 return -EOPNOTSUPP; 4041 4042 if (f2fs_readonly(sbi->sb)) 4043 return -EROFS; 4044 4045 ret = mnt_want_write_file(filp); 4046 if (ret) 4047 return ret; 4048 4049 f2fs_balance_fs(sbi, true); 4050 4051 inode_lock(inode); 4052 4053 if (!f2fs_compressed_file(inode) || 4054 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4055 ret = -EINVAL; 4056 goto unlock_inode; 4057 } 4058 4059 if (atomic_read(&fi->i_compr_blocks)) 4060 goto unlock_inode; 4061 4062 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 4063 filemap_invalidate_lock(inode->i_mapping); 4064 4065 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4066 4067 while (page_idx < last_idx) { 4068 struct dnode_of_data dn; 4069 pgoff_t end_offset, count; 4070 4071 f2fs_lock_op(sbi); 4072 4073 set_new_dnode(&dn, inode, NULL, NULL, 0); 4074 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 4075 if (ret) { 4076 f2fs_unlock_op(sbi); 4077 if (ret == -ENOENT) { 4078 page_idx = f2fs_get_next_page_offset(&dn, 4079 page_idx); 4080 ret = 0; 4081 continue; 4082 } 4083 break; 4084 } 4085 4086 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 4087 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 4088 count = round_up(count, fi->i_cluster_size); 4089 4090 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 4091 4092 f2fs_put_dnode(&dn); 4093 4094 f2fs_unlock_op(sbi); 4095 4096 if (ret < 0) 4097 break; 4098 4099 page_idx += count; 4100 } 4101 4102 filemap_invalidate_unlock(inode->i_mapping); 4103 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 4104 4105 if (!ret) { 4106 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 4107 inode_set_ctime_current(inode); 4108 f2fs_mark_inode_dirty_sync(inode, true); 4109 } 4110 unlock_inode: 4111 if (reserved_blocks) 4112 f2fs_update_time(sbi, REQ_TIME); 4113 inode_unlock(inode); 4114 mnt_drop_write_file(filp); 4115 4116 if (!ret) { 4117 ret = put_user(reserved_blocks, (u64 __user *)arg); 4118 } else if (reserved_blocks && 4119 atomic_read(&fi->i_compr_blocks)) { 4120 set_sbi_flag(sbi, SBI_NEED_FSCK); 4121 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx " 4122 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 4123 "run fsck to fix.", 4124 __func__, inode->i_ino, inode->i_blocks, 4125 reserved_blocks, 4126 atomic_read(&fi->i_compr_blocks)); 4127 } 4128 4129 return ret; 4130 } 4131 4132 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 4133 pgoff_t off, block_t block, block_t len, u32 flags) 4134 { 4135 sector_t sector = SECTOR_FROM_BLOCK(block); 4136 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 4137 int ret = 0; 4138 4139 if (flags & F2FS_TRIM_FILE_DISCARD) { 4140 if (bdev_max_secure_erase_sectors(bdev)) 4141 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 4142 GFP_NOFS); 4143 else 4144 ret = blkdev_issue_discard(bdev, sector, nr_sects, 4145 GFP_NOFS); 4146 } 4147 4148 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 4149 if (IS_ENCRYPTED(inode)) 4150 ret = fscrypt_zeroout_range(inode, off, block, len); 4151 else 4152 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 4153 GFP_NOFS, 0); 4154 } 4155 4156 return ret; 4157 } 4158 4159 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 4160 { 4161 struct inode *inode = file_inode(filp); 4162 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4163 struct address_space *mapping = inode->i_mapping; 4164 struct block_device *prev_bdev = NULL; 4165 struct f2fs_sectrim_range range; 4166 pgoff_t index, pg_end, prev_index = 0; 4167 block_t prev_block = 0, len = 0; 4168 loff_t end_addr; 4169 bool to_end = false; 4170 int ret = 0; 4171 4172 if (!(filp->f_mode & FMODE_WRITE)) 4173 return -EBADF; 4174 4175 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 4176 sizeof(range))) 4177 return -EFAULT; 4178 4179 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 4180 !S_ISREG(inode->i_mode)) 4181 return -EINVAL; 4182 4183 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 4184 !f2fs_hw_support_discard(sbi)) || 4185 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 4186 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 4187 return -EOPNOTSUPP; 4188 4189 ret = mnt_want_write_file(filp); 4190 if (ret) 4191 return ret; 4192 inode_lock(inode); 4193 4194 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 4195 range.start >= inode->i_size) { 4196 ret = -EINVAL; 4197 goto err; 4198 } 4199 4200 if (range.len == 0) 4201 goto err; 4202 4203 if (inode->i_size - range.start > range.len) { 4204 end_addr = range.start + range.len; 4205 } else { 4206 end_addr = range.len == (u64)-1 ? 4207 sbi->sb->s_maxbytes : inode->i_size; 4208 to_end = true; 4209 } 4210 4211 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 4212 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 4213 ret = -EINVAL; 4214 goto err; 4215 } 4216 4217 index = F2FS_BYTES_TO_BLK(range.start); 4218 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 4219 4220 ret = f2fs_convert_inline_inode(inode); 4221 if (ret) 4222 goto err; 4223 4224 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4225 filemap_invalidate_lock(mapping); 4226 4227 ret = filemap_write_and_wait_range(mapping, range.start, 4228 to_end ? LLONG_MAX : end_addr - 1); 4229 if (ret) 4230 goto out; 4231 4232 truncate_inode_pages_range(mapping, range.start, 4233 to_end ? -1 : end_addr - 1); 4234 4235 while (index < pg_end) { 4236 struct dnode_of_data dn; 4237 pgoff_t end_offset, count; 4238 int i; 4239 4240 set_new_dnode(&dn, inode, NULL, NULL, 0); 4241 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 4242 if (ret) { 4243 if (ret == -ENOENT) { 4244 index = f2fs_get_next_page_offset(&dn, index); 4245 continue; 4246 } 4247 goto out; 4248 } 4249 4250 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 4251 count = min(end_offset - dn.ofs_in_node, pg_end - index); 4252 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 4253 struct block_device *cur_bdev; 4254 block_t blkaddr = f2fs_data_blkaddr(&dn); 4255 4256 if (!__is_valid_data_blkaddr(blkaddr)) 4257 continue; 4258 4259 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 4260 DATA_GENERIC_ENHANCE)) { 4261 ret = -EFSCORRUPTED; 4262 f2fs_put_dnode(&dn); 4263 goto out; 4264 } 4265 4266 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 4267 if (f2fs_is_multi_device(sbi)) { 4268 int di = f2fs_target_device_index(sbi, blkaddr); 4269 4270 blkaddr -= FDEV(di).start_blk; 4271 } 4272 4273 if (len) { 4274 if (prev_bdev == cur_bdev && 4275 index == prev_index + len && 4276 blkaddr == prev_block + len) { 4277 len++; 4278 } else { 4279 ret = f2fs_secure_erase(prev_bdev, 4280 inode, prev_index, prev_block, 4281 len, range.flags); 4282 if (ret) { 4283 f2fs_put_dnode(&dn); 4284 goto out; 4285 } 4286 4287 len = 0; 4288 } 4289 } 4290 4291 if (!len) { 4292 prev_bdev = cur_bdev; 4293 prev_index = index; 4294 prev_block = blkaddr; 4295 len = 1; 4296 } 4297 } 4298 4299 f2fs_put_dnode(&dn); 4300 4301 if (fatal_signal_pending(current)) { 4302 ret = -EINTR; 4303 goto out; 4304 } 4305 cond_resched(); 4306 } 4307 4308 if (len) 4309 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 4310 prev_block, len, range.flags); 4311 f2fs_update_time(sbi, REQ_TIME); 4312 out: 4313 filemap_invalidate_unlock(mapping); 4314 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4315 err: 4316 inode_unlock(inode); 4317 mnt_drop_write_file(filp); 4318 4319 return ret; 4320 } 4321 4322 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 4323 { 4324 struct inode *inode = file_inode(filp); 4325 struct f2fs_comp_option option; 4326 4327 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4328 return -EOPNOTSUPP; 4329 4330 inode_lock_shared(inode); 4331 4332 if (!f2fs_compressed_file(inode)) { 4333 inode_unlock_shared(inode); 4334 return -ENODATA; 4335 } 4336 4337 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4338 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4339 4340 inode_unlock_shared(inode); 4341 4342 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4343 sizeof(option))) 4344 return -EFAULT; 4345 4346 return 0; 4347 } 4348 4349 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4350 { 4351 struct inode *inode = file_inode(filp); 4352 struct f2fs_inode_info *fi = F2FS_I(inode); 4353 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4354 struct f2fs_comp_option option; 4355 int ret = 0; 4356 4357 if (!f2fs_sb_has_compression(sbi)) 4358 return -EOPNOTSUPP; 4359 4360 if (!(filp->f_mode & FMODE_WRITE)) 4361 return -EBADF; 4362 4363 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4364 sizeof(option))) 4365 return -EFAULT; 4366 4367 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4368 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4369 option.algorithm >= COMPRESS_MAX) 4370 return -EINVAL; 4371 4372 ret = mnt_want_write_file(filp); 4373 if (ret) 4374 return ret; 4375 inode_lock(inode); 4376 4377 f2fs_down_write(&F2FS_I(inode)->i_sem); 4378 if (!f2fs_compressed_file(inode)) { 4379 ret = -EINVAL; 4380 goto out; 4381 } 4382 4383 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4384 ret = -EBUSY; 4385 goto out; 4386 } 4387 4388 if (F2FS_HAS_BLOCKS(inode)) { 4389 ret = -EFBIG; 4390 goto out; 4391 } 4392 4393 fi->i_compress_algorithm = option.algorithm; 4394 fi->i_log_cluster_size = option.log_cluster_size; 4395 fi->i_cluster_size = BIT(option.log_cluster_size); 4396 /* Set default level */ 4397 if (fi->i_compress_algorithm == COMPRESS_ZSTD) 4398 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4399 else 4400 fi->i_compress_level = 0; 4401 /* Adjust mount option level */ 4402 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4403 F2FS_OPTION(sbi).compress_level) 4404 fi->i_compress_level = F2FS_OPTION(sbi).compress_level; 4405 f2fs_mark_inode_dirty_sync(inode, true); 4406 4407 if (!f2fs_is_compress_backend_ready(inode)) 4408 f2fs_warn(sbi, "compression algorithm is successfully set, " 4409 "but current kernel doesn't support this algorithm."); 4410 out: 4411 f2fs_up_write(&fi->i_sem); 4412 inode_unlock(inode); 4413 mnt_drop_write_file(filp); 4414 4415 return ret; 4416 } 4417 4418 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4419 { 4420 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4421 struct address_space *mapping = inode->i_mapping; 4422 struct folio *folio; 4423 pgoff_t redirty_idx = page_idx; 4424 int page_len = 0, ret = 0; 4425 4426 page_cache_ra_unbounded(&ractl, len, 0); 4427 4428 do { 4429 folio = read_cache_folio(mapping, page_idx, NULL, NULL); 4430 if (IS_ERR(folio)) { 4431 ret = PTR_ERR(folio); 4432 break; 4433 } 4434 page_len += folio_nr_pages(folio) - (page_idx - folio->index); 4435 page_idx = folio_next_index(folio); 4436 } while (page_len < len); 4437 4438 do { 4439 folio = filemap_lock_folio(mapping, redirty_idx); 4440 4441 /* It will never fail, when folio has pinned above */ 4442 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio)); 4443 4444 f2fs_folio_wait_writeback(folio, DATA, true, true); 4445 4446 folio_mark_dirty(folio); 4447 folio_set_f2fs_gcing(folio); 4448 redirty_idx = folio_next_index(folio); 4449 folio_unlock(folio); 4450 folio_put_refs(folio, 2); 4451 } while (redirty_idx < page_idx); 4452 4453 return ret; 4454 } 4455 4456 static int f2fs_ioc_decompress_file(struct file *filp) 4457 { 4458 struct inode *inode = file_inode(filp); 4459 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4460 struct f2fs_inode_info *fi = F2FS_I(inode); 4461 pgoff_t page_idx = 0, last_idx, cluster_idx; 4462 int ret; 4463 4464 if (!f2fs_sb_has_compression(sbi) || 4465 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4466 return -EOPNOTSUPP; 4467 4468 if (!(filp->f_mode & FMODE_WRITE)) 4469 return -EBADF; 4470 4471 f2fs_balance_fs(sbi, true); 4472 4473 ret = mnt_want_write_file(filp); 4474 if (ret) 4475 return ret; 4476 inode_lock(inode); 4477 4478 if (!f2fs_is_compress_backend_ready(inode)) { 4479 ret = -EOPNOTSUPP; 4480 goto out; 4481 } 4482 4483 if (!f2fs_compressed_file(inode) || 4484 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4485 ret = -EINVAL; 4486 goto out; 4487 } 4488 4489 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4490 if (ret) 4491 goto out; 4492 4493 if (!atomic_read(&fi->i_compr_blocks)) 4494 goto out; 4495 4496 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4497 last_idx >>= fi->i_log_cluster_size; 4498 4499 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4500 page_idx = cluster_idx << fi->i_log_cluster_size; 4501 4502 if (!f2fs_is_compressed_cluster(inode, page_idx)) 4503 continue; 4504 4505 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4506 if (ret < 0) 4507 break; 4508 4509 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4510 ret = filemap_fdatawrite(inode->i_mapping); 4511 if (ret < 0) 4512 break; 4513 } 4514 4515 cond_resched(); 4516 if (fatal_signal_pending(current)) { 4517 ret = -EINTR; 4518 break; 4519 } 4520 } 4521 4522 if (!ret) 4523 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4524 LLONG_MAX); 4525 4526 if (ret) 4527 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4528 __func__, ret); 4529 f2fs_update_time(sbi, REQ_TIME); 4530 out: 4531 inode_unlock(inode); 4532 mnt_drop_write_file(filp); 4533 4534 return ret; 4535 } 4536 4537 static int f2fs_ioc_compress_file(struct file *filp) 4538 { 4539 struct inode *inode = file_inode(filp); 4540 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4541 struct f2fs_inode_info *fi = F2FS_I(inode); 4542 pgoff_t page_idx = 0, last_idx, cluster_idx; 4543 int ret; 4544 4545 if (!f2fs_sb_has_compression(sbi) || 4546 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4547 return -EOPNOTSUPP; 4548 4549 if (!(filp->f_mode & FMODE_WRITE)) 4550 return -EBADF; 4551 4552 f2fs_balance_fs(sbi, true); 4553 4554 ret = mnt_want_write_file(filp); 4555 if (ret) 4556 return ret; 4557 inode_lock(inode); 4558 4559 if (!f2fs_is_compress_backend_ready(inode)) { 4560 ret = -EOPNOTSUPP; 4561 goto out; 4562 } 4563 4564 if (!f2fs_compressed_file(inode) || 4565 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4566 ret = -EINVAL; 4567 goto out; 4568 } 4569 4570 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4571 if (ret) 4572 goto out; 4573 4574 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4575 4576 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4577 last_idx >>= fi->i_log_cluster_size; 4578 4579 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4580 page_idx = cluster_idx << fi->i_log_cluster_size; 4581 4582 if (f2fs_is_sparse_cluster(inode, page_idx)) 4583 continue; 4584 4585 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4586 if (ret < 0) 4587 break; 4588 4589 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4590 ret = filemap_fdatawrite(inode->i_mapping); 4591 if (ret < 0) 4592 break; 4593 } 4594 4595 cond_resched(); 4596 if (fatal_signal_pending(current)) { 4597 ret = -EINTR; 4598 break; 4599 } 4600 } 4601 4602 if (!ret) 4603 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4604 LLONG_MAX); 4605 4606 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4607 4608 if (ret) 4609 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4610 __func__, ret); 4611 f2fs_update_time(sbi, REQ_TIME); 4612 out: 4613 inode_unlock(inode); 4614 mnt_drop_write_file(filp); 4615 4616 return ret; 4617 } 4618 4619 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4620 { 4621 switch (cmd) { 4622 case FS_IOC_GETVERSION: 4623 return f2fs_ioc_getversion(filp, arg); 4624 case F2FS_IOC_START_ATOMIC_WRITE: 4625 return f2fs_ioc_start_atomic_write(filp, false); 4626 case F2FS_IOC_START_ATOMIC_REPLACE: 4627 return f2fs_ioc_start_atomic_write(filp, true); 4628 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4629 return f2fs_ioc_commit_atomic_write(filp); 4630 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4631 return f2fs_ioc_abort_atomic_write(filp); 4632 case F2FS_IOC_START_VOLATILE_WRITE: 4633 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4634 return -EOPNOTSUPP; 4635 case F2FS_IOC_SHUTDOWN: 4636 return f2fs_ioc_shutdown(filp, arg); 4637 case FITRIM: 4638 return f2fs_ioc_fitrim(filp, arg); 4639 case FS_IOC_SET_ENCRYPTION_POLICY: 4640 return f2fs_ioc_set_encryption_policy(filp, arg); 4641 case FS_IOC_GET_ENCRYPTION_POLICY: 4642 return f2fs_ioc_get_encryption_policy(filp, arg); 4643 case FS_IOC_GET_ENCRYPTION_PWSALT: 4644 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4645 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4646 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4647 case FS_IOC_ADD_ENCRYPTION_KEY: 4648 return f2fs_ioc_add_encryption_key(filp, arg); 4649 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4650 return f2fs_ioc_remove_encryption_key(filp, arg); 4651 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4652 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4653 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4654 return f2fs_ioc_get_encryption_key_status(filp, arg); 4655 case FS_IOC_GET_ENCRYPTION_NONCE: 4656 return f2fs_ioc_get_encryption_nonce(filp, arg); 4657 case F2FS_IOC_GARBAGE_COLLECT: 4658 return f2fs_ioc_gc(filp, arg); 4659 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4660 return f2fs_ioc_gc_range(filp, arg); 4661 case F2FS_IOC_WRITE_CHECKPOINT: 4662 return f2fs_ioc_write_checkpoint(filp); 4663 case F2FS_IOC_DEFRAGMENT: 4664 return f2fs_ioc_defragment(filp, arg); 4665 case F2FS_IOC_MOVE_RANGE: 4666 return f2fs_ioc_move_range(filp, arg); 4667 case F2FS_IOC_FLUSH_DEVICE: 4668 return f2fs_ioc_flush_device(filp, arg); 4669 case F2FS_IOC_GET_FEATURES: 4670 return f2fs_ioc_get_features(filp, arg); 4671 case F2FS_IOC_GET_PIN_FILE: 4672 return f2fs_ioc_get_pin_file(filp, arg); 4673 case F2FS_IOC_SET_PIN_FILE: 4674 return f2fs_ioc_set_pin_file(filp, arg); 4675 case F2FS_IOC_PRECACHE_EXTENTS: 4676 return f2fs_ioc_precache_extents(filp); 4677 case F2FS_IOC_RESIZE_FS: 4678 return f2fs_ioc_resize_fs(filp, arg); 4679 case FS_IOC_ENABLE_VERITY: 4680 return f2fs_ioc_enable_verity(filp, arg); 4681 case FS_IOC_MEASURE_VERITY: 4682 return f2fs_ioc_measure_verity(filp, arg); 4683 case FS_IOC_READ_VERITY_METADATA: 4684 return f2fs_ioc_read_verity_metadata(filp, arg); 4685 case FS_IOC_GETFSLABEL: 4686 return f2fs_ioc_getfslabel(filp, arg); 4687 case FS_IOC_SETFSLABEL: 4688 return f2fs_ioc_setfslabel(filp, arg); 4689 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4690 return f2fs_ioc_get_compress_blocks(filp, arg); 4691 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4692 return f2fs_release_compress_blocks(filp, arg); 4693 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4694 return f2fs_reserve_compress_blocks(filp, arg); 4695 case F2FS_IOC_SEC_TRIM_FILE: 4696 return f2fs_sec_trim_file(filp, arg); 4697 case F2FS_IOC_GET_COMPRESS_OPTION: 4698 return f2fs_ioc_get_compress_option(filp, arg); 4699 case F2FS_IOC_SET_COMPRESS_OPTION: 4700 return f2fs_ioc_set_compress_option(filp, arg); 4701 case F2FS_IOC_DECOMPRESS_FILE: 4702 return f2fs_ioc_decompress_file(filp); 4703 case F2FS_IOC_COMPRESS_FILE: 4704 return f2fs_ioc_compress_file(filp); 4705 case F2FS_IOC_GET_DEV_ALIAS_FILE: 4706 return f2fs_ioc_get_dev_alias_file(filp, arg); 4707 case F2FS_IOC_IO_PRIO: 4708 return f2fs_ioc_io_prio(filp, arg); 4709 default: 4710 return -ENOTTY; 4711 } 4712 } 4713 4714 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4715 { 4716 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4717 return -EIO; 4718 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4719 return -ENOSPC; 4720 4721 return __f2fs_ioctl(filp, cmd, arg); 4722 } 4723 4724 /* 4725 * Return %true if the given read or write request should use direct I/O, or 4726 * %false if it should use buffered I/O. 4727 */ 4728 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4729 struct iov_iter *iter) 4730 { 4731 unsigned int align; 4732 4733 if (!(iocb->ki_flags & IOCB_DIRECT)) 4734 return false; 4735 4736 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4737 return false; 4738 4739 /* 4740 * Direct I/O not aligned to the disk's logical_block_size will be 4741 * attempted, but will fail with -EINVAL. 4742 * 4743 * f2fs additionally requires that direct I/O be aligned to the 4744 * filesystem block size, which is often a stricter requirement. 4745 * However, f2fs traditionally falls back to buffered I/O on requests 4746 * that are logical_block_size-aligned but not fs-block aligned. 4747 * 4748 * The below logic implements this behavior. 4749 */ 4750 align = iocb->ki_pos | iov_iter_alignment(iter); 4751 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4752 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4753 return false; 4754 4755 return true; 4756 } 4757 4758 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4759 unsigned int flags) 4760 { 4761 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4762 4763 dec_page_count(sbi, F2FS_DIO_READ); 4764 if (error) 4765 return error; 4766 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4767 return 0; 4768 } 4769 4770 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4771 .end_io = f2fs_dio_read_end_io, 4772 }; 4773 4774 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4775 { 4776 struct file *file = iocb->ki_filp; 4777 struct inode *inode = file_inode(file); 4778 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4779 struct f2fs_inode_info *fi = F2FS_I(inode); 4780 const loff_t pos = iocb->ki_pos; 4781 const size_t count = iov_iter_count(to); 4782 struct iomap_dio *dio; 4783 ssize_t ret; 4784 4785 if (count == 0) 4786 return 0; /* skip atime update */ 4787 4788 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4789 4790 if (iocb->ki_flags & IOCB_NOWAIT) { 4791 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4792 ret = -EAGAIN; 4793 goto out; 4794 } 4795 } else { 4796 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4797 } 4798 4799 /* dio is not compatible w/ atomic file */ 4800 if (f2fs_is_atomic_file(inode)) { 4801 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4802 ret = -EOPNOTSUPP; 4803 goto out; 4804 } 4805 4806 /* 4807 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4808 * the higher-level function iomap_dio_rw() in order to ensure that the 4809 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4810 */ 4811 inc_page_count(sbi, F2FS_DIO_READ); 4812 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4813 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4814 if (IS_ERR_OR_NULL(dio)) { 4815 ret = PTR_ERR_OR_ZERO(dio); 4816 if (ret != -EIOCBQUEUED) 4817 dec_page_count(sbi, F2FS_DIO_READ); 4818 } else { 4819 ret = iomap_dio_complete(dio); 4820 } 4821 4822 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4823 4824 file_accessed(file); 4825 out: 4826 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4827 return ret; 4828 } 4829 4830 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4831 int rw) 4832 { 4833 struct inode *inode = file_inode(file); 4834 char *buf, *path; 4835 4836 buf = f2fs_getname(F2FS_I_SB(inode)); 4837 if (!buf) 4838 return; 4839 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4840 if (IS_ERR(path)) 4841 goto free_buf; 4842 if (rw == WRITE) 4843 trace_f2fs_datawrite_start(inode, pos, count, 4844 current->pid, path, current->comm); 4845 else 4846 trace_f2fs_dataread_start(inode, pos, count, 4847 current->pid, path, current->comm); 4848 free_buf: 4849 f2fs_putname(buf); 4850 } 4851 4852 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4853 { 4854 struct inode *inode = file_inode(iocb->ki_filp); 4855 const loff_t pos = iocb->ki_pos; 4856 ssize_t ret; 4857 bool dio; 4858 4859 if (!f2fs_is_compress_backend_ready(inode)) 4860 return -EOPNOTSUPP; 4861 4862 if (trace_f2fs_dataread_start_enabled()) 4863 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4864 iov_iter_count(to), READ); 4865 4866 dio = f2fs_should_use_dio(inode, iocb, to); 4867 4868 /* In LFS mode, if there is inflight dio, wait for its completion */ 4869 if (f2fs_lfs_mode(F2FS_I_SB(inode)) && 4870 get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE) && 4871 (!f2fs_is_pinned_file(inode) || !dio)) 4872 inode_dio_wait(inode); 4873 4874 if (dio) { 4875 ret = f2fs_dio_read_iter(iocb, to); 4876 } else { 4877 ret = filemap_read(iocb, to, 0); 4878 if (ret > 0) 4879 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4880 APP_BUFFERED_READ_IO, ret); 4881 } 4882 trace_f2fs_dataread_end(inode, pos, ret); 4883 return ret; 4884 } 4885 4886 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4887 struct pipe_inode_info *pipe, 4888 size_t len, unsigned int flags) 4889 { 4890 struct inode *inode = file_inode(in); 4891 const loff_t pos = *ppos; 4892 ssize_t ret; 4893 4894 if (!f2fs_is_compress_backend_ready(inode)) 4895 return -EOPNOTSUPP; 4896 4897 if (trace_f2fs_dataread_start_enabled()) 4898 f2fs_trace_rw_file_path(in, pos, len, READ); 4899 4900 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4901 if (ret > 0) 4902 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4903 APP_BUFFERED_READ_IO, ret); 4904 4905 trace_f2fs_dataread_end(inode, pos, ret); 4906 return ret; 4907 } 4908 4909 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4910 { 4911 struct file *file = iocb->ki_filp; 4912 struct inode *inode = file_inode(file); 4913 ssize_t count; 4914 int err; 4915 4916 if (IS_IMMUTABLE(inode)) 4917 return -EPERM; 4918 4919 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4920 return -EPERM; 4921 4922 count = generic_write_checks(iocb, from); 4923 if (count <= 0) 4924 return count; 4925 4926 err = file_modified(file); 4927 if (err) 4928 return err; 4929 4930 f2fs_zero_post_eof_page(inode, 4931 iocb->ki_pos + iov_iter_count(from), true); 4932 return count; 4933 } 4934 4935 /* 4936 * Preallocate blocks for a write request, if it is possible and helpful to do 4937 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4938 * blocks were preallocated, or a negative errno value if something went 4939 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4940 * requested blocks (not just some of them) have been allocated. 4941 */ 4942 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4943 bool dio) 4944 { 4945 struct inode *inode = file_inode(iocb->ki_filp); 4946 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4947 const loff_t pos = iocb->ki_pos; 4948 const size_t count = iov_iter_count(iter); 4949 struct f2fs_map_blocks map = {}; 4950 int flag; 4951 int ret; 4952 4953 /* If it will be an out-of-place direct write, don't bother. */ 4954 if (dio && f2fs_lfs_mode(sbi)) 4955 return 0; 4956 /* 4957 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4958 * buffered IO, if DIO meets any holes. 4959 */ 4960 if (dio && i_size_read(inode) && 4961 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4962 return 0; 4963 4964 /* No-wait I/O can't allocate blocks. */ 4965 if (iocb->ki_flags & IOCB_NOWAIT) 4966 return 0; 4967 4968 /* If it will be a short write, don't bother. */ 4969 if (fault_in_iov_iter_readable(iter, count)) 4970 return 0; 4971 4972 if (f2fs_has_inline_data(inode)) { 4973 /* If the data will fit inline, don't bother. */ 4974 if (pos + count <= MAX_INLINE_DATA(inode)) 4975 return 0; 4976 ret = f2fs_convert_inline_inode(inode); 4977 if (ret) 4978 return ret; 4979 } 4980 4981 /* Do not preallocate blocks that will be written partially in 4KB. */ 4982 map.m_lblk = F2FS_BLK_ALIGN(pos); 4983 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4984 if (map.m_len > map.m_lblk) 4985 map.m_len -= map.m_lblk; 4986 else 4987 return 0; 4988 4989 if (!IS_DEVICE_ALIASING(inode)) 4990 map.m_may_create = true; 4991 if (dio) { 4992 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi, 4993 inode->i_write_hint); 4994 flag = F2FS_GET_BLOCK_PRE_DIO; 4995 } else { 4996 map.m_seg_type = NO_CHECK_TYPE; 4997 flag = F2FS_GET_BLOCK_PRE_AIO; 4998 } 4999 5000 ret = f2fs_map_blocks(inode, &map, flag); 5001 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 5002 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 5003 return ret; 5004 if (ret == 0) 5005 set_inode_flag(inode, FI_PREALLOCATED_ALL); 5006 return map.m_len; 5007 } 5008 5009 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 5010 struct iov_iter *from) 5011 { 5012 struct file *file = iocb->ki_filp; 5013 struct inode *inode = file_inode(file); 5014 ssize_t ret; 5015 5016 if (iocb->ki_flags & IOCB_NOWAIT) 5017 return -EOPNOTSUPP; 5018 5019 ret = generic_perform_write(iocb, from); 5020 5021 if (ret > 0) { 5022 f2fs_update_iostat(F2FS_I_SB(inode), inode, 5023 APP_BUFFERED_IO, ret); 5024 } 5025 return ret; 5026 } 5027 5028 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 5029 unsigned int flags) 5030 { 5031 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 5032 5033 dec_page_count(sbi, F2FS_DIO_WRITE); 5034 if (error) 5035 return error; 5036 f2fs_update_time(sbi, REQ_TIME); 5037 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 5038 return 0; 5039 } 5040 5041 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter, 5042 struct bio *bio, loff_t file_offset) 5043 { 5044 struct inode *inode = iter->inode; 5045 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 5046 enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint); 5047 enum temp_type temp = f2fs_get_segment_temp(sbi, type); 5048 5049 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp); 5050 submit_bio(bio); 5051 } 5052 5053 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 5054 .end_io = f2fs_dio_write_end_io, 5055 .submit_io = f2fs_dio_write_submit_io, 5056 }; 5057 5058 static void f2fs_flush_buffered_write(struct address_space *mapping, 5059 loff_t start_pos, loff_t end_pos) 5060 { 5061 int ret; 5062 5063 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 5064 if (ret < 0) 5065 return; 5066 invalidate_mapping_pages(mapping, 5067 start_pos >> PAGE_SHIFT, 5068 end_pos >> PAGE_SHIFT); 5069 } 5070 5071 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 5072 bool *may_need_sync) 5073 { 5074 struct file *file = iocb->ki_filp; 5075 struct inode *inode = file_inode(file); 5076 struct f2fs_inode_info *fi = F2FS_I(inode); 5077 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 5078 const bool do_opu = f2fs_lfs_mode(sbi); 5079 const loff_t pos = iocb->ki_pos; 5080 const ssize_t count = iov_iter_count(from); 5081 unsigned int dio_flags; 5082 struct iomap_dio *dio; 5083 ssize_t ret; 5084 5085 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 5086 5087 if (iocb->ki_flags & IOCB_NOWAIT) { 5088 /* f2fs_convert_inline_inode() and block allocation can block */ 5089 if (f2fs_has_inline_data(inode) || 5090 !f2fs_overwrite_io(inode, pos, count)) { 5091 ret = -EAGAIN; 5092 goto out; 5093 } 5094 5095 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 5096 ret = -EAGAIN; 5097 goto out; 5098 } 5099 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 5100 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5101 ret = -EAGAIN; 5102 goto out; 5103 } 5104 } else { 5105 ret = f2fs_convert_inline_inode(inode); 5106 if (ret) 5107 goto out; 5108 5109 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 5110 if (do_opu) 5111 f2fs_down_read(&fi->i_gc_rwsem[READ]); 5112 } 5113 5114 /* 5115 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 5116 * the higher-level function iomap_dio_rw() in order to ensure that the 5117 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 5118 */ 5119 inc_page_count(sbi, F2FS_DIO_WRITE); 5120 dio_flags = 0; 5121 if (pos + count > inode->i_size) 5122 dio_flags |= IOMAP_DIO_FORCE_WAIT; 5123 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 5124 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 5125 if (IS_ERR_OR_NULL(dio)) { 5126 ret = PTR_ERR_OR_ZERO(dio); 5127 if (ret == -ENOTBLK) 5128 ret = 0; 5129 if (ret != -EIOCBQUEUED) 5130 dec_page_count(sbi, F2FS_DIO_WRITE); 5131 } else { 5132 ret = iomap_dio_complete(dio); 5133 } 5134 5135 if (do_opu) 5136 f2fs_up_read(&fi->i_gc_rwsem[READ]); 5137 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5138 5139 if (ret < 0) 5140 goto out; 5141 if (pos + ret > inode->i_size) 5142 f2fs_i_size_write(inode, pos + ret); 5143 if (!do_opu) 5144 set_inode_flag(inode, FI_UPDATE_WRITE); 5145 5146 if (iov_iter_count(from)) { 5147 ssize_t ret2; 5148 loff_t bufio_start_pos = iocb->ki_pos; 5149 5150 /* 5151 * The direct write was partial, so we need to fall back to a 5152 * buffered write for the remainder. 5153 */ 5154 5155 ret2 = f2fs_buffered_write_iter(iocb, from); 5156 if (iov_iter_count(from)) 5157 f2fs_write_failed(inode, iocb->ki_pos); 5158 if (ret2 < 0) 5159 goto out; 5160 5161 /* 5162 * Ensure that the pagecache pages are written to disk and 5163 * invalidated to preserve the expected O_DIRECT semantics. 5164 */ 5165 if (ret2 > 0) { 5166 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 5167 5168 ret += ret2; 5169 5170 f2fs_flush_buffered_write(file->f_mapping, 5171 bufio_start_pos, 5172 bufio_end_pos); 5173 } 5174 } else { 5175 /* iomap_dio_rw() already handled the generic_write_sync(). */ 5176 *may_need_sync = false; 5177 } 5178 out: 5179 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 5180 return ret; 5181 } 5182 5183 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 5184 { 5185 struct inode *inode = file_inode(iocb->ki_filp); 5186 const loff_t orig_pos = iocb->ki_pos; 5187 const size_t orig_count = iov_iter_count(from); 5188 loff_t target_size; 5189 bool dio; 5190 bool may_need_sync = true; 5191 int preallocated; 5192 const loff_t pos = iocb->ki_pos; 5193 const ssize_t count = iov_iter_count(from); 5194 ssize_t ret; 5195 5196 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 5197 ret = -EIO; 5198 goto out; 5199 } 5200 5201 if (!f2fs_is_compress_backend_ready(inode)) { 5202 ret = -EOPNOTSUPP; 5203 goto out; 5204 } 5205 5206 if (iocb->ki_flags & IOCB_NOWAIT) { 5207 if (!inode_trylock(inode)) { 5208 ret = -EAGAIN; 5209 goto out; 5210 } 5211 } else { 5212 inode_lock(inode); 5213 } 5214 5215 if (f2fs_is_pinned_file(inode) && 5216 !f2fs_overwrite_io(inode, pos, count)) { 5217 ret = -EIO; 5218 goto out_unlock; 5219 } 5220 5221 ret = f2fs_write_checks(iocb, from); 5222 if (ret <= 0) 5223 goto out_unlock; 5224 5225 /* Determine whether we will do a direct write or a buffered write. */ 5226 dio = f2fs_should_use_dio(inode, iocb, from); 5227 5228 /* dio is not compatible w/ atomic write */ 5229 if (dio && f2fs_is_atomic_file(inode)) { 5230 ret = -EOPNOTSUPP; 5231 goto out_unlock; 5232 } 5233 5234 /* Possibly preallocate the blocks for the write. */ 5235 target_size = iocb->ki_pos + iov_iter_count(from); 5236 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 5237 if (preallocated < 0) { 5238 ret = preallocated; 5239 } else { 5240 if (trace_f2fs_datawrite_start_enabled()) 5241 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 5242 orig_count, WRITE); 5243 5244 /* Do the actual write. */ 5245 ret = dio ? 5246 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 5247 f2fs_buffered_write_iter(iocb, from); 5248 5249 trace_f2fs_datawrite_end(inode, orig_pos, ret); 5250 } 5251 5252 /* Don't leave any preallocated blocks around past i_size. */ 5253 if (preallocated && i_size_read(inode) < target_size) { 5254 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5255 filemap_invalidate_lock(inode->i_mapping); 5256 if (!f2fs_truncate(inode)) 5257 file_dont_truncate(inode); 5258 filemap_invalidate_unlock(inode->i_mapping); 5259 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5260 } else { 5261 file_dont_truncate(inode); 5262 } 5263 5264 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 5265 out_unlock: 5266 inode_unlock(inode); 5267 out: 5268 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 5269 5270 if (ret > 0 && may_need_sync) 5271 ret = generic_write_sync(iocb, ret); 5272 5273 /* If buffered IO was forced, flush and drop the data from 5274 * the page cache to preserve O_DIRECT semantics 5275 */ 5276 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 5277 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 5278 orig_pos, 5279 orig_pos + ret - 1); 5280 5281 return ret; 5282 } 5283 5284 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 5285 int advice) 5286 { 5287 struct address_space *mapping; 5288 struct backing_dev_info *bdi; 5289 struct inode *inode = file_inode(filp); 5290 int err; 5291 5292 trace_f2fs_fadvise(inode, offset, len, advice); 5293 5294 if (advice == POSIX_FADV_SEQUENTIAL) { 5295 if (S_ISFIFO(inode->i_mode)) 5296 return -ESPIPE; 5297 5298 mapping = filp->f_mapping; 5299 if (!mapping || len < 0) 5300 return -EINVAL; 5301 5302 bdi = inode_to_bdi(mapping->host); 5303 filp->f_ra.ra_pages = bdi->ra_pages * 5304 F2FS_I_SB(inode)->seq_file_ra_mul; 5305 spin_lock(&filp->f_lock); 5306 filp->f_mode &= ~FMODE_RANDOM; 5307 spin_unlock(&filp->f_lock); 5308 return 0; 5309 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 5310 /* Load extent cache at the first readahead. */ 5311 f2fs_precache_extents(inode); 5312 } 5313 5314 err = generic_fadvise(filp, offset, len, advice); 5315 if (err) 5316 return err; 5317 5318 if (advice == POSIX_FADV_DONTNEED && 5319 (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 5320 f2fs_compressed_file(inode))) 5321 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 5322 else if (advice == POSIX_FADV_NOREUSE) 5323 err = f2fs_keep_noreuse_range(inode, offset, len); 5324 return err; 5325 } 5326 5327 #ifdef CONFIG_COMPAT 5328 struct compat_f2fs_gc_range { 5329 u32 sync; 5330 compat_u64 start; 5331 compat_u64 len; 5332 }; 5333 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 5334 struct compat_f2fs_gc_range) 5335 5336 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 5337 { 5338 struct compat_f2fs_gc_range __user *urange; 5339 struct f2fs_gc_range range; 5340 int err; 5341 5342 urange = compat_ptr(arg); 5343 err = get_user(range.sync, &urange->sync); 5344 err |= get_user(range.start, &urange->start); 5345 err |= get_user(range.len, &urange->len); 5346 if (err) 5347 return -EFAULT; 5348 5349 return __f2fs_ioc_gc_range(file, &range); 5350 } 5351 5352 struct compat_f2fs_move_range { 5353 u32 dst_fd; 5354 compat_u64 pos_in; 5355 compat_u64 pos_out; 5356 compat_u64 len; 5357 }; 5358 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 5359 struct compat_f2fs_move_range) 5360 5361 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 5362 { 5363 struct compat_f2fs_move_range __user *urange; 5364 struct f2fs_move_range range; 5365 int err; 5366 5367 urange = compat_ptr(arg); 5368 err = get_user(range.dst_fd, &urange->dst_fd); 5369 err |= get_user(range.pos_in, &urange->pos_in); 5370 err |= get_user(range.pos_out, &urange->pos_out); 5371 err |= get_user(range.len, &urange->len); 5372 if (err) 5373 return -EFAULT; 5374 5375 return __f2fs_ioc_move_range(file, &range); 5376 } 5377 5378 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5379 { 5380 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 5381 return -EIO; 5382 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 5383 return -ENOSPC; 5384 5385 switch (cmd) { 5386 case FS_IOC32_GETVERSION: 5387 cmd = FS_IOC_GETVERSION; 5388 break; 5389 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 5390 return f2fs_compat_ioc_gc_range(file, arg); 5391 case F2FS_IOC32_MOVE_RANGE: 5392 return f2fs_compat_ioc_move_range(file, arg); 5393 case F2FS_IOC_START_ATOMIC_WRITE: 5394 case F2FS_IOC_START_ATOMIC_REPLACE: 5395 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5396 case F2FS_IOC_START_VOLATILE_WRITE: 5397 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5398 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5399 case F2FS_IOC_SHUTDOWN: 5400 case FITRIM: 5401 case FS_IOC_SET_ENCRYPTION_POLICY: 5402 case FS_IOC_GET_ENCRYPTION_PWSALT: 5403 case FS_IOC_GET_ENCRYPTION_POLICY: 5404 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5405 case FS_IOC_ADD_ENCRYPTION_KEY: 5406 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5407 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5408 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5409 case FS_IOC_GET_ENCRYPTION_NONCE: 5410 case F2FS_IOC_GARBAGE_COLLECT: 5411 case F2FS_IOC_WRITE_CHECKPOINT: 5412 case F2FS_IOC_DEFRAGMENT: 5413 case F2FS_IOC_FLUSH_DEVICE: 5414 case F2FS_IOC_GET_FEATURES: 5415 case F2FS_IOC_GET_PIN_FILE: 5416 case F2FS_IOC_SET_PIN_FILE: 5417 case F2FS_IOC_PRECACHE_EXTENTS: 5418 case F2FS_IOC_RESIZE_FS: 5419 case FS_IOC_ENABLE_VERITY: 5420 case FS_IOC_MEASURE_VERITY: 5421 case FS_IOC_READ_VERITY_METADATA: 5422 case FS_IOC_GETFSLABEL: 5423 case FS_IOC_SETFSLABEL: 5424 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5425 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5426 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5427 case F2FS_IOC_SEC_TRIM_FILE: 5428 case F2FS_IOC_GET_COMPRESS_OPTION: 5429 case F2FS_IOC_SET_COMPRESS_OPTION: 5430 case F2FS_IOC_DECOMPRESS_FILE: 5431 case F2FS_IOC_COMPRESS_FILE: 5432 case F2FS_IOC_GET_DEV_ALIAS_FILE: 5433 case F2FS_IOC_IO_PRIO: 5434 break; 5435 default: 5436 return -ENOIOCTLCMD; 5437 } 5438 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5439 } 5440 #endif 5441 5442 const struct file_operations f2fs_file_operations = { 5443 .llseek = f2fs_llseek, 5444 .read_iter = f2fs_file_read_iter, 5445 .write_iter = f2fs_file_write_iter, 5446 .iopoll = iocb_bio_iopoll, 5447 .open = f2fs_file_open, 5448 .release = f2fs_release_file, 5449 .mmap_prepare = f2fs_file_mmap_prepare, 5450 .flush = f2fs_file_flush, 5451 .fsync = f2fs_sync_file, 5452 .fallocate = f2fs_fallocate, 5453 .unlocked_ioctl = f2fs_ioctl, 5454 #ifdef CONFIG_COMPAT 5455 .compat_ioctl = f2fs_compat_ioctl, 5456 #endif 5457 .splice_read = f2fs_file_splice_read, 5458 .splice_write = iter_file_splice_write, 5459 .fadvise = f2fs_file_fadvise, 5460 .fop_flags = FOP_BUFFER_RASYNC, 5461 .setlease = generic_setlease, 5462 }; 5463