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/types.h> 15 #include <linux/compat.h> 16 #include <linux/uaccess.h> 17 #include <linux/mount.h> 18 #include <linux/pagevec.h> 19 #include <linux/uio.h> 20 #include <linux/uuid.h> 21 #include <linux/file.h> 22 #include <linux/nls.h> 23 #include <linux/sched/signal.h> 24 #include <linux/fileattr.h> 25 #include <linux/fadvise.h> 26 #include <linux/iomap.h> 27 28 #include "f2fs.h" 29 #include "node.h" 30 #include "segment.h" 31 #include "xattr.h" 32 #include "acl.h" 33 #include "gc.h" 34 #include "iostat.h" 35 #include <trace/events/f2fs.h> 36 #include <uapi/linux/f2fs.h> 37 38 static void f2fs_zero_post_eof_page(struct inode *inode, 39 loff_t new_size, bool lock) 40 { 41 loff_t old_size = i_size_read(inode); 42 43 if (old_size >= new_size) 44 return; 45 46 if (mapping_empty(inode->i_mapping)) 47 return; 48 49 if (lock) 50 filemap_invalidate_lock(inode->i_mapping); 51 /* zero or drop pages only in range of [old_size, new_size] */ 52 truncate_inode_pages_range(inode->i_mapping, old_size, new_size); 53 if (lock) 54 filemap_invalidate_unlock(inode->i_mapping); 55 } 56 57 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 58 { 59 struct inode *inode = file_inode(vmf->vma->vm_file); 60 vm_flags_t flags = vmf->vma->vm_flags; 61 vm_fault_t ret; 62 63 ret = filemap_fault(vmf); 64 if (ret & VM_FAULT_LOCKED) 65 f2fs_update_iostat(F2FS_I_SB(inode), inode, 66 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 67 68 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret); 69 70 return ret; 71 } 72 73 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 74 { 75 struct folio *folio = page_folio(vmf->page); 76 struct inode *inode = file_inode(vmf->vma->vm_file); 77 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 78 struct dnode_of_data dn; 79 bool need_alloc = !f2fs_is_pinned_file(inode); 80 int err = 0; 81 vm_fault_t ret; 82 83 if (unlikely(IS_IMMUTABLE(inode))) 84 return VM_FAULT_SIGBUS; 85 86 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 87 err = -EIO; 88 goto out; 89 } 90 91 if (unlikely(f2fs_cp_error(sbi))) { 92 err = -EIO; 93 goto out; 94 } 95 96 if (!f2fs_is_checkpoint_ready(sbi)) { 97 err = -ENOSPC; 98 goto out; 99 } 100 101 err = f2fs_convert_inline_inode(inode); 102 if (err) 103 goto out; 104 105 #ifdef CONFIG_F2FS_FS_COMPRESSION 106 if (f2fs_compressed_file(inode)) { 107 int ret = f2fs_is_compressed_cluster(inode, folio->index); 108 109 if (ret < 0) { 110 err = ret; 111 goto out; 112 } else if (ret) { 113 need_alloc = false; 114 } 115 } 116 #endif 117 /* should do out of any locked page */ 118 if (need_alloc) 119 f2fs_balance_fs(sbi, true); 120 121 sb_start_pagefault(inode->i_sb); 122 123 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 124 125 f2fs_zero_post_eof_page(inode, (folio->index + 1) << PAGE_SHIFT, true); 126 127 file_update_time(vmf->vma->vm_file); 128 filemap_invalidate_lock_shared(inode->i_mapping); 129 130 folio_lock(folio); 131 if (unlikely(folio->mapping != inode->i_mapping || 132 folio_pos(folio) > i_size_read(inode) || 133 !folio_test_uptodate(folio))) { 134 folio_unlock(folio); 135 err = -EFAULT; 136 goto out_sem; 137 } 138 139 set_new_dnode(&dn, inode, NULL, NULL, 0); 140 if (need_alloc) { 141 /* block allocation */ 142 err = f2fs_get_block_locked(&dn, folio->index); 143 } else { 144 err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE); 145 f2fs_put_dnode(&dn); 146 if (f2fs_is_pinned_file(inode) && 147 !__is_valid_data_blkaddr(dn.data_blkaddr)) 148 err = -EIO; 149 } 150 151 if (err) { 152 folio_unlock(folio); 153 goto out_sem; 154 } 155 156 f2fs_folio_wait_writeback(folio, DATA, false, true); 157 158 /* wait for GCed page writeback via META_MAPPING */ 159 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 160 161 /* 162 * check to see if the page is mapped already (no holes) 163 */ 164 if (folio_test_mappedtodisk(folio)) 165 goto out_sem; 166 167 /* page is wholly or partially inside EOF */ 168 if (((loff_t)(folio->index + 1) << PAGE_SHIFT) > 169 i_size_read(inode)) { 170 loff_t offset; 171 172 offset = i_size_read(inode) & ~PAGE_MASK; 173 folio_zero_segment(folio, offset, folio_size(folio)); 174 } 175 folio_mark_dirty(folio); 176 177 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 178 f2fs_update_time(sbi, REQ_TIME); 179 180 out_sem: 181 filemap_invalidate_unlock_shared(inode->i_mapping); 182 183 sb_end_pagefault(inode->i_sb); 184 out: 185 ret = vmf_fs_error(err); 186 187 trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret); 188 return ret; 189 } 190 191 static const struct vm_operations_struct f2fs_file_vm_ops = { 192 .fault = f2fs_filemap_fault, 193 .map_pages = filemap_map_pages, 194 .page_mkwrite = f2fs_vm_page_mkwrite, 195 }; 196 197 static int get_parent_ino(struct inode *inode, nid_t *pino) 198 { 199 struct dentry *dentry; 200 201 /* 202 * Make sure to get the non-deleted alias. The alias associated with 203 * the open file descriptor being fsync()'ed may be deleted already. 204 */ 205 dentry = d_find_alias(inode); 206 if (!dentry) 207 return 0; 208 209 *pino = d_parent_ino(dentry); 210 dput(dentry); 211 return 1; 212 } 213 214 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 215 { 216 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 217 enum cp_reason_type cp_reason = CP_NO_NEEDED; 218 219 if (!S_ISREG(inode->i_mode)) 220 cp_reason = CP_NON_REGULAR; 221 else if (f2fs_compressed_file(inode)) 222 cp_reason = CP_COMPRESSED; 223 else if (inode->i_nlink != 1) 224 cp_reason = CP_HARDLINK; 225 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 226 cp_reason = CP_SB_NEED_CP; 227 else if (file_wrong_pino(inode)) 228 cp_reason = CP_WRONG_PINO; 229 else if (!f2fs_space_for_roll_forward(sbi)) 230 cp_reason = CP_NO_SPC_ROLL; 231 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 232 cp_reason = CP_NODE_NEED_CP; 233 else if (test_opt(sbi, FASTBOOT)) 234 cp_reason = CP_FASTBOOT_MODE; 235 else if (F2FS_OPTION(sbi).active_logs == 2) 236 cp_reason = CP_SPEC_LOG_NUM; 237 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 238 f2fs_need_dentry_mark(sbi, inode->i_ino) && 239 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 240 TRANS_DIR_INO)) 241 cp_reason = CP_RECOVER_DIR; 242 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 243 XATTR_DIR_INO)) 244 cp_reason = CP_XATTR_DIR; 245 246 return cp_reason; 247 } 248 249 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 250 { 251 struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino); 252 bool ret = false; 253 /* But we need to avoid that there are some inode updates */ 254 if ((!IS_ERR(i) && folio_test_dirty(i)) || 255 f2fs_need_inode_block_update(sbi, ino)) 256 ret = true; 257 f2fs_folio_put(i, false); 258 return ret; 259 } 260 261 static void try_to_fix_pino(struct inode *inode) 262 { 263 struct f2fs_inode_info *fi = F2FS_I(inode); 264 nid_t pino; 265 266 f2fs_down_write(&fi->i_sem); 267 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 268 get_parent_ino(inode, &pino)) { 269 f2fs_i_pino_write(inode, pino); 270 file_got_pino(inode); 271 } 272 f2fs_up_write(&fi->i_sem); 273 } 274 275 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 276 int datasync, bool atomic) 277 { 278 struct inode *inode = file->f_mapping->host; 279 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 280 nid_t ino = inode->i_ino; 281 int ret = 0; 282 enum cp_reason_type cp_reason = 0; 283 struct writeback_control wbc = { 284 .sync_mode = WB_SYNC_ALL, 285 .nr_to_write = LONG_MAX, 286 }; 287 unsigned int seq_id = 0; 288 289 if (unlikely(f2fs_readonly(inode->i_sb))) 290 return 0; 291 292 trace_f2fs_sync_file_enter(inode); 293 294 if (S_ISDIR(inode->i_mode)) 295 goto go_write; 296 297 /* if fdatasync is triggered, let's do in-place-update */ 298 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 299 set_inode_flag(inode, FI_NEED_IPU); 300 ret = file_write_and_wait_range(file, start, end); 301 clear_inode_flag(inode, FI_NEED_IPU); 302 303 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 304 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 305 return ret; 306 } 307 308 /* if the inode is dirty, let's recover all the time */ 309 if (!f2fs_skip_inode_update(inode, datasync)) { 310 f2fs_write_inode(inode, NULL); 311 goto go_write; 312 } 313 314 /* 315 * if there is no written data, don't waste time to write recovery info. 316 */ 317 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 318 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 319 320 /* it may call write_inode just prior to fsync */ 321 if (need_inode_page_update(sbi, ino)) 322 goto go_write; 323 324 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 325 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 326 goto flush_out; 327 goto out; 328 } else { 329 /* 330 * for OPU case, during fsync(), node can be persisted before 331 * data when lower device doesn't support write barrier, result 332 * in data corruption after SPO. 333 * So for strict fsync mode, force to use atomic write semantics 334 * to keep write order in between data/node and last node to 335 * avoid potential data corruption. 336 */ 337 if (F2FS_OPTION(sbi).fsync_mode == 338 FSYNC_MODE_STRICT && !atomic) 339 atomic = true; 340 } 341 go_write: 342 /* 343 * Both of fdatasync() and fsync() are able to be recovered from 344 * sudden-power-off. 345 */ 346 f2fs_down_read(&F2FS_I(inode)->i_sem); 347 cp_reason = need_do_checkpoint(inode); 348 f2fs_up_read(&F2FS_I(inode)->i_sem); 349 350 if (cp_reason) { 351 /* all the dirty node pages should be flushed for POR */ 352 ret = f2fs_sync_fs(inode->i_sb, 1); 353 354 /* 355 * We've secured consistency through sync_fs. Following pino 356 * will be used only for fsynced inodes after checkpoint. 357 */ 358 try_to_fix_pino(inode); 359 clear_inode_flag(inode, FI_APPEND_WRITE); 360 clear_inode_flag(inode, FI_UPDATE_WRITE); 361 goto out; 362 } 363 sync_nodes: 364 atomic_inc(&sbi->wb_sync_req[NODE]); 365 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 366 atomic_dec(&sbi->wb_sync_req[NODE]); 367 if (ret) 368 goto out; 369 370 /* if cp_error was enabled, we should avoid infinite loop */ 371 if (unlikely(f2fs_cp_error(sbi))) { 372 ret = -EIO; 373 goto out; 374 } 375 376 if (f2fs_need_inode_block_update(sbi, ino)) { 377 f2fs_mark_inode_dirty_sync(inode, true); 378 f2fs_write_inode(inode, NULL); 379 goto sync_nodes; 380 } 381 382 /* 383 * If it's atomic_write, it's just fine to keep write ordering. So 384 * here we don't need to wait for node write completion, since we use 385 * node chain which serializes node blocks. If one of node writes are 386 * reordered, we can see simply broken chain, resulting in stopping 387 * roll-forward recovery. It means we'll recover all or none node blocks 388 * given fsync mark. 389 */ 390 if (!atomic) { 391 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 392 if (ret) 393 goto out; 394 } 395 396 /* once recovery info is written, don't need to tack this */ 397 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 398 clear_inode_flag(inode, FI_APPEND_WRITE); 399 flush_out: 400 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 401 ret = f2fs_issue_flush(sbi, inode->i_ino); 402 if (!ret) { 403 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 404 clear_inode_flag(inode, FI_UPDATE_WRITE); 405 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 406 } 407 f2fs_update_time(sbi, REQ_TIME); 408 out: 409 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 410 return ret; 411 } 412 413 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 414 { 415 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 416 return -EIO; 417 return f2fs_do_sync_file(file, start, end, datasync, false); 418 } 419 420 static bool __found_offset(struct address_space *mapping, 421 struct dnode_of_data *dn, pgoff_t index, int whence) 422 { 423 block_t blkaddr = f2fs_data_blkaddr(dn); 424 struct inode *inode = mapping->host; 425 bool compressed_cluster = false; 426 427 if (f2fs_compressed_file(inode)) { 428 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio, 429 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size)); 430 431 compressed_cluster = first_blkaddr == COMPRESS_ADDR; 432 } 433 434 switch (whence) { 435 case SEEK_DATA: 436 if (__is_valid_data_blkaddr(blkaddr)) 437 return true; 438 if (blkaddr == NEW_ADDR && 439 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 440 return true; 441 if (compressed_cluster) 442 return true; 443 break; 444 case SEEK_HOLE: 445 if (compressed_cluster) 446 return false; 447 if (blkaddr == NULL_ADDR) 448 return true; 449 break; 450 } 451 return false; 452 } 453 454 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 455 { 456 struct inode *inode = file->f_mapping->host; 457 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 458 struct dnode_of_data dn; 459 pgoff_t pgofs, end_offset; 460 loff_t data_ofs = offset; 461 loff_t isize; 462 int err = 0; 463 464 inode_lock_shared(inode); 465 466 isize = i_size_read(inode); 467 if (offset >= isize) 468 goto fail; 469 470 /* handle inline data case */ 471 if (f2fs_has_inline_data(inode)) { 472 if (whence == SEEK_HOLE) { 473 data_ofs = isize; 474 goto found; 475 } else if (whence == SEEK_DATA) { 476 data_ofs = offset; 477 goto found; 478 } 479 } 480 481 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 482 483 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 484 set_new_dnode(&dn, inode, NULL, NULL, 0); 485 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 486 if (err && err != -ENOENT) { 487 goto fail; 488 } else if (err == -ENOENT) { 489 /* direct node does not exists */ 490 if (whence == SEEK_DATA) { 491 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 492 continue; 493 } else { 494 goto found; 495 } 496 } 497 498 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 499 500 /* find data/hole in dnode block */ 501 for (; dn.ofs_in_node < end_offset; 502 dn.ofs_in_node++, pgofs++, 503 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 504 block_t blkaddr; 505 506 blkaddr = f2fs_data_blkaddr(&dn); 507 508 if (__is_valid_data_blkaddr(blkaddr) && 509 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 510 blkaddr, DATA_GENERIC_ENHANCE)) { 511 f2fs_put_dnode(&dn); 512 goto fail; 513 } 514 515 if (__found_offset(file->f_mapping, &dn, 516 pgofs, whence)) { 517 f2fs_put_dnode(&dn); 518 goto found; 519 } 520 } 521 f2fs_put_dnode(&dn); 522 } 523 524 if (whence == SEEK_DATA) 525 goto fail; 526 found: 527 if (whence == SEEK_HOLE && data_ofs > isize) 528 data_ofs = isize; 529 inode_unlock_shared(inode); 530 return vfs_setpos(file, data_ofs, maxbytes); 531 fail: 532 inode_unlock_shared(inode); 533 return -ENXIO; 534 } 535 536 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 537 { 538 struct inode *inode = file->f_mapping->host; 539 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 540 541 switch (whence) { 542 case SEEK_SET: 543 case SEEK_CUR: 544 case SEEK_END: 545 return generic_file_llseek_size(file, offset, whence, 546 maxbytes, i_size_read(inode)); 547 case SEEK_DATA: 548 case SEEK_HOLE: 549 if (offset < 0) 550 return -ENXIO; 551 return f2fs_seek_block(file, offset, whence); 552 } 553 554 return -EINVAL; 555 } 556 557 static int f2fs_file_mmap_prepare(struct vm_area_desc *desc) 558 { 559 struct file *file = desc->file; 560 struct inode *inode = file_inode(file); 561 562 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 563 return -EIO; 564 565 if (!f2fs_is_compress_backend_ready(inode)) 566 return -EOPNOTSUPP; 567 568 file_accessed(file); 569 desc->vm_ops = &f2fs_file_vm_ops; 570 571 f2fs_down_read(&F2FS_I(inode)->i_sem); 572 set_inode_flag(inode, FI_MMAP_FILE); 573 f2fs_up_read(&F2FS_I(inode)->i_sem); 574 575 return 0; 576 } 577 578 static int finish_preallocate_blocks(struct inode *inode) 579 { 580 int ret = 0; 581 bool opened; 582 583 f2fs_down_read(&F2FS_I(inode)->i_sem); 584 opened = is_inode_flag_set(inode, FI_OPENED_FILE); 585 f2fs_up_read(&F2FS_I(inode)->i_sem); 586 if (opened) 587 return 0; 588 589 inode_lock(inode); 590 if (is_inode_flag_set(inode, FI_OPENED_FILE)) 591 goto out_unlock; 592 593 if (!file_should_truncate(inode)) 594 goto out_update; 595 596 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 597 filemap_invalidate_lock(inode->i_mapping); 598 599 truncate_setsize(inode, i_size_read(inode)); 600 ret = f2fs_truncate(inode); 601 602 filemap_invalidate_unlock(inode->i_mapping); 603 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 604 if (ret) 605 goto out_unlock; 606 607 file_dont_truncate(inode); 608 out_update: 609 f2fs_down_write(&F2FS_I(inode)->i_sem); 610 set_inode_flag(inode, FI_OPENED_FILE); 611 f2fs_up_write(&F2FS_I(inode)->i_sem); 612 out_unlock: 613 inode_unlock(inode); 614 return ret; 615 } 616 617 static int f2fs_file_open(struct inode *inode, struct file *filp) 618 { 619 int err = fscrypt_file_open(inode, filp); 620 621 if (err) 622 return err; 623 624 if (!f2fs_is_compress_backend_ready(inode)) 625 return -EOPNOTSUPP; 626 627 err = fsverity_file_open(inode, filp); 628 if (err) 629 return err; 630 631 filp->f_mode |= FMODE_NOWAIT; 632 filp->f_mode |= FMODE_CAN_ODIRECT; 633 634 err = dquot_file_open(inode, filp); 635 if (err) 636 return err; 637 638 err = finish_preallocate_blocks(inode); 639 if (!err) 640 atomic_inc(&F2FS_I(inode)->open_count); 641 return err; 642 } 643 644 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 645 { 646 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 647 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 648 __le32 *addr; 649 bool compressed_cluster = false; 650 int cluster_index = 0, valid_blocks = 0; 651 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 652 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 653 block_t blkstart; 654 int blklen = 0; 655 656 addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs; 657 blkstart = le32_to_cpu(*addr); 658 659 /* Assumption: truncation starts with cluster */ 660 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 661 block_t blkaddr = le32_to_cpu(*addr); 662 663 if (f2fs_compressed_file(dn->inode) && 664 !(cluster_index & (cluster_size - 1))) { 665 if (compressed_cluster) 666 f2fs_i_compr_blocks_update(dn->inode, 667 valid_blocks, false); 668 compressed_cluster = (blkaddr == COMPRESS_ADDR); 669 valid_blocks = 0; 670 } 671 672 if (blkaddr == NULL_ADDR) 673 goto next; 674 675 f2fs_set_data_blkaddr(dn, NULL_ADDR); 676 677 if (__is_valid_data_blkaddr(blkaddr)) { 678 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE)) 679 goto next; 680 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr, 681 DATA_GENERIC_ENHANCE)) 682 goto next; 683 if (compressed_cluster) 684 valid_blocks++; 685 } 686 687 if (blkstart + blklen == blkaddr) { 688 blklen++; 689 } else { 690 f2fs_invalidate_blocks(sbi, blkstart, blklen); 691 blkstart = blkaddr; 692 blklen = 1; 693 } 694 695 if (!released || blkaddr != COMPRESS_ADDR) 696 nr_free++; 697 698 continue; 699 700 next: 701 if (blklen) 702 f2fs_invalidate_blocks(sbi, blkstart, blklen); 703 704 blkstart = le32_to_cpu(*(addr + 1)); 705 blklen = 0; 706 } 707 708 if (blklen) 709 f2fs_invalidate_blocks(sbi, blkstart, blklen); 710 711 if (compressed_cluster) 712 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 713 714 if (nr_free) { 715 pgoff_t fofs; 716 /* 717 * once we invalidate valid blkaddr in range [ofs, ofs + count], 718 * we will invalidate all blkaddr in the whole range. 719 */ 720 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_folio), 721 dn->inode) + ofs; 722 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 723 f2fs_update_age_extent_cache_range(dn, fofs, len); 724 dec_valid_block_count(sbi, dn->inode, nr_free); 725 } 726 dn->ofs_in_node = ofs; 727 728 f2fs_update_time(sbi, REQ_TIME); 729 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 730 dn->ofs_in_node, nr_free); 731 } 732 733 static int truncate_partial_data_page(struct inode *inode, u64 from, 734 bool cache_only) 735 { 736 loff_t offset = from & (PAGE_SIZE - 1); 737 pgoff_t index = from >> PAGE_SHIFT; 738 struct address_space *mapping = inode->i_mapping; 739 struct folio *folio; 740 741 if (!offset && !cache_only) 742 return 0; 743 744 if (cache_only) { 745 folio = filemap_lock_folio(mapping, index); 746 if (IS_ERR(folio)) 747 return 0; 748 if (folio_test_uptodate(folio)) 749 goto truncate_out; 750 f2fs_folio_put(folio, true); 751 return 0; 752 } 753 754 folio = f2fs_get_lock_data_folio(inode, index, true); 755 if (IS_ERR(folio)) 756 return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio); 757 truncate_out: 758 f2fs_folio_wait_writeback(folio, DATA, true, true); 759 folio_zero_segment(folio, offset, folio_size(folio)); 760 761 /* An encrypted inode should have a key and truncate the last page. */ 762 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 763 if (!cache_only) 764 folio_mark_dirty(folio); 765 f2fs_folio_put(folio, true); 766 return 0; 767 } 768 769 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 770 { 771 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 772 struct dnode_of_data dn; 773 pgoff_t free_from; 774 int count = 0, err = 0; 775 struct folio *ifolio; 776 bool truncate_page = false; 777 778 trace_f2fs_truncate_blocks_enter(inode, from); 779 780 if (IS_DEVICE_ALIASING(inode) && from) { 781 err = -EINVAL; 782 goto out_err; 783 } 784 785 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 786 787 if (free_from >= max_file_blocks(inode)) 788 goto free_partial; 789 790 if (lock) 791 f2fs_lock_op(sbi); 792 793 ifolio = f2fs_get_inode_folio(sbi, inode->i_ino); 794 if (IS_ERR(ifolio)) { 795 err = PTR_ERR(ifolio); 796 goto out; 797 } 798 799 if (IS_DEVICE_ALIASING(inode)) { 800 struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ]; 801 struct extent_info ei = et->largest; 802 803 f2fs_invalidate_blocks(sbi, ei.blk, ei.len); 804 805 dec_valid_block_count(sbi, inode, ei.len); 806 f2fs_update_time(sbi, REQ_TIME); 807 808 f2fs_folio_put(ifolio, true); 809 goto out; 810 } 811 812 if (f2fs_has_inline_data(inode)) { 813 f2fs_truncate_inline_inode(inode, ifolio, from); 814 f2fs_folio_put(ifolio, true); 815 truncate_page = true; 816 goto out; 817 } 818 819 set_new_dnode(&dn, inode, ifolio, NULL, 0); 820 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 821 if (err) { 822 if (err == -ENOENT) 823 goto free_next; 824 goto out; 825 } 826 827 count = ADDRS_PER_PAGE(dn.node_folio, inode); 828 829 count -= dn.ofs_in_node; 830 f2fs_bug_on(sbi, count < 0); 831 832 if (dn.ofs_in_node || IS_INODE(dn.node_folio)) { 833 f2fs_truncate_data_blocks_range(&dn, count); 834 free_from += count; 835 } 836 837 f2fs_put_dnode(&dn); 838 free_next: 839 err = f2fs_truncate_inode_blocks(inode, free_from); 840 out: 841 if (lock) 842 f2fs_unlock_op(sbi); 843 free_partial: 844 /* lastly zero out the first data page */ 845 if (!err) 846 err = truncate_partial_data_page(inode, from, truncate_page); 847 out_err: 848 trace_f2fs_truncate_blocks_exit(inode, err); 849 return err; 850 } 851 852 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 853 { 854 u64 free_from = from; 855 int err; 856 857 #ifdef CONFIG_F2FS_FS_COMPRESSION 858 /* 859 * for compressed file, only support cluster size 860 * aligned truncation. 861 */ 862 if (f2fs_compressed_file(inode)) 863 free_from = round_up(from, 864 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 865 #endif 866 867 err = f2fs_do_truncate_blocks(inode, free_from, lock); 868 if (err) 869 return err; 870 871 #ifdef CONFIG_F2FS_FS_COMPRESSION 872 /* 873 * For compressed file, after release compress blocks, don't allow write 874 * direct, but we should allow write direct after truncate to zero. 875 */ 876 if (f2fs_compressed_file(inode) && !free_from 877 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 878 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 879 880 if (from != free_from) { 881 err = f2fs_truncate_partial_cluster(inode, from, lock); 882 if (err) 883 return err; 884 } 885 #endif 886 887 return 0; 888 } 889 890 int f2fs_truncate(struct inode *inode) 891 { 892 int err; 893 894 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 895 return -EIO; 896 897 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 898 S_ISLNK(inode->i_mode))) 899 return 0; 900 901 trace_f2fs_truncate(inode); 902 903 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 904 return -EIO; 905 906 err = f2fs_dquot_initialize(inode); 907 if (err) 908 return err; 909 910 /* we should check inline_data size */ 911 if (!f2fs_may_inline_data(inode)) { 912 err = f2fs_convert_inline_inode(inode); 913 if (err) { 914 /* 915 * Always truncate page #0 to avoid page cache 916 * leak in evict() path. 917 */ 918 truncate_inode_pages_range(inode->i_mapping, 919 F2FS_BLK_TO_BYTES(0), 920 F2FS_BLK_END_BYTES(0)); 921 return err; 922 } 923 } 924 925 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 926 if (err) 927 return err; 928 929 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 930 f2fs_mark_inode_dirty_sync(inode, false); 931 return 0; 932 } 933 934 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 935 { 936 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 937 938 if (!fscrypt_dio_supported(inode)) 939 return true; 940 if (fsverity_active(inode)) 941 return true; 942 if (f2fs_compressed_file(inode)) 943 return true; 944 /* 945 * only force direct read to use buffered IO, for direct write, 946 * it expects inline data conversion before committing IO. 947 */ 948 if (f2fs_has_inline_data(inode) && rw == READ) 949 return true; 950 951 /* disallow direct IO if any of devices has unaligned blksize */ 952 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 953 return true; 954 /* 955 * for blkzoned device, fallback direct IO to buffered IO, so 956 * all IOs can be serialized by log-structured write. 957 */ 958 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) && 959 !f2fs_is_pinned_file(inode)) 960 return true; 961 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 962 return true; 963 964 return false; 965 } 966 967 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 968 struct kstat *stat, u32 request_mask, unsigned int query_flags) 969 { 970 struct inode *inode = d_inode(path->dentry); 971 struct f2fs_inode_info *fi = F2FS_I(inode); 972 struct f2fs_inode *ri = NULL; 973 unsigned int flags; 974 975 if (f2fs_has_extra_attr(inode) && 976 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 977 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 978 stat->result_mask |= STATX_BTIME; 979 stat->btime.tv_sec = fi->i_crtime.tv_sec; 980 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 981 } 982 983 /* 984 * Return the DIO alignment restrictions if requested. We only return 985 * this information when requested, since on encrypted files it might 986 * take a fair bit of work to get if the file wasn't opened recently. 987 * 988 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 989 * cannot represent that, so in that case we report no DIO support. 990 */ 991 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 992 unsigned int bsize = i_blocksize(inode); 993 994 stat->result_mask |= STATX_DIOALIGN; 995 if (!f2fs_force_buffered_io(inode, WRITE)) { 996 stat->dio_mem_align = bsize; 997 stat->dio_offset_align = bsize; 998 } 999 } 1000 1001 flags = fi->i_flags; 1002 if (flags & F2FS_COMPR_FL) 1003 stat->attributes |= STATX_ATTR_COMPRESSED; 1004 if (flags & F2FS_APPEND_FL) 1005 stat->attributes |= STATX_ATTR_APPEND; 1006 if (IS_ENCRYPTED(inode)) 1007 stat->attributes |= STATX_ATTR_ENCRYPTED; 1008 if (flags & F2FS_IMMUTABLE_FL) 1009 stat->attributes |= STATX_ATTR_IMMUTABLE; 1010 if (flags & F2FS_NODUMP_FL) 1011 stat->attributes |= STATX_ATTR_NODUMP; 1012 if (IS_VERITY(inode)) 1013 stat->attributes |= STATX_ATTR_VERITY; 1014 1015 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 1016 STATX_ATTR_APPEND | 1017 STATX_ATTR_ENCRYPTED | 1018 STATX_ATTR_IMMUTABLE | 1019 STATX_ATTR_NODUMP | 1020 STATX_ATTR_VERITY); 1021 1022 generic_fillattr(idmap, request_mask, inode, stat); 1023 1024 /* we need to show initial sectors used for inline_data/dentries */ 1025 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 1026 f2fs_has_inline_dentry(inode)) 1027 stat->blocks += (stat->size + 511) >> 9; 1028 1029 return 0; 1030 } 1031 1032 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1033 static void __setattr_copy(struct mnt_idmap *idmap, 1034 struct inode *inode, const struct iattr *attr) 1035 { 1036 unsigned int ia_valid = attr->ia_valid; 1037 1038 i_uid_update(idmap, attr, inode); 1039 i_gid_update(idmap, attr, inode); 1040 if (ia_valid & ATTR_ATIME) 1041 inode_set_atime_to_ts(inode, attr->ia_atime); 1042 if (ia_valid & ATTR_MTIME) 1043 inode_set_mtime_to_ts(inode, attr->ia_mtime); 1044 if (ia_valid & ATTR_CTIME) 1045 inode_set_ctime_to_ts(inode, attr->ia_ctime); 1046 if (ia_valid & ATTR_MODE) { 1047 umode_t mode = attr->ia_mode; 1048 1049 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode))) 1050 mode &= ~S_ISGID; 1051 set_acl_inode(inode, mode); 1052 } 1053 } 1054 #else 1055 #define __setattr_copy setattr_copy 1056 #endif 1057 1058 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 1059 struct iattr *attr) 1060 { 1061 struct inode *inode = d_inode(dentry); 1062 struct f2fs_inode_info *fi = F2FS_I(inode); 1063 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1064 int err; 1065 1066 if (unlikely(f2fs_cp_error(sbi))) 1067 return -EIO; 1068 1069 err = setattr_prepare(idmap, dentry, attr); 1070 if (err) 1071 return err; 1072 1073 err = fscrypt_prepare_setattr(dentry, attr); 1074 if (err) 1075 return err; 1076 1077 err = fsverity_prepare_setattr(dentry, attr); 1078 if (err) 1079 return err; 1080 1081 if (unlikely(IS_IMMUTABLE(inode))) 1082 return -EPERM; 1083 1084 if (unlikely(IS_APPEND(inode) && 1085 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 1086 ATTR_GID | ATTR_TIMES_SET)))) 1087 return -EPERM; 1088 1089 if ((attr->ia_valid & ATTR_SIZE)) { 1090 if (!f2fs_is_compress_backend_ready(inode) || 1091 IS_DEVICE_ALIASING(inode)) 1092 return -EOPNOTSUPP; 1093 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 1094 !IS_ALIGNED(attr->ia_size, 1095 F2FS_BLK_TO_BYTES(fi->i_cluster_size))) 1096 return -EINVAL; 1097 /* 1098 * To prevent scattered pin block generation, we don't allow 1099 * smaller/equal size unaligned truncation for pinned file. 1100 * We only support overwrite IO to pinned file, so don't 1101 * care about larger size truncation. 1102 */ 1103 if (f2fs_is_pinned_file(inode) && 1104 attr->ia_size <= i_size_read(inode) && 1105 !IS_ALIGNED(attr->ia_size, 1106 F2FS_BLK_TO_BYTES(CAP_BLKS_PER_SEC(sbi)))) 1107 return -EINVAL; 1108 } 1109 1110 if (is_quota_modification(idmap, inode, attr)) { 1111 err = f2fs_dquot_initialize(inode); 1112 if (err) 1113 return err; 1114 } 1115 if (i_uid_needs_update(idmap, attr, inode) || 1116 i_gid_needs_update(idmap, attr, inode)) { 1117 f2fs_lock_op(sbi); 1118 err = dquot_transfer(idmap, inode, attr); 1119 if (err) { 1120 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 1121 f2fs_unlock_op(sbi); 1122 return err; 1123 } 1124 /* 1125 * update uid/gid under lock_op(), so that dquot and inode can 1126 * be updated atomically. 1127 */ 1128 i_uid_update(idmap, attr, inode); 1129 i_gid_update(idmap, attr, inode); 1130 f2fs_mark_inode_dirty_sync(inode, true); 1131 f2fs_unlock_op(sbi); 1132 } 1133 1134 if (attr->ia_valid & ATTR_SIZE) { 1135 loff_t old_size = i_size_read(inode); 1136 1137 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1138 /* 1139 * should convert inline inode before i_size_write to 1140 * keep smaller than inline_data size with inline flag. 1141 */ 1142 err = f2fs_convert_inline_inode(inode); 1143 if (err) 1144 return err; 1145 } 1146 1147 /* 1148 * wait for inflight dio, blocks should be removed after 1149 * IO completion. 1150 */ 1151 if (attr->ia_size < old_size) 1152 inode_dio_wait(inode); 1153 1154 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 1155 filemap_invalidate_lock(inode->i_mapping); 1156 1157 if (attr->ia_size > old_size) 1158 f2fs_zero_post_eof_page(inode, attr->ia_size, false); 1159 truncate_setsize(inode, attr->ia_size); 1160 1161 if (attr->ia_size <= old_size) 1162 err = f2fs_truncate(inode); 1163 /* 1164 * do not trim all blocks after i_size if target size is 1165 * larger than i_size. 1166 */ 1167 filemap_invalidate_unlock(inode->i_mapping); 1168 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 1169 if (err) 1170 return err; 1171 1172 spin_lock(&fi->i_size_lock); 1173 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1174 fi->last_disk_size = i_size_read(inode); 1175 spin_unlock(&fi->i_size_lock); 1176 } 1177 1178 __setattr_copy(idmap, inode, attr); 1179 1180 if (attr->ia_valid & ATTR_MODE) { 1181 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1182 1183 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1184 if (!err) 1185 inode->i_mode = fi->i_acl_mode; 1186 clear_inode_flag(inode, FI_ACL_MODE); 1187 } 1188 } 1189 1190 /* file size may changed here */ 1191 f2fs_mark_inode_dirty_sync(inode, true); 1192 1193 /* inode change will produce dirty node pages flushed by checkpoint */ 1194 f2fs_balance_fs(sbi, true); 1195 1196 return err; 1197 } 1198 1199 const struct inode_operations f2fs_file_inode_operations = { 1200 .getattr = f2fs_getattr, 1201 .setattr = f2fs_setattr, 1202 .get_inode_acl = f2fs_get_acl, 1203 .set_acl = f2fs_set_acl, 1204 .listxattr = f2fs_listxattr, 1205 .fiemap = f2fs_fiemap, 1206 .fileattr_get = f2fs_fileattr_get, 1207 .fileattr_set = f2fs_fileattr_set, 1208 }; 1209 1210 static int fill_zero(struct inode *inode, pgoff_t index, 1211 loff_t start, loff_t len) 1212 { 1213 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1214 struct folio *folio; 1215 1216 if (!len) 1217 return 0; 1218 1219 f2fs_balance_fs(sbi, true); 1220 1221 f2fs_lock_op(sbi); 1222 folio = f2fs_get_new_data_folio(inode, NULL, index, false); 1223 f2fs_unlock_op(sbi); 1224 1225 if (IS_ERR(folio)) 1226 return PTR_ERR(folio); 1227 1228 f2fs_folio_wait_writeback(folio, DATA, true, true); 1229 folio_zero_range(folio, start, len); 1230 folio_mark_dirty(folio); 1231 f2fs_folio_put(folio, true); 1232 return 0; 1233 } 1234 1235 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1236 { 1237 int err; 1238 1239 while (pg_start < pg_end) { 1240 struct dnode_of_data dn; 1241 pgoff_t end_offset, count; 1242 1243 set_new_dnode(&dn, inode, NULL, NULL, 0); 1244 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1245 if (err) { 1246 if (err == -ENOENT) { 1247 pg_start = f2fs_get_next_page_offset(&dn, 1248 pg_start); 1249 continue; 1250 } 1251 return err; 1252 } 1253 1254 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 1255 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1256 1257 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1258 1259 f2fs_truncate_data_blocks_range(&dn, count); 1260 f2fs_put_dnode(&dn); 1261 1262 pg_start += count; 1263 } 1264 return 0; 1265 } 1266 1267 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1268 { 1269 pgoff_t pg_start, pg_end; 1270 loff_t off_start, off_end; 1271 int ret; 1272 1273 ret = f2fs_convert_inline_inode(inode); 1274 if (ret) 1275 return ret; 1276 1277 f2fs_zero_post_eof_page(inode, offset + len, true); 1278 1279 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1280 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1281 1282 off_start = offset & (PAGE_SIZE - 1); 1283 off_end = (offset + len) & (PAGE_SIZE - 1); 1284 1285 if (pg_start == pg_end) { 1286 ret = fill_zero(inode, pg_start, off_start, 1287 off_end - off_start); 1288 if (ret) 1289 return ret; 1290 } else { 1291 if (off_start) { 1292 ret = fill_zero(inode, pg_start++, off_start, 1293 PAGE_SIZE - off_start); 1294 if (ret) 1295 return ret; 1296 } 1297 if (off_end) { 1298 ret = fill_zero(inode, pg_end, 0, off_end); 1299 if (ret) 1300 return ret; 1301 } 1302 1303 if (pg_start < pg_end) { 1304 loff_t blk_start, blk_end; 1305 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1306 1307 f2fs_balance_fs(sbi, true); 1308 1309 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1310 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1311 1312 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1313 filemap_invalidate_lock(inode->i_mapping); 1314 1315 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1316 1317 f2fs_lock_op(sbi); 1318 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1319 f2fs_unlock_op(sbi); 1320 1321 filemap_invalidate_unlock(inode->i_mapping); 1322 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1323 } 1324 } 1325 1326 return ret; 1327 } 1328 1329 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1330 int *do_replace, pgoff_t off, pgoff_t len) 1331 { 1332 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1333 struct dnode_of_data dn; 1334 int ret, done, i; 1335 1336 next_dnode: 1337 set_new_dnode(&dn, inode, NULL, NULL, 0); 1338 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1339 if (ret && ret != -ENOENT) { 1340 return ret; 1341 } else if (ret == -ENOENT) { 1342 if (dn.max_level == 0) 1343 return -ENOENT; 1344 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1345 dn.ofs_in_node, len); 1346 blkaddr += done; 1347 do_replace += done; 1348 goto next; 1349 } 1350 1351 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_folio, inode) - 1352 dn.ofs_in_node, len); 1353 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1354 *blkaddr = f2fs_data_blkaddr(&dn); 1355 1356 if (__is_valid_data_blkaddr(*blkaddr) && 1357 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1358 DATA_GENERIC_ENHANCE)) { 1359 f2fs_put_dnode(&dn); 1360 return -EFSCORRUPTED; 1361 } 1362 1363 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1364 1365 if (f2fs_lfs_mode(sbi)) { 1366 f2fs_put_dnode(&dn); 1367 return -EOPNOTSUPP; 1368 } 1369 1370 /* do not invalidate this block address */ 1371 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1372 *do_replace = 1; 1373 } 1374 } 1375 f2fs_put_dnode(&dn); 1376 next: 1377 len -= done; 1378 off += done; 1379 if (len) 1380 goto next_dnode; 1381 return 0; 1382 } 1383 1384 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1385 int *do_replace, pgoff_t off, int len) 1386 { 1387 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1388 struct dnode_of_data dn; 1389 int ret, i; 1390 1391 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1392 if (*do_replace == 0) 1393 continue; 1394 1395 set_new_dnode(&dn, inode, NULL, NULL, 0); 1396 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1397 if (ret) { 1398 dec_valid_block_count(sbi, inode, 1); 1399 f2fs_invalidate_blocks(sbi, *blkaddr, 1); 1400 } else { 1401 f2fs_update_data_blkaddr(&dn, *blkaddr); 1402 } 1403 f2fs_put_dnode(&dn); 1404 } 1405 return 0; 1406 } 1407 1408 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1409 block_t *blkaddr, int *do_replace, 1410 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1411 { 1412 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1413 pgoff_t i = 0; 1414 int ret; 1415 1416 while (i < len) { 1417 if (blkaddr[i] == NULL_ADDR && !full) { 1418 i++; 1419 continue; 1420 } 1421 1422 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1423 struct dnode_of_data dn; 1424 struct node_info ni; 1425 size_t new_size; 1426 pgoff_t ilen; 1427 1428 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1429 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1430 if (ret) 1431 return ret; 1432 1433 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1434 if (ret) { 1435 f2fs_put_dnode(&dn); 1436 return ret; 1437 } 1438 1439 ilen = min((pgoff_t) 1440 ADDRS_PER_PAGE(dn.node_folio, dst_inode) - 1441 dn.ofs_in_node, len - i); 1442 do { 1443 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1444 f2fs_truncate_data_blocks_range(&dn, 1); 1445 1446 if (do_replace[i]) { 1447 f2fs_i_blocks_write(src_inode, 1448 1, false, false); 1449 f2fs_i_blocks_write(dst_inode, 1450 1, true, false); 1451 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1452 blkaddr[i], ni.version, true, false); 1453 1454 do_replace[i] = 0; 1455 } 1456 dn.ofs_in_node++; 1457 i++; 1458 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1459 if (dst_inode->i_size < new_size) 1460 f2fs_i_size_write(dst_inode, new_size); 1461 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1462 1463 f2fs_put_dnode(&dn); 1464 } else { 1465 struct folio *fsrc, *fdst; 1466 1467 fsrc = f2fs_get_lock_data_folio(src_inode, 1468 src + i, true); 1469 if (IS_ERR(fsrc)) 1470 return PTR_ERR(fsrc); 1471 fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i, 1472 true); 1473 if (IS_ERR(fdst)) { 1474 f2fs_folio_put(fsrc, true); 1475 return PTR_ERR(fdst); 1476 } 1477 1478 f2fs_folio_wait_writeback(fdst, DATA, true, true); 1479 1480 memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE); 1481 folio_mark_dirty(fdst); 1482 folio_set_f2fs_gcing(fdst); 1483 f2fs_folio_put(fdst, true); 1484 f2fs_folio_put(fsrc, true); 1485 1486 ret = f2fs_truncate_hole(src_inode, 1487 src + i, src + i + 1); 1488 if (ret) 1489 return ret; 1490 i++; 1491 } 1492 } 1493 return 0; 1494 } 1495 1496 static int __exchange_data_block(struct inode *src_inode, 1497 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1498 pgoff_t len, bool full) 1499 { 1500 block_t *src_blkaddr; 1501 int *do_replace; 1502 pgoff_t olen; 1503 int ret; 1504 1505 while (len) { 1506 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1507 1508 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1509 array_size(olen, sizeof(block_t)), 1510 GFP_NOFS); 1511 if (!src_blkaddr) 1512 return -ENOMEM; 1513 1514 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1515 array_size(olen, sizeof(int)), 1516 GFP_NOFS); 1517 if (!do_replace) { 1518 kvfree(src_blkaddr); 1519 return -ENOMEM; 1520 } 1521 1522 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1523 do_replace, src, olen); 1524 if (ret) 1525 goto roll_back; 1526 1527 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1528 do_replace, src, dst, olen, full); 1529 if (ret) 1530 goto roll_back; 1531 1532 src += olen; 1533 dst += olen; 1534 len -= olen; 1535 1536 kvfree(src_blkaddr); 1537 kvfree(do_replace); 1538 } 1539 return 0; 1540 1541 roll_back: 1542 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1543 kvfree(src_blkaddr); 1544 kvfree(do_replace); 1545 return ret; 1546 } 1547 1548 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1549 { 1550 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1551 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1552 pgoff_t start = offset >> PAGE_SHIFT; 1553 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1554 int ret; 1555 1556 f2fs_balance_fs(sbi, true); 1557 1558 /* avoid gc operation during block exchange */ 1559 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1560 filemap_invalidate_lock(inode->i_mapping); 1561 1562 f2fs_zero_post_eof_page(inode, offset + len, false); 1563 1564 f2fs_lock_op(sbi); 1565 f2fs_drop_extent_tree(inode); 1566 truncate_pagecache(inode, offset); 1567 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1568 f2fs_unlock_op(sbi); 1569 1570 filemap_invalidate_unlock(inode->i_mapping); 1571 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1572 return ret; 1573 } 1574 1575 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1576 { 1577 loff_t new_size; 1578 int ret; 1579 1580 if (offset + len >= i_size_read(inode)) 1581 return -EINVAL; 1582 1583 /* collapse range should be aligned to block size of f2fs. */ 1584 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1585 return -EINVAL; 1586 1587 ret = f2fs_convert_inline_inode(inode); 1588 if (ret) 1589 return ret; 1590 1591 /* write out all dirty pages from offset */ 1592 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1593 if (ret) 1594 return ret; 1595 1596 ret = f2fs_do_collapse(inode, offset, len); 1597 if (ret) 1598 return ret; 1599 1600 /* write out all moved pages, if possible */ 1601 filemap_invalidate_lock(inode->i_mapping); 1602 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1603 truncate_pagecache(inode, offset); 1604 1605 new_size = i_size_read(inode) - len; 1606 ret = f2fs_truncate_blocks(inode, new_size, true); 1607 filemap_invalidate_unlock(inode->i_mapping); 1608 if (!ret) 1609 f2fs_i_size_write(inode, new_size); 1610 return ret; 1611 } 1612 1613 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1614 pgoff_t end) 1615 { 1616 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1617 pgoff_t index = start; 1618 unsigned int ofs_in_node = dn->ofs_in_node; 1619 blkcnt_t count = 0; 1620 int ret; 1621 1622 for (; index < end; index++, dn->ofs_in_node++) { 1623 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1624 count++; 1625 } 1626 1627 dn->ofs_in_node = ofs_in_node; 1628 ret = f2fs_reserve_new_blocks(dn, count); 1629 if (ret) 1630 return ret; 1631 1632 dn->ofs_in_node = ofs_in_node; 1633 for (index = start; index < end; index++, dn->ofs_in_node++) { 1634 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1635 /* 1636 * f2fs_reserve_new_blocks will not guarantee entire block 1637 * allocation. 1638 */ 1639 if (dn->data_blkaddr == NULL_ADDR) { 1640 ret = -ENOSPC; 1641 break; 1642 } 1643 1644 if (dn->data_blkaddr == NEW_ADDR) 1645 continue; 1646 1647 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1648 DATA_GENERIC_ENHANCE)) { 1649 ret = -EFSCORRUPTED; 1650 break; 1651 } 1652 1653 f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1); 1654 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1655 } 1656 1657 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1658 f2fs_update_age_extent_cache_range(dn, start, index - start); 1659 1660 return ret; 1661 } 1662 1663 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1664 int mode) 1665 { 1666 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1667 struct address_space *mapping = inode->i_mapping; 1668 pgoff_t index, pg_start, pg_end; 1669 loff_t new_size = i_size_read(inode); 1670 loff_t off_start, off_end; 1671 int ret = 0; 1672 1673 ret = inode_newsize_ok(inode, (len + offset)); 1674 if (ret) 1675 return ret; 1676 1677 ret = f2fs_convert_inline_inode(inode); 1678 if (ret) 1679 return ret; 1680 1681 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1682 if (ret) 1683 return ret; 1684 1685 f2fs_zero_post_eof_page(inode, offset + len, true); 1686 1687 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1688 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1689 1690 off_start = offset & (PAGE_SIZE - 1); 1691 off_end = (offset + len) & (PAGE_SIZE - 1); 1692 1693 if (pg_start == pg_end) { 1694 ret = fill_zero(inode, pg_start, off_start, 1695 off_end - off_start); 1696 if (ret) 1697 return ret; 1698 1699 new_size = max_t(loff_t, new_size, offset + len); 1700 } else { 1701 if (off_start) { 1702 ret = fill_zero(inode, pg_start++, off_start, 1703 PAGE_SIZE - off_start); 1704 if (ret) 1705 return ret; 1706 1707 new_size = max_t(loff_t, new_size, 1708 (loff_t)pg_start << PAGE_SHIFT); 1709 } 1710 1711 for (index = pg_start; index < pg_end;) { 1712 struct dnode_of_data dn; 1713 unsigned int end_offset; 1714 pgoff_t end; 1715 1716 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1717 filemap_invalidate_lock(mapping); 1718 1719 truncate_pagecache_range(inode, 1720 (loff_t)index << PAGE_SHIFT, 1721 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1722 1723 f2fs_lock_op(sbi); 1724 1725 set_new_dnode(&dn, inode, NULL, NULL, 0); 1726 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1727 if (ret) { 1728 f2fs_unlock_op(sbi); 1729 filemap_invalidate_unlock(mapping); 1730 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1731 goto out; 1732 } 1733 1734 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 1735 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1736 1737 ret = f2fs_do_zero_range(&dn, index, end); 1738 f2fs_put_dnode(&dn); 1739 1740 f2fs_unlock_op(sbi); 1741 filemap_invalidate_unlock(mapping); 1742 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1743 1744 f2fs_balance_fs(sbi, dn.node_changed); 1745 1746 if (ret) 1747 goto out; 1748 1749 index = end; 1750 new_size = max_t(loff_t, new_size, 1751 (loff_t)index << PAGE_SHIFT); 1752 } 1753 1754 if (off_end) { 1755 ret = fill_zero(inode, pg_end, 0, off_end); 1756 if (ret) 1757 goto out; 1758 1759 new_size = max_t(loff_t, new_size, offset + len); 1760 } 1761 } 1762 1763 out: 1764 if (new_size > i_size_read(inode)) { 1765 if (mode & FALLOC_FL_KEEP_SIZE) 1766 file_set_keep_isize(inode); 1767 else 1768 f2fs_i_size_write(inode, new_size); 1769 } 1770 return ret; 1771 } 1772 1773 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1774 { 1775 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1776 struct address_space *mapping = inode->i_mapping; 1777 pgoff_t nr, pg_start, pg_end, delta, idx; 1778 loff_t new_size; 1779 int ret = 0; 1780 1781 new_size = i_size_read(inode) + len; 1782 ret = inode_newsize_ok(inode, new_size); 1783 if (ret) 1784 return ret; 1785 1786 if (offset >= i_size_read(inode)) 1787 return -EINVAL; 1788 1789 /* insert range should be aligned to block size of f2fs. */ 1790 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1791 return -EINVAL; 1792 1793 ret = f2fs_convert_inline_inode(inode); 1794 if (ret) 1795 return ret; 1796 1797 f2fs_balance_fs(sbi, true); 1798 1799 filemap_invalidate_lock(mapping); 1800 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1801 filemap_invalidate_unlock(mapping); 1802 if (ret) 1803 return ret; 1804 1805 /* write out all dirty pages from offset */ 1806 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1807 if (ret) 1808 return ret; 1809 1810 pg_start = offset >> PAGE_SHIFT; 1811 pg_end = (offset + len) >> PAGE_SHIFT; 1812 delta = pg_end - pg_start; 1813 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1814 1815 /* avoid gc operation during block exchange */ 1816 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1817 filemap_invalidate_lock(mapping); 1818 1819 f2fs_zero_post_eof_page(inode, offset + len, false); 1820 truncate_pagecache(inode, offset); 1821 1822 while (!ret && idx > pg_start) { 1823 nr = idx - pg_start; 1824 if (nr > delta) 1825 nr = delta; 1826 idx -= nr; 1827 1828 f2fs_lock_op(sbi); 1829 f2fs_drop_extent_tree(inode); 1830 1831 ret = __exchange_data_block(inode, inode, idx, 1832 idx + delta, nr, false); 1833 f2fs_unlock_op(sbi); 1834 } 1835 filemap_invalidate_unlock(mapping); 1836 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1837 if (ret) 1838 return ret; 1839 1840 /* write out all moved pages, if possible */ 1841 filemap_invalidate_lock(mapping); 1842 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1843 truncate_pagecache(inode, offset); 1844 filemap_invalidate_unlock(mapping); 1845 1846 if (!ret) 1847 f2fs_i_size_write(inode, new_size); 1848 return ret; 1849 } 1850 1851 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1852 loff_t len, int mode) 1853 { 1854 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1855 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1856 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1857 .m_may_create = true }; 1858 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1859 .init_gc_type = FG_GC, 1860 .should_migrate_blocks = false, 1861 .err_gc_skipped = true, 1862 .nr_free_secs = 0 }; 1863 pgoff_t pg_start, pg_end; 1864 loff_t new_size; 1865 loff_t off_end; 1866 block_t expanded = 0; 1867 int err; 1868 1869 err = inode_newsize_ok(inode, (len + offset)); 1870 if (err) 1871 return err; 1872 1873 err = f2fs_convert_inline_inode(inode); 1874 if (err) 1875 return err; 1876 1877 f2fs_zero_post_eof_page(inode, offset + len, true); 1878 1879 f2fs_balance_fs(sbi, true); 1880 1881 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1882 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1883 off_end = (offset + len) & (PAGE_SIZE - 1); 1884 1885 map.m_lblk = pg_start; 1886 map.m_len = pg_end - pg_start; 1887 if (off_end) 1888 map.m_len++; 1889 1890 if (!map.m_len) 1891 return 0; 1892 1893 if (f2fs_is_pinned_file(inode)) { 1894 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1895 block_t sec_len = roundup(map.m_len, sec_blks); 1896 1897 map.m_len = sec_blks; 1898 next_alloc: 1899 f2fs_down_write(&sbi->pin_sem); 1900 1901 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1902 if (has_not_enough_free_secs(sbi, 0, 0)) { 1903 f2fs_up_write(&sbi->pin_sem); 1904 err = -ENOSPC; 1905 f2fs_warn_ratelimited(sbi, 1906 "ino:%lu, start:%lu, end:%lu, need to trigger GC to " 1907 "reclaim enough free segment when checkpoint is enabled", 1908 inode->i_ino, pg_start, pg_end); 1909 goto out_err; 1910 } 1911 } 1912 1913 if (has_not_enough_free_secs(sbi, 0, 1914 sbi->reserved_pin_section)) { 1915 f2fs_down_write(&sbi->gc_lock); 1916 stat_inc_gc_call_count(sbi, FOREGROUND); 1917 err = f2fs_gc(sbi, &gc_control); 1918 if (err && err != -ENODATA) { 1919 f2fs_up_write(&sbi->pin_sem); 1920 goto out_err; 1921 } 1922 } 1923 1924 err = f2fs_allocate_pinning_section(sbi); 1925 if (err) { 1926 f2fs_up_write(&sbi->pin_sem); 1927 goto out_err; 1928 } 1929 1930 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1931 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1932 file_dont_truncate(inode); 1933 1934 f2fs_up_write(&sbi->pin_sem); 1935 1936 expanded += map.m_len; 1937 sec_len -= map.m_len; 1938 map.m_lblk += map.m_len; 1939 if (!err && sec_len) 1940 goto next_alloc; 1941 1942 map.m_len = expanded; 1943 } else { 1944 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1945 expanded = map.m_len; 1946 } 1947 out_err: 1948 if (err) { 1949 pgoff_t last_off; 1950 1951 if (!expanded) 1952 return err; 1953 1954 last_off = pg_start + expanded - 1; 1955 1956 /* update new size to the failed position */ 1957 new_size = (last_off == pg_end) ? offset + len : 1958 (loff_t)(last_off + 1) << PAGE_SHIFT; 1959 } else { 1960 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1961 } 1962 1963 if (new_size > i_size_read(inode)) { 1964 if (mode & FALLOC_FL_KEEP_SIZE) 1965 file_set_keep_isize(inode); 1966 else 1967 f2fs_i_size_write(inode, new_size); 1968 } 1969 1970 return err; 1971 } 1972 1973 static long f2fs_fallocate(struct file *file, int mode, 1974 loff_t offset, loff_t len) 1975 { 1976 struct inode *inode = file_inode(file); 1977 long ret = 0; 1978 1979 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1980 return -EIO; 1981 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1982 return -ENOSPC; 1983 if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode)) 1984 return -EOPNOTSUPP; 1985 1986 /* f2fs only support ->fallocate for regular file */ 1987 if (!S_ISREG(inode->i_mode)) 1988 return -EINVAL; 1989 1990 if (IS_ENCRYPTED(inode) && 1991 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1992 return -EOPNOTSUPP; 1993 1994 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1995 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1996 FALLOC_FL_INSERT_RANGE)) 1997 return -EOPNOTSUPP; 1998 1999 inode_lock(inode); 2000 2001 /* 2002 * Pinned file should not support partial truncation since the block 2003 * can be used by applications. 2004 */ 2005 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 2006 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 2007 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 2008 ret = -EOPNOTSUPP; 2009 goto out; 2010 } 2011 2012 ret = file_modified(file); 2013 if (ret) 2014 goto out; 2015 2016 /* 2017 * wait for inflight dio, blocks should be removed after IO 2018 * completion. 2019 */ 2020 inode_dio_wait(inode); 2021 2022 if (mode & FALLOC_FL_PUNCH_HOLE) { 2023 if (offset >= inode->i_size) 2024 goto out; 2025 2026 ret = f2fs_punch_hole(inode, offset, len); 2027 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 2028 ret = f2fs_collapse_range(inode, offset, len); 2029 } else if (mode & FALLOC_FL_ZERO_RANGE) { 2030 ret = f2fs_zero_range(inode, offset, len, mode); 2031 } else if (mode & FALLOC_FL_INSERT_RANGE) { 2032 ret = f2fs_insert_range(inode, offset, len); 2033 } else { 2034 ret = f2fs_expand_inode_data(inode, offset, len, mode); 2035 } 2036 2037 if (!ret) { 2038 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 2039 f2fs_mark_inode_dirty_sync(inode, false); 2040 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2041 } 2042 2043 out: 2044 inode_unlock(inode); 2045 2046 trace_f2fs_fallocate(inode, mode, offset, len, ret); 2047 return ret; 2048 } 2049 2050 static int f2fs_release_file(struct inode *inode, struct file *filp) 2051 { 2052 if (atomic_dec_and_test(&F2FS_I(inode)->open_count)) 2053 f2fs_remove_donate_inode(inode); 2054 2055 /* 2056 * f2fs_release_file is called at every close calls. So we should 2057 * not drop any inmemory pages by close called by other process. 2058 */ 2059 if (!(filp->f_mode & FMODE_WRITE) || 2060 atomic_read(&inode->i_writecount) != 1) 2061 return 0; 2062 2063 inode_lock(inode); 2064 f2fs_abort_atomic_write(inode, true); 2065 inode_unlock(inode); 2066 2067 return 0; 2068 } 2069 2070 static int f2fs_file_flush(struct file *file, fl_owner_t id) 2071 { 2072 struct inode *inode = file_inode(file); 2073 2074 /* 2075 * If the process doing a transaction is crashed, we should do 2076 * roll-back. Otherwise, other reader/write can see corrupted database 2077 * until all the writers close its file. Since this should be done 2078 * before dropping file lock, it needs to do in ->flush. 2079 */ 2080 if (F2FS_I(inode)->atomic_write_task == current && 2081 (current->flags & PF_EXITING)) { 2082 inode_lock(inode); 2083 f2fs_abort_atomic_write(inode, true); 2084 inode_unlock(inode); 2085 } 2086 2087 return 0; 2088 } 2089 2090 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 2091 { 2092 struct f2fs_inode_info *fi = F2FS_I(inode); 2093 u32 masked_flags = fi->i_flags & mask; 2094 2095 /* mask can be shrunk by flags_valid selector */ 2096 iflags &= mask; 2097 2098 /* Is it quota file? Do not allow user to mess with it */ 2099 if (IS_NOQUOTA(inode)) 2100 return -EPERM; 2101 2102 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 2103 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 2104 return -EOPNOTSUPP; 2105 if (!f2fs_empty_dir(inode)) 2106 return -ENOTEMPTY; 2107 } 2108 2109 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 2110 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 2111 return -EOPNOTSUPP; 2112 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 2113 return -EINVAL; 2114 } 2115 2116 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 2117 if (masked_flags & F2FS_COMPR_FL) { 2118 if (!f2fs_disable_compressed_file(inode)) 2119 return -EINVAL; 2120 } else { 2121 /* try to convert inline_data to support compression */ 2122 int err = f2fs_convert_inline_inode(inode); 2123 if (err) 2124 return err; 2125 2126 f2fs_down_write(&fi->i_sem); 2127 if (!f2fs_may_compress(inode) || 2128 (S_ISREG(inode->i_mode) && 2129 F2FS_HAS_BLOCKS(inode))) { 2130 f2fs_up_write(&fi->i_sem); 2131 return -EINVAL; 2132 } 2133 err = set_compress_context(inode); 2134 f2fs_up_write(&fi->i_sem); 2135 2136 if (err) 2137 return err; 2138 } 2139 } 2140 2141 fi->i_flags = iflags | (fi->i_flags & ~mask); 2142 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 2143 (fi->i_flags & F2FS_NOCOMP_FL)); 2144 2145 if (fi->i_flags & F2FS_PROJINHERIT_FL) 2146 set_inode_flag(inode, FI_PROJ_INHERIT); 2147 else 2148 clear_inode_flag(inode, FI_PROJ_INHERIT); 2149 2150 inode_set_ctime_current(inode); 2151 f2fs_set_inode_flags(inode); 2152 f2fs_mark_inode_dirty_sync(inode, true); 2153 return 0; 2154 } 2155 2156 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 2157 2158 /* 2159 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 2160 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 2161 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 2162 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 2163 * 2164 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 2165 * FS_IOC_FSSETXATTR is done by the VFS. 2166 */ 2167 2168 static const struct { 2169 u32 iflag; 2170 u32 fsflag; 2171 } f2fs_fsflags_map[] = { 2172 { F2FS_COMPR_FL, FS_COMPR_FL }, 2173 { F2FS_SYNC_FL, FS_SYNC_FL }, 2174 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 2175 { F2FS_APPEND_FL, FS_APPEND_FL }, 2176 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 2177 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 2178 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 2179 { F2FS_INDEX_FL, FS_INDEX_FL }, 2180 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 2181 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2182 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2183 }; 2184 2185 #define F2FS_GETTABLE_FS_FL ( \ 2186 FS_COMPR_FL | \ 2187 FS_SYNC_FL | \ 2188 FS_IMMUTABLE_FL | \ 2189 FS_APPEND_FL | \ 2190 FS_NODUMP_FL | \ 2191 FS_NOATIME_FL | \ 2192 FS_NOCOMP_FL | \ 2193 FS_INDEX_FL | \ 2194 FS_DIRSYNC_FL | \ 2195 FS_PROJINHERIT_FL | \ 2196 FS_ENCRYPT_FL | \ 2197 FS_INLINE_DATA_FL | \ 2198 FS_NOCOW_FL | \ 2199 FS_VERITY_FL | \ 2200 FS_CASEFOLD_FL) 2201 2202 #define F2FS_SETTABLE_FS_FL ( \ 2203 FS_COMPR_FL | \ 2204 FS_SYNC_FL | \ 2205 FS_IMMUTABLE_FL | \ 2206 FS_APPEND_FL | \ 2207 FS_NODUMP_FL | \ 2208 FS_NOATIME_FL | \ 2209 FS_NOCOMP_FL | \ 2210 FS_DIRSYNC_FL | \ 2211 FS_PROJINHERIT_FL | \ 2212 FS_CASEFOLD_FL) 2213 2214 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2215 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2216 { 2217 u32 fsflags = 0; 2218 int i; 2219 2220 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2221 if (iflags & f2fs_fsflags_map[i].iflag) 2222 fsflags |= f2fs_fsflags_map[i].fsflag; 2223 2224 return fsflags; 2225 } 2226 2227 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2228 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2229 { 2230 u32 iflags = 0; 2231 int i; 2232 2233 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2234 if (fsflags & f2fs_fsflags_map[i].fsflag) 2235 iflags |= f2fs_fsflags_map[i].iflag; 2236 2237 return iflags; 2238 } 2239 2240 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2241 { 2242 struct inode *inode = file_inode(filp); 2243 2244 return put_user(inode->i_generation, (int __user *)arg); 2245 } 2246 2247 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2248 { 2249 struct inode *inode = file_inode(filp); 2250 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2251 struct f2fs_inode_info *fi = F2FS_I(inode); 2252 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2253 loff_t isize; 2254 int ret; 2255 2256 if (!(filp->f_mode & FMODE_WRITE)) 2257 return -EBADF; 2258 2259 if (!inode_owner_or_capable(idmap, inode)) 2260 return -EACCES; 2261 2262 if (!S_ISREG(inode->i_mode)) 2263 return -EINVAL; 2264 2265 if (filp->f_flags & O_DIRECT) 2266 return -EINVAL; 2267 2268 ret = mnt_want_write_file(filp); 2269 if (ret) 2270 return ret; 2271 2272 inode_lock(inode); 2273 2274 if (!f2fs_disable_compressed_file(inode) || 2275 f2fs_is_pinned_file(inode)) { 2276 ret = -EINVAL; 2277 goto out; 2278 } 2279 2280 if (f2fs_is_atomic_file(inode)) 2281 goto out; 2282 2283 ret = f2fs_convert_inline_inode(inode); 2284 if (ret) 2285 goto out; 2286 2287 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2288 f2fs_down_write(&fi->i_gc_rwsem[READ]); 2289 2290 /* 2291 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2292 * f2fs_is_atomic_file. 2293 */ 2294 if (get_dirty_pages(inode)) 2295 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2296 inode->i_ino, get_dirty_pages(inode)); 2297 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2298 if (ret) 2299 goto out_unlock; 2300 2301 /* Check if the inode already has a COW inode */ 2302 if (fi->cow_inode == NULL) { 2303 /* Create a COW inode for atomic write */ 2304 struct dentry *dentry = file_dentry(filp); 2305 struct inode *dir = d_inode(dentry->d_parent); 2306 2307 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode); 2308 if (ret) 2309 goto out_unlock; 2310 2311 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2312 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2313 2314 /* Set the COW inode's atomic_inode to the atomic inode */ 2315 F2FS_I(fi->cow_inode)->atomic_inode = inode; 2316 } else { 2317 /* Reuse the already created COW inode */ 2318 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode)); 2319 2320 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1); 2321 2322 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2323 if (ret) 2324 goto out_unlock; 2325 } 2326 2327 f2fs_write_inode(inode, NULL); 2328 2329 stat_inc_atomic_inode(inode); 2330 2331 set_inode_flag(inode, FI_ATOMIC_FILE); 2332 2333 isize = i_size_read(inode); 2334 fi->original_i_size = isize; 2335 if (truncate) { 2336 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2337 truncate_inode_pages_final(inode->i_mapping); 2338 f2fs_i_size_write(inode, 0); 2339 isize = 0; 2340 } 2341 f2fs_i_size_write(fi->cow_inode, isize); 2342 2343 out_unlock: 2344 f2fs_up_write(&fi->i_gc_rwsem[READ]); 2345 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2346 if (ret) 2347 goto out; 2348 2349 f2fs_update_time(sbi, REQ_TIME); 2350 fi->atomic_write_task = current; 2351 stat_update_max_atomic_write(inode); 2352 fi->atomic_write_cnt = 0; 2353 out: 2354 inode_unlock(inode); 2355 mnt_drop_write_file(filp); 2356 return ret; 2357 } 2358 2359 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2360 { 2361 struct inode *inode = file_inode(filp); 2362 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2363 int ret; 2364 2365 if (!(filp->f_mode & FMODE_WRITE)) 2366 return -EBADF; 2367 2368 if (!inode_owner_or_capable(idmap, inode)) 2369 return -EACCES; 2370 2371 ret = mnt_want_write_file(filp); 2372 if (ret) 2373 return ret; 2374 2375 f2fs_balance_fs(F2FS_I_SB(inode), true); 2376 2377 inode_lock(inode); 2378 2379 if (f2fs_is_atomic_file(inode)) { 2380 ret = f2fs_commit_atomic_write(inode); 2381 if (!ret) 2382 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2383 2384 f2fs_abort_atomic_write(inode, ret); 2385 } else { 2386 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2387 } 2388 2389 inode_unlock(inode); 2390 mnt_drop_write_file(filp); 2391 return ret; 2392 } 2393 2394 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2395 { 2396 struct inode *inode = file_inode(filp); 2397 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2398 int ret; 2399 2400 if (!(filp->f_mode & FMODE_WRITE)) 2401 return -EBADF; 2402 2403 if (!inode_owner_or_capable(idmap, inode)) 2404 return -EACCES; 2405 2406 ret = mnt_want_write_file(filp); 2407 if (ret) 2408 return ret; 2409 2410 inode_lock(inode); 2411 2412 f2fs_abort_atomic_write(inode, true); 2413 2414 inode_unlock(inode); 2415 2416 mnt_drop_write_file(filp); 2417 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2418 return ret; 2419 } 2420 2421 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag, 2422 bool readonly, bool need_lock) 2423 { 2424 struct super_block *sb = sbi->sb; 2425 int ret = 0; 2426 2427 switch (flag) { 2428 case F2FS_GOING_DOWN_FULLSYNC: 2429 ret = bdev_freeze(sb->s_bdev); 2430 if (ret) 2431 goto out; 2432 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2433 bdev_thaw(sb->s_bdev); 2434 break; 2435 case F2FS_GOING_DOWN_METASYNC: 2436 /* do checkpoint only */ 2437 ret = f2fs_sync_fs(sb, 1); 2438 if (ret) { 2439 if (ret == -EIO) 2440 ret = 0; 2441 goto out; 2442 } 2443 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2444 break; 2445 case F2FS_GOING_DOWN_NOSYNC: 2446 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2447 break; 2448 case F2FS_GOING_DOWN_METAFLUSH: 2449 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2450 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2451 break; 2452 case F2FS_GOING_DOWN_NEED_FSCK: 2453 set_sbi_flag(sbi, SBI_NEED_FSCK); 2454 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2455 set_sbi_flag(sbi, SBI_IS_DIRTY); 2456 /* do checkpoint only */ 2457 ret = f2fs_sync_fs(sb, 1); 2458 if (ret == -EIO) 2459 ret = 0; 2460 goto out; 2461 default: 2462 ret = -EINVAL; 2463 goto out; 2464 } 2465 2466 if (readonly) 2467 goto out; 2468 2469 /* 2470 * grab sb->s_umount to avoid racing w/ remount() and other shutdown 2471 * paths. 2472 */ 2473 if (need_lock) 2474 down_write(&sbi->sb->s_umount); 2475 2476 f2fs_stop_gc_thread(sbi); 2477 f2fs_stop_discard_thread(sbi); 2478 2479 f2fs_drop_discard_cmd(sbi); 2480 clear_opt(sbi, DISCARD); 2481 2482 if (need_lock) 2483 up_write(&sbi->sb->s_umount); 2484 2485 f2fs_update_time(sbi, REQ_TIME); 2486 out: 2487 2488 trace_f2fs_shutdown(sbi, flag, ret); 2489 2490 return ret; 2491 } 2492 2493 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2494 { 2495 struct inode *inode = file_inode(filp); 2496 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2497 __u32 in; 2498 int ret; 2499 bool need_drop = false, readonly = false; 2500 2501 if (!capable(CAP_SYS_ADMIN)) 2502 return -EPERM; 2503 2504 if (get_user(in, (__u32 __user *)arg)) 2505 return -EFAULT; 2506 2507 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2508 ret = mnt_want_write_file(filp); 2509 if (ret) { 2510 if (ret != -EROFS) 2511 return ret; 2512 2513 /* fallback to nosync shutdown for readonly fs */ 2514 in = F2FS_GOING_DOWN_NOSYNC; 2515 readonly = true; 2516 } else { 2517 need_drop = true; 2518 } 2519 } 2520 2521 ret = f2fs_do_shutdown(sbi, in, readonly, true); 2522 2523 if (need_drop) 2524 mnt_drop_write_file(filp); 2525 2526 return ret; 2527 } 2528 2529 static int f2fs_keep_noreuse_range(struct inode *inode, 2530 loff_t offset, loff_t len) 2531 { 2532 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2533 u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 2534 u64 start, end; 2535 int ret = 0; 2536 2537 if (!S_ISREG(inode->i_mode)) 2538 return 0; 2539 2540 if (offset >= max_bytes || len > max_bytes || 2541 (offset + len) > max_bytes) 2542 return 0; 2543 2544 start = offset >> PAGE_SHIFT; 2545 end = DIV_ROUND_UP(offset + len, PAGE_SIZE); 2546 2547 inode_lock(inode); 2548 if (f2fs_is_atomic_file(inode)) { 2549 inode_unlock(inode); 2550 return 0; 2551 } 2552 2553 spin_lock(&sbi->inode_lock[DONATE_INODE]); 2554 /* let's remove the range, if len = 0 */ 2555 if (!len) { 2556 if (!list_empty(&F2FS_I(inode)->gdonate_list)) { 2557 list_del_init(&F2FS_I(inode)->gdonate_list); 2558 sbi->donate_files--; 2559 if (is_inode_flag_set(inode, FI_DONATE_FINISHED)) 2560 ret = -EALREADY; 2561 else 2562 set_inode_flag(inode, FI_DONATE_FINISHED); 2563 } else 2564 ret = -ENOENT; 2565 } else { 2566 if (list_empty(&F2FS_I(inode)->gdonate_list)) { 2567 list_add_tail(&F2FS_I(inode)->gdonate_list, 2568 &sbi->inode_list[DONATE_INODE]); 2569 sbi->donate_files++; 2570 } else { 2571 list_move_tail(&F2FS_I(inode)->gdonate_list, 2572 &sbi->inode_list[DONATE_INODE]); 2573 } 2574 F2FS_I(inode)->donate_start = start; 2575 F2FS_I(inode)->donate_end = end - 1; 2576 clear_inode_flag(inode, FI_DONATE_FINISHED); 2577 } 2578 spin_unlock(&sbi->inode_lock[DONATE_INODE]); 2579 inode_unlock(inode); 2580 2581 return ret; 2582 } 2583 2584 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2585 { 2586 struct inode *inode = file_inode(filp); 2587 struct super_block *sb = inode->i_sb; 2588 struct fstrim_range range; 2589 int ret; 2590 2591 if (!capable(CAP_SYS_ADMIN)) 2592 return -EPERM; 2593 2594 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2595 return -EOPNOTSUPP; 2596 2597 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2598 sizeof(range))) 2599 return -EFAULT; 2600 2601 ret = mnt_want_write_file(filp); 2602 if (ret) 2603 return ret; 2604 2605 range.minlen = max((unsigned int)range.minlen, 2606 bdev_discard_granularity(sb->s_bdev)); 2607 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2608 mnt_drop_write_file(filp); 2609 if (ret < 0) 2610 return ret; 2611 2612 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2613 sizeof(range))) 2614 return -EFAULT; 2615 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2616 return 0; 2617 } 2618 2619 static bool uuid_is_nonzero(__u8 u[16]) 2620 { 2621 int i; 2622 2623 for (i = 0; i < 16; i++) 2624 if (u[i]) 2625 return true; 2626 return false; 2627 } 2628 2629 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2630 { 2631 struct inode *inode = file_inode(filp); 2632 int ret; 2633 2634 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2635 return -EOPNOTSUPP; 2636 2637 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2638 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2639 return ret; 2640 } 2641 2642 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2643 { 2644 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2645 return -EOPNOTSUPP; 2646 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2647 } 2648 2649 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2650 { 2651 struct inode *inode = file_inode(filp); 2652 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2653 u8 encrypt_pw_salt[16]; 2654 int err; 2655 2656 if (!f2fs_sb_has_encrypt(sbi)) 2657 return -EOPNOTSUPP; 2658 2659 err = mnt_want_write_file(filp); 2660 if (err) 2661 return err; 2662 2663 f2fs_down_write(&sbi->sb_lock); 2664 2665 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2666 goto got_it; 2667 2668 /* update superblock with uuid */ 2669 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2670 2671 err = f2fs_commit_super(sbi, false); 2672 if (err) { 2673 /* undo new data */ 2674 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2675 goto out_err; 2676 } 2677 got_it: 2678 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2679 out_err: 2680 f2fs_up_write(&sbi->sb_lock); 2681 mnt_drop_write_file(filp); 2682 2683 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2684 err = -EFAULT; 2685 2686 return err; 2687 } 2688 2689 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2690 unsigned long arg) 2691 { 2692 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2693 return -EOPNOTSUPP; 2694 2695 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2696 } 2697 2698 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2699 { 2700 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2701 return -EOPNOTSUPP; 2702 2703 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2704 } 2705 2706 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2707 { 2708 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2709 return -EOPNOTSUPP; 2710 2711 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2712 } 2713 2714 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2715 unsigned long arg) 2716 { 2717 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2718 return -EOPNOTSUPP; 2719 2720 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2721 } 2722 2723 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2724 unsigned long arg) 2725 { 2726 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2727 return -EOPNOTSUPP; 2728 2729 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2730 } 2731 2732 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2733 { 2734 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2735 return -EOPNOTSUPP; 2736 2737 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2738 } 2739 2740 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2741 { 2742 struct inode *inode = file_inode(filp); 2743 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2744 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2745 .no_bg_gc = false, 2746 .should_migrate_blocks = false, 2747 .nr_free_secs = 0 }; 2748 __u32 sync; 2749 int ret; 2750 2751 if (!capable(CAP_SYS_ADMIN)) 2752 return -EPERM; 2753 2754 if (get_user(sync, (__u32 __user *)arg)) 2755 return -EFAULT; 2756 2757 if (f2fs_readonly(sbi->sb)) 2758 return -EROFS; 2759 2760 ret = mnt_want_write_file(filp); 2761 if (ret) 2762 return ret; 2763 2764 if (!sync) { 2765 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2766 ret = -EBUSY; 2767 goto out; 2768 } 2769 } else { 2770 f2fs_down_write(&sbi->gc_lock); 2771 } 2772 2773 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2774 gc_control.err_gc_skipped = sync; 2775 stat_inc_gc_call_count(sbi, FOREGROUND); 2776 ret = f2fs_gc(sbi, &gc_control); 2777 out: 2778 mnt_drop_write_file(filp); 2779 return ret; 2780 } 2781 2782 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2783 { 2784 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2785 struct f2fs_gc_control gc_control = { 2786 .init_gc_type = range->sync ? FG_GC : BG_GC, 2787 .no_bg_gc = false, 2788 .should_migrate_blocks = false, 2789 .err_gc_skipped = range->sync, 2790 .nr_free_secs = 0 }; 2791 u64 end; 2792 int ret; 2793 2794 if (!capable(CAP_SYS_ADMIN)) 2795 return -EPERM; 2796 if (f2fs_readonly(sbi->sb)) 2797 return -EROFS; 2798 2799 end = range->start + range->len; 2800 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2801 end >= MAX_BLKADDR(sbi)) 2802 return -EINVAL; 2803 2804 ret = mnt_want_write_file(filp); 2805 if (ret) 2806 return ret; 2807 2808 do_more: 2809 if (!range->sync) { 2810 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2811 ret = -EBUSY; 2812 goto out; 2813 } 2814 } else { 2815 f2fs_down_write(&sbi->gc_lock); 2816 } 2817 2818 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2819 stat_inc_gc_call_count(sbi, FOREGROUND); 2820 ret = f2fs_gc(sbi, &gc_control); 2821 if (ret) { 2822 if (ret == -EBUSY) 2823 ret = -EAGAIN; 2824 goto out; 2825 } 2826 range->start += CAP_BLKS_PER_SEC(sbi); 2827 if (range->start <= end) 2828 goto do_more; 2829 out: 2830 mnt_drop_write_file(filp); 2831 return ret; 2832 } 2833 2834 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2835 { 2836 struct f2fs_gc_range range; 2837 2838 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2839 sizeof(range))) 2840 return -EFAULT; 2841 return __f2fs_ioc_gc_range(filp, &range); 2842 } 2843 2844 static int f2fs_ioc_write_checkpoint(struct file *filp) 2845 { 2846 struct inode *inode = file_inode(filp); 2847 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2848 int ret; 2849 2850 if (!capable(CAP_SYS_ADMIN)) 2851 return -EPERM; 2852 2853 if (f2fs_readonly(sbi->sb)) 2854 return -EROFS; 2855 2856 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2857 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2858 return -EINVAL; 2859 } 2860 2861 ret = mnt_want_write_file(filp); 2862 if (ret) 2863 return ret; 2864 2865 ret = f2fs_sync_fs(sbi->sb, 1); 2866 2867 mnt_drop_write_file(filp); 2868 return ret; 2869 } 2870 2871 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2872 struct file *filp, 2873 struct f2fs_defragment *range) 2874 { 2875 struct inode *inode = file_inode(filp); 2876 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2877 .m_seg_type = NO_CHECK_TYPE, 2878 .m_may_create = false }; 2879 struct extent_info ei = {}; 2880 pgoff_t pg_start, pg_end, next_pgofs; 2881 unsigned int total = 0, sec_num; 2882 block_t blk_end = 0; 2883 bool fragmented = false; 2884 int err; 2885 2886 f2fs_balance_fs(sbi, true); 2887 2888 inode_lock(inode); 2889 pg_start = range->start >> PAGE_SHIFT; 2890 pg_end = min_t(pgoff_t, 2891 (range->start + range->len) >> PAGE_SHIFT, 2892 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE)); 2893 2894 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) || 2895 f2fs_is_atomic_file(inode)) { 2896 err = -EINVAL; 2897 goto unlock_out; 2898 } 2899 2900 /* if in-place-update policy is enabled, don't waste time here */ 2901 set_inode_flag(inode, FI_OPU_WRITE); 2902 if (f2fs_should_update_inplace(inode, NULL)) { 2903 err = -EINVAL; 2904 goto out; 2905 } 2906 2907 /* writeback all dirty pages in the range */ 2908 err = filemap_write_and_wait_range(inode->i_mapping, 2909 pg_start << PAGE_SHIFT, 2910 (pg_end << PAGE_SHIFT) - 1); 2911 if (err) 2912 goto out; 2913 2914 /* 2915 * lookup mapping info in extent cache, skip defragmenting if physical 2916 * block addresses are continuous. 2917 */ 2918 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2919 if ((pgoff_t)ei.fofs + ei.len >= pg_end) 2920 goto out; 2921 } 2922 2923 map.m_lblk = pg_start; 2924 map.m_next_pgofs = &next_pgofs; 2925 2926 /* 2927 * lookup mapping info in dnode page cache, skip defragmenting if all 2928 * physical block addresses are continuous even if there are hole(s) 2929 * in logical blocks. 2930 */ 2931 while (map.m_lblk < pg_end) { 2932 map.m_len = pg_end - map.m_lblk; 2933 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2934 if (err) 2935 goto out; 2936 2937 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2938 map.m_lblk = next_pgofs; 2939 continue; 2940 } 2941 2942 if (blk_end && blk_end != map.m_pblk) 2943 fragmented = true; 2944 2945 /* record total count of block that we're going to move */ 2946 total += map.m_len; 2947 2948 blk_end = map.m_pblk + map.m_len; 2949 2950 map.m_lblk += map.m_len; 2951 } 2952 2953 if (!fragmented) { 2954 total = 0; 2955 goto out; 2956 } 2957 2958 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2959 2960 /* 2961 * make sure there are enough free section for LFS allocation, this can 2962 * avoid defragment running in SSR mode when free section are allocated 2963 * intensively 2964 */ 2965 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2966 err = -EAGAIN; 2967 goto out; 2968 } 2969 2970 map.m_lblk = pg_start; 2971 map.m_len = pg_end - pg_start; 2972 total = 0; 2973 2974 while (map.m_lblk < pg_end) { 2975 pgoff_t idx; 2976 int cnt = 0; 2977 2978 do_map: 2979 map.m_len = pg_end - map.m_lblk; 2980 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2981 if (err) 2982 goto clear_out; 2983 2984 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2985 map.m_lblk = next_pgofs; 2986 goto check; 2987 } 2988 2989 set_inode_flag(inode, FI_SKIP_WRITES); 2990 2991 idx = map.m_lblk; 2992 while (idx < map.m_lblk + map.m_len && 2993 cnt < BLKS_PER_SEG(sbi)) { 2994 struct folio *folio; 2995 2996 folio = f2fs_get_lock_data_folio(inode, idx, true); 2997 if (IS_ERR(folio)) { 2998 err = PTR_ERR(folio); 2999 goto clear_out; 3000 } 3001 3002 f2fs_folio_wait_writeback(folio, DATA, true, true); 3003 3004 folio_mark_dirty(folio); 3005 folio_set_f2fs_gcing(folio); 3006 f2fs_folio_put(folio, true); 3007 3008 idx++; 3009 cnt++; 3010 total++; 3011 } 3012 3013 map.m_lblk = idx; 3014 check: 3015 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 3016 goto do_map; 3017 3018 clear_inode_flag(inode, FI_SKIP_WRITES); 3019 3020 err = filemap_fdatawrite(inode->i_mapping); 3021 if (err) 3022 goto out; 3023 } 3024 clear_out: 3025 clear_inode_flag(inode, FI_SKIP_WRITES); 3026 out: 3027 clear_inode_flag(inode, FI_OPU_WRITE); 3028 unlock_out: 3029 inode_unlock(inode); 3030 if (!err) 3031 range->len = (u64)total << PAGE_SHIFT; 3032 return err; 3033 } 3034 3035 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 3036 { 3037 struct inode *inode = file_inode(filp); 3038 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3039 struct f2fs_defragment range; 3040 int err; 3041 3042 if (!capable(CAP_SYS_ADMIN)) 3043 return -EPERM; 3044 3045 if (!S_ISREG(inode->i_mode)) 3046 return -EINVAL; 3047 3048 if (f2fs_readonly(sbi->sb)) 3049 return -EROFS; 3050 3051 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 3052 sizeof(range))) 3053 return -EFAULT; 3054 3055 /* verify alignment of offset & size */ 3056 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 3057 return -EINVAL; 3058 3059 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 3060 max_file_blocks(inode))) 3061 return -EINVAL; 3062 3063 err = mnt_want_write_file(filp); 3064 if (err) 3065 return err; 3066 3067 err = f2fs_defragment_range(sbi, filp, &range); 3068 mnt_drop_write_file(filp); 3069 3070 if (range.len) 3071 f2fs_update_time(sbi, REQ_TIME); 3072 if (err < 0) 3073 return err; 3074 3075 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 3076 sizeof(range))) 3077 return -EFAULT; 3078 3079 return 0; 3080 } 3081 3082 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 3083 struct file *file_out, loff_t pos_out, size_t len) 3084 { 3085 struct inode *src = file_inode(file_in); 3086 struct inode *dst = file_inode(file_out); 3087 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 3088 size_t olen = len, dst_max_i_size = 0; 3089 size_t dst_osize; 3090 int ret; 3091 3092 if (file_in->f_path.mnt != file_out->f_path.mnt || 3093 src->i_sb != dst->i_sb) 3094 return -EXDEV; 3095 3096 if (unlikely(f2fs_readonly(src->i_sb))) 3097 return -EROFS; 3098 3099 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 3100 return -EINVAL; 3101 3102 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 3103 return -EOPNOTSUPP; 3104 3105 if (pos_out < 0 || pos_in < 0) 3106 return -EINVAL; 3107 3108 if (src == dst) { 3109 if (pos_in == pos_out) 3110 return 0; 3111 if (pos_out > pos_in && pos_out < pos_in + len) 3112 return -EINVAL; 3113 } 3114 3115 inode_lock(src); 3116 if (src != dst) { 3117 ret = -EBUSY; 3118 if (!inode_trylock(dst)) 3119 goto out; 3120 } 3121 3122 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) || 3123 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 3124 ret = -EOPNOTSUPP; 3125 goto out_unlock; 3126 } 3127 3128 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) { 3129 ret = -EINVAL; 3130 goto out_unlock; 3131 } 3132 3133 ret = -EINVAL; 3134 if (pos_in + len > src->i_size || pos_in + len < pos_in) 3135 goto out_unlock; 3136 if (len == 0) 3137 olen = len = src->i_size - pos_in; 3138 if (pos_in + len == src->i_size) 3139 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 3140 if (len == 0) { 3141 ret = 0; 3142 goto out_unlock; 3143 } 3144 3145 dst_osize = dst->i_size; 3146 if (pos_out + olen > dst->i_size) 3147 dst_max_i_size = pos_out + olen; 3148 3149 /* verify the end result is block aligned */ 3150 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 3151 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 3152 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 3153 goto out_unlock; 3154 3155 ret = f2fs_convert_inline_inode(src); 3156 if (ret) 3157 goto out_unlock; 3158 3159 ret = f2fs_convert_inline_inode(dst); 3160 if (ret) 3161 goto out_unlock; 3162 3163 /* write out all dirty pages from offset */ 3164 ret = filemap_write_and_wait_range(src->i_mapping, 3165 pos_in, pos_in + len); 3166 if (ret) 3167 goto out_unlock; 3168 3169 ret = filemap_write_and_wait_range(dst->i_mapping, 3170 pos_out, pos_out + len); 3171 if (ret) 3172 goto out_unlock; 3173 3174 f2fs_balance_fs(sbi, true); 3175 3176 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3177 if (src != dst) { 3178 ret = -EBUSY; 3179 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 3180 goto out_src; 3181 } 3182 3183 f2fs_lock_op(sbi); 3184 ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in), 3185 F2FS_BYTES_TO_BLK(pos_out), 3186 F2FS_BYTES_TO_BLK(len), false); 3187 3188 if (!ret) { 3189 if (dst_max_i_size) 3190 f2fs_i_size_write(dst, dst_max_i_size); 3191 else if (dst_osize != dst->i_size) 3192 f2fs_i_size_write(dst, dst_osize); 3193 } 3194 f2fs_unlock_op(sbi); 3195 3196 if (src != dst) 3197 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 3198 out_src: 3199 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3200 if (ret) 3201 goto out_unlock; 3202 3203 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 3204 f2fs_mark_inode_dirty_sync(src, false); 3205 if (src != dst) { 3206 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 3207 f2fs_mark_inode_dirty_sync(dst, false); 3208 } 3209 f2fs_update_time(sbi, REQ_TIME); 3210 3211 out_unlock: 3212 if (src != dst) 3213 inode_unlock(dst); 3214 out: 3215 inode_unlock(src); 3216 return ret; 3217 } 3218 3219 static int __f2fs_ioc_move_range(struct file *filp, 3220 struct f2fs_move_range *range) 3221 { 3222 int err; 3223 3224 if (!(filp->f_mode & FMODE_READ) || 3225 !(filp->f_mode & FMODE_WRITE)) 3226 return -EBADF; 3227 3228 CLASS(fd, dst)(range->dst_fd); 3229 if (fd_empty(dst)) 3230 return -EBADF; 3231 3232 if (!(fd_file(dst)->f_mode & FMODE_WRITE)) 3233 return -EBADF; 3234 3235 err = mnt_want_write_file(filp); 3236 if (err) 3237 return err; 3238 3239 err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst), 3240 range->pos_out, range->len); 3241 3242 mnt_drop_write_file(filp); 3243 return err; 3244 } 3245 3246 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 3247 { 3248 struct f2fs_move_range range; 3249 3250 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 3251 sizeof(range))) 3252 return -EFAULT; 3253 return __f2fs_ioc_move_range(filp, &range); 3254 } 3255 3256 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 3257 { 3258 struct inode *inode = file_inode(filp); 3259 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3260 struct sit_info *sm = SIT_I(sbi); 3261 unsigned int start_segno = 0, end_segno = 0; 3262 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3263 struct f2fs_flush_device range; 3264 struct f2fs_gc_control gc_control = { 3265 .init_gc_type = FG_GC, 3266 .should_migrate_blocks = true, 3267 .err_gc_skipped = true, 3268 .nr_free_secs = 0 }; 3269 int ret; 3270 3271 if (!capable(CAP_SYS_ADMIN)) 3272 return -EPERM; 3273 3274 if (f2fs_readonly(sbi->sb)) 3275 return -EROFS; 3276 3277 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3278 return -EINVAL; 3279 3280 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3281 sizeof(range))) 3282 return -EFAULT; 3283 3284 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3285 __is_large_section(sbi)) { 3286 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3287 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3288 return -EINVAL; 3289 } 3290 3291 ret = mnt_want_write_file(filp); 3292 if (ret) 3293 return ret; 3294 3295 if (range.dev_num != 0) 3296 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3297 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3298 3299 start_segno = sm->last_victim[FLUSH_DEVICE]; 3300 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3301 start_segno = dev_start_segno; 3302 end_segno = min(start_segno + range.segments, dev_end_segno); 3303 3304 while (start_segno < end_segno) { 3305 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3306 ret = -EBUSY; 3307 goto out; 3308 } 3309 sm->last_victim[GC_CB] = end_segno + 1; 3310 sm->last_victim[GC_GREEDY] = end_segno + 1; 3311 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3312 3313 gc_control.victim_segno = start_segno; 3314 stat_inc_gc_call_count(sbi, FOREGROUND); 3315 ret = f2fs_gc(sbi, &gc_control); 3316 if (ret == -EAGAIN) 3317 ret = 0; 3318 else if (ret < 0) 3319 break; 3320 start_segno++; 3321 } 3322 out: 3323 mnt_drop_write_file(filp); 3324 return ret; 3325 } 3326 3327 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3328 { 3329 struct inode *inode = file_inode(filp); 3330 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3331 3332 /* Must validate to set it with SQLite behavior in Android. */ 3333 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3334 3335 return put_user(sb_feature, (u32 __user *)arg); 3336 } 3337 3338 #ifdef CONFIG_QUOTA 3339 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3340 { 3341 struct dquot *transfer_to[MAXQUOTAS] = {}; 3342 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3343 struct super_block *sb = sbi->sb; 3344 int err; 3345 3346 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3347 if (IS_ERR(transfer_to[PRJQUOTA])) 3348 return PTR_ERR(transfer_to[PRJQUOTA]); 3349 3350 err = __dquot_transfer(inode, transfer_to); 3351 if (err) 3352 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3353 dqput(transfer_to[PRJQUOTA]); 3354 return err; 3355 } 3356 3357 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3358 { 3359 struct f2fs_inode_info *fi = F2FS_I(inode); 3360 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3361 struct f2fs_inode *ri = NULL; 3362 kprojid_t kprojid; 3363 int err; 3364 3365 if (!f2fs_sb_has_project_quota(sbi)) { 3366 if (projid != F2FS_DEF_PROJID) 3367 return -EOPNOTSUPP; 3368 else 3369 return 0; 3370 } 3371 3372 if (!f2fs_has_extra_attr(inode)) 3373 return -EOPNOTSUPP; 3374 3375 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3376 3377 if (projid_eq(kprojid, fi->i_projid)) 3378 return 0; 3379 3380 err = -EPERM; 3381 /* Is it quota file? Do not allow user to mess with it */ 3382 if (IS_NOQUOTA(inode)) 3383 return err; 3384 3385 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3386 return -EOVERFLOW; 3387 3388 err = f2fs_dquot_initialize(inode); 3389 if (err) 3390 return err; 3391 3392 f2fs_lock_op(sbi); 3393 err = f2fs_transfer_project_quota(inode, kprojid); 3394 if (err) 3395 goto out_unlock; 3396 3397 fi->i_projid = kprojid; 3398 inode_set_ctime_current(inode); 3399 f2fs_mark_inode_dirty_sync(inode, true); 3400 out_unlock: 3401 f2fs_unlock_op(sbi); 3402 return err; 3403 } 3404 #else 3405 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3406 { 3407 return 0; 3408 } 3409 3410 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3411 { 3412 if (projid != F2FS_DEF_PROJID) 3413 return -EOPNOTSUPP; 3414 return 0; 3415 } 3416 #endif 3417 3418 int f2fs_fileattr_get(struct dentry *dentry, struct file_kattr *fa) 3419 { 3420 struct inode *inode = d_inode(dentry); 3421 struct f2fs_inode_info *fi = F2FS_I(inode); 3422 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3423 3424 if (IS_ENCRYPTED(inode)) 3425 fsflags |= FS_ENCRYPT_FL; 3426 if (IS_VERITY(inode)) 3427 fsflags |= FS_VERITY_FL; 3428 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3429 fsflags |= FS_INLINE_DATA_FL; 3430 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3431 fsflags |= FS_NOCOW_FL; 3432 3433 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3434 3435 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3436 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3437 3438 return 0; 3439 } 3440 3441 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3442 struct dentry *dentry, struct file_kattr *fa) 3443 { 3444 struct inode *inode = d_inode(dentry); 3445 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3446 u32 iflags; 3447 int err; 3448 3449 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3450 return -EIO; 3451 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3452 return -ENOSPC; 3453 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3454 return -EOPNOTSUPP; 3455 fsflags &= F2FS_SETTABLE_FS_FL; 3456 if (!fa->flags_valid) 3457 mask &= FS_COMMON_FL; 3458 3459 iflags = f2fs_fsflags_to_iflags(fsflags); 3460 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3461 return -EOPNOTSUPP; 3462 3463 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3464 if (!err) 3465 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3466 3467 return err; 3468 } 3469 3470 int f2fs_pin_file_control(struct inode *inode, bool inc) 3471 { 3472 struct f2fs_inode_info *fi = F2FS_I(inode); 3473 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3474 3475 if (IS_DEVICE_ALIASING(inode)) 3476 return -EINVAL; 3477 3478 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) { 3479 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3480 __func__, inode->i_ino, fi->i_gc_failures); 3481 clear_inode_flag(inode, FI_PIN_FILE); 3482 return -EAGAIN; 3483 } 3484 3485 /* Use i_gc_failures for normal file as a risk signal. */ 3486 if (inc) 3487 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1); 3488 3489 return 0; 3490 } 3491 3492 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3493 { 3494 struct inode *inode = file_inode(filp); 3495 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3496 __u32 pin; 3497 int ret = 0; 3498 3499 if (get_user(pin, (__u32 __user *)arg)) 3500 return -EFAULT; 3501 3502 if (!S_ISREG(inode->i_mode)) 3503 return -EINVAL; 3504 3505 if (f2fs_readonly(sbi->sb)) 3506 return -EROFS; 3507 3508 if (!pin && IS_DEVICE_ALIASING(inode)) 3509 return -EOPNOTSUPP; 3510 3511 ret = mnt_want_write_file(filp); 3512 if (ret) 3513 return ret; 3514 3515 inode_lock(inode); 3516 3517 if (f2fs_is_atomic_file(inode)) { 3518 ret = -EINVAL; 3519 goto out; 3520 } 3521 3522 if (!pin) { 3523 clear_inode_flag(inode, FI_PIN_FILE); 3524 f2fs_i_gc_failures_write(inode, 0); 3525 goto done; 3526 } else if (f2fs_is_pinned_file(inode)) { 3527 goto done; 3528 } 3529 3530 if (F2FS_HAS_BLOCKS(inode)) { 3531 ret = -EFBIG; 3532 goto out; 3533 } 3534 3535 /* Let's allow file pinning on zoned device. */ 3536 if (!f2fs_sb_has_blkzoned(sbi) && 3537 f2fs_should_update_outplace(inode, NULL)) { 3538 ret = -EINVAL; 3539 goto out; 3540 } 3541 3542 if (f2fs_pin_file_control(inode, false)) { 3543 ret = -EAGAIN; 3544 goto out; 3545 } 3546 3547 ret = f2fs_convert_inline_inode(inode); 3548 if (ret) 3549 goto out; 3550 3551 if (!f2fs_disable_compressed_file(inode)) { 3552 ret = -EOPNOTSUPP; 3553 goto out; 3554 } 3555 3556 set_inode_flag(inode, FI_PIN_FILE); 3557 ret = F2FS_I(inode)->i_gc_failures; 3558 done: 3559 f2fs_update_time(sbi, REQ_TIME); 3560 out: 3561 inode_unlock(inode); 3562 mnt_drop_write_file(filp); 3563 return ret; 3564 } 3565 3566 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3567 { 3568 struct inode *inode = file_inode(filp); 3569 __u32 pin = 0; 3570 3571 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3572 pin = F2FS_I(inode)->i_gc_failures; 3573 return put_user(pin, (u32 __user *)arg); 3574 } 3575 3576 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg) 3577 { 3578 return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0, 3579 (u32 __user *)arg); 3580 } 3581 3582 static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg) 3583 { 3584 struct inode *inode = file_inode(filp); 3585 __u32 level; 3586 3587 if (get_user(level, (__u32 __user *)arg)) 3588 return -EFAULT; 3589 3590 if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX) 3591 return -EINVAL; 3592 3593 inode_lock(inode); 3594 F2FS_I(inode)->ioprio_hint = level; 3595 inode_unlock(inode); 3596 return 0; 3597 } 3598 3599 int f2fs_precache_extents(struct inode *inode) 3600 { 3601 struct f2fs_inode_info *fi = F2FS_I(inode); 3602 struct f2fs_map_blocks map; 3603 pgoff_t m_next_extent; 3604 loff_t end; 3605 int err; 3606 3607 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3608 return -EOPNOTSUPP; 3609 3610 map.m_lblk = 0; 3611 map.m_pblk = 0; 3612 map.m_next_pgofs = NULL; 3613 map.m_next_extent = &m_next_extent; 3614 map.m_seg_type = NO_CHECK_TYPE; 3615 map.m_may_create = false; 3616 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3617 3618 while (map.m_lblk < end) { 3619 map.m_len = end - map.m_lblk; 3620 3621 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3622 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3623 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3624 if (err || !map.m_len) 3625 return err; 3626 3627 map.m_lblk = m_next_extent; 3628 } 3629 3630 return 0; 3631 } 3632 3633 static int f2fs_ioc_precache_extents(struct file *filp) 3634 { 3635 return f2fs_precache_extents(file_inode(filp)); 3636 } 3637 3638 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3639 { 3640 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3641 __u64 block_count; 3642 3643 if (!capable(CAP_SYS_ADMIN)) 3644 return -EPERM; 3645 3646 if (f2fs_readonly(sbi->sb)) 3647 return -EROFS; 3648 3649 if (copy_from_user(&block_count, (void __user *)arg, 3650 sizeof(block_count))) 3651 return -EFAULT; 3652 3653 return f2fs_resize_fs(filp, block_count); 3654 } 3655 3656 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3657 { 3658 struct inode *inode = file_inode(filp); 3659 3660 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3661 3662 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3663 f2fs_warn(F2FS_I_SB(inode), 3664 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3665 inode->i_ino); 3666 return -EOPNOTSUPP; 3667 } 3668 3669 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3670 } 3671 3672 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3673 { 3674 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3675 return -EOPNOTSUPP; 3676 3677 return fsverity_ioctl_measure(filp, (void __user *)arg); 3678 } 3679 3680 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3681 { 3682 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3683 return -EOPNOTSUPP; 3684 3685 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3686 } 3687 3688 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3689 { 3690 struct inode *inode = file_inode(filp); 3691 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3692 char *vbuf; 3693 int count; 3694 int err = 0; 3695 3696 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3697 if (!vbuf) 3698 return -ENOMEM; 3699 3700 f2fs_down_read(&sbi->sb_lock); 3701 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3702 ARRAY_SIZE(sbi->raw_super->volume_name), 3703 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3704 f2fs_up_read(&sbi->sb_lock); 3705 3706 if (copy_to_user((char __user *)arg, vbuf, 3707 min(FSLABEL_MAX, count))) 3708 err = -EFAULT; 3709 3710 kfree(vbuf); 3711 return err; 3712 } 3713 3714 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3715 { 3716 struct inode *inode = file_inode(filp); 3717 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3718 char *vbuf; 3719 int err = 0; 3720 3721 if (!capable(CAP_SYS_ADMIN)) 3722 return -EPERM; 3723 3724 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3725 if (IS_ERR(vbuf)) 3726 return PTR_ERR(vbuf); 3727 3728 err = mnt_want_write_file(filp); 3729 if (err) 3730 goto out; 3731 3732 f2fs_down_write(&sbi->sb_lock); 3733 3734 memset(sbi->raw_super->volume_name, 0, 3735 sizeof(sbi->raw_super->volume_name)); 3736 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3737 sbi->raw_super->volume_name, 3738 ARRAY_SIZE(sbi->raw_super->volume_name)); 3739 3740 err = f2fs_commit_super(sbi, false); 3741 3742 f2fs_up_write(&sbi->sb_lock); 3743 3744 mnt_drop_write_file(filp); 3745 out: 3746 kfree(vbuf); 3747 return err; 3748 } 3749 3750 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3751 { 3752 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3753 return -EOPNOTSUPP; 3754 3755 if (!f2fs_compressed_file(inode)) 3756 return -EINVAL; 3757 3758 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3759 3760 return 0; 3761 } 3762 3763 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3764 { 3765 struct inode *inode = file_inode(filp); 3766 __u64 blocks; 3767 int ret; 3768 3769 ret = f2fs_get_compress_blocks(inode, &blocks); 3770 if (ret < 0) 3771 return ret; 3772 3773 return put_user(blocks, (u64 __user *)arg); 3774 } 3775 3776 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3777 { 3778 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3779 unsigned int released_blocks = 0; 3780 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3781 block_t blkaddr; 3782 int i; 3783 3784 for (i = 0; i < count; i++) { 3785 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3786 dn->ofs_in_node + i); 3787 3788 if (!__is_valid_data_blkaddr(blkaddr)) 3789 continue; 3790 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3791 DATA_GENERIC_ENHANCE))) 3792 return -EFSCORRUPTED; 3793 } 3794 3795 while (count) { 3796 int compr_blocks = 0; 3797 3798 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3799 blkaddr = f2fs_data_blkaddr(dn); 3800 3801 if (i == 0) { 3802 if (blkaddr == COMPRESS_ADDR) 3803 continue; 3804 dn->ofs_in_node += cluster_size; 3805 goto next; 3806 } 3807 3808 if (__is_valid_data_blkaddr(blkaddr)) 3809 compr_blocks++; 3810 3811 if (blkaddr != NEW_ADDR) 3812 continue; 3813 3814 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3815 } 3816 3817 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3818 dec_valid_block_count(sbi, dn->inode, 3819 cluster_size - compr_blocks); 3820 3821 released_blocks += cluster_size - compr_blocks; 3822 next: 3823 count -= cluster_size; 3824 } 3825 3826 return released_blocks; 3827 } 3828 3829 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3830 { 3831 struct inode *inode = file_inode(filp); 3832 struct f2fs_inode_info *fi = F2FS_I(inode); 3833 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3834 pgoff_t page_idx = 0, last_idx; 3835 unsigned int released_blocks = 0; 3836 int ret; 3837 int writecount; 3838 3839 if (!f2fs_sb_has_compression(sbi)) 3840 return -EOPNOTSUPP; 3841 3842 if (f2fs_readonly(sbi->sb)) 3843 return -EROFS; 3844 3845 ret = mnt_want_write_file(filp); 3846 if (ret) 3847 return ret; 3848 3849 f2fs_balance_fs(sbi, true); 3850 3851 inode_lock(inode); 3852 3853 writecount = atomic_read(&inode->i_writecount); 3854 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3855 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3856 ret = -EBUSY; 3857 goto out; 3858 } 3859 3860 if (!f2fs_compressed_file(inode) || 3861 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3862 ret = -EINVAL; 3863 goto out; 3864 } 3865 3866 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3867 if (ret) 3868 goto out; 3869 3870 if (!atomic_read(&fi->i_compr_blocks)) { 3871 ret = -EPERM; 3872 goto out; 3873 } 3874 3875 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3876 inode_set_ctime_current(inode); 3877 f2fs_mark_inode_dirty_sync(inode, true); 3878 3879 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3880 filemap_invalidate_lock(inode->i_mapping); 3881 3882 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3883 3884 while (page_idx < last_idx) { 3885 struct dnode_of_data dn; 3886 pgoff_t end_offset, count; 3887 3888 f2fs_lock_op(sbi); 3889 3890 set_new_dnode(&dn, inode, NULL, NULL, 0); 3891 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3892 if (ret) { 3893 f2fs_unlock_op(sbi); 3894 if (ret == -ENOENT) { 3895 page_idx = f2fs_get_next_page_offset(&dn, 3896 page_idx); 3897 ret = 0; 3898 continue; 3899 } 3900 break; 3901 } 3902 3903 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 3904 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3905 count = round_up(count, fi->i_cluster_size); 3906 3907 ret = release_compress_blocks(&dn, count); 3908 3909 f2fs_put_dnode(&dn); 3910 3911 f2fs_unlock_op(sbi); 3912 3913 if (ret < 0) 3914 break; 3915 3916 page_idx += count; 3917 released_blocks += ret; 3918 } 3919 3920 filemap_invalidate_unlock(inode->i_mapping); 3921 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3922 out: 3923 if (released_blocks) 3924 f2fs_update_time(sbi, REQ_TIME); 3925 inode_unlock(inode); 3926 3927 mnt_drop_write_file(filp); 3928 3929 if (ret >= 0) { 3930 ret = put_user(released_blocks, (u64 __user *)arg); 3931 } else if (released_blocks && 3932 atomic_read(&fi->i_compr_blocks)) { 3933 set_sbi_flag(sbi, SBI_NEED_FSCK); 3934 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3935 "iblocks=%llu, released=%u, compr_blocks=%u, " 3936 "run fsck to fix.", 3937 __func__, inode->i_ino, inode->i_blocks, 3938 released_blocks, 3939 atomic_read(&fi->i_compr_blocks)); 3940 } 3941 3942 return ret; 3943 } 3944 3945 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3946 unsigned int *reserved_blocks) 3947 { 3948 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3949 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3950 block_t blkaddr; 3951 int i; 3952 3953 for (i = 0; i < count; i++) { 3954 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3955 dn->ofs_in_node + i); 3956 3957 if (!__is_valid_data_blkaddr(blkaddr)) 3958 continue; 3959 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3960 DATA_GENERIC_ENHANCE))) 3961 return -EFSCORRUPTED; 3962 } 3963 3964 while (count) { 3965 int compr_blocks = 0; 3966 blkcnt_t reserved = 0; 3967 blkcnt_t to_reserved; 3968 int ret; 3969 3970 for (i = 0; i < cluster_size; i++) { 3971 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3972 dn->ofs_in_node + i); 3973 3974 if (i == 0) { 3975 if (blkaddr != COMPRESS_ADDR) { 3976 dn->ofs_in_node += cluster_size; 3977 goto next; 3978 } 3979 continue; 3980 } 3981 3982 /* 3983 * compressed cluster was not released due to it 3984 * fails in release_compress_blocks(), so NEW_ADDR 3985 * is a possible case. 3986 */ 3987 if (blkaddr == NEW_ADDR) { 3988 reserved++; 3989 continue; 3990 } 3991 if (__is_valid_data_blkaddr(blkaddr)) { 3992 compr_blocks++; 3993 continue; 3994 } 3995 } 3996 3997 to_reserved = cluster_size - compr_blocks - reserved; 3998 3999 /* for the case all blocks in cluster were reserved */ 4000 if (reserved && to_reserved == 1) { 4001 dn->ofs_in_node += cluster_size; 4002 goto next; 4003 } 4004 4005 ret = inc_valid_block_count(sbi, dn->inode, 4006 &to_reserved, false); 4007 if (unlikely(ret)) 4008 return ret; 4009 4010 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 4011 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 4012 f2fs_set_data_blkaddr(dn, NEW_ADDR); 4013 } 4014 4015 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 4016 4017 *reserved_blocks += to_reserved; 4018 next: 4019 count -= cluster_size; 4020 } 4021 4022 return 0; 4023 } 4024 4025 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 4026 { 4027 struct inode *inode = file_inode(filp); 4028 struct f2fs_inode_info *fi = F2FS_I(inode); 4029 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4030 pgoff_t page_idx = 0, last_idx; 4031 unsigned int reserved_blocks = 0; 4032 int ret; 4033 4034 if (!f2fs_sb_has_compression(sbi)) 4035 return -EOPNOTSUPP; 4036 4037 if (f2fs_readonly(sbi->sb)) 4038 return -EROFS; 4039 4040 ret = mnt_want_write_file(filp); 4041 if (ret) 4042 return ret; 4043 4044 f2fs_balance_fs(sbi, true); 4045 4046 inode_lock(inode); 4047 4048 if (!f2fs_compressed_file(inode) || 4049 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4050 ret = -EINVAL; 4051 goto unlock_inode; 4052 } 4053 4054 if (atomic_read(&fi->i_compr_blocks)) 4055 goto unlock_inode; 4056 4057 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 4058 filemap_invalidate_lock(inode->i_mapping); 4059 4060 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4061 4062 while (page_idx < last_idx) { 4063 struct dnode_of_data dn; 4064 pgoff_t end_offset, count; 4065 4066 f2fs_lock_op(sbi); 4067 4068 set_new_dnode(&dn, inode, NULL, NULL, 0); 4069 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 4070 if (ret) { 4071 f2fs_unlock_op(sbi); 4072 if (ret == -ENOENT) { 4073 page_idx = f2fs_get_next_page_offset(&dn, 4074 page_idx); 4075 ret = 0; 4076 continue; 4077 } 4078 break; 4079 } 4080 4081 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 4082 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 4083 count = round_up(count, fi->i_cluster_size); 4084 4085 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 4086 4087 f2fs_put_dnode(&dn); 4088 4089 f2fs_unlock_op(sbi); 4090 4091 if (ret < 0) 4092 break; 4093 4094 page_idx += count; 4095 } 4096 4097 filemap_invalidate_unlock(inode->i_mapping); 4098 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 4099 4100 if (!ret) { 4101 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 4102 inode_set_ctime_current(inode); 4103 f2fs_mark_inode_dirty_sync(inode, true); 4104 } 4105 unlock_inode: 4106 if (reserved_blocks) 4107 f2fs_update_time(sbi, REQ_TIME); 4108 inode_unlock(inode); 4109 mnt_drop_write_file(filp); 4110 4111 if (!ret) { 4112 ret = put_user(reserved_blocks, (u64 __user *)arg); 4113 } else if (reserved_blocks && 4114 atomic_read(&fi->i_compr_blocks)) { 4115 set_sbi_flag(sbi, SBI_NEED_FSCK); 4116 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx " 4117 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 4118 "run fsck to fix.", 4119 __func__, inode->i_ino, inode->i_blocks, 4120 reserved_blocks, 4121 atomic_read(&fi->i_compr_blocks)); 4122 } 4123 4124 return ret; 4125 } 4126 4127 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 4128 pgoff_t off, block_t block, block_t len, u32 flags) 4129 { 4130 sector_t sector = SECTOR_FROM_BLOCK(block); 4131 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 4132 int ret = 0; 4133 4134 if (flags & F2FS_TRIM_FILE_DISCARD) { 4135 if (bdev_max_secure_erase_sectors(bdev)) 4136 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 4137 GFP_NOFS); 4138 else 4139 ret = blkdev_issue_discard(bdev, sector, nr_sects, 4140 GFP_NOFS); 4141 } 4142 4143 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 4144 if (IS_ENCRYPTED(inode)) 4145 ret = fscrypt_zeroout_range(inode, off, block, len); 4146 else 4147 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 4148 GFP_NOFS, 0); 4149 } 4150 4151 return ret; 4152 } 4153 4154 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 4155 { 4156 struct inode *inode = file_inode(filp); 4157 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4158 struct address_space *mapping = inode->i_mapping; 4159 struct block_device *prev_bdev = NULL; 4160 struct f2fs_sectrim_range range; 4161 pgoff_t index, pg_end, prev_index = 0; 4162 block_t prev_block = 0, len = 0; 4163 loff_t end_addr; 4164 bool to_end = false; 4165 int ret = 0; 4166 4167 if (!(filp->f_mode & FMODE_WRITE)) 4168 return -EBADF; 4169 4170 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 4171 sizeof(range))) 4172 return -EFAULT; 4173 4174 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 4175 !S_ISREG(inode->i_mode)) 4176 return -EINVAL; 4177 4178 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 4179 !f2fs_hw_support_discard(sbi)) || 4180 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 4181 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 4182 return -EOPNOTSUPP; 4183 4184 ret = mnt_want_write_file(filp); 4185 if (ret) 4186 return ret; 4187 inode_lock(inode); 4188 4189 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 4190 range.start >= inode->i_size) { 4191 ret = -EINVAL; 4192 goto err; 4193 } 4194 4195 if (range.len == 0) 4196 goto err; 4197 4198 if (inode->i_size - range.start > range.len) { 4199 end_addr = range.start + range.len; 4200 } else { 4201 end_addr = range.len == (u64)-1 ? 4202 sbi->sb->s_maxbytes : inode->i_size; 4203 to_end = true; 4204 } 4205 4206 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 4207 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 4208 ret = -EINVAL; 4209 goto err; 4210 } 4211 4212 index = F2FS_BYTES_TO_BLK(range.start); 4213 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 4214 4215 ret = f2fs_convert_inline_inode(inode); 4216 if (ret) 4217 goto err; 4218 4219 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4220 filemap_invalidate_lock(mapping); 4221 4222 ret = filemap_write_and_wait_range(mapping, range.start, 4223 to_end ? LLONG_MAX : end_addr - 1); 4224 if (ret) 4225 goto out; 4226 4227 truncate_inode_pages_range(mapping, range.start, 4228 to_end ? -1 : end_addr - 1); 4229 4230 while (index < pg_end) { 4231 struct dnode_of_data dn; 4232 pgoff_t end_offset, count; 4233 int i; 4234 4235 set_new_dnode(&dn, inode, NULL, NULL, 0); 4236 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 4237 if (ret) { 4238 if (ret == -ENOENT) { 4239 index = f2fs_get_next_page_offset(&dn, index); 4240 continue; 4241 } 4242 goto out; 4243 } 4244 4245 end_offset = ADDRS_PER_PAGE(dn.node_folio, inode); 4246 count = min(end_offset - dn.ofs_in_node, pg_end - index); 4247 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 4248 struct block_device *cur_bdev; 4249 block_t blkaddr = f2fs_data_blkaddr(&dn); 4250 4251 if (!__is_valid_data_blkaddr(blkaddr)) 4252 continue; 4253 4254 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 4255 DATA_GENERIC_ENHANCE)) { 4256 ret = -EFSCORRUPTED; 4257 f2fs_put_dnode(&dn); 4258 goto out; 4259 } 4260 4261 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 4262 if (f2fs_is_multi_device(sbi)) { 4263 int di = f2fs_target_device_index(sbi, blkaddr); 4264 4265 blkaddr -= FDEV(di).start_blk; 4266 } 4267 4268 if (len) { 4269 if (prev_bdev == cur_bdev && 4270 index == prev_index + len && 4271 blkaddr == prev_block + len) { 4272 len++; 4273 } else { 4274 ret = f2fs_secure_erase(prev_bdev, 4275 inode, prev_index, prev_block, 4276 len, range.flags); 4277 if (ret) { 4278 f2fs_put_dnode(&dn); 4279 goto out; 4280 } 4281 4282 len = 0; 4283 } 4284 } 4285 4286 if (!len) { 4287 prev_bdev = cur_bdev; 4288 prev_index = index; 4289 prev_block = blkaddr; 4290 len = 1; 4291 } 4292 } 4293 4294 f2fs_put_dnode(&dn); 4295 4296 if (fatal_signal_pending(current)) { 4297 ret = -EINTR; 4298 goto out; 4299 } 4300 cond_resched(); 4301 } 4302 4303 if (len) 4304 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 4305 prev_block, len, range.flags); 4306 f2fs_update_time(sbi, REQ_TIME); 4307 out: 4308 filemap_invalidate_unlock(mapping); 4309 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4310 err: 4311 inode_unlock(inode); 4312 mnt_drop_write_file(filp); 4313 4314 return ret; 4315 } 4316 4317 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 4318 { 4319 struct inode *inode = file_inode(filp); 4320 struct f2fs_comp_option option; 4321 4322 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4323 return -EOPNOTSUPP; 4324 4325 inode_lock_shared(inode); 4326 4327 if (!f2fs_compressed_file(inode)) { 4328 inode_unlock_shared(inode); 4329 return -ENODATA; 4330 } 4331 4332 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4333 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4334 4335 inode_unlock_shared(inode); 4336 4337 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4338 sizeof(option))) 4339 return -EFAULT; 4340 4341 return 0; 4342 } 4343 4344 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4345 { 4346 struct inode *inode = file_inode(filp); 4347 struct f2fs_inode_info *fi = F2FS_I(inode); 4348 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4349 struct f2fs_comp_option option; 4350 int ret = 0; 4351 4352 if (!f2fs_sb_has_compression(sbi)) 4353 return -EOPNOTSUPP; 4354 4355 if (!(filp->f_mode & FMODE_WRITE)) 4356 return -EBADF; 4357 4358 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4359 sizeof(option))) 4360 return -EFAULT; 4361 4362 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4363 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4364 option.algorithm >= COMPRESS_MAX) 4365 return -EINVAL; 4366 4367 ret = mnt_want_write_file(filp); 4368 if (ret) 4369 return ret; 4370 inode_lock(inode); 4371 4372 f2fs_down_write(&F2FS_I(inode)->i_sem); 4373 if (!f2fs_compressed_file(inode)) { 4374 ret = -EINVAL; 4375 goto out; 4376 } 4377 4378 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4379 ret = -EBUSY; 4380 goto out; 4381 } 4382 4383 if (F2FS_HAS_BLOCKS(inode)) { 4384 ret = -EFBIG; 4385 goto out; 4386 } 4387 4388 fi->i_compress_algorithm = option.algorithm; 4389 fi->i_log_cluster_size = option.log_cluster_size; 4390 fi->i_cluster_size = BIT(option.log_cluster_size); 4391 /* Set default level */ 4392 if (fi->i_compress_algorithm == COMPRESS_ZSTD) 4393 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4394 else 4395 fi->i_compress_level = 0; 4396 /* Adjust mount option level */ 4397 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4398 F2FS_OPTION(sbi).compress_level) 4399 fi->i_compress_level = F2FS_OPTION(sbi).compress_level; 4400 f2fs_mark_inode_dirty_sync(inode, true); 4401 4402 if (!f2fs_is_compress_backend_ready(inode)) 4403 f2fs_warn(sbi, "compression algorithm is successfully set, " 4404 "but current kernel doesn't support this algorithm."); 4405 out: 4406 f2fs_up_write(&fi->i_sem); 4407 inode_unlock(inode); 4408 mnt_drop_write_file(filp); 4409 4410 return ret; 4411 } 4412 4413 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4414 { 4415 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4416 struct address_space *mapping = inode->i_mapping; 4417 struct folio *folio; 4418 pgoff_t redirty_idx = page_idx; 4419 int page_len = 0, ret = 0; 4420 4421 page_cache_ra_unbounded(&ractl, len, 0); 4422 4423 do { 4424 folio = read_cache_folio(mapping, page_idx, NULL, NULL); 4425 if (IS_ERR(folio)) { 4426 ret = PTR_ERR(folio); 4427 break; 4428 } 4429 page_len += folio_nr_pages(folio) - (page_idx - folio->index); 4430 page_idx = folio_next_index(folio); 4431 } while (page_len < len); 4432 4433 do { 4434 folio = filemap_lock_folio(mapping, redirty_idx); 4435 4436 /* It will never fail, when folio has pinned above */ 4437 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio)); 4438 4439 f2fs_folio_wait_writeback(folio, DATA, true, true); 4440 4441 folio_mark_dirty(folio); 4442 folio_set_f2fs_gcing(folio); 4443 redirty_idx = folio_next_index(folio); 4444 folio_unlock(folio); 4445 folio_put_refs(folio, 2); 4446 } while (redirty_idx < page_idx); 4447 4448 return ret; 4449 } 4450 4451 static int f2fs_ioc_decompress_file(struct file *filp) 4452 { 4453 struct inode *inode = file_inode(filp); 4454 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4455 struct f2fs_inode_info *fi = F2FS_I(inode); 4456 pgoff_t page_idx = 0, last_idx, cluster_idx; 4457 int ret; 4458 4459 if (!f2fs_sb_has_compression(sbi) || 4460 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4461 return -EOPNOTSUPP; 4462 4463 if (!(filp->f_mode & FMODE_WRITE)) 4464 return -EBADF; 4465 4466 f2fs_balance_fs(sbi, true); 4467 4468 ret = mnt_want_write_file(filp); 4469 if (ret) 4470 return ret; 4471 inode_lock(inode); 4472 4473 if (!f2fs_is_compress_backend_ready(inode)) { 4474 ret = -EOPNOTSUPP; 4475 goto out; 4476 } 4477 4478 if (!f2fs_compressed_file(inode) || 4479 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4480 ret = -EINVAL; 4481 goto out; 4482 } 4483 4484 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4485 if (ret) 4486 goto out; 4487 4488 if (!atomic_read(&fi->i_compr_blocks)) 4489 goto out; 4490 4491 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4492 last_idx >>= fi->i_log_cluster_size; 4493 4494 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4495 page_idx = cluster_idx << fi->i_log_cluster_size; 4496 4497 if (!f2fs_is_compressed_cluster(inode, page_idx)) 4498 continue; 4499 4500 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4501 if (ret < 0) 4502 break; 4503 4504 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4505 ret = filemap_fdatawrite(inode->i_mapping); 4506 if (ret < 0) 4507 break; 4508 } 4509 4510 cond_resched(); 4511 if (fatal_signal_pending(current)) { 4512 ret = -EINTR; 4513 break; 4514 } 4515 } 4516 4517 if (!ret) 4518 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4519 LLONG_MAX); 4520 4521 if (ret) 4522 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4523 __func__, ret); 4524 f2fs_update_time(sbi, REQ_TIME); 4525 out: 4526 inode_unlock(inode); 4527 mnt_drop_write_file(filp); 4528 4529 return ret; 4530 } 4531 4532 static int f2fs_ioc_compress_file(struct file *filp) 4533 { 4534 struct inode *inode = file_inode(filp); 4535 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4536 struct f2fs_inode_info *fi = F2FS_I(inode); 4537 pgoff_t page_idx = 0, last_idx, cluster_idx; 4538 int ret; 4539 4540 if (!f2fs_sb_has_compression(sbi) || 4541 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4542 return -EOPNOTSUPP; 4543 4544 if (!(filp->f_mode & FMODE_WRITE)) 4545 return -EBADF; 4546 4547 f2fs_balance_fs(sbi, true); 4548 4549 ret = mnt_want_write_file(filp); 4550 if (ret) 4551 return ret; 4552 inode_lock(inode); 4553 4554 if (!f2fs_is_compress_backend_ready(inode)) { 4555 ret = -EOPNOTSUPP; 4556 goto out; 4557 } 4558 4559 if (!f2fs_compressed_file(inode) || 4560 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4561 ret = -EINVAL; 4562 goto out; 4563 } 4564 4565 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4566 if (ret) 4567 goto out; 4568 4569 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4570 4571 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4572 last_idx >>= fi->i_log_cluster_size; 4573 4574 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4575 page_idx = cluster_idx << fi->i_log_cluster_size; 4576 4577 if (f2fs_is_sparse_cluster(inode, page_idx)) 4578 continue; 4579 4580 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4581 if (ret < 0) 4582 break; 4583 4584 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4585 ret = filemap_fdatawrite(inode->i_mapping); 4586 if (ret < 0) 4587 break; 4588 } 4589 4590 cond_resched(); 4591 if (fatal_signal_pending(current)) { 4592 ret = -EINTR; 4593 break; 4594 } 4595 } 4596 4597 if (!ret) 4598 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4599 LLONG_MAX); 4600 4601 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4602 4603 if (ret) 4604 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4605 __func__, ret); 4606 f2fs_update_time(sbi, REQ_TIME); 4607 out: 4608 inode_unlock(inode); 4609 mnt_drop_write_file(filp); 4610 4611 return ret; 4612 } 4613 4614 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4615 { 4616 switch (cmd) { 4617 case FS_IOC_GETVERSION: 4618 return f2fs_ioc_getversion(filp, arg); 4619 case F2FS_IOC_START_ATOMIC_WRITE: 4620 return f2fs_ioc_start_atomic_write(filp, false); 4621 case F2FS_IOC_START_ATOMIC_REPLACE: 4622 return f2fs_ioc_start_atomic_write(filp, true); 4623 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4624 return f2fs_ioc_commit_atomic_write(filp); 4625 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4626 return f2fs_ioc_abort_atomic_write(filp); 4627 case F2FS_IOC_START_VOLATILE_WRITE: 4628 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4629 return -EOPNOTSUPP; 4630 case F2FS_IOC_SHUTDOWN: 4631 return f2fs_ioc_shutdown(filp, arg); 4632 case FITRIM: 4633 return f2fs_ioc_fitrim(filp, arg); 4634 case FS_IOC_SET_ENCRYPTION_POLICY: 4635 return f2fs_ioc_set_encryption_policy(filp, arg); 4636 case FS_IOC_GET_ENCRYPTION_POLICY: 4637 return f2fs_ioc_get_encryption_policy(filp, arg); 4638 case FS_IOC_GET_ENCRYPTION_PWSALT: 4639 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4640 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4641 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4642 case FS_IOC_ADD_ENCRYPTION_KEY: 4643 return f2fs_ioc_add_encryption_key(filp, arg); 4644 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4645 return f2fs_ioc_remove_encryption_key(filp, arg); 4646 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4647 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4648 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4649 return f2fs_ioc_get_encryption_key_status(filp, arg); 4650 case FS_IOC_GET_ENCRYPTION_NONCE: 4651 return f2fs_ioc_get_encryption_nonce(filp, arg); 4652 case F2FS_IOC_GARBAGE_COLLECT: 4653 return f2fs_ioc_gc(filp, arg); 4654 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4655 return f2fs_ioc_gc_range(filp, arg); 4656 case F2FS_IOC_WRITE_CHECKPOINT: 4657 return f2fs_ioc_write_checkpoint(filp); 4658 case F2FS_IOC_DEFRAGMENT: 4659 return f2fs_ioc_defragment(filp, arg); 4660 case F2FS_IOC_MOVE_RANGE: 4661 return f2fs_ioc_move_range(filp, arg); 4662 case F2FS_IOC_FLUSH_DEVICE: 4663 return f2fs_ioc_flush_device(filp, arg); 4664 case F2FS_IOC_GET_FEATURES: 4665 return f2fs_ioc_get_features(filp, arg); 4666 case F2FS_IOC_GET_PIN_FILE: 4667 return f2fs_ioc_get_pin_file(filp, arg); 4668 case F2FS_IOC_SET_PIN_FILE: 4669 return f2fs_ioc_set_pin_file(filp, arg); 4670 case F2FS_IOC_PRECACHE_EXTENTS: 4671 return f2fs_ioc_precache_extents(filp); 4672 case F2FS_IOC_RESIZE_FS: 4673 return f2fs_ioc_resize_fs(filp, arg); 4674 case FS_IOC_ENABLE_VERITY: 4675 return f2fs_ioc_enable_verity(filp, arg); 4676 case FS_IOC_MEASURE_VERITY: 4677 return f2fs_ioc_measure_verity(filp, arg); 4678 case FS_IOC_READ_VERITY_METADATA: 4679 return f2fs_ioc_read_verity_metadata(filp, arg); 4680 case FS_IOC_GETFSLABEL: 4681 return f2fs_ioc_getfslabel(filp, arg); 4682 case FS_IOC_SETFSLABEL: 4683 return f2fs_ioc_setfslabel(filp, arg); 4684 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4685 return f2fs_ioc_get_compress_blocks(filp, arg); 4686 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4687 return f2fs_release_compress_blocks(filp, arg); 4688 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4689 return f2fs_reserve_compress_blocks(filp, arg); 4690 case F2FS_IOC_SEC_TRIM_FILE: 4691 return f2fs_sec_trim_file(filp, arg); 4692 case F2FS_IOC_GET_COMPRESS_OPTION: 4693 return f2fs_ioc_get_compress_option(filp, arg); 4694 case F2FS_IOC_SET_COMPRESS_OPTION: 4695 return f2fs_ioc_set_compress_option(filp, arg); 4696 case F2FS_IOC_DECOMPRESS_FILE: 4697 return f2fs_ioc_decompress_file(filp); 4698 case F2FS_IOC_COMPRESS_FILE: 4699 return f2fs_ioc_compress_file(filp); 4700 case F2FS_IOC_GET_DEV_ALIAS_FILE: 4701 return f2fs_ioc_get_dev_alias_file(filp, arg); 4702 case F2FS_IOC_IO_PRIO: 4703 return f2fs_ioc_io_prio(filp, arg); 4704 default: 4705 return -ENOTTY; 4706 } 4707 } 4708 4709 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4710 { 4711 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4712 return -EIO; 4713 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4714 return -ENOSPC; 4715 4716 return __f2fs_ioctl(filp, cmd, arg); 4717 } 4718 4719 /* 4720 * Return %true if the given read or write request should use direct I/O, or 4721 * %false if it should use buffered I/O. 4722 */ 4723 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4724 struct iov_iter *iter) 4725 { 4726 unsigned int align; 4727 4728 if (!(iocb->ki_flags & IOCB_DIRECT)) 4729 return false; 4730 4731 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4732 return false; 4733 4734 /* 4735 * Direct I/O not aligned to the disk's logical_block_size will be 4736 * attempted, but will fail with -EINVAL. 4737 * 4738 * f2fs additionally requires that direct I/O be aligned to the 4739 * filesystem block size, which is often a stricter requirement. 4740 * However, f2fs traditionally falls back to buffered I/O on requests 4741 * that are logical_block_size-aligned but not fs-block aligned. 4742 * 4743 * The below logic implements this behavior. 4744 */ 4745 align = iocb->ki_pos | iov_iter_alignment(iter); 4746 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4747 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4748 return false; 4749 4750 return true; 4751 } 4752 4753 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4754 unsigned int flags) 4755 { 4756 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4757 4758 dec_page_count(sbi, F2FS_DIO_READ); 4759 if (error) 4760 return error; 4761 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4762 return 0; 4763 } 4764 4765 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4766 .end_io = f2fs_dio_read_end_io, 4767 }; 4768 4769 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4770 { 4771 struct file *file = iocb->ki_filp; 4772 struct inode *inode = file_inode(file); 4773 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4774 struct f2fs_inode_info *fi = F2FS_I(inode); 4775 const loff_t pos = iocb->ki_pos; 4776 const size_t count = iov_iter_count(to); 4777 struct iomap_dio *dio; 4778 ssize_t ret; 4779 4780 if (count == 0) 4781 return 0; /* skip atime update */ 4782 4783 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4784 4785 if (iocb->ki_flags & IOCB_NOWAIT) { 4786 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4787 ret = -EAGAIN; 4788 goto out; 4789 } 4790 } else { 4791 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4792 } 4793 4794 /* dio is not compatible w/ atomic file */ 4795 if (f2fs_is_atomic_file(inode)) { 4796 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4797 ret = -EOPNOTSUPP; 4798 goto out; 4799 } 4800 4801 /* 4802 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4803 * the higher-level function iomap_dio_rw() in order to ensure that the 4804 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4805 */ 4806 inc_page_count(sbi, F2FS_DIO_READ); 4807 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4808 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4809 if (IS_ERR_OR_NULL(dio)) { 4810 ret = PTR_ERR_OR_ZERO(dio); 4811 if (ret != -EIOCBQUEUED) 4812 dec_page_count(sbi, F2FS_DIO_READ); 4813 } else { 4814 ret = iomap_dio_complete(dio); 4815 } 4816 4817 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4818 4819 file_accessed(file); 4820 out: 4821 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4822 return ret; 4823 } 4824 4825 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4826 int rw) 4827 { 4828 struct inode *inode = file_inode(file); 4829 char *buf, *path; 4830 4831 buf = f2fs_getname(F2FS_I_SB(inode)); 4832 if (!buf) 4833 return; 4834 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4835 if (IS_ERR(path)) 4836 goto free_buf; 4837 if (rw == WRITE) 4838 trace_f2fs_datawrite_start(inode, pos, count, 4839 current->pid, path, current->comm); 4840 else 4841 trace_f2fs_dataread_start(inode, pos, count, 4842 current->pid, path, current->comm); 4843 free_buf: 4844 f2fs_putname(buf); 4845 } 4846 4847 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4848 { 4849 struct inode *inode = file_inode(iocb->ki_filp); 4850 const loff_t pos = iocb->ki_pos; 4851 ssize_t ret; 4852 bool dio; 4853 4854 if (!f2fs_is_compress_backend_ready(inode)) 4855 return -EOPNOTSUPP; 4856 4857 if (trace_f2fs_dataread_start_enabled()) 4858 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4859 iov_iter_count(to), READ); 4860 4861 dio = f2fs_should_use_dio(inode, iocb, to); 4862 4863 /* In LFS mode, if there is inflight dio, wait for its completion */ 4864 if (f2fs_lfs_mode(F2FS_I_SB(inode)) && 4865 get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE) && 4866 (!f2fs_is_pinned_file(inode) || !dio)) 4867 inode_dio_wait(inode); 4868 4869 if (dio) { 4870 ret = f2fs_dio_read_iter(iocb, to); 4871 } else { 4872 ret = filemap_read(iocb, to, 0); 4873 if (ret > 0) 4874 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4875 APP_BUFFERED_READ_IO, ret); 4876 } 4877 trace_f2fs_dataread_end(inode, pos, ret); 4878 return ret; 4879 } 4880 4881 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4882 struct pipe_inode_info *pipe, 4883 size_t len, unsigned int flags) 4884 { 4885 struct inode *inode = file_inode(in); 4886 const loff_t pos = *ppos; 4887 ssize_t ret; 4888 4889 if (!f2fs_is_compress_backend_ready(inode)) 4890 return -EOPNOTSUPP; 4891 4892 if (trace_f2fs_dataread_start_enabled()) 4893 f2fs_trace_rw_file_path(in, pos, len, READ); 4894 4895 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4896 if (ret > 0) 4897 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4898 APP_BUFFERED_READ_IO, ret); 4899 4900 trace_f2fs_dataread_end(inode, pos, ret); 4901 return ret; 4902 } 4903 4904 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4905 { 4906 struct file *file = iocb->ki_filp; 4907 struct inode *inode = file_inode(file); 4908 ssize_t count; 4909 int err; 4910 4911 if (IS_IMMUTABLE(inode)) 4912 return -EPERM; 4913 4914 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4915 return -EPERM; 4916 4917 count = generic_write_checks(iocb, from); 4918 if (count <= 0) 4919 return count; 4920 4921 err = file_modified(file); 4922 if (err) 4923 return err; 4924 4925 f2fs_zero_post_eof_page(inode, 4926 iocb->ki_pos + iov_iter_count(from), true); 4927 return count; 4928 } 4929 4930 /* 4931 * Preallocate blocks for a write request, if it is possible and helpful to do 4932 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4933 * blocks were preallocated, or a negative errno value if something went 4934 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4935 * requested blocks (not just some of them) have been allocated. 4936 */ 4937 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4938 bool dio) 4939 { 4940 struct inode *inode = file_inode(iocb->ki_filp); 4941 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4942 const loff_t pos = iocb->ki_pos; 4943 const size_t count = iov_iter_count(iter); 4944 struct f2fs_map_blocks map = {}; 4945 int flag; 4946 int ret; 4947 4948 /* If it will be an out-of-place direct write, don't bother. */ 4949 if (dio && f2fs_lfs_mode(sbi)) 4950 return 0; 4951 /* 4952 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4953 * buffered IO, if DIO meets any holes. 4954 */ 4955 if (dio && i_size_read(inode) && 4956 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4957 return 0; 4958 4959 /* No-wait I/O can't allocate blocks. */ 4960 if (iocb->ki_flags & IOCB_NOWAIT) 4961 return 0; 4962 4963 /* If it will be a short write, don't bother. */ 4964 if (fault_in_iov_iter_readable(iter, count)) 4965 return 0; 4966 4967 if (f2fs_has_inline_data(inode)) { 4968 /* If the data will fit inline, don't bother. */ 4969 if (pos + count <= MAX_INLINE_DATA(inode)) 4970 return 0; 4971 ret = f2fs_convert_inline_inode(inode); 4972 if (ret) 4973 return ret; 4974 } 4975 4976 /* Do not preallocate blocks that will be written partially in 4KB. */ 4977 map.m_lblk = F2FS_BLK_ALIGN(pos); 4978 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4979 if (map.m_len > map.m_lblk) 4980 map.m_len -= map.m_lblk; 4981 else 4982 return 0; 4983 4984 if (!IS_DEVICE_ALIASING(inode)) 4985 map.m_may_create = true; 4986 if (dio) { 4987 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi, 4988 inode->i_write_hint); 4989 flag = F2FS_GET_BLOCK_PRE_DIO; 4990 } else { 4991 map.m_seg_type = NO_CHECK_TYPE; 4992 flag = F2FS_GET_BLOCK_PRE_AIO; 4993 } 4994 4995 ret = f2fs_map_blocks(inode, &map, flag); 4996 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4997 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4998 return ret; 4999 if (ret == 0) 5000 set_inode_flag(inode, FI_PREALLOCATED_ALL); 5001 return map.m_len; 5002 } 5003 5004 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 5005 struct iov_iter *from) 5006 { 5007 struct file *file = iocb->ki_filp; 5008 struct inode *inode = file_inode(file); 5009 ssize_t ret; 5010 5011 if (iocb->ki_flags & IOCB_NOWAIT) 5012 return -EOPNOTSUPP; 5013 5014 ret = generic_perform_write(iocb, from); 5015 5016 if (ret > 0) { 5017 f2fs_update_iostat(F2FS_I_SB(inode), inode, 5018 APP_BUFFERED_IO, ret); 5019 } 5020 return ret; 5021 } 5022 5023 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 5024 unsigned int flags) 5025 { 5026 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 5027 5028 dec_page_count(sbi, F2FS_DIO_WRITE); 5029 if (error) 5030 return error; 5031 f2fs_update_time(sbi, REQ_TIME); 5032 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 5033 return 0; 5034 } 5035 5036 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter, 5037 struct bio *bio, loff_t file_offset) 5038 { 5039 struct inode *inode = iter->inode; 5040 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 5041 enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint); 5042 enum temp_type temp = f2fs_get_segment_temp(sbi, type); 5043 5044 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp); 5045 submit_bio(bio); 5046 } 5047 5048 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 5049 .end_io = f2fs_dio_write_end_io, 5050 .submit_io = f2fs_dio_write_submit_io, 5051 }; 5052 5053 static void f2fs_flush_buffered_write(struct address_space *mapping, 5054 loff_t start_pos, loff_t end_pos) 5055 { 5056 int ret; 5057 5058 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 5059 if (ret < 0) 5060 return; 5061 invalidate_mapping_pages(mapping, 5062 start_pos >> PAGE_SHIFT, 5063 end_pos >> PAGE_SHIFT); 5064 } 5065 5066 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 5067 bool *may_need_sync) 5068 { 5069 struct file *file = iocb->ki_filp; 5070 struct inode *inode = file_inode(file); 5071 struct f2fs_inode_info *fi = F2FS_I(inode); 5072 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 5073 const bool do_opu = f2fs_lfs_mode(sbi); 5074 const loff_t pos = iocb->ki_pos; 5075 const ssize_t count = iov_iter_count(from); 5076 unsigned int dio_flags; 5077 struct iomap_dio *dio; 5078 ssize_t ret; 5079 5080 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 5081 5082 if (iocb->ki_flags & IOCB_NOWAIT) { 5083 /* f2fs_convert_inline_inode() and block allocation can block */ 5084 if (f2fs_has_inline_data(inode) || 5085 !f2fs_overwrite_io(inode, pos, count)) { 5086 ret = -EAGAIN; 5087 goto out; 5088 } 5089 5090 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 5091 ret = -EAGAIN; 5092 goto out; 5093 } 5094 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 5095 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5096 ret = -EAGAIN; 5097 goto out; 5098 } 5099 } else { 5100 ret = f2fs_convert_inline_inode(inode); 5101 if (ret) 5102 goto out; 5103 5104 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 5105 if (do_opu) 5106 f2fs_down_read(&fi->i_gc_rwsem[READ]); 5107 } 5108 5109 /* 5110 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 5111 * the higher-level function iomap_dio_rw() in order to ensure that the 5112 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 5113 */ 5114 inc_page_count(sbi, F2FS_DIO_WRITE); 5115 dio_flags = 0; 5116 if (pos + count > inode->i_size) 5117 dio_flags |= IOMAP_DIO_FORCE_WAIT; 5118 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 5119 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 5120 if (IS_ERR_OR_NULL(dio)) { 5121 ret = PTR_ERR_OR_ZERO(dio); 5122 if (ret == -ENOTBLK) 5123 ret = 0; 5124 if (ret != -EIOCBQUEUED) 5125 dec_page_count(sbi, F2FS_DIO_WRITE); 5126 } else { 5127 ret = iomap_dio_complete(dio); 5128 } 5129 5130 if (do_opu) 5131 f2fs_up_read(&fi->i_gc_rwsem[READ]); 5132 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5133 5134 if (ret < 0) 5135 goto out; 5136 if (pos + ret > inode->i_size) 5137 f2fs_i_size_write(inode, pos + ret); 5138 if (!do_opu) 5139 set_inode_flag(inode, FI_UPDATE_WRITE); 5140 5141 if (iov_iter_count(from)) { 5142 ssize_t ret2; 5143 loff_t bufio_start_pos = iocb->ki_pos; 5144 5145 /* 5146 * The direct write was partial, so we need to fall back to a 5147 * buffered write for the remainder. 5148 */ 5149 5150 ret2 = f2fs_buffered_write_iter(iocb, from); 5151 if (iov_iter_count(from)) 5152 f2fs_write_failed(inode, iocb->ki_pos); 5153 if (ret2 < 0) 5154 goto out; 5155 5156 /* 5157 * Ensure that the pagecache pages are written to disk and 5158 * invalidated to preserve the expected O_DIRECT semantics. 5159 */ 5160 if (ret2 > 0) { 5161 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 5162 5163 ret += ret2; 5164 5165 f2fs_flush_buffered_write(file->f_mapping, 5166 bufio_start_pos, 5167 bufio_end_pos); 5168 } 5169 } else { 5170 /* iomap_dio_rw() already handled the generic_write_sync(). */ 5171 *may_need_sync = false; 5172 } 5173 out: 5174 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 5175 return ret; 5176 } 5177 5178 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 5179 { 5180 struct inode *inode = file_inode(iocb->ki_filp); 5181 const loff_t orig_pos = iocb->ki_pos; 5182 const size_t orig_count = iov_iter_count(from); 5183 loff_t target_size; 5184 bool dio; 5185 bool may_need_sync = true; 5186 int preallocated; 5187 const loff_t pos = iocb->ki_pos; 5188 const ssize_t count = iov_iter_count(from); 5189 ssize_t ret; 5190 5191 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 5192 ret = -EIO; 5193 goto out; 5194 } 5195 5196 if (!f2fs_is_compress_backend_ready(inode)) { 5197 ret = -EOPNOTSUPP; 5198 goto out; 5199 } 5200 5201 if (iocb->ki_flags & IOCB_NOWAIT) { 5202 if (!inode_trylock(inode)) { 5203 ret = -EAGAIN; 5204 goto out; 5205 } 5206 } else { 5207 inode_lock(inode); 5208 } 5209 5210 if (f2fs_is_pinned_file(inode) && 5211 !f2fs_overwrite_io(inode, pos, count)) { 5212 ret = -EIO; 5213 goto out_unlock; 5214 } 5215 5216 ret = f2fs_write_checks(iocb, from); 5217 if (ret <= 0) 5218 goto out_unlock; 5219 5220 /* Determine whether we will do a direct write or a buffered write. */ 5221 dio = f2fs_should_use_dio(inode, iocb, from); 5222 5223 /* dio is not compatible w/ atomic write */ 5224 if (dio && f2fs_is_atomic_file(inode)) { 5225 ret = -EOPNOTSUPP; 5226 goto out_unlock; 5227 } 5228 5229 /* Possibly preallocate the blocks for the write. */ 5230 target_size = iocb->ki_pos + iov_iter_count(from); 5231 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 5232 if (preallocated < 0) { 5233 ret = preallocated; 5234 } else { 5235 if (trace_f2fs_datawrite_start_enabled()) 5236 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 5237 orig_count, WRITE); 5238 5239 /* Do the actual write. */ 5240 ret = dio ? 5241 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 5242 f2fs_buffered_write_iter(iocb, from); 5243 5244 trace_f2fs_datawrite_end(inode, orig_pos, ret); 5245 } 5246 5247 /* Don't leave any preallocated blocks around past i_size. */ 5248 if (preallocated && i_size_read(inode) < target_size) { 5249 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5250 filemap_invalidate_lock(inode->i_mapping); 5251 if (!f2fs_truncate(inode)) 5252 file_dont_truncate(inode); 5253 filemap_invalidate_unlock(inode->i_mapping); 5254 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5255 } else { 5256 file_dont_truncate(inode); 5257 } 5258 5259 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 5260 out_unlock: 5261 inode_unlock(inode); 5262 out: 5263 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 5264 5265 if (ret > 0 && may_need_sync) 5266 ret = generic_write_sync(iocb, ret); 5267 5268 /* If buffered IO was forced, flush and drop the data from 5269 * the page cache to preserve O_DIRECT semantics 5270 */ 5271 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 5272 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 5273 orig_pos, 5274 orig_pos + ret - 1); 5275 5276 return ret; 5277 } 5278 5279 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 5280 int advice) 5281 { 5282 struct address_space *mapping; 5283 struct backing_dev_info *bdi; 5284 struct inode *inode = file_inode(filp); 5285 int err; 5286 5287 if (advice == POSIX_FADV_SEQUENTIAL) { 5288 if (S_ISFIFO(inode->i_mode)) 5289 return -ESPIPE; 5290 5291 mapping = filp->f_mapping; 5292 if (!mapping || len < 0) 5293 return -EINVAL; 5294 5295 bdi = inode_to_bdi(mapping->host); 5296 filp->f_ra.ra_pages = bdi->ra_pages * 5297 F2FS_I_SB(inode)->seq_file_ra_mul; 5298 spin_lock(&filp->f_lock); 5299 filp->f_mode &= ~FMODE_RANDOM; 5300 spin_unlock(&filp->f_lock); 5301 return 0; 5302 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 5303 /* Load extent cache at the first readahead. */ 5304 f2fs_precache_extents(inode); 5305 } 5306 5307 err = generic_fadvise(filp, offset, len, advice); 5308 if (err) 5309 return err; 5310 5311 if (advice == POSIX_FADV_DONTNEED && 5312 (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 5313 f2fs_compressed_file(inode))) 5314 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 5315 else if (advice == POSIX_FADV_NOREUSE) 5316 err = f2fs_keep_noreuse_range(inode, offset, len); 5317 return err; 5318 } 5319 5320 #ifdef CONFIG_COMPAT 5321 struct compat_f2fs_gc_range { 5322 u32 sync; 5323 compat_u64 start; 5324 compat_u64 len; 5325 }; 5326 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 5327 struct compat_f2fs_gc_range) 5328 5329 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 5330 { 5331 struct compat_f2fs_gc_range __user *urange; 5332 struct f2fs_gc_range range; 5333 int err; 5334 5335 urange = compat_ptr(arg); 5336 err = get_user(range.sync, &urange->sync); 5337 err |= get_user(range.start, &urange->start); 5338 err |= get_user(range.len, &urange->len); 5339 if (err) 5340 return -EFAULT; 5341 5342 return __f2fs_ioc_gc_range(file, &range); 5343 } 5344 5345 struct compat_f2fs_move_range { 5346 u32 dst_fd; 5347 compat_u64 pos_in; 5348 compat_u64 pos_out; 5349 compat_u64 len; 5350 }; 5351 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 5352 struct compat_f2fs_move_range) 5353 5354 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 5355 { 5356 struct compat_f2fs_move_range __user *urange; 5357 struct f2fs_move_range range; 5358 int err; 5359 5360 urange = compat_ptr(arg); 5361 err = get_user(range.dst_fd, &urange->dst_fd); 5362 err |= get_user(range.pos_in, &urange->pos_in); 5363 err |= get_user(range.pos_out, &urange->pos_out); 5364 err |= get_user(range.len, &urange->len); 5365 if (err) 5366 return -EFAULT; 5367 5368 return __f2fs_ioc_move_range(file, &range); 5369 } 5370 5371 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5372 { 5373 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 5374 return -EIO; 5375 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 5376 return -ENOSPC; 5377 5378 switch (cmd) { 5379 case FS_IOC32_GETVERSION: 5380 cmd = FS_IOC_GETVERSION; 5381 break; 5382 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 5383 return f2fs_compat_ioc_gc_range(file, arg); 5384 case F2FS_IOC32_MOVE_RANGE: 5385 return f2fs_compat_ioc_move_range(file, arg); 5386 case F2FS_IOC_START_ATOMIC_WRITE: 5387 case F2FS_IOC_START_ATOMIC_REPLACE: 5388 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5389 case F2FS_IOC_START_VOLATILE_WRITE: 5390 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5391 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5392 case F2FS_IOC_SHUTDOWN: 5393 case FITRIM: 5394 case FS_IOC_SET_ENCRYPTION_POLICY: 5395 case FS_IOC_GET_ENCRYPTION_PWSALT: 5396 case FS_IOC_GET_ENCRYPTION_POLICY: 5397 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5398 case FS_IOC_ADD_ENCRYPTION_KEY: 5399 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5400 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5401 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5402 case FS_IOC_GET_ENCRYPTION_NONCE: 5403 case F2FS_IOC_GARBAGE_COLLECT: 5404 case F2FS_IOC_WRITE_CHECKPOINT: 5405 case F2FS_IOC_DEFRAGMENT: 5406 case F2FS_IOC_FLUSH_DEVICE: 5407 case F2FS_IOC_GET_FEATURES: 5408 case F2FS_IOC_GET_PIN_FILE: 5409 case F2FS_IOC_SET_PIN_FILE: 5410 case F2FS_IOC_PRECACHE_EXTENTS: 5411 case F2FS_IOC_RESIZE_FS: 5412 case FS_IOC_ENABLE_VERITY: 5413 case FS_IOC_MEASURE_VERITY: 5414 case FS_IOC_READ_VERITY_METADATA: 5415 case FS_IOC_GETFSLABEL: 5416 case FS_IOC_SETFSLABEL: 5417 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5418 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5419 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5420 case F2FS_IOC_SEC_TRIM_FILE: 5421 case F2FS_IOC_GET_COMPRESS_OPTION: 5422 case F2FS_IOC_SET_COMPRESS_OPTION: 5423 case F2FS_IOC_DECOMPRESS_FILE: 5424 case F2FS_IOC_COMPRESS_FILE: 5425 case F2FS_IOC_GET_DEV_ALIAS_FILE: 5426 case F2FS_IOC_IO_PRIO: 5427 break; 5428 default: 5429 return -ENOIOCTLCMD; 5430 } 5431 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5432 } 5433 #endif 5434 5435 const struct file_operations f2fs_file_operations = { 5436 .llseek = f2fs_llseek, 5437 .read_iter = f2fs_file_read_iter, 5438 .write_iter = f2fs_file_write_iter, 5439 .iopoll = iocb_bio_iopoll, 5440 .open = f2fs_file_open, 5441 .release = f2fs_release_file, 5442 .mmap_prepare = f2fs_file_mmap_prepare, 5443 .flush = f2fs_file_flush, 5444 .fsync = f2fs_sync_file, 5445 .fallocate = f2fs_fallocate, 5446 .unlocked_ioctl = f2fs_ioctl, 5447 #ifdef CONFIG_COMPAT 5448 .compat_ioctl = f2fs_compat_ioctl, 5449 #endif 5450 .splice_read = f2fs_file_splice_read, 5451 .splice_write = iter_file_splice_write, 5452 .fadvise = f2fs_file_fadvise, 5453 .fop_flags = FOP_BUFFER_RASYNC, 5454 }; 5455