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 vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 39 { 40 struct inode *inode = file_inode(vmf->vma->vm_file); 41 vm_flags_t flags = vmf->vma->vm_flags; 42 vm_fault_t ret; 43 44 ret = filemap_fault(vmf); 45 if (ret & VM_FAULT_LOCKED) 46 f2fs_update_iostat(F2FS_I_SB(inode), inode, 47 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 48 49 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret); 50 51 return ret; 52 } 53 54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 55 { 56 struct folio *folio = page_folio(vmf->page); 57 struct inode *inode = file_inode(vmf->vma->vm_file); 58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 59 struct dnode_of_data dn; 60 bool need_alloc = !f2fs_is_pinned_file(inode); 61 int err = 0; 62 vm_fault_t ret; 63 64 if (unlikely(IS_IMMUTABLE(inode))) 65 return VM_FAULT_SIGBUS; 66 67 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 68 err = -EIO; 69 goto out; 70 } 71 72 if (unlikely(f2fs_cp_error(sbi))) { 73 err = -EIO; 74 goto out; 75 } 76 77 if (!f2fs_is_checkpoint_ready(sbi)) { 78 err = -ENOSPC; 79 goto out; 80 } 81 82 err = f2fs_convert_inline_inode(inode); 83 if (err) 84 goto out; 85 86 #ifdef CONFIG_F2FS_FS_COMPRESSION 87 if (f2fs_compressed_file(inode)) { 88 int ret = f2fs_is_compressed_cluster(inode, folio->index); 89 90 if (ret < 0) { 91 err = ret; 92 goto out; 93 } else if (ret) { 94 need_alloc = false; 95 } 96 } 97 #endif 98 /* should do out of any locked page */ 99 if (need_alloc) 100 f2fs_balance_fs(sbi, true); 101 102 sb_start_pagefault(inode->i_sb); 103 104 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 105 106 file_update_time(vmf->vma->vm_file); 107 filemap_invalidate_lock_shared(inode->i_mapping); 108 folio_lock(folio); 109 if (unlikely(folio->mapping != inode->i_mapping || 110 folio_pos(folio) > i_size_read(inode) || 111 !folio_test_uptodate(folio))) { 112 folio_unlock(folio); 113 err = -EFAULT; 114 goto out_sem; 115 } 116 117 set_new_dnode(&dn, inode, NULL, NULL, 0); 118 if (need_alloc) { 119 /* block allocation */ 120 err = f2fs_get_block_locked(&dn, folio->index); 121 } else { 122 err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE); 123 f2fs_put_dnode(&dn); 124 if (f2fs_is_pinned_file(inode) && 125 !__is_valid_data_blkaddr(dn.data_blkaddr)) 126 err = -EIO; 127 } 128 129 if (err) { 130 folio_unlock(folio); 131 goto out_sem; 132 } 133 134 f2fs_folio_wait_writeback(folio, DATA, false, true); 135 136 /* wait for GCed page writeback via META_MAPPING */ 137 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 138 139 /* 140 * check to see if the page is mapped already (no holes) 141 */ 142 if (folio_test_mappedtodisk(folio)) 143 goto out_sem; 144 145 /* page is wholly or partially inside EOF */ 146 if (((loff_t)(folio->index + 1) << PAGE_SHIFT) > 147 i_size_read(inode)) { 148 loff_t offset; 149 150 offset = i_size_read(inode) & ~PAGE_MASK; 151 folio_zero_segment(folio, offset, folio_size(folio)); 152 } 153 folio_mark_dirty(folio); 154 155 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 156 f2fs_update_time(sbi, REQ_TIME); 157 158 out_sem: 159 filemap_invalidate_unlock_shared(inode->i_mapping); 160 161 sb_end_pagefault(inode->i_sb); 162 out: 163 ret = vmf_fs_error(err); 164 165 trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret); 166 return ret; 167 } 168 169 static const struct vm_operations_struct f2fs_file_vm_ops = { 170 .fault = f2fs_filemap_fault, 171 .map_pages = filemap_map_pages, 172 .page_mkwrite = f2fs_vm_page_mkwrite, 173 }; 174 175 static int get_parent_ino(struct inode *inode, nid_t *pino) 176 { 177 struct dentry *dentry; 178 179 /* 180 * Make sure to get the non-deleted alias. The alias associated with 181 * the open file descriptor being fsync()'ed may be deleted already. 182 */ 183 dentry = d_find_alias(inode); 184 if (!dentry) 185 return 0; 186 187 *pino = d_parent_ino(dentry); 188 dput(dentry); 189 return 1; 190 } 191 192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 193 { 194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 195 enum cp_reason_type cp_reason = CP_NO_NEEDED; 196 197 if (!S_ISREG(inode->i_mode)) 198 cp_reason = CP_NON_REGULAR; 199 else if (f2fs_compressed_file(inode)) 200 cp_reason = CP_COMPRESSED; 201 else if (inode->i_nlink != 1) 202 cp_reason = CP_HARDLINK; 203 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 204 cp_reason = CP_SB_NEED_CP; 205 else if (file_wrong_pino(inode)) 206 cp_reason = CP_WRONG_PINO; 207 else if (!f2fs_space_for_roll_forward(sbi)) 208 cp_reason = CP_NO_SPC_ROLL; 209 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 210 cp_reason = CP_NODE_NEED_CP; 211 else if (test_opt(sbi, FASTBOOT)) 212 cp_reason = CP_FASTBOOT_MODE; 213 else if (F2FS_OPTION(sbi).active_logs == 2) 214 cp_reason = CP_SPEC_LOG_NUM; 215 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 216 f2fs_need_dentry_mark(sbi, inode->i_ino) && 217 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 218 TRANS_DIR_INO)) 219 cp_reason = CP_RECOVER_DIR; 220 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 221 XATTR_DIR_INO)) 222 cp_reason = CP_XATTR_DIR; 223 224 return cp_reason; 225 } 226 227 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 228 { 229 struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino); 230 bool ret = false; 231 /* But we need to avoid that there are some inode updates */ 232 if ((!IS_ERR(i) && folio_test_dirty(i)) || 233 f2fs_need_inode_block_update(sbi, ino)) 234 ret = true; 235 f2fs_folio_put(i, false); 236 return ret; 237 } 238 239 static void try_to_fix_pino(struct inode *inode) 240 { 241 struct f2fs_inode_info *fi = F2FS_I(inode); 242 nid_t pino; 243 244 f2fs_down_write(&fi->i_sem); 245 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 246 get_parent_ino(inode, &pino)) { 247 f2fs_i_pino_write(inode, pino); 248 file_got_pino(inode); 249 } 250 f2fs_up_write(&fi->i_sem); 251 } 252 253 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 254 int datasync, bool atomic) 255 { 256 struct inode *inode = file->f_mapping->host; 257 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 258 nid_t ino = inode->i_ino; 259 int ret = 0; 260 enum cp_reason_type cp_reason = 0; 261 struct writeback_control wbc = { 262 .sync_mode = WB_SYNC_ALL, 263 .nr_to_write = LONG_MAX, 264 }; 265 unsigned int seq_id = 0; 266 267 if (unlikely(f2fs_readonly(inode->i_sb))) 268 return 0; 269 270 trace_f2fs_sync_file_enter(inode); 271 272 if (S_ISDIR(inode->i_mode)) 273 goto go_write; 274 275 /* if fdatasync is triggered, let's do in-place-update */ 276 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 277 set_inode_flag(inode, FI_NEED_IPU); 278 ret = file_write_and_wait_range(file, start, end); 279 clear_inode_flag(inode, FI_NEED_IPU); 280 281 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 282 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 283 return ret; 284 } 285 286 /* if the inode is dirty, let's recover all the time */ 287 if (!f2fs_skip_inode_update(inode, datasync)) { 288 f2fs_write_inode(inode, NULL); 289 goto go_write; 290 } 291 292 /* 293 * if there is no written data, don't waste time to write recovery info. 294 */ 295 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 296 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 297 298 /* it may call write_inode just prior to fsync */ 299 if (need_inode_page_update(sbi, ino)) 300 goto go_write; 301 302 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 303 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 304 goto flush_out; 305 goto out; 306 } else { 307 /* 308 * for OPU case, during fsync(), node can be persisted before 309 * data when lower device doesn't support write barrier, result 310 * in data corruption after SPO. 311 * So for strict fsync mode, force to use atomic write semantics 312 * to keep write order in between data/node and last node to 313 * avoid potential data corruption. 314 */ 315 if (F2FS_OPTION(sbi).fsync_mode == 316 FSYNC_MODE_STRICT && !atomic) 317 atomic = true; 318 } 319 go_write: 320 /* 321 * Both of fdatasync() and fsync() are able to be recovered from 322 * sudden-power-off. 323 */ 324 f2fs_down_read(&F2FS_I(inode)->i_sem); 325 cp_reason = need_do_checkpoint(inode); 326 f2fs_up_read(&F2FS_I(inode)->i_sem); 327 328 if (cp_reason) { 329 /* all the dirty node pages should be flushed for POR */ 330 ret = f2fs_sync_fs(inode->i_sb, 1); 331 332 /* 333 * We've secured consistency through sync_fs. Following pino 334 * will be used only for fsynced inodes after checkpoint. 335 */ 336 try_to_fix_pino(inode); 337 clear_inode_flag(inode, FI_APPEND_WRITE); 338 clear_inode_flag(inode, FI_UPDATE_WRITE); 339 goto out; 340 } 341 sync_nodes: 342 atomic_inc(&sbi->wb_sync_req[NODE]); 343 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 344 atomic_dec(&sbi->wb_sync_req[NODE]); 345 if (ret) 346 goto out; 347 348 /* if cp_error was enabled, we should avoid infinite loop */ 349 if (unlikely(f2fs_cp_error(sbi))) { 350 ret = -EIO; 351 goto out; 352 } 353 354 if (f2fs_need_inode_block_update(sbi, ino)) { 355 f2fs_mark_inode_dirty_sync(inode, true); 356 f2fs_write_inode(inode, NULL); 357 goto sync_nodes; 358 } 359 360 /* 361 * If it's atomic_write, it's just fine to keep write ordering. So 362 * here we don't need to wait for node write completion, since we use 363 * node chain which serializes node blocks. If one of node writes are 364 * reordered, we can see simply broken chain, resulting in stopping 365 * roll-forward recovery. It means we'll recover all or none node blocks 366 * given fsync mark. 367 */ 368 if (!atomic) { 369 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 370 if (ret) 371 goto out; 372 } 373 374 /* once recovery info is written, don't need to tack this */ 375 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 376 clear_inode_flag(inode, FI_APPEND_WRITE); 377 flush_out: 378 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 379 ret = f2fs_issue_flush(sbi, inode->i_ino); 380 if (!ret) { 381 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 382 clear_inode_flag(inode, FI_UPDATE_WRITE); 383 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 384 } 385 f2fs_update_time(sbi, REQ_TIME); 386 out: 387 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 388 return ret; 389 } 390 391 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 392 { 393 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 394 return -EIO; 395 return f2fs_do_sync_file(file, start, end, datasync, false); 396 } 397 398 static bool __found_offset(struct address_space *mapping, 399 struct dnode_of_data *dn, pgoff_t index, int whence) 400 { 401 block_t blkaddr = f2fs_data_blkaddr(dn); 402 struct inode *inode = mapping->host; 403 bool compressed_cluster = false; 404 405 if (f2fs_compressed_file(inode)) { 406 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio, 407 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size)); 408 409 compressed_cluster = first_blkaddr == COMPRESS_ADDR; 410 } 411 412 switch (whence) { 413 case SEEK_DATA: 414 if (__is_valid_data_blkaddr(blkaddr)) 415 return true; 416 if (blkaddr == NEW_ADDR && 417 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 418 return true; 419 if (compressed_cluster) 420 return true; 421 break; 422 case SEEK_HOLE: 423 if (compressed_cluster) 424 return false; 425 if (blkaddr == NULL_ADDR) 426 return true; 427 break; 428 } 429 return false; 430 } 431 432 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 433 { 434 struct inode *inode = file->f_mapping->host; 435 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 436 struct dnode_of_data dn; 437 pgoff_t pgofs, end_offset; 438 loff_t data_ofs = offset; 439 loff_t isize; 440 int err = 0; 441 442 inode_lock_shared(inode); 443 444 isize = i_size_read(inode); 445 if (offset >= isize) 446 goto fail; 447 448 /* handle inline data case */ 449 if (f2fs_has_inline_data(inode)) { 450 if (whence == SEEK_HOLE) { 451 data_ofs = isize; 452 goto found; 453 } else if (whence == SEEK_DATA) { 454 data_ofs = offset; 455 goto found; 456 } 457 } 458 459 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 460 461 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 462 set_new_dnode(&dn, inode, NULL, NULL, 0); 463 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 464 if (err && err != -ENOENT) { 465 goto fail; 466 } else if (err == -ENOENT) { 467 /* direct node does not exists */ 468 if (whence == SEEK_DATA) { 469 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 470 continue; 471 } else { 472 goto found; 473 } 474 } 475 476 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 477 478 /* find data/hole in dnode block */ 479 for (; dn.ofs_in_node < end_offset; 480 dn.ofs_in_node++, pgofs++, 481 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 482 block_t blkaddr; 483 484 blkaddr = f2fs_data_blkaddr(&dn); 485 486 if (__is_valid_data_blkaddr(blkaddr) && 487 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 488 blkaddr, DATA_GENERIC_ENHANCE)) { 489 f2fs_put_dnode(&dn); 490 goto fail; 491 } 492 493 if (__found_offset(file->f_mapping, &dn, 494 pgofs, whence)) { 495 f2fs_put_dnode(&dn); 496 goto found; 497 } 498 } 499 f2fs_put_dnode(&dn); 500 } 501 502 if (whence == SEEK_DATA) 503 goto fail; 504 found: 505 if (whence == SEEK_HOLE && data_ofs > isize) 506 data_ofs = isize; 507 inode_unlock_shared(inode); 508 return vfs_setpos(file, data_ofs, maxbytes); 509 fail: 510 inode_unlock_shared(inode); 511 return -ENXIO; 512 } 513 514 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 515 { 516 struct inode *inode = file->f_mapping->host; 517 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 518 519 switch (whence) { 520 case SEEK_SET: 521 case SEEK_CUR: 522 case SEEK_END: 523 return generic_file_llseek_size(file, offset, whence, 524 maxbytes, i_size_read(inode)); 525 case SEEK_DATA: 526 case SEEK_HOLE: 527 if (offset < 0) 528 return -ENXIO; 529 return f2fs_seek_block(file, offset, whence); 530 } 531 532 return -EINVAL; 533 } 534 535 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 536 { 537 struct inode *inode = file_inode(file); 538 539 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 540 return -EIO; 541 542 if (!f2fs_is_compress_backend_ready(inode)) 543 return -EOPNOTSUPP; 544 545 file_accessed(file); 546 vma->vm_ops = &f2fs_file_vm_ops; 547 548 f2fs_down_read(&F2FS_I(inode)->i_sem); 549 set_inode_flag(inode, FI_MMAP_FILE); 550 f2fs_up_read(&F2FS_I(inode)->i_sem); 551 552 return 0; 553 } 554 555 static int finish_preallocate_blocks(struct inode *inode) 556 { 557 int ret = 0; 558 bool opened; 559 560 f2fs_down_read(&F2FS_I(inode)->i_sem); 561 opened = is_inode_flag_set(inode, FI_OPENED_FILE); 562 f2fs_up_read(&F2FS_I(inode)->i_sem); 563 if (opened) 564 return 0; 565 566 inode_lock(inode); 567 if (is_inode_flag_set(inode, FI_OPENED_FILE)) 568 goto out_unlock; 569 570 if (!file_should_truncate(inode)) 571 goto out_update; 572 573 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 574 filemap_invalidate_lock(inode->i_mapping); 575 576 truncate_setsize(inode, i_size_read(inode)); 577 ret = f2fs_truncate(inode); 578 579 filemap_invalidate_unlock(inode->i_mapping); 580 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 581 if (ret) 582 goto out_unlock; 583 584 file_dont_truncate(inode); 585 out_update: 586 f2fs_down_write(&F2FS_I(inode)->i_sem); 587 set_inode_flag(inode, FI_OPENED_FILE); 588 f2fs_up_write(&F2FS_I(inode)->i_sem); 589 out_unlock: 590 inode_unlock(inode); 591 return ret; 592 } 593 594 static int f2fs_file_open(struct inode *inode, struct file *filp) 595 { 596 int err = fscrypt_file_open(inode, filp); 597 598 if (err) 599 return err; 600 601 if (!f2fs_is_compress_backend_ready(inode)) 602 return -EOPNOTSUPP; 603 604 err = fsverity_file_open(inode, filp); 605 if (err) 606 return err; 607 608 filp->f_mode |= FMODE_NOWAIT; 609 filp->f_mode |= FMODE_CAN_ODIRECT; 610 611 err = dquot_file_open(inode, filp); 612 if (err) 613 return err; 614 615 return finish_preallocate_blocks(inode); 616 } 617 618 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 619 { 620 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 621 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 622 __le32 *addr; 623 bool compressed_cluster = false; 624 int cluster_index = 0, valid_blocks = 0; 625 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 626 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 627 block_t blkstart; 628 int blklen = 0; 629 630 addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs; 631 blkstart = le32_to_cpu(*addr); 632 633 /* Assumption: truncation starts with cluster */ 634 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 635 block_t blkaddr = le32_to_cpu(*addr); 636 637 if (f2fs_compressed_file(dn->inode) && 638 !(cluster_index & (cluster_size - 1))) { 639 if (compressed_cluster) 640 f2fs_i_compr_blocks_update(dn->inode, 641 valid_blocks, false); 642 compressed_cluster = (blkaddr == COMPRESS_ADDR); 643 valid_blocks = 0; 644 } 645 646 if (blkaddr == NULL_ADDR) 647 goto next; 648 649 f2fs_set_data_blkaddr(dn, NULL_ADDR); 650 651 if (__is_valid_data_blkaddr(blkaddr)) { 652 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE)) 653 goto next; 654 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr, 655 DATA_GENERIC_ENHANCE)) 656 goto next; 657 if (compressed_cluster) 658 valid_blocks++; 659 } 660 661 if (blkstart + blklen == blkaddr) { 662 blklen++; 663 } else { 664 f2fs_invalidate_blocks(sbi, blkstart, blklen); 665 blkstart = blkaddr; 666 blklen = 1; 667 } 668 669 if (!released || blkaddr != COMPRESS_ADDR) 670 nr_free++; 671 672 continue; 673 674 next: 675 if (blklen) 676 f2fs_invalidate_blocks(sbi, blkstart, blklen); 677 678 blkstart = le32_to_cpu(*(addr + 1)); 679 blklen = 0; 680 } 681 682 if (blklen) 683 f2fs_invalidate_blocks(sbi, blkstart, blklen); 684 685 if (compressed_cluster) 686 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 687 688 if (nr_free) { 689 pgoff_t fofs; 690 /* 691 * once we invalidate valid blkaddr in range [ofs, ofs + count], 692 * we will invalidate all blkaddr in the whole range. 693 */ 694 fofs = f2fs_start_bidx_of_node(ofs_of_node(&dn->node_folio->page), 695 dn->inode) + ofs; 696 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 697 f2fs_update_age_extent_cache_range(dn, fofs, len); 698 dec_valid_block_count(sbi, dn->inode, nr_free); 699 } 700 dn->ofs_in_node = ofs; 701 702 f2fs_update_time(sbi, REQ_TIME); 703 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 704 dn->ofs_in_node, nr_free); 705 } 706 707 static int truncate_partial_data_page(struct inode *inode, u64 from, 708 bool cache_only) 709 { 710 loff_t offset = from & (PAGE_SIZE - 1); 711 pgoff_t index = from >> PAGE_SHIFT; 712 struct address_space *mapping = inode->i_mapping; 713 struct folio *folio; 714 715 if (!offset && !cache_only) 716 return 0; 717 718 if (cache_only) { 719 folio = filemap_lock_folio(mapping, index); 720 if (IS_ERR(folio)) 721 return 0; 722 if (folio_test_uptodate(folio)) 723 goto truncate_out; 724 f2fs_folio_put(folio, true); 725 return 0; 726 } 727 728 folio = f2fs_get_lock_data_folio(inode, index, true); 729 if (IS_ERR(folio)) 730 return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio); 731 truncate_out: 732 f2fs_folio_wait_writeback(folio, DATA, true, true); 733 folio_zero_segment(folio, offset, folio_size(folio)); 734 735 /* An encrypted inode should have a key and truncate the last page. */ 736 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 737 if (!cache_only) 738 folio_mark_dirty(folio); 739 f2fs_folio_put(folio, true); 740 return 0; 741 } 742 743 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 744 { 745 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 746 struct dnode_of_data dn; 747 pgoff_t free_from; 748 int count = 0, err = 0; 749 struct folio *ifolio; 750 bool truncate_page = false; 751 752 trace_f2fs_truncate_blocks_enter(inode, from); 753 754 if (IS_DEVICE_ALIASING(inode) && from) { 755 err = -EINVAL; 756 goto out_err; 757 } 758 759 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 760 761 if (free_from >= max_file_blocks(inode)) 762 goto free_partial; 763 764 if (lock) 765 f2fs_lock_op(sbi); 766 767 ifolio = f2fs_get_inode_folio(sbi, inode->i_ino); 768 if (IS_ERR(ifolio)) { 769 err = PTR_ERR(ifolio); 770 goto out; 771 } 772 773 if (IS_DEVICE_ALIASING(inode)) { 774 struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ]; 775 struct extent_info ei = et->largest; 776 777 f2fs_invalidate_blocks(sbi, ei.blk, ei.len); 778 779 dec_valid_block_count(sbi, inode, ei.len); 780 f2fs_update_time(sbi, REQ_TIME); 781 782 f2fs_folio_put(ifolio, true); 783 goto out; 784 } 785 786 if (f2fs_has_inline_data(inode)) { 787 f2fs_truncate_inline_inode(inode, ifolio, from); 788 f2fs_folio_put(ifolio, true); 789 truncate_page = true; 790 goto out; 791 } 792 793 set_new_dnode(&dn, inode, ifolio, NULL, 0); 794 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 795 if (err) { 796 if (err == -ENOENT) 797 goto free_next; 798 goto out; 799 } 800 801 count = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 802 803 count -= dn.ofs_in_node; 804 f2fs_bug_on(sbi, count < 0); 805 806 if (dn.ofs_in_node || IS_INODE(&dn.node_folio->page)) { 807 f2fs_truncate_data_blocks_range(&dn, count); 808 free_from += count; 809 } 810 811 f2fs_put_dnode(&dn); 812 free_next: 813 err = f2fs_truncate_inode_blocks(inode, free_from); 814 out: 815 if (lock) 816 f2fs_unlock_op(sbi); 817 free_partial: 818 /* lastly zero out the first data page */ 819 if (!err) 820 err = truncate_partial_data_page(inode, from, truncate_page); 821 out_err: 822 trace_f2fs_truncate_blocks_exit(inode, err); 823 return err; 824 } 825 826 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 827 { 828 u64 free_from = from; 829 int err; 830 831 #ifdef CONFIG_F2FS_FS_COMPRESSION 832 /* 833 * for compressed file, only support cluster size 834 * aligned truncation. 835 */ 836 if (f2fs_compressed_file(inode)) 837 free_from = round_up(from, 838 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 839 #endif 840 841 err = f2fs_do_truncate_blocks(inode, free_from, lock); 842 if (err) 843 return err; 844 845 #ifdef CONFIG_F2FS_FS_COMPRESSION 846 /* 847 * For compressed file, after release compress blocks, don't allow write 848 * direct, but we should allow write direct after truncate to zero. 849 */ 850 if (f2fs_compressed_file(inode) && !free_from 851 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 852 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 853 854 if (from != free_from) { 855 err = f2fs_truncate_partial_cluster(inode, from, lock); 856 if (err) 857 return err; 858 } 859 #endif 860 861 return 0; 862 } 863 864 int f2fs_truncate(struct inode *inode) 865 { 866 int err; 867 868 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 869 return -EIO; 870 871 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 872 S_ISLNK(inode->i_mode))) 873 return 0; 874 875 trace_f2fs_truncate(inode); 876 877 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 878 return -EIO; 879 880 err = f2fs_dquot_initialize(inode); 881 if (err) 882 return err; 883 884 /* we should check inline_data size */ 885 if (!f2fs_may_inline_data(inode)) { 886 err = f2fs_convert_inline_inode(inode); 887 if (err) 888 return err; 889 } 890 891 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 892 if (err) 893 return err; 894 895 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 896 f2fs_mark_inode_dirty_sync(inode, false); 897 return 0; 898 } 899 900 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 901 { 902 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 903 904 if (!fscrypt_dio_supported(inode)) 905 return true; 906 if (fsverity_active(inode)) 907 return true; 908 if (f2fs_compressed_file(inode)) 909 return true; 910 /* 911 * only force direct read to use buffered IO, for direct write, 912 * it expects inline data conversion before committing IO. 913 */ 914 if (f2fs_has_inline_data(inode) && rw == READ) 915 return true; 916 917 /* disallow direct IO if any of devices has unaligned blksize */ 918 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 919 return true; 920 /* 921 * for blkzoned device, fallback direct IO to buffered IO, so 922 * all IOs can be serialized by log-structured write. 923 */ 924 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) && 925 !f2fs_is_pinned_file(inode)) 926 return true; 927 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 928 return true; 929 930 return false; 931 } 932 933 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 934 struct kstat *stat, u32 request_mask, unsigned int query_flags) 935 { 936 struct inode *inode = d_inode(path->dentry); 937 struct f2fs_inode_info *fi = F2FS_I(inode); 938 struct f2fs_inode *ri = NULL; 939 unsigned int flags; 940 941 if (f2fs_has_extra_attr(inode) && 942 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 943 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 944 stat->result_mask |= STATX_BTIME; 945 stat->btime.tv_sec = fi->i_crtime.tv_sec; 946 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 947 } 948 949 /* 950 * Return the DIO alignment restrictions if requested. We only return 951 * this information when requested, since on encrypted files it might 952 * take a fair bit of work to get if the file wasn't opened recently. 953 * 954 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 955 * cannot represent that, so in that case we report no DIO support. 956 */ 957 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 958 unsigned int bsize = i_blocksize(inode); 959 960 stat->result_mask |= STATX_DIOALIGN; 961 if (!f2fs_force_buffered_io(inode, WRITE)) { 962 stat->dio_mem_align = bsize; 963 stat->dio_offset_align = bsize; 964 } 965 } 966 967 flags = fi->i_flags; 968 if (flags & F2FS_COMPR_FL) 969 stat->attributes |= STATX_ATTR_COMPRESSED; 970 if (flags & F2FS_APPEND_FL) 971 stat->attributes |= STATX_ATTR_APPEND; 972 if (IS_ENCRYPTED(inode)) 973 stat->attributes |= STATX_ATTR_ENCRYPTED; 974 if (flags & F2FS_IMMUTABLE_FL) 975 stat->attributes |= STATX_ATTR_IMMUTABLE; 976 if (flags & F2FS_NODUMP_FL) 977 stat->attributes |= STATX_ATTR_NODUMP; 978 if (IS_VERITY(inode)) 979 stat->attributes |= STATX_ATTR_VERITY; 980 981 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 982 STATX_ATTR_APPEND | 983 STATX_ATTR_ENCRYPTED | 984 STATX_ATTR_IMMUTABLE | 985 STATX_ATTR_NODUMP | 986 STATX_ATTR_VERITY); 987 988 generic_fillattr(idmap, request_mask, inode, stat); 989 990 /* we need to show initial sectors used for inline_data/dentries */ 991 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 992 f2fs_has_inline_dentry(inode)) 993 stat->blocks += (stat->size + 511) >> 9; 994 995 return 0; 996 } 997 998 #ifdef CONFIG_F2FS_FS_POSIX_ACL 999 static void __setattr_copy(struct mnt_idmap *idmap, 1000 struct inode *inode, const struct iattr *attr) 1001 { 1002 unsigned int ia_valid = attr->ia_valid; 1003 1004 i_uid_update(idmap, attr, inode); 1005 i_gid_update(idmap, attr, inode); 1006 if (ia_valid & ATTR_ATIME) 1007 inode_set_atime_to_ts(inode, attr->ia_atime); 1008 if (ia_valid & ATTR_MTIME) 1009 inode_set_mtime_to_ts(inode, attr->ia_mtime); 1010 if (ia_valid & ATTR_CTIME) 1011 inode_set_ctime_to_ts(inode, attr->ia_ctime); 1012 if (ia_valid & ATTR_MODE) { 1013 umode_t mode = attr->ia_mode; 1014 1015 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode))) 1016 mode &= ~S_ISGID; 1017 set_acl_inode(inode, mode); 1018 } 1019 } 1020 #else 1021 #define __setattr_copy setattr_copy 1022 #endif 1023 1024 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 1025 struct iattr *attr) 1026 { 1027 struct inode *inode = d_inode(dentry); 1028 struct f2fs_inode_info *fi = F2FS_I(inode); 1029 int err; 1030 1031 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1032 return -EIO; 1033 1034 if (unlikely(IS_IMMUTABLE(inode))) 1035 return -EPERM; 1036 1037 if (unlikely(IS_APPEND(inode) && 1038 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 1039 ATTR_GID | ATTR_TIMES_SET)))) 1040 return -EPERM; 1041 1042 if ((attr->ia_valid & ATTR_SIZE)) { 1043 if (!f2fs_is_compress_backend_ready(inode) || 1044 IS_DEVICE_ALIASING(inode)) 1045 return -EOPNOTSUPP; 1046 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 1047 !IS_ALIGNED(attr->ia_size, 1048 F2FS_BLK_TO_BYTES(fi->i_cluster_size))) 1049 return -EINVAL; 1050 } 1051 1052 err = setattr_prepare(idmap, dentry, attr); 1053 if (err) 1054 return err; 1055 1056 err = fscrypt_prepare_setattr(dentry, attr); 1057 if (err) 1058 return err; 1059 1060 err = fsverity_prepare_setattr(dentry, attr); 1061 if (err) 1062 return err; 1063 1064 if (is_quota_modification(idmap, inode, attr)) { 1065 err = f2fs_dquot_initialize(inode); 1066 if (err) 1067 return err; 1068 } 1069 if (i_uid_needs_update(idmap, attr, inode) || 1070 i_gid_needs_update(idmap, attr, inode)) { 1071 f2fs_lock_op(F2FS_I_SB(inode)); 1072 err = dquot_transfer(idmap, inode, attr); 1073 if (err) { 1074 set_sbi_flag(F2FS_I_SB(inode), 1075 SBI_QUOTA_NEED_REPAIR); 1076 f2fs_unlock_op(F2FS_I_SB(inode)); 1077 return err; 1078 } 1079 /* 1080 * update uid/gid under lock_op(), so that dquot and inode can 1081 * be updated atomically. 1082 */ 1083 i_uid_update(idmap, attr, inode); 1084 i_gid_update(idmap, attr, inode); 1085 f2fs_mark_inode_dirty_sync(inode, true); 1086 f2fs_unlock_op(F2FS_I_SB(inode)); 1087 } 1088 1089 if (attr->ia_valid & ATTR_SIZE) { 1090 loff_t old_size = i_size_read(inode); 1091 1092 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1093 /* 1094 * should convert inline inode before i_size_write to 1095 * keep smaller than inline_data size with inline flag. 1096 */ 1097 err = f2fs_convert_inline_inode(inode); 1098 if (err) 1099 return err; 1100 } 1101 1102 /* 1103 * wait for inflight dio, blocks should be removed after 1104 * IO completion. 1105 */ 1106 if (attr->ia_size < old_size) 1107 inode_dio_wait(inode); 1108 1109 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 1110 filemap_invalidate_lock(inode->i_mapping); 1111 1112 truncate_setsize(inode, attr->ia_size); 1113 1114 if (attr->ia_size <= old_size) 1115 err = f2fs_truncate(inode); 1116 /* 1117 * do not trim all blocks after i_size if target size is 1118 * larger than i_size. 1119 */ 1120 filemap_invalidate_unlock(inode->i_mapping); 1121 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 1122 if (err) 1123 return err; 1124 1125 spin_lock(&fi->i_size_lock); 1126 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1127 fi->last_disk_size = i_size_read(inode); 1128 spin_unlock(&fi->i_size_lock); 1129 } 1130 1131 __setattr_copy(idmap, inode, attr); 1132 1133 if (attr->ia_valid & ATTR_MODE) { 1134 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1135 1136 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1137 if (!err) 1138 inode->i_mode = fi->i_acl_mode; 1139 clear_inode_flag(inode, FI_ACL_MODE); 1140 } 1141 } 1142 1143 /* file size may changed here */ 1144 f2fs_mark_inode_dirty_sync(inode, true); 1145 1146 /* inode change will produce dirty node pages flushed by checkpoint */ 1147 f2fs_balance_fs(F2FS_I_SB(inode), true); 1148 1149 return err; 1150 } 1151 1152 const struct inode_operations f2fs_file_inode_operations = { 1153 .getattr = f2fs_getattr, 1154 .setattr = f2fs_setattr, 1155 .get_inode_acl = f2fs_get_acl, 1156 .set_acl = f2fs_set_acl, 1157 .listxattr = f2fs_listxattr, 1158 .fiemap = f2fs_fiemap, 1159 .fileattr_get = f2fs_fileattr_get, 1160 .fileattr_set = f2fs_fileattr_set, 1161 }; 1162 1163 static int fill_zero(struct inode *inode, pgoff_t index, 1164 loff_t start, loff_t len) 1165 { 1166 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1167 struct folio *folio; 1168 1169 if (!len) 1170 return 0; 1171 1172 f2fs_balance_fs(sbi, true); 1173 1174 f2fs_lock_op(sbi); 1175 folio = f2fs_get_new_data_folio(inode, NULL, index, false); 1176 f2fs_unlock_op(sbi); 1177 1178 if (IS_ERR(folio)) 1179 return PTR_ERR(folio); 1180 1181 f2fs_folio_wait_writeback(folio, DATA, true, true); 1182 folio_zero_range(folio, start, len); 1183 folio_mark_dirty(folio); 1184 f2fs_folio_put(folio, true); 1185 return 0; 1186 } 1187 1188 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1189 { 1190 int err; 1191 1192 while (pg_start < pg_end) { 1193 struct dnode_of_data dn; 1194 pgoff_t end_offset, count; 1195 1196 set_new_dnode(&dn, inode, NULL, NULL, 0); 1197 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1198 if (err) { 1199 if (err == -ENOENT) { 1200 pg_start = f2fs_get_next_page_offset(&dn, 1201 pg_start); 1202 continue; 1203 } 1204 return err; 1205 } 1206 1207 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 1208 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1209 1210 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1211 1212 f2fs_truncate_data_blocks_range(&dn, count); 1213 f2fs_put_dnode(&dn); 1214 1215 pg_start += count; 1216 } 1217 return 0; 1218 } 1219 1220 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1221 { 1222 pgoff_t pg_start, pg_end; 1223 loff_t off_start, off_end; 1224 int ret; 1225 1226 ret = f2fs_convert_inline_inode(inode); 1227 if (ret) 1228 return ret; 1229 1230 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1231 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1232 1233 off_start = offset & (PAGE_SIZE - 1); 1234 off_end = (offset + len) & (PAGE_SIZE - 1); 1235 1236 if (pg_start == pg_end) { 1237 ret = fill_zero(inode, pg_start, off_start, 1238 off_end - off_start); 1239 if (ret) 1240 return ret; 1241 } else { 1242 if (off_start) { 1243 ret = fill_zero(inode, pg_start++, off_start, 1244 PAGE_SIZE - off_start); 1245 if (ret) 1246 return ret; 1247 } 1248 if (off_end) { 1249 ret = fill_zero(inode, pg_end, 0, off_end); 1250 if (ret) 1251 return ret; 1252 } 1253 1254 if (pg_start < pg_end) { 1255 loff_t blk_start, blk_end; 1256 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1257 1258 f2fs_balance_fs(sbi, true); 1259 1260 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1261 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1262 1263 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1264 filemap_invalidate_lock(inode->i_mapping); 1265 1266 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1267 1268 f2fs_lock_op(sbi); 1269 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1270 f2fs_unlock_op(sbi); 1271 1272 filemap_invalidate_unlock(inode->i_mapping); 1273 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1274 } 1275 } 1276 1277 return ret; 1278 } 1279 1280 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1281 int *do_replace, pgoff_t off, pgoff_t len) 1282 { 1283 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1284 struct dnode_of_data dn; 1285 int ret, done, i; 1286 1287 next_dnode: 1288 set_new_dnode(&dn, inode, NULL, NULL, 0); 1289 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1290 if (ret && ret != -ENOENT) { 1291 return ret; 1292 } else if (ret == -ENOENT) { 1293 if (dn.max_level == 0) 1294 return -ENOENT; 1295 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1296 dn.ofs_in_node, len); 1297 blkaddr += done; 1298 do_replace += done; 1299 goto next; 1300 } 1301 1302 done = min((pgoff_t)ADDRS_PER_PAGE(&dn.node_folio->page, inode) - 1303 dn.ofs_in_node, len); 1304 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1305 *blkaddr = f2fs_data_blkaddr(&dn); 1306 1307 if (__is_valid_data_blkaddr(*blkaddr) && 1308 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1309 DATA_GENERIC_ENHANCE)) { 1310 f2fs_put_dnode(&dn); 1311 return -EFSCORRUPTED; 1312 } 1313 1314 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1315 1316 if (f2fs_lfs_mode(sbi)) { 1317 f2fs_put_dnode(&dn); 1318 return -EOPNOTSUPP; 1319 } 1320 1321 /* do not invalidate this block address */ 1322 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1323 *do_replace = 1; 1324 } 1325 } 1326 f2fs_put_dnode(&dn); 1327 next: 1328 len -= done; 1329 off += done; 1330 if (len) 1331 goto next_dnode; 1332 return 0; 1333 } 1334 1335 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1336 int *do_replace, pgoff_t off, int len) 1337 { 1338 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1339 struct dnode_of_data dn; 1340 int ret, i; 1341 1342 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1343 if (*do_replace == 0) 1344 continue; 1345 1346 set_new_dnode(&dn, inode, NULL, NULL, 0); 1347 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1348 if (ret) { 1349 dec_valid_block_count(sbi, inode, 1); 1350 f2fs_invalidate_blocks(sbi, *blkaddr, 1); 1351 } else { 1352 f2fs_update_data_blkaddr(&dn, *blkaddr); 1353 } 1354 f2fs_put_dnode(&dn); 1355 } 1356 return 0; 1357 } 1358 1359 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1360 block_t *blkaddr, int *do_replace, 1361 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1362 { 1363 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1364 pgoff_t i = 0; 1365 int ret; 1366 1367 while (i < len) { 1368 if (blkaddr[i] == NULL_ADDR && !full) { 1369 i++; 1370 continue; 1371 } 1372 1373 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1374 struct dnode_of_data dn; 1375 struct node_info ni; 1376 size_t new_size; 1377 pgoff_t ilen; 1378 1379 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1380 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1381 if (ret) 1382 return ret; 1383 1384 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1385 if (ret) { 1386 f2fs_put_dnode(&dn); 1387 return ret; 1388 } 1389 1390 ilen = min((pgoff_t) 1391 ADDRS_PER_PAGE(&dn.node_folio->page, dst_inode) - 1392 dn.ofs_in_node, len - i); 1393 do { 1394 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1395 f2fs_truncate_data_blocks_range(&dn, 1); 1396 1397 if (do_replace[i]) { 1398 f2fs_i_blocks_write(src_inode, 1399 1, false, false); 1400 f2fs_i_blocks_write(dst_inode, 1401 1, true, false); 1402 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1403 blkaddr[i], ni.version, true, false); 1404 1405 do_replace[i] = 0; 1406 } 1407 dn.ofs_in_node++; 1408 i++; 1409 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1410 if (dst_inode->i_size < new_size) 1411 f2fs_i_size_write(dst_inode, new_size); 1412 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1413 1414 f2fs_put_dnode(&dn); 1415 } else { 1416 struct folio *fsrc, *fdst; 1417 1418 fsrc = f2fs_get_lock_data_folio(src_inode, 1419 src + i, true); 1420 if (IS_ERR(fsrc)) 1421 return PTR_ERR(fsrc); 1422 fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i, 1423 true); 1424 if (IS_ERR(fdst)) { 1425 f2fs_folio_put(fsrc, true); 1426 return PTR_ERR(fdst); 1427 } 1428 1429 f2fs_folio_wait_writeback(fdst, DATA, true, true); 1430 1431 memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE); 1432 folio_mark_dirty(fdst); 1433 set_page_private_gcing(&fdst->page); 1434 f2fs_folio_put(fdst, true); 1435 f2fs_folio_put(fsrc, true); 1436 1437 ret = f2fs_truncate_hole(src_inode, 1438 src + i, src + i + 1); 1439 if (ret) 1440 return ret; 1441 i++; 1442 } 1443 } 1444 return 0; 1445 } 1446 1447 static int __exchange_data_block(struct inode *src_inode, 1448 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1449 pgoff_t len, bool full) 1450 { 1451 block_t *src_blkaddr; 1452 int *do_replace; 1453 pgoff_t olen; 1454 int ret; 1455 1456 while (len) { 1457 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1458 1459 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1460 array_size(olen, sizeof(block_t)), 1461 GFP_NOFS); 1462 if (!src_blkaddr) 1463 return -ENOMEM; 1464 1465 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1466 array_size(olen, sizeof(int)), 1467 GFP_NOFS); 1468 if (!do_replace) { 1469 kvfree(src_blkaddr); 1470 return -ENOMEM; 1471 } 1472 1473 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1474 do_replace, src, olen); 1475 if (ret) 1476 goto roll_back; 1477 1478 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1479 do_replace, src, dst, olen, full); 1480 if (ret) 1481 goto roll_back; 1482 1483 src += olen; 1484 dst += olen; 1485 len -= olen; 1486 1487 kvfree(src_blkaddr); 1488 kvfree(do_replace); 1489 } 1490 return 0; 1491 1492 roll_back: 1493 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1494 kvfree(src_blkaddr); 1495 kvfree(do_replace); 1496 return ret; 1497 } 1498 1499 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1500 { 1501 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1502 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1503 pgoff_t start = offset >> PAGE_SHIFT; 1504 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1505 int ret; 1506 1507 f2fs_balance_fs(sbi, true); 1508 1509 /* avoid gc operation during block exchange */ 1510 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1511 filemap_invalidate_lock(inode->i_mapping); 1512 1513 f2fs_lock_op(sbi); 1514 f2fs_drop_extent_tree(inode); 1515 truncate_pagecache(inode, offset); 1516 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1517 f2fs_unlock_op(sbi); 1518 1519 filemap_invalidate_unlock(inode->i_mapping); 1520 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1521 return ret; 1522 } 1523 1524 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1525 { 1526 loff_t new_size; 1527 int ret; 1528 1529 if (offset + len >= i_size_read(inode)) 1530 return -EINVAL; 1531 1532 /* collapse range should be aligned to block size of f2fs. */ 1533 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1534 return -EINVAL; 1535 1536 ret = f2fs_convert_inline_inode(inode); 1537 if (ret) 1538 return ret; 1539 1540 /* write out all dirty pages from offset */ 1541 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1542 if (ret) 1543 return ret; 1544 1545 ret = f2fs_do_collapse(inode, offset, len); 1546 if (ret) 1547 return ret; 1548 1549 /* write out all moved pages, if possible */ 1550 filemap_invalidate_lock(inode->i_mapping); 1551 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1552 truncate_pagecache(inode, offset); 1553 1554 new_size = i_size_read(inode) - len; 1555 ret = f2fs_truncate_blocks(inode, new_size, true); 1556 filemap_invalidate_unlock(inode->i_mapping); 1557 if (!ret) 1558 f2fs_i_size_write(inode, new_size); 1559 return ret; 1560 } 1561 1562 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1563 pgoff_t end) 1564 { 1565 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1566 pgoff_t index = start; 1567 unsigned int ofs_in_node = dn->ofs_in_node; 1568 blkcnt_t count = 0; 1569 int ret; 1570 1571 for (; index < end; index++, dn->ofs_in_node++) { 1572 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1573 count++; 1574 } 1575 1576 dn->ofs_in_node = ofs_in_node; 1577 ret = f2fs_reserve_new_blocks(dn, count); 1578 if (ret) 1579 return ret; 1580 1581 dn->ofs_in_node = ofs_in_node; 1582 for (index = start; index < end; index++, dn->ofs_in_node++) { 1583 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1584 /* 1585 * f2fs_reserve_new_blocks will not guarantee entire block 1586 * allocation. 1587 */ 1588 if (dn->data_blkaddr == NULL_ADDR) { 1589 ret = -ENOSPC; 1590 break; 1591 } 1592 1593 if (dn->data_blkaddr == NEW_ADDR) 1594 continue; 1595 1596 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1597 DATA_GENERIC_ENHANCE)) { 1598 ret = -EFSCORRUPTED; 1599 break; 1600 } 1601 1602 f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1); 1603 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1604 } 1605 1606 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1607 f2fs_update_age_extent_cache_range(dn, start, index - start); 1608 1609 return ret; 1610 } 1611 1612 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1613 int mode) 1614 { 1615 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1616 struct address_space *mapping = inode->i_mapping; 1617 pgoff_t index, pg_start, pg_end; 1618 loff_t new_size = i_size_read(inode); 1619 loff_t off_start, off_end; 1620 int ret = 0; 1621 1622 ret = inode_newsize_ok(inode, (len + offset)); 1623 if (ret) 1624 return ret; 1625 1626 ret = f2fs_convert_inline_inode(inode); 1627 if (ret) 1628 return ret; 1629 1630 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1631 if (ret) 1632 return ret; 1633 1634 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1635 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1636 1637 off_start = offset & (PAGE_SIZE - 1); 1638 off_end = (offset + len) & (PAGE_SIZE - 1); 1639 1640 if (pg_start == pg_end) { 1641 ret = fill_zero(inode, pg_start, off_start, 1642 off_end - off_start); 1643 if (ret) 1644 return ret; 1645 1646 new_size = max_t(loff_t, new_size, offset + len); 1647 } else { 1648 if (off_start) { 1649 ret = fill_zero(inode, pg_start++, off_start, 1650 PAGE_SIZE - off_start); 1651 if (ret) 1652 return ret; 1653 1654 new_size = max_t(loff_t, new_size, 1655 (loff_t)pg_start << PAGE_SHIFT); 1656 } 1657 1658 for (index = pg_start; index < pg_end;) { 1659 struct dnode_of_data dn; 1660 unsigned int end_offset; 1661 pgoff_t end; 1662 1663 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1664 filemap_invalidate_lock(mapping); 1665 1666 truncate_pagecache_range(inode, 1667 (loff_t)index << PAGE_SHIFT, 1668 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1669 1670 f2fs_lock_op(sbi); 1671 1672 set_new_dnode(&dn, inode, NULL, NULL, 0); 1673 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1674 if (ret) { 1675 f2fs_unlock_op(sbi); 1676 filemap_invalidate_unlock(mapping); 1677 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1678 goto out; 1679 } 1680 1681 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 1682 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1683 1684 ret = f2fs_do_zero_range(&dn, index, end); 1685 f2fs_put_dnode(&dn); 1686 1687 f2fs_unlock_op(sbi); 1688 filemap_invalidate_unlock(mapping); 1689 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1690 1691 f2fs_balance_fs(sbi, dn.node_changed); 1692 1693 if (ret) 1694 goto out; 1695 1696 index = end; 1697 new_size = max_t(loff_t, new_size, 1698 (loff_t)index << PAGE_SHIFT); 1699 } 1700 1701 if (off_end) { 1702 ret = fill_zero(inode, pg_end, 0, off_end); 1703 if (ret) 1704 goto out; 1705 1706 new_size = max_t(loff_t, new_size, offset + len); 1707 } 1708 } 1709 1710 out: 1711 if (new_size > i_size_read(inode)) { 1712 if (mode & FALLOC_FL_KEEP_SIZE) 1713 file_set_keep_isize(inode); 1714 else 1715 f2fs_i_size_write(inode, new_size); 1716 } 1717 return ret; 1718 } 1719 1720 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1721 { 1722 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1723 struct address_space *mapping = inode->i_mapping; 1724 pgoff_t nr, pg_start, pg_end, delta, idx; 1725 loff_t new_size; 1726 int ret = 0; 1727 1728 new_size = i_size_read(inode) + len; 1729 ret = inode_newsize_ok(inode, new_size); 1730 if (ret) 1731 return ret; 1732 1733 if (offset >= i_size_read(inode)) 1734 return -EINVAL; 1735 1736 /* insert range should be aligned to block size of f2fs. */ 1737 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1738 return -EINVAL; 1739 1740 ret = f2fs_convert_inline_inode(inode); 1741 if (ret) 1742 return ret; 1743 1744 f2fs_balance_fs(sbi, true); 1745 1746 filemap_invalidate_lock(mapping); 1747 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1748 filemap_invalidate_unlock(mapping); 1749 if (ret) 1750 return ret; 1751 1752 /* write out all dirty pages from offset */ 1753 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1754 if (ret) 1755 return ret; 1756 1757 pg_start = offset >> PAGE_SHIFT; 1758 pg_end = (offset + len) >> PAGE_SHIFT; 1759 delta = pg_end - pg_start; 1760 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1761 1762 /* avoid gc operation during block exchange */ 1763 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1764 filemap_invalidate_lock(mapping); 1765 truncate_pagecache(inode, offset); 1766 1767 while (!ret && idx > pg_start) { 1768 nr = idx - pg_start; 1769 if (nr > delta) 1770 nr = delta; 1771 idx -= nr; 1772 1773 f2fs_lock_op(sbi); 1774 f2fs_drop_extent_tree(inode); 1775 1776 ret = __exchange_data_block(inode, inode, idx, 1777 idx + delta, nr, false); 1778 f2fs_unlock_op(sbi); 1779 } 1780 filemap_invalidate_unlock(mapping); 1781 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1782 if (ret) 1783 return ret; 1784 1785 /* write out all moved pages, if possible */ 1786 filemap_invalidate_lock(mapping); 1787 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1788 truncate_pagecache(inode, offset); 1789 filemap_invalidate_unlock(mapping); 1790 1791 if (!ret) 1792 f2fs_i_size_write(inode, new_size); 1793 return ret; 1794 } 1795 1796 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1797 loff_t len, int mode) 1798 { 1799 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1800 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1801 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1802 .m_may_create = true }; 1803 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1804 .init_gc_type = FG_GC, 1805 .should_migrate_blocks = false, 1806 .err_gc_skipped = true, 1807 .nr_free_secs = 0 }; 1808 pgoff_t pg_start, pg_end; 1809 loff_t new_size; 1810 loff_t off_end; 1811 block_t expanded = 0; 1812 int err; 1813 1814 err = inode_newsize_ok(inode, (len + offset)); 1815 if (err) 1816 return err; 1817 1818 err = f2fs_convert_inline_inode(inode); 1819 if (err) 1820 return err; 1821 1822 f2fs_balance_fs(sbi, true); 1823 1824 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1825 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1826 off_end = (offset + len) & (PAGE_SIZE - 1); 1827 1828 map.m_lblk = pg_start; 1829 map.m_len = pg_end - pg_start; 1830 if (off_end) 1831 map.m_len++; 1832 1833 if (!map.m_len) 1834 return 0; 1835 1836 if (f2fs_is_pinned_file(inode)) { 1837 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1838 block_t sec_len = roundup(map.m_len, sec_blks); 1839 1840 map.m_len = sec_blks; 1841 next_alloc: 1842 f2fs_down_write(&sbi->pin_sem); 1843 1844 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1845 if (has_not_enough_free_secs(sbi, 0, 0)) { 1846 f2fs_up_write(&sbi->pin_sem); 1847 err = -ENOSPC; 1848 f2fs_warn_ratelimited(sbi, 1849 "ino:%lu, start:%lu, end:%lu, need to trigger GC to " 1850 "reclaim enough free segment when checkpoint is enabled", 1851 inode->i_ino, pg_start, pg_end); 1852 goto out_err; 1853 } 1854 } 1855 1856 if (has_not_enough_free_secs(sbi, 0, f2fs_sb_has_blkzoned(sbi) ? 1857 ZONED_PIN_SEC_REQUIRED_COUNT : 1858 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1859 f2fs_down_write(&sbi->gc_lock); 1860 stat_inc_gc_call_count(sbi, FOREGROUND); 1861 err = f2fs_gc(sbi, &gc_control); 1862 if (err && err != -ENODATA) { 1863 f2fs_up_write(&sbi->pin_sem); 1864 goto out_err; 1865 } 1866 } 1867 1868 err = f2fs_allocate_pinning_section(sbi); 1869 if (err) { 1870 f2fs_up_write(&sbi->pin_sem); 1871 goto out_err; 1872 } 1873 1874 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1875 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1876 file_dont_truncate(inode); 1877 1878 f2fs_up_write(&sbi->pin_sem); 1879 1880 expanded += map.m_len; 1881 sec_len -= map.m_len; 1882 map.m_lblk += map.m_len; 1883 if (!err && sec_len) 1884 goto next_alloc; 1885 1886 map.m_len = expanded; 1887 } else { 1888 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1889 expanded = map.m_len; 1890 } 1891 out_err: 1892 if (err) { 1893 pgoff_t last_off; 1894 1895 if (!expanded) 1896 return err; 1897 1898 last_off = pg_start + expanded - 1; 1899 1900 /* update new size to the failed position */ 1901 new_size = (last_off == pg_end) ? offset + len : 1902 (loff_t)(last_off + 1) << PAGE_SHIFT; 1903 } else { 1904 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1905 } 1906 1907 if (new_size > i_size_read(inode)) { 1908 if (mode & FALLOC_FL_KEEP_SIZE) 1909 file_set_keep_isize(inode); 1910 else 1911 f2fs_i_size_write(inode, new_size); 1912 } 1913 1914 return err; 1915 } 1916 1917 static long f2fs_fallocate(struct file *file, int mode, 1918 loff_t offset, loff_t len) 1919 { 1920 struct inode *inode = file_inode(file); 1921 long ret = 0; 1922 1923 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1924 return -EIO; 1925 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1926 return -ENOSPC; 1927 if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode)) 1928 return -EOPNOTSUPP; 1929 1930 /* f2fs only support ->fallocate for regular file */ 1931 if (!S_ISREG(inode->i_mode)) 1932 return -EINVAL; 1933 1934 if (IS_ENCRYPTED(inode) && 1935 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1936 return -EOPNOTSUPP; 1937 1938 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1939 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1940 FALLOC_FL_INSERT_RANGE)) 1941 return -EOPNOTSUPP; 1942 1943 inode_lock(inode); 1944 1945 /* 1946 * Pinned file should not support partial truncation since the block 1947 * can be used by applications. 1948 */ 1949 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1950 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1951 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 1952 ret = -EOPNOTSUPP; 1953 goto out; 1954 } 1955 1956 ret = file_modified(file); 1957 if (ret) 1958 goto out; 1959 1960 /* 1961 * wait for inflight dio, blocks should be removed after IO 1962 * completion. 1963 */ 1964 inode_dio_wait(inode); 1965 1966 if (mode & FALLOC_FL_PUNCH_HOLE) { 1967 if (offset >= inode->i_size) 1968 goto out; 1969 1970 ret = f2fs_punch_hole(inode, offset, len); 1971 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1972 ret = f2fs_collapse_range(inode, offset, len); 1973 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1974 ret = f2fs_zero_range(inode, offset, len, mode); 1975 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1976 ret = f2fs_insert_range(inode, offset, len); 1977 } else { 1978 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1979 } 1980 1981 if (!ret) { 1982 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1983 f2fs_mark_inode_dirty_sync(inode, false); 1984 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1985 } 1986 1987 out: 1988 inode_unlock(inode); 1989 1990 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1991 return ret; 1992 } 1993 1994 static int f2fs_release_file(struct inode *inode, struct file *filp) 1995 { 1996 /* 1997 * f2fs_release_file is called at every close calls. So we should 1998 * not drop any inmemory pages by close called by other process. 1999 */ 2000 if (!(filp->f_mode & FMODE_WRITE) || 2001 atomic_read(&inode->i_writecount) != 1) 2002 return 0; 2003 2004 inode_lock(inode); 2005 f2fs_abort_atomic_write(inode, true); 2006 inode_unlock(inode); 2007 2008 return 0; 2009 } 2010 2011 static int f2fs_file_flush(struct file *file, fl_owner_t id) 2012 { 2013 struct inode *inode = file_inode(file); 2014 2015 /* 2016 * If the process doing a transaction is crashed, we should do 2017 * roll-back. Otherwise, other reader/write can see corrupted database 2018 * until all the writers close its file. Since this should be done 2019 * before dropping file lock, it needs to do in ->flush. 2020 */ 2021 if (F2FS_I(inode)->atomic_write_task == current && 2022 (current->flags & PF_EXITING)) { 2023 inode_lock(inode); 2024 f2fs_abort_atomic_write(inode, true); 2025 inode_unlock(inode); 2026 } 2027 2028 return 0; 2029 } 2030 2031 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 2032 { 2033 struct f2fs_inode_info *fi = F2FS_I(inode); 2034 u32 masked_flags = fi->i_flags & mask; 2035 2036 /* mask can be shrunk by flags_valid selector */ 2037 iflags &= mask; 2038 2039 /* Is it quota file? Do not allow user to mess with it */ 2040 if (IS_NOQUOTA(inode)) 2041 return -EPERM; 2042 2043 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 2044 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 2045 return -EOPNOTSUPP; 2046 if (!f2fs_empty_dir(inode)) 2047 return -ENOTEMPTY; 2048 } 2049 2050 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 2051 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 2052 return -EOPNOTSUPP; 2053 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 2054 return -EINVAL; 2055 } 2056 2057 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 2058 if (masked_flags & F2FS_COMPR_FL) { 2059 if (!f2fs_disable_compressed_file(inode)) 2060 return -EINVAL; 2061 } else { 2062 /* try to convert inline_data to support compression */ 2063 int err = f2fs_convert_inline_inode(inode); 2064 if (err) 2065 return err; 2066 2067 f2fs_down_write(&fi->i_sem); 2068 if (!f2fs_may_compress(inode) || 2069 (S_ISREG(inode->i_mode) && 2070 F2FS_HAS_BLOCKS(inode))) { 2071 f2fs_up_write(&fi->i_sem); 2072 return -EINVAL; 2073 } 2074 err = set_compress_context(inode); 2075 f2fs_up_write(&fi->i_sem); 2076 2077 if (err) 2078 return err; 2079 } 2080 } 2081 2082 fi->i_flags = iflags | (fi->i_flags & ~mask); 2083 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 2084 (fi->i_flags & F2FS_NOCOMP_FL)); 2085 2086 if (fi->i_flags & F2FS_PROJINHERIT_FL) 2087 set_inode_flag(inode, FI_PROJ_INHERIT); 2088 else 2089 clear_inode_flag(inode, FI_PROJ_INHERIT); 2090 2091 inode_set_ctime_current(inode); 2092 f2fs_set_inode_flags(inode); 2093 f2fs_mark_inode_dirty_sync(inode, true); 2094 return 0; 2095 } 2096 2097 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 2098 2099 /* 2100 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 2101 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 2102 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 2103 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 2104 * 2105 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 2106 * FS_IOC_FSSETXATTR is done by the VFS. 2107 */ 2108 2109 static const struct { 2110 u32 iflag; 2111 u32 fsflag; 2112 } f2fs_fsflags_map[] = { 2113 { F2FS_COMPR_FL, FS_COMPR_FL }, 2114 { F2FS_SYNC_FL, FS_SYNC_FL }, 2115 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 2116 { F2FS_APPEND_FL, FS_APPEND_FL }, 2117 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 2118 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 2119 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 2120 { F2FS_INDEX_FL, FS_INDEX_FL }, 2121 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 2122 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2123 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2124 }; 2125 2126 #define F2FS_GETTABLE_FS_FL ( \ 2127 FS_COMPR_FL | \ 2128 FS_SYNC_FL | \ 2129 FS_IMMUTABLE_FL | \ 2130 FS_APPEND_FL | \ 2131 FS_NODUMP_FL | \ 2132 FS_NOATIME_FL | \ 2133 FS_NOCOMP_FL | \ 2134 FS_INDEX_FL | \ 2135 FS_DIRSYNC_FL | \ 2136 FS_PROJINHERIT_FL | \ 2137 FS_ENCRYPT_FL | \ 2138 FS_INLINE_DATA_FL | \ 2139 FS_NOCOW_FL | \ 2140 FS_VERITY_FL | \ 2141 FS_CASEFOLD_FL) 2142 2143 #define F2FS_SETTABLE_FS_FL ( \ 2144 FS_COMPR_FL | \ 2145 FS_SYNC_FL | \ 2146 FS_IMMUTABLE_FL | \ 2147 FS_APPEND_FL | \ 2148 FS_NODUMP_FL | \ 2149 FS_NOATIME_FL | \ 2150 FS_NOCOMP_FL | \ 2151 FS_DIRSYNC_FL | \ 2152 FS_PROJINHERIT_FL | \ 2153 FS_CASEFOLD_FL) 2154 2155 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2156 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2157 { 2158 u32 fsflags = 0; 2159 int i; 2160 2161 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2162 if (iflags & f2fs_fsflags_map[i].iflag) 2163 fsflags |= f2fs_fsflags_map[i].fsflag; 2164 2165 return fsflags; 2166 } 2167 2168 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2169 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2170 { 2171 u32 iflags = 0; 2172 int i; 2173 2174 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2175 if (fsflags & f2fs_fsflags_map[i].fsflag) 2176 iflags |= f2fs_fsflags_map[i].iflag; 2177 2178 return iflags; 2179 } 2180 2181 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2182 { 2183 struct inode *inode = file_inode(filp); 2184 2185 return put_user(inode->i_generation, (int __user *)arg); 2186 } 2187 2188 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2189 { 2190 struct inode *inode = file_inode(filp); 2191 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2192 struct f2fs_inode_info *fi = F2FS_I(inode); 2193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2194 loff_t isize; 2195 int ret; 2196 2197 if (!(filp->f_mode & FMODE_WRITE)) 2198 return -EBADF; 2199 2200 if (!inode_owner_or_capable(idmap, inode)) 2201 return -EACCES; 2202 2203 if (!S_ISREG(inode->i_mode)) 2204 return -EINVAL; 2205 2206 if (filp->f_flags & O_DIRECT) 2207 return -EINVAL; 2208 2209 ret = mnt_want_write_file(filp); 2210 if (ret) 2211 return ret; 2212 2213 inode_lock(inode); 2214 2215 if (!f2fs_disable_compressed_file(inode) || 2216 f2fs_is_pinned_file(inode)) { 2217 ret = -EINVAL; 2218 goto out; 2219 } 2220 2221 if (f2fs_is_atomic_file(inode)) 2222 goto out; 2223 2224 ret = f2fs_convert_inline_inode(inode); 2225 if (ret) 2226 goto out; 2227 2228 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2229 f2fs_down_write(&fi->i_gc_rwsem[READ]); 2230 2231 /* 2232 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2233 * f2fs_is_atomic_file. 2234 */ 2235 if (get_dirty_pages(inode)) 2236 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2237 inode->i_ino, get_dirty_pages(inode)); 2238 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2239 if (ret) 2240 goto out_unlock; 2241 2242 /* Check if the inode already has a COW inode */ 2243 if (fi->cow_inode == NULL) { 2244 /* Create a COW inode for atomic write */ 2245 struct dentry *dentry = file_dentry(filp); 2246 struct inode *dir = d_inode(dentry->d_parent); 2247 2248 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode); 2249 if (ret) 2250 goto out_unlock; 2251 2252 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2253 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2254 2255 /* Set the COW inode's atomic_inode to the atomic inode */ 2256 F2FS_I(fi->cow_inode)->atomic_inode = inode; 2257 } else { 2258 /* Reuse the already created COW inode */ 2259 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode)); 2260 2261 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1); 2262 2263 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2264 if (ret) 2265 goto out_unlock; 2266 } 2267 2268 f2fs_write_inode(inode, NULL); 2269 2270 stat_inc_atomic_inode(inode); 2271 2272 set_inode_flag(inode, FI_ATOMIC_FILE); 2273 2274 isize = i_size_read(inode); 2275 fi->original_i_size = isize; 2276 if (truncate) { 2277 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2278 truncate_inode_pages_final(inode->i_mapping); 2279 f2fs_i_size_write(inode, 0); 2280 isize = 0; 2281 } 2282 f2fs_i_size_write(fi->cow_inode, isize); 2283 2284 out_unlock: 2285 f2fs_up_write(&fi->i_gc_rwsem[READ]); 2286 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2287 if (ret) 2288 goto out; 2289 2290 f2fs_update_time(sbi, REQ_TIME); 2291 fi->atomic_write_task = current; 2292 stat_update_max_atomic_write(inode); 2293 fi->atomic_write_cnt = 0; 2294 out: 2295 inode_unlock(inode); 2296 mnt_drop_write_file(filp); 2297 return ret; 2298 } 2299 2300 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2301 { 2302 struct inode *inode = file_inode(filp); 2303 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2304 int ret; 2305 2306 if (!(filp->f_mode & FMODE_WRITE)) 2307 return -EBADF; 2308 2309 if (!inode_owner_or_capable(idmap, inode)) 2310 return -EACCES; 2311 2312 ret = mnt_want_write_file(filp); 2313 if (ret) 2314 return ret; 2315 2316 f2fs_balance_fs(F2FS_I_SB(inode), true); 2317 2318 inode_lock(inode); 2319 2320 if (f2fs_is_atomic_file(inode)) { 2321 ret = f2fs_commit_atomic_write(inode); 2322 if (!ret) 2323 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2324 2325 f2fs_abort_atomic_write(inode, ret); 2326 } else { 2327 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2328 } 2329 2330 inode_unlock(inode); 2331 mnt_drop_write_file(filp); 2332 return ret; 2333 } 2334 2335 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2336 { 2337 struct inode *inode = file_inode(filp); 2338 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2339 int ret; 2340 2341 if (!(filp->f_mode & FMODE_WRITE)) 2342 return -EBADF; 2343 2344 if (!inode_owner_or_capable(idmap, inode)) 2345 return -EACCES; 2346 2347 ret = mnt_want_write_file(filp); 2348 if (ret) 2349 return ret; 2350 2351 inode_lock(inode); 2352 2353 f2fs_abort_atomic_write(inode, true); 2354 2355 inode_unlock(inode); 2356 2357 mnt_drop_write_file(filp); 2358 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2359 return ret; 2360 } 2361 2362 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag, 2363 bool readonly, bool need_lock) 2364 { 2365 struct super_block *sb = sbi->sb; 2366 int ret = 0; 2367 2368 switch (flag) { 2369 case F2FS_GOING_DOWN_FULLSYNC: 2370 ret = bdev_freeze(sb->s_bdev); 2371 if (ret) 2372 goto out; 2373 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2374 bdev_thaw(sb->s_bdev); 2375 break; 2376 case F2FS_GOING_DOWN_METASYNC: 2377 /* do checkpoint only */ 2378 ret = f2fs_sync_fs(sb, 1); 2379 if (ret) { 2380 if (ret == -EIO) 2381 ret = 0; 2382 goto out; 2383 } 2384 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2385 break; 2386 case F2FS_GOING_DOWN_NOSYNC: 2387 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2388 break; 2389 case F2FS_GOING_DOWN_METAFLUSH: 2390 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2391 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2392 break; 2393 case F2FS_GOING_DOWN_NEED_FSCK: 2394 set_sbi_flag(sbi, SBI_NEED_FSCK); 2395 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2396 set_sbi_flag(sbi, SBI_IS_DIRTY); 2397 /* do checkpoint only */ 2398 ret = f2fs_sync_fs(sb, 1); 2399 if (ret == -EIO) 2400 ret = 0; 2401 goto out; 2402 default: 2403 ret = -EINVAL; 2404 goto out; 2405 } 2406 2407 if (readonly) 2408 goto out; 2409 2410 /* 2411 * grab sb->s_umount to avoid racing w/ remount() and other shutdown 2412 * paths. 2413 */ 2414 if (need_lock) 2415 down_write(&sbi->sb->s_umount); 2416 2417 f2fs_stop_gc_thread(sbi); 2418 f2fs_stop_discard_thread(sbi); 2419 2420 f2fs_drop_discard_cmd(sbi); 2421 clear_opt(sbi, DISCARD); 2422 2423 if (need_lock) 2424 up_write(&sbi->sb->s_umount); 2425 2426 f2fs_update_time(sbi, REQ_TIME); 2427 out: 2428 2429 trace_f2fs_shutdown(sbi, flag, ret); 2430 2431 return ret; 2432 } 2433 2434 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2435 { 2436 struct inode *inode = file_inode(filp); 2437 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2438 __u32 in; 2439 int ret; 2440 bool need_drop = false, readonly = false; 2441 2442 if (!capable(CAP_SYS_ADMIN)) 2443 return -EPERM; 2444 2445 if (get_user(in, (__u32 __user *)arg)) 2446 return -EFAULT; 2447 2448 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2449 ret = mnt_want_write_file(filp); 2450 if (ret) { 2451 if (ret != -EROFS) 2452 return ret; 2453 2454 /* fallback to nosync shutdown for readonly fs */ 2455 in = F2FS_GOING_DOWN_NOSYNC; 2456 readonly = true; 2457 } else { 2458 need_drop = true; 2459 } 2460 } 2461 2462 ret = f2fs_do_shutdown(sbi, in, readonly, true); 2463 2464 if (need_drop) 2465 mnt_drop_write_file(filp); 2466 2467 return ret; 2468 } 2469 2470 static int f2fs_keep_noreuse_range(struct inode *inode, 2471 loff_t offset, loff_t len) 2472 { 2473 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2474 u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 2475 u64 start, end; 2476 int ret = 0; 2477 2478 if (!S_ISREG(inode->i_mode)) 2479 return 0; 2480 2481 if (offset >= max_bytes || len > max_bytes || 2482 (offset + len) > max_bytes) 2483 return 0; 2484 2485 start = offset >> PAGE_SHIFT; 2486 end = DIV_ROUND_UP(offset + len, PAGE_SIZE); 2487 2488 inode_lock(inode); 2489 if (f2fs_is_atomic_file(inode)) { 2490 inode_unlock(inode); 2491 return 0; 2492 } 2493 2494 spin_lock(&sbi->inode_lock[DONATE_INODE]); 2495 /* let's remove the range, if len = 0 */ 2496 if (!len) { 2497 if (!list_empty(&F2FS_I(inode)->gdonate_list)) { 2498 list_del_init(&F2FS_I(inode)->gdonate_list); 2499 sbi->donate_files--; 2500 if (is_inode_flag_set(inode, FI_DONATE_FINISHED)) 2501 ret = -EALREADY; 2502 else 2503 set_inode_flag(inode, FI_DONATE_FINISHED); 2504 } else 2505 ret = -ENOENT; 2506 } else { 2507 if (list_empty(&F2FS_I(inode)->gdonate_list)) { 2508 list_add_tail(&F2FS_I(inode)->gdonate_list, 2509 &sbi->inode_list[DONATE_INODE]); 2510 sbi->donate_files++; 2511 } else { 2512 list_move_tail(&F2FS_I(inode)->gdonate_list, 2513 &sbi->inode_list[DONATE_INODE]); 2514 } 2515 F2FS_I(inode)->donate_start = start; 2516 F2FS_I(inode)->donate_end = end - 1; 2517 clear_inode_flag(inode, FI_DONATE_FINISHED); 2518 } 2519 spin_unlock(&sbi->inode_lock[DONATE_INODE]); 2520 inode_unlock(inode); 2521 2522 return ret; 2523 } 2524 2525 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2526 { 2527 struct inode *inode = file_inode(filp); 2528 struct super_block *sb = inode->i_sb; 2529 struct fstrim_range range; 2530 int ret; 2531 2532 if (!capable(CAP_SYS_ADMIN)) 2533 return -EPERM; 2534 2535 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2536 return -EOPNOTSUPP; 2537 2538 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2539 sizeof(range))) 2540 return -EFAULT; 2541 2542 ret = mnt_want_write_file(filp); 2543 if (ret) 2544 return ret; 2545 2546 range.minlen = max((unsigned int)range.minlen, 2547 bdev_discard_granularity(sb->s_bdev)); 2548 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2549 mnt_drop_write_file(filp); 2550 if (ret < 0) 2551 return ret; 2552 2553 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2554 sizeof(range))) 2555 return -EFAULT; 2556 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2557 return 0; 2558 } 2559 2560 static bool uuid_is_nonzero(__u8 u[16]) 2561 { 2562 int i; 2563 2564 for (i = 0; i < 16; i++) 2565 if (u[i]) 2566 return true; 2567 return false; 2568 } 2569 2570 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2571 { 2572 struct inode *inode = file_inode(filp); 2573 int ret; 2574 2575 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2576 return -EOPNOTSUPP; 2577 2578 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2579 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2580 return ret; 2581 } 2582 2583 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2584 { 2585 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2586 return -EOPNOTSUPP; 2587 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2588 } 2589 2590 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2591 { 2592 struct inode *inode = file_inode(filp); 2593 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2594 u8 encrypt_pw_salt[16]; 2595 int err; 2596 2597 if (!f2fs_sb_has_encrypt(sbi)) 2598 return -EOPNOTSUPP; 2599 2600 err = mnt_want_write_file(filp); 2601 if (err) 2602 return err; 2603 2604 f2fs_down_write(&sbi->sb_lock); 2605 2606 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2607 goto got_it; 2608 2609 /* update superblock with uuid */ 2610 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2611 2612 err = f2fs_commit_super(sbi, false); 2613 if (err) { 2614 /* undo new data */ 2615 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2616 goto out_err; 2617 } 2618 got_it: 2619 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2620 out_err: 2621 f2fs_up_write(&sbi->sb_lock); 2622 mnt_drop_write_file(filp); 2623 2624 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2625 err = -EFAULT; 2626 2627 return err; 2628 } 2629 2630 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2631 unsigned long arg) 2632 { 2633 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2634 return -EOPNOTSUPP; 2635 2636 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2637 } 2638 2639 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2640 { 2641 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2642 return -EOPNOTSUPP; 2643 2644 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2645 } 2646 2647 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2648 { 2649 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2650 return -EOPNOTSUPP; 2651 2652 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2653 } 2654 2655 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2656 unsigned long arg) 2657 { 2658 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2659 return -EOPNOTSUPP; 2660 2661 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2662 } 2663 2664 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2665 unsigned long arg) 2666 { 2667 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2668 return -EOPNOTSUPP; 2669 2670 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2671 } 2672 2673 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2674 { 2675 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2676 return -EOPNOTSUPP; 2677 2678 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2679 } 2680 2681 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2682 { 2683 struct inode *inode = file_inode(filp); 2684 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2685 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2686 .no_bg_gc = false, 2687 .should_migrate_blocks = false, 2688 .nr_free_secs = 0 }; 2689 __u32 sync; 2690 int ret; 2691 2692 if (!capable(CAP_SYS_ADMIN)) 2693 return -EPERM; 2694 2695 if (get_user(sync, (__u32 __user *)arg)) 2696 return -EFAULT; 2697 2698 if (f2fs_readonly(sbi->sb)) 2699 return -EROFS; 2700 2701 ret = mnt_want_write_file(filp); 2702 if (ret) 2703 return ret; 2704 2705 if (!sync) { 2706 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2707 ret = -EBUSY; 2708 goto out; 2709 } 2710 } else { 2711 f2fs_down_write(&sbi->gc_lock); 2712 } 2713 2714 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2715 gc_control.err_gc_skipped = sync; 2716 stat_inc_gc_call_count(sbi, FOREGROUND); 2717 ret = f2fs_gc(sbi, &gc_control); 2718 out: 2719 mnt_drop_write_file(filp); 2720 return ret; 2721 } 2722 2723 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2724 { 2725 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2726 struct f2fs_gc_control gc_control = { 2727 .init_gc_type = range->sync ? FG_GC : BG_GC, 2728 .no_bg_gc = false, 2729 .should_migrate_blocks = false, 2730 .err_gc_skipped = range->sync, 2731 .nr_free_secs = 0 }; 2732 u64 end; 2733 int ret; 2734 2735 if (!capable(CAP_SYS_ADMIN)) 2736 return -EPERM; 2737 if (f2fs_readonly(sbi->sb)) 2738 return -EROFS; 2739 2740 end = range->start + range->len; 2741 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2742 end >= MAX_BLKADDR(sbi)) 2743 return -EINVAL; 2744 2745 ret = mnt_want_write_file(filp); 2746 if (ret) 2747 return ret; 2748 2749 do_more: 2750 if (!range->sync) { 2751 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2752 ret = -EBUSY; 2753 goto out; 2754 } 2755 } else { 2756 f2fs_down_write(&sbi->gc_lock); 2757 } 2758 2759 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2760 stat_inc_gc_call_count(sbi, FOREGROUND); 2761 ret = f2fs_gc(sbi, &gc_control); 2762 if (ret) { 2763 if (ret == -EBUSY) 2764 ret = -EAGAIN; 2765 goto out; 2766 } 2767 range->start += CAP_BLKS_PER_SEC(sbi); 2768 if (range->start <= end) 2769 goto do_more; 2770 out: 2771 mnt_drop_write_file(filp); 2772 return ret; 2773 } 2774 2775 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2776 { 2777 struct f2fs_gc_range range; 2778 2779 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2780 sizeof(range))) 2781 return -EFAULT; 2782 return __f2fs_ioc_gc_range(filp, &range); 2783 } 2784 2785 static int f2fs_ioc_write_checkpoint(struct file *filp) 2786 { 2787 struct inode *inode = file_inode(filp); 2788 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2789 int ret; 2790 2791 if (!capable(CAP_SYS_ADMIN)) 2792 return -EPERM; 2793 2794 if (f2fs_readonly(sbi->sb)) 2795 return -EROFS; 2796 2797 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2798 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2799 return -EINVAL; 2800 } 2801 2802 ret = mnt_want_write_file(filp); 2803 if (ret) 2804 return ret; 2805 2806 ret = f2fs_sync_fs(sbi->sb, 1); 2807 2808 mnt_drop_write_file(filp); 2809 return ret; 2810 } 2811 2812 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2813 struct file *filp, 2814 struct f2fs_defragment *range) 2815 { 2816 struct inode *inode = file_inode(filp); 2817 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2818 .m_seg_type = NO_CHECK_TYPE, 2819 .m_may_create = false }; 2820 struct extent_info ei = {}; 2821 pgoff_t pg_start, pg_end, next_pgofs; 2822 unsigned int total = 0, sec_num; 2823 block_t blk_end = 0; 2824 bool fragmented = false; 2825 int err; 2826 2827 f2fs_balance_fs(sbi, true); 2828 2829 inode_lock(inode); 2830 pg_start = range->start >> PAGE_SHIFT; 2831 pg_end = min_t(pgoff_t, 2832 (range->start + range->len) >> PAGE_SHIFT, 2833 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE)); 2834 2835 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) || 2836 f2fs_is_atomic_file(inode)) { 2837 err = -EINVAL; 2838 goto unlock_out; 2839 } 2840 2841 /* if in-place-update policy is enabled, don't waste time here */ 2842 set_inode_flag(inode, FI_OPU_WRITE); 2843 if (f2fs_should_update_inplace(inode, NULL)) { 2844 err = -EINVAL; 2845 goto out; 2846 } 2847 2848 /* writeback all dirty pages in the range */ 2849 err = filemap_write_and_wait_range(inode->i_mapping, 2850 pg_start << PAGE_SHIFT, 2851 (pg_end << PAGE_SHIFT) - 1); 2852 if (err) 2853 goto out; 2854 2855 /* 2856 * lookup mapping info in extent cache, skip defragmenting if physical 2857 * block addresses are continuous. 2858 */ 2859 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2860 if ((pgoff_t)ei.fofs + ei.len >= pg_end) 2861 goto out; 2862 } 2863 2864 map.m_lblk = pg_start; 2865 map.m_next_pgofs = &next_pgofs; 2866 2867 /* 2868 * lookup mapping info in dnode page cache, skip defragmenting if all 2869 * physical block addresses are continuous even if there are hole(s) 2870 * in logical blocks. 2871 */ 2872 while (map.m_lblk < pg_end) { 2873 map.m_len = pg_end - map.m_lblk; 2874 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2875 if (err) 2876 goto out; 2877 2878 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2879 map.m_lblk = next_pgofs; 2880 continue; 2881 } 2882 2883 if (blk_end && blk_end != map.m_pblk) 2884 fragmented = true; 2885 2886 /* record total count of block that we're going to move */ 2887 total += map.m_len; 2888 2889 blk_end = map.m_pblk + map.m_len; 2890 2891 map.m_lblk += map.m_len; 2892 } 2893 2894 if (!fragmented) { 2895 total = 0; 2896 goto out; 2897 } 2898 2899 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2900 2901 /* 2902 * make sure there are enough free section for LFS allocation, this can 2903 * avoid defragment running in SSR mode when free section are allocated 2904 * intensively 2905 */ 2906 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2907 err = -EAGAIN; 2908 goto out; 2909 } 2910 2911 map.m_lblk = pg_start; 2912 map.m_len = pg_end - pg_start; 2913 total = 0; 2914 2915 while (map.m_lblk < pg_end) { 2916 pgoff_t idx; 2917 int cnt = 0; 2918 2919 do_map: 2920 map.m_len = pg_end - map.m_lblk; 2921 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2922 if (err) 2923 goto clear_out; 2924 2925 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2926 map.m_lblk = next_pgofs; 2927 goto check; 2928 } 2929 2930 set_inode_flag(inode, FI_SKIP_WRITES); 2931 2932 idx = map.m_lblk; 2933 while (idx < map.m_lblk + map.m_len && 2934 cnt < BLKS_PER_SEG(sbi)) { 2935 struct folio *folio; 2936 2937 folio = f2fs_get_lock_data_folio(inode, idx, true); 2938 if (IS_ERR(folio)) { 2939 err = PTR_ERR(folio); 2940 goto clear_out; 2941 } 2942 2943 f2fs_folio_wait_writeback(folio, DATA, true, true); 2944 2945 folio_mark_dirty(folio); 2946 set_page_private_gcing(&folio->page); 2947 f2fs_folio_put(folio, true); 2948 2949 idx++; 2950 cnt++; 2951 total++; 2952 } 2953 2954 map.m_lblk = idx; 2955 check: 2956 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 2957 goto do_map; 2958 2959 clear_inode_flag(inode, FI_SKIP_WRITES); 2960 2961 err = filemap_fdatawrite(inode->i_mapping); 2962 if (err) 2963 goto out; 2964 } 2965 clear_out: 2966 clear_inode_flag(inode, FI_SKIP_WRITES); 2967 out: 2968 clear_inode_flag(inode, FI_OPU_WRITE); 2969 unlock_out: 2970 inode_unlock(inode); 2971 if (!err) 2972 range->len = (u64)total << PAGE_SHIFT; 2973 return err; 2974 } 2975 2976 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2977 { 2978 struct inode *inode = file_inode(filp); 2979 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2980 struct f2fs_defragment range; 2981 int err; 2982 2983 if (!capable(CAP_SYS_ADMIN)) 2984 return -EPERM; 2985 2986 if (!S_ISREG(inode->i_mode)) 2987 return -EINVAL; 2988 2989 if (f2fs_readonly(sbi->sb)) 2990 return -EROFS; 2991 2992 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2993 sizeof(range))) 2994 return -EFAULT; 2995 2996 /* verify alignment of offset & size */ 2997 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2998 return -EINVAL; 2999 3000 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 3001 max_file_blocks(inode))) 3002 return -EINVAL; 3003 3004 err = mnt_want_write_file(filp); 3005 if (err) 3006 return err; 3007 3008 err = f2fs_defragment_range(sbi, filp, &range); 3009 mnt_drop_write_file(filp); 3010 3011 if (range.len) 3012 f2fs_update_time(sbi, REQ_TIME); 3013 if (err < 0) 3014 return err; 3015 3016 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 3017 sizeof(range))) 3018 return -EFAULT; 3019 3020 return 0; 3021 } 3022 3023 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 3024 struct file *file_out, loff_t pos_out, size_t len) 3025 { 3026 struct inode *src = file_inode(file_in); 3027 struct inode *dst = file_inode(file_out); 3028 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 3029 size_t olen = len, dst_max_i_size = 0; 3030 size_t dst_osize; 3031 int ret; 3032 3033 if (file_in->f_path.mnt != file_out->f_path.mnt || 3034 src->i_sb != dst->i_sb) 3035 return -EXDEV; 3036 3037 if (unlikely(f2fs_readonly(src->i_sb))) 3038 return -EROFS; 3039 3040 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 3041 return -EINVAL; 3042 3043 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 3044 return -EOPNOTSUPP; 3045 3046 if (pos_out < 0 || pos_in < 0) 3047 return -EINVAL; 3048 3049 if (src == dst) { 3050 if (pos_in == pos_out) 3051 return 0; 3052 if (pos_out > pos_in && pos_out < pos_in + len) 3053 return -EINVAL; 3054 } 3055 3056 inode_lock(src); 3057 if (src != dst) { 3058 ret = -EBUSY; 3059 if (!inode_trylock(dst)) 3060 goto out; 3061 } 3062 3063 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) || 3064 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 3065 ret = -EOPNOTSUPP; 3066 goto out_unlock; 3067 } 3068 3069 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) { 3070 ret = -EINVAL; 3071 goto out_unlock; 3072 } 3073 3074 ret = -EINVAL; 3075 if (pos_in + len > src->i_size || pos_in + len < pos_in) 3076 goto out_unlock; 3077 if (len == 0) 3078 olen = len = src->i_size - pos_in; 3079 if (pos_in + len == src->i_size) 3080 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 3081 if (len == 0) { 3082 ret = 0; 3083 goto out_unlock; 3084 } 3085 3086 dst_osize = dst->i_size; 3087 if (pos_out + olen > dst->i_size) 3088 dst_max_i_size = pos_out + olen; 3089 3090 /* verify the end result is block aligned */ 3091 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 3092 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 3093 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 3094 goto out_unlock; 3095 3096 ret = f2fs_convert_inline_inode(src); 3097 if (ret) 3098 goto out_unlock; 3099 3100 ret = f2fs_convert_inline_inode(dst); 3101 if (ret) 3102 goto out_unlock; 3103 3104 /* write out all dirty pages from offset */ 3105 ret = filemap_write_and_wait_range(src->i_mapping, 3106 pos_in, pos_in + len); 3107 if (ret) 3108 goto out_unlock; 3109 3110 ret = filemap_write_and_wait_range(dst->i_mapping, 3111 pos_out, pos_out + len); 3112 if (ret) 3113 goto out_unlock; 3114 3115 f2fs_balance_fs(sbi, true); 3116 3117 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3118 if (src != dst) { 3119 ret = -EBUSY; 3120 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 3121 goto out_src; 3122 } 3123 3124 f2fs_lock_op(sbi); 3125 ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in), 3126 F2FS_BYTES_TO_BLK(pos_out), 3127 F2FS_BYTES_TO_BLK(len), false); 3128 3129 if (!ret) { 3130 if (dst_max_i_size) 3131 f2fs_i_size_write(dst, dst_max_i_size); 3132 else if (dst_osize != dst->i_size) 3133 f2fs_i_size_write(dst, dst_osize); 3134 } 3135 f2fs_unlock_op(sbi); 3136 3137 if (src != dst) 3138 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 3139 out_src: 3140 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3141 if (ret) 3142 goto out_unlock; 3143 3144 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 3145 f2fs_mark_inode_dirty_sync(src, false); 3146 if (src != dst) { 3147 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 3148 f2fs_mark_inode_dirty_sync(dst, false); 3149 } 3150 f2fs_update_time(sbi, REQ_TIME); 3151 3152 out_unlock: 3153 if (src != dst) 3154 inode_unlock(dst); 3155 out: 3156 inode_unlock(src); 3157 return ret; 3158 } 3159 3160 static int __f2fs_ioc_move_range(struct file *filp, 3161 struct f2fs_move_range *range) 3162 { 3163 int err; 3164 3165 if (!(filp->f_mode & FMODE_READ) || 3166 !(filp->f_mode & FMODE_WRITE)) 3167 return -EBADF; 3168 3169 CLASS(fd, dst)(range->dst_fd); 3170 if (fd_empty(dst)) 3171 return -EBADF; 3172 3173 if (!(fd_file(dst)->f_mode & FMODE_WRITE)) 3174 return -EBADF; 3175 3176 err = mnt_want_write_file(filp); 3177 if (err) 3178 return err; 3179 3180 err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst), 3181 range->pos_out, range->len); 3182 3183 mnt_drop_write_file(filp); 3184 return err; 3185 } 3186 3187 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 3188 { 3189 struct f2fs_move_range range; 3190 3191 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 3192 sizeof(range))) 3193 return -EFAULT; 3194 return __f2fs_ioc_move_range(filp, &range); 3195 } 3196 3197 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 3198 { 3199 struct inode *inode = file_inode(filp); 3200 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3201 struct sit_info *sm = SIT_I(sbi); 3202 unsigned int start_segno = 0, end_segno = 0; 3203 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3204 struct f2fs_flush_device range; 3205 struct f2fs_gc_control gc_control = { 3206 .init_gc_type = FG_GC, 3207 .should_migrate_blocks = true, 3208 .err_gc_skipped = true, 3209 .nr_free_secs = 0 }; 3210 int ret; 3211 3212 if (!capable(CAP_SYS_ADMIN)) 3213 return -EPERM; 3214 3215 if (f2fs_readonly(sbi->sb)) 3216 return -EROFS; 3217 3218 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3219 return -EINVAL; 3220 3221 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3222 sizeof(range))) 3223 return -EFAULT; 3224 3225 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3226 __is_large_section(sbi)) { 3227 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3228 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3229 return -EINVAL; 3230 } 3231 3232 ret = mnt_want_write_file(filp); 3233 if (ret) 3234 return ret; 3235 3236 if (range.dev_num != 0) 3237 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3238 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3239 3240 start_segno = sm->last_victim[FLUSH_DEVICE]; 3241 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3242 start_segno = dev_start_segno; 3243 end_segno = min(start_segno + range.segments, dev_end_segno); 3244 3245 while (start_segno < end_segno) { 3246 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3247 ret = -EBUSY; 3248 goto out; 3249 } 3250 sm->last_victim[GC_CB] = end_segno + 1; 3251 sm->last_victim[GC_GREEDY] = end_segno + 1; 3252 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3253 3254 gc_control.victim_segno = start_segno; 3255 stat_inc_gc_call_count(sbi, FOREGROUND); 3256 ret = f2fs_gc(sbi, &gc_control); 3257 if (ret == -EAGAIN) 3258 ret = 0; 3259 else if (ret < 0) 3260 break; 3261 start_segno++; 3262 } 3263 out: 3264 mnt_drop_write_file(filp); 3265 return ret; 3266 } 3267 3268 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3269 { 3270 struct inode *inode = file_inode(filp); 3271 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3272 3273 /* Must validate to set it with SQLite behavior in Android. */ 3274 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3275 3276 return put_user(sb_feature, (u32 __user *)arg); 3277 } 3278 3279 #ifdef CONFIG_QUOTA 3280 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3281 { 3282 struct dquot *transfer_to[MAXQUOTAS] = {}; 3283 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3284 struct super_block *sb = sbi->sb; 3285 int err; 3286 3287 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3288 if (IS_ERR(transfer_to[PRJQUOTA])) 3289 return PTR_ERR(transfer_to[PRJQUOTA]); 3290 3291 err = __dquot_transfer(inode, transfer_to); 3292 if (err) 3293 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3294 dqput(transfer_to[PRJQUOTA]); 3295 return err; 3296 } 3297 3298 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3299 { 3300 struct f2fs_inode_info *fi = F2FS_I(inode); 3301 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3302 struct f2fs_inode *ri = NULL; 3303 kprojid_t kprojid; 3304 int err; 3305 3306 if (!f2fs_sb_has_project_quota(sbi)) { 3307 if (projid != F2FS_DEF_PROJID) 3308 return -EOPNOTSUPP; 3309 else 3310 return 0; 3311 } 3312 3313 if (!f2fs_has_extra_attr(inode)) 3314 return -EOPNOTSUPP; 3315 3316 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3317 3318 if (projid_eq(kprojid, fi->i_projid)) 3319 return 0; 3320 3321 err = -EPERM; 3322 /* Is it quota file? Do not allow user to mess with it */ 3323 if (IS_NOQUOTA(inode)) 3324 return err; 3325 3326 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3327 return -EOVERFLOW; 3328 3329 err = f2fs_dquot_initialize(inode); 3330 if (err) 3331 return err; 3332 3333 f2fs_lock_op(sbi); 3334 err = f2fs_transfer_project_quota(inode, kprojid); 3335 if (err) 3336 goto out_unlock; 3337 3338 fi->i_projid = kprojid; 3339 inode_set_ctime_current(inode); 3340 f2fs_mark_inode_dirty_sync(inode, true); 3341 out_unlock: 3342 f2fs_unlock_op(sbi); 3343 return err; 3344 } 3345 #else 3346 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3347 { 3348 return 0; 3349 } 3350 3351 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3352 { 3353 if (projid != F2FS_DEF_PROJID) 3354 return -EOPNOTSUPP; 3355 return 0; 3356 } 3357 #endif 3358 3359 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3360 { 3361 struct inode *inode = d_inode(dentry); 3362 struct f2fs_inode_info *fi = F2FS_I(inode); 3363 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3364 3365 if (IS_ENCRYPTED(inode)) 3366 fsflags |= FS_ENCRYPT_FL; 3367 if (IS_VERITY(inode)) 3368 fsflags |= FS_VERITY_FL; 3369 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3370 fsflags |= FS_INLINE_DATA_FL; 3371 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3372 fsflags |= FS_NOCOW_FL; 3373 3374 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3375 3376 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3377 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3378 3379 return 0; 3380 } 3381 3382 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3383 struct dentry *dentry, struct fileattr *fa) 3384 { 3385 struct inode *inode = d_inode(dentry); 3386 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3387 u32 iflags; 3388 int err; 3389 3390 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3391 return -EIO; 3392 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3393 return -ENOSPC; 3394 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3395 return -EOPNOTSUPP; 3396 fsflags &= F2FS_SETTABLE_FS_FL; 3397 if (!fa->flags_valid) 3398 mask &= FS_COMMON_FL; 3399 3400 iflags = f2fs_fsflags_to_iflags(fsflags); 3401 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3402 return -EOPNOTSUPP; 3403 3404 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3405 if (!err) 3406 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3407 3408 return err; 3409 } 3410 3411 int f2fs_pin_file_control(struct inode *inode, bool inc) 3412 { 3413 struct f2fs_inode_info *fi = F2FS_I(inode); 3414 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3415 3416 if (IS_DEVICE_ALIASING(inode)) 3417 return -EINVAL; 3418 3419 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) { 3420 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3421 __func__, inode->i_ino, fi->i_gc_failures); 3422 clear_inode_flag(inode, FI_PIN_FILE); 3423 return -EAGAIN; 3424 } 3425 3426 /* Use i_gc_failures for normal file as a risk signal. */ 3427 if (inc) 3428 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1); 3429 3430 return 0; 3431 } 3432 3433 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3434 { 3435 struct inode *inode = file_inode(filp); 3436 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3437 __u32 pin; 3438 int ret = 0; 3439 3440 if (get_user(pin, (__u32 __user *)arg)) 3441 return -EFAULT; 3442 3443 if (!S_ISREG(inode->i_mode)) 3444 return -EINVAL; 3445 3446 if (f2fs_readonly(sbi->sb)) 3447 return -EROFS; 3448 3449 if (!pin && IS_DEVICE_ALIASING(inode)) 3450 return -EOPNOTSUPP; 3451 3452 ret = mnt_want_write_file(filp); 3453 if (ret) 3454 return ret; 3455 3456 inode_lock(inode); 3457 3458 if (f2fs_is_atomic_file(inode)) { 3459 ret = -EINVAL; 3460 goto out; 3461 } 3462 3463 if (!pin) { 3464 clear_inode_flag(inode, FI_PIN_FILE); 3465 f2fs_i_gc_failures_write(inode, 0); 3466 goto done; 3467 } else if (f2fs_is_pinned_file(inode)) { 3468 goto done; 3469 } 3470 3471 if (F2FS_HAS_BLOCKS(inode)) { 3472 ret = -EFBIG; 3473 goto out; 3474 } 3475 3476 /* Let's allow file pinning on zoned device. */ 3477 if (!f2fs_sb_has_blkzoned(sbi) && 3478 f2fs_should_update_outplace(inode, NULL)) { 3479 ret = -EINVAL; 3480 goto out; 3481 } 3482 3483 if (f2fs_pin_file_control(inode, false)) { 3484 ret = -EAGAIN; 3485 goto out; 3486 } 3487 3488 ret = f2fs_convert_inline_inode(inode); 3489 if (ret) 3490 goto out; 3491 3492 if (!f2fs_disable_compressed_file(inode)) { 3493 ret = -EOPNOTSUPP; 3494 goto out; 3495 } 3496 3497 set_inode_flag(inode, FI_PIN_FILE); 3498 ret = F2FS_I(inode)->i_gc_failures; 3499 done: 3500 f2fs_update_time(sbi, REQ_TIME); 3501 out: 3502 inode_unlock(inode); 3503 mnt_drop_write_file(filp); 3504 return ret; 3505 } 3506 3507 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3508 { 3509 struct inode *inode = file_inode(filp); 3510 __u32 pin = 0; 3511 3512 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3513 pin = F2FS_I(inode)->i_gc_failures; 3514 return put_user(pin, (u32 __user *)arg); 3515 } 3516 3517 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg) 3518 { 3519 return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0, 3520 (u32 __user *)arg); 3521 } 3522 3523 static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg) 3524 { 3525 struct inode *inode = file_inode(filp); 3526 __u32 level; 3527 3528 if (get_user(level, (__u32 __user *)arg)) 3529 return -EFAULT; 3530 3531 if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX) 3532 return -EINVAL; 3533 3534 inode_lock(inode); 3535 F2FS_I(inode)->ioprio_hint = level; 3536 inode_unlock(inode); 3537 return 0; 3538 } 3539 3540 int f2fs_precache_extents(struct inode *inode) 3541 { 3542 struct f2fs_inode_info *fi = F2FS_I(inode); 3543 struct f2fs_map_blocks map; 3544 pgoff_t m_next_extent; 3545 loff_t end; 3546 int err; 3547 3548 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3549 return -EOPNOTSUPP; 3550 3551 map.m_lblk = 0; 3552 map.m_pblk = 0; 3553 map.m_next_pgofs = NULL; 3554 map.m_next_extent = &m_next_extent; 3555 map.m_seg_type = NO_CHECK_TYPE; 3556 map.m_may_create = false; 3557 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3558 3559 while (map.m_lblk < end) { 3560 map.m_len = end - map.m_lblk; 3561 3562 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3563 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3564 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3565 if (err || !map.m_len) 3566 return err; 3567 3568 map.m_lblk = m_next_extent; 3569 } 3570 3571 return 0; 3572 } 3573 3574 static int f2fs_ioc_precache_extents(struct file *filp) 3575 { 3576 return f2fs_precache_extents(file_inode(filp)); 3577 } 3578 3579 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3580 { 3581 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3582 __u64 block_count; 3583 3584 if (!capable(CAP_SYS_ADMIN)) 3585 return -EPERM; 3586 3587 if (f2fs_readonly(sbi->sb)) 3588 return -EROFS; 3589 3590 if (copy_from_user(&block_count, (void __user *)arg, 3591 sizeof(block_count))) 3592 return -EFAULT; 3593 3594 return f2fs_resize_fs(filp, block_count); 3595 } 3596 3597 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3598 { 3599 struct inode *inode = file_inode(filp); 3600 3601 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3602 3603 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3604 f2fs_warn(F2FS_I_SB(inode), 3605 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3606 inode->i_ino); 3607 return -EOPNOTSUPP; 3608 } 3609 3610 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3611 } 3612 3613 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3614 { 3615 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3616 return -EOPNOTSUPP; 3617 3618 return fsverity_ioctl_measure(filp, (void __user *)arg); 3619 } 3620 3621 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3622 { 3623 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3624 return -EOPNOTSUPP; 3625 3626 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3627 } 3628 3629 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3630 { 3631 struct inode *inode = file_inode(filp); 3632 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3633 char *vbuf; 3634 int count; 3635 int err = 0; 3636 3637 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3638 if (!vbuf) 3639 return -ENOMEM; 3640 3641 f2fs_down_read(&sbi->sb_lock); 3642 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3643 ARRAY_SIZE(sbi->raw_super->volume_name), 3644 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3645 f2fs_up_read(&sbi->sb_lock); 3646 3647 if (copy_to_user((char __user *)arg, vbuf, 3648 min(FSLABEL_MAX, count))) 3649 err = -EFAULT; 3650 3651 kfree(vbuf); 3652 return err; 3653 } 3654 3655 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3656 { 3657 struct inode *inode = file_inode(filp); 3658 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3659 char *vbuf; 3660 int err = 0; 3661 3662 if (!capable(CAP_SYS_ADMIN)) 3663 return -EPERM; 3664 3665 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3666 if (IS_ERR(vbuf)) 3667 return PTR_ERR(vbuf); 3668 3669 err = mnt_want_write_file(filp); 3670 if (err) 3671 goto out; 3672 3673 f2fs_down_write(&sbi->sb_lock); 3674 3675 memset(sbi->raw_super->volume_name, 0, 3676 sizeof(sbi->raw_super->volume_name)); 3677 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3678 sbi->raw_super->volume_name, 3679 ARRAY_SIZE(sbi->raw_super->volume_name)); 3680 3681 err = f2fs_commit_super(sbi, false); 3682 3683 f2fs_up_write(&sbi->sb_lock); 3684 3685 mnt_drop_write_file(filp); 3686 out: 3687 kfree(vbuf); 3688 return err; 3689 } 3690 3691 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3692 { 3693 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3694 return -EOPNOTSUPP; 3695 3696 if (!f2fs_compressed_file(inode)) 3697 return -EINVAL; 3698 3699 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3700 3701 return 0; 3702 } 3703 3704 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3705 { 3706 struct inode *inode = file_inode(filp); 3707 __u64 blocks; 3708 int ret; 3709 3710 ret = f2fs_get_compress_blocks(inode, &blocks); 3711 if (ret < 0) 3712 return ret; 3713 3714 return put_user(blocks, (u64 __user *)arg); 3715 } 3716 3717 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3718 { 3719 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3720 unsigned int released_blocks = 0; 3721 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3722 block_t blkaddr; 3723 int i; 3724 3725 for (i = 0; i < count; i++) { 3726 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3727 dn->ofs_in_node + i); 3728 3729 if (!__is_valid_data_blkaddr(blkaddr)) 3730 continue; 3731 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3732 DATA_GENERIC_ENHANCE))) 3733 return -EFSCORRUPTED; 3734 } 3735 3736 while (count) { 3737 int compr_blocks = 0; 3738 3739 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3740 blkaddr = f2fs_data_blkaddr(dn); 3741 3742 if (i == 0) { 3743 if (blkaddr == COMPRESS_ADDR) 3744 continue; 3745 dn->ofs_in_node += cluster_size; 3746 goto next; 3747 } 3748 3749 if (__is_valid_data_blkaddr(blkaddr)) 3750 compr_blocks++; 3751 3752 if (blkaddr != NEW_ADDR) 3753 continue; 3754 3755 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3756 } 3757 3758 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3759 dec_valid_block_count(sbi, dn->inode, 3760 cluster_size - compr_blocks); 3761 3762 released_blocks += cluster_size - compr_blocks; 3763 next: 3764 count -= cluster_size; 3765 } 3766 3767 return released_blocks; 3768 } 3769 3770 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3771 { 3772 struct inode *inode = file_inode(filp); 3773 struct f2fs_inode_info *fi = F2FS_I(inode); 3774 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3775 pgoff_t page_idx = 0, last_idx; 3776 unsigned int released_blocks = 0; 3777 int ret; 3778 int writecount; 3779 3780 if (!f2fs_sb_has_compression(sbi)) 3781 return -EOPNOTSUPP; 3782 3783 if (f2fs_readonly(sbi->sb)) 3784 return -EROFS; 3785 3786 ret = mnt_want_write_file(filp); 3787 if (ret) 3788 return ret; 3789 3790 f2fs_balance_fs(sbi, true); 3791 3792 inode_lock(inode); 3793 3794 writecount = atomic_read(&inode->i_writecount); 3795 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3796 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3797 ret = -EBUSY; 3798 goto out; 3799 } 3800 3801 if (!f2fs_compressed_file(inode) || 3802 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3803 ret = -EINVAL; 3804 goto out; 3805 } 3806 3807 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3808 if (ret) 3809 goto out; 3810 3811 if (!atomic_read(&fi->i_compr_blocks)) { 3812 ret = -EPERM; 3813 goto out; 3814 } 3815 3816 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3817 inode_set_ctime_current(inode); 3818 f2fs_mark_inode_dirty_sync(inode, true); 3819 3820 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3821 filemap_invalidate_lock(inode->i_mapping); 3822 3823 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3824 3825 while (page_idx < last_idx) { 3826 struct dnode_of_data dn; 3827 pgoff_t end_offset, count; 3828 3829 f2fs_lock_op(sbi); 3830 3831 set_new_dnode(&dn, inode, NULL, NULL, 0); 3832 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3833 if (ret) { 3834 f2fs_unlock_op(sbi); 3835 if (ret == -ENOENT) { 3836 page_idx = f2fs_get_next_page_offset(&dn, 3837 page_idx); 3838 ret = 0; 3839 continue; 3840 } 3841 break; 3842 } 3843 3844 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 3845 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3846 count = round_up(count, fi->i_cluster_size); 3847 3848 ret = release_compress_blocks(&dn, count); 3849 3850 f2fs_put_dnode(&dn); 3851 3852 f2fs_unlock_op(sbi); 3853 3854 if (ret < 0) 3855 break; 3856 3857 page_idx += count; 3858 released_blocks += ret; 3859 } 3860 3861 filemap_invalidate_unlock(inode->i_mapping); 3862 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3863 out: 3864 if (released_blocks) 3865 f2fs_update_time(sbi, REQ_TIME); 3866 inode_unlock(inode); 3867 3868 mnt_drop_write_file(filp); 3869 3870 if (ret >= 0) { 3871 ret = put_user(released_blocks, (u64 __user *)arg); 3872 } else if (released_blocks && 3873 atomic_read(&fi->i_compr_blocks)) { 3874 set_sbi_flag(sbi, SBI_NEED_FSCK); 3875 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3876 "iblocks=%llu, released=%u, compr_blocks=%u, " 3877 "run fsck to fix.", 3878 __func__, inode->i_ino, inode->i_blocks, 3879 released_blocks, 3880 atomic_read(&fi->i_compr_blocks)); 3881 } 3882 3883 return ret; 3884 } 3885 3886 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3887 unsigned int *reserved_blocks) 3888 { 3889 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3890 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3891 block_t blkaddr; 3892 int i; 3893 3894 for (i = 0; i < count; i++) { 3895 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3896 dn->ofs_in_node + i); 3897 3898 if (!__is_valid_data_blkaddr(blkaddr)) 3899 continue; 3900 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3901 DATA_GENERIC_ENHANCE))) 3902 return -EFSCORRUPTED; 3903 } 3904 3905 while (count) { 3906 int compr_blocks = 0; 3907 blkcnt_t reserved = 0; 3908 blkcnt_t to_reserved; 3909 int ret; 3910 3911 for (i = 0; i < cluster_size; i++) { 3912 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3913 dn->ofs_in_node + i); 3914 3915 if (i == 0) { 3916 if (blkaddr != COMPRESS_ADDR) { 3917 dn->ofs_in_node += cluster_size; 3918 goto next; 3919 } 3920 continue; 3921 } 3922 3923 /* 3924 * compressed cluster was not released due to it 3925 * fails in release_compress_blocks(), so NEW_ADDR 3926 * is a possible case. 3927 */ 3928 if (blkaddr == NEW_ADDR) { 3929 reserved++; 3930 continue; 3931 } 3932 if (__is_valid_data_blkaddr(blkaddr)) { 3933 compr_blocks++; 3934 continue; 3935 } 3936 } 3937 3938 to_reserved = cluster_size - compr_blocks - reserved; 3939 3940 /* for the case all blocks in cluster were reserved */ 3941 if (reserved && to_reserved == 1) { 3942 dn->ofs_in_node += cluster_size; 3943 goto next; 3944 } 3945 3946 ret = inc_valid_block_count(sbi, dn->inode, 3947 &to_reserved, false); 3948 if (unlikely(ret)) 3949 return ret; 3950 3951 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3952 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3953 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3954 } 3955 3956 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3957 3958 *reserved_blocks += to_reserved; 3959 next: 3960 count -= cluster_size; 3961 } 3962 3963 return 0; 3964 } 3965 3966 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3967 { 3968 struct inode *inode = file_inode(filp); 3969 struct f2fs_inode_info *fi = F2FS_I(inode); 3970 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3971 pgoff_t page_idx = 0, last_idx; 3972 unsigned int reserved_blocks = 0; 3973 int ret; 3974 3975 if (!f2fs_sb_has_compression(sbi)) 3976 return -EOPNOTSUPP; 3977 3978 if (f2fs_readonly(sbi->sb)) 3979 return -EROFS; 3980 3981 ret = mnt_want_write_file(filp); 3982 if (ret) 3983 return ret; 3984 3985 f2fs_balance_fs(sbi, true); 3986 3987 inode_lock(inode); 3988 3989 if (!f2fs_compressed_file(inode) || 3990 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3991 ret = -EINVAL; 3992 goto unlock_inode; 3993 } 3994 3995 if (atomic_read(&fi->i_compr_blocks)) 3996 goto unlock_inode; 3997 3998 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3999 filemap_invalidate_lock(inode->i_mapping); 4000 4001 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4002 4003 while (page_idx < last_idx) { 4004 struct dnode_of_data dn; 4005 pgoff_t end_offset, count; 4006 4007 f2fs_lock_op(sbi); 4008 4009 set_new_dnode(&dn, inode, NULL, NULL, 0); 4010 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 4011 if (ret) { 4012 f2fs_unlock_op(sbi); 4013 if (ret == -ENOENT) { 4014 page_idx = f2fs_get_next_page_offset(&dn, 4015 page_idx); 4016 ret = 0; 4017 continue; 4018 } 4019 break; 4020 } 4021 4022 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 4023 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 4024 count = round_up(count, fi->i_cluster_size); 4025 4026 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 4027 4028 f2fs_put_dnode(&dn); 4029 4030 f2fs_unlock_op(sbi); 4031 4032 if (ret < 0) 4033 break; 4034 4035 page_idx += count; 4036 } 4037 4038 filemap_invalidate_unlock(inode->i_mapping); 4039 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 4040 4041 if (!ret) { 4042 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 4043 inode_set_ctime_current(inode); 4044 f2fs_mark_inode_dirty_sync(inode, true); 4045 } 4046 unlock_inode: 4047 if (reserved_blocks) 4048 f2fs_update_time(sbi, REQ_TIME); 4049 inode_unlock(inode); 4050 mnt_drop_write_file(filp); 4051 4052 if (!ret) { 4053 ret = put_user(reserved_blocks, (u64 __user *)arg); 4054 } else if (reserved_blocks && 4055 atomic_read(&fi->i_compr_blocks)) { 4056 set_sbi_flag(sbi, SBI_NEED_FSCK); 4057 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx " 4058 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 4059 "run fsck to fix.", 4060 __func__, inode->i_ino, inode->i_blocks, 4061 reserved_blocks, 4062 atomic_read(&fi->i_compr_blocks)); 4063 } 4064 4065 return ret; 4066 } 4067 4068 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 4069 pgoff_t off, block_t block, block_t len, u32 flags) 4070 { 4071 sector_t sector = SECTOR_FROM_BLOCK(block); 4072 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 4073 int ret = 0; 4074 4075 if (flags & F2FS_TRIM_FILE_DISCARD) { 4076 if (bdev_max_secure_erase_sectors(bdev)) 4077 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 4078 GFP_NOFS); 4079 else 4080 ret = blkdev_issue_discard(bdev, sector, nr_sects, 4081 GFP_NOFS); 4082 } 4083 4084 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 4085 if (IS_ENCRYPTED(inode)) 4086 ret = fscrypt_zeroout_range(inode, off, block, len); 4087 else 4088 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 4089 GFP_NOFS, 0); 4090 } 4091 4092 return ret; 4093 } 4094 4095 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 4096 { 4097 struct inode *inode = file_inode(filp); 4098 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4099 struct address_space *mapping = inode->i_mapping; 4100 struct block_device *prev_bdev = NULL; 4101 struct f2fs_sectrim_range range; 4102 pgoff_t index, pg_end, prev_index = 0; 4103 block_t prev_block = 0, len = 0; 4104 loff_t end_addr; 4105 bool to_end = false; 4106 int ret = 0; 4107 4108 if (!(filp->f_mode & FMODE_WRITE)) 4109 return -EBADF; 4110 4111 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 4112 sizeof(range))) 4113 return -EFAULT; 4114 4115 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 4116 !S_ISREG(inode->i_mode)) 4117 return -EINVAL; 4118 4119 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 4120 !f2fs_hw_support_discard(sbi)) || 4121 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 4122 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 4123 return -EOPNOTSUPP; 4124 4125 ret = mnt_want_write_file(filp); 4126 if (ret) 4127 return ret; 4128 inode_lock(inode); 4129 4130 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 4131 range.start >= inode->i_size) { 4132 ret = -EINVAL; 4133 goto err; 4134 } 4135 4136 if (range.len == 0) 4137 goto err; 4138 4139 if (inode->i_size - range.start > range.len) { 4140 end_addr = range.start + range.len; 4141 } else { 4142 end_addr = range.len == (u64)-1 ? 4143 sbi->sb->s_maxbytes : inode->i_size; 4144 to_end = true; 4145 } 4146 4147 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 4148 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 4149 ret = -EINVAL; 4150 goto err; 4151 } 4152 4153 index = F2FS_BYTES_TO_BLK(range.start); 4154 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 4155 4156 ret = f2fs_convert_inline_inode(inode); 4157 if (ret) 4158 goto err; 4159 4160 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4161 filemap_invalidate_lock(mapping); 4162 4163 ret = filemap_write_and_wait_range(mapping, range.start, 4164 to_end ? LLONG_MAX : end_addr - 1); 4165 if (ret) 4166 goto out; 4167 4168 truncate_inode_pages_range(mapping, range.start, 4169 to_end ? -1 : end_addr - 1); 4170 4171 while (index < pg_end) { 4172 struct dnode_of_data dn; 4173 pgoff_t end_offset, count; 4174 int i; 4175 4176 set_new_dnode(&dn, inode, NULL, NULL, 0); 4177 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 4178 if (ret) { 4179 if (ret == -ENOENT) { 4180 index = f2fs_get_next_page_offset(&dn, index); 4181 continue; 4182 } 4183 goto out; 4184 } 4185 4186 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 4187 count = min(end_offset - dn.ofs_in_node, pg_end - index); 4188 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 4189 struct block_device *cur_bdev; 4190 block_t blkaddr = f2fs_data_blkaddr(&dn); 4191 4192 if (!__is_valid_data_blkaddr(blkaddr)) 4193 continue; 4194 4195 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 4196 DATA_GENERIC_ENHANCE)) { 4197 ret = -EFSCORRUPTED; 4198 f2fs_put_dnode(&dn); 4199 goto out; 4200 } 4201 4202 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 4203 if (f2fs_is_multi_device(sbi)) { 4204 int di = f2fs_target_device_index(sbi, blkaddr); 4205 4206 blkaddr -= FDEV(di).start_blk; 4207 } 4208 4209 if (len) { 4210 if (prev_bdev == cur_bdev && 4211 index == prev_index + len && 4212 blkaddr == prev_block + len) { 4213 len++; 4214 } else { 4215 ret = f2fs_secure_erase(prev_bdev, 4216 inode, prev_index, prev_block, 4217 len, range.flags); 4218 if (ret) { 4219 f2fs_put_dnode(&dn); 4220 goto out; 4221 } 4222 4223 len = 0; 4224 } 4225 } 4226 4227 if (!len) { 4228 prev_bdev = cur_bdev; 4229 prev_index = index; 4230 prev_block = blkaddr; 4231 len = 1; 4232 } 4233 } 4234 4235 f2fs_put_dnode(&dn); 4236 4237 if (fatal_signal_pending(current)) { 4238 ret = -EINTR; 4239 goto out; 4240 } 4241 cond_resched(); 4242 } 4243 4244 if (len) 4245 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 4246 prev_block, len, range.flags); 4247 f2fs_update_time(sbi, REQ_TIME); 4248 out: 4249 filemap_invalidate_unlock(mapping); 4250 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4251 err: 4252 inode_unlock(inode); 4253 mnt_drop_write_file(filp); 4254 4255 return ret; 4256 } 4257 4258 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 4259 { 4260 struct inode *inode = file_inode(filp); 4261 struct f2fs_comp_option option; 4262 4263 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4264 return -EOPNOTSUPP; 4265 4266 inode_lock_shared(inode); 4267 4268 if (!f2fs_compressed_file(inode)) { 4269 inode_unlock_shared(inode); 4270 return -ENODATA; 4271 } 4272 4273 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4274 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4275 4276 inode_unlock_shared(inode); 4277 4278 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4279 sizeof(option))) 4280 return -EFAULT; 4281 4282 return 0; 4283 } 4284 4285 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4286 { 4287 struct inode *inode = file_inode(filp); 4288 struct f2fs_inode_info *fi = F2FS_I(inode); 4289 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4290 struct f2fs_comp_option option; 4291 int ret = 0; 4292 4293 if (!f2fs_sb_has_compression(sbi)) 4294 return -EOPNOTSUPP; 4295 4296 if (!(filp->f_mode & FMODE_WRITE)) 4297 return -EBADF; 4298 4299 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4300 sizeof(option))) 4301 return -EFAULT; 4302 4303 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4304 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4305 option.algorithm >= COMPRESS_MAX) 4306 return -EINVAL; 4307 4308 ret = mnt_want_write_file(filp); 4309 if (ret) 4310 return ret; 4311 inode_lock(inode); 4312 4313 f2fs_down_write(&F2FS_I(inode)->i_sem); 4314 if (!f2fs_compressed_file(inode)) { 4315 ret = -EINVAL; 4316 goto out; 4317 } 4318 4319 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4320 ret = -EBUSY; 4321 goto out; 4322 } 4323 4324 if (F2FS_HAS_BLOCKS(inode)) { 4325 ret = -EFBIG; 4326 goto out; 4327 } 4328 4329 fi->i_compress_algorithm = option.algorithm; 4330 fi->i_log_cluster_size = option.log_cluster_size; 4331 fi->i_cluster_size = BIT(option.log_cluster_size); 4332 /* Set default level */ 4333 if (fi->i_compress_algorithm == COMPRESS_ZSTD) 4334 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4335 else 4336 fi->i_compress_level = 0; 4337 /* Adjust mount option level */ 4338 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4339 F2FS_OPTION(sbi).compress_level) 4340 fi->i_compress_level = F2FS_OPTION(sbi).compress_level; 4341 f2fs_mark_inode_dirty_sync(inode, true); 4342 4343 if (!f2fs_is_compress_backend_ready(inode)) 4344 f2fs_warn(sbi, "compression algorithm is successfully set, " 4345 "but current kernel doesn't support this algorithm."); 4346 out: 4347 f2fs_up_write(&fi->i_sem); 4348 inode_unlock(inode); 4349 mnt_drop_write_file(filp); 4350 4351 return ret; 4352 } 4353 4354 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4355 { 4356 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4357 struct address_space *mapping = inode->i_mapping; 4358 struct folio *folio; 4359 pgoff_t redirty_idx = page_idx; 4360 int page_len = 0, ret = 0; 4361 4362 page_cache_ra_unbounded(&ractl, len, 0); 4363 4364 do { 4365 folio = read_cache_folio(mapping, page_idx, NULL, NULL); 4366 if (IS_ERR(folio)) { 4367 ret = PTR_ERR(folio); 4368 break; 4369 } 4370 page_len += folio_nr_pages(folio) - (page_idx - folio->index); 4371 page_idx = folio_next_index(folio); 4372 } while (page_len < len); 4373 4374 do { 4375 folio = filemap_lock_folio(mapping, redirty_idx); 4376 4377 /* It will never fail, when folio has pinned above */ 4378 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio)); 4379 4380 f2fs_folio_wait_writeback(folio, DATA, true, true); 4381 4382 folio_mark_dirty(folio); 4383 set_page_private_gcing(&folio->page); 4384 redirty_idx = folio_next_index(folio); 4385 folio_unlock(folio); 4386 folio_put_refs(folio, 2); 4387 } while (redirty_idx < page_idx); 4388 4389 return ret; 4390 } 4391 4392 static int f2fs_ioc_decompress_file(struct file *filp) 4393 { 4394 struct inode *inode = file_inode(filp); 4395 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4396 struct f2fs_inode_info *fi = F2FS_I(inode); 4397 pgoff_t page_idx = 0, last_idx, cluster_idx; 4398 int ret; 4399 4400 if (!f2fs_sb_has_compression(sbi) || 4401 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4402 return -EOPNOTSUPP; 4403 4404 if (!(filp->f_mode & FMODE_WRITE)) 4405 return -EBADF; 4406 4407 f2fs_balance_fs(sbi, true); 4408 4409 ret = mnt_want_write_file(filp); 4410 if (ret) 4411 return ret; 4412 inode_lock(inode); 4413 4414 if (!f2fs_is_compress_backend_ready(inode)) { 4415 ret = -EOPNOTSUPP; 4416 goto out; 4417 } 4418 4419 if (!f2fs_compressed_file(inode) || 4420 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4421 ret = -EINVAL; 4422 goto out; 4423 } 4424 4425 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4426 if (ret) 4427 goto out; 4428 4429 if (!atomic_read(&fi->i_compr_blocks)) 4430 goto out; 4431 4432 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4433 last_idx >>= fi->i_log_cluster_size; 4434 4435 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4436 page_idx = cluster_idx << fi->i_log_cluster_size; 4437 4438 if (!f2fs_is_compressed_cluster(inode, page_idx)) 4439 continue; 4440 4441 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4442 if (ret < 0) 4443 break; 4444 4445 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4446 ret = filemap_fdatawrite(inode->i_mapping); 4447 if (ret < 0) 4448 break; 4449 } 4450 4451 cond_resched(); 4452 if (fatal_signal_pending(current)) { 4453 ret = -EINTR; 4454 break; 4455 } 4456 } 4457 4458 if (!ret) 4459 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4460 LLONG_MAX); 4461 4462 if (ret) 4463 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4464 __func__, ret); 4465 f2fs_update_time(sbi, REQ_TIME); 4466 out: 4467 inode_unlock(inode); 4468 mnt_drop_write_file(filp); 4469 4470 return ret; 4471 } 4472 4473 static int f2fs_ioc_compress_file(struct file *filp) 4474 { 4475 struct inode *inode = file_inode(filp); 4476 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4477 struct f2fs_inode_info *fi = F2FS_I(inode); 4478 pgoff_t page_idx = 0, last_idx, cluster_idx; 4479 int ret; 4480 4481 if (!f2fs_sb_has_compression(sbi) || 4482 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4483 return -EOPNOTSUPP; 4484 4485 if (!(filp->f_mode & FMODE_WRITE)) 4486 return -EBADF; 4487 4488 f2fs_balance_fs(sbi, true); 4489 4490 ret = mnt_want_write_file(filp); 4491 if (ret) 4492 return ret; 4493 inode_lock(inode); 4494 4495 if (!f2fs_is_compress_backend_ready(inode)) { 4496 ret = -EOPNOTSUPP; 4497 goto out; 4498 } 4499 4500 if (!f2fs_compressed_file(inode) || 4501 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4502 ret = -EINVAL; 4503 goto out; 4504 } 4505 4506 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4507 if (ret) 4508 goto out; 4509 4510 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4511 4512 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4513 last_idx >>= fi->i_log_cluster_size; 4514 4515 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4516 page_idx = cluster_idx << fi->i_log_cluster_size; 4517 4518 if (f2fs_is_sparse_cluster(inode, page_idx)) 4519 continue; 4520 4521 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4522 if (ret < 0) 4523 break; 4524 4525 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4526 ret = filemap_fdatawrite(inode->i_mapping); 4527 if (ret < 0) 4528 break; 4529 } 4530 4531 cond_resched(); 4532 if (fatal_signal_pending(current)) { 4533 ret = -EINTR; 4534 break; 4535 } 4536 } 4537 4538 if (!ret) 4539 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4540 LLONG_MAX); 4541 4542 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4543 4544 if (ret) 4545 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4546 __func__, ret); 4547 f2fs_update_time(sbi, REQ_TIME); 4548 out: 4549 inode_unlock(inode); 4550 mnt_drop_write_file(filp); 4551 4552 return ret; 4553 } 4554 4555 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4556 { 4557 switch (cmd) { 4558 case FS_IOC_GETVERSION: 4559 return f2fs_ioc_getversion(filp, arg); 4560 case F2FS_IOC_START_ATOMIC_WRITE: 4561 return f2fs_ioc_start_atomic_write(filp, false); 4562 case F2FS_IOC_START_ATOMIC_REPLACE: 4563 return f2fs_ioc_start_atomic_write(filp, true); 4564 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4565 return f2fs_ioc_commit_atomic_write(filp); 4566 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4567 return f2fs_ioc_abort_atomic_write(filp); 4568 case F2FS_IOC_START_VOLATILE_WRITE: 4569 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4570 return -EOPNOTSUPP; 4571 case F2FS_IOC_SHUTDOWN: 4572 return f2fs_ioc_shutdown(filp, arg); 4573 case FITRIM: 4574 return f2fs_ioc_fitrim(filp, arg); 4575 case FS_IOC_SET_ENCRYPTION_POLICY: 4576 return f2fs_ioc_set_encryption_policy(filp, arg); 4577 case FS_IOC_GET_ENCRYPTION_POLICY: 4578 return f2fs_ioc_get_encryption_policy(filp, arg); 4579 case FS_IOC_GET_ENCRYPTION_PWSALT: 4580 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4581 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4582 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4583 case FS_IOC_ADD_ENCRYPTION_KEY: 4584 return f2fs_ioc_add_encryption_key(filp, arg); 4585 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4586 return f2fs_ioc_remove_encryption_key(filp, arg); 4587 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4588 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4589 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4590 return f2fs_ioc_get_encryption_key_status(filp, arg); 4591 case FS_IOC_GET_ENCRYPTION_NONCE: 4592 return f2fs_ioc_get_encryption_nonce(filp, arg); 4593 case F2FS_IOC_GARBAGE_COLLECT: 4594 return f2fs_ioc_gc(filp, arg); 4595 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4596 return f2fs_ioc_gc_range(filp, arg); 4597 case F2FS_IOC_WRITE_CHECKPOINT: 4598 return f2fs_ioc_write_checkpoint(filp); 4599 case F2FS_IOC_DEFRAGMENT: 4600 return f2fs_ioc_defragment(filp, arg); 4601 case F2FS_IOC_MOVE_RANGE: 4602 return f2fs_ioc_move_range(filp, arg); 4603 case F2FS_IOC_FLUSH_DEVICE: 4604 return f2fs_ioc_flush_device(filp, arg); 4605 case F2FS_IOC_GET_FEATURES: 4606 return f2fs_ioc_get_features(filp, arg); 4607 case F2FS_IOC_GET_PIN_FILE: 4608 return f2fs_ioc_get_pin_file(filp, arg); 4609 case F2FS_IOC_SET_PIN_FILE: 4610 return f2fs_ioc_set_pin_file(filp, arg); 4611 case F2FS_IOC_PRECACHE_EXTENTS: 4612 return f2fs_ioc_precache_extents(filp); 4613 case F2FS_IOC_RESIZE_FS: 4614 return f2fs_ioc_resize_fs(filp, arg); 4615 case FS_IOC_ENABLE_VERITY: 4616 return f2fs_ioc_enable_verity(filp, arg); 4617 case FS_IOC_MEASURE_VERITY: 4618 return f2fs_ioc_measure_verity(filp, arg); 4619 case FS_IOC_READ_VERITY_METADATA: 4620 return f2fs_ioc_read_verity_metadata(filp, arg); 4621 case FS_IOC_GETFSLABEL: 4622 return f2fs_ioc_getfslabel(filp, arg); 4623 case FS_IOC_SETFSLABEL: 4624 return f2fs_ioc_setfslabel(filp, arg); 4625 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4626 return f2fs_ioc_get_compress_blocks(filp, arg); 4627 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4628 return f2fs_release_compress_blocks(filp, arg); 4629 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4630 return f2fs_reserve_compress_blocks(filp, arg); 4631 case F2FS_IOC_SEC_TRIM_FILE: 4632 return f2fs_sec_trim_file(filp, arg); 4633 case F2FS_IOC_GET_COMPRESS_OPTION: 4634 return f2fs_ioc_get_compress_option(filp, arg); 4635 case F2FS_IOC_SET_COMPRESS_OPTION: 4636 return f2fs_ioc_set_compress_option(filp, arg); 4637 case F2FS_IOC_DECOMPRESS_FILE: 4638 return f2fs_ioc_decompress_file(filp); 4639 case F2FS_IOC_COMPRESS_FILE: 4640 return f2fs_ioc_compress_file(filp); 4641 case F2FS_IOC_GET_DEV_ALIAS_FILE: 4642 return f2fs_ioc_get_dev_alias_file(filp, arg); 4643 case F2FS_IOC_IO_PRIO: 4644 return f2fs_ioc_io_prio(filp, arg); 4645 default: 4646 return -ENOTTY; 4647 } 4648 } 4649 4650 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4651 { 4652 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4653 return -EIO; 4654 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4655 return -ENOSPC; 4656 4657 return __f2fs_ioctl(filp, cmd, arg); 4658 } 4659 4660 /* 4661 * Return %true if the given read or write request should use direct I/O, or 4662 * %false if it should use buffered I/O. 4663 */ 4664 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4665 struct iov_iter *iter) 4666 { 4667 unsigned int align; 4668 4669 if (!(iocb->ki_flags & IOCB_DIRECT)) 4670 return false; 4671 4672 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4673 return false; 4674 4675 /* 4676 * Direct I/O not aligned to the disk's logical_block_size will be 4677 * attempted, but will fail with -EINVAL. 4678 * 4679 * f2fs additionally requires that direct I/O be aligned to the 4680 * filesystem block size, which is often a stricter requirement. 4681 * However, f2fs traditionally falls back to buffered I/O on requests 4682 * that are logical_block_size-aligned but not fs-block aligned. 4683 * 4684 * The below logic implements this behavior. 4685 */ 4686 align = iocb->ki_pos | iov_iter_alignment(iter); 4687 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4688 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4689 return false; 4690 4691 return true; 4692 } 4693 4694 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4695 unsigned int flags) 4696 { 4697 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4698 4699 dec_page_count(sbi, F2FS_DIO_READ); 4700 if (error) 4701 return error; 4702 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4703 return 0; 4704 } 4705 4706 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4707 .end_io = f2fs_dio_read_end_io, 4708 }; 4709 4710 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4711 { 4712 struct file *file = iocb->ki_filp; 4713 struct inode *inode = file_inode(file); 4714 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4715 struct f2fs_inode_info *fi = F2FS_I(inode); 4716 const loff_t pos = iocb->ki_pos; 4717 const size_t count = iov_iter_count(to); 4718 struct iomap_dio *dio; 4719 ssize_t ret; 4720 4721 if (count == 0) 4722 return 0; /* skip atime update */ 4723 4724 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4725 4726 if (iocb->ki_flags & IOCB_NOWAIT) { 4727 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4728 ret = -EAGAIN; 4729 goto out; 4730 } 4731 } else { 4732 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4733 } 4734 4735 /* dio is not compatible w/ atomic file */ 4736 if (f2fs_is_atomic_file(inode)) { 4737 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4738 ret = -EOPNOTSUPP; 4739 goto out; 4740 } 4741 4742 /* 4743 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4744 * the higher-level function iomap_dio_rw() in order to ensure that the 4745 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4746 */ 4747 inc_page_count(sbi, F2FS_DIO_READ); 4748 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4749 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4750 if (IS_ERR_OR_NULL(dio)) { 4751 ret = PTR_ERR_OR_ZERO(dio); 4752 if (ret != -EIOCBQUEUED) 4753 dec_page_count(sbi, F2FS_DIO_READ); 4754 } else { 4755 ret = iomap_dio_complete(dio); 4756 } 4757 4758 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4759 4760 file_accessed(file); 4761 out: 4762 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4763 return ret; 4764 } 4765 4766 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4767 int rw) 4768 { 4769 struct inode *inode = file_inode(file); 4770 char *buf, *path; 4771 4772 buf = f2fs_getname(F2FS_I_SB(inode)); 4773 if (!buf) 4774 return; 4775 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4776 if (IS_ERR(path)) 4777 goto free_buf; 4778 if (rw == WRITE) 4779 trace_f2fs_datawrite_start(inode, pos, count, 4780 current->pid, path, current->comm); 4781 else 4782 trace_f2fs_dataread_start(inode, pos, count, 4783 current->pid, path, current->comm); 4784 free_buf: 4785 f2fs_putname(buf); 4786 } 4787 4788 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4789 { 4790 struct inode *inode = file_inode(iocb->ki_filp); 4791 const loff_t pos = iocb->ki_pos; 4792 ssize_t ret; 4793 4794 if (!f2fs_is_compress_backend_ready(inode)) 4795 return -EOPNOTSUPP; 4796 4797 if (trace_f2fs_dataread_start_enabled()) 4798 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4799 iov_iter_count(to), READ); 4800 4801 /* In LFS mode, if there is inflight dio, wait for its completion */ 4802 if (f2fs_lfs_mode(F2FS_I_SB(inode)) && 4803 get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE)) 4804 inode_dio_wait(inode); 4805 4806 if (f2fs_should_use_dio(inode, iocb, to)) { 4807 ret = f2fs_dio_read_iter(iocb, to); 4808 } else { 4809 ret = filemap_read(iocb, to, 0); 4810 if (ret > 0) 4811 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4812 APP_BUFFERED_READ_IO, ret); 4813 } 4814 if (trace_f2fs_dataread_end_enabled()) 4815 trace_f2fs_dataread_end(inode, pos, ret); 4816 return ret; 4817 } 4818 4819 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4820 struct pipe_inode_info *pipe, 4821 size_t len, unsigned int flags) 4822 { 4823 struct inode *inode = file_inode(in); 4824 const loff_t pos = *ppos; 4825 ssize_t ret; 4826 4827 if (!f2fs_is_compress_backend_ready(inode)) 4828 return -EOPNOTSUPP; 4829 4830 if (trace_f2fs_dataread_start_enabled()) 4831 f2fs_trace_rw_file_path(in, pos, len, READ); 4832 4833 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4834 if (ret > 0) 4835 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4836 APP_BUFFERED_READ_IO, ret); 4837 4838 if (trace_f2fs_dataread_end_enabled()) 4839 trace_f2fs_dataread_end(inode, pos, ret); 4840 return ret; 4841 } 4842 4843 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4844 { 4845 struct file *file = iocb->ki_filp; 4846 struct inode *inode = file_inode(file); 4847 ssize_t count; 4848 int err; 4849 4850 if (IS_IMMUTABLE(inode)) 4851 return -EPERM; 4852 4853 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4854 return -EPERM; 4855 4856 count = generic_write_checks(iocb, from); 4857 if (count <= 0) 4858 return count; 4859 4860 err = file_modified(file); 4861 if (err) 4862 return err; 4863 return count; 4864 } 4865 4866 /* 4867 * Preallocate blocks for a write request, if it is possible and helpful to do 4868 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4869 * blocks were preallocated, or a negative errno value if something went 4870 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4871 * requested blocks (not just some of them) have been allocated. 4872 */ 4873 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4874 bool dio) 4875 { 4876 struct inode *inode = file_inode(iocb->ki_filp); 4877 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4878 const loff_t pos = iocb->ki_pos; 4879 const size_t count = iov_iter_count(iter); 4880 struct f2fs_map_blocks map = {}; 4881 int flag; 4882 int ret; 4883 4884 /* If it will be an out-of-place direct write, don't bother. */ 4885 if (dio && f2fs_lfs_mode(sbi)) 4886 return 0; 4887 /* 4888 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4889 * buffered IO, if DIO meets any holes. 4890 */ 4891 if (dio && i_size_read(inode) && 4892 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4893 return 0; 4894 4895 /* No-wait I/O can't allocate blocks. */ 4896 if (iocb->ki_flags & IOCB_NOWAIT) 4897 return 0; 4898 4899 /* If it will be a short write, don't bother. */ 4900 if (fault_in_iov_iter_readable(iter, count)) 4901 return 0; 4902 4903 if (f2fs_has_inline_data(inode)) { 4904 /* If the data will fit inline, don't bother. */ 4905 if (pos + count <= MAX_INLINE_DATA(inode)) 4906 return 0; 4907 ret = f2fs_convert_inline_inode(inode); 4908 if (ret) 4909 return ret; 4910 } 4911 4912 /* Do not preallocate blocks that will be written partially in 4KB. */ 4913 map.m_lblk = F2FS_BLK_ALIGN(pos); 4914 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4915 if (map.m_len > map.m_lblk) 4916 map.m_len -= map.m_lblk; 4917 else 4918 return 0; 4919 4920 if (!IS_DEVICE_ALIASING(inode)) 4921 map.m_may_create = true; 4922 if (dio) { 4923 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi, 4924 inode->i_write_hint); 4925 flag = F2FS_GET_BLOCK_PRE_DIO; 4926 } else { 4927 map.m_seg_type = NO_CHECK_TYPE; 4928 flag = F2FS_GET_BLOCK_PRE_AIO; 4929 } 4930 4931 ret = f2fs_map_blocks(inode, &map, flag); 4932 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4933 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4934 return ret; 4935 if (ret == 0) 4936 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4937 return map.m_len; 4938 } 4939 4940 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4941 struct iov_iter *from) 4942 { 4943 struct file *file = iocb->ki_filp; 4944 struct inode *inode = file_inode(file); 4945 ssize_t ret; 4946 4947 if (iocb->ki_flags & IOCB_NOWAIT) 4948 return -EOPNOTSUPP; 4949 4950 ret = generic_perform_write(iocb, from); 4951 4952 if (ret > 0) { 4953 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4954 APP_BUFFERED_IO, ret); 4955 } 4956 return ret; 4957 } 4958 4959 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4960 unsigned int flags) 4961 { 4962 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4963 4964 dec_page_count(sbi, F2FS_DIO_WRITE); 4965 if (error) 4966 return error; 4967 f2fs_update_time(sbi, REQ_TIME); 4968 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4969 return 0; 4970 } 4971 4972 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter, 4973 struct bio *bio, loff_t file_offset) 4974 { 4975 struct inode *inode = iter->inode; 4976 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4977 enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint); 4978 enum temp_type temp = f2fs_get_segment_temp(sbi, type); 4979 4980 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp); 4981 submit_bio(bio); 4982 } 4983 4984 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4985 .end_io = f2fs_dio_write_end_io, 4986 .submit_io = f2fs_dio_write_submit_io, 4987 }; 4988 4989 static void f2fs_flush_buffered_write(struct address_space *mapping, 4990 loff_t start_pos, loff_t end_pos) 4991 { 4992 int ret; 4993 4994 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4995 if (ret < 0) 4996 return; 4997 invalidate_mapping_pages(mapping, 4998 start_pos >> PAGE_SHIFT, 4999 end_pos >> PAGE_SHIFT); 5000 } 5001 5002 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 5003 bool *may_need_sync) 5004 { 5005 struct file *file = iocb->ki_filp; 5006 struct inode *inode = file_inode(file); 5007 struct f2fs_inode_info *fi = F2FS_I(inode); 5008 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 5009 const bool do_opu = f2fs_lfs_mode(sbi); 5010 const loff_t pos = iocb->ki_pos; 5011 const ssize_t count = iov_iter_count(from); 5012 unsigned int dio_flags; 5013 struct iomap_dio *dio; 5014 ssize_t ret; 5015 5016 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 5017 5018 if (iocb->ki_flags & IOCB_NOWAIT) { 5019 /* f2fs_convert_inline_inode() and block allocation can block */ 5020 if (f2fs_has_inline_data(inode) || 5021 !f2fs_overwrite_io(inode, pos, count)) { 5022 ret = -EAGAIN; 5023 goto out; 5024 } 5025 5026 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 5027 ret = -EAGAIN; 5028 goto out; 5029 } 5030 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 5031 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5032 ret = -EAGAIN; 5033 goto out; 5034 } 5035 } else { 5036 ret = f2fs_convert_inline_inode(inode); 5037 if (ret) 5038 goto out; 5039 5040 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 5041 if (do_opu) 5042 f2fs_down_read(&fi->i_gc_rwsem[READ]); 5043 } 5044 5045 /* 5046 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 5047 * the higher-level function iomap_dio_rw() in order to ensure that the 5048 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 5049 */ 5050 inc_page_count(sbi, F2FS_DIO_WRITE); 5051 dio_flags = 0; 5052 if (pos + count > inode->i_size) 5053 dio_flags |= IOMAP_DIO_FORCE_WAIT; 5054 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 5055 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 5056 if (IS_ERR_OR_NULL(dio)) { 5057 ret = PTR_ERR_OR_ZERO(dio); 5058 if (ret == -ENOTBLK) 5059 ret = 0; 5060 if (ret != -EIOCBQUEUED) 5061 dec_page_count(sbi, F2FS_DIO_WRITE); 5062 } else { 5063 ret = iomap_dio_complete(dio); 5064 } 5065 5066 if (do_opu) 5067 f2fs_up_read(&fi->i_gc_rwsem[READ]); 5068 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5069 5070 if (ret < 0) 5071 goto out; 5072 if (pos + ret > inode->i_size) 5073 f2fs_i_size_write(inode, pos + ret); 5074 if (!do_opu) 5075 set_inode_flag(inode, FI_UPDATE_WRITE); 5076 5077 if (iov_iter_count(from)) { 5078 ssize_t ret2; 5079 loff_t bufio_start_pos = iocb->ki_pos; 5080 5081 /* 5082 * The direct write was partial, so we need to fall back to a 5083 * buffered write for the remainder. 5084 */ 5085 5086 ret2 = f2fs_buffered_write_iter(iocb, from); 5087 if (iov_iter_count(from)) 5088 f2fs_write_failed(inode, iocb->ki_pos); 5089 if (ret2 < 0) 5090 goto out; 5091 5092 /* 5093 * Ensure that the pagecache pages are written to disk and 5094 * invalidated to preserve the expected O_DIRECT semantics. 5095 */ 5096 if (ret2 > 0) { 5097 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 5098 5099 ret += ret2; 5100 5101 f2fs_flush_buffered_write(file->f_mapping, 5102 bufio_start_pos, 5103 bufio_end_pos); 5104 } 5105 } else { 5106 /* iomap_dio_rw() already handled the generic_write_sync(). */ 5107 *may_need_sync = false; 5108 } 5109 out: 5110 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 5111 return ret; 5112 } 5113 5114 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 5115 { 5116 struct inode *inode = file_inode(iocb->ki_filp); 5117 const loff_t orig_pos = iocb->ki_pos; 5118 const size_t orig_count = iov_iter_count(from); 5119 loff_t target_size; 5120 bool dio; 5121 bool may_need_sync = true; 5122 int preallocated; 5123 const loff_t pos = iocb->ki_pos; 5124 const ssize_t count = iov_iter_count(from); 5125 ssize_t ret; 5126 5127 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 5128 ret = -EIO; 5129 goto out; 5130 } 5131 5132 if (!f2fs_is_compress_backend_ready(inode)) { 5133 ret = -EOPNOTSUPP; 5134 goto out; 5135 } 5136 5137 if (iocb->ki_flags & IOCB_NOWAIT) { 5138 if (!inode_trylock(inode)) { 5139 ret = -EAGAIN; 5140 goto out; 5141 } 5142 } else { 5143 inode_lock(inode); 5144 } 5145 5146 if (f2fs_is_pinned_file(inode) && 5147 !f2fs_overwrite_io(inode, pos, count)) { 5148 ret = -EIO; 5149 goto out_unlock; 5150 } 5151 5152 ret = f2fs_write_checks(iocb, from); 5153 if (ret <= 0) 5154 goto out_unlock; 5155 5156 /* Determine whether we will do a direct write or a buffered write. */ 5157 dio = f2fs_should_use_dio(inode, iocb, from); 5158 5159 /* dio is not compatible w/ atomic write */ 5160 if (dio && f2fs_is_atomic_file(inode)) { 5161 ret = -EOPNOTSUPP; 5162 goto out_unlock; 5163 } 5164 5165 /* Possibly preallocate the blocks for the write. */ 5166 target_size = iocb->ki_pos + iov_iter_count(from); 5167 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 5168 if (preallocated < 0) { 5169 ret = preallocated; 5170 } else { 5171 if (trace_f2fs_datawrite_start_enabled()) 5172 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 5173 orig_count, WRITE); 5174 5175 /* Do the actual write. */ 5176 ret = dio ? 5177 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 5178 f2fs_buffered_write_iter(iocb, from); 5179 5180 if (trace_f2fs_datawrite_end_enabled()) 5181 trace_f2fs_datawrite_end(inode, orig_pos, ret); 5182 } 5183 5184 /* Don't leave any preallocated blocks around past i_size. */ 5185 if (preallocated && i_size_read(inode) < target_size) { 5186 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5187 filemap_invalidate_lock(inode->i_mapping); 5188 if (!f2fs_truncate(inode)) 5189 file_dont_truncate(inode); 5190 filemap_invalidate_unlock(inode->i_mapping); 5191 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5192 } else { 5193 file_dont_truncate(inode); 5194 } 5195 5196 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 5197 out_unlock: 5198 inode_unlock(inode); 5199 out: 5200 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 5201 5202 if (ret > 0 && may_need_sync) 5203 ret = generic_write_sync(iocb, ret); 5204 5205 /* If buffered IO was forced, flush and drop the data from 5206 * the page cache to preserve O_DIRECT semantics 5207 */ 5208 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 5209 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 5210 orig_pos, 5211 orig_pos + ret - 1); 5212 5213 return ret; 5214 } 5215 5216 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 5217 int advice) 5218 { 5219 struct address_space *mapping; 5220 struct backing_dev_info *bdi; 5221 struct inode *inode = file_inode(filp); 5222 int err; 5223 5224 if (advice == POSIX_FADV_SEQUENTIAL) { 5225 if (S_ISFIFO(inode->i_mode)) 5226 return -ESPIPE; 5227 5228 mapping = filp->f_mapping; 5229 if (!mapping || len < 0) 5230 return -EINVAL; 5231 5232 bdi = inode_to_bdi(mapping->host); 5233 filp->f_ra.ra_pages = bdi->ra_pages * 5234 F2FS_I_SB(inode)->seq_file_ra_mul; 5235 spin_lock(&filp->f_lock); 5236 filp->f_mode &= ~FMODE_RANDOM; 5237 spin_unlock(&filp->f_lock); 5238 return 0; 5239 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 5240 /* Load extent cache at the first readahead. */ 5241 f2fs_precache_extents(inode); 5242 } 5243 5244 err = generic_fadvise(filp, offset, len, advice); 5245 if (err) 5246 return err; 5247 5248 if (advice == POSIX_FADV_DONTNEED && 5249 (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 5250 f2fs_compressed_file(inode))) 5251 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 5252 else if (advice == POSIX_FADV_NOREUSE) 5253 err = f2fs_keep_noreuse_range(inode, offset, len); 5254 return err; 5255 } 5256 5257 #ifdef CONFIG_COMPAT 5258 struct compat_f2fs_gc_range { 5259 u32 sync; 5260 compat_u64 start; 5261 compat_u64 len; 5262 }; 5263 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 5264 struct compat_f2fs_gc_range) 5265 5266 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 5267 { 5268 struct compat_f2fs_gc_range __user *urange; 5269 struct f2fs_gc_range range; 5270 int err; 5271 5272 urange = compat_ptr(arg); 5273 err = get_user(range.sync, &urange->sync); 5274 err |= get_user(range.start, &urange->start); 5275 err |= get_user(range.len, &urange->len); 5276 if (err) 5277 return -EFAULT; 5278 5279 return __f2fs_ioc_gc_range(file, &range); 5280 } 5281 5282 struct compat_f2fs_move_range { 5283 u32 dst_fd; 5284 compat_u64 pos_in; 5285 compat_u64 pos_out; 5286 compat_u64 len; 5287 }; 5288 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 5289 struct compat_f2fs_move_range) 5290 5291 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 5292 { 5293 struct compat_f2fs_move_range __user *urange; 5294 struct f2fs_move_range range; 5295 int err; 5296 5297 urange = compat_ptr(arg); 5298 err = get_user(range.dst_fd, &urange->dst_fd); 5299 err |= get_user(range.pos_in, &urange->pos_in); 5300 err |= get_user(range.pos_out, &urange->pos_out); 5301 err |= get_user(range.len, &urange->len); 5302 if (err) 5303 return -EFAULT; 5304 5305 return __f2fs_ioc_move_range(file, &range); 5306 } 5307 5308 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5309 { 5310 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 5311 return -EIO; 5312 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 5313 return -ENOSPC; 5314 5315 switch (cmd) { 5316 case FS_IOC32_GETVERSION: 5317 cmd = FS_IOC_GETVERSION; 5318 break; 5319 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 5320 return f2fs_compat_ioc_gc_range(file, arg); 5321 case F2FS_IOC32_MOVE_RANGE: 5322 return f2fs_compat_ioc_move_range(file, arg); 5323 case F2FS_IOC_START_ATOMIC_WRITE: 5324 case F2FS_IOC_START_ATOMIC_REPLACE: 5325 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5326 case F2FS_IOC_START_VOLATILE_WRITE: 5327 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5328 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5329 case F2FS_IOC_SHUTDOWN: 5330 case FITRIM: 5331 case FS_IOC_SET_ENCRYPTION_POLICY: 5332 case FS_IOC_GET_ENCRYPTION_PWSALT: 5333 case FS_IOC_GET_ENCRYPTION_POLICY: 5334 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5335 case FS_IOC_ADD_ENCRYPTION_KEY: 5336 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5337 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5338 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5339 case FS_IOC_GET_ENCRYPTION_NONCE: 5340 case F2FS_IOC_GARBAGE_COLLECT: 5341 case F2FS_IOC_WRITE_CHECKPOINT: 5342 case F2FS_IOC_DEFRAGMENT: 5343 case F2FS_IOC_FLUSH_DEVICE: 5344 case F2FS_IOC_GET_FEATURES: 5345 case F2FS_IOC_GET_PIN_FILE: 5346 case F2FS_IOC_SET_PIN_FILE: 5347 case F2FS_IOC_PRECACHE_EXTENTS: 5348 case F2FS_IOC_RESIZE_FS: 5349 case FS_IOC_ENABLE_VERITY: 5350 case FS_IOC_MEASURE_VERITY: 5351 case FS_IOC_READ_VERITY_METADATA: 5352 case FS_IOC_GETFSLABEL: 5353 case FS_IOC_SETFSLABEL: 5354 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5355 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5356 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5357 case F2FS_IOC_SEC_TRIM_FILE: 5358 case F2FS_IOC_GET_COMPRESS_OPTION: 5359 case F2FS_IOC_SET_COMPRESS_OPTION: 5360 case F2FS_IOC_DECOMPRESS_FILE: 5361 case F2FS_IOC_COMPRESS_FILE: 5362 case F2FS_IOC_GET_DEV_ALIAS_FILE: 5363 case F2FS_IOC_IO_PRIO: 5364 break; 5365 default: 5366 return -ENOIOCTLCMD; 5367 } 5368 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5369 } 5370 #endif 5371 5372 const struct file_operations f2fs_file_operations = { 5373 .llseek = f2fs_llseek, 5374 .read_iter = f2fs_file_read_iter, 5375 .write_iter = f2fs_file_write_iter, 5376 .iopoll = iocb_bio_iopoll, 5377 .open = f2fs_file_open, 5378 .release = f2fs_release_file, 5379 .mmap = f2fs_file_mmap, 5380 .flush = f2fs_file_flush, 5381 .fsync = f2fs_sync_file, 5382 .fallocate = f2fs_fallocate, 5383 .unlocked_ioctl = f2fs_ioctl, 5384 #ifdef CONFIG_COMPAT 5385 .compat_ioctl = f2fs_compat_ioctl, 5386 #endif 5387 .splice_read = f2fs_file_splice_read, 5388 .splice_write = iter_file_splice_write, 5389 .fadvise = f2fs_file_fadvise, 5390 .fop_flags = FOP_BUFFER_RASYNC, 5391 }; 5392