1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/file.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/fs.h> 9 #include <linux/f2fs_fs.h> 10 #include <linux/stat.h> 11 #include <linux/writeback.h> 12 #include <linux/blkdev.h> 13 #include <linux/falloc.h> 14 #include <linux/types.h> 15 #include <linux/compat.h> 16 #include <linux/uaccess.h> 17 #include <linux/mount.h> 18 #include <linux/pagevec.h> 19 #include <linux/uio.h> 20 #include <linux/uuid.h> 21 #include <linux/file.h> 22 #include <linux/nls.h> 23 #include <linux/sched/signal.h> 24 #include <linux/fileattr.h> 25 #include <linux/fadvise.h> 26 #include <linux/iomap.h> 27 28 #include "f2fs.h" 29 #include "node.h" 30 #include "segment.h" 31 #include "xattr.h" 32 #include "acl.h" 33 #include "gc.h" 34 #include "iostat.h" 35 #include <trace/events/f2fs.h> 36 #include <uapi/linux/f2fs.h> 37 38 static void f2fs_zero_post_eof_page(struct inode *inode, loff_t new_size) 39 { 40 loff_t old_size = i_size_read(inode); 41 42 if (old_size >= new_size) 43 return; 44 45 /* zero or drop pages only in range of [old_size, new_size] */ 46 truncate_pagecache(inode, old_size); 47 } 48 49 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 50 { 51 struct inode *inode = file_inode(vmf->vma->vm_file); 52 vm_flags_t flags = vmf->vma->vm_flags; 53 vm_fault_t ret; 54 55 ret = filemap_fault(vmf); 56 if (ret & VM_FAULT_LOCKED) 57 f2fs_update_iostat(F2FS_I_SB(inode), inode, 58 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 59 60 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret); 61 62 return ret; 63 } 64 65 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 66 { 67 struct folio *folio = page_folio(vmf->page); 68 struct inode *inode = file_inode(vmf->vma->vm_file); 69 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 70 struct dnode_of_data dn; 71 bool need_alloc = !f2fs_is_pinned_file(inode); 72 int err = 0; 73 vm_fault_t ret; 74 75 if (unlikely(IS_IMMUTABLE(inode))) 76 return VM_FAULT_SIGBUS; 77 78 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 79 err = -EIO; 80 goto out; 81 } 82 83 if (unlikely(f2fs_cp_error(sbi))) { 84 err = -EIO; 85 goto out; 86 } 87 88 if (!f2fs_is_checkpoint_ready(sbi)) { 89 err = -ENOSPC; 90 goto out; 91 } 92 93 err = f2fs_convert_inline_inode(inode); 94 if (err) 95 goto out; 96 97 #ifdef CONFIG_F2FS_FS_COMPRESSION 98 if (f2fs_compressed_file(inode)) { 99 int ret = f2fs_is_compressed_cluster(inode, folio->index); 100 101 if (ret < 0) { 102 err = ret; 103 goto out; 104 } else if (ret) { 105 need_alloc = false; 106 } 107 } 108 #endif 109 /* should do out of any locked page */ 110 if (need_alloc) 111 f2fs_balance_fs(sbi, true); 112 113 sb_start_pagefault(inode->i_sb); 114 115 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 116 117 filemap_invalidate_lock(inode->i_mapping); 118 f2fs_zero_post_eof_page(inode, (folio->index + 1) << PAGE_SHIFT); 119 filemap_invalidate_unlock(inode->i_mapping); 120 121 file_update_time(vmf->vma->vm_file); 122 filemap_invalidate_lock_shared(inode->i_mapping); 123 124 folio_lock(folio); 125 if (unlikely(folio->mapping != inode->i_mapping || 126 folio_pos(folio) > i_size_read(inode) || 127 !folio_test_uptodate(folio))) { 128 folio_unlock(folio); 129 err = -EFAULT; 130 goto out_sem; 131 } 132 133 set_new_dnode(&dn, inode, NULL, NULL, 0); 134 if (need_alloc) { 135 /* block allocation */ 136 err = f2fs_get_block_locked(&dn, folio->index); 137 } else { 138 err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE); 139 f2fs_put_dnode(&dn); 140 if (f2fs_is_pinned_file(inode) && 141 !__is_valid_data_blkaddr(dn.data_blkaddr)) 142 err = -EIO; 143 } 144 145 if (err) { 146 folio_unlock(folio); 147 goto out_sem; 148 } 149 150 f2fs_folio_wait_writeback(folio, DATA, false, true); 151 152 /* wait for GCed page writeback via META_MAPPING */ 153 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 154 155 /* 156 * check to see if the page is mapped already (no holes) 157 */ 158 if (folio_test_mappedtodisk(folio)) 159 goto out_sem; 160 161 /* page is wholly or partially inside EOF */ 162 if (((loff_t)(folio->index + 1) << PAGE_SHIFT) > 163 i_size_read(inode)) { 164 loff_t offset; 165 166 offset = i_size_read(inode) & ~PAGE_MASK; 167 folio_zero_segment(folio, offset, folio_size(folio)); 168 } 169 folio_mark_dirty(folio); 170 171 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 172 f2fs_update_time(sbi, REQ_TIME); 173 174 out_sem: 175 filemap_invalidate_unlock_shared(inode->i_mapping); 176 177 sb_end_pagefault(inode->i_sb); 178 out: 179 ret = vmf_fs_error(err); 180 181 trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret); 182 return ret; 183 } 184 185 static const struct vm_operations_struct f2fs_file_vm_ops = { 186 .fault = f2fs_filemap_fault, 187 .map_pages = filemap_map_pages, 188 .page_mkwrite = f2fs_vm_page_mkwrite, 189 }; 190 191 static int get_parent_ino(struct inode *inode, nid_t *pino) 192 { 193 struct dentry *dentry; 194 195 /* 196 * Make sure to get the non-deleted alias. The alias associated with 197 * the open file descriptor being fsync()'ed may be deleted already. 198 */ 199 dentry = d_find_alias(inode); 200 if (!dentry) 201 return 0; 202 203 *pino = d_parent_ino(dentry); 204 dput(dentry); 205 return 1; 206 } 207 208 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 209 { 210 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 211 enum cp_reason_type cp_reason = CP_NO_NEEDED; 212 213 if (!S_ISREG(inode->i_mode)) 214 cp_reason = CP_NON_REGULAR; 215 else if (f2fs_compressed_file(inode)) 216 cp_reason = CP_COMPRESSED; 217 else if (inode->i_nlink != 1) 218 cp_reason = CP_HARDLINK; 219 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 220 cp_reason = CP_SB_NEED_CP; 221 else if (file_wrong_pino(inode)) 222 cp_reason = CP_WRONG_PINO; 223 else if (!f2fs_space_for_roll_forward(sbi)) 224 cp_reason = CP_NO_SPC_ROLL; 225 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 226 cp_reason = CP_NODE_NEED_CP; 227 else if (test_opt(sbi, FASTBOOT)) 228 cp_reason = CP_FASTBOOT_MODE; 229 else if (F2FS_OPTION(sbi).active_logs == 2) 230 cp_reason = CP_SPEC_LOG_NUM; 231 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 232 f2fs_need_dentry_mark(sbi, inode->i_ino) && 233 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 234 TRANS_DIR_INO)) 235 cp_reason = CP_RECOVER_DIR; 236 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 237 XATTR_DIR_INO)) 238 cp_reason = CP_XATTR_DIR; 239 240 return cp_reason; 241 } 242 243 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 244 { 245 struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino); 246 bool ret = false; 247 /* But we need to avoid that there are some inode updates */ 248 if ((!IS_ERR(i) && folio_test_dirty(i)) || 249 f2fs_need_inode_block_update(sbi, ino)) 250 ret = true; 251 f2fs_folio_put(i, false); 252 return ret; 253 } 254 255 static void try_to_fix_pino(struct inode *inode) 256 { 257 struct f2fs_inode_info *fi = F2FS_I(inode); 258 nid_t pino; 259 260 f2fs_down_write(&fi->i_sem); 261 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 262 get_parent_ino(inode, &pino)) { 263 f2fs_i_pino_write(inode, pino); 264 file_got_pino(inode); 265 } 266 f2fs_up_write(&fi->i_sem); 267 } 268 269 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 270 int datasync, bool atomic) 271 { 272 struct inode *inode = file->f_mapping->host; 273 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 274 nid_t ino = inode->i_ino; 275 int ret = 0; 276 enum cp_reason_type cp_reason = 0; 277 struct writeback_control wbc = { 278 .sync_mode = WB_SYNC_ALL, 279 .nr_to_write = LONG_MAX, 280 }; 281 unsigned int seq_id = 0; 282 283 if (unlikely(f2fs_readonly(inode->i_sb))) 284 return 0; 285 286 trace_f2fs_sync_file_enter(inode); 287 288 if (S_ISDIR(inode->i_mode)) 289 goto go_write; 290 291 /* if fdatasync is triggered, let's do in-place-update */ 292 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 293 set_inode_flag(inode, FI_NEED_IPU); 294 ret = file_write_and_wait_range(file, start, end); 295 clear_inode_flag(inode, FI_NEED_IPU); 296 297 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 298 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 299 return ret; 300 } 301 302 /* if the inode is dirty, let's recover all the time */ 303 if (!f2fs_skip_inode_update(inode, datasync)) { 304 f2fs_write_inode(inode, NULL); 305 goto go_write; 306 } 307 308 /* 309 * if there is no written data, don't waste time to write recovery info. 310 */ 311 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 312 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 313 314 /* it may call write_inode just prior to fsync */ 315 if (need_inode_page_update(sbi, ino)) 316 goto go_write; 317 318 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 319 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 320 goto flush_out; 321 goto out; 322 } else { 323 /* 324 * for OPU case, during fsync(), node can be persisted before 325 * data when lower device doesn't support write barrier, result 326 * in data corruption after SPO. 327 * So for strict fsync mode, force to use atomic write semantics 328 * to keep write order in between data/node and last node to 329 * avoid potential data corruption. 330 */ 331 if (F2FS_OPTION(sbi).fsync_mode == 332 FSYNC_MODE_STRICT && !atomic) 333 atomic = true; 334 } 335 go_write: 336 /* 337 * Both of fdatasync() and fsync() are able to be recovered from 338 * sudden-power-off. 339 */ 340 f2fs_down_read(&F2FS_I(inode)->i_sem); 341 cp_reason = need_do_checkpoint(inode); 342 f2fs_up_read(&F2FS_I(inode)->i_sem); 343 344 if (cp_reason) { 345 /* all the dirty node pages should be flushed for POR */ 346 ret = f2fs_sync_fs(inode->i_sb, 1); 347 348 /* 349 * We've secured consistency through sync_fs. Following pino 350 * will be used only for fsynced inodes after checkpoint. 351 */ 352 try_to_fix_pino(inode); 353 clear_inode_flag(inode, FI_APPEND_WRITE); 354 clear_inode_flag(inode, FI_UPDATE_WRITE); 355 goto out; 356 } 357 sync_nodes: 358 atomic_inc(&sbi->wb_sync_req[NODE]); 359 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 360 atomic_dec(&sbi->wb_sync_req[NODE]); 361 if (ret) 362 goto out; 363 364 /* if cp_error was enabled, we should avoid infinite loop */ 365 if (unlikely(f2fs_cp_error(sbi))) { 366 ret = -EIO; 367 goto out; 368 } 369 370 if (f2fs_need_inode_block_update(sbi, ino)) { 371 f2fs_mark_inode_dirty_sync(inode, true); 372 f2fs_write_inode(inode, NULL); 373 goto sync_nodes; 374 } 375 376 /* 377 * If it's atomic_write, it's just fine to keep write ordering. So 378 * here we don't need to wait for node write completion, since we use 379 * node chain which serializes node blocks. If one of node writes are 380 * reordered, we can see simply broken chain, resulting in stopping 381 * roll-forward recovery. It means we'll recover all or none node blocks 382 * given fsync mark. 383 */ 384 if (!atomic) { 385 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 386 if (ret) 387 goto out; 388 } 389 390 /* once recovery info is written, don't need to tack this */ 391 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 392 clear_inode_flag(inode, FI_APPEND_WRITE); 393 flush_out: 394 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 395 ret = f2fs_issue_flush(sbi, inode->i_ino); 396 if (!ret) { 397 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 398 clear_inode_flag(inode, FI_UPDATE_WRITE); 399 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 400 } 401 f2fs_update_time(sbi, REQ_TIME); 402 out: 403 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 404 return ret; 405 } 406 407 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 408 { 409 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 410 return -EIO; 411 return f2fs_do_sync_file(file, start, end, datasync, false); 412 } 413 414 static bool __found_offset(struct address_space *mapping, 415 struct dnode_of_data *dn, pgoff_t index, int whence) 416 { 417 block_t blkaddr = f2fs_data_blkaddr(dn); 418 struct inode *inode = mapping->host; 419 bool compressed_cluster = false; 420 421 if (f2fs_compressed_file(inode)) { 422 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio, 423 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size)); 424 425 compressed_cluster = first_blkaddr == COMPRESS_ADDR; 426 } 427 428 switch (whence) { 429 case SEEK_DATA: 430 if (__is_valid_data_blkaddr(blkaddr)) 431 return true; 432 if (blkaddr == NEW_ADDR && 433 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 434 return true; 435 if (compressed_cluster) 436 return true; 437 break; 438 case SEEK_HOLE: 439 if (compressed_cluster) 440 return false; 441 if (blkaddr == NULL_ADDR) 442 return true; 443 break; 444 } 445 return false; 446 } 447 448 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 449 { 450 struct inode *inode = file->f_mapping->host; 451 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 452 struct dnode_of_data dn; 453 pgoff_t pgofs, end_offset; 454 loff_t data_ofs = offset; 455 loff_t isize; 456 int err = 0; 457 458 inode_lock_shared(inode); 459 460 isize = i_size_read(inode); 461 if (offset >= isize) 462 goto fail; 463 464 /* handle inline data case */ 465 if (f2fs_has_inline_data(inode)) { 466 if (whence == SEEK_HOLE) { 467 data_ofs = isize; 468 goto found; 469 } else if (whence == SEEK_DATA) { 470 data_ofs = offset; 471 goto found; 472 } 473 } 474 475 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 476 477 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 478 set_new_dnode(&dn, inode, NULL, NULL, 0); 479 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 480 if (err && err != -ENOENT) { 481 goto fail; 482 } else if (err == -ENOENT) { 483 /* direct node does not exists */ 484 if (whence == SEEK_DATA) { 485 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 486 continue; 487 } else { 488 goto found; 489 } 490 } 491 492 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 493 494 /* find data/hole in dnode block */ 495 for (; dn.ofs_in_node < end_offset; 496 dn.ofs_in_node++, pgofs++, 497 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 498 block_t blkaddr; 499 500 blkaddr = f2fs_data_blkaddr(&dn); 501 502 if (__is_valid_data_blkaddr(blkaddr) && 503 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 504 blkaddr, DATA_GENERIC_ENHANCE)) { 505 f2fs_put_dnode(&dn); 506 goto fail; 507 } 508 509 if (__found_offset(file->f_mapping, &dn, 510 pgofs, whence)) { 511 f2fs_put_dnode(&dn); 512 goto found; 513 } 514 } 515 f2fs_put_dnode(&dn); 516 } 517 518 if (whence == SEEK_DATA) 519 goto fail; 520 found: 521 if (whence == SEEK_HOLE && data_ofs > isize) 522 data_ofs = isize; 523 inode_unlock_shared(inode); 524 return vfs_setpos(file, data_ofs, maxbytes); 525 fail: 526 inode_unlock_shared(inode); 527 return -ENXIO; 528 } 529 530 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 531 { 532 struct inode *inode = file->f_mapping->host; 533 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 534 535 switch (whence) { 536 case SEEK_SET: 537 case SEEK_CUR: 538 case SEEK_END: 539 return generic_file_llseek_size(file, offset, whence, 540 maxbytes, i_size_read(inode)); 541 case SEEK_DATA: 542 case SEEK_HOLE: 543 if (offset < 0) 544 return -ENXIO; 545 return f2fs_seek_block(file, offset, whence); 546 } 547 548 return -EINVAL; 549 } 550 551 static int f2fs_file_mmap_prepare(struct vm_area_desc *desc) 552 { 553 struct file *file = desc->file; 554 struct inode *inode = file_inode(file); 555 556 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 557 return -EIO; 558 559 if (!f2fs_is_compress_backend_ready(inode)) 560 return -EOPNOTSUPP; 561 562 file_accessed(file); 563 desc->vm_ops = &f2fs_file_vm_ops; 564 565 f2fs_down_read(&F2FS_I(inode)->i_sem); 566 set_inode_flag(inode, FI_MMAP_FILE); 567 f2fs_up_read(&F2FS_I(inode)->i_sem); 568 569 return 0; 570 } 571 572 static int finish_preallocate_blocks(struct inode *inode) 573 { 574 int ret = 0; 575 bool opened; 576 577 f2fs_down_read(&F2FS_I(inode)->i_sem); 578 opened = is_inode_flag_set(inode, FI_OPENED_FILE); 579 f2fs_up_read(&F2FS_I(inode)->i_sem); 580 if (opened) 581 return 0; 582 583 inode_lock(inode); 584 if (is_inode_flag_set(inode, FI_OPENED_FILE)) 585 goto out_unlock; 586 587 if (!file_should_truncate(inode)) 588 goto out_update; 589 590 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 591 filemap_invalidate_lock(inode->i_mapping); 592 593 truncate_setsize(inode, i_size_read(inode)); 594 ret = f2fs_truncate(inode); 595 596 filemap_invalidate_unlock(inode->i_mapping); 597 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 598 if (ret) 599 goto out_unlock; 600 601 file_dont_truncate(inode); 602 out_update: 603 f2fs_down_write(&F2FS_I(inode)->i_sem); 604 set_inode_flag(inode, FI_OPENED_FILE); 605 f2fs_up_write(&F2FS_I(inode)->i_sem); 606 out_unlock: 607 inode_unlock(inode); 608 return ret; 609 } 610 611 static int f2fs_file_open(struct inode *inode, struct file *filp) 612 { 613 int err = fscrypt_file_open(inode, filp); 614 615 if (err) 616 return err; 617 618 if (!f2fs_is_compress_backend_ready(inode)) 619 return -EOPNOTSUPP; 620 621 err = fsverity_file_open(inode, filp); 622 if (err) 623 return err; 624 625 filp->f_mode |= FMODE_NOWAIT; 626 filp->f_mode |= FMODE_CAN_ODIRECT; 627 628 err = dquot_file_open(inode, filp); 629 if (err) 630 return err; 631 632 return finish_preallocate_blocks(inode); 633 } 634 635 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 636 { 637 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 638 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 639 __le32 *addr; 640 bool compressed_cluster = false; 641 int cluster_index = 0, valid_blocks = 0; 642 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 643 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 644 block_t blkstart; 645 int blklen = 0; 646 647 addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs; 648 blkstart = le32_to_cpu(*addr); 649 650 /* Assumption: truncation starts with cluster */ 651 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 652 block_t blkaddr = le32_to_cpu(*addr); 653 654 if (f2fs_compressed_file(dn->inode) && 655 !(cluster_index & (cluster_size - 1))) { 656 if (compressed_cluster) 657 f2fs_i_compr_blocks_update(dn->inode, 658 valid_blocks, false); 659 compressed_cluster = (blkaddr == COMPRESS_ADDR); 660 valid_blocks = 0; 661 } 662 663 if (blkaddr == NULL_ADDR) 664 goto next; 665 666 f2fs_set_data_blkaddr(dn, NULL_ADDR); 667 668 if (__is_valid_data_blkaddr(blkaddr)) { 669 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE)) 670 goto next; 671 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr, 672 DATA_GENERIC_ENHANCE)) 673 goto next; 674 if (compressed_cluster) 675 valid_blocks++; 676 } 677 678 if (blkstart + blklen == blkaddr) { 679 blklen++; 680 } else { 681 f2fs_invalidate_blocks(sbi, blkstart, blklen); 682 blkstart = blkaddr; 683 blklen = 1; 684 } 685 686 if (!released || blkaddr != COMPRESS_ADDR) 687 nr_free++; 688 689 continue; 690 691 next: 692 if (blklen) 693 f2fs_invalidate_blocks(sbi, blkstart, blklen); 694 695 blkstart = le32_to_cpu(*(addr + 1)); 696 blklen = 0; 697 } 698 699 if (blklen) 700 f2fs_invalidate_blocks(sbi, blkstart, blklen); 701 702 if (compressed_cluster) 703 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 704 705 if (nr_free) { 706 pgoff_t fofs; 707 /* 708 * once we invalidate valid blkaddr in range [ofs, ofs + count], 709 * we will invalidate all blkaddr in the whole range. 710 */ 711 fofs = f2fs_start_bidx_of_node(ofs_of_node(&dn->node_folio->page), 712 dn->inode) + ofs; 713 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 714 f2fs_update_age_extent_cache_range(dn, fofs, len); 715 dec_valid_block_count(sbi, dn->inode, nr_free); 716 } 717 dn->ofs_in_node = ofs; 718 719 f2fs_update_time(sbi, REQ_TIME); 720 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 721 dn->ofs_in_node, nr_free); 722 } 723 724 static int truncate_partial_data_page(struct inode *inode, u64 from, 725 bool cache_only) 726 { 727 loff_t offset = from & (PAGE_SIZE - 1); 728 pgoff_t index = from >> PAGE_SHIFT; 729 struct address_space *mapping = inode->i_mapping; 730 struct folio *folio; 731 732 if (!offset && !cache_only) 733 return 0; 734 735 if (cache_only) { 736 folio = filemap_lock_folio(mapping, index); 737 if (IS_ERR(folio)) 738 return 0; 739 if (folio_test_uptodate(folio)) 740 goto truncate_out; 741 f2fs_folio_put(folio, true); 742 return 0; 743 } 744 745 folio = f2fs_get_lock_data_folio(inode, index, true); 746 if (IS_ERR(folio)) 747 return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio); 748 truncate_out: 749 f2fs_folio_wait_writeback(folio, DATA, true, true); 750 folio_zero_segment(folio, offset, folio_size(folio)); 751 752 /* An encrypted inode should have a key and truncate the last page. */ 753 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 754 if (!cache_only) 755 folio_mark_dirty(folio); 756 f2fs_folio_put(folio, true); 757 return 0; 758 } 759 760 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 761 { 762 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 763 struct dnode_of_data dn; 764 pgoff_t free_from; 765 int count = 0, err = 0; 766 struct folio *ifolio; 767 bool truncate_page = false; 768 769 trace_f2fs_truncate_blocks_enter(inode, from); 770 771 if (IS_DEVICE_ALIASING(inode) && from) { 772 err = -EINVAL; 773 goto out_err; 774 } 775 776 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 777 778 if (free_from >= max_file_blocks(inode)) 779 goto free_partial; 780 781 if (lock) 782 f2fs_lock_op(sbi); 783 784 ifolio = f2fs_get_inode_folio(sbi, inode->i_ino); 785 if (IS_ERR(ifolio)) { 786 err = PTR_ERR(ifolio); 787 goto out; 788 } 789 790 if (IS_DEVICE_ALIASING(inode)) { 791 struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ]; 792 struct extent_info ei = et->largest; 793 794 f2fs_invalidate_blocks(sbi, ei.blk, ei.len); 795 796 dec_valid_block_count(sbi, inode, ei.len); 797 f2fs_update_time(sbi, REQ_TIME); 798 799 f2fs_folio_put(ifolio, true); 800 goto out; 801 } 802 803 if (f2fs_has_inline_data(inode)) { 804 f2fs_truncate_inline_inode(inode, ifolio, from); 805 f2fs_folio_put(ifolio, true); 806 truncate_page = true; 807 goto out; 808 } 809 810 set_new_dnode(&dn, inode, ifolio, NULL, 0); 811 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 812 if (err) { 813 if (err == -ENOENT) 814 goto free_next; 815 goto out; 816 } 817 818 count = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 819 820 count -= dn.ofs_in_node; 821 f2fs_bug_on(sbi, count < 0); 822 823 if (dn.ofs_in_node || IS_INODE(&dn.node_folio->page)) { 824 f2fs_truncate_data_blocks_range(&dn, count); 825 free_from += count; 826 } 827 828 f2fs_put_dnode(&dn); 829 free_next: 830 err = f2fs_truncate_inode_blocks(inode, free_from); 831 out: 832 if (lock) 833 f2fs_unlock_op(sbi); 834 free_partial: 835 /* lastly zero out the first data page */ 836 if (!err) 837 err = truncate_partial_data_page(inode, from, truncate_page); 838 out_err: 839 trace_f2fs_truncate_blocks_exit(inode, err); 840 return err; 841 } 842 843 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 844 { 845 u64 free_from = from; 846 int err; 847 848 #ifdef CONFIG_F2FS_FS_COMPRESSION 849 /* 850 * for compressed file, only support cluster size 851 * aligned truncation. 852 */ 853 if (f2fs_compressed_file(inode)) 854 free_from = round_up(from, 855 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 856 #endif 857 858 err = f2fs_do_truncate_blocks(inode, free_from, lock); 859 if (err) 860 return err; 861 862 #ifdef CONFIG_F2FS_FS_COMPRESSION 863 /* 864 * For compressed file, after release compress blocks, don't allow write 865 * direct, but we should allow write direct after truncate to zero. 866 */ 867 if (f2fs_compressed_file(inode) && !free_from 868 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 869 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 870 871 if (from != free_from) { 872 err = f2fs_truncate_partial_cluster(inode, from, lock); 873 if (err) 874 return err; 875 } 876 #endif 877 878 return 0; 879 } 880 881 int f2fs_truncate(struct inode *inode) 882 { 883 int err; 884 885 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 886 return -EIO; 887 888 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 889 S_ISLNK(inode->i_mode))) 890 return 0; 891 892 trace_f2fs_truncate(inode); 893 894 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 895 return -EIO; 896 897 err = f2fs_dquot_initialize(inode); 898 if (err) 899 return err; 900 901 /* we should check inline_data size */ 902 if (!f2fs_may_inline_data(inode)) { 903 err = f2fs_convert_inline_inode(inode); 904 if (err) 905 return err; 906 } 907 908 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 909 if (err) 910 return err; 911 912 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 913 f2fs_mark_inode_dirty_sync(inode, false); 914 return 0; 915 } 916 917 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 918 { 919 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 920 921 if (!fscrypt_dio_supported(inode)) 922 return true; 923 if (fsverity_active(inode)) 924 return true; 925 if (f2fs_compressed_file(inode)) 926 return true; 927 /* 928 * only force direct read to use buffered IO, for direct write, 929 * it expects inline data conversion before committing IO. 930 */ 931 if (f2fs_has_inline_data(inode) && rw == READ) 932 return true; 933 934 /* disallow direct IO if any of devices has unaligned blksize */ 935 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 936 return true; 937 /* 938 * for blkzoned device, fallback direct IO to buffered IO, so 939 * all IOs can be serialized by log-structured write. 940 */ 941 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) && 942 !f2fs_is_pinned_file(inode)) 943 return true; 944 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 945 return true; 946 947 return false; 948 } 949 950 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 951 struct kstat *stat, u32 request_mask, unsigned int query_flags) 952 { 953 struct inode *inode = d_inode(path->dentry); 954 struct f2fs_inode_info *fi = F2FS_I(inode); 955 struct f2fs_inode *ri = NULL; 956 unsigned int flags; 957 958 if (f2fs_has_extra_attr(inode) && 959 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 960 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 961 stat->result_mask |= STATX_BTIME; 962 stat->btime.tv_sec = fi->i_crtime.tv_sec; 963 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 964 } 965 966 /* 967 * Return the DIO alignment restrictions if requested. We only return 968 * this information when requested, since on encrypted files it might 969 * take a fair bit of work to get if the file wasn't opened recently. 970 * 971 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 972 * cannot represent that, so in that case we report no DIO support. 973 */ 974 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 975 unsigned int bsize = i_blocksize(inode); 976 977 stat->result_mask |= STATX_DIOALIGN; 978 if (!f2fs_force_buffered_io(inode, WRITE)) { 979 stat->dio_mem_align = bsize; 980 stat->dio_offset_align = bsize; 981 } 982 } 983 984 flags = fi->i_flags; 985 if (flags & F2FS_COMPR_FL) 986 stat->attributes |= STATX_ATTR_COMPRESSED; 987 if (flags & F2FS_APPEND_FL) 988 stat->attributes |= STATX_ATTR_APPEND; 989 if (IS_ENCRYPTED(inode)) 990 stat->attributes |= STATX_ATTR_ENCRYPTED; 991 if (flags & F2FS_IMMUTABLE_FL) 992 stat->attributes |= STATX_ATTR_IMMUTABLE; 993 if (flags & F2FS_NODUMP_FL) 994 stat->attributes |= STATX_ATTR_NODUMP; 995 if (IS_VERITY(inode)) 996 stat->attributes |= STATX_ATTR_VERITY; 997 998 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 999 STATX_ATTR_APPEND | 1000 STATX_ATTR_ENCRYPTED | 1001 STATX_ATTR_IMMUTABLE | 1002 STATX_ATTR_NODUMP | 1003 STATX_ATTR_VERITY); 1004 1005 generic_fillattr(idmap, request_mask, inode, stat); 1006 1007 /* we need to show initial sectors used for inline_data/dentries */ 1008 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 1009 f2fs_has_inline_dentry(inode)) 1010 stat->blocks += (stat->size + 511) >> 9; 1011 1012 return 0; 1013 } 1014 1015 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1016 static void __setattr_copy(struct mnt_idmap *idmap, 1017 struct inode *inode, const struct iattr *attr) 1018 { 1019 unsigned int ia_valid = attr->ia_valid; 1020 1021 i_uid_update(idmap, attr, inode); 1022 i_gid_update(idmap, attr, inode); 1023 if (ia_valid & ATTR_ATIME) 1024 inode_set_atime_to_ts(inode, attr->ia_atime); 1025 if (ia_valid & ATTR_MTIME) 1026 inode_set_mtime_to_ts(inode, attr->ia_mtime); 1027 if (ia_valid & ATTR_CTIME) 1028 inode_set_ctime_to_ts(inode, attr->ia_ctime); 1029 if (ia_valid & ATTR_MODE) { 1030 umode_t mode = attr->ia_mode; 1031 1032 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode))) 1033 mode &= ~S_ISGID; 1034 set_acl_inode(inode, mode); 1035 } 1036 } 1037 #else 1038 #define __setattr_copy setattr_copy 1039 #endif 1040 1041 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 1042 struct iattr *attr) 1043 { 1044 struct inode *inode = d_inode(dentry); 1045 struct f2fs_inode_info *fi = F2FS_I(inode); 1046 int err; 1047 1048 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1049 return -EIO; 1050 1051 if (unlikely(IS_IMMUTABLE(inode))) 1052 return -EPERM; 1053 1054 if (unlikely(IS_APPEND(inode) && 1055 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 1056 ATTR_GID | ATTR_TIMES_SET)))) 1057 return -EPERM; 1058 1059 if ((attr->ia_valid & ATTR_SIZE)) { 1060 if (!f2fs_is_compress_backend_ready(inode) || 1061 IS_DEVICE_ALIASING(inode)) 1062 return -EOPNOTSUPP; 1063 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 1064 !IS_ALIGNED(attr->ia_size, 1065 F2FS_BLK_TO_BYTES(fi->i_cluster_size))) 1066 return -EINVAL; 1067 } 1068 1069 err = setattr_prepare(idmap, dentry, attr); 1070 if (err) 1071 return err; 1072 1073 err = fscrypt_prepare_setattr(dentry, attr); 1074 if (err) 1075 return err; 1076 1077 err = fsverity_prepare_setattr(dentry, attr); 1078 if (err) 1079 return err; 1080 1081 if (is_quota_modification(idmap, inode, attr)) { 1082 err = f2fs_dquot_initialize(inode); 1083 if (err) 1084 return err; 1085 } 1086 if (i_uid_needs_update(idmap, attr, inode) || 1087 i_gid_needs_update(idmap, attr, inode)) { 1088 f2fs_lock_op(F2FS_I_SB(inode)); 1089 err = dquot_transfer(idmap, inode, attr); 1090 if (err) { 1091 set_sbi_flag(F2FS_I_SB(inode), 1092 SBI_QUOTA_NEED_REPAIR); 1093 f2fs_unlock_op(F2FS_I_SB(inode)); 1094 return err; 1095 } 1096 /* 1097 * update uid/gid under lock_op(), so that dquot and inode can 1098 * be updated atomically. 1099 */ 1100 i_uid_update(idmap, attr, inode); 1101 i_gid_update(idmap, attr, inode); 1102 f2fs_mark_inode_dirty_sync(inode, true); 1103 f2fs_unlock_op(F2FS_I_SB(inode)); 1104 } 1105 1106 if (attr->ia_valid & ATTR_SIZE) { 1107 loff_t old_size = i_size_read(inode); 1108 1109 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1110 /* 1111 * should convert inline inode before i_size_write to 1112 * keep smaller than inline_data size with inline flag. 1113 */ 1114 err = f2fs_convert_inline_inode(inode); 1115 if (err) 1116 return err; 1117 } 1118 1119 /* 1120 * wait for inflight dio, blocks should be removed after 1121 * IO completion. 1122 */ 1123 if (attr->ia_size < old_size) 1124 inode_dio_wait(inode); 1125 1126 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 1127 filemap_invalidate_lock(inode->i_mapping); 1128 1129 if (attr->ia_size > old_size) 1130 f2fs_zero_post_eof_page(inode, attr->ia_size); 1131 truncate_setsize(inode, attr->ia_size); 1132 1133 if (attr->ia_size <= old_size) 1134 err = f2fs_truncate(inode); 1135 /* 1136 * do not trim all blocks after i_size if target size is 1137 * larger than i_size. 1138 */ 1139 filemap_invalidate_unlock(inode->i_mapping); 1140 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 1141 if (err) 1142 return err; 1143 1144 spin_lock(&fi->i_size_lock); 1145 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1146 fi->last_disk_size = i_size_read(inode); 1147 spin_unlock(&fi->i_size_lock); 1148 } 1149 1150 __setattr_copy(idmap, inode, attr); 1151 1152 if (attr->ia_valid & ATTR_MODE) { 1153 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1154 1155 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1156 if (!err) 1157 inode->i_mode = fi->i_acl_mode; 1158 clear_inode_flag(inode, FI_ACL_MODE); 1159 } 1160 } 1161 1162 /* file size may changed here */ 1163 f2fs_mark_inode_dirty_sync(inode, true); 1164 1165 /* inode change will produce dirty node pages flushed by checkpoint */ 1166 f2fs_balance_fs(F2FS_I_SB(inode), true); 1167 1168 return err; 1169 } 1170 1171 const struct inode_operations f2fs_file_inode_operations = { 1172 .getattr = f2fs_getattr, 1173 .setattr = f2fs_setattr, 1174 .get_inode_acl = f2fs_get_acl, 1175 .set_acl = f2fs_set_acl, 1176 .listxattr = f2fs_listxattr, 1177 .fiemap = f2fs_fiemap, 1178 .fileattr_get = f2fs_fileattr_get, 1179 .fileattr_set = f2fs_fileattr_set, 1180 }; 1181 1182 static int fill_zero(struct inode *inode, pgoff_t index, 1183 loff_t start, loff_t len) 1184 { 1185 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1186 struct folio *folio; 1187 1188 if (!len) 1189 return 0; 1190 1191 f2fs_balance_fs(sbi, true); 1192 1193 f2fs_lock_op(sbi); 1194 folio = f2fs_get_new_data_folio(inode, NULL, index, false); 1195 f2fs_unlock_op(sbi); 1196 1197 if (IS_ERR(folio)) 1198 return PTR_ERR(folio); 1199 1200 f2fs_folio_wait_writeback(folio, DATA, true, true); 1201 folio_zero_range(folio, start, len); 1202 folio_mark_dirty(folio); 1203 f2fs_folio_put(folio, true); 1204 return 0; 1205 } 1206 1207 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1208 { 1209 int err; 1210 1211 while (pg_start < pg_end) { 1212 struct dnode_of_data dn; 1213 pgoff_t end_offset, count; 1214 1215 set_new_dnode(&dn, inode, NULL, NULL, 0); 1216 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1217 if (err) { 1218 if (err == -ENOENT) { 1219 pg_start = f2fs_get_next_page_offset(&dn, 1220 pg_start); 1221 continue; 1222 } 1223 return err; 1224 } 1225 1226 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 1227 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1228 1229 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1230 1231 f2fs_truncate_data_blocks_range(&dn, count); 1232 f2fs_put_dnode(&dn); 1233 1234 pg_start += count; 1235 } 1236 return 0; 1237 } 1238 1239 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1240 { 1241 pgoff_t pg_start, pg_end; 1242 loff_t off_start, off_end; 1243 int ret; 1244 1245 ret = f2fs_convert_inline_inode(inode); 1246 if (ret) 1247 return ret; 1248 1249 filemap_invalidate_lock(inode->i_mapping); 1250 f2fs_zero_post_eof_page(inode, offset + len); 1251 filemap_invalidate_unlock(inode->i_mapping); 1252 1253 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1254 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1255 1256 off_start = offset & (PAGE_SIZE - 1); 1257 off_end = (offset + len) & (PAGE_SIZE - 1); 1258 1259 if (pg_start == pg_end) { 1260 ret = fill_zero(inode, pg_start, off_start, 1261 off_end - off_start); 1262 if (ret) 1263 return ret; 1264 } else { 1265 if (off_start) { 1266 ret = fill_zero(inode, pg_start++, off_start, 1267 PAGE_SIZE - off_start); 1268 if (ret) 1269 return ret; 1270 } 1271 if (off_end) { 1272 ret = fill_zero(inode, pg_end, 0, off_end); 1273 if (ret) 1274 return ret; 1275 } 1276 1277 if (pg_start < pg_end) { 1278 loff_t blk_start, blk_end; 1279 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1280 1281 f2fs_balance_fs(sbi, true); 1282 1283 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1284 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1285 1286 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1287 filemap_invalidate_lock(inode->i_mapping); 1288 1289 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1290 1291 f2fs_lock_op(sbi); 1292 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1293 f2fs_unlock_op(sbi); 1294 1295 filemap_invalidate_unlock(inode->i_mapping); 1296 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1297 } 1298 } 1299 1300 return ret; 1301 } 1302 1303 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1304 int *do_replace, pgoff_t off, pgoff_t len) 1305 { 1306 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1307 struct dnode_of_data dn; 1308 int ret, done, i; 1309 1310 next_dnode: 1311 set_new_dnode(&dn, inode, NULL, NULL, 0); 1312 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1313 if (ret && ret != -ENOENT) { 1314 return ret; 1315 } else if (ret == -ENOENT) { 1316 if (dn.max_level == 0) 1317 return -ENOENT; 1318 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1319 dn.ofs_in_node, len); 1320 blkaddr += done; 1321 do_replace += done; 1322 goto next; 1323 } 1324 1325 done = min((pgoff_t)ADDRS_PER_PAGE(&dn.node_folio->page, inode) - 1326 dn.ofs_in_node, len); 1327 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1328 *blkaddr = f2fs_data_blkaddr(&dn); 1329 1330 if (__is_valid_data_blkaddr(*blkaddr) && 1331 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1332 DATA_GENERIC_ENHANCE)) { 1333 f2fs_put_dnode(&dn); 1334 return -EFSCORRUPTED; 1335 } 1336 1337 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1338 1339 if (f2fs_lfs_mode(sbi)) { 1340 f2fs_put_dnode(&dn); 1341 return -EOPNOTSUPP; 1342 } 1343 1344 /* do not invalidate this block address */ 1345 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1346 *do_replace = 1; 1347 } 1348 } 1349 f2fs_put_dnode(&dn); 1350 next: 1351 len -= done; 1352 off += done; 1353 if (len) 1354 goto next_dnode; 1355 return 0; 1356 } 1357 1358 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1359 int *do_replace, pgoff_t off, int len) 1360 { 1361 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1362 struct dnode_of_data dn; 1363 int ret, i; 1364 1365 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1366 if (*do_replace == 0) 1367 continue; 1368 1369 set_new_dnode(&dn, inode, NULL, NULL, 0); 1370 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1371 if (ret) { 1372 dec_valid_block_count(sbi, inode, 1); 1373 f2fs_invalidate_blocks(sbi, *blkaddr, 1); 1374 } else { 1375 f2fs_update_data_blkaddr(&dn, *blkaddr); 1376 } 1377 f2fs_put_dnode(&dn); 1378 } 1379 return 0; 1380 } 1381 1382 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1383 block_t *blkaddr, int *do_replace, 1384 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1385 { 1386 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1387 pgoff_t i = 0; 1388 int ret; 1389 1390 while (i < len) { 1391 if (blkaddr[i] == NULL_ADDR && !full) { 1392 i++; 1393 continue; 1394 } 1395 1396 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1397 struct dnode_of_data dn; 1398 struct node_info ni; 1399 size_t new_size; 1400 pgoff_t ilen; 1401 1402 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1403 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1404 if (ret) 1405 return ret; 1406 1407 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1408 if (ret) { 1409 f2fs_put_dnode(&dn); 1410 return ret; 1411 } 1412 1413 ilen = min((pgoff_t) 1414 ADDRS_PER_PAGE(&dn.node_folio->page, dst_inode) - 1415 dn.ofs_in_node, len - i); 1416 do { 1417 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1418 f2fs_truncate_data_blocks_range(&dn, 1); 1419 1420 if (do_replace[i]) { 1421 f2fs_i_blocks_write(src_inode, 1422 1, false, false); 1423 f2fs_i_blocks_write(dst_inode, 1424 1, true, false); 1425 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1426 blkaddr[i], ni.version, true, false); 1427 1428 do_replace[i] = 0; 1429 } 1430 dn.ofs_in_node++; 1431 i++; 1432 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1433 if (dst_inode->i_size < new_size) 1434 f2fs_i_size_write(dst_inode, new_size); 1435 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1436 1437 f2fs_put_dnode(&dn); 1438 } else { 1439 struct folio *fsrc, *fdst; 1440 1441 fsrc = f2fs_get_lock_data_folio(src_inode, 1442 src + i, true); 1443 if (IS_ERR(fsrc)) 1444 return PTR_ERR(fsrc); 1445 fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i, 1446 true); 1447 if (IS_ERR(fdst)) { 1448 f2fs_folio_put(fsrc, true); 1449 return PTR_ERR(fdst); 1450 } 1451 1452 f2fs_folio_wait_writeback(fdst, DATA, true, true); 1453 1454 memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE); 1455 folio_mark_dirty(fdst); 1456 set_page_private_gcing(&fdst->page); 1457 f2fs_folio_put(fdst, true); 1458 f2fs_folio_put(fsrc, true); 1459 1460 ret = f2fs_truncate_hole(src_inode, 1461 src + i, src + i + 1); 1462 if (ret) 1463 return ret; 1464 i++; 1465 } 1466 } 1467 return 0; 1468 } 1469 1470 static int __exchange_data_block(struct inode *src_inode, 1471 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1472 pgoff_t len, bool full) 1473 { 1474 block_t *src_blkaddr; 1475 int *do_replace; 1476 pgoff_t olen; 1477 int ret; 1478 1479 while (len) { 1480 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1481 1482 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1483 array_size(olen, sizeof(block_t)), 1484 GFP_NOFS); 1485 if (!src_blkaddr) 1486 return -ENOMEM; 1487 1488 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1489 array_size(olen, sizeof(int)), 1490 GFP_NOFS); 1491 if (!do_replace) { 1492 kvfree(src_blkaddr); 1493 return -ENOMEM; 1494 } 1495 1496 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1497 do_replace, src, olen); 1498 if (ret) 1499 goto roll_back; 1500 1501 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1502 do_replace, src, dst, olen, full); 1503 if (ret) 1504 goto roll_back; 1505 1506 src += olen; 1507 dst += olen; 1508 len -= olen; 1509 1510 kvfree(src_blkaddr); 1511 kvfree(do_replace); 1512 } 1513 return 0; 1514 1515 roll_back: 1516 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1517 kvfree(src_blkaddr); 1518 kvfree(do_replace); 1519 return ret; 1520 } 1521 1522 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1523 { 1524 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1525 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1526 pgoff_t start = offset >> PAGE_SHIFT; 1527 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1528 int ret; 1529 1530 f2fs_balance_fs(sbi, true); 1531 1532 /* avoid gc operation during block exchange */ 1533 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1534 filemap_invalidate_lock(inode->i_mapping); 1535 1536 f2fs_zero_post_eof_page(inode, offset + len); 1537 1538 f2fs_lock_op(sbi); 1539 f2fs_drop_extent_tree(inode); 1540 truncate_pagecache(inode, offset); 1541 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1542 f2fs_unlock_op(sbi); 1543 1544 filemap_invalidate_unlock(inode->i_mapping); 1545 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1546 return ret; 1547 } 1548 1549 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1550 { 1551 loff_t new_size; 1552 int ret; 1553 1554 if (offset + len >= i_size_read(inode)) 1555 return -EINVAL; 1556 1557 /* collapse range should be aligned to block size of f2fs. */ 1558 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1559 return -EINVAL; 1560 1561 ret = f2fs_convert_inline_inode(inode); 1562 if (ret) 1563 return ret; 1564 1565 /* write out all dirty pages from offset */ 1566 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1567 if (ret) 1568 return ret; 1569 1570 ret = f2fs_do_collapse(inode, offset, len); 1571 if (ret) 1572 return ret; 1573 1574 /* write out all moved pages, if possible */ 1575 filemap_invalidate_lock(inode->i_mapping); 1576 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1577 truncate_pagecache(inode, offset); 1578 1579 new_size = i_size_read(inode) - len; 1580 ret = f2fs_truncate_blocks(inode, new_size, true); 1581 filemap_invalidate_unlock(inode->i_mapping); 1582 if (!ret) 1583 f2fs_i_size_write(inode, new_size); 1584 return ret; 1585 } 1586 1587 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1588 pgoff_t end) 1589 { 1590 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1591 pgoff_t index = start; 1592 unsigned int ofs_in_node = dn->ofs_in_node; 1593 blkcnt_t count = 0; 1594 int ret; 1595 1596 for (; index < end; index++, dn->ofs_in_node++) { 1597 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1598 count++; 1599 } 1600 1601 dn->ofs_in_node = ofs_in_node; 1602 ret = f2fs_reserve_new_blocks(dn, count); 1603 if (ret) 1604 return ret; 1605 1606 dn->ofs_in_node = ofs_in_node; 1607 for (index = start; index < end; index++, dn->ofs_in_node++) { 1608 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1609 /* 1610 * f2fs_reserve_new_blocks will not guarantee entire block 1611 * allocation. 1612 */ 1613 if (dn->data_blkaddr == NULL_ADDR) { 1614 ret = -ENOSPC; 1615 break; 1616 } 1617 1618 if (dn->data_blkaddr == NEW_ADDR) 1619 continue; 1620 1621 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1622 DATA_GENERIC_ENHANCE)) { 1623 ret = -EFSCORRUPTED; 1624 break; 1625 } 1626 1627 f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1); 1628 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1629 } 1630 1631 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1632 f2fs_update_age_extent_cache_range(dn, start, index - start); 1633 1634 return ret; 1635 } 1636 1637 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1638 int mode) 1639 { 1640 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1641 struct address_space *mapping = inode->i_mapping; 1642 pgoff_t index, pg_start, pg_end; 1643 loff_t new_size = i_size_read(inode); 1644 loff_t off_start, off_end; 1645 int ret = 0; 1646 1647 ret = inode_newsize_ok(inode, (len + offset)); 1648 if (ret) 1649 return ret; 1650 1651 ret = f2fs_convert_inline_inode(inode); 1652 if (ret) 1653 return ret; 1654 1655 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1656 if (ret) 1657 return ret; 1658 1659 filemap_invalidate_lock(mapping); 1660 f2fs_zero_post_eof_page(inode, offset + len); 1661 filemap_invalidate_unlock(mapping); 1662 1663 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1664 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1665 1666 off_start = offset & (PAGE_SIZE - 1); 1667 off_end = (offset + len) & (PAGE_SIZE - 1); 1668 1669 if (pg_start == pg_end) { 1670 ret = fill_zero(inode, pg_start, off_start, 1671 off_end - off_start); 1672 if (ret) 1673 return ret; 1674 1675 new_size = max_t(loff_t, new_size, offset + len); 1676 } else { 1677 if (off_start) { 1678 ret = fill_zero(inode, pg_start++, off_start, 1679 PAGE_SIZE - off_start); 1680 if (ret) 1681 return ret; 1682 1683 new_size = max_t(loff_t, new_size, 1684 (loff_t)pg_start << PAGE_SHIFT); 1685 } 1686 1687 for (index = pg_start; index < pg_end;) { 1688 struct dnode_of_data dn; 1689 unsigned int end_offset; 1690 pgoff_t end; 1691 1692 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1693 filemap_invalidate_lock(mapping); 1694 1695 truncate_pagecache_range(inode, 1696 (loff_t)index << PAGE_SHIFT, 1697 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1698 1699 f2fs_lock_op(sbi); 1700 1701 set_new_dnode(&dn, inode, NULL, NULL, 0); 1702 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1703 if (ret) { 1704 f2fs_unlock_op(sbi); 1705 filemap_invalidate_unlock(mapping); 1706 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1707 goto out; 1708 } 1709 1710 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 1711 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1712 1713 ret = f2fs_do_zero_range(&dn, index, end); 1714 f2fs_put_dnode(&dn); 1715 1716 f2fs_unlock_op(sbi); 1717 filemap_invalidate_unlock(mapping); 1718 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1719 1720 f2fs_balance_fs(sbi, dn.node_changed); 1721 1722 if (ret) 1723 goto out; 1724 1725 index = end; 1726 new_size = max_t(loff_t, new_size, 1727 (loff_t)index << PAGE_SHIFT); 1728 } 1729 1730 if (off_end) { 1731 ret = fill_zero(inode, pg_end, 0, off_end); 1732 if (ret) 1733 goto out; 1734 1735 new_size = max_t(loff_t, new_size, offset + len); 1736 } 1737 } 1738 1739 out: 1740 if (new_size > i_size_read(inode)) { 1741 if (mode & FALLOC_FL_KEEP_SIZE) 1742 file_set_keep_isize(inode); 1743 else 1744 f2fs_i_size_write(inode, new_size); 1745 } 1746 return ret; 1747 } 1748 1749 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1750 { 1751 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1752 struct address_space *mapping = inode->i_mapping; 1753 pgoff_t nr, pg_start, pg_end, delta, idx; 1754 loff_t new_size; 1755 int ret = 0; 1756 1757 new_size = i_size_read(inode) + len; 1758 ret = inode_newsize_ok(inode, new_size); 1759 if (ret) 1760 return ret; 1761 1762 if (offset >= i_size_read(inode)) 1763 return -EINVAL; 1764 1765 /* insert range should be aligned to block size of f2fs. */ 1766 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1767 return -EINVAL; 1768 1769 ret = f2fs_convert_inline_inode(inode); 1770 if (ret) 1771 return ret; 1772 1773 f2fs_balance_fs(sbi, true); 1774 1775 filemap_invalidate_lock(mapping); 1776 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1777 filemap_invalidate_unlock(mapping); 1778 if (ret) 1779 return ret; 1780 1781 /* write out all dirty pages from offset */ 1782 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1783 if (ret) 1784 return ret; 1785 1786 pg_start = offset >> PAGE_SHIFT; 1787 pg_end = (offset + len) >> PAGE_SHIFT; 1788 delta = pg_end - pg_start; 1789 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1790 1791 /* avoid gc operation during block exchange */ 1792 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1793 filemap_invalidate_lock(mapping); 1794 1795 f2fs_zero_post_eof_page(inode, offset + len); 1796 truncate_pagecache(inode, offset); 1797 1798 while (!ret && idx > pg_start) { 1799 nr = idx - pg_start; 1800 if (nr > delta) 1801 nr = delta; 1802 idx -= nr; 1803 1804 f2fs_lock_op(sbi); 1805 f2fs_drop_extent_tree(inode); 1806 1807 ret = __exchange_data_block(inode, inode, idx, 1808 idx + delta, nr, false); 1809 f2fs_unlock_op(sbi); 1810 } 1811 filemap_invalidate_unlock(mapping); 1812 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1813 if (ret) 1814 return ret; 1815 1816 /* write out all moved pages, if possible */ 1817 filemap_invalidate_lock(mapping); 1818 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1819 truncate_pagecache(inode, offset); 1820 filemap_invalidate_unlock(mapping); 1821 1822 if (!ret) 1823 f2fs_i_size_write(inode, new_size); 1824 return ret; 1825 } 1826 1827 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1828 loff_t len, int mode) 1829 { 1830 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1831 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1832 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1833 .m_may_create = true }; 1834 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1835 .init_gc_type = FG_GC, 1836 .should_migrate_blocks = false, 1837 .err_gc_skipped = true, 1838 .nr_free_secs = 0 }; 1839 pgoff_t pg_start, pg_end; 1840 loff_t new_size; 1841 loff_t off_end; 1842 block_t expanded = 0; 1843 int err; 1844 1845 err = inode_newsize_ok(inode, (len + offset)); 1846 if (err) 1847 return err; 1848 1849 err = f2fs_convert_inline_inode(inode); 1850 if (err) 1851 return err; 1852 1853 filemap_invalidate_lock(inode->i_mapping); 1854 f2fs_zero_post_eof_page(inode, offset + len); 1855 filemap_invalidate_unlock(inode->i_mapping); 1856 1857 f2fs_balance_fs(sbi, true); 1858 1859 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1860 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1861 off_end = (offset + len) & (PAGE_SIZE - 1); 1862 1863 map.m_lblk = pg_start; 1864 map.m_len = pg_end - pg_start; 1865 if (off_end) 1866 map.m_len++; 1867 1868 if (!map.m_len) 1869 return 0; 1870 1871 if (f2fs_is_pinned_file(inode)) { 1872 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1873 block_t sec_len = roundup(map.m_len, sec_blks); 1874 1875 map.m_len = sec_blks; 1876 next_alloc: 1877 f2fs_down_write(&sbi->pin_sem); 1878 1879 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1880 if (has_not_enough_free_secs(sbi, 0, 0)) { 1881 f2fs_up_write(&sbi->pin_sem); 1882 err = -ENOSPC; 1883 f2fs_warn_ratelimited(sbi, 1884 "ino:%lu, start:%lu, end:%lu, need to trigger GC to " 1885 "reclaim enough free segment when checkpoint is enabled", 1886 inode->i_ino, pg_start, pg_end); 1887 goto out_err; 1888 } 1889 } 1890 1891 if (has_not_enough_free_secs(sbi, 0, f2fs_sb_has_blkzoned(sbi) ? 1892 ZONED_PIN_SEC_REQUIRED_COUNT : 1893 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1894 f2fs_down_write(&sbi->gc_lock); 1895 stat_inc_gc_call_count(sbi, FOREGROUND); 1896 err = f2fs_gc(sbi, &gc_control); 1897 if (err && err != -ENODATA) { 1898 f2fs_up_write(&sbi->pin_sem); 1899 goto out_err; 1900 } 1901 } 1902 1903 err = f2fs_allocate_pinning_section(sbi); 1904 if (err) { 1905 f2fs_up_write(&sbi->pin_sem); 1906 goto out_err; 1907 } 1908 1909 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1910 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1911 file_dont_truncate(inode); 1912 1913 f2fs_up_write(&sbi->pin_sem); 1914 1915 expanded += map.m_len; 1916 sec_len -= map.m_len; 1917 map.m_lblk += map.m_len; 1918 if (!err && sec_len) 1919 goto next_alloc; 1920 1921 map.m_len = expanded; 1922 } else { 1923 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1924 expanded = map.m_len; 1925 } 1926 out_err: 1927 if (err) { 1928 pgoff_t last_off; 1929 1930 if (!expanded) 1931 return err; 1932 1933 last_off = pg_start + expanded - 1; 1934 1935 /* update new size to the failed position */ 1936 new_size = (last_off == pg_end) ? offset + len : 1937 (loff_t)(last_off + 1) << PAGE_SHIFT; 1938 } else { 1939 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1940 } 1941 1942 if (new_size > i_size_read(inode)) { 1943 if (mode & FALLOC_FL_KEEP_SIZE) 1944 file_set_keep_isize(inode); 1945 else 1946 f2fs_i_size_write(inode, new_size); 1947 } 1948 1949 return err; 1950 } 1951 1952 static long f2fs_fallocate(struct file *file, int mode, 1953 loff_t offset, loff_t len) 1954 { 1955 struct inode *inode = file_inode(file); 1956 long ret = 0; 1957 1958 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1959 return -EIO; 1960 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1961 return -ENOSPC; 1962 if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode)) 1963 return -EOPNOTSUPP; 1964 1965 /* f2fs only support ->fallocate for regular file */ 1966 if (!S_ISREG(inode->i_mode)) 1967 return -EINVAL; 1968 1969 if (IS_ENCRYPTED(inode) && 1970 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1971 return -EOPNOTSUPP; 1972 1973 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1974 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1975 FALLOC_FL_INSERT_RANGE)) 1976 return -EOPNOTSUPP; 1977 1978 inode_lock(inode); 1979 1980 /* 1981 * Pinned file should not support partial truncation since the block 1982 * can be used by applications. 1983 */ 1984 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1985 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1986 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 1987 ret = -EOPNOTSUPP; 1988 goto out; 1989 } 1990 1991 ret = file_modified(file); 1992 if (ret) 1993 goto out; 1994 1995 /* 1996 * wait for inflight dio, blocks should be removed after IO 1997 * completion. 1998 */ 1999 inode_dio_wait(inode); 2000 2001 if (mode & FALLOC_FL_PUNCH_HOLE) { 2002 if (offset >= inode->i_size) 2003 goto out; 2004 2005 ret = f2fs_punch_hole(inode, offset, len); 2006 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 2007 ret = f2fs_collapse_range(inode, offset, len); 2008 } else if (mode & FALLOC_FL_ZERO_RANGE) { 2009 ret = f2fs_zero_range(inode, offset, len, mode); 2010 } else if (mode & FALLOC_FL_INSERT_RANGE) { 2011 ret = f2fs_insert_range(inode, offset, len); 2012 } else { 2013 ret = f2fs_expand_inode_data(inode, offset, len, mode); 2014 } 2015 2016 if (!ret) { 2017 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 2018 f2fs_mark_inode_dirty_sync(inode, false); 2019 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2020 } 2021 2022 out: 2023 inode_unlock(inode); 2024 2025 trace_f2fs_fallocate(inode, mode, offset, len, ret); 2026 return ret; 2027 } 2028 2029 static int f2fs_release_file(struct inode *inode, struct file *filp) 2030 { 2031 /* 2032 * f2fs_release_file is called at every close calls. So we should 2033 * not drop any inmemory pages by close called by other process. 2034 */ 2035 if (!(filp->f_mode & FMODE_WRITE) || 2036 atomic_read(&inode->i_writecount) != 1) 2037 return 0; 2038 2039 inode_lock(inode); 2040 f2fs_abort_atomic_write(inode, true); 2041 inode_unlock(inode); 2042 2043 return 0; 2044 } 2045 2046 static int f2fs_file_flush(struct file *file, fl_owner_t id) 2047 { 2048 struct inode *inode = file_inode(file); 2049 2050 /* 2051 * If the process doing a transaction is crashed, we should do 2052 * roll-back. Otherwise, other reader/write can see corrupted database 2053 * until all the writers close its file. Since this should be done 2054 * before dropping file lock, it needs to do in ->flush. 2055 */ 2056 if (F2FS_I(inode)->atomic_write_task == current && 2057 (current->flags & PF_EXITING)) { 2058 inode_lock(inode); 2059 f2fs_abort_atomic_write(inode, true); 2060 inode_unlock(inode); 2061 } 2062 2063 return 0; 2064 } 2065 2066 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 2067 { 2068 struct f2fs_inode_info *fi = F2FS_I(inode); 2069 u32 masked_flags = fi->i_flags & mask; 2070 2071 /* mask can be shrunk by flags_valid selector */ 2072 iflags &= mask; 2073 2074 /* Is it quota file? Do not allow user to mess with it */ 2075 if (IS_NOQUOTA(inode)) 2076 return -EPERM; 2077 2078 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 2079 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 2080 return -EOPNOTSUPP; 2081 if (!f2fs_empty_dir(inode)) 2082 return -ENOTEMPTY; 2083 } 2084 2085 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 2086 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 2087 return -EOPNOTSUPP; 2088 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 2089 return -EINVAL; 2090 } 2091 2092 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 2093 if (masked_flags & F2FS_COMPR_FL) { 2094 if (!f2fs_disable_compressed_file(inode)) 2095 return -EINVAL; 2096 } else { 2097 /* try to convert inline_data to support compression */ 2098 int err = f2fs_convert_inline_inode(inode); 2099 if (err) 2100 return err; 2101 2102 f2fs_down_write(&fi->i_sem); 2103 if (!f2fs_may_compress(inode) || 2104 (S_ISREG(inode->i_mode) && 2105 F2FS_HAS_BLOCKS(inode))) { 2106 f2fs_up_write(&fi->i_sem); 2107 return -EINVAL; 2108 } 2109 err = set_compress_context(inode); 2110 f2fs_up_write(&fi->i_sem); 2111 2112 if (err) 2113 return err; 2114 } 2115 } 2116 2117 fi->i_flags = iflags | (fi->i_flags & ~mask); 2118 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 2119 (fi->i_flags & F2FS_NOCOMP_FL)); 2120 2121 if (fi->i_flags & F2FS_PROJINHERIT_FL) 2122 set_inode_flag(inode, FI_PROJ_INHERIT); 2123 else 2124 clear_inode_flag(inode, FI_PROJ_INHERIT); 2125 2126 inode_set_ctime_current(inode); 2127 f2fs_set_inode_flags(inode); 2128 f2fs_mark_inode_dirty_sync(inode, true); 2129 return 0; 2130 } 2131 2132 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 2133 2134 /* 2135 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 2136 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 2137 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 2138 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 2139 * 2140 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 2141 * FS_IOC_FSSETXATTR is done by the VFS. 2142 */ 2143 2144 static const struct { 2145 u32 iflag; 2146 u32 fsflag; 2147 } f2fs_fsflags_map[] = { 2148 { F2FS_COMPR_FL, FS_COMPR_FL }, 2149 { F2FS_SYNC_FL, FS_SYNC_FL }, 2150 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 2151 { F2FS_APPEND_FL, FS_APPEND_FL }, 2152 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 2153 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 2154 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 2155 { F2FS_INDEX_FL, FS_INDEX_FL }, 2156 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 2157 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2158 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2159 }; 2160 2161 #define F2FS_GETTABLE_FS_FL ( \ 2162 FS_COMPR_FL | \ 2163 FS_SYNC_FL | \ 2164 FS_IMMUTABLE_FL | \ 2165 FS_APPEND_FL | \ 2166 FS_NODUMP_FL | \ 2167 FS_NOATIME_FL | \ 2168 FS_NOCOMP_FL | \ 2169 FS_INDEX_FL | \ 2170 FS_DIRSYNC_FL | \ 2171 FS_PROJINHERIT_FL | \ 2172 FS_ENCRYPT_FL | \ 2173 FS_INLINE_DATA_FL | \ 2174 FS_NOCOW_FL | \ 2175 FS_VERITY_FL | \ 2176 FS_CASEFOLD_FL) 2177 2178 #define F2FS_SETTABLE_FS_FL ( \ 2179 FS_COMPR_FL | \ 2180 FS_SYNC_FL | \ 2181 FS_IMMUTABLE_FL | \ 2182 FS_APPEND_FL | \ 2183 FS_NODUMP_FL | \ 2184 FS_NOATIME_FL | \ 2185 FS_NOCOMP_FL | \ 2186 FS_DIRSYNC_FL | \ 2187 FS_PROJINHERIT_FL | \ 2188 FS_CASEFOLD_FL) 2189 2190 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2191 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2192 { 2193 u32 fsflags = 0; 2194 int i; 2195 2196 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2197 if (iflags & f2fs_fsflags_map[i].iflag) 2198 fsflags |= f2fs_fsflags_map[i].fsflag; 2199 2200 return fsflags; 2201 } 2202 2203 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2204 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2205 { 2206 u32 iflags = 0; 2207 int i; 2208 2209 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2210 if (fsflags & f2fs_fsflags_map[i].fsflag) 2211 iflags |= f2fs_fsflags_map[i].iflag; 2212 2213 return iflags; 2214 } 2215 2216 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2217 { 2218 struct inode *inode = file_inode(filp); 2219 2220 return put_user(inode->i_generation, (int __user *)arg); 2221 } 2222 2223 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2224 { 2225 struct inode *inode = file_inode(filp); 2226 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2227 struct f2fs_inode_info *fi = F2FS_I(inode); 2228 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2229 loff_t isize; 2230 int ret; 2231 2232 if (!(filp->f_mode & FMODE_WRITE)) 2233 return -EBADF; 2234 2235 if (!inode_owner_or_capable(idmap, inode)) 2236 return -EACCES; 2237 2238 if (!S_ISREG(inode->i_mode)) 2239 return -EINVAL; 2240 2241 if (filp->f_flags & O_DIRECT) 2242 return -EINVAL; 2243 2244 ret = mnt_want_write_file(filp); 2245 if (ret) 2246 return ret; 2247 2248 inode_lock(inode); 2249 2250 if (!f2fs_disable_compressed_file(inode) || 2251 f2fs_is_pinned_file(inode)) { 2252 ret = -EINVAL; 2253 goto out; 2254 } 2255 2256 if (f2fs_is_atomic_file(inode)) 2257 goto out; 2258 2259 ret = f2fs_convert_inline_inode(inode); 2260 if (ret) 2261 goto out; 2262 2263 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2264 f2fs_down_write(&fi->i_gc_rwsem[READ]); 2265 2266 /* 2267 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2268 * f2fs_is_atomic_file. 2269 */ 2270 if (get_dirty_pages(inode)) 2271 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2272 inode->i_ino, get_dirty_pages(inode)); 2273 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2274 if (ret) 2275 goto out_unlock; 2276 2277 /* Check if the inode already has a COW inode */ 2278 if (fi->cow_inode == NULL) { 2279 /* Create a COW inode for atomic write */ 2280 struct dentry *dentry = file_dentry(filp); 2281 struct inode *dir = d_inode(dentry->d_parent); 2282 2283 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode); 2284 if (ret) 2285 goto out_unlock; 2286 2287 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2288 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2289 2290 /* Set the COW inode's atomic_inode to the atomic inode */ 2291 F2FS_I(fi->cow_inode)->atomic_inode = inode; 2292 } else { 2293 /* Reuse the already created COW inode */ 2294 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode)); 2295 2296 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1); 2297 2298 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2299 if (ret) 2300 goto out_unlock; 2301 } 2302 2303 f2fs_write_inode(inode, NULL); 2304 2305 stat_inc_atomic_inode(inode); 2306 2307 set_inode_flag(inode, FI_ATOMIC_FILE); 2308 2309 isize = i_size_read(inode); 2310 fi->original_i_size = isize; 2311 if (truncate) { 2312 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2313 truncate_inode_pages_final(inode->i_mapping); 2314 f2fs_i_size_write(inode, 0); 2315 isize = 0; 2316 } 2317 f2fs_i_size_write(fi->cow_inode, isize); 2318 2319 out_unlock: 2320 f2fs_up_write(&fi->i_gc_rwsem[READ]); 2321 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2322 if (ret) 2323 goto out; 2324 2325 f2fs_update_time(sbi, REQ_TIME); 2326 fi->atomic_write_task = current; 2327 stat_update_max_atomic_write(inode); 2328 fi->atomic_write_cnt = 0; 2329 out: 2330 inode_unlock(inode); 2331 mnt_drop_write_file(filp); 2332 return ret; 2333 } 2334 2335 static int f2fs_ioc_commit_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 f2fs_balance_fs(F2FS_I_SB(inode), true); 2352 2353 inode_lock(inode); 2354 2355 if (f2fs_is_atomic_file(inode)) { 2356 ret = f2fs_commit_atomic_write(inode); 2357 if (!ret) 2358 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2359 2360 f2fs_abort_atomic_write(inode, ret); 2361 } else { 2362 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2363 } 2364 2365 inode_unlock(inode); 2366 mnt_drop_write_file(filp); 2367 return ret; 2368 } 2369 2370 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2371 { 2372 struct inode *inode = file_inode(filp); 2373 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2374 int ret; 2375 2376 if (!(filp->f_mode & FMODE_WRITE)) 2377 return -EBADF; 2378 2379 if (!inode_owner_or_capable(idmap, inode)) 2380 return -EACCES; 2381 2382 ret = mnt_want_write_file(filp); 2383 if (ret) 2384 return ret; 2385 2386 inode_lock(inode); 2387 2388 f2fs_abort_atomic_write(inode, true); 2389 2390 inode_unlock(inode); 2391 2392 mnt_drop_write_file(filp); 2393 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2394 return ret; 2395 } 2396 2397 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag, 2398 bool readonly, bool need_lock) 2399 { 2400 struct super_block *sb = sbi->sb; 2401 int ret = 0; 2402 2403 switch (flag) { 2404 case F2FS_GOING_DOWN_FULLSYNC: 2405 ret = bdev_freeze(sb->s_bdev); 2406 if (ret) 2407 goto out; 2408 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2409 bdev_thaw(sb->s_bdev); 2410 break; 2411 case F2FS_GOING_DOWN_METASYNC: 2412 /* do checkpoint only */ 2413 ret = f2fs_sync_fs(sb, 1); 2414 if (ret) { 2415 if (ret == -EIO) 2416 ret = 0; 2417 goto out; 2418 } 2419 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2420 break; 2421 case F2FS_GOING_DOWN_NOSYNC: 2422 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2423 break; 2424 case F2FS_GOING_DOWN_METAFLUSH: 2425 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2426 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2427 break; 2428 case F2FS_GOING_DOWN_NEED_FSCK: 2429 set_sbi_flag(sbi, SBI_NEED_FSCK); 2430 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2431 set_sbi_flag(sbi, SBI_IS_DIRTY); 2432 /* do checkpoint only */ 2433 ret = f2fs_sync_fs(sb, 1); 2434 if (ret == -EIO) 2435 ret = 0; 2436 goto out; 2437 default: 2438 ret = -EINVAL; 2439 goto out; 2440 } 2441 2442 if (readonly) 2443 goto out; 2444 2445 /* 2446 * grab sb->s_umount to avoid racing w/ remount() and other shutdown 2447 * paths. 2448 */ 2449 if (need_lock) 2450 down_write(&sbi->sb->s_umount); 2451 2452 f2fs_stop_gc_thread(sbi); 2453 f2fs_stop_discard_thread(sbi); 2454 2455 f2fs_drop_discard_cmd(sbi); 2456 clear_opt(sbi, DISCARD); 2457 2458 if (need_lock) 2459 up_write(&sbi->sb->s_umount); 2460 2461 f2fs_update_time(sbi, REQ_TIME); 2462 out: 2463 2464 trace_f2fs_shutdown(sbi, flag, ret); 2465 2466 return ret; 2467 } 2468 2469 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2470 { 2471 struct inode *inode = file_inode(filp); 2472 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2473 __u32 in; 2474 int ret; 2475 bool need_drop = false, readonly = false; 2476 2477 if (!capable(CAP_SYS_ADMIN)) 2478 return -EPERM; 2479 2480 if (get_user(in, (__u32 __user *)arg)) 2481 return -EFAULT; 2482 2483 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2484 ret = mnt_want_write_file(filp); 2485 if (ret) { 2486 if (ret != -EROFS) 2487 return ret; 2488 2489 /* fallback to nosync shutdown for readonly fs */ 2490 in = F2FS_GOING_DOWN_NOSYNC; 2491 readonly = true; 2492 } else { 2493 need_drop = true; 2494 } 2495 } 2496 2497 ret = f2fs_do_shutdown(sbi, in, readonly, true); 2498 2499 if (need_drop) 2500 mnt_drop_write_file(filp); 2501 2502 return ret; 2503 } 2504 2505 static int f2fs_keep_noreuse_range(struct inode *inode, 2506 loff_t offset, loff_t len) 2507 { 2508 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2509 u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 2510 u64 start, end; 2511 int ret = 0; 2512 2513 if (!S_ISREG(inode->i_mode)) 2514 return 0; 2515 2516 if (offset >= max_bytes || len > max_bytes || 2517 (offset + len) > max_bytes) 2518 return 0; 2519 2520 start = offset >> PAGE_SHIFT; 2521 end = DIV_ROUND_UP(offset + len, PAGE_SIZE); 2522 2523 inode_lock(inode); 2524 if (f2fs_is_atomic_file(inode)) { 2525 inode_unlock(inode); 2526 return 0; 2527 } 2528 2529 spin_lock(&sbi->inode_lock[DONATE_INODE]); 2530 /* let's remove the range, if len = 0 */ 2531 if (!len) { 2532 if (!list_empty(&F2FS_I(inode)->gdonate_list)) { 2533 list_del_init(&F2FS_I(inode)->gdonate_list); 2534 sbi->donate_files--; 2535 if (is_inode_flag_set(inode, FI_DONATE_FINISHED)) 2536 ret = -EALREADY; 2537 else 2538 set_inode_flag(inode, FI_DONATE_FINISHED); 2539 } else 2540 ret = -ENOENT; 2541 } else { 2542 if (list_empty(&F2FS_I(inode)->gdonate_list)) { 2543 list_add_tail(&F2FS_I(inode)->gdonate_list, 2544 &sbi->inode_list[DONATE_INODE]); 2545 sbi->donate_files++; 2546 } else { 2547 list_move_tail(&F2FS_I(inode)->gdonate_list, 2548 &sbi->inode_list[DONATE_INODE]); 2549 } 2550 F2FS_I(inode)->donate_start = start; 2551 F2FS_I(inode)->donate_end = end - 1; 2552 clear_inode_flag(inode, FI_DONATE_FINISHED); 2553 } 2554 spin_unlock(&sbi->inode_lock[DONATE_INODE]); 2555 inode_unlock(inode); 2556 2557 return ret; 2558 } 2559 2560 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2561 { 2562 struct inode *inode = file_inode(filp); 2563 struct super_block *sb = inode->i_sb; 2564 struct fstrim_range range; 2565 int ret; 2566 2567 if (!capable(CAP_SYS_ADMIN)) 2568 return -EPERM; 2569 2570 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2571 return -EOPNOTSUPP; 2572 2573 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2574 sizeof(range))) 2575 return -EFAULT; 2576 2577 ret = mnt_want_write_file(filp); 2578 if (ret) 2579 return ret; 2580 2581 range.minlen = max((unsigned int)range.minlen, 2582 bdev_discard_granularity(sb->s_bdev)); 2583 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2584 mnt_drop_write_file(filp); 2585 if (ret < 0) 2586 return ret; 2587 2588 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2589 sizeof(range))) 2590 return -EFAULT; 2591 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2592 return 0; 2593 } 2594 2595 static bool uuid_is_nonzero(__u8 u[16]) 2596 { 2597 int i; 2598 2599 for (i = 0; i < 16; i++) 2600 if (u[i]) 2601 return true; 2602 return false; 2603 } 2604 2605 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2606 { 2607 struct inode *inode = file_inode(filp); 2608 int ret; 2609 2610 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2611 return -EOPNOTSUPP; 2612 2613 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2614 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2615 return ret; 2616 } 2617 2618 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2619 { 2620 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2621 return -EOPNOTSUPP; 2622 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2623 } 2624 2625 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2626 { 2627 struct inode *inode = file_inode(filp); 2628 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2629 u8 encrypt_pw_salt[16]; 2630 int err; 2631 2632 if (!f2fs_sb_has_encrypt(sbi)) 2633 return -EOPNOTSUPP; 2634 2635 err = mnt_want_write_file(filp); 2636 if (err) 2637 return err; 2638 2639 f2fs_down_write(&sbi->sb_lock); 2640 2641 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2642 goto got_it; 2643 2644 /* update superblock with uuid */ 2645 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2646 2647 err = f2fs_commit_super(sbi, false); 2648 if (err) { 2649 /* undo new data */ 2650 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2651 goto out_err; 2652 } 2653 got_it: 2654 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2655 out_err: 2656 f2fs_up_write(&sbi->sb_lock); 2657 mnt_drop_write_file(filp); 2658 2659 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2660 err = -EFAULT; 2661 2662 return err; 2663 } 2664 2665 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2666 unsigned long arg) 2667 { 2668 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2669 return -EOPNOTSUPP; 2670 2671 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2672 } 2673 2674 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2675 { 2676 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2677 return -EOPNOTSUPP; 2678 2679 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2680 } 2681 2682 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2683 { 2684 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2685 return -EOPNOTSUPP; 2686 2687 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2688 } 2689 2690 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2691 unsigned long arg) 2692 { 2693 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2694 return -EOPNOTSUPP; 2695 2696 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2697 } 2698 2699 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2700 unsigned long arg) 2701 { 2702 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2703 return -EOPNOTSUPP; 2704 2705 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2706 } 2707 2708 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2709 { 2710 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2711 return -EOPNOTSUPP; 2712 2713 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2714 } 2715 2716 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2717 { 2718 struct inode *inode = file_inode(filp); 2719 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2720 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2721 .no_bg_gc = false, 2722 .should_migrate_blocks = false, 2723 .nr_free_secs = 0 }; 2724 __u32 sync; 2725 int ret; 2726 2727 if (!capable(CAP_SYS_ADMIN)) 2728 return -EPERM; 2729 2730 if (get_user(sync, (__u32 __user *)arg)) 2731 return -EFAULT; 2732 2733 if (f2fs_readonly(sbi->sb)) 2734 return -EROFS; 2735 2736 ret = mnt_want_write_file(filp); 2737 if (ret) 2738 return ret; 2739 2740 if (!sync) { 2741 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2742 ret = -EBUSY; 2743 goto out; 2744 } 2745 } else { 2746 f2fs_down_write(&sbi->gc_lock); 2747 } 2748 2749 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2750 gc_control.err_gc_skipped = sync; 2751 stat_inc_gc_call_count(sbi, FOREGROUND); 2752 ret = f2fs_gc(sbi, &gc_control); 2753 out: 2754 mnt_drop_write_file(filp); 2755 return ret; 2756 } 2757 2758 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2759 { 2760 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2761 struct f2fs_gc_control gc_control = { 2762 .init_gc_type = range->sync ? FG_GC : BG_GC, 2763 .no_bg_gc = false, 2764 .should_migrate_blocks = false, 2765 .err_gc_skipped = range->sync, 2766 .nr_free_secs = 0 }; 2767 u64 end; 2768 int ret; 2769 2770 if (!capable(CAP_SYS_ADMIN)) 2771 return -EPERM; 2772 if (f2fs_readonly(sbi->sb)) 2773 return -EROFS; 2774 2775 end = range->start + range->len; 2776 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2777 end >= MAX_BLKADDR(sbi)) 2778 return -EINVAL; 2779 2780 ret = mnt_want_write_file(filp); 2781 if (ret) 2782 return ret; 2783 2784 do_more: 2785 if (!range->sync) { 2786 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2787 ret = -EBUSY; 2788 goto out; 2789 } 2790 } else { 2791 f2fs_down_write(&sbi->gc_lock); 2792 } 2793 2794 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2795 stat_inc_gc_call_count(sbi, FOREGROUND); 2796 ret = f2fs_gc(sbi, &gc_control); 2797 if (ret) { 2798 if (ret == -EBUSY) 2799 ret = -EAGAIN; 2800 goto out; 2801 } 2802 range->start += CAP_BLKS_PER_SEC(sbi); 2803 if (range->start <= end) 2804 goto do_more; 2805 out: 2806 mnt_drop_write_file(filp); 2807 return ret; 2808 } 2809 2810 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2811 { 2812 struct f2fs_gc_range range; 2813 2814 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2815 sizeof(range))) 2816 return -EFAULT; 2817 return __f2fs_ioc_gc_range(filp, &range); 2818 } 2819 2820 static int f2fs_ioc_write_checkpoint(struct file *filp) 2821 { 2822 struct inode *inode = file_inode(filp); 2823 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2824 int ret; 2825 2826 if (!capable(CAP_SYS_ADMIN)) 2827 return -EPERM; 2828 2829 if (f2fs_readonly(sbi->sb)) 2830 return -EROFS; 2831 2832 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2833 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2834 return -EINVAL; 2835 } 2836 2837 ret = mnt_want_write_file(filp); 2838 if (ret) 2839 return ret; 2840 2841 ret = f2fs_sync_fs(sbi->sb, 1); 2842 2843 mnt_drop_write_file(filp); 2844 return ret; 2845 } 2846 2847 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2848 struct file *filp, 2849 struct f2fs_defragment *range) 2850 { 2851 struct inode *inode = file_inode(filp); 2852 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2853 .m_seg_type = NO_CHECK_TYPE, 2854 .m_may_create = false }; 2855 struct extent_info ei = {}; 2856 pgoff_t pg_start, pg_end, next_pgofs; 2857 unsigned int total = 0, sec_num; 2858 block_t blk_end = 0; 2859 bool fragmented = false; 2860 int err; 2861 2862 f2fs_balance_fs(sbi, true); 2863 2864 inode_lock(inode); 2865 pg_start = range->start >> PAGE_SHIFT; 2866 pg_end = min_t(pgoff_t, 2867 (range->start + range->len) >> PAGE_SHIFT, 2868 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE)); 2869 2870 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) || 2871 f2fs_is_atomic_file(inode)) { 2872 err = -EINVAL; 2873 goto unlock_out; 2874 } 2875 2876 /* if in-place-update policy is enabled, don't waste time here */ 2877 set_inode_flag(inode, FI_OPU_WRITE); 2878 if (f2fs_should_update_inplace(inode, NULL)) { 2879 err = -EINVAL; 2880 goto out; 2881 } 2882 2883 /* writeback all dirty pages in the range */ 2884 err = filemap_write_and_wait_range(inode->i_mapping, 2885 pg_start << PAGE_SHIFT, 2886 (pg_end << PAGE_SHIFT) - 1); 2887 if (err) 2888 goto out; 2889 2890 /* 2891 * lookup mapping info in extent cache, skip defragmenting if physical 2892 * block addresses are continuous. 2893 */ 2894 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2895 if ((pgoff_t)ei.fofs + ei.len >= pg_end) 2896 goto out; 2897 } 2898 2899 map.m_lblk = pg_start; 2900 map.m_next_pgofs = &next_pgofs; 2901 2902 /* 2903 * lookup mapping info in dnode page cache, skip defragmenting if all 2904 * physical block addresses are continuous even if there are hole(s) 2905 * in logical blocks. 2906 */ 2907 while (map.m_lblk < pg_end) { 2908 map.m_len = pg_end - map.m_lblk; 2909 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2910 if (err) 2911 goto out; 2912 2913 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2914 map.m_lblk = next_pgofs; 2915 continue; 2916 } 2917 2918 if (blk_end && blk_end != map.m_pblk) 2919 fragmented = true; 2920 2921 /* record total count of block that we're going to move */ 2922 total += map.m_len; 2923 2924 blk_end = map.m_pblk + map.m_len; 2925 2926 map.m_lblk += map.m_len; 2927 } 2928 2929 if (!fragmented) { 2930 total = 0; 2931 goto out; 2932 } 2933 2934 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2935 2936 /* 2937 * make sure there are enough free section for LFS allocation, this can 2938 * avoid defragment running in SSR mode when free section are allocated 2939 * intensively 2940 */ 2941 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2942 err = -EAGAIN; 2943 goto out; 2944 } 2945 2946 map.m_lblk = pg_start; 2947 map.m_len = pg_end - pg_start; 2948 total = 0; 2949 2950 while (map.m_lblk < pg_end) { 2951 pgoff_t idx; 2952 int cnt = 0; 2953 2954 do_map: 2955 map.m_len = pg_end - map.m_lblk; 2956 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2957 if (err) 2958 goto clear_out; 2959 2960 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2961 map.m_lblk = next_pgofs; 2962 goto check; 2963 } 2964 2965 set_inode_flag(inode, FI_SKIP_WRITES); 2966 2967 idx = map.m_lblk; 2968 while (idx < map.m_lblk + map.m_len && 2969 cnt < BLKS_PER_SEG(sbi)) { 2970 struct folio *folio; 2971 2972 folio = f2fs_get_lock_data_folio(inode, idx, true); 2973 if (IS_ERR(folio)) { 2974 err = PTR_ERR(folio); 2975 goto clear_out; 2976 } 2977 2978 f2fs_folio_wait_writeback(folio, DATA, true, true); 2979 2980 folio_mark_dirty(folio); 2981 set_page_private_gcing(&folio->page); 2982 f2fs_folio_put(folio, true); 2983 2984 idx++; 2985 cnt++; 2986 total++; 2987 } 2988 2989 map.m_lblk = idx; 2990 check: 2991 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 2992 goto do_map; 2993 2994 clear_inode_flag(inode, FI_SKIP_WRITES); 2995 2996 err = filemap_fdatawrite(inode->i_mapping); 2997 if (err) 2998 goto out; 2999 } 3000 clear_out: 3001 clear_inode_flag(inode, FI_SKIP_WRITES); 3002 out: 3003 clear_inode_flag(inode, FI_OPU_WRITE); 3004 unlock_out: 3005 inode_unlock(inode); 3006 if (!err) 3007 range->len = (u64)total << PAGE_SHIFT; 3008 return err; 3009 } 3010 3011 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 3012 { 3013 struct inode *inode = file_inode(filp); 3014 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3015 struct f2fs_defragment range; 3016 int err; 3017 3018 if (!capable(CAP_SYS_ADMIN)) 3019 return -EPERM; 3020 3021 if (!S_ISREG(inode->i_mode)) 3022 return -EINVAL; 3023 3024 if (f2fs_readonly(sbi->sb)) 3025 return -EROFS; 3026 3027 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 3028 sizeof(range))) 3029 return -EFAULT; 3030 3031 /* verify alignment of offset & size */ 3032 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 3033 return -EINVAL; 3034 3035 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 3036 max_file_blocks(inode))) 3037 return -EINVAL; 3038 3039 err = mnt_want_write_file(filp); 3040 if (err) 3041 return err; 3042 3043 err = f2fs_defragment_range(sbi, filp, &range); 3044 mnt_drop_write_file(filp); 3045 3046 if (range.len) 3047 f2fs_update_time(sbi, REQ_TIME); 3048 if (err < 0) 3049 return err; 3050 3051 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 3052 sizeof(range))) 3053 return -EFAULT; 3054 3055 return 0; 3056 } 3057 3058 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 3059 struct file *file_out, loff_t pos_out, size_t len) 3060 { 3061 struct inode *src = file_inode(file_in); 3062 struct inode *dst = file_inode(file_out); 3063 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 3064 size_t olen = len, dst_max_i_size = 0; 3065 size_t dst_osize; 3066 int ret; 3067 3068 if (file_in->f_path.mnt != file_out->f_path.mnt || 3069 src->i_sb != dst->i_sb) 3070 return -EXDEV; 3071 3072 if (unlikely(f2fs_readonly(src->i_sb))) 3073 return -EROFS; 3074 3075 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 3076 return -EINVAL; 3077 3078 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 3079 return -EOPNOTSUPP; 3080 3081 if (pos_out < 0 || pos_in < 0) 3082 return -EINVAL; 3083 3084 if (src == dst) { 3085 if (pos_in == pos_out) 3086 return 0; 3087 if (pos_out > pos_in && pos_out < pos_in + len) 3088 return -EINVAL; 3089 } 3090 3091 inode_lock(src); 3092 if (src != dst) { 3093 ret = -EBUSY; 3094 if (!inode_trylock(dst)) 3095 goto out; 3096 } 3097 3098 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) || 3099 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 3100 ret = -EOPNOTSUPP; 3101 goto out_unlock; 3102 } 3103 3104 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) { 3105 ret = -EINVAL; 3106 goto out_unlock; 3107 } 3108 3109 ret = -EINVAL; 3110 if (pos_in + len > src->i_size || pos_in + len < pos_in) 3111 goto out_unlock; 3112 if (len == 0) 3113 olen = len = src->i_size - pos_in; 3114 if (pos_in + len == src->i_size) 3115 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 3116 if (len == 0) { 3117 ret = 0; 3118 goto out_unlock; 3119 } 3120 3121 dst_osize = dst->i_size; 3122 if (pos_out + olen > dst->i_size) 3123 dst_max_i_size = pos_out + olen; 3124 3125 /* verify the end result is block aligned */ 3126 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 3127 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 3128 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 3129 goto out_unlock; 3130 3131 ret = f2fs_convert_inline_inode(src); 3132 if (ret) 3133 goto out_unlock; 3134 3135 ret = f2fs_convert_inline_inode(dst); 3136 if (ret) 3137 goto out_unlock; 3138 3139 /* write out all dirty pages from offset */ 3140 ret = filemap_write_and_wait_range(src->i_mapping, 3141 pos_in, pos_in + len); 3142 if (ret) 3143 goto out_unlock; 3144 3145 ret = filemap_write_and_wait_range(dst->i_mapping, 3146 pos_out, pos_out + len); 3147 if (ret) 3148 goto out_unlock; 3149 3150 f2fs_balance_fs(sbi, true); 3151 3152 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3153 if (src != dst) { 3154 ret = -EBUSY; 3155 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 3156 goto out_src; 3157 } 3158 3159 f2fs_lock_op(sbi); 3160 ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in), 3161 F2FS_BYTES_TO_BLK(pos_out), 3162 F2FS_BYTES_TO_BLK(len), false); 3163 3164 if (!ret) { 3165 if (dst_max_i_size) 3166 f2fs_i_size_write(dst, dst_max_i_size); 3167 else if (dst_osize != dst->i_size) 3168 f2fs_i_size_write(dst, dst_osize); 3169 } 3170 f2fs_unlock_op(sbi); 3171 3172 if (src != dst) 3173 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 3174 out_src: 3175 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3176 if (ret) 3177 goto out_unlock; 3178 3179 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 3180 f2fs_mark_inode_dirty_sync(src, false); 3181 if (src != dst) { 3182 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 3183 f2fs_mark_inode_dirty_sync(dst, false); 3184 } 3185 f2fs_update_time(sbi, REQ_TIME); 3186 3187 out_unlock: 3188 if (src != dst) 3189 inode_unlock(dst); 3190 out: 3191 inode_unlock(src); 3192 return ret; 3193 } 3194 3195 static int __f2fs_ioc_move_range(struct file *filp, 3196 struct f2fs_move_range *range) 3197 { 3198 int err; 3199 3200 if (!(filp->f_mode & FMODE_READ) || 3201 !(filp->f_mode & FMODE_WRITE)) 3202 return -EBADF; 3203 3204 CLASS(fd, dst)(range->dst_fd); 3205 if (fd_empty(dst)) 3206 return -EBADF; 3207 3208 if (!(fd_file(dst)->f_mode & FMODE_WRITE)) 3209 return -EBADF; 3210 3211 err = mnt_want_write_file(filp); 3212 if (err) 3213 return err; 3214 3215 err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst), 3216 range->pos_out, range->len); 3217 3218 mnt_drop_write_file(filp); 3219 return err; 3220 } 3221 3222 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 3223 { 3224 struct f2fs_move_range range; 3225 3226 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 3227 sizeof(range))) 3228 return -EFAULT; 3229 return __f2fs_ioc_move_range(filp, &range); 3230 } 3231 3232 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 3233 { 3234 struct inode *inode = file_inode(filp); 3235 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3236 struct sit_info *sm = SIT_I(sbi); 3237 unsigned int start_segno = 0, end_segno = 0; 3238 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3239 struct f2fs_flush_device range; 3240 struct f2fs_gc_control gc_control = { 3241 .init_gc_type = FG_GC, 3242 .should_migrate_blocks = true, 3243 .err_gc_skipped = true, 3244 .nr_free_secs = 0 }; 3245 int ret; 3246 3247 if (!capable(CAP_SYS_ADMIN)) 3248 return -EPERM; 3249 3250 if (f2fs_readonly(sbi->sb)) 3251 return -EROFS; 3252 3253 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3254 return -EINVAL; 3255 3256 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3257 sizeof(range))) 3258 return -EFAULT; 3259 3260 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3261 __is_large_section(sbi)) { 3262 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3263 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3264 return -EINVAL; 3265 } 3266 3267 ret = mnt_want_write_file(filp); 3268 if (ret) 3269 return ret; 3270 3271 if (range.dev_num != 0) 3272 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3273 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3274 3275 start_segno = sm->last_victim[FLUSH_DEVICE]; 3276 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3277 start_segno = dev_start_segno; 3278 end_segno = min(start_segno + range.segments, dev_end_segno); 3279 3280 while (start_segno < end_segno) { 3281 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3282 ret = -EBUSY; 3283 goto out; 3284 } 3285 sm->last_victim[GC_CB] = end_segno + 1; 3286 sm->last_victim[GC_GREEDY] = end_segno + 1; 3287 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3288 3289 gc_control.victim_segno = start_segno; 3290 stat_inc_gc_call_count(sbi, FOREGROUND); 3291 ret = f2fs_gc(sbi, &gc_control); 3292 if (ret == -EAGAIN) 3293 ret = 0; 3294 else if (ret < 0) 3295 break; 3296 start_segno++; 3297 } 3298 out: 3299 mnt_drop_write_file(filp); 3300 return ret; 3301 } 3302 3303 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3304 { 3305 struct inode *inode = file_inode(filp); 3306 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3307 3308 /* Must validate to set it with SQLite behavior in Android. */ 3309 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3310 3311 return put_user(sb_feature, (u32 __user *)arg); 3312 } 3313 3314 #ifdef CONFIG_QUOTA 3315 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3316 { 3317 struct dquot *transfer_to[MAXQUOTAS] = {}; 3318 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3319 struct super_block *sb = sbi->sb; 3320 int err; 3321 3322 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3323 if (IS_ERR(transfer_to[PRJQUOTA])) 3324 return PTR_ERR(transfer_to[PRJQUOTA]); 3325 3326 err = __dquot_transfer(inode, transfer_to); 3327 if (err) 3328 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3329 dqput(transfer_to[PRJQUOTA]); 3330 return err; 3331 } 3332 3333 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3334 { 3335 struct f2fs_inode_info *fi = F2FS_I(inode); 3336 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3337 struct f2fs_inode *ri = NULL; 3338 kprojid_t kprojid; 3339 int err; 3340 3341 if (!f2fs_sb_has_project_quota(sbi)) { 3342 if (projid != F2FS_DEF_PROJID) 3343 return -EOPNOTSUPP; 3344 else 3345 return 0; 3346 } 3347 3348 if (!f2fs_has_extra_attr(inode)) 3349 return -EOPNOTSUPP; 3350 3351 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3352 3353 if (projid_eq(kprojid, fi->i_projid)) 3354 return 0; 3355 3356 err = -EPERM; 3357 /* Is it quota file? Do not allow user to mess with it */ 3358 if (IS_NOQUOTA(inode)) 3359 return err; 3360 3361 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3362 return -EOVERFLOW; 3363 3364 err = f2fs_dquot_initialize(inode); 3365 if (err) 3366 return err; 3367 3368 f2fs_lock_op(sbi); 3369 err = f2fs_transfer_project_quota(inode, kprojid); 3370 if (err) 3371 goto out_unlock; 3372 3373 fi->i_projid = kprojid; 3374 inode_set_ctime_current(inode); 3375 f2fs_mark_inode_dirty_sync(inode, true); 3376 out_unlock: 3377 f2fs_unlock_op(sbi); 3378 return err; 3379 } 3380 #else 3381 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3382 { 3383 return 0; 3384 } 3385 3386 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3387 { 3388 if (projid != F2FS_DEF_PROJID) 3389 return -EOPNOTSUPP; 3390 return 0; 3391 } 3392 #endif 3393 3394 int f2fs_fileattr_get(struct dentry *dentry, struct file_kattr *fa) 3395 { 3396 struct inode *inode = d_inode(dentry); 3397 struct f2fs_inode_info *fi = F2FS_I(inode); 3398 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3399 3400 if (IS_ENCRYPTED(inode)) 3401 fsflags |= FS_ENCRYPT_FL; 3402 if (IS_VERITY(inode)) 3403 fsflags |= FS_VERITY_FL; 3404 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3405 fsflags |= FS_INLINE_DATA_FL; 3406 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3407 fsflags |= FS_NOCOW_FL; 3408 3409 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3410 3411 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3412 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3413 3414 return 0; 3415 } 3416 3417 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3418 struct dentry *dentry, struct file_kattr *fa) 3419 { 3420 struct inode *inode = d_inode(dentry); 3421 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3422 u32 iflags; 3423 int err; 3424 3425 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3426 return -EIO; 3427 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3428 return -ENOSPC; 3429 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3430 return -EOPNOTSUPP; 3431 fsflags &= F2FS_SETTABLE_FS_FL; 3432 if (!fa->flags_valid) 3433 mask &= FS_COMMON_FL; 3434 3435 iflags = f2fs_fsflags_to_iflags(fsflags); 3436 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3437 return -EOPNOTSUPP; 3438 3439 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3440 if (!err) 3441 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3442 3443 return err; 3444 } 3445 3446 int f2fs_pin_file_control(struct inode *inode, bool inc) 3447 { 3448 struct f2fs_inode_info *fi = F2FS_I(inode); 3449 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3450 3451 if (IS_DEVICE_ALIASING(inode)) 3452 return -EINVAL; 3453 3454 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) { 3455 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3456 __func__, inode->i_ino, fi->i_gc_failures); 3457 clear_inode_flag(inode, FI_PIN_FILE); 3458 return -EAGAIN; 3459 } 3460 3461 /* Use i_gc_failures for normal file as a risk signal. */ 3462 if (inc) 3463 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1); 3464 3465 return 0; 3466 } 3467 3468 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3469 { 3470 struct inode *inode = file_inode(filp); 3471 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3472 __u32 pin; 3473 int ret = 0; 3474 3475 if (get_user(pin, (__u32 __user *)arg)) 3476 return -EFAULT; 3477 3478 if (!S_ISREG(inode->i_mode)) 3479 return -EINVAL; 3480 3481 if (f2fs_readonly(sbi->sb)) 3482 return -EROFS; 3483 3484 if (!pin && IS_DEVICE_ALIASING(inode)) 3485 return -EOPNOTSUPP; 3486 3487 ret = mnt_want_write_file(filp); 3488 if (ret) 3489 return ret; 3490 3491 inode_lock(inode); 3492 3493 if (f2fs_is_atomic_file(inode)) { 3494 ret = -EINVAL; 3495 goto out; 3496 } 3497 3498 if (!pin) { 3499 clear_inode_flag(inode, FI_PIN_FILE); 3500 f2fs_i_gc_failures_write(inode, 0); 3501 goto done; 3502 } else if (f2fs_is_pinned_file(inode)) { 3503 goto done; 3504 } 3505 3506 if (F2FS_HAS_BLOCKS(inode)) { 3507 ret = -EFBIG; 3508 goto out; 3509 } 3510 3511 /* Let's allow file pinning on zoned device. */ 3512 if (!f2fs_sb_has_blkzoned(sbi) && 3513 f2fs_should_update_outplace(inode, NULL)) { 3514 ret = -EINVAL; 3515 goto out; 3516 } 3517 3518 if (f2fs_pin_file_control(inode, false)) { 3519 ret = -EAGAIN; 3520 goto out; 3521 } 3522 3523 ret = f2fs_convert_inline_inode(inode); 3524 if (ret) 3525 goto out; 3526 3527 if (!f2fs_disable_compressed_file(inode)) { 3528 ret = -EOPNOTSUPP; 3529 goto out; 3530 } 3531 3532 set_inode_flag(inode, FI_PIN_FILE); 3533 ret = F2FS_I(inode)->i_gc_failures; 3534 done: 3535 f2fs_update_time(sbi, REQ_TIME); 3536 out: 3537 inode_unlock(inode); 3538 mnt_drop_write_file(filp); 3539 return ret; 3540 } 3541 3542 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3543 { 3544 struct inode *inode = file_inode(filp); 3545 __u32 pin = 0; 3546 3547 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3548 pin = F2FS_I(inode)->i_gc_failures; 3549 return put_user(pin, (u32 __user *)arg); 3550 } 3551 3552 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg) 3553 { 3554 return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0, 3555 (u32 __user *)arg); 3556 } 3557 3558 static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg) 3559 { 3560 struct inode *inode = file_inode(filp); 3561 __u32 level; 3562 3563 if (get_user(level, (__u32 __user *)arg)) 3564 return -EFAULT; 3565 3566 if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX) 3567 return -EINVAL; 3568 3569 inode_lock(inode); 3570 F2FS_I(inode)->ioprio_hint = level; 3571 inode_unlock(inode); 3572 return 0; 3573 } 3574 3575 int f2fs_precache_extents(struct inode *inode) 3576 { 3577 struct f2fs_inode_info *fi = F2FS_I(inode); 3578 struct f2fs_map_blocks map; 3579 pgoff_t m_next_extent; 3580 loff_t end; 3581 int err; 3582 3583 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3584 return -EOPNOTSUPP; 3585 3586 map.m_lblk = 0; 3587 map.m_pblk = 0; 3588 map.m_next_pgofs = NULL; 3589 map.m_next_extent = &m_next_extent; 3590 map.m_seg_type = NO_CHECK_TYPE; 3591 map.m_may_create = false; 3592 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3593 3594 while (map.m_lblk < end) { 3595 map.m_len = end - map.m_lblk; 3596 3597 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3598 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3599 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3600 if (err || !map.m_len) 3601 return err; 3602 3603 map.m_lblk = m_next_extent; 3604 } 3605 3606 return 0; 3607 } 3608 3609 static int f2fs_ioc_precache_extents(struct file *filp) 3610 { 3611 return f2fs_precache_extents(file_inode(filp)); 3612 } 3613 3614 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3615 { 3616 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3617 __u64 block_count; 3618 3619 if (!capable(CAP_SYS_ADMIN)) 3620 return -EPERM; 3621 3622 if (f2fs_readonly(sbi->sb)) 3623 return -EROFS; 3624 3625 if (copy_from_user(&block_count, (void __user *)arg, 3626 sizeof(block_count))) 3627 return -EFAULT; 3628 3629 return f2fs_resize_fs(filp, block_count); 3630 } 3631 3632 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3633 { 3634 struct inode *inode = file_inode(filp); 3635 3636 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3637 3638 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3639 f2fs_warn(F2FS_I_SB(inode), 3640 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3641 inode->i_ino); 3642 return -EOPNOTSUPP; 3643 } 3644 3645 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3646 } 3647 3648 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3649 { 3650 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3651 return -EOPNOTSUPP; 3652 3653 return fsverity_ioctl_measure(filp, (void __user *)arg); 3654 } 3655 3656 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3657 { 3658 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3659 return -EOPNOTSUPP; 3660 3661 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3662 } 3663 3664 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3665 { 3666 struct inode *inode = file_inode(filp); 3667 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3668 char *vbuf; 3669 int count; 3670 int err = 0; 3671 3672 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3673 if (!vbuf) 3674 return -ENOMEM; 3675 3676 f2fs_down_read(&sbi->sb_lock); 3677 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3678 ARRAY_SIZE(sbi->raw_super->volume_name), 3679 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3680 f2fs_up_read(&sbi->sb_lock); 3681 3682 if (copy_to_user((char __user *)arg, vbuf, 3683 min(FSLABEL_MAX, count))) 3684 err = -EFAULT; 3685 3686 kfree(vbuf); 3687 return err; 3688 } 3689 3690 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3691 { 3692 struct inode *inode = file_inode(filp); 3693 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3694 char *vbuf; 3695 int err = 0; 3696 3697 if (!capable(CAP_SYS_ADMIN)) 3698 return -EPERM; 3699 3700 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3701 if (IS_ERR(vbuf)) 3702 return PTR_ERR(vbuf); 3703 3704 err = mnt_want_write_file(filp); 3705 if (err) 3706 goto out; 3707 3708 f2fs_down_write(&sbi->sb_lock); 3709 3710 memset(sbi->raw_super->volume_name, 0, 3711 sizeof(sbi->raw_super->volume_name)); 3712 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3713 sbi->raw_super->volume_name, 3714 ARRAY_SIZE(sbi->raw_super->volume_name)); 3715 3716 err = f2fs_commit_super(sbi, false); 3717 3718 f2fs_up_write(&sbi->sb_lock); 3719 3720 mnt_drop_write_file(filp); 3721 out: 3722 kfree(vbuf); 3723 return err; 3724 } 3725 3726 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3727 { 3728 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3729 return -EOPNOTSUPP; 3730 3731 if (!f2fs_compressed_file(inode)) 3732 return -EINVAL; 3733 3734 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3735 3736 return 0; 3737 } 3738 3739 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3740 { 3741 struct inode *inode = file_inode(filp); 3742 __u64 blocks; 3743 int ret; 3744 3745 ret = f2fs_get_compress_blocks(inode, &blocks); 3746 if (ret < 0) 3747 return ret; 3748 3749 return put_user(blocks, (u64 __user *)arg); 3750 } 3751 3752 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3753 { 3754 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3755 unsigned int released_blocks = 0; 3756 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3757 block_t blkaddr; 3758 int i; 3759 3760 for (i = 0; i < count; i++) { 3761 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3762 dn->ofs_in_node + i); 3763 3764 if (!__is_valid_data_blkaddr(blkaddr)) 3765 continue; 3766 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3767 DATA_GENERIC_ENHANCE))) 3768 return -EFSCORRUPTED; 3769 } 3770 3771 while (count) { 3772 int compr_blocks = 0; 3773 3774 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3775 blkaddr = f2fs_data_blkaddr(dn); 3776 3777 if (i == 0) { 3778 if (blkaddr == COMPRESS_ADDR) 3779 continue; 3780 dn->ofs_in_node += cluster_size; 3781 goto next; 3782 } 3783 3784 if (__is_valid_data_blkaddr(blkaddr)) 3785 compr_blocks++; 3786 3787 if (blkaddr != NEW_ADDR) 3788 continue; 3789 3790 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3791 } 3792 3793 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3794 dec_valid_block_count(sbi, dn->inode, 3795 cluster_size - compr_blocks); 3796 3797 released_blocks += cluster_size - compr_blocks; 3798 next: 3799 count -= cluster_size; 3800 } 3801 3802 return released_blocks; 3803 } 3804 3805 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3806 { 3807 struct inode *inode = file_inode(filp); 3808 struct f2fs_inode_info *fi = F2FS_I(inode); 3809 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3810 pgoff_t page_idx = 0, last_idx; 3811 unsigned int released_blocks = 0; 3812 int ret; 3813 int writecount; 3814 3815 if (!f2fs_sb_has_compression(sbi)) 3816 return -EOPNOTSUPP; 3817 3818 if (f2fs_readonly(sbi->sb)) 3819 return -EROFS; 3820 3821 ret = mnt_want_write_file(filp); 3822 if (ret) 3823 return ret; 3824 3825 f2fs_balance_fs(sbi, true); 3826 3827 inode_lock(inode); 3828 3829 writecount = atomic_read(&inode->i_writecount); 3830 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3831 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3832 ret = -EBUSY; 3833 goto out; 3834 } 3835 3836 if (!f2fs_compressed_file(inode) || 3837 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3838 ret = -EINVAL; 3839 goto out; 3840 } 3841 3842 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3843 if (ret) 3844 goto out; 3845 3846 if (!atomic_read(&fi->i_compr_blocks)) { 3847 ret = -EPERM; 3848 goto out; 3849 } 3850 3851 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3852 inode_set_ctime_current(inode); 3853 f2fs_mark_inode_dirty_sync(inode, true); 3854 3855 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3856 filemap_invalidate_lock(inode->i_mapping); 3857 3858 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3859 3860 while (page_idx < last_idx) { 3861 struct dnode_of_data dn; 3862 pgoff_t end_offset, count; 3863 3864 f2fs_lock_op(sbi); 3865 3866 set_new_dnode(&dn, inode, NULL, NULL, 0); 3867 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3868 if (ret) { 3869 f2fs_unlock_op(sbi); 3870 if (ret == -ENOENT) { 3871 page_idx = f2fs_get_next_page_offset(&dn, 3872 page_idx); 3873 ret = 0; 3874 continue; 3875 } 3876 break; 3877 } 3878 3879 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 3880 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3881 count = round_up(count, fi->i_cluster_size); 3882 3883 ret = release_compress_blocks(&dn, count); 3884 3885 f2fs_put_dnode(&dn); 3886 3887 f2fs_unlock_op(sbi); 3888 3889 if (ret < 0) 3890 break; 3891 3892 page_idx += count; 3893 released_blocks += ret; 3894 } 3895 3896 filemap_invalidate_unlock(inode->i_mapping); 3897 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3898 out: 3899 if (released_blocks) 3900 f2fs_update_time(sbi, REQ_TIME); 3901 inode_unlock(inode); 3902 3903 mnt_drop_write_file(filp); 3904 3905 if (ret >= 0) { 3906 ret = put_user(released_blocks, (u64 __user *)arg); 3907 } else if (released_blocks && 3908 atomic_read(&fi->i_compr_blocks)) { 3909 set_sbi_flag(sbi, SBI_NEED_FSCK); 3910 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3911 "iblocks=%llu, released=%u, compr_blocks=%u, " 3912 "run fsck to fix.", 3913 __func__, inode->i_ino, inode->i_blocks, 3914 released_blocks, 3915 atomic_read(&fi->i_compr_blocks)); 3916 } 3917 3918 return ret; 3919 } 3920 3921 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3922 unsigned int *reserved_blocks) 3923 { 3924 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3925 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3926 block_t blkaddr; 3927 int i; 3928 3929 for (i = 0; i < count; i++) { 3930 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3931 dn->ofs_in_node + i); 3932 3933 if (!__is_valid_data_blkaddr(blkaddr)) 3934 continue; 3935 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3936 DATA_GENERIC_ENHANCE))) 3937 return -EFSCORRUPTED; 3938 } 3939 3940 while (count) { 3941 int compr_blocks = 0; 3942 blkcnt_t reserved = 0; 3943 blkcnt_t to_reserved; 3944 int ret; 3945 3946 for (i = 0; i < cluster_size; i++) { 3947 blkaddr = data_blkaddr(dn->inode, dn->node_folio, 3948 dn->ofs_in_node + i); 3949 3950 if (i == 0) { 3951 if (blkaddr != COMPRESS_ADDR) { 3952 dn->ofs_in_node += cluster_size; 3953 goto next; 3954 } 3955 continue; 3956 } 3957 3958 /* 3959 * compressed cluster was not released due to it 3960 * fails in release_compress_blocks(), so NEW_ADDR 3961 * is a possible case. 3962 */ 3963 if (blkaddr == NEW_ADDR) { 3964 reserved++; 3965 continue; 3966 } 3967 if (__is_valid_data_blkaddr(blkaddr)) { 3968 compr_blocks++; 3969 continue; 3970 } 3971 } 3972 3973 to_reserved = cluster_size - compr_blocks - reserved; 3974 3975 /* for the case all blocks in cluster were reserved */ 3976 if (reserved && to_reserved == 1) { 3977 dn->ofs_in_node += cluster_size; 3978 goto next; 3979 } 3980 3981 ret = inc_valid_block_count(sbi, dn->inode, 3982 &to_reserved, false); 3983 if (unlikely(ret)) 3984 return ret; 3985 3986 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3987 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3988 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3989 } 3990 3991 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3992 3993 *reserved_blocks += to_reserved; 3994 next: 3995 count -= cluster_size; 3996 } 3997 3998 return 0; 3999 } 4000 4001 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 4002 { 4003 struct inode *inode = file_inode(filp); 4004 struct f2fs_inode_info *fi = F2FS_I(inode); 4005 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4006 pgoff_t page_idx = 0, last_idx; 4007 unsigned int reserved_blocks = 0; 4008 int ret; 4009 4010 if (!f2fs_sb_has_compression(sbi)) 4011 return -EOPNOTSUPP; 4012 4013 if (f2fs_readonly(sbi->sb)) 4014 return -EROFS; 4015 4016 ret = mnt_want_write_file(filp); 4017 if (ret) 4018 return ret; 4019 4020 f2fs_balance_fs(sbi, true); 4021 4022 inode_lock(inode); 4023 4024 if (!f2fs_compressed_file(inode) || 4025 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4026 ret = -EINVAL; 4027 goto unlock_inode; 4028 } 4029 4030 if (atomic_read(&fi->i_compr_blocks)) 4031 goto unlock_inode; 4032 4033 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 4034 filemap_invalidate_lock(inode->i_mapping); 4035 4036 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4037 4038 while (page_idx < last_idx) { 4039 struct dnode_of_data dn; 4040 pgoff_t end_offset, count; 4041 4042 f2fs_lock_op(sbi); 4043 4044 set_new_dnode(&dn, inode, NULL, NULL, 0); 4045 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 4046 if (ret) { 4047 f2fs_unlock_op(sbi); 4048 if (ret == -ENOENT) { 4049 page_idx = f2fs_get_next_page_offset(&dn, 4050 page_idx); 4051 ret = 0; 4052 continue; 4053 } 4054 break; 4055 } 4056 4057 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 4058 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 4059 count = round_up(count, fi->i_cluster_size); 4060 4061 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 4062 4063 f2fs_put_dnode(&dn); 4064 4065 f2fs_unlock_op(sbi); 4066 4067 if (ret < 0) 4068 break; 4069 4070 page_idx += count; 4071 } 4072 4073 filemap_invalidate_unlock(inode->i_mapping); 4074 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 4075 4076 if (!ret) { 4077 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 4078 inode_set_ctime_current(inode); 4079 f2fs_mark_inode_dirty_sync(inode, true); 4080 } 4081 unlock_inode: 4082 if (reserved_blocks) 4083 f2fs_update_time(sbi, REQ_TIME); 4084 inode_unlock(inode); 4085 mnt_drop_write_file(filp); 4086 4087 if (!ret) { 4088 ret = put_user(reserved_blocks, (u64 __user *)arg); 4089 } else if (reserved_blocks && 4090 atomic_read(&fi->i_compr_blocks)) { 4091 set_sbi_flag(sbi, SBI_NEED_FSCK); 4092 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx " 4093 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 4094 "run fsck to fix.", 4095 __func__, inode->i_ino, inode->i_blocks, 4096 reserved_blocks, 4097 atomic_read(&fi->i_compr_blocks)); 4098 } 4099 4100 return ret; 4101 } 4102 4103 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 4104 pgoff_t off, block_t block, block_t len, u32 flags) 4105 { 4106 sector_t sector = SECTOR_FROM_BLOCK(block); 4107 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 4108 int ret = 0; 4109 4110 if (flags & F2FS_TRIM_FILE_DISCARD) { 4111 if (bdev_max_secure_erase_sectors(bdev)) 4112 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 4113 GFP_NOFS); 4114 else 4115 ret = blkdev_issue_discard(bdev, sector, nr_sects, 4116 GFP_NOFS); 4117 } 4118 4119 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 4120 if (IS_ENCRYPTED(inode)) 4121 ret = fscrypt_zeroout_range(inode, off, block, len); 4122 else 4123 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 4124 GFP_NOFS, 0); 4125 } 4126 4127 return ret; 4128 } 4129 4130 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 4131 { 4132 struct inode *inode = file_inode(filp); 4133 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4134 struct address_space *mapping = inode->i_mapping; 4135 struct block_device *prev_bdev = NULL; 4136 struct f2fs_sectrim_range range; 4137 pgoff_t index, pg_end, prev_index = 0; 4138 block_t prev_block = 0, len = 0; 4139 loff_t end_addr; 4140 bool to_end = false; 4141 int ret = 0; 4142 4143 if (!(filp->f_mode & FMODE_WRITE)) 4144 return -EBADF; 4145 4146 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 4147 sizeof(range))) 4148 return -EFAULT; 4149 4150 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 4151 !S_ISREG(inode->i_mode)) 4152 return -EINVAL; 4153 4154 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 4155 !f2fs_hw_support_discard(sbi)) || 4156 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 4157 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 4158 return -EOPNOTSUPP; 4159 4160 ret = mnt_want_write_file(filp); 4161 if (ret) 4162 return ret; 4163 inode_lock(inode); 4164 4165 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 4166 range.start >= inode->i_size) { 4167 ret = -EINVAL; 4168 goto err; 4169 } 4170 4171 if (range.len == 0) 4172 goto err; 4173 4174 if (inode->i_size - range.start > range.len) { 4175 end_addr = range.start + range.len; 4176 } else { 4177 end_addr = range.len == (u64)-1 ? 4178 sbi->sb->s_maxbytes : inode->i_size; 4179 to_end = true; 4180 } 4181 4182 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 4183 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 4184 ret = -EINVAL; 4185 goto err; 4186 } 4187 4188 index = F2FS_BYTES_TO_BLK(range.start); 4189 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 4190 4191 ret = f2fs_convert_inline_inode(inode); 4192 if (ret) 4193 goto err; 4194 4195 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4196 filemap_invalidate_lock(mapping); 4197 4198 ret = filemap_write_and_wait_range(mapping, range.start, 4199 to_end ? LLONG_MAX : end_addr - 1); 4200 if (ret) 4201 goto out; 4202 4203 truncate_inode_pages_range(mapping, range.start, 4204 to_end ? -1 : end_addr - 1); 4205 4206 while (index < pg_end) { 4207 struct dnode_of_data dn; 4208 pgoff_t end_offset, count; 4209 int i; 4210 4211 set_new_dnode(&dn, inode, NULL, NULL, 0); 4212 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 4213 if (ret) { 4214 if (ret == -ENOENT) { 4215 index = f2fs_get_next_page_offset(&dn, index); 4216 continue; 4217 } 4218 goto out; 4219 } 4220 4221 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode); 4222 count = min(end_offset - dn.ofs_in_node, pg_end - index); 4223 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 4224 struct block_device *cur_bdev; 4225 block_t blkaddr = f2fs_data_blkaddr(&dn); 4226 4227 if (!__is_valid_data_blkaddr(blkaddr)) 4228 continue; 4229 4230 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 4231 DATA_GENERIC_ENHANCE)) { 4232 ret = -EFSCORRUPTED; 4233 f2fs_put_dnode(&dn); 4234 goto out; 4235 } 4236 4237 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 4238 if (f2fs_is_multi_device(sbi)) { 4239 int di = f2fs_target_device_index(sbi, blkaddr); 4240 4241 blkaddr -= FDEV(di).start_blk; 4242 } 4243 4244 if (len) { 4245 if (prev_bdev == cur_bdev && 4246 index == prev_index + len && 4247 blkaddr == prev_block + len) { 4248 len++; 4249 } else { 4250 ret = f2fs_secure_erase(prev_bdev, 4251 inode, prev_index, prev_block, 4252 len, range.flags); 4253 if (ret) { 4254 f2fs_put_dnode(&dn); 4255 goto out; 4256 } 4257 4258 len = 0; 4259 } 4260 } 4261 4262 if (!len) { 4263 prev_bdev = cur_bdev; 4264 prev_index = index; 4265 prev_block = blkaddr; 4266 len = 1; 4267 } 4268 } 4269 4270 f2fs_put_dnode(&dn); 4271 4272 if (fatal_signal_pending(current)) { 4273 ret = -EINTR; 4274 goto out; 4275 } 4276 cond_resched(); 4277 } 4278 4279 if (len) 4280 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 4281 prev_block, len, range.flags); 4282 f2fs_update_time(sbi, REQ_TIME); 4283 out: 4284 filemap_invalidate_unlock(mapping); 4285 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4286 err: 4287 inode_unlock(inode); 4288 mnt_drop_write_file(filp); 4289 4290 return ret; 4291 } 4292 4293 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 4294 { 4295 struct inode *inode = file_inode(filp); 4296 struct f2fs_comp_option option; 4297 4298 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4299 return -EOPNOTSUPP; 4300 4301 inode_lock_shared(inode); 4302 4303 if (!f2fs_compressed_file(inode)) { 4304 inode_unlock_shared(inode); 4305 return -ENODATA; 4306 } 4307 4308 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4309 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4310 4311 inode_unlock_shared(inode); 4312 4313 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4314 sizeof(option))) 4315 return -EFAULT; 4316 4317 return 0; 4318 } 4319 4320 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4321 { 4322 struct inode *inode = file_inode(filp); 4323 struct f2fs_inode_info *fi = F2FS_I(inode); 4324 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4325 struct f2fs_comp_option option; 4326 int ret = 0; 4327 4328 if (!f2fs_sb_has_compression(sbi)) 4329 return -EOPNOTSUPP; 4330 4331 if (!(filp->f_mode & FMODE_WRITE)) 4332 return -EBADF; 4333 4334 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4335 sizeof(option))) 4336 return -EFAULT; 4337 4338 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4339 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4340 option.algorithm >= COMPRESS_MAX) 4341 return -EINVAL; 4342 4343 ret = mnt_want_write_file(filp); 4344 if (ret) 4345 return ret; 4346 inode_lock(inode); 4347 4348 f2fs_down_write(&F2FS_I(inode)->i_sem); 4349 if (!f2fs_compressed_file(inode)) { 4350 ret = -EINVAL; 4351 goto out; 4352 } 4353 4354 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4355 ret = -EBUSY; 4356 goto out; 4357 } 4358 4359 if (F2FS_HAS_BLOCKS(inode)) { 4360 ret = -EFBIG; 4361 goto out; 4362 } 4363 4364 fi->i_compress_algorithm = option.algorithm; 4365 fi->i_log_cluster_size = option.log_cluster_size; 4366 fi->i_cluster_size = BIT(option.log_cluster_size); 4367 /* Set default level */ 4368 if (fi->i_compress_algorithm == COMPRESS_ZSTD) 4369 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4370 else 4371 fi->i_compress_level = 0; 4372 /* Adjust mount option level */ 4373 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4374 F2FS_OPTION(sbi).compress_level) 4375 fi->i_compress_level = F2FS_OPTION(sbi).compress_level; 4376 f2fs_mark_inode_dirty_sync(inode, true); 4377 4378 if (!f2fs_is_compress_backend_ready(inode)) 4379 f2fs_warn(sbi, "compression algorithm is successfully set, " 4380 "but current kernel doesn't support this algorithm."); 4381 out: 4382 f2fs_up_write(&fi->i_sem); 4383 inode_unlock(inode); 4384 mnt_drop_write_file(filp); 4385 4386 return ret; 4387 } 4388 4389 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4390 { 4391 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4392 struct address_space *mapping = inode->i_mapping; 4393 struct folio *folio; 4394 pgoff_t redirty_idx = page_idx; 4395 int page_len = 0, ret = 0; 4396 4397 page_cache_ra_unbounded(&ractl, len, 0); 4398 4399 do { 4400 folio = read_cache_folio(mapping, page_idx, NULL, NULL); 4401 if (IS_ERR(folio)) { 4402 ret = PTR_ERR(folio); 4403 break; 4404 } 4405 page_len += folio_nr_pages(folio) - (page_idx - folio->index); 4406 page_idx = folio_next_index(folio); 4407 } while (page_len < len); 4408 4409 do { 4410 folio = filemap_lock_folio(mapping, redirty_idx); 4411 4412 /* It will never fail, when folio has pinned above */ 4413 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio)); 4414 4415 f2fs_folio_wait_writeback(folio, DATA, true, true); 4416 4417 folio_mark_dirty(folio); 4418 set_page_private_gcing(&folio->page); 4419 redirty_idx = folio_next_index(folio); 4420 folio_unlock(folio); 4421 folio_put_refs(folio, 2); 4422 } while (redirty_idx < page_idx); 4423 4424 return ret; 4425 } 4426 4427 static int f2fs_ioc_decompress_file(struct file *filp) 4428 { 4429 struct inode *inode = file_inode(filp); 4430 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4431 struct f2fs_inode_info *fi = F2FS_I(inode); 4432 pgoff_t page_idx = 0, last_idx, cluster_idx; 4433 int ret; 4434 4435 if (!f2fs_sb_has_compression(sbi) || 4436 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4437 return -EOPNOTSUPP; 4438 4439 if (!(filp->f_mode & FMODE_WRITE)) 4440 return -EBADF; 4441 4442 f2fs_balance_fs(sbi, true); 4443 4444 ret = mnt_want_write_file(filp); 4445 if (ret) 4446 return ret; 4447 inode_lock(inode); 4448 4449 if (!f2fs_is_compress_backend_ready(inode)) { 4450 ret = -EOPNOTSUPP; 4451 goto out; 4452 } 4453 4454 if (!f2fs_compressed_file(inode) || 4455 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4456 ret = -EINVAL; 4457 goto out; 4458 } 4459 4460 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4461 if (ret) 4462 goto out; 4463 4464 if (!atomic_read(&fi->i_compr_blocks)) 4465 goto out; 4466 4467 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4468 last_idx >>= fi->i_log_cluster_size; 4469 4470 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4471 page_idx = cluster_idx << fi->i_log_cluster_size; 4472 4473 if (!f2fs_is_compressed_cluster(inode, page_idx)) 4474 continue; 4475 4476 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4477 if (ret < 0) 4478 break; 4479 4480 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4481 ret = filemap_fdatawrite(inode->i_mapping); 4482 if (ret < 0) 4483 break; 4484 } 4485 4486 cond_resched(); 4487 if (fatal_signal_pending(current)) { 4488 ret = -EINTR; 4489 break; 4490 } 4491 } 4492 4493 if (!ret) 4494 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4495 LLONG_MAX); 4496 4497 if (ret) 4498 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4499 __func__, ret); 4500 f2fs_update_time(sbi, REQ_TIME); 4501 out: 4502 inode_unlock(inode); 4503 mnt_drop_write_file(filp); 4504 4505 return ret; 4506 } 4507 4508 static int f2fs_ioc_compress_file(struct file *filp) 4509 { 4510 struct inode *inode = file_inode(filp); 4511 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4512 struct f2fs_inode_info *fi = F2FS_I(inode); 4513 pgoff_t page_idx = 0, last_idx, cluster_idx; 4514 int ret; 4515 4516 if (!f2fs_sb_has_compression(sbi) || 4517 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4518 return -EOPNOTSUPP; 4519 4520 if (!(filp->f_mode & FMODE_WRITE)) 4521 return -EBADF; 4522 4523 f2fs_balance_fs(sbi, true); 4524 4525 ret = mnt_want_write_file(filp); 4526 if (ret) 4527 return ret; 4528 inode_lock(inode); 4529 4530 if (!f2fs_is_compress_backend_ready(inode)) { 4531 ret = -EOPNOTSUPP; 4532 goto out; 4533 } 4534 4535 if (!f2fs_compressed_file(inode) || 4536 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4537 ret = -EINVAL; 4538 goto out; 4539 } 4540 4541 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4542 if (ret) 4543 goto out; 4544 4545 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4546 4547 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4548 last_idx >>= fi->i_log_cluster_size; 4549 4550 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4551 page_idx = cluster_idx << fi->i_log_cluster_size; 4552 4553 if (f2fs_is_sparse_cluster(inode, page_idx)) 4554 continue; 4555 4556 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4557 if (ret < 0) 4558 break; 4559 4560 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4561 ret = filemap_fdatawrite(inode->i_mapping); 4562 if (ret < 0) 4563 break; 4564 } 4565 4566 cond_resched(); 4567 if (fatal_signal_pending(current)) { 4568 ret = -EINTR; 4569 break; 4570 } 4571 } 4572 4573 if (!ret) 4574 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4575 LLONG_MAX); 4576 4577 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4578 4579 if (ret) 4580 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4581 __func__, ret); 4582 f2fs_update_time(sbi, REQ_TIME); 4583 out: 4584 inode_unlock(inode); 4585 mnt_drop_write_file(filp); 4586 4587 return ret; 4588 } 4589 4590 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4591 { 4592 switch (cmd) { 4593 case FS_IOC_GETVERSION: 4594 return f2fs_ioc_getversion(filp, arg); 4595 case F2FS_IOC_START_ATOMIC_WRITE: 4596 return f2fs_ioc_start_atomic_write(filp, false); 4597 case F2FS_IOC_START_ATOMIC_REPLACE: 4598 return f2fs_ioc_start_atomic_write(filp, true); 4599 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4600 return f2fs_ioc_commit_atomic_write(filp); 4601 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4602 return f2fs_ioc_abort_atomic_write(filp); 4603 case F2FS_IOC_START_VOLATILE_WRITE: 4604 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4605 return -EOPNOTSUPP; 4606 case F2FS_IOC_SHUTDOWN: 4607 return f2fs_ioc_shutdown(filp, arg); 4608 case FITRIM: 4609 return f2fs_ioc_fitrim(filp, arg); 4610 case FS_IOC_SET_ENCRYPTION_POLICY: 4611 return f2fs_ioc_set_encryption_policy(filp, arg); 4612 case FS_IOC_GET_ENCRYPTION_POLICY: 4613 return f2fs_ioc_get_encryption_policy(filp, arg); 4614 case FS_IOC_GET_ENCRYPTION_PWSALT: 4615 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4616 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4617 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4618 case FS_IOC_ADD_ENCRYPTION_KEY: 4619 return f2fs_ioc_add_encryption_key(filp, arg); 4620 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4621 return f2fs_ioc_remove_encryption_key(filp, arg); 4622 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4623 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4624 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4625 return f2fs_ioc_get_encryption_key_status(filp, arg); 4626 case FS_IOC_GET_ENCRYPTION_NONCE: 4627 return f2fs_ioc_get_encryption_nonce(filp, arg); 4628 case F2FS_IOC_GARBAGE_COLLECT: 4629 return f2fs_ioc_gc(filp, arg); 4630 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4631 return f2fs_ioc_gc_range(filp, arg); 4632 case F2FS_IOC_WRITE_CHECKPOINT: 4633 return f2fs_ioc_write_checkpoint(filp); 4634 case F2FS_IOC_DEFRAGMENT: 4635 return f2fs_ioc_defragment(filp, arg); 4636 case F2FS_IOC_MOVE_RANGE: 4637 return f2fs_ioc_move_range(filp, arg); 4638 case F2FS_IOC_FLUSH_DEVICE: 4639 return f2fs_ioc_flush_device(filp, arg); 4640 case F2FS_IOC_GET_FEATURES: 4641 return f2fs_ioc_get_features(filp, arg); 4642 case F2FS_IOC_GET_PIN_FILE: 4643 return f2fs_ioc_get_pin_file(filp, arg); 4644 case F2FS_IOC_SET_PIN_FILE: 4645 return f2fs_ioc_set_pin_file(filp, arg); 4646 case F2FS_IOC_PRECACHE_EXTENTS: 4647 return f2fs_ioc_precache_extents(filp); 4648 case F2FS_IOC_RESIZE_FS: 4649 return f2fs_ioc_resize_fs(filp, arg); 4650 case FS_IOC_ENABLE_VERITY: 4651 return f2fs_ioc_enable_verity(filp, arg); 4652 case FS_IOC_MEASURE_VERITY: 4653 return f2fs_ioc_measure_verity(filp, arg); 4654 case FS_IOC_READ_VERITY_METADATA: 4655 return f2fs_ioc_read_verity_metadata(filp, arg); 4656 case FS_IOC_GETFSLABEL: 4657 return f2fs_ioc_getfslabel(filp, arg); 4658 case FS_IOC_SETFSLABEL: 4659 return f2fs_ioc_setfslabel(filp, arg); 4660 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4661 return f2fs_ioc_get_compress_blocks(filp, arg); 4662 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4663 return f2fs_release_compress_blocks(filp, arg); 4664 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4665 return f2fs_reserve_compress_blocks(filp, arg); 4666 case F2FS_IOC_SEC_TRIM_FILE: 4667 return f2fs_sec_trim_file(filp, arg); 4668 case F2FS_IOC_GET_COMPRESS_OPTION: 4669 return f2fs_ioc_get_compress_option(filp, arg); 4670 case F2FS_IOC_SET_COMPRESS_OPTION: 4671 return f2fs_ioc_set_compress_option(filp, arg); 4672 case F2FS_IOC_DECOMPRESS_FILE: 4673 return f2fs_ioc_decompress_file(filp); 4674 case F2FS_IOC_COMPRESS_FILE: 4675 return f2fs_ioc_compress_file(filp); 4676 case F2FS_IOC_GET_DEV_ALIAS_FILE: 4677 return f2fs_ioc_get_dev_alias_file(filp, arg); 4678 case F2FS_IOC_IO_PRIO: 4679 return f2fs_ioc_io_prio(filp, arg); 4680 default: 4681 return -ENOTTY; 4682 } 4683 } 4684 4685 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4686 { 4687 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4688 return -EIO; 4689 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4690 return -ENOSPC; 4691 4692 return __f2fs_ioctl(filp, cmd, arg); 4693 } 4694 4695 /* 4696 * Return %true if the given read or write request should use direct I/O, or 4697 * %false if it should use buffered I/O. 4698 */ 4699 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4700 struct iov_iter *iter) 4701 { 4702 unsigned int align; 4703 4704 if (!(iocb->ki_flags & IOCB_DIRECT)) 4705 return false; 4706 4707 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4708 return false; 4709 4710 /* 4711 * Direct I/O not aligned to the disk's logical_block_size will be 4712 * attempted, but will fail with -EINVAL. 4713 * 4714 * f2fs additionally requires that direct I/O be aligned to the 4715 * filesystem block size, which is often a stricter requirement. 4716 * However, f2fs traditionally falls back to buffered I/O on requests 4717 * that are logical_block_size-aligned but not fs-block aligned. 4718 * 4719 * The below logic implements this behavior. 4720 */ 4721 align = iocb->ki_pos | iov_iter_alignment(iter); 4722 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4723 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4724 return false; 4725 4726 return true; 4727 } 4728 4729 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4730 unsigned int flags) 4731 { 4732 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4733 4734 dec_page_count(sbi, F2FS_DIO_READ); 4735 if (error) 4736 return error; 4737 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4738 return 0; 4739 } 4740 4741 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4742 .end_io = f2fs_dio_read_end_io, 4743 }; 4744 4745 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4746 { 4747 struct file *file = iocb->ki_filp; 4748 struct inode *inode = file_inode(file); 4749 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4750 struct f2fs_inode_info *fi = F2FS_I(inode); 4751 const loff_t pos = iocb->ki_pos; 4752 const size_t count = iov_iter_count(to); 4753 struct iomap_dio *dio; 4754 ssize_t ret; 4755 4756 if (count == 0) 4757 return 0; /* skip atime update */ 4758 4759 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4760 4761 if (iocb->ki_flags & IOCB_NOWAIT) { 4762 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4763 ret = -EAGAIN; 4764 goto out; 4765 } 4766 } else { 4767 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4768 } 4769 4770 /* dio is not compatible w/ atomic file */ 4771 if (f2fs_is_atomic_file(inode)) { 4772 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4773 ret = -EOPNOTSUPP; 4774 goto out; 4775 } 4776 4777 /* 4778 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4779 * the higher-level function iomap_dio_rw() in order to ensure that the 4780 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4781 */ 4782 inc_page_count(sbi, F2FS_DIO_READ); 4783 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4784 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4785 if (IS_ERR_OR_NULL(dio)) { 4786 ret = PTR_ERR_OR_ZERO(dio); 4787 if (ret != -EIOCBQUEUED) 4788 dec_page_count(sbi, F2FS_DIO_READ); 4789 } else { 4790 ret = iomap_dio_complete(dio); 4791 } 4792 4793 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4794 4795 file_accessed(file); 4796 out: 4797 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4798 return ret; 4799 } 4800 4801 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4802 int rw) 4803 { 4804 struct inode *inode = file_inode(file); 4805 char *buf, *path; 4806 4807 buf = f2fs_getname(F2FS_I_SB(inode)); 4808 if (!buf) 4809 return; 4810 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4811 if (IS_ERR(path)) 4812 goto free_buf; 4813 if (rw == WRITE) 4814 trace_f2fs_datawrite_start(inode, pos, count, 4815 current->pid, path, current->comm); 4816 else 4817 trace_f2fs_dataread_start(inode, pos, count, 4818 current->pid, path, current->comm); 4819 free_buf: 4820 f2fs_putname(buf); 4821 } 4822 4823 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4824 { 4825 struct inode *inode = file_inode(iocb->ki_filp); 4826 const loff_t pos = iocb->ki_pos; 4827 ssize_t ret; 4828 4829 if (!f2fs_is_compress_backend_ready(inode)) 4830 return -EOPNOTSUPP; 4831 4832 if (trace_f2fs_dataread_start_enabled()) 4833 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4834 iov_iter_count(to), READ); 4835 4836 /* In LFS mode, if there is inflight dio, wait for its completion */ 4837 if (f2fs_lfs_mode(F2FS_I_SB(inode)) && 4838 get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE)) 4839 inode_dio_wait(inode); 4840 4841 if (f2fs_should_use_dio(inode, iocb, to)) { 4842 ret = f2fs_dio_read_iter(iocb, to); 4843 } else { 4844 ret = filemap_read(iocb, to, 0); 4845 if (ret > 0) 4846 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4847 APP_BUFFERED_READ_IO, ret); 4848 } 4849 if (trace_f2fs_dataread_end_enabled()) 4850 trace_f2fs_dataread_end(inode, pos, ret); 4851 return ret; 4852 } 4853 4854 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4855 struct pipe_inode_info *pipe, 4856 size_t len, unsigned int flags) 4857 { 4858 struct inode *inode = file_inode(in); 4859 const loff_t pos = *ppos; 4860 ssize_t ret; 4861 4862 if (!f2fs_is_compress_backend_ready(inode)) 4863 return -EOPNOTSUPP; 4864 4865 if (trace_f2fs_dataread_start_enabled()) 4866 f2fs_trace_rw_file_path(in, pos, len, READ); 4867 4868 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4869 if (ret > 0) 4870 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4871 APP_BUFFERED_READ_IO, ret); 4872 4873 if (trace_f2fs_dataread_end_enabled()) 4874 trace_f2fs_dataread_end(inode, pos, ret); 4875 return ret; 4876 } 4877 4878 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4879 { 4880 struct file *file = iocb->ki_filp; 4881 struct inode *inode = file_inode(file); 4882 ssize_t count; 4883 int err; 4884 4885 if (IS_IMMUTABLE(inode)) 4886 return -EPERM; 4887 4888 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4889 return -EPERM; 4890 4891 count = generic_write_checks(iocb, from); 4892 if (count <= 0) 4893 return count; 4894 4895 err = file_modified(file); 4896 if (err) 4897 return err; 4898 4899 filemap_invalidate_lock(inode->i_mapping); 4900 f2fs_zero_post_eof_page(inode, iocb->ki_pos + iov_iter_count(from)); 4901 filemap_invalidate_unlock(inode->i_mapping); 4902 return count; 4903 } 4904 4905 /* 4906 * Preallocate blocks for a write request, if it is possible and helpful to do 4907 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4908 * blocks were preallocated, or a negative errno value if something went 4909 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4910 * requested blocks (not just some of them) have been allocated. 4911 */ 4912 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4913 bool dio) 4914 { 4915 struct inode *inode = file_inode(iocb->ki_filp); 4916 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4917 const loff_t pos = iocb->ki_pos; 4918 const size_t count = iov_iter_count(iter); 4919 struct f2fs_map_blocks map = {}; 4920 int flag; 4921 int ret; 4922 4923 /* If it will be an out-of-place direct write, don't bother. */ 4924 if (dio && f2fs_lfs_mode(sbi)) 4925 return 0; 4926 /* 4927 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4928 * buffered IO, if DIO meets any holes. 4929 */ 4930 if (dio && i_size_read(inode) && 4931 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4932 return 0; 4933 4934 /* No-wait I/O can't allocate blocks. */ 4935 if (iocb->ki_flags & IOCB_NOWAIT) 4936 return 0; 4937 4938 /* If it will be a short write, don't bother. */ 4939 if (fault_in_iov_iter_readable(iter, count)) 4940 return 0; 4941 4942 if (f2fs_has_inline_data(inode)) { 4943 /* If the data will fit inline, don't bother. */ 4944 if (pos + count <= MAX_INLINE_DATA(inode)) 4945 return 0; 4946 ret = f2fs_convert_inline_inode(inode); 4947 if (ret) 4948 return ret; 4949 } 4950 4951 /* Do not preallocate blocks that will be written partially in 4KB. */ 4952 map.m_lblk = F2FS_BLK_ALIGN(pos); 4953 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4954 if (map.m_len > map.m_lblk) 4955 map.m_len -= map.m_lblk; 4956 else 4957 return 0; 4958 4959 if (!IS_DEVICE_ALIASING(inode)) 4960 map.m_may_create = true; 4961 if (dio) { 4962 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi, 4963 inode->i_write_hint); 4964 flag = F2FS_GET_BLOCK_PRE_DIO; 4965 } else { 4966 map.m_seg_type = NO_CHECK_TYPE; 4967 flag = F2FS_GET_BLOCK_PRE_AIO; 4968 } 4969 4970 ret = f2fs_map_blocks(inode, &map, flag); 4971 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4972 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4973 return ret; 4974 if (ret == 0) 4975 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4976 return map.m_len; 4977 } 4978 4979 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4980 struct iov_iter *from) 4981 { 4982 struct file *file = iocb->ki_filp; 4983 struct inode *inode = file_inode(file); 4984 ssize_t ret; 4985 4986 if (iocb->ki_flags & IOCB_NOWAIT) 4987 return -EOPNOTSUPP; 4988 4989 ret = generic_perform_write(iocb, from); 4990 4991 if (ret > 0) { 4992 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4993 APP_BUFFERED_IO, ret); 4994 } 4995 return ret; 4996 } 4997 4998 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4999 unsigned int flags) 5000 { 5001 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 5002 5003 dec_page_count(sbi, F2FS_DIO_WRITE); 5004 if (error) 5005 return error; 5006 f2fs_update_time(sbi, REQ_TIME); 5007 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 5008 return 0; 5009 } 5010 5011 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter, 5012 struct bio *bio, loff_t file_offset) 5013 { 5014 struct inode *inode = iter->inode; 5015 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 5016 enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint); 5017 enum temp_type temp = f2fs_get_segment_temp(sbi, type); 5018 5019 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp); 5020 submit_bio(bio); 5021 } 5022 5023 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 5024 .end_io = f2fs_dio_write_end_io, 5025 .submit_io = f2fs_dio_write_submit_io, 5026 }; 5027 5028 static void f2fs_flush_buffered_write(struct address_space *mapping, 5029 loff_t start_pos, loff_t end_pos) 5030 { 5031 int ret; 5032 5033 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 5034 if (ret < 0) 5035 return; 5036 invalidate_mapping_pages(mapping, 5037 start_pos >> PAGE_SHIFT, 5038 end_pos >> PAGE_SHIFT); 5039 } 5040 5041 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 5042 bool *may_need_sync) 5043 { 5044 struct file *file = iocb->ki_filp; 5045 struct inode *inode = file_inode(file); 5046 struct f2fs_inode_info *fi = F2FS_I(inode); 5047 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 5048 const bool do_opu = f2fs_lfs_mode(sbi); 5049 const loff_t pos = iocb->ki_pos; 5050 const ssize_t count = iov_iter_count(from); 5051 unsigned int dio_flags; 5052 struct iomap_dio *dio; 5053 ssize_t ret; 5054 5055 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 5056 5057 if (iocb->ki_flags & IOCB_NOWAIT) { 5058 /* f2fs_convert_inline_inode() and block allocation can block */ 5059 if (f2fs_has_inline_data(inode) || 5060 !f2fs_overwrite_io(inode, pos, count)) { 5061 ret = -EAGAIN; 5062 goto out; 5063 } 5064 5065 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 5066 ret = -EAGAIN; 5067 goto out; 5068 } 5069 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 5070 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5071 ret = -EAGAIN; 5072 goto out; 5073 } 5074 } else { 5075 ret = f2fs_convert_inline_inode(inode); 5076 if (ret) 5077 goto out; 5078 5079 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 5080 if (do_opu) 5081 f2fs_down_read(&fi->i_gc_rwsem[READ]); 5082 } 5083 5084 /* 5085 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 5086 * the higher-level function iomap_dio_rw() in order to ensure that the 5087 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 5088 */ 5089 inc_page_count(sbi, F2FS_DIO_WRITE); 5090 dio_flags = 0; 5091 if (pos + count > inode->i_size) 5092 dio_flags |= IOMAP_DIO_FORCE_WAIT; 5093 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 5094 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 5095 if (IS_ERR_OR_NULL(dio)) { 5096 ret = PTR_ERR_OR_ZERO(dio); 5097 if (ret == -ENOTBLK) 5098 ret = 0; 5099 if (ret != -EIOCBQUEUED) 5100 dec_page_count(sbi, F2FS_DIO_WRITE); 5101 } else { 5102 ret = iomap_dio_complete(dio); 5103 } 5104 5105 if (do_opu) 5106 f2fs_up_read(&fi->i_gc_rwsem[READ]); 5107 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5108 5109 if (ret < 0) 5110 goto out; 5111 if (pos + ret > inode->i_size) 5112 f2fs_i_size_write(inode, pos + ret); 5113 if (!do_opu) 5114 set_inode_flag(inode, FI_UPDATE_WRITE); 5115 5116 if (iov_iter_count(from)) { 5117 ssize_t ret2; 5118 loff_t bufio_start_pos = iocb->ki_pos; 5119 5120 /* 5121 * The direct write was partial, so we need to fall back to a 5122 * buffered write for the remainder. 5123 */ 5124 5125 ret2 = f2fs_buffered_write_iter(iocb, from); 5126 if (iov_iter_count(from)) 5127 f2fs_write_failed(inode, iocb->ki_pos); 5128 if (ret2 < 0) 5129 goto out; 5130 5131 /* 5132 * Ensure that the pagecache pages are written to disk and 5133 * invalidated to preserve the expected O_DIRECT semantics. 5134 */ 5135 if (ret2 > 0) { 5136 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 5137 5138 ret += ret2; 5139 5140 f2fs_flush_buffered_write(file->f_mapping, 5141 bufio_start_pos, 5142 bufio_end_pos); 5143 } 5144 } else { 5145 /* iomap_dio_rw() already handled the generic_write_sync(). */ 5146 *may_need_sync = false; 5147 } 5148 out: 5149 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 5150 return ret; 5151 } 5152 5153 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 5154 { 5155 struct inode *inode = file_inode(iocb->ki_filp); 5156 const loff_t orig_pos = iocb->ki_pos; 5157 const size_t orig_count = iov_iter_count(from); 5158 loff_t target_size; 5159 bool dio; 5160 bool may_need_sync = true; 5161 int preallocated; 5162 const loff_t pos = iocb->ki_pos; 5163 const ssize_t count = iov_iter_count(from); 5164 ssize_t ret; 5165 5166 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 5167 ret = -EIO; 5168 goto out; 5169 } 5170 5171 if (!f2fs_is_compress_backend_ready(inode)) { 5172 ret = -EOPNOTSUPP; 5173 goto out; 5174 } 5175 5176 if (iocb->ki_flags & IOCB_NOWAIT) { 5177 if (!inode_trylock(inode)) { 5178 ret = -EAGAIN; 5179 goto out; 5180 } 5181 } else { 5182 inode_lock(inode); 5183 } 5184 5185 if (f2fs_is_pinned_file(inode) && 5186 !f2fs_overwrite_io(inode, pos, count)) { 5187 ret = -EIO; 5188 goto out_unlock; 5189 } 5190 5191 ret = f2fs_write_checks(iocb, from); 5192 if (ret <= 0) 5193 goto out_unlock; 5194 5195 /* Determine whether we will do a direct write or a buffered write. */ 5196 dio = f2fs_should_use_dio(inode, iocb, from); 5197 5198 /* dio is not compatible w/ atomic write */ 5199 if (dio && f2fs_is_atomic_file(inode)) { 5200 ret = -EOPNOTSUPP; 5201 goto out_unlock; 5202 } 5203 5204 /* Possibly preallocate the blocks for the write. */ 5205 target_size = iocb->ki_pos + iov_iter_count(from); 5206 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 5207 if (preallocated < 0) { 5208 ret = preallocated; 5209 } else { 5210 if (trace_f2fs_datawrite_start_enabled()) 5211 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 5212 orig_count, WRITE); 5213 5214 /* Do the actual write. */ 5215 ret = dio ? 5216 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 5217 f2fs_buffered_write_iter(iocb, from); 5218 5219 if (trace_f2fs_datawrite_end_enabled()) 5220 trace_f2fs_datawrite_end(inode, orig_pos, ret); 5221 } 5222 5223 /* Don't leave any preallocated blocks around past i_size. */ 5224 if (preallocated && i_size_read(inode) < target_size) { 5225 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5226 filemap_invalidate_lock(inode->i_mapping); 5227 if (!f2fs_truncate(inode)) 5228 file_dont_truncate(inode); 5229 filemap_invalidate_unlock(inode->i_mapping); 5230 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5231 } else { 5232 file_dont_truncate(inode); 5233 } 5234 5235 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 5236 out_unlock: 5237 inode_unlock(inode); 5238 out: 5239 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 5240 5241 if (ret > 0 && may_need_sync) 5242 ret = generic_write_sync(iocb, ret); 5243 5244 /* If buffered IO was forced, flush and drop the data from 5245 * the page cache to preserve O_DIRECT semantics 5246 */ 5247 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 5248 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 5249 orig_pos, 5250 orig_pos + ret - 1); 5251 5252 return ret; 5253 } 5254 5255 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 5256 int advice) 5257 { 5258 struct address_space *mapping; 5259 struct backing_dev_info *bdi; 5260 struct inode *inode = file_inode(filp); 5261 int err; 5262 5263 if (advice == POSIX_FADV_SEQUENTIAL) { 5264 if (S_ISFIFO(inode->i_mode)) 5265 return -ESPIPE; 5266 5267 mapping = filp->f_mapping; 5268 if (!mapping || len < 0) 5269 return -EINVAL; 5270 5271 bdi = inode_to_bdi(mapping->host); 5272 filp->f_ra.ra_pages = bdi->ra_pages * 5273 F2FS_I_SB(inode)->seq_file_ra_mul; 5274 spin_lock(&filp->f_lock); 5275 filp->f_mode &= ~FMODE_RANDOM; 5276 spin_unlock(&filp->f_lock); 5277 return 0; 5278 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 5279 /* Load extent cache at the first readahead. */ 5280 f2fs_precache_extents(inode); 5281 } 5282 5283 err = generic_fadvise(filp, offset, len, advice); 5284 if (err) 5285 return err; 5286 5287 if (advice == POSIX_FADV_DONTNEED && 5288 (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 5289 f2fs_compressed_file(inode))) 5290 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 5291 else if (advice == POSIX_FADV_NOREUSE) 5292 err = f2fs_keep_noreuse_range(inode, offset, len); 5293 return err; 5294 } 5295 5296 #ifdef CONFIG_COMPAT 5297 struct compat_f2fs_gc_range { 5298 u32 sync; 5299 compat_u64 start; 5300 compat_u64 len; 5301 }; 5302 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 5303 struct compat_f2fs_gc_range) 5304 5305 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 5306 { 5307 struct compat_f2fs_gc_range __user *urange; 5308 struct f2fs_gc_range range; 5309 int err; 5310 5311 urange = compat_ptr(arg); 5312 err = get_user(range.sync, &urange->sync); 5313 err |= get_user(range.start, &urange->start); 5314 err |= get_user(range.len, &urange->len); 5315 if (err) 5316 return -EFAULT; 5317 5318 return __f2fs_ioc_gc_range(file, &range); 5319 } 5320 5321 struct compat_f2fs_move_range { 5322 u32 dst_fd; 5323 compat_u64 pos_in; 5324 compat_u64 pos_out; 5325 compat_u64 len; 5326 }; 5327 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 5328 struct compat_f2fs_move_range) 5329 5330 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 5331 { 5332 struct compat_f2fs_move_range __user *urange; 5333 struct f2fs_move_range range; 5334 int err; 5335 5336 urange = compat_ptr(arg); 5337 err = get_user(range.dst_fd, &urange->dst_fd); 5338 err |= get_user(range.pos_in, &urange->pos_in); 5339 err |= get_user(range.pos_out, &urange->pos_out); 5340 err |= get_user(range.len, &urange->len); 5341 if (err) 5342 return -EFAULT; 5343 5344 return __f2fs_ioc_move_range(file, &range); 5345 } 5346 5347 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5348 { 5349 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 5350 return -EIO; 5351 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 5352 return -ENOSPC; 5353 5354 switch (cmd) { 5355 case FS_IOC32_GETVERSION: 5356 cmd = FS_IOC_GETVERSION; 5357 break; 5358 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 5359 return f2fs_compat_ioc_gc_range(file, arg); 5360 case F2FS_IOC32_MOVE_RANGE: 5361 return f2fs_compat_ioc_move_range(file, arg); 5362 case F2FS_IOC_START_ATOMIC_WRITE: 5363 case F2FS_IOC_START_ATOMIC_REPLACE: 5364 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5365 case F2FS_IOC_START_VOLATILE_WRITE: 5366 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5367 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5368 case F2FS_IOC_SHUTDOWN: 5369 case FITRIM: 5370 case FS_IOC_SET_ENCRYPTION_POLICY: 5371 case FS_IOC_GET_ENCRYPTION_PWSALT: 5372 case FS_IOC_GET_ENCRYPTION_POLICY: 5373 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5374 case FS_IOC_ADD_ENCRYPTION_KEY: 5375 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5376 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5377 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5378 case FS_IOC_GET_ENCRYPTION_NONCE: 5379 case F2FS_IOC_GARBAGE_COLLECT: 5380 case F2FS_IOC_WRITE_CHECKPOINT: 5381 case F2FS_IOC_DEFRAGMENT: 5382 case F2FS_IOC_FLUSH_DEVICE: 5383 case F2FS_IOC_GET_FEATURES: 5384 case F2FS_IOC_GET_PIN_FILE: 5385 case F2FS_IOC_SET_PIN_FILE: 5386 case F2FS_IOC_PRECACHE_EXTENTS: 5387 case F2FS_IOC_RESIZE_FS: 5388 case FS_IOC_ENABLE_VERITY: 5389 case FS_IOC_MEASURE_VERITY: 5390 case FS_IOC_READ_VERITY_METADATA: 5391 case FS_IOC_GETFSLABEL: 5392 case FS_IOC_SETFSLABEL: 5393 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5394 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5395 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5396 case F2FS_IOC_SEC_TRIM_FILE: 5397 case F2FS_IOC_GET_COMPRESS_OPTION: 5398 case F2FS_IOC_SET_COMPRESS_OPTION: 5399 case F2FS_IOC_DECOMPRESS_FILE: 5400 case F2FS_IOC_COMPRESS_FILE: 5401 case F2FS_IOC_GET_DEV_ALIAS_FILE: 5402 case F2FS_IOC_IO_PRIO: 5403 break; 5404 default: 5405 return -ENOIOCTLCMD; 5406 } 5407 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5408 } 5409 #endif 5410 5411 const struct file_operations f2fs_file_operations = { 5412 .llseek = f2fs_llseek, 5413 .read_iter = f2fs_file_read_iter, 5414 .write_iter = f2fs_file_write_iter, 5415 .iopoll = iocb_bio_iopoll, 5416 .open = f2fs_file_open, 5417 .release = f2fs_release_file, 5418 .mmap_prepare = f2fs_file_mmap_prepare, 5419 .flush = f2fs_file_flush, 5420 .fsync = f2fs_sync_file, 5421 .fallocate = f2fs_fallocate, 5422 .unlocked_ioctl = f2fs_ioctl, 5423 #ifdef CONFIG_COMPAT 5424 .compat_ioctl = f2fs_compat_ioctl, 5425 #endif 5426 .splice_read = f2fs_file_splice_read, 5427 .splice_write = iter_file_splice_write, 5428 .fadvise = f2fs_file_fadvise, 5429 .fop_flags = FOP_BUFFER_RASYNC, 5430 }; 5431