1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/file.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/fs.h> 9 #include <linux/f2fs_fs.h> 10 #include <linux/stat.h> 11 #include <linux/writeback.h> 12 #include <linux/blkdev.h> 13 #include <linux/falloc.h> 14 #include <linux/types.h> 15 #include <linux/compat.h> 16 #include <linux/uaccess.h> 17 #include <linux/mount.h> 18 #include <linux/pagevec.h> 19 #include <linux/uio.h> 20 #include <linux/uuid.h> 21 #include <linux/file.h> 22 #include <linux/nls.h> 23 #include <linux/sched/signal.h> 24 #include <linux/fileattr.h> 25 #include <linux/fadvise.h> 26 #include <linux/iomap.h> 27 28 #include "f2fs.h" 29 #include "node.h" 30 #include "segment.h" 31 #include "xattr.h" 32 #include "acl.h" 33 #include "gc.h" 34 #include "iostat.h" 35 #include <trace/events/f2fs.h> 36 #include <uapi/linux/f2fs.h> 37 38 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 39 { 40 struct inode *inode = file_inode(vmf->vma->vm_file); 41 vm_flags_t flags = vmf->vma->vm_flags; 42 vm_fault_t ret; 43 44 ret = filemap_fault(vmf); 45 if (ret & VM_FAULT_LOCKED) 46 f2fs_update_iostat(F2FS_I_SB(inode), inode, 47 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 48 49 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret); 50 51 return ret; 52 } 53 54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 55 { 56 struct folio *folio = page_folio(vmf->page); 57 struct inode *inode = file_inode(vmf->vma->vm_file); 58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 59 struct dnode_of_data dn; 60 bool need_alloc = !f2fs_is_pinned_file(inode); 61 int err = 0; 62 vm_fault_t ret; 63 64 if (unlikely(IS_IMMUTABLE(inode))) 65 return VM_FAULT_SIGBUS; 66 67 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 68 err = -EIO; 69 goto out; 70 } 71 72 if (unlikely(f2fs_cp_error(sbi))) { 73 err = -EIO; 74 goto out; 75 } 76 77 if (!f2fs_is_checkpoint_ready(sbi)) { 78 err = -ENOSPC; 79 goto out; 80 } 81 82 err = f2fs_convert_inline_inode(inode); 83 if (err) 84 goto out; 85 86 #ifdef CONFIG_F2FS_FS_COMPRESSION 87 if (f2fs_compressed_file(inode)) { 88 int ret = f2fs_is_compressed_cluster(inode, folio->index); 89 90 if (ret < 0) { 91 err = ret; 92 goto out; 93 } else if (ret) { 94 need_alloc = false; 95 } 96 } 97 #endif 98 /* should do out of any locked page */ 99 if (need_alloc) 100 f2fs_balance_fs(sbi, true); 101 102 sb_start_pagefault(inode->i_sb); 103 104 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 105 106 file_update_time(vmf->vma->vm_file); 107 filemap_invalidate_lock_shared(inode->i_mapping); 108 folio_lock(folio); 109 if (unlikely(folio->mapping != inode->i_mapping || 110 folio_pos(folio) > i_size_read(inode) || 111 !folio_test_uptodate(folio))) { 112 folio_unlock(folio); 113 err = -EFAULT; 114 goto out_sem; 115 } 116 117 set_new_dnode(&dn, inode, NULL, NULL, 0); 118 if (need_alloc) { 119 /* block allocation */ 120 err = f2fs_get_block_locked(&dn, folio->index); 121 } else { 122 err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE); 123 f2fs_put_dnode(&dn); 124 if (f2fs_is_pinned_file(inode) && 125 !__is_valid_data_blkaddr(dn.data_blkaddr)) 126 err = -EIO; 127 } 128 129 if (err) { 130 folio_unlock(folio); 131 goto out_sem; 132 } 133 134 f2fs_wait_on_page_writeback(folio_page(folio, 0), DATA, false, true); 135 136 /* wait for GCed page writeback via META_MAPPING */ 137 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 138 139 /* 140 * check to see if the page is mapped already (no holes) 141 */ 142 if (folio_test_mappedtodisk(folio)) 143 goto out_sem; 144 145 /* page is wholly or partially inside EOF */ 146 if (((loff_t)(folio->index + 1) << PAGE_SHIFT) > 147 i_size_read(inode)) { 148 loff_t offset; 149 150 offset = i_size_read(inode) & ~PAGE_MASK; 151 folio_zero_segment(folio, offset, folio_size(folio)); 152 } 153 folio_mark_dirty(folio); 154 155 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 156 f2fs_update_time(sbi, REQ_TIME); 157 158 out_sem: 159 filemap_invalidate_unlock_shared(inode->i_mapping); 160 161 sb_end_pagefault(inode->i_sb); 162 out: 163 ret = vmf_fs_error(err); 164 165 trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret); 166 return ret; 167 } 168 169 static const struct vm_operations_struct f2fs_file_vm_ops = { 170 .fault = f2fs_filemap_fault, 171 .map_pages = filemap_map_pages, 172 .page_mkwrite = f2fs_vm_page_mkwrite, 173 }; 174 175 static int get_parent_ino(struct inode *inode, nid_t *pino) 176 { 177 struct dentry *dentry; 178 179 /* 180 * Make sure to get the non-deleted alias. The alias associated with 181 * the open file descriptor being fsync()'ed may be deleted already. 182 */ 183 dentry = d_find_alias(inode); 184 if (!dentry) 185 return 0; 186 187 *pino = d_parent_ino(dentry); 188 dput(dentry); 189 return 1; 190 } 191 192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 193 { 194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 195 enum cp_reason_type cp_reason = CP_NO_NEEDED; 196 197 if (!S_ISREG(inode->i_mode)) 198 cp_reason = CP_NON_REGULAR; 199 else if (f2fs_compressed_file(inode)) 200 cp_reason = CP_COMPRESSED; 201 else if (inode->i_nlink != 1) 202 cp_reason = CP_HARDLINK; 203 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 204 cp_reason = CP_SB_NEED_CP; 205 else if (file_wrong_pino(inode)) 206 cp_reason = CP_WRONG_PINO; 207 else if (!f2fs_space_for_roll_forward(sbi)) 208 cp_reason = CP_NO_SPC_ROLL; 209 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 210 cp_reason = CP_NODE_NEED_CP; 211 else if (test_opt(sbi, FASTBOOT)) 212 cp_reason = CP_FASTBOOT_MODE; 213 else if (F2FS_OPTION(sbi).active_logs == 2) 214 cp_reason = CP_SPEC_LOG_NUM; 215 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 216 f2fs_need_dentry_mark(sbi, inode->i_ino) && 217 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 218 TRANS_DIR_INO)) 219 cp_reason = CP_RECOVER_DIR; 220 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 221 XATTR_DIR_INO)) 222 cp_reason = CP_XATTR_DIR; 223 224 return cp_reason; 225 } 226 227 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 228 { 229 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 230 bool ret = false; 231 /* But we need to avoid that there are some inode updates */ 232 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 233 ret = true; 234 f2fs_put_page(i, 0); 235 return ret; 236 } 237 238 static void try_to_fix_pino(struct inode *inode) 239 { 240 struct f2fs_inode_info *fi = F2FS_I(inode); 241 nid_t pino; 242 243 f2fs_down_write(&fi->i_sem); 244 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 245 get_parent_ino(inode, &pino)) { 246 f2fs_i_pino_write(inode, pino); 247 file_got_pino(inode); 248 } 249 f2fs_up_write(&fi->i_sem); 250 } 251 252 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 253 int datasync, bool atomic) 254 { 255 struct inode *inode = file->f_mapping->host; 256 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 257 nid_t ino = inode->i_ino; 258 int ret = 0; 259 enum cp_reason_type cp_reason = 0; 260 struct writeback_control wbc = { 261 .sync_mode = WB_SYNC_ALL, 262 .nr_to_write = LONG_MAX, 263 .for_reclaim = 0, 264 }; 265 unsigned int seq_id = 0; 266 267 if (unlikely(f2fs_readonly(inode->i_sb))) 268 return 0; 269 270 trace_f2fs_sync_file_enter(inode); 271 272 if (S_ISDIR(inode->i_mode)) 273 goto go_write; 274 275 /* if fdatasync is triggered, let's do in-place-update */ 276 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 277 set_inode_flag(inode, FI_NEED_IPU); 278 ret = file_write_and_wait_range(file, start, end); 279 clear_inode_flag(inode, FI_NEED_IPU); 280 281 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 282 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 283 return ret; 284 } 285 286 /* if the inode is dirty, let's recover all the time */ 287 if (!f2fs_skip_inode_update(inode, datasync)) { 288 f2fs_write_inode(inode, NULL); 289 goto go_write; 290 } 291 292 /* 293 * if there is no written data, don't waste time to write recovery info. 294 */ 295 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 296 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 297 298 /* it may call write_inode just prior to fsync */ 299 if (need_inode_page_update(sbi, ino)) 300 goto go_write; 301 302 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 303 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 304 goto flush_out; 305 goto out; 306 } else { 307 /* 308 * for OPU case, during fsync(), node can be persisted before 309 * data when lower device doesn't support write barrier, result 310 * in data corruption after SPO. 311 * So for strict fsync mode, force to use atomic write semantics 312 * to keep write order in between data/node and last node to 313 * avoid potential data corruption. 314 */ 315 if (F2FS_OPTION(sbi).fsync_mode == 316 FSYNC_MODE_STRICT && !atomic) 317 atomic = true; 318 } 319 go_write: 320 /* 321 * Both of fdatasync() and fsync() are able to be recovered from 322 * sudden-power-off. 323 */ 324 f2fs_down_read(&F2FS_I(inode)->i_sem); 325 cp_reason = need_do_checkpoint(inode); 326 f2fs_up_read(&F2FS_I(inode)->i_sem); 327 328 if (cp_reason) { 329 /* all the dirty node pages should be flushed for POR */ 330 ret = f2fs_sync_fs(inode->i_sb, 1); 331 332 /* 333 * We've secured consistency through sync_fs. Following pino 334 * will be used only for fsynced inodes after checkpoint. 335 */ 336 try_to_fix_pino(inode); 337 clear_inode_flag(inode, FI_APPEND_WRITE); 338 clear_inode_flag(inode, FI_UPDATE_WRITE); 339 goto out; 340 } 341 sync_nodes: 342 atomic_inc(&sbi->wb_sync_req[NODE]); 343 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 344 atomic_dec(&sbi->wb_sync_req[NODE]); 345 if (ret) 346 goto out; 347 348 /* if cp_error was enabled, we should avoid infinite loop */ 349 if (unlikely(f2fs_cp_error(sbi))) { 350 ret = -EIO; 351 goto out; 352 } 353 354 if (f2fs_need_inode_block_update(sbi, ino)) { 355 f2fs_mark_inode_dirty_sync(inode, true); 356 f2fs_write_inode(inode, NULL); 357 goto sync_nodes; 358 } 359 360 /* 361 * If it's atomic_write, it's just fine to keep write ordering. So 362 * here we don't need to wait for node write completion, since we use 363 * node chain which serializes node blocks. If one of node writes are 364 * reordered, we can see simply broken chain, resulting in stopping 365 * roll-forward recovery. It means we'll recover all or none node blocks 366 * given fsync mark. 367 */ 368 if (!atomic) { 369 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 370 if (ret) 371 goto out; 372 } 373 374 /* once recovery info is written, don't need to tack this */ 375 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 376 clear_inode_flag(inode, FI_APPEND_WRITE); 377 flush_out: 378 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 379 ret = f2fs_issue_flush(sbi, inode->i_ino); 380 if (!ret) { 381 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 382 clear_inode_flag(inode, FI_UPDATE_WRITE); 383 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 384 } 385 f2fs_update_time(sbi, REQ_TIME); 386 out: 387 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 388 return ret; 389 } 390 391 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 392 { 393 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 394 return -EIO; 395 return f2fs_do_sync_file(file, start, end, datasync, false); 396 } 397 398 static bool __found_offset(struct address_space *mapping, 399 struct dnode_of_data *dn, pgoff_t index, int whence) 400 { 401 block_t blkaddr = f2fs_data_blkaddr(dn); 402 struct inode *inode = mapping->host; 403 bool compressed_cluster = false; 404 405 if (f2fs_compressed_file(inode)) { 406 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_page, 407 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size)); 408 409 compressed_cluster = first_blkaddr == COMPRESS_ADDR; 410 } 411 412 switch (whence) { 413 case SEEK_DATA: 414 if (__is_valid_data_blkaddr(blkaddr)) 415 return true; 416 if (blkaddr == NEW_ADDR && 417 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 418 return true; 419 if (compressed_cluster) 420 return true; 421 break; 422 case SEEK_HOLE: 423 if (compressed_cluster) 424 return false; 425 if (blkaddr == NULL_ADDR) 426 return true; 427 break; 428 } 429 return false; 430 } 431 432 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 433 { 434 struct inode *inode = file->f_mapping->host; 435 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 436 struct dnode_of_data dn; 437 pgoff_t pgofs, end_offset; 438 loff_t data_ofs = offset; 439 loff_t isize; 440 int err = 0; 441 442 inode_lock_shared(inode); 443 444 isize = i_size_read(inode); 445 if (offset >= isize) 446 goto fail; 447 448 /* handle inline data case */ 449 if (f2fs_has_inline_data(inode)) { 450 if (whence == SEEK_HOLE) { 451 data_ofs = isize; 452 goto found; 453 } else if (whence == SEEK_DATA) { 454 data_ofs = offset; 455 goto found; 456 } 457 } 458 459 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 460 461 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 462 set_new_dnode(&dn, inode, NULL, NULL, 0); 463 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 464 if (err && err != -ENOENT) { 465 goto fail; 466 } else if (err == -ENOENT) { 467 /* direct node does not exists */ 468 if (whence == SEEK_DATA) { 469 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 470 continue; 471 } else { 472 goto found; 473 } 474 } 475 476 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 477 478 /* find data/hole in dnode block */ 479 for (; dn.ofs_in_node < end_offset; 480 dn.ofs_in_node++, pgofs++, 481 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 482 block_t blkaddr; 483 484 blkaddr = f2fs_data_blkaddr(&dn); 485 486 if (__is_valid_data_blkaddr(blkaddr) && 487 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 488 blkaddr, DATA_GENERIC_ENHANCE)) { 489 f2fs_put_dnode(&dn); 490 goto fail; 491 } 492 493 if (__found_offset(file->f_mapping, &dn, 494 pgofs, whence)) { 495 f2fs_put_dnode(&dn); 496 goto found; 497 } 498 } 499 f2fs_put_dnode(&dn); 500 } 501 502 if (whence == SEEK_DATA) 503 goto fail; 504 found: 505 if (whence == SEEK_HOLE && data_ofs > isize) 506 data_ofs = isize; 507 inode_unlock_shared(inode); 508 return vfs_setpos(file, data_ofs, maxbytes); 509 fail: 510 inode_unlock_shared(inode); 511 return -ENXIO; 512 } 513 514 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 515 { 516 struct inode *inode = file->f_mapping->host; 517 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 518 519 switch (whence) { 520 case SEEK_SET: 521 case SEEK_CUR: 522 case SEEK_END: 523 return generic_file_llseek_size(file, offset, whence, 524 maxbytes, i_size_read(inode)); 525 case SEEK_DATA: 526 case SEEK_HOLE: 527 if (offset < 0) 528 return -ENXIO; 529 return f2fs_seek_block(file, offset, whence); 530 } 531 532 return -EINVAL; 533 } 534 535 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 536 { 537 struct inode *inode = file_inode(file); 538 539 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 540 return -EIO; 541 542 if (!f2fs_is_compress_backend_ready(inode)) 543 return -EOPNOTSUPP; 544 545 file_accessed(file); 546 vma->vm_ops = &f2fs_file_vm_ops; 547 548 f2fs_down_read(&F2FS_I(inode)->i_sem); 549 set_inode_flag(inode, FI_MMAP_FILE); 550 f2fs_up_read(&F2FS_I(inode)->i_sem); 551 552 return 0; 553 } 554 555 static int finish_preallocate_blocks(struct inode *inode) 556 { 557 int ret; 558 559 inode_lock(inode); 560 if (is_inode_flag_set(inode, FI_OPENED_FILE)) { 561 inode_unlock(inode); 562 return 0; 563 } 564 565 if (!file_should_truncate(inode)) { 566 set_inode_flag(inode, FI_OPENED_FILE); 567 inode_unlock(inode); 568 return 0; 569 } 570 571 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 572 filemap_invalidate_lock(inode->i_mapping); 573 574 truncate_setsize(inode, i_size_read(inode)); 575 ret = f2fs_truncate(inode); 576 577 filemap_invalidate_unlock(inode->i_mapping); 578 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 579 580 if (!ret) 581 set_inode_flag(inode, FI_OPENED_FILE); 582 583 inode_unlock(inode); 584 if (ret) 585 return ret; 586 587 file_dont_truncate(inode); 588 return 0; 589 } 590 591 static int f2fs_file_open(struct inode *inode, struct file *filp) 592 { 593 int err = fscrypt_file_open(inode, filp); 594 595 if (err) 596 return err; 597 598 if (!f2fs_is_compress_backend_ready(inode)) 599 return -EOPNOTSUPP; 600 601 err = fsverity_file_open(inode, filp); 602 if (err) 603 return err; 604 605 filp->f_mode |= FMODE_NOWAIT; 606 filp->f_mode |= FMODE_CAN_ODIRECT; 607 608 err = dquot_file_open(inode, filp); 609 if (err) 610 return err; 611 612 return finish_preallocate_blocks(inode); 613 } 614 615 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 616 { 617 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 618 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 619 __le32 *addr; 620 bool compressed_cluster = false; 621 int cluster_index = 0, valid_blocks = 0; 622 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 623 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 624 block_t blkstart; 625 int blklen = 0; 626 627 addr = get_dnode_addr(dn->inode, dn->node_page) + ofs; 628 blkstart = le32_to_cpu(*addr); 629 630 /* Assumption: truncation starts with cluster */ 631 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 632 block_t blkaddr = le32_to_cpu(*addr); 633 634 if (f2fs_compressed_file(dn->inode) && 635 !(cluster_index & (cluster_size - 1))) { 636 if (compressed_cluster) 637 f2fs_i_compr_blocks_update(dn->inode, 638 valid_blocks, false); 639 compressed_cluster = (blkaddr == COMPRESS_ADDR); 640 valid_blocks = 0; 641 } 642 643 if (blkaddr == NULL_ADDR) 644 goto next; 645 646 f2fs_set_data_blkaddr(dn, NULL_ADDR); 647 648 if (__is_valid_data_blkaddr(blkaddr)) { 649 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE)) 650 goto next; 651 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr, 652 DATA_GENERIC_ENHANCE)) 653 goto next; 654 if (compressed_cluster) 655 valid_blocks++; 656 } 657 658 if (blkstart + blklen == blkaddr) { 659 blklen++; 660 } else { 661 f2fs_invalidate_blocks(sbi, blkstart, blklen); 662 blkstart = blkaddr; 663 blklen = 1; 664 } 665 666 if (!released || blkaddr != COMPRESS_ADDR) 667 nr_free++; 668 669 continue; 670 671 next: 672 if (blklen) 673 f2fs_invalidate_blocks(sbi, blkstart, blklen); 674 675 blkstart = le32_to_cpu(*(addr + 1)); 676 blklen = 0; 677 } 678 679 if (blklen) 680 f2fs_invalidate_blocks(sbi, blkstart, blklen); 681 682 if (compressed_cluster) 683 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 684 685 if (nr_free) { 686 pgoff_t fofs; 687 /* 688 * once we invalidate valid blkaddr in range [ofs, ofs + count], 689 * we will invalidate all blkaddr in the whole range. 690 */ 691 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 692 dn->inode) + ofs; 693 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 694 f2fs_update_age_extent_cache_range(dn, fofs, len); 695 dec_valid_block_count(sbi, dn->inode, nr_free); 696 } 697 dn->ofs_in_node = ofs; 698 699 f2fs_update_time(sbi, REQ_TIME); 700 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 701 dn->ofs_in_node, nr_free); 702 } 703 704 static int truncate_partial_data_page(struct inode *inode, u64 from, 705 bool cache_only) 706 { 707 loff_t offset = from & (PAGE_SIZE - 1); 708 pgoff_t index = from >> PAGE_SHIFT; 709 struct address_space *mapping = inode->i_mapping; 710 struct folio *folio; 711 712 if (!offset && !cache_only) 713 return 0; 714 715 if (cache_only) { 716 folio = filemap_lock_folio(mapping, index); 717 if (IS_ERR(folio)) 718 return 0; 719 if (folio_test_uptodate(folio)) 720 goto truncate_out; 721 f2fs_folio_put(folio, true); 722 return 0; 723 } 724 725 folio = f2fs_get_lock_data_folio(inode, index, true); 726 if (IS_ERR(folio)) 727 return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio); 728 truncate_out: 729 f2fs_folio_wait_writeback(folio, DATA, true, true); 730 folio_zero_segment(folio, offset, folio_size(folio)); 731 732 /* An encrypted inode should have a key and truncate the last page. */ 733 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 734 if (!cache_only) 735 folio_mark_dirty(folio); 736 f2fs_folio_put(folio, true); 737 return 0; 738 } 739 740 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 741 { 742 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 743 struct dnode_of_data dn; 744 pgoff_t free_from; 745 int count = 0, err = 0; 746 struct page *ipage; 747 bool truncate_page = false; 748 749 trace_f2fs_truncate_blocks_enter(inode, from); 750 751 if (IS_DEVICE_ALIASING(inode) && from) { 752 err = -EINVAL; 753 goto out_err; 754 } 755 756 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 757 758 if (free_from >= max_file_blocks(inode)) 759 goto free_partial; 760 761 if (lock) 762 f2fs_lock_op(sbi); 763 764 ipage = f2fs_get_inode_page(sbi, inode->i_ino); 765 if (IS_ERR(ipage)) { 766 err = PTR_ERR(ipage); 767 goto out; 768 } 769 770 if (IS_DEVICE_ALIASING(inode)) { 771 struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ]; 772 struct extent_info ei = et->largest; 773 774 f2fs_invalidate_blocks(sbi, ei.blk, ei.len); 775 776 dec_valid_block_count(sbi, inode, ei.len); 777 f2fs_update_time(sbi, REQ_TIME); 778 779 f2fs_put_page(ipage, 1); 780 goto out; 781 } 782 783 if (f2fs_has_inline_data(inode)) { 784 f2fs_truncate_inline_inode(inode, ipage, from); 785 f2fs_put_page(ipage, 1); 786 truncate_page = true; 787 goto out; 788 } 789 790 set_new_dnode(&dn, inode, ipage, NULL, 0); 791 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 792 if (err) { 793 if (err == -ENOENT) 794 goto free_next; 795 goto out; 796 } 797 798 count = ADDRS_PER_PAGE(dn.node_page, inode); 799 800 count -= dn.ofs_in_node; 801 f2fs_bug_on(sbi, count < 0); 802 803 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 804 f2fs_truncate_data_blocks_range(&dn, count); 805 free_from += count; 806 } 807 808 f2fs_put_dnode(&dn); 809 free_next: 810 err = f2fs_truncate_inode_blocks(inode, free_from); 811 out: 812 if (lock) 813 f2fs_unlock_op(sbi); 814 free_partial: 815 /* lastly zero out the first data page */ 816 if (!err) 817 err = truncate_partial_data_page(inode, from, truncate_page); 818 out_err: 819 trace_f2fs_truncate_blocks_exit(inode, err); 820 return err; 821 } 822 823 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 824 { 825 u64 free_from = from; 826 int err; 827 828 #ifdef CONFIG_F2FS_FS_COMPRESSION 829 /* 830 * for compressed file, only support cluster size 831 * aligned truncation. 832 */ 833 if (f2fs_compressed_file(inode)) 834 free_from = round_up(from, 835 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 836 #endif 837 838 err = f2fs_do_truncate_blocks(inode, free_from, lock); 839 if (err) 840 return err; 841 842 #ifdef CONFIG_F2FS_FS_COMPRESSION 843 /* 844 * For compressed file, after release compress blocks, don't allow write 845 * direct, but we should allow write direct after truncate to zero. 846 */ 847 if (f2fs_compressed_file(inode) && !free_from 848 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 849 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 850 851 if (from != free_from) { 852 err = f2fs_truncate_partial_cluster(inode, from, lock); 853 if (err) 854 return err; 855 } 856 #endif 857 858 return 0; 859 } 860 861 int f2fs_truncate(struct inode *inode) 862 { 863 int err; 864 865 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 866 return -EIO; 867 868 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 869 S_ISLNK(inode->i_mode))) 870 return 0; 871 872 trace_f2fs_truncate(inode); 873 874 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 875 return -EIO; 876 877 err = f2fs_dquot_initialize(inode); 878 if (err) 879 return err; 880 881 /* we should check inline_data size */ 882 if (!f2fs_may_inline_data(inode)) { 883 err = f2fs_convert_inline_inode(inode); 884 if (err) 885 return err; 886 } 887 888 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 889 if (err) 890 return err; 891 892 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 893 f2fs_mark_inode_dirty_sync(inode, false); 894 return 0; 895 } 896 897 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 898 { 899 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 900 901 if (!fscrypt_dio_supported(inode)) 902 return true; 903 if (fsverity_active(inode)) 904 return true; 905 if (f2fs_compressed_file(inode)) 906 return true; 907 /* 908 * only force direct read to use buffered IO, for direct write, 909 * it expects inline data conversion before committing IO. 910 */ 911 if (f2fs_has_inline_data(inode) && rw == READ) 912 return true; 913 914 /* disallow direct IO if any of devices has unaligned blksize */ 915 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 916 return true; 917 /* 918 * for blkzoned device, fallback direct IO to buffered IO, so 919 * all IOs can be serialized by log-structured write. 920 */ 921 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) && 922 !f2fs_is_pinned_file(inode)) 923 return true; 924 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 925 return true; 926 927 return false; 928 } 929 930 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 931 struct kstat *stat, u32 request_mask, unsigned int query_flags) 932 { 933 struct inode *inode = d_inode(path->dentry); 934 struct f2fs_inode_info *fi = F2FS_I(inode); 935 struct f2fs_inode *ri = NULL; 936 unsigned int flags; 937 938 if (f2fs_has_extra_attr(inode) && 939 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 940 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 941 stat->result_mask |= STATX_BTIME; 942 stat->btime.tv_sec = fi->i_crtime.tv_sec; 943 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 944 } 945 946 /* 947 * Return the DIO alignment restrictions if requested. We only return 948 * this information when requested, since on encrypted files it might 949 * take a fair bit of work to get if the file wasn't opened recently. 950 * 951 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 952 * cannot represent that, so in that case we report no DIO support. 953 */ 954 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 955 unsigned int bsize = i_blocksize(inode); 956 957 stat->result_mask |= STATX_DIOALIGN; 958 if (!f2fs_force_buffered_io(inode, WRITE)) { 959 stat->dio_mem_align = bsize; 960 stat->dio_offset_align = bsize; 961 } 962 } 963 964 flags = fi->i_flags; 965 if (flags & F2FS_COMPR_FL) 966 stat->attributes |= STATX_ATTR_COMPRESSED; 967 if (flags & F2FS_APPEND_FL) 968 stat->attributes |= STATX_ATTR_APPEND; 969 if (IS_ENCRYPTED(inode)) 970 stat->attributes |= STATX_ATTR_ENCRYPTED; 971 if (flags & F2FS_IMMUTABLE_FL) 972 stat->attributes |= STATX_ATTR_IMMUTABLE; 973 if (flags & F2FS_NODUMP_FL) 974 stat->attributes |= STATX_ATTR_NODUMP; 975 if (IS_VERITY(inode)) 976 stat->attributes |= STATX_ATTR_VERITY; 977 978 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 979 STATX_ATTR_APPEND | 980 STATX_ATTR_ENCRYPTED | 981 STATX_ATTR_IMMUTABLE | 982 STATX_ATTR_NODUMP | 983 STATX_ATTR_VERITY); 984 985 generic_fillattr(idmap, request_mask, inode, stat); 986 987 /* we need to show initial sectors used for inline_data/dentries */ 988 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 989 f2fs_has_inline_dentry(inode)) 990 stat->blocks += (stat->size + 511) >> 9; 991 992 return 0; 993 } 994 995 #ifdef CONFIG_F2FS_FS_POSIX_ACL 996 static void __setattr_copy(struct mnt_idmap *idmap, 997 struct inode *inode, const struct iattr *attr) 998 { 999 unsigned int ia_valid = attr->ia_valid; 1000 1001 i_uid_update(idmap, attr, inode); 1002 i_gid_update(idmap, attr, inode); 1003 if (ia_valid & ATTR_ATIME) 1004 inode_set_atime_to_ts(inode, attr->ia_atime); 1005 if (ia_valid & ATTR_MTIME) 1006 inode_set_mtime_to_ts(inode, attr->ia_mtime); 1007 if (ia_valid & ATTR_CTIME) 1008 inode_set_ctime_to_ts(inode, attr->ia_ctime); 1009 if (ia_valid & ATTR_MODE) { 1010 umode_t mode = attr->ia_mode; 1011 1012 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode))) 1013 mode &= ~S_ISGID; 1014 set_acl_inode(inode, mode); 1015 } 1016 } 1017 #else 1018 #define __setattr_copy setattr_copy 1019 #endif 1020 1021 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 1022 struct iattr *attr) 1023 { 1024 struct inode *inode = d_inode(dentry); 1025 struct f2fs_inode_info *fi = F2FS_I(inode); 1026 int err; 1027 1028 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1029 return -EIO; 1030 1031 if (unlikely(IS_IMMUTABLE(inode))) 1032 return -EPERM; 1033 1034 if (unlikely(IS_APPEND(inode) && 1035 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 1036 ATTR_GID | ATTR_TIMES_SET)))) 1037 return -EPERM; 1038 1039 if ((attr->ia_valid & ATTR_SIZE)) { 1040 if (!f2fs_is_compress_backend_ready(inode) || 1041 IS_DEVICE_ALIASING(inode)) 1042 return -EOPNOTSUPP; 1043 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 1044 !IS_ALIGNED(attr->ia_size, 1045 F2FS_BLK_TO_BYTES(fi->i_cluster_size))) 1046 return -EINVAL; 1047 } 1048 1049 err = setattr_prepare(idmap, dentry, attr); 1050 if (err) 1051 return err; 1052 1053 err = fscrypt_prepare_setattr(dentry, attr); 1054 if (err) 1055 return err; 1056 1057 err = fsverity_prepare_setattr(dentry, attr); 1058 if (err) 1059 return err; 1060 1061 if (is_quota_modification(idmap, inode, attr)) { 1062 err = f2fs_dquot_initialize(inode); 1063 if (err) 1064 return err; 1065 } 1066 if (i_uid_needs_update(idmap, attr, inode) || 1067 i_gid_needs_update(idmap, attr, inode)) { 1068 f2fs_lock_op(F2FS_I_SB(inode)); 1069 err = dquot_transfer(idmap, inode, attr); 1070 if (err) { 1071 set_sbi_flag(F2FS_I_SB(inode), 1072 SBI_QUOTA_NEED_REPAIR); 1073 f2fs_unlock_op(F2FS_I_SB(inode)); 1074 return err; 1075 } 1076 /* 1077 * update uid/gid under lock_op(), so that dquot and inode can 1078 * be updated atomically. 1079 */ 1080 i_uid_update(idmap, attr, inode); 1081 i_gid_update(idmap, attr, inode); 1082 f2fs_mark_inode_dirty_sync(inode, true); 1083 f2fs_unlock_op(F2FS_I_SB(inode)); 1084 } 1085 1086 if (attr->ia_valid & ATTR_SIZE) { 1087 loff_t old_size = i_size_read(inode); 1088 1089 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1090 /* 1091 * should convert inline inode before i_size_write to 1092 * keep smaller than inline_data size with inline flag. 1093 */ 1094 err = f2fs_convert_inline_inode(inode); 1095 if (err) 1096 return err; 1097 } 1098 1099 /* 1100 * wait for inflight dio, blocks should be removed after 1101 * IO completion. 1102 */ 1103 if (attr->ia_size < old_size) 1104 inode_dio_wait(inode); 1105 1106 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 1107 filemap_invalidate_lock(inode->i_mapping); 1108 1109 truncate_setsize(inode, attr->ia_size); 1110 1111 if (attr->ia_size <= old_size) 1112 err = f2fs_truncate(inode); 1113 /* 1114 * do not trim all blocks after i_size if target size is 1115 * larger than i_size. 1116 */ 1117 filemap_invalidate_unlock(inode->i_mapping); 1118 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 1119 if (err) 1120 return err; 1121 1122 spin_lock(&fi->i_size_lock); 1123 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1124 fi->last_disk_size = i_size_read(inode); 1125 spin_unlock(&fi->i_size_lock); 1126 } 1127 1128 __setattr_copy(idmap, inode, attr); 1129 1130 if (attr->ia_valid & ATTR_MODE) { 1131 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1132 1133 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1134 if (!err) 1135 inode->i_mode = fi->i_acl_mode; 1136 clear_inode_flag(inode, FI_ACL_MODE); 1137 } 1138 } 1139 1140 /* file size may changed here */ 1141 f2fs_mark_inode_dirty_sync(inode, true); 1142 1143 /* inode change will produce dirty node pages flushed by checkpoint */ 1144 f2fs_balance_fs(F2FS_I_SB(inode), true); 1145 1146 return err; 1147 } 1148 1149 const struct inode_operations f2fs_file_inode_operations = { 1150 .getattr = f2fs_getattr, 1151 .setattr = f2fs_setattr, 1152 .get_inode_acl = f2fs_get_acl, 1153 .set_acl = f2fs_set_acl, 1154 .listxattr = f2fs_listxattr, 1155 .fiemap = f2fs_fiemap, 1156 .fileattr_get = f2fs_fileattr_get, 1157 .fileattr_set = f2fs_fileattr_set, 1158 }; 1159 1160 static int fill_zero(struct inode *inode, pgoff_t index, 1161 loff_t start, loff_t len) 1162 { 1163 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1164 struct page *page; 1165 1166 if (!len) 1167 return 0; 1168 1169 f2fs_balance_fs(sbi, true); 1170 1171 f2fs_lock_op(sbi); 1172 page = f2fs_get_new_data_page(inode, NULL, index, false); 1173 f2fs_unlock_op(sbi); 1174 1175 if (IS_ERR(page)) 1176 return PTR_ERR(page); 1177 1178 f2fs_wait_on_page_writeback(page, DATA, true, true); 1179 zero_user(page, start, len); 1180 set_page_dirty(page); 1181 f2fs_put_page(page, 1); 1182 return 0; 1183 } 1184 1185 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1186 { 1187 int err; 1188 1189 while (pg_start < pg_end) { 1190 struct dnode_of_data dn; 1191 pgoff_t end_offset, count; 1192 1193 set_new_dnode(&dn, inode, NULL, NULL, 0); 1194 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1195 if (err) { 1196 if (err == -ENOENT) { 1197 pg_start = f2fs_get_next_page_offset(&dn, 1198 pg_start); 1199 continue; 1200 } 1201 return err; 1202 } 1203 1204 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1205 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1206 1207 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1208 1209 f2fs_truncate_data_blocks_range(&dn, count); 1210 f2fs_put_dnode(&dn); 1211 1212 pg_start += count; 1213 } 1214 return 0; 1215 } 1216 1217 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1218 { 1219 pgoff_t pg_start, pg_end; 1220 loff_t off_start, off_end; 1221 int ret; 1222 1223 ret = f2fs_convert_inline_inode(inode); 1224 if (ret) 1225 return ret; 1226 1227 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1228 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1229 1230 off_start = offset & (PAGE_SIZE - 1); 1231 off_end = (offset + len) & (PAGE_SIZE - 1); 1232 1233 if (pg_start == pg_end) { 1234 ret = fill_zero(inode, pg_start, off_start, 1235 off_end - off_start); 1236 if (ret) 1237 return ret; 1238 } else { 1239 if (off_start) { 1240 ret = fill_zero(inode, pg_start++, off_start, 1241 PAGE_SIZE - off_start); 1242 if (ret) 1243 return ret; 1244 } 1245 if (off_end) { 1246 ret = fill_zero(inode, pg_end, 0, off_end); 1247 if (ret) 1248 return ret; 1249 } 1250 1251 if (pg_start < pg_end) { 1252 loff_t blk_start, blk_end; 1253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1254 1255 f2fs_balance_fs(sbi, true); 1256 1257 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1258 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1259 1260 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1261 filemap_invalidate_lock(inode->i_mapping); 1262 1263 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1264 1265 f2fs_lock_op(sbi); 1266 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1267 f2fs_unlock_op(sbi); 1268 1269 filemap_invalidate_unlock(inode->i_mapping); 1270 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1271 } 1272 } 1273 1274 return ret; 1275 } 1276 1277 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1278 int *do_replace, pgoff_t off, pgoff_t len) 1279 { 1280 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1281 struct dnode_of_data dn; 1282 int ret, done, i; 1283 1284 next_dnode: 1285 set_new_dnode(&dn, inode, NULL, NULL, 0); 1286 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1287 if (ret && ret != -ENOENT) { 1288 return ret; 1289 } else if (ret == -ENOENT) { 1290 if (dn.max_level == 0) 1291 return -ENOENT; 1292 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1293 dn.ofs_in_node, len); 1294 blkaddr += done; 1295 do_replace += done; 1296 goto next; 1297 } 1298 1299 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1300 dn.ofs_in_node, len); 1301 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1302 *blkaddr = f2fs_data_blkaddr(&dn); 1303 1304 if (__is_valid_data_blkaddr(*blkaddr) && 1305 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1306 DATA_GENERIC_ENHANCE)) { 1307 f2fs_put_dnode(&dn); 1308 return -EFSCORRUPTED; 1309 } 1310 1311 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1312 1313 if (f2fs_lfs_mode(sbi)) { 1314 f2fs_put_dnode(&dn); 1315 return -EOPNOTSUPP; 1316 } 1317 1318 /* do not invalidate this block address */ 1319 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1320 *do_replace = 1; 1321 } 1322 } 1323 f2fs_put_dnode(&dn); 1324 next: 1325 len -= done; 1326 off += done; 1327 if (len) 1328 goto next_dnode; 1329 return 0; 1330 } 1331 1332 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1333 int *do_replace, pgoff_t off, int len) 1334 { 1335 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1336 struct dnode_of_data dn; 1337 int ret, i; 1338 1339 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1340 if (*do_replace == 0) 1341 continue; 1342 1343 set_new_dnode(&dn, inode, NULL, NULL, 0); 1344 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1345 if (ret) { 1346 dec_valid_block_count(sbi, inode, 1); 1347 f2fs_invalidate_blocks(sbi, *blkaddr, 1); 1348 } else { 1349 f2fs_update_data_blkaddr(&dn, *blkaddr); 1350 } 1351 f2fs_put_dnode(&dn); 1352 } 1353 return 0; 1354 } 1355 1356 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1357 block_t *blkaddr, int *do_replace, 1358 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1359 { 1360 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1361 pgoff_t i = 0; 1362 int ret; 1363 1364 while (i < len) { 1365 if (blkaddr[i] == NULL_ADDR && !full) { 1366 i++; 1367 continue; 1368 } 1369 1370 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1371 struct dnode_of_data dn; 1372 struct node_info ni; 1373 size_t new_size; 1374 pgoff_t ilen; 1375 1376 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1377 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1378 if (ret) 1379 return ret; 1380 1381 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1382 if (ret) { 1383 f2fs_put_dnode(&dn); 1384 return ret; 1385 } 1386 1387 ilen = min((pgoff_t) 1388 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1389 dn.ofs_in_node, len - i); 1390 do { 1391 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1392 f2fs_truncate_data_blocks_range(&dn, 1); 1393 1394 if (do_replace[i]) { 1395 f2fs_i_blocks_write(src_inode, 1396 1, false, false); 1397 f2fs_i_blocks_write(dst_inode, 1398 1, true, false); 1399 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1400 blkaddr[i], ni.version, true, false); 1401 1402 do_replace[i] = 0; 1403 } 1404 dn.ofs_in_node++; 1405 i++; 1406 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1407 if (dst_inode->i_size < new_size) 1408 f2fs_i_size_write(dst_inode, new_size); 1409 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1410 1411 f2fs_put_dnode(&dn); 1412 } else { 1413 struct page *psrc, *pdst; 1414 1415 psrc = f2fs_get_lock_data_page(src_inode, 1416 src + i, true); 1417 if (IS_ERR(psrc)) 1418 return PTR_ERR(psrc); 1419 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1420 true); 1421 if (IS_ERR(pdst)) { 1422 f2fs_put_page(psrc, 1); 1423 return PTR_ERR(pdst); 1424 } 1425 1426 f2fs_wait_on_page_writeback(pdst, DATA, true, true); 1427 1428 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1429 set_page_dirty(pdst); 1430 set_page_private_gcing(pdst); 1431 f2fs_put_page(pdst, 1); 1432 f2fs_put_page(psrc, 1); 1433 1434 ret = f2fs_truncate_hole(src_inode, 1435 src + i, src + i + 1); 1436 if (ret) 1437 return ret; 1438 i++; 1439 } 1440 } 1441 return 0; 1442 } 1443 1444 static int __exchange_data_block(struct inode *src_inode, 1445 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1446 pgoff_t len, bool full) 1447 { 1448 block_t *src_blkaddr; 1449 int *do_replace; 1450 pgoff_t olen; 1451 int ret; 1452 1453 while (len) { 1454 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1455 1456 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1457 array_size(olen, sizeof(block_t)), 1458 GFP_NOFS); 1459 if (!src_blkaddr) 1460 return -ENOMEM; 1461 1462 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1463 array_size(olen, sizeof(int)), 1464 GFP_NOFS); 1465 if (!do_replace) { 1466 kvfree(src_blkaddr); 1467 return -ENOMEM; 1468 } 1469 1470 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1471 do_replace, src, olen); 1472 if (ret) 1473 goto roll_back; 1474 1475 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1476 do_replace, src, dst, olen, full); 1477 if (ret) 1478 goto roll_back; 1479 1480 src += olen; 1481 dst += olen; 1482 len -= olen; 1483 1484 kvfree(src_blkaddr); 1485 kvfree(do_replace); 1486 } 1487 return 0; 1488 1489 roll_back: 1490 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1491 kvfree(src_blkaddr); 1492 kvfree(do_replace); 1493 return ret; 1494 } 1495 1496 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1497 { 1498 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1499 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1500 pgoff_t start = offset >> PAGE_SHIFT; 1501 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1502 int ret; 1503 1504 f2fs_balance_fs(sbi, true); 1505 1506 /* avoid gc operation during block exchange */ 1507 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1508 filemap_invalidate_lock(inode->i_mapping); 1509 1510 f2fs_lock_op(sbi); 1511 f2fs_drop_extent_tree(inode); 1512 truncate_pagecache(inode, offset); 1513 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1514 f2fs_unlock_op(sbi); 1515 1516 filemap_invalidate_unlock(inode->i_mapping); 1517 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1518 return ret; 1519 } 1520 1521 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1522 { 1523 loff_t new_size; 1524 int ret; 1525 1526 if (offset + len >= i_size_read(inode)) 1527 return -EINVAL; 1528 1529 /* collapse range should be aligned to block size of f2fs. */ 1530 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1531 return -EINVAL; 1532 1533 ret = f2fs_convert_inline_inode(inode); 1534 if (ret) 1535 return ret; 1536 1537 /* write out all dirty pages from offset */ 1538 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1539 if (ret) 1540 return ret; 1541 1542 ret = f2fs_do_collapse(inode, offset, len); 1543 if (ret) 1544 return ret; 1545 1546 /* write out all moved pages, if possible */ 1547 filemap_invalidate_lock(inode->i_mapping); 1548 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1549 truncate_pagecache(inode, offset); 1550 1551 new_size = i_size_read(inode) - len; 1552 ret = f2fs_truncate_blocks(inode, new_size, true); 1553 filemap_invalidate_unlock(inode->i_mapping); 1554 if (!ret) 1555 f2fs_i_size_write(inode, new_size); 1556 return ret; 1557 } 1558 1559 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1560 pgoff_t end) 1561 { 1562 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1563 pgoff_t index = start; 1564 unsigned int ofs_in_node = dn->ofs_in_node; 1565 blkcnt_t count = 0; 1566 int ret; 1567 1568 for (; index < end; index++, dn->ofs_in_node++) { 1569 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1570 count++; 1571 } 1572 1573 dn->ofs_in_node = ofs_in_node; 1574 ret = f2fs_reserve_new_blocks(dn, count); 1575 if (ret) 1576 return ret; 1577 1578 dn->ofs_in_node = ofs_in_node; 1579 for (index = start; index < end; index++, dn->ofs_in_node++) { 1580 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1581 /* 1582 * f2fs_reserve_new_blocks will not guarantee entire block 1583 * allocation. 1584 */ 1585 if (dn->data_blkaddr == NULL_ADDR) { 1586 ret = -ENOSPC; 1587 break; 1588 } 1589 1590 if (dn->data_blkaddr == NEW_ADDR) 1591 continue; 1592 1593 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1594 DATA_GENERIC_ENHANCE)) { 1595 ret = -EFSCORRUPTED; 1596 break; 1597 } 1598 1599 f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1); 1600 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1601 } 1602 1603 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1604 f2fs_update_age_extent_cache_range(dn, start, index - start); 1605 1606 return ret; 1607 } 1608 1609 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1610 int mode) 1611 { 1612 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1613 struct address_space *mapping = inode->i_mapping; 1614 pgoff_t index, pg_start, pg_end; 1615 loff_t new_size = i_size_read(inode); 1616 loff_t off_start, off_end; 1617 int ret = 0; 1618 1619 ret = inode_newsize_ok(inode, (len + offset)); 1620 if (ret) 1621 return ret; 1622 1623 ret = f2fs_convert_inline_inode(inode); 1624 if (ret) 1625 return ret; 1626 1627 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1628 if (ret) 1629 return ret; 1630 1631 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1632 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1633 1634 off_start = offset & (PAGE_SIZE - 1); 1635 off_end = (offset + len) & (PAGE_SIZE - 1); 1636 1637 if (pg_start == pg_end) { 1638 ret = fill_zero(inode, pg_start, off_start, 1639 off_end - off_start); 1640 if (ret) 1641 return ret; 1642 1643 new_size = max_t(loff_t, new_size, offset + len); 1644 } else { 1645 if (off_start) { 1646 ret = fill_zero(inode, pg_start++, off_start, 1647 PAGE_SIZE - off_start); 1648 if (ret) 1649 return ret; 1650 1651 new_size = max_t(loff_t, new_size, 1652 (loff_t)pg_start << PAGE_SHIFT); 1653 } 1654 1655 for (index = pg_start; index < pg_end;) { 1656 struct dnode_of_data dn; 1657 unsigned int end_offset; 1658 pgoff_t end; 1659 1660 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1661 filemap_invalidate_lock(mapping); 1662 1663 truncate_pagecache_range(inode, 1664 (loff_t)index << PAGE_SHIFT, 1665 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1666 1667 f2fs_lock_op(sbi); 1668 1669 set_new_dnode(&dn, inode, NULL, NULL, 0); 1670 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1671 if (ret) { 1672 f2fs_unlock_op(sbi); 1673 filemap_invalidate_unlock(mapping); 1674 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1675 goto out; 1676 } 1677 1678 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1679 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1680 1681 ret = f2fs_do_zero_range(&dn, index, end); 1682 f2fs_put_dnode(&dn); 1683 1684 f2fs_unlock_op(sbi); 1685 filemap_invalidate_unlock(mapping); 1686 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1687 1688 f2fs_balance_fs(sbi, dn.node_changed); 1689 1690 if (ret) 1691 goto out; 1692 1693 index = end; 1694 new_size = max_t(loff_t, new_size, 1695 (loff_t)index << PAGE_SHIFT); 1696 } 1697 1698 if (off_end) { 1699 ret = fill_zero(inode, pg_end, 0, off_end); 1700 if (ret) 1701 goto out; 1702 1703 new_size = max_t(loff_t, new_size, offset + len); 1704 } 1705 } 1706 1707 out: 1708 if (new_size > i_size_read(inode)) { 1709 if (mode & FALLOC_FL_KEEP_SIZE) 1710 file_set_keep_isize(inode); 1711 else 1712 f2fs_i_size_write(inode, new_size); 1713 } 1714 return ret; 1715 } 1716 1717 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1718 { 1719 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1720 struct address_space *mapping = inode->i_mapping; 1721 pgoff_t nr, pg_start, pg_end, delta, idx; 1722 loff_t new_size; 1723 int ret = 0; 1724 1725 new_size = i_size_read(inode) + len; 1726 ret = inode_newsize_ok(inode, new_size); 1727 if (ret) 1728 return ret; 1729 1730 if (offset >= i_size_read(inode)) 1731 return -EINVAL; 1732 1733 /* insert range should be aligned to block size of f2fs. */ 1734 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1735 return -EINVAL; 1736 1737 ret = f2fs_convert_inline_inode(inode); 1738 if (ret) 1739 return ret; 1740 1741 f2fs_balance_fs(sbi, true); 1742 1743 filemap_invalidate_lock(mapping); 1744 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1745 filemap_invalidate_unlock(mapping); 1746 if (ret) 1747 return ret; 1748 1749 /* write out all dirty pages from offset */ 1750 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1751 if (ret) 1752 return ret; 1753 1754 pg_start = offset >> PAGE_SHIFT; 1755 pg_end = (offset + len) >> PAGE_SHIFT; 1756 delta = pg_end - pg_start; 1757 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1758 1759 /* avoid gc operation during block exchange */ 1760 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1761 filemap_invalidate_lock(mapping); 1762 truncate_pagecache(inode, offset); 1763 1764 while (!ret && idx > pg_start) { 1765 nr = idx - pg_start; 1766 if (nr > delta) 1767 nr = delta; 1768 idx -= nr; 1769 1770 f2fs_lock_op(sbi); 1771 f2fs_drop_extent_tree(inode); 1772 1773 ret = __exchange_data_block(inode, inode, idx, 1774 idx + delta, nr, false); 1775 f2fs_unlock_op(sbi); 1776 } 1777 filemap_invalidate_unlock(mapping); 1778 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1779 if (ret) 1780 return ret; 1781 1782 /* write out all moved pages, if possible */ 1783 filemap_invalidate_lock(mapping); 1784 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1785 truncate_pagecache(inode, offset); 1786 filemap_invalidate_unlock(mapping); 1787 1788 if (!ret) 1789 f2fs_i_size_write(inode, new_size); 1790 return ret; 1791 } 1792 1793 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1794 loff_t len, int mode) 1795 { 1796 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1797 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1798 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1799 .m_may_create = true }; 1800 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1801 .init_gc_type = FG_GC, 1802 .should_migrate_blocks = false, 1803 .err_gc_skipped = true, 1804 .nr_free_secs = 0 }; 1805 pgoff_t pg_start, pg_end; 1806 loff_t new_size; 1807 loff_t off_end; 1808 block_t expanded = 0; 1809 int err; 1810 1811 err = inode_newsize_ok(inode, (len + offset)); 1812 if (err) 1813 return err; 1814 1815 err = f2fs_convert_inline_inode(inode); 1816 if (err) 1817 return err; 1818 1819 f2fs_balance_fs(sbi, true); 1820 1821 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1822 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1823 off_end = (offset + len) & (PAGE_SIZE - 1); 1824 1825 map.m_lblk = pg_start; 1826 map.m_len = pg_end - pg_start; 1827 if (off_end) 1828 map.m_len++; 1829 1830 if (!map.m_len) 1831 return 0; 1832 1833 if (f2fs_is_pinned_file(inode)) { 1834 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1835 block_t sec_len = roundup(map.m_len, sec_blks); 1836 1837 map.m_len = sec_blks; 1838 next_alloc: 1839 f2fs_down_write(&sbi->pin_sem); 1840 1841 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1842 if (has_not_enough_free_secs(sbi, 0, 0)) { 1843 f2fs_up_write(&sbi->pin_sem); 1844 err = -ENOSPC; 1845 f2fs_warn_ratelimited(sbi, 1846 "ino:%lu, start:%lu, end:%lu, need to trigger GC to " 1847 "reclaim enough free segment when checkpoint is enabled", 1848 inode->i_ino, pg_start, pg_end); 1849 goto out_err; 1850 } 1851 } 1852 1853 if (has_not_enough_free_secs(sbi, 0, f2fs_sb_has_blkzoned(sbi) ? 1854 ZONED_PIN_SEC_REQUIRED_COUNT : 1855 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1856 f2fs_down_write(&sbi->gc_lock); 1857 stat_inc_gc_call_count(sbi, FOREGROUND); 1858 err = f2fs_gc(sbi, &gc_control); 1859 if (err && err != -ENODATA) { 1860 f2fs_up_write(&sbi->pin_sem); 1861 goto out_err; 1862 } 1863 } 1864 1865 err = f2fs_allocate_pinning_section(sbi); 1866 if (err) { 1867 f2fs_up_write(&sbi->pin_sem); 1868 goto out_err; 1869 } 1870 1871 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1872 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1873 file_dont_truncate(inode); 1874 1875 f2fs_up_write(&sbi->pin_sem); 1876 1877 expanded += map.m_len; 1878 sec_len -= map.m_len; 1879 map.m_lblk += map.m_len; 1880 if (!err && sec_len) 1881 goto next_alloc; 1882 1883 map.m_len = expanded; 1884 } else { 1885 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1886 expanded = map.m_len; 1887 } 1888 out_err: 1889 if (err) { 1890 pgoff_t last_off; 1891 1892 if (!expanded) 1893 return err; 1894 1895 last_off = pg_start + expanded - 1; 1896 1897 /* update new size to the failed position */ 1898 new_size = (last_off == pg_end) ? offset + len : 1899 (loff_t)(last_off + 1) << PAGE_SHIFT; 1900 } else { 1901 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1902 } 1903 1904 if (new_size > i_size_read(inode)) { 1905 if (mode & FALLOC_FL_KEEP_SIZE) 1906 file_set_keep_isize(inode); 1907 else 1908 f2fs_i_size_write(inode, new_size); 1909 } 1910 1911 return err; 1912 } 1913 1914 static long f2fs_fallocate(struct file *file, int mode, 1915 loff_t offset, loff_t len) 1916 { 1917 struct inode *inode = file_inode(file); 1918 long ret = 0; 1919 1920 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1921 return -EIO; 1922 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1923 return -ENOSPC; 1924 if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode)) 1925 return -EOPNOTSUPP; 1926 1927 /* f2fs only support ->fallocate for regular file */ 1928 if (!S_ISREG(inode->i_mode)) 1929 return -EINVAL; 1930 1931 if (IS_ENCRYPTED(inode) && 1932 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1933 return -EOPNOTSUPP; 1934 1935 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1936 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1937 FALLOC_FL_INSERT_RANGE)) 1938 return -EOPNOTSUPP; 1939 1940 inode_lock(inode); 1941 1942 /* 1943 * Pinned file should not support partial truncation since the block 1944 * can be used by applications. 1945 */ 1946 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1947 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1948 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 1949 ret = -EOPNOTSUPP; 1950 goto out; 1951 } 1952 1953 ret = file_modified(file); 1954 if (ret) 1955 goto out; 1956 1957 /* 1958 * wait for inflight dio, blocks should be removed after IO 1959 * completion. 1960 */ 1961 inode_dio_wait(inode); 1962 1963 if (mode & FALLOC_FL_PUNCH_HOLE) { 1964 if (offset >= inode->i_size) 1965 goto out; 1966 1967 ret = f2fs_punch_hole(inode, offset, len); 1968 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1969 ret = f2fs_collapse_range(inode, offset, len); 1970 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1971 ret = f2fs_zero_range(inode, offset, len, mode); 1972 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1973 ret = f2fs_insert_range(inode, offset, len); 1974 } else { 1975 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1976 } 1977 1978 if (!ret) { 1979 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1980 f2fs_mark_inode_dirty_sync(inode, false); 1981 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1982 } 1983 1984 out: 1985 inode_unlock(inode); 1986 1987 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1988 return ret; 1989 } 1990 1991 static int f2fs_release_file(struct inode *inode, struct file *filp) 1992 { 1993 /* 1994 * f2fs_release_file is called at every close calls. So we should 1995 * not drop any inmemory pages by close called by other process. 1996 */ 1997 if (!(filp->f_mode & FMODE_WRITE) || 1998 atomic_read(&inode->i_writecount) != 1) 1999 return 0; 2000 2001 inode_lock(inode); 2002 f2fs_abort_atomic_write(inode, true); 2003 inode_unlock(inode); 2004 2005 return 0; 2006 } 2007 2008 static int f2fs_file_flush(struct file *file, fl_owner_t id) 2009 { 2010 struct inode *inode = file_inode(file); 2011 2012 /* 2013 * If the process doing a transaction is crashed, we should do 2014 * roll-back. Otherwise, other reader/write can see corrupted database 2015 * until all the writers close its file. Since this should be done 2016 * before dropping file lock, it needs to do in ->flush. 2017 */ 2018 if (F2FS_I(inode)->atomic_write_task == current && 2019 (current->flags & PF_EXITING)) { 2020 inode_lock(inode); 2021 f2fs_abort_atomic_write(inode, true); 2022 inode_unlock(inode); 2023 } 2024 2025 return 0; 2026 } 2027 2028 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 2029 { 2030 struct f2fs_inode_info *fi = F2FS_I(inode); 2031 u32 masked_flags = fi->i_flags & mask; 2032 2033 /* mask can be shrunk by flags_valid selector */ 2034 iflags &= mask; 2035 2036 /* Is it quota file? Do not allow user to mess with it */ 2037 if (IS_NOQUOTA(inode)) 2038 return -EPERM; 2039 2040 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 2041 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 2042 return -EOPNOTSUPP; 2043 if (!f2fs_empty_dir(inode)) 2044 return -ENOTEMPTY; 2045 } 2046 2047 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 2048 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 2049 return -EOPNOTSUPP; 2050 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 2051 return -EINVAL; 2052 } 2053 2054 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 2055 if (masked_flags & F2FS_COMPR_FL) { 2056 if (!f2fs_disable_compressed_file(inode)) 2057 return -EINVAL; 2058 } else { 2059 /* try to convert inline_data to support compression */ 2060 int err = f2fs_convert_inline_inode(inode); 2061 if (err) 2062 return err; 2063 2064 f2fs_down_write(&fi->i_sem); 2065 if (!f2fs_may_compress(inode) || 2066 (S_ISREG(inode->i_mode) && 2067 F2FS_HAS_BLOCKS(inode))) { 2068 f2fs_up_write(&fi->i_sem); 2069 return -EINVAL; 2070 } 2071 err = set_compress_context(inode); 2072 f2fs_up_write(&fi->i_sem); 2073 2074 if (err) 2075 return err; 2076 } 2077 } 2078 2079 fi->i_flags = iflags | (fi->i_flags & ~mask); 2080 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 2081 (fi->i_flags & F2FS_NOCOMP_FL)); 2082 2083 if (fi->i_flags & F2FS_PROJINHERIT_FL) 2084 set_inode_flag(inode, FI_PROJ_INHERIT); 2085 else 2086 clear_inode_flag(inode, FI_PROJ_INHERIT); 2087 2088 inode_set_ctime_current(inode); 2089 f2fs_set_inode_flags(inode); 2090 f2fs_mark_inode_dirty_sync(inode, true); 2091 return 0; 2092 } 2093 2094 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 2095 2096 /* 2097 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 2098 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 2099 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 2100 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 2101 * 2102 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 2103 * FS_IOC_FSSETXATTR is done by the VFS. 2104 */ 2105 2106 static const struct { 2107 u32 iflag; 2108 u32 fsflag; 2109 } f2fs_fsflags_map[] = { 2110 { F2FS_COMPR_FL, FS_COMPR_FL }, 2111 { F2FS_SYNC_FL, FS_SYNC_FL }, 2112 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 2113 { F2FS_APPEND_FL, FS_APPEND_FL }, 2114 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 2115 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 2116 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 2117 { F2FS_INDEX_FL, FS_INDEX_FL }, 2118 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 2119 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2120 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2121 }; 2122 2123 #define F2FS_GETTABLE_FS_FL ( \ 2124 FS_COMPR_FL | \ 2125 FS_SYNC_FL | \ 2126 FS_IMMUTABLE_FL | \ 2127 FS_APPEND_FL | \ 2128 FS_NODUMP_FL | \ 2129 FS_NOATIME_FL | \ 2130 FS_NOCOMP_FL | \ 2131 FS_INDEX_FL | \ 2132 FS_DIRSYNC_FL | \ 2133 FS_PROJINHERIT_FL | \ 2134 FS_ENCRYPT_FL | \ 2135 FS_INLINE_DATA_FL | \ 2136 FS_NOCOW_FL | \ 2137 FS_VERITY_FL | \ 2138 FS_CASEFOLD_FL) 2139 2140 #define F2FS_SETTABLE_FS_FL ( \ 2141 FS_COMPR_FL | \ 2142 FS_SYNC_FL | \ 2143 FS_IMMUTABLE_FL | \ 2144 FS_APPEND_FL | \ 2145 FS_NODUMP_FL | \ 2146 FS_NOATIME_FL | \ 2147 FS_NOCOMP_FL | \ 2148 FS_DIRSYNC_FL | \ 2149 FS_PROJINHERIT_FL | \ 2150 FS_CASEFOLD_FL) 2151 2152 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2153 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2154 { 2155 u32 fsflags = 0; 2156 int i; 2157 2158 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2159 if (iflags & f2fs_fsflags_map[i].iflag) 2160 fsflags |= f2fs_fsflags_map[i].fsflag; 2161 2162 return fsflags; 2163 } 2164 2165 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2166 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2167 { 2168 u32 iflags = 0; 2169 int i; 2170 2171 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2172 if (fsflags & f2fs_fsflags_map[i].fsflag) 2173 iflags |= f2fs_fsflags_map[i].iflag; 2174 2175 return iflags; 2176 } 2177 2178 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2179 { 2180 struct inode *inode = file_inode(filp); 2181 2182 return put_user(inode->i_generation, (int __user *)arg); 2183 } 2184 2185 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2186 { 2187 struct inode *inode = file_inode(filp); 2188 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2189 struct f2fs_inode_info *fi = F2FS_I(inode); 2190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2191 loff_t isize; 2192 int ret; 2193 2194 if (!(filp->f_mode & FMODE_WRITE)) 2195 return -EBADF; 2196 2197 if (!inode_owner_or_capable(idmap, inode)) 2198 return -EACCES; 2199 2200 if (!S_ISREG(inode->i_mode)) 2201 return -EINVAL; 2202 2203 if (filp->f_flags & O_DIRECT) 2204 return -EINVAL; 2205 2206 ret = mnt_want_write_file(filp); 2207 if (ret) 2208 return ret; 2209 2210 inode_lock(inode); 2211 2212 if (!f2fs_disable_compressed_file(inode) || 2213 f2fs_is_pinned_file(inode)) { 2214 ret = -EINVAL; 2215 goto out; 2216 } 2217 2218 if (f2fs_is_atomic_file(inode)) 2219 goto out; 2220 2221 ret = f2fs_convert_inline_inode(inode); 2222 if (ret) 2223 goto out; 2224 2225 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2226 f2fs_down_write(&fi->i_gc_rwsem[READ]); 2227 2228 /* 2229 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2230 * f2fs_is_atomic_file. 2231 */ 2232 if (get_dirty_pages(inode)) 2233 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2234 inode->i_ino, get_dirty_pages(inode)); 2235 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2236 if (ret) 2237 goto out_unlock; 2238 2239 /* Check if the inode already has a COW inode */ 2240 if (fi->cow_inode == NULL) { 2241 /* Create a COW inode for atomic write */ 2242 struct dentry *dentry = file_dentry(filp); 2243 struct inode *dir = d_inode(dentry->d_parent); 2244 2245 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode); 2246 if (ret) 2247 goto out_unlock; 2248 2249 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2250 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2251 2252 /* Set the COW inode's atomic_inode to the atomic inode */ 2253 F2FS_I(fi->cow_inode)->atomic_inode = inode; 2254 } else { 2255 /* Reuse the already created COW inode */ 2256 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode)); 2257 2258 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1); 2259 2260 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2261 if (ret) 2262 goto out_unlock; 2263 } 2264 2265 f2fs_write_inode(inode, NULL); 2266 2267 stat_inc_atomic_inode(inode); 2268 2269 set_inode_flag(inode, FI_ATOMIC_FILE); 2270 2271 isize = i_size_read(inode); 2272 fi->original_i_size = isize; 2273 if (truncate) { 2274 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2275 truncate_inode_pages_final(inode->i_mapping); 2276 f2fs_i_size_write(inode, 0); 2277 isize = 0; 2278 } 2279 f2fs_i_size_write(fi->cow_inode, isize); 2280 2281 out_unlock: 2282 f2fs_up_write(&fi->i_gc_rwsem[READ]); 2283 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2284 if (ret) 2285 goto out; 2286 2287 f2fs_update_time(sbi, REQ_TIME); 2288 fi->atomic_write_task = current; 2289 stat_update_max_atomic_write(inode); 2290 fi->atomic_write_cnt = 0; 2291 out: 2292 inode_unlock(inode); 2293 mnt_drop_write_file(filp); 2294 return ret; 2295 } 2296 2297 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2298 { 2299 struct inode *inode = file_inode(filp); 2300 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2301 int ret; 2302 2303 if (!(filp->f_mode & FMODE_WRITE)) 2304 return -EBADF; 2305 2306 if (!inode_owner_or_capable(idmap, inode)) 2307 return -EACCES; 2308 2309 ret = mnt_want_write_file(filp); 2310 if (ret) 2311 return ret; 2312 2313 f2fs_balance_fs(F2FS_I_SB(inode), true); 2314 2315 inode_lock(inode); 2316 2317 if (f2fs_is_atomic_file(inode)) { 2318 ret = f2fs_commit_atomic_write(inode); 2319 if (!ret) 2320 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2321 2322 f2fs_abort_atomic_write(inode, ret); 2323 } else { 2324 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2325 } 2326 2327 inode_unlock(inode); 2328 mnt_drop_write_file(filp); 2329 return ret; 2330 } 2331 2332 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2333 { 2334 struct inode *inode = file_inode(filp); 2335 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2336 int ret; 2337 2338 if (!(filp->f_mode & FMODE_WRITE)) 2339 return -EBADF; 2340 2341 if (!inode_owner_or_capable(idmap, inode)) 2342 return -EACCES; 2343 2344 ret = mnt_want_write_file(filp); 2345 if (ret) 2346 return ret; 2347 2348 inode_lock(inode); 2349 2350 f2fs_abort_atomic_write(inode, true); 2351 2352 inode_unlock(inode); 2353 2354 mnt_drop_write_file(filp); 2355 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2356 return ret; 2357 } 2358 2359 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag, 2360 bool readonly, bool need_lock) 2361 { 2362 struct super_block *sb = sbi->sb; 2363 int ret = 0; 2364 2365 switch (flag) { 2366 case F2FS_GOING_DOWN_FULLSYNC: 2367 ret = bdev_freeze(sb->s_bdev); 2368 if (ret) 2369 goto out; 2370 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2371 bdev_thaw(sb->s_bdev); 2372 break; 2373 case F2FS_GOING_DOWN_METASYNC: 2374 /* do checkpoint only */ 2375 ret = f2fs_sync_fs(sb, 1); 2376 if (ret) { 2377 if (ret == -EIO) 2378 ret = 0; 2379 goto out; 2380 } 2381 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2382 break; 2383 case F2FS_GOING_DOWN_NOSYNC: 2384 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2385 break; 2386 case F2FS_GOING_DOWN_METAFLUSH: 2387 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2388 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2389 break; 2390 case F2FS_GOING_DOWN_NEED_FSCK: 2391 set_sbi_flag(sbi, SBI_NEED_FSCK); 2392 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2393 set_sbi_flag(sbi, SBI_IS_DIRTY); 2394 /* do checkpoint only */ 2395 ret = f2fs_sync_fs(sb, 1); 2396 if (ret == -EIO) 2397 ret = 0; 2398 goto out; 2399 default: 2400 ret = -EINVAL; 2401 goto out; 2402 } 2403 2404 if (readonly) 2405 goto out; 2406 2407 /* 2408 * grab sb->s_umount to avoid racing w/ remount() and other shutdown 2409 * paths. 2410 */ 2411 if (need_lock) 2412 down_write(&sbi->sb->s_umount); 2413 2414 f2fs_stop_gc_thread(sbi); 2415 f2fs_stop_discard_thread(sbi); 2416 2417 f2fs_drop_discard_cmd(sbi); 2418 clear_opt(sbi, DISCARD); 2419 2420 if (need_lock) 2421 up_write(&sbi->sb->s_umount); 2422 2423 f2fs_update_time(sbi, REQ_TIME); 2424 out: 2425 2426 trace_f2fs_shutdown(sbi, flag, ret); 2427 2428 return ret; 2429 } 2430 2431 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2432 { 2433 struct inode *inode = file_inode(filp); 2434 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2435 __u32 in; 2436 int ret; 2437 bool need_drop = false, readonly = false; 2438 2439 if (!capable(CAP_SYS_ADMIN)) 2440 return -EPERM; 2441 2442 if (get_user(in, (__u32 __user *)arg)) 2443 return -EFAULT; 2444 2445 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2446 ret = mnt_want_write_file(filp); 2447 if (ret) { 2448 if (ret != -EROFS) 2449 return ret; 2450 2451 /* fallback to nosync shutdown for readonly fs */ 2452 in = F2FS_GOING_DOWN_NOSYNC; 2453 readonly = true; 2454 } else { 2455 need_drop = true; 2456 } 2457 } 2458 2459 ret = f2fs_do_shutdown(sbi, in, readonly, true); 2460 2461 if (need_drop) 2462 mnt_drop_write_file(filp); 2463 2464 return ret; 2465 } 2466 2467 static void f2fs_keep_noreuse_range(struct inode *inode, 2468 loff_t offset, loff_t len) 2469 { 2470 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2471 u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode)); 2472 u64 start, end; 2473 2474 if (!S_ISREG(inode->i_mode)) 2475 return; 2476 2477 if (offset >= max_bytes || len > max_bytes || 2478 (offset + len) > max_bytes) 2479 return; 2480 2481 start = offset >> PAGE_SHIFT; 2482 end = DIV_ROUND_UP(offset + len, PAGE_SIZE); 2483 2484 inode_lock(inode); 2485 if (f2fs_is_atomic_file(inode)) { 2486 inode_unlock(inode); 2487 return; 2488 } 2489 2490 spin_lock(&sbi->inode_lock[DONATE_INODE]); 2491 /* let's remove the range, if len = 0 */ 2492 if (!len) { 2493 if (!list_empty(&F2FS_I(inode)->gdonate_list)) { 2494 list_del_init(&F2FS_I(inode)->gdonate_list); 2495 sbi->donate_files--; 2496 } 2497 } else { 2498 if (list_empty(&F2FS_I(inode)->gdonate_list)) { 2499 list_add_tail(&F2FS_I(inode)->gdonate_list, 2500 &sbi->inode_list[DONATE_INODE]); 2501 sbi->donate_files++; 2502 } else { 2503 list_move_tail(&F2FS_I(inode)->gdonate_list, 2504 &sbi->inode_list[DONATE_INODE]); 2505 } 2506 F2FS_I(inode)->donate_start = start; 2507 F2FS_I(inode)->donate_end = end - 1; 2508 } 2509 spin_unlock(&sbi->inode_lock[DONATE_INODE]); 2510 inode_unlock(inode); 2511 } 2512 2513 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2514 { 2515 struct inode *inode = file_inode(filp); 2516 struct super_block *sb = inode->i_sb; 2517 struct fstrim_range range; 2518 int ret; 2519 2520 if (!capable(CAP_SYS_ADMIN)) 2521 return -EPERM; 2522 2523 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2524 return -EOPNOTSUPP; 2525 2526 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2527 sizeof(range))) 2528 return -EFAULT; 2529 2530 ret = mnt_want_write_file(filp); 2531 if (ret) 2532 return ret; 2533 2534 range.minlen = max((unsigned int)range.minlen, 2535 bdev_discard_granularity(sb->s_bdev)); 2536 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2537 mnt_drop_write_file(filp); 2538 if (ret < 0) 2539 return ret; 2540 2541 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2542 sizeof(range))) 2543 return -EFAULT; 2544 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2545 return 0; 2546 } 2547 2548 static bool uuid_is_nonzero(__u8 u[16]) 2549 { 2550 int i; 2551 2552 for (i = 0; i < 16; i++) 2553 if (u[i]) 2554 return true; 2555 return false; 2556 } 2557 2558 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2559 { 2560 struct inode *inode = file_inode(filp); 2561 int ret; 2562 2563 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2564 return -EOPNOTSUPP; 2565 2566 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2567 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2568 return ret; 2569 } 2570 2571 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2572 { 2573 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2574 return -EOPNOTSUPP; 2575 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2576 } 2577 2578 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2579 { 2580 struct inode *inode = file_inode(filp); 2581 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2582 u8 encrypt_pw_salt[16]; 2583 int err; 2584 2585 if (!f2fs_sb_has_encrypt(sbi)) 2586 return -EOPNOTSUPP; 2587 2588 err = mnt_want_write_file(filp); 2589 if (err) 2590 return err; 2591 2592 f2fs_down_write(&sbi->sb_lock); 2593 2594 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2595 goto got_it; 2596 2597 /* update superblock with uuid */ 2598 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2599 2600 err = f2fs_commit_super(sbi, false); 2601 if (err) { 2602 /* undo new data */ 2603 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2604 goto out_err; 2605 } 2606 got_it: 2607 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2608 out_err: 2609 f2fs_up_write(&sbi->sb_lock); 2610 mnt_drop_write_file(filp); 2611 2612 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2613 err = -EFAULT; 2614 2615 return err; 2616 } 2617 2618 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2619 unsigned long arg) 2620 { 2621 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2622 return -EOPNOTSUPP; 2623 2624 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2625 } 2626 2627 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2628 { 2629 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2630 return -EOPNOTSUPP; 2631 2632 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2633 } 2634 2635 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2636 { 2637 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2638 return -EOPNOTSUPP; 2639 2640 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2641 } 2642 2643 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2644 unsigned long arg) 2645 { 2646 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2647 return -EOPNOTSUPP; 2648 2649 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2650 } 2651 2652 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2653 unsigned long arg) 2654 { 2655 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2656 return -EOPNOTSUPP; 2657 2658 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2659 } 2660 2661 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2662 { 2663 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2664 return -EOPNOTSUPP; 2665 2666 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2667 } 2668 2669 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2670 { 2671 struct inode *inode = file_inode(filp); 2672 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2673 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2674 .no_bg_gc = false, 2675 .should_migrate_blocks = false, 2676 .nr_free_secs = 0 }; 2677 __u32 sync; 2678 int ret; 2679 2680 if (!capable(CAP_SYS_ADMIN)) 2681 return -EPERM; 2682 2683 if (get_user(sync, (__u32 __user *)arg)) 2684 return -EFAULT; 2685 2686 if (f2fs_readonly(sbi->sb)) 2687 return -EROFS; 2688 2689 ret = mnt_want_write_file(filp); 2690 if (ret) 2691 return ret; 2692 2693 if (!sync) { 2694 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2695 ret = -EBUSY; 2696 goto out; 2697 } 2698 } else { 2699 f2fs_down_write(&sbi->gc_lock); 2700 } 2701 2702 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2703 gc_control.err_gc_skipped = sync; 2704 stat_inc_gc_call_count(sbi, FOREGROUND); 2705 ret = f2fs_gc(sbi, &gc_control); 2706 out: 2707 mnt_drop_write_file(filp); 2708 return ret; 2709 } 2710 2711 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2712 { 2713 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2714 struct f2fs_gc_control gc_control = { 2715 .init_gc_type = range->sync ? FG_GC : BG_GC, 2716 .no_bg_gc = false, 2717 .should_migrate_blocks = false, 2718 .err_gc_skipped = range->sync, 2719 .nr_free_secs = 0 }; 2720 u64 end; 2721 int ret; 2722 2723 if (!capable(CAP_SYS_ADMIN)) 2724 return -EPERM; 2725 if (f2fs_readonly(sbi->sb)) 2726 return -EROFS; 2727 2728 end = range->start + range->len; 2729 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2730 end >= MAX_BLKADDR(sbi)) 2731 return -EINVAL; 2732 2733 ret = mnt_want_write_file(filp); 2734 if (ret) 2735 return ret; 2736 2737 do_more: 2738 if (!range->sync) { 2739 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2740 ret = -EBUSY; 2741 goto out; 2742 } 2743 } else { 2744 f2fs_down_write(&sbi->gc_lock); 2745 } 2746 2747 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2748 stat_inc_gc_call_count(sbi, FOREGROUND); 2749 ret = f2fs_gc(sbi, &gc_control); 2750 if (ret) { 2751 if (ret == -EBUSY) 2752 ret = -EAGAIN; 2753 goto out; 2754 } 2755 range->start += CAP_BLKS_PER_SEC(sbi); 2756 if (range->start <= end) 2757 goto do_more; 2758 out: 2759 mnt_drop_write_file(filp); 2760 return ret; 2761 } 2762 2763 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2764 { 2765 struct f2fs_gc_range range; 2766 2767 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2768 sizeof(range))) 2769 return -EFAULT; 2770 return __f2fs_ioc_gc_range(filp, &range); 2771 } 2772 2773 static int f2fs_ioc_write_checkpoint(struct file *filp) 2774 { 2775 struct inode *inode = file_inode(filp); 2776 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2777 int ret; 2778 2779 if (!capable(CAP_SYS_ADMIN)) 2780 return -EPERM; 2781 2782 if (f2fs_readonly(sbi->sb)) 2783 return -EROFS; 2784 2785 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2786 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2787 return -EINVAL; 2788 } 2789 2790 ret = mnt_want_write_file(filp); 2791 if (ret) 2792 return ret; 2793 2794 ret = f2fs_sync_fs(sbi->sb, 1); 2795 2796 mnt_drop_write_file(filp); 2797 return ret; 2798 } 2799 2800 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2801 struct file *filp, 2802 struct f2fs_defragment *range) 2803 { 2804 struct inode *inode = file_inode(filp); 2805 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2806 .m_seg_type = NO_CHECK_TYPE, 2807 .m_may_create = false }; 2808 struct extent_info ei = {}; 2809 pgoff_t pg_start, pg_end, next_pgofs; 2810 unsigned int total = 0, sec_num; 2811 block_t blk_end = 0; 2812 bool fragmented = false; 2813 int err; 2814 2815 f2fs_balance_fs(sbi, true); 2816 2817 inode_lock(inode); 2818 pg_start = range->start >> PAGE_SHIFT; 2819 pg_end = min_t(pgoff_t, 2820 (range->start + range->len) >> PAGE_SHIFT, 2821 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE)); 2822 2823 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) || 2824 f2fs_is_atomic_file(inode)) { 2825 err = -EINVAL; 2826 goto unlock_out; 2827 } 2828 2829 /* if in-place-update policy is enabled, don't waste time here */ 2830 set_inode_flag(inode, FI_OPU_WRITE); 2831 if (f2fs_should_update_inplace(inode, NULL)) { 2832 err = -EINVAL; 2833 goto out; 2834 } 2835 2836 /* writeback all dirty pages in the range */ 2837 err = filemap_write_and_wait_range(inode->i_mapping, 2838 pg_start << PAGE_SHIFT, 2839 (pg_end << PAGE_SHIFT) - 1); 2840 if (err) 2841 goto out; 2842 2843 /* 2844 * lookup mapping info in extent cache, skip defragmenting if physical 2845 * block addresses are continuous. 2846 */ 2847 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2848 if ((pgoff_t)ei.fofs + ei.len >= pg_end) 2849 goto out; 2850 } 2851 2852 map.m_lblk = pg_start; 2853 map.m_next_pgofs = &next_pgofs; 2854 2855 /* 2856 * lookup mapping info in dnode page cache, skip defragmenting if all 2857 * physical block addresses are continuous even if there are hole(s) 2858 * in logical blocks. 2859 */ 2860 while (map.m_lblk < pg_end) { 2861 map.m_len = pg_end - map.m_lblk; 2862 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2863 if (err) 2864 goto out; 2865 2866 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2867 map.m_lblk = next_pgofs; 2868 continue; 2869 } 2870 2871 if (blk_end && blk_end != map.m_pblk) 2872 fragmented = true; 2873 2874 /* record total count of block that we're going to move */ 2875 total += map.m_len; 2876 2877 blk_end = map.m_pblk + map.m_len; 2878 2879 map.m_lblk += map.m_len; 2880 } 2881 2882 if (!fragmented) { 2883 total = 0; 2884 goto out; 2885 } 2886 2887 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2888 2889 /* 2890 * make sure there are enough free section for LFS allocation, this can 2891 * avoid defragment running in SSR mode when free section are allocated 2892 * intensively 2893 */ 2894 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2895 err = -EAGAIN; 2896 goto out; 2897 } 2898 2899 map.m_lblk = pg_start; 2900 map.m_len = pg_end - pg_start; 2901 total = 0; 2902 2903 while (map.m_lblk < pg_end) { 2904 pgoff_t idx; 2905 int cnt = 0; 2906 2907 do_map: 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 clear_out; 2912 2913 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2914 map.m_lblk = next_pgofs; 2915 goto check; 2916 } 2917 2918 set_inode_flag(inode, FI_SKIP_WRITES); 2919 2920 idx = map.m_lblk; 2921 while (idx < map.m_lblk + map.m_len && 2922 cnt < BLKS_PER_SEG(sbi)) { 2923 struct page *page; 2924 2925 page = f2fs_get_lock_data_page(inode, idx, true); 2926 if (IS_ERR(page)) { 2927 err = PTR_ERR(page); 2928 goto clear_out; 2929 } 2930 2931 f2fs_wait_on_page_writeback(page, DATA, true, true); 2932 2933 set_page_dirty(page); 2934 set_page_private_gcing(page); 2935 f2fs_put_page(page, 1); 2936 2937 idx++; 2938 cnt++; 2939 total++; 2940 } 2941 2942 map.m_lblk = idx; 2943 check: 2944 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 2945 goto do_map; 2946 2947 clear_inode_flag(inode, FI_SKIP_WRITES); 2948 2949 err = filemap_fdatawrite(inode->i_mapping); 2950 if (err) 2951 goto out; 2952 } 2953 clear_out: 2954 clear_inode_flag(inode, FI_SKIP_WRITES); 2955 out: 2956 clear_inode_flag(inode, FI_OPU_WRITE); 2957 unlock_out: 2958 inode_unlock(inode); 2959 if (!err) 2960 range->len = (u64)total << PAGE_SHIFT; 2961 return err; 2962 } 2963 2964 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2965 { 2966 struct inode *inode = file_inode(filp); 2967 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2968 struct f2fs_defragment range; 2969 int err; 2970 2971 if (!capable(CAP_SYS_ADMIN)) 2972 return -EPERM; 2973 2974 if (!S_ISREG(inode->i_mode)) 2975 return -EINVAL; 2976 2977 if (f2fs_readonly(sbi->sb)) 2978 return -EROFS; 2979 2980 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2981 sizeof(range))) 2982 return -EFAULT; 2983 2984 /* verify alignment of offset & size */ 2985 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2986 return -EINVAL; 2987 2988 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2989 max_file_blocks(inode))) 2990 return -EINVAL; 2991 2992 err = mnt_want_write_file(filp); 2993 if (err) 2994 return err; 2995 2996 err = f2fs_defragment_range(sbi, filp, &range); 2997 mnt_drop_write_file(filp); 2998 2999 if (range.len) 3000 f2fs_update_time(sbi, REQ_TIME); 3001 if (err < 0) 3002 return err; 3003 3004 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 3005 sizeof(range))) 3006 return -EFAULT; 3007 3008 return 0; 3009 } 3010 3011 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 3012 struct file *file_out, loff_t pos_out, size_t len) 3013 { 3014 struct inode *src = file_inode(file_in); 3015 struct inode *dst = file_inode(file_out); 3016 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 3017 size_t olen = len, dst_max_i_size = 0; 3018 size_t dst_osize; 3019 int ret; 3020 3021 if (file_in->f_path.mnt != file_out->f_path.mnt || 3022 src->i_sb != dst->i_sb) 3023 return -EXDEV; 3024 3025 if (unlikely(f2fs_readonly(src->i_sb))) 3026 return -EROFS; 3027 3028 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 3029 return -EINVAL; 3030 3031 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 3032 return -EOPNOTSUPP; 3033 3034 if (pos_out < 0 || pos_in < 0) 3035 return -EINVAL; 3036 3037 if (src == dst) { 3038 if (pos_in == pos_out) 3039 return 0; 3040 if (pos_out > pos_in && pos_out < pos_in + len) 3041 return -EINVAL; 3042 } 3043 3044 inode_lock(src); 3045 if (src != dst) { 3046 ret = -EBUSY; 3047 if (!inode_trylock(dst)) 3048 goto out; 3049 } 3050 3051 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) || 3052 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 3053 ret = -EOPNOTSUPP; 3054 goto out_unlock; 3055 } 3056 3057 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) { 3058 ret = -EINVAL; 3059 goto out_unlock; 3060 } 3061 3062 ret = -EINVAL; 3063 if (pos_in + len > src->i_size || pos_in + len < pos_in) 3064 goto out_unlock; 3065 if (len == 0) 3066 olen = len = src->i_size - pos_in; 3067 if (pos_in + len == src->i_size) 3068 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 3069 if (len == 0) { 3070 ret = 0; 3071 goto out_unlock; 3072 } 3073 3074 dst_osize = dst->i_size; 3075 if (pos_out + olen > dst->i_size) 3076 dst_max_i_size = pos_out + olen; 3077 3078 /* verify the end result is block aligned */ 3079 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 3080 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 3081 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 3082 goto out_unlock; 3083 3084 ret = f2fs_convert_inline_inode(src); 3085 if (ret) 3086 goto out_unlock; 3087 3088 ret = f2fs_convert_inline_inode(dst); 3089 if (ret) 3090 goto out_unlock; 3091 3092 /* write out all dirty pages from offset */ 3093 ret = filemap_write_and_wait_range(src->i_mapping, 3094 pos_in, pos_in + len); 3095 if (ret) 3096 goto out_unlock; 3097 3098 ret = filemap_write_and_wait_range(dst->i_mapping, 3099 pos_out, pos_out + len); 3100 if (ret) 3101 goto out_unlock; 3102 3103 f2fs_balance_fs(sbi, true); 3104 3105 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3106 if (src != dst) { 3107 ret = -EBUSY; 3108 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 3109 goto out_src; 3110 } 3111 3112 f2fs_lock_op(sbi); 3113 ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in), 3114 F2FS_BYTES_TO_BLK(pos_out), 3115 F2FS_BYTES_TO_BLK(len), false); 3116 3117 if (!ret) { 3118 if (dst_max_i_size) 3119 f2fs_i_size_write(dst, dst_max_i_size); 3120 else if (dst_osize != dst->i_size) 3121 f2fs_i_size_write(dst, dst_osize); 3122 } 3123 f2fs_unlock_op(sbi); 3124 3125 if (src != dst) 3126 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 3127 out_src: 3128 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 3129 if (ret) 3130 goto out_unlock; 3131 3132 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 3133 f2fs_mark_inode_dirty_sync(src, false); 3134 if (src != dst) { 3135 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 3136 f2fs_mark_inode_dirty_sync(dst, false); 3137 } 3138 f2fs_update_time(sbi, REQ_TIME); 3139 3140 out_unlock: 3141 if (src != dst) 3142 inode_unlock(dst); 3143 out: 3144 inode_unlock(src); 3145 return ret; 3146 } 3147 3148 static int __f2fs_ioc_move_range(struct file *filp, 3149 struct f2fs_move_range *range) 3150 { 3151 int err; 3152 3153 if (!(filp->f_mode & FMODE_READ) || 3154 !(filp->f_mode & FMODE_WRITE)) 3155 return -EBADF; 3156 3157 CLASS(fd, dst)(range->dst_fd); 3158 if (fd_empty(dst)) 3159 return -EBADF; 3160 3161 if (!(fd_file(dst)->f_mode & FMODE_WRITE)) 3162 return -EBADF; 3163 3164 err = mnt_want_write_file(filp); 3165 if (err) 3166 return err; 3167 3168 err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst), 3169 range->pos_out, range->len); 3170 3171 mnt_drop_write_file(filp); 3172 return err; 3173 } 3174 3175 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 3176 { 3177 struct f2fs_move_range range; 3178 3179 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 3180 sizeof(range))) 3181 return -EFAULT; 3182 return __f2fs_ioc_move_range(filp, &range); 3183 } 3184 3185 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 3186 { 3187 struct inode *inode = file_inode(filp); 3188 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3189 struct sit_info *sm = SIT_I(sbi); 3190 unsigned int start_segno = 0, end_segno = 0; 3191 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3192 struct f2fs_flush_device range; 3193 struct f2fs_gc_control gc_control = { 3194 .init_gc_type = FG_GC, 3195 .should_migrate_blocks = true, 3196 .err_gc_skipped = true, 3197 .nr_free_secs = 0 }; 3198 int ret; 3199 3200 if (!capable(CAP_SYS_ADMIN)) 3201 return -EPERM; 3202 3203 if (f2fs_readonly(sbi->sb)) 3204 return -EROFS; 3205 3206 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3207 return -EINVAL; 3208 3209 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3210 sizeof(range))) 3211 return -EFAULT; 3212 3213 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3214 __is_large_section(sbi)) { 3215 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3216 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3217 return -EINVAL; 3218 } 3219 3220 ret = mnt_want_write_file(filp); 3221 if (ret) 3222 return ret; 3223 3224 if (range.dev_num != 0) 3225 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3226 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3227 3228 start_segno = sm->last_victim[FLUSH_DEVICE]; 3229 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3230 start_segno = dev_start_segno; 3231 end_segno = min(start_segno + range.segments, dev_end_segno); 3232 3233 while (start_segno < end_segno) { 3234 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3235 ret = -EBUSY; 3236 goto out; 3237 } 3238 sm->last_victim[GC_CB] = end_segno + 1; 3239 sm->last_victim[GC_GREEDY] = end_segno + 1; 3240 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3241 3242 gc_control.victim_segno = start_segno; 3243 stat_inc_gc_call_count(sbi, FOREGROUND); 3244 ret = f2fs_gc(sbi, &gc_control); 3245 if (ret == -EAGAIN) 3246 ret = 0; 3247 else if (ret < 0) 3248 break; 3249 start_segno++; 3250 } 3251 out: 3252 mnt_drop_write_file(filp); 3253 return ret; 3254 } 3255 3256 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3257 { 3258 struct inode *inode = file_inode(filp); 3259 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3260 3261 /* Must validate to set it with SQLite behavior in Android. */ 3262 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3263 3264 return put_user(sb_feature, (u32 __user *)arg); 3265 } 3266 3267 #ifdef CONFIG_QUOTA 3268 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3269 { 3270 struct dquot *transfer_to[MAXQUOTAS] = {}; 3271 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3272 struct super_block *sb = sbi->sb; 3273 int err; 3274 3275 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3276 if (IS_ERR(transfer_to[PRJQUOTA])) 3277 return PTR_ERR(transfer_to[PRJQUOTA]); 3278 3279 err = __dquot_transfer(inode, transfer_to); 3280 if (err) 3281 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3282 dqput(transfer_to[PRJQUOTA]); 3283 return err; 3284 } 3285 3286 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3287 { 3288 struct f2fs_inode_info *fi = F2FS_I(inode); 3289 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3290 struct f2fs_inode *ri = NULL; 3291 kprojid_t kprojid; 3292 int err; 3293 3294 if (!f2fs_sb_has_project_quota(sbi)) { 3295 if (projid != F2FS_DEF_PROJID) 3296 return -EOPNOTSUPP; 3297 else 3298 return 0; 3299 } 3300 3301 if (!f2fs_has_extra_attr(inode)) 3302 return -EOPNOTSUPP; 3303 3304 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3305 3306 if (projid_eq(kprojid, fi->i_projid)) 3307 return 0; 3308 3309 err = -EPERM; 3310 /* Is it quota file? Do not allow user to mess with it */ 3311 if (IS_NOQUOTA(inode)) 3312 return err; 3313 3314 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3315 return -EOVERFLOW; 3316 3317 err = f2fs_dquot_initialize(inode); 3318 if (err) 3319 return err; 3320 3321 f2fs_lock_op(sbi); 3322 err = f2fs_transfer_project_quota(inode, kprojid); 3323 if (err) 3324 goto out_unlock; 3325 3326 fi->i_projid = kprojid; 3327 inode_set_ctime_current(inode); 3328 f2fs_mark_inode_dirty_sync(inode, true); 3329 out_unlock: 3330 f2fs_unlock_op(sbi); 3331 return err; 3332 } 3333 #else 3334 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3335 { 3336 return 0; 3337 } 3338 3339 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3340 { 3341 if (projid != F2FS_DEF_PROJID) 3342 return -EOPNOTSUPP; 3343 return 0; 3344 } 3345 #endif 3346 3347 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3348 { 3349 struct inode *inode = d_inode(dentry); 3350 struct f2fs_inode_info *fi = F2FS_I(inode); 3351 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3352 3353 if (IS_ENCRYPTED(inode)) 3354 fsflags |= FS_ENCRYPT_FL; 3355 if (IS_VERITY(inode)) 3356 fsflags |= FS_VERITY_FL; 3357 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3358 fsflags |= FS_INLINE_DATA_FL; 3359 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3360 fsflags |= FS_NOCOW_FL; 3361 3362 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3363 3364 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3365 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3366 3367 return 0; 3368 } 3369 3370 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3371 struct dentry *dentry, struct fileattr *fa) 3372 { 3373 struct inode *inode = d_inode(dentry); 3374 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3375 u32 iflags; 3376 int err; 3377 3378 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3379 return -EIO; 3380 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3381 return -ENOSPC; 3382 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3383 return -EOPNOTSUPP; 3384 fsflags &= F2FS_SETTABLE_FS_FL; 3385 if (!fa->flags_valid) 3386 mask &= FS_COMMON_FL; 3387 3388 iflags = f2fs_fsflags_to_iflags(fsflags); 3389 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3390 return -EOPNOTSUPP; 3391 3392 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3393 if (!err) 3394 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3395 3396 return err; 3397 } 3398 3399 int f2fs_pin_file_control(struct inode *inode, bool inc) 3400 { 3401 struct f2fs_inode_info *fi = F2FS_I(inode); 3402 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3403 3404 if (IS_DEVICE_ALIASING(inode)) 3405 return -EINVAL; 3406 3407 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) { 3408 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3409 __func__, inode->i_ino, fi->i_gc_failures); 3410 clear_inode_flag(inode, FI_PIN_FILE); 3411 return -EAGAIN; 3412 } 3413 3414 /* Use i_gc_failures for normal file as a risk signal. */ 3415 if (inc) 3416 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1); 3417 3418 return 0; 3419 } 3420 3421 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3422 { 3423 struct inode *inode = file_inode(filp); 3424 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3425 __u32 pin; 3426 int ret = 0; 3427 3428 if (get_user(pin, (__u32 __user *)arg)) 3429 return -EFAULT; 3430 3431 if (!S_ISREG(inode->i_mode)) 3432 return -EINVAL; 3433 3434 if (f2fs_readonly(sbi->sb)) 3435 return -EROFS; 3436 3437 if (!pin && IS_DEVICE_ALIASING(inode)) 3438 return -EOPNOTSUPP; 3439 3440 ret = mnt_want_write_file(filp); 3441 if (ret) 3442 return ret; 3443 3444 inode_lock(inode); 3445 3446 if (f2fs_is_atomic_file(inode)) { 3447 ret = -EINVAL; 3448 goto out; 3449 } 3450 3451 if (!pin) { 3452 clear_inode_flag(inode, FI_PIN_FILE); 3453 f2fs_i_gc_failures_write(inode, 0); 3454 goto done; 3455 } else if (f2fs_is_pinned_file(inode)) { 3456 goto done; 3457 } 3458 3459 if (F2FS_HAS_BLOCKS(inode)) { 3460 ret = -EFBIG; 3461 goto out; 3462 } 3463 3464 /* Let's allow file pinning on zoned device. */ 3465 if (!f2fs_sb_has_blkzoned(sbi) && 3466 f2fs_should_update_outplace(inode, NULL)) { 3467 ret = -EINVAL; 3468 goto out; 3469 } 3470 3471 if (f2fs_pin_file_control(inode, false)) { 3472 ret = -EAGAIN; 3473 goto out; 3474 } 3475 3476 ret = f2fs_convert_inline_inode(inode); 3477 if (ret) 3478 goto out; 3479 3480 if (!f2fs_disable_compressed_file(inode)) { 3481 ret = -EOPNOTSUPP; 3482 goto out; 3483 } 3484 3485 set_inode_flag(inode, FI_PIN_FILE); 3486 ret = F2FS_I(inode)->i_gc_failures; 3487 done: 3488 f2fs_update_time(sbi, REQ_TIME); 3489 out: 3490 inode_unlock(inode); 3491 mnt_drop_write_file(filp); 3492 return ret; 3493 } 3494 3495 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3496 { 3497 struct inode *inode = file_inode(filp); 3498 __u32 pin = 0; 3499 3500 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3501 pin = F2FS_I(inode)->i_gc_failures; 3502 return put_user(pin, (u32 __user *)arg); 3503 } 3504 3505 static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg) 3506 { 3507 return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0, 3508 (u32 __user *)arg); 3509 } 3510 3511 static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg) 3512 { 3513 struct inode *inode = file_inode(filp); 3514 __u32 level; 3515 3516 if (get_user(level, (__u32 __user *)arg)) 3517 return -EFAULT; 3518 3519 if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX) 3520 return -EINVAL; 3521 3522 inode_lock(inode); 3523 F2FS_I(inode)->ioprio_hint = level; 3524 inode_unlock(inode); 3525 return 0; 3526 } 3527 3528 int f2fs_precache_extents(struct inode *inode) 3529 { 3530 struct f2fs_inode_info *fi = F2FS_I(inode); 3531 struct f2fs_map_blocks map; 3532 pgoff_t m_next_extent; 3533 loff_t end; 3534 int err; 3535 3536 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3537 return -EOPNOTSUPP; 3538 3539 map.m_lblk = 0; 3540 map.m_pblk = 0; 3541 map.m_next_pgofs = NULL; 3542 map.m_next_extent = &m_next_extent; 3543 map.m_seg_type = NO_CHECK_TYPE; 3544 map.m_may_create = false; 3545 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3546 3547 while (map.m_lblk < end) { 3548 map.m_len = end - map.m_lblk; 3549 3550 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3551 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3552 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3553 if (err || !map.m_len) 3554 return err; 3555 3556 map.m_lblk = m_next_extent; 3557 } 3558 3559 return 0; 3560 } 3561 3562 static int f2fs_ioc_precache_extents(struct file *filp) 3563 { 3564 return f2fs_precache_extents(file_inode(filp)); 3565 } 3566 3567 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3568 { 3569 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3570 __u64 block_count; 3571 3572 if (!capable(CAP_SYS_ADMIN)) 3573 return -EPERM; 3574 3575 if (f2fs_readonly(sbi->sb)) 3576 return -EROFS; 3577 3578 if (copy_from_user(&block_count, (void __user *)arg, 3579 sizeof(block_count))) 3580 return -EFAULT; 3581 3582 return f2fs_resize_fs(filp, block_count); 3583 } 3584 3585 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3586 { 3587 struct inode *inode = file_inode(filp); 3588 3589 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3590 3591 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3592 f2fs_warn(F2FS_I_SB(inode), 3593 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3594 inode->i_ino); 3595 return -EOPNOTSUPP; 3596 } 3597 3598 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3599 } 3600 3601 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3602 { 3603 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3604 return -EOPNOTSUPP; 3605 3606 return fsverity_ioctl_measure(filp, (void __user *)arg); 3607 } 3608 3609 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3610 { 3611 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3612 return -EOPNOTSUPP; 3613 3614 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3615 } 3616 3617 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3618 { 3619 struct inode *inode = file_inode(filp); 3620 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3621 char *vbuf; 3622 int count; 3623 int err = 0; 3624 3625 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3626 if (!vbuf) 3627 return -ENOMEM; 3628 3629 f2fs_down_read(&sbi->sb_lock); 3630 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3631 ARRAY_SIZE(sbi->raw_super->volume_name), 3632 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3633 f2fs_up_read(&sbi->sb_lock); 3634 3635 if (copy_to_user((char __user *)arg, vbuf, 3636 min(FSLABEL_MAX, count))) 3637 err = -EFAULT; 3638 3639 kfree(vbuf); 3640 return err; 3641 } 3642 3643 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3644 { 3645 struct inode *inode = file_inode(filp); 3646 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3647 char *vbuf; 3648 int err = 0; 3649 3650 if (!capable(CAP_SYS_ADMIN)) 3651 return -EPERM; 3652 3653 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3654 if (IS_ERR(vbuf)) 3655 return PTR_ERR(vbuf); 3656 3657 err = mnt_want_write_file(filp); 3658 if (err) 3659 goto out; 3660 3661 f2fs_down_write(&sbi->sb_lock); 3662 3663 memset(sbi->raw_super->volume_name, 0, 3664 sizeof(sbi->raw_super->volume_name)); 3665 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3666 sbi->raw_super->volume_name, 3667 ARRAY_SIZE(sbi->raw_super->volume_name)); 3668 3669 err = f2fs_commit_super(sbi, false); 3670 3671 f2fs_up_write(&sbi->sb_lock); 3672 3673 mnt_drop_write_file(filp); 3674 out: 3675 kfree(vbuf); 3676 return err; 3677 } 3678 3679 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3680 { 3681 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3682 return -EOPNOTSUPP; 3683 3684 if (!f2fs_compressed_file(inode)) 3685 return -EINVAL; 3686 3687 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3688 3689 return 0; 3690 } 3691 3692 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3693 { 3694 struct inode *inode = file_inode(filp); 3695 __u64 blocks; 3696 int ret; 3697 3698 ret = f2fs_get_compress_blocks(inode, &blocks); 3699 if (ret < 0) 3700 return ret; 3701 3702 return put_user(blocks, (u64 __user *)arg); 3703 } 3704 3705 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3706 { 3707 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3708 unsigned int released_blocks = 0; 3709 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3710 block_t blkaddr; 3711 int i; 3712 3713 for (i = 0; i < count; i++) { 3714 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3715 dn->ofs_in_node + i); 3716 3717 if (!__is_valid_data_blkaddr(blkaddr)) 3718 continue; 3719 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3720 DATA_GENERIC_ENHANCE))) 3721 return -EFSCORRUPTED; 3722 } 3723 3724 while (count) { 3725 int compr_blocks = 0; 3726 3727 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3728 blkaddr = f2fs_data_blkaddr(dn); 3729 3730 if (i == 0) { 3731 if (blkaddr == COMPRESS_ADDR) 3732 continue; 3733 dn->ofs_in_node += cluster_size; 3734 goto next; 3735 } 3736 3737 if (__is_valid_data_blkaddr(blkaddr)) 3738 compr_blocks++; 3739 3740 if (blkaddr != NEW_ADDR) 3741 continue; 3742 3743 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3744 } 3745 3746 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3747 dec_valid_block_count(sbi, dn->inode, 3748 cluster_size - compr_blocks); 3749 3750 released_blocks += cluster_size - compr_blocks; 3751 next: 3752 count -= cluster_size; 3753 } 3754 3755 return released_blocks; 3756 } 3757 3758 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3759 { 3760 struct inode *inode = file_inode(filp); 3761 struct f2fs_inode_info *fi = F2FS_I(inode); 3762 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3763 pgoff_t page_idx = 0, last_idx; 3764 unsigned int released_blocks = 0; 3765 int ret; 3766 int writecount; 3767 3768 if (!f2fs_sb_has_compression(sbi)) 3769 return -EOPNOTSUPP; 3770 3771 if (f2fs_readonly(sbi->sb)) 3772 return -EROFS; 3773 3774 ret = mnt_want_write_file(filp); 3775 if (ret) 3776 return ret; 3777 3778 f2fs_balance_fs(sbi, true); 3779 3780 inode_lock(inode); 3781 3782 writecount = atomic_read(&inode->i_writecount); 3783 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3784 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3785 ret = -EBUSY; 3786 goto out; 3787 } 3788 3789 if (!f2fs_compressed_file(inode) || 3790 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3791 ret = -EINVAL; 3792 goto out; 3793 } 3794 3795 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3796 if (ret) 3797 goto out; 3798 3799 if (!atomic_read(&fi->i_compr_blocks)) { 3800 ret = -EPERM; 3801 goto out; 3802 } 3803 3804 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3805 inode_set_ctime_current(inode); 3806 f2fs_mark_inode_dirty_sync(inode, true); 3807 3808 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3809 filemap_invalidate_lock(inode->i_mapping); 3810 3811 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3812 3813 while (page_idx < last_idx) { 3814 struct dnode_of_data dn; 3815 pgoff_t end_offset, count; 3816 3817 f2fs_lock_op(sbi); 3818 3819 set_new_dnode(&dn, inode, NULL, NULL, 0); 3820 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3821 if (ret) { 3822 f2fs_unlock_op(sbi); 3823 if (ret == -ENOENT) { 3824 page_idx = f2fs_get_next_page_offset(&dn, 3825 page_idx); 3826 ret = 0; 3827 continue; 3828 } 3829 break; 3830 } 3831 3832 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3833 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3834 count = round_up(count, fi->i_cluster_size); 3835 3836 ret = release_compress_blocks(&dn, count); 3837 3838 f2fs_put_dnode(&dn); 3839 3840 f2fs_unlock_op(sbi); 3841 3842 if (ret < 0) 3843 break; 3844 3845 page_idx += count; 3846 released_blocks += ret; 3847 } 3848 3849 filemap_invalidate_unlock(inode->i_mapping); 3850 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3851 out: 3852 if (released_blocks) 3853 f2fs_update_time(sbi, REQ_TIME); 3854 inode_unlock(inode); 3855 3856 mnt_drop_write_file(filp); 3857 3858 if (ret >= 0) { 3859 ret = put_user(released_blocks, (u64 __user *)arg); 3860 } else if (released_blocks && 3861 atomic_read(&fi->i_compr_blocks)) { 3862 set_sbi_flag(sbi, SBI_NEED_FSCK); 3863 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3864 "iblocks=%llu, released=%u, compr_blocks=%u, " 3865 "run fsck to fix.", 3866 __func__, inode->i_ino, inode->i_blocks, 3867 released_blocks, 3868 atomic_read(&fi->i_compr_blocks)); 3869 } 3870 3871 return ret; 3872 } 3873 3874 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3875 unsigned int *reserved_blocks) 3876 { 3877 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3878 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3879 block_t blkaddr; 3880 int i; 3881 3882 for (i = 0; i < count; i++) { 3883 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3884 dn->ofs_in_node + i); 3885 3886 if (!__is_valid_data_blkaddr(blkaddr)) 3887 continue; 3888 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3889 DATA_GENERIC_ENHANCE))) 3890 return -EFSCORRUPTED; 3891 } 3892 3893 while (count) { 3894 int compr_blocks = 0; 3895 blkcnt_t reserved = 0; 3896 blkcnt_t to_reserved; 3897 int ret; 3898 3899 for (i = 0; i < cluster_size; i++) { 3900 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3901 dn->ofs_in_node + i); 3902 3903 if (i == 0) { 3904 if (blkaddr != COMPRESS_ADDR) { 3905 dn->ofs_in_node += cluster_size; 3906 goto next; 3907 } 3908 continue; 3909 } 3910 3911 /* 3912 * compressed cluster was not released due to it 3913 * fails in release_compress_blocks(), so NEW_ADDR 3914 * is a possible case. 3915 */ 3916 if (blkaddr == NEW_ADDR) { 3917 reserved++; 3918 continue; 3919 } 3920 if (__is_valid_data_blkaddr(blkaddr)) { 3921 compr_blocks++; 3922 continue; 3923 } 3924 } 3925 3926 to_reserved = cluster_size - compr_blocks - reserved; 3927 3928 /* for the case all blocks in cluster were reserved */ 3929 if (reserved && to_reserved == 1) { 3930 dn->ofs_in_node += cluster_size; 3931 goto next; 3932 } 3933 3934 ret = inc_valid_block_count(sbi, dn->inode, 3935 &to_reserved, false); 3936 if (unlikely(ret)) 3937 return ret; 3938 3939 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3940 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3941 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3942 } 3943 3944 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3945 3946 *reserved_blocks += to_reserved; 3947 next: 3948 count -= cluster_size; 3949 } 3950 3951 return 0; 3952 } 3953 3954 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3955 { 3956 struct inode *inode = file_inode(filp); 3957 struct f2fs_inode_info *fi = F2FS_I(inode); 3958 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3959 pgoff_t page_idx = 0, last_idx; 3960 unsigned int reserved_blocks = 0; 3961 int ret; 3962 3963 if (!f2fs_sb_has_compression(sbi)) 3964 return -EOPNOTSUPP; 3965 3966 if (f2fs_readonly(sbi->sb)) 3967 return -EROFS; 3968 3969 ret = mnt_want_write_file(filp); 3970 if (ret) 3971 return ret; 3972 3973 f2fs_balance_fs(sbi, true); 3974 3975 inode_lock(inode); 3976 3977 if (!f2fs_compressed_file(inode) || 3978 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3979 ret = -EINVAL; 3980 goto unlock_inode; 3981 } 3982 3983 if (atomic_read(&fi->i_compr_blocks)) 3984 goto unlock_inode; 3985 3986 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3987 filemap_invalidate_lock(inode->i_mapping); 3988 3989 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3990 3991 while (page_idx < last_idx) { 3992 struct dnode_of_data dn; 3993 pgoff_t end_offset, count; 3994 3995 f2fs_lock_op(sbi); 3996 3997 set_new_dnode(&dn, inode, NULL, NULL, 0); 3998 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3999 if (ret) { 4000 f2fs_unlock_op(sbi); 4001 if (ret == -ENOENT) { 4002 page_idx = f2fs_get_next_page_offset(&dn, 4003 page_idx); 4004 ret = 0; 4005 continue; 4006 } 4007 break; 4008 } 4009 4010 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 4011 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 4012 count = round_up(count, fi->i_cluster_size); 4013 4014 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 4015 4016 f2fs_put_dnode(&dn); 4017 4018 f2fs_unlock_op(sbi); 4019 4020 if (ret < 0) 4021 break; 4022 4023 page_idx += count; 4024 } 4025 4026 filemap_invalidate_unlock(inode->i_mapping); 4027 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 4028 4029 if (!ret) { 4030 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 4031 inode_set_ctime_current(inode); 4032 f2fs_mark_inode_dirty_sync(inode, true); 4033 } 4034 unlock_inode: 4035 if (reserved_blocks) 4036 f2fs_update_time(sbi, REQ_TIME); 4037 inode_unlock(inode); 4038 mnt_drop_write_file(filp); 4039 4040 if (!ret) { 4041 ret = put_user(reserved_blocks, (u64 __user *)arg); 4042 } else if (reserved_blocks && 4043 atomic_read(&fi->i_compr_blocks)) { 4044 set_sbi_flag(sbi, SBI_NEED_FSCK); 4045 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx " 4046 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 4047 "run fsck to fix.", 4048 __func__, inode->i_ino, inode->i_blocks, 4049 reserved_blocks, 4050 atomic_read(&fi->i_compr_blocks)); 4051 } 4052 4053 return ret; 4054 } 4055 4056 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 4057 pgoff_t off, block_t block, block_t len, u32 flags) 4058 { 4059 sector_t sector = SECTOR_FROM_BLOCK(block); 4060 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 4061 int ret = 0; 4062 4063 if (flags & F2FS_TRIM_FILE_DISCARD) { 4064 if (bdev_max_secure_erase_sectors(bdev)) 4065 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 4066 GFP_NOFS); 4067 else 4068 ret = blkdev_issue_discard(bdev, sector, nr_sects, 4069 GFP_NOFS); 4070 } 4071 4072 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 4073 if (IS_ENCRYPTED(inode)) 4074 ret = fscrypt_zeroout_range(inode, off, block, len); 4075 else 4076 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 4077 GFP_NOFS, 0); 4078 } 4079 4080 return ret; 4081 } 4082 4083 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 4084 { 4085 struct inode *inode = file_inode(filp); 4086 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4087 struct address_space *mapping = inode->i_mapping; 4088 struct block_device *prev_bdev = NULL; 4089 struct f2fs_sectrim_range range; 4090 pgoff_t index, pg_end, prev_index = 0; 4091 block_t prev_block = 0, len = 0; 4092 loff_t end_addr; 4093 bool to_end = false; 4094 int ret = 0; 4095 4096 if (!(filp->f_mode & FMODE_WRITE)) 4097 return -EBADF; 4098 4099 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 4100 sizeof(range))) 4101 return -EFAULT; 4102 4103 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 4104 !S_ISREG(inode->i_mode)) 4105 return -EINVAL; 4106 4107 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 4108 !f2fs_hw_support_discard(sbi)) || 4109 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 4110 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 4111 return -EOPNOTSUPP; 4112 4113 ret = mnt_want_write_file(filp); 4114 if (ret) 4115 return ret; 4116 inode_lock(inode); 4117 4118 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 4119 range.start >= inode->i_size) { 4120 ret = -EINVAL; 4121 goto err; 4122 } 4123 4124 if (range.len == 0) 4125 goto err; 4126 4127 if (inode->i_size - range.start > range.len) { 4128 end_addr = range.start + range.len; 4129 } else { 4130 end_addr = range.len == (u64)-1 ? 4131 sbi->sb->s_maxbytes : inode->i_size; 4132 to_end = true; 4133 } 4134 4135 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 4136 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 4137 ret = -EINVAL; 4138 goto err; 4139 } 4140 4141 index = F2FS_BYTES_TO_BLK(range.start); 4142 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 4143 4144 ret = f2fs_convert_inline_inode(inode); 4145 if (ret) 4146 goto err; 4147 4148 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4149 filemap_invalidate_lock(mapping); 4150 4151 ret = filemap_write_and_wait_range(mapping, range.start, 4152 to_end ? LLONG_MAX : end_addr - 1); 4153 if (ret) 4154 goto out; 4155 4156 truncate_inode_pages_range(mapping, range.start, 4157 to_end ? -1 : end_addr - 1); 4158 4159 while (index < pg_end) { 4160 struct dnode_of_data dn; 4161 pgoff_t end_offset, count; 4162 int i; 4163 4164 set_new_dnode(&dn, inode, NULL, NULL, 0); 4165 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 4166 if (ret) { 4167 if (ret == -ENOENT) { 4168 index = f2fs_get_next_page_offset(&dn, index); 4169 continue; 4170 } 4171 goto out; 4172 } 4173 4174 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 4175 count = min(end_offset - dn.ofs_in_node, pg_end - index); 4176 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 4177 struct block_device *cur_bdev; 4178 block_t blkaddr = f2fs_data_blkaddr(&dn); 4179 4180 if (!__is_valid_data_blkaddr(blkaddr)) 4181 continue; 4182 4183 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 4184 DATA_GENERIC_ENHANCE)) { 4185 ret = -EFSCORRUPTED; 4186 f2fs_put_dnode(&dn); 4187 goto out; 4188 } 4189 4190 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 4191 if (f2fs_is_multi_device(sbi)) { 4192 int di = f2fs_target_device_index(sbi, blkaddr); 4193 4194 blkaddr -= FDEV(di).start_blk; 4195 } 4196 4197 if (len) { 4198 if (prev_bdev == cur_bdev && 4199 index == prev_index + len && 4200 blkaddr == prev_block + len) { 4201 len++; 4202 } else { 4203 ret = f2fs_secure_erase(prev_bdev, 4204 inode, prev_index, prev_block, 4205 len, range.flags); 4206 if (ret) { 4207 f2fs_put_dnode(&dn); 4208 goto out; 4209 } 4210 4211 len = 0; 4212 } 4213 } 4214 4215 if (!len) { 4216 prev_bdev = cur_bdev; 4217 prev_index = index; 4218 prev_block = blkaddr; 4219 len = 1; 4220 } 4221 } 4222 4223 f2fs_put_dnode(&dn); 4224 4225 if (fatal_signal_pending(current)) { 4226 ret = -EINTR; 4227 goto out; 4228 } 4229 cond_resched(); 4230 } 4231 4232 if (len) 4233 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 4234 prev_block, len, range.flags); 4235 f2fs_update_time(sbi, REQ_TIME); 4236 out: 4237 filemap_invalidate_unlock(mapping); 4238 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4239 err: 4240 inode_unlock(inode); 4241 mnt_drop_write_file(filp); 4242 4243 return ret; 4244 } 4245 4246 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 4247 { 4248 struct inode *inode = file_inode(filp); 4249 struct f2fs_comp_option option; 4250 4251 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4252 return -EOPNOTSUPP; 4253 4254 inode_lock_shared(inode); 4255 4256 if (!f2fs_compressed_file(inode)) { 4257 inode_unlock_shared(inode); 4258 return -ENODATA; 4259 } 4260 4261 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4262 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4263 4264 inode_unlock_shared(inode); 4265 4266 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4267 sizeof(option))) 4268 return -EFAULT; 4269 4270 return 0; 4271 } 4272 4273 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4274 { 4275 struct inode *inode = file_inode(filp); 4276 struct f2fs_inode_info *fi = F2FS_I(inode); 4277 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4278 struct f2fs_comp_option option; 4279 int ret = 0; 4280 4281 if (!f2fs_sb_has_compression(sbi)) 4282 return -EOPNOTSUPP; 4283 4284 if (!(filp->f_mode & FMODE_WRITE)) 4285 return -EBADF; 4286 4287 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4288 sizeof(option))) 4289 return -EFAULT; 4290 4291 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4292 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4293 option.algorithm >= COMPRESS_MAX) 4294 return -EINVAL; 4295 4296 ret = mnt_want_write_file(filp); 4297 if (ret) 4298 return ret; 4299 inode_lock(inode); 4300 4301 f2fs_down_write(&F2FS_I(inode)->i_sem); 4302 if (!f2fs_compressed_file(inode)) { 4303 ret = -EINVAL; 4304 goto out; 4305 } 4306 4307 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4308 ret = -EBUSY; 4309 goto out; 4310 } 4311 4312 if (F2FS_HAS_BLOCKS(inode)) { 4313 ret = -EFBIG; 4314 goto out; 4315 } 4316 4317 fi->i_compress_algorithm = option.algorithm; 4318 fi->i_log_cluster_size = option.log_cluster_size; 4319 fi->i_cluster_size = BIT(option.log_cluster_size); 4320 /* Set default level */ 4321 if (fi->i_compress_algorithm == COMPRESS_ZSTD) 4322 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4323 else 4324 fi->i_compress_level = 0; 4325 /* Adjust mount option level */ 4326 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4327 F2FS_OPTION(sbi).compress_level) 4328 fi->i_compress_level = F2FS_OPTION(sbi).compress_level; 4329 f2fs_mark_inode_dirty_sync(inode, true); 4330 4331 if (!f2fs_is_compress_backend_ready(inode)) 4332 f2fs_warn(sbi, "compression algorithm is successfully set, " 4333 "but current kernel doesn't support this algorithm."); 4334 out: 4335 f2fs_up_write(&fi->i_sem); 4336 inode_unlock(inode); 4337 mnt_drop_write_file(filp); 4338 4339 return ret; 4340 } 4341 4342 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4343 { 4344 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4345 struct address_space *mapping = inode->i_mapping; 4346 struct page *page; 4347 pgoff_t redirty_idx = page_idx; 4348 int i, page_len = 0, ret = 0; 4349 4350 page_cache_ra_unbounded(&ractl, len, 0); 4351 4352 for (i = 0; i < len; i++, page_idx++) { 4353 page = read_cache_page(mapping, page_idx, NULL, NULL); 4354 if (IS_ERR(page)) { 4355 ret = PTR_ERR(page); 4356 break; 4357 } 4358 page_len++; 4359 } 4360 4361 for (i = 0; i < page_len; i++, redirty_idx++) { 4362 page = find_lock_page(mapping, redirty_idx); 4363 4364 /* It will never fail, when page has pinned above */ 4365 f2fs_bug_on(F2FS_I_SB(inode), !page); 4366 4367 f2fs_wait_on_page_writeback(page, DATA, true, true); 4368 4369 set_page_dirty(page); 4370 set_page_private_gcing(page); 4371 f2fs_put_page(page, 1); 4372 f2fs_put_page(page, 0); 4373 } 4374 4375 return ret; 4376 } 4377 4378 static int f2fs_ioc_decompress_file(struct file *filp) 4379 { 4380 struct inode *inode = file_inode(filp); 4381 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4382 struct f2fs_inode_info *fi = F2FS_I(inode); 4383 pgoff_t page_idx = 0, last_idx, cluster_idx; 4384 int ret; 4385 4386 if (!f2fs_sb_has_compression(sbi) || 4387 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4388 return -EOPNOTSUPP; 4389 4390 if (!(filp->f_mode & FMODE_WRITE)) 4391 return -EBADF; 4392 4393 f2fs_balance_fs(sbi, true); 4394 4395 ret = mnt_want_write_file(filp); 4396 if (ret) 4397 return ret; 4398 inode_lock(inode); 4399 4400 if (!f2fs_is_compress_backend_ready(inode)) { 4401 ret = -EOPNOTSUPP; 4402 goto out; 4403 } 4404 4405 if (!f2fs_compressed_file(inode) || 4406 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4407 ret = -EINVAL; 4408 goto out; 4409 } 4410 4411 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4412 if (ret) 4413 goto out; 4414 4415 if (!atomic_read(&fi->i_compr_blocks)) 4416 goto out; 4417 4418 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4419 last_idx >>= fi->i_log_cluster_size; 4420 4421 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4422 page_idx = cluster_idx << fi->i_log_cluster_size; 4423 4424 if (!f2fs_is_compressed_cluster(inode, page_idx)) 4425 continue; 4426 4427 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4428 if (ret < 0) 4429 break; 4430 4431 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4432 ret = filemap_fdatawrite(inode->i_mapping); 4433 if (ret < 0) 4434 break; 4435 } 4436 4437 cond_resched(); 4438 if (fatal_signal_pending(current)) { 4439 ret = -EINTR; 4440 break; 4441 } 4442 } 4443 4444 if (!ret) 4445 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4446 LLONG_MAX); 4447 4448 if (ret) 4449 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4450 __func__, ret); 4451 f2fs_update_time(sbi, REQ_TIME); 4452 out: 4453 inode_unlock(inode); 4454 mnt_drop_write_file(filp); 4455 4456 return ret; 4457 } 4458 4459 static int f2fs_ioc_compress_file(struct file *filp) 4460 { 4461 struct inode *inode = file_inode(filp); 4462 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4463 struct f2fs_inode_info *fi = F2FS_I(inode); 4464 pgoff_t page_idx = 0, last_idx, cluster_idx; 4465 int ret; 4466 4467 if (!f2fs_sb_has_compression(sbi) || 4468 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4469 return -EOPNOTSUPP; 4470 4471 if (!(filp->f_mode & FMODE_WRITE)) 4472 return -EBADF; 4473 4474 f2fs_balance_fs(sbi, true); 4475 4476 ret = mnt_want_write_file(filp); 4477 if (ret) 4478 return ret; 4479 inode_lock(inode); 4480 4481 if (!f2fs_is_compress_backend_ready(inode)) { 4482 ret = -EOPNOTSUPP; 4483 goto out; 4484 } 4485 4486 if (!f2fs_compressed_file(inode) || 4487 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4488 ret = -EINVAL; 4489 goto out; 4490 } 4491 4492 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4493 if (ret) 4494 goto out; 4495 4496 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4497 4498 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4499 last_idx >>= fi->i_log_cluster_size; 4500 4501 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) { 4502 page_idx = cluster_idx << fi->i_log_cluster_size; 4503 4504 if (f2fs_is_sparse_cluster(inode, page_idx)) 4505 continue; 4506 4507 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size); 4508 if (ret < 0) 4509 break; 4510 4511 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4512 ret = filemap_fdatawrite(inode->i_mapping); 4513 if (ret < 0) 4514 break; 4515 } 4516 4517 cond_resched(); 4518 if (fatal_signal_pending(current)) { 4519 ret = -EINTR; 4520 break; 4521 } 4522 } 4523 4524 if (!ret) 4525 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4526 LLONG_MAX); 4527 4528 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4529 4530 if (ret) 4531 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4532 __func__, ret); 4533 f2fs_update_time(sbi, REQ_TIME); 4534 out: 4535 inode_unlock(inode); 4536 mnt_drop_write_file(filp); 4537 4538 return ret; 4539 } 4540 4541 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4542 { 4543 switch (cmd) { 4544 case FS_IOC_GETVERSION: 4545 return f2fs_ioc_getversion(filp, arg); 4546 case F2FS_IOC_START_ATOMIC_WRITE: 4547 return f2fs_ioc_start_atomic_write(filp, false); 4548 case F2FS_IOC_START_ATOMIC_REPLACE: 4549 return f2fs_ioc_start_atomic_write(filp, true); 4550 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4551 return f2fs_ioc_commit_atomic_write(filp); 4552 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4553 return f2fs_ioc_abort_atomic_write(filp); 4554 case F2FS_IOC_START_VOLATILE_WRITE: 4555 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4556 return -EOPNOTSUPP; 4557 case F2FS_IOC_SHUTDOWN: 4558 return f2fs_ioc_shutdown(filp, arg); 4559 case FITRIM: 4560 return f2fs_ioc_fitrim(filp, arg); 4561 case FS_IOC_SET_ENCRYPTION_POLICY: 4562 return f2fs_ioc_set_encryption_policy(filp, arg); 4563 case FS_IOC_GET_ENCRYPTION_POLICY: 4564 return f2fs_ioc_get_encryption_policy(filp, arg); 4565 case FS_IOC_GET_ENCRYPTION_PWSALT: 4566 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4567 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4568 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4569 case FS_IOC_ADD_ENCRYPTION_KEY: 4570 return f2fs_ioc_add_encryption_key(filp, arg); 4571 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4572 return f2fs_ioc_remove_encryption_key(filp, arg); 4573 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4574 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4575 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4576 return f2fs_ioc_get_encryption_key_status(filp, arg); 4577 case FS_IOC_GET_ENCRYPTION_NONCE: 4578 return f2fs_ioc_get_encryption_nonce(filp, arg); 4579 case F2FS_IOC_GARBAGE_COLLECT: 4580 return f2fs_ioc_gc(filp, arg); 4581 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4582 return f2fs_ioc_gc_range(filp, arg); 4583 case F2FS_IOC_WRITE_CHECKPOINT: 4584 return f2fs_ioc_write_checkpoint(filp); 4585 case F2FS_IOC_DEFRAGMENT: 4586 return f2fs_ioc_defragment(filp, arg); 4587 case F2FS_IOC_MOVE_RANGE: 4588 return f2fs_ioc_move_range(filp, arg); 4589 case F2FS_IOC_FLUSH_DEVICE: 4590 return f2fs_ioc_flush_device(filp, arg); 4591 case F2FS_IOC_GET_FEATURES: 4592 return f2fs_ioc_get_features(filp, arg); 4593 case F2FS_IOC_GET_PIN_FILE: 4594 return f2fs_ioc_get_pin_file(filp, arg); 4595 case F2FS_IOC_SET_PIN_FILE: 4596 return f2fs_ioc_set_pin_file(filp, arg); 4597 case F2FS_IOC_PRECACHE_EXTENTS: 4598 return f2fs_ioc_precache_extents(filp); 4599 case F2FS_IOC_RESIZE_FS: 4600 return f2fs_ioc_resize_fs(filp, arg); 4601 case FS_IOC_ENABLE_VERITY: 4602 return f2fs_ioc_enable_verity(filp, arg); 4603 case FS_IOC_MEASURE_VERITY: 4604 return f2fs_ioc_measure_verity(filp, arg); 4605 case FS_IOC_READ_VERITY_METADATA: 4606 return f2fs_ioc_read_verity_metadata(filp, arg); 4607 case FS_IOC_GETFSLABEL: 4608 return f2fs_ioc_getfslabel(filp, arg); 4609 case FS_IOC_SETFSLABEL: 4610 return f2fs_ioc_setfslabel(filp, arg); 4611 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4612 return f2fs_ioc_get_compress_blocks(filp, arg); 4613 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4614 return f2fs_release_compress_blocks(filp, arg); 4615 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4616 return f2fs_reserve_compress_blocks(filp, arg); 4617 case F2FS_IOC_SEC_TRIM_FILE: 4618 return f2fs_sec_trim_file(filp, arg); 4619 case F2FS_IOC_GET_COMPRESS_OPTION: 4620 return f2fs_ioc_get_compress_option(filp, arg); 4621 case F2FS_IOC_SET_COMPRESS_OPTION: 4622 return f2fs_ioc_set_compress_option(filp, arg); 4623 case F2FS_IOC_DECOMPRESS_FILE: 4624 return f2fs_ioc_decompress_file(filp); 4625 case F2FS_IOC_COMPRESS_FILE: 4626 return f2fs_ioc_compress_file(filp); 4627 case F2FS_IOC_GET_DEV_ALIAS_FILE: 4628 return f2fs_ioc_get_dev_alias_file(filp, arg); 4629 case F2FS_IOC_IO_PRIO: 4630 return f2fs_ioc_io_prio(filp, arg); 4631 default: 4632 return -ENOTTY; 4633 } 4634 } 4635 4636 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4637 { 4638 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4639 return -EIO; 4640 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4641 return -ENOSPC; 4642 4643 return __f2fs_ioctl(filp, cmd, arg); 4644 } 4645 4646 /* 4647 * Return %true if the given read or write request should use direct I/O, or 4648 * %false if it should use buffered I/O. 4649 */ 4650 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4651 struct iov_iter *iter) 4652 { 4653 unsigned int align; 4654 4655 if (!(iocb->ki_flags & IOCB_DIRECT)) 4656 return false; 4657 4658 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4659 return false; 4660 4661 /* 4662 * Direct I/O not aligned to the disk's logical_block_size will be 4663 * attempted, but will fail with -EINVAL. 4664 * 4665 * f2fs additionally requires that direct I/O be aligned to the 4666 * filesystem block size, which is often a stricter requirement. 4667 * However, f2fs traditionally falls back to buffered I/O on requests 4668 * that are logical_block_size-aligned but not fs-block aligned. 4669 * 4670 * The below logic implements this behavior. 4671 */ 4672 align = iocb->ki_pos | iov_iter_alignment(iter); 4673 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4674 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4675 return false; 4676 4677 return true; 4678 } 4679 4680 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4681 unsigned int flags) 4682 { 4683 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4684 4685 dec_page_count(sbi, F2FS_DIO_READ); 4686 if (error) 4687 return error; 4688 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4689 return 0; 4690 } 4691 4692 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4693 .end_io = f2fs_dio_read_end_io, 4694 }; 4695 4696 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4697 { 4698 struct file *file = iocb->ki_filp; 4699 struct inode *inode = file_inode(file); 4700 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4701 struct f2fs_inode_info *fi = F2FS_I(inode); 4702 const loff_t pos = iocb->ki_pos; 4703 const size_t count = iov_iter_count(to); 4704 struct iomap_dio *dio; 4705 ssize_t ret; 4706 4707 if (count == 0) 4708 return 0; /* skip atime update */ 4709 4710 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4711 4712 if (iocb->ki_flags & IOCB_NOWAIT) { 4713 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4714 ret = -EAGAIN; 4715 goto out; 4716 } 4717 } else { 4718 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4719 } 4720 4721 /* dio is not compatible w/ atomic file */ 4722 if (f2fs_is_atomic_file(inode)) { 4723 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4724 ret = -EOPNOTSUPP; 4725 goto out; 4726 } 4727 4728 /* 4729 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4730 * the higher-level function iomap_dio_rw() in order to ensure that the 4731 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4732 */ 4733 inc_page_count(sbi, F2FS_DIO_READ); 4734 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4735 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4736 if (IS_ERR_OR_NULL(dio)) { 4737 ret = PTR_ERR_OR_ZERO(dio); 4738 if (ret != -EIOCBQUEUED) 4739 dec_page_count(sbi, F2FS_DIO_READ); 4740 } else { 4741 ret = iomap_dio_complete(dio); 4742 } 4743 4744 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4745 4746 file_accessed(file); 4747 out: 4748 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4749 return ret; 4750 } 4751 4752 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4753 int rw) 4754 { 4755 struct inode *inode = file_inode(file); 4756 char *buf, *path; 4757 4758 buf = f2fs_getname(F2FS_I_SB(inode)); 4759 if (!buf) 4760 return; 4761 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4762 if (IS_ERR(path)) 4763 goto free_buf; 4764 if (rw == WRITE) 4765 trace_f2fs_datawrite_start(inode, pos, count, 4766 current->pid, path, current->comm); 4767 else 4768 trace_f2fs_dataread_start(inode, pos, count, 4769 current->pid, path, current->comm); 4770 free_buf: 4771 f2fs_putname(buf); 4772 } 4773 4774 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4775 { 4776 struct inode *inode = file_inode(iocb->ki_filp); 4777 const loff_t pos = iocb->ki_pos; 4778 ssize_t ret; 4779 4780 if (!f2fs_is_compress_backend_ready(inode)) 4781 return -EOPNOTSUPP; 4782 4783 if (trace_f2fs_dataread_start_enabled()) 4784 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4785 iov_iter_count(to), READ); 4786 4787 /* In LFS mode, if there is inflight dio, wait for its completion */ 4788 if (f2fs_lfs_mode(F2FS_I_SB(inode)) && 4789 get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE)) 4790 inode_dio_wait(inode); 4791 4792 if (f2fs_should_use_dio(inode, iocb, to)) { 4793 ret = f2fs_dio_read_iter(iocb, to); 4794 } else { 4795 ret = filemap_read(iocb, to, 0); 4796 if (ret > 0) 4797 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4798 APP_BUFFERED_READ_IO, ret); 4799 } 4800 if (trace_f2fs_dataread_end_enabled()) 4801 trace_f2fs_dataread_end(inode, pos, ret); 4802 return ret; 4803 } 4804 4805 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4806 struct pipe_inode_info *pipe, 4807 size_t len, unsigned int flags) 4808 { 4809 struct inode *inode = file_inode(in); 4810 const loff_t pos = *ppos; 4811 ssize_t ret; 4812 4813 if (!f2fs_is_compress_backend_ready(inode)) 4814 return -EOPNOTSUPP; 4815 4816 if (trace_f2fs_dataread_start_enabled()) 4817 f2fs_trace_rw_file_path(in, pos, len, READ); 4818 4819 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4820 if (ret > 0) 4821 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4822 APP_BUFFERED_READ_IO, ret); 4823 4824 if (trace_f2fs_dataread_end_enabled()) 4825 trace_f2fs_dataread_end(inode, pos, ret); 4826 return ret; 4827 } 4828 4829 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4830 { 4831 struct file *file = iocb->ki_filp; 4832 struct inode *inode = file_inode(file); 4833 ssize_t count; 4834 int err; 4835 4836 if (IS_IMMUTABLE(inode)) 4837 return -EPERM; 4838 4839 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4840 return -EPERM; 4841 4842 count = generic_write_checks(iocb, from); 4843 if (count <= 0) 4844 return count; 4845 4846 err = file_modified(file); 4847 if (err) 4848 return err; 4849 return count; 4850 } 4851 4852 /* 4853 * Preallocate blocks for a write request, if it is possible and helpful to do 4854 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4855 * blocks were preallocated, or a negative errno value if something went 4856 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4857 * requested blocks (not just some of them) have been allocated. 4858 */ 4859 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4860 bool dio) 4861 { 4862 struct inode *inode = file_inode(iocb->ki_filp); 4863 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4864 const loff_t pos = iocb->ki_pos; 4865 const size_t count = iov_iter_count(iter); 4866 struct f2fs_map_blocks map = {}; 4867 int flag; 4868 int ret; 4869 4870 /* If it will be an out-of-place direct write, don't bother. */ 4871 if (dio && f2fs_lfs_mode(sbi)) 4872 return 0; 4873 /* 4874 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4875 * buffered IO, if DIO meets any holes. 4876 */ 4877 if (dio && i_size_read(inode) && 4878 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4879 return 0; 4880 4881 /* No-wait I/O can't allocate blocks. */ 4882 if (iocb->ki_flags & IOCB_NOWAIT) 4883 return 0; 4884 4885 /* If it will be a short write, don't bother. */ 4886 if (fault_in_iov_iter_readable(iter, count)) 4887 return 0; 4888 4889 if (f2fs_has_inline_data(inode)) { 4890 /* If the data will fit inline, don't bother. */ 4891 if (pos + count <= MAX_INLINE_DATA(inode)) 4892 return 0; 4893 ret = f2fs_convert_inline_inode(inode); 4894 if (ret) 4895 return ret; 4896 } 4897 4898 /* Do not preallocate blocks that will be written partially in 4KB. */ 4899 map.m_lblk = F2FS_BLK_ALIGN(pos); 4900 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4901 if (map.m_len > map.m_lblk) 4902 map.m_len -= map.m_lblk; 4903 else 4904 return 0; 4905 4906 if (!IS_DEVICE_ALIASING(inode)) 4907 map.m_may_create = true; 4908 if (dio) { 4909 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi, 4910 inode->i_write_hint); 4911 flag = F2FS_GET_BLOCK_PRE_DIO; 4912 } else { 4913 map.m_seg_type = NO_CHECK_TYPE; 4914 flag = F2FS_GET_BLOCK_PRE_AIO; 4915 } 4916 4917 ret = f2fs_map_blocks(inode, &map, flag); 4918 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4919 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4920 return ret; 4921 if (ret == 0) 4922 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4923 return map.m_len; 4924 } 4925 4926 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4927 struct iov_iter *from) 4928 { 4929 struct file *file = iocb->ki_filp; 4930 struct inode *inode = file_inode(file); 4931 ssize_t ret; 4932 4933 if (iocb->ki_flags & IOCB_NOWAIT) 4934 return -EOPNOTSUPP; 4935 4936 ret = generic_perform_write(iocb, from); 4937 4938 if (ret > 0) { 4939 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4940 APP_BUFFERED_IO, ret); 4941 } 4942 return ret; 4943 } 4944 4945 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4946 unsigned int flags) 4947 { 4948 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4949 4950 dec_page_count(sbi, F2FS_DIO_WRITE); 4951 if (error) 4952 return error; 4953 f2fs_update_time(sbi, REQ_TIME); 4954 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4955 return 0; 4956 } 4957 4958 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter, 4959 struct bio *bio, loff_t file_offset) 4960 { 4961 struct inode *inode = iter->inode; 4962 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4963 enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint); 4964 enum temp_type temp = f2fs_get_segment_temp(sbi, type); 4965 4966 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp); 4967 submit_bio(bio); 4968 } 4969 4970 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4971 .end_io = f2fs_dio_write_end_io, 4972 .submit_io = f2fs_dio_write_submit_io, 4973 }; 4974 4975 static void f2fs_flush_buffered_write(struct address_space *mapping, 4976 loff_t start_pos, loff_t end_pos) 4977 { 4978 int ret; 4979 4980 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4981 if (ret < 0) 4982 return; 4983 invalidate_mapping_pages(mapping, 4984 start_pos >> PAGE_SHIFT, 4985 end_pos >> PAGE_SHIFT); 4986 } 4987 4988 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4989 bool *may_need_sync) 4990 { 4991 struct file *file = iocb->ki_filp; 4992 struct inode *inode = file_inode(file); 4993 struct f2fs_inode_info *fi = F2FS_I(inode); 4994 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4995 const bool do_opu = f2fs_lfs_mode(sbi); 4996 const loff_t pos = iocb->ki_pos; 4997 const ssize_t count = iov_iter_count(from); 4998 unsigned int dio_flags; 4999 struct iomap_dio *dio; 5000 ssize_t ret; 5001 5002 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 5003 5004 if (iocb->ki_flags & IOCB_NOWAIT) { 5005 /* f2fs_convert_inline_inode() and block allocation can block */ 5006 if (f2fs_has_inline_data(inode) || 5007 !f2fs_overwrite_io(inode, pos, count)) { 5008 ret = -EAGAIN; 5009 goto out; 5010 } 5011 5012 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 5013 ret = -EAGAIN; 5014 goto out; 5015 } 5016 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 5017 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5018 ret = -EAGAIN; 5019 goto out; 5020 } 5021 } else { 5022 ret = f2fs_convert_inline_inode(inode); 5023 if (ret) 5024 goto out; 5025 5026 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 5027 if (do_opu) 5028 f2fs_down_read(&fi->i_gc_rwsem[READ]); 5029 } 5030 5031 /* 5032 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 5033 * the higher-level function iomap_dio_rw() in order to ensure that the 5034 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 5035 */ 5036 inc_page_count(sbi, F2FS_DIO_WRITE); 5037 dio_flags = 0; 5038 if (pos + count > inode->i_size) 5039 dio_flags |= IOMAP_DIO_FORCE_WAIT; 5040 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 5041 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 5042 if (IS_ERR_OR_NULL(dio)) { 5043 ret = PTR_ERR_OR_ZERO(dio); 5044 if (ret == -ENOTBLK) 5045 ret = 0; 5046 if (ret != -EIOCBQUEUED) 5047 dec_page_count(sbi, F2FS_DIO_WRITE); 5048 } else { 5049 ret = iomap_dio_complete(dio); 5050 } 5051 5052 if (do_opu) 5053 f2fs_up_read(&fi->i_gc_rwsem[READ]); 5054 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 5055 5056 if (ret < 0) 5057 goto out; 5058 if (pos + ret > inode->i_size) 5059 f2fs_i_size_write(inode, pos + ret); 5060 if (!do_opu) 5061 set_inode_flag(inode, FI_UPDATE_WRITE); 5062 5063 if (iov_iter_count(from)) { 5064 ssize_t ret2; 5065 loff_t bufio_start_pos = iocb->ki_pos; 5066 5067 /* 5068 * The direct write was partial, so we need to fall back to a 5069 * buffered write for the remainder. 5070 */ 5071 5072 ret2 = f2fs_buffered_write_iter(iocb, from); 5073 if (iov_iter_count(from)) 5074 f2fs_write_failed(inode, iocb->ki_pos); 5075 if (ret2 < 0) 5076 goto out; 5077 5078 /* 5079 * Ensure that the pagecache pages are written to disk and 5080 * invalidated to preserve the expected O_DIRECT semantics. 5081 */ 5082 if (ret2 > 0) { 5083 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 5084 5085 ret += ret2; 5086 5087 f2fs_flush_buffered_write(file->f_mapping, 5088 bufio_start_pos, 5089 bufio_end_pos); 5090 } 5091 } else { 5092 /* iomap_dio_rw() already handled the generic_write_sync(). */ 5093 *may_need_sync = false; 5094 } 5095 out: 5096 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 5097 return ret; 5098 } 5099 5100 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 5101 { 5102 struct inode *inode = file_inode(iocb->ki_filp); 5103 const loff_t orig_pos = iocb->ki_pos; 5104 const size_t orig_count = iov_iter_count(from); 5105 loff_t target_size; 5106 bool dio; 5107 bool may_need_sync = true; 5108 int preallocated; 5109 const loff_t pos = iocb->ki_pos; 5110 const ssize_t count = iov_iter_count(from); 5111 ssize_t ret; 5112 5113 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 5114 ret = -EIO; 5115 goto out; 5116 } 5117 5118 if (!f2fs_is_compress_backend_ready(inode)) { 5119 ret = -EOPNOTSUPP; 5120 goto out; 5121 } 5122 5123 if (iocb->ki_flags & IOCB_NOWAIT) { 5124 if (!inode_trylock(inode)) { 5125 ret = -EAGAIN; 5126 goto out; 5127 } 5128 } else { 5129 inode_lock(inode); 5130 } 5131 5132 if (f2fs_is_pinned_file(inode) && 5133 !f2fs_overwrite_io(inode, pos, count)) { 5134 ret = -EIO; 5135 goto out_unlock; 5136 } 5137 5138 ret = f2fs_write_checks(iocb, from); 5139 if (ret <= 0) 5140 goto out_unlock; 5141 5142 /* Determine whether we will do a direct write or a buffered write. */ 5143 dio = f2fs_should_use_dio(inode, iocb, from); 5144 5145 /* dio is not compatible w/ atomic write */ 5146 if (dio && f2fs_is_atomic_file(inode)) { 5147 ret = -EOPNOTSUPP; 5148 goto out_unlock; 5149 } 5150 5151 /* Possibly preallocate the blocks for the write. */ 5152 target_size = iocb->ki_pos + iov_iter_count(from); 5153 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 5154 if (preallocated < 0) { 5155 ret = preallocated; 5156 } else { 5157 if (trace_f2fs_datawrite_start_enabled()) 5158 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 5159 orig_count, WRITE); 5160 5161 /* Do the actual write. */ 5162 ret = dio ? 5163 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 5164 f2fs_buffered_write_iter(iocb, from); 5165 5166 if (trace_f2fs_datawrite_end_enabled()) 5167 trace_f2fs_datawrite_end(inode, orig_pos, ret); 5168 } 5169 5170 /* Don't leave any preallocated blocks around past i_size. */ 5171 if (preallocated && i_size_read(inode) < target_size) { 5172 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5173 filemap_invalidate_lock(inode->i_mapping); 5174 if (!f2fs_truncate(inode)) 5175 file_dont_truncate(inode); 5176 filemap_invalidate_unlock(inode->i_mapping); 5177 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 5178 } else { 5179 file_dont_truncate(inode); 5180 } 5181 5182 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 5183 out_unlock: 5184 inode_unlock(inode); 5185 out: 5186 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 5187 5188 if (ret > 0 && may_need_sync) 5189 ret = generic_write_sync(iocb, ret); 5190 5191 /* If buffered IO was forced, flush and drop the data from 5192 * the page cache to preserve O_DIRECT semantics 5193 */ 5194 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 5195 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 5196 orig_pos, 5197 orig_pos + ret - 1); 5198 5199 return ret; 5200 } 5201 5202 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 5203 int advice) 5204 { 5205 struct address_space *mapping; 5206 struct backing_dev_info *bdi; 5207 struct inode *inode = file_inode(filp); 5208 int err; 5209 5210 if (advice == POSIX_FADV_SEQUENTIAL) { 5211 if (S_ISFIFO(inode->i_mode)) 5212 return -ESPIPE; 5213 5214 mapping = filp->f_mapping; 5215 if (!mapping || len < 0) 5216 return -EINVAL; 5217 5218 bdi = inode_to_bdi(mapping->host); 5219 filp->f_ra.ra_pages = bdi->ra_pages * 5220 F2FS_I_SB(inode)->seq_file_ra_mul; 5221 spin_lock(&filp->f_lock); 5222 filp->f_mode &= ~FMODE_RANDOM; 5223 spin_unlock(&filp->f_lock); 5224 return 0; 5225 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 5226 /* Load extent cache at the first readahead. */ 5227 f2fs_precache_extents(inode); 5228 } 5229 5230 err = generic_fadvise(filp, offset, len, advice); 5231 if (err) 5232 return err; 5233 5234 if (advice == POSIX_FADV_DONTNEED && 5235 (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 5236 f2fs_compressed_file(inode))) 5237 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 5238 else if (advice == POSIX_FADV_NOREUSE) 5239 f2fs_keep_noreuse_range(inode, offset, len); 5240 return 0; 5241 } 5242 5243 #ifdef CONFIG_COMPAT 5244 struct compat_f2fs_gc_range { 5245 u32 sync; 5246 compat_u64 start; 5247 compat_u64 len; 5248 }; 5249 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 5250 struct compat_f2fs_gc_range) 5251 5252 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 5253 { 5254 struct compat_f2fs_gc_range __user *urange; 5255 struct f2fs_gc_range range; 5256 int err; 5257 5258 urange = compat_ptr(arg); 5259 err = get_user(range.sync, &urange->sync); 5260 err |= get_user(range.start, &urange->start); 5261 err |= get_user(range.len, &urange->len); 5262 if (err) 5263 return -EFAULT; 5264 5265 return __f2fs_ioc_gc_range(file, &range); 5266 } 5267 5268 struct compat_f2fs_move_range { 5269 u32 dst_fd; 5270 compat_u64 pos_in; 5271 compat_u64 pos_out; 5272 compat_u64 len; 5273 }; 5274 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 5275 struct compat_f2fs_move_range) 5276 5277 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 5278 { 5279 struct compat_f2fs_move_range __user *urange; 5280 struct f2fs_move_range range; 5281 int err; 5282 5283 urange = compat_ptr(arg); 5284 err = get_user(range.dst_fd, &urange->dst_fd); 5285 err |= get_user(range.pos_in, &urange->pos_in); 5286 err |= get_user(range.pos_out, &urange->pos_out); 5287 err |= get_user(range.len, &urange->len); 5288 if (err) 5289 return -EFAULT; 5290 5291 return __f2fs_ioc_move_range(file, &range); 5292 } 5293 5294 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5295 { 5296 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 5297 return -EIO; 5298 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 5299 return -ENOSPC; 5300 5301 switch (cmd) { 5302 case FS_IOC32_GETVERSION: 5303 cmd = FS_IOC_GETVERSION; 5304 break; 5305 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 5306 return f2fs_compat_ioc_gc_range(file, arg); 5307 case F2FS_IOC32_MOVE_RANGE: 5308 return f2fs_compat_ioc_move_range(file, arg); 5309 case F2FS_IOC_START_ATOMIC_WRITE: 5310 case F2FS_IOC_START_ATOMIC_REPLACE: 5311 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5312 case F2FS_IOC_START_VOLATILE_WRITE: 5313 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5314 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5315 case F2FS_IOC_SHUTDOWN: 5316 case FITRIM: 5317 case FS_IOC_SET_ENCRYPTION_POLICY: 5318 case FS_IOC_GET_ENCRYPTION_PWSALT: 5319 case FS_IOC_GET_ENCRYPTION_POLICY: 5320 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5321 case FS_IOC_ADD_ENCRYPTION_KEY: 5322 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5323 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5324 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5325 case FS_IOC_GET_ENCRYPTION_NONCE: 5326 case F2FS_IOC_GARBAGE_COLLECT: 5327 case F2FS_IOC_WRITE_CHECKPOINT: 5328 case F2FS_IOC_DEFRAGMENT: 5329 case F2FS_IOC_FLUSH_DEVICE: 5330 case F2FS_IOC_GET_FEATURES: 5331 case F2FS_IOC_GET_PIN_FILE: 5332 case F2FS_IOC_SET_PIN_FILE: 5333 case F2FS_IOC_PRECACHE_EXTENTS: 5334 case F2FS_IOC_RESIZE_FS: 5335 case FS_IOC_ENABLE_VERITY: 5336 case FS_IOC_MEASURE_VERITY: 5337 case FS_IOC_READ_VERITY_METADATA: 5338 case FS_IOC_GETFSLABEL: 5339 case FS_IOC_SETFSLABEL: 5340 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5341 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5342 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5343 case F2FS_IOC_SEC_TRIM_FILE: 5344 case F2FS_IOC_GET_COMPRESS_OPTION: 5345 case F2FS_IOC_SET_COMPRESS_OPTION: 5346 case F2FS_IOC_DECOMPRESS_FILE: 5347 case F2FS_IOC_COMPRESS_FILE: 5348 case F2FS_IOC_GET_DEV_ALIAS_FILE: 5349 case F2FS_IOC_IO_PRIO: 5350 break; 5351 default: 5352 return -ENOIOCTLCMD; 5353 } 5354 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5355 } 5356 #endif 5357 5358 const struct file_operations f2fs_file_operations = { 5359 .llseek = f2fs_llseek, 5360 .read_iter = f2fs_file_read_iter, 5361 .write_iter = f2fs_file_write_iter, 5362 .iopoll = iocb_bio_iopoll, 5363 .open = f2fs_file_open, 5364 .release = f2fs_release_file, 5365 .mmap = f2fs_file_mmap, 5366 .flush = f2fs_file_flush, 5367 .fsync = f2fs_sync_file, 5368 .fallocate = f2fs_fallocate, 5369 .unlocked_ioctl = f2fs_ioctl, 5370 #ifdef CONFIG_COMPAT 5371 .compat_ioctl = f2fs_compat_ioctl, 5372 #endif 5373 .splice_read = f2fs_file_splice_read, 5374 .splice_write = iter_file_splice_write, 5375 .fadvise = f2fs_file_fadvise, 5376 .fop_flags = FOP_BUFFER_RASYNC, 5377 }; 5378