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/buffer_head.h> 12 #include <linux/writeback.h> 13 #include <linux/blkdev.h> 14 #include <linux/falloc.h> 15 #include <linux/types.h> 16 #include <linux/compat.h> 17 #include <linux/uaccess.h> 18 #include <linux/mount.h> 19 #include <linux/pagevec.h> 20 #include <linux/uio.h> 21 #include <linux/uuid.h> 22 #include <linux/file.h> 23 #include <linux/nls.h> 24 #include <linux/sched/signal.h> 25 #include <linux/fileattr.h> 26 #include <linux/fadvise.h> 27 #include <linux/iomap.h> 28 29 #include "f2fs.h" 30 #include "node.h" 31 #include "segment.h" 32 #include "xattr.h" 33 #include "acl.h" 34 #include "gc.h" 35 #include "iostat.h" 36 #include <trace/events/f2fs.h> 37 #include <uapi/linux/f2fs.h> 38 39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 40 { 41 struct inode *inode = file_inode(vmf->vma->vm_file); 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, (unsigned long)ret); 50 51 return ret; 52 } 53 54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 55 { 56 struct page *page = 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 = true; 61 int err = 0; 62 63 if (unlikely(IS_IMMUTABLE(inode))) 64 return VM_FAULT_SIGBUS; 65 66 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 67 return VM_FAULT_SIGBUS; 68 69 if (unlikely(f2fs_cp_error(sbi))) { 70 err = -EIO; 71 goto err; 72 } 73 74 if (!f2fs_is_checkpoint_ready(sbi)) { 75 err = -ENOSPC; 76 goto err; 77 } 78 79 err = f2fs_convert_inline_inode(inode); 80 if (err) 81 goto err; 82 83 #ifdef CONFIG_F2FS_FS_COMPRESSION 84 if (f2fs_compressed_file(inode)) { 85 int ret = f2fs_is_compressed_cluster(inode, page->index); 86 87 if (ret < 0) { 88 err = ret; 89 goto err; 90 } else if (ret) { 91 need_alloc = false; 92 } 93 } 94 #endif 95 /* should do out of any locked page */ 96 if (need_alloc) 97 f2fs_balance_fs(sbi, true); 98 99 sb_start_pagefault(inode->i_sb); 100 101 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 102 103 file_update_time(vmf->vma->vm_file); 104 filemap_invalidate_lock_shared(inode->i_mapping); 105 lock_page(page); 106 if (unlikely(page->mapping != inode->i_mapping || 107 page_offset(page) > i_size_read(inode) || 108 !PageUptodate(page))) { 109 unlock_page(page); 110 err = -EFAULT; 111 goto out_sem; 112 } 113 114 if (need_alloc) { 115 /* block allocation */ 116 set_new_dnode(&dn, inode, NULL, NULL, 0); 117 err = f2fs_get_block_locked(&dn, page->index); 118 } 119 120 #ifdef CONFIG_F2FS_FS_COMPRESSION 121 if (!need_alloc) { 122 set_new_dnode(&dn, inode, NULL, NULL, 0); 123 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE); 124 f2fs_put_dnode(&dn); 125 } 126 #endif 127 if (err) { 128 unlock_page(page); 129 goto out_sem; 130 } 131 132 f2fs_wait_on_page_writeback(page, DATA, false, true); 133 134 /* wait for GCed page writeback via META_MAPPING */ 135 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 136 137 /* 138 * check to see if the page is mapped already (no holes) 139 */ 140 if (PageMappedToDisk(page)) 141 goto out_sem; 142 143 /* page is wholly or partially inside EOF */ 144 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 145 i_size_read(inode)) { 146 loff_t offset; 147 148 offset = i_size_read(inode) & ~PAGE_MASK; 149 zero_user_segment(page, offset, PAGE_SIZE); 150 } 151 set_page_dirty(page); 152 153 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 154 f2fs_update_time(sbi, REQ_TIME); 155 156 trace_f2fs_vm_page_mkwrite(page, DATA); 157 out_sem: 158 filemap_invalidate_unlock_shared(inode->i_mapping); 159 160 sb_end_pagefault(inode->i_sb); 161 err: 162 return vmf_fs_error(err); 163 } 164 165 static const struct vm_operations_struct f2fs_file_vm_ops = { 166 .fault = f2fs_filemap_fault, 167 .map_pages = filemap_map_pages, 168 .page_mkwrite = f2fs_vm_page_mkwrite, 169 }; 170 171 static int get_parent_ino(struct inode *inode, nid_t *pino) 172 { 173 struct dentry *dentry; 174 175 /* 176 * Make sure to get the non-deleted alias. The alias associated with 177 * the open file descriptor being fsync()'ed may be deleted already. 178 */ 179 dentry = d_find_alias(inode); 180 if (!dentry) 181 return 0; 182 183 *pino = parent_ino(dentry); 184 dput(dentry); 185 return 1; 186 } 187 188 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 189 { 190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 191 enum cp_reason_type cp_reason = CP_NO_NEEDED; 192 193 if (!S_ISREG(inode->i_mode)) 194 cp_reason = CP_NON_REGULAR; 195 else if (f2fs_compressed_file(inode)) 196 cp_reason = CP_COMPRESSED; 197 else if (inode->i_nlink != 1) 198 cp_reason = CP_HARDLINK; 199 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 200 cp_reason = CP_SB_NEED_CP; 201 else if (file_wrong_pino(inode)) 202 cp_reason = CP_WRONG_PINO; 203 else if (!f2fs_space_for_roll_forward(sbi)) 204 cp_reason = CP_NO_SPC_ROLL; 205 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 206 cp_reason = CP_NODE_NEED_CP; 207 else if (test_opt(sbi, FASTBOOT)) 208 cp_reason = CP_FASTBOOT_MODE; 209 else if (F2FS_OPTION(sbi).active_logs == 2) 210 cp_reason = CP_SPEC_LOG_NUM; 211 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 212 f2fs_need_dentry_mark(sbi, inode->i_ino) && 213 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 214 TRANS_DIR_INO)) 215 cp_reason = CP_RECOVER_DIR; 216 217 return cp_reason; 218 } 219 220 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 221 { 222 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 223 bool ret = false; 224 /* But we need to avoid that there are some inode updates */ 225 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 226 ret = true; 227 f2fs_put_page(i, 0); 228 return ret; 229 } 230 231 static void try_to_fix_pino(struct inode *inode) 232 { 233 struct f2fs_inode_info *fi = F2FS_I(inode); 234 nid_t pino; 235 236 f2fs_down_write(&fi->i_sem); 237 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 238 get_parent_ino(inode, &pino)) { 239 f2fs_i_pino_write(inode, pino); 240 file_got_pino(inode); 241 } 242 f2fs_up_write(&fi->i_sem); 243 } 244 245 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 246 int datasync, bool atomic) 247 { 248 struct inode *inode = file->f_mapping->host; 249 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 250 nid_t ino = inode->i_ino; 251 int ret = 0; 252 enum cp_reason_type cp_reason = 0; 253 struct writeback_control wbc = { 254 .sync_mode = WB_SYNC_ALL, 255 .nr_to_write = LONG_MAX, 256 .for_reclaim = 0, 257 }; 258 unsigned int seq_id = 0; 259 260 if (unlikely(f2fs_readonly(inode->i_sb))) 261 return 0; 262 263 trace_f2fs_sync_file_enter(inode); 264 265 if (S_ISDIR(inode->i_mode)) 266 goto go_write; 267 268 /* if fdatasync is triggered, let's do in-place-update */ 269 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 270 set_inode_flag(inode, FI_NEED_IPU); 271 ret = file_write_and_wait_range(file, start, end); 272 clear_inode_flag(inode, FI_NEED_IPU); 273 274 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 275 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 276 return ret; 277 } 278 279 /* if the inode is dirty, let's recover all the time */ 280 if (!f2fs_skip_inode_update(inode, datasync)) { 281 f2fs_write_inode(inode, NULL); 282 goto go_write; 283 } 284 285 /* 286 * if there is no written data, don't waste time to write recovery info. 287 */ 288 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 289 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 290 291 /* it may call write_inode just prior to fsync */ 292 if (need_inode_page_update(sbi, ino)) 293 goto go_write; 294 295 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 296 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 297 goto flush_out; 298 goto out; 299 } else { 300 /* 301 * for OPU case, during fsync(), node can be persisted before 302 * data when lower device doesn't support write barrier, result 303 * in data corruption after SPO. 304 * So for strict fsync mode, force to use atomic write semantics 305 * to keep write order in between data/node and last node to 306 * avoid potential data corruption. 307 */ 308 if (F2FS_OPTION(sbi).fsync_mode == 309 FSYNC_MODE_STRICT && !atomic) 310 atomic = true; 311 } 312 go_write: 313 /* 314 * Both of fdatasync() and fsync() are able to be recovered from 315 * sudden-power-off. 316 */ 317 f2fs_down_read(&F2FS_I(inode)->i_sem); 318 cp_reason = need_do_checkpoint(inode); 319 f2fs_up_read(&F2FS_I(inode)->i_sem); 320 321 if (cp_reason) { 322 /* all the dirty node pages should be flushed for POR */ 323 ret = f2fs_sync_fs(inode->i_sb, 1); 324 325 /* 326 * We've secured consistency through sync_fs. Following pino 327 * will be used only for fsynced inodes after checkpoint. 328 */ 329 try_to_fix_pino(inode); 330 clear_inode_flag(inode, FI_APPEND_WRITE); 331 clear_inode_flag(inode, FI_UPDATE_WRITE); 332 goto out; 333 } 334 sync_nodes: 335 atomic_inc(&sbi->wb_sync_req[NODE]); 336 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 337 atomic_dec(&sbi->wb_sync_req[NODE]); 338 if (ret) 339 goto out; 340 341 /* if cp_error was enabled, we should avoid infinite loop */ 342 if (unlikely(f2fs_cp_error(sbi))) { 343 ret = -EIO; 344 goto out; 345 } 346 347 if (f2fs_need_inode_block_update(sbi, ino)) { 348 f2fs_mark_inode_dirty_sync(inode, true); 349 f2fs_write_inode(inode, NULL); 350 goto sync_nodes; 351 } 352 353 /* 354 * If it's atomic_write, it's just fine to keep write ordering. So 355 * here we don't need to wait for node write completion, since we use 356 * node chain which serializes node blocks. If one of node writes are 357 * reordered, we can see simply broken chain, resulting in stopping 358 * roll-forward recovery. It means we'll recover all or none node blocks 359 * given fsync mark. 360 */ 361 if (!atomic) { 362 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 363 if (ret) 364 goto out; 365 } 366 367 /* once recovery info is written, don't need to tack this */ 368 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 369 clear_inode_flag(inode, FI_APPEND_WRITE); 370 flush_out: 371 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) || 372 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi))) 373 ret = f2fs_issue_flush(sbi, inode->i_ino); 374 if (!ret) { 375 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 376 clear_inode_flag(inode, FI_UPDATE_WRITE); 377 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 378 } 379 f2fs_update_time(sbi, REQ_TIME); 380 out: 381 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 382 return ret; 383 } 384 385 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 386 { 387 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 388 return -EIO; 389 return f2fs_do_sync_file(file, start, end, datasync, false); 390 } 391 392 static bool __found_offset(struct address_space *mapping, block_t blkaddr, 393 pgoff_t index, int whence) 394 { 395 switch (whence) { 396 case SEEK_DATA: 397 if (__is_valid_data_blkaddr(blkaddr)) 398 return true; 399 if (blkaddr == NEW_ADDR && 400 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 401 return true; 402 break; 403 case SEEK_HOLE: 404 if (blkaddr == NULL_ADDR) 405 return true; 406 break; 407 } 408 return false; 409 } 410 411 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 412 { 413 struct inode *inode = file->f_mapping->host; 414 loff_t maxbytes = inode->i_sb->s_maxbytes; 415 struct dnode_of_data dn; 416 pgoff_t pgofs, end_offset; 417 loff_t data_ofs = offset; 418 loff_t isize; 419 int err = 0; 420 421 inode_lock(inode); 422 423 isize = i_size_read(inode); 424 if (offset >= isize) 425 goto fail; 426 427 /* handle inline data case */ 428 if (f2fs_has_inline_data(inode)) { 429 if (whence == SEEK_HOLE) { 430 data_ofs = isize; 431 goto found; 432 } else if (whence == SEEK_DATA) { 433 data_ofs = offset; 434 goto found; 435 } 436 } 437 438 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 439 440 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 441 set_new_dnode(&dn, inode, NULL, NULL, 0); 442 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 443 if (err && err != -ENOENT) { 444 goto fail; 445 } else if (err == -ENOENT) { 446 /* direct node does not exists */ 447 if (whence == SEEK_DATA) { 448 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 449 continue; 450 } else { 451 goto found; 452 } 453 } 454 455 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 456 457 /* find data/hole in dnode block */ 458 for (; dn.ofs_in_node < end_offset; 459 dn.ofs_in_node++, pgofs++, 460 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 461 block_t blkaddr; 462 463 blkaddr = f2fs_data_blkaddr(&dn); 464 465 if (__is_valid_data_blkaddr(blkaddr) && 466 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 467 blkaddr, DATA_GENERIC_ENHANCE)) { 468 f2fs_put_dnode(&dn); 469 goto fail; 470 } 471 472 if (__found_offset(file->f_mapping, blkaddr, 473 pgofs, whence)) { 474 f2fs_put_dnode(&dn); 475 goto found; 476 } 477 } 478 f2fs_put_dnode(&dn); 479 } 480 481 if (whence == SEEK_DATA) 482 goto fail; 483 found: 484 if (whence == SEEK_HOLE && data_ofs > isize) 485 data_ofs = isize; 486 inode_unlock(inode); 487 return vfs_setpos(file, data_ofs, maxbytes); 488 fail: 489 inode_unlock(inode); 490 return -ENXIO; 491 } 492 493 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 494 { 495 struct inode *inode = file->f_mapping->host; 496 loff_t maxbytes = inode->i_sb->s_maxbytes; 497 498 if (f2fs_compressed_file(inode)) 499 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; 500 501 switch (whence) { 502 case SEEK_SET: 503 case SEEK_CUR: 504 case SEEK_END: 505 return generic_file_llseek_size(file, offset, whence, 506 maxbytes, i_size_read(inode)); 507 case SEEK_DATA: 508 case SEEK_HOLE: 509 if (offset < 0) 510 return -ENXIO; 511 return f2fs_seek_block(file, offset, whence); 512 } 513 514 return -EINVAL; 515 } 516 517 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 518 { 519 struct inode *inode = file_inode(file); 520 521 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 522 return -EIO; 523 524 if (!f2fs_is_compress_backend_ready(inode)) 525 return -EOPNOTSUPP; 526 527 file_accessed(file); 528 vma->vm_ops = &f2fs_file_vm_ops; 529 530 f2fs_down_read(&F2FS_I(inode)->i_sem); 531 set_inode_flag(inode, FI_MMAP_FILE); 532 f2fs_up_read(&F2FS_I(inode)->i_sem); 533 534 return 0; 535 } 536 537 static int finish_preallocate_blocks(struct inode *inode) 538 { 539 int ret; 540 541 inode_lock(inode); 542 if (is_inode_flag_set(inode, FI_OPENED_FILE)) { 543 inode_unlock(inode); 544 return 0; 545 } 546 547 if (!file_should_truncate(inode)) { 548 set_inode_flag(inode, FI_OPENED_FILE); 549 inode_unlock(inode); 550 return 0; 551 } 552 553 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 554 filemap_invalidate_lock(inode->i_mapping); 555 556 truncate_setsize(inode, i_size_read(inode)); 557 ret = f2fs_truncate(inode); 558 559 filemap_invalidate_unlock(inode->i_mapping); 560 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 561 562 if (!ret) 563 set_inode_flag(inode, FI_OPENED_FILE); 564 565 inode_unlock(inode); 566 if (ret) 567 return ret; 568 569 file_dont_truncate(inode); 570 return 0; 571 } 572 573 static int f2fs_file_open(struct inode *inode, struct file *filp) 574 { 575 int err = fscrypt_file_open(inode, filp); 576 577 if (err) 578 return err; 579 580 if (!f2fs_is_compress_backend_ready(inode)) 581 return -EOPNOTSUPP; 582 583 err = fsverity_file_open(inode, filp); 584 if (err) 585 return err; 586 587 filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC; 588 filp->f_mode |= FMODE_CAN_ODIRECT; 589 590 err = dquot_file_open(inode, filp); 591 if (err) 592 return err; 593 594 return finish_preallocate_blocks(inode); 595 } 596 597 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 598 { 599 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 600 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 601 __le32 *addr; 602 bool compressed_cluster = false; 603 int cluster_index = 0, valid_blocks = 0; 604 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 605 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 606 607 addr = get_dnode_addr(dn->inode, dn->node_page) + ofs; 608 609 /* Assumption: truncation starts with cluster */ 610 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 611 block_t blkaddr = le32_to_cpu(*addr); 612 613 if (f2fs_compressed_file(dn->inode) && 614 !(cluster_index & (cluster_size - 1))) { 615 if (compressed_cluster) 616 f2fs_i_compr_blocks_update(dn->inode, 617 valid_blocks, false); 618 compressed_cluster = (blkaddr == COMPRESS_ADDR); 619 valid_blocks = 0; 620 } 621 622 if (blkaddr == NULL_ADDR) 623 continue; 624 625 f2fs_set_data_blkaddr(dn, NULL_ADDR); 626 627 if (__is_valid_data_blkaddr(blkaddr)) { 628 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 629 DATA_GENERIC_ENHANCE)) 630 continue; 631 if (compressed_cluster) 632 valid_blocks++; 633 } 634 635 f2fs_invalidate_blocks(sbi, blkaddr); 636 637 if (!released || blkaddr != COMPRESS_ADDR) 638 nr_free++; 639 } 640 641 if (compressed_cluster) 642 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 643 644 if (nr_free) { 645 pgoff_t fofs; 646 /* 647 * once we invalidate valid blkaddr in range [ofs, ofs + count], 648 * we will invalidate all blkaddr in the whole range. 649 */ 650 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 651 dn->inode) + ofs; 652 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 653 f2fs_update_age_extent_cache_range(dn, fofs, len); 654 dec_valid_block_count(sbi, dn->inode, nr_free); 655 } 656 dn->ofs_in_node = ofs; 657 658 f2fs_update_time(sbi, REQ_TIME); 659 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 660 dn->ofs_in_node, nr_free); 661 } 662 663 static int truncate_partial_data_page(struct inode *inode, u64 from, 664 bool cache_only) 665 { 666 loff_t offset = from & (PAGE_SIZE - 1); 667 pgoff_t index = from >> PAGE_SHIFT; 668 struct address_space *mapping = inode->i_mapping; 669 struct page *page; 670 671 if (!offset && !cache_only) 672 return 0; 673 674 if (cache_only) { 675 page = find_lock_page(mapping, index); 676 if (page && PageUptodate(page)) 677 goto truncate_out; 678 f2fs_put_page(page, 1); 679 return 0; 680 } 681 682 page = f2fs_get_lock_data_page(inode, index, true); 683 if (IS_ERR(page)) 684 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 685 truncate_out: 686 f2fs_wait_on_page_writeback(page, DATA, true, true); 687 zero_user(page, offset, PAGE_SIZE - offset); 688 689 /* An encrypted inode should have a key and truncate the last page. */ 690 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 691 if (!cache_only) 692 set_page_dirty(page); 693 f2fs_put_page(page, 1); 694 return 0; 695 } 696 697 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 698 { 699 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 700 struct dnode_of_data dn; 701 pgoff_t free_from; 702 int count = 0, err = 0; 703 struct page *ipage; 704 bool truncate_page = false; 705 706 trace_f2fs_truncate_blocks_enter(inode, from); 707 708 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 709 710 if (free_from >= max_file_blocks(inode)) 711 goto free_partial; 712 713 if (lock) 714 f2fs_lock_op(sbi); 715 716 ipage = f2fs_get_node_page(sbi, inode->i_ino); 717 if (IS_ERR(ipage)) { 718 err = PTR_ERR(ipage); 719 goto out; 720 } 721 722 if (f2fs_has_inline_data(inode)) { 723 f2fs_truncate_inline_inode(inode, ipage, from); 724 f2fs_put_page(ipage, 1); 725 truncate_page = true; 726 goto out; 727 } 728 729 set_new_dnode(&dn, inode, ipage, NULL, 0); 730 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 731 if (err) { 732 if (err == -ENOENT) 733 goto free_next; 734 goto out; 735 } 736 737 count = ADDRS_PER_PAGE(dn.node_page, inode); 738 739 count -= dn.ofs_in_node; 740 f2fs_bug_on(sbi, count < 0); 741 742 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 743 f2fs_truncate_data_blocks_range(&dn, count); 744 free_from += count; 745 } 746 747 f2fs_put_dnode(&dn); 748 free_next: 749 err = f2fs_truncate_inode_blocks(inode, free_from); 750 out: 751 if (lock) 752 f2fs_unlock_op(sbi); 753 free_partial: 754 /* lastly zero out the first data page */ 755 if (!err) 756 err = truncate_partial_data_page(inode, from, truncate_page); 757 758 trace_f2fs_truncate_blocks_exit(inode, err); 759 return err; 760 } 761 762 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 763 { 764 u64 free_from = from; 765 int err; 766 767 #ifdef CONFIG_F2FS_FS_COMPRESSION 768 /* 769 * for compressed file, only support cluster size 770 * aligned truncation. 771 */ 772 if (f2fs_compressed_file(inode)) 773 free_from = round_up(from, 774 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 775 #endif 776 777 err = f2fs_do_truncate_blocks(inode, free_from, lock); 778 if (err) 779 return err; 780 781 #ifdef CONFIG_F2FS_FS_COMPRESSION 782 /* 783 * For compressed file, after release compress blocks, don't allow write 784 * direct, but we should allow write direct after truncate to zero. 785 */ 786 if (f2fs_compressed_file(inode) && !free_from 787 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 788 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 789 790 if (from != free_from) { 791 err = f2fs_truncate_partial_cluster(inode, from, lock); 792 if (err) 793 return err; 794 } 795 #endif 796 797 return 0; 798 } 799 800 int f2fs_truncate(struct inode *inode) 801 { 802 int err; 803 804 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 805 return -EIO; 806 807 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 808 S_ISLNK(inode->i_mode))) 809 return 0; 810 811 trace_f2fs_truncate(inode); 812 813 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 814 return -EIO; 815 816 err = f2fs_dquot_initialize(inode); 817 if (err) 818 return err; 819 820 /* we should check inline_data size */ 821 if (!f2fs_may_inline_data(inode)) { 822 err = f2fs_convert_inline_inode(inode); 823 if (err) 824 return err; 825 } 826 827 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 828 if (err) 829 return err; 830 831 inode->i_mtime = inode_set_ctime_current(inode); 832 f2fs_mark_inode_dirty_sync(inode, false); 833 return 0; 834 } 835 836 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 837 { 838 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 839 840 if (!fscrypt_dio_supported(inode)) 841 return true; 842 if (fsverity_active(inode)) 843 return true; 844 if (f2fs_compressed_file(inode)) 845 return true; 846 if (f2fs_has_inline_data(inode)) 847 return true; 848 849 /* disallow direct IO if any of devices has unaligned blksize */ 850 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 851 return true; 852 /* 853 * for blkzoned device, fallback direct IO to buffered IO, so 854 * all IOs can be serialized by log-structured write. 855 */ 856 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE)) 857 return true; 858 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 859 return true; 860 861 return false; 862 } 863 864 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 865 struct kstat *stat, u32 request_mask, unsigned int query_flags) 866 { 867 struct inode *inode = d_inode(path->dentry); 868 struct f2fs_inode_info *fi = F2FS_I(inode); 869 struct f2fs_inode *ri = NULL; 870 unsigned int flags; 871 872 if (f2fs_has_extra_attr(inode) && 873 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 874 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 875 stat->result_mask |= STATX_BTIME; 876 stat->btime.tv_sec = fi->i_crtime.tv_sec; 877 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 878 } 879 880 /* 881 * Return the DIO alignment restrictions if requested. We only return 882 * this information when requested, since on encrypted files it might 883 * take a fair bit of work to get if the file wasn't opened recently. 884 * 885 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 886 * cannot represent that, so in that case we report no DIO support. 887 */ 888 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 889 unsigned int bsize = i_blocksize(inode); 890 891 stat->result_mask |= STATX_DIOALIGN; 892 if (!f2fs_force_buffered_io(inode, WRITE)) { 893 stat->dio_mem_align = bsize; 894 stat->dio_offset_align = bsize; 895 } 896 } 897 898 flags = fi->i_flags; 899 if (flags & F2FS_COMPR_FL) 900 stat->attributes |= STATX_ATTR_COMPRESSED; 901 if (flags & F2FS_APPEND_FL) 902 stat->attributes |= STATX_ATTR_APPEND; 903 if (IS_ENCRYPTED(inode)) 904 stat->attributes |= STATX_ATTR_ENCRYPTED; 905 if (flags & F2FS_IMMUTABLE_FL) 906 stat->attributes |= STATX_ATTR_IMMUTABLE; 907 if (flags & F2FS_NODUMP_FL) 908 stat->attributes |= STATX_ATTR_NODUMP; 909 if (IS_VERITY(inode)) 910 stat->attributes |= STATX_ATTR_VERITY; 911 912 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 913 STATX_ATTR_APPEND | 914 STATX_ATTR_ENCRYPTED | 915 STATX_ATTR_IMMUTABLE | 916 STATX_ATTR_NODUMP | 917 STATX_ATTR_VERITY); 918 919 generic_fillattr(idmap, request_mask, inode, stat); 920 921 /* we need to show initial sectors used for inline_data/dentries */ 922 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 923 f2fs_has_inline_dentry(inode)) 924 stat->blocks += (stat->size + 511) >> 9; 925 926 return 0; 927 } 928 929 #ifdef CONFIG_F2FS_FS_POSIX_ACL 930 static void __setattr_copy(struct mnt_idmap *idmap, 931 struct inode *inode, const struct iattr *attr) 932 { 933 unsigned int ia_valid = attr->ia_valid; 934 935 i_uid_update(idmap, attr, inode); 936 i_gid_update(idmap, attr, inode); 937 if (ia_valid & ATTR_ATIME) 938 inode->i_atime = attr->ia_atime; 939 if (ia_valid & ATTR_MTIME) 940 inode->i_mtime = attr->ia_mtime; 941 if (ia_valid & ATTR_CTIME) 942 inode_set_ctime_to_ts(inode, attr->ia_ctime); 943 if (ia_valid & ATTR_MODE) { 944 umode_t mode = attr->ia_mode; 945 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode); 946 947 if (!vfsgid_in_group_p(vfsgid) && 948 !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) 949 mode &= ~S_ISGID; 950 set_acl_inode(inode, mode); 951 } 952 } 953 #else 954 #define __setattr_copy setattr_copy 955 #endif 956 957 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 958 struct iattr *attr) 959 { 960 struct inode *inode = d_inode(dentry); 961 int err; 962 963 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 964 return -EIO; 965 966 if (unlikely(IS_IMMUTABLE(inode))) 967 return -EPERM; 968 969 if (unlikely(IS_APPEND(inode) && 970 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 971 ATTR_GID | ATTR_TIMES_SET)))) 972 return -EPERM; 973 974 if ((attr->ia_valid & ATTR_SIZE)) { 975 if (!f2fs_is_compress_backend_ready(inode)) 976 return -EOPNOTSUPP; 977 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 978 !IS_ALIGNED(attr->ia_size, 979 F2FS_BLK_TO_BYTES(F2FS_I(inode)->i_cluster_size))) 980 return -EINVAL; 981 } 982 983 err = setattr_prepare(idmap, dentry, attr); 984 if (err) 985 return err; 986 987 err = fscrypt_prepare_setattr(dentry, attr); 988 if (err) 989 return err; 990 991 err = fsverity_prepare_setattr(dentry, attr); 992 if (err) 993 return err; 994 995 if (is_quota_modification(idmap, inode, attr)) { 996 err = f2fs_dquot_initialize(inode); 997 if (err) 998 return err; 999 } 1000 if (i_uid_needs_update(idmap, attr, inode) || 1001 i_gid_needs_update(idmap, attr, inode)) { 1002 f2fs_lock_op(F2FS_I_SB(inode)); 1003 err = dquot_transfer(idmap, inode, attr); 1004 if (err) { 1005 set_sbi_flag(F2FS_I_SB(inode), 1006 SBI_QUOTA_NEED_REPAIR); 1007 f2fs_unlock_op(F2FS_I_SB(inode)); 1008 return err; 1009 } 1010 /* 1011 * update uid/gid under lock_op(), so that dquot and inode can 1012 * be updated atomically. 1013 */ 1014 i_uid_update(idmap, attr, inode); 1015 i_gid_update(idmap, attr, inode); 1016 f2fs_mark_inode_dirty_sync(inode, true); 1017 f2fs_unlock_op(F2FS_I_SB(inode)); 1018 } 1019 1020 if (attr->ia_valid & ATTR_SIZE) { 1021 loff_t old_size = i_size_read(inode); 1022 1023 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1024 /* 1025 * should convert inline inode before i_size_write to 1026 * keep smaller than inline_data size with inline flag. 1027 */ 1028 err = f2fs_convert_inline_inode(inode); 1029 if (err) 1030 return err; 1031 } 1032 1033 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1034 filemap_invalidate_lock(inode->i_mapping); 1035 1036 truncate_setsize(inode, attr->ia_size); 1037 1038 if (attr->ia_size <= old_size) 1039 err = f2fs_truncate(inode); 1040 /* 1041 * do not trim all blocks after i_size if target size is 1042 * larger than i_size. 1043 */ 1044 filemap_invalidate_unlock(inode->i_mapping); 1045 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1046 if (err) 1047 return err; 1048 1049 spin_lock(&F2FS_I(inode)->i_size_lock); 1050 inode->i_mtime = inode_set_ctime_current(inode); 1051 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1052 spin_unlock(&F2FS_I(inode)->i_size_lock); 1053 } 1054 1055 __setattr_copy(idmap, inode, attr); 1056 1057 if (attr->ia_valid & ATTR_MODE) { 1058 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1059 1060 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1061 if (!err) 1062 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1063 clear_inode_flag(inode, FI_ACL_MODE); 1064 } 1065 } 1066 1067 /* file size may changed here */ 1068 f2fs_mark_inode_dirty_sync(inode, true); 1069 1070 /* inode change will produce dirty node pages flushed by checkpoint */ 1071 f2fs_balance_fs(F2FS_I_SB(inode), true); 1072 1073 return err; 1074 } 1075 1076 const struct inode_operations f2fs_file_inode_operations = { 1077 .getattr = f2fs_getattr, 1078 .setattr = f2fs_setattr, 1079 .get_inode_acl = f2fs_get_acl, 1080 .set_acl = f2fs_set_acl, 1081 .listxattr = f2fs_listxattr, 1082 .fiemap = f2fs_fiemap, 1083 .fileattr_get = f2fs_fileattr_get, 1084 .fileattr_set = f2fs_fileattr_set, 1085 }; 1086 1087 static int fill_zero(struct inode *inode, pgoff_t index, 1088 loff_t start, loff_t len) 1089 { 1090 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1091 struct page *page; 1092 1093 if (!len) 1094 return 0; 1095 1096 f2fs_balance_fs(sbi, true); 1097 1098 f2fs_lock_op(sbi); 1099 page = f2fs_get_new_data_page(inode, NULL, index, false); 1100 f2fs_unlock_op(sbi); 1101 1102 if (IS_ERR(page)) 1103 return PTR_ERR(page); 1104 1105 f2fs_wait_on_page_writeback(page, DATA, true, true); 1106 zero_user(page, start, len); 1107 set_page_dirty(page); 1108 f2fs_put_page(page, 1); 1109 return 0; 1110 } 1111 1112 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1113 { 1114 int err; 1115 1116 while (pg_start < pg_end) { 1117 struct dnode_of_data dn; 1118 pgoff_t end_offset, count; 1119 1120 set_new_dnode(&dn, inode, NULL, NULL, 0); 1121 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1122 if (err) { 1123 if (err == -ENOENT) { 1124 pg_start = f2fs_get_next_page_offset(&dn, 1125 pg_start); 1126 continue; 1127 } 1128 return err; 1129 } 1130 1131 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1132 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1133 1134 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1135 1136 f2fs_truncate_data_blocks_range(&dn, count); 1137 f2fs_put_dnode(&dn); 1138 1139 pg_start += count; 1140 } 1141 return 0; 1142 } 1143 1144 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1145 { 1146 pgoff_t pg_start, pg_end; 1147 loff_t off_start, off_end; 1148 int ret; 1149 1150 ret = f2fs_convert_inline_inode(inode); 1151 if (ret) 1152 return ret; 1153 1154 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1155 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1156 1157 off_start = offset & (PAGE_SIZE - 1); 1158 off_end = (offset + len) & (PAGE_SIZE - 1); 1159 1160 if (pg_start == pg_end) { 1161 ret = fill_zero(inode, pg_start, off_start, 1162 off_end - off_start); 1163 if (ret) 1164 return ret; 1165 } else { 1166 if (off_start) { 1167 ret = fill_zero(inode, pg_start++, off_start, 1168 PAGE_SIZE - off_start); 1169 if (ret) 1170 return ret; 1171 } 1172 if (off_end) { 1173 ret = fill_zero(inode, pg_end, 0, off_end); 1174 if (ret) 1175 return ret; 1176 } 1177 1178 if (pg_start < pg_end) { 1179 loff_t blk_start, blk_end; 1180 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1181 1182 f2fs_balance_fs(sbi, true); 1183 1184 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1185 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1186 1187 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1188 filemap_invalidate_lock(inode->i_mapping); 1189 1190 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1191 1192 f2fs_lock_op(sbi); 1193 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1194 f2fs_unlock_op(sbi); 1195 1196 filemap_invalidate_unlock(inode->i_mapping); 1197 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1198 } 1199 } 1200 1201 return ret; 1202 } 1203 1204 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1205 int *do_replace, pgoff_t off, pgoff_t len) 1206 { 1207 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1208 struct dnode_of_data dn; 1209 int ret, done, i; 1210 1211 next_dnode: 1212 set_new_dnode(&dn, inode, NULL, NULL, 0); 1213 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1214 if (ret && ret != -ENOENT) { 1215 return ret; 1216 } else if (ret == -ENOENT) { 1217 if (dn.max_level == 0) 1218 return -ENOENT; 1219 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1220 dn.ofs_in_node, len); 1221 blkaddr += done; 1222 do_replace += done; 1223 goto next; 1224 } 1225 1226 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1227 dn.ofs_in_node, len); 1228 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1229 *blkaddr = f2fs_data_blkaddr(&dn); 1230 1231 if (__is_valid_data_blkaddr(*blkaddr) && 1232 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1233 DATA_GENERIC_ENHANCE)) { 1234 f2fs_put_dnode(&dn); 1235 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1236 return -EFSCORRUPTED; 1237 } 1238 1239 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1240 1241 if (f2fs_lfs_mode(sbi)) { 1242 f2fs_put_dnode(&dn); 1243 return -EOPNOTSUPP; 1244 } 1245 1246 /* do not invalidate this block address */ 1247 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1248 *do_replace = 1; 1249 } 1250 } 1251 f2fs_put_dnode(&dn); 1252 next: 1253 len -= done; 1254 off += done; 1255 if (len) 1256 goto next_dnode; 1257 return 0; 1258 } 1259 1260 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1261 int *do_replace, pgoff_t off, int len) 1262 { 1263 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1264 struct dnode_of_data dn; 1265 int ret, i; 1266 1267 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1268 if (*do_replace == 0) 1269 continue; 1270 1271 set_new_dnode(&dn, inode, NULL, NULL, 0); 1272 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1273 if (ret) { 1274 dec_valid_block_count(sbi, inode, 1); 1275 f2fs_invalidate_blocks(sbi, *blkaddr); 1276 } else { 1277 f2fs_update_data_blkaddr(&dn, *blkaddr); 1278 } 1279 f2fs_put_dnode(&dn); 1280 } 1281 return 0; 1282 } 1283 1284 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1285 block_t *blkaddr, int *do_replace, 1286 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1287 { 1288 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1289 pgoff_t i = 0; 1290 int ret; 1291 1292 while (i < len) { 1293 if (blkaddr[i] == NULL_ADDR && !full) { 1294 i++; 1295 continue; 1296 } 1297 1298 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1299 struct dnode_of_data dn; 1300 struct node_info ni; 1301 size_t new_size; 1302 pgoff_t ilen; 1303 1304 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1305 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1306 if (ret) 1307 return ret; 1308 1309 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1310 if (ret) { 1311 f2fs_put_dnode(&dn); 1312 return ret; 1313 } 1314 1315 ilen = min((pgoff_t) 1316 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1317 dn.ofs_in_node, len - i); 1318 do { 1319 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1320 f2fs_truncate_data_blocks_range(&dn, 1); 1321 1322 if (do_replace[i]) { 1323 f2fs_i_blocks_write(src_inode, 1324 1, false, false); 1325 f2fs_i_blocks_write(dst_inode, 1326 1, true, false); 1327 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1328 blkaddr[i], ni.version, true, false); 1329 1330 do_replace[i] = 0; 1331 } 1332 dn.ofs_in_node++; 1333 i++; 1334 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1335 if (dst_inode->i_size < new_size) 1336 f2fs_i_size_write(dst_inode, new_size); 1337 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1338 1339 f2fs_put_dnode(&dn); 1340 } else { 1341 struct page *psrc, *pdst; 1342 1343 psrc = f2fs_get_lock_data_page(src_inode, 1344 src + i, true); 1345 if (IS_ERR(psrc)) 1346 return PTR_ERR(psrc); 1347 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1348 true); 1349 if (IS_ERR(pdst)) { 1350 f2fs_put_page(psrc, 1); 1351 return PTR_ERR(pdst); 1352 } 1353 1354 f2fs_wait_on_page_writeback(pdst, DATA, true, true); 1355 1356 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1357 set_page_dirty(pdst); 1358 set_page_private_gcing(pdst); 1359 f2fs_put_page(pdst, 1); 1360 f2fs_put_page(psrc, 1); 1361 1362 ret = f2fs_truncate_hole(src_inode, 1363 src + i, src + i + 1); 1364 if (ret) 1365 return ret; 1366 i++; 1367 } 1368 } 1369 return 0; 1370 } 1371 1372 static int __exchange_data_block(struct inode *src_inode, 1373 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1374 pgoff_t len, bool full) 1375 { 1376 block_t *src_blkaddr; 1377 int *do_replace; 1378 pgoff_t olen; 1379 int ret; 1380 1381 while (len) { 1382 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1383 1384 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1385 array_size(olen, sizeof(block_t)), 1386 GFP_NOFS); 1387 if (!src_blkaddr) 1388 return -ENOMEM; 1389 1390 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1391 array_size(olen, sizeof(int)), 1392 GFP_NOFS); 1393 if (!do_replace) { 1394 kvfree(src_blkaddr); 1395 return -ENOMEM; 1396 } 1397 1398 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1399 do_replace, src, olen); 1400 if (ret) 1401 goto roll_back; 1402 1403 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1404 do_replace, src, dst, olen, full); 1405 if (ret) 1406 goto roll_back; 1407 1408 src += olen; 1409 dst += olen; 1410 len -= olen; 1411 1412 kvfree(src_blkaddr); 1413 kvfree(do_replace); 1414 } 1415 return 0; 1416 1417 roll_back: 1418 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1419 kvfree(src_blkaddr); 1420 kvfree(do_replace); 1421 return ret; 1422 } 1423 1424 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1425 { 1426 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1427 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1428 pgoff_t start = offset >> PAGE_SHIFT; 1429 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1430 int ret; 1431 1432 f2fs_balance_fs(sbi, true); 1433 1434 /* avoid gc operation during block exchange */ 1435 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1436 filemap_invalidate_lock(inode->i_mapping); 1437 1438 f2fs_lock_op(sbi); 1439 f2fs_drop_extent_tree(inode); 1440 truncate_pagecache(inode, offset); 1441 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1442 f2fs_unlock_op(sbi); 1443 1444 filemap_invalidate_unlock(inode->i_mapping); 1445 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1446 return ret; 1447 } 1448 1449 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1450 { 1451 loff_t new_size; 1452 int ret; 1453 1454 if (offset + len >= i_size_read(inode)) 1455 return -EINVAL; 1456 1457 /* collapse range should be aligned to block size of f2fs. */ 1458 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1459 return -EINVAL; 1460 1461 ret = f2fs_convert_inline_inode(inode); 1462 if (ret) 1463 return ret; 1464 1465 /* write out all dirty pages from offset */ 1466 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1467 if (ret) 1468 return ret; 1469 1470 ret = f2fs_do_collapse(inode, offset, len); 1471 if (ret) 1472 return ret; 1473 1474 /* write out all moved pages, if possible */ 1475 filemap_invalidate_lock(inode->i_mapping); 1476 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1477 truncate_pagecache(inode, offset); 1478 1479 new_size = i_size_read(inode) - len; 1480 ret = f2fs_truncate_blocks(inode, new_size, true); 1481 filemap_invalidate_unlock(inode->i_mapping); 1482 if (!ret) 1483 f2fs_i_size_write(inode, new_size); 1484 return ret; 1485 } 1486 1487 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1488 pgoff_t end) 1489 { 1490 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1491 pgoff_t index = start; 1492 unsigned int ofs_in_node = dn->ofs_in_node; 1493 blkcnt_t count = 0; 1494 int ret; 1495 1496 for (; index < end; index++, dn->ofs_in_node++) { 1497 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1498 count++; 1499 } 1500 1501 dn->ofs_in_node = ofs_in_node; 1502 ret = f2fs_reserve_new_blocks(dn, count); 1503 if (ret) 1504 return ret; 1505 1506 dn->ofs_in_node = ofs_in_node; 1507 for (index = start; index < end; index++, dn->ofs_in_node++) { 1508 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1509 /* 1510 * f2fs_reserve_new_blocks will not guarantee entire block 1511 * allocation. 1512 */ 1513 if (dn->data_blkaddr == NULL_ADDR) { 1514 ret = -ENOSPC; 1515 break; 1516 } 1517 1518 if (dn->data_blkaddr == NEW_ADDR) 1519 continue; 1520 1521 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1522 DATA_GENERIC_ENHANCE)) { 1523 ret = -EFSCORRUPTED; 1524 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1525 break; 1526 } 1527 1528 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1529 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1530 } 1531 1532 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1533 f2fs_update_age_extent_cache_range(dn, start, index - start); 1534 1535 return ret; 1536 } 1537 1538 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1539 int mode) 1540 { 1541 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1542 struct address_space *mapping = inode->i_mapping; 1543 pgoff_t index, pg_start, pg_end; 1544 loff_t new_size = i_size_read(inode); 1545 loff_t off_start, off_end; 1546 int ret = 0; 1547 1548 ret = inode_newsize_ok(inode, (len + offset)); 1549 if (ret) 1550 return ret; 1551 1552 ret = f2fs_convert_inline_inode(inode); 1553 if (ret) 1554 return ret; 1555 1556 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1557 if (ret) 1558 return ret; 1559 1560 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1561 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1562 1563 off_start = offset & (PAGE_SIZE - 1); 1564 off_end = (offset + len) & (PAGE_SIZE - 1); 1565 1566 if (pg_start == pg_end) { 1567 ret = fill_zero(inode, pg_start, off_start, 1568 off_end - off_start); 1569 if (ret) 1570 return ret; 1571 1572 new_size = max_t(loff_t, new_size, offset + len); 1573 } else { 1574 if (off_start) { 1575 ret = fill_zero(inode, pg_start++, off_start, 1576 PAGE_SIZE - off_start); 1577 if (ret) 1578 return ret; 1579 1580 new_size = max_t(loff_t, new_size, 1581 (loff_t)pg_start << PAGE_SHIFT); 1582 } 1583 1584 for (index = pg_start; index < pg_end;) { 1585 struct dnode_of_data dn; 1586 unsigned int end_offset; 1587 pgoff_t end; 1588 1589 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1590 filemap_invalidate_lock(mapping); 1591 1592 truncate_pagecache_range(inode, 1593 (loff_t)index << PAGE_SHIFT, 1594 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1595 1596 f2fs_lock_op(sbi); 1597 1598 set_new_dnode(&dn, inode, NULL, NULL, 0); 1599 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1600 if (ret) { 1601 f2fs_unlock_op(sbi); 1602 filemap_invalidate_unlock(mapping); 1603 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1604 goto out; 1605 } 1606 1607 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1608 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1609 1610 ret = f2fs_do_zero_range(&dn, index, end); 1611 f2fs_put_dnode(&dn); 1612 1613 f2fs_unlock_op(sbi); 1614 filemap_invalidate_unlock(mapping); 1615 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1616 1617 f2fs_balance_fs(sbi, dn.node_changed); 1618 1619 if (ret) 1620 goto out; 1621 1622 index = end; 1623 new_size = max_t(loff_t, new_size, 1624 (loff_t)index << PAGE_SHIFT); 1625 } 1626 1627 if (off_end) { 1628 ret = fill_zero(inode, pg_end, 0, off_end); 1629 if (ret) 1630 goto out; 1631 1632 new_size = max_t(loff_t, new_size, offset + len); 1633 } 1634 } 1635 1636 out: 1637 if (new_size > i_size_read(inode)) { 1638 if (mode & FALLOC_FL_KEEP_SIZE) 1639 file_set_keep_isize(inode); 1640 else 1641 f2fs_i_size_write(inode, new_size); 1642 } 1643 return ret; 1644 } 1645 1646 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1647 { 1648 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1649 struct address_space *mapping = inode->i_mapping; 1650 pgoff_t nr, pg_start, pg_end, delta, idx; 1651 loff_t new_size; 1652 int ret = 0; 1653 1654 new_size = i_size_read(inode) + len; 1655 ret = inode_newsize_ok(inode, new_size); 1656 if (ret) 1657 return ret; 1658 1659 if (offset >= i_size_read(inode)) 1660 return -EINVAL; 1661 1662 /* insert range should be aligned to block size of f2fs. */ 1663 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1664 return -EINVAL; 1665 1666 ret = f2fs_convert_inline_inode(inode); 1667 if (ret) 1668 return ret; 1669 1670 f2fs_balance_fs(sbi, true); 1671 1672 filemap_invalidate_lock(mapping); 1673 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1674 filemap_invalidate_unlock(mapping); 1675 if (ret) 1676 return ret; 1677 1678 /* write out all dirty pages from offset */ 1679 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1680 if (ret) 1681 return ret; 1682 1683 pg_start = offset >> PAGE_SHIFT; 1684 pg_end = (offset + len) >> PAGE_SHIFT; 1685 delta = pg_end - pg_start; 1686 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1687 1688 /* avoid gc operation during block exchange */ 1689 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1690 filemap_invalidate_lock(mapping); 1691 truncate_pagecache(inode, offset); 1692 1693 while (!ret && idx > pg_start) { 1694 nr = idx - pg_start; 1695 if (nr > delta) 1696 nr = delta; 1697 idx -= nr; 1698 1699 f2fs_lock_op(sbi); 1700 f2fs_drop_extent_tree(inode); 1701 1702 ret = __exchange_data_block(inode, inode, idx, 1703 idx + delta, nr, false); 1704 f2fs_unlock_op(sbi); 1705 } 1706 filemap_invalidate_unlock(mapping); 1707 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1708 1709 /* write out all moved pages, if possible */ 1710 filemap_invalidate_lock(mapping); 1711 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1712 truncate_pagecache(inode, offset); 1713 filemap_invalidate_unlock(mapping); 1714 1715 if (!ret) 1716 f2fs_i_size_write(inode, new_size); 1717 return ret; 1718 } 1719 1720 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1721 loff_t len, int mode) 1722 { 1723 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1724 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1725 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1726 .m_may_create = true }; 1727 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1728 .init_gc_type = FG_GC, 1729 .should_migrate_blocks = false, 1730 .err_gc_skipped = true, 1731 .nr_free_secs = 0 }; 1732 pgoff_t pg_start, pg_end; 1733 loff_t new_size; 1734 loff_t off_end; 1735 block_t expanded = 0; 1736 int err; 1737 1738 err = inode_newsize_ok(inode, (len + offset)); 1739 if (err) 1740 return err; 1741 1742 err = f2fs_convert_inline_inode(inode); 1743 if (err) 1744 return err; 1745 1746 f2fs_balance_fs(sbi, true); 1747 1748 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1749 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1750 off_end = (offset + len) & (PAGE_SIZE - 1); 1751 1752 map.m_lblk = pg_start; 1753 map.m_len = pg_end - pg_start; 1754 if (off_end) 1755 map.m_len++; 1756 1757 if (!map.m_len) 1758 return 0; 1759 1760 if (f2fs_is_pinned_file(inode)) { 1761 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1762 block_t sec_len = roundup(map.m_len, sec_blks); 1763 1764 map.m_len = sec_blks; 1765 next_alloc: 1766 if (has_not_enough_free_secs(sbi, 0, 1767 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1768 f2fs_down_write(&sbi->gc_lock); 1769 stat_inc_gc_call_count(sbi, FOREGROUND); 1770 err = f2fs_gc(sbi, &gc_control); 1771 if (err && err != -ENODATA) 1772 goto out_err; 1773 } 1774 1775 f2fs_down_write(&sbi->pin_sem); 1776 1777 err = f2fs_allocate_pinning_section(sbi); 1778 if (err) { 1779 f2fs_up_write(&sbi->pin_sem); 1780 goto out_err; 1781 } 1782 1783 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1784 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1785 file_dont_truncate(inode); 1786 1787 f2fs_up_write(&sbi->pin_sem); 1788 1789 expanded += map.m_len; 1790 sec_len -= map.m_len; 1791 map.m_lblk += map.m_len; 1792 if (!err && sec_len) 1793 goto next_alloc; 1794 1795 map.m_len = expanded; 1796 } else { 1797 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1798 expanded = map.m_len; 1799 } 1800 out_err: 1801 if (err) { 1802 pgoff_t last_off; 1803 1804 if (!expanded) 1805 return err; 1806 1807 last_off = pg_start + expanded - 1; 1808 1809 /* update new size to the failed position */ 1810 new_size = (last_off == pg_end) ? offset + len : 1811 (loff_t)(last_off + 1) << PAGE_SHIFT; 1812 } else { 1813 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1814 } 1815 1816 if (new_size > i_size_read(inode)) { 1817 if (mode & FALLOC_FL_KEEP_SIZE) 1818 file_set_keep_isize(inode); 1819 else 1820 f2fs_i_size_write(inode, new_size); 1821 } 1822 1823 return err; 1824 } 1825 1826 static long f2fs_fallocate(struct file *file, int mode, 1827 loff_t offset, loff_t len) 1828 { 1829 struct inode *inode = file_inode(file); 1830 long ret = 0; 1831 1832 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1833 return -EIO; 1834 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1835 return -ENOSPC; 1836 if (!f2fs_is_compress_backend_ready(inode)) 1837 return -EOPNOTSUPP; 1838 1839 /* f2fs only support ->fallocate for regular file */ 1840 if (!S_ISREG(inode->i_mode)) 1841 return -EINVAL; 1842 1843 if (IS_ENCRYPTED(inode) && 1844 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1845 return -EOPNOTSUPP; 1846 1847 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1848 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1849 FALLOC_FL_INSERT_RANGE)) 1850 return -EOPNOTSUPP; 1851 1852 inode_lock(inode); 1853 1854 /* 1855 * Pinned file should not support partial truncation since the block 1856 * can be used by applications. 1857 */ 1858 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1859 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1860 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 1861 ret = -EOPNOTSUPP; 1862 goto out; 1863 } 1864 1865 ret = file_modified(file); 1866 if (ret) 1867 goto out; 1868 1869 if (mode & FALLOC_FL_PUNCH_HOLE) { 1870 if (offset >= inode->i_size) 1871 goto out; 1872 1873 ret = f2fs_punch_hole(inode, offset, len); 1874 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1875 ret = f2fs_collapse_range(inode, offset, len); 1876 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1877 ret = f2fs_zero_range(inode, offset, len, mode); 1878 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1879 ret = f2fs_insert_range(inode, offset, len); 1880 } else { 1881 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1882 } 1883 1884 if (!ret) { 1885 inode->i_mtime = inode_set_ctime_current(inode); 1886 f2fs_mark_inode_dirty_sync(inode, false); 1887 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1888 } 1889 1890 out: 1891 inode_unlock(inode); 1892 1893 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1894 return ret; 1895 } 1896 1897 static int f2fs_release_file(struct inode *inode, struct file *filp) 1898 { 1899 /* 1900 * f2fs_release_file is called at every close calls. So we should 1901 * not drop any inmemory pages by close called by other process. 1902 */ 1903 if (!(filp->f_mode & FMODE_WRITE) || 1904 atomic_read(&inode->i_writecount) != 1) 1905 return 0; 1906 1907 inode_lock(inode); 1908 f2fs_abort_atomic_write(inode, true); 1909 inode_unlock(inode); 1910 1911 return 0; 1912 } 1913 1914 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1915 { 1916 struct inode *inode = file_inode(file); 1917 1918 /* 1919 * If the process doing a transaction is crashed, we should do 1920 * roll-back. Otherwise, other reader/write can see corrupted database 1921 * until all the writers close its file. Since this should be done 1922 * before dropping file lock, it needs to do in ->flush. 1923 */ 1924 if (F2FS_I(inode)->atomic_write_task == current && 1925 (current->flags & PF_EXITING)) { 1926 inode_lock(inode); 1927 f2fs_abort_atomic_write(inode, true); 1928 inode_unlock(inode); 1929 } 1930 1931 return 0; 1932 } 1933 1934 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1935 { 1936 struct f2fs_inode_info *fi = F2FS_I(inode); 1937 u32 masked_flags = fi->i_flags & mask; 1938 1939 /* mask can be shrunk by flags_valid selector */ 1940 iflags &= mask; 1941 1942 /* Is it quota file? Do not allow user to mess with it */ 1943 if (IS_NOQUOTA(inode)) 1944 return -EPERM; 1945 1946 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1947 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1948 return -EOPNOTSUPP; 1949 if (!f2fs_empty_dir(inode)) 1950 return -ENOTEMPTY; 1951 } 1952 1953 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1954 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1955 return -EOPNOTSUPP; 1956 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1957 return -EINVAL; 1958 } 1959 1960 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1961 if (masked_flags & F2FS_COMPR_FL) { 1962 if (!f2fs_disable_compressed_file(inode)) 1963 return -EINVAL; 1964 } else { 1965 /* try to convert inline_data to support compression */ 1966 int err = f2fs_convert_inline_inode(inode); 1967 if (err) 1968 return err; 1969 1970 f2fs_down_write(&F2FS_I(inode)->i_sem); 1971 if (!f2fs_may_compress(inode) || 1972 (S_ISREG(inode->i_mode) && 1973 F2FS_HAS_BLOCKS(inode))) { 1974 f2fs_up_write(&F2FS_I(inode)->i_sem); 1975 return -EINVAL; 1976 } 1977 err = set_compress_context(inode); 1978 f2fs_up_write(&F2FS_I(inode)->i_sem); 1979 1980 if (err) 1981 return err; 1982 } 1983 } 1984 1985 fi->i_flags = iflags | (fi->i_flags & ~mask); 1986 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1987 (fi->i_flags & F2FS_NOCOMP_FL)); 1988 1989 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1990 set_inode_flag(inode, FI_PROJ_INHERIT); 1991 else 1992 clear_inode_flag(inode, FI_PROJ_INHERIT); 1993 1994 inode_set_ctime_current(inode); 1995 f2fs_set_inode_flags(inode); 1996 f2fs_mark_inode_dirty_sync(inode, true); 1997 return 0; 1998 } 1999 2000 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 2001 2002 /* 2003 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 2004 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 2005 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 2006 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 2007 * 2008 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 2009 * FS_IOC_FSSETXATTR is done by the VFS. 2010 */ 2011 2012 static const struct { 2013 u32 iflag; 2014 u32 fsflag; 2015 } f2fs_fsflags_map[] = { 2016 { F2FS_COMPR_FL, FS_COMPR_FL }, 2017 { F2FS_SYNC_FL, FS_SYNC_FL }, 2018 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 2019 { F2FS_APPEND_FL, FS_APPEND_FL }, 2020 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 2021 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 2022 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 2023 { F2FS_INDEX_FL, FS_INDEX_FL }, 2024 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 2025 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2026 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2027 }; 2028 2029 #define F2FS_GETTABLE_FS_FL ( \ 2030 FS_COMPR_FL | \ 2031 FS_SYNC_FL | \ 2032 FS_IMMUTABLE_FL | \ 2033 FS_APPEND_FL | \ 2034 FS_NODUMP_FL | \ 2035 FS_NOATIME_FL | \ 2036 FS_NOCOMP_FL | \ 2037 FS_INDEX_FL | \ 2038 FS_DIRSYNC_FL | \ 2039 FS_PROJINHERIT_FL | \ 2040 FS_ENCRYPT_FL | \ 2041 FS_INLINE_DATA_FL | \ 2042 FS_NOCOW_FL | \ 2043 FS_VERITY_FL | \ 2044 FS_CASEFOLD_FL) 2045 2046 #define F2FS_SETTABLE_FS_FL ( \ 2047 FS_COMPR_FL | \ 2048 FS_SYNC_FL | \ 2049 FS_IMMUTABLE_FL | \ 2050 FS_APPEND_FL | \ 2051 FS_NODUMP_FL | \ 2052 FS_NOATIME_FL | \ 2053 FS_NOCOMP_FL | \ 2054 FS_DIRSYNC_FL | \ 2055 FS_PROJINHERIT_FL | \ 2056 FS_CASEFOLD_FL) 2057 2058 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2059 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2060 { 2061 u32 fsflags = 0; 2062 int i; 2063 2064 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2065 if (iflags & f2fs_fsflags_map[i].iflag) 2066 fsflags |= f2fs_fsflags_map[i].fsflag; 2067 2068 return fsflags; 2069 } 2070 2071 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2072 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2073 { 2074 u32 iflags = 0; 2075 int i; 2076 2077 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2078 if (fsflags & f2fs_fsflags_map[i].fsflag) 2079 iflags |= f2fs_fsflags_map[i].iflag; 2080 2081 return iflags; 2082 } 2083 2084 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2085 { 2086 struct inode *inode = file_inode(filp); 2087 2088 return put_user(inode->i_generation, (int __user *)arg); 2089 } 2090 2091 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2092 { 2093 struct inode *inode = file_inode(filp); 2094 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2095 struct f2fs_inode_info *fi = F2FS_I(inode); 2096 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2097 struct inode *pinode; 2098 loff_t isize; 2099 int ret; 2100 2101 if (!inode_owner_or_capable(idmap, inode)) 2102 return -EACCES; 2103 2104 if (!S_ISREG(inode->i_mode)) 2105 return -EINVAL; 2106 2107 if (filp->f_flags & O_DIRECT) 2108 return -EINVAL; 2109 2110 ret = mnt_want_write_file(filp); 2111 if (ret) 2112 return ret; 2113 2114 inode_lock(inode); 2115 2116 if (!f2fs_disable_compressed_file(inode)) { 2117 ret = -EINVAL; 2118 goto out; 2119 } 2120 2121 if (f2fs_is_atomic_file(inode)) 2122 goto out; 2123 2124 ret = f2fs_convert_inline_inode(inode); 2125 if (ret) 2126 goto out; 2127 2128 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2129 2130 /* 2131 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2132 * f2fs_is_atomic_file. 2133 */ 2134 if (get_dirty_pages(inode)) 2135 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2136 inode->i_ino, get_dirty_pages(inode)); 2137 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2138 if (ret) { 2139 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2140 goto out; 2141 } 2142 2143 /* Check if the inode already has a COW inode */ 2144 if (fi->cow_inode == NULL) { 2145 /* Create a COW inode for atomic write */ 2146 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2147 if (IS_ERR(pinode)) { 2148 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2149 ret = PTR_ERR(pinode); 2150 goto out; 2151 } 2152 2153 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2154 iput(pinode); 2155 if (ret) { 2156 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2157 goto out; 2158 } 2159 2160 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2161 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2162 2163 /* Set the COW inode's atomic_inode to the atomic inode */ 2164 F2FS_I(fi->cow_inode)->atomic_inode = inode; 2165 } else { 2166 /* Reuse the already created COW inode */ 2167 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2168 if (ret) { 2169 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2170 goto out; 2171 } 2172 } 2173 2174 f2fs_write_inode(inode, NULL); 2175 2176 stat_inc_atomic_inode(inode); 2177 2178 set_inode_flag(inode, FI_ATOMIC_FILE); 2179 2180 isize = i_size_read(inode); 2181 fi->original_i_size = isize; 2182 if (truncate) { 2183 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2184 truncate_inode_pages_final(inode->i_mapping); 2185 f2fs_i_size_write(inode, 0); 2186 isize = 0; 2187 } 2188 f2fs_i_size_write(fi->cow_inode, isize); 2189 2190 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2191 2192 f2fs_update_time(sbi, REQ_TIME); 2193 fi->atomic_write_task = current; 2194 stat_update_max_atomic_write(inode); 2195 fi->atomic_write_cnt = 0; 2196 out: 2197 inode_unlock(inode); 2198 mnt_drop_write_file(filp); 2199 return ret; 2200 } 2201 2202 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2203 { 2204 struct inode *inode = file_inode(filp); 2205 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2206 int ret; 2207 2208 if (!inode_owner_or_capable(idmap, inode)) 2209 return -EACCES; 2210 2211 ret = mnt_want_write_file(filp); 2212 if (ret) 2213 return ret; 2214 2215 f2fs_balance_fs(F2FS_I_SB(inode), true); 2216 2217 inode_lock(inode); 2218 2219 if (f2fs_is_atomic_file(inode)) { 2220 ret = f2fs_commit_atomic_write(inode); 2221 if (!ret) 2222 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2223 2224 f2fs_abort_atomic_write(inode, ret); 2225 } else { 2226 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2227 } 2228 2229 inode_unlock(inode); 2230 mnt_drop_write_file(filp); 2231 return ret; 2232 } 2233 2234 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2235 { 2236 struct inode *inode = file_inode(filp); 2237 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2238 int ret; 2239 2240 if (!inode_owner_or_capable(idmap, inode)) 2241 return -EACCES; 2242 2243 ret = mnt_want_write_file(filp); 2244 if (ret) 2245 return ret; 2246 2247 inode_lock(inode); 2248 2249 f2fs_abort_atomic_write(inode, true); 2250 2251 inode_unlock(inode); 2252 2253 mnt_drop_write_file(filp); 2254 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2255 return ret; 2256 } 2257 2258 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2259 { 2260 struct inode *inode = file_inode(filp); 2261 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2262 struct super_block *sb = sbi->sb; 2263 __u32 in; 2264 int ret = 0; 2265 2266 if (!capable(CAP_SYS_ADMIN)) 2267 return -EPERM; 2268 2269 if (get_user(in, (__u32 __user *)arg)) 2270 return -EFAULT; 2271 2272 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2273 ret = mnt_want_write_file(filp); 2274 if (ret) { 2275 if (ret == -EROFS) { 2276 ret = 0; 2277 f2fs_stop_checkpoint(sbi, false, 2278 STOP_CP_REASON_SHUTDOWN); 2279 trace_f2fs_shutdown(sbi, in, ret); 2280 } 2281 return ret; 2282 } 2283 } 2284 2285 switch (in) { 2286 case F2FS_GOING_DOWN_FULLSYNC: 2287 ret = freeze_bdev(sb->s_bdev); 2288 if (ret) 2289 goto out; 2290 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2291 thaw_bdev(sb->s_bdev); 2292 break; 2293 case F2FS_GOING_DOWN_METASYNC: 2294 /* do checkpoint only */ 2295 ret = f2fs_sync_fs(sb, 1); 2296 if (ret) 2297 goto out; 2298 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2299 break; 2300 case F2FS_GOING_DOWN_NOSYNC: 2301 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2302 break; 2303 case F2FS_GOING_DOWN_METAFLUSH: 2304 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2305 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2306 break; 2307 case F2FS_GOING_DOWN_NEED_FSCK: 2308 set_sbi_flag(sbi, SBI_NEED_FSCK); 2309 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2310 set_sbi_flag(sbi, SBI_IS_DIRTY); 2311 /* do checkpoint only */ 2312 ret = f2fs_sync_fs(sb, 1); 2313 goto out; 2314 default: 2315 ret = -EINVAL; 2316 goto out; 2317 } 2318 2319 f2fs_stop_gc_thread(sbi); 2320 f2fs_stop_discard_thread(sbi); 2321 2322 f2fs_drop_discard_cmd(sbi); 2323 clear_opt(sbi, DISCARD); 2324 2325 f2fs_update_time(sbi, REQ_TIME); 2326 out: 2327 if (in != F2FS_GOING_DOWN_FULLSYNC) 2328 mnt_drop_write_file(filp); 2329 2330 trace_f2fs_shutdown(sbi, in, ret); 2331 2332 return ret; 2333 } 2334 2335 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2336 { 2337 struct inode *inode = file_inode(filp); 2338 struct super_block *sb = inode->i_sb; 2339 struct fstrim_range range; 2340 int ret; 2341 2342 if (!capable(CAP_SYS_ADMIN)) 2343 return -EPERM; 2344 2345 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2346 return -EOPNOTSUPP; 2347 2348 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2349 sizeof(range))) 2350 return -EFAULT; 2351 2352 ret = mnt_want_write_file(filp); 2353 if (ret) 2354 return ret; 2355 2356 range.minlen = max((unsigned int)range.minlen, 2357 bdev_discard_granularity(sb->s_bdev)); 2358 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2359 mnt_drop_write_file(filp); 2360 if (ret < 0) 2361 return ret; 2362 2363 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2364 sizeof(range))) 2365 return -EFAULT; 2366 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2367 return 0; 2368 } 2369 2370 static bool uuid_is_nonzero(__u8 u[16]) 2371 { 2372 int i; 2373 2374 for (i = 0; i < 16; i++) 2375 if (u[i]) 2376 return true; 2377 return false; 2378 } 2379 2380 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2381 { 2382 struct inode *inode = file_inode(filp); 2383 2384 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2385 return -EOPNOTSUPP; 2386 2387 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2388 2389 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2390 } 2391 2392 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2393 { 2394 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2395 return -EOPNOTSUPP; 2396 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2397 } 2398 2399 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2400 { 2401 struct inode *inode = file_inode(filp); 2402 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2403 u8 encrypt_pw_salt[16]; 2404 int err; 2405 2406 if (!f2fs_sb_has_encrypt(sbi)) 2407 return -EOPNOTSUPP; 2408 2409 err = mnt_want_write_file(filp); 2410 if (err) 2411 return err; 2412 2413 f2fs_down_write(&sbi->sb_lock); 2414 2415 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2416 goto got_it; 2417 2418 /* update superblock with uuid */ 2419 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2420 2421 err = f2fs_commit_super(sbi, false); 2422 if (err) { 2423 /* undo new data */ 2424 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2425 goto out_err; 2426 } 2427 got_it: 2428 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2429 out_err: 2430 f2fs_up_write(&sbi->sb_lock); 2431 mnt_drop_write_file(filp); 2432 2433 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2434 err = -EFAULT; 2435 2436 return err; 2437 } 2438 2439 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2440 unsigned long arg) 2441 { 2442 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2443 return -EOPNOTSUPP; 2444 2445 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2446 } 2447 2448 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2449 { 2450 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2451 return -EOPNOTSUPP; 2452 2453 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2454 } 2455 2456 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2457 { 2458 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2459 return -EOPNOTSUPP; 2460 2461 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2462 } 2463 2464 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2465 unsigned long arg) 2466 { 2467 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2468 return -EOPNOTSUPP; 2469 2470 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2471 } 2472 2473 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2474 unsigned long arg) 2475 { 2476 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2477 return -EOPNOTSUPP; 2478 2479 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2480 } 2481 2482 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2483 { 2484 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2485 return -EOPNOTSUPP; 2486 2487 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2488 } 2489 2490 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2491 { 2492 struct inode *inode = file_inode(filp); 2493 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2494 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2495 .no_bg_gc = false, 2496 .should_migrate_blocks = false, 2497 .nr_free_secs = 0 }; 2498 __u32 sync; 2499 int ret; 2500 2501 if (!capable(CAP_SYS_ADMIN)) 2502 return -EPERM; 2503 2504 if (get_user(sync, (__u32 __user *)arg)) 2505 return -EFAULT; 2506 2507 if (f2fs_readonly(sbi->sb)) 2508 return -EROFS; 2509 2510 ret = mnt_want_write_file(filp); 2511 if (ret) 2512 return ret; 2513 2514 if (!sync) { 2515 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2516 ret = -EBUSY; 2517 goto out; 2518 } 2519 } else { 2520 f2fs_down_write(&sbi->gc_lock); 2521 } 2522 2523 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2524 gc_control.err_gc_skipped = sync; 2525 stat_inc_gc_call_count(sbi, FOREGROUND); 2526 ret = f2fs_gc(sbi, &gc_control); 2527 out: 2528 mnt_drop_write_file(filp); 2529 return ret; 2530 } 2531 2532 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2533 { 2534 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2535 struct f2fs_gc_control gc_control = { 2536 .init_gc_type = range->sync ? FG_GC : BG_GC, 2537 .no_bg_gc = false, 2538 .should_migrate_blocks = false, 2539 .err_gc_skipped = range->sync, 2540 .nr_free_secs = 0 }; 2541 u64 end; 2542 int ret; 2543 2544 if (!capable(CAP_SYS_ADMIN)) 2545 return -EPERM; 2546 if (f2fs_readonly(sbi->sb)) 2547 return -EROFS; 2548 2549 end = range->start + range->len; 2550 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2551 end >= MAX_BLKADDR(sbi)) 2552 return -EINVAL; 2553 2554 ret = mnt_want_write_file(filp); 2555 if (ret) 2556 return ret; 2557 2558 do_more: 2559 if (!range->sync) { 2560 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2561 ret = -EBUSY; 2562 goto out; 2563 } 2564 } else { 2565 f2fs_down_write(&sbi->gc_lock); 2566 } 2567 2568 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2569 stat_inc_gc_call_count(sbi, FOREGROUND); 2570 ret = f2fs_gc(sbi, &gc_control); 2571 if (ret) { 2572 if (ret == -EBUSY) 2573 ret = -EAGAIN; 2574 goto out; 2575 } 2576 range->start += CAP_BLKS_PER_SEC(sbi); 2577 if (range->start <= end) 2578 goto do_more; 2579 out: 2580 mnt_drop_write_file(filp); 2581 return ret; 2582 } 2583 2584 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2585 { 2586 struct f2fs_gc_range range; 2587 2588 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2589 sizeof(range))) 2590 return -EFAULT; 2591 return __f2fs_ioc_gc_range(filp, &range); 2592 } 2593 2594 static int f2fs_ioc_write_checkpoint(struct file *filp) 2595 { 2596 struct inode *inode = file_inode(filp); 2597 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2598 int ret; 2599 2600 if (!capable(CAP_SYS_ADMIN)) 2601 return -EPERM; 2602 2603 if (f2fs_readonly(sbi->sb)) 2604 return -EROFS; 2605 2606 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2607 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2608 return -EINVAL; 2609 } 2610 2611 ret = mnt_want_write_file(filp); 2612 if (ret) 2613 return ret; 2614 2615 ret = f2fs_sync_fs(sbi->sb, 1); 2616 2617 mnt_drop_write_file(filp); 2618 return ret; 2619 } 2620 2621 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2622 struct file *filp, 2623 struct f2fs_defragment *range) 2624 { 2625 struct inode *inode = file_inode(filp); 2626 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2627 .m_seg_type = NO_CHECK_TYPE, 2628 .m_may_create = false }; 2629 struct extent_info ei = {}; 2630 pgoff_t pg_start, pg_end, next_pgofs; 2631 unsigned int total = 0, sec_num; 2632 block_t blk_end = 0; 2633 bool fragmented = false; 2634 int err; 2635 2636 pg_start = range->start >> PAGE_SHIFT; 2637 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2638 2639 f2fs_balance_fs(sbi, true); 2640 2641 inode_lock(inode); 2642 2643 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2644 err = -EINVAL; 2645 goto unlock_out; 2646 } 2647 2648 /* if in-place-update policy is enabled, don't waste time here */ 2649 set_inode_flag(inode, FI_OPU_WRITE); 2650 if (f2fs_should_update_inplace(inode, NULL)) { 2651 err = -EINVAL; 2652 goto out; 2653 } 2654 2655 /* writeback all dirty pages in the range */ 2656 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2657 range->start + range->len - 1); 2658 if (err) 2659 goto out; 2660 2661 /* 2662 * lookup mapping info in extent cache, skip defragmenting if physical 2663 * block addresses are continuous. 2664 */ 2665 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2666 if (ei.fofs + ei.len >= pg_end) 2667 goto out; 2668 } 2669 2670 map.m_lblk = pg_start; 2671 map.m_next_pgofs = &next_pgofs; 2672 2673 /* 2674 * lookup mapping info in dnode page cache, skip defragmenting if all 2675 * physical block addresses are continuous even if there are hole(s) 2676 * in logical blocks. 2677 */ 2678 while (map.m_lblk < pg_end) { 2679 map.m_len = pg_end - map.m_lblk; 2680 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2681 if (err) 2682 goto out; 2683 2684 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2685 map.m_lblk = next_pgofs; 2686 continue; 2687 } 2688 2689 if (blk_end && blk_end != map.m_pblk) 2690 fragmented = true; 2691 2692 /* record total count of block that we're going to move */ 2693 total += map.m_len; 2694 2695 blk_end = map.m_pblk + map.m_len; 2696 2697 map.m_lblk += map.m_len; 2698 } 2699 2700 if (!fragmented) { 2701 total = 0; 2702 goto out; 2703 } 2704 2705 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2706 2707 /* 2708 * make sure there are enough free section for LFS allocation, this can 2709 * avoid defragment running in SSR mode when free section are allocated 2710 * intensively 2711 */ 2712 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2713 err = -EAGAIN; 2714 goto out; 2715 } 2716 2717 map.m_lblk = pg_start; 2718 map.m_len = pg_end - pg_start; 2719 total = 0; 2720 2721 while (map.m_lblk < pg_end) { 2722 pgoff_t idx; 2723 int cnt = 0; 2724 2725 do_map: 2726 map.m_len = pg_end - map.m_lblk; 2727 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2728 if (err) 2729 goto clear_out; 2730 2731 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2732 map.m_lblk = next_pgofs; 2733 goto check; 2734 } 2735 2736 set_inode_flag(inode, FI_SKIP_WRITES); 2737 2738 idx = map.m_lblk; 2739 while (idx < map.m_lblk + map.m_len && 2740 cnt < BLKS_PER_SEG(sbi)) { 2741 struct page *page; 2742 2743 page = f2fs_get_lock_data_page(inode, idx, true); 2744 if (IS_ERR(page)) { 2745 err = PTR_ERR(page); 2746 goto clear_out; 2747 } 2748 2749 set_page_dirty(page); 2750 set_page_private_gcing(page); 2751 f2fs_put_page(page, 1); 2752 2753 idx++; 2754 cnt++; 2755 total++; 2756 } 2757 2758 map.m_lblk = idx; 2759 check: 2760 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 2761 goto do_map; 2762 2763 clear_inode_flag(inode, FI_SKIP_WRITES); 2764 2765 err = filemap_fdatawrite(inode->i_mapping); 2766 if (err) 2767 goto out; 2768 } 2769 clear_out: 2770 clear_inode_flag(inode, FI_SKIP_WRITES); 2771 out: 2772 clear_inode_flag(inode, FI_OPU_WRITE); 2773 unlock_out: 2774 inode_unlock(inode); 2775 if (!err) 2776 range->len = (u64)total << PAGE_SHIFT; 2777 return err; 2778 } 2779 2780 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2781 { 2782 struct inode *inode = file_inode(filp); 2783 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2784 struct f2fs_defragment range; 2785 int err; 2786 2787 if (!capable(CAP_SYS_ADMIN)) 2788 return -EPERM; 2789 2790 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2791 return -EINVAL; 2792 2793 if (f2fs_readonly(sbi->sb)) 2794 return -EROFS; 2795 2796 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2797 sizeof(range))) 2798 return -EFAULT; 2799 2800 /* verify alignment of offset & size */ 2801 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2802 return -EINVAL; 2803 2804 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2805 max_file_blocks(inode))) 2806 return -EINVAL; 2807 2808 err = mnt_want_write_file(filp); 2809 if (err) 2810 return err; 2811 2812 err = f2fs_defragment_range(sbi, filp, &range); 2813 mnt_drop_write_file(filp); 2814 2815 f2fs_update_time(sbi, REQ_TIME); 2816 if (err < 0) 2817 return err; 2818 2819 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2820 sizeof(range))) 2821 return -EFAULT; 2822 2823 return 0; 2824 } 2825 2826 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2827 struct file *file_out, loff_t pos_out, size_t len) 2828 { 2829 struct inode *src = file_inode(file_in); 2830 struct inode *dst = file_inode(file_out); 2831 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2832 size_t olen = len, dst_max_i_size = 0; 2833 size_t dst_osize; 2834 int ret; 2835 2836 if (file_in->f_path.mnt != file_out->f_path.mnt || 2837 src->i_sb != dst->i_sb) 2838 return -EXDEV; 2839 2840 if (unlikely(f2fs_readonly(src->i_sb))) 2841 return -EROFS; 2842 2843 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2844 return -EINVAL; 2845 2846 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2847 return -EOPNOTSUPP; 2848 2849 if (pos_out < 0 || pos_in < 0) 2850 return -EINVAL; 2851 2852 if (src == dst) { 2853 if (pos_in == pos_out) 2854 return 0; 2855 if (pos_out > pos_in && pos_out < pos_in + len) 2856 return -EINVAL; 2857 } 2858 2859 inode_lock(src); 2860 if (src != dst) { 2861 ret = -EBUSY; 2862 if (!inode_trylock(dst)) 2863 goto out; 2864 } 2865 2866 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) || 2867 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 2868 ret = -EOPNOTSUPP; 2869 goto out_unlock; 2870 } 2871 2872 ret = -EINVAL; 2873 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2874 goto out_unlock; 2875 if (len == 0) 2876 olen = len = src->i_size - pos_in; 2877 if (pos_in + len == src->i_size) 2878 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2879 if (len == 0) { 2880 ret = 0; 2881 goto out_unlock; 2882 } 2883 2884 dst_osize = dst->i_size; 2885 if (pos_out + olen > dst->i_size) 2886 dst_max_i_size = pos_out + olen; 2887 2888 /* verify the end result is block aligned */ 2889 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2890 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2891 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2892 goto out_unlock; 2893 2894 ret = f2fs_convert_inline_inode(src); 2895 if (ret) 2896 goto out_unlock; 2897 2898 ret = f2fs_convert_inline_inode(dst); 2899 if (ret) 2900 goto out_unlock; 2901 2902 /* write out all dirty pages from offset */ 2903 ret = filemap_write_and_wait_range(src->i_mapping, 2904 pos_in, pos_in + len); 2905 if (ret) 2906 goto out_unlock; 2907 2908 ret = filemap_write_and_wait_range(dst->i_mapping, 2909 pos_out, pos_out + len); 2910 if (ret) 2911 goto out_unlock; 2912 2913 f2fs_balance_fs(sbi, true); 2914 2915 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2916 if (src != dst) { 2917 ret = -EBUSY; 2918 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2919 goto out_src; 2920 } 2921 2922 f2fs_lock_op(sbi); 2923 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2924 pos_out >> F2FS_BLKSIZE_BITS, 2925 len >> F2FS_BLKSIZE_BITS, false); 2926 2927 if (!ret) { 2928 if (dst_max_i_size) 2929 f2fs_i_size_write(dst, dst_max_i_size); 2930 else if (dst_osize != dst->i_size) 2931 f2fs_i_size_write(dst, dst_osize); 2932 } 2933 f2fs_unlock_op(sbi); 2934 2935 if (src != dst) 2936 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2937 out_src: 2938 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2939 if (ret) 2940 goto out_unlock; 2941 2942 src->i_mtime = inode_set_ctime_current(src); 2943 f2fs_mark_inode_dirty_sync(src, false); 2944 if (src != dst) { 2945 dst->i_mtime = inode_set_ctime_current(dst); 2946 f2fs_mark_inode_dirty_sync(dst, false); 2947 } 2948 f2fs_update_time(sbi, REQ_TIME); 2949 2950 out_unlock: 2951 if (src != dst) 2952 inode_unlock(dst); 2953 out: 2954 inode_unlock(src); 2955 return ret; 2956 } 2957 2958 static int __f2fs_ioc_move_range(struct file *filp, 2959 struct f2fs_move_range *range) 2960 { 2961 struct fd dst; 2962 int err; 2963 2964 if (!(filp->f_mode & FMODE_READ) || 2965 !(filp->f_mode & FMODE_WRITE)) 2966 return -EBADF; 2967 2968 dst = fdget(range->dst_fd); 2969 if (!dst.file) 2970 return -EBADF; 2971 2972 if (!(dst.file->f_mode & FMODE_WRITE)) { 2973 err = -EBADF; 2974 goto err_out; 2975 } 2976 2977 err = mnt_want_write_file(filp); 2978 if (err) 2979 goto err_out; 2980 2981 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2982 range->pos_out, range->len); 2983 2984 mnt_drop_write_file(filp); 2985 err_out: 2986 fdput(dst); 2987 return err; 2988 } 2989 2990 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2991 { 2992 struct f2fs_move_range range; 2993 2994 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2995 sizeof(range))) 2996 return -EFAULT; 2997 return __f2fs_ioc_move_range(filp, &range); 2998 } 2999 3000 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 3001 { 3002 struct inode *inode = file_inode(filp); 3003 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3004 struct sit_info *sm = SIT_I(sbi); 3005 unsigned int start_segno = 0, end_segno = 0; 3006 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3007 struct f2fs_flush_device range; 3008 struct f2fs_gc_control gc_control = { 3009 .init_gc_type = FG_GC, 3010 .should_migrate_blocks = true, 3011 .err_gc_skipped = true, 3012 .nr_free_secs = 0 }; 3013 int ret; 3014 3015 if (!capable(CAP_SYS_ADMIN)) 3016 return -EPERM; 3017 3018 if (f2fs_readonly(sbi->sb)) 3019 return -EROFS; 3020 3021 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3022 return -EINVAL; 3023 3024 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3025 sizeof(range))) 3026 return -EFAULT; 3027 3028 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3029 __is_large_section(sbi)) { 3030 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3031 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3032 return -EINVAL; 3033 } 3034 3035 ret = mnt_want_write_file(filp); 3036 if (ret) 3037 return ret; 3038 3039 if (range.dev_num != 0) 3040 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3041 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3042 3043 start_segno = sm->last_victim[FLUSH_DEVICE]; 3044 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3045 start_segno = dev_start_segno; 3046 end_segno = min(start_segno + range.segments, dev_end_segno); 3047 3048 while (start_segno < end_segno) { 3049 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3050 ret = -EBUSY; 3051 goto out; 3052 } 3053 sm->last_victim[GC_CB] = end_segno + 1; 3054 sm->last_victim[GC_GREEDY] = end_segno + 1; 3055 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3056 3057 gc_control.victim_segno = start_segno; 3058 stat_inc_gc_call_count(sbi, FOREGROUND); 3059 ret = f2fs_gc(sbi, &gc_control); 3060 if (ret == -EAGAIN) 3061 ret = 0; 3062 else if (ret < 0) 3063 break; 3064 start_segno++; 3065 } 3066 out: 3067 mnt_drop_write_file(filp); 3068 return ret; 3069 } 3070 3071 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3072 { 3073 struct inode *inode = file_inode(filp); 3074 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3075 3076 /* Must validate to set it with SQLite behavior in Android. */ 3077 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3078 3079 return put_user(sb_feature, (u32 __user *)arg); 3080 } 3081 3082 #ifdef CONFIG_QUOTA 3083 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3084 { 3085 struct dquot *transfer_to[MAXQUOTAS] = {}; 3086 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3087 struct super_block *sb = sbi->sb; 3088 int err; 3089 3090 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3091 if (IS_ERR(transfer_to[PRJQUOTA])) 3092 return PTR_ERR(transfer_to[PRJQUOTA]); 3093 3094 err = __dquot_transfer(inode, transfer_to); 3095 if (err) 3096 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3097 dqput(transfer_to[PRJQUOTA]); 3098 return err; 3099 } 3100 3101 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3102 { 3103 struct f2fs_inode_info *fi = F2FS_I(inode); 3104 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3105 struct f2fs_inode *ri = NULL; 3106 kprojid_t kprojid; 3107 int err; 3108 3109 if (!f2fs_sb_has_project_quota(sbi)) { 3110 if (projid != F2FS_DEF_PROJID) 3111 return -EOPNOTSUPP; 3112 else 3113 return 0; 3114 } 3115 3116 if (!f2fs_has_extra_attr(inode)) 3117 return -EOPNOTSUPP; 3118 3119 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3120 3121 if (projid_eq(kprojid, fi->i_projid)) 3122 return 0; 3123 3124 err = -EPERM; 3125 /* Is it quota file? Do not allow user to mess with it */ 3126 if (IS_NOQUOTA(inode)) 3127 return err; 3128 3129 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3130 return -EOVERFLOW; 3131 3132 err = f2fs_dquot_initialize(inode); 3133 if (err) 3134 return err; 3135 3136 f2fs_lock_op(sbi); 3137 err = f2fs_transfer_project_quota(inode, kprojid); 3138 if (err) 3139 goto out_unlock; 3140 3141 fi->i_projid = kprojid; 3142 inode_set_ctime_current(inode); 3143 f2fs_mark_inode_dirty_sync(inode, true); 3144 out_unlock: 3145 f2fs_unlock_op(sbi); 3146 return err; 3147 } 3148 #else 3149 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3150 { 3151 return 0; 3152 } 3153 3154 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3155 { 3156 if (projid != F2FS_DEF_PROJID) 3157 return -EOPNOTSUPP; 3158 return 0; 3159 } 3160 #endif 3161 3162 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3163 { 3164 struct inode *inode = d_inode(dentry); 3165 struct f2fs_inode_info *fi = F2FS_I(inode); 3166 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3167 3168 if (IS_ENCRYPTED(inode)) 3169 fsflags |= FS_ENCRYPT_FL; 3170 if (IS_VERITY(inode)) 3171 fsflags |= FS_VERITY_FL; 3172 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3173 fsflags |= FS_INLINE_DATA_FL; 3174 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3175 fsflags |= FS_NOCOW_FL; 3176 3177 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3178 3179 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3180 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3181 3182 return 0; 3183 } 3184 3185 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3186 struct dentry *dentry, struct fileattr *fa) 3187 { 3188 struct inode *inode = d_inode(dentry); 3189 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3190 u32 iflags; 3191 int err; 3192 3193 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3194 return -EIO; 3195 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3196 return -ENOSPC; 3197 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3198 return -EOPNOTSUPP; 3199 fsflags &= F2FS_SETTABLE_FS_FL; 3200 if (!fa->flags_valid) 3201 mask &= FS_COMMON_FL; 3202 3203 iflags = f2fs_fsflags_to_iflags(fsflags); 3204 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3205 return -EOPNOTSUPP; 3206 3207 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3208 if (!err) 3209 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3210 3211 return err; 3212 } 3213 3214 int f2fs_pin_file_control(struct inode *inode, bool inc) 3215 { 3216 struct f2fs_inode_info *fi = F2FS_I(inode); 3217 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3218 3219 /* Use i_gc_failures for normal file as a risk signal. */ 3220 if (inc) 3221 f2fs_i_gc_failures_write(inode, 3222 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3223 3224 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3225 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3226 __func__, inode->i_ino, 3227 fi->i_gc_failures[GC_FAILURE_PIN]); 3228 clear_inode_flag(inode, FI_PIN_FILE); 3229 return -EAGAIN; 3230 } 3231 return 0; 3232 } 3233 3234 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3235 { 3236 struct inode *inode = file_inode(filp); 3237 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3238 __u32 pin; 3239 int ret = 0; 3240 3241 if (get_user(pin, (__u32 __user *)arg)) 3242 return -EFAULT; 3243 3244 if (!S_ISREG(inode->i_mode)) 3245 return -EINVAL; 3246 3247 if (f2fs_readonly(sbi->sb)) 3248 return -EROFS; 3249 3250 ret = mnt_want_write_file(filp); 3251 if (ret) 3252 return ret; 3253 3254 inode_lock(inode); 3255 3256 if (!pin) { 3257 clear_inode_flag(inode, FI_PIN_FILE); 3258 f2fs_i_gc_failures_write(inode, 0); 3259 goto done; 3260 } else if (f2fs_is_pinned_file(inode)) { 3261 goto done; 3262 } 3263 3264 if (f2fs_sb_has_blkzoned(sbi) && F2FS_HAS_BLOCKS(inode)) { 3265 ret = -EFBIG; 3266 goto out; 3267 } 3268 3269 /* Let's allow file pinning on zoned device. */ 3270 if (!f2fs_sb_has_blkzoned(sbi) && 3271 f2fs_should_update_outplace(inode, NULL)) { 3272 ret = -EINVAL; 3273 goto out; 3274 } 3275 3276 if (f2fs_pin_file_control(inode, false)) { 3277 ret = -EAGAIN; 3278 goto out; 3279 } 3280 3281 ret = f2fs_convert_inline_inode(inode); 3282 if (ret) 3283 goto out; 3284 3285 if (!f2fs_disable_compressed_file(inode)) { 3286 ret = -EOPNOTSUPP; 3287 goto out; 3288 } 3289 3290 set_inode_flag(inode, FI_PIN_FILE); 3291 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3292 done: 3293 f2fs_update_time(sbi, REQ_TIME); 3294 out: 3295 inode_unlock(inode); 3296 mnt_drop_write_file(filp); 3297 return ret; 3298 } 3299 3300 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3301 { 3302 struct inode *inode = file_inode(filp); 3303 __u32 pin = 0; 3304 3305 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3306 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3307 return put_user(pin, (u32 __user *)arg); 3308 } 3309 3310 int f2fs_precache_extents(struct inode *inode) 3311 { 3312 struct f2fs_inode_info *fi = F2FS_I(inode); 3313 struct f2fs_map_blocks map; 3314 pgoff_t m_next_extent; 3315 loff_t end; 3316 int err; 3317 3318 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3319 return -EOPNOTSUPP; 3320 3321 map.m_lblk = 0; 3322 map.m_pblk = 0; 3323 map.m_next_pgofs = NULL; 3324 map.m_next_extent = &m_next_extent; 3325 map.m_seg_type = NO_CHECK_TYPE; 3326 map.m_may_create = false; 3327 end = max_file_blocks(inode); 3328 3329 while (map.m_lblk < end) { 3330 map.m_len = end - map.m_lblk; 3331 3332 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3333 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3334 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3335 if (err) 3336 return err; 3337 3338 map.m_lblk = m_next_extent; 3339 } 3340 3341 return 0; 3342 } 3343 3344 static int f2fs_ioc_precache_extents(struct file *filp) 3345 { 3346 return f2fs_precache_extents(file_inode(filp)); 3347 } 3348 3349 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3350 { 3351 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3352 __u64 block_count; 3353 3354 if (!capable(CAP_SYS_ADMIN)) 3355 return -EPERM; 3356 3357 if (f2fs_readonly(sbi->sb)) 3358 return -EROFS; 3359 3360 if (copy_from_user(&block_count, (void __user *)arg, 3361 sizeof(block_count))) 3362 return -EFAULT; 3363 3364 return f2fs_resize_fs(filp, block_count); 3365 } 3366 3367 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3368 { 3369 struct inode *inode = file_inode(filp); 3370 3371 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3372 3373 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3374 f2fs_warn(F2FS_I_SB(inode), 3375 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3376 inode->i_ino); 3377 return -EOPNOTSUPP; 3378 } 3379 3380 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3381 } 3382 3383 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3384 { 3385 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3386 return -EOPNOTSUPP; 3387 3388 return fsverity_ioctl_measure(filp, (void __user *)arg); 3389 } 3390 3391 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3392 { 3393 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3394 return -EOPNOTSUPP; 3395 3396 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3397 } 3398 3399 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3400 { 3401 struct inode *inode = file_inode(filp); 3402 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3403 char *vbuf; 3404 int count; 3405 int err = 0; 3406 3407 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3408 if (!vbuf) 3409 return -ENOMEM; 3410 3411 f2fs_down_read(&sbi->sb_lock); 3412 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3413 ARRAY_SIZE(sbi->raw_super->volume_name), 3414 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3415 f2fs_up_read(&sbi->sb_lock); 3416 3417 if (copy_to_user((char __user *)arg, vbuf, 3418 min(FSLABEL_MAX, count))) 3419 err = -EFAULT; 3420 3421 kfree(vbuf); 3422 return err; 3423 } 3424 3425 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3426 { 3427 struct inode *inode = file_inode(filp); 3428 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3429 char *vbuf; 3430 int err = 0; 3431 3432 if (!capable(CAP_SYS_ADMIN)) 3433 return -EPERM; 3434 3435 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3436 if (IS_ERR(vbuf)) 3437 return PTR_ERR(vbuf); 3438 3439 err = mnt_want_write_file(filp); 3440 if (err) 3441 goto out; 3442 3443 f2fs_down_write(&sbi->sb_lock); 3444 3445 memset(sbi->raw_super->volume_name, 0, 3446 sizeof(sbi->raw_super->volume_name)); 3447 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3448 sbi->raw_super->volume_name, 3449 ARRAY_SIZE(sbi->raw_super->volume_name)); 3450 3451 err = f2fs_commit_super(sbi, false); 3452 3453 f2fs_up_write(&sbi->sb_lock); 3454 3455 mnt_drop_write_file(filp); 3456 out: 3457 kfree(vbuf); 3458 return err; 3459 } 3460 3461 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3462 { 3463 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3464 return -EOPNOTSUPP; 3465 3466 if (!f2fs_compressed_file(inode)) 3467 return -EINVAL; 3468 3469 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3470 3471 return 0; 3472 } 3473 3474 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3475 { 3476 struct inode *inode = file_inode(filp); 3477 __u64 blocks; 3478 int ret; 3479 3480 ret = f2fs_get_compress_blocks(inode, &blocks); 3481 if (ret < 0) 3482 return ret; 3483 3484 return put_user(blocks, (u64 __user *)arg); 3485 } 3486 3487 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3488 { 3489 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3490 unsigned int released_blocks = 0; 3491 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3492 block_t blkaddr; 3493 int i; 3494 3495 for (i = 0; i < count; i++) { 3496 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3497 dn->ofs_in_node + i); 3498 3499 if (!__is_valid_data_blkaddr(blkaddr)) 3500 continue; 3501 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3502 DATA_GENERIC_ENHANCE))) { 3503 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3504 return -EFSCORRUPTED; 3505 } 3506 } 3507 3508 while (count) { 3509 int compr_blocks = 0; 3510 3511 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3512 blkaddr = f2fs_data_blkaddr(dn); 3513 3514 if (i == 0) { 3515 if (blkaddr == COMPRESS_ADDR) 3516 continue; 3517 dn->ofs_in_node += cluster_size; 3518 goto next; 3519 } 3520 3521 if (__is_valid_data_blkaddr(blkaddr)) 3522 compr_blocks++; 3523 3524 if (blkaddr != NEW_ADDR) 3525 continue; 3526 3527 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3528 } 3529 3530 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3531 dec_valid_block_count(sbi, dn->inode, 3532 cluster_size - compr_blocks); 3533 3534 released_blocks += cluster_size - compr_blocks; 3535 next: 3536 count -= cluster_size; 3537 } 3538 3539 return released_blocks; 3540 } 3541 3542 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3543 { 3544 struct inode *inode = file_inode(filp); 3545 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3546 pgoff_t page_idx = 0, last_idx; 3547 unsigned int released_blocks = 0; 3548 int ret; 3549 int writecount; 3550 3551 if (!f2fs_sb_has_compression(sbi)) 3552 return -EOPNOTSUPP; 3553 3554 if (f2fs_readonly(sbi->sb)) 3555 return -EROFS; 3556 3557 ret = mnt_want_write_file(filp); 3558 if (ret) 3559 return ret; 3560 3561 f2fs_balance_fs(sbi, true); 3562 3563 inode_lock(inode); 3564 3565 writecount = atomic_read(&inode->i_writecount); 3566 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3567 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3568 ret = -EBUSY; 3569 goto out; 3570 } 3571 3572 if (!f2fs_compressed_file(inode) || 3573 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3574 ret = -EINVAL; 3575 goto out; 3576 } 3577 3578 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3579 if (ret) 3580 goto out; 3581 3582 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3583 ret = -EPERM; 3584 goto out; 3585 } 3586 3587 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3588 inode_set_ctime_current(inode); 3589 f2fs_mark_inode_dirty_sync(inode, true); 3590 3591 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3592 filemap_invalidate_lock(inode->i_mapping); 3593 3594 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3595 3596 while (page_idx < last_idx) { 3597 struct dnode_of_data dn; 3598 pgoff_t end_offset, count; 3599 3600 f2fs_lock_op(sbi); 3601 3602 set_new_dnode(&dn, inode, NULL, NULL, 0); 3603 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3604 if (ret) { 3605 f2fs_unlock_op(sbi); 3606 if (ret == -ENOENT) { 3607 page_idx = f2fs_get_next_page_offset(&dn, 3608 page_idx); 3609 ret = 0; 3610 continue; 3611 } 3612 break; 3613 } 3614 3615 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3616 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3617 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3618 3619 ret = release_compress_blocks(&dn, count); 3620 3621 f2fs_put_dnode(&dn); 3622 3623 f2fs_unlock_op(sbi); 3624 3625 if (ret < 0) 3626 break; 3627 3628 page_idx += count; 3629 released_blocks += ret; 3630 } 3631 3632 filemap_invalidate_unlock(inode->i_mapping); 3633 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3634 out: 3635 inode_unlock(inode); 3636 3637 mnt_drop_write_file(filp); 3638 3639 if (ret >= 0) { 3640 ret = put_user(released_blocks, (u64 __user *)arg); 3641 } else if (released_blocks && 3642 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3643 set_sbi_flag(sbi, SBI_NEED_FSCK); 3644 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3645 "iblocks=%llu, released=%u, compr_blocks=%u, " 3646 "run fsck to fix.", 3647 __func__, inode->i_ino, inode->i_blocks, 3648 released_blocks, 3649 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3650 } 3651 3652 return ret; 3653 } 3654 3655 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3656 unsigned int *reserved_blocks) 3657 { 3658 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3659 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3660 block_t blkaddr; 3661 int i; 3662 3663 for (i = 0; i < count; i++) { 3664 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3665 dn->ofs_in_node + i); 3666 3667 if (!__is_valid_data_blkaddr(blkaddr)) 3668 continue; 3669 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3670 DATA_GENERIC_ENHANCE))) { 3671 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3672 return -EFSCORRUPTED; 3673 } 3674 } 3675 3676 while (count) { 3677 int compr_blocks = 0; 3678 blkcnt_t reserved = 0; 3679 blkcnt_t to_reserved; 3680 int ret; 3681 3682 for (i = 0; i < cluster_size; i++) { 3683 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3684 dn->ofs_in_node + i); 3685 3686 if (i == 0) { 3687 if (blkaddr != COMPRESS_ADDR) { 3688 dn->ofs_in_node += cluster_size; 3689 goto next; 3690 } 3691 continue; 3692 } 3693 3694 /* 3695 * compressed cluster was not released due to it 3696 * fails in release_compress_blocks(), so NEW_ADDR 3697 * is a possible case. 3698 */ 3699 if (blkaddr == NEW_ADDR) { 3700 reserved++; 3701 continue; 3702 } 3703 if (__is_valid_data_blkaddr(blkaddr)) { 3704 compr_blocks++; 3705 continue; 3706 } 3707 } 3708 3709 to_reserved = cluster_size - compr_blocks - reserved; 3710 3711 /* for the case all blocks in cluster were reserved */ 3712 if (to_reserved == 1) { 3713 dn->ofs_in_node += cluster_size; 3714 goto next; 3715 } 3716 3717 ret = inc_valid_block_count(sbi, dn->inode, 3718 &to_reserved, false); 3719 if (unlikely(ret)) 3720 return ret; 3721 3722 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3723 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3724 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3725 } 3726 3727 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3728 3729 *reserved_blocks += to_reserved; 3730 next: 3731 count -= cluster_size; 3732 } 3733 3734 return 0; 3735 } 3736 3737 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3738 { 3739 struct inode *inode = file_inode(filp); 3740 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3741 pgoff_t page_idx = 0, last_idx; 3742 unsigned int reserved_blocks = 0; 3743 int ret; 3744 3745 if (!f2fs_sb_has_compression(sbi)) 3746 return -EOPNOTSUPP; 3747 3748 if (f2fs_readonly(sbi->sb)) 3749 return -EROFS; 3750 3751 ret = mnt_want_write_file(filp); 3752 if (ret) 3753 return ret; 3754 3755 f2fs_balance_fs(sbi, true); 3756 3757 inode_lock(inode); 3758 3759 if (!f2fs_compressed_file(inode) || 3760 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3761 ret = -EINVAL; 3762 goto unlock_inode; 3763 } 3764 3765 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3766 goto unlock_inode; 3767 3768 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3769 filemap_invalidate_lock(inode->i_mapping); 3770 3771 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3772 3773 while (page_idx < last_idx) { 3774 struct dnode_of_data dn; 3775 pgoff_t end_offset, count; 3776 3777 f2fs_lock_op(sbi); 3778 3779 set_new_dnode(&dn, inode, NULL, NULL, 0); 3780 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3781 if (ret) { 3782 f2fs_unlock_op(sbi); 3783 if (ret == -ENOENT) { 3784 page_idx = f2fs_get_next_page_offset(&dn, 3785 page_idx); 3786 ret = 0; 3787 continue; 3788 } 3789 break; 3790 } 3791 3792 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3793 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3794 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3795 3796 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 3797 3798 f2fs_put_dnode(&dn); 3799 3800 f2fs_unlock_op(sbi); 3801 3802 if (ret < 0) 3803 break; 3804 3805 page_idx += count; 3806 } 3807 3808 filemap_invalidate_unlock(inode->i_mapping); 3809 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3810 3811 if (!ret) { 3812 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3813 inode_set_ctime_current(inode); 3814 f2fs_mark_inode_dirty_sync(inode, true); 3815 } 3816 unlock_inode: 3817 inode_unlock(inode); 3818 mnt_drop_write_file(filp); 3819 3820 if (!ret) { 3821 ret = put_user(reserved_blocks, (u64 __user *)arg); 3822 } else if (reserved_blocks && 3823 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3824 set_sbi_flag(sbi, SBI_NEED_FSCK); 3825 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3826 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3827 "run fsck to fix.", 3828 __func__, inode->i_ino, inode->i_blocks, 3829 reserved_blocks, 3830 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3831 } 3832 3833 return ret; 3834 } 3835 3836 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3837 pgoff_t off, block_t block, block_t len, u32 flags) 3838 { 3839 sector_t sector = SECTOR_FROM_BLOCK(block); 3840 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3841 int ret = 0; 3842 3843 if (flags & F2FS_TRIM_FILE_DISCARD) { 3844 if (bdev_max_secure_erase_sectors(bdev)) 3845 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3846 GFP_NOFS); 3847 else 3848 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3849 GFP_NOFS); 3850 } 3851 3852 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3853 if (IS_ENCRYPTED(inode)) 3854 ret = fscrypt_zeroout_range(inode, off, block, len); 3855 else 3856 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3857 GFP_NOFS, 0); 3858 } 3859 3860 return ret; 3861 } 3862 3863 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3864 { 3865 struct inode *inode = file_inode(filp); 3866 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3867 struct address_space *mapping = inode->i_mapping; 3868 struct block_device *prev_bdev = NULL; 3869 struct f2fs_sectrim_range range; 3870 pgoff_t index, pg_end, prev_index = 0; 3871 block_t prev_block = 0, len = 0; 3872 loff_t end_addr; 3873 bool to_end = false; 3874 int ret = 0; 3875 3876 if (!(filp->f_mode & FMODE_WRITE)) 3877 return -EBADF; 3878 3879 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3880 sizeof(range))) 3881 return -EFAULT; 3882 3883 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3884 !S_ISREG(inode->i_mode)) 3885 return -EINVAL; 3886 3887 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3888 !f2fs_hw_support_discard(sbi)) || 3889 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3890 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3891 return -EOPNOTSUPP; 3892 3893 file_start_write(filp); 3894 inode_lock(inode); 3895 3896 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3897 range.start >= inode->i_size) { 3898 ret = -EINVAL; 3899 goto err; 3900 } 3901 3902 if (range.len == 0) 3903 goto err; 3904 3905 if (inode->i_size - range.start > range.len) { 3906 end_addr = range.start + range.len; 3907 } else { 3908 end_addr = range.len == (u64)-1 ? 3909 sbi->sb->s_maxbytes : inode->i_size; 3910 to_end = true; 3911 } 3912 3913 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3914 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3915 ret = -EINVAL; 3916 goto err; 3917 } 3918 3919 index = F2FS_BYTES_TO_BLK(range.start); 3920 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3921 3922 ret = f2fs_convert_inline_inode(inode); 3923 if (ret) 3924 goto err; 3925 3926 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3927 filemap_invalidate_lock(mapping); 3928 3929 ret = filemap_write_and_wait_range(mapping, range.start, 3930 to_end ? LLONG_MAX : end_addr - 1); 3931 if (ret) 3932 goto out; 3933 3934 truncate_inode_pages_range(mapping, range.start, 3935 to_end ? -1 : end_addr - 1); 3936 3937 while (index < pg_end) { 3938 struct dnode_of_data dn; 3939 pgoff_t end_offset, count; 3940 int i; 3941 3942 set_new_dnode(&dn, inode, NULL, NULL, 0); 3943 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3944 if (ret) { 3945 if (ret == -ENOENT) { 3946 index = f2fs_get_next_page_offset(&dn, index); 3947 continue; 3948 } 3949 goto out; 3950 } 3951 3952 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3953 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3954 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3955 struct block_device *cur_bdev; 3956 block_t blkaddr = f2fs_data_blkaddr(&dn); 3957 3958 if (!__is_valid_data_blkaddr(blkaddr)) 3959 continue; 3960 3961 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3962 DATA_GENERIC_ENHANCE)) { 3963 ret = -EFSCORRUPTED; 3964 f2fs_put_dnode(&dn); 3965 f2fs_handle_error(sbi, 3966 ERROR_INVALID_BLKADDR); 3967 goto out; 3968 } 3969 3970 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3971 if (f2fs_is_multi_device(sbi)) { 3972 int di = f2fs_target_device_index(sbi, blkaddr); 3973 3974 blkaddr -= FDEV(di).start_blk; 3975 } 3976 3977 if (len) { 3978 if (prev_bdev == cur_bdev && 3979 index == prev_index + len && 3980 blkaddr == prev_block + len) { 3981 len++; 3982 } else { 3983 ret = f2fs_secure_erase(prev_bdev, 3984 inode, prev_index, prev_block, 3985 len, range.flags); 3986 if (ret) { 3987 f2fs_put_dnode(&dn); 3988 goto out; 3989 } 3990 3991 len = 0; 3992 } 3993 } 3994 3995 if (!len) { 3996 prev_bdev = cur_bdev; 3997 prev_index = index; 3998 prev_block = blkaddr; 3999 len = 1; 4000 } 4001 } 4002 4003 f2fs_put_dnode(&dn); 4004 4005 if (fatal_signal_pending(current)) { 4006 ret = -EINTR; 4007 goto out; 4008 } 4009 cond_resched(); 4010 } 4011 4012 if (len) 4013 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 4014 prev_block, len, range.flags); 4015 out: 4016 filemap_invalidate_unlock(mapping); 4017 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4018 err: 4019 inode_unlock(inode); 4020 file_end_write(filp); 4021 4022 return ret; 4023 } 4024 4025 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 4026 { 4027 struct inode *inode = file_inode(filp); 4028 struct f2fs_comp_option option; 4029 4030 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4031 return -EOPNOTSUPP; 4032 4033 inode_lock_shared(inode); 4034 4035 if (!f2fs_compressed_file(inode)) { 4036 inode_unlock_shared(inode); 4037 return -ENODATA; 4038 } 4039 4040 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4041 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4042 4043 inode_unlock_shared(inode); 4044 4045 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4046 sizeof(option))) 4047 return -EFAULT; 4048 4049 return 0; 4050 } 4051 4052 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4053 { 4054 struct inode *inode = file_inode(filp); 4055 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4056 struct f2fs_comp_option option; 4057 int ret = 0; 4058 4059 if (!f2fs_sb_has_compression(sbi)) 4060 return -EOPNOTSUPP; 4061 4062 if (!(filp->f_mode & FMODE_WRITE)) 4063 return -EBADF; 4064 4065 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4066 sizeof(option))) 4067 return -EFAULT; 4068 4069 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4070 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4071 option.algorithm >= COMPRESS_MAX) 4072 return -EINVAL; 4073 4074 file_start_write(filp); 4075 inode_lock(inode); 4076 4077 f2fs_down_write(&F2FS_I(inode)->i_sem); 4078 if (!f2fs_compressed_file(inode)) { 4079 ret = -EINVAL; 4080 goto out; 4081 } 4082 4083 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4084 ret = -EBUSY; 4085 goto out; 4086 } 4087 4088 if (F2FS_HAS_BLOCKS(inode)) { 4089 ret = -EFBIG; 4090 goto out; 4091 } 4092 4093 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4094 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4095 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4096 /* Set default level */ 4097 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4098 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4099 else 4100 F2FS_I(inode)->i_compress_level = 0; 4101 /* Adjust mount option level */ 4102 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4103 F2FS_OPTION(sbi).compress_level) 4104 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4105 f2fs_mark_inode_dirty_sync(inode, true); 4106 4107 if (!f2fs_is_compress_backend_ready(inode)) 4108 f2fs_warn(sbi, "compression algorithm is successfully set, " 4109 "but current kernel doesn't support this algorithm."); 4110 out: 4111 f2fs_up_write(&F2FS_I(inode)->i_sem); 4112 inode_unlock(inode); 4113 file_end_write(filp); 4114 4115 return ret; 4116 } 4117 4118 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4119 { 4120 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4121 struct address_space *mapping = inode->i_mapping; 4122 struct page *page; 4123 pgoff_t redirty_idx = page_idx; 4124 int i, page_len = 0, ret = 0; 4125 4126 page_cache_ra_unbounded(&ractl, len, 0); 4127 4128 for (i = 0; i < len; i++, page_idx++) { 4129 page = read_cache_page(mapping, page_idx, NULL, NULL); 4130 if (IS_ERR(page)) { 4131 ret = PTR_ERR(page); 4132 break; 4133 } 4134 page_len++; 4135 } 4136 4137 for (i = 0; i < page_len; i++, redirty_idx++) { 4138 page = find_lock_page(mapping, redirty_idx); 4139 4140 /* It will never fail, when page has pinned above */ 4141 f2fs_bug_on(F2FS_I_SB(inode), !page); 4142 4143 set_page_dirty(page); 4144 set_page_private_gcing(page); 4145 f2fs_put_page(page, 1); 4146 f2fs_put_page(page, 0); 4147 } 4148 4149 return ret; 4150 } 4151 4152 static int f2fs_ioc_decompress_file(struct file *filp) 4153 { 4154 struct inode *inode = file_inode(filp); 4155 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4156 struct f2fs_inode_info *fi = F2FS_I(inode); 4157 pgoff_t page_idx = 0, last_idx; 4158 int cluster_size = fi->i_cluster_size; 4159 int count, ret; 4160 4161 if (!f2fs_sb_has_compression(sbi) || 4162 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4163 return -EOPNOTSUPP; 4164 4165 if (!(filp->f_mode & FMODE_WRITE)) 4166 return -EBADF; 4167 4168 f2fs_balance_fs(sbi, true); 4169 4170 file_start_write(filp); 4171 inode_lock(inode); 4172 4173 if (!f2fs_is_compress_backend_ready(inode)) { 4174 ret = -EOPNOTSUPP; 4175 goto out; 4176 } 4177 4178 if (!f2fs_compressed_file(inode) || 4179 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4180 ret = -EINVAL; 4181 goto out; 4182 } 4183 4184 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4185 if (ret) 4186 goto out; 4187 4188 if (!atomic_read(&fi->i_compr_blocks)) 4189 goto out; 4190 4191 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4192 4193 count = last_idx - page_idx; 4194 while (count && count >= cluster_size) { 4195 ret = redirty_blocks(inode, page_idx, cluster_size); 4196 if (ret < 0) 4197 break; 4198 4199 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4200 ret = filemap_fdatawrite(inode->i_mapping); 4201 if (ret < 0) 4202 break; 4203 } 4204 4205 count -= cluster_size; 4206 page_idx += cluster_size; 4207 4208 cond_resched(); 4209 if (fatal_signal_pending(current)) { 4210 ret = -EINTR; 4211 break; 4212 } 4213 } 4214 4215 if (!ret) 4216 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4217 LLONG_MAX); 4218 4219 if (ret) 4220 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4221 __func__, ret); 4222 out: 4223 inode_unlock(inode); 4224 file_end_write(filp); 4225 4226 return ret; 4227 } 4228 4229 static int f2fs_ioc_compress_file(struct file *filp) 4230 { 4231 struct inode *inode = file_inode(filp); 4232 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4233 pgoff_t page_idx = 0, last_idx; 4234 int cluster_size = F2FS_I(inode)->i_cluster_size; 4235 int count, ret; 4236 4237 if (!f2fs_sb_has_compression(sbi) || 4238 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4239 return -EOPNOTSUPP; 4240 4241 if (!(filp->f_mode & FMODE_WRITE)) 4242 return -EBADF; 4243 4244 f2fs_balance_fs(sbi, true); 4245 4246 file_start_write(filp); 4247 inode_lock(inode); 4248 4249 if (!f2fs_is_compress_backend_ready(inode)) { 4250 ret = -EOPNOTSUPP; 4251 goto out; 4252 } 4253 4254 if (!f2fs_compressed_file(inode) || 4255 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4256 ret = -EINVAL; 4257 goto out; 4258 } 4259 4260 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4261 if (ret) 4262 goto out; 4263 4264 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4265 4266 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4267 4268 count = last_idx - page_idx; 4269 while (count && count >= cluster_size) { 4270 ret = redirty_blocks(inode, page_idx, cluster_size); 4271 if (ret < 0) 4272 break; 4273 4274 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4275 ret = filemap_fdatawrite(inode->i_mapping); 4276 if (ret < 0) 4277 break; 4278 } 4279 4280 count -= cluster_size; 4281 page_idx += cluster_size; 4282 4283 cond_resched(); 4284 if (fatal_signal_pending(current)) { 4285 ret = -EINTR; 4286 break; 4287 } 4288 } 4289 4290 if (!ret) 4291 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4292 LLONG_MAX); 4293 4294 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4295 4296 if (ret) 4297 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4298 __func__, ret); 4299 out: 4300 inode_unlock(inode); 4301 file_end_write(filp); 4302 4303 return ret; 4304 } 4305 4306 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4307 { 4308 switch (cmd) { 4309 case FS_IOC_GETVERSION: 4310 return f2fs_ioc_getversion(filp, arg); 4311 case F2FS_IOC_START_ATOMIC_WRITE: 4312 return f2fs_ioc_start_atomic_write(filp, false); 4313 case F2FS_IOC_START_ATOMIC_REPLACE: 4314 return f2fs_ioc_start_atomic_write(filp, true); 4315 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4316 return f2fs_ioc_commit_atomic_write(filp); 4317 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4318 return f2fs_ioc_abort_atomic_write(filp); 4319 case F2FS_IOC_START_VOLATILE_WRITE: 4320 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4321 return -EOPNOTSUPP; 4322 case F2FS_IOC_SHUTDOWN: 4323 return f2fs_ioc_shutdown(filp, arg); 4324 case FITRIM: 4325 return f2fs_ioc_fitrim(filp, arg); 4326 case FS_IOC_SET_ENCRYPTION_POLICY: 4327 return f2fs_ioc_set_encryption_policy(filp, arg); 4328 case FS_IOC_GET_ENCRYPTION_POLICY: 4329 return f2fs_ioc_get_encryption_policy(filp, arg); 4330 case FS_IOC_GET_ENCRYPTION_PWSALT: 4331 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4332 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4333 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4334 case FS_IOC_ADD_ENCRYPTION_KEY: 4335 return f2fs_ioc_add_encryption_key(filp, arg); 4336 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4337 return f2fs_ioc_remove_encryption_key(filp, arg); 4338 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4339 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4340 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4341 return f2fs_ioc_get_encryption_key_status(filp, arg); 4342 case FS_IOC_GET_ENCRYPTION_NONCE: 4343 return f2fs_ioc_get_encryption_nonce(filp, arg); 4344 case F2FS_IOC_GARBAGE_COLLECT: 4345 return f2fs_ioc_gc(filp, arg); 4346 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4347 return f2fs_ioc_gc_range(filp, arg); 4348 case F2FS_IOC_WRITE_CHECKPOINT: 4349 return f2fs_ioc_write_checkpoint(filp); 4350 case F2FS_IOC_DEFRAGMENT: 4351 return f2fs_ioc_defragment(filp, arg); 4352 case F2FS_IOC_MOVE_RANGE: 4353 return f2fs_ioc_move_range(filp, arg); 4354 case F2FS_IOC_FLUSH_DEVICE: 4355 return f2fs_ioc_flush_device(filp, arg); 4356 case F2FS_IOC_GET_FEATURES: 4357 return f2fs_ioc_get_features(filp, arg); 4358 case F2FS_IOC_GET_PIN_FILE: 4359 return f2fs_ioc_get_pin_file(filp, arg); 4360 case F2FS_IOC_SET_PIN_FILE: 4361 return f2fs_ioc_set_pin_file(filp, arg); 4362 case F2FS_IOC_PRECACHE_EXTENTS: 4363 return f2fs_ioc_precache_extents(filp); 4364 case F2FS_IOC_RESIZE_FS: 4365 return f2fs_ioc_resize_fs(filp, arg); 4366 case FS_IOC_ENABLE_VERITY: 4367 return f2fs_ioc_enable_verity(filp, arg); 4368 case FS_IOC_MEASURE_VERITY: 4369 return f2fs_ioc_measure_verity(filp, arg); 4370 case FS_IOC_READ_VERITY_METADATA: 4371 return f2fs_ioc_read_verity_metadata(filp, arg); 4372 case FS_IOC_GETFSLABEL: 4373 return f2fs_ioc_getfslabel(filp, arg); 4374 case FS_IOC_SETFSLABEL: 4375 return f2fs_ioc_setfslabel(filp, arg); 4376 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4377 return f2fs_ioc_get_compress_blocks(filp, arg); 4378 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4379 return f2fs_release_compress_blocks(filp, arg); 4380 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4381 return f2fs_reserve_compress_blocks(filp, arg); 4382 case F2FS_IOC_SEC_TRIM_FILE: 4383 return f2fs_sec_trim_file(filp, arg); 4384 case F2FS_IOC_GET_COMPRESS_OPTION: 4385 return f2fs_ioc_get_compress_option(filp, arg); 4386 case F2FS_IOC_SET_COMPRESS_OPTION: 4387 return f2fs_ioc_set_compress_option(filp, arg); 4388 case F2FS_IOC_DECOMPRESS_FILE: 4389 return f2fs_ioc_decompress_file(filp); 4390 case F2FS_IOC_COMPRESS_FILE: 4391 return f2fs_ioc_compress_file(filp); 4392 default: 4393 return -ENOTTY; 4394 } 4395 } 4396 4397 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4398 { 4399 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4400 return -EIO; 4401 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4402 return -ENOSPC; 4403 4404 return __f2fs_ioctl(filp, cmd, arg); 4405 } 4406 4407 /* 4408 * Return %true if the given read or write request should use direct I/O, or 4409 * %false if it should use buffered I/O. 4410 */ 4411 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4412 struct iov_iter *iter) 4413 { 4414 unsigned int align; 4415 4416 if (!(iocb->ki_flags & IOCB_DIRECT)) 4417 return false; 4418 4419 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4420 return false; 4421 4422 /* 4423 * Direct I/O not aligned to the disk's logical_block_size will be 4424 * attempted, but will fail with -EINVAL. 4425 * 4426 * f2fs additionally requires that direct I/O be aligned to the 4427 * filesystem block size, which is often a stricter requirement. 4428 * However, f2fs traditionally falls back to buffered I/O on requests 4429 * that are logical_block_size-aligned but not fs-block aligned. 4430 * 4431 * The below logic implements this behavior. 4432 */ 4433 align = iocb->ki_pos | iov_iter_alignment(iter); 4434 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4435 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4436 return false; 4437 4438 return true; 4439 } 4440 4441 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4442 unsigned int flags) 4443 { 4444 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4445 4446 dec_page_count(sbi, F2FS_DIO_READ); 4447 if (error) 4448 return error; 4449 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4450 return 0; 4451 } 4452 4453 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4454 .end_io = f2fs_dio_read_end_io, 4455 }; 4456 4457 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4458 { 4459 struct file *file = iocb->ki_filp; 4460 struct inode *inode = file_inode(file); 4461 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4462 struct f2fs_inode_info *fi = F2FS_I(inode); 4463 const loff_t pos = iocb->ki_pos; 4464 const size_t count = iov_iter_count(to); 4465 struct iomap_dio *dio; 4466 ssize_t ret; 4467 4468 if (count == 0) 4469 return 0; /* skip atime update */ 4470 4471 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4472 4473 if (iocb->ki_flags & IOCB_NOWAIT) { 4474 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4475 ret = -EAGAIN; 4476 goto out; 4477 } 4478 } else { 4479 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4480 } 4481 4482 /* 4483 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4484 * the higher-level function iomap_dio_rw() in order to ensure that the 4485 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4486 */ 4487 inc_page_count(sbi, F2FS_DIO_READ); 4488 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4489 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4490 if (IS_ERR_OR_NULL(dio)) { 4491 ret = PTR_ERR_OR_ZERO(dio); 4492 if (ret != -EIOCBQUEUED) 4493 dec_page_count(sbi, F2FS_DIO_READ); 4494 } else { 4495 ret = iomap_dio_complete(dio); 4496 } 4497 4498 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4499 4500 file_accessed(file); 4501 out: 4502 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4503 return ret; 4504 } 4505 4506 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4507 int rw) 4508 { 4509 struct inode *inode = file_inode(file); 4510 char *buf, *path; 4511 4512 buf = f2fs_getname(F2FS_I_SB(inode)); 4513 if (!buf) 4514 return; 4515 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4516 if (IS_ERR(path)) 4517 goto free_buf; 4518 if (rw == WRITE) 4519 trace_f2fs_datawrite_start(inode, pos, count, 4520 current->pid, path, current->comm); 4521 else 4522 trace_f2fs_dataread_start(inode, pos, count, 4523 current->pid, path, current->comm); 4524 free_buf: 4525 f2fs_putname(buf); 4526 } 4527 4528 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4529 { 4530 struct inode *inode = file_inode(iocb->ki_filp); 4531 const loff_t pos = iocb->ki_pos; 4532 ssize_t ret; 4533 4534 if (!f2fs_is_compress_backend_ready(inode)) 4535 return -EOPNOTSUPP; 4536 4537 if (trace_f2fs_dataread_start_enabled()) 4538 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4539 iov_iter_count(to), READ); 4540 4541 if (f2fs_should_use_dio(inode, iocb, to)) { 4542 ret = f2fs_dio_read_iter(iocb, to); 4543 } else { 4544 ret = filemap_read(iocb, to, 0); 4545 if (ret > 0) 4546 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4547 APP_BUFFERED_READ_IO, ret); 4548 } 4549 if (trace_f2fs_dataread_end_enabled()) 4550 trace_f2fs_dataread_end(inode, pos, ret); 4551 return ret; 4552 } 4553 4554 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4555 struct pipe_inode_info *pipe, 4556 size_t len, unsigned int flags) 4557 { 4558 struct inode *inode = file_inode(in); 4559 const loff_t pos = *ppos; 4560 ssize_t ret; 4561 4562 if (!f2fs_is_compress_backend_ready(inode)) 4563 return -EOPNOTSUPP; 4564 4565 if (trace_f2fs_dataread_start_enabled()) 4566 f2fs_trace_rw_file_path(in, pos, len, READ); 4567 4568 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4569 if (ret > 0) 4570 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4571 APP_BUFFERED_READ_IO, ret); 4572 4573 if (trace_f2fs_dataread_end_enabled()) 4574 trace_f2fs_dataread_end(inode, pos, ret); 4575 return ret; 4576 } 4577 4578 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4579 { 4580 struct file *file = iocb->ki_filp; 4581 struct inode *inode = file_inode(file); 4582 ssize_t count; 4583 int err; 4584 4585 if (IS_IMMUTABLE(inode)) 4586 return -EPERM; 4587 4588 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4589 return -EPERM; 4590 4591 count = generic_write_checks(iocb, from); 4592 if (count <= 0) 4593 return count; 4594 4595 err = file_modified(file); 4596 if (err) 4597 return err; 4598 return count; 4599 } 4600 4601 /* 4602 * Preallocate blocks for a write request, if it is possible and helpful to do 4603 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4604 * blocks were preallocated, or a negative errno value if something went 4605 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4606 * requested blocks (not just some of them) have been allocated. 4607 */ 4608 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4609 bool dio) 4610 { 4611 struct inode *inode = file_inode(iocb->ki_filp); 4612 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4613 const loff_t pos = iocb->ki_pos; 4614 const size_t count = iov_iter_count(iter); 4615 struct f2fs_map_blocks map = {}; 4616 int flag; 4617 int ret; 4618 4619 /* If it will be an out-of-place direct write, don't bother. */ 4620 if (dio && f2fs_lfs_mode(sbi)) 4621 return 0; 4622 /* 4623 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4624 * buffered IO, if DIO meets any holes. 4625 */ 4626 if (dio && i_size_read(inode) && 4627 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4628 return 0; 4629 4630 /* No-wait I/O can't allocate blocks. */ 4631 if (iocb->ki_flags & IOCB_NOWAIT) 4632 return 0; 4633 4634 /* If it will be a short write, don't bother. */ 4635 if (fault_in_iov_iter_readable(iter, count)) 4636 return 0; 4637 4638 if (f2fs_has_inline_data(inode)) { 4639 /* If the data will fit inline, don't bother. */ 4640 if (pos + count <= MAX_INLINE_DATA(inode)) 4641 return 0; 4642 ret = f2fs_convert_inline_inode(inode); 4643 if (ret) 4644 return ret; 4645 } 4646 4647 /* Do not preallocate blocks that will be written partially in 4KB. */ 4648 map.m_lblk = F2FS_BLK_ALIGN(pos); 4649 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4650 if (map.m_len > map.m_lblk) 4651 map.m_len -= map.m_lblk; 4652 else 4653 map.m_len = 0; 4654 map.m_may_create = true; 4655 if (dio) { 4656 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4657 flag = F2FS_GET_BLOCK_PRE_DIO; 4658 } else { 4659 map.m_seg_type = NO_CHECK_TYPE; 4660 flag = F2FS_GET_BLOCK_PRE_AIO; 4661 } 4662 4663 ret = f2fs_map_blocks(inode, &map, flag); 4664 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4665 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4666 return ret; 4667 if (ret == 0) 4668 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4669 return map.m_len; 4670 } 4671 4672 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4673 struct iov_iter *from) 4674 { 4675 struct file *file = iocb->ki_filp; 4676 struct inode *inode = file_inode(file); 4677 ssize_t ret; 4678 4679 if (iocb->ki_flags & IOCB_NOWAIT) 4680 return -EOPNOTSUPP; 4681 4682 ret = generic_perform_write(iocb, from); 4683 4684 if (ret > 0) { 4685 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4686 APP_BUFFERED_IO, ret); 4687 } 4688 return ret; 4689 } 4690 4691 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4692 unsigned int flags) 4693 { 4694 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4695 4696 dec_page_count(sbi, F2FS_DIO_WRITE); 4697 if (error) 4698 return error; 4699 f2fs_update_time(sbi, REQ_TIME); 4700 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4701 return 0; 4702 } 4703 4704 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4705 .end_io = f2fs_dio_write_end_io, 4706 }; 4707 4708 static void f2fs_flush_buffered_write(struct address_space *mapping, 4709 loff_t start_pos, loff_t end_pos) 4710 { 4711 int ret; 4712 4713 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4714 if (ret < 0) 4715 return; 4716 invalidate_mapping_pages(mapping, 4717 start_pos >> PAGE_SHIFT, 4718 end_pos >> PAGE_SHIFT); 4719 } 4720 4721 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4722 bool *may_need_sync) 4723 { 4724 struct file *file = iocb->ki_filp; 4725 struct inode *inode = file_inode(file); 4726 struct f2fs_inode_info *fi = F2FS_I(inode); 4727 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4728 const bool do_opu = f2fs_lfs_mode(sbi); 4729 const loff_t pos = iocb->ki_pos; 4730 const ssize_t count = iov_iter_count(from); 4731 unsigned int dio_flags; 4732 struct iomap_dio *dio; 4733 ssize_t ret; 4734 4735 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4736 4737 if (iocb->ki_flags & IOCB_NOWAIT) { 4738 /* f2fs_convert_inline_inode() and block allocation can block */ 4739 if (f2fs_has_inline_data(inode) || 4740 !f2fs_overwrite_io(inode, pos, count)) { 4741 ret = -EAGAIN; 4742 goto out; 4743 } 4744 4745 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4746 ret = -EAGAIN; 4747 goto out; 4748 } 4749 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4750 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4751 ret = -EAGAIN; 4752 goto out; 4753 } 4754 } else { 4755 ret = f2fs_convert_inline_inode(inode); 4756 if (ret) 4757 goto out; 4758 4759 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4760 if (do_opu) 4761 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4762 } 4763 4764 /* 4765 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4766 * the higher-level function iomap_dio_rw() in order to ensure that the 4767 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4768 */ 4769 inc_page_count(sbi, F2FS_DIO_WRITE); 4770 dio_flags = 0; 4771 if (pos + count > inode->i_size) 4772 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4773 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4774 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4775 if (IS_ERR_OR_NULL(dio)) { 4776 ret = PTR_ERR_OR_ZERO(dio); 4777 if (ret == -ENOTBLK) 4778 ret = 0; 4779 if (ret != -EIOCBQUEUED) 4780 dec_page_count(sbi, F2FS_DIO_WRITE); 4781 } else { 4782 ret = iomap_dio_complete(dio); 4783 } 4784 4785 if (do_opu) 4786 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4787 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4788 4789 if (ret < 0) 4790 goto out; 4791 if (pos + ret > inode->i_size) 4792 f2fs_i_size_write(inode, pos + ret); 4793 if (!do_opu) 4794 set_inode_flag(inode, FI_UPDATE_WRITE); 4795 4796 if (iov_iter_count(from)) { 4797 ssize_t ret2; 4798 loff_t bufio_start_pos = iocb->ki_pos; 4799 4800 /* 4801 * The direct write was partial, so we need to fall back to a 4802 * buffered write for the remainder. 4803 */ 4804 4805 ret2 = f2fs_buffered_write_iter(iocb, from); 4806 if (iov_iter_count(from)) 4807 f2fs_write_failed(inode, iocb->ki_pos); 4808 if (ret2 < 0) 4809 goto out; 4810 4811 /* 4812 * Ensure that the pagecache pages are written to disk and 4813 * invalidated to preserve the expected O_DIRECT semantics. 4814 */ 4815 if (ret2 > 0) { 4816 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4817 4818 ret += ret2; 4819 4820 f2fs_flush_buffered_write(file->f_mapping, 4821 bufio_start_pos, 4822 bufio_end_pos); 4823 } 4824 } else { 4825 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4826 *may_need_sync = false; 4827 } 4828 out: 4829 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4830 return ret; 4831 } 4832 4833 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4834 { 4835 struct inode *inode = file_inode(iocb->ki_filp); 4836 const loff_t orig_pos = iocb->ki_pos; 4837 const size_t orig_count = iov_iter_count(from); 4838 loff_t target_size; 4839 bool dio; 4840 bool may_need_sync = true; 4841 int preallocated; 4842 ssize_t ret; 4843 4844 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4845 ret = -EIO; 4846 goto out; 4847 } 4848 4849 if (!f2fs_is_compress_backend_ready(inode)) { 4850 ret = -EOPNOTSUPP; 4851 goto out; 4852 } 4853 4854 if (iocb->ki_flags & IOCB_NOWAIT) { 4855 if (!inode_trylock(inode)) { 4856 ret = -EAGAIN; 4857 goto out; 4858 } 4859 } else { 4860 inode_lock(inode); 4861 } 4862 4863 ret = f2fs_write_checks(iocb, from); 4864 if (ret <= 0) 4865 goto out_unlock; 4866 4867 /* Determine whether we will do a direct write or a buffered write. */ 4868 dio = f2fs_should_use_dio(inode, iocb, from); 4869 4870 /* Possibly preallocate the blocks for the write. */ 4871 target_size = iocb->ki_pos + iov_iter_count(from); 4872 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4873 if (preallocated < 0) { 4874 ret = preallocated; 4875 } else { 4876 if (trace_f2fs_datawrite_start_enabled()) 4877 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4878 orig_count, WRITE); 4879 4880 /* Do the actual write. */ 4881 ret = dio ? 4882 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4883 f2fs_buffered_write_iter(iocb, from); 4884 4885 if (trace_f2fs_datawrite_end_enabled()) 4886 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4887 } 4888 4889 /* Don't leave any preallocated blocks around past i_size. */ 4890 if (preallocated && i_size_read(inode) < target_size) { 4891 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4892 filemap_invalidate_lock(inode->i_mapping); 4893 if (!f2fs_truncate(inode)) 4894 file_dont_truncate(inode); 4895 filemap_invalidate_unlock(inode->i_mapping); 4896 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4897 } else { 4898 file_dont_truncate(inode); 4899 } 4900 4901 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4902 out_unlock: 4903 inode_unlock(inode); 4904 out: 4905 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4906 4907 if (ret > 0 && may_need_sync) 4908 ret = generic_write_sync(iocb, ret); 4909 4910 /* If buffered IO was forced, flush and drop the data from 4911 * the page cache to preserve O_DIRECT semantics 4912 */ 4913 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4914 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4915 orig_pos, 4916 orig_pos + ret - 1); 4917 4918 return ret; 4919 } 4920 4921 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4922 int advice) 4923 { 4924 struct address_space *mapping; 4925 struct backing_dev_info *bdi; 4926 struct inode *inode = file_inode(filp); 4927 int err; 4928 4929 if (advice == POSIX_FADV_SEQUENTIAL) { 4930 if (S_ISFIFO(inode->i_mode)) 4931 return -ESPIPE; 4932 4933 mapping = filp->f_mapping; 4934 if (!mapping || len < 0) 4935 return -EINVAL; 4936 4937 bdi = inode_to_bdi(mapping->host); 4938 filp->f_ra.ra_pages = bdi->ra_pages * 4939 F2FS_I_SB(inode)->seq_file_ra_mul; 4940 spin_lock(&filp->f_lock); 4941 filp->f_mode &= ~FMODE_RANDOM; 4942 spin_unlock(&filp->f_lock); 4943 return 0; 4944 } 4945 4946 err = generic_fadvise(filp, offset, len, advice); 4947 if (!err && advice == POSIX_FADV_DONTNEED && 4948 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4949 f2fs_compressed_file(inode)) 4950 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4951 4952 return err; 4953 } 4954 4955 #ifdef CONFIG_COMPAT 4956 struct compat_f2fs_gc_range { 4957 u32 sync; 4958 compat_u64 start; 4959 compat_u64 len; 4960 }; 4961 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4962 struct compat_f2fs_gc_range) 4963 4964 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4965 { 4966 struct compat_f2fs_gc_range __user *urange; 4967 struct f2fs_gc_range range; 4968 int err; 4969 4970 urange = compat_ptr(arg); 4971 err = get_user(range.sync, &urange->sync); 4972 err |= get_user(range.start, &urange->start); 4973 err |= get_user(range.len, &urange->len); 4974 if (err) 4975 return -EFAULT; 4976 4977 return __f2fs_ioc_gc_range(file, &range); 4978 } 4979 4980 struct compat_f2fs_move_range { 4981 u32 dst_fd; 4982 compat_u64 pos_in; 4983 compat_u64 pos_out; 4984 compat_u64 len; 4985 }; 4986 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4987 struct compat_f2fs_move_range) 4988 4989 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4990 { 4991 struct compat_f2fs_move_range __user *urange; 4992 struct f2fs_move_range range; 4993 int err; 4994 4995 urange = compat_ptr(arg); 4996 err = get_user(range.dst_fd, &urange->dst_fd); 4997 err |= get_user(range.pos_in, &urange->pos_in); 4998 err |= get_user(range.pos_out, &urange->pos_out); 4999 err |= get_user(range.len, &urange->len); 5000 if (err) 5001 return -EFAULT; 5002 5003 return __f2fs_ioc_move_range(file, &range); 5004 } 5005 5006 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5007 { 5008 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 5009 return -EIO; 5010 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 5011 return -ENOSPC; 5012 5013 switch (cmd) { 5014 case FS_IOC32_GETVERSION: 5015 cmd = FS_IOC_GETVERSION; 5016 break; 5017 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 5018 return f2fs_compat_ioc_gc_range(file, arg); 5019 case F2FS_IOC32_MOVE_RANGE: 5020 return f2fs_compat_ioc_move_range(file, arg); 5021 case F2FS_IOC_START_ATOMIC_WRITE: 5022 case F2FS_IOC_START_ATOMIC_REPLACE: 5023 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5024 case F2FS_IOC_START_VOLATILE_WRITE: 5025 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5026 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5027 case F2FS_IOC_SHUTDOWN: 5028 case FITRIM: 5029 case FS_IOC_SET_ENCRYPTION_POLICY: 5030 case FS_IOC_GET_ENCRYPTION_PWSALT: 5031 case FS_IOC_GET_ENCRYPTION_POLICY: 5032 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5033 case FS_IOC_ADD_ENCRYPTION_KEY: 5034 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5035 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5036 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5037 case FS_IOC_GET_ENCRYPTION_NONCE: 5038 case F2FS_IOC_GARBAGE_COLLECT: 5039 case F2FS_IOC_WRITE_CHECKPOINT: 5040 case F2FS_IOC_DEFRAGMENT: 5041 case F2FS_IOC_FLUSH_DEVICE: 5042 case F2FS_IOC_GET_FEATURES: 5043 case F2FS_IOC_GET_PIN_FILE: 5044 case F2FS_IOC_SET_PIN_FILE: 5045 case F2FS_IOC_PRECACHE_EXTENTS: 5046 case F2FS_IOC_RESIZE_FS: 5047 case FS_IOC_ENABLE_VERITY: 5048 case FS_IOC_MEASURE_VERITY: 5049 case FS_IOC_READ_VERITY_METADATA: 5050 case FS_IOC_GETFSLABEL: 5051 case FS_IOC_SETFSLABEL: 5052 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5053 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5054 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5055 case F2FS_IOC_SEC_TRIM_FILE: 5056 case F2FS_IOC_GET_COMPRESS_OPTION: 5057 case F2FS_IOC_SET_COMPRESS_OPTION: 5058 case F2FS_IOC_DECOMPRESS_FILE: 5059 case F2FS_IOC_COMPRESS_FILE: 5060 break; 5061 default: 5062 return -ENOIOCTLCMD; 5063 } 5064 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5065 } 5066 #endif 5067 5068 const struct file_operations f2fs_file_operations = { 5069 .llseek = f2fs_llseek, 5070 .read_iter = f2fs_file_read_iter, 5071 .write_iter = f2fs_file_write_iter, 5072 .iopoll = iocb_bio_iopoll, 5073 .open = f2fs_file_open, 5074 .release = f2fs_release_file, 5075 .mmap = f2fs_file_mmap, 5076 .flush = f2fs_file_flush, 5077 .fsync = f2fs_sync_file, 5078 .fallocate = f2fs_fallocate, 5079 .unlocked_ioctl = f2fs_ioctl, 5080 #ifdef CONFIG_COMPAT 5081 .compat_ioctl = f2fs_compat_ioctl, 5082 #endif 5083 .splice_read = f2fs_file_splice_read, 5084 .splice_write = iter_file_splice_write, 5085 .fadvise = f2fs_file_fadvise, 5086 }; 5087