개념 설명 전체 · v6.6 / fs/f2fs/file.c
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) 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 f2fs_file_open(struct inode *inode, struct file *filp) 538 { 539 int err = fscrypt_file_open(inode, filp); 540 541 if (err) 542 return err; 543 544 if (!f2fs_is_compress_backend_ready(inode)) 545 return -EOPNOTSUPP; 546 547 err = fsverity_file_open(inode, filp); 548 if (err) 549 return err; 550 551 filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC; 552 filp->f_mode |= FMODE_CAN_ODIRECT; 553 554 return dquot_file_open(inode, filp); 555 } 556 557 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 558 { 559 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 560 struct f2fs_node *raw_node; 561 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 562 __le32 *addr; 563 int base = 0; 564 bool compressed_cluster = false; 565 int cluster_index = 0, valid_blocks = 0; 566 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 567 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 568 569 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode)) 570 base = get_extra_isize(dn->inode); 571 572 raw_node = F2FS_NODE(dn->node_page); 573 addr = blkaddr_in_node(raw_node) + base + ofs; 574 575 /* Assumption: truncation starts with cluster */ 576 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 577 block_t blkaddr = le32_to_cpu(*addr); 578 579 if (f2fs_compressed_file(dn->inode) && 580 !(cluster_index & (cluster_size - 1))) { 581 if (compressed_cluster) 582 f2fs_i_compr_blocks_update(dn->inode, 583 valid_blocks, false); 584 compressed_cluster = (blkaddr == COMPRESS_ADDR); 585 valid_blocks = 0; 586 } 587 588 if (blkaddr == NULL_ADDR) 589 continue; 590 591 dn->data_blkaddr = NULL_ADDR; 592 f2fs_set_data_blkaddr(dn); 593 594 if (__is_valid_data_blkaddr(blkaddr)) { 595 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 596 DATA_GENERIC_ENHANCE)) 597 continue; 598 if (compressed_cluster) 599 valid_blocks++; 600 } 601 602 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 603 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 604 605 f2fs_invalidate_blocks(sbi, blkaddr); 606 607 if (!released || blkaddr != COMPRESS_ADDR) 608 nr_free++; 609 } 610 611 if (compressed_cluster) 612 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 613 614 if (nr_free) { 615 pgoff_t fofs; 616 /* 617 * once we invalidate valid blkaddr in range [ofs, ofs + count], 618 * we will invalidate all blkaddr in the whole range. 619 */ 620 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 621 dn->inode) + ofs; 622 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 623 f2fs_update_age_extent_cache_range(dn, fofs, len); 624 dec_valid_block_count(sbi, dn->inode, nr_free); 625 } 626 dn->ofs_in_node = ofs; 627 628 f2fs_update_time(sbi, REQ_TIME); 629 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 630 dn->ofs_in_node, nr_free); 631 } 632 633 static int truncate_partial_data_page(struct inode *inode, u64 from, 634 bool cache_only) 635 { 636 loff_t offset = from & (PAGE_SIZE - 1); 637 pgoff_t index = from >> PAGE_SHIFT; 638 struct address_space *mapping = inode->i_mapping; 639 struct page *page; 640 641 if (!offset && !cache_only) 642 return 0; 643 644 if (cache_only) { 645 page = find_lock_page(mapping, index); 646 if (page && PageUptodate(page)) 647 goto truncate_out; 648 f2fs_put_page(page, 1); 649 return 0; 650 } 651 652 page = f2fs_get_lock_data_page(inode, index, true); 653 if (IS_ERR(page)) 654 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 655 truncate_out: 656 f2fs_wait_on_page_writeback(page, DATA, true, true); 657 zero_user(page, offset, PAGE_SIZE - offset); 658 659 /* An encrypted inode should have a key and truncate the last page. */ 660 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 661 if (!cache_only) 662 set_page_dirty(page); 663 f2fs_put_page(page, 1); 664 return 0; 665 } 666 667 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 668 { 669 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 670 struct dnode_of_data dn; 671 pgoff_t free_from; 672 int count = 0, err = 0; 673 struct page *ipage; 674 bool truncate_page = false; 675 676 trace_f2fs_truncate_blocks_enter(inode, from); 677 678 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 679 680 if (free_from >= max_file_blocks(inode)) 681 goto free_partial; 682 683 if (lock) 684 f2fs_lock_op(sbi); 685 686 ipage = f2fs_get_node_page(sbi, inode->i_ino); 687 if (IS_ERR(ipage)) { 688 err = PTR_ERR(ipage); 689 goto out; 690 } 691 692 if (f2fs_has_inline_data(inode)) { 693 f2fs_truncate_inline_inode(inode, ipage, from); 694 f2fs_put_page(ipage, 1); 695 truncate_page = true; 696 goto out; 697 } 698 699 set_new_dnode(&dn, inode, ipage, NULL, 0); 700 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 701 if (err) { 702 if (err == -ENOENT) 703 goto free_next; 704 goto out; 705 } 706 707 count = ADDRS_PER_PAGE(dn.node_page, inode); 708 709 count -= dn.ofs_in_node; 710 f2fs_bug_on(sbi, count < 0); 711 712 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 713 f2fs_truncate_data_blocks_range(&dn, count); 714 free_from += count; 715 } 716 717 f2fs_put_dnode(&dn); 718 free_next: 719 err = f2fs_truncate_inode_blocks(inode, free_from); 720 out: 721 if (lock) 722 f2fs_unlock_op(sbi); 723 free_partial: 724 /* lastly zero out the first data page */ 725 if (!err) 726 err = truncate_partial_data_page(inode, from, truncate_page); 727 728 trace_f2fs_truncate_blocks_exit(inode, err); 729 return err; 730 } 731 732 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 733 { 734 u64 free_from = from; 735 int err; 736 737 #ifdef CONFIG_F2FS_FS_COMPRESSION 738 /* 739 * for compressed file, only support cluster size 740 * aligned truncation. 741 */ 742 if (f2fs_compressed_file(inode)) 743 free_from = round_up(from, 744 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 745 #endif 746 747 err = f2fs_do_truncate_blocks(inode, free_from, lock); 748 if (err) 749 return err; 750 751 #ifdef CONFIG_F2FS_FS_COMPRESSION 752 /* 753 * For compressed file, after release compress blocks, don't allow write 754 * direct, but we should allow write direct after truncate to zero. 755 */ 756 if (f2fs_compressed_file(inode) && !free_from 757 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 758 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 759 760 if (from != free_from) { 761 err = f2fs_truncate_partial_cluster(inode, from, lock); 762 if (err) 763 return err; 764 } 765 #endif 766 767 return 0; 768 } 769 770 int f2fs_truncate(struct inode *inode) 771 { 772 int err; 773 774 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 775 return -EIO; 776 777 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 778 S_ISLNK(inode->i_mode))) 779 return 0; 780 781 trace_f2fs_truncate(inode); 782 783 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 784 return -EIO; 785 786 err = f2fs_dquot_initialize(inode); 787 if (err) 788 return err; 789 790 /* we should check inline_data size */ 791 if (!f2fs_may_inline_data(inode)) { 792 err = f2fs_convert_inline_inode(inode); 793 if (err) 794 return err; 795 } 796 797 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 798 if (err) 799 return err; 800 801 inode->i_mtime = inode_set_ctime_current(inode); 802 f2fs_mark_inode_dirty_sync(inode, false); 803 return 0; 804 } 805 806 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 807 { 808 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 809 810 if (!fscrypt_dio_supported(inode)) 811 return true; 812 if (fsverity_active(inode)) 813 return true; 814 if (f2fs_compressed_file(inode)) 815 return true; 816 817 /* disallow direct IO if any of devices has unaligned blksize */ 818 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 819 return true; 820 /* 821 * for blkzoned device, fallback direct IO to buffered IO, so 822 * all IOs can be serialized by log-structured write. 823 */ 824 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE)) 825 return true; 826 if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi)) 827 return true; 828 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 829 return true; 830 831 return false; 832 } 833 834 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 835 struct kstat *stat, u32 request_mask, unsigned int query_flags) 836 { 837 struct inode *inode = d_inode(path->dentry); 838 struct f2fs_inode_info *fi = F2FS_I(inode); 839 struct f2fs_inode *ri = NULL; 840 unsigned int flags; 841 842 if (f2fs_has_extra_attr(inode) && 843 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 844 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 845 stat->result_mask |= STATX_BTIME; 846 stat->btime.tv_sec = fi->i_crtime.tv_sec; 847 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 848 } 849 850 /* 851 * Return the DIO alignment restrictions if requested. We only return 852 * this information when requested, since on encrypted files it might 853 * take a fair bit of work to get if the file wasn't opened recently. 854 * 855 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 856 * cannot represent that, so in that case we report no DIO support. 857 */ 858 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 859 unsigned int bsize = i_blocksize(inode); 860 861 stat->result_mask |= STATX_DIOALIGN; 862 if (!f2fs_force_buffered_io(inode, WRITE)) { 863 stat->dio_mem_align = bsize; 864 stat->dio_offset_align = bsize; 865 } 866 } 867 868 flags = fi->i_flags; 869 if (flags & F2FS_COMPR_FL) 870 stat->attributes |= STATX_ATTR_COMPRESSED; 871 if (flags & F2FS_APPEND_FL) 872 stat->attributes |= STATX_ATTR_APPEND; 873 if (IS_ENCRYPTED(inode)) 874 stat->attributes |= STATX_ATTR_ENCRYPTED; 875 if (flags & F2FS_IMMUTABLE_FL) 876 stat->attributes |= STATX_ATTR_IMMUTABLE; 877 if (flags & F2FS_NODUMP_FL) 878 stat->attributes |= STATX_ATTR_NODUMP; 879 if (IS_VERITY(inode)) 880 stat->attributes |= STATX_ATTR_VERITY; 881 882 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 883 STATX_ATTR_APPEND | 884 STATX_ATTR_ENCRYPTED | 885 STATX_ATTR_IMMUTABLE | 886 STATX_ATTR_NODUMP | 887 STATX_ATTR_VERITY); 888 889 generic_fillattr(idmap, request_mask, inode, stat); 890 891 /* we need to show initial sectors used for inline_data/dentries */ 892 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 893 f2fs_has_inline_dentry(inode)) 894 stat->blocks += (stat->size + 511) >> 9; 895 896 return 0; 897 } 898 899 #ifdef CONFIG_F2FS_FS_POSIX_ACL 900 static void __setattr_copy(struct mnt_idmap *idmap, 901 struct inode *inode, const struct iattr *attr) 902 { 903 unsigned int ia_valid = attr->ia_valid; 904 905 i_uid_update(idmap, attr, inode); 906 i_gid_update(idmap, attr, inode); 907 if (ia_valid & ATTR_ATIME) 908 inode->i_atime = attr->ia_atime; 909 if (ia_valid & ATTR_MTIME) 910 inode->i_mtime = attr->ia_mtime; 911 if (ia_valid & ATTR_CTIME) 912 inode_set_ctime_to_ts(inode, attr->ia_ctime); 913 if (ia_valid & ATTR_MODE) { 914 umode_t mode = attr->ia_mode; 915 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode); 916 917 if (!vfsgid_in_group_p(vfsgid) && 918 !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) 919 mode &= ~S_ISGID; 920 set_acl_inode(inode, mode); 921 } 922 } 923 #else 924 #define __setattr_copy setattr_copy 925 #endif 926 927 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 928 struct iattr *attr) 929 { 930 struct inode *inode = d_inode(dentry); 931 int err; 932 933 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 934 return -EIO; 935 936 if (unlikely(IS_IMMUTABLE(inode))) 937 return -EPERM; 938 939 if (unlikely(IS_APPEND(inode) && 940 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 941 ATTR_GID | ATTR_TIMES_SET)))) 942 return -EPERM; 943 944 if ((attr->ia_valid & ATTR_SIZE) && 945 !f2fs_is_compress_backend_ready(inode)) 946 return -EOPNOTSUPP; 947 948 err = setattr_prepare(idmap, dentry, attr); 949 if (err) 950 return err; 951 952 err = fscrypt_prepare_setattr(dentry, attr); 953 if (err) 954 return err; 955 956 err = fsverity_prepare_setattr(dentry, attr); 957 if (err) 958 return err; 959 960 if (is_quota_modification(idmap, inode, attr)) { 961 err = f2fs_dquot_initialize(inode); 962 if (err) 963 return err; 964 } 965 if (i_uid_needs_update(idmap, attr, inode) || 966 i_gid_needs_update(idmap, attr, inode)) { 967 f2fs_lock_op(F2FS_I_SB(inode)); 968 err = dquot_transfer(idmap, inode, attr); 969 if (err) { 970 set_sbi_flag(F2FS_I_SB(inode), 971 SBI_QUOTA_NEED_REPAIR); 972 f2fs_unlock_op(F2FS_I_SB(inode)); 973 return err; 974 } 975 /* 976 * update uid/gid under lock_op(), so that dquot and inode can 977 * be updated atomically. 978 */ 979 i_uid_update(idmap, attr, inode); 980 i_gid_update(idmap, attr, inode); 981 f2fs_mark_inode_dirty_sync(inode, true); 982 f2fs_unlock_op(F2FS_I_SB(inode)); 983 } 984 985 if (attr->ia_valid & ATTR_SIZE) { 986 loff_t old_size = i_size_read(inode); 987 988 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 989 /* 990 * should convert inline inode before i_size_write to 991 * keep smaller than inline_data size with inline flag. 992 */ 993 err = f2fs_convert_inline_inode(inode); 994 if (err) 995 return err; 996 } 997 998 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 999 filemap_invalidate_lock(inode->i_mapping); 1000 1001 truncate_setsize(inode, attr->ia_size); 1002 1003 if (attr->ia_size <= old_size) 1004 err = f2fs_truncate(inode); 1005 /* 1006 * do not trim all blocks after i_size if target size is 1007 * larger than i_size. 1008 */ 1009 filemap_invalidate_unlock(inode->i_mapping); 1010 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1011 if (err) 1012 return err; 1013 1014 spin_lock(&F2FS_I(inode)->i_size_lock); 1015 inode->i_mtime = inode_set_ctime_current(inode); 1016 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1017 spin_unlock(&F2FS_I(inode)->i_size_lock); 1018 } 1019 1020 __setattr_copy(idmap, inode, attr); 1021 1022 if (attr->ia_valid & ATTR_MODE) { 1023 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1024 1025 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1026 if (!err) 1027 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1028 clear_inode_flag(inode, FI_ACL_MODE); 1029 } 1030 } 1031 1032 /* file size may changed here */ 1033 f2fs_mark_inode_dirty_sync(inode, true); 1034 1035 /* inode change will produce dirty node pages flushed by checkpoint */ 1036 f2fs_balance_fs(F2FS_I_SB(inode), true); 1037 1038 return err; 1039 } 1040 1041 const struct inode_operations f2fs_file_inode_operations = { 1042 .getattr = f2fs_getattr, 1043 .setattr = f2fs_setattr, 1044 .get_inode_acl = f2fs_get_acl, 1045 .set_acl = f2fs_set_acl, 1046 .listxattr = f2fs_listxattr, 1047 .fiemap = f2fs_fiemap, 1048 .fileattr_get = f2fs_fileattr_get, 1049 .fileattr_set = f2fs_fileattr_set, 1050 }; 1051 1052 static int fill_zero(struct inode *inode, pgoff_t index, 1053 loff_t start, loff_t len) 1054 { 1055 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1056 struct page *page; 1057 1058 if (!len) 1059 return 0; 1060 1061 f2fs_balance_fs(sbi, true); 1062 1063 f2fs_lock_op(sbi); 1064 page = f2fs_get_new_data_page(inode, NULL, index, false); 1065 f2fs_unlock_op(sbi); 1066 1067 if (IS_ERR(page)) 1068 return PTR_ERR(page); 1069 1070 f2fs_wait_on_page_writeback(page, DATA, true, true); 1071 zero_user(page, start, len); 1072 set_page_dirty(page); 1073 f2fs_put_page(page, 1); 1074 return 0; 1075 } 1076 1077 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1078 { 1079 int err; 1080 1081 while (pg_start < pg_end) { 1082 struct dnode_of_data dn; 1083 pgoff_t end_offset, count; 1084 1085 set_new_dnode(&dn, inode, NULL, NULL, 0); 1086 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1087 if (err) { 1088 if (err == -ENOENT) { 1089 pg_start = f2fs_get_next_page_offset(&dn, 1090 pg_start); 1091 continue; 1092 } 1093 return err; 1094 } 1095 1096 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1097 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1098 1099 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1100 1101 f2fs_truncate_data_blocks_range(&dn, count); 1102 f2fs_put_dnode(&dn); 1103 1104 pg_start += count; 1105 } 1106 return 0; 1107 } 1108 1109 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1110 { 1111 pgoff_t pg_start, pg_end; 1112 loff_t off_start, off_end; 1113 int ret; 1114 1115 ret = f2fs_convert_inline_inode(inode); 1116 if (ret) 1117 return ret; 1118 1119 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1120 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1121 1122 off_start = offset & (PAGE_SIZE - 1); 1123 off_end = (offset + len) & (PAGE_SIZE - 1); 1124 1125 if (pg_start == pg_end) { 1126 ret = fill_zero(inode, pg_start, off_start, 1127 off_end - off_start); 1128 if (ret) 1129 return ret; 1130 } else { 1131 if (off_start) { 1132 ret = fill_zero(inode, pg_start++, off_start, 1133 PAGE_SIZE - off_start); 1134 if (ret) 1135 return ret; 1136 } 1137 if (off_end) { 1138 ret = fill_zero(inode, pg_end, 0, off_end); 1139 if (ret) 1140 return ret; 1141 } 1142 1143 if (pg_start < pg_end) { 1144 loff_t blk_start, blk_end; 1145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1146 1147 f2fs_balance_fs(sbi, true); 1148 1149 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1150 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1151 1152 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1153 filemap_invalidate_lock(inode->i_mapping); 1154 1155 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1156 1157 f2fs_lock_op(sbi); 1158 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1159 f2fs_unlock_op(sbi); 1160 1161 filemap_invalidate_unlock(inode->i_mapping); 1162 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1163 } 1164 } 1165 1166 return ret; 1167 } 1168 1169 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1170 int *do_replace, pgoff_t off, pgoff_t len) 1171 { 1172 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1173 struct dnode_of_data dn; 1174 int ret, done, i; 1175 1176 next_dnode: 1177 set_new_dnode(&dn, inode, NULL, NULL, 0); 1178 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1179 if (ret && ret != -ENOENT) { 1180 return ret; 1181 } else if (ret == -ENOENT) { 1182 if (dn.max_level == 0) 1183 return -ENOENT; 1184 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1185 dn.ofs_in_node, len); 1186 blkaddr += done; 1187 do_replace += done; 1188 goto next; 1189 } 1190 1191 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1192 dn.ofs_in_node, len); 1193 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1194 *blkaddr = f2fs_data_blkaddr(&dn); 1195 1196 if (__is_valid_data_blkaddr(*blkaddr) && 1197 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1198 DATA_GENERIC_ENHANCE)) { 1199 f2fs_put_dnode(&dn); 1200 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1201 return -EFSCORRUPTED; 1202 } 1203 1204 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1205 1206 if (f2fs_lfs_mode(sbi)) { 1207 f2fs_put_dnode(&dn); 1208 return -EOPNOTSUPP; 1209 } 1210 1211 /* do not invalidate this block address */ 1212 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1213 *do_replace = 1; 1214 } 1215 } 1216 f2fs_put_dnode(&dn); 1217 next: 1218 len -= done; 1219 off += done; 1220 if (len) 1221 goto next_dnode; 1222 return 0; 1223 } 1224 1225 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1226 int *do_replace, pgoff_t off, int len) 1227 { 1228 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1229 struct dnode_of_data dn; 1230 int ret, i; 1231 1232 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1233 if (*do_replace == 0) 1234 continue; 1235 1236 set_new_dnode(&dn, inode, NULL, NULL, 0); 1237 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1238 if (ret) { 1239 dec_valid_block_count(sbi, inode, 1); 1240 f2fs_invalidate_blocks(sbi, *blkaddr); 1241 } else { 1242 f2fs_update_data_blkaddr(&dn, *blkaddr); 1243 } 1244 f2fs_put_dnode(&dn); 1245 } 1246 return 0; 1247 } 1248 1249 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1250 block_t *blkaddr, int *do_replace, 1251 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1252 { 1253 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1254 pgoff_t i = 0; 1255 int ret; 1256 1257 while (i < len) { 1258 if (blkaddr[i] == NULL_ADDR && !full) { 1259 i++; 1260 continue; 1261 } 1262 1263 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1264 struct dnode_of_data dn; 1265 struct node_info ni; 1266 size_t new_size; 1267 pgoff_t ilen; 1268 1269 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1270 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1271 if (ret) 1272 return ret; 1273 1274 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1275 if (ret) { 1276 f2fs_put_dnode(&dn); 1277 return ret; 1278 } 1279 1280 ilen = min((pgoff_t) 1281 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1282 dn.ofs_in_node, len - i); 1283 do { 1284 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1285 f2fs_truncate_data_blocks_range(&dn, 1); 1286 1287 if (do_replace[i]) { 1288 f2fs_i_blocks_write(src_inode, 1289 1, false, false); 1290 f2fs_i_blocks_write(dst_inode, 1291 1, true, false); 1292 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1293 blkaddr[i], ni.version, true, false); 1294 1295 do_replace[i] = 0; 1296 } 1297 dn.ofs_in_node++; 1298 i++; 1299 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1300 if (dst_inode->i_size < new_size) 1301 f2fs_i_size_write(dst_inode, new_size); 1302 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1303 1304 f2fs_put_dnode(&dn); 1305 } else { 1306 struct page *psrc, *pdst; 1307 1308 psrc = f2fs_get_lock_data_page(src_inode, 1309 src + i, true); 1310 if (IS_ERR(psrc)) 1311 return PTR_ERR(psrc); 1312 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1313 true); 1314 if (IS_ERR(pdst)) { 1315 f2fs_put_page(psrc, 1); 1316 return PTR_ERR(pdst); 1317 } 1318 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1319 set_page_dirty(pdst); 1320 f2fs_put_page(pdst, 1); 1321 f2fs_put_page(psrc, 1); 1322 1323 ret = f2fs_truncate_hole(src_inode, 1324 src + i, src + i + 1); 1325 if (ret) 1326 return ret; 1327 i++; 1328 } 1329 } 1330 return 0; 1331 } 1332 1333 static int __exchange_data_block(struct inode *src_inode, 1334 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1335 pgoff_t len, bool full) 1336 { 1337 block_t *src_blkaddr; 1338 int *do_replace; 1339 pgoff_t olen; 1340 int ret; 1341 1342 while (len) { 1343 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1344 1345 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1346 array_size(olen, sizeof(block_t)), 1347 GFP_NOFS); 1348 if (!src_blkaddr) 1349 return -ENOMEM; 1350 1351 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1352 array_size(olen, sizeof(int)), 1353 GFP_NOFS); 1354 if (!do_replace) { 1355 kvfree(src_blkaddr); 1356 return -ENOMEM; 1357 } 1358 1359 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1360 do_replace, src, olen); 1361 if (ret) 1362 goto roll_back; 1363 1364 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1365 do_replace, src, dst, olen, full); 1366 if (ret) 1367 goto roll_back; 1368 1369 src += olen; 1370 dst += olen; 1371 len -= olen; 1372 1373 kvfree(src_blkaddr); 1374 kvfree(do_replace); 1375 } 1376 return 0; 1377 1378 roll_back: 1379 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1380 kvfree(src_blkaddr); 1381 kvfree(do_replace); 1382 return ret; 1383 } 1384 1385 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1386 { 1387 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1388 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1389 pgoff_t start = offset >> PAGE_SHIFT; 1390 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1391 int ret; 1392 1393 f2fs_balance_fs(sbi, true); 1394 1395 /* avoid gc operation during block exchange */ 1396 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1397 filemap_invalidate_lock(inode->i_mapping); 1398 1399 f2fs_lock_op(sbi); 1400 f2fs_drop_extent_tree(inode); 1401 truncate_pagecache(inode, offset); 1402 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1403 f2fs_unlock_op(sbi); 1404 1405 filemap_invalidate_unlock(inode->i_mapping); 1406 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1407 return ret; 1408 } 1409 1410 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1411 { 1412 loff_t new_size; 1413 int ret; 1414 1415 if (offset + len >= i_size_read(inode)) 1416 return -EINVAL; 1417 1418 /* collapse range should be aligned to block size of f2fs. */ 1419 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1420 return -EINVAL; 1421 1422 ret = f2fs_convert_inline_inode(inode); 1423 if (ret) 1424 return ret; 1425 1426 /* write out all dirty pages from offset */ 1427 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1428 if (ret) 1429 return ret; 1430 1431 ret = f2fs_do_collapse(inode, offset, len); 1432 if (ret) 1433 return ret; 1434 1435 /* write out all moved pages, if possible */ 1436 filemap_invalidate_lock(inode->i_mapping); 1437 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1438 truncate_pagecache(inode, offset); 1439 1440 new_size = i_size_read(inode) - len; 1441 ret = f2fs_truncate_blocks(inode, new_size, true); 1442 filemap_invalidate_unlock(inode->i_mapping); 1443 if (!ret) 1444 f2fs_i_size_write(inode, new_size); 1445 return ret; 1446 } 1447 1448 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1449 pgoff_t end) 1450 { 1451 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1452 pgoff_t index = start; 1453 unsigned int ofs_in_node = dn->ofs_in_node; 1454 blkcnt_t count = 0; 1455 int ret; 1456 1457 for (; index < end; index++, dn->ofs_in_node++) { 1458 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1459 count++; 1460 } 1461 1462 dn->ofs_in_node = ofs_in_node; 1463 ret = f2fs_reserve_new_blocks(dn, count); 1464 if (ret) 1465 return ret; 1466 1467 dn->ofs_in_node = ofs_in_node; 1468 for (index = start; index < end; index++, dn->ofs_in_node++) { 1469 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1470 /* 1471 * f2fs_reserve_new_blocks will not guarantee entire block 1472 * allocation. 1473 */ 1474 if (dn->data_blkaddr == NULL_ADDR) { 1475 ret = -ENOSPC; 1476 break; 1477 } 1478 1479 if (dn->data_blkaddr == NEW_ADDR) 1480 continue; 1481 1482 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1483 DATA_GENERIC_ENHANCE)) { 1484 ret = -EFSCORRUPTED; 1485 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1486 break; 1487 } 1488 1489 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1490 dn->data_blkaddr = NEW_ADDR; 1491 f2fs_set_data_blkaddr(dn); 1492 } 1493 1494 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1495 f2fs_update_age_extent_cache_range(dn, start, index - start); 1496 1497 return ret; 1498 } 1499 1500 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1501 int mode) 1502 { 1503 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1504 struct address_space *mapping = inode->i_mapping; 1505 pgoff_t index, pg_start, pg_end; 1506 loff_t new_size = i_size_read(inode); 1507 loff_t off_start, off_end; 1508 int ret = 0; 1509 1510 ret = inode_newsize_ok(inode, (len + offset)); 1511 if (ret) 1512 return ret; 1513 1514 ret = f2fs_convert_inline_inode(inode); 1515 if (ret) 1516 return ret; 1517 1518 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1519 if (ret) 1520 return ret; 1521 1522 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1523 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1524 1525 off_start = offset & (PAGE_SIZE - 1); 1526 off_end = (offset + len) & (PAGE_SIZE - 1); 1527 1528 if (pg_start == pg_end) { 1529 ret = fill_zero(inode, pg_start, off_start, 1530 off_end - off_start); 1531 if (ret) 1532 return ret; 1533 1534 new_size = max_t(loff_t, new_size, offset + len); 1535 } else { 1536 if (off_start) { 1537 ret = fill_zero(inode, pg_start++, off_start, 1538 PAGE_SIZE - off_start); 1539 if (ret) 1540 return ret; 1541 1542 new_size = max_t(loff_t, new_size, 1543 (loff_t)pg_start << PAGE_SHIFT); 1544 } 1545 1546 for (index = pg_start; index < pg_end;) { 1547 struct dnode_of_data dn; 1548 unsigned int end_offset; 1549 pgoff_t end; 1550 1551 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1552 filemap_invalidate_lock(mapping); 1553 1554 truncate_pagecache_range(inode, 1555 (loff_t)index << PAGE_SHIFT, 1556 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1557 1558 f2fs_lock_op(sbi); 1559 1560 set_new_dnode(&dn, inode, NULL, NULL, 0); 1561 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1562 if (ret) { 1563 f2fs_unlock_op(sbi); 1564 filemap_invalidate_unlock(mapping); 1565 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1566 goto out; 1567 } 1568 1569 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1570 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1571 1572 ret = f2fs_do_zero_range(&dn, index, end); 1573 f2fs_put_dnode(&dn); 1574 1575 f2fs_unlock_op(sbi); 1576 filemap_invalidate_unlock(mapping); 1577 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1578 1579 f2fs_balance_fs(sbi, dn.node_changed); 1580 1581 if (ret) 1582 goto out; 1583 1584 index = end; 1585 new_size = max_t(loff_t, new_size, 1586 (loff_t)index << PAGE_SHIFT); 1587 } 1588 1589 if (off_end) { 1590 ret = fill_zero(inode, pg_end, 0, off_end); 1591 if (ret) 1592 goto out; 1593 1594 new_size = max_t(loff_t, new_size, offset + len); 1595 } 1596 } 1597 1598 out: 1599 if (new_size > i_size_read(inode)) { 1600 if (mode & FALLOC_FL_KEEP_SIZE) 1601 file_set_keep_isize(inode); 1602 else 1603 f2fs_i_size_write(inode, new_size); 1604 } 1605 return ret; 1606 } 1607 1608 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1609 { 1610 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1611 struct address_space *mapping = inode->i_mapping; 1612 pgoff_t nr, pg_start, pg_end, delta, idx; 1613 loff_t new_size; 1614 int ret = 0; 1615 1616 new_size = i_size_read(inode) + len; 1617 ret = inode_newsize_ok(inode, new_size); 1618 if (ret) 1619 return ret; 1620 1621 if (offset >= i_size_read(inode)) 1622 return -EINVAL; 1623 1624 /* insert range should be aligned to block size of f2fs. */ 1625 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1626 return -EINVAL; 1627 1628 ret = f2fs_convert_inline_inode(inode); 1629 if (ret) 1630 return ret; 1631 1632 f2fs_balance_fs(sbi, true); 1633 1634 filemap_invalidate_lock(mapping); 1635 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1636 filemap_invalidate_unlock(mapping); 1637 if (ret) 1638 return ret; 1639 1640 /* write out all dirty pages from offset */ 1641 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1642 if (ret) 1643 return ret; 1644 1645 pg_start = offset >> PAGE_SHIFT; 1646 pg_end = (offset + len) >> PAGE_SHIFT; 1647 delta = pg_end - pg_start; 1648 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1649 1650 /* avoid gc operation during block exchange */ 1651 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1652 filemap_invalidate_lock(mapping); 1653 truncate_pagecache(inode, offset); 1654 1655 while (!ret && idx > pg_start) { 1656 nr = idx - pg_start; 1657 if (nr > delta) 1658 nr = delta; 1659 idx -= nr; 1660 1661 f2fs_lock_op(sbi); 1662 f2fs_drop_extent_tree(inode); 1663 1664 ret = __exchange_data_block(inode, inode, idx, 1665 idx + delta, nr, false); 1666 f2fs_unlock_op(sbi); 1667 } 1668 filemap_invalidate_unlock(mapping); 1669 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1670 1671 /* write out all moved pages, if possible */ 1672 filemap_invalidate_lock(mapping); 1673 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1674 truncate_pagecache(inode, offset); 1675 filemap_invalidate_unlock(mapping); 1676 1677 if (!ret) 1678 f2fs_i_size_write(inode, new_size); 1679 return ret; 1680 } 1681 1682 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1683 loff_t len, int mode) 1684 { 1685 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1686 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1687 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1688 .m_may_create = true }; 1689 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1690 .init_gc_type = FG_GC, 1691 .should_migrate_blocks = false, 1692 .err_gc_skipped = true, 1693 .nr_free_secs = 0 }; 1694 pgoff_t pg_start, pg_end; 1695 loff_t new_size; 1696 loff_t off_end; 1697 block_t expanded = 0; 1698 int err; 1699 1700 err = inode_newsize_ok(inode, (len + offset)); 1701 if (err) 1702 return err; 1703 1704 err = f2fs_convert_inline_inode(inode); 1705 if (err) 1706 return err; 1707 1708 f2fs_balance_fs(sbi, true); 1709 1710 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1711 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1712 off_end = (offset + len) & (PAGE_SIZE - 1); 1713 1714 map.m_lblk = pg_start; 1715 map.m_len = pg_end - pg_start; 1716 if (off_end) 1717 map.m_len++; 1718 1719 if (!map.m_len) 1720 return 0; 1721 1722 if (f2fs_is_pinned_file(inode)) { 1723 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1724 block_t sec_len = roundup(map.m_len, sec_blks); 1725 1726 map.m_len = sec_blks; 1727 next_alloc: 1728 if (has_not_enough_free_secs(sbi, 0, 1729 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1730 f2fs_down_write(&sbi->gc_lock); 1731 stat_inc_gc_call_count(sbi, FOREGROUND); 1732 err = f2fs_gc(sbi, &gc_control); 1733 if (err && err != -ENODATA) 1734 goto out_err; 1735 } 1736 1737 f2fs_down_write(&sbi->pin_sem); 1738 1739 f2fs_lock_op(sbi); 1740 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false); 1741 f2fs_unlock_op(sbi); 1742 1743 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1744 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1745 file_dont_truncate(inode); 1746 1747 f2fs_up_write(&sbi->pin_sem); 1748 1749 expanded += map.m_len; 1750 sec_len -= map.m_len; 1751 map.m_lblk += map.m_len; 1752 if (!err && sec_len) 1753 goto next_alloc; 1754 1755 map.m_len = expanded; 1756 } else { 1757 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1758 expanded = map.m_len; 1759 } 1760 out_err: 1761 if (err) { 1762 pgoff_t last_off; 1763 1764 if (!expanded) 1765 return err; 1766 1767 last_off = pg_start + expanded - 1; 1768 1769 /* update new size to the failed position */ 1770 new_size = (last_off == pg_end) ? offset + len : 1771 (loff_t)(last_off + 1) << PAGE_SHIFT; 1772 } else { 1773 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1774 } 1775 1776 if (new_size > i_size_read(inode)) { 1777 if (mode & FALLOC_FL_KEEP_SIZE) 1778 file_set_keep_isize(inode); 1779 else 1780 f2fs_i_size_write(inode, new_size); 1781 } 1782 1783 return err; 1784 } 1785 1786 static long f2fs_fallocate(struct file *file, int mode, 1787 loff_t offset, loff_t len) 1788 { 1789 struct inode *inode = file_inode(file); 1790 long ret = 0; 1791 1792 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1793 return -EIO; 1794 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1795 return -ENOSPC; 1796 if (!f2fs_is_compress_backend_ready(inode)) 1797 return -EOPNOTSUPP; 1798 1799 /* f2fs only support ->fallocate for regular file */ 1800 if (!S_ISREG(inode->i_mode)) 1801 return -EINVAL; 1802 1803 if (IS_ENCRYPTED(inode) && 1804 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1805 return -EOPNOTSUPP; 1806 1807 /* 1808 * Pinned file should not support partial truncation since the block 1809 * can be used by applications. 1810 */ 1811 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1812 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1813 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1814 return -EOPNOTSUPP; 1815 1816 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1817 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1818 FALLOC_FL_INSERT_RANGE)) 1819 return -EOPNOTSUPP; 1820 1821 inode_lock(inode); 1822 1823 ret = file_modified(file); 1824 if (ret) 1825 goto out; 1826 1827 if (mode & FALLOC_FL_PUNCH_HOLE) { 1828 if (offset >= inode->i_size) 1829 goto out; 1830 1831 ret = f2fs_punch_hole(inode, offset, len); 1832 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1833 ret = f2fs_collapse_range(inode, offset, len); 1834 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1835 ret = f2fs_zero_range(inode, offset, len, mode); 1836 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1837 ret = f2fs_insert_range(inode, offset, len); 1838 } else { 1839 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1840 } 1841 1842 if (!ret) { 1843 inode->i_mtime = inode_set_ctime_current(inode); 1844 f2fs_mark_inode_dirty_sync(inode, false); 1845 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1846 } 1847 1848 out: 1849 inode_unlock(inode); 1850 1851 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1852 return ret; 1853 } 1854 1855 static int f2fs_release_file(struct inode *inode, struct file *filp) 1856 { 1857 /* 1858 * f2fs_release_file is called at every close calls. So we should 1859 * not drop any inmemory pages by close called by other process. 1860 */ 1861 if (!(filp->f_mode & FMODE_WRITE) || 1862 atomic_read(&inode->i_writecount) != 1) 1863 return 0; 1864 1865 inode_lock(inode); 1866 f2fs_abort_atomic_write(inode, true); 1867 inode_unlock(inode); 1868 1869 return 0; 1870 } 1871 1872 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1873 { 1874 struct inode *inode = file_inode(file); 1875 1876 /* 1877 * If the process doing a transaction is crashed, we should do 1878 * roll-back. Otherwise, other reader/write can see corrupted database 1879 * until all the writers close its file. Since this should be done 1880 * before dropping file lock, it needs to do in ->flush. 1881 */ 1882 if (F2FS_I(inode)->atomic_write_task == current && 1883 (current->flags & PF_EXITING)) { 1884 inode_lock(inode); 1885 f2fs_abort_atomic_write(inode, true); 1886 inode_unlock(inode); 1887 } 1888 1889 return 0; 1890 } 1891 1892 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1893 { 1894 struct f2fs_inode_info *fi = F2FS_I(inode); 1895 u32 masked_flags = fi->i_flags & mask; 1896 1897 /* mask can be shrunk by flags_valid selector */ 1898 iflags &= mask; 1899 1900 /* Is it quota file? Do not allow user to mess with it */ 1901 if (IS_NOQUOTA(inode)) 1902 return -EPERM; 1903 1904 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1905 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1906 return -EOPNOTSUPP; 1907 if (!f2fs_empty_dir(inode)) 1908 return -ENOTEMPTY; 1909 } 1910 1911 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1912 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1913 return -EOPNOTSUPP; 1914 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1915 return -EINVAL; 1916 } 1917 1918 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1919 if (masked_flags & F2FS_COMPR_FL) { 1920 if (!f2fs_disable_compressed_file(inode)) 1921 return -EINVAL; 1922 } else { 1923 /* try to convert inline_data to support compression */ 1924 int err = f2fs_convert_inline_inode(inode); 1925 if (err) 1926 return err; 1927 1928 f2fs_down_write(&F2FS_I(inode)->i_sem); 1929 if (!f2fs_may_compress(inode) || 1930 (S_ISREG(inode->i_mode) && 1931 F2FS_HAS_BLOCKS(inode))) { 1932 f2fs_up_write(&F2FS_I(inode)->i_sem); 1933 return -EINVAL; 1934 } 1935 err = set_compress_context(inode); 1936 f2fs_up_write(&F2FS_I(inode)->i_sem); 1937 1938 if (err) 1939 return err; 1940 } 1941 } 1942 1943 fi->i_flags = iflags | (fi->i_flags & ~mask); 1944 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1945 (fi->i_flags & F2FS_NOCOMP_FL)); 1946 1947 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1948 set_inode_flag(inode, FI_PROJ_INHERIT); 1949 else 1950 clear_inode_flag(inode, FI_PROJ_INHERIT); 1951 1952 inode_set_ctime_current(inode); 1953 f2fs_set_inode_flags(inode); 1954 f2fs_mark_inode_dirty_sync(inode, true); 1955 return 0; 1956 } 1957 1958 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1959 1960 /* 1961 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1962 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1963 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1964 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1965 * 1966 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1967 * FS_IOC_FSSETXATTR is done by the VFS. 1968 */ 1969 1970 static const struct { 1971 u32 iflag; 1972 u32 fsflag; 1973 } f2fs_fsflags_map[] = { 1974 { F2FS_COMPR_FL, FS_COMPR_FL }, 1975 { F2FS_SYNC_FL, FS_SYNC_FL }, 1976 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1977 { F2FS_APPEND_FL, FS_APPEND_FL }, 1978 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1979 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1980 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1981 { F2FS_INDEX_FL, FS_INDEX_FL }, 1982 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1983 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1984 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1985 }; 1986 1987 #define F2FS_GETTABLE_FS_FL ( \ 1988 FS_COMPR_FL | \ 1989 FS_SYNC_FL | \ 1990 FS_IMMUTABLE_FL | \ 1991 FS_APPEND_FL | \ 1992 FS_NODUMP_FL | \ 1993 FS_NOATIME_FL | \ 1994 FS_NOCOMP_FL | \ 1995 FS_INDEX_FL | \ 1996 FS_DIRSYNC_FL | \ 1997 FS_PROJINHERIT_FL | \ 1998 FS_ENCRYPT_FL | \ 1999 FS_INLINE_DATA_FL | \ 2000 FS_NOCOW_FL | \ 2001 FS_VERITY_FL | \ 2002 FS_CASEFOLD_FL) 2003 2004 #define F2FS_SETTABLE_FS_FL ( \ 2005 FS_COMPR_FL | \ 2006 FS_SYNC_FL | \ 2007 FS_IMMUTABLE_FL | \ 2008 FS_APPEND_FL | \ 2009 FS_NODUMP_FL | \ 2010 FS_NOATIME_FL | \ 2011 FS_NOCOMP_FL | \ 2012 FS_DIRSYNC_FL | \ 2013 FS_PROJINHERIT_FL | \ 2014 FS_CASEFOLD_FL) 2015 2016 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2017 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2018 { 2019 u32 fsflags = 0; 2020 int i; 2021 2022 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2023 if (iflags & f2fs_fsflags_map[i].iflag) 2024 fsflags |= f2fs_fsflags_map[i].fsflag; 2025 2026 return fsflags; 2027 } 2028 2029 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2030 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2031 { 2032 u32 iflags = 0; 2033 int i; 2034 2035 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2036 if (fsflags & f2fs_fsflags_map[i].fsflag) 2037 iflags |= f2fs_fsflags_map[i].iflag; 2038 2039 return iflags; 2040 } 2041 2042 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2043 { 2044 struct inode *inode = file_inode(filp); 2045 2046 return put_user(inode->i_generation, (int __user *)arg); 2047 } 2048 2049 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2050 { 2051 struct inode *inode = file_inode(filp); 2052 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2053 struct f2fs_inode_info *fi = F2FS_I(inode); 2054 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2055 struct inode *pinode; 2056 loff_t isize; 2057 int ret; 2058 2059 if (!inode_owner_or_capable(idmap, inode)) 2060 return -EACCES; 2061 2062 if (!S_ISREG(inode->i_mode)) 2063 return -EINVAL; 2064 2065 if (filp->f_flags & O_DIRECT) 2066 return -EINVAL; 2067 2068 ret = mnt_want_write_file(filp); 2069 if (ret) 2070 return ret; 2071 2072 inode_lock(inode); 2073 2074 if (!f2fs_disable_compressed_file(inode)) { 2075 ret = -EINVAL; 2076 goto out; 2077 } 2078 2079 if (f2fs_is_atomic_file(inode)) 2080 goto out; 2081 2082 ret = f2fs_convert_inline_inode(inode); 2083 if (ret) 2084 goto out; 2085 2086 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2087 2088 /* 2089 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2090 * f2fs_is_atomic_file. 2091 */ 2092 if (get_dirty_pages(inode)) 2093 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2094 inode->i_ino, get_dirty_pages(inode)); 2095 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2096 if (ret) { 2097 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2098 goto out; 2099 } 2100 2101 /* Check if the inode already has a COW inode */ 2102 if (fi->cow_inode == NULL) { 2103 /* Create a COW inode for atomic write */ 2104 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2105 if (IS_ERR(pinode)) { 2106 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2107 ret = PTR_ERR(pinode); 2108 goto out; 2109 } 2110 2111 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2112 iput(pinode); 2113 if (ret) { 2114 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2115 goto out; 2116 } 2117 2118 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2119 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2120 } else { 2121 /* Reuse the already created COW inode */ 2122 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2123 if (ret) { 2124 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2125 goto out; 2126 } 2127 } 2128 2129 f2fs_write_inode(inode, NULL); 2130 2131 stat_inc_atomic_inode(inode); 2132 2133 set_inode_flag(inode, FI_ATOMIC_FILE); 2134 2135 isize = i_size_read(inode); 2136 fi->original_i_size = isize; 2137 if (truncate) { 2138 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2139 truncate_inode_pages_final(inode->i_mapping); 2140 f2fs_i_size_write(inode, 0); 2141 isize = 0; 2142 } 2143 f2fs_i_size_write(fi->cow_inode, isize); 2144 2145 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2146 2147 f2fs_update_time(sbi, REQ_TIME); 2148 fi->atomic_write_task = current; 2149 stat_update_max_atomic_write(inode); 2150 fi->atomic_write_cnt = 0; 2151 out: 2152 inode_unlock(inode); 2153 mnt_drop_write_file(filp); 2154 return ret; 2155 } 2156 2157 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2158 { 2159 struct inode *inode = file_inode(filp); 2160 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2161 int ret; 2162 2163 if (!inode_owner_or_capable(idmap, inode)) 2164 return -EACCES; 2165 2166 ret = mnt_want_write_file(filp); 2167 if (ret) 2168 return ret; 2169 2170 f2fs_balance_fs(F2FS_I_SB(inode), true); 2171 2172 inode_lock(inode); 2173 2174 if (f2fs_is_atomic_file(inode)) { 2175 ret = f2fs_commit_atomic_write(inode); 2176 if (!ret) 2177 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2178 2179 f2fs_abort_atomic_write(inode, ret); 2180 } else { 2181 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2182 } 2183 2184 inode_unlock(inode); 2185 mnt_drop_write_file(filp); 2186 return ret; 2187 } 2188 2189 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2190 { 2191 struct inode *inode = file_inode(filp); 2192 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2193 int ret; 2194 2195 if (!inode_owner_or_capable(idmap, inode)) 2196 return -EACCES; 2197 2198 ret = mnt_want_write_file(filp); 2199 if (ret) 2200 return ret; 2201 2202 inode_lock(inode); 2203 2204 f2fs_abort_atomic_write(inode, true); 2205 2206 inode_unlock(inode); 2207 2208 mnt_drop_write_file(filp); 2209 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2210 return ret; 2211 } 2212 2213 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2214 { 2215 struct inode *inode = file_inode(filp); 2216 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2217 struct super_block *sb = sbi->sb; 2218 __u32 in; 2219 int ret = 0; 2220 2221 if (!capable(CAP_SYS_ADMIN)) 2222 return -EPERM; 2223 2224 if (get_user(in, (__u32 __user *)arg)) 2225 return -EFAULT; 2226 2227 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2228 ret = mnt_want_write_file(filp); 2229 if (ret) { 2230 if (ret == -EROFS) { 2231 ret = 0; 2232 f2fs_stop_checkpoint(sbi, false, 2233 STOP_CP_REASON_SHUTDOWN); 2234 trace_f2fs_shutdown(sbi, in, ret); 2235 } 2236 return ret; 2237 } 2238 } 2239 2240 switch (in) { 2241 case F2FS_GOING_DOWN_FULLSYNC: 2242 ret = freeze_bdev(sb->s_bdev); 2243 if (ret) 2244 goto out; 2245 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2246 thaw_bdev(sb->s_bdev); 2247 break; 2248 case F2FS_GOING_DOWN_METASYNC: 2249 /* do checkpoint only */ 2250 ret = f2fs_sync_fs(sb, 1); 2251 if (ret) 2252 goto out; 2253 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2254 break; 2255 case F2FS_GOING_DOWN_NOSYNC: 2256 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2257 break; 2258 case F2FS_GOING_DOWN_METAFLUSH: 2259 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2260 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2261 break; 2262 case F2FS_GOING_DOWN_NEED_FSCK: 2263 set_sbi_flag(sbi, SBI_NEED_FSCK); 2264 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2265 set_sbi_flag(sbi, SBI_IS_DIRTY); 2266 /* do checkpoint only */ 2267 ret = f2fs_sync_fs(sb, 1); 2268 goto out; 2269 default: 2270 ret = -EINVAL; 2271 goto out; 2272 } 2273 2274 f2fs_stop_gc_thread(sbi); 2275 f2fs_stop_discard_thread(sbi); 2276 2277 f2fs_drop_discard_cmd(sbi); 2278 clear_opt(sbi, DISCARD); 2279 2280 f2fs_update_time(sbi, REQ_TIME); 2281 out: 2282 if (in != F2FS_GOING_DOWN_FULLSYNC) 2283 mnt_drop_write_file(filp); 2284 2285 trace_f2fs_shutdown(sbi, in, ret); 2286 2287 return ret; 2288 } 2289 2290 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2291 { 2292 struct inode *inode = file_inode(filp); 2293 struct super_block *sb = inode->i_sb; 2294 struct fstrim_range range; 2295 int ret; 2296 2297 if (!capable(CAP_SYS_ADMIN)) 2298 return -EPERM; 2299 2300 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2301 return -EOPNOTSUPP; 2302 2303 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2304 sizeof(range))) 2305 return -EFAULT; 2306 2307 ret = mnt_want_write_file(filp); 2308 if (ret) 2309 return ret; 2310 2311 range.minlen = max((unsigned int)range.minlen, 2312 bdev_discard_granularity(sb->s_bdev)); 2313 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2314 mnt_drop_write_file(filp); 2315 if (ret < 0) 2316 return ret; 2317 2318 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2319 sizeof(range))) 2320 return -EFAULT; 2321 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2322 return 0; 2323 } 2324 2325 static bool uuid_is_nonzero(__u8 u[16]) 2326 { 2327 int i; 2328 2329 for (i = 0; i < 16; i++) 2330 if (u[i]) 2331 return true; 2332 return false; 2333 } 2334 2335 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2336 { 2337 struct inode *inode = file_inode(filp); 2338 2339 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2340 return -EOPNOTSUPP; 2341 2342 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2343 2344 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2345 } 2346 2347 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2348 { 2349 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2350 return -EOPNOTSUPP; 2351 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2352 } 2353 2354 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2355 { 2356 struct inode *inode = file_inode(filp); 2357 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2358 u8 encrypt_pw_salt[16]; 2359 int err; 2360 2361 if (!f2fs_sb_has_encrypt(sbi)) 2362 return -EOPNOTSUPP; 2363 2364 err = mnt_want_write_file(filp); 2365 if (err) 2366 return err; 2367 2368 f2fs_down_write(&sbi->sb_lock); 2369 2370 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2371 goto got_it; 2372 2373 /* update superblock with uuid */ 2374 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2375 2376 err = f2fs_commit_super(sbi, false); 2377 if (err) { 2378 /* undo new data */ 2379 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2380 goto out_err; 2381 } 2382 got_it: 2383 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2384 out_err: 2385 f2fs_up_write(&sbi->sb_lock); 2386 mnt_drop_write_file(filp); 2387 2388 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2389 err = -EFAULT; 2390 2391 return err; 2392 } 2393 2394 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2395 unsigned long arg) 2396 { 2397 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2398 return -EOPNOTSUPP; 2399 2400 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2401 } 2402 2403 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2404 { 2405 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2406 return -EOPNOTSUPP; 2407 2408 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2409 } 2410 2411 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2412 { 2413 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2414 return -EOPNOTSUPP; 2415 2416 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2417 } 2418 2419 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2420 unsigned long arg) 2421 { 2422 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2423 return -EOPNOTSUPP; 2424 2425 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2426 } 2427 2428 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2429 unsigned long arg) 2430 { 2431 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2432 return -EOPNOTSUPP; 2433 2434 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2435 } 2436 2437 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2438 { 2439 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2440 return -EOPNOTSUPP; 2441 2442 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2443 } 2444 2445 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2446 { 2447 struct inode *inode = file_inode(filp); 2448 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2449 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2450 .no_bg_gc = false, 2451 .should_migrate_blocks = false, 2452 .nr_free_secs = 0 }; 2453 __u32 sync; 2454 int ret; 2455 2456 if (!capable(CAP_SYS_ADMIN)) 2457 return -EPERM; 2458 2459 if (get_user(sync, (__u32 __user *)arg)) 2460 return -EFAULT; 2461 2462 if (f2fs_readonly(sbi->sb)) 2463 return -EROFS; 2464 2465 ret = mnt_want_write_file(filp); 2466 if (ret) 2467 return ret; 2468 2469 if (!sync) { 2470 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2471 ret = -EBUSY; 2472 goto out; 2473 } 2474 } else { 2475 f2fs_down_write(&sbi->gc_lock); 2476 } 2477 2478 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2479 gc_control.err_gc_skipped = sync; 2480 stat_inc_gc_call_count(sbi, FOREGROUND); 2481 ret = f2fs_gc(sbi, &gc_control); 2482 out: 2483 mnt_drop_write_file(filp); 2484 return ret; 2485 } 2486 2487 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2488 { 2489 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2490 struct f2fs_gc_control gc_control = { 2491 .init_gc_type = range->sync ? FG_GC : BG_GC, 2492 .no_bg_gc = false, 2493 .should_migrate_blocks = false, 2494 .err_gc_skipped = range->sync, 2495 .nr_free_secs = 0 }; 2496 u64 end; 2497 int ret; 2498 2499 if (!capable(CAP_SYS_ADMIN)) 2500 return -EPERM; 2501 if (f2fs_readonly(sbi->sb)) 2502 return -EROFS; 2503 2504 end = range->start + range->len; 2505 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2506 end >= MAX_BLKADDR(sbi)) 2507 return -EINVAL; 2508 2509 ret = mnt_want_write_file(filp); 2510 if (ret) 2511 return ret; 2512 2513 do_more: 2514 if (!range->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.victim_segno = GET_SEGNO(sbi, range->start); 2524 stat_inc_gc_call_count(sbi, FOREGROUND); 2525 ret = f2fs_gc(sbi, &gc_control); 2526 if (ret) { 2527 if (ret == -EBUSY) 2528 ret = -EAGAIN; 2529 goto out; 2530 } 2531 range->start += CAP_BLKS_PER_SEC(sbi); 2532 if (range->start <= end) 2533 goto do_more; 2534 out: 2535 mnt_drop_write_file(filp); 2536 return ret; 2537 } 2538 2539 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2540 { 2541 struct f2fs_gc_range range; 2542 2543 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2544 sizeof(range))) 2545 return -EFAULT; 2546 return __f2fs_ioc_gc_range(filp, &range); 2547 } 2548 2549 static int f2fs_ioc_write_checkpoint(struct file *filp) 2550 { 2551 struct inode *inode = file_inode(filp); 2552 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2553 int ret; 2554 2555 if (!capable(CAP_SYS_ADMIN)) 2556 return -EPERM; 2557 2558 if (f2fs_readonly(sbi->sb)) 2559 return -EROFS; 2560 2561 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2562 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2563 return -EINVAL; 2564 } 2565 2566 ret = mnt_want_write_file(filp); 2567 if (ret) 2568 return ret; 2569 2570 ret = f2fs_sync_fs(sbi->sb, 1); 2571 2572 mnt_drop_write_file(filp); 2573 return ret; 2574 } 2575 2576 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2577 struct file *filp, 2578 struct f2fs_defragment *range) 2579 { 2580 struct inode *inode = file_inode(filp); 2581 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2582 .m_seg_type = NO_CHECK_TYPE, 2583 .m_may_create = false }; 2584 struct extent_info ei = {}; 2585 pgoff_t pg_start, pg_end, next_pgofs; 2586 unsigned int blk_per_seg = sbi->blocks_per_seg; 2587 unsigned int total = 0, sec_num; 2588 block_t blk_end = 0; 2589 bool fragmented = false; 2590 int err; 2591 2592 pg_start = range->start >> PAGE_SHIFT; 2593 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2594 2595 f2fs_balance_fs(sbi, true); 2596 2597 inode_lock(inode); 2598 2599 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2600 err = -EINVAL; 2601 goto unlock_out; 2602 } 2603 2604 /* if in-place-update policy is enabled, don't waste time here */ 2605 set_inode_flag(inode, FI_OPU_WRITE); 2606 if (f2fs_should_update_inplace(inode, NULL)) { 2607 err = -EINVAL; 2608 goto out; 2609 } 2610 2611 /* writeback all dirty pages in the range */ 2612 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2613 range->start + range->len - 1); 2614 if (err) 2615 goto out; 2616 2617 /* 2618 * lookup mapping info in extent cache, skip defragmenting if physical 2619 * block addresses are continuous. 2620 */ 2621 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2622 if (ei.fofs + ei.len >= pg_end) 2623 goto out; 2624 } 2625 2626 map.m_lblk = pg_start; 2627 map.m_next_pgofs = &next_pgofs; 2628 2629 /* 2630 * lookup mapping info in dnode page cache, skip defragmenting if all 2631 * physical block addresses are continuous even if there are hole(s) 2632 * in logical blocks. 2633 */ 2634 while (map.m_lblk < pg_end) { 2635 map.m_len = pg_end - map.m_lblk; 2636 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2637 if (err) 2638 goto out; 2639 2640 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2641 map.m_lblk = next_pgofs; 2642 continue; 2643 } 2644 2645 if (blk_end && blk_end != map.m_pblk) 2646 fragmented = true; 2647 2648 /* record total count of block that we're going to move */ 2649 total += map.m_len; 2650 2651 blk_end = map.m_pblk + map.m_len; 2652 2653 map.m_lblk += map.m_len; 2654 } 2655 2656 if (!fragmented) { 2657 total = 0; 2658 goto out; 2659 } 2660 2661 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2662 2663 /* 2664 * make sure there are enough free section for LFS allocation, this can 2665 * avoid defragment running in SSR mode when free section are allocated 2666 * intensively 2667 */ 2668 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2669 err = -EAGAIN; 2670 goto out; 2671 } 2672 2673 map.m_lblk = pg_start; 2674 map.m_len = pg_end - pg_start; 2675 total = 0; 2676 2677 while (map.m_lblk < pg_end) { 2678 pgoff_t idx; 2679 int cnt = 0; 2680 2681 do_map: 2682 map.m_len = pg_end - map.m_lblk; 2683 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2684 if (err) 2685 goto clear_out; 2686 2687 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2688 map.m_lblk = next_pgofs; 2689 goto check; 2690 } 2691 2692 set_inode_flag(inode, FI_SKIP_WRITES); 2693 2694 idx = map.m_lblk; 2695 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2696 struct page *page; 2697 2698 page = f2fs_get_lock_data_page(inode, idx, true); 2699 if (IS_ERR(page)) { 2700 err = PTR_ERR(page); 2701 goto clear_out; 2702 } 2703 2704 set_page_dirty(page); 2705 set_page_private_gcing(page); 2706 f2fs_put_page(page, 1); 2707 2708 idx++; 2709 cnt++; 2710 total++; 2711 } 2712 2713 map.m_lblk = idx; 2714 check: 2715 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2716 goto do_map; 2717 2718 clear_inode_flag(inode, FI_SKIP_WRITES); 2719 2720 err = filemap_fdatawrite(inode->i_mapping); 2721 if (err) 2722 goto out; 2723 } 2724 clear_out: 2725 clear_inode_flag(inode, FI_SKIP_WRITES); 2726 out: 2727 clear_inode_flag(inode, FI_OPU_WRITE); 2728 unlock_out: 2729 inode_unlock(inode); 2730 if (!err) 2731 range->len = (u64)total << PAGE_SHIFT; 2732 return err; 2733 } 2734 2735 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2736 { 2737 struct inode *inode = file_inode(filp); 2738 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2739 struct f2fs_defragment range; 2740 int err; 2741 2742 if (!capable(CAP_SYS_ADMIN)) 2743 return -EPERM; 2744 2745 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2746 return -EINVAL; 2747 2748 if (f2fs_readonly(sbi->sb)) 2749 return -EROFS; 2750 2751 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2752 sizeof(range))) 2753 return -EFAULT; 2754 2755 /* verify alignment of offset & size */ 2756 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2757 return -EINVAL; 2758 2759 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2760 max_file_blocks(inode))) 2761 return -EINVAL; 2762 2763 err = mnt_want_write_file(filp); 2764 if (err) 2765 return err; 2766 2767 err = f2fs_defragment_range(sbi, filp, &range); 2768 mnt_drop_write_file(filp); 2769 2770 f2fs_update_time(sbi, REQ_TIME); 2771 if (err < 0) 2772 return err; 2773 2774 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2775 sizeof(range))) 2776 return -EFAULT; 2777 2778 return 0; 2779 } 2780 2781 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2782 struct file *file_out, loff_t pos_out, size_t len) 2783 { 2784 struct inode *src = file_inode(file_in); 2785 struct inode *dst = file_inode(file_out); 2786 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2787 size_t olen = len, dst_max_i_size = 0; 2788 size_t dst_osize; 2789 int ret; 2790 2791 if (file_in->f_path.mnt != file_out->f_path.mnt || 2792 src->i_sb != dst->i_sb) 2793 return -EXDEV; 2794 2795 if (unlikely(f2fs_readonly(src->i_sb))) 2796 return -EROFS; 2797 2798 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2799 return -EINVAL; 2800 2801 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2802 return -EOPNOTSUPP; 2803 2804 if (pos_out < 0 || pos_in < 0) 2805 return -EINVAL; 2806 2807 if (src == dst) { 2808 if (pos_in == pos_out) 2809 return 0; 2810 if (pos_out > pos_in && pos_out < pos_in + len) 2811 return -EINVAL; 2812 } 2813 2814 inode_lock(src); 2815 if (src != dst) { 2816 ret = -EBUSY; 2817 if (!inode_trylock(dst)) 2818 goto out; 2819 } 2820 2821 ret = -EINVAL; 2822 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2823 goto out_unlock; 2824 if (len == 0) 2825 olen = len = src->i_size - pos_in; 2826 if (pos_in + len == src->i_size) 2827 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2828 if (len == 0) { 2829 ret = 0; 2830 goto out_unlock; 2831 } 2832 2833 dst_osize = dst->i_size; 2834 if (pos_out + olen > dst->i_size) 2835 dst_max_i_size = pos_out + olen; 2836 2837 /* verify the end result is block aligned */ 2838 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2839 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2840 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2841 goto out_unlock; 2842 2843 ret = f2fs_convert_inline_inode(src); 2844 if (ret) 2845 goto out_unlock; 2846 2847 ret = f2fs_convert_inline_inode(dst); 2848 if (ret) 2849 goto out_unlock; 2850 2851 /* write out all dirty pages from offset */ 2852 ret = filemap_write_and_wait_range(src->i_mapping, 2853 pos_in, pos_in + len); 2854 if (ret) 2855 goto out_unlock; 2856 2857 ret = filemap_write_and_wait_range(dst->i_mapping, 2858 pos_out, pos_out + len); 2859 if (ret) 2860 goto out_unlock; 2861 2862 f2fs_balance_fs(sbi, true); 2863 2864 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2865 if (src != dst) { 2866 ret = -EBUSY; 2867 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2868 goto out_src; 2869 } 2870 2871 f2fs_lock_op(sbi); 2872 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2873 pos_out >> F2FS_BLKSIZE_BITS, 2874 len >> F2FS_BLKSIZE_BITS, false); 2875 2876 if (!ret) { 2877 if (dst_max_i_size) 2878 f2fs_i_size_write(dst, dst_max_i_size); 2879 else if (dst_osize != dst->i_size) 2880 f2fs_i_size_write(dst, dst_osize); 2881 } 2882 f2fs_unlock_op(sbi); 2883 2884 if (src != dst) 2885 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2886 out_src: 2887 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2888 if (ret) 2889 goto out_unlock; 2890 2891 src->i_mtime = inode_set_ctime_current(src); 2892 f2fs_mark_inode_dirty_sync(src, false); 2893 if (src != dst) { 2894 dst->i_mtime = inode_set_ctime_current(dst); 2895 f2fs_mark_inode_dirty_sync(dst, false); 2896 } 2897 f2fs_update_time(sbi, REQ_TIME); 2898 2899 out_unlock: 2900 if (src != dst) 2901 inode_unlock(dst); 2902 out: 2903 inode_unlock(src); 2904 return ret; 2905 } 2906 2907 static int __f2fs_ioc_move_range(struct file *filp, 2908 struct f2fs_move_range *range) 2909 { 2910 struct fd dst; 2911 int err; 2912 2913 if (!(filp->f_mode & FMODE_READ) || 2914 !(filp->f_mode & FMODE_WRITE)) 2915 return -EBADF; 2916 2917 dst = fdget(range->dst_fd); 2918 if (!dst.file) 2919 return -EBADF; 2920 2921 if (!(dst.file->f_mode & FMODE_WRITE)) { 2922 err = -EBADF; 2923 goto err_out; 2924 } 2925 2926 err = mnt_want_write_file(filp); 2927 if (err) 2928 goto err_out; 2929 2930 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2931 range->pos_out, range->len); 2932 2933 mnt_drop_write_file(filp); 2934 err_out: 2935 fdput(dst); 2936 return err; 2937 } 2938 2939 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2940 { 2941 struct f2fs_move_range range; 2942 2943 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2944 sizeof(range))) 2945 return -EFAULT; 2946 return __f2fs_ioc_move_range(filp, &range); 2947 } 2948 2949 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2950 { 2951 struct inode *inode = file_inode(filp); 2952 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2953 struct sit_info *sm = SIT_I(sbi); 2954 unsigned int start_segno = 0, end_segno = 0; 2955 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2956 struct f2fs_flush_device range; 2957 struct f2fs_gc_control gc_control = { 2958 .init_gc_type = FG_GC, 2959 .should_migrate_blocks = true, 2960 .err_gc_skipped = true, 2961 .nr_free_secs = 0 }; 2962 int ret; 2963 2964 if (!capable(CAP_SYS_ADMIN)) 2965 return -EPERM; 2966 2967 if (f2fs_readonly(sbi->sb)) 2968 return -EROFS; 2969 2970 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2971 return -EINVAL; 2972 2973 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2974 sizeof(range))) 2975 return -EFAULT; 2976 2977 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2978 __is_large_section(sbi)) { 2979 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2980 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2981 return -EINVAL; 2982 } 2983 2984 ret = mnt_want_write_file(filp); 2985 if (ret) 2986 return ret; 2987 2988 if (range.dev_num != 0) 2989 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2990 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2991 2992 start_segno = sm->last_victim[FLUSH_DEVICE]; 2993 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2994 start_segno = dev_start_segno; 2995 end_segno = min(start_segno + range.segments, dev_end_segno); 2996 2997 while (start_segno < end_segno) { 2998 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2999 ret = -EBUSY; 3000 goto out; 3001 } 3002 sm->last_victim[GC_CB] = end_segno + 1; 3003 sm->last_victim[GC_GREEDY] = end_segno + 1; 3004 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3005 3006 gc_control.victim_segno = start_segno; 3007 stat_inc_gc_call_count(sbi, FOREGROUND); 3008 ret = f2fs_gc(sbi, &gc_control); 3009 if (ret == -EAGAIN) 3010 ret = 0; 3011 else if (ret < 0) 3012 break; 3013 start_segno++; 3014 } 3015 out: 3016 mnt_drop_write_file(filp); 3017 return ret; 3018 } 3019 3020 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3021 { 3022 struct inode *inode = file_inode(filp); 3023 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3024 3025 /* Must validate to set it with SQLite behavior in Android. */ 3026 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3027 3028 return put_user(sb_feature, (u32 __user *)arg); 3029 } 3030 3031 #ifdef CONFIG_QUOTA 3032 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3033 { 3034 struct dquot *transfer_to[MAXQUOTAS] = {}; 3035 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3036 struct super_block *sb = sbi->sb; 3037 int err; 3038 3039 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3040 if (IS_ERR(transfer_to[PRJQUOTA])) 3041 return PTR_ERR(transfer_to[PRJQUOTA]); 3042 3043 err = __dquot_transfer(inode, transfer_to); 3044 if (err) 3045 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3046 dqput(transfer_to[PRJQUOTA]); 3047 return err; 3048 } 3049 3050 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3051 { 3052 struct f2fs_inode_info *fi = F2FS_I(inode); 3053 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3054 struct f2fs_inode *ri = NULL; 3055 kprojid_t kprojid; 3056 int err; 3057 3058 if (!f2fs_sb_has_project_quota(sbi)) { 3059 if (projid != F2FS_DEF_PROJID) 3060 return -EOPNOTSUPP; 3061 else 3062 return 0; 3063 } 3064 3065 if (!f2fs_has_extra_attr(inode)) 3066 return -EOPNOTSUPP; 3067 3068 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3069 3070 if (projid_eq(kprojid, fi->i_projid)) 3071 return 0; 3072 3073 err = -EPERM; 3074 /* Is it quota file? Do not allow user to mess with it */ 3075 if (IS_NOQUOTA(inode)) 3076 return err; 3077 3078 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3079 return -EOVERFLOW; 3080 3081 err = f2fs_dquot_initialize(inode); 3082 if (err) 3083 return err; 3084 3085 f2fs_lock_op(sbi); 3086 err = f2fs_transfer_project_quota(inode, kprojid); 3087 if (err) 3088 goto out_unlock; 3089 3090 fi->i_projid = kprojid; 3091 inode_set_ctime_current(inode); 3092 f2fs_mark_inode_dirty_sync(inode, true); 3093 out_unlock: 3094 f2fs_unlock_op(sbi); 3095 return err; 3096 } 3097 #else 3098 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3099 { 3100 return 0; 3101 } 3102 3103 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3104 { 3105 if (projid != F2FS_DEF_PROJID) 3106 return -EOPNOTSUPP; 3107 return 0; 3108 } 3109 #endif 3110 3111 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3112 { 3113 struct inode *inode = d_inode(dentry); 3114 struct f2fs_inode_info *fi = F2FS_I(inode); 3115 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3116 3117 if (IS_ENCRYPTED(inode)) 3118 fsflags |= FS_ENCRYPT_FL; 3119 if (IS_VERITY(inode)) 3120 fsflags |= FS_VERITY_FL; 3121 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3122 fsflags |= FS_INLINE_DATA_FL; 3123 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3124 fsflags |= FS_NOCOW_FL; 3125 3126 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3127 3128 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3129 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3130 3131 return 0; 3132 } 3133 3134 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3135 struct dentry *dentry, struct fileattr *fa) 3136 { 3137 struct inode *inode = d_inode(dentry); 3138 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3139 u32 iflags; 3140 int err; 3141 3142 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3143 return -EIO; 3144 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3145 return -ENOSPC; 3146 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3147 return -EOPNOTSUPP; 3148 fsflags &= F2FS_SETTABLE_FS_FL; 3149 if (!fa->flags_valid) 3150 mask &= FS_COMMON_FL; 3151 3152 iflags = f2fs_fsflags_to_iflags(fsflags); 3153 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3154 return -EOPNOTSUPP; 3155 3156 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3157 if (!err) 3158 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3159 3160 return err; 3161 } 3162 3163 int f2fs_pin_file_control(struct inode *inode, bool inc) 3164 { 3165 struct f2fs_inode_info *fi = F2FS_I(inode); 3166 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3167 3168 /* Use i_gc_failures for normal file as a risk signal. */ 3169 if (inc) 3170 f2fs_i_gc_failures_write(inode, 3171 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3172 3173 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3174 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3175 __func__, inode->i_ino, 3176 fi->i_gc_failures[GC_FAILURE_PIN]); 3177 clear_inode_flag(inode, FI_PIN_FILE); 3178 return -EAGAIN; 3179 } 3180 return 0; 3181 } 3182 3183 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3184 { 3185 struct inode *inode = file_inode(filp); 3186 __u32 pin; 3187 int ret = 0; 3188 3189 if (get_user(pin, (__u32 __user *)arg)) 3190 return -EFAULT; 3191 3192 if (!S_ISREG(inode->i_mode)) 3193 return -EINVAL; 3194 3195 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3196 return -EROFS; 3197 3198 ret = mnt_want_write_file(filp); 3199 if (ret) 3200 return ret; 3201 3202 inode_lock(inode); 3203 3204 if (!pin) { 3205 clear_inode_flag(inode, FI_PIN_FILE); 3206 f2fs_i_gc_failures_write(inode, 0); 3207 goto done; 3208 } 3209 3210 if (f2fs_should_update_outplace(inode, NULL)) { 3211 ret = -EINVAL; 3212 goto out; 3213 } 3214 3215 if (f2fs_pin_file_control(inode, false)) { 3216 ret = -EAGAIN; 3217 goto out; 3218 } 3219 3220 ret = f2fs_convert_inline_inode(inode); 3221 if (ret) 3222 goto out; 3223 3224 if (!f2fs_disable_compressed_file(inode)) { 3225 ret = -EOPNOTSUPP; 3226 goto out; 3227 } 3228 3229 set_inode_flag(inode, FI_PIN_FILE); 3230 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3231 done: 3232 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3233 out: 3234 inode_unlock(inode); 3235 mnt_drop_write_file(filp); 3236 return ret; 3237 } 3238 3239 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3240 { 3241 struct inode *inode = file_inode(filp); 3242 __u32 pin = 0; 3243 3244 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3245 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3246 return put_user(pin, (u32 __user *)arg); 3247 } 3248 3249 int f2fs_precache_extents(struct inode *inode) 3250 { 3251 struct f2fs_inode_info *fi = F2FS_I(inode); 3252 struct f2fs_map_blocks map; 3253 pgoff_t m_next_extent; 3254 loff_t end; 3255 int err; 3256 3257 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3258 return -EOPNOTSUPP; 3259 3260 map.m_lblk = 0; 3261 map.m_next_pgofs = NULL; 3262 map.m_next_extent = &m_next_extent; 3263 map.m_seg_type = NO_CHECK_TYPE; 3264 map.m_may_create = false; 3265 end = max_file_blocks(inode); 3266 3267 while (map.m_lblk < end) { 3268 map.m_len = end - map.m_lblk; 3269 3270 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3271 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3272 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3273 if (err) 3274 return err; 3275 3276 map.m_lblk = m_next_extent; 3277 } 3278 3279 return 0; 3280 } 3281 3282 static int f2fs_ioc_precache_extents(struct file *filp) 3283 { 3284 return f2fs_precache_extents(file_inode(filp)); 3285 } 3286 3287 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3288 { 3289 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3290 __u64 block_count; 3291 3292 if (!capable(CAP_SYS_ADMIN)) 3293 return -EPERM; 3294 3295 if (f2fs_readonly(sbi->sb)) 3296 return -EROFS; 3297 3298 if (copy_from_user(&block_count, (void __user *)arg, 3299 sizeof(block_count))) 3300 return -EFAULT; 3301 3302 return f2fs_resize_fs(filp, block_count); 3303 } 3304 3305 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3306 { 3307 struct inode *inode = file_inode(filp); 3308 3309 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3310 3311 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3312 f2fs_warn(F2FS_I_SB(inode), 3313 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3314 inode->i_ino); 3315 return -EOPNOTSUPP; 3316 } 3317 3318 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3319 } 3320 3321 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3322 { 3323 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3324 return -EOPNOTSUPP; 3325 3326 return fsverity_ioctl_measure(filp, (void __user *)arg); 3327 } 3328 3329 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3330 { 3331 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3332 return -EOPNOTSUPP; 3333 3334 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3335 } 3336 3337 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3338 { 3339 struct inode *inode = file_inode(filp); 3340 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3341 char *vbuf; 3342 int count; 3343 int err = 0; 3344 3345 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3346 if (!vbuf) 3347 return -ENOMEM; 3348 3349 f2fs_down_read(&sbi->sb_lock); 3350 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3351 ARRAY_SIZE(sbi->raw_super->volume_name), 3352 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3353 f2fs_up_read(&sbi->sb_lock); 3354 3355 if (copy_to_user((char __user *)arg, vbuf, 3356 min(FSLABEL_MAX, count))) 3357 err = -EFAULT; 3358 3359 kfree(vbuf); 3360 return err; 3361 } 3362 3363 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3364 { 3365 struct inode *inode = file_inode(filp); 3366 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3367 char *vbuf; 3368 int err = 0; 3369 3370 if (!capable(CAP_SYS_ADMIN)) 3371 return -EPERM; 3372 3373 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3374 if (IS_ERR(vbuf)) 3375 return PTR_ERR(vbuf); 3376 3377 err = mnt_want_write_file(filp); 3378 if (err) 3379 goto out; 3380 3381 f2fs_down_write(&sbi->sb_lock); 3382 3383 memset(sbi->raw_super->volume_name, 0, 3384 sizeof(sbi->raw_super->volume_name)); 3385 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3386 sbi->raw_super->volume_name, 3387 ARRAY_SIZE(sbi->raw_super->volume_name)); 3388 3389 err = f2fs_commit_super(sbi, false); 3390 3391 f2fs_up_write(&sbi->sb_lock); 3392 3393 mnt_drop_write_file(filp); 3394 out: 3395 kfree(vbuf); 3396 return err; 3397 } 3398 3399 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3400 { 3401 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3402 return -EOPNOTSUPP; 3403 3404 if (!f2fs_compressed_file(inode)) 3405 return -EINVAL; 3406 3407 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3408 3409 return 0; 3410 } 3411 3412 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3413 { 3414 struct inode *inode = file_inode(filp); 3415 __u64 blocks; 3416 int ret; 3417 3418 ret = f2fs_get_compress_blocks(inode, &blocks); 3419 if (ret < 0) 3420 return ret; 3421 3422 return put_user(blocks, (u64 __user *)arg); 3423 } 3424 3425 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3426 { 3427 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3428 unsigned int released_blocks = 0; 3429 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3430 block_t blkaddr; 3431 int i; 3432 3433 for (i = 0; i < count; i++) { 3434 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3435 dn->ofs_in_node + i); 3436 3437 if (!__is_valid_data_blkaddr(blkaddr)) 3438 continue; 3439 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3440 DATA_GENERIC_ENHANCE))) { 3441 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3442 return -EFSCORRUPTED; 3443 } 3444 } 3445 3446 while (count) { 3447 int compr_blocks = 0; 3448 3449 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3450 blkaddr = f2fs_data_blkaddr(dn); 3451 3452 if (i == 0) { 3453 if (blkaddr == COMPRESS_ADDR) 3454 continue; 3455 dn->ofs_in_node += cluster_size; 3456 goto next; 3457 } 3458 3459 if (__is_valid_data_blkaddr(blkaddr)) 3460 compr_blocks++; 3461 3462 if (blkaddr != NEW_ADDR) 3463 continue; 3464 3465 dn->data_blkaddr = NULL_ADDR; 3466 f2fs_set_data_blkaddr(dn); 3467 } 3468 3469 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3470 dec_valid_block_count(sbi, dn->inode, 3471 cluster_size - compr_blocks); 3472 3473 released_blocks += cluster_size - compr_blocks; 3474 next: 3475 count -= cluster_size; 3476 } 3477 3478 return released_blocks; 3479 } 3480 3481 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3482 { 3483 struct inode *inode = file_inode(filp); 3484 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3485 pgoff_t page_idx = 0, last_idx; 3486 unsigned int released_blocks = 0; 3487 int ret; 3488 int writecount; 3489 3490 if (!f2fs_sb_has_compression(sbi)) 3491 return -EOPNOTSUPP; 3492 3493 if (!f2fs_compressed_file(inode)) 3494 return -EINVAL; 3495 3496 if (f2fs_readonly(sbi->sb)) 3497 return -EROFS; 3498 3499 ret = mnt_want_write_file(filp); 3500 if (ret) 3501 return ret; 3502 3503 f2fs_balance_fs(sbi, true); 3504 3505 inode_lock(inode); 3506 3507 writecount = atomic_read(&inode->i_writecount); 3508 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3509 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3510 ret = -EBUSY; 3511 goto out; 3512 } 3513 3514 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3515 ret = -EINVAL; 3516 goto out; 3517 } 3518 3519 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3520 if (ret) 3521 goto out; 3522 3523 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3524 ret = -EPERM; 3525 goto out; 3526 } 3527 3528 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3529 inode_set_ctime_current(inode); 3530 f2fs_mark_inode_dirty_sync(inode, true); 3531 3532 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3533 filemap_invalidate_lock(inode->i_mapping); 3534 3535 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3536 3537 while (page_idx < last_idx) { 3538 struct dnode_of_data dn; 3539 pgoff_t end_offset, count; 3540 3541 set_new_dnode(&dn, inode, NULL, NULL, 0); 3542 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3543 if (ret) { 3544 if (ret == -ENOENT) { 3545 page_idx = f2fs_get_next_page_offset(&dn, 3546 page_idx); 3547 ret = 0; 3548 continue; 3549 } 3550 break; 3551 } 3552 3553 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3554 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3555 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3556 3557 ret = release_compress_blocks(&dn, count); 3558 3559 f2fs_put_dnode(&dn); 3560 3561 if (ret < 0) 3562 break; 3563 3564 page_idx += count; 3565 released_blocks += ret; 3566 } 3567 3568 filemap_invalidate_unlock(inode->i_mapping); 3569 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3570 out: 3571 inode_unlock(inode); 3572 3573 mnt_drop_write_file(filp); 3574 3575 if (ret >= 0) { 3576 ret = put_user(released_blocks, (u64 __user *)arg); 3577 } else if (released_blocks && 3578 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3579 set_sbi_flag(sbi, SBI_NEED_FSCK); 3580 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3581 "iblocks=%llu, released=%u, compr_blocks=%u, " 3582 "run fsck to fix.", 3583 __func__, inode->i_ino, inode->i_blocks, 3584 released_blocks, 3585 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3586 } 3587 3588 return ret; 3589 } 3590 3591 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3592 { 3593 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3594 unsigned int reserved_blocks = 0; 3595 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3596 block_t blkaddr; 3597 int i; 3598 3599 for (i = 0; i < count; i++) { 3600 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3601 dn->ofs_in_node + i); 3602 3603 if (!__is_valid_data_blkaddr(blkaddr)) 3604 continue; 3605 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3606 DATA_GENERIC_ENHANCE))) { 3607 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3608 return -EFSCORRUPTED; 3609 } 3610 } 3611 3612 while (count) { 3613 int compr_blocks = 0; 3614 blkcnt_t reserved; 3615 int ret; 3616 3617 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3618 blkaddr = f2fs_data_blkaddr(dn); 3619 3620 if (i == 0) { 3621 if (blkaddr == COMPRESS_ADDR) 3622 continue; 3623 dn->ofs_in_node += cluster_size; 3624 goto next; 3625 } 3626 3627 if (__is_valid_data_blkaddr(blkaddr)) { 3628 compr_blocks++; 3629 continue; 3630 } 3631 3632 dn->data_blkaddr = NEW_ADDR; 3633 f2fs_set_data_blkaddr(dn); 3634 } 3635 3636 reserved = cluster_size - compr_blocks; 3637 ret = inc_valid_block_count(sbi, dn->inode, &reserved); 3638 if (ret) 3639 return ret; 3640 3641 if (reserved != cluster_size - compr_blocks) 3642 return -ENOSPC; 3643 3644 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3645 3646 reserved_blocks += reserved; 3647 next: 3648 count -= cluster_size; 3649 } 3650 3651 return reserved_blocks; 3652 } 3653 3654 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3655 { 3656 struct inode *inode = file_inode(filp); 3657 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3658 pgoff_t page_idx = 0, last_idx; 3659 unsigned int reserved_blocks = 0; 3660 int ret; 3661 3662 if (!f2fs_sb_has_compression(sbi)) 3663 return -EOPNOTSUPP; 3664 3665 if (!f2fs_compressed_file(inode)) 3666 return -EINVAL; 3667 3668 if (f2fs_readonly(sbi->sb)) 3669 return -EROFS; 3670 3671 ret = mnt_want_write_file(filp); 3672 if (ret) 3673 return ret; 3674 3675 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3676 goto out; 3677 3678 f2fs_balance_fs(sbi, true); 3679 3680 inode_lock(inode); 3681 3682 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3683 ret = -EINVAL; 3684 goto unlock_inode; 3685 } 3686 3687 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3688 filemap_invalidate_lock(inode->i_mapping); 3689 3690 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3691 3692 while (page_idx < last_idx) { 3693 struct dnode_of_data dn; 3694 pgoff_t end_offset, count; 3695 3696 set_new_dnode(&dn, inode, NULL, NULL, 0); 3697 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3698 if (ret) { 3699 if (ret == -ENOENT) { 3700 page_idx = f2fs_get_next_page_offset(&dn, 3701 page_idx); 3702 ret = 0; 3703 continue; 3704 } 3705 break; 3706 } 3707 3708 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3709 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3710 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3711 3712 ret = reserve_compress_blocks(&dn, count); 3713 3714 f2fs_put_dnode(&dn); 3715 3716 if (ret < 0) 3717 break; 3718 3719 page_idx += count; 3720 reserved_blocks += ret; 3721 } 3722 3723 filemap_invalidate_unlock(inode->i_mapping); 3724 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3725 3726 if (ret >= 0) { 3727 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3728 inode_set_ctime_current(inode); 3729 f2fs_mark_inode_dirty_sync(inode, true); 3730 } 3731 unlock_inode: 3732 inode_unlock(inode); 3733 out: 3734 mnt_drop_write_file(filp); 3735 3736 if (ret >= 0) { 3737 ret = put_user(reserved_blocks, (u64 __user *)arg); 3738 } else if (reserved_blocks && 3739 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3740 set_sbi_flag(sbi, SBI_NEED_FSCK); 3741 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3742 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3743 "run fsck to fix.", 3744 __func__, inode->i_ino, inode->i_blocks, 3745 reserved_blocks, 3746 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3747 } 3748 3749 return ret; 3750 } 3751 3752 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3753 pgoff_t off, block_t block, block_t len, u32 flags) 3754 { 3755 sector_t sector = SECTOR_FROM_BLOCK(block); 3756 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3757 int ret = 0; 3758 3759 if (flags & F2FS_TRIM_FILE_DISCARD) { 3760 if (bdev_max_secure_erase_sectors(bdev)) 3761 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3762 GFP_NOFS); 3763 else 3764 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3765 GFP_NOFS); 3766 } 3767 3768 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3769 if (IS_ENCRYPTED(inode)) 3770 ret = fscrypt_zeroout_range(inode, off, block, len); 3771 else 3772 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3773 GFP_NOFS, 0); 3774 } 3775 3776 return ret; 3777 } 3778 3779 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3780 { 3781 struct inode *inode = file_inode(filp); 3782 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3783 struct address_space *mapping = inode->i_mapping; 3784 struct block_device *prev_bdev = NULL; 3785 struct f2fs_sectrim_range range; 3786 pgoff_t index, pg_end, prev_index = 0; 3787 block_t prev_block = 0, len = 0; 3788 loff_t end_addr; 3789 bool to_end = false; 3790 int ret = 0; 3791 3792 if (!(filp->f_mode & FMODE_WRITE)) 3793 return -EBADF; 3794 3795 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3796 sizeof(range))) 3797 return -EFAULT; 3798 3799 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3800 !S_ISREG(inode->i_mode)) 3801 return -EINVAL; 3802 3803 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3804 !f2fs_hw_support_discard(sbi)) || 3805 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3806 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3807 return -EOPNOTSUPP; 3808 3809 file_start_write(filp); 3810 inode_lock(inode); 3811 3812 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3813 range.start >= inode->i_size) { 3814 ret = -EINVAL; 3815 goto err; 3816 } 3817 3818 if (range.len == 0) 3819 goto err; 3820 3821 if (inode->i_size - range.start > range.len) { 3822 end_addr = range.start + range.len; 3823 } else { 3824 end_addr = range.len == (u64)-1 ? 3825 sbi->sb->s_maxbytes : inode->i_size; 3826 to_end = true; 3827 } 3828 3829 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3830 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3831 ret = -EINVAL; 3832 goto err; 3833 } 3834 3835 index = F2FS_BYTES_TO_BLK(range.start); 3836 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3837 3838 ret = f2fs_convert_inline_inode(inode); 3839 if (ret) 3840 goto err; 3841 3842 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3843 filemap_invalidate_lock(mapping); 3844 3845 ret = filemap_write_and_wait_range(mapping, range.start, 3846 to_end ? LLONG_MAX : end_addr - 1); 3847 if (ret) 3848 goto out; 3849 3850 truncate_inode_pages_range(mapping, range.start, 3851 to_end ? -1 : end_addr - 1); 3852 3853 while (index < pg_end) { 3854 struct dnode_of_data dn; 3855 pgoff_t end_offset, count; 3856 int i; 3857 3858 set_new_dnode(&dn, inode, NULL, NULL, 0); 3859 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3860 if (ret) { 3861 if (ret == -ENOENT) { 3862 index = f2fs_get_next_page_offset(&dn, index); 3863 continue; 3864 } 3865 goto out; 3866 } 3867 3868 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3869 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3870 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3871 struct block_device *cur_bdev; 3872 block_t blkaddr = f2fs_data_blkaddr(&dn); 3873 3874 if (!__is_valid_data_blkaddr(blkaddr)) 3875 continue; 3876 3877 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3878 DATA_GENERIC_ENHANCE)) { 3879 ret = -EFSCORRUPTED; 3880 f2fs_put_dnode(&dn); 3881 f2fs_handle_error(sbi, 3882 ERROR_INVALID_BLKADDR); 3883 goto out; 3884 } 3885 3886 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3887 if (f2fs_is_multi_device(sbi)) { 3888 int di = f2fs_target_device_index(sbi, blkaddr); 3889 3890 blkaddr -= FDEV(di).start_blk; 3891 } 3892 3893 if (len) { 3894 if (prev_bdev == cur_bdev && 3895 index == prev_index + len && 3896 blkaddr == prev_block + len) { 3897 len++; 3898 } else { 3899 ret = f2fs_secure_erase(prev_bdev, 3900 inode, prev_index, prev_block, 3901 len, range.flags); 3902 if (ret) { 3903 f2fs_put_dnode(&dn); 3904 goto out; 3905 } 3906 3907 len = 0; 3908 } 3909 } 3910 3911 if (!len) { 3912 prev_bdev = cur_bdev; 3913 prev_index = index; 3914 prev_block = blkaddr; 3915 len = 1; 3916 } 3917 } 3918 3919 f2fs_put_dnode(&dn); 3920 3921 if (fatal_signal_pending(current)) { 3922 ret = -EINTR; 3923 goto out; 3924 } 3925 cond_resched(); 3926 } 3927 3928 if (len) 3929 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3930 prev_block, len, range.flags); 3931 out: 3932 filemap_invalidate_unlock(mapping); 3933 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3934 err: 3935 inode_unlock(inode); 3936 file_end_write(filp); 3937 3938 return ret; 3939 } 3940 3941 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3942 { 3943 struct inode *inode = file_inode(filp); 3944 struct f2fs_comp_option option; 3945 3946 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3947 return -EOPNOTSUPP; 3948 3949 inode_lock_shared(inode); 3950 3951 if (!f2fs_compressed_file(inode)) { 3952 inode_unlock_shared(inode); 3953 return -ENODATA; 3954 } 3955 3956 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3957 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3958 3959 inode_unlock_shared(inode); 3960 3961 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3962 sizeof(option))) 3963 return -EFAULT; 3964 3965 return 0; 3966 } 3967 3968 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3969 { 3970 struct inode *inode = file_inode(filp); 3971 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3972 struct f2fs_comp_option option; 3973 int ret = 0; 3974 3975 if (!f2fs_sb_has_compression(sbi)) 3976 return -EOPNOTSUPP; 3977 3978 if (!(filp->f_mode & FMODE_WRITE)) 3979 return -EBADF; 3980 3981 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3982 sizeof(option))) 3983 return -EFAULT; 3984 3985 if (!f2fs_compressed_file(inode) || 3986 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3987 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3988 option.algorithm >= COMPRESS_MAX) 3989 return -EINVAL; 3990 3991 file_start_write(filp); 3992 inode_lock(inode); 3993 3994 f2fs_down_write(&F2FS_I(inode)->i_sem); 3995 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 3996 ret = -EBUSY; 3997 goto out; 3998 } 3999 4000 if (F2FS_HAS_BLOCKS(inode)) { 4001 ret = -EFBIG; 4002 goto out; 4003 } 4004 4005 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4006 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4007 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4008 f2fs_mark_inode_dirty_sync(inode, true); 4009 4010 if (!f2fs_is_compress_backend_ready(inode)) 4011 f2fs_warn(sbi, "compression algorithm is successfully set, " 4012 "but current kernel doesn't support this algorithm."); 4013 out: 4014 f2fs_up_write(&F2FS_I(inode)->i_sem); 4015 inode_unlock(inode); 4016 file_end_write(filp); 4017 4018 return ret; 4019 } 4020 4021 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4022 { 4023 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4024 struct address_space *mapping = inode->i_mapping; 4025 struct page *page; 4026 pgoff_t redirty_idx = page_idx; 4027 int i, page_len = 0, ret = 0; 4028 4029 page_cache_ra_unbounded(&ractl, len, 0); 4030 4031 for (i = 0; i < len; i++, page_idx++) { 4032 page = read_cache_page(mapping, page_idx, NULL, NULL); 4033 if (IS_ERR(page)) { 4034 ret = PTR_ERR(page); 4035 break; 4036 } 4037 page_len++; 4038 } 4039 4040 for (i = 0; i < page_len; i++, redirty_idx++) { 4041 page = find_lock_page(mapping, redirty_idx); 4042 4043 /* It will never fail, when page has pinned above */ 4044 f2fs_bug_on(F2FS_I_SB(inode), !page); 4045 4046 set_page_dirty(page); 4047 f2fs_put_page(page, 1); 4048 f2fs_put_page(page, 0); 4049 } 4050 4051 return ret; 4052 } 4053 4054 static int f2fs_ioc_decompress_file(struct file *filp) 4055 { 4056 struct inode *inode = file_inode(filp); 4057 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4058 struct f2fs_inode_info *fi = F2FS_I(inode); 4059 pgoff_t page_idx = 0, last_idx; 4060 unsigned int blk_per_seg = sbi->blocks_per_seg; 4061 int cluster_size = fi->i_cluster_size; 4062 int count, ret; 4063 4064 if (!f2fs_sb_has_compression(sbi) || 4065 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4066 return -EOPNOTSUPP; 4067 4068 if (!(filp->f_mode & FMODE_WRITE)) 4069 return -EBADF; 4070 4071 if (!f2fs_compressed_file(inode)) 4072 return -EINVAL; 4073 4074 f2fs_balance_fs(sbi, true); 4075 4076 file_start_write(filp); 4077 inode_lock(inode); 4078 4079 if (!f2fs_is_compress_backend_ready(inode)) { 4080 ret = -EOPNOTSUPP; 4081 goto out; 4082 } 4083 4084 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4085 ret = -EINVAL; 4086 goto out; 4087 } 4088 4089 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4090 if (ret) 4091 goto out; 4092 4093 if (!atomic_read(&fi->i_compr_blocks)) 4094 goto out; 4095 4096 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4097 4098 count = last_idx - page_idx; 4099 while (count && count >= cluster_size) { 4100 ret = redirty_blocks(inode, page_idx, cluster_size); 4101 if (ret < 0) 4102 break; 4103 4104 if (get_dirty_pages(inode) >= blk_per_seg) { 4105 ret = filemap_fdatawrite(inode->i_mapping); 4106 if (ret < 0) 4107 break; 4108 } 4109 4110 count -= cluster_size; 4111 page_idx += cluster_size; 4112 4113 cond_resched(); 4114 if (fatal_signal_pending(current)) { 4115 ret = -EINTR; 4116 break; 4117 } 4118 } 4119 4120 if (!ret) 4121 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4122 LLONG_MAX); 4123 4124 if (ret) 4125 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4126 __func__, ret); 4127 out: 4128 inode_unlock(inode); 4129 file_end_write(filp); 4130 4131 return ret; 4132 } 4133 4134 static int f2fs_ioc_compress_file(struct file *filp) 4135 { 4136 struct inode *inode = file_inode(filp); 4137 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4138 pgoff_t page_idx = 0, last_idx; 4139 unsigned int blk_per_seg = sbi->blocks_per_seg; 4140 int cluster_size = F2FS_I(inode)->i_cluster_size; 4141 int count, ret; 4142 4143 if (!f2fs_sb_has_compression(sbi) || 4144 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4145 return -EOPNOTSUPP; 4146 4147 if (!(filp->f_mode & FMODE_WRITE)) 4148 return -EBADF; 4149 4150 if (!f2fs_compressed_file(inode)) 4151 return -EINVAL; 4152 4153 f2fs_balance_fs(sbi, true); 4154 4155 file_start_write(filp); 4156 inode_lock(inode); 4157 4158 if (!f2fs_is_compress_backend_ready(inode)) { 4159 ret = -EOPNOTSUPP; 4160 goto out; 4161 } 4162 4163 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4164 ret = -EINVAL; 4165 goto out; 4166 } 4167 4168 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4169 if (ret) 4170 goto out; 4171 4172 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4173 4174 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4175 4176 count = last_idx - page_idx; 4177 while (count && count >= cluster_size) { 4178 ret = redirty_blocks(inode, page_idx, cluster_size); 4179 if (ret < 0) 4180 break; 4181 4182 if (get_dirty_pages(inode) >= blk_per_seg) { 4183 ret = filemap_fdatawrite(inode->i_mapping); 4184 if (ret < 0) 4185 break; 4186 } 4187 4188 count -= cluster_size; 4189 page_idx += cluster_size; 4190 4191 cond_resched(); 4192 if (fatal_signal_pending(current)) { 4193 ret = -EINTR; 4194 break; 4195 } 4196 } 4197 4198 if (!ret) 4199 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4200 LLONG_MAX); 4201 4202 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4203 4204 if (ret) 4205 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4206 __func__, ret); 4207 out: 4208 inode_unlock(inode); 4209 file_end_write(filp); 4210 4211 return ret; 4212 } 4213 4214 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4215 { 4216 switch (cmd) { 4217 case FS_IOC_GETVERSION: 4218 return f2fs_ioc_getversion(filp, arg); 4219 case F2FS_IOC_START_ATOMIC_WRITE: 4220 return f2fs_ioc_start_atomic_write(filp, false); 4221 case F2FS_IOC_START_ATOMIC_REPLACE: 4222 return f2fs_ioc_start_atomic_write(filp, true); 4223 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4224 return f2fs_ioc_commit_atomic_write(filp); 4225 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4226 return f2fs_ioc_abort_atomic_write(filp); 4227 case F2FS_IOC_START_VOLATILE_WRITE: 4228 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4229 return -EOPNOTSUPP; 4230 case F2FS_IOC_SHUTDOWN: 4231 return f2fs_ioc_shutdown(filp, arg); 4232 case FITRIM: 4233 return f2fs_ioc_fitrim(filp, arg); 4234 case FS_IOC_SET_ENCRYPTION_POLICY: 4235 return f2fs_ioc_set_encryption_policy(filp, arg); 4236 case FS_IOC_GET_ENCRYPTION_POLICY: 4237 return f2fs_ioc_get_encryption_policy(filp, arg); 4238 case FS_IOC_GET_ENCRYPTION_PWSALT: 4239 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4240 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4241 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4242 case FS_IOC_ADD_ENCRYPTION_KEY: 4243 return f2fs_ioc_add_encryption_key(filp, arg); 4244 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4245 return f2fs_ioc_remove_encryption_key(filp, arg); 4246 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4247 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4248 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4249 return f2fs_ioc_get_encryption_key_status(filp, arg); 4250 case FS_IOC_GET_ENCRYPTION_NONCE: 4251 return f2fs_ioc_get_encryption_nonce(filp, arg); 4252 case F2FS_IOC_GARBAGE_COLLECT: 4253 return f2fs_ioc_gc(filp, arg); 4254 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4255 return f2fs_ioc_gc_range(filp, arg); 4256 case F2FS_IOC_WRITE_CHECKPOINT: 4257 return f2fs_ioc_write_checkpoint(filp); 4258 case F2FS_IOC_DEFRAGMENT: 4259 return f2fs_ioc_defragment(filp, arg); 4260 case F2FS_IOC_MOVE_RANGE: 4261 return f2fs_ioc_move_range(filp, arg); 4262 case F2FS_IOC_FLUSH_DEVICE: 4263 return f2fs_ioc_flush_device(filp, arg); 4264 case F2FS_IOC_GET_FEATURES: 4265 return f2fs_ioc_get_features(filp, arg); 4266 case F2FS_IOC_GET_PIN_FILE: 4267 return f2fs_ioc_get_pin_file(filp, arg); 4268 case F2FS_IOC_SET_PIN_FILE: 4269 return f2fs_ioc_set_pin_file(filp, arg); 4270 case F2FS_IOC_PRECACHE_EXTENTS: 4271 return f2fs_ioc_precache_extents(filp); 4272 case F2FS_IOC_RESIZE_FS: 4273 return f2fs_ioc_resize_fs(filp, arg); 4274 case FS_IOC_ENABLE_VERITY: 4275 return f2fs_ioc_enable_verity(filp, arg); 4276 case FS_IOC_MEASURE_VERITY: 4277 return f2fs_ioc_measure_verity(filp, arg); 4278 case FS_IOC_READ_VERITY_METADATA: 4279 return f2fs_ioc_read_verity_metadata(filp, arg); 4280 case FS_IOC_GETFSLABEL: 4281 return f2fs_ioc_getfslabel(filp, arg); 4282 case FS_IOC_SETFSLABEL: 4283 return f2fs_ioc_setfslabel(filp, arg); 4284 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4285 return f2fs_ioc_get_compress_blocks(filp, arg); 4286 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4287 return f2fs_release_compress_blocks(filp, arg); 4288 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4289 return f2fs_reserve_compress_blocks(filp, arg); 4290 case F2FS_IOC_SEC_TRIM_FILE: 4291 return f2fs_sec_trim_file(filp, arg); 4292 case F2FS_IOC_GET_COMPRESS_OPTION: 4293 return f2fs_ioc_get_compress_option(filp, arg); 4294 case F2FS_IOC_SET_COMPRESS_OPTION: 4295 return f2fs_ioc_set_compress_option(filp, arg); 4296 case F2FS_IOC_DECOMPRESS_FILE: 4297 return f2fs_ioc_decompress_file(filp); 4298 case F2FS_IOC_COMPRESS_FILE: 4299 return f2fs_ioc_compress_file(filp); 4300 default: 4301 return -ENOTTY; 4302 } 4303 } 4304 4305 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4306 { 4307 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4308 return -EIO; 4309 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4310 return -ENOSPC; 4311 4312 return __f2fs_ioctl(filp, cmd, arg); 4313 } 4314 4315 /* 4316 * Return %true if the given read or write request should use direct I/O, or 4317 * %false if it should use buffered I/O. 4318 */ 4319 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4320 struct iov_iter *iter) 4321 { 4322 unsigned int align; 4323 4324 if (!(iocb->ki_flags & IOCB_DIRECT)) 4325 return false; 4326 4327 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4328 return false; 4329 4330 /* 4331 * Direct I/O not aligned to the disk's logical_block_size will be 4332 * attempted, but will fail with -EINVAL. 4333 * 4334 * f2fs additionally requires that direct I/O be aligned to the 4335 * filesystem block size, which is often a stricter requirement. 4336 * However, f2fs traditionally falls back to buffered I/O on requests 4337 * that are logical_block_size-aligned but not fs-block aligned. 4338 * 4339 * The below logic implements this behavior. 4340 */ 4341 align = iocb->ki_pos | iov_iter_alignment(iter); 4342 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4343 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4344 return false; 4345 4346 return true; 4347 } 4348 4349 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4350 unsigned int flags) 4351 { 4352 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4353 4354 dec_page_count(sbi, F2FS_DIO_READ); 4355 if (error) 4356 return error; 4357 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4358 return 0; 4359 } 4360 4361 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4362 .end_io = f2fs_dio_read_end_io, 4363 }; 4364 4365 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4366 { 4367 struct file *file = iocb->ki_filp; 4368 struct inode *inode = file_inode(file); 4369 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4370 struct f2fs_inode_info *fi = F2FS_I(inode); 4371 const loff_t pos = iocb->ki_pos; 4372 const size_t count = iov_iter_count(to); 4373 struct iomap_dio *dio; 4374 ssize_t ret; 4375 4376 if (count == 0) 4377 return 0; /* skip atime update */ 4378 4379 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4380 4381 if (iocb->ki_flags & IOCB_NOWAIT) { 4382 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4383 ret = -EAGAIN; 4384 goto out; 4385 } 4386 } else { 4387 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4388 } 4389 4390 /* 4391 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4392 * the higher-level function iomap_dio_rw() in order to ensure that the 4393 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4394 */ 4395 inc_page_count(sbi, F2FS_DIO_READ); 4396 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4397 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4398 if (IS_ERR_OR_NULL(dio)) { 4399 ret = PTR_ERR_OR_ZERO(dio); 4400 if (ret != -EIOCBQUEUED) 4401 dec_page_count(sbi, F2FS_DIO_READ); 4402 } else { 4403 ret = iomap_dio_complete(dio); 4404 } 4405 4406 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4407 4408 file_accessed(file); 4409 out: 4410 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4411 return ret; 4412 } 4413 4414 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4415 int rw) 4416 { 4417 struct inode *inode = file_inode(file); 4418 char *buf, *path; 4419 4420 buf = f2fs_getname(F2FS_I_SB(inode)); 4421 if (!buf) 4422 return; 4423 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4424 if (IS_ERR(path)) 4425 goto free_buf; 4426 if (rw == WRITE) 4427 trace_f2fs_datawrite_start(inode, pos, count, 4428 current->pid, path, current->comm); 4429 else 4430 trace_f2fs_dataread_start(inode, pos, count, 4431 current->pid, path, current->comm); 4432 free_buf: 4433 f2fs_putname(buf); 4434 } 4435 4436 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4437 { 4438 struct inode *inode = file_inode(iocb->ki_filp); 4439 const loff_t pos = iocb->ki_pos; 4440 ssize_t ret; 4441 4442 if (!f2fs_is_compress_backend_ready(inode)) 4443 return -EOPNOTSUPP; 4444 4445 if (trace_f2fs_dataread_start_enabled()) 4446 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4447 iov_iter_count(to), READ); 4448 4449 if (f2fs_should_use_dio(inode, iocb, to)) { 4450 ret = f2fs_dio_read_iter(iocb, to); 4451 } else { 4452 ret = filemap_read(iocb, to, 0); 4453 if (ret > 0) 4454 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4455 APP_BUFFERED_READ_IO, ret); 4456 } 4457 if (trace_f2fs_dataread_end_enabled()) 4458 trace_f2fs_dataread_end(inode, pos, ret); 4459 return ret; 4460 } 4461 4462 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4463 struct pipe_inode_info *pipe, 4464 size_t len, unsigned int flags) 4465 { 4466 struct inode *inode = file_inode(in); 4467 const loff_t pos = *ppos; 4468 ssize_t ret; 4469 4470 if (!f2fs_is_compress_backend_ready(inode)) 4471 return -EOPNOTSUPP; 4472 4473 if (trace_f2fs_dataread_start_enabled()) 4474 f2fs_trace_rw_file_path(in, pos, len, READ); 4475 4476 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4477 if (ret > 0) 4478 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4479 APP_BUFFERED_READ_IO, ret); 4480 4481 if (trace_f2fs_dataread_end_enabled()) 4482 trace_f2fs_dataread_end(inode, pos, ret); 4483 return ret; 4484 } 4485 4486 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4487 { 4488 struct file *file = iocb->ki_filp; 4489 struct inode *inode = file_inode(file); 4490 ssize_t count; 4491 int err; 4492 4493 if (IS_IMMUTABLE(inode)) 4494 return -EPERM; 4495 4496 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4497 return -EPERM; 4498 4499 count = generic_write_checks(iocb, from); 4500 if (count <= 0) 4501 return count; 4502 4503 err = file_modified(file); 4504 if (err) 4505 return err; 4506 return count; 4507 } 4508 4509 /* 4510 * Preallocate blocks for a write request, if it is possible and helpful to do 4511 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4512 * blocks were preallocated, or a negative errno value if something went 4513 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4514 * requested blocks (not just some of them) have been allocated. 4515 */ 4516 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4517 bool dio) 4518 { 4519 struct inode *inode = file_inode(iocb->ki_filp); 4520 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4521 const loff_t pos = iocb->ki_pos; 4522 const size_t count = iov_iter_count(iter); 4523 struct f2fs_map_blocks map = {}; 4524 int flag; 4525 int ret; 4526 4527 /* If it will be an out-of-place direct write, don't bother. */ 4528 if (dio && f2fs_lfs_mode(sbi)) 4529 return 0; 4530 /* 4531 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4532 * buffered IO, if DIO meets any holes. 4533 */ 4534 if (dio && i_size_read(inode) && 4535 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4536 return 0; 4537 4538 /* No-wait I/O can't allocate blocks. */ 4539 if (iocb->ki_flags & IOCB_NOWAIT) 4540 return 0; 4541 4542 /* If it will be a short write, don't bother. */ 4543 if (fault_in_iov_iter_readable(iter, count)) 4544 return 0; 4545 4546 if (f2fs_has_inline_data(inode)) { 4547 /* If the data will fit inline, don't bother. */ 4548 if (pos + count <= MAX_INLINE_DATA(inode)) 4549 return 0; 4550 ret = f2fs_convert_inline_inode(inode); 4551 if (ret) 4552 return ret; 4553 } 4554 4555 /* Do not preallocate blocks that will be written partially in 4KB. */ 4556 map.m_lblk = F2FS_BLK_ALIGN(pos); 4557 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4558 if (map.m_len > map.m_lblk) 4559 map.m_len -= map.m_lblk; 4560 else 4561 map.m_len = 0; 4562 map.m_may_create = true; 4563 if (dio) { 4564 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4565 flag = F2FS_GET_BLOCK_PRE_DIO; 4566 } else { 4567 map.m_seg_type = NO_CHECK_TYPE; 4568 flag = F2FS_GET_BLOCK_PRE_AIO; 4569 } 4570 4571 ret = f2fs_map_blocks(inode, &map, flag); 4572 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4573 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4574 return ret; 4575 if (ret == 0) 4576 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4577 return map.m_len; 4578 } 4579 4580 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4581 struct iov_iter *from) 4582 { 4583 struct file *file = iocb->ki_filp; 4584 struct inode *inode = file_inode(file); 4585 ssize_t ret; 4586 4587 if (iocb->ki_flags & IOCB_NOWAIT) 4588 return -EOPNOTSUPP; 4589 4590 ret = generic_perform_write(iocb, from); 4591 4592 if (ret > 0) { 4593 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4594 APP_BUFFERED_IO, ret); 4595 } 4596 return ret; 4597 } 4598 4599 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4600 unsigned int flags) 4601 { 4602 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4603 4604 dec_page_count(sbi, F2FS_DIO_WRITE); 4605 if (error) 4606 return error; 4607 f2fs_update_time(sbi, REQ_TIME); 4608 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4609 return 0; 4610 } 4611 4612 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4613 .end_io = f2fs_dio_write_end_io, 4614 }; 4615 4616 static void f2fs_flush_buffered_write(struct address_space *mapping, 4617 loff_t start_pos, loff_t end_pos) 4618 { 4619 int ret; 4620 4621 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4622 if (ret < 0) 4623 return; 4624 invalidate_mapping_pages(mapping, 4625 start_pos >> PAGE_SHIFT, 4626 end_pos >> PAGE_SHIFT); 4627 } 4628 4629 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4630 bool *may_need_sync) 4631 { 4632 struct file *file = iocb->ki_filp; 4633 struct inode *inode = file_inode(file); 4634 struct f2fs_inode_info *fi = F2FS_I(inode); 4635 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4636 const bool do_opu = f2fs_lfs_mode(sbi); 4637 const loff_t pos = iocb->ki_pos; 4638 const ssize_t count = iov_iter_count(from); 4639 unsigned int dio_flags; 4640 struct iomap_dio *dio; 4641 ssize_t ret; 4642 4643 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4644 4645 if (iocb->ki_flags & IOCB_NOWAIT) { 4646 /* f2fs_convert_inline_inode() and block allocation can block */ 4647 if (f2fs_has_inline_data(inode) || 4648 !f2fs_overwrite_io(inode, pos, count)) { 4649 ret = -EAGAIN; 4650 goto out; 4651 } 4652 4653 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4654 ret = -EAGAIN; 4655 goto out; 4656 } 4657 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4658 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4659 ret = -EAGAIN; 4660 goto out; 4661 } 4662 } else { 4663 ret = f2fs_convert_inline_inode(inode); 4664 if (ret) 4665 goto out; 4666 4667 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4668 if (do_opu) 4669 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4670 } 4671 4672 /* 4673 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4674 * the higher-level function iomap_dio_rw() in order to ensure that the 4675 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4676 */ 4677 inc_page_count(sbi, F2FS_DIO_WRITE); 4678 dio_flags = 0; 4679 if (pos + count > inode->i_size) 4680 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4681 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4682 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4683 if (IS_ERR_OR_NULL(dio)) { 4684 ret = PTR_ERR_OR_ZERO(dio); 4685 if (ret == -ENOTBLK) 4686 ret = 0; 4687 if (ret != -EIOCBQUEUED) 4688 dec_page_count(sbi, F2FS_DIO_WRITE); 4689 } else { 4690 ret = iomap_dio_complete(dio); 4691 } 4692 4693 if (do_opu) 4694 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4695 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4696 4697 if (ret < 0) 4698 goto out; 4699 if (pos + ret > inode->i_size) 4700 f2fs_i_size_write(inode, pos + ret); 4701 if (!do_opu) 4702 set_inode_flag(inode, FI_UPDATE_WRITE); 4703 4704 if (iov_iter_count(from)) { 4705 ssize_t ret2; 4706 loff_t bufio_start_pos = iocb->ki_pos; 4707 4708 /* 4709 * The direct write was partial, so we need to fall back to a 4710 * buffered write for the remainder. 4711 */ 4712 4713 ret2 = f2fs_buffered_write_iter(iocb, from); 4714 if (iov_iter_count(from)) 4715 f2fs_write_failed(inode, iocb->ki_pos); 4716 if (ret2 < 0) 4717 goto out; 4718 4719 /* 4720 * Ensure that the pagecache pages are written to disk and 4721 * invalidated to preserve the expected O_DIRECT semantics. 4722 */ 4723 if (ret2 > 0) { 4724 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4725 4726 ret += ret2; 4727 4728 f2fs_flush_buffered_write(file->f_mapping, 4729 bufio_start_pos, 4730 bufio_end_pos); 4731 } 4732 } else { 4733 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4734 *may_need_sync = false; 4735 } 4736 out: 4737 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4738 return ret; 4739 } 4740 4741 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4742 { 4743 struct inode *inode = file_inode(iocb->ki_filp); 4744 const loff_t orig_pos = iocb->ki_pos; 4745 const size_t orig_count = iov_iter_count(from); 4746 loff_t target_size; 4747 bool dio; 4748 bool may_need_sync = true; 4749 int preallocated; 4750 ssize_t ret; 4751 4752 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4753 ret = -EIO; 4754 goto out; 4755 } 4756 4757 if (!f2fs_is_compress_backend_ready(inode)) { 4758 ret = -EOPNOTSUPP; 4759 goto out; 4760 } 4761 4762 if (iocb->ki_flags & IOCB_NOWAIT) { 4763 if (!inode_trylock(inode)) { 4764 ret = -EAGAIN; 4765 goto out; 4766 } 4767 } else { 4768 inode_lock(inode); 4769 } 4770 4771 ret = f2fs_write_checks(iocb, from); 4772 if (ret <= 0) 4773 goto out_unlock; 4774 4775 /* Determine whether we will do a direct write or a buffered write. */ 4776 dio = f2fs_should_use_dio(inode, iocb, from); 4777 4778 /* Possibly preallocate the blocks for the write. */ 4779 target_size = iocb->ki_pos + iov_iter_count(from); 4780 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4781 if (preallocated < 0) { 4782 ret = preallocated; 4783 } else { 4784 if (trace_f2fs_datawrite_start_enabled()) 4785 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4786 orig_count, WRITE); 4787 4788 /* Do the actual write. */ 4789 ret = dio ? 4790 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4791 f2fs_buffered_write_iter(iocb, from); 4792 4793 if (trace_f2fs_datawrite_end_enabled()) 4794 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4795 } 4796 4797 /* Don't leave any preallocated blocks around past i_size. */ 4798 if (preallocated && i_size_read(inode) < target_size) { 4799 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4800 filemap_invalidate_lock(inode->i_mapping); 4801 if (!f2fs_truncate(inode)) 4802 file_dont_truncate(inode); 4803 filemap_invalidate_unlock(inode->i_mapping); 4804 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4805 } else { 4806 file_dont_truncate(inode); 4807 } 4808 4809 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4810 out_unlock: 4811 inode_unlock(inode); 4812 out: 4813 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4814 4815 if (ret > 0 && may_need_sync) 4816 ret = generic_write_sync(iocb, ret); 4817 4818 /* If buffered IO was forced, flush and drop the data from 4819 * the page cache to preserve O_DIRECT semantics 4820 */ 4821 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4822 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4823 orig_pos, 4824 orig_pos + ret - 1); 4825 4826 return ret; 4827 } 4828 4829 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4830 int advice) 4831 { 4832 struct address_space *mapping; 4833 struct backing_dev_info *bdi; 4834 struct inode *inode = file_inode(filp); 4835 int err; 4836 4837 if (advice == POSIX_FADV_SEQUENTIAL) { 4838 if (S_ISFIFO(inode->i_mode)) 4839 return -ESPIPE; 4840 4841 mapping = filp->f_mapping; 4842 if (!mapping || len < 0) 4843 return -EINVAL; 4844 4845 bdi = inode_to_bdi(mapping->host); 4846 filp->f_ra.ra_pages = bdi->ra_pages * 4847 F2FS_I_SB(inode)->seq_file_ra_mul; 4848 spin_lock(&filp->f_lock); 4849 filp->f_mode &= ~FMODE_RANDOM; 4850 spin_unlock(&filp->f_lock); 4851 return 0; 4852 } 4853 4854 err = generic_fadvise(filp, offset, len, advice); 4855 if (!err && advice == POSIX_FADV_DONTNEED && 4856 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4857 f2fs_compressed_file(inode)) 4858 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4859 4860 return err; 4861 } 4862 4863 #ifdef CONFIG_COMPAT 4864 struct compat_f2fs_gc_range { 4865 u32 sync; 4866 compat_u64 start; 4867 compat_u64 len; 4868 }; 4869 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4870 struct compat_f2fs_gc_range) 4871 4872 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4873 { 4874 struct compat_f2fs_gc_range __user *urange; 4875 struct f2fs_gc_range range; 4876 int err; 4877 4878 urange = compat_ptr(arg); 4879 err = get_user(range.sync, &urange->sync); 4880 err |= get_user(range.start, &urange->start); 4881 err |= get_user(range.len, &urange->len); 4882 if (err) 4883 return -EFAULT; 4884 4885 return __f2fs_ioc_gc_range(file, &range); 4886 } 4887 4888 struct compat_f2fs_move_range { 4889 u32 dst_fd; 4890 compat_u64 pos_in; 4891 compat_u64 pos_out; 4892 compat_u64 len; 4893 }; 4894 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4895 struct compat_f2fs_move_range) 4896 4897 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4898 { 4899 struct compat_f2fs_move_range __user *urange; 4900 struct f2fs_move_range range; 4901 int err; 4902 4903 urange = compat_ptr(arg); 4904 err = get_user(range.dst_fd, &urange->dst_fd); 4905 err |= get_user(range.pos_in, &urange->pos_in); 4906 err |= get_user(range.pos_out, &urange->pos_out); 4907 err |= get_user(range.len, &urange->len); 4908 if (err) 4909 return -EFAULT; 4910 4911 return __f2fs_ioc_move_range(file, &range); 4912 } 4913 4914 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4915 { 4916 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4917 return -EIO; 4918 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4919 return -ENOSPC; 4920 4921 switch (cmd) { 4922 case FS_IOC32_GETVERSION: 4923 cmd = FS_IOC_GETVERSION; 4924 break; 4925 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4926 return f2fs_compat_ioc_gc_range(file, arg); 4927 case F2FS_IOC32_MOVE_RANGE: 4928 return f2fs_compat_ioc_move_range(file, arg); 4929 case F2FS_IOC_START_ATOMIC_WRITE: 4930 case F2FS_IOC_START_ATOMIC_REPLACE: 4931 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4932 case F2FS_IOC_START_VOLATILE_WRITE: 4933 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4934 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4935 case F2FS_IOC_SHUTDOWN: 4936 case FITRIM: 4937 case FS_IOC_SET_ENCRYPTION_POLICY: 4938 case FS_IOC_GET_ENCRYPTION_PWSALT: 4939 case FS_IOC_GET_ENCRYPTION_POLICY: 4940 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4941 case FS_IOC_ADD_ENCRYPTION_KEY: 4942 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4943 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4944 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4945 case FS_IOC_GET_ENCRYPTION_NONCE: 4946 case F2FS_IOC_GARBAGE_COLLECT: 4947 case F2FS_IOC_WRITE_CHECKPOINT: 4948 case F2FS_IOC_DEFRAGMENT: 4949 case F2FS_IOC_FLUSH_DEVICE: 4950 case F2FS_IOC_GET_FEATURES: 4951 case F2FS_IOC_GET_PIN_FILE: 4952 case F2FS_IOC_SET_PIN_FILE: 4953 case F2FS_IOC_PRECACHE_EXTENTS: 4954 case F2FS_IOC_RESIZE_FS: 4955 case FS_IOC_ENABLE_VERITY: 4956 case FS_IOC_MEASURE_VERITY: 4957 case FS_IOC_READ_VERITY_METADATA: 4958 case FS_IOC_GETFSLABEL: 4959 case FS_IOC_SETFSLABEL: 4960 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4961 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4962 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4963 case F2FS_IOC_SEC_TRIM_FILE: 4964 case F2FS_IOC_GET_COMPRESS_OPTION: 4965 case F2FS_IOC_SET_COMPRESS_OPTION: 4966 case F2FS_IOC_DECOMPRESS_FILE: 4967 case F2FS_IOC_COMPRESS_FILE: 4968 break; 4969 default: 4970 return -ENOIOCTLCMD; 4971 } 4972 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4973 } 4974 #endif 4975 4976 const struct file_operations f2fs_file_operations = { 4977 .llseek = f2fs_llseek, 4978 .read_iter = f2fs_file_read_iter, 4979 .write_iter = f2fs_file_write_iter, 4980 .iopoll = iocb_bio_iopoll, 4981 .open = f2fs_file_open, 4982 .release = f2fs_release_file, 4983 .mmap = f2fs_file_mmap, 4984 .flush = f2fs_file_flush, 4985 .fsync = f2fs_sync_file, 4986 .fallocate = f2fs_fallocate, 4987 .unlocked_ioctl = f2fs_ioctl, 4988 #ifdef CONFIG_COMPAT 4989 .compat_ioctl = f2fs_compat_ioctl, 4990 #endif 4991 .splice_read = f2fs_file_splice_read, 4992 .splice_write = iter_file_splice_write, 4993 .fadvise = f2fs_file_fadvise, 4994 };