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_flags_t flags = vmf->vma->vm_flags; 43 vm_fault_t ret; 44 45 ret = filemap_fault(vmf); 46 if (ret & VM_FAULT_LOCKED) 47 f2fs_update_iostat(F2FS_I_SB(inode), inode, 48 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 49 50 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret); 51 52 return ret; 53 } 54 55 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 56 { 57 struct page *page = vmf->page; 58 struct inode *inode = file_inode(vmf->vma->vm_file); 59 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 60 struct dnode_of_data dn; 61 bool need_alloc = true; 62 int err = 0; 63 vm_fault_t ret; 64 65 if (unlikely(IS_IMMUTABLE(inode))) 66 return VM_FAULT_SIGBUS; 67 68 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 69 err = -EIO; 70 goto out; 71 } 72 73 if (unlikely(f2fs_cp_error(sbi))) { 74 err = -EIO; 75 goto out; 76 } 77 78 if (!f2fs_is_checkpoint_ready(sbi)) { 79 err = -ENOSPC; 80 goto out; 81 } 82 83 err = f2fs_convert_inline_inode(inode); 84 if (err) 85 goto out; 86 87 #ifdef CONFIG_F2FS_FS_COMPRESSION 88 if (f2fs_compressed_file(inode)) { 89 int ret = f2fs_is_compressed_cluster(inode, page->index); 90 91 if (ret < 0) { 92 err = ret; 93 goto out; 94 } else if (ret) { 95 need_alloc = false; 96 } 97 } 98 #endif 99 /* should do out of any locked page */ 100 if (need_alloc) 101 f2fs_balance_fs(sbi, true); 102 103 sb_start_pagefault(inode->i_sb); 104 105 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 106 107 file_update_time(vmf->vma->vm_file); 108 filemap_invalidate_lock_shared(inode->i_mapping); 109 lock_page(page); 110 if (unlikely(page->mapping != inode->i_mapping || 111 page_offset(page) > i_size_read(inode) || 112 !PageUptodate(page))) { 113 unlock_page(page); 114 err = -EFAULT; 115 goto out_sem; 116 } 117 118 if (need_alloc) { 119 /* block allocation */ 120 set_new_dnode(&dn, inode, NULL, NULL, 0); 121 err = f2fs_get_block_locked(&dn, page->index); 122 } 123 124 #ifdef CONFIG_F2FS_FS_COMPRESSION 125 if (!need_alloc) { 126 set_new_dnode(&dn, inode, NULL, NULL, 0); 127 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE); 128 f2fs_put_dnode(&dn); 129 } 130 #endif 131 if (err) { 132 unlock_page(page); 133 goto out_sem; 134 } 135 136 f2fs_wait_on_page_writeback(page, DATA, false, true); 137 138 /* wait for GCed page writeback via META_MAPPING */ 139 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 140 141 /* 142 * check to see if the page is mapped already (no holes) 143 */ 144 if (PageMappedToDisk(page)) 145 goto out_sem; 146 147 /* page is wholly or partially inside EOF */ 148 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 149 i_size_read(inode)) { 150 loff_t offset; 151 152 offset = i_size_read(inode) & ~PAGE_MASK; 153 zero_user_segment(page, offset, PAGE_SIZE); 154 } 155 set_page_dirty(page); 156 157 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 158 f2fs_update_time(sbi, REQ_TIME); 159 160 out_sem: 161 filemap_invalidate_unlock_shared(inode->i_mapping); 162 163 sb_end_pagefault(inode->i_sb); 164 out: 165 ret = vmf_fs_error(err); 166 167 trace_f2fs_vm_page_mkwrite(inode, page->index, vmf->vma->vm_flags, ret); 168 return ret; 169 } 170 171 static const struct vm_operations_struct f2fs_file_vm_ops = { 172 .fault = f2fs_filemap_fault, 173 .map_pages = filemap_map_pages, 174 .page_mkwrite = f2fs_vm_page_mkwrite, 175 }; 176 177 static int get_parent_ino(struct inode *inode, nid_t *pino) 178 { 179 struct dentry *dentry; 180 181 /* 182 * Make sure to get the non-deleted alias. The alias associated with 183 * the open file descriptor being fsync()'ed may be deleted already. 184 */ 185 dentry = d_find_alias(inode); 186 if (!dentry) 187 return 0; 188 189 *pino = parent_ino(dentry); 190 dput(dentry); 191 return 1; 192 } 193 194 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 195 { 196 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 197 enum cp_reason_type cp_reason = CP_NO_NEEDED; 198 199 if (!S_ISREG(inode->i_mode)) 200 cp_reason = CP_NON_REGULAR; 201 else if (f2fs_compressed_file(inode)) 202 cp_reason = CP_COMPRESSED; 203 else if (inode->i_nlink != 1) 204 cp_reason = CP_HARDLINK; 205 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 206 cp_reason = CP_SB_NEED_CP; 207 else if (file_wrong_pino(inode)) 208 cp_reason = CP_WRONG_PINO; 209 else if (!f2fs_space_for_roll_forward(sbi)) 210 cp_reason = CP_NO_SPC_ROLL; 211 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 212 cp_reason = CP_NODE_NEED_CP; 213 else if (test_opt(sbi, FASTBOOT)) 214 cp_reason = CP_FASTBOOT_MODE; 215 else if (F2FS_OPTION(sbi).active_logs == 2) 216 cp_reason = CP_SPEC_LOG_NUM; 217 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 218 f2fs_need_dentry_mark(sbi, inode->i_ino) && 219 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 220 TRANS_DIR_INO)) 221 cp_reason = CP_RECOVER_DIR; 222 223 return cp_reason; 224 } 225 226 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 227 { 228 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 229 bool ret = false; 230 /* But we need to avoid that there are some inode updates */ 231 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 232 ret = true; 233 f2fs_put_page(i, 0); 234 return ret; 235 } 236 237 static void try_to_fix_pino(struct inode *inode) 238 { 239 struct f2fs_inode_info *fi = F2FS_I(inode); 240 nid_t pino; 241 242 f2fs_down_write(&fi->i_sem); 243 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 244 get_parent_ino(inode, &pino)) { 245 f2fs_i_pino_write(inode, pino); 246 file_got_pino(inode); 247 } 248 f2fs_up_write(&fi->i_sem); 249 } 250 251 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 252 int datasync, bool atomic) 253 { 254 struct inode *inode = file->f_mapping->host; 255 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 256 nid_t ino = inode->i_ino; 257 int ret = 0; 258 enum cp_reason_type cp_reason = 0; 259 struct writeback_control wbc = { 260 .sync_mode = WB_SYNC_ALL, 261 .nr_to_write = LONG_MAX, 262 .for_reclaim = 0, 263 }; 264 unsigned int seq_id = 0; 265 266 if (unlikely(f2fs_readonly(inode->i_sb))) 267 return 0; 268 269 trace_f2fs_sync_file_enter(inode); 270 271 if (S_ISDIR(inode->i_mode)) 272 goto go_write; 273 274 /* if fdatasync is triggered, let's do in-place-update */ 275 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 276 set_inode_flag(inode, FI_NEED_IPU); 277 ret = file_write_and_wait_range(file, start, end); 278 clear_inode_flag(inode, FI_NEED_IPU); 279 280 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 281 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 282 return ret; 283 } 284 285 /* if the inode is dirty, let's recover all the time */ 286 if (!f2fs_skip_inode_update(inode, datasync)) { 287 f2fs_write_inode(inode, NULL); 288 goto go_write; 289 } 290 291 /* 292 * if there is no written data, don't waste time to write recovery info. 293 */ 294 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 295 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 296 297 /* it may call write_inode just prior to fsync */ 298 if (need_inode_page_update(sbi, ino)) 299 goto go_write; 300 301 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 302 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 303 goto flush_out; 304 goto out; 305 } else { 306 /* 307 * for OPU case, during fsync(), node can be persisted before 308 * data when lower device doesn't support write barrier, result 309 * in data corruption after SPO. 310 * So for strict fsync mode, force to use atomic write semantics 311 * to keep write order in between data/node and last node to 312 * avoid potential data corruption. 313 */ 314 if (F2FS_OPTION(sbi).fsync_mode == 315 FSYNC_MODE_STRICT && !atomic) 316 atomic = true; 317 } 318 go_write: 319 /* 320 * Both of fdatasync() and fsync() are able to be recovered from 321 * sudden-power-off. 322 */ 323 f2fs_down_read(&F2FS_I(inode)->i_sem); 324 cp_reason = need_do_checkpoint(inode); 325 f2fs_up_read(&F2FS_I(inode)->i_sem); 326 327 if (cp_reason) { 328 /* all the dirty node pages should be flushed for POR */ 329 ret = f2fs_sync_fs(inode->i_sb, 1); 330 331 /* 332 * We've secured consistency through sync_fs. Following pino 333 * will be used only for fsynced inodes after checkpoint. 334 */ 335 try_to_fix_pino(inode); 336 clear_inode_flag(inode, FI_APPEND_WRITE); 337 clear_inode_flag(inode, FI_UPDATE_WRITE); 338 goto out; 339 } 340 sync_nodes: 341 atomic_inc(&sbi->wb_sync_req[NODE]); 342 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 343 atomic_dec(&sbi->wb_sync_req[NODE]); 344 if (ret) 345 goto out; 346 347 /* if cp_error was enabled, we should avoid infinite loop */ 348 if (unlikely(f2fs_cp_error(sbi))) { 349 ret = -EIO; 350 goto out; 351 } 352 353 if (f2fs_need_inode_block_update(sbi, ino)) { 354 f2fs_mark_inode_dirty_sync(inode, true); 355 f2fs_write_inode(inode, NULL); 356 goto sync_nodes; 357 } 358 359 /* 360 * If it's atomic_write, it's just fine to keep write ordering. So 361 * here we don't need to wait for node write completion, since we use 362 * node chain which serializes node blocks. If one of node writes are 363 * reordered, we can see simply broken chain, resulting in stopping 364 * roll-forward recovery. It means we'll recover all or none node blocks 365 * given fsync mark. 366 */ 367 if (!atomic) { 368 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 369 if (ret) 370 goto out; 371 } 372 373 /* once recovery info is written, don't need to tack this */ 374 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 375 clear_inode_flag(inode, FI_APPEND_WRITE); 376 flush_out: 377 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) || 378 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi))) 379 ret = f2fs_issue_flush(sbi, inode->i_ino); 380 if (!ret) { 381 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 382 clear_inode_flag(inode, FI_UPDATE_WRITE); 383 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 384 } 385 f2fs_update_time(sbi, REQ_TIME); 386 out: 387 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 388 return ret; 389 } 390 391 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 392 { 393 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 394 return -EIO; 395 return f2fs_do_sync_file(file, start, end, datasync, false); 396 } 397 398 static bool __found_offset(struct address_space *mapping, 399 struct dnode_of_data *dn, pgoff_t index, int whence) 400 { 401 block_t blkaddr = f2fs_data_blkaddr(dn); 402 struct inode *inode = mapping->host; 403 bool compressed_cluster = false; 404 405 if (f2fs_compressed_file(inode)) { 406 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_page, 407 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size)); 408 409 compressed_cluster = first_blkaddr == COMPRESS_ADDR; 410 } 411 412 switch (whence) { 413 case SEEK_DATA: 414 if (__is_valid_data_blkaddr(blkaddr)) 415 return true; 416 if (blkaddr == NEW_ADDR && 417 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 418 return true; 419 if (compressed_cluster) 420 return true; 421 break; 422 case SEEK_HOLE: 423 if (compressed_cluster) 424 return false; 425 if (blkaddr == NULL_ADDR) 426 return true; 427 break; 428 } 429 return false; 430 } 431 432 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 433 { 434 struct inode *inode = file->f_mapping->host; 435 loff_t maxbytes = inode->i_sb->s_maxbytes; 436 struct dnode_of_data dn; 437 pgoff_t pgofs, end_offset; 438 loff_t data_ofs = offset; 439 loff_t isize; 440 int err = 0; 441 442 inode_lock_shared(inode); 443 444 isize = i_size_read(inode); 445 if (offset >= isize) 446 goto fail; 447 448 /* handle inline data case */ 449 if (f2fs_has_inline_data(inode)) { 450 if (whence == SEEK_HOLE) { 451 data_ofs = isize; 452 goto found; 453 } else if (whence == SEEK_DATA) { 454 data_ofs = offset; 455 goto found; 456 } 457 } 458 459 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 460 461 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 462 set_new_dnode(&dn, inode, NULL, NULL, 0); 463 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 464 if (err && err != -ENOENT) { 465 goto fail; 466 } else if (err == -ENOENT) { 467 /* direct node does not exists */ 468 if (whence == SEEK_DATA) { 469 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 470 continue; 471 } else { 472 goto found; 473 } 474 } 475 476 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 477 478 /* find data/hole in dnode block */ 479 for (; dn.ofs_in_node < end_offset; 480 dn.ofs_in_node++, pgofs++, 481 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 482 block_t blkaddr; 483 484 blkaddr = f2fs_data_blkaddr(&dn); 485 486 if (__is_valid_data_blkaddr(blkaddr) && 487 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 488 blkaddr, DATA_GENERIC_ENHANCE)) { 489 f2fs_put_dnode(&dn); 490 goto fail; 491 } 492 493 if (__found_offset(file->f_mapping, &dn, 494 pgofs, whence)) { 495 f2fs_put_dnode(&dn); 496 goto found; 497 } 498 } 499 f2fs_put_dnode(&dn); 500 } 501 502 if (whence == SEEK_DATA) 503 goto fail; 504 found: 505 if (whence == SEEK_HOLE && data_ofs > isize) 506 data_ofs = isize; 507 inode_unlock_shared(inode); 508 return vfs_setpos(file, data_ofs, maxbytes); 509 fail: 510 inode_unlock_shared(inode); 511 return -ENXIO; 512 } 513 514 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 515 { 516 struct inode *inode = file->f_mapping->host; 517 loff_t maxbytes = inode->i_sb->s_maxbytes; 518 519 if (f2fs_compressed_file(inode)) 520 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; 521 522 switch (whence) { 523 case SEEK_SET: 524 case SEEK_CUR: 525 case SEEK_END: 526 return generic_file_llseek_size(file, offset, whence, 527 maxbytes, i_size_read(inode)); 528 case SEEK_DATA: 529 case SEEK_HOLE: 530 if (offset < 0) 531 return -ENXIO; 532 return f2fs_seek_block(file, offset, whence); 533 } 534 535 return -EINVAL; 536 } 537 538 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 539 { 540 struct inode *inode = file_inode(file); 541 542 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 543 return -EIO; 544 545 if (!f2fs_is_compress_backend_ready(inode)) 546 return -EOPNOTSUPP; 547 548 file_accessed(file); 549 vma->vm_ops = &f2fs_file_vm_ops; 550 551 f2fs_down_read(&F2FS_I(inode)->i_sem); 552 set_inode_flag(inode, FI_MMAP_FILE); 553 f2fs_up_read(&F2FS_I(inode)->i_sem); 554 555 return 0; 556 } 557 558 static int f2fs_file_open(struct inode *inode, struct file *filp) 559 { 560 int err = fscrypt_file_open(inode, filp); 561 562 if (err) 563 return err; 564 565 if (!f2fs_is_compress_backend_ready(inode)) 566 return -EOPNOTSUPP; 567 568 err = fsverity_file_open(inode, filp); 569 if (err) 570 return err; 571 572 filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC; 573 filp->f_mode |= FMODE_CAN_ODIRECT; 574 575 return dquot_file_open(inode, filp); 576 } 577 578 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 579 { 580 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 581 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 582 __le32 *addr; 583 bool compressed_cluster = false; 584 int cluster_index = 0, valid_blocks = 0; 585 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 586 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 587 588 addr = get_dnode_addr(dn->inode, dn->node_page) + ofs; 589 590 /* Assumption: truncation starts with cluster */ 591 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 592 block_t blkaddr = le32_to_cpu(*addr); 593 594 if (f2fs_compressed_file(dn->inode) && 595 !(cluster_index & (cluster_size - 1))) { 596 if (compressed_cluster) 597 f2fs_i_compr_blocks_update(dn->inode, 598 valid_blocks, false); 599 compressed_cluster = (blkaddr == COMPRESS_ADDR); 600 valid_blocks = 0; 601 } 602 603 if (blkaddr == NULL_ADDR) 604 continue; 605 606 f2fs_set_data_blkaddr(dn, NULL_ADDR); 607 608 if (__is_valid_data_blkaddr(blkaddr)) { 609 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE)) 610 continue; 611 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr, 612 DATA_GENERIC_ENHANCE)) 613 continue; 614 if (compressed_cluster) 615 valid_blocks++; 616 } 617 618 f2fs_invalidate_blocks(sbi, blkaddr); 619 620 if (!released || blkaddr != COMPRESS_ADDR) 621 nr_free++; 622 } 623 624 if (compressed_cluster) 625 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 626 627 if (nr_free) { 628 pgoff_t fofs; 629 /* 630 * once we invalidate valid blkaddr in range [ofs, ofs + count], 631 * we will invalidate all blkaddr in the whole range. 632 */ 633 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 634 dn->inode) + ofs; 635 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 636 f2fs_update_age_extent_cache_range(dn, fofs, len); 637 dec_valid_block_count(sbi, dn->inode, nr_free); 638 } 639 dn->ofs_in_node = ofs; 640 641 f2fs_update_time(sbi, REQ_TIME); 642 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 643 dn->ofs_in_node, nr_free); 644 } 645 646 static int truncate_partial_data_page(struct inode *inode, u64 from, 647 bool cache_only) 648 { 649 loff_t offset = from & (PAGE_SIZE - 1); 650 pgoff_t index = from >> PAGE_SHIFT; 651 struct address_space *mapping = inode->i_mapping; 652 struct page *page; 653 654 if (!offset && !cache_only) 655 return 0; 656 657 if (cache_only) { 658 page = find_lock_page(mapping, index); 659 if (page && PageUptodate(page)) 660 goto truncate_out; 661 f2fs_put_page(page, 1); 662 return 0; 663 } 664 665 page = f2fs_get_lock_data_page(inode, index, true); 666 if (IS_ERR(page)) 667 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 668 truncate_out: 669 f2fs_wait_on_page_writeback(page, DATA, true, true); 670 zero_user(page, offset, PAGE_SIZE - offset); 671 672 /* An encrypted inode should have a key and truncate the last page. */ 673 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 674 if (!cache_only) 675 set_page_dirty(page); 676 f2fs_put_page(page, 1); 677 return 0; 678 } 679 680 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 681 { 682 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 683 struct dnode_of_data dn; 684 pgoff_t free_from; 685 int count = 0, err = 0; 686 struct page *ipage; 687 bool truncate_page = false; 688 689 trace_f2fs_truncate_blocks_enter(inode, from); 690 691 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 692 693 if (free_from >= max_file_blocks(inode)) 694 goto free_partial; 695 696 if (lock) 697 f2fs_lock_op(sbi); 698 699 ipage = f2fs_get_node_page(sbi, inode->i_ino); 700 if (IS_ERR(ipage)) { 701 err = PTR_ERR(ipage); 702 goto out; 703 } 704 705 if (f2fs_has_inline_data(inode)) { 706 f2fs_truncate_inline_inode(inode, ipage, from); 707 f2fs_put_page(ipage, 1); 708 truncate_page = true; 709 goto out; 710 } 711 712 set_new_dnode(&dn, inode, ipage, NULL, 0); 713 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 714 if (err) { 715 if (err == -ENOENT) 716 goto free_next; 717 goto out; 718 } 719 720 count = ADDRS_PER_PAGE(dn.node_page, inode); 721 722 count -= dn.ofs_in_node; 723 f2fs_bug_on(sbi, count < 0); 724 725 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 726 f2fs_truncate_data_blocks_range(&dn, count); 727 free_from += count; 728 } 729 730 f2fs_put_dnode(&dn); 731 free_next: 732 err = f2fs_truncate_inode_blocks(inode, free_from); 733 out: 734 if (lock) 735 f2fs_unlock_op(sbi); 736 free_partial: 737 /* lastly zero out the first data page */ 738 if (!err) 739 err = truncate_partial_data_page(inode, from, truncate_page); 740 741 trace_f2fs_truncate_blocks_exit(inode, err); 742 return err; 743 } 744 745 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 746 { 747 u64 free_from = from; 748 int err; 749 750 #ifdef CONFIG_F2FS_FS_COMPRESSION 751 /* 752 * for compressed file, only support cluster size 753 * aligned truncation. 754 */ 755 if (f2fs_compressed_file(inode)) 756 free_from = round_up(from, 757 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 758 #endif 759 760 err = f2fs_do_truncate_blocks(inode, free_from, lock); 761 if (err) 762 return err; 763 764 #ifdef CONFIG_F2FS_FS_COMPRESSION 765 /* 766 * For compressed file, after release compress blocks, don't allow write 767 * direct, but we should allow write direct after truncate to zero. 768 */ 769 if (f2fs_compressed_file(inode) && !free_from 770 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 771 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 772 773 if (from != free_from) { 774 err = f2fs_truncate_partial_cluster(inode, from, lock); 775 if (err) 776 return err; 777 } 778 #endif 779 780 return 0; 781 } 782 783 int f2fs_truncate(struct inode *inode) 784 { 785 int err; 786 787 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 788 return -EIO; 789 790 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 791 S_ISLNK(inode->i_mode))) 792 return 0; 793 794 trace_f2fs_truncate(inode); 795 796 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 797 return -EIO; 798 799 err = f2fs_dquot_initialize(inode); 800 if (err) 801 return err; 802 803 /* we should check inline_data size */ 804 if (!f2fs_may_inline_data(inode)) { 805 err = f2fs_convert_inline_inode(inode); 806 if (err) 807 return err; 808 } 809 810 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 811 if (err) 812 return err; 813 814 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 815 f2fs_mark_inode_dirty_sync(inode, false); 816 return 0; 817 } 818 819 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 820 { 821 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 822 823 if (!fscrypt_dio_supported(inode)) 824 return true; 825 if (fsverity_active(inode)) 826 return true; 827 if (f2fs_compressed_file(inode)) 828 return true; 829 830 /* disallow direct IO if any of devices has unaligned blksize */ 831 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 832 return true; 833 /* 834 * for blkzoned device, fallback direct IO to buffered IO, so 835 * all IOs can be serialized by log-structured write. 836 */ 837 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE)) 838 return true; 839 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 840 return true; 841 842 return false; 843 } 844 845 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 846 struct kstat *stat, u32 request_mask, unsigned int query_flags) 847 { 848 struct inode *inode = d_inode(path->dentry); 849 struct f2fs_inode_info *fi = F2FS_I(inode); 850 struct f2fs_inode *ri = NULL; 851 unsigned int flags; 852 853 if (f2fs_has_extra_attr(inode) && 854 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 855 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 856 stat->result_mask |= STATX_BTIME; 857 stat->btime.tv_sec = fi->i_crtime.tv_sec; 858 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 859 } 860 861 /* 862 * Return the DIO alignment restrictions if requested. We only return 863 * this information when requested, since on encrypted files it might 864 * take a fair bit of work to get if the file wasn't opened recently. 865 * 866 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 867 * cannot represent that, so in that case we report no DIO support. 868 */ 869 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 870 unsigned int bsize = i_blocksize(inode); 871 872 stat->result_mask |= STATX_DIOALIGN; 873 if (!f2fs_force_buffered_io(inode, WRITE)) { 874 stat->dio_mem_align = bsize; 875 stat->dio_offset_align = bsize; 876 } 877 } 878 879 flags = fi->i_flags; 880 if (flags & F2FS_COMPR_FL) 881 stat->attributes |= STATX_ATTR_COMPRESSED; 882 if (flags & F2FS_APPEND_FL) 883 stat->attributes |= STATX_ATTR_APPEND; 884 if (IS_ENCRYPTED(inode)) 885 stat->attributes |= STATX_ATTR_ENCRYPTED; 886 if (flags & F2FS_IMMUTABLE_FL) 887 stat->attributes |= STATX_ATTR_IMMUTABLE; 888 if (flags & F2FS_NODUMP_FL) 889 stat->attributes |= STATX_ATTR_NODUMP; 890 if (IS_VERITY(inode)) 891 stat->attributes |= STATX_ATTR_VERITY; 892 893 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 894 STATX_ATTR_APPEND | 895 STATX_ATTR_ENCRYPTED | 896 STATX_ATTR_IMMUTABLE | 897 STATX_ATTR_NODUMP | 898 STATX_ATTR_VERITY); 899 900 generic_fillattr(idmap, request_mask, inode, stat); 901 902 /* we need to show initial sectors used for inline_data/dentries */ 903 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 904 f2fs_has_inline_dentry(inode)) 905 stat->blocks += (stat->size + 511) >> 9; 906 907 return 0; 908 } 909 910 #ifdef CONFIG_F2FS_FS_POSIX_ACL 911 static void __setattr_copy(struct mnt_idmap *idmap, 912 struct inode *inode, const struct iattr *attr) 913 { 914 unsigned int ia_valid = attr->ia_valid; 915 916 i_uid_update(idmap, attr, inode); 917 i_gid_update(idmap, attr, inode); 918 if (ia_valid & ATTR_ATIME) 919 inode_set_atime_to_ts(inode, attr->ia_atime); 920 if (ia_valid & ATTR_MTIME) 921 inode_set_mtime_to_ts(inode, attr->ia_mtime); 922 if (ia_valid & ATTR_CTIME) 923 inode_set_ctime_to_ts(inode, attr->ia_ctime); 924 if (ia_valid & ATTR_MODE) { 925 umode_t mode = attr->ia_mode; 926 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode); 927 928 if (!vfsgid_in_group_p(vfsgid) && 929 !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) 930 mode &= ~S_ISGID; 931 set_acl_inode(inode, mode); 932 } 933 } 934 #else 935 #define __setattr_copy setattr_copy 936 #endif 937 938 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 939 struct iattr *attr) 940 { 941 struct inode *inode = d_inode(dentry); 942 int err; 943 944 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 945 return -EIO; 946 947 if (unlikely(IS_IMMUTABLE(inode))) 948 return -EPERM; 949 950 if (unlikely(IS_APPEND(inode) && 951 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 952 ATTR_GID | ATTR_TIMES_SET)))) 953 return -EPERM; 954 955 if ((attr->ia_valid & ATTR_SIZE) && 956 !f2fs_is_compress_backend_ready(inode)) 957 return -EOPNOTSUPP; 958 959 err = setattr_prepare(idmap, dentry, attr); 960 if (err) 961 return err; 962 963 err = fscrypt_prepare_setattr(dentry, attr); 964 if (err) 965 return err; 966 967 err = fsverity_prepare_setattr(dentry, attr); 968 if (err) 969 return err; 970 971 if (is_quota_modification(idmap, inode, attr)) { 972 err = f2fs_dquot_initialize(inode); 973 if (err) 974 return err; 975 } 976 if (i_uid_needs_update(idmap, attr, inode) || 977 i_gid_needs_update(idmap, attr, inode)) { 978 f2fs_lock_op(F2FS_I_SB(inode)); 979 err = dquot_transfer(idmap, inode, attr); 980 if (err) { 981 set_sbi_flag(F2FS_I_SB(inode), 982 SBI_QUOTA_NEED_REPAIR); 983 f2fs_unlock_op(F2FS_I_SB(inode)); 984 return err; 985 } 986 /* 987 * update uid/gid under lock_op(), so that dquot and inode can 988 * be updated atomically. 989 */ 990 i_uid_update(idmap, attr, inode); 991 i_gid_update(idmap, attr, inode); 992 f2fs_mark_inode_dirty_sync(inode, true); 993 f2fs_unlock_op(F2FS_I_SB(inode)); 994 } 995 996 if (attr->ia_valid & ATTR_SIZE) { 997 loff_t old_size = i_size_read(inode); 998 999 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1000 /* 1001 * should convert inline inode before i_size_write to 1002 * keep smaller than inline_data size with inline flag. 1003 */ 1004 err = f2fs_convert_inline_inode(inode); 1005 if (err) 1006 return err; 1007 } 1008 1009 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1010 filemap_invalidate_lock(inode->i_mapping); 1011 1012 truncate_setsize(inode, attr->ia_size); 1013 1014 if (attr->ia_size <= old_size) 1015 err = f2fs_truncate(inode); 1016 /* 1017 * do not trim all blocks after i_size if target size is 1018 * larger than i_size. 1019 */ 1020 filemap_invalidate_unlock(inode->i_mapping); 1021 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1022 if (err) 1023 return err; 1024 1025 spin_lock(&F2FS_I(inode)->i_size_lock); 1026 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1027 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1028 spin_unlock(&F2FS_I(inode)->i_size_lock); 1029 } 1030 1031 __setattr_copy(idmap, inode, attr); 1032 1033 if (attr->ia_valid & ATTR_MODE) { 1034 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1035 1036 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1037 if (!err) 1038 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1039 clear_inode_flag(inode, FI_ACL_MODE); 1040 } 1041 } 1042 1043 /* file size may changed here */ 1044 f2fs_mark_inode_dirty_sync(inode, true); 1045 1046 /* inode change will produce dirty node pages flushed by checkpoint */ 1047 f2fs_balance_fs(F2FS_I_SB(inode), true); 1048 1049 return err; 1050 } 1051 1052 const struct inode_operations f2fs_file_inode_operations = { 1053 .getattr = f2fs_getattr, 1054 .setattr = f2fs_setattr, 1055 .get_inode_acl = f2fs_get_acl, 1056 .set_acl = f2fs_set_acl, 1057 .listxattr = f2fs_listxattr, 1058 .fiemap = f2fs_fiemap, 1059 .fileattr_get = f2fs_fileattr_get, 1060 .fileattr_set = f2fs_fileattr_set, 1061 }; 1062 1063 static int fill_zero(struct inode *inode, pgoff_t index, 1064 loff_t start, loff_t len) 1065 { 1066 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1067 struct page *page; 1068 1069 if (!len) 1070 return 0; 1071 1072 f2fs_balance_fs(sbi, true); 1073 1074 f2fs_lock_op(sbi); 1075 page = f2fs_get_new_data_page(inode, NULL, index, false); 1076 f2fs_unlock_op(sbi); 1077 1078 if (IS_ERR(page)) 1079 return PTR_ERR(page); 1080 1081 f2fs_wait_on_page_writeback(page, DATA, true, true); 1082 zero_user(page, start, len); 1083 set_page_dirty(page); 1084 f2fs_put_page(page, 1); 1085 return 0; 1086 } 1087 1088 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1089 { 1090 int err; 1091 1092 while (pg_start < pg_end) { 1093 struct dnode_of_data dn; 1094 pgoff_t end_offset, count; 1095 1096 set_new_dnode(&dn, inode, NULL, NULL, 0); 1097 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1098 if (err) { 1099 if (err == -ENOENT) { 1100 pg_start = f2fs_get_next_page_offset(&dn, 1101 pg_start); 1102 continue; 1103 } 1104 return err; 1105 } 1106 1107 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1108 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1109 1110 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1111 1112 f2fs_truncate_data_blocks_range(&dn, count); 1113 f2fs_put_dnode(&dn); 1114 1115 pg_start += count; 1116 } 1117 return 0; 1118 } 1119 1120 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1121 { 1122 pgoff_t pg_start, pg_end; 1123 loff_t off_start, off_end; 1124 int ret; 1125 1126 ret = f2fs_convert_inline_inode(inode); 1127 if (ret) 1128 return ret; 1129 1130 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1131 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1132 1133 off_start = offset & (PAGE_SIZE - 1); 1134 off_end = (offset + len) & (PAGE_SIZE - 1); 1135 1136 if (pg_start == pg_end) { 1137 ret = fill_zero(inode, pg_start, off_start, 1138 off_end - off_start); 1139 if (ret) 1140 return ret; 1141 } else { 1142 if (off_start) { 1143 ret = fill_zero(inode, pg_start++, off_start, 1144 PAGE_SIZE - off_start); 1145 if (ret) 1146 return ret; 1147 } 1148 if (off_end) { 1149 ret = fill_zero(inode, pg_end, 0, off_end); 1150 if (ret) 1151 return ret; 1152 } 1153 1154 if (pg_start < pg_end) { 1155 loff_t blk_start, blk_end; 1156 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1157 1158 f2fs_balance_fs(sbi, true); 1159 1160 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1161 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1162 1163 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1164 filemap_invalidate_lock(inode->i_mapping); 1165 1166 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1167 1168 f2fs_lock_op(sbi); 1169 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1170 f2fs_unlock_op(sbi); 1171 1172 filemap_invalidate_unlock(inode->i_mapping); 1173 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1174 } 1175 } 1176 1177 return ret; 1178 } 1179 1180 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1181 int *do_replace, pgoff_t off, pgoff_t len) 1182 { 1183 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1184 struct dnode_of_data dn; 1185 int ret, done, i; 1186 1187 next_dnode: 1188 set_new_dnode(&dn, inode, NULL, NULL, 0); 1189 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1190 if (ret && ret != -ENOENT) { 1191 return ret; 1192 } else if (ret == -ENOENT) { 1193 if (dn.max_level == 0) 1194 return -ENOENT; 1195 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1196 dn.ofs_in_node, len); 1197 blkaddr += done; 1198 do_replace += done; 1199 goto next; 1200 } 1201 1202 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1203 dn.ofs_in_node, len); 1204 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1205 *blkaddr = f2fs_data_blkaddr(&dn); 1206 1207 if (__is_valid_data_blkaddr(*blkaddr) && 1208 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1209 DATA_GENERIC_ENHANCE)) { 1210 f2fs_put_dnode(&dn); 1211 return -EFSCORRUPTED; 1212 } 1213 1214 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1215 1216 if (f2fs_lfs_mode(sbi)) { 1217 f2fs_put_dnode(&dn); 1218 return -EOPNOTSUPP; 1219 } 1220 1221 /* do not invalidate this block address */ 1222 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1223 *do_replace = 1; 1224 } 1225 } 1226 f2fs_put_dnode(&dn); 1227 next: 1228 len -= done; 1229 off += done; 1230 if (len) 1231 goto next_dnode; 1232 return 0; 1233 } 1234 1235 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1236 int *do_replace, pgoff_t off, int len) 1237 { 1238 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1239 struct dnode_of_data dn; 1240 int ret, i; 1241 1242 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1243 if (*do_replace == 0) 1244 continue; 1245 1246 set_new_dnode(&dn, inode, NULL, NULL, 0); 1247 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1248 if (ret) { 1249 dec_valid_block_count(sbi, inode, 1); 1250 f2fs_invalidate_blocks(sbi, *blkaddr); 1251 } else { 1252 f2fs_update_data_blkaddr(&dn, *blkaddr); 1253 } 1254 f2fs_put_dnode(&dn); 1255 } 1256 return 0; 1257 } 1258 1259 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1260 block_t *blkaddr, int *do_replace, 1261 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1262 { 1263 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1264 pgoff_t i = 0; 1265 int ret; 1266 1267 while (i < len) { 1268 if (blkaddr[i] == NULL_ADDR && !full) { 1269 i++; 1270 continue; 1271 } 1272 1273 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1274 struct dnode_of_data dn; 1275 struct node_info ni; 1276 size_t new_size; 1277 pgoff_t ilen; 1278 1279 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1280 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1281 if (ret) 1282 return ret; 1283 1284 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1285 if (ret) { 1286 f2fs_put_dnode(&dn); 1287 return ret; 1288 } 1289 1290 ilen = min((pgoff_t) 1291 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1292 dn.ofs_in_node, len - i); 1293 do { 1294 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1295 f2fs_truncate_data_blocks_range(&dn, 1); 1296 1297 if (do_replace[i]) { 1298 f2fs_i_blocks_write(src_inode, 1299 1, false, false); 1300 f2fs_i_blocks_write(dst_inode, 1301 1, true, false); 1302 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1303 blkaddr[i], ni.version, true, false); 1304 1305 do_replace[i] = 0; 1306 } 1307 dn.ofs_in_node++; 1308 i++; 1309 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1310 if (dst_inode->i_size < new_size) 1311 f2fs_i_size_write(dst_inode, new_size); 1312 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1313 1314 f2fs_put_dnode(&dn); 1315 } else { 1316 struct page *psrc, *pdst; 1317 1318 psrc = f2fs_get_lock_data_page(src_inode, 1319 src + i, true); 1320 if (IS_ERR(psrc)) 1321 return PTR_ERR(psrc); 1322 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1323 true); 1324 if (IS_ERR(pdst)) { 1325 f2fs_put_page(psrc, 1); 1326 return PTR_ERR(pdst); 1327 } 1328 1329 f2fs_wait_on_page_writeback(pdst, DATA, true, true); 1330 1331 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1332 set_page_dirty(pdst); 1333 set_page_private_gcing(pdst); 1334 f2fs_put_page(pdst, 1); 1335 f2fs_put_page(psrc, 1); 1336 1337 ret = f2fs_truncate_hole(src_inode, 1338 src + i, src + i + 1); 1339 if (ret) 1340 return ret; 1341 i++; 1342 } 1343 } 1344 return 0; 1345 } 1346 1347 static int __exchange_data_block(struct inode *src_inode, 1348 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1349 pgoff_t len, bool full) 1350 { 1351 block_t *src_blkaddr; 1352 int *do_replace; 1353 pgoff_t olen; 1354 int ret; 1355 1356 while (len) { 1357 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1358 1359 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1360 array_size(olen, sizeof(block_t)), 1361 GFP_NOFS); 1362 if (!src_blkaddr) 1363 return -ENOMEM; 1364 1365 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1366 array_size(olen, sizeof(int)), 1367 GFP_NOFS); 1368 if (!do_replace) { 1369 kvfree(src_blkaddr); 1370 return -ENOMEM; 1371 } 1372 1373 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1374 do_replace, src, olen); 1375 if (ret) 1376 goto roll_back; 1377 1378 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1379 do_replace, src, dst, olen, full); 1380 if (ret) 1381 goto roll_back; 1382 1383 src += olen; 1384 dst += olen; 1385 len -= olen; 1386 1387 kvfree(src_blkaddr); 1388 kvfree(do_replace); 1389 } 1390 return 0; 1391 1392 roll_back: 1393 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1394 kvfree(src_blkaddr); 1395 kvfree(do_replace); 1396 return ret; 1397 } 1398 1399 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1400 { 1401 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1402 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1403 pgoff_t start = offset >> PAGE_SHIFT; 1404 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1405 int ret; 1406 1407 f2fs_balance_fs(sbi, true); 1408 1409 /* avoid gc operation during block exchange */ 1410 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1411 filemap_invalidate_lock(inode->i_mapping); 1412 1413 f2fs_lock_op(sbi); 1414 f2fs_drop_extent_tree(inode); 1415 truncate_pagecache(inode, offset); 1416 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1417 f2fs_unlock_op(sbi); 1418 1419 filemap_invalidate_unlock(inode->i_mapping); 1420 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1421 return ret; 1422 } 1423 1424 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1425 { 1426 loff_t new_size; 1427 int ret; 1428 1429 if (offset + len >= i_size_read(inode)) 1430 return -EINVAL; 1431 1432 /* collapse range should be aligned to block size of f2fs. */ 1433 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1434 return -EINVAL; 1435 1436 ret = f2fs_convert_inline_inode(inode); 1437 if (ret) 1438 return ret; 1439 1440 /* write out all dirty pages from offset */ 1441 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1442 if (ret) 1443 return ret; 1444 1445 ret = f2fs_do_collapse(inode, offset, len); 1446 if (ret) 1447 return ret; 1448 1449 /* write out all moved pages, if possible */ 1450 filemap_invalidate_lock(inode->i_mapping); 1451 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1452 truncate_pagecache(inode, offset); 1453 1454 new_size = i_size_read(inode) - len; 1455 ret = f2fs_truncate_blocks(inode, new_size, true); 1456 filemap_invalidate_unlock(inode->i_mapping); 1457 if (!ret) 1458 f2fs_i_size_write(inode, new_size); 1459 return ret; 1460 } 1461 1462 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1463 pgoff_t end) 1464 { 1465 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1466 pgoff_t index = start; 1467 unsigned int ofs_in_node = dn->ofs_in_node; 1468 blkcnt_t count = 0; 1469 int ret; 1470 1471 for (; index < end; index++, dn->ofs_in_node++) { 1472 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1473 count++; 1474 } 1475 1476 dn->ofs_in_node = ofs_in_node; 1477 ret = f2fs_reserve_new_blocks(dn, count); 1478 if (ret) 1479 return ret; 1480 1481 dn->ofs_in_node = ofs_in_node; 1482 for (index = start; index < end; index++, dn->ofs_in_node++) { 1483 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1484 /* 1485 * f2fs_reserve_new_blocks will not guarantee entire block 1486 * allocation. 1487 */ 1488 if (dn->data_blkaddr == NULL_ADDR) { 1489 ret = -ENOSPC; 1490 break; 1491 } 1492 1493 if (dn->data_blkaddr == NEW_ADDR) 1494 continue; 1495 1496 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1497 DATA_GENERIC_ENHANCE)) { 1498 ret = -EFSCORRUPTED; 1499 break; 1500 } 1501 1502 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1503 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1504 } 1505 1506 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1507 f2fs_update_age_extent_cache_range(dn, start, index - start); 1508 1509 return ret; 1510 } 1511 1512 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1513 int mode) 1514 { 1515 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1516 struct address_space *mapping = inode->i_mapping; 1517 pgoff_t index, pg_start, pg_end; 1518 loff_t new_size = i_size_read(inode); 1519 loff_t off_start, off_end; 1520 int ret = 0; 1521 1522 ret = inode_newsize_ok(inode, (len + offset)); 1523 if (ret) 1524 return ret; 1525 1526 ret = f2fs_convert_inline_inode(inode); 1527 if (ret) 1528 return ret; 1529 1530 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1531 if (ret) 1532 return ret; 1533 1534 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1535 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1536 1537 off_start = offset & (PAGE_SIZE - 1); 1538 off_end = (offset + len) & (PAGE_SIZE - 1); 1539 1540 if (pg_start == pg_end) { 1541 ret = fill_zero(inode, pg_start, off_start, 1542 off_end - off_start); 1543 if (ret) 1544 return ret; 1545 1546 new_size = max_t(loff_t, new_size, offset + len); 1547 } else { 1548 if (off_start) { 1549 ret = fill_zero(inode, pg_start++, off_start, 1550 PAGE_SIZE - off_start); 1551 if (ret) 1552 return ret; 1553 1554 new_size = max_t(loff_t, new_size, 1555 (loff_t)pg_start << PAGE_SHIFT); 1556 } 1557 1558 for (index = pg_start; index < pg_end;) { 1559 struct dnode_of_data dn; 1560 unsigned int end_offset; 1561 pgoff_t end; 1562 1563 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1564 filemap_invalidate_lock(mapping); 1565 1566 truncate_pagecache_range(inode, 1567 (loff_t)index << PAGE_SHIFT, 1568 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1569 1570 f2fs_lock_op(sbi); 1571 1572 set_new_dnode(&dn, inode, NULL, NULL, 0); 1573 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1574 if (ret) { 1575 f2fs_unlock_op(sbi); 1576 filemap_invalidate_unlock(mapping); 1577 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1578 goto out; 1579 } 1580 1581 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1582 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1583 1584 ret = f2fs_do_zero_range(&dn, index, end); 1585 f2fs_put_dnode(&dn); 1586 1587 f2fs_unlock_op(sbi); 1588 filemap_invalidate_unlock(mapping); 1589 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1590 1591 f2fs_balance_fs(sbi, dn.node_changed); 1592 1593 if (ret) 1594 goto out; 1595 1596 index = end; 1597 new_size = max_t(loff_t, new_size, 1598 (loff_t)index << PAGE_SHIFT); 1599 } 1600 1601 if (off_end) { 1602 ret = fill_zero(inode, pg_end, 0, off_end); 1603 if (ret) 1604 goto out; 1605 1606 new_size = max_t(loff_t, new_size, offset + len); 1607 } 1608 } 1609 1610 out: 1611 if (new_size > i_size_read(inode)) { 1612 if (mode & FALLOC_FL_KEEP_SIZE) 1613 file_set_keep_isize(inode); 1614 else 1615 f2fs_i_size_write(inode, new_size); 1616 } 1617 return ret; 1618 } 1619 1620 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1621 { 1622 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1623 struct address_space *mapping = inode->i_mapping; 1624 pgoff_t nr, pg_start, pg_end, delta, idx; 1625 loff_t new_size; 1626 int ret = 0; 1627 1628 new_size = i_size_read(inode) + len; 1629 ret = inode_newsize_ok(inode, new_size); 1630 if (ret) 1631 return ret; 1632 1633 if (offset >= i_size_read(inode)) 1634 return -EINVAL; 1635 1636 /* insert range should be aligned to block size of f2fs. */ 1637 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1638 return -EINVAL; 1639 1640 ret = f2fs_convert_inline_inode(inode); 1641 if (ret) 1642 return ret; 1643 1644 f2fs_balance_fs(sbi, true); 1645 1646 filemap_invalidate_lock(mapping); 1647 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1648 filemap_invalidate_unlock(mapping); 1649 if (ret) 1650 return ret; 1651 1652 /* write out all dirty pages from offset */ 1653 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1654 if (ret) 1655 return ret; 1656 1657 pg_start = offset >> PAGE_SHIFT; 1658 pg_end = (offset + len) >> PAGE_SHIFT; 1659 delta = pg_end - pg_start; 1660 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1661 1662 /* avoid gc operation during block exchange */ 1663 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1664 filemap_invalidate_lock(mapping); 1665 truncate_pagecache(inode, offset); 1666 1667 while (!ret && idx > pg_start) { 1668 nr = idx - pg_start; 1669 if (nr > delta) 1670 nr = delta; 1671 idx -= nr; 1672 1673 f2fs_lock_op(sbi); 1674 f2fs_drop_extent_tree(inode); 1675 1676 ret = __exchange_data_block(inode, inode, idx, 1677 idx + delta, nr, false); 1678 f2fs_unlock_op(sbi); 1679 } 1680 filemap_invalidate_unlock(mapping); 1681 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1682 if (ret) 1683 return ret; 1684 1685 /* write out all moved pages, if possible */ 1686 filemap_invalidate_lock(mapping); 1687 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1688 truncate_pagecache(inode, offset); 1689 filemap_invalidate_unlock(mapping); 1690 1691 if (!ret) 1692 f2fs_i_size_write(inode, new_size); 1693 return ret; 1694 } 1695 1696 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1697 loff_t len, int mode) 1698 { 1699 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1700 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1701 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1702 .m_may_create = true }; 1703 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1704 .init_gc_type = FG_GC, 1705 .should_migrate_blocks = false, 1706 .err_gc_skipped = true, 1707 .nr_free_secs = 0 }; 1708 pgoff_t pg_start, pg_end; 1709 loff_t new_size; 1710 loff_t off_end; 1711 block_t expanded = 0; 1712 int err; 1713 1714 err = inode_newsize_ok(inode, (len + offset)); 1715 if (err) 1716 return err; 1717 1718 err = f2fs_convert_inline_inode(inode); 1719 if (err) 1720 return err; 1721 1722 f2fs_balance_fs(sbi, true); 1723 1724 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1725 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1726 off_end = (offset + len) & (PAGE_SIZE - 1); 1727 1728 map.m_lblk = pg_start; 1729 map.m_len = pg_end - pg_start; 1730 if (off_end) 1731 map.m_len++; 1732 1733 if (!map.m_len) 1734 return 0; 1735 1736 if (f2fs_is_pinned_file(inode)) { 1737 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1738 block_t sec_len = roundup(map.m_len, sec_blks); 1739 1740 map.m_len = sec_blks; 1741 next_alloc: 1742 if (has_not_enough_free_secs(sbi, 0, 1743 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1744 f2fs_down_write(&sbi->gc_lock); 1745 stat_inc_gc_call_count(sbi, FOREGROUND); 1746 err = f2fs_gc(sbi, &gc_control); 1747 if (err && err != -ENODATA) 1748 goto out_err; 1749 } 1750 1751 f2fs_down_write(&sbi->pin_sem); 1752 1753 err = f2fs_allocate_pinning_section(sbi); 1754 if (err) { 1755 f2fs_up_write(&sbi->pin_sem); 1756 goto out_err; 1757 } 1758 1759 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1760 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1761 file_dont_truncate(inode); 1762 1763 f2fs_up_write(&sbi->pin_sem); 1764 1765 expanded += map.m_len; 1766 sec_len -= map.m_len; 1767 map.m_lblk += map.m_len; 1768 if (!err && sec_len) 1769 goto next_alloc; 1770 1771 map.m_len = expanded; 1772 } else { 1773 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1774 expanded = map.m_len; 1775 } 1776 out_err: 1777 if (err) { 1778 pgoff_t last_off; 1779 1780 if (!expanded) 1781 return err; 1782 1783 last_off = pg_start + expanded - 1; 1784 1785 /* update new size to the failed position */ 1786 new_size = (last_off == pg_end) ? offset + len : 1787 (loff_t)(last_off + 1) << PAGE_SHIFT; 1788 } else { 1789 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1790 } 1791 1792 if (new_size > i_size_read(inode)) { 1793 if (mode & FALLOC_FL_KEEP_SIZE) 1794 file_set_keep_isize(inode); 1795 else 1796 f2fs_i_size_write(inode, new_size); 1797 } 1798 1799 return err; 1800 } 1801 1802 static long f2fs_fallocate(struct file *file, int mode, 1803 loff_t offset, loff_t len) 1804 { 1805 struct inode *inode = file_inode(file); 1806 long ret = 0; 1807 1808 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1809 return -EIO; 1810 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1811 return -ENOSPC; 1812 if (!f2fs_is_compress_backend_ready(inode)) 1813 return -EOPNOTSUPP; 1814 1815 /* f2fs only support ->fallocate for regular file */ 1816 if (!S_ISREG(inode->i_mode)) 1817 return -EINVAL; 1818 1819 if (IS_ENCRYPTED(inode) && 1820 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1821 return -EOPNOTSUPP; 1822 1823 /* 1824 * Pinned file should not support partial truncation since the block 1825 * can be used by applications. 1826 */ 1827 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1828 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1829 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1830 return -EOPNOTSUPP; 1831 1832 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1833 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1834 FALLOC_FL_INSERT_RANGE)) 1835 return -EOPNOTSUPP; 1836 1837 inode_lock(inode); 1838 1839 ret = file_modified(file); 1840 if (ret) 1841 goto out; 1842 1843 if (mode & FALLOC_FL_PUNCH_HOLE) { 1844 if (offset >= inode->i_size) 1845 goto out; 1846 1847 ret = f2fs_punch_hole(inode, offset, len); 1848 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1849 ret = f2fs_collapse_range(inode, offset, len); 1850 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1851 ret = f2fs_zero_range(inode, offset, len, mode); 1852 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1853 ret = f2fs_insert_range(inode, offset, len); 1854 } else { 1855 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1856 } 1857 1858 if (!ret) { 1859 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1860 f2fs_mark_inode_dirty_sync(inode, false); 1861 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1862 } 1863 1864 out: 1865 inode_unlock(inode); 1866 1867 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1868 return ret; 1869 } 1870 1871 static int f2fs_release_file(struct inode *inode, struct file *filp) 1872 { 1873 /* 1874 * f2fs_release_file is called at every close calls. So we should 1875 * not drop any inmemory pages by close called by other process. 1876 */ 1877 if (!(filp->f_mode & FMODE_WRITE) || 1878 atomic_read(&inode->i_writecount) != 1) 1879 return 0; 1880 1881 inode_lock(inode); 1882 f2fs_abort_atomic_write(inode, true); 1883 inode_unlock(inode); 1884 1885 return 0; 1886 } 1887 1888 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1889 { 1890 struct inode *inode = file_inode(file); 1891 1892 /* 1893 * If the process doing a transaction is crashed, we should do 1894 * roll-back. Otherwise, other reader/write can see corrupted database 1895 * until all the writers close its file. Since this should be done 1896 * before dropping file lock, it needs to do in ->flush. 1897 */ 1898 if (F2FS_I(inode)->atomic_write_task == current && 1899 (current->flags & PF_EXITING)) { 1900 inode_lock(inode); 1901 f2fs_abort_atomic_write(inode, true); 1902 inode_unlock(inode); 1903 } 1904 1905 return 0; 1906 } 1907 1908 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1909 { 1910 struct f2fs_inode_info *fi = F2FS_I(inode); 1911 u32 masked_flags = fi->i_flags & mask; 1912 1913 /* mask can be shrunk by flags_valid selector */ 1914 iflags &= mask; 1915 1916 /* Is it quota file? Do not allow user to mess with it */ 1917 if (IS_NOQUOTA(inode)) 1918 return -EPERM; 1919 1920 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1921 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1922 return -EOPNOTSUPP; 1923 if (!f2fs_empty_dir(inode)) 1924 return -ENOTEMPTY; 1925 } 1926 1927 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1928 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1929 return -EOPNOTSUPP; 1930 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1931 return -EINVAL; 1932 } 1933 1934 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1935 if (masked_flags & F2FS_COMPR_FL) { 1936 if (!f2fs_disable_compressed_file(inode)) 1937 return -EINVAL; 1938 } else { 1939 /* try to convert inline_data to support compression */ 1940 int err = f2fs_convert_inline_inode(inode); 1941 if (err) 1942 return err; 1943 1944 f2fs_down_write(&F2FS_I(inode)->i_sem); 1945 if (!f2fs_may_compress(inode) || 1946 (S_ISREG(inode->i_mode) && 1947 F2FS_HAS_BLOCKS(inode))) { 1948 f2fs_up_write(&F2FS_I(inode)->i_sem); 1949 return -EINVAL; 1950 } 1951 err = set_compress_context(inode); 1952 f2fs_up_write(&F2FS_I(inode)->i_sem); 1953 1954 if (err) 1955 return err; 1956 } 1957 } 1958 1959 fi->i_flags = iflags | (fi->i_flags & ~mask); 1960 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1961 (fi->i_flags & F2FS_NOCOMP_FL)); 1962 1963 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1964 set_inode_flag(inode, FI_PROJ_INHERIT); 1965 else 1966 clear_inode_flag(inode, FI_PROJ_INHERIT); 1967 1968 inode_set_ctime_current(inode); 1969 f2fs_set_inode_flags(inode); 1970 f2fs_mark_inode_dirty_sync(inode, true); 1971 return 0; 1972 } 1973 1974 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1975 1976 /* 1977 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1978 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1979 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1980 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1981 * 1982 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1983 * FS_IOC_FSSETXATTR is done by the VFS. 1984 */ 1985 1986 static const struct { 1987 u32 iflag; 1988 u32 fsflag; 1989 } f2fs_fsflags_map[] = { 1990 { F2FS_COMPR_FL, FS_COMPR_FL }, 1991 { F2FS_SYNC_FL, FS_SYNC_FL }, 1992 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1993 { F2FS_APPEND_FL, FS_APPEND_FL }, 1994 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1995 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1996 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1997 { F2FS_INDEX_FL, FS_INDEX_FL }, 1998 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1999 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2000 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2001 }; 2002 2003 #define F2FS_GETTABLE_FS_FL ( \ 2004 FS_COMPR_FL | \ 2005 FS_SYNC_FL | \ 2006 FS_IMMUTABLE_FL | \ 2007 FS_APPEND_FL | \ 2008 FS_NODUMP_FL | \ 2009 FS_NOATIME_FL | \ 2010 FS_NOCOMP_FL | \ 2011 FS_INDEX_FL | \ 2012 FS_DIRSYNC_FL | \ 2013 FS_PROJINHERIT_FL | \ 2014 FS_ENCRYPT_FL | \ 2015 FS_INLINE_DATA_FL | \ 2016 FS_NOCOW_FL | \ 2017 FS_VERITY_FL | \ 2018 FS_CASEFOLD_FL) 2019 2020 #define F2FS_SETTABLE_FS_FL ( \ 2021 FS_COMPR_FL | \ 2022 FS_SYNC_FL | \ 2023 FS_IMMUTABLE_FL | \ 2024 FS_APPEND_FL | \ 2025 FS_NODUMP_FL | \ 2026 FS_NOATIME_FL | \ 2027 FS_NOCOMP_FL | \ 2028 FS_DIRSYNC_FL | \ 2029 FS_PROJINHERIT_FL | \ 2030 FS_CASEFOLD_FL) 2031 2032 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2033 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2034 { 2035 u32 fsflags = 0; 2036 int i; 2037 2038 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2039 if (iflags & f2fs_fsflags_map[i].iflag) 2040 fsflags |= f2fs_fsflags_map[i].fsflag; 2041 2042 return fsflags; 2043 } 2044 2045 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2046 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2047 { 2048 u32 iflags = 0; 2049 int i; 2050 2051 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2052 if (fsflags & f2fs_fsflags_map[i].fsflag) 2053 iflags |= f2fs_fsflags_map[i].iflag; 2054 2055 return iflags; 2056 } 2057 2058 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2059 { 2060 struct inode *inode = file_inode(filp); 2061 2062 return put_user(inode->i_generation, (int __user *)arg); 2063 } 2064 2065 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2066 { 2067 struct inode *inode = file_inode(filp); 2068 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2069 struct f2fs_inode_info *fi = F2FS_I(inode); 2070 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2071 struct inode *pinode; 2072 loff_t isize; 2073 int ret; 2074 2075 if (!inode_owner_or_capable(idmap, inode)) 2076 return -EACCES; 2077 2078 if (!S_ISREG(inode->i_mode)) 2079 return -EINVAL; 2080 2081 if (filp->f_flags & O_DIRECT) 2082 return -EINVAL; 2083 2084 ret = mnt_want_write_file(filp); 2085 if (ret) 2086 return ret; 2087 2088 inode_lock(inode); 2089 2090 if (!f2fs_disable_compressed_file(inode) || 2091 f2fs_is_pinned_file(inode)) { 2092 ret = -EINVAL; 2093 goto out; 2094 } 2095 2096 if (f2fs_is_atomic_file(inode)) 2097 goto out; 2098 2099 ret = f2fs_convert_inline_inode(inode); 2100 if (ret) 2101 goto out; 2102 2103 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2104 2105 /* 2106 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2107 * f2fs_is_atomic_file. 2108 */ 2109 if (get_dirty_pages(inode)) 2110 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2111 inode->i_ino, get_dirty_pages(inode)); 2112 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2113 if (ret) { 2114 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2115 goto out; 2116 } 2117 2118 /* Check if the inode already has a COW inode */ 2119 if (fi->cow_inode == NULL) { 2120 /* Create a COW inode for atomic write */ 2121 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2122 if (IS_ERR(pinode)) { 2123 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2124 ret = PTR_ERR(pinode); 2125 goto out; 2126 } 2127 2128 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2129 iput(pinode); 2130 if (ret) { 2131 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2132 goto out; 2133 } 2134 2135 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2136 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2137 } else { 2138 /* Reuse the already created COW inode */ 2139 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2140 if (ret) { 2141 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2142 goto out; 2143 } 2144 } 2145 2146 f2fs_write_inode(inode, NULL); 2147 2148 stat_inc_atomic_inode(inode); 2149 2150 set_inode_flag(inode, FI_ATOMIC_FILE); 2151 2152 isize = i_size_read(inode); 2153 fi->original_i_size = isize; 2154 if (truncate) { 2155 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2156 truncate_inode_pages_final(inode->i_mapping); 2157 f2fs_i_size_write(inode, 0); 2158 isize = 0; 2159 } 2160 f2fs_i_size_write(fi->cow_inode, isize); 2161 2162 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2163 2164 f2fs_update_time(sbi, REQ_TIME); 2165 fi->atomic_write_task = current; 2166 stat_update_max_atomic_write(inode); 2167 fi->atomic_write_cnt = 0; 2168 out: 2169 inode_unlock(inode); 2170 mnt_drop_write_file(filp); 2171 return ret; 2172 } 2173 2174 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2175 { 2176 struct inode *inode = file_inode(filp); 2177 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2178 int ret; 2179 2180 if (!inode_owner_or_capable(idmap, inode)) 2181 return -EACCES; 2182 2183 ret = mnt_want_write_file(filp); 2184 if (ret) 2185 return ret; 2186 2187 f2fs_balance_fs(F2FS_I_SB(inode), true); 2188 2189 inode_lock(inode); 2190 2191 if (f2fs_is_atomic_file(inode)) { 2192 ret = f2fs_commit_atomic_write(inode); 2193 if (!ret) 2194 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2195 2196 f2fs_abort_atomic_write(inode, ret); 2197 } else { 2198 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2199 } 2200 2201 inode_unlock(inode); 2202 mnt_drop_write_file(filp); 2203 return ret; 2204 } 2205 2206 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2207 { 2208 struct inode *inode = file_inode(filp); 2209 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2210 int ret; 2211 2212 if (!inode_owner_or_capable(idmap, inode)) 2213 return -EACCES; 2214 2215 ret = mnt_want_write_file(filp); 2216 if (ret) 2217 return ret; 2218 2219 inode_lock(inode); 2220 2221 f2fs_abort_atomic_write(inode, true); 2222 2223 inode_unlock(inode); 2224 2225 mnt_drop_write_file(filp); 2226 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2227 return ret; 2228 } 2229 2230 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag, 2231 bool readonly) 2232 { 2233 struct super_block *sb = sbi->sb; 2234 int ret = 0; 2235 2236 switch (flag) { 2237 case F2FS_GOING_DOWN_FULLSYNC: 2238 ret = bdev_freeze(sb->s_bdev); 2239 if (ret) 2240 goto out; 2241 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2242 bdev_thaw(sb->s_bdev); 2243 break; 2244 case F2FS_GOING_DOWN_METASYNC: 2245 /* do checkpoint only */ 2246 ret = f2fs_sync_fs(sb, 1); 2247 if (ret) { 2248 if (ret == -EIO) 2249 ret = 0; 2250 goto out; 2251 } 2252 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2253 break; 2254 case F2FS_GOING_DOWN_NOSYNC: 2255 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2256 break; 2257 case F2FS_GOING_DOWN_METAFLUSH: 2258 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2259 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2260 break; 2261 case F2FS_GOING_DOWN_NEED_FSCK: 2262 set_sbi_flag(sbi, SBI_NEED_FSCK); 2263 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2264 set_sbi_flag(sbi, SBI_IS_DIRTY); 2265 /* do checkpoint only */ 2266 ret = f2fs_sync_fs(sb, 1); 2267 if (ret == -EIO) 2268 ret = 0; 2269 goto out; 2270 default: 2271 ret = -EINVAL; 2272 goto out; 2273 } 2274 2275 if (readonly) 2276 goto out; 2277 2278 f2fs_stop_gc_thread(sbi); 2279 f2fs_stop_discard_thread(sbi); 2280 2281 f2fs_drop_discard_cmd(sbi); 2282 clear_opt(sbi, DISCARD); 2283 2284 f2fs_update_time(sbi, REQ_TIME); 2285 out: 2286 2287 trace_f2fs_shutdown(sbi, flag, ret); 2288 2289 return ret; 2290 } 2291 2292 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2293 { 2294 struct inode *inode = file_inode(filp); 2295 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2296 __u32 in; 2297 int ret; 2298 bool need_drop = false, readonly = false; 2299 2300 if (!capable(CAP_SYS_ADMIN)) 2301 return -EPERM; 2302 2303 if (get_user(in, (__u32 __user *)arg)) 2304 return -EFAULT; 2305 2306 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2307 ret = mnt_want_write_file(filp); 2308 if (ret) { 2309 if (ret != -EROFS) 2310 return ret; 2311 2312 /* fallback to nosync shutdown for readonly fs */ 2313 in = F2FS_GOING_DOWN_NOSYNC; 2314 readonly = true; 2315 } else { 2316 need_drop = true; 2317 } 2318 } 2319 2320 ret = f2fs_do_shutdown(sbi, in, readonly); 2321 2322 if (need_drop) 2323 mnt_drop_write_file(filp); 2324 2325 return ret; 2326 } 2327 2328 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2329 { 2330 struct inode *inode = file_inode(filp); 2331 struct super_block *sb = inode->i_sb; 2332 struct fstrim_range range; 2333 int ret; 2334 2335 if (!capable(CAP_SYS_ADMIN)) 2336 return -EPERM; 2337 2338 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2339 return -EOPNOTSUPP; 2340 2341 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2342 sizeof(range))) 2343 return -EFAULT; 2344 2345 ret = mnt_want_write_file(filp); 2346 if (ret) 2347 return ret; 2348 2349 range.minlen = max((unsigned int)range.minlen, 2350 bdev_discard_granularity(sb->s_bdev)); 2351 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2352 mnt_drop_write_file(filp); 2353 if (ret < 0) 2354 return ret; 2355 2356 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2357 sizeof(range))) 2358 return -EFAULT; 2359 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2360 return 0; 2361 } 2362 2363 static bool uuid_is_nonzero(__u8 u[16]) 2364 { 2365 int i; 2366 2367 for (i = 0; i < 16; i++) 2368 if (u[i]) 2369 return true; 2370 return false; 2371 } 2372 2373 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2374 { 2375 struct inode *inode = file_inode(filp); 2376 2377 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2378 return -EOPNOTSUPP; 2379 2380 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2381 2382 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2383 } 2384 2385 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2386 { 2387 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2388 return -EOPNOTSUPP; 2389 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2390 } 2391 2392 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2393 { 2394 struct inode *inode = file_inode(filp); 2395 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2396 u8 encrypt_pw_salt[16]; 2397 int err; 2398 2399 if (!f2fs_sb_has_encrypt(sbi)) 2400 return -EOPNOTSUPP; 2401 2402 err = mnt_want_write_file(filp); 2403 if (err) 2404 return err; 2405 2406 f2fs_down_write(&sbi->sb_lock); 2407 2408 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2409 goto got_it; 2410 2411 /* update superblock with uuid */ 2412 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2413 2414 err = f2fs_commit_super(sbi, false); 2415 if (err) { 2416 /* undo new data */ 2417 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2418 goto out_err; 2419 } 2420 got_it: 2421 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2422 out_err: 2423 f2fs_up_write(&sbi->sb_lock); 2424 mnt_drop_write_file(filp); 2425 2426 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2427 err = -EFAULT; 2428 2429 return err; 2430 } 2431 2432 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2433 unsigned long arg) 2434 { 2435 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2436 return -EOPNOTSUPP; 2437 2438 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2439 } 2440 2441 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2442 { 2443 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2444 return -EOPNOTSUPP; 2445 2446 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2447 } 2448 2449 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2450 { 2451 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2452 return -EOPNOTSUPP; 2453 2454 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2455 } 2456 2457 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2458 unsigned long arg) 2459 { 2460 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2461 return -EOPNOTSUPP; 2462 2463 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2464 } 2465 2466 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2467 unsigned long arg) 2468 { 2469 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2470 return -EOPNOTSUPP; 2471 2472 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2473 } 2474 2475 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2476 { 2477 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2478 return -EOPNOTSUPP; 2479 2480 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2481 } 2482 2483 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2484 { 2485 struct inode *inode = file_inode(filp); 2486 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2487 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2488 .no_bg_gc = false, 2489 .should_migrate_blocks = false, 2490 .nr_free_secs = 0 }; 2491 __u32 sync; 2492 int ret; 2493 2494 if (!capable(CAP_SYS_ADMIN)) 2495 return -EPERM; 2496 2497 if (get_user(sync, (__u32 __user *)arg)) 2498 return -EFAULT; 2499 2500 if (f2fs_readonly(sbi->sb)) 2501 return -EROFS; 2502 2503 ret = mnt_want_write_file(filp); 2504 if (ret) 2505 return ret; 2506 2507 if (!sync) { 2508 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2509 ret = -EBUSY; 2510 goto out; 2511 } 2512 } else { 2513 f2fs_down_write(&sbi->gc_lock); 2514 } 2515 2516 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2517 gc_control.err_gc_skipped = sync; 2518 stat_inc_gc_call_count(sbi, FOREGROUND); 2519 ret = f2fs_gc(sbi, &gc_control); 2520 out: 2521 mnt_drop_write_file(filp); 2522 return ret; 2523 } 2524 2525 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2526 { 2527 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2528 struct f2fs_gc_control gc_control = { 2529 .init_gc_type = range->sync ? FG_GC : BG_GC, 2530 .no_bg_gc = false, 2531 .should_migrate_blocks = false, 2532 .err_gc_skipped = range->sync, 2533 .nr_free_secs = 0 }; 2534 u64 end; 2535 int ret; 2536 2537 if (!capable(CAP_SYS_ADMIN)) 2538 return -EPERM; 2539 if (f2fs_readonly(sbi->sb)) 2540 return -EROFS; 2541 2542 end = range->start + range->len; 2543 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2544 end >= MAX_BLKADDR(sbi)) 2545 return -EINVAL; 2546 2547 ret = mnt_want_write_file(filp); 2548 if (ret) 2549 return ret; 2550 2551 do_more: 2552 if (!range->sync) { 2553 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2554 ret = -EBUSY; 2555 goto out; 2556 } 2557 } else { 2558 f2fs_down_write(&sbi->gc_lock); 2559 } 2560 2561 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2562 stat_inc_gc_call_count(sbi, FOREGROUND); 2563 ret = f2fs_gc(sbi, &gc_control); 2564 if (ret) { 2565 if (ret == -EBUSY) 2566 ret = -EAGAIN; 2567 goto out; 2568 } 2569 range->start += CAP_BLKS_PER_SEC(sbi); 2570 if (range->start <= end) 2571 goto do_more; 2572 out: 2573 mnt_drop_write_file(filp); 2574 return ret; 2575 } 2576 2577 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2578 { 2579 struct f2fs_gc_range range; 2580 2581 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2582 sizeof(range))) 2583 return -EFAULT; 2584 return __f2fs_ioc_gc_range(filp, &range); 2585 } 2586 2587 static int f2fs_ioc_write_checkpoint(struct file *filp) 2588 { 2589 struct inode *inode = file_inode(filp); 2590 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2591 int ret; 2592 2593 if (!capable(CAP_SYS_ADMIN)) 2594 return -EPERM; 2595 2596 if (f2fs_readonly(sbi->sb)) 2597 return -EROFS; 2598 2599 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2600 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2601 return -EINVAL; 2602 } 2603 2604 ret = mnt_want_write_file(filp); 2605 if (ret) 2606 return ret; 2607 2608 ret = f2fs_sync_fs(sbi->sb, 1); 2609 2610 mnt_drop_write_file(filp); 2611 return ret; 2612 } 2613 2614 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2615 struct file *filp, 2616 struct f2fs_defragment *range) 2617 { 2618 struct inode *inode = file_inode(filp); 2619 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2620 .m_seg_type = NO_CHECK_TYPE, 2621 .m_may_create = false }; 2622 struct extent_info ei = {}; 2623 pgoff_t pg_start, pg_end, next_pgofs; 2624 unsigned int total = 0, sec_num; 2625 block_t blk_end = 0; 2626 bool fragmented = false; 2627 int err; 2628 2629 f2fs_balance_fs(sbi, true); 2630 2631 inode_lock(inode); 2632 pg_start = range->start >> PAGE_SHIFT; 2633 pg_end = min_t(pgoff_t, 2634 (range->start + range->len) >> PAGE_SHIFT, 2635 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE)); 2636 2637 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2638 err = -EINVAL; 2639 goto unlock_out; 2640 } 2641 2642 /* if in-place-update policy is enabled, don't waste time here */ 2643 set_inode_flag(inode, FI_OPU_WRITE); 2644 if (f2fs_should_update_inplace(inode, NULL)) { 2645 err = -EINVAL; 2646 goto out; 2647 } 2648 2649 /* writeback all dirty pages in the range */ 2650 err = filemap_write_and_wait_range(inode->i_mapping, 2651 pg_start << PAGE_SHIFT, 2652 (pg_end << PAGE_SHIFT) - 1); 2653 if (err) 2654 goto out; 2655 2656 /* 2657 * lookup mapping info in extent cache, skip defragmenting if physical 2658 * block addresses are continuous. 2659 */ 2660 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2661 if (ei.fofs + ei.len >= pg_end) 2662 goto out; 2663 } 2664 2665 map.m_lblk = pg_start; 2666 map.m_next_pgofs = &next_pgofs; 2667 2668 /* 2669 * lookup mapping info in dnode page cache, skip defragmenting if all 2670 * physical block addresses are continuous even if there are hole(s) 2671 * in logical blocks. 2672 */ 2673 while (map.m_lblk < pg_end) { 2674 map.m_len = pg_end - map.m_lblk; 2675 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2676 if (err) 2677 goto out; 2678 2679 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2680 map.m_lblk = next_pgofs; 2681 continue; 2682 } 2683 2684 if (blk_end && blk_end != map.m_pblk) 2685 fragmented = true; 2686 2687 /* record total count of block that we're going to move */ 2688 total += map.m_len; 2689 2690 blk_end = map.m_pblk + map.m_len; 2691 2692 map.m_lblk += map.m_len; 2693 } 2694 2695 if (!fragmented) { 2696 total = 0; 2697 goto out; 2698 } 2699 2700 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2701 2702 /* 2703 * make sure there are enough free section for LFS allocation, this can 2704 * avoid defragment running in SSR mode when free section are allocated 2705 * intensively 2706 */ 2707 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2708 err = -EAGAIN; 2709 goto out; 2710 } 2711 2712 map.m_lblk = pg_start; 2713 map.m_len = pg_end - pg_start; 2714 total = 0; 2715 2716 while (map.m_lblk < pg_end) { 2717 pgoff_t idx; 2718 int cnt = 0; 2719 2720 do_map: 2721 map.m_len = pg_end - map.m_lblk; 2722 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2723 if (err) 2724 goto clear_out; 2725 2726 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2727 map.m_lblk = next_pgofs; 2728 goto check; 2729 } 2730 2731 set_inode_flag(inode, FI_SKIP_WRITES); 2732 2733 idx = map.m_lblk; 2734 while (idx < map.m_lblk + map.m_len && 2735 cnt < BLKS_PER_SEG(sbi)) { 2736 struct page *page; 2737 2738 page = f2fs_get_lock_data_page(inode, idx, true); 2739 if (IS_ERR(page)) { 2740 err = PTR_ERR(page); 2741 goto clear_out; 2742 } 2743 2744 set_page_dirty(page); 2745 set_page_private_gcing(page); 2746 f2fs_put_page(page, 1); 2747 2748 idx++; 2749 cnt++; 2750 total++; 2751 } 2752 2753 map.m_lblk = idx; 2754 check: 2755 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 2756 goto do_map; 2757 2758 clear_inode_flag(inode, FI_SKIP_WRITES); 2759 2760 err = filemap_fdatawrite(inode->i_mapping); 2761 if (err) 2762 goto out; 2763 } 2764 clear_out: 2765 clear_inode_flag(inode, FI_SKIP_WRITES); 2766 out: 2767 clear_inode_flag(inode, FI_OPU_WRITE); 2768 unlock_out: 2769 inode_unlock(inode); 2770 if (!err) 2771 range->len = (u64)total << PAGE_SHIFT; 2772 return err; 2773 } 2774 2775 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2776 { 2777 struct inode *inode = file_inode(filp); 2778 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2779 struct f2fs_defragment range; 2780 int err; 2781 2782 if (!capable(CAP_SYS_ADMIN)) 2783 return -EPERM; 2784 2785 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2786 return -EINVAL; 2787 2788 if (f2fs_readonly(sbi->sb)) 2789 return -EROFS; 2790 2791 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2792 sizeof(range))) 2793 return -EFAULT; 2794 2795 /* verify alignment of offset & size */ 2796 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2797 return -EINVAL; 2798 2799 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2800 max_file_blocks(inode))) 2801 return -EINVAL; 2802 2803 err = mnt_want_write_file(filp); 2804 if (err) 2805 return err; 2806 2807 err = f2fs_defragment_range(sbi, filp, &range); 2808 mnt_drop_write_file(filp); 2809 2810 f2fs_update_time(sbi, REQ_TIME); 2811 if (err < 0) 2812 return err; 2813 2814 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2815 sizeof(range))) 2816 return -EFAULT; 2817 2818 return 0; 2819 } 2820 2821 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2822 struct file *file_out, loff_t pos_out, size_t len) 2823 { 2824 struct inode *src = file_inode(file_in); 2825 struct inode *dst = file_inode(file_out); 2826 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2827 size_t olen = len, dst_max_i_size = 0; 2828 size_t dst_osize; 2829 int ret; 2830 2831 if (file_in->f_path.mnt != file_out->f_path.mnt || 2832 src->i_sb != dst->i_sb) 2833 return -EXDEV; 2834 2835 if (unlikely(f2fs_readonly(src->i_sb))) 2836 return -EROFS; 2837 2838 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2839 return -EINVAL; 2840 2841 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2842 return -EOPNOTSUPP; 2843 2844 if (pos_out < 0 || pos_in < 0) 2845 return -EINVAL; 2846 2847 if (src == dst) { 2848 if (pos_in == pos_out) 2849 return 0; 2850 if (pos_out > pos_in && pos_out < pos_in + len) 2851 return -EINVAL; 2852 } 2853 2854 inode_lock(src); 2855 if (src != dst) { 2856 ret = -EBUSY; 2857 if (!inode_trylock(dst)) 2858 goto out; 2859 } 2860 2861 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) { 2862 ret = -EOPNOTSUPP; 2863 goto out_unlock; 2864 } 2865 2866 ret = -EINVAL; 2867 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2868 goto out_unlock; 2869 if (len == 0) 2870 olen = len = src->i_size - pos_in; 2871 if (pos_in + len == src->i_size) 2872 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2873 if (len == 0) { 2874 ret = 0; 2875 goto out_unlock; 2876 } 2877 2878 dst_osize = dst->i_size; 2879 if (pos_out + olen > dst->i_size) 2880 dst_max_i_size = pos_out + olen; 2881 2882 /* verify the end result is block aligned */ 2883 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2884 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2885 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2886 goto out_unlock; 2887 2888 ret = f2fs_convert_inline_inode(src); 2889 if (ret) 2890 goto out_unlock; 2891 2892 ret = f2fs_convert_inline_inode(dst); 2893 if (ret) 2894 goto out_unlock; 2895 2896 /* write out all dirty pages from offset */ 2897 ret = filemap_write_and_wait_range(src->i_mapping, 2898 pos_in, pos_in + len); 2899 if (ret) 2900 goto out_unlock; 2901 2902 ret = filemap_write_and_wait_range(dst->i_mapping, 2903 pos_out, pos_out + len); 2904 if (ret) 2905 goto out_unlock; 2906 2907 f2fs_balance_fs(sbi, true); 2908 2909 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2910 if (src != dst) { 2911 ret = -EBUSY; 2912 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2913 goto out_src; 2914 } 2915 2916 f2fs_lock_op(sbi); 2917 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2918 pos_out >> F2FS_BLKSIZE_BITS, 2919 len >> F2FS_BLKSIZE_BITS, false); 2920 2921 if (!ret) { 2922 if (dst_max_i_size) 2923 f2fs_i_size_write(dst, dst_max_i_size); 2924 else if (dst_osize != dst->i_size) 2925 f2fs_i_size_write(dst, dst_osize); 2926 } 2927 f2fs_unlock_op(sbi); 2928 2929 if (src != dst) 2930 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2931 out_src: 2932 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2933 if (ret) 2934 goto out_unlock; 2935 2936 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 2937 f2fs_mark_inode_dirty_sync(src, false); 2938 if (src != dst) { 2939 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 2940 f2fs_mark_inode_dirty_sync(dst, false); 2941 } 2942 f2fs_update_time(sbi, REQ_TIME); 2943 2944 out_unlock: 2945 if (src != dst) 2946 inode_unlock(dst); 2947 out: 2948 inode_unlock(src); 2949 return ret; 2950 } 2951 2952 static int __f2fs_ioc_move_range(struct file *filp, 2953 struct f2fs_move_range *range) 2954 { 2955 struct fd dst; 2956 int err; 2957 2958 if (!(filp->f_mode & FMODE_READ) || 2959 !(filp->f_mode & FMODE_WRITE)) 2960 return -EBADF; 2961 2962 dst = fdget(range->dst_fd); 2963 if (!dst.file) 2964 return -EBADF; 2965 2966 if (!(dst.file->f_mode & FMODE_WRITE)) { 2967 err = -EBADF; 2968 goto err_out; 2969 } 2970 2971 err = mnt_want_write_file(filp); 2972 if (err) 2973 goto err_out; 2974 2975 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2976 range->pos_out, range->len); 2977 2978 mnt_drop_write_file(filp); 2979 err_out: 2980 fdput(dst); 2981 return err; 2982 } 2983 2984 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2985 { 2986 struct f2fs_move_range range; 2987 2988 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2989 sizeof(range))) 2990 return -EFAULT; 2991 return __f2fs_ioc_move_range(filp, &range); 2992 } 2993 2994 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2995 { 2996 struct inode *inode = file_inode(filp); 2997 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2998 struct sit_info *sm = SIT_I(sbi); 2999 unsigned int start_segno = 0, end_segno = 0; 3000 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3001 struct f2fs_flush_device range; 3002 struct f2fs_gc_control gc_control = { 3003 .init_gc_type = FG_GC, 3004 .should_migrate_blocks = true, 3005 .err_gc_skipped = true, 3006 .nr_free_secs = 0 }; 3007 int ret; 3008 3009 if (!capable(CAP_SYS_ADMIN)) 3010 return -EPERM; 3011 3012 if (f2fs_readonly(sbi->sb)) 3013 return -EROFS; 3014 3015 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3016 return -EINVAL; 3017 3018 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3019 sizeof(range))) 3020 return -EFAULT; 3021 3022 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3023 __is_large_section(sbi)) { 3024 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3025 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3026 return -EINVAL; 3027 } 3028 3029 ret = mnt_want_write_file(filp); 3030 if (ret) 3031 return ret; 3032 3033 if (range.dev_num != 0) 3034 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3035 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3036 3037 start_segno = sm->last_victim[FLUSH_DEVICE]; 3038 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3039 start_segno = dev_start_segno; 3040 end_segno = min(start_segno + range.segments, dev_end_segno); 3041 3042 while (start_segno < end_segno) { 3043 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3044 ret = -EBUSY; 3045 goto out; 3046 } 3047 sm->last_victim[GC_CB] = end_segno + 1; 3048 sm->last_victim[GC_GREEDY] = end_segno + 1; 3049 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3050 3051 gc_control.victim_segno = start_segno; 3052 stat_inc_gc_call_count(sbi, FOREGROUND); 3053 ret = f2fs_gc(sbi, &gc_control); 3054 if (ret == -EAGAIN) 3055 ret = 0; 3056 else if (ret < 0) 3057 break; 3058 start_segno++; 3059 } 3060 out: 3061 mnt_drop_write_file(filp); 3062 return ret; 3063 } 3064 3065 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3066 { 3067 struct inode *inode = file_inode(filp); 3068 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3069 3070 /* Must validate to set it with SQLite behavior in Android. */ 3071 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3072 3073 return put_user(sb_feature, (u32 __user *)arg); 3074 } 3075 3076 #ifdef CONFIG_QUOTA 3077 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3078 { 3079 struct dquot *transfer_to[MAXQUOTAS] = {}; 3080 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3081 struct super_block *sb = sbi->sb; 3082 int err; 3083 3084 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3085 if (IS_ERR(transfer_to[PRJQUOTA])) 3086 return PTR_ERR(transfer_to[PRJQUOTA]); 3087 3088 err = __dquot_transfer(inode, transfer_to); 3089 if (err) 3090 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3091 dqput(transfer_to[PRJQUOTA]); 3092 return err; 3093 } 3094 3095 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3096 { 3097 struct f2fs_inode_info *fi = F2FS_I(inode); 3098 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3099 struct f2fs_inode *ri = NULL; 3100 kprojid_t kprojid; 3101 int err; 3102 3103 if (!f2fs_sb_has_project_quota(sbi)) { 3104 if (projid != F2FS_DEF_PROJID) 3105 return -EOPNOTSUPP; 3106 else 3107 return 0; 3108 } 3109 3110 if (!f2fs_has_extra_attr(inode)) 3111 return -EOPNOTSUPP; 3112 3113 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3114 3115 if (projid_eq(kprojid, fi->i_projid)) 3116 return 0; 3117 3118 err = -EPERM; 3119 /* Is it quota file? Do not allow user to mess with it */ 3120 if (IS_NOQUOTA(inode)) 3121 return err; 3122 3123 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3124 return -EOVERFLOW; 3125 3126 err = f2fs_dquot_initialize(inode); 3127 if (err) 3128 return err; 3129 3130 f2fs_lock_op(sbi); 3131 err = f2fs_transfer_project_quota(inode, kprojid); 3132 if (err) 3133 goto out_unlock; 3134 3135 fi->i_projid = kprojid; 3136 inode_set_ctime_current(inode); 3137 f2fs_mark_inode_dirty_sync(inode, true); 3138 out_unlock: 3139 f2fs_unlock_op(sbi); 3140 return err; 3141 } 3142 #else 3143 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3144 { 3145 return 0; 3146 } 3147 3148 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3149 { 3150 if (projid != F2FS_DEF_PROJID) 3151 return -EOPNOTSUPP; 3152 return 0; 3153 } 3154 #endif 3155 3156 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3157 { 3158 struct inode *inode = d_inode(dentry); 3159 struct f2fs_inode_info *fi = F2FS_I(inode); 3160 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3161 3162 if (IS_ENCRYPTED(inode)) 3163 fsflags |= FS_ENCRYPT_FL; 3164 if (IS_VERITY(inode)) 3165 fsflags |= FS_VERITY_FL; 3166 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3167 fsflags |= FS_INLINE_DATA_FL; 3168 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3169 fsflags |= FS_NOCOW_FL; 3170 3171 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3172 3173 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3174 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3175 3176 return 0; 3177 } 3178 3179 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3180 struct dentry *dentry, struct fileattr *fa) 3181 { 3182 struct inode *inode = d_inode(dentry); 3183 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3184 u32 iflags; 3185 int err; 3186 3187 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3188 return -EIO; 3189 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3190 return -ENOSPC; 3191 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3192 return -EOPNOTSUPP; 3193 fsflags &= F2FS_SETTABLE_FS_FL; 3194 if (!fa->flags_valid) 3195 mask &= FS_COMMON_FL; 3196 3197 iflags = f2fs_fsflags_to_iflags(fsflags); 3198 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3199 return -EOPNOTSUPP; 3200 3201 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3202 if (!err) 3203 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3204 3205 return err; 3206 } 3207 3208 int f2fs_pin_file_control(struct inode *inode, bool inc) 3209 { 3210 struct f2fs_inode_info *fi = F2FS_I(inode); 3211 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3212 3213 /* Use i_gc_failures for normal file as a risk signal. */ 3214 if (inc) 3215 f2fs_i_gc_failures_write(inode, 3216 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3217 3218 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3219 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3220 __func__, inode->i_ino, 3221 fi->i_gc_failures[GC_FAILURE_PIN]); 3222 clear_inode_flag(inode, FI_PIN_FILE); 3223 return -EAGAIN; 3224 } 3225 return 0; 3226 } 3227 3228 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3229 { 3230 struct inode *inode = file_inode(filp); 3231 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3232 __u32 pin; 3233 int ret = 0; 3234 3235 if (get_user(pin, (__u32 __user *)arg)) 3236 return -EFAULT; 3237 3238 if (!S_ISREG(inode->i_mode)) 3239 return -EINVAL; 3240 3241 if (f2fs_readonly(sbi->sb)) 3242 return -EROFS; 3243 3244 ret = mnt_want_write_file(filp); 3245 if (ret) 3246 return ret; 3247 3248 inode_lock(inode); 3249 3250 if (!pin) { 3251 clear_inode_flag(inode, FI_PIN_FILE); 3252 f2fs_i_gc_failures_write(inode, 0); 3253 goto done; 3254 } else if (f2fs_is_pinned_file(inode)) { 3255 goto done; 3256 } 3257 3258 if (f2fs_sb_has_blkzoned(sbi) && F2FS_HAS_BLOCKS(inode)) { 3259 ret = -EFBIG; 3260 goto out; 3261 } 3262 3263 /* Let's allow file pinning on zoned device. */ 3264 if (!f2fs_sb_has_blkzoned(sbi) && 3265 f2fs_should_update_outplace(inode, NULL)) { 3266 ret = -EINVAL; 3267 goto out; 3268 } 3269 3270 if (f2fs_pin_file_control(inode, false)) { 3271 ret = -EAGAIN; 3272 goto out; 3273 } 3274 3275 ret = f2fs_convert_inline_inode(inode); 3276 if (ret) 3277 goto out; 3278 3279 if (!f2fs_disable_compressed_file(inode)) { 3280 ret = -EOPNOTSUPP; 3281 goto out; 3282 } 3283 3284 set_inode_flag(inode, FI_PIN_FILE); 3285 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3286 done: 3287 f2fs_update_time(sbi, REQ_TIME); 3288 out: 3289 inode_unlock(inode); 3290 mnt_drop_write_file(filp); 3291 return ret; 3292 } 3293 3294 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3295 { 3296 struct inode *inode = file_inode(filp); 3297 __u32 pin = 0; 3298 3299 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3300 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3301 return put_user(pin, (u32 __user *)arg); 3302 } 3303 3304 int f2fs_precache_extents(struct inode *inode) 3305 { 3306 struct f2fs_inode_info *fi = F2FS_I(inode); 3307 struct f2fs_map_blocks map; 3308 pgoff_t m_next_extent; 3309 loff_t end; 3310 int err; 3311 3312 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3313 return -EOPNOTSUPP; 3314 3315 map.m_lblk = 0; 3316 map.m_pblk = 0; 3317 map.m_next_pgofs = NULL; 3318 map.m_next_extent = &m_next_extent; 3319 map.m_seg_type = NO_CHECK_TYPE; 3320 map.m_may_create = false; 3321 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3322 3323 while (map.m_lblk < end) { 3324 map.m_len = end - map.m_lblk; 3325 3326 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3327 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3328 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3329 if (err || !map.m_len) 3330 return err; 3331 3332 map.m_lblk = m_next_extent; 3333 } 3334 3335 return 0; 3336 } 3337 3338 static int f2fs_ioc_precache_extents(struct file *filp) 3339 { 3340 return f2fs_precache_extents(file_inode(filp)); 3341 } 3342 3343 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3344 { 3345 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3346 __u64 block_count; 3347 3348 if (!capable(CAP_SYS_ADMIN)) 3349 return -EPERM; 3350 3351 if (f2fs_readonly(sbi->sb)) 3352 return -EROFS; 3353 3354 if (copy_from_user(&block_count, (void __user *)arg, 3355 sizeof(block_count))) 3356 return -EFAULT; 3357 3358 return f2fs_resize_fs(filp, block_count); 3359 } 3360 3361 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3362 { 3363 struct inode *inode = file_inode(filp); 3364 3365 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3366 3367 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3368 f2fs_warn(F2FS_I_SB(inode), 3369 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3370 inode->i_ino); 3371 return -EOPNOTSUPP; 3372 } 3373 3374 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3375 } 3376 3377 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3378 { 3379 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3380 return -EOPNOTSUPP; 3381 3382 return fsverity_ioctl_measure(filp, (void __user *)arg); 3383 } 3384 3385 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3386 { 3387 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3388 return -EOPNOTSUPP; 3389 3390 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3391 } 3392 3393 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3394 { 3395 struct inode *inode = file_inode(filp); 3396 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3397 char *vbuf; 3398 int count; 3399 int err = 0; 3400 3401 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3402 if (!vbuf) 3403 return -ENOMEM; 3404 3405 f2fs_down_read(&sbi->sb_lock); 3406 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3407 ARRAY_SIZE(sbi->raw_super->volume_name), 3408 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3409 f2fs_up_read(&sbi->sb_lock); 3410 3411 if (copy_to_user((char __user *)arg, vbuf, 3412 min(FSLABEL_MAX, count))) 3413 err = -EFAULT; 3414 3415 kfree(vbuf); 3416 return err; 3417 } 3418 3419 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3420 { 3421 struct inode *inode = file_inode(filp); 3422 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3423 char *vbuf; 3424 int err = 0; 3425 3426 if (!capable(CAP_SYS_ADMIN)) 3427 return -EPERM; 3428 3429 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3430 if (IS_ERR(vbuf)) 3431 return PTR_ERR(vbuf); 3432 3433 err = mnt_want_write_file(filp); 3434 if (err) 3435 goto out; 3436 3437 f2fs_down_write(&sbi->sb_lock); 3438 3439 memset(sbi->raw_super->volume_name, 0, 3440 sizeof(sbi->raw_super->volume_name)); 3441 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3442 sbi->raw_super->volume_name, 3443 ARRAY_SIZE(sbi->raw_super->volume_name)); 3444 3445 err = f2fs_commit_super(sbi, false); 3446 3447 f2fs_up_write(&sbi->sb_lock); 3448 3449 mnt_drop_write_file(filp); 3450 out: 3451 kfree(vbuf); 3452 return err; 3453 } 3454 3455 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3456 { 3457 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3458 return -EOPNOTSUPP; 3459 3460 if (!f2fs_compressed_file(inode)) 3461 return -EINVAL; 3462 3463 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3464 3465 return 0; 3466 } 3467 3468 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3469 { 3470 struct inode *inode = file_inode(filp); 3471 __u64 blocks; 3472 int ret; 3473 3474 ret = f2fs_get_compress_blocks(inode, &blocks); 3475 if (ret < 0) 3476 return ret; 3477 3478 return put_user(blocks, (u64 __user *)arg); 3479 } 3480 3481 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3482 { 3483 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3484 unsigned int released_blocks = 0; 3485 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3486 block_t blkaddr; 3487 int i; 3488 3489 for (i = 0; i < count; i++) { 3490 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3491 dn->ofs_in_node + i); 3492 3493 if (!__is_valid_data_blkaddr(blkaddr)) 3494 continue; 3495 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3496 DATA_GENERIC_ENHANCE))) 3497 return -EFSCORRUPTED; 3498 } 3499 3500 while (count) { 3501 int compr_blocks = 0; 3502 3503 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3504 blkaddr = f2fs_data_blkaddr(dn); 3505 3506 if (i == 0) { 3507 if (blkaddr == COMPRESS_ADDR) 3508 continue; 3509 dn->ofs_in_node += cluster_size; 3510 goto next; 3511 } 3512 3513 if (__is_valid_data_blkaddr(blkaddr)) 3514 compr_blocks++; 3515 3516 if (blkaddr != NEW_ADDR) 3517 continue; 3518 3519 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3520 } 3521 3522 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3523 dec_valid_block_count(sbi, dn->inode, 3524 cluster_size - compr_blocks); 3525 3526 released_blocks += cluster_size - compr_blocks; 3527 next: 3528 count -= cluster_size; 3529 } 3530 3531 return released_blocks; 3532 } 3533 3534 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3535 { 3536 struct inode *inode = file_inode(filp); 3537 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3538 pgoff_t page_idx = 0, last_idx; 3539 unsigned int released_blocks = 0; 3540 int ret; 3541 int writecount; 3542 3543 if (!f2fs_sb_has_compression(sbi)) 3544 return -EOPNOTSUPP; 3545 3546 if (f2fs_readonly(sbi->sb)) 3547 return -EROFS; 3548 3549 ret = mnt_want_write_file(filp); 3550 if (ret) 3551 return ret; 3552 3553 f2fs_balance_fs(sbi, true); 3554 3555 inode_lock(inode); 3556 3557 writecount = atomic_read(&inode->i_writecount); 3558 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3559 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3560 ret = -EBUSY; 3561 goto out; 3562 } 3563 3564 if (!f2fs_compressed_file(inode) || 3565 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3566 ret = -EINVAL; 3567 goto out; 3568 } 3569 3570 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3571 if (ret) 3572 goto out; 3573 3574 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3575 ret = -EPERM; 3576 goto out; 3577 } 3578 3579 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3580 inode_set_ctime_current(inode); 3581 f2fs_mark_inode_dirty_sync(inode, true); 3582 3583 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3584 filemap_invalidate_lock(inode->i_mapping); 3585 3586 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3587 3588 while (page_idx < last_idx) { 3589 struct dnode_of_data dn; 3590 pgoff_t end_offset, count; 3591 3592 set_new_dnode(&dn, inode, NULL, NULL, 0); 3593 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3594 if (ret) { 3595 if (ret == -ENOENT) { 3596 page_idx = f2fs_get_next_page_offset(&dn, 3597 page_idx); 3598 ret = 0; 3599 continue; 3600 } 3601 break; 3602 } 3603 3604 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3605 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3606 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3607 3608 ret = release_compress_blocks(&dn, count); 3609 3610 f2fs_put_dnode(&dn); 3611 3612 if (ret < 0) 3613 break; 3614 3615 page_idx += count; 3616 released_blocks += ret; 3617 } 3618 3619 filemap_invalidate_unlock(inode->i_mapping); 3620 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3621 out: 3622 inode_unlock(inode); 3623 3624 mnt_drop_write_file(filp); 3625 3626 if (ret >= 0) { 3627 ret = put_user(released_blocks, (u64 __user *)arg); 3628 } else if (released_blocks && 3629 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3630 set_sbi_flag(sbi, SBI_NEED_FSCK); 3631 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3632 "iblocks=%llu, released=%u, compr_blocks=%u, " 3633 "run fsck to fix.", 3634 __func__, inode->i_ino, inode->i_blocks, 3635 released_blocks, 3636 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3637 } 3638 3639 return ret; 3640 } 3641 3642 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3643 unsigned int *reserved_blocks) 3644 { 3645 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3646 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3647 block_t blkaddr; 3648 int i; 3649 3650 for (i = 0; i < count; i++) { 3651 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3652 dn->ofs_in_node + i); 3653 3654 if (!__is_valid_data_blkaddr(blkaddr)) 3655 continue; 3656 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3657 DATA_GENERIC_ENHANCE))) 3658 return -EFSCORRUPTED; 3659 } 3660 3661 while (count) { 3662 int compr_blocks = 0; 3663 blkcnt_t reserved; 3664 int ret; 3665 3666 for (i = 0; i < cluster_size; i++) { 3667 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3668 dn->ofs_in_node + i); 3669 3670 if (i == 0) { 3671 if (blkaddr != COMPRESS_ADDR) { 3672 dn->ofs_in_node += cluster_size; 3673 goto next; 3674 } 3675 continue; 3676 } 3677 3678 /* 3679 * compressed cluster was not released due to it 3680 * fails in release_compress_blocks(), so NEW_ADDR 3681 * is a possible case. 3682 */ 3683 if (blkaddr == NEW_ADDR || 3684 __is_valid_data_blkaddr(blkaddr)) { 3685 compr_blocks++; 3686 continue; 3687 } 3688 } 3689 3690 reserved = cluster_size - compr_blocks; 3691 3692 /* for the case all blocks in cluster were reserved */ 3693 if (reserved == 1) 3694 goto next; 3695 3696 ret = inc_valid_block_count(sbi, dn->inode, &reserved, false); 3697 if (unlikely(ret)) 3698 return ret; 3699 3700 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3701 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3702 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3703 } 3704 3705 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3706 3707 *reserved_blocks += reserved; 3708 next: 3709 count -= cluster_size; 3710 } 3711 3712 return 0; 3713 } 3714 3715 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3716 { 3717 struct inode *inode = file_inode(filp); 3718 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3719 pgoff_t page_idx = 0, last_idx; 3720 unsigned int reserved_blocks = 0; 3721 int ret; 3722 3723 if (!f2fs_sb_has_compression(sbi)) 3724 return -EOPNOTSUPP; 3725 3726 if (f2fs_readonly(sbi->sb)) 3727 return -EROFS; 3728 3729 ret = mnt_want_write_file(filp); 3730 if (ret) 3731 return ret; 3732 3733 f2fs_balance_fs(sbi, true); 3734 3735 inode_lock(inode); 3736 3737 if (!f2fs_compressed_file(inode) || 3738 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3739 ret = -EINVAL; 3740 goto unlock_inode; 3741 } 3742 3743 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3744 goto unlock_inode; 3745 3746 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3747 filemap_invalidate_lock(inode->i_mapping); 3748 3749 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3750 3751 while (page_idx < last_idx) { 3752 struct dnode_of_data dn; 3753 pgoff_t end_offset, count; 3754 3755 set_new_dnode(&dn, inode, NULL, NULL, 0); 3756 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3757 if (ret) { 3758 if (ret == -ENOENT) { 3759 page_idx = f2fs_get_next_page_offset(&dn, 3760 page_idx); 3761 ret = 0; 3762 continue; 3763 } 3764 break; 3765 } 3766 3767 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3768 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3769 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3770 3771 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 3772 3773 f2fs_put_dnode(&dn); 3774 3775 if (ret < 0) 3776 break; 3777 3778 page_idx += count; 3779 } 3780 3781 filemap_invalidate_unlock(inode->i_mapping); 3782 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3783 3784 if (!ret) { 3785 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3786 inode_set_ctime_current(inode); 3787 f2fs_mark_inode_dirty_sync(inode, true); 3788 } 3789 unlock_inode: 3790 inode_unlock(inode); 3791 mnt_drop_write_file(filp); 3792 3793 if (!ret) { 3794 ret = put_user(reserved_blocks, (u64 __user *)arg); 3795 } else if (reserved_blocks && 3796 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3797 set_sbi_flag(sbi, SBI_NEED_FSCK); 3798 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3799 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3800 "run fsck to fix.", 3801 __func__, inode->i_ino, inode->i_blocks, 3802 reserved_blocks, 3803 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3804 } 3805 3806 return ret; 3807 } 3808 3809 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3810 pgoff_t off, block_t block, block_t len, u32 flags) 3811 { 3812 sector_t sector = SECTOR_FROM_BLOCK(block); 3813 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3814 int ret = 0; 3815 3816 if (flags & F2FS_TRIM_FILE_DISCARD) { 3817 if (bdev_max_secure_erase_sectors(bdev)) 3818 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3819 GFP_NOFS); 3820 else 3821 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3822 GFP_NOFS); 3823 } 3824 3825 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3826 if (IS_ENCRYPTED(inode)) 3827 ret = fscrypt_zeroout_range(inode, off, block, len); 3828 else 3829 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3830 GFP_NOFS, 0); 3831 } 3832 3833 return ret; 3834 } 3835 3836 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3837 { 3838 struct inode *inode = file_inode(filp); 3839 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3840 struct address_space *mapping = inode->i_mapping; 3841 struct block_device *prev_bdev = NULL; 3842 struct f2fs_sectrim_range range; 3843 pgoff_t index, pg_end, prev_index = 0; 3844 block_t prev_block = 0, len = 0; 3845 loff_t end_addr; 3846 bool to_end = false; 3847 int ret = 0; 3848 3849 if (!(filp->f_mode & FMODE_WRITE)) 3850 return -EBADF; 3851 3852 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3853 sizeof(range))) 3854 return -EFAULT; 3855 3856 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3857 !S_ISREG(inode->i_mode)) 3858 return -EINVAL; 3859 3860 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3861 !f2fs_hw_support_discard(sbi)) || 3862 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3863 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3864 return -EOPNOTSUPP; 3865 3866 file_start_write(filp); 3867 inode_lock(inode); 3868 3869 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3870 range.start >= inode->i_size) { 3871 ret = -EINVAL; 3872 goto err; 3873 } 3874 3875 if (range.len == 0) 3876 goto err; 3877 3878 if (inode->i_size - range.start > range.len) { 3879 end_addr = range.start + range.len; 3880 } else { 3881 end_addr = range.len == (u64)-1 ? 3882 sbi->sb->s_maxbytes : inode->i_size; 3883 to_end = true; 3884 } 3885 3886 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3887 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3888 ret = -EINVAL; 3889 goto err; 3890 } 3891 3892 index = F2FS_BYTES_TO_BLK(range.start); 3893 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3894 3895 ret = f2fs_convert_inline_inode(inode); 3896 if (ret) 3897 goto err; 3898 3899 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3900 filemap_invalidate_lock(mapping); 3901 3902 ret = filemap_write_and_wait_range(mapping, range.start, 3903 to_end ? LLONG_MAX : end_addr - 1); 3904 if (ret) 3905 goto out; 3906 3907 truncate_inode_pages_range(mapping, range.start, 3908 to_end ? -1 : end_addr - 1); 3909 3910 while (index < pg_end) { 3911 struct dnode_of_data dn; 3912 pgoff_t end_offset, count; 3913 int i; 3914 3915 set_new_dnode(&dn, inode, NULL, NULL, 0); 3916 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3917 if (ret) { 3918 if (ret == -ENOENT) { 3919 index = f2fs_get_next_page_offset(&dn, index); 3920 continue; 3921 } 3922 goto out; 3923 } 3924 3925 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3926 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3927 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3928 struct block_device *cur_bdev; 3929 block_t blkaddr = f2fs_data_blkaddr(&dn); 3930 3931 if (!__is_valid_data_blkaddr(blkaddr)) 3932 continue; 3933 3934 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3935 DATA_GENERIC_ENHANCE)) { 3936 ret = -EFSCORRUPTED; 3937 f2fs_put_dnode(&dn); 3938 goto out; 3939 } 3940 3941 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3942 if (f2fs_is_multi_device(sbi)) { 3943 int di = f2fs_target_device_index(sbi, blkaddr); 3944 3945 blkaddr -= FDEV(di).start_blk; 3946 } 3947 3948 if (len) { 3949 if (prev_bdev == cur_bdev && 3950 index == prev_index + len && 3951 blkaddr == prev_block + len) { 3952 len++; 3953 } else { 3954 ret = f2fs_secure_erase(prev_bdev, 3955 inode, prev_index, prev_block, 3956 len, range.flags); 3957 if (ret) { 3958 f2fs_put_dnode(&dn); 3959 goto out; 3960 } 3961 3962 len = 0; 3963 } 3964 } 3965 3966 if (!len) { 3967 prev_bdev = cur_bdev; 3968 prev_index = index; 3969 prev_block = blkaddr; 3970 len = 1; 3971 } 3972 } 3973 3974 f2fs_put_dnode(&dn); 3975 3976 if (fatal_signal_pending(current)) { 3977 ret = -EINTR; 3978 goto out; 3979 } 3980 cond_resched(); 3981 } 3982 3983 if (len) 3984 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3985 prev_block, len, range.flags); 3986 out: 3987 filemap_invalidate_unlock(mapping); 3988 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3989 err: 3990 inode_unlock(inode); 3991 file_end_write(filp); 3992 3993 return ret; 3994 } 3995 3996 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3997 { 3998 struct inode *inode = file_inode(filp); 3999 struct f2fs_comp_option option; 4000 4001 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4002 return -EOPNOTSUPP; 4003 4004 inode_lock_shared(inode); 4005 4006 if (!f2fs_compressed_file(inode)) { 4007 inode_unlock_shared(inode); 4008 return -ENODATA; 4009 } 4010 4011 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4012 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4013 4014 inode_unlock_shared(inode); 4015 4016 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4017 sizeof(option))) 4018 return -EFAULT; 4019 4020 return 0; 4021 } 4022 4023 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4024 { 4025 struct inode *inode = file_inode(filp); 4026 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4027 struct f2fs_comp_option option; 4028 int ret = 0; 4029 4030 if (!f2fs_sb_has_compression(sbi)) 4031 return -EOPNOTSUPP; 4032 4033 if (!(filp->f_mode & FMODE_WRITE)) 4034 return -EBADF; 4035 4036 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4037 sizeof(option))) 4038 return -EFAULT; 4039 4040 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4041 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4042 option.algorithm >= COMPRESS_MAX) 4043 return -EINVAL; 4044 4045 file_start_write(filp); 4046 inode_lock(inode); 4047 4048 f2fs_down_write(&F2FS_I(inode)->i_sem); 4049 if (!f2fs_compressed_file(inode)) { 4050 ret = -EINVAL; 4051 goto out; 4052 } 4053 4054 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4055 ret = -EBUSY; 4056 goto out; 4057 } 4058 4059 if (F2FS_HAS_BLOCKS(inode)) { 4060 ret = -EFBIG; 4061 goto out; 4062 } 4063 4064 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4065 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4066 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4067 /* Set default level */ 4068 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4069 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4070 else 4071 F2FS_I(inode)->i_compress_level = 0; 4072 /* Adjust mount option level */ 4073 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4074 F2FS_OPTION(sbi).compress_level) 4075 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4076 f2fs_mark_inode_dirty_sync(inode, true); 4077 4078 if (!f2fs_is_compress_backend_ready(inode)) 4079 f2fs_warn(sbi, "compression algorithm is successfully set, " 4080 "but current kernel doesn't support this algorithm."); 4081 out: 4082 f2fs_up_write(&F2FS_I(inode)->i_sem); 4083 inode_unlock(inode); 4084 file_end_write(filp); 4085 4086 return ret; 4087 } 4088 4089 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4090 { 4091 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4092 struct address_space *mapping = inode->i_mapping; 4093 struct page *page; 4094 pgoff_t redirty_idx = page_idx; 4095 int i, page_len = 0, ret = 0; 4096 4097 page_cache_ra_unbounded(&ractl, len, 0); 4098 4099 for (i = 0; i < len; i++, page_idx++) { 4100 page = read_cache_page(mapping, page_idx, NULL, NULL); 4101 if (IS_ERR(page)) { 4102 ret = PTR_ERR(page); 4103 break; 4104 } 4105 page_len++; 4106 } 4107 4108 for (i = 0; i < page_len; i++, redirty_idx++) { 4109 page = find_lock_page(mapping, redirty_idx); 4110 4111 /* It will never fail, when page has pinned above */ 4112 f2fs_bug_on(F2FS_I_SB(inode), !page); 4113 4114 set_page_dirty(page); 4115 set_page_private_gcing(page); 4116 f2fs_put_page(page, 1); 4117 f2fs_put_page(page, 0); 4118 } 4119 4120 return ret; 4121 } 4122 4123 static int f2fs_ioc_decompress_file(struct file *filp) 4124 { 4125 struct inode *inode = file_inode(filp); 4126 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4127 struct f2fs_inode_info *fi = F2FS_I(inode); 4128 pgoff_t page_idx = 0, last_idx; 4129 int cluster_size = fi->i_cluster_size; 4130 int count, ret; 4131 4132 if (!f2fs_sb_has_compression(sbi) || 4133 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4134 return -EOPNOTSUPP; 4135 4136 if (!(filp->f_mode & FMODE_WRITE)) 4137 return -EBADF; 4138 4139 if (!f2fs_compressed_file(inode)) 4140 return -EINVAL; 4141 4142 f2fs_balance_fs(sbi, true); 4143 4144 file_start_write(filp); 4145 inode_lock(inode); 4146 4147 if (!f2fs_is_compress_backend_ready(inode)) { 4148 ret = -EOPNOTSUPP; 4149 goto out; 4150 } 4151 4152 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4153 ret = -EINVAL; 4154 goto out; 4155 } 4156 4157 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4158 if (ret) 4159 goto out; 4160 4161 if (!atomic_read(&fi->i_compr_blocks)) 4162 goto out; 4163 4164 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4165 4166 count = last_idx - page_idx; 4167 while (count && count >= cluster_size) { 4168 ret = redirty_blocks(inode, page_idx, cluster_size); 4169 if (ret < 0) 4170 break; 4171 4172 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4173 ret = filemap_fdatawrite(inode->i_mapping); 4174 if (ret < 0) 4175 break; 4176 } 4177 4178 count -= cluster_size; 4179 page_idx += cluster_size; 4180 4181 cond_resched(); 4182 if (fatal_signal_pending(current)) { 4183 ret = -EINTR; 4184 break; 4185 } 4186 } 4187 4188 if (!ret) 4189 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4190 LLONG_MAX); 4191 4192 if (ret) 4193 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4194 __func__, ret); 4195 out: 4196 inode_unlock(inode); 4197 file_end_write(filp); 4198 4199 return ret; 4200 } 4201 4202 static int f2fs_ioc_compress_file(struct file *filp) 4203 { 4204 struct inode *inode = file_inode(filp); 4205 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4206 pgoff_t page_idx = 0, last_idx; 4207 int cluster_size = F2FS_I(inode)->i_cluster_size; 4208 int count, ret; 4209 4210 if (!f2fs_sb_has_compression(sbi) || 4211 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4212 return -EOPNOTSUPP; 4213 4214 if (!(filp->f_mode & FMODE_WRITE)) 4215 return -EBADF; 4216 4217 if (!f2fs_compressed_file(inode)) 4218 return -EINVAL; 4219 4220 f2fs_balance_fs(sbi, true); 4221 4222 file_start_write(filp); 4223 inode_lock(inode); 4224 4225 if (!f2fs_is_compress_backend_ready(inode)) { 4226 ret = -EOPNOTSUPP; 4227 goto out; 4228 } 4229 4230 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4231 ret = -EINVAL; 4232 goto out; 4233 } 4234 4235 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4236 if (ret) 4237 goto out; 4238 4239 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4240 4241 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4242 4243 count = last_idx - page_idx; 4244 while (count && count >= cluster_size) { 4245 ret = redirty_blocks(inode, page_idx, cluster_size); 4246 if (ret < 0) 4247 break; 4248 4249 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4250 ret = filemap_fdatawrite(inode->i_mapping); 4251 if (ret < 0) 4252 break; 4253 } 4254 4255 count -= cluster_size; 4256 page_idx += cluster_size; 4257 4258 cond_resched(); 4259 if (fatal_signal_pending(current)) { 4260 ret = -EINTR; 4261 break; 4262 } 4263 } 4264 4265 if (!ret) 4266 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4267 LLONG_MAX); 4268 4269 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4270 4271 if (ret) 4272 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4273 __func__, ret); 4274 out: 4275 inode_unlock(inode); 4276 file_end_write(filp); 4277 4278 return ret; 4279 } 4280 4281 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4282 { 4283 switch (cmd) { 4284 case FS_IOC_GETVERSION: 4285 return f2fs_ioc_getversion(filp, arg); 4286 case F2FS_IOC_START_ATOMIC_WRITE: 4287 return f2fs_ioc_start_atomic_write(filp, false); 4288 case F2FS_IOC_START_ATOMIC_REPLACE: 4289 return f2fs_ioc_start_atomic_write(filp, true); 4290 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4291 return f2fs_ioc_commit_atomic_write(filp); 4292 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4293 return f2fs_ioc_abort_atomic_write(filp); 4294 case F2FS_IOC_START_VOLATILE_WRITE: 4295 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4296 return -EOPNOTSUPP; 4297 case F2FS_IOC_SHUTDOWN: 4298 return f2fs_ioc_shutdown(filp, arg); 4299 case FITRIM: 4300 return f2fs_ioc_fitrim(filp, arg); 4301 case FS_IOC_SET_ENCRYPTION_POLICY: 4302 return f2fs_ioc_set_encryption_policy(filp, arg); 4303 case FS_IOC_GET_ENCRYPTION_POLICY: 4304 return f2fs_ioc_get_encryption_policy(filp, arg); 4305 case FS_IOC_GET_ENCRYPTION_PWSALT: 4306 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4307 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4308 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4309 case FS_IOC_ADD_ENCRYPTION_KEY: 4310 return f2fs_ioc_add_encryption_key(filp, arg); 4311 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4312 return f2fs_ioc_remove_encryption_key(filp, arg); 4313 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4314 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4315 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4316 return f2fs_ioc_get_encryption_key_status(filp, arg); 4317 case FS_IOC_GET_ENCRYPTION_NONCE: 4318 return f2fs_ioc_get_encryption_nonce(filp, arg); 4319 case F2FS_IOC_GARBAGE_COLLECT: 4320 return f2fs_ioc_gc(filp, arg); 4321 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4322 return f2fs_ioc_gc_range(filp, arg); 4323 case F2FS_IOC_WRITE_CHECKPOINT: 4324 return f2fs_ioc_write_checkpoint(filp); 4325 case F2FS_IOC_DEFRAGMENT: 4326 return f2fs_ioc_defragment(filp, arg); 4327 case F2FS_IOC_MOVE_RANGE: 4328 return f2fs_ioc_move_range(filp, arg); 4329 case F2FS_IOC_FLUSH_DEVICE: 4330 return f2fs_ioc_flush_device(filp, arg); 4331 case F2FS_IOC_GET_FEATURES: 4332 return f2fs_ioc_get_features(filp, arg); 4333 case F2FS_IOC_GET_PIN_FILE: 4334 return f2fs_ioc_get_pin_file(filp, arg); 4335 case F2FS_IOC_SET_PIN_FILE: 4336 return f2fs_ioc_set_pin_file(filp, arg); 4337 case F2FS_IOC_PRECACHE_EXTENTS: 4338 return f2fs_ioc_precache_extents(filp); 4339 case F2FS_IOC_RESIZE_FS: 4340 return f2fs_ioc_resize_fs(filp, arg); 4341 case FS_IOC_ENABLE_VERITY: 4342 return f2fs_ioc_enable_verity(filp, arg); 4343 case FS_IOC_MEASURE_VERITY: 4344 return f2fs_ioc_measure_verity(filp, arg); 4345 case FS_IOC_READ_VERITY_METADATA: 4346 return f2fs_ioc_read_verity_metadata(filp, arg); 4347 case FS_IOC_GETFSLABEL: 4348 return f2fs_ioc_getfslabel(filp, arg); 4349 case FS_IOC_SETFSLABEL: 4350 return f2fs_ioc_setfslabel(filp, arg); 4351 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4352 return f2fs_ioc_get_compress_blocks(filp, arg); 4353 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4354 return f2fs_release_compress_blocks(filp, arg); 4355 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4356 return f2fs_reserve_compress_blocks(filp, arg); 4357 case F2FS_IOC_SEC_TRIM_FILE: 4358 return f2fs_sec_trim_file(filp, arg); 4359 case F2FS_IOC_GET_COMPRESS_OPTION: 4360 return f2fs_ioc_get_compress_option(filp, arg); 4361 case F2FS_IOC_SET_COMPRESS_OPTION: 4362 return f2fs_ioc_set_compress_option(filp, arg); 4363 case F2FS_IOC_DECOMPRESS_FILE: 4364 return f2fs_ioc_decompress_file(filp); 4365 case F2FS_IOC_COMPRESS_FILE: 4366 return f2fs_ioc_compress_file(filp); 4367 default: 4368 return -ENOTTY; 4369 } 4370 } 4371 4372 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4373 { 4374 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4375 return -EIO; 4376 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4377 return -ENOSPC; 4378 4379 return __f2fs_ioctl(filp, cmd, arg); 4380 } 4381 4382 /* 4383 * Return %true if the given read or write request should use direct I/O, or 4384 * %false if it should use buffered I/O. 4385 */ 4386 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4387 struct iov_iter *iter) 4388 { 4389 unsigned int align; 4390 4391 if (!(iocb->ki_flags & IOCB_DIRECT)) 4392 return false; 4393 4394 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4395 return false; 4396 4397 /* 4398 * Direct I/O not aligned to the disk's logical_block_size will be 4399 * attempted, but will fail with -EINVAL. 4400 * 4401 * f2fs additionally requires that direct I/O be aligned to the 4402 * filesystem block size, which is often a stricter requirement. 4403 * However, f2fs traditionally falls back to buffered I/O on requests 4404 * that are logical_block_size-aligned but not fs-block aligned. 4405 * 4406 * The below logic implements this behavior. 4407 */ 4408 align = iocb->ki_pos | iov_iter_alignment(iter); 4409 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4410 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4411 return false; 4412 4413 return true; 4414 } 4415 4416 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4417 unsigned int flags) 4418 { 4419 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4420 4421 dec_page_count(sbi, F2FS_DIO_READ); 4422 if (error) 4423 return error; 4424 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4425 return 0; 4426 } 4427 4428 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4429 .end_io = f2fs_dio_read_end_io, 4430 }; 4431 4432 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4433 { 4434 struct file *file = iocb->ki_filp; 4435 struct inode *inode = file_inode(file); 4436 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4437 struct f2fs_inode_info *fi = F2FS_I(inode); 4438 const loff_t pos = iocb->ki_pos; 4439 const size_t count = iov_iter_count(to); 4440 struct iomap_dio *dio; 4441 ssize_t ret; 4442 4443 if (count == 0) 4444 return 0; /* skip atime update */ 4445 4446 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4447 4448 if (iocb->ki_flags & IOCB_NOWAIT) { 4449 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4450 ret = -EAGAIN; 4451 goto out; 4452 } 4453 } else { 4454 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4455 } 4456 4457 /* 4458 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4459 * the higher-level function iomap_dio_rw() in order to ensure that the 4460 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4461 */ 4462 inc_page_count(sbi, F2FS_DIO_READ); 4463 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4464 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4465 if (IS_ERR_OR_NULL(dio)) { 4466 ret = PTR_ERR_OR_ZERO(dio); 4467 if (ret != -EIOCBQUEUED) 4468 dec_page_count(sbi, F2FS_DIO_READ); 4469 } else { 4470 ret = iomap_dio_complete(dio); 4471 } 4472 4473 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4474 4475 file_accessed(file); 4476 out: 4477 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4478 return ret; 4479 } 4480 4481 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4482 int rw) 4483 { 4484 struct inode *inode = file_inode(file); 4485 char *buf, *path; 4486 4487 buf = f2fs_getname(F2FS_I_SB(inode)); 4488 if (!buf) 4489 return; 4490 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4491 if (IS_ERR(path)) 4492 goto free_buf; 4493 if (rw == WRITE) 4494 trace_f2fs_datawrite_start(inode, pos, count, 4495 current->pid, path, current->comm); 4496 else 4497 trace_f2fs_dataread_start(inode, pos, count, 4498 current->pid, path, current->comm); 4499 free_buf: 4500 f2fs_putname(buf); 4501 } 4502 4503 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4504 { 4505 struct inode *inode = file_inode(iocb->ki_filp); 4506 const loff_t pos = iocb->ki_pos; 4507 ssize_t ret; 4508 4509 if (!f2fs_is_compress_backend_ready(inode)) 4510 return -EOPNOTSUPP; 4511 4512 if (trace_f2fs_dataread_start_enabled()) 4513 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4514 iov_iter_count(to), READ); 4515 4516 if (f2fs_should_use_dio(inode, iocb, to)) { 4517 ret = f2fs_dio_read_iter(iocb, to); 4518 } else { 4519 ret = filemap_read(iocb, to, 0); 4520 if (ret > 0) 4521 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4522 APP_BUFFERED_READ_IO, ret); 4523 } 4524 if (trace_f2fs_dataread_end_enabled()) 4525 trace_f2fs_dataread_end(inode, pos, ret); 4526 return ret; 4527 } 4528 4529 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4530 struct pipe_inode_info *pipe, 4531 size_t len, unsigned int flags) 4532 { 4533 struct inode *inode = file_inode(in); 4534 const loff_t pos = *ppos; 4535 ssize_t ret; 4536 4537 if (!f2fs_is_compress_backend_ready(inode)) 4538 return -EOPNOTSUPP; 4539 4540 if (trace_f2fs_dataread_start_enabled()) 4541 f2fs_trace_rw_file_path(in, pos, len, READ); 4542 4543 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4544 if (ret > 0) 4545 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4546 APP_BUFFERED_READ_IO, ret); 4547 4548 if (trace_f2fs_dataread_end_enabled()) 4549 trace_f2fs_dataread_end(inode, pos, ret); 4550 return ret; 4551 } 4552 4553 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4554 { 4555 struct file *file = iocb->ki_filp; 4556 struct inode *inode = file_inode(file); 4557 ssize_t count; 4558 int err; 4559 4560 if (IS_IMMUTABLE(inode)) 4561 return -EPERM; 4562 4563 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4564 return -EPERM; 4565 4566 count = generic_write_checks(iocb, from); 4567 if (count <= 0) 4568 return count; 4569 4570 err = file_modified(file); 4571 if (err) 4572 return err; 4573 return count; 4574 } 4575 4576 /* 4577 * Preallocate blocks for a write request, if it is possible and helpful to do 4578 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4579 * blocks were preallocated, or a negative errno value if something went 4580 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4581 * requested blocks (not just some of them) have been allocated. 4582 */ 4583 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4584 bool dio) 4585 { 4586 struct inode *inode = file_inode(iocb->ki_filp); 4587 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4588 const loff_t pos = iocb->ki_pos; 4589 const size_t count = iov_iter_count(iter); 4590 struct f2fs_map_blocks map = {}; 4591 int flag; 4592 int ret; 4593 4594 /* If it will be an out-of-place direct write, don't bother. */ 4595 if (dio && f2fs_lfs_mode(sbi)) 4596 return 0; 4597 /* 4598 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4599 * buffered IO, if DIO meets any holes. 4600 */ 4601 if (dio && i_size_read(inode) && 4602 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4603 return 0; 4604 4605 /* No-wait I/O can't allocate blocks. */ 4606 if (iocb->ki_flags & IOCB_NOWAIT) 4607 return 0; 4608 4609 /* If it will be a short write, don't bother. */ 4610 if (fault_in_iov_iter_readable(iter, count)) 4611 return 0; 4612 4613 if (f2fs_has_inline_data(inode)) { 4614 /* If the data will fit inline, don't bother. */ 4615 if (pos + count <= MAX_INLINE_DATA(inode)) 4616 return 0; 4617 ret = f2fs_convert_inline_inode(inode); 4618 if (ret) 4619 return ret; 4620 } 4621 4622 /* Do not preallocate blocks that will be written partially in 4KB. */ 4623 map.m_lblk = F2FS_BLK_ALIGN(pos); 4624 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4625 if (map.m_len > map.m_lblk) 4626 map.m_len -= map.m_lblk; 4627 else 4628 return 0; 4629 4630 map.m_may_create = true; 4631 if (dio) { 4632 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4633 flag = F2FS_GET_BLOCK_PRE_DIO; 4634 } else { 4635 map.m_seg_type = NO_CHECK_TYPE; 4636 flag = F2FS_GET_BLOCK_PRE_AIO; 4637 } 4638 4639 ret = f2fs_map_blocks(inode, &map, flag); 4640 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4641 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4642 return ret; 4643 if (ret == 0) 4644 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4645 return map.m_len; 4646 } 4647 4648 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4649 struct iov_iter *from) 4650 { 4651 struct file *file = iocb->ki_filp; 4652 struct inode *inode = file_inode(file); 4653 ssize_t ret; 4654 4655 if (iocb->ki_flags & IOCB_NOWAIT) 4656 return -EOPNOTSUPP; 4657 4658 ret = generic_perform_write(iocb, from); 4659 4660 if (ret > 0) { 4661 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4662 APP_BUFFERED_IO, ret); 4663 } 4664 return ret; 4665 } 4666 4667 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4668 unsigned int flags) 4669 { 4670 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4671 4672 dec_page_count(sbi, F2FS_DIO_WRITE); 4673 if (error) 4674 return error; 4675 f2fs_update_time(sbi, REQ_TIME); 4676 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4677 return 0; 4678 } 4679 4680 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4681 .end_io = f2fs_dio_write_end_io, 4682 }; 4683 4684 static void f2fs_flush_buffered_write(struct address_space *mapping, 4685 loff_t start_pos, loff_t end_pos) 4686 { 4687 int ret; 4688 4689 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4690 if (ret < 0) 4691 return; 4692 invalidate_mapping_pages(mapping, 4693 start_pos >> PAGE_SHIFT, 4694 end_pos >> PAGE_SHIFT); 4695 } 4696 4697 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4698 bool *may_need_sync) 4699 { 4700 struct file *file = iocb->ki_filp; 4701 struct inode *inode = file_inode(file); 4702 struct f2fs_inode_info *fi = F2FS_I(inode); 4703 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4704 const bool do_opu = f2fs_lfs_mode(sbi); 4705 const loff_t pos = iocb->ki_pos; 4706 const ssize_t count = iov_iter_count(from); 4707 unsigned int dio_flags; 4708 struct iomap_dio *dio; 4709 ssize_t ret; 4710 4711 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4712 4713 if (iocb->ki_flags & IOCB_NOWAIT) { 4714 /* f2fs_convert_inline_inode() and block allocation can block */ 4715 if (f2fs_has_inline_data(inode) || 4716 !f2fs_overwrite_io(inode, pos, count)) { 4717 ret = -EAGAIN; 4718 goto out; 4719 } 4720 4721 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4722 ret = -EAGAIN; 4723 goto out; 4724 } 4725 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4726 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4727 ret = -EAGAIN; 4728 goto out; 4729 } 4730 } else { 4731 ret = f2fs_convert_inline_inode(inode); 4732 if (ret) 4733 goto out; 4734 4735 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4736 if (do_opu) 4737 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4738 } 4739 4740 /* 4741 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4742 * the higher-level function iomap_dio_rw() in order to ensure that the 4743 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4744 */ 4745 inc_page_count(sbi, F2FS_DIO_WRITE); 4746 dio_flags = 0; 4747 if (pos + count > inode->i_size) 4748 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4749 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4750 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4751 if (IS_ERR_OR_NULL(dio)) { 4752 ret = PTR_ERR_OR_ZERO(dio); 4753 if (ret == -ENOTBLK) 4754 ret = 0; 4755 if (ret != -EIOCBQUEUED) 4756 dec_page_count(sbi, F2FS_DIO_WRITE); 4757 } else { 4758 ret = iomap_dio_complete(dio); 4759 } 4760 4761 if (do_opu) 4762 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4763 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4764 4765 if (ret < 0) 4766 goto out; 4767 if (pos + ret > inode->i_size) 4768 f2fs_i_size_write(inode, pos + ret); 4769 if (!do_opu) 4770 set_inode_flag(inode, FI_UPDATE_WRITE); 4771 4772 if (iov_iter_count(from)) { 4773 ssize_t ret2; 4774 loff_t bufio_start_pos = iocb->ki_pos; 4775 4776 /* 4777 * The direct write was partial, so we need to fall back to a 4778 * buffered write for the remainder. 4779 */ 4780 4781 ret2 = f2fs_buffered_write_iter(iocb, from); 4782 if (iov_iter_count(from)) 4783 f2fs_write_failed(inode, iocb->ki_pos); 4784 if (ret2 < 0) 4785 goto out; 4786 4787 /* 4788 * Ensure that the pagecache pages are written to disk and 4789 * invalidated to preserve the expected O_DIRECT semantics. 4790 */ 4791 if (ret2 > 0) { 4792 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4793 4794 ret += ret2; 4795 4796 f2fs_flush_buffered_write(file->f_mapping, 4797 bufio_start_pos, 4798 bufio_end_pos); 4799 } 4800 } else { 4801 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4802 *may_need_sync = false; 4803 } 4804 out: 4805 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4806 return ret; 4807 } 4808 4809 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4810 { 4811 struct inode *inode = file_inode(iocb->ki_filp); 4812 const loff_t orig_pos = iocb->ki_pos; 4813 const size_t orig_count = iov_iter_count(from); 4814 loff_t target_size; 4815 bool dio; 4816 bool may_need_sync = true; 4817 int preallocated; 4818 ssize_t ret; 4819 4820 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4821 ret = -EIO; 4822 goto out; 4823 } 4824 4825 if (!f2fs_is_compress_backend_ready(inode)) { 4826 ret = -EOPNOTSUPP; 4827 goto out; 4828 } 4829 4830 if (iocb->ki_flags & IOCB_NOWAIT) { 4831 if (!inode_trylock(inode)) { 4832 ret = -EAGAIN; 4833 goto out; 4834 } 4835 } else { 4836 inode_lock(inode); 4837 } 4838 4839 ret = f2fs_write_checks(iocb, from); 4840 if (ret <= 0) 4841 goto out_unlock; 4842 4843 /* Determine whether we will do a direct write or a buffered write. */ 4844 dio = f2fs_should_use_dio(inode, iocb, from); 4845 4846 /* Possibly preallocate the blocks for the write. */ 4847 target_size = iocb->ki_pos + iov_iter_count(from); 4848 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4849 if (preallocated < 0) { 4850 ret = preallocated; 4851 } else { 4852 if (trace_f2fs_datawrite_start_enabled()) 4853 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4854 orig_count, WRITE); 4855 4856 /* Do the actual write. */ 4857 ret = dio ? 4858 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4859 f2fs_buffered_write_iter(iocb, from); 4860 4861 if (trace_f2fs_datawrite_end_enabled()) 4862 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4863 } 4864 4865 /* Don't leave any preallocated blocks around past i_size. */ 4866 if (preallocated && i_size_read(inode) < target_size) { 4867 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4868 filemap_invalidate_lock(inode->i_mapping); 4869 if (!f2fs_truncate(inode)) 4870 file_dont_truncate(inode); 4871 filemap_invalidate_unlock(inode->i_mapping); 4872 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4873 } else { 4874 file_dont_truncate(inode); 4875 } 4876 4877 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4878 out_unlock: 4879 inode_unlock(inode); 4880 out: 4881 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4882 4883 if (ret > 0 && may_need_sync) 4884 ret = generic_write_sync(iocb, ret); 4885 4886 /* If buffered IO was forced, flush and drop the data from 4887 * the page cache to preserve O_DIRECT semantics 4888 */ 4889 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4890 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4891 orig_pos, 4892 orig_pos + ret - 1); 4893 4894 return ret; 4895 } 4896 4897 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4898 int advice) 4899 { 4900 struct address_space *mapping; 4901 struct backing_dev_info *bdi; 4902 struct inode *inode = file_inode(filp); 4903 int err; 4904 4905 if (advice == POSIX_FADV_SEQUENTIAL) { 4906 if (S_ISFIFO(inode->i_mode)) 4907 return -ESPIPE; 4908 4909 mapping = filp->f_mapping; 4910 if (!mapping || len < 0) 4911 return -EINVAL; 4912 4913 bdi = inode_to_bdi(mapping->host); 4914 filp->f_ra.ra_pages = bdi->ra_pages * 4915 F2FS_I_SB(inode)->seq_file_ra_mul; 4916 spin_lock(&filp->f_lock); 4917 filp->f_mode &= ~FMODE_RANDOM; 4918 spin_unlock(&filp->f_lock); 4919 return 0; 4920 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 4921 /* Load extent cache at the first readahead. */ 4922 f2fs_precache_extents(inode); 4923 } 4924 4925 err = generic_fadvise(filp, offset, len, advice); 4926 if (!err && advice == POSIX_FADV_DONTNEED && 4927 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4928 f2fs_compressed_file(inode)) 4929 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4930 4931 return err; 4932 } 4933 4934 #ifdef CONFIG_COMPAT 4935 struct compat_f2fs_gc_range { 4936 u32 sync; 4937 compat_u64 start; 4938 compat_u64 len; 4939 }; 4940 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4941 struct compat_f2fs_gc_range) 4942 4943 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4944 { 4945 struct compat_f2fs_gc_range __user *urange; 4946 struct f2fs_gc_range range; 4947 int err; 4948 4949 urange = compat_ptr(arg); 4950 err = get_user(range.sync, &urange->sync); 4951 err |= get_user(range.start, &urange->start); 4952 err |= get_user(range.len, &urange->len); 4953 if (err) 4954 return -EFAULT; 4955 4956 return __f2fs_ioc_gc_range(file, &range); 4957 } 4958 4959 struct compat_f2fs_move_range { 4960 u32 dst_fd; 4961 compat_u64 pos_in; 4962 compat_u64 pos_out; 4963 compat_u64 len; 4964 }; 4965 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4966 struct compat_f2fs_move_range) 4967 4968 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4969 { 4970 struct compat_f2fs_move_range __user *urange; 4971 struct f2fs_move_range range; 4972 int err; 4973 4974 urange = compat_ptr(arg); 4975 err = get_user(range.dst_fd, &urange->dst_fd); 4976 err |= get_user(range.pos_in, &urange->pos_in); 4977 err |= get_user(range.pos_out, &urange->pos_out); 4978 err |= get_user(range.len, &urange->len); 4979 if (err) 4980 return -EFAULT; 4981 4982 return __f2fs_ioc_move_range(file, &range); 4983 } 4984 4985 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4986 { 4987 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4988 return -EIO; 4989 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4990 return -ENOSPC; 4991 4992 switch (cmd) { 4993 case FS_IOC32_GETVERSION: 4994 cmd = FS_IOC_GETVERSION; 4995 break; 4996 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4997 return f2fs_compat_ioc_gc_range(file, arg); 4998 case F2FS_IOC32_MOVE_RANGE: 4999 return f2fs_compat_ioc_move_range(file, arg); 5000 case F2FS_IOC_START_ATOMIC_WRITE: 5001 case F2FS_IOC_START_ATOMIC_REPLACE: 5002 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5003 case F2FS_IOC_START_VOLATILE_WRITE: 5004 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5005 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5006 case F2FS_IOC_SHUTDOWN: 5007 case FITRIM: 5008 case FS_IOC_SET_ENCRYPTION_POLICY: 5009 case FS_IOC_GET_ENCRYPTION_PWSALT: 5010 case FS_IOC_GET_ENCRYPTION_POLICY: 5011 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5012 case FS_IOC_ADD_ENCRYPTION_KEY: 5013 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5014 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5015 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5016 case FS_IOC_GET_ENCRYPTION_NONCE: 5017 case F2FS_IOC_GARBAGE_COLLECT: 5018 case F2FS_IOC_WRITE_CHECKPOINT: 5019 case F2FS_IOC_DEFRAGMENT: 5020 case F2FS_IOC_FLUSH_DEVICE: 5021 case F2FS_IOC_GET_FEATURES: 5022 case F2FS_IOC_GET_PIN_FILE: 5023 case F2FS_IOC_SET_PIN_FILE: 5024 case F2FS_IOC_PRECACHE_EXTENTS: 5025 case F2FS_IOC_RESIZE_FS: 5026 case FS_IOC_ENABLE_VERITY: 5027 case FS_IOC_MEASURE_VERITY: 5028 case FS_IOC_READ_VERITY_METADATA: 5029 case FS_IOC_GETFSLABEL: 5030 case FS_IOC_SETFSLABEL: 5031 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5032 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5033 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5034 case F2FS_IOC_SEC_TRIM_FILE: 5035 case F2FS_IOC_GET_COMPRESS_OPTION: 5036 case F2FS_IOC_SET_COMPRESS_OPTION: 5037 case F2FS_IOC_DECOMPRESS_FILE: 5038 case F2FS_IOC_COMPRESS_FILE: 5039 break; 5040 default: 5041 return -ENOIOCTLCMD; 5042 } 5043 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5044 } 5045 #endif 5046 5047 const struct file_operations f2fs_file_operations = { 5048 .llseek = f2fs_llseek, 5049 .read_iter = f2fs_file_read_iter, 5050 .write_iter = f2fs_file_write_iter, 5051 .iopoll = iocb_bio_iopoll, 5052 .open = f2fs_file_open, 5053 .release = f2fs_release_file, 5054 .mmap = f2fs_file_mmap, 5055 .flush = f2fs_file_flush, 5056 .fsync = f2fs_sync_file, 5057 .fallocate = f2fs_fallocate, 5058 .unlocked_ioctl = f2fs_ioctl, 5059 #ifdef CONFIG_COMPAT 5060 .compat_ioctl = f2fs_compat_ioctl, 5061 #endif 5062 .splice_read = f2fs_file_splice_read, 5063 .splice_write = iter_file_splice_write, 5064 .fadvise = f2fs_file_fadvise, 5065 }; 5066