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 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1329 set_page_dirty(pdst); 1330 set_page_private_gcing(pdst); 1331 f2fs_put_page(pdst, 1); 1332 f2fs_put_page(psrc, 1); 1333 1334 ret = f2fs_truncate_hole(src_inode, 1335 src + i, src + i + 1); 1336 if (ret) 1337 return ret; 1338 i++; 1339 } 1340 } 1341 return 0; 1342 } 1343 1344 static int __exchange_data_block(struct inode *src_inode, 1345 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1346 pgoff_t len, bool full) 1347 { 1348 block_t *src_blkaddr; 1349 int *do_replace; 1350 pgoff_t olen; 1351 int ret; 1352 1353 while (len) { 1354 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1355 1356 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1357 array_size(olen, sizeof(block_t)), 1358 GFP_NOFS); 1359 if (!src_blkaddr) 1360 return -ENOMEM; 1361 1362 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1363 array_size(olen, sizeof(int)), 1364 GFP_NOFS); 1365 if (!do_replace) { 1366 kvfree(src_blkaddr); 1367 return -ENOMEM; 1368 } 1369 1370 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1371 do_replace, src, olen); 1372 if (ret) 1373 goto roll_back; 1374 1375 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1376 do_replace, src, dst, olen, full); 1377 if (ret) 1378 goto roll_back; 1379 1380 src += olen; 1381 dst += olen; 1382 len -= olen; 1383 1384 kvfree(src_blkaddr); 1385 kvfree(do_replace); 1386 } 1387 return 0; 1388 1389 roll_back: 1390 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1391 kvfree(src_blkaddr); 1392 kvfree(do_replace); 1393 return ret; 1394 } 1395 1396 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1397 { 1398 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1399 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1400 pgoff_t start = offset >> PAGE_SHIFT; 1401 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1402 int ret; 1403 1404 f2fs_balance_fs(sbi, true); 1405 1406 /* avoid gc operation during block exchange */ 1407 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1408 filemap_invalidate_lock(inode->i_mapping); 1409 1410 f2fs_lock_op(sbi); 1411 f2fs_drop_extent_tree(inode); 1412 truncate_pagecache(inode, offset); 1413 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1414 f2fs_unlock_op(sbi); 1415 1416 filemap_invalidate_unlock(inode->i_mapping); 1417 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1418 return ret; 1419 } 1420 1421 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1422 { 1423 loff_t new_size; 1424 int ret; 1425 1426 if (offset + len >= i_size_read(inode)) 1427 return -EINVAL; 1428 1429 /* collapse range should be aligned to block size of f2fs. */ 1430 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1431 return -EINVAL; 1432 1433 ret = f2fs_convert_inline_inode(inode); 1434 if (ret) 1435 return ret; 1436 1437 /* write out all dirty pages from offset */ 1438 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1439 if (ret) 1440 return ret; 1441 1442 ret = f2fs_do_collapse(inode, offset, len); 1443 if (ret) 1444 return ret; 1445 1446 /* write out all moved pages, if possible */ 1447 filemap_invalidate_lock(inode->i_mapping); 1448 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1449 truncate_pagecache(inode, offset); 1450 1451 new_size = i_size_read(inode) - len; 1452 ret = f2fs_truncate_blocks(inode, new_size, true); 1453 filemap_invalidate_unlock(inode->i_mapping); 1454 if (!ret) 1455 f2fs_i_size_write(inode, new_size); 1456 return ret; 1457 } 1458 1459 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1460 pgoff_t end) 1461 { 1462 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1463 pgoff_t index = start; 1464 unsigned int ofs_in_node = dn->ofs_in_node; 1465 blkcnt_t count = 0; 1466 int ret; 1467 1468 for (; index < end; index++, dn->ofs_in_node++) { 1469 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1470 count++; 1471 } 1472 1473 dn->ofs_in_node = ofs_in_node; 1474 ret = f2fs_reserve_new_blocks(dn, count); 1475 if (ret) 1476 return ret; 1477 1478 dn->ofs_in_node = ofs_in_node; 1479 for (index = start; index < end; index++, dn->ofs_in_node++) { 1480 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1481 /* 1482 * f2fs_reserve_new_blocks will not guarantee entire block 1483 * allocation. 1484 */ 1485 if (dn->data_blkaddr == NULL_ADDR) { 1486 ret = -ENOSPC; 1487 break; 1488 } 1489 1490 if (dn->data_blkaddr == NEW_ADDR) 1491 continue; 1492 1493 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1494 DATA_GENERIC_ENHANCE)) { 1495 ret = -EFSCORRUPTED; 1496 break; 1497 } 1498 1499 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1500 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1501 } 1502 1503 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1504 f2fs_update_age_extent_cache_range(dn, start, index - start); 1505 1506 return ret; 1507 } 1508 1509 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1510 int mode) 1511 { 1512 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1513 struct address_space *mapping = inode->i_mapping; 1514 pgoff_t index, pg_start, pg_end; 1515 loff_t new_size = i_size_read(inode); 1516 loff_t off_start, off_end; 1517 int ret = 0; 1518 1519 ret = inode_newsize_ok(inode, (len + offset)); 1520 if (ret) 1521 return ret; 1522 1523 ret = f2fs_convert_inline_inode(inode); 1524 if (ret) 1525 return ret; 1526 1527 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1528 if (ret) 1529 return ret; 1530 1531 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1532 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1533 1534 off_start = offset & (PAGE_SIZE - 1); 1535 off_end = (offset + len) & (PAGE_SIZE - 1); 1536 1537 if (pg_start == pg_end) { 1538 ret = fill_zero(inode, pg_start, off_start, 1539 off_end - off_start); 1540 if (ret) 1541 return ret; 1542 1543 new_size = max_t(loff_t, new_size, offset + len); 1544 } else { 1545 if (off_start) { 1546 ret = fill_zero(inode, pg_start++, off_start, 1547 PAGE_SIZE - off_start); 1548 if (ret) 1549 return ret; 1550 1551 new_size = max_t(loff_t, new_size, 1552 (loff_t)pg_start << PAGE_SHIFT); 1553 } 1554 1555 for (index = pg_start; index < pg_end;) { 1556 struct dnode_of_data dn; 1557 unsigned int end_offset; 1558 pgoff_t end; 1559 1560 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1561 filemap_invalidate_lock(mapping); 1562 1563 truncate_pagecache_range(inode, 1564 (loff_t)index << PAGE_SHIFT, 1565 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1566 1567 f2fs_lock_op(sbi); 1568 1569 set_new_dnode(&dn, inode, NULL, NULL, 0); 1570 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1571 if (ret) { 1572 f2fs_unlock_op(sbi); 1573 filemap_invalidate_unlock(mapping); 1574 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1575 goto out; 1576 } 1577 1578 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1579 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1580 1581 ret = f2fs_do_zero_range(&dn, index, end); 1582 f2fs_put_dnode(&dn); 1583 1584 f2fs_unlock_op(sbi); 1585 filemap_invalidate_unlock(mapping); 1586 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1587 1588 f2fs_balance_fs(sbi, dn.node_changed); 1589 1590 if (ret) 1591 goto out; 1592 1593 index = end; 1594 new_size = max_t(loff_t, new_size, 1595 (loff_t)index << PAGE_SHIFT); 1596 } 1597 1598 if (off_end) { 1599 ret = fill_zero(inode, pg_end, 0, off_end); 1600 if (ret) 1601 goto out; 1602 1603 new_size = max_t(loff_t, new_size, offset + len); 1604 } 1605 } 1606 1607 out: 1608 if (new_size > i_size_read(inode)) { 1609 if (mode & FALLOC_FL_KEEP_SIZE) 1610 file_set_keep_isize(inode); 1611 else 1612 f2fs_i_size_write(inode, new_size); 1613 } 1614 return ret; 1615 } 1616 1617 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1618 { 1619 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1620 struct address_space *mapping = inode->i_mapping; 1621 pgoff_t nr, pg_start, pg_end, delta, idx; 1622 loff_t new_size; 1623 int ret = 0; 1624 1625 new_size = i_size_read(inode) + len; 1626 ret = inode_newsize_ok(inode, new_size); 1627 if (ret) 1628 return ret; 1629 1630 if (offset >= i_size_read(inode)) 1631 return -EINVAL; 1632 1633 /* insert range should be aligned to block size of f2fs. */ 1634 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1635 return -EINVAL; 1636 1637 ret = f2fs_convert_inline_inode(inode); 1638 if (ret) 1639 return ret; 1640 1641 f2fs_balance_fs(sbi, true); 1642 1643 filemap_invalidate_lock(mapping); 1644 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1645 filemap_invalidate_unlock(mapping); 1646 if (ret) 1647 return ret; 1648 1649 /* write out all dirty pages from offset */ 1650 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1651 if (ret) 1652 return ret; 1653 1654 pg_start = offset >> PAGE_SHIFT; 1655 pg_end = (offset + len) >> PAGE_SHIFT; 1656 delta = pg_end - pg_start; 1657 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1658 1659 /* avoid gc operation during block exchange */ 1660 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1661 filemap_invalidate_lock(mapping); 1662 truncate_pagecache(inode, offset); 1663 1664 while (!ret && idx > pg_start) { 1665 nr = idx - pg_start; 1666 if (nr > delta) 1667 nr = delta; 1668 idx -= nr; 1669 1670 f2fs_lock_op(sbi); 1671 f2fs_drop_extent_tree(inode); 1672 1673 ret = __exchange_data_block(inode, inode, idx, 1674 idx + delta, nr, false); 1675 f2fs_unlock_op(sbi); 1676 } 1677 filemap_invalidate_unlock(mapping); 1678 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1679 if (ret) 1680 return ret; 1681 1682 /* write out all moved pages, if possible */ 1683 filemap_invalidate_lock(mapping); 1684 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1685 truncate_pagecache(inode, offset); 1686 filemap_invalidate_unlock(mapping); 1687 1688 if (!ret) 1689 f2fs_i_size_write(inode, new_size); 1690 return ret; 1691 } 1692 1693 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1694 loff_t len, int mode) 1695 { 1696 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1697 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1698 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1699 .m_may_create = true }; 1700 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1701 .init_gc_type = FG_GC, 1702 .should_migrate_blocks = false, 1703 .err_gc_skipped = true, 1704 .nr_free_secs = 0 }; 1705 pgoff_t pg_start, pg_end; 1706 loff_t new_size; 1707 loff_t off_end; 1708 block_t expanded = 0; 1709 int err; 1710 1711 err = inode_newsize_ok(inode, (len + offset)); 1712 if (err) 1713 return err; 1714 1715 err = f2fs_convert_inline_inode(inode); 1716 if (err) 1717 return err; 1718 1719 f2fs_balance_fs(sbi, true); 1720 1721 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1722 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1723 off_end = (offset + len) & (PAGE_SIZE - 1); 1724 1725 map.m_lblk = pg_start; 1726 map.m_len = pg_end - pg_start; 1727 if (off_end) 1728 map.m_len++; 1729 1730 if (!map.m_len) 1731 return 0; 1732 1733 if (f2fs_is_pinned_file(inode)) { 1734 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1735 block_t sec_len = roundup(map.m_len, sec_blks); 1736 1737 map.m_len = sec_blks; 1738 next_alloc: 1739 if (has_not_enough_free_secs(sbi, 0, 1740 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1741 f2fs_down_write(&sbi->gc_lock); 1742 stat_inc_gc_call_count(sbi, FOREGROUND); 1743 err = f2fs_gc(sbi, &gc_control); 1744 if (err && err != -ENODATA) 1745 goto out_err; 1746 } 1747 1748 f2fs_down_write(&sbi->pin_sem); 1749 1750 err = f2fs_allocate_pinning_section(sbi); 1751 if (err) { 1752 f2fs_up_write(&sbi->pin_sem); 1753 goto out_err; 1754 } 1755 1756 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1757 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1758 file_dont_truncate(inode); 1759 1760 f2fs_up_write(&sbi->pin_sem); 1761 1762 expanded += map.m_len; 1763 sec_len -= map.m_len; 1764 map.m_lblk += map.m_len; 1765 if (!err && sec_len) 1766 goto next_alloc; 1767 1768 map.m_len = expanded; 1769 } else { 1770 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1771 expanded = map.m_len; 1772 } 1773 out_err: 1774 if (err) { 1775 pgoff_t last_off; 1776 1777 if (!expanded) 1778 return err; 1779 1780 last_off = pg_start + expanded - 1; 1781 1782 /* update new size to the failed position */ 1783 new_size = (last_off == pg_end) ? offset + len : 1784 (loff_t)(last_off + 1) << PAGE_SHIFT; 1785 } else { 1786 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1787 } 1788 1789 if (new_size > i_size_read(inode)) { 1790 if (mode & FALLOC_FL_KEEP_SIZE) 1791 file_set_keep_isize(inode); 1792 else 1793 f2fs_i_size_write(inode, new_size); 1794 } 1795 1796 return err; 1797 } 1798 1799 static long f2fs_fallocate(struct file *file, int mode, 1800 loff_t offset, loff_t len) 1801 { 1802 struct inode *inode = file_inode(file); 1803 long ret = 0; 1804 1805 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1806 return -EIO; 1807 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1808 return -ENOSPC; 1809 if (!f2fs_is_compress_backend_ready(inode)) 1810 return -EOPNOTSUPP; 1811 1812 /* f2fs only support ->fallocate for regular file */ 1813 if (!S_ISREG(inode->i_mode)) 1814 return -EINVAL; 1815 1816 if (IS_ENCRYPTED(inode) && 1817 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1818 return -EOPNOTSUPP; 1819 1820 /* 1821 * Pinned file should not support partial truncation since the block 1822 * can be used by applications. 1823 */ 1824 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1825 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1826 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1827 return -EOPNOTSUPP; 1828 1829 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1830 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1831 FALLOC_FL_INSERT_RANGE)) 1832 return -EOPNOTSUPP; 1833 1834 inode_lock(inode); 1835 1836 ret = file_modified(file); 1837 if (ret) 1838 goto out; 1839 1840 if (mode & FALLOC_FL_PUNCH_HOLE) { 1841 if (offset >= inode->i_size) 1842 goto out; 1843 1844 ret = f2fs_punch_hole(inode, offset, len); 1845 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1846 ret = f2fs_collapse_range(inode, offset, len); 1847 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1848 ret = f2fs_zero_range(inode, offset, len, mode); 1849 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1850 ret = f2fs_insert_range(inode, offset, len); 1851 } else { 1852 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1853 } 1854 1855 if (!ret) { 1856 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1857 f2fs_mark_inode_dirty_sync(inode, false); 1858 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1859 } 1860 1861 out: 1862 inode_unlock(inode); 1863 1864 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1865 return ret; 1866 } 1867 1868 static int f2fs_release_file(struct inode *inode, struct file *filp) 1869 { 1870 /* 1871 * f2fs_release_file is called at every close calls. So we should 1872 * not drop any inmemory pages by close called by other process. 1873 */ 1874 if (!(filp->f_mode & FMODE_WRITE) || 1875 atomic_read(&inode->i_writecount) != 1) 1876 return 0; 1877 1878 inode_lock(inode); 1879 f2fs_abort_atomic_write(inode, true); 1880 inode_unlock(inode); 1881 1882 return 0; 1883 } 1884 1885 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1886 { 1887 struct inode *inode = file_inode(file); 1888 1889 /* 1890 * If the process doing a transaction is crashed, we should do 1891 * roll-back. Otherwise, other reader/write can see corrupted database 1892 * until all the writers close its file. Since this should be done 1893 * before dropping file lock, it needs to do in ->flush. 1894 */ 1895 if (F2FS_I(inode)->atomic_write_task == current && 1896 (current->flags & PF_EXITING)) { 1897 inode_lock(inode); 1898 f2fs_abort_atomic_write(inode, true); 1899 inode_unlock(inode); 1900 } 1901 1902 return 0; 1903 } 1904 1905 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1906 { 1907 struct f2fs_inode_info *fi = F2FS_I(inode); 1908 u32 masked_flags = fi->i_flags & mask; 1909 1910 /* mask can be shrunk by flags_valid selector */ 1911 iflags &= mask; 1912 1913 /* Is it quota file? Do not allow user to mess with it */ 1914 if (IS_NOQUOTA(inode)) 1915 return -EPERM; 1916 1917 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1918 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1919 return -EOPNOTSUPP; 1920 if (!f2fs_empty_dir(inode)) 1921 return -ENOTEMPTY; 1922 } 1923 1924 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1925 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1926 return -EOPNOTSUPP; 1927 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1928 return -EINVAL; 1929 } 1930 1931 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1932 if (masked_flags & F2FS_COMPR_FL) { 1933 if (!f2fs_disable_compressed_file(inode)) 1934 return -EINVAL; 1935 } else { 1936 /* try to convert inline_data to support compression */ 1937 int err = f2fs_convert_inline_inode(inode); 1938 if (err) 1939 return err; 1940 1941 f2fs_down_write(&F2FS_I(inode)->i_sem); 1942 if (!f2fs_may_compress(inode) || 1943 (S_ISREG(inode->i_mode) && 1944 F2FS_HAS_BLOCKS(inode))) { 1945 f2fs_up_write(&F2FS_I(inode)->i_sem); 1946 return -EINVAL; 1947 } 1948 err = set_compress_context(inode); 1949 f2fs_up_write(&F2FS_I(inode)->i_sem); 1950 1951 if (err) 1952 return err; 1953 } 1954 } 1955 1956 fi->i_flags = iflags | (fi->i_flags & ~mask); 1957 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1958 (fi->i_flags & F2FS_NOCOMP_FL)); 1959 1960 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1961 set_inode_flag(inode, FI_PROJ_INHERIT); 1962 else 1963 clear_inode_flag(inode, FI_PROJ_INHERIT); 1964 1965 inode_set_ctime_current(inode); 1966 f2fs_set_inode_flags(inode); 1967 f2fs_mark_inode_dirty_sync(inode, true); 1968 return 0; 1969 } 1970 1971 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1972 1973 /* 1974 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1975 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1976 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1977 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1978 * 1979 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1980 * FS_IOC_FSSETXATTR is done by the VFS. 1981 */ 1982 1983 static const struct { 1984 u32 iflag; 1985 u32 fsflag; 1986 } f2fs_fsflags_map[] = { 1987 { F2FS_COMPR_FL, FS_COMPR_FL }, 1988 { F2FS_SYNC_FL, FS_SYNC_FL }, 1989 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1990 { F2FS_APPEND_FL, FS_APPEND_FL }, 1991 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1992 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1993 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1994 { F2FS_INDEX_FL, FS_INDEX_FL }, 1995 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1996 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1997 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1998 }; 1999 2000 #define F2FS_GETTABLE_FS_FL ( \ 2001 FS_COMPR_FL | \ 2002 FS_SYNC_FL | \ 2003 FS_IMMUTABLE_FL | \ 2004 FS_APPEND_FL | \ 2005 FS_NODUMP_FL | \ 2006 FS_NOATIME_FL | \ 2007 FS_NOCOMP_FL | \ 2008 FS_INDEX_FL | \ 2009 FS_DIRSYNC_FL | \ 2010 FS_PROJINHERIT_FL | \ 2011 FS_ENCRYPT_FL | \ 2012 FS_INLINE_DATA_FL | \ 2013 FS_NOCOW_FL | \ 2014 FS_VERITY_FL | \ 2015 FS_CASEFOLD_FL) 2016 2017 #define F2FS_SETTABLE_FS_FL ( \ 2018 FS_COMPR_FL | \ 2019 FS_SYNC_FL | \ 2020 FS_IMMUTABLE_FL | \ 2021 FS_APPEND_FL | \ 2022 FS_NODUMP_FL | \ 2023 FS_NOATIME_FL | \ 2024 FS_NOCOMP_FL | \ 2025 FS_DIRSYNC_FL | \ 2026 FS_PROJINHERIT_FL | \ 2027 FS_CASEFOLD_FL) 2028 2029 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2030 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2031 { 2032 u32 fsflags = 0; 2033 int i; 2034 2035 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2036 if (iflags & f2fs_fsflags_map[i].iflag) 2037 fsflags |= f2fs_fsflags_map[i].fsflag; 2038 2039 return fsflags; 2040 } 2041 2042 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2043 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2044 { 2045 u32 iflags = 0; 2046 int i; 2047 2048 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2049 if (fsflags & f2fs_fsflags_map[i].fsflag) 2050 iflags |= f2fs_fsflags_map[i].iflag; 2051 2052 return iflags; 2053 } 2054 2055 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2056 { 2057 struct inode *inode = file_inode(filp); 2058 2059 return put_user(inode->i_generation, (int __user *)arg); 2060 } 2061 2062 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2063 { 2064 struct inode *inode = file_inode(filp); 2065 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2066 struct f2fs_inode_info *fi = F2FS_I(inode); 2067 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2068 struct inode *pinode; 2069 loff_t isize; 2070 int ret; 2071 2072 if (!inode_owner_or_capable(idmap, inode)) 2073 return -EACCES; 2074 2075 if (!S_ISREG(inode->i_mode)) 2076 return -EINVAL; 2077 2078 if (filp->f_flags & O_DIRECT) 2079 return -EINVAL; 2080 2081 ret = mnt_want_write_file(filp); 2082 if (ret) 2083 return ret; 2084 2085 inode_lock(inode); 2086 2087 if (!f2fs_disable_compressed_file(inode) || 2088 f2fs_is_pinned_file(inode)) { 2089 ret = -EINVAL; 2090 goto out; 2091 } 2092 2093 if (f2fs_is_atomic_file(inode)) 2094 goto out; 2095 2096 ret = f2fs_convert_inline_inode(inode); 2097 if (ret) 2098 goto out; 2099 2100 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2101 2102 /* 2103 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2104 * f2fs_is_atomic_file. 2105 */ 2106 if (get_dirty_pages(inode)) 2107 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2108 inode->i_ino, get_dirty_pages(inode)); 2109 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2110 if (ret) { 2111 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2112 goto out; 2113 } 2114 2115 /* Check if the inode already has a COW inode */ 2116 if (fi->cow_inode == NULL) { 2117 /* Create a COW inode for atomic write */ 2118 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2119 if (IS_ERR(pinode)) { 2120 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2121 ret = PTR_ERR(pinode); 2122 goto out; 2123 } 2124 2125 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2126 iput(pinode); 2127 if (ret) { 2128 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2129 goto out; 2130 } 2131 2132 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2133 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2134 } else { 2135 /* Reuse the already created COW inode */ 2136 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2137 if (ret) { 2138 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2139 goto out; 2140 } 2141 } 2142 2143 f2fs_write_inode(inode, NULL); 2144 2145 stat_inc_atomic_inode(inode); 2146 2147 set_inode_flag(inode, FI_ATOMIC_FILE); 2148 2149 isize = i_size_read(inode); 2150 fi->original_i_size = isize; 2151 if (truncate) { 2152 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2153 truncate_inode_pages_final(inode->i_mapping); 2154 f2fs_i_size_write(inode, 0); 2155 isize = 0; 2156 } 2157 f2fs_i_size_write(fi->cow_inode, isize); 2158 2159 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2160 2161 f2fs_update_time(sbi, REQ_TIME); 2162 fi->atomic_write_task = current; 2163 stat_update_max_atomic_write(inode); 2164 fi->atomic_write_cnt = 0; 2165 out: 2166 inode_unlock(inode); 2167 mnt_drop_write_file(filp); 2168 return ret; 2169 } 2170 2171 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2172 { 2173 struct inode *inode = file_inode(filp); 2174 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2175 int ret; 2176 2177 if (!inode_owner_or_capable(idmap, inode)) 2178 return -EACCES; 2179 2180 ret = mnt_want_write_file(filp); 2181 if (ret) 2182 return ret; 2183 2184 f2fs_balance_fs(F2FS_I_SB(inode), true); 2185 2186 inode_lock(inode); 2187 2188 if (f2fs_is_atomic_file(inode)) { 2189 ret = f2fs_commit_atomic_write(inode); 2190 if (!ret) 2191 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2192 2193 f2fs_abort_atomic_write(inode, ret); 2194 } else { 2195 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2196 } 2197 2198 inode_unlock(inode); 2199 mnt_drop_write_file(filp); 2200 return ret; 2201 } 2202 2203 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2204 { 2205 struct inode *inode = file_inode(filp); 2206 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2207 int ret; 2208 2209 if (!inode_owner_or_capable(idmap, inode)) 2210 return -EACCES; 2211 2212 ret = mnt_want_write_file(filp); 2213 if (ret) 2214 return ret; 2215 2216 inode_lock(inode); 2217 2218 f2fs_abort_atomic_write(inode, true); 2219 2220 inode_unlock(inode); 2221 2222 mnt_drop_write_file(filp); 2223 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2224 return ret; 2225 } 2226 2227 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2228 { 2229 struct inode *inode = file_inode(filp); 2230 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2231 struct super_block *sb = sbi->sb; 2232 __u32 in; 2233 int ret = 0; 2234 2235 if (!capable(CAP_SYS_ADMIN)) 2236 return -EPERM; 2237 2238 if (get_user(in, (__u32 __user *)arg)) 2239 return -EFAULT; 2240 2241 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2242 ret = mnt_want_write_file(filp); 2243 if (ret) { 2244 if (ret == -EROFS) { 2245 ret = 0; 2246 f2fs_stop_checkpoint(sbi, false, 2247 STOP_CP_REASON_SHUTDOWN); 2248 trace_f2fs_shutdown(sbi, in, ret); 2249 } 2250 return ret; 2251 } 2252 } 2253 2254 switch (in) { 2255 case F2FS_GOING_DOWN_FULLSYNC: 2256 ret = bdev_freeze(sb->s_bdev); 2257 if (ret) 2258 goto out; 2259 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2260 bdev_thaw(sb->s_bdev); 2261 break; 2262 case F2FS_GOING_DOWN_METASYNC: 2263 /* do checkpoint only */ 2264 ret = f2fs_sync_fs(sb, 1); 2265 if (ret) { 2266 if (ret == -EIO) 2267 ret = 0; 2268 goto out; 2269 } 2270 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2271 break; 2272 case F2FS_GOING_DOWN_NOSYNC: 2273 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2274 break; 2275 case F2FS_GOING_DOWN_METAFLUSH: 2276 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2277 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2278 break; 2279 case F2FS_GOING_DOWN_NEED_FSCK: 2280 set_sbi_flag(sbi, SBI_NEED_FSCK); 2281 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2282 set_sbi_flag(sbi, SBI_IS_DIRTY); 2283 /* do checkpoint only */ 2284 ret = f2fs_sync_fs(sb, 1); 2285 if (ret == -EIO) 2286 ret = 0; 2287 goto out; 2288 default: 2289 ret = -EINVAL; 2290 goto out; 2291 } 2292 2293 f2fs_stop_gc_thread(sbi); 2294 f2fs_stop_discard_thread(sbi); 2295 2296 f2fs_drop_discard_cmd(sbi); 2297 clear_opt(sbi, DISCARD); 2298 2299 f2fs_update_time(sbi, REQ_TIME); 2300 out: 2301 if (in != F2FS_GOING_DOWN_FULLSYNC) 2302 mnt_drop_write_file(filp); 2303 2304 trace_f2fs_shutdown(sbi, in, ret); 2305 2306 return ret; 2307 } 2308 2309 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2310 { 2311 struct inode *inode = file_inode(filp); 2312 struct super_block *sb = inode->i_sb; 2313 struct fstrim_range range; 2314 int ret; 2315 2316 if (!capable(CAP_SYS_ADMIN)) 2317 return -EPERM; 2318 2319 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2320 return -EOPNOTSUPP; 2321 2322 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2323 sizeof(range))) 2324 return -EFAULT; 2325 2326 ret = mnt_want_write_file(filp); 2327 if (ret) 2328 return ret; 2329 2330 range.minlen = max((unsigned int)range.minlen, 2331 bdev_discard_granularity(sb->s_bdev)); 2332 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2333 mnt_drop_write_file(filp); 2334 if (ret < 0) 2335 return ret; 2336 2337 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2338 sizeof(range))) 2339 return -EFAULT; 2340 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2341 return 0; 2342 } 2343 2344 static bool uuid_is_nonzero(__u8 u[16]) 2345 { 2346 int i; 2347 2348 for (i = 0; i < 16; i++) 2349 if (u[i]) 2350 return true; 2351 return false; 2352 } 2353 2354 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2355 { 2356 struct inode *inode = file_inode(filp); 2357 2358 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2359 return -EOPNOTSUPP; 2360 2361 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2362 2363 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2364 } 2365 2366 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2367 { 2368 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2369 return -EOPNOTSUPP; 2370 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2371 } 2372 2373 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2374 { 2375 struct inode *inode = file_inode(filp); 2376 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2377 u8 encrypt_pw_salt[16]; 2378 int err; 2379 2380 if (!f2fs_sb_has_encrypt(sbi)) 2381 return -EOPNOTSUPP; 2382 2383 err = mnt_want_write_file(filp); 2384 if (err) 2385 return err; 2386 2387 f2fs_down_write(&sbi->sb_lock); 2388 2389 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2390 goto got_it; 2391 2392 /* update superblock with uuid */ 2393 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2394 2395 err = f2fs_commit_super(sbi, false); 2396 if (err) { 2397 /* undo new data */ 2398 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2399 goto out_err; 2400 } 2401 got_it: 2402 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2403 out_err: 2404 f2fs_up_write(&sbi->sb_lock); 2405 mnt_drop_write_file(filp); 2406 2407 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2408 err = -EFAULT; 2409 2410 return err; 2411 } 2412 2413 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2414 unsigned long arg) 2415 { 2416 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2417 return -EOPNOTSUPP; 2418 2419 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2420 } 2421 2422 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2423 { 2424 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2425 return -EOPNOTSUPP; 2426 2427 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2428 } 2429 2430 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2431 { 2432 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2433 return -EOPNOTSUPP; 2434 2435 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2436 } 2437 2438 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2439 unsigned long arg) 2440 { 2441 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2442 return -EOPNOTSUPP; 2443 2444 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2445 } 2446 2447 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2448 unsigned long arg) 2449 { 2450 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2451 return -EOPNOTSUPP; 2452 2453 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2454 } 2455 2456 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2457 { 2458 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2459 return -EOPNOTSUPP; 2460 2461 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2462 } 2463 2464 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2465 { 2466 struct inode *inode = file_inode(filp); 2467 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2468 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2469 .no_bg_gc = false, 2470 .should_migrate_blocks = false, 2471 .nr_free_secs = 0 }; 2472 __u32 sync; 2473 int ret; 2474 2475 if (!capable(CAP_SYS_ADMIN)) 2476 return -EPERM; 2477 2478 if (get_user(sync, (__u32 __user *)arg)) 2479 return -EFAULT; 2480 2481 if (f2fs_readonly(sbi->sb)) 2482 return -EROFS; 2483 2484 ret = mnt_want_write_file(filp); 2485 if (ret) 2486 return ret; 2487 2488 if (!sync) { 2489 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2490 ret = -EBUSY; 2491 goto out; 2492 } 2493 } else { 2494 f2fs_down_write(&sbi->gc_lock); 2495 } 2496 2497 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2498 gc_control.err_gc_skipped = sync; 2499 stat_inc_gc_call_count(sbi, FOREGROUND); 2500 ret = f2fs_gc(sbi, &gc_control); 2501 out: 2502 mnt_drop_write_file(filp); 2503 return ret; 2504 } 2505 2506 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2507 { 2508 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2509 struct f2fs_gc_control gc_control = { 2510 .init_gc_type = range->sync ? FG_GC : BG_GC, 2511 .no_bg_gc = false, 2512 .should_migrate_blocks = false, 2513 .err_gc_skipped = range->sync, 2514 .nr_free_secs = 0 }; 2515 u64 end; 2516 int ret; 2517 2518 if (!capable(CAP_SYS_ADMIN)) 2519 return -EPERM; 2520 if (f2fs_readonly(sbi->sb)) 2521 return -EROFS; 2522 2523 end = range->start + range->len; 2524 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2525 end >= MAX_BLKADDR(sbi)) 2526 return -EINVAL; 2527 2528 ret = mnt_want_write_file(filp); 2529 if (ret) 2530 return ret; 2531 2532 do_more: 2533 if (!range->sync) { 2534 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2535 ret = -EBUSY; 2536 goto out; 2537 } 2538 } else { 2539 f2fs_down_write(&sbi->gc_lock); 2540 } 2541 2542 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2543 stat_inc_gc_call_count(sbi, FOREGROUND); 2544 ret = f2fs_gc(sbi, &gc_control); 2545 if (ret) { 2546 if (ret == -EBUSY) 2547 ret = -EAGAIN; 2548 goto out; 2549 } 2550 range->start += CAP_BLKS_PER_SEC(sbi); 2551 if (range->start <= end) 2552 goto do_more; 2553 out: 2554 mnt_drop_write_file(filp); 2555 return ret; 2556 } 2557 2558 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2559 { 2560 struct f2fs_gc_range range; 2561 2562 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2563 sizeof(range))) 2564 return -EFAULT; 2565 return __f2fs_ioc_gc_range(filp, &range); 2566 } 2567 2568 static int f2fs_ioc_write_checkpoint(struct file *filp) 2569 { 2570 struct inode *inode = file_inode(filp); 2571 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2572 int ret; 2573 2574 if (!capable(CAP_SYS_ADMIN)) 2575 return -EPERM; 2576 2577 if (f2fs_readonly(sbi->sb)) 2578 return -EROFS; 2579 2580 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2581 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2582 return -EINVAL; 2583 } 2584 2585 ret = mnt_want_write_file(filp); 2586 if (ret) 2587 return ret; 2588 2589 ret = f2fs_sync_fs(sbi->sb, 1); 2590 2591 mnt_drop_write_file(filp); 2592 return ret; 2593 } 2594 2595 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2596 struct file *filp, 2597 struct f2fs_defragment *range) 2598 { 2599 struct inode *inode = file_inode(filp); 2600 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2601 .m_seg_type = NO_CHECK_TYPE, 2602 .m_may_create = false }; 2603 struct extent_info ei = {}; 2604 pgoff_t pg_start, pg_end, next_pgofs; 2605 unsigned int total = 0, sec_num; 2606 block_t blk_end = 0; 2607 bool fragmented = false; 2608 int err; 2609 2610 pg_start = range->start >> PAGE_SHIFT; 2611 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2612 2613 f2fs_balance_fs(sbi, true); 2614 2615 inode_lock(inode); 2616 2617 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2618 err = -EINVAL; 2619 goto unlock_out; 2620 } 2621 2622 /* if in-place-update policy is enabled, don't waste time here */ 2623 set_inode_flag(inode, FI_OPU_WRITE); 2624 if (f2fs_should_update_inplace(inode, NULL)) { 2625 err = -EINVAL; 2626 goto out; 2627 } 2628 2629 /* writeback all dirty pages in the range */ 2630 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2631 range->start + range->len - 1); 2632 if (err) 2633 goto out; 2634 2635 /* 2636 * lookup mapping info in extent cache, skip defragmenting if physical 2637 * block addresses are continuous. 2638 */ 2639 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2640 if (ei.fofs + ei.len >= pg_end) 2641 goto out; 2642 } 2643 2644 map.m_lblk = pg_start; 2645 map.m_next_pgofs = &next_pgofs; 2646 2647 /* 2648 * lookup mapping info in dnode page cache, skip defragmenting if all 2649 * physical block addresses are continuous even if there are hole(s) 2650 * in logical blocks. 2651 */ 2652 while (map.m_lblk < pg_end) { 2653 map.m_len = pg_end - map.m_lblk; 2654 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2655 if (err) 2656 goto out; 2657 2658 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2659 map.m_lblk = next_pgofs; 2660 continue; 2661 } 2662 2663 if (blk_end && blk_end != map.m_pblk) 2664 fragmented = true; 2665 2666 /* record total count of block that we're going to move */ 2667 total += map.m_len; 2668 2669 blk_end = map.m_pblk + map.m_len; 2670 2671 map.m_lblk += map.m_len; 2672 } 2673 2674 if (!fragmented) { 2675 total = 0; 2676 goto out; 2677 } 2678 2679 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2680 2681 /* 2682 * make sure there are enough free section for LFS allocation, this can 2683 * avoid defragment running in SSR mode when free section are allocated 2684 * intensively 2685 */ 2686 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2687 err = -EAGAIN; 2688 goto out; 2689 } 2690 2691 map.m_lblk = pg_start; 2692 map.m_len = pg_end - pg_start; 2693 total = 0; 2694 2695 while (map.m_lblk < pg_end) { 2696 pgoff_t idx; 2697 int cnt = 0; 2698 2699 do_map: 2700 map.m_len = pg_end - map.m_lblk; 2701 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2702 if (err) 2703 goto clear_out; 2704 2705 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2706 map.m_lblk = next_pgofs; 2707 goto check; 2708 } 2709 2710 set_inode_flag(inode, FI_SKIP_WRITES); 2711 2712 idx = map.m_lblk; 2713 while (idx < map.m_lblk + map.m_len && 2714 cnt < BLKS_PER_SEG(sbi)) { 2715 struct page *page; 2716 2717 page = f2fs_get_lock_data_page(inode, idx, true); 2718 if (IS_ERR(page)) { 2719 err = PTR_ERR(page); 2720 goto clear_out; 2721 } 2722 2723 set_page_dirty(page); 2724 set_page_private_gcing(page); 2725 f2fs_put_page(page, 1); 2726 2727 idx++; 2728 cnt++; 2729 total++; 2730 } 2731 2732 map.m_lblk = idx; 2733 check: 2734 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 2735 goto do_map; 2736 2737 clear_inode_flag(inode, FI_SKIP_WRITES); 2738 2739 err = filemap_fdatawrite(inode->i_mapping); 2740 if (err) 2741 goto out; 2742 } 2743 clear_out: 2744 clear_inode_flag(inode, FI_SKIP_WRITES); 2745 out: 2746 clear_inode_flag(inode, FI_OPU_WRITE); 2747 unlock_out: 2748 inode_unlock(inode); 2749 if (!err) 2750 range->len = (u64)total << PAGE_SHIFT; 2751 return err; 2752 } 2753 2754 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2755 { 2756 struct inode *inode = file_inode(filp); 2757 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2758 struct f2fs_defragment range; 2759 int err; 2760 2761 if (!capable(CAP_SYS_ADMIN)) 2762 return -EPERM; 2763 2764 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2765 return -EINVAL; 2766 2767 if (f2fs_readonly(sbi->sb)) 2768 return -EROFS; 2769 2770 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2771 sizeof(range))) 2772 return -EFAULT; 2773 2774 /* verify alignment of offset & size */ 2775 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2776 return -EINVAL; 2777 2778 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2779 max_file_blocks(inode))) 2780 return -EINVAL; 2781 2782 err = mnt_want_write_file(filp); 2783 if (err) 2784 return err; 2785 2786 err = f2fs_defragment_range(sbi, filp, &range); 2787 mnt_drop_write_file(filp); 2788 2789 f2fs_update_time(sbi, REQ_TIME); 2790 if (err < 0) 2791 return err; 2792 2793 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2794 sizeof(range))) 2795 return -EFAULT; 2796 2797 return 0; 2798 } 2799 2800 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2801 struct file *file_out, loff_t pos_out, size_t len) 2802 { 2803 struct inode *src = file_inode(file_in); 2804 struct inode *dst = file_inode(file_out); 2805 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2806 size_t olen = len, dst_max_i_size = 0; 2807 size_t dst_osize; 2808 int ret; 2809 2810 if (file_in->f_path.mnt != file_out->f_path.mnt || 2811 src->i_sb != dst->i_sb) 2812 return -EXDEV; 2813 2814 if (unlikely(f2fs_readonly(src->i_sb))) 2815 return -EROFS; 2816 2817 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2818 return -EINVAL; 2819 2820 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2821 return -EOPNOTSUPP; 2822 2823 if (pos_out < 0 || pos_in < 0) 2824 return -EINVAL; 2825 2826 if (src == dst) { 2827 if (pos_in == pos_out) 2828 return 0; 2829 if (pos_out > pos_in && pos_out < pos_in + len) 2830 return -EINVAL; 2831 } 2832 2833 inode_lock(src); 2834 if (src != dst) { 2835 ret = -EBUSY; 2836 if (!inode_trylock(dst)) 2837 goto out; 2838 } 2839 2840 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) { 2841 ret = -EOPNOTSUPP; 2842 goto out_unlock; 2843 } 2844 2845 ret = -EINVAL; 2846 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2847 goto out_unlock; 2848 if (len == 0) 2849 olen = len = src->i_size - pos_in; 2850 if (pos_in + len == src->i_size) 2851 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2852 if (len == 0) { 2853 ret = 0; 2854 goto out_unlock; 2855 } 2856 2857 dst_osize = dst->i_size; 2858 if (pos_out + olen > dst->i_size) 2859 dst_max_i_size = pos_out + olen; 2860 2861 /* verify the end result is block aligned */ 2862 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2863 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2864 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2865 goto out_unlock; 2866 2867 ret = f2fs_convert_inline_inode(src); 2868 if (ret) 2869 goto out_unlock; 2870 2871 ret = f2fs_convert_inline_inode(dst); 2872 if (ret) 2873 goto out_unlock; 2874 2875 /* write out all dirty pages from offset */ 2876 ret = filemap_write_and_wait_range(src->i_mapping, 2877 pos_in, pos_in + len); 2878 if (ret) 2879 goto out_unlock; 2880 2881 ret = filemap_write_and_wait_range(dst->i_mapping, 2882 pos_out, pos_out + len); 2883 if (ret) 2884 goto out_unlock; 2885 2886 f2fs_balance_fs(sbi, true); 2887 2888 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2889 if (src != dst) { 2890 ret = -EBUSY; 2891 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2892 goto out_src; 2893 } 2894 2895 f2fs_lock_op(sbi); 2896 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2897 pos_out >> F2FS_BLKSIZE_BITS, 2898 len >> F2FS_BLKSIZE_BITS, false); 2899 2900 if (!ret) { 2901 if (dst_max_i_size) 2902 f2fs_i_size_write(dst, dst_max_i_size); 2903 else if (dst_osize != dst->i_size) 2904 f2fs_i_size_write(dst, dst_osize); 2905 } 2906 f2fs_unlock_op(sbi); 2907 2908 if (src != dst) 2909 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2910 out_src: 2911 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2912 if (ret) 2913 goto out_unlock; 2914 2915 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 2916 f2fs_mark_inode_dirty_sync(src, false); 2917 if (src != dst) { 2918 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 2919 f2fs_mark_inode_dirty_sync(dst, false); 2920 } 2921 f2fs_update_time(sbi, REQ_TIME); 2922 2923 out_unlock: 2924 if (src != dst) 2925 inode_unlock(dst); 2926 out: 2927 inode_unlock(src); 2928 return ret; 2929 } 2930 2931 static int __f2fs_ioc_move_range(struct file *filp, 2932 struct f2fs_move_range *range) 2933 { 2934 struct fd dst; 2935 int err; 2936 2937 if (!(filp->f_mode & FMODE_READ) || 2938 !(filp->f_mode & FMODE_WRITE)) 2939 return -EBADF; 2940 2941 dst = fdget(range->dst_fd); 2942 if (!dst.file) 2943 return -EBADF; 2944 2945 if (!(dst.file->f_mode & FMODE_WRITE)) { 2946 err = -EBADF; 2947 goto err_out; 2948 } 2949 2950 err = mnt_want_write_file(filp); 2951 if (err) 2952 goto err_out; 2953 2954 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2955 range->pos_out, range->len); 2956 2957 mnt_drop_write_file(filp); 2958 err_out: 2959 fdput(dst); 2960 return err; 2961 } 2962 2963 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2964 { 2965 struct f2fs_move_range range; 2966 2967 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2968 sizeof(range))) 2969 return -EFAULT; 2970 return __f2fs_ioc_move_range(filp, &range); 2971 } 2972 2973 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2974 { 2975 struct inode *inode = file_inode(filp); 2976 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2977 struct sit_info *sm = SIT_I(sbi); 2978 unsigned int start_segno = 0, end_segno = 0; 2979 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2980 struct f2fs_flush_device range; 2981 struct f2fs_gc_control gc_control = { 2982 .init_gc_type = FG_GC, 2983 .should_migrate_blocks = true, 2984 .err_gc_skipped = true, 2985 .nr_free_secs = 0 }; 2986 int ret; 2987 2988 if (!capable(CAP_SYS_ADMIN)) 2989 return -EPERM; 2990 2991 if (f2fs_readonly(sbi->sb)) 2992 return -EROFS; 2993 2994 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2995 return -EINVAL; 2996 2997 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2998 sizeof(range))) 2999 return -EFAULT; 3000 3001 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3002 __is_large_section(sbi)) { 3003 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3004 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3005 return -EINVAL; 3006 } 3007 3008 ret = mnt_want_write_file(filp); 3009 if (ret) 3010 return ret; 3011 3012 if (range.dev_num != 0) 3013 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3014 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3015 3016 start_segno = sm->last_victim[FLUSH_DEVICE]; 3017 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3018 start_segno = dev_start_segno; 3019 end_segno = min(start_segno + range.segments, dev_end_segno); 3020 3021 while (start_segno < end_segno) { 3022 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3023 ret = -EBUSY; 3024 goto out; 3025 } 3026 sm->last_victim[GC_CB] = end_segno + 1; 3027 sm->last_victim[GC_GREEDY] = end_segno + 1; 3028 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3029 3030 gc_control.victim_segno = start_segno; 3031 stat_inc_gc_call_count(sbi, FOREGROUND); 3032 ret = f2fs_gc(sbi, &gc_control); 3033 if (ret == -EAGAIN) 3034 ret = 0; 3035 else if (ret < 0) 3036 break; 3037 start_segno++; 3038 } 3039 out: 3040 mnt_drop_write_file(filp); 3041 return ret; 3042 } 3043 3044 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3045 { 3046 struct inode *inode = file_inode(filp); 3047 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3048 3049 /* Must validate to set it with SQLite behavior in Android. */ 3050 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3051 3052 return put_user(sb_feature, (u32 __user *)arg); 3053 } 3054 3055 #ifdef CONFIG_QUOTA 3056 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3057 { 3058 struct dquot *transfer_to[MAXQUOTAS] = {}; 3059 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3060 struct super_block *sb = sbi->sb; 3061 int err; 3062 3063 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3064 if (IS_ERR(transfer_to[PRJQUOTA])) 3065 return PTR_ERR(transfer_to[PRJQUOTA]); 3066 3067 err = __dquot_transfer(inode, transfer_to); 3068 if (err) 3069 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3070 dqput(transfer_to[PRJQUOTA]); 3071 return err; 3072 } 3073 3074 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3075 { 3076 struct f2fs_inode_info *fi = F2FS_I(inode); 3077 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3078 struct f2fs_inode *ri = NULL; 3079 kprojid_t kprojid; 3080 int err; 3081 3082 if (!f2fs_sb_has_project_quota(sbi)) { 3083 if (projid != F2FS_DEF_PROJID) 3084 return -EOPNOTSUPP; 3085 else 3086 return 0; 3087 } 3088 3089 if (!f2fs_has_extra_attr(inode)) 3090 return -EOPNOTSUPP; 3091 3092 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3093 3094 if (projid_eq(kprojid, fi->i_projid)) 3095 return 0; 3096 3097 err = -EPERM; 3098 /* Is it quota file? Do not allow user to mess with it */ 3099 if (IS_NOQUOTA(inode)) 3100 return err; 3101 3102 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3103 return -EOVERFLOW; 3104 3105 err = f2fs_dquot_initialize(inode); 3106 if (err) 3107 return err; 3108 3109 f2fs_lock_op(sbi); 3110 err = f2fs_transfer_project_quota(inode, kprojid); 3111 if (err) 3112 goto out_unlock; 3113 3114 fi->i_projid = kprojid; 3115 inode_set_ctime_current(inode); 3116 f2fs_mark_inode_dirty_sync(inode, true); 3117 out_unlock: 3118 f2fs_unlock_op(sbi); 3119 return err; 3120 } 3121 #else 3122 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3123 { 3124 return 0; 3125 } 3126 3127 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3128 { 3129 if (projid != F2FS_DEF_PROJID) 3130 return -EOPNOTSUPP; 3131 return 0; 3132 } 3133 #endif 3134 3135 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3136 { 3137 struct inode *inode = d_inode(dentry); 3138 struct f2fs_inode_info *fi = F2FS_I(inode); 3139 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3140 3141 if (IS_ENCRYPTED(inode)) 3142 fsflags |= FS_ENCRYPT_FL; 3143 if (IS_VERITY(inode)) 3144 fsflags |= FS_VERITY_FL; 3145 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3146 fsflags |= FS_INLINE_DATA_FL; 3147 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3148 fsflags |= FS_NOCOW_FL; 3149 3150 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3151 3152 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3153 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3154 3155 return 0; 3156 } 3157 3158 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3159 struct dentry *dentry, struct fileattr *fa) 3160 { 3161 struct inode *inode = d_inode(dentry); 3162 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3163 u32 iflags; 3164 int err; 3165 3166 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3167 return -EIO; 3168 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3169 return -ENOSPC; 3170 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3171 return -EOPNOTSUPP; 3172 fsflags &= F2FS_SETTABLE_FS_FL; 3173 if (!fa->flags_valid) 3174 mask &= FS_COMMON_FL; 3175 3176 iflags = f2fs_fsflags_to_iflags(fsflags); 3177 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3178 return -EOPNOTSUPP; 3179 3180 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3181 if (!err) 3182 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3183 3184 return err; 3185 } 3186 3187 int f2fs_pin_file_control(struct inode *inode, bool inc) 3188 { 3189 struct f2fs_inode_info *fi = F2FS_I(inode); 3190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3191 3192 /* Use i_gc_failures for normal file as a risk signal. */ 3193 if (inc) 3194 f2fs_i_gc_failures_write(inode, 3195 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3196 3197 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3198 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3199 __func__, inode->i_ino, 3200 fi->i_gc_failures[GC_FAILURE_PIN]); 3201 clear_inode_flag(inode, FI_PIN_FILE); 3202 return -EAGAIN; 3203 } 3204 return 0; 3205 } 3206 3207 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3208 { 3209 struct inode *inode = file_inode(filp); 3210 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3211 __u32 pin; 3212 int ret = 0; 3213 3214 if (get_user(pin, (__u32 __user *)arg)) 3215 return -EFAULT; 3216 3217 if (!S_ISREG(inode->i_mode)) 3218 return -EINVAL; 3219 3220 if (f2fs_readonly(sbi->sb)) 3221 return -EROFS; 3222 3223 ret = mnt_want_write_file(filp); 3224 if (ret) 3225 return ret; 3226 3227 inode_lock(inode); 3228 3229 if (!pin) { 3230 clear_inode_flag(inode, FI_PIN_FILE); 3231 f2fs_i_gc_failures_write(inode, 0); 3232 goto done; 3233 } else if (f2fs_is_pinned_file(inode)) { 3234 goto done; 3235 } 3236 3237 if (f2fs_sb_has_blkzoned(sbi) && F2FS_HAS_BLOCKS(inode)) { 3238 ret = -EFBIG; 3239 goto out; 3240 } 3241 3242 /* Let's allow file pinning on zoned device. */ 3243 if (!f2fs_sb_has_blkzoned(sbi) && 3244 f2fs_should_update_outplace(inode, NULL)) { 3245 ret = -EINVAL; 3246 goto out; 3247 } 3248 3249 if (f2fs_pin_file_control(inode, false)) { 3250 ret = -EAGAIN; 3251 goto out; 3252 } 3253 3254 ret = f2fs_convert_inline_inode(inode); 3255 if (ret) 3256 goto out; 3257 3258 if (!f2fs_disable_compressed_file(inode)) { 3259 ret = -EOPNOTSUPP; 3260 goto out; 3261 } 3262 3263 set_inode_flag(inode, FI_PIN_FILE); 3264 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3265 done: 3266 f2fs_update_time(sbi, REQ_TIME); 3267 out: 3268 inode_unlock(inode); 3269 mnt_drop_write_file(filp); 3270 return ret; 3271 } 3272 3273 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3274 { 3275 struct inode *inode = file_inode(filp); 3276 __u32 pin = 0; 3277 3278 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3279 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3280 return put_user(pin, (u32 __user *)arg); 3281 } 3282 3283 int f2fs_precache_extents(struct inode *inode) 3284 { 3285 struct f2fs_inode_info *fi = F2FS_I(inode); 3286 struct f2fs_map_blocks map; 3287 pgoff_t m_next_extent; 3288 loff_t end; 3289 int err; 3290 3291 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3292 return -EOPNOTSUPP; 3293 3294 map.m_lblk = 0; 3295 map.m_pblk = 0; 3296 map.m_next_pgofs = NULL; 3297 map.m_next_extent = &m_next_extent; 3298 map.m_seg_type = NO_CHECK_TYPE; 3299 map.m_may_create = false; 3300 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3301 3302 while (map.m_lblk < end) { 3303 map.m_len = end - map.m_lblk; 3304 3305 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3306 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3307 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3308 if (err || !map.m_len) 3309 return err; 3310 3311 map.m_lblk = m_next_extent; 3312 } 3313 3314 return 0; 3315 } 3316 3317 static int f2fs_ioc_precache_extents(struct file *filp) 3318 { 3319 return f2fs_precache_extents(file_inode(filp)); 3320 } 3321 3322 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3323 { 3324 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3325 __u64 block_count; 3326 3327 if (!capable(CAP_SYS_ADMIN)) 3328 return -EPERM; 3329 3330 if (f2fs_readonly(sbi->sb)) 3331 return -EROFS; 3332 3333 if (copy_from_user(&block_count, (void __user *)arg, 3334 sizeof(block_count))) 3335 return -EFAULT; 3336 3337 return f2fs_resize_fs(filp, block_count); 3338 } 3339 3340 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3341 { 3342 struct inode *inode = file_inode(filp); 3343 3344 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3345 3346 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3347 f2fs_warn(F2FS_I_SB(inode), 3348 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3349 inode->i_ino); 3350 return -EOPNOTSUPP; 3351 } 3352 3353 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3354 } 3355 3356 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3357 { 3358 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3359 return -EOPNOTSUPP; 3360 3361 return fsverity_ioctl_measure(filp, (void __user *)arg); 3362 } 3363 3364 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3365 { 3366 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3367 return -EOPNOTSUPP; 3368 3369 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3370 } 3371 3372 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3373 { 3374 struct inode *inode = file_inode(filp); 3375 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3376 char *vbuf; 3377 int count; 3378 int err = 0; 3379 3380 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3381 if (!vbuf) 3382 return -ENOMEM; 3383 3384 f2fs_down_read(&sbi->sb_lock); 3385 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3386 ARRAY_SIZE(sbi->raw_super->volume_name), 3387 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3388 f2fs_up_read(&sbi->sb_lock); 3389 3390 if (copy_to_user((char __user *)arg, vbuf, 3391 min(FSLABEL_MAX, count))) 3392 err = -EFAULT; 3393 3394 kfree(vbuf); 3395 return err; 3396 } 3397 3398 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3399 { 3400 struct inode *inode = file_inode(filp); 3401 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3402 char *vbuf; 3403 int err = 0; 3404 3405 if (!capable(CAP_SYS_ADMIN)) 3406 return -EPERM; 3407 3408 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3409 if (IS_ERR(vbuf)) 3410 return PTR_ERR(vbuf); 3411 3412 err = mnt_want_write_file(filp); 3413 if (err) 3414 goto out; 3415 3416 f2fs_down_write(&sbi->sb_lock); 3417 3418 memset(sbi->raw_super->volume_name, 0, 3419 sizeof(sbi->raw_super->volume_name)); 3420 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3421 sbi->raw_super->volume_name, 3422 ARRAY_SIZE(sbi->raw_super->volume_name)); 3423 3424 err = f2fs_commit_super(sbi, false); 3425 3426 f2fs_up_write(&sbi->sb_lock); 3427 3428 mnt_drop_write_file(filp); 3429 out: 3430 kfree(vbuf); 3431 return err; 3432 } 3433 3434 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3435 { 3436 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3437 return -EOPNOTSUPP; 3438 3439 if (!f2fs_compressed_file(inode)) 3440 return -EINVAL; 3441 3442 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3443 3444 return 0; 3445 } 3446 3447 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3448 { 3449 struct inode *inode = file_inode(filp); 3450 __u64 blocks; 3451 int ret; 3452 3453 ret = f2fs_get_compress_blocks(inode, &blocks); 3454 if (ret < 0) 3455 return ret; 3456 3457 return put_user(blocks, (u64 __user *)arg); 3458 } 3459 3460 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3461 { 3462 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3463 unsigned int released_blocks = 0; 3464 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3465 block_t blkaddr; 3466 int i; 3467 3468 for (i = 0; i < count; i++) { 3469 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3470 dn->ofs_in_node + i); 3471 3472 if (!__is_valid_data_blkaddr(blkaddr)) 3473 continue; 3474 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3475 DATA_GENERIC_ENHANCE))) 3476 return -EFSCORRUPTED; 3477 } 3478 3479 while (count) { 3480 int compr_blocks = 0; 3481 3482 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3483 blkaddr = f2fs_data_blkaddr(dn); 3484 3485 if (i == 0) { 3486 if (blkaddr == COMPRESS_ADDR) 3487 continue; 3488 dn->ofs_in_node += cluster_size; 3489 goto next; 3490 } 3491 3492 if (__is_valid_data_blkaddr(blkaddr)) 3493 compr_blocks++; 3494 3495 if (blkaddr != NEW_ADDR) 3496 continue; 3497 3498 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3499 } 3500 3501 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3502 dec_valid_block_count(sbi, dn->inode, 3503 cluster_size - compr_blocks); 3504 3505 released_blocks += cluster_size - compr_blocks; 3506 next: 3507 count -= cluster_size; 3508 } 3509 3510 return released_blocks; 3511 } 3512 3513 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3514 { 3515 struct inode *inode = file_inode(filp); 3516 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3517 pgoff_t page_idx = 0, last_idx; 3518 unsigned int released_blocks = 0; 3519 int ret; 3520 int writecount; 3521 3522 if (!f2fs_sb_has_compression(sbi)) 3523 return -EOPNOTSUPP; 3524 3525 if (!f2fs_compressed_file(inode)) 3526 return -EINVAL; 3527 3528 if (f2fs_readonly(sbi->sb)) 3529 return -EROFS; 3530 3531 ret = mnt_want_write_file(filp); 3532 if (ret) 3533 return ret; 3534 3535 f2fs_balance_fs(sbi, true); 3536 3537 inode_lock(inode); 3538 3539 writecount = atomic_read(&inode->i_writecount); 3540 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3541 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3542 ret = -EBUSY; 3543 goto out; 3544 } 3545 3546 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3547 ret = -EINVAL; 3548 goto out; 3549 } 3550 3551 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3552 if (ret) 3553 goto out; 3554 3555 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3556 ret = -EPERM; 3557 goto out; 3558 } 3559 3560 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3561 inode_set_ctime_current(inode); 3562 f2fs_mark_inode_dirty_sync(inode, true); 3563 3564 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3565 filemap_invalidate_lock(inode->i_mapping); 3566 3567 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3568 3569 while (page_idx < last_idx) { 3570 struct dnode_of_data dn; 3571 pgoff_t end_offset, count; 3572 3573 set_new_dnode(&dn, inode, NULL, NULL, 0); 3574 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3575 if (ret) { 3576 if (ret == -ENOENT) { 3577 page_idx = f2fs_get_next_page_offset(&dn, 3578 page_idx); 3579 ret = 0; 3580 continue; 3581 } 3582 break; 3583 } 3584 3585 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3586 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3587 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3588 3589 ret = release_compress_blocks(&dn, count); 3590 3591 f2fs_put_dnode(&dn); 3592 3593 if (ret < 0) 3594 break; 3595 3596 page_idx += count; 3597 released_blocks += ret; 3598 } 3599 3600 filemap_invalidate_unlock(inode->i_mapping); 3601 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3602 out: 3603 inode_unlock(inode); 3604 3605 mnt_drop_write_file(filp); 3606 3607 if (ret >= 0) { 3608 ret = put_user(released_blocks, (u64 __user *)arg); 3609 } else if (released_blocks && 3610 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3611 set_sbi_flag(sbi, SBI_NEED_FSCK); 3612 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3613 "iblocks=%llu, released=%u, compr_blocks=%u, " 3614 "run fsck to fix.", 3615 __func__, inode->i_ino, inode->i_blocks, 3616 released_blocks, 3617 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3618 } 3619 3620 return ret; 3621 } 3622 3623 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3624 unsigned int *reserved_blocks) 3625 { 3626 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3627 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3628 block_t blkaddr; 3629 int i; 3630 3631 for (i = 0; i < count; i++) { 3632 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3633 dn->ofs_in_node + i); 3634 3635 if (!__is_valid_data_blkaddr(blkaddr)) 3636 continue; 3637 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3638 DATA_GENERIC_ENHANCE))) 3639 return -EFSCORRUPTED; 3640 } 3641 3642 while (count) { 3643 int compr_blocks = 0; 3644 blkcnt_t reserved; 3645 int ret; 3646 3647 for (i = 0; i < cluster_size; i++) { 3648 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3649 dn->ofs_in_node + i); 3650 3651 if (i == 0) { 3652 if (blkaddr != COMPRESS_ADDR) { 3653 dn->ofs_in_node += cluster_size; 3654 goto next; 3655 } 3656 continue; 3657 } 3658 3659 /* 3660 * compressed cluster was not released due to it 3661 * fails in release_compress_blocks(), so NEW_ADDR 3662 * is a possible case. 3663 */ 3664 if (blkaddr == NEW_ADDR || 3665 __is_valid_data_blkaddr(blkaddr)) { 3666 compr_blocks++; 3667 continue; 3668 } 3669 } 3670 3671 reserved = cluster_size - compr_blocks; 3672 3673 /* for the case all blocks in cluster were reserved */ 3674 if (reserved == 1) 3675 goto next; 3676 3677 ret = inc_valid_block_count(sbi, dn->inode, &reserved, false); 3678 if (unlikely(ret)) 3679 return ret; 3680 3681 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3682 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3683 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3684 } 3685 3686 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3687 3688 *reserved_blocks += reserved; 3689 next: 3690 count -= cluster_size; 3691 } 3692 3693 return 0; 3694 } 3695 3696 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3697 { 3698 struct inode *inode = file_inode(filp); 3699 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3700 pgoff_t page_idx = 0, last_idx; 3701 unsigned int reserved_blocks = 0; 3702 int ret; 3703 3704 if (!f2fs_sb_has_compression(sbi)) 3705 return -EOPNOTSUPP; 3706 3707 if (!f2fs_compressed_file(inode)) 3708 return -EINVAL; 3709 3710 if (f2fs_readonly(sbi->sb)) 3711 return -EROFS; 3712 3713 ret = mnt_want_write_file(filp); 3714 if (ret) 3715 return ret; 3716 3717 f2fs_balance_fs(sbi, true); 3718 3719 inode_lock(inode); 3720 3721 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3722 ret = -EINVAL; 3723 goto unlock_inode; 3724 } 3725 3726 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3727 goto unlock_inode; 3728 3729 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3730 filemap_invalidate_lock(inode->i_mapping); 3731 3732 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3733 3734 while (page_idx < last_idx) { 3735 struct dnode_of_data dn; 3736 pgoff_t end_offset, count; 3737 3738 set_new_dnode(&dn, inode, NULL, NULL, 0); 3739 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3740 if (ret) { 3741 if (ret == -ENOENT) { 3742 page_idx = f2fs_get_next_page_offset(&dn, 3743 page_idx); 3744 ret = 0; 3745 continue; 3746 } 3747 break; 3748 } 3749 3750 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3751 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3752 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3753 3754 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 3755 3756 f2fs_put_dnode(&dn); 3757 3758 if (ret < 0) 3759 break; 3760 3761 page_idx += count; 3762 } 3763 3764 filemap_invalidate_unlock(inode->i_mapping); 3765 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3766 3767 if (!ret) { 3768 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3769 inode_set_ctime_current(inode); 3770 f2fs_mark_inode_dirty_sync(inode, true); 3771 } 3772 unlock_inode: 3773 inode_unlock(inode); 3774 mnt_drop_write_file(filp); 3775 3776 if (!ret) { 3777 ret = put_user(reserved_blocks, (u64 __user *)arg); 3778 } else if (reserved_blocks && 3779 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3780 set_sbi_flag(sbi, SBI_NEED_FSCK); 3781 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3782 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3783 "run fsck to fix.", 3784 __func__, inode->i_ino, inode->i_blocks, 3785 reserved_blocks, 3786 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3787 } 3788 3789 return ret; 3790 } 3791 3792 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3793 pgoff_t off, block_t block, block_t len, u32 flags) 3794 { 3795 sector_t sector = SECTOR_FROM_BLOCK(block); 3796 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3797 int ret = 0; 3798 3799 if (flags & F2FS_TRIM_FILE_DISCARD) { 3800 if (bdev_max_secure_erase_sectors(bdev)) 3801 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3802 GFP_NOFS); 3803 else 3804 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3805 GFP_NOFS); 3806 } 3807 3808 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3809 if (IS_ENCRYPTED(inode)) 3810 ret = fscrypt_zeroout_range(inode, off, block, len); 3811 else 3812 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3813 GFP_NOFS, 0); 3814 } 3815 3816 return ret; 3817 } 3818 3819 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3820 { 3821 struct inode *inode = file_inode(filp); 3822 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3823 struct address_space *mapping = inode->i_mapping; 3824 struct block_device *prev_bdev = NULL; 3825 struct f2fs_sectrim_range range; 3826 pgoff_t index, pg_end, prev_index = 0; 3827 block_t prev_block = 0, len = 0; 3828 loff_t end_addr; 3829 bool to_end = false; 3830 int ret = 0; 3831 3832 if (!(filp->f_mode & FMODE_WRITE)) 3833 return -EBADF; 3834 3835 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3836 sizeof(range))) 3837 return -EFAULT; 3838 3839 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3840 !S_ISREG(inode->i_mode)) 3841 return -EINVAL; 3842 3843 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3844 !f2fs_hw_support_discard(sbi)) || 3845 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3846 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3847 return -EOPNOTSUPP; 3848 3849 file_start_write(filp); 3850 inode_lock(inode); 3851 3852 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3853 range.start >= inode->i_size) { 3854 ret = -EINVAL; 3855 goto err; 3856 } 3857 3858 if (range.len == 0) 3859 goto err; 3860 3861 if (inode->i_size - range.start > range.len) { 3862 end_addr = range.start + range.len; 3863 } else { 3864 end_addr = range.len == (u64)-1 ? 3865 sbi->sb->s_maxbytes : inode->i_size; 3866 to_end = true; 3867 } 3868 3869 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3870 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3871 ret = -EINVAL; 3872 goto err; 3873 } 3874 3875 index = F2FS_BYTES_TO_BLK(range.start); 3876 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3877 3878 ret = f2fs_convert_inline_inode(inode); 3879 if (ret) 3880 goto err; 3881 3882 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3883 filemap_invalidate_lock(mapping); 3884 3885 ret = filemap_write_and_wait_range(mapping, range.start, 3886 to_end ? LLONG_MAX : end_addr - 1); 3887 if (ret) 3888 goto out; 3889 3890 truncate_inode_pages_range(mapping, range.start, 3891 to_end ? -1 : end_addr - 1); 3892 3893 while (index < pg_end) { 3894 struct dnode_of_data dn; 3895 pgoff_t end_offset, count; 3896 int i; 3897 3898 set_new_dnode(&dn, inode, NULL, NULL, 0); 3899 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3900 if (ret) { 3901 if (ret == -ENOENT) { 3902 index = f2fs_get_next_page_offset(&dn, index); 3903 continue; 3904 } 3905 goto out; 3906 } 3907 3908 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3909 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3910 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3911 struct block_device *cur_bdev; 3912 block_t blkaddr = f2fs_data_blkaddr(&dn); 3913 3914 if (!__is_valid_data_blkaddr(blkaddr)) 3915 continue; 3916 3917 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3918 DATA_GENERIC_ENHANCE)) { 3919 ret = -EFSCORRUPTED; 3920 f2fs_put_dnode(&dn); 3921 goto out; 3922 } 3923 3924 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3925 if (f2fs_is_multi_device(sbi)) { 3926 int di = f2fs_target_device_index(sbi, blkaddr); 3927 3928 blkaddr -= FDEV(di).start_blk; 3929 } 3930 3931 if (len) { 3932 if (prev_bdev == cur_bdev && 3933 index == prev_index + len && 3934 blkaddr == prev_block + len) { 3935 len++; 3936 } else { 3937 ret = f2fs_secure_erase(prev_bdev, 3938 inode, prev_index, prev_block, 3939 len, range.flags); 3940 if (ret) { 3941 f2fs_put_dnode(&dn); 3942 goto out; 3943 } 3944 3945 len = 0; 3946 } 3947 } 3948 3949 if (!len) { 3950 prev_bdev = cur_bdev; 3951 prev_index = index; 3952 prev_block = blkaddr; 3953 len = 1; 3954 } 3955 } 3956 3957 f2fs_put_dnode(&dn); 3958 3959 if (fatal_signal_pending(current)) { 3960 ret = -EINTR; 3961 goto out; 3962 } 3963 cond_resched(); 3964 } 3965 3966 if (len) 3967 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3968 prev_block, len, range.flags); 3969 out: 3970 filemap_invalidate_unlock(mapping); 3971 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3972 err: 3973 inode_unlock(inode); 3974 file_end_write(filp); 3975 3976 return ret; 3977 } 3978 3979 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3980 { 3981 struct inode *inode = file_inode(filp); 3982 struct f2fs_comp_option option; 3983 3984 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3985 return -EOPNOTSUPP; 3986 3987 inode_lock_shared(inode); 3988 3989 if (!f2fs_compressed_file(inode)) { 3990 inode_unlock_shared(inode); 3991 return -ENODATA; 3992 } 3993 3994 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3995 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3996 3997 inode_unlock_shared(inode); 3998 3999 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4000 sizeof(option))) 4001 return -EFAULT; 4002 4003 return 0; 4004 } 4005 4006 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4007 { 4008 struct inode *inode = file_inode(filp); 4009 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4010 struct f2fs_comp_option option; 4011 int ret = 0; 4012 4013 if (!f2fs_sb_has_compression(sbi)) 4014 return -EOPNOTSUPP; 4015 4016 if (!(filp->f_mode & FMODE_WRITE)) 4017 return -EBADF; 4018 4019 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4020 sizeof(option))) 4021 return -EFAULT; 4022 4023 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4024 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4025 option.algorithm >= COMPRESS_MAX) 4026 return -EINVAL; 4027 4028 file_start_write(filp); 4029 inode_lock(inode); 4030 4031 f2fs_down_write(&F2FS_I(inode)->i_sem); 4032 if (!f2fs_compressed_file(inode)) { 4033 ret = -EINVAL; 4034 goto out; 4035 } 4036 4037 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4038 ret = -EBUSY; 4039 goto out; 4040 } 4041 4042 if (F2FS_HAS_BLOCKS(inode)) { 4043 ret = -EFBIG; 4044 goto out; 4045 } 4046 4047 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4048 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4049 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4050 /* Set default level */ 4051 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4052 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4053 else 4054 F2FS_I(inode)->i_compress_level = 0; 4055 /* Adjust mount option level */ 4056 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4057 F2FS_OPTION(sbi).compress_level) 4058 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4059 f2fs_mark_inode_dirty_sync(inode, true); 4060 4061 if (!f2fs_is_compress_backend_ready(inode)) 4062 f2fs_warn(sbi, "compression algorithm is successfully set, " 4063 "but current kernel doesn't support this algorithm."); 4064 out: 4065 f2fs_up_write(&F2FS_I(inode)->i_sem); 4066 inode_unlock(inode); 4067 file_end_write(filp); 4068 4069 return ret; 4070 } 4071 4072 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4073 { 4074 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4075 struct address_space *mapping = inode->i_mapping; 4076 struct page *page; 4077 pgoff_t redirty_idx = page_idx; 4078 int i, page_len = 0, ret = 0; 4079 4080 page_cache_ra_unbounded(&ractl, len, 0); 4081 4082 for (i = 0; i < len; i++, page_idx++) { 4083 page = read_cache_page(mapping, page_idx, NULL, NULL); 4084 if (IS_ERR(page)) { 4085 ret = PTR_ERR(page); 4086 break; 4087 } 4088 page_len++; 4089 } 4090 4091 for (i = 0; i < page_len; i++, redirty_idx++) { 4092 page = find_lock_page(mapping, redirty_idx); 4093 4094 /* It will never fail, when page has pinned above */ 4095 f2fs_bug_on(F2FS_I_SB(inode), !page); 4096 4097 set_page_dirty(page); 4098 set_page_private_gcing(page); 4099 f2fs_put_page(page, 1); 4100 f2fs_put_page(page, 0); 4101 } 4102 4103 return ret; 4104 } 4105 4106 static int f2fs_ioc_decompress_file(struct file *filp) 4107 { 4108 struct inode *inode = file_inode(filp); 4109 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4110 struct f2fs_inode_info *fi = F2FS_I(inode); 4111 pgoff_t page_idx = 0, last_idx; 4112 int cluster_size = fi->i_cluster_size; 4113 int count, ret; 4114 4115 if (!f2fs_sb_has_compression(sbi) || 4116 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4117 return -EOPNOTSUPP; 4118 4119 if (!(filp->f_mode & FMODE_WRITE)) 4120 return -EBADF; 4121 4122 if (!f2fs_compressed_file(inode)) 4123 return -EINVAL; 4124 4125 f2fs_balance_fs(sbi, true); 4126 4127 file_start_write(filp); 4128 inode_lock(inode); 4129 4130 if (!f2fs_is_compress_backend_ready(inode)) { 4131 ret = -EOPNOTSUPP; 4132 goto out; 4133 } 4134 4135 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4136 ret = -EINVAL; 4137 goto out; 4138 } 4139 4140 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4141 if (ret) 4142 goto out; 4143 4144 if (!atomic_read(&fi->i_compr_blocks)) 4145 goto out; 4146 4147 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4148 4149 count = last_idx - page_idx; 4150 while (count && count >= cluster_size) { 4151 ret = redirty_blocks(inode, page_idx, cluster_size); 4152 if (ret < 0) 4153 break; 4154 4155 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4156 ret = filemap_fdatawrite(inode->i_mapping); 4157 if (ret < 0) 4158 break; 4159 } 4160 4161 count -= cluster_size; 4162 page_idx += cluster_size; 4163 4164 cond_resched(); 4165 if (fatal_signal_pending(current)) { 4166 ret = -EINTR; 4167 break; 4168 } 4169 } 4170 4171 if (!ret) 4172 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4173 LLONG_MAX); 4174 4175 if (ret) 4176 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4177 __func__, ret); 4178 out: 4179 inode_unlock(inode); 4180 file_end_write(filp); 4181 4182 return ret; 4183 } 4184 4185 static int f2fs_ioc_compress_file(struct file *filp) 4186 { 4187 struct inode *inode = file_inode(filp); 4188 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4189 pgoff_t page_idx = 0, last_idx; 4190 int cluster_size = F2FS_I(inode)->i_cluster_size; 4191 int count, ret; 4192 4193 if (!f2fs_sb_has_compression(sbi) || 4194 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4195 return -EOPNOTSUPP; 4196 4197 if (!(filp->f_mode & FMODE_WRITE)) 4198 return -EBADF; 4199 4200 if (!f2fs_compressed_file(inode)) 4201 return -EINVAL; 4202 4203 f2fs_balance_fs(sbi, true); 4204 4205 file_start_write(filp); 4206 inode_lock(inode); 4207 4208 if (!f2fs_is_compress_backend_ready(inode)) { 4209 ret = -EOPNOTSUPP; 4210 goto out; 4211 } 4212 4213 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4214 ret = -EINVAL; 4215 goto out; 4216 } 4217 4218 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4219 if (ret) 4220 goto out; 4221 4222 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4223 4224 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4225 4226 count = last_idx - page_idx; 4227 while (count && count >= cluster_size) { 4228 ret = redirty_blocks(inode, page_idx, cluster_size); 4229 if (ret < 0) 4230 break; 4231 4232 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4233 ret = filemap_fdatawrite(inode->i_mapping); 4234 if (ret < 0) 4235 break; 4236 } 4237 4238 count -= cluster_size; 4239 page_idx += cluster_size; 4240 4241 cond_resched(); 4242 if (fatal_signal_pending(current)) { 4243 ret = -EINTR; 4244 break; 4245 } 4246 } 4247 4248 if (!ret) 4249 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4250 LLONG_MAX); 4251 4252 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4253 4254 if (ret) 4255 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4256 __func__, ret); 4257 out: 4258 inode_unlock(inode); 4259 file_end_write(filp); 4260 4261 return ret; 4262 } 4263 4264 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4265 { 4266 switch (cmd) { 4267 case FS_IOC_GETVERSION: 4268 return f2fs_ioc_getversion(filp, arg); 4269 case F2FS_IOC_START_ATOMIC_WRITE: 4270 return f2fs_ioc_start_atomic_write(filp, false); 4271 case F2FS_IOC_START_ATOMIC_REPLACE: 4272 return f2fs_ioc_start_atomic_write(filp, true); 4273 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4274 return f2fs_ioc_commit_atomic_write(filp); 4275 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4276 return f2fs_ioc_abort_atomic_write(filp); 4277 case F2FS_IOC_START_VOLATILE_WRITE: 4278 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4279 return -EOPNOTSUPP; 4280 case F2FS_IOC_SHUTDOWN: 4281 return f2fs_ioc_shutdown(filp, arg); 4282 case FITRIM: 4283 return f2fs_ioc_fitrim(filp, arg); 4284 case FS_IOC_SET_ENCRYPTION_POLICY: 4285 return f2fs_ioc_set_encryption_policy(filp, arg); 4286 case FS_IOC_GET_ENCRYPTION_POLICY: 4287 return f2fs_ioc_get_encryption_policy(filp, arg); 4288 case FS_IOC_GET_ENCRYPTION_PWSALT: 4289 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4290 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4291 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4292 case FS_IOC_ADD_ENCRYPTION_KEY: 4293 return f2fs_ioc_add_encryption_key(filp, arg); 4294 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4295 return f2fs_ioc_remove_encryption_key(filp, arg); 4296 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4297 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4298 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4299 return f2fs_ioc_get_encryption_key_status(filp, arg); 4300 case FS_IOC_GET_ENCRYPTION_NONCE: 4301 return f2fs_ioc_get_encryption_nonce(filp, arg); 4302 case F2FS_IOC_GARBAGE_COLLECT: 4303 return f2fs_ioc_gc(filp, arg); 4304 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4305 return f2fs_ioc_gc_range(filp, arg); 4306 case F2FS_IOC_WRITE_CHECKPOINT: 4307 return f2fs_ioc_write_checkpoint(filp); 4308 case F2FS_IOC_DEFRAGMENT: 4309 return f2fs_ioc_defragment(filp, arg); 4310 case F2FS_IOC_MOVE_RANGE: 4311 return f2fs_ioc_move_range(filp, arg); 4312 case F2FS_IOC_FLUSH_DEVICE: 4313 return f2fs_ioc_flush_device(filp, arg); 4314 case F2FS_IOC_GET_FEATURES: 4315 return f2fs_ioc_get_features(filp, arg); 4316 case F2FS_IOC_GET_PIN_FILE: 4317 return f2fs_ioc_get_pin_file(filp, arg); 4318 case F2FS_IOC_SET_PIN_FILE: 4319 return f2fs_ioc_set_pin_file(filp, arg); 4320 case F2FS_IOC_PRECACHE_EXTENTS: 4321 return f2fs_ioc_precache_extents(filp); 4322 case F2FS_IOC_RESIZE_FS: 4323 return f2fs_ioc_resize_fs(filp, arg); 4324 case FS_IOC_ENABLE_VERITY: 4325 return f2fs_ioc_enable_verity(filp, arg); 4326 case FS_IOC_MEASURE_VERITY: 4327 return f2fs_ioc_measure_verity(filp, arg); 4328 case FS_IOC_READ_VERITY_METADATA: 4329 return f2fs_ioc_read_verity_metadata(filp, arg); 4330 case FS_IOC_GETFSLABEL: 4331 return f2fs_ioc_getfslabel(filp, arg); 4332 case FS_IOC_SETFSLABEL: 4333 return f2fs_ioc_setfslabel(filp, arg); 4334 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4335 return f2fs_ioc_get_compress_blocks(filp, arg); 4336 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4337 return f2fs_release_compress_blocks(filp, arg); 4338 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4339 return f2fs_reserve_compress_blocks(filp, arg); 4340 case F2FS_IOC_SEC_TRIM_FILE: 4341 return f2fs_sec_trim_file(filp, arg); 4342 case F2FS_IOC_GET_COMPRESS_OPTION: 4343 return f2fs_ioc_get_compress_option(filp, arg); 4344 case F2FS_IOC_SET_COMPRESS_OPTION: 4345 return f2fs_ioc_set_compress_option(filp, arg); 4346 case F2FS_IOC_DECOMPRESS_FILE: 4347 return f2fs_ioc_decompress_file(filp); 4348 case F2FS_IOC_COMPRESS_FILE: 4349 return f2fs_ioc_compress_file(filp); 4350 default: 4351 return -ENOTTY; 4352 } 4353 } 4354 4355 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4356 { 4357 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4358 return -EIO; 4359 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4360 return -ENOSPC; 4361 4362 return __f2fs_ioctl(filp, cmd, arg); 4363 } 4364 4365 /* 4366 * Return %true if the given read or write request should use direct I/O, or 4367 * %false if it should use buffered I/O. 4368 */ 4369 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4370 struct iov_iter *iter) 4371 { 4372 unsigned int align; 4373 4374 if (!(iocb->ki_flags & IOCB_DIRECT)) 4375 return false; 4376 4377 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4378 return false; 4379 4380 /* 4381 * Direct I/O not aligned to the disk's logical_block_size will be 4382 * attempted, but will fail with -EINVAL. 4383 * 4384 * f2fs additionally requires that direct I/O be aligned to the 4385 * filesystem block size, which is often a stricter requirement. 4386 * However, f2fs traditionally falls back to buffered I/O on requests 4387 * that are logical_block_size-aligned but not fs-block aligned. 4388 * 4389 * The below logic implements this behavior. 4390 */ 4391 align = iocb->ki_pos | iov_iter_alignment(iter); 4392 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4393 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4394 return false; 4395 4396 return true; 4397 } 4398 4399 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4400 unsigned int flags) 4401 { 4402 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4403 4404 dec_page_count(sbi, F2FS_DIO_READ); 4405 if (error) 4406 return error; 4407 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4408 return 0; 4409 } 4410 4411 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4412 .end_io = f2fs_dio_read_end_io, 4413 }; 4414 4415 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4416 { 4417 struct file *file = iocb->ki_filp; 4418 struct inode *inode = file_inode(file); 4419 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4420 struct f2fs_inode_info *fi = F2FS_I(inode); 4421 const loff_t pos = iocb->ki_pos; 4422 const size_t count = iov_iter_count(to); 4423 struct iomap_dio *dio; 4424 ssize_t ret; 4425 4426 if (count == 0) 4427 return 0; /* skip atime update */ 4428 4429 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4430 4431 if (iocb->ki_flags & IOCB_NOWAIT) { 4432 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4433 ret = -EAGAIN; 4434 goto out; 4435 } 4436 } else { 4437 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4438 } 4439 4440 /* 4441 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4442 * the higher-level function iomap_dio_rw() in order to ensure that the 4443 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4444 */ 4445 inc_page_count(sbi, F2FS_DIO_READ); 4446 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4447 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4448 if (IS_ERR_OR_NULL(dio)) { 4449 ret = PTR_ERR_OR_ZERO(dio); 4450 if (ret != -EIOCBQUEUED) 4451 dec_page_count(sbi, F2FS_DIO_READ); 4452 } else { 4453 ret = iomap_dio_complete(dio); 4454 } 4455 4456 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4457 4458 file_accessed(file); 4459 out: 4460 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4461 return ret; 4462 } 4463 4464 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4465 int rw) 4466 { 4467 struct inode *inode = file_inode(file); 4468 char *buf, *path; 4469 4470 buf = f2fs_getname(F2FS_I_SB(inode)); 4471 if (!buf) 4472 return; 4473 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4474 if (IS_ERR(path)) 4475 goto free_buf; 4476 if (rw == WRITE) 4477 trace_f2fs_datawrite_start(inode, pos, count, 4478 current->pid, path, current->comm); 4479 else 4480 trace_f2fs_dataread_start(inode, pos, count, 4481 current->pid, path, current->comm); 4482 free_buf: 4483 f2fs_putname(buf); 4484 } 4485 4486 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4487 { 4488 struct inode *inode = file_inode(iocb->ki_filp); 4489 const loff_t pos = iocb->ki_pos; 4490 ssize_t ret; 4491 4492 if (!f2fs_is_compress_backend_ready(inode)) 4493 return -EOPNOTSUPP; 4494 4495 if (trace_f2fs_dataread_start_enabled()) 4496 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4497 iov_iter_count(to), READ); 4498 4499 if (f2fs_should_use_dio(inode, iocb, to)) { 4500 ret = f2fs_dio_read_iter(iocb, to); 4501 } else { 4502 ret = filemap_read(iocb, to, 0); 4503 if (ret > 0) 4504 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4505 APP_BUFFERED_READ_IO, ret); 4506 } 4507 if (trace_f2fs_dataread_end_enabled()) 4508 trace_f2fs_dataread_end(inode, pos, ret); 4509 return ret; 4510 } 4511 4512 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4513 struct pipe_inode_info *pipe, 4514 size_t len, unsigned int flags) 4515 { 4516 struct inode *inode = file_inode(in); 4517 const loff_t pos = *ppos; 4518 ssize_t ret; 4519 4520 if (!f2fs_is_compress_backend_ready(inode)) 4521 return -EOPNOTSUPP; 4522 4523 if (trace_f2fs_dataread_start_enabled()) 4524 f2fs_trace_rw_file_path(in, pos, len, READ); 4525 4526 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4527 if (ret > 0) 4528 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4529 APP_BUFFERED_READ_IO, ret); 4530 4531 if (trace_f2fs_dataread_end_enabled()) 4532 trace_f2fs_dataread_end(inode, pos, ret); 4533 return ret; 4534 } 4535 4536 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4537 { 4538 struct file *file = iocb->ki_filp; 4539 struct inode *inode = file_inode(file); 4540 ssize_t count; 4541 int err; 4542 4543 if (IS_IMMUTABLE(inode)) 4544 return -EPERM; 4545 4546 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4547 return -EPERM; 4548 4549 count = generic_write_checks(iocb, from); 4550 if (count <= 0) 4551 return count; 4552 4553 err = file_modified(file); 4554 if (err) 4555 return err; 4556 return count; 4557 } 4558 4559 /* 4560 * Preallocate blocks for a write request, if it is possible and helpful to do 4561 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4562 * blocks were preallocated, or a negative errno value if something went 4563 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4564 * requested blocks (not just some of them) have been allocated. 4565 */ 4566 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4567 bool dio) 4568 { 4569 struct inode *inode = file_inode(iocb->ki_filp); 4570 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4571 const loff_t pos = iocb->ki_pos; 4572 const size_t count = iov_iter_count(iter); 4573 struct f2fs_map_blocks map = {}; 4574 int flag; 4575 int ret; 4576 4577 /* If it will be an out-of-place direct write, don't bother. */ 4578 if (dio && f2fs_lfs_mode(sbi)) 4579 return 0; 4580 /* 4581 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4582 * buffered IO, if DIO meets any holes. 4583 */ 4584 if (dio && i_size_read(inode) && 4585 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4586 return 0; 4587 4588 /* No-wait I/O can't allocate blocks. */ 4589 if (iocb->ki_flags & IOCB_NOWAIT) 4590 return 0; 4591 4592 /* If it will be a short write, don't bother. */ 4593 if (fault_in_iov_iter_readable(iter, count)) 4594 return 0; 4595 4596 if (f2fs_has_inline_data(inode)) { 4597 /* If the data will fit inline, don't bother. */ 4598 if (pos + count <= MAX_INLINE_DATA(inode)) 4599 return 0; 4600 ret = f2fs_convert_inline_inode(inode); 4601 if (ret) 4602 return ret; 4603 } 4604 4605 /* Do not preallocate blocks that will be written partially in 4KB. */ 4606 map.m_lblk = F2FS_BLK_ALIGN(pos); 4607 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4608 if (map.m_len > map.m_lblk) 4609 map.m_len -= map.m_lblk; 4610 else 4611 return 0; 4612 4613 map.m_may_create = true; 4614 if (dio) { 4615 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4616 flag = F2FS_GET_BLOCK_PRE_DIO; 4617 } else { 4618 map.m_seg_type = NO_CHECK_TYPE; 4619 flag = F2FS_GET_BLOCK_PRE_AIO; 4620 } 4621 4622 ret = f2fs_map_blocks(inode, &map, flag); 4623 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4624 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4625 return ret; 4626 if (ret == 0) 4627 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4628 return map.m_len; 4629 } 4630 4631 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4632 struct iov_iter *from) 4633 { 4634 struct file *file = iocb->ki_filp; 4635 struct inode *inode = file_inode(file); 4636 ssize_t ret; 4637 4638 if (iocb->ki_flags & IOCB_NOWAIT) 4639 return -EOPNOTSUPP; 4640 4641 ret = generic_perform_write(iocb, from); 4642 4643 if (ret > 0) { 4644 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4645 APP_BUFFERED_IO, ret); 4646 } 4647 return ret; 4648 } 4649 4650 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4651 unsigned int flags) 4652 { 4653 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4654 4655 dec_page_count(sbi, F2FS_DIO_WRITE); 4656 if (error) 4657 return error; 4658 f2fs_update_time(sbi, REQ_TIME); 4659 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4660 return 0; 4661 } 4662 4663 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4664 .end_io = f2fs_dio_write_end_io, 4665 }; 4666 4667 static void f2fs_flush_buffered_write(struct address_space *mapping, 4668 loff_t start_pos, loff_t end_pos) 4669 { 4670 int ret; 4671 4672 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4673 if (ret < 0) 4674 return; 4675 invalidate_mapping_pages(mapping, 4676 start_pos >> PAGE_SHIFT, 4677 end_pos >> PAGE_SHIFT); 4678 } 4679 4680 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4681 bool *may_need_sync) 4682 { 4683 struct file *file = iocb->ki_filp; 4684 struct inode *inode = file_inode(file); 4685 struct f2fs_inode_info *fi = F2FS_I(inode); 4686 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4687 const bool do_opu = f2fs_lfs_mode(sbi); 4688 const loff_t pos = iocb->ki_pos; 4689 const ssize_t count = iov_iter_count(from); 4690 unsigned int dio_flags; 4691 struct iomap_dio *dio; 4692 ssize_t ret; 4693 4694 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4695 4696 if (iocb->ki_flags & IOCB_NOWAIT) { 4697 /* f2fs_convert_inline_inode() and block allocation can block */ 4698 if (f2fs_has_inline_data(inode) || 4699 !f2fs_overwrite_io(inode, pos, count)) { 4700 ret = -EAGAIN; 4701 goto out; 4702 } 4703 4704 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4705 ret = -EAGAIN; 4706 goto out; 4707 } 4708 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4709 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4710 ret = -EAGAIN; 4711 goto out; 4712 } 4713 } else { 4714 ret = f2fs_convert_inline_inode(inode); 4715 if (ret) 4716 goto out; 4717 4718 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4719 if (do_opu) 4720 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4721 } 4722 4723 /* 4724 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4725 * the higher-level function iomap_dio_rw() in order to ensure that the 4726 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4727 */ 4728 inc_page_count(sbi, F2FS_DIO_WRITE); 4729 dio_flags = 0; 4730 if (pos + count > inode->i_size) 4731 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4732 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4733 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4734 if (IS_ERR_OR_NULL(dio)) { 4735 ret = PTR_ERR_OR_ZERO(dio); 4736 if (ret == -ENOTBLK) 4737 ret = 0; 4738 if (ret != -EIOCBQUEUED) 4739 dec_page_count(sbi, F2FS_DIO_WRITE); 4740 } else { 4741 ret = iomap_dio_complete(dio); 4742 } 4743 4744 if (do_opu) 4745 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4746 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4747 4748 if (ret < 0) 4749 goto out; 4750 if (pos + ret > inode->i_size) 4751 f2fs_i_size_write(inode, pos + ret); 4752 if (!do_opu) 4753 set_inode_flag(inode, FI_UPDATE_WRITE); 4754 4755 if (iov_iter_count(from)) { 4756 ssize_t ret2; 4757 loff_t bufio_start_pos = iocb->ki_pos; 4758 4759 /* 4760 * The direct write was partial, so we need to fall back to a 4761 * buffered write for the remainder. 4762 */ 4763 4764 ret2 = f2fs_buffered_write_iter(iocb, from); 4765 if (iov_iter_count(from)) 4766 f2fs_write_failed(inode, iocb->ki_pos); 4767 if (ret2 < 0) 4768 goto out; 4769 4770 /* 4771 * Ensure that the pagecache pages are written to disk and 4772 * invalidated to preserve the expected O_DIRECT semantics. 4773 */ 4774 if (ret2 > 0) { 4775 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4776 4777 ret += ret2; 4778 4779 f2fs_flush_buffered_write(file->f_mapping, 4780 bufio_start_pos, 4781 bufio_end_pos); 4782 } 4783 } else { 4784 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4785 *may_need_sync = false; 4786 } 4787 out: 4788 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4789 return ret; 4790 } 4791 4792 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4793 { 4794 struct inode *inode = file_inode(iocb->ki_filp); 4795 const loff_t orig_pos = iocb->ki_pos; 4796 const size_t orig_count = iov_iter_count(from); 4797 loff_t target_size; 4798 bool dio; 4799 bool may_need_sync = true; 4800 int preallocated; 4801 ssize_t ret; 4802 4803 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4804 ret = -EIO; 4805 goto out; 4806 } 4807 4808 if (!f2fs_is_compress_backend_ready(inode)) { 4809 ret = -EOPNOTSUPP; 4810 goto out; 4811 } 4812 4813 if (iocb->ki_flags & IOCB_NOWAIT) { 4814 if (!inode_trylock(inode)) { 4815 ret = -EAGAIN; 4816 goto out; 4817 } 4818 } else { 4819 inode_lock(inode); 4820 } 4821 4822 ret = f2fs_write_checks(iocb, from); 4823 if (ret <= 0) 4824 goto out_unlock; 4825 4826 /* Determine whether we will do a direct write or a buffered write. */ 4827 dio = f2fs_should_use_dio(inode, iocb, from); 4828 4829 /* Possibly preallocate the blocks for the write. */ 4830 target_size = iocb->ki_pos + iov_iter_count(from); 4831 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4832 if (preallocated < 0) { 4833 ret = preallocated; 4834 } else { 4835 if (trace_f2fs_datawrite_start_enabled()) 4836 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4837 orig_count, WRITE); 4838 4839 /* Do the actual write. */ 4840 ret = dio ? 4841 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4842 f2fs_buffered_write_iter(iocb, from); 4843 4844 if (trace_f2fs_datawrite_end_enabled()) 4845 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4846 } 4847 4848 /* Don't leave any preallocated blocks around past i_size. */ 4849 if (preallocated && i_size_read(inode) < target_size) { 4850 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4851 filemap_invalidate_lock(inode->i_mapping); 4852 if (!f2fs_truncate(inode)) 4853 file_dont_truncate(inode); 4854 filemap_invalidate_unlock(inode->i_mapping); 4855 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4856 } else { 4857 file_dont_truncate(inode); 4858 } 4859 4860 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4861 out_unlock: 4862 inode_unlock(inode); 4863 out: 4864 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4865 4866 if (ret > 0 && may_need_sync) 4867 ret = generic_write_sync(iocb, ret); 4868 4869 /* If buffered IO was forced, flush and drop the data from 4870 * the page cache to preserve O_DIRECT semantics 4871 */ 4872 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4873 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4874 orig_pos, 4875 orig_pos + ret - 1); 4876 4877 return ret; 4878 } 4879 4880 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4881 int advice) 4882 { 4883 struct address_space *mapping; 4884 struct backing_dev_info *bdi; 4885 struct inode *inode = file_inode(filp); 4886 int err; 4887 4888 if (advice == POSIX_FADV_SEQUENTIAL) { 4889 if (S_ISFIFO(inode->i_mode)) 4890 return -ESPIPE; 4891 4892 mapping = filp->f_mapping; 4893 if (!mapping || len < 0) 4894 return -EINVAL; 4895 4896 bdi = inode_to_bdi(mapping->host); 4897 filp->f_ra.ra_pages = bdi->ra_pages * 4898 F2FS_I_SB(inode)->seq_file_ra_mul; 4899 spin_lock(&filp->f_lock); 4900 filp->f_mode &= ~FMODE_RANDOM; 4901 spin_unlock(&filp->f_lock); 4902 return 0; 4903 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 4904 /* Load extent cache at the first readahead. */ 4905 f2fs_precache_extents(inode); 4906 } 4907 4908 err = generic_fadvise(filp, offset, len, advice); 4909 if (!err && advice == POSIX_FADV_DONTNEED && 4910 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4911 f2fs_compressed_file(inode)) 4912 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4913 4914 return err; 4915 } 4916 4917 #ifdef CONFIG_COMPAT 4918 struct compat_f2fs_gc_range { 4919 u32 sync; 4920 compat_u64 start; 4921 compat_u64 len; 4922 }; 4923 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4924 struct compat_f2fs_gc_range) 4925 4926 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4927 { 4928 struct compat_f2fs_gc_range __user *urange; 4929 struct f2fs_gc_range range; 4930 int err; 4931 4932 urange = compat_ptr(arg); 4933 err = get_user(range.sync, &urange->sync); 4934 err |= get_user(range.start, &urange->start); 4935 err |= get_user(range.len, &urange->len); 4936 if (err) 4937 return -EFAULT; 4938 4939 return __f2fs_ioc_gc_range(file, &range); 4940 } 4941 4942 struct compat_f2fs_move_range { 4943 u32 dst_fd; 4944 compat_u64 pos_in; 4945 compat_u64 pos_out; 4946 compat_u64 len; 4947 }; 4948 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4949 struct compat_f2fs_move_range) 4950 4951 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4952 { 4953 struct compat_f2fs_move_range __user *urange; 4954 struct f2fs_move_range range; 4955 int err; 4956 4957 urange = compat_ptr(arg); 4958 err = get_user(range.dst_fd, &urange->dst_fd); 4959 err |= get_user(range.pos_in, &urange->pos_in); 4960 err |= get_user(range.pos_out, &urange->pos_out); 4961 err |= get_user(range.len, &urange->len); 4962 if (err) 4963 return -EFAULT; 4964 4965 return __f2fs_ioc_move_range(file, &range); 4966 } 4967 4968 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4969 { 4970 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4971 return -EIO; 4972 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4973 return -ENOSPC; 4974 4975 switch (cmd) { 4976 case FS_IOC32_GETVERSION: 4977 cmd = FS_IOC_GETVERSION; 4978 break; 4979 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4980 return f2fs_compat_ioc_gc_range(file, arg); 4981 case F2FS_IOC32_MOVE_RANGE: 4982 return f2fs_compat_ioc_move_range(file, arg); 4983 case F2FS_IOC_START_ATOMIC_WRITE: 4984 case F2FS_IOC_START_ATOMIC_REPLACE: 4985 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4986 case F2FS_IOC_START_VOLATILE_WRITE: 4987 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4988 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4989 case F2FS_IOC_SHUTDOWN: 4990 case FITRIM: 4991 case FS_IOC_SET_ENCRYPTION_POLICY: 4992 case FS_IOC_GET_ENCRYPTION_PWSALT: 4993 case FS_IOC_GET_ENCRYPTION_POLICY: 4994 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4995 case FS_IOC_ADD_ENCRYPTION_KEY: 4996 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4997 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4998 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4999 case FS_IOC_GET_ENCRYPTION_NONCE: 5000 case F2FS_IOC_GARBAGE_COLLECT: 5001 case F2FS_IOC_WRITE_CHECKPOINT: 5002 case F2FS_IOC_DEFRAGMENT: 5003 case F2FS_IOC_FLUSH_DEVICE: 5004 case F2FS_IOC_GET_FEATURES: 5005 case F2FS_IOC_GET_PIN_FILE: 5006 case F2FS_IOC_SET_PIN_FILE: 5007 case F2FS_IOC_PRECACHE_EXTENTS: 5008 case F2FS_IOC_RESIZE_FS: 5009 case FS_IOC_ENABLE_VERITY: 5010 case FS_IOC_MEASURE_VERITY: 5011 case FS_IOC_READ_VERITY_METADATA: 5012 case FS_IOC_GETFSLABEL: 5013 case FS_IOC_SETFSLABEL: 5014 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5015 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5016 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5017 case F2FS_IOC_SEC_TRIM_FILE: 5018 case F2FS_IOC_GET_COMPRESS_OPTION: 5019 case F2FS_IOC_SET_COMPRESS_OPTION: 5020 case F2FS_IOC_DECOMPRESS_FILE: 5021 case F2FS_IOC_COMPRESS_FILE: 5022 break; 5023 default: 5024 return -ENOIOCTLCMD; 5025 } 5026 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5027 } 5028 #endif 5029 5030 const struct file_operations f2fs_file_operations = { 5031 .llseek = f2fs_llseek, 5032 .read_iter = f2fs_file_read_iter, 5033 .write_iter = f2fs_file_write_iter, 5034 .iopoll = iocb_bio_iopoll, 5035 .open = f2fs_file_open, 5036 .release = f2fs_release_file, 5037 .mmap = f2fs_file_mmap, 5038 .flush = f2fs_file_flush, 5039 .fsync = f2fs_sync_file, 5040 .fallocate = f2fs_fallocate, 5041 .unlocked_ioctl = f2fs_ioctl, 5042 #ifdef CONFIG_COMPAT 5043 .compat_ioctl = f2fs_compat_ioctl, 5044 #endif 5045 .splice_read = f2fs_file_splice_read, 5046 .splice_write = iter_file_splice_write, 5047 .fadvise = f2fs_file_fadvise, 5048 }; 5049