1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/file.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/fs.h> 9 #include <linux/f2fs_fs.h> 10 #include <linux/stat.h> 11 #include <linux/buffer_head.h> 12 #include <linux/writeback.h> 13 #include <linux/blkdev.h> 14 #include <linux/falloc.h> 15 #include <linux/types.h> 16 #include <linux/compat.h> 17 #include <linux/uaccess.h> 18 #include <linux/mount.h> 19 #include <linux/pagevec.h> 20 #include <linux/uio.h> 21 #include <linux/uuid.h> 22 #include <linux/file.h> 23 #include <linux/nls.h> 24 #include <linux/sched/signal.h> 25 #include <linux/fileattr.h> 26 #include <linux/fadvise.h> 27 #include <linux/iomap.h> 28 29 #include "f2fs.h" 30 #include "node.h" 31 #include "segment.h" 32 #include "xattr.h" 33 #include "acl.h" 34 #include "gc.h" 35 #include "iostat.h" 36 #include <trace/events/f2fs.h> 37 #include <uapi/linux/f2fs.h> 38 39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 40 { 41 struct inode *inode = file_inode(vmf->vma->vm_file); 42 vm_fault_t ret; 43 44 ret = filemap_fault(vmf); 45 if (ret & VM_FAULT_LOCKED) 46 f2fs_update_iostat(F2FS_I_SB(inode), inode, 47 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 48 49 trace_f2fs_filemap_fault(inode, vmf->pgoff, vmf->vma->vm_flags, ret); 50 51 return ret; 52 } 53 54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 55 { 56 struct page *page = vmf->page; 57 struct inode *inode = file_inode(vmf->vma->vm_file); 58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 59 struct dnode_of_data dn; 60 bool need_alloc = true; 61 int err = 0; 62 vm_fault_t ret; 63 64 if (unlikely(IS_IMMUTABLE(inode))) 65 return VM_FAULT_SIGBUS; 66 67 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 68 err = -EIO; 69 goto out; 70 } 71 72 if (unlikely(f2fs_cp_error(sbi))) { 73 err = -EIO; 74 goto out; 75 } 76 77 if (!f2fs_is_checkpoint_ready(sbi)) { 78 err = -ENOSPC; 79 goto out; 80 } 81 82 err = f2fs_convert_inline_inode(inode); 83 if (err) 84 goto out; 85 86 #ifdef CONFIG_F2FS_FS_COMPRESSION 87 if (f2fs_compressed_file(inode)) { 88 int ret = f2fs_is_compressed_cluster(inode, page->index); 89 90 if (ret < 0) { 91 err = ret; 92 goto out; 93 } else if (ret) { 94 need_alloc = false; 95 } 96 } 97 #endif 98 /* should do out of any locked page */ 99 if (need_alloc) 100 f2fs_balance_fs(sbi, true); 101 102 sb_start_pagefault(inode->i_sb); 103 104 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 105 106 file_update_time(vmf->vma->vm_file); 107 filemap_invalidate_lock_shared(inode->i_mapping); 108 lock_page(page); 109 if (unlikely(page->mapping != inode->i_mapping || 110 page_offset(page) > i_size_read(inode) || 111 !PageUptodate(page))) { 112 unlock_page(page); 113 err = -EFAULT; 114 goto out_sem; 115 } 116 117 if (need_alloc) { 118 /* block allocation */ 119 set_new_dnode(&dn, inode, NULL, NULL, 0); 120 err = f2fs_get_block_locked(&dn, page->index); 121 } 122 123 #ifdef CONFIG_F2FS_FS_COMPRESSION 124 if (!need_alloc) { 125 set_new_dnode(&dn, inode, NULL, NULL, 0); 126 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE); 127 f2fs_put_dnode(&dn); 128 } 129 #endif 130 if (err) { 131 unlock_page(page); 132 goto out_sem; 133 } 134 135 f2fs_wait_on_page_writeback(page, DATA, false, true); 136 137 /* wait for GCed page writeback via META_MAPPING */ 138 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 139 140 /* 141 * check to see if the page is mapped already (no holes) 142 */ 143 if (PageMappedToDisk(page)) 144 goto out_sem; 145 146 /* page is wholly or partially inside EOF */ 147 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 148 i_size_read(inode)) { 149 loff_t offset; 150 151 offset = i_size_read(inode) & ~PAGE_MASK; 152 zero_user_segment(page, offset, PAGE_SIZE); 153 } 154 set_page_dirty(page); 155 156 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 157 f2fs_update_time(sbi, REQ_TIME); 158 159 out_sem: 160 filemap_invalidate_unlock_shared(inode->i_mapping); 161 162 sb_end_pagefault(inode->i_sb); 163 out: 164 ret = vmf_fs_error(err); 165 166 trace_f2fs_vm_page_mkwrite(inode, page->index, vmf->vma->vm_flags, ret); 167 return ret; 168 } 169 170 static const struct vm_operations_struct f2fs_file_vm_ops = { 171 .fault = f2fs_filemap_fault, 172 .map_pages = filemap_map_pages, 173 .page_mkwrite = f2fs_vm_page_mkwrite, 174 }; 175 176 static int get_parent_ino(struct inode *inode, nid_t *pino) 177 { 178 struct dentry *dentry; 179 180 /* 181 * Make sure to get the non-deleted alias. The alias associated with 182 * the open file descriptor being fsync()'ed may be deleted already. 183 */ 184 dentry = d_find_alias(inode); 185 if (!dentry) 186 return 0; 187 188 *pino = parent_ino(dentry); 189 dput(dentry); 190 return 1; 191 } 192 193 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 194 { 195 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 196 enum cp_reason_type cp_reason = CP_NO_NEEDED; 197 198 if (!S_ISREG(inode->i_mode)) 199 cp_reason = CP_NON_REGULAR; 200 else if (f2fs_compressed_file(inode)) 201 cp_reason = CP_COMPRESSED; 202 else if (inode->i_nlink != 1) 203 cp_reason = CP_HARDLINK; 204 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 205 cp_reason = CP_SB_NEED_CP; 206 else if (file_wrong_pino(inode)) 207 cp_reason = CP_WRONG_PINO; 208 else if (!f2fs_space_for_roll_forward(sbi)) 209 cp_reason = CP_NO_SPC_ROLL; 210 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 211 cp_reason = CP_NODE_NEED_CP; 212 else if (test_opt(sbi, FASTBOOT)) 213 cp_reason = CP_FASTBOOT_MODE; 214 else if (F2FS_OPTION(sbi).active_logs == 2) 215 cp_reason = CP_SPEC_LOG_NUM; 216 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 217 f2fs_need_dentry_mark(sbi, inode->i_ino) && 218 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 219 TRANS_DIR_INO)) 220 cp_reason = CP_RECOVER_DIR; 221 222 return cp_reason; 223 } 224 225 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 226 { 227 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 228 bool ret = false; 229 /* But we need to avoid that there are some inode updates */ 230 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 231 ret = true; 232 f2fs_put_page(i, 0); 233 return ret; 234 } 235 236 static void try_to_fix_pino(struct inode *inode) 237 { 238 struct f2fs_inode_info *fi = F2FS_I(inode); 239 nid_t pino; 240 241 f2fs_down_write(&fi->i_sem); 242 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 243 get_parent_ino(inode, &pino)) { 244 f2fs_i_pino_write(inode, pino); 245 file_got_pino(inode); 246 } 247 f2fs_up_write(&fi->i_sem); 248 } 249 250 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 251 int datasync, bool atomic) 252 { 253 struct inode *inode = file->f_mapping->host; 254 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 255 nid_t ino = inode->i_ino; 256 int ret = 0; 257 enum cp_reason_type cp_reason = 0; 258 struct writeback_control wbc = { 259 .sync_mode = WB_SYNC_ALL, 260 .nr_to_write = LONG_MAX, 261 .for_reclaim = 0, 262 }; 263 unsigned int seq_id = 0; 264 265 if (unlikely(f2fs_readonly(inode->i_sb))) 266 return 0; 267 268 trace_f2fs_sync_file_enter(inode); 269 270 if (S_ISDIR(inode->i_mode)) 271 goto go_write; 272 273 /* if fdatasync is triggered, let's do in-place-update */ 274 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 275 set_inode_flag(inode, FI_NEED_IPU); 276 ret = file_write_and_wait_range(file, start, end); 277 clear_inode_flag(inode, FI_NEED_IPU); 278 279 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 280 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 281 return ret; 282 } 283 284 /* if the inode is dirty, let's recover all the time */ 285 if (!f2fs_skip_inode_update(inode, datasync)) { 286 f2fs_write_inode(inode, NULL); 287 goto go_write; 288 } 289 290 /* 291 * if there is no written data, don't waste time to write recovery info. 292 */ 293 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 294 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 295 296 /* it may call write_inode just prior to fsync */ 297 if (need_inode_page_update(sbi, ino)) 298 goto go_write; 299 300 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 301 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 302 goto flush_out; 303 goto out; 304 } else { 305 /* 306 * for OPU case, during fsync(), node can be persisted before 307 * data when lower device doesn't support write barrier, result 308 * in data corruption after SPO. 309 * So for strict fsync mode, force to use atomic write semantics 310 * to keep write order in between data/node and last node to 311 * avoid potential data corruption. 312 */ 313 if (F2FS_OPTION(sbi).fsync_mode == 314 FSYNC_MODE_STRICT && !atomic) 315 atomic = true; 316 } 317 go_write: 318 /* 319 * Both of fdatasync() and fsync() are able to be recovered from 320 * sudden-power-off. 321 */ 322 f2fs_down_read(&F2FS_I(inode)->i_sem); 323 cp_reason = need_do_checkpoint(inode); 324 f2fs_up_read(&F2FS_I(inode)->i_sem); 325 326 if (cp_reason) { 327 /* all the dirty node pages should be flushed for POR */ 328 ret = f2fs_sync_fs(inode->i_sb, 1); 329 330 /* 331 * We've secured consistency through sync_fs. Following pino 332 * will be used only for fsynced inodes after checkpoint. 333 */ 334 try_to_fix_pino(inode); 335 clear_inode_flag(inode, FI_APPEND_WRITE); 336 clear_inode_flag(inode, FI_UPDATE_WRITE); 337 goto out; 338 } 339 sync_nodes: 340 atomic_inc(&sbi->wb_sync_req[NODE]); 341 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 342 atomic_dec(&sbi->wb_sync_req[NODE]); 343 if (ret) 344 goto out; 345 346 /* if cp_error was enabled, we should avoid infinite loop */ 347 if (unlikely(f2fs_cp_error(sbi))) { 348 ret = -EIO; 349 goto out; 350 } 351 352 if (f2fs_need_inode_block_update(sbi, ino)) { 353 f2fs_mark_inode_dirty_sync(inode, true); 354 f2fs_write_inode(inode, NULL); 355 goto sync_nodes; 356 } 357 358 /* 359 * If it's atomic_write, it's just fine to keep write ordering. So 360 * here we don't need to wait for node write completion, since we use 361 * node chain which serializes node blocks. If one of node writes are 362 * reordered, we can see simply broken chain, resulting in stopping 363 * roll-forward recovery. It means we'll recover all or none node blocks 364 * given fsync mark. 365 */ 366 if (!atomic) { 367 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 368 if (ret) 369 goto out; 370 } 371 372 /* once recovery info is written, don't need to tack this */ 373 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 374 clear_inode_flag(inode, FI_APPEND_WRITE); 375 flush_out: 376 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) || 377 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi))) 378 ret = f2fs_issue_flush(sbi, inode->i_ino); 379 if (!ret) { 380 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 381 clear_inode_flag(inode, FI_UPDATE_WRITE); 382 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 383 } 384 f2fs_update_time(sbi, REQ_TIME); 385 out: 386 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 387 return ret; 388 } 389 390 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 391 { 392 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 393 return -EIO; 394 return f2fs_do_sync_file(file, start, end, datasync, false); 395 } 396 397 static bool __found_offset(struct address_space *mapping, block_t blkaddr, 398 pgoff_t index, int whence) 399 { 400 switch (whence) { 401 case SEEK_DATA: 402 if (__is_valid_data_blkaddr(blkaddr)) 403 return true; 404 if (blkaddr == NEW_ADDR && 405 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 406 return true; 407 break; 408 case SEEK_HOLE: 409 if (blkaddr == NULL_ADDR) 410 return true; 411 break; 412 } 413 return false; 414 } 415 416 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 417 { 418 struct inode *inode = file->f_mapping->host; 419 loff_t maxbytes = inode->i_sb->s_maxbytes; 420 struct dnode_of_data dn; 421 pgoff_t pgofs, end_offset; 422 loff_t data_ofs = offset; 423 loff_t isize; 424 int err = 0; 425 426 inode_lock_shared(inode); 427 428 isize = i_size_read(inode); 429 if (offset >= isize) 430 goto fail; 431 432 /* handle inline data case */ 433 if (f2fs_has_inline_data(inode)) { 434 if (whence == SEEK_HOLE) { 435 data_ofs = isize; 436 goto found; 437 } else if (whence == SEEK_DATA) { 438 data_ofs = offset; 439 goto found; 440 } 441 } 442 443 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 444 445 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 446 set_new_dnode(&dn, inode, NULL, NULL, 0); 447 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 448 if (err && err != -ENOENT) { 449 goto fail; 450 } else if (err == -ENOENT) { 451 /* direct node does not exists */ 452 if (whence == SEEK_DATA) { 453 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 454 continue; 455 } else { 456 goto found; 457 } 458 } 459 460 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 461 462 /* find data/hole in dnode block */ 463 for (; dn.ofs_in_node < end_offset; 464 dn.ofs_in_node++, pgofs++, 465 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 466 block_t blkaddr; 467 468 blkaddr = f2fs_data_blkaddr(&dn); 469 470 if (__is_valid_data_blkaddr(blkaddr) && 471 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 472 blkaddr, DATA_GENERIC_ENHANCE)) { 473 f2fs_put_dnode(&dn); 474 goto fail; 475 } 476 477 if (__found_offset(file->f_mapping, blkaddr, 478 pgofs, whence)) { 479 f2fs_put_dnode(&dn); 480 goto found; 481 } 482 } 483 f2fs_put_dnode(&dn); 484 } 485 486 if (whence == SEEK_DATA) 487 goto fail; 488 found: 489 if (whence == SEEK_HOLE && data_ofs > isize) 490 data_ofs = isize; 491 inode_unlock_shared(inode); 492 return vfs_setpos(file, data_ofs, maxbytes); 493 fail: 494 inode_unlock_shared(inode); 495 return -ENXIO; 496 } 497 498 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 499 { 500 struct inode *inode = file->f_mapping->host; 501 loff_t maxbytes = inode->i_sb->s_maxbytes; 502 503 if (f2fs_compressed_file(inode)) 504 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; 505 506 switch (whence) { 507 case SEEK_SET: 508 case SEEK_CUR: 509 case SEEK_END: 510 return generic_file_llseek_size(file, offset, whence, 511 maxbytes, i_size_read(inode)); 512 case SEEK_DATA: 513 case SEEK_HOLE: 514 if (offset < 0) 515 return -ENXIO; 516 return f2fs_seek_block(file, offset, whence); 517 } 518 519 return -EINVAL; 520 } 521 522 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 523 { 524 struct inode *inode = file_inode(file); 525 526 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 527 return -EIO; 528 529 if (!f2fs_is_compress_backend_ready(inode)) 530 return -EOPNOTSUPP; 531 532 file_accessed(file); 533 vma->vm_ops = &f2fs_file_vm_ops; 534 535 f2fs_down_read(&F2FS_I(inode)->i_sem); 536 set_inode_flag(inode, FI_MMAP_FILE); 537 f2fs_up_read(&F2FS_I(inode)->i_sem); 538 539 return 0; 540 } 541 542 static int f2fs_file_open(struct inode *inode, struct file *filp) 543 { 544 int err = fscrypt_file_open(inode, filp); 545 546 if (err) 547 return err; 548 549 if (!f2fs_is_compress_backend_ready(inode)) 550 return -EOPNOTSUPP; 551 552 err = fsverity_file_open(inode, filp); 553 if (err) 554 return err; 555 556 filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC; 557 filp->f_mode |= FMODE_CAN_ODIRECT; 558 559 return dquot_file_open(inode, filp); 560 } 561 562 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 563 { 564 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 565 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 566 __le32 *addr; 567 bool compressed_cluster = false; 568 int cluster_index = 0, valid_blocks = 0; 569 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 570 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 571 572 addr = get_dnode_addr(dn->inode, dn->node_page) + ofs; 573 574 /* Assumption: truncation starts with cluster */ 575 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 576 block_t blkaddr = le32_to_cpu(*addr); 577 578 if (f2fs_compressed_file(dn->inode) && 579 !(cluster_index & (cluster_size - 1))) { 580 if (compressed_cluster) 581 f2fs_i_compr_blocks_update(dn->inode, 582 valid_blocks, false); 583 compressed_cluster = (blkaddr == COMPRESS_ADDR); 584 valid_blocks = 0; 585 } 586 587 if (blkaddr == NULL_ADDR) 588 continue; 589 590 f2fs_set_data_blkaddr(dn, NULL_ADDR); 591 592 if (__is_valid_data_blkaddr(blkaddr)) { 593 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 594 DATA_GENERIC_ENHANCE)) 595 continue; 596 if (compressed_cluster) 597 valid_blocks++; 598 } 599 600 f2fs_invalidate_blocks(sbi, blkaddr); 601 602 if (!released || blkaddr != COMPRESS_ADDR) 603 nr_free++; 604 } 605 606 if (compressed_cluster) 607 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 608 609 if (nr_free) { 610 pgoff_t fofs; 611 /* 612 * once we invalidate valid blkaddr in range [ofs, ofs + count], 613 * we will invalidate all blkaddr in the whole range. 614 */ 615 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 616 dn->inode) + ofs; 617 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 618 f2fs_update_age_extent_cache_range(dn, fofs, len); 619 dec_valid_block_count(sbi, dn->inode, nr_free); 620 } 621 dn->ofs_in_node = ofs; 622 623 f2fs_update_time(sbi, REQ_TIME); 624 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 625 dn->ofs_in_node, nr_free); 626 } 627 628 static int truncate_partial_data_page(struct inode *inode, u64 from, 629 bool cache_only) 630 { 631 loff_t offset = from & (PAGE_SIZE - 1); 632 pgoff_t index = from >> PAGE_SHIFT; 633 struct address_space *mapping = inode->i_mapping; 634 struct page *page; 635 636 if (!offset && !cache_only) 637 return 0; 638 639 if (cache_only) { 640 page = find_lock_page(mapping, index); 641 if (page && PageUptodate(page)) 642 goto truncate_out; 643 f2fs_put_page(page, 1); 644 return 0; 645 } 646 647 page = f2fs_get_lock_data_page(inode, index, true); 648 if (IS_ERR(page)) 649 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 650 truncate_out: 651 f2fs_wait_on_page_writeback(page, DATA, true, true); 652 zero_user(page, offset, PAGE_SIZE - offset); 653 654 /* An encrypted inode should have a key and truncate the last page. */ 655 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 656 if (!cache_only) 657 set_page_dirty(page); 658 f2fs_put_page(page, 1); 659 return 0; 660 } 661 662 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 663 { 664 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 665 struct dnode_of_data dn; 666 pgoff_t free_from; 667 int count = 0, err = 0; 668 struct page *ipage; 669 bool truncate_page = false; 670 671 trace_f2fs_truncate_blocks_enter(inode, from); 672 673 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 674 675 if (free_from >= max_file_blocks(inode)) 676 goto free_partial; 677 678 if (lock) 679 f2fs_lock_op(sbi); 680 681 ipage = f2fs_get_node_page(sbi, inode->i_ino); 682 if (IS_ERR(ipage)) { 683 err = PTR_ERR(ipage); 684 goto out; 685 } 686 687 if (f2fs_has_inline_data(inode)) { 688 f2fs_truncate_inline_inode(inode, ipage, from); 689 f2fs_put_page(ipage, 1); 690 truncate_page = true; 691 goto out; 692 } 693 694 set_new_dnode(&dn, inode, ipage, NULL, 0); 695 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 696 if (err) { 697 if (err == -ENOENT) 698 goto free_next; 699 goto out; 700 } 701 702 count = ADDRS_PER_PAGE(dn.node_page, inode); 703 704 count -= dn.ofs_in_node; 705 f2fs_bug_on(sbi, count < 0); 706 707 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 708 f2fs_truncate_data_blocks_range(&dn, count); 709 free_from += count; 710 } 711 712 f2fs_put_dnode(&dn); 713 free_next: 714 err = f2fs_truncate_inode_blocks(inode, free_from); 715 out: 716 if (lock) 717 f2fs_unlock_op(sbi); 718 free_partial: 719 /* lastly zero out the first data page */ 720 if (!err) 721 err = truncate_partial_data_page(inode, from, truncate_page); 722 723 trace_f2fs_truncate_blocks_exit(inode, err); 724 return err; 725 } 726 727 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 728 { 729 u64 free_from = from; 730 int err; 731 732 #ifdef CONFIG_F2FS_FS_COMPRESSION 733 /* 734 * for compressed file, only support cluster size 735 * aligned truncation. 736 */ 737 if (f2fs_compressed_file(inode)) 738 free_from = round_up(from, 739 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 740 #endif 741 742 err = f2fs_do_truncate_blocks(inode, free_from, lock); 743 if (err) 744 return err; 745 746 #ifdef CONFIG_F2FS_FS_COMPRESSION 747 /* 748 * For compressed file, after release compress blocks, don't allow write 749 * direct, but we should allow write direct after truncate to zero. 750 */ 751 if (f2fs_compressed_file(inode) && !free_from 752 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 753 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 754 755 if (from != free_from) { 756 err = f2fs_truncate_partial_cluster(inode, from, lock); 757 if (err) 758 return err; 759 } 760 #endif 761 762 return 0; 763 } 764 765 int f2fs_truncate(struct inode *inode) 766 { 767 int err; 768 769 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 770 return -EIO; 771 772 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 773 S_ISLNK(inode->i_mode))) 774 return 0; 775 776 trace_f2fs_truncate(inode); 777 778 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 779 return -EIO; 780 781 err = f2fs_dquot_initialize(inode); 782 if (err) 783 return err; 784 785 /* we should check inline_data size */ 786 if (!f2fs_may_inline_data(inode)) { 787 err = f2fs_convert_inline_inode(inode); 788 if (err) 789 return err; 790 } 791 792 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 793 if (err) 794 return err; 795 796 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 797 f2fs_mark_inode_dirty_sync(inode, false); 798 return 0; 799 } 800 801 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 802 { 803 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 804 805 if (!fscrypt_dio_supported(inode)) 806 return true; 807 if (fsverity_active(inode)) 808 return true; 809 if (f2fs_compressed_file(inode)) 810 return true; 811 812 /* disallow direct IO if any of devices has unaligned blksize */ 813 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 814 return true; 815 /* 816 * for blkzoned device, fallback direct IO to buffered IO, so 817 * all IOs can be serialized by log-structured write. 818 */ 819 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE)) 820 return true; 821 if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi)) 822 return true; 823 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 824 return true; 825 826 return false; 827 } 828 829 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 830 struct kstat *stat, u32 request_mask, unsigned int query_flags) 831 { 832 struct inode *inode = d_inode(path->dentry); 833 struct f2fs_inode_info *fi = F2FS_I(inode); 834 struct f2fs_inode *ri = NULL; 835 unsigned int flags; 836 837 if (f2fs_has_extra_attr(inode) && 838 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 839 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 840 stat->result_mask |= STATX_BTIME; 841 stat->btime.tv_sec = fi->i_crtime.tv_sec; 842 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 843 } 844 845 /* 846 * Return the DIO alignment restrictions if requested. We only return 847 * this information when requested, since on encrypted files it might 848 * take a fair bit of work to get if the file wasn't opened recently. 849 * 850 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 851 * cannot represent that, so in that case we report no DIO support. 852 */ 853 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 854 unsigned int bsize = i_blocksize(inode); 855 856 stat->result_mask |= STATX_DIOALIGN; 857 if (!f2fs_force_buffered_io(inode, WRITE)) { 858 stat->dio_mem_align = bsize; 859 stat->dio_offset_align = bsize; 860 } 861 } 862 863 flags = fi->i_flags; 864 if (flags & F2FS_COMPR_FL) 865 stat->attributes |= STATX_ATTR_COMPRESSED; 866 if (flags & F2FS_APPEND_FL) 867 stat->attributes |= STATX_ATTR_APPEND; 868 if (IS_ENCRYPTED(inode)) 869 stat->attributes |= STATX_ATTR_ENCRYPTED; 870 if (flags & F2FS_IMMUTABLE_FL) 871 stat->attributes |= STATX_ATTR_IMMUTABLE; 872 if (flags & F2FS_NODUMP_FL) 873 stat->attributes |= STATX_ATTR_NODUMP; 874 if (IS_VERITY(inode)) 875 stat->attributes |= STATX_ATTR_VERITY; 876 877 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 878 STATX_ATTR_APPEND | 879 STATX_ATTR_ENCRYPTED | 880 STATX_ATTR_IMMUTABLE | 881 STATX_ATTR_NODUMP | 882 STATX_ATTR_VERITY); 883 884 generic_fillattr(idmap, request_mask, inode, stat); 885 886 /* we need to show initial sectors used for inline_data/dentries */ 887 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 888 f2fs_has_inline_dentry(inode)) 889 stat->blocks += (stat->size + 511) >> 9; 890 891 return 0; 892 } 893 894 #ifdef CONFIG_F2FS_FS_POSIX_ACL 895 static void __setattr_copy(struct mnt_idmap *idmap, 896 struct inode *inode, const struct iattr *attr) 897 { 898 unsigned int ia_valid = attr->ia_valid; 899 900 i_uid_update(idmap, attr, inode); 901 i_gid_update(idmap, attr, inode); 902 if (ia_valid & ATTR_ATIME) 903 inode_set_atime_to_ts(inode, attr->ia_atime); 904 if (ia_valid & ATTR_MTIME) 905 inode_set_mtime_to_ts(inode, attr->ia_mtime); 906 if (ia_valid & ATTR_CTIME) 907 inode_set_ctime_to_ts(inode, attr->ia_ctime); 908 if (ia_valid & ATTR_MODE) { 909 umode_t mode = attr->ia_mode; 910 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode); 911 912 if (!vfsgid_in_group_p(vfsgid) && 913 !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) 914 mode &= ~S_ISGID; 915 set_acl_inode(inode, mode); 916 } 917 } 918 #else 919 #define __setattr_copy setattr_copy 920 #endif 921 922 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 923 struct iattr *attr) 924 { 925 struct inode *inode = d_inode(dentry); 926 int err; 927 928 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 929 return -EIO; 930 931 if (unlikely(IS_IMMUTABLE(inode))) 932 return -EPERM; 933 934 if (unlikely(IS_APPEND(inode) && 935 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 936 ATTR_GID | ATTR_TIMES_SET)))) 937 return -EPERM; 938 939 if ((attr->ia_valid & ATTR_SIZE) && 940 !f2fs_is_compress_backend_ready(inode)) 941 return -EOPNOTSUPP; 942 943 err = setattr_prepare(idmap, dentry, attr); 944 if (err) 945 return err; 946 947 err = fscrypt_prepare_setattr(dentry, attr); 948 if (err) 949 return err; 950 951 err = fsverity_prepare_setattr(dentry, attr); 952 if (err) 953 return err; 954 955 if (is_quota_modification(idmap, inode, attr)) { 956 err = f2fs_dquot_initialize(inode); 957 if (err) 958 return err; 959 } 960 if (i_uid_needs_update(idmap, attr, inode) || 961 i_gid_needs_update(idmap, attr, inode)) { 962 f2fs_lock_op(F2FS_I_SB(inode)); 963 err = dquot_transfer(idmap, inode, attr); 964 if (err) { 965 set_sbi_flag(F2FS_I_SB(inode), 966 SBI_QUOTA_NEED_REPAIR); 967 f2fs_unlock_op(F2FS_I_SB(inode)); 968 return err; 969 } 970 /* 971 * update uid/gid under lock_op(), so that dquot and inode can 972 * be updated atomically. 973 */ 974 i_uid_update(idmap, attr, inode); 975 i_gid_update(idmap, attr, inode); 976 f2fs_mark_inode_dirty_sync(inode, true); 977 f2fs_unlock_op(F2FS_I_SB(inode)); 978 } 979 980 if (attr->ia_valid & ATTR_SIZE) { 981 loff_t old_size = i_size_read(inode); 982 983 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 984 /* 985 * should convert inline inode before i_size_write to 986 * keep smaller than inline_data size with inline flag. 987 */ 988 err = f2fs_convert_inline_inode(inode); 989 if (err) 990 return err; 991 } 992 993 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 994 filemap_invalidate_lock(inode->i_mapping); 995 996 truncate_setsize(inode, attr->ia_size); 997 998 if (attr->ia_size <= old_size) 999 err = f2fs_truncate(inode); 1000 /* 1001 * do not trim all blocks after i_size if target size is 1002 * larger than i_size. 1003 */ 1004 filemap_invalidate_unlock(inode->i_mapping); 1005 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1006 if (err) 1007 return err; 1008 1009 spin_lock(&F2FS_I(inode)->i_size_lock); 1010 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1011 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1012 spin_unlock(&F2FS_I(inode)->i_size_lock); 1013 } 1014 1015 __setattr_copy(idmap, inode, attr); 1016 1017 if (attr->ia_valid & ATTR_MODE) { 1018 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1019 1020 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1021 if (!err) 1022 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1023 clear_inode_flag(inode, FI_ACL_MODE); 1024 } 1025 } 1026 1027 /* file size may changed here */ 1028 f2fs_mark_inode_dirty_sync(inode, true); 1029 1030 /* inode change will produce dirty node pages flushed by checkpoint */ 1031 f2fs_balance_fs(F2FS_I_SB(inode), true); 1032 1033 return err; 1034 } 1035 1036 const struct inode_operations f2fs_file_inode_operations = { 1037 .getattr = f2fs_getattr, 1038 .setattr = f2fs_setattr, 1039 .get_inode_acl = f2fs_get_acl, 1040 .set_acl = f2fs_set_acl, 1041 .listxattr = f2fs_listxattr, 1042 .fiemap = f2fs_fiemap, 1043 .fileattr_get = f2fs_fileattr_get, 1044 .fileattr_set = f2fs_fileattr_set, 1045 }; 1046 1047 static int fill_zero(struct inode *inode, pgoff_t index, 1048 loff_t start, loff_t len) 1049 { 1050 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1051 struct page *page; 1052 1053 if (!len) 1054 return 0; 1055 1056 f2fs_balance_fs(sbi, true); 1057 1058 f2fs_lock_op(sbi); 1059 page = f2fs_get_new_data_page(inode, NULL, index, false); 1060 f2fs_unlock_op(sbi); 1061 1062 if (IS_ERR(page)) 1063 return PTR_ERR(page); 1064 1065 f2fs_wait_on_page_writeback(page, DATA, true, true); 1066 zero_user(page, start, len); 1067 set_page_dirty(page); 1068 f2fs_put_page(page, 1); 1069 return 0; 1070 } 1071 1072 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1073 { 1074 int err; 1075 1076 while (pg_start < pg_end) { 1077 struct dnode_of_data dn; 1078 pgoff_t end_offset, count; 1079 1080 set_new_dnode(&dn, inode, NULL, NULL, 0); 1081 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1082 if (err) { 1083 if (err == -ENOENT) { 1084 pg_start = f2fs_get_next_page_offset(&dn, 1085 pg_start); 1086 continue; 1087 } 1088 return err; 1089 } 1090 1091 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1092 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1093 1094 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1095 1096 f2fs_truncate_data_blocks_range(&dn, count); 1097 f2fs_put_dnode(&dn); 1098 1099 pg_start += count; 1100 } 1101 return 0; 1102 } 1103 1104 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1105 { 1106 pgoff_t pg_start, pg_end; 1107 loff_t off_start, off_end; 1108 int ret; 1109 1110 ret = f2fs_convert_inline_inode(inode); 1111 if (ret) 1112 return ret; 1113 1114 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1115 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1116 1117 off_start = offset & (PAGE_SIZE - 1); 1118 off_end = (offset + len) & (PAGE_SIZE - 1); 1119 1120 if (pg_start == pg_end) { 1121 ret = fill_zero(inode, pg_start, off_start, 1122 off_end - off_start); 1123 if (ret) 1124 return ret; 1125 } else { 1126 if (off_start) { 1127 ret = fill_zero(inode, pg_start++, off_start, 1128 PAGE_SIZE - off_start); 1129 if (ret) 1130 return ret; 1131 } 1132 if (off_end) { 1133 ret = fill_zero(inode, pg_end, 0, off_end); 1134 if (ret) 1135 return ret; 1136 } 1137 1138 if (pg_start < pg_end) { 1139 loff_t blk_start, blk_end; 1140 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1141 1142 f2fs_balance_fs(sbi, true); 1143 1144 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1145 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1146 1147 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1148 filemap_invalidate_lock(inode->i_mapping); 1149 1150 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1151 1152 f2fs_lock_op(sbi); 1153 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1154 f2fs_unlock_op(sbi); 1155 1156 filemap_invalidate_unlock(inode->i_mapping); 1157 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1158 } 1159 } 1160 1161 return ret; 1162 } 1163 1164 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1165 int *do_replace, pgoff_t off, pgoff_t len) 1166 { 1167 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1168 struct dnode_of_data dn; 1169 int ret, done, i; 1170 1171 next_dnode: 1172 set_new_dnode(&dn, inode, NULL, NULL, 0); 1173 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1174 if (ret && ret != -ENOENT) { 1175 return ret; 1176 } else if (ret == -ENOENT) { 1177 if (dn.max_level == 0) 1178 return -ENOENT; 1179 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1180 dn.ofs_in_node, len); 1181 blkaddr += done; 1182 do_replace += done; 1183 goto next; 1184 } 1185 1186 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1187 dn.ofs_in_node, len); 1188 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1189 *blkaddr = f2fs_data_blkaddr(&dn); 1190 1191 if (__is_valid_data_blkaddr(*blkaddr) && 1192 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1193 DATA_GENERIC_ENHANCE)) { 1194 f2fs_put_dnode(&dn); 1195 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1196 return -EFSCORRUPTED; 1197 } 1198 1199 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1200 1201 if (f2fs_lfs_mode(sbi)) { 1202 f2fs_put_dnode(&dn); 1203 return -EOPNOTSUPP; 1204 } 1205 1206 /* do not invalidate this block address */ 1207 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1208 *do_replace = 1; 1209 } 1210 } 1211 f2fs_put_dnode(&dn); 1212 next: 1213 len -= done; 1214 off += done; 1215 if (len) 1216 goto next_dnode; 1217 return 0; 1218 } 1219 1220 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1221 int *do_replace, pgoff_t off, int len) 1222 { 1223 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1224 struct dnode_of_data dn; 1225 int ret, i; 1226 1227 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1228 if (*do_replace == 0) 1229 continue; 1230 1231 set_new_dnode(&dn, inode, NULL, NULL, 0); 1232 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1233 if (ret) { 1234 dec_valid_block_count(sbi, inode, 1); 1235 f2fs_invalidate_blocks(sbi, *blkaddr); 1236 } else { 1237 f2fs_update_data_blkaddr(&dn, *blkaddr); 1238 } 1239 f2fs_put_dnode(&dn); 1240 } 1241 return 0; 1242 } 1243 1244 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1245 block_t *blkaddr, int *do_replace, 1246 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1247 { 1248 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1249 pgoff_t i = 0; 1250 int ret; 1251 1252 while (i < len) { 1253 if (blkaddr[i] == NULL_ADDR && !full) { 1254 i++; 1255 continue; 1256 } 1257 1258 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1259 struct dnode_of_data dn; 1260 struct node_info ni; 1261 size_t new_size; 1262 pgoff_t ilen; 1263 1264 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1265 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1266 if (ret) 1267 return ret; 1268 1269 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1270 if (ret) { 1271 f2fs_put_dnode(&dn); 1272 return ret; 1273 } 1274 1275 ilen = min((pgoff_t) 1276 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1277 dn.ofs_in_node, len - i); 1278 do { 1279 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1280 f2fs_truncate_data_blocks_range(&dn, 1); 1281 1282 if (do_replace[i]) { 1283 f2fs_i_blocks_write(src_inode, 1284 1, false, false); 1285 f2fs_i_blocks_write(dst_inode, 1286 1, true, false); 1287 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1288 blkaddr[i], ni.version, true, false); 1289 1290 do_replace[i] = 0; 1291 } 1292 dn.ofs_in_node++; 1293 i++; 1294 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1295 if (dst_inode->i_size < new_size) 1296 f2fs_i_size_write(dst_inode, new_size); 1297 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1298 1299 f2fs_put_dnode(&dn); 1300 } else { 1301 struct page *psrc, *pdst; 1302 1303 psrc = f2fs_get_lock_data_page(src_inode, 1304 src + i, true); 1305 if (IS_ERR(psrc)) 1306 return PTR_ERR(psrc); 1307 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1308 true); 1309 if (IS_ERR(pdst)) { 1310 f2fs_put_page(psrc, 1); 1311 return PTR_ERR(pdst); 1312 } 1313 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1314 set_page_dirty(pdst); 1315 set_page_private_gcing(pdst); 1316 f2fs_put_page(pdst, 1); 1317 f2fs_put_page(psrc, 1); 1318 1319 ret = f2fs_truncate_hole(src_inode, 1320 src + i, src + i + 1); 1321 if (ret) 1322 return ret; 1323 i++; 1324 } 1325 } 1326 return 0; 1327 } 1328 1329 static int __exchange_data_block(struct inode *src_inode, 1330 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1331 pgoff_t len, bool full) 1332 { 1333 block_t *src_blkaddr; 1334 int *do_replace; 1335 pgoff_t olen; 1336 int ret; 1337 1338 while (len) { 1339 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1340 1341 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1342 array_size(olen, sizeof(block_t)), 1343 GFP_NOFS); 1344 if (!src_blkaddr) 1345 return -ENOMEM; 1346 1347 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1348 array_size(olen, sizeof(int)), 1349 GFP_NOFS); 1350 if (!do_replace) { 1351 kvfree(src_blkaddr); 1352 return -ENOMEM; 1353 } 1354 1355 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1356 do_replace, src, olen); 1357 if (ret) 1358 goto roll_back; 1359 1360 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1361 do_replace, src, dst, olen, full); 1362 if (ret) 1363 goto roll_back; 1364 1365 src += olen; 1366 dst += olen; 1367 len -= olen; 1368 1369 kvfree(src_blkaddr); 1370 kvfree(do_replace); 1371 } 1372 return 0; 1373 1374 roll_back: 1375 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1376 kvfree(src_blkaddr); 1377 kvfree(do_replace); 1378 return ret; 1379 } 1380 1381 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1382 { 1383 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1384 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1385 pgoff_t start = offset >> PAGE_SHIFT; 1386 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1387 int ret; 1388 1389 f2fs_balance_fs(sbi, true); 1390 1391 /* avoid gc operation during block exchange */ 1392 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1393 filemap_invalidate_lock(inode->i_mapping); 1394 1395 f2fs_lock_op(sbi); 1396 f2fs_drop_extent_tree(inode); 1397 truncate_pagecache(inode, offset); 1398 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1399 f2fs_unlock_op(sbi); 1400 1401 filemap_invalidate_unlock(inode->i_mapping); 1402 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1403 return ret; 1404 } 1405 1406 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1407 { 1408 loff_t new_size; 1409 int ret; 1410 1411 if (offset + len >= i_size_read(inode)) 1412 return -EINVAL; 1413 1414 /* collapse range should be aligned to block size of f2fs. */ 1415 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1416 return -EINVAL; 1417 1418 ret = f2fs_convert_inline_inode(inode); 1419 if (ret) 1420 return ret; 1421 1422 /* write out all dirty pages from offset */ 1423 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1424 if (ret) 1425 return ret; 1426 1427 ret = f2fs_do_collapse(inode, offset, len); 1428 if (ret) 1429 return ret; 1430 1431 /* write out all moved pages, if possible */ 1432 filemap_invalidate_lock(inode->i_mapping); 1433 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1434 truncate_pagecache(inode, offset); 1435 1436 new_size = i_size_read(inode) - len; 1437 ret = f2fs_truncate_blocks(inode, new_size, true); 1438 filemap_invalidate_unlock(inode->i_mapping); 1439 if (!ret) 1440 f2fs_i_size_write(inode, new_size); 1441 return ret; 1442 } 1443 1444 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1445 pgoff_t end) 1446 { 1447 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1448 pgoff_t index = start; 1449 unsigned int ofs_in_node = dn->ofs_in_node; 1450 blkcnt_t count = 0; 1451 int ret; 1452 1453 for (; index < end; index++, dn->ofs_in_node++) { 1454 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1455 count++; 1456 } 1457 1458 dn->ofs_in_node = ofs_in_node; 1459 ret = f2fs_reserve_new_blocks(dn, count); 1460 if (ret) 1461 return ret; 1462 1463 dn->ofs_in_node = ofs_in_node; 1464 for (index = start; index < end; index++, dn->ofs_in_node++) { 1465 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1466 /* 1467 * f2fs_reserve_new_blocks will not guarantee entire block 1468 * allocation. 1469 */ 1470 if (dn->data_blkaddr == NULL_ADDR) { 1471 ret = -ENOSPC; 1472 break; 1473 } 1474 1475 if (dn->data_blkaddr == NEW_ADDR) 1476 continue; 1477 1478 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1479 DATA_GENERIC_ENHANCE)) { 1480 ret = -EFSCORRUPTED; 1481 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1482 break; 1483 } 1484 1485 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1486 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1487 } 1488 1489 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1490 f2fs_update_age_extent_cache_range(dn, start, index - start); 1491 1492 return ret; 1493 } 1494 1495 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1496 int mode) 1497 { 1498 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1499 struct address_space *mapping = inode->i_mapping; 1500 pgoff_t index, pg_start, pg_end; 1501 loff_t new_size = i_size_read(inode); 1502 loff_t off_start, off_end; 1503 int ret = 0; 1504 1505 ret = inode_newsize_ok(inode, (len + offset)); 1506 if (ret) 1507 return ret; 1508 1509 ret = f2fs_convert_inline_inode(inode); 1510 if (ret) 1511 return ret; 1512 1513 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1514 if (ret) 1515 return ret; 1516 1517 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1518 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1519 1520 off_start = offset & (PAGE_SIZE - 1); 1521 off_end = (offset + len) & (PAGE_SIZE - 1); 1522 1523 if (pg_start == pg_end) { 1524 ret = fill_zero(inode, pg_start, off_start, 1525 off_end - off_start); 1526 if (ret) 1527 return ret; 1528 1529 new_size = max_t(loff_t, new_size, offset + len); 1530 } else { 1531 if (off_start) { 1532 ret = fill_zero(inode, pg_start++, off_start, 1533 PAGE_SIZE - off_start); 1534 if (ret) 1535 return ret; 1536 1537 new_size = max_t(loff_t, new_size, 1538 (loff_t)pg_start << PAGE_SHIFT); 1539 } 1540 1541 for (index = pg_start; index < pg_end;) { 1542 struct dnode_of_data dn; 1543 unsigned int end_offset; 1544 pgoff_t end; 1545 1546 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1547 filemap_invalidate_lock(mapping); 1548 1549 truncate_pagecache_range(inode, 1550 (loff_t)index << PAGE_SHIFT, 1551 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1552 1553 f2fs_lock_op(sbi); 1554 1555 set_new_dnode(&dn, inode, NULL, NULL, 0); 1556 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1557 if (ret) { 1558 f2fs_unlock_op(sbi); 1559 filemap_invalidate_unlock(mapping); 1560 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1561 goto out; 1562 } 1563 1564 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1565 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1566 1567 ret = f2fs_do_zero_range(&dn, index, end); 1568 f2fs_put_dnode(&dn); 1569 1570 f2fs_unlock_op(sbi); 1571 filemap_invalidate_unlock(mapping); 1572 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1573 1574 f2fs_balance_fs(sbi, dn.node_changed); 1575 1576 if (ret) 1577 goto out; 1578 1579 index = end; 1580 new_size = max_t(loff_t, new_size, 1581 (loff_t)index << PAGE_SHIFT); 1582 } 1583 1584 if (off_end) { 1585 ret = fill_zero(inode, pg_end, 0, off_end); 1586 if (ret) 1587 goto out; 1588 1589 new_size = max_t(loff_t, new_size, offset + len); 1590 } 1591 } 1592 1593 out: 1594 if (new_size > i_size_read(inode)) { 1595 if (mode & FALLOC_FL_KEEP_SIZE) 1596 file_set_keep_isize(inode); 1597 else 1598 f2fs_i_size_write(inode, new_size); 1599 } 1600 return ret; 1601 } 1602 1603 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1604 { 1605 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1606 struct address_space *mapping = inode->i_mapping; 1607 pgoff_t nr, pg_start, pg_end, delta, idx; 1608 loff_t new_size; 1609 int ret = 0; 1610 1611 new_size = i_size_read(inode) + len; 1612 ret = inode_newsize_ok(inode, new_size); 1613 if (ret) 1614 return ret; 1615 1616 if (offset >= i_size_read(inode)) 1617 return -EINVAL; 1618 1619 /* insert range should be aligned to block size of f2fs. */ 1620 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1621 return -EINVAL; 1622 1623 ret = f2fs_convert_inline_inode(inode); 1624 if (ret) 1625 return ret; 1626 1627 f2fs_balance_fs(sbi, true); 1628 1629 filemap_invalidate_lock(mapping); 1630 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1631 filemap_invalidate_unlock(mapping); 1632 if (ret) 1633 return ret; 1634 1635 /* write out all dirty pages from offset */ 1636 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1637 if (ret) 1638 return ret; 1639 1640 pg_start = offset >> PAGE_SHIFT; 1641 pg_end = (offset + len) >> PAGE_SHIFT; 1642 delta = pg_end - pg_start; 1643 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1644 1645 /* avoid gc operation during block exchange */ 1646 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1647 filemap_invalidate_lock(mapping); 1648 truncate_pagecache(inode, offset); 1649 1650 while (!ret && idx > pg_start) { 1651 nr = idx - pg_start; 1652 if (nr > delta) 1653 nr = delta; 1654 idx -= nr; 1655 1656 f2fs_lock_op(sbi); 1657 f2fs_drop_extent_tree(inode); 1658 1659 ret = __exchange_data_block(inode, inode, idx, 1660 idx + delta, nr, false); 1661 f2fs_unlock_op(sbi); 1662 } 1663 filemap_invalidate_unlock(mapping); 1664 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1665 1666 /* write out all moved pages, if possible */ 1667 filemap_invalidate_lock(mapping); 1668 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1669 truncate_pagecache(inode, offset); 1670 filemap_invalidate_unlock(mapping); 1671 1672 if (!ret) 1673 f2fs_i_size_write(inode, new_size); 1674 return ret; 1675 } 1676 1677 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1678 loff_t len, int mode) 1679 { 1680 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1681 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1682 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1683 .m_may_create = true }; 1684 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1685 .init_gc_type = FG_GC, 1686 .should_migrate_blocks = false, 1687 .err_gc_skipped = true, 1688 .nr_free_secs = 0 }; 1689 pgoff_t pg_start, pg_end; 1690 loff_t new_size; 1691 loff_t off_end; 1692 block_t expanded = 0; 1693 int err; 1694 1695 err = inode_newsize_ok(inode, (len + offset)); 1696 if (err) 1697 return err; 1698 1699 err = f2fs_convert_inline_inode(inode); 1700 if (err) 1701 return err; 1702 1703 f2fs_balance_fs(sbi, true); 1704 1705 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1706 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1707 off_end = (offset + len) & (PAGE_SIZE - 1); 1708 1709 map.m_lblk = pg_start; 1710 map.m_len = pg_end - pg_start; 1711 if (off_end) 1712 map.m_len++; 1713 1714 if (!map.m_len) 1715 return 0; 1716 1717 if (f2fs_is_pinned_file(inode)) { 1718 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1719 block_t sec_len = roundup(map.m_len, sec_blks); 1720 1721 map.m_len = sec_blks; 1722 next_alloc: 1723 if (has_not_enough_free_secs(sbi, 0, 1724 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1725 f2fs_down_write(&sbi->gc_lock); 1726 stat_inc_gc_call_count(sbi, FOREGROUND); 1727 err = f2fs_gc(sbi, &gc_control); 1728 if (err && err != -ENODATA) 1729 goto out_err; 1730 } 1731 1732 f2fs_down_write(&sbi->pin_sem); 1733 1734 f2fs_lock_op(sbi); 1735 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false); 1736 f2fs_unlock_op(sbi); 1737 1738 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1739 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1740 file_dont_truncate(inode); 1741 1742 f2fs_up_write(&sbi->pin_sem); 1743 1744 expanded += map.m_len; 1745 sec_len -= map.m_len; 1746 map.m_lblk += map.m_len; 1747 if (!err && sec_len) 1748 goto next_alloc; 1749 1750 map.m_len = expanded; 1751 } else { 1752 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1753 expanded = map.m_len; 1754 } 1755 out_err: 1756 if (err) { 1757 pgoff_t last_off; 1758 1759 if (!expanded) 1760 return err; 1761 1762 last_off = pg_start + expanded - 1; 1763 1764 /* update new size to the failed position */ 1765 new_size = (last_off == pg_end) ? offset + len : 1766 (loff_t)(last_off + 1) << PAGE_SHIFT; 1767 } else { 1768 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1769 } 1770 1771 if (new_size > i_size_read(inode)) { 1772 if (mode & FALLOC_FL_KEEP_SIZE) 1773 file_set_keep_isize(inode); 1774 else 1775 f2fs_i_size_write(inode, new_size); 1776 } 1777 1778 return err; 1779 } 1780 1781 static long f2fs_fallocate(struct file *file, int mode, 1782 loff_t offset, loff_t len) 1783 { 1784 struct inode *inode = file_inode(file); 1785 long ret = 0; 1786 1787 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1788 return -EIO; 1789 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1790 return -ENOSPC; 1791 if (!f2fs_is_compress_backend_ready(inode)) 1792 return -EOPNOTSUPP; 1793 1794 /* f2fs only support ->fallocate for regular file */ 1795 if (!S_ISREG(inode->i_mode)) 1796 return -EINVAL; 1797 1798 if (IS_ENCRYPTED(inode) && 1799 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1800 return -EOPNOTSUPP; 1801 1802 /* 1803 * Pinned file should not support partial truncation since the block 1804 * can be used by applications. 1805 */ 1806 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1807 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1808 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1809 return -EOPNOTSUPP; 1810 1811 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1812 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1813 FALLOC_FL_INSERT_RANGE)) 1814 return -EOPNOTSUPP; 1815 1816 inode_lock(inode); 1817 1818 ret = file_modified(file); 1819 if (ret) 1820 goto out; 1821 1822 if (mode & FALLOC_FL_PUNCH_HOLE) { 1823 if (offset >= inode->i_size) 1824 goto out; 1825 1826 ret = f2fs_punch_hole(inode, offset, len); 1827 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1828 ret = f2fs_collapse_range(inode, offset, len); 1829 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1830 ret = f2fs_zero_range(inode, offset, len, mode); 1831 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1832 ret = f2fs_insert_range(inode, offset, len); 1833 } else { 1834 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1835 } 1836 1837 if (!ret) { 1838 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1839 f2fs_mark_inode_dirty_sync(inode, false); 1840 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1841 } 1842 1843 out: 1844 inode_unlock(inode); 1845 1846 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1847 return ret; 1848 } 1849 1850 static int f2fs_release_file(struct inode *inode, struct file *filp) 1851 { 1852 /* 1853 * f2fs_release_file is called at every close calls. So we should 1854 * not drop any inmemory pages by close called by other process. 1855 */ 1856 if (!(filp->f_mode & FMODE_WRITE) || 1857 atomic_read(&inode->i_writecount) != 1) 1858 return 0; 1859 1860 inode_lock(inode); 1861 f2fs_abort_atomic_write(inode, true); 1862 inode_unlock(inode); 1863 1864 return 0; 1865 } 1866 1867 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1868 { 1869 struct inode *inode = file_inode(file); 1870 1871 /* 1872 * If the process doing a transaction is crashed, we should do 1873 * roll-back. Otherwise, other reader/write can see corrupted database 1874 * until all the writers close its file. Since this should be done 1875 * before dropping file lock, it needs to do in ->flush. 1876 */ 1877 if (F2FS_I(inode)->atomic_write_task == current && 1878 (current->flags & PF_EXITING)) { 1879 inode_lock(inode); 1880 f2fs_abort_atomic_write(inode, true); 1881 inode_unlock(inode); 1882 } 1883 1884 return 0; 1885 } 1886 1887 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1888 { 1889 struct f2fs_inode_info *fi = F2FS_I(inode); 1890 u32 masked_flags = fi->i_flags & mask; 1891 1892 /* mask can be shrunk by flags_valid selector */ 1893 iflags &= mask; 1894 1895 /* Is it quota file? Do not allow user to mess with it */ 1896 if (IS_NOQUOTA(inode)) 1897 return -EPERM; 1898 1899 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1900 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1901 return -EOPNOTSUPP; 1902 if (!f2fs_empty_dir(inode)) 1903 return -ENOTEMPTY; 1904 } 1905 1906 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1907 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1908 return -EOPNOTSUPP; 1909 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1910 return -EINVAL; 1911 } 1912 1913 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1914 if (masked_flags & F2FS_COMPR_FL) { 1915 if (!f2fs_disable_compressed_file(inode)) 1916 return -EINVAL; 1917 } else { 1918 /* try to convert inline_data to support compression */ 1919 int err = f2fs_convert_inline_inode(inode); 1920 if (err) 1921 return err; 1922 1923 f2fs_down_write(&F2FS_I(inode)->i_sem); 1924 if (!f2fs_may_compress(inode) || 1925 (S_ISREG(inode->i_mode) && 1926 F2FS_HAS_BLOCKS(inode))) { 1927 f2fs_up_write(&F2FS_I(inode)->i_sem); 1928 return -EINVAL; 1929 } 1930 err = set_compress_context(inode); 1931 f2fs_up_write(&F2FS_I(inode)->i_sem); 1932 1933 if (err) 1934 return err; 1935 } 1936 } 1937 1938 fi->i_flags = iflags | (fi->i_flags & ~mask); 1939 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1940 (fi->i_flags & F2FS_NOCOMP_FL)); 1941 1942 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1943 set_inode_flag(inode, FI_PROJ_INHERIT); 1944 else 1945 clear_inode_flag(inode, FI_PROJ_INHERIT); 1946 1947 inode_set_ctime_current(inode); 1948 f2fs_set_inode_flags(inode); 1949 f2fs_mark_inode_dirty_sync(inode, true); 1950 return 0; 1951 } 1952 1953 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1954 1955 /* 1956 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1957 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1958 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1959 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1960 * 1961 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1962 * FS_IOC_FSSETXATTR is done by the VFS. 1963 */ 1964 1965 static const struct { 1966 u32 iflag; 1967 u32 fsflag; 1968 } f2fs_fsflags_map[] = { 1969 { F2FS_COMPR_FL, FS_COMPR_FL }, 1970 { F2FS_SYNC_FL, FS_SYNC_FL }, 1971 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1972 { F2FS_APPEND_FL, FS_APPEND_FL }, 1973 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1974 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1975 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1976 { F2FS_INDEX_FL, FS_INDEX_FL }, 1977 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1978 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1979 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1980 }; 1981 1982 #define F2FS_GETTABLE_FS_FL ( \ 1983 FS_COMPR_FL | \ 1984 FS_SYNC_FL | \ 1985 FS_IMMUTABLE_FL | \ 1986 FS_APPEND_FL | \ 1987 FS_NODUMP_FL | \ 1988 FS_NOATIME_FL | \ 1989 FS_NOCOMP_FL | \ 1990 FS_INDEX_FL | \ 1991 FS_DIRSYNC_FL | \ 1992 FS_PROJINHERIT_FL | \ 1993 FS_ENCRYPT_FL | \ 1994 FS_INLINE_DATA_FL | \ 1995 FS_NOCOW_FL | \ 1996 FS_VERITY_FL | \ 1997 FS_CASEFOLD_FL) 1998 1999 #define F2FS_SETTABLE_FS_FL ( \ 2000 FS_COMPR_FL | \ 2001 FS_SYNC_FL | \ 2002 FS_IMMUTABLE_FL | \ 2003 FS_APPEND_FL | \ 2004 FS_NODUMP_FL | \ 2005 FS_NOATIME_FL | \ 2006 FS_NOCOMP_FL | \ 2007 FS_DIRSYNC_FL | \ 2008 FS_PROJINHERIT_FL | \ 2009 FS_CASEFOLD_FL) 2010 2011 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2012 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2013 { 2014 u32 fsflags = 0; 2015 int i; 2016 2017 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2018 if (iflags & f2fs_fsflags_map[i].iflag) 2019 fsflags |= f2fs_fsflags_map[i].fsflag; 2020 2021 return fsflags; 2022 } 2023 2024 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2025 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2026 { 2027 u32 iflags = 0; 2028 int i; 2029 2030 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2031 if (fsflags & f2fs_fsflags_map[i].fsflag) 2032 iflags |= f2fs_fsflags_map[i].iflag; 2033 2034 return iflags; 2035 } 2036 2037 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2038 { 2039 struct inode *inode = file_inode(filp); 2040 2041 return put_user(inode->i_generation, (int __user *)arg); 2042 } 2043 2044 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2045 { 2046 struct inode *inode = file_inode(filp); 2047 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2048 struct f2fs_inode_info *fi = F2FS_I(inode); 2049 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2050 struct inode *pinode; 2051 loff_t isize; 2052 int ret; 2053 2054 if (!inode_owner_or_capable(idmap, inode)) 2055 return -EACCES; 2056 2057 if (!S_ISREG(inode->i_mode)) 2058 return -EINVAL; 2059 2060 if (filp->f_flags & O_DIRECT) 2061 return -EINVAL; 2062 2063 ret = mnt_want_write_file(filp); 2064 if (ret) 2065 return ret; 2066 2067 inode_lock(inode); 2068 2069 if (!f2fs_disable_compressed_file(inode)) { 2070 ret = -EINVAL; 2071 goto out; 2072 } 2073 2074 if (f2fs_is_atomic_file(inode)) 2075 goto out; 2076 2077 ret = f2fs_convert_inline_inode(inode); 2078 if (ret) 2079 goto out; 2080 2081 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2082 2083 /* 2084 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2085 * f2fs_is_atomic_file. 2086 */ 2087 if (get_dirty_pages(inode)) 2088 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2089 inode->i_ino, get_dirty_pages(inode)); 2090 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2091 if (ret) { 2092 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2093 goto out; 2094 } 2095 2096 /* Check if the inode already has a COW inode */ 2097 if (fi->cow_inode == NULL) { 2098 /* Create a COW inode for atomic write */ 2099 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2100 if (IS_ERR(pinode)) { 2101 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2102 ret = PTR_ERR(pinode); 2103 goto out; 2104 } 2105 2106 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2107 iput(pinode); 2108 if (ret) { 2109 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2110 goto out; 2111 } 2112 2113 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2114 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2115 } else { 2116 /* Reuse the already created COW inode */ 2117 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2118 if (ret) { 2119 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2120 goto out; 2121 } 2122 } 2123 2124 f2fs_write_inode(inode, NULL); 2125 2126 stat_inc_atomic_inode(inode); 2127 2128 set_inode_flag(inode, FI_ATOMIC_FILE); 2129 2130 isize = i_size_read(inode); 2131 fi->original_i_size = isize; 2132 if (truncate) { 2133 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2134 truncate_inode_pages_final(inode->i_mapping); 2135 f2fs_i_size_write(inode, 0); 2136 isize = 0; 2137 } 2138 f2fs_i_size_write(fi->cow_inode, isize); 2139 2140 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2141 2142 f2fs_update_time(sbi, REQ_TIME); 2143 fi->atomic_write_task = current; 2144 stat_update_max_atomic_write(inode); 2145 fi->atomic_write_cnt = 0; 2146 out: 2147 inode_unlock(inode); 2148 mnt_drop_write_file(filp); 2149 return ret; 2150 } 2151 2152 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2153 { 2154 struct inode *inode = file_inode(filp); 2155 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2156 int ret; 2157 2158 if (!inode_owner_or_capable(idmap, inode)) 2159 return -EACCES; 2160 2161 ret = mnt_want_write_file(filp); 2162 if (ret) 2163 return ret; 2164 2165 f2fs_balance_fs(F2FS_I_SB(inode), true); 2166 2167 inode_lock(inode); 2168 2169 if (f2fs_is_atomic_file(inode)) { 2170 ret = f2fs_commit_atomic_write(inode); 2171 if (!ret) 2172 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2173 2174 f2fs_abort_atomic_write(inode, ret); 2175 } else { 2176 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2177 } 2178 2179 inode_unlock(inode); 2180 mnt_drop_write_file(filp); 2181 return ret; 2182 } 2183 2184 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2185 { 2186 struct inode *inode = file_inode(filp); 2187 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2188 int ret; 2189 2190 if (!inode_owner_or_capable(idmap, inode)) 2191 return -EACCES; 2192 2193 ret = mnt_want_write_file(filp); 2194 if (ret) 2195 return ret; 2196 2197 inode_lock(inode); 2198 2199 f2fs_abort_atomic_write(inode, true); 2200 2201 inode_unlock(inode); 2202 2203 mnt_drop_write_file(filp); 2204 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2205 return ret; 2206 } 2207 2208 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2209 { 2210 struct inode *inode = file_inode(filp); 2211 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2212 struct super_block *sb = sbi->sb; 2213 __u32 in; 2214 int ret = 0; 2215 2216 if (!capable(CAP_SYS_ADMIN)) 2217 return -EPERM; 2218 2219 if (get_user(in, (__u32 __user *)arg)) 2220 return -EFAULT; 2221 2222 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2223 ret = mnt_want_write_file(filp); 2224 if (ret) { 2225 if (ret == -EROFS) { 2226 ret = 0; 2227 f2fs_stop_checkpoint(sbi, false, 2228 STOP_CP_REASON_SHUTDOWN); 2229 trace_f2fs_shutdown(sbi, in, ret); 2230 } 2231 return ret; 2232 } 2233 } 2234 2235 switch (in) { 2236 case F2FS_GOING_DOWN_FULLSYNC: 2237 ret = bdev_freeze(sb->s_bdev); 2238 if (ret) 2239 goto out; 2240 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2241 bdev_thaw(sb->s_bdev); 2242 break; 2243 case F2FS_GOING_DOWN_METASYNC: 2244 /* do checkpoint only */ 2245 ret = f2fs_sync_fs(sb, 1); 2246 if (ret) 2247 goto out; 2248 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2249 break; 2250 case F2FS_GOING_DOWN_NOSYNC: 2251 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2252 break; 2253 case F2FS_GOING_DOWN_METAFLUSH: 2254 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2255 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2256 break; 2257 case F2FS_GOING_DOWN_NEED_FSCK: 2258 set_sbi_flag(sbi, SBI_NEED_FSCK); 2259 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2260 set_sbi_flag(sbi, SBI_IS_DIRTY); 2261 /* do checkpoint only */ 2262 ret = f2fs_sync_fs(sb, 1); 2263 goto out; 2264 default: 2265 ret = -EINVAL; 2266 goto out; 2267 } 2268 2269 f2fs_stop_gc_thread(sbi); 2270 f2fs_stop_discard_thread(sbi); 2271 2272 f2fs_drop_discard_cmd(sbi); 2273 clear_opt(sbi, DISCARD); 2274 2275 f2fs_update_time(sbi, REQ_TIME); 2276 out: 2277 if (in != F2FS_GOING_DOWN_FULLSYNC) 2278 mnt_drop_write_file(filp); 2279 2280 trace_f2fs_shutdown(sbi, in, ret); 2281 2282 return ret; 2283 } 2284 2285 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2286 { 2287 struct inode *inode = file_inode(filp); 2288 struct super_block *sb = inode->i_sb; 2289 struct fstrim_range range; 2290 int ret; 2291 2292 if (!capable(CAP_SYS_ADMIN)) 2293 return -EPERM; 2294 2295 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2296 return -EOPNOTSUPP; 2297 2298 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2299 sizeof(range))) 2300 return -EFAULT; 2301 2302 ret = mnt_want_write_file(filp); 2303 if (ret) 2304 return ret; 2305 2306 range.minlen = max((unsigned int)range.minlen, 2307 bdev_discard_granularity(sb->s_bdev)); 2308 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2309 mnt_drop_write_file(filp); 2310 if (ret < 0) 2311 return ret; 2312 2313 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2314 sizeof(range))) 2315 return -EFAULT; 2316 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2317 return 0; 2318 } 2319 2320 static bool uuid_is_nonzero(__u8 u[16]) 2321 { 2322 int i; 2323 2324 for (i = 0; i < 16; i++) 2325 if (u[i]) 2326 return true; 2327 return false; 2328 } 2329 2330 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2331 { 2332 struct inode *inode = file_inode(filp); 2333 2334 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2335 return -EOPNOTSUPP; 2336 2337 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2338 2339 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2340 } 2341 2342 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2343 { 2344 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2345 return -EOPNOTSUPP; 2346 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2347 } 2348 2349 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2350 { 2351 struct inode *inode = file_inode(filp); 2352 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2353 u8 encrypt_pw_salt[16]; 2354 int err; 2355 2356 if (!f2fs_sb_has_encrypt(sbi)) 2357 return -EOPNOTSUPP; 2358 2359 err = mnt_want_write_file(filp); 2360 if (err) 2361 return err; 2362 2363 f2fs_down_write(&sbi->sb_lock); 2364 2365 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2366 goto got_it; 2367 2368 /* update superblock with uuid */ 2369 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2370 2371 err = f2fs_commit_super(sbi, false); 2372 if (err) { 2373 /* undo new data */ 2374 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2375 goto out_err; 2376 } 2377 got_it: 2378 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2379 out_err: 2380 f2fs_up_write(&sbi->sb_lock); 2381 mnt_drop_write_file(filp); 2382 2383 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2384 err = -EFAULT; 2385 2386 return err; 2387 } 2388 2389 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2390 unsigned long arg) 2391 { 2392 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2393 return -EOPNOTSUPP; 2394 2395 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2396 } 2397 2398 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2399 { 2400 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2401 return -EOPNOTSUPP; 2402 2403 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2404 } 2405 2406 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2407 { 2408 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2409 return -EOPNOTSUPP; 2410 2411 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2412 } 2413 2414 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2415 unsigned long arg) 2416 { 2417 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2418 return -EOPNOTSUPP; 2419 2420 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2421 } 2422 2423 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2424 unsigned long arg) 2425 { 2426 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2427 return -EOPNOTSUPP; 2428 2429 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2430 } 2431 2432 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2433 { 2434 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2435 return -EOPNOTSUPP; 2436 2437 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2438 } 2439 2440 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2441 { 2442 struct inode *inode = file_inode(filp); 2443 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2444 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2445 .no_bg_gc = false, 2446 .should_migrate_blocks = false, 2447 .nr_free_secs = 0 }; 2448 __u32 sync; 2449 int ret; 2450 2451 if (!capable(CAP_SYS_ADMIN)) 2452 return -EPERM; 2453 2454 if (get_user(sync, (__u32 __user *)arg)) 2455 return -EFAULT; 2456 2457 if (f2fs_readonly(sbi->sb)) 2458 return -EROFS; 2459 2460 ret = mnt_want_write_file(filp); 2461 if (ret) 2462 return ret; 2463 2464 if (!sync) { 2465 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2466 ret = -EBUSY; 2467 goto out; 2468 } 2469 } else { 2470 f2fs_down_write(&sbi->gc_lock); 2471 } 2472 2473 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2474 gc_control.err_gc_skipped = sync; 2475 stat_inc_gc_call_count(sbi, FOREGROUND); 2476 ret = f2fs_gc(sbi, &gc_control); 2477 out: 2478 mnt_drop_write_file(filp); 2479 return ret; 2480 } 2481 2482 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2483 { 2484 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2485 struct f2fs_gc_control gc_control = { 2486 .init_gc_type = range->sync ? FG_GC : BG_GC, 2487 .no_bg_gc = false, 2488 .should_migrate_blocks = false, 2489 .err_gc_skipped = range->sync, 2490 .nr_free_secs = 0 }; 2491 u64 end; 2492 int ret; 2493 2494 if (!capable(CAP_SYS_ADMIN)) 2495 return -EPERM; 2496 if (f2fs_readonly(sbi->sb)) 2497 return -EROFS; 2498 2499 end = range->start + range->len; 2500 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2501 end >= MAX_BLKADDR(sbi)) 2502 return -EINVAL; 2503 2504 ret = mnt_want_write_file(filp); 2505 if (ret) 2506 return ret; 2507 2508 do_more: 2509 if (!range->sync) { 2510 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2511 ret = -EBUSY; 2512 goto out; 2513 } 2514 } else { 2515 f2fs_down_write(&sbi->gc_lock); 2516 } 2517 2518 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2519 stat_inc_gc_call_count(sbi, FOREGROUND); 2520 ret = f2fs_gc(sbi, &gc_control); 2521 if (ret) { 2522 if (ret == -EBUSY) 2523 ret = -EAGAIN; 2524 goto out; 2525 } 2526 range->start += CAP_BLKS_PER_SEC(sbi); 2527 if (range->start <= end) 2528 goto do_more; 2529 out: 2530 mnt_drop_write_file(filp); 2531 return ret; 2532 } 2533 2534 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2535 { 2536 struct f2fs_gc_range range; 2537 2538 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2539 sizeof(range))) 2540 return -EFAULT; 2541 return __f2fs_ioc_gc_range(filp, &range); 2542 } 2543 2544 static int f2fs_ioc_write_checkpoint(struct file *filp) 2545 { 2546 struct inode *inode = file_inode(filp); 2547 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2548 int ret; 2549 2550 if (!capable(CAP_SYS_ADMIN)) 2551 return -EPERM; 2552 2553 if (f2fs_readonly(sbi->sb)) 2554 return -EROFS; 2555 2556 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2557 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2558 return -EINVAL; 2559 } 2560 2561 ret = mnt_want_write_file(filp); 2562 if (ret) 2563 return ret; 2564 2565 ret = f2fs_sync_fs(sbi->sb, 1); 2566 2567 mnt_drop_write_file(filp); 2568 return ret; 2569 } 2570 2571 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2572 struct file *filp, 2573 struct f2fs_defragment *range) 2574 { 2575 struct inode *inode = file_inode(filp); 2576 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2577 .m_seg_type = NO_CHECK_TYPE, 2578 .m_may_create = false }; 2579 struct extent_info ei = {}; 2580 pgoff_t pg_start, pg_end, next_pgofs; 2581 unsigned int blk_per_seg = sbi->blocks_per_seg; 2582 unsigned int total = 0, sec_num; 2583 block_t blk_end = 0; 2584 bool fragmented = false; 2585 int err; 2586 2587 pg_start = range->start >> PAGE_SHIFT; 2588 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2589 2590 f2fs_balance_fs(sbi, true); 2591 2592 inode_lock(inode); 2593 2594 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2595 err = -EINVAL; 2596 goto unlock_out; 2597 } 2598 2599 /* if in-place-update policy is enabled, don't waste time here */ 2600 set_inode_flag(inode, FI_OPU_WRITE); 2601 if (f2fs_should_update_inplace(inode, NULL)) { 2602 err = -EINVAL; 2603 goto out; 2604 } 2605 2606 /* writeback all dirty pages in the range */ 2607 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2608 range->start + range->len - 1); 2609 if (err) 2610 goto out; 2611 2612 /* 2613 * lookup mapping info in extent cache, skip defragmenting if physical 2614 * block addresses are continuous. 2615 */ 2616 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2617 if (ei.fofs + ei.len >= pg_end) 2618 goto out; 2619 } 2620 2621 map.m_lblk = pg_start; 2622 map.m_next_pgofs = &next_pgofs; 2623 2624 /* 2625 * lookup mapping info in dnode page cache, skip defragmenting if all 2626 * physical block addresses are continuous even if there are hole(s) 2627 * in logical blocks. 2628 */ 2629 while (map.m_lblk < pg_end) { 2630 map.m_len = pg_end - map.m_lblk; 2631 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2632 if (err) 2633 goto out; 2634 2635 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2636 map.m_lblk = next_pgofs; 2637 continue; 2638 } 2639 2640 if (blk_end && blk_end != map.m_pblk) 2641 fragmented = true; 2642 2643 /* record total count of block that we're going to move */ 2644 total += map.m_len; 2645 2646 blk_end = map.m_pblk + map.m_len; 2647 2648 map.m_lblk += map.m_len; 2649 } 2650 2651 if (!fragmented) { 2652 total = 0; 2653 goto out; 2654 } 2655 2656 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2657 2658 /* 2659 * make sure there are enough free section for LFS allocation, this can 2660 * avoid defragment running in SSR mode when free section are allocated 2661 * intensively 2662 */ 2663 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2664 err = -EAGAIN; 2665 goto out; 2666 } 2667 2668 map.m_lblk = pg_start; 2669 map.m_len = pg_end - pg_start; 2670 total = 0; 2671 2672 while (map.m_lblk < pg_end) { 2673 pgoff_t idx; 2674 int cnt = 0; 2675 2676 do_map: 2677 map.m_len = pg_end - map.m_lblk; 2678 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2679 if (err) 2680 goto clear_out; 2681 2682 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2683 map.m_lblk = next_pgofs; 2684 goto check; 2685 } 2686 2687 set_inode_flag(inode, FI_SKIP_WRITES); 2688 2689 idx = map.m_lblk; 2690 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2691 struct page *page; 2692 2693 page = f2fs_get_lock_data_page(inode, idx, true); 2694 if (IS_ERR(page)) { 2695 err = PTR_ERR(page); 2696 goto clear_out; 2697 } 2698 2699 set_page_dirty(page); 2700 set_page_private_gcing(page); 2701 f2fs_put_page(page, 1); 2702 2703 idx++; 2704 cnt++; 2705 total++; 2706 } 2707 2708 map.m_lblk = idx; 2709 check: 2710 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2711 goto do_map; 2712 2713 clear_inode_flag(inode, FI_SKIP_WRITES); 2714 2715 err = filemap_fdatawrite(inode->i_mapping); 2716 if (err) 2717 goto out; 2718 } 2719 clear_out: 2720 clear_inode_flag(inode, FI_SKIP_WRITES); 2721 out: 2722 clear_inode_flag(inode, FI_OPU_WRITE); 2723 unlock_out: 2724 inode_unlock(inode); 2725 if (!err) 2726 range->len = (u64)total << PAGE_SHIFT; 2727 return err; 2728 } 2729 2730 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2731 { 2732 struct inode *inode = file_inode(filp); 2733 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2734 struct f2fs_defragment range; 2735 int err; 2736 2737 if (!capable(CAP_SYS_ADMIN)) 2738 return -EPERM; 2739 2740 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2741 return -EINVAL; 2742 2743 if (f2fs_readonly(sbi->sb)) 2744 return -EROFS; 2745 2746 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2747 sizeof(range))) 2748 return -EFAULT; 2749 2750 /* verify alignment of offset & size */ 2751 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2752 return -EINVAL; 2753 2754 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2755 max_file_blocks(inode))) 2756 return -EINVAL; 2757 2758 err = mnt_want_write_file(filp); 2759 if (err) 2760 return err; 2761 2762 err = f2fs_defragment_range(sbi, filp, &range); 2763 mnt_drop_write_file(filp); 2764 2765 f2fs_update_time(sbi, REQ_TIME); 2766 if (err < 0) 2767 return err; 2768 2769 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2770 sizeof(range))) 2771 return -EFAULT; 2772 2773 return 0; 2774 } 2775 2776 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2777 struct file *file_out, loff_t pos_out, size_t len) 2778 { 2779 struct inode *src = file_inode(file_in); 2780 struct inode *dst = file_inode(file_out); 2781 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2782 size_t olen = len, dst_max_i_size = 0; 2783 size_t dst_osize; 2784 int ret; 2785 2786 if (file_in->f_path.mnt != file_out->f_path.mnt || 2787 src->i_sb != dst->i_sb) 2788 return -EXDEV; 2789 2790 if (unlikely(f2fs_readonly(src->i_sb))) 2791 return -EROFS; 2792 2793 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2794 return -EINVAL; 2795 2796 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2797 return -EOPNOTSUPP; 2798 2799 if (pos_out < 0 || pos_in < 0) 2800 return -EINVAL; 2801 2802 if (src == dst) { 2803 if (pos_in == pos_out) 2804 return 0; 2805 if (pos_out > pos_in && pos_out < pos_in + len) 2806 return -EINVAL; 2807 } 2808 2809 inode_lock(src); 2810 if (src != dst) { 2811 ret = -EBUSY; 2812 if (!inode_trylock(dst)) 2813 goto out; 2814 } 2815 2816 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) { 2817 ret = -EOPNOTSUPP; 2818 goto out_unlock; 2819 } 2820 2821 ret = -EINVAL; 2822 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2823 goto out_unlock; 2824 if (len == 0) 2825 olen = len = src->i_size - pos_in; 2826 if (pos_in + len == src->i_size) 2827 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2828 if (len == 0) { 2829 ret = 0; 2830 goto out_unlock; 2831 } 2832 2833 dst_osize = dst->i_size; 2834 if (pos_out + olen > dst->i_size) 2835 dst_max_i_size = pos_out + olen; 2836 2837 /* verify the end result is block aligned */ 2838 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2839 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2840 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2841 goto out_unlock; 2842 2843 ret = f2fs_convert_inline_inode(src); 2844 if (ret) 2845 goto out_unlock; 2846 2847 ret = f2fs_convert_inline_inode(dst); 2848 if (ret) 2849 goto out_unlock; 2850 2851 /* write out all dirty pages from offset */ 2852 ret = filemap_write_and_wait_range(src->i_mapping, 2853 pos_in, pos_in + len); 2854 if (ret) 2855 goto out_unlock; 2856 2857 ret = filemap_write_and_wait_range(dst->i_mapping, 2858 pos_out, pos_out + len); 2859 if (ret) 2860 goto out_unlock; 2861 2862 f2fs_balance_fs(sbi, true); 2863 2864 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2865 if (src != dst) { 2866 ret = -EBUSY; 2867 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2868 goto out_src; 2869 } 2870 2871 f2fs_lock_op(sbi); 2872 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2873 pos_out >> F2FS_BLKSIZE_BITS, 2874 len >> F2FS_BLKSIZE_BITS, false); 2875 2876 if (!ret) { 2877 if (dst_max_i_size) 2878 f2fs_i_size_write(dst, dst_max_i_size); 2879 else if (dst_osize != dst->i_size) 2880 f2fs_i_size_write(dst, dst_osize); 2881 } 2882 f2fs_unlock_op(sbi); 2883 2884 if (src != dst) 2885 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2886 out_src: 2887 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2888 if (ret) 2889 goto out_unlock; 2890 2891 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 2892 f2fs_mark_inode_dirty_sync(src, false); 2893 if (src != dst) { 2894 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 2895 f2fs_mark_inode_dirty_sync(dst, false); 2896 } 2897 f2fs_update_time(sbi, REQ_TIME); 2898 2899 out_unlock: 2900 if (src != dst) 2901 inode_unlock(dst); 2902 out: 2903 inode_unlock(src); 2904 return ret; 2905 } 2906 2907 static int __f2fs_ioc_move_range(struct file *filp, 2908 struct f2fs_move_range *range) 2909 { 2910 struct fd dst; 2911 int err; 2912 2913 if (!(filp->f_mode & FMODE_READ) || 2914 !(filp->f_mode & FMODE_WRITE)) 2915 return -EBADF; 2916 2917 dst = fdget(range->dst_fd); 2918 if (!dst.file) 2919 return -EBADF; 2920 2921 if (!(dst.file->f_mode & FMODE_WRITE)) { 2922 err = -EBADF; 2923 goto err_out; 2924 } 2925 2926 err = mnt_want_write_file(filp); 2927 if (err) 2928 goto err_out; 2929 2930 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2931 range->pos_out, range->len); 2932 2933 mnt_drop_write_file(filp); 2934 err_out: 2935 fdput(dst); 2936 return err; 2937 } 2938 2939 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2940 { 2941 struct f2fs_move_range range; 2942 2943 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2944 sizeof(range))) 2945 return -EFAULT; 2946 return __f2fs_ioc_move_range(filp, &range); 2947 } 2948 2949 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2950 { 2951 struct inode *inode = file_inode(filp); 2952 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2953 struct sit_info *sm = SIT_I(sbi); 2954 unsigned int start_segno = 0, end_segno = 0; 2955 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2956 struct f2fs_flush_device range; 2957 struct f2fs_gc_control gc_control = { 2958 .init_gc_type = FG_GC, 2959 .should_migrate_blocks = true, 2960 .err_gc_skipped = true, 2961 .nr_free_secs = 0 }; 2962 int ret; 2963 2964 if (!capable(CAP_SYS_ADMIN)) 2965 return -EPERM; 2966 2967 if (f2fs_readonly(sbi->sb)) 2968 return -EROFS; 2969 2970 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2971 return -EINVAL; 2972 2973 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2974 sizeof(range))) 2975 return -EFAULT; 2976 2977 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2978 __is_large_section(sbi)) { 2979 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2980 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2981 return -EINVAL; 2982 } 2983 2984 ret = mnt_want_write_file(filp); 2985 if (ret) 2986 return ret; 2987 2988 if (range.dev_num != 0) 2989 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2990 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2991 2992 start_segno = sm->last_victim[FLUSH_DEVICE]; 2993 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2994 start_segno = dev_start_segno; 2995 end_segno = min(start_segno + range.segments, dev_end_segno); 2996 2997 while (start_segno < end_segno) { 2998 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2999 ret = -EBUSY; 3000 goto out; 3001 } 3002 sm->last_victim[GC_CB] = end_segno + 1; 3003 sm->last_victim[GC_GREEDY] = end_segno + 1; 3004 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3005 3006 gc_control.victim_segno = start_segno; 3007 stat_inc_gc_call_count(sbi, FOREGROUND); 3008 ret = f2fs_gc(sbi, &gc_control); 3009 if (ret == -EAGAIN) 3010 ret = 0; 3011 else if (ret < 0) 3012 break; 3013 start_segno++; 3014 } 3015 out: 3016 mnt_drop_write_file(filp); 3017 return ret; 3018 } 3019 3020 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3021 { 3022 struct inode *inode = file_inode(filp); 3023 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3024 3025 /* Must validate to set it with SQLite behavior in Android. */ 3026 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3027 3028 return put_user(sb_feature, (u32 __user *)arg); 3029 } 3030 3031 #ifdef CONFIG_QUOTA 3032 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3033 { 3034 struct dquot *transfer_to[MAXQUOTAS] = {}; 3035 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3036 struct super_block *sb = sbi->sb; 3037 int err; 3038 3039 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3040 if (IS_ERR(transfer_to[PRJQUOTA])) 3041 return PTR_ERR(transfer_to[PRJQUOTA]); 3042 3043 err = __dquot_transfer(inode, transfer_to); 3044 if (err) 3045 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3046 dqput(transfer_to[PRJQUOTA]); 3047 return err; 3048 } 3049 3050 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3051 { 3052 struct f2fs_inode_info *fi = F2FS_I(inode); 3053 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3054 struct f2fs_inode *ri = NULL; 3055 kprojid_t kprojid; 3056 int err; 3057 3058 if (!f2fs_sb_has_project_quota(sbi)) { 3059 if (projid != F2FS_DEF_PROJID) 3060 return -EOPNOTSUPP; 3061 else 3062 return 0; 3063 } 3064 3065 if (!f2fs_has_extra_attr(inode)) 3066 return -EOPNOTSUPP; 3067 3068 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3069 3070 if (projid_eq(kprojid, fi->i_projid)) 3071 return 0; 3072 3073 err = -EPERM; 3074 /* Is it quota file? Do not allow user to mess with it */ 3075 if (IS_NOQUOTA(inode)) 3076 return err; 3077 3078 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3079 return -EOVERFLOW; 3080 3081 err = f2fs_dquot_initialize(inode); 3082 if (err) 3083 return err; 3084 3085 f2fs_lock_op(sbi); 3086 err = f2fs_transfer_project_quota(inode, kprojid); 3087 if (err) 3088 goto out_unlock; 3089 3090 fi->i_projid = kprojid; 3091 inode_set_ctime_current(inode); 3092 f2fs_mark_inode_dirty_sync(inode, true); 3093 out_unlock: 3094 f2fs_unlock_op(sbi); 3095 return err; 3096 } 3097 #else 3098 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3099 { 3100 return 0; 3101 } 3102 3103 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3104 { 3105 if (projid != F2FS_DEF_PROJID) 3106 return -EOPNOTSUPP; 3107 return 0; 3108 } 3109 #endif 3110 3111 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3112 { 3113 struct inode *inode = d_inode(dentry); 3114 struct f2fs_inode_info *fi = F2FS_I(inode); 3115 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3116 3117 if (IS_ENCRYPTED(inode)) 3118 fsflags |= FS_ENCRYPT_FL; 3119 if (IS_VERITY(inode)) 3120 fsflags |= FS_VERITY_FL; 3121 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3122 fsflags |= FS_INLINE_DATA_FL; 3123 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3124 fsflags |= FS_NOCOW_FL; 3125 3126 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3127 3128 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3129 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3130 3131 return 0; 3132 } 3133 3134 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3135 struct dentry *dentry, struct fileattr *fa) 3136 { 3137 struct inode *inode = d_inode(dentry); 3138 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3139 u32 iflags; 3140 int err; 3141 3142 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3143 return -EIO; 3144 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3145 return -ENOSPC; 3146 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3147 return -EOPNOTSUPP; 3148 fsflags &= F2FS_SETTABLE_FS_FL; 3149 if (!fa->flags_valid) 3150 mask &= FS_COMMON_FL; 3151 3152 iflags = f2fs_fsflags_to_iflags(fsflags); 3153 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3154 return -EOPNOTSUPP; 3155 3156 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3157 if (!err) 3158 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3159 3160 return err; 3161 } 3162 3163 int f2fs_pin_file_control(struct inode *inode, bool inc) 3164 { 3165 struct f2fs_inode_info *fi = F2FS_I(inode); 3166 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3167 3168 /* Use i_gc_failures for normal file as a risk signal. */ 3169 if (inc) 3170 f2fs_i_gc_failures_write(inode, 3171 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3172 3173 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3174 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3175 __func__, inode->i_ino, 3176 fi->i_gc_failures[GC_FAILURE_PIN]); 3177 clear_inode_flag(inode, FI_PIN_FILE); 3178 return -EAGAIN; 3179 } 3180 return 0; 3181 } 3182 3183 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3184 { 3185 struct inode *inode = file_inode(filp); 3186 __u32 pin; 3187 int ret = 0; 3188 3189 if (get_user(pin, (__u32 __user *)arg)) 3190 return -EFAULT; 3191 3192 if (!S_ISREG(inode->i_mode)) 3193 return -EINVAL; 3194 3195 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3196 return -EROFS; 3197 3198 ret = mnt_want_write_file(filp); 3199 if (ret) 3200 return ret; 3201 3202 inode_lock(inode); 3203 3204 if (!pin) { 3205 clear_inode_flag(inode, FI_PIN_FILE); 3206 f2fs_i_gc_failures_write(inode, 0); 3207 goto done; 3208 } 3209 3210 if (f2fs_should_update_outplace(inode, NULL)) { 3211 ret = -EINVAL; 3212 goto out; 3213 } 3214 3215 if (f2fs_pin_file_control(inode, false)) { 3216 ret = -EAGAIN; 3217 goto out; 3218 } 3219 3220 ret = f2fs_convert_inline_inode(inode); 3221 if (ret) 3222 goto out; 3223 3224 if (!f2fs_disable_compressed_file(inode)) { 3225 ret = -EOPNOTSUPP; 3226 goto out; 3227 } 3228 3229 set_inode_flag(inode, FI_PIN_FILE); 3230 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3231 done: 3232 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3233 out: 3234 inode_unlock(inode); 3235 mnt_drop_write_file(filp); 3236 return ret; 3237 } 3238 3239 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3240 { 3241 struct inode *inode = file_inode(filp); 3242 __u32 pin = 0; 3243 3244 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3245 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3246 return put_user(pin, (u32 __user *)arg); 3247 } 3248 3249 int f2fs_precache_extents(struct inode *inode) 3250 { 3251 struct f2fs_inode_info *fi = F2FS_I(inode); 3252 struct f2fs_map_blocks map; 3253 pgoff_t m_next_extent; 3254 loff_t end; 3255 int err; 3256 3257 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3258 return -EOPNOTSUPP; 3259 3260 map.m_lblk = 0; 3261 map.m_pblk = 0; 3262 map.m_next_pgofs = NULL; 3263 map.m_next_extent = &m_next_extent; 3264 map.m_seg_type = NO_CHECK_TYPE; 3265 map.m_may_create = false; 3266 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3267 3268 while (map.m_lblk < end) { 3269 map.m_len = end - map.m_lblk; 3270 3271 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3272 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3273 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3274 if (err || !map.m_len) 3275 return err; 3276 3277 map.m_lblk = m_next_extent; 3278 } 3279 3280 return 0; 3281 } 3282 3283 static int f2fs_ioc_precache_extents(struct file *filp) 3284 { 3285 return f2fs_precache_extents(file_inode(filp)); 3286 } 3287 3288 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3289 { 3290 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3291 __u64 block_count; 3292 3293 if (!capable(CAP_SYS_ADMIN)) 3294 return -EPERM; 3295 3296 if (f2fs_readonly(sbi->sb)) 3297 return -EROFS; 3298 3299 if (copy_from_user(&block_count, (void __user *)arg, 3300 sizeof(block_count))) 3301 return -EFAULT; 3302 3303 return f2fs_resize_fs(filp, block_count); 3304 } 3305 3306 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3307 { 3308 struct inode *inode = file_inode(filp); 3309 3310 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3311 3312 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3313 f2fs_warn(F2FS_I_SB(inode), 3314 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3315 inode->i_ino); 3316 return -EOPNOTSUPP; 3317 } 3318 3319 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3320 } 3321 3322 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3323 { 3324 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3325 return -EOPNOTSUPP; 3326 3327 return fsverity_ioctl_measure(filp, (void __user *)arg); 3328 } 3329 3330 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3331 { 3332 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3333 return -EOPNOTSUPP; 3334 3335 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3336 } 3337 3338 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3339 { 3340 struct inode *inode = file_inode(filp); 3341 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3342 char *vbuf; 3343 int count; 3344 int err = 0; 3345 3346 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3347 if (!vbuf) 3348 return -ENOMEM; 3349 3350 f2fs_down_read(&sbi->sb_lock); 3351 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3352 ARRAY_SIZE(sbi->raw_super->volume_name), 3353 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3354 f2fs_up_read(&sbi->sb_lock); 3355 3356 if (copy_to_user((char __user *)arg, vbuf, 3357 min(FSLABEL_MAX, count))) 3358 err = -EFAULT; 3359 3360 kfree(vbuf); 3361 return err; 3362 } 3363 3364 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3365 { 3366 struct inode *inode = file_inode(filp); 3367 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3368 char *vbuf; 3369 int err = 0; 3370 3371 if (!capable(CAP_SYS_ADMIN)) 3372 return -EPERM; 3373 3374 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3375 if (IS_ERR(vbuf)) 3376 return PTR_ERR(vbuf); 3377 3378 err = mnt_want_write_file(filp); 3379 if (err) 3380 goto out; 3381 3382 f2fs_down_write(&sbi->sb_lock); 3383 3384 memset(sbi->raw_super->volume_name, 0, 3385 sizeof(sbi->raw_super->volume_name)); 3386 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3387 sbi->raw_super->volume_name, 3388 ARRAY_SIZE(sbi->raw_super->volume_name)); 3389 3390 err = f2fs_commit_super(sbi, false); 3391 3392 f2fs_up_write(&sbi->sb_lock); 3393 3394 mnt_drop_write_file(filp); 3395 out: 3396 kfree(vbuf); 3397 return err; 3398 } 3399 3400 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3401 { 3402 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3403 return -EOPNOTSUPP; 3404 3405 if (!f2fs_compressed_file(inode)) 3406 return -EINVAL; 3407 3408 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3409 3410 return 0; 3411 } 3412 3413 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3414 { 3415 struct inode *inode = file_inode(filp); 3416 __u64 blocks; 3417 int ret; 3418 3419 ret = f2fs_get_compress_blocks(inode, &blocks); 3420 if (ret < 0) 3421 return ret; 3422 3423 return put_user(blocks, (u64 __user *)arg); 3424 } 3425 3426 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3427 { 3428 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3429 unsigned int released_blocks = 0; 3430 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3431 block_t blkaddr; 3432 int i; 3433 3434 for (i = 0; i < count; i++) { 3435 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3436 dn->ofs_in_node + i); 3437 3438 if (!__is_valid_data_blkaddr(blkaddr)) 3439 continue; 3440 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3441 DATA_GENERIC_ENHANCE))) { 3442 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3443 return -EFSCORRUPTED; 3444 } 3445 } 3446 3447 while (count) { 3448 int compr_blocks = 0; 3449 3450 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3451 blkaddr = f2fs_data_blkaddr(dn); 3452 3453 if (i == 0) { 3454 if (blkaddr == COMPRESS_ADDR) 3455 continue; 3456 dn->ofs_in_node += cluster_size; 3457 goto next; 3458 } 3459 3460 if (__is_valid_data_blkaddr(blkaddr)) 3461 compr_blocks++; 3462 3463 if (blkaddr != NEW_ADDR) 3464 continue; 3465 3466 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3467 } 3468 3469 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3470 dec_valid_block_count(sbi, dn->inode, 3471 cluster_size - compr_blocks); 3472 3473 released_blocks += cluster_size - compr_blocks; 3474 next: 3475 count -= cluster_size; 3476 } 3477 3478 return released_blocks; 3479 } 3480 3481 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3482 { 3483 struct inode *inode = file_inode(filp); 3484 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3485 pgoff_t page_idx = 0, last_idx; 3486 unsigned int released_blocks = 0; 3487 int ret; 3488 int writecount; 3489 3490 if (!f2fs_sb_has_compression(sbi)) 3491 return -EOPNOTSUPP; 3492 3493 if (!f2fs_compressed_file(inode)) 3494 return -EINVAL; 3495 3496 if (f2fs_readonly(sbi->sb)) 3497 return -EROFS; 3498 3499 ret = mnt_want_write_file(filp); 3500 if (ret) 3501 return ret; 3502 3503 f2fs_balance_fs(sbi, true); 3504 3505 inode_lock(inode); 3506 3507 writecount = atomic_read(&inode->i_writecount); 3508 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3509 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3510 ret = -EBUSY; 3511 goto out; 3512 } 3513 3514 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3515 ret = -EINVAL; 3516 goto out; 3517 } 3518 3519 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3520 if (ret) 3521 goto out; 3522 3523 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3524 ret = -EPERM; 3525 goto out; 3526 } 3527 3528 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3529 inode_set_ctime_current(inode); 3530 f2fs_mark_inode_dirty_sync(inode, true); 3531 3532 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3533 filemap_invalidate_lock(inode->i_mapping); 3534 3535 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3536 3537 while (page_idx < last_idx) { 3538 struct dnode_of_data dn; 3539 pgoff_t end_offset, count; 3540 3541 set_new_dnode(&dn, inode, NULL, NULL, 0); 3542 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3543 if (ret) { 3544 if (ret == -ENOENT) { 3545 page_idx = f2fs_get_next_page_offset(&dn, 3546 page_idx); 3547 ret = 0; 3548 continue; 3549 } 3550 break; 3551 } 3552 3553 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3554 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3555 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3556 3557 ret = release_compress_blocks(&dn, count); 3558 3559 f2fs_put_dnode(&dn); 3560 3561 if (ret < 0) 3562 break; 3563 3564 page_idx += count; 3565 released_blocks += ret; 3566 } 3567 3568 filemap_invalidate_unlock(inode->i_mapping); 3569 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3570 out: 3571 inode_unlock(inode); 3572 3573 mnt_drop_write_file(filp); 3574 3575 if (ret >= 0) { 3576 ret = put_user(released_blocks, (u64 __user *)arg); 3577 } else if (released_blocks && 3578 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3579 set_sbi_flag(sbi, SBI_NEED_FSCK); 3580 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3581 "iblocks=%llu, released=%u, compr_blocks=%u, " 3582 "run fsck to fix.", 3583 __func__, inode->i_ino, inode->i_blocks, 3584 released_blocks, 3585 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3586 } 3587 3588 return ret; 3589 } 3590 3591 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3592 { 3593 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3594 unsigned int reserved_blocks = 0; 3595 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3596 block_t blkaddr; 3597 int i; 3598 3599 for (i = 0; i < count; i++) { 3600 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3601 dn->ofs_in_node + i); 3602 3603 if (!__is_valid_data_blkaddr(blkaddr)) 3604 continue; 3605 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3606 DATA_GENERIC_ENHANCE))) { 3607 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3608 return -EFSCORRUPTED; 3609 } 3610 } 3611 3612 while (count) { 3613 int compr_blocks = 0; 3614 blkcnt_t reserved; 3615 int ret; 3616 3617 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3618 blkaddr = f2fs_data_blkaddr(dn); 3619 3620 if (i == 0) { 3621 if (blkaddr == COMPRESS_ADDR) 3622 continue; 3623 dn->ofs_in_node += cluster_size; 3624 goto next; 3625 } 3626 3627 if (__is_valid_data_blkaddr(blkaddr)) { 3628 compr_blocks++; 3629 continue; 3630 } 3631 3632 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3633 } 3634 3635 reserved = cluster_size - compr_blocks; 3636 ret = inc_valid_block_count(sbi, dn->inode, &reserved); 3637 if (ret) 3638 return ret; 3639 3640 if (reserved != cluster_size - compr_blocks) 3641 return -ENOSPC; 3642 3643 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3644 3645 reserved_blocks += reserved; 3646 next: 3647 count -= cluster_size; 3648 } 3649 3650 return reserved_blocks; 3651 } 3652 3653 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3654 { 3655 struct inode *inode = file_inode(filp); 3656 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3657 pgoff_t page_idx = 0, last_idx; 3658 unsigned int reserved_blocks = 0; 3659 int ret; 3660 3661 if (!f2fs_sb_has_compression(sbi)) 3662 return -EOPNOTSUPP; 3663 3664 if (!f2fs_compressed_file(inode)) 3665 return -EINVAL; 3666 3667 if (f2fs_readonly(sbi->sb)) 3668 return -EROFS; 3669 3670 ret = mnt_want_write_file(filp); 3671 if (ret) 3672 return ret; 3673 3674 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3675 goto out; 3676 3677 f2fs_balance_fs(sbi, true); 3678 3679 inode_lock(inode); 3680 3681 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3682 ret = -EINVAL; 3683 goto unlock_inode; 3684 } 3685 3686 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3687 filemap_invalidate_lock(inode->i_mapping); 3688 3689 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3690 3691 while (page_idx < last_idx) { 3692 struct dnode_of_data dn; 3693 pgoff_t end_offset, count; 3694 3695 set_new_dnode(&dn, inode, NULL, NULL, 0); 3696 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3697 if (ret) { 3698 if (ret == -ENOENT) { 3699 page_idx = f2fs_get_next_page_offset(&dn, 3700 page_idx); 3701 ret = 0; 3702 continue; 3703 } 3704 break; 3705 } 3706 3707 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3708 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3709 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3710 3711 ret = reserve_compress_blocks(&dn, count); 3712 3713 f2fs_put_dnode(&dn); 3714 3715 if (ret < 0) 3716 break; 3717 3718 page_idx += count; 3719 reserved_blocks += ret; 3720 } 3721 3722 filemap_invalidate_unlock(inode->i_mapping); 3723 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3724 3725 if (ret >= 0) { 3726 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3727 inode_set_ctime_current(inode); 3728 f2fs_mark_inode_dirty_sync(inode, true); 3729 } 3730 unlock_inode: 3731 inode_unlock(inode); 3732 out: 3733 mnt_drop_write_file(filp); 3734 3735 if (ret >= 0) { 3736 ret = put_user(reserved_blocks, (u64 __user *)arg); 3737 } else if (reserved_blocks && 3738 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3739 set_sbi_flag(sbi, SBI_NEED_FSCK); 3740 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3741 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3742 "run fsck to fix.", 3743 __func__, inode->i_ino, inode->i_blocks, 3744 reserved_blocks, 3745 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3746 } 3747 3748 return ret; 3749 } 3750 3751 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3752 pgoff_t off, block_t block, block_t len, u32 flags) 3753 { 3754 sector_t sector = SECTOR_FROM_BLOCK(block); 3755 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3756 int ret = 0; 3757 3758 if (flags & F2FS_TRIM_FILE_DISCARD) { 3759 if (bdev_max_secure_erase_sectors(bdev)) 3760 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3761 GFP_NOFS); 3762 else 3763 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3764 GFP_NOFS); 3765 } 3766 3767 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3768 if (IS_ENCRYPTED(inode)) 3769 ret = fscrypt_zeroout_range(inode, off, block, len); 3770 else 3771 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3772 GFP_NOFS, 0); 3773 } 3774 3775 return ret; 3776 } 3777 3778 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3779 { 3780 struct inode *inode = file_inode(filp); 3781 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3782 struct address_space *mapping = inode->i_mapping; 3783 struct block_device *prev_bdev = NULL; 3784 struct f2fs_sectrim_range range; 3785 pgoff_t index, pg_end, prev_index = 0; 3786 block_t prev_block = 0, len = 0; 3787 loff_t end_addr; 3788 bool to_end = false; 3789 int ret = 0; 3790 3791 if (!(filp->f_mode & FMODE_WRITE)) 3792 return -EBADF; 3793 3794 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3795 sizeof(range))) 3796 return -EFAULT; 3797 3798 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3799 !S_ISREG(inode->i_mode)) 3800 return -EINVAL; 3801 3802 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3803 !f2fs_hw_support_discard(sbi)) || 3804 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3805 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3806 return -EOPNOTSUPP; 3807 3808 file_start_write(filp); 3809 inode_lock(inode); 3810 3811 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3812 range.start >= inode->i_size) { 3813 ret = -EINVAL; 3814 goto err; 3815 } 3816 3817 if (range.len == 0) 3818 goto err; 3819 3820 if (inode->i_size - range.start > range.len) { 3821 end_addr = range.start + range.len; 3822 } else { 3823 end_addr = range.len == (u64)-1 ? 3824 sbi->sb->s_maxbytes : inode->i_size; 3825 to_end = true; 3826 } 3827 3828 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3829 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3830 ret = -EINVAL; 3831 goto err; 3832 } 3833 3834 index = F2FS_BYTES_TO_BLK(range.start); 3835 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3836 3837 ret = f2fs_convert_inline_inode(inode); 3838 if (ret) 3839 goto err; 3840 3841 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3842 filemap_invalidate_lock(mapping); 3843 3844 ret = filemap_write_and_wait_range(mapping, range.start, 3845 to_end ? LLONG_MAX : end_addr - 1); 3846 if (ret) 3847 goto out; 3848 3849 truncate_inode_pages_range(mapping, range.start, 3850 to_end ? -1 : end_addr - 1); 3851 3852 while (index < pg_end) { 3853 struct dnode_of_data dn; 3854 pgoff_t end_offset, count; 3855 int i; 3856 3857 set_new_dnode(&dn, inode, NULL, NULL, 0); 3858 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3859 if (ret) { 3860 if (ret == -ENOENT) { 3861 index = f2fs_get_next_page_offset(&dn, index); 3862 continue; 3863 } 3864 goto out; 3865 } 3866 3867 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3868 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3869 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3870 struct block_device *cur_bdev; 3871 block_t blkaddr = f2fs_data_blkaddr(&dn); 3872 3873 if (!__is_valid_data_blkaddr(blkaddr)) 3874 continue; 3875 3876 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3877 DATA_GENERIC_ENHANCE)) { 3878 ret = -EFSCORRUPTED; 3879 f2fs_put_dnode(&dn); 3880 f2fs_handle_error(sbi, 3881 ERROR_INVALID_BLKADDR); 3882 goto out; 3883 } 3884 3885 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3886 if (f2fs_is_multi_device(sbi)) { 3887 int di = f2fs_target_device_index(sbi, blkaddr); 3888 3889 blkaddr -= FDEV(di).start_blk; 3890 } 3891 3892 if (len) { 3893 if (prev_bdev == cur_bdev && 3894 index == prev_index + len && 3895 blkaddr == prev_block + len) { 3896 len++; 3897 } else { 3898 ret = f2fs_secure_erase(prev_bdev, 3899 inode, prev_index, prev_block, 3900 len, range.flags); 3901 if (ret) { 3902 f2fs_put_dnode(&dn); 3903 goto out; 3904 } 3905 3906 len = 0; 3907 } 3908 } 3909 3910 if (!len) { 3911 prev_bdev = cur_bdev; 3912 prev_index = index; 3913 prev_block = blkaddr; 3914 len = 1; 3915 } 3916 } 3917 3918 f2fs_put_dnode(&dn); 3919 3920 if (fatal_signal_pending(current)) { 3921 ret = -EINTR; 3922 goto out; 3923 } 3924 cond_resched(); 3925 } 3926 3927 if (len) 3928 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3929 prev_block, len, range.flags); 3930 out: 3931 filemap_invalidate_unlock(mapping); 3932 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3933 err: 3934 inode_unlock(inode); 3935 file_end_write(filp); 3936 3937 return ret; 3938 } 3939 3940 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3941 { 3942 struct inode *inode = file_inode(filp); 3943 struct f2fs_comp_option option; 3944 3945 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3946 return -EOPNOTSUPP; 3947 3948 inode_lock_shared(inode); 3949 3950 if (!f2fs_compressed_file(inode)) { 3951 inode_unlock_shared(inode); 3952 return -ENODATA; 3953 } 3954 3955 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3956 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3957 3958 inode_unlock_shared(inode); 3959 3960 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3961 sizeof(option))) 3962 return -EFAULT; 3963 3964 return 0; 3965 } 3966 3967 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3968 { 3969 struct inode *inode = file_inode(filp); 3970 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3971 struct f2fs_comp_option option; 3972 int ret = 0; 3973 3974 if (!f2fs_sb_has_compression(sbi)) 3975 return -EOPNOTSUPP; 3976 3977 if (!(filp->f_mode & FMODE_WRITE)) 3978 return -EBADF; 3979 3980 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3981 sizeof(option))) 3982 return -EFAULT; 3983 3984 if (!f2fs_compressed_file(inode) || 3985 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3986 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3987 option.algorithm >= COMPRESS_MAX) 3988 return -EINVAL; 3989 3990 file_start_write(filp); 3991 inode_lock(inode); 3992 3993 f2fs_down_write(&F2FS_I(inode)->i_sem); 3994 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 3995 ret = -EBUSY; 3996 goto out; 3997 } 3998 3999 if (F2FS_HAS_BLOCKS(inode)) { 4000 ret = -EFBIG; 4001 goto out; 4002 } 4003 4004 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4005 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4006 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4007 /* Set default level */ 4008 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4009 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4010 else 4011 F2FS_I(inode)->i_compress_level = 0; 4012 /* Adjust mount option level */ 4013 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4014 F2FS_OPTION(sbi).compress_level) 4015 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4016 f2fs_mark_inode_dirty_sync(inode, true); 4017 4018 if (!f2fs_is_compress_backend_ready(inode)) 4019 f2fs_warn(sbi, "compression algorithm is successfully set, " 4020 "but current kernel doesn't support this algorithm."); 4021 out: 4022 f2fs_up_write(&F2FS_I(inode)->i_sem); 4023 inode_unlock(inode); 4024 file_end_write(filp); 4025 4026 return ret; 4027 } 4028 4029 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4030 { 4031 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4032 struct address_space *mapping = inode->i_mapping; 4033 struct page *page; 4034 pgoff_t redirty_idx = page_idx; 4035 int i, page_len = 0, ret = 0; 4036 4037 page_cache_ra_unbounded(&ractl, len, 0); 4038 4039 for (i = 0; i < len; i++, page_idx++) { 4040 page = read_cache_page(mapping, page_idx, NULL, NULL); 4041 if (IS_ERR(page)) { 4042 ret = PTR_ERR(page); 4043 break; 4044 } 4045 page_len++; 4046 } 4047 4048 for (i = 0; i < page_len; i++, redirty_idx++) { 4049 page = find_lock_page(mapping, redirty_idx); 4050 4051 /* It will never fail, when page has pinned above */ 4052 f2fs_bug_on(F2FS_I_SB(inode), !page); 4053 4054 set_page_dirty(page); 4055 set_page_private_gcing(page); 4056 f2fs_put_page(page, 1); 4057 f2fs_put_page(page, 0); 4058 } 4059 4060 return ret; 4061 } 4062 4063 static int f2fs_ioc_decompress_file(struct file *filp) 4064 { 4065 struct inode *inode = file_inode(filp); 4066 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4067 struct f2fs_inode_info *fi = F2FS_I(inode); 4068 pgoff_t page_idx = 0, last_idx; 4069 unsigned int blk_per_seg = sbi->blocks_per_seg; 4070 int cluster_size = fi->i_cluster_size; 4071 int count, ret; 4072 4073 if (!f2fs_sb_has_compression(sbi) || 4074 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4075 return -EOPNOTSUPP; 4076 4077 if (!(filp->f_mode & FMODE_WRITE)) 4078 return -EBADF; 4079 4080 if (!f2fs_compressed_file(inode)) 4081 return -EINVAL; 4082 4083 f2fs_balance_fs(sbi, true); 4084 4085 file_start_write(filp); 4086 inode_lock(inode); 4087 4088 if (!f2fs_is_compress_backend_ready(inode)) { 4089 ret = -EOPNOTSUPP; 4090 goto out; 4091 } 4092 4093 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4094 ret = -EINVAL; 4095 goto out; 4096 } 4097 4098 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4099 if (ret) 4100 goto out; 4101 4102 if (!atomic_read(&fi->i_compr_blocks)) 4103 goto out; 4104 4105 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4106 4107 count = last_idx - page_idx; 4108 while (count && count >= cluster_size) { 4109 ret = redirty_blocks(inode, page_idx, cluster_size); 4110 if (ret < 0) 4111 break; 4112 4113 if (get_dirty_pages(inode) >= blk_per_seg) { 4114 ret = filemap_fdatawrite(inode->i_mapping); 4115 if (ret < 0) 4116 break; 4117 } 4118 4119 count -= cluster_size; 4120 page_idx += cluster_size; 4121 4122 cond_resched(); 4123 if (fatal_signal_pending(current)) { 4124 ret = -EINTR; 4125 break; 4126 } 4127 } 4128 4129 if (!ret) 4130 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4131 LLONG_MAX); 4132 4133 if (ret) 4134 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4135 __func__, ret); 4136 out: 4137 inode_unlock(inode); 4138 file_end_write(filp); 4139 4140 return ret; 4141 } 4142 4143 static int f2fs_ioc_compress_file(struct file *filp) 4144 { 4145 struct inode *inode = file_inode(filp); 4146 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4147 pgoff_t page_idx = 0, last_idx; 4148 unsigned int blk_per_seg = sbi->blocks_per_seg; 4149 int cluster_size = F2FS_I(inode)->i_cluster_size; 4150 int count, ret; 4151 4152 if (!f2fs_sb_has_compression(sbi) || 4153 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4154 return -EOPNOTSUPP; 4155 4156 if (!(filp->f_mode & FMODE_WRITE)) 4157 return -EBADF; 4158 4159 if (!f2fs_compressed_file(inode)) 4160 return -EINVAL; 4161 4162 f2fs_balance_fs(sbi, true); 4163 4164 file_start_write(filp); 4165 inode_lock(inode); 4166 4167 if (!f2fs_is_compress_backend_ready(inode)) { 4168 ret = -EOPNOTSUPP; 4169 goto out; 4170 } 4171 4172 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4173 ret = -EINVAL; 4174 goto out; 4175 } 4176 4177 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4178 if (ret) 4179 goto out; 4180 4181 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4182 4183 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4184 4185 count = last_idx - page_idx; 4186 while (count && count >= cluster_size) { 4187 ret = redirty_blocks(inode, page_idx, cluster_size); 4188 if (ret < 0) 4189 break; 4190 4191 if (get_dirty_pages(inode) >= blk_per_seg) { 4192 ret = filemap_fdatawrite(inode->i_mapping); 4193 if (ret < 0) 4194 break; 4195 } 4196 4197 count -= cluster_size; 4198 page_idx += cluster_size; 4199 4200 cond_resched(); 4201 if (fatal_signal_pending(current)) { 4202 ret = -EINTR; 4203 break; 4204 } 4205 } 4206 4207 if (!ret) 4208 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4209 LLONG_MAX); 4210 4211 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4212 4213 if (ret) 4214 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4215 __func__, ret); 4216 out: 4217 inode_unlock(inode); 4218 file_end_write(filp); 4219 4220 return ret; 4221 } 4222 4223 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4224 { 4225 switch (cmd) { 4226 case FS_IOC_GETVERSION: 4227 return f2fs_ioc_getversion(filp, arg); 4228 case F2FS_IOC_START_ATOMIC_WRITE: 4229 return f2fs_ioc_start_atomic_write(filp, false); 4230 case F2FS_IOC_START_ATOMIC_REPLACE: 4231 return f2fs_ioc_start_atomic_write(filp, true); 4232 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4233 return f2fs_ioc_commit_atomic_write(filp); 4234 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4235 return f2fs_ioc_abort_atomic_write(filp); 4236 case F2FS_IOC_START_VOLATILE_WRITE: 4237 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4238 return -EOPNOTSUPP; 4239 case F2FS_IOC_SHUTDOWN: 4240 return f2fs_ioc_shutdown(filp, arg); 4241 case FITRIM: 4242 return f2fs_ioc_fitrim(filp, arg); 4243 case FS_IOC_SET_ENCRYPTION_POLICY: 4244 return f2fs_ioc_set_encryption_policy(filp, arg); 4245 case FS_IOC_GET_ENCRYPTION_POLICY: 4246 return f2fs_ioc_get_encryption_policy(filp, arg); 4247 case FS_IOC_GET_ENCRYPTION_PWSALT: 4248 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4249 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4250 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4251 case FS_IOC_ADD_ENCRYPTION_KEY: 4252 return f2fs_ioc_add_encryption_key(filp, arg); 4253 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4254 return f2fs_ioc_remove_encryption_key(filp, arg); 4255 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4256 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4257 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4258 return f2fs_ioc_get_encryption_key_status(filp, arg); 4259 case FS_IOC_GET_ENCRYPTION_NONCE: 4260 return f2fs_ioc_get_encryption_nonce(filp, arg); 4261 case F2FS_IOC_GARBAGE_COLLECT: 4262 return f2fs_ioc_gc(filp, arg); 4263 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4264 return f2fs_ioc_gc_range(filp, arg); 4265 case F2FS_IOC_WRITE_CHECKPOINT: 4266 return f2fs_ioc_write_checkpoint(filp); 4267 case F2FS_IOC_DEFRAGMENT: 4268 return f2fs_ioc_defragment(filp, arg); 4269 case F2FS_IOC_MOVE_RANGE: 4270 return f2fs_ioc_move_range(filp, arg); 4271 case F2FS_IOC_FLUSH_DEVICE: 4272 return f2fs_ioc_flush_device(filp, arg); 4273 case F2FS_IOC_GET_FEATURES: 4274 return f2fs_ioc_get_features(filp, arg); 4275 case F2FS_IOC_GET_PIN_FILE: 4276 return f2fs_ioc_get_pin_file(filp, arg); 4277 case F2FS_IOC_SET_PIN_FILE: 4278 return f2fs_ioc_set_pin_file(filp, arg); 4279 case F2FS_IOC_PRECACHE_EXTENTS: 4280 return f2fs_ioc_precache_extents(filp); 4281 case F2FS_IOC_RESIZE_FS: 4282 return f2fs_ioc_resize_fs(filp, arg); 4283 case FS_IOC_ENABLE_VERITY: 4284 return f2fs_ioc_enable_verity(filp, arg); 4285 case FS_IOC_MEASURE_VERITY: 4286 return f2fs_ioc_measure_verity(filp, arg); 4287 case FS_IOC_READ_VERITY_METADATA: 4288 return f2fs_ioc_read_verity_metadata(filp, arg); 4289 case FS_IOC_GETFSLABEL: 4290 return f2fs_ioc_getfslabel(filp, arg); 4291 case FS_IOC_SETFSLABEL: 4292 return f2fs_ioc_setfslabel(filp, arg); 4293 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4294 return f2fs_ioc_get_compress_blocks(filp, arg); 4295 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4296 return f2fs_release_compress_blocks(filp, arg); 4297 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4298 return f2fs_reserve_compress_blocks(filp, arg); 4299 case F2FS_IOC_SEC_TRIM_FILE: 4300 return f2fs_sec_trim_file(filp, arg); 4301 case F2FS_IOC_GET_COMPRESS_OPTION: 4302 return f2fs_ioc_get_compress_option(filp, arg); 4303 case F2FS_IOC_SET_COMPRESS_OPTION: 4304 return f2fs_ioc_set_compress_option(filp, arg); 4305 case F2FS_IOC_DECOMPRESS_FILE: 4306 return f2fs_ioc_decompress_file(filp); 4307 case F2FS_IOC_COMPRESS_FILE: 4308 return f2fs_ioc_compress_file(filp); 4309 default: 4310 return -ENOTTY; 4311 } 4312 } 4313 4314 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4315 { 4316 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4317 return -EIO; 4318 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4319 return -ENOSPC; 4320 4321 return __f2fs_ioctl(filp, cmd, arg); 4322 } 4323 4324 /* 4325 * Return %true if the given read or write request should use direct I/O, or 4326 * %false if it should use buffered I/O. 4327 */ 4328 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4329 struct iov_iter *iter) 4330 { 4331 unsigned int align; 4332 4333 if (!(iocb->ki_flags & IOCB_DIRECT)) 4334 return false; 4335 4336 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4337 return false; 4338 4339 /* 4340 * Direct I/O not aligned to the disk's logical_block_size will be 4341 * attempted, but will fail with -EINVAL. 4342 * 4343 * f2fs additionally requires that direct I/O be aligned to the 4344 * filesystem block size, which is often a stricter requirement. 4345 * However, f2fs traditionally falls back to buffered I/O on requests 4346 * that are logical_block_size-aligned but not fs-block aligned. 4347 * 4348 * The below logic implements this behavior. 4349 */ 4350 align = iocb->ki_pos | iov_iter_alignment(iter); 4351 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4352 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4353 return false; 4354 4355 return true; 4356 } 4357 4358 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4359 unsigned int flags) 4360 { 4361 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4362 4363 dec_page_count(sbi, F2FS_DIO_READ); 4364 if (error) 4365 return error; 4366 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4367 return 0; 4368 } 4369 4370 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4371 .end_io = f2fs_dio_read_end_io, 4372 }; 4373 4374 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4375 { 4376 struct file *file = iocb->ki_filp; 4377 struct inode *inode = file_inode(file); 4378 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4379 struct f2fs_inode_info *fi = F2FS_I(inode); 4380 const loff_t pos = iocb->ki_pos; 4381 const size_t count = iov_iter_count(to); 4382 struct iomap_dio *dio; 4383 ssize_t ret; 4384 4385 if (count == 0) 4386 return 0; /* skip atime update */ 4387 4388 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4389 4390 if (iocb->ki_flags & IOCB_NOWAIT) { 4391 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4392 ret = -EAGAIN; 4393 goto out; 4394 } 4395 } else { 4396 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4397 } 4398 4399 /* 4400 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4401 * the higher-level function iomap_dio_rw() in order to ensure that the 4402 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4403 */ 4404 inc_page_count(sbi, F2FS_DIO_READ); 4405 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4406 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4407 if (IS_ERR_OR_NULL(dio)) { 4408 ret = PTR_ERR_OR_ZERO(dio); 4409 if (ret != -EIOCBQUEUED) 4410 dec_page_count(sbi, F2FS_DIO_READ); 4411 } else { 4412 ret = iomap_dio_complete(dio); 4413 } 4414 4415 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4416 4417 file_accessed(file); 4418 out: 4419 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4420 return ret; 4421 } 4422 4423 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4424 int rw) 4425 { 4426 struct inode *inode = file_inode(file); 4427 char *buf, *path; 4428 4429 buf = f2fs_getname(F2FS_I_SB(inode)); 4430 if (!buf) 4431 return; 4432 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4433 if (IS_ERR(path)) 4434 goto free_buf; 4435 if (rw == WRITE) 4436 trace_f2fs_datawrite_start(inode, pos, count, 4437 current->pid, path, current->comm); 4438 else 4439 trace_f2fs_dataread_start(inode, pos, count, 4440 current->pid, path, current->comm); 4441 free_buf: 4442 f2fs_putname(buf); 4443 } 4444 4445 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4446 { 4447 struct inode *inode = file_inode(iocb->ki_filp); 4448 const loff_t pos = iocb->ki_pos; 4449 ssize_t ret; 4450 4451 if (!f2fs_is_compress_backend_ready(inode)) 4452 return -EOPNOTSUPP; 4453 4454 if (trace_f2fs_dataread_start_enabled()) 4455 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4456 iov_iter_count(to), READ); 4457 4458 if (f2fs_should_use_dio(inode, iocb, to)) { 4459 ret = f2fs_dio_read_iter(iocb, to); 4460 } else { 4461 ret = filemap_read(iocb, to, 0); 4462 if (ret > 0) 4463 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4464 APP_BUFFERED_READ_IO, ret); 4465 } 4466 if (trace_f2fs_dataread_end_enabled()) 4467 trace_f2fs_dataread_end(inode, pos, ret); 4468 return ret; 4469 } 4470 4471 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4472 struct pipe_inode_info *pipe, 4473 size_t len, unsigned int flags) 4474 { 4475 struct inode *inode = file_inode(in); 4476 const loff_t pos = *ppos; 4477 ssize_t ret; 4478 4479 if (!f2fs_is_compress_backend_ready(inode)) 4480 return -EOPNOTSUPP; 4481 4482 if (trace_f2fs_dataread_start_enabled()) 4483 f2fs_trace_rw_file_path(in, pos, len, READ); 4484 4485 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4486 if (ret > 0) 4487 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4488 APP_BUFFERED_READ_IO, ret); 4489 4490 if (trace_f2fs_dataread_end_enabled()) 4491 trace_f2fs_dataread_end(inode, pos, ret); 4492 return ret; 4493 } 4494 4495 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4496 { 4497 struct file *file = iocb->ki_filp; 4498 struct inode *inode = file_inode(file); 4499 ssize_t count; 4500 int err; 4501 4502 if (IS_IMMUTABLE(inode)) 4503 return -EPERM; 4504 4505 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4506 return -EPERM; 4507 4508 count = generic_write_checks(iocb, from); 4509 if (count <= 0) 4510 return count; 4511 4512 err = file_modified(file); 4513 if (err) 4514 return err; 4515 return count; 4516 } 4517 4518 /* 4519 * Preallocate blocks for a write request, if it is possible and helpful to do 4520 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4521 * blocks were preallocated, or a negative errno value if something went 4522 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4523 * requested blocks (not just some of them) have been allocated. 4524 */ 4525 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4526 bool dio) 4527 { 4528 struct inode *inode = file_inode(iocb->ki_filp); 4529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4530 const loff_t pos = iocb->ki_pos; 4531 const size_t count = iov_iter_count(iter); 4532 struct f2fs_map_blocks map = {}; 4533 int flag; 4534 int ret; 4535 4536 /* If it will be an out-of-place direct write, don't bother. */ 4537 if (dio && f2fs_lfs_mode(sbi)) 4538 return 0; 4539 /* 4540 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4541 * buffered IO, if DIO meets any holes. 4542 */ 4543 if (dio && i_size_read(inode) && 4544 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4545 return 0; 4546 4547 /* No-wait I/O can't allocate blocks. */ 4548 if (iocb->ki_flags & IOCB_NOWAIT) 4549 return 0; 4550 4551 /* If it will be a short write, don't bother. */ 4552 if (fault_in_iov_iter_readable(iter, count)) 4553 return 0; 4554 4555 if (f2fs_has_inline_data(inode)) { 4556 /* If the data will fit inline, don't bother. */ 4557 if (pos + count <= MAX_INLINE_DATA(inode)) 4558 return 0; 4559 ret = f2fs_convert_inline_inode(inode); 4560 if (ret) 4561 return ret; 4562 } 4563 4564 /* Do not preallocate blocks that will be written partially in 4KB. */ 4565 map.m_lblk = F2FS_BLK_ALIGN(pos); 4566 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4567 if (map.m_len > map.m_lblk) 4568 map.m_len -= map.m_lblk; 4569 else 4570 return 0; 4571 4572 map.m_may_create = true; 4573 if (dio) { 4574 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4575 flag = F2FS_GET_BLOCK_PRE_DIO; 4576 } else { 4577 map.m_seg_type = NO_CHECK_TYPE; 4578 flag = F2FS_GET_BLOCK_PRE_AIO; 4579 } 4580 4581 ret = f2fs_map_blocks(inode, &map, flag); 4582 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4583 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4584 return ret; 4585 if (ret == 0) 4586 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4587 return map.m_len; 4588 } 4589 4590 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4591 struct iov_iter *from) 4592 { 4593 struct file *file = iocb->ki_filp; 4594 struct inode *inode = file_inode(file); 4595 ssize_t ret; 4596 4597 if (iocb->ki_flags & IOCB_NOWAIT) 4598 return -EOPNOTSUPP; 4599 4600 ret = generic_perform_write(iocb, from); 4601 4602 if (ret > 0) { 4603 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4604 APP_BUFFERED_IO, ret); 4605 } 4606 return ret; 4607 } 4608 4609 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4610 unsigned int flags) 4611 { 4612 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4613 4614 dec_page_count(sbi, F2FS_DIO_WRITE); 4615 if (error) 4616 return error; 4617 f2fs_update_time(sbi, REQ_TIME); 4618 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4619 return 0; 4620 } 4621 4622 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4623 .end_io = f2fs_dio_write_end_io, 4624 }; 4625 4626 static void f2fs_flush_buffered_write(struct address_space *mapping, 4627 loff_t start_pos, loff_t end_pos) 4628 { 4629 int ret; 4630 4631 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4632 if (ret < 0) 4633 return; 4634 invalidate_mapping_pages(mapping, 4635 start_pos >> PAGE_SHIFT, 4636 end_pos >> PAGE_SHIFT); 4637 } 4638 4639 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4640 bool *may_need_sync) 4641 { 4642 struct file *file = iocb->ki_filp; 4643 struct inode *inode = file_inode(file); 4644 struct f2fs_inode_info *fi = F2FS_I(inode); 4645 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4646 const bool do_opu = f2fs_lfs_mode(sbi); 4647 const loff_t pos = iocb->ki_pos; 4648 const ssize_t count = iov_iter_count(from); 4649 unsigned int dio_flags; 4650 struct iomap_dio *dio; 4651 ssize_t ret; 4652 4653 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4654 4655 if (iocb->ki_flags & IOCB_NOWAIT) { 4656 /* f2fs_convert_inline_inode() and block allocation can block */ 4657 if (f2fs_has_inline_data(inode) || 4658 !f2fs_overwrite_io(inode, pos, count)) { 4659 ret = -EAGAIN; 4660 goto out; 4661 } 4662 4663 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4664 ret = -EAGAIN; 4665 goto out; 4666 } 4667 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4668 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4669 ret = -EAGAIN; 4670 goto out; 4671 } 4672 } else { 4673 ret = f2fs_convert_inline_inode(inode); 4674 if (ret) 4675 goto out; 4676 4677 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4678 if (do_opu) 4679 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4680 } 4681 4682 /* 4683 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4684 * the higher-level function iomap_dio_rw() in order to ensure that the 4685 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4686 */ 4687 inc_page_count(sbi, F2FS_DIO_WRITE); 4688 dio_flags = 0; 4689 if (pos + count > inode->i_size) 4690 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4691 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4692 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4693 if (IS_ERR_OR_NULL(dio)) { 4694 ret = PTR_ERR_OR_ZERO(dio); 4695 if (ret == -ENOTBLK) 4696 ret = 0; 4697 if (ret != -EIOCBQUEUED) 4698 dec_page_count(sbi, F2FS_DIO_WRITE); 4699 } else { 4700 ret = iomap_dio_complete(dio); 4701 } 4702 4703 if (do_opu) 4704 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4705 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4706 4707 if (ret < 0) 4708 goto out; 4709 if (pos + ret > inode->i_size) 4710 f2fs_i_size_write(inode, pos + ret); 4711 if (!do_opu) 4712 set_inode_flag(inode, FI_UPDATE_WRITE); 4713 4714 if (iov_iter_count(from)) { 4715 ssize_t ret2; 4716 loff_t bufio_start_pos = iocb->ki_pos; 4717 4718 /* 4719 * The direct write was partial, so we need to fall back to a 4720 * buffered write for the remainder. 4721 */ 4722 4723 ret2 = f2fs_buffered_write_iter(iocb, from); 4724 if (iov_iter_count(from)) 4725 f2fs_write_failed(inode, iocb->ki_pos); 4726 if (ret2 < 0) 4727 goto out; 4728 4729 /* 4730 * Ensure that the pagecache pages are written to disk and 4731 * invalidated to preserve the expected O_DIRECT semantics. 4732 */ 4733 if (ret2 > 0) { 4734 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4735 4736 ret += ret2; 4737 4738 f2fs_flush_buffered_write(file->f_mapping, 4739 bufio_start_pos, 4740 bufio_end_pos); 4741 } 4742 } else { 4743 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4744 *may_need_sync = false; 4745 } 4746 out: 4747 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4748 return ret; 4749 } 4750 4751 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4752 { 4753 struct inode *inode = file_inode(iocb->ki_filp); 4754 const loff_t orig_pos = iocb->ki_pos; 4755 const size_t orig_count = iov_iter_count(from); 4756 loff_t target_size; 4757 bool dio; 4758 bool may_need_sync = true; 4759 int preallocated; 4760 ssize_t ret; 4761 4762 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4763 ret = -EIO; 4764 goto out; 4765 } 4766 4767 if (!f2fs_is_compress_backend_ready(inode)) { 4768 ret = -EOPNOTSUPP; 4769 goto out; 4770 } 4771 4772 if (iocb->ki_flags & IOCB_NOWAIT) { 4773 if (!inode_trylock(inode)) { 4774 ret = -EAGAIN; 4775 goto out; 4776 } 4777 } else { 4778 inode_lock(inode); 4779 } 4780 4781 ret = f2fs_write_checks(iocb, from); 4782 if (ret <= 0) 4783 goto out_unlock; 4784 4785 /* Determine whether we will do a direct write or a buffered write. */ 4786 dio = f2fs_should_use_dio(inode, iocb, from); 4787 4788 /* Possibly preallocate the blocks for the write. */ 4789 target_size = iocb->ki_pos + iov_iter_count(from); 4790 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4791 if (preallocated < 0) { 4792 ret = preallocated; 4793 } else { 4794 if (trace_f2fs_datawrite_start_enabled()) 4795 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4796 orig_count, WRITE); 4797 4798 /* Do the actual write. */ 4799 ret = dio ? 4800 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4801 f2fs_buffered_write_iter(iocb, from); 4802 4803 if (trace_f2fs_datawrite_end_enabled()) 4804 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4805 } 4806 4807 /* Don't leave any preallocated blocks around past i_size. */ 4808 if (preallocated && i_size_read(inode) < target_size) { 4809 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4810 filemap_invalidate_lock(inode->i_mapping); 4811 if (!f2fs_truncate(inode)) 4812 file_dont_truncate(inode); 4813 filemap_invalidate_unlock(inode->i_mapping); 4814 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4815 } else { 4816 file_dont_truncate(inode); 4817 } 4818 4819 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4820 out_unlock: 4821 inode_unlock(inode); 4822 out: 4823 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4824 4825 if (ret > 0 && may_need_sync) 4826 ret = generic_write_sync(iocb, ret); 4827 4828 /* If buffered IO was forced, flush and drop the data from 4829 * the page cache to preserve O_DIRECT semantics 4830 */ 4831 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4832 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4833 orig_pos, 4834 orig_pos + ret - 1); 4835 4836 return ret; 4837 } 4838 4839 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4840 int advice) 4841 { 4842 struct address_space *mapping; 4843 struct backing_dev_info *bdi; 4844 struct inode *inode = file_inode(filp); 4845 int err; 4846 4847 if (advice == POSIX_FADV_SEQUENTIAL) { 4848 if (S_ISFIFO(inode->i_mode)) 4849 return -ESPIPE; 4850 4851 mapping = filp->f_mapping; 4852 if (!mapping || len < 0) 4853 return -EINVAL; 4854 4855 bdi = inode_to_bdi(mapping->host); 4856 filp->f_ra.ra_pages = bdi->ra_pages * 4857 F2FS_I_SB(inode)->seq_file_ra_mul; 4858 spin_lock(&filp->f_lock); 4859 filp->f_mode &= ~FMODE_RANDOM; 4860 spin_unlock(&filp->f_lock); 4861 return 0; 4862 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 4863 /* Load extent cache at the first readahead. */ 4864 f2fs_precache_extents(inode); 4865 } 4866 4867 err = generic_fadvise(filp, offset, len, advice); 4868 if (!err && advice == POSIX_FADV_DONTNEED && 4869 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4870 f2fs_compressed_file(inode)) 4871 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4872 4873 return err; 4874 } 4875 4876 #ifdef CONFIG_COMPAT 4877 struct compat_f2fs_gc_range { 4878 u32 sync; 4879 compat_u64 start; 4880 compat_u64 len; 4881 }; 4882 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4883 struct compat_f2fs_gc_range) 4884 4885 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4886 { 4887 struct compat_f2fs_gc_range __user *urange; 4888 struct f2fs_gc_range range; 4889 int err; 4890 4891 urange = compat_ptr(arg); 4892 err = get_user(range.sync, &urange->sync); 4893 err |= get_user(range.start, &urange->start); 4894 err |= get_user(range.len, &urange->len); 4895 if (err) 4896 return -EFAULT; 4897 4898 return __f2fs_ioc_gc_range(file, &range); 4899 } 4900 4901 struct compat_f2fs_move_range { 4902 u32 dst_fd; 4903 compat_u64 pos_in; 4904 compat_u64 pos_out; 4905 compat_u64 len; 4906 }; 4907 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4908 struct compat_f2fs_move_range) 4909 4910 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4911 { 4912 struct compat_f2fs_move_range __user *urange; 4913 struct f2fs_move_range range; 4914 int err; 4915 4916 urange = compat_ptr(arg); 4917 err = get_user(range.dst_fd, &urange->dst_fd); 4918 err |= get_user(range.pos_in, &urange->pos_in); 4919 err |= get_user(range.pos_out, &urange->pos_out); 4920 err |= get_user(range.len, &urange->len); 4921 if (err) 4922 return -EFAULT; 4923 4924 return __f2fs_ioc_move_range(file, &range); 4925 } 4926 4927 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4928 { 4929 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4930 return -EIO; 4931 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4932 return -ENOSPC; 4933 4934 switch (cmd) { 4935 case FS_IOC32_GETVERSION: 4936 cmd = FS_IOC_GETVERSION; 4937 break; 4938 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4939 return f2fs_compat_ioc_gc_range(file, arg); 4940 case F2FS_IOC32_MOVE_RANGE: 4941 return f2fs_compat_ioc_move_range(file, arg); 4942 case F2FS_IOC_START_ATOMIC_WRITE: 4943 case F2FS_IOC_START_ATOMIC_REPLACE: 4944 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4945 case F2FS_IOC_START_VOLATILE_WRITE: 4946 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4947 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4948 case F2FS_IOC_SHUTDOWN: 4949 case FITRIM: 4950 case FS_IOC_SET_ENCRYPTION_POLICY: 4951 case FS_IOC_GET_ENCRYPTION_PWSALT: 4952 case FS_IOC_GET_ENCRYPTION_POLICY: 4953 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4954 case FS_IOC_ADD_ENCRYPTION_KEY: 4955 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4956 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4957 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4958 case FS_IOC_GET_ENCRYPTION_NONCE: 4959 case F2FS_IOC_GARBAGE_COLLECT: 4960 case F2FS_IOC_WRITE_CHECKPOINT: 4961 case F2FS_IOC_DEFRAGMENT: 4962 case F2FS_IOC_FLUSH_DEVICE: 4963 case F2FS_IOC_GET_FEATURES: 4964 case F2FS_IOC_GET_PIN_FILE: 4965 case F2FS_IOC_SET_PIN_FILE: 4966 case F2FS_IOC_PRECACHE_EXTENTS: 4967 case F2FS_IOC_RESIZE_FS: 4968 case FS_IOC_ENABLE_VERITY: 4969 case FS_IOC_MEASURE_VERITY: 4970 case FS_IOC_READ_VERITY_METADATA: 4971 case FS_IOC_GETFSLABEL: 4972 case FS_IOC_SETFSLABEL: 4973 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4974 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4975 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4976 case F2FS_IOC_SEC_TRIM_FILE: 4977 case F2FS_IOC_GET_COMPRESS_OPTION: 4978 case F2FS_IOC_SET_COMPRESS_OPTION: 4979 case F2FS_IOC_DECOMPRESS_FILE: 4980 case F2FS_IOC_COMPRESS_FILE: 4981 break; 4982 default: 4983 return -ENOIOCTLCMD; 4984 } 4985 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4986 } 4987 #endif 4988 4989 const struct file_operations f2fs_file_operations = { 4990 .llseek = f2fs_llseek, 4991 .read_iter = f2fs_file_read_iter, 4992 .write_iter = f2fs_file_write_iter, 4993 .iopoll = iocb_bio_iopoll, 4994 .open = f2fs_file_open, 4995 .release = f2fs_release_file, 4996 .mmap = f2fs_file_mmap, 4997 .flush = f2fs_file_flush, 4998 .fsync = f2fs_sync_file, 4999 .fallocate = f2fs_fallocate, 5000 .unlocked_ioctl = f2fs_ioctl, 5001 #ifdef CONFIG_COMPAT 5002 .compat_ioctl = f2fs_compat_ioctl, 5003 #endif 5004 .splice_read = f2fs_file_splice_read, 5005 .splice_write = iter_file_splice_write, 5006 .fadvise = f2fs_file_fadvise, 5007 }; 5008