1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/file.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/fs.h> 9 #include <linux/f2fs_fs.h> 10 #include <linux/stat.h> 11 #include <linux/buffer_head.h> 12 #include <linux/writeback.h> 13 #include <linux/blkdev.h> 14 #include <linux/falloc.h> 15 #include <linux/types.h> 16 #include <linux/compat.h> 17 #include <linux/uaccess.h> 18 #include <linux/mount.h> 19 #include <linux/pagevec.h> 20 #include <linux/uio.h> 21 #include <linux/uuid.h> 22 #include <linux/file.h> 23 #include <linux/nls.h> 24 #include <linux/sched/signal.h> 25 #include <linux/fileattr.h> 26 #include <linux/fadvise.h> 27 #include <linux/iomap.h> 28 29 #include "f2fs.h" 30 #include "node.h" 31 #include "segment.h" 32 #include "xattr.h" 33 #include "acl.h" 34 #include "gc.h" 35 #include "iostat.h" 36 #include <trace/events/f2fs.h> 37 #include <uapi/linux/f2fs.h> 38 39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 40 { 41 struct inode *inode = file_inode(vmf->vma->vm_file); 42 vm_fault_t ret; 43 44 ret = filemap_fault(vmf); 45 if (ret & VM_FAULT_LOCKED) 46 f2fs_update_iostat(F2FS_I_SB(inode), inode, 47 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 48 49 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret); 50 51 return ret; 52 } 53 54 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 55 { 56 struct page *page = vmf->page; 57 struct inode *inode = file_inode(vmf->vma->vm_file); 58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 59 struct dnode_of_data dn; 60 bool need_alloc = true; 61 int err = 0; 62 63 if (unlikely(IS_IMMUTABLE(inode))) 64 return VM_FAULT_SIGBUS; 65 66 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 67 return VM_FAULT_SIGBUS; 68 69 if (unlikely(f2fs_cp_error(sbi))) { 70 err = -EIO; 71 goto err; 72 } 73 74 if (!f2fs_is_checkpoint_ready(sbi)) { 75 err = -ENOSPC; 76 goto err; 77 } 78 79 err = f2fs_convert_inline_inode(inode); 80 if (err) 81 goto err; 82 83 #ifdef CONFIG_F2FS_FS_COMPRESSION 84 if (f2fs_compressed_file(inode)) { 85 int ret = f2fs_is_compressed_cluster(inode, page->index); 86 87 if (ret < 0) { 88 err = ret; 89 goto err; 90 } else if (ret) { 91 need_alloc = false; 92 } 93 } 94 #endif 95 /* should do out of any locked page */ 96 if (need_alloc) 97 f2fs_balance_fs(sbi, true); 98 99 sb_start_pagefault(inode->i_sb); 100 101 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 102 103 file_update_time(vmf->vma->vm_file); 104 filemap_invalidate_lock_shared(inode->i_mapping); 105 lock_page(page); 106 if (unlikely(page->mapping != inode->i_mapping || 107 page_offset(page) > i_size_read(inode) || 108 !PageUptodate(page))) { 109 unlock_page(page); 110 err = -EFAULT; 111 goto out_sem; 112 } 113 114 if (need_alloc) { 115 /* block allocation */ 116 set_new_dnode(&dn, inode, NULL, NULL, 0); 117 err = f2fs_get_block_locked(&dn, page->index); 118 } 119 120 #ifdef CONFIG_F2FS_FS_COMPRESSION 121 if (!need_alloc) { 122 set_new_dnode(&dn, inode, NULL, NULL, 0); 123 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE); 124 f2fs_put_dnode(&dn); 125 } 126 #endif 127 if (err) { 128 unlock_page(page); 129 goto out_sem; 130 } 131 132 f2fs_wait_on_page_writeback(page, DATA, false, true); 133 134 /* wait for GCed page writeback via META_MAPPING */ 135 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 136 137 /* 138 * check to see if the page is mapped already (no holes) 139 */ 140 if (PageMappedToDisk(page)) 141 goto out_sem; 142 143 /* page is wholly or partially inside EOF */ 144 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 145 i_size_read(inode)) { 146 loff_t offset; 147 148 offset = i_size_read(inode) & ~PAGE_MASK; 149 zero_user_segment(page, offset, PAGE_SIZE); 150 } 151 set_page_dirty(page); 152 153 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 154 f2fs_update_time(sbi, REQ_TIME); 155 156 trace_f2fs_vm_page_mkwrite(page, DATA); 157 out_sem: 158 filemap_invalidate_unlock_shared(inode->i_mapping); 159 160 sb_end_pagefault(inode->i_sb); 161 err: 162 return vmf_fs_error(err); 163 } 164 165 static const struct vm_operations_struct f2fs_file_vm_ops = { 166 .fault = f2fs_filemap_fault, 167 .map_pages = filemap_map_pages, 168 .page_mkwrite = f2fs_vm_page_mkwrite, 169 }; 170 171 static int get_parent_ino(struct inode *inode, nid_t *pino) 172 { 173 struct dentry *dentry; 174 175 /* 176 * Make sure to get the non-deleted alias. The alias associated with 177 * the open file descriptor being fsync()'ed may be deleted already. 178 */ 179 dentry = d_find_alias(inode); 180 if (!dentry) 181 return 0; 182 183 *pino = parent_ino(dentry); 184 dput(dentry); 185 return 1; 186 } 187 188 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 189 { 190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 191 enum cp_reason_type cp_reason = CP_NO_NEEDED; 192 193 if (!S_ISREG(inode->i_mode)) 194 cp_reason = CP_NON_REGULAR; 195 else if (f2fs_compressed_file(inode)) 196 cp_reason = CP_COMPRESSED; 197 else if (inode->i_nlink != 1) 198 cp_reason = CP_HARDLINK; 199 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 200 cp_reason = CP_SB_NEED_CP; 201 else if (file_wrong_pino(inode)) 202 cp_reason = CP_WRONG_PINO; 203 else if (!f2fs_space_for_roll_forward(sbi)) 204 cp_reason = CP_NO_SPC_ROLL; 205 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 206 cp_reason = CP_NODE_NEED_CP; 207 else if (test_opt(sbi, FASTBOOT)) 208 cp_reason = CP_FASTBOOT_MODE; 209 else if (F2FS_OPTION(sbi).active_logs == 2) 210 cp_reason = CP_SPEC_LOG_NUM; 211 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 212 f2fs_need_dentry_mark(sbi, inode->i_ino) && 213 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 214 TRANS_DIR_INO)) 215 cp_reason = CP_RECOVER_DIR; 216 217 return cp_reason; 218 } 219 220 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 221 { 222 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 223 bool ret = false; 224 /* But we need to avoid that there are some inode updates */ 225 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 226 ret = true; 227 f2fs_put_page(i, 0); 228 return ret; 229 } 230 231 static void try_to_fix_pino(struct inode *inode) 232 { 233 struct f2fs_inode_info *fi = F2FS_I(inode); 234 nid_t pino; 235 236 f2fs_down_write(&fi->i_sem); 237 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 238 get_parent_ino(inode, &pino)) { 239 f2fs_i_pino_write(inode, pino); 240 file_got_pino(inode); 241 } 242 f2fs_up_write(&fi->i_sem); 243 } 244 245 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 246 int datasync, bool atomic) 247 { 248 struct inode *inode = file->f_mapping->host; 249 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 250 nid_t ino = inode->i_ino; 251 int ret = 0; 252 enum cp_reason_type cp_reason = 0; 253 struct writeback_control wbc = { 254 .sync_mode = WB_SYNC_ALL, 255 .nr_to_write = LONG_MAX, 256 .for_reclaim = 0, 257 }; 258 unsigned int seq_id = 0; 259 260 if (unlikely(f2fs_readonly(inode->i_sb))) 261 return 0; 262 263 trace_f2fs_sync_file_enter(inode); 264 265 if (S_ISDIR(inode->i_mode)) 266 goto go_write; 267 268 /* if fdatasync is triggered, let's do in-place-update */ 269 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 270 set_inode_flag(inode, FI_NEED_IPU); 271 ret = file_write_and_wait_range(file, start, end); 272 clear_inode_flag(inode, FI_NEED_IPU); 273 274 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 275 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 276 return ret; 277 } 278 279 /* if the inode is dirty, let's recover all the time */ 280 if (!f2fs_skip_inode_update(inode, datasync)) { 281 f2fs_write_inode(inode, NULL); 282 goto go_write; 283 } 284 285 /* 286 * if there is no written data, don't waste time to write recovery info. 287 */ 288 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 289 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 290 291 /* it may call write_inode just prior to fsync */ 292 if (need_inode_page_update(sbi, ino)) 293 goto go_write; 294 295 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 296 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 297 goto flush_out; 298 goto out; 299 } else { 300 /* 301 * for OPU case, during fsync(), node can be persisted before 302 * data when lower device doesn't support write barrier, result 303 * in data corruption after SPO. 304 * So for strict fsync mode, force to use atomic write semantics 305 * to keep write order in between data/node and last node to 306 * avoid potential data corruption. 307 */ 308 if (F2FS_OPTION(sbi).fsync_mode == 309 FSYNC_MODE_STRICT && !atomic) 310 atomic = true; 311 } 312 go_write: 313 /* 314 * Both of fdatasync() and fsync() are able to be recovered from 315 * sudden-power-off. 316 */ 317 f2fs_down_read(&F2FS_I(inode)->i_sem); 318 cp_reason = need_do_checkpoint(inode); 319 f2fs_up_read(&F2FS_I(inode)->i_sem); 320 321 if (cp_reason) { 322 /* all the dirty node pages should be flushed for POR */ 323 ret = f2fs_sync_fs(inode->i_sb, 1); 324 325 /* 326 * We've secured consistency through sync_fs. Following pino 327 * will be used only for fsynced inodes after checkpoint. 328 */ 329 try_to_fix_pino(inode); 330 clear_inode_flag(inode, FI_APPEND_WRITE); 331 clear_inode_flag(inode, FI_UPDATE_WRITE); 332 goto out; 333 } 334 sync_nodes: 335 atomic_inc(&sbi->wb_sync_req[NODE]); 336 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 337 atomic_dec(&sbi->wb_sync_req[NODE]); 338 if (ret) 339 goto out; 340 341 /* if cp_error was enabled, we should avoid infinite loop */ 342 if (unlikely(f2fs_cp_error(sbi))) { 343 ret = -EIO; 344 goto out; 345 } 346 347 if (f2fs_need_inode_block_update(sbi, ino)) { 348 f2fs_mark_inode_dirty_sync(inode, true); 349 f2fs_write_inode(inode, NULL); 350 goto sync_nodes; 351 } 352 353 /* 354 * If it's atomic_write, it's just fine to keep write ordering. So 355 * here we don't need to wait for node write completion, since we use 356 * node chain which serializes node blocks. If one of node writes are 357 * reordered, we can see simply broken chain, resulting in stopping 358 * roll-forward recovery. It means we'll recover all or none node blocks 359 * given fsync mark. 360 */ 361 if (!atomic) { 362 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 363 if (ret) 364 goto out; 365 } 366 367 /* once recovery info is written, don't need to tack this */ 368 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 369 clear_inode_flag(inode, FI_APPEND_WRITE); 370 flush_out: 371 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) || 372 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi))) 373 ret = f2fs_issue_flush(sbi, inode->i_ino); 374 if (!ret) { 375 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 376 clear_inode_flag(inode, FI_UPDATE_WRITE); 377 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 378 } 379 f2fs_update_time(sbi, REQ_TIME); 380 out: 381 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 382 return ret; 383 } 384 385 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 386 { 387 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 388 return -EIO; 389 return f2fs_do_sync_file(file, start, end, datasync, false); 390 } 391 392 static bool __found_offset(struct address_space *mapping, block_t blkaddr, 393 pgoff_t index, int whence) 394 { 395 switch (whence) { 396 case SEEK_DATA: 397 if (__is_valid_data_blkaddr(blkaddr)) 398 return true; 399 if (blkaddr == NEW_ADDR && 400 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 401 return true; 402 break; 403 case SEEK_HOLE: 404 if (blkaddr == NULL_ADDR) 405 return true; 406 break; 407 } 408 return false; 409 } 410 411 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 412 { 413 struct inode *inode = file->f_mapping->host; 414 loff_t maxbytes = inode->i_sb->s_maxbytes; 415 struct dnode_of_data dn; 416 pgoff_t pgofs, end_offset; 417 loff_t data_ofs = offset; 418 loff_t isize; 419 int err = 0; 420 421 inode_lock(inode); 422 423 isize = i_size_read(inode); 424 if (offset >= isize) 425 goto fail; 426 427 /* handle inline data case */ 428 if (f2fs_has_inline_data(inode)) { 429 if (whence == SEEK_HOLE) { 430 data_ofs = isize; 431 goto found; 432 } else if (whence == SEEK_DATA) { 433 data_ofs = offset; 434 goto found; 435 } 436 } 437 438 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 439 440 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 441 set_new_dnode(&dn, inode, NULL, NULL, 0); 442 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 443 if (err && err != -ENOENT) { 444 goto fail; 445 } else if (err == -ENOENT) { 446 /* direct node does not exists */ 447 if (whence == SEEK_DATA) { 448 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 449 continue; 450 } else { 451 goto found; 452 } 453 } 454 455 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 456 457 /* find data/hole in dnode block */ 458 for (; dn.ofs_in_node < end_offset; 459 dn.ofs_in_node++, pgofs++, 460 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 461 block_t blkaddr; 462 463 blkaddr = f2fs_data_blkaddr(&dn); 464 465 if (__is_valid_data_blkaddr(blkaddr) && 466 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 467 blkaddr, DATA_GENERIC_ENHANCE)) { 468 f2fs_put_dnode(&dn); 469 goto fail; 470 } 471 472 if (__found_offset(file->f_mapping, blkaddr, 473 pgofs, whence)) { 474 f2fs_put_dnode(&dn); 475 goto found; 476 } 477 } 478 f2fs_put_dnode(&dn); 479 } 480 481 if (whence == SEEK_DATA) 482 goto fail; 483 found: 484 if (whence == SEEK_HOLE && data_ofs > isize) 485 data_ofs = isize; 486 inode_unlock(inode); 487 return vfs_setpos(file, data_ofs, maxbytes); 488 fail: 489 inode_unlock(inode); 490 return -ENXIO; 491 } 492 493 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 494 { 495 struct inode *inode = file->f_mapping->host; 496 loff_t maxbytes = inode->i_sb->s_maxbytes; 497 498 if (f2fs_compressed_file(inode)) 499 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; 500 501 switch (whence) { 502 case SEEK_SET: 503 case SEEK_CUR: 504 case SEEK_END: 505 return generic_file_llseek_size(file, offset, whence, 506 maxbytes, i_size_read(inode)); 507 case SEEK_DATA: 508 case SEEK_HOLE: 509 if (offset < 0) 510 return -ENXIO; 511 return f2fs_seek_block(file, offset, whence); 512 } 513 514 return -EINVAL; 515 } 516 517 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 518 { 519 struct inode *inode = file_inode(file); 520 521 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 522 return -EIO; 523 524 if (!f2fs_is_compress_backend_ready(inode)) 525 return -EOPNOTSUPP; 526 527 file_accessed(file); 528 vma->vm_ops = &f2fs_file_vm_ops; 529 530 f2fs_down_read(&F2FS_I(inode)->i_sem); 531 set_inode_flag(inode, FI_MMAP_FILE); 532 f2fs_up_read(&F2FS_I(inode)->i_sem); 533 534 return 0; 535 } 536 537 static int f2fs_file_open(struct inode *inode, struct file *filp) 538 { 539 int err = fscrypt_file_open(inode, filp); 540 541 if (err) 542 return err; 543 544 if (!f2fs_is_compress_backend_ready(inode)) 545 return -EOPNOTSUPP; 546 547 err = fsverity_file_open(inode, filp); 548 if (err) 549 return err; 550 551 filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC; 552 filp->f_mode |= FMODE_CAN_ODIRECT; 553 554 return dquot_file_open(inode, filp); 555 } 556 557 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 558 { 559 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 560 struct f2fs_node *raw_node; 561 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 562 __le32 *addr; 563 int base = 0; 564 bool compressed_cluster = false; 565 int cluster_index = 0, valid_blocks = 0; 566 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 567 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 568 569 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode)) 570 base = get_extra_isize(dn->inode); 571 572 raw_node = F2FS_NODE(dn->node_page); 573 addr = blkaddr_in_node(raw_node) + base + ofs; 574 575 /* Assumption: truncation starts with cluster */ 576 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 577 block_t blkaddr = le32_to_cpu(*addr); 578 579 if (f2fs_compressed_file(dn->inode) && 580 !(cluster_index & (cluster_size - 1))) { 581 if (compressed_cluster) 582 f2fs_i_compr_blocks_update(dn->inode, 583 valid_blocks, false); 584 compressed_cluster = (blkaddr == COMPRESS_ADDR); 585 valid_blocks = 0; 586 } 587 588 if (blkaddr == NULL_ADDR) 589 continue; 590 591 dn->data_blkaddr = NULL_ADDR; 592 f2fs_set_data_blkaddr(dn); 593 594 if (__is_valid_data_blkaddr(blkaddr)) { 595 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 596 DATA_GENERIC_ENHANCE)) 597 continue; 598 if (compressed_cluster) 599 valid_blocks++; 600 } 601 602 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 603 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 604 605 f2fs_invalidate_blocks(sbi, blkaddr); 606 607 if (!released || blkaddr != COMPRESS_ADDR) 608 nr_free++; 609 } 610 611 if (compressed_cluster) 612 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 613 614 if (nr_free) { 615 pgoff_t fofs; 616 /* 617 * once we invalidate valid blkaddr in range [ofs, ofs + count], 618 * we will invalidate all blkaddr in the whole range. 619 */ 620 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 621 dn->inode) + ofs; 622 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 623 f2fs_update_age_extent_cache_range(dn, fofs, len); 624 dec_valid_block_count(sbi, dn->inode, nr_free); 625 } 626 dn->ofs_in_node = ofs; 627 628 f2fs_update_time(sbi, REQ_TIME); 629 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 630 dn->ofs_in_node, nr_free); 631 } 632 633 static int truncate_partial_data_page(struct inode *inode, u64 from, 634 bool cache_only) 635 { 636 loff_t offset = from & (PAGE_SIZE - 1); 637 pgoff_t index = from >> PAGE_SHIFT; 638 struct address_space *mapping = inode->i_mapping; 639 struct page *page; 640 641 if (!offset && !cache_only) 642 return 0; 643 644 if (cache_only) { 645 page = find_lock_page(mapping, index); 646 if (page && PageUptodate(page)) 647 goto truncate_out; 648 f2fs_put_page(page, 1); 649 return 0; 650 } 651 652 page = f2fs_get_lock_data_page(inode, index, true); 653 if (IS_ERR(page)) 654 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 655 truncate_out: 656 f2fs_wait_on_page_writeback(page, DATA, true, true); 657 zero_user(page, offset, PAGE_SIZE - offset); 658 659 /* An encrypted inode should have a key and truncate the last page. */ 660 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 661 if (!cache_only) 662 set_page_dirty(page); 663 f2fs_put_page(page, 1); 664 return 0; 665 } 666 667 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 668 { 669 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 670 struct dnode_of_data dn; 671 pgoff_t free_from; 672 int count = 0, err = 0; 673 struct page *ipage; 674 bool truncate_page = false; 675 676 trace_f2fs_truncate_blocks_enter(inode, from); 677 678 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 679 680 if (free_from >= max_file_blocks(inode)) 681 goto free_partial; 682 683 if (lock) 684 f2fs_lock_op(sbi); 685 686 ipage = f2fs_get_node_page(sbi, inode->i_ino); 687 if (IS_ERR(ipage)) { 688 err = PTR_ERR(ipage); 689 goto out; 690 } 691 692 if (f2fs_has_inline_data(inode)) { 693 f2fs_truncate_inline_inode(inode, ipage, from); 694 f2fs_put_page(ipage, 1); 695 truncate_page = true; 696 goto out; 697 } 698 699 set_new_dnode(&dn, inode, ipage, NULL, 0); 700 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 701 if (err) { 702 if (err == -ENOENT) 703 goto free_next; 704 goto out; 705 } 706 707 count = ADDRS_PER_PAGE(dn.node_page, inode); 708 709 count -= dn.ofs_in_node; 710 f2fs_bug_on(sbi, count < 0); 711 712 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 713 f2fs_truncate_data_blocks_range(&dn, count); 714 free_from += count; 715 } 716 717 f2fs_put_dnode(&dn); 718 free_next: 719 err = f2fs_truncate_inode_blocks(inode, free_from); 720 out: 721 if (lock) 722 f2fs_unlock_op(sbi); 723 free_partial: 724 /* lastly zero out the first data page */ 725 if (!err) 726 err = truncate_partial_data_page(inode, from, truncate_page); 727 728 trace_f2fs_truncate_blocks_exit(inode, err); 729 return err; 730 } 731 732 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 733 { 734 u64 free_from = from; 735 int err; 736 737 #ifdef CONFIG_F2FS_FS_COMPRESSION 738 /* 739 * for compressed file, only support cluster size 740 * aligned truncation. 741 */ 742 if (f2fs_compressed_file(inode)) 743 free_from = round_up(from, 744 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 745 #endif 746 747 err = f2fs_do_truncate_blocks(inode, free_from, lock); 748 if (err) 749 return err; 750 751 #ifdef CONFIG_F2FS_FS_COMPRESSION 752 /* 753 * For compressed file, after release compress blocks, don't allow write 754 * direct, but we should allow write direct after truncate to zero. 755 */ 756 if (f2fs_compressed_file(inode) && !free_from 757 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 758 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 759 760 if (from != free_from) { 761 err = f2fs_truncate_partial_cluster(inode, from, lock); 762 if (err) 763 return err; 764 } 765 #endif 766 767 return 0; 768 } 769 770 int f2fs_truncate(struct inode *inode) 771 { 772 int err; 773 774 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 775 return -EIO; 776 777 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 778 S_ISLNK(inode->i_mode))) 779 return 0; 780 781 trace_f2fs_truncate(inode); 782 783 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 784 return -EIO; 785 786 err = f2fs_dquot_initialize(inode); 787 if (err) 788 return err; 789 790 /* we should check inline_data size */ 791 if (!f2fs_may_inline_data(inode)) { 792 err = f2fs_convert_inline_inode(inode); 793 if (err) 794 return err; 795 } 796 797 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 798 if (err) 799 return err; 800 801 inode->i_mtime = inode_set_ctime_current(inode); 802 f2fs_mark_inode_dirty_sync(inode, false); 803 return 0; 804 } 805 806 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 807 { 808 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 809 810 if (!fscrypt_dio_supported(inode)) 811 return true; 812 if (fsverity_active(inode)) 813 return true; 814 if (f2fs_compressed_file(inode)) 815 return true; 816 817 /* disallow direct IO if any of devices has unaligned blksize */ 818 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 819 return true; 820 /* 821 * for blkzoned device, fallback direct IO to buffered IO, so 822 * all IOs can be serialized by log-structured write. 823 */ 824 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE)) 825 return true; 826 if (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi)) 827 return true; 828 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 829 return true; 830 831 return false; 832 } 833 834 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 835 struct kstat *stat, u32 request_mask, unsigned int query_flags) 836 { 837 struct inode *inode = d_inode(path->dentry); 838 struct f2fs_inode_info *fi = F2FS_I(inode); 839 struct f2fs_inode *ri = NULL; 840 unsigned int flags; 841 842 if (f2fs_has_extra_attr(inode) && 843 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 844 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 845 stat->result_mask |= STATX_BTIME; 846 stat->btime.tv_sec = fi->i_crtime.tv_sec; 847 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 848 } 849 850 /* 851 * Return the DIO alignment restrictions if requested. We only return 852 * this information when requested, since on encrypted files it might 853 * take a fair bit of work to get if the file wasn't opened recently. 854 * 855 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 856 * cannot represent that, so in that case we report no DIO support. 857 */ 858 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 859 unsigned int bsize = i_blocksize(inode); 860 861 stat->result_mask |= STATX_DIOALIGN; 862 if (!f2fs_force_buffered_io(inode, WRITE)) { 863 stat->dio_mem_align = bsize; 864 stat->dio_offset_align = bsize; 865 } 866 } 867 868 flags = fi->i_flags; 869 if (flags & F2FS_COMPR_FL) 870 stat->attributes |= STATX_ATTR_COMPRESSED; 871 if (flags & F2FS_APPEND_FL) 872 stat->attributes |= STATX_ATTR_APPEND; 873 if (IS_ENCRYPTED(inode)) 874 stat->attributes |= STATX_ATTR_ENCRYPTED; 875 if (flags & F2FS_IMMUTABLE_FL) 876 stat->attributes |= STATX_ATTR_IMMUTABLE; 877 if (flags & F2FS_NODUMP_FL) 878 stat->attributes |= STATX_ATTR_NODUMP; 879 if (IS_VERITY(inode)) 880 stat->attributes |= STATX_ATTR_VERITY; 881 882 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 883 STATX_ATTR_APPEND | 884 STATX_ATTR_ENCRYPTED | 885 STATX_ATTR_IMMUTABLE | 886 STATX_ATTR_NODUMP | 887 STATX_ATTR_VERITY); 888 889 generic_fillattr(idmap, request_mask, inode, stat); 890 891 /* we need to show initial sectors used for inline_data/dentries */ 892 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 893 f2fs_has_inline_dentry(inode)) 894 stat->blocks += (stat->size + 511) >> 9; 895 896 return 0; 897 } 898 899 #ifdef CONFIG_F2FS_FS_POSIX_ACL 900 static void __setattr_copy(struct mnt_idmap *idmap, 901 struct inode *inode, const struct iattr *attr) 902 { 903 unsigned int ia_valid = attr->ia_valid; 904 905 i_uid_update(idmap, attr, inode); 906 i_gid_update(idmap, attr, inode); 907 if (ia_valid & ATTR_ATIME) 908 inode->i_atime = attr->ia_atime; 909 if (ia_valid & ATTR_MTIME) 910 inode->i_mtime = attr->ia_mtime; 911 if (ia_valid & ATTR_CTIME) 912 inode_set_ctime_to_ts(inode, attr->ia_ctime); 913 if (ia_valid & ATTR_MODE) { 914 umode_t mode = attr->ia_mode; 915 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode); 916 917 if (!vfsgid_in_group_p(vfsgid) && 918 !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) 919 mode &= ~S_ISGID; 920 set_acl_inode(inode, mode); 921 } 922 } 923 #else 924 #define __setattr_copy setattr_copy 925 #endif 926 927 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 928 struct iattr *attr) 929 { 930 struct inode *inode = d_inode(dentry); 931 int err; 932 933 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 934 return -EIO; 935 936 if (unlikely(IS_IMMUTABLE(inode))) 937 return -EPERM; 938 939 if (unlikely(IS_APPEND(inode) && 940 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 941 ATTR_GID | ATTR_TIMES_SET)))) 942 return -EPERM; 943 944 if ((attr->ia_valid & ATTR_SIZE) && 945 !f2fs_is_compress_backend_ready(inode)) 946 return -EOPNOTSUPP; 947 948 err = setattr_prepare(idmap, dentry, attr); 949 if (err) 950 return err; 951 952 err = fscrypt_prepare_setattr(dentry, attr); 953 if (err) 954 return err; 955 956 err = fsverity_prepare_setattr(dentry, attr); 957 if (err) 958 return err; 959 960 if (is_quota_modification(idmap, inode, attr)) { 961 err = f2fs_dquot_initialize(inode); 962 if (err) 963 return err; 964 } 965 if (i_uid_needs_update(idmap, attr, inode) || 966 i_gid_needs_update(idmap, attr, inode)) { 967 f2fs_lock_op(F2FS_I_SB(inode)); 968 err = dquot_transfer(idmap, inode, attr); 969 if (err) { 970 set_sbi_flag(F2FS_I_SB(inode), 971 SBI_QUOTA_NEED_REPAIR); 972 f2fs_unlock_op(F2FS_I_SB(inode)); 973 return err; 974 } 975 /* 976 * update uid/gid under lock_op(), so that dquot and inode can 977 * be updated atomically. 978 */ 979 i_uid_update(idmap, attr, inode); 980 i_gid_update(idmap, attr, inode); 981 f2fs_mark_inode_dirty_sync(inode, true); 982 f2fs_unlock_op(F2FS_I_SB(inode)); 983 } 984 985 if (attr->ia_valid & ATTR_SIZE) { 986 loff_t old_size = i_size_read(inode); 987 988 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 989 /* 990 * should convert inline inode before i_size_write to 991 * keep smaller than inline_data size with inline flag. 992 */ 993 err = f2fs_convert_inline_inode(inode); 994 if (err) 995 return err; 996 } 997 998 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 999 filemap_invalidate_lock(inode->i_mapping); 1000 1001 truncate_setsize(inode, attr->ia_size); 1002 1003 if (attr->ia_size <= old_size) 1004 err = f2fs_truncate(inode); 1005 /* 1006 * do not trim all blocks after i_size if target size is 1007 * larger than i_size. 1008 */ 1009 filemap_invalidate_unlock(inode->i_mapping); 1010 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1011 if (err) 1012 return err; 1013 1014 spin_lock(&F2FS_I(inode)->i_size_lock); 1015 inode->i_mtime = inode_set_ctime_current(inode); 1016 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1017 spin_unlock(&F2FS_I(inode)->i_size_lock); 1018 } 1019 1020 __setattr_copy(idmap, inode, attr); 1021 1022 if (attr->ia_valid & ATTR_MODE) { 1023 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1024 1025 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1026 if (!err) 1027 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1028 clear_inode_flag(inode, FI_ACL_MODE); 1029 } 1030 } 1031 1032 /* file size may changed here */ 1033 f2fs_mark_inode_dirty_sync(inode, true); 1034 1035 /* inode change will produce dirty node pages flushed by checkpoint */ 1036 f2fs_balance_fs(F2FS_I_SB(inode), true); 1037 1038 return err; 1039 } 1040 1041 const struct inode_operations f2fs_file_inode_operations = { 1042 .getattr = f2fs_getattr, 1043 .setattr = f2fs_setattr, 1044 .get_inode_acl = f2fs_get_acl, 1045 .set_acl = f2fs_set_acl, 1046 .listxattr = f2fs_listxattr, 1047 .fiemap = f2fs_fiemap, 1048 .fileattr_get = f2fs_fileattr_get, 1049 .fileattr_set = f2fs_fileattr_set, 1050 }; 1051 1052 static int fill_zero(struct inode *inode, pgoff_t index, 1053 loff_t start, loff_t len) 1054 { 1055 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1056 struct page *page; 1057 1058 if (!len) 1059 return 0; 1060 1061 f2fs_balance_fs(sbi, true); 1062 1063 f2fs_lock_op(sbi); 1064 page = f2fs_get_new_data_page(inode, NULL, index, false); 1065 f2fs_unlock_op(sbi); 1066 1067 if (IS_ERR(page)) 1068 return PTR_ERR(page); 1069 1070 f2fs_wait_on_page_writeback(page, DATA, true, true); 1071 zero_user(page, start, len); 1072 set_page_dirty(page); 1073 f2fs_put_page(page, 1); 1074 return 0; 1075 } 1076 1077 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1078 { 1079 int err; 1080 1081 while (pg_start < pg_end) { 1082 struct dnode_of_data dn; 1083 pgoff_t end_offset, count; 1084 1085 set_new_dnode(&dn, inode, NULL, NULL, 0); 1086 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1087 if (err) { 1088 if (err == -ENOENT) { 1089 pg_start = f2fs_get_next_page_offset(&dn, 1090 pg_start); 1091 continue; 1092 } 1093 return err; 1094 } 1095 1096 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1097 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1098 1099 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1100 1101 f2fs_truncate_data_blocks_range(&dn, count); 1102 f2fs_put_dnode(&dn); 1103 1104 pg_start += count; 1105 } 1106 return 0; 1107 } 1108 1109 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1110 { 1111 pgoff_t pg_start, pg_end; 1112 loff_t off_start, off_end; 1113 int ret; 1114 1115 ret = f2fs_convert_inline_inode(inode); 1116 if (ret) 1117 return ret; 1118 1119 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1120 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1121 1122 off_start = offset & (PAGE_SIZE - 1); 1123 off_end = (offset + len) & (PAGE_SIZE - 1); 1124 1125 if (pg_start == pg_end) { 1126 ret = fill_zero(inode, pg_start, off_start, 1127 off_end - off_start); 1128 if (ret) 1129 return ret; 1130 } else { 1131 if (off_start) { 1132 ret = fill_zero(inode, pg_start++, off_start, 1133 PAGE_SIZE - off_start); 1134 if (ret) 1135 return ret; 1136 } 1137 if (off_end) { 1138 ret = fill_zero(inode, pg_end, 0, off_end); 1139 if (ret) 1140 return ret; 1141 } 1142 1143 if (pg_start < pg_end) { 1144 loff_t blk_start, blk_end; 1145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1146 1147 f2fs_balance_fs(sbi, true); 1148 1149 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1150 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1151 1152 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1153 filemap_invalidate_lock(inode->i_mapping); 1154 1155 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1156 1157 f2fs_lock_op(sbi); 1158 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1159 f2fs_unlock_op(sbi); 1160 1161 filemap_invalidate_unlock(inode->i_mapping); 1162 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1163 } 1164 } 1165 1166 return ret; 1167 } 1168 1169 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1170 int *do_replace, pgoff_t off, pgoff_t len) 1171 { 1172 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1173 struct dnode_of_data dn; 1174 int ret, done, i; 1175 1176 next_dnode: 1177 set_new_dnode(&dn, inode, NULL, NULL, 0); 1178 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1179 if (ret && ret != -ENOENT) { 1180 return ret; 1181 } else if (ret == -ENOENT) { 1182 if (dn.max_level == 0) 1183 return -ENOENT; 1184 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1185 dn.ofs_in_node, len); 1186 blkaddr += done; 1187 do_replace += done; 1188 goto next; 1189 } 1190 1191 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1192 dn.ofs_in_node, len); 1193 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1194 *blkaddr = f2fs_data_blkaddr(&dn); 1195 1196 if (__is_valid_data_blkaddr(*blkaddr) && 1197 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1198 DATA_GENERIC_ENHANCE)) { 1199 f2fs_put_dnode(&dn); 1200 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1201 return -EFSCORRUPTED; 1202 } 1203 1204 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1205 1206 if (f2fs_lfs_mode(sbi)) { 1207 f2fs_put_dnode(&dn); 1208 return -EOPNOTSUPP; 1209 } 1210 1211 /* do not invalidate this block address */ 1212 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1213 *do_replace = 1; 1214 } 1215 } 1216 f2fs_put_dnode(&dn); 1217 next: 1218 len -= done; 1219 off += done; 1220 if (len) 1221 goto next_dnode; 1222 return 0; 1223 } 1224 1225 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1226 int *do_replace, pgoff_t off, int len) 1227 { 1228 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1229 struct dnode_of_data dn; 1230 int ret, i; 1231 1232 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1233 if (*do_replace == 0) 1234 continue; 1235 1236 set_new_dnode(&dn, inode, NULL, NULL, 0); 1237 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1238 if (ret) { 1239 dec_valid_block_count(sbi, inode, 1); 1240 f2fs_invalidate_blocks(sbi, *blkaddr); 1241 } else { 1242 f2fs_update_data_blkaddr(&dn, *blkaddr); 1243 } 1244 f2fs_put_dnode(&dn); 1245 } 1246 return 0; 1247 } 1248 1249 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1250 block_t *blkaddr, int *do_replace, 1251 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1252 { 1253 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1254 pgoff_t i = 0; 1255 int ret; 1256 1257 while (i < len) { 1258 if (blkaddr[i] == NULL_ADDR && !full) { 1259 i++; 1260 continue; 1261 } 1262 1263 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1264 struct dnode_of_data dn; 1265 struct node_info ni; 1266 size_t new_size; 1267 pgoff_t ilen; 1268 1269 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1270 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1271 if (ret) 1272 return ret; 1273 1274 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1275 if (ret) { 1276 f2fs_put_dnode(&dn); 1277 return ret; 1278 } 1279 1280 ilen = min((pgoff_t) 1281 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1282 dn.ofs_in_node, len - i); 1283 do { 1284 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1285 f2fs_truncate_data_blocks_range(&dn, 1); 1286 1287 if (do_replace[i]) { 1288 f2fs_i_blocks_write(src_inode, 1289 1, false, false); 1290 f2fs_i_blocks_write(dst_inode, 1291 1, true, false); 1292 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1293 blkaddr[i], ni.version, true, false); 1294 1295 do_replace[i] = 0; 1296 } 1297 dn.ofs_in_node++; 1298 i++; 1299 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1300 if (dst_inode->i_size < new_size) 1301 f2fs_i_size_write(dst_inode, new_size); 1302 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1303 1304 f2fs_put_dnode(&dn); 1305 } else { 1306 struct page *psrc, *pdst; 1307 1308 psrc = f2fs_get_lock_data_page(src_inode, 1309 src + i, true); 1310 if (IS_ERR(psrc)) 1311 return PTR_ERR(psrc); 1312 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1313 true); 1314 if (IS_ERR(pdst)) { 1315 f2fs_put_page(psrc, 1); 1316 return PTR_ERR(pdst); 1317 } 1318 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1319 set_page_dirty(pdst); 1320 set_page_private_gcing(pdst); 1321 f2fs_put_page(pdst, 1); 1322 f2fs_put_page(psrc, 1); 1323 1324 ret = f2fs_truncate_hole(src_inode, 1325 src + i, src + i + 1); 1326 if (ret) 1327 return ret; 1328 i++; 1329 } 1330 } 1331 return 0; 1332 } 1333 1334 static int __exchange_data_block(struct inode *src_inode, 1335 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1336 pgoff_t len, bool full) 1337 { 1338 block_t *src_blkaddr; 1339 int *do_replace; 1340 pgoff_t olen; 1341 int ret; 1342 1343 while (len) { 1344 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1345 1346 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1347 array_size(olen, sizeof(block_t)), 1348 GFP_NOFS); 1349 if (!src_blkaddr) 1350 return -ENOMEM; 1351 1352 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1353 array_size(olen, sizeof(int)), 1354 GFP_NOFS); 1355 if (!do_replace) { 1356 kvfree(src_blkaddr); 1357 return -ENOMEM; 1358 } 1359 1360 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1361 do_replace, src, olen); 1362 if (ret) 1363 goto roll_back; 1364 1365 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1366 do_replace, src, dst, olen, full); 1367 if (ret) 1368 goto roll_back; 1369 1370 src += olen; 1371 dst += olen; 1372 len -= olen; 1373 1374 kvfree(src_blkaddr); 1375 kvfree(do_replace); 1376 } 1377 return 0; 1378 1379 roll_back: 1380 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1381 kvfree(src_blkaddr); 1382 kvfree(do_replace); 1383 return ret; 1384 } 1385 1386 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1387 { 1388 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1389 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1390 pgoff_t start = offset >> PAGE_SHIFT; 1391 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1392 int ret; 1393 1394 f2fs_balance_fs(sbi, true); 1395 1396 /* avoid gc operation during block exchange */ 1397 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1398 filemap_invalidate_lock(inode->i_mapping); 1399 1400 f2fs_lock_op(sbi); 1401 f2fs_drop_extent_tree(inode); 1402 truncate_pagecache(inode, offset); 1403 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1404 f2fs_unlock_op(sbi); 1405 1406 filemap_invalidate_unlock(inode->i_mapping); 1407 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1408 return ret; 1409 } 1410 1411 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1412 { 1413 loff_t new_size; 1414 int ret; 1415 1416 if (offset + len >= i_size_read(inode)) 1417 return -EINVAL; 1418 1419 /* collapse range should be aligned to block size of f2fs. */ 1420 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1421 return -EINVAL; 1422 1423 ret = f2fs_convert_inline_inode(inode); 1424 if (ret) 1425 return ret; 1426 1427 /* write out all dirty pages from offset */ 1428 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1429 if (ret) 1430 return ret; 1431 1432 ret = f2fs_do_collapse(inode, offset, len); 1433 if (ret) 1434 return ret; 1435 1436 /* write out all moved pages, if possible */ 1437 filemap_invalidate_lock(inode->i_mapping); 1438 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1439 truncate_pagecache(inode, offset); 1440 1441 new_size = i_size_read(inode) - len; 1442 ret = f2fs_truncate_blocks(inode, new_size, true); 1443 filemap_invalidate_unlock(inode->i_mapping); 1444 if (!ret) 1445 f2fs_i_size_write(inode, new_size); 1446 return ret; 1447 } 1448 1449 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1450 pgoff_t end) 1451 { 1452 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1453 pgoff_t index = start; 1454 unsigned int ofs_in_node = dn->ofs_in_node; 1455 blkcnt_t count = 0; 1456 int ret; 1457 1458 for (; index < end; index++, dn->ofs_in_node++) { 1459 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1460 count++; 1461 } 1462 1463 dn->ofs_in_node = ofs_in_node; 1464 ret = f2fs_reserve_new_blocks(dn, count); 1465 if (ret) 1466 return ret; 1467 1468 dn->ofs_in_node = ofs_in_node; 1469 for (index = start; index < end; index++, dn->ofs_in_node++) { 1470 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1471 /* 1472 * f2fs_reserve_new_blocks will not guarantee entire block 1473 * allocation. 1474 */ 1475 if (dn->data_blkaddr == NULL_ADDR) { 1476 ret = -ENOSPC; 1477 break; 1478 } 1479 1480 if (dn->data_blkaddr == NEW_ADDR) 1481 continue; 1482 1483 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1484 DATA_GENERIC_ENHANCE)) { 1485 ret = -EFSCORRUPTED; 1486 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1487 break; 1488 } 1489 1490 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1491 dn->data_blkaddr = NEW_ADDR; 1492 f2fs_set_data_blkaddr(dn); 1493 } 1494 1495 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1496 f2fs_update_age_extent_cache_range(dn, start, index - start); 1497 1498 return ret; 1499 } 1500 1501 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1502 int mode) 1503 { 1504 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1505 struct address_space *mapping = inode->i_mapping; 1506 pgoff_t index, pg_start, pg_end; 1507 loff_t new_size = i_size_read(inode); 1508 loff_t off_start, off_end; 1509 int ret = 0; 1510 1511 ret = inode_newsize_ok(inode, (len + offset)); 1512 if (ret) 1513 return ret; 1514 1515 ret = f2fs_convert_inline_inode(inode); 1516 if (ret) 1517 return ret; 1518 1519 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1520 if (ret) 1521 return ret; 1522 1523 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1524 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1525 1526 off_start = offset & (PAGE_SIZE - 1); 1527 off_end = (offset + len) & (PAGE_SIZE - 1); 1528 1529 if (pg_start == pg_end) { 1530 ret = fill_zero(inode, pg_start, off_start, 1531 off_end - off_start); 1532 if (ret) 1533 return ret; 1534 1535 new_size = max_t(loff_t, new_size, offset + len); 1536 } else { 1537 if (off_start) { 1538 ret = fill_zero(inode, pg_start++, off_start, 1539 PAGE_SIZE - off_start); 1540 if (ret) 1541 return ret; 1542 1543 new_size = max_t(loff_t, new_size, 1544 (loff_t)pg_start << PAGE_SHIFT); 1545 } 1546 1547 for (index = pg_start; index < pg_end;) { 1548 struct dnode_of_data dn; 1549 unsigned int end_offset; 1550 pgoff_t end; 1551 1552 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1553 filemap_invalidate_lock(mapping); 1554 1555 truncate_pagecache_range(inode, 1556 (loff_t)index << PAGE_SHIFT, 1557 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1558 1559 f2fs_lock_op(sbi); 1560 1561 set_new_dnode(&dn, inode, NULL, NULL, 0); 1562 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1563 if (ret) { 1564 f2fs_unlock_op(sbi); 1565 filemap_invalidate_unlock(mapping); 1566 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1567 goto out; 1568 } 1569 1570 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1571 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1572 1573 ret = f2fs_do_zero_range(&dn, index, end); 1574 f2fs_put_dnode(&dn); 1575 1576 f2fs_unlock_op(sbi); 1577 filemap_invalidate_unlock(mapping); 1578 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1579 1580 f2fs_balance_fs(sbi, dn.node_changed); 1581 1582 if (ret) 1583 goto out; 1584 1585 index = end; 1586 new_size = max_t(loff_t, new_size, 1587 (loff_t)index << PAGE_SHIFT); 1588 } 1589 1590 if (off_end) { 1591 ret = fill_zero(inode, pg_end, 0, off_end); 1592 if (ret) 1593 goto out; 1594 1595 new_size = max_t(loff_t, new_size, offset + len); 1596 } 1597 } 1598 1599 out: 1600 if (new_size > i_size_read(inode)) { 1601 if (mode & FALLOC_FL_KEEP_SIZE) 1602 file_set_keep_isize(inode); 1603 else 1604 f2fs_i_size_write(inode, new_size); 1605 } 1606 return ret; 1607 } 1608 1609 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1610 { 1611 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1612 struct address_space *mapping = inode->i_mapping; 1613 pgoff_t nr, pg_start, pg_end, delta, idx; 1614 loff_t new_size; 1615 int ret = 0; 1616 1617 new_size = i_size_read(inode) + len; 1618 ret = inode_newsize_ok(inode, new_size); 1619 if (ret) 1620 return ret; 1621 1622 if (offset >= i_size_read(inode)) 1623 return -EINVAL; 1624 1625 /* insert range should be aligned to block size of f2fs. */ 1626 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1627 return -EINVAL; 1628 1629 ret = f2fs_convert_inline_inode(inode); 1630 if (ret) 1631 return ret; 1632 1633 f2fs_balance_fs(sbi, true); 1634 1635 filemap_invalidate_lock(mapping); 1636 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1637 filemap_invalidate_unlock(mapping); 1638 if (ret) 1639 return ret; 1640 1641 /* write out all dirty pages from offset */ 1642 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1643 if (ret) 1644 return ret; 1645 1646 pg_start = offset >> PAGE_SHIFT; 1647 pg_end = (offset + len) >> PAGE_SHIFT; 1648 delta = pg_end - pg_start; 1649 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1650 1651 /* avoid gc operation during block exchange */ 1652 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1653 filemap_invalidate_lock(mapping); 1654 truncate_pagecache(inode, offset); 1655 1656 while (!ret && idx > pg_start) { 1657 nr = idx - pg_start; 1658 if (nr > delta) 1659 nr = delta; 1660 idx -= nr; 1661 1662 f2fs_lock_op(sbi); 1663 f2fs_drop_extent_tree(inode); 1664 1665 ret = __exchange_data_block(inode, inode, idx, 1666 idx + delta, nr, false); 1667 f2fs_unlock_op(sbi); 1668 } 1669 filemap_invalidate_unlock(mapping); 1670 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1671 1672 /* write out all moved pages, if possible */ 1673 filemap_invalidate_lock(mapping); 1674 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1675 truncate_pagecache(inode, offset); 1676 filemap_invalidate_unlock(mapping); 1677 1678 if (!ret) 1679 f2fs_i_size_write(inode, new_size); 1680 return ret; 1681 } 1682 1683 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1684 loff_t len, int mode) 1685 { 1686 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1687 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1688 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1689 .m_may_create = true }; 1690 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1691 .init_gc_type = FG_GC, 1692 .should_migrate_blocks = false, 1693 .err_gc_skipped = true, 1694 .nr_free_secs = 0 }; 1695 pgoff_t pg_start, pg_end; 1696 loff_t new_size; 1697 loff_t off_end; 1698 block_t expanded = 0; 1699 int err; 1700 1701 err = inode_newsize_ok(inode, (len + offset)); 1702 if (err) 1703 return err; 1704 1705 err = f2fs_convert_inline_inode(inode); 1706 if (err) 1707 return err; 1708 1709 f2fs_balance_fs(sbi, true); 1710 1711 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1712 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1713 off_end = (offset + len) & (PAGE_SIZE - 1); 1714 1715 map.m_lblk = pg_start; 1716 map.m_len = pg_end - pg_start; 1717 if (off_end) 1718 map.m_len++; 1719 1720 if (!map.m_len) 1721 return 0; 1722 1723 if (f2fs_is_pinned_file(inode)) { 1724 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1725 block_t sec_len = roundup(map.m_len, sec_blks); 1726 1727 map.m_len = sec_blks; 1728 next_alloc: 1729 if (has_not_enough_free_secs(sbi, 0, 1730 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1731 f2fs_down_write(&sbi->gc_lock); 1732 stat_inc_gc_call_count(sbi, FOREGROUND); 1733 err = f2fs_gc(sbi, &gc_control); 1734 if (err && err != -ENODATA) 1735 goto out_err; 1736 } 1737 1738 f2fs_down_write(&sbi->pin_sem); 1739 1740 f2fs_lock_op(sbi); 1741 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false); 1742 f2fs_unlock_op(sbi); 1743 1744 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1745 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1746 file_dont_truncate(inode); 1747 1748 f2fs_up_write(&sbi->pin_sem); 1749 1750 expanded += map.m_len; 1751 sec_len -= map.m_len; 1752 map.m_lblk += map.m_len; 1753 if (!err && sec_len) 1754 goto next_alloc; 1755 1756 map.m_len = expanded; 1757 } else { 1758 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1759 expanded = map.m_len; 1760 } 1761 out_err: 1762 if (err) { 1763 pgoff_t last_off; 1764 1765 if (!expanded) 1766 return err; 1767 1768 last_off = pg_start + expanded - 1; 1769 1770 /* update new size to the failed position */ 1771 new_size = (last_off == pg_end) ? offset + len : 1772 (loff_t)(last_off + 1) << PAGE_SHIFT; 1773 } else { 1774 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1775 } 1776 1777 if (new_size > i_size_read(inode)) { 1778 if (mode & FALLOC_FL_KEEP_SIZE) 1779 file_set_keep_isize(inode); 1780 else 1781 f2fs_i_size_write(inode, new_size); 1782 } 1783 1784 return err; 1785 } 1786 1787 static long f2fs_fallocate(struct file *file, int mode, 1788 loff_t offset, loff_t len) 1789 { 1790 struct inode *inode = file_inode(file); 1791 long ret = 0; 1792 1793 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1794 return -EIO; 1795 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1796 return -ENOSPC; 1797 if (!f2fs_is_compress_backend_ready(inode)) 1798 return -EOPNOTSUPP; 1799 1800 /* f2fs only support ->fallocate for regular file */ 1801 if (!S_ISREG(inode->i_mode)) 1802 return -EINVAL; 1803 1804 if (IS_ENCRYPTED(inode) && 1805 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1806 return -EOPNOTSUPP; 1807 1808 /* 1809 * Pinned file should not support partial truncation since the block 1810 * can be used by applications. 1811 */ 1812 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1813 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1814 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1815 return -EOPNOTSUPP; 1816 1817 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1818 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1819 FALLOC_FL_INSERT_RANGE)) 1820 return -EOPNOTSUPP; 1821 1822 inode_lock(inode); 1823 1824 ret = file_modified(file); 1825 if (ret) 1826 goto out; 1827 1828 if (mode & FALLOC_FL_PUNCH_HOLE) { 1829 if (offset >= inode->i_size) 1830 goto out; 1831 1832 ret = f2fs_punch_hole(inode, offset, len); 1833 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1834 ret = f2fs_collapse_range(inode, offset, len); 1835 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1836 ret = f2fs_zero_range(inode, offset, len, mode); 1837 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1838 ret = f2fs_insert_range(inode, offset, len); 1839 } else { 1840 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1841 } 1842 1843 if (!ret) { 1844 inode->i_mtime = inode_set_ctime_current(inode); 1845 f2fs_mark_inode_dirty_sync(inode, false); 1846 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1847 } 1848 1849 out: 1850 inode_unlock(inode); 1851 1852 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1853 return ret; 1854 } 1855 1856 static int f2fs_release_file(struct inode *inode, struct file *filp) 1857 { 1858 /* 1859 * f2fs_release_file is called at every close calls. So we should 1860 * not drop any inmemory pages by close called by other process. 1861 */ 1862 if (!(filp->f_mode & FMODE_WRITE) || 1863 atomic_read(&inode->i_writecount) != 1) 1864 return 0; 1865 1866 inode_lock(inode); 1867 f2fs_abort_atomic_write(inode, true); 1868 inode_unlock(inode); 1869 1870 return 0; 1871 } 1872 1873 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1874 { 1875 struct inode *inode = file_inode(file); 1876 1877 /* 1878 * If the process doing a transaction is crashed, we should do 1879 * roll-back. Otherwise, other reader/write can see corrupted database 1880 * until all the writers close its file. Since this should be done 1881 * before dropping file lock, it needs to do in ->flush. 1882 */ 1883 if (F2FS_I(inode)->atomic_write_task == current && 1884 (current->flags & PF_EXITING)) { 1885 inode_lock(inode); 1886 f2fs_abort_atomic_write(inode, true); 1887 inode_unlock(inode); 1888 } 1889 1890 return 0; 1891 } 1892 1893 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1894 { 1895 struct f2fs_inode_info *fi = F2FS_I(inode); 1896 u32 masked_flags = fi->i_flags & mask; 1897 1898 /* mask can be shrunk by flags_valid selector */ 1899 iflags &= mask; 1900 1901 /* Is it quota file? Do not allow user to mess with it */ 1902 if (IS_NOQUOTA(inode)) 1903 return -EPERM; 1904 1905 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1906 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1907 return -EOPNOTSUPP; 1908 if (!f2fs_empty_dir(inode)) 1909 return -ENOTEMPTY; 1910 } 1911 1912 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1913 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1914 return -EOPNOTSUPP; 1915 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1916 return -EINVAL; 1917 } 1918 1919 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1920 if (masked_flags & F2FS_COMPR_FL) { 1921 if (!f2fs_disable_compressed_file(inode)) 1922 return -EINVAL; 1923 } else { 1924 /* try to convert inline_data to support compression */ 1925 int err = f2fs_convert_inline_inode(inode); 1926 if (err) 1927 return err; 1928 1929 f2fs_down_write(&F2FS_I(inode)->i_sem); 1930 if (!f2fs_may_compress(inode) || 1931 (S_ISREG(inode->i_mode) && 1932 F2FS_HAS_BLOCKS(inode))) { 1933 f2fs_up_write(&F2FS_I(inode)->i_sem); 1934 return -EINVAL; 1935 } 1936 err = set_compress_context(inode); 1937 f2fs_up_write(&F2FS_I(inode)->i_sem); 1938 1939 if (err) 1940 return err; 1941 } 1942 } 1943 1944 fi->i_flags = iflags | (fi->i_flags & ~mask); 1945 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1946 (fi->i_flags & F2FS_NOCOMP_FL)); 1947 1948 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1949 set_inode_flag(inode, FI_PROJ_INHERIT); 1950 else 1951 clear_inode_flag(inode, FI_PROJ_INHERIT); 1952 1953 inode_set_ctime_current(inode); 1954 f2fs_set_inode_flags(inode); 1955 f2fs_mark_inode_dirty_sync(inode, true); 1956 return 0; 1957 } 1958 1959 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1960 1961 /* 1962 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1963 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1964 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1965 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1966 * 1967 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1968 * FS_IOC_FSSETXATTR is done by the VFS. 1969 */ 1970 1971 static const struct { 1972 u32 iflag; 1973 u32 fsflag; 1974 } f2fs_fsflags_map[] = { 1975 { F2FS_COMPR_FL, FS_COMPR_FL }, 1976 { F2FS_SYNC_FL, FS_SYNC_FL }, 1977 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1978 { F2FS_APPEND_FL, FS_APPEND_FL }, 1979 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1980 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1981 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1982 { F2FS_INDEX_FL, FS_INDEX_FL }, 1983 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1984 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1985 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1986 }; 1987 1988 #define F2FS_GETTABLE_FS_FL ( \ 1989 FS_COMPR_FL | \ 1990 FS_SYNC_FL | \ 1991 FS_IMMUTABLE_FL | \ 1992 FS_APPEND_FL | \ 1993 FS_NODUMP_FL | \ 1994 FS_NOATIME_FL | \ 1995 FS_NOCOMP_FL | \ 1996 FS_INDEX_FL | \ 1997 FS_DIRSYNC_FL | \ 1998 FS_PROJINHERIT_FL | \ 1999 FS_ENCRYPT_FL | \ 2000 FS_INLINE_DATA_FL | \ 2001 FS_NOCOW_FL | \ 2002 FS_VERITY_FL | \ 2003 FS_CASEFOLD_FL) 2004 2005 #define F2FS_SETTABLE_FS_FL ( \ 2006 FS_COMPR_FL | \ 2007 FS_SYNC_FL | \ 2008 FS_IMMUTABLE_FL | \ 2009 FS_APPEND_FL | \ 2010 FS_NODUMP_FL | \ 2011 FS_NOATIME_FL | \ 2012 FS_NOCOMP_FL | \ 2013 FS_DIRSYNC_FL | \ 2014 FS_PROJINHERIT_FL | \ 2015 FS_CASEFOLD_FL) 2016 2017 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2018 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2019 { 2020 u32 fsflags = 0; 2021 int i; 2022 2023 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2024 if (iflags & f2fs_fsflags_map[i].iflag) 2025 fsflags |= f2fs_fsflags_map[i].fsflag; 2026 2027 return fsflags; 2028 } 2029 2030 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2031 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2032 { 2033 u32 iflags = 0; 2034 int i; 2035 2036 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2037 if (fsflags & f2fs_fsflags_map[i].fsflag) 2038 iflags |= f2fs_fsflags_map[i].iflag; 2039 2040 return iflags; 2041 } 2042 2043 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2044 { 2045 struct inode *inode = file_inode(filp); 2046 2047 return put_user(inode->i_generation, (int __user *)arg); 2048 } 2049 2050 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2051 { 2052 struct inode *inode = file_inode(filp); 2053 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2054 struct f2fs_inode_info *fi = F2FS_I(inode); 2055 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2056 struct inode *pinode; 2057 loff_t isize; 2058 int ret; 2059 2060 if (!inode_owner_or_capable(idmap, inode)) 2061 return -EACCES; 2062 2063 if (!S_ISREG(inode->i_mode)) 2064 return -EINVAL; 2065 2066 if (filp->f_flags & O_DIRECT) 2067 return -EINVAL; 2068 2069 ret = mnt_want_write_file(filp); 2070 if (ret) 2071 return ret; 2072 2073 inode_lock(inode); 2074 2075 if (!f2fs_disable_compressed_file(inode)) { 2076 ret = -EINVAL; 2077 goto out; 2078 } 2079 2080 if (f2fs_is_atomic_file(inode)) 2081 goto out; 2082 2083 ret = f2fs_convert_inline_inode(inode); 2084 if (ret) 2085 goto out; 2086 2087 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2088 2089 /* 2090 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2091 * f2fs_is_atomic_file. 2092 */ 2093 if (get_dirty_pages(inode)) 2094 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2095 inode->i_ino, get_dirty_pages(inode)); 2096 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2097 if (ret) { 2098 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2099 goto out; 2100 } 2101 2102 /* Check if the inode already has a COW inode */ 2103 if (fi->cow_inode == NULL) { 2104 /* Create a COW inode for atomic write */ 2105 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2106 if (IS_ERR(pinode)) { 2107 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2108 ret = PTR_ERR(pinode); 2109 goto out; 2110 } 2111 2112 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2113 iput(pinode); 2114 if (ret) { 2115 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2116 goto out; 2117 } 2118 2119 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2120 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2121 } else { 2122 /* Reuse the already created COW inode */ 2123 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2124 if (ret) { 2125 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2126 goto out; 2127 } 2128 } 2129 2130 f2fs_write_inode(inode, NULL); 2131 2132 stat_inc_atomic_inode(inode); 2133 2134 set_inode_flag(inode, FI_ATOMIC_FILE); 2135 2136 isize = i_size_read(inode); 2137 fi->original_i_size = isize; 2138 if (truncate) { 2139 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2140 truncate_inode_pages_final(inode->i_mapping); 2141 f2fs_i_size_write(inode, 0); 2142 isize = 0; 2143 } 2144 f2fs_i_size_write(fi->cow_inode, isize); 2145 2146 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2147 2148 f2fs_update_time(sbi, REQ_TIME); 2149 fi->atomic_write_task = current; 2150 stat_update_max_atomic_write(inode); 2151 fi->atomic_write_cnt = 0; 2152 out: 2153 inode_unlock(inode); 2154 mnt_drop_write_file(filp); 2155 return ret; 2156 } 2157 2158 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2159 { 2160 struct inode *inode = file_inode(filp); 2161 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2162 int ret; 2163 2164 if (!inode_owner_or_capable(idmap, inode)) 2165 return -EACCES; 2166 2167 ret = mnt_want_write_file(filp); 2168 if (ret) 2169 return ret; 2170 2171 f2fs_balance_fs(F2FS_I_SB(inode), true); 2172 2173 inode_lock(inode); 2174 2175 if (f2fs_is_atomic_file(inode)) { 2176 ret = f2fs_commit_atomic_write(inode); 2177 if (!ret) 2178 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2179 2180 f2fs_abort_atomic_write(inode, ret); 2181 } else { 2182 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2183 } 2184 2185 inode_unlock(inode); 2186 mnt_drop_write_file(filp); 2187 return ret; 2188 } 2189 2190 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2191 { 2192 struct inode *inode = file_inode(filp); 2193 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2194 int ret; 2195 2196 if (!inode_owner_or_capable(idmap, inode)) 2197 return -EACCES; 2198 2199 ret = mnt_want_write_file(filp); 2200 if (ret) 2201 return ret; 2202 2203 inode_lock(inode); 2204 2205 f2fs_abort_atomic_write(inode, true); 2206 2207 inode_unlock(inode); 2208 2209 mnt_drop_write_file(filp); 2210 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2211 return ret; 2212 } 2213 2214 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2215 { 2216 struct inode *inode = file_inode(filp); 2217 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2218 struct super_block *sb = sbi->sb; 2219 __u32 in; 2220 int ret = 0; 2221 2222 if (!capable(CAP_SYS_ADMIN)) 2223 return -EPERM; 2224 2225 if (get_user(in, (__u32 __user *)arg)) 2226 return -EFAULT; 2227 2228 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2229 ret = mnt_want_write_file(filp); 2230 if (ret) { 2231 if (ret == -EROFS) { 2232 ret = 0; 2233 f2fs_stop_checkpoint(sbi, false, 2234 STOP_CP_REASON_SHUTDOWN); 2235 trace_f2fs_shutdown(sbi, in, ret); 2236 } 2237 return ret; 2238 } 2239 } 2240 2241 switch (in) { 2242 case F2FS_GOING_DOWN_FULLSYNC: 2243 ret = freeze_bdev(sb->s_bdev); 2244 if (ret) 2245 goto out; 2246 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2247 thaw_bdev(sb->s_bdev); 2248 break; 2249 case F2FS_GOING_DOWN_METASYNC: 2250 /* do checkpoint only */ 2251 ret = f2fs_sync_fs(sb, 1); 2252 if (ret) 2253 goto out; 2254 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2255 break; 2256 case F2FS_GOING_DOWN_NOSYNC: 2257 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2258 break; 2259 case F2FS_GOING_DOWN_METAFLUSH: 2260 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2261 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2262 break; 2263 case F2FS_GOING_DOWN_NEED_FSCK: 2264 set_sbi_flag(sbi, SBI_NEED_FSCK); 2265 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2266 set_sbi_flag(sbi, SBI_IS_DIRTY); 2267 /* do checkpoint only */ 2268 ret = f2fs_sync_fs(sb, 1); 2269 goto out; 2270 default: 2271 ret = -EINVAL; 2272 goto out; 2273 } 2274 2275 f2fs_stop_gc_thread(sbi); 2276 f2fs_stop_discard_thread(sbi); 2277 2278 f2fs_drop_discard_cmd(sbi); 2279 clear_opt(sbi, DISCARD); 2280 2281 f2fs_update_time(sbi, REQ_TIME); 2282 out: 2283 if (in != F2FS_GOING_DOWN_FULLSYNC) 2284 mnt_drop_write_file(filp); 2285 2286 trace_f2fs_shutdown(sbi, in, ret); 2287 2288 return ret; 2289 } 2290 2291 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2292 { 2293 struct inode *inode = file_inode(filp); 2294 struct super_block *sb = inode->i_sb; 2295 struct fstrim_range range; 2296 int ret; 2297 2298 if (!capable(CAP_SYS_ADMIN)) 2299 return -EPERM; 2300 2301 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2302 return -EOPNOTSUPP; 2303 2304 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2305 sizeof(range))) 2306 return -EFAULT; 2307 2308 ret = mnt_want_write_file(filp); 2309 if (ret) 2310 return ret; 2311 2312 range.minlen = max((unsigned int)range.minlen, 2313 bdev_discard_granularity(sb->s_bdev)); 2314 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2315 mnt_drop_write_file(filp); 2316 if (ret < 0) 2317 return ret; 2318 2319 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2320 sizeof(range))) 2321 return -EFAULT; 2322 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2323 return 0; 2324 } 2325 2326 static bool uuid_is_nonzero(__u8 u[16]) 2327 { 2328 int i; 2329 2330 for (i = 0; i < 16; i++) 2331 if (u[i]) 2332 return true; 2333 return false; 2334 } 2335 2336 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2337 { 2338 struct inode *inode = file_inode(filp); 2339 2340 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2341 return -EOPNOTSUPP; 2342 2343 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2344 2345 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2346 } 2347 2348 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2349 { 2350 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2351 return -EOPNOTSUPP; 2352 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2353 } 2354 2355 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2356 { 2357 struct inode *inode = file_inode(filp); 2358 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2359 u8 encrypt_pw_salt[16]; 2360 int err; 2361 2362 if (!f2fs_sb_has_encrypt(sbi)) 2363 return -EOPNOTSUPP; 2364 2365 err = mnt_want_write_file(filp); 2366 if (err) 2367 return err; 2368 2369 f2fs_down_write(&sbi->sb_lock); 2370 2371 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2372 goto got_it; 2373 2374 /* update superblock with uuid */ 2375 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2376 2377 err = f2fs_commit_super(sbi, false); 2378 if (err) { 2379 /* undo new data */ 2380 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2381 goto out_err; 2382 } 2383 got_it: 2384 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2385 out_err: 2386 f2fs_up_write(&sbi->sb_lock); 2387 mnt_drop_write_file(filp); 2388 2389 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2390 err = -EFAULT; 2391 2392 return err; 2393 } 2394 2395 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2396 unsigned long arg) 2397 { 2398 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2399 return -EOPNOTSUPP; 2400 2401 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2402 } 2403 2404 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2405 { 2406 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2407 return -EOPNOTSUPP; 2408 2409 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2410 } 2411 2412 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2413 { 2414 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2415 return -EOPNOTSUPP; 2416 2417 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2418 } 2419 2420 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2421 unsigned long arg) 2422 { 2423 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2424 return -EOPNOTSUPP; 2425 2426 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2427 } 2428 2429 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2430 unsigned long arg) 2431 { 2432 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2433 return -EOPNOTSUPP; 2434 2435 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2436 } 2437 2438 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2439 { 2440 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2441 return -EOPNOTSUPP; 2442 2443 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2444 } 2445 2446 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2447 { 2448 struct inode *inode = file_inode(filp); 2449 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2450 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2451 .no_bg_gc = false, 2452 .should_migrate_blocks = false, 2453 .nr_free_secs = 0 }; 2454 __u32 sync; 2455 int ret; 2456 2457 if (!capable(CAP_SYS_ADMIN)) 2458 return -EPERM; 2459 2460 if (get_user(sync, (__u32 __user *)arg)) 2461 return -EFAULT; 2462 2463 if (f2fs_readonly(sbi->sb)) 2464 return -EROFS; 2465 2466 ret = mnt_want_write_file(filp); 2467 if (ret) 2468 return ret; 2469 2470 if (!sync) { 2471 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2472 ret = -EBUSY; 2473 goto out; 2474 } 2475 } else { 2476 f2fs_down_write(&sbi->gc_lock); 2477 } 2478 2479 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2480 gc_control.err_gc_skipped = sync; 2481 stat_inc_gc_call_count(sbi, FOREGROUND); 2482 ret = f2fs_gc(sbi, &gc_control); 2483 out: 2484 mnt_drop_write_file(filp); 2485 return ret; 2486 } 2487 2488 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2489 { 2490 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2491 struct f2fs_gc_control gc_control = { 2492 .init_gc_type = range->sync ? FG_GC : BG_GC, 2493 .no_bg_gc = false, 2494 .should_migrate_blocks = false, 2495 .err_gc_skipped = range->sync, 2496 .nr_free_secs = 0 }; 2497 u64 end; 2498 int ret; 2499 2500 if (!capable(CAP_SYS_ADMIN)) 2501 return -EPERM; 2502 if (f2fs_readonly(sbi->sb)) 2503 return -EROFS; 2504 2505 end = range->start + range->len; 2506 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2507 end >= MAX_BLKADDR(sbi)) 2508 return -EINVAL; 2509 2510 ret = mnt_want_write_file(filp); 2511 if (ret) 2512 return ret; 2513 2514 do_more: 2515 if (!range->sync) { 2516 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2517 ret = -EBUSY; 2518 goto out; 2519 } 2520 } else { 2521 f2fs_down_write(&sbi->gc_lock); 2522 } 2523 2524 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2525 stat_inc_gc_call_count(sbi, FOREGROUND); 2526 ret = f2fs_gc(sbi, &gc_control); 2527 if (ret) { 2528 if (ret == -EBUSY) 2529 ret = -EAGAIN; 2530 goto out; 2531 } 2532 range->start += CAP_BLKS_PER_SEC(sbi); 2533 if (range->start <= end) 2534 goto do_more; 2535 out: 2536 mnt_drop_write_file(filp); 2537 return ret; 2538 } 2539 2540 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2541 { 2542 struct f2fs_gc_range range; 2543 2544 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2545 sizeof(range))) 2546 return -EFAULT; 2547 return __f2fs_ioc_gc_range(filp, &range); 2548 } 2549 2550 static int f2fs_ioc_write_checkpoint(struct file *filp) 2551 { 2552 struct inode *inode = file_inode(filp); 2553 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2554 int ret; 2555 2556 if (!capable(CAP_SYS_ADMIN)) 2557 return -EPERM; 2558 2559 if (f2fs_readonly(sbi->sb)) 2560 return -EROFS; 2561 2562 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2563 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2564 return -EINVAL; 2565 } 2566 2567 ret = mnt_want_write_file(filp); 2568 if (ret) 2569 return ret; 2570 2571 ret = f2fs_sync_fs(sbi->sb, 1); 2572 2573 mnt_drop_write_file(filp); 2574 return ret; 2575 } 2576 2577 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2578 struct file *filp, 2579 struct f2fs_defragment *range) 2580 { 2581 struct inode *inode = file_inode(filp); 2582 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2583 .m_seg_type = NO_CHECK_TYPE, 2584 .m_may_create = false }; 2585 struct extent_info ei = {}; 2586 pgoff_t pg_start, pg_end, next_pgofs; 2587 unsigned int blk_per_seg = sbi->blocks_per_seg; 2588 unsigned int total = 0, sec_num; 2589 block_t blk_end = 0; 2590 bool fragmented = false; 2591 int err; 2592 2593 pg_start = range->start >> PAGE_SHIFT; 2594 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2595 2596 f2fs_balance_fs(sbi, true); 2597 2598 inode_lock(inode); 2599 2600 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2601 err = -EINVAL; 2602 goto unlock_out; 2603 } 2604 2605 /* if in-place-update policy is enabled, don't waste time here */ 2606 set_inode_flag(inode, FI_OPU_WRITE); 2607 if (f2fs_should_update_inplace(inode, NULL)) { 2608 err = -EINVAL; 2609 goto out; 2610 } 2611 2612 /* writeback all dirty pages in the range */ 2613 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2614 range->start + range->len - 1); 2615 if (err) 2616 goto out; 2617 2618 /* 2619 * lookup mapping info in extent cache, skip defragmenting if physical 2620 * block addresses are continuous. 2621 */ 2622 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2623 if (ei.fofs + ei.len >= pg_end) 2624 goto out; 2625 } 2626 2627 map.m_lblk = pg_start; 2628 map.m_next_pgofs = &next_pgofs; 2629 2630 /* 2631 * lookup mapping info in dnode page cache, skip defragmenting if all 2632 * physical block addresses are continuous even if there are hole(s) 2633 * in logical blocks. 2634 */ 2635 while (map.m_lblk < pg_end) { 2636 map.m_len = pg_end - map.m_lblk; 2637 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2638 if (err) 2639 goto out; 2640 2641 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2642 map.m_lblk = next_pgofs; 2643 continue; 2644 } 2645 2646 if (blk_end && blk_end != map.m_pblk) 2647 fragmented = true; 2648 2649 /* record total count of block that we're going to move */ 2650 total += map.m_len; 2651 2652 blk_end = map.m_pblk + map.m_len; 2653 2654 map.m_lblk += map.m_len; 2655 } 2656 2657 if (!fragmented) { 2658 total = 0; 2659 goto out; 2660 } 2661 2662 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2663 2664 /* 2665 * make sure there are enough free section for LFS allocation, this can 2666 * avoid defragment running in SSR mode when free section are allocated 2667 * intensively 2668 */ 2669 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2670 err = -EAGAIN; 2671 goto out; 2672 } 2673 2674 map.m_lblk = pg_start; 2675 map.m_len = pg_end - pg_start; 2676 total = 0; 2677 2678 while (map.m_lblk < pg_end) { 2679 pgoff_t idx; 2680 int cnt = 0; 2681 2682 do_map: 2683 map.m_len = pg_end - map.m_lblk; 2684 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2685 if (err) 2686 goto clear_out; 2687 2688 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2689 map.m_lblk = next_pgofs; 2690 goto check; 2691 } 2692 2693 set_inode_flag(inode, FI_SKIP_WRITES); 2694 2695 idx = map.m_lblk; 2696 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2697 struct page *page; 2698 2699 page = f2fs_get_lock_data_page(inode, idx, true); 2700 if (IS_ERR(page)) { 2701 err = PTR_ERR(page); 2702 goto clear_out; 2703 } 2704 2705 set_page_dirty(page); 2706 set_page_private_gcing(page); 2707 f2fs_put_page(page, 1); 2708 2709 idx++; 2710 cnt++; 2711 total++; 2712 } 2713 2714 map.m_lblk = idx; 2715 check: 2716 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2717 goto do_map; 2718 2719 clear_inode_flag(inode, FI_SKIP_WRITES); 2720 2721 err = filemap_fdatawrite(inode->i_mapping); 2722 if (err) 2723 goto out; 2724 } 2725 clear_out: 2726 clear_inode_flag(inode, FI_SKIP_WRITES); 2727 out: 2728 clear_inode_flag(inode, FI_OPU_WRITE); 2729 unlock_out: 2730 inode_unlock(inode); 2731 if (!err) 2732 range->len = (u64)total << PAGE_SHIFT; 2733 return err; 2734 } 2735 2736 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2737 { 2738 struct inode *inode = file_inode(filp); 2739 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2740 struct f2fs_defragment range; 2741 int err; 2742 2743 if (!capable(CAP_SYS_ADMIN)) 2744 return -EPERM; 2745 2746 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2747 return -EINVAL; 2748 2749 if (f2fs_readonly(sbi->sb)) 2750 return -EROFS; 2751 2752 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2753 sizeof(range))) 2754 return -EFAULT; 2755 2756 /* verify alignment of offset & size */ 2757 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2758 return -EINVAL; 2759 2760 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2761 max_file_blocks(inode))) 2762 return -EINVAL; 2763 2764 err = mnt_want_write_file(filp); 2765 if (err) 2766 return err; 2767 2768 err = f2fs_defragment_range(sbi, filp, &range); 2769 mnt_drop_write_file(filp); 2770 2771 f2fs_update_time(sbi, REQ_TIME); 2772 if (err < 0) 2773 return err; 2774 2775 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2776 sizeof(range))) 2777 return -EFAULT; 2778 2779 return 0; 2780 } 2781 2782 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2783 struct file *file_out, loff_t pos_out, size_t len) 2784 { 2785 struct inode *src = file_inode(file_in); 2786 struct inode *dst = file_inode(file_out); 2787 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2788 size_t olen = len, dst_max_i_size = 0; 2789 size_t dst_osize; 2790 int ret; 2791 2792 if (file_in->f_path.mnt != file_out->f_path.mnt || 2793 src->i_sb != dst->i_sb) 2794 return -EXDEV; 2795 2796 if (unlikely(f2fs_readonly(src->i_sb))) 2797 return -EROFS; 2798 2799 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2800 return -EINVAL; 2801 2802 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2803 return -EOPNOTSUPP; 2804 2805 if (pos_out < 0 || pos_in < 0) 2806 return -EINVAL; 2807 2808 if (src == dst) { 2809 if (pos_in == pos_out) 2810 return 0; 2811 if (pos_out > pos_in && pos_out < pos_in + len) 2812 return -EINVAL; 2813 } 2814 2815 inode_lock(src); 2816 if (src != dst) { 2817 ret = -EBUSY; 2818 if (!inode_trylock(dst)) 2819 goto out; 2820 } 2821 2822 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) { 2823 ret = -EOPNOTSUPP; 2824 goto out_unlock; 2825 } 2826 2827 ret = -EINVAL; 2828 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2829 goto out_unlock; 2830 if (len == 0) 2831 olen = len = src->i_size - pos_in; 2832 if (pos_in + len == src->i_size) 2833 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2834 if (len == 0) { 2835 ret = 0; 2836 goto out_unlock; 2837 } 2838 2839 dst_osize = dst->i_size; 2840 if (pos_out + olen > dst->i_size) 2841 dst_max_i_size = pos_out + olen; 2842 2843 /* verify the end result is block aligned */ 2844 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2845 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2846 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2847 goto out_unlock; 2848 2849 ret = f2fs_convert_inline_inode(src); 2850 if (ret) 2851 goto out_unlock; 2852 2853 ret = f2fs_convert_inline_inode(dst); 2854 if (ret) 2855 goto out_unlock; 2856 2857 /* write out all dirty pages from offset */ 2858 ret = filemap_write_and_wait_range(src->i_mapping, 2859 pos_in, pos_in + len); 2860 if (ret) 2861 goto out_unlock; 2862 2863 ret = filemap_write_and_wait_range(dst->i_mapping, 2864 pos_out, pos_out + len); 2865 if (ret) 2866 goto out_unlock; 2867 2868 f2fs_balance_fs(sbi, true); 2869 2870 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2871 if (src != dst) { 2872 ret = -EBUSY; 2873 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2874 goto out_src; 2875 } 2876 2877 f2fs_lock_op(sbi); 2878 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2879 pos_out >> F2FS_BLKSIZE_BITS, 2880 len >> F2FS_BLKSIZE_BITS, false); 2881 2882 if (!ret) { 2883 if (dst_max_i_size) 2884 f2fs_i_size_write(dst, dst_max_i_size); 2885 else if (dst_osize != dst->i_size) 2886 f2fs_i_size_write(dst, dst_osize); 2887 } 2888 f2fs_unlock_op(sbi); 2889 2890 if (src != dst) 2891 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2892 out_src: 2893 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2894 if (ret) 2895 goto out_unlock; 2896 2897 src->i_mtime = inode_set_ctime_current(src); 2898 f2fs_mark_inode_dirty_sync(src, false); 2899 if (src != dst) { 2900 dst->i_mtime = inode_set_ctime_current(dst); 2901 f2fs_mark_inode_dirty_sync(dst, false); 2902 } 2903 f2fs_update_time(sbi, REQ_TIME); 2904 2905 out_unlock: 2906 if (src != dst) 2907 inode_unlock(dst); 2908 out: 2909 inode_unlock(src); 2910 return ret; 2911 } 2912 2913 static int __f2fs_ioc_move_range(struct file *filp, 2914 struct f2fs_move_range *range) 2915 { 2916 struct fd dst; 2917 int err; 2918 2919 if (!(filp->f_mode & FMODE_READ) || 2920 !(filp->f_mode & FMODE_WRITE)) 2921 return -EBADF; 2922 2923 dst = fdget(range->dst_fd); 2924 if (!dst.file) 2925 return -EBADF; 2926 2927 if (!(dst.file->f_mode & FMODE_WRITE)) { 2928 err = -EBADF; 2929 goto err_out; 2930 } 2931 2932 err = mnt_want_write_file(filp); 2933 if (err) 2934 goto err_out; 2935 2936 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2937 range->pos_out, range->len); 2938 2939 mnt_drop_write_file(filp); 2940 err_out: 2941 fdput(dst); 2942 return err; 2943 } 2944 2945 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2946 { 2947 struct f2fs_move_range range; 2948 2949 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2950 sizeof(range))) 2951 return -EFAULT; 2952 return __f2fs_ioc_move_range(filp, &range); 2953 } 2954 2955 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2956 { 2957 struct inode *inode = file_inode(filp); 2958 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2959 struct sit_info *sm = SIT_I(sbi); 2960 unsigned int start_segno = 0, end_segno = 0; 2961 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2962 struct f2fs_flush_device range; 2963 struct f2fs_gc_control gc_control = { 2964 .init_gc_type = FG_GC, 2965 .should_migrate_blocks = true, 2966 .err_gc_skipped = true, 2967 .nr_free_secs = 0 }; 2968 int ret; 2969 2970 if (!capable(CAP_SYS_ADMIN)) 2971 return -EPERM; 2972 2973 if (f2fs_readonly(sbi->sb)) 2974 return -EROFS; 2975 2976 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2977 return -EINVAL; 2978 2979 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2980 sizeof(range))) 2981 return -EFAULT; 2982 2983 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2984 __is_large_section(sbi)) { 2985 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2986 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2987 return -EINVAL; 2988 } 2989 2990 ret = mnt_want_write_file(filp); 2991 if (ret) 2992 return ret; 2993 2994 if (range.dev_num != 0) 2995 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2996 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2997 2998 start_segno = sm->last_victim[FLUSH_DEVICE]; 2999 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3000 start_segno = dev_start_segno; 3001 end_segno = min(start_segno + range.segments, dev_end_segno); 3002 3003 while (start_segno < end_segno) { 3004 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3005 ret = -EBUSY; 3006 goto out; 3007 } 3008 sm->last_victim[GC_CB] = end_segno + 1; 3009 sm->last_victim[GC_GREEDY] = end_segno + 1; 3010 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3011 3012 gc_control.victim_segno = start_segno; 3013 stat_inc_gc_call_count(sbi, FOREGROUND); 3014 ret = f2fs_gc(sbi, &gc_control); 3015 if (ret == -EAGAIN) 3016 ret = 0; 3017 else if (ret < 0) 3018 break; 3019 start_segno++; 3020 } 3021 out: 3022 mnt_drop_write_file(filp); 3023 return ret; 3024 } 3025 3026 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3027 { 3028 struct inode *inode = file_inode(filp); 3029 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3030 3031 /* Must validate to set it with SQLite behavior in Android. */ 3032 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3033 3034 return put_user(sb_feature, (u32 __user *)arg); 3035 } 3036 3037 #ifdef CONFIG_QUOTA 3038 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3039 { 3040 struct dquot *transfer_to[MAXQUOTAS] = {}; 3041 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3042 struct super_block *sb = sbi->sb; 3043 int err; 3044 3045 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3046 if (IS_ERR(transfer_to[PRJQUOTA])) 3047 return PTR_ERR(transfer_to[PRJQUOTA]); 3048 3049 err = __dquot_transfer(inode, transfer_to); 3050 if (err) 3051 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3052 dqput(transfer_to[PRJQUOTA]); 3053 return err; 3054 } 3055 3056 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3057 { 3058 struct f2fs_inode_info *fi = F2FS_I(inode); 3059 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3060 struct f2fs_inode *ri = NULL; 3061 kprojid_t kprojid; 3062 int err; 3063 3064 if (!f2fs_sb_has_project_quota(sbi)) { 3065 if (projid != F2FS_DEF_PROJID) 3066 return -EOPNOTSUPP; 3067 else 3068 return 0; 3069 } 3070 3071 if (!f2fs_has_extra_attr(inode)) 3072 return -EOPNOTSUPP; 3073 3074 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3075 3076 if (projid_eq(kprojid, fi->i_projid)) 3077 return 0; 3078 3079 err = -EPERM; 3080 /* Is it quota file? Do not allow user to mess with it */ 3081 if (IS_NOQUOTA(inode)) 3082 return err; 3083 3084 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3085 return -EOVERFLOW; 3086 3087 err = f2fs_dquot_initialize(inode); 3088 if (err) 3089 return err; 3090 3091 f2fs_lock_op(sbi); 3092 err = f2fs_transfer_project_quota(inode, kprojid); 3093 if (err) 3094 goto out_unlock; 3095 3096 fi->i_projid = kprojid; 3097 inode_set_ctime_current(inode); 3098 f2fs_mark_inode_dirty_sync(inode, true); 3099 out_unlock: 3100 f2fs_unlock_op(sbi); 3101 return err; 3102 } 3103 #else 3104 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3105 { 3106 return 0; 3107 } 3108 3109 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3110 { 3111 if (projid != F2FS_DEF_PROJID) 3112 return -EOPNOTSUPP; 3113 return 0; 3114 } 3115 #endif 3116 3117 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3118 { 3119 struct inode *inode = d_inode(dentry); 3120 struct f2fs_inode_info *fi = F2FS_I(inode); 3121 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3122 3123 if (IS_ENCRYPTED(inode)) 3124 fsflags |= FS_ENCRYPT_FL; 3125 if (IS_VERITY(inode)) 3126 fsflags |= FS_VERITY_FL; 3127 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3128 fsflags |= FS_INLINE_DATA_FL; 3129 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3130 fsflags |= FS_NOCOW_FL; 3131 3132 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3133 3134 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3135 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3136 3137 return 0; 3138 } 3139 3140 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3141 struct dentry *dentry, struct fileattr *fa) 3142 { 3143 struct inode *inode = d_inode(dentry); 3144 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3145 u32 iflags; 3146 int err; 3147 3148 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3149 return -EIO; 3150 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3151 return -ENOSPC; 3152 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3153 return -EOPNOTSUPP; 3154 fsflags &= F2FS_SETTABLE_FS_FL; 3155 if (!fa->flags_valid) 3156 mask &= FS_COMMON_FL; 3157 3158 iflags = f2fs_fsflags_to_iflags(fsflags); 3159 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3160 return -EOPNOTSUPP; 3161 3162 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3163 if (!err) 3164 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3165 3166 return err; 3167 } 3168 3169 int f2fs_pin_file_control(struct inode *inode, bool inc) 3170 { 3171 struct f2fs_inode_info *fi = F2FS_I(inode); 3172 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3173 3174 /* Use i_gc_failures for normal file as a risk signal. */ 3175 if (inc) 3176 f2fs_i_gc_failures_write(inode, 3177 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3178 3179 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3180 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3181 __func__, inode->i_ino, 3182 fi->i_gc_failures[GC_FAILURE_PIN]); 3183 clear_inode_flag(inode, FI_PIN_FILE); 3184 return -EAGAIN; 3185 } 3186 return 0; 3187 } 3188 3189 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3190 { 3191 struct inode *inode = file_inode(filp); 3192 __u32 pin; 3193 int ret = 0; 3194 3195 if (get_user(pin, (__u32 __user *)arg)) 3196 return -EFAULT; 3197 3198 if (!S_ISREG(inode->i_mode)) 3199 return -EINVAL; 3200 3201 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3202 return -EROFS; 3203 3204 ret = mnt_want_write_file(filp); 3205 if (ret) 3206 return ret; 3207 3208 inode_lock(inode); 3209 3210 if (!pin) { 3211 clear_inode_flag(inode, FI_PIN_FILE); 3212 f2fs_i_gc_failures_write(inode, 0); 3213 goto done; 3214 } 3215 3216 if (f2fs_should_update_outplace(inode, NULL)) { 3217 ret = -EINVAL; 3218 goto out; 3219 } 3220 3221 if (f2fs_pin_file_control(inode, false)) { 3222 ret = -EAGAIN; 3223 goto out; 3224 } 3225 3226 ret = f2fs_convert_inline_inode(inode); 3227 if (ret) 3228 goto out; 3229 3230 if (!f2fs_disable_compressed_file(inode)) { 3231 ret = -EOPNOTSUPP; 3232 goto out; 3233 } 3234 3235 set_inode_flag(inode, FI_PIN_FILE); 3236 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3237 done: 3238 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3239 out: 3240 inode_unlock(inode); 3241 mnt_drop_write_file(filp); 3242 return ret; 3243 } 3244 3245 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3246 { 3247 struct inode *inode = file_inode(filp); 3248 __u32 pin = 0; 3249 3250 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3251 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3252 return put_user(pin, (u32 __user *)arg); 3253 } 3254 3255 int f2fs_precache_extents(struct inode *inode) 3256 { 3257 struct f2fs_inode_info *fi = F2FS_I(inode); 3258 struct f2fs_map_blocks map; 3259 pgoff_t m_next_extent; 3260 loff_t end; 3261 int err; 3262 3263 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3264 return -EOPNOTSUPP; 3265 3266 map.m_lblk = 0; 3267 map.m_pblk = 0; 3268 map.m_next_pgofs = NULL; 3269 map.m_next_extent = &m_next_extent; 3270 map.m_seg_type = NO_CHECK_TYPE; 3271 map.m_may_create = false; 3272 end = max_file_blocks(inode); 3273 3274 while (map.m_lblk < end) { 3275 map.m_len = end - map.m_lblk; 3276 3277 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3278 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3279 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3280 if (err) 3281 return err; 3282 3283 map.m_lblk = m_next_extent; 3284 } 3285 3286 return 0; 3287 } 3288 3289 static int f2fs_ioc_precache_extents(struct file *filp) 3290 { 3291 return f2fs_precache_extents(file_inode(filp)); 3292 } 3293 3294 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3295 { 3296 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3297 __u64 block_count; 3298 3299 if (!capable(CAP_SYS_ADMIN)) 3300 return -EPERM; 3301 3302 if (f2fs_readonly(sbi->sb)) 3303 return -EROFS; 3304 3305 if (copy_from_user(&block_count, (void __user *)arg, 3306 sizeof(block_count))) 3307 return -EFAULT; 3308 3309 return f2fs_resize_fs(filp, block_count); 3310 } 3311 3312 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3313 { 3314 struct inode *inode = file_inode(filp); 3315 3316 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3317 3318 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3319 f2fs_warn(F2FS_I_SB(inode), 3320 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3321 inode->i_ino); 3322 return -EOPNOTSUPP; 3323 } 3324 3325 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3326 } 3327 3328 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3329 { 3330 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3331 return -EOPNOTSUPP; 3332 3333 return fsverity_ioctl_measure(filp, (void __user *)arg); 3334 } 3335 3336 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3337 { 3338 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3339 return -EOPNOTSUPP; 3340 3341 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3342 } 3343 3344 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3345 { 3346 struct inode *inode = file_inode(filp); 3347 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3348 char *vbuf; 3349 int count; 3350 int err = 0; 3351 3352 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3353 if (!vbuf) 3354 return -ENOMEM; 3355 3356 f2fs_down_read(&sbi->sb_lock); 3357 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3358 ARRAY_SIZE(sbi->raw_super->volume_name), 3359 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3360 f2fs_up_read(&sbi->sb_lock); 3361 3362 if (copy_to_user((char __user *)arg, vbuf, 3363 min(FSLABEL_MAX, count))) 3364 err = -EFAULT; 3365 3366 kfree(vbuf); 3367 return err; 3368 } 3369 3370 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3371 { 3372 struct inode *inode = file_inode(filp); 3373 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3374 char *vbuf; 3375 int err = 0; 3376 3377 if (!capable(CAP_SYS_ADMIN)) 3378 return -EPERM; 3379 3380 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3381 if (IS_ERR(vbuf)) 3382 return PTR_ERR(vbuf); 3383 3384 err = mnt_want_write_file(filp); 3385 if (err) 3386 goto out; 3387 3388 f2fs_down_write(&sbi->sb_lock); 3389 3390 memset(sbi->raw_super->volume_name, 0, 3391 sizeof(sbi->raw_super->volume_name)); 3392 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3393 sbi->raw_super->volume_name, 3394 ARRAY_SIZE(sbi->raw_super->volume_name)); 3395 3396 err = f2fs_commit_super(sbi, false); 3397 3398 f2fs_up_write(&sbi->sb_lock); 3399 3400 mnt_drop_write_file(filp); 3401 out: 3402 kfree(vbuf); 3403 return err; 3404 } 3405 3406 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3407 { 3408 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3409 return -EOPNOTSUPP; 3410 3411 if (!f2fs_compressed_file(inode)) 3412 return -EINVAL; 3413 3414 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3415 3416 return 0; 3417 } 3418 3419 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3420 { 3421 struct inode *inode = file_inode(filp); 3422 __u64 blocks; 3423 int ret; 3424 3425 ret = f2fs_get_compress_blocks(inode, &blocks); 3426 if (ret < 0) 3427 return ret; 3428 3429 return put_user(blocks, (u64 __user *)arg); 3430 } 3431 3432 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3433 { 3434 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3435 unsigned int released_blocks = 0; 3436 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3437 block_t blkaddr; 3438 int i; 3439 3440 for (i = 0; i < count; i++) { 3441 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3442 dn->ofs_in_node + i); 3443 3444 if (!__is_valid_data_blkaddr(blkaddr)) 3445 continue; 3446 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3447 DATA_GENERIC_ENHANCE))) { 3448 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3449 return -EFSCORRUPTED; 3450 } 3451 } 3452 3453 while (count) { 3454 int compr_blocks = 0; 3455 3456 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3457 blkaddr = f2fs_data_blkaddr(dn); 3458 3459 if (i == 0) { 3460 if (blkaddr == COMPRESS_ADDR) 3461 continue; 3462 dn->ofs_in_node += cluster_size; 3463 goto next; 3464 } 3465 3466 if (__is_valid_data_blkaddr(blkaddr)) 3467 compr_blocks++; 3468 3469 if (blkaddr != NEW_ADDR) 3470 continue; 3471 3472 dn->data_blkaddr = NULL_ADDR; 3473 f2fs_set_data_blkaddr(dn); 3474 } 3475 3476 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3477 dec_valid_block_count(sbi, dn->inode, 3478 cluster_size - compr_blocks); 3479 3480 released_blocks += cluster_size - compr_blocks; 3481 next: 3482 count -= cluster_size; 3483 } 3484 3485 return released_blocks; 3486 } 3487 3488 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3489 { 3490 struct inode *inode = file_inode(filp); 3491 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3492 pgoff_t page_idx = 0, last_idx; 3493 unsigned int released_blocks = 0; 3494 int ret; 3495 int writecount; 3496 3497 if (!f2fs_sb_has_compression(sbi)) 3498 return -EOPNOTSUPP; 3499 3500 if (!f2fs_compressed_file(inode)) 3501 return -EINVAL; 3502 3503 if (f2fs_readonly(sbi->sb)) 3504 return -EROFS; 3505 3506 ret = mnt_want_write_file(filp); 3507 if (ret) 3508 return ret; 3509 3510 f2fs_balance_fs(sbi, true); 3511 3512 inode_lock(inode); 3513 3514 writecount = atomic_read(&inode->i_writecount); 3515 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3516 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3517 ret = -EBUSY; 3518 goto out; 3519 } 3520 3521 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3522 ret = -EINVAL; 3523 goto out; 3524 } 3525 3526 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3527 if (ret) 3528 goto out; 3529 3530 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3531 ret = -EPERM; 3532 goto out; 3533 } 3534 3535 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3536 inode_set_ctime_current(inode); 3537 f2fs_mark_inode_dirty_sync(inode, true); 3538 3539 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3540 filemap_invalidate_lock(inode->i_mapping); 3541 3542 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3543 3544 while (page_idx < last_idx) { 3545 struct dnode_of_data dn; 3546 pgoff_t end_offset, count; 3547 3548 set_new_dnode(&dn, inode, NULL, NULL, 0); 3549 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3550 if (ret) { 3551 if (ret == -ENOENT) { 3552 page_idx = f2fs_get_next_page_offset(&dn, 3553 page_idx); 3554 ret = 0; 3555 continue; 3556 } 3557 break; 3558 } 3559 3560 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3561 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3562 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3563 3564 ret = release_compress_blocks(&dn, count); 3565 3566 f2fs_put_dnode(&dn); 3567 3568 if (ret < 0) 3569 break; 3570 3571 page_idx += count; 3572 released_blocks += ret; 3573 } 3574 3575 filemap_invalidate_unlock(inode->i_mapping); 3576 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3577 out: 3578 inode_unlock(inode); 3579 3580 mnt_drop_write_file(filp); 3581 3582 if (ret >= 0) { 3583 ret = put_user(released_blocks, (u64 __user *)arg); 3584 } else if (released_blocks && 3585 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3586 set_sbi_flag(sbi, SBI_NEED_FSCK); 3587 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3588 "iblocks=%llu, released=%u, compr_blocks=%u, " 3589 "run fsck to fix.", 3590 __func__, inode->i_ino, inode->i_blocks, 3591 released_blocks, 3592 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3593 } 3594 3595 return ret; 3596 } 3597 3598 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3599 { 3600 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3601 unsigned int reserved_blocks = 0; 3602 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3603 block_t blkaddr; 3604 int i; 3605 3606 for (i = 0; i < count; i++) { 3607 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3608 dn->ofs_in_node + i); 3609 3610 if (!__is_valid_data_blkaddr(blkaddr)) 3611 continue; 3612 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3613 DATA_GENERIC_ENHANCE))) { 3614 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3615 return -EFSCORRUPTED; 3616 } 3617 } 3618 3619 while (count) { 3620 int compr_blocks = 0; 3621 blkcnt_t reserved; 3622 int ret; 3623 3624 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3625 blkaddr = f2fs_data_blkaddr(dn); 3626 3627 if (i == 0) { 3628 if (blkaddr == COMPRESS_ADDR) 3629 continue; 3630 dn->ofs_in_node += cluster_size; 3631 goto next; 3632 } 3633 3634 if (__is_valid_data_blkaddr(blkaddr)) { 3635 compr_blocks++; 3636 continue; 3637 } 3638 3639 dn->data_blkaddr = NEW_ADDR; 3640 f2fs_set_data_blkaddr(dn); 3641 } 3642 3643 reserved = cluster_size - compr_blocks; 3644 ret = inc_valid_block_count(sbi, dn->inode, &reserved); 3645 if (ret) 3646 return ret; 3647 3648 if (reserved != cluster_size - compr_blocks) 3649 return -ENOSPC; 3650 3651 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3652 3653 reserved_blocks += reserved; 3654 next: 3655 count -= cluster_size; 3656 } 3657 3658 return reserved_blocks; 3659 } 3660 3661 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3662 { 3663 struct inode *inode = file_inode(filp); 3664 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3665 pgoff_t page_idx = 0, last_idx; 3666 unsigned int reserved_blocks = 0; 3667 int ret; 3668 3669 if (!f2fs_sb_has_compression(sbi)) 3670 return -EOPNOTSUPP; 3671 3672 if (!f2fs_compressed_file(inode)) 3673 return -EINVAL; 3674 3675 if (f2fs_readonly(sbi->sb)) 3676 return -EROFS; 3677 3678 ret = mnt_want_write_file(filp); 3679 if (ret) 3680 return ret; 3681 3682 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3683 goto out; 3684 3685 f2fs_balance_fs(sbi, true); 3686 3687 inode_lock(inode); 3688 3689 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3690 ret = -EINVAL; 3691 goto unlock_inode; 3692 } 3693 3694 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3695 filemap_invalidate_lock(inode->i_mapping); 3696 3697 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3698 3699 while (page_idx < last_idx) { 3700 struct dnode_of_data dn; 3701 pgoff_t end_offset, count; 3702 3703 set_new_dnode(&dn, inode, NULL, NULL, 0); 3704 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3705 if (ret) { 3706 if (ret == -ENOENT) { 3707 page_idx = f2fs_get_next_page_offset(&dn, 3708 page_idx); 3709 ret = 0; 3710 continue; 3711 } 3712 break; 3713 } 3714 3715 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3716 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3717 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3718 3719 ret = reserve_compress_blocks(&dn, count); 3720 3721 f2fs_put_dnode(&dn); 3722 3723 if (ret < 0) 3724 break; 3725 3726 page_idx += count; 3727 reserved_blocks += ret; 3728 } 3729 3730 filemap_invalidate_unlock(inode->i_mapping); 3731 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3732 3733 if (ret >= 0) { 3734 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3735 inode_set_ctime_current(inode); 3736 f2fs_mark_inode_dirty_sync(inode, true); 3737 } 3738 unlock_inode: 3739 inode_unlock(inode); 3740 out: 3741 mnt_drop_write_file(filp); 3742 3743 if (ret >= 0) { 3744 ret = put_user(reserved_blocks, (u64 __user *)arg); 3745 } else if (reserved_blocks && 3746 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3747 set_sbi_flag(sbi, SBI_NEED_FSCK); 3748 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3749 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3750 "run fsck to fix.", 3751 __func__, inode->i_ino, inode->i_blocks, 3752 reserved_blocks, 3753 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3754 } 3755 3756 return ret; 3757 } 3758 3759 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3760 pgoff_t off, block_t block, block_t len, u32 flags) 3761 { 3762 sector_t sector = SECTOR_FROM_BLOCK(block); 3763 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3764 int ret = 0; 3765 3766 if (flags & F2FS_TRIM_FILE_DISCARD) { 3767 if (bdev_max_secure_erase_sectors(bdev)) 3768 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3769 GFP_NOFS); 3770 else 3771 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3772 GFP_NOFS); 3773 } 3774 3775 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3776 if (IS_ENCRYPTED(inode)) 3777 ret = fscrypt_zeroout_range(inode, off, block, len); 3778 else 3779 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3780 GFP_NOFS, 0); 3781 } 3782 3783 return ret; 3784 } 3785 3786 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3787 { 3788 struct inode *inode = file_inode(filp); 3789 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3790 struct address_space *mapping = inode->i_mapping; 3791 struct block_device *prev_bdev = NULL; 3792 struct f2fs_sectrim_range range; 3793 pgoff_t index, pg_end, prev_index = 0; 3794 block_t prev_block = 0, len = 0; 3795 loff_t end_addr; 3796 bool to_end = false; 3797 int ret = 0; 3798 3799 if (!(filp->f_mode & FMODE_WRITE)) 3800 return -EBADF; 3801 3802 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3803 sizeof(range))) 3804 return -EFAULT; 3805 3806 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3807 !S_ISREG(inode->i_mode)) 3808 return -EINVAL; 3809 3810 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3811 !f2fs_hw_support_discard(sbi)) || 3812 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3813 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3814 return -EOPNOTSUPP; 3815 3816 file_start_write(filp); 3817 inode_lock(inode); 3818 3819 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3820 range.start >= inode->i_size) { 3821 ret = -EINVAL; 3822 goto err; 3823 } 3824 3825 if (range.len == 0) 3826 goto err; 3827 3828 if (inode->i_size - range.start > range.len) { 3829 end_addr = range.start + range.len; 3830 } else { 3831 end_addr = range.len == (u64)-1 ? 3832 sbi->sb->s_maxbytes : inode->i_size; 3833 to_end = true; 3834 } 3835 3836 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3837 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3838 ret = -EINVAL; 3839 goto err; 3840 } 3841 3842 index = F2FS_BYTES_TO_BLK(range.start); 3843 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3844 3845 ret = f2fs_convert_inline_inode(inode); 3846 if (ret) 3847 goto err; 3848 3849 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3850 filemap_invalidate_lock(mapping); 3851 3852 ret = filemap_write_and_wait_range(mapping, range.start, 3853 to_end ? LLONG_MAX : end_addr - 1); 3854 if (ret) 3855 goto out; 3856 3857 truncate_inode_pages_range(mapping, range.start, 3858 to_end ? -1 : end_addr - 1); 3859 3860 while (index < pg_end) { 3861 struct dnode_of_data dn; 3862 pgoff_t end_offset, count; 3863 int i; 3864 3865 set_new_dnode(&dn, inode, NULL, NULL, 0); 3866 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3867 if (ret) { 3868 if (ret == -ENOENT) { 3869 index = f2fs_get_next_page_offset(&dn, index); 3870 continue; 3871 } 3872 goto out; 3873 } 3874 3875 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3876 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3877 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3878 struct block_device *cur_bdev; 3879 block_t blkaddr = f2fs_data_blkaddr(&dn); 3880 3881 if (!__is_valid_data_blkaddr(blkaddr)) 3882 continue; 3883 3884 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3885 DATA_GENERIC_ENHANCE)) { 3886 ret = -EFSCORRUPTED; 3887 f2fs_put_dnode(&dn); 3888 f2fs_handle_error(sbi, 3889 ERROR_INVALID_BLKADDR); 3890 goto out; 3891 } 3892 3893 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3894 if (f2fs_is_multi_device(sbi)) { 3895 int di = f2fs_target_device_index(sbi, blkaddr); 3896 3897 blkaddr -= FDEV(di).start_blk; 3898 } 3899 3900 if (len) { 3901 if (prev_bdev == cur_bdev && 3902 index == prev_index + len && 3903 blkaddr == prev_block + len) { 3904 len++; 3905 } else { 3906 ret = f2fs_secure_erase(prev_bdev, 3907 inode, prev_index, prev_block, 3908 len, range.flags); 3909 if (ret) { 3910 f2fs_put_dnode(&dn); 3911 goto out; 3912 } 3913 3914 len = 0; 3915 } 3916 } 3917 3918 if (!len) { 3919 prev_bdev = cur_bdev; 3920 prev_index = index; 3921 prev_block = blkaddr; 3922 len = 1; 3923 } 3924 } 3925 3926 f2fs_put_dnode(&dn); 3927 3928 if (fatal_signal_pending(current)) { 3929 ret = -EINTR; 3930 goto out; 3931 } 3932 cond_resched(); 3933 } 3934 3935 if (len) 3936 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3937 prev_block, len, range.flags); 3938 out: 3939 filemap_invalidate_unlock(mapping); 3940 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3941 err: 3942 inode_unlock(inode); 3943 file_end_write(filp); 3944 3945 return ret; 3946 } 3947 3948 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3949 { 3950 struct inode *inode = file_inode(filp); 3951 struct f2fs_comp_option option; 3952 3953 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3954 return -EOPNOTSUPP; 3955 3956 inode_lock_shared(inode); 3957 3958 if (!f2fs_compressed_file(inode)) { 3959 inode_unlock_shared(inode); 3960 return -ENODATA; 3961 } 3962 3963 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3964 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3965 3966 inode_unlock_shared(inode); 3967 3968 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3969 sizeof(option))) 3970 return -EFAULT; 3971 3972 return 0; 3973 } 3974 3975 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3976 { 3977 struct inode *inode = file_inode(filp); 3978 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3979 struct f2fs_comp_option option; 3980 int ret = 0; 3981 3982 if (!f2fs_sb_has_compression(sbi)) 3983 return -EOPNOTSUPP; 3984 3985 if (!(filp->f_mode & FMODE_WRITE)) 3986 return -EBADF; 3987 3988 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3989 sizeof(option))) 3990 return -EFAULT; 3991 3992 if (!f2fs_compressed_file(inode) || 3993 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3994 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3995 option.algorithm >= COMPRESS_MAX) 3996 return -EINVAL; 3997 3998 file_start_write(filp); 3999 inode_lock(inode); 4000 4001 f2fs_down_write(&F2FS_I(inode)->i_sem); 4002 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4003 ret = -EBUSY; 4004 goto out; 4005 } 4006 4007 if (F2FS_HAS_BLOCKS(inode)) { 4008 ret = -EFBIG; 4009 goto out; 4010 } 4011 4012 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4013 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4014 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4015 /* Set default level */ 4016 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4017 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4018 else 4019 F2FS_I(inode)->i_compress_level = 0; 4020 /* Adjust mount option level */ 4021 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4022 F2FS_OPTION(sbi).compress_level) 4023 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4024 f2fs_mark_inode_dirty_sync(inode, true); 4025 4026 if (!f2fs_is_compress_backend_ready(inode)) 4027 f2fs_warn(sbi, "compression algorithm is successfully set, " 4028 "but current kernel doesn't support this algorithm."); 4029 out: 4030 f2fs_up_write(&F2FS_I(inode)->i_sem); 4031 inode_unlock(inode); 4032 file_end_write(filp); 4033 4034 return ret; 4035 } 4036 4037 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4038 { 4039 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4040 struct address_space *mapping = inode->i_mapping; 4041 struct page *page; 4042 pgoff_t redirty_idx = page_idx; 4043 int i, page_len = 0, ret = 0; 4044 4045 page_cache_ra_unbounded(&ractl, len, 0); 4046 4047 for (i = 0; i < len; i++, page_idx++) { 4048 page = read_cache_page(mapping, page_idx, NULL, NULL); 4049 if (IS_ERR(page)) { 4050 ret = PTR_ERR(page); 4051 break; 4052 } 4053 page_len++; 4054 } 4055 4056 for (i = 0; i < page_len; i++, redirty_idx++) { 4057 page = find_lock_page(mapping, redirty_idx); 4058 4059 /* It will never fail, when page has pinned above */ 4060 f2fs_bug_on(F2FS_I_SB(inode), !page); 4061 4062 set_page_dirty(page); 4063 set_page_private_gcing(page); 4064 f2fs_put_page(page, 1); 4065 f2fs_put_page(page, 0); 4066 } 4067 4068 return ret; 4069 } 4070 4071 static int f2fs_ioc_decompress_file(struct file *filp) 4072 { 4073 struct inode *inode = file_inode(filp); 4074 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4075 struct f2fs_inode_info *fi = F2FS_I(inode); 4076 pgoff_t page_idx = 0, last_idx; 4077 unsigned int blk_per_seg = sbi->blocks_per_seg; 4078 int cluster_size = fi->i_cluster_size; 4079 int count, ret; 4080 4081 if (!f2fs_sb_has_compression(sbi) || 4082 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4083 return -EOPNOTSUPP; 4084 4085 if (!(filp->f_mode & FMODE_WRITE)) 4086 return -EBADF; 4087 4088 if (!f2fs_compressed_file(inode)) 4089 return -EINVAL; 4090 4091 f2fs_balance_fs(sbi, true); 4092 4093 file_start_write(filp); 4094 inode_lock(inode); 4095 4096 if (!f2fs_is_compress_backend_ready(inode)) { 4097 ret = -EOPNOTSUPP; 4098 goto out; 4099 } 4100 4101 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4102 ret = -EINVAL; 4103 goto out; 4104 } 4105 4106 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4107 if (ret) 4108 goto out; 4109 4110 if (!atomic_read(&fi->i_compr_blocks)) 4111 goto out; 4112 4113 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4114 4115 count = last_idx - page_idx; 4116 while (count && count >= cluster_size) { 4117 ret = redirty_blocks(inode, page_idx, cluster_size); 4118 if (ret < 0) 4119 break; 4120 4121 if (get_dirty_pages(inode) >= blk_per_seg) { 4122 ret = filemap_fdatawrite(inode->i_mapping); 4123 if (ret < 0) 4124 break; 4125 } 4126 4127 count -= cluster_size; 4128 page_idx += cluster_size; 4129 4130 cond_resched(); 4131 if (fatal_signal_pending(current)) { 4132 ret = -EINTR; 4133 break; 4134 } 4135 } 4136 4137 if (!ret) 4138 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4139 LLONG_MAX); 4140 4141 if (ret) 4142 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4143 __func__, ret); 4144 out: 4145 inode_unlock(inode); 4146 file_end_write(filp); 4147 4148 return ret; 4149 } 4150 4151 static int f2fs_ioc_compress_file(struct file *filp) 4152 { 4153 struct inode *inode = file_inode(filp); 4154 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4155 pgoff_t page_idx = 0, last_idx; 4156 unsigned int blk_per_seg = sbi->blocks_per_seg; 4157 int cluster_size = F2FS_I(inode)->i_cluster_size; 4158 int count, ret; 4159 4160 if (!f2fs_sb_has_compression(sbi) || 4161 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4162 return -EOPNOTSUPP; 4163 4164 if (!(filp->f_mode & FMODE_WRITE)) 4165 return -EBADF; 4166 4167 if (!f2fs_compressed_file(inode)) 4168 return -EINVAL; 4169 4170 f2fs_balance_fs(sbi, true); 4171 4172 file_start_write(filp); 4173 inode_lock(inode); 4174 4175 if (!f2fs_is_compress_backend_ready(inode)) { 4176 ret = -EOPNOTSUPP; 4177 goto out; 4178 } 4179 4180 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4181 ret = -EINVAL; 4182 goto out; 4183 } 4184 4185 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4186 if (ret) 4187 goto out; 4188 4189 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4190 4191 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4192 4193 count = last_idx - page_idx; 4194 while (count && count >= cluster_size) { 4195 ret = redirty_blocks(inode, page_idx, cluster_size); 4196 if (ret < 0) 4197 break; 4198 4199 if (get_dirty_pages(inode) >= blk_per_seg) { 4200 ret = filemap_fdatawrite(inode->i_mapping); 4201 if (ret < 0) 4202 break; 4203 } 4204 4205 count -= cluster_size; 4206 page_idx += cluster_size; 4207 4208 cond_resched(); 4209 if (fatal_signal_pending(current)) { 4210 ret = -EINTR; 4211 break; 4212 } 4213 } 4214 4215 if (!ret) 4216 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4217 LLONG_MAX); 4218 4219 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4220 4221 if (ret) 4222 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4223 __func__, ret); 4224 out: 4225 inode_unlock(inode); 4226 file_end_write(filp); 4227 4228 return ret; 4229 } 4230 4231 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4232 { 4233 switch (cmd) { 4234 case FS_IOC_GETVERSION: 4235 return f2fs_ioc_getversion(filp, arg); 4236 case F2FS_IOC_START_ATOMIC_WRITE: 4237 return f2fs_ioc_start_atomic_write(filp, false); 4238 case F2FS_IOC_START_ATOMIC_REPLACE: 4239 return f2fs_ioc_start_atomic_write(filp, true); 4240 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4241 return f2fs_ioc_commit_atomic_write(filp); 4242 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4243 return f2fs_ioc_abort_atomic_write(filp); 4244 case F2FS_IOC_START_VOLATILE_WRITE: 4245 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4246 return -EOPNOTSUPP; 4247 case F2FS_IOC_SHUTDOWN: 4248 return f2fs_ioc_shutdown(filp, arg); 4249 case FITRIM: 4250 return f2fs_ioc_fitrim(filp, arg); 4251 case FS_IOC_SET_ENCRYPTION_POLICY: 4252 return f2fs_ioc_set_encryption_policy(filp, arg); 4253 case FS_IOC_GET_ENCRYPTION_POLICY: 4254 return f2fs_ioc_get_encryption_policy(filp, arg); 4255 case FS_IOC_GET_ENCRYPTION_PWSALT: 4256 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4257 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4258 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4259 case FS_IOC_ADD_ENCRYPTION_KEY: 4260 return f2fs_ioc_add_encryption_key(filp, arg); 4261 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4262 return f2fs_ioc_remove_encryption_key(filp, arg); 4263 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4264 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4265 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4266 return f2fs_ioc_get_encryption_key_status(filp, arg); 4267 case FS_IOC_GET_ENCRYPTION_NONCE: 4268 return f2fs_ioc_get_encryption_nonce(filp, arg); 4269 case F2FS_IOC_GARBAGE_COLLECT: 4270 return f2fs_ioc_gc(filp, arg); 4271 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4272 return f2fs_ioc_gc_range(filp, arg); 4273 case F2FS_IOC_WRITE_CHECKPOINT: 4274 return f2fs_ioc_write_checkpoint(filp); 4275 case F2FS_IOC_DEFRAGMENT: 4276 return f2fs_ioc_defragment(filp, arg); 4277 case F2FS_IOC_MOVE_RANGE: 4278 return f2fs_ioc_move_range(filp, arg); 4279 case F2FS_IOC_FLUSH_DEVICE: 4280 return f2fs_ioc_flush_device(filp, arg); 4281 case F2FS_IOC_GET_FEATURES: 4282 return f2fs_ioc_get_features(filp, arg); 4283 case F2FS_IOC_GET_PIN_FILE: 4284 return f2fs_ioc_get_pin_file(filp, arg); 4285 case F2FS_IOC_SET_PIN_FILE: 4286 return f2fs_ioc_set_pin_file(filp, arg); 4287 case F2FS_IOC_PRECACHE_EXTENTS: 4288 return f2fs_ioc_precache_extents(filp); 4289 case F2FS_IOC_RESIZE_FS: 4290 return f2fs_ioc_resize_fs(filp, arg); 4291 case FS_IOC_ENABLE_VERITY: 4292 return f2fs_ioc_enable_verity(filp, arg); 4293 case FS_IOC_MEASURE_VERITY: 4294 return f2fs_ioc_measure_verity(filp, arg); 4295 case FS_IOC_READ_VERITY_METADATA: 4296 return f2fs_ioc_read_verity_metadata(filp, arg); 4297 case FS_IOC_GETFSLABEL: 4298 return f2fs_ioc_getfslabel(filp, arg); 4299 case FS_IOC_SETFSLABEL: 4300 return f2fs_ioc_setfslabel(filp, arg); 4301 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4302 return f2fs_ioc_get_compress_blocks(filp, arg); 4303 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4304 return f2fs_release_compress_blocks(filp, arg); 4305 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4306 return f2fs_reserve_compress_blocks(filp, arg); 4307 case F2FS_IOC_SEC_TRIM_FILE: 4308 return f2fs_sec_trim_file(filp, arg); 4309 case F2FS_IOC_GET_COMPRESS_OPTION: 4310 return f2fs_ioc_get_compress_option(filp, arg); 4311 case F2FS_IOC_SET_COMPRESS_OPTION: 4312 return f2fs_ioc_set_compress_option(filp, arg); 4313 case F2FS_IOC_DECOMPRESS_FILE: 4314 return f2fs_ioc_decompress_file(filp); 4315 case F2FS_IOC_COMPRESS_FILE: 4316 return f2fs_ioc_compress_file(filp); 4317 default: 4318 return -ENOTTY; 4319 } 4320 } 4321 4322 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4323 { 4324 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4325 return -EIO; 4326 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4327 return -ENOSPC; 4328 4329 return __f2fs_ioctl(filp, cmd, arg); 4330 } 4331 4332 /* 4333 * Return %true if the given read or write request should use direct I/O, or 4334 * %false if it should use buffered I/O. 4335 */ 4336 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4337 struct iov_iter *iter) 4338 { 4339 unsigned int align; 4340 4341 if (!(iocb->ki_flags & IOCB_DIRECT)) 4342 return false; 4343 4344 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4345 return false; 4346 4347 /* 4348 * Direct I/O not aligned to the disk's logical_block_size will be 4349 * attempted, but will fail with -EINVAL. 4350 * 4351 * f2fs additionally requires that direct I/O be aligned to the 4352 * filesystem block size, which is often a stricter requirement. 4353 * However, f2fs traditionally falls back to buffered I/O on requests 4354 * that are logical_block_size-aligned but not fs-block aligned. 4355 * 4356 * The below logic implements this behavior. 4357 */ 4358 align = iocb->ki_pos | iov_iter_alignment(iter); 4359 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4360 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4361 return false; 4362 4363 return true; 4364 } 4365 4366 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4367 unsigned int flags) 4368 { 4369 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4370 4371 dec_page_count(sbi, F2FS_DIO_READ); 4372 if (error) 4373 return error; 4374 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4375 return 0; 4376 } 4377 4378 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4379 .end_io = f2fs_dio_read_end_io, 4380 }; 4381 4382 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4383 { 4384 struct file *file = iocb->ki_filp; 4385 struct inode *inode = file_inode(file); 4386 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4387 struct f2fs_inode_info *fi = F2FS_I(inode); 4388 const loff_t pos = iocb->ki_pos; 4389 const size_t count = iov_iter_count(to); 4390 struct iomap_dio *dio; 4391 ssize_t ret; 4392 4393 if (count == 0) 4394 return 0; /* skip atime update */ 4395 4396 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4397 4398 if (iocb->ki_flags & IOCB_NOWAIT) { 4399 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4400 ret = -EAGAIN; 4401 goto out; 4402 } 4403 } else { 4404 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4405 } 4406 4407 /* 4408 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4409 * the higher-level function iomap_dio_rw() in order to ensure that the 4410 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4411 */ 4412 inc_page_count(sbi, F2FS_DIO_READ); 4413 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4414 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4415 if (IS_ERR_OR_NULL(dio)) { 4416 ret = PTR_ERR_OR_ZERO(dio); 4417 if (ret != -EIOCBQUEUED) 4418 dec_page_count(sbi, F2FS_DIO_READ); 4419 } else { 4420 ret = iomap_dio_complete(dio); 4421 } 4422 4423 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4424 4425 file_accessed(file); 4426 out: 4427 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4428 return ret; 4429 } 4430 4431 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4432 int rw) 4433 { 4434 struct inode *inode = file_inode(file); 4435 char *buf, *path; 4436 4437 buf = f2fs_getname(F2FS_I_SB(inode)); 4438 if (!buf) 4439 return; 4440 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4441 if (IS_ERR(path)) 4442 goto free_buf; 4443 if (rw == WRITE) 4444 trace_f2fs_datawrite_start(inode, pos, count, 4445 current->pid, path, current->comm); 4446 else 4447 trace_f2fs_dataread_start(inode, pos, count, 4448 current->pid, path, current->comm); 4449 free_buf: 4450 f2fs_putname(buf); 4451 } 4452 4453 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4454 { 4455 struct inode *inode = file_inode(iocb->ki_filp); 4456 const loff_t pos = iocb->ki_pos; 4457 ssize_t ret; 4458 4459 if (!f2fs_is_compress_backend_ready(inode)) 4460 return -EOPNOTSUPP; 4461 4462 if (trace_f2fs_dataread_start_enabled()) 4463 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4464 iov_iter_count(to), READ); 4465 4466 if (f2fs_should_use_dio(inode, iocb, to)) { 4467 ret = f2fs_dio_read_iter(iocb, to); 4468 } else { 4469 ret = filemap_read(iocb, to, 0); 4470 if (ret > 0) 4471 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4472 APP_BUFFERED_READ_IO, ret); 4473 } 4474 if (trace_f2fs_dataread_end_enabled()) 4475 trace_f2fs_dataread_end(inode, pos, ret); 4476 return ret; 4477 } 4478 4479 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4480 struct pipe_inode_info *pipe, 4481 size_t len, unsigned int flags) 4482 { 4483 struct inode *inode = file_inode(in); 4484 const loff_t pos = *ppos; 4485 ssize_t ret; 4486 4487 if (!f2fs_is_compress_backend_ready(inode)) 4488 return -EOPNOTSUPP; 4489 4490 if (trace_f2fs_dataread_start_enabled()) 4491 f2fs_trace_rw_file_path(in, pos, len, READ); 4492 4493 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4494 if (ret > 0) 4495 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4496 APP_BUFFERED_READ_IO, ret); 4497 4498 if (trace_f2fs_dataread_end_enabled()) 4499 trace_f2fs_dataread_end(inode, pos, ret); 4500 return ret; 4501 } 4502 4503 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4504 { 4505 struct file *file = iocb->ki_filp; 4506 struct inode *inode = file_inode(file); 4507 ssize_t count; 4508 int err; 4509 4510 if (IS_IMMUTABLE(inode)) 4511 return -EPERM; 4512 4513 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4514 return -EPERM; 4515 4516 count = generic_write_checks(iocb, from); 4517 if (count <= 0) 4518 return count; 4519 4520 err = file_modified(file); 4521 if (err) 4522 return err; 4523 return count; 4524 } 4525 4526 /* 4527 * Preallocate blocks for a write request, if it is possible and helpful to do 4528 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4529 * blocks were preallocated, or a negative errno value if something went 4530 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4531 * requested blocks (not just some of them) have been allocated. 4532 */ 4533 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4534 bool dio) 4535 { 4536 struct inode *inode = file_inode(iocb->ki_filp); 4537 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4538 const loff_t pos = iocb->ki_pos; 4539 const size_t count = iov_iter_count(iter); 4540 struct f2fs_map_blocks map = {}; 4541 int flag; 4542 int ret; 4543 4544 /* If it will be an out-of-place direct write, don't bother. */ 4545 if (dio && f2fs_lfs_mode(sbi)) 4546 return 0; 4547 /* 4548 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4549 * buffered IO, if DIO meets any holes. 4550 */ 4551 if (dio && i_size_read(inode) && 4552 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4553 return 0; 4554 4555 /* No-wait I/O can't allocate blocks. */ 4556 if (iocb->ki_flags & IOCB_NOWAIT) 4557 return 0; 4558 4559 /* If it will be a short write, don't bother. */ 4560 if (fault_in_iov_iter_readable(iter, count)) 4561 return 0; 4562 4563 if (f2fs_has_inline_data(inode)) { 4564 /* If the data will fit inline, don't bother. */ 4565 if (pos + count <= MAX_INLINE_DATA(inode)) 4566 return 0; 4567 ret = f2fs_convert_inline_inode(inode); 4568 if (ret) 4569 return ret; 4570 } 4571 4572 /* Do not preallocate blocks that will be written partially in 4KB. */ 4573 map.m_lblk = F2FS_BLK_ALIGN(pos); 4574 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4575 if (map.m_len > map.m_lblk) 4576 map.m_len -= map.m_lblk; 4577 else 4578 map.m_len = 0; 4579 map.m_may_create = true; 4580 if (dio) { 4581 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4582 flag = F2FS_GET_BLOCK_PRE_DIO; 4583 } else { 4584 map.m_seg_type = NO_CHECK_TYPE; 4585 flag = F2FS_GET_BLOCK_PRE_AIO; 4586 } 4587 4588 ret = f2fs_map_blocks(inode, &map, flag); 4589 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4590 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4591 return ret; 4592 if (ret == 0) 4593 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4594 return map.m_len; 4595 } 4596 4597 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4598 struct iov_iter *from) 4599 { 4600 struct file *file = iocb->ki_filp; 4601 struct inode *inode = file_inode(file); 4602 ssize_t ret; 4603 4604 if (iocb->ki_flags & IOCB_NOWAIT) 4605 return -EOPNOTSUPP; 4606 4607 ret = generic_perform_write(iocb, from); 4608 4609 if (ret > 0) { 4610 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4611 APP_BUFFERED_IO, ret); 4612 } 4613 return ret; 4614 } 4615 4616 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4617 unsigned int flags) 4618 { 4619 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4620 4621 dec_page_count(sbi, F2FS_DIO_WRITE); 4622 if (error) 4623 return error; 4624 f2fs_update_time(sbi, REQ_TIME); 4625 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4626 return 0; 4627 } 4628 4629 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4630 .end_io = f2fs_dio_write_end_io, 4631 }; 4632 4633 static void f2fs_flush_buffered_write(struct address_space *mapping, 4634 loff_t start_pos, loff_t end_pos) 4635 { 4636 int ret; 4637 4638 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4639 if (ret < 0) 4640 return; 4641 invalidate_mapping_pages(mapping, 4642 start_pos >> PAGE_SHIFT, 4643 end_pos >> PAGE_SHIFT); 4644 } 4645 4646 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4647 bool *may_need_sync) 4648 { 4649 struct file *file = iocb->ki_filp; 4650 struct inode *inode = file_inode(file); 4651 struct f2fs_inode_info *fi = F2FS_I(inode); 4652 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4653 const bool do_opu = f2fs_lfs_mode(sbi); 4654 const loff_t pos = iocb->ki_pos; 4655 const ssize_t count = iov_iter_count(from); 4656 unsigned int dio_flags; 4657 struct iomap_dio *dio; 4658 ssize_t ret; 4659 4660 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4661 4662 if (iocb->ki_flags & IOCB_NOWAIT) { 4663 /* f2fs_convert_inline_inode() and block allocation can block */ 4664 if (f2fs_has_inline_data(inode) || 4665 !f2fs_overwrite_io(inode, pos, count)) { 4666 ret = -EAGAIN; 4667 goto out; 4668 } 4669 4670 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4671 ret = -EAGAIN; 4672 goto out; 4673 } 4674 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4675 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4676 ret = -EAGAIN; 4677 goto out; 4678 } 4679 } else { 4680 ret = f2fs_convert_inline_inode(inode); 4681 if (ret) 4682 goto out; 4683 4684 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4685 if (do_opu) 4686 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4687 } 4688 4689 /* 4690 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4691 * the higher-level function iomap_dio_rw() in order to ensure that the 4692 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4693 */ 4694 inc_page_count(sbi, F2FS_DIO_WRITE); 4695 dio_flags = 0; 4696 if (pos + count > inode->i_size) 4697 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4698 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4699 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4700 if (IS_ERR_OR_NULL(dio)) { 4701 ret = PTR_ERR_OR_ZERO(dio); 4702 if (ret == -ENOTBLK) 4703 ret = 0; 4704 if (ret != -EIOCBQUEUED) 4705 dec_page_count(sbi, F2FS_DIO_WRITE); 4706 } else { 4707 ret = iomap_dio_complete(dio); 4708 } 4709 4710 if (do_opu) 4711 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4712 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4713 4714 if (ret < 0) 4715 goto out; 4716 if (pos + ret > inode->i_size) 4717 f2fs_i_size_write(inode, pos + ret); 4718 if (!do_opu) 4719 set_inode_flag(inode, FI_UPDATE_WRITE); 4720 4721 if (iov_iter_count(from)) { 4722 ssize_t ret2; 4723 loff_t bufio_start_pos = iocb->ki_pos; 4724 4725 /* 4726 * The direct write was partial, so we need to fall back to a 4727 * buffered write for the remainder. 4728 */ 4729 4730 ret2 = f2fs_buffered_write_iter(iocb, from); 4731 if (iov_iter_count(from)) 4732 f2fs_write_failed(inode, iocb->ki_pos); 4733 if (ret2 < 0) 4734 goto out; 4735 4736 /* 4737 * Ensure that the pagecache pages are written to disk and 4738 * invalidated to preserve the expected O_DIRECT semantics. 4739 */ 4740 if (ret2 > 0) { 4741 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4742 4743 ret += ret2; 4744 4745 f2fs_flush_buffered_write(file->f_mapping, 4746 bufio_start_pos, 4747 bufio_end_pos); 4748 } 4749 } else { 4750 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4751 *may_need_sync = false; 4752 } 4753 out: 4754 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4755 return ret; 4756 } 4757 4758 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4759 { 4760 struct inode *inode = file_inode(iocb->ki_filp); 4761 const loff_t orig_pos = iocb->ki_pos; 4762 const size_t orig_count = iov_iter_count(from); 4763 loff_t target_size; 4764 bool dio; 4765 bool may_need_sync = true; 4766 int preallocated; 4767 ssize_t ret; 4768 4769 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4770 ret = -EIO; 4771 goto out; 4772 } 4773 4774 if (!f2fs_is_compress_backend_ready(inode)) { 4775 ret = -EOPNOTSUPP; 4776 goto out; 4777 } 4778 4779 if (iocb->ki_flags & IOCB_NOWAIT) { 4780 if (!inode_trylock(inode)) { 4781 ret = -EAGAIN; 4782 goto out; 4783 } 4784 } else { 4785 inode_lock(inode); 4786 } 4787 4788 ret = f2fs_write_checks(iocb, from); 4789 if (ret <= 0) 4790 goto out_unlock; 4791 4792 /* Determine whether we will do a direct write or a buffered write. */ 4793 dio = f2fs_should_use_dio(inode, iocb, from); 4794 4795 /* Possibly preallocate the blocks for the write. */ 4796 target_size = iocb->ki_pos + iov_iter_count(from); 4797 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4798 if (preallocated < 0) { 4799 ret = preallocated; 4800 } else { 4801 if (trace_f2fs_datawrite_start_enabled()) 4802 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4803 orig_count, WRITE); 4804 4805 /* Do the actual write. */ 4806 ret = dio ? 4807 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4808 f2fs_buffered_write_iter(iocb, from); 4809 4810 if (trace_f2fs_datawrite_end_enabled()) 4811 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4812 } 4813 4814 /* Don't leave any preallocated blocks around past i_size. */ 4815 if (preallocated && i_size_read(inode) < target_size) { 4816 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4817 filemap_invalidate_lock(inode->i_mapping); 4818 if (!f2fs_truncate(inode)) 4819 file_dont_truncate(inode); 4820 filemap_invalidate_unlock(inode->i_mapping); 4821 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4822 } else { 4823 file_dont_truncate(inode); 4824 } 4825 4826 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4827 out_unlock: 4828 inode_unlock(inode); 4829 out: 4830 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4831 4832 if (ret > 0 && may_need_sync) 4833 ret = generic_write_sync(iocb, ret); 4834 4835 /* If buffered IO was forced, flush and drop the data from 4836 * the page cache to preserve O_DIRECT semantics 4837 */ 4838 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4839 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4840 orig_pos, 4841 orig_pos + ret - 1); 4842 4843 return ret; 4844 } 4845 4846 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4847 int advice) 4848 { 4849 struct address_space *mapping; 4850 struct backing_dev_info *bdi; 4851 struct inode *inode = file_inode(filp); 4852 int err; 4853 4854 if (advice == POSIX_FADV_SEQUENTIAL) { 4855 if (S_ISFIFO(inode->i_mode)) 4856 return -ESPIPE; 4857 4858 mapping = filp->f_mapping; 4859 if (!mapping || len < 0) 4860 return -EINVAL; 4861 4862 bdi = inode_to_bdi(mapping->host); 4863 filp->f_ra.ra_pages = bdi->ra_pages * 4864 F2FS_I_SB(inode)->seq_file_ra_mul; 4865 spin_lock(&filp->f_lock); 4866 filp->f_mode &= ~FMODE_RANDOM; 4867 spin_unlock(&filp->f_lock); 4868 return 0; 4869 } 4870 4871 err = generic_fadvise(filp, offset, len, advice); 4872 if (!err && advice == POSIX_FADV_DONTNEED && 4873 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4874 f2fs_compressed_file(inode)) 4875 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4876 4877 return err; 4878 } 4879 4880 #ifdef CONFIG_COMPAT 4881 struct compat_f2fs_gc_range { 4882 u32 sync; 4883 compat_u64 start; 4884 compat_u64 len; 4885 }; 4886 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4887 struct compat_f2fs_gc_range) 4888 4889 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4890 { 4891 struct compat_f2fs_gc_range __user *urange; 4892 struct f2fs_gc_range range; 4893 int err; 4894 4895 urange = compat_ptr(arg); 4896 err = get_user(range.sync, &urange->sync); 4897 err |= get_user(range.start, &urange->start); 4898 err |= get_user(range.len, &urange->len); 4899 if (err) 4900 return -EFAULT; 4901 4902 return __f2fs_ioc_gc_range(file, &range); 4903 } 4904 4905 struct compat_f2fs_move_range { 4906 u32 dst_fd; 4907 compat_u64 pos_in; 4908 compat_u64 pos_out; 4909 compat_u64 len; 4910 }; 4911 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4912 struct compat_f2fs_move_range) 4913 4914 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4915 { 4916 struct compat_f2fs_move_range __user *urange; 4917 struct f2fs_move_range range; 4918 int err; 4919 4920 urange = compat_ptr(arg); 4921 err = get_user(range.dst_fd, &urange->dst_fd); 4922 err |= get_user(range.pos_in, &urange->pos_in); 4923 err |= get_user(range.pos_out, &urange->pos_out); 4924 err |= get_user(range.len, &urange->len); 4925 if (err) 4926 return -EFAULT; 4927 4928 return __f2fs_ioc_move_range(file, &range); 4929 } 4930 4931 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4932 { 4933 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4934 return -EIO; 4935 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4936 return -ENOSPC; 4937 4938 switch (cmd) { 4939 case FS_IOC32_GETVERSION: 4940 cmd = FS_IOC_GETVERSION; 4941 break; 4942 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4943 return f2fs_compat_ioc_gc_range(file, arg); 4944 case F2FS_IOC32_MOVE_RANGE: 4945 return f2fs_compat_ioc_move_range(file, arg); 4946 case F2FS_IOC_START_ATOMIC_WRITE: 4947 case F2FS_IOC_START_ATOMIC_REPLACE: 4948 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4949 case F2FS_IOC_START_VOLATILE_WRITE: 4950 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4951 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4952 case F2FS_IOC_SHUTDOWN: 4953 case FITRIM: 4954 case FS_IOC_SET_ENCRYPTION_POLICY: 4955 case FS_IOC_GET_ENCRYPTION_PWSALT: 4956 case FS_IOC_GET_ENCRYPTION_POLICY: 4957 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4958 case FS_IOC_ADD_ENCRYPTION_KEY: 4959 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4960 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4961 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4962 case FS_IOC_GET_ENCRYPTION_NONCE: 4963 case F2FS_IOC_GARBAGE_COLLECT: 4964 case F2FS_IOC_WRITE_CHECKPOINT: 4965 case F2FS_IOC_DEFRAGMENT: 4966 case F2FS_IOC_FLUSH_DEVICE: 4967 case F2FS_IOC_GET_FEATURES: 4968 case F2FS_IOC_GET_PIN_FILE: 4969 case F2FS_IOC_SET_PIN_FILE: 4970 case F2FS_IOC_PRECACHE_EXTENTS: 4971 case F2FS_IOC_RESIZE_FS: 4972 case FS_IOC_ENABLE_VERITY: 4973 case FS_IOC_MEASURE_VERITY: 4974 case FS_IOC_READ_VERITY_METADATA: 4975 case FS_IOC_GETFSLABEL: 4976 case FS_IOC_SETFSLABEL: 4977 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4978 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4979 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4980 case F2FS_IOC_SEC_TRIM_FILE: 4981 case F2FS_IOC_GET_COMPRESS_OPTION: 4982 case F2FS_IOC_SET_COMPRESS_OPTION: 4983 case F2FS_IOC_DECOMPRESS_FILE: 4984 case F2FS_IOC_COMPRESS_FILE: 4985 break; 4986 default: 4987 return -ENOIOCTLCMD; 4988 } 4989 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4990 } 4991 #endif 4992 4993 const struct file_operations f2fs_file_operations = { 4994 .llseek = f2fs_llseek, 4995 .read_iter = f2fs_file_read_iter, 4996 .write_iter = f2fs_file_write_iter, 4997 .iopoll = iocb_bio_iopoll, 4998 .open = f2fs_file_open, 4999 .release = f2fs_release_file, 5000 .mmap = f2fs_file_mmap, 5001 .flush = f2fs_file_flush, 5002 .fsync = f2fs_sync_file, 5003 .fallocate = f2fs_fallocate, 5004 .unlocked_ioctl = f2fs_ioctl, 5005 #ifdef CONFIG_COMPAT 5006 .compat_ioctl = f2fs_compat_ioctl, 5007 #endif 5008 .splice_read = f2fs_file_splice_read, 5009 .splice_write = iter_file_splice_write, 5010 .fadvise = f2fs_file_fadvise, 5011 }; 5012