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 f2fs_put_page(pdst, 1); 1321 f2fs_put_page(psrc, 1); 1322 1323 ret = f2fs_truncate_hole(src_inode, 1324 src + i, src + i + 1); 1325 if (ret) 1326 return ret; 1327 i++; 1328 } 1329 } 1330 return 0; 1331 } 1332 1333 static int __exchange_data_block(struct inode *src_inode, 1334 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1335 pgoff_t len, bool full) 1336 { 1337 block_t *src_blkaddr; 1338 int *do_replace; 1339 pgoff_t olen; 1340 int ret; 1341 1342 while (len) { 1343 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1344 1345 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1346 array_size(olen, sizeof(block_t)), 1347 GFP_NOFS); 1348 if (!src_blkaddr) 1349 return -ENOMEM; 1350 1351 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1352 array_size(olen, sizeof(int)), 1353 GFP_NOFS); 1354 if (!do_replace) { 1355 kvfree(src_blkaddr); 1356 return -ENOMEM; 1357 } 1358 1359 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1360 do_replace, src, olen); 1361 if (ret) 1362 goto roll_back; 1363 1364 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1365 do_replace, src, dst, olen, full); 1366 if (ret) 1367 goto roll_back; 1368 1369 src += olen; 1370 dst += olen; 1371 len -= olen; 1372 1373 kvfree(src_blkaddr); 1374 kvfree(do_replace); 1375 } 1376 return 0; 1377 1378 roll_back: 1379 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1380 kvfree(src_blkaddr); 1381 kvfree(do_replace); 1382 return ret; 1383 } 1384 1385 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1386 { 1387 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1388 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1389 pgoff_t start = offset >> PAGE_SHIFT; 1390 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1391 int ret; 1392 1393 f2fs_balance_fs(sbi, true); 1394 1395 /* avoid gc operation during block exchange */ 1396 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1397 filemap_invalidate_lock(inode->i_mapping); 1398 1399 f2fs_lock_op(sbi); 1400 f2fs_drop_extent_tree(inode); 1401 truncate_pagecache(inode, offset); 1402 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1403 f2fs_unlock_op(sbi); 1404 1405 filemap_invalidate_unlock(inode->i_mapping); 1406 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1407 return ret; 1408 } 1409 1410 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1411 { 1412 loff_t new_size; 1413 int ret; 1414 1415 if (offset + len >= i_size_read(inode)) 1416 return -EINVAL; 1417 1418 /* collapse range should be aligned to block size of f2fs. */ 1419 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1420 return -EINVAL; 1421 1422 ret = f2fs_convert_inline_inode(inode); 1423 if (ret) 1424 return ret; 1425 1426 /* write out all dirty pages from offset */ 1427 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1428 if (ret) 1429 return ret; 1430 1431 ret = f2fs_do_collapse(inode, offset, len); 1432 if (ret) 1433 return ret; 1434 1435 /* write out all moved pages, if possible */ 1436 filemap_invalidate_lock(inode->i_mapping); 1437 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1438 truncate_pagecache(inode, offset); 1439 1440 new_size = i_size_read(inode) - len; 1441 ret = f2fs_truncate_blocks(inode, new_size, true); 1442 filemap_invalidate_unlock(inode->i_mapping); 1443 if (!ret) 1444 f2fs_i_size_write(inode, new_size); 1445 return ret; 1446 } 1447 1448 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1449 pgoff_t end) 1450 { 1451 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1452 pgoff_t index = start; 1453 unsigned int ofs_in_node = dn->ofs_in_node; 1454 blkcnt_t count = 0; 1455 int ret; 1456 1457 for (; index < end; index++, dn->ofs_in_node++) { 1458 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1459 count++; 1460 } 1461 1462 dn->ofs_in_node = ofs_in_node; 1463 ret = f2fs_reserve_new_blocks(dn, count); 1464 if (ret) 1465 return ret; 1466 1467 dn->ofs_in_node = ofs_in_node; 1468 for (index = start; index < end; index++, dn->ofs_in_node++) { 1469 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1470 /* 1471 * f2fs_reserve_new_blocks will not guarantee entire block 1472 * allocation. 1473 */ 1474 if (dn->data_blkaddr == NULL_ADDR) { 1475 ret = -ENOSPC; 1476 break; 1477 } 1478 1479 if (dn->data_blkaddr == NEW_ADDR) 1480 continue; 1481 1482 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1483 DATA_GENERIC_ENHANCE)) { 1484 ret = -EFSCORRUPTED; 1485 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1486 break; 1487 } 1488 1489 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1490 dn->data_blkaddr = NEW_ADDR; 1491 f2fs_set_data_blkaddr(dn); 1492 } 1493 1494 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1495 f2fs_update_age_extent_cache_range(dn, start, index - start); 1496 1497 return ret; 1498 } 1499 1500 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1501 int mode) 1502 { 1503 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1504 struct address_space *mapping = inode->i_mapping; 1505 pgoff_t index, pg_start, pg_end; 1506 loff_t new_size = i_size_read(inode); 1507 loff_t off_start, off_end; 1508 int ret = 0; 1509 1510 ret = inode_newsize_ok(inode, (len + offset)); 1511 if (ret) 1512 return ret; 1513 1514 ret = f2fs_convert_inline_inode(inode); 1515 if (ret) 1516 return ret; 1517 1518 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1519 if (ret) 1520 return ret; 1521 1522 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1523 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1524 1525 off_start = offset & (PAGE_SIZE - 1); 1526 off_end = (offset + len) & (PAGE_SIZE - 1); 1527 1528 if (pg_start == pg_end) { 1529 ret = fill_zero(inode, pg_start, off_start, 1530 off_end - off_start); 1531 if (ret) 1532 return ret; 1533 1534 new_size = max_t(loff_t, new_size, offset + len); 1535 } else { 1536 if (off_start) { 1537 ret = fill_zero(inode, pg_start++, off_start, 1538 PAGE_SIZE - off_start); 1539 if (ret) 1540 return ret; 1541 1542 new_size = max_t(loff_t, new_size, 1543 (loff_t)pg_start << PAGE_SHIFT); 1544 } 1545 1546 for (index = pg_start; index < pg_end;) { 1547 struct dnode_of_data dn; 1548 unsigned int end_offset; 1549 pgoff_t end; 1550 1551 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1552 filemap_invalidate_lock(mapping); 1553 1554 truncate_pagecache_range(inode, 1555 (loff_t)index << PAGE_SHIFT, 1556 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1557 1558 f2fs_lock_op(sbi); 1559 1560 set_new_dnode(&dn, inode, NULL, NULL, 0); 1561 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1562 if (ret) { 1563 f2fs_unlock_op(sbi); 1564 filemap_invalidate_unlock(mapping); 1565 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1566 goto out; 1567 } 1568 1569 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1570 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1571 1572 ret = f2fs_do_zero_range(&dn, index, end); 1573 f2fs_put_dnode(&dn); 1574 1575 f2fs_unlock_op(sbi); 1576 filemap_invalidate_unlock(mapping); 1577 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1578 1579 f2fs_balance_fs(sbi, dn.node_changed); 1580 1581 if (ret) 1582 goto out; 1583 1584 index = end; 1585 new_size = max_t(loff_t, new_size, 1586 (loff_t)index << PAGE_SHIFT); 1587 } 1588 1589 if (off_end) { 1590 ret = fill_zero(inode, pg_end, 0, off_end); 1591 if (ret) 1592 goto out; 1593 1594 new_size = max_t(loff_t, new_size, offset + len); 1595 } 1596 } 1597 1598 out: 1599 if (new_size > i_size_read(inode)) { 1600 if (mode & FALLOC_FL_KEEP_SIZE) 1601 file_set_keep_isize(inode); 1602 else 1603 f2fs_i_size_write(inode, new_size); 1604 } 1605 return ret; 1606 } 1607 1608 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1609 { 1610 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1611 struct address_space *mapping = inode->i_mapping; 1612 pgoff_t nr, pg_start, pg_end, delta, idx; 1613 loff_t new_size; 1614 int ret = 0; 1615 1616 new_size = i_size_read(inode) + len; 1617 ret = inode_newsize_ok(inode, new_size); 1618 if (ret) 1619 return ret; 1620 1621 if (offset >= i_size_read(inode)) 1622 return -EINVAL; 1623 1624 /* insert range should be aligned to block size of f2fs. */ 1625 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1626 return -EINVAL; 1627 1628 ret = f2fs_convert_inline_inode(inode); 1629 if (ret) 1630 return ret; 1631 1632 f2fs_balance_fs(sbi, true); 1633 1634 filemap_invalidate_lock(mapping); 1635 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1636 filemap_invalidate_unlock(mapping); 1637 if (ret) 1638 return ret; 1639 1640 /* write out all dirty pages from offset */ 1641 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1642 if (ret) 1643 return ret; 1644 1645 pg_start = offset >> PAGE_SHIFT; 1646 pg_end = (offset + len) >> PAGE_SHIFT; 1647 delta = pg_end - pg_start; 1648 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1649 1650 /* avoid gc operation during block exchange */ 1651 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1652 filemap_invalidate_lock(mapping); 1653 truncate_pagecache(inode, offset); 1654 1655 while (!ret && idx > pg_start) { 1656 nr = idx - pg_start; 1657 if (nr > delta) 1658 nr = delta; 1659 idx -= nr; 1660 1661 f2fs_lock_op(sbi); 1662 f2fs_drop_extent_tree(inode); 1663 1664 ret = __exchange_data_block(inode, inode, idx, 1665 idx + delta, nr, false); 1666 f2fs_unlock_op(sbi); 1667 } 1668 filemap_invalidate_unlock(mapping); 1669 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1670 1671 /* write out all moved pages, if possible */ 1672 filemap_invalidate_lock(mapping); 1673 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1674 truncate_pagecache(inode, offset); 1675 filemap_invalidate_unlock(mapping); 1676 1677 if (!ret) 1678 f2fs_i_size_write(inode, new_size); 1679 return ret; 1680 } 1681 1682 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1683 loff_t len, int mode) 1684 { 1685 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1686 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1687 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1688 .m_may_create = true }; 1689 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1690 .init_gc_type = FG_GC, 1691 .should_migrate_blocks = false, 1692 .err_gc_skipped = true, 1693 .nr_free_secs = 0 }; 1694 pgoff_t pg_start, pg_end; 1695 loff_t new_size; 1696 loff_t off_end; 1697 block_t expanded = 0; 1698 int err; 1699 1700 err = inode_newsize_ok(inode, (len + offset)); 1701 if (err) 1702 return err; 1703 1704 err = f2fs_convert_inline_inode(inode); 1705 if (err) 1706 return err; 1707 1708 f2fs_balance_fs(sbi, true); 1709 1710 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1711 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1712 off_end = (offset + len) & (PAGE_SIZE - 1); 1713 1714 map.m_lblk = pg_start; 1715 map.m_len = pg_end - pg_start; 1716 if (off_end) 1717 map.m_len++; 1718 1719 if (!map.m_len) 1720 return 0; 1721 1722 if (f2fs_is_pinned_file(inode)) { 1723 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1724 block_t sec_len = roundup(map.m_len, sec_blks); 1725 1726 map.m_len = sec_blks; 1727 next_alloc: 1728 if (has_not_enough_free_secs(sbi, 0, 1729 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1730 f2fs_down_write(&sbi->gc_lock); 1731 stat_inc_gc_call_count(sbi, FOREGROUND); 1732 err = f2fs_gc(sbi, &gc_control); 1733 if (err && err != -ENODATA) 1734 goto out_err; 1735 } 1736 1737 f2fs_down_write(&sbi->pin_sem); 1738 1739 f2fs_lock_op(sbi); 1740 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false); 1741 f2fs_unlock_op(sbi); 1742 1743 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1744 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1745 file_dont_truncate(inode); 1746 1747 f2fs_up_write(&sbi->pin_sem); 1748 1749 expanded += map.m_len; 1750 sec_len -= map.m_len; 1751 map.m_lblk += map.m_len; 1752 if (!err && sec_len) 1753 goto next_alloc; 1754 1755 map.m_len = expanded; 1756 } else { 1757 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1758 expanded = map.m_len; 1759 } 1760 out_err: 1761 if (err) { 1762 pgoff_t last_off; 1763 1764 if (!expanded) 1765 return err; 1766 1767 last_off = pg_start + expanded - 1; 1768 1769 /* update new size to the failed position */ 1770 new_size = (last_off == pg_end) ? offset + len : 1771 (loff_t)(last_off + 1) << PAGE_SHIFT; 1772 } else { 1773 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1774 } 1775 1776 if (new_size > i_size_read(inode)) { 1777 if (mode & FALLOC_FL_KEEP_SIZE) 1778 file_set_keep_isize(inode); 1779 else 1780 f2fs_i_size_write(inode, new_size); 1781 } 1782 1783 return err; 1784 } 1785 1786 static long f2fs_fallocate(struct file *file, int mode, 1787 loff_t offset, loff_t len) 1788 { 1789 struct inode *inode = file_inode(file); 1790 long ret = 0; 1791 1792 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1793 return -EIO; 1794 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1795 return -ENOSPC; 1796 if (!f2fs_is_compress_backend_ready(inode)) 1797 return -EOPNOTSUPP; 1798 1799 /* f2fs only support ->fallocate for regular file */ 1800 if (!S_ISREG(inode->i_mode)) 1801 return -EINVAL; 1802 1803 if (IS_ENCRYPTED(inode) && 1804 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1805 return -EOPNOTSUPP; 1806 1807 /* 1808 * Pinned file should not support partial truncation since the block 1809 * can be used by applications. 1810 */ 1811 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1812 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1813 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1814 return -EOPNOTSUPP; 1815 1816 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1817 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1818 FALLOC_FL_INSERT_RANGE)) 1819 return -EOPNOTSUPP; 1820 1821 inode_lock(inode); 1822 1823 ret = file_modified(file); 1824 if (ret) 1825 goto out; 1826 1827 if (mode & FALLOC_FL_PUNCH_HOLE) { 1828 if (offset >= inode->i_size) 1829 goto out; 1830 1831 ret = f2fs_punch_hole(inode, offset, len); 1832 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1833 ret = f2fs_collapse_range(inode, offset, len); 1834 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1835 ret = f2fs_zero_range(inode, offset, len, mode); 1836 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1837 ret = f2fs_insert_range(inode, offset, len); 1838 } else { 1839 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1840 } 1841 1842 if (!ret) { 1843 inode->i_mtime = inode_set_ctime_current(inode); 1844 f2fs_mark_inode_dirty_sync(inode, false); 1845 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1846 } 1847 1848 out: 1849 inode_unlock(inode); 1850 1851 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1852 return ret; 1853 } 1854 1855 static int f2fs_release_file(struct inode *inode, struct file *filp) 1856 { 1857 /* 1858 * f2fs_release_file is called at every close calls. So we should 1859 * not drop any inmemory pages by close called by other process. 1860 */ 1861 if (!(filp->f_mode & FMODE_WRITE) || 1862 atomic_read(&inode->i_writecount) != 1) 1863 return 0; 1864 1865 inode_lock(inode); 1866 f2fs_abort_atomic_write(inode, true); 1867 inode_unlock(inode); 1868 1869 return 0; 1870 } 1871 1872 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1873 { 1874 struct inode *inode = file_inode(file); 1875 1876 /* 1877 * If the process doing a transaction is crashed, we should do 1878 * roll-back. Otherwise, other reader/write can see corrupted database 1879 * until all the writers close its file. Since this should be done 1880 * before dropping file lock, it needs to do in ->flush. 1881 */ 1882 if (F2FS_I(inode)->atomic_write_task == current && 1883 (current->flags & PF_EXITING)) { 1884 inode_lock(inode); 1885 f2fs_abort_atomic_write(inode, true); 1886 inode_unlock(inode); 1887 } 1888 1889 return 0; 1890 } 1891 1892 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1893 { 1894 struct f2fs_inode_info *fi = F2FS_I(inode); 1895 u32 masked_flags = fi->i_flags & mask; 1896 1897 /* mask can be shrunk by flags_valid selector */ 1898 iflags &= mask; 1899 1900 /* Is it quota file? Do not allow user to mess with it */ 1901 if (IS_NOQUOTA(inode)) 1902 return -EPERM; 1903 1904 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1905 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1906 return -EOPNOTSUPP; 1907 if (!f2fs_empty_dir(inode)) 1908 return -ENOTEMPTY; 1909 } 1910 1911 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1912 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1913 return -EOPNOTSUPP; 1914 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1915 return -EINVAL; 1916 } 1917 1918 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1919 if (masked_flags & F2FS_COMPR_FL) { 1920 if (!f2fs_disable_compressed_file(inode)) 1921 return -EINVAL; 1922 } else { 1923 /* try to convert inline_data to support compression */ 1924 int err = f2fs_convert_inline_inode(inode); 1925 if (err) 1926 return err; 1927 1928 f2fs_down_write(&F2FS_I(inode)->i_sem); 1929 if (!f2fs_may_compress(inode) || 1930 (S_ISREG(inode->i_mode) && 1931 F2FS_HAS_BLOCKS(inode))) { 1932 f2fs_up_write(&F2FS_I(inode)->i_sem); 1933 return -EINVAL; 1934 } 1935 err = set_compress_context(inode); 1936 f2fs_up_write(&F2FS_I(inode)->i_sem); 1937 1938 if (err) 1939 return err; 1940 } 1941 } 1942 1943 fi->i_flags = iflags | (fi->i_flags & ~mask); 1944 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1945 (fi->i_flags & F2FS_NOCOMP_FL)); 1946 1947 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1948 set_inode_flag(inode, FI_PROJ_INHERIT); 1949 else 1950 clear_inode_flag(inode, FI_PROJ_INHERIT); 1951 1952 inode_set_ctime_current(inode); 1953 f2fs_set_inode_flags(inode); 1954 f2fs_mark_inode_dirty_sync(inode, true); 1955 return 0; 1956 } 1957 1958 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1959 1960 /* 1961 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1962 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1963 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1964 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1965 * 1966 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1967 * FS_IOC_FSSETXATTR is done by the VFS. 1968 */ 1969 1970 static const struct { 1971 u32 iflag; 1972 u32 fsflag; 1973 } f2fs_fsflags_map[] = { 1974 { F2FS_COMPR_FL, FS_COMPR_FL }, 1975 { F2FS_SYNC_FL, FS_SYNC_FL }, 1976 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1977 { F2FS_APPEND_FL, FS_APPEND_FL }, 1978 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1979 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1980 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1981 { F2FS_INDEX_FL, FS_INDEX_FL }, 1982 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1983 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1984 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1985 }; 1986 1987 #define F2FS_GETTABLE_FS_FL ( \ 1988 FS_COMPR_FL | \ 1989 FS_SYNC_FL | \ 1990 FS_IMMUTABLE_FL | \ 1991 FS_APPEND_FL | \ 1992 FS_NODUMP_FL | \ 1993 FS_NOATIME_FL | \ 1994 FS_NOCOMP_FL | \ 1995 FS_INDEX_FL | \ 1996 FS_DIRSYNC_FL | \ 1997 FS_PROJINHERIT_FL | \ 1998 FS_ENCRYPT_FL | \ 1999 FS_INLINE_DATA_FL | \ 2000 FS_NOCOW_FL | \ 2001 FS_VERITY_FL | \ 2002 FS_CASEFOLD_FL) 2003 2004 #define F2FS_SETTABLE_FS_FL ( \ 2005 FS_COMPR_FL | \ 2006 FS_SYNC_FL | \ 2007 FS_IMMUTABLE_FL | \ 2008 FS_APPEND_FL | \ 2009 FS_NODUMP_FL | \ 2010 FS_NOATIME_FL | \ 2011 FS_NOCOMP_FL | \ 2012 FS_DIRSYNC_FL | \ 2013 FS_PROJINHERIT_FL | \ 2014 FS_CASEFOLD_FL) 2015 2016 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2017 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2018 { 2019 u32 fsflags = 0; 2020 int i; 2021 2022 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2023 if (iflags & f2fs_fsflags_map[i].iflag) 2024 fsflags |= f2fs_fsflags_map[i].fsflag; 2025 2026 return fsflags; 2027 } 2028 2029 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2030 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2031 { 2032 u32 iflags = 0; 2033 int i; 2034 2035 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2036 if (fsflags & f2fs_fsflags_map[i].fsflag) 2037 iflags |= f2fs_fsflags_map[i].iflag; 2038 2039 return iflags; 2040 } 2041 2042 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2043 { 2044 struct inode *inode = file_inode(filp); 2045 2046 return put_user(inode->i_generation, (int __user *)arg); 2047 } 2048 2049 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2050 { 2051 struct inode *inode = file_inode(filp); 2052 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2053 struct f2fs_inode_info *fi = F2FS_I(inode); 2054 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2055 struct inode *pinode; 2056 loff_t isize; 2057 int ret; 2058 2059 if (!inode_owner_or_capable(idmap, inode)) 2060 return -EACCES; 2061 2062 if (!S_ISREG(inode->i_mode)) 2063 return -EINVAL; 2064 2065 if (filp->f_flags & O_DIRECT) 2066 return -EINVAL; 2067 2068 ret = mnt_want_write_file(filp); 2069 if (ret) 2070 return ret; 2071 2072 inode_lock(inode); 2073 2074 if (!f2fs_disable_compressed_file(inode)) { 2075 ret = -EINVAL; 2076 goto out; 2077 } 2078 2079 if (f2fs_is_atomic_file(inode)) 2080 goto out; 2081 2082 ret = f2fs_convert_inline_inode(inode); 2083 if (ret) 2084 goto out; 2085 2086 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2087 2088 /* 2089 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2090 * f2fs_is_atomic_file. 2091 */ 2092 if (get_dirty_pages(inode)) 2093 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2094 inode->i_ino, get_dirty_pages(inode)); 2095 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2096 if (ret) { 2097 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2098 goto out; 2099 } 2100 2101 /* Check if the inode already has a COW inode */ 2102 if (fi->cow_inode == NULL) { 2103 /* Create a COW inode for atomic write */ 2104 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2105 if (IS_ERR(pinode)) { 2106 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2107 ret = PTR_ERR(pinode); 2108 goto out; 2109 } 2110 2111 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2112 iput(pinode); 2113 if (ret) { 2114 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2115 goto out; 2116 } 2117 2118 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2119 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2120 } else { 2121 /* Reuse the already created COW inode */ 2122 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2123 if (ret) { 2124 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2125 goto out; 2126 } 2127 } 2128 2129 f2fs_write_inode(inode, NULL); 2130 2131 stat_inc_atomic_inode(inode); 2132 2133 set_inode_flag(inode, FI_ATOMIC_FILE); 2134 2135 isize = i_size_read(inode); 2136 fi->original_i_size = isize; 2137 if (truncate) { 2138 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2139 truncate_inode_pages_final(inode->i_mapping); 2140 f2fs_i_size_write(inode, 0); 2141 isize = 0; 2142 } 2143 f2fs_i_size_write(fi->cow_inode, isize); 2144 2145 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2146 2147 f2fs_update_time(sbi, REQ_TIME); 2148 fi->atomic_write_task = current; 2149 stat_update_max_atomic_write(inode); 2150 fi->atomic_write_cnt = 0; 2151 out: 2152 inode_unlock(inode); 2153 mnt_drop_write_file(filp); 2154 return ret; 2155 } 2156 2157 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2158 { 2159 struct inode *inode = file_inode(filp); 2160 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2161 int ret; 2162 2163 if (!inode_owner_or_capable(idmap, inode)) 2164 return -EACCES; 2165 2166 ret = mnt_want_write_file(filp); 2167 if (ret) 2168 return ret; 2169 2170 f2fs_balance_fs(F2FS_I_SB(inode), true); 2171 2172 inode_lock(inode); 2173 2174 if (f2fs_is_atomic_file(inode)) { 2175 ret = f2fs_commit_atomic_write(inode); 2176 if (!ret) 2177 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2178 2179 f2fs_abort_atomic_write(inode, ret); 2180 } else { 2181 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2182 } 2183 2184 inode_unlock(inode); 2185 mnt_drop_write_file(filp); 2186 return ret; 2187 } 2188 2189 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2190 { 2191 struct inode *inode = file_inode(filp); 2192 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2193 int ret; 2194 2195 if (!inode_owner_or_capable(idmap, inode)) 2196 return -EACCES; 2197 2198 ret = mnt_want_write_file(filp); 2199 if (ret) 2200 return ret; 2201 2202 inode_lock(inode); 2203 2204 f2fs_abort_atomic_write(inode, true); 2205 2206 inode_unlock(inode); 2207 2208 mnt_drop_write_file(filp); 2209 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2210 return ret; 2211 } 2212 2213 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2214 { 2215 struct inode *inode = file_inode(filp); 2216 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2217 struct super_block *sb = sbi->sb; 2218 __u32 in; 2219 int ret = 0; 2220 2221 if (!capable(CAP_SYS_ADMIN)) 2222 return -EPERM; 2223 2224 if (get_user(in, (__u32 __user *)arg)) 2225 return -EFAULT; 2226 2227 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2228 ret = mnt_want_write_file(filp); 2229 if (ret) { 2230 if (ret == -EROFS) { 2231 ret = 0; 2232 f2fs_stop_checkpoint(sbi, false, 2233 STOP_CP_REASON_SHUTDOWN); 2234 trace_f2fs_shutdown(sbi, in, ret); 2235 } 2236 return ret; 2237 } 2238 } 2239 2240 switch (in) { 2241 case F2FS_GOING_DOWN_FULLSYNC: 2242 ret = freeze_bdev(sb->s_bdev); 2243 if (ret) 2244 goto out; 2245 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2246 thaw_bdev(sb->s_bdev); 2247 break; 2248 case F2FS_GOING_DOWN_METASYNC: 2249 /* do checkpoint only */ 2250 ret = f2fs_sync_fs(sb, 1); 2251 if (ret) 2252 goto out; 2253 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2254 break; 2255 case F2FS_GOING_DOWN_NOSYNC: 2256 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2257 break; 2258 case F2FS_GOING_DOWN_METAFLUSH: 2259 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2260 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2261 break; 2262 case F2FS_GOING_DOWN_NEED_FSCK: 2263 set_sbi_flag(sbi, SBI_NEED_FSCK); 2264 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2265 set_sbi_flag(sbi, SBI_IS_DIRTY); 2266 /* do checkpoint only */ 2267 ret = f2fs_sync_fs(sb, 1); 2268 goto out; 2269 default: 2270 ret = -EINVAL; 2271 goto out; 2272 } 2273 2274 f2fs_stop_gc_thread(sbi); 2275 f2fs_stop_discard_thread(sbi); 2276 2277 f2fs_drop_discard_cmd(sbi); 2278 clear_opt(sbi, DISCARD); 2279 2280 f2fs_update_time(sbi, REQ_TIME); 2281 out: 2282 if (in != F2FS_GOING_DOWN_FULLSYNC) 2283 mnt_drop_write_file(filp); 2284 2285 trace_f2fs_shutdown(sbi, in, ret); 2286 2287 return ret; 2288 } 2289 2290 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2291 { 2292 struct inode *inode = file_inode(filp); 2293 struct super_block *sb = inode->i_sb; 2294 struct fstrim_range range; 2295 int ret; 2296 2297 if (!capable(CAP_SYS_ADMIN)) 2298 return -EPERM; 2299 2300 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2301 return -EOPNOTSUPP; 2302 2303 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2304 sizeof(range))) 2305 return -EFAULT; 2306 2307 ret = mnt_want_write_file(filp); 2308 if (ret) 2309 return ret; 2310 2311 range.minlen = max((unsigned int)range.minlen, 2312 bdev_discard_granularity(sb->s_bdev)); 2313 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2314 mnt_drop_write_file(filp); 2315 if (ret < 0) 2316 return ret; 2317 2318 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2319 sizeof(range))) 2320 return -EFAULT; 2321 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2322 return 0; 2323 } 2324 2325 static bool uuid_is_nonzero(__u8 u[16]) 2326 { 2327 int i; 2328 2329 for (i = 0; i < 16; i++) 2330 if (u[i]) 2331 return true; 2332 return false; 2333 } 2334 2335 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2336 { 2337 struct inode *inode = file_inode(filp); 2338 2339 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2340 return -EOPNOTSUPP; 2341 2342 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2343 2344 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2345 } 2346 2347 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2348 { 2349 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2350 return -EOPNOTSUPP; 2351 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2352 } 2353 2354 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2355 { 2356 struct inode *inode = file_inode(filp); 2357 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2358 u8 encrypt_pw_salt[16]; 2359 int err; 2360 2361 if (!f2fs_sb_has_encrypt(sbi)) 2362 return -EOPNOTSUPP; 2363 2364 err = mnt_want_write_file(filp); 2365 if (err) 2366 return err; 2367 2368 f2fs_down_write(&sbi->sb_lock); 2369 2370 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2371 goto got_it; 2372 2373 /* update superblock with uuid */ 2374 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2375 2376 err = f2fs_commit_super(sbi, false); 2377 if (err) { 2378 /* undo new data */ 2379 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2380 goto out_err; 2381 } 2382 got_it: 2383 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2384 out_err: 2385 f2fs_up_write(&sbi->sb_lock); 2386 mnt_drop_write_file(filp); 2387 2388 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2389 err = -EFAULT; 2390 2391 return err; 2392 } 2393 2394 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2395 unsigned long arg) 2396 { 2397 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2398 return -EOPNOTSUPP; 2399 2400 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2401 } 2402 2403 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2404 { 2405 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2406 return -EOPNOTSUPP; 2407 2408 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2409 } 2410 2411 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2412 { 2413 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2414 return -EOPNOTSUPP; 2415 2416 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2417 } 2418 2419 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2420 unsigned long arg) 2421 { 2422 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2423 return -EOPNOTSUPP; 2424 2425 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2426 } 2427 2428 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2429 unsigned long arg) 2430 { 2431 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2432 return -EOPNOTSUPP; 2433 2434 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2435 } 2436 2437 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2438 { 2439 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2440 return -EOPNOTSUPP; 2441 2442 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2443 } 2444 2445 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2446 { 2447 struct inode *inode = file_inode(filp); 2448 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2449 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2450 .no_bg_gc = false, 2451 .should_migrate_blocks = false, 2452 .nr_free_secs = 0 }; 2453 __u32 sync; 2454 int ret; 2455 2456 if (!capable(CAP_SYS_ADMIN)) 2457 return -EPERM; 2458 2459 if (get_user(sync, (__u32 __user *)arg)) 2460 return -EFAULT; 2461 2462 if (f2fs_readonly(sbi->sb)) 2463 return -EROFS; 2464 2465 ret = mnt_want_write_file(filp); 2466 if (ret) 2467 return ret; 2468 2469 if (!sync) { 2470 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2471 ret = -EBUSY; 2472 goto out; 2473 } 2474 } else { 2475 f2fs_down_write(&sbi->gc_lock); 2476 } 2477 2478 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2479 gc_control.err_gc_skipped = sync; 2480 stat_inc_gc_call_count(sbi, FOREGROUND); 2481 ret = f2fs_gc(sbi, &gc_control); 2482 out: 2483 mnt_drop_write_file(filp); 2484 return ret; 2485 } 2486 2487 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2488 { 2489 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2490 struct f2fs_gc_control gc_control = { 2491 .init_gc_type = range->sync ? FG_GC : BG_GC, 2492 .no_bg_gc = false, 2493 .should_migrate_blocks = false, 2494 .err_gc_skipped = range->sync, 2495 .nr_free_secs = 0 }; 2496 u64 end; 2497 int ret; 2498 2499 if (!capable(CAP_SYS_ADMIN)) 2500 return -EPERM; 2501 if (f2fs_readonly(sbi->sb)) 2502 return -EROFS; 2503 2504 end = range->start + range->len; 2505 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2506 end >= MAX_BLKADDR(sbi)) 2507 return -EINVAL; 2508 2509 ret = mnt_want_write_file(filp); 2510 if (ret) 2511 return ret; 2512 2513 do_more: 2514 if (!range->sync) { 2515 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2516 ret = -EBUSY; 2517 goto out; 2518 } 2519 } else { 2520 f2fs_down_write(&sbi->gc_lock); 2521 } 2522 2523 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2524 stat_inc_gc_call_count(sbi, FOREGROUND); 2525 ret = f2fs_gc(sbi, &gc_control); 2526 if (ret) { 2527 if (ret == -EBUSY) 2528 ret = -EAGAIN; 2529 goto out; 2530 } 2531 range->start += CAP_BLKS_PER_SEC(sbi); 2532 if (range->start <= end) 2533 goto do_more; 2534 out: 2535 mnt_drop_write_file(filp); 2536 return ret; 2537 } 2538 2539 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2540 { 2541 struct f2fs_gc_range range; 2542 2543 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2544 sizeof(range))) 2545 return -EFAULT; 2546 return __f2fs_ioc_gc_range(filp, &range); 2547 } 2548 2549 static int f2fs_ioc_write_checkpoint(struct file *filp) 2550 { 2551 struct inode *inode = file_inode(filp); 2552 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2553 int ret; 2554 2555 if (!capable(CAP_SYS_ADMIN)) 2556 return -EPERM; 2557 2558 if (f2fs_readonly(sbi->sb)) 2559 return -EROFS; 2560 2561 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2562 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2563 return -EINVAL; 2564 } 2565 2566 ret = mnt_want_write_file(filp); 2567 if (ret) 2568 return ret; 2569 2570 ret = f2fs_sync_fs(sbi->sb, 1); 2571 2572 mnt_drop_write_file(filp); 2573 return ret; 2574 } 2575 2576 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2577 struct file *filp, 2578 struct f2fs_defragment *range) 2579 { 2580 struct inode *inode = file_inode(filp); 2581 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2582 .m_seg_type = NO_CHECK_TYPE, 2583 .m_may_create = false }; 2584 struct extent_info ei = {}; 2585 pgoff_t pg_start, pg_end, next_pgofs; 2586 unsigned int blk_per_seg = sbi->blocks_per_seg; 2587 unsigned int total = 0, sec_num; 2588 block_t blk_end = 0; 2589 bool fragmented = false; 2590 int err; 2591 2592 pg_start = range->start >> PAGE_SHIFT; 2593 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2594 2595 f2fs_balance_fs(sbi, true); 2596 2597 inode_lock(inode); 2598 2599 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2600 err = -EINVAL; 2601 goto unlock_out; 2602 } 2603 2604 /* if in-place-update policy is enabled, don't waste time here */ 2605 set_inode_flag(inode, FI_OPU_WRITE); 2606 if (f2fs_should_update_inplace(inode, NULL)) { 2607 err = -EINVAL; 2608 goto out; 2609 } 2610 2611 /* writeback all dirty pages in the range */ 2612 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2613 range->start + range->len - 1); 2614 if (err) 2615 goto out; 2616 2617 /* 2618 * lookup mapping info in extent cache, skip defragmenting if physical 2619 * block addresses are continuous. 2620 */ 2621 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2622 if (ei.fofs + ei.len >= pg_end) 2623 goto out; 2624 } 2625 2626 map.m_lblk = pg_start; 2627 map.m_next_pgofs = &next_pgofs; 2628 2629 /* 2630 * lookup mapping info in dnode page cache, skip defragmenting if all 2631 * physical block addresses are continuous even if there are hole(s) 2632 * in logical blocks. 2633 */ 2634 while (map.m_lblk < pg_end) { 2635 map.m_len = pg_end - map.m_lblk; 2636 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2637 if (err) 2638 goto out; 2639 2640 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2641 map.m_lblk = next_pgofs; 2642 continue; 2643 } 2644 2645 if (blk_end && blk_end != map.m_pblk) 2646 fragmented = true; 2647 2648 /* record total count of block that we're going to move */ 2649 total += map.m_len; 2650 2651 blk_end = map.m_pblk + map.m_len; 2652 2653 map.m_lblk += map.m_len; 2654 } 2655 2656 if (!fragmented) { 2657 total = 0; 2658 goto out; 2659 } 2660 2661 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2662 2663 /* 2664 * make sure there are enough free section for LFS allocation, this can 2665 * avoid defragment running in SSR mode when free section are allocated 2666 * intensively 2667 */ 2668 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2669 err = -EAGAIN; 2670 goto out; 2671 } 2672 2673 map.m_lblk = pg_start; 2674 map.m_len = pg_end - pg_start; 2675 total = 0; 2676 2677 while (map.m_lblk < pg_end) { 2678 pgoff_t idx; 2679 int cnt = 0; 2680 2681 do_map: 2682 map.m_len = pg_end - map.m_lblk; 2683 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2684 if (err) 2685 goto clear_out; 2686 2687 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2688 map.m_lblk = next_pgofs; 2689 goto check; 2690 } 2691 2692 set_inode_flag(inode, FI_SKIP_WRITES); 2693 2694 idx = map.m_lblk; 2695 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2696 struct page *page; 2697 2698 page = f2fs_get_lock_data_page(inode, idx, true); 2699 if (IS_ERR(page)) { 2700 err = PTR_ERR(page); 2701 goto clear_out; 2702 } 2703 2704 set_page_dirty(page); 2705 set_page_private_gcing(page); 2706 f2fs_put_page(page, 1); 2707 2708 idx++; 2709 cnt++; 2710 total++; 2711 } 2712 2713 map.m_lblk = idx; 2714 check: 2715 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2716 goto do_map; 2717 2718 clear_inode_flag(inode, FI_SKIP_WRITES); 2719 2720 err = filemap_fdatawrite(inode->i_mapping); 2721 if (err) 2722 goto out; 2723 } 2724 clear_out: 2725 clear_inode_flag(inode, FI_SKIP_WRITES); 2726 out: 2727 clear_inode_flag(inode, FI_OPU_WRITE); 2728 unlock_out: 2729 inode_unlock(inode); 2730 if (!err) 2731 range->len = (u64)total << PAGE_SHIFT; 2732 return err; 2733 } 2734 2735 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2736 { 2737 struct inode *inode = file_inode(filp); 2738 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2739 struct f2fs_defragment range; 2740 int err; 2741 2742 if (!capable(CAP_SYS_ADMIN)) 2743 return -EPERM; 2744 2745 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2746 return -EINVAL; 2747 2748 if (f2fs_readonly(sbi->sb)) 2749 return -EROFS; 2750 2751 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2752 sizeof(range))) 2753 return -EFAULT; 2754 2755 /* verify alignment of offset & size */ 2756 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2757 return -EINVAL; 2758 2759 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2760 max_file_blocks(inode))) 2761 return -EINVAL; 2762 2763 err = mnt_want_write_file(filp); 2764 if (err) 2765 return err; 2766 2767 err = f2fs_defragment_range(sbi, filp, &range); 2768 mnt_drop_write_file(filp); 2769 2770 f2fs_update_time(sbi, REQ_TIME); 2771 if (err < 0) 2772 return err; 2773 2774 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2775 sizeof(range))) 2776 return -EFAULT; 2777 2778 return 0; 2779 } 2780 2781 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2782 struct file *file_out, loff_t pos_out, size_t len) 2783 { 2784 struct inode *src = file_inode(file_in); 2785 struct inode *dst = file_inode(file_out); 2786 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2787 size_t olen = len, dst_max_i_size = 0; 2788 size_t dst_osize; 2789 int ret; 2790 2791 if (file_in->f_path.mnt != file_out->f_path.mnt || 2792 src->i_sb != dst->i_sb) 2793 return -EXDEV; 2794 2795 if (unlikely(f2fs_readonly(src->i_sb))) 2796 return -EROFS; 2797 2798 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2799 return -EINVAL; 2800 2801 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2802 return -EOPNOTSUPP; 2803 2804 if (pos_out < 0 || pos_in < 0) 2805 return -EINVAL; 2806 2807 if (src == dst) { 2808 if (pos_in == pos_out) 2809 return 0; 2810 if (pos_out > pos_in && pos_out < pos_in + len) 2811 return -EINVAL; 2812 } 2813 2814 inode_lock(src); 2815 if (src != dst) { 2816 ret = -EBUSY; 2817 if (!inode_trylock(dst)) 2818 goto out; 2819 } 2820 2821 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) { 2822 ret = -EOPNOTSUPP; 2823 goto out_unlock; 2824 } 2825 2826 ret = -EINVAL; 2827 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2828 goto out_unlock; 2829 if (len == 0) 2830 olen = len = src->i_size - pos_in; 2831 if (pos_in + len == src->i_size) 2832 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2833 if (len == 0) { 2834 ret = 0; 2835 goto out_unlock; 2836 } 2837 2838 dst_osize = dst->i_size; 2839 if (pos_out + olen > dst->i_size) 2840 dst_max_i_size = pos_out + olen; 2841 2842 /* verify the end result is block aligned */ 2843 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2844 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2845 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2846 goto out_unlock; 2847 2848 ret = f2fs_convert_inline_inode(src); 2849 if (ret) 2850 goto out_unlock; 2851 2852 ret = f2fs_convert_inline_inode(dst); 2853 if (ret) 2854 goto out_unlock; 2855 2856 /* write out all dirty pages from offset */ 2857 ret = filemap_write_and_wait_range(src->i_mapping, 2858 pos_in, pos_in + len); 2859 if (ret) 2860 goto out_unlock; 2861 2862 ret = filemap_write_and_wait_range(dst->i_mapping, 2863 pos_out, pos_out + len); 2864 if (ret) 2865 goto out_unlock; 2866 2867 f2fs_balance_fs(sbi, true); 2868 2869 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2870 if (src != dst) { 2871 ret = -EBUSY; 2872 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2873 goto out_src; 2874 } 2875 2876 f2fs_lock_op(sbi); 2877 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2878 pos_out >> F2FS_BLKSIZE_BITS, 2879 len >> F2FS_BLKSIZE_BITS, false); 2880 2881 if (!ret) { 2882 if (dst_max_i_size) 2883 f2fs_i_size_write(dst, dst_max_i_size); 2884 else if (dst_osize != dst->i_size) 2885 f2fs_i_size_write(dst, dst_osize); 2886 } 2887 f2fs_unlock_op(sbi); 2888 2889 if (src != dst) 2890 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2891 out_src: 2892 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2893 if (ret) 2894 goto out_unlock; 2895 2896 src->i_mtime = inode_set_ctime_current(src); 2897 f2fs_mark_inode_dirty_sync(src, false); 2898 if (src != dst) { 2899 dst->i_mtime = inode_set_ctime_current(dst); 2900 f2fs_mark_inode_dirty_sync(dst, false); 2901 } 2902 f2fs_update_time(sbi, REQ_TIME); 2903 2904 out_unlock: 2905 if (src != dst) 2906 inode_unlock(dst); 2907 out: 2908 inode_unlock(src); 2909 return ret; 2910 } 2911 2912 static int __f2fs_ioc_move_range(struct file *filp, 2913 struct f2fs_move_range *range) 2914 { 2915 struct fd dst; 2916 int err; 2917 2918 if (!(filp->f_mode & FMODE_READ) || 2919 !(filp->f_mode & FMODE_WRITE)) 2920 return -EBADF; 2921 2922 dst = fdget(range->dst_fd); 2923 if (!dst.file) 2924 return -EBADF; 2925 2926 if (!(dst.file->f_mode & FMODE_WRITE)) { 2927 err = -EBADF; 2928 goto err_out; 2929 } 2930 2931 err = mnt_want_write_file(filp); 2932 if (err) 2933 goto err_out; 2934 2935 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2936 range->pos_out, range->len); 2937 2938 mnt_drop_write_file(filp); 2939 err_out: 2940 fdput(dst); 2941 return err; 2942 } 2943 2944 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2945 { 2946 struct f2fs_move_range range; 2947 2948 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2949 sizeof(range))) 2950 return -EFAULT; 2951 return __f2fs_ioc_move_range(filp, &range); 2952 } 2953 2954 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2955 { 2956 struct inode *inode = file_inode(filp); 2957 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2958 struct sit_info *sm = SIT_I(sbi); 2959 unsigned int start_segno = 0, end_segno = 0; 2960 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2961 struct f2fs_flush_device range; 2962 struct f2fs_gc_control gc_control = { 2963 .init_gc_type = FG_GC, 2964 .should_migrate_blocks = true, 2965 .err_gc_skipped = true, 2966 .nr_free_secs = 0 }; 2967 int ret; 2968 2969 if (!capable(CAP_SYS_ADMIN)) 2970 return -EPERM; 2971 2972 if (f2fs_readonly(sbi->sb)) 2973 return -EROFS; 2974 2975 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2976 return -EINVAL; 2977 2978 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2979 sizeof(range))) 2980 return -EFAULT; 2981 2982 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2983 __is_large_section(sbi)) { 2984 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2985 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2986 return -EINVAL; 2987 } 2988 2989 ret = mnt_want_write_file(filp); 2990 if (ret) 2991 return ret; 2992 2993 if (range.dev_num != 0) 2994 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2995 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2996 2997 start_segno = sm->last_victim[FLUSH_DEVICE]; 2998 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2999 start_segno = dev_start_segno; 3000 end_segno = min(start_segno + range.segments, dev_end_segno); 3001 3002 while (start_segno < end_segno) { 3003 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3004 ret = -EBUSY; 3005 goto out; 3006 } 3007 sm->last_victim[GC_CB] = end_segno + 1; 3008 sm->last_victim[GC_GREEDY] = end_segno + 1; 3009 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3010 3011 gc_control.victim_segno = start_segno; 3012 stat_inc_gc_call_count(sbi, FOREGROUND); 3013 ret = f2fs_gc(sbi, &gc_control); 3014 if (ret == -EAGAIN) 3015 ret = 0; 3016 else if (ret < 0) 3017 break; 3018 start_segno++; 3019 } 3020 out: 3021 mnt_drop_write_file(filp); 3022 return ret; 3023 } 3024 3025 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3026 { 3027 struct inode *inode = file_inode(filp); 3028 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3029 3030 /* Must validate to set it with SQLite behavior in Android. */ 3031 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3032 3033 return put_user(sb_feature, (u32 __user *)arg); 3034 } 3035 3036 #ifdef CONFIG_QUOTA 3037 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3038 { 3039 struct dquot *transfer_to[MAXQUOTAS] = {}; 3040 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3041 struct super_block *sb = sbi->sb; 3042 int err; 3043 3044 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3045 if (IS_ERR(transfer_to[PRJQUOTA])) 3046 return PTR_ERR(transfer_to[PRJQUOTA]); 3047 3048 err = __dquot_transfer(inode, transfer_to); 3049 if (err) 3050 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3051 dqput(transfer_to[PRJQUOTA]); 3052 return err; 3053 } 3054 3055 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3056 { 3057 struct f2fs_inode_info *fi = F2FS_I(inode); 3058 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3059 struct f2fs_inode *ri = NULL; 3060 kprojid_t kprojid; 3061 int err; 3062 3063 if (!f2fs_sb_has_project_quota(sbi)) { 3064 if (projid != F2FS_DEF_PROJID) 3065 return -EOPNOTSUPP; 3066 else 3067 return 0; 3068 } 3069 3070 if (!f2fs_has_extra_attr(inode)) 3071 return -EOPNOTSUPP; 3072 3073 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3074 3075 if (projid_eq(kprojid, fi->i_projid)) 3076 return 0; 3077 3078 err = -EPERM; 3079 /* Is it quota file? Do not allow user to mess with it */ 3080 if (IS_NOQUOTA(inode)) 3081 return err; 3082 3083 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3084 return -EOVERFLOW; 3085 3086 err = f2fs_dquot_initialize(inode); 3087 if (err) 3088 return err; 3089 3090 f2fs_lock_op(sbi); 3091 err = f2fs_transfer_project_quota(inode, kprojid); 3092 if (err) 3093 goto out_unlock; 3094 3095 fi->i_projid = kprojid; 3096 inode_set_ctime_current(inode); 3097 f2fs_mark_inode_dirty_sync(inode, true); 3098 out_unlock: 3099 f2fs_unlock_op(sbi); 3100 return err; 3101 } 3102 #else 3103 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3104 { 3105 return 0; 3106 } 3107 3108 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3109 { 3110 if (projid != F2FS_DEF_PROJID) 3111 return -EOPNOTSUPP; 3112 return 0; 3113 } 3114 #endif 3115 3116 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3117 { 3118 struct inode *inode = d_inode(dentry); 3119 struct f2fs_inode_info *fi = F2FS_I(inode); 3120 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3121 3122 if (IS_ENCRYPTED(inode)) 3123 fsflags |= FS_ENCRYPT_FL; 3124 if (IS_VERITY(inode)) 3125 fsflags |= FS_VERITY_FL; 3126 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3127 fsflags |= FS_INLINE_DATA_FL; 3128 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3129 fsflags |= FS_NOCOW_FL; 3130 3131 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3132 3133 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3134 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3135 3136 return 0; 3137 } 3138 3139 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3140 struct dentry *dentry, struct fileattr *fa) 3141 { 3142 struct inode *inode = d_inode(dentry); 3143 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3144 u32 iflags; 3145 int err; 3146 3147 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3148 return -EIO; 3149 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3150 return -ENOSPC; 3151 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3152 return -EOPNOTSUPP; 3153 fsflags &= F2FS_SETTABLE_FS_FL; 3154 if (!fa->flags_valid) 3155 mask &= FS_COMMON_FL; 3156 3157 iflags = f2fs_fsflags_to_iflags(fsflags); 3158 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3159 return -EOPNOTSUPP; 3160 3161 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3162 if (!err) 3163 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3164 3165 return err; 3166 } 3167 3168 int f2fs_pin_file_control(struct inode *inode, bool inc) 3169 { 3170 struct f2fs_inode_info *fi = F2FS_I(inode); 3171 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3172 3173 /* Use i_gc_failures for normal file as a risk signal. */ 3174 if (inc) 3175 f2fs_i_gc_failures_write(inode, 3176 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3177 3178 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3179 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3180 __func__, inode->i_ino, 3181 fi->i_gc_failures[GC_FAILURE_PIN]); 3182 clear_inode_flag(inode, FI_PIN_FILE); 3183 return -EAGAIN; 3184 } 3185 return 0; 3186 } 3187 3188 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3189 { 3190 struct inode *inode = file_inode(filp); 3191 __u32 pin; 3192 int ret = 0; 3193 3194 if (get_user(pin, (__u32 __user *)arg)) 3195 return -EFAULT; 3196 3197 if (!S_ISREG(inode->i_mode)) 3198 return -EINVAL; 3199 3200 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3201 return -EROFS; 3202 3203 ret = mnt_want_write_file(filp); 3204 if (ret) 3205 return ret; 3206 3207 inode_lock(inode); 3208 3209 if (!pin) { 3210 clear_inode_flag(inode, FI_PIN_FILE); 3211 f2fs_i_gc_failures_write(inode, 0); 3212 goto done; 3213 } 3214 3215 if (f2fs_should_update_outplace(inode, NULL)) { 3216 ret = -EINVAL; 3217 goto out; 3218 } 3219 3220 if (f2fs_pin_file_control(inode, false)) { 3221 ret = -EAGAIN; 3222 goto out; 3223 } 3224 3225 ret = f2fs_convert_inline_inode(inode); 3226 if (ret) 3227 goto out; 3228 3229 if (!f2fs_disable_compressed_file(inode)) { 3230 ret = -EOPNOTSUPP; 3231 goto out; 3232 } 3233 3234 set_inode_flag(inode, FI_PIN_FILE); 3235 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3236 done: 3237 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3238 out: 3239 inode_unlock(inode); 3240 mnt_drop_write_file(filp); 3241 return ret; 3242 } 3243 3244 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3245 { 3246 struct inode *inode = file_inode(filp); 3247 __u32 pin = 0; 3248 3249 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3250 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3251 return put_user(pin, (u32 __user *)arg); 3252 } 3253 3254 int f2fs_precache_extents(struct inode *inode) 3255 { 3256 struct f2fs_inode_info *fi = F2FS_I(inode); 3257 struct f2fs_map_blocks map; 3258 pgoff_t m_next_extent; 3259 loff_t end; 3260 int err; 3261 3262 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3263 return -EOPNOTSUPP; 3264 3265 map.m_lblk = 0; 3266 map.m_pblk = 0; 3267 map.m_next_pgofs = NULL; 3268 map.m_next_extent = &m_next_extent; 3269 map.m_seg_type = NO_CHECK_TYPE; 3270 map.m_may_create = false; 3271 end = max_file_blocks(inode); 3272 3273 while (map.m_lblk < end) { 3274 map.m_len = end - map.m_lblk; 3275 3276 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3277 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3278 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3279 if (err) 3280 return err; 3281 3282 map.m_lblk = m_next_extent; 3283 } 3284 3285 return 0; 3286 } 3287 3288 static int f2fs_ioc_precache_extents(struct file *filp) 3289 { 3290 return f2fs_precache_extents(file_inode(filp)); 3291 } 3292 3293 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3294 { 3295 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3296 __u64 block_count; 3297 3298 if (!capable(CAP_SYS_ADMIN)) 3299 return -EPERM; 3300 3301 if (f2fs_readonly(sbi->sb)) 3302 return -EROFS; 3303 3304 if (copy_from_user(&block_count, (void __user *)arg, 3305 sizeof(block_count))) 3306 return -EFAULT; 3307 3308 return f2fs_resize_fs(filp, block_count); 3309 } 3310 3311 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3312 { 3313 struct inode *inode = file_inode(filp); 3314 3315 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3316 3317 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3318 f2fs_warn(F2FS_I_SB(inode), 3319 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3320 inode->i_ino); 3321 return -EOPNOTSUPP; 3322 } 3323 3324 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3325 } 3326 3327 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3328 { 3329 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3330 return -EOPNOTSUPP; 3331 3332 return fsverity_ioctl_measure(filp, (void __user *)arg); 3333 } 3334 3335 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3336 { 3337 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3338 return -EOPNOTSUPP; 3339 3340 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3341 } 3342 3343 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3344 { 3345 struct inode *inode = file_inode(filp); 3346 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3347 char *vbuf; 3348 int count; 3349 int err = 0; 3350 3351 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3352 if (!vbuf) 3353 return -ENOMEM; 3354 3355 f2fs_down_read(&sbi->sb_lock); 3356 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3357 ARRAY_SIZE(sbi->raw_super->volume_name), 3358 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3359 f2fs_up_read(&sbi->sb_lock); 3360 3361 if (copy_to_user((char __user *)arg, vbuf, 3362 min(FSLABEL_MAX, count))) 3363 err = -EFAULT; 3364 3365 kfree(vbuf); 3366 return err; 3367 } 3368 3369 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3370 { 3371 struct inode *inode = file_inode(filp); 3372 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3373 char *vbuf; 3374 int err = 0; 3375 3376 if (!capable(CAP_SYS_ADMIN)) 3377 return -EPERM; 3378 3379 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3380 if (IS_ERR(vbuf)) 3381 return PTR_ERR(vbuf); 3382 3383 err = mnt_want_write_file(filp); 3384 if (err) 3385 goto out; 3386 3387 f2fs_down_write(&sbi->sb_lock); 3388 3389 memset(sbi->raw_super->volume_name, 0, 3390 sizeof(sbi->raw_super->volume_name)); 3391 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3392 sbi->raw_super->volume_name, 3393 ARRAY_SIZE(sbi->raw_super->volume_name)); 3394 3395 err = f2fs_commit_super(sbi, false); 3396 3397 f2fs_up_write(&sbi->sb_lock); 3398 3399 mnt_drop_write_file(filp); 3400 out: 3401 kfree(vbuf); 3402 return err; 3403 } 3404 3405 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3406 { 3407 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3408 return -EOPNOTSUPP; 3409 3410 if (!f2fs_compressed_file(inode)) 3411 return -EINVAL; 3412 3413 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3414 3415 return 0; 3416 } 3417 3418 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3419 { 3420 struct inode *inode = file_inode(filp); 3421 __u64 blocks; 3422 int ret; 3423 3424 ret = f2fs_get_compress_blocks(inode, &blocks); 3425 if (ret < 0) 3426 return ret; 3427 3428 return put_user(blocks, (u64 __user *)arg); 3429 } 3430 3431 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3432 { 3433 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3434 unsigned int released_blocks = 0; 3435 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3436 block_t blkaddr; 3437 int i; 3438 3439 for (i = 0; i < count; i++) { 3440 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3441 dn->ofs_in_node + i); 3442 3443 if (!__is_valid_data_blkaddr(blkaddr)) 3444 continue; 3445 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3446 DATA_GENERIC_ENHANCE))) { 3447 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3448 return -EFSCORRUPTED; 3449 } 3450 } 3451 3452 while (count) { 3453 int compr_blocks = 0; 3454 3455 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3456 blkaddr = f2fs_data_blkaddr(dn); 3457 3458 if (i == 0) { 3459 if (blkaddr == COMPRESS_ADDR) 3460 continue; 3461 dn->ofs_in_node += cluster_size; 3462 goto next; 3463 } 3464 3465 if (__is_valid_data_blkaddr(blkaddr)) 3466 compr_blocks++; 3467 3468 if (blkaddr != NEW_ADDR) 3469 continue; 3470 3471 dn->data_blkaddr = NULL_ADDR; 3472 f2fs_set_data_blkaddr(dn); 3473 } 3474 3475 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3476 dec_valid_block_count(sbi, dn->inode, 3477 cluster_size - compr_blocks); 3478 3479 released_blocks += cluster_size - compr_blocks; 3480 next: 3481 count -= cluster_size; 3482 } 3483 3484 return released_blocks; 3485 } 3486 3487 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3488 { 3489 struct inode *inode = file_inode(filp); 3490 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3491 pgoff_t page_idx = 0, last_idx; 3492 unsigned int released_blocks = 0; 3493 int ret; 3494 int writecount; 3495 3496 if (!f2fs_sb_has_compression(sbi)) 3497 return -EOPNOTSUPP; 3498 3499 if (!f2fs_compressed_file(inode)) 3500 return -EINVAL; 3501 3502 if (f2fs_readonly(sbi->sb)) 3503 return -EROFS; 3504 3505 ret = mnt_want_write_file(filp); 3506 if (ret) 3507 return ret; 3508 3509 f2fs_balance_fs(sbi, true); 3510 3511 inode_lock(inode); 3512 3513 writecount = atomic_read(&inode->i_writecount); 3514 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3515 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3516 ret = -EBUSY; 3517 goto out; 3518 } 3519 3520 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3521 ret = -EINVAL; 3522 goto out; 3523 } 3524 3525 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3526 if (ret) 3527 goto out; 3528 3529 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3530 ret = -EPERM; 3531 goto out; 3532 } 3533 3534 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3535 inode_set_ctime_current(inode); 3536 f2fs_mark_inode_dirty_sync(inode, true); 3537 3538 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3539 filemap_invalidate_lock(inode->i_mapping); 3540 3541 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3542 3543 while (page_idx < last_idx) { 3544 struct dnode_of_data dn; 3545 pgoff_t end_offset, count; 3546 3547 set_new_dnode(&dn, inode, NULL, NULL, 0); 3548 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3549 if (ret) { 3550 if (ret == -ENOENT) { 3551 page_idx = f2fs_get_next_page_offset(&dn, 3552 page_idx); 3553 ret = 0; 3554 continue; 3555 } 3556 break; 3557 } 3558 3559 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3560 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3561 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3562 3563 ret = release_compress_blocks(&dn, count); 3564 3565 f2fs_put_dnode(&dn); 3566 3567 if (ret < 0) 3568 break; 3569 3570 page_idx += count; 3571 released_blocks += ret; 3572 } 3573 3574 filemap_invalidate_unlock(inode->i_mapping); 3575 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3576 out: 3577 inode_unlock(inode); 3578 3579 mnt_drop_write_file(filp); 3580 3581 if (ret >= 0) { 3582 ret = put_user(released_blocks, (u64 __user *)arg); 3583 } else if (released_blocks && 3584 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3585 set_sbi_flag(sbi, SBI_NEED_FSCK); 3586 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3587 "iblocks=%llu, released=%u, compr_blocks=%u, " 3588 "run fsck to fix.", 3589 __func__, inode->i_ino, inode->i_blocks, 3590 released_blocks, 3591 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3592 } 3593 3594 return ret; 3595 } 3596 3597 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3598 { 3599 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3600 unsigned int reserved_blocks = 0; 3601 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3602 block_t blkaddr; 3603 int i; 3604 3605 for (i = 0; i < count; i++) { 3606 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3607 dn->ofs_in_node + i); 3608 3609 if (!__is_valid_data_blkaddr(blkaddr)) 3610 continue; 3611 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3612 DATA_GENERIC_ENHANCE))) { 3613 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3614 return -EFSCORRUPTED; 3615 } 3616 } 3617 3618 while (count) { 3619 int compr_blocks = 0; 3620 blkcnt_t reserved; 3621 int ret; 3622 3623 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3624 blkaddr = f2fs_data_blkaddr(dn); 3625 3626 if (i == 0) { 3627 if (blkaddr == COMPRESS_ADDR) 3628 continue; 3629 dn->ofs_in_node += cluster_size; 3630 goto next; 3631 } 3632 3633 if (__is_valid_data_blkaddr(blkaddr)) { 3634 compr_blocks++; 3635 continue; 3636 } 3637 3638 dn->data_blkaddr = NEW_ADDR; 3639 f2fs_set_data_blkaddr(dn); 3640 } 3641 3642 reserved = cluster_size - compr_blocks; 3643 ret = inc_valid_block_count(sbi, dn->inode, &reserved); 3644 if (ret) 3645 return ret; 3646 3647 if (reserved != cluster_size - compr_blocks) 3648 return -ENOSPC; 3649 3650 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3651 3652 reserved_blocks += reserved; 3653 next: 3654 count -= cluster_size; 3655 } 3656 3657 return reserved_blocks; 3658 } 3659 3660 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3661 { 3662 struct inode *inode = file_inode(filp); 3663 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3664 pgoff_t page_idx = 0, last_idx; 3665 unsigned int reserved_blocks = 0; 3666 int ret; 3667 3668 if (!f2fs_sb_has_compression(sbi)) 3669 return -EOPNOTSUPP; 3670 3671 if (!f2fs_compressed_file(inode)) 3672 return -EINVAL; 3673 3674 if (f2fs_readonly(sbi->sb)) 3675 return -EROFS; 3676 3677 ret = mnt_want_write_file(filp); 3678 if (ret) 3679 return ret; 3680 3681 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3682 goto out; 3683 3684 f2fs_balance_fs(sbi, true); 3685 3686 inode_lock(inode); 3687 3688 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3689 ret = -EINVAL; 3690 goto unlock_inode; 3691 } 3692 3693 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3694 filemap_invalidate_lock(inode->i_mapping); 3695 3696 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3697 3698 while (page_idx < last_idx) { 3699 struct dnode_of_data dn; 3700 pgoff_t end_offset, count; 3701 3702 set_new_dnode(&dn, inode, NULL, NULL, 0); 3703 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3704 if (ret) { 3705 if (ret == -ENOENT) { 3706 page_idx = f2fs_get_next_page_offset(&dn, 3707 page_idx); 3708 ret = 0; 3709 continue; 3710 } 3711 break; 3712 } 3713 3714 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3715 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3716 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3717 3718 ret = reserve_compress_blocks(&dn, count); 3719 3720 f2fs_put_dnode(&dn); 3721 3722 if (ret < 0) 3723 break; 3724 3725 page_idx += count; 3726 reserved_blocks += ret; 3727 } 3728 3729 filemap_invalidate_unlock(inode->i_mapping); 3730 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3731 3732 if (ret >= 0) { 3733 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3734 inode_set_ctime_current(inode); 3735 f2fs_mark_inode_dirty_sync(inode, true); 3736 } 3737 unlock_inode: 3738 inode_unlock(inode); 3739 out: 3740 mnt_drop_write_file(filp); 3741 3742 if (ret >= 0) { 3743 ret = put_user(reserved_blocks, (u64 __user *)arg); 3744 } else if (reserved_blocks && 3745 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3746 set_sbi_flag(sbi, SBI_NEED_FSCK); 3747 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3748 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3749 "run fsck to fix.", 3750 __func__, inode->i_ino, inode->i_blocks, 3751 reserved_blocks, 3752 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3753 } 3754 3755 return ret; 3756 } 3757 3758 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3759 pgoff_t off, block_t block, block_t len, u32 flags) 3760 { 3761 sector_t sector = SECTOR_FROM_BLOCK(block); 3762 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3763 int ret = 0; 3764 3765 if (flags & F2FS_TRIM_FILE_DISCARD) { 3766 if (bdev_max_secure_erase_sectors(bdev)) 3767 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3768 GFP_NOFS); 3769 else 3770 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3771 GFP_NOFS); 3772 } 3773 3774 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3775 if (IS_ENCRYPTED(inode)) 3776 ret = fscrypt_zeroout_range(inode, off, block, len); 3777 else 3778 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3779 GFP_NOFS, 0); 3780 } 3781 3782 return ret; 3783 } 3784 3785 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3786 { 3787 struct inode *inode = file_inode(filp); 3788 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3789 struct address_space *mapping = inode->i_mapping; 3790 struct block_device *prev_bdev = NULL; 3791 struct f2fs_sectrim_range range; 3792 pgoff_t index, pg_end, prev_index = 0; 3793 block_t prev_block = 0, len = 0; 3794 loff_t end_addr; 3795 bool to_end = false; 3796 int ret = 0; 3797 3798 if (!(filp->f_mode & FMODE_WRITE)) 3799 return -EBADF; 3800 3801 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3802 sizeof(range))) 3803 return -EFAULT; 3804 3805 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3806 !S_ISREG(inode->i_mode)) 3807 return -EINVAL; 3808 3809 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3810 !f2fs_hw_support_discard(sbi)) || 3811 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3812 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3813 return -EOPNOTSUPP; 3814 3815 file_start_write(filp); 3816 inode_lock(inode); 3817 3818 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3819 range.start >= inode->i_size) { 3820 ret = -EINVAL; 3821 goto err; 3822 } 3823 3824 if (range.len == 0) 3825 goto err; 3826 3827 if (inode->i_size - range.start > range.len) { 3828 end_addr = range.start + range.len; 3829 } else { 3830 end_addr = range.len == (u64)-1 ? 3831 sbi->sb->s_maxbytes : inode->i_size; 3832 to_end = true; 3833 } 3834 3835 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3836 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3837 ret = -EINVAL; 3838 goto err; 3839 } 3840 3841 index = F2FS_BYTES_TO_BLK(range.start); 3842 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3843 3844 ret = f2fs_convert_inline_inode(inode); 3845 if (ret) 3846 goto err; 3847 3848 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3849 filemap_invalidate_lock(mapping); 3850 3851 ret = filemap_write_and_wait_range(mapping, range.start, 3852 to_end ? LLONG_MAX : end_addr - 1); 3853 if (ret) 3854 goto out; 3855 3856 truncate_inode_pages_range(mapping, range.start, 3857 to_end ? -1 : end_addr - 1); 3858 3859 while (index < pg_end) { 3860 struct dnode_of_data dn; 3861 pgoff_t end_offset, count; 3862 int i; 3863 3864 set_new_dnode(&dn, inode, NULL, NULL, 0); 3865 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3866 if (ret) { 3867 if (ret == -ENOENT) { 3868 index = f2fs_get_next_page_offset(&dn, index); 3869 continue; 3870 } 3871 goto out; 3872 } 3873 3874 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3875 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3876 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3877 struct block_device *cur_bdev; 3878 block_t blkaddr = f2fs_data_blkaddr(&dn); 3879 3880 if (!__is_valid_data_blkaddr(blkaddr)) 3881 continue; 3882 3883 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3884 DATA_GENERIC_ENHANCE)) { 3885 ret = -EFSCORRUPTED; 3886 f2fs_put_dnode(&dn); 3887 f2fs_handle_error(sbi, 3888 ERROR_INVALID_BLKADDR); 3889 goto out; 3890 } 3891 3892 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3893 if (f2fs_is_multi_device(sbi)) { 3894 int di = f2fs_target_device_index(sbi, blkaddr); 3895 3896 blkaddr -= FDEV(di).start_blk; 3897 } 3898 3899 if (len) { 3900 if (prev_bdev == cur_bdev && 3901 index == prev_index + len && 3902 blkaddr == prev_block + len) { 3903 len++; 3904 } else { 3905 ret = f2fs_secure_erase(prev_bdev, 3906 inode, prev_index, prev_block, 3907 len, range.flags); 3908 if (ret) { 3909 f2fs_put_dnode(&dn); 3910 goto out; 3911 } 3912 3913 len = 0; 3914 } 3915 } 3916 3917 if (!len) { 3918 prev_bdev = cur_bdev; 3919 prev_index = index; 3920 prev_block = blkaddr; 3921 len = 1; 3922 } 3923 } 3924 3925 f2fs_put_dnode(&dn); 3926 3927 if (fatal_signal_pending(current)) { 3928 ret = -EINTR; 3929 goto out; 3930 } 3931 cond_resched(); 3932 } 3933 3934 if (len) 3935 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3936 prev_block, len, range.flags); 3937 out: 3938 filemap_invalidate_unlock(mapping); 3939 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3940 err: 3941 inode_unlock(inode); 3942 file_end_write(filp); 3943 3944 return ret; 3945 } 3946 3947 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3948 { 3949 struct inode *inode = file_inode(filp); 3950 struct f2fs_comp_option option; 3951 3952 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3953 return -EOPNOTSUPP; 3954 3955 inode_lock_shared(inode); 3956 3957 if (!f2fs_compressed_file(inode)) { 3958 inode_unlock_shared(inode); 3959 return -ENODATA; 3960 } 3961 3962 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3963 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3964 3965 inode_unlock_shared(inode); 3966 3967 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3968 sizeof(option))) 3969 return -EFAULT; 3970 3971 return 0; 3972 } 3973 3974 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3975 { 3976 struct inode *inode = file_inode(filp); 3977 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3978 struct f2fs_comp_option option; 3979 int ret = 0; 3980 3981 if (!f2fs_sb_has_compression(sbi)) 3982 return -EOPNOTSUPP; 3983 3984 if (!(filp->f_mode & FMODE_WRITE)) 3985 return -EBADF; 3986 3987 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3988 sizeof(option))) 3989 return -EFAULT; 3990 3991 if (!f2fs_compressed_file(inode) || 3992 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3993 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3994 option.algorithm >= COMPRESS_MAX) 3995 return -EINVAL; 3996 3997 file_start_write(filp); 3998 inode_lock(inode); 3999 4000 f2fs_down_write(&F2FS_I(inode)->i_sem); 4001 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4002 ret = -EBUSY; 4003 goto out; 4004 } 4005 4006 if (F2FS_HAS_BLOCKS(inode)) { 4007 ret = -EFBIG; 4008 goto out; 4009 } 4010 4011 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4012 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4013 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4014 /* Set default level */ 4015 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4016 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4017 else 4018 F2FS_I(inode)->i_compress_level = 0; 4019 /* Adjust mount option level */ 4020 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4021 F2FS_OPTION(sbi).compress_level) 4022 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4023 f2fs_mark_inode_dirty_sync(inode, true); 4024 4025 if (!f2fs_is_compress_backend_ready(inode)) 4026 f2fs_warn(sbi, "compression algorithm is successfully set, " 4027 "but current kernel doesn't support this algorithm."); 4028 out: 4029 f2fs_up_write(&F2FS_I(inode)->i_sem); 4030 inode_unlock(inode); 4031 file_end_write(filp); 4032 4033 return ret; 4034 } 4035 4036 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4037 { 4038 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4039 struct address_space *mapping = inode->i_mapping; 4040 struct page *page; 4041 pgoff_t redirty_idx = page_idx; 4042 int i, page_len = 0, ret = 0; 4043 4044 page_cache_ra_unbounded(&ractl, len, 0); 4045 4046 for (i = 0; i < len; i++, page_idx++) { 4047 page = read_cache_page(mapping, page_idx, NULL, NULL); 4048 if (IS_ERR(page)) { 4049 ret = PTR_ERR(page); 4050 break; 4051 } 4052 page_len++; 4053 } 4054 4055 for (i = 0; i < page_len; i++, redirty_idx++) { 4056 page = find_lock_page(mapping, redirty_idx); 4057 4058 /* It will never fail, when page has pinned above */ 4059 f2fs_bug_on(F2FS_I_SB(inode), !page); 4060 4061 set_page_dirty(page); 4062 f2fs_put_page(page, 1); 4063 f2fs_put_page(page, 0); 4064 } 4065 4066 return ret; 4067 } 4068 4069 static int f2fs_ioc_decompress_file(struct file *filp) 4070 { 4071 struct inode *inode = file_inode(filp); 4072 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4073 struct f2fs_inode_info *fi = F2FS_I(inode); 4074 pgoff_t page_idx = 0, last_idx; 4075 unsigned int blk_per_seg = sbi->blocks_per_seg; 4076 int cluster_size = fi->i_cluster_size; 4077 int count, ret; 4078 4079 if (!f2fs_sb_has_compression(sbi) || 4080 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4081 return -EOPNOTSUPP; 4082 4083 if (!(filp->f_mode & FMODE_WRITE)) 4084 return -EBADF; 4085 4086 if (!f2fs_compressed_file(inode)) 4087 return -EINVAL; 4088 4089 f2fs_balance_fs(sbi, true); 4090 4091 file_start_write(filp); 4092 inode_lock(inode); 4093 4094 if (!f2fs_is_compress_backend_ready(inode)) { 4095 ret = -EOPNOTSUPP; 4096 goto out; 4097 } 4098 4099 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4100 ret = -EINVAL; 4101 goto out; 4102 } 4103 4104 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4105 if (ret) 4106 goto out; 4107 4108 if (!atomic_read(&fi->i_compr_blocks)) 4109 goto out; 4110 4111 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4112 4113 count = last_idx - page_idx; 4114 while (count && count >= cluster_size) { 4115 ret = redirty_blocks(inode, page_idx, cluster_size); 4116 if (ret < 0) 4117 break; 4118 4119 if (get_dirty_pages(inode) >= blk_per_seg) { 4120 ret = filemap_fdatawrite(inode->i_mapping); 4121 if (ret < 0) 4122 break; 4123 } 4124 4125 count -= cluster_size; 4126 page_idx += cluster_size; 4127 4128 cond_resched(); 4129 if (fatal_signal_pending(current)) { 4130 ret = -EINTR; 4131 break; 4132 } 4133 } 4134 4135 if (!ret) 4136 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4137 LLONG_MAX); 4138 4139 if (ret) 4140 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4141 __func__, ret); 4142 out: 4143 inode_unlock(inode); 4144 file_end_write(filp); 4145 4146 return ret; 4147 } 4148 4149 static int f2fs_ioc_compress_file(struct file *filp) 4150 { 4151 struct inode *inode = file_inode(filp); 4152 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4153 pgoff_t page_idx = 0, last_idx; 4154 unsigned int blk_per_seg = sbi->blocks_per_seg; 4155 int cluster_size = F2FS_I(inode)->i_cluster_size; 4156 int count, ret; 4157 4158 if (!f2fs_sb_has_compression(sbi) || 4159 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4160 return -EOPNOTSUPP; 4161 4162 if (!(filp->f_mode & FMODE_WRITE)) 4163 return -EBADF; 4164 4165 if (!f2fs_compressed_file(inode)) 4166 return -EINVAL; 4167 4168 f2fs_balance_fs(sbi, true); 4169 4170 file_start_write(filp); 4171 inode_lock(inode); 4172 4173 if (!f2fs_is_compress_backend_ready(inode)) { 4174 ret = -EOPNOTSUPP; 4175 goto out; 4176 } 4177 4178 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4179 ret = -EINVAL; 4180 goto out; 4181 } 4182 4183 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4184 if (ret) 4185 goto out; 4186 4187 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4188 4189 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4190 4191 count = last_idx - page_idx; 4192 while (count && count >= cluster_size) { 4193 ret = redirty_blocks(inode, page_idx, cluster_size); 4194 if (ret < 0) 4195 break; 4196 4197 if (get_dirty_pages(inode) >= blk_per_seg) { 4198 ret = filemap_fdatawrite(inode->i_mapping); 4199 if (ret < 0) 4200 break; 4201 } 4202 4203 count -= cluster_size; 4204 page_idx += cluster_size; 4205 4206 cond_resched(); 4207 if (fatal_signal_pending(current)) { 4208 ret = -EINTR; 4209 break; 4210 } 4211 } 4212 4213 if (!ret) 4214 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4215 LLONG_MAX); 4216 4217 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4218 4219 if (ret) 4220 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4221 __func__, ret); 4222 out: 4223 inode_unlock(inode); 4224 file_end_write(filp); 4225 4226 return ret; 4227 } 4228 4229 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4230 { 4231 switch (cmd) { 4232 case FS_IOC_GETVERSION: 4233 return f2fs_ioc_getversion(filp, arg); 4234 case F2FS_IOC_START_ATOMIC_WRITE: 4235 return f2fs_ioc_start_atomic_write(filp, false); 4236 case F2FS_IOC_START_ATOMIC_REPLACE: 4237 return f2fs_ioc_start_atomic_write(filp, true); 4238 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4239 return f2fs_ioc_commit_atomic_write(filp); 4240 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4241 return f2fs_ioc_abort_atomic_write(filp); 4242 case F2FS_IOC_START_VOLATILE_WRITE: 4243 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4244 return -EOPNOTSUPP; 4245 case F2FS_IOC_SHUTDOWN: 4246 return f2fs_ioc_shutdown(filp, arg); 4247 case FITRIM: 4248 return f2fs_ioc_fitrim(filp, arg); 4249 case FS_IOC_SET_ENCRYPTION_POLICY: 4250 return f2fs_ioc_set_encryption_policy(filp, arg); 4251 case FS_IOC_GET_ENCRYPTION_POLICY: 4252 return f2fs_ioc_get_encryption_policy(filp, arg); 4253 case FS_IOC_GET_ENCRYPTION_PWSALT: 4254 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4255 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4256 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4257 case FS_IOC_ADD_ENCRYPTION_KEY: 4258 return f2fs_ioc_add_encryption_key(filp, arg); 4259 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4260 return f2fs_ioc_remove_encryption_key(filp, arg); 4261 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4262 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4263 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4264 return f2fs_ioc_get_encryption_key_status(filp, arg); 4265 case FS_IOC_GET_ENCRYPTION_NONCE: 4266 return f2fs_ioc_get_encryption_nonce(filp, arg); 4267 case F2FS_IOC_GARBAGE_COLLECT: 4268 return f2fs_ioc_gc(filp, arg); 4269 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4270 return f2fs_ioc_gc_range(filp, arg); 4271 case F2FS_IOC_WRITE_CHECKPOINT: 4272 return f2fs_ioc_write_checkpoint(filp); 4273 case F2FS_IOC_DEFRAGMENT: 4274 return f2fs_ioc_defragment(filp, arg); 4275 case F2FS_IOC_MOVE_RANGE: 4276 return f2fs_ioc_move_range(filp, arg); 4277 case F2FS_IOC_FLUSH_DEVICE: 4278 return f2fs_ioc_flush_device(filp, arg); 4279 case F2FS_IOC_GET_FEATURES: 4280 return f2fs_ioc_get_features(filp, arg); 4281 case F2FS_IOC_GET_PIN_FILE: 4282 return f2fs_ioc_get_pin_file(filp, arg); 4283 case F2FS_IOC_SET_PIN_FILE: 4284 return f2fs_ioc_set_pin_file(filp, arg); 4285 case F2FS_IOC_PRECACHE_EXTENTS: 4286 return f2fs_ioc_precache_extents(filp); 4287 case F2FS_IOC_RESIZE_FS: 4288 return f2fs_ioc_resize_fs(filp, arg); 4289 case FS_IOC_ENABLE_VERITY: 4290 return f2fs_ioc_enable_verity(filp, arg); 4291 case FS_IOC_MEASURE_VERITY: 4292 return f2fs_ioc_measure_verity(filp, arg); 4293 case FS_IOC_READ_VERITY_METADATA: 4294 return f2fs_ioc_read_verity_metadata(filp, arg); 4295 case FS_IOC_GETFSLABEL: 4296 return f2fs_ioc_getfslabel(filp, arg); 4297 case FS_IOC_SETFSLABEL: 4298 return f2fs_ioc_setfslabel(filp, arg); 4299 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4300 return f2fs_ioc_get_compress_blocks(filp, arg); 4301 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4302 return f2fs_release_compress_blocks(filp, arg); 4303 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4304 return f2fs_reserve_compress_blocks(filp, arg); 4305 case F2FS_IOC_SEC_TRIM_FILE: 4306 return f2fs_sec_trim_file(filp, arg); 4307 case F2FS_IOC_GET_COMPRESS_OPTION: 4308 return f2fs_ioc_get_compress_option(filp, arg); 4309 case F2FS_IOC_SET_COMPRESS_OPTION: 4310 return f2fs_ioc_set_compress_option(filp, arg); 4311 case F2FS_IOC_DECOMPRESS_FILE: 4312 return f2fs_ioc_decompress_file(filp); 4313 case F2FS_IOC_COMPRESS_FILE: 4314 return f2fs_ioc_compress_file(filp); 4315 default: 4316 return -ENOTTY; 4317 } 4318 } 4319 4320 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4321 { 4322 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4323 return -EIO; 4324 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4325 return -ENOSPC; 4326 4327 return __f2fs_ioctl(filp, cmd, arg); 4328 } 4329 4330 /* 4331 * Return %true if the given read or write request should use direct I/O, or 4332 * %false if it should use buffered I/O. 4333 */ 4334 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4335 struct iov_iter *iter) 4336 { 4337 unsigned int align; 4338 4339 if (!(iocb->ki_flags & IOCB_DIRECT)) 4340 return false; 4341 4342 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4343 return false; 4344 4345 /* 4346 * Direct I/O not aligned to the disk's logical_block_size will be 4347 * attempted, but will fail with -EINVAL. 4348 * 4349 * f2fs additionally requires that direct I/O be aligned to the 4350 * filesystem block size, which is often a stricter requirement. 4351 * However, f2fs traditionally falls back to buffered I/O on requests 4352 * that are logical_block_size-aligned but not fs-block aligned. 4353 * 4354 * The below logic implements this behavior. 4355 */ 4356 align = iocb->ki_pos | iov_iter_alignment(iter); 4357 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4358 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4359 return false; 4360 4361 return true; 4362 } 4363 4364 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4365 unsigned int flags) 4366 { 4367 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4368 4369 dec_page_count(sbi, F2FS_DIO_READ); 4370 if (error) 4371 return error; 4372 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4373 return 0; 4374 } 4375 4376 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4377 .end_io = f2fs_dio_read_end_io, 4378 }; 4379 4380 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4381 { 4382 struct file *file = iocb->ki_filp; 4383 struct inode *inode = file_inode(file); 4384 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4385 struct f2fs_inode_info *fi = F2FS_I(inode); 4386 const loff_t pos = iocb->ki_pos; 4387 const size_t count = iov_iter_count(to); 4388 struct iomap_dio *dio; 4389 ssize_t ret; 4390 4391 if (count == 0) 4392 return 0; /* skip atime update */ 4393 4394 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4395 4396 if (iocb->ki_flags & IOCB_NOWAIT) { 4397 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4398 ret = -EAGAIN; 4399 goto out; 4400 } 4401 } else { 4402 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4403 } 4404 4405 /* 4406 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4407 * the higher-level function iomap_dio_rw() in order to ensure that the 4408 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4409 */ 4410 inc_page_count(sbi, F2FS_DIO_READ); 4411 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4412 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4413 if (IS_ERR_OR_NULL(dio)) { 4414 ret = PTR_ERR_OR_ZERO(dio); 4415 if (ret != -EIOCBQUEUED) 4416 dec_page_count(sbi, F2FS_DIO_READ); 4417 } else { 4418 ret = iomap_dio_complete(dio); 4419 } 4420 4421 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4422 4423 file_accessed(file); 4424 out: 4425 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4426 return ret; 4427 } 4428 4429 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4430 int rw) 4431 { 4432 struct inode *inode = file_inode(file); 4433 char *buf, *path; 4434 4435 buf = f2fs_getname(F2FS_I_SB(inode)); 4436 if (!buf) 4437 return; 4438 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4439 if (IS_ERR(path)) 4440 goto free_buf; 4441 if (rw == WRITE) 4442 trace_f2fs_datawrite_start(inode, pos, count, 4443 current->pid, path, current->comm); 4444 else 4445 trace_f2fs_dataread_start(inode, pos, count, 4446 current->pid, path, current->comm); 4447 free_buf: 4448 f2fs_putname(buf); 4449 } 4450 4451 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4452 { 4453 struct inode *inode = file_inode(iocb->ki_filp); 4454 const loff_t pos = iocb->ki_pos; 4455 ssize_t ret; 4456 4457 if (!f2fs_is_compress_backend_ready(inode)) 4458 return -EOPNOTSUPP; 4459 4460 if (trace_f2fs_dataread_start_enabled()) 4461 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4462 iov_iter_count(to), READ); 4463 4464 if (f2fs_should_use_dio(inode, iocb, to)) { 4465 ret = f2fs_dio_read_iter(iocb, to); 4466 } else { 4467 ret = filemap_read(iocb, to, 0); 4468 if (ret > 0) 4469 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4470 APP_BUFFERED_READ_IO, ret); 4471 } 4472 if (trace_f2fs_dataread_end_enabled()) 4473 trace_f2fs_dataread_end(inode, pos, ret); 4474 return ret; 4475 } 4476 4477 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4478 struct pipe_inode_info *pipe, 4479 size_t len, unsigned int flags) 4480 { 4481 struct inode *inode = file_inode(in); 4482 const loff_t pos = *ppos; 4483 ssize_t ret; 4484 4485 if (!f2fs_is_compress_backend_ready(inode)) 4486 return -EOPNOTSUPP; 4487 4488 if (trace_f2fs_dataread_start_enabled()) 4489 f2fs_trace_rw_file_path(in, pos, len, READ); 4490 4491 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4492 if (ret > 0) 4493 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4494 APP_BUFFERED_READ_IO, ret); 4495 4496 if (trace_f2fs_dataread_end_enabled()) 4497 trace_f2fs_dataread_end(inode, pos, ret); 4498 return ret; 4499 } 4500 4501 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4502 { 4503 struct file *file = iocb->ki_filp; 4504 struct inode *inode = file_inode(file); 4505 ssize_t count; 4506 int err; 4507 4508 if (IS_IMMUTABLE(inode)) 4509 return -EPERM; 4510 4511 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4512 return -EPERM; 4513 4514 count = generic_write_checks(iocb, from); 4515 if (count <= 0) 4516 return count; 4517 4518 err = file_modified(file); 4519 if (err) 4520 return err; 4521 return count; 4522 } 4523 4524 /* 4525 * Preallocate blocks for a write request, if it is possible and helpful to do 4526 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4527 * blocks were preallocated, or a negative errno value if something went 4528 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4529 * requested blocks (not just some of them) have been allocated. 4530 */ 4531 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4532 bool dio) 4533 { 4534 struct inode *inode = file_inode(iocb->ki_filp); 4535 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4536 const loff_t pos = iocb->ki_pos; 4537 const size_t count = iov_iter_count(iter); 4538 struct f2fs_map_blocks map = {}; 4539 int flag; 4540 int ret; 4541 4542 /* If it will be an out-of-place direct write, don't bother. */ 4543 if (dio && f2fs_lfs_mode(sbi)) 4544 return 0; 4545 /* 4546 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4547 * buffered IO, if DIO meets any holes. 4548 */ 4549 if (dio && i_size_read(inode) && 4550 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4551 return 0; 4552 4553 /* No-wait I/O can't allocate blocks. */ 4554 if (iocb->ki_flags & IOCB_NOWAIT) 4555 return 0; 4556 4557 /* If it will be a short write, don't bother. */ 4558 if (fault_in_iov_iter_readable(iter, count)) 4559 return 0; 4560 4561 if (f2fs_has_inline_data(inode)) { 4562 /* If the data will fit inline, don't bother. */ 4563 if (pos + count <= MAX_INLINE_DATA(inode)) 4564 return 0; 4565 ret = f2fs_convert_inline_inode(inode); 4566 if (ret) 4567 return ret; 4568 } 4569 4570 /* Do not preallocate blocks that will be written partially in 4KB. */ 4571 map.m_lblk = F2FS_BLK_ALIGN(pos); 4572 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4573 if (map.m_len > map.m_lblk) 4574 map.m_len -= map.m_lblk; 4575 else 4576 map.m_len = 0; 4577 map.m_may_create = true; 4578 if (dio) { 4579 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4580 flag = F2FS_GET_BLOCK_PRE_DIO; 4581 } else { 4582 map.m_seg_type = NO_CHECK_TYPE; 4583 flag = F2FS_GET_BLOCK_PRE_AIO; 4584 } 4585 4586 ret = f2fs_map_blocks(inode, &map, flag); 4587 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4588 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4589 return ret; 4590 if (ret == 0) 4591 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4592 return map.m_len; 4593 } 4594 4595 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4596 struct iov_iter *from) 4597 { 4598 struct file *file = iocb->ki_filp; 4599 struct inode *inode = file_inode(file); 4600 ssize_t ret; 4601 4602 if (iocb->ki_flags & IOCB_NOWAIT) 4603 return -EOPNOTSUPP; 4604 4605 ret = generic_perform_write(iocb, from); 4606 4607 if (ret > 0) { 4608 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4609 APP_BUFFERED_IO, ret); 4610 } 4611 return ret; 4612 } 4613 4614 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4615 unsigned int flags) 4616 { 4617 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4618 4619 dec_page_count(sbi, F2FS_DIO_WRITE); 4620 if (error) 4621 return error; 4622 f2fs_update_time(sbi, REQ_TIME); 4623 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4624 return 0; 4625 } 4626 4627 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4628 .end_io = f2fs_dio_write_end_io, 4629 }; 4630 4631 static void f2fs_flush_buffered_write(struct address_space *mapping, 4632 loff_t start_pos, loff_t end_pos) 4633 { 4634 int ret; 4635 4636 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4637 if (ret < 0) 4638 return; 4639 invalidate_mapping_pages(mapping, 4640 start_pos >> PAGE_SHIFT, 4641 end_pos >> PAGE_SHIFT); 4642 } 4643 4644 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4645 bool *may_need_sync) 4646 { 4647 struct file *file = iocb->ki_filp; 4648 struct inode *inode = file_inode(file); 4649 struct f2fs_inode_info *fi = F2FS_I(inode); 4650 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4651 const bool do_opu = f2fs_lfs_mode(sbi); 4652 const loff_t pos = iocb->ki_pos; 4653 const ssize_t count = iov_iter_count(from); 4654 unsigned int dio_flags; 4655 struct iomap_dio *dio; 4656 ssize_t ret; 4657 4658 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4659 4660 if (iocb->ki_flags & IOCB_NOWAIT) { 4661 /* f2fs_convert_inline_inode() and block allocation can block */ 4662 if (f2fs_has_inline_data(inode) || 4663 !f2fs_overwrite_io(inode, pos, count)) { 4664 ret = -EAGAIN; 4665 goto out; 4666 } 4667 4668 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4669 ret = -EAGAIN; 4670 goto out; 4671 } 4672 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4673 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4674 ret = -EAGAIN; 4675 goto out; 4676 } 4677 } else { 4678 ret = f2fs_convert_inline_inode(inode); 4679 if (ret) 4680 goto out; 4681 4682 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4683 if (do_opu) 4684 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4685 } 4686 4687 /* 4688 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4689 * the higher-level function iomap_dio_rw() in order to ensure that the 4690 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4691 */ 4692 inc_page_count(sbi, F2FS_DIO_WRITE); 4693 dio_flags = 0; 4694 if (pos + count > inode->i_size) 4695 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4696 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4697 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4698 if (IS_ERR_OR_NULL(dio)) { 4699 ret = PTR_ERR_OR_ZERO(dio); 4700 if (ret == -ENOTBLK) 4701 ret = 0; 4702 if (ret != -EIOCBQUEUED) 4703 dec_page_count(sbi, F2FS_DIO_WRITE); 4704 } else { 4705 ret = iomap_dio_complete(dio); 4706 } 4707 4708 if (do_opu) 4709 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4710 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4711 4712 if (ret < 0) 4713 goto out; 4714 if (pos + ret > inode->i_size) 4715 f2fs_i_size_write(inode, pos + ret); 4716 if (!do_opu) 4717 set_inode_flag(inode, FI_UPDATE_WRITE); 4718 4719 if (iov_iter_count(from)) { 4720 ssize_t ret2; 4721 loff_t bufio_start_pos = iocb->ki_pos; 4722 4723 /* 4724 * The direct write was partial, so we need to fall back to a 4725 * buffered write for the remainder. 4726 */ 4727 4728 ret2 = f2fs_buffered_write_iter(iocb, from); 4729 if (iov_iter_count(from)) 4730 f2fs_write_failed(inode, iocb->ki_pos); 4731 if (ret2 < 0) 4732 goto out; 4733 4734 /* 4735 * Ensure that the pagecache pages are written to disk and 4736 * invalidated to preserve the expected O_DIRECT semantics. 4737 */ 4738 if (ret2 > 0) { 4739 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4740 4741 ret += ret2; 4742 4743 f2fs_flush_buffered_write(file->f_mapping, 4744 bufio_start_pos, 4745 bufio_end_pos); 4746 } 4747 } else { 4748 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4749 *may_need_sync = false; 4750 } 4751 out: 4752 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4753 return ret; 4754 } 4755 4756 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4757 { 4758 struct inode *inode = file_inode(iocb->ki_filp); 4759 const loff_t orig_pos = iocb->ki_pos; 4760 const size_t orig_count = iov_iter_count(from); 4761 loff_t target_size; 4762 bool dio; 4763 bool may_need_sync = true; 4764 int preallocated; 4765 ssize_t ret; 4766 4767 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4768 ret = -EIO; 4769 goto out; 4770 } 4771 4772 if (!f2fs_is_compress_backend_ready(inode)) { 4773 ret = -EOPNOTSUPP; 4774 goto out; 4775 } 4776 4777 if (iocb->ki_flags & IOCB_NOWAIT) { 4778 if (!inode_trylock(inode)) { 4779 ret = -EAGAIN; 4780 goto out; 4781 } 4782 } else { 4783 inode_lock(inode); 4784 } 4785 4786 ret = f2fs_write_checks(iocb, from); 4787 if (ret <= 0) 4788 goto out_unlock; 4789 4790 /* Determine whether we will do a direct write or a buffered write. */ 4791 dio = f2fs_should_use_dio(inode, iocb, from); 4792 4793 /* Possibly preallocate the blocks for the write. */ 4794 target_size = iocb->ki_pos + iov_iter_count(from); 4795 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4796 if (preallocated < 0) { 4797 ret = preallocated; 4798 } else { 4799 if (trace_f2fs_datawrite_start_enabled()) 4800 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4801 orig_count, WRITE); 4802 4803 /* Do the actual write. */ 4804 ret = dio ? 4805 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4806 f2fs_buffered_write_iter(iocb, from); 4807 4808 if (trace_f2fs_datawrite_end_enabled()) 4809 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4810 } 4811 4812 /* Don't leave any preallocated blocks around past i_size. */ 4813 if (preallocated && i_size_read(inode) < target_size) { 4814 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4815 filemap_invalidate_lock(inode->i_mapping); 4816 if (!f2fs_truncate(inode)) 4817 file_dont_truncate(inode); 4818 filemap_invalidate_unlock(inode->i_mapping); 4819 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4820 } else { 4821 file_dont_truncate(inode); 4822 } 4823 4824 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4825 out_unlock: 4826 inode_unlock(inode); 4827 out: 4828 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4829 4830 if (ret > 0 && may_need_sync) 4831 ret = generic_write_sync(iocb, ret); 4832 4833 /* If buffered IO was forced, flush and drop the data from 4834 * the page cache to preserve O_DIRECT semantics 4835 */ 4836 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4837 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4838 orig_pos, 4839 orig_pos + ret - 1); 4840 4841 return ret; 4842 } 4843 4844 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4845 int advice) 4846 { 4847 struct address_space *mapping; 4848 struct backing_dev_info *bdi; 4849 struct inode *inode = file_inode(filp); 4850 int err; 4851 4852 if (advice == POSIX_FADV_SEQUENTIAL) { 4853 if (S_ISFIFO(inode->i_mode)) 4854 return -ESPIPE; 4855 4856 mapping = filp->f_mapping; 4857 if (!mapping || len < 0) 4858 return -EINVAL; 4859 4860 bdi = inode_to_bdi(mapping->host); 4861 filp->f_ra.ra_pages = bdi->ra_pages * 4862 F2FS_I_SB(inode)->seq_file_ra_mul; 4863 spin_lock(&filp->f_lock); 4864 filp->f_mode &= ~FMODE_RANDOM; 4865 spin_unlock(&filp->f_lock); 4866 return 0; 4867 } 4868 4869 err = generic_fadvise(filp, offset, len, advice); 4870 if (!err && advice == POSIX_FADV_DONTNEED && 4871 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4872 f2fs_compressed_file(inode)) 4873 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4874 4875 return err; 4876 } 4877 4878 #ifdef CONFIG_COMPAT 4879 struct compat_f2fs_gc_range { 4880 u32 sync; 4881 compat_u64 start; 4882 compat_u64 len; 4883 }; 4884 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4885 struct compat_f2fs_gc_range) 4886 4887 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4888 { 4889 struct compat_f2fs_gc_range __user *urange; 4890 struct f2fs_gc_range range; 4891 int err; 4892 4893 urange = compat_ptr(arg); 4894 err = get_user(range.sync, &urange->sync); 4895 err |= get_user(range.start, &urange->start); 4896 err |= get_user(range.len, &urange->len); 4897 if (err) 4898 return -EFAULT; 4899 4900 return __f2fs_ioc_gc_range(file, &range); 4901 } 4902 4903 struct compat_f2fs_move_range { 4904 u32 dst_fd; 4905 compat_u64 pos_in; 4906 compat_u64 pos_out; 4907 compat_u64 len; 4908 }; 4909 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4910 struct compat_f2fs_move_range) 4911 4912 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4913 { 4914 struct compat_f2fs_move_range __user *urange; 4915 struct f2fs_move_range range; 4916 int err; 4917 4918 urange = compat_ptr(arg); 4919 err = get_user(range.dst_fd, &urange->dst_fd); 4920 err |= get_user(range.pos_in, &urange->pos_in); 4921 err |= get_user(range.pos_out, &urange->pos_out); 4922 err |= get_user(range.len, &urange->len); 4923 if (err) 4924 return -EFAULT; 4925 4926 return __f2fs_ioc_move_range(file, &range); 4927 } 4928 4929 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4930 { 4931 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4932 return -EIO; 4933 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4934 return -ENOSPC; 4935 4936 switch (cmd) { 4937 case FS_IOC32_GETVERSION: 4938 cmd = FS_IOC_GETVERSION; 4939 break; 4940 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4941 return f2fs_compat_ioc_gc_range(file, arg); 4942 case F2FS_IOC32_MOVE_RANGE: 4943 return f2fs_compat_ioc_move_range(file, arg); 4944 case F2FS_IOC_START_ATOMIC_WRITE: 4945 case F2FS_IOC_START_ATOMIC_REPLACE: 4946 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4947 case F2FS_IOC_START_VOLATILE_WRITE: 4948 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4949 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4950 case F2FS_IOC_SHUTDOWN: 4951 case FITRIM: 4952 case FS_IOC_SET_ENCRYPTION_POLICY: 4953 case FS_IOC_GET_ENCRYPTION_PWSALT: 4954 case FS_IOC_GET_ENCRYPTION_POLICY: 4955 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4956 case FS_IOC_ADD_ENCRYPTION_KEY: 4957 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4958 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4959 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4960 case FS_IOC_GET_ENCRYPTION_NONCE: 4961 case F2FS_IOC_GARBAGE_COLLECT: 4962 case F2FS_IOC_WRITE_CHECKPOINT: 4963 case F2FS_IOC_DEFRAGMENT: 4964 case F2FS_IOC_FLUSH_DEVICE: 4965 case F2FS_IOC_GET_FEATURES: 4966 case F2FS_IOC_GET_PIN_FILE: 4967 case F2FS_IOC_SET_PIN_FILE: 4968 case F2FS_IOC_PRECACHE_EXTENTS: 4969 case F2FS_IOC_RESIZE_FS: 4970 case FS_IOC_ENABLE_VERITY: 4971 case FS_IOC_MEASURE_VERITY: 4972 case FS_IOC_READ_VERITY_METADATA: 4973 case FS_IOC_GETFSLABEL: 4974 case FS_IOC_SETFSLABEL: 4975 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4976 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4977 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4978 case F2FS_IOC_SEC_TRIM_FILE: 4979 case F2FS_IOC_GET_COMPRESS_OPTION: 4980 case F2FS_IOC_SET_COMPRESS_OPTION: 4981 case F2FS_IOC_DECOMPRESS_FILE: 4982 case F2FS_IOC_COMPRESS_FILE: 4983 break; 4984 default: 4985 return -ENOIOCTLCMD; 4986 } 4987 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4988 } 4989 #endif 4990 4991 const struct file_operations f2fs_file_operations = { 4992 .llseek = f2fs_llseek, 4993 .read_iter = f2fs_file_read_iter, 4994 .write_iter = f2fs_file_write_iter, 4995 .iopoll = iocb_bio_iopoll, 4996 .open = f2fs_file_open, 4997 .release = f2fs_release_file, 4998 .mmap = f2fs_file_mmap, 4999 .flush = f2fs_file_flush, 5000 .fsync = f2fs_sync_file, 5001 .fallocate = f2fs_fallocate, 5002 .unlocked_ioctl = f2fs_ioctl, 5003 #ifdef CONFIG_COMPAT 5004 .compat_ioctl = f2fs_compat_ioctl, 5005 #endif 5006 .splice_read = f2fs_file_splice_read, 5007 .splice_write = iter_file_splice_write, 5008 .fadvise = f2fs_file_fadvise, 5009 }; 5010