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 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 561 __le32 *addr; 562 bool compressed_cluster = false; 563 int cluster_index = 0, valid_blocks = 0; 564 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 565 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 566 567 addr = get_dnode_addr(dn->inode, dn->node_page) + ofs; 568 569 /* Assumption: truncation starts with cluster */ 570 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 571 block_t blkaddr = le32_to_cpu(*addr); 572 573 if (f2fs_compressed_file(dn->inode) && 574 !(cluster_index & (cluster_size - 1))) { 575 if (compressed_cluster) 576 f2fs_i_compr_blocks_update(dn->inode, 577 valid_blocks, false); 578 compressed_cluster = (blkaddr == COMPRESS_ADDR); 579 valid_blocks = 0; 580 } 581 582 if (blkaddr == NULL_ADDR) 583 continue; 584 585 f2fs_set_data_blkaddr(dn, NULL_ADDR); 586 587 if (__is_valid_data_blkaddr(blkaddr)) { 588 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 589 DATA_GENERIC_ENHANCE)) 590 continue; 591 if (compressed_cluster) 592 valid_blocks++; 593 } 594 595 f2fs_invalidate_blocks(sbi, blkaddr); 596 597 if (!released || blkaddr != COMPRESS_ADDR) 598 nr_free++; 599 } 600 601 if (compressed_cluster) 602 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 603 604 if (nr_free) { 605 pgoff_t fofs; 606 /* 607 * once we invalidate valid blkaddr in range [ofs, ofs + count], 608 * we will invalidate all blkaddr in the whole range. 609 */ 610 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 611 dn->inode) + ofs; 612 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 613 f2fs_update_age_extent_cache_range(dn, fofs, len); 614 dec_valid_block_count(sbi, dn->inode, nr_free); 615 } 616 dn->ofs_in_node = ofs; 617 618 f2fs_update_time(sbi, REQ_TIME); 619 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 620 dn->ofs_in_node, nr_free); 621 } 622 623 static int truncate_partial_data_page(struct inode *inode, u64 from, 624 bool cache_only) 625 { 626 loff_t offset = from & (PAGE_SIZE - 1); 627 pgoff_t index = from >> PAGE_SHIFT; 628 struct address_space *mapping = inode->i_mapping; 629 struct page *page; 630 631 if (!offset && !cache_only) 632 return 0; 633 634 if (cache_only) { 635 page = find_lock_page(mapping, index); 636 if (page && PageUptodate(page)) 637 goto truncate_out; 638 f2fs_put_page(page, 1); 639 return 0; 640 } 641 642 page = f2fs_get_lock_data_page(inode, index, true); 643 if (IS_ERR(page)) 644 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 645 truncate_out: 646 f2fs_wait_on_page_writeback(page, DATA, true, true); 647 zero_user(page, offset, PAGE_SIZE - offset); 648 649 /* An encrypted inode should have a key and truncate the last page. */ 650 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 651 if (!cache_only) 652 set_page_dirty(page); 653 f2fs_put_page(page, 1); 654 return 0; 655 } 656 657 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 658 { 659 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 660 struct dnode_of_data dn; 661 pgoff_t free_from; 662 int count = 0, err = 0; 663 struct page *ipage; 664 bool truncate_page = false; 665 666 trace_f2fs_truncate_blocks_enter(inode, from); 667 668 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 669 670 if (free_from >= max_file_blocks(inode)) 671 goto free_partial; 672 673 if (lock) 674 f2fs_lock_op(sbi); 675 676 ipage = f2fs_get_node_page(sbi, inode->i_ino); 677 if (IS_ERR(ipage)) { 678 err = PTR_ERR(ipage); 679 goto out; 680 } 681 682 if (f2fs_has_inline_data(inode)) { 683 f2fs_truncate_inline_inode(inode, ipage, from); 684 f2fs_put_page(ipage, 1); 685 truncate_page = true; 686 goto out; 687 } 688 689 set_new_dnode(&dn, inode, ipage, NULL, 0); 690 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 691 if (err) { 692 if (err == -ENOENT) 693 goto free_next; 694 goto out; 695 } 696 697 count = ADDRS_PER_PAGE(dn.node_page, inode); 698 699 count -= dn.ofs_in_node; 700 f2fs_bug_on(sbi, count < 0); 701 702 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 703 f2fs_truncate_data_blocks_range(&dn, count); 704 free_from += count; 705 } 706 707 f2fs_put_dnode(&dn); 708 free_next: 709 err = f2fs_truncate_inode_blocks(inode, free_from); 710 out: 711 if (lock) 712 f2fs_unlock_op(sbi); 713 free_partial: 714 /* lastly zero out the first data page */ 715 if (!err) 716 err = truncate_partial_data_page(inode, from, truncate_page); 717 718 trace_f2fs_truncate_blocks_exit(inode, err); 719 return err; 720 } 721 722 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 723 { 724 u64 free_from = from; 725 int err; 726 727 #ifdef CONFIG_F2FS_FS_COMPRESSION 728 /* 729 * for compressed file, only support cluster size 730 * aligned truncation. 731 */ 732 if (f2fs_compressed_file(inode)) 733 free_from = round_up(from, 734 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 735 #endif 736 737 err = f2fs_do_truncate_blocks(inode, free_from, lock); 738 if (err) 739 return err; 740 741 #ifdef CONFIG_F2FS_FS_COMPRESSION 742 /* 743 * For compressed file, after release compress blocks, don't allow write 744 * direct, but we should allow write direct after truncate to zero. 745 */ 746 if (f2fs_compressed_file(inode) && !free_from 747 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 748 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 749 750 if (from != free_from) { 751 err = f2fs_truncate_partial_cluster(inode, from, lock); 752 if (err) 753 return err; 754 } 755 #endif 756 757 return 0; 758 } 759 760 int f2fs_truncate(struct inode *inode) 761 { 762 int err; 763 764 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 765 return -EIO; 766 767 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 768 S_ISLNK(inode->i_mode))) 769 return 0; 770 771 trace_f2fs_truncate(inode); 772 773 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 774 return -EIO; 775 776 err = f2fs_dquot_initialize(inode); 777 if (err) 778 return err; 779 780 /* we should check inline_data size */ 781 if (!f2fs_may_inline_data(inode)) { 782 err = f2fs_convert_inline_inode(inode); 783 if (err) 784 return err; 785 } 786 787 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 788 if (err) 789 return err; 790 791 inode->i_mtime = inode_set_ctime_current(inode); 792 f2fs_mark_inode_dirty_sync(inode, false); 793 return 0; 794 } 795 796 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 797 { 798 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 799 800 if (!fscrypt_dio_supported(inode)) 801 return true; 802 if (fsverity_active(inode)) 803 return true; 804 if (f2fs_compressed_file(inode)) 805 return true; 806 807 /* disallow direct IO if any of devices has unaligned blksize */ 808 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 809 return true; 810 /* 811 * for blkzoned device, fallback direct IO to buffered IO, so 812 * all IOs can be serialized by log-structured write. 813 */ 814 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE)) 815 return true; 816 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 817 return true; 818 819 return false; 820 } 821 822 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 823 struct kstat *stat, u32 request_mask, unsigned int query_flags) 824 { 825 struct inode *inode = d_inode(path->dentry); 826 struct f2fs_inode_info *fi = F2FS_I(inode); 827 struct f2fs_inode *ri = NULL; 828 unsigned int flags; 829 830 if (f2fs_has_extra_attr(inode) && 831 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 832 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 833 stat->result_mask |= STATX_BTIME; 834 stat->btime.tv_sec = fi->i_crtime.tv_sec; 835 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 836 } 837 838 /* 839 * Return the DIO alignment restrictions if requested. We only return 840 * this information when requested, since on encrypted files it might 841 * take a fair bit of work to get if the file wasn't opened recently. 842 * 843 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 844 * cannot represent that, so in that case we report no DIO support. 845 */ 846 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 847 unsigned int bsize = i_blocksize(inode); 848 849 stat->result_mask |= STATX_DIOALIGN; 850 if (!f2fs_force_buffered_io(inode, WRITE)) { 851 stat->dio_mem_align = bsize; 852 stat->dio_offset_align = bsize; 853 } 854 } 855 856 flags = fi->i_flags; 857 if (flags & F2FS_COMPR_FL) 858 stat->attributes |= STATX_ATTR_COMPRESSED; 859 if (flags & F2FS_APPEND_FL) 860 stat->attributes |= STATX_ATTR_APPEND; 861 if (IS_ENCRYPTED(inode)) 862 stat->attributes |= STATX_ATTR_ENCRYPTED; 863 if (flags & F2FS_IMMUTABLE_FL) 864 stat->attributes |= STATX_ATTR_IMMUTABLE; 865 if (flags & F2FS_NODUMP_FL) 866 stat->attributes |= STATX_ATTR_NODUMP; 867 if (IS_VERITY(inode)) 868 stat->attributes |= STATX_ATTR_VERITY; 869 870 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 871 STATX_ATTR_APPEND | 872 STATX_ATTR_ENCRYPTED | 873 STATX_ATTR_IMMUTABLE | 874 STATX_ATTR_NODUMP | 875 STATX_ATTR_VERITY); 876 877 generic_fillattr(idmap, request_mask, inode, stat); 878 879 /* we need to show initial sectors used for inline_data/dentries */ 880 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 881 f2fs_has_inline_dentry(inode)) 882 stat->blocks += (stat->size + 511) >> 9; 883 884 return 0; 885 } 886 887 #ifdef CONFIG_F2FS_FS_POSIX_ACL 888 static void __setattr_copy(struct mnt_idmap *idmap, 889 struct inode *inode, const struct iattr *attr) 890 { 891 unsigned int ia_valid = attr->ia_valid; 892 893 i_uid_update(idmap, attr, inode); 894 i_gid_update(idmap, attr, inode); 895 if (ia_valid & ATTR_ATIME) 896 inode->i_atime = attr->ia_atime; 897 if (ia_valid & ATTR_MTIME) 898 inode->i_mtime = attr->ia_mtime; 899 if (ia_valid & ATTR_CTIME) 900 inode_set_ctime_to_ts(inode, attr->ia_ctime); 901 if (ia_valid & ATTR_MODE) { 902 umode_t mode = attr->ia_mode; 903 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode); 904 905 if (!vfsgid_in_group_p(vfsgid) && 906 !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) 907 mode &= ~S_ISGID; 908 set_acl_inode(inode, mode); 909 } 910 } 911 #else 912 #define __setattr_copy setattr_copy 913 #endif 914 915 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 916 struct iattr *attr) 917 { 918 struct inode *inode = d_inode(dentry); 919 int err; 920 921 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 922 return -EIO; 923 924 if (unlikely(IS_IMMUTABLE(inode))) 925 return -EPERM; 926 927 if (unlikely(IS_APPEND(inode) && 928 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 929 ATTR_GID | ATTR_TIMES_SET)))) 930 return -EPERM; 931 932 if ((attr->ia_valid & ATTR_SIZE)) { 933 if (!f2fs_is_compress_backend_ready(inode)) 934 return -EOPNOTSUPP; 935 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 936 !IS_ALIGNED(attr->ia_size, 937 F2FS_BLK_TO_BYTES(F2FS_I(inode)->i_cluster_size))) 938 return -EINVAL; 939 } 940 941 err = setattr_prepare(idmap, dentry, attr); 942 if (err) 943 return err; 944 945 err = fscrypt_prepare_setattr(dentry, attr); 946 if (err) 947 return err; 948 949 err = fsverity_prepare_setattr(dentry, attr); 950 if (err) 951 return err; 952 953 if (is_quota_modification(idmap, inode, attr)) { 954 err = f2fs_dquot_initialize(inode); 955 if (err) 956 return err; 957 } 958 if (i_uid_needs_update(idmap, attr, inode) || 959 i_gid_needs_update(idmap, attr, inode)) { 960 f2fs_lock_op(F2FS_I_SB(inode)); 961 err = dquot_transfer(idmap, inode, attr); 962 if (err) { 963 set_sbi_flag(F2FS_I_SB(inode), 964 SBI_QUOTA_NEED_REPAIR); 965 f2fs_unlock_op(F2FS_I_SB(inode)); 966 return err; 967 } 968 /* 969 * update uid/gid under lock_op(), so that dquot and inode can 970 * be updated atomically. 971 */ 972 i_uid_update(idmap, attr, inode); 973 i_gid_update(idmap, attr, inode); 974 f2fs_mark_inode_dirty_sync(inode, true); 975 f2fs_unlock_op(F2FS_I_SB(inode)); 976 } 977 978 if (attr->ia_valid & ATTR_SIZE) { 979 loff_t old_size = i_size_read(inode); 980 981 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 982 /* 983 * should convert inline inode before i_size_write to 984 * keep smaller than inline_data size with inline flag. 985 */ 986 err = f2fs_convert_inline_inode(inode); 987 if (err) 988 return err; 989 } 990 991 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 992 filemap_invalidate_lock(inode->i_mapping); 993 994 truncate_setsize(inode, attr->ia_size); 995 996 if (attr->ia_size <= old_size) 997 err = f2fs_truncate(inode); 998 /* 999 * do not trim all blocks after i_size if target size is 1000 * larger than i_size. 1001 */ 1002 filemap_invalidate_unlock(inode->i_mapping); 1003 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1004 if (err) 1005 return err; 1006 1007 spin_lock(&F2FS_I(inode)->i_size_lock); 1008 inode->i_mtime = inode_set_ctime_current(inode); 1009 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1010 spin_unlock(&F2FS_I(inode)->i_size_lock); 1011 } 1012 1013 __setattr_copy(idmap, inode, attr); 1014 1015 if (attr->ia_valid & ATTR_MODE) { 1016 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1017 1018 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1019 if (!err) 1020 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1021 clear_inode_flag(inode, FI_ACL_MODE); 1022 } 1023 } 1024 1025 /* file size may changed here */ 1026 f2fs_mark_inode_dirty_sync(inode, true); 1027 1028 /* inode change will produce dirty node pages flushed by checkpoint */ 1029 f2fs_balance_fs(F2FS_I_SB(inode), true); 1030 1031 return err; 1032 } 1033 1034 const struct inode_operations f2fs_file_inode_operations = { 1035 .getattr = f2fs_getattr, 1036 .setattr = f2fs_setattr, 1037 .get_inode_acl = f2fs_get_acl, 1038 .set_acl = f2fs_set_acl, 1039 .listxattr = f2fs_listxattr, 1040 .fiemap = f2fs_fiemap, 1041 .fileattr_get = f2fs_fileattr_get, 1042 .fileattr_set = f2fs_fileattr_set, 1043 }; 1044 1045 static int fill_zero(struct inode *inode, pgoff_t index, 1046 loff_t start, loff_t len) 1047 { 1048 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1049 struct page *page; 1050 1051 if (!len) 1052 return 0; 1053 1054 f2fs_balance_fs(sbi, true); 1055 1056 f2fs_lock_op(sbi); 1057 page = f2fs_get_new_data_page(inode, NULL, index, false); 1058 f2fs_unlock_op(sbi); 1059 1060 if (IS_ERR(page)) 1061 return PTR_ERR(page); 1062 1063 f2fs_wait_on_page_writeback(page, DATA, true, true); 1064 zero_user(page, start, len); 1065 set_page_dirty(page); 1066 f2fs_put_page(page, 1); 1067 return 0; 1068 } 1069 1070 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1071 { 1072 int err; 1073 1074 while (pg_start < pg_end) { 1075 struct dnode_of_data dn; 1076 pgoff_t end_offset, count; 1077 1078 set_new_dnode(&dn, inode, NULL, NULL, 0); 1079 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1080 if (err) { 1081 if (err == -ENOENT) { 1082 pg_start = f2fs_get_next_page_offset(&dn, 1083 pg_start); 1084 continue; 1085 } 1086 return err; 1087 } 1088 1089 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1090 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1091 1092 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1093 1094 f2fs_truncate_data_blocks_range(&dn, count); 1095 f2fs_put_dnode(&dn); 1096 1097 pg_start += count; 1098 } 1099 return 0; 1100 } 1101 1102 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1103 { 1104 pgoff_t pg_start, pg_end; 1105 loff_t off_start, off_end; 1106 int ret; 1107 1108 ret = f2fs_convert_inline_inode(inode); 1109 if (ret) 1110 return ret; 1111 1112 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1113 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1114 1115 off_start = offset & (PAGE_SIZE - 1); 1116 off_end = (offset + len) & (PAGE_SIZE - 1); 1117 1118 if (pg_start == pg_end) { 1119 ret = fill_zero(inode, pg_start, off_start, 1120 off_end - off_start); 1121 if (ret) 1122 return ret; 1123 } else { 1124 if (off_start) { 1125 ret = fill_zero(inode, pg_start++, off_start, 1126 PAGE_SIZE - off_start); 1127 if (ret) 1128 return ret; 1129 } 1130 if (off_end) { 1131 ret = fill_zero(inode, pg_end, 0, off_end); 1132 if (ret) 1133 return ret; 1134 } 1135 1136 if (pg_start < pg_end) { 1137 loff_t blk_start, blk_end; 1138 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1139 1140 f2fs_balance_fs(sbi, true); 1141 1142 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1143 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1144 1145 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1146 filemap_invalidate_lock(inode->i_mapping); 1147 1148 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1149 1150 f2fs_lock_op(sbi); 1151 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1152 f2fs_unlock_op(sbi); 1153 1154 filemap_invalidate_unlock(inode->i_mapping); 1155 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1156 } 1157 } 1158 1159 return ret; 1160 } 1161 1162 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1163 int *do_replace, pgoff_t off, pgoff_t len) 1164 { 1165 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1166 struct dnode_of_data dn; 1167 int ret, done, i; 1168 1169 next_dnode: 1170 set_new_dnode(&dn, inode, NULL, NULL, 0); 1171 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1172 if (ret && ret != -ENOENT) { 1173 return ret; 1174 } else if (ret == -ENOENT) { 1175 if (dn.max_level == 0) 1176 return -ENOENT; 1177 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1178 dn.ofs_in_node, len); 1179 blkaddr += done; 1180 do_replace += done; 1181 goto next; 1182 } 1183 1184 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1185 dn.ofs_in_node, len); 1186 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1187 *blkaddr = f2fs_data_blkaddr(&dn); 1188 1189 if (__is_valid_data_blkaddr(*blkaddr) && 1190 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1191 DATA_GENERIC_ENHANCE)) { 1192 f2fs_put_dnode(&dn); 1193 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1194 return -EFSCORRUPTED; 1195 } 1196 1197 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1198 1199 if (f2fs_lfs_mode(sbi)) { 1200 f2fs_put_dnode(&dn); 1201 return -EOPNOTSUPP; 1202 } 1203 1204 /* do not invalidate this block address */ 1205 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1206 *do_replace = 1; 1207 } 1208 } 1209 f2fs_put_dnode(&dn); 1210 next: 1211 len -= done; 1212 off += done; 1213 if (len) 1214 goto next_dnode; 1215 return 0; 1216 } 1217 1218 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1219 int *do_replace, pgoff_t off, int len) 1220 { 1221 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1222 struct dnode_of_data dn; 1223 int ret, i; 1224 1225 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1226 if (*do_replace == 0) 1227 continue; 1228 1229 set_new_dnode(&dn, inode, NULL, NULL, 0); 1230 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1231 if (ret) { 1232 dec_valid_block_count(sbi, inode, 1); 1233 f2fs_invalidate_blocks(sbi, *blkaddr); 1234 } else { 1235 f2fs_update_data_blkaddr(&dn, *blkaddr); 1236 } 1237 f2fs_put_dnode(&dn); 1238 } 1239 return 0; 1240 } 1241 1242 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1243 block_t *blkaddr, int *do_replace, 1244 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1245 { 1246 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1247 pgoff_t i = 0; 1248 int ret; 1249 1250 while (i < len) { 1251 if (blkaddr[i] == NULL_ADDR && !full) { 1252 i++; 1253 continue; 1254 } 1255 1256 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1257 struct dnode_of_data dn; 1258 struct node_info ni; 1259 size_t new_size; 1260 pgoff_t ilen; 1261 1262 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1263 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1264 if (ret) 1265 return ret; 1266 1267 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1268 if (ret) { 1269 f2fs_put_dnode(&dn); 1270 return ret; 1271 } 1272 1273 ilen = min((pgoff_t) 1274 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1275 dn.ofs_in_node, len - i); 1276 do { 1277 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1278 f2fs_truncate_data_blocks_range(&dn, 1); 1279 1280 if (do_replace[i]) { 1281 f2fs_i_blocks_write(src_inode, 1282 1, false, false); 1283 f2fs_i_blocks_write(dst_inode, 1284 1, true, false); 1285 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1286 blkaddr[i], ni.version, true, false); 1287 1288 do_replace[i] = 0; 1289 } 1290 dn.ofs_in_node++; 1291 i++; 1292 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1293 if (dst_inode->i_size < new_size) 1294 f2fs_i_size_write(dst_inode, new_size); 1295 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1296 1297 f2fs_put_dnode(&dn); 1298 } else { 1299 struct page *psrc, *pdst; 1300 1301 psrc = f2fs_get_lock_data_page(src_inode, 1302 src + i, true); 1303 if (IS_ERR(psrc)) 1304 return PTR_ERR(psrc); 1305 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1306 true); 1307 if (IS_ERR(pdst)) { 1308 f2fs_put_page(psrc, 1); 1309 return PTR_ERR(pdst); 1310 } 1311 1312 f2fs_wait_on_page_writeback(pdst, DATA, true, true); 1313 1314 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1315 set_page_dirty(pdst); 1316 set_page_private_gcing(pdst); 1317 f2fs_put_page(pdst, 1); 1318 f2fs_put_page(psrc, 1); 1319 1320 ret = f2fs_truncate_hole(src_inode, 1321 src + i, src + i + 1); 1322 if (ret) 1323 return ret; 1324 i++; 1325 } 1326 } 1327 return 0; 1328 } 1329 1330 static int __exchange_data_block(struct inode *src_inode, 1331 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1332 pgoff_t len, bool full) 1333 { 1334 block_t *src_blkaddr; 1335 int *do_replace; 1336 pgoff_t olen; 1337 int ret; 1338 1339 while (len) { 1340 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1341 1342 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1343 array_size(olen, sizeof(block_t)), 1344 GFP_NOFS); 1345 if (!src_blkaddr) 1346 return -ENOMEM; 1347 1348 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1349 array_size(olen, sizeof(int)), 1350 GFP_NOFS); 1351 if (!do_replace) { 1352 kvfree(src_blkaddr); 1353 return -ENOMEM; 1354 } 1355 1356 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1357 do_replace, src, olen); 1358 if (ret) 1359 goto roll_back; 1360 1361 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1362 do_replace, src, dst, olen, full); 1363 if (ret) 1364 goto roll_back; 1365 1366 src += olen; 1367 dst += olen; 1368 len -= olen; 1369 1370 kvfree(src_blkaddr); 1371 kvfree(do_replace); 1372 } 1373 return 0; 1374 1375 roll_back: 1376 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1377 kvfree(src_blkaddr); 1378 kvfree(do_replace); 1379 return ret; 1380 } 1381 1382 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1383 { 1384 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1385 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1386 pgoff_t start = offset >> PAGE_SHIFT; 1387 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1388 int ret; 1389 1390 f2fs_balance_fs(sbi, true); 1391 1392 /* avoid gc operation during block exchange */ 1393 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1394 filemap_invalidate_lock(inode->i_mapping); 1395 1396 f2fs_lock_op(sbi); 1397 f2fs_drop_extent_tree(inode); 1398 truncate_pagecache(inode, offset); 1399 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1400 f2fs_unlock_op(sbi); 1401 1402 filemap_invalidate_unlock(inode->i_mapping); 1403 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1404 return ret; 1405 } 1406 1407 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1408 { 1409 loff_t new_size; 1410 int ret; 1411 1412 if (offset + len >= i_size_read(inode)) 1413 return -EINVAL; 1414 1415 /* collapse range should be aligned to block size of f2fs. */ 1416 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1417 return -EINVAL; 1418 1419 ret = f2fs_convert_inline_inode(inode); 1420 if (ret) 1421 return ret; 1422 1423 /* write out all dirty pages from offset */ 1424 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1425 if (ret) 1426 return ret; 1427 1428 ret = f2fs_do_collapse(inode, offset, len); 1429 if (ret) 1430 return ret; 1431 1432 /* write out all moved pages, if possible */ 1433 filemap_invalidate_lock(inode->i_mapping); 1434 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1435 truncate_pagecache(inode, offset); 1436 1437 new_size = i_size_read(inode) - len; 1438 ret = f2fs_truncate_blocks(inode, new_size, true); 1439 filemap_invalidate_unlock(inode->i_mapping); 1440 if (!ret) 1441 f2fs_i_size_write(inode, new_size); 1442 return ret; 1443 } 1444 1445 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1446 pgoff_t end) 1447 { 1448 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1449 pgoff_t index = start; 1450 unsigned int ofs_in_node = dn->ofs_in_node; 1451 blkcnt_t count = 0; 1452 int ret; 1453 1454 for (; index < end; index++, dn->ofs_in_node++) { 1455 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1456 count++; 1457 } 1458 1459 dn->ofs_in_node = ofs_in_node; 1460 ret = f2fs_reserve_new_blocks(dn, count); 1461 if (ret) 1462 return ret; 1463 1464 dn->ofs_in_node = ofs_in_node; 1465 for (index = start; index < end; index++, dn->ofs_in_node++) { 1466 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1467 /* 1468 * f2fs_reserve_new_blocks will not guarantee entire block 1469 * allocation. 1470 */ 1471 if (dn->data_blkaddr == NULL_ADDR) { 1472 ret = -ENOSPC; 1473 break; 1474 } 1475 1476 if (dn->data_blkaddr == NEW_ADDR) 1477 continue; 1478 1479 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1480 DATA_GENERIC_ENHANCE)) { 1481 ret = -EFSCORRUPTED; 1482 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1483 break; 1484 } 1485 1486 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1487 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1488 } 1489 1490 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1491 f2fs_update_age_extent_cache_range(dn, start, index - start); 1492 1493 return ret; 1494 } 1495 1496 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1497 int mode) 1498 { 1499 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1500 struct address_space *mapping = inode->i_mapping; 1501 pgoff_t index, pg_start, pg_end; 1502 loff_t new_size = i_size_read(inode); 1503 loff_t off_start, off_end; 1504 int ret = 0; 1505 1506 ret = inode_newsize_ok(inode, (len + offset)); 1507 if (ret) 1508 return ret; 1509 1510 ret = f2fs_convert_inline_inode(inode); 1511 if (ret) 1512 return ret; 1513 1514 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1515 if (ret) 1516 return ret; 1517 1518 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1519 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1520 1521 off_start = offset & (PAGE_SIZE - 1); 1522 off_end = (offset + len) & (PAGE_SIZE - 1); 1523 1524 if (pg_start == pg_end) { 1525 ret = fill_zero(inode, pg_start, off_start, 1526 off_end - off_start); 1527 if (ret) 1528 return ret; 1529 1530 new_size = max_t(loff_t, new_size, offset + len); 1531 } else { 1532 if (off_start) { 1533 ret = fill_zero(inode, pg_start++, off_start, 1534 PAGE_SIZE - off_start); 1535 if (ret) 1536 return ret; 1537 1538 new_size = max_t(loff_t, new_size, 1539 (loff_t)pg_start << PAGE_SHIFT); 1540 } 1541 1542 for (index = pg_start; index < pg_end;) { 1543 struct dnode_of_data dn; 1544 unsigned int end_offset; 1545 pgoff_t end; 1546 1547 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1548 filemap_invalidate_lock(mapping); 1549 1550 truncate_pagecache_range(inode, 1551 (loff_t)index << PAGE_SHIFT, 1552 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1553 1554 f2fs_lock_op(sbi); 1555 1556 set_new_dnode(&dn, inode, NULL, NULL, 0); 1557 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1558 if (ret) { 1559 f2fs_unlock_op(sbi); 1560 filemap_invalidate_unlock(mapping); 1561 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1562 goto out; 1563 } 1564 1565 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1566 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1567 1568 ret = f2fs_do_zero_range(&dn, index, end); 1569 f2fs_put_dnode(&dn); 1570 1571 f2fs_unlock_op(sbi); 1572 filemap_invalidate_unlock(mapping); 1573 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1574 1575 f2fs_balance_fs(sbi, dn.node_changed); 1576 1577 if (ret) 1578 goto out; 1579 1580 index = end; 1581 new_size = max_t(loff_t, new_size, 1582 (loff_t)index << PAGE_SHIFT); 1583 } 1584 1585 if (off_end) { 1586 ret = fill_zero(inode, pg_end, 0, off_end); 1587 if (ret) 1588 goto out; 1589 1590 new_size = max_t(loff_t, new_size, offset + len); 1591 } 1592 } 1593 1594 out: 1595 if (new_size > i_size_read(inode)) { 1596 if (mode & FALLOC_FL_KEEP_SIZE) 1597 file_set_keep_isize(inode); 1598 else 1599 f2fs_i_size_write(inode, new_size); 1600 } 1601 return ret; 1602 } 1603 1604 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1605 { 1606 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1607 struct address_space *mapping = inode->i_mapping; 1608 pgoff_t nr, pg_start, pg_end, delta, idx; 1609 loff_t new_size; 1610 int ret = 0; 1611 1612 new_size = i_size_read(inode) + len; 1613 ret = inode_newsize_ok(inode, new_size); 1614 if (ret) 1615 return ret; 1616 1617 if (offset >= i_size_read(inode)) 1618 return -EINVAL; 1619 1620 /* insert range should be aligned to block size of f2fs. */ 1621 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1622 return -EINVAL; 1623 1624 ret = f2fs_convert_inline_inode(inode); 1625 if (ret) 1626 return ret; 1627 1628 f2fs_balance_fs(sbi, true); 1629 1630 filemap_invalidate_lock(mapping); 1631 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1632 filemap_invalidate_unlock(mapping); 1633 if (ret) 1634 return ret; 1635 1636 /* write out all dirty pages from offset */ 1637 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1638 if (ret) 1639 return ret; 1640 1641 pg_start = offset >> PAGE_SHIFT; 1642 pg_end = (offset + len) >> PAGE_SHIFT; 1643 delta = pg_end - pg_start; 1644 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1645 1646 /* avoid gc operation during block exchange */ 1647 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1648 filemap_invalidate_lock(mapping); 1649 truncate_pagecache(inode, offset); 1650 1651 while (!ret && idx > pg_start) { 1652 nr = idx - pg_start; 1653 if (nr > delta) 1654 nr = delta; 1655 idx -= nr; 1656 1657 f2fs_lock_op(sbi); 1658 f2fs_drop_extent_tree(inode); 1659 1660 ret = __exchange_data_block(inode, inode, idx, 1661 idx + delta, nr, false); 1662 f2fs_unlock_op(sbi); 1663 } 1664 filemap_invalidate_unlock(mapping); 1665 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1666 1667 /* write out all moved pages, if possible */ 1668 filemap_invalidate_lock(mapping); 1669 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1670 truncate_pagecache(inode, offset); 1671 filemap_invalidate_unlock(mapping); 1672 1673 if (!ret) 1674 f2fs_i_size_write(inode, new_size); 1675 return ret; 1676 } 1677 1678 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1679 loff_t len, int mode) 1680 { 1681 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1682 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1683 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1684 .m_may_create = true }; 1685 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1686 .init_gc_type = FG_GC, 1687 .should_migrate_blocks = false, 1688 .err_gc_skipped = true, 1689 .nr_free_secs = 0 }; 1690 pgoff_t pg_start, pg_end; 1691 loff_t new_size; 1692 loff_t off_end; 1693 block_t expanded = 0; 1694 int err; 1695 1696 err = inode_newsize_ok(inode, (len + offset)); 1697 if (err) 1698 return err; 1699 1700 err = f2fs_convert_inline_inode(inode); 1701 if (err) 1702 return err; 1703 1704 f2fs_balance_fs(sbi, true); 1705 1706 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1707 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1708 off_end = (offset + len) & (PAGE_SIZE - 1); 1709 1710 map.m_lblk = pg_start; 1711 map.m_len = pg_end - pg_start; 1712 if (off_end) 1713 map.m_len++; 1714 1715 if (!map.m_len) 1716 return 0; 1717 1718 if (f2fs_is_pinned_file(inode)) { 1719 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1720 block_t sec_len = roundup(map.m_len, sec_blks); 1721 1722 map.m_len = sec_blks; 1723 next_alloc: 1724 if (has_not_enough_free_secs(sbi, 0, 1725 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1726 f2fs_down_write(&sbi->gc_lock); 1727 stat_inc_gc_call_count(sbi, FOREGROUND); 1728 err = f2fs_gc(sbi, &gc_control); 1729 if (err && err != -ENODATA) 1730 goto out_err; 1731 } 1732 1733 f2fs_down_write(&sbi->pin_sem); 1734 1735 err = f2fs_allocate_pinning_section(sbi); 1736 if (err) { 1737 f2fs_up_write(&sbi->pin_sem); 1738 goto out_err; 1739 } 1740 1741 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1742 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1743 file_dont_truncate(inode); 1744 1745 f2fs_up_write(&sbi->pin_sem); 1746 1747 expanded += map.m_len; 1748 sec_len -= map.m_len; 1749 map.m_lblk += map.m_len; 1750 if (!err && sec_len) 1751 goto next_alloc; 1752 1753 map.m_len = expanded; 1754 } else { 1755 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1756 expanded = map.m_len; 1757 } 1758 out_err: 1759 if (err) { 1760 pgoff_t last_off; 1761 1762 if (!expanded) 1763 return err; 1764 1765 last_off = pg_start + expanded - 1; 1766 1767 /* update new size to the failed position */ 1768 new_size = (last_off == pg_end) ? offset + len : 1769 (loff_t)(last_off + 1) << PAGE_SHIFT; 1770 } else { 1771 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1772 } 1773 1774 if (new_size > i_size_read(inode)) { 1775 if (mode & FALLOC_FL_KEEP_SIZE) 1776 file_set_keep_isize(inode); 1777 else 1778 f2fs_i_size_write(inode, new_size); 1779 } 1780 1781 return err; 1782 } 1783 1784 static long f2fs_fallocate(struct file *file, int mode, 1785 loff_t offset, loff_t len) 1786 { 1787 struct inode *inode = file_inode(file); 1788 long ret = 0; 1789 1790 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1791 return -EIO; 1792 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1793 return -ENOSPC; 1794 if (!f2fs_is_compress_backend_ready(inode)) 1795 return -EOPNOTSUPP; 1796 1797 /* f2fs only support ->fallocate for regular file */ 1798 if (!S_ISREG(inode->i_mode)) 1799 return -EINVAL; 1800 1801 if (IS_ENCRYPTED(inode) && 1802 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1803 return -EOPNOTSUPP; 1804 1805 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1806 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1807 FALLOC_FL_INSERT_RANGE)) 1808 return -EOPNOTSUPP; 1809 1810 inode_lock(inode); 1811 1812 /* 1813 * Pinned file should not support partial truncation since the block 1814 * can be used by applications. 1815 */ 1816 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1817 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1818 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 1819 ret = -EOPNOTSUPP; 1820 goto out; 1821 } 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 total = 0, sec_num; 2587 block_t blk_end = 0; 2588 bool fragmented = false; 2589 int err; 2590 2591 pg_start = range->start >> PAGE_SHIFT; 2592 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2593 2594 f2fs_balance_fs(sbi, true); 2595 2596 inode_lock(inode); 2597 2598 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2599 err = -EINVAL; 2600 goto unlock_out; 2601 } 2602 2603 /* if in-place-update policy is enabled, don't waste time here */ 2604 set_inode_flag(inode, FI_OPU_WRITE); 2605 if (f2fs_should_update_inplace(inode, NULL)) { 2606 err = -EINVAL; 2607 goto out; 2608 } 2609 2610 /* writeback all dirty pages in the range */ 2611 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2612 range->start + range->len - 1); 2613 if (err) 2614 goto out; 2615 2616 /* 2617 * lookup mapping info in extent cache, skip defragmenting if physical 2618 * block addresses are continuous. 2619 */ 2620 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2621 if (ei.fofs + ei.len >= pg_end) 2622 goto out; 2623 } 2624 2625 map.m_lblk = pg_start; 2626 map.m_next_pgofs = &next_pgofs; 2627 2628 /* 2629 * lookup mapping info in dnode page cache, skip defragmenting if all 2630 * physical block addresses are continuous even if there are hole(s) 2631 * in logical blocks. 2632 */ 2633 while (map.m_lblk < pg_end) { 2634 map.m_len = pg_end - map.m_lblk; 2635 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2636 if (err) 2637 goto out; 2638 2639 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2640 map.m_lblk = next_pgofs; 2641 continue; 2642 } 2643 2644 if (blk_end && blk_end != map.m_pblk) 2645 fragmented = true; 2646 2647 /* record total count of block that we're going to move */ 2648 total += map.m_len; 2649 2650 blk_end = map.m_pblk + map.m_len; 2651 2652 map.m_lblk += map.m_len; 2653 } 2654 2655 if (!fragmented) { 2656 total = 0; 2657 goto out; 2658 } 2659 2660 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2661 2662 /* 2663 * make sure there are enough free section for LFS allocation, this can 2664 * avoid defragment running in SSR mode when free section are allocated 2665 * intensively 2666 */ 2667 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2668 err = -EAGAIN; 2669 goto out; 2670 } 2671 2672 map.m_lblk = pg_start; 2673 map.m_len = pg_end - pg_start; 2674 total = 0; 2675 2676 while (map.m_lblk < pg_end) { 2677 pgoff_t idx; 2678 int cnt = 0; 2679 2680 do_map: 2681 map.m_len = pg_end - map.m_lblk; 2682 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2683 if (err) 2684 goto clear_out; 2685 2686 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2687 map.m_lblk = next_pgofs; 2688 goto check; 2689 } 2690 2691 set_inode_flag(inode, FI_SKIP_WRITES); 2692 2693 idx = map.m_lblk; 2694 while (idx < map.m_lblk + map.m_len && 2695 cnt < BLKS_PER_SEG(sbi)) { 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 < BLKS_PER_SEG(sbi)) 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 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 2823 ret = -EOPNOTSUPP; 2824 goto out_unlock; 2825 } 2826 2827 ret = -EINVAL; 2828 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2829 goto out_unlock; 2830 if (len == 0) 2831 olen = len = src->i_size - pos_in; 2832 if (pos_in + len == src->i_size) 2833 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2834 if (len == 0) { 2835 ret = 0; 2836 goto out_unlock; 2837 } 2838 2839 dst_osize = dst->i_size; 2840 if (pos_out + olen > dst->i_size) 2841 dst_max_i_size = pos_out + olen; 2842 2843 /* verify the end result is block aligned */ 2844 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2845 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2846 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2847 goto out_unlock; 2848 2849 ret = f2fs_convert_inline_inode(src); 2850 if (ret) 2851 goto out_unlock; 2852 2853 ret = f2fs_convert_inline_inode(dst); 2854 if (ret) 2855 goto out_unlock; 2856 2857 /* write out all dirty pages from offset */ 2858 ret = filemap_write_and_wait_range(src->i_mapping, 2859 pos_in, pos_in + len); 2860 if (ret) 2861 goto out_unlock; 2862 2863 ret = filemap_write_and_wait_range(dst->i_mapping, 2864 pos_out, pos_out + len); 2865 if (ret) 2866 goto out_unlock; 2867 2868 f2fs_balance_fs(sbi, true); 2869 2870 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2871 if (src != dst) { 2872 ret = -EBUSY; 2873 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2874 goto out_src; 2875 } 2876 2877 f2fs_lock_op(sbi); 2878 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2879 pos_out >> F2FS_BLKSIZE_BITS, 2880 len >> F2FS_BLKSIZE_BITS, false); 2881 2882 if (!ret) { 2883 if (dst_max_i_size) 2884 f2fs_i_size_write(dst, dst_max_i_size); 2885 else if (dst_osize != dst->i_size) 2886 f2fs_i_size_write(dst, dst_osize); 2887 } 2888 f2fs_unlock_op(sbi); 2889 2890 if (src != dst) 2891 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2892 out_src: 2893 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2894 if (ret) 2895 goto out_unlock; 2896 2897 src->i_mtime = inode_set_ctime_current(src); 2898 f2fs_mark_inode_dirty_sync(src, false); 2899 if (src != dst) { 2900 dst->i_mtime = inode_set_ctime_current(dst); 2901 f2fs_mark_inode_dirty_sync(dst, false); 2902 } 2903 f2fs_update_time(sbi, REQ_TIME); 2904 2905 out_unlock: 2906 if (src != dst) 2907 inode_unlock(dst); 2908 out: 2909 inode_unlock(src); 2910 return ret; 2911 } 2912 2913 static int __f2fs_ioc_move_range(struct file *filp, 2914 struct f2fs_move_range *range) 2915 { 2916 struct fd dst; 2917 int err; 2918 2919 if (!(filp->f_mode & FMODE_READ) || 2920 !(filp->f_mode & FMODE_WRITE)) 2921 return -EBADF; 2922 2923 dst = fdget(range->dst_fd); 2924 if (!dst.file) 2925 return -EBADF; 2926 2927 if (!(dst.file->f_mode & FMODE_WRITE)) { 2928 err = -EBADF; 2929 goto err_out; 2930 } 2931 2932 err = mnt_want_write_file(filp); 2933 if (err) 2934 goto err_out; 2935 2936 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2937 range->pos_out, range->len); 2938 2939 mnt_drop_write_file(filp); 2940 err_out: 2941 fdput(dst); 2942 return err; 2943 } 2944 2945 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2946 { 2947 struct f2fs_move_range range; 2948 2949 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2950 sizeof(range))) 2951 return -EFAULT; 2952 return __f2fs_ioc_move_range(filp, &range); 2953 } 2954 2955 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2956 { 2957 struct inode *inode = file_inode(filp); 2958 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2959 struct sit_info *sm = SIT_I(sbi); 2960 unsigned int start_segno = 0, end_segno = 0; 2961 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2962 struct f2fs_flush_device range; 2963 struct f2fs_gc_control gc_control = { 2964 .init_gc_type = FG_GC, 2965 .should_migrate_blocks = true, 2966 .err_gc_skipped = true, 2967 .nr_free_secs = 0 }; 2968 int ret; 2969 2970 if (!capable(CAP_SYS_ADMIN)) 2971 return -EPERM; 2972 2973 if (f2fs_readonly(sbi->sb)) 2974 return -EROFS; 2975 2976 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2977 return -EINVAL; 2978 2979 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2980 sizeof(range))) 2981 return -EFAULT; 2982 2983 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2984 __is_large_section(sbi)) { 2985 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 2986 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 2987 return -EINVAL; 2988 } 2989 2990 ret = mnt_want_write_file(filp); 2991 if (ret) 2992 return ret; 2993 2994 if (range.dev_num != 0) 2995 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2996 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2997 2998 start_segno = sm->last_victim[FLUSH_DEVICE]; 2999 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3000 start_segno = dev_start_segno; 3001 end_segno = min(start_segno + range.segments, dev_end_segno); 3002 3003 while (start_segno < end_segno) { 3004 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3005 ret = -EBUSY; 3006 goto out; 3007 } 3008 sm->last_victim[GC_CB] = end_segno + 1; 3009 sm->last_victim[GC_GREEDY] = end_segno + 1; 3010 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3011 3012 gc_control.victim_segno = start_segno; 3013 stat_inc_gc_call_count(sbi, FOREGROUND); 3014 ret = f2fs_gc(sbi, &gc_control); 3015 if (ret == -EAGAIN) 3016 ret = 0; 3017 else if (ret < 0) 3018 break; 3019 start_segno++; 3020 } 3021 out: 3022 mnt_drop_write_file(filp); 3023 return ret; 3024 } 3025 3026 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3027 { 3028 struct inode *inode = file_inode(filp); 3029 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3030 3031 /* Must validate to set it with SQLite behavior in Android. */ 3032 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3033 3034 return put_user(sb_feature, (u32 __user *)arg); 3035 } 3036 3037 #ifdef CONFIG_QUOTA 3038 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3039 { 3040 struct dquot *transfer_to[MAXQUOTAS] = {}; 3041 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3042 struct super_block *sb = sbi->sb; 3043 int err; 3044 3045 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3046 if (IS_ERR(transfer_to[PRJQUOTA])) 3047 return PTR_ERR(transfer_to[PRJQUOTA]); 3048 3049 err = __dquot_transfer(inode, transfer_to); 3050 if (err) 3051 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3052 dqput(transfer_to[PRJQUOTA]); 3053 return err; 3054 } 3055 3056 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3057 { 3058 struct f2fs_inode_info *fi = F2FS_I(inode); 3059 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3060 struct f2fs_inode *ri = NULL; 3061 kprojid_t kprojid; 3062 int err; 3063 3064 if (!f2fs_sb_has_project_quota(sbi)) { 3065 if (projid != F2FS_DEF_PROJID) 3066 return -EOPNOTSUPP; 3067 else 3068 return 0; 3069 } 3070 3071 if (!f2fs_has_extra_attr(inode)) 3072 return -EOPNOTSUPP; 3073 3074 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3075 3076 if (projid_eq(kprojid, fi->i_projid)) 3077 return 0; 3078 3079 err = -EPERM; 3080 /* Is it quota file? Do not allow user to mess with it */ 3081 if (IS_NOQUOTA(inode)) 3082 return err; 3083 3084 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3085 return -EOVERFLOW; 3086 3087 err = f2fs_dquot_initialize(inode); 3088 if (err) 3089 return err; 3090 3091 f2fs_lock_op(sbi); 3092 err = f2fs_transfer_project_quota(inode, kprojid); 3093 if (err) 3094 goto out_unlock; 3095 3096 fi->i_projid = kprojid; 3097 inode_set_ctime_current(inode); 3098 f2fs_mark_inode_dirty_sync(inode, true); 3099 out_unlock: 3100 f2fs_unlock_op(sbi); 3101 return err; 3102 } 3103 #else 3104 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3105 { 3106 return 0; 3107 } 3108 3109 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3110 { 3111 if (projid != F2FS_DEF_PROJID) 3112 return -EOPNOTSUPP; 3113 return 0; 3114 } 3115 #endif 3116 3117 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3118 { 3119 struct inode *inode = d_inode(dentry); 3120 struct f2fs_inode_info *fi = F2FS_I(inode); 3121 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3122 3123 if (IS_ENCRYPTED(inode)) 3124 fsflags |= FS_ENCRYPT_FL; 3125 if (IS_VERITY(inode)) 3126 fsflags |= FS_VERITY_FL; 3127 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3128 fsflags |= FS_INLINE_DATA_FL; 3129 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3130 fsflags |= FS_NOCOW_FL; 3131 3132 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3133 3134 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3135 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3136 3137 return 0; 3138 } 3139 3140 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3141 struct dentry *dentry, struct fileattr *fa) 3142 { 3143 struct inode *inode = d_inode(dentry); 3144 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3145 u32 iflags; 3146 int err; 3147 3148 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3149 return -EIO; 3150 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3151 return -ENOSPC; 3152 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3153 return -EOPNOTSUPP; 3154 fsflags &= F2FS_SETTABLE_FS_FL; 3155 if (!fa->flags_valid) 3156 mask &= FS_COMMON_FL; 3157 3158 iflags = f2fs_fsflags_to_iflags(fsflags); 3159 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3160 return -EOPNOTSUPP; 3161 3162 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3163 if (!err) 3164 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3165 3166 return err; 3167 } 3168 3169 int f2fs_pin_file_control(struct inode *inode, bool inc) 3170 { 3171 struct f2fs_inode_info *fi = F2FS_I(inode); 3172 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3173 3174 /* Use i_gc_failures for normal file as a risk signal. */ 3175 if (inc) 3176 f2fs_i_gc_failures_write(inode, 3177 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3178 3179 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3180 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3181 __func__, inode->i_ino, 3182 fi->i_gc_failures[GC_FAILURE_PIN]); 3183 clear_inode_flag(inode, FI_PIN_FILE); 3184 return -EAGAIN; 3185 } 3186 return 0; 3187 } 3188 3189 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3190 { 3191 struct inode *inode = file_inode(filp); 3192 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3193 __u32 pin; 3194 int ret = 0; 3195 3196 if (get_user(pin, (__u32 __user *)arg)) 3197 return -EFAULT; 3198 3199 if (!S_ISREG(inode->i_mode)) 3200 return -EINVAL; 3201 3202 if (f2fs_readonly(sbi->sb)) 3203 return -EROFS; 3204 3205 ret = mnt_want_write_file(filp); 3206 if (ret) 3207 return ret; 3208 3209 inode_lock(inode); 3210 3211 if (!pin) { 3212 clear_inode_flag(inode, FI_PIN_FILE); 3213 f2fs_i_gc_failures_write(inode, 0); 3214 goto done; 3215 } else if (f2fs_is_pinned_file(inode)) { 3216 goto done; 3217 } 3218 3219 if (f2fs_sb_has_blkzoned(sbi) && F2FS_HAS_BLOCKS(inode)) { 3220 ret = -EFBIG; 3221 goto out; 3222 } 3223 3224 /* Let's allow file pinning on zoned device. */ 3225 if (!f2fs_sb_has_blkzoned(sbi) && 3226 f2fs_should_update_outplace(inode, NULL)) { 3227 ret = -EINVAL; 3228 goto out; 3229 } 3230 3231 if (f2fs_pin_file_control(inode, false)) { 3232 ret = -EAGAIN; 3233 goto out; 3234 } 3235 3236 ret = f2fs_convert_inline_inode(inode); 3237 if (ret) 3238 goto out; 3239 3240 if (!f2fs_disable_compressed_file(inode)) { 3241 ret = -EOPNOTSUPP; 3242 goto out; 3243 } 3244 3245 set_inode_flag(inode, FI_PIN_FILE); 3246 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3247 done: 3248 f2fs_update_time(sbi, REQ_TIME); 3249 out: 3250 inode_unlock(inode); 3251 mnt_drop_write_file(filp); 3252 return ret; 3253 } 3254 3255 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3256 { 3257 struct inode *inode = file_inode(filp); 3258 __u32 pin = 0; 3259 3260 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3261 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3262 return put_user(pin, (u32 __user *)arg); 3263 } 3264 3265 int f2fs_precache_extents(struct inode *inode) 3266 { 3267 struct f2fs_inode_info *fi = F2FS_I(inode); 3268 struct f2fs_map_blocks map; 3269 pgoff_t m_next_extent; 3270 loff_t end; 3271 int err; 3272 3273 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3274 return -EOPNOTSUPP; 3275 3276 map.m_lblk = 0; 3277 map.m_pblk = 0; 3278 map.m_next_pgofs = NULL; 3279 map.m_next_extent = &m_next_extent; 3280 map.m_seg_type = NO_CHECK_TYPE; 3281 map.m_may_create = false; 3282 end = max_file_blocks(inode); 3283 3284 while (map.m_lblk < end) { 3285 map.m_len = end - map.m_lblk; 3286 3287 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3288 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3289 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3290 if (err) 3291 return err; 3292 3293 map.m_lblk = m_next_extent; 3294 } 3295 3296 return 0; 3297 } 3298 3299 static int f2fs_ioc_precache_extents(struct file *filp) 3300 { 3301 return f2fs_precache_extents(file_inode(filp)); 3302 } 3303 3304 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3305 { 3306 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3307 __u64 block_count; 3308 3309 if (!capable(CAP_SYS_ADMIN)) 3310 return -EPERM; 3311 3312 if (f2fs_readonly(sbi->sb)) 3313 return -EROFS; 3314 3315 if (copy_from_user(&block_count, (void __user *)arg, 3316 sizeof(block_count))) 3317 return -EFAULT; 3318 3319 return f2fs_resize_fs(filp, block_count); 3320 } 3321 3322 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3323 { 3324 struct inode *inode = file_inode(filp); 3325 3326 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3327 3328 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3329 f2fs_warn(F2FS_I_SB(inode), 3330 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3331 inode->i_ino); 3332 return -EOPNOTSUPP; 3333 } 3334 3335 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3336 } 3337 3338 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3339 { 3340 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3341 return -EOPNOTSUPP; 3342 3343 return fsverity_ioctl_measure(filp, (void __user *)arg); 3344 } 3345 3346 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3347 { 3348 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3349 return -EOPNOTSUPP; 3350 3351 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3352 } 3353 3354 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3355 { 3356 struct inode *inode = file_inode(filp); 3357 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3358 char *vbuf; 3359 int count; 3360 int err = 0; 3361 3362 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3363 if (!vbuf) 3364 return -ENOMEM; 3365 3366 f2fs_down_read(&sbi->sb_lock); 3367 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3368 ARRAY_SIZE(sbi->raw_super->volume_name), 3369 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3370 f2fs_up_read(&sbi->sb_lock); 3371 3372 if (copy_to_user((char __user *)arg, vbuf, 3373 min(FSLABEL_MAX, count))) 3374 err = -EFAULT; 3375 3376 kfree(vbuf); 3377 return err; 3378 } 3379 3380 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3381 { 3382 struct inode *inode = file_inode(filp); 3383 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3384 char *vbuf; 3385 int err = 0; 3386 3387 if (!capable(CAP_SYS_ADMIN)) 3388 return -EPERM; 3389 3390 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3391 if (IS_ERR(vbuf)) 3392 return PTR_ERR(vbuf); 3393 3394 err = mnt_want_write_file(filp); 3395 if (err) 3396 goto out; 3397 3398 f2fs_down_write(&sbi->sb_lock); 3399 3400 memset(sbi->raw_super->volume_name, 0, 3401 sizeof(sbi->raw_super->volume_name)); 3402 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3403 sbi->raw_super->volume_name, 3404 ARRAY_SIZE(sbi->raw_super->volume_name)); 3405 3406 err = f2fs_commit_super(sbi, false); 3407 3408 f2fs_up_write(&sbi->sb_lock); 3409 3410 mnt_drop_write_file(filp); 3411 out: 3412 kfree(vbuf); 3413 return err; 3414 } 3415 3416 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3417 { 3418 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3419 return -EOPNOTSUPP; 3420 3421 if (!f2fs_compressed_file(inode)) 3422 return -EINVAL; 3423 3424 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3425 3426 return 0; 3427 } 3428 3429 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3430 { 3431 struct inode *inode = file_inode(filp); 3432 __u64 blocks; 3433 int ret; 3434 3435 ret = f2fs_get_compress_blocks(inode, &blocks); 3436 if (ret < 0) 3437 return ret; 3438 3439 return put_user(blocks, (u64 __user *)arg); 3440 } 3441 3442 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3443 { 3444 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3445 unsigned int released_blocks = 0; 3446 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3447 block_t blkaddr; 3448 int i; 3449 3450 for (i = 0; i < count; i++) { 3451 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3452 dn->ofs_in_node + i); 3453 3454 if (!__is_valid_data_blkaddr(blkaddr)) 3455 continue; 3456 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3457 DATA_GENERIC_ENHANCE))) { 3458 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3459 return -EFSCORRUPTED; 3460 } 3461 } 3462 3463 while (count) { 3464 int compr_blocks = 0; 3465 3466 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3467 blkaddr = f2fs_data_blkaddr(dn); 3468 3469 if (i == 0) { 3470 if (blkaddr == COMPRESS_ADDR) 3471 continue; 3472 dn->ofs_in_node += cluster_size; 3473 goto next; 3474 } 3475 3476 if (__is_valid_data_blkaddr(blkaddr)) 3477 compr_blocks++; 3478 3479 if (blkaddr != NEW_ADDR) 3480 continue; 3481 3482 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3483 } 3484 3485 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3486 dec_valid_block_count(sbi, dn->inode, 3487 cluster_size - compr_blocks); 3488 3489 released_blocks += cluster_size - compr_blocks; 3490 next: 3491 count -= cluster_size; 3492 } 3493 3494 return released_blocks; 3495 } 3496 3497 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3498 { 3499 struct inode *inode = file_inode(filp); 3500 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3501 pgoff_t page_idx = 0, last_idx; 3502 unsigned int released_blocks = 0; 3503 int ret; 3504 int writecount; 3505 3506 if (!f2fs_sb_has_compression(sbi)) 3507 return -EOPNOTSUPP; 3508 3509 if (f2fs_readonly(sbi->sb)) 3510 return -EROFS; 3511 3512 ret = mnt_want_write_file(filp); 3513 if (ret) 3514 return ret; 3515 3516 f2fs_balance_fs(sbi, true); 3517 3518 inode_lock(inode); 3519 3520 writecount = atomic_read(&inode->i_writecount); 3521 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3522 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3523 ret = -EBUSY; 3524 goto out; 3525 } 3526 3527 if (!f2fs_compressed_file(inode) || 3528 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3529 ret = -EINVAL; 3530 goto out; 3531 } 3532 3533 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3534 if (ret) 3535 goto out; 3536 3537 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3538 ret = -EPERM; 3539 goto out; 3540 } 3541 3542 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3543 inode_set_ctime_current(inode); 3544 f2fs_mark_inode_dirty_sync(inode, true); 3545 3546 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3547 filemap_invalidate_lock(inode->i_mapping); 3548 3549 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3550 3551 while (page_idx < last_idx) { 3552 struct dnode_of_data dn; 3553 pgoff_t end_offset, count; 3554 3555 f2fs_lock_op(sbi); 3556 3557 set_new_dnode(&dn, inode, NULL, NULL, 0); 3558 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3559 if (ret) { 3560 f2fs_unlock_op(sbi); 3561 if (ret == -ENOENT) { 3562 page_idx = f2fs_get_next_page_offset(&dn, 3563 page_idx); 3564 ret = 0; 3565 continue; 3566 } 3567 break; 3568 } 3569 3570 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3571 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3572 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3573 3574 ret = release_compress_blocks(&dn, count); 3575 3576 f2fs_put_dnode(&dn); 3577 3578 f2fs_unlock_op(sbi); 3579 3580 if (ret < 0) 3581 break; 3582 3583 page_idx += count; 3584 released_blocks += ret; 3585 } 3586 3587 filemap_invalidate_unlock(inode->i_mapping); 3588 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3589 out: 3590 inode_unlock(inode); 3591 3592 mnt_drop_write_file(filp); 3593 3594 if (ret >= 0) { 3595 ret = put_user(released_blocks, (u64 __user *)arg); 3596 } else if (released_blocks && 3597 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3598 set_sbi_flag(sbi, SBI_NEED_FSCK); 3599 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3600 "iblocks=%llu, released=%u, compr_blocks=%u, " 3601 "run fsck to fix.", 3602 __func__, inode->i_ino, inode->i_blocks, 3603 released_blocks, 3604 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3605 } 3606 3607 return ret; 3608 } 3609 3610 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3611 unsigned int *reserved_blocks) 3612 { 3613 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3614 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3615 block_t blkaddr; 3616 int i; 3617 3618 for (i = 0; i < count; i++) { 3619 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3620 dn->ofs_in_node + i); 3621 3622 if (!__is_valid_data_blkaddr(blkaddr)) 3623 continue; 3624 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3625 DATA_GENERIC_ENHANCE))) { 3626 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3627 return -EFSCORRUPTED; 3628 } 3629 } 3630 3631 while (count) { 3632 int compr_blocks = 0; 3633 blkcnt_t reserved = 0; 3634 blkcnt_t to_reserved; 3635 int ret; 3636 3637 for (i = 0; i < cluster_size; i++) { 3638 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3639 dn->ofs_in_node + i); 3640 3641 if (i == 0) { 3642 if (blkaddr != COMPRESS_ADDR) { 3643 dn->ofs_in_node += cluster_size; 3644 goto next; 3645 } 3646 continue; 3647 } 3648 3649 /* 3650 * compressed cluster was not released due to it 3651 * fails in release_compress_blocks(), so NEW_ADDR 3652 * is a possible case. 3653 */ 3654 if (blkaddr == NEW_ADDR) { 3655 reserved++; 3656 continue; 3657 } 3658 if (__is_valid_data_blkaddr(blkaddr)) { 3659 compr_blocks++; 3660 continue; 3661 } 3662 } 3663 3664 to_reserved = cluster_size - compr_blocks - reserved; 3665 3666 /* for the case all blocks in cluster were reserved */ 3667 if (to_reserved == 1) { 3668 dn->ofs_in_node += cluster_size; 3669 goto next; 3670 } 3671 3672 ret = inc_valid_block_count(sbi, dn->inode, 3673 &to_reserved, false); 3674 if (unlikely(ret)) 3675 return ret; 3676 3677 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3678 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3679 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3680 } 3681 3682 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3683 3684 *reserved_blocks += to_reserved; 3685 next: 3686 count -= cluster_size; 3687 } 3688 3689 return 0; 3690 } 3691 3692 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3693 { 3694 struct inode *inode = file_inode(filp); 3695 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3696 pgoff_t page_idx = 0, last_idx; 3697 unsigned int reserved_blocks = 0; 3698 int ret; 3699 3700 if (!f2fs_sb_has_compression(sbi)) 3701 return -EOPNOTSUPP; 3702 3703 if (f2fs_readonly(sbi->sb)) 3704 return -EROFS; 3705 3706 ret = mnt_want_write_file(filp); 3707 if (ret) 3708 return ret; 3709 3710 f2fs_balance_fs(sbi, true); 3711 3712 inode_lock(inode); 3713 3714 if (!f2fs_compressed_file(inode) || 3715 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3716 ret = -EINVAL; 3717 goto unlock_inode; 3718 } 3719 3720 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3721 goto unlock_inode; 3722 3723 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3724 filemap_invalidate_lock(inode->i_mapping); 3725 3726 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3727 3728 while (page_idx < last_idx) { 3729 struct dnode_of_data dn; 3730 pgoff_t end_offset, count; 3731 3732 f2fs_lock_op(sbi); 3733 3734 set_new_dnode(&dn, inode, NULL, NULL, 0); 3735 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3736 if (ret) { 3737 f2fs_unlock_op(sbi); 3738 if (ret == -ENOENT) { 3739 page_idx = f2fs_get_next_page_offset(&dn, 3740 page_idx); 3741 ret = 0; 3742 continue; 3743 } 3744 break; 3745 } 3746 3747 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3748 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3749 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3750 3751 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 3752 3753 f2fs_put_dnode(&dn); 3754 3755 f2fs_unlock_op(sbi); 3756 3757 if (ret < 0) 3758 break; 3759 3760 page_idx += count; 3761 } 3762 3763 filemap_invalidate_unlock(inode->i_mapping); 3764 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3765 3766 if (!ret) { 3767 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3768 inode_set_ctime_current(inode); 3769 f2fs_mark_inode_dirty_sync(inode, true); 3770 } 3771 unlock_inode: 3772 inode_unlock(inode); 3773 mnt_drop_write_file(filp); 3774 3775 if (!ret) { 3776 ret = put_user(reserved_blocks, (u64 __user *)arg); 3777 } else if (reserved_blocks && 3778 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3779 set_sbi_flag(sbi, SBI_NEED_FSCK); 3780 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3781 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3782 "run fsck to fix.", 3783 __func__, inode->i_ino, inode->i_blocks, 3784 reserved_blocks, 3785 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3786 } 3787 3788 return ret; 3789 } 3790 3791 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3792 pgoff_t off, block_t block, block_t len, u32 flags) 3793 { 3794 sector_t sector = SECTOR_FROM_BLOCK(block); 3795 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3796 int ret = 0; 3797 3798 if (flags & F2FS_TRIM_FILE_DISCARD) { 3799 if (bdev_max_secure_erase_sectors(bdev)) 3800 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3801 GFP_NOFS); 3802 else 3803 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3804 GFP_NOFS); 3805 } 3806 3807 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3808 if (IS_ENCRYPTED(inode)) 3809 ret = fscrypt_zeroout_range(inode, off, block, len); 3810 else 3811 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3812 GFP_NOFS, 0); 3813 } 3814 3815 return ret; 3816 } 3817 3818 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3819 { 3820 struct inode *inode = file_inode(filp); 3821 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3822 struct address_space *mapping = inode->i_mapping; 3823 struct block_device *prev_bdev = NULL; 3824 struct f2fs_sectrim_range range; 3825 pgoff_t index, pg_end, prev_index = 0; 3826 block_t prev_block = 0, len = 0; 3827 loff_t end_addr; 3828 bool to_end = false; 3829 int ret = 0; 3830 3831 if (!(filp->f_mode & FMODE_WRITE)) 3832 return -EBADF; 3833 3834 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3835 sizeof(range))) 3836 return -EFAULT; 3837 3838 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3839 !S_ISREG(inode->i_mode)) 3840 return -EINVAL; 3841 3842 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3843 !f2fs_hw_support_discard(sbi)) || 3844 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3845 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3846 return -EOPNOTSUPP; 3847 3848 file_start_write(filp); 3849 inode_lock(inode); 3850 3851 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3852 range.start >= inode->i_size) { 3853 ret = -EINVAL; 3854 goto err; 3855 } 3856 3857 if (range.len == 0) 3858 goto err; 3859 3860 if (inode->i_size - range.start > range.len) { 3861 end_addr = range.start + range.len; 3862 } else { 3863 end_addr = range.len == (u64)-1 ? 3864 sbi->sb->s_maxbytes : inode->i_size; 3865 to_end = true; 3866 } 3867 3868 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3869 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3870 ret = -EINVAL; 3871 goto err; 3872 } 3873 3874 index = F2FS_BYTES_TO_BLK(range.start); 3875 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3876 3877 ret = f2fs_convert_inline_inode(inode); 3878 if (ret) 3879 goto err; 3880 3881 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3882 filemap_invalidate_lock(mapping); 3883 3884 ret = filemap_write_and_wait_range(mapping, range.start, 3885 to_end ? LLONG_MAX : end_addr - 1); 3886 if (ret) 3887 goto out; 3888 3889 truncate_inode_pages_range(mapping, range.start, 3890 to_end ? -1 : end_addr - 1); 3891 3892 while (index < pg_end) { 3893 struct dnode_of_data dn; 3894 pgoff_t end_offset, count; 3895 int i; 3896 3897 set_new_dnode(&dn, inode, NULL, NULL, 0); 3898 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3899 if (ret) { 3900 if (ret == -ENOENT) { 3901 index = f2fs_get_next_page_offset(&dn, index); 3902 continue; 3903 } 3904 goto out; 3905 } 3906 3907 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3908 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3909 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3910 struct block_device *cur_bdev; 3911 block_t blkaddr = f2fs_data_blkaddr(&dn); 3912 3913 if (!__is_valid_data_blkaddr(blkaddr)) 3914 continue; 3915 3916 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3917 DATA_GENERIC_ENHANCE)) { 3918 ret = -EFSCORRUPTED; 3919 f2fs_put_dnode(&dn); 3920 f2fs_handle_error(sbi, 3921 ERROR_INVALID_BLKADDR); 3922 goto out; 3923 } 3924 3925 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3926 if (f2fs_is_multi_device(sbi)) { 3927 int di = f2fs_target_device_index(sbi, blkaddr); 3928 3929 blkaddr -= FDEV(di).start_blk; 3930 } 3931 3932 if (len) { 3933 if (prev_bdev == cur_bdev && 3934 index == prev_index + len && 3935 blkaddr == prev_block + len) { 3936 len++; 3937 } else { 3938 ret = f2fs_secure_erase(prev_bdev, 3939 inode, prev_index, prev_block, 3940 len, range.flags); 3941 if (ret) { 3942 f2fs_put_dnode(&dn); 3943 goto out; 3944 } 3945 3946 len = 0; 3947 } 3948 } 3949 3950 if (!len) { 3951 prev_bdev = cur_bdev; 3952 prev_index = index; 3953 prev_block = blkaddr; 3954 len = 1; 3955 } 3956 } 3957 3958 f2fs_put_dnode(&dn); 3959 3960 if (fatal_signal_pending(current)) { 3961 ret = -EINTR; 3962 goto out; 3963 } 3964 cond_resched(); 3965 } 3966 3967 if (len) 3968 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3969 prev_block, len, range.flags); 3970 out: 3971 filemap_invalidate_unlock(mapping); 3972 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3973 err: 3974 inode_unlock(inode); 3975 file_end_write(filp); 3976 3977 return ret; 3978 } 3979 3980 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3981 { 3982 struct inode *inode = file_inode(filp); 3983 struct f2fs_comp_option option; 3984 3985 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3986 return -EOPNOTSUPP; 3987 3988 inode_lock_shared(inode); 3989 3990 if (!f2fs_compressed_file(inode)) { 3991 inode_unlock_shared(inode); 3992 return -ENODATA; 3993 } 3994 3995 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3996 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3997 3998 inode_unlock_shared(inode); 3999 4000 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4001 sizeof(option))) 4002 return -EFAULT; 4003 4004 return 0; 4005 } 4006 4007 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4008 { 4009 struct inode *inode = file_inode(filp); 4010 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4011 struct f2fs_comp_option option; 4012 int ret = 0; 4013 4014 if (!f2fs_sb_has_compression(sbi)) 4015 return -EOPNOTSUPP; 4016 4017 if (!(filp->f_mode & FMODE_WRITE)) 4018 return -EBADF; 4019 4020 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4021 sizeof(option))) 4022 return -EFAULT; 4023 4024 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4025 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4026 option.algorithm >= COMPRESS_MAX) 4027 return -EINVAL; 4028 4029 file_start_write(filp); 4030 inode_lock(inode); 4031 4032 f2fs_down_write(&F2FS_I(inode)->i_sem); 4033 if (!f2fs_compressed_file(inode)) { 4034 ret = -EINVAL; 4035 goto out; 4036 } 4037 4038 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4039 ret = -EBUSY; 4040 goto out; 4041 } 4042 4043 if (F2FS_HAS_BLOCKS(inode)) { 4044 ret = -EFBIG; 4045 goto out; 4046 } 4047 4048 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4049 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4050 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4051 /* Set default level */ 4052 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4053 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4054 else 4055 F2FS_I(inode)->i_compress_level = 0; 4056 /* Adjust mount option level */ 4057 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4058 F2FS_OPTION(sbi).compress_level) 4059 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4060 f2fs_mark_inode_dirty_sync(inode, true); 4061 4062 if (!f2fs_is_compress_backend_ready(inode)) 4063 f2fs_warn(sbi, "compression algorithm is successfully set, " 4064 "but current kernel doesn't support this algorithm."); 4065 out: 4066 f2fs_up_write(&F2FS_I(inode)->i_sem); 4067 inode_unlock(inode); 4068 file_end_write(filp); 4069 4070 return ret; 4071 } 4072 4073 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4074 { 4075 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4076 struct address_space *mapping = inode->i_mapping; 4077 struct page *page; 4078 pgoff_t redirty_idx = page_idx; 4079 int i, page_len = 0, ret = 0; 4080 4081 page_cache_ra_unbounded(&ractl, len, 0); 4082 4083 for (i = 0; i < len; i++, page_idx++) { 4084 page = read_cache_page(mapping, page_idx, NULL, NULL); 4085 if (IS_ERR(page)) { 4086 ret = PTR_ERR(page); 4087 break; 4088 } 4089 page_len++; 4090 } 4091 4092 for (i = 0; i < page_len; i++, redirty_idx++) { 4093 page = find_lock_page(mapping, redirty_idx); 4094 4095 /* It will never fail, when page has pinned above */ 4096 f2fs_bug_on(F2FS_I_SB(inode), !page); 4097 4098 set_page_dirty(page); 4099 set_page_private_gcing(page); 4100 f2fs_put_page(page, 1); 4101 f2fs_put_page(page, 0); 4102 } 4103 4104 return ret; 4105 } 4106 4107 static int f2fs_ioc_decompress_file(struct file *filp) 4108 { 4109 struct inode *inode = file_inode(filp); 4110 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4111 struct f2fs_inode_info *fi = F2FS_I(inode); 4112 pgoff_t page_idx = 0, last_idx; 4113 int cluster_size = fi->i_cluster_size; 4114 int count, ret; 4115 4116 if (!f2fs_sb_has_compression(sbi) || 4117 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4118 return -EOPNOTSUPP; 4119 4120 if (!(filp->f_mode & FMODE_WRITE)) 4121 return -EBADF; 4122 4123 f2fs_balance_fs(sbi, true); 4124 4125 file_start_write(filp); 4126 inode_lock(inode); 4127 4128 if (!f2fs_is_compress_backend_ready(inode)) { 4129 ret = -EOPNOTSUPP; 4130 goto out; 4131 } 4132 4133 if (!f2fs_compressed_file(inode) || 4134 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4135 ret = -EINVAL; 4136 goto out; 4137 } 4138 4139 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4140 if (ret) 4141 goto out; 4142 4143 if (!atomic_read(&fi->i_compr_blocks)) 4144 goto out; 4145 4146 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4147 4148 count = last_idx - page_idx; 4149 while (count && count >= cluster_size) { 4150 ret = redirty_blocks(inode, page_idx, cluster_size); 4151 if (ret < 0) 4152 break; 4153 4154 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4155 ret = filemap_fdatawrite(inode->i_mapping); 4156 if (ret < 0) 4157 break; 4158 } 4159 4160 count -= cluster_size; 4161 page_idx += cluster_size; 4162 4163 cond_resched(); 4164 if (fatal_signal_pending(current)) { 4165 ret = -EINTR; 4166 break; 4167 } 4168 } 4169 4170 if (!ret) 4171 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4172 LLONG_MAX); 4173 4174 if (ret) 4175 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4176 __func__, ret); 4177 out: 4178 inode_unlock(inode); 4179 file_end_write(filp); 4180 4181 return ret; 4182 } 4183 4184 static int f2fs_ioc_compress_file(struct file *filp) 4185 { 4186 struct inode *inode = file_inode(filp); 4187 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4188 pgoff_t page_idx = 0, last_idx; 4189 int cluster_size = F2FS_I(inode)->i_cluster_size; 4190 int count, ret; 4191 4192 if (!f2fs_sb_has_compression(sbi) || 4193 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4194 return -EOPNOTSUPP; 4195 4196 if (!(filp->f_mode & FMODE_WRITE)) 4197 return -EBADF; 4198 4199 f2fs_balance_fs(sbi, true); 4200 4201 file_start_write(filp); 4202 inode_lock(inode); 4203 4204 if (!f2fs_is_compress_backend_ready(inode)) { 4205 ret = -EOPNOTSUPP; 4206 goto out; 4207 } 4208 4209 if (!f2fs_compressed_file(inode) || 4210 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4211 ret = -EINVAL; 4212 goto out; 4213 } 4214 4215 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4216 if (ret) 4217 goto out; 4218 4219 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4220 4221 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4222 4223 count = last_idx - page_idx; 4224 while (count && count >= cluster_size) { 4225 ret = redirty_blocks(inode, page_idx, cluster_size); 4226 if (ret < 0) 4227 break; 4228 4229 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4230 ret = filemap_fdatawrite(inode->i_mapping); 4231 if (ret < 0) 4232 break; 4233 } 4234 4235 count -= cluster_size; 4236 page_idx += cluster_size; 4237 4238 cond_resched(); 4239 if (fatal_signal_pending(current)) { 4240 ret = -EINTR; 4241 break; 4242 } 4243 } 4244 4245 if (!ret) 4246 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4247 LLONG_MAX); 4248 4249 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4250 4251 if (ret) 4252 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4253 __func__, ret); 4254 out: 4255 inode_unlock(inode); 4256 file_end_write(filp); 4257 4258 return ret; 4259 } 4260 4261 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4262 { 4263 switch (cmd) { 4264 case FS_IOC_GETVERSION: 4265 return f2fs_ioc_getversion(filp, arg); 4266 case F2FS_IOC_START_ATOMIC_WRITE: 4267 return f2fs_ioc_start_atomic_write(filp, false); 4268 case F2FS_IOC_START_ATOMIC_REPLACE: 4269 return f2fs_ioc_start_atomic_write(filp, true); 4270 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4271 return f2fs_ioc_commit_atomic_write(filp); 4272 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4273 return f2fs_ioc_abort_atomic_write(filp); 4274 case F2FS_IOC_START_VOLATILE_WRITE: 4275 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4276 return -EOPNOTSUPP; 4277 case F2FS_IOC_SHUTDOWN: 4278 return f2fs_ioc_shutdown(filp, arg); 4279 case FITRIM: 4280 return f2fs_ioc_fitrim(filp, arg); 4281 case FS_IOC_SET_ENCRYPTION_POLICY: 4282 return f2fs_ioc_set_encryption_policy(filp, arg); 4283 case FS_IOC_GET_ENCRYPTION_POLICY: 4284 return f2fs_ioc_get_encryption_policy(filp, arg); 4285 case FS_IOC_GET_ENCRYPTION_PWSALT: 4286 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4287 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4288 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4289 case FS_IOC_ADD_ENCRYPTION_KEY: 4290 return f2fs_ioc_add_encryption_key(filp, arg); 4291 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4292 return f2fs_ioc_remove_encryption_key(filp, arg); 4293 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4294 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4295 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4296 return f2fs_ioc_get_encryption_key_status(filp, arg); 4297 case FS_IOC_GET_ENCRYPTION_NONCE: 4298 return f2fs_ioc_get_encryption_nonce(filp, arg); 4299 case F2FS_IOC_GARBAGE_COLLECT: 4300 return f2fs_ioc_gc(filp, arg); 4301 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4302 return f2fs_ioc_gc_range(filp, arg); 4303 case F2FS_IOC_WRITE_CHECKPOINT: 4304 return f2fs_ioc_write_checkpoint(filp); 4305 case F2FS_IOC_DEFRAGMENT: 4306 return f2fs_ioc_defragment(filp, arg); 4307 case F2FS_IOC_MOVE_RANGE: 4308 return f2fs_ioc_move_range(filp, arg); 4309 case F2FS_IOC_FLUSH_DEVICE: 4310 return f2fs_ioc_flush_device(filp, arg); 4311 case F2FS_IOC_GET_FEATURES: 4312 return f2fs_ioc_get_features(filp, arg); 4313 case F2FS_IOC_GET_PIN_FILE: 4314 return f2fs_ioc_get_pin_file(filp, arg); 4315 case F2FS_IOC_SET_PIN_FILE: 4316 return f2fs_ioc_set_pin_file(filp, arg); 4317 case F2FS_IOC_PRECACHE_EXTENTS: 4318 return f2fs_ioc_precache_extents(filp); 4319 case F2FS_IOC_RESIZE_FS: 4320 return f2fs_ioc_resize_fs(filp, arg); 4321 case FS_IOC_ENABLE_VERITY: 4322 return f2fs_ioc_enable_verity(filp, arg); 4323 case FS_IOC_MEASURE_VERITY: 4324 return f2fs_ioc_measure_verity(filp, arg); 4325 case FS_IOC_READ_VERITY_METADATA: 4326 return f2fs_ioc_read_verity_metadata(filp, arg); 4327 case FS_IOC_GETFSLABEL: 4328 return f2fs_ioc_getfslabel(filp, arg); 4329 case FS_IOC_SETFSLABEL: 4330 return f2fs_ioc_setfslabel(filp, arg); 4331 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4332 return f2fs_ioc_get_compress_blocks(filp, arg); 4333 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4334 return f2fs_release_compress_blocks(filp, arg); 4335 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4336 return f2fs_reserve_compress_blocks(filp, arg); 4337 case F2FS_IOC_SEC_TRIM_FILE: 4338 return f2fs_sec_trim_file(filp, arg); 4339 case F2FS_IOC_GET_COMPRESS_OPTION: 4340 return f2fs_ioc_get_compress_option(filp, arg); 4341 case F2FS_IOC_SET_COMPRESS_OPTION: 4342 return f2fs_ioc_set_compress_option(filp, arg); 4343 case F2FS_IOC_DECOMPRESS_FILE: 4344 return f2fs_ioc_decompress_file(filp); 4345 case F2FS_IOC_COMPRESS_FILE: 4346 return f2fs_ioc_compress_file(filp); 4347 default: 4348 return -ENOTTY; 4349 } 4350 } 4351 4352 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4353 { 4354 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4355 return -EIO; 4356 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4357 return -ENOSPC; 4358 4359 return __f2fs_ioctl(filp, cmd, arg); 4360 } 4361 4362 /* 4363 * Return %true if the given read or write request should use direct I/O, or 4364 * %false if it should use buffered I/O. 4365 */ 4366 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4367 struct iov_iter *iter) 4368 { 4369 unsigned int align; 4370 4371 if (!(iocb->ki_flags & IOCB_DIRECT)) 4372 return false; 4373 4374 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4375 return false; 4376 4377 /* 4378 * Direct I/O not aligned to the disk's logical_block_size will be 4379 * attempted, but will fail with -EINVAL. 4380 * 4381 * f2fs additionally requires that direct I/O be aligned to the 4382 * filesystem block size, which is often a stricter requirement. 4383 * However, f2fs traditionally falls back to buffered I/O on requests 4384 * that are logical_block_size-aligned but not fs-block aligned. 4385 * 4386 * The below logic implements this behavior. 4387 */ 4388 align = iocb->ki_pos | iov_iter_alignment(iter); 4389 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4390 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4391 return false; 4392 4393 return true; 4394 } 4395 4396 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4397 unsigned int flags) 4398 { 4399 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4400 4401 dec_page_count(sbi, F2FS_DIO_READ); 4402 if (error) 4403 return error; 4404 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4405 return 0; 4406 } 4407 4408 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4409 .end_io = f2fs_dio_read_end_io, 4410 }; 4411 4412 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4413 { 4414 struct file *file = iocb->ki_filp; 4415 struct inode *inode = file_inode(file); 4416 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4417 struct f2fs_inode_info *fi = F2FS_I(inode); 4418 const loff_t pos = iocb->ki_pos; 4419 const size_t count = iov_iter_count(to); 4420 struct iomap_dio *dio; 4421 ssize_t ret; 4422 4423 if (count == 0) 4424 return 0; /* skip atime update */ 4425 4426 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4427 4428 if (iocb->ki_flags & IOCB_NOWAIT) { 4429 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4430 ret = -EAGAIN; 4431 goto out; 4432 } 4433 } else { 4434 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4435 } 4436 4437 /* 4438 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4439 * the higher-level function iomap_dio_rw() in order to ensure that the 4440 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4441 */ 4442 inc_page_count(sbi, F2FS_DIO_READ); 4443 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4444 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4445 if (IS_ERR_OR_NULL(dio)) { 4446 ret = PTR_ERR_OR_ZERO(dio); 4447 if (ret != -EIOCBQUEUED) 4448 dec_page_count(sbi, F2FS_DIO_READ); 4449 } else { 4450 ret = iomap_dio_complete(dio); 4451 } 4452 4453 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4454 4455 file_accessed(file); 4456 out: 4457 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4458 return ret; 4459 } 4460 4461 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4462 int rw) 4463 { 4464 struct inode *inode = file_inode(file); 4465 char *buf, *path; 4466 4467 buf = f2fs_getname(F2FS_I_SB(inode)); 4468 if (!buf) 4469 return; 4470 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4471 if (IS_ERR(path)) 4472 goto free_buf; 4473 if (rw == WRITE) 4474 trace_f2fs_datawrite_start(inode, pos, count, 4475 current->pid, path, current->comm); 4476 else 4477 trace_f2fs_dataread_start(inode, pos, count, 4478 current->pid, path, current->comm); 4479 free_buf: 4480 f2fs_putname(buf); 4481 } 4482 4483 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4484 { 4485 struct inode *inode = file_inode(iocb->ki_filp); 4486 const loff_t pos = iocb->ki_pos; 4487 ssize_t ret; 4488 4489 if (!f2fs_is_compress_backend_ready(inode)) 4490 return -EOPNOTSUPP; 4491 4492 if (trace_f2fs_dataread_start_enabled()) 4493 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4494 iov_iter_count(to), READ); 4495 4496 if (f2fs_should_use_dio(inode, iocb, to)) { 4497 ret = f2fs_dio_read_iter(iocb, to); 4498 } else { 4499 ret = filemap_read(iocb, to, 0); 4500 if (ret > 0) 4501 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4502 APP_BUFFERED_READ_IO, ret); 4503 } 4504 if (trace_f2fs_dataread_end_enabled()) 4505 trace_f2fs_dataread_end(inode, pos, ret); 4506 return ret; 4507 } 4508 4509 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4510 struct pipe_inode_info *pipe, 4511 size_t len, unsigned int flags) 4512 { 4513 struct inode *inode = file_inode(in); 4514 const loff_t pos = *ppos; 4515 ssize_t ret; 4516 4517 if (!f2fs_is_compress_backend_ready(inode)) 4518 return -EOPNOTSUPP; 4519 4520 if (trace_f2fs_dataread_start_enabled()) 4521 f2fs_trace_rw_file_path(in, pos, len, READ); 4522 4523 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4524 if (ret > 0) 4525 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4526 APP_BUFFERED_READ_IO, ret); 4527 4528 if (trace_f2fs_dataread_end_enabled()) 4529 trace_f2fs_dataread_end(inode, pos, ret); 4530 return ret; 4531 } 4532 4533 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4534 { 4535 struct file *file = iocb->ki_filp; 4536 struct inode *inode = file_inode(file); 4537 ssize_t count; 4538 int err; 4539 4540 if (IS_IMMUTABLE(inode)) 4541 return -EPERM; 4542 4543 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4544 return -EPERM; 4545 4546 count = generic_write_checks(iocb, from); 4547 if (count <= 0) 4548 return count; 4549 4550 err = file_modified(file); 4551 if (err) 4552 return err; 4553 return count; 4554 } 4555 4556 /* 4557 * Preallocate blocks for a write request, if it is possible and helpful to do 4558 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4559 * blocks were preallocated, or a negative errno value if something went 4560 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4561 * requested blocks (not just some of them) have been allocated. 4562 */ 4563 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4564 bool dio) 4565 { 4566 struct inode *inode = file_inode(iocb->ki_filp); 4567 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4568 const loff_t pos = iocb->ki_pos; 4569 const size_t count = iov_iter_count(iter); 4570 struct f2fs_map_blocks map = {}; 4571 int flag; 4572 int ret; 4573 4574 /* If it will be an out-of-place direct write, don't bother. */ 4575 if (dio && f2fs_lfs_mode(sbi)) 4576 return 0; 4577 /* 4578 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4579 * buffered IO, if DIO meets any holes. 4580 */ 4581 if (dio && i_size_read(inode) && 4582 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4583 return 0; 4584 4585 /* No-wait I/O can't allocate blocks. */ 4586 if (iocb->ki_flags & IOCB_NOWAIT) 4587 return 0; 4588 4589 /* If it will be a short write, don't bother. */ 4590 if (fault_in_iov_iter_readable(iter, count)) 4591 return 0; 4592 4593 if (f2fs_has_inline_data(inode)) { 4594 /* If the data will fit inline, don't bother. */ 4595 if (pos + count <= MAX_INLINE_DATA(inode)) 4596 return 0; 4597 ret = f2fs_convert_inline_inode(inode); 4598 if (ret) 4599 return ret; 4600 } 4601 4602 /* Do not preallocate blocks that will be written partially in 4KB. */ 4603 map.m_lblk = F2FS_BLK_ALIGN(pos); 4604 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4605 if (map.m_len > map.m_lblk) 4606 map.m_len -= map.m_lblk; 4607 else 4608 map.m_len = 0; 4609 map.m_may_create = true; 4610 if (dio) { 4611 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4612 flag = F2FS_GET_BLOCK_PRE_DIO; 4613 } else { 4614 map.m_seg_type = NO_CHECK_TYPE; 4615 flag = F2FS_GET_BLOCK_PRE_AIO; 4616 } 4617 4618 ret = f2fs_map_blocks(inode, &map, flag); 4619 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4620 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4621 return ret; 4622 if (ret == 0) 4623 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4624 return map.m_len; 4625 } 4626 4627 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4628 struct iov_iter *from) 4629 { 4630 struct file *file = iocb->ki_filp; 4631 struct inode *inode = file_inode(file); 4632 ssize_t ret; 4633 4634 if (iocb->ki_flags & IOCB_NOWAIT) 4635 return -EOPNOTSUPP; 4636 4637 ret = generic_perform_write(iocb, from); 4638 4639 if (ret > 0) { 4640 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4641 APP_BUFFERED_IO, ret); 4642 } 4643 return ret; 4644 } 4645 4646 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4647 unsigned int flags) 4648 { 4649 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4650 4651 dec_page_count(sbi, F2FS_DIO_WRITE); 4652 if (error) 4653 return error; 4654 f2fs_update_time(sbi, REQ_TIME); 4655 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4656 return 0; 4657 } 4658 4659 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4660 .end_io = f2fs_dio_write_end_io, 4661 }; 4662 4663 static void f2fs_flush_buffered_write(struct address_space *mapping, 4664 loff_t start_pos, loff_t end_pos) 4665 { 4666 int ret; 4667 4668 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4669 if (ret < 0) 4670 return; 4671 invalidate_mapping_pages(mapping, 4672 start_pos >> PAGE_SHIFT, 4673 end_pos >> PAGE_SHIFT); 4674 } 4675 4676 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4677 bool *may_need_sync) 4678 { 4679 struct file *file = iocb->ki_filp; 4680 struct inode *inode = file_inode(file); 4681 struct f2fs_inode_info *fi = F2FS_I(inode); 4682 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4683 const bool do_opu = f2fs_lfs_mode(sbi); 4684 const loff_t pos = iocb->ki_pos; 4685 const ssize_t count = iov_iter_count(from); 4686 unsigned int dio_flags; 4687 struct iomap_dio *dio; 4688 ssize_t ret; 4689 4690 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4691 4692 if (iocb->ki_flags & IOCB_NOWAIT) { 4693 /* f2fs_convert_inline_inode() and block allocation can block */ 4694 if (f2fs_has_inline_data(inode) || 4695 !f2fs_overwrite_io(inode, pos, count)) { 4696 ret = -EAGAIN; 4697 goto out; 4698 } 4699 4700 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4701 ret = -EAGAIN; 4702 goto out; 4703 } 4704 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4705 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4706 ret = -EAGAIN; 4707 goto out; 4708 } 4709 } else { 4710 ret = f2fs_convert_inline_inode(inode); 4711 if (ret) 4712 goto out; 4713 4714 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4715 if (do_opu) 4716 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4717 } 4718 4719 /* 4720 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4721 * the higher-level function iomap_dio_rw() in order to ensure that the 4722 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4723 */ 4724 inc_page_count(sbi, F2FS_DIO_WRITE); 4725 dio_flags = 0; 4726 if (pos + count > inode->i_size) 4727 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4728 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4729 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4730 if (IS_ERR_OR_NULL(dio)) { 4731 ret = PTR_ERR_OR_ZERO(dio); 4732 if (ret == -ENOTBLK) 4733 ret = 0; 4734 if (ret != -EIOCBQUEUED) 4735 dec_page_count(sbi, F2FS_DIO_WRITE); 4736 } else { 4737 ret = iomap_dio_complete(dio); 4738 } 4739 4740 if (do_opu) 4741 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4742 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4743 4744 if (ret < 0) 4745 goto out; 4746 if (pos + ret > inode->i_size) 4747 f2fs_i_size_write(inode, pos + ret); 4748 if (!do_opu) 4749 set_inode_flag(inode, FI_UPDATE_WRITE); 4750 4751 if (iov_iter_count(from)) { 4752 ssize_t ret2; 4753 loff_t bufio_start_pos = iocb->ki_pos; 4754 4755 /* 4756 * The direct write was partial, so we need to fall back to a 4757 * buffered write for the remainder. 4758 */ 4759 4760 ret2 = f2fs_buffered_write_iter(iocb, from); 4761 if (iov_iter_count(from)) 4762 f2fs_write_failed(inode, iocb->ki_pos); 4763 if (ret2 < 0) 4764 goto out; 4765 4766 /* 4767 * Ensure that the pagecache pages are written to disk and 4768 * invalidated to preserve the expected O_DIRECT semantics. 4769 */ 4770 if (ret2 > 0) { 4771 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4772 4773 ret += ret2; 4774 4775 f2fs_flush_buffered_write(file->f_mapping, 4776 bufio_start_pos, 4777 bufio_end_pos); 4778 } 4779 } else { 4780 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4781 *may_need_sync = false; 4782 } 4783 out: 4784 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4785 return ret; 4786 } 4787 4788 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4789 { 4790 struct inode *inode = file_inode(iocb->ki_filp); 4791 const loff_t orig_pos = iocb->ki_pos; 4792 const size_t orig_count = iov_iter_count(from); 4793 loff_t target_size; 4794 bool dio; 4795 bool may_need_sync = true; 4796 int preallocated; 4797 ssize_t ret; 4798 4799 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4800 ret = -EIO; 4801 goto out; 4802 } 4803 4804 if (!f2fs_is_compress_backend_ready(inode)) { 4805 ret = -EOPNOTSUPP; 4806 goto out; 4807 } 4808 4809 if (iocb->ki_flags & IOCB_NOWAIT) { 4810 if (!inode_trylock(inode)) { 4811 ret = -EAGAIN; 4812 goto out; 4813 } 4814 } else { 4815 inode_lock(inode); 4816 } 4817 4818 ret = f2fs_write_checks(iocb, from); 4819 if (ret <= 0) 4820 goto out_unlock; 4821 4822 /* Determine whether we will do a direct write or a buffered write. */ 4823 dio = f2fs_should_use_dio(inode, iocb, from); 4824 4825 /* Possibly preallocate the blocks for the write. */ 4826 target_size = iocb->ki_pos + iov_iter_count(from); 4827 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4828 if (preallocated < 0) { 4829 ret = preallocated; 4830 } else { 4831 if (trace_f2fs_datawrite_start_enabled()) 4832 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4833 orig_count, WRITE); 4834 4835 /* Do the actual write. */ 4836 ret = dio ? 4837 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4838 f2fs_buffered_write_iter(iocb, from); 4839 4840 if (trace_f2fs_datawrite_end_enabled()) 4841 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4842 } 4843 4844 /* Don't leave any preallocated blocks around past i_size. */ 4845 if (preallocated && i_size_read(inode) < target_size) { 4846 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4847 filemap_invalidate_lock(inode->i_mapping); 4848 if (!f2fs_truncate(inode)) 4849 file_dont_truncate(inode); 4850 filemap_invalidate_unlock(inode->i_mapping); 4851 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4852 } else { 4853 file_dont_truncate(inode); 4854 } 4855 4856 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4857 out_unlock: 4858 inode_unlock(inode); 4859 out: 4860 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4861 4862 if (ret > 0 && may_need_sync) 4863 ret = generic_write_sync(iocb, ret); 4864 4865 /* If buffered IO was forced, flush and drop the data from 4866 * the page cache to preserve O_DIRECT semantics 4867 */ 4868 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4869 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4870 orig_pos, 4871 orig_pos + ret - 1); 4872 4873 return ret; 4874 } 4875 4876 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4877 int advice) 4878 { 4879 struct address_space *mapping; 4880 struct backing_dev_info *bdi; 4881 struct inode *inode = file_inode(filp); 4882 int err; 4883 4884 if (advice == POSIX_FADV_SEQUENTIAL) { 4885 if (S_ISFIFO(inode->i_mode)) 4886 return -ESPIPE; 4887 4888 mapping = filp->f_mapping; 4889 if (!mapping || len < 0) 4890 return -EINVAL; 4891 4892 bdi = inode_to_bdi(mapping->host); 4893 filp->f_ra.ra_pages = bdi->ra_pages * 4894 F2FS_I_SB(inode)->seq_file_ra_mul; 4895 spin_lock(&filp->f_lock); 4896 filp->f_mode &= ~FMODE_RANDOM; 4897 spin_unlock(&filp->f_lock); 4898 return 0; 4899 } 4900 4901 err = generic_fadvise(filp, offset, len, advice); 4902 if (!err && advice == POSIX_FADV_DONTNEED && 4903 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4904 f2fs_compressed_file(inode)) 4905 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4906 4907 return err; 4908 } 4909 4910 #ifdef CONFIG_COMPAT 4911 struct compat_f2fs_gc_range { 4912 u32 sync; 4913 compat_u64 start; 4914 compat_u64 len; 4915 }; 4916 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4917 struct compat_f2fs_gc_range) 4918 4919 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4920 { 4921 struct compat_f2fs_gc_range __user *urange; 4922 struct f2fs_gc_range range; 4923 int err; 4924 4925 urange = compat_ptr(arg); 4926 err = get_user(range.sync, &urange->sync); 4927 err |= get_user(range.start, &urange->start); 4928 err |= get_user(range.len, &urange->len); 4929 if (err) 4930 return -EFAULT; 4931 4932 return __f2fs_ioc_gc_range(file, &range); 4933 } 4934 4935 struct compat_f2fs_move_range { 4936 u32 dst_fd; 4937 compat_u64 pos_in; 4938 compat_u64 pos_out; 4939 compat_u64 len; 4940 }; 4941 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4942 struct compat_f2fs_move_range) 4943 4944 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4945 { 4946 struct compat_f2fs_move_range __user *urange; 4947 struct f2fs_move_range range; 4948 int err; 4949 4950 urange = compat_ptr(arg); 4951 err = get_user(range.dst_fd, &urange->dst_fd); 4952 err |= get_user(range.pos_in, &urange->pos_in); 4953 err |= get_user(range.pos_out, &urange->pos_out); 4954 err |= get_user(range.len, &urange->len); 4955 if (err) 4956 return -EFAULT; 4957 4958 return __f2fs_ioc_move_range(file, &range); 4959 } 4960 4961 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4962 { 4963 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4964 return -EIO; 4965 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4966 return -ENOSPC; 4967 4968 switch (cmd) { 4969 case FS_IOC32_GETVERSION: 4970 cmd = FS_IOC_GETVERSION; 4971 break; 4972 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4973 return f2fs_compat_ioc_gc_range(file, arg); 4974 case F2FS_IOC32_MOVE_RANGE: 4975 return f2fs_compat_ioc_move_range(file, arg); 4976 case F2FS_IOC_START_ATOMIC_WRITE: 4977 case F2FS_IOC_START_ATOMIC_REPLACE: 4978 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4979 case F2FS_IOC_START_VOLATILE_WRITE: 4980 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4981 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4982 case F2FS_IOC_SHUTDOWN: 4983 case FITRIM: 4984 case FS_IOC_SET_ENCRYPTION_POLICY: 4985 case FS_IOC_GET_ENCRYPTION_PWSALT: 4986 case FS_IOC_GET_ENCRYPTION_POLICY: 4987 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4988 case FS_IOC_ADD_ENCRYPTION_KEY: 4989 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4990 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4991 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4992 case FS_IOC_GET_ENCRYPTION_NONCE: 4993 case F2FS_IOC_GARBAGE_COLLECT: 4994 case F2FS_IOC_WRITE_CHECKPOINT: 4995 case F2FS_IOC_DEFRAGMENT: 4996 case F2FS_IOC_FLUSH_DEVICE: 4997 case F2FS_IOC_GET_FEATURES: 4998 case F2FS_IOC_GET_PIN_FILE: 4999 case F2FS_IOC_SET_PIN_FILE: 5000 case F2FS_IOC_PRECACHE_EXTENTS: 5001 case F2FS_IOC_RESIZE_FS: 5002 case FS_IOC_ENABLE_VERITY: 5003 case FS_IOC_MEASURE_VERITY: 5004 case FS_IOC_READ_VERITY_METADATA: 5005 case FS_IOC_GETFSLABEL: 5006 case FS_IOC_SETFSLABEL: 5007 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5008 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5009 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5010 case F2FS_IOC_SEC_TRIM_FILE: 5011 case F2FS_IOC_GET_COMPRESS_OPTION: 5012 case F2FS_IOC_SET_COMPRESS_OPTION: 5013 case F2FS_IOC_DECOMPRESS_FILE: 5014 case F2FS_IOC_COMPRESS_FILE: 5015 break; 5016 default: 5017 return -ENOIOCTLCMD; 5018 } 5019 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5020 } 5021 #endif 5022 5023 const struct file_operations f2fs_file_operations = { 5024 .llseek = f2fs_llseek, 5025 .read_iter = f2fs_file_read_iter, 5026 .write_iter = f2fs_file_write_iter, 5027 .iopoll = iocb_bio_iopoll, 5028 .open = f2fs_file_open, 5029 .release = f2fs_release_file, 5030 .mmap = f2fs_file_mmap, 5031 .flush = f2fs_file_flush, 5032 .fsync = f2fs_sync_file, 5033 .fallocate = f2fs_fallocate, 5034 .unlocked_ioctl = f2fs_ioctl, 5035 #ifdef CONFIG_COMPAT 5036 .compat_ioctl = f2fs_compat_ioctl, 5037 #endif 5038 .splice_read = f2fs_file_splice_read, 5039 .splice_write = iter_file_splice_write, 5040 .fadvise = f2fs_file_fadvise, 5041 }; 5042