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 (f2fs_lfs_mode(sbi) && rw == WRITE && F2FS_IO_ALIGNED(sbi)) 817 return true; 818 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 819 return true; 820 821 return false; 822 } 823 824 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 825 struct kstat *stat, u32 request_mask, unsigned int query_flags) 826 { 827 struct inode *inode = d_inode(path->dentry); 828 struct f2fs_inode_info *fi = F2FS_I(inode); 829 struct f2fs_inode *ri = NULL; 830 unsigned int flags; 831 832 if (f2fs_has_extra_attr(inode) && 833 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 834 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 835 stat->result_mask |= STATX_BTIME; 836 stat->btime.tv_sec = fi->i_crtime.tv_sec; 837 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 838 } 839 840 /* 841 * Return the DIO alignment restrictions if requested. We only return 842 * this information when requested, since on encrypted files it might 843 * take a fair bit of work to get if the file wasn't opened recently. 844 * 845 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 846 * cannot represent that, so in that case we report no DIO support. 847 */ 848 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 849 unsigned int bsize = i_blocksize(inode); 850 851 stat->result_mask |= STATX_DIOALIGN; 852 if (!f2fs_force_buffered_io(inode, WRITE)) { 853 stat->dio_mem_align = bsize; 854 stat->dio_offset_align = bsize; 855 } 856 } 857 858 flags = fi->i_flags; 859 if (flags & F2FS_COMPR_FL) 860 stat->attributes |= STATX_ATTR_COMPRESSED; 861 if (flags & F2FS_APPEND_FL) 862 stat->attributes |= STATX_ATTR_APPEND; 863 if (IS_ENCRYPTED(inode)) 864 stat->attributes |= STATX_ATTR_ENCRYPTED; 865 if (flags & F2FS_IMMUTABLE_FL) 866 stat->attributes |= STATX_ATTR_IMMUTABLE; 867 if (flags & F2FS_NODUMP_FL) 868 stat->attributes |= STATX_ATTR_NODUMP; 869 if (IS_VERITY(inode)) 870 stat->attributes |= STATX_ATTR_VERITY; 871 872 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 873 STATX_ATTR_APPEND | 874 STATX_ATTR_ENCRYPTED | 875 STATX_ATTR_IMMUTABLE | 876 STATX_ATTR_NODUMP | 877 STATX_ATTR_VERITY); 878 879 generic_fillattr(idmap, request_mask, inode, stat); 880 881 /* we need to show initial sectors used for inline_data/dentries */ 882 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 883 f2fs_has_inline_dentry(inode)) 884 stat->blocks += (stat->size + 511) >> 9; 885 886 return 0; 887 } 888 889 #ifdef CONFIG_F2FS_FS_POSIX_ACL 890 static void __setattr_copy(struct mnt_idmap *idmap, 891 struct inode *inode, const struct iattr *attr) 892 { 893 unsigned int ia_valid = attr->ia_valid; 894 895 i_uid_update(idmap, attr, inode); 896 i_gid_update(idmap, attr, inode); 897 if (ia_valid & ATTR_ATIME) 898 inode->i_atime = attr->ia_atime; 899 if (ia_valid & ATTR_MTIME) 900 inode->i_mtime = attr->ia_mtime; 901 if (ia_valid & ATTR_CTIME) 902 inode_set_ctime_to_ts(inode, attr->ia_ctime); 903 if (ia_valid & ATTR_MODE) { 904 umode_t mode = attr->ia_mode; 905 vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode); 906 907 if (!vfsgid_in_group_p(vfsgid) && 908 !capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) 909 mode &= ~S_ISGID; 910 set_acl_inode(inode, mode); 911 } 912 } 913 #else 914 #define __setattr_copy setattr_copy 915 #endif 916 917 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 918 struct iattr *attr) 919 { 920 struct inode *inode = d_inode(dentry); 921 int err; 922 923 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 924 return -EIO; 925 926 if (unlikely(IS_IMMUTABLE(inode))) 927 return -EPERM; 928 929 if (unlikely(IS_APPEND(inode) && 930 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 931 ATTR_GID | ATTR_TIMES_SET)))) 932 return -EPERM; 933 934 if ((attr->ia_valid & ATTR_SIZE) && 935 !f2fs_is_compress_backend_ready(inode)) 936 return -EOPNOTSUPP; 937 938 err = setattr_prepare(idmap, dentry, attr); 939 if (err) 940 return err; 941 942 err = fscrypt_prepare_setattr(dentry, attr); 943 if (err) 944 return err; 945 946 err = fsverity_prepare_setattr(dentry, attr); 947 if (err) 948 return err; 949 950 if (is_quota_modification(idmap, inode, attr)) { 951 err = f2fs_dquot_initialize(inode); 952 if (err) 953 return err; 954 } 955 if (i_uid_needs_update(idmap, attr, inode) || 956 i_gid_needs_update(idmap, attr, inode)) { 957 f2fs_lock_op(F2FS_I_SB(inode)); 958 err = dquot_transfer(idmap, inode, attr); 959 if (err) { 960 set_sbi_flag(F2FS_I_SB(inode), 961 SBI_QUOTA_NEED_REPAIR); 962 f2fs_unlock_op(F2FS_I_SB(inode)); 963 return err; 964 } 965 /* 966 * update uid/gid under lock_op(), so that dquot and inode can 967 * be updated atomically. 968 */ 969 i_uid_update(idmap, attr, inode); 970 i_gid_update(idmap, attr, inode); 971 f2fs_mark_inode_dirty_sync(inode, true); 972 f2fs_unlock_op(F2FS_I_SB(inode)); 973 } 974 975 if (attr->ia_valid & ATTR_SIZE) { 976 loff_t old_size = i_size_read(inode); 977 978 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 979 /* 980 * should convert inline inode before i_size_write to 981 * keep smaller than inline_data size with inline flag. 982 */ 983 err = f2fs_convert_inline_inode(inode); 984 if (err) 985 return err; 986 } 987 988 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 989 filemap_invalidate_lock(inode->i_mapping); 990 991 truncate_setsize(inode, attr->ia_size); 992 993 if (attr->ia_size <= old_size) 994 err = f2fs_truncate(inode); 995 /* 996 * do not trim all blocks after i_size if target size is 997 * larger than i_size. 998 */ 999 filemap_invalidate_unlock(inode->i_mapping); 1000 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1001 if (err) 1002 return err; 1003 1004 spin_lock(&F2FS_I(inode)->i_size_lock); 1005 inode->i_mtime = inode_set_ctime_current(inode); 1006 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1007 spin_unlock(&F2FS_I(inode)->i_size_lock); 1008 } 1009 1010 __setattr_copy(idmap, inode, attr); 1011 1012 if (attr->ia_valid & ATTR_MODE) { 1013 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1014 1015 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1016 if (!err) 1017 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1018 clear_inode_flag(inode, FI_ACL_MODE); 1019 } 1020 } 1021 1022 /* file size may changed here */ 1023 f2fs_mark_inode_dirty_sync(inode, true); 1024 1025 /* inode change will produce dirty node pages flushed by checkpoint */ 1026 f2fs_balance_fs(F2FS_I_SB(inode), true); 1027 1028 return err; 1029 } 1030 1031 const struct inode_operations f2fs_file_inode_operations = { 1032 .getattr = f2fs_getattr, 1033 .setattr = f2fs_setattr, 1034 .get_inode_acl = f2fs_get_acl, 1035 .set_acl = f2fs_set_acl, 1036 .listxattr = f2fs_listxattr, 1037 .fiemap = f2fs_fiemap, 1038 .fileattr_get = f2fs_fileattr_get, 1039 .fileattr_set = f2fs_fileattr_set, 1040 }; 1041 1042 static int fill_zero(struct inode *inode, pgoff_t index, 1043 loff_t start, loff_t len) 1044 { 1045 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1046 struct page *page; 1047 1048 if (!len) 1049 return 0; 1050 1051 f2fs_balance_fs(sbi, true); 1052 1053 f2fs_lock_op(sbi); 1054 page = f2fs_get_new_data_page(inode, NULL, index, false); 1055 f2fs_unlock_op(sbi); 1056 1057 if (IS_ERR(page)) 1058 return PTR_ERR(page); 1059 1060 f2fs_wait_on_page_writeback(page, DATA, true, true); 1061 zero_user(page, start, len); 1062 set_page_dirty(page); 1063 f2fs_put_page(page, 1); 1064 return 0; 1065 } 1066 1067 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1068 { 1069 int err; 1070 1071 while (pg_start < pg_end) { 1072 struct dnode_of_data dn; 1073 pgoff_t end_offset, count; 1074 1075 set_new_dnode(&dn, inode, NULL, NULL, 0); 1076 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1077 if (err) { 1078 if (err == -ENOENT) { 1079 pg_start = f2fs_get_next_page_offset(&dn, 1080 pg_start); 1081 continue; 1082 } 1083 return err; 1084 } 1085 1086 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1087 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1088 1089 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1090 1091 f2fs_truncate_data_blocks_range(&dn, count); 1092 f2fs_put_dnode(&dn); 1093 1094 pg_start += count; 1095 } 1096 return 0; 1097 } 1098 1099 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1100 { 1101 pgoff_t pg_start, pg_end; 1102 loff_t off_start, off_end; 1103 int ret; 1104 1105 ret = f2fs_convert_inline_inode(inode); 1106 if (ret) 1107 return ret; 1108 1109 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1110 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1111 1112 off_start = offset & (PAGE_SIZE - 1); 1113 off_end = (offset + len) & (PAGE_SIZE - 1); 1114 1115 if (pg_start == pg_end) { 1116 ret = fill_zero(inode, pg_start, off_start, 1117 off_end - off_start); 1118 if (ret) 1119 return ret; 1120 } else { 1121 if (off_start) { 1122 ret = fill_zero(inode, pg_start++, off_start, 1123 PAGE_SIZE - off_start); 1124 if (ret) 1125 return ret; 1126 } 1127 if (off_end) { 1128 ret = fill_zero(inode, pg_end, 0, off_end); 1129 if (ret) 1130 return ret; 1131 } 1132 1133 if (pg_start < pg_end) { 1134 loff_t blk_start, blk_end; 1135 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1136 1137 f2fs_balance_fs(sbi, true); 1138 1139 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1140 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1141 1142 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1143 filemap_invalidate_lock(inode->i_mapping); 1144 1145 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1146 1147 f2fs_lock_op(sbi); 1148 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1149 f2fs_unlock_op(sbi); 1150 1151 filemap_invalidate_unlock(inode->i_mapping); 1152 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1153 } 1154 } 1155 1156 return ret; 1157 } 1158 1159 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1160 int *do_replace, pgoff_t off, pgoff_t len) 1161 { 1162 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1163 struct dnode_of_data dn; 1164 int ret, done, i; 1165 1166 next_dnode: 1167 set_new_dnode(&dn, inode, NULL, NULL, 0); 1168 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1169 if (ret && ret != -ENOENT) { 1170 return ret; 1171 } else if (ret == -ENOENT) { 1172 if (dn.max_level == 0) 1173 return -ENOENT; 1174 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1175 dn.ofs_in_node, len); 1176 blkaddr += done; 1177 do_replace += done; 1178 goto next; 1179 } 1180 1181 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1182 dn.ofs_in_node, len); 1183 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1184 *blkaddr = f2fs_data_blkaddr(&dn); 1185 1186 if (__is_valid_data_blkaddr(*blkaddr) && 1187 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1188 DATA_GENERIC_ENHANCE)) { 1189 f2fs_put_dnode(&dn); 1190 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1191 return -EFSCORRUPTED; 1192 } 1193 1194 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1195 1196 if (f2fs_lfs_mode(sbi)) { 1197 f2fs_put_dnode(&dn); 1198 return -EOPNOTSUPP; 1199 } 1200 1201 /* do not invalidate this block address */ 1202 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1203 *do_replace = 1; 1204 } 1205 } 1206 f2fs_put_dnode(&dn); 1207 next: 1208 len -= done; 1209 off += done; 1210 if (len) 1211 goto next_dnode; 1212 return 0; 1213 } 1214 1215 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1216 int *do_replace, pgoff_t off, int len) 1217 { 1218 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1219 struct dnode_of_data dn; 1220 int ret, i; 1221 1222 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1223 if (*do_replace == 0) 1224 continue; 1225 1226 set_new_dnode(&dn, inode, NULL, NULL, 0); 1227 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1228 if (ret) { 1229 dec_valid_block_count(sbi, inode, 1); 1230 f2fs_invalidate_blocks(sbi, *blkaddr); 1231 } else { 1232 f2fs_update_data_blkaddr(&dn, *blkaddr); 1233 } 1234 f2fs_put_dnode(&dn); 1235 } 1236 return 0; 1237 } 1238 1239 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1240 block_t *blkaddr, int *do_replace, 1241 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1242 { 1243 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1244 pgoff_t i = 0; 1245 int ret; 1246 1247 while (i < len) { 1248 if (blkaddr[i] == NULL_ADDR && !full) { 1249 i++; 1250 continue; 1251 } 1252 1253 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1254 struct dnode_of_data dn; 1255 struct node_info ni; 1256 size_t new_size; 1257 pgoff_t ilen; 1258 1259 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1260 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1261 if (ret) 1262 return ret; 1263 1264 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1265 if (ret) { 1266 f2fs_put_dnode(&dn); 1267 return ret; 1268 } 1269 1270 ilen = min((pgoff_t) 1271 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1272 dn.ofs_in_node, len - i); 1273 do { 1274 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1275 f2fs_truncate_data_blocks_range(&dn, 1); 1276 1277 if (do_replace[i]) { 1278 f2fs_i_blocks_write(src_inode, 1279 1, false, false); 1280 f2fs_i_blocks_write(dst_inode, 1281 1, true, false); 1282 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1283 blkaddr[i], ni.version, true, false); 1284 1285 do_replace[i] = 0; 1286 } 1287 dn.ofs_in_node++; 1288 i++; 1289 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1290 if (dst_inode->i_size < new_size) 1291 f2fs_i_size_write(dst_inode, new_size); 1292 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1293 1294 f2fs_put_dnode(&dn); 1295 } else { 1296 struct page *psrc, *pdst; 1297 1298 psrc = f2fs_get_lock_data_page(src_inode, 1299 src + i, true); 1300 if (IS_ERR(psrc)) 1301 return PTR_ERR(psrc); 1302 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1303 true); 1304 if (IS_ERR(pdst)) { 1305 f2fs_put_page(psrc, 1); 1306 return PTR_ERR(pdst); 1307 } 1308 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1309 set_page_dirty(pdst); 1310 set_page_private_gcing(pdst); 1311 f2fs_put_page(pdst, 1); 1312 f2fs_put_page(psrc, 1); 1313 1314 ret = f2fs_truncate_hole(src_inode, 1315 src + i, src + i + 1); 1316 if (ret) 1317 return ret; 1318 i++; 1319 } 1320 } 1321 return 0; 1322 } 1323 1324 static int __exchange_data_block(struct inode *src_inode, 1325 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1326 pgoff_t len, bool full) 1327 { 1328 block_t *src_blkaddr; 1329 int *do_replace; 1330 pgoff_t olen; 1331 int ret; 1332 1333 while (len) { 1334 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1335 1336 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1337 array_size(olen, sizeof(block_t)), 1338 GFP_NOFS); 1339 if (!src_blkaddr) 1340 return -ENOMEM; 1341 1342 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1343 array_size(olen, sizeof(int)), 1344 GFP_NOFS); 1345 if (!do_replace) { 1346 kvfree(src_blkaddr); 1347 return -ENOMEM; 1348 } 1349 1350 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1351 do_replace, src, olen); 1352 if (ret) 1353 goto roll_back; 1354 1355 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1356 do_replace, src, dst, olen, full); 1357 if (ret) 1358 goto roll_back; 1359 1360 src += olen; 1361 dst += olen; 1362 len -= olen; 1363 1364 kvfree(src_blkaddr); 1365 kvfree(do_replace); 1366 } 1367 return 0; 1368 1369 roll_back: 1370 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1371 kvfree(src_blkaddr); 1372 kvfree(do_replace); 1373 return ret; 1374 } 1375 1376 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1377 { 1378 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1379 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1380 pgoff_t start = offset >> PAGE_SHIFT; 1381 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1382 int ret; 1383 1384 f2fs_balance_fs(sbi, true); 1385 1386 /* avoid gc operation during block exchange */ 1387 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1388 filemap_invalidate_lock(inode->i_mapping); 1389 1390 f2fs_lock_op(sbi); 1391 f2fs_drop_extent_tree(inode); 1392 truncate_pagecache(inode, offset); 1393 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1394 f2fs_unlock_op(sbi); 1395 1396 filemap_invalidate_unlock(inode->i_mapping); 1397 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1398 return ret; 1399 } 1400 1401 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1402 { 1403 loff_t new_size; 1404 int ret; 1405 1406 if (offset + len >= i_size_read(inode)) 1407 return -EINVAL; 1408 1409 /* collapse range should be aligned to block size of f2fs. */ 1410 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1411 return -EINVAL; 1412 1413 ret = f2fs_convert_inline_inode(inode); 1414 if (ret) 1415 return ret; 1416 1417 /* write out all dirty pages from offset */ 1418 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1419 if (ret) 1420 return ret; 1421 1422 ret = f2fs_do_collapse(inode, offset, len); 1423 if (ret) 1424 return ret; 1425 1426 /* write out all moved pages, if possible */ 1427 filemap_invalidate_lock(inode->i_mapping); 1428 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1429 truncate_pagecache(inode, offset); 1430 1431 new_size = i_size_read(inode) - len; 1432 ret = f2fs_truncate_blocks(inode, new_size, true); 1433 filemap_invalidate_unlock(inode->i_mapping); 1434 if (!ret) 1435 f2fs_i_size_write(inode, new_size); 1436 return ret; 1437 } 1438 1439 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1440 pgoff_t end) 1441 { 1442 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1443 pgoff_t index = start; 1444 unsigned int ofs_in_node = dn->ofs_in_node; 1445 blkcnt_t count = 0; 1446 int ret; 1447 1448 for (; index < end; index++, dn->ofs_in_node++) { 1449 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1450 count++; 1451 } 1452 1453 dn->ofs_in_node = ofs_in_node; 1454 ret = f2fs_reserve_new_blocks(dn, count); 1455 if (ret) 1456 return ret; 1457 1458 dn->ofs_in_node = ofs_in_node; 1459 for (index = start; index < end; index++, dn->ofs_in_node++) { 1460 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1461 /* 1462 * f2fs_reserve_new_blocks will not guarantee entire block 1463 * allocation. 1464 */ 1465 if (dn->data_blkaddr == NULL_ADDR) { 1466 ret = -ENOSPC; 1467 break; 1468 } 1469 1470 if (dn->data_blkaddr == NEW_ADDR) 1471 continue; 1472 1473 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1474 DATA_GENERIC_ENHANCE)) { 1475 ret = -EFSCORRUPTED; 1476 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 1477 break; 1478 } 1479 1480 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1481 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1482 } 1483 1484 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1485 f2fs_update_age_extent_cache_range(dn, start, index - start); 1486 1487 return ret; 1488 } 1489 1490 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1491 int mode) 1492 { 1493 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1494 struct address_space *mapping = inode->i_mapping; 1495 pgoff_t index, pg_start, pg_end; 1496 loff_t new_size = i_size_read(inode); 1497 loff_t off_start, off_end; 1498 int ret = 0; 1499 1500 ret = inode_newsize_ok(inode, (len + offset)); 1501 if (ret) 1502 return ret; 1503 1504 ret = f2fs_convert_inline_inode(inode); 1505 if (ret) 1506 return ret; 1507 1508 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1509 if (ret) 1510 return ret; 1511 1512 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1513 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1514 1515 off_start = offset & (PAGE_SIZE - 1); 1516 off_end = (offset + len) & (PAGE_SIZE - 1); 1517 1518 if (pg_start == pg_end) { 1519 ret = fill_zero(inode, pg_start, off_start, 1520 off_end - off_start); 1521 if (ret) 1522 return ret; 1523 1524 new_size = max_t(loff_t, new_size, offset + len); 1525 } else { 1526 if (off_start) { 1527 ret = fill_zero(inode, pg_start++, off_start, 1528 PAGE_SIZE - off_start); 1529 if (ret) 1530 return ret; 1531 1532 new_size = max_t(loff_t, new_size, 1533 (loff_t)pg_start << PAGE_SHIFT); 1534 } 1535 1536 for (index = pg_start; index < pg_end;) { 1537 struct dnode_of_data dn; 1538 unsigned int end_offset; 1539 pgoff_t end; 1540 1541 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1542 filemap_invalidate_lock(mapping); 1543 1544 truncate_pagecache_range(inode, 1545 (loff_t)index << PAGE_SHIFT, 1546 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1547 1548 f2fs_lock_op(sbi); 1549 1550 set_new_dnode(&dn, inode, NULL, NULL, 0); 1551 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1552 if (ret) { 1553 f2fs_unlock_op(sbi); 1554 filemap_invalidate_unlock(mapping); 1555 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1556 goto out; 1557 } 1558 1559 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1560 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1561 1562 ret = f2fs_do_zero_range(&dn, index, end); 1563 f2fs_put_dnode(&dn); 1564 1565 f2fs_unlock_op(sbi); 1566 filemap_invalidate_unlock(mapping); 1567 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1568 1569 f2fs_balance_fs(sbi, dn.node_changed); 1570 1571 if (ret) 1572 goto out; 1573 1574 index = end; 1575 new_size = max_t(loff_t, new_size, 1576 (loff_t)index << PAGE_SHIFT); 1577 } 1578 1579 if (off_end) { 1580 ret = fill_zero(inode, pg_end, 0, off_end); 1581 if (ret) 1582 goto out; 1583 1584 new_size = max_t(loff_t, new_size, offset + len); 1585 } 1586 } 1587 1588 out: 1589 if (new_size > i_size_read(inode)) { 1590 if (mode & FALLOC_FL_KEEP_SIZE) 1591 file_set_keep_isize(inode); 1592 else 1593 f2fs_i_size_write(inode, new_size); 1594 } 1595 return ret; 1596 } 1597 1598 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1599 { 1600 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1601 struct address_space *mapping = inode->i_mapping; 1602 pgoff_t nr, pg_start, pg_end, delta, idx; 1603 loff_t new_size; 1604 int ret = 0; 1605 1606 new_size = i_size_read(inode) + len; 1607 ret = inode_newsize_ok(inode, new_size); 1608 if (ret) 1609 return ret; 1610 1611 if (offset >= i_size_read(inode)) 1612 return -EINVAL; 1613 1614 /* insert range should be aligned to block size of f2fs. */ 1615 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1616 return -EINVAL; 1617 1618 ret = f2fs_convert_inline_inode(inode); 1619 if (ret) 1620 return ret; 1621 1622 f2fs_balance_fs(sbi, true); 1623 1624 filemap_invalidate_lock(mapping); 1625 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1626 filemap_invalidate_unlock(mapping); 1627 if (ret) 1628 return ret; 1629 1630 /* write out all dirty pages from offset */ 1631 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1632 if (ret) 1633 return ret; 1634 1635 pg_start = offset >> PAGE_SHIFT; 1636 pg_end = (offset + len) >> PAGE_SHIFT; 1637 delta = pg_end - pg_start; 1638 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1639 1640 /* avoid gc operation during block exchange */ 1641 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1642 filemap_invalidate_lock(mapping); 1643 truncate_pagecache(inode, offset); 1644 1645 while (!ret && idx > pg_start) { 1646 nr = idx - pg_start; 1647 if (nr > delta) 1648 nr = delta; 1649 idx -= nr; 1650 1651 f2fs_lock_op(sbi); 1652 f2fs_drop_extent_tree(inode); 1653 1654 ret = __exchange_data_block(inode, inode, idx, 1655 idx + delta, nr, false); 1656 f2fs_unlock_op(sbi); 1657 } 1658 filemap_invalidate_unlock(mapping); 1659 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1660 1661 /* write out all moved pages, if possible */ 1662 filemap_invalidate_lock(mapping); 1663 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1664 truncate_pagecache(inode, offset); 1665 filemap_invalidate_unlock(mapping); 1666 1667 if (!ret) 1668 f2fs_i_size_write(inode, new_size); 1669 return ret; 1670 } 1671 1672 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1673 loff_t len, int mode) 1674 { 1675 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1676 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1677 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1678 .m_may_create = true }; 1679 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1680 .init_gc_type = FG_GC, 1681 .should_migrate_blocks = false, 1682 .err_gc_skipped = true, 1683 .nr_free_secs = 0 }; 1684 pgoff_t pg_start, pg_end; 1685 loff_t new_size; 1686 loff_t off_end; 1687 block_t expanded = 0; 1688 int err; 1689 1690 err = inode_newsize_ok(inode, (len + offset)); 1691 if (err) 1692 return err; 1693 1694 err = f2fs_convert_inline_inode(inode); 1695 if (err) 1696 return err; 1697 1698 f2fs_balance_fs(sbi, true); 1699 1700 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1701 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1702 off_end = (offset + len) & (PAGE_SIZE - 1); 1703 1704 map.m_lblk = pg_start; 1705 map.m_len = pg_end - pg_start; 1706 if (off_end) 1707 map.m_len++; 1708 1709 if (!map.m_len) 1710 return 0; 1711 1712 if (f2fs_is_pinned_file(inode)) { 1713 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1714 block_t sec_len = roundup(map.m_len, sec_blks); 1715 1716 map.m_len = sec_blks; 1717 next_alloc: 1718 if (has_not_enough_free_secs(sbi, 0, 1719 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1720 f2fs_down_write(&sbi->gc_lock); 1721 stat_inc_gc_call_count(sbi, FOREGROUND); 1722 err = f2fs_gc(sbi, &gc_control); 1723 if (err && err != -ENODATA) 1724 goto out_err; 1725 } 1726 1727 f2fs_down_write(&sbi->pin_sem); 1728 1729 f2fs_lock_op(sbi); 1730 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false); 1731 f2fs_unlock_op(sbi); 1732 1733 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1734 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1735 file_dont_truncate(inode); 1736 1737 f2fs_up_write(&sbi->pin_sem); 1738 1739 expanded += map.m_len; 1740 sec_len -= map.m_len; 1741 map.m_lblk += map.m_len; 1742 if (!err && sec_len) 1743 goto next_alloc; 1744 1745 map.m_len = expanded; 1746 } else { 1747 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1748 expanded = map.m_len; 1749 } 1750 out_err: 1751 if (err) { 1752 pgoff_t last_off; 1753 1754 if (!expanded) 1755 return err; 1756 1757 last_off = pg_start + expanded - 1; 1758 1759 /* update new size to the failed position */ 1760 new_size = (last_off == pg_end) ? offset + len : 1761 (loff_t)(last_off + 1) << PAGE_SHIFT; 1762 } else { 1763 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1764 } 1765 1766 if (new_size > i_size_read(inode)) { 1767 if (mode & FALLOC_FL_KEEP_SIZE) 1768 file_set_keep_isize(inode); 1769 else 1770 f2fs_i_size_write(inode, new_size); 1771 } 1772 1773 return err; 1774 } 1775 1776 static long f2fs_fallocate(struct file *file, int mode, 1777 loff_t offset, loff_t len) 1778 { 1779 struct inode *inode = file_inode(file); 1780 long ret = 0; 1781 1782 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1783 return -EIO; 1784 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1785 return -ENOSPC; 1786 if (!f2fs_is_compress_backend_ready(inode)) 1787 return -EOPNOTSUPP; 1788 1789 /* f2fs only support ->fallocate for regular file */ 1790 if (!S_ISREG(inode->i_mode)) 1791 return -EINVAL; 1792 1793 if (IS_ENCRYPTED(inode) && 1794 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1795 return -EOPNOTSUPP; 1796 1797 /* 1798 * Pinned file should not support partial truncation since the block 1799 * can be used by applications. 1800 */ 1801 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1802 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1803 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1804 return -EOPNOTSUPP; 1805 1806 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1807 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1808 FALLOC_FL_INSERT_RANGE)) 1809 return -EOPNOTSUPP; 1810 1811 inode_lock(inode); 1812 1813 ret = file_modified(file); 1814 if (ret) 1815 goto out; 1816 1817 if (mode & FALLOC_FL_PUNCH_HOLE) { 1818 if (offset >= inode->i_size) 1819 goto out; 1820 1821 ret = f2fs_punch_hole(inode, offset, len); 1822 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1823 ret = f2fs_collapse_range(inode, offset, len); 1824 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1825 ret = f2fs_zero_range(inode, offset, len, mode); 1826 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1827 ret = f2fs_insert_range(inode, offset, len); 1828 } else { 1829 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1830 } 1831 1832 if (!ret) { 1833 inode->i_mtime = inode_set_ctime_current(inode); 1834 f2fs_mark_inode_dirty_sync(inode, false); 1835 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1836 } 1837 1838 out: 1839 inode_unlock(inode); 1840 1841 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1842 return ret; 1843 } 1844 1845 static int f2fs_release_file(struct inode *inode, struct file *filp) 1846 { 1847 /* 1848 * f2fs_release_file is called at every close calls. So we should 1849 * not drop any inmemory pages by close called by other process. 1850 */ 1851 if (!(filp->f_mode & FMODE_WRITE) || 1852 atomic_read(&inode->i_writecount) != 1) 1853 return 0; 1854 1855 inode_lock(inode); 1856 f2fs_abort_atomic_write(inode, true); 1857 inode_unlock(inode); 1858 1859 return 0; 1860 } 1861 1862 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1863 { 1864 struct inode *inode = file_inode(file); 1865 1866 /* 1867 * If the process doing a transaction is crashed, we should do 1868 * roll-back. Otherwise, other reader/write can see corrupted database 1869 * until all the writers close its file. Since this should be done 1870 * before dropping file lock, it needs to do in ->flush. 1871 */ 1872 if (F2FS_I(inode)->atomic_write_task == current && 1873 (current->flags & PF_EXITING)) { 1874 inode_lock(inode); 1875 f2fs_abort_atomic_write(inode, true); 1876 inode_unlock(inode); 1877 } 1878 1879 return 0; 1880 } 1881 1882 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1883 { 1884 struct f2fs_inode_info *fi = F2FS_I(inode); 1885 u32 masked_flags = fi->i_flags & mask; 1886 1887 /* mask can be shrunk by flags_valid selector */ 1888 iflags &= mask; 1889 1890 /* Is it quota file? Do not allow user to mess with it */ 1891 if (IS_NOQUOTA(inode)) 1892 return -EPERM; 1893 1894 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1895 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1896 return -EOPNOTSUPP; 1897 if (!f2fs_empty_dir(inode)) 1898 return -ENOTEMPTY; 1899 } 1900 1901 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1902 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1903 return -EOPNOTSUPP; 1904 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1905 return -EINVAL; 1906 } 1907 1908 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1909 if (masked_flags & F2FS_COMPR_FL) { 1910 if (!f2fs_disable_compressed_file(inode)) 1911 return -EINVAL; 1912 } else { 1913 /* try to convert inline_data to support compression */ 1914 int err = f2fs_convert_inline_inode(inode); 1915 if (err) 1916 return err; 1917 1918 f2fs_down_write(&F2FS_I(inode)->i_sem); 1919 if (!f2fs_may_compress(inode) || 1920 (S_ISREG(inode->i_mode) && 1921 F2FS_HAS_BLOCKS(inode))) { 1922 f2fs_up_write(&F2FS_I(inode)->i_sem); 1923 return -EINVAL; 1924 } 1925 err = set_compress_context(inode); 1926 f2fs_up_write(&F2FS_I(inode)->i_sem); 1927 1928 if (err) 1929 return err; 1930 } 1931 } 1932 1933 fi->i_flags = iflags | (fi->i_flags & ~mask); 1934 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1935 (fi->i_flags & F2FS_NOCOMP_FL)); 1936 1937 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1938 set_inode_flag(inode, FI_PROJ_INHERIT); 1939 else 1940 clear_inode_flag(inode, FI_PROJ_INHERIT); 1941 1942 inode_set_ctime_current(inode); 1943 f2fs_set_inode_flags(inode); 1944 f2fs_mark_inode_dirty_sync(inode, true); 1945 return 0; 1946 } 1947 1948 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1949 1950 /* 1951 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1952 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1953 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1954 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1955 * 1956 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1957 * FS_IOC_FSSETXATTR is done by the VFS. 1958 */ 1959 1960 static const struct { 1961 u32 iflag; 1962 u32 fsflag; 1963 } f2fs_fsflags_map[] = { 1964 { F2FS_COMPR_FL, FS_COMPR_FL }, 1965 { F2FS_SYNC_FL, FS_SYNC_FL }, 1966 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1967 { F2FS_APPEND_FL, FS_APPEND_FL }, 1968 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1969 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1970 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1971 { F2FS_INDEX_FL, FS_INDEX_FL }, 1972 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1973 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1974 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1975 }; 1976 1977 #define F2FS_GETTABLE_FS_FL ( \ 1978 FS_COMPR_FL | \ 1979 FS_SYNC_FL | \ 1980 FS_IMMUTABLE_FL | \ 1981 FS_APPEND_FL | \ 1982 FS_NODUMP_FL | \ 1983 FS_NOATIME_FL | \ 1984 FS_NOCOMP_FL | \ 1985 FS_INDEX_FL | \ 1986 FS_DIRSYNC_FL | \ 1987 FS_PROJINHERIT_FL | \ 1988 FS_ENCRYPT_FL | \ 1989 FS_INLINE_DATA_FL | \ 1990 FS_NOCOW_FL | \ 1991 FS_VERITY_FL | \ 1992 FS_CASEFOLD_FL) 1993 1994 #define F2FS_SETTABLE_FS_FL ( \ 1995 FS_COMPR_FL | \ 1996 FS_SYNC_FL | \ 1997 FS_IMMUTABLE_FL | \ 1998 FS_APPEND_FL | \ 1999 FS_NODUMP_FL | \ 2000 FS_NOATIME_FL | \ 2001 FS_NOCOMP_FL | \ 2002 FS_DIRSYNC_FL | \ 2003 FS_PROJINHERIT_FL | \ 2004 FS_CASEFOLD_FL) 2005 2006 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2007 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2008 { 2009 u32 fsflags = 0; 2010 int i; 2011 2012 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2013 if (iflags & f2fs_fsflags_map[i].iflag) 2014 fsflags |= f2fs_fsflags_map[i].fsflag; 2015 2016 return fsflags; 2017 } 2018 2019 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2020 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2021 { 2022 u32 iflags = 0; 2023 int i; 2024 2025 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2026 if (fsflags & f2fs_fsflags_map[i].fsflag) 2027 iflags |= f2fs_fsflags_map[i].iflag; 2028 2029 return iflags; 2030 } 2031 2032 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2033 { 2034 struct inode *inode = file_inode(filp); 2035 2036 return put_user(inode->i_generation, (int __user *)arg); 2037 } 2038 2039 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2040 { 2041 struct inode *inode = file_inode(filp); 2042 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2043 struct f2fs_inode_info *fi = F2FS_I(inode); 2044 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2045 struct inode *pinode; 2046 loff_t isize; 2047 int ret; 2048 2049 if (!inode_owner_or_capable(idmap, inode)) 2050 return -EACCES; 2051 2052 if (!S_ISREG(inode->i_mode)) 2053 return -EINVAL; 2054 2055 if (filp->f_flags & O_DIRECT) 2056 return -EINVAL; 2057 2058 ret = mnt_want_write_file(filp); 2059 if (ret) 2060 return ret; 2061 2062 inode_lock(inode); 2063 2064 if (!f2fs_disable_compressed_file(inode)) { 2065 ret = -EINVAL; 2066 goto out; 2067 } 2068 2069 if (f2fs_is_atomic_file(inode)) 2070 goto out; 2071 2072 ret = f2fs_convert_inline_inode(inode); 2073 if (ret) 2074 goto out; 2075 2076 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2077 2078 /* 2079 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2080 * f2fs_is_atomic_file. 2081 */ 2082 if (get_dirty_pages(inode)) 2083 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2084 inode->i_ino, get_dirty_pages(inode)); 2085 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2086 if (ret) { 2087 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2088 goto out; 2089 } 2090 2091 /* Check if the inode already has a COW inode */ 2092 if (fi->cow_inode == NULL) { 2093 /* Create a COW inode for atomic write */ 2094 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2095 if (IS_ERR(pinode)) { 2096 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2097 ret = PTR_ERR(pinode); 2098 goto out; 2099 } 2100 2101 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2102 iput(pinode); 2103 if (ret) { 2104 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2105 goto out; 2106 } 2107 2108 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2109 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2110 } else { 2111 /* Reuse the already created COW inode */ 2112 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2113 if (ret) { 2114 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2115 goto out; 2116 } 2117 } 2118 2119 f2fs_write_inode(inode, NULL); 2120 2121 stat_inc_atomic_inode(inode); 2122 2123 set_inode_flag(inode, FI_ATOMIC_FILE); 2124 2125 isize = i_size_read(inode); 2126 fi->original_i_size = isize; 2127 if (truncate) { 2128 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2129 truncate_inode_pages_final(inode->i_mapping); 2130 f2fs_i_size_write(inode, 0); 2131 isize = 0; 2132 } 2133 f2fs_i_size_write(fi->cow_inode, isize); 2134 2135 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2136 2137 f2fs_update_time(sbi, REQ_TIME); 2138 fi->atomic_write_task = current; 2139 stat_update_max_atomic_write(inode); 2140 fi->atomic_write_cnt = 0; 2141 out: 2142 inode_unlock(inode); 2143 mnt_drop_write_file(filp); 2144 return ret; 2145 } 2146 2147 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2148 { 2149 struct inode *inode = file_inode(filp); 2150 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2151 int ret; 2152 2153 if (!inode_owner_or_capable(idmap, inode)) 2154 return -EACCES; 2155 2156 ret = mnt_want_write_file(filp); 2157 if (ret) 2158 return ret; 2159 2160 f2fs_balance_fs(F2FS_I_SB(inode), true); 2161 2162 inode_lock(inode); 2163 2164 if (f2fs_is_atomic_file(inode)) { 2165 ret = f2fs_commit_atomic_write(inode); 2166 if (!ret) 2167 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2168 2169 f2fs_abort_atomic_write(inode, ret); 2170 } else { 2171 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2172 } 2173 2174 inode_unlock(inode); 2175 mnt_drop_write_file(filp); 2176 return ret; 2177 } 2178 2179 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2180 { 2181 struct inode *inode = file_inode(filp); 2182 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2183 int ret; 2184 2185 if (!inode_owner_or_capable(idmap, inode)) 2186 return -EACCES; 2187 2188 ret = mnt_want_write_file(filp); 2189 if (ret) 2190 return ret; 2191 2192 inode_lock(inode); 2193 2194 f2fs_abort_atomic_write(inode, true); 2195 2196 inode_unlock(inode); 2197 2198 mnt_drop_write_file(filp); 2199 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2200 return ret; 2201 } 2202 2203 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2204 { 2205 struct inode *inode = file_inode(filp); 2206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2207 struct super_block *sb = sbi->sb; 2208 __u32 in; 2209 int ret = 0; 2210 2211 if (!capable(CAP_SYS_ADMIN)) 2212 return -EPERM; 2213 2214 if (get_user(in, (__u32 __user *)arg)) 2215 return -EFAULT; 2216 2217 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2218 ret = mnt_want_write_file(filp); 2219 if (ret) { 2220 if (ret == -EROFS) { 2221 ret = 0; 2222 f2fs_stop_checkpoint(sbi, false, 2223 STOP_CP_REASON_SHUTDOWN); 2224 trace_f2fs_shutdown(sbi, in, ret); 2225 } 2226 return ret; 2227 } 2228 } 2229 2230 switch (in) { 2231 case F2FS_GOING_DOWN_FULLSYNC: 2232 ret = freeze_bdev(sb->s_bdev); 2233 if (ret) 2234 goto out; 2235 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2236 thaw_bdev(sb->s_bdev); 2237 break; 2238 case F2FS_GOING_DOWN_METASYNC: 2239 /* do checkpoint only */ 2240 ret = f2fs_sync_fs(sb, 1); 2241 if (ret) 2242 goto out; 2243 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2244 break; 2245 case F2FS_GOING_DOWN_NOSYNC: 2246 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2247 break; 2248 case F2FS_GOING_DOWN_METAFLUSH: 2249 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2250 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2251 break; 2252 case F2FS_GOING_DOWN_NEED_FSCK: 2253 set_sbi_flag(sbi, SBI_NEED_FSCK); 2254 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2255 set_sbi_flag(sbi, SBI_IS_DIRTY); 2256 /* do checkpoint only */ 2257 ret = f2fs_sync_fs(sb, 1); 2258 goto out; 2259 default: 2260 ret = -EINVAL; 2261 goto out; 2262 } 2263 2264 f2fs_stop_gc_thread(sbi); 2265 f2fs_stop_discard_thread(sbi); 2266 2267 f2fs_drop_discard_cmd(sbi); 2268 clear_opt(sbi, DISCARD); 2269 2270 f2fs_update_time(sbi, REQ_TIME); 2271 out: 2272 if (in != F2FS_GOING_DOWN_FULLSYNC) 2273 mnt_drop_write_file(filp); 2274 2275 trace_f2fs_shutdown(sbi, in, ret); 2276 2277 return ret; 2278 } 2279 2280 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2281 { 2282 struct inode *inode = file_inode(filp); 2283 struct super_block *sb = inode->i_sb; 2284 struct fstrim_range range; 2285 int ret; 2286 2287 if (!capable(CAP_SYS_ADMIN)) 2288 return -EPERM; 2289 2290 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2291 return -EOPNOTSUPP; 2292 2293 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2294 sizeof(range))) 2295 return -EFAULT; 2296 2297 ret = mnt_want_write_file(filp); 2298 if (ret) 2299 return ret; 2300 2301 range.minlen = max((unsigned int)range.minlen, 2302 bdev_discard_granularity(sb->s_bdev)); 2303 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2304 mnt_drop_write_file(filp); 2305 if (ret < 0) 2306 return ret; 2307 2308 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2309 sizeof(range))) 2310 return -EFAULT; 2311 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2312 return 0; 2313 } 2314 2315 static bool uuid_is_nonzero(__u8 u[16]) 2316 { 2317 int i; 2318 2319 for (i = 0; i < 16; i++) 2320 if (u[i]) 2321 return true; 2322 return false; 2323 } 2324 2325 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2326 { 2327 struct inode *inode = file_inode(filp); 2328 2329 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2330 return -EOPNOTSUPP; 2331 2332 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2333 2334 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2335 } 2336 2337 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2338 { 2339 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2340 return -EOPNOTSUPP; 2341 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2342 } 2343 2344 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2345 { 2346 struct inode *inode = file_inode(filp); 2347 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2348 u8 encrypt_pw_salt[16]; 2349 int err; 2350 2351 if (!f2fs_sb_has_encrypt(sbi)) 2352 return -EOPNOTSUPP; 2353 2354 err = mnt_want_write_file(filp); 2355 if (err) 2356 return err; 2357 2358 f2fs_down_write(&sbi->sb_lock); 2359 2360 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2361 goto got_it; 2362 2363 /* update superblock with uuid */ 2364 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2365 2366 err = f2fs_commit_super(sbi, false); 2367 if (err) { 2368 /* undo new data */ 2369 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2370 goto out_err; 2371 } 2372 got_it: 2373 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2374 out_err: 2375 f2fs_up_write(&sbi->sb_lock); 2376 mnt_drop_write_file(filp); 2377 2378 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2379 err = -EFAULT; 2380 2381 return err; 2382 } 2383 2384 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2385 unsigned long arg) 2386 { 2387 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2388 return -EOPNOTSUPP; 2389 2390 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2391 } 2392 2393 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2394 { 2395 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2396 return -EOPNOTSUPP; 2397 2398 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2399 } 2400 2401 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2402 { 2403 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2404 return -EOPNOTSUPP; 2405 2406 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2407 } 2408 2409 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2410 unsigned long arg) 2411 { 2412 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2413 return -EOPNOTSUPP; 2414 2415 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2416 } 2417 2418 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2419 unsigned long arg) 2420 { 2421 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2422 return -EOPNOTSUPP; 2423 2424 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2425 } 2426 2427 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2428 { 2429 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2430 return -EOPNOTSUPP; 2431 2432 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2433 } 2434 2435 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2436 { 2437 struct inode *inode = file_inode(filp); 2438 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2439 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2440 .no_bg_gc = false, 2441 .should_migrate_blocks = false, 2442 .nr_free_secs = 0 }; 2443 __u32 sync; 2444 int ret; 2445 2446 if (!capable(CAP_SYS_ADMIN)) 2447 return -EPERM; 2448 2449 if (get_user(sync, (__u32 __user *)arg)) 2450 return -EFAULT; 2451 2452 if (f2fs_readonly(sbi->sb)) 2453 return -EROFS; 2454 2455 ret = mnt_want_write_file(filp); 2456 if (ret) 2457 return ret; 2458 2459 if (!sync) { 2460 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2461 ret = -EBUSY; 2462 goto out; 2463 } 2464 } else { 2465 f2fs_down_write(&sbi->gc_lock); 2466 } 2467 2468 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2469 gc_control.err_gc_skipped = sync; 2470 stat_inc_gc_call_count(sbi, FOREGROUND); 2471 ret = f2fs_gc(sbi, &gc_control); 2472 out: 2473 mnt_drop_write_file(filp); 2474 return ret; 2475 } 2476 2477 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2478 { 2479 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2480 struct f2fs_gc_control gc_control = { 2481 .init_gc_type = range->sync ? FG_GC : BG_GC, 2482 .no_bg_gc = false, 2483 .should_migrate_blocks = false, 2484 .err_gc_skipped = range->sync, 2485 .nr_free_secs = 0 }; 2486 u64 end; 2487 int ret; 2488 2489 if (!capable(CAP_SYS_ADMIN)) 2490 return -EPERM; 2491 if (f2fs_readonly(sbi->sb)) 2492 return -EROFS; 2493 2494 end = range->start + range->len; 2495 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2496 end >= MAX_BLKADDR(sbi)) 2497 return -EINVAL; 2498 2499 ret = mnt_want_write_file(filp); 2500 if (ret) 2501 return ret; 2502 2503 do_more: 2504 if (!range->sync) { 2505 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2506 ret = -EBUSY; 2507 goto out; 2508 } 2509 } else { 2510 f2fs_down_write(&sbi->gc_lock); 2511 } 2512 2513 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2514 stat_inc_gc_call_count(sbi, FOREGROUND); 2515 ret = f2fs_gc(sbi, &gc_control); 2516 if (ret) { 2517 if (ret == -EBUSY) 2518 ret = -EAGAIN; 2519 goto out; 2520 } 2521 range->start += CAP_BLKS_PER_SEC(sbi); 2522 if (range->start <= end) 2523 goto do_more; 2524 out: 2525 mnt_drop_write_file(filp); 2526 return ret; 2527 } 2528 2529 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2530 { 2531 struct f2fs_gc_range range; 2532 2533 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2534 sizeof(range))) 2535 return -EFAULT; 2536 return __f2fs_ioc_gc_range(filp, &range); 2537 } 2538 2539 static int f2fs_ioc_write_checkpoint(struct file *filp) 2540 { 2541 struct inode *inode = file_inode(filp); 2542 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2543 int ret; 2544 2545 if (!capable(CAP_SYS_ADMIN)) 2546 return -EPERM; 2547 2548 if (f2fs_readonly(sbi->sb)) 2549 return -EROFS; 2550 2551 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2552 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2553 return -EINVAL; 2554 } 2555 2556 ret = mnt_want_write_file(filp); 2557 if (ret) 2558 return ret; 2559 2560 ret = f2fs_sync_fs(sbi->sb, 1); 2561 2562 mnt_drop_write_file(filp); 2563 return ret; 2564 } 2565 2566 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2567 struct file *filp, 2568 struct f2fs_defragment *range) 2569 { 2570 struct inode *inode = file_inode(filp); 2571 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2572 .m_seg_type = NO_CHECK_TYPE, 2573 .m_may_create = false }; 2574 struct extent_info ei = {}; 2575 pgoff_t pg_start, pg_end, next_pgofs; 2576 unsigned int blk_per_seg = sbi->blocks_per_seg; 2577 unsigned int total = 0, sec_num; 2578 block_t blk_end = 0; 2579 bool fragmented = false; 2580 int err; 2581 2582 pg_start = range->start >> PAGE_SHIFT; 2583 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2584 2585 f2fs_balance_fs(sbi, true); 2586 2587 inode_lock(inode); 2588 2589 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2590 err = -EINVAL; 2591 goto unlock_out; 2592 } 2593 2594 /* if in-place-update policy is enabled, don't waste time here */ 2595 set_inode_flag(inode, FI_OPU_WRITE); 2596 if (f2fs_should_update_inplace(inode, NULL)) { 2597 err = -EINVAL; 2598 goto out; 2599 } 2600 2601 /* writeback all dirty pages in the range */ 2602 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2603 range->start + range->len - 1); 2604 if (err) 2605 goto out; 2606 2607 /* 2608 * lookup mapping info in extent cache, skip defragmenting if physical 2609 * block addresses are continuous. 2610 */ 2611 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2612 if (ei.fofs + ei.len >= pg_end) 2613 goto out; 2614 } 2615 2616 map.m_lblk = pg_start; 2617 map.m_next_pgofs = &next_pgofs; 2618 2619 /* 2620 * lookup mapping info in dnode page cache, skip defragmenting if all 2621 * physical block addresses are continuous even if there are hole(s) 2622 * in logical blocks. 2623 */ 2624 while (map.m_lblk < pg_end) { 2625 map.m_len = pg_end - map.m_lblk; 2626 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2627 if (err) 2628 goto out; 2629 2630 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2631 map.m_lblk = next_pgofs; 2632 continue; 2633 } 2634 2635 if (blk_end && blk_end != map.m_pblk) 2636 fragmented = true; 2637 2638 /* record total count of block that we're going to move */ 2639 total += map.m_len; 2640 2641 blk_end = map.m_pblk + map.m_len; 2642 2643 map.m_lblk += map.m_len; 2644 } 2645 2646 if (!fragmented) { 2647 total = 0; 2648 goto out; 2649 } 2650 2651 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2652 2653 /* 2654 * make sure there are enough free section for LFS allocation, this can 2655 * avoid defragment running in SSR mode when free section are allocated 2656 * intensively 2657 */ 2658 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2659 err = -EAGAIN; 2660 goto out; 2661 } 2662 2663 map.m_lblk = pg_start; 2664 map.m_len = pg_end - pg_start; 2665 total = 0; 2666 2667 while (map.m_lblk < pg_end) { 2668 pgoff_t idx; 2669 int cnt = 0; 2670 2671 do_map: 2672 map.m_len = pg_end - map.m_lblk; 2673 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2674 if (err) 2675 goto clear_out; 2676 2677 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2678 map.m_lblk = next_pgofs; 2679 goto check; 2680 } 2681 2682 set_inode_flag(inode, FI_SKIP_WRITES); 2683 2684 idx = map.m_lblk; 2685 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2686 struct page *page; 2687 2688 page = f2fs_get_lock_data_page(inode, idx, true); 2689 if (IS_ERR(page)) { 2690 err = PTR_ERR(page); 2691 goto clear_out; 2692 } 2693 2694 set_page_dirty(page); 2695 set_page_private_gcing(page); 2696 f2fs_put_page(page, 1); 2697 2698 idx++; 2699 cnt++; 2700 total++; 2701 } 2702 2703 map.m_lblk = idx; 2704 check: 2705 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2706 goto do_map; 2707 2708 clear_inode_flag(inode, FI_SKIP_WRITES); 2709 2710 err = filemap_fdatawrite(inode->i_mapping); 2711 if (err) 2712 goto out; 2713 } 2714 clear_out: 2715 clear_inode_flag(inode, FI_SKIP_WRITES); 2716 out: 2717 clear_inode_flag(inode, FI_OPU_WRITE); 2718 unlock_out: 2719 inode_unlock(inode); 2720 if (!err) 2721 range->len = (u64)total << PAGE_SHIFT; 2722 return err; 2723 } 2724 2725 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2726 { 2727 struct inode *inode = file_inode(filp); 2728 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2729 struct f2fs_defragment range; 2730 int err; 2731 2732 if (!capable(CAP_SYS_ADMIN)) 2733 return -EPERM; 2734 2735 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2736 return -EINVAL; 2737 2738 if (f2fs_readonly(sbi->sb)) 2739 return -EROFS; 2740 2741 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2742 sizeof(range))) 2743 return -EFAULT; 2744 2745 /* verify alignment of offset & size */ 2746 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2747 return -EINVAL; 2748 2749 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2750 max_file_blocks(inode))) 2751 return -EINVAL; 2752 2753 err = mnt_want_write_file(filp); 2754 if (err) 2755 return err; 2756 2757 err = f2fs_defragment_range(sbi, filp, &range); 2758 mnt_drop_write_file(filp); 2759 2760 f2fs_update_time(sbi, REQ_TIME); 2761 if (err < 0) 2762 return err; 2763 2764 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2765 sizeof(range))) 2766 return -EFAULT; 2767 2768 return 0; 2769 } 2770 2771 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2772 struct file *file_out, loff_t pos_out, size_t len) 2773 { 2774 struct inode *src = file_inode(file_in); 2775 struct inode *dst = file_inode(file_out); 2776 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2777 size_t olen = len, dst_max_i_size = 0; 2778 size_t dst_osize; 2779 int ret; 2780 2781 if (file_in->f_path.mnt != file_out->f_path.mnt || 2782 src->i_sb != dst->i_sb) 2783 return -EXDEV; 2784 2785 if (unlikely(f2fs_readonly(src->i_sb))) 2786 return -EROFS; 2787 2788 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2789 return -EINVAL; 2790 2791 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2792 return -EOPNOTSUPP; 2793 2794 if (pos_out < 0 || pos_in < 0) 2795 return -EINVAL; 2796 2797 if (src == dst) { 2798 if (pos_in == pos_out) 2799 return 0; 2800 if (pos_out > pos_in && pos_out < pos_in + len) 2801 return -EINVAL; 2802 } 2803 2804 inode_lock(src); 2805 if (src != dst) { 2806 ret = -EBUSY; 2807 if (!inode_trylock(dst)) 2808 goto out; 2809 } 2810 2811 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst)) { 2812 ret = -EOPNOTSUPP; 2813 goto out_unlock; 2814 } 2815 2816 ret = -EINVAL; 2817 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2818 goto out_unlock; 2819 if (len == 0) 2820 olen = len = src->i_size - pos_in; 2821 if (pos_in + len == src->i_size) 2822 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2823 if (len == 0) { 2824 ret = 0; 2825 goto out_unlock; 2826 } 2827 2828 dst_osize = dst->i_size; 2829 if (pos_out + olen > dst->i_size) 2830 dst_max_i_size = pos_out + olen; 2831 2832 /* verify the end result is block aligned */ 2833 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2834 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2835 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2836 goto out_unlock; 2837 2838 ret = f2fs_convert_inline_inode(src); 2839 if (ret) 2840 goto out_unlock; 2841 2842 ret = f2fs_convert_inline_inode(dst); 2843 if (ret) 2844 goto out_unlock; 2845 2846 /* write out all dirty pages from offset */ 2847 ret = filemap_write_and_wait_range(src->i_mapping, 2848 pos_in, pos_in + len); 2849 if (ret) 2850 goto out_unlock; 2851 2852 ret = filemap_write_and_wait_range(dst->i_mapping, 2853 pos_out, pos_out + len); 2854 if (ret) 2855 goto out_unlock; 2856 2857 f2fs_balance_fs(sbi, true); 2858 2859 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2860 if (src != dst) { 2861 ret = -EBUSY; 2862 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2863 goto out_src; 2864 } 2865 2866 f2fs_lock_op(sbi); 2867 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2868 pos_out >> F2FS_BLKSIZE_BITS, 2869 len >> F2FS_BLKSIZE_BITS, false); 2870 2871 if (!ret) { 2872 if (dst_max_i_size) 2873 f2fs_i_size_write(dst, dst_max_i_size); 2874 else if (dst_osize != dst->i_size) 2875 f2fs_i_size_write(dst, dst_osize); 2876 } 2877 f2fs_unlock_op(sbi); 2878 2879 if (src != dst) 2880 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2881 out_src: 2882 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2883 if (ret) 2884 goto out_unlock; 2885 2886 src->i_mtime = inode_set_ctime_current(src); 2887 f2fs_mark_inode_dirty_sync(src, false); 2888 if (src != dst) { 2889 dst->i_mtime = inode_set_ctime_current(dst); 2890 f2fs_mark_inode_dirty_sync(dst, false); 2891 } 2892 f2fs_update_time(sbi, REQ_TIME); 2893 2894 out_unlock: 2895 if (src != dst) 2896 inode_unlock(dst); 2897 out: 2898 inode_unlock(src); 2899 return ret; 2900 } 2901 2902 static int __f2fs_ioc_move_range(struct file *filp, 2903 struct f2fs_move_range *range) 2904 { 2905 struct fd dst; 2906 int err; 2907 2908 if (!(filp->f_mode & FMODE_READ) || 2909 !(filp->f_mode & FMODE_WRITE)) 2910 return -EBADF; 2911 2912 dst = fdget(range->dst_fd); 2913 if (!dst.file) 2914 return -EBADF; 2915 2916 if (!(dst.file->f_mode & FMODE_WRITE)) { 2917 err = -EBADF; 2918 goto err_out; 2919 } 2920 2921 err = mnt_want_write_file(filp); 2922 if (err) 2923 goto err_out; 2924 2925 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2926 range->pos_out, range->len); 2927 2928 mnt_drop_write_file(filp); 2929 err_out: 2930 fdput(dst); 2931 return err; 2932 } 2933 2934 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2935 { 2936 struct f2fs_move_range range; 2937 2938 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2939 sizeof(range))) 2940 return -EFAULT; 2941 return __f2fs_ioc_move_range(filp, &range); 2942 } 2943 2944 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2945 { 2946 struct inode *inode = file_inode(filp); 2947 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2948 struct sit_info *sm = SIT_I(sbi); 2949 unsigned int start_segno = 0, end_segno = 0; 2950 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2951 struct f2fs_flush_device range; 2952 struct f2fs_gc_control gc_control = { 2953 .init_gc_type = FG_GC, 2954 .should_migrate_blocks = true, 2955 .err_gc_skipped = true, 2956 .nr_free_secs = 0 }; 2957 int ret; 2958 2959 if (!capable(CAP_SYS_ADMIN)) 2960 return -EPERM; 2961 2962 if (f2fs_readonly(sbi->sb)) 2963 return -EROFS; 2964 2965 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2966 return -EINVAL; 2967 2968 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2969 sizeof(range))) 2970 return -EFAULT; 2971 2972 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2973 __is_large_section(sbi)) { 2974 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2975 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2976 return -EINVAL; 2977 } 2978 2979 ret = mnt_want_write_file(filp); 2980 if (ret) 2981 return ret; 2982 2983 if (range.dev_num != 0) 2984 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2985 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2986 2987 start_segno = sm->last_victim[FLUSH_DEVICE]; 2988 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2989 start_segno = dev_start_segno; 2990 end_segno = min(start_segno + range.segments, dev_end_segno); 2991 2992 while (start_segno < end_segno) { 2993 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2994 ret = -EBUSY; 2995 goto out; 2996 } 2997 sm->last_victim[GC_CB] = end_segno + 1; 2998 sm->last_victim[GC_GREEDY] = end_segno + 1; 2999 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3000 3001 gc_control.victim_segno = start_segno; 3002 stat_inc_gc_call_count(sbi, FOREGROUND); 3003 ret = f2fs_gc(sbi, &gc_control); 3004 if (ret == -EAGAIN) 3005 ret = 0; 3006 else if (ret < 0) 3007 break; 3008 start_segno++; 3009 } 3010 out: 3011 mnt_drop_write_file(filp); 3012 return ret; 3013 } 3014 3015 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3016 { 3017 struct inode *inode = file_inode(filp); 3018 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3019 3020 /* Must validate to set it with SQLite behavior in Android. */ 3021 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3022 3023 return put_user(sb_feature, (u32 __user *)arg); 3024 } 3025 3026 #ifdef CONFIG_QUOTA 3027 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3028 { 3029 struct dquot *transfer_to[MAXQUOTAS] = {}; 3030 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3031 struct super_block *sb = sbi->sb; 3032 int err; 3033 3034 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3035 if (IS_ERR(transfer_to[PRJQUOTA])) 3036 return PTR_ERR(transfer_to[PRJQUOTA]); 3037 3038 err = __dquot_transfer(inode, transfer_to); 3039 if (err) 3040 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3041 dqput(transfer_to[PRJQUOTA]); 3042 return err; 3043 } 3044 3045 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3046 { 3047 struct f2fs_inode_info *fi = F2FS_I(inode); 3048 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3049 struct f2fs_inode *ri = NULL; 3050 kprojid_t kprojid; 3051 int err; 3052 3053 if (!f2fs_sb_has_project_quota(sbi)) { 3054 if (projid != F2FS_DEF_PROJID) 3055 return -EOPNOTSUPP; 3056 else 3057 return 0; 3058 } 3059 3060 if (!f2fs_has_extra_attr(inode)) 3061 return -EOPNOTSUPP; 3062 3063 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3064 3065 if (projid_eq(kprojid, fi->i_projid)) 3066 return 0; 3067 3068 err = -EPERM; 3069 /* Is it quota file? Do not allow user to mess with it */ 3070 if (IS_NOQUOTA(inode)) 3071 return err; 3072 3073 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3074 return -EOVERFLOW; 3075 3076 err = f2fs_dquot_initialize(inode); 3077 if (err) 3078 return err; 3079 3080 f2fs_lock_op(sbi); 3081 err = f2fs_transfer_project_quota(inode, kprojid); 3082 if (err) 3083 goto out_unlock; 3084 3085 fi->i_projid = kprojid; 3086 inode_set_ctime_current(inode); 3087 f2fs_mark_inode_dirty_sync(inode, true); 3088 out_unlock: 3089 f2fs_unlock_op(sbi); 3090 return err; 3091 } 3092 #else 3093 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3094 { 3095 return 0; 3096 } 3097 3098 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3099 { 3100 if (projid != F2FS_DEF_PROJID) 3101 return -EOPNOTSUPP; 3102 return 0; 3103 } 3104 #endif 3105 3106 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3107 { 3108 struct inode *inode = d_inode(dentry); 3109 struct f2fs_inode_info *fi = F2FS_I(inode); 3110 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3111 3112 if (IS_ENCRYPTED(inode)) 3113 fsflags |= FS_ENCRYPT_FL; 3114 if (IS_VERITY(inode)) 3115 fsflags |= FS_VERITY_FL; 3116 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3117 fsflags |= FS_INLINE_DATA_FL; 3118 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3119 fsflags |= FS_NOCOW_FL; 3120 3121 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3122 3123 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3124 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3125 3126 return 0; 3127 } 3128 3129 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3130 struct dentry *dentry, struct fileattr *fa) 3131 { 3132 struct inode *inode = d_inode(dentry); 3133 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3134 u32 iflags; 3135 int err; 3136 3137 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3138 return -EIO; 3139 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3140 return -ENOSPC; 3141 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3142 return -EOPNOTSUPP; 3143 fsflags &= F2FS_SETTABLE_FS_FL; 3144 if (!fa->flags_valid) 3145 mask &= FS_COMMON_FL; 3146 3147 iflags = f2fs_fsflags_to_iflags(fsflags); 3148 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3149 return -EOPNOTSUPP; 3150 3151 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3152 if (!err) 3153 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3154 3155 return err; 3156 } 3157 3158 int f2fs_pin_file_control(struct inode *inode, bool inc) 3159 { 3160 struct f2fs_inode_info *fi = F2FS_I(inode); 3161 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3162 3163 /* Use i_gc_failures for normal file as a risk signal. */ 3164 if (inc) 3165 f2fs_i_gc_failures_write(inode, 3166 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3167 3168 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3169 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3170 __func__, inode->i_ino, 3171 fi->i_gc_failures[GC_FAILURE_PIN]); 3172 clear_inode_flag(inode, FI_PIN_FILE); 3173 return -EAGAIN; 3174 } 3175 return 0; 3176 } 3177 3178 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3179 { 3180 struct inode *inode = file_inode(filp); 3181 __u32 pin; 3182 int ret = 0; 3183 3184 if (get_user(pin, (__u32 __user *)arg)) 3185 return -EFAULT; 3186 3187 if (!S_ISREG(inode->i_mode)) 3188 return -EINVAL; 3189 3190 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3191 return -EROFS; 3192 3193 ret = mnt_want_write_file(filp); 3194 if (ret) 3195 return ret; 3196 3197 inode_lock(inode); 3198 3199 if (!pin) { 3200 clear_inode_flag(inode, FI_PIN_FILE); 3201 f2fs_i_gc_failures_write(inode, 0); 3202 goto done; 3203 } 3204 3205 if (f2fs_should_update_outplace(inode, NULL)) { 3206 ret = -EINVAL; 3207 goto out; 3208 } 3209 3210 if (f2fs_pin_file_control(inode, false)) { 3211 ret = -EAGAIN; 3212 goto out; 3213 } 3214 3215 ret = f2fs_convert_inline_inode(inode); 3216 if (ret) 3217 goto out; 3218 3219 if (!f2fs_disable_compressed_file(inode)) { 3220 ret = -EOPNOTSUPP; 3221 goto out; 3222 } 3223 3224 set_inode_flag(inode, FI_PIN_FILE); 3225 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3226 done: 3227 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3228 out: 3229 inode_unlock(inode); 3230 mnt_drop_write_file(filp); 3231 return ret; 3232 } 3233 3234 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3235 { 3236 struct inode *inode = file_inode(filp); 3237 __u32 pin = 0; 3238 3239 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3240 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3241 return put_user(pin, (u32 __user *)arg); 3242 } 3243 3244 int f2fs_precache_extents(struct inode *inode) 3245 { 3246 struct f2fs_inode_info *fi = F2FS_I(inode); 3247 struct f2fs_map_blocks map; 3248 pgoff_t m_next_extent; 3249 loff_t end; 3250 int err; 3251 3252 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3253 return -EOPNOTSUPP; 3254 3255 map.m_lblk = 0; 3256 map.m_pblk = 0; 3257 map.m_next_pgofs = NULL; 3258 map.m_next_extent = &m_next_extent; 3259 map.m_seg_type = NO_CHECK_TYPE; 3260 map.m_may_create = false; 3261 end = max_file_blocks(inode); 3262 3263 while (map.m_lblk < end) { 3264 map.m_len = end - map.m_lblk; 3265 3266 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3267 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3268 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3269 if (err) 3270 return err; 3271 3272 map.m_lblk = m_next_extent; 3273 } 3274 3275 return 0; 3276 } 3277 3278 static int f2fs_ioc_precache_extents(struct file *filp) 3279 { 3280 return f2fs_precache_extents(file_inode(filp)); 3281 } 3282 3283 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3284 { 3285 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3286 __u64 block_count; 3287 3288 if (!capable(CAP_SYS_ADMIN)) 3289 return -EPERM; 3290 3291 if (f2fs_readonly(sbi->sb)) 3292 return -EROFS; 3293 3294 if (copy_from_user(&block_count, (void __user *)arg, 3295 sizeof(block_count))) 3296 return -EFAULT; 3297 3298 return f2fs_resize_fs(filp, block_count); 3299 } 3300 3301 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3302 { 3303 struct inode *inode = file_inode(filp); 3304 3305 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3306 3307 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3308 f2fs_warn(F2FS_I_SB(inode), 3309 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3310 inode->i_ino); 3311 return -EOPNOTSUPP; 3312 } 3313 3314 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3315 } 3316 3317 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3318 { 3319 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3320 return -EOPNOTSUPP; 3321 3322 return fsverity_ioctl_measure(filp, (void __user *)arg); 3323 } 3324 3325 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3326 { 3327 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3328 return -EOPNOTSUPP; 3329 3330 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3331 } 3332 3333 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3334 { 3335 struct inode *inode = file_inode(filp); 3336 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3337 char *vbuf; 3338 int count; 3339 int err = 0; 3340 3341 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3342 if (!vbuf) 3343 return -ENOMEM; 3344 3345 f2fs_down_read(&sbi->sb_lock); 3346 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3347 ARRAY_SIZE(sbi->raw_super->volume_name), 3348 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3349 f2fs_up_read(&sbi->sb_lock); 3350 3351 if (copy_to_user((char __user *)arg, vbuf, 3352 min(FSLABEL_MAX, count))) 3353 err = -EFAULT; 3354 3355 kfree(vbuf); 3356 return err; 3357 } 3358 3359 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3360 { 3361 struct inode *inode = file_inode(filp); 3362 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3363 char *vbuf; 3364 int err = 0; 3365 3366 if (!capable(CAP_SYS_ADMIN)) 3367 return -EPERM; 3368 3369 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3370 if (IS_ERR(vbuf)) 3371 return PTR_ERR(vbuf); 3372 3373 err = mnt_want_write_file(filp); 3374 if (err) 3375 goto out; 3376 3377 f2fs_down_write(&sbi->sb_lock); 3378 3379 memset(sbi->raw_super->volume_name, 0, 3380 sizeof(sbi->raw_super->volume_name)); 3381 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3382 sbi->raw_super->volume_name, 3383 ARRAY_SIZE(sbi->raw_super->volume_name)); 3384 3385 err = f2fs_commit_super(sbi, false); 3386 3387 f2fs_up_write(&sbi->sb_lock); 3388 3389 mnt_drop_write_file(filp); 3390 out: 3391 kfree(vbuf); 3392 return err; 3393 } 3394 3395 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3396 { 3397 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3398 return -EOPNOTSUPP; 3399 3400 if (!f2fs_compressed_file(inode)) 3401 return -EINVAL; 3402 3403 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3404 3405 return 0; 3406 } 3407 3408 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3409 { 3410 struct inode *inode = file_inode(filp); 3411 __u64 blocks; 3412 int ret; 3413 3414 ret = f2fs_get_compress_blocks(inode, &blocks); 3415 if (ret < 0) 3416 return ret; 3417 3418 return put_user(blocks, (u64 __user *)arg); 3419 } 3420 3421 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3422 { 3423 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3424 unsigned int released_blocks = 0; 3425 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3426 block_t blkaddr; 3427 int i; 3428 3429 for (i = 0; i < count; i++) { 3430 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3431 dn->ofs_in_node + i); 3432 3433 if (!__is_valid_data_blkaddr(blkaddr)) 3434 continue; 3435 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3436 DATA_GENERIC_ENHANCE))) { 3437 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3438 return -EFSCORRUPTED; 3439 } 3440 } 3441 3442 while (count) { 3443 int compr_blocks = 0; 3444 3445 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3446 blkaddr = f2fs_data_blkaddr(dn); 3447 3448 if (i == 0) { 3449 if (blkaddr == COMPRESS_ADDR) 3450 continue; 3451 dn->ofs_in_node += cluster_size; 3452 goto next; 3453 } 3454 3455 if (__is_valid_data_blkaddr(blkaddr)) 3456 compr_blocks++; 3457 3458 if (blkaddr != NEW_ADDR) 3459 continue; 3460 3461 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3462 } 3463 3464 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3465 dec_valid_block_count(sbi, dn->inode, 3466 cluster_size - compr_blocks); 3467 3468 released_blocks += cluster_size - compr_blocks; 3469 next: 3470 count -= cluster_size; 3471 } 3472 3473 return released_blocks; 3474 } 3475 3476 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3477 { 3478 struct inode *inode = file_inode(filp); 3479 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3480 pgoff_t page_idx = 0, last_idx; 3481 unsigned int released_blocks = 0; 3482 int ret; 3483 int writecount; 3484 3485 if (!f2fs_sb_has_compression(sbi)) 3486 return -EOPNOTSUPP; 3487 3488 if (!f2fs_compressed_file(inode)) 3489 return -EINVAL; 3490 3491 if (f2fs_readonly(sbi->sb)) 3492 return -EROFS; 3493 3494 ret = mnt_want_write_file(filp); 3495 if (ret) 3496 return ret; 3497 3498 f2fs_balance_fs(sbi, true); 3499 3500 inode_lock(inode); 3501 3502 writecount = atomic_read(&inode->i_writecount); 3503 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3504 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3505 ret = -EBUSY; 3506 goto out; 3507 } 3508 3509 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3510 ret = -EINVAL; 3511 goto out; 3512 } 3513 3514 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3515 if (ret) 3516 goto out; 3517 3518 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3519 ret = -EPERM; 3520 goto out; 3521 } 3522 3523 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3524 inode_set_ctime_current(inode); 3525 f2fs_mark_inode_dirty_sync(inode, true); 3526 3527 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3528 filemap_invalidate_lock(inode->i_mapping); 3529 3530 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3531 3532 while (page_idx < last_idx) { 3533 struct dnode_of_data dn; 3534 pgoff_t end_offset, count; 3535 3536 set_new_dnode(&dn, inode, NULL, NULL, 0); 3537 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3538 if (ret) { 3539 if (ret == -ENOENT) { 3540 page_idx = f2fs_get_next_page_offset(&dn, 3541 page_idx); 3542 ret = 0; 3543 continue; 3544 } 3545 break; 3546 } 3547 3548 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3549 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3550 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3551 3552 ret = release_compress_blocks(&dn, count); 3553 3554 f2fs_put_dnode(&dn); 3555 3556 if (ret < 0) 3557 break; 3558 3559 page_idx += count; 3560 released_blocks += ret; 3561 } 3562 3563 filemap_invalidate_unlock(inode->i_mapping); 3564 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3565 out: 3566 inode_unlock(inode); 3567 3568 mnt_drop_write_file(filp); 3569 3570 if (ret >= 0) { 3571 ret = put_user(released_blocks, (u64 __user *)arg); 3572 } else if (released_blocks && 3573 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3574 set_sbi_flag(sbi, SBI_NEED_FSCK); 3575 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3576 "iblocks=%llu, released=%u, compr_blocks=%u, " 3577 "run fsck to fix.", 3578 __func__, inode->i_ino, inode->i_blocks, 3579 released_blocks, 3580 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3581 } 3582 3583 return ret; 3584 } 3585 3586 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3587 unsigned int *reserved_blocks) 3588 { 3589 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3590 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3591 block_t blkaddr; 3592 int i; 3593 3594 for (i = 0; i < count; i++) { 3595 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3596 dn->ofs_in_node + i); 3597 3598 if (!__is_valid_data_blkaddr(blkaddr)) 3599 continue; 3600 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3601 DATA_GENERIC_ENHANCE))) { 3602 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); 3603 return -EFSCORRUPTED; 3604 } 3605 } 3606 3607 while (count) { 3608 int compr_blocks = 0; 3609 blkcnt_t reserved; 3610 int ret; 3611 3612 for (i = 0; i < cluster_size; i++) { 3613 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3614 dn->ofs_in_node + i); 3615 3616 if (i == 0) { 3617 if (blkaddr != COMPRESS_ADDR) { 3618 dn->ofs_in_node += cluster_size; 3619 goto next; 3620 } 3621 continue; 3622 } 3623 3624 /* 3625 * compressed cluster was not released due to it 3626 * fails in release_compress_blocks(), so NEW_ADDR 3627 * is a possible case. 3628 */ 3629 if (blkaddr == NEW_ADDR || 3630 __is_valid_data_blkaddr(blkaddr)) { 3631 compr_blocks++; 3632 continue; 3633 } 3634 } 3635 3636 reserved = cluster_size - compr_blocks; 3637 3638 /* for the case all blocks in cluster were reserved */ 3639 if (reserved == 1) 3640 goto next; 3641 3642 ret = inc_valid_block_count(sbi, dn->inode, &reserved, false); 3643 if (unlikely(ret)) 3644 return ret; 3645 3646 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3647 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3648 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3649 } 3650 3651 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3652 3653 *reserved_blocks += reserved; 3654 next: 3655 count -= cluster_size; 3656 } 3657 3658 return 0; 3659 } 3660 3661 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3662 { 3663 struct inode *inode = file_inode(filp); 3664 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3665 pgoff_t page_idx = 0, last_idx; 3666 unsigned int reserved_blocks = 0; 3667 int ret; 3668 3669 if (!f2fs_sb_has_compression(sbi)) 3670 return -EOPNOTSUPP; 3671 3672 if (!f2fs_compressed_file(inode)) 3673 return -EINVAL; 3674 3675 if (f2fs_readonly(sbi->sb)) 3676 return -EROFS; 3677 3678 ret = mnt_want_write_file(filp); 3679 if (ret) 3680 return ret; 3681 3682 f2fs_balance_fs(sbi, true); 3683 3684 inode_lock(inode); 3685 3686 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3687 ret = -EINVAL; 3688 goto unlock_inode; 3689 } 3690 3691 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3692 goto unlock_inode; 3693 3694 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3695 filemap_invalidate_lock(inode->i_mapping); 3696 3697 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3698 3699 while (page_idx < last_idx) { 3700 struct dnode_of_data dn; 3701 pgoff_t end_offset, count; 3702 3703 set_new_dnode(&dn, inode, NULL, NULL, 0); 3704 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3705 if (ret) { 3706 if (ret == -ENOENT) { 3707 page_idx = f2fs_get_next_page_offset(&dn, 3708 page_idx); 3709 ret = 0; 3710 continue; 3711 } 3712 break; 3713 } 3714 3715 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3716 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3717 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3718 3719 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 3720 3721 f2fs_put_dnode(&dn); 3722 3723 if (ret < 0) 3724 break; 3725 3726 page_idx += count; 3727 } 3728 3729 filemap_invalidate_unlock(inode->i_mapping); 3730 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3731 3732 if (!ret) { 3733 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3734 inode_set_ctime_current(inode); 3735 f2fs_mark_inode_dirty_sync(inode, true); 3736 } 3737 unlock_inode: 3738 inode_unlock(inode); 3739 mnt_drop_write_file(filp); 3740 3741 if (!ret) { 3742 ret = put_user(reserved_blocks, (u64 __user *)arg); 3743 } else if (reserved_blocks && 3744 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3745 set_sbi_flag(sbi, SBI_NEED_FSCK); 3746 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3747 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3748 "run fsck to fix.", 3749 __func__, inode->i_ino, inode->i_blocks, 3750 reserved_blocks, 3751 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3752 } 3753 3754 return ret; 3755 } 3756 3757 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3758 pgoff_t off, block_t block, block_t len, u32 flags) 3759 { 3760 sector_t sector = SECTOR_FROM_BLOCK(block); 3761 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3762 int ret = 0; 3763 3764 if (flags & F2FS_TRIM_FILE_DISCARD) { 3765 if (bdev_max_secure_erase_sectors(bdev)) 3766 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3767 GFP_NOFS); 3768 else 3769 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3770 GFP_NOFS); 3771 } 3772 3773 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3774 if (IS_ENCRYPTED(inode)) 3775 ret = fscrypt_zeroout_range(inode, off, block, len); 3776 else 3777 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3778 GFP_NOFS, 0); 3779 } 3780 3781 return ret; 3782 } 3783 3784 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3785 { 3786 struct inode *inode = file_inode(filp); 3787 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3788 struct address_space *mapping = inode->i_mapping; 3789 struct block_device *prev_bdev = NULL; 3790 struct f2fs_sectrim_range range; 3791 pgoff_t index, pg_end, prev_index = 0; 3792 block_t prev_block = 0, len = 0; 3793 loff_t end_addr; 3794 bool to_end = false; 3795 int ret = 0; 3796 3797 if (!(filp->f_mode & FMODE_WRITE)) 3798 return -EBADF; 3799 3800 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3801 sizeof(range))) 3802 return -EFAULT; 3803 3804 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3805 !S_ISREG(inode->i_mode)) 3806 return -EINVAL; 3807 3808 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3809 !f2fs_hw_support_discard(sbi)) || 3810 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3811 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3812 return -EOPNOTSUPP; 3813 3814 file_start_write(filp); 3815 inode_lock(inode); 3816 3817 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3818 range.start >= inode->i_size) { 3819 ret = -EINVAL; 3820 goto err; 3821 } 3822 3823 if (range.len == 0) 3824 goto err; 3825 3826 if (inode->i_size - range.start > range.len) { 3827 end_addr = range.start + range.len; 3828 } else { 3829 end_addr = range.len == (u64)-1 ? 3830 sbi->sb->s_maxbytes : inode->i_size; 3831 to_end = true; 3832 } 3833 3834 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3835 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3836 ret = -EINVAL; 3837 goto err; 3838 } 3839 3840 index = F2FS_BYTES_TO_BLK(range.start); 3841 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3842 3843 ret = f2fs_convert_inline_inode(inode); 3844 if (ret) 3845 goto err; 3846 3847 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3848 filemap_invalidate_lock(mapping); 3849 3850 ret = filemap_write_and_wait_range(mapping, range.start, 3851 to_end ? LLONG_MAX : end_addr - 1); 3852 if (ret) 3853 goto out; 3854 3855 truncate_inode_pages_range(mapping, range.start, 3856 to_end ? -1 : end_addr - 1); 3857 3858 while (index < pg_end) { 3859 struct dnode_of_data dn; 3860 pgoff_t end_offset, count; 3861 int i; 3862 3863 set_new_dnode(&dn, inode, NULL, NULL, 0); 3864 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3865 if (ret) { 3866 if (ret == -ENOENT) { 3867 index = f2fs_get_next_page_offset(&dn, index); 3868 continue; 3869 } 3870 goto out; 3871 } 3872 3873 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3874 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3875 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3876 struct block_device *cur_bdev; 3877 block_t blkaddr = f2fs_data_blkaddr(&dn); 3878 3879 if (!__is_valid_data_blkaddr(blkaddr)) 3880 continue; 3881 3882 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3883 DATA_GENERIC_ENHANCE)) { 3884 ret = -EFSCORRUPTED; 3885 f2fs_put_dnode(&dn); 3886 f2fs_handle_error(sbi, 3887 ERROR_INVALID_BLKADDR); 3888 goto out; 3889 } 3890 3891 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3892 if (f2fs_is_multi_device(sbi)) { 3893 int di = f2fs_target_device_index(sbi, blkaddr); 3894 3895 blkaddr -= FDEV(di).start_blk; 3896 } 3897 3898 if (len) { 3899 if (prev_bdev == cur_bdev && 3900 index == prev_index + len && 3901 blkaddr == prev_block + len) { 3902 len++; 3903 } else { 3904 ret = f2fs_secure_erase(prev_bdev, 3905 inode, prev_index, prev_block, 3906 len, range.flags); 3907 if (ret) { 3908 f2fs_put_dnode(&dn); 3909 goto out; 3910 } 3911 3912 len = 0; 3913 } 3914 } 3915 3916 if (!len) { 3917 prev_bdev = cur_bdev; 3918 prev_index = index; 3919 prev_block = blkaddr; 3920 len = 1; 3921 } 3922 } 3923 3924 f2fs_put_dnode(&dn); 3925 3926 if (fatal_signal_pending(current)) { 3927 ret = -EINTR; 3928 goto out; 3929 } 3930 cond_resched(); 3931 } 3932 3933 if (len) 3934 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3935 prev_block, len, range.flags); 3936 out: 3937 filemap_invalidate_unlock(mapping); 3938 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3939 err: 3940 inode_unlock(inode); 3941 file_end_write(filp); 3942 3943 return ret; 3944 } 3945 3946 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3947 { 3948 struct inode *inode = file_inode(filp); 3949 struct f2fs_comp_option option; 3950 3951 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3952 return -EOPNOTSUPP; 3953 3954 inode_lock_shared(inode); 3955 3956 if (!f2fs_compressed_file(inode)) { 3957 inode_unlock_shared(inode); 3958 return -ENODATA; 3959 } 3960 3961 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3962 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3963 3964 inode_unlock_shared(inode); 3965 3966 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3967 sizeof(option))) 3968 return -EFAULT; 3969 3970 return 0; 3971 } 3972 3973 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3974 { 3975 struct inode *inode = file_inode(filp); 3976 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3977 struct f2fs_comp_option option; 3978 int ret = 0; 3979 3980 if (!f2fs_sb_has_compression(sbi)) 3981 return -EOPNOTSUPP; 3982 3983 if (!(filp->f_mode & FMODE_WRITE)) 3984 return -EBADF; 3985 3986 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3987 sizeof(option))) 3988 return -EFAULT; 3989 3990 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3991 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3992 option.algorithm >= COMPRESS_MAX) 3993 return -EINVAL; 3994 3995 file_start_write(filp); 3996 inode_lock(inode); 3997 3998 f2fs_down_write(&F2FS_I(inode)->i_sem); 3999 if (!f2fs_compressed_file(inode)) { 4000 ret = -EINVAL; 4001 goto out; 4002 } 4003 4004 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4005 ret = -EBUSY; 4006 goto out; 4007 } 4008 4009 if (F2FS_HAS_BLOCKS(inode)) { 4010 ret = -EFBIG; 4011 goto out; 4012 } 4013 4014 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4015 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4016 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4017 /* Set default level */ 4018 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4019 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4020 else 4021 F2FS_I(inode)->i_compress_level = 0; 4022 /* Adjust mount option level */ 4023 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4024 F2FS_OPTION(sbi).compress_level) 4025 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4026 f2fs_mark_inode_dirty_sync(inode, true); 4027 4028 if (!f2fs_is_compress_backend_ready(inode)) 4029 f2fs_warn(sbi, "compression algorithm is successfully set, " 4030 "but current kernel doesn't support this algorithm."); 4031 out: 4032 f2fs_up_write(&F2FS_I(inode)->i_sem); 4033 inode_unlock(inode); 4034 file_end_write(filp); 4035 4036 return ret; 4037 } 4038 4039 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4040 { 4041 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4042 struct address_space *mapping = inode->i_mapping; 4043 struct page *page; 4044 pgoff_t redirty_idx = page_idx; 4045 int i, page_len = 0, ret = 0; 4046 4047 page_cache_ra_unbounded(&ractl, len, 0); 4048 4049 for (i = 0; i < len; i++, page_idx++) { 4050 page = read_cache_page(mapping, page_idx, NULL, NULL); 4051 if (IS_ERR(page)) { 4052 ret = PTR_ERR(page); 4053 break; 4054 } 4055 page_len++; 4056 } 4057 4058 for (i = 0; i < page_len; i++, redirty_idx++) { 4059 page = find_lock_page(mapping, redirty_idx); 4060 4061 /* It will never fail, when page has pinned above */ 4062 f2fs_bug_on(F2FS_I_SB(inode), !page); 4063 4064 set_page_dirty(page); 4065 set_page_private_gcing(page); 4066 f2fs_put_page(page, 1); 4067 f2fs_put_page(page, 0); 4068 } 4069 4070 return ret; 4071 } 4072 4073 static int f2fs_ioc_decompress_file(struct file *filp) 4074 { 4075 struct inode *inode = file_inode(filp); 4076 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4077 struct f2fs_inode_info *fi = F2FS_I(inode); 4078 pgoff_t page_idx = 0, last_idx; 4079 unsigned int blk_per_seg = sbi->blocks_per_seg; 4080 int cluster_size = fi->i_cluster_size; 4081 int count, ret; 4082 4083 if (!f2fs_sb_has_compression(sbi) || 4084 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4085 return -EOPNOTSUPP; 4086 4087 if (!(filp->f_mode & FMODE_WRITE)) 4088 return -EBADF; 4089 4090 if (!f2fs_compressed_file(inode)) 4091 return -EINVAL; 4092 4093 f2fs_balance_fs(sbi, true); 4094 4095 file_start_write(filp); 4096 inode_lock(inode); 4097 4098 if (!f2fs_is_compress_backend_ready(inode)) { 4099 ret = -EOPNOTSUPP; 4100 goto out; 4101 } 4102 4103 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4104 ret = -EINVAL; 4105 goto out; 4106 } 4107 4108 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4109 if (ret) 4110 goto out; 4111 4112 if (!atomic_read(&fi->i_compr_blocks)) 4113 goto out; 4114 4115 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4116 4117 count = last_idx - page_idx; 4118 while (count && count >= cluster_size) { 4119 ret = redirty_blocks(inode, page_idx, cluster_size); 4120 if (ret < 0) 4121 break; 4122 4123 if (get_dirty_pages(inode) >= blk_per_seg) { 4124 ret = filemap_fdatawrite(inode->i_mapping); 4125 if (ret < 0) 4126 break; 4127 } 4128 4129 count -= cluster_size; 4130 page_idx += cluster_size; 4131 4132 cond_resched(); 4133 if (fatal_signal_pending(current)) { 4134 ret = -EINTR; 4135 break; 4136 } 4137 } 4138 4139 if (!ret) 4140 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4141 LLONG_MAX); 4142 4143 if (ret) 4144 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4145 __func__, ret); 4146 out: 4147 inode_unlock(inode); 4148 file_end_write(filp); 4149 4150 return ret; 4151 } 4152 4153 static int f2fs_ioc_compress_file(struct file *filp) 4154 { 4155 struct inode *inode = file_inode(filp); 4156 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4157 pgoff_t page_idx = 0, last_idx; 4158 unsigned int blk_per_seg = sbi->blocks_per_seg; 4159 int cluster_size = F2FS_I(inode)->i_cluster_size; 4160 int count, ret; 4161 4162 if (!f2fs_sb_has_compression(sbi) || 4163 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4164 return -EOPNOTSUPP; 4165 4166 if (!(filp->f_mode & FMODE_WRITE)) 4167 return -EBADF; 4168 4169 if (!f2fs_compressed_file(inode)) 4170 return -EINVAL; 4171 4172 f2fs_balance_fs(sbi, true); 4173 4174 file_start_write(filp); 4175 inode_lock(inode); 4176 4177 if (!f2fs_is_compress_backend_ready(inode)) { 4178 ret = -EOPNOTSUPP; 4179 goto out; 4180 } 4181 4182 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4183 ret = -EINVAL; 4184 goto out; 4185 } 4186 4187 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4188 if (ret) 4189 goto out; 4190 4191 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4192 4193 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4194 4195 count = last_idx - page_idx; 4196 while (count && count >= cluster_size) { 4197 ret = redirty_blocks(inode, page_idx, cluster_size); 4198 if (ret < 0) 4199 break; 4200 4201 if (get_dirty_pages(inode) >= blk_per_seg) { 4202 ret = filemap_fdatawrite(inode->i_mapping); 4203 if (ret < 0) 4204 break; 4205 } 4206 4207 count -= cluster_size; 4208 page_idx += cluster_size; 4209 4210 cond_resched(); 4211 if (fatal_signal_pending(current)) { 4212 ret = -EINTR; 4213 break; 4214 } 4215 } 4216 4217 if (!ret) 4218 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4219 LLONG_MAX); 4220 4221 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4222 4223 if (ret) 4224 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4225 __func__, ret); 4226 out: 4227 inode_unlock(inode); 4228 file_end_write(filp); 4229 4230 return ret; 4231 } 4232 4233 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4234 { 4235 switch (cmd) { 4236 case FS_IOC_GETVERSION: 4237 return f2fs_ioc_getversion(filp, arg); 4238 case F2FS_IOC_START_ATOMIC_WRITE: 4239 return f2fs_ioc_start_atomic_write(filp, false); 4240 case F2FS_IOC_START_ATOMIC_REPLACE: 4241 return f2fs_ioc_start_atomic_write(filp, true); 4242 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4243 return f2fs_ioc_commit_atomic_write(filp); 4244 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4245 return f2fs_ioc_abort_atomic_write(filp); 4246 case F2FS_IOC_START_VOLATILE_WRITE: 4247 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4248 return -EOPNOTSUPP; 4249 case F2FS_IOC_SHUTDOWN: 4250 return f2fs_ioc_shutdown(filp, arg); 4251 case FITRIM: 4252 return f2fs_ioc_fitrim(filp, arg); 4253 case FS_IOC_SET_ENCRYPTION_POLICY: 4254 return f2fs_ioc_set_encryption_policy(filp, arg); 4255 case FS_IOC_GET_ENCRYPTION_POLICY: 4256 return f2fs_ioc_get_encryption_policy(filp, arg); 4257 case FS_IOC_GET_ENCRYPTION_PWSALT: 4258 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4259 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4260 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4261 case FS_IOC_ADD_ENCRYPTION_KEY: 4262 return f2fs_ioc_add_encryption_key(filp, arg); 4263 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4264 return f2fs_ioc_remove_encryption_key(filp, arg); 4265 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4266 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4267 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4268 return f2fs_ioc_get_encryption_key_status(filp, arg); 4269 case FS_IOC_GET_ENCRYPTION_NONCE: 4270 return f2fs_ioc_get_encryption_nonce(filp, arg); 4271 case F2FS_IOC_GARBAGE_COLLECT: 4272 return f2fs_ioc_gc(filp, arg); 4273 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4274 return f2fs_ioc_gc_range(filp, arg); 4275 case F2FS_IOC_WRITE_CHECKPOINT: 4276 return f2fs_ioc_write_checkpoint(filp); 4277 case F2FS_IOC_DEFRAGMENT: 4278 return f2fs_ioc_defragment(filp, arg); 4279 case F2FS_IOC_MOVE_RANGE: 4280 return f2fs_ioc_move_range(filp, arg); 4281 case F2FS_IOC_FLUSH_DEVICE: 4282 return f2fs_ioc_flush_device(filp, arg); 4283 case F2FS_IOC_GET_FEATURES: 4284 return f2fs_ioc_get_features(filp, arg); 4285 case F2FS_IOC_GET_PIN_FILE: 4286 return f2fs_ioc_get_pin_file(filp, arg); 4287 case F2FS_IOC_SET_PIN_FILE: 4288 return f2fs_ioc_set_pin_file(filp, arg); 4289 case F2FS_IOC_PRECACHE_EXTENTS: 4290 return f2fs_ioc_precache_extents(filp); 4291 case F2FS_IOC_RESIZE_FS: 4292 return f2fs_ioc_resize_fs(filp, arg); 4293 case FS_IOC_ENABLE_VERITY: 4294 return f2fs_ioc_enable_verity(filp, arg); 4295 case FS_IOC_MEASURE_VERITY: 4296 return f2fs_ioc_measure_verity(filp, arg); 4297 case FS_IOC_READ_VERITY_METADATA: 4298 return f2fs_ioc_read_verity_metadata(filp, arg); 4299 case FS_IOC_GETFSLABEL: 4300 return f2fs_ioc_getfslabel(filp, arg); 4301 case FS_IOC_SETFSLABEL: 4302 return f2fs_ioc_setfslabel(filp, arg); 4303 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4304 return f2fs_ioc_get_compress_blocks(filp, arg); 4305 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4306 return f2fs_release_compress_blocks(filp, arg); 4307 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4308 return f2fs_reserve_compress_blocks(filp, arg); 4309 case F2FS_IOC_SEC_TRIM_FILE: 4310 return f2fs_sec_trim_file(filp, arg); 4311 case F2FS_IOC_GET_COMPRESS_OPTION: 4312 return f2fs_ioc_get_compress_option(filp, arg); 4313 case F2FS_IOC_SET_COMPRESS_OPTION: 4314 return f2fs_ioc_set_compress_option(filp, arg); 4315 case F2FS_IOC_DECOMPRESS_FILE: 4316 return f2fs_ioc_decompress_file(filp); 4317 case F2FS_IOC_COMPRESS_FILE: 4318 return f2fs_ioc_compress_file(filp); 4319 default: 4320 return -ENOTTY; 4321 } 4322 } 4323 4324 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4325 { 4326 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4327 return -EIO; 4328 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4329 return -ENOSPC; 4330 4331 return __f2fs_ioctl(filp, cmd, arg); 4332 } 4333 4334 /* 4335 * Return %true if the given read or write request should use direct I/O, or 4336 * %false if it should use buffered I/O. 4337 */ 4338 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4339 struct iov_iter *iter) 4340 { 4341 unsigned int align; 4342 4343 if (!(iocb->ki_flags & IOCB_DIRECT)) 4344 return false; 4345 4346 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4347 return false; 4348 4349 /* 4350 * Direct I/O not aligned to the disk's logical_block_size will be 4351 * attempted, but will fail with -EINVAL. 4352 * 4353 * f2fs additionally requires that direct I/O be aligned to the 4354 * filesystem block size, which is often a stricter requirement. 4355 * However, f2fs traditionally falls back to buffered I/O on requests 4356 * that are logical_block_size-aligned but not fs-block aligned. 4357 * 4358 * The below logic implements this behavior. 4359 */ 4360 align = iocb->ki_pos | iov_iter_alignment(iter); 4361 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4362 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4363 return false; 4364 4365 return true; 4366 } 4367 4368 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4369 unsigned int flags) 4370 { 4371 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4372 4373 dec_page_count(sbi, F2FS_DIO_READ); 4374 if (error) 4375 return error; 4376 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4377 return 0; 4378 } 4379 4380 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4381 .end_io = f2fs_dio_read_end_io, 4382 }; 4383 4384 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4385 { 4386 struct file *file = iocb->ki_filp; 4387 struct inode *inode = file_inode(file); 4388 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4389 struct f2fs_inode_info *fi = F2FS_I(inode); 4390 const loff_t pos = iocb->ki_pos; 4391 const size_t count = iov_iter_count(to); 4392 struct iomap_dio *dio; 4393 ssize_t ret; 4394 4395 if (count == 0) 4396 return 0; /* skip atime update */ 4397 4398 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4399 4400 if (iocb->ki_flags & IOCB_NOWAIT) { 4401 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4402 ret = -EAGAIN; 4403 goto out; 4404 } 4405 } else { 4406 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4407 } 4408 4409 /* 4410 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4411 * the higher-level function iomap_dio_rw() in order to ensure that the 4412 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4413 */ 4414 inc_page_count(sbi, F2FS_DIO_READ); 4415 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4416 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4417 if (IS_ERR_OR_NULL(dio)) { 4418 ret = PTR_ERR_OR_ZERO(dio); 4419 if (ret != -EIOCBQUEUED) 4420 dec_page_count(sbi, F2FS_DIO_READ); 4421 } else { 4422 ret = iomap_dio_complete(dio); 4423 } 4424 4425 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4426 4427 file_accessed(file); 4428 out: 4429 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4430 return ret; 4431 } 4432 4433 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4434 int rw) 4435 { 4436 struct inode *inode = file_inode(file); 4437 char *buf, *path; 4438 4439 buf = f2fs_getname(F2FS_I_SB(inode)); 4440 if (!buf) 4441 return; 4442 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4443 if (IS_ERR(path)) 4444 goto free_buf; 4445 if (rw == WRITE) 4446 trace_f2fs_datawrite_start(inode, pos, count, 4447 current->pid, path, current->comm); 4448 else 4449 trace_f2fs_dataread_start(inode, pos, count, 4450 current->pid, path, current->comm); 4451 free_buf: 4452 f2fs_putname(buf); 4453 } 4454 4455 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4456 { 4457 struct inode *inode = file_inode(iocb->ki_filp); 4458 const loff_t pos = iocb->ki_pos; 4459 ssize_t ret; 4460 4461 if (!f2fs_is_compress_backend_ready(inode)) 4462 return -EOPNOTSUPP; 4463 4464 if (trace_f2fs_dataread_start_enabled()) 4465 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4466 iov_iter_count(to), READ); 4467 4468 if (f2fs_should_use_dio(inode, iocb, to)) { 4469 ret = f2fs_dio_read_iter(iocb, to); 4470 } else { 4471 ret = filemap_read(iocb, to, 0); 4472 if (ret > 0) 4473 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4474 APP_BUFFERED_READ_IO, ret); 4475 } 4476 if (trace_f2fs_dataread_end_enabled()) 4477 trace_f2fs_dataread_end(inode, pos, ret); 4478 return ret; 4479 } 4480 4481 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4482 struct pipe_inode_info *pipe, 4483 size_t len, unsigned int flags) 4484 { 4485 struct inode *inode = file_inode(in); 4486 const loff_t pos = *ppos; 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(in, pos, len, READ); 4494 4495 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4496 if (ret > 0) 4497 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4498 APP_BUFFERED_READ_IO, ret); 4499 4500 if (trace_f2fs_dataread_end_enabled()) 4501 trace_f2fs_dataread_end(inode, pos, ret); 4502 return ret; 4503 } 4504 4505 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4506 { 4507 struct file *file = iocb->ki_filp; 4508 struct inode *inode = file_inode(file); 4509 ssize_t count; 4510 int err; 4511 4512 if (IS_IMMUTABLE(inode)) 4513 return -EPERM; 4514 4515 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4516 return -EPERM; 4517 4518 count = generic_write_checks(iocb, from); 4519 if (count <= 0) 4520 return count; 4521 4522 err = file_modified(file); 4523 if (err) 4524 return err; 4525 return count; 4526 } 4527 4528 /* 4529 * Preallocate blocks for a write request, if it is possible and helpful to do 4530 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4531 * blocks were preallocated, or a negative errno value if something went 4532 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4533 * requested blocks (not just some of them) have been allocated. 4534 */ 4535 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4536 bool dio) 4537 { 4538 struct inode *inode = file_inode(iocb->ki_filp); 4539 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4540 const loff_t pos = iocb->ki_pos; 4541 const size_t count = iov_iter_count(iter); 4542 struct f2fs_map_blocks map = {}; 4543 int flag; 4544 int ret; 4545 4546 /* If it will be an out-of-place direct write, don't bother. */ 4547 if (dio && f2fs_lfs_mode(sbi)) 4548 return 0; 4549 /* 4550 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4551 * buffered IO, if DIO meets any holes. 4552 */ 4553 if (dio && i_size_read(inode) && 4554 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4555 return 0; 4556 4557 /* No-wait I/O can't allocate blocks. */ 4558 if (iocb->ki_flags & IOCB_NOWAIT) 4559 return 0; 4560 4561 /* If it will be a short write, don't bother. */ 4562 if (fault_in_iov_iter_readable(iter, count)) 4563 return 0; 4564 4565 if (f2fs_has_inline_data(inode)) { 4566 /* If the data will fit inline, don't bother. */ 4567 if (pos + count <= MAX_INLINE_DATA(inode)) 4568 return 0; 4569 ret = f2fs_convert_inline_inode(inode); 4570 if (ret) 4571 return ret; 4572 } 4573 4574 /* Do not preallocate blocks that will be written partially in 4KB. */ 4575 map.m_lblk = F2FS_BLK_ALIGN(pos); 4576 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4577 if (map.m_len > map.m_lblk) 4578 map.m_len -= map.m_lblk; 4579 else 4580 map.m_len = 0; 4581 map.m_may_create = true; 4582 if (dio) { 4583 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4584 flag = F2FS_GET_BLOCK_PRE_DIO; 4585 } else { 4586 map.m_seg_type = NO_CHECK_TYPE; 4587 flag = F2FS_GET_BLOCK_PRE_AIO; 4588 } 4589 4590 ret = f2fs_map_blocks(inode, &map, flag); 4591 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4592 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4593 return ret; 4594 if (ret == 0) 4595 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4596 return map.m_len; 4597 } 4598 4599 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4600 struct iov_iter *from) 4601 { 4602 struct file *file = iocb->ki_filp; 4603 struct inode *inode = file_inode(file); 4604 ssize_t ret; 4605 4606 if (iocb->ki_flags & IOCB_NOWAIT) 4607 return -EOPNOTSUPP; 4608 4609 ret = generic_perform_write(iocb, from); 4610 4611 if (ret > 0) { 4612 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4613 APP_BUFFERED_IO, ret); 4614 } 4615 return ret; 4616 } 4617 4618 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4619 unsigned int flags) 4620 { 4621 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4622 4623 dec_page_count(sbi, F2FS_DIO_WRITE); 4624 if (error) 4625 return error; 4626 f2fs_update_time(sbi, REQ_TIME); 4627 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4628 return 0; 4629 } 4630 4631 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4632 .end_io = f2fs_dio_write_end_io, 4633 }; 4634 4635 static void f2fs_flush_buffered_write(struct address_space *mapping, 4636 loff_t start_pos, loff_t end_pos) 4637 { 4638 int ret; 4639 4640 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4641 if (ret < 0) 4642 return; 4643 invalidate_mapping_pages(mapping, 4644 start_pos >> PAGE_SHIFT, 4645 end_pos >> PAGE_SHIFT); 4646 } 4647 4648 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4649 bool *may_need_sync) 4650 { 4651 struct file *file = iocb->ki_filp; 4652 struct inode *inode = file_inode(file); 4653 struct f2fs_inode_info *fi = F2FS_I(inode); 4654 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4655 const bool do_opu = f2fs_lfs_mode(sbi); 4656 const loff_t pos = iocb->ki_pos; 4657 const ssize_t count = iov_iter_count(from); 4658 unsigned int dio_flags; 4659 struct iomap_dio *dio; 4660 ssize_t ret; 4661 4662 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4663 4664 if (iocb->ki_flags & IOCB_NOWAIT) { 4665 /* f2fs_convert_inline_inode() and block allocation can block */ 4666 if (f2fs_has_inline_data(inode) || 4667 !f2fs_overwrite_io(inode, pos, count)) { 4668 ret = -EAGAIN; 4669 goto out; 4670 } 4671 4672 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4673 ret = -EAGAIN; 4674 goto out; 4675 } 4676 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4677 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4678 ret = -EAGAIN; 4679 goto out; 4680 } 4681 } else { 4682 ret = f2fs_convert_inline_inode(inode); 4683 if (ret) 4684 goto out; 4685 4686 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4687 if (do_opu) 4688 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4689 } 4690 4691 /* 4692 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4693 * the higher-level function iomap_dio_rw() in order to ensure that the 4694 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4695 */ 4696 inc_page_count(sbi, F2FS_DIO_WRITE); 4697 dio_flags = 0; 4698 if (pos + count > inode->i_size) 4699 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4700 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4701 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4702 if (IS_ERR_OR_NULL(dio)) { 4703 ret = PTR_ERR_OR_ZERO(dio); 4704 if (ret == -ENOTBLK) 4705 ret = 0; 4706 if (ret != -EIOCBQUEUED) 4707 dec_page_count(sbi, F2FS_DIO_WRITE); 4708 } else { 4709 ret = iomap_dio_complete(dio); 4710 } 4711 4712 if (do_opu) 4713 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4714 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4715 4716 if (ret < 0) 4717 goto out; 4718 if (pos + ret > inode->i_size) 4719 f2fs_i_size_write(inode, pos + ret); 4720 if (!do_opu) 4721 set_inode_flag(inode, FI_UPDATE_WRITE); 4722 4723 if (iov_iter_count(from)) { 4724 ssize_t ret2; 4725 loff_t bufio_start_pos = iocb->ki_pos; 4726 4727 /* 4728 * The direct write was partial, so we need to fall back to a 4729 * buffered write for the remainder. 4730 */ 4731 4732 ret2 = f2fs_buffered_write_iter(iocb, from); 4733 if (iov_iter_count(from)) 4734 f2fs_write_failed(inode, iocb->ki_pos); 4735 if (ret2 < 0) 4736 goto out; 4737 4738 /* 4739 * Ensure that the pagecache pages are written to disk and 4740 * invalidated to preserve the expected O_DIRECT semantics. 4741 */ 4742 if (ret2 > 0) { 4743 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4744 4745 ret += ret2; 4746 4747 f2fs_flush_buffered_write(file->f_mapping, 4748 bufio_start_pos, 4749 bufio_end_pos); 4750 } 4751 } else { 4752 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4753 *may_need_sync = false; 4754 } 4755 out: 4756 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4757 return ret; 4758 } 4759 4760 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4761 { 4762 struct inode *inode = file_inode(iocb->ki_filp); 4763 const loff_t orig_pos = iocb->ki_pos; 4764 const size_t orig_count = iov_iter_count(from); 4765 loff_t target_size; 4766 bool dio; 4767 bool may_need_sync = true; 4768 int preallocated; 4769 ssize_t ret; 4770 4771 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4772 ret = -EIO; 4773 goto out; 4774 } 4775 4776 if (!f2fs_is_compress_backend_ready(inode)) { 4777 ret = -EOPNOTSUPP; 4778 goto out; 4779 } 4780 4781 if (iocb->ki_flags & IOCB_NOWAIT) { 4782 if (!inode_trylock(inode)) { 4783 ret = -EAGAIN; 4784 goto out; 4785 } 4786 } else { 4787 inode_lock(inode); 4788 } 4789 4790 ret = f2fs_write_checks(iocb, from); 4791 if (ret <= 0) 4792 goto out_unlock; 4793 4794 /* Determine whether we will do a direct write or a buffered write. */ 4795 dio = f2fs_should_use_dio(inode, iocb, from); 4796 4797 /* Possibly preallocate the blocks for the write. */ 4798 target_size = iocb->ki_pos + iov_iter_count(from); 4799 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4800 if (preallocated < 0) { 4801 ret = preallocated; 4802 } else { 4803 if (trace_f2fs_datawrite_start_enabled()) 4804 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4805 orig_count, WRITE); 4806 4807 /* Do the actual write. */ 4808 ret = dio ? 4809 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4810 f2fs_buffered_write_iter(iocb, from); 4811 4812 if (trace_f2fs_datawrite_end_enabled()) 4813 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4814 } 4815 4816 /* Don't leave any preallocated blocks around past i_size. */ 4817 if (preallocated && i_size_read(inode) < target_size) { 4818 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4819 filemap_invalidate_lock(inode->i_mapping); 4820 if (!f2fs_truncate(inode)) 4821 file_dont_truncate(inode); 4822 filemap_invalidate_unlock(inode->i_mapping); 4823 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4824 } else { 4825 file_dont_truncate(inode); 4826 } 4827 4828 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4829 out_unlock: 4830 inode_unlock(inode); 4831 out: 4832 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4833 4834 if (ret > 0 && may_need_sync) 4835 ret = generic_write_sync(iocb, ret); 4836 4837 /* If buffered IO was forced, flush and drop the data from 4838 * the page cache to preserve O_DIRECT semantics 4839 */ 4840 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4841 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4842 orig_pos, 4843 orig_pos + ret - 1); 4844 4845 return ret; 4846 } 4847 4848 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4849 int advice) 4850 { 4851 struct address_space *mapping; 4852 struct backing_dev_info *bdi; 4853 struct inode *inode = file_inode(filp); 4854 int err; 4855 4856 if (advice == POSIX_FADV_SEQUENTIAL) { 4857 if (S_ISFIFO(inode->i_mode)) 4858 return -ESPIPE; 4859 4860 mapping = filp->f_mapping; 4861 if (!mapping || len < 0) 4862 return -EINVAL; 4863 4864 bdi = inode_to_bdi(mapping->host); 4865 filp->f_ra.ra_pages = bdi->ra_pages * 4866 F2FS_I_SB(inode)->seq_file_ra_mul; 4867 spin_lock(&filp->f_lock); 4868 filp->f_mode &= ~FMODE_RANDOM; 4869 spin_unlock(&filp->f_lock); 4870 return 0; 4871 } 4872 4873 err = generic_fadvise(filp, offset, len, advice); 4874 if (!err && advice == POSIX_FADV_DONTNEED && 4875 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4876 f2fs_compressed_file(inode)) 4877 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4878 4879 return err; 4880 } 4881 4882 #ifdef CONFIG_COMPAT 4883 struct compat_f2fs_gc_range { 4884 u32 sync; 4885 compat_u64 start; 4886 compat_u64 len; 4887 }; 4888 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4889 struct compat_f2fs_gc_range) 4890 4891 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4892 { 4893 struct compat_f2fs_gc_range __user *urange; 4894 struct f2fs_gc_range range; 4895 int err; 4896 4897 urange = compat_ptr(arg); 4898 err = get_user(range.sync, &urange->sync); 4899 err |= get_user(range.start, &urange->start); 4900 err |= get_user(range.len, &urange->len); 4901 if (err) 4902 return -EFAULT; 4903 4904 return __f2fs_ioc_gc_range(file, &range); 4905 } 4906 4907 struct compat_f2fs_move_range { 4908 u32 dst_fd; 4909 compat_u64 pos_in; 4910 compat_u64 pos_out; 4911 compat_u64 len; 4912 }; 4913 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4914 struct compat_f2fs_move_range) 4915 4916 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4917 { 4918 struct compat_f2fs_move_range __user *urange; 4919 struct f2fs_move_range range; 4920 int err; 4921 4922 urange = compat_ptr(arg); 4923 err = get_user(range.dst_fd, &urange->dst_fd); 4924 err |= get_user(range.pos_in, &urange->pos_in); 4925 err |= get_user(range.pos_out, &urange->pos_out); 4926 err |= get_user(range.len, &urange->len); 4927 if (err) 4928 return -EFAULT; 4929 4930 return __f2fs_ioc_move_range(file, &range); 4931 } 4932 4933 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4934 { 4935 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4936 return -EIO; 4937 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4938 return -ENOSPC; 4939 4940 switch (cmd) { 4941 case FS_IOC32_GETVERSION: 4942 cmd = FS_IOC_GETVERSION; 4943 break; 4944 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4945 return f2fs_compat_ioc_gc_range(file, arg); 4946 case F2FS_IOC32_MOVE_RANGE: 4947 return f2fs_compat_ioc_move_range(file, arg); 4948 case F2FS_IOC_START_ATOMIC_WRITE: 4949 case F2FS_IOC_START_ATOMIC_REPLACE: 4950 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4951 case F2FS_IOC_START_VOLATILE_WRITE: 4952 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4953 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4954 case F2FS_IOC_SHUTDOWN: 4955 case FITRIM: 4956 case FS_IOC_SET_ENCRYPTION_POLICY: 4957 case FS_IOC_GET_ENCRYPTION_PWSALT: 4958 case FS_IOC_GET_ENCRYPTION_POLICY: 4959 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4960 case FS_IOC_ADD_ENCRYPTION_KEY: 4961 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4962 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4963 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4964 case FS_IOC_GET_ENCRYPTION_NONCE: 4965 case F2FS_IOC_GARBAGE_COLLECT: 4966 case F2FS_IOC_WRITE_CHECKPOINT: 4967 case F2FS_IOC_DEFRAGMENT: 4968 case F2FS_IOC_FLUSH_DEVICE: 4969 case F2FS_IOC_GET_FEATURES: 4970 case F2FS_IOC_GET_PIN_FILE: 4971 case F2FS_IOC_SET_PIN_FILE: 4972 case F2FS_IOC_PRECACHE_EXTENTS: 4973 case F2FS_IOC_RESIZE_FS: 4974 case FS_IOC_ENABLE_VERITY: 4975 case FS_IOC_MEASURE_VERITY: 4976 case FS_IOC_READ_VERITY_METADATA: 4977 case FS_IOC_GETFSLABEL: 4978 case FS_IOC_SETFSLABEL: 4979 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4980 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4981 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4982 case F2FS_IOC_SEC_TRIM_FILE: 4983 case F2FS_IOC_GET_COMPRESS_OPTION: 4984 case F2FS_IOC_SET_COMPRESS_OPTION: 4985 case F2FS_IOC_DECOMPRESS_FILE: 4986 case F2FS_IOC_COMPRESS_FILE: 4987 break; 4988 default: 4989 return -ENOIOCTLCMD; 4990 } 4991 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4992 } 4993 #endif 4994 4995 const struct file_operations f2fs_file_operations = { 4996 .llseek = f2fs_llseek, 4997 .read_iter = f2fs_file_read_iter, 4998 .write_iter = f2fs_file_write_iter, 4999 .iopoll = iocb_bio_iopoll, 5000 .open = f2fs_file_open, 5001 .release = f2fs_release_file, 5002 .mmap = f2fs_file_mmap, 5003 .flush = f2fs_file_flush, 5004 .fsync = f2fs_sync_file, 5005 .fallocate = f2fs_fallocate, 5006 .unlocked_ioctl = f2fs_ioctl, 5007 #ifdef CONFIG_COMPAT 5008 .compat_ioctl = f2fs_compat_ioctl, 5009 #endif 5010 .splice_read = f2fs_file_splice_read, 5011 .splice_write = iter_file_splice_write, 5012 .fadvise = f2fs_file_fadvise, 5013 }; 5014