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) 46 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO, 47 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 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true); 117 set_new_dnode(&dn, inode, NULL, NULL, 0); 118 err = f2fs_get_block(&dn, page->index); 119 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false); 120 } 121 122 #ifdef CONFIG_F2FS_FS_COMPRESSION 123 if (!need_alloc) { 124 set_new_dnode(&dn, inode, NULL, NULL, 0); 125 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE); 126 f2fs_put_dnode(&dn); 127 } 128 #endif 129 if (err) { 130 unlock_page(page); 131 goto out_sem; 132 } 133 134 f2fs_wait_on_page_writeback(page, DATA, false, true); 135 136 /* wait for GCed page writeback via META_MAPPING */ 137 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 138 139 /* 140 * check to see if the page is mapped already (no holes) 141 */ 142 if (PageMappedToDisk(page)) 143 goto out_sem; 144 145 /* page is wholly or partially inside EOF */ 146 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 147 i_size_read(inode)) { 148 loff_t offset; 149 150 offset = i_size_read(inode) & ~PAGE_MASK; 151 zero_user_segment(page, offset, PAGE_SIZE); 152 } 153 set_page_dirty(page); 154 if (!PageUptodate(page)) 155 SetPageUptodate(page); 156 157 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE); 158 f2fs_update_time(sbi, REQ_TIME); 159 160 trace_f2fs_vm_page_mkwrite(page, DATA); 161 out_sem: 162 filemap_invalidate_unlock_shared(inode->i_mapping); 163 164 sb_end_pagefault(inode->i_sb); 165 err: 166 return block_page_mkwrite_return(err); 167 } 168 169 static const struct vm_operations_struct f2fs_file_vm_ops = { 170 .fault = f2fs_filemap_fault, 171 .map_pages = filemap_map_pages, 172 .page_mkwrite = f2fs_vm_page_mkwrite, 173 }; 174 175 static int get_parent_ino(struct inode *inode, nid_t *pino) 176 { 177 struct dentry *dentry; 178 179 /* 180 * Make sure to get the non-deleted alias. The alias associated with 181 * the open file descriptor being fsync()'ed may be deleted already. 182 */ 183 dentry = d_find_alias(inode); 184 if (!dentry) 185 return 0; 186 187 *pino = parent_ino(dentry); 188 dput(dentry); 189 return 1; 190 } 191 192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 193 { 194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 195 enum cp_reason_type cp_reason = CP_NO_NEEDED; 196 197 if (!S_ISREG(inode->i_mode)) 198 cp_reason = CP_NON_REGULAR; 199 else if (f2fs_compressed_file(inode)) 200 cp_reason = CP_COMPRESSED; 201 else if (inode->i_nlink != 1) 202 cp_reason = CP_HARDLINK; 203 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 204 cp_reason = CP_SB_NEED_CP; 205 else if (file_wrong_pino(inode)) 206 cp_reason = CP_WRONG_PINO; 207 else if (!f2fs_space_for_roll_forward(sbi)) 208 cp_reason = CP_NO_SPC_ROLL; 209 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 210 cp_reason = CP_NODE_NEED_CP; 211 else if (test_opt(sbi, FASTBOOT)) 212 cp_reason = CP_FASTBOOT_MODE; 213 else if (F2FS_OPTION(sbi).active_logs == 2) 214 cp_reason = CP_SPEC_LOG_NUM; 215 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 216 f2fs_need_dentry_mark(sbi, inode->i_ino) && 217 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 218 TRANS_DIR_INO)) 219 cp_reason = CP_RECOVER_DIR; 220 221 return cp_reason; 222 } 223 224 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 225 { 226 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 227 bool ret = false; 228 /* But we need to avoid that there are some inode updates */ 229 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 230 ret = true; 231 f2fs_put_page(i, 0); 232 return ret; 233 } 234 235 static void try_to_fix_pino(struct inode *inode) 236 { 237 struct f2fs_inode_info *fi = F2FS_I(inode); 238 nid_t pino; 239 240 f2fs_down_write(&fi->i_sem); 241 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 242 get_parent_ino(inode, &pino)) { 243 f2fs_i_pino_write(inode, pino); 244 file_got_pino(inode); 245 } 246 f2fs_up_write(&fi->i_sem); 247 } 248 249 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 250 int datasync, bool atomic) 251 { 252 struct inode *inode = file->f_mapping->host; 253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 254 nid_t ino = inode->i_ino; 255 int ret = 0; 256 enum cp_reason_type cp_reason = 0; 257 struct writeback_control wbc = { 258 .sync_mode = WB_SYNC_ALL, 259 .nr_to_write = LONG_MAX, 260 .for_reclaim = 0, 261 }; 262 unsigned int seq_id = 0; 263 264 if (unlikely(f2fs_readonly(inode->i_sb))) 265 return 0; 266 267 trace_f2fs_sync_file_enter(inode); 268 269 if (S_ISDIR(inode->i_mode)) 270 goto go_write; 271 272 /* if fdatasync is triggered, let's do in-place-update */ 273 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 274 set_inode_flag(inode, FI_NEED_IPU); 275 ret = file_write_and_wait_range(file, start, end); 276 clear_inode_flag(inode, FI_NEED_IPU); 277 278 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 279 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 280 return ret; 281 } 282 283 /* if the inode is dirty, let's recover all the time */ 284 if (!f2fs_skip_inode_update(inode, datasync)) { 285 f2fs_write_inode(inode, NULL); 286 goto go_write; 287 } 288 289 /* 290 * if there is no written data, don't waste time to write recovery info. 291 */ 292 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 293 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 294 295 /* it may call write_inode just prior to fsync */ 296 if (need_inode_page_update(sbi, ino)) 297 goto go_write; 298 299 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 300 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 301 goto flush_out; 302 goto out; 303 } else { 304 /* 305 * for OPU case, during fsync(), node can be persisted before 306 * data when lower device doesn't support write barrier, result 307 * in data corruption after SPO. 308 * So for strict fsync mode, force to use atomic write sematics 309 * to keep write order in between data/node and last node to 310 * avoid potential data corruption. 311 */ 312 if (F2FS_OPTION(sbi).fsync_mode == 313 FSYNC_MODE_STRICT && !atomic) 314 atomic = true; 315 } 316 go_write: 317 /* 318 * Both of fdatasync() and fsync() are able to be recovered from 319 * sudden-power-off. 320 */ 321 f2fs_down_read(&F2FS_I(inode)->i_sem); 322 cp_reason = need_do_checkpoint(inode); 323 f2fs_up_read(&F2FS_I(inode)->i_sem); 324 325 if (cp_reason) { 326 /* all the dirty node pages should be flushed for POR */ 327 ret = f2fs_sync_fs(inode->i_sb, 1); 328 329 /* 330 * We've secured consistency through sync_fs. Following pino 331 * will be used only for fsynced inodes after checkpoint. 332 */ 333 try_to_fix_pino(inode); 334 clear_inode_flag(inode, FI_APPEND_WRITE); 335 clear_inode_flag(inode, FI_UPDATE_WRITE); 336 goto out; 337 } 338 sync_nodes: 339 atomic_inc(&sbi->wb_sync_req[NODE]); 340 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 341 atomic_dec(&sbi->wb_sync_req[NODE]); 342 if (ret) 343 goto out; 344 345 /* if cp_error was enabled, we should avoid infinite loop */ 346 if (unlikely(f2fs_cp_error(sbi))) { 347 ret = -EIO; 348 goto out; 349 } 350 351 if (f2fs_need_inode_block_update(sbi, ino)) { 352 f2fs_mark_inode_dirty_sync(inode, true); 353 f2fs_write_inode(inode, NULL); 354 goto sync_nodes; 355 } 356 357 /* 358 * If it's atomic_write, it's just fine to keep write ordering. So 359 * here we don't need to wait for node write completion, since we use 360 * node chain which serializes node blocks. If one of node writes are 361 * reordered, we can see simply broken chain, resulting in stopping 362 * roll-forward recovery. It means we'll recover all or none node blocks 363 * given fsync mark. 364 */ 365 if (!atomic) { 366 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 367 if (ret) 368 goto out; 369 } 370 371 /* once recovery info is written, don't need to tack this */ 372 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 373 clear_inode_flag(inode, FI_APPEND_WRITE); 374 flush_out: 375 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 376 ret = f2fs_issue_flush(sbi, inode->i_ino); 377 if (!ret) { 378 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 379 clear_inode_flag(inode, FI_UPDATE_WRITE); 380 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 381 } 382 f2fs_update_time(sbi, REQ_TIME); 383 out: 384 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 385 return ret; 386 } 387 388 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 389 { 390 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 391 return -EIO; 392 return f2fs_do_sync_file(file, start, end, datasync, false); 393 } 394 395 static bool __found_offset(struct address_space *mapping, block_t blkaddr, 396 pgoff_t index, int whence) 397 { 398 switch (whence) { 399 case SEEK_DATA: 400 if (__is_valid_data_blkaddr(blkaddr)) 401 return true; 402 if (blkaddr == NEW_ADDR && 403 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 404 return true; 405 break; 406 case SEEK_HOLE: 407 if (blkaddr == NULL_ADDR) 408 return true; 409 break; 410 } 411 return false; 412 } 413 414 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 415 { 416 struct inode *inode = file->f_mapping->host; 417 loff_t maxbytes = inode->i_sb->s_maxbytes; 418 struct dnode_of_data dn; 419 pgoff_t pgofs, end_offset; 420 loff_t data_ofs = offset; 421 loff_t isize; 422 int err = 0; 423 424 inode_lock(inode); 425 426 isize = i_size_read(inode); 427 if (offset >= isize) 428 goto fail; 429 430 /* handle inline data case */ 431 if (f2fs_has_inline_data(inode)) { 432 if (whence == SEEK_HOLE) { 433 data_ofs = isize; 434 goto found; 435 } else if (whence == SEEK_DATA) { 436 data_ofs = offset; 437 goto found; 438 } 439 } 440 441 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 442 443 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 444 set_new_dnode(&dn, inode, NULL, NULL, 0); 445 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 446 if (err && err != -ENOENT) { 447 goto fail; 448 } else if (err == -ENOENT) { 449 /* direct node does not exists */ 450 if (whence == SEEK_DATA) { 451 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 452 continue; 453 } else { 454 goto found; 455 } 456 } 457 458 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 459 460 /* find data/hole in dnode block */ 461 for (; dn.ofs_in_node < end_offset; 462 dn.ofs_in_node++, pgofs++, 463 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 464 block_t blkaddr; 465 466 blkaddr = f2fs_data_blkaddr(&dn); 467 468 if (__is_valid_data_blkaddr(blkaddr) && 469 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 470 blkaddr, DATA_GENERIC_ENHANCE)) { 471 f2fs_put_dnode(&dn); 472 goto fail; 473 } 474 475 if (__found_offset(file->f_mapping, blkaddr, 476 pgofs, whence)) { 477 f2fs_put_dnode(&dn); 478 goto found; 479 } 480 } 481 f2fs_put_dnode(&dn); 482 } 483 484 if (whence == SEEK_DATA) 485 goto fail; 486 found: 487 if (whence == SEEK_HOLE && data_ofs > isize) 488 data_ofs = isize; 489 inode_unlock(inode); 490 return vfs_setpos(file, data_ofs, maxbytes); 491 fail: 492 inode_unlock(inode); 493 return -ENXIO; 494 } 495 496 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 497 { 498 struct inode *inode = file->f_mapping->host; 499 loff_t maxbytes = inode->i_sb->s_maxbytes; 500 501 if (f2fs_compressed_file(inode)) 502 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; 503 504 switch (whence) { 505 case SEEK_SET: 506 case SEEK_CUR: 507 case SEEK_END: 508 return generic_file_llseek_size(file, offset, whence, 509 maxbytes, i_size_read(inode)); 510 case SEEK_DATA: 511 case SEEK_HOLE: 512 if (offset < 0) 513 return -ENXIO; 514 return f2fs_seek_block(file, offset, whence); 515 } 516 517 return -EINVAL; 518 } 519 520 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 521 { 522 struct inode *inode = file_inode(file); 523 524 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 525 return -EIO; 526 527 if (!f2fs_is_compress_backend_ready(inode)) 528 return -EOPNOTSUPP; 529 530 file_accessed(file); 531 vma->vm_ops = &f2fs_file_vm_ops; 532 set_inode_flag(inode, FI_MMAP_FILE); 533 return 0; 534 } 535 536 static int f2fs_file_open(struct inode *inode, struct file *filp) 537 { 538 int err = fscrypt_file_open(inode, filp); 539 540 if (err) 541 return err; 542 543 if (!f2fs_is_compress_backend_ready(inode)) 544 return -EOPNOTSUPP; 545 546 err = fsverity_file_open(inode, filp); 547 if (err) 548 return err; 549 550 filp->f_mode |= FMODE_NOWAIT; 551 552 return dquot_file_open(inode, filp); 553 } 554 555 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 556 { 557 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 558 struct f2fs_node *raw_node; 559 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 560 __le32 *addr; 561 int base = 0; 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 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode)) 568 base = get_extra_isize(dn->inode); 569 570 raw_node = F2FS_NODE(dn->node_page); 571 addr = blkaddr_in_node(raw_node) + base + ofs; 572 573 /* Assumption: truncateion starts with cluster */ 574 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 575 block_t blkaddr = le32_to_cpu(*addr); 576 577 if (f2fs_compressed_file(dn->inode) && 578 !(cluster_index & (cluster_size - 1))) { 579 if (compressed_cluster) 580 f2fs_i_compr_blocks_update(dn->inode, 581 valid_blocks, false); 582 compressed_cluster = (blkaddr == COMPRESS_ADDR); 583 valid_blocks = 0; 584 } 585 586 if (blkaddr == NULL_ADDR) 587 continue; 588 589 dn->data_blkaddr = NULL_ADDR; 590 f2fs_set_data_blkaddr(dn); 591 592 if (__is_valid_data_blkaddr(blkaddr)) { 593 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 594 DATA_GENERIC_ENHANCE)) 595 continue; 596 if (compressed_cluster) 597 valid_blocks++; 598 } 599 600 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 601 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 602 603 f2fs_invalidate_blocks(sbi, blkaddr); 604 605 if (!released || blkaddr != COMPRESS_ADDR) 606 nr_free++; 607 } 608 609 if (compressed_cluster) 610 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 611 612 if (nr_free) { 613 pgoff_t fofs; 614 /* 615 * once we invalidate valid blkaddr in range [ofs, ofs + count], 616 * we will invalidate all blkaddr in the whole range. 617 */ 618 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 619 dn->inode) + ofs; 620 f2fs_update_extent_cache_range(dn, fofs, 0, len); 621 dec_valid_block_count(sbi, dn->inode, nr_free); 622 } 623 dn->ofs_in_node = ofs; 624 625 f2fs_update_time(sbi, REQ_TIME); 626 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 627 dn->ofs_in_node, nr_free); 628 } 629 630 void f2fs_truncate_data_blocks(struct dnode_of_data *dn) 631 { 632 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode)); 633 } 634 635 static int truncate_partial_data_page(struct inode *inode, u64 from, 636 bool cache_only) 637 { 638 loff_t offset = from & (PAGE_SIZE - 1); 639 pgoff_t index = from >> PAGE_SHIFT; 640 struct address_space *mapping = inode->i_mapping; 641 struct page *page; 642 643 if (!offset && !cache_only) 644 return 0; 645 646 if (cache_only) { 647 page = find_lock_page(mapping, index); 648 if (page && PageUptodate(page)) 649 goto truncate_out; 650 f2fs_put_page(page, 1); 651 return 0; 652 } 653 654 page = f2fs_get_lock_data_page(inode, index, true); 655 if (IS_ERR(page)) 656 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 657 truncate_out: 658 f2fs_wait_on_page_writeback(page, DATA, true, true); 659 zero_user(page, offset, PAGE_SIZE - offset); 660 661 /* An encrypted inode should have a key and truncate the last page. */ 662 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 663 if (!cache_only) 664 set_page_dirty(page); 665 f2fs_put_page(page, 1); 666 return 0; 667 } 668 669 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 670 { 671 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 672 struct dnode_of_data dn; 673 pgoff_t free_from; 674 int count = 0, err = 0; 675 struct page *ipage; 676 bool truncate_page = false; 677 678 trace_f2fs_truncate_blocks_enter(inode, from); 679 680 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 681 682 if (free_from >= max_file_blocks(inode)) 683 goto free_partial; 684 685 if (lock) 686 f2fs_lock_op(sbi); 687 688 ipage = f2fs_get_node_page(sbi, inode->i_ino); 689 if (IS_ERR(ipage)) { 690 err = PTR_ERR(ipage); 691 goto out; 692 } 693 694 if (f2fs_has_inline_data(inode)) { 695 f2fs_truncate_inline_inode(inode, ipage, from); 696 f2fs_put_page(ipage, 1); 697 truncate_page = true; 698 goto out; 699 } 700 701 set_new_dnode(&dn, inode, ipage, NULL, 0); 702 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 703 if (err) { 704 if (err == -ENOENT) 705 goto free_next; 706 goto out; 707 } 708 709 count = ADDRS_PER_PAGE(dn.node_page, inode); 710 711 count -= dn.ofs_in_node; 712 f2fs_bug_on(sbi, count < 0); 713 714 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 715 f2fs_truncate_data_blocks_range(&dn, count); 716 free_from += count; 717 } 718 719 f2fs_put_dnode(&dn); 720 free_next: 721 err = f2fs_truncate_inode_blocks(inode, free_from); 722 out: 723 if (lock) 724 f2fs_unlock_op(sbi); 725 free_partial: 726 /* lastly zero out the first data page */ 727 if (!err) 728 err = truncate_partial_data_page(inode, from, truncate_page); 729 730 trace_f2fs_truncate_blocks_exit(inode, err); 731 return err; 732 } 733 734 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 735 { 736 u64 free_from = from; 737 int err; 738 739 #ifdef CONFIG_F2FS_FS_COMPRESSION 740 /* 741 * for compressed file, only support cluster size 742 * aligned truncation. 743 */ 744 if (f2fs_compressed_file(inode)) 745 free_from = round_up(from, 746 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 747 #endif 748 749 err = f2fs_do_truncate_blocks(inode, free_from, lock); 750 if (err) 751 return err; 752 753 #ifdef CONFIG_F2FS_FS_COMPRESSION 754 /* 755 * For compressed file, after release compress blocks, don't allow write 756 * direct, but we should allow write direct after truncate to zero. 757 */ 758 if (f2fs_compressed_file(inode) && !free_from 759 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 760 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 761 762 if (from != free_from) { 763 err = f2fs_truncate_partial_cluster(inode, from, lock); 764 if (err) 765 return err; 766 } 767 #endif 768 769 return 0; 770 } 771 772 int f2fs_truncate(struct inode *inode) 773 { 774 int err; 775 776 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 777 return -EIO; 778 779 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 780 S_ISLNK(inode->i_mode))) 781 return 0; 782 783 trace_f2fs_truncate(inode); 784 785 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) { 786 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE); 787 return -EIO; 788 } 789 790 err = f2fs_dquot_initialize(inode); 791 if (err) 792 return err; 793 794 /* we should check inline_data size */ 795 if (!f2fs_may_inline_data(inode)) { 796 err = f2fs_convert_inline_inode(inode); 797 if (err) 798 return err; 799 } 800 801 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 802 if (err) 803 return err; 804 805 inode->i_mtime = inode->i_ctime = current_time(inode); 806 f2fs_mark_inode_dirty_sync(inode, false); 807 return 0; 808 } 809 810 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path, 811 struct kstat *stat, u32 request_mask, unsigned int query_flags) 812 { 813 struct inode *inode = d_inode(path->dentry); 814 struct f2fs_inode_info *fi = F2FS_I(inode); 815 struct f2fs_inode *ri = NULL; 816 unsigned int flags; 817 818 if (f2fs_has_extra_attr(inode) && 819 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 820 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 821 stat->result_mask |= STATX_BTIME; 822 stat->btime.tv_sec = fi->i_crtime.tv_sec; 823 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 824 } 825 826 flags = fi->i_flags; 827 if (flags & F2FS_COMPR_FL) 828 stat->attributes |= STATX_ATTR_COMPRESSED; 829 if (flags & F2FS_APPEND_FL) 830 stat->attributes |= STATX_ATTR_APPEND; 831 if (IS_ENCRYPTED(inode)) 832 stat->attributes |= STATX_ATTR_ENCRYPTED; 833 if (flags & F2FS_IMMUTABLE_FL) 834 stat->attributes |= STATX_ATTR_IMMUTABLE; 835 if (flags & F2FS_NODUMP_FL) 836 stat->attributes |= STATX_ATTR_NODUMP; 837 if (IS_VERITY(inode)) 838 stat->attributes |= STATX_ATTR_VERITY; 839 840 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 841 STATX_ATTR_APPEND | 842 STATX_ATTR_ENCRYPTED | 843 STATX_ATTR_IMMUTABLE | 844 STATX_ATTR_NODUMP | 845 STATX_ATTR_VERITY); 846 847 generic_fillattr(mnt_userns, inode, stat); 848 849 /* we need to show initial sectors used for inline_data/dentries */ 850 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 851 f2fs_has_inline_dentry(inode)) 852 stat->blocks += (stat->size + 511) >> 9; 853 854 return 0; 855 } 856 857 #ifdef CONFIG_F2FS_FS_POSIX_ACL 858 static void __setattr_copy(struct user_namespace *mnt_userns, 859 struct inode *inode, const struct iattr *attr) 860 { 861 unsigned int ia_valid = attr->ia_valid; 862 863 if (ia_valid & ATTR_UID) 864 inode->i_uid = attr->ia_uid; 865 if (ia_valid & ATTR_GID) 866 inode->i_gid = attr->ia_gid; 867 if (ia_valid & ATTR_ATIME) 868 inode->i_atime = attr->ia_atime; 869 if (ia_valid & ATTR_MTIME) 870 inode->i_mtime = attr->ia_mtime; 871 if (ia_valid & ATTR_CTIME) 872 inode->i_ctime = attr->ia_ctime; 873 if (ia_valid & ATTR_MODE) { 874 umode_t mode = attr->ia_mode; 875 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode); 876 877 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID)) 878 mode &= ~S_ISGID; 879 set_acl_inode(inode, mode); 880 } 881 } 882 #else 883 #define __setattr_copy setattr_copy 884 #endif 885 886 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry, 887 struct iattr *attr) 888 { 889 struct inode *inode = d_inode(dentry); 890 int err; 891 892 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 893 return -EIO; 894 895 if (unlikely(IS_IMMUTABLE(inode))) 896 return -EPERM; 897 898 if (unlikely(IS_APPEND(inode) && 899 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 900 ATTR_GID | ATTR_TIMES_SET)))) 901 return -EPERM; 902 903 if ((attr->ia_valid & ATTR_SIZE) && 904 !f2fs_is_compress_backend_ready(inode)) 905 return -EOPNOTSUPP; 906 907 err = setattr_prepare(mnt_userns, dentry, attr); 908 if (err) 909 return err; 910 911 err = fscrypt_prepare_setattr(dentry, attr); 912 if (err) 913 return err; 914 915 err = fsverity_prepare_setattr(dentry, attr); 916 if (err) 917 return err; 918 919 if (is_quota_modification(inode, attr)) { 920 err = f2fs_dquot_initialize(inode); 921 if (err) 922 return err; 923 } 924 if ((attr->ia_valid & ATTR_UID && 925 !uid_eq(attr->ia_uid, inode->i_uid)) || 926 (attr->ia_valid & ATTR_GID && 927 !gid_eq(attr->ia_gid, inode->i_gid))) { 928 f2fs_lock_op(F2FS_I_SB(inode)); 929 err = dquot_transfer(inode, attr); 930 if (err) { 931 set_sbi_flag(F2FS_I_SB(inode), 932 SBI_QUOTA_NEED_REPAIR); 933 f2fs_unlock_op(F2FS_I_SB(inode)); 934 return err; 935 } 936 /* 937 * update uid/gid under lock_op(), so that dquot and inode can 938 * be updated atomically. 939 */ 940 if (attr->ia_valid & ATTR_UID) 941 inode->i_uid = attr->ia_uid; 942 if (attr->ia_valid & ATTR_GID) 943 inode->i_gid = attr->ia_gid; 944 f2fs_mark_inode_dirty_sync(inode, true); 945 f2fs_unlock_op(F2FS_I_SB(inode)); 946 } 947 948 if (attr->ia_valid & ATTR_SIZE) { 949 loff_t old_size = i_size_read(inode); 950 951 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 952 /* 953 * should convert inline inode before i_size_write to 954 * keep smaller than inline_data size with inline flag. 955 */ 956 err = f2fs_convert_inline_inode(inode); 957 if (err) 958 return err; 959 } 960 961 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 962 filemap_invalidate_lock(inode->i_mapping); 963 964 truncate_setsize(inode, attr->ia_size); 965 966 if (attr->ia_size <= old_size) 967 err = f2fs_truncate(inode); 968 /* 969 * do not trim all blocks after i_size if target size is 970 * larger than i_size. 971 */ 972 filemap_invalidate_unlock(inode->i_mapping); 973 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 974 if (err) 975 return err; 976 977 spin_lock(&F2FS_I(inode)->i_size_lock); 978 inode->i_mtime = inode->i_ctime = current_time(inode); 979 F2FS_I(inode)->last_disk_size = i_size_read(inode); 980 spin_unlock(&F2FS_I(inode)->i_size_lock); 981 } 982 983 __setattr_copy(mnt_userns, inode, attr); 984 985 if (attr->ia_valid & ATTR_MODE) { 986 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode)); 987 988 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 989 if (!err) 990 inode->i_mode = F2FS_I(inode)->i_acl_mode; 991 clear_inode_flag(inode, FI_ACL_MODE); 992 } 993 } 994 995 /* file size may changed here */ 996 f2fs_mark_inode_dirty_sync(inode, true); 997 998 /* inode change will produce dirty node pages flushed by checkpoint */ 999 f2fs_balance_fs(F2FS_I_SB(inode), true); 1000 1001 return err; 1002 } 1003 1004 const struct inode_operations f2fs_file_inode_operations = { 1005 .getattr = f2fs_getattr, 1006 .setattr = f2fs_setattr, 1007 .get_acl = f2fs_get_acl, 1008 .set_acl = f2fs_set_acl, 1009 .listxattr = f2fs_listxattr, 1010 .fiemap = f2fs_fiemap, 1011 .fileattr_get = f2fs_fileattr_get, 1012 .fileattr_set = f2fs_fileattr_set, 1013 }; 1014 1015 static int fill_zero(struct inode *inode, pgoff_t index, 1016 loff_t start, loff_t len) 1017 { 1018 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1019 struct page *page; 1020 1021 if (!len) 1022 return 0; 1023 1024 f2fs_balance_fs(sbi, true); 1025 1026 f2fs_lock_op(sbi); 1027 page = f2fs_get_new_data_page(inode, NULL, index, false); 1028 f2fs_unlock_op(sbi); 1029 1030 if (IS_ERR(page)) 1031 return PTR_ERR(page); 1032 1033 f2fs_wait_on_page_writeback(page, DATA, true, true); 1034 zero_user(page, start, len); 1035 set_page_dirty(page); 1036 f2fs_put_page(page, 1); 1037 return 0; 1038 } 1039 1040 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1041 { 1042 int err; 1043 1044 while (pg_start < pg_end) { 1045 struct dnode_of_data dn; 1046 pgoff_t end_offset, count; 1047 1048 set_new_dnode(&dn, inode, NULL, NULL, 0); 1049 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1050 if (err) { 1051 if (err == -ENOENT) { 1052 pg_start = f2fs_get_next_page_offset(&dn, 1053 pg_start); 1054 continue; 1055 } 1056 return err; 1057 } 1058 1059 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1060 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1061 1062 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1063 1064 f2fs_truncate_data_blocks_range(&dn, count); 1065 f2fs_put_dnode(&dn); 1066 1067 pg_start += count; 1068 } 1069 return 0; 1070 } 1071 1072 static int punch_hole(struct inode *inode, loff_t offset, loff_t len) 1073 { 1074 pgoff_t pg_start, pg_end; 1075 loff_t off_start, off_end; 1076 int ret; 1077 1078 ret = f2fs_convert_inline_inode(inode); 1079 if (ret) 1080 return ret; 1081 1082 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1083 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1084 1085 off_start = offset & (PAGE_SIZE - 1); 1086 off_end = (offset + len) & (PAGE_SIZE - 1); 1087 1088 if (pg_start == pg_end) { 1089 ret = fill_zero(inode, pg_start, off_start, 1090 off_end - off_start); 1091 if (ret) 1092 return ret; 1093 } else { 1094 if (off_start) { 1095 ret = fill_zero(inode, pg_start++, off_start, 1096 PAGE_SIZE - off_start); 1097 if (ret) 1098 return ret; 1099 } 1100 if (off_end) { 1101 ret = fill_zero(inode, pg_end, 0, off_end); 1102 if (ret) 1103 return ret; 1104 } 1105 1106 if (pg_start < pg_end) { 1107 loff_t blk_start, blk_end; 1108 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1109 1110 f2fs_balance_fs(sbi, true); 1111 1112 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1113 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1114 1115 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1116 filemap_invalidate_lock(inode->i_mapping); 1117 1118 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1119 1120 f2fs_lock_op(sbi); 1121 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1122 f2fs_unlock_op(sbi); 1123 1124 filemap_invalidate_unlock(inode->i_mapping); 1125 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1126 } 1127 } 1128 1129 return ret; 1130 } 1131 1132 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1133 int *do_replace, pgoff_t off, pgoff_t len) 1134 { 1135 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1136 struct dnode_of_data dn; 1137 int ret, done, i; 1138 1139 next_dnode: 1140 set_new_dnode(&dn, inode, NULL, NULL, 0); 1141 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1142 if (ret && ret != -ENOENT) { 1143 return ret; 1144 } else if (ret == -ENOENT) { 1145 if (dn.max_level == 0) 1146 return -ENOENT; 1147 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1148 dn.ofs_in_node, len); 1149 blkaddr += done; 1150 do_replace += done; 1151 goto next; 1152 } 1153 1154 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1155 dn.ofs_in_node, len); 1156 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1157 *blkaddr = f2fs_data_blkaddr(&dn); 1158 1159 if (__is_valid_data_blkaddr(*blkaddr) && 1160 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1161 DATA_GENERIC_ENHANCE)) { 1162 f2fs_put_dnode(&dn); 1163 return -EFSCORRUPTED; 1164 } 1165 1166 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1167 1168 if (f2fs_lfs_mode(sbi)) { 1169 f2fs_put_dnode(&dn); 1170 return -EOPNOTSUPP; 1171 } 1172 1173 /* do not invalidate this block address */ 1174 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1175 *do_replace = 1; 1176 } 1177 } 1178 f2fs_put_dnode(&dn); 1179 next: 1180 len -= done; 1181 off += done; 1182 if (len) 1183 goto next_dnode; 1184 return 0; 1185 } 1186 1187 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1188 int *do_replace, pgoff_t off, int len) 1189 { 1190 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1191 struct dnode_of_data dn; 1192 int ret, i; 1193 1194 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1195 if (*do_replace == 0) 1196 continue; 1197 1198 set_new_dnode(&dn, inode, NULL, NULL, 0); 1199 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1200 if (ret) { 1201 dec_valid_block_count(sbi, inode, 1); 1202 f2fs_invalidate_blocks(sbi, *blkaddr); 1203 } else { 1204 f2fs_update_data_blkaddr(&dn, *blkaddr); 1205 } 1206 f2fs_put_dnode(&dn); 1207 } 1208 return 0; 1209 } 1210 1211 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1212 block_t *blkaddr, int *do_replace, 1213 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1214 { 1215 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1216 pgoff_t i = 0; 1217 int ret; 1218 1219 while (i < len) { 1220 if (blkaddr[i] == NULL_ADDR && !full) { 1221 i++; 1222 continue; 1223 } 1224 1225 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1226 struct dnode_of_data dn; 1227 struct node_info ni; 1228 size_t new_size; 1229 pgoff_t ilen; 1230 1231 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1232 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1233 if (ret) 1234 return ret; 1235 1236 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1237 if (ret) { 1238 f2fs_put_dnode(&dn); 1239 return ret; 1240 } 1241 1242 ilen = min((pgoff_t) 1243 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1244 dn.ofs_in_node, len - i); 1245 do { 1246 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1247 f2fs_truncate_data_blocks_range(&dn, 1); 1248 1249 if (do_replace[i]) { 1250 f2fs_i_blocks_write(src_inode, 1251 1, false, false); 1252 f2fs_i_blocks_write(dst_inode, 1253 1, true, false); 1254 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1255 blkaddr[i], ni.version, true, false); 1256 1257 do_replace[i] = 0; 1258 } 1259 dn.ofs_in_node++; 1260 i++; 1261 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1262 if (dst_inode->i_size < new_size) 1263 f2fs_i_size_write(dst_inode, new_size); 1264 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1265 1266 f2fs_put_dnode(&dn); 1267 } else { 1268 struct page *psrc, *pdst; 1269 1270 psrc = f2fs_get_lock_data_page(src_inode, 1271 src + i, true); 1272 if (IS_ERR(psrc)) 1273 return PTR_ERR(psrc); 1274 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1275 true); 1276 if (IS_ERR(pdst)) { 1277 f2fs_put_page(psrc, 1); 1278 return PTR_ERR(pdst); 1279 } 1280 f2fs_copy_page(psrc, pdst); 1281 set_page_dirty(pdst); 1282 f2fs_put_page(pdst, 1); 1283 f2fs_put_page(psrc, 1); 1284 1285 ret = f2fs_truncate_hole(src_inode, 1286 src + i, src + i + 1); 1287 if (ret) 1288 return ret; 1289 i++; 1290 } 1291 } 1292 return 0; 1293 } 1294 1295 static int __exchange_data_block(struct inode *src_inode, 1296 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1297 pgoff_t len, bool full) 1298 { 1299 block_t *src_blkaddr; 1300 int *do_replace; 1301 pgoff_t olen; 1302 int ret; 1303 1304 while (len) { 1305 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1306 1307 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1308 array_size(olen, sizeof(block_t)), 1309 GFP_NOFS); 1310 if (!src_blkaddr) 1311 return -ENOMEM; 1312 1313 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1314 array_size(olen, sizeof(int)), 1315 GFP_NOFS); 1316 if (!do_replace) { 1317 kvfree(src_blkaddr); 1318 return -ENOMEM; 1319 } 1320 1321 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1322 do_replace, src, olen); 1323 if (ret) 1324 goto roll_back; 1325 1326 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1327 do_replace, src, dst, olen, full); 1328 if (ret) 1329 goto roll_back; 1330 1331 src += olen; 1332 dst += olen; 1333 len -= olen; 1334 1335 kvfree(src_blkaddr); 1336 kvfree(do_replace); 1337 } 1338 return 0; 1339 1340 roll_back: 1341 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1342 kvfree(src_blkaddr); 1343 kvfree(do_replace); 1344 return ret; 1345 } 1346 1347 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1348 { 1349 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1350 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1351 pgoff_t start = offset >> PAGE_SHIFT; 1352 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1353 int ret; 1354 1355 f2fs_balance_fs(sbi, true); 1356 1357 /* avoid gc operation during block exchange */ 1358 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1359 filemap_invalidate_lock(inode->i_mapping); 1360 1361 f2fs_lock_op(sbi); 1362 f2fs_drop_extent_tree(inode); 1363 truncate_pagecache(inode, offset); 1364 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1365 f2fs_unlock_op(sbi); 1366 1367 filemap_invalidate_unlock(inode->i_mapping); 1368 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1369 return ret; 1370 } 1371 1372 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1373 { 1374 loff_t new_size; 1375 int ret; 1376 1377 if (offset + len >= i_size_read(inode)) 1378 return -EINVAL; 1379 1380 /* collapse range should be aligned to block size of f2fs. */ 1381 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1382 return -EINVAL; 1383 1384 ret = f2fs_convert_inline_inode(inode); 1385 if (ret) 1386 return ret; 1387 1388 /* write out all dirty pages from offset */ 1389 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1390 if (ret) 1391 return ret; 1392 1393 ret = f2fs_do_collapse(inode, offset, len); 1394 if (ret) 1395 return ret; 1396 1397 /* write out all moved pages, if possible */ 1398 filemap_invalidate_lock(inode->i_mapping); 1399 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1400 truncate_pagecache(inode, offset); 1401 1402 new_size = i_size_read(inode) - len; 1403 ret = f2fs_truncate_blocks(inode, new_size, true); 1404 filemap_invalidate_unlock(inode->i_mapping); 1405 if (!ret) 1406 f2fs_i_size_write(inode, new_size); 1407 return ret; 1408 } 1409 1410 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1411 pgoff_t end) 1412 { 1413 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1414 pgoff_t index = start; 1415 unsigned int ofs_in_node = dn->ofs_in_node; 1416 blkcnt_t count = 0; 1417 int ret; 1418 1419 for (; index < end; index++, dn->ofs_in_node++) { 1420 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1421 count++; 1422 } 1423 1424 dn->ofs_in_node = ofs_in_node; 1425 ret = f2fs_reserve_new_blocks(dn, count); 1426 if (ret) 1427 return ret; 1428 1429 dn->ofs_in_node = ofs_in_node; 1430 for (index = start; index < end; index++, dn->ofs_in_node++) { 1431 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1432 /* 1433 * f2fs_reserve_new_blocks will not guarantee entire block 1434 * allocation. 1435 */ 1436 if (dn->data_blkaddr == NULL_ADDR) { 1437 ret = -ENOSPC; 1438 break; 1439 } 1440 if (dn->data_blkaddr != NEW_ADDR) { 1441 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1442 dn->data_blkaddr = NEW_ADDR; 1443 f2fs_set_data_blkaddr(dn); 1444 } 1445 } 1446 1447 f2fs_update_extent_cache_range(dn, start, 0, index - start); 1448 1449 return ret; 1450 } 1451 1452 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1453 int mode) 1454 { 1455 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1456 struct address_space *mapping = inode->i_mapping; 1457 pgoff_t index, pg_start, pg_end; 1458 loff_t new_size = i_size_read(inode); 1459 loff_t off_start, off_end; 1460 int ret = 0; 1461 1462 ret = inode_newsize_ok(inode, (len + offset)); 1463 if (ret) 1464 return ret; 1465 1466 ret = f2fs_convert_inline_inode(inode); 1467 if (ret) 1468 return ret; 1469 1470 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1471 if (ret) 1472 return ret; 1473 1474 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1475 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1476 1477 off_start = offset & (PAGE_SIZE - 1); 1478 off_end = (offset + len) & (PAGE_SIZE - 1); 1479 1480 if (pg_start == pg_end) { 1481 ret = fill_zero(inode, pg_start, off_start, 1482 off_end - off_start); 1483 if (ret) 1484 return ret; 1485 1486 new_size = max_t(loff_t, new_size, offset + len); 1487 } else { 1488 if (off_start) { 1489 ret = fill_zero(inode, pg_start++, off_start, 1490 PAGE_SIZE - off_start); 1491 if (ret) 1492 return ret; 1493 1494 new_size = max_t(loff_t, new_size, 1495 (loff_t)pg_start << PAGE_SHIFT); 1496 } 1497 1498 for (index = pg_start; index < pg_end;) { 1499 struct dnode_of_data dn; 1500 unsigned int end_offset; 1501 pgoff_t end; 1502 1503 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1504 filemap_invalidate_lock(mapping); 1505 1506 truncate_pagecache_range(inode, 1507 (loff_t)index << PAGE_SHIFT, 1508 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1509 1510 f2fs_lock_op(sbi); 1511 1512 set_new_dnode(&dn, inode, NULL, NULL, 0); 1513 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1514 if (ret) { 1515 f2fs_unlock_op(sbi); 1516 filemap_invalidate_unlock(mapping); 1517 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1518 goto out; 1519 } 1520 1521 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1522 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1523 1524 ret = f2fs_do_zero_range(&dn, index, end); 1525 f2fs_put_dnode(&dn); 1526 1527 f2fs_unlock_op(sbi); 1528 filemap_invalidate_unlock(mapping); 1529 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1530 1531 f2fs_balance_fs(sbi, dn.node_changed); 1532 1533 if (ret) 1534 goto out; 1535 1536 index = end; 1537 new_size = max_t(loff_t, new_size, 1538 (loff_t)index << PAGE_SHIFT); 1539 } 1540 1541 if (off_end) { 1542 ret = fill_zero(inode, pg_end, 0, off_end); 1543 if (ret) 1544 goto out; 1545 1546 new_size = max_t(loff_t, new_size, offset + len); 1547 } 1548 } 1549 1550 out: 1551 if (new_size > i_size_read(inode)) { 1552 if (mode & FALLOC_FL_KEEP_SIZE) 1553 file_set_keep_isize(inode); 1554 else 1555 f2fs_i_size_write(inode, new_size); 1556 } 1557 return ret; 1558 } 1559 1560 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1561 { 1562 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1563 struct address_space *mapping = inode->i_mapping; 1564 pgoff_t nr, pg_start, pg_end, delta, idx; 1565 loff_t new_size; 1566 int ret = 0; 1567 1568 new_size = i_size_read(inode) + len; 1569 ret = inode_newsize_ok(inode, new_size); 1570 if (ret) 1571 return ret; 1572 1573 if (offset >= i_size_read(inode)) 1574 return -EINVAL; 1575 1576 /* insert range should be aligned to block size of f2fs. */ 1577 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1578 return -EINVAL; 1579 1580 ret = f2fs_convert_inline_inode(inode); 1581 if (ret) 1582 return ret; 1583 1584 f2fs_balance_fs(sbi, true); 1585 1586 filemap_invalidate_lock(mapping); 1587 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1588 filemap_invalidate_unlock(mapping); 1589 if (ret) 1590 return ret; 1591 1592 /* write out all dirty pages from offset */ 1593 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1594 if (ret) 1595 return ret; 1596 1597 pg_start = offset >> PAGE_SHIFT; 1598 pg_end = (offset + len) >> PAGE_SHIFT; 1599 delta = pg_end - pg_start; 1600 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1601 1602 /* avoid gc operation during block exchange */ 1603 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1604 filemap_invalidate_lock(mapping); 1605 truncate_pagecache(inode, offset); 1606 1607 while (!ret && idx > pg_start) { 1608 nr = idx - pg_start; 1609 if (nr > delta) 1610 nr = delta; 1611 idx -= nr; 1612 1613 f2fs_lock_op(sbi); 1614 f2fs_drop_extent_tree(inode); 1615 1616 ret = __exchange_data_block(inode, inode, idx, 1617 idx + delta, nr, false); 1618 f2fs_unlock_op(sbi); 1619 } 1620 filemap_invalidate_unlock(mapping); 1621 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1622 1623 /* write out all moved pages, if possible */ 1624 filemap_invalidate_lock(mapping); 1625 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1626 truncate_pagecache(inode, offset); 1627 filemap_invalidate_unlock(mapping); 1628 1629 if (!ret) 1630 f2fs_i_size_write(inode, new_size); 1631 return ret; 1632 } 1633 1634 static int expand_inode_data(struct inode *inode, loff_t offset, 1635 loff_t len, int mode) 1636 { 1637 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1638 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1639 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1640 .m_may_create = true }; 1641 pgoff_t pg_start, pg_end; 1642 loff_t new_size = i_size_read(inode); 1643 loff_t off_end; 1644 block_t expanded = 0; 1645 int err; 1646 1647 err = inode_newsize_ok(inode, (len + offset)); 1648 if (err) 1649 return err; 1650 1651 err = f2fs_convert_inline_inode(inode); 1652 if (err) 1653 return err; 1654 1655 f2fs_balance_fs(sbi, true); 1656 1657 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1658 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1659 off_end = (offset + len) & (PAGE_SIZE - 1); 1660 1661 map.m_lblk = pg_start; 1662 map.m_len = pg_end - pg_start; 1663 if (off_end) 1664 map.m_len++; 1665 1666 if (!map.m_len) 1667 return 0; 1668 1669 if (f2fs_is_pinned_file(inode)) { 1670 block_t sec_blks = BLKS_PER_SEC(sbi); 1671 block_t sec_len = roundup(map.m_len, sec_blks); 1672 1673 map.m_len = sec_blks; 1674 next_alloc: 1675 if (has_not_enough_free_secs(sbi, 0, 1676 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1677 f2fs_down_write(&sbi->gc_lock); 1678 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO); 1679 if (err && err != -ENODATA && err != -EAGAIN) 1680 goto out_err; 1681 } 1682 1683 f2fs_down_write(&sbi->pin_sem); 1684 1685 f2fs_lock_op(sbi); 1686 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false); 1687 f2fs_unlock_op(sbi); 1688 1689 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1690 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO); 1691 file_dont_truncate(inode); 1692 1693 f2fs_up_write(&sbi->pin_sem); 1694 1695 expanded += map.m_len; 1696 sec_len -= map.m_len; 1697 map.m_lblk += map.m_len; 1698 if (!err && sec_len) 1699 goto next_alloc; 1700 1701 map.m_len = expanded; 1702 } else { 1703 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); 1704 expanded = map.m_len; 1705 } 1706 out_err: 1707 if (err) { 1708 pgoff_t last_off; 1709 1710 if (!expanded) 1711 return err; 1712 1713 last_off = pg_start + expanded - 1; 1714 1715 /* update new size to the failed position */ 1716 new_size = (last_off == pg_end) ? offset + len : 1717 (loff_t)(last_off + 1) << PAGE_SHIFT; 1718 } else { 1719 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1720 } 1721 1722 if (new_size > i_size_read(inode)) { 1723 if (mode & FALLOC_FL_KEEP_SIZE) 1724 file_set_keep_isize(inode); 1725 else 1726 f2fs_i_size_write(inode, new_size); 1727 } 1728 1729 return err; 1730 } 1731 1732 static long f2fs_fallocate(struct file *file, int mode, 1733 loff_t offset, loff_t len) 1734 { 1735 struct inode *inode = file_inode(file); 1736 long ret = 0; 1737 1738 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1739 return -EIO; 1740 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1741 return -ENOSPC; 1742 if (!f2fs_is_compress_backend_ready(inode)) 1743 return -EOPNOTSUPP; 1744 1745 /* f2fs only support ->fallocate for regular file */ 1746 if (!S_ISREG(inode->i_mode)) 1747 return -EINVAL; 1748 1749 if (IS_ENCRYPTED(inode) && 1750 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1751 return -EOPNOTSUPP; 1752 1753 /* 1754 * Pinned file should not support partial trucation since the block 1755 * can be used by applications. 1756 */ 1757 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1758 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1759 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1760 return -EOPNOTSUPP; 1761 1762 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1763 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1764 FALLOC_FL_INSERT_RANGE)) 1765 return -EOPNOTSUPP; 1766 1767 inode_lock(inode); 1768 1769 if (mode & FALLOC_FL_PUNCH_HOLE) { 1770 if (offset >= inode->i_size) 1771 goto out; 1772 1773 ret = punch_hole(inode, offset, len); 1774 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1775 ret = f2fs_collapse_range(inode, offset, len); 1776 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1777 ret = f2fs_zero_range(inode, offset, len, mode); 1778 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1779 ret = f2fs_insert_range(inode, offset, len); 1780 } else { 1781 ret = expand_inode_data(inode, offset, len, mode); 1782 } 1783 1784 if (!ret) { 1785 inode->i_mtime = inode->i_ctime = current_time(inode); 1786 f2fs_mark_inode_dirty_sync(inode, false); 1787 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1788 } 1789 1790 out: 1791 inode_unlock(inode); 1792 1793 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1794 return ret; 1795 } 1796 1797 static int f2fs_release_file(struct inode *inode, struct file *filp) 1798 { 1799 /* 1800 * f2fs_relase_file is called at every close calls. So we should 1801 * not drop any inmemory pages by close called by other process. 1802 */ 1803 if (!(filp->f_mode & FMODE_WRITE) || 1804 atomic_read(&inode->i_writecount) != 1) 1805 return 0; 1806 1807 /* some remained atomic pages should discarded */ 1808 if (f2fs_is_atomic_file(inode)) 1809 f2fs_drop_inmem_pages(inode); 1810 if (f2fs_is_volatile_file(inode)) { 1811 set_inode_flag(inode, FI_DROP_CACHE); 1812 filemap_fdatawrite(inode->i_mapping); 1813 clear_inode_flag(inode, FI_DROP_CACHE); 1814 clear_inode_flag(inode, FI_VOLATILE_FILE); 1815 stat_dec_volatile_write(inode); 1816 } 1817 return 0; 1818 } 1819 1820 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1821 { 1822 struct inode *inode = file_inode(file); 1823 1824 /* 1825 * If the process doing a transaction is crashed, we should do 1826 * roll-back. Otherwise, other reader/write can see corrupted database 1827 * until all the writers close its file. Since this should be done 1828 * before dropping file lock, it needs to do in ->flush. 1829 */ 1830 if (f2fs_is_atomic_file(inode) && 1831 F2FS_I(inode)->inmem_task == current) 1832 f2fs_drop_inmem_pages(inode); 1833 return 0; 1834 } 1835 1836 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1837 { 1838 struct f2fs_inode_info *fi = F2FS_I(inode); 1839 u32 masked_flags = fi->i_flags & mask; 1840 1841 /* mask can be shrunk by flags_valid selector */ 1842 iflags &= mask; 1843 1844 /* Is it quota file? Do not allow user to mess with it */ 1845 if (IS_NOQUOTA(inode)) 1846 return -EPERM; 1847 1848 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1849 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1850 return -EOPNOTSUPP; 1851 if (!f2fs_empty_dir(inode)) 1852 return -ENOTEMPTY; 1853 } 1854 1855 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1856 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1857 return -EOPNOTSUPP; 1858 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1859 return -EINVAL; 1860 } 1861 1862 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1863 if (masked_flags & F2FS_COMPR_FL) { 1864 if (!f2fs_disable_compressed_file(inode)) 1865 return -EINVAL; 1866 } 1867 if (iflags & F2FS_NOCOMP_FL) 1868 return -EINVAL; 1869 if (iflags & F2FS_COMPR_FL) { 1870 if (!f2fs_may_compress(inode)) 1871 return -EINVAL; 1872 if (S_ISREG(inode->i_mode) && inode->i_size) 1873 return -EINVAL; 1874 1875 set_compress_context(inode); 1876 } 1877 } 1878 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) { 1879 if (masked_flags & F2FS_COMPR_FL) 1880 return -EINVAL; 1881 } 1882 1883 fi->i_flags = iflags | (fi->i_flags & ~mask); 1884 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1885 (fi->i_flags & F2FS_NOCOMP_FL)); 1886 1887 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1888 set_inode_flag(inode, FI_PROJ_INHERIT); 1889 else 1890 clear_inode_flag(inode, FI_PROJ_INHERIT); 1891 1892 inode->i_ctime = current_time(inode); 1893 f2fs_set_inode_flags(inode); 1894 f2fs_mark_inode_dirty_sync(inode, true); 1895 return 0; 1896 } 1897 1898 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1899 1900 /* 1901 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1902 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1903 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1904 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1905 * 1906 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1907 * FS_IOC_FSSETXATTR is done by the VFS. 1908 */ 1909 1910 static const struct { 1911 u32 iflag; 1912 u32 fsflag; 1913 } f2fs_fsflags_map[] = { 1914 { F2FS_COMPR_FL, FS_COMPR_FL }, 1915 { F2FS_SYNC_FL, FS_SYNC_FL }, 1916 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1917 { F2FS_APPEND_FL, FS_APPEND_FL }, 1918 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1919 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1920 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1921 { F2FS_INDEX_FL, FS_INDEX_FL }, 1922 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1923 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1924 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1925 }; 1926 1927 #define F2FS_GETTABLE_FS_FL ( \ 1928 FS_COMPR_FL | \ 1929 FS_SYNC_FL | \ 1930 FS_IMMUTABLE_FL | \ 1931 FS_APPEND_FL | \ 1932 FS_NODUMP_FL | \ 1933 FS_NOATIME_FL | \ 1934 FS_NOCOMP_FL | \ 1935 FS_INDEX_FL | \ 1936 FS_DIRSYNC_FL | \ 1937 FS_PROJINHERIT_FL | \ 1938 FS_ENCRYPT_FL | \ 1939 FS_INLINE_DATA_FL | \ 1940 FS_NOCOW_FL | \ 1941 FS_VERITY_FL | \ 1942 FS_CASEFOLD_FL) 1943 1944 #define F2FS_SETTABLE_FS_FL ( \ 1945 FS_COMPR_FL | \ 1946 FS_SYNC_FL | \ 1947 FS_IMMUTABLE_FL | \ 1948 FS_APPEND_FL | \ 1949 FS_NODUMP_FL | \ 1950 FS_NOATIME_FL | \ 1951 FS_NOCOMP_FL | \ 1952 FS_DIRSYNC_FL | \ 1953 FS_PROJINHERIT_FL | \ 1954 FS_CASEFOLD_FL) 1955 1956 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 1957 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 1958 { 1959 u32 fsflags = 0; 1960 int i; 1961 1962 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1963 if (iflags & f2fs_fsflags_map[i].iflag) 1964 fsflags |= f2fs_fsflags_map[i].fsflag; 1965 1966 return fsflags; 1967 } 1968 1969 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 1970 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 1971 { 1972 u32 iflags = 0; 1973 int i; 1974 1975 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1976 if (fsflags & f2fs_fsflags_map[i].fsflag) 1977 iflags |= f2fs_fsflags_map[i].iflag; 1978 1979 return iflags; 1980 } 1981 1982 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1983 { 1984 struct inode *inode = file_inode(filp); 1985 1986 return put_user(inode->i_generation, (int __user *)arg); 1987 } 1988 1989 static int f2fs_ioc_start_atomic_write(struct file *filp) 1990 { 1991 struct inode *inode = file_inode(filp); 1992 struct user_namespace *mnt_userns = file_mnt_user_ns(filp); 1993 struct f2fs_inode_info *fi = F2FS_I(inode); 1994 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1995 int ret; 1996 1997 if (!inode_owner_or_capable(mnt_userns, inode)) 1998 return -EACCES; 1999 2000 if (!S_ISREG(inode->i_mode)) 2001 return -EINVAL; 2002 2003 if (filp->f_flags & O_DIRECT) 2004 return -EINVAL; 2005 2006 ret = mnt_want_write_file(filp); 2007 if (ret) 2008 return ret; 2009 2010 inode_lock(inode); 2011 2012 if (!f2fs_disable_compressed_file(inode)) { 2013 ret = -EINVAL; 2014 goto out; 2015 } 2016 2017 if (f2fs_is_atomic_file(inode)) { 2018 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) 2019 ret = -EINVAL; 2020 goto out; 2021 } 2022 2023 ret = f2fs_convert_inline_inode(inode); 2024 if (ret) 2025 goto out; 2026 2027 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2028 2029 /* 2030 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2031 * f2fs_is_atomic_file. 2032 */ 2033 if (get_dirty_pages(inode)) 2034 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2035 inode->i_ino, get_dirty_pages(inode)); 2036 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2037 if (ret) { 2038 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2039 goto out; 2040 } 2041 2042 spin_lock(&sbi->inode_lock[ATOMIC_FILE]); 2043 if (list_empty(&fi->inmem_ilist)) 2044 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]); 2045 sbi->atomic_files++; 2046 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]); 2047 2048 /* add inode in inmem_list first and set atomic_file */ 2049 set_inode_flag(inode, FI_ATOMIC_FILE); 2050 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2051 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2052 2053 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2054 F2FS_I(inode)->inmem_task = current; 2055 stat_update_max_atomic_write(inode); 2056 out: 2057 inode_unlock(inode); 2058 mnt_drop_write_file(filp); 2059 return ret; 2060 } 2061 2062 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2063 { 2064 struct inode *inode = file_inode(filp); 2065 struct user_namespace *mnt_userns = file_mnt_user_ns(filp); 2066 int ret; 2067 2068 if (!inode_owner_or_capable(mnt_userns, inode)) 2069 return -EACCES; 2070 2071 ret = mnt_want_write_file(filp); 2072 if (ret) 2073 return ret; 2074 2075 f2fs_balance_fs(F2FS_I_SB(inode), true); 2076 2077 inode_lock(inode); 2078 2079 if (f2fs_is_volatile_file(inode)) { 2080 ret = -EINVAL; 2081 goto err_out; 2082 } 2083 2084 if (f2fs_is_atomic_file(inode)) { 2085 ret = f2fs_commit_inmem_pages(inode); 2086 if (ret) 2087 goto err_out; 2088 2089 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2090 if (!ret) 2091 f2fs_drop_inmem_pages(inode); 2092 } else { 2093 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2094 } 2095 err_out: 2096 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) { 2097 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2098 ret = -EINVAL; 2099 } 2100 inode_unlock(inode); 2101 mnt_drop_write_file(filp); 2102 return ret; 2103 } 2104 2105 static int f2fs_ioc_start_volatile_write(struct file *filp) 2106 { 2107 struct inode *inode = file_inode(filp); 2108 struct user_namespace *mnt_userns = file_mnt_user_ns(filp); 2109 int ret; 2110 2111 if (!inode_owner_or_capable(mnt_userns, inode)) 2112 return -EACCES; 2113 2114 if (!S_ISREG(inode->i_mode)) 2115 return -EINVAL; 2116 2117 ret = mnt_want_write_file(filp); 2118 if (ret) 2119 return ret; 2120 2121 inode_lock(inode); 2122 2123 if (f2fs_is_volatile_file(inode)) 2124 goto out; 2125 2126 ret = f2fs_convert_inline_inode(inode); 2127 if (ret) 2128 goto out; 2129 2130 stat_inc_volatile_write(inode); 2131 stat_update_max_volatile_write(inode); 2132 2133 set_inode_flag(inode, FI_VOLATILE_FILE); 2134 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2135 out: 2136 inode_unlock(inode); 2137 mnt_drop_write_file(filp); 2138 return ret; 2139 } 2140 2141 static int f2fs_ioc_release_volatile_write(struct file *filp) 2142 { 2143 struct inode *inode = file_inode(filp); 2144 struct user_namespace *mnt_userns = file_mnt_user_ns(filp); 2145 int ret; 2146 2147 if (!inode_owner_or_capable(mnt_userns, inode)) 2148 return -EACCES; 2149 2150 ret = mnt_want_write_file(filp); 2151 if (ret) 2152 return ret; 2153 2154 inode_lock(inode); 2155 2156 if (!f2fs_is_volatile_file(inode)) 2157 goto out; 2158 2159 if (!f2fs_is_first_block_written(inode)) { 2160 ret = truncate_partial_data_page(inode, 0, true); 2161 goto out; 2162 } 2163 2164 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 2165 out: 2166 inode_unlock(inode); 2167 mnt_drop_write_file(filp); 2168 return ret; 2169 } 2170 2171 static int f2fs_ioc_abort_volatile_write(struct file *filp) 2172 { 2173 struct inode *inode = file_inode(filp); 2174 struct user_namespace *mnt_userns = file_mnt_user_ns(filp); 2175 int ret; 2176 2177 if (!inode_owner_or_capable(mnt_userns, inode)) 2178 return -EACCES; 2179 2180 ret = mnt_want_write_file(filp); 2181 if (ret) 2182 return ret; 2183 2184 inode_lock(inode); 2185 2186 if (f2fs_is_atomic_file(inode)) 2187 f2fs_drop_inmem_pages(inode); 2188 if (f2fs_is_volatile_file(inode)) { 2189 clear_inode_flag(inode, FI_VOLATILE_FILE); 2190 stat_dec_volatile_write(inode); 2191 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2192 } 2193 2194 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 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 set_sbi_flag(sbi, SBI_IS_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); 2236 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2237 thaw_bdev(sb->s_bdev); 2238 break; 2239 case F2FS_GOING_DOWN_METASYNC: 2240 /* do checkpoint only */ 2241 ret = f2fs_sync_fs(sb, 1); 2242 if (ret) 2243 goto out; 2244 f2fs_stop_checkpoint(sbi, false); 2245 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2246 break; 2247 case F2FS_GOING_DOWN_NOSYNC: 2248 f2fs_stop_checkpoint(sbi, false); 2249 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2250 break; 2251 case F2FS_GOING_DOWN_METAFLUSH: 2252 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2253 f2fs_stop_checkpoint(sbi, false); 2254 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2255 break; 2256 case F2FS_GOING_DOWN_NEED_FSCK: 2257 set_sbi_flag(sbi, SBI_NEED_FSCK); 2258 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2259 set_sbi_flag(sbi, SBI_IS_DIRTY); 2260 /* do checkpoint only */ 2261 ret = f2fs_sync_fs(sb, 1); 2262 goto out; 2263 default: 2264 ret = -EINVAL; 2265 goto out; 2266 } 2267 2268 f2fs_stop_gc_thread(sbi); 2269 f2fs_stop_discard_thread(sbi); 2270 2271 f2fs_drop_discard_cmd(sbi); 2272 clear_opt(sbi, DISCARD); 2273 2274 f2fs_update_time(sbi, REQ_TIME); 2275 out: 2276 if (in != F2FS_GOING_DOWN_FULLSYNC) 2277 mnt_drop_write_file(filp); 2278 2279 trace_f2fs_shutdown(sbi, in, ret); 2280 2281 return ret; 2282 } 2283 2284 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2285 { 2286 struct inode *inode = file_inode(filp); 2287 struct super_block *sb = inode->i_sb; 2288 struct fstrim_range range; 2289 int ret; 2290 2291 if (!capable(CAP_SYS_ADMIN)) 2292 return -EPERM; 2293 2294 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2295 return -EOPNOTSUPP; 2296 2297 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2298 sizeof(range))) 2299 return -EFAULT; 2300 2301 ret = mnt_want_write_file(filp); 2302 if (ret) 2303 return ret; 2304 2305 range.minlen = max((unsigned int)range.minlen, 2306 bdev_discard_granularity(sb->s_bdev)); 2307 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2308 mnt_drop_write_file(filp); 2309 if (ret < 0) 2310 return ret; 2311 2312 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2313 sizeof(range))) 2314 return -EFAULT; 2315 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2316 return 0; 2317 } 2318 2319 static bool uuid_is_nonzero(__u8 u[16]) 2320 { 2321 int i; 2322 2323 for (i = 0; i < 16; i++) 2324 if (u[i]) 2325 return true; 2326 return false; 2327 } 2328 2329 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2330 { 2331 struct inode *inode = file_inode(filp); 2332 2333 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2334 return -EOPNOTSUPP; 2335 2336 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2337 2338 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2339 } 2340 2341 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2342 { 2343 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2344 return -EOPNOTSUPP; 2345 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2346 } 2347 2348 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2349 { 2350 struct inode *inode = file_inode(filp); 2351 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2352 int err; 2353 2354 if (!f2fs_sb_has_encrypt(sbi)) 2355 return -EOPNOTSUPP; 2356 2357 err = mnt_want_write_file(filp); 2358 if (err) 2359 return err; 2360 2361 f2fs_down_write(&sbi->sb_lock); 2362 2363 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2364 goto got_it; 2365 2366 /* update superblock with uuid */ 2367 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2368 2369 err = f2fs_commit_super(sbi, false); 2370 if (err) { 2371 /* undo new data */ 2372 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2373 goto out_err; 2374 } 2375 got_it: 2376 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 2377 16)) 2378 err = -EFAULT; 2379 out_err: 2380 f2fs_up_write(&sbi->sb_lock); 2381 mnt_drop_write_file(filp); 2382 return err; 2383 } 2384 2385 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2386 unsigned long arg) 2387 { 2388 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2389 return -EOPNOTSUPP; 2390 2391 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2392 } 2393 2394 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2395 { 2396 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2397 return -EOPNOTSUPP; 2398 2399 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2400 } 2401 2402 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2403 { 2404 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2405 return -EOPNOTSUPP; 2406 2407 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2408 } 2409 2410 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2411 unsigned long arg) 2412 { 2413 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2414 return -EOPNOTSUPP; 2415 2416 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2417 } 2418 2419 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2420 unsigned long arg) 2421 { 2422 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2423 return -EOPNOTSUPP; 2424 2425 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2426 } 2427 2428 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2429 { 2430 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2431 return -EOPNOTSUPP; 2432 2433 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2434 } 2435 2436 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2437 { 2438 struct inode *inode = file_inode(filp); 2439 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2440 __u32 sync; 2441 int ret; 2442 2443 if (!capable(CAP_SYS_ADMIN)) 2444 return -EPERM; 2445 2446 if (get_user(sync, (__u32 __user *)arg)) 2447 return -EFAULT; 2448 2449 if (f2fs_readonly(sbi->sb)) 2450 return -EROFS; 2451 2452 ret = mnt_want_write_file(filp); 2453 if (ret) 2454 return ret; 2455 2456 if (!sync) { 2457 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2458 ret = -EBUSY; 2459 goto out; 2460 } 2461 } else { 2462 f2fs_down_write(&sbi->gc_lock); 2463 } 2464 2465 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO); 2466 out: 2467 mnt_drop_write_file(filp); 2468 return ret; 2469 } 2470 2471 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2472 { 2473 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2474 u64 end; 2475 int ret; 2476 2477 if (!capable(CAP_SYS_ADMIN)) 2478 return -EPERM; 2479 if (f2fs_readonly(sbi->sb)) 2480 return -EROFS; 2481 2482 end = range->start + range->len; 2483 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2484 end >= MAX_BLKADDR(sbi)) 2485 return -EINVAL; 2486 2487 ret = mnt_want_write_file(filp); 2488 if (ret) 2489 return ret; 2490 2491 do_more: 2492 if (!range->sync) { 2493 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2494 ret = -EBUSY; 2495 goto out; 2496 } 2497 } else { 2498 f2fs_down_write(&sbi->gc_lock); 2499 } 2500 2501 ret = f2fs_gc(sbi, range->sync, true, false, 2502 GET_SEGNO(sbi, range->start)); 2503 if (ret) { 2504 if (ret == -EBUSY) 2505 ret = -EAGAIN; 2506 goto out; 2507 } 2508 range->start += BLKS_PER_SEC(sbi); 2509 if (range->start <= end) 2510 goto do_more; 2511 out: 2512 mnt_drop_write_file(filp); 2513 return ret; 2514 } 2515 2516 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2517 { 2518 struct f2fs_gc_range range; 2519 2520 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2521 sizeof(range))) 2522 return -EFAULT; 2523 return __f2fs_ioc_gc_range(filp, &range); 2524 } 2525 2526 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg) 2527 { 2528 struct inode *inode = file_inode(filp); 2529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2530 int ret; 2531 2532 if (!capable(CAP_SYS_ADMIN)) 2533 return -EPERM; 2534 2535 if (f2fs_readonly(sbi->sb)) 2536 return -EROFS; 2537 2538 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2539 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2540 return -EINVAL; 2541 } 2542 2543 ret = mnt_want_write_file(filp); 2544 if (ret) 2545 return ret; 2546 2547 ret = f2fs_sync_fs(sbi->sb, 1); 2548 2549 mnt_drop_write_file(filp); 2550 return ret; 2551 } 2552 2553 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2554 struct file *filp, 2555 struct f2fs_defragment *range) 2556 { 2557 struct inode *inode = file_inode(filp); 2558 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2559 .m_seg_type = NO_CHECK_TYPE, 2560 .m_may_create = false }; 2561 struct extent_info ei = {0, 0, 0}; 2562 pgoff_t pg_start, pg_end, next_pgofs; 2563 unsigned int blk_per_seg = sbi->blocks_per_seg; 2564 unsigned int total = 0, sec_num; 2565 block_t blk_end = 0; 2566 bool fragmented = false; 2567 int err; 2568 2569 pg_start = range->start >> PAGE_SHIFT; 2570 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2571 2572 f2fs_balance_fs(sbi, true); 2573 2574 inode_lock(inode); 2575 2576 /* if in-place-update policy is enabled, don't waste time here */ 2577 set_inode_flag(inode, FI_OPU_WRITE); 2578 if (f2fs_should_update_inplace(inode, NULL)) { 2579 err = -EINVAL; 2580 goto out; 2581 } 2582 2583 /* writeback all dirty pages in the range */ 2584 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2585 range->start + range->len - 1); 2586 if (err) 2587 goto out; 2588 2589 /* 2590 * lookup mapping info in extent cache, skip defragmenting if physical 2591 * block addresses are continuous. 2592 */ 2593 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) { 2594 if (ei.fofs + ei.len >= pg_end) 2595 goto out; 2596 } 2597 2598 map.m_lblk = pg_start; 2599 map.m_next_pgofs = &next_pgofs; 2600 2601 /* 2602 * lookup mapping info in dnode page cache, skip defragmenting if all 2603 * physical block addresses are continuous even if there are hole(s) 2604 * in logical blocks. 2605 */ 2606 while (map.m_lblk < pg_end) { 2607 map.m_len = pg_end - map.m_lblk; 2608 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2609 if (err) 2610 goto out; 2611 2612 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2613 map.m_lblk = next_pgofs; 2614 continue; 2615 } 2616 2617 if (blk_end && blk_end != map.m_pblk) 2618 fragmented = true; 2619 2620 /* record total count of block that we're going to move */ 2621 total += map.m_len; 2622 2623 blk_end = map.m_pblk + map.m_len; 2624 2625 map.m_lblk += map.m_len; 2626 } 2627 2628 if (!fragmented) { 2629 total = 0; 2630 goto out; 2631 } 2632 2633 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi)); 2634 2635 /* 2636 * make sure there are enough free section for LFS allocation, this can 2637 * avoid defragment running in SSR mode when free section are allocated 2638 * intensively 2639 */ 2640 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2641 err = -EAGAIN; 2642 goto out; 2643 } 2644 2645 map.m_lblk = pg_start; 2646 map.m_len = pg_end - pg_start; 2647 total = 0; 2648 2649 while (map.m_lblk < pg_end) { 2650 pgoff_t idx; 2651 int cnt = 0; 2652 2653 do_map: 2654 map.m_len = pg_end - map.m_lblk; 2655 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2656 if (err) 2657 goto clear_out; 2658 2659 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2660 map.m_lblk = next_pgofs; 2661 goto check; 2662 } 2663 2664 set_inode_flag(inode, FI_SKIP_WRITES); 2665 2666 idx = map.m_lblk; 2667 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2668 struct page *page; 2669 2670 page = f2fs_get_lock_data_page(inode, idx, true); 2671 if (IS_ERR(page)) { 2672 err = PTR_ERR(page); 2673 goto clear_out; 2674 } 2675 2676 set_page_dirty(page); 2677 f2fs_put_page(page, 1); 2678 2679 idx++; 2680 cnt++; 2681 total++; 2682 } 2683 2684 map.m_lblk = idx; 2685 check: 2686 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2687 goto do_map; 2688 2689 clear_inode_flag(inode, FI_SKIP_WRITES); 2690 2691 err = filemap_fdatawrite(inode->i_mapping); 2692 if (err) 2693 goto out; 2694 } 2695 clear_out: 2696 clear_inode_flag(inode, FI_SKIP_WRITES); 2697 out: 2698 clear_inode_flag(inode, FI_OPU_WRITE); 2699 inode_unlock(inode); 2700 if (!err) 2701 range->len = (u64)total << PAGE_SHIFT; 2702 return err; 2703 } 2704 2705 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2706 { 2707 struct inode *inode = file_inode(filp); 2708 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2709 struct f2fs_defragment range; 2710 int err; 2711 2712 if (!capable(CAP_SYS_ADMIN)) 2713 return -EPERM; 2714 2715 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2716 return -EINVAL; 2717 2718 if (f2fs_readonly(sbi->sb)) 2719 return -EROFS; 2720 2721 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2722 sizeof(range))) 2723 return -EFAULT; 2724 2725 /* verify alignment of offset & size */ 2726 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2727 return -EINVAL; 2728 2729 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2730 max_file_blocks(inode))) 2731 return -EINVAL; 2732 2733 err = mnt_want_write_file(filp); 2734 if (err) 2735 return err; 2736 2737 err = f2fs_defragment_range(sbi, filp, &range); 2738 mnt_drop_write_file(filp); 2739 2740 f2fs_update_time(sbi, REQ_TIME); 2741 if (err < 0) 2742 return err; 2743 2744 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2745 sizeof(range))) 2746 return -EFAULT; 2747 2748 return 0; 2749 } 2750 2751 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2752 struct file *file_out, loff_t pos_out, size_t len) 2753 { 2754 struct inode *src = file_inode(file_in); 2755 struct inode *dst = file_inode(file_out); 2756 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2757 size_t olen = len, dst_max_i_size = 0; 2758 size_t dst_osize; 2759 int ret; 2760 2761 if (file_in->f_path.mnt != file_out->f_path.mnt || 2762 src->i_sb != dst->i_sb) 2763 return -EXDEV; 2764 2765 if (unlikely(f2fs_readonly(src->i_sb))) 2766 return -EROFS; 2767 2768 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2769 return -EINVAL; 2770 2771 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2772 return -EOPNOTSUPP; 2773 2774 if (pos_out < 0 || pos_in < 0) 2775 return -EINVAL; 2776 2777 if (src == dst) { 2778 if (pos_in == pos_out) 2779 return 0; 2780 if (pos_out > pos_in && pos_out < pos_in + len) 2781 return -EINVAL; 2782 } 2783 2784 inode_lock(src); 2785 if (src != dst) { 2786 ret = -EBUSY; 2787 if (!inode_trylock(dst)) 2788 goto out; 2789 } 2790 2791 ret = -EINVAL; 2792 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2793 goto out_unlock; 2794 if (len == 0) 2795 olen = len = src->i_size - pos_in; 2796 if (pos_in + len == src->i_size) 2797 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2798 if (len == 0) { 2799 ret = 0; 2800 goto out_unlock; 2801 } 2802 2803 dst_osize = dst->i_size; 2804 if (pos_out + olen > dst->i_size) 2805 dst_max_i_size = pos_out + olen; 2806 2807 /* verify the end result is block aligned */ 2808 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2809 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2810 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2811 goto out_unlock; 2812 2813 ret = f2fs_convert_inline_inode(src); 2814 if (ret) 2815 goto out_unlock; 2816 2817 ret = f2fs_convert_inline_inode(dst); 2818 if (ret) 2819 goto out_unlock; 2820 2821 /* write out all dirty pages from offset */ 2822 ret = filemap_write_and_wait_range(src->i_mapping, 2823 pos_in, pos_in + len); 2824 if (ret) 2825 goto out_unlock; 2826 2827 ret = filemap_write_and_wait_range(dst->i_mapping, 2828 pos_out, pos_out + len); 2829 if (ret) 2830 goto out_unlock; 2831 2832 f2fs_balance_fs(sbi, true); 2833 2834 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2835 if (src != dst) { 2836 ret = -EBUSY; 2837 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2838 goto out_src; 2839 } 2840 2841 f2fs_lock_op(sbi); 2842 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2843 pos_out >> F2FS_BLKSIZE_BITS, 2844 len >> F2FS_BLKSIZE_BITS, false); 2845 2846 if (!ret) { 2847 if (dst_max_i_size) 2848 f2fs_i_size_write(dst, dst_max_i_size); 2849 else if (dst_osize != dst->i_size) 2850 f2fs_i_size_write(dst, dst_osize); 2851 } 2852 f2fs_unlock_op(sbi); 2853 2854 if (src != dst) 2855 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2856 out_src: 2857 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2858 out_unlock: 2859 if (src != dst) 2860 inode_unlock(dst); 2861 out: 2862 inode_unlock(src); 2863 return ret; 2864 } 2865 2866 static int __f2fs_ioc_move_range(struct file *filp, 2867 struct f2fs_move_range *range) 2868 { 2869 struct fd dst; 2870 int err; 2871 2872 if (!(filp->f_mode & FMODE_READ) || 2873 !(filp->f_mode & FMODE_WRITE)) 2874 return -EBADF; 2875 2876 dst = fdget(range->dst_fd); 2877 if (!dst.file) 2878 return -EBADF; 2879 2880 if (!(dst.file->f_mode & FMODE_WRITE)) { 2881 err = -EBADF; 2882 goto err_out; 2883 } 2884 2885 err = mnt_want_write_file(filp); 2886 if (err) 2887 goto err_out; 2888 2889 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2890 range->pos_out, range->len); 2891 2892 mnt_drop_write_file(filp); 2893 err_out: 2894 fdput(dst); 2895 return err; 2896 } 2897 2898 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2899 { 2900 struct f2fs_move_range range; 2901 2902 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2903 sizeof(range))) 2904 return -EFAULT; 2905 return __f2fs_ioc_move_range(filp, &range); 2906 } 2907 2908 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2909 { 2910 struct inode *inode = file_inode(filp); 2911 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2912 struct sit_info *sm = SIT_I(sbi); 2913 unsigned int start_segno = 0, end_segno = 0; 2914 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2915 struct f2fs_flush_device range; 2916 int ret; 2917 2918 if (!capable(CAP_SYS_ADMIN)) 2919 return -EPERM; 2920 2921 if (f2fs_readonly(sbi->sb)) 2922 return -EROFS; 2923 2924 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2925 return -EINVAL; 2926 2927 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2928 sizeof(range))) 2929 return -EFAULT; 2930 2931 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2932 __is_large_section(sbi)) { 2933 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2934 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2935 return -EINVAL; 2936 } 2937 2938 ret = mnt_want_write_file(filp); 2939 if (ret) 2940 return ret; 2941 2942 if (range.dev_num != 0) 2943 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2944 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2945 2946 start_segno = sm->last_victim[FLUSH_DEVICE]; 2947 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2948 start_segno = dev_start_segno; 2949 end_segno = min(start_segno + range.segments, dev_end_segno); 2950 2951 while (start_segno < end_segno) { 2952 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2953 ret = -EBUSY; 2954 goto out; 2955 } 2956 sm->last_victim[GC_CB] = end_segno + 1; 2957 sm->last_victim[GC_GREEDY] = end_segno + 1; 2958 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 2959 ret = f2fs_gc(sbi, true, true, true, start_segno); 2960 if (ret == -EAGAIN) 2961 ret = 0; 2962 else if (ret < 0) 2963 break; 2964 start_segno++; 2965 } 2966 out: 2967 mnt_drop_write_file(filp); 2968 return ret; 2969 } 2970 2971 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 2972 { 2973 struct inode *inode = file_inode(filp); 2974 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 2975 2976 /* Must validate to set it with SQLite behavior in Android. */ 2977 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 2978 2979 return put_user(sb_feature, (u32 __user *)arg); 2980 } 2981 2982 #ifdef CONFIG_QUOTA 2983 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 2984 { 2985 struct dquot *transfer_to[MAXQUOTAS] = {}; 2986 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2987 struct super_block *sb = sbi->sb; 2988 int err = 0; 2989 2990 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 2991 if (!IS_ERR(transfer_to[PRJQUOTA])) { 2992 err = __dquot_transfer(inode, transfer_to); 2993 if (err) 2994 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2995 dqput(transfer_to[PRJQUOTA]); 2996 } 2997 return err; 2998 } 2999 3000 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3001 { 3002 struct f2fs_inode_info *fi = F2FS_I(inode); 3003 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3004 struct f2fs_inode *ri = NULL; 3005 kprojid_t kprojid; 3006 int err; 3007 3008 if (!f2fs_sb_has_project_quota(sbi)) { 3009 if (projid != F2FS_DEF_PROJID) 3010 return -EOPNOTSUPP; 3011 else 3012 return 0; 3013 } 3014 3015 if (!f2fs_has_extra_attr(inode)) 3016 return -EOPNOTSUPP; 3017 3018 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3019 3020 if (projid_eq(kprojid, F2FS_I(inode)->i_projid)) 3021 return 0; 3022 3023 err = -EPERM; 3024 /* Is it quota file? Do not allow user to mess with it */ 3025 if (IS_NOQUOTA(inode)) 3026 return err; 3027 3028 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3029 return -EOVERFLOW; 3030 3031 err = f2fs_dquot_initialize(inode); 3032 if (err) 3033 return err; 3034 3035 f2fs_lock_op(sbi); 3036 err = f2fs_transfer_project_quota(inode, kprojid); 3037 if (err) 3038 goto out_unlock; 3039 3040 F2FS_I(inode)->i_projid = kprojid; 3041 inode->i_ctime = current_time(inode); 3042 f2fs_mark_inode_dirty_sync(inode, true); 3043 out_unlock: 3044 f2fs_unlock_op(sbi); 3045 return err; 3046 } 3047 #else 3048 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3049 { 3050 return 0; 3051 } 3052 3053 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3054 { 3055 if (projid != F2FS_DEF_PROJID) 3056 return -EOPNOTSUPP; 3057 return 0; 3058 } 3059 #endif 3060 3061 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3062 { 3063 struct inode *inode = d_inode(dentry); 3064 struct f2fs_inode_info *fi = F2FS_I(inode); 3065 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3066 3067 if (IS_ENCRYPTED(inode)) 3068 fsflags |= FS_ENCRYPT_FL; 3069 if (IS_VERITY(inode)) 3070 fsflags |= FS_VERITY_FL; 3071 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3072 fsflags |= FS_INLINE_DATA_FL; 3073 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3074 fsflags |= FS_NOCOW_FL; 3075 3076 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3077 3078 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3079 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3080 3081 return 0; 3082 } 3083 3084 int f2fs_fileattr_set(struct user_namespace *mnt_userns, 3085 struct dentry *dentry, struct fileattr *fa) 3086 { 3087 struct inode *inode = d_inode(dentry); 3088 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3089 u32 iflags; 3090 int err; 3091 3092 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3093 return -EIO; 3094 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3095 return -ENOSPC; 3096 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3097 return -EOPNOTSUPP; 3098 fsflags &= F2FS_SETTABLE_FS_FL; 3099 if (!fa->flags_valid) 3100 mask &= FS_COMMON_FL; 3101 3102 iflags = f2fs_fsflags_to_iflags(fsflags); 3103 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3104 return -EOPNOTSUPP; 3105 3106 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3107 if (!err) 3108 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3109 3110 return err; 3111 } 3112 3113 int f2fs_pin_file_control(struct inode *inode, bool inc) 3114 { 3115 struct f2fs_inode_info *fi = F2FS_I(inode); 3116 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3117 3118 /* Use i_gc_failures for normal file as a risk signal. */ 3119 if (inc) 3120 f2fs_i_gc_failures_write(inode, 3121 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3122 3123 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3124 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3125 __func__, inode->i_ino, 3126 fi->i_gc_failures[GC_FAILURE_PIN]); 3127 clear_inode_flag(inode, FI_PIN_FILE); 3128 return -EAGAIN; 3129 } 3130 return 0; 3131 } 3132 3133 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3134 { 3135 struct inode *inode = file_inode(filp); 3136 __u32 pin; 3137 int ret = 0; 3138 3139 if (get_user(pin, (__u32 __user *)arg)) 3140 return -EFAULT; 3141 3142 if (!S_ISREG(inode->i_mode)) 3143 return -EINVAL; 3144 3145 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3146 return -EROFS; 3147 3148 ret = mnt_want_write_file(filp); 3149 if (ret) 3150 return ret; 3151 3152 inode_lock(inode); 3153 3154 if (!pin) { 3155 clear_inode_flag(inode, FI_PIN_FILE); 3156 f2fs_i_gc_failures_write(inode, 0); 3157 goto done; 3158 } 3159 3160 if (f2fs_should_update_outplace(inode, NULL)) { 3161 ret = -EINVAL; 3162 goto out; 3163 } 3164 3165 if (f2fs_pin_file_control(inode, false)) { 3166 ret = -EAGAIN; 3167 goto out; 3168 } 3169 3170 ret = f2fs_convert_inline_inode(inode); 3171 if (ret) 3172 goto out; 3173 3174 if (!f2fs_disable_compressed_file(inode)) { 3175 ret = -EOPNOTSUPP; 3176 goto out; 3177 } 3178 3179 set_inode_flag(inode, FI_PIN_FILE); 3180 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3181 done: 3182 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3183 out: 3184 inode_unlock(inode); 3185 mnt_drop_write_file(filp); 3186 return ret; 3187 } 3188 3189 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3190 { 3191 struct inode *inode = file_inode(filp); 3192 __u32 pin = 0; 3193 3194 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3195 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3196 return put_user(pin, (u32 __user *)arg); 3197 } 3198 3199 int f2fs_precache_extents(struct inode *inode) 3200 { 3201 struct f2fs_inode_info *fi = F2FS_I(inode); 3202 struct f2fs_map_blocks map; 3203 pgoff_t m_next_extent; 3204 loff_t end; 3205 int err; 3206 3207 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3208 return -EOPNOTSUPP; 3209 3210 map.m_lblk = 0; 3211 map.m_next_pgofs = NULL; 3212 map.m_next_extent = &m_next_extent; 3213 map.m_seg_type = NO_CHECK_TYPE; 3214 map.m_may_create = false; 3215 end = max_file_blocks(inode); 3216 3217 while (map.m_lblk < end) { 3218 map.m_len = end - map.m_lblk; 3219 3220 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3221 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE); 3222 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3223 if (err) 3224 return err; 3225 3226 map.m_lblk = m_next_extent; 3227 } 3228 3229 return 0; 3230 } 3231 3232 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg) 3233 { 3234 return f2fs_precache_extents(file_inode(filp)); 3235 } 3236 3237 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3238 { 3239 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3240 __u64 block_count; 3241 3242 if (!capable(CAP_SYS_ADMIN)) 3243 return -EPERM; 3244 3245 if (f2fs_readonly(sbi->sb)) 3246 return -EROFS; 3247 3248 if (copy_from_user(&block_count, (void __user *)arg, 3249 sizeof(block_count))) 3250 return -EFAULT; 3251 3252 return f2fs_resize_fs(sbi, block_count); 3253 } 3254 3255 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3256 { 3257 struct inode *inode = file_inode(filp); 3258 3259 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3260 3261 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3262 f2fs_warn(F2FS_I_SB(inode), 3263 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3264 inode->i_ino); 3265 return -EOPNOTSUPP; 3266 } 3267 3268 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3269 } 3270 3271 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3272 { 3273 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3274 return -EOPNOTSUPP; 3275 3276 return fsverity_ioctl_measure(filp, (void __user *)arg); 3277 } 3278 3279 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3280 { 3281 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3282 return -EOPNOTSUPP; 3283 3284 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3285 } 3286 3287 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3288 { 3289 struct inode *inode = file_inode(filp); 3290 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3291 char *vbuf; 3292 int count; 3293 int err = 0; 3294 3295 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3296 if (!vbuf) 3297 return -ENOMEM; 3298 3299 f2fs_down_read(&sbi->sb_lock); 3300 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3301 ARRAY_SIZE(sbi->raw_super->volume_name), 3302 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3303 f2fs_up_read(&sbi->sb_lock); 3304 3305 if (copy_to_user((char __user *)arg, vbuf, 3306 min(FSLABEL_MAX, count))) 3307 err = -EFAULT; 3308 3309 kfree(vbuf); 3310 return err; 3311 } 3312 3313 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3314 { 3315 struct inode *inode = file_inode(filp); 3316 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3317 char *vbuf; 3318 int err = 0; 3319 3320 if (!capable(CAP_SYS_ADMIN)) 3321 return -EPERM; 3322 3323 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3324 if (IS_ERR(vbuf)) 3325 return PTR_ERR(vbuf); 3326 3327 err = mnt_want_write_file(filp); 3328 if (err) 3329 goto out; 3330 3331 f2fs_down_write(&sbi->sb_lock); 3332 3333 memset(sbi->raw_super->volume_name, 0, 3334 sizeof(sbi->raw_super->volume_name)); 3335 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3336 sbi->raw_super->volume_name, 3337 ARRAY_SIZE(sbi->raw_super->volume_name)); 3338 3339 err = f2fs_commit_super(sbi, false); 3340 3341 f2fs_up_write(&sbi->sb_lock); 3342 3343 mnt_drop_write_file(filp); 3344 out: 3345 kfree(vbuf); 3346 return err; 3347 } 3348 3349 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg) 3350 { 3351 struct inode *inode = file_inode(filp); 3352 __u64 blocks; 3353 3354 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3355 return -EOPNOTSUPP; 3356 3357 if (!f2fs_compressed_file(inode)) 3358 return -EINVAL; 3359 3360 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3361 return put_user(blocks, (u64 __user *)arg); 3362 } 3363 3364 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3365 { 3366 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3367 unsigned int released_blocks = 0; 3368 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3369 block_t blkaddr; 3370 int i; 3371 3372 for (i = 0; i < count; i++) { 3373 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3374 dn->ofs_in_node + i); 3375 3376 if (!__is_valid_data_blkaddr(blkaddr)) 3377 continue; 3378 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3379 DATA_GENERIC_ENHANCE))) 3380 return -EFSCORRUPTED; 3381 } 3382 3383 while (count) { 3384 int compr_blocks = 0; 3385 3386 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3387 blkaddr = f2fs_data_blkaddr(dn); 3388 3389 if (i == 0) { 3390 if (blkaddr == COMPRESS_ADDR) 3391 continue; 3392 dn->ofs_in_node += cluster_size; 3393 goto next; 3394 } 3395 3396 if (__is_valid_data_blkaddr(blkaddr)) 3397 compr_blocks++; 3398 3399 if (blkaddr != NEW_ADDR) 3400 continue; 3401 3402 dn->data_blkaddr = NULL_ADDR; 3403 f2fs_set_data_blkaddr(dn); 3404 } 3405 3406 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3407 dec_valid_block_count(sbi, dn->inode, 3408 cluster_size - compr_blocks); 3409 3410 released_blocks += cluster_size - compr_blocks; 3411 next: 3412 count -= cluster_size; 3413 } 3414 3415 return released_blocks; 3416 } 3417 3418 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3419 { 3420 struct inode *inode = file_inode(filp); 3421 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3422 pgoff_t page_idx = 0, last_idx; 3423 unsigned int released_blocks = 0; 3424 int ret; 3425 int writecount; 3426 3427 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3428 return -EOPNOTSUPP; 3429 3430 if (!f2fs_compressed_file(inode)) 3431 return -EINVAL; 3432 3433 if (f2fs_readonly(sbi->sb)) 3434 return -EROFS; 3435 3436 ret = mnt_want_write_file(filp); 3437 if (ret) 3438 return ret; 3439 3440 f2fs_balance_fs(F2FS_I_SB(inode), true); 3441 3442 inode_lock(inode); 3443 3444 writecount = atomic_read(&inode->i_writecount); 3445 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3446 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3447 ret = -EBUSY; 3448 goto out; 3449 } 3450 3451 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3452 ret = -EINVAL; 3453 goto out; 3454 } 3455 3456 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3457 if (ret) 3458 goto out; 3459 3460 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3461 inode->i_ctime = current_time(inode); 3462 f2fs_mark_inode_dirty_sync(inode, true); 3463 3464 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3465 goto out; 3466 3467 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3468 filemap_invalidate_lock(inode->i_mapping); 3469 3470 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3471 3472 while (page_idx < last_idx) { 3473 struct dnode_of_data dn; 3474 pgoff_t end_offset, count; 3475 3476 set_new_dnode(&dn, inode, NULL, NULL, 0); 3477 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3478 if (ret) { 3479 if (ret == -ENOENT) { 3480 page_idx = f2fs_get_next_page_offset(&dn, 3481 page_idx); 3482 ret = 0; 3483 continue; 3484 } 3485 break; 3486 } 3487 3488 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3489 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3490 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3491 3492 ret = release_compress_blocks(&dn, count); 3493 3494 f2fs_put_dnode(&dn); 3495 3496 if (ret < 0) 3497 break; 3498 3499 page_idx += count; 3500 released_blocks += ret; 3501 } 3502 3503 filemap_invalidate_unlock(inode->i_mapping); 3504 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3505 out: 3506 inode_unlock(inode); 3507 3508 mnt_drop_write_file(filp); 3509 3510 if (ret >= 0) { 3511 ret = put_user(released_blocks, (u64 __user *)arg); 3512 } else if (released_blocks && 3513 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3514 set_sbi_flag(sbi, SBI_NEED_FSCK); 3515 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3516 "iblocks=%llu, released=%u, compr_blocks=%u, " 3517 "run fsck to fix.", 3518 __func__, inode->i_ino, inode->i_blocks, 3519 released_blocks, 3520 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3521 } 3522 3523 return ret; 3524 } 3525 3526 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3527 { 3528 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3529 unsigned int reserved_blocks = 0; 3530 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3531 block_t blkaddr; 3532 int i; 3533 3534 for (i = 0; i < count; i++) { 3535 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3536 dn->ofs_in_node + i); 3537 3538 if (!__is_valid_data_blkaddr(blkaddr)) 3539 continue; 3540 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3541 DATA_GENERIC_ENHANCE))) 3542 return -EFSCORRUPTED; 3543 } 3544 3545 while (count) { 3546 int compr_blocks = 0; 3547 blkcnt_t reserved; 3548 int ret; 3549 3550 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3551 blkaddr = f2fs_data_blkaddr(dn); 3552 3553 if (i == 0) { 3554 if (blkaddr == COMPRESS_ADDR) 3555 continue; 3556 dn->ofs_in_node += cluster_size; 3557 goto next; 3558 } 3559 3560 if (__is_valid_data_blkaddr(blkaddr)) { 3561 compr_blocks++; 3562 continue; 3563 } 3564 3565 dn->data_blkaddr = NEW_ADDR; 3566 f2fs_set_data_blkaddr(dn); 3567 } 3568 3569 reserved = cluster_size - compr_blocks; 3570 ret = inc_valid_block_count(sbi, dn->inode, &reserved); 3571 if (ret) 3572 return ret; 3573 3574 if (reserved != cluster_size - compr_blocks) 3575 return -ENOSPC; 3576 3577 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3578 3579 reserved_blocks += reserved; 3580 next: 3581 count -= cluster_size; 3582 } 3583 3584 return reserved_blocks; 3585 } 3586 3587 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3588 { 3589 struct inode *inode = file_inode(filp); 3590 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3591 pgoff_t page_idx = 0, last_idx; 3592 unsigned int reserved_blocks = 0; 3593 int ret; 3594 3595 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3596 return -EOPNOTSUPP; 3597 3598 if (!f2fs_compressed_file(inode)) 3599 return -EINVAL; 3600 3601 if (f2fs_readonly(sbi->sb)) 3602 return -EROFS; 3603 3604 ret = mnt_want_write_file(filp); 3605 if (ret) 3606 return ret; 3607 3608 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3609 goto out; 3610 3611 f2fs_balance_fs(F2FS_I_SB(inode), true); 3612 3613 inode_lock(inode); 3614 3615 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3616 ret = -EINVAL; 3617 goto unlock_inode; 3618 } 3619 3620 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3621 filemap_invalidate_lock(inode->i_mapping); 3622 3623 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3624 3625 while (page_idx < last_idx) { 3626 struct dnode_of_data dn; 3627 pgoff_t end_offset, count; 3628 3629 set_new_dnode(&dn, inode, NULL, NULL, 0); 3630 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3631 if (ret) { 3632 if (ret == -ENOENT) { 3633 page_idx = f2fs_get_next_page_offset(&dn, 3634 page_idx); 3635 ret = 0; 3636 continue; 3637 } 3638 break; 3639 } 3640 3641 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3642 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3643 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3644 3645 ret = reserve_compress_blocks(&dn, count); 3646 3647 f2fs_put_dnode(&dn); 3648 3649 if (ret < 0) 3650 break; 3651 3652 page_idx += count; 3653 reserved_blocks += ret; 3654 } 3655 3656 filemap_invalidate_unlock(inode->i_mapping); 3657 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3658 3659 if (ret >= 0) { 3660 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3661 inode->i_ctime = current_time(inode); 3662 f2fs_mark_inode_dirty_sync(inode, true); 3663 } 3664 unlock_inode: 3665 inode_unlock(inode); 3666 out: 3667 mnt_drop_write_file(filp); 3668 3669 if (ret >= 0) { 3670 ret = put_user(reserved_blocks, (u64 __user *)arg); 3671 } else if (reserved_blocks && 3672 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3673 set_sbi_flag(sbi, SBI_NEED_FSCK); 3674 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3675 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3676 "run fsck to fix.", 3677 __func__, inode->i_ino, inode->i_blocks, 3678 reserved_blocks, 3679 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3680 } 3681 3682 return ret; 3683 } 3684 3685 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3686 pgoff_t off, block_t block, block_t len, u32 flags) 3687 { 3688 sector_t sector = SECTOR_FROM_BLOCK(block); 3689 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3690 int ret = 0; 3691 3692 if (flags & F2FS_TRIM_FILE_DISCARD) { 3693 if (bdev_max_secure_erase_sectors(bdev)) 3694 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3695 GFP_NOFS); 3696 else 3697 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3698 GFP_NOFS); 3699 } 3700 3701 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3702 if (IS_ENCRYPTED(inode)) 3703 ret = fscrypt_zeroout_range(inode, off, block, len); 3704 else 3705 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3706 GFP_NOFS, 0); 3707 } 3708 3709 return ret; 3710 } 3711 3712 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3713 { 3714 struct inode *inode = file_inode(filp); 3715 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3716 struct address_space *mapping = inode->i_mapping; 3717 struct block_device *prev_bdev = NULL; 3718 struct f2fs_sectrim_range range; 3719 pgoff_t index, pg_end, prev_index = 0; 3720 block_t prev_block = 0, len = 0; 3721 loff_t end_addr; 3722 bool to_end = false; 3723 int ret = 0; 3724 3725 if (!(filp->f_mode & FMODE_WRITE)) 3726 return -EBADF; 3727 3728 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3729 sizeof(range))) 3730 return -EFAULT; 3731 3732 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3733 !S_ISREG(inode->i_mode)) 3734 return -EINVAL; 3735 3736 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3737 !f2fs_hw_support_discard(sbi)) || 3738 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3739 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3740 return -EOPNOTSUPP; 3741 3742 file_start_write(filp); 3743 inode_lock(inode); 3744 3745 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3746 range.start >= inode->i_size) { 3747 ret = -EINVAL; 3748 goto err; 3749 } 3750 3751 if (range.len == 0) 3752 goto err; 3753 3754 if (inode->i_size - range.start > range.len) { 3755 end_addr = range.start + range.len; 3756 } else { 3757 end_addr = range.len == (u64)-1 ? 3758 sbi->sb->s_maxbytes : inode->i_size; 3759 to_end = true; 3760 } 3761 3762 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3763 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3764 ret = -EINVAL; 3765 goto err; 3766 } 3767 3768 index = F2FS_BYTES_TO_BLK(range.start); 3769 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3770 3771 ret = f2fs_convert_inline_inode(inode); 3772 if (ret) 3773 goto err; 3774 3775 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3776 filemap_invalidate_lock(mapping); 3777 3778 ret = filemap_write_and_wait_range(mapping, range.start, 3779 to_end ? LLONG_MAX : end_addr - 1); 3780 if (ret) 3781 goto out; 3782 3783 truncate_inode_pages_range(mapping, range.start, 3784 to_end ? -1 : end_addr - 1); 3785 3786 while (index < pg_end) { 3787 struct dnode_of_data dn; 3788 pgoff_t end_offset, count; 3789 int i; 3790 3791 set_new_dnode(&dn, inode, NULL, NULL, 0); 3792 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3793 if (ret) { 3794 if (ret == -ENOENT) { 3795 index = f2fs_get_next_page_offset(&dn, index); 3796 continue; 3797 } 3798 goto out; 3799 } 3800 3801 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3802 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3803 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3804 struct block_device *cur_bdev; 3805 block_t blkaddr = f2fs_data_blkaddr(&dn); 3806 3807 if (!__is_valid_data_blkaddr(blkaddr)) 3808 continue; 3809 3810 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3811 DATA_GENERIC_ENHANCE)) { 3812 ret = -EFSCORRUPTED; 3813 f2fs_put_dnode(&dn); 3814 goto out; 3815 } 3816 3817 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3818 if (f2fs_is_multi_device(sbi)) { 3819 int di = f2fs_target_device_index(sbi, blkaddr); 3820 3821 blkaddr -= FDEV(di).start_blk; 3822 } 3823 3824 if (len) { 3825 if (prev_bdev == cur_bdev && 3826 index == prev_index + len && 3827 blkaddr == prev_block + len) { 3828 len++; 3829 } else { 3830 ret = f2fs_secure_erase(prev_bdev, 3831 inode, prev_index, prev_block, 3832 len, range.flags); 3833 if (ret) { 3834 f2fs_put_dnode(&dn); 3835 goto out; 3836 } 3837 3838 len = 0; 3839 } 3840 } 3841 3842 if (!len) { 3843 prev_bdev = cur_bdev; 3844 prev_index = index; 3845 prev_block = blkaddr; 3846 len = 1; 3847 } 3848 } 3849 3850 f2fs_put_dnode(&dn); 3851 3852 if (fatal_signal_pending(current)) { 3853 ret = -EINTR; 3854 goto out; 3855 } 3856 cond_resched(); 3857 } 3858 3859 if (len) 3860 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3861 prev_block, len, range.flags); 3862 out: 3863 filemap_invalidate_unlock(mapping); 3864 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3865 err: 3866 inode_unlock(inode); 3867 file_end_write(filp); 3868 3869 return ret; 3870 } 3871 3872 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3873 { 3874 struct inode *inode = file_inode(filp); 3875 struct f2fs_comp_option option; 3876 3877 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3878 return -EOPNOTSUPP; 3879 3880 inode_lock_shared(inode); 3881 3882 if (!f2fs_compressed_file(inode)) { 3883 inode_unlock_shared(inode); 3884 return -ENODATA; 3885 } 3886 3887 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3888 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3889 3890 inode_unlock_shared(inode); 3891 3892 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3893 sizeof(option))) 3894 return -EFAULT; 3895 3896 return 0; 3897 } 3898 3899 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3900 { 3901 struct inode *inode = file_inode(filp); 3902 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3903 struct f2fs_comp_option option; 3904 int ret = 0; 3905 3906 if (!f2fs_sb_has_compression(sbi)) 3907 return -EOPNOTSUPP; 3908 3909 if (!(filp->f_mode & FMODE_WRITE)) 3910 return -EBADF; 3911 3912 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3913 sizeof(option))) 3914 return -EFAULT; 3915 3916 if (!f2fs_compressed_file(inode) || 3917 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3918 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3919 option.algorithm >= COMPRESS_MAX) 3920 return -EINVAL; 3921 3922 file_start_write(filp); 3923 inode_lock(inode); 3924 3925 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 3926 ret = -EBUSY; 3927 goto out; 3928 } 3929 3930 if (inode->i_size != 0) { 3931 ret = -EFBIG; 3932 goto out; 3933 } 3934 3935 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 3936 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 3937 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size; 3938 f2fs_mark_inode_dirty_sync(inode, true); 3939 3940 if (!f2fs_is_compress_backend_ready(inode)) 3941 f2fs_warn(sbi, "compression algorithm is successfully set, " 3942 "but current kernel doesn't support this algorithm."); 3943 out: 3944 inode_unlock(inode); 3945 file_end_write(filp); 3946 3947 return ret; 3948 } 3949 3950 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 3951 { 3952 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 3953 struct address_space *mapping = inode->i_mapping; 3954 struct page *page; 3955 pgoff_t redirty_idx = page_idx; 3956 int i, page_len = 0, ret = 0; 3957 3958 page_cache_ra_unbounded(&ractl, len, 0); 3959 3960 for (i = 0; i < len; i++, page_idx++) { 3961 page = read_cache_page(mapping, page_idx, NULL, NULL); 3962 if (IS_ERR(page)) { 3963 ret = PTR_ERR(page); 3964 break; 3965 } 3966 page_len++; 3967 } 3968 3969 for (i = 0; i < page_len; i++, redirty_idx++) { 3970 page = find_lock_page(mapping, redirty_idx); 3971 if (!page) { 3972 ret = -ENOMEM; 3973 break; 3974 } 3975 set_page_dirty(page); 3976 f2fs_put_page(page, 1); 3977 f2fs_put_page(page, 0); 3978 } 3979 3980 return ret; 3981 } 3982 3983 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg) 3984 { 3985 struct inode *inode = file_inode(filp); 3986 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3987 struct f2fs_inode_info *fi = F2FS_I(inode); 3988 pgoff_t page_idx = 0, last_idx; 3989 unsigned int blk_per_seg = sbi->blocks_per_seg; 3990 int cluster_size = F2FS_I(inode)->i_cluster_size; 3991 int count, ret; 3992 3993 if (!f2fs_sb_has_compression(sbi) || 3994 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 3995 return -EOPNOTSUPP; 3996 3997 if (!(filp->f_mode & FMODE_WRITE)) 3998 return -EBADF; 3999 4000 if (!f2fs_compressed_file(inode)) 4001 return -EINVAL; 4002 4003 f2fs_balance_fs(F2FS_I_SB(inode), true); 4004 4005 file_start_write(filp); 4006 inode_lock(inode); 4007 4008 if (!f2fs_is_compress_backend_ready(inode)) { 4009 ret = -EOPNOTSUPP; 4010 goto out; 4011 } 4012 4013 if (f2fs_is_mmap_file(inode)) { 4014 ret = -EBUSY; 4015 goto out; 4016 } 4017 4018 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4019 if (ret) 4020 goto out; 4021 4022 if (!atomic_read(&fi->i_compr_blocks)) 4023 goto out; 4024 4025 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4026 4027 count = last_idx - page_idx; 4028 while (count) { 4029 int len = min(cluster_size, count); 4030 4031 ret = redirty_blocks(inode, page_idx, len); 4032 if (ret < 0) 4033 break; 4034 4035 if (get_dirty_pages(inode) >= blk_per_seg) 4036 filemap_fdatawrite(inode->i_mapping); 4037 4038 count -= len; 4039 page_idx += len; 4040 } 4041 4042 if (!ret) 4043 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4044 LLONG_MAX); 4045 4046 if (ret) 4047 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4048 __func__, ret); 4049 out: 4050 inode_unlock(inode); 4051 file_end_write(filp); 4052 4053 return ret; 4054 } 4055 4056 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg) 4057 { 4058 struct inode *inode = file_inode(filp); 4059 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4060 pgoff_t page_idx = 0, last_idx; 4061 unsigned int blk_per_seg = sbi->blocks_per_seg; 4062 int cluster_size = F2FS_I(inode)->i_cluster_size; 4063 int count, ret; 4064 4065 if (!f2fs_sb_has_compression(sbi) || 4066 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4067 return -EOPNOTSUPP; 4068 4069 if (!(filp->f_mode & FMODE_WRITE)) 4070 return -EBADF; 4071 4072 if (!f2fs_compressed_file(inode)) 4073 return -EINVAL; 4074 4075 f2fs_balance_fs(F2FS_I_SB(inode), true); 4076 4077 file_start_write(filp); 4078 inode_lock(inode); 4079 4080 if (!f2fs_is_compress_backend_ready(inode)) { 4081 ret = -EOPNOTSUPP; 4082 goto out; 4083 } 4084 4085 if (f2fs_is_mmap_file(inode)) { 4086 ret = -EBUSY; 4087 goto out; 4088 } 4089 4090 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4091 if (ret) 4092 goto out; 4093 4094 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4095 4096 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4097 4098 count = last_idx - page_idx; 4099 while (count) { 4100 int len = min(cluster_size, count); 4101 4102 ret = redirty_blocks(inode, page_idx, len); 4103 if (ret < 0) 4104 break; 4105 4106 if (get_dirty_pages(inode) >= blk_per_seg) 4107 filemap_fdatawrite(inode->i_mapping); 4108 4109 count -= len; 4110 page_idx += len; 4111 } 4112 4113 if (!ret) 4114 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4115 LLONG_MAX); 4116 4117 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4118 4119 if (ret) 4120 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4121 __func__, ret); 4122 out: 4123 inode_unlock(inode); 4124 file_end_write(filp); 4125 4126 return ret; 4127 } 4128 4129 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4130 { 4131 switch (cmd) { 4132 case FS_IOC_GETVERSION: 4133 return f2fs_ioc_getversion(filp, arg); 4134 case F2FS_IOC_START_ATOMIC_WRITE: 4135 return f2fs_ioc_start_atomic_write(filp); 4136 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4137 return f2fs_ioc_commit_atomic_write(filp); 4138 case F2FS_IOC_START_VOLATILE_WRITE: 4139 return f2fs_ioc_start_volatile_write(filp); 4140 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4141 return f2fs_ioc_release_volatile_write(filp); 4142 case F2FS_IOC_ABORT_VOLATILE_WRITE: 4143 return f2fs_ioc_abort_volatile_write(filp); 4144 case F2FS_IOC_SHUTDOWN: 4145 return f2fs_ioc_shutdown(filp, arg); 4146 case FITRIM: 4147 return f2fs_ioc_fitrim(filp, arg); 4148 case FS_IOC_SET_ENCRYPTION_POLICY: 4149 return f2fs_ioc_set_encryption_policy(filp, arg); 4150 case FS_IOC_GET_ENCRYPTION_POLICY: 4151 return f2fs_ioc_get_encryption_policy(filp, arg); 4152 case FS_IOC_GET_ENCRYPTION_PWSALT: 4153 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4154 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4155 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4156 case FS_IOC_ADD_ENCRYPTION_KEY: 4157 return f2fs_ioc_add_encryption_key(filp, arg); 4158 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4159 return f2fs_ioc_remove_encryption_key(filp, arg); 4160 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4161 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4162 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4163 return f2fs_ioc_get_encryption_key_status(filp, arg); 4164 case FS_IOC_GET_ENCRYPTION_NONCE: 4165 return f2fs_ioc_get_encryption_nonce(filp, arg); 4166 case F2FS_IOC_GARBAGE_COLLECT: 4167 return f2fs_ioc_gc(filp, arg); 4168 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4169 return f2fs_ioc_gc_range(filp, arg); 4170 case F2FS_IOC_WRITE_CHECKPOINT: 4171 return f2fs_ioc_write_checkpoint(filp, arg); 4172 case F2FS_IOC_DEFRAGMENT: 4173 return f2fs_ioc_defragment(filp, arg); 4174 case F2FS_IOC_MOVE_RANGE: 4175 return f2fs_ioc_move_range(filp, arg); 4176 case F2FS_IOC_FLUSH_DEVICE: 4177 return f2fs_ioc_flush_device(filp, arg); 4178 case F2FS_IOC_GET_FEATURES: 4179 return f2fs_ioc_get_features(filp, arg); 4180 case F2FS_IOC_GET_PIN_FILE: 4181 return f2fs_ioc_get_pin_file(filp, arg); 4182 case F2FS_IOC_SET_PIN_FILE: 4183 return f2fs_ioc_set_pin_file(filp, arg); 4184 case F2FS_IOC_PRECACHE_EXTENTS: 4185 return f2fs_ioc_precache_extents(filp, arg); 4186 case F2FS_IOC_RESIZE_FS: 4187 return f2fs_ioc_resize_fs(filp, arg); 4188 case FS_IOC_ENABLE_VERITY: 4189 return f2fs_ioc_enable_verity(filp, arg); 4190 case FS_IOC_MEASURE_VERITY: 4191 return f2fs_ioc_measure_verity(filp, arg); 4192 case FS_IOC_READ_VERITY_METADATA: 4193 return f2fs_ioc_read_verity_metadata(filp, arg); 4194 case FS_IOC_GETFSLABEL: 4195 return f2fs_ioc_getfslabel(filp, arg); 4196 case FS_IOC_SETFSLABEL: 4197 return f2fs_ioc_setfslabel(filp, arg); 4198 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4199 return f2fs_get_compress_blocks(filp, arg); 4200 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4201 return f2fs_release_compress_blocks(filp, arg); 4202 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4203 return f2fs_reserve_compress_blocks(filp, arg); 4204 case F2FS_IOC_SEC_TRIM_FILE: 4205 return f2fs_sec_trim_file(filp, arg); 4206 case F2FS_IOC_GET_COMPRESS_OPTION: 4207 return f2fs_ioc_get_compress_option(filp, arg); 4208 case F2FS_IOC_SET_COMPRESS_OPTION: 4209 return f2fs_ioc_set_compress_option(filp, arg); 4210 case F2FS_IOC_DECOMPRESS_FILE: 4211 return f2fs_ioc_decompress_file(filp, arg); 4212 case F2FS_IOC_COMPRESS_FILE: 4213 return f2fs_ioc_compress_file(filp, arg); 4214 default: 4215 return -ENOTTY; 4216 } 4217 } 4218 4219 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4220 { 4221 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4222 return -EIO; 4223 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4224 return -ENOSPC; 4225 4226 return __f2fs_ioctl(filp, cmd, arg); 4227 } 4228 4229 /* 4230 * Return %true if the given read or write request should use direct I/O, or 4231 * %false if it should use buffered I/O. 4232 */ 4233 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4234 struct iov_iter *iter) 4235 { 4236 unsigned int align; 4237 4238 if (!(iocb->ki_flags & IOCB_DIRECT)) 4239 return false; 4240 4241 if (f2fs_force_buffered_io(inode, iocb, iter)) 4242 return false; 4243 4244 /* 4245 * Direct I/O not aligned to the disk's logical_block_size will be 4246 * attempted, but will fail with -EINVAL. 4247 * 4248 * f2fs additionally requires that direct I/O be aligned to the 4249 * filesystem block size, which is often a stricter requirement. 4250 * However, f2fs traditionally falls back to buffered I/O on requests 4251 * that are logical_block_size-aligned but not fs-block aligned. 4252 * 4253 * The below logic implements this behavior. 4254 */ 4255 align = iocb->ki_pos | iov_iter_alignment(iter); 4256 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4257 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4258 return false; 4259 4260 return true; 4261 } 4262 4263 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4264 unsigned int flags) 4265 { 4266 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4267 4268 dec_page_count(sbi, F2FS_DIO_READ); 4269 if (error) 4270 return error; 4271 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size); 4272 return 0; 4273 } 4274 4275 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4276 .end_io = f2fs_dio_read_end_io, 4277 }; 4278 4279 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4280 { 4281 struct file *file = iocb->ki_filp; 4282 struct inode *inode = file_inode(file); 4283 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4284 struct f2fs_inode_info *fi = F2FS_I(inode); 4285 const loff_t pos = iocb->ki_pos; 4286 const size_t count = iov_iter_count(to); 4287 struct iomap_dio *dio; 4288 ssize_t ret; 4289 4290 if (count == 0) 4291 return 0; /* skip atime update */ 4292 4293 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4294 4295 if (iocb->ki_flags & IOCB_NOWAIT) { 4296 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4297 ret = -EAGAIN; 4298 goto out; 4299 } 4300 } else { 4301 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4302 } 4303 4304 /* 4305 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4306 * the higher-level function iomap_dio_rw() in order to ensure that the 4307 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4308 */ 4309 inc_page_count(sbi, F2FS_DIO_READ); 4310 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4311 &f2fs_iomap_dio_read_ops, 0, 0); 4312 if (IS_ERR_OR_NULL(dio)) { 4313 ret = PTR_ERR_OR_ZERO(dio); 4314 if (ret != -EIOCBQUEUED) 4315 dec_page_count(sbi, F2FS_DIO_READ); 4316 } else { 4317 ret = iomap_dio_complete(dio); 4318 } 4319 4320 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4321 4322 file_accessed(file); 4323 out: 4324 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4325 return ret; 4326 } 4327 4328 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4329 { 4330 struct inode *inode = file_inode(iocb->ki_filp); 4331 ssize_t ret; 4332 4333 if (!f2fs_is_compress_backend_ready(inode)) 4334 return -EOPNOTSUPP; 4335 4336 if (f2fs_should_use_dio(inode, iocb, to)) 4337 return f2fs_dio_read_iter(iocb, to); 4338 4339 ret = filemap_read(iocb, to, 0); 4340 if (ret > 0) 4341 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret); 4342 return ret; 4343 } 4344 4345 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4346 { 4347 struct file *file = iocb->ki_filp; 4348 struct inode *inode = file_inode(file); 4349 ssize_t count; 4350 int err; 4351 4352 if (IS_IMMUTABLE(inode)) 4353 return -EPERM; 4354 4355 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4356 return -EPERM; 4357 4358 count = generic_write_checks(iocb, from); 4359 if (count <= 0) 4360 return count; 4361 4362 err = file_modified(file); 4363 if (err) 4364 return err; 4365 return count; 4366 } 4367 4368 /* 4369 * Preallocate blocks for a write request, if it is possible and helpful to do 4370 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4371 * blocks were preallocated, or a negative errno value if something went 4372 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4373 * requested blocks (not just some of them) have been allocated. 4374 */ 4375 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4376 bool dio) 4377 { 4378 struct inode *inode = file_inode(iocb->ki_filp); 4379 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4380 const loff_t pos = iocb->ki_pos; 4381 const size_t count = iov_iter_count(iter); 4382 struct f2fs_map_blocks map = {}; 4383 int flag; 4384 int ret; 4385 4386 /* If it will be an out-of-place direct write, don't bother. */ 4387 if (dio && f2fs_lfs_mode(sbi)) 4388 return 0; 4389 /* 4390 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4391 * buffered IO, if DIO meets any holes. 4392 */ 4393 if (dio && i_size_read(inode) && 4394 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4395 return 0; 4396 4397 /* No-wait I/O can't allocate blocks. */ 4398 if (iocb->ki_flags & IOCB_NOWAIT) 4399 return 0; 4400 4401 /* If it will be a short write, don't bother. */ 4402 if (fault_in_iov_iter_readable(iter, count)) 4403 return 0; 4404 4405 if (f2fs_has_inline_data(inode)) { 4406 /* If the data will fit inline, don't bother. */ 4407 if (pos + count <= MAX_INLINE_DATA(inode)) 4408 return 0; 4409 ret = f2fs_convert_inline_inode(inode); 4410 if (ret) 4411 return ret; 4412 } 4413 4414 /* Do not preallocate blocks that will be written partially in 4KB. */ 4415 map.m_lblk = F2FS_BLK_ALIGN(pos); 4416 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4417 if (map.m_len > map.m_lblk) 4418 map.m_len -= map.m_lblk; 4419 else 4420 map.m_len = 0; 4421 map.m_may_create = true; 4422 if (dio) { 4423 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4424 flag = F2FS_GET_BLOCK_PRE_DIO; 4425 } else { 4426 map.m_seg_type = NO_CHECK_TYPE; 4427 flag = F2FS_GET_BLOCK_PRE_AIO; 4428 } 4429 4430 ret = f2fs_map_blocks(inode, &map, 1, flag); 4431 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4432 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4433 return ret; 4434 if (ret == 0) 4435 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4436 return map.m_len; 4437 } 4438 4439 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4440 struct iov_iter *from) 4441 { 4442 struct file *file = iocb->ki_filp; 4443 struct inode *inode = file_inode(file); 4444 ssize_t ret; 4445 4446 if (iocb->ki_flags & IOCB_NOWAIT) 4447 return -EOPNOTSUPP; 4448 4449 current->backing_dev_info = inode_to_bdi(inode); 4450 ret = generic_perform_write(iocb, from); 4451 current->backing_dev_info = NULL; 4452 4453 if (ret > 0) { 4454 iocb->ki_pos += ret; 4455 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret); 4456 } 4457 return ret; 4458 } 4459 4460 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4461 unsigned int flags) 4462 { 4463 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4464 4465 dec_page_count(sbi, F2FS_DIO_WRITE); 4466 if (error) 4467 return error; 4468 f2fs_update_iostat(sbi, APP_DIRECT_IO, size); 4469 return 0; 4470 } 4471 4472 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4473 .end_io = f2fs_dio_write_end_io, 4474 }; 4475 4476 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4477 bool *may_need_sync) 4478 { 4479 struct file *file = iocb->ki_filp; 4480 struct inode *inode = file_inode(file); 4481 struct f2fs_inode_info *fi = F2FS_I(inode); 4482 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4483 const bool do_opu = f2fs_lfs_mode(sbi); 4484 const loff_t pos = iocb->ki_pos; 4485 const ssize_t count = iov_iter_count(from); 4486 unsigned int dio_flags; 4487 struct iomap_dio *dio; 4488 ssize_t ret; 4489 4490 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4491 4492 if (iocb->ki_flags & IOCB_NOWAIT) { 4493 /* f2fs_convert_inline_inode() and block allocation can block */ 4494 if (f2fs_has_inline_data(inode) || 4495 !f2fs_overwrite_io(inode, pos, count)) { 4496 ret = -EAGAIN; 4497 goto out; 4498 } 4499 4500 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4501 ret = -EAGAIN; 4502 goto out; 4503 } 4504 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4505 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4506 ret = -EAGAIN; 4507 goto out; 4508 } 4509 } else { 4510 ret = f2fs_convert_inline_inode(inode); 4511 if (ret) 4512 goto out; 4513 4514 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4515 if (do_opu) 4516 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4517 } 4518 4519 /* 4520 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4521 * the higher-level function iomap_dio_rw() in order to ensure that the 4522 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4523 */ 4524 inc_page_count(sbi, F2FS_DIO_WRITE); 4525 dio_flags = 0; 4526 if (pos + count > inode->i_size) 4527 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4528 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4529 &f2fs_iomap_dio_write_ops, dio_flags, 0); 4530 if (IS_ERR_OR_NULL(dio)) { 4531 ret = PTR_ERR_OR_ZERO(dio); 4532 if (ret == -ENOTBLK) 4533 ret = 0; 4534 if (ret != -EIOCBQUEUED) 4535 dec_page_count(sbi, F2FS_DIO_WRITE); 4536 } else { 4537 ret = iomap_dio_complete(dio); 4538 } 4539 4540 if (do_opu) 4541 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4542 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4543 4544 if (ret < 0) 4545 goto out; 4546 if (pos + ret > inode->i_size) 4547 f2fs_i_size_write(inode, pos + ret); 4548 if (!do_opu) 4549 set_inode_flag(inode, FI_UPDATE_WRITE); 4550 4551 if (iov_iter_count(from)) { 4552 ssize_t ret2; 4553 loff_t bufio_start_pos = iocb->ki_pos; 4554 4555 /* 4556 * The direct write was partial, so we need to fall back to a 4557 * buffered write for the remainder. 4558 */ 4559 4560 ret2 = f2fs_buffered_write_iter(iocb, from); 4561 if (iov_iter_count(from)) 4562 f2fs_write_failed(inode, iocb->ki_pos); 4563 if (ret2 < 0) 4564 goto out; 4565 4566 /* 4567 * Ensure that the pagecache pages are written to disk and 4568 * invalidated to preserve the expected O_DIRECT semantics. 4569 */ 4570 if (ret2 > 0) { 4571 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4572 4573 ret += ret2; 4574 4575 ret2 = filemap_write_and_wait_range(file->f_mapping, 4576 bufio_start_pos, 4577 bufio_end_pos); 4578 if (ret2 < 0) 4579 goto out; 4580 invalidate_mapping_pages(file->f_mapping, 4581 bufio_start_pos >> PAGE_SHIFT, 4582 bufio_end_pos >> PAGE_SHIFT); 4583 } 4584 } else { 4585 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4586 *may_need_sync = false; 4587 } 4588 out: 4589 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4590 return ret; 4591 } 4592 4593 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4594 { 4595 struct inode *inode = file_inode(iocb->ki_filp); 4596 const loff_t orig_pos = iocb->ki_pos; 4597 const size_t orig_count = iov_iter_count(from); 4598 loff_t target_size; 4599 bool dio; 4600 bool may_need_sync = true; 4601 int preallocated; 4602 ssize_t ret; 4603 4604 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4605 ret = -EIO; 4606 goto out; 4607 } 4608 4609 if (!f2fs_is_compress_backend_ready(inode)) { 4610 ret = -EOPNOTSUPP; 4611 goto out; 4612 } 4613 4614 if (iocb->ki_flags & IOCB_NOWAIT) { 4615 if (!inode_trylock(inode)) { 4616 ret = -EAGAIN; 4617 goto out; 4618 } 4619 } else { 4620 inode_lock(inode); 4621 } 4622 4623 ret = f2fs_write_checks(iocb, from); 4624 if (ret <= 0) 4625 goto out_unlock; 4626 4627 /* Determine whether we will do a direct write or a buffered write. */ 4628 dio = f2fs_should_use_dio(inode, iocb, from); 4629 4630 /* Possibly preallocate the blocks for the write. */ 4631 target_size = iocb->ki_pos + iov_iter_count(from); 4632 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4633 if (preallocated < 0) 4634 ret = preallocated; 4635 else 4636 /* Do the actual write. */ 4637 ret = dio ? 4638 f2fs_dio_write_iter(iocb, from, &may_need_sync): 4639 f2fs_buffered_write_iter(iocb, from); 4640 4641 /* Don't leave any preallocated blocks around past i_size. */ 4642 if (preallocated && i_size_read(inode) < target_size) { 4643 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4644 filemap_invalidate_lock(inode->i_mapping); 4645 if (!f2fs_truncate(inode)) 4646 file_dont_truncate(inode); 4647 filemap_invalidate_unlock(inode->i_mapping); 4648 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4649 } else { 4650 file_dont_truncate(inode); 4651 } 4652 4653 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4654 out_unlock: 4655 inode_unlock(inode); 4656 out: 4657 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4658 if (ret > 0 && may_need_sync) 4659 ret = generic_write_sync(iocb, ret); 4660 return ret; 4661 } 4662 4663 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4664 int advice) 4665 { 4666 struct address_space *mapping; 4667 struct backing_dev_info *bdi; 4668 struct inode *inode = file_inode(filp); 4669 int err; 4670 4671 if (advice == POSIX_FADV_SEQUENTIAL) { 4672 if (S_ISFIFO(inode->i_mode)) 4673 return -ESPIPE; 4674 4675 mapping = filp->f_mapping; 4676 if (!mapping || len < 0) 4677 return -EINVAL; 4678 4679 bdi = inode_to_bdi(mapping->host); 4680 filp->f_ra.ra_pages = bdi->ra_pages * 4681 F2FS_I_SB(inode)->seq_file_ra_mul; 4682 spin_lock(&filp->f_lock); 4683 filp->f_mode &= ~FMODE_RANDOM; 4684 spin_unlock(&filp->f_lock); 4685 return 0; 4686 } 4687 4688 err = generic_fadvise(filp, offset, len, advice); 4689 if (!err && advice == POSIX_FADV_DONTNEED && 4690 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4691 f2fs_compressed_file(inode)) 4692 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4693 4694 return err; 4695 } 4696 4697 #ifdef CONFIG_COMPAT 4698 struct compat_f2fs_gc_range { 4699 u32 sync; 4700 compat_u64 start; 4701 compat_u64 len; 4702 }; 4703 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4704 struct compat_f2fs_gc_range) 4705 4706 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4707 { 4708 struct compat_f2fs_gc_range __user *urange; 4709 struct f2fs_gc_range range; 4710 int err; 4711 4712 urange = compat_ptr(arg); 4713 err = get_user(range.sync, &urange->sync); 4714 err |= get_user(range.start, &urange->start); 4715 err |= get_user(range.len, &urange->len); 4716 if (err) 4717 return -EFAULT; 4718 4719 return __f2fs_ioc_gc_range(file, &range); 4720 } 4721 4722 struct compat_f2fs_move_range { 4723 u32 dst_fd; 4724 compat_u64 pos_in; 4725 compat_u64 pos_out; 4726 compat_u64 len; 4727 }; 4728 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4729 struct compat_f2fs_move_range) 4730 4731 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4732 { 4733 struct compat_f2fs_move_range __user *urange; 4734 struct f2fs_move_range range; 4735 int err; 4736 4737 urange = compat_ptr(arg); 4738 err = get_user(range.dst_fd, &urange->dst_fd); 4739 err |= get_user(range.pos_in, &urange->pos_in); 4740 err |= get_user(range.pos_out, &urange->pos_out); 4741 err |= get_user(range.len, &urange->len); 4742 if (err) 4743 return -EFAULT; 4744 4745 return __f2fs_ioc_move_range(file, &range); 4746 } 4747 4748 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4749 { 4750 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4751 return -EIO; 4752 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4753 return -ENOSPC; 4754 4755 switch (cmd) { 4756 case FS_IOC32_GETVERSION: 4757 cmd = FS_IOC_GETVERSION; 4758 break; 4759 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4760 return f2fs_compat_ioc_gc_range(file, arg); 4761 case F2FS_IOC32_MOVE_RANGE: 4762 return f2fs_compat_ioc_move_range(file, arg); 4763 case F2FS_IOC_START_ATOMIC_WRITE: 4764 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4765 case F2FS_IOC_START_VOLATILE_WRITE: 4766 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4767 case F2FS_IOC_ABORT_VOLATILE_WRITE: 4768 case F2FS_IOC_SHUTDOWN: 4769 case FITRIM: 4770 case FS_IOC_SET_ENCRYPTION_POLICY: 4771 case FS_IOC_GET_ENCRYPTION_PWSALT: 4772 case FS_IOC_GET_ENCRYPTION_POLICY: 4773 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4774 case FS_IOC_ADD_ENCRYPTION_KEY: 4775 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4776 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4777 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4778 case FS_IOC_GET_ENCRYPTION_NONCE: 4779 case F2FS_IOC_GARBAGE_COLLECT: 4780 case F2FS_IOC_WRITE_CHECKPOINT: 4781 case F2FS_IOC_DEFRAGMENT: 4782 case F2FS_IOC_FLUSH_DEVICE: 4783 case F2FS_IOC_GET_FEATURES: 4784 case F2FS_IOC_GET_PIN_FILE: 4785 case F2FS_IOC_SET_PIN_FILE: 4786 case F2FS_IOC_PRECACHE_EXTENTS: 4787 case F2FS_IOC_RESIZE_FS: 4788 case FS_IOC_ENABLE_VERITY: 4789 case FS_IOC_MEASURE_VERITY: 4790 case FS_IOC_READ_VERITY_METADATA: 4791 case FS_IOC_GETFSLABEL: 4792 case FS_IOC_SETFSLABEL: 4793 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4794 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4795 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4796 case F2FS_IOC_SEC_TRIM_FILE: 4797 case F2FS_IOC_GET_COMPRESS_OPTION: 4798 case F2FS_IOC_SET_COMPRESS_OPTION: 4799 case F2FS_IOC_DECOMPRESS_FILE: 4800 case F2FS_IOC_COMPRESS_FILE: 4801 break; 4802 default: 4803 return -ENOIOCTLCMD; 4804 } 4805 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4806 } 4807 #endif 4808 4809 const struct file_operations f2fs_file_operations = { 4810 .llseek = f2fs_llseek, 4811 .read_iter = f2fs_file_read_iter, 4812 .write_iter = f2fs_file_write_iter, 4813 .open = f2fs_file_open, 4814 .release = f2fs_release_file, 4815 .mmap = f2fs_file_mmap, 4816 .flush = f2fs_file_flush, 4817 .fsync = f2fs_sync_file, 4818 .fallocate = f2fs_fallocate, 4819 .unlocked_ioctl = f2fs_ioctl, 4820 #ifdef CONFIG_COMPAT 4821 .compat_ioctl = f2fs_compat_ioctl, 4822 #endif 4823 .splice_read = generic_file_splice_read, 4824 .splice_write = iter_file_splice_write, 4825 .fadvise = f2fs_file_fadvise, 4826 }; 4827