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 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 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 down_read(&F2FS_I(inode)->i_sem); 322 cp_reason = need_do_checkpoint(inode); 323 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; 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(&init_user_ns, 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(&init_user_ns, 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 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 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(&init_user_ns, inode, attr); 984 985 if (attr->ia_valid & ATTR_MODE) { 986 err = posix_acl_chmod(&init_user_ns, 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 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 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); 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 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 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 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 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 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 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 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 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 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 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 if (f2fs_compressed_file(inode) && 1754 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1755 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1756 return -EOPNOTSUPP; 1757 1758 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1759 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1760 FALLOC_FL_INSERT_RANGE)) 1761 return -EOPNOTSUPP; 1762 1763 inode_lock(inode); 1764 1765 if (mode & FALLOC_FL_PUNCH_HOLE) { 1766 if (offset >= inode->i_size) 1767 goto out; 1768 1769 ret = punch_hole(inode, offset, len); 1770 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1771 ret = f2fs_collapse_range(inode, offset, len); 1772 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1773 ret = f2fs_zero_range(inode, offset, len, mode); 1774 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1775 ret = f2fs_insert_range(inode, offset, len); 1776 } else { 1777 ret = expand_inode_data(inode, offset, len, mode); 1778 } 1779 1780 if (!ret) { 1781 inode->i_mtime = inode->i_ctime = current_time(inode); 1782 f2fs_mark_inode_dirty_sync(inode, false); 1783 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1784 } 1785 1786 out: 1787 inode_unlock(inode); 1788 1789 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1790 return ret; 1791 } 1792 1793 static int f2fs_release_file(struct inode *inode, struct file *filp) 1794 { 1795 /* 1796 * f2fs_relase_file is called at every close calls. So we should 1797 * not drop any inmemory pages by close called by other process. 1798 */ 1799 if (!(filp->f_mode & FMODE_WRITE) || 1800 atomic_read(&inode->i_writecount) != 1) 1801 return 0; 1802 1803 /* some remained atomic pages should discarded */ 1804 if (f2fs_is_atomic_file(inode)) 1805 f2fs_drop_inmem_pages(inode); 1806 if (f2fs_is_volatile_file(inode)) { 1807 set_inode_flag(inode, FI_DROP_CACHE); 1808 filemap_fdatawrite(inode->i_mapping); 1809 clear_inode_flag(inode, FI_DROP_CACHE); 1810 clear_inode_flag(inode, FI_VOLATILE_FILE); 1811 stat_dec_volatile_write(inode); 1812 } 1813 return 0; 1814 } 1815 1816 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1817 { 1818 struct inode *inode = file_inode(file); 1819 1820 /* 1821 * If the process doing a transaction is crashed, we should do 1822 * roll-back. Otherwise, other reader/write can see corrupted database 1823 * until all the writers close its file. Since this should be done 1824 * before dropping file lock, it needs to do in ->flush. 1825 */ 1826 if (f2fs_is_atomic_file(inode) && 1827 F2FS_I(inode)->inmem_task == current) 1828 f2fs_drop_inmem_pages(inode); 1829 return 0; 1830 } 1831 1832 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1833 { 1834 struct f2fs_inode_info *fi = F2FS_I(inode); 1835 u32 masked_flags = fi->i_flags & mask; 1836 1837 /* mask can be shrunk by flags_valid selector */ 1838 iflags &= mask; 1839 1840 /* Is it quota file? Do not allow user to mess with it */ 1841 if (IS_NOQUOTA(inode)) 1842 return -EPERM; 1843 1844 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1845 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1846 return -EOPNOTSUPP; 1847 if (!f2fs_empty_dir(inode)) 1848 return -ENOTEMPTY; 1849 } 1850 1851 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1852 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1853 return -EOPNOTSUPP; 1854 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1855 return -EINVAL; 1856 } 1857 1858 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1859 if (masked_flags & F2FS_COMPR_FL) { 1860 if (!f2fs_disable_compressed_file(inode)) 1861 return -EINVAL; 1862 } 1863 if (iflags & F2FS_NOCOMP_FL) 1864 return -EINVAL; 1865 if (iflags & F2FS_COMPR_FL) { 1866 if (!f2fs_may_compress(inode)) 1867 return -EINVAL; 1868 if (S_ISREG(inode->i_mode) && inode->i_size) 1869 return -EINVAL; 1870 1871 set_compress_context(inode); 1872 } 1873 } 1874 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) { 1875 if (masked_flags & F2FS_COMPR_FL) 1876 return -EINVAL; 1877 } 1878 1879 fi->i_flags = iflags | (fi->i_flags & ~mask); 1880 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1881 (fi->i_flags & F2FS_NOCOMP_FL)); 1882 1883 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1884 set_inode_flag(inode, FI_PROJ_INHERIT); 1885 else 1886 clear_inode_flag(inode, FI_PROJ_INHERIT); 1887 1888 inode->i_ctime = current_time(inode); 1889 f2fs_set_inode_flags(inode); 1890 f2fs_mark_inode_dirty_sync(inode, true); 1891 return 0; 1892 } 1893 1894 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1895 1896 /* 1897 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1898 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1899 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1900 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1901 * 1902 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1903 * FS_IOC_FSSETXATTR is done by the VFS. 1904 */ 1905 1906 static const struct { 1907 u32 iflag; 1908 u32 fsflag; 1909 } f2fs_fsflags_map[] = { 1910 { F2FS_COMPR_FL, FS_COMPR_FL }, 1911 { F2FS_SYNC_FL, FS_SYNC_FL }, 1912 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1913 { F2FS_APPEND_FL, FS_APPEND_FL }, 1914 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1915 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1916 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1917 { F2FS_INDEX_FL, FS_INDEX_FL }, 1918 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1919 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1920 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1921 }; 1922 1923 #define F2FS_GETTABLE_FS_FL ( \ 1924 FS_COMPR_FL | \ 1925 FS_SYNC_FL | \ 1926 FS_IMMUTABLE_FL | \ 1927 FS_APPEND_FL | \ 1928 FS_NODUMP_FL | \ 1929 FS_NOATIME_FL | \ 1930 FS_NOCOMP_FL | \ 1931 FS_INDEX_FL | \ 1932 FS_DIRSYNC_FL | \ 1933 FS_PROJINHERIT_FL | \ 1934 FS_ENCRYPT_FL | \ 1935 FS_INLINE_DATA_FL | \ 1936 FS_NOCOW_FL | \ 1937 FS_VERITY_FL | \ 1938 FS_CASEFOLD_FL) 1939 1940 #define F2FS_SETTABLE_FS_FL ( \ 1941 FS_COMPR_FL | \ 1942 FS_SYNC_FL | \ 1943 FS_IMMUTABLE_FL | \ 1944 FS_APPEND_FL | \ 1945 FS_NODUMP_FL | \ 1946 FS_NOATIME_FL | \ 1947 FS_NOCOMP_FL | \ 1948 FS_DIRSYNC_FL | \ 1949 FS_PROJINHERIT_FL | \ 1950 FS_CASEFOLD_FL) 1951 1952 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 1953 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 1954 { 1955 u32 fsflags = 0; 1956 int i; 1957 1958 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1959 if (iflags & f2fs_fsflags_map[i].iflag) 1960 fsflags |= f2fs_fsflags_map[i].fsflag; 1961 1962 return fsflags; 1963 } 1964 1965 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 1966 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 1967 { 1968 u32 iflags = 0; 1969 int i; 1970 1971 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1972 if (fsflags & f2fs_fsflags_map[i].fsflag) 1973 iflags |= f2fs_fsflags_map[i].iflag; 1974 1975 return iflags; 1976 } 1977 1978 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1979 { 1980 struct inode *inode = file_inode(filp); 1981 1982 return put_user(inode->i_generation, (int __user *)arg); 1983 } 1984 1985 static int f2fs_ioc_start_atomic_write(struct file *filp) 1986 { 1987 struct inode *inode = file_inode(filp); 1988 struct f2fs_inode_info *fi = F2FS_I(inode); 1989 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1990 int ret; 1991 1992 if (!inode_owner_or_capable(&init_user_ns, inode)) 1993 return -EACCES; 1994 1995 if (!S_ISREG(inode->i_mode)) 1996 return -EINVAL; 1997 1998 if (filp->f_flags & O_DIRECT) 1999 return -EINVAL; 2000 2001 ret = mnt_want_write_file(filp); 2002 if (ret) 2003 return ret; 2004 2005 inode_lock(inode); 2006 2007 f2fs_disable_compressed_file(inode); 2008 2009 if (f2fs_is_atomic_file(inode)) { 2010 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) 2011 ret = -EINVAL; 2012 goto out; 2013 } 2014 2015 ret = f2fs_convert_inline_inode(inode); 2016 if (ret) 2017 goto out; 2018 2019 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2020 2021 /* 2022 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2023 * f2fs_is_atomic_file. 2024 */ 2025 if (get_dirty_pages(inode)) 2026 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2027 inode->i_ino, get_dirty_pages(inode)); 2028 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2029 if (ret) { 2030 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2031 goto out; 2032 } 2033 2034 spin_lock(&sbi->inode_lock[ATOMIC_FILE]); 2035 if (list_empty(&fi->inmem_ilist)) 2036 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]); 2037 sbi->atomic_files++; 2038 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]); 2039 2040 /* add inode in inmem_list first and set atomic_file */ 2041 set_inode_flag(inode, FI_ATOMIC_FILE); 2042 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2043 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2044 2045 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2046 F2FS_I(inode)->inmem_task = current; 2047 stat_update_max_atomic_write(inode); 2048 out: 2049 inode_unlock(inode); 2050 mnt_drop_write_file(filp); 2051 return ret; 2052 } 2053 2054 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2055 { 2056 struct inode *inode = file_inode(filp); 2057 int ret; 2058 2059 if (!inode_owner_or_capable(&init_user_ns, inode)) 2060 return -EACCES; 2061 2062 ret = mnt_want_write_file(filp); 2063 if (ret) 2064 return ret; 2065 2066 f2fs_balance_fs(F2FS_I_SB(inode), true); 2067 2068 inode_lock(inode); 2069 2070 if (f2fs_is_volatile_file(inode)) { 2071 ret = -EINVAL; 2072 goto err_out; 2073 } 2074 2075 if (f2fs_is_atomic_file(inode)) { 2076 ret = f2fs_commit_inmem_pages(inode); 2077 if (ret) 2078 goto err_out; 2079 2080 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2081 if (!ret) 2082 f2fs_drop_inmem_pages(inode); 2083 } else { 2084 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2085 } 2086 err_out: 2087 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) { 2088 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2089 ret = -EINVAL; 2090 } 2091 inode_unlock(inode); 2092 mnt_drop_write_file(filp); 2093 return ret; 2094 } 2095 2096 static int f2fs_ioc_start_volatile_write(struct file *filp) 2097 { 2098 struct inode *inode = file_inode(filp); 2099 int ret; 2100 2101 if (!inode_owner_or_capable(&init_user_ns, inode)) 2102 return -EACCES; 2103 2104 if (!S_ISREG(inode->i_mode)) 2105 return -EINVAL; 2106 2107 ret = mnt_want_write_file(filp); 2108 if (ret) 2109 return ret; 2110 2111 inode_lock(inode); 2112 2113 if (f2fs_is_volatile_file(inode)) 2114 goto out; 2115 2116 ret = f2fs_convert_inline_inode(inode); 2117 if (ret) 2118 goto out; 2119 2120 stat_inc_volatile_write(inode); 2121 stat_update_max_volatile_write(inode); 2122 2123 set_inode_flag(inode, FI_VOLATILE_FILE); 2124 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2125 out: 2126 inode_unlock(inode); 2127 mnt_drop_write_file(filp); 2128 return ret; 2129 } 2130 2131 static int f2fs_ioc_release_volatile_write(struct file *filp) 2132 { 2133 struct inode *inode = file_inode(filp); 2134 int ret; 2135 2136 if (!inode_owner_or_capable(&init_user_ns, inode)) 2137 return -EACCES; 2138 2139 ret = mnt_want_write_file(filp); 2140 if (ret) 2141 return ret; 2142 2143 inode_lock(inode); 2144 2145 if (!f2fs_is_volatile_file(inode)) 2146 goto out; 2147 2148 if (!f2fs_is_first_block_written(inode)) { 2149 ret = truncate_partial_data_page(inode, 0, true); 2150 goto out; 2151 } 2152 2153 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 2154 out: 2155 inode_unlock(inode); 2156 mnt_drop_write_file(filp); 2157 return ret; 2158 } 2159 2160 static int f2fs_ioc_abort_volatile_write(struct file *filp) 2161 { 2162 struct inode *inode = file_inode(filp); 2163 int ret; 2164 2165 if (!inode_owner_or_capable(&init_user_ns, inode)) 2166 return -EACCES; 2167 2168 ret = mnt_want_write_file(filp); 2169 if (ret) 2170 return ret; 2171 2172 inode_lock(inode); 2173 2174 if (f2fs_is_atomic_file(inode)) 2175 f2fs_drop_inmem_pages(inode); 2176 if (f2fs_is_volatile_file(inode)) { 2177 clear_inode_flag(inode, FI_VOLATILE_FILE); 2178 stat_dec_volatile_write(inode); 2179 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2180 } 2181 2182 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2183 2184 inode_unlock(inode); 2185 2186 mnt_drop_write_file(filp); 2187 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2188 return ret; 2189 } 2190 2191 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2192 { 2193 struct inode *inode = file_inode(filp); 2194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2195 struct super_block *sb = sbi->sb; 2196 __u32 in; 2197 int ret = 0; 2198 2199 if (!capable(CAP_SYS_ADMIN)) 2200 return -EPERM; 2201 2202 if (get_user(in, (__u32 __user *)arg)) 2203 return -EFAULT; 2204 2205 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2206 ret = mnt_want_write_file(filp); 2207 if (ret) { 2208 if (ret == -EROFS) { 2209 ret = 0; 2210 f2fs_stop_checkpoint(sbi, false); 2211 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2212 trace_f2fs_shutdown(sbi, in, ret); 2213 } 2214 return ret; 2215 } 2216 } 2217 2218 switch (in) { 2219 case F2FS_GOING_DOWN_FULLSYNC: 2220 ret = freeze_bdev(sb->s_bdev); 2221 if (ret) 2222 goto out; 2223 f2fs_stop_checkpoint(sbi, false); 2224 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2225 thaw_bdev(sb->s_bdev); 2226 break; 2227 case F2FS_GOING_DOWN_METASYNC: 2228 /* do checkpoint only */ 2229 ret = f2fs_sync_fs(sb, 1); 2230 if (ret) 2231 goto out; 2232 f2fs_stop_checkpoint(sbi, false); 2233 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2234 break; 2235 case F2FS_GOING_DOWN_NOSYNC: 2236 f2fs_stop_checkpoint(sbi, false); 2237 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2238 break; 2239 case F2FS_GOING_DOWN_METAFLUSH: 2240 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2241 f2fs_stop_checkpoint(sbi, false); 2242 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2243 break; 2244 case F2FS_GOING_DOWN_NEED_FSCK: 2245 set_sbi_flag(sbi, SBI_NEED_FSCK); 2246 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2247 set_sbi_flag(sbi, SBI_IS_DIRTY); 2248 /* do checkpoint only */ 2249 ret = f2fs_sync_fs(sb, 1); 2250 goto out; 2251 default: 2252 ret = -EINVAL; 2253 goto out; 2254 } 2255 2256 f2fs_stop_gc_thread(sbi); 2257 f2fs_stop_discard_thread(sbi); 2258 2259 f2fs_drop_discard_cmd(sbi); 2260 clear_opt(sbi, DISCARD); 2261 2262 f2fs_update_time(sbi, REQ_TIME); 2263 out: 2264 if (in != F2FS_GOING_DOWN_FULLSYNC) 2265 mnt_drop_write_file(filp); 2266 2267 trace_f2fs_shutdown(sbi, in, ret); 2268 2269 return ret; 2270 } 2271 2272 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2273 { 2274 struct inode *inode = file_inode(filp); 2275 struct super_block *sb = inode->i_sb; 2276 struct request_queue *q = bdev_get_queue(sb->s_bdev); 2277 struct fstrim_range range; 2278 int ret; 2279 2280 if (!capable(CAP_SYS_ADMIN)) 2281 return -EPERM; 2282 2283 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2284 return -EOPNOTSUPP; 2285 2286 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2287 sizeof(range))) 2288 return -EFAULT; 2289 2290 ret = mnt_want_write_file(filp); 2291 if (ret) 2292 return ret; 2293 2294 range.minlen = max((unsigned int)range.minlen, 2295 q->limits.discard_granularity); 2296 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2297 mnt_drop_write_file(filp); 2298 if (ret < 0) 2299 return ret; 2300 2301 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2302 sizeof(range))) 2303 return -EFAULT; 2304 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2305 return 0; 2306 } 2307 2308 static bool uuid_is_nonzero(__u8 u[16]) 2309 { 2310 int i; 2311 2312 for (i = 0; i < 16; i++) 2313 if (u[i]) 2314 return true; 2315 return false; 2316 } 2317 2318 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2319 { 2320 struct inode *inode = file_inode(filp); 2321 2322 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2323 return -EOPNOTSUPP; 2324 2325 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2326 2327 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2328 } 2329 2330 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2331 { 2332 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2333 return -EOPNOTSUPP; 2334 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2335 } 2336 2337 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2338 { 2339 struct inode *inode = file_inode(filp); 2340 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2341 int err; 2342 2343 if (!f2fs_sb_has_encrypt(sbi)) 2344 return -EOPNOTSUPP; 2345 2346 err = mnt_want_write_file(filp); 2347 if (err) 2348 return err; 2349 2350 down_write(&sbi->sb_lock); 2351 2352 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2353 goto got_it; 2354 2355 /* update superblock with uuid */ 2356 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2357 2358 err = f2fs_commit_super(sbi, false); 2359 if (err) { 2360 /* undo new data */ 2361 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2362 goto out_err; 2363 } 2364 got_it: 2365 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 2366 16)) 2367 err = -EFAULT; 2368 out_err: 2369 up_write(&sbi->sb_lock); 2370 mnt_drop_write_file(filp); 2371 return err; 2372 } 2373 2374 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2375 unsigned long arg) 2376 { 2377 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2378 return -EOPNOTSUPP; 2379 2380 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2381 } 2382 2383 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2384 { 2385 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2386 return -EOPNOTSUPP; 2387 2388 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2389 } 2390 2391 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2392 { 2393 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2394 return -EOPNOTSUPP; 2395 2396 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2397 } 2398 2399 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2400 unsigned long arg) 2401 { 2402 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2403 return -EOPNOTSUPP; 2404 2405 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2406 } 2407 2408 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2409 unsigned long arg) 2410 { 2411 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2412 return -EOPNOTSUPP; 2413 2414 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2415 } 2416 2417 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2418 { 2419 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2420 return -EOPNOTSUPP; 2421 2422 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2423 } 2424 2425 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2426 { 2427 struct inode *inode = file_inode(filp); 2428 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2429 __u32 sync; 2430 int ret; 2431 2432 if (!capable(CAP_SYS_ADMIN)) 2433 return -EPERM; 2434 2435 if (get_user(sync, (__u32 __user *)arg)) 2436 return -EFAULT; 2437 2438 if (f2fs_readonly(sbi->sb)) 2439 return -EROFS; 2440 2441 ret = mnt_want_write_file(filp); 2442 if (ret) 2443 return ret; 2444 2445 if (!sync) { 2446 if (!down_write_trylock(&sbi->gc_lock)) { 2447 ret = -EBUSY; 2448 goto out; 2449 } 2450 } else { 2451 down_write(&sbi->gc_lock); 2452 } 2453 2454 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO); 2455 out: 2456 mnt_drop_write_file(filp); 2457 return ret; 2458 } 2459 2460 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2461 { 2462 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2463 u64 end; 2464 int ret; 2465 2466 if (!capable(CAP_SYS_ADMIN)) 2467 return -EPERM; 2468 if (f2fs_readonly(sbi->sb)) 2469 return -EROFS; 2470 2471 end = range->start + range->len; 2472 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2473 end >= MAX_BLKADDR(sbi)) 2474 return -EINVAL; 2475 2476 ret = mnt_want_write_file(filp); 2477 if (ret) 2478 return ret; 2479 2480 do_more: 2481 if (!range->sync) { 2482 if (!down_write_trylock(&sbi->gc_lock)) { 2483 ret = -EBUSY; 2484 goto out; 2485 } 2486 } else { 2487 down_write(&sbi->gc_lock); 2488 } 2489 2490 ret = f2fs_gc(sbi, range->sync, true, false, 2491 GET_SEGNO(sbi, range->start)); 2492 if (ret) { 2493 if (ret == -EBUSY) 2494 ret = -EAGAIN; 2495 goto out; 2496 } 2497 range->start += BLKS_PER_SEC(sbi); 2498 if (range->start <= end) 2499 goto do_more; 2500 out: 2501 mnt_drop_write_file(filp); 2502 return ret; 2503 } 2504 2505 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2506 { 2507 struct f2fs_gc_range range; 2508 2509 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2510 sizeof(range))) 2511 return -EFAULT; 2512 return __f2fs_ioc_gc_range(filp, &range); 2513 } 2514 2515 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg) 2516 { 2517 struct inode *inode = file_inode(filp); 2518 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2519 int ret; 2520 2521 if (!capable(CAP_SYS_ADMIN)) 2522 return -EPERM; 2523 2524 if (f2fs_readonly(sbi->sb)) 2525 return -EROFS; 2526 2527 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2528 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2529 return -EINVAL; 2530 } 2531 2532 ret = mnt_want_write_file(filp); 2533 if (ret) 2534 return ret; 2535 2536 ret = f2fs_sync_fs(sbi->sb, 1); 2537 2538 mnt_drop_write_file(filp); 2539 return ret; 2540 } 2541 2542 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2543 struct file *filp, 2544 struct f2fs_defragment *range) 2545 { 2546 struct inode *inode = file_inode(filp); 2547 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2548 .m_seg_type = NO_CHECK_TYPE, 2549 .m_may_create = false }; 2550 struct extent_info ei = {0, 0, 0}; 2551 pgoff_t pg_start, pg_end, next_pgofs; 2552 unsigned int blk_per_seg = sbi->blocks_per_seg; 2553 unsigned int total = 0, sec_num; 2554 block_t blk_end = 0; 2555 bool fragmented = false; 2556 int err; 2557 2558 /* if in-place-update policy is enabled, don't waste time here */ 2559 if (f2fs_should_update_inplace(inode, NULL)) 2560 return -EINVAL; 2561 2562 pg_start = range->start >> PAGE_SHIFT; 2563 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2564 2565 f2fs_balance_fs(sbi, true); 2566 2567 inode_lock(inode); 2568 2569 /* writeback all dirty pages in the range */ 2570 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2571 range->start + range->len - 1); 2572 if (err) 2573 goto out; 2574 2575 /* 2576 * lookup mapping info in extent cache, skip defragmenting if physical 2577 * block addresses are continuous. 2578 */ 2579 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) { 2580 if (ei.fofs + ei.len >= pg_end) 2581 goto out; 2582 } 2583 2584 map.m_lblk = pg_start; 2585 map.m_next_pgofs = &next_pgofs; 2586 2587 /* 2588 * lookup mapping info in dnode page cache, skip defragmenting if all 2589 * physical block addresses are continuous even if there are hole(s) 2590 * in logical blocks. 2591 */ 2592 while (map.m_lblk < pg_end) { 2593 map.m_len = pg_end - map.m_lblk; 2594 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2595 if (err) 2596 goto out; 2597 2598 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2599 map.m_lblk = next_pgofs; 2600 continue; 2601 } 2602 2603 if (blk_end && blk_end != map.m_pblk) 2604 fragmented = true; 2605 2606 /* record total count of block that we're going to move */ 2607 total += map.m_len; 2608 2609 blk_end = map.m_pblk + map.m_len; 2610 2611 map.m_lblk += map.m_len; 2612 } 2613 2614 if (!fragmented) { 2615 total = 0; 2616 goto out; 2617 } 2618 2619 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi)); 2620 2621 /* 2622 * make sure there are enough free section for LFS allocation, this can 2623 * avoid defragment running in SSR mode when free section are allocated 2624 * intensively 2625 */ 2626 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2627 err = -EAGAIN; 2628 goto out; 2629 } 2630 2631 map.m_lblk = pg_start; 2632 map.m_len = pg_end - pg_start; 2633 total = 0; 2634 2635 while (map.m_lblk < pg_end) { 2636 pgoff_t idx; 2637 int cnt = 0; 2638 2639 do_map: 2640 map.m_len = pg_end - map.m_lblk; 2641 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2642 if (err) 2643 goto clear_out; 2644 2645 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2646 map.m_lblk = next_pgofs; 2647 goto check; 2648 } 2649 2650 set_inode_flag(inode, FI_DO_DEFRAG); 2651 2652 idx = map.m_lblk; 2653 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2654 struct page *page; 2655 2656 page = f2fs_get_lock_data_page(inode, idx, true); 2657 if (IS_ERR(page)) { 2658 err = PTR_ERR(page); 2659 goto clear_out; 2660 } 2661 2662 set_page_dirty(page); 2663 f2fs_put_page(page, 1); 2664 2665 idx++; 2666 cnt++; 2667 total++; 2668 } 2669 2670 map.m_lblk = idx; 2671 check: 2672 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2673 goto do_map; 2674 2675 clear_inode_flag(inode, FI_DO_DEFRAG); 2676 2677 err = filemap_fdatawrite(inode->i_mapping); 2678 if (err) 2679 goto out; 2680 } 2681 clear_out: 2682 clear_inode_flag(inode, FI_DO_DEFRAG); 2683 out: 2684 inode_unlock(inode); 2685 if (!err) 2686 range->len = (u64)total << PAGE_SHIFT; 2687 return err; 2688 } 2689 2690 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2691 { 2692 struct inode *inode = file_inode(filp); 2693 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2694 struct f2fs_defragment range; 2695 int err; 2696 2697 if (!capable(CAP_SYS_ADMIN)) 2698 return -EPERM; 2699 2700 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2701 return -EINVAL; 2702 2703 if (f2fs_readonly(sbi->sb)) 2704 return -EROFS; 2705 2706 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2707 sizeof(range))) 2708 return -EFAULT; 2709 2710 /* verify alignment of offset & size */ 2711 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2712 return -EINVAL; 2713 2714 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2715 max_file_blocks(inode))) 2716 return -EINVAL; 2717 2718 err = mnt_want_write_file(filp); 2719 if (err) 2720 return err; 2721 2722 err = f2fs_defragment_range(sbi, filp, &range); 2723 mnt_drop_write_file(filp); 2724 2725 f2fs_update_time(sbi, REQ_TIME); 2726 if (err < 0) 2727 return err; 2728 2729 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2730 sizeof(range))) 2731 return -EFAULT; 2732 2733 return 0; 2734 } 2735 2736 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2737 struct file *file_out, loff_t pos_out, size_t len) 2738 { 2739 struct inode *src = file_inode(file_in); 2740 struct inode *dst = file_inode(file_out); 2741 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2742 size_t olen = len, dst_max_i_size = 0; 2743 size_t dst_osize; 2744 int ret; 2745 2746 if (file_in->f_path.mnt != file_out->f_path.mnt || 2747 src->i_sb != dst->i_sb) 2748 return -EXDEV; 2749 2750 if (unlikely(f2fs_readonly(src->i_sb))) 2751 return -EROFS; 2752 2753 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2754 return -EINVAL; 2755 2756 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2757 return -EOPNOTSUPP; 2758 2759 if (pos_out < 0 || pos_in < 0) 2760 return -EINVAL; 2761 2762 if (src == dst) { 2763 if (pos_in == pos_out) 2764 return 0; 2765 if (pos_out > pos_in && pos_out < pos_in + len) 2766 return -EINVAL; 2767 } 2768 2769 inode_lock(src); 2770 if (src != dst) { 2771 ret = -EBUSY; 2772 if (!inode_trylock(dst)) 2773 goto out; 2774 } 2775 2776 ret = -EINVAL; 2777 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2778 goto out_unlock; 2779 if (len == 0) 2780 olen = len = src->i_size - pos_in; 2781 if (pos_in + len == src->i_size) 2782 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2783 if (len == 0) { 2784 ret = 0; 2785 goto out_unlock; 2786 } 2787 2788 dst_osize = dst->i_size; 2789 if (pos_out + olen > dst->i_size) 2790 dst_max_i_size = pos_out + olen; 2791 2792 /* verify the end result is block aligned */ 2793 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2794 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2795 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2796 goto out_unlock; 2797 2798 ret = f2fs_convert_inline_inode(src); 2799 if (ret) 2800 goto out_unlock; 2801 2802 ret = f2fs_convert_inline_inode(dst); 2803 if (ret) 2804 goto out_unlock; 2805 2806 /* write out all dirty pages from offset */ 2807 ret = filemap_write_and_wait_range(src->i_mapping, 2808 pos_in, pos_in + len); 2809 if (ret) 2810 goto out_unlock; 2811 2812 ret = filemap_write_and_wait_range(dst->i_mapping, 2813 pos_out, pos_out + len); 2814 if (ret) 2815 goto out_unlock; 2816 2817 f2fs_balance_fs(sbi, true); 2818 2819 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2820 if (src != dst) { 2821 ret = -EBUSY; 2822 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2823 goto out_src; 2824 } 2825 2826 f2fs_lock_op(sbi); 2827 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2828 pos_out >> F2FS_BLKSIZE_BITS, 2829 len >> F2FS_BLKSIZE_BITS, false); 2830 2831 if (!ret) { 2832 if (dst_max_i_size) 2833 f2fs_i_size_write(dst, dst_max_i_size); 2834 else if (dst_osize != dst->i_size) 2835 f2fs_i_size_write(dst, dst_osize); 2836 } 2837 f2fs_unlock_op(sbi); 2838 2839 if (src != dst) 2840 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2841 out_src: 2842 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2843 out_unlock: 2844 if (src != dst) 2845 inode_unlock(dst); 2846 out: 2847 inode_unlock(src); 2848 return ret; 2849 } 2850 2851 static int __f2fs_ioc_move_range(struct file *filp, 2852 struct f2fs_move_range *range) 2853 { 2854 struct fd dst; 2855 int err; 2856 2857 if (!(filp->f_mode & FMODE_READ) || 2858 !(filp->f_mode & FMODE_WRITE)) 2859 return -EBADF; 2860 2861 dst = fdget(range->dst_fd); 2862 if (!dst.file) 2863 return -EBADF; 2864 2865 if (!(dst.file->f_mode & FMODE_WRITE)) { 2866 err = -EBADF; 2867 goto err_out; 2868 } 2869 2870 err = mnt_want_write_file(filp); 2871 if (err) 2872 goto err_out; 2873 2874 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2875 range->pos_out, range->len); 2876 2877 mnt_drop_write_file(filp); 2878 err_out: 2879 fdput(dst); 2880 return err; 2881 } 2882 2883 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2884 { 2885 struct f2fs_move_range range; 2886 2887 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2888 sizeof(range))) 2889 return -EFAULT; 2890 return __f2fs_ioc_move_range(filp, &range); 2891 } 2892 2893 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2894 { 2895 struct inode *inode = file_inode(filp); 2896 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2897 struct sit_info *sm = SIT_I(sbi); 2898 unsigned int start_segno = 0, end_segno = 0; 2899 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2900 struct f2fs_flush_device range; 2901 int ret; 2902 2903 if (!capable(CAP_SYS_ADMIN)) 2904 return -EPERM; 2905 2906 if (f2fs_readonly(sbi->sb)) 2907 return -EROFS; 2908 2909 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2910 return -EINVAL; 2911 2912 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2913 sizeof(range))) 2914 return -EFAULT; 2915 2916 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2917 __is_large_section(sbi)) { 2918 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2919 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2920 return -EINVAL; 2921 } 2922 2923 ret = mnt_want_write_file(filp); 2924 if (ret) 2925 return ret; 2926 2927 if (range.dev_num != 0) 2928 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2929 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2930 2931 start_segno = sm->last_victim[FLUSH_DEVICE]; 2932 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2933 start_segno = dev_start_segno; 2934 end_segno = min(start_segno + range.segments, dev_end_segno); 2935 2936 while (start_segno < end_segno) { 2937 if (!down_write_trylock(&sbi->gc_lock)) { 2938 ret = -EBUSY; 2939 goto out; 2940 } 2941 sm->last_victim[GC_CB] = end_segno + 1; 2942 sm->last_victim[GC_GREEDY] = end_segno + 1; 2943 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 2944 ret = f2fs_gc(sbi, true, true, true, start_segno); 2945 if (ret == -EAGAIN) 2946 ret = 0; 2947 else if (ret < 0) 2948 break; 2949 start_segno++; 2950 } 2951 out: 2952 mnt_drop_write_file(filp); 2953 return ret; 2954 } 2955 2956 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 2957 { 2958 struct inode *inode = file_inode(filp); 2959 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 2960 2961 /* Must validate to set it with SQLite behavior in Android. */ 2962 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 2963 2964 return put_user(sb_feature, (u32 __user *)arg); 2965 } 2966 2967 #ifdef CONFIG_QUOTA 2968 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 2969 { 2970 struct dquot *transfer_to[MAXQUOTAS] = {}; 2971 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2972 struct super_block *sb = sbi->sb; 2973 int err = 0; 2974 2975 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 2976 if (!IS_ERR(transfer_to[PRJQUOTA])) { 2977 err = __dquot_transfer(inode, transfer_to); 2978 if (err) 2979 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2980 dqput(transfer_to[PRJQUOTA]); 2981 } 2982 return err; 2983 } 2984 2985 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 2986 { 2987 struct f2fs_inode_info *fi = F2FS_I(inode); 2988 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2989 struct page *ipage; 2990 kprojid_t kprojid; 2991 int err; 2992 2993 if (!f2fs_sb_has_project_quota(sbi)) { 2994 if (projid != F2FS_DEF_PROJID) 2995 return -EOPNOTSUPP; 2996 else 2997 return 0; 2998 } 2999 3000 if (!f2fs_has_extra_attr(inode)) 3001 return -EOPNOTSUPP; 3002 3003 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3004 3005 if (projid_eq(kprojid, F2FS_I(inode)->i_projid)) 3006 return 0; 3007 3008 err = -EPERM; 3009 /* Is it quota file? Do not allow user to mess with it */ 3010 if (IS_NOQUOTA(inode)) 3011 return err; 3012 3013 ipage = f2fs_get_node_page(sbi, inode->i_ino); 3014 if (IS_ERR(ipage)) 3015 return PTR_ERR(ipage); 3016 3017 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize, 3018 i_projid)) { 3019 err = -EOVERFLOW; 3020 f2fs_put_page(ipage, 1); 3021 return err; 3022 } 3023 f2fs_put_page(ipage, 1); 3024 3025 err = f2fs_dquot_initialize(inode); 3026 if (err) 3027 return err; 3028 3029 f2fs_lock_op(sbi); 3030 err = f2fs_transfer_project_quota(inode, kprojid); 3031 if (err) 3032 goto out_unlock; 3033 3034 F2FS_I(inode)->i_projid = kprojid; 3035 inode->i_ctime = current_time(inode); 3036 f2fs_mark_inode_dirty_sync(inode, true); 3037 out_unlock: 3038 f2fs_unlock_op(sbi); 3039 return err; 3040 } 3041 #else 3042 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3043 { 3044 return 0; 3045 } 3046 3047 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3048 { 3049 if (projid != F2FS_DEF_PROJID) 3050 return -EOPNOTSUPP; 3051 return 0; 3052 } 3053 #endif 3054 3055 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3056 { 3057 struct inode *inode = d_inode(dentry); 3058 struct f2fs_inode_info *fi = F2FS_I(inode); 3059 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3060 3061 if (IS_ENCRYPTED(inode)) 3062 fsflags |= FS_ENCRYPT_FL; 3063 if (IS_VERITY(inode)) 3064 fsflags |= FS_VERITY_FL; 3065 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3066 fsflags |= FS_INLINE_DATA_FL; 3067 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3068 fsflags |= FS_NOCOW_FL; 3069 3070 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3071 3072 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3073 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3074 3075 return 0; 3076 } 3077 3078 int f2fs_fileattr_set(struct user_namespace *mnt_userns, 3079 struct dentry *dentry, struct fileattr *fa) 3080 { 3081 struct inode *inode = d_inode(dentry); 3082 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3083 u32 iflags; 3084 int err; 3085 3086 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3087 return -EIO; 3088 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3089 return -ENOSPC; 3090 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3091 return -EOPNOTSUPP; 3092 fsflags &= F2FS_SETTABLE_FS_FL; 3093 if (!fa->flags_valid) 3094 mask &= FS_COMMON_FL; 3095 3096 iflags = f2fs_fsflags_to_iflags(fsflags); 3097 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3098 return -EOPNOTSUPP; 3099 3100 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3101 if (!err) 3102 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3103 3104 return err; 3105 } 3106 3107 int f2fs_pin_file_control(struct inode *inode, bool inc) 3108 { 3109 struct f2fs_inode_info *fi = F2FS_I(inode); 3110 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3111 3112 /* Use i_gc_failures for normal file as a risk signal. */ 3113 if (inc) 3114 f2fs_i_gc_failures_write(inode, 3115 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3116 3117 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3118 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3119 __func__, inode->i_ino, 3120 fi->i_gc_failures[GC_FAILURE_PIN]); 3121 clear_inode_flag(inode, FI_PIN_FILE); 3122 return -EAGAIN; 3123 } 3124 return 0; 3125 } 3126 3127 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3128 { 3129 struct inode *inode = file_inode(filp); 3130 __u32 pin; 3131 int ret = 0; 3132 3133 if (get_user(pin, (__u32 __user *)arg)) 3134 return -EFAULT; 3135 3136 if (!S_ISREG(inode->i_mode)) 3137 return -EINVAL; 3138 3139 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3140 return -EROFS; 3141 3142 ret = mnt_want_write_file(filp); 3143 if (ret) 3144 return ret; 3145 3146 inode_lock(inode); 3147 3148 if (f2fs_should_update_outplace(inode, NULL)) { 3149 ret = -EINVAL; 3150 goto out; 3151 } 3152 3153 if (!pin) { 3154 clear_inode_flag(inode, FI_PIN_FILE); 3155 f2fs_i_gc_failures_write(inode, 0); 3156 goto done; 3157 } 3158 3159 if (f2fs_pin_file_control(inode, false)) { 3160 ret = -EAGAIN; 3161 goto out; 3162 } 3163 3164 ret = f2fs_convert_inline_inode(inode); 3165 if (ret) 3166 goto out; 3167 3168 if (!f2fs_disable_compressed_file(inode)) { 3169 ret = -EOPNOTSUPP; 3170 goto out; 3171 } 3172 3173 set_inode_flag(inode, FI_PIN_FILE); 3174 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3175 done: 3176 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3177 out: 3178 inode_unlock(inode); 3179 mnt_drop_write_file(filp); 3180 return ret; 3181 } 3182 3183 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3184 { 3185 struct inode *inode = file_inode(filp); 3186 __u32 pin = 0; 3187 3188 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3189 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3190 return put_user(pin, (u32 __user *)arg); 3191 } 3192 3193 int f2fs_precache_extents(struct inode *inode) 3194 { 3195 struct f2fs_inode_info *fi = F2FS_I(inode); 3196 struct f2fs_map_blocks map; 3197 pgoff_t m_next_extent; 3198 loff_t end; 3199 int err; 3200 3201 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3202 return -EOPNOTSUPP; 3203 3204 map.m_lblk = 0; 3205 map.m_next_pgofs = NULL; 3206 map.m_next_extent = &m_next_extent; 3207 map.m_seg_type = NO_CHECK_TYPE; 3208 map.m_may_create = false; 3209 end = max_file_blocks(inode); 3210 3211 while (map.m_lblk < end) { 3212 map.m_len = end - map.m_lblk; 3213 3214 down_write(&fi->i_gc_rwsem[WRITE]); 3215 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE); 3216 up_write(&fi->i_gc_rwsem[WRITE]); 3217 if (err) 3218 return err; 3219 3220 map.m_lblk = m_next_extent; 3221 } 3222 3223 return 0; 3224 } 3225 3226 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg) 3227 { 3228 return f2fs_precache_extents(file_inode(filp)); 3229 } 3230 3231 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3232 { 3233 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3234 __u64 block_count; 3235 3236 if (!capable(CAP_SYS_ADMIN)) 3237 return -EPERM; 3238 3239 if (f2fs_readonly(sbi->sb)) 3240 return -EROFS; 3241 3242 if (copy_from_user(&block_count, (void __user *)arg, 3243 sizeof(block_count))) 3244 return -EFAULT; 3245 3246 return f2fs_resize_fs(sbi, block_count); 3247 } 3248 3249 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3250 { 3251 struct inode *inode = file_inode(filp); 3252 3253 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3254 3255 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3256 f2fs_warn(F2FS_I_SB(inode), 3257 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3258 inode->i_ino); 3259 return -EOPNOTSUPP; 3260 } 3261 3262 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3263 } 3264 3265 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3266 { 3267 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3268 return -EOPNOTSUPP; 3269 3270 return fsverity_ioctl_measure(filp, (void __user *)arg); 3271 } 3272 3273 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3274 { 3275 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3276 return -EOPNOTSUPP; 3277 3278 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3279 } 3280 3281 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3282 { 3283 struct inode *inode = file_inode(filp); 3284 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3285 char *vbuf; 3286 int count; 3287 int err = 0; 3288 3289 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3290 if (!vbuf) 3291 return -ENOMEM; 3292 3293 down_read(&sbi->sb_lock); 3294 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3295 ARRAY_SIZE(sbi->raw_super->volume_name), 3296 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3297 up_read(&sbi->sb_lock); 3298 3299 if (copy_to_user((char __user *)arg, vbuf, 3300 min(FSLABEL_MAX, count))) 3301 err = -EFAULT; 3302 3303 kfree(vbuf); 3304 return err; 3305 } 3306 3307 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3308 { 3309 struct inode *inode = file_inode(filp); 3310 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3311 char *vbuf; 3312 int err = 0; 3313 3314 if (!capable(CAP_SYS_ADMIN)) 3315 return -EPERM; 3316 3317 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3318 if (IS_ERR(vbuf)) 3319 return PTR_ERR(vbuf); 3320 3321 err = mnt_want_write_file(filp); 3322 if (err) 3323 goto out; 3324 3325 down_write(&sbi->sb_lock); 3326 3327 memset(sbi->raw_super->volume_name, 0, 3328 sizeof(sbi->raw_super->volume_name)); 3329 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3330 sbi->raw_super->volume_name, 3331 ARRAY_SIZE(sbi->raw_super->volume_name)); 3332 3333 err = f2fs_commit_super(sbi, false); 3334 3335 up_write(&sbi->sb_lock); 3336 3337 mnt_drop_write_file(filp); 3338 out: 3339 kfree(vbuf); 3340 return err; 3341 } 3342 3343 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg) 3344 { 3345 struct inode *inode = file_inode(filp); 3346 __u64 blocks; 3347 3348 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3349 return -EOPNOTSUPP; 3350 3351 if (!f2fs_compressed_file(inode)) 3352 return -EINVAL; 3353 3354 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3355 return put_user(blocks, (u64 __user *)arg); 3356 } 3357 3358 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3359 { 3360 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3361 unsigned int released_blocks = 0; 3362 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3363 block_t blkaddr; 3364 int i; 3365 3366 for (i = 0; i < count; i++) { 3367 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3368 dn->ofs_in_node + i); 3369 3370 if (!__is_valid_data_blkaddr(blkaddr)) 3371 continue; 3372 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3373 DATA_GENERIC_ENHANCE))) 3374 return -EFSCORRUPTED; 3375 } 3376 3377 while (count) { 3378 int compr_blocks = 0; 3379 3380 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3381 blkaddr = f2fs_data_blkaddr(dn); 3382 3383 if (i == 0) { 3384 if (blkaddr == COMPRESS_ADDR) 3385 continue; 3386 dn->ofs_in_node += cluster_size; 3387 goto next; 3388 } 3389 3390 if (__is_valid_data_blkaddr(blkaddr)) 3391 compr_blocks++; 3392 3393 if (blkaddr != NEW_ADDR) 3394 continue; 3395 3396 dn->data_blkaddr = NULL_ADDR; 3397 f2fs_set_data_blkaddr(dn); 3398 } 3399 3400 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3401 dec_valid_block_count(sbi, dn->inode, 3402 cluster_size - compr_blocks); 3403 3404 released_blocks += cluster_size - compr_blocks; 3405 next: 3406 count -= cluster_size; 3407 } 3408 3409 return released_blocks; 3410 } 3411 3412 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3413 { 3414 struct inode *inode = file_inode(filp); 3415 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3416 pgoff_t page_idx = 0, last_idx; 3417 unsigned int released_blocks = 0; 3418 int ret; 3419 int writecount; 3420 3421 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3422 return -EOPNOTSUPP; 3423 3424 if (!f2fs_compressed_file(inode)) 3425 return -EINVAL; 3426 3427 if (f2fs_readonly(sbi->sb)) 3428 return -EROFS; 3429 3430 ret = mnt_want_write_file(filp); 3431 if (ret) 3432 return ret; 3433 3434 f2fs_balance_fs(F2FS_I_SB(inode), true); 3435 3436 inode_lock(inode); 3437 3438 writecount = atomic_read(&inode->i_writecount); 3439 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3440 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3441 ret = -EBUSY; 3442 goto out; 3443 } 3444 3445 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3446 ret = -EINVAL; 3447 goto out; 3448 } 3449 3450 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3451 if (ret) 3452 goto out; 3453 3454 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3455 inode->i_ctime = current_time(inode); 3456 f2fs_mark_inode_dirty_sync(inode, true); 3457 3458 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3459 goto out; 3460 3461 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3462 filemap_invalidate_lock(inode->i_mapping); 3463 3464 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3465 3466 while (page_idx < last_idx) { 3467 struct dnode_of_data dn; 3468 pgoff_t end_offset, count; 3469 3470 set_new_dnode(&dn, inode, NULL, NULL, 0); 3471 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3472 if (ret) { 3473 if (ret == -ENOENT) { 3474 page_idx = f2fs_get_next_page_offset(&dn, 3475 page_idx); 3476 ret = 0; 3477 continue; 3478 } 3479 break; 3480 } 3481 3482 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3483 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3484 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3485 3486 ret = release_compress_blocks(&dn, count); 3487 3488 f2fs_put_dnode(&dn); 3489 3490 if (ret < 0) 3491 break; 3492 3493 page_idx += count; 3494 released_blocks += ret; 3495 } 3496 3497 filemap_invalidate_unlock(inode->i_mapping); 3498 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3499 out: 3500 inode_unlock(inode); 3501 3502 mnt_drop_write_file(filp); 3503 3504 if (ret >= 0) { 3505 ret = put_user(released_blocks, (u64 __user *)arg); 3506 } else if (released_blocks && 3507 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3508 set_sbi_flag(sbi, SBI_NEED_FSCK); 3509 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3510 "iblocks=%llu, released=%u, compr_blocks=%u, " 3511 "run fsck to fix.", 3512 __func__, inode->i_ino, inode->i_blocks, 3513 released_blocks, 3514 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3515 } 3516 3517 return ret; 3518 } 3519 3520 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3521 { 3522 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3523 unsigned int reserved_blocks = 0; 3524 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3525 block_t blkaddr; 3526 int i; 3527 3528 for (i = 0; i < count; i++) { 3529 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3530 dn->ofs_in_node + i); 3531 3532 if (!__is_valid_data_blkaddr(blkaddr)) 3533 continue; 3534 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3535 DATA_GENERIC_ENHANCE))) 3536 return -EFSCORRUPTED; 3537 } 3538 3539 while (count) { 3540 int compr_blocks = 0; 3541 blkcnt_t reserved; 3542 int ret; 3543 3544 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3545 blkaddr = f2fs_data_blkaddr(dn); 3546 3547 if (i == 0) { 3548 if (blkaddr == COMPRESS_ADDR) 3549 continue; 3550 dn->ofs_in_node += cluster_size; 3551 goto next; 3552 } 3553 3554 if (__is_valid_data_blkaddr(blkaddr)) { 3555 compr_blocks++; 3556 continue; 3557 } 3558 3559 dn->data_blkaddr = NEW_ADDR; 3560 f2fs_set_data_blkaddr(dn); 3561 } 3562 3563 reserved = cluster_size - compr_blocks; 3564 ret = inc_valid_block_count(sbi, dn->inode, &reserved); 3565 if (ret) 3566 return ret; 3567 3568 if (reserved != cluster_size - compr_blocks) 3569 return -ENOSPC; 3570 3571 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3572 3573 reserved_blocks += reserved; 3574 next: 3575 count -= cluster_size; 3576 } 3577 3578 return reserved_blocks; 3579 } 3580 3581 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3582 { 3583 struct inode *inode = file_inode(filp); 3584 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3585 pgoff_t page_idx = 0, last_idx; 3586 unsigned int reserved_blocks = 0; 3587 int ret; 3588 3589 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3590 return -EOPNOTSUPP; 3591 3592 if (!f2fs_compressed_file(inode)) 3593 return -EINVAL; 3594 3595 if (f2fs_readonly(sbi->sb)) 3596 return -EROFS; 3597 3598 ret = mnt_want_write_file(filp); 3599 if (ret) 3600 return ret; 3601 3602 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3603 goto out; 3604 3605 f2fs_balance_fs(F2FS_I_SB(inode), true); 3606 3607 inode_lock(inode); 3608 3609 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3610 ret = -EINVAL; 3611 goto unlock_inode; 3612 } 3613 3614 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3615 filemap_invalidate_lock(inode->i_mapping); 3616 3617 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3618 3619 while (page_idx < last_idx) { 3620 struct dnode_of_data dn; 3621 pgoff_t end_offset, count; 3622 3623 set_new_dnode(&dn, inode, NULL, NULL, 0); 3624 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3625 if (ret) { 3626 if (ret == -ENOENT) { 3627 page_idx = f2fs_get_next_page_offset(&dn, 3628 page_idx); 3629 ret = 0; 3630 continue; 3631 } 3632 break; 3633 } 3634 3635 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3636 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3637 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3638 3639 ret = reserve_compress_blocks(&dn, count); 3640 3641 f2fs_put_dnode(&dn); 3642 3643 if (ret < 0) 3644 break; 3645 3646 page_idx += count; 3647 reserved_blocks += ret; 3648 } 3649 3650 filemap_invalidate_unlock(inode->i_mapping); 3651 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3652 3653 if (ret >= 0) { 3654 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3655 inode->i_ctime = current_time(inode); 3656 f2fs_mark_inode_dirty_sync(inode, true); 3657 } 3658 unlock_inode: 3659 inode_unlock(inode); 3660 out: 3661 mnt_drop_write_file(filp); 3662 3663 if (ret >= 0) { 3664 ret = put_user(reserved_blocks, (u64 __user *)arg); 3665 } else if (reserved_blocks && 3666 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3667 set_sbi_flag(sbi, SBI_NEED_FSCK); 3668 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3669 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3670 "run fsck to fix.", 3671 __func__, inode->i_ino, inode->i_blocks, 3672 reserved_blocks, 3673 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3674 } 3675 3676 return ret; 3677 } 3678 3679 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3680 pgoff_t off, block_t block, block_t len, u32 flags) 3681 { 3682 struct request_queue *q = bdev_get_queue(bdev); 3683 sector_t sector = SECTOR_FROM_BLOCK(block); 3684 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3685 int ret = 0; 3686 3687 if (!q) 3688 return -ENXIO; 3689 3690 if (flags & F2FS_TRIM_FILE_DISCARD) 3691 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS, 3692 blk_queue_secure_erase(q) ? 3693 BLKDEV_DISCARD_SECURE : 0); 3694 3695 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3696 if (IS_ENCRYPTED(inode)) 3697 ret = fscrypt_zeroout_range(inode, off, block, len); 3698 else 3699 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3700 GFP_NOFS, 0); 3701 } 3702 3703 return ret; 3704 } 3705 3706 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3707 { 3708 struct inode *inode = file_inode(filp); 3709 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3710 struct address_space *mapping = inode->i_mapping; 3711 struct block_device *prev_bdev = NULL; 3712 struct f2fs_sectrim_range range; 3713 pgoff_t index, pg_end, prev_index = 0; 3714 block_t prev_block = 0, len = 0; 3715 loff_t end_addr; 3716 bool to_end = false; 3717 int ret = 0; 3718 3719 if (!(filp->f_mode & FMODE_WRITE)) 3720 return -EBADF; 3721 3722 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3723 sizeof(range))) 3724 return -EFAULT; 3725 3726 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3727 !S_ISREG(inode->i_mode)) 3728 return -EINVAL; 3729 3730 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3731 !f2fs_hw_support_discard(sbi)) || 3732 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3733 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3734 return -EOPNOTSUPP; 3735 3736 file_start_write(filp); 3737 inode_lock(inode); 3738 3739 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3740 range.start >= inode->i_size) { 3741 ret = -EINVAL; 3742 goto err; 3743 } 3744 3745 if (range.len == 0) 3746 goto err; 3747 3748 if (inode->i_size - range.start > range.len) { 3749 end_addr = range.start + range.len; 3750 } else { 3751 end_addr = range.len == (u64)-1 ? 3752 sbi->sb->s_maxbytes : inode->i_size; 3753 to_end = true; 3754 } 3755 3756 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3757 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3758 ret = -EINVAL; 3759 goto err; 3760 } 3761 3762 index = F2FS_BYTES_TO_BLK(range.start); 3763 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3764 3765 ret = f2fs_convert_inline_inode(inode); 3766 if (ret) 3767 goto err; 3768 3769 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3770 filemap_invalidate_lock(mapping); 3771 3772 ret = filemap_write_and_wait_range(mapping, range.start, 3773 to_end ? LLONG_MAX : end_addr - 1); 3774 if (ret) 3775 goto out; 3776 3777 truncate_inode_pages_range(mapping, range.start, 3778 to_end ? -1 : end_addr - 1); 3779 3780 while (index < pg_end) { 3781 struct dnode_of_data dn; 3782 pgoff_t end_offset, count; 3783 int i; 3784 3785 set_new_dnode(&dn, inode, NULL, NULL, 0); 3786 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3787 if (ret) { 3788 if (ret == -ENOENT) { 3789 index = f2fs_get_next_page_offset(&dn, index); 3790 continue; 3791 } 3792 goto out; 3793 } 3794 3795 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3796 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3797 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3798 struct block_device *cur_bdev; 3799 block_t blkaddr = f2fs_data_blkaddr(&dn); 3800 3801 if (!__is_valid_data_blkaddr(blkaddr)) 3802 continue; 3803 3804 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3805 DATA_GENERIC_ENHANCE)) { 3806 ret = -EFSCORRUPTED; 3807 f2fs_put_dnode(&dn); 3808 goto out; 3809 } 3810 3811 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3812 if (f2fs_is_multi_device(sbi)) { 3813 int di = f2fs_target_device_index(sbi, blkaddr); 3814 3815 blkaddr -= FDEV(di).start_blk; 3816 } 3817 3818 if (len) { 3819 if (prev_bdev == cur_bdev && 3820 index == prev_index + len && 3821 blkaddr == prev_block + len) { 3822 len++; 3823 } else { 3824 ret = f2fs_secure_erase(prev_bdev, 3825 inode, prev_index, prev_block, 3826 len, range.flags); 3827 if (ret) { 3828 f2fs_put_dnode(&dn); 3829 goto out; 3830 } 3831 3832 len = 0; 3833 } 3834 } 3835 3836 if (!len) { 3837 prev_bdev = cur_bdev; 3838 prev_index = index; 3839 prev_block = blkaddr; 3840 len = 1; 3841 } 3842 } 3843 3844 f2fs_put_dnode(&dn); 3845 3846 if (fatal_signal_pending(current)) { 3847 ret = -EINTR; 3848 goto out; 3849 } 3850 cond_resched(); 3851 } 3852 3853 if (len) 3854 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3855 prev_block, len, range.flags); 3856 out: 3857 filemap_invalidate_unlock(mapping); 3858 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3859 err: 3860 inode_unlock(inode); 3861 file_end_write(filp); 3862 3863 return ret; 3864 } 3865 3866 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3867 { 3868 struct inode *inode = file_inode(filp); 3869 struct f2fs_comp_option option; 3870 3871 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3872 return -EOPNOTSUPP; 3873 3874 inode_lock_shared(inode); 3875 3876 if (!f2fs_compressed_file(inode)) { 3877 inode_unlock_shared(inode); 3878 return -ENODATA; 3879 } 3880 3881 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3882 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3883 3884 inode_unlock_shared(inode); 3885 3886 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3887 sizeof(option))) 3888 return -EFAULT; 3889 3890 return 0; 3891 } 3892 3893 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3894 { 3895 struct inode *inode = file_inode(filp); 3896 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3897 struct f2fs_comp_option option; 3898 int ret = 0; 3899 3900 if (!f2fs_sb_has_compression(sbi)) 3901 return -EOPNOTSUPP; 3902 3903 if (!(filp->f_mode & FMODE_WRITE)) 3904 return -EBADF; 3905 3906 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3907 sizeof(option))) 3908 return -EFAULT; 3909 3910 if (!f2fs_compressed_file(inode) || 3911 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3912 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3913 option.algorithm >= COMPRESS_MAX) 3914 return -EINVAL; 3915 3916 file_start_write(filp); 3917 inode_lock(inode); 3918 3919 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 3920 ret = -EBUSY; 3921 goto out; 3922 } 3923 3924 if (inode->i_size != 0) { 3925 ret = -EFBIG; 3926 goto out; 3927 } 3928 3929 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 3930 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 3931 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size; 3932 f2fs_mark_inode_dirty_sync(inode, true); 3933 3934 if (!f2fs_is_compress_backend_ready(inode)) 3935 f2fs_warn(sbi, "compression algorithm is successfully set, " 3936 "but current kernel doesn't support this algorithm."); 3937 out: 3938 inode_unlock(inode); 3939 file_end_write(filp); 3940 3941 return ret; 3942 } 3943 3944 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 3945 { 3946 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 3947 struct address_space *mapping = inode->i_mapping; 3948 struct page *page; 3949 pgoff_t redirty_idx = page_idx; 3950 int i, page_len = 0, ret = 0; 3951 3952 page_cache_ra_unbounded(&ractl, len, 0); 3953 3954 for (i = 0; i < len; i++, page_idx++) { 3955 page = read_cache_page(mapping, page_idx, NULL, NULL); 3956 if (IS_ERR(page)) { 3957 ret = PTR_ERR(page); 3958 break; 3959 } 3960 page_len++; 3961 } 3962 3963 for (i = 0; i < page_len; i++, redirty_idx++) { 3964 page = find_lock_page(mapping, redirty_idx); 3965 if (!page) { 3966 ret = -ENOMEM; 3967 break; 3968 } 3969 set_page_dirty(page); 3970 f2fs_put_page(page, 1); 3971 f2fs_put_page(page, 0); 3972 } 3973 3974 return ret; 3975 } 3976 3977 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg) 3978 { 3979 struct inode *inode = file_inode(filp); 3980 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3981 struct f2fs_inode_info *fi = F2FS_I(inode); 3982 pgoff_t page_idx = 0, last_idx; 3983 unsigned int blk_per_seg = sbi->blocks_per_seg; 3984 int cluster_size = F2FS_I(inode)->i_cluster_size; 3985 int count, ret; 3986 3987 if (!f2fs_sb_has_compression(sbi) || 3988 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 3989 return -EOPNOTSUPP; 3990 3991 if (!(filp->f_mode & FMODE_WRITE)) 3992 return -EBADF; 3993 3994 if (!f2fs_compressed_file(inode)) 3995 return -EINVAL; 3996 3997 f2fs_balance_fs(F2FS_I_SB(inode), true); 3998 3999 file_start_write(filp); 4000 inode_lock(inode); 4001 4002 if (!f2fs_is_compress_backend_ready(inode)) { 4003 ret = -EOPNOTSUPP; 4004 goto out; 4005 } 4006 4007 if (f2fs_is_mmap_file(inode)) { 4008 ret = -EBUSY; 4009 goto out; 4010 } 4011 4012 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4013 if (ret) 4014 goto out; 4015 4016 if (!atomic_read(&fi->i_compr_blocks)) 4017 goto out; 4018 4019 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4020 4021 count = last_idx - page_idx; 4022 while (count) { 4023 int len = min(cluster_size, count); 4024 4025 ret = redirty_blocks(inode, page_idx, len); 4026 if (ret < 0) 4027 break; 4028 4029 if (get_dirty_pages(inode) >= blk_per_seg) 4030 filemap_fdatawrite(inode->i_mapping); 4031 4032 count -= len; 4033 page_idx += len; 4034 } 4035 4036 if (!ret) 4037 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4038 LLONG_MAX); 4039 4040 if (ret) 4041 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4042 __func__, ret); 4043 out: 4044 inode_unlock(inode); 4045 file_end_write(filp); 4046 4047 return ret; 4048 } 4049 4050 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg) 4051 { 4052 struct inode *inode = file_inode(filp); 4053 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4054 pgoff_t page_idx = 0, last_idx; 4055 unsigned int blk_per_seg = sbi->blocks_per_seg; 4056 int cluster_size = F2FS_I(inode)->i_cluster_size; 4057 int count, ret; 4058 4059 if (!f2fs_sb_has_compression(sbi) || 4060 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4061 return -EOPNOTSUPP; 4062 4063 if (!(filp->f_mode & FMODE_WRITE)) 4064 return -EBADF; 4065 4066 if (!f2fs_compressed_file(inode)) 4067 return -EINVAL; 4068 4069 f2fs_balance_fs(F2FS_I_SB(inode), true); 4070 4071 file_start_write(filp); 4072 inode_lock(inode); 4073 4074 if (!f2fs_is_compress_backend_ready(inode)) { 4075 ret = -EOPNOTSUPP; 4076 goto out; 4077 } 4078 4079 if (f2fs_is_mmap_file(inode)) { 4080 ret = -EBUSY; 4081 goto out; 4082 } 4083 4084 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4085 if (ret) 4086 goto out; 4087 4088 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4089 4090 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4091 4092 count = last_idx - page_idx; 4093 while (count) { 4094 int len = min(cluster_size, count); 4095 4096 ret = redirty_blocks(inode, page_idx, len); 4097 if (ret < 0) 4098 break; 4099 4100 if (get_dirty_pages(inode) >= blk_per_seg) 4101 filemap_fdatawrite(inode->i_mapping); 4102 4103 count -= len; 4104 page_idx += len; 4105 } 4106 4107 if (!ret) 4108 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4109 LLONG_MAX); 4110 4111 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4112 4113 if (ret) 4114 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4115 __func__, ret); 4116 out: 4117 inode_unlock(inode); 4118 file_end_write(filp); 4119 4120 return ret; 4121 } 4122 4123 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4124 { 4125 switch (cmd) { 4126 case FS_IOC_GETVERSION: 4127 return f2fs_ioc_getversion(filp, arg); 4128 case F2FS_IOC_START_ATOMIC_WRITE: 4129 return f2fs_ioc_start_atomic_write(filp); 4130 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4131 return f2fs_ioc_commit_atomic_write(filp); 4132 case F2FS_IOC_START_VOLATILE_WRITE: 4133 return f2fs_ioc_start_volatile_write(filp); 4134 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4135 return f2fs_ioc_release_volatile_write(filp); 4136 case F2FS_IOC_ABORT_VOLATILE_WRITE: 4137 return f2fs_ioc_abort_volatile_write(filp); 4138 case F2FS_IOC_SHUTDOWN: 4139 return f2fs_ioc_shutdown(filp, arg); 4140 case FITRIM: 4141 return f2fs_ioc_fitrim(filp, arg); 4142 case FS_IOC_SET_ENCRYPTION_POLICY: 4143 return f2fs_ioc_set_encryption_policy(filp, arg); 4144 case FS_IOC_GET_ENCRYPTION_POLICY: 4145 return f2fs_ioc_get_encryption_policy(filp, arg); 4146 case FS_IOC_GET_ENCRYPTION_PWSALT: 4147 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4148 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4149 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4150 case FS_IOC_ADD_ENCRYPTION_KEY: 4151 return f2fs_ioc_add_encryption_key(filp, arg); 4152 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4153 return f2fs_ioc_remove_encryption_key(filp, arg); 4154 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4155 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4156 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4157 return f2fs_ioc_get_encryption_key_status(filp, arg); 4158 case FS_IOC_GET_ENCRYPTION_NONCE: 4159 return f2fs_ioc_get_encryption_nonce(filp, arg); 4160 case F2FS_IOC_GARBAGE_COLLECT: 4161 return f2fs_ioc_gc(filp, arg); 4162 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4163 return f2fs_ioc_gc_range(filp, arg); 4164 case F2FS_IOC_WRITE_CHECKPOINT: 4165 return f2fs_ioc_write_checkpoint(filp, arg); 4166 case F2FS_IOC_DEFRAGMENT: 4167 return f2fs_ioc_defragment(filp, arg); 4168 case F2FS_IOC_MOVE_RANGE: 4169 return f2fs_ioc_move_range(filp, arg); 4170 case F2FS_IOC_FLUSH_DEVICE: 4171 return f2fs_ioc_flush_device(filp, arg); 4172 case F2FS_IOC_GET_FEATURES: 4173 return f2fs_ioc_get_features(filp, arg); 4174 case F2FS_IOC_GET_PIN_FILE: 4175 return f2fs_ioc_get_pin_file(filp, arg); 4176 case F2FS_IOC_SET_PIN_FILE: 4177 return f2fs_ioc_set_pin_file(filp, arg); 4178 case F2FS_IOC_PRECACHE_EXTENTS: 4179 return f2fs_ioc_precache_extents(filp, arg); 4180 case F2FS_IOC_RESIZE_FS: 4181 return f2fs_ioc_resize_fs(filp, arg); 4182 case FS_IOC_ENABLE_VERITY: 4183 return f2fs_ioc_enable_verity(filp, arg); 4184 case FS_IOC_MEASURE_VERITY: 4185 return f2fs_ioc_measure_verity(filp, arg); 4186 case FS_IOC_READ_VERITY_METADATA: 4187 return f2fs_ioc_read_verity_metadata(filp, arg); 4188 case FS_IOC_GETFSLABEL: 4189 return f2fs_ioc_getfslabel(filp, arg); 4190 case FS_IOC_SETFSLABEL: 4191 return f2fs_ioc_setfslabel(filp, arg); 4192 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4193 return f2fs_get_compress_blocks(filp, arg); 4194 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4195 return f2fs_release_compress_blocks(filp, arg); 4196 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4197 return f2fs_reserve_compress_blocks(filp, arg); 4198 case F2FS_IOC_SEC_TRIM_FILE: 4199 return f2fs_sec_trim_file(filp, arg); 4200 case F2FS_IOC_GET_COMPRESS_OPTION: 4201 return f2fs_ioc_get_compress_option(filp, arg); 4202 case F2FS_IOC_SET_COMPRESS_OPTION: 4203 return f2fs_ioc_set_compress_option(filp, arg); 4204 case F2FS_IOC_DECOMPRESS_FILE: 4205 return f2fs_ioc_decompress_file(filp, arg); 4206 case F2FS_IOC_COMPRESS_FILE: 4207 return f2fs_ioc_compress_file(filp, arg); 4208 default: 4209 return -ENOTTY; 4210 } 4211 } 4212 4213 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4214 { 4215 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4216 return -EIO; 4217 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4218 return -ENOSPC; 4219 4220 return __f2fs_ioctl(filp, cmd, arg); 4221 } 4222 4223 /* 4224 * Return %true if the given read or write request should use direct I/O, or 4225 * %false if it should use buffered I/O. 4226 */ 4227 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4228 struct iov_iter *iter) 4229 { 4230 unsigned int align; 4231 4232 if (!(iocb->ki_flags & IOCB_DIRECT)) 4233 return false; 4234 4235 if (f2fs_force_buffered_io(inode, iocb, iter)) 4236 return false; 4237 4238 /* 4239 * Direct I/O not aligned to the disk's logical_block_size will be 4240 * attempted, but will fail with -EINVAL. 4241 * 4242 * f2fs additionally requires that direct I/O be aligned to the 4243 * filesystem block size, which is often a stricter requirement. 4244 * However, f2fs traditionally falls back to buffered I/O on requests 4245 * that are logical_block_size-aligned but not fs-block aligned. 4246 * 4247 * The below logic implements this behavior. 4248 */ 4249 align = iocb->ki_pos | iov_iter_alignment(iter); 4250 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4251 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4252 return false; 4253 4254 return true; 4255 } 4256 4257 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4258 unsigned int flags) 4259 { 4260 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4261 4262 dec_page_count(sbi, F2FS_DIO_READ); 4263 if (error) 4264 return error; 4265 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size); 4266 return 0; 4267 } 4268 4269 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4270 .end_io = f2fs_dio_read_end_io, 4271 }; 4272 4273 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4274 { 4275 struct file *file = iocb->ki_filp; 4276 struct inode *inode = file_inode(file); 4277 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4278 struct f2fs_inode_info *fi = F2FS_I(inode); 4279 const loff_t pos = iocb->ki_pos; 4280 const size_t count = iov_iter_count(to); 4281 struct iomap_dio *dio; 4282 ssize_t ret; 4283 4284 if (count == 0) 4285 return 0; /* skip atime update */ 4286 4287 trace_f2fs_direct_IO_enter(inode, pos, count, READ); 4288 4289 if (iocb->ki_flags & IOCB_NOWAIT) { 4290 if (!down_read_trylock(&fi->i_gc_rwsem[READ])) { 4291 ret = -EAGAIN; 4292 goto out; 4293 } 4294 } else { 4295 down_read(&fi->i_gc_rwsem[READ]); 4296 } 4297 4298 /* 4299 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4300 * the higher-level function iomap_dio_rw() in order to ensure that the 4301 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4302 */ 4303 inc_page_count(sbi, F2FS_DIO_READ); 4304 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4305 &f2fs_iomap_dio_read_ops, 0, 0); 4306 if (IS_ERR_OR_NULL(dio)) { 4307 ret = PTR_ERR_OR_ZERO(dio); 4308 if (ret != -EIOCBQUEUED) 4309 dec_page_count(sbi, F2FS_DIO_READ); 4310 } else { 4311 ret = iomap_dio_complete(dio); 4312 } 4313 4314 up_read(&fi->i_gc_rwsem[READ]); 4315 4316 file_accessed(file); 4317 out: 4318 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4319 return ret; 4320 } 4321 4322 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4323 { 4324 struct inode *inode = file_inode(iocb->ki_filp); 4325 ssize_t ret; 4326 4327 if (!f2fs_is_compress_backend_ready(inode)) 4328 return -EOPNOTSUPP; 4329 4330 if (f2fs_should_use_dio(inode, iocb, to)) 4331 return f2fs_dio_read_iter(iocb, to); 4332 4333 ret = filemap_read(iocb, to, 0); 4334 if (ret > 0) 4335 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret); 4336 return ret; 4337 } 4338 4339 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4340 { 4341 struct file *file = iocb->ki_filp; 4342 struct inode *inode = file_inode(file); 4343 ssize_t count; 4344 int err; 4345 4346 if (IS_IMMUTABLE(inode)) 4347 return -EPERM; 4348 4349 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4350 return -EPERM; 4351 4352 count = generic_write_checks(iocb, from); 4353 if (count <= 0) 4354 return count; 4355 4356 err = file_modified(file); 4357 if (err) 4358 return err; 4359 return count; 4360 } 4361 4362 /* 4363 * Preallocate blocks for a write request, if it is possible and helpful to do 4364 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4365 * blocks were preallocated, or a negative errno value if something went 4366 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4367 * requested blocks (not just some of them) have been allocated. 4368 */ 4369 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4370 bool dio) 4371 { 4372 struct inode *inode = file_inode(iocb->ki_filp); 4373 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4374 const loff_t pos = iocb->ki_pos; 4375 const size_t count = iov_iter_count(iter); 4376 struct f2fs_map_blocks map = {}; 4377 int flag; 4378 int ret; 4379 4380 /* If it will be an out-of-place direct write, don't bother. */ 4381 if (dio && f2fs_lfs_mode(sbi)) 4382 return 0; 4383 /* 4384 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4385 * buffered IO, if DIO meets any holes. 4386 */ 4387 if (dio && i_size_read(inode) && 4388 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4389 return 0; 4390 4391 /* No-wait I/O can't allocate blocks. */ 4392 if (iocb->ki_flags & IOCB_NOWAIT) 4393 return 0; 4394 4395 /* If it will be a short write, don't bother. */ 4396 if (fault_in_iov_iter_readable(iter, count)) 4397 return 0; 4398 4399 if (f2fs_has_inline_data(inode)) { 4400 /* If the data will fit inline, don't bother. */ 4401 if (pos + count <= MAX_INLINE_DATA(inode)) 4402 return 0; 4403 ret = f2fs_convert_inline_inode(inode); 4404 if (ret) 4405 return ret; 4406 } 4407 4408 /* Do not preallocate blocks that will be written partially in 4KB. */ 4409 map.m_lblk = F2FS_BLK_ALIGN(pos); 4410 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4411 if (map.m_len > map.m_lblk) 4412 map.m_len -= map.m_lblk; 4413 else 4414 map.m_len = 0; 4415 map.m_may_create = true; 4416 if (dio) { 4417 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint); 4418 flag = F2FS_GET_BLOCK_PRE_DIO; 4419 } else { 4420 map.m_seg_type = NO_CHECK_TYPE; 4421 flag = F2FS_GET_BLOCK_PRE_AIO; 4422 } 4423 4424 ret = f2fs_map_blocks(inode, &map, 1, flag); 4425 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4426 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4427 return ret; 4428 if (ret == 0) 4429 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4430 return map.m_len; 4431 } 4432 4433 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4434 struct iov_iter *from) 4435 { 4436 struct file *file = iocb->ki_filp; 4437 struct inode *inode = file_inode(file); 4438 ssize_t ret; 4439 4440 if (iocb->ki_flags & IOCB_NOWAIT) 4441 return -EOPNOTSUPP; 4442 4443 current->backing_dev_info = inode_to_bdi(inode); 4444 ret = generic_perform_write(file, from, iocb->ki_pos); 4445 current->backing_dev_info = NULL; 4446 4447 if (ret > 0) { 4448 iocb->ki_pos += ret; 4449 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret); 4450 } 4451 return ret; 4452 } 4453 4454 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4455 unsigned int flags) 4456 { 4457 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4458 4459 dec_page_count(sbi, F2FS_DIO_WRITE); 4460 if (error) 4461 return error; 4462 f2fs_update_iostat(sbi, APP_DIRECT_IO, size); 4463 return 0; 4464 } 4465 4466 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4467 .end_io = f2fs_dio_write_end_io, 4468 }; 4469 4470 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4471 bool *may_need_sync) 4472 { 4473 struct file *file = iocb->ki_filp; 4474 struct inode *inode = file_inode(file); 4475 struct f2fs_inode_info *fi = F2FS_I(inode); 4476 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4477 const bool do_opu = f2fs_lfs_mode(sbi); 4478 const int whint_mode = F2FS_OPTION(sbi).whint_mode; 4479 const loff_t pos = iocb->ki_pos; 4480 const ssize_t count = iov_iter_count(from); 4481 const enum rw_hint hint = iocb->ki_hint; 4482 unsigned int dio_flags; 4483 struct iomap_dio *dio; 4484 ssize_t ret; 4485 4486 trace_f2fs_direct_IO_enter(inode, pos, count, WRITE); 4487 4488 if (iocb->ki_flags & IOCB_NOWAIT) { 4489 /* f2fs_convert_inline_inode() and block allocation can block */ 4490 if (f2fs_has_inline_data(inode) || 4491 !f2fs_overwrite_io(inode, pos, count)) { 4492 ret = -EAGAIN; 4493 goto out; 4494 } 4495 4496 if (!down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4497 ret = -EAGAIN; 4498 goto out; 4499 } 4500 if (do_opu && !down_read_trylock(&fi->i_gc_rwsem[READ])) { 4501 up_read(&fi->i_gc_rwsem[WRITE]); 4502 ret = -EAGAIN; 4503 goto out; 4504 } 4505 } else { 4506 ret = f2fs_convert_inline_inode(inode); 4507 if (ret) 4508 goto out; 4509 4510 down_read(&fi->i_gc_rwsem[WRITE]); 4511 if (do_opu) 4512 down_read(&fi->i_gc_rwsem[READ]); 4513 } 4514 if (whint_mode == WHINT_MODE_OFF) 4515 iocb->ki_hint = WRITE_LIFE_NOT_SET; 4516 4517 /* 4518 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4519 * the higher-level function iomap_dio_rw() in order to ensure that the 4520 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4521 */ 4522 inc_page_count(sbi, F2FS_DIO_WRITE); 4523 dio_flags = 0; 4524 if (pos + count > inode->i_size) 4525 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4526 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4527 &f2fs_iomap_dio_write_ops, dio_flags, 0); 4528 if (IS_ERR_OR_NULL(dio)) { 4529 ret = PTR_ERR_OR_ZERO(dio); 4530 if (ret == -ENOTBLK) 4531 ret = 0; 4532 if (ret != -EIOCBQUEUED) 4533 dec_page_count(sbi, F2FS_DIO_WRITE); 4534 } else { 4535 ret = iomap_dio_complete(dio); 4536 } 4537 4538 if (whint_mode == WHINT_MODE_OFF) 4539 iocb->ki_hint = hint; 4540 if (do_opu) 4541 up_read(&fi->i_gc_rwsem[READ]); 4542 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 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 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 inode *inode; 4667 struct address_space *mapping; 4668 struct backing_dev_info *bdi; 4669 4670 if (advice == POSIX_FADV_SEQUENTIAL) { 4671 inode = file_inode(filp); 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 return generic_fadvise(filp, offset, len, advice); 4689 } 4690 4691 #ifdef CONFIG_COMPAT 4692 struct compat_f2fs_gc_range { 4693 u32 sync; 4694 compat_u64 start; 4695 compat_u64 len; 4696 }; 4697 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4698 struct compat_f2fs_gc_range) 4699 4700 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4701 { 4702 struct compat_f2fs_gc_range __user *urange; 4703 struct f2fs_gc_range range; 4704 int err; 4705 4706 urange = compat_ptr(arg); 4707 err = get_user(range.sync, &urange->sync); 4708 err |= get_user(range.start, &urange->start); 4709 err |= get_user(range.len, &urange->len); 4710 if (err) 4711 return -EFAULT; 4712 4713 return __f2fs_ioc_gc_range(file, &range); 4714 } 4715 4716 struct compat_f2fs_move_range { 4717 u32 dst_fd; 4718 compat_u64 pos_in; 4719 compat_u64 pos_out; 4720 compat_u64 len; 4721 }; 4722 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4723 struct compat_f2fs_move_range) 4724 4725 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4726 { 4727 struct compat_f2fs_move_range __user *urange; 4728 struct f2fs_move_range range; 4729 int err; 4730 4731 urange = compat_ptr(arg); 4732 err = get_user(range.dst_fd, &urange->dst_fd); 4733 err |= get_user(range.pos_in, &urange->pos_in); 4734 err |= get_user(range.pos_out, &urange->pos_out); 4735 err |= get_user(range.len, &urange->len); 4736 if (err) 4737 return -EFAULT; 4738 4739 return __f2fs_ioc_move_range(file, &range); 4740 } 4741 4742 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4743 { 4744 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4745 return -EIO; 4746 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4747 return -ENOSPC; 4748 4749 switch (cmd) { 4750 case FS_IOC32_GETVERSION: 4751 cmd = FS_IOC_GETVERSION; 4752 break; 4753 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4754 return f2fs_compat_ioc_gc_range(file, arg); 4755 case F2FS_IOC32_MOVE_RANGE: 4756 return f2fs_compat_ioc_move_range(file, arg); 4757 case F2FS_IOC_START_ATOMIC_WRITE: 4758 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4759 case F2FS_IOC_START_VOLATILE_WRITE: 4760 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4761 case F2FS_IOC_ABORT_VOLATILE_WRITE: 4762 case F2FS_IOC_SHUTDOWN: 4763 case FITRIM: 4764 case FS_IOC_SET_ENCRYPTION_POLICY: 4765 case FS_IOC_GET_ENCRYPTION_PWSALT: 4766 case FS_IOC_GET_ENCRYPTION_POLICY: 4767 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4768 case FS_IOC_ADD_ENCRYPTION_KEY: 4769 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4770 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4771 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4772 case FS_IOC_GET_ENCRYPTION_NONCE: 4773 case F2FS_IOC_GARBAGE_COLLECT: 4774 case F2FS_IOC_WRITE_CHECKPOINT: 4775 case F2FS_IOC_DEFRAGMENT: 4776 case F2FS_IOC_FLUSH_DEVICE: 4777 case F2FS_IOC_GET_FEATURES: 4778 case F2FS_IOC_GET_PIN_FILE: 4779 case F2FS_IOC_SET_PIN_FILE: 4780 case F2FS_IOC_PRECACHE_EXTENTS: 4781 case F2FS_IOC_RESIZE_FS: 4782 case FS_IOC_ENABLE_VERITY: 4783 case FS_IOC_MEASURE_VERITY: 4784 case FS_IOC_READ_VERITY_METADATA: 4785 case FS_IOC_GETFSLABEL: 4786 case FS_IOC_SETFSLABEL: 4787 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4788 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4789 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4790 case F2FS_IOC_SEC_TRIM_FILE: 4791 case F2FS_IOC_GET_COMPRESS_OPTION: 4792 case F2FS_IOC_SET_COMPRESS_OPTION: 4793 case F2FS_IOC_DECOMPRESS_FILE: 4794 case F2FS_IOC_COMPRESS_FILE: 4795 break; 4796 default: 4797 return -ENOIOCTLCMD; 4798 } 4799 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4800 } 4801 #endif 4802 4803 const struct file_operations f2fs_file_operations = { 4804 .llseek = f2fs_llseek, 4805 .read_iter = f2fs_file_read_iter, 4806 .write_iter = f2fs_file_write_iter, 4807 .open = f2fs_file_open, 4808 .release = f2fs_release_file, 4809 .mmap = f2fs_file_mmap, 4810 .flush = f2fs_file_flush, 4811 .fsync = f2fs_sync_file, 4812 .fallocate = f2fs_fallocate, 4813 .unlocked_ioctl = f2fs_ioctl, 4814 #ifdef CONFIG_COMPAT 4815 .compat_ioctl = f2fs_compat_ioctl, 4816 #endif 4817 .splice_read = generic_file_splice_read, 4818 .splice_write = iter_file_splice_write, 4819 .fadvise = f2fs_file_fadvise, 4820 }; 4821