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 28 #include "f2fs.h" 29 #include "node.h" 30 #include "segment.h" 31 #include "xattr.h" 32 #include "acl.h" 33 #include "gc.h" 34 #include "iostat.h" 35 #include <trace/events/f2fs.h> 36 #include <uapi/linux/f2fs.h> 37 38 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 39 { 40 struct inode *inode = file_inode(vmf->vma->vm_file); 41 vm_fault_t ret; 42 43 ret = filemap_fault(vmf); 44 if (!ret) 45 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO, 46 F2FS_BLKSIZE); 47 48 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret); 49 50 return ret; 51 } 52 53 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 54 { 55 struct page *page = vmf->page; 56 struct inode *inode = file_inode(vmf->vma->vm_file); 57 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 58 struct dnode_of_data dn; 59 bool need_alloc = true; 60 int err = 0; 61 62 if (unlikely(IS_IMMUTABLE(inode))) 63 return VM_FAULT_SIGBUS; 64 65 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 66 return VM_FAULT_SIGBUS; 67 68 if (unlikely(f2fs_cp_error(sbi))) { 69 err = -EIO; 70 goto err; 71 } 72 73 if (!f2fs_is_checkpoint_ready(sbi)) { 74 err = -ENOSPC; 75 goto err; 76 } 77 78 err = f2fs_convert_inline_inode(inode); 79 if (err) 80 goto err; 81 82 #ifdef CONFIG_F2FS_FS_COMPRESSION 83 if (f2fs_compressed_file(inode)) { 84 int ret = f2fs_is_compressed_cluster(inode, page->index); 85 86 if (ret < 0) { 87 err = ret; 88 goto err; 89 } else if (ret) { 90 need_alloc = false; 91 } 92 } 93 #endif 94 /* should do out of any locked page */ 95 if (need_alloc) 96 f2fs_balance_fs(sbi, true); 97 98 sb_start_pagefault(inode->i_sb); 99 100 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 101 102 file_update_time(vmf->vma->vm_file); 103 filemap_invalidate_lock_shared(inode->i_mapping); 104 lock_page(page); 105 if (unlikely(page->mapping != inode->i_mapping || 106 page_offset(page) > i_size_read(inode) || 107 !PageUptodate(page))) { 108 unlock_page(page); 109 err = -EFAULT; 110 goto out_sem; 111 } 112 113 if (need_alloc) { 114 /* block allocation */ 115 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true); 116 set_new_dnode(&dn, inode, NULL, NULL, 0); 117 err = f2fs_get_block(&dn, page->index); 118 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false); 119 } 120 121 #ifdef CONFIG_F2FS_FS_COMPRESSION 122 if (!need_alloc) { 123 set_new_dnode(&dn, inode, NULL, NULL, 0); 124 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE); 125 f2fs_put_dnode(&dn); 126 } 127 #endif 128 if (err) { 129 unlock_page(page); 130 goto out_sem; 131 } 132 133 f2fs_wait_on_page_writeback(page, DATA, false, true); 134 135 /* wait for GCed page writeback via META_MAPPING */ 136 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 137 138 /* 139 * check to see if the page is mapped already (no holes) 140 */ 141 if (PageMappedToDisk(page)) 142 goto out_sem; 143 144 /* page is wholly or partially inside EOF */ 145 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 146 i_size_read(inode)) { 147 loff_t offset; 148 149 offset = i_size_read(inode) & ~PAGE_MASK; 150 zero_user_segment(page, offset, PAGE_SIZE); 151 } 152 set_page_dirty(page); 153 if (!PageUptodate(page)) 154 SetPageUptodate(page); 155 156 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE); 157 f2fs_update_time(sbi, REQ_TIME); 158 159 trace_f2fs_vm_page_mkwrite(page, DATA); 160 out_sem: 161 filemap_invalidate_unlock_shared(inode->i_mapping); 162 163 sb_end_pagefault(inode->i_sb); 164 err: 165 return block_page_mkwrite_return(err); 166 } 167 168 static const struct vm_operations_struct f2fs_file_vm_ops = { 169 .fault = f2fs_filemap_fault, 170 .map_pages = filemap_map_pages, 171 .page_mkwrite = f2fs_vm_page_mkwrite, 172 }; 173 174 static int get_parent_ino(struct inode *inode, nid_t *pino) 175 { 176 struct dentry *dentry; 177 178 /* 179 * Make sure to get the non-deleted alias. The alias associated with 180 * the open file descriptor being fsync()'ed may be deleted already. 181 */ 182 dentry = d_find_alias(inode); 183 if (!dentry) 184 return 0; 185 186 *pino = parent_ino(dentry); 187 dput(dentry); 188 return 1; 189 } 190 191 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 192 { 193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 194 enum cp_reason_type cp_reason = CP_NO_NEEDED; 195 196 if (!S_ISREG(inode->i_mode)) 197 cp_reason = CP_NON_REGULAR; 198 else if (f2fs_compressed_file(inode)) 199 cp_reason = CP_COMPRESSED; 200 else if (inode->i_nlink != 1) 201 cp_reason = CP_HARDLINK; 202 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 203 cp_reason = CP_SB_NEED_CP; 204 else if (file_wrong_pino(inode)) 205 cp_reason = CP_WRONG_PINO; 206 else if (!f2fs_space_for_roll_forward(sbi)) 207 cp_reason = CP_NO_SPC_ROLL; 208 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 209 cp_reason = CP_NODE_NEED_CP; 210 else if (test_opt(sbi, FASTBOOT)) 211 cp_reason = CP_FASTBOOT_MODE; 212 else if (F2FS_OPTION(sbi).active_logs == 2) 213 cp_reason = CP_SPEC_LOG_NUM; 214 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 215 f2fs_need_dentry_mark(sbi, inode->i_ino) && 216 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 217 TRANS_DIR_INO)) 218 cp_reason = CP_RECOVER_DIR; 219 220 return cp_reason; 221 } 222 223 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 224 { 225 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 226 bool ret = false; 227 /* But we need to avoid that there are some inode updates */ 228 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 229 ret = true; 230 f2fs_put_page(i, 0); 231 return ret; 232 } 233 234 static void try_to_fix_pino(struct inode *inode) 235 { 236 struct f2fs_inode_info *fi = F2FS_I(inode); 237 nid_t pino; 238 239 down_write(&fi->i_sem); 240 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 241 get_parent_ino(inode, &pino)) { 242 f2fs_i_pino_write(inode, pino); 243 file_got_pino(inode); 244 } 245 up_write(&fi->i_sem); 246 } 247 248 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 249 int datasync, bool atomic) 250 { 251 struct inode *inode = file->f_mapping->host; 252 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 253 nid_t ino = inode->i_ino; 254 int ret = 0; 255 enum cp_reason_type cp_reason = 0; 256 struct writeback_control wbc = { 257 .sync_mode = WB_SYNC_ALL, 258 .nr_to_write = LONG_MAX, 259 .for_reclaim = 0, 260 }; 261 unsigned int seq_id = 0; 262 263 if (unlikely(f2fs_readonly(inode->i_sb))) 264 return 0; 265 266 trace_f2fs_sync_file_enter(inode); 267 268 if (S_ISDIR(inode->i_mode)) 269 goto go_write; 270 271 /* if fdatasync is triggered, let's do in-place-update */ 272 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 273 set_inode_flag(inode, FI_NEED_IPU); 274 ret = file_write_and_wait_range(file, start, end); 275 clear_inode_flag(inode, FI_NEED_IPU); 276 277 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 278 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 279 return ret; 280 } 281 282 /* if the inode is dirty, let's recover all the time */ 283 if (!f2fs_skip_inode_update(inode, datasync)) { 284 f2fs_write_inode(inode, NULL); 285 goto go_write; 286 } 287 288 /* 289 * if there is no written data, don't waste time to write recovery info. 290 */ 291 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 292 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 293 294 /* it may call write_inode just prior to fsync */ 295 if (need_inode_page_update(sbi, ino)) 296 goto go_write; 297 298 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 299 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 300 goto flush_out; 301 goto out; 302 } else { 303 /* 304 * for OPU case, during fsync(), node can be persisted before 305 * data when lower device doesn't support write barrier, result 306 * in data corruption after SPO. 307 * So for strict fsync mode, force to use atomic write sematics 308 * to keep write order in between data/node and last node to 309 * avoid potential data corruption. 310 */ 311 if (F2FS_OPTION(sbi).fsync_mode == 312 FSYNC_MODE_STRICT && !atomic) 313 atomic = true; 314 } 315 go_write: 316 /* 317 * Both of fdatasync() and fsync() are able to be recovered from 318 * sudden-power-off. 319 */ 320 down_read(&F2FS_I(inode)->i_sem); 321 cp_reason = need_do_checkpoint(inode); 322 up_read(&F2FS_I(inode)->i_sem); 323 324 if (cp_reason) { 325 /* all the dirty node pages should be flushed for POR */ 326 ret = f2fs_sync_fs(inode->i_sb, 1); 327 328 /* 329 * We've secured consistency through sync_fs. Following pino 330 * will be used only for fsynced inodes after checkpoint. 331 */ 332 try_to_fix_pino(inode); 333 clear_inode_flag(inode, FI_APPEND_WRITE); 334 clear_inode_flag(inode, FI_UPDATE_WRITE); 335 goto out; 336 } 337 sync_nodes: 338 atomic_inc(&sbi->wb_sync_req[NODE]); 339 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 340 atomic_dec(&sbi->wb_sync_req[NODE]); 341 if (ret) 342 goto out; 343 344 /* if cp_error was enabled, we should avoid infinite loop */ 345 if (unlikely(f2fs_cp_error(sbi))) { 346 ret = -EIO; 347 goto out; 348 } 349 350 if (f2fs_need_inode_block_update(sbi, ino)) { 351 f2fs_mark_inode_dirty_sync(inode, true); 352 f2fs_write_inode(inode, NULL); 353 goto sync_nodes; 354 } 355 356 /* 357 * If it's atomic_write, it's just fine to keep write ordering. So 358 * here we don't need to wait for node write completion, since we use 359 * node chain which serializes node blocks. If one of node writes are 360 * reordered, we can see simply broken chain, resulting in stopping 361 * roll-forward recovery. It means we'll recover all or none node blocks 362 * given fsync mark. 363 */ 364 if (!atomic) { 365 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 366 if (ret) 367 goto out; 368 } 369 370 /* once recovery info is written, don't need to tack this */ 371 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 372 clear_inode_flag(inode, FI_APPEND_WRITE); 373 flush_out: 374 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 375 ret = f2fs_issue_flush(sbi, inode->i_ino); 376 if (!ret) { 377 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 378 clear_inode_flag(inode, FI_UPDATE_WRITE); 379 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 380 } 381 f2fs_update_time(sbi, REQ_TIME); 382 out: 383 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 384 return ret; 385 } 386 387 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 388 { 389 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 390 return -EIO; 391 return f2fs_do_sync_file(file, start, end, datasync, false); 392 } 393 394 static bool __found_offset(struct address_space *mapping, block_t blkaddr, 395 pgoff_t index, int whence) 396 { 397 switch (whence) { 398 case SEEK_DATA: 399 if (__is_valid_data_blkaddr(blkaddr)) 400 return true; 401 if (blkaddr == NEW_ADDR && 402 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 403 return true; 404 break; 405 case SEEK_HOLE: 406 if (blkaddr == NULL_ADDR) 407 return true; 408 break; 409 } 410 return false; 411 } 412 413 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 414 { 415 struct inode *inode = file->f_mapping->host; 416 loff_t maxbytes = inode->i_sb->s_maxbytes; 417 struct dnode_of_data dn; 418 pgoff_t pgofs, end_offset; 419 loff_t data_ofs = offset; 420 loff_t isize; 421 int err = 0; 422 423 inode_lock(inode); 424 425 isize = i_size_read(inode); 426 if (offset >= isize) 427 goto fail; 428 429 /* handle inline data case */ 430 if (f2fs_has_inline_data(inode)) { 431 if (whence == SEEK_HOLE) { 432 data_ofs = isize; 433 goto found; 434 } else if (whence == SEEK_DATA) { 435 data_ofs = offset; 436 goto found; 437 } 438 } 439 440 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 441 442 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 443 set_new_dnode(&dn, inode, NULL, NULL, 0); 444 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 445 if (err && err != -ENOENT) { 446 goto fail; 447 } else if (err == -ENOENT) { 448 /* direct node does not exists */ 449 if (whence == SEEK_DATA) { 450 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 451 continue; 452 } else { 453 goto found; 454 } 455 } 456 457 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 458 459 /* find data/hole in dnode block */ 460 for (; dn.ofs_in_node < end_offset; 461 dn.ofs_in_node++, pgofs++, 462 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 463 block_t blkaddr; 464 465 blkaddr = f2fs_data_blkaddr(&dn); 466 467 if (__is_valid_data_blkaddr(blkaddr) && 468 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 469 blkaddr, DATA_GENERIC_ENHANCE)) { 470 f2fs_put_dnode(&dn); 471 goto fail; 472 } 473 474 if (__found_offset(file->f_mapping, blkaddr, 475 pgofs, whence)) { 476 f2fs_put_dnode(&dn); 477 goto found; 478 } 479 } 480 f2fs_put_dnode(&dn); 481 } 482 483 if (whence == SEEK_DATA) 484 goto fail; 485 found: 486 if (whence == SEEK_HOLE && data_ofs > isize) 487 data_ofs = isize; 488 inode_unlock(inode); 489 return vfs_setpos(file, data_ofs, maxbytes); 490 fail: 491 inode_unlock(inode); 492 return -ENXIO; 493 } 494 495 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 496 { 497 struct inode *inode = file->f_mapping->host; 498 loff_t maxbytes = inode->i_sb->s_maxbytes; 499 500 if (f2fs_compressed_file(inode)) 501 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; 502 503 switch (whence) { 504 case SEEK_SET: 505 case SEEK_CUR: 506 case SEEK_END: 507 return generic_file_llseek_size(file, offset, whence, 508 maxbytes, i_size_read(inode)); 509 case SEEK_DATA: 510 case SEEK_HOLE: 511 if (offset < 0) 512 return -ENXIO; 513 return f2fs_seek_block(file, offset, whence); 514 } 515 516 return -EINVAL; 517 } 518 519 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 520 { 521 struct inode *inode = file_inode(file); 522 523 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 524 return -EIO; 525 526 if (!f2fs_is_compress_backend_ready(inode)) 527 return -EOPNOTSUPP; 528 529 file_accessed(file); 530 vma->vm_ops = &f2fs_file_vm_ops; 531 set_inode_flag(inode, FI_MMAP_FILE); 532 return 0; 533 } 534 535 static int f2fs_file_open(struct inode *inode, struct file *filp) 536 { 537 int err = fscrypt_file_open(inode, filp); 538 539 if (err) 540 return err; 541 542 if (!f2fs_is_compress_backend_ready(inode)) 543 return -EOPNOTSUPP; 544 545 err = fsverity_file_open(inode, filp); 546 if (err) 547 return err; 548 549 filp->f_mode |= FMODE_NOWAIT; 550 551 return dquot_file_open(inode, filp); 552 } 553 554 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 555 { 556 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 557 struct f2fs_node *raw_node; 558 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 559 __le32 *addr; 560 int base = 0; 561 bool compressed_cluster = false; 562 int cluster_index = 0, valid_blocks = 0; 563 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 564 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 565 566 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode)) 567 base = get_extra_isize(dn->inode); 568 569 raw_node = F2FS_NODE(dn->node_page); 570 addr = blkaddr_in_node(raw_node) + base + ofs; 571 572 /* Assumption: truncateion starts with cluster */ 573 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 574 block_t blkaddr = le32_to_cpu(*addr); 575 576 if (f2fs_compressed_file(dn->inode) && 577 !(cluster_index & (cluster_size - 1))) { 578 if (compressed_cluster) 579 f2fs_i_compr_blocks_update(dn->inode, 580 valid_blocks, false); 581 compressed_cluster = (blkaddr == COMPRESS_ADDR); 582 valid_blocks = 0; 583 } 584 585 if (blkaddr == NULL_ADDR) 586 continue; 587 588 dn->data_blkaddr = NULL_ADDR; 589 f2fs_set_data_blkaddr(dn); 590 591 if (__is_valid_data_blkaddr(blkaddr)) { 592 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 593 DATA_GENERIC_ENHANCE)) 594 continue; 595 if (compressed_cluster) 596 valid_blocks++; 597 } 598 599 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 600 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 601 602 f2fs_invalidate_blocks(sbi, blkaddr); 603 604 if (!released || blkaddr != COMPRESS_ADDR) 605 nr_free++; 606 } 607 608 if (compressed_cluster) 609 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 610 611 if (nr_free) { 612 pgoff_t fofs; 613 /* 614 * once we invalidate valid blkaddr in range [ofs, ofs + count], 615 * we will invalidate all blkaddr in the whole range. 616 */ 617 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 618 dn->inode) + ofs; 619 f2fs_update_extent_cache_range(dn, fofs, 0, len); 620 dec_valid_block_count(sbi, dn->inode, nr_free); 621 } 622 dn->ofs_in_node = ofs; 623 624 f2fs_update_time(sbi, REQ_TIME); 625 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 626 dn->ofs_in_node, nr_free); 627 } 628 629 void f2fs_truncate_data_blocks(struct dnode_of_data *dn) 630 { 631 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode)); 632 } 633 634 static int truncate_partial_data_page(struct inode *inode, u64 from, 635 bool cache_only) 636 { 637 loff_t offset = from & (PAGE_SIZE - 1); 638 pgoff_t index = from >> PAGE_SHIFT; 639 struct address_space *mapping = inode->i_mapping; 640 struct page *page; 641 642 if (!offset && !cache_only) 643 return 0; 644 645 if (cache_only) { 646 page = find_lock_page(mapping, index); 647 if (page && PageUptodate(page)) 648 goto truncate_out; 649 f2fs_put_page(page, 1); 650 return 0; 651 } 652 653 page = f2fs_get_lock_data_page(inode, index, true); 654 if (IS_ERR(page)) 655 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 656 truncate_out: 657 f2fs_wait_on_page_writeback(page, DATA, true, true); 658 zero_user(page, offset, PAGE_SIZE - offset); 659 660 /* An encrypted inode should have a key and truncate the last page. */ 661 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 662 if (!cache_only) 663 set_page_dirty(page); 664 f2fs_put_page(page, 1); 665 return 0; 666 } 667 668 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 669 { 670 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 671 struct dnode_of_data dn; 672 pgoff_t free_from; 673 int count = 0, err = 0; 674 struct page *ipage; 675 bool truncate_page = false; 676 677 trace_f2fs_truncate_blocks_enter(inode, from); 678 679 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 680 681 if (free_from >= max_file_blocks(inode)) 682 goto free_partial; 683 684 if (lock) 685 f2fs_lock_op(sbi); 686 687 ipage = f2fs_get_node_page(sbi, inode->i_ino); 688 if (IS_ERR(ipage)) { 689 err = PTR_ERR(ipage); 690 goto out; 691 } 692 693 if (f2fs_has_inline_data(inode)) { 694 f2fs_truncate_inline_inode(inode, ipage, from); 695 f2fs_put_page(ipage, 1); 696 truncate_page = true; 697 goto out; 698 } 699 700 set_new_dnode(&dn, inode, ipage, NULL, 0); 701 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 702 if (err) { 703 if (err == -ENOENT) 704 goto free_next; 705 goto out; 706 } 707 708 count = ADDRS_PER_PAGE(dn.node_page, inode); 709 710 count -= dn.ofs_in_node; 711 f2fs_bug_on(sbi, count < 0); 712 713 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 714 f2fs_truncate_data_blocks_range(&dn, count); 715 free_from += count; 716 } 717 718 f2fs_put_dnode(&dn); 719 free_next: 720 err = f2fs_truncate_inode_blocks(inode, free_from); 721 out: 722 if (lock) 723 f2fs_unlock_op(sbi); 724 free_partial: 725 /* lastly zero out the first data page */ 726 if (!err) 727 err = truncate_partial_data_page(inode, from, truncate_page); 728 729 trace_f2fs_truncate_blocks_exit(inode, err); 730 return err; 731 } 732 733 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 734 { 735 u64 free_from = from; 736 int err; 737 738 #ifdef CONFIG_F2FS_FS_COMPRESSION 739 /* 740 * for compressed file, only support cluster size 741 * aligned truncation. 742 */ 743 if (f2fs_compressed_file(inode)) 744 free_from = round_up(from, 745 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 746 #endif 747 748 err = f2fs_do_truncate_blocks(inode, free_from, lock); 749 if (err) 750 return err; 751 752 #ifdef CONFIG_F2FS_FS_COMPRESSION 753 /* 754 * For compressed file, after release compress blocks, don't allow write 755 * direct, but we should allow write direct after truncate to zero. 756 */ 757 if (f2fs_compressed_file(inode) && !free_from 758 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 759 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 760 761 if (from != free_from) { 762 err = f2fs_truncate_partial_cluster(inode, from, lock); 763 if (err) 764 return err; 765 } 766 #endif 767 768 return 0; 769 } 770 771 int f2fs_truncate(struct inode *inode) 772 { 773 int err; 774 775 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 776 return -EIO; 777 778 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 779 S_ISLNK(inode->i_mode))) 780 return 0; 781 782 trace_f2fs_truncate(inode); 783 784 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) { 785 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE); 786 return -EIO; 787 } 788 789 err = dquot_initialize(inode); 790 if (err) 791 return err; 792 793 /* we should check inline_data size */ 794 if (!f2fs_may_inline_data(inode)) { 795 err = f2fs_convert_inline_inode(inode); 796 if (err) 797 return err; 798 } 799 800 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 801 if (err) 802 return err; 803 804 inode->i_mtime = inode->i_ctime = current_time(inode); 805 f2fs_mark_inode_dirty_sync(inode, false); 806 return 0; 807 } 808 809 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path, 810 struct kstat *stat, u32 request_mask, unsigned int query_flags) 811 { 812 struct inode *inode = d_inode(path->dentry); 813 struct f2fs_inode_info *fi = F2FS_I(inode); 814 struct f2fs_inode *ri; 815 unsigned int flags; 816 817 if (f2fs_has_extra_attr(inode) && 818 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 819 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 820 stat->result_mask |= STATX_BTIME; 821 stat->btime.tv_sec = fi->i_crtime.tv_sec; 822 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 823 } 824 825 flags = fi->i_flags; 826 if (flags & F2FS_COMPR_FL) 827 stat->attributes |= STATX_ATTR_COMPRESSED; 828 if (flags & F2FS_APPEND_FL) 829 stat->attributes |= STATX_ATTR_APPEND; 830 if (IS_ENCRYPTED(inode)) 831 stat->attributes |= STATX_ATTR_ENCRYPTED; 832 if (flags & F2FS_IMMUTABLE_FL) 833 stat->attributes |= STATX_ATTR_IMMUTABLE; 834 if (flags & F2FS_NODUMP_FL) 835 stat->attributes |= STATX_ATTR_NODUMP; 836 if (IS_VERITY(inode)) 837 stat->attributes |= STATX_ATTR_VERITY; 838 839 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 840 STATX_ATTR_APPEND | 841 STATX_ATTR_ENCRYPTED | 842 STATX_ATTR_IMMUTABLE | 843 STATX_ATTR_NODUMP | 844 STATX_ATTR_VERITY); 845 846 generic_fillattr(&init_user_ns, inode, stat); 847 848 /* we need to show initial sectors used for inline_data/dentries */ 849 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 850 f2fs_has_inline_dentry(inode)) 851 stat->blocks += (stat->size + 511) >> 9; 852 853 return 0; 854 } 855 856 #ifdef CONFIG_F2FS_FS_POSIX_ACL 857 static void __setattr_copy(struct user_namespace *mnt_userns, 858 struct inode *inode, const struct iattr *attr) 859 { 860 unsigned int ia_valid = attr->ia_valid; 861 862 if (ia_valid & ATTR_UID) 863 inode->i_uid = attr->ia_uid; 864 if (ia_valid & ATTR_GID) 865 inode->i_gid = attr->ia_gid; 866 if (ia_valid & ATTR_ATIME) 867 inode->i_atime = attr->ia_atime; 868 if (ia_valid & ATTR_MTIME) 869 inode->i_mtime = attr->ia_mtime; 870 if (ia_valid & ATTR_CTIME) 871 inode->i_ctime = attr->ia_ctime; 872 if (ia_valid & ATTR_MODE) { 873 umode_t mode = attr->ia_mode; 874 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode); 875 876 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID)) 877 mode &= ~S_ISGID; 878 set_acl_inode(inode, mode); 879 } 880 } 881 #else 882 #define __setattr_copy setattr_copy 883 #endif 884 885 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry, 886 struct iattr *attr) 887 { 888 struct inode *inode = d_inode(dentry); 889 int err; 890 891 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 892 return -EIO; 893 894 if (unlikely(IS_IMMUTABLE(inode))) 895 return -EPERM; 896 897 if (unlikely(IS_APPEND(inode) && 898 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 899 ATTR_GID | ATTR_TIMES_SET)))) 900 return -EPERM; 901 902 if ((attr->ia_valid & ATTR_SIZE) && 903 !f2fs_is_compress_backend_ready(inode)) 904 return -EOPNOTSUPP; 905 906 err = setattr_prepare(&init_user_ns, dentry, attr); 907 if (err) 908 return err; 909 910 err = fscrypt_prepare_setattr(dentry, attr); 911 if (err) 912 return err; 913 914 err = fsverity_prepare_setattr(dentry, attr); 915 if (err) 916 return err; 917 918 if (is_quota_modification(inode, attr)) { 919 err = dquot_initialize(inode); 920 if (err) 921 return err; 922 } 923 if ((attr->ia_valid & ATTR_UID && 924 !uid_eq(attr->ia_uid, inode->i_uid)) || 925 (attr->ia_valid & ATTR_GID && 926 !gid_eq(attr->ia_gid, inode->i_gid))) { 927 f2fs_lock_op(F2FS_I_SB(inode)); 928 err = dquot_transfer(inode, attr); 929 if (err) { 930 set_sbi_flag(F2FS_I_SB(inode), 931 SBI_QUOTA_NEED_REPAIR); 932 f2fs_unlock_op(F2FS_I_SB(inode)); 933 return err; 934 } 935 /* 936 * update uid/gid under lock_op(), so that dquot and inode can 937 * be updated atomically. 938 */ 939 if (attr->ia_valid & ATTR_UID) 940 inode->i_uid = attr->ia_uid; 941 if (attr->ia_valid & ATTR_GID) 942 inode->i_gid = attr->ia_gid; 943 f2fs_mark_inode_dirty_sync(inode, true); 944 f2fs_unlock_op(F2FS_I_SB(inode)); 945 } 946 947 if (attr->ia_valid & ATTR_SIZE) { 948 loff_t old_size = i_size_read(inode); 949 950 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 951 /* 952 * should convert inline inode before i_size_write to 953 * keep smaller than inline_data size with inline flag. 954 */ 955 err = f2fs_convert_inline_inode(inode); 956 if (err) 957 return err; 958 } 959 960 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 961 filemap_invalidate_lock(inode->i_mapping); 962 963 truncate_setsize(inode, attr->ia_size); 964 965 if (attr->ia_size <= old_size) 966 err = f2fs_truncate(inode); 967 /* 968 * do not trim all blocks after i_size if target size is 969 * larger than i_size. 970 */ 971 filemap_invalidate_unlock(inode->i_mapping); 972 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 973 if (err) 974 return err; 975 976 spin_lock(&F2FS_I(inode)->i_size_lock); 977 inode->i_mtime = inode->i_ctime = current_time(inode); 978 F2FS_I(inode)->last_disk_size = i_size_read(inode); 979 spin_unlock(&F2FS_I(inode)->i_size_lock); 980 } 981 982 __setattr_copy(&init_user_ns, inode, attr); 983 984 if (attr->ia_valid & ATTR_MODE) { 985 err = posix_acl_chmod(&init_user_ns, inode, f2fs_get_inode_mode(inode)); 986 987 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 988 if (!err) 989 inode->i_mode = F2FS_I(inode)->i_acl_mode; 990 clear_inode_flag(inode, FI_ACL_MODE); 991 } 992 } 993 994 /* file size may changed here */ 995 f2fs_mark_inode_dirty_sync(inode, true); 996 997 /* inode change will produce dirty node pages flushed by checkpoint */ 998 f2fs_balance_fs(F2FS_I_SB(inode), true); 999 1000 return err; 1001 } 1002 1003 const struct inode_operations f2fs_file_inode_operations = { 1004 .getattr = f2fs_getattr, 1005 .setattr = f2fs_setattr, 1006 .get_acl = f2fs_get_acl, 1007 .set_acl = f2fs_set_acl, 1008 .listxattr = f2fs_listxattr, 1009 .fiemap = f2fs_fiemap, 1010 .fileattr_get = f2fs_fileattr_get, 1011 .fileattr_set = f2fs_fileattr_set, 1012 }; 1013 1014 static int fill_zero(struct inode *inode, pgoff_t index, 1015 loff_t start, loff_t len) 1016 { 1017 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1018 struct page *page; 1019 1020 if (!len) 1021 return 0; 1022 1023 f2fs_balance_fs(sbi, true); 1024 1025 f2fs_lock_op(sbi); 1026 page = f2fs_get_new_data_page(inode, NULL, index, false); 1027 f2fs_unlock_op(sbi); 1028 1029 if (IS_ERR(page)) 1030 return PTR_ERR(page); 1031 1032 f2fs_wait_on_page_writeback(page, DATA, true, true); 1033 zero_user(page, start, len); 1034 set_page_dirty(page); 1035 f2fs_put_page(page, 1); 1036 return 0; 1037 } 1038 1039 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1040 { 1041 int err; 1042 1043 while (pg_start < pg_end) { 1044 struct dnode_of_data dn; 1045 pgoff_t end_offset, count; 1046 1047 set_new_dnode(&dn, inode, NULL, NULL, 0); 1048 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1049 if (err) { 1050 if (err == -ENOENT) { 1051 pg_start = f2fs_get_next_page_offset(&dn, 1052 pg_start); 1053 continue; 1054 } 1055 return err; 1056 } 1057 1058 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1059 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1060 1061 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1062 1063 f2fs_truncate_data_blocks_range(&dn, count); 1064 f2fs_put_dnode(&dn); 1065 1066 pg_start += count; 1067 } 1068 return 0; 1069 } 1070 1071 static int punch_hole(struct inode *inode, loff_t offset, loff_t len) 1072 { 1073 pgoff_t pg_start, pg_end; 1074 loff_t off_start, off_end; 1075 int ret; 1076 1077 ret = f2fs_convert_inline_inode(inode); 1078 if (ret) 1079 return ret; 1080 1081 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1082 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1083 1084 off_start = offset & (PAGE_SIZE - 1); 1085 off_end = (offset + len) & (PAGE_SIZE - 1); 1086 1087 if (pg_start == pg_end) { 1088 ret = fill_zero(inode, pg_start, off_start, 1089 off_end - off_start); 1090 if (ret) 1091 return ret; 1092 } else { 1093 if (off_start) { 1094 ret = fill_zero(inode, pg_start++, off_start, 1095 PAGE_SIZE - off_start); 1096 if (ret) 1097 return ret; 1098 } 1099 if (off_end) { 1100 ret = fill_zero(inode, pg_end, 0, off_end); 1101 if (ret) 1102 return ret; 1103 } 1104 1105 if (pg_start < pg_end) { 1106 loff_t blk_start, blk_end; 1107 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1108 1109 f2fs_balance_fs(sbi, true); 1110 1111 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1112 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1113 1114 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1115 filemap_invalidate_lock(inode->i_mapping); 1116 1117 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1118 1119 f2fs_lock_op(sbi); 1120 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1121 f2fs_unlock_op(sbi); 1122 1123 filemap_invalidate_unlock(inode->i_mapping); 1124 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1125 } 1126 } 1127 1128 return ret; 1129 } 1130 1131 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1132 int *do_replace, pgoff_t off, pgoff_t len) 1133 { 1134 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1135 struct dnode_of_data dn; 1136 int ret, done, i; 1137 1138 next_dnode: 1139 set_new_dnode(&dn, inode, NULL, NULL, 0); 1140 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1141 if (ret && ret != -ENOENT) { 1142 return ret; 1143 } else if (ret == -ENOENT) { 1144 if (dn.max_level == 0) 1145 return -ENOENT; 1146 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1147 dn.ofs_in_node, len); 1148 blkaddr += done; 1149 do_replace += done; 1150 goto next; 1151 } 1152 1153 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1154 dn.ofs_in_node, len); 1155 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1156 *blkaddr = f2fs_data_blkaddr(&dn); 1157 1158 if (__is_valid_data_blkaddr(*blkaddr) && 1159 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1160 DATA_GENERIC_ENHANCE)) { 1161 f2fs_put_dnode(&dn); 1162 return -EFSCORRUPTED; 1163 } 1164 1165 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1166 1167 if (f2fs_lfs_mode(sbi)) { 1168 f2fs_put_dnode(&dn); 1169 return -EOPNOTSUPP; 1170 } 1171 1172 /* do not invalidate this block address */ 1173 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1174 *do_replace = 1; 1175 } 1176 } 1177 f2fs_put_dnode(&dn); 1178 next: 1179 len -= done; 1180 off += done; 1181 if (len) 1182 goto next_dnode; 1183 return 0; 1184 } 1185 1186 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1187 int *do_replace, pgoff_t off, int len) 1188 { 1189 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1190 struct dnode_of_data dn; 1191 int ret, i; 1192 1193 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1194 if (*do_replace == 0) 1195 continue; 1196 1197 set_new_dnode(&dn, inode, NULL, NULL, 0); 1198 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1199 if (ret) { 1200 dec_valid_block_count(sbi, inode, 1); 1201 f2fs_invalidate_blocks(sbi, *blkaddr); 1202 } else { 1203 f2fs_update_data_blkaddr(&dn, *blkaddr); 1204 } 1205 f2fs_put_dnode(&dn); 1206 } 1207 return 0; 1208 } 1209 1210 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1211 block_t *blkaddr, int *do_replace, 1212 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1213 { 1214 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1215 pgoff_t i = 0; 1216 int ret; 1217 1218 while (i < len) { 1219 if (blkaddr[i] == NULL_ADDR && !full) { 1220 i++; 1221 continue; 1222 } 1223 1224 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1225 struct dnode_of_data dn; 1226 struct node_info ni; 1227 size_t new_size; 1228 pgoff_t ilen; 1229 1230 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1231 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1232 if (ret) 1233 return ret; 1234 1235 ret = f2fs_get_node_info(sbi, dn.nid, &ni); 1236 if (ret) { 1237 f2fs_put_dnode(&dn); 1238 return ret; 1239 } 1240 1241 ilen = min((pgoff_t) 1242 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1243 dn.ofs_in_node, len - i); 1244 do { 1245 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1246 f2fs_truncate_data_blocks_range(&dn, 1); 1247 1248 if (do_replace[i]) { 1249 f2fs_i_blocks_write(src_inode, 1250 1, false, false); 1251 f2fs_i_blocks_write(dst_inode, 1252 1, true, false); 1253 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1254 blkaddr[i], ni.version, true, false); 1255 1256 do_replace[i] = 0; 1257 } 1258 dn.ofs_in_node++; 1259 i++; 1260 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1261 if (dst_inode->i_size < new_size) 1262 f2fs_i_size_write(dst_inode, new_size); 1263 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1264 1265 f2fs_put_dnode(&dn); 1266 } else { 1267 struct page *psrc, *pdst; 1268 1269 psrc = f2fs_get_lock_data_page(src_inode, 1270 src + i, true); 1271 if (IS_ERR(psrc)) 1272 return PTR_ERR(psrc); 1273 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1274 true); 1275 if (IS_ERR(pdst)) { 1276 f2fs_put_page(psrc, 1); 1277 return PTR_ERR(pdst); 1278 } 1279 f2fs_copy_page(psrc, pdst); 1280 set_page_dirty(pdst); 1281 f2fs_put_page(pdst, 1); 1282 f2fs_put_page(psrc, 1); 1283 1284 ret = f2fs_truncate_hole(src_inode, 1285 src + i, src + i + 1); 1286 if (ret) 1287 return ret; 1288 i++; 1289 } 1290 } 1291 return 0; 1292 } 1293 1294 static int __exchange_data_block(struct inode *src_inode, 1295 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1296 pgoff_t len, bool full) 1297 { 1298 block_t *src_blkaddr; 1299 int *do_replace; 1300 pgoff_t olen; 1301 int ret; 1302 1303 while (len) { 1304 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1305 1306 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1307 array_size(olen, sizeof(block_t)), 1308 GFP_NOFS); 1309 if (!src_blkaddr) 1310 return -ENOMEM; 1311 1312 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1313 array_size(olen, sizeof(int)), 1314 GFP_NOFS); 1315 if (!do_replace) { 1316 kvfree(src_blkaddr); 1317 return -ENOMEM; 1318 } 1319 1320 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1321 do_replace, src, olen); 1322 if (ret) 1323 goto roll_back; 1324 1325 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1326 do_replace, src, dst, olen, full); 1327 if (ret) 1328 goto roll_back; 1329 1330 src += olen; 1331 dst += olen; 1332 len -= olen; 1333 1334 kvfree(src_blkaddr); 1335 kvfree(do_replace); 1336 } 1337 return 0; 1338 1339 roll_back: 1340 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1341 kvfree(src_blkaddr); 1342 kvfree(do_replace); 1343 return ret; 1344 } 1345 1346 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1347 { 1348 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1349 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1350 pgoff_t start = offset >> PAGE_SHIFT; 1351 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1352 int ret; 1353 1354 f2fs_balance_fs(sbi, true); 1355 1356 /* avoid gc operation during block exchange */ 1357 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1358 filemap_invalidate_lock(inode->i_mapping); 1359 1360 f2fs_lock_op(sbi); 1361 f2fs_drop_extent_tree(inode); 1362 truncate_pagecache(inode, offset); 1363 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1364 f2fs_unlock_op(sbi); 1365 1366 filemap_invalidate_unlock(inode->i_mapping); 1367 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1368 return ret; 1369 } 1370 1371 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1372 { 1373 loff_t new_size; 1374 int ret; 1375 1376 if (offset + len >= i_size_read(inode)) 1377 return -EINVAL; 1378 1379 /* collapse range should be aligned to block size of f2fs. */ 1380 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1381 return -EINVAL; 1382 1383 ret = f2fs_convert_inline_inode(inode); 1384 if (ret) 1385 return ret; 1386 1387 /* write out all dirty pages from offset */ 1388 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1389 if (ret) 1390 return ret; 1391 1392 ret = f2fs_do_collapse(inode, offset, len); 1393 if (ret) 1394 return ret; 1395 1396 /* write out all moved pages, if possible */ 1397 filemap_invalidate_lock(inode->i_mapping); 1398 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1399 truncate_pagecache(inode, offset); 1400 1401 new_size = i_size_read(inode) - len; 1402 ret = f2fs_truncate_blocks(inode, new_size, true); 1403 filemap_invalidate_unlock(inode->i_mapping); 1404 if (!ret) 1405 f2fs_i_size_write(inode, new_size); 1406 return ret; 1407 } 1408 1409 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1410 pgoff_t end) 1411 { 1412 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1413 pgoff_t index = start; 1414 unsigned int ofs_in_node = dn->ofs_in_node; 1415 blkcnt_t count = 0; 1416 int ret; 1417 1418 for (; index < end; index++, dn->ofs_in_node++) { 1419 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1420 count++; 1421 } 1422 1423 dn->ofs_in_node = ofs_in_node; 1424 ret = f2fs_reserve_new_blocks(dn, count); 1425 if (ret) 1426 return ret; 1427 1428 dn->ofs_in_node = ofs_in_node; 1429 for (index = start; index < end; index++, dn->ofs_in_node++) { 1430 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1431 /* 1432 * f2fs_reserve_new_blocks will not guarantee entire block 1433 * allocation. 1434 */ 1435 if (dn->data_blkaddr == NULL_ADDR) { 1436 ret = -ENOSPC; 1437 break; 1438 } 1439 if (dn->data_blkaddr != NEW_ADDR) { 1440 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1441 dn->data_blkaddr = NEW_ADDR; 1442 f2fs_set_data_blkaddr(dn); 1443 } 1444 } 1445 1446 f2fs_update_extent_cache_range(dn, start, 0, index - start); 1447 1448 return ret; 1449 } 1450 1451 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1452 int mode) 1453 { 1454 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1455 struct address_space *mapping = inode->i_mapping; 1456 pgoff_t index, pg_start, pg_end; 1457 loff_t new_size = i_size_read(inode); 1458 loff_t off_start, off_end; 1459 int ret = 0; 1460 1461 ret = inode_newsize_ok(inode, (len + offset)); 1462 if (ret) 1463 return ret; 1464 1465 ret = f2fs_convert_inline_inode(inode); 1466 if (ret) 1467 return ret; 1468 1469 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1470 if (ret) 1471 return ret; 1472 1473 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1474 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1475 1476 off_start = offset & (PAGE_SIZE - 1); 1477 off_end = (offset + len) & (PAGE_SIZE - 1); 1478 1479 if (pg_start == pg_end) { 1480 ret = fill_zero(inode, pg_start, off_start, 1481 off_end - off_start); 1482 if (ret) 1483 return ret; 1484 1485 new_size = max_t(loff_t, new_size, offset + len); 1486 } else { 1487 if (off_start) { 1488 ret = fill_zero(inode, pg_start++, off_start, 1489 PAGE_SIZE - off_start); 1490 if (ret) 1491 return ret; 1492 1493 new_size = max_t(loff_t, new_size, 1494 (loff_t)pg_start << PAGE_SHIFT); 1495 } 1496 1497 for (index = pg_start; index < pg_end;) { 1498 struct dnode_of_data dn; 1499 unsigned int end_offset; 1500 pgoff_t end; 1501 1502 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1503 filemap_invalidate_lock(mapping); 1504 1505 truncate_pagecache_range(inode, 1506 (loff_t)index << PAGE_SHIFT, 1507 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1508 1509 f2fs_lock_op(sbi); 1510 1511 set_new_dnode(&dn, inode, NULL, NULL, 0); 1512 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1513 if (ret) { 1514 f2fs_unlock_op(sbi); 1515 filemap_invalidate_unlock(mapping); 1516 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1517 goto out; 1518 } 1519 1520 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1521 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1522 1523 ret = f2fs_do_zero_range(&dn, index, end); 1524 f2fs_put_dnode(&dn); 1525 1526 f2fs_unlock_op(sbi); 1527 filemap_invalidate_unlock(mapping); 1528 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1529 1530 f2fs_balance_fs(sbi, dn.node_changed); 1531 1532 if (ret) 1533 goto out; 1534 1535 index = end; 1536 new_size = max_t(loff_t, new_size, 1537 (loff_t)index << PAGE_SHIFT); 1538 } 1539 1540 if (off_end) { 1541 ret = fill_zero(inode, pg_end, 0, off_end); 1542 if (ret) 1543 goto out; 1544 1545 new_size = max_t(loff_t, new_size, offset + len); 1546 } 1547 } 1548 1549 out: 1550 if (new_size > i_size_read(inode)) { 1551 if (mode & FALLOC_FL_KEEP_SIZE) 1552 file_set_keep_isize(inode); 1553 else 1554 f2fs_i_size_write(inode, new_size); 1555 } 1556 return ret; 1557 } 1558 1559 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1560 { 1561 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1562 struct address_space *mapping = inode->i_mapping; 1563 pgoff_t nr, pg_start, pg_end, delta, idx; 1564 loff_t new_size; 1565 int ret = 0; 1566 1567 new_size = i_size_read(inode) + len; 1568 ret = inode_newsize_ok(inode, new_size); 1569 if (ret) 1570 return ret; 1571 1572 if (offset >= i_size_read(inode)) 1573 return -EINVAL; 1574 1575 /* insert range should be aligned to block size of f2fs. */ 1576 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1577 return -EINVAL; 1578 1579 ret = f2fs_convert_inline_inode(inode); 1580 if (ret) 1581 return ret; 1582 1583 f2fs_balance_fs(sbi, true); 1584 1585 filemap_invalidate_lock(mapping); 1586 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1587 filemap_invalidate_unlock(mapping); 1588 if (ret) 1589 return ret; 1590 1591 /* write out all dirty pages from offset */ 1592 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1593 if (ret) 1594 return ret; 1595 1596 pg_start = offset >> PAGE_SHIFT; 1597 pg_end = (offset + len) >> PAGE_SHIFT; 1598 delta = pg_end - pg_start; 1599 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1600 1601 /* avoid gc operation during block exchange */ 1602 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1603 filemap_invalidate_lock(mapping); 1604 truncate_pagecache(inode, offset); 1605 1606 while (!ret && idx > pg_start) { 1607 nr = idx - pg_start; 1608 if (nr > delta) 1609 nr = delta; 1610 idx -= nr; 1611 1612 f2fs_lock_op(sbi); 1613 f2fs_drop_extent_tree(inode); 1614 1615 ret = __exchange_data_block(inode, inode, idx, 1616 idx + delta, nr, false); 1617 f2fs_unlock_op(sbi); 1618 } 1619 filemap_invalidate_unlock(mapping); 1620 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1621 1622 /* write out all moved pages, if possible */ 1623 filemap_invalidate_lock(mapping); 1624 filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1625 truncate_pagecache(inode, offset); 1626 filemap_invalidate_unlock(mapping); 1627 1628 if (!ret) 1629 f2fs_i_size_write(inode, new_size); 1630 return ret; 1631 } 1632 1633 static int expand_inode_data(struct inode *inode, loff_t offset, 1634 loff_t len, int mode) 1635 { 1636 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1637 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1638 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1639 .m_may_create = true }; 1640 pgoff_t pg_start, pg_end; 1641 loff_t new_size = i_size_read(inode); 1642 loff_t off_end; 1643 block_t expanded = 0; 1644 int err; 1645 1646 err = inode_newsize_ok(inode, (len + offset)); 1647 if (err) 1648 return err; 1649 1650 err = f2fs_convert_inline_inode(inode); 1651 if (err) 1652 return err; 1653 1654 f2fs_balance_fs(sbi, true); 1655 1656 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1657 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1658 off_end = (offset + len) & (PAGE_SIZE - 1); 1659 1660 map.m_lblk = pg_start; 1661 map.m_len = pg_end - pg_start; 1662 if (off_end) 1663 map.m_len++; 1664 1665 if (!map.m_len) 1666 return 0; 1667 1668 if (f2fs_is_pinned_file(inode)) { 1669 block_t sec_blks = BLKS_PER_SEC(sbi); 1670 block_t sec_len = roundup(map.m_len, sec_blks); 1671 1672 map.m_len = sec_blks; 1673 next_alloc: 1674 if (has_not_enough_free_secs(sbi, 0, 1675 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1676 down_write(&sbi->gc_lock); 1677 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO); 1678 if (err && err != -ENODATA && err != -EAGAIN) 1679 goto out_err; 1680 } 1681 1682 down_write(&sbi->pin_sem); 1683 1684 f2fs_lock_op(sbi); 1685 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false); 1686 f2fs_unlock_op(sbi); 1687 1688 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1689 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO); 1690 1691 up_write(&sbi->pin_sem); 1692 1693 expanded += map.m_len; 1694 sec_len -= map.m_len; 1695 map.m_lblk += map.m_len; 1696 if (!err && sec_len) 1697 goto next_alloc; 1698 1699 map.m_len = expanded; 1700 } else { 1701 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); 1702 expanded = map.m_len; 1703 } 1704 out_err: 1705 if (err) { 1706 pgoff_t last_off; 1707 1708 if (!expanded) 1709 return err; 1710 1711 last_off = pg_start + expanded - 1; 1712 1713 /* update new size to the failed position */ 1714 new_size = (last_off == pg_end) ? offset + len : 1715 (loff_t)(last_off + 1) << PAGE_SHIFT; 1716 } else { 1717 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1718 } 1719 1720 if (new_size > i_size_read(inode)) { 1721 if (mode & FALLOC_FL_KEEP_SIZE) 1722 file_set_keep_isize(inode); 1723 else 1724 f2fs_i_size_write(inode, new_size); 1725 } 1726 1727 return err; 1728 } 1729 1730 static long f2fs_fallocate(struct file *file, int mode, 1731 loff_t offset, loff_t len) 1732 { 1733 struct inode *inode = file_inode(file); 1734 long ret = 0; 1735 1736 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1737 return -EIO; 1738 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1739 return -ENOSPC; 1740 if (!f2fs_is_compress_backend_ready(inode)) 1741 return -EOPNOTSUPP; 1742 1743 /* f2fs only support ->fallocate for regular file */ 1744 if (!S_ISREG(inode->i_mode)) 1745 return -EINVAL; 1746 1747 if (IS_ENCRYPTED(inode) && 1748 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1749 return -EOPNOTSUPP; 1750 1751 if (f2fs_compressed_file(inode) && 1752 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1753 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1754 return -EOPNOTSUPP; 1755 1756 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1757 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1758 FALLOC_FL_INSERT_RANGE)) 1759 return -EOPNOTSUPP; 1760 1761 inode_lock(inode); 1762 1763 if (mode & FALLOC_FL_PUNCH_HOLE) { 1764 if (offset >= inode->i_size) 1765 goto out; 1766 1767 ret = punch_hole(inode, offset, len); 1768 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1769 ret = f2fs_collapse_range(inode, offset, len); 1770 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1771 ret = f2fs_zero_range(inode, offset, len, mode); 1772 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1773 ret = f2fs_insert_range(inode, offset, len); 1774 } else { 1775 ret = expand_inode_data(inode, offset, len, mode); 1776 } 1777 1778 if (!ret) { 1779 inode->i_mtime = inode->i_ctime = current_time(inode); 1780 f2fs_mark_inode_dirty_sync(inode, false); 1781 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1782 } 1783 1784 out: 1785 inode_unlock(inode); 1786 1787 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1788 return ret; 1789 } 1790 1791 static int f2fs_release_file(struct inode *inode, struct file *filp) 1792 { 1793 /* 1794 * f2fs_relase_file is called at every close calls. So we should 1795 * not drop any inmemory pages by close called by other process. 1796 */ 1797 if (!(filp->f_mode & FMODE_WRITE) || 1798 atomic_read(&inode->i_writecount) != 1) 1799 return 0; 1800 1801 /* some remained atomic pages should discarded */ 1802 if (f2fs_is_atomic_file(inode)) 1803 f2fs_drop_inmem_pages(inode); 1804 if (f2fs_is_volatile_file(inode)) { 1805 set_inode_flag(inode, FI_DROP_CACHE); 1806 filemap_fdatawrite(inode->i_mapping); 1807 clear_inode_flag(inode, FI_DROP_CACHE); 1808 clear_inode_flag(inode, FI_VOLATILE_FILE); 1809 stat_dec_volatile_write(inode); 1810 } 1811 return 0; 1812 } 1813 1814 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1815 { 1816 struct inode *inode = file_inode(file); 1817 1818 /* 1819 * If the process doing a transaction is crashed, we should do 1820 * roll-back. Otherwise, other reader/write can see corrupted database 1821 * until all the writers close its file. Since this should be done 1822 * before dropping file lock, it needs to do in ->flush. 1823 */ 1824 if (f2fs_is_atomic_file(inode) && 1825 F2FS_I(inode)->inmem_task == current) 1826 f2fs_drop_inmem_pages(inode); 1827 return 0; 1828 } 1829 1830 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1831 { 1832 struct f2fs_inode_info *fi = F2FS_I(inode); 1833 u32 masked_flags = fi->i_flags & mask; 1834 1835 /* mask can be shrunk by flags_valid selector */ 1836 iflags &= mask; 1837 1838 /* Is it quota file? Do not allow user to mess with it */ 1839 if (IS_NOQUOTA(inode)) 1840 return -EPERM; 1841 1842 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1843 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1844 return -EOPNOTSUPP; 1845 if (!f2fs_empty_dir(inode)) 1846 return -ENOTEMPTY; 1847 } 1848 1849 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1850 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1851 return -EOPNOTSUPP; 1852 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1853 return -EINVAL; 1854 } 1855 1856 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1857 if (masked_flags & F2FS_COMPR_FL) { 1858 if (!f2fs_disable_compressed_file(inode)) 1859 return -EINVAL; 1860 } 1861 if (iflags & F2FS_NOCOMP_FL) 1862 return -EINVAL; 1863 if (iflags & F2FS_COMPR_FL) { 1864 if (!f2fs_may_compress(inode)) 1865 return -EINVAL; 1866 if (S_ISREG(inode->i_mode) && inode->i_size) 1867 return -EINVAL; 1868 1869 set_compress_context(inode); 1870 } 1871 } 1872 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) { 1873 if (masked_flags & F2FS_COMPR_FL) 1874 return -EINVAL; 1875 } 1876 1877 fi->i_flags = iflags | (fi->i_flags & ~mask); 1878 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1879 (fi->i_flags & F2FS_NOCOMP_FL)); 1880 1881 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1882 set_inode_flag(inode, FI_PROJ_INHERIT); 1883 else 1884 clear_inode_flag(inode, FI_PROJ_INHERIT); 1885 1886 inode->i_ctime = current_time(inode); 1887 f2fs_set_inode_flags(inode); 1888 f2fs_mark_inode_dirty_sync(inode, true); 1889 return 0; 1890 } 1891 1892 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1893 1894 /* 1895 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1896 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1897 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1898 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1899 * 1900 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1901 * FS_IOC_FSSETXATTR is done by the VFS. 1902 */ 1903 1904 static const struct { 1905 u32 iflag; 1906 u32 fsflag; 1907 } f2fs_fsflags_map[] = { 1908 { F2FS_COMPR_FL, FS_COMPR_FL }, 1909 { F2FS_SYNC_FL, FS_SYNC_FL }, 1910 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1911 { F2FS_APPEND_FL, FS_APPEND_FL }, 1912 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1913 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1914 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1915 { F2FS_INDEX_FL, FS_INDEX_FL }, 1916 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1917 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1918 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1919 }; 1920 1921 #define F2FS_GETTABLE_FS_FL ( \ 1922 FS_COMPR_FL | \ 1923 FS_SYNC_FL | \ 1924 FS_IMMUTABLE_FL | \ 1925 FS_APPEND_FL | \ 1926 FS_NODUMP_FL | \ 1927 FS_NOATIME_FL | \ 1928 FS_NOCOMP_FL | \ 1929 FS_INDEX_FL | \ 1930 FS_DIRSYNC_FL | \ 1931 FS_PROJINHERIT_FL | \ 1932 FS_ENCRYPT_FL | \ 1933 FS_INLINE_DATA_FL | \ 1934 FS_NOCOW_FL | \ 1935 FS_VERITY_FL | \ 1936 FS_CASEFOLD_FL) 1937 1938 #define F2FS_SETTABLE_FS_FL ( \ 1939 FS_COMPR_FL | \ 1940 FS_SYNC_FL | \ 1941 FS_IMMUTABLE_FL | \ 1942 FS_APPEND_FL | \ 1943 FS_NODUMP_FL | \ 1944 FS_NOATIME_FL | \ 1945 FS_NOCOMP_FL | \ 1946 FS_DIRSYNC_FL | \ 1947 FS_PROJINHERIT_FL | \ 1948 FS_CASEFOLD_FL) 1949 1950 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 1951 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 1952 { 1953 u32 fsflags = 0; 1954 int i; 1955 1956 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1957 if (iflags & f2fs_fsflags_map[i].iflag) 1958 fsflags |= f2fs_fsflags_map[i].fsflag; 1959 1960 return fsflags; 1961 } 1962 1963 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 1964 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 1965 { 1966 u32 iflags = 0; 1967 int i; 1968 1969 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1970 if (fsflags & f2fs_fsflags_map[i].fsflag) 1971 iflags |= f2fs_fsflags_map[i].iflag; 1972 1973 return iflags; 1974 } 1975 1976 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1977 { 1978 struct inode *inode = file_inode(filp); 1979 1980 return put_user(inode->i_generation, (int __user *)arg); 1981 } 1982 1983 static int f2fs_ioc_start_atomic_write(struct file *filp) 1984 { 1985 struct inode *inode = file_inode(filp); 1986 struct f2fs_inode_info *fi = F2FS_I(inode); 1987 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1988 int ret; 1989 1990 if (!inode_owner_or_capable(&init_user_ns, inode)) 1991 return -EACCES; 1992 1993 if (!S_ISREG(inode->i_mode)) 1994 return -EINVAL; 1995 1996 if (filp->f_flags & O_DIRECT) 1997 return -EINVAL; 1998 1999 ret = mnt_want_write_file(filp); 2000 if (ret) 2001 return ret; 2002 2003 inode_lock(inode); 2004 2005 f2fs_disable_compressed_file(inode); 2006 2007 if (f2fs_is_atomic_file(inode)) { 2008 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) 2009 ret = -EINVAL; 2010 goto out; 2011 } 2012 2013 ret = f2fs_convert_inline_inode(inode); 2014 if (ret) 2015 goto out; 2016 2017 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2018 2019 /* 2020 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2021 * f2fs_is_atomic_file. 2022 */ 2023 if (get_dirty_pages(inode)) 2024 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2025 inode->i_ino, get_dirty_pages(inode)); 2026 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2027 if (ret) { 2028 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2029 goto out; 2030 } 2031 2032 spin_lock(&sbi->inode_lock[ATOMIC_FILE]); 2033 if (list_empty(&fi->inmem_ilist)) 2034 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]); 2035 sbi->atomic_files++; 2036 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]); 2037 2038 /* add inode in inmem_list first and set atomic_file */ 2039 set_inode_flag(inode, FI_ATOMIC_FILE); 2040 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2041 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2042 2043 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2044 F2FS_I(inode)->inmem_task = current; 2045 stat_update_max_atomic_write(inode); 2046 out: 2047 inode_unlock(inode); 2048 mnt_drop_write_file(filp); 2049 return ret; 2050 } 2051 2052 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2053 { 2054 struct inode *inode = file_inode(filp); 2055 int ret; 2056 2057 if (!inode_owner_or_capable(&init_user_ns, inode)) 2058 return -EACCES; 2059 2060 ret = mnt_want_write_file(filp); 2061 if (ret) 2062 return ret; 2063 2064 f2fs_balance_fs(F2FS_I_SB(inode), true); 2065 2066 inode_lock(inode); 2067 2068 if (f2fs_is_volatile_file(inode)) { 2069 ret = -EINVAL; 2070 goto err_out; 2071 } 2072 2073 if (f2fs_is_atomic_file(inode)) { 2074 ret = f2fs_commit_inmem_pages(inode); 2075 if (ret) 2076 goto err_out; 2077 2078 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2079 if (!ret) 2080 f2fs_drop_inmem_pages(inode); 2081 } else { 2082 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2083 } 2084 err_out: 2085 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) { 2086 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2087 ret = -EINVAL; 2088 } 2089 inode_unlock(inode); 2090 mnt_drop_write_file(filp); 2091 return ret; 2092 } 2093 2094 static int f2fs_ioc_start_volatile_write(struct file *filp) 2095 { 2096 struct inode *inode = file_inode(filp); 2097 int ret; 2098 2099 if (!inode_owner_or_capable(&init_user_ns, inode)) 2100 return -EACCES; 2101 2102 if (!S_ISREG(inode->i_mode)) 2103 return -EINVAL; 2104 2105 ret = mnt_want_write_file(filp); 2106 if (ret) 2107 return ret; 2108 2109 inode_lock(inode); 2110 2111 if (f2fs_is_volatile_file(inode)) 2112 goto out; 2113 2114 ret = f2fs_convert_inline_inode(inode); 2115 if (ret) 2116 goto out; 2117 2118 stat_inc_volatile_write(inode); 2119 stat_update_max_volatile_write(inode); 2120 2121 set_inode_flag(inode, FI_VOLATILE_FILE); 2122 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2123 out: 2124 inode_unlock(inode); 2125 mnt_drop_write_file(filp); 2126 return ret; 2127 } 2128 2129 static int f2fs_ioc_release_volatile_write(struct file *filp) 2130 { 2131 struct inode *inode = file_inode(filp); 2132 int ret; 2133 2134 if (!inode_owner_or_capable(&init_user_ns, inode)) 2135 return -EACCES; 2136 2137 ret = mnt_want_write_file(filp); 2138 if (ret) 2139 return ret; 2140 2141 inode_lock(inode); 2142 2143 if (!f2fs_is_volatile_file(inode)) 2144 goto out; 2145 2146 if (!f2fs_is_first_block_written(inode)) { 2147 ret = truncate_partial_data_page(inode, 0, true); 2148 goto out; 2149 } 2150 2151 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 2152 out: 2153 inode_unlock(inode); 2154 mnt_drop_write_file(filp); 2155 return ret; 2156 } 2157 2158 static int f2fs_ioc_abort_volatile_write(struct file *filp) 2159 { 2160 struct inode *inode = file_inode(filp); 2161 int ret; 2162 2163 if (!inode_owner_or_capable(&init_user_ns, inode)) 2164 return -EACCES; 2165 2166 ret = mnt_want_write_file(filp); 2167 if (ret) 2168 return ret; 2169 2170 inode_lock(inode); 2171 2172 if (f2fs_is_atomic_file(inode)) 2173 f2fs_drop_inmem_pages(inode); 2174 if (f2fs_is_volatile_file(inode)) { 2175 clear_inode_flag(inode, FI_VOLATILE_FILE); 2176 stat_dec_volatile_write(inode); 2177 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2178 } 2179 2180 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2181 2182 inode_unlock(inode); 2183 2184 mnt_drop_write_file(filp); 2185 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2186 return ret; 2187 } 2188 2189 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2190 { 2191 struct inode *inode = file_inode(filp); 2192 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2193 struct super_block *sb = sbi->sb; 2194 __u32 in; 2195 int ret = 0; 2196 2197 if (!capable(CAP_SYS_ADMIN)) 2198 return -EPERM; 2199 2200 if (get_user(in, (__u32 __user *)arg)) 2201 return -EFAULT; 2202 2203 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2204 ret = mnt_want_write_file(filp); 2205 if (ret) { 2206 if (ret == -EROFS) { 2207 ret = 0; 2208 f2fs_stop_checkpoint(sbi, false); 2209 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2210 trace_f2fs_shutdown(sbi, in, ret); 2211 } 2212 return ret; 2213 } 2214 } 2215 2216 switch (in) { 2217 case F2FS_GOING_DOWN_FULLSYNC: 2218 ret = freeze_bdev(sb->s_bdev); 2219 if (ret) 2220 goto out; 2221 f2fs_stop_checkpoint(sbi, false); 2222 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2223 thaw_bdev(sb->s_bdev); 2224 break; 2225 case F2FS_GOING_DOWN_METASYNC: 2226 /* do checkpoint only */ 2227 ret = f2fs_sync_fs(sb, 1); 2228 if (ret) 2229 goto out; 2230 f2fs_stop_checkpoint(sbi, false); 2231 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2232 break; 2233 case F2FS_GOING_DOWN_NOSYNC: 2234 f2fs_stop_checkpoint(sbi, false); 2235 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2236 break; 2237 case F2FS_GOING_DOWN_METAFLUSH: 2238 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2239 f2fs_stop_checkpoint(sbi, false); 2240 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2241 break; 2242 case F2FS_GOING_DOWN_NEED_FSCK: 2243 set_sbi_flag(sbi, SBI_NEED_FSCK); 2244 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2245 set_sbi_flag(sbi, SBI_IS_DIRTY); 2246 /* do checkpoint only */ 2247 ret = f2fs_sync_fs(sb, 1); 2248 goto out; 2249 default: 2250 ret = -EINVAL; 2251 goto out; 2252 } 2253 2254 f2fs_stop_gc_thread(sbi); 2255 f2fs_stop_discard_thread(sbi); 2256 2257 f2fs_drop_discard_cmd(sbi); 2258 clear_opt(sbi, DISCARD); 2259 2260 f2fs_update_time(sbi, REQ_TIME); 2261 out: 2262 if (in != F2FS_GOING_DOWN_FULLSYNC) 2263 mnt_drop_write_file(filp); 2264 2265 trace_f2fs_shutdown(sbi, in, ret); 2266 2267 return ret; 2268 } 2269 2270 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2271 { 2272 struct inode *inode = file_inode(filp); 2273 struct super_block *sb = inode->i_sb; 2274 struct request_queue *q = bdev_get_queue(sb->s_bdev); 2275 struct fstrim_range range; 2276 int ret; 2277 2278 if (!capable(CAP_SYS_ADMIN)) 2279 return -EPERM; 2280 2281 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2282 return -EOPNOTSUPP; 2283 2284 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2285 sizeof(range))) 2286 return -EFAULT; 2287 2288 ret = mnt_want_write_file(filp); 2289 if (ret) 2290 return ret; 2291 2292 range.minlen = max((unsigned int)range.minlen, 2293 q->limits.discard_granularity); 2294 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2295 mnt_drop_write_file(filp); 2296 if (ret < 0) 2297 return ret; 2298 2299 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2300 sizeof(range))) 2301 return -EFAULT; 2302 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2303 return 0; 2304 } 2305 2306 static bool uuid_is_nonzero(__u8 u[16]) 2307 { 2308 int i; 2309 2310 for (i = 0; i < 16; i++) 2311 if (u[i]) 2312 return true; 2313 return false; 2314 } 2315 2316 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2317 { 2318 struct inode *inode = file_inode(filp); 2319 2320 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2321 return -EOPNOTSUPP; 2322 2323 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2324 2325 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2326 } 2327 2328 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2329 { 2330 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2331 return -EOPNOTSUPP; 2332 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2333 } 2334 2335 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2336 { 2337 struct inode *inode = file_inode(filp); 2338 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2339 int err; 2340 2341 if (!f2fs_sb_has_encrypt(sbi)) 2342 return -EOPNOTSUPP; 2343 2344 err = mnt_want_write_file(filp); 2345 if (err) 2346 return err; 2347 2348 down_write(&sbi->sb_lock); 2349 2350 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2351 goto got_it; 2352 2353 /* update superblock with uuid */ 2354 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2355 2356 err = f2fs_commit_super(sbi, false); 2357 if (err) { 2358 /* undo new data */ 2359 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2360 goto out_err; 2361 } 2362 got_it: 2363 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 2364 16)) 2365 err = -EFAULT; 2366 out_err: 2367 up_write(&sbi->sb_lock); 2368 mnt_drop_write_file(filp); 2369 return err; 2370 } 2371 2372 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2373 unsigned long arg) 2374 { 2375 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2376 return -EOPNOTSUPP; 2377 2378 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2379 } 2380 2381 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2382 { 2383 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2384 return -EOPNOTSUPP; 2385 2386 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2387 } 2388 2389 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2390 { 2391 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2392 return -EOPNOTSUPP; 2393 2394 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2395 } 2396 2397 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2398 unsigned long arg) 2399 { 2400 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2401 return -EOPNOTSUPP; 2402 2403 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2404 } 2405 2406 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2407 unsigned long arg) 2408 { 2409 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2410 return -EOPNOTSUPP; 2411 2412 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2413 } 2414 2415 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2416 { 2417 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2418 return -EOPNOTSUPP; 2419 2420 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2421 } 2422 2423 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2424 { 2425 struct inode *inode = file_inode(filp); 2426 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2427 __u32 sync; 2428 int ret; 2429 2430 if (!capable(CAP_SYS_ADMIN)) 2431 return -EPERM; 2432 2433 if (get_user(sync, (__u32 __user *)arg)) 2434 return -EFAULT; 2435 2436 if (f2fs_readonly(sbi->sb)) 2437 return -EROFS; 2438 2439 ret = mnt_want_write_file(filp); 2440 if (ret) 2441 return ret; 2442 2443 if (!sync) { 2444 if (!down_write_trylock(&sbi->gc_lock)) { 2445 ret = -EBUSY; 2446 goto out; 2447 } 2448 } else { 2449 down_write(&sbi->gc_lock); 2450 } 2451 2452 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO); 2453 out: 2454 mnt_drop_write_file(filp); 2455 return ret; 2456 } 2457 2458 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2459 { 2460 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2461 u64 end; 2462 int ret; 2463 2464 if (!capable(CAP_SYS_ADMIN)) 2465 return -EPERM; 2466 if (f2fs_readonly(sbi->sb)) 2467 return -EROFS; 2468 2469 end = range->start + range->len; 2470 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2471 end >= MAX_BLKADDR(sbi)) 2472 return -EINVAL; 2473 2474 ret = mnt_want_write_file(filp); 2475 if (ret) 2476 return ret; 2477 2478 do_more: 2479 if (!range->sync) { 2480 if (!down_write_trylock(&sbi->gc_lock)) { 2481 ret = -EBUSY; 2482 goto out; 2483 } 2484 } else { 2485 down_write(&sbi->gc_lock); 2486 } 2487 2488 ret = f2fs_gc(sbi, range->sync, true, false, 2489 GET_SEGNO(sbi, range->start)); 2490 if (ret) { 2491 if (ret == -EBUSY) 2492 ret = -EAGAIN; 2493 goto out; 2494 } 2495 range->start += BLKS_PER_SEC(sbi); 2496 if (range->start <= end) 2497 goto do_more; 2498 out: 2499 mnt_drop_write_file(filp); 2500 return ret; 2501 } 2502 2503 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2504 { 2505 struct f2fs_gc_range range; 2506 2507 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2508 sizeof(range))) 2509 return -EFAULT; 2510 return __f2fs_ioc_gc_range(filp, &range); 2511 } 2512 2513 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg) 2514 { 2515 struct inode *inode = file_inode(filp); 2516 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2517 int ret; 2518 2519 if (!capable(CAP_SYS_ADMIN)) 2520 return -EPERM; 2521 2522 if (f2fs_readonly(sbi->sb)) 2523 return -EROFS; 2524 2525 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2526 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2527 return -EINVAL; 2528 } 2529 2530 ret = mnt_want_write_file(filp); 2531 if (ret) 2532 return ret; 2533 2534 ret = f2fs_sync_fs(sbi->sb, 1); 2535 2536 mnt_drop_write_file(filp); 2537 return ret; 2538 } 2539 2540 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2541 struct file *filp, 2542 struct f2fs_defragment *range) 2543 { 2544 struct inode *inode = file_inode(filp); 2545 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2546 .m_seg_type = NO_CHECK_TYPE, 2547 .m_may_create = false }; 2548 struct extent_info ei = {0, 0, 0}; 2549 pgoff_t pg_start, pg_end, next_pgofs; 2550 unsigned int blk_per_seg = sbi->blocks_per_seg; 2551 unsigned int total = 0, sec_num; 2552 block_t blk_end = 0; 2553 bool fragmented = false; 2554 int err; 2555 2556 /* if in-place-update policy is enabled, don't waste time here */ 2557 if (f2fs_should_update_inplace(inode, NULL)) 2558 return -EINVAL; 2559 2560 pg_start = range->start >> PAGE_SHIFT; 2561 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2562 2563 f2fs_balance_fs(sbi, true); 2564 2565 inode_lock(inode); 2566 2567 /* writeback all dirty pages in the range */ 2568 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2569 range->start + range->len - 1); 2570 if (err) 2571 goto out; 2572 2573 /* 2574 * lookup mapping info in extent cache, skip defragmenting if physical 2575 * block addresses are continuous. 2576 */ 2577 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) { 2578 if (ei.fofs + ei.len >= pg_end) 2579 goto out; 2580 } 2581 2582 map.m_lblk = pg_start; 2583 map.m_next_pgofs = &next_pgofs; 2584 2585 /* 2586 * lookup mapping info in dnode page cache, skip defragmenting if all 2587 * physical block addresses are continuous even if there are hole(s) 2588 * in logical blocks. 2589 */ 2590 while (map.m_lblk < pg_end) { 2591 map.m_len = pg_end - map.m_lblk; 2592 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2593 if (err) 2594 goto out; 2595 2596 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2597 map.m_lblk = next_pgofs; 2598 continue; 2599 } 2600 2601 if (blk_end && blk_end != map.m_pblk) 2602 fragmented = true; 2603 2604 /* record total count of block that we're going to move */ 2605 total += map.m_len; 2606 2607 blk_end = map.m_pblk + map.m_len; 2608 2609 map.m_lblk += map.m_len; 2610 } 2611 2612 if (!fragmented) { 2613 total = 0; 2614 goto out; 2615 } 2616 2617 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi)); 2618 2619 /* 2620 * make sure there are enough free section for LFS allocation, this can 2621 * avoid defragment running in SSR mode when free section are allocated 2622 * intensively 2623 */ 2624 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2625 err = -EAGAIN; 2626 goto out; 2627 } 2628 2629 map.m_lblk = pg_start; 2630 map.m_len = pg_end - pg_start; 2631 total = 0; 2632 2633 while (map.m_lblk < pg_end) { 2634 pgoff_t idx; 2635 int cnt = 0; 2636 2637 do_map: 2638 map.m_len = pg_end - map.m_lblk; 2639 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2640 if (err) 2641 goto clear_out; 2642 2643 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2644 map.m_lblk = next_pgofs; 2645 goto check; 2646 } 2647 2648 set_inode_flag(inode, FI_DO_DEFRAG); 2649 2650 idx = map.m_lblk; 2651 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2652 struct page *page; 2653 2654 page = f2fs_get_lock_data_page(inode, idx, true); 2655 if (IS_ERR(page)) { 2656 err = PTR_ERR(page); 2657 goto clear_out; 2658 } 2659 2660 set_page_dirty(page); 2661 f2fs_put_page(page, 1); 2662 2663 idx++; 2664 cnt++; 2665 total++; 2666 } 2667 2668 map.m_lblk = idx; 2669 check: 2670 if (map.m_lblk < pg_end && cnt < blk_per_seg) 2671 goto do_map; 2672 2673 clear_inode_flag(inode, FI_DO_DEFRAG); 2674 2675 err = filemap_fdatawrite(inode->i_mapping); 2676 if (err) 2677 goto out; 2678 } 2679 clear_out: 2680 clear_inode_flag(inode, FI_DO_DEFRAG); 2681 out: 2682 inode_unlock(inode); 2683 if (!err) 2684 range->len = (u64)total << PAGE_SHIFT; 2685 return err; 2686 } 2687 2688 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2689 { 2690 struct inode *inode = file_inode(filp); 2691 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2692 struct f2fs_defragment range; 2693 int err; 2694 2695 if (!capable(CAP_SYS_ADMIN)) 2696 return -EPERM; 2697 2698 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2699 return -EINVAL; 2700 2701 if (f2fs_readonly(sbi->sb)) 2702 return -EROFS; 2703 2704 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2705 sizeof(range))) 2706 return -EFAULT; 2707 2708 /* verify alignment of offset & size */ 2709 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2710 return -EINVAL; 2711 2712 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2713 max_file_blocks(inode))) 2714 return -EINVAL; 2715 2716 err = mnt_want_write_file(filp); 2717 if (err) 2718 return err; 2719 2720 err = f2fs_defragment_range(sbi, filp, &range); 2721 mnt_drop_write_file(filp); 2722 2723 f2fs_update_time(sbi, REQ_TIME); 2724 if (err < 0) 2725 return err; 2726 2727 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2728 sizeof(range))) 2729 return -EFAULT; 2730 2731 return 0; 2732 } 2733 2734 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2735 struct file *file_out, loff_t pos_out, size_t len) 2736 { 2737 struct inode *src = file_inode(file_in); 2738 struct inode *dst = file_inode(file_out); 2739 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2740 size_t olen = len, dst_max_i_size = 0; 2741 size_t dst_osize; 2742 int ret; 2743 2744 if (file_in->f_path.mnt != file_out->f_path.mnt || 2745 src->i_sb != dst->i_sb) 2746 return -EXDEV; 2747 2748 if (unlikely(f2fs_readonly(src->i_sb))) 2749 return -EROFS; 2750 2751 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2752 return -EINVAL; 2753 2754 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2755 return -EOPNOTSUPP; 2756 2757 if (pos_out < 0 || pos_in < 0) 2758 return -EINVAL; 2759 2760 if (src == dst) { 2761 if (pos_in == pos_out) 2762 return 0; 2763 if (pos_out > pos_in && pos_out < pos_in + len) 2764 return -EINVAL; 2765 } 2766 2767 inode_lock(src); 2768 if (src != dst) { 2769 ret = -EBUSY; 2770 if (!inode_trylock(dst)) 2771 goto out; 2772 } 2773 2774 ret = -EINVAL; 2775 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2776 goto out_unlock; 2777 if (len == 0) 2778 olen = len = src->i_size - pos_in; 2779 if (pos_in + len == src->i_size) 2780 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2781 if (len == 0) { 2782 ret = 0; 2783 goto out_unlock; 2784 } 2785 2786 dst_osize = dst->i_size; 2787 if (pos_out + olen > dst->i_size) 2788 dst_max_i_size = pos_out + olen; 2789 2790 /* verify the end result is block aligned */ 2791 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2792 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2793 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2794 goto out_unlock; 2795 2796 ret = f2fs_convert_inline_inode(src); 2797 if (ret) 2798 goto out_unlock; 2799 2800 ret = f2fs_convert_inline_inode(dst); 2801 if (ret) 2802 goto out_unlock; 2803 2804 /* write out all dirty pages from offset */ 2805 ret = filemap_write_and_wait_range(src->i_mapping, 2806 pos_in, pos_in + len); 2807 if (ret) 2808 goto out_unlock; 2809 2810 ret = filemap_write_and_wait_range(dst->i_mapping, 2811 pos_out, pos_out + len); 2812 if (ret) 2813 goto out_unlock; 2814 2815 f2fs_balance_fs(sbi, true); 2816 2817 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2818 if (src != dst) { 2819 ret = -EBUSY; 2820 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2821 goto out_src; 2822 } 2823 2824 f2fs_lock_op(sbi); 2825 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2826 pos_out >> F2FS_BLKSIZE_BITS, 2827 len >> F2FS_BLKSIZE_BITS, false); 2828 2829 if (!ret) { 2830 if (dst_max_i_size) 2831 f2fs_i_size_write(dst, dst_max_i_size); 2832 else if (dst_osize != dst->i_size) 2833 f2fs_i_size_write(dst, dst_osize); 2834 } 2835 f2fs_unlock_op(sbi); 2836 2837 if (src != dst) 2838 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2839 out_src: 2840 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2841 out_unlock: 2842 if (src != dst) 2843 inode_unlock(dst); 2844 out: 2845 inode_unlock(src); 2846 return ret; 2847 } 2848 2849 static int __f2fs_ioc_move_range(struct file *filp, 2850 struct f2fs_move_range *range) 2851 { 2852 struct fd dst; 2853 int err; 2854 2855 if (!(filp->f_mode & FMODE_READ) || 2856 !(filp->f_mode & FMODE_WRITE)) 2857 return -EBADF; 2858 2859 dst = fdget(range->dst_fd); 2860 if (!dst.file) 2861 return -EBADF; 2862 2863 if (!(dst.file->f_mode & FMODE_WRITE)) { 2864 err = -EBADF; 2865 goto err_out; 2866 } 2867 2868 err = mnt_want_write_file(filp); 2869 if (err) 2870 goto err_out; 2871 2872 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2873 range->pos_out, range->len); 2874 2875 mnt_drop_write_file(filp); 2876 err_out: 2877 fdput(dst); 2878 return err; 2879 } 2880 2881 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2882 { 2883 struct f2fs_move_range range; 2884 2885 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2886 sizeof(range))) 2887 return -EFAULT; 2888 return __f2fs_ioc_move_range(filp, &range); 2889 } 2890 2891 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2892 { 2893 struct inode *inode = file_inode(filp); 2894 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2895 struct sit_info *sm = SIT_I(sbi); 2896 unsigned int start_segno = 0, end_segno = 0; 2897 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2898 struct f2fs_flush_device range; 2899 int ret; 2900 2901 if (!capable(CAP_SYS_ADMIN)) 2902 return -EPERM; 2903 2904 if (f2fs_readonly(sbi->sb)) 2905 return -EROFS; 2906 2907 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2908 return -EINVAL; 2909 2910 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2911 sizeof(range))) 2912 return -EFAULT; 2913 2914 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2915 __is_large_section(sbi)) { 2916 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2917 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2918 return -EINVAL; 2919 } 2920 2921 ret = mnt_want_write_file(filp); 2922 if (ret) 2923 return ret; 2924 2925 if (range.dev_num != 0) 2926 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2927 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2928 2929 start_segno = sm->last_victim[FLUSH_DEVICE]; 2930 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2931 start_segno = dev_start_segno; 2932 end_segno = min(start_segno + range.segments, dev_end_segno); 2933 2934 while (start_segno < end_segno) { 2935 if (!down_write_trylock(&sbi->gc_lock)) { 2936 ret = -EBUSY; 2937 goto out; 2938 } 2939 sm->last_victim[GC_CB] = end_segno + 1; 2940 sm->last_victim[GC_GREEDY] = end_segno + 1; 2941 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 2942 ret = f2fs_gc(sbi, true, true, true, start_segno); 2943 if (ret == -EAGAIN) 2944 ret = 0; 2945 else if (ret < 0) 2946 break; 2947 start_segno++; 2948 } 2949 out: 2950 mnt_drop_write_file(filp); 2951 return ret; 2952 } 2953 2954 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 2955 { 2956 struct inode *inode = file_inode(filp); 2957 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 2958 2959 /* Must validate to set it with SQLite behavior in Android. */ 2960 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 2961 2962 return put_user(sb_feature, (u32 __user *)arg); 2963 } 2964 2965 #ifdef CONFIG_QUOTA 2966 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 2967 { 2968 struct dquot *transfer_to[MAXQUOTAS] = {}; 2969 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2970 struct super_block *sb = sbi->sb; 2971 int err = 0; 2972 2973 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 2974 if (!IS_ERR(transfer_to[PRJQUOTA])) { 2975 err = __dquot_transfer(inode, transfer_to); 2976 if (err) 2977 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2978 dqput(transfer_to[PRJQUOTA]); 2979 } 2980 return err; 2981 } 2982 2983 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 2984 { 2985 struct f2fs_inode_info *fi = F2FS_I(inode); 2986 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2987 struct page *ipage; 2988 kprojid_t kprojid; 2989 int err; 2990 2991 if (!f2fs_sb_has_project_quota(sbi)) { 2992 if (projid != F2FS_DEF_PROJID) 2993 return -EOPNOTSUPP; 2994 else 2995 return 0; 2996 } 2997 2998 if (!f2fs_has_extra_attr(inode)) 2999 return -EOPNOTSUPP; 3000 3001 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3002 3003 if (projid_eq(kprojid, F2FS_I(inode)->i_projid)) 3004 return 0; 3005 3006 err = -EPERM; 3007 /* Is it quota file? Do not allow user to mess with it */ 3008 if (IS_NOQUOTA(inode)) 3009 return err; 3010 3011 ipage = f2fs_get_node_page(sbi, inode->i_ino); 3012 if (IS_ERR(ipage)) 3013 return PTR_ERR(ipage); 3014 3015 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize, 3016 i_projid)) { 3017 err = -EOVERFLOW; 3018 f2fs_put_page(ipage, 1); 3019 return err; 3020 } 3021 f2fs_put_page(ipage, 1); 3022 3023 err = dquot_initialize(inode); 3024 if (err) 3025 return err; 3026 3027 f2fs_lock_op(sbi); 3028 err = f2fs_transfer_project_quota(inode, kprojid); 3029 if (err) 3030 goto out_unlock; 3031 3032 F2FS_I(inode)->i_projid = kprojid; 3033 inode->i_ctime = current_time(inode); 3034 f2fs_mark_inode_dirty_sync(inode, true); 3035 out_unlock: 3036 f2fs_unlock_op(sbi); 3037 return err; 3038 } 3039 #else 3040 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3041 { 3042 return 0; 3043 } 3044 3045 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3046 { 3047 if (projid != F2FS_DEF_PROJID) 3048 return -EOPNOTSUPP; 3049 return 0; 3050 } 3051 #endif 3052 3053 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3054 { 3055 struct inode *inode = d_inode(dentry); 3056 struct f2fs_inode_info *fi = F2FS_I(inode); 3057 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3058 3059 if (IS_ENCRYPTED(inode)) 3060 fsflags |= FS_ENCRYPT_FL; 3061 if (IS_VERITY(inode)) 3062 fsflags |= FS_VERITY_FL; 3063 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3064 fsflags |= FS_INLINE_DATA_FL; 3065 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3066 fsflags |= FS_NOCOW_FL; 3067 3068 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3069 3070 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3071 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3072 3073 return 0; 3074 } 3075 3076 int f2fs_fileattr_set(struct user_namespace *mnt_userns, 3077 struct dentry *dentry, struct fileattr *fa) 3078 { 3079 struct inode *inode = d_inode(dentry); 3080 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3081 u32 iflags; 3082 int err; 3083 3084 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3085 return -EIO; 3086 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3087 return -ENOSPC; 3088 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3089 return -EOPNOTSUPP; 3090 fsflags &= F2FS_SETTABLE_FS_FL; 3091 if (!fa->flags_valid) 3092 mask &= FS_COMMON_FL; 3093 3094 iflags = f2fs_fsflags_to_iflags(fsflags); 3095 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3096 return -EOPNOTSUPP; 3097 3098 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3099 if (!err) 3100 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3101 3102 return err; 3103 } 3104 3105 int f2fs_pin_file_control(struct inode *inode, bool inc) 3106 { 3107 struct f2fs_inode_info *fi = F2FS_I(inode); 3108 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3109 3110 /* Use i_gc_failures for normal file as a risk signal. */ 3111 if (inc) 3112 f2fs_i_gc_failures_write(inode, 3113 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3114 3115 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3116 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3117 __func__, inode->i_ino, 3118 fi->i_gc_failures[GC_FAILURE_PIN]); 3119 clear_inode_flag(inode, FI_PIN_FILE); 3120 return -EAGAIN; 3121 } 3122 return 0; 3123 } 3124 3125 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3126 { 3127 struct inode *inode = file_inode(filp); 3128 __u32 pin; 3129 int ret = 0; 3130 3131 if (get_user(pin, (__u32 __user *)arg)) 3132 return -EFAULT; 3133 3134 if (!S_ISREG(inode->i_mode)) 3135 return -EINVAL; 3136 3137 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3138 return -EROFS; 3139 3140 ret = mnt_want_write_file(filp); 3141 if (ret) 3142 return ret; 3143 3144 inode_lock(inode); 3145 3146 if (f2fs_should_update_outplace(inode, NULL)) { 3147 ret = -EINVAL; 3148 goto out; 3149 } 3150 3151 if (!pin) { 3152 clear_inode_flag(inode, FI_PIN_FILE); 3153 f2fs_i_gc_failures_write(inode, 0); 3154 goto done; 3155 } 3156 3157 if (f2fs_pin_file_control(inode, false)) { 3158 ret = -EAGAIN; 3159 goto out; 3160 } 3161 3162 ret = f2fs_convert_inline_inode(inode); 3163 if (ret) 3164 goto out; 3165 3166 if (!f2fs_disable_compressed_file(inode)) { 3167 ret = -EOPNOTSUPP; 3168 goto out; 3169 } 3170 3171 set_inode_flag(inode, FI_PIN_FILE); 3172 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3173 done: 3174 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3175 out: 3176 inode_unlock(inode); 3177 mnt_drop_write_file(filp); 3178 return ret; 3179 } 3180 3181 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3182 { 3183 struct inode *inode = file_inode(filp); 3184 __u32 pin = 0; 3185 3186 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3187 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3188 return put_user(pin, (u32 __user *)arg); 3189 } 3190 3191 int f2fs_precache_extents(struct inode *inode) 3192 { 3193 struct f2fs_inode_info *fi = F2FS_I(inode); 3194 struct f2fs_map_blocks map; 3195 pgoff_t m_next_extent; 3196 loff_t end; 3197 int err; 3198 3199 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3200 return -EOPNOTSUPP; 3201 3202 map.m_lblk = 0; 3203 map.m_next_pgofs = NULL; 3204 map.m_next_extent = &m_next_extent; 3205 map.m_seg_type = NO_CHECK_TYPE; 3206 map.m_may_create = false; 3207 end = max_file_blocks(inode); 3208 3209 while (map.m_lblk < end) { 3210 map.m_len = end - map.m_lblk; 3211 3212 down_write(&fi->i_gc_rwsem[WRITE]); 3213 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE); 3214 up_write(&fi->i_gc_rwsem[WRITE]); 3215 if (err) 3216 return err; 3217 3218 map.m_lblk = m_next_extent; 3219 } 3220 3221 return 0; 3222 } 3223 3224 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg) 3225 { 3226 return f2fs_precache_extents(file_inode(filp)); 3227 } 3228 3229 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3230 { 3231 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3232 __u64 block_count; 3233 3234 if (!capable(CAP_SYS_ADMIN)) 3235 return -EPERM; 3236 3237 if (f2fs_readonly(sbi->sb)) 3238 return -EROFS; 3239 3240 if (copy_from_user(&block_count, (void __user *)arg, 3241 sizeof(block_count))) 3242 return -EFAULT; 3243 3244 return f2fs_resize_fs(sbi, block_count); 3245 } 3246 3247 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3248 { 3249 struct inode *inode = file_inode(filp); 3250 3251 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3252 3253 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3254 f2fs_warn(F2FS_I_SB(inode), 3255 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3256 inode->i_ino); 3257 return -EOPNOTSUPP; 3258 } 3259 3260 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3261 } 3262 3263 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3264 { 3265 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3266 return -EOPNOTSUPP; 3267 3268 return fsverity_ioctl_measure(filp, (void __user *)arg); 3269 } 3270 3271 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3272 { 3273 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3274 return -EOPNOTSUPP; 3275 3276 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3277 } 3278 3279 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3280 { 3281 struct inode *inode = file_inode(filp); 3282 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3283 char *vbuf; 3284 int count; 3285 int err = 0; 3286 3287 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3288 if (!vbuf) 3289 return -ENOMEM; 3290 3291 down_read(&sbi->sb_lock); 3292 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3293 ARRAY_SIZE(sbi->raw_super->volume_name), 3294 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3295 up_read(&sbi->sb_lock); 3296 3297 if (copy_to_user((char __user *)arg, vbuf, 3298 min(FSLABEL_MAX, count))) 3299 err = -EFAULT; 3300 3301 kfree(vbuf); 3302 return err; 3303 } 3304 3305 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3306 { 3307 struct inode *inode = file_inode(filp); 3308 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3309 char *vbuf; 3310 int err = 0; 3311 3312 if (!capable(CAP_SYS_ADMIN)) 3313 return -EPERM; 3314 3315 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3316 if (IS_ERR(vbuf)) 3317 return PTR_ERR(vbuf); 3318 3319 err = mnt_want_write_file(filp); 3320 if (err) 3321 goto out; 3322 3323 down_write(&sbi->sb_lock); 3324 3325 memset(sbi->raw_super->volume_name, 0, 3326 sizeof(sbi->raw_super->volume_name)); 3327 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3328 sbi->raw_super->volume_name, 3329 ARRAY_SIZE(sbi->raw_super->volume_name)); 3330 3331 err = f2fs_commit_super(sbi, false); 3332 3333 up_write(&sbi->sb_lock); 3334 3335 mnt_drop_write_file(filp); 3336 out: 3337 kfree(vbuf); 3338 return err; 3339 } 3340 3341 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg) 3342 { 3343 struct inode *inode = file_inode(filp); 3344 __u64 blocks; 3345 3346 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3347 return -EOPNOTSUPP; 3348 3349 if (!f2fs_compressed_file(inode)) 3350 return -EINVAL; 3351 3352 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3353 return put_user(blocks, (u64 __user *)arg); 3354 } 3355 3356 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3357 { 3358 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3359 unsigned int released_blocks = 0; 3360 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3361 block_t blkaddr; 3362 int i; 3363 3364 for (i = 0; i < count; i++) { 3365 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3366 dn->ofs_in_node + i); 3367 3368 if (!__is_valid_data_blkaddr(blkaddr)) 3369 continue; 3370 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3371 DATA_GENERIC_ENHANCE))) 3372 return -EFSCORRUPTED; 3373 } 3374 3375 while (count) { 3376 int compr_blocks = 0; 3377 3378 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3379 blkaddr = f2fs_data_blkaddr(dn); 3380 3381 if (i == 0) { 3382 if (blkaddr == COMPRESS_ADDR) 3383 continue; 3384 dn->ofs_in_node += cluster_size; 3385 goto next; 3386 } 3387 3388 if (__is_valid_data_blkaddr(blkaddr)) 3389 compr_blocks++; 3390 3391 if (blkaddr != NEW_ADDR) 3392 continue; 3393 3394 dn->data_blkaddr = NULL_ADDR; 3395 f2fs_set_data_blkaddr(dn); 3396 } 3397 3398 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3399 dec_valid_block_count(sbi, dn->inode, 3400 cluster_size - compr_blocks); 3401 3402 released_blocks += cluster_size - compr_blocks; 3403 next: 3404 count -= cluster_size; 3405 } 3406 3407 return released_blocks; 3408 } 3409 3410 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3411 { 3412 struct inode *inode = file_inode(filp); 3413 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3414 pgoff_t page_idx = 0, last_idx; 3415 unsigned int released_blocks = 0; 3416 int ret; 3417 int writecount; 3418 3419 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3420 return -EOPNOTSUPP; 3421 3422 if (!f2fs_compressed_file(inode)) 3423 return -EINVAL; 3424 3425 if (f2fs_readonly(sbi->sb)) 3426 return -EROFS; 3427 3428 ret = mnt_want_write_file(filp); 3429 if (ret) 3430 return ret; 3431 3432 f2fs_balance_fs(F2FS_I_SB(inode), true); 3433 3434 inode_lock(inode); 3435 3436 writecount = atomic_read(&inode->i_writecount); 3437 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3438 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3439 ret = -EBUSY; 3440 goto out; 3441 } 3442 3443 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3444 ret = -EINVAL; 3445 goto out; 3446 } 3447 3448 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3449 if (ret) 3450 goto out; 3451 3452 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3453 inode->i_ctime = current_time(inode); 3454 f2fs_mark_inode_dirty_sync(inode, true); 3455 3456 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3457 goto out; 3458 3459 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3460 filemap_invalidate_lock(inode->i_mapping); 3461 3462 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3463 3464 while (page_idx < last_idx) { 3465 struct dnode_of_data dn; 3466 pgoff_t end_offset, count; 3467 3468 set_new_dnode(&dn, inode, NULL, NULL, 0); 3469 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3470 if (ret) { 3471 if (ret == -ENOENT) { 3472 page_idx = f2fs_get_next_page_offset(&dn, 3473 page_idx); 3474 ret = 0; 3475 continue; 3476 } 3477 break; 3478 } 3479 3480 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3481 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3482 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3483 3484 ret = release_compress_blocks(&dn, count); 3485 3486 f2fs_put_dnode(&dn); 3487 3488 if (ret < 0) 3489 break; 3490 3491 page_idx += count; 3492 released_blocks += ret; 3493 } 3494 3495 filemap_invalidate_unlock(inode->i_mapping); 3496 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3497 out: 3498 inode_unlock(inode); 3499 3500 mnt_drop_write_file(filp); 3501 3502 if (ret >= 0) { 3503 ret = put_user(released_blocks, (u64 __user *)arg); 3504 } else if (released_blocks && 3505 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3506 set_sbi_flag(sbi, SBI_NEED_FSCK); 3507 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3508 "iblocks=%llu, released=%u, compr_blocks=%u, " 3509 "run fsck to fix.", 3510 __func__, inode->i_ino, inode->i_blocks, 3511 released_blocks, 3512 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3513 } 3514 3515 return ret; 3516 } 3517 3518 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3519 { 3520 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3521 unsigned int reserved_blocks = 0; 3522 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3523 block_t blkaddr; 3524 int i; 3525 3526 for (i = 0; i < count; i++) { 3527 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3528 dn->ofs_in_node + i); 3529 3530 if (!__is_valid_data_blkaddr(blkaddr)) 3531 continue; 3532 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3533 DATA_GENERIC_ENHANCE))) 3534 return -EFSCORRUPTED; 3535 } 3536 3537 while (count) { 3538 int compr_blocks = 0; 3539 blkcnt_t reserved; 3540 int ret; 3541 3542 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3543 blkaddr = f2fs_data_blkaddr(dn); 3544 3545 if (i == 0) { 3546 if (blkaddr == COMPRESS_ADDR) 3547 continue; 3548 dn->ofs_in_node += cluster_size; 3549 goto next; 3550 } 3551 3552 if (__is_valid_data_blkaddr(blkaddr)) { 3553 compr_blocks++; 3554 continue; 3555 } 3556 3557 dn->data_blkaddr = NEW_ADDR; 3558 f2fs_set_data_blkaddr(dn); 3559 } 3560 3561 reserved = cluster_size - compr_blocks; 3562 ret = inc_valid_block_count(sbi, dn->inode, &reserved); 3563 if (ret) 3564 return ret; 3565 3566 if (reserved != cluster_size - compr_blocks) 3567 return -ENOSPC; 3568 3569 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3570 3571 reserved_blocks += reserved; 3572 next: 3573 count -= cluster_size; 3574 } 3575 3576 return reserved_blocks; 3577 } 3578 3579 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3580 { 3581 struct inode *inode = file_inode(filp); 3582 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3583 pgoff_t page_idx = 0, last_idx; 3584 unsigned int reserved_blocks = 0; 3585 int ret; 3586 3587 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3588 return -EOPNOTSUPP; 3589 3590 if (!f2fs_compressed_file(inode)) 3591 return -EINVAL; 3592 3593 if (f2fs_readonly(sbi->sb)) 3594 return -EROFS; 3595 3596 ret = mnt_want_write_file(filp); 3597 if (ret) 3598 return ret; 3599 3600 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3601 goto out; 3602 3603 f2fs_balance_fs(F2FS_I_SB(inode), true); 3604 3605 inode_lock(inode); 3606 3607 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3608 ret = -EINVAL; 3609 goto unlock_inode; 3610 } 3611 3612 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3613 filemap_invalidate_lock(inode->i_mapping); 3614 3615 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3616 3617 while (page_idx < last_idx) { 3618 struct dnode_of_data dn; 3619 pgoff_t end_offset, count; 3620 3621 set_new_dnode(&dn, inode, NULL, NULL, 0); 3622 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3623 if (ret) { 3624 if (ret == -ENOENT) { 3625 page_idx = f2fs_get_next_page_offset(&dn, 3626 page_idx); 3627 ret = 0; 3628 continue; 3629 } 3630 break; 3631 } 3632 3633 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3634 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3635 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3636 3637 ret = reserve_compress_blocks(&dn, count); 3638 3639 f2fs_put_dnode(&dn); 3640 3641 if (ret < 0) 3642 break; 3643 3644 page_idx += count; 3645 reserved_blocks += ret; 3646 } 3647 3648 filemap_invalidate_unlock(inode->i_mapping); 3649 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3650 3651 if (ret >= 0) { 3652 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3653 inode->i_ctime = current_time(inode); 3654 f2fs_mark_inode_dirty_sync(inode, true); 3655 } 3656 unlock_inode: 3657 inode_unlock(inode); 3658 out: 3659 mnt_drop_write_file(filp); 3660 3661 if (ret >= 0) { 3662 ret = put_user(reserved_blocks, (u64 __user *)arg); 3663 } else if (reserved_blocks && 3664 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3665 set_sbi_flag(sbi, SBI_NEED_FSCK); 3666 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3667 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3668 "run fsck to fix.", 3669 __func__, inode->i_ino, inode->i_blocks, 3670 reserved_blocks, 3671 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3672 } 3673 3674 return ret; 3675 } 3676 3677 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3678 pgoff_t off, block_t block, block_t len, u32 flags) 3679 { 3680 struct request_queue *q = bdev_get_queue(bdev); 3681 sector_t sector = SECTOR_FROM_BLOCK(block); 3682 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3683 int ret = 0; 3684 3685 if (!q) 3686 return -ENXIO; 3687 3688 if (flags & F2FS_TRIM_FILE_DISCARD) 3689 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS, 3690 blk_queue_secure_erase(q) ? 3691 BLKDEV_DISCARD_SECURE : 0); 3692 3693 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3694 if (IS_ENCRYPTED(inode)) 3695 ret = fscrypt_zeroout_range(inode, off, block, len); 3696 else 3697 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3698 GFP_NOFS, 0); 3699 } 3700 3701 return ret; 3702 } 3703 3704 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3705 { 3706 struct inode *inode = file_inode(filp); 3707 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3708 struct address_space *mapping = inode->i_mapping; 3709 struct block_device *prev_bdev = NULL; 3710 struct f2fs_sectrim_range range; 3711 pgoff_t index, pg_end, prev_index = 0; 3712 block_t prev_block = 0, len = 0; 3713 loff_t end_addr; 3714 bool to_end = false; 3715 int ret = 0; 3716 3717 if (!(filp->f_mode & FMODE_WRITE)) 3718 return -EBADF; 3719 3720 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3721 sizeof(range))) 3722 return -EFAULT; 3723 3724 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3725 !S_ISREG(inode->i_mode)) 3726 return -EINVAL; 3727 3728 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3729 !f2fs_hw_support_discard(sbi)) || 3730 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3731 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3732 return -EOPNOTSUPP; 3733 3734 file_start_write(filp); 3735 inode_lock(inode); 3736 3737 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3738 range.start >= inode->i_size) { 3739 ret = -EINVAL; 3740 goto err; 3741 } 3742 3743 if (range.len == 0) 3744 goto err; 3745 3746 if (inode->i_size - range.start > range.len) { 3747 end_addr = range.start + range.len; 3748 } else { 3749 end_addr = range.len == (u64)-1 ? 3750 sbi->sb->s_maxbytes : inode->i_size; 3751 to_end = true; 3752 } 3753 3754 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3755 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3756 ret = -EINVAL; 3757 goto err; 3758 } 3759 3760 index = F2FS_BYTES_TO_BLK(range.start); 3761 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3762 3763 ret = f2fs_convert_inline_inode(inode); 3764 if (ret) 3765 goto err; 3766 3767 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3768 filemap_invalidate_lock(mapping); 3769 3770 ret = filemap_write_and_wait_range(mapping, range.start, 3771 to_end ? LLONG_MAX : end_addr - 1); 3772 if (ret) 3773 goto out; 3774 3775 truncate_inode_pages_range(mapping, range.start, 3776 to_end ? -1 : end_addr - 1); 3777 3778 while (index < pg_end) { 3779 struct dnode_of_data dn; 3780 pgoff_t end_offset, count; 3781 int i; 3782 3783 set_new_dnode(&dn, inode, NULL, NULL, 0); 3784 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3785 if (ret) { 3786 if (ret == -ENOENT) { 3787 index = f2fs_get_next_page_offset(&dn, index); 3788 continue; 3789 } 3790 goto out; 3791 } 3792 3793 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3794 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3795 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3796 struct block_device *cur_bdev; 3797 block_t blkaddr = f2fs_data_blkaddr(&dn); 3798 3799 if (!__is_valid_data_blkaddr(blkaddr)) 3800 continue; 3801 3802 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3803 DATA_GENERIC_ENHANCE)) { 3804 ret = -EFSCORRUPTED; 3805 f2fs_put_dnode(&dn); 3806 goto out; 3807 } 3808 3809 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3810 if (f2fs_is_multi_device(sbi)) { 3811 int di = f2fs_target_device_index(sbi, blkaddr); 3812 3813 blkaddr -= FDEV(di).start_blk; 3814 } 3815 3816 if (len) { 3817 if (prev_bdev == cur_bdev && 3818 index == prev_index + len && 3819 blkaddr == prev_block + len) { 3820 len++; 3821 } else { 3822 ret = f2fs_secure_erase(prev_bdev, 3823 inode, prev_index, prev_block, 3824 len, range.flags); 3825 if (ret) { 3826 f2fs_put_dnode(&dn); 3827 goto out; 3828 } 3829 3830 len = 0; 3831 } 3832 } 3833 3834 if (!len) { 3835 prev_bdev = cur_bdev; 3836 prev_index = index; 3837 prev_block = blkaddr; 3838 len = 1; 3839 } 3840 } 3841 3842 f2fs_put_dnode(&dn); 3843 3844 if (fatal_signal_pending(current)) { 3845 ret = -EINTR; 3846 goto out; 3847 } 3848 cond_resched(); 3849 } 3850 3851 if (len) 3852 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 3853 prev_block, len, range.flags); 3854 out: 3855 filemap_invalidate_unlock(mapping); 3856 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3857 err: 3858 inode_unlock(inode); 3859 file_end_write(filp); 3860 3861 return ret; 3862 } 3863 3864 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 3865 { 3866 struct inode *inode = file_inode(filp); 3867 struct f2fs_comp_option option; 3868 3869 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3870 return -EOPNOTSUPP; 3871 3872 inode_lock_shared(inode); 3873 3874 if (!f2fs_compressed_file(inode)) { 3875 inode_unlock_shared(inode); 3876 return -ENODATA; 3877 } 3878 3879 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 3880 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 3881 3882 inode_unlock_shared(inode); 3883 3884 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 3885 sizeof(option))) 3886 return -EFAULT; 3887 3888 return 0; 3889 } 3890 3891 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 3892 { 3893 struct inode *inode = file_inode(filp); 3894 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3895 struct f2fs_comp_option option; 3896 int ret = 0; 3897 3898 if (!f2fs_sb_has_compression(sbi)) 3899 return -EOPNOTSUPP; 3900 3901 if (!(filp->f_mode & FMODE_WRITE)) 3902 return -EBADF; 3903 3904 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 3905 sizeof(option))) 3906 return -EFAULT; 3907 3908 if (!f2fs_compressed_file(inode) || 3909 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 3910 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 3911 option.algorithm >= COMPRESS_MAX) 3912 return -EINVAL; 3913 3914 file_start_write(filp); 3915 inode_lock(inode); 3916 3917 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 3918 ret = -EBUSY; 3919 goto out; 3920 } 3921 3922 if (inode->i_size != 0) { 3923 ret = -EFBIG; 3924 goto out; 3925 } 3926 3927 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 3928 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 3929 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size; 3930 f2fs_mark_inode_dirty_sync(inode, true); 3931 3932 if (!f2fs_is_compress_backend_ready(inode)) 3933 f2fs_warn(sbi, "compression algorithm is successfully set, " 3934 "but current kernel doesn't support this algorithm."); 3935 out: 3936 inode_unlock(inode); 3937 file_end_write(filp); 3938 3939 return ret; 3940 } 3941 3942 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 3943 { 3944 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 3945 struct address_space *mapping = inode->i_mapping; 3946 struct page *page; 3947 pgoff_t redirty_idx = page_idx; 3948 int i, page_len = 0, ret = 0; 3949 3950 page_cache_ra_unbounded(&ractl, len, 0); 3951 3952 for (i = 0; i < len; i++, page_idx++) { 3953 page = read_cache_page(mapping, page_idx, NULL, NULL); 3954 if (IS_ERR(page)) { 3955 ret = PTR_ERR(page); 3956 break; 3957 } 3958 page_len++; 3959 } 3960 3961 for (i = 0; i < page_len; i++, redirty_idx++) { 3962 page = find_lock_page(mapping, redirty_idx); 3963 if (!page) { 3964 ret = -ENOMEM; 3965 break; 3966 } 3967 set_page_dirty(page); 3968 f2fs_put_page(page, 1); 3969 f2fs_put_page(page, 0); 3970 } 3971 3972 return ret; 3973 } 3974 3975 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg) 3976 { 3977 struct inode *inode = file_inode(filp); 3978 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3979 struct f2fs_inode_info *fi = F2FS_I(inode); 3980 pgoff_t page_idx = 0, last_idx; 3981 unsigned int blk_per_seg = sbi->blocks_per_seg; 3982 int cluster_size = F2FS_I(inode)->i_cluster_size; 3983 int count, ret; 3984 3985 if (!f2fs_sb_has_compression(sbi) || 3986 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 3987 return -EOPNOTSUPP; 3988 3989 if (!(filp->f_mode & FMODE_WRITE)) 3990 return -EBADF; 3991 3992 if (!f2fs_compressed_file(inode)) 3993 return -EINVAL; 3994 3995 f2fs_balance_fs(F2FS_I_SB(inode), true); 3996 3997 file_start_write(filp); 3998 inode_lock(inode); 3999 4000 if (!f2fs_is_compress_backend_ready(inode)) { 4001 ret = -EOPNOTSUPP; 4002 goto out; 4003 } 4004 4005 if (f2fs_is_mmap_file(inode)) { 4006 ret = -EBUSY; 4007 goto out; 4008 } 4009 4010 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4011 if (ret) 4012 goto out; 4013 4014 if (!atomic_read(&fi->i_compr_blocks)) 4015 goto out; 4016 4017 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4018 4019 count = last_idx - page_idx; 4020 while (count) { 4021 int len = min(cluster_size, count); 4022 4023 ret = redirty_blocks(inode, page_idx, len); 4024 if (ret < 0) 4025 break; 4026 4027 if (get_dirty_pages(inode) >= blk_per_seg) 4028 filemap_fdatawrite(inode->i_mapping); 4029 4030 count -= len; 4031 page_idx += len; 4032 } 4033 4034 if (!ret) 4035 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4036 LLONG_MAX); 4037 4038 if (ret) 4039 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4040 __func__, ret); 4041 out: 4042 inode_unlock(inode); 4043 file_end_write(filp); 4044 4045 return ret; 4046 } 4047 4048 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg) 4049 { 4050 struct inode *inode = file_inode(filp); 4051 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4052 pgoff_t page_idx = 0, last_idx; 4053 unsigned int blk_per_seg = sbi->blocks_per_seg; 4054 int cluster_size = F2FS_I(inode)->i_cluster_size; 4055 int count, ret; 4056 4057 if (!f2fs_sb_has_compression(sbi) || 4058 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4059 return -EOPNOTSUPP; 4060 4061 if (!(filp->f_mode & FMODE_WRITE)) 4062 return -EBADF; 4063 4064 if (!f2fs_compressed_file(inode)) 4065 return -EINVAL; 4066 4067 f2fs_balance_fs(F2FS_I_SB(inode), true); 4068 4069 file_start_write(filp); 4070 inode_lock(inode); 4071 4072 if (!f2fs_is_compress_backend_ready(inode)) { 4073 ret = -EOPNOTSUPP; 4074 goto out; 4075 } 4076 4077 if (f2fs_is_mmap_file(inode)) { 4078 ret = -EBUSY; 4079 goto out; 4080 } 4081 4082 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4083 if (ret) 4084 goto out; 4085 4086 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4087 4088 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4089 4090 count = last_idx - page_idx; 4091 while (count) { 4092 int len = min(cluster_size, count); 4093 4094 ret = redirty_blocks(inode, page_idx, len); 4095 if (ret < 0) 4096 break; 4097 4098 if (get_dirty_pages(inode) >= blk_per_seg) 4099 filemap_fdatawrite(inode->i_mapping); 4100 4101 count -= len; 4102 page_idx += len; 4103 } 4104 4105 if (!ret) 4106 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4107 LLONG_MAX); 4108 4109 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4110 4111 if (ret) 4112 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4113 __func__, ret); 4114 out: 4115 inode_unlock(inode); 4116 file_end_write(filp); 4117 4118 return ret; 4119 } 4120 4121 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4122 { 4123 switch (cmd) { 4124 case FS_IOC_GETVERSION: 4125 return f2fs_ioc_getversion(filp, arg); 4126 case F2FS_IOC_START_ATOMIC_WRITE: 4127 return f2fs_ioc_start_atomic_write(filp); 4128 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4129 return f2fs_ioc_commit_atomic_write(filp); 4130 case F2FS_IOC_START_VOLATILE_WRITE: 4131 return f2fs_ioc_start_volatile_write(filp); 4132 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4133 return f2fs_ioc_release_volatile_write(filp); 4134 case F2FS_IOC_ABORT_VOLATILE_WRITE: 4135 return f2fs_ioc_abort_volatile_write(filp); 4136 case F2FS_IOC_SHUTDOWN: 4137 return f2fs_ioc_shutdown(filp, arg); 4138 case FITRIM: 4139 return f2fs_ioc_fitrim(filp, arg); 4140 case FS_IOC_SET_ENCRYPTION_POLICY: 4141 return f2fs_ioc_set_encryption_policy(filp, arg); 4142 case FS_IOC_GET_ENCRYPTION_POLICY: 4143 return f2fs_ioc_get_encryption_policy(filp, arg); 4144 case FS_IOC_GET_ENCRYPTION_PWSALT: 4145 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4146 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4147 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4148 case FS_IOC_ADD_ENCRYPTION_KEY: 4149 return f2fs_ioc_add_encryption_key(filp, arg); 4150 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4151 return f2fs_ioc_remove_encryption_key(filp, arg); 4152 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4153 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4154 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4155 return f2fs_ioc_get_encryption_key_status(filp, arg); 4156 case FS_IOC_GET_ENCRYPTION_NONCE: 4157 return f2fs_ioc_get_encryption_nonce(filp, arg); 4158 case F2FS_IOC_GARBAGE_COLLECT: 4159 return f2fs_ioc_gc(filp, arg); 4160 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4161 return f2fs_ioc_gc_range(filp, arg); 4162 case F2FS_IOC_WRITE_CHECKPOINT: 4163 return f2fs_ioc_write_checkpoint(filp, arg); 4164 case F2FS_IOC_DEFRAGMENT: 4165 return f2fs_ioc_defragment(filp, arg); 4166 case F2FS_IOC_MOVE_RANGE: 4167 return f2fs_ioc_move_range(filp, arg); 4168 case F2FS_IOC_FLUSH_DEVICE: 4169 return f2fs_ioc_flush_device(filp, arg); 4170 case F2FS_IOC_GET_FEATURES: 4171 return f2fs_ioc_get_features(filp, arg); 4172 case F2FS_IOC_GET_PIN_FILE: 4173 return f2fs_ioc_get_pin_file(filp, arg); 4174 case F2FS_IOC_SET_PIN_FILE: 4175 return f2fs_ioc_set_pin_file(filp, arg); 4176 case F2FS_IOC_PRECACHE_EXTENTS: 4177 return f2fs_ioc_precache_extents(filp, arg); 4178 case F2FS_IOC_RESIZE_FS: 4179 return f2fs_ioc_resize_fs(filp, arg); 4180 case FS_IOC_ENABLE_VERITY: 4181 return f2fs_ioc_enable_verity(filp, arg); 4182 case FS_IOC_MEASURE_VERITY: 4183 return f2fs_ioc_measure_verity(filp, arg); 4184 case FS_IOC_READ_VERITY_METADATA: 4185 return f2fs_ioc_read_verity_metadata(filp, arg); 4186 case FS_IOC_GETFSLABEL: 4187 return f2fs_ioc_getfslabel(filp, arg); 4188 case FS_IOC_SETFSLABEL: 4189 return f2fs_ioc_setfslabel(filp, arg); 4190 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4191 return f2fs_get_compress_blocks(filp, arg); 4192 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4193 return f2fs_release_compress_blocks(filp, arg); 4194 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4195 return f2fs_reserve_compress_blocks(filp, arg); 4196 case F2FS_IOC_SEC_TRIM_FILE: 4197 return f2fs_sec_trim_file(filp, arg); 4198 case F2FS_IOC_GET_COMPRESS_OPTION: 4199 return f2fs_ioc_get_compress_option(filp, arg); 4200 case F2FS_IOC_SET_COMPRESS_OPTION: 4201 return f2fs_ioc_set_compress_option(filp, arg); 4202 case F2FS_IOC_DECOMPRESS_FILE: 4203 return f2fs_ioc_decompress_file(filp, arg); 4204 case F2FS_IOC_COMPRESS_FILE: 4205 return f2fs_ioc_compress_file(filp, arg); 4206 default: 4207 return -ENOTTY; 4208 } 4209 } 4210 4211 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4212 { 4213 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4214 return -EIO; 4215 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4216 return -ENOSPC; 4217 4218 return __f2fs_ioctl(filp, cmd, arg); 4219 } 4220 4221 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter) 4222 { 4223 struct file *file = iocb->ki_filp; 4224 struct inode *inode = file_inode(file); 4225 int ret; 4226 4227 if (!f2fs_is_compress_backend_ready(inode)) 4228 return -EOPNOTSUPP; 4229 4230 ret = generic_file_read_iter(iocb, iter); 4231 4232 if (ret > 0) 4233 f2fs_update_iostat(F2FS_I_SB(inode), APP_READ_IO, ret); 4234 4235 return ret; 4236 } 4237 4238 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4239 { 4240 struct file *file = iocb->ki_filp; 4241 struct inode *inode = file_inode(file); 4242 ssize_t ret; 4243 4244 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4245 ret = -EIO; 4246 goto out; 4247 } 4248 4249 if (!f2fs_is_compress_backend_ready(inode)) { 4250 ret = -EOPNOTSUPP; 4251 goto out; 4252 } 4253 4254 if (iocb->ki_flags & IOCB_NOWAIT) { 4255 if (!inode_trylock(inode)) { 4256 ret = -EAGAIN; 4257 goto out; 4258 } 4259 } else { 4260 inode_lock(inode); 4261 } 4262 4263 if (unlikely(IS_IMMUTABLE(inode))) { 4264 ret = -EPERM; 4265 goto unlock; 4266 } 4267 4268 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4269 ret = -EPERM; 4270 goto unlock; 4271 } 4272 4273 ret = generic_write_checks(iocb, from); 4274 if (ret > 0) { 4275 bool preallocated = false; 4276 size_t target_size = 0; 4277 int err; 4278 4279 if (iov_iter_fault_in_readable(from, iov_iter_count(from))) 4280 set_inode_flag(inode, FI_NO_PREALLOC); 4281 4282 if ((iocb->ki_flags & IOCB_NOWAIT)) { 4283 if (!f2fs_overwrite_io(inode, iocb->ki_pos, 4284 iov_iter_count(from)) || 4285 f2fs_has_inline_data(inode) || 4286 f2fs_force_buffered_io(inode, iocb, from)) { 4287 clear_inode_flag(inode, FI_NO_PREALLOC); 4288 inode_unlock(inode); 4289 ret = -EAGAIN; 4290 goto out; 4291 } 4292 goto write; 4293 } 4294 4295 if (is_inode_flag_set(inode, FI_NO_PREALLOC)) 4296 goto write; 4297 4298 if (iocb->ki_flags & IOCB_DIRECT) { 4299 /* 4300 * Convert inline data for Direct I/O before entering 4301 * f2fs_direct_IO(). 4302 */ 4303 err = f2fs_convert_inline_inode(inode); 4304 if (err) 4305 goto out_err; 4306 /* 4307 * If force_buffere_io() is true, we have to allocate 4308 * blocks all the time, since f2fs_direct_IO will fall 4309 * back to buffered IO. 4310 */ 4311 if (!f2fs_force_buffered_io(inode, iocb, from) && 4312 f2fs_lfs_mode(F2FS_I_SB(inode))) 4313 goto write; 4314 } 4315 preallocated = true; 4316 target_size = iocb->ki_pos + iov_iter_count(from); 4317 4318 err = f2fs_preallocate_blocks(iocb, from); 4319 if (err) { 4320 out_err: 4321 clear_inode_flag(inode, FI_NO_PREALLOC); 4322 inode_unlock(inode); 4323 ret = err; 4324 goto out; 4325 } 4326 write: 4327 ret = __generic_file_write_iter(iocb, from); 4328 clear_inode_flag(inode, FI_NO_PREALLOC); 4329 4330 /* if we couldn't write data, we should deallocate blocks. */ 4331 if (preallocated && i_size_read(inode) < target_size) { 4332 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4333 filemap_invalidate_lock(inode->i_mapping); 4334 f2fs_truncate(inode); 4335 filemap_invalidate_unlock(inode->i_mapping); 4336 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4337 } 4338 4339 if (ret > 0) 4340 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret); 4341 } 4342 unlock: 4343 inode_unlock(inode); 4344 out: 4345 trace_f2fs_file_write_iter(inode, iocb->ki_pos, 4346 iov_iter_count(from), ret); 4347 if (ret > 0) 4348 ret = generic_write_sync(iocb, ret); 4349 return ret; 4350 } 4351 4352 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4353 int advice) 4354 { 4355 struct inode *inode; 4356 struct address_space *mapping; 4357 struct backing_dev_info *bdi; 4358 4359 if (advice == POSIX_FADV_SEQUENTIAL) { 4360 inode = file_inode(filp); 4361 if (S_ISFIFO(inode->i_mode)) 4362 return -ESPIPE; 4363 4364 mapping = filp->f_mapping; 4365 if (!mapping || len < 0) 4366 return -EINVAL; 4367 4368 bdi = inode_to_bdi(mapping->host); 4369 filp->f_ra.ra_pages = bdi->ra_pages * 4370 F2FS_I_SB(inode)->seq_file_ra_mul; 4371 spin_lock(&filp->f_lock); 4372 filp->f_mode &= ~FMODE_RANDOM; 4373 spin_unlock(&filp->f_lock); 4374 return 0; 4375 } 4376 4377 return generic_fadvise(filp, offset, len, advice); 4378 } 4379 4380 #ifdef CONFIG_COMPAT 4381 struct compat_f2fs_gc_range { 4382 u32 sync; 4383 compat_u64 start; 4384 compat_u64 len; 4385 }; 4386 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4387 struct compat_f2fs_gc_range) 4388 4389 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4390 { 4391 struct compat_f2fs_gc_range __user *urange; 4392 struct f2fs_gc_range range; 4393 int err; 4394 4395 urange = compat_ptr(arg); 4396 err = get_user(range.sync, &urange->sync); 4397 err |= get_user(range.start, &urange->start); 4398 err |= get_user(range.len, &urange->len); 4399 if (err) 4400 return -EFAULT; 4401 4402 return __f2fs_ioc_gc_range(file, &range); 4403 } 4404 4405 struct compat_f2fs_move_range { 4406 u32 dst_fd; 4407 compat_u64 pos_in; 4408 compat_u64 pos_out; 4409 compat_u64 len; 4410 }; 4411 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 4412 struct compat_f2fs_move_range) 4413 4414 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 4415 { 4416 struct compat_f2fs_move_range __user *urange; 4417 struct f2fs_move_range range; 4418 int err; 4419 4420 urange = compat_ptr(arg); 4421 err = get_user(range.dst_fd, &urange->dst_fd); 4422 err |= get_user(range.pos_in, &urange->pos_in); 4423 err |= get_user(range.pos_out, &urange->pos_out); 4424 err |= get_user(range.len, &urange->len); 4425 if (err) 4426 return -EFAULT; 4427 4428 return __f2fs_ioc_move_range(file, &range); 4429 } 4430 4431 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 4432 { 4433 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 4434 return -EIO; 4435 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 4436 return -ENOSPC; 4437 4438 switch (cmd) { 4439 case FS_IOC32_GETVERSION: 4440 cmd = FS_IOC_GETVERSION; 4441 break; 4442 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 4443 return f2fs_compat_ioc_gc_range(file, arg); 4444 case F2FS_IOC32_MOVE_RANGE: 4445 return f2fs_compat_ioc_move_range(file, arg); 4446 case F2FS_IOC_START_ATOMIC_WRITE: 4447 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4448 case F2FS_IOC_START_VOLATILE_WRITE: 4449 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4450 case F2FS_IOC_ABORT_VOLATILE_WRITE: 4451 case F2FS_IOC_SHUTDOWN: 4452 case FITRIM: 4453 case FS_IOC_SET_ENCRYPTION_POLICY: 4454 case FS_IOC_GET_ENCRYPTION_PWSALT: 4455 case FS_IOC_GET_ENCRYPTION_POLICY: 4456 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4457 case FS_IOC_ADD_ENCRYPTION_KEY: 4458 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4459 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4460 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4461 case FS_IOC_GET_ENCRYPTION_NONCE: 4462 case F2FS_IOC_GARBAGE_COLLECT: 4463 case F2FS_IOC_WRITE_CHECKPOINT: 4464 case F2FS_IOC_DEFRAGMENT: 4465 case F2FS_IOC_FLUSH_DEVICE: 4466 case F2FS_IOC_GET_FEATURES: 4467 case F2FS_IOC_GET_PIN_FILE: 4468 case F2FS_IOC_SET_PIN_FILE: 4469 case F2FS_IOC_PRECACHE_EXTENTS: 4470 case F2FS_IOC_RESIZE_FS: 4471 case FS_IOC_ENABLE_VERITY: 4472 case FS_IOC_MEASURE_VERITY: 4473 case FS_IOC_READ_VERITY_METADATA: 4474 case FS_IOC_GETFSLABEL: 4475 case FS_IOC_SETFSLABEL: 4476 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4477 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4478 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4479 case F2FS_IOC_SEC_TRIM_FILE: 4480 case F2FS_IOC_GET_COMPRESS_OPTION: 4481 case F2FS_IOC_SET_COMPRESS_OPTION: 4482 case F2FS_IOC_DECOMPRESS_FILE: 4483 case F2FS_IOC_COMPRESS_FILE: 4484 break; 4485 default: 4486 return -ENOIOCTLCMD; 4487 } 4488 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 4489 } 4490 #endif 4491 4492 const struct file_operations f2fs_file_operations = { 4493 .llseek = f2fs_llseek, 4494 .read_iter = f2fs_file_read_iter, 4495 .write_iter = f2fs_file_write_iter, 4496 .open = f2fs_file_open, 4497 .release = f2fs_release_file, 4498 .mmap = f2fs_file_mmap, 4499 .flush = f2fs_file_flush, 4500 .fsync = f2fs_sync_file, 4501 .fallocate = f2fs_fallocate, 4502 .unlocked_ioctl = f2fs_ioctl, 4503 #ifdef CONFIG_COMPAT 4504 .compat_ioctl = f2fs_compat_ioctl, 4505 #endif 4506 .splice_read = generic_file_splice_read, 4507 .splice_write = iter_file_splice_write, 4508 .fadvise = f2fs_file_fadvise, 4509 }; 4510