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 25 #include "f2fs.h" 26 #include "node.h" 27 #include "segment.h" 28 #include "xattr.h" 29 #include "acl.h" 30 #include "gc.h" 31 #include "trace.h" 32 #include <trace/events/f2fs.h> 33 34 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 35 { 36 struct inode *inode = file_inode(vmf->vma->vm_file); 37 vm_fault_t ret; 38 39 down_read(&F2FS_I(inode)->i_mmap_sem); 40 ret = filemap_fault(vmf); 41 up_read(&F2FS_I(inode)->i_mmap_sem); 42 43 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret); 44 45 return ret; 46 } 47 48 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 49 { 50 struct page *page = vmf->page; 51 struct inode *inode = file_inode(vmf->vma->vm_file); 52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 53 struct dnode_of_data dn; 54 bool need_alloc = true; 55 int err = 0; 56 57 if (unlikely(f2fs_cp_error(sbi))) { 58 err = -EIO; 59 goto err; 60 } 61 62 if (!f2fs_is_checkpoint_ready(sbi)) { 63 err = -ENOSPC; 64 goto err; 65 } 66 67 #ifdef CONFIG_F2FS_FS_COMPRESSION 68 if (f2fs_compressed_file(inode)) { 69 int ret = f2fs_is_compressed_cluster(inode, page->index); 70 71 if (ret < 0) { 72 err = ret; 73 goto err; 74 } else if (ret) { 75 if (ret < F2FS_I(inode)->i_cluster_size) { 76 err = -EAGAIN; 77 goto err; 78 } 79 need_alloc = false; 80 } 81 } 82 #endif 83 /* should do out of any locked page */ 84 if (need_alloc) 85 f2fs_balance_fs(sbi, true); 86 87 sb_start_pagefault(inode->i_sb); 88 89 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 90 91 file_update_time(vmf->vma->vm_file); 92 down_read(&F2FS_I(inode)->i_mmap_sem); 93 lock_page(page); 94 if (unlikely(page->mapping != inode->i_mapping || 95 page_offset(page) > i_size_read(inode) || 96 !PageUptodate(page))) { 97 unlock_page(page); 98 err = -EFAULT; 99 goto out_sem; 100 } 101 102 if (need_alloc) { 103 /* block allocation */ 104 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true); 105 set_new_dnode(&dn, inode, NULL, NULL, 0); 106 err = f2fs_get_block(&dn, page->index); 107 f2fs_put_dnode(&dn); 108 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false); 109 if (err) { 110 unlock_page(page); 111 goto out_sem; 112 } 113 } 114 115 /* fill the page */ 116 f2fs_wait_on_page_writeback(page, DATA, false, true); 117 118 /* wait for GCed page writeback via META_MAPPING */ 119 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 120 121 /* 122 * check to see if the page is mapped already (no holes) 123 */ 124 if (PageMappedToDisk(page)) 125 goto out_sem; 126 127 /* page is wholly or partially inside EOF */ 128 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 129 i_size_read(inode)) { 130 loff_t offset; 131 132 offset = i_size_read(inode) & ~PAGE_MASK; 133 zero_user_segment(page, offset, PAGE_SIZE); 134 } 135 set_page_dirty(page); 136 if (!PageUptodate(page)) 137 SetPageUptodate(page); 138 139 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE); 140 f2fs_update_time(sbi, REQ_TIME); 141 142 trace_f2fs_vm_page_mkwrite(page, DATA); 143 out_sem: 144 up_read(&F2FS_I(inode)->i_mmap_sem); 145 146 sb_end_pagefault(inode->i_sb); 147 err: 148 return block_page_mkwrite_return(err); 149 } 150 151 static const struct vm_operations_struct f2fs_file_vm_ops = { 152 .fault = f2fs_filemap_fault, 153 .map_pages = filemap_map_pages, 154 .page_mkwrite = f2fs_vm_page_mkwrite, 155 }; 156 157 static int get_parent_ino(struct inode *inode, nid_t *pino) 158 { 159 struct dentry *dentry; 160 161 inode = igrab(inode); 162 dentry = d_find_any_alias(inode); 163 iput(inode); 164 if (!dentry) 165 return 0; 166 167 *pino = parent_ino(dentry); 168 dput(dentry); 169 return 1; 170 } 171 172 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 173 { 174 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 175 enum cp_reason_type cp_reason = CP_NO_NEEDED; 176 177 if (!S_ISREG(inode->i_mode)) 178 cp_reason = CP_NON_REGULAR; 179 else if (f2fs_compressed_file(inode)) 180 cp_reason = CP_COMPRESSED; 181 else if (inode->i_nlink != 1) 182 cp_reason = CP_HARDLINK; 183 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 184 cp_reason = CP_SB_NEED_CP; 185 else if (file_wrong_pino(inode)) 186 cp_reason = CP_WRONG_PINO; 187 else if (!f2fs_space_for_roll_forward(sbi)) 188 cp_reason = CP_NO_SPC_ROLL; 189 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 190 cp_reason = CP_NODE_NEED_CP; 191 else if (test_opt(sbi, FASTBOOT)) 192 cp_reason = CP_FASTBOOT_MODE; 193 else if (F2FS_OPTION(sbi).active_logs == 2) 194 cp_reason = CP_SPEC_LOG_NUM; 195 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 196 f2fs_need_dentry_mark(sbi, inode->i_ino) && 197 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 198 TRANS_DIR_INO)) 199 cp_reason = CP_RECOVER_DIR; 200 201 return cp_reason; 202 } 203 204 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 205 { 206 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 207 bool ret = false; 208 /* But we need to avoid that there are some inode updates */ 209 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 210 ret = true; 211 f2fs_put_page(i, 0); 212 return ret; 213 } 214 215 static void try_to_fix_pino(struct inode *inode) 216 { 217 struct f2fs_inode_info *fi = F2FS_I(inode); 218 nid_t pino; 219 220 down_write(&fi->i_sem); 221 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 222 get_parent_ino(inode, &pino)) { 223 f2fs_i_pino_write(inode, pino); 224 file_got_pino(inode); 225 } 226 up_write(&fi->i_sem); 227 } 228 229 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 230 int datasync, bool atomic) 231 { 232 struct inode *inode = file->f_mapping->host; 233 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 234 nid_t ino = inode->i_ino; 235 int ret = 0; 236 enum cp_reason_type cp_reason = 0; 237 struct writeback_control wbc = { 238 .sync_mode = WB_SYNC_ALL, 239 .nr_to_write = LONG_MAX, 240 .for_reclaim = 0, 241 }; 242 unsigned int seq_id = 0; 243 244 if (unlikely(f2fs_readonly(inode->i_sb) || 245 is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 246 return 0; 247 248 trace_f2fs_sync_file_enter(inode); 249 250 if (S_ISDIR(inode->i_mode)) 251 goto go_write; 252 253 /* if fdatasync is triggered, let's do in-place-update */ 254 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 255 set_inode_flag(inode, FI_NEED_IPU); 256 ret = file_write_and_wait_range(file, start, end); 257 clear_inode_flag(inode, FI_NEED_IPU); 258 259 if (ret) { 260 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 261 return ret; 262 } 263 264 /* if the inode is dirty, let's recover all the time */ 265 if (!f2fs_skip_inode_update(inode, datasync)) { 266 f2fs_write_inode(inode, NULL); 267 goto go_write; 268 } 269 270 /* 271 * if there is no written data, don't waste time to write recovery info. 272 */ 273 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 274 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 275 276 /* it may call write_inode just prior to fsync */ 277 if (need_inode_page_update(sbi, ino)) 278 goto go_write; 279 280 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 281 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 282 goto flush_out; 283 goto out; 284 } 285 go_write: 286 /* 287 * Both of fdatasync() and fsync() are able to be recovered from 288 * sudden-power-off. 289 */ 290 down_read(&F2FS_I(inode)->i_sem); 291 cp_reason = need_do_checkpoint(inode); 292 up_read(&F2FS_I(inode)->i_sem); 293 294 if (cp_reason) { 295 /* all the dirty node pages should be flushed for POR */ 296 ret = f2fs_sync_fs(inode->i_sb, 1); 297 298 /* 299 * We've secured consistency through sync_fs. Following pino 300 * will be used only for fsynced inodes after checkpoint. 301 */ 302 try_to_fix_pino(inode); 303 clear_inode_flag(inode, FI_APPEND_WRITE); 304 clear_inode_flag(inode, FI_UPDATE_WRITE); 305 goto out; 306 } 307 sync_nodes: 308 atomic_inc(&sbi->wb_sync_req[NODE]); 309 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 310 atomic_dec(&sbi->wb_sync_req[NODE]); 311 if (ret) 312 goto out; 313 314 /* if cp_error was enabled, we should avoid infinite loop */ 315 if (unlikely(f2fs_cp_error(sbi))) { 316 ret = -EIO; 317 goto out; 318 } 319 320 if (f2fs_need_inode_block_update(sbi, ino)) { 321 f2fs_mark_inode_dirty_sync(inode, true); 322 f2fs_write_inode(inode, NULL); 323 goto sync_nodes; 324 } 325 326 /* 327 * If it's atomic_write, it's just fine to keep write ordering. So 328 * here we don't need to wait for node write completion, since we use 329 * node chain which serializes node blocks. If one of node writes are 330 * reordered, we can see simply broken chain, resulting in stopping 331 * roll-forward recovery. It means we'll recover all or none node blocks 332 * given fsync mark. 333 */ 334 if (!atomic) { 335 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 336 if (ret) 337 goto out; 338 } 339 340 /* once recovery info is written, don't need to tack this */ 341 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 342 clear_inode_flag(inode, FI_APPEND_WRITE); 343 flush_out: 344 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 345 ret = f2fs_issue_flush(sbi, inode->i_ino); 346 if (!ret) { 347 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 348 clear_inode_flag(inode, FI_UPDATE_WRITE); 349 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 350 } 351 f2fs_update_time(sbi, REQ_TIME); 352 out: 353 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 354 f2fs_trace_ios(NULL, 1); 355 return ret; 356 } 357 358 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 359 { 360 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 361 return -EIO; 362 return f2fs_do_sync_file(file, start, end, datasync, false); 363 } 364 365 static pgoff_t __get_first_dirty_index(struct address_space *mapping, 366 pgoff_t pgofs, int whence) 367 { 368 struct page *page; 369 int nr_pages; 370 371 if (whence != SEEK_DATA) 372 return 0; 373 374 /* find first dirty page index */ 375 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY, 376 1, &page); 377 if (!nr_pages) 378 return ULONG_MAX; 379 pgofs = page->index; 380 put_page(page); 381 return pgofs; 382 } 383 384 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr, 385 pgoff_t dirty, pgoff_t pgofs, int whence) 386 { 387 switch (whence) { 388 case SEEK_DATA: 389 if ((blkaddr == NEW_ADDR && dirty == pgofs) || 390 __is_valid_data_blkaddr(blkaddr)) 391 return true; 392 break; 393 case SEEK_HOLE: 394 if (blkaddr == NULL_ADDR) 395 return true; 396 break; 397 } 398 return false; 399 } 400 401 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 402 { 403 struct inode *inode = file->f_mapping->host; 404 loff_t maxbytes = inode->i_sb->s_maxbytes; 405 struct dnode_of_data dn; 406 pgoff_t pgofs, end_offset, dirty; 407 loff_t data_ofs = offset; 408 loff_t isize; 409 int err = 0; 410 411 inode_lock(inode); 412 413 isize = i_size_read(inode); 414 if (offset >= isize) 415 goto fail; 416 417 /* handle inline data case */ 418 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) { 419 if (whence == SEEK_HOLE) 420 data_ofs = isize; 421 goto found; 422 } 423 424 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 425 426 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence); 427 428 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 429 set_new_dnode(&dn, inode, NULL, NULL, 0); 430 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 431 if (err && err != -ENOENT) { 432 goto fail; 433 } else if (err == -ENOENT) { 434 /* direct node does not exists */ 435 if (whence == SEEK_DATA) { 436 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 437 continue; 438 } else { 439 goto found; 440 } 441 } 442 443 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 444 445 /* find data/hole in dnode block */ 446 for (; dn.ofs_in_node < end_offset; 447 dn.ofs_in_node++, pgofs++, 448 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 449 block_t blkaddr; 450 451 blkaddr = datablock_addr(dn.inode, 452 dn.node_page, dn.ofs_in_node); 453 454 if (__is_valid_data_blkaddr(blkaddr) && 455 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 456 blkaddr, DATA_GENERIC_ENHANCE)) { 457 f2fs_put_dnode(&dn); 458 goto fail; 459 } 460 461 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty, 462 pgofs, whence)) { 463 f2fs_put_dnode(&dn); 464 goto found; 465 } 466 } 467 f2fs_put_dnode(&dn); 468 } 469 470 if (whence == SEEK_DATA) 471 goto fail; 472 found: 473 if (whence == SEEK_HOLE && data_ofs > isize) 474 data_ofs = isize; 475 inode_unlock(inode); 476 return vfs_setpos(file, data_ofs, maxbytes); 477 fail: 478 inode_unlock(inode); 479 return -ENXIO; 480 } 481 482 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 483 { 484 struct inode *inode = file->f_mapping->host; 485 loff_t maxbytes = inode->i_sb->s_maxbytes; 486 487 switch (whence) { 488 case SEEK_SET: 489 case SEEK_CUR: 490 case SEEK_END: 491 return generic_file_llseek_size(file, offset, whence, 492 maxbytes, i_size_read(inode)); 493 case SEEK_DATA: 494 case SEEK_HOLE: 495 if (offset < 0) 496 return -ENXIO; 497 return f2fs_seek_block(file, offset, whence); 498 } 499 500 return -EINVAL; 501 } 502 503 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 504 { 505 struct inode *inode = file_inode(file); 506 int err; 507 508 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 509 return -EIO; 510 511 if (!f2fs_is_compress_backend_ready(inode)) 512 return -EOPNOTSUPP; 513 514 /* we don't need to use inline_data strictly */ 515 err = f2fs_convert_inline_inode(inode); 516 if (err) 517 return err; 518 519 file_accessed(file); 520 vma->vm_ops = &f2fs_file_vm_ops; 521 set_inode_flag(inode, FI_MMAP_FILE); 522 return 0; 523 } 524 525 static int f2fs_file_open(struct inode *inode, struct file *filp) 526 { 527 int err = fscrypt_file_open(inode, filp); 528 529 if (err) 530 return err; 531 532 if (!f2fs_is_compress_backend_ready(inode)) 533 return -EOPNOTSUPP; 534 535 err = fsverity_file_open(inode, filp); 536 if (err) 537 return err; 538 539 filp->f_mode |= FMODE_NOWAIT; 540 541 return dquot_file_open(inode, filp); 542 } 543 544 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 545 { 546 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 547 struct f2fs_node *raw_node; 548 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 549 __le32 *addr; 550 int base = 0; 551 bool compressed_cluster = false; 552 int cluster_index = 0, valid_blocks = 0; 553 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 554 555 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode)) 556 base = get_extra_isize(dn->inode); 557 558 raw_node = F2FS_NODE(dn->node_page); 559 addr = blkaddr_in_node(raw_node) + base + ofs; 560 561 /* Assumption: truncateion starts with cluster */ 562 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 563 block_t blkaddr = le32_to_cpu(*addr); 564 565 if (f2fs_compressed_file(dn->inode) && 566 !(cluster_index & (cluster_size - 1))) { 567 if (compressed_cluster) 568 f2fs_i_compr_blocks_update(dn->inode, 569 valid_blocks, false); 570 compressed_cluster = (blkaddr == COMPRESS_ADDR); 571 valid_blocks = 0; 572 } 573 574 if (blkaddr == NULL_ADDR) 575 continue; 576 577 dn->data_blkaddr = NULL_ADDR; 578 f2fs_set_data_blkaddr(dn); 579 580 if (__is_valid_data_blkaddr(blkaddr)) { 581 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 582 DATA_GENERIC_ENHANCE)) 583 continue; 584 if (compressed_cluster) 585 valid_blocks++; 586 } 587 588 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 589 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 590 591 f2fs_invalidate_blocks(sbi, blkaddr); 592 nr_free++; 593 } 594 595 if (compressed_cluster) 596 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 597 598 if (nr_free) { 599 pgoff_t fofs; 600 /* 601 * once we invalidate valid blkaddr in range [ofs, ofs + count], 602 * we will invalidate all blkaddr in the whole range. 603 */ 604 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 605 dn->inode) + ofs; 606 f2fs_update_extent_cache_range(dn, fofs, 0, len); 607 dec_valid_block_count(sbi, dn->inode, nr_free); 608 } 609 dn->ofs_in_node = ofs; 610 611 f2fs_update_time(sbi, REQ_TIME); 612 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 613 dn->ofs_in_node, nr_free); 614 } 615 616 void f2fs_truncate_data_blocks(struct dnode_of_data *dn) 617 { 618 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode)); 619 } 620 621 static int truncate_partial_data_page(struct inode *inode, u64 from, 622 bool cache_only) 623 { 624 loff_t offset = from & (PAGE_SIZE - 1); 625 pgoff_t index = from >> PAGE_SHIFT; 626 struct address_space *mapping = inode->i_mapping; 627 struct page *page; 628 629 if (!offset && !cache_only) 630 return 0; 631 632 if (cache_only) { 633 page = find_lock_page(mapping, index); 634 if (page && PageUptodate(page)) 635 goto truncate_out; 636 f2fs_put_page(page, 1); 637 return 0; 638 } 639 640 if (f2fs_compressed_file(inode)) 641 return 0; 642 643 page = f2fs_get_lock_data_page(inode, index, true); 644 if (IS_ERR(page)) 645 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 646 truncate_out: 647 f2fs_wait_on_page_writeback(page, DATA, true, true); 648 zero_user(page, offset, PAGE_SIZE - offset); 649 650 /* An encrypted inode should have a key and truncate the last page. */ 651 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 652 if (!cache_only) 653 set_page_dirty(page); 654 f2fs_put_page(page, 1); 655 return 0; 656 } 657 658 static int do_truncate_blocks(struct inode *inode, u64 from, bool lock) 659 { 660 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 661 struct dnode_of_data dn; 662 pgoff_t free_from; 663 int count = 0, err = 0; 664 struct page *ipage; 665 bool truncate_page = false; 666 667 trace_f2fs_truncate_blocks_enter(inode, from); 668 669 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 670 671 if (free_from >= sbi->max_file_blocks) 672 goto free_partial; 673 674 if (lock) 675 f2fs_lock_op(sbi); 676 677 ipage = f2fs_get_node_page(sbi, inode->i_ino); 678 if (IS_ERR(ipage)) { 679 err = PTR_ERR(ipage); 680 goto out; 681 } 682 683 if (f2fs_has_inline_data(inode)) { 684 f2fs_truncate_inline_inode(inode, ipage, from); 685 f2fs_put_page(ipage, 1); 686 truncate_page = true; 687 goto out; 688 } 689 690 set_new_dnode(&dn, inode, ipage, NULL, 0); 691 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 692 if (err) { 693 if (err == -ENOENT) 694 goto free_next; 695 goto out; 696 } 697 698 count = ADDRS_PER_PAGE(dn.node_page, inode); 699 700 count -= dn.ofs_in_node; 701 f2fs_bug_on(sbi, count < 0); 702 703 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 704 f2fs_truncate_data_blocks_range(&dn, count); 705 free_from += count; 706 } 707 708 f2fs_put_dnode(&dn); 709 free_next: 710 err = f2fs_truncate_inode_blocks(inode, free_from); 711 out: 712 if (lock) 713 f2fs_unlock_op(sbi); 714 free_partial: 715 /* lastly zero out the first data page */ 716 if (!err) 717 err = truncate_partial_data_page(inode, from, truncate_page); 718 719 trace_f2fs_truncate_blocks_exit(inode, err); 720 return err; 721 } 722 723 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 724 { 725 u64 free_from = from; 726 727 /* 728 * for compressed file, only support cluster size 729 * aligned truncation. 730 */ 731 if (f2fs_compressed_file(inode)) { 732 size_t cluster_shift = PAGE_SHIFT + 733 F2FS_I(inode)->i_log_cluster_size; 734 size_t cluster_mask = (1 << cluster_shift) - 1; 735 736 free_from = from >> cluster_shift; 737 if (from & cluster_mask) 738 free_from++; 739 free_from <<= cluster_shift; 740 } 741 742 return do_truncate_blocks(inode, free_from, lock); 743 } 744 745 int f2fs_truncate(struct inode *inode) 746 { 747 int err; 748 749 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 750 return -EIO; 751 752 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 753 S_ISLNK(inode->i_mode))) 754 return 0; 755 756 trace_f2fs_truncate(inode); 757 758 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) { 759 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE); 760 return -EIO; 761 } 762 763 /* we should check inline_data size */ 764 if (!f2fs_may_inline_data(inode)) { 765 err = f2fs_convert_inline_inode(inode); 766 if (err) 767 return err; 768 } 769 770 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 771 if (err) 772 return err; 773 774 inode->i_mtime = inode->i_ctime = current_time(inode); 775 f2fs_mark_inode_dirty_sync(inode, false); 776 return 0; 777 } 778 779 int f2fs_getattr(const struct path *path, struct kstat *stat, 780 u32 request_mask, unsigned int query_flags) 781 { 782 struct inode *inode = d_inode(path->dentry); 783 struct f2fs_inode_info *fi = F2FS_I(inode); 784 struct f2fs_inode *ri; 785 unsigned int flags; 786 787 if (f2fs_has_extra_attr(inode) && 788 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 789 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 790 stat->result_mask |= STATX_BTIME; 791 stat->btime.tv_sec = fi->i_crtime.tv_sec; 792 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 793 } 794 795 flags = fi->i_flags; 796 if (flags & F2FS_APPEND_FL) 797 stat->attributes |= STATX_ATTR_APPEND; 798 if (IS_ENCRYPTED(inode)) 799 stat->attributes |= STATX_ATTR_ENCRYPTED; 800 if (flags & F2FS_IMMUTABLE_FL) 801 stat->attributes |= STATX_ATTR_IMMUTABLE; 802 if (flags & F2FS_NODUMP_FL) 803 stat->attributes |= STATX_ATTR_NODUMP; 804 if (IS_VERITY(inode)) 805 stat->attributes |= STATX_ATTR_VERITY; 806 807 stat->attributes_mask |= (STATX_ATTR_APPEND | 808 STATX_ATTR_ENCRYPTED | 809 STATX_ATTR_IMMUTABLE | 810 STATX_ATTR_NODUMP | 811 STATX_ATTR_VERITY); 812 813 generic_fillattr(inode, stat); 814 815 /* we need to show initial sectors used for inline_data/dentries */ 816 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 817 f2fs_has_inline_dentry(inode)) 818 stat->blocks += (stat->size + 511) >> 9; 819 820 return 0; 821 } 822 823 #ifdef CONFIG_F2FS_FS_POSIX_ACL 824 static void __setattr_copy(struct inode *inode, const struct iattr *attr) 825 { 826 unsigned int ia_valid = attr->ia_valid; 827 828 if (ia_valid & ATTR_UID) 829 inode->i_uid = attr->ia_uid; 830 if (ia_valid & ATTR_GID) 831 inode->i_gid = attr->ia_gid; 832 if (ia_valid & ATTR_ATIME) { 833 inode->i_atime = timestamp_truncate(attr->ia_atime, 834 inode); 835 } 836 if (ia_valid & ATTR_MTIME) { 837 inode->i_mtime = timestamp_truncate(attr->ia_mtime, 838 inode); 839 } 840 if (ia_valid & ATTR_CTIME) { 841 inode->i_ctime = timestamp_truncate(attr->ia_ctime, 842 inode); 843 } 844 if (ia_valid & ATTR_MODE) { 845 umode_t mode = attr->ia_mode; 846 847 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) 848 mode &= ~S_ISGID; 849 set_acl_inode(inode, mode); 850 } 851 } 852 #else 853 #define __setattr_copy setattr_copy 854 #endif 855 856 int f2fs_setattr(struct dentry *dentry, struct iattr *attr) 857 { 858 struct inode *inode = d_inode(dentry); 859 int err; 860 861 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 862 return -EIO; 863 864 if ((attr->ia_valid & ATTR_SIZE) && 865 !f2fs_is_compress_backend_ready(inode)) 866 return -EOPNOTSUPP; 867 868 err = setattr_prepare(dentry, attr); 869 if (err) 870 return err; 871 872 err = fscrypt_prepare_setattr(dentry, attr); 873 if (err) 874 return err; 875 876 err = fsverity_prepare_setattr(dentry, attr); 877 if (err) 878 return err; 879 880 if (is_quota_modification(inode, attr)) { 881 err = dquot_initialize(inode); 882 if (err) 883 return err; 884 } 885 if ((attr->ia_valid & ATTR_UID && 886 !uid_eq(attr->ia_uid, inode->i_uid)) || 887 (attr->ia_valid & ATTR_GID && 888 !gid_eq(attr->ia_gid, inode->i_gid))) { 889 f2fs_lock_op(F2FS_I_SB(inode)); 890 err = dquot_transfer(inode, attr); 891 if (err) { 892 set_sbi_flag(F2FS_I_SB(inode), 893 SBI_QUOTA_NEED_REPAIR); 894 f2fs_unlock_op(F2FS_I_SB(inode)); 895 return err; 896 } 897 /* 898 * update uid/gid under lock_op(), so that dquot and inode can 899 * be updated atomically. 900 */ 901 if (attr->ia_valid & ATTR_UID) 902 inode->i_uid = attr->ia_uid; 903 if (attr->ia_valid & ATTR_GID) 904 inode->i_gid = attr->ia_gid; 905 f2fs_mark_inode_dirty_sync(inode, true); 906 f2fs_unlock_op(F2FS_I_SB(inode)); 907 } 908 909 if (attr->ia_valid & ATTR_SIZE) { 910 loff_t old_size = i_size_read(inode); 911 912 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 913 /* 914 * should convert inline inode before i_size_write to 915 * keep smaller than inline_data size with inline flag. 916 */ 917 err = f2fs_convert_inline_inode(inode); 918 if (err) 919 return err; 920 } 921 922 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 923 down_write(&F2FS_I(inode)->i_mmap_sem); 924 925 truncate_setsize(inode, attr->ia_size); 926 927 if (attr->ia_size <= old_size) 928 err = f2fs_truncate(inode); 929 /* 930 * do not trim all blocks after i_size if target size is 931 * larger than i_size. 932 */ 933 up_write(&F2FS_I(inode)->i_mmap_sem); 934 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 935 if (err) 936 return err; 937 938 down_write(&F2FS_I(inode)->i_sem); 939 inode->i_mtime = inode->i_ctime = current_time(inode); 940 F2FS_I(inode)->last_disk_size = i_size_read(inode); 941 up_write(&F2FS_I(inode)->i_sem); 942 } 943 944 __setattr_copy(inode, attr); 945 946 if (attr->ia_valid & ATTR_MODE) { 947 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode)); 948 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) { 949 inode->i_mode = F2FS_I(inode)->i_acl_mode; 950 clear_inode_flag(inode, FI_ACL_MODE); 951 } 952 } 953 954 /* file size may changed here */ 955 f2fs_mark_inode_dirty_sync(inode, true); 956 957 /* inode change will produce dirty node pages flushed by checkpoint */ 958 f2fs_balance_fs(F2FS_I_SB(inode), true); 959 960 return err; 961 } 962 963 const struct inode_operations f2fs_file_inode_operations = { 964 .getattr = f2fs_getattr, 965 .setattr = f2fs_setattr, 966 .get_acl = f2fs_get_acl, 967 .set_acl = f2fs_set_acl, 968 #ifdef CONFIG_F2FS_FS_XATTR 969 .listxattr = f2fs_listxattr, 970 #endif 971 .fiemap = f2fs_fiemap, 972 }; 973 974 static int fill_zero(struct inode *inode, pgoff_t index, 975 loff_t start, loff_t len) 976 { 977 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 978 struct page *page; 979 980 if (!len) 981 return 0; 982 983 f2fs_balance_fs(sbi, true); 984 985 f2fs_lock_op(sbi); 986 page = f2fs_get_new_data_page(inode, NULL, index, false); 987 f2fs_unlock_op(sbi); 988 989 if (IS_ERR(page)) 990 return PTR_ERR(page); 991 992 f2fs_wait_on_page_writeback(page, DATA, true, true); 993 zero_user(page, start, len); 994 set_page_dirty(page); 995 f2fs_put_page(page, 1); 996 return 0; 997 } 998 999 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1000 { 1001 int err; 1002 1003 while (pg_start < pg_end) { 1004 struct dnode_of_data dn; 1005 pgoff_t end_offset, count; 1006 1007 set_new_dnode(&dn, inode, NULL, NULL, 0); 1008 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1009 if (err) { 1010 if (err == -ENOENT) { 1011 pg_start = f2fs_get_next_page_offset(&dn, 1012 pg_start); 1013 continue; 1014 } 1015 return err; 1016 } 1017 1018 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1019 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1020 1021 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1022 1023 f2fs_truncate_data_blocks_range(&dn, count); 1024 f2fs_put_dnode(&dn); 1025 1026 pg_start += count; 1027 } 1028 return 0; 1029 } 1030 1031 static int punch_hole(struct inode *inode, loff_t offset, loff_t len) 1032 { 1033 pgoff_t pg_start, pg_end; 1034 loff_t off_start, off_end; 1035 int ret; 1036 1037 ret = f2fs_convert_inline_inode(inode); 1038 if (ret) 1039 return ret; 1040 1041 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1042 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1043 1044 off_start = offset & (PAGE_SIZE - 1); 1045 off_end = (offset + len) & (PAGE_SIZE - 1); 1046 1047 if (pg_start == pg_end) { 1048 ret = fill_zero(inode, pg_start, off_start, 1049 off_end - off_start); 1050 if (ret) 1051 return ret; 1052 } else { 1053 if (off_start) { 1054 ret = fill_zero(inode, pg_start++, off_start, 1055 PAGE_SIZE - off_start); 1056 if (ret) 1057 return ret; 1058 } 1059 if (off_end) { 1060 ret = fill_zero(inode, pg_end, 0, off_end); 1061 if (ret) 1062 return ret; 1063 } 1064 1065 if (pg_start < pg_end) { 1066 struct address_space *mapping = inode->i_mapping; 1067 loff_t blk_start, blk_end; 1068 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1069 1070 f2fs_balance_fs(sbi, true); 1071 1072 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1073 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1074 1075 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1076 down_write(&F2FS_I(inode)->i_mmap_sem); 1077 1078 truncate_inode_pages_range(mapping, blk_start, 1079 blk_end - 1); 1080 1081 f2fs_lock_op(sbi); 1082 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1083 f2fs_unlock_op(sbi); 1084 1085 up_write(&F2FS_I(inode)->i_mmap_sem); 1086 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1087 } 1088 } 1089 1090 return ret; 1091 } 1092 1093 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1094 int *do_replace, pgoff_t off, pgoff_t len) 1095 { 1096 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1097 struct dnode_of_data dn; 1098 int ret, done, i; 1099 1100 next_dnode: 1101 set_new_dnode(&dn, inode, NULL, NULL, 0); 1102 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1103 if (ret && ret != -ENOENT) { 1104 return ret; 1105 } else if (ret == -ENOENT) { 1106 if (dn.max_level == 0) 1107 return -ENOENT; 1108 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1109 dn.ofs_in_node, len); 1110 blkaddr += done; 1111 do_replace += done; 1112 goto next; 1113 } 1114 1115 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1116 dn.ofs_in_node, len); 1117 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1118 *blkaddr = datablock_addr(dn.inode, 1119 dn.node_page, dn.ofs_in_node); 1120 1121 if (__is_valid_data_blkaddr(*blkaddr) && 1122 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1123 DATA_GENERIC_ENHANCE)) { 1124 f2fs_put_dnode(&dn); 1125 return -EFSCORRUPTED; 1126 } 1127 1128 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1129 1130 if (test_opt(sbi, LFS)) { 1131 f2fs_put_dnode(&dn); 1132 return -EOPNOTSUPP; 1133 } 1134 1135 /* do not invalidate this block address */ 1136 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1137 *do_replace = 1; 1138 } 1139 } 1140 f2fs_put_dnode(&dn); 1141 next: 1142 len -= done; 1143 off += done; 1144 if (len) 1145 goto next_dnode; 1146 return 0; 1147 } 1148 1149 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1150 int *do_replace, pgoff_t off, int len) 1151 { 1152 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1153 struct dnode_of_data dn; 1154 int ret, i; 1155 1156 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1157 if (*do_replace == 0) 1158 continue; 1159 1160 set_new_dnode(&dn, inode, NULL, NULL, 0); 1161 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1162 if (ret) { 1163 dec_valid_block_count(sbi, inode, 1); 1164 f2fs_invalidate_blocks(sbi, *blkaddr); 1165 } else { 1166 f2fs_update_data_blkaddr(&dn, *blkaddr); 1167 } 1168 f2fs_put_dnode(&dn); 1169 } 1170 return 0; 1171 } 1172 1173 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1174 block_t *blkaddr, int *do_replace, 1175 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1176 { 1177 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1178 pgoff_t i = 0; 1179 int ret; 1180 1181 while (i < len) { 1182 if (blkaddr[i] == NULL_ADDR && !full) { 1183 i++; 1184 continue; 1185 } 1186 1187 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1188 struct dnode_of_data dn; 1189 struct node_info ni; 1190 size_t new_size; 1191 pgoff_t ilen; 1192 1193 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1194 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1195 if (ret) 1196 return ret; 1197 1198 ret = f2fs_get_node_info(sbi, dn.nid, &ni); 1199 if (ret) { 1200 f2fs_put_dnode(&dn); 1201 return ret; 1202 } 1203 1204 ilen = min((pgoff_t) 1205 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1206 dn.ofs_in_node, len - i); 1207 do { 1208 dn.data_blkaddr = datablock_addr(dn.inode, 1209 dn.node_page, dn.ofs_in_node); 1210 f2fs_truncate_data_blocks_range(&dn, 1); 1211 1212 if (do_replace[i]) { 1213 f2fs_i_blocks_write(src_inode, 1214 1, false, false); 1215 f2fs_i_blocks_write(dst_inode, 1216 1, true, false); 1217 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1218 blkaddr[i], ni.version, true, false); 1219 1220 do_replace[i] = 0; 1221 } 1222 dn.ofs_in_node++; 1223 i++; 1224 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1225 if (dst_inode->i_size < new_size) 1226 f2fs_i_size_write(dst_inode, new_size); 1227 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1228 1229 f2fs_put_dnode(&dn); 1230 } else { 1231 struct page *psrc, *pdst; 1232 1233 psrc = f2fs_get_lock_data_page(src_inode, 1234 src + i, true); 1235 if (IS_ERR(psrc)) 1236 return PTR_ERR(psrc); 1237 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1238 true); 1239 if (IS_ERR(pdst)) { 1240 f2fs_put_page(psrc, 1); 1241 return PTR_ERR(pdst); 1242 } 1243 f2fs_copy_page(psrc, pdst); 1244 set_page_dirty(pdst); 1245 f2fs_put_page(pdst, 1); 1246 f2fs_put_page(psrc, 1); 1247 1248 ret = f2fs_truncate_hole(src_inode, 1249 src + i, src + i + 1); 1250 if (ret) 1251 return ret; 1252 i++; 1253 } 1254 } 1255 return 0; 1256 } 1257 1258 static int __exchange_data_block(struct inode *src_inode, 1259 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1260 pgoff_t len, bool full) 1261 { 1262 block_t *src_blkaddr; 1263 int *do_replace; 1264 pgoff_t olen; 1265 int ret; 1266 1267 while (len) { 1268 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1269 1270 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1271 array_size(olen, sizeof(block_t)), 1272 GFP_NOFS); 1273 if (!src_blkaddr) 1274 return -ENOMEM; 1275 1276 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1277 array_size(olen, sizeof(int)), 1278 GFP_NOFS); 1279 if (!do_replace) { 1280 kvfree(src_blkaddr); 1281 return -ENOMEM; 1282 } 1283 1284 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1285 do_replace, src, olen); 1286 if (ret) 1287 goto roll_back; 1288 1289 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1290 do_replace, src, dst, olen, full); 1291 if (ret) 1292 goto roll_back; 1293 1294 src += olen; 1295 dst += olen; 1296 len -= olen; 1297 1298 kvfree(src_blkaddr); 1299 kvfree(do_replace); 1300 } 1301 return 0; 1302 1303 roll_back: 1304 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1305 kvfree(src_blkaddr); 1306 kvfree(do_replace); 1307 return ret; 1308 } 1309 1310 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1311 { 1312 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1313 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1314 pgoff_t start = offset >> PAGE_SHIFT; 1315 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1316 int ret; 1317 1318 f2fs_balance_fs(sbi, true); 1319 1320 /* avoid gc operation during block exchange */ 1321 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1322 down_write(&F2FS_I(inode)->i_mmap_sem); 1323 1324 f2fs_lock_op(sbi); 1325 f2fs_drop_extent_tree(inode); 1326 truncate_pagecache(inode, offset); 1327 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1328 f2fs_unlock_op(sbi); 1329 1330 up_write(&F2FS_I(inode)->i_mmap_sem); 1331 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1332 return ret; 1333 } 1334 1335 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1336 { 1337 loff_t new_size; 1338 int ret; 1339 1340 if (offset + len >= i_size_read(inode)) 1341 return -EINVAL; 1342 1343 /* collapse range should be aligned to block size of f2fs. */ 1344 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1345 return -EINVAL; 1346 1347 ret = f2fs_convert_inline_inode(inode); 1348 if (ret) 1349 return ret; 1350 1351 /* write out all dirty pages from offset */ 1352 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1353 if (ret) 1354 return ret; 1355 1356 ret = f2fs_do_collapse(inode, offset, len); 1357 if (ret) 1358 return ret; 1359 1360 /* write out all moved pages, if possible */ 1361 down_write(&F2FS_I(inode)->i_mmap_sem); 1362 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1363 truncate_pagecache(inode, offset); 1364 1365 new_size = i_size_read(inode) - len; 1366 truncate_pagecache(inode, new_size); 1367 1368 ret = f2fs_truncate_blocks(inode, new_size, true); 1369 up_write(&F2FS_I(inode)->i_mmap_sem); 1370 if (!ret) 1371 f2fs_i_size_write(inode, new_size); 1372 return ret; 1373 } 1374 1375 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1376 pgoff_t end) 1377 { 1378 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1379 pgoff_t index = start; 1380 unsigned int ofs_in_node = dn->ofs_in_node; 1381 blkcnt_t count = 0; 1382 int ret; 1383 1384 for (; index < end; index++, dn->ofs_in_node++) { 1385 if (datablock_addr(dn->inode, dn->node_page, 1386 dn->ofs_in_node) == NULL_ADDR) 1387 count++; 1388 } 1389 1390 dn->ofs_in_node = ofs_in_node; 1391 ret = f2fs_reserve_new_blocks(dn, count); 1392 if (ret) 1393 return ret; 1394 1395 dn->ofs_in_node = ofs_in_node; 1396 for (index = start; index < end; index++, dn->ofs_in_node++) { 1397 dn->data_blkaddr = datablock_addr(dn->inode, 1398 dn->node_page, dn->ofs_in_node); 1399 /* 1400 * f2fs_reserve_new_blocks will not guarantee entire block 1401 * allocation. 1402 */ 1403 if (dn->data_blkaddr == NULL_ADDR) { 1404 ret = -ENOSPC; 1405 break; 1406 } 1407 if (dn->data_blkaddr != NEW_ADDR) { 1408 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1409 dn->data_blkaddr = NEW_ADDR; 1410 f2fs_set_data_blkaddr(dn); 1411 } 1412 } 1413 1414 f2fs_update_extent_cache_range(dn, start, 0, index - start); 1415 1416 return ret; 1417 } 1418 1419 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1420 int mode) 1421 { 1422 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1423 struct address_space *mapping = inode->i_mapping; 1424 pgoff_t index, pg_start, pg_end; 1425 loff_t new_size = i_size_read(inode); 1426 loff_t off_start, off_end; 1427 int ret = 0; 1428 1429 ret = inode_newsize_ok(inode, (len + offset)); 1430 if (ret) 1431 return ret; 1432 1433 ret = f2fs_convert_inline_inode(inode); 1434 if (ret) 1435 return ret; 1436 1437 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1438 if (ret) 1439 return ret; 1440 1441 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1442 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1443 1444 off_start = offset & (PAGE_SIZE - 1); 1445 off_end = (offset + len) & (PAGE_SIZE - 1); 1446 1447 if (pg_start == pg_end) { 1448 ret = fill_zero(inode, pg_start, off_start, 1449 off_end - off_start); 1450 if (ret) 1451 return ret; 1452 1453 new_size = max_t(loff_t, new_size, offset + len); 1454 } else { 1455 if (off_start) { 1456 ret = fill_zero(inode, pg_start++, off_start, 1457 PAGE_SIZE - off_start); 1458 if (ret) 1459 return ret; 1460 1461 new_size = max_t(loff_t, new_size, 1462 (loff_t)pg_start << PAGE_SHIFT); 1463 } 1464 1465 for (index = pg_start; index < pg_end;) { 1466 struct dnode_of_data dn; 1467 unsigned int end_offset; 1468 pgoff_t end; 1469 1470 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1471 down_write(&F2FS_I(inode)->i_mmap_sem); 1472 1473 truncate_pagecache_range(inode, 1474 (loff_t)index << PAGE_SHIFT, 1475 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1476 1477 f2fs_lock_op(sbi); 1478 1479 set_new_dnode(&dn, inode, NULL, NULL, 0); 1480 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1481 if (ret) { 1482 f2fs_unlock_op(sbi); 1483 up_write(&F2FS_I(inode)->i_mmap_sem); 1484 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1485 goto out; 1486 } 1487 1488 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1489 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1490 1491 ret = f2fs_do_zero_range(&dn, index, end); 1492 f2fs_put_dnode(&dn); 1493 1494 f2fs_unlock_op(sbi); 1495 up_write(&F2FS_I(inode)->i_mmap_sem); 1496 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1497 1498 f2fs_balance_fs(sbi, dn.node_changed); 1499 1500 if (ret) 1501 goto out; 1502 1503 index = end; 1504 new_size = max_t(loff_t, new_size, 1505 (loff_t)index << PAGE_SHIFT); 1506 } 1507 1508 if (off_end) { 1509 ret = fill_zero(inode, pg_end, 0, off_end); 1510 if (ret) 1511 goto out; 1512 1513 new_size = max_t(loff_t, new_size, offset + len); 1514 } 1515 } 1516 1517 out: 1518 if (new_size > i_size_read(inode)) { 1519 if (mode & FALLOC_FL_KEEP_SIZE) 1520 file_set_keep_isize(inode); 1521 else 1522 f2fs_i_size_write(inode, new_size); 1523 } 1524 return ret; 1525 } 1526 1527 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1528 { 1529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1530 pgoff_t nr, pg_start, pg_end, delta, idx; 1531 loff_t new_size; 1532 int ret = 0; 1533 1534 new_size = i_size_read(inode) + len; 1535 ret = inode_newsize_ok(inode, new_size); 1536 if (ret) 1537 return ret; 1538 1539 if (offset >= i_size_read(inode)) 1540 return -EINVAL; 1541 1542 /* insert range should be aligned to block size of f2fs. */ 1543 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1544 return -EINVAL; 1545 1546 ret = f2fs_convert_inline_inode(inode); 1547 if (ret) 1548 return ret; 1549 1550 f2fs_balance_fs(sbi, true); 1551 1552 down_write(&F2FS_I(inode)->i_mmap_sem); 1553 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1554 up_write(&F2FS_I(inode)->i_mmap_sem); 1555 if (ret) 1556 return ret; 1557 1558 /* write out all dirty pages from offset */ 1559 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1560 if (ret) 1561 return ret; 1562 1563 pg_start = offset >> PAGE_SHIFT; 1564 pg_end = (offset + len) >> PAGE_SHIFT; 1565 delta = pg_end - pg_start; 1566 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1567 1568 /* avoid gc operation during block exchange */ 1569 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1570 down_write(&F2FS_I(inode)->i_mmap_sem); 1571 truncate_pagecache(inode, offset); 1572 1573 while (!ret && idx > pg_start) { 1574 nr = idx - pg_start; 1575 if (nr > delta) 1576 nr = delta; 1577 idx -= nr; 1578 1579 f2fs_lock_op(sbi); 1580 f2fs_drop_extent_tree(inode); 1581 1582 ret = __exchange_data_block(inode, inode, idx, 1583 idx + delta, nr, false); 1584 f2fs_unlock_op(sbi); 1585 } 1586 up_write(&F2FS_I(inode)->i_mmap_sem); 1587 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1588 1589 /* write out all moved pages, if possible */ 1590 down_write(&F2FS_I(inode)->i_mmap_sem); 1591 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1592 truncate_pagecache(inode, offset); 1593 up_write(&F2FS_I(inode)->i_mmap_sem); 1594 1595 if (!ret) 1596 f2fs_i_size_write(inode, new_size); 1597 return ret; 1598 } 1599 1600 static int expand_inode_data(struct inode *inode, loff_t offset, 1601 loff_t len, int mode) 1602 { 1603 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1604 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1605 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1606 .m_may_create = true }; 1607 pgoff_t pg_end; 1608 loff_t new_size = i_size_read(inode); 1609 loff_t off_end; 1610 int err; 1611 1612 err = inode_newsize_ok(inode, (len + offset)); 1613 if (err) 1614 return err; 1615 1616 err = f2fs_convert_inline_inode(inode); 1617 if (err) 1618 return err; 1619 1620 f2fs_balance_fs(sbi, true); 1621 1622 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1623 off_end = (offset + len) & (PAGE_SIZE - 1); 1624 1625 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT; 1626 map.m_len = pg_end - map.m_lblk; 1627 if (off_end) 1628 map.m_len++; 1629 1630 if (!map.m_len) 1631 return 0; 1632 1633 if (f2fs_is_pinned_file(inode)) { 1634 block_t len = (map.m_len >> sbi->log_blocks_per_seg) << 1635 sbi->log_blocks_per_seg; 1636 block_t done = 0; 1637 1638 if (map.m_len % sbi->blocks_per_seg) 1639 len += sbi->blocks_per_seg; 1640 1641 map.m_len = sbi->blocks_per_seg; 1642 next_alloc: 1643 if (has_not_enough_free_secs(sbi, 0, 1644 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1645 down_write(&sbi->gc_lock); 1646 err = f2fs_gc(sbi, true, false, NULL_SEGNO); 1647 if (err && err != -ENODATA && err != -EAGAIN) 1648 goto out_err; 1649 } 1650 1651 down_write(&sbi->pin_sem); 1652 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1653 f2fs_allocate_new_segments(sbi, CURSEG_COLD_DATA); 1654 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO); 1655 up_write(&sbi->pin_sem); 1656 1657 done += map.m_len; 1658 len -= map.m_len; 1659 map.m_lblk += map.m_len; 1660 if (!err && len) 1661 goto next_alloc; 1662 1663 map.m_len = done; 1664 } else { 1665 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); 1666 } 1667 out_err: 1668 if (err) { 1669 pgoff_t last_off; 1670 1671 if (!map.m_len) 1672 return err; 1673 1674 last_off = map.m_lblk + map.m_len - 1; 1675 1676 /* update new size to the failed position */ 1677 new_size = (last_off == pg_end) ? offset + len : 1678 (loff_t)(last_off + 1) << PAGE_SHIFT; 1679 } else { 1680 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1681 } 1682 1683 if (new_size > i_size_read(inode)) { 1684 if (mode & FALLOC_FL_KEEP_SIZE) 1685 file_set_keep_isize(inode); 1686 else 1687 f2fs_i_size_write(inode, new_size); 1688 } 1689 1690 return err; 1691 } 1692 1693 static long f2fs_fallocate(struct file *file, int mode, 1694 loff_t offset, loff_t len) 1695 { 1696 struct inode *inode = file_inode(file); 1697 long ret = 0; 1698 1699 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1700 return -EIO; 1701 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1702 return -ENOSPC; 1703 if (!f2fs_is_compress_backend_ready(inode)) 1704 return -EOPNOTSUPP; 1705 1706 /* f2fs only support ->fallocate for regular file */ 1707 if (!S_ISREG(inode->i_mode)) 1708 return -EINVAL; 1709 1710 if (IS_ENCRYPTED(inode) && 1711 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1712 return -EOPNOTSUPP; 1713 1714 if (f2fs_compressed_file(inode) && 1715 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1716 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) 1717 return -EOPNOTSUPP; 1718 1719 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1720 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1721 FALLOC_FL_INSERT_RANGE)) 1722 return -EOPNOTSUPP; 1723 1724 inode_lock(inode); 1725 1726 if (mode & FALLOC_FL_PUNCH_HOLE) { 1727 if (offset >= inode->i_size) 1728 goto out; 1729 1730 ret = punch_hole(inode, offset, len); 1731 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1732 ret = f2fs_collapse_range(inode, offset, len); 1733 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1734 ret = f2fs_zero_range(inode, offset, len, mode); 1735 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1736 ret = f2fs_insert_range(inode, offset, len); 1737 } else { 1738 ret = expand_inode_data(inode, offset, len, mode); 1739 } 1740 1741 if (!ret) { 1742 inode->i_mtime = inode->i_ctime = current_time(inode); 1743 f2fs_mark_inode_dirty_sync(inode, false); 1744 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1745 } 1746 1747 out: 1748 inode_unlock(inode); 1749 1750 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1751 return ret; 1752 } 1753 1754 static int f2fs_release_file(struct inode *inode, struct file *filp) 1755 { 1756 /* 1757 * f2fs_relase_file is called at every close calls. So we should 1758 * not drop any inmemory pages by close called by other process. 1759 */ 1760 if (!(filp->f_mode & FMODE_WRITE) || 1761 atomic_read(&inode->i_writecount) != 1) 1762 return 0; 1763 1764 /* some remained atomic pages should discarded */ 1765 if (f2fs_is_atomic_file(inode)) 1766 f2fs_drop_inmem_pages(inode); 1767 if (f2fs_is_volatile_file(inode)) { 1768 set_inode_flag(inode, FI_DROP_CACHE); 1769 filemap_fdatawrite(inode->i_mapping); 1770 clear_inode_flag(inode, FI_DROP_CACHE); 1771 clear_inode_flag(inode, FI_VOLATILE_FILE); 1772 stat_dec_volatile_write(inode); 1773 } 1774 return 0; 1775 } 1776 1777 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1778 { 1779 struct inode *inode = file_inode(file); 1780 1781 /* 1782 * If the process doing a transaction is crashed, we should do 1783 * roll-back. Otherwise, other reader/write can see corrupted database 1784 * until all the writers close its file. Since this should be done 1785 * before dropping file lock, it needs to do in ->flush. 1786 */ 1787 if (f2fs_is_atomic_file(inode) && 1788 F2FS_I(inode)->inmem_task == current) 1789 f2fs_drop_inmem_pages(inode); 1790 return 0; 1791 } 1792 1793 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1794 { 1795 struct f2fs_inode_info *fi = F2FS_I(inode); 1796 1797 /* Is it quota file? Do not allow user to mess with it */ 1798 if (IS_NOQUOTA(inode)) 1799 return -EPERM; 1800 1801 if ((iflags ^ fi->i_flags) & F2FS_CASEFOLD_FL) { 1802 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1803 return -EOPNOTSUPP; 1804 if (!f2fs_empty_dir(inode)) 1805 return -ENOTEMPTY; 1806 } 1807 1808 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1809 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1810 return -EOPNOTSUPP; 1811 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1812 return -EINVAL; 1813 } 1814 1815 if ((iflags ^ fi->i_flags) & F2FS_COMPR_FL) { 1816 if (S_ISREG(inode->i_mode) && 1817 (fi->i_flags & F2FS_COMPR_FL || i_size_read(inode) || 1818 F2FS_HAS_BLOCKS(inode))) 1819 return -EINVAL; 1820 if (iflags & F2FS_NOCOMP_FL) 1821 return -EINVAL; 1822 if (iflags & F2FS_COMPR_FL) { 1823 int err = f2fs_convert_inline_inode(inode); 1824 1825 if (err) 1826 return err; 1827 1828 if (!f2fs_may_compress(inode)) 1829 return -EINVAL; 1830 1831 set_compress_context(inode); 1832 } 1833 } 1834 if ((iflags ^ fi->i_flags) & F2FS_NOCOMP_FL) { 1835 if (fi->i_flags & F2FS_COMPR_FL) 1836 return -EINVAL; 1837 } 1838 1839 fi->i_flags = iflags | (fi->i_flags & ~mask); 1840 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1841 (fi->i_flags & F2FS_NOCOMP_FL)); 1842 1843 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1844 set_inode_flag(inode, FI_PROJ_INHERIT); 1845 else 1846 clear_inode_flag(inode, FI_PROJ_INHERIT); 1847 1848 inode->i_ctime = current_time(inode); 1849 f2fs_set_inode_flags(inode); 1850 f2fs_mark_inode_dirty_sync(inode, true); 1851 return 0; 1852 } 1853 1854 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */ 1855 1856 /* 1857 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1858 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1859 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1860 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1861 */ 1862 1863 static const struct { 1864 u32 iflag; 1865 u32 fsflag; 1866 } f2fs_fsflags_map[] = { 1867 { F2FS_COMPR_FL, FS_COMPR_FL }, 1868 { F2FS_SYNC_FL, FS_SYNC_FL }, 1869 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1870 { F2FS_APPEND_FL, FS_APPEND_FL }, 1871 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 1872 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 1873 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 1874 { F2FS_INDEX_FL, FS_INDEX_FL }, 1875 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 1876 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 1877 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 1878 }; 1879 1880 #define F2FS_GETTABLE_FS_FL ( \ 1881 FS_COMPR_FL | \ 1882 FS_SYNC_FL | \ 1883 FS_IMMUTABLE_FL | \ 1884 FS_APPEND_FL | \ 1885 FS_NODUMP_FL | \ 1886 FS_NOATIME_FL | \ 1887 FS_NOCOMP_FL | \ 1888 FS_INDEX_FL | \ 1889 FS_DIRSYNC_FL | \ 1890 FS_PROJINHERIT_FL | \ 1891 FS_ENCRYPT_FL | \ 1892 FS_INLINE_DATA_FL | \ 1893 FS_NOCOW_FL | \ 1894 FS_VERITY_FL | \ 1895 FS_CASEFOLD_FL) 1896 1897 #define F2FS_SETTABLE_FS_FL ( \ 1898 FS_COMPR_FL | \ 1899 FS_SYNC_FL | \ 1900 FS_IMMUTABLE_FL | \ 1901 FS_APPEND_FL | \ 1902 FS_NODUMP_FL | \ 1903 FS_NOATIME_FL | \ 1904 FS_NOCOMP_FL | \ 1905 FS_DIRSYNC_FL | \ 1906 FS_PROJINHERIT_FL | \ 1907 FS_CASEFOLD_FL) 1908 1909 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 1910 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 1911 { 1912 u32 fsflags = 0; 1913 int i; 1914 1915 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1916 if (iflags & f2fs_fsflags_map[i].iflag) 1917 fsflags |= f2fs_fsflags_map[i].fsflag; 1918 1919 return fsflags; 1920 } 1921 1922 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 1923 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 1924 { 1925 u32 iflags = 0; 1926 int i; 1927 1928 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 1929 if (fsflags & f2fs_fsflags_map[i].fsflag) 1930 iflags |= f2fs_fsflags_map[i].iflag; 1931 1932 return iflags; 1933 } 1934 1935 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg) 1936 { 1937 struct inode *inode = file_inode(filp); 1938 struct f2fs_inode_info *fi = F2FS_I(inode); 1939 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 1940 1941 if (IS_ENCRYPTED(inode)) 1942 fsflags |= FS_ENCRYPT_FL; 1943 if (IS_VERITY(inode)) 1944 fsflags |= FS_VERITY_FL; 1945 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 1946 fsflags |= FS_INLINE_DATA_FL; 1947 if (is_inode_flag_set(inode, FI_PIN_FILE)) 1948 fsflags |= FS_NOCOW_FL; 1949 1950 fsflags &= F2FS_GETTABLE_FS_FL; 1951 1952 return put_user(fsflags, (int __user *)arg); 1953 } 1954 1955 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg) 1956 { 1957 struct inode *inode = file_inode(filp); 1958 struct f2fs_inode_info *fi = F2FS_I(inode); 1959 u32 fsflags, old_fsflags; 1960 u32 iflags; 1961 int ret; 1962 1963 if (!inode_owner_or_capable(inode)) 1964 return -EACCES; 1965 1966 if (get_user(fsflags, (int __user *)arg)) 1967 return -EFAULT; 1968 1969 if (fsflags & ~F2FS_GETTABLE_FS_FL) 1970 return -EOPNOTSUPP; 1971 fsflags &= F2FS_SETTABLE_FS_FL; 1972 1973 iflags = f2fs_fsflags_to_iflags(fsflags); 1974 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 1975 return -EOPNOTSUPP; 1976 1977 ret = mnt_want_write_file(filp); 1978 if (ret) 1979 return ret; 1980 1981 inode_lock(inode); 1982 1983 old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 1984 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags); 1985 if (ret) 1986 goto out; 1987 1988 ret = f2fs_setflags_common(inode, iflags, 1989 f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL)); 1990 out: 1991 inode_unlock(inode); 1992 mnt_drop_write_file(filp); 1993 return ret; 1994 } 1995 1996 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1997 { 1998 struct inode *inode = file_inode(filp); 1999 2000 return put_user(inode->i_generation, (int __user *)arg); 2001 } 2002 2003 static int f2fs_ioc_start_atomic_write(struct file *filp) 2004 { 2005 struct inode *inode = file_inode(filp); 2006 struct f2fs_inode_info *fi = F2FS_I(inode); 2007 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2008 int ret; 2009 2010 if (!inode_owner_or_capable(inode)) 2011 return -EACCES; 2012 2013 if (!S_ISREG(inode->i_mode)) 2014 return -EINVAL; 2015 2016 if (filp->f_flags & O_DIRECT) 2017 return -EINVAL; 2018 2019 ret = mnt_want_write_file(filp); 2020 if (ret) 2021 return ret; 2022 2023 inode_lock(inode); 2024 2025 f2fs_disable_compressed_file(inode); 2026 2027 if (f2fs_is_atomic_file(inode)) { 2028 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) 2029 ret = -EINVAL; 2030 goto out; 2031 } 2032 2033 ret = f2fs_convert_inline_inode(inode); 2034 if (ret) 2035 goto out; 2036 2037 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2038 2039 /* 2040 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2041 * f2fs_is_atomic_file. 2042 */ 2043 if (get_dirty_pages(inode)) 2044 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2045 inode->i_ino, get_dirty_pages(inode)); 2046 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2047 if (ret) { 2048 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2049 goto out; 2050 } 2051 2052 spin_lock(&sbi->inode_lock[ATOMIC_FILE]); 2053 if (list_empty(&fi->inmem_ilist)) 2054 list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]); 2055 sbi->atomic_files++; 2056 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]); 2057 2058 /* add inode in inmem_list first and set atomic_file */ 2059 set_inode_flag(inode, FI_ATOMIC_FILE); 2060 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2061 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 2062 2063 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2064 F2FS_I(inode)->inmem_task = current; 2065 stat_update_max_atomic_write(inode); 2066 out: 2067 inode_unlock(inode); 2068 mnt_drop_write_file(filp); 2069 return ret; 2070 } 2071 2072 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2073 { 2074 struct inode *inode = file_inode(filp); 2075 int ret; 2076 2077 if (!inode_owner_or_capable(inode)) 2078 return -EACCES; 2079 2080 ret = mnt_want_write_file(filp); 2081 if (ret) 2082 return ret; 2083 2084 f2fs_balance_fs(F2FS_I_SB(inode), true); 2085 2086 inode_lock(inode); 2087 2088 if (f2fs_is_volatile_file(inode)) { 2089 ret = -EINVAL; 2090 goto err_out; 2091 } 2092 2093 if (f2fs_is_atomic_file(inode)) { 2094 ret = f2fs_commit_inmem_pages(inode); 2095 if (ret) 2096 goto err_out; 2097 2098 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2099 if (!ret) 2100 f2fs_drop_inmem_pages(inode); 2101 } else { 2102 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2103 } 2104 err_out: 2105 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) { 2106 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2107 ret = -EINVAL; 2108 } 2109 inode_unlock(inode); 2110 mnt_drop_write_file(filp); 2111 return ret; 2112 } 2113 2114 static int f2fs_ioc_start_volatile_write(struct file *filp) 2115 { 2116 struct inode *inode = file_inode(filp); 2117 int ret; 2118 2119 if (!inode_owner_or_capable(inode)) 2120 return -EACCES; 2121 2122 if (!S_ISREG(inode->i_mode)) 2123 return -EINVAL; 2124 2125 ret = mnt_want_write_file(filp); 2126 if (ret) 2127 return ret; 2128 2129 inode_lock(inode); 2130 2131 if (f2fs_is_volatile_file(inode)) 2132 goto out; 2133 2134 ret = f2fs_convert_inline_inode(inode); 2135 if (ret) 2136 goto out; 2137 2138 stat_inc_volatile_write(inode); 2139 stat_update_max_volatile_write(inode); 2140 2141 set_inode_flag(inode, FI_VOLATILE_FILE); 2142 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2143 out: 2144 inode_unlock(inode); 2145 mnt_drop_write_file(filp); 2146 return ret; 2147 } 2148 2149 static int f2fs_ioc_release_volatile_write(struct file *filp) 2150 { 2151 struct inode *inode = file_inode(filp); 2152 int ret; 2153 2154 if (!inode_owner_or_capable(inode)) 2155 return -EACCES; 2156 2157 ret = mnt_want_write_file(filp); 2158 if (ret) 2159 return ret; 2160 2161 inode_lock(inode); 2162 2163 if (!f2fs_is_volatile_file(inode)) 2164 goto out; 2165 2166 if (!f2fs_is_first_block_written(inode)) { 2167 ret = truncate_partial_data_page(inode, 0, true); 2168 goto out; 2169 } 2170 2171 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 2172 out: 2173 inode_unlock(inode); 2174 mnt_drop_write_file(filp); 2175 return ret; 2176 } 2177 2178 static int f2fs_ioc_abort_volatile_write(struct file *filp) 2179 { 2180 struct inode *inode = file_inode(filp); 2181 int ret; 2182 2183 if (!inode_owner_or_capable(inode)) 2184 return -EACCES; 2185 2186 ret = mnt_want_write_file(filp); 2187 if (ret) 2188 return ret; 2189 2190 inode_lock(inode); 2191 2192 if (f2fs_is_atomic_file(inode)) 2193 f2fs_drop_inmem_pages(inode); 2194 if (f2fs_is_volatile_file(inode)) { 2195 clear_inode_flag(inode, FI_VOLATILE_FILE); 2196 stat_dec_volatile_write(inode); 2197 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2198 } 2199 2200 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 2201 2202 inode_unlock(inode); 2203 2204 mnt_drop_write_file(filp); 2205 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2206 return ret; 2207 } 2208 2209 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2210 { 2211 struct inode *inode = file_inode(filp); 2212 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2213 struct super_block *sb = sbi->sb; 2214 __u32 in; 2215 int ret = 0; 2216 2217 if (!capable(CAP_SYS_ADMIN)) 2218 return -EPERM; 2219 2220 if (get_user(in, (__u32 __user *)arg)) 2221 return -EFAULT; 2222 2223 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2224 ret = mnt_want_write_file(filp); 2225 if (ret) 2226 return ret; 2227 } 2228 2229 switch (in) { 2230 case F2FS_GOING_DOWN_FULLSYNC: 2231 sb = freeze_bdev(sb->s_bdev); 2232 if (IS_ERR(sb)) { 2233 ret = PTR_ERR(sb); 2234 goto out; 2235 } 2236 if (sb) { 2237 f2fs_stop_checkpoint(sbi, false); 2238 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2239 thaw_bdev(sb->s_bdev, sb); 2240 } 2241 break; 2242 case F2FS_GOING_DOWN_METASYNC: 2243 /* do checkpoint only */ 2244 ret = f2fs_sync_fs(sb, 1); 2245 if (ret) 2246 goto out; 2247 f2fs_stop_checkpoint(sbi, false); 2248 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2249 break; 2250 case F2FS_GOING_DOWN_NOSYNC: 2251 f2fs_stop_checkpoint(sbi, false); 2252 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2253 break; 2254 case F2FS_GOING_DOWN_METAFLUSH: 2255 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2256 f2fs_stop_checkpoint(sbi, false); 2257 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 2258 break; 2259 case F2FS_GOING_DOWN_NEED_FSCK: 2260 set_sbi_flag(sbi, SBI_NEED_FSCK); 2261 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2262 set_sbi_flag(sbi, SBI_IS_DIRTY); 2263 /* do checkpoint only */ 2264 ret = f2fs_sync_fs(sb, 1); 2265 goto out; 2266 default: 2267 ret = -EINVAL; 2268 goto out; 2269 } 2270 2271 f2fs_stop_gc_thread(sbi); 2272 f2fs_stop_discard_thread(sbi); 2273 2274 f2fs_drop_discard_cmd(sbi); 2275 clear_opt(sbi, DISCARD); 2276 2277 f2fs_update_time(sbi, REQ_TIME); 2278 out: 2279 if (in != F2FS_GOING_DOWN_FULLSYNC) 2280 mnt_drop_write_file(filp); 2281 2282 trace_f2fs_shutdown(sbi, in, ret); 2283 2284 return ret; 2285 } 2286 2287 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2288 { 2289 struct inode *inode = file_inode(filp); 2290 struct super_block *sb = inode->i_sb; 2291 struct request_queue *q = bdev_get_queue(sb->s_bdev); 2292 struct fstrim_range range; 2293 int ret; 2294 2295 if (!capable(CAP_SYS_ADMIN)) 2296 return -EPERM; 2297 2298 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2299 return -EOPNOTSUPP; 2300 2301 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2302 sizeof(range))) 2303 return -EFAULT; 2304 2305 ret = mnt_want_write_file(filp); 2306 if (ret) 2307 return ret; 2308 2309 range.minlen = max((unsigned int)range.minlen, 2310 q->limits.discard_granularity); 2311 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2312 mnt_drop_write_file(filp); 2313 if (ret < 0) 2314 return ret; 2315 2316 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2317 sizeof(range))) 2318 return -EFAULT; 2319 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2320 return 0; 2321 } 2322 2323 static bool uuid_is_nonzero(__u8 u[16]) 2324 { 2325 int i; 2326 2327 for (i = 0; i < 16; i++) 2328 if (u[i]) 2329 return true; 2330 return false; 2331 } 2332 2333 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2334 { 2335 struct inode *inode = file_inode(filp); 2336 2337 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2338 return -EOPNOTSUPP; 2339 2340 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2341 2342 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2343 } 2344 2345 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2346 { 2347 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2348 return -EOPNOTSUPP; 2349 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2350 } 2351 2352 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2353 { 2354 struct inode *inode = file_inode(filp); 2355 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2356 int err; 2357 2358 if (!f2fs_sb_has_encrypt(sbi)) 2359 return -EOPNOTSUPP; 2360 2361 err = mnt_want_write_file(filp); 2362 if (err) 2363 return err; 2364 2365 down_write(&sbi->sb_lock); 2366 2367 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2368 goto got_it; 2369 2370 /* update superblock with uuid */ 2371 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2372 2373 err = f2fs_commit_super(sbi, false); 2374 if (err) { 2375 /* undo new data */ 2376 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2377 goto out_err; 2378 } 2379 got_it: 2380 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 2381 16)) 2382 err = -EFAULT; 2383 out_err: 2384 up_write(&sbi->sb_lock); 2385 mnt_drop_write_file(filp); 2386 return err; 2387 } 2388 2389 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2390 unsigned long arg) 2391 { 2392 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2393 return -EOPNOTSUPP; 2394 2395 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2396 } 2397 2398 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2399 { 2400 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2401 return -EOPNOTSUPP; 2402 2403 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2404 } 2405 2406 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2407 { 2408 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2409 return -EOPNOTSUPP; 2410 2411 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2412 } 2413 2414 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2415 unsigned long arg) 2416 { 2417 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2418 return -EOPNOTSUPP; 2419 2420 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2421 } 2422 2423 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2424 unsigned long arg) 2425 { 2426 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2427 return -EOPNOTSUPP; 2428 2429 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2430 } 2431 2432 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2433 { 2434 struct inode *inode = file_inode(filp); 2435 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2436 __u32 sync; 2437 int ret; 2438 2439 if (!capable(CAP_SYS_ADMIN)) 2440 return -EPERM; 2441 2442 if (get_user(sync, (__u32 __user *)arg)) 2443 return -EFAULT; 2444 2445 if (f2fs_readonly(sbi->sb)) 2446 return -EROFS; 2447 2448 ret = mnt_want_write_file(filp); 2449 if (ret) 2450 return ret; 2451 2452 if (!sync) { 2453 if (!down_write_trylock(&sbi->gc_lock)) { 2454 ret = -EBUSY; 2455 goto out; 2456 } 2457 } else { 2458 down_write(&sbi->gc_lock); 2459 } 2460 2461 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO); 2462 out: 2463 mnt_drop_write_file(filp); 2464 return ret; 2465 } 2466 2467 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2468 { 2469 struct inode *inode = file_inode(filp); 2470 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2471 struct f2fs_gc_range range; 2472 u64 end; 2473 int ret; 2474 2475 if (!capable(CAP_SYS_ADMIN)) 2476 return -EPERM; 2477 2478 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2479 sizeof(range))) 2480 return -EFAULT; 2481 2482 if (f2fs_readonly(sbi->sb)) 2483 return -EROFS; 2484 2485 end = range.start + range.len; 2486 if (end < range.start || range.start < MAIN_BLKADDR(sbi) || 2487 end >= MAX_BLKADDR(sbi)) 2488 return -EINVAL; 2489 2490 ret = mnt_want_write_file(filp); 2491 if (ret) 2492 return ret; 2493 2494 do_more: 2495 if (!range.sync) { 2496 if (!down_write_trylock(&sbi->gc_lock)) { 2497 ret = -EBUSY; 2498 goto out; 2499 } 2500 } else { 2501 down_write(&sbi->gc_lock); 2502 } 2503 2504 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start)); 2505 range.start += BLKS_PER_SEC(sbi); 2506 if (range.start <= end) 2507 goto do_more; 2508 out: 2509 mnt_drop_write_file(filp); 2510 return ret; 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 sbi->max_file_blocks)) 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 (src == dst) { 2758 if (pos_in == pos_out) 2759 return 0; 2760 if (pos_out > pos_in && pos_out < pos_in + len) 2761 return -EINVAL; 2762 } 2763 2764 inode_lock(src); 2765 if (src != dst) { 2766 ret = -EBUSY; 2767 if (!inode_trylock(dst)) 2768 goto out; 2769 } 2770 2771 ret = -EINVAL; 2772 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2773 goto out_unlock; 2774 if (len == 0) 2775 olen = len = src->i_size - pos_in; 2776 if (pos_in + len == src->i_size) 2777 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2778 if (len == 0) { 2779 ret = 0; 2780 goto out_unlock; 2781 } 2782 2783 dst_osize = dst->i_size; 2784 if (pos_out + olen > dst->i_size) 2785 dst_max_i_size = pos_out + olen; 2786 2787 /* verify the end result is block aligned */ 2788 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2789 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2790 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2791 goto out_unlock; 2792 2793 ret = f2fs_convert_inline_inode(src); 2794 if (ret) 2795 goto out_unlock; 2796 2797 ret = f2fs_convert_inline_inode(dst); 2798 if (ret) 2799 goto out_unlock; 2800 2801 /* write out all dirty pages from offset */ 2802 ret = filemap_write_and_wait_range(src->i_mapping, 2803 pos_in, pos_in + len); 2804 if (ret) 2805 goto out_unlock; 2806 2807 ret = filemap_write_and_wait_range(dst->i_mapping, 2808 pos_out, pos_out + len); 2809 if (ret) 2810 goto out_unlock; 2811 2812 f2fs_balance_fs(sbi, true); 2813 2814 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2815 if (src != dst) { 2816 ret = -EBUSY; 2817 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2818 goto out_src; 2819 } 2820 2821 f2fs_lock_op(sbi); 2822 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2823 pos_out >> F2FS_BLKSIZE_BITS, 2824 len >> F2FS_BLKSIZE_BITS, false); 2825 2826 if (!ret) { 2827 if (dst_max_i_size) 2828 f2fs_i_size_write(dst, dst_max_i_size); 2829 else if (dst_osize != dst->i_size) 2830 f2fs_i_size_write(dst, dst_osize); 2831 } 2832 f2fs_unlock_op(sbi); 2833 2834 if (src != dst) 2835 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2836 out_src: 2837 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2838 out_unlock: 2839 if (src != dst) 2840 inode_unlock(dst); 2841 out: 2842 inode_unlock(src); 2843 return ret; 2844 } 2845 2846 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2847 { 2848 struct f2fs_move_range range; 2849 struct fd dst; 2850 int err; 2851 2852 if (!(filp->f_mode & FMODE_READ) || 2853 !(filp->f_mode & FMODE_WRITE)) 2854 return -EBADF; 2855 2856 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2857 sizeof(range))) 2858 return -EFAULT; 2859 2860 dst = fdget(range.dst_fd); 2861 if (!dst.file) 2862 return -EBADF; 2863 2864 if (!(dst.file->f_mode & FMODE_WRITE)) { 2865 err = -EBADF; 2866 goto err_out; 2867 } 2868 2869 err = mnt_want_write_file(filp); 2870 if (err) 2871 goto err_out; 2872 2873 err = f2fs_move_file_range(filp, range.pos_in, dst.file, 2874 range.pos_out, range.len); 2875 2876 mnt_drop_write_file(filp); 2877 if (err) 2878 goto err_out; 2879 2880 if (copy_to_user((struct f2fs_move_range __user *)arg, 2881 &range, sizeof(range))) 2882 err = -EFAULT; 2883 err_out: 2884 fdput(dst); 2885 return err; 2886 } 2887 2888 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2889 { 2890 struct inode *inode = file_inode(filp); 2891 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2892 struct sit_info *sm = SIT_I(sbi); 2893 unsigned int start_segno = 0, end_segno = 0; 2894 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2895 struct f2fs_flush_device range; 2896 int ret; 2897 2898 if (!capable(CAP_SYS_ADMIN)) 2899 return -EPERM; 2900 2901 if (f2fs_readonly(sbi->sb)) 2902 return -EROFS; 2903 2904 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2905 return -EINVAL; 2906 2907 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2908 sizeof(range))) 2909 return -EFAULT; 2910 2911 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 2912 __is_large_section(sbi)) { 2913 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1", 2914 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec); 2915 return -EINVAL; 2916 } 2917 2918 ret = mnt_want_write_file(filp); 2919 if (ret) 2920 return ret; 2921 2922 if (range.dev_num != 0) 2923 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2924 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2925 2926 start_segno = sm->last_victim[FLUSH_DEVICE]; 2927 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2928 start_segno = dev_start_segno; 2929 end_segno = min(start_segno + range.segments, dev_end_segno); 2930 2931 while (start_segno < end_segno) { 2932 if (!down_write_trylock(&sbi->gc_lock)) { 2933 ret = -EBUSY; 2934 goto out; 2935 } 2936 sm->last_victim[GC_CB] = end_segno + 1; 2937 sm->last_victim[GC_GREEDY] = end_segno + 1; 2938 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 2939 ret = f2fs_gc(sbi, true, true, start_segno); 2940 if (ret == -EAGAIN) 2941 ret = 0; 2942 else if (ret < 0) 2943 break; 2944 start_segno++; 2945 } 2946 out: 2947 mnt_drop_write_file(filp); 2948 return ret; 2949 } 2950 2951 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 2952 { 2953 struct inode *inode = file_inode(filp); 2954 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 2955 2956 /* Must validate to set it with SQLite behavior in Android. */ 2957 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 2958 2959 return put_user(sb_feature, (u32 __user *)arg); 2960 } 2961 2962 #ifdef CONFIG_QUOTA 2963 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 2964 { 2965 struct dquot *transfer_to[MAXQUOTAS] = {}; 2966 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2967 struct super_block *sb = sbi->sb; 2968 int err = 0; 2969 2970 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 2971 if (!IS_ERR(transfer_to[PRJQUOTA])) { 2972 err = __dquot_transfer(inode, transfer_to); 2973 if (err) 2974 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2975 dqput(transfer_to[PRJQUOTA]); 2976 } 2977 return err; 2978 } 2979 2980 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 2981 { 2982 struct inode *inode = file_inode(filp); 2983 struct f2fs_inode_info *fi = F2FS_I(inode); 2984 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2985 struct page *ipage; 2986 kprojid_t kprojid; 2987 int err; 2988 2989 if (!f2fs_sb_has_project_quota(sbi)) { 2990 if (projid != F2FS_DEF_PROJID) 2991 return -EOPNOTSUPP; 2992 else 2993 return 0; 2994 } 2995 2996 if (!f2fs_has_extra_attr(inode)) 2997 return -EOPNOTSUPP; 2998 2999 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3000 3001 if (projid_eq(kprojid, F2FS_I(inode)->i_projid)) 3002 return 0; 3003 3004 err = -EPERM; 3005 /* Is it quota file? Do not allow user to mess with it */ 3006 if (IS_NOQUOTA(inode)) 3007 return err; 3008 3009 ipage = f2fs_get_node_page(sbi, inode->i_ino); 3010 if (IS_ERR(ipage)) 3011 return PTR_ERR(ipage); 3012 3013 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize, 3014 i_projid)) { 3015 err = -EOVERFLOW; 3016 f2fs_put_page(ipage, 1); 3017 return err; 3018 } 3019 f2fs_put_page(ipage, 1); 3020 3021 err = dquot_initialize(inode); 3022 if (err) 3023 return err; 3024 3025 f2fs_lock_op(sbi); 3026 err = f2fs_transfer_project_quota(inode, kprojid); 3027 if (err) 3028 goto out_unlock; 3029 3030 F2FS_I(inode)->i_projid = kprojid; 3031 inode->i_ctime = current_time(inode); 3032 f2fs_mark_inode_dirty_sync(inode, true); 3033 out_unlock: 3034 f2fs_unlock_op(sbi); 3035 return err; 3036 } 3037 #else 3038 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3039 { 3040 return 0; 3041 } 3042 3043 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 3044 { 3045 if (projid != F2FS_DEF_PROJID) 3046 return -EOPNOTSUPP; 3047 return 0; 3048 } 3049 #endif 3050 3051 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */ 3052 3053 /* 3054 * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable 3055 * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its 3056 * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS. 3057 */ 3058 3059 static const struct { 3060 u32 iflag; 3061 u32 xflag; 3062 } f2fs_xflags_map[] = { 3063 { F2FS_SYNC_FL, FS_XFLAG_SYNC }, 3064 { F2FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE }, 3065 { F2FS_APPEND_FL, FS_XFLAG_APPEND }, 3066 { F2FS_NODUMP_FL, FS_XFLAG_NODUMP }, 3067 { F2FS_NOATIME_FL, FS_XFLAG_NOATIME }, 3068 { F2FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT }, 3069 }; 3070 3071 #define F2FS_SUPPORTED_XFLAGS ( \ 3072 FS_XFLAG_SYNC | \ 3073 FS_XFLAG_IMMUTABLE | \ 3074 FS_XFLAG_APPEND | \ 3075 FS_XFLAG_NODUMP | \ 3076 FS_XFLAG_NOATIME | \ 3077 FS_XFLAG_PROJINHERIT) 3078 3079 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */ 3080 static inline u32 f2fs_iflags_to_xflags(u32 iflags) 3081 { 3082 u32 xflags = 0; 3083 int i; 3084 3085 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++) 3086 if (iflags & f2fs_xflags_map[i].iflag) 3087 xflags |= f2fs_xflags_map[i].xflag; 3088 3089 return xflags; 3090 } 3091 3092 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */ 3093 static inline u32 f2fs_xflags_to_iflags(u32 xflags) 3094 { 3095 u32 iflags = 0; 3096 int i; 3097 3098 for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++) 3099 if (xflags & f2fs_xflags_map[i].xflag) 3100 iflags |= f2fs_xflags_map[i].iflag; 3101 3102 return iflags; 3103 } 3104 3105 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa) 3106 { 3107 struct f2fs_inode_info *fi = F2FS_I(inode); 3108 3109 simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags)); 3110 3111 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3112 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3113 } 3114 3115 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg) 3116 { 3117 struct inode *inode = file_inode(filp); 3118 struct fsxattr fa; 3119 3120 f2fs_fill_fsxattr(inode, &fa); 3121 3122 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa))) 3123 return -EFAULT; 3124 return 0; 3125 } 3126 3127 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg) 3128 { 3129 struct inode *inode = file_inode(filp); 3130 struct fsxattr fa, old_fa; 3131 u32 iflags; 3132 int err; 3133 3134 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa))) 3135 return -EFAULT; 3136 3137 /* Make sure caller has proper permission */ 3138 if (!inode_owner_or_capable(inode)) 3139 return -EACCES; 3140 3141 if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS) 3142 return -EOPNOTSUPP; 3143 3144 iflags = f2fs_xflags_to_iflags(fa.fsx_xflags); 3145 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3146 return -EOPNOTSUPP; 3147 3148 err = mnt_want_write_file(filp); 3149 if (err) 3150 return err; 3151 3152 inode_lock(inode); 3153 3154 f2fs_fill_fsxattr(inode, &old_fa); 3155 err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa); 3156 if (err) 3157 goto out; 3158 3159 err = f2fs_setflags_common(inode, iflags, 3160 f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS)); 3161 if (err) 3162 goto out; 3163 3164 err = f2fs_ioc_setproject(filp, fa.fsx_projid); 3165 out: 3166 inode_unlock(inode); 3167 mnt_drop_write_file(filp); 3168 return err; 3169 } 3170 3171 int f2fs_pin_file_control(struct inode *inode, bool inc) 3172 { 3173 struct f2fs_inode_info *fi = F2FS_I(inode); 3174 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3175 3176 /* Use i_gc_failures for normal file as a risk signal. */ 3177 if (inc) 3178 f2fs_i_gc_failures_write(inode, 3179 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 3180 3181 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 3182 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3183 __func__, inode->i_ino, 3184 fi->i_gc_failures[GC_FAILURE_PIN]); 3185 clear_inode_flag(inode, FI_PIN_FILE); 3186 return -EAGAIN; 3187 } 3188 return 0; 3189 } 3190 3191 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3192 { 3193 struct inode *inode = file_inode(filp); 3194 __u32 pin; 3195 int ret = 0; 3196 3197 if (get_user(pin, (__u32 __user *)arg)) 3198 return -EFAULT; 3199 3200 if (!S_ISREG(inode->i_mode)) 3201 return -EINVAL; 3202 3203 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 3204 return -EROFS; 3205 3206 ret = mnt_want_write_file(filp); 3207 if (ret) 3208 return ret; 3209 3210 inode_lock(inode); 3211 3212 if (f2fs_should_update_outplace(inode, NULL)) { 3213 ret = -EINVAL; 3214 goto out; 3215 } 3216 3217 if (!pin) { 3218 clear_inode_flag(inode, FI_PIN_FILE); 3219 f2fs_i_gc_failures_write(inode, 0); 3220 goto done; 3221 } 3222 3223 if (f2fs_pin_file_control(inode, false)) { 3224 ret = -EAGAIN; 3225 goto out; 3226 } 3227 3228 ret = f2fs_convert_inline_inode(inode); 3229 if (ret) 3230 goto out; 3231 3232 if (f2fs_disable_compressed_file(inode)) { 3233 ret = -EOPNOTSUPP; 3234 goto out; 3235 } 3236 3237 set_inode_flag(inode, FI_PIN_FILE); 3238 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3239 done: 3240 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3241 out: 3242 inode_unlock(inode); 3243 mnt_drop_write_file(filp); 3244 return ret; 3245 } 3246 3247 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3248 { 3249 struct inode *inode = file_inode(filp); 3250 __u32 pin = 0; 3251 3252 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3253 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 3254 return put_user(pin, (u32 __user *)arg); 3255 } 3256 3257 int f2fs_precache_extents(struct inode *inode) 3258 { 3259 struct f2fs_inode_info *fi = F2FS_I(inode); 3260 struct f2fs_map_blocks map; 3261 pgoff_t m_next_extent; 3262 loff_t end; 3263 int err; 3264 3265 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3266 return -EOPNOTSUPP; 3267 3268 map.m_lblk = 0; 3269 map.m_next_pgofs = NULL; 3270 map.m_next_extent = &m_next_extent; 3271 map.m_seg_type = NO_CHECK_TYPE; 3272 map.m_may_create = false; 3273 end = F2FS_I_SB(inode)->max_file_blocks; 3274 3275 while (map.m_lblk < end) { 3276 map.m_len = end - map.m_lblk; 3277 3278 down_write(&fi->i_gc_rwsem[WRITE]); 3279 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE); 3280 up_write(&fi->i_gc_rwsem[WRITE]); 3281 if (err) 3282 return err; 3283 3284 map.m_lblk = m_next_extent; 3285 } 3286 3287 return err; 3288 } 3289 3290 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg) 3291 { 3292 return f2fs_precache_extents(file_inode(filp)); 3293 } 3294 3295 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3296 { 3297 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3298 __u64 block_count; 3299 int ret; 3300 3301 if (!capable(CAP_SYS_ADMIN)) 3302 return -EPERM; 3303 3304 if (f2fs_readonly(sbi->sb)) 3305 return -EROFS; 3306 3307 if (copy_from_user(&block_count, (void __user *)arg, 3308 sizeof(block_count))) 3309 return -EFAULT; 3310 3311 ret = f2fs_resize_fs(sbi, block_count); 3312 3313 return ret; 3314 } 3315 3316 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3317 { 3318 struct inode *inode = file_inode(filp); 3319 3320 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3321 3322 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3323 f2fs_warn(F2FS_I_SB(inode), 3324 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n", 3325 inode->i_ino); 3326 return -EOPNOTSUPP; 3327 } 3328 3329 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3330 } 3331 3332 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3333 { 3334 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3335 return -EOPNOTSUPP; 3336 3337 return fsverity_ioctl_measure(filp, (void __user *)arg); 3338 } 3339 3340 static int f2fs_get_volume_name(struct file *filp, unsigned long arg) 3341 { 3342 struct inode *inode = file_inode(filp); 3343 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3344 char *vbuf; 3345 int count; 3346 int err = 0; 3347 3348 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3349 if (!vbuf) 3350 return -ENOMEM; 3351 3352 down_read(&sbi->sb_lock); 3353 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3354 ARRAY_SIZE(sbi->raw_super->volume_name), 3355 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3356 up_read(&sbi->sb_lock); 3357 3358 if (copy_to_user((char __user *)arg, vbuf, 3359 min(FSLABEL_MAX, count))) 3360 err = -EFAULT; 3361 3362 kvfree(vbuf); 3363 return err; 3364 } 3365 3366 static int f2fs_set_volume_name(struct file *filp, unsigned long arg) 3367 { 3368 struct inode *inode = file_inode(filp); 3369 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3370 char *vbuf; 3371 int err = 0; 3372 3373 if (!capable(CAP_SYS_ADMIN)) 3374 return -EPERM; 3375 3376 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3377 if (IS_ERR(vbuf)) 3378 return PTR_ERR(vbuf); 3379 3380 err = mnt_want_write_file(filp); 3381 if (err) 3382 goto out; 3383 3384 down_write(&sbi->sb_lock); 3385 3386 memset(sbi->raw_super->volume_name, 0, 3387 sizeof(sbi->raw_super->volume_name)); 3388 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3389 sbi->raw_super->volume_name, 3390 ARRAY_SIZE(sbi->raw_super->volume_name)); 3391 3392 err = f2fs_commit_super(sbi, false); 3393 3394 up_write(&sbi->sb_lock); 3395 3396 mnt_drop_write_file(filp); 3397 out: 3398 kfree(vbuf); 3399 return err; 3400 } 3401 3402 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 3403 { 3404 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 3405 return -EIO; 3406 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 3407 return -ENOSPC; 3408 3409 switch (cmd) { 3410 case F2FS_IOC_GETFLAGS: 3411 return f2fs_ioc_getflags(filp, arg); 3412 case F2FS_IOC_SETFLAGS: 3413 return f2fs_ioc_setflags(filp, arg); 3414 case F2FS_IOC_GETVERSION: 3415 return f2fs_ioc_getversion(filp, arg); 3416 case F2FS_IOC_START_ATOMIC_WRITE: 3417 return f2fs_ioc_start_atomic_write(filp); 3418 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 3419 return f2fs_ioc_commit_atomic_write(filp); 3420 case F2FS_IOC_START_VOLATILE_WRITE: 3421 return f2fs_ioc_start_volatile_write(filp); 3422 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 3423 return f2fs_ioc_release_volatile_write(filp); 3424 case F2FS_IOC_ABORT_VOLATILE_WRITE: 3425 return f2fs_ioc_abort_volatile_write(filp); 3426 case F2FS_IOC_SHUTDOWN: 3427 return f2fs_ioc_shutdown(filp, arg); 3428 case FITRIM: 3429 return f2fs_ioc_fitrim(filp, arg); 3430 case F2FS_IOC_SET_ENCRYPTION_POLICY: 3431 return f2fs_ioc_set_encryption_policy(filp, arg); 3432 case F2FS_IOC_GET_ENCRYPTION_POLICY: 3433 return f2fs_ioc_get_encryption_policy(filp, arg); 3434 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 3435 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 3436 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 3437 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 3438 case FS_IOC_ADD_ENCRYPTION_KEY: 3439 return f2fs_ioc_add_encryption_key(filp, arg); 3440 case FS_IOC_REMOVE_ENCRYPTION_KEY: 3441 return f2fs_ioc_remove_encryption_key(filp, arg); 3442 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 3443 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 3444 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 3445 return f2fs_ioc_get_encryption_key_status(filp, arg); 3446 case F2FS_IOC_GARBAGE_COLLECT: 3447 return f2fs_ioc_gc(filp, arg); 3448 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 3449 return f2fs_ioc_gc_range(filp, arg); 3450 case F2FS_IOC_WRITE_CHECKPOINT: 3451 return f2fs_ioc_write_checkpoint(filp, arg); 3452 case F2FS_IOC_DEFRAGMENT: 3453 return f2fs_ioc_defragment(filp, arg); 3454 case F2FS_IOC_MOVE_RANGE: 3455 return f2fs_ioc_move_range(filp, arg); 3456 case F2FS_IOC_FLUSH_DEVICE: 3457 return f2fs_ioc_flush_device(filp, arg); 3458 case F2FS_IOC_GET_FEATURES: 3459 return f2fs_ioc_get_features(filp, arg); 3460 case F2FS_IOC_FSGETXATTR: 3461 return f2fs_ioc_fsgetxattr(filp, arg); 3462 case F2FS_IOC_FSSETXATTR: 3463 return f2fs_ioc_fssetxattr(filp, arg); 3464 case F2FS_IOC_GET_PIN_FILE: 3465 return f2fs_ioc_get_pin_file(filp, arg); 3466 case F2FS_IOC_SET_PIN_FILE: 3467 return f2fs_ioc_set_pin_file(filp, arg); 3468 case F2FS_IOC_PRECACHE_EXTENTS: 3469 return f2fs_ioc_precache_extents(filp, arg); 3470 case F2FS_IOC_RESIZE_FS: 3471 return f2fs_ioc_resize_fs(filp, arg); 3472 case FS_IOC_ENABLE_VERITY: 3473 return f2fs_ioc_enable_verity(filp, arg); 3474 case FS_IOC_MEASURE_VERITY: 3475 return f2fs_ioc_measure_verity(filp, arg); 3476 case F2FS_IOC_GET_VOLUME_NAME: 3477 return f2fs_get_volume_name(filp, arg); 3478 case F2FS_IOC_SET_VOLUME_NAME: 3479 return f2fs_set_volume_name(filp, arg); 3480 default: 3481 return -ENOTTY; 3482 } 3483 } 3484 3485 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter) 3486 { 3487 struct file *file = iocb->ki_filp; 3488 struct inode *inode = file_inode(file); 3489 3490 if (!f2fs_is_compress_backend_ready(inode)) 3491 return -EOPNOTSUPP; 3492 3493 return generic_file_read_iter(iocb, iter); 3494 } 3495 3496 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 3497 { 3498 struct file *file = iocb->ki_filp; 3499 struct inode *inode = file_inode(file); 3500 ssize_t ret; 3501 3502 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 3503 ret = -EIO; 3504 goto out; 3505 } 3506 3507 if (!f2fs_is_compress_backend_ready(inode)) 3508 return -EOPNOTSUPP; 3509 3510 if (iocb->ki_flags & IOCB_NOWAIT) { 3511 if (!inode_trylock(inode)) { 3512 ret = -EAGAIN; 3513 goto out; 3514 } 3515 } else { 3516 inode_lock(inode); 3517 } 3518 3519 ret = generic_write_checks(iocb, from); 3520 if (ret > 0) { 3521 bool preallocated = false; 3522 size_t target_size = 0; 3523 int err; 3524 3525 if (iov_iter_fault_in_readable(from, iov_iter_count(from))) 3526 set_inode_flag(inode, FI_NO_PREALLOC); 3527 3528 if ((iocb->ki_flags & IOCB_NOWAIT)) { 3529 if (!f2fs_overwrite_io(inode, iocb->ki_pos, 3530 iov_iter_count(from)) || 3531 f2fs_has_inline_data(inode) || 3532 f2fs_force_buffered_io(inode, iocb, from)) { 3533 clear_inode_flag(inode, FI_NO_PREALLOC); 3534 inode_unlock(inode); 3535 ret = -EAGAIN; 3536 goto out; 3537 } 3538 goto write; 3539 } 3540 3541 if (is_inode_flag_set(inode, FI_NO_PREALLOC)) 3542 goto write; 3543 3544 if (iocb->ki_flags & IOCB_DIRECT) { 3545 /* 3546 * Convert inline data for Direct I/O before entering 3547 * f2fs_direct_IO(). 3548 */ 3549 err = f2fs_convert_inline_inode(inode); 3550 if (err) 3551 goto out_err; 3552 /* 3553 * If force_buffere_io() is true, we have to allocate 3554 * blocks all the time, since f2fs_direct_IO will fall 3555 * back to buffered IO. 3556 */ 3557 if (!f2fs_force_buffered_io(inode, iocb, from) && 3558 allow_outplace_dio(inode, iocb, from)) 3559 goto write; 3560 } 3561 preallocated = true; 3562 target_size = iocb->ki_pos + iov_iter_count(from); 3563 3564 err = f2fs_preallocate_blocks(iocb, from); 3565 if (err) { 3566 out_err: 3567 clear_inode_flag(inode, FI_NO_PREALLOC); 3568 inode_unlock(inode); 3569 ret = err; 3570 goto out; 3571 } 3572 write: 3573 ret = __generic_file_write_iter(iocb, from); 3574 clear_inode_flag(inode, FI_NO_PREALLOC); 3575 3576 /* if we couldn't write data, we should deallocate blocks. */ 3577 if (preallocated && i_size_read(inode) < target_size) 3578 f2fs_truncate(inode); 3579 3580 if (ret > 0) 3581 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret); 3582 } 3583 inode_unlock(inode); 3584 out: 3585 trace_f2fs_file_write_iter(inode, iocb->ki_pos, 3586 iov_iter_count(from), ret); 3587 if (ret > 0) 3588 ret = generic_write_sync(iocb, ret); 3589 return ret; 3590 } 3591 3592 #ifdef CONFIG_COMPAT 3593 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 3594 { 3595 switch (cmd) { 3596 case F2FS_IOC32_GETFLAGS: 3597 cmd = F2FS_IOC_GETFLAGS; 3598 break; 3599 case F2FS_IOC32_SETFLAGS: 3600 cmd = F2FS_IOC_SETFLAGS; 3601 break; 3602 case F2FS_IOC32_GETVERSION: 3603 cmd = F2FS_IOC_GETVERSION; 3604 break; 3605 case F2FS_IOC_START_ATOMIC_WRITE: 3606 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 3607 case F2FS_IOC_START_VOLATILE_WRITE: 3608 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 3609 case F2FS_IOC_ABORT_VOLATILE_WRITE: 3610 case F2FS_IOC_SHUTDOWN: 3611 case FITRIM: 3612 case F2FS_IOC_SET_ENCRYPTION_POLICY: 3613 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 3614 case F2FS_IOC_GET_ENCRYPTION_POLICY: 3615 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 3616 case FS_IOC_ADD_ENCRYPTION_KEY: 3617 case FS_IOC_REMOVE_ENCRYPTION_KEY: 3618 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 3619 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 3620 case F2FS_IOC_GARBAGE_COLLECT: 3621 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 3622 case F2FS_IOC_WRITE_CHECKPOINT: 3623 case F2FS_IOC_DEFRAGMENT: 3624 case F2FS_IOC_MOVE_RANGE: 3625 case F2FS_IOC_FLUSH_DEVICE: 3626 case F2FS_IOC_GET_FEATURES: 3627 case F2FS_IOC_FSGETXATTR: 3628 case F2FS_IOC_FSSETXATTR: 3629 case F2FS_IOC_GET_PIN_FILE: 3630 case F2FS_IOC_SET_PIN_FILE: 3631 case F2FS_IOC_PRECACHE_EXTENTS: 3632 case F2FS_IOC_RESIZE_FS: 3633 case FS_IOC_ENABLE_VERITY: 3634 case FS_IOC_MEASURE_VERITY: 3635 case F2FS_IOC_GET_VOLUME_NAME: 3636 case F2FS_IOC_SET_VOLUME_NAME: 3637 break; 3638 default: 3639 return -ENOIOCTLCMD; 3640 } 3641 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 3642 } 3643 #endif 3644 3645 const struct file_operations f2fs_file_operations = { 3646 .llseek = f2fs_llseek, 3647 .read_iter = f2fs_file_read_iter, 3648 .write_iter = f2fs_file_write_iter, 3649 .open = f2fs_file_open, 3650 .release = f2fs_release_file, 3651 .mmap = f2fs_file_mmap, 3652 .flush = f2fs_file_flush, 3653 .fsync = f2fs_sync_file, 3654 .fallocate = f2fs_fallocate, 3655 .unlocked_ioctl = f2fs_ioctl, 3656 #ifdef CONFIG_COMPAT 3657 .compat_ioctl = f2fs_compat_ioctl, 3658 #endif 3659 .splice_read = generic_file_splice_read, 3660 .splice_write = iter_file_splice_write, 3661 }; 3662