1 /* 2 * fs/f2fs/file.c 3 * 4 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 5 * http://www.samsung.com/ 6 * 7 * This program is free software; you can redistribute it and/or modify 8 * it under the terms of the GNU General Public License version 2 as 9 * published by the Free Software Foundation. 10 */ 11 #include <linux/fs.h> 12 #include <linux/f2fs_fs.h> 13 #include <linux/stat.h> 14 #include <linux/buffer_head.h> 15 #include <linux/writeback.h> 16 #include <linux/blkdev.h> 17 #include <linux/falloc.h> 18 #include <linux/types.h> 19 #include <linux/compat.h> 20 #include <linux/uaccess.h> 21 #include <linux/mount.h> 22 #include <linux/pagevec.h> 23 #include <linux/uio.h> 24 #include <linux/uuid.h> 25 #include <linux/file.h> 26 27 #include "f2fs.h" 28 #include "node.h" 29 #include "segment.h" 30 #include "xattr.h" 31 #include "acl.h" 32 #include "gc.h" 33 #include "trace.h" 34 #include <trace/events/f2fs.h> 35 36 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 37 { 38 struct inode *inode = file_inode(vmf->vma->vm_file); 39 vm_fault_t ret; 40 41 down_read(&F2FS_I(inode)->i_mmap_sem); 42 ret = filemap_fault(vmf); 43 up_read(&F2FS_I(inode)->i_mmap_sem); 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 int err; 55 56 if (unlikely(f2fs_cp_error(sbi))) { 57 err = -EIO; 58 goto err; 59 } 60 61 sb_start_pagefault(inode->i_sb); 62 63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 64 65 /* block allocation */ 66 f2fs_lock_op(sbi); 67 set_new_dnode(&dn, inode, NULL, NULL, 0); 68 err = f2fs_reserve_block(&dn, page->index); 69 if (err) { 70 f2fs_unlock_op(sbi); 71 goto out; 72 } 73 f2fs_put_dnode(&dn); 74 f2fs_unlock_op(sbi); 75 76 f2fs_balance_fs(sbi, dn.node_changed); 77 78 file_update_time(vmf->vma->vm_file); 79 down_read(&F2FS_I(inode)->i_mmap_sem); 80 lock_page(page); 81 if (unlikely(page->mapping != inode->i_mapping || 82 page_offset(page) > i_size_read(inode) || 83 !PageUptodate(page))) { 84 unlock_page(page); 85 err = -EFAULT; 86 goto out_sem; 87 } 88 89 /* 90 * check to see if the page is mapped already (no holes) 91 */ 92 if (PageMappedToDisk(page)) 93 goto mapped; 94 95 /* page is wholly or partially inside EOF */ 96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 97 i_size_read(inode)) { 98 loff_t offset; 99 100 offset = i_size_read(inode) & ~PAGE_MASK; 101 zero_user_segment(page, offset, PAGE_SIZE); 102 } 103 set_page_dirty(page); 104 if (!PageUptodate(page)) 105 SetPageUptodate(page); 106 107 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE); 108 109 trace_f2fs_vm_page_mkwrite(page, DATA); 110 mapped: 111 /* fill the page */ 112 f2fs_wait_on_page_writeback(page, DATA, false); 113 114 /* wait for GCed page writeback via META_MAPPING */ 115 if (f2fs_post_read_required(inode)) 116 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr); 117 118 out_sem: 119 up_read(&F2FS_I(inode)->i_mmap_sem); 120 out: 121 sb_end_pagefault(inode->i_sb); 122 f2fs_update_time(sbi, REQ_TIME); 123 err: 124 return block_page_mkwrite_return(err); 125 } 126 127 static const struct vm_operations_struct f2fs_file_vm_ops = { 128 .fault = f2fs_filemap_fault, 129 .map_pages = filemap_map_pages, 130 .page_mkwrite = f2fs_vm_page_mkwrite, 131 }; 132 133 static int get_parent_ino(struct inode *inode, nid_t *pino) 134 { 135 struct dentry *dentry; 136 137 inode = igrab(inode); 138 dentry = d_find_any_alias(inode); 139 iput(inode); 140 if (!dentry) 141 return 0; 142 143 *pino = parent_ino(dentry); 144 dput(dentry); 145 return 1; 146 } 147 148 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 149 { 150 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 151 enum cp_reason_type cp_reason = CP_NO_NEEDED; 152 153 if (!S_ISREG(inode->i_mode)) 154 cp_reason = CP_NON_REGULAR; 155 else if (inode->i_nlink != 1) 156 cp_reason = CP_HARDLINK; 157 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 158 cp_reason = CP_SB_NEED_CP; 159 else if (file_wrong_pino(inode)) 160 cp_reason = CP_WRONG_PINO; 161 else if (!f2fs_space_for_roll_forward(sbi)) 162 cp_reason = CP_NO_SPC_ROLL; 163 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 164 cp_reason = CP_NODE_NEED_CP; 165 else if (test_opt(sbi, FASTBOOT)) 166 cp_reason = CP_FASTBOOT_MODE; 167 else if (F2FS_OPTION(sbi).active_logs == 2) 168 cp_reason = CP_SPEC_LOG_NUM; 169 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 170 f2fs_need_dentry_mark(sbi, inode->i_ino) && 171 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 172 TRANS_DIR_INO)) 173 cp_reason = CP_RECOVER_DIR; 174 175 return cp_reason; 176 } 177 178 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 179 { 180 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 181 bool ret = false; 182 /* But we need to avoid that there are some inode updates */ 183 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 184 ret = true; 185 f2fs_put_page(i, 0); 186 return ret; 187 } 188 189 static void try_to_fix_pino(struct inode *inode) 190 { 191 struct f2fs_inode_info *fi = F2FS_I(inode); 192 nid_t pino; 193 194 down_write(&fi->i_sem); 195 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 196 get_parent_ino(inode, &pino)) { 197 f2fs_i_pino_write(inode, pino); 198 file_got_pino(inode); 199 } 200 up_write(&fi->i_sem); 201 } 202 203 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 204 int datasync, bool atomic) 205 { 206 struct inode *inode = file->f_mapping->host; 207 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 208 nid_t ino = inode->i_ino; 209 int ret = 0; 210 enum cp_reason_type cp_reason = 0; 211 struct writeback_control wbc = { 212 .sync_mode = WB_SYNC_ALL, 213 .nr_to_write = LONG_MAX, 214 .for_reclaim = 0, 215 }; 216 217 if (unlikely(f2fs_readonly(inode->i_sb))) 218 return 0; 219 220 trace_f2fs_sync_file_enter(inode); 221 222 /* if fdatasync is triggered, let's do in-place-update */ 223 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 224 set_inode_flag(inode, FI_NEED_IPU); 225 ret = file_write_and_wait_range(file, start, end); 226 clear_inode_flag(inode, FI_NEED_IPU); 227 228 if (ret) { 229 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 230 return ret; 231 } 232 233 /* if the inode is dirty, let's recover all the time */ 234 if (!f2fs_skip_inode_update(inode, datasync)) { 235 f2fs_write_inode(inode, NULL); 236 goto go_write; 237 } 238 239 /* 240 * if there is no written data, don't waste time to write recovery info. 241 */ 242 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 243 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 244 245 /* it may call write_inode just prior to fsync */ 246 if (need_inode_page_update(sbi, ino)) 247 goto go_write; 248 249 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 250 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 251 goto flush_out; 252 goto out; 253 } 254 go_write: 255 /* 256 * Both of fdatasync() and fsync() are able to be recovered from 257 * sudden-power-off. 258 */ 259 down_read(&F2FS_I(inode)->i_sem); 260 cp_reason = need_do_checkpoint(inode); 261 up_read(&F2FS_I(inode)->i_sem); 262 263 if (cp_reason) { 264 /* all the dirty node pages should be flushed for POR */ 265 ret = f2fs_sync_fs(inode->i_sb, 1); 266 267 /* 268 * We've secured consistency through sync_fs. Following pino 269 * will be used only for fsynced inodes after checkpoint. 270 */ 271 try_to_fix_pino(inode); 272 clear_inode_flag(inode, FI_APPEND_WRITE); 273 clear_inode_flag(inode, FI_UPDATE_WRITE); 274 goto out; 275 } 276 sync_nodes: 277 atomic_inc(&sbi->wb_sync_req[NODE]); 278 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic); 279 atomic_dec(&sbi->wb_sync_req[NODE]); 280 if (ret) 281 goto out; 282 283 /* if cp_error was enabled, we should avoid infinite loop */ 284 if (unlikely(f2fs_cp_error(sbi))) { 285 ret = -EIO; 286 goto out; 287 } 288 289 if (f2fs_need_inode_block_update(sbi, ino)) { 290 f2fs_mark_inode_dirty_sync(inode, true); 291 f2fs_write_inode(inode, NULL); 292 goto sync_nodes; 293 } 294 295 /* 296 * If it's atomic_write, it's just fine to keep write ordering. So 297 * here we don't need to wait for node write completion, since we use 298 * node chain which serializes node blocks. If one of node writes are 299 * reordered, we can see simply broken chain, resulting in stopping 300 * roll-forward recovery. It means we'll recover all or none node blocks 301 * given fsync mark. 302 */ 303 if (!atomic) { 304 ret = f2fs_wait_on_node_pages_writeback(sbi, ino); 305 if (ret) 306 goto out; 307 } 308 309 /* once recovery info is written, don't need to tack this */ 310 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 311 clear_inode_flag(inode, FI_APPEND_WRITE); 312 flush_out: 313 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 314 ret = f2fs_issue_flush(sbi, inode->i_ino); 315 if (!ret) { 316 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 317 clear_inode_flag(inode, FI_UPDATE_WRITE); 318 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 319 } 320 f2fs_update_time(sbi, REQ_TIME); 321 out: 322 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 323 f2fs_trace_ios(NULL, 1); 324 return ret; 325 } 326 327 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 328 { 329 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 330 return -EIO; 331 return f2fs_do_sync_file(file, start, end, datasync, false); 332 } 333 334 static pgoff_t __get_first_dirty_index(struct address_space *mapping, 335 pgoff_t pgofs, int whence) 336 { 337 struct page *page; 338 int nr_pages; 339 340 if (whence != SEEK_DATA) 341 return 0; 342 343 /* find first dirty page index */ 344 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY, 345 1, &page); 346 if (!nr_pages) 347 return ULONG_MAX; 348 pgofs = page->index; 349 put_page(page); 350 return pgofs; 351 } 352 353 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs, 354 int whence) 355 { 356 switch (whence) { 357 case SEEK_DATA: 358 if ((blkaddr == NEW_ADDR && dirty == pgofs) || 359 is_valid_blkaddr(blkaddr)) 360 return true; 361 break; 362 case SEEK_HOLE: 363 if (blkaddr == NULL_ADDR) 364 return true; 365 break; 366 } 367 return false; 368 } 369 370 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 371 { 372 struct inode *inode = file->f_mapping->host; 373 loff_t maxbytes = inode->i_sb->s_maxbytes; 374 struct dnode_of_data dn; 375 pgoff_t pgofs, end_offset, dirty; 376 loff_t data_ofs = offset; 377 loff_t isize; 378 int err = 0; 379 380 inode_lock(inode); 381 382 isize = i_size_read(inode); 383 if (offset >= isize) 384 goto fail; 385 386 /* handle inline data case */ 387 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) { 388 if (whence == SEEK_HOLE) 389 data_ofs = isize; 390 goto found; 391 } 392 393 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 394 395 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence); 396 397 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 398 set_new_dnode(&dn, inode, NULL, NULL, 0); 399 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 400 if (err && err != -ENOENT) { 401 goto fail; 402 } else if (err == -ENOENT) { 403 /* direct node does not exists */ 404 if (whence == SEEK_DATA) { 405 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 406 continue; 407 } else { 408 goto found; 409 } 410 } 411 412 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 413 414 /* find data/hole in dnode block */ 415 for (; dn.ofs_in_node < end_offset; 416 dn.ofs_in_node++, pgofs++, 417 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 418 block_t blkaddr; 419 420 blkaddr = datablock_addr(dn.inode, 421 dn.node_page, dn.ofs_in_node); 422 423 if (__found_offset(blkaddr, dirty, pgofs, whence)) { 424 f2fs_put_dnode(&dn); 425 goto found; 426 } 427 } 428 f2fs_put_dnode(&dn); 429 } 430 431 if (whence == SEEK_DATA) 432 goto fail; 433 found: 434 if (whence == SEEK_HOLE && data_ofs > isize) 435 data_ofs = isize; 436 inode_unlock(inode); 437 return vfs_setpos(file, data_ofs, maxbytes); 438 fail: 439 inode_unlock(inode); 440 return -ENXIO; 441 } 442 443 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 444 { 445 struct inode *inode = file->f_mapping->host; 446 loff_t maxbytes = inode->i_sb->s_maxbytes; 447 448 switch (whence) { 449 case SEEK_SET: 450 case SEEK_CUR: 451 case SEEK_END: 452 return generic_file_llseek_size(file, offset, whence, 453 maxbytes, i_size_read(inode)); 454 case SEEK_DATA: 455 case SEEK_HOLE: 456 if (offset < 0) 457 return -ENXIO; 458 return f2fs_seek_block(file, offset, whence); 459 } 460 461 return -EINVAL; 462 } 463 464 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 465 { 466 struct inode *inode = file_inode(file); 467 int err; 468 469 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 470 return -EIO; 471 472 /* we don't need to use inline_data strictly */ 473 err = f2fs_convert_inline_inode(inode); 474 if (err) 475 return err; 476 477 file_accessed(file); 478 vma->vm_ops = &f2fs_file_vm_ops; 479 return 0; 480 } 481 482 static int f2fs_file_open(struct inode *inode, struct file *filp) 483 { 484 int err = fscrypt_file_open(inode, filp); 485 486 if (err) 487 return err; 488 489 filp->f_mode |= FMODE_NOWAIT; 490 491 return dquot_file_open(inode, filp); 492 } 493 494 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 495 { 496 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 497 struct f2fs_node *raw_node; 498 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 499 __le32 *addr; 500 int base = 0; 501 502 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode)) 503 base = get_extra_isize(dn->inode); 504 505 raw_node = F2FS_NODE(dn->node_page); 506 addr = blkaddr_in_node(raw_node) + base + ofs; 507 508 for (; count > 0; count--, addr++, dn->ofs_in_node++) { 509 block_t blkaddr = le32_to_cpu(*addr); 510 511 if (blkaddr == NULL_ADDR) 512 continue; 513 514 dn->data_blkaddr = NULL_ADDR; 515 f2fs_set_data_blkaddr(dn); 516 f2fs_invalidate_blocks(sbi, blkaddr); 517 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 518 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 519 nr_free++; 520 } 521 522 if (nr_free) { 523 pgoff_t fofs; 524 /* 525 * once we invalidate valid blkaddr in range [ofs, ofs + count], 526 * we will invalidate all blkaddr in the whole range. 527 */ 528 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 529 dn->inode) + ofs; 530 f2fs_update_extent_cache_range(dn, fofs, 0, len); 531 dec_valid_block_count(sbi, dn->inode, nr_free); 532 } 533 dn->ofs_in_node = ofs; 534 535 f2fs_update_time(sbi, REQ_TIME); 536 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 537 dn->ofs_in_node, nr_free); 538 } 539 540 void f2fs_truncate_data_blocks(struct dnode_of_data *dn) 541 { 542 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK); 543 } 544 545 static int truncate_partial_data_page(struct inode *inode, u64 from, 546 bool cache_only) 547 { 548 loff_t offset = from & (PAGE_SIZE - 1); 549 pgoff_t index = from >> PAGE_SHIFT; 550 struct address_space *mapping = inode->i_mapping; 551 struct page *page; 552 553 if (!offset && !cache_only) 554 return 0; 555 556 if (cache_only) { 557 page = find_lock_page(mapping, index); 558 if (page && PageUptodate(page)) 559 goto truncate_out; 560 f2fs_put_page(page, 1); 561 return 0; 562 } 563 564 page = f2fs_get_lock_data_page(inode, index, true); 565 if (IS_ERR(page)) 566 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 567 truncate_out: 568 f2fs_wait_on_page_writeback(page, DATA, true); 569 zero_user(page, offset, PAGE_SIZE - offset); 570 571 /* An encrypted inode should have a key and truncate the last page. */ 572 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode)); 573 if (!cache_only) 574 set_page_dirty(page); 575 f2fs_put_page(page, 1); 576 return 0; 577 } 578 579 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 580 { 581 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 582 struct dnode_of_data dn; 583 pgoff_t free_from; 584 int count = 0, err = 0; 585 struct page *ipage; 586 bool truncate_page = false; 587 588 trace_f2fs_truncate_blocks_enter(inode, from); 589 590 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 591 592 if (free_from >= sbi->max_file_blocks) 593 goto free_partial; 594 595 if (lock) 596 f2fs_lock_op(sbi); 597 598 ipage = f2fs_get_node_page(sbi, inode->i_ino); 599 if (IS_ERR(ipage)) { 600 err = PTR_ERR(ipage); 601 goto out; 602 } 603 604 if (f2fs_has_inline_data(inode)) { 605 f2fs_truncate_inline_inode(inode, ipage, from); 606 f2fs_put_page(ipage, 1); 607 truncate_page = true; 608 goto out; 609 } 610 611 set_new_dnode(&dn, inode, ipage, NULL, 0); 612 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 613 if (err) { 614 if (err == -ENOENT) 615 goto free_next; 616 goto out; 617 } 618 619 count = ADDRS_PER_PAGE(dn.node_page, inode); 620 621 count -= dn.ofs_in_node; 622 f2fs_bug_on(sbi, count < 0); 623 624 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 625 f2fs_truncate_data_blocks_range(&dn, count); 626 free_from += count; 627 } 628 629 f2fs_put_dnode(&dn); 630 free_next: 631 err = f2fs_truncate_inode_blocks(inode, free_from); 632 out: 633 if (lock) 634 f2fs_unlock_op(sbi); 635 free_partial: 636 /* lastly zero out the first data page */ 637 if (!err) 638 err = truncate_partial_data_page(inode, from, truncate_page); 639 640 trace_f2fs_truncate_blocks_exit(inode, err); 641 return err; 642 } 643 644 int f2fs_truncate(struct inode *inode) 645 { 646 int err; 647 648 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 649 return -EIO; 650 651 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 652 S_ISLNK(inode->i_mode))) 653 return 0; 654 655 trace_f2fs_truncate(inode); 656 657 #ifdef CONFIG_F2FS_FAULT_INJECTION 658 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) { 659 f2fs_show_injection_info(FAULT_TRUNCATE); 660 return -EIO; 661 } 662 #endif 663 /* we should check inline_data size */ 664 if (!f2fs_may_inline_data(inode)) { 665 err = f2fs_convert_inline_inode(inode); 666 if (err) 667 return err; 668 } 669 670 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 671 if (err) 672 return err; 673 674 inode->i_mtime = inode->i_ctime = current_time(inode); 675 f2fs_mark_inode_dirty_sync(inode, false); 676 return 0; 677 } 678 679 int f2fs_getattr(const struct path *path, struct kstat *stat, 680 u32 request_mask, unsigned int query_flags) 681 { 682 struct inode *inode = d_inode(path->dentry); 683 struct f2fs_inode_info *fi = F2FS_I(inode); 684 struct f2fs_inode *ri; 685 unsigned int flags; 686 687 if (f2fs_has_extra_attr(inode) && 688 f2fs_sb_has_inode_crtime(inode->i_sb) && 689 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 690 stat->result_mask |= STATX_BTIME; 691 stat->btime.tv_sec = fi->i_crtime.tv_sec; 692 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 693 } 694 695 flags = fi->i_flags & F2FS_FL_USER_VISIBLE; 696 if (flags & F2FS_APPEND_FL) 697 stat->attributes |= STATX_ATTR_APPEND; 698 if (flags & F2FS_COMPR_FL) 699 stat->attributes |= STATX_ATTR_COMPRESSED; 700 if (f2fs_encrypted_inode(inode)) 701 stat->attributes |= STATX_ATTR_ENCRYPTED; 702 if (flags & F2FS_IMMUTABLE_FL) 703 stat->attributes |= STATX_ATTR_IMMUTABLE; 704 if (flags & F2FS_NODUMP_FL) 705 stat->attributes |= STATX_ATTR_NODUMP; 706 707 stat->attributes_mask |= (STATX_ATTR_APPEND | 708 STATX_ATTR_COMPRESSED | 709 STATX_ATTR_ENCRYPTED | 710 STATX_ATTR_IMMUTABLE | 711 STATX_ATTR_NODUMP); 712 713 generic_fillattr(inode, stat); 714 715 /* we need to show initial sectors used for inline_data/dentries */ 716 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 717 f2fs_has_inline_dentry(inode)) 718 stat->blocks += (stat->size + 511) >> 9; 719 720 return 0; 721 } 722 723 #ifdef CONFIG_F2FS_FS_POSIX_ACL 724 static void __setattr_copy(struct inode *inode, const struct iattr *attr) 725 { 726 unsigned int ia_valid = attr->ia_valid; 727 728 if (ia_valid & ATTR_UID) 729 inode->i_uid = attr->ia_uid; 730 if (ia_valid & ATTR_GID) 731 inode->i_gid = attr->ia_gid; 732 if (ia_valid & ATTR_ATIME) 733 inode->i_atime = timespec_trunc(attr->ia_atime, 734 inode->i_sb->s_time_gran); 735 if (ia_valid & ATTR_MTIME) 736 inode->i_mtime = timespec_trunc(attr->ia_mtime, 737 inode->i_sb->s_time_gran); 738 if (ia_valid & ATTR_CTIME) 739 inode->i_ctime = timespec_trunc(attr->ia_ctime, 740 inode->i_sb->s_time_gran); 741 if (ia_valid & ATTR_MODE) { 742 umode_t mode = attr->ia_mode; 743 744 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) 745 mode &= ~S_ISGID; 746 set_acl_inode(inode, mode); 747 } 748 } 749 #else 750 #define __setattr_copy setattr_copy 751 #endif 752 753 int f2fs_setattr(struct dentry *dentry, struct iattr *attr) 754 { 755 struct inode *inode = d_inode(dentry); 756 int err; 757 bool size_changed = false; 758 759 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 760 return -EIO; 761 762 err = setattr_prepare(dentry, attr); 763 if (err) 764 return err; 765 766 err = fscrypt_prepare_setattr(dentry, attr); 767 if (err) 768 return err; 769 770 if (is_quota_modification(inode, attr)) { 771 err = dquot_initialize(inode); 772 if (err) 773 return err; 774 } 775 if ((attr->ia_valid & ATTR_UID && 776 !uid_eq(attr->ia_uid, inode->i_uid)) || 777 (attr->ia_valid & ATTR_GID && 778 !gid_eq(attr->ia_gid, inode->i_gid))) { 779 err = dquot_transfer(inode, attr); 780 if (err) 781 return err; 782 } 783 784 if (attr->ia_valid & ATTR_SIZE) { 785 if (attr->ia_size <= i_size_read(inode)) { 786 down_write(&F2FS_I(inode)->i_mmap_sem); 787 truncate_setsize(inode, attr->ia_size); 788 err = f2fs_truncate(inode); 789 up_write(&F2FS_I(inode)->i_mmap_sem); 790 if (err) 791 return err; 792 } else { 793 /* 794 * do not trim all blocks after i_size if target size is 795 * larger than i_size. 796 */ 797 down_write(&F2FS_I(inode)->i_mmap_sem); 798 truncate_setsize(inode, attr->ia_size); 799 up_write(&F2FS_I(inode)->i_mmap_sem); 800 801 /* should convert inline inode here */ 802 if (!f2fs_may_inline_data(inode)) { 803 err = f2fs_convert_inline_inode(inode); 804 if (err) 805 return err; 806 } 807 inode->i_mtime = inode->i_ctime = current_time(inode); 808 } 809 810 down_write(&F2FS_I(inode)->i_sem); 811 F2FS_I(inode)->last_disk_size = i_size_read(inode); 812 up_write(&F2FS_I(inode)->i_sem); 813 814 size_changed = true; 815 } 816 817 __setattr_copy(inode, attr); 818 819 if (attr->ia_valid & ATTR_MODE) { 820 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode)); 821 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) { 822 inode->i_mode = F2FS_I(inode)->i_acl_mode; 823 clear_inode_flag(inode, FI_ACL_MODE); 824 } 825 } 826 827 /* file size may changed here */ 828 f2fs_mark_inode_dirty_sync(inode, size_changed); 829 830 /* inode change will produce dirty node pages flushed by checkpoint */ 831 f2fs_balance_fs(F2FS_I_SB(inode), true); 832 833 return err; 834 } 835 836 const struct inode_operations f2fs_file_inode_operations = { 837 .getattr = f2fs_getattr, 838 .setattr = f2fs_setattr, 839 .get_acl = f2fs_get_acl, 840 .set_acl = f2fs_set_acl, 841 #ifdef CONFIG_F2FS_FS_XATTR 842 .listxattr = f2fs_listxattr, 843 #endif 844 .fiemap = f2fs_fiemap, 845 }; 846 847 static int fill_zero(struct inode *inode, pgoff_t index, 848 loff_t start, loff_t len) 849 { 850 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 851 struct page *page; 852 853 if (!len) 854 return 0; 855 856 f2fs_balance_fs(sbi, true); 857 858 f2fs_lock_op(sbi); 859 page = f2fs_get_new_data_page(inode, NULL, index, false); 860 f2fs_unlock_op(sbi); 861 862 if (IS_ERR(page)) 863 return PTR_ERR(page); 864 865 f2fs_wait_on_page_writeback(page, DATA, true); 866 zero_user(page, start, len); 867 set_page_dirty(page); 868 f2fs_put_page(page, 1); 869 return 0; 870 } 871 872 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 873 { 874 int err; 875 876 while (pg_start < pg_end) { 877 struct dnode_of_data dn; 878 pgoff_t end_offset, count; 879 880 set_new_dnode(&dn, inode, NULL, NULL, 0); 881 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 882 if (err) { 883 if (err == -ENOENT) { 884 pg_start = f2fs_get_next_page_offset(&dn, 885 pg_start); 886 continue; 887 } 888 return err; 889 } 890 891 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 892 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 893 894 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 895 896 f2fs_truncate_data_blocks_range(&dn, count); 897 f2fs_put_dnode(&dn); 898 899 pg_start += count; 900 } 901 return 0; 902 } 903 904 static int punch_hole(struct inode *inode, loff_t offset, loff_t len) 905 { 906 pgoff_t pg_start, pg_end; 907 loff_t off_start, off_end; 908 int ret; 909 910 ret = f2fs_convert_inline_inode(inode); 911 if (ret) 912 return ret; 913 914 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 915 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 916 917 off_start = offset & (PAGE_SIZE - 1); 918 off_end = (offset + len) & (PAGE_SIZE - 1); 919 920 if (pg_start == pg_end) { 921 ret = fill_zero(inode, pg_start, off_start, 922 off_end - off_start); 923 if (ret) 924 return ret; 925 } else { 926 if (off_start) { 927 ret = fill_zero(inode, pg_start++, off_start, 928 PAGE_SIZE - off_start); 929 if (ret) 930 return ret; 931 } 932 if (off_end) { 933 ret = fill_zero(inode, pg_end, 0, off_end); 934 if (ret) 935 return ret; 936 } 937 938 if (pg_start < pg_end) { 939 struct address_space *mapping = inode->i_mapping; 940 loff_t blk_start, blk_end; 941 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 942 943 f2fs_balance_fs(sbi, true); 944 945 blk_start = (loff_t)pg_start << PAGE_SHIFT; 946 blk_end = (loff_t)pg_end << PAGE_SHIFT; 947 down_write(&F2FS_I(inode)->i_mmap_sem); 948 truncate_inode_pages_range(mapping, blk_start, 949 blk_end - 1); 950 951 f2fs_lock_op(sbi); 952 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 953 f2fs_unlock_op(sbi); 954 up_write(&F2FS_I(inode)->i_mmap_sem); 955 } 956 } 957 958 return ret; 959 } 960 961 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 962 int *do_replace, pgoff_t off, pgoff_t len) 963 { 964 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 965 struct dnode_of_data dn; 966 int ret, done, i; 967 968 next_dnode: 969 set_new_dnode(&dn, inode, NULL, NULL, 0); 970 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 971 if (ret && ret != -ENOENT) { 972 return ret; 973 } else if (ret == -ENOENT) { 974 if (dn.max_level == 0) 975 return -ENOENT; 976 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len); 977 blkaddr += done; 978 do_replace += done; 979 goto next; 980 } 981 982 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 983 dn.ofs_in_node, len); 984 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 985 *blkaddr = datablock_addr(dn.inode, 986 dn.node_page, dn.ofs_in_node); 987 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 988 989 if (test_opt(sbi, LFS)) { 990 f2fs_put_dnode(&dn); 991 return -ENOTSUPP; 992 } 993 994 /* do not invalidate this block address */ 995 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 996 *do_replace = 1; 997 } 998 } 999 f2fs_put_dnode(&dn); 1000 next: 1001 len -= done; 1002 off += done; 1003 if (len) 1004 goto next_dnode; 1005 return 0; 1006 } 1007 1008 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1009 int *do_replace, pgoff_t off, int len) 1010 { 1011 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1012 struct dnode_of_data dn; 1013 int ret, i; 1014 1015 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1016 if (*do_replace == 0) 1017 continue; 1018 1019 set_new_dnode(&dn, inode, NULL, NULL, 0); 1020 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1021 if (ret) { 1022 dec_valid_block_count(sbi, inode, 1); 1023 f2fs_invalidate_blocks(sbi, *blkaddr); 1024 } else { 1025 f2fs_update_data_blkaddr(&dn, *blkaddr); 1026 } 1027 f2fs_put_dnode(&dn); 1028 } 1029 return 0; 1030 } 1031 1032 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1033 block_t *blkaddr, int *do_replace, 1034 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1035 { 1036 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1037 pgoff_t i = 0; 1038 int ret; 1039 1040 while (i < len) { 1041 if (blkaddr[i] == NULL_ADDR && !full) { 1042 i++; 1043 continue; 1044 } 1045 1046 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1047 struct dnode_of_data dn; 1048 struct node_info ni; 1049 size_t new_size; 1050 pgoff_t ilen; 1051 1052 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1053 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1054 if (ret) 1055 return ret; 1056 1057 f2fs_get_node_info(sbi, dn.nid, &ni); 1058 ilen = min((pgoff_t) 1059 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1060 dn.ofs_in_node, len - i); 1061 do { 1062 dn.data_blkaddr = datablock_addr(dn.inode, 1063 dn.node_page, dn.ofs_in_node); 1064 f2fs_truncate_data_blocks_range(&dn, 1); 1065 1066 if (do_replace[i]) { 1067 f2fs_i_blocks_write(src_inode, 1068 1, false, false); 1069 f2fs_i_blocks_write(dst_inode, 1070 1, true, false); 1071 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1072 blkaddr[i], ni.version, true, false); 1073 1074 do_replace[i] = 0; 1075 } 1076 dn.ofs_in_node++; 1077 i++; 1078 new_size = (dst + i) << PAGE_SHIFT; 1079 if (dst_inode->i_size < new_size) 1080 f2fs_i_size_write(dst_inode, new_size); 1081 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1082 1083 f2fs_put_dnode(&dn); 1084 } else { 1085 struct page *psrc, *pdst; 1086 1087 psrc = f2fs_get_lock_data_page(src_inode, 1088 src + i, true); 1089 if (IS_ERR(psrc)) 1090 return PTR_ERR(psrc); 1091 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1092 true); 1093 if (IS_ERR(pdst)) { 1094 f2fs_put_page(psrc, 1); 1095 return PTR_ERR(pdst); 1096 } 1097 f2fs_copy_page(psrc, pdst); 1098 set_page_dirty(pdst); 1099 f2fs_put_page(pdst, 1); 1100 f2fs_put_page(psrc, 1); 1101 1102 ret = f2fs_truncate_hole(src_inode, 1103 src + i, src + i + 1); 1104 if (ret) 1105 return ret; 1106 i++; 1107 } 1108 } 1109 return 0; 1110 } 1111 1112 static int __exchange_data_block(struct inode *src_inode, 1113 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1114 pgoff_t len, bool full) 1115 { 1116 block_t *src_blkaddr; 1117 int *do_replace; 1118 pgoff_t olen; 1119 int ret; 1120 1121 while (len) { 1122 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len); 1123 1124 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1125 sizeof(block_t) * olen, GFP_KERNEL); 1126 if (!src_blkaddr) 1127 return -ENOMEM; 1128 1129 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1130 sizeof(int) * olen, GFP_KERNEL); 1131 if (!do_replace) { 1132 kvfree(src_blkaddr); 1133 return -ENOMEM; 1134 } 1135 1136 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1137 do_replace, src, olen); 1138 if (ret) 1139 goto roll_back; 1140 1141 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1142 do_replace, src, dst, olen, full); 1143 if (ret) 1144 goto roll_back; 1145 1146 src += olen; 1147 dst += olen; 1148 len -= olen; 1149 1150 kvfree(src_blkaddr); 1151 kvfree(do_replace); 1152 } 1153 return 0; 1154 1155 roll_back: 1156 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1157 kvfree(src_blkaddr); 1158 kvfree(do_replace); 1159 return ret; 1160 } 1161 1162 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end) 1163 { 1164 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1165 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1166 int ret; 1167 1168 f2fs_balance_fs(sbi, true); 1169 f2fs_lock_op(sbi); 1170 1171 f2fs_drop_extent_tree(inode); 1172 1173 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1174 f2fs_unlock_op(sbi); 1175 return ret; 1176 } 1177 1178 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1179 { 1180 pgoff_t pg_start, pg_end; 1181 loff_t new_size; 1182 int ret; 1183 1184 if (offset + len >= i_size_read(inode)) 1185 return -EINVAL; 1186 1187 /* collapse range should be aligned to block size of f2fs. */ 1188 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1189 return -EINVAL; 1190 1191 ret = f2fs_convert_inline_inode(inode); 1192 if (ret) 1193 return ret; 1194 1195 pg_start = offset >> PAGE_SHIFT; 1196 pg_end = (offset + len) >> PAGE_SHIFT; 1197 1198 /* avoid gc operation during block exchange */ 1199 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1200 1201 down_write(&F2FS_I(inode)->i_mmap_sem); 1202 /* write out all dirty pages from offset */ 1203 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1204 if (ret) 1205 goto out_unlock; 1206 1207 truncate_pagecache(inode, offset); 1208 1209 ret = f2fs_do_collapse(inode, pg_start, pg_end); 1210 if (ret) 1211 goto out_unlock; 1212 1213 /* write out all moved pages, if possible */ 1214 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1215 truncate_pagecache(inode, offset); 1216 1217 new_size = i_size_read(inode) - len; 1218 truncate_pagecache(inode, new_size); 1219 1220 ret = f2fs_truncate_blocks(inode, new_size, true); 1221 if (!ret) 1222 f2fs_i_size_write(inode, new_size); 1223 out_unlock: 1224 up_write(&F2FS_I(inode)->i_mmap_sem); 1225 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1226 return ret; 1227 } 1228 1229 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1230 pgoff_t end) 1231 { 1232 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1233 pgoff_t index = start; 1234 unsigned int ofs_in_node = dn->ofs_in_node; 1235 blkcnt_t count = 0; 1236 int ret; 1237 1238 for (; index < end; index++, dn->ofs_in_node++) { 1239 if (datablock_addr(dn->inode, dn->node_page, 1240 dn->ofs_in_node) == NULL_ADDR) 1241 count++; 1242 } 1243 1244 dn->ofs_in_node = ofs_in_node; 1245 ret = f2fs_reserve_new_blocks(dn, count); 1246 if (ret) 1247 return ret; 1248 1249 dn->ofs_in_node = ofs_in_node; 1250 for (index = start; index < end; index++, dn->ofs_in_node++) { 1251 dn->data_blkaddr = datablock_addr(dn->inode, 1252 dn->node_page, dn->ofs_in_node); 1253 /* 1254 * f2fs_reserve_new_blocks will not guarantee entire block 1255 * allocation. 1256 */ 1257 if (dn->data_blkaddr == NULL_ADDR) { 1258 ret = -ENOSPC; 1259 break; 1260 } 1261 if (dn->data_blkaddr != NEW_ADDR) { 1262 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1263 dn->data_blkaddr = NEW_ADDR; 1264 f2fs_set_data_blkaddr(dn); 1265 } 1266 } 1267 1268 f2fs_update_extent_cache_range(dn, start, 0, index - start); 1269 1270 return ret; 1271 } 1272 1273 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1274 int mode) 1275 { 1276 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1277 struct address_space *mapping = inode->i_mapping; 1278 pgoff_t index, pg_start, pg_end; 1279 loff_t new_size = i_size_read(inode); 1280 loff_t off_start, off_end; 1281 int ret = 0; 1282 1283 ret = inode_newsize_ok(inode, (len + offset)); 1284 if (ret) 1285 return ret; 1286 1287 ret = f2fs_convert_inline_inode(inode); 1288 if (ret) 1289 return ret; 1290 1291 down_write(&F2FS_I(inode)->i_mmap_sem); 1292 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1293 if (ret) 1294 goto out_sem; 1295 1296 truncate_pagecache_range(inode, offset, offset + len - 1); 1297 1298 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1299 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1300 1301 off_start = offset & (PAGE_SIZE - 1); 1302 off_end = (offset + len) & (PAGE_SIZE - 1); 1303 1304 if (pg_start == pg_end) { 1305 ret = fill_zero(inode, pg_start, off_start, 1306 off_end - off_start); 1307 if (ret) 1308 goto out_sem; 1309 1310 new_size = max_t(loff_t, new_size, offset + len); 1311 } else { 1312 if (off_start) { 1313 ret = fill_zero(inode, pg_start++, off_start, 1314 PAGE_SIZE - off_start); 1315 if (ret) 1316 goto out_sem; 1317 1318 new_size = max_t(loff_t, new_size, 1319 (loff_t)pg_start << PAGE_SHIFT); 1320 } 1321 1322 for (index = pg_start; index < pg_end;) { 1323 struct dnode_of_data dn; 1324 unsigned int end_offset; 1325 pgoff_t end; 1326 1327 f2fs_lock_op(sbi); 1328 1329 set_new_dnode(&dn, inode, NULL, NULL, 0); 1330 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1331 if (ret) { 1332 f2fs_unlock_op(sbi); 1333 goto out; 1334 } 1335 1336 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1337 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1338 1339 ret = f2fs_do_zero_range(&dn, index, end); 1340 f2fs_put_dnode(&dn); 1341 f2fs_unlock_op(sbi); 1342 1343 f2fs_balance_fs(sbi, dn.node_changed); 1344 1345 if (ret) 1346 goto out; 1347 1348 index = end; 1349 new_size = max_t(loff_t, new_size, 1350 (loff_t)index << PAGE_SHIFT); 1351 } 1352 1353 if (off_end) { 1354 ret = fill_zero(inode, pg_end, 0, off_end); 1355 if (ret) 1356 goto out; 1357 1358 new_size = max_t(loff_t, new_size, offset + len); 1359 } 1360 } 1361 1362 out: 1363 if (new_size > i_size_read(inode)) { 1364 if (mode & FALLOC_FL_KEEP_SIZE) 1365 file_set_keep_isize(inode); 1366 else 1367 f2fs_i_size_write(inode, new_size); 1368 } 1369 out_sem: 1370 up_write(&F2FS_I(inode)->i_mmap_sem); 1371 1372 return ret; 1373 } 1374 1375 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1376 { 1377 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1378 pgoff_t nr, pg_start, pg_end, delta, idx; 1379 loff_t new_size; 1380 int ret = 0; 1381 1382 new_size = i_size_read(inode) + len; 1383 ret = inode_newsize_ok(inode, new_size); 1384 if (ret) 1385 return ret; 1386 1387 if (offset >= i_size_read(inode)) 1388 return -EINVAL; 1389 1390 /* insert range should be aligned to block size of f2fs. */ 1391 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1392 return -EINVAL; 1393 1394 ret = f2fs_convert_inline_inode(inode); 1395 if (ret) 1396 return ret; 1397 1398 f2fs_balance_fs(sbi, true); 1399 1400 /* avoid gc operation during block exchange */ 1401 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1402 1403 down_write(&F2FS_I(inode)->i_mmap_sem); 1404 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1405 if (ret) 1406 goto out; 1407 1408 /* write out all dirty pages from offset */ 1409 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1410 if (ret) 1411 goto out; 1412 1413 truncate_pagecache(inode, offset); 1414 1415 pg_start = offset >> PAGE_SHIFT; 1416 pg_end = (offset + len) >> PAGE_SHIFT; 1417 delta = pg_end - pg_start; 1418 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1419 1420 while (!ret && idx > pg_start) { 1421 nr = idx - pg_start; 1422 if (nr > delta) 1423 nr = delta; 1424 idx -= nr; 1425 1426 f2fs_lock_op(sbi); 1427 f2fs_drop_extent_tree(inode); 1428 1429 ret = __exchange_data_block(inode, inode, idx, 1430 idx + delta, nr, false); 1431 f2fs_unlock_op(sbi); 1432 } 1433 1434 /* write out all moved pages, if possible */ 1435 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1436 truncate_pagecache(inode, offset); 1437 1438 if (!ret) 1439 f2fs_i_size_write(inode, new_size); 1440 out: 1441 up_write(&F2FS_I(inode)->i_mmap_sem); 1442 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1443 return ret; 1444 } 1445 1446 static int expand_inode_data(struct inode *inode, loff_t offset, 1447 loff_t len, int mode) 1448 { 1449 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1450 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1451 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE }; 1452 pgoff_t pg_end; 1453 loff_t new_size = i_size_read(inode); 1454 loff_t off_end; 1455 int err; 1456 1457 err = inode_newsize_ok(inode, (len + offset)); 1458 if (err) 1459 return err; 1460 1461 err = f2fs_convert_inline_inode(inode); 1462 if (err) 1463 return err; 1464 1465 f2fs_balance_fs(sbi, true); 1466 1467 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1468 off_end = (offset + len) & (PAGE_SIZE - 1); 1469 1470 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT; 1471 map.m_len = pg_end - map.m_lblk; 1472 if (off_end) 1473 map.m_len++; 1474 1475 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); 1476 if (err) { 1477 pgoff_t last_off; 1478 1479 if (!map.m_len) 1480 return err; 1481 1482 last_off = map.m_lblk + map.m_len - 1; 1483 1484 /* update new size to the failed position */ 1485 new_size = (last_off == pg_end) ? offset + len : 1486 (loff_t)(last_off + 1) << PAGE_SHIFT; 1487 } else { 1488 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1489 } 1490 1491 if (new_size > i_size_read(inode)) { 1492 if (mode & FALLOC_FL_KEEP_SIZE) 1493 file_set_keep_isize(inode); 1494 else 1495 f2fs_i_size_write(inode, new_size); 1496 } 1497 1498 return err; 1499 } 1500 1501 static long f2fs_fallocate(struct file *file, int mode, 1502 loff_t offset, loff_t len) 1503 { 1504 struct inode *inode = file_inode(file); 1505 long ret = 0; 1506 1507 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1508 return -EIO; 1509 1510 /* f2fs only support ->fallocate for regular file */ 1511 if (!S_ISREG(inode->i_mode)) 1512 return -EINVAL; 1513 1514 if (f2fs_encrypted_inode(inode) && 1515 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1516 return -EOPNOTSUPP; 1517 1518 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1519 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1520 FALLOC_FL_INSERT_RANGE)) 1521 return -EOPNOTSUPP; 1522 1523 inode_lock(inode); 1524 1525 if (mode & FALLOC_FL_PUNCH_HOLE) { 1526 if (offset >= inode->i_size) 1527 goto out; 1528 1529 ret = punch_hole(inode, offset, len); 1530 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1531 ret = f2fs_collapse_range(inode, offset, len); 1532 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1533 ret = f2fs_zero_range(inode, offset, len, mode); 1534 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1535 ret = f2fs_insert_range(inode, offset, len); 1536 } else { 1537 ret = expand_inode_data(inode, offset, len, mode); 1538 } 1539 1540 if (!ret) { 1541 inode->i_mtime = inode->i_ctime = current_time(inode); 1542 f2fs_mark_inode_dirty_sync(inode, false); 1543 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1544 } 1545 1546 out: 1547 inode_unlock(inode); 1548 1549 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1550 return ret; 1551 } 1552 1553 static int f2fs_release_file(struct inode *inode, struct file *filp) 1554 { 1555 /* 1556 * f2fs_relase_file is called at every close calls. So we should 1557 * not drop any inmemory pages by close called by other process. 1558 */ 1559 if (!(filp->f_mode & FMODE_WRITE) || 1560 atomic_read(&inode->i_writecount) != 1) 1561 return 0; 1562 1563 /* some remained atomic pages should discarded */ 1564 if (f2fs_is_atomic_file(inode)) 1565 f2fs_drop_inmem_pages(inode); 1566 if (f2fs_is_volatile_file(inode)) { 1567 set_inode_flag(inode, FI_DROP_CACHE); 1568 filemap_fdatawrite(inode->i_mapping); 1569 clear_inode_flag(inode, FI_DROP_CACHE); 1570 clear_inode_flag(inode, FI_VOLATILE_FILE); 1571 stat_dec_volatile_write(inode); 1572 } 1573 return 0; 1574 } 1575 1576 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1577 { 1578 struct inode *inode = file_inode(file); 1579 1580 /* 1581 * If the process doing a transaction is crashed, we should do 1582 * roll-back. Otherwise, other reader/write can see corrupted database 1583 * until all the writers close its file. Since this should be done 1584 * before dropping file lock, it needs to do in ->flush. 1585 */ 1586 if (f2fs_is_atomic_file(inode) && 1587 F2FS_I(inode)->inmem_task == current) 1588 f2fs_drop_inmem_pages(inode); 1589 return 0; 1590 } 1591 1592 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg) 1593 { 1594 struct inode *inode = file_inode(filp); 1595 struct f2fs_inode_info *fi = F2FS_I(inode); 1596 unsigned int flags = fi->i_flags; 1597 1598 if (file_is_encrypt(inode)) 1599 flags |= F2FS_ENCRYPT_FL; 1600 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 1601 flags |= F2FS_INLINE_DATA_FL; 1602 1603 flags &= F2FS_FL_USER_VISIBLE; 1604 1605 return put_user(flags, (int __user *)arg); 1606 } 1607 1608 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags) 1609 { 1610 struct f2fs_inode_info *fi = F2FS_I(inode); 1611 unsigned int oldflags; 1612 1613 /* Is it quota file? Do not allow user to mess with it */ 1614 if (IS_NOQUOTA(inode)) 1615 return -EPERM; 1616 1617 flags = f2fs_mask_flags(inode->i_mode, flags); 1618 1619 oldflags = fi->i_flags; 1620 1621 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL)) 1622 if (!capable(CAP_LINUX_IMMUTABLE)) 1623 return -EPERM; 1624 1625 flags = flags & F2FS_FL_USER_MODIFIABLE; 1626 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE; 1627 fi->i_flags = flags; 1628 1629 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1630 set_inode_flag(inode, FI_PROJ_INHERIT); 1631 else 1632 clear_inode_flag(inode, FI_PROJ_INHERIT); 1633 1634 inode->i_ctime = current_time(inode); 1635 f2fs_set_inode_flags(inode); 1636 f2fs_mark_inode_dirty_sync(inode, false); 1637 return 0; 1638 } 1639 1640 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg) 1641 { 1642 struct inode *inode = file_inode(filp); 1643 unsigned int flags; 1644 int ret; 1645 1646 if (!inode_owner_or_capable(inode)) 1647 return -EACCES; 1648 1649 if (get_user(flags, (int __user *)arg)) 1650 return -EFAULT; 1651 1652 ret = mnt_want_write_file(filp); 1653 if (ret) 1654 return ret; 1655 1656 inode_lock(inode); 1657 1658 ret = __f2fs_ioc_setflags(inode, flags); 1659 1660 inode_unlock(inode); 1661 mnt_drop_write_file(filp); 1662 return ret; 1663 } 1664 1665 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1666 { 1667 struct inode *inode = file_inode(filp); 1668 1669 return put_user(inode->i_generation, (int __user *)arg); 1670 } 1671 1672 static int f2fs_ioc_start_atomic_write(struct file *filp) 1673 { 1674 struct inode *inode = file_inode(filp); 1675 int ret; 1676 1677 if (!inode_owner_or_capable(inode)) 1678 return -EACCES; 1679 1680 if (!S_ISREG(inode->i_mode)) 1681 return -EINVAL; 1682 1683 ret = mnt_want_write_file(filp); 1684 if (ret) 1685 return ret; 1686 1687 inode_lock(inode); 1688 1689 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1690 1691 if (f2fs_is_atomic_file(inode)) 1692 goto out; 1693 1694 ret = f2fs_convert_inline_inode(inode); 1695 if (ret) 1696 goto out; 1697 1698 if (!get_dirty_pages(inode)) 1699 goto skip_flush; 1700 1701 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING, 1702 "Unexpected flush for atomic writes: ino=%lu, npages=%u", 1703 inode->i_ino, get_dirty_pages(inode)); 1704 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 1705 if (ret) 1706 goto out; 1707 skip_flush: 1708 set_inode_flag(inode, FI_ATOMIC_FILE); 1709 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1710 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1711 1712 F2FS_I(inode)->inmem_task = current; 1713 stat_inc_atomic_write(inode); 1714 stat_update_max_atomic_write(inode); 1715 out: 1716 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1717 inode_unlock(inode); 1718 mnt_drop_write_file(filp); 1719 return ret; 1720 } 1721 1722 static int f2fs_ioc_commit_atomic_write(struct file *filp) 1723 { 1724 struct inode *inode = file_inode(filp); 1725 int ret; 1726 1727 if (!inode_owner_or_capable(inode)) 1728 return -EACCES; 1729 1730 ret = mnt_want_write_file(filp); 1731 if (ret) 1732 return ret; 1733 1734 inode_lock(inode); 1735 1736 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1737 1738 if (f2fs_is_volatile_file(inode)) { 1739 ret = -EINVAL; 1740 goto err_out; 1741 } 1742 1743 if (f2fs_is_atomic_file(inode)) { 1744 ret = f2fs_commit_inmem_pages(inode); 1745 if (ret) 1746 goto err_out; 1747 1748 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1749 if (!ret) { 1750 clear_inode_flag(inode, FI_ATOMIC_FILE); 1751 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0; 1752 stat_dec_atomic_write(inode); 1753 } 1754 } else { 1755 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 1756 } 1757 err_out: 1758 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) { 1759 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1760 ret = -EINVAL; 1761 } 1762 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1763 inode_unlock(inode); 1764 mnt_drop_write_file(filp); 1765 return ret; 1766 } 1767 1768 static int f2fs_ioc_start_volatile_write(struct file *filp) 1769 { 1770 struct inode *inode = file_inode(filp); 1771 int ret; 1772 1773 if (!inode_owner_or_capable(inode)) 1774 return -EACCES; 1775 1776 if (!S_ISREG(inode->i_mode)) 1777 return -EINVAL; 1778 1779 ret = mnt_want_write_file(filp); 1780 if (ret) 1781 return ret; 1782 1783 inode_lock(inode); 1784 1785 if (f2fs_is_volatile_file(inode)) 1786 goto out; 1787 1788 ret = f2fs_convert_inline_inode(inode); 1789 if (ret) 1790 goto out; 1791 1792 stat_inc_volatile_write(inode); 1793 stat_update_max_volatile_write(inode); 1794 1795 set_inode_flag(inode, FI_VOLATILE_FILE); 1796 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1797 out: 1798 inode_unlock(inode); 1799 mnt_drop_write_file(filp); 1800 return ret; 1801 } 1802 1803 static int f2fs_ioc_release_volatile_write(struct file *filp) 1804 { 1805 struct inode *inode = file_inode(filp); 1806 int ret; 1807 1808 if (!inode_owner_or_capable(inode)) 1809 return -EACCES; 1810 1811 ret = mnt_want_write_file(filp); 1812 if (ret) 1813 return ret; 1814 1815 inode_lock(inode); 1816 1817 if (!f2fs_is_volatile_file(inode)) 1818 goto out; 1819 1820 if (!f2fs_is_first_block_written(inode)) { 1821 ret = truncate_partial_data_page(inode, 0, true); 1822 goto out; 1823 } 1824 1825 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 1826 out: 1827 inode_unlock(inode); 1828 mnt_drop_write_file(filp); 1829 return ret; 1830 } 1831 1832 static int f2fs_ioc_abort_volatile_write(struct file *filp) 1833 { 1834 struct inode *inode = file_inode(filp); 1835 int ret; 1836 1837 if (!inode_owner_or_capable(inode)) 1838 return -EACCES; 1839 1840 ret = mnt_want_write_file(filp); 1841 if (ret) 1842 return ret; 1843 1844 inode_lock(inode); 1845 1846 if (f2fs_is_atomic_file(inode)) 1847 f2fs_drop_inmem_pages(inode); 1848 if (f2fs_is_volatile_file(inode)) { 1849 clear_inode_flag(inode, FI_VOLATILE_FILE); 1850 stat_dec_volatile_write(inode); 1851 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1852 } 1853 1854 inode_unlock(inode); 1855 1856 mnt_drop_write_file(filp); 1857 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1858 return ret; 1859 } 1860 1861 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 1862 { 1863 struct inode *inode = file_inode(filp); 1864 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1865 struct super_block *sb = sbi->sb; 1866 __u32 in; 1867 int ret; 1868 1869 if (!capable(CAP_SYS_ADMIN)) 1870 return -EPERM; 1871 1872 if (get_user(in, (__u32 __user *)arg)) 1873 return -EFAULT; 1874 1875 if (in != F2FS_GOING_DOWN_FULLSYNC) { 1876 ret = mnt_want_write_file(filp); 1877 if (ret) 1878 return ret; 1879 } 1880 1881 switch (in) { 1882 case F2FS_GOING_DOWN_FULLSYNC: 1883 sb = freeze_bdev(sb->s_bdev); 1884 if (IS_ERR(sb)) { 1885 ret = PTR_ERR(sb); 1886 goto out; 1887 } 1888 if (sb) { 1889 f2fs_stop_checkpoint(sbi, false); 1890 thaw_bdev(sb->s_bdev, sb); 1891 } 1892 break; 1893 case F2FS_GOING_DOWN_METASYNC: 1894 /* do checkpoint only */ 1895 ret = f2fs_sync_fs(sb, 1); 1896 if (ret) 1897 goto out; 1898 f2fs_stop_checkpoint(sbi, false); 1899 break; 1900 case F2FS_GOING_DOWN_NOSYNC: 1901 f2fs_stop_checkpoint(sbi, false); 1902 break; 1903 case F2FS_GOING_DOWN_METAFLUSH: 1904 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 1905 f2fs_stop_checkpoint(sbi, false); 1906 break; 1907 default: 1908 ret = -EINVAL; 1909 goto out; 1910 } 1911 1912 f2fs_stop_gc_thread(sbi); 1913 f2fs_stop_discard_thread(sbi); 1914 1915 f2fs_drop_discard_cmd(sbi); 1916 clear_opt(sbi, DISCARD); 1917 1918 f2fs_update_time(sbi, REQ_TIME); 1919 out: 1920 if (in != F2FS_GOING_DOWN_FULLSYNC) 1921 mnt_drop_write_file(filp); 1922 return ret; 1923 } 1924 1925 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 1926 { 1927 struct inode *inode = file_inode(filp); 1928 struct super_block *sb = inode->i_sb; 1929 struct request_queue *q = bdev_get_queue(sb->s_bdev); 1930 struct fstrim_range range; 1931 int ret; 1932 1933 if (!capable(CAP_SYS_ADMIN)) 1934 return -EPERM; 1935 1936 if (!blk_queue_discard(q)) 1937 return -EOPNOTSUPP; 1938 1939 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 1940 sizeof(range))) 1941 return -EFAULT; 1942 1943 ret = mnt_want_write_file(filp); 1944 if (ret) 1945 return ret; 1946 1947 range.minlen = max((unsigned int)range.minlen, 1948 q->limits.discard_granularity); 1949 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 1950 mnt_drop_write_file(filp); 1951 if (ret < 0) 1952 return ret; 1953 1954 if (copy_to_user((struct fstrim_range __user *)arg, &range, 1955 sizeof(range))) 1956 return -EFAULT; 1957 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1958 return 0; 1959 } 1960 1961 static bool uuid_is_nonzero(__u8 u[16]) 1962 { 1963 int i; 1964 1965 for (i = 0; i < 16; i++) 1966 if (u[i]) 1967 return true; 1968 return false; 1969 } 1970 1971 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 1972 { 1973 struct inode *inode = file_inode(filp); 1974 1975 if (!f2fs_sb_has_encrypt(inode->i_sb)) 1976 return -EOPNOTSUPP; 1977 1978 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1979 1980 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 1981 } 1982 1983 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 1984 { 1985 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb)) 1986 return -EOPNOTSUPP; 1987 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 1988 } 1989 1990 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 1991 { 1992 struct inode *inode = file_inode(filp); 1993 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1994 int err; 1995 1996 if (!f2fs_sb_has_encrypt(inode->i_sb)) 1997 return -EOPNOTSUPP; 1998 1999 err = mnt_want_write_file(filp); 2000 if (err) 2001 return err; 2002 2003 down_write(&sbi->sb_lock); 2004 2005 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2006 goto got_it; 2007 2008 /* update superblock with uuid */ 2009 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2010 2011 err = f2fs_commit_super(sbi, false); 2012 if (err) { 2013 /* undo new data */ 2014 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2015 goto out_err; 2016 } 2017 got_it: 2018 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 2019 16)) 2020 err = -EFAULT; 2021 out_err: 2022 up_write(&sbi->sb_lock); 2023 mnt_drop_write_file(filp); 2024 return err; 2025 } 2026 2027 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2028 { 2029 struct inode *inode = file_inode(filp); 2030 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2031 __u32 sync; 2032 int ret; 2033 2034 if (!capable(CAP_SYS_ADMIN)) 2035 return -EPERM; 2036 2037 if (get_user(sync, (__u32 __user *)arg)) 2038 return -EFAULT; 2039 2040 if (f2fs_readonly(sbi->sb)) 2041 return -EROFS; 2042 2043 ret = mnt_want_write_file(filp); 2044 if (ret) 2045 return ret; 2046 2047 if (!sync) { 2048 if (!mutex_trylock(&sbi->gc_mutex)) { 2049 ret = -EBUSY; 2050 goto out; 2051 } 2052 } else { 2053 mutex_lock(&sbi->gc_mutex); 2054 } 2055 2056 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO); 2057 out: 2058 mnt_drop_write_file(filp); 2059 return ret; 2060 } 2061 2062 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2063 { 2064 struct inode *inode = file_inode(filp); 2065 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2066 struct f2fs_gc_range range; 2067 u64 end; 2068 int ret; 2069 2070 if (!capable(CAP_SYS_ADMIN)) 2071 return -EPERM; 2072 2073 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2074 sizeof(range))) 2075 return -EFAULT; 2076 2077 if (f2fs_readonly(sbi->sb)) 2078 return -EROFS; 2079 2080 end = range.start + range.len; 2081 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) { 2082 return -EINVAL; 2083 } 2084 2085 ret = mnt_want_write_file(filp); 2086 if (ret) 2087 return ret; 2088 2089 do_more: 2090 if (!range.sync) { 2091 if (!mutex_trylock(&sbi->gc_mutex)) { 2092 ret = -EBUSY; 2093 goto out; 2094 } 2095 } else { 2096 mutex_lock(&sbi->gc_mutex); 2097 } 2098 2099 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start)); 2100 range.start += sbi->blocks_per_seg; 2101 if (range.start <= end) 2102 goto do_more; 2103 out: 2104 mnt_drop_write_file(filp); 2105 return ret; 2106 } 2107 2108 static int f2fs_ioc_f2fs_write_checkpoint(struct file *filp, unsigned long arg) 2109 { 2110 struct inode *inode = file_inode(filp); 2111 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2112 int ret; 2113 2114 if (!capable(CAP_SYS_ADMIN)) 2115 return -EPERM; 2116 2117 if (f2fs_readonly(sbi->sb)) 2118 return -EROFS; 2119 2120 ret = mnt_want_write_file(filp); 2121 if (ret) 2122 return ret; 2123 2124 ret = f2fs_sync_fs(sbi->sb, 1); 2125 2126 mnt_drop_write_file(filp); 2127 return ret; 2128 } 2129 2130 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2131 struct file *filp, 2132 struct f2fs_defragment *range) 2133 { 2134 struct inode *inode = file_inode(filp); 2135 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2136 .m_seg_type = NO_CHECK_TYPE }; 2137 struct extent_info ei = {0, 0, 0}; 2138 pgoff_t pg_start, pg_end, next_pgofs; 2139 unsigned int blk_per_seg = sbi->blocks_per_seg; 2140 unsigned int total = 0, sec_num; 2141 block_t blk_end = 0; 2142 bool fragmented = false; 2143 int err; 2144 2145 /* if in-place-update policy is enabled, don't waste time here */ 2146 if (f2fs_should_update_inplace(inode, NULL)) 2147 return -EINVAL; 2148 2149 pg_start = range->start >> PAGE_SHIFT; 2150 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2151 2152 f2fs_balance_fs(sbi, true); 2153 2154 inode_lock(inode); 2155 2156 /* writeback all dirty pages in the range */ 2157 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2158 range->start + range->len - 1); 2159 if (err) 2160 goto out; 2161 2162 /* 2163 * lookup mapping info in extent cache, skip defragmenting if physical 2164 * block addresses are continuous. 2165 */ 2166 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) { 2167 if (ei.fofs + ei.len >= pg_end) 2168 goto out; 2169 } 2170 2171 map.m_lblk = pg_start; 2172 map.m_next_pgofs = &next_pgofs; 2173 2174 /* 2175 * lookup mapping info in dnode page cache, skip defragmenting if all 2176 * physical block addresses are continuous even if there are hole(s) 2177 * in logical blocks. 2178 */ 2179 while (map.m_lblk < pg_end) { 2180 map.m_len = pg_end - map.m_lblk; 2181 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2182 if (err) 2183 goto out; 2184 2185 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2186 map.m_lblk = next_pgofs; 2187 continue; 2188 } 2189 2190 if (blk_end && blk_end != map.m_pblk) 2191 fragmented = true; 2192 2193 /* record total count of block that we're going to move */ 2194 total += map.m_len; 2195 2196 blk_end = map.m_pblk + map.m_len; 2197 2198 map.m_lblk += map.m_len; 2199 } 2200 2201 if (!fragmented) 2202 goto out; 2203 2204 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi); 2205 2206 /* 2207 * make sure there are enough free section for LFS allocation, this can 2208 * avoid defragment running in SSR mode when free section are allocated 2209 * intensively 2210 */ 2211 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2212 err = -EAGAIN; 2213 goto out; 2214 } 2215 2216 map.m_lblk = pg_start; 2217 map.m_len = pg_end - pg_start; 2218 total = 0; 2219 2220 while (map.m_lblk < pg_end) { 2221 pgoff_t idx; 2222 int cnt = 0; 2223 2224 do_map: 2225 map.m_len = pg_end - map.m_lblk; 2226 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2227 if (err) 2228 goto clear_out; 2229 2230 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2231 map.m_lblk = next_pgofs; 2232 continue; 2233 } 2234 2235 set_inode_flag(inode, FI_DO_DEFRAG); 2236 2237 idx = map.m_lblk; 2238 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2239 struct page *page; 2240 2241 page = f2fs_get_lock_data_page(inode, idx, true); 2242 if (IS_ERR(page)) { 2243 err = PTR_ERR(page); 2244 goto clear_out; 2245 } 2246 2247 set_page_dirty(page); 2248 f2fs_put_page(page, 1); 2249 2250 idx++; 2251 cnt++; 2252 total++; 2253 } 2254 2255 map.m_lblk = idx; 2256 2257 if (idx < pg_end && cnt < blk_per_seg) 2258 goto do_map; 2259 2260 clear_inode_flag(inode, FI_DO_DEFRAG); 2261 2262 err = filemap_fdatawrite(inode->i_mapping); 2263 if (err) 2264 goto out; 2265 } 2266 clear_out: 2267 clear_inode_flag(inode, FI_DO_DEFRAG); 2268 out: 2269 inode_unlock(inode); 2270 if (!err) 2271 range->len = (u64)total << PAGE_SHIFT; 2272 return err; 2273 } 2274 2275 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2276 { 2277 struct inode *inode = file_inode(filp); 2278 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2279 struct f2fs_defragment range; 2280 int err; 2281 2282 if (!capable(CAP_SYS_ADMIN)) 2283 return -EPERM; 2284 2285 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2286 return -EINVAL; 2287 2288 if (f2fs_readonly(sbi->sb)) 2289 return -EROFS; 2290 2291 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2292 sizeof(range))) 2293 return -EFAULT; 2294 2295 /* verify alignment of offset & size */ 2296 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2297 return -EINVAL; 2298 2299 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2300 sbi->max_file_blocks)) 2301 return -EINVAL; 2302 2303 err = mnt_want_write_file(filp); 2304 if (err) 2305 return err; 2306 2307 err = f2fs_defragment_range(sbi, filp, &range); 2308 mnt_drop_write_file(filp); 2309 2310 f2fs_update_time(sbi, REQ_TIME); 2311 if (err < 0) 2312 return err; 2313 2314 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2315 sizeof(range))) 2316 return -EFAULT; 2317 2318 return 0; 2319 } 2320 2321 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2322 struct file *file_out, loff_t pos_out, size_t len) 2323 { 2324 struct inode *src = file_inode(file_in); 2325 struct inode *dst = file_inode(file_out); 2326 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2327 size_t olen = len, dst_max_i_size = 0; 2328 size_t dst_osize; 2329 int ret; 2330 2331 if (file_in->f_path.mnt != file_out->f_path.mnt || 2332 src->i_sb != dst->i_sb) 2333 return -EXDEV; 2334 2335 if (unlikely(f2fs_readonly(src->i_sb))) 2336 return -EROFS; 2337 2338 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2339 return -EINVAL; 2340 2341 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst)) 2342 return -EOPNOTSUPP; 2343 2344 if (src == dst) { 2345 if (pos_in == pos_out) 2346 return 0; 2347 if (pos_out > pos_in && pos_out < pos_in + len) 2348 return -EINVAL; 2349 } 2350 2351 inode_lock(src); 2352 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2353 if (src != dst) { 2354 ret = -EBUSY; 2355 if (!inode_trylock(dst)) 2356 goto out; 2357 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) { 2358 inode_unlock(dst); 2359 goto out; 2360 } 2361 } 2362 2363 ret = -EINVAL; 2364 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2365 goto out_unlock; 2366 if (len == 0) 2367 olen = len = src->i_size - pos_in; 2368 if (pos_in + len == src->i_size) 2369 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2370 if (len == 0) { 2371 ret = 0; 2372 goto out_unlock; 2373 } 2374 2375 dst_osize = dst->i_size; 2376 if (pos_out + olen > dst->i_size) 2377 dst_max_i_size = pos_out + olen; 2378 2379 /* verify the end result is block aligned */ 2380 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2381 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2382 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2383 goto out_unlock; 2384 2385 ret = f2fs_convert_inline_inode(src); 2386 if (ret) 2387 goto out_unlock; 2388 2389 ret = f2fs_convert_inline_inode(dst); 2390 if (ret) 2391 goto out_unlock; 2392 2393 /* write out all dirty pages from offset */ 2394 ret = filemap_write_and_wait_range(src->i_mapping, 2395 pos_in, pos_in + len); 2396 if (ret) 2397 goto out_unlock; 2398 2399 ret = filemap_write_and_wait_range(dst->i_mapping, 2400 pos_out, pos_out + len); 2401 if (ret) 2402 goto out_unlock; 2403 2404 f2fs_balance_fs(sbi, true); 2405 f2fs_lock_op(sbi); 2406 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2407 pos_out >> F2FS_BLKSIZE_BITS, 2408 len >> F2FS_BLKSIZE_BITS, false); 2409 2410 if (!ret) { 2411 if (dst_max_i_size) 2412 f2fs_i_size_write(dst, dst_max_i_size); 2413 else if (dst_osize != dst->i_size) 2414 f2fs_i_size_write(dst, dst_osize); 2415 } 2416 f2fs_unlock_op(sbi); 2417 out_unlock: 2418 if (src != dst) { 2419 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2420 inode_unlock(dst); 2421 } 2422 out: 2423 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2424 inode_unlock(src); 2425 return ret; 2426 } 2427 2428 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2429 { 2430 struct f2fs_move_range range; 2431 struct fd dst; 2432 int err; 2433 2434 if (!(filp->f_mode & FMODE_READ) || 2435 !(filp->f_mode & FMODE_WRITE)) 2436 return -EBADF; 2437 2438 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2439 sizeof(range))) 2440 return -EFAULT; 2441 2442 dst = fdget(range.dst_fd); 2443 if (!dst.file) 2444 return -EBADF; 2445 2446 if (!(dst.file->f_mode & FMODE_WRITE)) { 2447 err = -EBADF; 2448 goto err_out; 2449 } 2450 2451 err = mnt_want_write_file(filp); 2452 if (err) 2453 goto err_out; 2454 2455 err = f2fs_move_file_range(filp, range.pos_in, dst.file, 2456 range.pos_out, range.len); 2457 2458 mnt_drop_write_file(filp); 2459 if (err) 2460 goto err_out; 2461 2462 if (copy_to_user((struct f2fs_move_range __user *)arg, 2463 &range, sizeof(range))) 2464 err = -EFAULT; 2465 err_out: 2466 fdput(dst); 2467 return err; 2468 } 2469 2470 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2471 { 2472 struct inode *inode = file_inode(filp); 2473 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2474 struct sit_info *sm = SIT_I(sbi); 2475 unsigned int start_segno = 0, end_segno = 0; 2476 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2477 struct f2fs_flush_device range; 2478 int ret; 2479 2480 if (!capable(CAP_SYS_ADMIN)) 2481 return -EPERM; 2482 2483 if (f2fs_readonly(sbi->sb)) 2484 return -EROFS; 2485 2486 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2487 sizeof(range))) 2488 return -EFAULT; 2489 2490 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num || 2491 sbi->segs_per_sec != 1) { 2492 f2fs_msg(sbi->sb, KERN_WARNING, 2493 "Can't flush %u in %d for segs_per_sec %u != 1\n", 2494 range.dev_num, sbi->s_ndevs, 2495 sbi->segs_per_sec); 2496 return -EINVAL; 2497 } 2498 2499 ret = mnt_want_write_file(filp); 2500 if (ret) 2501 return ret; 2502 2503 if (range.dev_num != 0) 2504 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2505 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2506 2507 start_segno = sm->last_victim[FLUSH_DEVICE]; 2508 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2509 start_segno = dev_start_segno; 2510 end_segno = min(start_segno + range.segments, dev_end_segno); 2511 2512 while (start_segno < end_segno) { 2513 if (!mutex_trylock(&sbi->gc_mutex)) { 2514 ret = -EBUSY; 2515 goto out; 2516 } 2517 sm->last_victim[GC_CB] = end_segno + 1; 2518 sm->last_victim[GC_GREEDY] = end_segno + 1; 2519 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 2520 ret = f2fs_gc(sbi, true, true, start_segno); 2521 if (ret == -EAGAIN) 2522 ret = 0; 2523 else if (ret < 0) 2524 break; 2525 start_segno++; 2526 } 2527 out: 2528 mnt_drop_write_file(filp); 2529 return ret; 2530 } 2531 2532 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 2533 { 2534 struct inode *inode = file_inode(filp); 2535 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 2536 2537 /* Must validate to set it with SQLite behavior in Android. */ 2538 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 2539 2540 return put_user(sb_feature, (u32 __user *)arg); 2541 } 2542 2543 #ifdef CONFIG_QUOTA 2544 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 2545 { 2546 struct inode *inode = file_inode(filp); 2547 struct f2fs_inode_info *fi = F2FS_I(inode); 2548 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2549 struct super_block *sb = sbi->sb; 2550 struct dquot *transfer_to[MAXQUOTAS] = {}; 2551 struct page *ipage; 2552 kprojid_t kprojid; 2553 int err; 2554 2555 if (!f2fs_sb_has_project_quota(sb)) { 2556 if (projid != F2FS_DEF_PROJID) 2557 return -EOPNOTSUPP; 2558 else 2559 return 0; 2560 } 2561 2562 if (!f2fs_has_extra_attr(inode)) 2563 return -EOPNOTSUPP; 2564 2565 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 2566 2567 if (projid_eq(kprojid, F2FS_I(inode)->i_projid)) 2568 return 0; 2569 2570 err = mnt_want_write_file(filp); 2571 if (err) 2572 return err; 2573 2574 err = -EPERM; 2575 inode_lock(inode); 2576 2577 /* Is it quota file? Do not allow user to mess with it */ 2578 if (IS_NOQUOTA(inode)) 2579 goto out_unlock; 2580 2581 ipage = f2fs_get_node_page(sbi, inode->i_ino); 2582 if (IS_ERR(ipage)) { 2583 err = PTR_ERR(ipage); 2584 goto out_unlock; 2585 } 2586 2587 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize, 2588 i_projid)) { 2589 err = -EOVERFLOW; 2590 f2fs_put_page(ipage, 1); 2591 goto out_unlock; 2592 } 2593 f2fs_put_page(ipage, 1); 2594 2595 err = dquot_initialize(inode); 2596 if (err) 2597 goto out_unlock; 2598 2599 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 2600 if (!IS_ERR(transfer_to[PRJQUOTA])) { 2601 err = __dquot_transfer(inode, transfer_to); 2602 dqput(transfer_to[PRJQUOTA]); 2603 if (err) 2604 goto out_dirty; 2605 } 2606 2607 F2FS_I(inode)->i_projid = kprojid; 2608 inode->i_ctime = current_time(inode); 2609 out_dirty: 2610 f2fs_mark_inode_dirty_sync(inode, true); 2611 out_unlock: 2612 inode_unlock(inode); 2613 mnt_drop_write_file(filp); 2614 return err; 2615 } 2616 #else 2617 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 2618 { 2619 if (projid != F2FS_DEF_PROJID) 2620 return -EOPNOTSUPP; 2621 return 0; 2622 } 2623 #endif 2624 2625 /* Transfer internal flags to xflags */ 2626 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags) 2627 { 2628 __u32 xflags = 0; 2629 2630 if (iflags & F2FS_SYNC_FL) 2631 xflags |= FS_XFLAG_SYNC; 2632 if (iflags & F2FS_IMMUTABLE_FL) 2633 xflags |= FS_XFLAG_IMMUTABLE; 2634 if (iflags & F2FS_APPEND_FL) 2635 xflags |= FS_XFLAG_APPEND; 2636 if (iflags & F2FS_NODUMP_FL) 2637 xflags |= FS_XFLAG_NODUMP; 2638 if (iflags & F2FS_NOATIME_FL) 2639 xflags |= FS_XFLAG_NOATIME; 2640 if (iflags & F2FS_PROJINHERIT_FL) 2641 xflags |= FS_XFLAG_PROJINHERIT; 2642 return xflags; 2643 } 2644 2645 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \ 2646 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \ 2647 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT) 2648 2649 /* Transfer xflags flags to internal */ 2650 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags) 2651 { 2652 unsigned long iflags = 0; 2653 2654 if (xflags & FS_XFLAG_SYNC) 2655 iflags |= F2FS_SYNC_FL; 2656 if (xflags & FS_XFLAG_IMMUTABLE) 2657 iflags |= F2FS_IMMUTABLE_FL; 2658 if (xflags & FS_XFLAG_APPEND) 2659 iflags |= F2FS_APPEND_FL; 2660 if (xflags & FS_XFLAG_NODUMP) 2661 iflags |= F2FS_NODUMP_FL; 2662 if (xflags & FS_XFLAG_NOATIME) 2663 iflags |= F2FS_NOATIME_FL; 2664 if (xflags & FS_XFLAG_PROJINHERIT) 2665 iflags |= F2FS_PROJINHERIT_FL; 2666 2667 return iflags; 2668 } 2669 2670 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg) 2671 { 2672 struct inode *inode = file_inode(filp); 2673 struct f2fs_inode_info *fi = F2FS_I(inode); 2674 struct fsxattr fa; 2675 2676 memset(&fa, 0, sizeof(struct fsxattr)); 2677 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags & 2678 F2FS_FL_USER_VISIBLE); 2679 2680 if (f2fs_sb_has_project_quota(inode->i_sb)) 2681 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns, 2682 fi->i_projid); 2683 2684 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa))) 2685 return -EFAULT; 2686 return 0; 2687 } 2688 2689 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg) 2690 { 2691 struct inode *inode = file_inode(filp); 2692 struct f2fs_inode_info *fi = F2FS_I(inode); 2693 struct fsxattr fa; 2694 unsigned int flags; 2695 int err; 2696 2697 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa))) 2698 return -EFAULT; 2699 2700 /* Make sure caller has proper permission */ 2701 if (!inode_owner_or_capable(inode)) 2702 return -EACCES; 2703 2704 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS) 2705 return -EOPNOTSUPP; 2706 2707 flags = f2fs_xflags_to_iflags(fa.fsx_xflags); 2708 if (f2fs_mask_flags(inode->i_mode, flags) != flags) 2709 return -EOPNOTSUPP; 2710 2711 err = mnt_want_write_file(filp); 2712 if (err) 2713 return err; 2714 2715 inode_lock(inode); 2716 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) | 2717 (flags & F2FS_FL_XFLAG_VISIBLE); 2718 err = __f2fs_ioc_setflags(inode, flags); 2719 inode_unlock(inode); 2720 mnt_drop_write_file(filp); 2721 if (err) 2722 return err; 2723 2724 err = f2fs_ioc_setproject(filp, fa.fsx_projid); 2725 if (err) 2726 return err; 2727 2728 return 0; 2729 } 2730 2731 int f2fs_pin_file_control(struct inode *inode, bool inc) 2732 { 2733 struct f2fs_inode_info *fi = F2FS_I(inode); 2734 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2735 2736 /* Use i_gc_failures for normal file as a risk signal. */ 2737 if (inc) 2738 f2fs_i_gc_failures_write(inode, 2739 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 2740 2741 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 2742 f2fs_msg(sbi->sb, KERN_WARNING, 2743 "%s: Enable GC = ino %lx after %x GC trials\n", 2744 __func__, inode->i_ino, 2745 fi->i_gc_failures[GC_FAILURE_PIN]); 2746 clear_inode_flag(inode, FI_PIN_FILE); 2747 return -EAGAIN; 2748 } 2749 return 0; 2750 } 2751 2752 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 2753 { 2754 struct inode *inode = file_inode(filp); 2755 __u32 pin; 2756 int ret = 0; 2757 2758 if (!inode_owner_or_capable(inode)) 2759 return -EACCES; 2760 2761 if (get_user(pin, (__u32 __user *)arg)) 2762 return -EFAULT; 2763 2764 if (!S_ISREG(inode->i_mode)) 2765 return -EINVAL; 2766 2767 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 2768 return -EROFS; 2769 2770 ret = mnt_want_write_file(filp); 2771 if (ret) 2772 return ret; 2773 2774 inode_lock(inode); 2775 2776 if (f2fs_should_update_outplace(inode, NULL)) { 2777 ret = -EINVAL; 2778 goto out; 2779 } 2780 2781 if (!pin) { 2782 clear_inode_flag(inode, FI_PIN_FILE); 2783 F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN] = 1; 2784 goto done; 2785 } 2786 2787 if (f2fs_pin_file_control(inode, false)) { 2788 ret = -EAGAIN; 2789 goto out; 2790 } 2791 ret = f2fs_convert_inline_inode(inode); 2792 if (ret) 2793 goto out; 2794 2795 set_inode_flag(inode, FI_PIN_FILE); 2796 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 2797 done: 2798 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2799 out: 2800 inode_unlock(inode); 2801 mnt_drop_write_file(filp); 2802 return ret; 2803 } 2804 2805 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 2806 { 2807 struct inode *inode = file_inode(filp); 2808 __u32 pin = 0; 2809 2810 if (is_inode_flag_set(inode, FI_PIN_FILE)) 2811 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 2812 return put_user(pin, (u32 __user *)arg); 2813 } 2814 2815 int f2fs_precache_extents(struct inode *inode) 2816 { 2817 struct f2fs_inode_info *fi = F2FS_I(inode); 2818 struct f2fs_map_blocks map; 2819 pgoff_t m_next_extent; 2820 loff_t end; 2821 int err; 2822 2823 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 2824 return -EOPNOTSUPP; 2825 2826 map.m_lblk = 0; 2827 map.m_next_pgofs = NULL; 2828 map.m_next_extent = &m_next_extent; 2829 map.m_seg_type = NO_CHECK_TYPE; 2830 end = F2FS_I_SB(inode)->max_file_blocks; 2831 2832 while (map.m_lblk < end) { 2833 map.m_len = end - map.m_lblk; 2834 2835 down_write(&fi->i_gc_rwsem[WRITE]); 2836 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE); 2837 up_write(&fi->i_gc_rwsem[WRITE]); 2838 if (err) 2839 return err; 2840 2841 map.m_lblk = m_next_extent; 2842 } 2843 2844 return err; 2845 } 2846 2847 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg) 2848 { 2849 return f2fs_precache_extents(file_inode(filp)); 2850 } 2851 2852 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 2853 { 2854 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 2855 return -EIO; 2856 2857 switch (cmd) { 2858 case F2FS_IOC_GETFLAGS: 2859 return f2fs_ioc_getflags(filp, arg); 2860 case F2FS_IOC_SETFLAGS: 2861 return f2fs_ioc_setflags(filp, arg); 2862 case F2FS_IOC_GETVERSION: 2863 return f2fs_ioc_getversion(filp, arg); 2864 case F2FS_IOC_START_ATOMIC_WRITE: 2865 return f2fs_ioc_start_atomic_write(filp); 2866 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 2867 return f2fs_ioc_commit_atomic_write(filp); 2868 case F2FS_IOC_START_VOLATILE_WRITE: 2869 return f2fs_ioc_start_volatile_write(filp); 2870 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 2871 return f2fs_ioc_release_volatile_write(filp); 2872 case F2FS_IOC_ABORT_VOLATILE_WRITE: 2873 return f2fs_ioc_abort_volatile_write(filp); 2874 case F2FS_IOC_SHUTDOWN: 2875 return f2fs_ioc_shutdown(filp, arg); 2876 case FITRIM: 2877 return f2fs_ioc_fitrim(filp, arg); 2878 case F2FS_IOC_SET_ENCRYPTION_POLICY: 2879 return f2fs_ioc_set_encryption_policy(filp, arg); 2880 case F2FS_IOC_GET_ENCRYPTION_POLICY: 2881 return f2fs_ioc_get_encryption_policy(filp, arg); 2882 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 2883 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 2884 case F2FS_IOC_GARBAGE_COLLECT: 2885 return f2fs_ioc_gc(filp, arg); 2886 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 2887 return f2fs_ioc_gc_range(filp, arg); 2888 case F2FS_IOC_WRITE_CHECKPOINT: 2889 return f2fs_ioc_f2fs_write_checkpoint(filp, arg); 2890 case F2FS_IOC_DEFRAGMENT: 2891 return f2fs_ioc_defragment(filp, arg); 2892 case F2FS_IOC_MOVE_RANGE: 2893 return f2fs_ioc_move_range(filp, arg); 2894 case F2FS_IOC_FLUSH_DEVICE: 2895 return f2fs_ioc_flush_device(filp, arg); 2896 case F2FS_IOC_GET_FEATURES: 2897 return f2fs_ioc_get_features(filp, arg); 2898 case F2FS_IOC_FSGETXATTR: 2899 return f2fs_ioc_fsgetxattr(filp, arg); 2900 case F2FS_IOC_FSSETXATTR: 2901 return f2fs_ioc_fssetxattr(filp, arg); 2902 case F2FS_IOC_GET_PIN_FILE: 2903 return f2fs_ioc_get_pin_file(filp, arg); 2904 case F2FS_IOC_SET_PIN_FILE: 2905 return f2fs_ioc_set_pin_file(filp, arg); 2906 case F2FS_IOC_PRECACHE_EXTENTS: 2907 return f2fs_ioc_precache_extents(filp, arg); 2908 default: 2909 return -ENOTTY; 2910 } 2911 } 2912 2913 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 2914 { 2915 struct file *file = iocb->ki_filp; 2916 struct inode *inode = file_inode(file); 2917 ssize_t ret; 2918 2919 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 2920 return -EIO; 2921 2922 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) 2923 return -EINVAL; 2924 2925 if (!inode_trylock(inode)) { 2926 if (iocb->ki_flags & IOCB_NOWAIT) 2927 return -EAGAIN; 2928 inode_lock(inode); 2929 } 2930 2931 ret = generic_write_checks(iocb, from); 2932 if (ret > 0) { 2933 bool preallocated = false; 2934 size_t target_size = 0; 2935 int err; 2936 2937 if (iov_iter_fault_in_readable(from, iov_iter_count(from))) 2938 set_inode_flag(inode, FI_NO_PREALLOC); 2939 2940 if ((iocb->ki_flags & IOCB_NOWAIT) && 2941 (iocb->ki_flags & IOCB_DIRECT)) { 2942 if (!f2fs_overwrite_io(inode, iocb->ki_pos, 2943 iov_iter_count(from)) || 2944 f2fs_has_inline_data(inode) || 2945 f2fs_force_buffered_io(inode, WRITE)) { 2946 clear_inode_flag(inode, 2947 FI_NO_PREALLOC); 2948 inode_unlock(inode); 2949 return -EAGAIN; 2950 } 2951 2952 } else { 2953 preallocated = true; 2954 target_size = iocb->ki_pos + iov_iter_count(from); 2955 2956 err = f2fs_preallocate_blocks(iocb, from); 2957 if (err) { 2958 clear_inode_flag(inode, FI_NO_PREALLOC); 2959 inode_unlock(inode); 2960 return err; 2961 } 2962 } 2963 ret = __generic_file_write_iter(iocb, from); 2964 clear_inode_flag(inode, FI_NO_PREALLOC); 2965 2966 /* if we couldn't write data, we should deallocate blocks. */ 2967 if (preallocated && i_size_read(inode) < target_size) 2968 f2fs_truncate(inode); 2969 2970 if (ret > 0) 2971 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret); 2972 } 2973 inode_unlock(inode); 2974 2975 if (ret > 0) 2976 ret = generic_write_sync(iocb, ret); 2977 return ret; 2978 } 2979 2980 #ifdef CONFIG_COMPAT 2981 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 2982 { 2983 switch (cmd) { 2984 case F2FS_IOC32_GETFLAGS: 2985 cmd = F2FS_IOC_GETFLAGS; 2986 break; 2987 case F2FS_IOC32_SETFLAGS: 2988 cmd = F2FS_IOC_SETFLAGS; 2989 break; 2990 case F2FS_IOC32_GETVERSION: 2991 cmd = F2FS_IOC_GETVERSION; 2992 break; 2993 case F2FS_IOC_START_ATOMIC_WRITE: 2994 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 2995 case F2FS_IOC_START_VOLATILE_WRITE: 2996 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 2997 case F2FS_IOC_ABORT_VOLATILE_WRITE: 2998 case F2FS_IOC_SHUTDOWN: 2999 case F2FS_IOC_SET_ENCRYPTION_POLICY: 3000 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 3001 case F2FS_IOC_GET_ENCRYPTION_POLICY: 3002 case F2FS_IOC_GARBAGE_COLLECT: 3003 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 3004 case F2FS_IOC_WRITE_CHECKPOINT: 3005 case F2FS_IOC_DEFRAGMENT: 3006 case F2FS_IOC_MOVE_RANGE: 3007 case F2FS_IOC_FLUSH_DEVICE: 3008 case F2FS_IOC_GET_FEATURES: 3009 case F2FS_IOC_FSGETXATTR: 3010 case F2FS_IOC_FSSETXATTR: 3011 case F2FS_IOC_GET_PIN_FILE: 3012 case F2FS_IOC_SET_PIN_FILE: 3013 case F2FS_IOC_PRECACHE_EXTENTS: 3014 break; 3015 default: 3016 return -ENOIOCTLCMD; 3017 } 3018 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 3019 } 3020 #endif 3021 3022 const struct file_operations f2fs_file_operations = { 3023 .llseek = f2fs_llseek, 3024 .read_iter = generic_file_read_iter, 3025 .write_iter = f2fs_file_write_iter, 3026 .open = f2fs_file_open, 3027 .release = f2fs_release_file, 3028 .mmap = f2fs_file_mmap, 3029 .flush = f2fs_file_flush, 3030 .fsync = f2fs_sync_file, 3031 .fallocate = f2fs_fallocate, 3032 .unlocked_ioctl = f2fs_ioctl, 3033 #ifdef CONFIG_COMPAT 3034 .compat_ioctl = f2fs_compat_ioctl, 3035 #endif 3036 .splice_read = generic_file_splice_read, 3037 .splice_write = iter_file_splice_write, 3038 }; 3039