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 24 #include "f2fs.h" 25 #include "node.h" 26 #include "segment.h" 27 #include "xattr.h" 28 #include "acl.h" 29 #include "gc.h" 30 #include "trace.h" 31 #include <trace/events/f2fs.h> 32 33 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 34 { 35 struct inode *inode = file_inode(vmf->vma->vm_file); 36 vm_fault_t ret; 37 38 down_read(&F2FS_I(inode)->i_mmap_sem); 39 ret = filemap_fault(vmf); 40 up_read(&F2FS_I(inode)->i_mmap_sem); 41 42 return ret; 43 } 44 45 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 46 { 47 struct page *page = vmf->page; 48 struct inode *inode = file_inode(vmf->vma->vm_file); 49 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 50 struct dnode_of_data dn = { .node_changed = false }; 51 int err; 52 53 if (unlikely(f2fs_cp_error(sbi))) { 54 err = -EIO; 55 goto err; 56 } 57 58 sb_start_pagefault(inode->i_sb); 59 60 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 61 62 file_update_time(vmf->vma->vm_file); 63 down_read(&F2FS_I(inode)->i_mmap_sem); 64 lock_page(page); 65 if (unlikely(page->mapping != inode->i_mapping || 66 page_offset(page) > i_size_read(inode) || 67 !PageUptodate(page))) { 68 unlock_page(page); 69 err = -EFAULT; 70 goto out_sem; 71 } 72 73 /* block allocation */ 74 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true); 75 set_new_dnode(&dn, inode, NULL, NULL, 0); 76 err = f2fs_get_block(&dn, page->index); 77 f2fs_put_dnode(&dn); 78 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false); 79 if (err) { 80 unlock_page(page); 81 goto out_sem; 82 } 83 84 /* fill the page */ 85 f2fs_wait_on_page_writeback(page, DATA, false); 86 87 /* wait for GCed page writeback via META_MAPPING */ 88 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 89 90 /* 91 * check to see if the page is mapped already (no holes) 92 */ 93 if (PageMappedToDisk(page)) 94 goto out_sem; 95 96 /* page is wholly or partially inside EOF */ 97 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 98 i_size_read(inode)) { 99 loff_t offset; 100 101 offset = i_size_read(inode) & ~PAGE_MASK; 102 zero_user_segment(page, offset, PAGE_SIZE); 103 } 104 set_page_dirty(page); 105 if (!PageUptodate(page)) 106 SetPageUptodate(page); 107 108 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE); 109 110 trace_f2fs_vm_page_mkwrite(page, DATA); 111 out_sem: 112 up_read(&F2FS_I(inode)->i_mmap_sem); 113 114 f2fs_balance_fs(sbi, dn.node_changed); 115 116 sb_end_pagefault(inode->i_sb); 117 f2fs_update_time(sbi, REQ_TIME); 118 err: 119 return block_page_mkwrite_return(err); 120 } 121 122 static const struct vm_operations_struct f2fs_file_vm_ops = { 123 .fault = f2fs_filemap_fault, 124 .map_pages = filemap_map_pages, 125 .page_mkwrite = f2fs_vm_page_mkwrite, 126 }; 127 128 static int get_parent_ino(struct inode *inode, nid_t *pino) 129 { 130 struct dentry *dentry; 131 132 inode = igrab(inode); 133 dentry = d_find_any_alias(inode); 134 iput(inode); 135 if (!dentry) 136 return 0; 137 138 *pino = parent_ino(dentry); 139 dput(dentry); 140 return 1; 141 } 142 143 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 144 { 145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 146 enum cp_reason_type cp_reason = CP_NO_NEEDED; 147 148 if (!S_ISREG(inode->i_mode)) 149 cp_reason = CP_NON_REGULAR; 150 else if (inode->i_nlink != 1) 151 cp_reason = CP_HARDLINK; 152 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 153 cp_reason = CP_SB_NEED_CP; 154 else if (file_wrong_pino(inode)) 155 cp_reason = CP_WRONG_PINO; 156 else if (!f2fs_space_for_roll_forward(sbi)) 157 cp_reason = CP_NO_SPC_ROLL; 158 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 159 cp_reason = CP_NODE_NEED_CP; 160 else if (test_opt(sbi, FASTBOOT)) 161 cp_reason = CP_FASTBOOT_MODE; 162 else if (F2FS_OPTION(sbi).active_logs == 2) 163 cp_reason = CP_SPEC_LOG_NUM; 164 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 165 f2fs_need_dentry_mark(sbi, inode->i_ino) && 166 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 167 TRANS_DIR_INO)) 168 cp_reason = CP_RECOVER_DIR; 169 170 return cp_reason; 171 } 172 173 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 174 { 175 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 176 bool ret = false; 177 /* But we need to avoid that there are some inode updates */ 178 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 179 ret = true; 180 f2fs_put_page(i, 0); 181 return ret; 182 } 183 184 static void try_to_fix_pino(struct inode *inode) 185 { 186 struct f2fs_inode_info *fi = F2FS_I(inode); 187 nid_t pino; 188 189 down_write(&fi->i_sem); 190 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 191 get_parent_ino(inode, &pino)) { 192 f2fs_i_pino_write(inode, pino); 193 file_got_pino(inode); 194 } 195 up_write(&fi->i_sem); 196 } 197 198 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 199 int datasync, bool atomic) 200 { 201 struct inode *inode = file->f_mapping->host; 202 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 203 nid_t ino = inode->i_ino; 204 int ret = 0; 205 enum cp_reason_type cp_reason = 0; 206 struct writeback_control wbc = { 207 .sync_mode = WB_SYNC_ALL, 208 .nr_to_write = LONG_MAX, 209 .for_reclaim = 0, 210 }; 211 unsigned int seq_id = 0; 212 213 if (unlikely(f2fs_readonly(inode->i_sb) || 214 is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 215 return 0; 216 217 trace_f2fs_sync_file_enter(inode); 218 219 /* if fdatasync is triggered, let's do in-place-update */ 220 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 221 set_inode_flag(inode, FI_NEED_IPU); 222 ret = file_write_and_wait_range(file, start, end); 223 clear_inode_flag(inode, FI_NEED_IPU); 224 225 if (ret) { 226 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 227 return ret; 228 } 229 230 /* if the inode is dirty, let's recover all the time */ 231 if (!f2fs_skip_inode_update(inode, datasync)) { 232 f2fs_write_inode(inode, NULL); 233 goto go_write; 234 } 235 236 /* 237 * if there is no written data, don't waste time to write recovery info. 238 */ 239 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 240 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 241 242 /* it may call write_inode just prior to fsync */ 243 if (need_inode_page_update(sbi, ino)) 244 goto go_write; 245 246 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 247 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 248 goto flush_out; 249 goto out; 250 } 251 go_write: 252 /* 253 * Both of fdatasync() and fsync() are able to be recovered from 254 * sudden-power-off. 255 */ 256 down_read(&F2FS_I(inode)->i_sem); 257 cp_reason = need_do_checkpoint(inode); 258 up_read(&F2FS_I(inode)->i_sem); 259 260 if (cp_reason) { 261 /* all the dirty node pages should be flushed for POR */ 262 ret = f2fs_sync_fs(inode->i_sb, 1); 263 264 /* 265 * We've secured consistency through sync_fs. Following pino 266 * will be used only for fsynced inodes after checkpoint. 267 */ 268 try_to_fix_pino(inode); 269 clear_inode_flag(inode, FI_APPEND_WRITE); 270 clear_inode_flag(inode, FI_UPDATE_WRITE); 271 goto out; 272 } 273 sync_nodes: 274 atomic_inc(&sbi->wb_sync_req[NODE]); 275 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 276 atomic_dec(&sbi->wb_sync_req[NODE]); 277 if (ret) 278 goto out; 279 280 /* if cp_error was enabled, we should avoid infinite loop */ 281 if (unlikely(f2fs_cp_error(sbi))) { 282 ret = -EIO; 283 goto out; 284 } 285 286 if (f2fs_need_inode_block_update(sbi, ino)) { 287 f2fs_mark_inode_dirty_sync(inode, true); 288 f2fs_write_inode(inode, NULL); 289 goto sync_nodes; 290 } 291 292 /* 293 * If it's atomic_write, it's just fine to keep write ordering. So 294 * here we don't need to wait for node write completion, since we use 295 * node chain which serializes node blocks. If one of node writes are 296 * reordered, we can see simply broken chain, resulting in stopping 297 * roll-forward recovery. It means we'll recover all or none node blocks 298 * given fsync mark. 299 */ 300 if (!atomic) { 301 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 302 if (ret) 303 goto out; 304 } 305 306 /* once recovery info is written, don't need to tack this */ 307 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 308 clear_inode_flag(inode, FI_APPEND_WRITE); 309 flush_out: 310 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) 311 ret = f2fs_issue_flush(sbi, inode->i_ino); 312 if (!ret) { 313 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 314 clear_inode_flag(inode, FI_UPDATE_WRITE); 315 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 316 } 317 f2fs_update_time(sbi, REQ_TIME); 318 out: 319 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 320 f2fs_trace_ios(NULL, 1); 321 return ret; 322 } 323 324 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 325 { 326 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 327 return -EIO; 328 return f2fs_do_sync_file(file, start, end, datasync, false); 329 } 330 331 static pgoff_t __get_first_dirty_index(struct address_space *mapping, 332 pgoff_t pgofs, int whence) 333 { 334 struct page *page; 335 int nr_pages; 336 337 if (whence != SEEK_DATA) 338 return 0; 339 340 /* find first dirty page index */ 341 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY, 342 1, &page); 343 if (!nr_pages) 344 return ULONG_MAX; 345 pgofs = page->index; 346 put_page(page); 347 return pgofs; 348 } 349 350 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr, 351 pgoff_t dirty, pgoff_t pgofs, int whence) 352 { 353 switch (whence) { 354 case SEEK_DATA: 355 if ((blkaddr == NEW_ADDR && dirty == pgofs) || 356 is_valid_data_blkaddr(sbi, blkaddr)) 357 return true; 358 break; 359 case SEEK_HOLE: 360 if (blkaddr == NULL_ADDR) 361 return true; 362 break; 363 } 364 return false; 365 } 366 367 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 368 { 369 struct inode *inode = file->f_mapping->host; 370 loff_t maxbytes = inode->i_sb->s_maxbytes; 371 struct dnode_of_data dn; 372 pgoff_t pgofs, end_offset, dirty; 373 loff_t data_ofs = offset; 374 loff_t isize; 375 int err = 0; 376 377 inode_lock(inode); 378 379 isize = i_size_read(inode); 380 if (offset >= isize) 381 goto fail; 382 383 /* handle inline data case */ 384 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) { 385 if (whence == SEEK_HOLE) 386 data_ofs = isize; 387 goto found; 388 } 389 390 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 391 392 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence); 393 394 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 395 set_new_dnode(&dn, inode, NULL, NULL, 0); 396 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 397 if (err && err != -ENOENT) { 398 goto fail; 399 } else if (err == -ENOENT) { 400 /* direct node does not exists */ 401 if (whence == SEEK_DATA) { 402 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 403 continue; 404 } else { 405 goto found; 406 } 407 } 408 409 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 410 411 /* find data/hole in dnode block */ 412 for (; dn.ofs_in_node < end_offset; 413 dn.ofs_in_node++, pgofs++, 414 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 415 block_t blkaddr; 416 417 blkaddr = datablock_addr(dn.inode, 418 dn.node_page, dn.ofs_in_node); 419 420 if (__is_valid_data_blkaddr(blkaddr) && 421 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 422 blkaddr, DATA_GENERIC)) { 423 f2fs_put_dnode(&dn); 424 goto fail; 425 } 426 427 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty, 428 pgofs, whence)) { 429 f2fs_put_dnode(&dn); 430 goto found; 431 } 432 } 433 f2fs_put_dnode(&dn); 434 } 435 436 if (whence == SEEK_DATA) 437 goto fail; 438 found: 439 if (whence == SEEK_HOLE && data_ofs > isize) 440 data_ofs = isize; 441 inode_unlock(inode); 442 return vfs_setpos(file, data_ofs, maxbytes); 443 fail: 444 inode_unlock(inode); 445 return -ENXIO; 446 } 447 448 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 449 { 450 struct inode *inode = file->f_mapping->host; 451 loff_t maxbytes = inode->i_sb->s_maxbytes; 452 453 switch (whence) { 454 case SEEK_SET: 455 case SEEK_CUR: 456 case SEEK_END: 457 return generic_file_llseek_size(file, offset, whence, 458 maxbytes, i_size_read(inode)); 459 case SEEK_DATA: 460 case SEEK_HOLE: 461 if (offset < 0) 462 return -ENXIO; 463 return f2fs_seek_block(file, offset, whence); 464 } 465 466 return -EINVAL; 467 } 468 469 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 470 { 471 struct inode *inode = file_inode(file); 472 int err; 473 474 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 475 return -EIO; 476 477 /* we don't need to use inline_data strictly */ 478 err = f2fs_convert_inline_inode(inode); 479 if (err) 480 return err; 481 482 file_accessed(file); 483 vma->vm_ops = &f2fs_file_vm_ops; 484 return 0; 485 } 486 487 static int f2fs_file_open(struct inode *inode, struct file *filp) 488 { 489 int err = fscrypt_file_open(inode, filp); 490 491 if (err) 492 return err; 493 494 filp->f_mode |= FMODE_NOWAIT; 495 496 return dquot_file_open(inode, filp); 497 } 498 499 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 500 { 501 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 502 struct f2fs_node *raw_node; 503 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 504 __le32 *addr; 505 int base = 0; 506 507 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode)) 508 base = get_extra_isize(dn->inode); 509 510 raw_node = F2FS_NODE(dn->node_page); 511 addr = blkaddr_in_node(raw_node) + base + ofs; 512 513 for (; count > 0; count--, addr++, dn->ofs_in_node++) { 514 block_t blkaddr = le32_to_cpu(*addr); 515 516 if (blkaddr == NULL_ADDR) 517 continue; 518 519 dn->data_blkaddr = NULL_ADDR; 520 f2fs_set_data_blkaddr(dn); 521 522 if (__is_valid_data_blkaddr(blkaddr) && 523 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) 524 continue; 525 526 f2fs_invalidate_blocks(sbi, blkaddr); 527 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) 528 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN); 529 nr_free++; 530 } 531 532 if (nr_free) { 533 pgoff_t fofs; 534 /* 535 * once we invalidate valid blkaddr in range [ofs, ofs + count], 536 * we will invalidate all blkaddr in the whole range. 537 */ 538 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 539 dn->inode) + ofs; 540 f2fs_update_extent_cache_range(dn, fofs, 0, len); 541 dec_valid_block_count(sbi, dn->inode, nr_free); 542 } 543 dn->ofs_in_node = ofs; 544 545 f2fs_update_time(sbi, REQ_TIME); 546 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 547 dn->ofs_in_node, nr_free); 548 } 549 550 void f2fs_truncate_data_blocks(struct dnode_of_data *dn) 551 { 552 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK); 553 } 554 555 static int truncate_partial_data_page(struct inode *inode, u64 from, 556 bool cache_only) 557 { 558 loff_t offset = from & (PAGE_SIZE - 1); 559 pgoff_t index = from >> PAGE_SHIFT; 560 struct address_space *mapping = inode->i_mapping; 561 struct page *page; 562 563 if (!offset && !cache_only) 564 return 0; 565 566 if (cache_only) { 567 page = find_lock_page(mapping, index); 568 if (page && PageUptodate(page)) 569 goto truncate_out; 570 f2fs_put_page(page, 1); 571 return 0; 572 } 573 574 page = f2fs_get_lock_data_page(inode, index, true); 575 if (IS_ERR(page)) 576 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 577 truncate_out: 578 f2fs_wait_on_page_writeback(page, DATA, true); 579 zero_user(page, offset, PAGE_SIZE - offset); 580 581 /* An encrypted inode should have a key and truncate the last page. */ 582 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode)); 583 if (!cache_only) 584 set_page_dirty(page); 585 f2fs_put_page(page, 1); 586 return 0; 587 } 588 589 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 590 { 591 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 592 struct dnode_of_data dn; 593 pgoff_t free_from; 594 int count = 0, err = 0; 595 struct page *ipage; 596 bool truncate_page = false; 597 598 trace_f2fs_truncate_blocks_enter(inode, from); 599 600 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 601 602 if (free_from >= sbi->max_file_blocks) 603 goto free_partial; 604 605 if (lock) 606 f2fs_lock_op(sbi); 607 608 ipage = f2fs_get_node_page(sbi, inode->i_ino); 609 if (IS_ERR(ipage)) { 610 err = PTR_ERR(ipage); 611 goto out; 612 } 613 614 if (f2fs_has_inline_data(inode)) { 615 f2fs_truncate_inline_inode(inode, ipage, from); 616 f2fs_put_page(ipage, 1); 617 truncate_page = true; 618 goto out; 619 } 620 621 set_new_dnode(&dn, inode, ipage, NULL, 0); 622 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 623 if (err) { 624 if (err == -ENOENT) 625 goto free_next; 626 goto out; 627 } 628 629 count = ADDRS_PER_PAGE(dn.node_page, inode); 630 631 count -= dn.ofs_in_node; 632 f2fs_bug_on(sbi, count < 0); 633 634 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 635 f2fs_truncate_data_blocks_range(&dn, count); 636 free_from += count; 637 } 638 639 f2fs_put_dnode(&dn); 640 free_next: 641 err = f2fs_truncate_inode_blocks(inode, free_from); 642 out: 643 if (lock) 644 f2fs_unlock_op(sbi); 645 free_partial: 646 /* lastly zero out the first data page */ 647 if (!err) 648 err = truncate_partial_data_page(inode, from, truncate_page); 649 650 trace_f2fs_truncate_blocks_exit(inode, err); 651 return err; 652 } 653 654 int f2fs_truncate(struct inode *inode) 655 { 656 int err; 657 658 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 659 return -EIO; 660 661 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 662 S_ISLNK(inode->i_mode))) 663 return 0; 664 665 trace_f2fs_truncate(inode); 666 667 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) { 668 f2fs_show_injection_info(FAULT_TRUNCATE); 669 return -EIO; 670 } 671 672 /* we should check inline_data size */ 673 if (!f2fs_may_inline_data(inode)) { 674 err = f2fs_convert_inline_inode(inode); 675 if (err) 676 return err; 677 } 678 679 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 680 if (err) 681 return err; 682 683 inode->i_mtime = inode->i_ctime = current_time(inode); 684 f2fs_mark_inode_dirty_sync(inode, false); 685 return 0; 686 } 687 688 int f2fs_getattr(const struct path *path, struct kstat *stat, 689 u32 request_mask, unsigned int query_flags) 690 { 691 struct inode *inode = d_inode(path->dentry); 692 struct f2fs_inode_info *fi = F2FS_I(inode); 693 struct f2fs_inode *ri; 694 unsigned int flags; 695 696 if (f2fs_has_extra_attr(inode) && 697 f2fs_sb_has_inode_crtime(inode->i_sb) && 698 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 699 stat->result_mask |= STATX_BTIME; 700 stat->btime.tv_sec = fi->i_crtime.tv_sec; 701 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 702 } 703 704 flags = fi->i_flags & F2FS_FL_USER_VISIBLE; 705 if (flags & F2FS_APPEND_FL) 706 stat->attributes |= STATX_ATTR_APPEND; 707 if (flags & F2FS_COMPR_FL) 708 stat->attributes |= STATX_ATTR_COMPRESSED; 709 if (f2fs_encrypted_inode(inode)) 710 stat->attributes |= STATX_ATTR_ENCRYPTED; 711 if (flags & F2FS_IMMUTABLE_FL) 712 stat->attributes |= STATX_ATTR_IMMUTABLE; 713 if (flags & F2FS_NODUMP_FL) 714 stat->attributes |= STATX_ATTR_NODUMP; 715 716 stat->attributes_mask |= (STATX_ATTR_APPEND | 717 STATX_ATTR_COMPRESSED | 718 STATX_ATTR_ENCRYPTED | 719 STATX_ATTR_IMMUTABLE | 720 STATX_ATTR_NODUMP); 721 722 generic_fillattr(inode, stat); 723 724 /* we need to show initial sectors used for inline_data/dentries */ 725 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 726 f2fs_has_inline_dentry(inode)) 727 stat->blocks += (stat->size + 511) >> 9; 728 729 return 0; 730 } 731 732 #ifdef CONFIG_F2FS_FS_POSIX_ACL 733 static void __setattr_copy(struct inode *inode, const struct iattr *attr) 734 { 735 unsigned int ia_valid = attr->ia_valid; 736 737 if (ia_valid & ATTR_UID) 738 inode->i_uid = attr->ia_uid; 739 if (ia_valid & ATTR_GID) 740 inode->i_gid = attr->ia_gid; 741 if (ia_valid & ATTR_ATIME) 742 inode->i_atime = timespec64_trunc(attr->ia_atime, 743 inode->i_sb->s_time_gran); 744 if (ia_valid & ATTR_MTIME) 745 inode->i_mtime = timespec64_trunc(attr->ia_mtime, 746 inode->i_sb->s_time_gran); 747 if (ia_valid & ATTR_CTIME) 748 inode->i_ctime = timespec64_trunc(attr->ia_ctime, 749 inode->i_sb->s_time_gran); 750 if (ia_valid & ATTR_MODE) { 751 umode_t mode = attr->ia_mode; 752 753 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) 754 mode &= ~S_ISGID; 755 set_acl_inode(inode, mode); 756 } 757 } 758 #else 759 #define __setattr_copy setattr_copy 760 #endif 761 762 int f2fs_setattr(struct dentry *dentry, struct iattr *attr) 763 { 764 struct inode *inode = d_inode(dentry); 765 int err; 766 bool size_changed = false; 767 768 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 769 return -EIO; 770 771 err = setattr_prepare(dentry, attr); 772 if (err) 773 return err; 774 775 err = fscrypt_prepare_setattr(dentry, attr); 776 if (err) 777 return err; 778 779 if (is_quota_modification(inode, attr)) { 780 err = dquot_initialize(inode); 781 if (err) 782 return err; 783 } 784 if ((attr->ia_valid & ATTR_UID && 785 !uid_eq(attr->ia_uid, inode->i_uid)) || 786 (attr->ia_valid & ATTR_GID && 787 !gid_eq(attr->ia_gid, inode->i_gid))) { 788 err = dquot_transfer(inode, attr); 789 if (err) 790 return err; 791 } 792 793 if (attr->ia_valid & ATTR_SIZE) { 794 bool to_smaller = (attr->ia_size <= i_size_read(inode)); 795 796 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 797 down_write(&F2FS_I(inode)->i_mmap_sem); 798 799 truncate_setsize(inode, attr->ia_size); 800 801 if (to_smaller) 802 err = f2fs_truncate(inode); 803 /* 804 * do not trim all blocks after i_size if target size is 805 * larger than i_size. 806 */ 807 up_write(&F2FS_I(inode)->i_mmap_sem); 808 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 809 810 if (err) 811 return err; 812 813 if (!to_smaller) { 814 /* should convert inline inode here */ 815 if (!f2fs_may_inline_data(inode)) { 816 err = f2fs_convert_inline_inode(inode); 817 if (err) 818 return err; 819 } 820 inode->i_mtime = inode->i_ctime = current_time(inode); 821 } 822 823 down_write(&F2FS_I(inode)->i_sem); 824 F2FS_I(inode)->last_disk_size = i_size_read(inode); 825 up_write(&F2FS_I(inode)->i_sem); 826 827 size_changed = true; 828 } 829 830 __setattr_copy(inode, attr); 831 832 if (attr->ia_valid & ATTR_MODE) { 833 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode)); 834 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) { 835 inode->i_mode = F2FS_I(inode)->i_acl_mode; 836 clear_inode_flag(inode, FI_ACL_MODE); 837 } 838 } 839 840 /* file size may changed here */ 841 f2fs_mark_inode_dirty_sync(inode, size_changed); 842 843 /* inode change will produce dirty node pages flushed by checkpoint */ 844 f2fs_balance_fs(F2FS_I_SB(inode), true); 845 846 return err; 847 } 848 849 const struct inode_operations f2fs_file_inode_operations = { 850 .getattr = f2fs_getattr, 851 .setattr = f2fs_setattr, 852 .get_acl = f2fs_get_acl, 853 .set_acl = f2fs_set_acl, 854 #ifdef CONFIG_F2FS_FS_XATTR 855 .listxattr = f2fs_listxattr, 856 #endif 857 .fiemap = f2fs_fiemap, 858 }; 859 860 static int fill_zero(struct inode *inode, pgoff_t index, 861 loff_t start, loff_t len) 862 { 863 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 864 struct page *page; 865 866 if (!len) 867 return 0; 868 869 f2fs_balance_fs(sbi, true); 870 871 f2fs_lock_op(sbi); 872 page = f2fs_get_new_data_page(inode, NULL, index, false); 873 f2fs_unlock_op(sbi); 874 875 if (IS_ERR(page)) 876 return PTR_ERR(page); 877 878 f2fs_wait_on_page_writeback(page, DATA, true); 879 zero_user(page, start, len); 880 set_page_dirty(page); 881 f2fs_put_page(page, 1); 882 return 0; 883 } 884 885 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 886 { 887 int err; 888 889 while (pg_start < pg_end) { 890 struct dnode_of_data dn; 891 pgoff_t end_offset, count; 892 893 set_new_dnode(&dn, inode, NULL, NULL, 0); 894 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 895 if (err) { 896 if (err == -ENOENT) { 897 pg_start = f2fs_get_next_page_offset(&dn, 898 pg_start); 899 continue; 900 } 901 return err; 902 } 903 904 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 905 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 906 907 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 908 909 f2fs_truncate_data_blocks_range(&dn, count); 910 f2fs_put_dnode(&dn); 911 912 pg_start += count; 913 } 914 return 0; 915 } 916 917 static int punch_hole(struct inode *inode, loff_t offset, loff_t len) 918 { 919 pgoff_t pg_start, pg_end; 920 loff_t off_start, off_end; 921 int ret; 922 923 ret = f2fs_convert_inline_inode(inode); 924 if (ret) 925 return ret; 926 927 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 928 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 929 930 off_start = offset & (PAGE_SIZE - 1); 931 off_end = (offset + len) & (PAGE_SIZE - 1); 932 933 if (pg_start == pg_end) { 934 ret = fill_zero(inode, pg_start, off_start, 935 off_end - off_start); 936 if (ret) 937 return ret; 938 } else { 939 if (off_start) { 940 ret = fill_zero(inode, pg_start++, off_start, 941 PAGE_SIZE - off_start); 942 if (ret) 943 return ret; 944 } 945 if (off_end) { 946 ret = fill_zero(inode, pg_end, 0, off_end); 947 if (ret) 948 return ret; 949 } 950 951 if (pg_start < pg_end) { 952 struct address_space *mapping = inode->i_mapping; 953 loff_t blk_start, blk_end; 954 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 955 956 f2fs_balance_fs(sbi, true); 957 958 blk_start = (loff_t)pg_start << PAGE_SHIFT; 959 blk_end = (loff_t)pg_end << PAGE_SHIFT; 960 961 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 962 down_write(&F2FS_I(inode)->i_mmap_sem); 963 964 truncate_inode_pages_range(mapping, blk_start, 965 blk_end - 1); 966 967 f2fs_lock_op(sbi); 968 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 969 f2fs_unlock_op(sbi); 970 971 up_write(&F2FS_I(inode)->i_mmap_sem); 972 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 973 } 974 } 975 976 return ret; 977 } 978 979 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 980 int *do_replace, pgoff_t off, pgoff_t len) 981 { 982 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 983 struct dnode_of_data dn; 984 int ret, done, i; 985 986 next_dnode: 987 set_new_dnode(&dn, inode, NULL, NULL, 0); 988 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 989 if (ret && ret != -ENOENT) { 990 return ret; 991 } else if (ret == -ENOENT) { 992 if (dn.max_level == 0) 993 return -ENOENT; 994 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len); 995 blkaddr += done; 996 do_replace += done; 997 goto next; 998 } 999 1000 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1001 dn.ofs_in_node, len); 1002 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1003 *blkaddr = datablock_addr(dn.inode, 1004 dn.node_page, dn.ofs_in_node); 1005 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1006 1007 if (test_opt(sbi, LFS)) { 1008 f2fs_put_dnode(&dn); 1009 return -ENOTSUPP; 1010 } 1011 1012 /* do not invalidate this block address */ 1013 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1014 *do_replace = 1; 1015 } 1016 } 1017 f2fs_put_dnode(&dn); 1018 next: 1019 len -= done; 1020 off += done; 1021 if (len) 1022 goto next_dnode; 1023 return 0; 1024 } 1025 1026 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1027 int *do_replace, pgoff_t off, int len) 1028 { 1029 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1030 struct dnode_of_data dn; 1031 int ret, i; 1032 1033 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1034 if (*do_replace == 0) 1035 continue; 1036 1037 set_new_dnode(&dn, inode, NULL, NULL, 0); 1038 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1039 if (ret) { 1040 dec_valid_block_count(sbi, inode, 1); 1041 f2fs_invalidate_blocks(sbi, *blkaddr); 1042 } else { 1043 f2fs_update_data_blkaddr(&dn, *blkaddr); 1044 } 1045 f2fs_put_dnode(&dn); 1046 } 1047 return 0; 1048 } 1049 1050 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1051 block_t *blkaddr, int *do_replace, 1052 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1053 { 1054 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1055 pgoff_t i = 0; 1056 int ret; 1057 1058 while (i < len) { 1059 if (blkaddr[i] == NULL_ADDR && !full) { 1060 i++; 1061 continue; 1062 } 1063 1064 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1065 struct dnode_of_data dn; 1066 struct node_info ni; 1067 size_t new_size; 1068 pgoff_t ilen; 1069 1070 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1071 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1072 if (ret) 1073 return ret; 1074 1075 ret = f2fs_get_node_info(sbi, dn.nid, &ni); 1076 if (ret) { 1077 f2fs_put_dnode(&dn); 1078 return ret; 1079 } 1080 1081 ilen = min((pgoff_t) 1082 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1083 dn.ofs_in_node, len - i); 1084 do { 1085 dn.data_blkaddr = datablock_addr(dn.inode, 1086 dn.node_page, dn.ofs_in_node); 1087 f2fs_truncate_data_blocks_range(&dn, 1); 1088 1089 if (do_replace[i]) { 1090 f2fs_i_blocks_write(src_inode, 1091 1, false, false); 1092 f2fs_i_blocks_write(dst_inode, 1093 1, true, false); 1094 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1095 blkaddr[i], ni.version, true, false); 1096 1097 do_replace[i] = 0; 1098 } 1099 dn.ofs_in_node++; 1100 i++; 1101 new_size = (dst + i) << PAGE_SHIFT; 1102 if (dst_inode->i_size < new_size) 1103 f2fs_i_size_write(dst_inode, new_size); 1104 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1105 1106 f2fs_put_dnode(&dn); 1107 } else { 1108 struct page *psrc, *pdst; 1109 1110 psrc = f2fs_get_lock_data_page(src_inode, 1111 src + i, true); 1112 if (IS_ERR(psrc)) 1113 return PTR_ERR(psrc); 1114 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1115 true); 1116 if (IS_ERR(pdst)) { 1117 f2fs_put_page(psrc, 1); 1118 return PTR_ERR(pdst); 1119 } 1120 f2fs_copy_page(psrc, pdst); 1121 set_page_dirty(pdst); 1122 f2fs_put_page(pdst, 1); 1123 f2fs_put_page(psrc, 1); 1124 1125 ret = f2fs_truncate_hole(src_inode, 1126 src + i, src + i + 1); 1127 if (ret) 1128 return ret; 1129 i++; 1130 } 1131 } 1132 return 0; 1133 } 1134 1135 static int __exchange_data_block(struct inode *src_inode, 1136 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1137 pgoff_t len, bool full) 1138 { 1139 block_t *src_blkaddr; 1140 int *do_replace; 1141 pgoff_t olen; 1142 int ret; 1143 1144 while (len) { 1145 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len); 1146 1147 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1148 array_size(olen, sizeof(block_t)), 1149 GFP_KERNEL); 1150 if (!src_blkaddr) 1151 return -ENOMEM; 1152 1153 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1154 array_size(olen, sizeof(int)), 1155 GFP_KERNEL); 1156 if (!do_replace) { 1157 kvfree(src_blkaddr); 1158 return -ENOMEM; 1159 } 1160 1161 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1162 do_replace, src, olen); 1163 if (ret) 1164 goto roll_back; 1165 1166 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1167 do_replace, src, dst, olen, full); 1168 if (ret) 1169 goto roll_back; 1170 1171 src += olen; 1172 dst += olen; 1173 len -= olen; 1174 1175 kvfree(src_blkaddr); 1176 kvfree(do_replace); 1177 } 1178 return 0; 1179 1180 roll_back: 1181 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1182 kvfree(src_blkaddr); 1183 kvfree(do_replace); 1184 return ret; 1185 } 1186 1187 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1188 { 1189 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1190 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1191 pgoff_t start = offset >> PAGE_SHIFT; 1192 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1193 int ret; 1194 1195 f2fs_balance_fs(sbi, true); 1196 1197 /* avoid gc operation during block exchange */ 1198 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1199 down_write(&F2FS_I(inode)->i_mmap_sem); 1200 1201 f2fs_lock_op(sbi); 1202 f2fs_drop_extent_tree(inode); 1203 truncate_pagecache(inode, offset); 1204 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1205 f2fs_unlock_op(sbi); 1206 1207 up_write(&F2FS_I(inode)->i_mmap_sem); 1208 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1209 return ret; 1210 } 1211 1212 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1213 { 1214 loff_t new_size; 1215 int ret; 1216 1217 if (offset + len >= i_size_read(inode)) 1218 return -EINVAL; 1219 1220 /* collapse range should be aligned to block size of f2fs. */ 1221 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1222 return -EINVAL; 1223 1224 ret = f2fs_convert_inline_inode(inode); 1225 if (ret) 1226 return ret; 1227 1228 /* write out all dirty pages from offset */ 1229 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1230 if (ret) 1231 return ret; 1232 1233 ret = f2fs_do_collapse(inode, offset, len); 1234 if (ret) 1235 return ret; 1236 1237 /* write out all moved pages, if possible */ 1238 down_write(&F2FS_I(inode)->i_mmap_sem); 1239 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1240 truncate_pagecache(inode, offset); 1241 1242 new_size = i_size_read(inode) - len; 1243 truncate_pagecache(inode, new_size); 1244 1245 ret = f2fs_truncate_blocks(inode, new_size, true); 1246 up_write(&F2FS_I(inode)->i_mmap_sem); 1247 if (!ret) 1248 f2fs_i_size_write(inode, new_size); 1249 return ret; 1250 } 1251 1252 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1253 pgoff_t end) 1254 { 1255 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1256 pgoff_t index = start; 1257 unsigned int ofs_in_node = dn->ofs_in_node; 1258 blkcnt_t count = 0; 1259 int ret; 1260 1261 for (; index < end; index++, dn->ofs_in_node++) { 1262 if (datablock_addr(dn->inode, dn->node_page, 1263 dn->ofs_in_node) == NULL_ADDR) 1264 count++; 1265 } 1266 1267 dn->ofs_in_node = ofs_in_node; 1268 ret = f2fs_reserve_new_blocks(dn, count); 1269 if (ret) 1270 return ret; 1271 1272 dn->ofs_in_node = ofs_in_node; 1273 for (index = start; index < end; index++, dn->ofs_in_node++) { 1274 dn->data_blkaddr = datablock_addr(dn->inode, 1275 dn->node_page, dn->ofs_in_node); 1276 /* 1277 * f2fs_reserve_new_blocks will not guarantee entire block 1278 * allocation. 1279 */ 1280 if (dn->data_blkaddr == NULL_ADDR) { 1281 ret = -ENOSPC; 1282 break; 1283 } 1284 if (dn->data_blkaddr != NEW_ADDR) { 1285 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1286 dn->data_blkaddr = NEW_ADDR; 1287 f2fs_set_data_blkaddr(dn); 1288 } 1289 } 1290 1291 f2fs_update_extent_cache_range(dn, start, 0, index - start); 1292 1293 return ret; 1294 } 1295 1296 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1297 int mode) 1298 { 1299 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1300 struct address_space *mapping = inode->i_mapping; 1301 pgoff_t index, pg_start, pg_end; 1302 loff_t new_size = i_size_read(inode); 1303 loff_t off_start, off_end; 1304 int ret = 0; 1305 1306 ret = inode_newsize_ok(inode, (len + offset)); 1307 if (ret) 1308 return ret; 1309 1310 ret = f2fs_convert_inline_inode(inode); 1311 if (ret) 1312 return ret; 1313 1314 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1315 if (ret) 1316 return ret; 1317 1318 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1319 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1320 1321 off_start = offset & (PAGE_SIZE - 1); 1322 off_end = (offset + len) & (PAGE_SIZE - 1); 1323 1324 if (pg_start == pg_end) { 1325 ret = fill_zero(inode, pg_start, off_start, 1326 off_end - off_start); 1327 if (ret) 1328 return ret; 1329 1330 new_size = max_t(loff_t, new_size, offset + len); 1331 } else { 1332 if (off_start) { 1333 ret = fill_zero(inode, pg_start++, off_start, 1334 PAGE_SIZE - off_start); 1335 if (ret) 1336 return ret; 1337 1338 new_size = max_t(loff_t, new_size, 1339 (loff_t)pg_start << PAGE_SHIFT); 1340 } 1341 1342 for (index = pg_start; index < pg_end;) { 1343 struct dnode_of_data dn; 1344 unsigned int end_offset; 1345 pgoff_t end; 1346 1347 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1348 down_write(&F2FS_I(inode)->i_mmap_sem); 1349 1350 truncate_pagecache_range(inode, 1351 (loff_t)index << PAGE_SHIFT, 1352 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1353 1354 f2fs_lock_op(sbi); 1355 1356 set_new_dnode(&dn, inode, NULL, NULL, 0); 1357 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1358 if (ret) { 1359 f2fs_unlock_op(sbi); 1360 up_write(&F2FS_I(inode)->i_mmap_sem); 1361 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1362 goto out; 1363 } 1364 1365 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1366 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1367 1368 ret = f2fs_do_zero_range(&dn, index, end); 1369 f2fs_put_dnode(&dn); 1370 1371 f2fs_unlock_op(sbi); 1372 up_write(&F2FS_I(inode)->i_mmap_sem); 1373 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1374 1375 f2fs_balance_fs(sbi, dn.node_changed); 1376 1377 if (ret) 1378 goto out; 1379 1380 index = end; 1381 new_size = max_t(loff_t, new_size, 1382 (loff_t)index << PAGE_SHIFT); 1383 } 1384 1385 if (off_end) { 1386 ret = fill_zero(inode, pg_end, 0, off_end); 1387 if (ret) 1388 goto out; 1389 1390 new_size = max_t(loff_t, new_size, offset + len); 1391 } 1392 } 1393 1394 out: 1395 if (new_size > i_size_read(inode)) { 1396 if (mode & FALLOC_FL_KEEP_SIZE) 1397 file_set_keep_isize(inode); 1398 else 1399 f2fs_i_size_write(inode, new_size); 1400 } 1401 return ret; 1402 } 1403 1404 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1405 { 1406 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1407 pgoff_t nr, pg_start, pg_end, delta, idx; 1408 loff_t new_size; 1409 int ret = 0; 1410 1411 new_size = i_size_read(inode) + len; 1412 ret = inode_newsize_ok(inode, new_size); 1413 if (ret) 1414 return ret; 1415 1416 if (offset >= i_size_read(inode)) 1417 return -EINVAL; 1418 1419 /* insert range should be aligned to block size of f2fs. */ 1420 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1421 return -EINVAL; 1422 1423 ret = f2fs_convert_inline_inode(inode); 1424 if (ret) 1425 return ret; 1426 1427 f2fs_balance_fs(sbi, true); 1428 1429 down_write(&F2FS_I(inode)->i_mmap_sem); 1430 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1431 up_write(&F2FS_I(inode)->i_mmap_sem); 1432 if (ret) 1433 return ret; 1434 1435 /* write out all dirty pages from offset */ 1436 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1437 if (ret) 1438 return ret; 1439 1440 pg_start = offset >> PAGE_SHIFT; 1441 pg_end = (offset + len) >> PAGE_SHIFT; 1442 delta = pg_end - pg_start; 1443 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1444 1445 /* avoid gc operation during block exchange */ 1446 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1447 down_write(&F2FS_I(inode)->i_mmap_sem); 1448 truncate_pagecache(inode, offset); 1449 1450 while (!ret && idx > pg_start) { 1451 nr = idx - pg_start; 1452 if (nr > delta) 1453 nr = delta; 1454 idx -= nr; 1455 1456 f2fs_lock_op(sbi); 1457 f2fs_drop_extent_tree(inode); 1458 1459 ret = __exchange_data_block(inode, inode, idx, 1460 idx + delta, nr, false); 1461 f2fs_unlock_op(sbi); 1462 } 1463 up_write(&F2FS_I(inode)->i_mmap_sem); 1464 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1465 1466 /* write out all moved pages, if possible */ 1467 down_write(&F2FS_I(inode)->i_mmap_sem); 1468 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1469 truncate_pagecache(inode, offset); 1470 up_write(&F2FS_I(inode)->i_mmap_sem); 1471 1472 if (!ret) 1473 f2fs_i_size_write(inode, new_size); 1474 return ret; 1475 } 1476 1477 static int expand_inode_data(struct inode *inode, loff_t offset, 1478 loff_t len, int mode) 1479 { 1480 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1481 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1482 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE }; 1483 pgoff_t pg_end; 1484 loff_t new_size = i_size_read(inode); 1485 loff_t off_end; 1486 int err; 1487 1488 err = inode_newsize_ok(inode, (len + offset)); 1489 if (err) 1490 return err; 1491 1492 err = f2fs_convert_inline_inode(inode); 1493 if (err) 1494 return err; 1495 1496 f2fs_balance_fs(sbi, true); 1497 1498 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1499 off_end = (offset + len) & (PAGE_SIZE - 1); 1500 1501 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT; 1502 map.m_len = pg_end - map.m_lblk; 1503 if (off_end) 1504 map.m_len++; 1505 1506 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); 1507 if (err) { 1508 pgoff_t last_off; 1509 1510 if (!map.m_len) 1511 return err; 1512 1513 last_off = map.m_lblk + map.m_len - 1; 1514 1515 /* update new size to the failed position */ 1516 new_size = (last_off == pg_end) ? offset + len : 1517 (loff_t)(last_off + 1) << PAGE_SHIFT; 1518 } else { 1519 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1520 } 1521 1522 if (new_size > i_size_read(inode)) { 1523 if (mode & FALLOC_FL_KEEP_SIZE) 1524 file_set_keep_isize(inode); 1525 else 1526 f2fs_i_size_write(inode, new_size); 1527 } 1528 1529 return err; 1530 } 1531 1532 static long f2fs_fallocate(struct file *file, int mode, 1533 loff_t offset, loff_t len) 1534 { 1535 struct inode *inode = file_inode(file); 1536 long ret = 0; 1537 1538 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1539 return -EIO; 1540 1541 /* f2fs only support ->fallocate for regular file */ 1542 if (!S_ISREG(inode->i_mode)) 1543 return -EINVAL; 1544 1545 if (f2fs_encrypted_inode(inode) && 1546 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1547 return -EOPNOTSUPP; 1548 1549 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1550 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1551 FALLOC_FL_INSERT_RANGE)) 1552 return -EOPNOTSUPP; 1553 1554 inode_lock(inode); 1555 1556 if (mode & FALLOC_FL_PUNCH_HOLE) { 1557 if (offset >= inode->i_size) 1558 goto out; 1559 1560 ret = punch_hole(inode, offset, len); 1561 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1562 ret = f2fs_collapse_range(inode, offset, len); 1563 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1564 ret = f2fs_zero_range(inode, offset, len, mode); 1565 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1566 ret = f2fs_insert_range(inode, offset, len); 1567 } else { 1568 ret = expand_inode_data(inode, offset, len, mode); 1569 } 1570 1571 if (!ret) { 1572 inode->i_mtime = inode->i_ctime = current_time(inode); 1573 f2fs_mark_inode_dirty_sync(inode, false); 1574 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1575 } 1576 1577 out: 1578 inode_unlock(inode); 1579 1580 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1581 return ret; 1582 } 1583 1584 static int f2fs_release_file(struct inode *inode, struct file *filp) 1585 { 1586 /* 1587 * f2fs_relase_file is called at every close calls. So we should 1588 * not drop any inmemory pages by close called by other process. 1589 */ 1590 if (!(filp->f_mode & FMODE_WRITE) || 1591 atomic_read(&inode->i_writecount) != 1) 1592 return 0; 1593 1594 /* some remained atomic pages should discarded */ 1595 if (f2fs_is_atomic_file(inode)) 1596 f2fs_drop_inmem_pages(inode); 1597 if (f2fs_is_volatile_file(inode)) { 1598 set_inode_flag(inode, FI_DROP_CACHE); 1599 filemap_fdatawrite(inode->i_mapping); 1600 clear_inode_flag(inode, FI_DROP_CACHE); 1601 clear_inode_flag(inode, FI_VOLATILE_FILE); 1602 stat_dec_volatile_write(inode); 1603 } 1604 return 0; 1605 } 1606 1607 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1608 { 1609 struct inode *inode = file_inode(file); 1610 1611 /* 1612 * If the process doing a transaction is crashed, we should do 1613 * roll-back. Otherwise, other reader/write can see corrupted database 1614 * until all the writers close its file. Since this should be done 1615 * before dropping file lock, it needs to do in ->flush. 1616 */ 1617 if (f2fs_is_atomic_file(inode) && 1618 F2FS_I(inode)->inmem_task == current) 1619 f2fs_drop_inmem_pages(inode); 1620 return 0; 1621 } 1622 1623 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg) 1624 { 1625 struct inode *inode = file_inode(filp); 1626 struct f2fs_inode_info *fi = F2FS_I(inode); 1627 unsigned int flags = fi->i_flags; 1628 1629 if (f2fs_encrypted_inode(inode)) 1630 flags |= F2FS_ENCRYPT_FL; 1631 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 1632 flags |= F2FS_INLINE_DATA_FL; 1633 1634 flags &= F2FS_FL_USER_VISIBLE; 1635 1636 return put_user(flags, (int __user *)arg); 1637 } 1638 1639 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags) 1640 { 1641 struct f2fs_inode_info *fi = F2FS_I(inode); 1642 unsigned int oldflags; 1643 1644 /* Is it quota file? Do not allow user to mess with it */ 1645 if (IS_NOQUOTA(inode)) 1646 return -EPERM; 1647 1648 flags = f2fs_mask_flags(inode->i_mode, flags); 1649 1650 oldflags = fi->i_flags; 1651 1652 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL)) 1653 if (!capable(CAP_LINUX_IMMUTABLE)) 1654 return -EPERM; 1655 1656 flags = flags & F2FS_FL_USER_MODIFIABLE; 1657 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE; 1658 fi->i_flags = flags; 1659 1660 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1661 set_inode_flag(inode, FI_PROJ_INHERIT); 1662 else 1663 clear_inode_flag(inode, FI_PROJ_INHERIT); 1664 1665 inode->i_ctime = current_time(inode); 1666 f2fs_set_inode_flags(inode); 1667 f2fs_mark_inode_dirty_sync(inode, false); 1668 return 0; 1669 } 1670 1671 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg) 1672 { 1673 struct inode *inode = file_inode(filp); 1674 unsigned int flags; 1675 int ret; 1676 1677 if (!inode_owner_or_capable(inode)) 1678 return -EACCES; 1679 1680 if (get_user(flags, (int __user *)arg)) 1681 return -EFAULT; 1682 1683 ret = mnt_want_write_file(filp); 1684 if (ret) 1685 return ret; 1686 1687 inode_lock(inode); 1688 1689 ret = __f2fs_ioc_setflags(inode, flags); 1690 1691 inode_unlock(inode); 1692 mnt_drop_write_file(filp); 1693 return ret; 1694 } 1695 1696 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1697 { 1698 struct inode *inode = file_inode(filp); 1699 1700 return put_user(inode->i_generation, (int __user *)arg); 1701 } 1702 1703 static int f2fs_ioc_start_atomic_write(struct file *filp) 1704 { 1705 struct inode *inode = file_inode(filp); 1706 int ret; 1707 1708 if (!inode_owner_or_capable(inode)) 1709 return -EACCES; 1710 1711 if (!S_ISREG(inode->i_mode)) 1712 return -EINVAL; 1713 1714 ret = mnt_want_write_file(filp); 1715 if (ret) 1716 return ret; 1717 1718 inode_lock(inode); 1719 1720 if (f2fs_is_atomic_file(inode)) { 1721 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) 1722 ret = -EINVAL; 1723 goto out; 1724 } 1725 1726 ret = f2fs_convert_inline_inode(inode); 1727 if (ret) 1728 goto out; 1729 1730 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1731 1732 if (!get_dirty_pages(inode)) 1733 goto skip_flush; 1734 1735 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING, 1736 "Unexpected flush for atomic writes: ino=%lu, npages=%u", 1737 inode->i_ino, get_dirty_pages(inode)); 1738 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 1739 if (ret) { 1740 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1741 goto out; 1742 } 1743 skip_flush: 1744 set_inode_flag(inode, FI_ATOMIC_FILE); 1745 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1746 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1747 1748 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1749 F2FS_I(inode)->inmem_task = current; 1750 stat_inc_atomic_write(inode); 1751 stat_update_max_atomic_write(inode); 1752 out: 1753 inode_unlock(inode); 1754 mnt_drop_write_file(filp); 1755 return ret; 1756 } 1757 1758 static int f2fs_ioc_commit_atomic_write(struct file *filp) 1759 { 1760 struct inode *inode = file_inode(filp); 1761 int ret; 1762 1763 if (!inode_owner_or_capable(inode)) 1764 return -EACCES; 1765 1766 ret = mnt_want_write_file(filp); 1767 if (ret) 1768 return ret; 1769 1770 f2fs_balance_fs(F2FS_I_SB(inode), true); 1771 1772 inode_lock(inode); 1773 1774 if (f2fs_is_volatile_file(inode)) { 1775 ret = -EINVAL; 1776 goto err_out; 1777 } 1778 1779 if (f2fs_is_atomic_file(inode)) { 1780 ret = f2fs_commit_inmem_pages(inode); 1781 if (ret) 1782 goto err_out; 1783 1784 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1785 if (!ret) { 1786 clear_inode_flag(inode, FI_ATOMIC_FILE); 1787 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0; 1788 stat_dec_atomic_write(inode); 1789 } 1790 } else { 1791 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 1792 } 1793 err_out: 1794 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) { 1795 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1796 ret = -EINVAL; 1797 } 1798 inode_unlock(inode); 1799 mnt_drop_write_file(filp); 1800 return ret; 1801 } 1802 1803 static int f2fs_ioc_start_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 if (!S_ISREG(inode->i_mode)) 1812 return -EINVAL; 1813 1814 ret = mnt_want_write_file(filp); 1815 if (ret) 1816 return ret; 1817 1818 inode_lock(inode); 1819 1820 if (f2fs_is_volatile_file(inode)) 1821 goto out; 1822 1823 ret = f2fs_convert_inline_inode(inode); 1824 if (ret) 1825 goto out; 1826 1827 stat_inc_volatile_write(inode); 1828 stat_update_max_volatile_write(inode); 1829 1830 set_inode_flag(inode, FI_VOLATILE_FILE); 1831 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1832 out: 1833 inode_unlock(inode); 1834 mnt_drop_write_file(filp); 1835 return ret; 1836 } 1837 1838 static int f2fs_ioc_release_volatile_write(struct file *filp) 1839 { 1840 struct inode *inode = file_inode(filp); 1841 int ret; 1842 1843 if (!inode_owner_or_capable(inode)) 1844 return -EACCES; 1845 1846 ret = mnt_want_write_file(filp); 1847 if (ret) 1848 return ret; 1849 1850 inode_lock(inode); 1851 1852 if (!f2fs_is_volatile_file(inode)) 1853 goto out; 1854 1855 if (!f2fs_is_first_block_written(inode)) { 1856 ret = truncate_partial_data_page(inode, 0, true); 1857 goto out; 1858 } 1859 1860 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 1861 out: 1862 inode_unlock(inode); 1863 mnt_drop_write_file(filp); 1864 return ret; 1865 } 1866 1867 static int f2fs_ioc_abort_volatile_write(struct file *filp) 1868 { 1869 struct inode *inode = file_inode(filp); 1870 int ret; 1871 1872 if (!inode_owner_or_capable(inode)) 1873 return -EACCES; 1874 1875 ret = mnt_want_write_file(filp); 1876 if (ret) 1877 return ret; 1878 1879 inode_lock(inode); 1880 1881 if (f2fs_is_atomic_file(inode)) 1882 f2fs_drop_inmem_pages(inode); 1883 if (f2fs_is_volatile_file(inode)) { 1884 clear_inode_flag(inode, FI_VOLATILE_FILE); 1885 stat_dec_volatile_write(inode); 1886 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1887 } 1888 1889 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1890 1891 inode_unlock(inode); 1892 1893 mnt_drop_write_file(filp); 1894 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1895 return ret; 1896 } 1897 1898 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 1899 { 1900 struct inode *inode = file_inode(filp); 1901 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1902 struct super_block *sb = sbi->sb; 1903 __u32 in; 1904 int ret = 0; 1905 1906 if (!capable(CAP_SYS_ADMIN)) 1907 return -EPERM; 1908 1909 if (get_user(in, (__u32 __user *)arg)) 1910 return -EFAULT; 1911 1912 if (in != F2FS_GOING_DOWN_FULLSYNC) { 1913 ret = mnt_want_write_file(filp); 1914 if (ret) 1915 return ret; 1916 } 1917 1918 switch (in) { 1919 case F2FS_GOING_DOWN_FULLSYNC: 1920 sb = freeze_bdev(sb->s_bdev); 1921 if (IS_ERR(sb)) { 1922 ret = PTR_ERR(sb); 1923 goto out; 1924 } 1925 if (sb) { 1926 f2fs_stop_checkpoint(sbi, false); 1927 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1928 thaw_bdev(sb->s_bdev, sb); 1929 } 1930 break; 1931 case F2FS_GOING_DOWN_METASYNC: 1932 /* do checkpoint only */ 1933 ret = f2fs_sync_fs(sb, 1); 1934 if (ret) 1935 goto out; 1936 f2fs_stop_checkpoint(sbi, false); 1937 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1938 break; 1939 case F2FS_GOING_DOWN_NOSYNC: 1940 f2fs_stop_checkpoint(sbi, false); 1941 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1942 break; 1943 case F2FS_GOING_DOWN_METAFLUSH: 1944 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 1945 f2fs_stop_checkpoint(sbi, false); 1946 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1947 break; 1948 default: 1949 ret = -EINVAL; 1950 goto out; 1951 } 1952 1953 f2fs_stop_gc_thread(sbi); 1954 f2fs_stop_discard_thread(sbi); 1955 1956 f2fs_drop_discard_cmd(sbi); 1957 clear_opt(sbi, DISCARD); 1958 1959 f2fs_update_time(sbi, REQ_TIME); 1960 out: 1961 if (in != F2FS_GOING_DOWN_FULLSYNC) 1962 mnt_drop_write_file(filp); 1963 return ret; 1964 } 1965 1966 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 1967 { 1968 struct inode *inode = file_inode(filp); 1969 struct super_block *sb = inode->i_sb; 1970 struct request_queue *q = bdev_get_queue(sb->s_bdev); 1971 struct fstrim_range range; 1972 int ret; 1973 1974 if (!capable(CAP_SYS_ADMIN)) 1975 return -EPERM; 1976 1977 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 1978 return -EOPNOTSUPP; 1979 1980 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 1981 sizeof(range))) 1982 return -EFAULT; 1983 1984 ret = mnt_want_write_file(filp); 1985 if (ret) 1986 return ret; 1987 1988 range.minlen = max((unsigned int)range.minlen, 1989 q->limits.discard_granularity); 1990 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 1991 mnt_drop_write_file(filp); 1992 if (ret < 0) 1993 return ret; 1994 1995 if (copy_to_user((struct fstrim_range __user *)arg, &range, 1996 sizeof(range))) 1997 return -EFAULT; 1998 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1999 return 0; 2000 } 2001 2002 static bool uuid_is_nonzero(__u8 u[16]) 2003 { 2004 int i; 2005 2006 for (i = 0; i < 16; i++) 2007 if (u[i]) 2008 return true; 2009 return false; 2010 } 2011 2012 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2013 { 2014 struct inode *inode = file_inode(filp); 2015 2016 if (!f2fs_sb_has_encrypt(inode->i_sb)) 2017 return -EOPNOTSUPP; 2018 2019 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2020 2021 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2022 } 2023 2024 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2025 { 2026 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb)) 2027 return -EOPNOTSUPP; 2028 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2029 } 2030 2031 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2032 { 2033 struct inode *inode = file_inode(filp); 2034 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2035 int err; 2036 2037 if (!f2fs_sb_has_encrypt(inode->i_sb)) 2038 return -EOPNOTSUPP; 2039 2040 err = mnt_want_write_file(filp); 2041 if (err) 2042 return err; 2043 2044 down_write(&sbi->sb_lock); 2045 2046 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2047 goto got_it; 2048 2049 /* update superblock with uuid */ 2050 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2051 2052 err = f2fs_commit_super(sbi, false); 2053 if (err) { 2054 /* undo new data */ 2055 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2056 goto out_err; 2057 } 2058 got_it: 2059 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 2060 16)) 2061 err = -EFAULT; 2062 out_err: 2063 up_write(&sbi->sb_lock); 2064 mnt_drop_write_file(filp); 2065 return err; 2066 } 2067 2068 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2069 { 2070 struct inode *inode = file_inode(filp); 2071 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2072 __u32 sync; 2073 int ret; 2074 2075 if (!capable(CAP_SYS_ADMIN)) 2076 return -EPERM; 2077 2078 if (get_user(sync, (__u32 __user *)arg)) 2079 return -EFAULT; 2080 2081 if (f2fs_readonly(sbi->sb)) 2082 return -EROFS; 2083 2084 ret = mnt_want_write_file(filp); 2085 if (ret) 2086 return ret; 2087 2088 if (!sync) { 2089 if (!mutex_trylock(&sbi->gc_mutex)) { 2090 ret = -EBUSY; 2091 goto out; 2092 } 2093 } else { 2094 mutex_lock(&sbi->gc_mutex); 2095 } 2096 2097 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO); 2098 out: 2099 mnt_drop_write_file(filp); 2100 return ret; 2101 } 2102 2103 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2104 { 2105 struct inode *inode = file_inode(filp); 2106 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2107 struct f2fs_gc_range range; 2108 u64 end; 2109 int ret; 2110 2111 if (!capable(CAP_SYS_ADMIN)) 2112 return -EPERM; 2113 2114 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2115 sizeof(range))) 2116 return -EFAULT; 2117 2118 if (f2fs_readonly(sbi->sb)) 2119 return -EROFS; 2120 2121 end = range.start + range.len; 2122 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) { 2123 return -EINVAL; 2124 } 2125 2126 ret = mnt_want_write_file(filp); 2127 if (ret) 2128 return ret; 2129 2130 do_more: 2131 if (!range.sync) { 2132 if (!mutex_trylock(&sbi->gc_mutex)) { 2133 ret = -EBUSY; 2134 goto out; 2135 } 2136 } else { 2137 mutex_lock(&sbi->gc_mutex); 2138 } 2139 2140 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start)); 2141 range.start += sbi->blocks_per_seg; 2142 if (range.start <= end) 2143 goto do_more; 2144 out: 2145 mnt_drop_write_file(filp); 2146 return ret; 2147 } 2148 2149 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg) 2150 { 2151 struct inode *inode = file_inode(filp); 2152 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2153 int ret; 2154 2155 if (!capable(CAP_SYS_ADMIN)) 2156 return -EPERM; 2157 2158 if (f2fs_readonly(sbi->sb)) 2159 return -EROFS; 2160 2161 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2162 f2fs_msg(sbi->sb, KERN_INFO, 2163 "Skipping Checkpoint. Checkpoints currently disabled."); 2164 return -EINVAL; 2165 } 2166 2167 ret = mnt_want_write_file(filp); 2168 if (ret) 2169 return ret; 2170 2171 ret = f2fs_sync_fs(sbi->sb, 1); 2172 2173 mnt_drop_write_file(filp); 2174 return ret; 2175 } 2176 2177 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2178 struct file *filp, 2179 struct f2fs_defragment *range) 2180 { 2181 struct inode *inode = file_inode(filp); 2182 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2183 .m_seg_type = NO_CHECK_TYPE }; 2184 struct extent_info ei = {0, 0, 0}; 2185 pgoff_t pg_start, pg_end, next_pgofs; 2186 unsigned int blk_per_seg = sbi->blocks_per_seg; 2187 unsigned int total = 0, sec_num; 2188 block_t blk_end = 0; 2189 bool fragmented = false; 2190 int err; 2191 2192 /* if in-place-update policy is enabled, don't waste time here */ 2193 if (f2fs_should_update_inplace(inode, NULL)) 2194 return -EINVAL; 2195 2196 pg_start = range->start >> PAGE_SHIFT; 2197 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2198 2199 f2fs_balance_fs(sbi, true); 2200 2201 inode_lock(inode); 2202 2203 /* writeback all dirty pages in the range */ 2204 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2205 range->start + range->len - 1); 2206 if (err) 2207 goto out; 2208 2209 /* 2210 * lookup mapping info in extent cache, skip defragmenting if physical 2211 * block addresses are continuous. 2212 */ 2213 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) { 2214 if (ei.fofs + ei.len >= pg_end) 2215 goto out; 2216 } 2217 2218 map.m_lblk = pg_start; 2219 map.m_next_pgofs = &next_pgofs; 2220 2221 /* 2222 * lookup mapping info in dnode page cache, skip defragmenting if all 2223 * physical block addresses are continuous even if there are hole(s) 2224 * in logical blocks. 2225 */ 2226 while (map.m_lblk < pg_end) { 2227 map.m_len = pg_end - map.m_lblk; 2228 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2229 if (err) 2230 goto out; 2231 2232 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2233 map.m_lblk = next_pgofs; 2234 continue; 2235 } 2236 2237 if (blk_end && blk_end != map.m_pblk) 2238 fragmented = true; 2239 2240 /* record total count of block that we're going to move */ 2241 total += map.m_len; 2242 2243 blk_end = map.m_pblk + map.m_len; 2244 2245 map.m_lblk += map.m_len; 2246 } 2247 2248 if (!fragmented) 2249 goto out; 2250 2251 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi); 2252 2253 /* 2254 * make sure there are enough free section for LFS allocation, this can 2255 * avoid defragment running in SSR mode when free section are allocated 2256 * intensively 2257 */ 2258 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2259 err = -EAGAIN; 2260 goto out; 2261 } 2262 2263 map.m_lblk = pg_start; 2264 map.m_len = pg_end - pg_start; 2265 total = 0; 2266 2267 while (map.m_lblk < pg_end) { 2268 pgoff_t idx; 2269 int cnt = 0; 2270 2271 do_map: 2272 map.m_len = pg_end - map.m_lblk; 2273 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2274 if (err) 2275 goto clear_out; 2276 2277 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2278 map.m_lblk = next_pgofs; 2279 continue; 2280 } 2281 2282 set_inode_flag(inode, FI_DO_DEFRAG); 2283 2284 idx = map.m_lblk; 2285 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2286 struct page *page; 2287 2288 page = f2fs_get_lock_data_page(inode, idx, true); 2289 if (IS_ERR(page)) { 2290 err = PTR_ERR(page); 2291 goto clear_out; 2292 } 2293 2294 set_page_dirty(page); 2295 f2fs_put_page(page, 1); 2296 2297 idx++; 2298 cnt++; 2299 total++; 2300 } 2301 2302 map.m_lblk = idx; 2303 2304 if (idx < pg_end && cnt < blk_per_seg) 2305 goto do_map; 2306 2307 clear_inode_flag(inode, FI_DO_DEFRAG); 2308 2309 err = filemap_fdatawrite(inode->i_mapping); 2310 if (err) 2311 goto out; 2312 } 2313 clear_out: 2314 clear_inode_flag(inode, FI_DO_DEFRAG); 2315 out: 2316 inode_unlock(inode); 2317 if (!err) 2318 range->len = (u64)total << PAGE_SHIFT; 2319 return err; 2320 } 2321 2322 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2323 { 2324 struct inode *inode = file_inode(filp); 2325 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2326 struct f2fs_defragment range; 2327 int err; 2328 2329 if (!capable(CAP_SYS_ADMIN)) 2330 return -EPERM; 2331 2332 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2333 return -EINVAL; 2334 2335 if (f2fs_readonly(sbi->sb)) 2336 return -EROFS; 2337 2338 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2339 sizeof(range))) 2340 return -EFAULT; 2341 2342 /* verify alignment of offset & size */ 2343 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2344 return -EINVAL; 2345 2346 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2347 sbi->max_file_blocks)) 2348 return -EINVAL; 2349 2350 err = mnt_want_write_file(filp); 2351 if (err) 2352 return err; 2353 2354 err = f2fs_defragment_range(sbi, filp, &range); 2355 mnt_drop_write_file(filp); 2356 2357 f2fs_update_time(sbi, REQ_TIME); 2358 if (err < 0) 2359 return err; 2360 2361 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2362 sizeof(range))) 2363 return -EFAULT; 2364 2365 return 0; 2366 } 2367 2368 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2369 struct file *file_out, loff_t pos_out, size_t len) 2370 { 2371 struct inode *src = file_inode(file_in); 2372 struct inode *dst = file_inode(file_out); 2373 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2374 size_t olen = len, dst_max_i_size = 0; 2375 size_t dst_osize; 2376 int ret; 2377 2378 if (file_in->f_path.mnt != file_out->f_path.mnt || 2379 src->i_sb != dst->i_sb) 2380 return -EXDEV; 2381 2382 if (unlikely(f2fs_readonly(src->i_sb))) 2383 return -EROFS; 2384 2385 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2386 return -EINVAL; 2387 2388 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst)) 2389 return -EOPNOTSUPP; 2390 2391 if (src == dst) { 2392 if (pos_in == pos_out) 2393 return 0; 2394 if (pos_out > pos_in && pos_out < pos_in + len) 2395 return -EINVAL; 2396 } 2397 2398 inode_lock(src); 2399 if (src != dst) { 2400 ret = -EBUSY; 2401 if (!inode_trylock(dst)) 2402 goto out; 2403 } 2404 2405 ret = -EINVAL; 2406 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2407 goto out_unlock; 2408 if (len == 0) 2409 olen = len = src->i_size - pos_in; 2410 if (pos_in + len == src->i_size) 2411 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2412 if (len == 0) { 2413 ret = 0; 2414 goto out_unlock; 2415 } 2416 2417 dst_osize = dst->i_size; 2418 if (pos_out + olen > dst->i_size) 2419 dst_max_i_size = pos_out + olen; 2420 2421 /* verify the end result is block aligned */ 2422 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2423 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2424 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2425 goto out_unlock; 2426 2427 ret = f2fs_convert_inline_inode(src); 2428 if (ret) 2429 goto out_unlock; 2430 2431 ret = f2fs_convert_inline_inode(dst); 2432 if (ret) 2433 goto out_unlock; 2434 2435 /* write out all dirty pages from offset */ 2436 ret = filemap_write_and_wait_range(src->i_mapping, 2437 pos_in, pos_in + len); 2438 if (ret) 2439 goto out_unlock; 2440 2441 ret = filemap_write_and_wait_range(dst->i_mapping, 2442 pos_out, pos_out + len); 2443 if (ret) 2444 goto out_unlock; 2445 2446 f2fs_balance_fs(sbi, true); 2447 2448 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2449 if (src != dst) { 2450 ret = -EBUSY; 2451 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2452 goto out_src; 2453 } 2454 2455 f2fs_lock_op(sbi); 2456 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2457 pos_out >> F2FS_BLKSIZE_BITS, 2458 len >> F2FS_BLKSIZE_BITS, false); 2459 2460 if (!ret) { 2461 if (dst_max_i_size) 2462 f2fs_i_size_write(dst, dst_max_i_size); 2463 else if (dst_osize != dst->i_size) 2464 f2fs_i_size_write(dst, dst_osize); 2465 } 2466 f2fs_unlock_op(sbi); 2467 2468 if (src != dst) 2469 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2470 out_src: 2471 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2472 out_unlock: 2473 if (src != dst) 2474 inode_unlock(dst); 2475 out: 2476 inode_unlock(src); 2477 return ret; 2478 } 2479 2480 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2481 { 2482 struct f2fs_move_range range; 2483 struct fd dst; 2484 int err; 2485 2486 if (!(filp->f_mode & FMODE_READ) || 2487 !(filp->f_mode & FMODE_WRITE)) 2488 return -EBADF; 2489 2490 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2491 sizeof(range))) 2492 return -EFAULT; 2493 2494 dst = fdget(range.dst_fd); 2495 if (!dst.file) 2496 return -EBADF; 2497 2498 if (!(dst.file->f_mode & FMODE_WRITE)) { 2499 err = -EBADF; 2500 goto err_out; 2501 } 2502 2503 err = mnt_want_write_file(filp); 2504 if (err) 2505 goto err_out; 2506 2507 err = f2fs_move_file_range(filp, range.pos_in, dst.file, 2508 range.pos_out, range.len); 2509 2510 mnt_drop_write_file(filp); 2511 if (err) 2512 goto err_out; 2513 2514 if (copy_to_user((struct f2fs_move_range __user *)arg, 2515 &range, sizeof(range))) 2516 err = -EFAULT; 2517 err_out: 2518 fdput(dst); 2519 return err; 2520 } 2521 2522 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2523 { 2524 struct inode *inode = file_inode(filp); 2525 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2526 struct sit_info *sm = SIT_I(sbi); 2527 unsigned int start_segno = 0, end_segno = 0; 2528 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2529 struct f2fs_flush_device range; 2530 int ret; 2531 2532 if (!capable(CAP_SYS_ADMIN)) 2533 return -EPERM; 2534 2535 if (f2fs_readonly(sbi->sb)) 2536 return -EROFS; 2537 2538 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2539 return -EINVAL; 2540 2541 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2542 sizeof(range))) 2543 return -EFAULT; 2544 2545 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num || 2546 sbi->segs_per_sec != 1) { 2547 f2fs_msg(sbi->sb, KERN_WARNING, 2548 "Can't flush %u in %d for segs_per_sec %u != 1\n", 2549 range.dev_num, sbi->s_ndevs, 2550 sbi->segs_per_sec); 2551 return -EINVAL; 2552 } 2553 2554 ret = mnt_want_write_file(filp); 2555 if (ret) 2556 return ret; 2557 2558 if (range.dev_num != 0) 2559 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2560 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2561 2562 start_segno = sm->last_victim[FLUSH_DEVICE]; 2563 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2564 start_segno = dev_start_segno; 2565 end_segno = min(start_segno + range.segments, dev_end_segno); 2566 2567 while (start_segno < end_segno) { 2568 if (!mutex_trylock(&sbi->gc_mutex)) { 2569 ret = -EBUSY; 2570 goto out; 2571 } 2572 sm->last_victim[GC_CB] = end_segno + 1; 2573 sm->last_victim[GC_GREEDY] = end_segno + 1; 2574 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 2575 ret = f2fs_gc(sbi, true, true, start_segno); 2576 if (ret == -EAGAIN) 2577 ret = 0; 2578 else if (ret < 0) 2579 break; 2580 start_segno++; 2581 } 2582 out: 2583 mnt_drop_write_file(filp); 2584 return ret; 2585 } 2586 2587 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 2588 { 2589 struct inode *inode = file_inode(filp); 2590 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 2591 2592 /* Must validate to set it with SQLite behavior in Android. */ 2593 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 2594 2595 return put_user(sb_feature, (u32 __user *)arg); 2596 } 2597 2598 #ifdef CONFIG_QUOTA 2599 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 2600 { 2601 struct inode *inode = file_inode(filp); 2602 struct f2fs_inode_info *fi = F2FS_I(inode); 2603 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2604 struct super_block *sb = sbi->sb; 2605 struct dquot *transfer_to[MAXQUOTAS] = {}; 2606 struct page *ipage; 2607 kprojid_t kprojid; 2608 int err; 2609 2610 if (!f2fs_sb_has_project_quota(sb)) { 2611 if (projid != F2FS_DEF_PROJID) 2612 return -EOPNOTSUPP; 2613 else 2614 return 0; 2615 } 2616 2617 if (!f2fs_has_extra_attr(inode)) 2618 return -EOPNOTSUPP; 2619 2620 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 2621 2622 if (projid_eq(kprojid, F2FS_I(inode)->i_projid)) 2623 return 0; 2624 2625 err = -EPERM; 2626 /* Is it quota file? Do not allow user to mess with it */ 2627 if (IS_NOQUOTA(inode)) 2628 return err; 2629 2630 ipage = f2fs_get_node_page(sbi, inode->i_ino); 2631 if (IS_ERR(ipage)) 2632 return PTR_ERR(ipage); 2633 2634 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize, 2635 i_projid)) { 2636 err = -EOVERFLOW; 2637 f2fs_put_page(ipage, 1); 2638 return err; 2639 } 2640 f2fs_put_page(ipage, 1); 2641 2642 err = dquot_initialize(inode); 2643 if (err) 2644 return err; 2645 2646 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 2647 if (!IS_ERR(transfer_to[PRJQUOTA])) { 2648 err = __dquot_transfer(inode, transfer_to); 2649 dqput(transfer_to[PRJQUOTA]); 2650 if (err) 2651 goto out_dirty; 2652 } 2653 2654 F2FS_I(inode)->i_projid = kprojid; 2655 inode->i_ctime = current_time(inode); 2656 out_dirty: 2657 f2fs_mark_inode_dirty_sync(inode, true); 2658 return err; 2659 } 2660 #else 2661 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 2662 { 2663 if (projid != F2FS_DEF_PROJID) 2664 return -EOPNOTSUPP; 2665 return 0; 2666 } 2667 #endif 2668 2669 /* Transfer internal flags to xflags */ 2670 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags) 2671 { 2672 __u32 xflags = 0; 2673 2674 if (iflags & F2FS_SYNC_FL) 2675 xflags |= FS_XFLAG_SYNC; 2676 if (iflags & F2FS_IMMUTABLE_FL) 2677 xflags |= FS_XFLAG_IMMUTABLE; 2678 if (iflags & F2FS_APPEND_FL) 2679 xflags |= FS_XFLAG_APPEND; 2680 if (iflags & F2FS_NODUMP_FL) 2681 xflags |= FS_XFLAG_NODUMP; 2682 if (iflags & F2FS_NOATIME_FL) 2683 xflags |= FS_XFLAG_NOATIME; 2684 if (iflags & F2FS_PROJINHERIT_FL) 2685 xflags |= FS_XFLAG_PROJINHERIT; 2686 return xflags; 2687 } 2688 2689 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \ 2690 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \ 2691 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT) 2692 2693 /* Transfer xflags flags to internal */ 2694 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags) 2695 { 2696 unsigned long iflags = 0; 2697 2698 if (xflags & FS_XFLAG_SYNC) 2699 iflags |= F2FS_SYNC_FL; 2700 if (xflags & FS_XFLAG_IMMUTABLE) 2701 iflags |= F2FS_IMMUTABLE_FL; 2702 if (xflags & FS_XFLAG_APPEND) 2703 iflags |= F2FS_APPEND_FL; 2704 if (xflags & FS_XFLAG_NODUMP) 2705 iflags |= F2FS_NODUMP_FL; 2706 if (xflags & FS_XFLAG_NOATIME) 2707 iflags |= F2FS_NOATIME_FL; 2708 if (xflags & FS_XFLAG_PROJINHERIT) 2709 iflags |= F2FS_PROJINHERIT_FL; 2710 2711 return iflags; 2712 } 2713 2714 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg) 2715 { 2716 struct inode *inode = file_inode(filp); 2717 struct f2fs_inode_info *fi = F2FS_I(inode); 2718 struct fsxattr fa; 2719 2720 memset(&fa, 0, sizeof(struct fsxattr)); 2721 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags & 2722 F2FS_FL_USER_VISIBLE); 2723 2724 if (f2fs_sb_has_project_quota(inode->i_sb)) 2725 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns, 2726 fi->i_projid); 2727 2728 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa))) 2729 return -EFAULT; 2730 return 0; 2731 } 2732 2733 static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa) 2734 { 2735 /* 2736 * Project Quota ID state is only allowed to change from within the init 2737 * namespace. Enforce that restriction only if we are trying to change 2738 * the quota ID state. Everything else is allowed in user namespaces. 2739 */ 2740 if (current_user_ns() == &init_user_ns) 2741 return 0; 2742 2743 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid) 2744 return -EINVAL; 2745 2746 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) { 2747 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT)) 2748 return -EINVAL; 2749 } else { 2750 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT) 2751 return -EINVAL; 2752 } 2753 2754 return 0; 2755 } 2756 2757 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg) 2758 { 2759 struct inode *inode = file_inode(filp); 2760 struct f2fs_inode_info *fi = F2FS_I(inode); 2761 struct fsxattr fa; 2762 unsigned int flags; 2763 int err; 2764 2765 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa))) 2766 return -EFAULT; 2767 2768 /* Make sure caller has proper permission */ 2769 if (!inode_owner_or_capable(inode)) 2770 return -EACCES; 2771 2772 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS) 2773 return -EOPNOTSUPP; 2774 2775 flags = f2fs_xflags_to_iflags(fa.fsx_xflags); 2776 if (f2fs_mask_flags(inode->i_mode, flags) != flags) 2777 return -EOPNOTSUPP; 2778 2779 err = mnt_want_write_file(filp); 2780 if (err) 2781 return err; 2782 2783 inode_lock(inode); 2784 err = f2fs_ioctl_check_project(inode, &fa); 2785 if (err) 2786 goto out; 2787 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) | 2788 (flags & F2FS_FL_XFLAG_VISIBLE); 2789 err = __f2fs_ioc_setflags(inode, flags); 2790 if (err) 2791 goto out; 2792 2793 err = f2fs_ioc_setproject(filp, fa.fsx_projid); 2794 out: 2795 inode_unlock(inode); 2796 mnt_drop_write_file(filp); 2797 return err; 2798 } 2799 2800 int f2fs_pin_file_control(struct inode *inode, bool inc) 2801 { 2802 struct f2fs_inode_info *fi = F2FS_I(inode); 2803 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2804 2805 /* Use i_gc_failures for normal file as a risk signal. */ 2806 if (inc) 2807 f2fs_i_gc_failures_write(inode, 2808 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 2809 2810 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 2811 f2fs_msg(sbi->sb, KERN_WARNING, 2812 "%s: Enable GC = ino %lx after %x GC trials\n", 2813 __func__, inode->i_ino, 2814 fi->i_gc_failures[GC_FAILURE_PIN]); 2815 clear_inode_flag(inode, FI_PIN_FILE); 2816 return -EAGAIN; 2817 } 2818 return 0; 2819 } 2820 2821 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 2822 { 2823 struct inode *inode = file_inode(filp); 2824 __u32 pin; 2825 int ret = 0; 2826 2827 if (!inode_owner_or_capable(inode)) 2828 return -EACCES; 2829 2830 if (get_user(pin, (__u32 __user *)arg)) 2831 return -EFAULT; 2832 2833 if (!S_ISREG(inode->i_mode)) 2834 return -EINVAL; 2835 2836 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 2837 return -EROFS; 2838 2839 ret = mnt_want_write_file(filp); 2840 if (ret) 2841 return ret; 2842 2843 inode_lock(inode); 2844 2845 if (f2fs_should_update_outplace(inode, NULL)) { 2846 ret = -EINVAL; 2847 goto out; 2848 } 2849 2850 if (!pin) { 2851 clear_inode_flag(inode, FI_PIN_FILE); 2852 f2fs_i_gc_failures_write(inode, 0); 2853 goto done; 2854 } 2855 2856 if (f2fs_pin_file_control(inode, false)) { 2857 ret = -EAGAIN; 2858 goto out; 2859 } 2860 ret = f2fs_convert_inline_inode(inode); 2861 if (ret) 2862 goto out; 2863 2864 set_inode_flag(inode, FI_PIN_FILE); 2865 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 2866 done: 2867 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2868 out: 2869 inode_unlock(inode); 2870 mnt_drop_write_file(filp); 2871 return ret; 2872 } 2873 2874 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 2875 { 2876 struct inode *inode = file_inode(filp); 2877 __u32 pin = 0; 2878 2879 if (is_inode_flag_set(inode, FI_PIN_FILE)) 2880 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 2881 return put_user(pin, (u32 __user *)arg); 2882 } 2883 2884 int f2fs_precache_extents(struct inode *inode) 2885 { 2886 struct f2fs_inode_info *fi = F2FS_I(inode); 2887 struct f2fs_map_blocks map; 2888 pgoff_t m_next_extent; 2889 loff_t end; 2890 int err; 2891 2892 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 2893 return -EOPNOTSUPP; 2894 2895 map.m_lblk = 0; 2896 map.m_next_pgofs = NULL; 2897 map.m_next_extent = &m_next_extent; 2898 map.m_seg_type = NO_CHECK_TYPE; 2899 end = F2FS_I_SB(inode)->max_file_blocks; 2900 2901 while (map.m_lblk < end) { 2902 map.m_len = end - map.m_lblk; 2903 2904 down_write(&fi->i_gc_rwsem[WRITE]); 2905 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE); 2906 up_write(&fi->i_gc_rwsem[WRITE]); 2907 if (err) 2908 return err; 2909 2910 map.m_lblk = m_next_extent; 2911 } 2912 2913 return err; 2914 } 2915 2916 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg) 2917 { 2918 return f2fs_precache_extents(file_inode(filp)); 2919 } 2920 2921 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 2922 { 2923 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 2924 return -EIO; 2925 2926 switch (cmd) { 2927 case F2FS_IOC_GETFLAGS: 2928 return f2fs_ioc_getflags(filp, arg); 2929 case F2FS_IOC_SETFLAGS: 2930 return f2fs_ioc_setflags(filp, arg); 2931 case F2FS_IOC_GETVERSION: 2932 return f2fs_ioc_getversion(filp, arg); 2933 case F2FS_IOC_START_ATOMIC_WRITE: 2934 return f2fs_ioc_start_atomic_write(filp); 2935 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 2936 return f2fs_ioc_commit_atomic_write(filp); 2937 case F2FS_IOC_START_VOLATILE_WRITE: 2938 return f2fs_ioc_start_volatile_write(filp); 2939 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 2940 return f2fs_ioc_release_volatile_write(filp); 2941 case F2FS_IOC_ABORT_VOLATILE_WRITE: 2942 return f2fs_ioc_abort_volatile_write(filp); 2943 case F2FS_IOC_SHUTDOWN: 2944 return f2fs_ioc_shutdown(filp, arg); 2945 case FITRIM: 2946 return f2fs_ioc_fitrim(filp, arg); 2947 case F2FS_IOC_SET_ENCRYPTION_POLICY: 2948 return f2fs_ioc_set_encryption_policy(filp, arg); 2949 case F2FS_IOC_GET_ENCRYPTION_POLICY: 2950 return f2fs_ioc_get_encryption_policy(filp, arg); 2951 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 2952 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 2953 case F2FS_IOC_GARBAGE_COLLECT: 2954 return f2fs_ioc_gc(filp, arg); 2955 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 2956 return f2fs_ioc_gc_range(filp, arg); 2957 case F2FS_IOC_WRITE_CHECKPOINT: 2958 return f2fs_ioc_write_checkpoint(filp, arg); 2959 case F2FS_IOC_DEFRAGMENT: 2960 return f2fs_ioc_defragment(filp, arg); 2961 case F2FS_IOC_MOVE_RANGE: 2962 return f2fs_ioc_move_range(filp, arg); 2963 case F2FS_IOC_FLUSH_DEVICE: 2964 return f2fs_ioc_flush_device(filp, arg); 2965 case F2FS_IOC_GET_FEATURES: 2966 return f2fs_ioc_get_features(filp, arg); 2967 case F2FS_IOC_FSGETXATTR: 2968 return f2fs_ioc_fsgetxattr(filp, arg); 2969 case F2FS_IOC_FSSETXATTR: 2970 return f2fs_ioc_fssetxattr(filp, arg); 2971 case F2FS_IOC_GET_PIN_FILE: 2972 return f2fs_ioc_get_pin_file(filp, arg); 2973 case F2FS_IOC_SET_PIN_FILE: 2974 return f2fs_ioc_set_pin_file(filp, arg); 2975 case F2FS_IOC_PRECACHE_EXTENTS: 2976 return f2fs_ioc_precache_extents(filp, arg); 2977 default: 2978 return -ENOTTY; 2979 } 2980 } 2981 2982 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 2983 { 2984 struct file *file = iocb->ki_filp; 2985 struct inode *inode = file_inode(file); 2986 ssize_t ret; 2987 2988 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 2989 return -EIO; 2990 2991 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) 2992 return -EINVAL; 2993 2994 if (!inode_trylock(inode)) { 2995 if (iocb->ki_flags & IOCB_NOWAIT) 2996 return -EAGAIN; 2997 inode_lock(inode); 2998 } 2999 3000 ret = generic_write_checks(iocb, from); 3001 if (ret > 0) { 3002 bool preallocated = false; 3003 size_t target_size = 0; 3004 int err; 3005 3006 if (iov_iter_fault_in_readable(from, iov_iter_count(from))) 3007 set_inode_flag(inode, FI_NO_PREALLOC); 3008 3009 if ((iocb->ki_flags & IOCB_NOWAIT) && 3010 (iocb->ki_flags & IOCB_DIRECT)) { 3011 if (!f2fs_overwrite_io(inode, iocb->ki_pos, 3012 iov_iter_count(from)) || 3013 f2fs_has_inline_data(inode) || 3014 f2fs_force_buffered_io(inode, 3015 iocb, from)) { 3016 clear_inode_flag(inode, 3017 FI_NO_PREALLOC); 3018 inode_unlock(inode); 3019 return -EAGAIN; 3020 } 3021 3022 } else { 3023 preallocated = true; 3024 target_size = iocb->ki_pos + iov_iter_count(from); 3025 3026 err = f2fs_preallocate_blocks(iocb, from); 3027 if (err) { 3028 clear_inode_flag(inode, FI_NO_PREALLOC); 3029 inode_unlock(inode); 3030 return err; 3031 } 3032 } 3033 ret = __generic_file_write_iter(iocb, from); 3034 clear_inode_flag(inode, FI_NO_PREALLOC); 3035 3036 /* if we couldn't write data, we should deallocate blocks. */ 3037 if (preallocated && i_size_read(inode) < target_size) 3038 f2fs_truncate(inode); 3039 3040 if (ret > 0) 3041 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret); 3042 } 3043 inode_unlock(inode); 3044 3045 if (ret > 0) 3046 ret = generic_write_sync(iocb, ret); 3047 return ret; 3048 } 3049 3050 #ifdef CONFIG_COMPAT 3051 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 3052 { 3053 switch (cmd) { 3054 case F2FS_IOC32_GETFLAGS: 3055 cmd = F2FS_IOC_GETFLAGS; 3056 break; 3057 case F2FS_IOC32_SETFLAGS: 3058 cmd = F2FS_IOC_SETFLAGS; 3059 break; 3060 case F2FS_IOC32_GETVERSION: 3061 cmd = F2FS_IOC_GETVERSION; 3062 break; 3063 case F2FS_IOC_START_ATOMIC_WRITE: 3064 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 3065 case F2FS_IOC_START_VOLATILE_WRITE: 3066 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 3067 case F2FS_IOC_ABORT_VOLATILE_WRITE: 3068 case F2FS_IOC_SHUTDOWN: 3069 case F2FS_IOC_SET_ENCRYPTION_POLICY: 3070 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 3071 case F2FS_IOC_GET_ENCRYPTION_POLICY: 3072 case F2FS_IOC_GARBAGE_COLLECT: 3073 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 3074 case F2FS_IOC_WRITE_CHECKPOINT: 3075 case F2FS_IOC_DEFRAGMENT: 3076 case F2FS_IOC_MOVE_RANGE: 3077 case F2FS_IOC_FLUSH_DEVICE: 3078 case F2FS_IOC_GET_FEATURES: 3079 case F2FS_IOC_FSGETXATTR: 3080 case F2FS_IOC_FSSETXATTR: 3081 case F2FS_IOC_GET_PIN_FILE: 3082 case F2FS_IOC_SET_PIN_FILE: 3083 case F2FS_IOC_PRECACHE_EXTENTS: 3084 break; 3085 default: 3086 return -ENOIOCTLCMD; 3087 } 3088 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 3089 } 3090 #endif 3091 3092 const struct file_operations f2fs_file_operations = { 3093 .llseek = f2fs_llseek, 3094 .read_iter = generic_file_read_iter, 3095 .write_iter = f2fs_file_write_iter, 3096 .open = f2fs_file_open, 3097 .release = f2fs_release_file, 3098 .mmap = f2fs_file_mmap, 3099 .flush = f2fs_file_flush, 3100 .fsync = f2fs_sync_file, 3101 .fallocate = f2fs_fallocate, 3102 .unlocked_ioctl = f2fs_ioctl, 3103 #ifdef CONFIG_COMPAT 3104 .compat_ioctl = f2fs_compat_ioctl, 3105 #endif 3106 .splice_read = generic_file_splice_read, 3107 .splice_write = iter_file_splice_write, 3108 }; 3109