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 f2fs_update_time(sbi, REQ_TIME); 110 111 trace_f2fs_vm_page_mkwrite(page, DATA); 112 out_sem: 113 up_read(&F2FS_I(inode)->i_mmap_sem); 114 115 f2fs_balance_fs(sbi, dn.node_changed); 116 117 sb_end_pagefault(inode->i_sb); 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 bool buf_write) 591 { 592 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 593 struct dnode_of_data dn; 594 pgoff_t free_from; 595 int count = 0, err = 0; 596 struct page *ipage; 597 bool truncate_page = false; 598 int flag = buf_write ? F2FS_GET_BLOCK_PRE_AIO : F2FS_GET_BLOCK_PRE_DIO; 599 600 trace_f2fs_truncate_blocks_enter(inode, from); 601 602 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 603 604 if (free_from >= sbi->max_file_blocks) 605 goto free_partial; 606 607 if (lock) 608 __do_map_lock(sbi, flag, true); 609 610 ipage = f2fs_get_node_page(sbi, inode->i_ino); 611 if (IS_ERR(ipage)) { 612 err = PTR_ERR(ipage); 613 goto out; 614 } 615 616 if (f2fs_has_inline_data(inode)) { 617 f2fs_truncate_inline_inode(inode, ipage, from); 618 f2fs_put_page(ipage, 1); 619 truncate_page = true; 620 goto out; 621 } 622 623 set_new_dnode(&dn, inode, ipage, NULL, 0); 624 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 625 if (err) { 626 if (err == -ENOENT) 627 goto free_next; 628 goto out; 629 } 630 631 count = ADDRS_PER_PAGE(dn.node_page, inode); 632 633 count -= dn.ofs_in_node; 634 f2fs_bug_on(sbi, count < 0); 635 636 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 637 f2fs_truncate_data_blocks_range(&dn, count); 638 free_from += count; 639 } 640 641 f2fs_put_dnode(&dn); 642 free_next: 643 err = f2fs_truncate_inode_blocks(inode, free_from); 644 out: 645 if (lock) 646 __do_map_lock(sbi, flag, false); 647 free_partial: 648 /* lastly zero out the first data page */ 649 if (!err) 650 err = truncate_partial_data_page(inode, from, truncate_page); 651 652 trace_f2fs_truncate_blocks_exit(inode, err); 653 return err; 654 } 655 656 int f2fs_truncate(struct inode *inode) 657 { 658 int err; 659 660 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 661 return -EIO; 662 663 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 664 S_ISLNK(inode->i_mode))) 665 return 0; 666 667 trace_f2fs_truncate(inode); 668 669 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) { 670 f2fs_show_injection_info(FAULT_TRUNCATE); 671 return -EIO; 672 } 673 674 /* we should check inline_data size */ 675 if (!f2fs_may_inline_data(inode)) { 676 err = f2fs_convert_inline_inode(inode); 677 if (err) 678 return err; 679 } 680 681 err = f2fs_truncate_blocks(inode, i_size_read(inode), true, false); 682 if (err) 683 return err; 684 685 inode->i_mtime = inode->i_ctime = current_time(inode); 686 f2fs_mark_inode_dirty_sync(inode, false); 687 return 0; 688 } 689 690 int f2fs_getattr(const struct path *path, struct kstat *stat, 691 u32 request_mask, unsigned int query_flags) 692 { 693 struct inode *inode = d_inode(path->dentry); 694 struct f2fs_inode_info *fi = F2FS_I(inode); 695 struct f2fs_inode *ri; 696 unsigned int flags; 697 698 if (f2fs_has_extra_attr(inode) && 699 f2fs_sb_has_inode_crtime(inode->i_sb) && 700 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 701 stat->result_mask |= STATX_BTIME; 702 stat->btime.tv_sec = fi->i_crtime.tv_sec; 703 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 704 } 705 706 flags = fi->i_flags & F2FS_FL_USER_VISIBLE; 707 if (flags & F2FS_APPEND_FL) 708 stat->attributes |= STATX_ATTR_APPEND; 709 if (flags & F2FS_COMPR_FL) 710 stat->attributes |= STATX_ATTR_COMPRESSED; 711 if (f2fs_encrypted_inode(inode)) 712 stat->attributes |= STATX_ATTR_ENCRYPTED; 713 if (flags & F2FS_IMMUTABLE_FL) 714 stat->attributes |= STATX_ATTR_IMMUTABLE; 715 if (flags & F2FS_NODUMP_FL) 716 stat->attributes |= STATX_ATTR_NODUMP; 717 718 stat->attributes_mask |= (STATX_ATTR_APPEND | 719 STATX_ATTR_COMPRESSED | 720 STATX_ATTR_ENCRYPTED | 721 STATX_ATTR_IMMUTABLE | 722 STATX_ATTR_NODUMP); 723 724 generic_fillattr(inode, stat); 725 726 /* we need to show initial sectors used for inline_data/dentries */ 727 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 728 f2fs_has_inline_dentry(inode)) 729 stat->blocks += (stat->size + 511) >> 9; 730 731 return 0; 732 } 733 734 #ifdef CONFIG_F2FS_FS_POSIX_ACL 735 static void __setattr_copy(struct inode *inode, const struct iattr *attr) 736 { 737 unsigned int ia_valid = attr->ia_valid; 738 739 if (ia_valid & ATTR_UID) 740 inode->i_uid = attr->ia_uid; 741 if (ia_valid & ATTR_GID) 742 inode->i_gid = attr->ia_gid; 743 if (ia_valid & ATTR_ATIME) 744 inode->i_atime = timespec64_trunc(attr->ia_atime, 745 inode->i_sb->s_time_gran); 746 if (ia_valid & ATTR_MTIME) 747 inode->i_mtime = timespec64_trunc(attr->ia_mtime, 748 inode->i_sb->s_time_gran); 749 if (ia_valid & ATTR_CTIME) 750 inode->i_ctime = timespec64_trunc(attr->ia_ctime, 751 inode->i_sb->s_time_gran); 752 if (ia_valid & ATTR_MODE) { 753 umode_t mode = attr->ia_mode; 754 755 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) 756 mode &= ~S_ISGID; 757 set_acl_inode(inode, mode); 758 } 759 } 760 #else 761 #define __setattr_copy setattr_copy 762 #endif 763 764 int f2fs_setattr(struct dentry *dentry, struct iattr *attr) 765 { 766 struct inode *inode = d_inode(dentry); 767 int err; 768 bool size_changed = false; 769 770 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 771 return -EIO; 772 773 err = setattr_prepare(dentry, attr); 774 if (err) 775 return err; 776 777 err = fscrypt_prepare_setattr(dentry, attr); 778 if (err) 779 return err; 780 781 if (is_quota_modification(inode, attr)) { 782 err = dquot_initialize(inode); 783 if (err) 784 return err; 785 } 786 if ((attr->ia_valid & ATTR_UID && 787 !uid_eq(attr->ia_uid, inode->i_uid)) || 788 (attr->ia_valid & ATTR_GID && 789 !gid_eq(attr->ia_gid, inode->i_gid))) { 790 f2fs_lock_op(F2FS_I_SB(inode)); 791 err = dquot_transfer(inode, attr); 792 if (err) { 793 set_sbi_flag(F2FS_I_SB(inode), 794 SBI_QUOTA_NEED_REPAIR); 795 f2fs_unlock_op(F2FS_I_SB(inode)); 796 return err; 797 } 798 /* 799 * update uid/gid under lock_op(), so that dquot and inode can 800 * be updated atomically. 801 */ 802 if (attr->ia_valid & ATTR_UID) 803 inode->i_uid = attr->ia_uid; 804 if (attr->ia_valid & ATTR_GID) 805 inode->i_gid = attr->ia_gid; 806 f2fs_mark_inode_dirty_sync(inode, true); 807 f2fs_unlock_op(F2FS_I_SB(inode)); 808 } 809 810 if (attr->ia_valid & ATTR_SIZE) { 811 bool to_smaller = (attr->ia_size <= i_size_read(inode)); 812 813 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 814 down_write(&F2FS_I(inode)->i_mmap_sem); 815 816 truncate_setsize(inode, attr->ia_size); 817 818 if (to_smaller) 819 err = f2fs_truncate(inode); 820 /* 821 * do not trim all blocks after i_size if target size is 822 * larger than i_size. 823 */ 824 up_write(&F2FS_I(inode)->i_mmap_sem); 825 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 826 827 if (err) 828 return err; 829 830 if (!to_smaller) { 831 /* should convert inline inode here */ 832 if (!f2fs_may_inline_data(inode)) { 833 err = f2fs_convert_inline_inode(inode); 834 if (err) 835 return err; 836 } 837 inode->i_mtime = inode->i_ctime = current_time(inode); 838 } 839 840 down_write(&F2FS_I(inode)->i_sem); 841 F2FS_I(inode)->last_disk_size = i_size_read(inode); 842 up_write(&F2FS_I(inode)->i_sem); 843 844 size_changed = true; 845 } 846 847 __setattr_copy(inode, attr); 848 849 if (attr->ia_valid & ATTR_MODE) { 850 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode)); 851 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) { 852 inode->i_mode = F2FS_I(inode)->i_acl_mode; 853 clear_inode_flag(inode, FI_ACL_MODE); 854 } 855 } 856 857 /* file size may changed here */ 858 f2fs_mark_inode_dirty_sync(inode, size_changed); 859 860 /* inode change will produce dirty node pages flushed by checkpoint */ 861 f2fs_balance_fs(F2FS_I_SB(inode), true); 862 863 return err; 864 } 865 866 const struct inode_operations f2fs_file_inode_operations = { 867 .getattr = f2fs_getattr, 868 .setattr = f2fs_setattr, 869 .get_acl = f2fs_get_acl, 870 .set_acl = f2fs_set_acl, 871 #ifdef CONFIG_F2FS_FS_XATTR 872 .listxattr = f2fs_listxattr, 873 #endif 874 .fiemap = f2fs_fiemap, 875 }; 876 877 static int fill_zero(struct inode *inode, pgoff_t index, 878 loff_t start, loff_t len) 879 { 880 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 881 struct page *page; 882 883 if (!len) 884 return 0; 885 886 f2fs_balance_fs(sbi, true); 887 888 f2fs_lock_op(sbi); 889 page = f2fs_get_new_data_page(inode, NULL, index, false); 890 f2fs_unlock_op(sbi); 891 892 if (IS_ERR(page)) 893 return PTR_ERR(page); 894 895 f2fs_wait_on_page_writeback(page, DATA, true); 896 zero_user(page, start, len); 897 set_page_dirty(page); 898 f2fs_put_page(page, 1); 899 return 0; 900 } 901 902 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 903 { 904 int err; 905 906 while (pg_start < pg_end) { 907 struct dnode_of_data dn; 908 pgoff_t end_offset, count; 909 910 set_new_dnode(&dn, inode, NULL, NULL, 0); 911 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 912 if (err) { 913 if (err == -ENOENT) { 914 pg_start = f2fs_get_next_page_offset(&dn, 915 pg_start); 916 continue; 917 } 918 return err; 919 } 920 921 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 922 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 923 924 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 925 926 f2fs_truncate_data_blocks_range(&dn, count); 927 f2fs_put_dnode(&dn); 928 929 pg_start += count; 930 } 931 return 0; 932 } 933 934 static int punch_hole(struct inode *inode, loff_t offset, loff_t len) 935 { 936 pgoff_t pg_start, pg_end; 937 loff_t off_start, off_end; 938 int ret; 939 940 ret = f2fs_convert_inline_inode(inode); 941 if (ret) 942 return ret; 943 944 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 945 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 946 947 off_start = offset & (PAGE_SIZE - 1); 948 off_end = (offset + len) & (PAGE_SIZE - 1); 949 950 if (pg_start == pg_end) { 951 ret = fill_zero(inode, pg_start, off_start, 952 off_end - off_start); 953 if (ret) 954 return ret; 955 } else { 956 if (off_start) { 957 ret = fill_zero(inode, pg_start++, off_start, 958 PAGE_SIZE - off_start); 959 if (ret) 960 return ret; 961 } 962 if (off_end) { 963 ret = fill_zero(inode, pg_end, 0, off_end); 964 if (ret) 965 return ret; 966 } 967 968 if (pg_start < pg_end) { 969 struct address_space *mapping = inode->i_mapping; 970 loff_t blk_start, blk_end; 971 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 972 973 f2fs_balance_fs(sbi, true); 974 975 blk_start = (loff_t)pg_start << PAGE_SHIFT; 976 blk_end = (loff_t)pg_end << PAGE_SHIFT; 977 978 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 979 down_write(&F2FS_I(inode)->i_mmap_sem); 980 981 truncate_inode_pages_range(mapping, blk_start, 982 blk_end - 1); 983 984 f2fs_lock_op(sbi); 985 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 986 f2fs_unlock_op(sbi); 987 988 up_write(&F2FS_I(inode)->i_mmap_sem); 989 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 990 } 991 } 992 993 return ret; 994 } 995 996 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 997 int *do_replace, pgoff_t off, pgoff_t len) 998 { 999 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1000 struct dnode_of_data dn; 1001 int ret, done, i; 1002 1003 next_dnode: 1004 set_new_dnode(&dn, inode, NULL, NULL, 0); 1005 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1006 if (ret && ret != -ENOENT) { 1007 return ret; 1008 } else if (ret == -ENOENT) { 1009 if (dn.max_level == 0) 1010 return -ENOENT; 1011 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len); 1012 blkaddr += done; 1013 do_replace += done; 1014 goto next; 1015 } 1016 1017 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1018 dn.ofs_in_node, len); 1019 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1020 *blkaddr = datablock_addr(dn.inode, 1021 dn.node_page, dn.ofs_in_node); 1022 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1023 1024 if (test_opt(sbi, LFS)) { 1025 f2fs_put_dnode(&dn); 1026 return -ENOTSUPP; 1027 } 1028 1029 /* do not invalidate this block address */ 1030 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1031 *do_replace = 1; 1032 } 1033 } 1034 f2fs_put_dnode(&dn); 1035 next: 1036 len -= done; 1037 off += done; 1038 if (len) 1039 goto next_dnode; 1040 return 0; 1041 } 1042 1043 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1044 int *do_replace, pgoff_t off, int len) 1045 { 1046 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1047 struct dnode_of_data dn; 1048 int ret, i; 1049 1050 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1051 if (*do_replace == 0) 1052 continue; 1053 1054 set_new_dnode(&dn, inode, NULL, NULL, 0); 1055 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1056 if (ret) { 1057 dec_valid_block_count(sbi, inode, 1); 1058 f2fs_invalidate_blocks(sbi, *blkaddr); 1059 } else { 1060 f2fs_update_data_blkaddr(&dn, *blkaddr); 1061 } 1062 f2fs_put_dnode(&dn); 1063 } 1064 return 0; 1065 } 1066 1067 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1068 block_t *blkaddr, int *do_replace, 1069 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1070 { 1071 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1072 pgoff_t i = 0; 1073 int ret; 1074 1075 while (i < len) { 1076 if (blkaddr[i] == NULL_ADDR && !full) { 1077 i++; 1078 continue; 1079 } 1080 1081 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1082 struct dnode_of_data dn; 1083 struct node_info ni; 1084 size_t new_size; 1085 pgoff_t ilen; 1086 1087 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1088 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1089 if (ret) 1090 return ret; 1091 1092 ret = f2fs_get_node_info(sbi, dn.nid, &ni); 1093 if (ret) { 1094 f2fs_put_dnode(&dn); 1095 return ret; 1096 } 1097 1098 ilen = min((pgoff_t) 1099 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1100 dn.ofs_in_node, len - i); 1101 do { 1102 dn.data_blkaddr = datablock_addr(dn.inode, 1103 dn.node_page, dn.ofs_in_node); 1104 f2fs_truncate_data_blocks_range(&dn, 1); 1105 1106 if (do_replace[i]) { 1107 f2fs_i_blocks_write(src_inode, 1108 1, false, false); 1109 f2fs_i_blocks_write(dst_inode, 1110 1, true, false); 1111 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1112 blkaddr[i], ni.version, true, false); 1113 1114 do_replace[i] = 0; 1115 } 1116 dn.ofs_in_node++; 1117 i++; 1118 new_size = (dst + i) << PAGE_SHIFT; 1119 if (dst_inode->i_size < new_size) 1120 f2fs_i_size_write(dst_inode, new_size); 1121 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1122 1123 f2fs_put_dnode(&dn); 1124 } else { 1125 struct page *psrc, *pdst; 1126 1127 psrc = f2fs_get_lock_data_page(src_inode, 1128 src + i, true); 1129 if (IS_ERR(psrc)) 1130 return PTR_ERR(psrc); 1131 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1132 true); 1133 if (IS_ERR(pdst)) { 1134 f2fs_put_page(psrc, 1); 1135 return PTR_ERR(pdst); 1136 } 1137 f2fs_copy_page(psrc, pdst); 1138 set_page_dirty(pdst); 1139 f2fs_put_page(pdst, 1); 1140 f2fs_put_page(psrc, 1); 1141 1142 ret = f2fs_truncate_hole(src_inode, 1143 src + i, src + i + 1); 1144 if (ret) 1145 return ret; 1146 i++; 1147 } 1148 } 1149 return 0; 1150 } 1151 1152 static int __exchange_data_block(struct inode *src_inode, 1153 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1154 pgoff_t len, bool full) 1155 { 1156 block_t *src_blkaddr; 1157 int *do_replace; 1158 pgoff_t olen; 1159 int ret; 1160 1161 while (len) { 1162 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len); 1163 1164 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1165 array_size(olen, sizeof(block_t)), 1166 GFP_KERNEL); 1167 if (!src_blkaddr) 1168 return -ENOMEM; 1169 1170 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1171 array_size(olen, sizeof(int)), 1172 GFP_KERNEL); 1173 if (!do_replace) { 1174 kvfree(src_blkaddr); 1175 return -ENOMEM; 1176 } 1177 1178 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1179 do_replace, src, olen); 1180 if (ret) 1181 goto roll_back; 1182 1183 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1184 do_replace, src, dst, olen, full); 1185 if (ret) 1186 goto roll_back; 1187 1188 src += olen; 1189 dst += olen; 1190 len -= olen; 1191 1192 kvfree(src_blkaddr); 1193 kvfree(do_replace); 1194 } 1195 return 0; 1196 1197 roll_back: 1198 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1199 kvfree(src_blkaddr); 1200 kvfree(do_replace); 1201 return ret; 1202 } 1203 1204 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1205 { 1206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1207 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1208 pgoff_t start = offset >> PAGE_SHIFT; 1209 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1210 int ret; 1211 1212 f2fs_balance_fs(sbi, true); 1213 1214 /* avoid gc operation during block exchange */ 1215 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1216 down_write(&F2FS_I(inode)->i_mmap_sem); 1217 1218 f2fs_lock_op(sbi); 1219 f2fs_drop_extent_tree(inode); 1220 truncate_pagecache(inode, offset); 1221 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1222 f2fs_unlock_op(sbi); 1223 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_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1230 { 1231 loff_t new_size; 1232 int ret; 1233 1234 if (offset + len >= i_size_read(inode)) 1235 return -EINVAL; 1236 1237 /* collapse range should be aligned to block size of f2fs. */ 1238 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1239 return -EINVAL; 1240 1241 ret = f2fs_convert_inline_inode(inode); 1242 if (ret) 1243 return ret; 1244 1245 /* write out all dirty pages from offset */ 1246 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1247 if (ret) 1248 return ret; 1249 1250 ret = f2fs_do_collapse(inode, offset, len); 1251 if (ret) 1252 return ret; 1253 1254 /* write out all moved pages, if possible */ 1255 down_write(&F2FS_I(inode)->i_mmap_sem); 1256 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1257 truncate_pagecache(inode, offset); 1258 1259 new_size = i_size_read(inode) - len; 1260 truncate_pagecache(inode, new_size); 1261 1262 ret = f2fs_truncate_blocks(inode, new_size, true, false); 1263 up_write(&F2FS_I(inode)->i_mmap_sem); 1264 if (!ret) 1265 f2fs_i_size_write(inode, new_size); 1266 return ret; 1267 } 1268 1269 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1270 pgoff_t end) 1271 { 1272 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1273 pgoff_t index = start; 1274 unsigned int ofs_in_node = dn->ofs_in_node; 1275 blkcnt_t count = 0; 1276 int ret; 1277 1278 for (; index < end; index++, dn->ofs_in_node++) { 1279 if (datablock_addr(dn->inode, dn->node_page, 1280 dn->ofs_in_node) == NULL_ADDR) 1281 count++; 1282 } 1283 1284 dn->ofs_in_node = ofs_in_node; 1285 ret = f2fs_reserve_new_blocks(dn, count); 1286 if (ret) 1287 return ret; 1288 1289 dn->ofs_in_node = ofs_in_node; 1290 for (index = start; index < end; index++, dn->ofs_in_node++) { 1291 dn->data_blkaddr = datablock_addr(dn->inode, 1292 dn->node_page, dn->ofs_in_node); 1293 /* 1294 * f2fs_reserve_new_blocks will not guarantee entire block 1295 * allocation. 1296 */ 1297 if (dn->data_blkaddr == NULL_ADDR) { 1298 ret = -ENOSPC; 1299 break; 1300 } 1301 if (dn->data_blkaddr != NEW_ADDR) { 1302 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1303 dn->data_blkaddr = NEW_ADDR; 1304 f2fs_set_data_blkaddr(dn); 1305 } 1306 } 1307 1308 f2fs_update_extent_cache_range(dn, start, 0, index - start); 1309 1310 return ret; 1311 } 1312 1313 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1314 int mode) 1315 { 1316 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1317 struct address_space *mapping = inode->i_mapping; 1318 pgoff_t index, pg_start, pg_end; 1319 loff_t new_size = i_size_read(inode); 1320 loff_t off_start, off_end; 1321 int ret = 0; 1322 1323 ret = inode_newsize_ok(inode, (len + offset)); 1324 if (ret) 1325 return ret; 1326 1327 ret = f2fs_convert_inline_inode(inode); 1328 if (ret) 1329 return ret; 1330 1331 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1332 if (ret) 1333 return ret; 1334 1335 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1336 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1337 1338 off_start = offset & (PAGE_SIZE - 1); 1339 off_end = (offset + len) & (PAGE_SIZE - 1); 1340 1341 if (pg_start == pg_end) { 1342 ret = fill_zero(inode, pg_start, off_start, 1343 off_end - off_start); 1344 if (ret) 1345 return ret; 1346 1347 new_size = max_t(loff_t, new_size, offset + len); 1348 } else { 1349 if (off_start) { 1350 ret = fill_zero(inode, pg_start++, off_start, 1351 PAGE_SIZE - off_start); 1352 if (ret) 1353 return ret; 1354 1355 new_size = max_t(loff_t, new_size, 1356 (loff_t)pg_start << PAGE_SHIFT); 1357 } 1358 1359 for (index = pg_start; index < pg_end;) { 1360 struct dnode_of_data dn; 1361 unsigned int end_offset; 1362 pgoff_t end; 1363 1364 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1365 down_write(&F2FS_I(inode)->i_mmap_sem); 1366 1367 truncate_pagecache_range(inode, 1368 (loff_t)index << PAGE_SHIFT, 1369 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1370 1371 f2fs_lock_op(sbi); 1372 1373 set_new_dnode(&dn, inode, NULL, NULL, 0); 1374 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1375 if (ret) { 1376 f2fs_unlock_op(sbi); 1377 up_write(&F2FS_I(inode)->i_mmap_sem); 1378 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1379 goto out; 1380 } 1381 1382 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1383 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1384 1385 ret = f2fs_do_zero_range(&dn, index, end); 1386 f2fs_put_dnode(&dn); 1387 1388 f2fs_unlock_op(sbi); 1389 up_write(&F2FS_I(inode)->i_mmap_sem); 1390 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1391 1392 f2fs_balance_fs(sbi, dn.node_changed); 1393 1394 if (ret) 1395 goto out; 1396 1397 index = end; 1398 new_size = max_t(loff_t, new_size, 1399 (loff_t)index << PAGE_SHIFT); 1400 } 1401 1402 if (off_end) { 1403 ret = fill_zero(inode, pg_end, 0, off_end); 1404 if (ret) 1405 goto out; 1406 1407 new_size = max_t(loff_t, new_size, offset + len); 1408 } 1409 } 1410 1411 out: 1412 if (new_size > i_size_read(inode)) { 1413 if (mode & FALLOC_FL_KEEP_SIZE) 1414 file_set_keep_isize(inode); 1415 else 1416 f2fs_i_size_write(inode, new_size); 1417 } 1418 return ret; 1419 } 1420 1421 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1422 { 1423 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1424 pgoff_t nr, pg_start, pg_end, delta, idx; 1425 loff_t new_size; 1426 int ret = 0; 1427 1428 new_size = i_size_read(inode) + len; 1429 ret = inode_newsize_ok(inode, new_size); 1430 if (ret) 1431 return ret; 1432 1433 if (offset >= i_size_read(inode)) 1434 return -EINVAL; 1435 1436 /* insert range should be aligned to block size of f2fs. */ 1437 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1438 return -EINVAL; 1439 1440 ret = f2fs_convert_inline_inode(inode); 1441 if (ret) 1442 return ret; 1443 1444 f2fs_balance_fs(sbi, true); 1445 1446 down_write(&F2FS_I(inode)->i_mmap_sem); 1447 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true, false); 1448 up_write(&F2FS_I(inode)->i_mmap_sem); 1449 if (ret) 1450 return ret; 1451 1452 /* write out all dirty pages from offset */ 1453 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1454 if (ret) 1455 return ret; 1456 1457 pg_start = offset >> PAGE_SHIFT; 1458 pg_end = (offset + len) >> PAGE_SHIFT; 1459 delta = pg_end - pg_start; 1460 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE; 1461 1462 /* avoid gc operation during block exchange */ 1463 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1464 down_write(&F2FS_I(inode)->i_mmap_sem); 1465 truncate_pagecache(inode, offset); 1466 1467 while (!ret && idx > pg_start) { 1468 nr = idx - pg_start; 1469 if (nr > delta) 1470 nr = delta; 1471 idx -= nr; 1472 1473 f2fs_lock_op(sbi); 1474 f2fs_drop_extent_tree(inode); 1475 1476 ret = __exchange_data_block(inode, inode, idx, 1477 idx + delta, nr, false); 1478 f2fs_unlock_op(sbi); 1479 } 1480 up_write(&F2FS_I(inode)->i_mmap_sem); 1481 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1482 1483 /* write out all moved pages, if possible */ 1484 down_write(&F2FS_I(inode)->i_mmap_sem); 1485 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1486 truncate_pagecache(inode, offset); 1487 up_write(&F2FS_I(inode)->i_mmap_sem); 1488 1489 if (!ret) 1490 f2fs_i_size_write(inode, new_size); 1491 return ret; 1492 } 1493 1494 static int expand_inode_data(struct inode *inode, loff_t offset, 1495 loff_t len, int mode) 1496 { 1497 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1498 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1499 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE }; 1500 pgoff_t pg_end; 1501 loff_t new_size = i_size_read(inode); 1502 loff_t off_end; 1503 int err; 1504 1505 err = inode_newsize_ok(inode, (len + offset)); 1506 if (err) 1507 return err; 1508 1509 err = f2fs_convert_inline_inode(inode); 1510 if (err) 1511 return err; 1512 1513 f2fs_balance_fs(sbi, true); 1514 1515 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1516 off_end = (offset + len) & (PAGE_SIZE - 1); 1517 1518 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT; 1519 map.m_len = pg_end - map.m_lblk; 1520 if (off_end) 1521 map.m_len++; 1522 1523 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); 1524 if (err) { 1525 pgoff_t last_off; 1526 1527 if (!map.m_len) 1528 return err; 1529 1530 last_off = map.m_lblk + map.m_len - 1; 1531 1532 /* update new size to the failed position */ 1533 new_size = (last_off == pg_end) ? offset + len : 1534 (loff_t)(last_off + 1) << PAGE_SHIFT; 1535 } else { 1536 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1537 } 1538 1539 if (new_size > i_size_read(inode)) { 1540 if (mode & FALLOC_FL_KEEP_SIZE) 1541 file_set_keep_isize(inode); 1542 else 1543 f2fs_i_size_write(inode, new_size); 1544 } 1545 1546 return err; 1547 } 1548 1549 static long f2fs_fallocate(struct file *file, int mode, 1550 loff_t offset, loff_t len) 1551 { 1552 struct inode *inode = file_inode(file); 1553 long ret = 0; 1554 1555 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1556 return -EIO; 1557 1558 /* f2fs only support ->fallocate for regular file */ 1559 if (!S_ISREG(inode->i_mode)) 1560 return -EINVAL; 1561 1562 if (f2fs_encrypted_inode(inode) && 1563 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1564 return -EOPNOTSUPP; 1565 1566 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1567 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1568 FALLOC_FL_INSERT_RANGE)) 1569 return -EOPNOTSUPP; 1570 1571 inode_lock(inode); 1572 1573 if (mode & FALLOC_FL_PUNCH_HOLE) { 1574 if (offset >= inode->i_size) 1575 goto out; 1576 1577 ret = punch_hole(inode, offset, len); 1578 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1579 ret = f2fs_collapse_range(inode, offset, len); 1580 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1581 ret = f2fs_zero_range(inode, offset, len, mode); 1582 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1583 ret = f2fs_insert_range(inode, offset, len); 1584 } else { 1585 ret = expand_inode_data(inode, offset, len, mode); 1586 } 1587 1588 if (!ret) { 1589 inode->i_mtime = inode->i_ctime = current_time(inode); 1590 f2fs_mark_inode_dirty_sync(inode, false); 1591 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1592 } 1593 1594 out: 1595 inode_unlock(inode); 1596 1597 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1598 return ret; 1599 } 1600 1601 static int f2fs_release_file(struct inode *inode, struct file *filp) 1602 { 1603 /* 1604 * f2fs_relase_file is called at every close calls. So we should 1605 * not drop any inmemory pages by close called by other process. 1606 */ 1607 if (!(filp->f_mode & FMODE_WRITE) || 1608 atomic_read(&inode->i_writecount) != 1) 1609 return 0; 1610 1611 /* some remained atomic pages should discarded */ 1612 if (f2fs_is_atomic_file(inode)) 1613 f2fs_drop_inmem_pages(inode); 1614 if (f2fs_is_volatile_file(inode)) { 1615 set_inode_flag(inode, FI_DROP_CACHE); 1616 filemap_fdatawrite(inode->i_mapping); 1617 clear_inode_flag(inode, FI_DROP_CACHE); 1618 clear_inode_flag(inode, FI_VOLATILE_FILE); 1619 stat_dec_volatile_write(inode); 1620 } 1621 return 0; 1622 } 1623 1624 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1625 { 1626 struct inode *inode = file_inode(file); 1627 1628 /* 1629 * If the process doing a transaction is crashed, we should do 1630 * roll-back. Otherwise, other reader/write can see corrupted database 1631 * until all the writers close its file. Since this should be done 1632 * before dropping file lock, it needs to do in ->flush. 1633 */ 1634 if (f2fs_is_atomic_file(inode) && 1635 F2FS_I(inode)->inmem_task == current) 1636 f2fs_drop_inmem_pages(inode); 1637 return 0; 1638 } 1639 1640 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg) 1641 { 1642 struct inode *inode = file_inode(filp); 1643 struct f2fs_inode_info *fi = F2FS_I(inode); 1644 unsigned int flags = fi->i_flags; 1645 1646 if (f2fs_encrypted_inode(inode)) 1647 flags |= F2FS_ENCRYPT_FL; 1648 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 1649 flags |= F2FS_INLINE_DATA_FL; 1650 1651 flags &= F2FS_FL_USER_VISIBLE; 1652 1653 return put_user(flags, (int __user *)arg); 1654 } 1655 1656 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags) 1657 { 1658 struct f2fs_inode_info *fi = F2FS_I(inode); 1659 unsigned int oldflags; 1660 1661 /* Is it quota file? Do not allow user to mess with it */ 1662 if (IS_NOQUOTA(inode)) 1663 return -EPERM; 1664 1665 flags = f2fs_mask_flags(inode->i_mode, flags); 1666 1667 oldflags = fi->i_flags; 1668 1669 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL)) 1670 if (!capable(CAP_LINUX_IMMUTABLE)) 1671 return -EPERM; 1672 1673 flags = flags & F2FS_FL_USER_MODIFIABLE; 1674 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE; 1675 fi->i_flags = flags; 1676 1677 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1678 set_inode_flag(inode, FI_PROJ_INHERIT); 1679 else 1680 clear_inode_flag(inode, FI_PROJ_INHERIT); 1681 1682 inode->i_ctime = current_time(inode); 1683 f2fs_set_inode_flags(inode); 1684 f2fs_mark_inode_dirty_sync(inode, false); 1685 return 0; 1686 } 1687 1688 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg) 1689 { 1690 struct inode *inode = file_inode(filp); 1691 unsigned int flags; 1692 int ret; 1693 1694 if (!inode_owner_or_capable(inode)) 1695 return -EACCES; 1696 1697 if (get_user(flags, (int __user *)arg)) 1698 return -EFAULT; 1699 1700 ret = mnt_want_write_file(filp); 1701 if (ret) 1702 return ret; 1703 1704 inode_lock(inode); 1705 1706 ret = __f2fs_ioc_setflags(inode, flags); 1707 1708 inode_unlock(inode); 1709 mnt_drop_write_file(filp); 1710 return ret; 1711 } 1712 1713 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 1714 { 1715 struct inode *inode = file_inode(filp); 1716 1717 return put_user(inode->i_generation, (int __user *)arg); 1718 } 1719 1720 static int f2fs_ioc_start_atomic_write(struct file *filp) 1721 { 1722 struct inode *inode = file_inode(filp); 1723 int ret; 1724 1725 if (!inode_owner_or_capable(inode)) 1726 return -EACCES; 1727 1728 if (!S_ISREG(inode->i_mode)) 1729 return -EINVAL; 1730 1731 ret = mnt_want_write_file(filp); 1732 if (ret) 1733 return ret; 1734 1735 inode_lock(inode); 1736 1737 if (f2fs_is_atomic_file(inode)) { 1738 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) 1739 ret = -EINVAL; 1740 goto out; 1741 } 1742 1743 ret = f2fs_convert_inline_inode(inode); 1744 if (ret) 1745 goto out; 1746 1747 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1748 1749 if (!get_dirty_pages(inode)) 1750 goto skip_flush; 1751 1752 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING, 1753 "Unexpected flush for atomic writes: ino=%lu, npages=%u", 1754 inode->i_ino, get_dirty_pages(inode)); 1755 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 1756 if (ret) { 1757 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1758 goto out; 1759 } 1760 skip_flush: 1761 set_inode_flag(inode, FI_ATOMIC_FILE); 1762 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1763 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1764 1765 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1766 F2FS_I(inode)->inmem_task = current; 1767 stat_inc_atomic_write(inode); 1768 stat_update_max_atomic_write(inode); 1769 out: 1770 inode_unlock(inode); 1771 mnt_drop_write_file(filp); 1772 return ret; 1773 } 1774 1775 static int f2fs_ioc_commit_atomic_write(struct file *filp) 1776 { 1777 struct inode *inode = file_inode(filp); 1778 int ret; 1779 1780 if (!inode_owner_or_capable(inode)) 1781 return -EACCES; 1782 1783 ret = mnt_want_write_file(filp); 1784 if (ret) 1785 return ret; 1786 1787 f2fs_balance_fs(F2FS_I_SB(inode), true); 1788 1789 inode_lock(inode); 1790 1791 if (f2fs_is_volatile_file(inode)) { 1792 ret = -EINVAL; 1793 goto err_out; 1794 } 1795 1796 if (f2fs_is_atomic_file(inode)) { 1797 ret = f2fs_commit_inmem_pages(inode); 1798 if (ret) 1799 goto err_out; 1800 1801 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1802 if (!ret) { 1803 clear_inode_flag(inode, FI_ATOMIC_FILE); 1804 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0; 1805 stat_dec_atomic_write(inode); 1806 } 1807 } else { 1808 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 1809 } 1810 err_out: 1811 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) { 1812 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1813 ret = -EINVAL; 1814 } 1815 inode_unlock(inode); 1816 mnt_drop_write_file(filp); 1817 return ret; 1818 } 1819 1820 static int f2fs_ioc_start_volatile_write(struct file *filp) 1821 { 1822 struct inode *inode = file_inode(filp); 1823 int ret; 1824 1825 if (!inode_owner_or_capable(inode)) 1826 return -EACCES; 1827 1828 if (!S_ISREG(inode->i_mode)) 1829 return -EINVAL; 1830 1831 ret = mnt_want_write_file(filp); 1832 if (ret) 1833 return ret; 1834 1835 inode_lock(inode); 1836 1837 if (f2fs_is_volatile_file(inode)) 1838 goto out; 1839 1840 ret = f2fs_convert_inline_inode(inode); 1841 if (ret) 1842 goto out; 1843 1844 stat_inc_volatile_write(inode); 1845 stat_update_max_volatile_write(inode); 1846 1847 set_inode_flag(inode, FI_VOLATILE_FILE); 1848 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1849 out: 1850 inode_unlock(inode); 1851 mnt_drop_write_file(filp); 1852 return ret; 1853 } 1854 1855 static int f2fs_ioc_release_volatile_write(struct file *filp) 1856 { 1857 struct inode *inode = file_inode(filp); 1858 int ret; 1859 1860 if (!inode_owner_or_capable(inode)) 1861 return -EACCES; 1862 1863 ret = mnt_want_write_file(filp); 1864 if (ret) 1865 return ret; 1866 1867 inode_lock(inode); 1868 1869 if (!f2fs_is_volatile_file(inode)) 1870 goto out; 1871 1872 if (!f2fs_is_first_block_written(inode)) { 1873 ret = truncate_partial_data_page(inode, 0, true); 1874 goto out; 1875 } 1876 1877 ret = punch_hole(inode, 0, F2FS_BLKSIZE); 1878 out: 1879 inode_unlock(inode); 1880 mnt_drop_write_file(filp); 1881 return ret; 1882 } 1883 1884 static int f2fs_ioc_abort_volatile_write(struct file *filp) 1885 { 1886 struct inode *inode = file_inode(filp); 1887 int ret; 1888 1889 if (!inode_owner_or_capable(inode)) 1890 return -EACCES; 1891 1892 ret = mnt_want_write_file(filp); 1893 if (ret) 1894 return ret; 1895 1896 inode_lock(inode); 1897 1898 if (f2fs_is_atomic_file(inode)) 1899 f2fs_drop_inmem_pages(inode); 1900 if (f2fs_is_volatile_file(inode)) { 1901 clear_inode_flag(inode, FI_VOLATILE_FILE); 1902 stat_dec_volatile_write(inode); 1903 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 1904 } 1905 1906 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST); 1907 1908 inode_unlock(inode); 1909 1910 mnt_drop_write_file(filp); 1911 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1912 return ret; 1913 } 1914 1915 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 1916 { 1917 struct inode *inode = file_inode(filp); 1918 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1919 struct super_block *sb = sbi->sb; 1920 __u32 in; 1921 int ret = 0; 1922 1923 if (!capable(CAP_SYS_ADMIN)) 1924 return -EPERM; 1925 1926 if (get_user(in, (__u32 __user *)arg)) 1927 return -EFAULT; 1928 1929 if (in != F2FS_GOING_DOWN_FULLSYNC) { 1930 ret = mnt_want_write_file(filp); 1931 if (ret) 1932 return ret; 1933 } 1934 1935 switch (in) { 1936 case F2FS_GOING_DOWN_FULLSYNC: 1937 sb = freeze_bdev(sb->s_bdev); 1938 if (IS_ERR(sb)) { 1939 ret = PTR_ERR(sb); 1940 goto out; 1941 } 1942 if (sb) { 1943 f2fs_stop_checkpoint(sbi, false); 1944 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1945 thaw_bdev(sb->s_bdev, sb); 1946 } 1947 break; 1948 case F2FS_GOING_DOWN_METASYNC: 1949 /* do checkpoint only */ 1950 ret = f2fs_sync_fs(sb, 1); 1951 if (ret) 1952 goto out; 1953 f2fs_stop_checkpoint(sbi, false); 1954 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1955 break; 1956 case F2FS_GOING_DOWN_NOSYNC: 1957 f2fs_stop_checkpoint(sbi, false); 1958 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1959 break; 1960 case F2FS_GOING_DOWN_METAFLUSH: 1961 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 1962 f2fs_stop_checkpoint(sbi, false); 1963 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 1964 break; 1965 default: 1966 ret = -EINVAL; 1967 goto out; 1968 } 1969 1970 f2fs_stop_gc_thread(sbi); 1971 f2fs_stop_discard_thread(sbi); 1972 1973 f2fs_drop_discard_cmd(sbi); 1974 clear_opt(sbi, DISCARD); 1975 1976 f2fs_update_time(sbi, REQ_TIME); 1977 out: 1978 if (in != F2FS_GOING_DOWN_FULLSYNC) 1979 mnt_drop_write_file(filp); 1980 return ret; 1981 } 1982 1983 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 1984 { 1985 struct inode *inode = file_inode(filp); 1986 struct super_block *sb = inode->i_sb; 1987 struct request_queue *q = bdev_get_queue(sb->s_bdev); 1988 struct fstrim_range range; 1989 int ret; 1990 1991 if (!capable(CAP_SYS_ADMIN)) 1992 return -EPERM; 1993 1994 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 1995 return -EOPNOTSUPP; 1996 1997 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 1998 sizeof(range))) 1999 return -EFAULT; 2000 2001 ret = mnt_want_write_file(filp); 2002 if (ret) 2003 return ret; 2004 2005 range.minlen = max((unsigned int)range.minlen, 2006 q->limits.discard_granularity); 2007 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2008 mnt_drop_write_file(filp); 2009 if (ret < 0) 2010 return ret; 2011 2012 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2013 sizeof(range))) 2014 return -EFAULT; 2015 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2016 return 0; 2017 } 2018 2019 static bool uuid_is_nonzero(__u8 u[16]) 2020 { 2021 int i; 2022 2023 for (i = 0; i < 16; i++) 2024 if (u[i]) 2025 return true; 2026 return false; 2027 } 2028 2029 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2030 { 2031 struct inode *inode = file_inode(filp); 2032 2033 if (!f2fs_sb_has_encrypt(inode->i_sb)) 2034 return -EOPNOTSUPP; 2035 2036 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2037 2038 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2039 } 2040 2041 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2042 { 2043 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb)) 2044 return -EOPNOTSUPP; 2045 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2046 } 2047 2048 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2049 { 2050 struct inode *inode = file_inode(filp); 2051 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2052 int err; 2053 2054 if (!f2fs_sb_has_encrypt(inode->i_sb)) 2055 return -EOPNOTSUPP; 2056 2057 err = mnt_want_write_file(filp); 2058 if (err) 2059 return err; 2060 2061 down_write(&sbi->sb_lock); 2062 2063 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2064 goto got_it; 2065 2066 /* update superblock with uuid */ 2067 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2068 2069 err = f2fs_commit_super(sbi, false); 2070 if (err) { 2071 /* undo new data */ 2072 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2073 goto out_err; 2074 } 2075 got_it: 2076 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt, 2077 16)) 2078 err = -EFAULT; 2079 out_err: 2080 up_write(&sbi->sb_lock); 2081 mnt_drop_write_file(filp); 2082 return err; 2083 } 2084 2085 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2086 { 2087 struct inode *inode = file_inode(filp); 2088 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2089 __u32 sync; 2090 int ret; 2091 2092 if (!capable(CAP_SYS_ADMIN)) 2093 return -EPERM; 2094 2095 if (get_user(sync, (__u32 __user *)arg)) 2096 return -EFAULT; 2097 2098 if (f2fs_readonly(sbi->sb)) 2099 return -EROFS; 2100 2101 ret = mnt_want_write_file(filp); 2102 if (ret) 2103 return ret; 2104 2105 if (!sync) { 2106 if (!mutex_trylock(&sbi->gc_mutex)) { 2107 ret = -EBUSY; 2108 goto out; 2109 } 2110 } else { 2111 mutex_lock(&sbi->gc_mutex); 2112 } 2113 2114 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO); 2115 out: 2116 mnt_drop_write_file(filp); 2117 return ret; 2118 } 2119 2120 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2121 { 2122 struct inode *inode = file_inode(filp); 2123 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2124 struct f2fs_gc_range range; 2125 u64 end; 2126 int ret; 2127 2128 if (!capable(CAP_SYS_ADMIN)) 2129 return -EPERM; 2130 2131 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2132 sizeof(range))) 2133 return -EFAULT; 2134 2135 if (f2fs_readonly(sbi->sb)) 2136 return -EROFS; 2137 2138 end = range.start + range.len; 2139 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) { 2140 return -EINVAL; 2141 } 2142 2143 ret = mnt_want_write_file(filp); 2144 if (ret) 2145 return ret; 2146 2147 do_more: 2148 if (!range.sync) { 2149 if (!mutex_trylock(&sbi->gc_mutex)) { 2150 ret = -EBUSY; 2151 goto out; 2152 } 2153 } else { 2154 mutex_lock(&sbi->gc_mutex); 2155 } 2156 2157 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start)); 2158 range.start += sbi->blocks_per_seg; 2159 if (range.start <= end) 2160 goto do_more; 2161 out: 2162 mnt_drop_write_file(filp); 2163 return ret; 2164 } 2165 2166 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg) 2167 { 2168 struct inode *inode = file_inode(filp); 2169 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2170 int ret; 2171 2172 if (!capable(CAP_SYS_ADMIN)) 2173 return -EPERM; 2174 2175 if (f2fs_readonly(sbi->sb)) 2176 return -EROFS; 2177 2178 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2179 f2fs_msg(sbi->sb, KERN_INFO, 2180 "Skipping Checkpoint. Checkpoints currently disabled."); 2181 return -EINVAL; 2182 } 2183 2184 ret = mnt_want_write_file(filp); 2185 if (ret) 2186 return ret; 2187 2188 ret = f2fs_sync_fs(sbi->sb, 1); 2189 2190 mnt_drop_write_file(filp); 2191 return ret; 2192 } 2193 2194 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2195 struct file *filp, 2196 struct f2fs_defragment *range) 2197 { 2198 struct inode *inode = file_inode(filp); 2199 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2200 .m_seg_type = NO_CHECK_TYPE }; 2201 struct extent_info ei = {0, 0, 0}; 2202 pgoff_t pg_start, pg_end, next_pgofs; 2203 unsigned int blk_per_seg = sbi->blocks_per_seg; 2204 unsigned int total = 0, sec_num; 2205 block_t blk_end = 0; 2206 bool fragmented = false; 2207 int err; 2208 2209 /* if in-place-update policy is enabled, don't waste time here */ 2210 if (f2fs_should_update_inplace(inode, NULL)) 2211 return -EINVAL; 2212 2213 pg_start = range->start >> PAGE_SHIFT; 2214 pg_end = (range->start + range->len) >> PAGE_SHIFT; 2215 2216 f2fs_balance_fs(sbi, true); 2217 2218 inode_lock(inode); 2219 2220 /* writeback all dirty pages in the range */ 2221 err = filemap_write_and_wait_range(inode->i_mapping, range->start, 2222 range->start + range->len - 1); 2223 if (err) 2224 goto out; 2225 2226 /* 2227 * lookup mapping info in extent cache, skip defragmenting if physical 2228 * block addresses are continuous. 2229 */ 2230 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) { 2231 if (ei.fofs + ei.len >= pg_end) 2232 goto out; 2233 } 2234 2235 map.m_lblk = pg_start; 2236 map.m_next_pgofs = &next_pgofs; 2237 2238 /* 2239 * lookup mapping info in dnode page cache, skip defragmenting if all 2240 * physical block addresses are continuous even if there are hole(s) 2241 * in logical blocks. 2242 */ 2243 while (map.m_lblk < pg_end) { 2244 map.m_len = pg_end - map.m_lblk; 2245 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2246 if (err) 2247 goto out; 2248 2249 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2250 map.m_lblk = next_pgofs; 2251 continue; 2252 } 2253 2254 if (blk_end && blk_end != map.m_pblk) 2255 fragmented = true; 2256 2257 /* record total count of block that we're going to move */ 2258 total += map.m_len; 2259 2260 blk_end = map.m_pblk + map.m_len; 2261 2262 map.m_lblk += map.m_len; 2263 } 2264 2265 if (!fragmented) 2266 goto out; 2267 2268 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi); 2269 2270 /* 2271 * make sure there are enough free section for LFS allocation, this can 2272 * avoid defragment running in SSR mode when free section are allocated 2273 * intensively 2274 */ 2275 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2276 err = -EAGAIN; 2277 goto out; 2278 } 2279 2280 map.m_lblk = pg_start; 2281 map.m_len = pg_end - pg_start; 2282 total = 0; 2283 2284 while (map.m_lblk < pg_end) { 2285 pgoff_t idx; 2286 int cnt = 0; 2287 2288 do_map: 2289 map.m_len = pg_end - map.m_lblk; 2290 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT); 2291 if (err) 2292 goto clear_out; 2293 2294 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2295 map.m_lblk = next_pgofs; 2296 continue; 2297 } 2298 2299 set_inode_flag(inode, FI_DO_DEFRAG); 2300 2301 idx = map.m_lblk; 2302 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) { 2303 struct page *page; 2304 2305 page = f2fs_get_lock_data_page(inode, idx, true); 2306 if (IS_ERR(page)) { 2307 err = PTR_ERR(page); 2308 goto clear_out; 2309 } 2310 2311 set_page_dirty(page); 2312 f2fs_put_page(page, 1); 2313 2314 idx++; 2315 cnt++; 2316 total++; 2317 } 2318 2319 map.m_lblk = idx; 2320 2321 if (idx < pg_end && cnt < blk_per_seg) 2322 goto do_map; 2323 2324 clear_inode_flag(inode, FI_DO_DEFRAG); 2325 2326 err = filemap_fdatawrite(inode->i_mapping); 2327 if (err) 2328 goto out; 2329 } 2330 clear_out: 2331 clear_inode_flag(inode, FI_DO_DEFRAG); 2332 out: 2333 inode_unlock(inode); 2334 if (!err) 2335 range->len = (u64)total << PAGE_SHIFT; 2336 return err; 2337 } 2338 2339 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2340 { 2341 struct inode *inode = file_inode(filp); 2342 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2343 struct f2fs_defragment range; 2344 int err; 2345 2346 if (!capable(CAP_SYS_ADMIN)) 2347 return -EPERM; 2348 2349 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2350 return -EINVAL; 2351 2352 if (f2fs_readonly(sbi->sb)) 2353 return -EROFS; 2354 2355 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2356 sizeof(range))) 2357 return -EFAULT; 2358 2359 /* verify alignment of offset & size */ 2360 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2361 return -EINVAL; 2362 2363 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2364 sbi->max_file_blocks)) 2365 return -EINVAL; 2366 2367 err = mnt_want_write_file(filp); 2368 if (err) 2369 return err; 2370 2371 err = f2fs_defragment_range(sbi, filp, &range); 2372 mnt_drop_write_file(filp); 2373 2374 f2fs_update_time(sbi, REQ_TIME); 2375 if (err < 0) 2376 return err; 2377 2378 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2379 sizeof(range))) 2380 return -EFAULT; 2381 2382 return 0; 2383 } 2384 2385 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2386 struct file *file_out, loff_t pos_out, size_t len) 2387 { 2388 struct inode *src = file_inode(file_in); 2389 struct inode *dst = file_inode(file_out); 2390 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2391 size_t olen = len, dst_max_i_size = 0; 2392 size_t dst_osize; 2393 int ret; 2394 2395 if (file_in->f_path.mnt != file_out->f_path.mnt || 2396 src->i_sb != dst->i_sb) 2397 return -EXDEV; 2398 2399 if (unlikely(f2fs_readonly(src->i_sb))) 2400 return -EROFS; 2401 2402 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2403 return -EINVAL; 2404 2405 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst)) 2406 return -EOPNOTSUPP; 2407 2408 if (src == dst) { 2409 if (pos_in == pos_out) 2410 return 0; 2411 if (pos_out > pos_in && pos_out < pos_in + len) 2412 return -EINVAL; 2413 } 2414 2415 inode_lock(src); 2416 if (src != dst) { 2417 ret = -EBUSY; 2418 if (!inode_trylock(dst)) 2419 goto out; 2420 } 2421 2422 ret = -EINVAL; 2423 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2424 goto out_unlock; 2425 if (len == 0) 2426 olen = len = src->i_size - pos_in; 2427 if (pos_in + len == src->i_size) 2428 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2429 if (len == 0) { 2430 ret = 0; 2431 goto out_unlock; 2432 } 2433 2434 dst_osize = dst->i_size; 2435 if (pos_out + olen > dst->i_size) 2436 dst_max_i_size = pos_out + olen; 2437 2438 /* verify the end result is block aligned */ 2439 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2440 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2441 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2442 goto out_unlock; 2443 2444 ret = f2fs_convert_inline_inode(src); 2445 if (ret) 2446 goto out_unlock; 2447 2448 ret = f2fs_convert_inline_inode(dst); 2449 if (ret) 2450 goto out_unlock; 2451 2452 /* write out all dirty pages from offset */ 2453 ret = filemap_write_and_wait_range(src->i_mapping, 2454 pos_in, pos_in + len); 2455 if (ret) 2456 goto out_unlock; 2457 2458 ret = filemap_write_and_wait_range(dst->i_mapping, 2459 pos_out, pos_out + len); 2460 if (ret) 2461 goto out_unlock; 2462 2463 f2fs_balance_fs(sbi, true); 2464 2465 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2466 if (src != dst) { 2467 ret = -EBUSY; 2468 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2469 goto out_src; 2470 } 2471 2472 f2fs_lock_op(sbi); 2473 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2474 pos_out >> F2FS_BLKSIZE_BITS, 2475 len >> F2FS_BLKSIZE_BITS, false); 2476 2477 if (!ret) { 2478 if (dst_max_i_size) 2479 f2fs_i_size_write(dst, dst_max_i_size); 2480 else if (dst_osize != dst->i_size) 2481 f2fs_i_size_write(dst, dst_osize); 2482 } 2483 f2fs_unlock_op(sbi); 2484 2485 if (src != dst) 2486 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2487 out_src: 2488 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2489 out_unlock: 2490 if (src != dst) 2491 inode_unlock(dst); 2492 out: 2493 inode_unlock(src); 2494 return ret; 2495 } 2496 2497 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2498 { 2499 struct f2fs_move_range range; 2500 struct fd dst; 2501 int err; 2502 2503 if (!(filp->f_mode & FMODE_READ) || 2504 !(filp->f_mode & FMODE_WRITE)) 2505 return -EBADF; 2506 2507 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2508 sizeof(range))) 2509 return -EFAULT; 2510 2511 dst = fdget(range.dst_fd); 2512 if (!dst.file) 2513 return -EBADF; 2514 2515 if (!(dst.file->f_mode & FMODE_WRITE)) { 2516 err = -EBADF; 2517 goto err_out; 2518 } 2519 2520 err = mnt_want_write_file(filp); 2521 if (err) 2522 goto err_out; 2523 2524 err = f2fs_move_file_range(filp, range.pos_in, dst.file, 2525 range.pos_out, range.len); 2526 2527 mnt_drop_write_file(filp); 2528 if (err) 2529 goto err_out; 2530 2531 if (copy_to_user((struct f2fs_move_range __user *)arg, 2532 &range, sizeof(range))) 2533 err = -EFAULT; 2534 err_out: 2535 fdput(dst); 2536 return err; 2537 } 2538 2539 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 2540 { 2541 struct inode *inode = file_inode(filp); 2542 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2543 struct sit_info *sm = SIT_I(sbi); 2544 unsigned int start_segno = 0, end_segno = 0; 2545 unsigned int dev_start_segno = 0, dev_end_segno = 0; 2546 struct f2fs_flush_device range; 2547 int ret; 2548 2549 if (!capable(CAP_SYS_ADMIN)) 2550 return -EPERM; 2551 2552 if (f2fs_readonly(sbi->sb)) 2553 return -EROFS; 2554 2555 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2556 return -EINVAL; 2557 2558 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 2559 sizeof(range))) 2560 return -EFAULT; 2561 2562 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num || 2563 sbi->segs_per_sec != 1) { 2564 f2fs_msg(sbi->sb, KERN_WARNING, 2565 "Can't flush %u in %d for segs_per_sec %u != 1\n", 2566 range.dev_num, sbi->s_ndevs, 2567 sbi->segs_per_sec); 2568 return -EINVAL; 2569 } 2570 2571 ret = mnt_want_write_file(filp); 2572 if (ret) 2573 return ret; 2574 2575 if (range.dev_num != 0) 2576 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 2577 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 2578 2579 start_segno = sm->last_victim[FLUSH_DEVICE]; 2580 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 2581 start_segno = dev_start_segno; 2582 end_segno = min(start_segno + range.segments, dev_end_segno); 2583 2584 while (start_segno < end_segno) { 2585 if (!mutex_trylock(&sbi->gc_mutex)) { 2586 ret = -EBUSY; 2587 goto out; 2588 } 2589 sm->last_victim[GC_CB] = end_segno + 1; 2590 sm->last_victim[GC_GREEDY] = end_segno + 1; 2591 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 2592 ret = f2fs_gc(sbi, true, true, start_segno); 2593 if (ret == -EAGAIN) 2594 ret = 0; 2595 else if (ret < 0) 2596 break; 2597 start_segno++; 2598 } 2599 out: 2600 mnt_drop_write_file(filp); 2601 return ret; 2602 } 2603 2604 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 2605 { 2606 struct inode *inode = file_inode(filp); 2607 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 2608 2609 /* Must validate to set it with SQLite behavior in Android. */ 2610 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 2611 2612 return put_user(sb_feature, (u32 __user *)arg); 2613 } 2614 2615 #ifdef CONFIG_QUOTA 2616 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 2617 { 2618 struct dquot *transfer_to[MAXQUOTAS] = {}; 2619 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2620 struct super_block *sb = sbi->sb; 2621 int err = 0; 2622 2623 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 2624 if (!IS_ERR(transfer_to[PRJQUOTA])) { 2625 err = __dquot_transfer(inode, transfer_to); 2626 if (err) 2627 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2628 dqput(transfer_to[PRJQUOTA]); 2629 } 2630 return err; 2631 } 2632 2633 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 2634 { 2635 struct inode *inode = file_inode(filp); 2636 struct f2fs_inode_info *fi = F2FS_I(inode); 2637 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2638 struct super_block *sb = sbi->sb; 2639 struct page *ipage; 2640 kprojid_t kprojid; 2641 int err; 2642 2643 if (!f2fs_sb_has_project_quota(sb)) { 2644 if (projid != F2FS_DEF_PROJID) 2645 return -EOPNOTSUPP; 2646 else 2647 return 0; 2648 } 2649 2650 if (!f2fs_has_extra_attr(inode)) 2651 return -EOPNOTSUPP; 2652 2653 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 2654 2655 if (projid_eq(kprojid, F2FS_I(inode)->i_projid)) 2656 return 0; 2657 2658 err = -EPERM; 2659 /* Is it quota file? Do not allow user to mess with it */ 2660 if (IS_NOQUOTA(inode)) 2661 return err; 2662 2663 ipage = f2fs_get_node_page(sbi, inode->i_ino); 2664 if (IS_ERR(ipage)) 2665 return PTR_ERR(ipage); 2666 2667 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize, 2668 i_projid)) { 2669 err = -EOVERFLOW; 2670 f2fs_put_page(ipage, 1); 2671 return err; 2672 } 2673 f2fs_put_page(ipage, 1); 2674 2675 err = dquot_initialize(inode); 2676 if (err) 2677 return err; 2678 2679 f2fs_lock_op(sbi); 2680 err = f2fs_transfer_project_quota(inode, kprojid); 2681 if (err) 2682 goto out_unlock; 2683 2684 F2FS_I(inode)->i_projid = kprojid; 2685 inode->i_ctime = current_time(inode); 2686 f2fs_mark_inode_dirty_sync(inode, true); 2687 out_unlock: 2688 f2fs_unlock_op(sbi); 2689 return err; 2690 } 2691 #else 2692 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 2693 { 2694 return 0; 2695 } 2696 2697 static int f2fs_ioc_setproject(struct file *filp, __u32 projid) 2698 { 2699 if (projid != F2FS_DEF_PROJID) 2700 return -EOPNOTSUPP; 2701 return 0; 2702 } 2703 #endif 2704 2705 /* Transfer internal flags to xflags */ 2706 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags) 2707 { 2708 __u32 xflags = 0; 2709 2710 if (iflags & F2FS_SYNC_FL) 2711 xflags |= FS_XFLAG_SYNC; 2712 if (iflags & F2FS_IMMUTABLE_FL) 2713 xflags |= FS_XFLAG_IMMUTABLE; 2714 if (iflags & F2FS_APPEND_FL) 2715 xflags |= FS_XFLAG_APPEND; 2716 if (iflags & F2FS_NODUMP_FL) 2717 xflags |= FS_XFLAG_NODUMP; 2718 if (iflags & F2FS_NOATIME_FL) 2719 xflags |= FS_XFLAG_NOATIME; 2720 if (iflags & F2FS_PROJINHERIT_FL) 2721 xflags |= FS_XFLAG_PROJINHERIT; 2722 return xflags; 2723 } 2724 2725 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \ 2726 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \ 2727 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT) 2728 2729 /* Transfer xflags flags to internal */ 2730 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags) 2731 { 2732 unsigned long iflags = 0; 2733 2734 if (xflags & FS_XFLAG_SYNC) 2735 iflags |= F2FS_SYNC_FL; 2736 if (xflags & FS_XFLAG_IMMUTABLE) 2737 iflags |= F2FS_IMMUTABLE_FL; 2738 if (xflags & FS_XFLAG_APPEND) 2739 iflags |= F2FS_APPEND_FL; 2740 if (xflags & FS_XFLAG_NODUMP) 2741 iflags |= F2FS_NODUMP_FL; 2742 if (xflags & FS_XFLAG_NOATIME) 2743 iflags |= F2FS_NOATIME_FL; 2744 if (xflags & FS_XFLAG_PROJINHERIT) 2745 iflags |= F2FS_PROJINHERIT_FL; 2746 2747 return iflags; 2748 } 2749 2750 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg) 2751 { 2752 struct inode *inode = file_inode(filp); 2753 struct f2fs_inode_info *fi = F2FS_I(inode); 2754 struct fsxattr fa; 2755 2756 memset(&fa, 0, sizeof(struct fsxattr)); 2757 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags & 2758 F2FS_FL_USER_VISIBLE); 2759 2760 if (f2fs_sb_has_project_quota(inode->i_sb)) 2761 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns, 2762 fi->i_projid); 2763 2764 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa))) 2765 return -EFAULT; 2766 return 0; 2767 } 2768 2769 static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa) 2770 { 2771 /* 2772 * Project Quota ID state is only allowed to change from within the init 2773 * namespace. Enforce that restriction only if we are trying to change 2774 * the quota ID state. Everything else is allowed in user namespaces. 2775 */ 2776 if (current_user_ns() == &init_user_ns) 2777 return 0; 2778 2779 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid) 2780 return -EINVAL; 2781 2782 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) { 2783 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT)) 2784 return -EINVAL; 2785 } else { 2786 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT) 2787 return -EINVAL; 2788 } 2789 2790 return 0; 2791 } 2792 2793 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg) 2794 { 2795 struct inode *inode = file_inode(filp); 2796 struct f2fs_inode_info *fi = F2FS_I(inode); 2797 struct fsxattr fa; 2798 unsigned int flags; 2799 int err; 2800 2801 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa))) 2802 return -EFAULT; 2803 2804 /* Make sure caller has proper permission */ 2805 if (!inode_owner_or_capable(inode)) 2806 return -EACCES; 2807 2808 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS) 2809 return -EOPNOTSUPP; 2810 2811 flags = f2fs_xflags_to_iflags(fa.fsx_xflags); 2812 if (f2fs_mask_flags(inode->i_mode, flags) != flags) 2813 return -EOPNOTSUPP; 2814 2815 err = mnt_want_write_file(filp); 2816 if (err) 2817 return err; 2818 2819 inode_lock(inode); 2820 err = f2fs_ioctl_check_project(inode, &fa); 2821 if (err) 2822 goto out; 2823 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) | 2824 (flags & F2FS_FL_XFLAG_VISIBLE); 2825 err = __f2fs_ioc_setflags(inode, flags); 2826 if (err) 2827 goto out; 2828 2829 err = f2fs_ioc_setproject(filp, fa.fsx_projid); 2830 out: 2831 inode_unlock(inode); 2832 mnt_drop_write_file(filp); 2833 return err; 2834 } 2835 2836 int f2fs_pin_file_control(struct inode *inode, bool inc) 2837 { 2838 struct f2fs_inode_info *fi = F2FS_I(inode); 2839 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2840 2841 /* Use i_gc_failures for normal file as a risk signal. */ 2842 if (inc) 2843 f2fs_i_gc_failures_write(inode, 2844 fi->i_gc_failures[GC_FAILURE_PIN] + 1); 2845 2846 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) { 2847 f2fs_msg(sbi->sb, KERN_WARNING, 2848 "%s: Enable GC = ino %lx after %x GC trials\n", 2849 __func__, inode->i_ino, 2850 fi->i_gc_failures[GC_FAILURE_PIN]); 2851 clear_inode_flag(inode, FI_PIN_FILE); 2852 return -EAGAIN; 2853 } 2854 return 0; 2855 } 2856 2857 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 2858 { 2859 struct inode *inode = file_inode(filp); 2860 __u32 pin; 2861 int ret = 0; 2862 2863 if (!inode_owner_or_capable(inode)) 2864 return -EACCES; 2865 2866 if (get_user(pin, (__u32 __user *)arg)) 2867 return -EFAULT; 2868 2869 if (!S_ISREG(inode->i_mode)) 2870 return -EINVAL; 2871 2872 if (f2fs_readonly(F2FS_I_SB(inode)->sb)) 2873 return -EROFS; 2874 2875 ret = mnt_want_write_file(filp); 2876 if (ret) 2877 return ret; 2878 2879 inode_lock(inode); 2880 2881 if (f2fs_should_update_outplace(inode, NULL)) { 2882 ret = -EINVAL; 2883 goto out; 2884 } 2885 2886 if (!pin) { 2887 clear_inode_flag(inode, FI_PIN_FILE); 2888 f2fs_i_gc_failures_write(inode, 0); 2889 goto done; 2890 } 2891 2892 if (f2fs_pin_file_control(inode, false)) { 2893 ret = -EAGAIN; 2894 goto out; 2895 } 2896 ret = f2fs_convert_inline_inode(inode); 2897 if (ret) 2898 goto out; 2899 2900 set_inode_flag(inode, FI_PIN_FILE); 2901 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 2902 done: 2903 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2904 out: 2905 inode_unlock(inode); 2906 mnt_drop_write_file(filp); 2907 return ret; 2908 } 2909 2910 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 2911 { 2912 struct inode *inode = file_inode(filp); 2913 __u32 pin = 0; 2914 2915 if (is_inode_flag_set(inode, FI_PIN_FILE)) 2916 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN]; 2917 return put_user(pin, (u32 __user *)arg); 2918 } 2919 2920 int f2fs_precache_extents(struct inode *inode) 2921 { 2922 struct f2fs_inode_info *fi = F2FS_I(inode); 2923 struct f2fs_map_blocks map; 2924 pgoff_t m_next_extent; 2925 loff_t end; 2926 int err; 2927 2928 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 2929 return -EOPNOTSUPP; 2930 2931 map.m_lblk = 0; 2932 map.m_next_pgofs = NULL; 2933 map.m_next_extent = &m_next_extent; 2934 map.m_seg_type = NO_CHECK_TYPE; 2935 end = F2FS_I_SB(inode)->max_file_blocks; 2936 2937 while (map.m_lblk < end) { 2938 map.m_len = end - map.m_lblk; 2939 2940 down_write(&fi->i_gc_rwsem[WRITE]); 2941 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE); 2942 up_write(&fi->i_gc_rwsem[WRITE]); 2943 if (err) 2944 return err; 2945 2946 map.m_lblk = m_next_extent; 2947 } 2948 2949 return err; 2950 } 2951 2952 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg) 2953 { 2954 return f2fs_precache_extents(file_inode(filp)); 2955 } 2956 2957 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 2958 { 2959 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 2960 return -EIO; 2961 2962 switch (cmd) { 2963 case F2FS_IOC_GETFLAGS: 2964 return f2fs_ioc_getflags(filp, arg); 2965 case F2FS_IOC_SETFLAGS: 2966 return f2fs_ioc_setflags(filp, arg); 2967 case F2FS_IOC_GETVERSION: 2968 return f2fs_ioc_getversion(filp, arg); 2969 case F2FS_IOC_START_ATOMIC_WRITE: 2970 return f2fs_ioc_start_atomic_write(filp); 2971 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 2972 return f2fs_ioc_commit_atomic_write(filp); 2973 case F2FS_IOC_START_VOLATILE_WRITE: 2974 return f2fs_ioc_start_volatile_write(filp); 2975 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 2976 return f2fs_ioc_release_volatile_write(filp); 2977 case F2FS_IOC_ABORT_VOLATILE_WRITE: 2978 return f2fs_ioc_abort_volatile_write(filp); 2979 case F2FS_IOC_SHUTDOWN: 2980 return f2fs_ioc_shutdown(filp, arg); 2981 case FITRIM: 2982 return f2fs_ioc_fitrim(filp, arg); 2983 case F2FS_IOC_SET_ENCRYPTION_POLICY: 2984 return f2fs_ioc_set_encryption_policy(filp, arg); 2985 case F2FS_IOC_GET_ENCRYPTION_POLICY: 2986 return f2fs_ioc_get_encryption_policy(filp, arg); 2987 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 2988 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 2989 case F2FS_IOC_GARBAGE_COLLECT: 2990 return f2fs_ioc_gc(filp, arg); 2991 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 2992 return f2fs_ioc_gc_range(filp, arg); 2993 case F2FS_IOC_WRITE_CHECKPOINT: 2994 return f2fs_ioc_write_checkpoint(filp, arg); 2995 case F2FS_IOC_DEFRAGMENT: 2996 return f2fs_ioc_defragment(filp, arg); 2997 case F2FS_IOC_MOVE_RANGE: 2998 return f2fs_ioc_move_range(filp, arg); 2999 case F2FS_IOC_FLUSH_DEVICE: 3000 return f2fs_ioc_flush_device(filp, arg); 3001 case F2FS_IOC_GET_FEATURES: 3002 return f2fs_ioc_get_features(filp, arg); 3003 case F2FS_IOC_FSGETXATTR: 3004 return f2fs_ioc_fsgetxattr(filp, arg); 3005 case F2FS_IOC_FSSETXATTR: 3006 return f2fs_ioc_fssetxattr(filp, arg); 3007 case F2FS_IOC_GET_PIN_FILE: 3008 return f2fs_ioc_get_pin_file(filp, arg); 3009 case F2FS_IOC_SET_PIN_FILE: 3010 return f2fs_ioc_set_pin_file(filp, arg); 3011 case F2FS_IOC_PRECACHE_EXTENTS: 3012 return f2fs_ioc_precache_extents(filp, arg); 3013 default: 3014 return -ENOTTY; 3015 } 3016 } 3017 3018 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 3019 { 3020 struct file *file = iocb->ki_filp; 3021 struct inode *inode = file_inode(file); 3022 ssize_t ret; 3023 3024 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3025 return -EIO; 3026 3027 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT)) 3028 return -EINVAL; 3029 3030 if (!inode_trylock(inode)) { 3031 if (iocb->ki_flags & IOCB_NOWAIT) 3032 return -EAGAIN; 3033 inode_lock(inode); 3034 } 3035 3036 ret = generic_write_checks(iocb, from); 3037 if (ret > 0) { 3038 bool preallocated = false; 3039 size_t target_size = 0; 3040 int err; 3041 3042 if (iov_iter_fault_in_readable(from, iov_iter_count(from))) 3043 set_inode_flag(inode, FI_NO_PREALLOC); 3044 3045 if ((iocb->ki_flags & IOCB_NOWAIT) && 3046 (iocb->ki_flags & IOCB_DIRECT)) { 3047 if (!f2fs_overwrite_io(inode, iocb->ki_pos, 3048 iov_iter_count(from)) || 3049 f2fs_has_inline_data(inode) || 3050 f2fs_force_buffered_io(inode, 3051 iocb, from)) { 3052 clear_inode_flag(inode, 3053 FI_NO_PREALLOC); 3054 inode_unlock(inode); 3055 return -EAGAIN; 3056 } 3057 3058 } else { 3059 preallocated = true; 3060 target_size = iocb->ki_pos + iov_iter_count(from); 3061 3062 err = f2fs_preallocate_blocks(iocb, from); 3063 if (err) { 3064 clear_inode_flag(inode, FI_NO_PREALLOC); 3065 inode_unlock(inode); 3066 return err; 3067 } 3068 } 3069 ret = __generic_file_write_iter(iocb, from); 3070 clear_inode_flag(inode, FI_NO_PREALLOC); 3071 3072 /* if we couldn't write data, we should deallocate blocks. */ 3073 if (preallocated && i_size_read(inode) < target_size) 3074 f2fs_truncate(inode); 3075 3076 if (ret > 0) 3077 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret); 3078 } 3079 inode_unlock(inode); 3080 3081 if (ret > 0) 3082 ret = generic_write_sync(iocb, ret); 3083 return ret; 3084 } 3085 3086 #ifdef CONFIG_COMPAT 3087 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 3088 { 3089 switch (cmd) { 3090 case F2FS_IOC32_GETFLAGS: 3091 cmd = F2FS_IOC_GETFLAGS; 3092 break; 3093 case F2FS_IOC32_SETFLAGS: 3094 cmd = F2FS_IOC_SETFLAGS; 3095 break; 3096 case F2FS_IOC32_GETVERSION: 3097 cmd = F2FS_IOC_GETVERSION; 3098 break; 3099 case F2FS_IOC_START_ATOMIC_WRITE: 3100 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 3101 case F2FS_IOC_START_VOLATILE_WRITE: 3102 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 3103 case F2FS_IOC_ABORT_VOLATILE_WRITE: 3104 case F2FS_IOC_SHUTDOWN: 3105 case F2FS_IOC_SET_ENCRYPTION_POLICY: 3106 case F2FS_IOC_GET_ENCRYPTION_PWSALT: 3107 case F2FS_IOC_GET_ENCRYPTION_POLICY: 3108 case F2FS_IOC_GARBAGE_COLLECT: 3109 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 3110 case F2FS_IOC_WRITE_CHECKPOINT: 3111 case F2FS_IOC_DEFRAGMENT: 3112 case F2FS_IOC_MOVE_RANGE: 3113 case F2FS_IOC_FLUSH_DEVICE: 3114 case F2FS_IOC_GET_FEATURES: 3115 case F2FS_IOC_FSGETXATTR: 3116 case F2FS_IOC_FSSETXATTR: 3117 case F2FS_IOC_GET_PIN_FILE: 3118 case F2FS_IOC_SET_PIN_FILE: 3119 case F2FS_IOC_PRECACHE_EXTENTS: 3120 break; 3121 default: 3122 return -ENOIOCTLCMD; 3123 } 3124 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 3125 } 3126 #endif 3127 3128 const struct file_operations f2fs_file_operations = { 3129 .llseek = f2fs_llseek, 3130 .read_iter = generic_file_read_iter, 3131 .write_iter = f2fs_file_write_iter, 3132 .open = f2fs_file_open, 3133 .release = f2fs_release_file, 3134 .mmap = f2fs_file_mmap, 3135 .flush = f2fs_file_flush, 3136 .fsync = f2fs_sync_file, 3137 .fallocate = f2fs_fallocate, 3138 .unlocked_ioctl = f2fs_ioctl, 3139 #ifdef CONFIG_COMPAT 3140 .compat_ioctl = f2fs_compat_ioctl, 3141 #endif 3142 .splice_read = generic_file_splice_read, 3143 .splice_write = iter_file_splice_write, 3144 }; 3145