1 /* 2 * fs/f2fs/file.c 3 * 4 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 5 * http://www.samsung.com/ 6 * 7 * This program is free software; you can redistribute it and/or modify 8 * it under the terms of the GNU General Public License version 2 as 9 * published by the Free Software Foundation. 10 */ 11 #include <linux/fs.h> 12 #include <linux/f2fs_fs.h> 13 #include <linux/stat.h> 14 #include <linux/buffer_head.h> 15 #include <linux/writeback.h> 16 #include <linux/falloc.h> 17 #include <linux/types.h> 18 #include <linux/uaccess.h> 19 #include <linux/mount.h> 20 21 #include "f2fs.h" 22 #include "node.h" 23 #include "segment.h" 24 #include "xattr.h" 25 #include "acl.h" 26 27 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma, 28 struct vm_fault *vmf) 29 { 30 struct page *page = vmf->page; 31 struct inode *inode = vma->vm_file->f_path.dentry->d_inode; 32 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 33 block_t old_blk_addr; 34 struct dnode_of_data dn; 35 int err; 36 37 f2fs_balance_fs(sbi); 38 39 sb_start_pagefault(inode->i_sb); 40 41 mutex_lock_op(sbi, DATA_NEW); 42 43 /* block allocation */ 44 set_new_dnode(&dn, inode, NULL, NULL, 0); 45 err = get_dnode_of_data(&dn, page->index, 0); 46 if (err) { 47 mutex_unlock_op(sbi, DATA_NEW); 48 goto out; 49 } 50 51 old_blk_addr = dn.data_blkaddr; 52 53 if (old_blk_addr == NULL_ADDR) { 54 err = reserve_new_block(&dn); 55 if (err) { 56 f2fs_put_dnode(&dn); 57 mutex_unlock_op(sbi, DATA_NEW); 58 goto out; 59 } 60 } 61 f2fs_put_dnode(&dn); 62 63 mutex_unlock_op(sbi, DATA_NEW); 64 65 lock_page(page); 66 if (page->mapping != inode->i_mapping || 67 page_offset(page) >= i_size_read(inode) || 68 !PageUptodate(page)) { 69 unlock_page(page); 70 err = -EFAULT; 71 goto out; 72 } 73 74 /* 75 * check to see if the page is mapped already (no holes) 76 */ 77 if (PageMappedToDisk(page)) 78 goto out; 79 80 /* fill the page */ 81 wait_on_page_writeback(page); 82 83 /* page is wholly or partially inside EOF */ 84 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) { 85 unsigned offset; 86 offset = i_size_read(inode) & ~PAGE_CACHE_MASK; 87 zero_user_segment(page, offset, PAGE_CACHE_SIZE); 88 } 89 set_page_dirty(page); 90 SetPageUptodate(page); 91 92 file_update_time(vma->vm_file); 93 out: 94 sb_end_pagefault(inode->i_sb); 95 return block_page_mkwrite_return(err); 96 } 97 98 static const struct vm_operations_struct f2fs_file_vm_ops = { 99 .fault = filemap_fault, 100 .page_mkwrite = f2fs_vm_page_mkwrite, 101 }; 102 103 static int need_to_sync_dir(struct f2fs_sb_info *sbi, struct inode *inode) 104 { 105 struct dentry *dentry; 106 nid_t pino; 107 108 inode = igrab(inode); 109 dentry = d_find_any_alias(inode); 110 if (!dentry) { 111 iput(inode); 112 return 0; 113 } 114 pino = dentry->d_parent->d_inode->i_ino; 115 dput(dentry); 116 iput(inode); 117 return !is_checkpointed_node(sbi, pino); 118 } 119 120 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 121 { 122 struct inode *inode = file->f_mapping->host; 123 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 124 unsigned long long cur_version; 125 int ret = 0; 126 bool need_cp = false; 127 struct writeback_control wbc = { 128 .sync_mode = WB_SYNC_ALL, 129 .nr_to_write = LONG_MAX, 130 .for_reclaim = 0, 131 }; 132 133 if (inode->i_sb->s_flags & MS_RDONLY) 134 return 0; 135 136 ret = filemap_write_and_wait_range(inode->i_mapping, start, end); 137 if (ret) 138 return ret; 139 140 mutex_lock(&inode->i_mutex); 141 142 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) 143 goto out; 144 145 mutex_lock(&sbi->cp_mutex); 146 cur_version = le64_to_cpu(F2FS_CKPT(sbi)->checkpoint_ver); 147 mutex_unlock(&sbi->cp_mutex); 148 149 if (F2FS_I(inode)->data_version != cur_version && 150 !(inode->i_state & I_DIRTY)) 151 goto out; 152 F2FS_I(inode)->data_version--; 153 154 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1) 155 need_cp = true; 156 if (is_inode_flag_set(F2FS_I(inode), FI_NEED_CP)) 157 need_cp = true; 158 if (!space_for_roll_forward(sbi)) 159 need_cp = true; 160 if (need_to_sync_dir(sbi, inode)) 161 need_cp = true; 162 163 if (need_cp) { 164 /* all the dirty node pages should be flushed for POR */ 165 ret = f2fs_sync_fs(inode->i_sb, 1); 166 clear_inode_flag(F2FS_I(inode), FI_NEED_CP); 167 } else { 168 /* if there is no written node page, write its inode page */ 169 while (!sync_node_pages(sbi, inode->i_ino, &wbc)) { 170 ret = f2fs_write_inode(inode, NULL); 171 if (ret) 172 goto out; 173 } 174 filemap_fdatawait_range(sbi->node_inode->i_mapping, 175 0, LONG_MAX); 176 } 177 out: 178 mutex_unlock(&inode->i_mutex); 179 return ret; 180 } 181 182 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 183 { 184 file_accessed(file); 185 vma->vm_ops = &f2fs_file_vm_ops; 186 return 0; 187 } 188 189 static int truncate_data_blocks_range(struct dnode_of_data *dn, int count) 190 { 191 int nr_free = 0, ofs = dn->ofs_in_node; 192 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb); 193 struct f2fs_node *raw_node; 194 __le32 *addr; 195 196 raw_node = page_address(dn->node_page); 197 addr = blkaddr_in_node(raw_node) + ofs; 198 199 for ( ; count > 0; count--, addr++, dn->ofs_in_node++) { 200 block_t blkaddr = le32_to_cpu(*addr); 201 if (blkaddr == NULL_ADDR) 202 continue; 203 204 update_extent_cache(NULL_ADDR, dn); 205 invalidate_blocks(sbi, blkaddr); 206 dec_valid_block_count(sbi, dn->inode, 1); 207 nr_free++; 208 } 209 if (nr_free) { 210 set_page_dirty(dn->node_page); 211 sync_inode_page(dn); 212 } 213 dn->ofs_in_node = ofs; 214 return nr_free; 215 } 216 217 void truncate_data_blocks(struct dnode_of_data *dn) 218 { 219 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK); 220 } 221 222 static void truncate_partial_data_page(struct inode *inode, u64 from) 223 { 224 unsigned offset = from & (PAGE_CACHE_SIZE - 1); 225 struct page *page; 226 227 if (!offset) 228 return; 229 230 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT); 231 if (IS_ERR(page)) 232 return; 233 234 lock_page(page); 235 wait_on_page_writeback(page); 236 zero_user(page, offset, PAGE_CACHE_SIZE - offset); 237 set_page_dirty(page); 238 f2fs_put_page(page, 1); 239 } 240 241 static int truncate_blocks(struct inode *inode, u64 from) 242 { 243 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 244 unsigned int blocksize = inode->i_sb->s_blocksize; 245 struct dnode_of_data dn; 246 pgoff_t free_from; 247 int count = 0; 248 int err; 249 250 free_from = (pgoff_t) 251 ((from + blocksize - 1) >> (sbi->log_blocksize)); 252 253 mutex_lock_op(sbi, DATA_TRUNC); 254 255 set_new_dnode(&dn, inode, NULL, NULL, 0); 256 err = get_dnode_of_data(&dn, free_from, RDONLY_NODE); 257 if (err) { 258 if (err == -ENOENT) 259 goto free_next; 260 mutex_unlock_op(sbi, DATA_TRUNC); 261 return err; 262 } 263 264 if (IS_INODE(dn.node_page)) 265 count = ADDRS_PER_INODE; 266 else 267 count = ADDRS_PER_BLOCK; 268 269 count -= dn.ofs_in_node; 270 BUG_ON(count < 0); 271 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 272 truncate_data_blocks_range(&dn, count); 273 free_from += count; 274 } 275 276 f2fs_put_dnode(&dn); 277 free_next: 278 err = truncate_inode_blocks(inode, free_from); 279 mutex_unlock_op(sbi, DATA_TRUNC); 280 281 /* lastly zero out the first data page */ 282 truncate_partial_data_page(inode, from); 283 284 return err; 285 } 286 287 void f2fs_truncate(struct inode *inode) 288 { 289 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 290 S_ISLNK(inode->i_mode))) 291 return; 292 293 if (!truncate_blocks(inode, i_size_read(inode))) { 294 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 295 mark_inode_dirty(inode); 296 } 297 298 f2fs_balance_fs(F2FS_SB(inode->i_sb)); 299 } 300 301 static int f2fs_getattr(struct vfsmount *mnt, 302 struct dentry *dentry, struct kstat *stat) 303 { 304 struct inode *inode = dentry->d_inode; 305 generic_fillattr(inode, stat); 306 stat->blocks <<= 3; 307 return 0; 308 } 309 310 #ifdef CONFIG_F2FS_FS_POSIX_ACL 311 static void __setattr_copy(struct inode *inode, const struct iattr *attr) 312 { 313 struct f2fs_inode_info *fi = F2FS_I(inode); 314 unsigned int ia_valid = attr->ia_valid; 315 316 if (ia_valid & ATTR_UID) 317 inode->i_uid = attr->ia_uid; 318 if (ia_valid & ATTR_GID) 319 inode->i_gid = attr->ia_gid; 320 if (ia_valid & ATTR_ATIME) 321 inode->i_atime = timespec_trunc(attr->ia_atime, 322 inode->i_sb->s_time_gran); 323 if (ia_valid & ATTR_MTIME) 324 inode->i_mtime = timespec_trunc(attr->ia_mtime, 325 inode->i_sb->s_time_gran); 326 if (ia_valid & ATTR_CTIME) 327 inode->i_ctime = timespec_trunc(attr->ia_ctime, 328 inode->i_sb->s_time_gran); 329 if (ia_valid & ATTR_MODE) { 330 umode_t mode = attr->ia_mode; 331 332 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) 333 mode &= ~S_ISGID; 334 set_acl_inode(fi, mode); 335 } 336 } 337 #else 338 #define __setattr_copy setattr_copy 339 #endif 340 341 int f2fs_setattr(struct dentry *dentry, struct iattr *attr) 342 { 343 struct inode *inode = dentry->d_inode; 344 struct f2fs_inode_info *fi = F2FS_I(inode); 345 int err; 346 347 err = inode_change_ok(inode, attr); 348 if (err) 349 return err; 350 351 if ((attr->ia_valid & ATTR_SIZE) && 352 attr->ia_size != i_size_read(inode)) { 353 truncate_setsize(inode, attr->ia_size); 354 f2fs_truncate(inode); 355 } 356 357 __setattr_copy(inode, attr); 358 359 if (attr->ia_valid & ATTR_MODE) { 360 err = f2fs_acl_chmod(inode); 361 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) { 362 inode->i_mode = fi->i_acl_mode; 363 clear_inode_flag(fi, FI_ACL_MODE); 364 } 365 } 366 367 mark_inode_dirty(inode); 368 return err; 369 } 370 371 const struct inode_operations f2fs_file_inode_operations = { 372 .getattr = f2fs_getattr, 373 .setattr = f2fs_setattr, 374 .get_acl = f2fs_get_acl, 375 #ifdef CONFIG_F2FS_FS_XATTR 376 .setxattr = generic_setxattr, 377 .getxattr = generic_getxattr, 378 .listxattr = f2fs_listxattr, 379 .removexattr = generic_removexattr, 380 #endif 381 }; 382 383 static void fill_zero(struct inode *inode, pgoff_t index, 384 loff_t start, loff_t len) 385 { 386 struct page *page; 387 388 if (!len) 389 return; 390 391 page = get_new_data_page(inode, index, false); 392 393 if (!IS_ERR(page)) { 394 wait_on_page_writeback(page); 395 zero_user(page, start, len); 396 set_page_dirty(page); 397 f2fs_put_page(page, 1); 398 } 399 } 400 401 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 402 { 403 pgoff_t index; 404 int err; 405 406 for (index = pg_start; index < pg_end; index++) { 407 struct dnode_of_data dn; 408 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 409 410 mutex_lock_op(sbi, DATA_TRUNC); 411 set_new_dnode(&dn, inode, NULL, NULL, 0); 412 err = get_dnode_of_data(&dn, index, RDONLY_NODE); 413 if (err) { 414 mutex_unlock_op(sbi, DATA_TRUNC); 415 if (err == -ENOENT) 416 continue; 417 return err; 418 } 419 420 if (dn.data_blkaddr != NULL_ADDR) 421 truncate_data_blocks_range(&dn, 1); 422 f2fs_put_dnode(&dn); 423 mutex_unlock_op(sbi, DATA_TRUNC); 424 } 425 return 0; 426 } 427 428 static int punch_hole(struct inode *inode, loff_t offset, loff_t len, int mode) 429 { 430 pgoff_t pg_start, pg_end; 431 loff_t off_start, off_end; 432 int ret = 0; 433 434 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT; 435 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT; 436 437 off_start = offset & (PAGE_CACHE_SIZE - 1); 438 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1); 439 440 if (pg_start == pg_end) { 441 fill_zero(inode, pg_start, off_start, 442 off_end - off_start); 443 } else { 444 if (off_start) 445 fill_zero(inode, pg_start++, off_start, 446 PAGE_CACHE_SIZE - off_start); 447 if (off_end) 448 fill_zero(inode, pg_end, 0, off_end); 449 450 if (pg_start < pg_end) { 451 struct address_space *mapping = inode->i_mapping; 452 loff_t blk_start, blk_end; 453 454 blk_start = pg_start << PAGE_CACHE_SHIFT; 455 blk_end = pg_end << PAGE_CACHE_SHIFT; 456 truncate_inode_pages_range(mapping, blk_start, 457 blk_end - 1); 458 ret = truncate_hole(inode, pg_start, pg_end); 459 } 460 } 461 462 if (!(mode & FALLOC_FL_KEEP_SIZE) && 463 i_size_read(inode) <= (offset + len)) { 464 i_size_write(inode, offset); 465 mark_inode_dirty(inode); 466 } 467 468 return ret; 469 } 470 471 static int expand_inode_data(struct inode *inode, loff_t offset, 472 loff_t len, int mode) 473 { 474 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 475 pgoff_t index, pg_start, pg_end; 476 loff_t new_size = i_size_read(inode); 477 loff_t off_start, off_end; 478 int ret = 0; 479 480 ret = inode_newsize_ok(inode, (len + offset)); 481 if (ret) 482 return ret; 483 484 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT; 485 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT; 486 487 off_start = offset & (PAGE_CACHE_SIZE - 1); 488 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1); 489 490 for (index = pg_start; index <= pg_end; index++) { 491 struct dnode_of_data dn; 492 493 mutex_lock_op(sbi, DATA_NEW); 494 495 set_new_dnode(&dn, inode, NULL, NULL, 0); 496 ret = get_dnode_of_data(&dn, index, 0); 497 if (ret) { 498 mutex_unlock_op(sbi, DATA_NEW); 499 break; 500 } 501 502 if (dn.data_blkaddr == NULL_ADDR) { 503 ret = reserve_new_block(&dn); 504 if (ret) { 505 f2fs_put_dnode(&dn); 506 mutex_unlock_op(sbi, DATA_NEW); 507 break; 508 } 509 } 510 f2fs_put_dnode(&dn); 511 512 mutex_unlock_op(sbi, DATA_NEW); 513 514 if (pg_start == pg_end) 515 new_size = offset + len; 516 else if (index == pg_start && off_start) 517 new_size = (index + 1) << PAGE_CACHE_SHIFT; 518 else if (index == pg_end) 519 new_size = (index << PAGE_CACHE_SHIFT) + off_end; 520 else 521 new_size += PAGE_CACHE_SIZE; 522 } 523 524 if (!(mode & FALLOC_FL_KEEP_SIZE) && 525 i_size_read(inode) < new_size) { 526 i_size_write(inode, new_size); 527 mark_inode_dirty(inode); 528 } 529 530 return ret; 531 } 532 533 static long f2fs_fallocate(struct file *file, int mode, 534 loff_t offset, loff_t len) 535 { 536 struct inode *inode = file->f_path.dentry->d_inode; 537 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 538 long ret; 539 540 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) 541 return -EOPNOTSUPP; 542 543 if (mode & FALLOC_FL_PUNCH_HOLE) 544 ret = punch_hole(inode, offset, len, mode); 545 else 546 ret = expand_inode_data(inode, offset, len, mode); 547 548 f2fs_balance_fs(sbi); 549 return ret; 550 } 551 552 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL)) 553 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL) 554 555 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags) 556 { 557 if (S_ISDIR(mode)) 558 return flags; 559 else if (S_ISREG(mode)) 560 return flags & F2FS_REG_FLMASK; 561 else 562 return flags & F2FS_OTHER_FLMASK; 563 } 564 565 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 566 { 567 struct inode *inode = filp->f_dentry->d_inode; 568 struct f2fs_inode_info *fi = F2FS_I(inode); 569 unsigned int flags; 570 int ret; 571 572 switch (cmd) { 573 case FS_IOC_GETFLAGS: 574 flags = fi->i_flags & FS_FL_USER_VISIBLE; 575 return put_user(flags, (int __user *) arg); 576 case FS_IOC_SETFLAGS: 577 { 578 unsigned int oldflags; 579 580 ret = mnt_want_write(filp->f_path.mnt); 581 if (ret) 582 return ret; 583 584 if (!inode_owner_or_capable(inode)) { 585 ret = -EACCES; 586 goto out; 587 } 588 589 if (get_user(flags, (int __user *) arg)) { 590 ret = -EFAULT; 591 goto out; 592 } 593 594 flags = f2fs_mask_flags(inode->i_mode, flags); 595 596 mutex_lock(&inode->i_mutex); 597 598 oldflags = fi->i_flags; 599 600 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 601 if (!capable(CAP_LINUX_IMMUTABLE)) { 602 mutex_unlock(&inode->i_mutex); 603 ret = -EPERM; 604 goto out; 605 } 606 } 607 608 flags = flags & FS_FL_USER_MODIFIABLE; 609 flags |= oldflags & ~FS_FL_USER_MODIFIABLE; 610 fi->i_flags = flags; 611 mutex_unlock(&inode->i_mutex); 612 613 f2fs_set_inode_flags(inode); 614 inode->i_ctime = CURRENT_TIME; 615 mark_inode_dirty(inode); 616 out: 617 mnt_drop_write(filp->f_path.mnt); 618 return ret; 619 } 620 default: 621 return -ENOTTY; 622 } 623 } 624 625 const struct file_operations f2fs_file_operations = { 626 .llseek = generic_file_llseek, 627 .read = do_sync_read, 628 .write = do_sync_write, 629 .aio_read = generic_file_aio_read, 630 .aio_write = generic_file_aio_write, 631 .open = generic_file_open, 632 .mmap = f2fs_file_mmap, 633 .fsync = f2fs_sync_file, 634 .fallocate = f2fs_fallocate, 635 .unlocked_ioctl = f2fs_ioctl, 636 .splice_read = generic_file_splice_read, 637 .splice_write = generic_file_splice_write, 638 }; 639