xfs: fix type usage
[linux/fpc-iii.git] / fs / f2fs / file.c
blob517e112c8a9a943467492435c9f15c1668e079a8
1 /*
2 * fs/f2fs/file.c
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
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.
11 #include <linux/fs.h>
12 #include <linux/f2fs_fs.h>
13 #include <linux/stat.h>
14 #include <linux/buffer_head.h>
15 #include <linux/writeback.h>
16 #include <linux/blkdev.h>
17 #include <linux/falloc.h>
18 #include <linux/types.h>
19 #include <linux/compat.h>
20 #include <linux/uaccess.h>
21 #include <linux/mount.h>
22 #include <linux/pagevec.h>
23 #include <linux/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "acl.h"
32 #include "gc.h"
33 #include "trace.h"
34 #include <trace/events/f2fs.h>
36 static int f2fs_filemap_fault(struct vm_fault *vmf)
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 int err;
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 err = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
45 return err;
48 static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn;
54 int err;
56 sb_start_pagefault(inode->i_sb);
58 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
60 /* block allocation */
61 f2fs_lock_op(sbi);
62 set_new_dnode(&dn, inode, NULL, NULL, 0);
63 err = f2fs_reserve_block(&dn, page->index);
64 if (err) {
65 f2fs_unlock_op(sbi);
66 goto out;
68 f2fs_put_dnode(&dn);
69 f2fs_unlock_op(sbi);
71 f2fs_balance_fs(sbi, dn.node_changed);
73 file_update_time(vmf->vma->vm_file);
74 down_read(&F2FS_I(inode)->i_mmap_sem);
75 lock_page(page);
76 if (unlikely(page->mapping != inode->i_mapping ||
77 page_offset(page) > i_size_read(inode) ||
78 !PageUptodate(page))) {
79 unlock_page(page);
80 err = -EFAULT;
81 goto out_sem;
85 * check to see if the page is mapped already (no holes)
87 if (PageMappedToDisk(page))
88 goto mapped;
90 /* page is wholly or partially inside EOF */
91 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
92 i_size_read(inode)) {
93 unsigned offset;
94 offset = i_size_read(inode) & ~PAGE_MASK;
95 zero_user_segment(page, offset, PAGE_SIZE);
97 set_page_dirty(page);
98 if (!PageUptodate(page))
99 SetPageUptodate(page);
101 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
103 trace_f2fs_vm_page_mkwrite(page, DATA);
104 mapped:
105 /* fill the page */
106 f2fs_wait_on_page_writeback(page, DATA, false);
108 /* wait for GCed encrypted page writeback */
109 if (f2fs_encrypted_file(inode))
110 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
112 out_sem:
113 up_read(&F2FS_I(inode)->i_mmap_sem);
114 out:
115 sb_end_pagefault(inode->i_sb);
116 f2fs_update_time(sbi, REQ_TIME);
117 return block_page_mkwrite_return(err);
120 static const struct vm_operations_struct f2fs_file_vm_ops = {
121 .fault = f2fs_filemap_fault,
122 .map_pages = filemap_map_pages,
123 .page_mkwrite = f2fs_vm_page_mkwrite,
126 static int get_parent_ino(struct inode *inode, nid_t *pino)
128 struct dentry *dentry;
130 inode = igrab(inode);
131 dentry = d_find_any_alias(inode);
132 iput(inode);
133 if (!dentry)
134 return 0;
136 *pino = parent_ino(dentry);
137 dput(dentry);
138 return 1;
141 static inline bool need_do_checkpoint(struct inode *inode)
143 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
144 bool need_cp = false;
146 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
147 need_cp = true;
148 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
149 need_cp = true;
150 else if (file_wrong_pino(inode))
151 need_cp = true;
152 else if (!space_for_roll_forward(sbi))
153 need_cp = true;
154 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
155 need_cp = true;
156 else if (test_opt(sbi, FASTBOOT))
157 need_cp = true;
158 else if (sbi->active_logs == 2)
159 need_cp = true;
161 return need_cp;
164 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
166 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
167 bool ret = false;
168 /* But we need to avoid that there are some inode updates */
169 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
170 ret = true;
171 f2fs_put_page(i, 0);
172 return ret;
175 static void try_to_fix_pino(struct inode *inode)
177 struct f2fs_inode_info *fi = F2FS_I(inode);
178 nid_t pino;
180 down_write(&fi->i_sem);
181 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
182 get_parent_ino(inode, &pino)) {
183 f2fs_i_pino_write(inode, pino);
184 file_got_pino(inode);
186 up_write(&fi->i_sem);
189 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
190 int datasync, bool atomic)
192 struct inode *inode = file->f_mapping->host;
193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
194 nid_t ino = inode->i_ino;
195 int ret = 0;
196 bool need_cp = false;
197 struct writeback_control wbc = {
198 .sync_mode = WB_SYNC_ALL,
199 .nr_to_write = LONG_MAX,
200 .for_reclaim = 0,
203 if (unlikely(f2fs_readonly(inode->i_sb)))
204 return 0;
206 trace_f2fs_sync_file_enter(inode);
208 /* if fdatasync is triggered, let's do in-place-update */
209 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
210 set_inode_flag(inode, FI_NEED_IPU);
211 ret = file_write_and_wait_range(file, start, end);
212 clear_inode_flag(inode, FI_NEED_IPU);
214 if (ret) {
215 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
216 return ret;
219 /* if the inode is dirty, let's recover all the time */
220 if (!f2fs_skip_inode_update(inode, datasync)) {
221 f2fs_write_inode(inode, NULL);
222 goto go_write;
226 * if there is no written data, don't waste time to write recovery info.
228 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
229 !exist_written_data(sbi, ino, APPEND_INO)) {
231 /* it may call write_inode just prior to fsync */
232 if (need_inode_page_update(sbi, ino))
233 goto go_write;
235 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
236 exist_written_data(sbi, ino, UPDATE_INO))
237 goto flush_out;
238 goto out;
240 go_write:
242 * Both of fdatasync() and fsync() are able to be recovered from
243 * sudden-power-off.
245 down_read(&F2FS_I(inode)->i_sem);
246 need_cp = need_do_checkpoint(inode);
247 up_read(&F2FS_I(inode)->i_sem);
249 if (need_cp) {
250 /* all the dirty node pages should be flushed for POR */
251 ret = f2fs_sync_fs(inode->i_sb, 1);
254 * We've secured consistency through sync_fs. Following pino
255 * will be used only for fsynced inodes after checkpoint.
257 try_to_fix_pino(inode);
258 clear_inode_flag(inode, FI_APPEND_WRITE);
259 clear_inode_flag(inode, FI_UPDATE_WRITE);
260 goto out;
262 sync_nodes:
263 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
264 if (ret)
265 goto out;
267 /* if cp_error was enabled, we should avoid infinite loop */
268 if (unlikely(f2fs_cp_error(sbi))) {
269 ret = -EIO;
270 goto out;
273 if (need_inode_block_update(sbi, ino)) {
274 f2fs_mark_inode_dirty_sync(inode, true);
275 f2fs_write_inode(inode, NULL);
276 goto sync_nodes;
280 * If it's atomic_write, it's just fine to keep write ordering. So
281 * here we don't need to wait for node write completion, since we use
282 * node chain which serializes node blocks. If one of node writes are
283 * reordered, we can see simply broken chain, resulting in stopping
284 * roll-forward recovery. It means we'll recover all or none node blocks
285 * given fsync mark.
287 if (!atomic) {
288 ret = wait_on_node_pages_writeback(sbi, ino);
289 if (ret)
290 goto out;
293 /* once recovery info is written, don't need to tack this */
294 remove_ino_entry(sbi, ino, APPEND_INO);
295 clear_inode_flag(inode, FI_APPEND_WRITE);
296 flush_out:
297 remove_ino_entry(sbi, ino, UPDATE_INO);
298 clear_inode_flag(inode, FI_UPDATE_WRITE);
299 if (!atomic)
300 ret = f2fs_issue_flush(sbi);
301 f2fs_update_time(sbi, REQ_TIME);
302 out:
303 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
304 f2fs_trace_ios(NULL, 1);
305 return ret;
308 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
310 return f2fs_do_sync_file(file, start, end, datasync, false);
313 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
314 pgoff_t pgofs, int whence)
316 struct pagevec pvec;
317 int nr_pages;
319 if (whence != SEEK_DATA)
320 return 0;
322 /* find first dirty page index */
323 pagevec_init(&pvec, 0);
324 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
325 PAGECACHE_TAG_DIRTY, 1);
326 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
327 pagevec_release(&pvec);
328 return pgofs;
331 static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
332 int whence)
334 switch (whence) {
335 case SEEK_DATA:
336 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
337 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
338 return true;
339 break;
340 case SEEK_HOLE:
341 if (blkaddr == NULL_ADDR)
342 return true;
343 break;
345 return false;
348 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
350 struct inode *inode = file->f_mapping->host;
351 loff_t maxbytes = inode->i_sb->s_maxbytes;
352 struct dnode_of_data dn;
353 pgoff_t pgofs, end_offset, dirty;
354 loff_t data_ofs = offset;
355 loff_t isize;
356 int err = 0;
358 inode_lock(inode);
360 isize = i_size_read(inode);
361 if (offset >= isize)
362 goto fail;
364 /* handle inline data case */
365 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
366 if (whence == SEEK_HOLE)
367 data_ofs = isize;
368 goto found;
371 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
373 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
375 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
376 set_new_dnode(&dn, inode, NULL, NULL, 0);
377 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
378 if (err && err != -ENOENT) {
379 goto fail;
380 } else if (err == -ENOENT) {
381 /* direct node does not exists */
382 if (whence == SEEK_DATA) {
383 pgofs = get_next_page_offset(&dn, pgofs);
384 continue;
385 } else {
386 goto found;
390 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
392 /* find data/hole in dnode block */
393 for (; dn.ofs_in_node < end_offset;
394 dn.ofs_in_node++, pgofs++,
395 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
396 block_t blkaddr;
397 blkaddr = datablock_addr(dn.inode,
398 dn.node_page, dn.ofs_in_node);
400 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
401 f2fs_put_dnode(&dn);
402 goto found;
405 f2fs_put_dnode(&dn);
408 if (whence == SEEK_DATA)
409 goto fail;
410 found:
411 if (whence == SEEK_HOLE && data_ofs > isize)
412 data_ofs = isize;
413 inode_unlock(inode);
414 return vfs_setpos(file, data_ofs, maxbytes);
415 fail:
416 inode_unlock(inode);
417 return -ENXIO;
420 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
422 struct inode *inode = file->f_mapping->host;
423 loff_t maxbytes = inode->i_sb->s_maxbytes;
425 switch (whence) {
426 case SEEK_SET:
427 case SEEK_CUR:
428 case SEEK_END:
429 return generic_file_llseek_size(file, offset, whence,
430 maxbytes, i_size_read(inode));
431 case SEEK_DATA:
432 case SEEK_HOLE:
433 if (offset < 0)
434 return -ENXIO;
435 return f2fs_seek_block(file, offset, whence);
438 return -EINVAL;
441 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
443 struct inode *inode = file_inode(file);
444 int err;
446 /* we don't need to use inline_data strictly */
447 err = f2fs_convert_inline_inode(inode);
448 if (err)
449 return err;
451 file_accessed(file);
452 vma->vm_ops = &f2fs_file_vm_ops;
453 return 0;
456 static int f2fs_file_open(struct inode *inode, struct file *filp)
458 struct dentry *dir;
460 if (f2fs_encrypted_inode(inode)) {
461 int ret = fscrypt_get_encryption_info(inode);
462 if (ret)
463 return -EACCES;
464 if (!fscrypt_has_encryption_key(inode))
465 return -ENOKEY;
467 dir = dget_parent(file_dentry(filp));
468 if (f2fs_encrypted_inode(d_inode(dir)) &&
469 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
470 dput(dir);
471 return -EPERM;
473 dput(dir);
474 return dquot_file_open(inode, filp);
477 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
479 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
480 struct f2fs_node *raw_node;
481 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
482 __le32 *addr;
483 int base = 0;
485 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
486 base = get_extra_isize(dn->inode);
488 raw_node = F2FS_NODE(dn->node_page);
489 addr = blkaddr_in_node(raw_node) + base + ofs;
491 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
492 block_t blkaddr = le32_to_cpu(*addr);
493 if (blkaddr == NULL_ADDR)
494 continue;
496 dn->data_blkaddr = NULL_ADDR;
497 set_data_blkaddr(dn);
498 invalidate_blocks(sbi, blkaddr);
499 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
500 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
501 nr_free++;
504 if (nr_free) {
505 pgoff_t fofs;
507 * once we invalidate valid blkaddr in range [ofs, ofs + count],
508 * we will invalidate all blkaddr in the whole range.
510 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
511 dn->inode) + ofs;
512 f2fs_update_extent_cache_range(dn, fofs, 0, len);
513 dec_valid_block_count(sbi, dn->inode, nr_free);
515 dn->ofs_in_node = ofs;
517 f2fs_update_time(sbi, REQ_TIME);
518 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
519 dn->ofs_in_node, nr_free);
520 return nr_free;
523 void truncate_data_blocks(struct dnode_of_data *dn)
525 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
528 static int truncate_partial_data_page(struct inode *inode, u64 from,
529 bool cache_only)
531 unsigned offset = from & (PAGE_SIZE - 1);
532 pgoff_t index = from >> PAGE_SHIFT;
533 struct address_space *mapping = inode->i_mapping;
534 struct page *page;
536 if (!offset && !cache_only)
537 return 0;
539 if (cache_only) {
540 page = find_lock_page(mapping, index);
541 if (page && PageUptodate(page))
542 goto truncate_out;
543 f2fs_put_page(page, 1);
544 return 0;
547 page = get_lock_data_page(inode, index, true);
548 if (IS_ERR(page))
549 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
550 truncate_out:
551 f2fs_wait_on_page_writeback(page, DATA, true);
552 zero_user(page, offset, PAGE_SIZE - offset);
554 /* An encrypted inode should have a key and truncate the last page. */
555 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
556 if (!cache_only)
557 set_page_dirty(page);
558 f2fs_put_page(page, 1);
559 return 0;
562 int truncate_blocks(struct inode *inode, u64 from, bool lock)
564 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
565 unsigned int blocksize = inode->i_sb->s_blocksize;
566 struct dnode_of_data dn;
567 pgoff_t free_from;
568 int count = 0, err = 0;
569 struct page *ipage;
570 bool truncate_page = false;
572 trace_f2fs_truncate_blocks_enter(inode, from);
574 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
576 if (free_from >= sbi->max_file_blocks)
577 goto free_partial;
579 if (lock)
580 f2fs_lock_op(sbi);
582 ipage = get_node_page(sbi, inode->i_ino);
583 if (IS_ERR(ipage)) {
584 err = PTR_ERR(ipage);
585 goto out;
588 if (f2fs_has_inline_data(inode)) {
589 truncate_inline_inode(inode, ipage, from);
590 f2fs_put_page(ipage, 1);
591 truncate_page = true;
592 goto out;
595 set_new_dnode(&dn, inode, ipage, NULL, 0);
596 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
597 if (err) {
598 if (err == -ENOENT)
599 goto free_next;
600 goto out;
603 count = ADDRS_PER_PAGE(dn.node_page, inode);
605 count -= dn.ofs_in_node;
606 f2fs_bug_on(sbi, count < 0);
608 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
609 truncate_data_blocks_range(&dn, count);
610 free_from += count;
613 f2fs_put_dnode(&dn);
614 free_next:
615 err = truncate_inode_blocks(inode, free_from);
616 out:
617 if (lock)
618 f2fs_unlock_op(sbi);
619 free_partial:
620 /* lastly zero out the first data page */
621 if (!err)
622 err = truncate_partial_data_page(inode, from, truncate_page);
624 trace_f2fs_truncate_blocks_exit(inode, err);
625 return err;
628 int f2fs_truncate(struct inode *inode)
630 int err;
632 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
633 S_ISLNK(inode->i_mode)))
634 return 0;
636 trace_f2fs_truncate(inode);
638 #ifdef CONFIG_F2FS_FAULT_INJECTION
639 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
640 f2fs_show_injection_info(FAULT_TRUNCATE);
641 return -EIO;
643 #endif
644 /* we should check inline_data size */
645 if (!f2fs_may_inline_data(inode)) {
646 err = f2fs_convert_inline_inode(inode);
647 if (err)
648 return err;
651 err = truncate_blocks(inode, i_size_read(inode), true);
652 if (err)
653 return err;
655 inode->i_mtime = inode->i_ctime = current_time(inode);
656 f2fs_mark_inode_dirty_sync(inode, false);
657 return 0;
660 int f2fs_getattr(const struct path *path, struct kstat *stat,
661 u32 request_mask, unsigned int query_flags)
663 struct inode *inode = d_inode(path->dentry);
664 struct f2fs_inode_info *fi = F2FS_I(inode);
665 unsigned int flags;
667 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
668 if (flags & FS_APPEND_FL)
669 stat->attributes |= STATX_ATTR_APPEND;
670 if (flags & FS_COMPR_FL)
671 stat->attributes |= STATX_ATTR_COMPRESSED;
672 if (f2fs_encrypted_inode(inode))
673 stat->attributes |= STATX_ATTR_ENCRYPTED;
674 if (flags & FS_IMMUTABLE_FL)
675 stat->attributes |= STATX_ATTR_IMMUTABLE;
676 if (flags & FS_NODUMP_FL)
677 stat->attributes |= STATX_ATTR_NODUMP;
679 stat->attributes_mask |= (STATX_ATTR_APPEND |
680 STATX_ATTR_COMPRESSED |
681 STATX_ATTR_ENCRYPTED |
682 STATX_ATTR_IMMUTABLE |
683 STATX_ATTR_NODUMP);
685 generic_fillattr(inode, stat);
686 return 0;
689 #ifdef CONFIG_F2FS_FS_POSIX_ACL
690 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
692 unsigned int ia_valid = attr->ia_valid;
694 if (ia_valid & ATTR_UID)
695 inode->i_uid = attr->ia_uid;
696 if (ia_valid & ATTR_GID)
697 inode->i_gid = attr->ia_gid;
698 if (ia_valid & ATTR_ATIME)
699 inode->i_atime = timespec_trunc(attr->ia_atime,
700 inode->i_sb->s_time_gran);
701 if (ia_valid & ATTR_MTIME)
702 inode->i_mtime = timespec_trunc(attr->ia_mtime,
703 inode->i_sb->s_time_gran);
704 if (ia_valid & ATTR_CTIME)
705 inode->i_ctime = timespec_trunc(attr->ia_ctime,
706 inode->i_sb->s_time_gran);
707 if (ia_valid & ATTR_MODE) {
708 umode_t mode = attr->ia_mode;
710 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
711 mode &= ~S_ISGID;
712 set_acl_inode(inode, mode);
715 #else
716 #define __setattr_copy setattr_copy
717 #endif
719 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
721 struct inode *inode = d_inode(dentry);
722 int err;
723 bool size_changed = false;
725 err = setattr_prepare(dentry, attr);
726 if (err)
727 return err;
729 if (is_quota_modification(inode, attr)) {
730 err = dquot_initialize(inode);
731 if (err)
732 return err;
734 if ((attr->ia_valid & ATTR_UID &&
735 !uid_eq(attr->ia_uid, inode->i_uid)) ||
736 (attr->ia_valid & ATTR_GID &&
737 !gid_eq(attr->ia_gid, inode->i_gid))) {
738 err = dquot_transfer(inode, attr);
739 if (err)
740 return err;
743 if (attr->ia_valid & ATTR_SIZE) {
744 if (f2fs_encrypted_inode(inode)) {
745 err = fscrypt_get_encryption_info(inode);
746 if (err)
747 return err;
748 if (!fscrypt_has_encryption_key(inode))
749 return -ENOKEY;
752 if (attr->ia_size <= i_size_read(inode)) {
753 down_write(&F2FS_I(inode)->i_mmap_sem);
754 truncate_setsize(inode, attr->ia_size);
755 err = f2fs_truncate(inode);
756 up_write(&F2FS_I(inode)->i_mmap_sem);
757 if (err)
758 return err;
759 } else {
761 * do not trim all blocks after i_size if target size is
762 * larger than i_size.
764 down_write(&F2FS_I(inode)->i_mmap_sem);
765 truncate_setsize(inode, attr->ia_size);
766 up_write(&F2FS_I(inode)->i_mmap_sem);
768 /* should convert inline inode here */
769 if (!f2fs_may_inline_data(inode)) {
770 err = f2fs_convert_inline_inode(inode);
771 if (err)
772 return err;
774 inode->i_mtime = inode->i_ctime = current_time(inode);
777 size_changed = true;
780 __setattr_copy(inode, attr);
782 if (attr->ia_valid & ATTR_MODE) {
783 err = posix_acl_chmod(inode, get_inode_mode(inode));
784 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
785 inode->i_mode = F2FS_I(inode)->i_acl_mode;
786 clear_inode_flag(inode, FI_ACL_MODE);
790 /* file size may changed here */
791 f2fs_mark_inode_dirty_sync(inode, size_changed);
793 /* inode change will produce dirty node pages flushed by checkpoint */
794 f2fs_balance_fs(F2FS_I_SB(inode), true);
796 return err;
799 const struct inode_operations f2fs_file_inode_operations = {
800 .getattr = f2fs_getattr,
801 .setattr = f2fs_setattr,
802 .get_acl = f2fs_get_acl,
803 .set_acl = f2fs_set_acl,
804 #ifdef CONFIG_F2FS_FS_XATTR
805 .listxattr = f2fs_listxattr,
806 #endif
807 .fiemap = f2fs_fiemap,
810 static int fill_zero(struct inode *inode, pgoff_t index,
811 loff_t start, loff_t len)
813 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
814 struct page *page;
816 if (!len)
817 return 0;
819 f2fs_balance_fs(sbi, true);
821 f2fs_lock_op(sbi);
822 page = get_new_data_page(inode, NULL, index, false);
823 f2fs_unlock_op(sbi);
825 if (IS_ERR(page))
826 return PTR_ERR(page);
828 f2fs_wait_on_page_writeback(page, DATA, true);
829 zero_user(page, start, len);
830 set_page_dirty(page);
831 f2fs_put_page(page, 1);
832 return 0;
835 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
837 int err;
839 while (pg_start < pg_end) {
840 struct dnode_of_data dn;
841 pgoff_t end_offset, count;
843 set_new_dnode(&dn, inode, NULL, NULL, 0);
844 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
845 if (err) {
846 if (err == -ENOENT) {
847 pg_start++;
848 continue;
850 return err;
853 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
854 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
856 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
858 truncate_data_blocks_range(&dn, count);
859 f2fs_put_dnode(&dn);
861 pg_start += count;
863 return 0;
866 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
868 pgoff_t pg_start, pg_end;
869 loff_t off_start, off_end;
870 int ret;
872 ret = f2fs_convert_inline_inode(inode);
873 if (ret)
874 return ret;
876 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
877 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
879 off_start = offset & (PAGE_SIZE - 1);
880 off_end = (offset + len) & (PAGE_SIZE - 1);
882 if (pg_start == pg_end) {
883 ret = fill_zero(inode, pg_start, off_start,
884 off_end - off_start);
885 if (ret)
886 return ret;
887 } else {
888 if (off_start) {
889 ret = fill_zero(inode, pg_start++, off_start,
890 PAGE_SIZE - off_start);
891 if (ret)
892 return ret;
894 if (off_end) {
895 ret = fill_zero(inode, pg_end, 0, off_end);
896 if (ret)
897 return ret;
900 if (pg_start < pg_end) {
901 struct address_space *mapping = inode->i_mapping;
902 loff_t blk_start, blk_end;
903 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
905 f2fs_balance_fs(sbi, true);
907 blk_start = (loff_t)pg_start << PAGE_SHIFT;
908 blk_end = (loff_t)pg_end << PAGE_SHIFT;
909 down_write(&F2FS_I(inode)->i_mmap_sem);
910 truncate_inode_pages_range(mapping, blk_start,
911 blk_end - 1);
913 f2fs_lock_op(sbi);
914 ret = truncate_hole(inode, pg_start, pg_end);
915 f2fs_unlock_op(sbi);
916 up_write(&F2FS_I(inode)->i_mmap_sem);
920 return ret;
923 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
924 int *do_replace, pgoff_t off, pgoff_t len)
926 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
927 struct dnode_of_data dn;
928 int ret, done, i;
930 next_dnode:
931 set_new_dnode(&dn, inode, NULL, NULL, 0);
932 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
933 if (ret && ret != -ENOENT) {
934 return ret;
935 } else if (ret == -ENOENT) {
936 if (dn.max_level == 0)
937 return -ENOENT;
938 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
939 blkaddr += done;
940 do_replace += done;
941 goto next;
944 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
945 dn.ofs_in_node, len);
946 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
947 *blkaddr = datablock_addr(dn.inode,
948 dn.node_page, dn.ofs_in_node);
949 if (!is_checkpointed_data(sbi, *blkaddr)) {
951 if (test_opt(sbi, LFS)) {
952 f2fs_put_dnode(&dn);
953 return -ENOTSUPP;
956 /* do not invalidate this block address */
957 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
958 *do_replace = 1;
961 f2fs_put_dnode(&dn);
962 next:
963 len -= done;
964 off += done;
965 if (len)
966 goto next_dnode;
967 return 0;
970 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
971 int *do_replace, pgoff_t off, int len)
973 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
974 struct dnode_of_data dn;
975 int ret, i;
977 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
978 if (*do_replace == 0)
979 continue;
981 set_new_dnode(&dn, inode, NULL, NULL, 0);
982 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
983 if (ret) {
984 dec_valid_block_count(sbi, inode, 1);
985 invalidate_blocks(sbi, *blkaddr);
986 } else {
987 f2fs_update_data_blkaddr(&dn, *blkaddr);
989 f2fs_put_dnode(&dn);
991 return 0;
994 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
995 block_t *blkaddr, int *do_replace,
996 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
998 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
999 pgoff_t i = 0;
1000 int ret;
1002 while (i < len) {
1003 if (blkaddr[i] == NULL_ADDR && !full) {
1004 i++;
1005 continue;
1008 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1009 struct dnode_of_data dn;
1010 struct node_info ni;
1011 size_t new_size;
1012 pgoff_t ilen;
1014 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1015 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1016 if (ret)
1017 return ret;
1019 get_node_info(sbi, dn.nid, &ni);
1020 ilen = min((pgoff_t)
1021 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1022 dn.ofs_in_node, len - i);
1023 do {
1024 dn.data_blkaddr = datablock_addr(dn.inode,
1025 dn.node_page, dn.ofs_in_node);
1026 truncate_data_blocks_range(&dn, 1);
1028 if (do_replace[i]) {
1029 f2fs_i_blocks_write(src_inode,
1030 1, false, false);
1031 f2fs_i_blocks_write(dst_inode,
1032 1, true, false);
1033 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1034 blkaddr[i], ni.version, true, false);
1036 do_replace[i] = 0;
1038 dn.ofs_in_node++;
1039 i++;
1040 new_size = (dst + i) << PAGE_SHIFT;
1041 if (dst_inode->i_size < new_size)
1042 f2fs_i_size_write(dst_inode, new_size);
1043 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1045 f2fs_put_dnode(&dn);
1046 } else {
1047 struct page *psrc, *pdst;
1049 psrc = get_lock_data_page(src_inode, src + i, true);
1050 if (IS_ERR(psrc))
1051 return PTR_ERR(psrc);
1052 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1053 true);
1054 if (IS_ERR(pdst)) {
1055 f2fs_put_page(psrc, 1);
1056 return PTR_ERR(pdst);
1058 f2fs_copy_page(psrc, pdst);
1059 set_page_dirty(pdst);
1060 f2fs_put_page(pdst, 1);
1061 f2fs_put_page(psrc, 1);
1063 ret = truncate_hole(src_inode, src + i, src + i + 1);
1064 if (ret)
1065 return ret;
1066 i++;
1069 return 0;
1072 static int __exchange_data_block(struct inode *src_inode,
1073 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1074 pgoff_t len, bool full)
1076 block_t *src_blkaddr;
1077 int *do_replace;
1078 pgoff_t olen;
1079 int ret;
1081 while (len) {
1082 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1084 src_blkaddr = kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
1085 if (!src_blkaddr)
1086 return -ENOMEM;
1088 do_replace = kvzalloc(sizeof(int) * olen, GFP_KERNEL);
1089 if (!do_replace) {
1090 kvfree(src_blkaddr);
1091 return -ENOMEM;
1094 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1095 do_replace, src, olen);
1096 if (ret)
1097 goto roll_back;
1099 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1100 do_replace, src, dst, olen, full);
1101 if (ret)
1102 goto roll_back;
1104 src += olen;
1105 dst += olen;
1106 len -= olen;
1108 kvfree(src_blkaddr);
1109 kvfree(do_replace);
1111 return 0;
1113 roll_back:
1114 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1115 kvfree(src_blkaddr);
1116 kvfree(do_replace);
1117 return ret;
1120 static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1122 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1123 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1124 int ret;
1126 f2fs_balance_fs(sbi, true);
1127 f2fs_lock_op(sbi);
1129 f2fs_drop_extent_tree(inode);
1131 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1132 f2fs_unlock_op(sbi);
1133 return ret;
1136 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1138 pgoff_t pg_start, pg_end;
1139 loff_t new_size;
1140 int ret;
1142 if (offset + len >= i_size_read(inode))
1143 return -EINVAL;
1145 /* collapse range should be aligned to block size of f2fs. */
1146 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1147 return -EINVAL;
1149 ret = f2fs_convert_inline_inode(inode);
1150 if (ret)
1151 return ret;
1153 pg_start = offset >> PAGE_SHIFT;
1154 pg_end = (offset + len) >> PAGE_SHIFT;
1156 down_write(&F2FS_I(inode)->i_mmap_sem);
1157 /* write out all dirty pages from offset */
1158 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1159 if (ret)
1160 goto out;
1162 truncate_pagecache(inode, offset);
1164 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1165 if (ret)
1166 goto out;
1168 /* write out all moved pages, if possible */
1169 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1170 truncate_pagecache(inode, offset);
1172 new_size = i_size_read(inode) - len;
1173 truncate_pagecache(inode, new_size);
1175 ret = truncate_blocks(inode, new_size, true);
1176 if (!ret)
1177 f2fs_i_size_write(inode, new_size);
1179 out:
1180 up_write(&F2FS_I(inode)->i_mmap_sem);
1181 return ret;
1184 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1185 pgoff_t end)
1187 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1188 pgoff_t index = start;
1189 unsigned int ofs_in_node = dn->ofs_in_node;
1190 blkcnt_t count = 0;
1191 int ret;
1193 for (; index < end; index++, dn->ofs_in_node++) {
1194 if (datablock_addr(dn->inode, dn->node_page,
1195 dn->ofs_in_node) == NULL_ADDR)
1196 count++;
1199 dn->ofs_in_node = ofs_in_node;
1200 ret = reserve_new_blocks(dn, count);
1201 if (ret)
1202 return ret;
1204 dn->ofs_in_node = ofs_in_node;
1205 for (index = start; index < end; index++, dn->ofs_in_node++) {
1206 dn->data_blkaddr = datablock_addr(dn->inode,
1207 dn->node_page, dn->ofs_in_node);
1209 * reserve_new_blocks will not guarantee entire block
1210 * allocation.
1212 if (dn->data_blkaddr == NULL_ADDR) {
1213 ret = -ENOSPC;
1214 break;
1216 if (dn->data_blkaddr != NEW_ADDR) {
1217 invalidate_blocks(sbi, dn->data_blkaddr);
1218 dn->data_blkaddr = NEW_ADDR;
1219 set_data_blkaddr(dn);
1223 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1225 return ret;
1228 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1229 int mode)
1231 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1232 struct address_space *mapping = inode->i_mapping;
1233 pgoff_t index, pg_start, pg_end;
1234 loff_t new_size = i_size_read(inode);
1235 loff_t off_start, off_end;
1236 int ret = 0;
1238 ret = inode_newsize_ok(inode, (len + offset));
1239 if (ret)
1240 return ret;
1242 ret = f2fs_convert_inline_inode(inode);
1243 if (ret)
1244 return ret;
1246 down_write(&F2FS_I(inode)->i_mmap_sem);
1247 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1248 if (ret)
1249 goto out_sem;
1251 truncate_pagecache_range(inode, offset, offset + len - 1);
1253 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1254 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1256 off_start = offset & (PAGE_SIZE - 1);
1257 off_end = (offset + len) & (PAGE_SIZE - 1);
1259 if (pg_start == pg_end) {
1260 ret = fill_zero(inode, pg_start, off_start,
1261 off_end - off_start);
1262 if (ret)
1263 goto out_sem;
1265 new_size = max_t(loff_t, new_size, offset + len);
1266 } else {
1267 if (off_start) {
1268 ret = fill_zero(inode, pg_start++, off_start,
1269 PAGE_SIZE - off_start);
1270 if (ret)
1271 goto out_sem;
1273 new_size = max_t(loff_t, new_size,
1274 (loff_t)pg_start << PAGE_SHIFT);
1277 for (index = pg_start; index < pg_end;) {
1278 struct dnode_of_data dn;
1279 unsigned int end_offset;
1280 pgoff_t end;
1282 f2fs_lock_op(sbi);
1284 set_new_dnode(&dn, inode, NULL, NULL, 0);
1285 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
1286 if (ret) {
1287 f2fs_unlock_op(sbi);
1288 goto out;
1291 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1292 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1294 ret = f2fs_do_zero_range(&dn, index, end);
1295 f2fs_put_dnode(&dn);
1296 f2fs_unlock_op(sbi);
1298 f2fs_balance_fs(sbi, dn.node_changed);
1300 if (ret)
1301 goto out;
1303 index = end;
1304 new_size = max_t(loff_t, new_size,
1305 (loff_t)index << PAGE_SHIFT);
1308 if (off_end) {
1309 ret = fill_zero(inode, pg_end, 0, off_end);
1310 if (ret)
1311 goto out;
1313 new_size = max_t(loff_t, new_size, offset + len);
1317 out:
1318 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1319 f2fs_i_size_write(inode, new_size);
1320 out_sem:
1321 up_write(&F2FS_I(inode)->i_mmap_sem);
1323 return ret;
1326 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1328 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1329 pgoff_t nr, pg_start, pg_end, delta, idx;
1330 loff_t new_size;
1331 int ret = 0;
1333 new_size = i_size_read(inode) + len;
1334 ret = inode_newsize_ok(inode, new_size);
1335 if (ret)
1336 return ret;
1338 if (offset >= i_size_read(inode))
1339 return -EINVAL;
1341 /* insert range should be aligned to block size of f2fs. */
1342 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1343 return -EINVAL;
1345 ret = f2fs_convert_inline_inode(inode);
1346 if (ret)
1347 return ret;
1349 f2fs_balance_fs(sbi, true);
1351 down_write(&F2FS_I(inode)->i_mmap_sem);
1352 ret = truncate_blocks(inode, i_size_read(inode), true);
1353 if (ret)
1354 goto out;
1356 /* write out all dirty pages from offset */
1357 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1358 if (ret)
1359 goto out;
1361 truncate_pagecache(inode, offset);
1363 pg_start = offset >> PAGE_SHIFT;
1364 pg_end = (offset + len) >> PAGE_SHIFT;
1365 delta = pg_end - pg_start;
1366 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1368 while (!ret && idx > pg_start) {
1369 nr = idx - pg_start;
1370 if (nr > delta)
1371 nr = delta;
1372 idx -= nr;
1374 f2fs_lock_op(sbi);
1375 f2fs_drop_extent_tree(inode);
1377 ret = __exchange_data_block(inode, inode, idx,
1378 idx + delta, nr, false);
1379 f2fs_unlock_op(sbi);
1382 /* write out all moved pages, if possible */
1383 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1384 truncate_pagecache(inode, offset);
1386 if (!ret)
1387 f2fs_i_size_write(inode, new_size);
1388 out:
1389 up_write(&F2FS_I(inode)->i_mmap_sem);
1390 return ret;
1393 static int expand_inode_data(struct inode *inode, loff_t offset,
1394 loff_t len, int mode)
1396 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1397 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1398 pgoff_t pg_end;
1399 loff_t new_size = i_size_read(inode);
1400 loff_t off_end;
1401 int err;
1403 err = inode_newsize_ok(inode, (len + offset));
1404 if (err)
1405 return err;
1407 err = f2fs_convert_inline_inode(inode);
1408 if (err)
1409 return err;
1411 f2fs_balance_fs(sbi, true);
1413 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1414 off_end = (offset + len) & (PAGE_SIZE - 1);
1416 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1417 map.m_len = pg_end - map.m_lblk;
1418 if (off_end)
1419 map.m_len++;
1421 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1422 if (err) {
1423 pgoff_t last_off;
1425 if (!map.m_len)
1426 return err;
1428 last_off = map.m_lblk + map.m_len - 1;
1430 /* update new size to the failed position */
1431 new_size = (last_off == pg_end) ? offset + len:
1432 (loff_t)(last_off + 1) << PAGE_SHIFT;
1433 } else {
1434 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1437 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
1438 f2fs_i_size_write(inode, new_size);
1440 return err;
1443 static long f2fs_fallocate(struct file *file, int mode,
1444 loff_t offset, loff_t len)
1446 struct inode *inode = file_inode(file);
1447 long ret = 0;
1449 /* f2fs only support ->fallocate for regular file */
1450 if (!S_ISREG(inode->i_mode))
1451 return -EINVAL;
1453 if (f2fs_encrypted_inode(inode) &&
1454 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1455 return -EOPNOTSUPP;
1457 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1458 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1459 FALLOC_FL_INSERT_RANGE))
1460 return -EOPNOTSUPP;
1462 inode_lock(inode);
1464 if (mode & FALLOC_FL_PUNCH_HOLE) {
1465 if (offset >= inode->i_size)
1466 goto out;
1468 ret = punch_hole(inode, offset, len);
1469 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1470 ret = f2fs_collapse_range(inode, offset, len);
1471 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1472 ret = f2fs_zero_range(inode, offset, len, mode);
1473 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1474 ret = f2fs_insert_range(inode, offset, len);
1475 } else {
1476 ret = expand_inode_data(inode, offset, len, mode);
1479 if (!ret) {
1480 inode->i_mtime = inode->i_ctime = current_time(inode);
1481 f2fs_mark_inode_dirty_sync(inode, false);
1482 if (mode & FALLOC_FL_KEEP_SIZE)
1483 file_set_keep_isize(inode);
1484 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1487 out:
1488 inode_unlock(inode);
1490 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1491 return ret;
1494 static int f2fs_release_file(struct inode *inode, struct file *filp)
1497 * f2fs_relase_file is called at every close calls. So we should
1498 * not drop any inmemory pages by close called by other process.
1500 if (!(filp->f_mode & FMODE_WRITE) ||
1501 atomic_read(&inode->i_writecount) != 1)
1502 return 0;
1504 /* some remained atomic pages should discarded */
1505 if (f2fs_is_atomic_file(inode))
1506 drop_inmem_pages(inode);
1507 if (f2fs_is_volatile_file(inode)) {
1508 clear_inode_flag(inode, FI_VOLATILE_FILE);
1509 stat_dec_volatile_write(inode);
1510 set_inode_flag(inode, FI_DROP_CACHE);
1511 filemap_fdatawrite(inode->i_mapping);
1512 clear_inode_flag(inode, FI_DROP_CACHE);
1514 return 0;
1517 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1519 struct inode *inode = file_inode(file);
1522 * If the process doing a transaction is crashed, we should do
1523 * roll-back. Otherwise, other reader/write can see corrupted database
1524 * until all the writers close its file. Since this should be done
1525 * before dropping file lock, it needs to do in ->flush.
1527 if (f2fs_is_atomic_file(inode) &&
1528 F2FS_I(inode)->inmem_task == current)
1529 drop_inmem_pages(inode);
1530 return 0;
1533 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1535 struct inode *inode = file_inode(filp);
1536 struct f2fs_inode_info *fi = F2FS_I(inode);
1537 unsigned int flags = fi->i_flags &
1538 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1539 return put_user(flags, (int __user *)arg);
1542 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1544 struct f2fs_inode_info *fi = F2FS_I(inode);
1545 unsigned int oldflags;
1547 /* Is it quota file? Do not allow user to mess with it */
1548 if (IS_NOQUOTA(inode))
1549 return -EPERM;
1551 flags = f2fs_mask_flags(inode->i_mode, flags);
1553 oldflags = fi->i_flags;
1555 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1556 if (!capable(CAP_LINUX_IMMUTABLE))
1557 return -EPERM;
1559 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1560 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1561 fi->i_flags = flags;
1563 if (fi->i_flags & FS_PROJINHERIT_FL)
1564 set_inode_flag(inode, FI_PROJ_INHERIT);
1565 else
1566 clear_inode_flag(inode, FI_PROJ_INHERIT);
1568 inode->i_ctime = current_time(inode);
1569 f2fs_set_inode_flags(inode);
1570 f2fs_mark_inode_dirty_sync(inode, false);
1571 return 0;
1574 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1576 struct inode *inode = file_inode(filp);
1577 unsigned int flags;
1578 int ret;
1580 if (!inode_owner_or_capable(inode))
1581 return -EACCES;
1583 if (get_user(flags, (int __user *)arg))
1584 return -EFAULT;
1586 ret = mnt_want_write_file(filp);
1587 if (ret)
1588 return ret;
1590 inode_lock(inode);
1592 ret = __f2fs_ioc_setflags(inode, flags);
1594 inode_unlock(inode);
1595 mnt_drop_write_file(filp);
1596 return ret;
1599 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1601 struct inode *inode = file_inode(filp);
1603 return put_user(inode->i_generation, (int __user *)arg);
1606 static int f2fs_ioc_start_atomic_write(struct file *filp)
1608 struct inode *inode = file_inode(filp);
1609 int ret;
1611 if (!inode_owner_or_capable(inode))
1612 return -EACCES;
1614 if (!S_ISREG(inode->i_mode))
1615 return -EINVAL;
1617 ret = mnt_want_write_file(filp);
1618 if (ret)
1619 return ret;
1621 inode_lock(inode);
1623 if (f2fs_is_atomic_file(inode))
1624 goto out;
1626 ret = f2fs_convert_inline_inode(inode);
1627 if (ret)
1628 goto out;
1630 set_inode_flag(inode, FI_ATOMIC_FILE);
1631 set_inode_flag(inode, FI_HOT_DATA);
1632 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1634 if (!get_dirty_pages(inode))
1635 goto inc_stat;
1637 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1638 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1639 inode->i_ino, get_dirty_pages(inode));
1640 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1641 if (ret) {
1642 clear_inode_flag(inode, FI_ATOMIC_FILE);
1643 clear_inode_flag(inode, FI_HOT_DATA);
1644 goto out;
1647 inc_stat:
1648 F2FS_I(inode)->inmem_task = current;
1649 stat_inc_atomic_write(inode);
1650 stat_update_max_atomic_write(inode);
1651 out:
1652 inode_unlock(inode);
1653 mnt_drop_write_file(filp);
1654 return ret;
1657 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1659 struct inode *inode = file_inode(filp);
1660 int ret;
1662 if (!inode_owner_or_capable(inode))
1663 return -EACCES;
1665 ret = mnt_want_write_file(filp);
1666 if (ret)
1667 return ret;
1669 inode_lock(inode);
1671 if (f2fs_is_volatile_file(inode))
1672 goto err_out;
1674 if (f2fs_is_atomic_file(inode)) {
1675 ret = commit_inmem_pages(inode);
1676 if (ret)
1677 goto err_out;
1679 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1680 if (!ret) {
1681 clear_inode_flag(inode, FI_ATOMIC_FILE);
1682 clear_inode_flag(inode, FI_HOT_DATA);
1683 stat_dec_atomic_write(inode);
1685 } else {
1686 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1688 err_out:
1689 inode_unlock(inode);
1690 mnt_drop_write_file(filp);
1691 return ret;
1694 static int f2fs_ioc_start_volatile_write(struct file *filp)
1696 struct inode *inode = file_inode(filp);
1697 int ret;
1699 if (!inode_owner_or_capable(inode))
1700 return -EACCES;
1702 if (!S_ISREG(inode->i_mode))
1703 return -EINVAL;
1705 ret = mnt_want_write_file(filp);
1706 if (ret)
1707 return ret;
1709 inode_lock(inode);
1711 if (f2fs_is_volatile_file(inode))
1712 goto out;
1714 ret = f2fs_convert_inline_inode(inode);
1715 if (ret)
1716 goto out;
1718 stat_inc_volatile_write(inode);
1719 stat_update_max_volatile_write(inode);
1721 set_inode_flag(inode, FI_VOLATILE_FILE);
1722 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1723 out:
1724 inode_unlock(inode);
1725 mnt_drop_write_file(filp);
1726 return ret;
1729 static int f2fs_ioc_release_volatile_write(struct file *filp)
1731 struct inode *inode = file_inode(filp);
1732 int ret;
1734 if (!inode_owner_or_capable(inode))
1735 return -EACCES;
1737 ret = mnt_want_write_file(filp);
1738 if (ret)
1739 return ret;
1741 inode_lock(inode);
1743 if (!f2fs_is_volatile_file(inode))
1744 goto out;
1746 if (!f2fs_is_first_block_written(inode)) {
1747 ret = truncate_partial_data_page(inode, 0, true);
1748 goto out;
1751 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1752 out:
1753 inode_unlock(inode);
1754 mnt_drop_write_file(filp);
1755 return ret;
1758 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1760 struct inode *inode = file_inode(filp);
1761 int ret;
1763 if (!inode_owner_or_capable(inode))
1764 return -EACCES;
1766 ret = mnt_want_write_file(filp);
1767 if (ret)
1768 return ret;
1770 inode_lock(inode);
1772 if (f2fs_is_atomic_file(inode))
1773 drop_inmem_pages(inode);
1774 if (f2fs_is_volatile_file(inode)) {
1775 clear_inode_flag(inode, FI_VOLATILE_FILE);
1776 stat_dec_volatile_write(inode);
1777 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1780 inode_unlock(inode);
1782 mnt_drop_write_file(filp);
1783 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1784 return ret;
1787 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1789 struct inode *inode = file_inode(filp);
1790 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1791 struct super_block *sb = sbi->sb;
1792 __u32 in;
1793 int ret;
1795 if (!capable(CAP_SYS_ADMIN))
1796 return -EPERM;
1798 if (get_user(in, (__u32 __user *)arg))
1799 return -EFAULT;
1801 ret = mnt_want_write_file(filp);
1802 if (ret)
1803 return ret;
1805 switch (in) {
1806 case F2FS_GOING_DOWN_FULLSYNC:
1807 sb = freeze_bdev(sb->s_bdev);
1808 if (sb && !IS_ERR(sb)) {
1809 f2fs_stop_checkpoint(sbi, false);
1810 thaw_bdev(sb->s_bdev, sb);
1812 break;
1813 case F2FS_GOING_DOWN_METASYNC:
1814 /* do checkpoint only */
1815 f2fs_sync_fs(sb, 1);
1816 f2fs_stop_checkpoint(sbi, false);
1817 break;
1818 case F2FS_GOING_DOWN_NOSYNC:
1819 f2fs_stop_checkpoint(sbi, false);
1820 break;
1821 case F2FS_GOING_DOWN_METAFLUSH:
1822 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1823 f2fs_stop_checkpoint(sbi, false);
1824 break;
1825 default:
1826 ret = -EINVAL;
1827 goto out;
1829 f2fs_update_time(sbi, REQ_TIME);
1830 out:
1831 mnt_drop_write_file(filp);
1832 return ret;
1835 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1837 struct inode *inode = file_inode(filp);
1838 struct super_block *sb = inode->i_sb;
1839 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1840 struct fstrim_range range;
1841 int ret;
1843 if (!capable(CAP_SYS_ADMIN))
1844 return -EPERM;
1846 if (!blk_queue_discard(q))
1847 return -EOPNOTSUPP;
1849 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1850 sizeof(range)))
1851 return -EFAULT;
1853 ret = mnt_want_write_file(filp);
1854 if (ret)
1855 return ret;
1857 range.minlen = max((unsigned int)range.minlen,
1858 q->limits.discard_granularity);
1859 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1860 mnt_drop_write_file(filp);
1861 if (ret < 0)
1862 return ret;
1864 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1865 sizeof(range)))
1866 return -EFAULT;
1867 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1868 return 0;
1871 static bool uuid_is_nonzero(__u8 u[16])
1873 int i;
1875 for (i = 0; i < 16; i++)
1876 if (u[i])
1877 return true;
1878 return false;
1881 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1883 struct inode *inode = file_inode(filp);
1885 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1887 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1890 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1892 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1895 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1897 struct inode *inode = file_inode(filp);
1898 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1899 int err;
1901 if (!f2fs_sb_has_crypto(inode->i_sb))
1902 return -EOPNOTSUPP;
1904 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1905 goto got_it;
1907 err = mnt_want_write_file(filp);
1908 if (err)
1909 return err;
1911 /* update superblock with uuid */
1912 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1914 err = f2fs_commit_super(sbi, false);
1915 if (err) {
1916 /* undo new data */
1917 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
1918 mnt_drop_write_file(filp);
1919 return err;
1921 mnt_drop_write_file(filp);
1922 got_it:
1923 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1924 16))
1925 return -EFAULT;
1926 return 0;
1929 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1931 struct inode *inode = file_inode(filp);
1932 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1933 __u32 sync;
1934 int ret;
1936 if (!capable(CAP_SYS_ADMIN))
1937 return -EPERM;
1939 if (get_user(sync, (__u32 __user *)arg))
1940 return -EFAULT;
1942 if (f2fs_readonly(sbi->sb))
1943 return -EROFS;
1945 ret = mnt_want_write_file(filp);
1946 if (ret)
1947 return ret;
1949 if (!sync) {
1950 if (!mutex_trylock(&sbi->gc_mutex)) {
1951 ret = -EBUSY;
1952 goto out;
1954 } else {
1955 mutex_lock(&sbi->gc_mutex);
1958 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
1959 out:
1960 mnt_drop_write_file(filp);
1961 return ret;
1964 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
1966 struct inode *inode = file_inode(filp);
1967 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1968 struct f2fs_gc_range range;
1969 u64 end;
1970 int ret;
1972 if (!capable(CAP_SYS_ADMIN))
1973 return -EPERM;
1975 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
1976 sizeof(range)))
1977 return -EFAULT;
1979 if (f2fs_readonly(sbi->sb))
1980 return -EROFS;
1982 ret = mnt_want_write_file(filp);
1983 if (ret)
1984 return ret;
1986 end = range.start + range.len;
1987 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
1988 return -EINVAL;
1989 do_more:
1990 if (!range.sync) {
1991 if (!mutex_trylock(&sbi->gc_mutex)) {
1992 ret = -EBUSY;
1993 goto out;
1995 } else {
1996 mutex_lock(&sbi->gc_mutex);
1999 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2000 range.start += sbi->blocks_per_seg;
2001 if (range.start <= end)
2002 goto do_more;
2003 out:
2004 mnt_drop_write_file(filp);
2005 return ret;
2008 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2010 struct inode *inode = file_inode(filp);
2011 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2012 int ret;
2014 if (!capable(CAP_SYS_ADMIN))
2015 return -EPERM;
2017 if (f2fs_readonly(sbi->sb))
2018 return -EROFS;
2020 ret = mnt_want_write_file(filp);
2021 if (ret)
2022 return ret;
2024 ret = f2fs_sync_fs(sbi->sb, 1);
2026 mnt_drop_write_file(filp);
2027 return ret;
2030 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2031 struct file *filp,
2032 struct f2fs_defragment *range)
2034 struct inode *inode = file_inode(filp);
2035 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
2036 struct extent_info ei = {0,0,0};
2037 pgoff_t pg_start, pg_end;
2038 unsigned int blk_per_seg = sbi->blocks_per_seg;
2039 unsigned int total = 0, sec_num;
2040 block_t blk_end = 0;
2041 bool fragmented = false;
2042 int err;
2044 /* if in-place-update policy is enabled, don't waste time here */
2045 if (need_inplace_update_policy(inode, NULL))
2046 return -EINVAL;
2048 pg_start = range->start >> PAGE_SHIFT;
2049 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2051 f2fs_balance_fs(sbi, true);
2053 inode_lock(inode);
2055 /* writeback all dirty pages in the range */
2056 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2057 range->start + range->len - 1);
2058 if (err)
2059 goto out;
2062 * lookup mapping info in extent cache, skip defragmenting if physical
2063 * block addresses are continuous.
2065 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2066 if (ei.fofs + ei.len >= pg_end)
2067 goto out;
2070 map.m_lblk = pg_start;
2073 * lookup mapping info in dnode page cache, skip defragmenting if all
2074 * physical block addresses are continuous even if there are hole(s)
2075 * in logical blocks.
2077 while (map.m_lblk < pg_end) {
2078 map.m_len = pg_end - map.m_lblk;
2079 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2080 if (err)
2081 goto out;
2083 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2084 map.m_lblk++;
2085 continue;
2088 if (blk_end && blk_end != map.m_pblk) {
2089 fragmented = true;
2090 break;
2092 blk_end = map.m_pblk + map.m_len;
2094 map.m_lblk += map.m_len;
2097 if (!fragmented)
2098 goto out;
2100 map.m_lblk = pg_start;
2101 map.m_len = pg_end - pg_start;
2103 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2106 * make sure there are enough free section for LFS allocation, this can
2107 * avoid defragment running in SSR mode when free section are allocated
2108 * intensively
2110 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2111 err = -EAGAIN;
2112 goto out;
2115 while (map.m_lblk < pg_end) {
2116 pgoff_t idx;
2117 int cnt = 0;
2119 do_map:
2120 map.m_len = pg_end - map.m_lblk;
2121 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2122 if (err)
2123 goto clear_out;
2125 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2126 map.m_lblk++;
2127 continue;
2130 set_inode_flag(inode, FI_DO_DEFRAG);
2132 idx = map.m_lblk;
2133 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2134 struct page *page;
2136 page = get_lock_data_page(inode, idx, true);
2137 if (IS_ERR(page)) {
2138 err = PTR_ERR(page);
2139 goto clear_out;
2142 set_page_dirty(page);
2143 f2fs_put_page(page, 1);
2145 idx++;
2146 cnt++;
2147 total++;
2150 map.m_lblk = idx;
2152 if (idx < pg_end && cnt < blk_per_seg)
2153 goto do_map;
2155 clear_inode_flag(inode, FI_DO_DEFRAG);
2157 err = filemap_fdatawrite(inode->i_mapping);
2158 if (err)
2159 goto out;
2161 clear_out:
2162 clear_inode_flag(inode, FI_DO_DEFRAG);
2163 out:
2164 inode_unlock(inode);
2165 if (!err)
2166 range->len = (u64)total << PAGE_SHIFT;
2167 return err;
2170 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2172 struct inode *inode = file_inode(filp);
2173 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2174 struct f2fs_defragment range;
2175 int err;
2177 if (!capable(CAP_SYS_ADMIN))
2178 return -EPERM;
2180 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2181 return -EINVAL;
2183 if (f2fs_readonly(sbi->sb))
2184 return -EROFS;
2186 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2187 sizeof(range)))
2188 return -EFAULT;
2190 /* verify alignment of offset & size */
2191 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2192 return -EINVAL;
2194 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2195 sbi->max_file_blocks))
2196 return -EINVAL;
2198 err = mnt_want_write_file(filp);
2199 if (err)
2200 return err;
2202 err = f2fs_defragment_range(sbi, filp, &range);
2203 mnt_drop_write_file(filp);
2205 f2fs_update_time(sbi, REQ_TIME);
2206 if (err < 0)
2207 return err;
2209 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2210 sizeof(range)))
2211 return -EFAULT;
2213 return 0;
2216 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2217 struct file *file_out, loff_t pos_out, size_t len)
2219 struct inode *src = file_inode(file_in);
2220 struct inode *dst = file_inode(file_out);
2221 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2222 size_t olen = len, dst_max_i_size = 0;
2223 size_t dst_osize;
2224 int ret;
2226 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2227 src->i_sb != dst->i_sb)
2228 return -EXDEV;
2230 if (unlikely(f2fs_readonly(src->i_sb)))
2231 return -EROFS;
2233 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2234 return -EINVAL;
2236 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2237 return -EOPNOTSUPP;
2239 if (src == dst) {
2240 if (pos_in == pos_out)
2241 return 0;
2242 if (pos_out > pos_in && pos_out < pos_in + len)
2243 return -EINVAL;
2246 inode_lock(src);
2247 if (src != dst) {
2248 if (!inode_trylock(dst)) {
2249 ret = -EBUSY;
2250 goto out;
2254 ret = -EINVAL;
2255 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2256 goto out_unlock;
2257 if (len == 0)
2258 olen = len = src->i_size - pos_in;
2259 if (pos_in + len == src->i_size)
2260 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2261 if (len == 0) {
2262 ret = 0;
2263 goto out_unlock;
2266 dst_osize = dst->i_size;
2267 if (pos_out + olen > dst->i_size)
2268 dst_max_i_size = pos_out + olen;
2270 /* verify the end result is block aligned */
2271 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2272 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2273 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2274 goto out_unlock;
2276 ret = f2fs_convert_inline_inode(src);
2277 if (ret)
2278 goto out_unlock;
2280 ret = f2fs_convert_inline_inode(dst);
2281 if (ret)
2282 goto out_unlock;
2284 /* write out all dirty pages from offset */
2285 ret = filemap_write_and_wait_range(src->i_mapping,
2286 pos_in, pos_in + len);
2287 if (ret)
2288 goto out_unlock;
2290 ret = filemap_write_and_wait_range(dst->i_mapping,
2291 pos_out, pos_out + len);
2292 if (ret)
2293 goto out_unlock;
2295 f2fs_balance_fs(sbi, true);
2296 f2fs_lock_op(sbi);
2297 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2298 pos_out >> F2FS_BLKSIZE_BITS,
2299 len >> F2FS_BLKSIZE_BITS, false);
2301 if (!ret) {
2302 if (dst_max_i_size)
2303 f2fs_i_size_write(dst, dst_max_i_size);
2304 else if (dst_osize != dst->i_size)
2305 f2fs_i_size_write(dst, dst_osize);
2307 f2fs_unlock_op(sbi);
2308 out_unlock:
2309 if (src != dst)
2310 inode_unlock(dst);
2311 out:
2312 inode_unlock(src);
2313 return ret;
2316 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2318 struct f2fs_move_range range;
2319 struct fd dst;
2320 int err;
2322 if (!(filp->f_mode & FMODE_READ) ||
2323 !(filp->f_mode & FMODE_WRITE))
2324 return -EBADF;
2326 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2327 sizeof(range)))
2328 return -EFAULT;
2330 dst = fdget(range.dst_fd);
2331 if (!dst.file)
2332 return -EBADF;
2334 if (!(dst.file->f_mode & FMODE_WRITE)) {
2335 err = -EBADF;
2336 goto err_out;
2339 err = mnt_want_write_file(filp);
2340 if (err)
2341 goto err_out;
2343 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2344 range.pos_out, range.len);
2346 mnt_drop_write_file(filp);
2347 if (err)
2348 goto err_out;
2350 if (copy_to_user((struct f2fs_move_range __user *)arg,
2351 &range, sizeof(range)))
2352 err = -EFAULT;
2353 err_out:
2354 fdput(dst);
2355 return err;
2358 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2360 struct inode *inode = file_inode(filp);
2361 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2362 struct sit_info *sm = SIT_I(sbi);
2363 unsigned int start_segno = 0, end_segno = 0;
2364 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2365 struct f2fs_flush_device range;
2366 int ret;
2368 if (!capable(CAP_SYS_ADMIN))
2369 return -EPERM;
2371 if (f2fs_readonly(sbi->sb))
2372 return -EROFS;
2374 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2375 sizeof(range)))
2376 return -EFAULT;
2378 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2379 sbi->segs_per_sec != 1) {
2380 f2fs_msg(sbi->sb, KERN_WARNING,
2381 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2382 range.dev_num, sbi->s_ndevs,
2383 sbi->segs_per_sec);
2384 return -EINVAL;
2387 ret = mnt_want_write_file(filp);
2388 if (ret)
2389 return ret;
2391 if (range.dev_num != 0)
2392 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2393 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2395 start_segno = sm->last_victim[FLUSH_DEVICE];
2396 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2397 start_segno = dev_start_segno;
2398 end_segno = min(start_segno + range.segments, dev_end_segno);
2400 while (start_segno < end_segno) {
2401 if (!mutex_trylock(&sbi->gc_mutex)) {
2402 ret = -EBUSY;
2403 goto out;
2405 sm->last_victim[GC_CB] = end_segno + 1;
2406 sm->last_victim[GC_GREEDY] = end_segno + 1;
2407 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2408 ret = f2fs_gc(sbi, true, true, start_segno);
2409 if (ret == -EAGAIN)
2410 ret = 0;
2411 else if (ret < 0)
2412 break;
2413 start_segno++;
2415 out:
2416 mnt_drop_write_file(filp);
2417 return ret;
2420 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2422 struct inode *inode = file_inode(filp);
2423 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2425 /* Must validate to set it with SQLite behavior in Android. */
2426 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2428 return put_user(sb_feature, (u32 __user *)arg);
2431 #ifdef CONFIG_QUOTA
2432 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2434 struct inode *inode = file_inode(filp);
2435 struct f2fs_inode_info *fi = F2FS_I(inode);
2436 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2437 struct super_block *sb = sbi->sb;
2438 struct dquot *transfer_to[MAXQUOTAS] = {};
2439 struct page *ipage;
2440 kprojid_t kprojid;
2441 int err;
2443 if (!f2fs_sb_has_project_quota(sb)) {
2444 if (projid != F2FS_DEF_PROJID)
2445 return -EOPNOTSUPP;
2446 else
2447 return 0;
2450 if (!f2fs_has_extra_attr(inode))
2451 return -EOPNOTSUPP;
2453 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2455 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2456 return 0;
2458 err = mnt_want_write_file(filp);
2459 if (err)
2460 return err;
2462 err = -EPERM;
2463 inode_lock(inode);
2465 /* Is it quota file? Do not allow user to mess with it */
2466 if (IS_NOQUOTA(inode))
2467 goto out_unlock;
2469 ipage = get_node_page(sbi, inode->i_ino);
2470 if (IS_ERR(ipage)) {
2471 err = PTR_ERR(ipage);
2472 goto out_unlock;
2475 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2476 i_projid)) {
2477 err = -EOVERFLOW;
2478 f2fs_put_page(ipage, 1);
2479 goto out_unlock;
2481 f2fs_put_page(ipage, 1);
2483 dquot_initialize(inode);
2485 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2486 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2487 err = __dquot_transfer(inode, transfer_to);
2488 dqput(transfer_to[PRJQUOTA]);
2489 if (err)
2490 goto out_dirty;
2493 F2FS_I(inode)->i_projid = kprojid;
2494 inode->i_ctime = current_time(inode);
2495 out_dirty:
2496 f2fs_mark_inode_dirty_sync(inode, true);
2497 out_unlock:
2498 inode_unlock(inode);
2499 mnt_drop_write_file(filp);
2500 return err;
2502 #else
2503 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2505 if (projid != F2FS_DEF_PROJID)
2506 return -EOPNOTSUPP;
2507 return 0;
2509 #endif
2511 /* Transfer internal flags to xflags */
2512 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2514 __u32 xflags = 0;
2516 if (iflags & FS_SYNC_FL)
2517 xflags |= FS_XFLAG_SYNC;
2518 if (iflags & FS_IMMUTABLE_FL)
2519 xflags |= FS_XFLAG_IMMUTABLE;
2520 if (iflags & FS_APPEND_FL)
2521 xflags |= FS_XFLAG_APPEND;
2522 if (iflags & FS_NODUMP_FL)
2523 xflags |= FS_XFLAG_NODUMP;
2524 if (iflags & FS_NOATIME_FL)
2525 xflags |= FS_XFLAG_NOATIME;
2526 if (iflags & FS_PROJINHERIT_FL)
2527 xflags |= FS_XFLAG_PROJINHERIT;
2528 return xflags;
2531 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2532 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2533 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2535 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2536 #define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2537 FS_IMMUTABLE_FL | \
2538 FS_APPEND_FL | \
2539 FS_NODUMP_FL | \
2540 FS_NOATIME_FL | \
2541 FS_PROJINHERIT_FL)
2543 /* Transfer xflags flags to internal */
2544 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2546 unsigned long iflags = 0;
2548 if (xflags & FS_XFLAG_SYNC)
2549 iflags |= FS_SYNC_FL;
2550 if (xflags & FS_XFLAG_IMMUTABLE)
2551 iflags |= FS_IMMUTABLE_FL;
2552 if (xflags & FS_XFLAG_APPEND)
2553 iflags |= FS_APPEND_FL;
2554 if (xflags & FS_XFLAG_NODUMP)
2555 iflags |= FS_NODUMP_FL;
2556 if (xflags & FS_XFLAG_NOATIME)
2557 iflags |= FS_NOATIME_FL;
2558 if (xflags & FS_XFLAG_PROJINHERIT)
2559 iflags |= FS_PROJINHERIT_FL;
2561 return iflags;
2564 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2566 struct inode *inode = file_inode(filp);
2567 struct f2fs_inode_info *fi = F2FS_I(inode);
2568 struct fsxattr fa;
2570 memset(&fa, 0, sizeof(struct fsxattr));
2571 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2572 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2574 if (f2fs_sb_has_project_quota(inode->i_sb))
2575 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2576 fi->i_projid);
2578 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2579 return -EFAULT;
2580 return 0;
2583 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2585 struct inode *inode = file_inode(filp);
2586 struct f2fs_inode_info *fi = F2FS_I(inode);
2587 struct fsxattr fa;
2588 unsigned int flags;
2589 int err;
2591 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2592 return -EFAULT;
2594 /* Make sure caller has proper permission */
2595 if (!inode_owner_or_capable(inode))
2596 return -EACCES;
2598 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2599 return -EOPNOTSUPP;
2601 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2602 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2603 return -EOPNOTSUPP;
2605 err = mnt_want_write_file(filp);
2606 if (err)
2607 return err;
2609 inode_lock(inode);
2610 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2611 (flags & F2FS_FL_XFLAG_VISIBLE);
2612 err = __f2fs_ioc_setflags(inode, flags);
2613 inode_unlock(inode);
2614 mnt_drop_write_file(filp);
2615 if (err)
2616 return err;
2618 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2619 if (err)
2620 return err;
2622 return 0;
2625 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2627 switch (cmd) {
2628 case F2FS_IOC_GETFLAGS:
2629 return f2fs_ioc_getflags(filp, arg);
2630 case F2FS_IOC_SETFLAGS:
2631 return f2fs_ioc_setflags(filp, arg);
2632 case F2FS_IOC_GETVERSION:
2633 return f2fs_ioc_getversion(filp, arg);
2634 case F2FS_IOC_START_ATOMIC_WRITE:
2635 return f2fs_ioc_start_atomic_write(filp);
2636 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2637 return f2fs_ioc_commit_atomic_write(filp);
2638 case F2FS_IOC_START_VOLATILE_WRITE:
2639 return f2fs_ioc_start_volatile_write(filp);
2640 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2641 return f2fs_ioc_release_volatile_write(filp);
2642 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2643 return f2fs_ioc_abort_volatile_write(filp);
2644 case F2FS_IOC_SHUTDOWN:
2645 return f2fs_ioc_shutdown(filp, arg);
2646 case FITRIM:
2647 return f2fs_ioc_fitrim(filp, arg);
2648 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2649 return f2fs_ioc_set_encryption_policy(filp, arg);
2650 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2651 return f2fs_ioc_get_encryption_policy(filp, arg);
2652 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2653 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2654 case F2FS_IOC_GARBAGE_COLLECT:
2655 return f2fs_ioc_gc(filp, arg);
2656 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2657 return f2fs_ioc_gc_range(filp, arg);
2658 case F2FS_IOC_WRITE_CHECKPOINT:
2659 return f2fs_ioc_write_checkpoint(filp, arg);
2660 case F2FS_IOC_DEFRAGMENT:
2661 return f2fs_ioc_defragment(filp, arg);
2662 case F2FS_IOC_MOVE_RANGE:
2663 return f2fs_ioc_move_range(filp, arg);
2664 case F2FS_IOC_FLUSH_DEVICE:
2665 return f2fs_ioc_flush_device(filp, arg);
2666 case F2FS_IOC_GET_FEATURES:
2667 return f2fs_ioc_get_features(filp, arg);
2668 case F2FS_IOC_FSGETXATTR:
2669 return f2fs_ioc_fsgetxattr(filp, arg);
2670 case F2FS_IOC_FSSETXATTR:
2671 return f2fs_ioc_fssetxattr(filp, arg);
2672 default:
2673 return -ENOTTY;
2677 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2679 struct file *file = iocb->ki_filp;
2680 struct inode *inode = file_inode(file);
2681 struct blk_plug plug;
2682 ssize_t ret;
2684 inode_lock(inode);
2685 ret = generic_write_checks(iocb, from);
2686 if (ret > 0) {
2687 int err;
2689 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2690 set_inode_flag(inode, FI_NO_PREALLOC);
2692 err = f2fs_preallocate_blocks(iocb, from);
2693 if (err) {
2694 inode_unlock(inode);
2695 return err;
2697 blk_start_plug(&plug);
2698 ret = __generic_file_write_iter(iocb, from);
2699 blk_finish_plug(&plug);
2700 clear_inode_flag(inode, FI_NO_PREALLOC);
2702 if (ret > 0)
2703 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
2705 inode_unlock(inode);
2707 if (ret > 0)
2708 ret = generic_write_sync(iocb, ret);
2709 return ret;
2712 #ifdef CONFIG_COMPAT
2713 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2715 switch (cmd) {
2716 case F2FS_IOC32_GETFLAGS:
2717 cmd = F2FS_IOC_GETFLAGS;
2718 break;
2719 case F2FS_IOC32_SETFLAGS:
2720 cmd = F2FS_IOC_SETFLAGS;
2721 break;
2722 case F2FS_IOC32_GETVERSION:
2723 cmd = F2FS_IOC_GETVERSION;
2724 break;
2725 case F2FS_IOC_START_ATOMIC_WRITE:
2726 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2727 case F2FS_IOC_START_VOLATILE_WRITE:
2728 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2729 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2730 case F2FS_IOC_SHUTDOWN:
2731 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2732 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2733 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2734 case F2FS_IOC_GARBAGE_COLLECT:
2735 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2736 case F2FS_IOC_WRITE_CHECKPOINT:
2737 case F2FS_IOC_DEFRAGMENT:
2738 case F2FS_IOC_MOVE_RANGE:
2739 case F2FS_IOC_FLUSH_DEVICE:
2740 case F2FS_IOC_GET_FEATURES:
2741 case F2FS_IOC_FSGETXATTR:
2742 case F2FS_IOC_FSSETXATTR:
2743 break;
2744 default:
2745 return -ENOIOCTLCMD;
2747 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2749 #endif
2751 const struct file_operations f2fs_file_operations = {
2752 .llseek = f2fs_llseek,
2753 .read_iter = generic_file_read_iter,
2754 .write_iter = f2fs_file_write_iter,
2755 .open = f2fs_file_open,
2756 .release = f2fs_release_file,
2757 .mmap = f2fs_file_mmap,
2758 .flush = f2fs_file_flush,
2759 .fsync = f2fs_sync_file,
2760 .fallocate = f2fs_fallocate,
2761 .unlocked_ioctl = f2fs_ioctl,
2762 #ifdef CONFIG_COMPAT
2763 .compat_ioctl = f2fs_compat_ioctl,
2764 #endif
2765 .splice_read = generic_file_splice_read,
2766 .splice_write = iter_file_splice_write,