projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: avoid stale fi->gdirty_list pointer
[linux-block.git] / fs / f2fs / file.c
CommitLineData
0a8165d7 1/*
fbfa2cc5
JK		 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>
ae51fb31 16#include <linux/blkdev.h>
fbfa2cc5
JK		 17#include <linux/falloc.h>
18#include <linux/types.h>
e9750824 19#include <linux/compat.h>
fbfa2cc5
JK		 20#include <linux/uaccess.h>
21#include <linux/mount.h>
7f7670fe 22#include <linux/pagevec.h>
dc91de78 23#include <linux/uio.h>
8da4b8c4 24#include <linux/uuid.h>
4dd6f977 25#include <linux/file.h>
fbfa2cc5
JK		 26
27#include "f2fs.h"
28#include "node.h"
29#include "segment.h"
30#include "xattr.h"
31#include "acl.h"
c1c1b583 32#include "gc.h"
db9f7c1a 33#include "trace.h"
a2a4a7e4 34#include <trace/events/f2fs.h>
fbfa2cc5 35
5a3a2d83
QS		 36static int f2fs_filemap_fault(struct vm_fault *vmf)
37{
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 int err;
40
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 err = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
44
45 return err;
46}
47
11bac800 48static int f2fs_vm_page_mkwrite(struct vm_fault *vmf)
fbfa2cc5
JK		 49{
50 struct page *page = vmf->page;
11bac800 51 struct inode *inode = file_inode(vmf->vma->vm_file);
4081363f 52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5 53 struct dnode_of_data dn;
e479556b 54 int err;
fbfa2cc5 55
fbfa2cc5 56 sb_start_pagefault(inode->i_sb);
b3d208f9
JK		 57
58 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
b067ba1f 59
fbfa2cc5 60 /* block allocation */
e479556b 61 f2fs_lock_op(sbi);
b3d208f9 62 set_new_dnode(&dn, inode, NULL, NULL, 0);
b600965c 63 err = f2fs_reserve_block(&dn, page->index);
9ba69cf9
JK		 64 if (err) {
65 f2fs_unlock_op(sbi);
b600965c 66 goto out;
9ba69cf9
JK		 67 }
68 f2fs_put_dnode(&dn);
69 f2fs_unlock_op(sbi);
fbfa2cc5 70
2c4db1a6 71 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 72
11bac800 73 file_update_time(vmf->vma->vm_file);
5a3a2d83 74 down_read(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5 75 lock_page(page);
6bacf52f 76 if (unlikely(page->mapping != inode->i_mapping ||
9851e6e1 77 page_offset(page) > i_size_read(inode) ||
6bacf52f 78 !PageUptodate(page))) {
fbfa2cc5
JK		 79 unlock_page(page);
80 err = -EFAULT;
5a3a2d83 81 goto out_sem;
fbfa2cc5
JK		 82 }
83
84 /*
85 * check to see if the page is mapped already (no holes)
86 */
87 if (PageMappedToDisk(page))
9851e6e1 88 goto mapped;
fbfa2cc5
JK		 89
90 /* page is wholly or partially inside EOF */
09cbfeaf 91 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
9edcdabf 92 i_size_read(inode)) {
fbfa2cc5 93 unsigned offset;
09cbfeaf
KS		 94 offset = i_size_read(inode) & ~PAGE_MASK;
95 zero_user_segment(page, offset, PAGE_SIZE);
fbfa2cc5
JK		 96 }
97 set_page_dirty(page);
237c0790
JK		 98 if (!PageUptodate(page))
99 SetPageUptodate(page);
fbfa2cc5 100
b0af6d49
CY		 101 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
102
e943a10d 103 trace_f2fs_vm_page_mkwrite(page, DATA);
9851e6e1
NJ		 104mapped:
105 /* fill the page */
fec1d657 106 f2fs_wait_on_page_writeback(page, DATA, false);
08b39fbd
CY		 107
108 /* wait for GCed encrypted page writeback */
1958593e 109 if (f2fs_encrypted_file(inode))
d4c759ee 110 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
08b39fbd 111
5a3a2d83
QS		 112out_sem:
113 up_read(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5
JK		 114out:
115 sb_end_pagefault(inode->i_sb);
d0239e1b 116 f2fs_update_time(sbi, REQ_TIME);
fbfa2cc5
JK		 117 return block_page_mkwrite_return(err);
118}
119
120static const struct vm_operations_struct f2fs_file_vm_ops = {
5a3a2d83 121 .fault = f2fs_filemap_fault,
f1820361 122 .map_pages = filemap_map_pages,
692bb55d 123 .page_mkwrite = f2fs_vm_page_mkwrite,
fbfa2cc5
JK		 124};
125
354a3399
JK		 126static int get_parent_ino(struct inode *inode, nid_t *pino)
127{
128 struct dentry *dentry;
129
130 inode = igrab(inode);
131 dentry = d_find_any_alias(inode);
132 iput(inode);
133 if (!dentry)
134 return 0;
135
f0947e5c
JK		 136 *pino = parent_ino(dentry);
137 dput(dentry);
354a3399
JK		 138 return 1;
139}
140
9d1589ef
CY		 141static inline bool need_do_checkpoint(struct inode *inode)
142{
4081363f 143 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9d1589ef
CY		 144 bool need_cp = false;
145
146 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
147 need_cp = true;
bbf156f7 148 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
e7d55452 149 need_cp = true;
9d1589ef
CY		 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;
d5053a34
JK		 156 else if (test_opt(sbi, FASTBOOT))
157 need_cp = true;
a344b9fd
JK		 158 else if (sbi->active_logs == 2)
159 need_cp = true;
9d1589ef
CY		 160
161 return need_cp;
162}
163
9c7bb702
CL		 164static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
165{
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;
173}
174
51455b19
CL		 175static void try_to_fix_pino(struct inode *inode)
176{
177 struct f2fs_inode_info *fi = F2FS_I(inode);
178 nid_t pino;
179
180 down_write(&fi->i_sem);
51455b19
CL		 181 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
182 get_parent_ino(inode, &pino)) {
205b9822 183 f2fs_i_pino_write(inode, pino);
51455b19 184 file_got_pino(inode);
51455b19 185 }
ee6d182f 186 up_write(&fi->i_sem);
51455b19
CL		 187}
188
608514de
JK		 189static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
190 int datasync, bool atomic)
fbfa2cc5
JK		 191{
192 struct inode *inode = file->f_mapping->host;
4081363f 193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2403c155 194 nid_t ino = inode->i_ino;
fbfa2cc5
JK		 195 int ret = 0;
196 bool need_cp = false;
197 struct writeback_control wbc = {
c81bf1c8 198 .sync_mode = WB_SYNC_ALL,
fbfa2cc5
JK		 199 .nr_to_write = LONG_MAX,
200 .for_reclaim = 0,
201 };
202
6bacf52f 203 if (unlikely(f2fs_readonly(inode->i_sb)))
1fa95b0b
NJ		 204 return 0;
205
a2a4a7e4 206 trace_f2fs_sync_file_enter(inode);
ea1aa12c
JK		 207
208 /* if fdatasync is triggered, let's do in-place-update */
c46a155b 209 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
91942321 210 set_inode_flag(inode, FI_NEED_IPU);
3b49c9a1 211 ret = file_write_and_wait_range(file, start, end);
91942321 212 clear_inode_flag(inode, FI_NEED_IPU);
c1ce1b02 213
a2a4a7e4
NJ		 214 if (ret) {
215 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
fbfa2cc5 216 return ret;
a2a4a7e4 217 }
fbfa2cc5 218
9c7bb702 219 /* if the inode is dirty, let's recover all the time */
281518c6 220 if (!f2fs_skip_inode_update(inode, datasync)) {
2286c020 221 f2fs_write_inode(inode, NULL);
9c7bb702
CL		 222 goto go_write;
223 }
224
6d99ba41
JK		 225 /*
226 * if there is no written data, don't waste time to write recovery info.
227 */
91942321 228 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
2403c155 229 !exist_written_data(sbi, ino, APPEND_INO)) {
19c9c466 230
9c7bb702
CL		 231 /* it may call write_inode just prior to fsync */
232 if (need_inode_page_update(sbi, ino))
19c9c466 233 goto go_write;
19c9c466 234
91942321 235 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
2403c155 236 exist_written_data(sbi, ino, UPDATE_INO))
6d99ba41
JK		 237 goto flush_out;
238 goto out;
239 }
19c9c466 240go_write:
e5d2385e
JK		 241 /*
242 * Both of fdatasync() and fsync() are able to be recovered from
243 * sudden-power-off.
244 */
91942321 245 down_read(&F2FS_I(inode)->i_sem);
9d1589ef 246 need_cp = need_do_checkpoint(inode);
91942321 247 up_read(&F2FS_I(inode)->i_sem);
d928bfbf 248
fbfa2cc5
JK		 249 if (need_cp) {
250 /* all the dirty node pages should be flushed for POR */
251 ret = f2fs_sync_fs(inode->i_sb, 1);
d928bfbf 252
51455b19
CL		 253 /*
254 * We've secured consistency through sync_fs. Following pino
255 * will be used only for fsynced inodes after checkpoint.
256 */
257 try_to_fix_pino(inode);
91942321
JK		 258 clear_inode_flag(inode, FI_APPEND_WRITE);
259 clear_inode_flag(inode, FI_UPDATE_WRITE);
51455b19
CL		 260 goto out;
261 }
88bd02c9 262sync_nodes:
26de9b11 263 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
c267ec15
JK		 264 if (ret)
265 goto out;
51455b19 266
871f599f 267 /* if cp_error was enabled, we should avoid infinite loop */
6d5a1495
CY		 268 if (unlikely(f2fs_cp_error(sbi))) {
269 ret = -EIO;
871f599f 270 goto out;
6d5a1495 271 }
871f599f 272
51455b19 273 if (need_inode_block_update(sbi, ino)) {
7c45729a 274 f2fs_mark_inode_dirty_sync(inode, true);
51455b19
CL		 275 f2fs_write_inode(inode, NULL);
276 goto sync_nodes;
fbfa2cc5 277 }
51455b19 278
b6a245eb
JK		 279 /*
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.
286 */
287 if (!atomic) {
288 ret = wait_on_node_pages_writeback(sbi, ino);
289 if (ret)
290 goto out;
291 }
51455b19
CL		 292
293 /* once recovery info is written, don't need to tack this */
a49324f1 294 remove_ino_entry(sbi, ino, APPEND_INO);
91942321 295 clear_inode_flag(inode, FI_APPEND_WRITE);
51455b19 296flush_out:
e7c75ab0 297 if (!atomic)
39d787be 298 ret = f2fs_issue_flush(sbi, inode->i_ino);
3f06252f
CY		 299 if (!ret) {
300 remove_ino_entry(sbi, ino, UPDATE_INO);
301 clear_inode_flag(inode, FI_UPDATE_WRITE);
39d787be 302 remove_ino_entry(sbi, ino, FLUSH_INO);
3f06252f 303 }
d0239e1b 304 f2fs_update_time(sbi, REQ_TIME);
fbfa2cc5 305out:
a2a4a7e4 306 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
05ca3632 307 f2fs_trace_ios(NULL, 1);
fbfa2cc5
JK		 308 return ret;
309}
310
608514de
JK		 311int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
312{
313 return f2fs_do_sync_file(file, start, end, datasync, false);
314}
315
7f7670fe
JK		 316static pgoff_t __get_first_dirty_index(struct address_space *mapping,
317 pgoff_t pgofs, int whence)
318{
319 struct pagevec pvec;
320 int nr_pages;
321
322 if (whence != SEEK_DATA)
323 return 0;
324
325 /* find first dirty page index */
326 pagevec_init(&pvec, 0);
65b85ccc
JK		 327 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
328 PAGECACHE_TAG_DIRTY, 1);
80dd9c0e 329 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
7f7670fe
JK		 330 pagevec_release(&pvec);
331 return pgofs;
332}
333
334static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
335 int whence)
336{
337 switch (whence) {
338 case SEEK_DATA:
339 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
340 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
341 return true;
342 break;
343 case SEEK_HOLE:
344 if (blkaddr == NULL_ADDR)
345 return true;
346 break;
347 }
348 return false;
349}
350
267378d4
CY		 351static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
352{
353 struct inode *inode = file->f_mapping->host;
354 loff_t maxbytes = inode->i_sb->s_maxbytes;
355 struct dnode_of_data dn;
7f7670fe
JK		 356 pgoff_t pgofs, end_offset, dirty;
357 loff_t data_ofs = offset;
358 loff_t isize;
267378d4
CY		 359 int err = 0;
360
5955102c 361 inode_lock(inode);
267378d4
CY		 362
363 isize = i_size_read(inode);
364 if (offset >= isize)
365 goto fail;
366
367 /* handle inline data case */
622f28ae 368 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
267378d4
CY		 369 if (whence == SEEK_HOLE)
370 data_ofs = isize;
371 goto found;
372 }
373
09cbfeaf 374 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
267378d4 375
7f7670fe
JK		 376 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
377
09cbfeaf 378 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 379 set_new_dnode(&dn, inode, NULL, NULL, 0);
ac6f1999 380 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
267378d4
CY		 381 if (err && err != -ENOENT) {
382 goto fail;
383 } else if (err == -ENOENT) {
e1c42045 384 /* direct node does not exists */
267378d4 385 if (whence == SEEK_DATA) {
3cf45747 386 pgofs = get_next_page_offset(&dn, pgofs);
267378d4
CY		 387 continue;
388 } else {
389 goto found;
390 }
391 }
392
81ca7350 393 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
267378d4
CY		 394
395 /* find data/hole in dnode block */
396 for (; dn.ofs_in_node < end_offset;
397 dn.ofs_in_node++, pgofs++,
09cbfeaf 398 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 399 block_t blkaddr;
7a2af766
CY		 400 blkaddr = datablock_addr(dn.inode,
401 dn.node_page, dn.ofs_in_node);
267378d4 402
7f7670fe 403 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
267378d4
CY		 404 f2fs_put_dnode(&dn);
405 goto found;
406 }
407 }
408 f2fs_put_dnode(&dn);
409 }
410
411 if (whence == SEEK_DATA)
412 goto fail;
267378d4 413found:
fe369bc8
JK		 414 if (whence == SEEK_HOLE && data_ofs > isize)
415 data_ofs = isize;
5955102c 416 inode_unlock(inode);
267378d4
CY		 417 return vfs_setpos(file, data_ofs, maxbytes);
418fail:
5955102c 419 inode_unlock(inode);
267378d4
CY		 420 return -ENXIO;
421}
422
423static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
424{
425 struct inode *inode = file->f_mapping->host;
426 loff_t maxbytes = inode->i_sb->s_maxbytes;
427
428 switch (whence) {
429 case SEEK_SET:
430 case SEEK_CUR:
431 case SEEK_END:
432 return generic_file_llseek_size(file, offset, whence,
433 maxbytes, i_size_read(inode));
434 case SEEK_DATA:
435 case SEEK_HOLE:
0b4c5afd
JK		 436 if (offset < 0)
437 return -ENXIO;
267378d4
CY		 438 return f2fs_seek_block(file, offset, whence);
439 }
440
441 return -EINVAL;
442}
443
fbfa2cc5
JK		 444static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
445{
b3d208f9 446 struct inode *inode = file_inode(file);
b9d777b8 447 int err;
b3d208f9
JK		 448
449 /* we don't need to use inline_data strictly */
b9d777b8
JK		 450 err = f2fs_convert_inline_inode(inode);
451 if (err)
452 return err;
b3d208f9 453
fbfa2cc5
JK		 454 file_accessed(file);
455 vma->vm_ops = &f2fs_file_vm_ops;
456 return 0;
457}
458
fcc85a4d
JK		 459static int f2fs_file_open(struct inode *inode, struct file *filp)
460{
33b13951 461 struct dentry *dir;
fcc85a4d 462
0abd675e
CY		 463 if (f2fs_encrypted_inode(inode)) {
464 int ret = fscrypt_get_encryption_info(inode);
fcc85a4d 465 if (ret)
ae108668 466 return -EACCES;
0b81d077 467 if (!fscrypt_has_encryption_key(inode))
ae108668 468 return -ENOKEY;
fcc85a4d 469 }
33b13951
JK		 470 dir = dget_parent(file_dentry(filp));
471 if (f2fs_encrypted_inode(d_inode(dir)) &&
472 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
473 dput(dir);
8074bb51 474 return -EPERM;
33b13951
JK		 475 }
476 dput(dir);
0abd675e 477 return dquot_file_open(inode, filp);
fcc85a4d
JK		 478}
479
b292dcab 480int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
fbfa2cc5 481{
4081363f 482 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
fbfa2cc5 483 struct f2fs_node *raw_node;
19b2c30d 484 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
fbfa2cc5 485 __le32 *addr;
7a2af766
CY		 486 int base = 0;
487
488 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
489 base = get_extra_isize(dn->inode);
fbfa2cc5 490
45590710 491 raw_node = F2FS_NODE(dn->node_page);
7a2af766 492 addr = blkaddr_in_node(raw_node) + base + ofs;
fbfa2cc5 493
6c311ec6 494 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
fbfa2cc5
JK		 495 block_t blkaddr = le32_to_cpu(*addr);
496 if (blkaddr == NULL_ADDR)
497 continue;
498
e1509cf2 499 dn->data_blkaddr = NULL_ADDR;
216a620a 500 set_data_blkaddr(dn);
fbfa2cc5 501 invalidate_blocks(sbi, blkaddr);
3c6c2beb 502 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
91942321 503 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
fbfa2cc5
JK		 504 nr_free++;
505 }
19b2c30d 506
fbfa2cc5 507 if (nr_free) {
19b2c30d
CY		 508 pgoff_t fofs;
509 /*
510 * once we invalidate valid blkaddr in range [ofs, ofs + count],
511 * we will invalidate all blkaddr in the whole range.
512 */
513 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
81ca7350 514 dn->inode) + ofs;
19b2c30d 515 f2fs_update_extent_cache_range(dn, fofs, 0, len);
d7cc950b 516 dec_valid_block_count(sbi, dn->inode, nr_free);
fbfa2cc5
JK		 517 }
518 dn->ofs_in_node = ofs;
51dd6249 519
d0239e1b 520 f2fs_update_time(sbi, REQ_TIME);
51dd6249
NJ		 521 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
522 dn->ofs_in_node, nr_free);
fbfa2cc5
JK		 523 return nr_free;
524}
525
526void truncate_data_blocks(struct dnode_of_data *dn)
527{
528 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
529}
530
0bfcfcca 531static int truncate_partial_data_page(struct inode *inode, u64 from,
43f3eae1 532 bool cache_only)
fbfa2cc5 533{
09cbfeaf
KS		 534 unsigned offset = from & (PAGE_SIZE - 1);
535 pgoff_t index = from >> PAGE_SHIFT;
43f3eae1 536 struct address_space *mapping = inode->i_mapping;
fbfa2cc5
JK		 537 struct page *page;
538
43f3eae1 539 if (!offset && !cache_only)
b3d208f9 540 return 0;
fbfa2cc5 541
43f3eae1 542 if (cache_only) {
34b5d5c2 543 page = find_lock_page(mapping, index);
43f3eae1
JK		 544 if (page && PageUptodate(page))
545 goto truncate_out;
546 f2fs_put_page(page, 1);
b3d208f9 547 return 0;
43f3eae1 548 }
fbfa2cc5 549
a56c7c6f 550 page = get_lock_data_page(inode, index, true);
43f3eae1 551 if (IS_ERR(page))
a78aaa2c 552 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
43f3eae1 553truncate_out:
fec1d657 554 f2fs_wait_on_page_writeback(page, DATA, true);
09cbfeaf 555 zero_user(page, offset, PAGE_SIZE - offset);
a9bcf9bc
JK		 556
557 /* An encrypted inode should have a key and truncate the last page. */
558 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
559 if (!cache_only)
0bfcfcca 560 set_page_dirty(page);
fbfa2cc5 561 f2fs_put_page(page, 1);
b3d208f9 562 return 0;
fbfa2cc5
JK		 563}
564
764aa3e9 565int truncate_blocks(struct inode *inode, u64 from, bool lock)
fbfa2cc5 566{
4081363f 567 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 568 unsigned int blocksize = inode->i_sb->s_blocksize;
569 struct dnode_of_data dn;
570 pgoff_t free_from;
9ffe0fb5 571 int count = 0, err = 0;
b3d208f9 572 struct page *ipage;
0bfcfcca 573 bool truncate_page = false;
fbfa2cc5 574
51dd6249
NJ		 575 trace_f2fs_truncate_blocks_enter(inode, from);
576
f7ef9b83 577 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
fbfa2cc5 578
09210c97
CY		 579 if (free_from >= sbi->max_file_blocks)
580 goto free_partial;
581
764aa3e9
JK		 582 if (lock)
583 f2fs_lock_op(sbi);
9ffe0fb5 584
b3d208f9
JK		 585 ipage = get_node_page(sbi, inode->i_ino);
586 if (IS_ERR(ipage)) {
587 err = PTR_ERR(ipage);
588 goto out;
589 }
590
591 if (f2fs_has_inline_data(inode)) {
bd4667cb 592 truncate_inline_inode(inode, ipage, from);
b3d208f9 593 f2fs_put_page(ipage, 1);
0bfcfcca 594 truncate_page = true;
b3d208f9
JK		 595 goto out;
596 }
597
598 set_new_dnode(&dn, inode, ipage, NULL, 0);
79344efb 599 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
fbfa2cc5
JK		 600 if (err) {
601 if (err == -ENOENT)
602 goto free_next;
b3d208f9 603 goto out;
1ce86bf6
JK		 604 }
605
81ca7350 606 count = ADDRS_PER_PAGE(dn.node_page, inode);
fbfa2cc5
JK		 607
608 count -= dn.ofs_in_node;
9850cf4a 609 f2fs_bug_on(sbi, count < 0);
39936837 610
fbfa2cc5
JK		 611 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
612 truncate_data_blocks_range(&dn, count);
613 free_from += count;
614 }
615
616 f2fs_put_dnode(&dn);
617free_next:
618 err = truncate_inode_blocks(inode, free_from);
764d2c80
JK		 619out:
620 if (lock)
621 f2fs_unlock_op(sbi);
09210c97 622free_partial:
b3d208f9
JK		 623 /* lastly zero out the first data page */
624 if (!err)
0bfcfcca 625 err = truncate_partial_data_page(inode, from, truncate_page);
fbfa2cc5 626
51dd6249 627 trace_f2fs_truncate_blocks_exit(inode, err);
fbfa2cc5
JK		 628 return err;
629}
630
9a449e9c 631int f2fs_truncate(struct inode *inode)
fbfa2cc5 632{
b0154891
CY		 633 int err;
634
fbfa2cc5
JK		 635 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
636 S_ISLNK(inode->i_mode)))
b0154891 637 return 0;
fbfa2cc5 638
51dd6249
NJ		 639 trace_f2fs_truncate(inode);
640
14b44d23
JK		 641#ifdef CONFIG_F2FS_FAULT_INJECTION
642 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
643 f2fs_show_injection_info(FAULT_TRUNCATE);
644 return -EIO;
645 }
646#endif
92dffd01 647 /* we should check inline_data size */
b9d777b8 648 if (!f2fs_may_inline_data(inode)) {
b0154891
CY		 649 err = f2fs_convert_inline_inode(inode);
650 if (err)
651 return err;
92dffd01
JK		 652 }
653
9a449e9c 654 err = truncate_blocks(inode, i_size_read(inode), true);
b0154891
CY		 655 if (err)
656 return err;
657
078cd827 658 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 659 f2fs_mark_inode_dirty_sync(inode, false);
b0154891 660 return 0;
fbfa2cc5
JK		 661}
662
a528d35e 663int f2fs_getattr(const struct path *path, struct kstat *stat,
1c6d8ee4 664 u32 request_mask, unsigned int query_flags)
fbfa2cc5 665{
a528d35e 666 struct inode *inode = d_inode(path->dentry);
1c6d8ee4
CY		 667 struct f2fs_inode_info *fi = F2FS_I(inode);
668 unsigned int flags;
669
2c1d0305 670 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1c6d8ee4
CY		 671 if (flags & FS_APPEND_FL)
672 stat->attributes |= STATX_ATTR_APPEND;
673 if (flags & FS_COMPR_FL)
674 stat->attributes |= STATX_ATTR_COMPRESSED;
675 if (f2fs_encrypted_inode(inode))
676 stat->attributes |= STATX_ATTR_ENCRYPTED;
677 if (flags & FS_IMMUTABLE_FL)
678 stat->attributes |= STATX_ATTR_IMMUTABLE;
679 if (flags & FS_NODUMP_FL)
680 stat->attributes |= STATX_ATTR_NODUMP;
681
682 stat->attributes_mask |= (STATX_ATTR_APPEND |
683 STATX_ATTR_COMPRESSED |
684 STATX_ATTR_ENCRYPTED |
685 STATX_ATTR_IMMUTABLE |
686 STATX_ATTR_NODUMP);
687
fbfa2cc5 688 generic_fillattr(inode, stat);
fbfa2cc5
JK		 689 return 0;
690}
691
692#ifdef CONFIG_F2FS_FS_POSIX_ACL
693static void __setattr_copy(struct inode *inode, const struct iattr *attr)
694{
fbfa2cc5
JK		 695 unsigned int ia_valid = attr->ia_valid;
696
697 if (ia_valid & ATTR_UID)
698 inode->i_uid = attr->ia_uid;
699 if (ia_valid & ATTR_GID)
700 inode->i_gid = attr->ia_gid;
701 if (ia_valid & ATTR_ATIME)
702 inode->i_atime = timespec_trunc(attr->ia_atime,
703 inode->i_sb->s_time_gran);
704 if (ia_valid & ATTR_MTIME)
705 inode->i_mtime = timespec_trunc(attr->ia_mtime,
706 inode->i_sb->s_time_gran);
707 if (ia_valid & ATTR_CTIME)
708 inode->i_ctime = timespec_trunc(attr->ia_ctime,
709 inode->i_sb->s_time_gran);
710 if (ia_valid & ATTR_MODE) {
711 umode_t mode = attr->ia_mode;
712
713 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
714 mode &= ~S_ISGID;
91942321 715 set_acl_inode(inode, mode);
fbfa2cc5
JK		 716 }
717}
718#else
719#define __setattr_copy setattr_copy
720#endif
721
722int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
723{
2b0143b5 724 struct inode *inode = d_inode(dentry);
fbfa2cc5 725 int err;
c0ed4405 726 bool size_changed = false;
fbfa2cc5 727
31051c85 728 err = setattr_prepare(dentry, attr);
fbfa2cc5
JK		 729 if (err)
730 return err;
731
0abd675e
CY		 732 if (is_quota_modification(inode, attr)) {
733 err = dquot_initialize(inode);
734 if (err)
735 return err;
736 }
737 if ((attr->ia_valid & ATTR_UID &&
738 !uid_eq(attr->ia_uid, inode->i_uid)) ||
739 (attr->ia_valid & ATTR_GID &&
740 !gid_eq(attr->ia_gid, inode->i_gid))) {
741 err = dquot_transfer(inode, attr);
742 if (err)
743 return err;
744 }
745
09db6a2e 746 if (attr->ia_valid & ATTR_SIZE) {
67773a1f
EB		 747 if (f2fs_encrypted_inode(inode)) {
748 err = fscrypt_get_encryption_info(inode);
749 if (err)
750 return err;
751 if (!fscrypt_has_encryption_key(inode))
752 return -ENOKEY;
753 }
fcc85a4d 754
3c454145 755 if (attr->ia_size <= i_size_read(inode)) {
5a3a2d83 756 down_write(&F2FS_I(inode)->i_mmap_sem);
09db6a2e 757 truncate_setsize(inode, attr->ia_size);
9a449e9c 758 err = f2fs_truncate(inode);
5a3a2d83 759 up_write(&F2FS_I(inode)->i_mmap_sem);
b0154891
CY		 760 if (err)
761 return err;
09db6a2e
CY		 762 } else {
763 /*
3c454145
CY		 764 * do not trim all blocks after i_size if target size is
765 * larger than i_size.
09db6a2e 766 */
5a3a2d83 767 down_write(&F2FS_I(inode)->i_mmap_sem);
3c454145 768 truncate_setsize(inode, attr->ia_size);
5a3a2d83 769 up_write(&F2FS_I(inode)->i_mmap_sem);
0cab80ee
CY		 770
771 /* should convert inline inode here */
b9d777b8 772 if (!f2fs_may_inline_data(inode)) {
0cab80ee
CY		 773 err = f2fs_convert_inline_inode(inode);
774 if (err)
775 return err;
776 }
078cd827 777 inode->i_mtime = inode->i_ctime = current_time(inode);
09db6a2e 778 }
c0ed4405 779
a0d00fad
CY		 780 down_write(&F2FS_I(inode)->i_sem);
781 F2FS_I(inode)->last_disk_size = i_size_read(inode);
782 up_write(&F2FS_I(inode)->i_sem);
783
c0ed4405 784 size_changed = true;
fbfa2cc5
JK		 785 }
786
787 __setattr_copy(inode, attr);
788
789 if (attr->ia_valid & ATTR_MODE) {
a6dda0e6 790 err = posix_acl_chmod(inode, get_inode_mode(inode));
91942321
JK		 791 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
792 inode->i_mode = F2FS_I(inode)->i_acl_mode;
793 clear_inode_flag(inode, FI_ACL_MODE);
fbfa2cc5
JK		 794 }
795 }
796
c0ed4405
YH		 797 /* file size may changed here */
798 f2fs_mark_inode_dirty_sync(inode, size_changed);
15d04354
JK		 799
800 /* inode change will produce dirty node pages flushed by checkpoint */
801 f2fs_balance_fs(F2FS_I_SB(inode), true);
802
fbfa2cc5
JK		 803 return err;
804}
805
806const struct inode_operations f2fs_file_inode_operations = {
807 .getattr = f2fs_getattr,
808 .setattr = f2fs_setattr,
809 .get_acl = f2fs_get_acl,
a6dda0e6 810 .set_acl = f2fs_set_acl,
fbfa2cc5 811#ifdef CONFIG_F2FS_FS_XATTR
fbfa2cc5 812 .listxattr = f2fs_listxattr,
fbfa2cc5 813#endif
9ab70134 814 .fiemap = f2fs_fiemap,
fbfa2cc5
JK		 815};
816
6394328a 817static int fill_zero(struct inode *inode, pgoff_t index,
fbfa2cc5
JK		 818 loff_t start, loff_t len)
819{
4081363f 820 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 821 struct page *page;
822
823 if (!len)
6394328a 824 return 0;
fbfa2cc5 825
2c4db1a6 826 f2fs_balance_fs(sbi, true);
bd43df02 827
e479556b 828 f2fs_lock_op(sbi);
64aa7ed9 829 page = get_new_data_page(inode, NULL, index, false);
e479556b 830 f2fs_unlock_op(sbi);
fbfa2cc5 831
6394328a
CY		 832 if (IS_ERR(page))
833 return PTR_ERR(page);
834
fec1d657 835 f2fs_wait_on_page_writeback(page, DATA, true);
6394328a
CY		 836 zero_user(page, start, len);
837 set_page_dirty(page);
838 f2fs_put_page(page, 1);
839 return 0;
fbfa2cc5
JK		 840}
841
842int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
843{
fbfa2cc5
JK		 844 int err;
845
ea58711e 846 while (pg_start < pg_end) {
fbfa2cc5 847 struct dnode_of_data dn;
ea58711e 848 pgoff_t end_offset, count;
9eaeba70 849
fbfa2cc5 850 set_new_dnode(&dn, inode, NULL, NULL, 0);
ea58711e 851 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
fbfa2cc5 852 if (err) {
ea58711e
CY		 853 if (err == -ENOENT) {
854 pg_start++;
fbfa2cc5 855 continue;
ea58711e 856 }
fbfa2cc5
JK		 857 return err;
858 }
859
81ca7350 860 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
ea58711e
CY		 861 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
862
863 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
864
865 truncate_data_blocks_range(&dn, count);
fbfa2cc5 866 f2fs_put_dnode(&dn);
ea58711e
CY		 867
868 pg_start += count;
fbfa2cc5
JK		 869 }
870 return 0;
871}
872
a66c7b2f 873static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
fbfa2cc5
JK		 874{
875 pgoff_t pg_start, pg_end;
876 loff_t off_start, off_end;
b9d777b8 877 int ret;
fbfa2cc5 878
b9d777b8
JK		 879 ret = f2fs_convert_inline_inode(inode);
880 if (ret)
881 return ret;
9ffe0fb5 882
09cbfeaf
KS		 883 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
884 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
fbfa2cc5 885
09cbfeaf
KS		 886 off_start = offset & (PAGE_SIZE - 1);
887 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5
JK		 888
889 if (pg_start == pg_end) {
6394328a 890 ret = fill_zero(inode, pg_start, off_start,
fbfa2cc5 891 off_end - off_start);
6394328a
CY		 892 if (ret)
893 return ret;
fbfa2cc5 894 } else {
6394328a
CY		 895 if (off_start) {
896 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 897 PAGE_SIZE - off_start);
6394328a
CY		 898 if (ret)
899 return ret;
900 }
901 if (off_end) {
902 ret = fill_zero(inode, pg_end, 0, off_end);
903 if (ret)
904 return ret;
905 }
fbfa2cc5
JK		 906
907 if (pg_start < pg_end) {
908 struct address_space *mapping = inode->i_mapping;
909 loff_t blk_start, blk_end;
4081363f 910 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127a3d4 911
2c4db1a6 912 f2fs_balance_fs(sbi, true);
fbfa2cc5 913
09cbfeaf
KS		 914 blk_start = (loff_t)pg_start << PAGE_SHIFT;
915 blk_end = (loff_t)pg_end << PAGE_SHIFT;
5a3a2d83 916 down_write(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5
JK		 917 truncate_inode_pages_range(mapping, blk_start,
918 blk_end - 1);
39936837 919
e479556b 920 f2fs_lock_op(sbi);
fbfa2cc5 921 ret = truncate_hole(inode, pg_start, pg_end);
e479556b 922 f2fs_unlock_op(sbi);
5a3a2d83 923 up_write(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5
JK		 924 }
925 }
926
fbfa2cc5
JK		 927 return ret;
928}
929
0a2aa8fb
JK		 930static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
931 int *do_replace, pgoff_t off, pgoff_t len)
b4ace337
CY		 932{
933 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
934 struct dnode_of_data dn;
0a2aa8fb 935 int ret, done, i;
ecbaa406 936
0a2aa8fb 937next_dnode:
6e2c64ad 938 set_new_dnode(&dn, inode, NULL, NULL, 0);
0a2aa8fb 939 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
6e2c64ad
JK		 940 if (ret && ret != -ENOENT) {
941 return ret;
942 } else if (ret == -ENOENT) {
0a2aa8fb
JK		 943 if (dn.max_level == 0)
944 return -ENOENT;
945 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
946 blkaddr += done;
947 do_replace += done;
948 goto next;
949 }
950
951 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
952 dn.ofs_in_node, len);
953 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
7a2af766
CY		 954 *blkaddr = datablock_addr(dn.inode,
955 dn.node_page, dn.ofs_in_node);
0a2aa8fb
JK		 956 if (!is_checkpointed_data(sbi, *blkaddr)) {
957
958 if (test_opt(sbi, LFS)) {
959 f2fs_put_dnode(&dn);
960 return -ENOTSUPP;
961 }
962
6e2c64ad 963 /* do not invalidate this block address */
f28b3434 964 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
0a2aa8fb 965 *do_replace = 1;
b4ace337 966 }
6e2c64ad 967 }
0a2aa8fb
JK		 968 f2fs_put_dnode(&dn);
969next:
970 len -= done;
971 off += done;
972 if (len)
973 goto next_dnode;
974 return 0;
975}
b4ace337 976
0a2aa8fb
JK		 977static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
978 int *do_replace, pgoff_t off, int len)
979{
980 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
981 struct dnode_of_data dn;
982 int ret, i;
b4ace337 983
0a2aa8fb
JK		 984 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
985 if (*do_replace == 0)
986 continue;
b4ace337 987
0a2aa8fb
JK		 988 set_new_dnode(&dn, inode, NULL, NULL, 0);
989 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
990 if (ret) {
991 dec_valid_block_count(sbi, inode, 1);
992 invalidate_blocks(sbi, *blkaddr);
993 } else {
994 f2fs_update_data_blkaddr(&dn, *blkaddr);
36abef4e 995 }
0a2aa8fb
JK		 996 f2fs_put_dnode(&dn);
997 }
998 return 0;
999}
1000
1001static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1002 block_t *blkaddr, int *do_replace,
1003 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1004{
1005 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1006 pgoff_t i = 0;
1007 int ret;
36abef4e 1008
0a2aa8fb
JK		 1009 while (i < len) {
1010 if (blkaddr[i] == NULL_ADDR && !full) {
1011 i++;
1012 continue;
6e2c64ad 1013 }
b4ace337 1014
0a2aa8fb
JK		 1015 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1016 struct dnode_of_data dn;
1017 struct node_info ni;
1018 size_t new_size;
1019 pgoff_t ilen;
b4ace337 1020
0a2aa8fb
JK		 1021 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1022 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1023 if (ret)
1024 return ret;
b4ace337 1025
0a2aa8fb
JK		 1026 get_node_info(sbi, dn.nid, &ni);
1027 ilen = min((pgoff_t)
1028 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1029 dn.ofs_in_node, len - i);
1030 do {
7a2af766
CY		 1031 dn.data_blkaddr = datablock_addr(dn.inode,
1032 dn.node_page, dn.ofs_in_node);
0a2aa8fb
JK		 1033 truncate_data_blocks_range(&dn, 1);
1034
1035 if (do_replace[i]) {
1036 f2fs_i_blocks_write(src_inode,
0abd675e 1037 1, false, false);
0a2aa8fb 1038 f2fs_i_blocks_write(dst_inode,
0abd675e 1039 1, true, false);
0a2aa8fb
JK		 1040 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1041 blkaddr[i], ni.version, true, false);
1042
1043 do_replace[i] = 0;
1044 }
1045 dn.ofs_in_node++;
1046 i++;
1047 new_size = (dst + i) << PAGE_SHIFT;
1048 if (dst_inode->i_size < new_size)
1049 f2fs_i_size_write(dst_inode, new_size);
e87f7329 1050 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
6e2c64ad 1051
0a2aa8fb
JK		 1052 f2fs_put_dnode(&dn);
1053 } else {
1054 struct page *psrc, *pdst;
1055
1056 psrc = get_lock_data_page(src_inode, src + i, true);
1057 if (IS_ERR(psrc))
1058 return PTR_ERR(psrc);
1059 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1060 true);
1061 if (IS_ERR(pdst)) {
1062 f2fs_put_page(psrc, 1);
1063 return PTR_ERR(pdst);
1064 }
1065 f2fs_copy_page(psrc, pdst);
1066 set_page_dirty(pdst);
1067 f2fs_put_page(pdst, 1);
6e2c64ad 1068 f2fs_put_page(psrc, 1);
b4ace337 1069
0a2aa8fb
JK		 1070 ret = truncate_hole(src_inode, src + i, src + i + 1);
1071 if (ret)
1072 return ret;
1073 i++;
1074 }
6e2c64ad
JK		 1075 }
1076 return 0;
0a2aa8fb 1077}
b4ace337 1078
0a2aa8fb
JK		 1079static int __exchange_data_block(struct inode *src_inode,
1080 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
363cad7f 1081 pgoff_t len, bool full)
0a2aa8fb
JK		 1082{
1083 block_t *src_blkaddr;
1084 int *do_replace;
363cad7f 1085 pgoff_t olen;
0a2aa8fb
JK		 1086 int ret;
1087
363cad7f
JK		 1088 while (len) {
1089 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
0a2aa8fb 1090
a7c3e901 1091 src_blkaddr = kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
363cad7f
JK		 1092 if (!src_blkaddr)
1093 return -ENOMEM;
0a2aa8fb 1094
a7c3e901 1095 do_replace = kvzalloc(sizeof(int) * olen, GFP_KERNEL);
363cad7f
JK		 1096 if (!do_replace) {
1097 kvfree(src_blkaddr);
1098 return -ENOMEM;
1099 }
0a2aa8fb 1100
363cad7f
JK		 1101 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1102 do_replace, src, olen);
1103 if (ret)
1104 goto roll_back;
0a2aa8fb 1105
363cad7f
JK		 1106 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1107 do_replace, src, dst, olen, full);
1108 if (ret)
1109 goto roll_back;
1110
1111 src += olen;
1112 dst += olen;
1113 len -= olen;
1114
1115 kvfree(src_blkaddr);
1116 kvfree(do_replace);
1117 }
0a2aa8fb
JK		 1118 return 0;
1119
1120roll_back:
1121 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1122 kvfree(src_blkaddr);
1123 kvfree(do_replace);
6e2c64ad
JK		 1124 return ret;
1125}
b4ace337 1126
6e2c64ad
JK		 1127static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1128{
1129 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1130 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
0a2aa8fb 1131 int ret;
6e2c64ad 1132
0a2aa8fb
JK		 1133 f2fs_balance_fs(sbi, true);
1134 f2fs_lock_op(sbi);
5f281fab
JK		 1135
1136 f2fs_drop_extent_tree(inode);
1137
0a2aa8fb
JK		 1138 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1139 f2fs_unlock_op(sbi);
b4ace337
CY		 1140 return ret;
1141}
1142
1143static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1144{
1145 pgoff_t pg_start, pg_end;
1146 loff_t new_size;
1147 int ret;
1148
b4ace337
CY		 1149 if (offset + len >= i_size_read(inode))
1150 return -EINVAL;
1151
1152 /* collapse range should be aligned to block size of f2fs. */
1153 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1154 return -EINVAL;
1155
b9d777b8
JK		 1156 ret = f2fs_convert_inline_inode(inode);
1157 if (ret)
1158 return ret;
97a7b2c2 1159
09cbfeaf
KS		 1160 pg_start = offset >> PAGE_SHIFT;
1161 pg_end = (offset + len) >> PAGE_SHIFT;
b4ace337 1162
5a3a2d83 1163 down_write(&F2FS_I(inode)->i_mmap_sem);
b4ace337
CY		 1164 /* write out all dirty pages from offset */
1165 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1166 if (ret)
5a3a2d83 1167 goto out;
b4ace337
CY		 1168
1169 truncate_pagecache(inode, offset);
1170
1171 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1172 if (ret)
5a3a2d83 1173 goto out;
b4ace337 1174
6e2c64ad
JK		 1175 /* write out all moved pages, if possible */
1176 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1177 truncate_pagecache(inode, offset);
1178
b4ace337 1179 new_size = i_size_read(inode) - len;
6e2c64ad 1180 truncate_pagecache(inode, new_size);
b4ace337
CY		 1181
1182 ret = truncate_blocks(inode, new_size, true);
1183 if (!ret)
fc9581c8 1184 f2fs_i_size_write(inode, new_size);
b4ace337 1185
5a3a2d83
QS		 1186out:
1187 up_write(&F2FS_I(inode)->i_mmap_sem);
b4ace337
CY		 1188 return ret;
1189}
1190
6e961949
CY		 1191static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1192 pgoff_t end)
1193{
1194 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1195 pgoff_t index = start;
1196 unsigned int ofs_in_node = dn->ofs_in_node;
1197 blkcnt_t count = 0;
1198 int ret;
1199
1200 for (; index < end; index++, dn->ofs_in_node++) {
7a2af766
CY		 1201 if (datablock_addr(dn->inode, dn->node_page,
1202 dn->ofs_in_node) == NULL_ADDR)
6e961949
CY		 1203 count++;
1204 }
1205
1206 dn->ofs_in_node = ofs_in_node;
1207 ret = reserve_new_blocks(dn, count);
1208 if (ret)
1209 return ret;
1210
1211 dn->ofs_in_node = ofs_in_node;
1212 for (index = start; index < end; index++, dn->ofs_in_node++) {
7a2af766
CY		 1213 dn->data_blkaddr = datablock_addr(dn->inode,
1214 dn->node_page, dn->ofs_in_node);
6e961949
CY		 1215 /*
1216 * reserve_new_blocks will not guarantee entire block
1217 * allocation.
1218 */
1219 if (dn->data_blkaddr == NULL_ADDR) {
1220 ret = -ENOSPC;
1221 break;
1222 }
1223 if (dn->data_blkaddr != NEW_ADDR) {
1224 invalidate_blocks(sbi, dn->data_blkaddr);
1225 dn->data_blkaddr = NEW_ADDR;
1226 set_data_blkaddr(dn);
1227 }
1228 }
1229
1230 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1231
1232 return ret;
1233}
1234
75cd4e09
CY		 1235static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1236 int mode)
1237{
1238 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1239 struct address_space *mapping = inode->i_mapping;
1240 pgoff_t index, pg_start, pg_end;
1241 loff_t new_size = i_size_read(inode);
1242 loff_t off_start, off_end;
1243 int ret = 0;
1244
75cd4e09
CY		 1245 ret = inode_newsize_ok(inode, (len + offset));
1246 if (ret)
1247 return ret;
1248
b9d777b8
JK		 1249 ret = f2fs_convert_inline_inode(inode);
1250 if (ret)
1251 return ret;
75cd4e09 1252
5a3a2d83 1253 down_write(&F2FS_I(inode)->i_mmap_sem);
75cd4e09
CY		 1254 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1255 if (ret)
5a3a2d83 1256 goto out_sem;
75cd4e09
CY		 1257
1258 truncate_pagecache_range(inode, offset, offset + len - 1);
1259
09cbfeaf
KS		 1260 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1261 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
75cd4e09 1262
09cbfeaf
KS		 1263 off_start = offset & (PAGE_SIZE - 1);
1264 off_end = (offset + len) & (PAGE_SIZE - 1);
75cd4e09
CY		 1265
1266 if (pg_start == pg_end) {
6394328a
CY		 1267 ret = fill_zero(inode, pg_start, off_start,
1268 off_end - off_start);
1269 if (ret)
5a3a2d83 1270 goto out_sem;
6394328a 1271
75cd4e09
CY		 1272 new_size = max_t(loff_t, new_size, offset + len);
1273 } else {
1274 if (off_start) {
6394328a 1275 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 1276 PAGE_SIZE - off_start);
6394328a 1277 if (ret)
5a3a2d83 1278 goto out_sem;
6394328a 1279
75cd4e09 1280 new_size = max_t(loff_t, new_size,
09cbfeaf 1281 (loff_t)pg_start << PAGE_SHIFT);
75cd4e09
CY		 1282 }
1283
6e961949 1284 for (index = pg_start; index < pg_end;) {
75cd4e09 1285 struct dnode_of_data dn;
6e961949
CY		 1286 unsigned int end_offset;
1287 pgoff_t end;
75cd4e09
CY		 1288
1289 f2fs_lock_op(sbi);
1290
6e961949
CY		 1291 set_new_dnode(&dn, inode, NULL, NULL, 0);
1292 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
75cd4e09
CY		 1293 if (ret) {
1294 f2fs_unlock_op(sbi);
1295 goto out;
1296 }
1297
6e961949
CY		 1298 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1299 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1300
1301 ret = f2fs_do_zero_range(&dn, index, end);
75cd4e09
CY		 1302 f2fs_put_dnode(&dn);
1303 f2fs_unlock_op(sbi);
9434fcde
CY		 1304
1305 f2fs_balance_fs(sbi, dn.node_changed);
1306
6e961949
CY		 1307 if (ret)
1308 goto out;
75cd4e09 1309
6e961949 1310 index = end;
75cd4e09 1311 new_size = max_t(loff_t, new_size,
6e961949 1312 (loff_t)index << PAGE_SHIFT);
75cd4e09
CY		 1313 }
1314
1315 if (off_end) {
6394328a
CY		 1316 ret = fill_zero(inode, pg_end, 0, off_end);
1317 if (ret)
1318 goto out;
1319
75cd4e09
CY		 1320 new_size = max_t(loff_t, new_size, offset + len);
1321 }
1322 }
1323
1324out:
ee6d182f 1325 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
fc9581c8 1326 f2fs_i_size_write(inode, new_size);
5a3a2d83
QS		 1327out_sem:
1328 up_write(&F2FS_I(inode)->i_mmap_sem);
75cd4e09
CY		 1329
1330 return ret;
1331}
1332
f62185d0
CY		 1333static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1334{
1335 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
0a2aa8fb 1336 pgoff_t nr, pg_start, pg_end, delta, idx;
f62185d0 1337 loff_t new_size;
6e2c64ad 1338 int ret = 0;
f62185d0 1339
f62185d0 1340 new_size = i_size_read(inode) + len;
46e82fb1
KM		 1341 ret = inode_newsize_ok(inode, new_size);
1342 if (ret)
1343 return ret;
f62185d0
CY		 1344
1345 if (offset >= i_size_read(inode))
1346 return -EINVAL;
1347
1348 /* insert range should be aligned to block size of f2fs. */
1349 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1350 return -EINVAL;
1351
b9d777b8
JK		 1352 ret = f2fs_convert_inline_inode(inode);
1353 if (ret)
1354 return ret;
97a7b2c2 1355
2c4db1a6 1356 f2fs_balance_fs(sbi, true);
2a340760 1357
5a3a2d83 1358 down_write(&F2FS_I(inode)->i_mmap_sem);
f62185d0
CY		 1359 ret = truncate_blocks(inode, i_size_read(inode), true);
1360 if (ret)
5a3a2d83 1361 goto out;
f62185d0
CY		 1362
1363 /* write out all dirty pages from offset */
1364 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1365 if (ret)
5a3a2d83 1366 goto out;
f62185d0
CY		 1367
1368 truncate_pagecache(inode, offset);
1369
09cbfeaf
KS		 1370 pg_start = offset >> PAGE_SHIFT;
1371 pg_end = (offset + len) >> PAGE_SHIFT;
f62185d0 1372 delta = pg_end - pg_start;
0a2aa8fb
JK		 1373 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1374
1375 while (!ret && idx > pg_start) {
1376 nr = idx - pg_start;
1377 if (nr > delta)
1378 nr = delta;
1379 idx -= nr;
f62185d0 1380
f62185d0 1381 f2fs_lock_op(sbi);
5f281fab
JK		 1382 f2fs_drop_extent_tree(inode);
1383
0a2aa8fb
JK		 1384 ret = __exchange_data_block(inode, inode, idx,
1385 idx + delta, nr, false);
f62185d0
CY		 1386 f2fs_unlock_op(sbi);
1387 }
1388
6e2c64ad
JK		 1389 /* write out all moved pages, if possible */
1390 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1391 truncate_pagecache(inode, offset);
1392
1393 if (!ret)
fc9581c8 1394 f2fs_i_size_write(inode, new_size);
5a3a2d83
QS		 1395out:
1396 up_write(&F2FS_I(inode)->i_mmap_sem);
f62185d0
CY		 1397 return ret;
1398}
1399
fbfa2cc5
JK		 1400static int expand_inode_data(struct inode *inode, loff_t offset,
1401 loff_t len, int mode)
1402{
4081363f 1403 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
e12dd7bd
JK		 1404 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1405 pgoff_t pg_end;
fbfa2cc5 1406 loff_t new_size = i_size_read(inode);
e12dd7bd 1407 loff_t off_end;
a7de6086 1408 int err;
fbfa2cc5 1409
a7de6086
JK		 1410 err = inode_newsize_ok(inode, (len + offset));
1411 if (err)
1412 return err;
fbfa2cc5 1413
a7de6086
JK		 1414 err = f2fs_convert_inline_inode(inode);
1415 if (err)
1416 return err;
9e09fc85 1417
2c4db1a6 1418 f2fs_balance_fs(sbi, true);
2a340760 1419
e12dd7bd 1420 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
09cbfeaf 1421 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5 1422
e12dd7bd
JK		 1423 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1424 map.m_len = pg_end - map.m_lblk;
1425 if (off_end)
1426 map.m_len++;
ead43275 1427
a7de6086
JK		 1428 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1429 if (err) {
e12dd7bd 1430 pgoff_t last_off;
fbfa2cc5 1431
e12dd7bd 1432 if (!map.m_len)
a7de6086 1433 return err;
98397ff3 1434
e12dd7bd
JK		 1435 last_off = map.m_lblk + map.m_len - 1;
1436
1437 /* update new size to the failed position */
1438 new_size = (last_off == pg_end) ? offset + len:
1439 (loff_t)(last_off + 1) << PAGE_SHIFT;
1440 } else {
1441 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
fbfa2cc5
JK		 1442 }
1443
ee6d182f 1444 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
fc9581c8 1445 f2fs_i_size_write(inode, new_size);
fbfa2cc5 1446
a7de6086 1447 return err;
fbfa2cc5
JK		 1448}
1449
1450static long f2fs_fallocate(struct file *file, int mode,
1451 loff_t offset, loff_t len)
1452{
6131ffaa 1453 struct inode *inode = file_inode(file);
587c0a42 1454 long ret = 0;
fbfa2cc5 1455
c998012b
CY		 1456 /* f2fs only support ->fallocate for regular file */
1457 if (!S_ISREG(inode->i_mode))
1458 return -EINVAL;
1459
f62185d0
CY		 1460 if (f2fs_encrypted_inode(inode) &&
1461 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
fcc85a4d
JK		 1462 return -EOPNOTSUPP;
1463
b4ace337 1464 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
f62185d0
CY		 1465 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1466 FALLOC_FL_INSERT_RANGE))
fbfa2cc5
JK		 1467 return -EOPNOTSUPP;
1468
5955102c 1469 inode_lock(inode);
3375f696 1470
587c0a42
TY		 1471 if (mode & FALLOC_FL_PUNCH_HOLE) {
1472 if (offset >= inode->i_size)
1473 goto out;
1474
a66c7b2f 1475 ret = punch_hole(inode, offset, len);
b4ace337
CY		 1476 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1477 ret = f2fs_collapse_range(inode, offset, len);
75cd4e09
CY		 1478 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1479 ret = f2fs_zero_range(inode, offset, len, mode);
f62185d0
CY		 1480 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1481 ret = f2fs_insert_range(inode, offset, len);
b4ace337 1482 } else {
fbfa2cc5 1483 ret = expand_inode_data(inode, offset, len, mode);
b4ace337 1484 }
fbfa2cc5 1485
3af60a49 1486 if (!ret) {
078cd827 1487 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 1488 f2fs_mark_inode_dirty_sync(inode, false);
26787236
JK		 1489 if (mode & FALLOC_FL_KEEP_SIZE)
1490 file_set_keep_isize(inode);
d0239e1b 1491 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3af60a49 1492 }
3375f696 1493
587c0a42 1494out:
5955102c 1495 inode_unlock(inode);
3375f696 1496
c01e2853 1497 trace_f2fs_fallocate(inode, mode, offset, len, ret);
fbfa2cc5
JK		 1498 return ret;
1499}
1500
1e84371f
JK		 1501static int f2fs_release_file(struct inode *inode, struct file *filp)
1502{
de5307e4
JK		 1503 /*
1504 * f2fs_relase_file is called at every close calls. So we should
1505 * not drop any inmemory pages by close called by other process.
1506 */
1507 if (!(filp->f_mode & FMODE_WRITE) ||
1508 atomic_read(&inode->i_writecount) != 1)
1509 return 0;
1510
1e84371f
JK		 1511 /* some remained atomic pages should discarded */
1512 if (f2fs_is_atomic_file(inode))
29b96b54 1513 drop_inmem_pages(inode);
1e84371f 1514 if (f2fs_is_volatile_file(inode)) {
91942321 1515 clear_inode_flag(inode, FI_VOLATILE_FILE);
648d50ba 1516 stat_dec_volatile_write(inode);
91942321 1517 set_inode_flag(inode, FI_DROP_CACHE);
1e84371f 1518 filemap_fdatawrite(inode->i_mapping);
91942321 1519 clear_inode_flag(inode, FI_DROP_CACHE);
1e84371f
JK		 1520 }
1521 return 0;
1522}
1523
7a10f017 1524static int f2fs_file_flush(struct file *file, fl_owner_t id)
fbfa2cc5 1525{
7a10f017
JK		 1526 struct inode *inode = file_inode(file);
1527
1528 /*
1529 * If the process doing a transaction is crashed, we should do
1530 * roll-back. Otherwise, other reader/write can see corrupted database
1531 * until all the writers close its file. Since this should be done
1532 * before dropping file lock, it needs to do in ->flush.
1533 */
1534 if (f2fs_is_atomic_file(inode) &&
1535 F2FS_I(inode)->inmem_task == current)
1536 drop_inmem_pages(inode);
1537 return 0;
fbfa2cc5
JK		 1538}
1539
52656e6c 1540static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
fbfa2cc5 1541{
6131ffaa 1542 struct inode *inode = file_inode(filp);
fbfa2cc5 1543 struct f2fs_inode_info *fi = F2FS_I(inode);
2c1d0305
CY		 1544 unsigned int flags = fi->i_flags &
1545 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
52656e6c
JK		 1546 return put_user(flags, (int __user *)arg);
1547}
fbfa2cc5 1548
2c1d0305
CY		 1549static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1550{
1551 struct f2fs_inode_info *fi = F2FS_I(inode);
1552 unsigned int oldflags;
1553
1554 /* Is it quota file? Do not allow user to mess with it */
1555 if (IS_NOQUOTA(inode))
1556 return -EPERM;
1557
1558 flags = f2fs_mask_flags(inode->i_mode, flags);
1559
1560 oldflags = fi->i_flags;
1561
1562 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1563 if (!capable(CAP_LINUX_IMMUTABLE))
1564 return -EPERM;
1565
1566 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1567 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1568 fi->i_flags = flags;
1569
1570 if (fi->i_flags & FS_PROJINHERIT_FL)
1571 set_inode_flag(inode, FI_PROJ_INHERIT);
1572 else
1573 clear_inode_flag(inode, FI_PROJ_INHERIT);
1574
1575 inode->i_ctime = current_time(inode);
1576 f2fs_set_inode_flags(inode);
1577 f2fs_mark_inode_dirty_sync(inode, false);
1578 return 0;
1579}
1580
52656e6c
JK		 1581static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1582{
1583 struct inode *inode = file_inode(filp);
69494229 1584 unsigned int flags;
52656e6c 1585 int ret;
fbfa2cc5 1586
7fb17fe4
CY		 1587 if (!inode_owner_or_capable(inode))
1588 return -EACCES;
1589
1590 if (get_user(flags, (int __user *)arg))
1591 return -EFAULT;
1592
52656e6c
JK		 1593 ret = mnt_want_write_file(filp);
1594 if (ret)
1595 return ret;
fbfa2cc5 1596
5955102c 1597 inode_lock(inode);
fbfa2cc5 1598
2c1d0305 1599 ret = __f2fs_ioc_setflags(inode, flags);
fbfa2cc5 1600
a72d4b97 1601 inode_unlock(inode);
52656e6c
JK		 1602 mnt_drop_write_file(filp);
1603 return ret;
1604}
4b2fecc8 1605
d49f3e89
CY		 1606static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1607{
1608 struct inode *inode = file_inode(filp);
1609
1610 return put_user(inode->i_generation, (int __user *)arg);
1611}
1612
88b88a66
JK		 1613static int f2fs_ioc_start_atomic_write(struct file *filp)
1614{
1615 struct inode *inode = file_inode(filp);
f4c9c743 1616 int ret;
88b88a66
JK		 1617
1618 if (!inode_owner_or_capable(inode))
1619 return -EACCES;
1620
e811898c
JK		 1621 if (!S_ISREG(inode->i_mode))
1622 return -EINVAL;
1623
7fb17fe4
CY		 1624 ret = mnt_want_write_file(filp);
1625 if (ret)
1626 return ret;
1627
0fac558b
CY		 1628 inode_lock(inode);
1629
1e84371f 1630 if (f2fs_is_atomic_file(inode))
7fb17fe4 1631 goto out;
88b88a66 1632
f4c9c743
CY		 1633 ret = f2fs_convert_inline_inode(inode);
1634 if (ret)
7fb17fe4 1635 goto out;
88b88a66 1636
91942321 1637 set_inode_flag(inode, FI_ATOMIC_FILE);
6c3acd97 1638 set_inode_flag(inode, FI_HOT_DATA);
d0239e1b
JK		 1639 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1640
c27753d6 1641 if (!get_dirty_pages(inode))
684ca7e5 1642 goto inc_stat;
c27753d6
JK		 1643
1644 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
204706c7 1645 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
c27753d6
JK		 1646 inode->i_ino, get_dirty_pages(inode));
1647 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
684ca7e5 1648 if (ret) {
91942321 1649 clear_inode_flag(inode, FI_ATOMIC_FILE);
84a23fbe 1650 clear_inode_flag(inode, FI_HOT_DATA);
684ca7e5
KM		 1651 goto out;
1652 }
1653
1654inc_stat:
7a10f017 1655 F2FS_I(inode)->inmem_task = current;
26a28a0c
JK		 1656 stat_inc_atomic_write(inode);
1657 stat_update_max_atomic_write(inode);
684ca7e5 1658out:
0fac558b 1659 inode_unlock(inode);
7fb17fe4 1660 mnt_drop_write_file(filp);
c27753d6 1661 return ret;
88b88a66
JK		 1662}
1663
1664static int f2fs_ioc_commit_atomic_write(struct file *filp)
1665{
1666 struct inode *inode = file_inode(filp);
1667 int ret;
1668
1669 if (!inode_owner_or_capable(inode))
1670 return -EACCES;
1671
1672 ret = mnt_want_write_file(filp);
1673 if (ret)
1674 return ret;
1675
0fac558b
CY		 1676 inode_lock(inode);
1677
7fb17fe4
CY		 1678 if (f2fs_is_volatile_file(inode))
1679 goto err_out;
1680
6282adbf 1681 if (f2fs_is_atomic_file(inode)) {
29b96b54 1682 ret = commit_inmem_pages(inode);
5fe45743 1683 if (ret)
edb27dee 1684 goto err_out;
5fe45743 1685
26a28a0c 1686 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
5fe45743
CY		 1687 if (!ret) {
1688 clear_inode_flag(inode, FI_ATOMIC_FILE);
84a23fbe 1689 clear_inode_flag(inode, FI_HOT_DATA);
5fe45743 1690 stat_dec_atomic_write(inode);
447135a8 1691 }
26a28a0c 1692 } else {
774e1b78 1693 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
6282adbf 1694 }
edb27dee 1695err_out:
0fac558b 1696 inode_unlock(inode);
88b88a66
JK		 1697 mnt_drop_write_file(filp);
1698 return ret;
1699}
1700
02a1335f
JK		 1701static int f2fs_ioc_start_volatile_write(struct file *filp)
1702{
1703 struct inode *inode = file_inode(filp);
f4c9c743 1704 int ret;
02a1335f
JK		 1705
1706 if (!inode_owner_or_capable(inode))
1707 return -EACCES;
1708
8ff0971f
CY		 1709 if (!S_ISREG(inode->i_mode))
1710 return -EINVAL;
1711
7fb17fe4
CY		 1712 ret = mnt_want_write_file(filp);
1713 if (ret)
1714 return ret;
1715
0fac558b
CY		 1716 inode_lock(inode);
1717
1e84371f 1718 if (f2fs_is_volatile_file(inode))
7fb17fe4 1719 goto out;
1e84371f 1720
f4c9c743
CY		 1721 ret = f2fs_convert_inline_inode(inode);
1722 if (ret)
7fb17fe4 1723 goto out;
b3d208f9 1724
648d50ba
CY		 1725 stat_inc_volatile_write(inode);
1726 stat_update_max_volatile_write(inode);
1727
91942321 1728 set_inode_flag(inode, FI_VOLATILE_FILE);
d0239e1b 1729 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 1730out:
0fac558b 1731 inode_unlock(inode);
7fb17fe4
CY		 1732 mnt_drop_write_file(filp);
1733 return ret;
02a1335f
JK		 1734}
1735
1e84371f
JK		 1736static int f2fs_ioc_release_volatile_write(struct file *filp)
1737{
1738 struct inode *inode = file_inode(filp);
7fb17fe4 1739 int ret;
1e84371f
JK		 1740
1741 if (!inode_owner_or_capable(inode))
1742 return -EACCES;
1743
7fb17fe4
CY		 1744 ret = mnt_want_write_file(filp);
1745 if (ret)
1746 return ret;
1747
0fac558b
CY		 1748 inode_lock(inode);
1749
1e84371f 1750 if (!f2fs_is_volatile_file(inode))
7fb17fe4 1751 goto out;
1e84371f 1752
7fb17fe4
CY		 1753 if (!f2fs_is_first_block_written(inode)) {
1754 ret = truncate_partial_data_page(inode, 0, true);
1755 goto out;
1756 }
3c6c2beb 1757
7fb17fe4
CY		 1758 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1759out:
0fac558b 1760 inode_unlock(inode);
7fb17fe4
CY		 1761 mnt_drop_write_file(filp);
1762 return ret;
1e84371f
JK		 1763}
1764
1765static int f2fs_ioc_abort_volatile_write(struct file *filp)
1766{
1767 struct inode *inode = file_inode(filp);
1768 int ret;
1769
1770 if (!inode_owner_or_capable(inode))
1771 return -EACCES;
1772
1773 ret = mnt_want_write_file(filp);
1774 if (ret)
1775 return ret;
1776
0fac558b
CY		 1777 inode_lock(inode);
1778
26dc3d44 1779 if (f2fs_is_atomic_file(inode))
29b96b54 1780 drop_inmem_pages(inode);
732d5648 1781 if (f2fs_is_volatile_file(inode)) {
91942321 1782 clear_inode_flag(inode, FI_VOLATILE_FILE);
648d50ba 1783 stat_dec_volatile_write(inode);
608514de 1784 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
732d5648 1785 }
de6a8ec9 1786
0fac558b
CY		 1787 inode_unlock(inode);
1788
1e84371f 1789 mnt_drop_write_file(filp);
d0239e1b 1790 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1e84371f
JK		 1791 return ret;
1792}
1793
1abff93d
JK		 1794static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1795{
1796 struct inode *inode = file_inode(filp);
1797 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1798 struct super_block *sb = sbi->sb;
1799 __u32 in;
7fb17fe4 1800 int ret;
1abff93d
JK		 1801
1802 if (!capable(CAP_SYS_ADMIN))
1803 return -EPERM;
1804
1805 if (get_user(in, (__u32 __user *)arg))
1806 return -EFAULT;
1807
7fb17fe4
CY		 1808 ret = mnt_want_write_file(filp);
1809 if (ret)
1810 return ret;
1811
1abff93d
JK		 1812 switch (in) {
1813 case F2FS_GOING_DOWN_FULLSYNC:
1814 sb = freeze_bdev(sb->s_bdev);
1815 if (sb && !IS_ERR(sb)) {
38f91ca8 1816 f2fs_stop_checkpoint(sbi, false);
1abff93d
JK		 1817 thaw_bdev(sb->s_bdev, sb);
1818 }
1819 break;
1820 case F2FS_GOING_DOWN_METASYNC:
1821 /* do checkpoint only */
1822 f2fs_sync_fs(sb, 1);
38f91ca8 1823 f2fs_stop_checkpoint(sbi, false);
1abff93d
JK		 1824 break;
1825 case F2FS_GOING_DOWN_NOSYNC:
38f91ca8 1826 f2fs_stop_checkpoint(sbi, false);
1abff93d 1827 break;
c912a829 1828 case F2FS_GOING_DOWN_METAFLUSH:
b0af6d49 1829 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
38f91ca8 1830 f2fs_stop_checkpoint(sbi, false);
c912a829 1831 break;
1abff93d 1832 default:
7fb17fe4
CY		 1833 ret = -EINVAL;
1834 goto out;
1abff93d 1835 }
d0239e1b 1836 f2fs_update_time(sbi, REQ_TIME);
7fb17fe4
CY		 1837out:
1838 mnt_drop_write_file(filp);
1839 return ret;
1abff93d
JK		 1840}
1841
52656e6c
JK		 1842static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1843{
1844 struct inode *inode = file_inode(filp);
1845 struct super_block *sb = inode->i_sb;
1846 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1847 struct fstrim_range range;
1848 int ret;
4b2fecc8 1849
52656e6c
JK		 1850 if (!capable(CAP_SYS_ADMIN))
1851 return -EPERM;
4b2fecc8 1852
52656e6c
JK		 1853 if (!blk_queue_discard(q))
1854 return -EOPNOTSUPP;
4b2fecc8 1855
52656e6c
JK		 1856 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1857 sizeof(range)))
1858 return -EFAULT;
4b2fecc8 1859
7fb17fe4
CY		 1860 ret = mnt_want_write_file(filp);
1861 if (ret)
1862 return ret;
1863
52656e6c
JK		 1864 range.minlen = max((unsigned int)range.minlen,
1865 q->limits.discard_granularity);
1866 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
7fb17fe4 1867 mnt_drop_write_file(filp);
52656e6c
JK		 1868 if (ret < 0)
1869 return ret;
4b2fecc8 1870
52656e6c
JK		 1871 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1872 sizeof(range)))
1873 return -EFAULT;
d0239e1b 1874 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
52656e6c
JK		 1875 return 0;
1876}
1877
f424f664
JK		 1878static bool uuid_is_nonzero(__u8 u[16])
1879{
1880 int i;
1881
1882 for (i = 0; i < 16; i++)
1883 if (u[i])
1884 return true;
1885 return false;
1886}
1887
1888static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1889{
f424f664
JK		 1890 struct inode *inode = file_inode(filp);
1891
d0239e1b 1892 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 1893
db717d8e 1894 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
f424f664
JK		 1895}
1896
1897static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1898{
db717d8e 1899 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
f424f664
JK		 1900}
1901
1902static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1903{
1904 struct inode *inode = file_inode(filp);
1905 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1906 int err;
1907
1908 if (!f2fs_sb_has_crypto(inode->i_sb))
1909 return -EOPNOTSUPP;
1910
1911 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1912 goto got_it;
1913
1914 err = mnt_want_write_file(filp);
1915 if (err)
1916 return err;
1917
1918 /* update superblock with uuid */
1919 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1920
c5bda1c8 1921 err = f2fs_commit_super(sbi, false);
f424f664
JK		 1922 if (err) {
1923 /* undo new data */
1924 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
e8240f65 1925 mnt_drop_write_file(filp);
f424f664
JK		 1926 return err;
1927 }
e8240f65 1928 mnt_drop_write_file(filp);
f424f664
JK		 1929got_it:
1930 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1931 16))
1932 return -EFAULT;
1933 return 0;
1934}
1935
c1c1b583
CY		 1936static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1937{
1938 struct inode *inode = file_inode(filp);
1939 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d530d4d8 1940 __u32 sync;
7fb17fe4 1941 int ret;
c1c1b583
CY		 1942
1943 if (!capable(CAP_SYS_ADMIN))
1944 return -EPERM;
1945
d530d4d8 1946 if (get_user(sync, (__u32 __user *)arg))
c1c1b583
CY		 1947 return -EFAULT;
1948
d530d4d8
CY		 1949 if (f2fs_readonly(sbi->sb))
1950 return -EROFS;
c1c1b583 1951
7fb17fe4
CY		 1952 ret = mnt_want_write_file(filp);
1953 if (ret)
1954 return ret;
1955
d530d4d8 1956 if (!sync) {
7fb17fe4
CY		 1957 if (!mutex_trylock(&sbi->gc_mutex)) {
1958 ret = -EBUSY;
1959 goto out;
1960 }
d530d4d8
CY		 1961 } else {
1962 mutex_lock(&sbi->gc_mutex);
c1c1b583
CY		 1963 }
1964
e066b83c 1965 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
7fb17fe4
CY		 1966out:
1967 mnt_drop_write_file(filp);
1968 return ret;
c1c1b583
CY		 1969}
1970
34dc77ad
JK		 1971static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
1972{
1973 struct inode *inode = file_inode(filp);
1974 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1975 struct f2fs_gc_range range;
1976 u64 end;
1977 int ret;
1978
1979 if (!capable(CAP_SYS_ADMIN))
1980 return -EPERM;
1981
1982 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
1983 sizeof(range)))
1984 return -EFAULT;
1985
1986 if (f2fs_readonly(sbi->sb))
1987 return -EROFS;
1988
1989 ret = mnt_want_write_file(filp);
1990 if (ret)
1991 return ret;
1992
1993 end = range.start + range.len;
1994 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
1995 return -EINVAL;
1996do_more:
1997 if (!range.sync) {
1998 if (!mutex_trylock(&sbi->gc_mutex)) {
1999 ret = -EBUSY;
2000 goto out;
2001 }
2002 } else {
2003 mutex_lock(&sbi->gc_mutex);
2004 }
2005
2006 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2007 range.start += sbi->blocks_per_seg;
2008 if (range.start <= end)
2009 goto do_more;
2010out:
2011 mnt_drop_write_file(filp);
2012 return ret;
2013}
2014
456b88e4
CY		 2015static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2016{
2017 struct inode *inode = file_inode(filp);
2018 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
7fb17fe4 2019 int ret;
456b88e4
CY		 2020
2021 if (!capable(CAP_SYS_ADMIN))
2022 return -EPERM;
2023
2024 if (f2fs_readonly(sbi->sb))
2025 return -EROFS;
2026
7fb17fe4
CY		 2027 ret = mnt_want_write_file(filp);
2028 if (ret)
2029 return ret;
2030
2031 ret = f2fs_sync_fs(sbi->sb, 1);
2032
2033 mnt_drop_write_file(filp);
2034 return ret;
456b88e4
CY		 2035}
2036
d323d005
CY		 2037static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2038 struct file *filp,
2039 struct f2fs_defragment *range)
2040{
2041 struct inode *inode = file_inode(filp);
da85985c 2042 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
e15882b6 2043 struct extent_info ei = {0,0,0};
d323d005 2044 pgoff_t pg_start, pg_end;
3519e3f9 2045 unsigned int blk_per_seg = sbi->blocks_per_seg;
d323d005 2046 unsigned int total = 0, sec_num;
d323d005
CY		 2047 block_t blk_end = 0;
2048 bool fragmented = false;
2049 int err;
2050
2051 /* if in-place-update policy is enabled, don't waste time here */
7eab0c0d 2052 if (need_inplace_update_policy(inode, NULL))
d323d005
CY		 2053 return -EINVAL;
2054
09cbfeaf
KS		 2055 pg_start = range->start >> PAGE_SHIFT;
2056 pg_end = (range->start + range->len) >> PAGE_SHIFT;
d323d005 2057
2c4db1a6 2058 f2fs_balance_fs(sbi, true);
d323d005 2059
5955102c 2060 inode_lock(inode);
d323d005
CY		 2061
2062 /* writeback all dirty pages in the range */
2063 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
d8fe4f0e 2064 range->start + range->len - 1);
d323d005
CY		 2065 if (err)
2066 goto out;
2067
2068 /*
2069 * lookup mapping info in extent cache, skip defragmenting if physical
2070 * block addresses are continuous.
2071 */
2072 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2073 if (ei.fofs + ei.len >= pg_end)
2074 goto out;
2075 }
2076
2077 map.m_lblk = pg_start;
d323d005
CY		 2078
2079 /*
2080 * lookup mapping info in dnode page cache, skip defragmenting if all
2081 * physical block addresses are continuous even if there are hole(s)
2082 * in logical blocks.
2083 */
2084 while (map.m_lblk < pg_end) {
a1c1e9b7 2085 map.m_len = pg_end - map.m_lblk;
f2220c7f 2086 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2087 if (err)
2088 goto out;
2089
2090 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2091 map.m_lblk++;
d323d005
CY		 2092 continue;
2093 }
2094
2095 if (blk_end && blk_end != map.m_pblk) {
2096 fragmented = true;
2097 break;
2098 }
2099 blk_end = map.m_pblk + map.m_len;
2100
2101 map.m_lblk += map.m_len;
d323d005
CY		 2102 }
2103
2104 if (!fragmented)
2105 goto out;
2106
2107 map.m_lblk = pg_start;
2108 map.m_len = pg_end - pg_start;
2109
4ddb1a4d 2110 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
d323d005
CY		 2111
2112 /*
2113 * make sure there are enough free section for LFS allocation, this can
2114 * avoid defragment running in SSR mode when free section are allocated
2115 * intensively
2116 */
7f3037a5 2117 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
d323d005
CY		 2118 err = -EAGAIN;
2119 goto out;
2120 }
2121
2122 while (map.m_lblk < pg_end) {
2123 pgoff_t idx;
2124 int cnt = 0;
2125
2126do_map:
a1c1e9b7 2127 map.m_len = pg_end - map.m_lblk;
f2220c7f 2128 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2129 if (err)
2130 goto clear_out;
2131
2132 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2133 map.m_lblk++;
2134 continue;
2135 }
2136
91942321 2137 set_inode_flag(inode, FI_DO_DEFRAG);
d323d005
CY		 2138
2139 idx = map.m_lblk;
2140 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2141 struct page *page;
2142
2143 page = get_lock_data_page(inode, idx, true);
2144 if (IS_ERR(page)) {
2145 err = PTR_ERR(page);
2146 goto clear_out;
2147 }
2148
2149 set_page_dirty(page);
2150 f2fs_put_page(page, 1);
2151
2152 idx++;
2153 cnt++;
2154 total++;
2155 }
2156
2157 map.m_lblk = idx;
d323d005
CY		 2158
2159 if (idx < pg_end && cnt < blk_per_seg)
2160 goto do_map;
2161
91942321 2162 clear_inode_flag(inode, FI_DO_DEFRAG);
d323d005
CY		 2163
2164 err = filemap_fdatawrite(inode->i_mapping);
2165 if (err)
2166 goto out;
2167 }
2168clear_out:
91942321 2169 clear_inode_flag(inode, FI_DO_DEFRAG);
d323d005 2170out:
5955102c 2171 inode_unlock(inode);
d323d005 2172 if (!err)
09cbfeaf 2173 range->len = (u64)total << PAGE_SHIFT;
d323d005
CY		 2174 return err;
2175}
2176
2177static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2178{
2179 struct inode *inode = file_inode(filp);
2180 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2181 struct f2fs_defragment range;
2182 int err;
2183
2184 if (!capable(CAP_SYS_ADMIN))
2185 return -EPERM;
2186
7eab0c0d 2187 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
d323d005
CY		 2188 return -EINVAL;
2189
d7563861
KM		 2190 if (f2fs_readonly(sbi->sb))
2191 return -EROFS;
d323d005
CY		 2192
2193 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
d7563861
KM		 2194 sizeof(range)))
2195 return -EFAULT;
d323d005
CY		 2196
2197 /* verify alignment of offset & size */
d7563861
KM		 2198 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2199 return -EINVAL;
d323d005 2200
1941d7bc 2201 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
d7563861
KM		 2202 sbi->max_file_blocks))
2203 return -EINVAL;
2204
2205 err = mnt_want_write_file(filp);
2206 if (err)
2207 return err;
1941d7bc 2208
d323d005 2209 err = f2fs_defragment_range(sbi, filp, &range);
d7563861
KM		 2210 mnt_drop_write_file(filp);
2211
d0239e1b 2212 f2fs_update_time(sbi, REQ_TIME);
d323d005 2213 if (err < 0)
d7563861 2214 return err;
d323d005
CY		 2215
2216 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2217 sizeof(range)))
d7563861
KM		 2218 return -EFAULT;
2219
2220 return 0;
d323d005
CY		 2221}
2222
4dd6f977
JK		 2223static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2224 struct file *file_out, loff_t pos_out, size_t len)
2225{
2226 struct inode *src = file_inode(file_in);
2227 struct inode *dst = file_inode(file_out);
2228 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2229 size_t olen = len, dst_max_i_size = 0;
2230 size_t dst_osize;
2231 int ret;
2232
2233 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2234 src->i_sb != dst->i_sb)
2235 return -EXDEV;
2236
2237 if (unlikely(f2fs_readonly(src->i_sb)))
2238 return -EROFS;
2239
fe8494bf
CY		 2240 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2241 return -EINVAL;
4dd6f977
JK		 2242
2243 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2244 return -EOPNOTSUPP;
2245
d95fd91c
FL		 2246 if (src == dst) {
2247 if (pos_in == pos_out)
2248 return 0;
2249 if (pos_out > pos_in && pos_out < pos_in + len)
2250 return -EINVAL;
2251 }
2252
4dd6f977 2253 inode_lock(src);
20a3d61d
CY		 2254 if (src != dst) {
2255 if (!inode_trylock(dst)) {
2256 ret = -EBUSY;
2257 goto out;
2258 }
2259 }
4dd6f977
JK		 2260
2261 ret = -EINVAL;
2262 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2263 goto out_unlock;
2264 if (len == 0)
2265 olen = len = src->i_size - pos_in;
2266 if (pos_in + len == src->i_size)
2267 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2268 if (len == 0) {
2269 ret = 0;
2270 goto out_unlock;
2271 }
2272
2273 dst_osize = dst->i_size;
2274 if (pos_out + olen > dst->i_size)
2275 dst_max_i_size = pos_out + olen;
2276
2277 /* verify the end result is block aligned */
2278 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2279 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2280 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2281 goto out_unlock;
2282
2283 ret = f2fs_convert_inline_inode(src);
2284 if (ret)
2285 goto out_unlock;
2286
2287 ret = f2fs_convert_inline_inode(dst);
2288 if (ret)
2289 goto out_unlock;
2290
2291 /* write out all dirty pages from offset */
2292 ret = filemap_write_and_wait_range(src->i_mapping,
2293 pos_in, pos_in + len);
2294 if (ret)
2295 goto out_unlock;
2296
2297 ret = filemap_write_and_wait_range(dst->i_mapping,
2298 pos_out, pos_out + len);
2299 if (ret)
2300 goto out_unlock;
2301
2302 f2fs_balance_fs(sbi, true);
2303 f2fs_lock_op(sbi);
61e4da11
FL		 2304 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2305 pos_out >> F2FS_BLKSIZE_BITS,
2306 len >> F2FS_BLKSIZE_BITS, false);
4dd6f977
JK		 2307
2308 if (!ret) {
2309 if (dst_max_i_size)
2310 f2fs_i_size_write(dst, dst_max_i_size);
2311 else if (dst_osize != dst->i_size)
2312 f2fs_i_size_write(dst, dst_osize);
2313 }
2314 f2fs_unlock_op(sbi);
2315out_unlock:
2316 if (src != dst)
2317 inode_unlock(dst);
20a3d61d 2318out:
4dd6f977
JK		 2319 inode_unlock(src);
2320 return ret;
2321}
2322
2323static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2324{
2325 struct f2fs_move_range range;
2326 struct fd dst;
2327 int err;
2328
2329 if (!(filp->f_mode & FMODE_READ) ||
2330 !(filp->f_mode & FMODE_WRITE))
2331 return -EBADF;
2332
2333 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2334 sizeof(range)))
2335 return -EFAULT;
2336
2337 dst = fdget(range.dst_fd);
2338 if (!dst.file)
2339 return -EBADF;
2340
2341 if (!(dst.file->f_mode & FMODE_WRITE)) {
2342 err = -EBADF;
2343 goto err_out;
2344 }
2345
2346 err = mnt_want_write_file(filp);
2347 if (err)
2348 goto err_out;
2349
2350 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2351 range.pos_out, range.len);
2352
2353 mnt_drop_write_file(filp);
3cecfa5f
KM		 2354 if (err)
2355 goto err_out;
4dd6f977
JK		 2356
2357 if (copy_to_user((struct f2fs_move_range __user *)arg,
2358 &range, sizeof(range)))
2359 err = -EFAULT;
2360err_out:
2361 fdput(dst);
2362 return err;
2363}
2364
e066b83c
JK		 2365static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2366{
2367 struct inode *inode = file_inode(filp);
2368 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2369 struct sit_info *sm = SIT_I(sbi);
2370 unsigned int start_segno = 0, end_segno = 0;
2371 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2372 struct f2fs_flush_device range;
2373 int ret;
2374
2375 if (!capable(CAP_SYS_ADMIN))
2376 return -EPERM;
2377
2378 if (f2fs_readonly(sbi->sb))
2379 return -EROFS;
2380
2381 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2382 sizeof(range)))
2383 return -EFAULT;
2384
2385 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2386 sbi->segs_per_sec != 1) {
2387 f2fs_msg(sbi->sb, KERN_WARNING,
2388 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2389 range.dev_num, sbi->s_ndevs,
2390 sbi->segs_per_sec);
2391 return -EINVAL;
2392 }
2393
2394 ret = mnt_want_write_file(filp);
2395 if (ret)
2396 return ret;
2397
2398 if (range.dev_num != 0)
2399 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2400 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2401
2402 start_segno = sm->last_victim[FLUSH_DEVICE];
2403 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2404 start_segno = dev_start_segno;
2405 end_segno = min(start_segno + range.segments, dev_end_segno);
2406
2407 while (start_segno < end_segno) {
2408 if (!mutex_trylock(&sbi->gc_mutex)) {
2409 ret = -EBUSY;
2410 goto out;
2411 }
2412 sm->last_victim[GC_CB] = end_segno + 1;
2413 sm->last_victim[GC_GREEDY] = end_segno + 1;
2414 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2415 ret = f2fs_gc(sbi, true, true, start_segno);
2416 if (ret == -EAGAIN)
2417 ret = 0;
2418 else if (ret < 0)
2419 break;
2420 start_segno++;
2421 }
2422out:
2423 mnt_drop_write_file(filp);
2424 return ret;
2425}
2426
e65ef207
JK		 2427static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2428{
2429 struct inode *inode = file_inode(filp);
2430 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2431
2432 /* Must validate to set it with SQLite behavior in Android. */
2433 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2434
2435 return put_user(sb_feature, (u32 __user *)arg);
2436}
e066b83c 2437
2c1d0305
CY		 2438#ifdef CONFIG_QUOTA
2439static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2440{
2441 struct inode *inode = file_inode(filp);
2442 struct f2fs_inode_info *fi = F2FS_I(inode);
2443 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2444 struct super_block *sb = sbi->sb;
2445 struct dquot *transfer_to[MAXQUOTAS] = {};
2446 struct page *ipage;
2447 kprojid_t kprojid;
2448 int err;
2449
2450 if (!f2fs_sb_has_project_quota(sb)) {
2451 if (projid != F2FS_DEF_PROJID)
2452 return -EOPNOTSUPP;
2453 else
2454 return 0;
2455 }
2456
2457 if (!f2fs_has_extra_attr(inode))
2458 return -EOPNOTSUPP;
2459
2460 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2461
2462 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2463 return 0;
2464
2465 err = mnt_want_write_file(filp);
2466 if (err)
2467 return err;
2468
2469 err = -EPERM;
2470 inode_lock(inode);
2471
2472 /* Is it quota file? Do not allow user to mess with it */
2473 if (IS_NOQUOTA(inode))
2474 goto out_unlock;
2475
2476 ipage = get_node_page(sbi, inode->i_ino);
2477 if (IS_ERR(ipage)) {
2478 err = PTR_ERR(ipage);
2479 goto out_unlock;
2480 }
2481
2482 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2483 i_projid)) {
2484 err = -EOVERFLOW;
2485 f2fs_put_page(ipage, 1);
2486 goto out_unlock;
2487 }
2488 f2fs_put_page(ipage, 1);
2489
2490 dquot_initialize(inode);
2491
2492 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2493 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2494 err = __dquot_transfer(inode, transfer_to);
2495 dqput(transfer_to[PRJQUOTA]);
2496 if (err)
2497 goto out_dirty;
2498 }
2499
2500 F2FS_I(inode)->i_projid = kprojid;
2501 inode->i_ctime = current_time(inode);
2502out_dirty:
2503 f2fs_mark_inode_dirty_sync(inode, true);
2504out_unlock:
2505 inode_unlock(inode);
2506 mnt_drop_write_file(filp);
2507 return err;
2508}
2509#else
2510static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2511{
2512 if (projid != F2FS_DEF_PROJID)
2513 return -EOPNOTSUPP;
2514 return 0;
2515}
2516#endif
2517
2518/* Transfer internal flags to xflags */
2519static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2520{
2521 __u32 xflags = 0;
2522
2523 if (iflags & FS_SYNC_FL)
2524 xflags |= FS_XFLAG_SYNC;
2525 if (iflags & FS_IMMUTABLE_FL)
2526 xflags |= FS_XFLAG_IMMUTABLE;
2527 if (iflags & FS_APPEND_FL)
2528 xflags |= FS_XFLAG_APPEND;
2529 if (iflags & FS_NODUMP_FL)
2530 xflags |= FS_XFLAG_NODUMP;
2531 if (iflags & FS_NOATIME_FL)
2532 xflags |= FS_XFLAG_NOATIME;
2533 if (iflags & FS_PROJINHERIT_FL)
2534 xflags |= FS_XFLAG_PROJINHERIT;
2535 return xflags;
2536}
2537
2538#define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2539 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2540 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2541
2542/* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2543#define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2544 FS_IMMUTABLE_FL | \
2545 FS_APPEND_FL | \
2546 FS_NODUMP_FL | \
2547 FS_NOATIME_FL | \
2548 FS_PROJINHERIT_FL)
2549
2550/* Transfer xflags flags to internal */
2551static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2552{
2553 unsigned long iflags = 0;
2554
2555 if (xflags & FS_XFLAG_SYNC)
2556 iflags |= FS_SYNC_FL;
2557 if (xflags & FS_XFLAG_IMMUTABLE)
2558 iflags |= FS_IMMUTABLE_FL;
2559 if (xflags & FS_XFLAG_APPEND)
2560 iflags |= FS_APPEND_FL;
2561 if (xflags & FS_XFLAG_NODUMP)
2562 iflags |= FS_NODUMP_FL;
2563 if (xflags & FS_XFLAG_NOATIME)
2564 iflags |= FS_NOATIME_FL;
2565 if (xflags & FS_XFLAG_PROJINHERIT)
2566 iflags |= FS_PROJINHERIT_FL;
2567
2568 return iflags;
2569}
2570
2571static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2572{
2573 struct inode *inode = file_inode(filp);
2574 struct f2fs_inode_info *fi = F2FS_I(inode);
2575 struct fsxattr fa;
2576
2577 memset(&fa, 0, sizeof(struct fsxattr));
2578 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2579 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2580
2581 if (f2fs_sb_has_project_quota(inode->i_sb))
2582 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2583 fi->i_projid);
2584
2585 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2586 return -EFAULT;
2587 return 0;
2588}
2589
2590static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2591{
2592 struct inode *inode = file_inode(filp);
2593 struct f2fs_inode_info *fi = F2FS_I(inode);
2594 struct fsxattr fa;
2595 unsigned int flags;
2596 int err;
2597
2598 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2599 return -EFAULT;
2600
2601 /* Make sure caller has proper permission */
2602 if (!inode_owner_or_capable(inode))
2603 return -EACCES;
2604
2605 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2606 return -EOPNOTSUPP;
2607
2608 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2609 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2610 return -EOPNOTSUPP;
2611
2612 err = mnt_want_write_file(filp);
2613 if (err)
2614 return err;
2615
2616 inode_lock(inode);
2617 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2618 (flags & F2FS_FL_XFLAG_VISIBLE);
2619 err = __f2fs_ioc_setflags(inode, flags);
2620 inode_unlock(inode);
2621 mnt_drop_write_file(filp);
2622 if (err)
2623 return err;
2624
2625 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2626 if (err)
2627 return err;
2628
2629 return 0;
2630}
e066b83c 2631
52656e6c
JK		 2632long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2633{
2634 switch (cmd) {
2635 case F2FS_IOC_GETFLAGS:
2636 return f2fs_ioc_getflags(filp, arg);
2637 case F2FS_IOC_SETFLAGS:
2638 return f2fs_ioc_setflags(filp, arg);
d49f3e89
CY		 2639 case F2FS_IOC_GETVERSION:
2640 return f2fs_ioc_getversion(filp, arg);
88b88a66
JK		 2641 case F2FS_IOC_START_ATOMIC_WRITE:
2642 return f2fs_ioc_start_atomic_write(filp);
2643 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2644 return f2fs_ioc_commit_atomic_write(filp);
02a1335f
JK		 2645 case F2FS_IOC_START_VOLATILE_WRITE:
2646 return f2fs_ioc_start_volatile_write(filp);
1e84371f
JK		 2647 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2648 return f2fs_ioc_release_volatile_write(filp);
2649 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2650 return f2fs_ioc_abort_volatile_write(filp);
1abff93d
JK		 2651 case F2FS_IOC_SHUTDOWN:
2652 return f2fs_ioc_shutdown(filp, arg);
52656e6c
JK		 2653 case FITRIM:
2654 return f2fs_ioc_fitrim(filp, arg);
f424f664
JK		 2655 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2656 return f2fs_ioc_set_encryption_policy(filp, arg);
2657 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2658 return f2fs_ioc_get_encryption_policy(filp, arg);
2659 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2660 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
c1c1b583
CY		 2661 case F2FS_IOC_GARBAGE_COLLECT:
2662 return f2fs_ioc_gc(filp, arg);
34dc77ad
JK		 2663 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2664 return f2fs_ioc_gc_range(filp, arg);
456b88e4
CY		 2665 case F2FS_IOC_WRITE_CHECKPOINT:
2666 return f2fs_ioc_write_checkpoint(filp, arg);
d323d005
CY		 2667 case F2FS_IOC_DEFRAGMENT:
2668 return f2fs_ioc_defragment(filp, arg);
4dd6f977
JK		 2669 case F2FS_IOC_MOVE_RANGE:
2670 return f2fs_ioc_move_range(filp, arg);
e066b83c
JK		 2671 case F2FS_IOC_FLUSH_DEVICE:
2672 return f2fs_ioc_flush_device(filp, arg);
e65ef207
JK		 2673 case F2FS_IOC_GET_FEATURES:
2674 return f2fs_ioc_get_features(filp, arg);
2c1d0305
CY		 2675 case F2FS_IOC_FSGETXATTR:
2676 return f2fs_ioc_fsgetxattr(filp, arg);
2677 case F2FS_IOC_FSSETXATTR:
2678 return f2fs_ioc_fssetxattr(filp, arg);
fbfa2cc5
JK		 2679 default:
2680 return -ENOTTY;
2681 }
2682}
2683
fcc85a4d
JK		 2684static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2685{
b439b103
JK		 2686 struct file *file = iocb->ki_filp;
2687 struct inode *inode = file_inode(file);
9dfa1baf 2688 struct blk_plug plug;
b439b103 2689 ssize_t ret;
fcc85a4d 2690
b439b103
JK		 2691 inode_lock(inode);
2692 ret = generic_write_checks(iocb, from);
2693 if (ret > 0) {
dc91de78 2694 int err;
a7de6086 2695
dc91de78
JK		 2696 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2697 set_inode_flag(inode, FI_NO_PREALLOC);
a7de6086 2698
dc91de78 2699 err = f2fs_preallocate_blocks(iocb, from);
a7de6086
JK		 2700 if (err) {
2701 inode_unlock(inode);
2702 return err;
9dfa1baf 2703 }
a7de6086
JK		 2704 blk_start_plug(&plug);
2705 ret = __generic_file_write_iter(iocb, from);
2706 blk_finish_plug(&plug);
dc91de78 2707 clear_inode_flag(inode, FI_NO_PREALLOC);
b0af6d49
CY		 2708
2709 if (ret > 0)
2710 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
b439b103
JK		 2711 }
2712 inode_unlock(inode);
2713
e2592217
CH		 2714 if (ret > 0)
2715 ret = generic_write_sync(iocb, ret);
b439b103 2716 return ret;
fcc85a4d
JK		 2717}
2718
e9750824
NJ		 2719#ifdef CONFIG_COMPAT
2720long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2721{
2722 switch (cmd) {
2723 case F2FS_IOC32_GETFLAGS:
2724 cmd = F2FS_IOC_GETFLAGS;
2725 break;
2726 case F2FS_IOC32_SETFLAGS:
2727 cmd = F2FS_IOC_SETFLAGS;
2728 break;
04ef4b62
CY		 2729 case F2FS_IOC32_GETVERSION:
2730 cmd = F2FS_IOC_GETVERSION;
2731 break;
2732 case F2FS_IOC_START_ATOMIC_WRITE:
2733 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2734 case F2FS_IOC_START_VOLATILE_WRITE:
2735 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2736 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2737 case F2FS_IOC_SHUTDOWN:
2738 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2739 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2740 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2741 case F2FS_IOC_GARBAGE_COLLECT:
34dc77ad 2742 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
04ef4b62
CY		 2743 case F2FS_IOC_WRITE_CHECKPOINT:
2744 case F2FS_IOC_DEFRAGMENT:
4dd6f977 2745 case F2FS_IOC_MOVE_RANGE:
e066b83c 2746 case F2FS_IOC_FLUSH_DEVICE:
e65ef207 2747 case F2FS_IOC_GET_FEATURES:
2c1d0305
CY		 2748 case F2FS_IOC_FSGETXATTR:
2749 case F2FS_IOC_FSSETXATTR:
4dd6f977 2750 break;
e9750824
NJ		 2751 default:
2752 return -ENOIOCTLCMD;
2753 }
2754 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2755}
2756#endif
2757
fbfa2cc5 2758const struct file_operations f2fs_file_operations = {
267378d4 2759 .llseek = f2fs_llseek,
aad4f8bb 2760 .read_iter = generic_file_read_iter,
fcc85a4d
JK		 2761 .write_iter = f2fs_file_write_iter,
2762 .open = f2fs_file_open,
12662234 2763 .release = f2fs_release_file,
fbfa2cc5 2764 .mmap = f2fs_file_mmap,
7a10f017 2765 .flush = f2fs_file_flush,
fbfa2cc5
JK		 2766 .fsync = f2fs_sync_file,
2767 .fallocate = f2fs_fallocate,
2768 .unlocked_ioctl = f2fs_ioctl,
e9750824
NJ		 2769#ifdef CONFIG_COMPAT
2770 .compat_ioctl = f2fs_compat_ioctl,
2771#endif
fbfa2cc5 2772 .splice_read = generic_file_splice_read,
8d020765 2773 .splice_write = iter_file_splice_write,
fbfa2cc5 2774};
Linux 6.x block layer and io_uring tree(s)