projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: save a multiplication for last_nid calculation
[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
1f227a3e
JK		 56 if (unlikely(f2fs_cp_error(sbi))) {
57 err = -EIO;
58 goto err;
59 }
60
fbfa2cc5 61 sb_start_pagefault(inode->i_sb);
b3d208f9
JK		 62
63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
b067ba1f 64
fbfa2cc5 65 /* block allocation */
e479556b 66 f2fs_lock_op(sbi);
b3d208f9 67 set_new_dnode(&dn, inode, NULL, NULL, 0);
b600965c 68 err = f2fs_reserve_block(&dn, page->index);
9ba69cf9
JK		 69 if (err) {
70 f2fs_unlock_op(sbi);
b600965c 71 goto out;
9ba69cf9
JK		 72 }
73 f2fs_put_dnode(&dn);
74 f2fs_unlock_op(sbi);
fbfa2cc5 75
2c4db1a6 76 f2fs_balance_fs(sbi, dn.node_changed);
2a340760 77
11bac800 78 file_update_time(vmf->vma->vm_file);
5a3a2d83 79 down_read(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5 80 lock_page(page);
6bacf52f 81 if (unlikely(page->mapping != inode->i_mapping ||
9851e6e1 82 page_offset(page) > i_size_read(inode) ||
6bacf52f 83 !PageUptodate(page))) {
fbfa2cc5
JK		 84 unlock_page(page);
85 err = -EFAULT;
5a3a2d83 86 goto out_sem;
fbfa2cc5
JK		 87 }
88
89 /*
90 * check to see if the page is mapped already (no holes)
91 */
92 if (PageMappedToDisk(page))
9851e6e1 93 goto mapped;
fbfa2cc5
JK		 94
95 /* page is wholly or partially inside EOF */
09cbfeaf 96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
9edcdabf 97 i_size_read(inode)) {
fbfa2cc5 98 unsigned offset;
09cbfeaf
KS		 99 offset = i_size_read(inode) & ~PAGE_MASK;
100 zero_user_segment(page, offset, PAGE_SIZE);
fbfa2cc5
JK		 101 }
102 set_page_dirty(page);
237c0790
JK		 103 if (!PageUptodate(page))
104 SetPageUptodate(page);
fbfa2cc5 105
b0af6d49
CY		 106 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
107
e943a10d 108 trace_f2fs_vm_page_mkwrite(page, DATA);
9851e6e1
NJ		 109mapped:
110 /* fill the page */
fec1d657 111 f2fs_wait_on_page_writeback(page, DATA, false);
08b39fbd
CY		 112
113 /* wait for GCed encrypted page writeback */
1958593e 114 if (f2fs_encrypted_file(inode))
d4c759ee 115 f2fs_wait_on_block_writeback(sbi, dn.data_blkaddr);
08b39fbd 116
5a3a2d83
QS		 117out_sem:
118 up_read(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5
JK		 119out:
120 sb_end_pagefault(inode->i_sb);
d0239e1b 121 f2fs_update_time(sbi, REQ_TIME);
1f227a3e 122err:
fbfa2cc5
JK		 123 return block_page_mkwrite_return(err);
124}
125
126static const struct vm_operations_struct f2fs_file_vm_ops = {
5a3a2d83 127 .fault = f2fs_filemap_fault,
f1820361 128 .map_pages = filemap_map_pages,
692bb55d 129 .page_mkwrite = f2fs_vm_page_mkwrite,
fbfa2cc5
JK		 130};
131
354a3399
JK		 132static int get_parent_ino(struct inode *inode, nid_t *pino)
133{
134 struct dentry *dentry;
135
136 inode = igrab(inode);
137 dentry = d_find_any_alias(inode);
138 iput(inode);
139 if (!dentry)
140 return 0;
141
f0947e5c
JK		 142 *pino = parent_ino(dentry);
143 dput(dentry);
354a3399
JK		 144 return 1;
145}
146
9d1589ef
CY		 147static inline bool need_do_checkpoint(struct inode *inode)
148{
4081363f 149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9d1589ef
CY		 150 bool need_cp = false;
151
152 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
153 need_cp = true;
bbf156f7 154 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
e7d55452 155 need_cp = true;
9d1589ef
CY		 156 else if (file_wrong_pino(inode))
157 need_cp = true;
158 else if (!space_for_roll_forward(sbi))
159 need_cp = true;
160 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
161 need_cp = true;
d5053a34
JK		 162 else if (test_opt(sbi, FASTBOOT))
163 need_cp = true;
a344b9fd
JK		 164 else if (sbi->active_logs == 2)
165 need_cp = true;
9d1589ef
CY		 166
167 return need_cp;
168}
169
9c7bb702
CL		 170static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
171{
172 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
173 bool ret = false;
174 /* But we need to avoid that there are some inode updates */
175 if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino))
176 ret = true;
177 f2fs_put_page(i, 0);
178 return ret;
179}
180
51455b19
CL		 181static void try_to_fix_pino(struct inode *inode)
182{
183 struct f2fs_inode_info *fi = F2FS_I(inode);
184 nid_t pino;
185
186 down_write(&fi->i_sem);
51455b19
CL		 187 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
188 get_parent_ino(inode, &pino)) {
205b9822 189 f2fs_i_pino_write(inode, pino);
51455b19 190 file_got_pino(inode);
51455b19 191 }
ee6d182f 192 up_write(&fi->i_sem);
51455b19
CL		 193}
194
608514de
JK		 195static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
196 int datasync, bool atomic)
fbfa2cc5
JK		 197{
198 struct inode *inode = file->f_mapping->host;
4081363f 199 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2403c155 200 nid_t ino = inode->i_ino;
fbfa2cc5
JK		 201 int ret = 0;
202 bool need_cp = false;
203 struct writeback_control wbc = {
c81bf1c8 204 .sync_mode = WB_SYNC_ALL,
fbfa2cc5
JK		 205 .nr_to_write = LONG_MAX,
206 .for_reclaim = 0,
207 };
208
6bacf52f 209 if (unlikely(f2fs_readonly(inode->i_sb)))
1fa95b0b
NJ		 210 return 0;
211
a2a4a7e4 212 trace_f2fs_sync_file_enter(inode);
ea1aa12c
JK		 213
214 /* if fdatasync is triggered, let's do in-place-update */
c46a155b 215 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
91942321 216 set_inode_flag(inode, FI_NEED_IPU);
3b49c9a1 217 ret = file_write_and_wait_range(file, start, end);
91942321 218 clear_inode_flag(inode, FI_NEED_IPU);
c1ce1b02 219
a2a4a7e4
NJ		 220 if (ret) {
221 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
fbfa2cc5 222 return ret;
a2a4a7e4 223 }
fbfa2cc5 224
9c7bb702 225 /* if the inode is dirty, let's recover all the time */
281518c6 226 if (!f2fs_skip_inode_update(inode, datasync)) {
2286c020 227 f2fs_write_inode(inode, NULL);
9c7bb702
CL		 228 goto go_write;
229 }
230
6d99ba41
JK		 231 /*
232 * if there is no written data, don't waste time to write recovery info.
233 */
91942321 234 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
2403c155 235 !exist_written_data(sbi, ino, APPEND_INO)) {
19c9c466 236
9c7bb702
CL		 237 /* it may call write_inode just prior to fsync */
238 if (need_inode_page_update(sbi, ino))
19c9c466 239 goto go_write;
19c9c466 240
91942321 241 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
2403c155 242 exist_written_data(sbi, ino, UPDATE_INO))
6d99ba41
JK		 243 goto flush_out;
244 goto out;
245 }
19c9c466 246go_write:
e5d2385e
JK		 247 /*
248 * Both of fdatasync() and fsync() are able to be recovered from
249 * sudden-power-off.
250 */
91942321 251 down_read(&F2FS_I(inode)->i_sem);
9d1589ef 252 need_cp = need_do_checkpoint(inode);
91942321 253 up_read(&F2FS_I(inode)->i_sem);
d928bfbf 254
fbfa2cc5
JK		 255 if (need_cp) {
256 /* all the dirty node pages should be flushed for POR */
257 ret = f2fs_sync_fs(inode->i_sb, 1);
d928bfbf 258
51455b19
CL		 259 /*
260 * We've secured consistency through sync_fs. Following pino
261 * will be used only for fsynced inodes after checkpoint.
262 */
263 try_to_fix_pino(inode);
91942321
JK		 264 clear_inode_flag(inode, FI_APPEND_WRITE);
265 clear_inode_flag(inode, FI_UPDATE_WRITE);
51455b19
CL		 266 goto out;
267 }
88bd02c9 268sync_nodes:
26de9b11 269 ret = fsync_node_pages(sbi, inode, &wbc, atomic);
c267ec15
JK		 270 if (ret)
271 goto out;
51455b19 272
871f599f 273 /* if cp_error was enabled, we should avoid infinite loop */
6d5a1495
CY		 274 if (unlikely(f2fs_cp_error(sbi))) {
275 ret = -EIO;
871f599f 276 goto out;
6d5a1495 277 }
871f599f 278
51455b19 279 if (need_inode_block_update(sbi, ino)) {
7c45729a 280 f2fs_mark_inode_dirty_sync(inode, true);
51455b19
CL		 281 f2fs_write_inode(inode, NULL);
282 goto sync_nodes;
fbfa2cc5 283 }
51455b19 284
b6a245eb
JK		 285 /*
286 * If it's atomic_write, it's just fine to keep write ordering. So
287 * here we don't need to wait for node write completion, since we use
288 * node chain which serializes node blocks. If one of node writes are
289 * reordered, we can see simply broken chain, resulting in stopping
290 * roll-forward recovery. It means we'll recover all or none node blocks
291 * given fsync mark.
292 */
293 if (!atomic) {
294 ret = wait_on_node_pages_writeback(sbi, ino);
295 if (ret)
296 goto out;
297 }
51455b19
CL		 298
299 /* once recovery info is written, don't need to tack this */
a49324f1 300 remove_ino_entry(sbi, ino, APPEND_INO);
91942321 301 clear_inode_flag(inode, FI_APPEND_WRITE);
51455b19 302flush_out:
e7c75ab0 303 if (!atomic)
39d787be 304 ret = f2fs_issue_flush(sbi, inode->i_ino);
3f06252f
CY		 305 if (!ret) {
306 remove_ino_entry(sbi, ino, UPDATE_INO);
307 clear_inode_flag(inode, FI_UPDATE_WRITE);
39d787be 308 remove_ino_entry(sbi, ino, FLUSH_INO);
3f06252f 309 }
d0239e1b 310 f2fs_update_time(sbi, REQ_TIME);
fbfa2cc5 311out:
a2a4a7e4 312 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
05ca3632 313 f2fs_trace_ios(NULL, 1);
fbfa2cc5
JK		 314 return ret;
315}
316
608514de
JK		 317int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
318{
1f227a3e
JK		 319 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
320 return -EIO;
608514de
JK		 321 return f2fs_do_sync_file(file, start, end, datasync, false);
322}
323
7f7670fe
JK		 324static pgoff_t __get_first_dirty_index(struct address_space *mapping,
325 pgoff_t pgofs, int whence)
326{
327 struct pagevec pvec;
328 int nr_pages;
329
330 if (whence != SEEK_DATA)
331 return 0;
332
333 /* find first dirty page index */
334 pagevec_init(&pvec, 0);
65b85ccc
JK		 335 nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
336 PAGECACHE_TAG_DIRTY, 1);
80dd9c0e 337 pgofs = nr_pages ? pvec.pages[0]->index : ULONG_MAX;
7f7670fe
JK		 338 pagevec_release(&pvec);
339 return pgofs;
340}
341
342static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs,
343 int whence)
344{
345 switch (whence) {
346 case SEEK_DATA:
347 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
348 (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR))
349 return true;
350 break;
351 case SEEK_HOLE:
352 if (blkaddr == NULL_ADDR)
353 return true;
354 break;
355 }
356 return false;
357}
358
267378d4
CY		 359static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
360{
361 struct inode *inode = file->f_mapping->host;
362 loff_t maxbytes = inode->i_sb->s_maxbytes;
363 struct dnode_of_data dn;
7f7670fe
JK		 364 pgoff_t pgofs, end_offset, dirty;
365 loff_t data_ofs = offset;
366 loff_t isize;
267378d4
CY		 367 int err = 0;
368
5955102c 369 inode_lock(inode);
267378d4
CY		 370
371 isize = i_size_read(inode);
372 if (offset >= isize)
373 goto fail;
374
375 /* handle inline data case */
622f28ae 376 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
267378d4
CY		 377 if (whence == SEEK_HOLE)
378 data_ofs = isize;
379 goto found;
380 }
381
09cbfeaf 382 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
267378d4 383
7f7670fe
JK		 384 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
385
09cbfeaf 386 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 387 set_new_dnode(&dn, inode, NULL, NULL, 0);
ac6f1999 388 err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
267378d4
CY		 389 if (err && err != -ENOENT) {
390 goto fail;
391 } else if (err == -ENOENT) {
e1c42045 392 /* direct node does not exists */
267378d4 393 if (whence == SEEK_DATA) {
3cf45747 394 pgofs = get_next_page_offset(&dn, pgofs);
267378d4
CY		 395 continue;
396 } else {
397 goto found;
398 }
399 }
400
81ca7350 401 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
267378d4
CY		 402
403 /* find data/hole in dnode block */
404 for (; dn.ofs_in_node < end_offset;
405 dn.ofs_in_node++, pgofs++,
09cbfeaf 406 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 407 block_t blkaddr;
7a2af766
CY		 408 blkaddr = datablock_addr(dn.inode,
409 dn.node_page, dn.ofs_in_node);
267378d4 410
7f7670fe 411 if (__found_offset(blkaddr, dirty, pgofs, whence)) {
267378d4
CY		 412 f2fs_put_dnode(&dn);
413 goto found;
414 }
415 }
416 f2fs_put_dnode(&dn);
417 }
418
419 if (whence == SEEK_DATA)
420 goto fail;
267378d4 421found:
fe369bc8
JK		 422 if (whence == SEEK_HOLE && data_ofs > isize)
423 data_ofs = isize;
5955102c 424 inode_unlock(inode);
267378d4
CY		 425 return vfs_setpos(file, data_ofs, maxbytes);
426fail:
5955102c 427 inode_unlock(inode);
267378d4
CY		 428 return -ENXIO;
429}
430
431static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
432{
433 struct inode *inode = file->f_mapping->host;
434 loff_t maxbytes = inode->i_sb->s_maxbytes;
435
436 switch (whence) {
437 case SEEK_SET:
438 case SEEK_CUR:
439 case SEEK_END:
440 return generic_file_llseek_size(file, offset, whence,
441 maxbytes, i_size_read(inode));
442 case SEEK_DATA:
443 case SEEK_HOLE:
0b4c5afd
JK		 444 if (offset < 0)
445 return -ENXIO;
267378d4
CY		 446 return f2fs_seek_block(file, offset, whence);
447 }
448
449 return -EINVAL;
450}
451
fbfa2cc5
JK		 452static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
453{
b3d208f9 454 struct inode *inode = file_inode(file);
b9d777b8 455 int err;
b3d208f9 456
1f227a3e
JK		 457 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
458 return -EIO;
459
b3d208f9 460 /* we don't need to use inline_data strictly */
b9d777b8
JK		 461 err = f2fs_convert_inline_inode(inode);
462 if (err)
463 return err;
b3d208f9 464
fbfa2cc5
JK		 465 file_accessed(file);
466 vma->vm_ops = &f2fs_file_vm_ops;
467 return 0;
468}
469
fcc85a4d
JK		 470static int f2fs_file_open(struct inode *inode, struct file *filp)
471{
33b13951 472 struct dentry *dir;
fcc85a4d 473
0abd675e
CY		 474 if (f2fs_encrypted_inode(inode)) {
475 int ret = fscrypt_get_encryption_info(inode);
fcc85a4d 476 if (ret)
ae108668 477 return -EACCES;
0b81d077 478 if (!fscrypt_has_encryption_key(inode))
ae108668 479 return -ENOKEY;
fcc85a4d 480 }
33b13951
JK		 481 dir = dget_parent(file_dentry(filp));
482 if (f2fs_encrypted_inode(d_inode(dir)) &&
483 !fscrypt_has_permitted_context(d_inode(dir), inode)) {
484 dput(dir);
8074bb51 485 return -EPERM;
33b13951
JK		 486 }
487 dput(dir);
0abd675e 488 return dquot_file_open(inode, filp);
fcc85a4d
JK		 489}
490
b292dcab 491int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
fbfa2cc5 492{
4081363f 493 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
fbfa2cc5 494 struct f2fs_node *raw_node;
19b2c30d 495 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
fbfa2cc5 496 __le32 *addr;
7a2af766
CY		 497 int base = 0;
498
499 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
500 base = get_extra_isize(dn->inode);
fbfa2cc5 501
45590710 502 raw_node = F2FS_NODE(dn->node_page);
7a2af766 503 addr = blkaddr_in_node(raw_node) + base + ofs;
fbfa2cc5 504
6c311ec6 505 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
fbfa2cc5
JK		 506 block_t blkaddr = le32_to_cpu(*addr);
507 if (blkaddr == NULL_ADDR)
508 continue;
509
e1509cf2 510 dn->data_blkaddr = NULL_ADDR;
216a620a 511 set_data_blkaddr(dn);
fbfa2cc5 512 invalidate_blocks(sbi, blkaddr);
3c6c2beb 513 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
91942321 514 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
fbfa2cc5
JK		 515 nr_free++;
516 }
19b2c30d 517
fbfa2cc5 518 if (nr_free) {
19b2c30d
CY		 519 pgoff_t fofs;
520 /*
521 * once we invalidate valid blkaddr in range [ofs, ofs + count],
522 * we will invalidate all blkaddr in the whole range.
523 */
524 fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
81ca7350 525 dn->inode) + ofs;
19b2c30d 526 f2fs_update_extent_cache_range(dn, fofs, 0, len);
d7cc950b 527 dec_valid_block_count(sbi, dn->inode, nr_free);
fbfa2cc5
JK		 528 }
529 dn->ofs_in_node = ofs;
51dd6249 530
d0239e1b 531 f2fs_update_time(sbi, REQ_TIME);
51dd6249
NJ		 532 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
533 dn->ofs_in_node, nr_free);
fbfa2cc5
JK		 534 return nr_free;
535}
536
537void truncate_data_blocks(struct dnode_of_data *dn)
538{
539 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
540}
541
0bfcfcca 542static int truncate_partial_data_page(struct inode *inode, u64 from,
43f3eae1 543 bool cache_only)
fbfa2cc5 544{
09cbfeaf
KS		 545 unsigned offset = from & (PAGE_SIZE - 1);
546 pgoff_t index = from >> PAGE_SHIFT;
43f3eae1 547 struct address_space *mapping = inode->i_mapping;
fbfa2cc5
JK		 548 struct page *page;
549
43f3eae1 550 if (!offset && !cache_only)
b3d208f9 551 return 0;
fbfa2cc5 552
43f3eae1 553 if (cache_only) {
34b5d5c2 554 page = find_lock_page(mapping, index);
43f3eae1
JK		 555 if (page && PageUptodate(page))
556 goto truncate_out;
557 f2fs_put_page(page, 1);
b3d208f9 558 return 0;
43f3eae1 559 }
fbfa2cc5 560
a56c7c6f 561 page = get_lock_data_page(inode, index, true);
43f3eae1 562 if (IS_ERR(page))
a78aaa2c 563 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
43f3eae1 564truncate_out:
fec1d657 565 f2fs_wait_on_page_writeback(page, DATA, true);
09cbfeaf 566 zero_user(page, offset, PAGE_SIZE - offset);
a9bcf9bc
JK		 567
568 /* An encrypted inode should have a key and truncate the last page. */
569 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
570 if (!cache_only)
0bfcfcca 571 set_page_dirty(page);
fbfa2cc5 572 f2fs_put_page(page, 1);
b3d208f9 573 return 0;
fbfa2cc5
JK		 574}
575
764aa3e9 576int truncate_blocks(struct inode *inode, u64 from, bool lock)
fbfa2cc5 577{
4081363f 578 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 579 unsigned int blocksize = inode->i_sb->s_blocksize;
580 struct dnode_of_data dn;
581 pgoff_t free_from;
9ffe0fb5 582 int count = 0, err = 0;
b3d208f9 583 struct page *ipage;
0bfcfcca 584 bool truncate_page = false;
fbfa2cc5 585
51dd6249
NJ		 586 trace_f2fs_truncate_blocks_enter(inode, from);
587
f7ef9b83 588 free_from = (pgoff_t)F2FS_BYTES_TO_BLK(from + blocksize - 1);
fbfa2cc5 589
09210c97
CY		 590 if (free_from >= sbi->max_file_blocks)
591 goto free_partial;
592
764aa3e9
JK		 593 if (lock)
594 f2fs_lock_op(sbi);
9ffe0fb5 595
b3d208f9
JK		 596 ipage = get_node_page(sbi, inode->i_ino);
597 if (IS_ERR(ipage)) {
598 err = PTR_ERR(ipage);
599 goto out;
600 }
601
602 if (f2fs_has_inline_data(inode)) {
bd4667cb 603 truncate_inline_inode(inode, ipage, from);
b3d208f9 604 f2fs_put_page(ipage, 1);
0bfcfcca 605 truncate_page = true;
b3d208f9
JK		 606 goto out;
607 }
608
609 set_new_dnode(&dn, inode, ipage, NULL, 0);
79344efb 610 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
fbfa2cc5
JK		 611 if (err) {
612 if (err == -ENOENT)
613 goto free_next;
b3d208f9 614 goto out;
1ce86bf6
JK		 615 }
616
81ca7350 617 count = ADDRS_PER_PAGE(dn.node_page, inode);
fbfa2cc5
JK		 618
619 count -= dn.ofs_in_node;
9850cf4a 620 f2fs_bug_on(sbi, count < 0);
39936837 621
fbfa2cc5
JK		 622 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
623 truncate_data_blocks_range(&dn, count);
624 free_from += count;
625 }
626
627 f2fs_put_dnode(&dn);
628free_next:
629 err = truncate_inode_blocks(inode, free_from);
764d2c80
JK		 630out:
631 if (lock)
632 f2fs_unlock_op(sbi);
09210c97 633free_partial:
b3d208f9
JK		 634 /* lastly zero out the first data page */
635 if (!err)
0bfcfcca 636 err = truncate_partial_data_page(inode, from, truncate_page);
fbfa2cc5 637
51dd6249 638 trace_f2fs_truncate_blocks_exit(inode, err);
fbfa2cc5
JK		 639 return err;
640}
641
9a449e9c 642int f2fs_truncate(struct inode *inode)
fbfa2cc5 643{
b0154891
CY		 644 int err;
645
1f227a3e
JK		 646 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
647 return -EIO;
648
fbfa2cc5
JK		 649 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
650 S_ISLNK(inode->i_mode)))
b0154891 651 return 0;
fbfa2cc5 652
51dd6249
NJ		 653 trace_f2fs_truncate(inode);
654
14b44d23
JK		 655#ifdef CONFIG_F2FS_FAULT_INJECTION
656 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
657 f2fs_show_injection_info(FAULT_TRUNCATE);
658 return -EIO;
659 }
660#endif
92dffd01 661 /* we should check inline_data size */
b9d777b8 662 if (!f2fs_may_inline_data(inode)) {
b0154891
CY		 663 err = f2fs_convert_inline_inode(inode);
664 if (err)
665 return err;
92dffd01
JK		 666 }
667
9a449e9c 668 err = truncate_blocks(inode, i_size_read(inode), true);
b0154891
CY		 669 if (err)
670 return err;
671
078cd827 672 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 673 f2fs_mark_inode_dirty_sync(inode, false);
b0154891 674 return 0;
fbfa2cc5
JK		 675}
676
a528d35e 677int f2fs_getattr(const struct path *path, struct kstat *stat,
1c6d8ee4 678 u32 request_mask, unsigned int query_flags)
fbfa2cc5 679{
a528d35e 680 struct inode *inode = d_inode(path->dentry);
1c6d8ee4
CY		 681 struct f2fs_inode_info *fi = F2FS_I(inode);
682 unsigned int flags;
683
2c1d0305 684 flags = fi->i_flags & (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
1c6d8ee4
CY		 685 if (flags & FS_APPEND_FL)
686 stat->attributes |= STATX_ATTR_APPEND;
687 if (flags & FS_COMPR_FL)
688 stat->attributes |= STATX_ATTR_COMPRESSED;
689 if (f2fs_encrypted_inode(inode))
690 stat->attributes |= STATX_ATTR_ENCRYPTED;
691 if (flags & FS_IMMUTABLE_FL)
692 stat->attributes |= STATX_ATTR_IMMUTABLE;
693 if (flags & FS_NODUMP_FL)
694 stat->attributes |= STATX_ATTR_NODUMP;
695
696 stat->attributes_mask |= (STATX_ATTR_APPEND |
697 STATX_ATTR_COMPRESSED |
698 STATX_ATTR_ENCRYPTED |
699 STATX_ATTR_IMMUTABLE |
700 STATX_ATTR_NODUMP);
701
fbfa2cc5 702 generic_fillattr(inode, stat);
5b4267d1
JK		 703
704 /* we need to show initial sectors used for inline_data/dentries */
705 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
706 f2fs_has_inline_dentry(inode))
707 stat->blocks += (stat->size + 511) >> 9;
708
fbfa2cc5
JK		 709 return 0;
710}
711
712#ifdef CONFIG_F2FS_FS_POSIX_ACL
713static void __setattr_copy(struct inode *inode, const struct iattr *attr)
714{
fbfa2cc5
JK		 715 unsigned int ia_valid = attr->ia_valid;
716
717 if (ia_valid & ATTR_UID)
718 inode->i_uid = attr->ia_uid;
719 if (ia_valid & ATTR_GID)
720 inode->i_gid = attr->ia_gid;
721 if (ia_valid & ATTR_ATIME)
722 inode->i_atime = timespec_trunc(attr->ia_atime,
723 inode->i_sb->s_time_gran);
724 if (ia_valid & ATTR_MTIME)
725 inode->i_mtime = timespec_trunc(attr->ia_mtime,
726 inode->i_sb->s_time_gran);
727 if (ia_valid & ATTR_CTIME)
728 inode->i_ctime = timespec_trunc(attr->ia_ctime,
729 inode->i_sb->s_time_gran);
730 if (ia_valid & ATTR_MODE) {
731 umode_t mode = attr->ia_mode;
732
733 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
734 mode &= ~S_ISGID;
91942321 735 set_acl_inode(inode, mode);
fbfa2cc5
JK		 736 }
737}
738#else
739#define __setattr_copy setattr_copy
740#endif
741
742int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
743{
2b0143b5 744 struct inode *inode = d_inode(dentry);
fbfa2cc5 745 int err;
c0ed4405 746 bool size_changed = false;
fbfa2cc5 747
1f227a3e
JK		 748 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
749 return -EIO;
750
31051c85 751 err = setattr_prepare(dentry, attr);
fbfa2cc5
JK		 752 if (err)
753 return err;
754
0abd675e
CY		 755 if (is_quota_modification(inode, attr)) {
756 err = dquot_initialize(inode);
757 if (err)
758 return err;
759 }
760 if ((attr->ia_valid & ATTR_UID &&
761 !uid_eq(attr->ia_uid, inode->i_uid)) ||
762 (attr->ia_valid & ATTR_GID &&
763 !gid_eq(attr->ia_gid, inode->i_gid))) {
764 err = dquot_transfer(inode, attr);
765 if (err)
766 return err;
767 }
768
09db6a2e 769 if (attr->ia_valid & ATTR_SIZE) {
67773a1f
EB		 770 if (f2fs_encrypted_inode(inode)) {
771 err = fscrypt_get_encryption_info(inode);
772 if (err)
773 return err;
774 if (!fscrypt_has_encryption_key(inode))
775 return -ENOKEY;
776 }
fcc85a4d 777
3c454145 778 if (attr->ia_size <= i_size_read(inode)) {
5a3a2d83 779 down_write(&F2FS_I(inode)->i_mmap_sem);
09db6a2e 780 truncate_setsize(inode, attr->ia_size);
9a449e9c 781 err = f2fs_truncate(inode);
5a3a2d83 782 up_write(&F2FS_I(inode)->i_mmap_sem);
b0154891
CY		 783 if (err)
784 return err;
09db6a2e
CY		 785 } else {
786 /*
3c454145
CY		 787 * do not trim all blocks after i_size if target size is
788 * larger than i_size.
09db6a2e 789 */
5a3a2d83 790 down_write(&F2FS_I(inode)->i_mmap_sem);
3c454145 791 truncate_setsize(inode, attr->ia_size);
5a3a2d83 792 up_write(&F2FS_I(inode)->i_mmap_sem);
0cab80ee
CY		 793
794 /* should convert inline inode here */
b9d777b8 795 if (!f2fs_may_inline_data(inode)) {
0cab80ee
CY		 796 err = f2fs_convert_inline_inode(inode);
797 if (err)
798 return err;
799 }
078cd827 800 inode->i_mtime = inode->i_ctime = current_time(inode);
09db6a2e 801 }
c0ed4405 802
a0d00fad
CY		 803 down_write(&F2FS_I(inode)->i_sem);
804 F2FS_I(inode)->last_disk_size = i_size_read(inode);
805 up_write(&F2FS_I(inode)->i_sem);
806
c0ed4405 807 size_changed = true;
fbfa2cc5
JK		 808 }
809
810 __setattr_copy(inode, attr);
811
812 if (attr->ia_valid & ATTR_MODE) {
a6dda0e6 813 err = posix_acl_chmod(inode, get_inode_mode(inode));
91942321
JK		 814 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
815 inode->i_mode = F2FS_I(inode)->i_acl_mode;
816 clear_inode_flag(inode, FI_ACL_MODE);
fbfa2cc5
JK		 817 }
818 }
819
c0ed4405
YH		 820 /* file size may changed here */
821 f2fs_mark_inode_dirty_sync(inode, size_changed);
15d04354
JK		 822
823 /* inode change will produce dirty node pages flushed by checkpoint */
824 f2fs_balance_fs(F2FS_I_SB(inode), true);
825
fbfa2cc5
JK		 826 return err;
827}
828
829const struct inode_operations f2fs_file_inode_operations = {
830 .getattr = f2fs_getattr,
831 .setattr = f2fs_setattr,
832 .get_acl = f2fs_get_acl,
a6dda0e6 833 .set_acl = f2fs_set_acl,
fbfa2cc5 834#ifdef CONFIG_F2FS_FS_XATTR
fbfa2cc5 835 .listxattr = f2fs_listxattr,
fbfa2cc5 836#endif
9ab70134 837 .fiemap = f2fs_fiemap,
fbfa2cc5
JK		 838};
839
6394328a 840static int fill_zero(struct inode *inode, pgoff_t index,
fbfa2cc5
JK		 841 loff_t start, loff_t len)
842{
4081363f 843 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 844 struct page *page;
845
846 if (!len)
6394328a 847 return 0;
fbfa2cc5 848
2c4db1a6 849 f2fs_balance_fs(sbi, true);
bd43df02 850
e479556b 851 f2fs_lock_op(sbi);
64aa7ed9 852 page = get_new_data_page(inode, NULL, index, false);
e479556b 853 f2fs_unlock_op(sbi);
fbfa2cc5 854
6394328a
CY		 855 if (IS_ERR(page))
856 return PTR_ERR(page);
857
fec1d657 858 f2fs_wait_on_page_writeback(page, DATA, true);
6394328a
CY		 859 zero_user(page, start, len);
860 set_page_dirty(page);
861 f2fs_put_page(page, 1);
862 return 0;
fbfa2cc5
JK		 863}
864
865int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
866{
fbfa2cc5
JK		 867 int err;
868
ea58711e 869 while (pg_start < pg_end) {
fbfa2cc5 870 struct dnode_of_data dn;
ea58711e 871 pgoff_t end_offset, count;
9eaeba70 872
fbfa2cc5 873 set_new_dnode(&dn, inode, NULL, NULL, 0);
ea58711e 874 err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
fbfa2cc5 875 if (err) {
ea58711e 876 if (err == -ENOENT) {
48ab25f4 877 pg_start = get_next_page_offset(&dn, pg_start);
fbfa2cc5 878 continue;
ea58711e 879 }
fbfa2cc5
JK		 880 return err;
881 }
882
81ca7350 883 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
ea58711e
CY		 884 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
885
886 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
887
888 truncate_data_blocks_range(&dn, count);
fbfa2cc5 889 f2fs_put_dnode(&dn);
ea58711e
CY		 890
891 pg_start += count;
fbfa2cc5
JK		 892 }
893 return 0;
894}
895
a66c7b2f 896static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
fbfa2cc5
JK		 897{
898 pgoff_t pg_start, pg_end;
899 loff_t off_start, off_end;
b9d777b8 900 int ret;
fbfa2cc5 901
b9d777b8
JK		 902 ret = f2fs_convert_inline_inode(inode);
903 if (ret)
904 return ret;
9ffe0fb5 905
09cbfeaf
KS		 906 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
907 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
fbfa2cc5 908
09cbfeaf
KS		 909 off_start = offset & (PAGE_SIZE - 1);
910 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5
JK		 911
912 if (pg_start == pg_end) {
6394328a 913 ret = fill_zero(inode, pg_start, off_start,
fbfa2cc5 914 off_end - off_start);
6394328a
CY		 915 if (ret)
916 return ret;
fbfa2cc5 917 } else {
6394328a
CY		 918 if (off_start) {
919 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 920 PAGE_SIZE - off_start);
6394328a
CY		 921 if (ret)
922 return ret;
923 }
924 if (off_end) {
925 ret = fill_zero(inode, pg_end, 0, off_end);
926 if (ret)
927 return ret;
928 }
fbfa2cc5
JK		 929
930 if (pg_start < pg_end) {
931 struct address_space *mapping = inode->i_mapping;
932 loff_t blk_start, blk_end;
4081363f 933 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127a3d4 934
2c4db1a6 935 f2fs_balance_fs(sbi, true);
fbfa2cc5 936
09cbfeaf
KS		 937 blk_start = (loff_t)pg_start << PAGE_SHIFT;
938 blk_end = (loff_t)pg_end << PAGE_SHIFT;
5a3a2d83 939 down_write(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5
JK		 940 truncate_inode_pages_range(mapping, blk_start,
941 blk_end - 1);
39936837 942
e479556b 943 f2fs_lock_op(sbi);
fbfa2cc5 944 ret = truncate_hole(inode, pg_start, pg_end);
e479556b 945 f2fs_unlock_op(sbi);
5a3a2d83 946 up_write(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5
JK		 947 }
948 }
949
fbfa2cc5
JK		 950 return ret;
951}
952
0a2aa8fb
JK		 953static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
954 int *do_replace, pgoff_t off, pgoff_t len)
b4ace337
CY		 955{
956 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
957 struct dnode_of_data dn;
0a2aa8fb 958 int ret, done, i;
ecbaa406 959
0a2aa8fb 960next_dnode:
6e2c64ad 961 set_new_dnode(&dn, inode, NULL, NULL, 0);
0a2aa8fb 962 ret = get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
6e2c64ad
JK		 963 if (ret && ret != -ENOENT) {
964 return ret;
965 } else if (ret == -ENOENT) {
0a2aa8fb
JK		 966 if (dn.max_level == 0)
967 return -ENOENT;
968 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
969 blkaddr += done;
970 do_replace += done;
971 goto next;
972 }
973
974 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
975 dn.ofs_in_node, len);
976 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
7a2af766
CY		 977 *blkaddr = datablock_addr(dn.inode,
978 dn.node_page, dn.ofs_in_node);
0a2aa8fb
JK		 979 if (!is_checkpointed_data(sbi, *blkaddr)) {
980
981 if (test_opt(sbi, LFS)) {
982 f2fs_put_dnode(&dn);
983 return -ENOTSUPP;
984 }
985
6e2c64ad 986 /* do not invalidate this block address */
f28b3434 987 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
0a2aa8fb 988 *do_replace = 1;
b4ace337 989 }
6e2c64ad 990 }
0a2aa8fb
JK		 991 f2fs_put_dnode(&dn);
992next:
993 len -= done;
994 off += done;
995 if (len)
996 goto next_dnode;
997 return 0;
998}
b4ace337 999
0a2aa8fb
JK		 1000static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1001 int *do_replace, pgoff_t off, int len)
1002{
1003 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1004 struct dnode_of_data dn;
1005 int ret, i;
b4ace337 1006
0a2aa8fb
JK		 1007 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1008 if (*do_replace == 0)
1009 continue;
b4ace337 1010
0a2aa8fb
JK		 1011 set_new_dnode(&dn, inode, NULL, NULL, 0);
1012 ret = get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1013 if (ret) {
1014 dec_valid_block_count(sbi, inode, 1);
1015 invalidate_blocks(sbi, *blkaddr);
1016 } else {
1017 f2fs_update_data_blkaddr(&dn, *blkaddr);
36abef4e 1018 }
0a2aa8fb
JK		 1019 f2fs_put_dnode(&dn);
1020 }
1021 return 0;
1022}
1023
1024static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1025 block_t *blkaddr, int *do_replace,
1026 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1027{
1028 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1029 pgoff_t i = 0;
1030 int ret;
36abef4e 1031
0a2aa8fb
JK		 1032 while (i < len) {
1033 if (blkaddr[i] == NULL_ADDR && !full) {
1034 i++;
1035 continue;
6e2c64ad 1036 }
b4ace337 1037
0a2aa8fb
JK		 1038 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1039 struct dnode_of_data dn;
1040 struct node_info ni;
1041 size_t new_size;
1042 pgoff_t ilen;
b4ace337 1043
0a2aa8fb
JK		 1044 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1045 ret = get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1046 if (ret)
1047 return ret;
b4ace337 1048
0a2aa8fb
JK		 1049 get_node_info(sbi, dn.nid, &ni);
1050 ilen = min((pgoff_t)
1051 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1052 dn.ofs_in_node, len - i);
1053 do {
7a2af766
CY		 1054 dn.data_blkaddr = datablock_addr(dn.inode,
1055 dn.node_page, dn.ofs_in_node);
0a2aa8fb
JK		 1056 truncate_data_blocks_range(&dn, 1);
1057
1058 if (do_replace[i]) {
1059 f2fs_i_blocks_write(src_inode,
0abd675e 1060 1, false, false);
0a2aa8fb 1061 f2fs_i_blocks_write(dst_inode,
0abd675e 1062 1, true, false);
0a2aa8fb
JK		 1063 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1064 blkaddr[i], ni.version, true, false);
1065
1066 do_replace[i] = 0;
1067 }
1068 dn.ofs_in_node++;
1069 i++;
1070 new_size = (dst + i) << PAGE_SHIFT;
1071 if (dst_inode->i_size < new_size)
1072 f2fs_i_size_write(dst_inode, new_size);
e87f7329 1073 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
6e2c64ad 1074
0a2aa8fb
JK		 1075 f2fs_put_dnode(&dn);
1076 } else {
1077 struct page *psrc, *pdst;
1078
1079 psrc = get_lock_data_page(src_inode, src + i, true);
1080 if (IS_ERR(psrc))
1081 return PTR_ERR(psrc);
1082 pdst = get_new_data_page(dst_inode, NULL, dst + i,
1083 true);
1084 if (IS_ERR(pdst)) {
1085 f2fs_put_page(psrc, 1);
1086 return PTR_ERR(pdst);
1087 }
1088 f2fs_copy_page(psrc, pdst);
1089 set_page_dirty(pdst);
1090 f2fs_put_page(pdst, 1);
6e2c64ad 1091 f2fs_put_page(psrc, 1);
b4ace337 1092
0a2aa8fb
JK		 1093 ret = truncate_hole(src_inode, src + i, src + i + 1);
1094 if (ret)
1095 return ret;
1096 i++;
1097 }
6e2c64ad
JK		 1098 }
1099 return 0;
0a2aa8fb 1100}
b4ace337 1101
0a2aa8fb
JK		 1102static int __exchange_data_block(struct inode *src_inode,
1103 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
363cad7f 1104 pgoff_t len, bool full)
0a2aa8fb
JK		 1105{
1106 block_t *src_blkaddr;
1107 int *do_replace;
363cad7f 1108 pgoff_t olen;
0a2aa8fb
JK		 1109 int ret;
1110
363cad7f
JK		 1111 while (len) {
1112 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
0a2aa8fb 1113
a7c3e901 1114 src_blkaddr = kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
363cad7f
JK		 1115 if (!src_blkaddr)
1116 return -ENOMEM;
0a2aa8fb 1117
a7c3e901 1118 do_replace = kvzalloc(sizeof(int) * olen, GFP_KERNEL);
363cad7f
JK		 1119 if (!do_replace) {
1120 kvfree(src_blkaddr);
1121 return -ENOMEM;
1122 }
0a2aa8fb 1123
363cad7f
JK		 1124 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1125 do_replace, src, olen);
1126 if (ret)
1127 goto roll_back;
0a2aa8fb 1128
363cad7f
JK		 1129 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1130 do_replace, src, dst, olen, full);
1131 if (ret)
1132 goto roll_back;
1133
1134 src += olen;
1135 dst += olen;
1136 len -= olen;
1137
1138 kvfree(src_blkaddr);
1139 kvfree(do_replace);
1140 }
0a2aa8fb
JK		 1141 return 0;
1142
1143roll_back:
1144 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
1145 kvfree(src_blkaddr);
1146 kvfree(do_replace);
6e2c64ad
JK		 1147 return ret;
1148}
b4ace337 1149
6e2c64ad
JK		 1150static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
1151{
1152 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1153 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
0a2aa8fb 1154 int ret;
6e2c64ad 1155
0a2aa8fb
JK		 1156 f2fs_balance_fs(sbi, true);
1157 f2fs_lock_op(sbi);
5f281fab
JK		 1158
1159 f2fs_drop_extent_tree(inode);
1160
0a2aa8fb
JK		 1161 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1162 f2fs_unlock_op(sbi);
b4ace337
CY		 1163 return ret;
1164}
1165
1166static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1167{
1168 pgoff_t pg_start, pg_end;
1169 loff_t new_size;
1170 int ret;
1171
b4ace337
CY		 1172 if (offset + len >= i_size_read(inode))
1173 return -EINVAL;
1174
1175 /* collapse range should be aligned to block size of f2fs. */
1176 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1177 return -EINVAL;
1178
b9d777b8
JK		 1179 ret = f2fs_convert_inline_inode(inode);
1180 if (ret)
1181 return ret;
97a7b2c2 1182
09cbfeaf
KS		 1183 pg_start = offset >> PAGE_SHIFT;
1184 pg_end = (offset + len) >> PAGE_SHIFT;
b4ace337 1185
5a3a2d83 1186 down_write(&F2FS_I(inode)->i_mmap_sem);
b4ace337
CY		 1187 /* write out all dirty pages from offset */
1188 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1189 if (ret)
5a3a2d83 1190 goto out;
b4ace337
CY		 1191
1192 truncate_pagecache(inode, offset);
1193
1194 ret = f2fs_do_collapse(inode, pg_start, pg_end);
1195 if (ret)
5a3a2d83 1196 goto out;
b4ace337 1197
6e2c64ad
JK		 1198 /* write out all moved pages, if possible */
1199 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1200 truncate_pagecache(inode, offset);
1201
b4ace337 1202 new_size = i_size_read(inode) - len;
6e2c64ad 1203 truncate_pagecache(inode, new_size);
b4ace337
CY		 1204
1205 ret = truncate_blocks(inode, new_size, true);
1206 if (!ret)
fc9581c8 1207 f2fs_i_size_write(inode, new_size);
b4ace337 1208
5a3a2d83
QS		 1209out:
1210 up_write(&F2FS_I(inode)->i_mmap_sem);
b4ace337
CY		 1211 return ret;
1212}
1213
6e961949
CY		 1214static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1215 pgoff_t end)
1216{
1217 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1218 pgoff_t index = start;
1219 unsigned int ofs_in_node = dn->ofs_in_node;
1220 blkcnt_t count = 0;
1221 int ret;
1222
1223 for (; index < end; index++, dn->ofs_in_node++) {
7a2af766
CY		 1224 if (datablock_addr(dn->inode, dn->node_page,
1225 dn->ofs_in_node) == NULL_ADDR)
6e961949
CY		 1226 count++;
1227 }
1228
1229 dn->ofs_in_node = ofs_in_node;
1230 ret = reserve_new_blocks(dn, count);
1231 if (ret)
1232 return ret;
1233
1234 dn->ofs_in_node = ofs_in_node;
1235 for (index = start; index < end; index++, dn->ofs_in_node++) {
7a2af766
CY		 1236 dn->data_blkaddr = datablock_addr(dn->inode,
1237 dn->node_page, dn->ofs_in_node);
6e961949
CY		 1238 /*
1239 * reserve_new_blocks will not guarantee entire block
1240 * allocation.
1241 */
1242 if (dn->data_blkaddr == NULL_ADDR) {
1243 ret = -ENOSPC;
1244 break;
1245 }
1246 if (dn->data_blkaddr != NEW_ADDR) {
1247 invalidate_blocks(sbi, dn->data_blkaddr);
1248 dn->data_blkaddr = NEW_ADDR;
1249 set_data_blkaddr(dn);
1250 }
1251 }
1252
1253 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1254
1255 return ret;
1256}
1257
75cd4e09
CY		 1258static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1259 int mode)
1260{
1261 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1262 struct address_space *mapping = inode->i_mapping;
1263 pgoff_t index, pg_start, pg_end;
1264 loff_t new_size = i_size_read(inode);
1265 loff_t off_start, off_end;
1266 int ret = 0;
1267
75cd4e09
CY		 1268 ret = inode_newsize_ok(inode, (len + offset));
1269 if (ret)
1270 return ret;
1271
b9d777b8
JK		 1272 ret = f2fs_convert_inline_inode(inode);
1273 if (ret)
1274 return ret;
75cd4e09 1275
5a3a2d83 1276 down_write(&F2FS_I(inode)->i_mmap_sem);
75cd4e09
CY		 1277 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1278 if (ret)
5a3a2d83 1279 goto out_sem;
75cd4e09
CY		 1280
1281 truncate_pagecache_range(inode, offset, offset + len - 1);
1282
09cbfeaf
KS		 1283 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1284 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
75cd4e09 1285
09cbfeaf
KS		 1286 off_start = offset & (PAGE_SIZE - 1);
1287 off_end = (offset + len) & (PAGE_SIZE - 1);
75cd4e09
CY		 1288
1289 if (pg_start == pg_end) {
6394328a
CY		 1290 ret = fill_zero(inode, pg_start, off_start,
1291 off_end - off_start);
1292 if (ret)
5a3a2d83 1293 goto out_sem;
6394328a 1294
75cd4e09
CY		 1295 new_size = max_t(loff_t, new_size, offset + len);
1296 } else {
1297 if (off_start) {
6394328a 1298 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 1299 PAGE_SIZE - off_start);
6394328a 1300 if (ret)
5a3a2d83 1301 goto out_sem;
6394328a 1302
75cd4e09 1303 new_size = max_t(loff_t, new_size,
09cbfeaf 1304 (loff_t)pg_start << PAGE_SHIFT);
75cd4e09
CY		 1305 }
1306
6e961949 1307 for (index = pg_start; index < pg_end;) {
75cd4e09 1308 struct dnode_of_data dn;
6e961949
CY		 1309 unsigned int end_offset;
1310 pgoff_t end;
75cd4e09
CY		 1311
1312 f2fs_lock_op(sbi);
1313
6e961949
CY		 1314 set_new_dnode(&dn, inode, NULL, NULL, 0);
1315 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
75cd4e09
CY		 1316 if (ret) {
1317 f2fs_unlock_op(sbi);
1318 goto out;
1319 }
1320
6e961949
CY		 1321 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1322 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1323
1324 ret = f2fs_do_zero_range(&dn, index, end);
75cd4e09
CY		 1325 f2fs_put_dnode(&dn);
1326 f2fs_unlock_op(sbi);
9434fcde
CY		 1327
1328 f2fs_balance_fs(sbi, dn.node_changed);
1329
6e961949
CY		 1330 if (ret)
1331 goto out;
75cd4e09 1332
6e961949 1333 index = end;
75cd4e09 1334 new_size = max_t(loff_t, new_size,
6e961949 1335 (loff_t)index << PAGE_SHIFT);
75cd4e09
CY		 1336 }
1337
1338 if (off_end) {
6394328a
CY		 1339 ret = fill_zero(inode, pg_end, 0, off_end);
1340 if (ret)
1341 goto out;
1342
75cd4e09
CY		 1343 new_size = max_t(loff_t, new_size, offset + len);
1344 }
1345 }
1346
1347out:
ee6d182f 1348 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
fc9581c8 1349 f2fs_i_size_write(inode, new_size);
5a3a2d83
QS		 1350out_sem:
1351 up_write(&F2FS_I(inode)->i_mmap_sem);
75cd4e09
CY		 1352
1353 return ret;
1354}
1355
f62185d0
CY		 1356static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1357{
1358 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
0a2aa8fb 1359 pgoff_t nr, pg_start, pg_end, delta, idx;
f62185d0 1360 loff_t new_size;
6e2c64ad 1361 int ret = 0;
f62185d0 1362
f62185d0 1363 new_size = i_size_read(inode) + len;
46e82fb1
KM		 1364 ret = inode_newsize_ok(inode, new_size);
1365 if (ret)
1366 return ret;
f62185d0
CY		 1367
1368 if (offset >= i_size_read(inode))
1369 return -EINVAL;
1370
1371 /* insert range should be aligned to block size of f2fs. */
1372 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1373 return -EINVAL;
1374
b9d777b8
JK		 1375 ret = f2fs_convert_inline_inode(inode);
1376 if (ret)
1377 return ret;
97a7b2c2 1378
2c4db1a6 1379 f2fs_balance_fs(sbi, true);
2a340760 1380
5a3a2d83 1381 down_write(&F2FS_I(inode)->i_mmap_sem);
f62185d0
CY		 1382 ret = truncate_blocks(inode, i_size_read(inode), true);
1383 if (ret)
5a3a2d83 1384 goto out;
f62185d0
CY		 1385
1386 /* write out all dirty pages from offset */
1387 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1388 if (ret)
5a3a2d83 1389 goto out;
f62185d0
CY		 1390
1391 truncate_pagecache(inode, offset);
1392
09cbfeaf
KS		 1393 pg_start = offset >> PAGE_SHIFT;
1394 pg_end = (offset + len) >> PAGE_SHIFT;
f62185d0 1395 delta = pg_end - pg_start;
0a2aa8fb
JK		 1396 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1397
1398 while (!ret && idx > pg_start) {
1399 nr = idx - pg_start;
1400 if (nr > delta)
1401 nr = delta;
1402 idx -= nr;
f62185d0 1403
f62185d0 1404 f2fs_lock_op(sbi);
5f281fab
JK		 1405 f2fs_drop_extent_tree(inode);
1406
0a2aa8fb
JK		 1407 ret = __exchange_data_block(inode, inode, idx,
1408 idx + delta, nr, false);
f62185d0
CY		 1409 f2fs_unlock_op(sbi);
1410 }
1411
6e2c64ad
JK		 1412 /* write out all moved pages, if possible */
1413 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1414 truncate_pagecache(inode, offset);
1415
1416 if (!ret)
fc9581c8 1417 f2fs_i_size_write(inode, new_size);
5a3a2d83
QS		 1418out:
1419 up_write(&F2FS_I(inode)->i_mmap_sem);
f62185d0
CY		 1420 return ret;
1421}
1422
fbfa2cc5
JK		 1423static int expand_inode_data(struct inode *inode, loff_t offset,
1424 loff_t len, int mode)
1425{
4081363f 1426 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
e12dd7bd
JK		 1427 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
1428 pgoff_t pg_end;
fbfa2cc5 1429 loff_t new_size = i_size_read(inode);
e12dd7bd 1430 loff_t off_end;
a7de6086 1431 int err;
fbfa2cc5 1432
a7de6086
JK		 1433 err = inode_newsize_ok(inode, (len + offset));
1434 if (err)
1435 return err;
fbfa2cc5 1436
a7de6086
JK		 1437 err = f2fs_convert_inline_inode(inode);
1438 if (err)
1439 return err;
9e09fc85 1440
2c4db1a6 1441 f2fs_balance_fs(sbi, true);
2a340760 1442
e12dd7bd 1443 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
09cbfeaf 1444 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5 1445
e12dd7bd
JK		 1446 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1447 map.m_len = pg_end - map.m_lblk;
1448 if (off_end)
1449 map.m_len++;
ead43275 1450
a7de6086
JK		 1451 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1452 if (err) {
e12dd7bd 1453 pgoff_t last_off;
fbfa2cc5 1454
e12dd7bd 1455 if (!map.m_len)
a7de6086 1456 return err;
98397ff3 1457
e12dd7bd
JK		 1458 last_off = map.m_lblk + map.m_len - 1;
1459
1460 /* update new size to the failed position */
1461 new_size = (last_off == pg_end) ? offset + len:
1462 (loff_t)(last_off + 1) << PAGE_SHIFT;
1463 } else {
1464 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
fbfa2cc5
JK		 1465 }
1466
ee6d182f 1467 if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size)
fc9581c8 1468 f2fs_i_size_write(inode, new_size);
fbfa2cc5 1469
a7de6086 1470 return err;
fbfa2cc5
JK		 1471}
1472
1473static long f2fs_fallocate(struct file *file, int mode,
1474 loff_t offset, loff_t len)
1475{
6131ffaa 1476 struct inode *inode = file_inode(file);
587c0a42 1477 long ret = 0;
fbfa2cc5 1478
1f227a3e
JK		 1479 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1480 return -EIO;
1481
c998012b
CY		 1482 /* f2fs only support ->fallocate for regular file */
1483 if (!S_ISREG(inode->i_mode))
1484 return -EINVAL;
1485
f62185d0
CY		 1486 if (f2fs_encrypted_inode(inode) &&
1487 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
fcc85a4d
JK		 1488 return -EOPNOTSUPP;
1489
b4ace337 1490 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
f62185d0
CY		 1491 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1492 FALLOC_FL_INSERT_RANGE))
fbfa2cc5
JK		 1493 return -EOPNOTSUPP;
1494
5955102c 1495 inode_lock(inode);
3375f696 1496
587c0a42
TY		 1497 if (mode & FALLOC_FL_PUNCH_HOLE) {
1498 if (offset >= inode->i_size)
1499 goto out;
1500
a66c7b2f 1501 ret = punch_hole(inode, offset, len);
b4ace337
CY		 1502 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1503 ret = f2fs_collapse_range(inode, offset, len);
75cd4e09
CY		 1504 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1505 ret = f2fs_zero_range(inode, offset, len, mode);
f62185d0
CY		 1506 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1507 ret = f2fs_insert_range(inode, offset, len);
b4ace337 1508 } else {
fbfa2cc5 1509 ret = expand_inode_data(inode, offset, len, mode);
b4ace337 1510 }
fbfa2cc5 1511
3af60a49 1512 if (!ret) {
078cd827 1513 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 1514 f2fs_mark_inode_dirty_sync(inode, false);
26787236
JK		 1515 if (mode & FALLOC_FL_KEEP_SIZE)
1516 file_set_keep_isize(inode);
d0239e1b 1517 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3af60a49 1518 }
3375f696 1519
587c0a42 1520out:
5955102c 1521 inode_unlock(inode);
3375f696 1522
c01e2853 1523 trace_f2fs_fallocate(inode, mode, offset, len, ret);
fbfa2cc5
JK		 1524 return ret;
1525}
1526
1e84371f
JK		 1527static int f2fs_release_file(struct inode *inode, struct file *filp)
1528{
de5307e4
JK		 1529 /*
1530 * f2fs_relase_file is called at every close calls. So we should
1531 * not drop any inmemory pages by close called by other process.
1532 */
1533 if (!(filp->f_mode & FMODE_WRITE) ||
1534 atomic_read(&inode->i_writecount) != 1)
1535 return 0;
1536
1e84371f
JK		 1537 /* some remained atomic pages should discarded */
1538 if (f2fs_is_atomic_file(inode))
29b96b54 1539 drop_inmem_pages(inode);
1e84371f 1540 if (f2fs_is_volatile_file(inode)) {
91942321 1541 clear_inode_flag(inode, FI_VOLATILE_FILE);
648d50ba 1542 stat_dec_volatile_write(inode);
91942321 1543 set_inode_flag(inode, FI_DROP_CACHE);
1e84371f 1544 filemap_fdatawrite(inode->i_mapping);
91942321 1545 clear_inode_flag(inode, FI_DROP_CACHE);
1e84371f
JK		 1546 }
1547 return 0;
1548}
1549
7a10f017 1550static int f2fs_file_flush(struct file *file, fl_owner_t id)
fbfa2cc5 1551{
7a10f017
JK		 1552 struct inode *inode = file_inode(file);
1553
1554 /*
1555 * If the process doing a transaction is crashed, we should do
1556 * roll-back. Otherwise, other reader/write can see corrupted database
1557 * until all the writers close its file. Since this should be done
1558 * before dropping file lock, it needs to do in ->flush.
1559 */
1560 if (f2fs_is_atomic_file(inode) &&
1561 F2FS_I(inode)->inmem_task == current)
1562 drop_inmem_pages(inode);
1563 return 0;
fbfa2cc5
JK		 1564}
1565
52656e6c 1566static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
fbfa2cc5 1567{
6131ffaa 1568 struct inode *inode = file_inode(filp);
fbfa2cc5 1569 struct f2fs_inode_info *fi = F2FS_I(inode);
2c1d0305
CY		 1570 unsigned int flags = fi->i_flags &
1571 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL);
52656e6c
JK		 1572 return put_user(flags, (int __user *)arg);
1573}
fbfa2cc5 1574
2c1d0305
CY		 1575static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1576{
1577 struct f2fs_inode_info *fi = F2FS_I(inode);
1578 unsigned int oldflags;
1579
1580 /* Is it quota file? Do not allow user to mess with it */
1581 if (IS_NOQUOTA(inode))
1582 return -EPERM;
1583
1584 flags = f2fs_mask_flags(inode->i_mode, flags);
1585
1586 oldflags = fi->i_flags;
1587
1588 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL))
1589 if (!capable(CAP_LINUX_IMMUTABLE))
1590 return -EPERM;
1591
1592 flags = flags & (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1593 flags |= oldflags & ~(FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL);
1594 fi->i_flags = flags;
1595
1596 if (fi->i_flags & FS_PROJINHERIT_FL)
1597 set_inode_flag(inode, FI_PROJ_INHERIT);
1598 else
1599 clear_inode_flag(inode, FI_PROJ_INHERIT);
1600
1601 inode->i_ctime = current_time(inode);
1602 f2fs_set_inode_flags(inode);
1603 f2fs_mark_inode_dirty_sync(inode, false);
1604 return 0;
1605}
1606
52656e6c
JK		 1607static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1608{
1609 struct inode *inode = file_inode(filp);
69494229 1610 unsigned int flags;
52656e6c 1611 int ret;
fbfa2cc5 1612
7fb17fe4
CY		 1613 if (!inode_owner_or_capable(inode))
1614 return -EACCES;
1615
1616 if (get_user(flags, (int __user *)arg))
1617 return -EFAULT;
1618
52656e6c
JK		 1619 ret = mnt_want_write_file(filp);
1620 if (ret)
1621 return ret;
fbfa2cc5 1622
5955102c 1623 inode_lock(inode);
fbfa2cc5 1624
2c1d0305 1625 ret = __f2fs_ioc_setflags(inode, flags);
fbfa2cc5 1626
a72d4b97 1627 inode_unlock(inode);
52656e6c
JK		 1628 mnt_drop_write_file(filp);
1629 return ret;
1630}
4b2fecc8 1631
d49f3e89
CY		 1632static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1633{
1634 struct inode *inode = file_inode(filp);
1635
1636 return put_user(inode->i_generation, (int __user *)arg);
1637}
1638
88b88a66
JK		 1639static int f2fs_ioc_start_atomic_write(struct file *filp)
1640{
1641 struct inode *inode = file_inode(filp);
f4c9c743 1642 int ret;
88b88a66
JK		 1643
1644 if (!inode_owner_or_capable(inode))
1645 return -EACCES;
1646
e811898c
JK		 1647 if (!S_ISREG(inode->i_mode))
1648 return -EINVAL;
1649
7fb17fe4
CY		 1650 ret = mnt_want_write_file(filp);
1651 if (ret)
1652 return ret;
1653
0fac558b
CY		 1654 inode_lock(inode);
1655
1e84371f 1656 if (f2fs_is_atomic_file(inode))
7fb17fe4 1657 goto out;
88b88a66 1658
f4c9c743
CY		 1659 ret = f2fs_convert_inline_inode(inode);
1660 if (ret)
7fb17fe4 1661 goto out;
88b88a66 1662
91942321 1663 set_inode_flag(inode, FI_ATOMIC_FILE);
6c3acd97 1664 set_inode_flag(inode, FI_HOT_DATA);
d0239e1b
JK		 1665 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1666
c27753d6 1667 if (!get_dirty_pages(inode))
684ca7e5 1668 goto inc_stat;
c27753d6
JK		 1669
1670 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
204706c7 1671 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
c27753d6
JK		 1672 inode->i_ino, get_dirty_pages(inode));
1673 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
684ca7e5 1674 if (ret) {
91942321 1675 clear_inode_flag(inode, FI_ATOMIC_FILE);
84a23fbe 1676 clear_inode_flag(inode, FI_HOT_DATA);
684ca7e5
KM		 1677 goto out;
1678 }
1679
1680inc_stat:
7a10f017 1681 F2FS_I(inode)->inmem_task = current;
26a28a0c
JK		 1682 stat_inc_atomic_write(inode);
1683 stat_update_max_atomic_write(inode);
684ca7e5 1684out:
0fac558b 1685 inode_unlock(inode);
7fb17fe4 1686 mnt_drop_write_file(filp);
c27753d6 1687 return ret;
88b88a66
JK		 1688}
1689
1690static int f2fs_ioc_commit_atomic_write(struct file *filp)
1691{
1692 struct inode *inode = file_inode(filp);
1693 int ret;
1694
1695 if (!inode_owner_or_capable(inode))
1696 return -EACCES;
1697
1698 ret = mnt_want_write_file(filp);
1699 if (ret)
1700 return ret;
1701
0fac558b
CY		 1702 inode_lock(inode);
1703
7fb17fe4
CY		 1704 if (f2fs_is_volatile_file(inode))
1705 goto err_out;
1706
6282adbf 1707 if (f2fs_is_atomic_file(inode)) {
29b96b54 1708 ret = commit_inmem_pages(inode);
5fe45743 1709 if (ret)
edb27dee 1710 goto err_out;
5fe45743 1711
26a28a0c 1712 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
5fe45743
CY		 1713 if (!ret) {
1714 clear_inode_flag(inode, FI_ATOMIC_FILE);
84a23fbe 1715 clear_inode_flag(inode, FI_HOT_DATA);
5fe45743 1716 stat_dec_atomic_write(inode);
447135a8 1717 }
26a28a0c 1718 } else {
774e1b78 1719 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
6282adbf 1720 }
edb27dee 1721err_out:
0fac558b 1722 inode_unlock(inode);
88b88a66
JK		 1723 mnt_drop_write_file(filp);
1724 return ret;
1725}
1726
02a1335f
JK		 1727static int f2fs_ioc_start_volatile_write(struct file *filp)
1728{
1729 struct inode *inode = file_inode(filp);
f4c9c743 1730 int ret;
02a1335f
JK		 1731
1732 if (!inode_owner_or_capable(inode))
1733 return -EACCES;
1734
8ff0971f
CY		 1735 if (!S_ISREG(inode->i_mode))
1736 return -EINVAL;
1737
7fb17fe4
CY		 1738 ret = mnt_want_write_file(filp);
1739 if (ret)
1740 return ret;
1741
0fac558b
CY		 1742 inode_lock(inode);
1743
1e84371f 1744 if (f2fs_is_volatile_file(inode))
7fb17fe4 1745 goto out;
1e84371f 1746
f4c9c743
CY		 1747 ret = f2fs_convert_inline_inode(inode);
1748 if (ret)
7fb17fe4 1749 goto out;
b3d208f9 1750
648d50ba
CY		 1751 stat_inc_volatile_write(inode);
1752 stat_update_max_volatile_write(inode);
1753
91942321 1754 set_inode_flag(inode, FI_VOLATILE_FILE);
d0239e1b 1755 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 1756out:
0fac558b 1757 inode_unlock(inode);
7fb17fe4
CY		 1758 mnt_drop_write_file(filp);
1759 return ret;
02a1335f
JK		 1760}
1761
1e84371f
JK		 1762static int f2fs_ioc_release_volatile_write(struct file *filp)
1763{
1764 struct inode *inode = file_inode(filp);
7fb17fe4 1765 int ret;
1e84371f
JK		 1766
1767 if (!inode_owner_or_capable(inode))
1768 return -EACCES;
1769
7fb17fe4
CY		 1770 ret = mnt_want_write_file(filp);
1771 if (ret)
1772 return ret;
1773
0fac558b
CY		 1774 inode_lock(inode);
1775
1e84371f 1776 if (!f2fs_is_volatile_file(inode))
7fb17fe4 1777 goto out;
1e84371f 1778
7fb17fe4
CY		 1779 if (!f2fs_is_first_block_written(inode)) {
1780 ret = truncate_partial_data_page(inode, 0, true);
1781 goto out;
1782 }
3c6c2beb 1783
7fb17fe4
CY		 1784 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1785out:
0fac558b 1786 inode_unlock(inode);
7fb17fe4
CY		 1787 mnt_drop_write_file(filp);
1788 return ret;
1e84371f
JK		 1789}
1790
1791static int f2fs_ioc_abort_volatile_write(struct file *filp)
1792{
1793 struct inode *inode = file_inode(filp);
1794 int ret;
1795
1796 if (!inode_owner_or_capable(inode))
1797 return -EACCES;
1798
1799 ret = mnt_want_write_file(filp);
1800 if (ret)
1801 return ret;
1802
0fac558b
CY		 1803 inode_lock(inode);
1804
26dc3d44 1805 if (f2fs_is_atomic_file(inode))
29b96b54 1806 drop_inmem_pages(inode);
732d5648 1807 if (f2fs_is_volatile_file(inode)) {
91942321 1808 clear_inode_flag(inode, FI_VOLATILE_FILE);
648d50ba 1809 stat_dec_volatile_write(inode);
608514de 1810 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
732d5648 1811 }
de6a8ec9 1812
0fac558b
CY		 1813 inode_unlock(inode);
1814
1e84371f 1815 mnt_drop_write_file(filp);
d0239e1b 1816 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1e84371f
JK		 1817 return ret;
1818}
1819
1abff93d
JK		 1820static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1821{
1822 struct inode *inode = file_inode(filp);
1823 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1824 struct super_block *sb = sbi->sb;
1825 __u32 in;
7fb17fe4 1826 int ret;
1abff93d
JK		 1827
1828 if (!capable(CAP_SYS_ADMIN))
1829 return -EPERM;
1830
1831 if (get_user(in, (__u32 __user *)arg))
1832 return -EFAULT;
1833
7fb17fe4
CY		 1834 ret = mnt_want_write_file(filp);
1835 if (ret)
1836 return ret;
1837
1abff93d
JK		 1838 switch (in) {
1839 case F2FS_GOING_DOWN_FULLSYNC:
1840 sb = freeze_bdev(sb->s_bdev);
1841 if (sb && !IS_ERR(sb)) {
38f91ca8 1842 f2fs_stop_checkpoint(sbi, false);
1abff93d
JK		 1843 thaw_bdev(sb->s_bdev, sb);
1844 }
1845 break;
1846 case F2FS_GOING_DOWN_METASYNC:
1847 /* do checkpoint only */
1848 f2fs_sync_fs(sb, 1);
38f91ca8 1849 f2fs_stop_checkpoint(sbi, false);
1abff93d
JK		 1850 break;
1851 case F2FS_GOING_DOWN_NOSYNC:
38f91ca8 1852 f2fs_stop_checkpoint(sbi, false);
1abff93d 1853 break;
c912a829 1854 case F2FS_GOING_DOWN_METAFLUSH:
b0af6d49 1855 sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
38f91ca8 1856 f2fs_stop_checkpoint(sbi, false);
c912a829 1857 break;
1abff93d 1858 default:
7fb17fe4
CY		 1859 ret = -EINVAL;
1860 goto out;
1abff93d 1861 }
d0239e1b 1862 f2fs_update_time(sbi, REQ_TIME);
7fb17fe4
CY		 1863out:
1864 mnt_drop_write_file(filp);
1865 return ret;
1abff93d
JK		 1866}
1867
52656e6c
JK		 1868static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1869{
1870 struct inode *inode = file_inode(filp);
1871 struct super_block *sb = inode->i_sb;
1872 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1873 struct fstrim_range range;
1874 int ret;
4b2fecc8 1875
52656e6c
JK		 1876 if (!capable(CAP_SYS_ADMIN))
1877 return -EPERM;
4b2fecc8 1878
52656e6c
JK		 1879 if (!blk_queue_discard(q))
1880 return -EOPNOTSUPP;
4b2fecc8 1881
52656e6c
JK		 1882 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1883 sizeof(range)))
1884 return -EFAULT;
4b2fecc8 1885
7fb17fe4
CY		 1886 ret = mnt_want_write_file(filp);
1887 if (ret)
1888 return ret;
1889
52656e6c
JK		 1890 range.minlen = max((unsigned int)range.minlen,
1891 q->limits.discard_granularity);
1892 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
7fb17fe4 1893 mnt_drop_write_file(filp);
52656e6c
JK		 1894 if (ret < 0)
1895 return ret;
4b2fecc8 1896
52656e6c
JK		 1897 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1898 sizeof(range)))
1899 return -EFAULT;
d0239e1b 1900 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
52656e6c
JK		 1901 return 0;
1902}
1903
f424f664
JK		 1904static bool uuid_is_nonzero(__u8 u[16])
1905{
1906 int i;
1907
1908 for (i = 0; i < 16; i++)
1909 if (u[i])
1910 return true;
1911 return false;
1912}
1913
1914static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
1915{
f424f664
JK		 1916 struct inode *inode = file_inode(filp);
1917
d0239e1b 1918 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 1919
db717d8e 1920 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
f424f664
JK		 1921}
1922
1923static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
1924{
db717d8e 1925 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
f424f664
JK		 1926}
1927
1928static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
1929{
1930 struct inode *inode = file_inode(filp);
1931 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1932 int err;
1933
1934 if (!f2fs_sb_has_crypto(inode->i_sb))
1935 return -EOPNOTSUPP;
1936
1937 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
1938 goto got_it;
1939
1940 err = mnt_want_write_file(filp);
1941 if (err)
1942 return err;
1943
1944 /* update superblock with uuid */
1945 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
1946
c5bda1c8 1947 err = f2fs_commit_super(sbi, false);
f424f664
JK		 1948 if (err) {
1949 /* undo new data */
1950 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
e8240f65 1951 mnt_drop_write_file(filp);
f424f664
JK		 1952 return err;
1953 }
e8240f65 1954 mnt_drop_write_file(filp);
f424f664
JK		 1955got_it:
1956 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
1957 16))
1958 return -EFAULT;
1959 return 0;
1960}
1961
c1c1b583
CY		 1962static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
1963{
1964 struct inode *inode = file_inode(filp);
1965 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d530d4d8 1966 __u32 sync;
7fb17fe4 1967 int ret;
c1c1b583
CY		 1968
1969 if (!capable(CAP_SYS_ADMIN))
1970 return -EPERM;
1971
d530d4d8 1972 if (get_user(sync, (__u32 __user *)arg))
c1c1b583
CY		 1973 return -EFAULT;
1974
d530d4d8
CY		 1975 if (f2fs_readonly(sbi->sb))
1976 return -EROFS;
c1c1b583 1977
7fb17fe4
CY		 1978 ret = mnt_want_write_file(filp);
1979 if (ret)
1980 return ret;
1981
d530d4d8 1982 if (!sync) {
7fb17fe4
CY		 1983 if (!mutex_trylock(&sbi->gc_mutex)) {
1984 ret = -EBUSY;
1985 goto out;
1986 }
d530d4d8
CY		 1987 } else {
1988 mutex_lock(&sbi->gc_mutex);
c1c1b583
CY		 1989 }
1990
e066b83c 1991 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
7fb17fe4
CY		 1992out:
1993 mnt_drop_write_file(filp);
1994 return ret;
c1c1b583
CY		 1995}
1996
34dc77ad
JK		 1997static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
1998{
1999 struct inode *inode = file_inode(filp);
2000 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2001 struct f2fs_gc_range range;
2002 u64 end;
2003 int ret;
2004
2005 if (!capable(CAP_SYS_ADMIN))
2006 return -EPERM;
2007
2008 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2009 sizeof(range)))
2010 return -EFAULT;
2011
2012 if (f2fs_readonly(sbi->sb))
2013 return -EROFS;
2014
2015 ret = mnt_want_write_file(filp);
2016 if (ret)
2017 return ret;
2018
2019 end = range.start + range.len;
2020 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi))
2021 return -EINVAL;
2022do_more:
2023 if (!range.sync) {
2024 if (!mutex_trylock(&sbi->gc_mutex)) {
2025 ret = -EBUSY;
2026 goto out;
2027 }
2028 } else {
2029 mutex_lock(&sbi->gc_mutex);
2030 }
2031
2032 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2033 range.start += sbi->blocks_per_seg;
2034 if (range.start <= end)
2035 goto do_more;
2036out:
2037 mnt_drop_write_file(filp);
2038 return ret;
2039}
2040
456b88e4
CY		 2041static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2042{
2043 struct inode *inode = file_inode(filp);
2044 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
7fb17fe4 2045 int ret;
456b88e4
CY		 2046
2047 if (!capable(CAP_SYS_ADMIN))
2048 return -EPERM;
2049
2050 if (f2fs_readonly(sbi->sb))
2051 return -EROFS;
2052
7fb17fe4
CY		 2053 ret = mnt_want_write_file(filp);
2054 if (ret)
2055 return ret;
2056
2057 ret = f2fs_sync_fs(sbi->sb, 1);
2058
2059 mnt_drop_write_file(filp);
2060 return ret;
456b88e4
CY		 2061}
2062
d323d005
CY		 2063static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2064 struct file *filp,
2065 struct f2fs_defragment *range)
2066{
2067 struct inode *inode = file_inode(filp);
da85985c 2068 struct f2fs_map_blocks map = { .m_next_pgofs = NULL };
e15882b6 2069 struct extent_info ei = {0,0,0};
d323d005 2070 pgoff_t pg_start, pg_end;
3519e3f9 2071 unsigned int blk_per_seg = sbi->blocks_per_seg;
d323d005 2072 unsigned int total = 0, sec_num;
d323d005
CY		 2073 block_t blk_end = 0;
2074 bool fragmented = false;
2075 int err;
2076
2077 /* if in-place-update policy is enabled, don't waste time here */
7eab0c0d 2078 if (need_inplace_update_policy(inode, NULL))
d323d005
CY		 2079 return -EINVAL;
2080
09cbfeaf
KS		 2081 pg_start = range->start >> PAGE_SHIFT;
2082 pg_end = (range->start + range->len) >> PAGE_SHIFT;
d323d005 2083
2c4db1a6 2084 f2fs_balance_fs(sbi, true);
d323d005 2085
5955102c 2086 inode_lock(inode);
d323d005
CY		 2087
2088 /* writeback all dirty pages in the range */
2089 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
d8fe4f0e 2090 range->start + range->len - 1);
d323d005
CY		 2091 if (err)
2092 goto out;
2093
2094 /*
2095 * lookup mapping info in extent cache, skip defragmenting if physical
2096 * block addresses are continuous.
2097 */
2098 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2099 if (ei.fofs + ei.len >= pg_end)
2100 goto out;
2101 }
2102
2103 map.m_lblk = pg_start;
d323d005
CY		 2104
2105 /*
2106 * lookup mapping info in dnode page cache, skip defragmenting if all
2107 * physical block addresses are continuous even if there are hole(s)
2108 * in logical blocks.
2109 */
2110 while (map.m_lblk < pg_end) {
a1c1e9b7 2111 map.m_len = pg_end - map.m_lblk;
f2220c7f 2112 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2113 if (err)
2114 goto out;
2115
2116 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2117 map.m_lblk++;
d323d005
CY		 2118 continue;
2119 }
2120
2121 if (blk_end && blk_end != map.m_pblk) {
2122 fragmented = true;
2123 break;
2124 }
2125 blk_end = map.m_pblk + map.m_len;
2126
2127 map.m_lblk += map.m_len;
d323d005
CY		 2128 }
2129
2130 if (!fragmented)
2131 goto out;
2132
2133 map.m_lblk = pg_start;
2134 map.m_len = pg_end - pg_start;
2135
4ddb1a4d 2136 sec_num = (map.m_len + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
d323d005
CY		 2137
2138 /*
2139 * make sure there are enough free section for LFS allocation, this can
2140 * avoid defragment running in SSR mode when free section are allocated
2141 * intensively
2142 */
7f3037a5 2143 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
d323d005
CY		 2144 err = -EAGAIN;
2145 goto out;
2146 }
2147
2148 while (map.m_lblk < pg_end) {
2149 pgoff_t idx;
2150 int cnt = 0;
2151
2152do_map:
a1c1e9b7 2153 map.m_len = pg_end - map.m_lblk;
f2220c7f 2154 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2155 if (err)
2156 goto clear_out;
2157
2158 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2159 map.m_lblk++;
2160 continue;
2161 }
2162
91942321 2163 set_inode_flag(inode, FI_DO_DEFRAG);
d323d005
CY		 2164
2165 idx = map.m_lblk;
2166 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2167 struct page *page;
2168
2169 page = get_lock_data_page(inode, idx, true);
2170 if (IS_ERR(page)) {
2171 err = PTR_ERR(page);
2172 goto clear_out;
2173 }
2174
2175 set_page_dirty(page);
2176 f2fs_put_page(page, 1);
2177
2178 idx++;
2179 cnt++;
2180 total++;
2181 }
2182
2183 map.m_lblk = idx;
d323d005
CY		 2184
2185 if (idx < pg_end && cnt < blk_per_seg)
2186 goto do_map;
2187
91942321 2188 clear_inode_flag(inode, FI_DO_DEFRAG);
d323d005
CY		 2189
2190 err = filemap_fdatawrite(inode->i_mapping);
2191 if (err)
2192 goto out;
2193 }
2194clear_out:
91942321 2195 clear_inode_flag(inode, FI_DO_DEFRAG);
d323d005 2196out:
5955102c 2197 inode_unlock(inode);
d323d005 2198 if (!err)
09cbfeaf 2199 range->len = (u64)total << PAGE_SHIFT;
d323d005
CY		 2200 return err;
2201}
2202
2203static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2204{
2205 struct inode *inode = file_inode(filp);
2206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2207 struct f2fs_defragment range;
2208 int err;
2209
2210 if (!capable(CAP_SYS_ADMIN))
2211 return -EPERM;
2212
7eab0c0d 2213 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
d323d005
CY		 2214 return -EINVAL;
2215
d7563861
KM		 2216 if (f2fs_readonly(sbi->sb))
2217 return -EROFS;
d323d005
CY		 2218
2219 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
d7563861
KM		 2220 sizeof(range)))
2221 return -EFAULT;
d323d005
CY		 2222
2223 /* verify alignment of offset & size */
d7563861
KM		 2224 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2225 return -EINVAL;
d323d005 2226
1941d7bc 2227 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
d7563861
KM		 2228 sbi->max_file_blocks))
2229 return -EINVAL;
2230
2231 err = mnt_want_write_file(filp);
2232 if (err)
2233 return err;
1941d7bc 2234
d323d005 2235 err = f2fs_defragment_range(sbi, filp, &range);
d7563861
KM		 2236 mnt_drop_write_file(filp);
2237
d0239e1b 2238 f2fs_update_time(sbi, REQ_TIME);
d323d005 2239 if (err < 0)
d7563861 2240 return err;
d323d005
CY		 2241
2242 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2243 sizeof(range)))
d7563861
KM		 2244 return -EFAULT;
2245
2246 return 0;
d323d005
CY		 2247}
2248
4dd6f977
JK		 2249static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2250 struct file *file_out, loff_t pos_out, size_t len)
2251{
2252 struct inode *src = file_inode(file_in);
2253 struct inode *dst = file_inode(file_out);
2254 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2255 size_t olen = len, dst_max_i_size = 0;
2256 size_t dst_osize;
2257 int ret;
2258
2259 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2260 src->i_sb != dst->i_sb)
2261 return -EXDEV;
2262
2263 if (unlikely(f2fs_readonly(src->i_sb)))
2264 return -EROFS;
2265
fe8494bf
CY		 2266 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2267 return -EINVAL;
4dd6f977
JK		 2268
2269 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2270 return -EOPNOTSUPP;
2271
d95fd91c
FL		 2272 if (src == dst) {
2273 if (pos_in == pos_out)
2274 return 0;
2275 if (pos_out > pos_in && pos_out < pos_in + len)
2276 return -EINVAL;
2277 }
2278
4dd6f977 2279 inode_lock(src);
20a3d61d
CY		 2280 if (src != dst) {
2281 if (!inode_trylock(dst)) {
2282 ret = -EBUSY;
2283 goto out;
2284 }
2285 }
4dd6f977
JK		 2286
2287 ret = -EINVAL;
2288 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2289 goto out_unlock;
2290 if (len == 0)
2291 olen = len = src->i_size - pos_in;
2292 if (pos_in + len == src->i_size)
2293 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2294 if (len == 0) {
2295 ret = 0;
2296 goto out_unlock;
2297 }
2298
2299 dst_osize = dst->i_size;
2300 if (pos_out + olen > dst->i_size)
2301 dst_max_i_size = pos_out + olen;
2302
2303 /* verify the end result is block aligned */
2304 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2305 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2306 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2307 goto out_unlock;
2308
2309 ret = f2fs_convert_inline_inode(src);
2310 if (ret)
2311 goto out_unlock;
2312
2313 ret = f2fs_convert_inline_inode(dst);
2314 if (ret)
2315 goto out_unlock;
2316
2317 /* write out all dirty pages from offset */
2318 ret = filemap_write_and_wait_range(src->i_mapping,
2319 pos_in, pos_in + len);
2320 if (ret)
2321 goto out_unlock;
2322
2323 ret = filemap_write_and_wait_range(dst->i_mapping,
2324 pos_out, pos_out + len);
2325 if (ret)
2326 goto out_unlock;
2327
2328 f2fs_balance_fs(sbi, true);
2329 f2fs_lock_op(sbi);
61e4da11
FL		 2330 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2331 pos_out >> F2FS_BLKSIZE_BITS,
2332 len >> F2FS_BLKSIZE_BITS, false);
4dd6f977
JK		 2333
2334 if (!ret) {
2335 if (dst_max_i_size)
2336 f2fs_i_size_write(dst, dst_max_i_size);
2337 else if (dst_osize != dst->i_size)
2338 f2fs_i_size_write(dst, dst_osize);
2339 }
2340 f2fs_unlock_op(sbi);
2341out_unlock:
2342 if (src != dst)
2343 inode_unlock(dst);
20a3d61d 2344out:
4dd6f977
JK		 2345 inode_unlock(src);
2346 return ret;
2347}
2348
2349static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2350{
2351 struct f2fs_move_range range;
2352 struct fd dst;
2353 int err;
2354
2355 if (!(filp->f_mode & FMODE_READ) ||
2356 !(filp->f_mode & FMODE_WRITE))
2357 return -EBADF;
2358
2359 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2360 sizeof(range)))
2361 return -EFAULT;
2362
2363 dst = fdget(range.dst_fd);
2364 if (!dst.file)
2365 return -EBADF;
2366
2367 if (!(dst.file->f_mode & FMODE_WRITE)) {
2368 err = -EBADF;
2369 goto err_out;
2370 }
2371
2372 err = mnt_want_write_file(filp);
2373 if (err)
2374 goto err_out;
2375
2376 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2377 range.pos_out, range.len);
2378
2379 mnt_drop_write_file(filp);
3cecfa5f
KM		 2380 if (err)
2381 goto err_out;
4dd6f977
JK		 2382
2383 if (copy_to_user((struct f2fs_move_range __user *)arg,
2384 &range, sizeof(range)))
2385 err = -EFAULT;
2386err_out:
2387 fdput(dst);
2388 return err;
2389}
2390
e066b83c
JK		 2391static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2392{
2393 struct inode *inode = file_inode(filp);
2394 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2395 struct sit_info *sm = SIT_I(sbi);
2396 unsigned int start_segno = 0, end_segno = 0;
2397 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2398 struct f2fs_flush_device range;
2399 int ret;
2400
2401 if (!capable(CAP_SYS_ADMIN))
2402 return -EPERM;
2403
2404 if (f2fs_readonly(sbi->sb))
2405 return -EROFS;
2406
2407 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2408 sizeof(range)))
2409 return -EFAULT;
2410
2411 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2412 sbi->segs_per_sec != 1) {
2413 f2fs_msg(sbi->sb, KERN_WARNING,
2414 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2415 range.dev_num, sbi->s_ndevs,
2416 sbi->segs_per_sec);
2417 return -EINVAL;
2418 }
2419
2420 ret = mnt_want_write_file(filp);
2421 if (ret)
2422 return ret;
2423
2424 if (range.dev_num != 0)
2425 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2426 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2427
2428 start_segno = sm->last_victim[FLUSH_DEVICE];
2429 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2430 start_segno = dev_start_segno;
2431 end_segno = min(start_segno + range.segments, dev_end_segno);
2432
2433 while (start_segno < end_segno) {
2434 if (!mutex_trylock(&sbi->gc_mutex)) {
2435 ret = -EBUSY;
2436 goto out;
2437 }
2438 sm->last_victim[GC_CB] = end_segno + 1;
2439 sm->last_victim[GC_GREEDY] = end_segno + 1;
2440 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2441 ret = f2fs_gc(sbi, true, true, start_segno);
2442 if (ret == -EAGAIN)
2443 ret = 0;
2444 else if (ret < 0)
2445 break;
2446 start_segno++;
2447 }
2448out:
2449 mnt_drop_write_file(filp);
2450 return ret;
2451}
2452
e65ef207
JK		 2453static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2454{
2455 struct inode *inode = file_inode(filp);
2456 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2457
2458 /* Must validate to set it with SQLite behavior in Android. */
2459 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2460
2461 return put_user(sb_feature, (u32 __user *)arg);
2462}
e066b83c 2463
2c1d0305
CY		 2464#ifdef CONFIG_QUOTA
2465static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2466{
2467 struct inode *inode = file_inode(filp);
2468 struct f2fs_inode_info *fi = F2FS_I(inode);
2469 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2470 struct super_block *sb = sbi->sb;
2471 struct dquot *transfer_to[MAXQUOTAS] = {};
2472 struct page *ipage;
2473 kprojid_t kprojid;
2474 int err;
2475
2476 if (!f2fs_sb_has_project_quota(sb)) {
2477 if (projid != F2FS_DEF_PROJID)
2478 return -EOPNOTSUPP;
2479 else
2480 return 0;
2481 }
2482
2483 if (!f2fs_has_extra_attr(inode))
2484 return -EOPNOTSUPP;
2485
2486 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2487
2488 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2489 return 0;
2490
2491 err = mnt_want_write_file(filp);
2492 if (err)
2493 return err;
2494
2495 err = -EPERM;
2496 inode_lock(inode);
2497
2498 /* Is it quota file? Do not allow user to mess with it */
2499 if (IS_NOQUOTA(inode))
2500 goto out_unlock;
2501
2502 ipage = get_node_page(sbi, inode->i_ino);
2503 if (IS_ERR(ipage)) {
2504 err = PTR_ERR(ipage);
2505 goto out_unlock;
2506 }
2507
2508 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2509 i_projid)) {
2510 err = -EOVERFLOW;
2511 f2fs_put_page(ipage, 1);
2512 goto out_unlock;
2513 }
2514 f2fs_put_page(ipage, 1);
2515
2516 dquot_initialize(inode);
2517
2518 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2519 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2520 err = __dquot_transfer(inode, transfer_to);
2521 dqput(transfer_to[PRJQUOTA]);
2522 if (err)
2523 goto out_dirty;
2524 }
2525
2526 F2FS_I(inode)->i_projid = kprojid;
2527 inode->i_ctime = current_time(inode);
2528out_dirty:
2529 f2fs_mark_inode_dirty_sync(inode, true);
2530out_unlock:
2531 inode_unlock(inode);
2532 mnt_drop_write_file(filp);
2533 return err;
2534}
2535#else
2536static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2537{
2538 if (projid != F2FS_DEF_PROJID)
2539 return -EOPNOTSUPP;
2540 return 0;
2541}
2542#endif
2543
2544/* Transfer internal flags to xflags */
2545static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2546{
2547 __u32 xflags = 0;
2548
2549 if (iflags & FS_SYNC_FL)
2550 xflags |= FS_XFLAG_SYNC;
2551 if (iflags & FS_IMMUTABLE_FL)
2552 xflags |= FS_XFLAG_IMMUTABLE;
2553 if (iflags & FS_APPEND_FL)
2554 xflags |= FS_XFLAG_APPEND;
2555 if (iflags & FS_NODUMP_FL)
2556 xflags |= FS_XFLAG_NODUMP;
2557 if (iflags & FS_NOATIME_FL)
2558 xflags |= FS_XFLAG_NOATIME;
2559 if (iflags & FS_PROJINHERIT_FL)
2560 xflags |= FS_XFLAG_PROJINHERIT;
2561 return xflags;
2562}
2563
2564#define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2565 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2566 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2567
2568/* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
2569#define F2FS_FL_XFLAG_VISIBLE (FS_SYNC_FL | \
2570 FS_IMMUTABLE_FL | \
2571 FS_APPEND_FL | \
2572 FS_NODUMP_FL | \
2573 FS_NOATIME_FL | \
2574 FS_PROJINHERIT_FL)
2575
2576/* Transfer xflags flags to internal */
2577static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2578{
2579 unsigned long iflags = 0;
2580
2581 if (xflags & FS_XFLAG_SYNC)
2582 iflags |= FS_SYNC_FL;
2583 if (xflags & FS_XFLAG_IMMUTABLE)
2584 iflags |= FS_IMMUTABLE_FL;
2585 if (xflags & FS_XFLAG_APPEND)
2586 iflags |= FS_APPEND_FL;
2587 if (xflags & FS_XFLAG_NODUMP)
2588 iflags |= FS_NODUMP_FL;
2589 if (xflags & FS_XFLAG_NOATIME)
2590 iflags |= FS_NOATIME_FL;
2591 if (xflags & FS_XFLAG_PROJINHERIT)
2592 iflags |= FS_PROJINHERIT_FL;
2593
2594 return iflags;
2595}
2596
2597static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2598{
2599 struct inode *inode = file_inode(filp);
2600 struct f2fs_inode_info *fi = F2FS_I(inode);
2601 struct fsxattr fa;
2602
2603 memset(&fa, 0, sizeof(struct fsxattr));
2604 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2605 (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL));
2606
2607 if (f2fs_sb_has_project_quota(inode->i_sb))
2608 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2609 fi->i_projid);
2610
2611 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2612 return -EFAULT;
2613 return 0;
2614}
2615
2616static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2617{
2618 struct inode *inode = file_inode(filp);
2619 struct f2fs_inode_info *fi = F2FS_I(inode);
2620 struct fsxattr fa;
2621 unsigned int flags;
2622 int err;
2623
2624 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2625 return -EFAULT;
2626
2627 /* Make sure caller has proper permission */
2628 if (!inode_owner_or_capable(inode))
2629 return -EACCES;
2630
2631 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2632 return -EOPNOTSUPP;
2633
2634 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2635 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2636 return -EOPNOTSUPP;
2637
2638 err = mnt_want_write_file(filp);
2639 if (err)
2640 return err;
2641
2642 inode_lock(inode);
2643 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2644 (flags & F2FS_FL_XFLAG_VISIBLE);
2645 err = __f2fs_ioc_setflags(inode, flags);
2646 inode_unlock(inode);
2647 mnt_drop_write_file(filp);
2648 if (err)
2649 return err;
2650
2651 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2652 if (err)
2653 return err;
2654
2655 return 0;
2656}
e066b83c 2657
52656e6c
JK		 2658long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2659{
1f227a3e
JK		 2660 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2661 return -EIO;
2662
52656e6c
JK		 2663 switch (cmd) {
2664 case F2FS_IOC_GETFLAGS:
2665 return f2fs_ioc_getflags(filp, arg);
2666 case F2FS_IOC_SETFLAGS:
2667 return f2fs_ioc_setflags(filp, arg);
d49f3e89
CY		 2668 case F2FS_IOC_GETVERSION:
2669 return f2fs_ioc_getversion(filp, arg);
88b88a66
JK		 2670 case F2FS_IOC_START_ATOMIC_WRITE:
2671 return f2fs_ioc_start_atomic_write(filp);
2672 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2673 return f2fs_ioc_commit_atomic_write(filp);
02a1335f
JK		 2674 case F2FS_IOC_START_VOLATILE_WRITE:
2675 return f2fs_ioc_start_volatile_write(filp);
1e84371f
JK		 2676 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2677 return f2fs_ioc_release_volatile_write(filp);
2678 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2679 return f2fs_ioc_abort_volatile_write(filp);
1abff93d
JK		 2680 case F2FS_IOC_SHUTDOWN:
2681 return f2fs_ioc_shutdown(filp, arg);
52656e6c
JK		 2682 case FITRIM:
2683 return f2fs_ioc_fitrim(filp, arg);
f424f664
JK		 2684 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2685 return f2fs_ioc_set_encryption_policy(filp, arg);
2686 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2687 return f2fs_ioc_get_encryption_policy(filp, arg);
2688 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2689 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
c1c1b583
CY		 2690 case F2FS_IOC_GARBAGE_COLLECT:
2691 return f2fs_ioc_gc(filp, arg);
34dc77ad
JK		 2692 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2693 return f2fs_ioc_gc_range(filp, arg);
456b88e4
CY		 2694 case F2FS_IOC_WRITE_CHECKPOINT:
2695 return f2fs_ioc_write_checkpoint(filp, arg);
d323d005
CY		 2696 case F2FS_IOC_DEFRAGMENT:
2697 return f2fs_ioc_defragment(filp, arg);
4dd6f977
JK		 2698 case F2FS_IOC_MOVE_RANGE:
2699 return f2fs_ioc_move_range(filp, arg);
e066b83c
JK		 2700 case F2FS_IOC_FLUSH_DEVICE:
2701 return f2fs_ioc_flush_device(filp, arg);
e65ef207
JK		 2702 case F2FS_IOC_GET_FEATURES:
2703 return f2fs_ioc_get_features(filp, arg);
2c1d0305
CY		 2704 case F2FS_IOC_FSGETXATTR:
2705 return f2fs_ioc_fsgetxattr(filp, arg);
2706 case F2FS_IOC_FSSETXATTR:
2707 return f2fs_ioc_fssetxattr(filp, arg);
fbfa2cc5
JK		 2708 default:
2709 return -ENOTTY;
2710 }
2711}
2712
fcc85a4d
JK		 2713static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2714{
b439b103
JK		 2715 struct file *file = iocb->ki_filp;
2716 struct inode *inode = file_inode(file);
9dfa1baf 2717 struct blk_plug plug;
b439b103 2718 ssize_t ret;
fcc85a4d 2719
1f227a3e
JK		 2720 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2721 return -EIO;
2722
b439b103
JK		 2723 inode_lock(inode);
2724 ret = generic_write_checks(iocb, from);
2725 if (ret > 0) {
dc91de78 2726 int err;
a7de6086 2727
dc91de78
JK		 2728 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
2729 set_inode_flag(inode, FI_NO_PREALLOC);
a7de6086 2730
dc91de78 2731 err = f2fs_preallocate_blocks(iocb, from);
a7de6086
JK		 2732 if (err) {
2733 inode_unlock(inode);
2734 return err;
9dfa1baf 2735 }
a7de6086
JK		 2736 blk_start_plug(&plug);
2737 ret = __generic_file_write_iter(iocb, from);
2738 blk_finish_plug(&plug);
dc91de78 2739 clear_inode_flag(inode, FI_NO_PREALLOC);
b0af6d49
CY		 2740
2741 if (ret > 0)
2742 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
b439b103
JK		 2743 }
2744 inode_unlock(inode);
2745
e2592217
CH		 2746 if (ret > 0)
2747 ret = generic_write_sync(iocb, ret);
b439b103 2748 return ret;
fcc85a4d
JK		 2749}
2750
e9750824
NJ		 2751#ifdef CONFIG_COMPAT
2752long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
2753{
2754 switch (cmd) {
2755 case F2FS_IOC32_GETFLAGS:
2756 cmd = F2FS_IOC_GETFLAGS;
2757 break;
2758 case F2FS_IOC32_SETFLAGS:
2759 cmd = F2FS_IOC_SETFLAGS;
2760 break;
04ef4b62
CY		 2761 case F2FS_IOC32_GETVERSION:
2762 cmd = F2FS_IOC_GETVERSION;
2763 break;
2764 case F2FS_IOC_START_ATOMIC_WRITE:
2765 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2766 case F2FS_IOC_START_VOLATILE_WRITE:
2767 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2768 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2769 case F2FS_IOC_SHUTDOWN:
2770 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2771 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2772 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2773 case F2FS_IOC_GARBAGE_COLLECT:
34dc77ad 2774 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
04ef4b62
CY		 2775 case F2FS_IOC_WRITE_CHECKPOINT:
2776 case F2FS_IOC_DEFRAGMENT:
4dd6f977 2777 case F2FS_IOC_MOVE_RANGE:
e066b83c 2778 case F2FS_IOC_FLUSH_DEVICE:
e65ef207 2779 case F2FS_IOC_GET_FEATURES:
2c1d0305
CY		 2780 case F2FS_IOC_FSGETXATTR:
2781 case F2FS_IOC_FSSETXATTR:
4dd6f977 2782 break;
e9750824
NJ		 2783 default:
2784 return -ENOIOCTLCMD;
2785 }
2786 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2787}
2788#endif
2789
fbfa2cc5 2790const struct file_operations f2fs_file_operations = {
267378d4 2791 .llseek = f2fs_llseek,
aad4f8bb 2792 .read_iter = generic_file_read_iter,
fcc85a4d
JK		 2793 .write_iter = f2fs_file_write_iter,
2794 .open = f2fs_file_open,
12662234 2795 .release = f2fs_release_file,
fbfa2cc5 2796 .mmap = f2fs_file_mmap,
7a10f017 2797 .flush = f2fs_file_flush,
fbfa2cc5
JK		 2798 .fsync = f2fs_sync_file,
2799 .fallocate = f2fs_fallocate,
2800 .unlocked_ioctl = f2fs_ioctl,
e9750824
NJ		 2801#ifdef CONFIG_COMPAT
2802 .compat_ioctl = f2fs_compat_ioctl,
2803#endif
fbfa2cc5 2804 .splice_read = generic_file_splice_read,
8d020765 2805 .splice_write = iter_file_splice_write,
fbfa2cc5 2806};
Linux 6.x block layer and io_uring tree(s)