projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: reject test_dummy_encryption when !CONFIG_FS_ENCRYPTION
[linux-block.git] / fs / f2fs / file.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
fbfa2cc5
JK		 3 * fs/f2fs/file.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
fbfa2cc5
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/stat.h>
11#include <linux/buffer_head.h>
12#include <linux/writeback.h>
ae51fb31 13#include <linux/blkdev.h>
fbfa2cc5
JK		 14#include <linux/falloc.h>
15#include <linux/types.h>
e9750824 16#include <linux/compat.h>
fbfa2cc5
JK		 17#include <linux/uaccess.h>
18#include <linux/mount.h>
7f7670fe 19#include <linux/pagevec.h>
dc91de78 20#include <linux/uio.h>
8da4b8c4 21#include <linux/uuid.h>
4dd6f977 22#include <linux/file.h>
4507847c 23#include <linux/nls.h>
9af84648 24#include <linux/sched/signal.h>
9b1bb01c 25#include <linux/fileattr.h>
0f6b56ec 26#include <linux/fadvise.h>
a1e09b03 27#include <linux/iomap.h>
fbfa2cc5
JK		 28
29#include "f2fs.h"
30#include "node.h"
31#include "segment.h"
32#include "xattr.h"
33#include "acl.h"
c1c1b583 34#include "gc.h"
52118743 35#include "iostat.h"
a2a4a7e4 36#include <trace/events/f2fs.h>
fa4320ce 37#include <uapi/linux/f2fs.h>
fbfa2cc5 38
ea4d479b 39static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
5a3a2d83
QS		 40{
41 struct inode *inode = file_inode(vmf->vma->vm_file);
ea4d479b 42 vm_fault_t ret;
5a3a2d83 43
ea4d479b 44 ret = filemap_fault(vmf);
8b83ac81
CY		 45 if (!ret)
46 f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
47 F2FS_BLKSIZE);
48
d7648343
CY		 49 trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
50
ea4d479b 51 return ret;
5a3a2d83
QS		 52}
53
ea4d479b 54static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
fbfa2cc5
JK		 55{
56 struct page *page = vmf->page;
11bac800 57 struct inode *inode = file_inode(vmf->vma->vm_file);
4081363f 58 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
bdf03299 59 struct dnode_of_data dn;
4c8ff709
CY		 60 bool need_alloc = true;
61 int err = 0;
fbfa2cc5 62
e0fcd015
CY		 63 if (unlikely(IS_IMMUTABLE(inode)))
64 return VM_FAULT_SIGBUS;
65
c6140415
JK		 66 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
67 return VM_FAULT_SIGBUS;
68
1f227a3e
JK		 69 if (unlikely(f2fs_cp_error(sbi))) {
70 err = -EIO;
71 goto err;
72 }
73
00e09c0b
CY		 74 if (!f2fs_is_checkpoint_ready(sbi)) {
75 err = -ENOSPC;
955ebcd3 76 goto err;
00e09c0b 77 }
1f227a3e 78
c8e43d55
CY		 79 err = f2fs_convert_inline_inode(inode);
80 if (err)
81 goto err;
82
4c8ff709
CY		 83#ifdef CONFIG_F2FS_FS_COMPRESSION
84 if (f2fs_compressed_file(inode)) {
85 int ret = f2fs_is_compressed_cluster(inode, page->index);
86
87 if (ret < 0) {
88 err = ret;
89 goto err;
90 } else if (ret) {
4c8ff709
CY		 91 need_alloc = false;
92 }
93 }
94#endif
bdf03299 95 /* should do out of any locked page */
4c8ff709
CY		 96 if (need_alloc)
97 f2fs_balance_fs(sbi, true);
bdf03299 98
fbfa2cc5 99 sb_start_pagefault(inode->i_sb);
b3d208f9
JK		 100
101 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
b067ba1f 102
11bac800 103 file_update_time(vmf->vma->vm_file);
edc6d01b 104 filemap_invalidate_lock_shared(inode->i_mapping);
fbfa2cc5 105 lock_page(page);
6bacf52f 106 if (unlikely(page->mapping != inode->i_mapping ||
9851e6e1 107 page_offset(page) > i_size_read(inode) ||
6bacf52f 108 !PageUptodate(page))) {
fbfa2cc5
JK		 109 unlock_page(page);
110 err = -EFAULT;
5a3a2d83 111 goto out_sem;
fbfa2cc5
JK		 112 }
113
4c8ff709
CY		 114 if (need_alloc) {
115 /* block allocation */
0ef81833 116 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
4c8ff709
CY		 117 set_new_dnode(&dn, inode, NULL, NULL, 0);
118 err = f2fs_get_block(&dn, page->index);
0ef81833 119 f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
39a86958
CY		 120 }
121
06c7540f
CY		 122#ifdef CONFIG_F2FS_FS_COMPRESSION
123 if (!need_alloc) {
124 set_new_dnode(&dn, inode, NULL, NULL, 0);
125 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
126 f2fs_put_dnode(&dn);
127 }
128#endif
129 if (err) {
130 unlock_page(page);
131 goto out_sem;
39a86958
CY		 132 }
133
bae0ee7a 134 f2fs_wait_on_page_writeback(page, DATA, false, true);
39a86958
CY		 135
136 /* wait for GCed page writeback via META_MAPPING */
137 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
138
fbfa2cc5
JK		 139 /*
140 * check to see if the page is mapped already (no holes)
141 */
142 if (PageMappedToDisk(page))
39a86958 143 goto out_sem;
fbfa2cc5
JK		 144
145 /* page is wholly or partially inside EOF */
09cbfeaf 146 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
9edcdabf 147 i_size_read(inode)) {
193bea1d 148 loff_t offset;
f11e98bd 149
09cbfeaf
KS		 150 offset = i_size_read(inode) & ~PAGE_MASK;
151 zero_user_segment(page, offset, PAGE_SIZE);
fbfa2cc5
JK		 152 }
153 set_page_dirty(page);
237c0790
JK		 154 if (!PageUptodate(page))
155 SetPageUptodate(page);
fbfa2cc5 156
b0af6d49 157 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
c75f2feb 158 f2fs_update_time(sbi, REQ_TIME);
b0af6d49 159
e943a10d 160 trace_f2fs_vm_page_mkwrite(page, DATA);
5a3a2d83 161out_sem:
edc6d01b 162 filemap_invalidate_unlock_shared(inode->i_mapping);
39a86958 163
fbfa2cc5 164 sb_end_pagefault(inode->i_sb);
1f227a3e 165err:
fbfa2cc5
JK		 166 return block_page_mkwrite_return(err);
167}
168
169static const struct vm_operations_struct f2fs_file_vm_ops = {
5a3a2d83 170 .fault = f2fs_filemap_fault,
f1820361 171 .map_pages = filemap_map_pages,
692bb55d 172 .page_mkwrite = f2fs_vm_page_mkwrite,
fbfa2cc5
JK		 173};
174
354a3399
JK		 175static int get_parent_ino(struct inode *inode, nid_t *pino)
176{
177 struct dentry *dentry;
178
84c9c2de
EB		 179 /*
180 * Make sure to get the non-deleted alias. The alias associated with
181 * the open file descriptor being fsync()'ed may be deleted already.
182 */
183 dentry = d_find_alias(inode);
354a3399
JK		 184 if (!dentry)
185 return 0;
186
f0947e5c
JK		 187 *pino = parent_ino(dentry);
188 dput(dentry);
354a3399
JK		 189 return 1;
190}
191
a5fd5050 192static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
9d1589ef 193{
4081363f 194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a5fd5050 195 enum cp_reason_type cp_reason = CP_NO_NEEDED;
9d1589ef 196
a5fd5050
CY		 197 if (!S_ISREG(inode->i_mode))
198 cp_reason = CP_NON_REGULAR;
4c8ff709
CY		 199 else if (f2fs_compressed_file(inode))
200 cp_reason = CP_COMPRESSED;
a5fd5050
CY		 201 else if (inode->i_nlink != 1)
202 cp_reason = CP_HARDLINK;
bbf156f7 203 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
a5fd5050 204 cp_reason = CP_SB_NEED_CP;
9d1589ef 205 else if (file_wrong_pino(inode))
a5fd5050 206 cp_reason = CP_WRONG_PINO;
4d57b86d 207 else if (!f2fs_space_for_roll_forward(sbi))
a5fd5050 208 cp_reason = CP_NO_SPC_ROLL;
4d57b86d 209 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
a5fd5050 210 cp_reason = CP_NODE_NEED_CP;
d5053a34 211 else if (test_opt(sbi, FASTBOOT))
a5fd5050 212 cp_reason = CP_FASTBOOT_MODE;
63189b78 213 else if (F2FS_OPTION(sbi).active_logs == 2)
a5fd5050 214 cp_reason = CP_SPEC_LOG_NUM;
63189b78 215 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
4d57b86d
CY		 216 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
217 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
218 TRANS_DIR_INO))
0a007b97 219 cp_reason = CP_RECOVER_DIR;
9d1589ef 220
a5fd5050 221 return cp_reason;
9d1589ef
CY		 222}
223
9c7bb702
CL		 224static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
225{
226 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
227 bool ret = false;
228 /* But we need to avoid that there are some inode updates */
4d57b86d 229 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
9c7bb702
CL		 230 ret = true;
231 f2fs_put_page(i, 0);
232 return ret;
233}
234
51455b19
CL		 235static void try_to_fix_pino(struct inode *inode)
236{
237 struct f2fs_inode_info *fi = F2FS_I(inode);
238 nid_t pino;
239
e4544b63 240 f2fs_down_write(&fi->i_sem);
51455b19
CL		 241 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
242 get_parent_ino(inode, &pino)) {
205b9822 243 f2fs_i_pino_write(inode, pino);
51455b19 244 file_got_pino(inode);
51455b19 245 }
e4544b63 246 f2fs_up_write(&fi->i_sem);
51455b19
CL		 247}
248
608514de
JK		 249static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
250 int datasync, bool atomic)
fbfa2cc5
JK		 251{
252 struct inode *inode = file->f_mapping->host;
4081363f 253 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2403c155 254 nid_t ino = inode->i_ino;
fbfa2cc5 255 int ret = 0;
a5fd5050 256 enum cp_reason_type cp_reason = 0;
fbfa2cc5 257 struct writeback_control wbc = {
c81bf1c8 258 .sync_mode = WB_SYNC_ALL,
fbfa2cc5
JK		 259 .nr_to_write = LONG_MAX,
260 .for_reclaim = 0,
261 };
50fa53ec 262 unsigned int seq_id = 0;
fbfa2cc5 263
dddd3d65 264 if (unlikely(f2fs_readonly(inode->i_sb)))
1fa95b0b
NJ		 265 return 0;
266
a2a4a7e4 267 trace_f2fs_sync_file_enter(inode);
ea1aa12c 268
b61ac5b7
YH		 269 if (S_ISDIR(inode->i_mode))
270 goto go_write;
271
ea1aa12c 272 /* if fdatasync is triggered, let's do in-place-update */
c46a155b 273 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
91942321 274 set_inode_flag(inode, FI_NEED_IPU);
3b49c9a1 275 ret = file_write_and_wait_range(file, start, end);
91942321 276 clear_inode_flag(inode, FI_NEED_IPU);
c1ce1b02 277
dddd3d65 278 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
a5fd5050 279 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
fbfa2cc5 280 return ret;
a2a4a7e4 281 }
fbfa2cc5 282
9c7bb702 283 /* if the inode is dirty, let's recover all the time */
281518c6 284 if (!f2fs_skip_inode_update(inode, datasync)) {
2286c020 285 f2fs_write_inode(inode, NULL);
9c7bb702
CL		 286 goto go_write;
287 }
288
6d99ba41
JK		 289 /*
290 * if there is no written data, don't waste time to write recovery info.
291 */
91942321 292 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
4d57b86d 293 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
19c9c466 294
9c7bb702
CL		 295 /* it may call write_inode just prior to fsync */
296 if (need_inode_page_update(sbi, ino))
19c9c466 297 goto go_write;
19c9c466 298
91942321 299 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
4d57b86d 300 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
6d99ba41
JK		 301 goto flush_out;
302 goto out;
27879915
CY		 303 } else {
304 /*
305 * for OPU case, during fsync(), node can be persisted before
306 * data when lower device doesn't support write barrier, result
307 * in data corruption after SPO.
308 * So for strict fsync mode, force to use atomic write sematics
309 * to keep write order in between data/node and last node to
310 * avoid potential data corruption.
311 */
312 if (F2FS_OPTION(sbi).fsync_mode ==
313 FSYNC_MODE_STRICT && !atomic)
314 atomic = true;
6d99ba41 315 }
19c9c466 316go_write:
e5d2385e
JK		 317 /*
318 * Both of fdatasync() and fsync() are able to be recovered from
319 * sudden-power-off.
320 */
e4544b63 321 f2fs_down_read(&F2FS_I(inode)->i_sem);
a5fd5050 322 cp_reason = need_do_checkpoint(inode);
e4544b63 323 f2fs_up_read(&F2FS_I(inode)->i_sem);
d928bfbf 324
a5fd5050 325 if (cp_reason) {
fbfa2cc5
JK		 326 /* all the dirty node pages should be flushed for POR */
327 ret = f2fs_sync_fs(inode->i_sb, 1);
d928bfbf 328
51455b19
CL		 329 /*
330 * We've secured consistency through sync_fs. Following pino
331 * will be used only for fsynced inodes after checkpoint.
332 */
333 try_to_fix_pino(inode);
91942321
JK		 334 clear_inode_flag(inode, FI_APPEND_WRITE);
335 clear_inode_flag(inode, FI_UPDATE_WRITE);
51455b19
CL		 336 goto out;
337 }
88bd02c9 338sync_nodes:
c29fd0c0 339 atomic_inc(&sbi->wb_sync_req[NODE]);
50fa53ec 340 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
c29fd0c0 341 atomic_dec(&sbi->wb_sync_req[NODE]);
c267ec15
JK		 342 if (ret)
343 goto out;
51455b19 344
871f599f 345 /* if cp_error was enabled, we should avoid infinite loop */
6d5a1495
CY		 346 if (unlikely(f2fs_cp_error(sbi))) {
347 ret = -EIO;
871f599f 348 goto out;
6d5a1495 349 }
871f599f 350
4d57b86d 351 if (f2fs_need_inode_block_update(sbi, ino)) {
7c45729a 352 f2fs_mark_inode_dirty_sync(inode, true);
51455b19
CL		 353 f2fs_write_inode(inode, NULL);
354 goto sync_nodes;
fbfa2cc5 355 }
51455b19 356
b6a245eb
JK		 357 /*
358 * If it's atomic_write, it's just fine to keep write ordering. So
359 * here we don't need to wait for node write completion, since we use
360 * node chain which serializes node blocks. If one of node writes are
361 * reordered, we can see simply broken chain, resulting in stopping
362 * roll-forward recovery. It means we'll recover all or none node blocks
363 * given fsync mark.
364 */
365 if (!atomic) {
50fa53ec 366 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
b6a245eb
JK		 367 if (ret)
368 goto out;
369 }
51455b19
CL		 370
371 /* once recovery info is written, don't need to tack this */
4d57b86d 372 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
91942321 373 clear_inode_flag(inode, FI_APPEND_WRITE);
51455b19 374flush_out:
c550e25b
JK		 375 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
376 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
39d787be 377 ret = f2fs_issue_flush(sbi, inode->i_ino);
3f06252f 378 if (!ret) {
4d57b86d 379 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
3f06252f 380 clear_inode_flag(inode, FI_UPDATE_WRITE);
4d57b86d 381 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
3f06252f 382 }
d0239e1b 383 f2fs_update_time(sbi, REQ_TIME);
fbfa2cc5 384out:
a5fd5050 385 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
fbfa2cc5
JK		 386 return ret;
387}
388
608514de
JK		 389int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
390{
1f227a3e
JK		 391 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
392 return -EIO;
608514de
JK		 393 return f2fs_do_sync_file(file, start, end, datasync, false);
394}
395
4cb03fec
MWO		 396static bool __found_offset(struct address_space *mapping, block_t blkaddr,
397 pgoff_t index, int whence)
7f7670fe
JK		 398{
399 switch (whence) {
400 case SEEK_DATA:
4cb03fec
MWO		 401 if (__is_valid_data_blkaddr(blkaddr))
402 return true;
403 if (blkaddr == NEW_ADDR &&
404 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
7f7670fe
JK		 405 return true;
406 break;
407 case SEEK_HOLE:
408 if (blkaddr == NULL_ADDR)
409 return true;
410 break;
411 }
412 return false;
413}
414
267378d4
CY		 415static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
416{
417 struct inode *inode = file->f_mapping->host;
418 loff_t maxbytes = inode->i_sb->s_maxbytes;
419 struct dnode_of_data dn;
4cb03fec 420 pgoff_t pgofs, end_offset;
7f7670fe
JK		 421 loff_t data_ofs = offset;
422 loff_t isize;
267378d4
CY		 423 int err = 0;
424
5955102c 425 inode_lock(inode);
267378d4
CY		 426
427 isize = i_size_read(inode);
428 if (offset >= isize)
429 goto fail;
430
431 /* handle inline data case */
7a6e59d7
CY		 432 if (f2fs_has_inline_data(inode)) {
433 if (whence == SEEK_HOLE) {
434 data_ofs = isize;
435 goto found;
436 } else if (whence == SEEK_DATA) {
437 data_ofs = offset;
438 goto found;
439 }
267378d4
CY		 440 }
441
09cbfeaf 442 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
267378d4 443
09cbfeaf 444 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 445 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 446 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
267378d4
CY		 447 if (err && err != -ENOENT) {
448 goto fail;
449 } else if (err == -ENOENT) {
e1c42045 450 /* direct node does not exists */
267378d4 451 if (whence == SEEK_DATA) {
4d57b86d 452 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
267378d4
CY		 453 continue;
454 } else {
455 goto found;
456 }
457 }
458
81ca7350 459 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
267378d4
CY		 460
461 /* find data/hole in dnode block */
462 for (; dn.ofs_in_node < end_offset;
463 dn.ofs_in_node++, pgofs++,
09cbfeaf 464 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 465 block_t blkaddr;
f11e98bd 466
a2ced1ce 467 blkaddr = f2fs_data_blkaddr(&dn);
267378d4 468
c9b60788
CY		 469 if (__is_valid_data_blkaddr(blkaddr) &&
470 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
93770ab7 471 blkaddr, DATA_GENERIC_ENHANCE)) {
c9b60788
CY		 472 f2fs_put_dnode(&dn);
473 goto fail;
474 }
475
4cb03fec 476 if (__found_offset(file->f_mapping, blkaddr,
e1da7872 477 pgofs, whence)) {
267378d4
CY		 478 f2fs_put_dnode(&dn);
479 goto found;
480 }
481 }
482 f2fs_put_dnode(&dn);
483 }
484
485 if (whence == SEEK_DATA)
486 goto fail;
267378d4 487found:
fe369bc8
JK		 488 if (whence == SEEK_HOLE && data_ofs > isize)
489 data_ofs = isize;
5955102c 490 inode_unlock(inode);
267378d4
CY		 491 return vfs_setpos(file, data_ofs, maxbytes);
492fail:
5955102c 493 inode_unlock(inode);
267378d4
CY		 494 return -ENXIO;
495}
496
497static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
498{
499 struct inode *inode = file->f_mapping->host;
500 loff_t maxbytes = inode->i_sb->s_maxbytes;
501
6d1451bf
CX		 502 if (f2fs_compressed_file(inode))
503 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
504
267378d4
CY		 505 switch (whence) {
506 case SEEK_SET:
507 case SEEK_CUR:
508 case SEEK_END:
509 return generic_file_llseek_size(file, offset, whence,
510 maxbytes, i_size_read(inode));
511 case SEEK_DATA:
512 case SEEK_HOLE:
0b4c5afd
JK		 513 if (offset < 0)
514 return -ENXIO;
267378d4
CY		 515 return f2fs_seek_block(file, offset, whence);
516 }
517
518 return -EINVAL;
519}
520
fbfa2cc5
JK		 521static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
522{
b3d208f9
JK		 523 struct inode *inode = file_inode(file);
524
1f227a3e
JK		 525 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
526 return -EIO;
527
4c8ff709
CY		 528 if (!f2fs_is_compress_backend_ready(inode))
529 return -EOPNOTSUPP;
530
fbfa2cc5
JK		 531 file_accessed(file);
532 vma->vm_ops = &f2fs_file_vm_ops;
4c8ff709 533 set_inode_flag(inode, FI_MMAP_FILE);
fbfa2cc5
JK		 534 return 0;
535}
536
fcc85a4d
JK		 537static int f2fs_file_open(struct inode *inode, struct file *filp)
538{
2e168c82 539 int err = fscrypt_file_open(inode, filp);
fcc85a4d 540
95ae251f
EB		 541 if (err)
542 return err;
543
4c8ff709
CY		 544 if (!f2fs_is_compress_backend_ready(inode))
545 return -EOPNOTSUPP;
546
95ae251f 547 err = fsverity_file_open(inode, filp);
2e168c82
EB		 548 if (err)
549 return err;
b91050a8
HL		 550
551 filp->f_mode |= FMODE_NOWAIT;
552
0abd675e 553 return dquot_file_open(inode, filp);
fcc85a4d
JK		 554}
555
4d57b86d 556void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
fbfa2cc5 557{
4081363f 558 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
fbfa2cc5 559 struct f2fs_node *raw_node;
19b2c30d 560 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
fbfa2cc5 561 __le32 *addr;
7a2af766 562 int base = 0;
4c8ff709
CY		 563 bool compressed_cluster = false;
564 int cluster_index = 0, valid_blocks = 0;
565 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
c2759eba 566 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
7a2af766
CY		 567
568 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
569 base = get_extra_isize(dn->inode);
fbfa2cc5 570
45590710 571 raw_node = F2FS_NODE(dn->node_page);
7a2af766 572 addr = blkaddr_in_node(raw_node) + base + ofs;
fbfa2cc5 573
4c8ff709
CY		 574 /* Assumption: truncateion starts with cluster */
575 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
fbfa2cc5 576 block_t blkaddr = le32_to_cpu(*addr);
f11e98bd 577
4c8ff709
CY		 578 if (f2fs_compressed_file(dn->inode) &&
579 !(cluster_index & (cluster_size - 1))) {
580 if (compressed_cluster)
581 f2fs_i_compr_blocks_update(dn->inode,
582 valid_blocks, false);
583 compressed_cluster = (blkaddr == COMPRESS_ADDR);
584 valid_blocks = 0;
585 }
586
fbfa2cc5
JK		 587 if (blkaddr == NULL_ADDR)
588 continue;
589
e1509cf2 590 dn->data_blkaddr = NULL_ADDR;
4d57b86d 591 f2fs_set_data_blkaddr(dn);
c9b60788 592
4c8ff709
CY		 593 if (__is_valid_data_blkaddr(blkaddr)) {
594 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
93770ab7 595 DATA_GENERIC_ENHANCE))
4c8ff709
CY		 596 continue;
597 if (compressed_cluster)
598 valid_blocks++;
599 }
c9b60788 600
3c6c2beb 601 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
91942321 602 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
4c8ff709
CY		 603
604 f2fs_invalidate_blocks(sbi, blkaddr);
ef8d563f
CY		 605
606 if (!released || blkaddr != COMPRESS_ADDR)
607 nr_free++;
fbfa2cc5 608 }
19b2c30d 609
4c8ff709
CY		 610 if (compressed_cluster)
611 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
612
fbfa2cc5 613 if (nr_free) {
19b2c30d
CY		 614 pgoff_t fofs;
615 /*
616 * once we invalidate valid blkaddr in range [ofs, ofs + count],
617 * we will invalidate all blkaddr in the whole range.
618 */
4d57b86d 619 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
81ca7350 620 dn->inode) + ofs;
19b2c30d 621 f2fs_update_extent_cache_range(dn, fofs, 0, len);
d7cc950b 622 dec_valid_block_count(sbi, dn->inode, nr_free);
fbfa2cc5
JK		 623 }
624 dn->ofs_in_node = ofs;
51dd6249 625
d0239e1b 626 f2fs_update_time(sbi, REQ_TIME);
51dd6249
NJ		 627 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
628 dn->ofs_in_node, nr_free);
fbfa2cc5
JK		 629}
630
4d57b86d 631void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
fbfa2cc5 632{
d02a6e61 633 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
fbfa2cc5
JK		 634}
635
0bfcfcca 636static int truncate_partial_data_page(struct inode *inode, u64 from,
43f3eae1 637 bool cache_only)
fbfa2cc5 638{
193bea1d 639 loff_t offset = from & (PAGE_SIZE - 1);
09cbfeaf 640 pgoff_t index = from >> PAGE_SHIFT;
43f3eae1 641 struct address_space *mapping = inode->i_mapping;
fbfa2cc5
JK		 642 struct page *page;
643
43f3eae1 644 if (!offset && !cache_only)
b3d208f9 645 return 0;
fbfa2cc5 646
43f3eae1 647 if (cache_only) {
34b5d5c2 648 page = find_lock_page(mapping, index);
43f3eae1
JK		 649 if (page && PageUptodate(page))
650 goto truncate_out;
651 f2fs_put_page(page, 1);
b3d208f9 652 return 0;
43f3eae1 653 }
fbfa2cc5 654
4d57b86d 655 page = f2fs_get_lock_data_page(inode, index, true);
43f3eae1 656 if (IS_ERR(page))
a78aaa2c 657 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
43f3eae1 658truncate_out:
bae0ee7a 659 f2fs_wait_on_page_writeback(page, DATA, true, true);
09cbfeaf 660 zero_user(page, offset, PAGE_SIZE - offset);
a9bcf9bc
JK		 661
662 /* An encrypted inode should have a key and truncate the last page. */
62230e0d 663 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
a9bcf9bc 664 if (!cache_only)
0bfcfcca 665 set_page_dirty(page);
fbfa2cc5 666 f2fs_put_page(page, 1);
b3d208f9 667 return 0;
fbfa2cc5
JK		 668}
669
3265d3db 670int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
fbfa2cc5 671{
4081363f 672 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 673 struct dnode_of_data dn;
674 pgoff_t free_from;
9ffe0fb5 675 int count = 0, err = 0;
b3d208f9 676 struct page *ipage;
0bfcfcca 677 bool truncate_page = false;
fbfa2cc5 678
51dd6249
NJ		 679 trace_f2fs_truncate_blocks_enter(inode, from);
680
df033caf 681 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
fbfa2cc5 682
6d1451bf 683 if (free_from >= max_file_blocks(inode))
09210c97
CY		 684 goto free_partial;
685
764aa3e9 686 if (lock)
c42d28ce 687 f2fs_lock_op(sbi);
9ffe0fb5 688
4d57b86d 689 ipage = f2fs_get_node_page(sbi, inode->i_ino);
b3d208f9
JK		 690 if (IS_ERR(ipage)) {
691 err = PTR_ERR(ipage);
692 goto out;
693 }
694
695 if (f2fs_has_inline_data(inode)) {
4d57b86d 696 f2fs_truncate_inline_inode(inode, ipage, from);
b3d208f9 697 f2fs_put_page(ipage, 1);
0bfcfcca 698 truncate_page = true;
b3d208f9
JK		 699 goto out;
700 }
701
702 set_new_dnode(&dn, inode, ipage, NULL, 0);
4d57b86d 703 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
fbfa2cc5
JK		 704 if (err) {
705 if (err == -ENOENT)
706 goto free_next;
b3d208f9 707 goto out;
1ce86bf6
JK		 708 }
709
81ca7350 710 count = ADDRS_PER_PAGE(dn.node_page, inode);
fbfa2cc5
JK		 711
712 count -= dn.ofs_in_node;
9850cf4a 713 f2fs_bug_on(sbi, count < 0);
39936837 714
fbfa2cc5 715 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
4d57b86d 716 f2fs_truncate_data_blocks_range(&dn, count);
fbfa2cc5
JK		 717 free_from += count;
718 }
719
720 f2fs_put_dnode(&dn);
721free_next:
4d57b86d 722 err = f2fs_truncate_inode_blocks(inode, free_from);
764d2c80
JK		 723out:
724 if (lock)
c42d28ce 725 f2fs_unlock_op(sbi);
09210c97 726free_partial:
b3d208f9
JK		 727 /* lastly zero out the first data page */
728 if (!err)
0bfcfcca 729 err = truncate_partial_data_page(inode, from, truncate_page);
fbfa2cc5 730
51dd6249 731 trace_f2fs_truncate_blocks_exit(inode, err);
fbfa2cc5
JK		 732 return err;
733}
734
4c8ff709
CY		 735int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
736{
737 u64 free_from = from;
3265d3db 738 int err;
4c8ff709 739
3265d3db 740#ifdef CONFIG_F2FS_FS_COMPRESSION
4c8ff709
CY		 741 /*
742 * for compressed file, only support cluster size
743 * aligned truncation.
744 */
4fec3fc0
CY		 745 if (f2fs_compressed_file(inode))
746 free_from = round_up(from,
747 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
3265d3db
CY		 748#endif
749
750 err = f2fs_do_truncate_blocks(inode, free_from, lock);
751 if (err)
752 return err;
753
754#ifdef CONFIG_F2FS_FS_COMPRESSION
4a4fc043
FC		 755 /*
756 * For compressed file, after release compress blocks, don't allow write
757 * direct, but we should allow write direct after truncate to zero.
758 */
759 if (f2fs_compressed_file(inode) && !free_from
760 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
761 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
762
17d7648d 763 if (from != free_from) {
3265d3db 764 err = f2fs_truncate_partial_cluster(inode, from, lock);
17d7648d
CY		 765 if (err)
766 return err;
767 }
3265d3db 768#endif
4c8ff709 769
17d7648d 770 return 0;
4c8ff709
CY		 771}
772
9a449e9c 773int f2fs_truncate(struct inode *inode)
fbfa2cc5 774{
b0154891
CY		 775 int err;
776
1f227a3e
JK		 777 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
778 return -EIO;
779
fbfa2cc5
JK		 780 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
781 S_ISLNK(inode->i_mode)))
b0154891 782 return 0;
fbfa2cc5 783
51dd6249
NJ		 784 trace_f2fs_truncate(inode);
785
14b44d23 786 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
c45d6002 787 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
14b44d23
JK		 788 return -EIO;
789 }
7fa750a1 790
10a26878 791 err = f2fs_dquot_initialize(inode);
25fb04db
YC		 792 if (err)
793 return err;
794
92dffd01 795 /* we should check inline_data size */
b9d777b8 796 if (!f2fs_may_inline_data(inode)) {
b0154891
CY		 797 err = f2fs_convert_inline_inode(inode);
798 if (err)
799 return err;
92dffd01
JK		 800 }
801
c42d28ce 802 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
b0154891
CY		 803 if (err)
804 return err;
805
078cd827 806 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 807 f2fs_mark_inode_dirty_sync(inode, false);
b0154891 808 return 0;
fbfa2cc5
JK		 809}
810
549c7297
CB		 811int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
812 struct kstat *stat, u32 request_mask, unsigned int query_flags)
fbfa2cc5 813{
a528d35e 814 struct inode *inode = d_inode(path->dentry);
1c6d8ee4 815 struct f2fs_inode_info *fi = F2FS_I(inode);
d13732cc 816 struct f2fs_inode *ri = NULL;
1c6d8ee4
CY		 817 unsigned int flags;
818
1c1d35df 819 if (f2fs_has_extra_attr(inode) &&
7beb01f7 820 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
1c1d35df
CY		 821 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
822 stat->result_mask |= STATX_BTIME;
823 stat->btime.tv_sec = fi->i_crtime.tv_sec;
824 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
825 }
826
36098557 827 flags = fi->i_flags;
fd26725f
CY		 828 if (flags & F2FS_COMPR_FL)
829 stat->attributes |= STATX_ATTR_COMPRESSED;
59c84408 830 if (flags & F2FS_APPEND_FL)
1c6d8ee4 831 stat->attributes |= STATX_ATTR_APPEND;
62230e0d 832 if (IS_ENCRYPTED(inode))
1c6d8ee4 833 stat->attributes |= STATX_ATTR_ENCRYPTED;
59c84408 834 if (flags & F2FS_IMMUTABLE_FL)
1c6d8ee4 835 stat->attributes |= STATX_ATTR_IMMUTABLE;
59c84408 836 if (flags & F2FS_NODUMP_FL)
1c6d8ee4 837 stat->attributes |= STATX_ATTR_NODUMP;
924e3194
EB		 838 if (IS_VERITY(inode))
839 stat->attributes |= STATX_ATTR_VERITY;
1c6d8ee4 840
fd26725f
CY		 841 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
842 STATX_ATTR_APPEND |
1c6d8ee4
CY		 843 STATX_ATTR_ENCRYPTED |
844 STATX_ATTR_IMMUTABLE |
924e3194
EB		 845 STATX_ATTR_NODUMP |
846 STATX_ATTR_VERITY);
1c6d8ee4 847
984fc4e7 848 generic_fillattr(mnt_userns, inode, stat);
5b4267d1
JK		 849
850 /* we need to show initial sectors used for inline_data/dentries */
851 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
852 f2fs_has_inline_dentry(inode))
853 stat->blocks += (stat->size + 511) >> 9;
854
fbfa2cc5
JK		 855 return 0;
856}
857
858#ifdef CONFIG_F2FS_FS_POSIX_ACL
e65ce2a5
CB		 859static void __setattr_copy(struct user_namespace *mnt_userns,
860 struct inode *inode, const struct iattr *attr)
fbfa2cc5 861{
fbfa2cc5
JK		 862 unsigned int ia_valid = attr->ia_valid;
863
864 if (ia_valid & ATTR_UID)
865 inode->i_uid = attr->ia_uid;
866 if (ia_valid & ATTR_GID)
867 inode->i_gid = attr->ia_gid;
eb31e2f6
AG		 868 if (ia_valid & ATTR_ATIME)
869 inode->i_atime = attr->ia_atime;
870 if (ia_valid & ATTR_MTIME)
871 inode->i_mtime = attr->ia_mtime;
872 if (ia_valid & ATTR_CTIME)
873 inode->i_ctime = attr->ia_ctime;
fbfa2cc5
JK		 874 if (ia_valid & ATTR_MODE) {
875 umode_t mode = attr->ia_mode;
2f221d6f 876 kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
fbfa2cc5 877
7d6beb71 878 if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
fbfa2cc5 879 mode &= ~S_ISGID;
91942321 880 set_acl_inode(inode, mode);
fbfa2cc5
JK		 881 }
882}
883#else
884#define __setattr_copy setattr_copy
885#endif
886
549c7297
CB		 887int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
888 struct iattr *attr)
fbfa2cc5 889{
2b0143b5 890 struct inode *inode = d_inode(dentry);
fbfa2cc5
JK		 891 int err;
892
1f227a3e
JK		 893 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
894 return -EIO;
895
e0fcd015
CY		 896 if (unlikely(IS_IMMUTABLE(inode)))
897 return -EPERM;
898
899 if (unlikely(IS_APPEND(inode) &&
900 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
901 ATTR_GID | ATTR_TIMES_SET))))
902 return -EPERM;
903
4c8ff709
CY		 904 if ((attr->ia_valid & ATTR_SIZE) &&
905 !f2fs_is_compress_backend_ready(inode))
906 return -EOPNOTSUPP;
907
984fc4e7 908 err = setattr_prepare(mnt_userns, dentry, attr);
fbfa2cc5
JK		 909 if (err)
910 return err;
911
20bb2479
EB		 912 err = fscrypt_prepare_setattr(dentry, attr);
913 if (err)
914 return err;
915
95ae251f
EB		 916 err = fsverity_prepare_setattr(dentry, attr);
917 if (err)
918 return err;
919
0abd675e 920 if (is_quota_modification(inode, attr)) {
10a26878 921 err = f2fs_dquot_initialize(inode);
0abd675e
CY		 922 if (err)
923 return err;
924 }
925 if ((attr->ia_valid & ATTR_UID &&
926 !uid_eq(attr->ia_uid, inode->i_uid)) ||
927 (attr->ia_valid & ATTR_GID &&
928 !gid_eq(attr->ia_gid, inode->i_gid))) {
af033b2a 929 f2fs_lock_op(F2FS_I_SB(inode));
0abd675e 930 err = dquot_transfer(inode, attr);
af033b2a
CY		 931 if (err) {
932 set_sbi_flag(F2FS_I_SB(inode),
933 SBI_QUOTA_NEED_REPAIR);
934 f2fs_unlock_op(F2FS_I_SB(inode));
0abd675e 935 return err;
af033b2a
CY		 936 }
937 /*
938 * update uid/gid under lock_op(), so that dquot and inode can
939 * be updated atomically.
940 */
941 if (attr->ia_valid & ATTR_UID)
942 inode->i_uid = attr->ia_uid;
943 if (attr->ia_valid & ATTR_GID)
944 inode->i_gid = attr->ia_gid;
945 f2fs_mark_inode_dirty_sync(inode, true);
946 f2fs_unlock_op(F2FS_I_SB(inode));
0abd675e
CY		 947 }
948
09db6a2e 949 if (attr->ia_valid & ATTR_SIZE) {
cfb9a34d
JK		 950 loff_t old_size = i_size_read(inode);
951
952 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
953 /*
954 * should convert inline inode before i_size_write to
955 * keep smaller than inline_data size with inline flag.
956 */
957 err = f2fs_convert_inline_inode(inode);
958 if (err)
959 return err;
960 }
a33c1502 961
e4544b63 962 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 963 filemap_invalidate_lock(inode->i_mapping);
a33c1502
CY		 964
965 truncate_setsize(inode, attr->ia_size);
966
cfb9a34d 967 if (attr->ia_size <= old_size)
9a449e9c 968 err = f2fs_truncate(inode);
a33c1502
CY		 969 /*
970 * do not trim all blocks after i_size if target size is
971 * larger than i_size.
972 */
edc6d01b 973 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 974 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
a33c1502
CY		 975 if (err)
976 return err;
0cab80ee 977
c10c9820 978 spin_lock(&F2FS_I(inode)->i_size_lock);
cfb9a34d 979 inode->i_mtime = inode->i_ctime = current_time(inode);
a0d00fad 980 F2FS_I(inode)->last_disk_size = i_size_read(inode);
c10c9820 981 spin_unlock(&F2FS_I(inode)->i_size_lock);
fbfa2cc5
JK		 982 }
983
984fc4e7 984 __setattr_copy(mnt_userns, inode, attr);
fbfa2cc5
JK		 985
986 if (attr->ia_valid & ATTR_MODE) {
984fc4e7 987 err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
17232e83
CY		 988
989 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
990 if (!err)
991 inode->i_mode = F2FS_I(inode)->i_acl_mode;
91942321 992 clear_inode_flag(inode, FI_ACL_MODE);
fbfa2cc5
JK		 993 }
994 }
995
c0ed4405 996 /* file size may changed here */
ca597bdd 997 f2fs_mark_inode_dirty_sync(inode, true);
15d04354
JK		 998
999 /* inode change will produce dirty node pages flushed by checkpoint */
1000 f2fs_balance_fs(F2FS_I_SB(inode), true);
1001
fbfa2cc5
JK		 1002 return err;
1003}
1004
1005const struct inode_operations f2fs_file_inode_operations = {
1006 .getattr = f2fs_getattr,
1007 .setattr = f2fs_setattr,
1008 .get_acl = f2fs_get_acl,
a6dda0e6 1009 .set_acl = f2fs_set_acl,
fbfa2cc5 1010 .listxattr = f2fs_listxattr,
9ab70134 1011 .fiemap = f2fs_fiemap,
9b1bb01c
MS		 1012 .fileattr_get = f2fs_fileattr_get,
1013 .fileattr_set = f2fs_fileattr_set,
fbfa2cc5
JK		 1014};
1015
6394328a 1016static int fill_zero(struct inode *inode, pgoff_t index,
fbfa2cc5
JK		 1017 loff_t start, loff_t len)
1018{
4081363f 1019 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 1020 struct page *page;
1021
1022 if (!len)
6394328a 1023 return 0;
fbfa2cc5 1024
2c4db1a6 1025 f2fs_balance_fs(sbi, true);
bd43df02 1026
e479556b 1027 f2fs_lock_op(sbi);
4d57b86d 1028 page = f2fs_get_new_data_page(inode, NULL, index, false);
e479556b 1029 f2fs_unlock_op(sbi);
fbfa2cc5 1030
6394328a
CY		 1031 if (IS_ERR(page))
1032 return PTR_ERR(page);
1033
bae0ee7a 1034 f2fs_wait_on_page_writeback(page, DATA, true, true);
6394328a
CY		 1035 zero_user(page, start, len);
1036 set_page_dirty(page);
1037 f2fs_put_page(page, 1);
1038 return 0;
fbfa2cc5
JK		 1039}
1040
4d57b86d 1041int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
fbfa2cc5 1042{
fbfa2cc5
JK		 1043 int err;
1044
ea58711e 1045 while (pg_start < pg_end) {
fbfa2cc5 1046 struct dnode_of_data dn;
ea58711e 1047 pgoff_t end_offset, count;
9eaeba70 1048
fbfa2cc5 1049 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1050 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
fbfa2cc5 1051 if (err) {
ea58711e 1052 if (err == -ENOENT) {
4d57b86d
CY		 1053 pg_start = f2fs_get_next_page_offset(&dn,
1054 pg_start);
fbfa2cc5 1055 continue;
ea58711e 1056 }
fbfa2cc5
JK		 1057 return err;
1058 }
1059
81ca7350 1060 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
ea58711e
CY		 1061 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1062
1063 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1064
4d57b86d 1065 f2fs_truncate_data_blocks_range(&dn, count);
fbfa2cc5 1066 f2fs_put_dnode(&dn);
ea58711e
CY		 1067
1068 pg_start += count;
fbfa2cc5
JK		 1069 }
1070 return 0;
1071}
1072
a66c7b2f 1073static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
fbfa2cc5
JK		 1074{
1075 pgoff_t pg_start, pg_end;
1076 loff_t off_start, off_end;
b9d777b8 1077 int ret;
fbfa2cc5 1078
b9d777b8
JK		 1079 ret = f2fs_convert_inline_inode(inode);
1080 if (ret)
1081 return ret;
9ffe0fb5 1082
09cbfeaf
KS		 1083 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1084 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
fbfa2cc5 1085
09cbfeaf
KS		 1086 off_start = offset & (PAGE_SIZE - 1);
1087 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5
JK		 1088
1089 if (pg_start == pg_end) {
6394328a 1090 ret = fill_zero(inode, pg_start, off_start,
fbfa2cc5 1091 off_end - off_start);
6394328a
CY		 1092 if (ret)
1093 return ret;
fbfa2cc5 1094 } else {
6394328a
CY		 1095 if (off_start) {
1096 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 1097 PAGE_SIZE - off_start);
6394328a
CY		 1098 if (ret)
1099 return ret;
1100 }
1101 if (off_end) {
1102 ret = fill_zero(inode, pg_end, 0, off_end);
1103 if (ret)
1104 return ret;
1105 }
fbfa2cc5
JK		 1106
1107 if (pg_start < pg_end) {
fbfa2cc5 1108 loff_t blk_start, blk_end;
4081363f 1109 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127a3d4 1110
2c4db1a6 1111 f2fs_balance_fs(sbi, true);
fbfa2cc5 1112
09cbfeaf
KS		 1113 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1114 blk_end = (loff_t)pg_end << PAGE_SHIFT;
a33c1502 1115
e4544b63 1116 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6abaa83c 1117 filemap_invalidate_lock(inode->i_mapping);
a33c1502 1118
c8dc3047 1119 truncate_pagecache_range(inode, blk_start, blk_end - 1);
39936837 1120
e479556b 1121 f2fs_lock_op(sbi);
4d57b86d 1122 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
e479556b 1123 f2fs_unlock_op(sbi);
a33c1502 1124
6abaa83c 1125 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 1126 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
fbfa2cc5
JK		 1127 }
1128 }
1129
fbfa2cc5
JK		 1130 return ret;
1131}
1132
0a2aa8fb
JK		 1133static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1134 int *do_replace, pgoff_t off, pgoff_t len)
b4ace337
CY		 1135{
1136 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1137 struct dnode_of_data dn;
0a2aa8fb 1138 int ret, done, i;
ecbaa406 1139
0a2aa8fb 1140next_dnode:
6e2c64ad 1141 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1142 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
6e2c64ad
JK		 1143 if (ret && ret != -ENOENT) {
1144 return ret;
1145 } else if (ret == -ENOENT) {
0a2aa8fb
JK		 1146 if (dn.max_level == 0)
1147 return -ENOENT;
4c8ff709
CY		 1148 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1149 dn.ofs_in_node, len);
0a2aa8fb
JK		 1150 blkaddr += done;
1151 do_replace += done;
1152 goto next;
1153 }
1154
1155 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1156 dn.ofs_in_node, len);
1157 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
a2ced1ce 1158 *blkaddr = f2fs_data_blkaddr(&dn);
93770ab7
CY		 1159
1160 if (__is_valid_data_blkaddr(*blkaddr) &&
1161 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1162 DATA_GENERIC_ENHANCE)) {
1163 f2fs_put_dnode(&dn);
10f966bb 1164 return -EFSCORRUPTED;
93770ab7
CY		 1165 }
1166
4d57b86d 1167 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
0a2aa8fb 1168
b0332a0f 1169 if (f2fs_lfs_mode(sbi)) {
0a2aa8fb 1170 f2fs_put_dnode(&dn);
fd114ab2 1171 return -EOPNOTSUPP;
0a2aa8fb
JK		 1172 }
1173
6e2c64ad 1174 /* do not invalidate this block address */
f28b3434 1175 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
0a2aa8fb 1176 *do_replace = 1;
b4ace337 1177 }
6e2c64ad 1178 }
0a2aa8fb
JK		 1179 f2fs_put_dnode(&dn);
1180next:
1181 len -= done;
1182 off += done;
1183 if (len)
1184 goto next_dnode;
1185 return 0;
1186}
b4ace337 1187
0a2aa8fb
JK		 1188static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1189 int *do_replace, pgoff_t off, int len)
1190{
1191 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1192 struct dnode_of_data dn;
1193 int ret, i;
b4ace337 1194
0a2aa8fb
JK		 1195 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1196 if (*do_replace == 0)
1197 continue;
b4ace337 1198
0a2aa8fb 1199 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1200 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
0a2aa8fb
JK		 1201 if (ret) {
1202 dec_valid_block_count(sbi, inode, 1);
4d57b86d 1203 f2fs_invalidate_blocks(sbi, *blkaddr);
0a2aa8fb
JK		 1204 } else {
1205 f2fs_update_data_blkaddr(&dn, *blkaddr);
36abef4e 1206 }
0a2aa8fb
JK		 1207 f2fs_put_dnode(&dn);
1208 }
1209 return 0;
1210}
1211
1212static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1213 block_t *blkaddr, int *do_replace,
1214 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1215{
1216 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1217 pgoff_t i = 0;
1218 int ret;
36abef4e 1219
0a2aa8fb
JK		 1220 while (i < len) {
1221 if (blkaddr[i] == NULL_ADDR && !full) {
1222 i++;
1223 continue;
6e2c64ad 1224 }
b4ace337 1225
0a2aa8fb
JK		 1226 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1227 struct dnode_of_data dn;
1228 struct node_info ni;
1229 size_t new_size;
1230 pgoff_t ilen;
b4ace337 1231
0a2aa8fb 1232 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
4d57b86d 1233 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
0a2aa8fb
JK		 1234 if (ret)
1235 return ret;
b4ace337 1236
a9419b63 1237 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
7735730d
CY		 1238 if (ret) {
1239 f2fs_put_dnode(&dn);
1240 return ret;
1241 }
1242
0a2aa8fb
JK		 1243 ilen = min((pgoff_t)
1244 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1245 dn.ofs_in_node, len - i);
1246 do {
a2ced1ce 1247 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
4d57b86d 1248 f2fs_truncate_data_blocks_range(&dn, 1);
0a2aa8fb
JK		 1249
1250 if (do_replace[i]) {
1251 f2fs_i_blocks_write(src_inode,
0abd675e 1252 1, false, false);
0a2aa8fb 1253 f2fs_i_blocks_write(dst_inode,
0abd675e 1254 1, true, false);
0a2aa8fb
JK		 1255 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1256 blkaddr[i], ni.version, true, false);
1257
1258 do_replace[i] = 0;
1259 }
1260 dn.ofs_in_node++;
1261 i++;
1f0d5c91 1262 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
0a2aa8fb
JK		 1263 if (dst_inode->i_size < new_size)
1264 f2fs_i_size_write(dst_inode, new_size);
e87f7329 1265 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
6e2c64ad 1266
0a2aa8fb
JK		 1267 f2fs_put_dnode(&dn);
1268 } else {
1269 struct page *psrc, *pdst;
1270
4d57b86d
CY		 1271 psrc = f2fs_get_lock_data_page(src_inode,
1272 src + i, true);
0a2aa8fb
JK		 1273 if (IS_ERR(psrc))
1274 return PTR_ERR(psrc);
4d57b86d 1275 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
0a2aa8fb
JK		 1276 true);
1277 if (IS_ERR(pdst)) {
1278 f2fs_put_page(psrc, 1);
1279 return PTR_ERR(pdst);
1280 }
1281 f2fs_copy_page(psrc, pdst);
1282 set_page_dirty(pdst);
1283 f2fs_put_page(pdst, 1);
6e2c64ad 1284 f2fs_put_page(psrc, 1);
b4ace337 1285
4d57b86d
CY		 1286 ret = f2fs_truncate_hole(src_inode,
1287 src + i, src + i + 1);
0a2aa8fb
JK		 1288 if (ret)
1289 return ret;
1290 i++;
1291 }
6e2c64ad
JK		 1292 }
1293 return 0;
0a2aa8fb 1294}
b4ace337 1295
0a2aa8fb
JK		 1296static int __exchange_data_block(struct inode *src_inode,
1297 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
363cad7f 1298 pgoff_t len, bool full)
0a2aa8fb
JK		 1299{
1300 block_t *src_blkaddr;
1301 int *do_replace;
363cad7f 1302 pgoff_t olen;
0a2aa8fb
JK		 1303 int ret;
1304
363cad7f 1305 while (len) {
d02a6e61 1306 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
0a2aa8fb 1307
628b3d14 1308 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
9d2a789c 1309 array_size(olen, sizeof(block_t)),
4f4460c0 1310 GFP_NOFS);
363cad7f
JK		 1311 if (!src_blkaddr)
1312 return -ENOMEM;
0a2aa8fb 1313
628b3d14 1314 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
9d2a789c 1315 array_size(olen, sizeof(int)),
4f4460c0 1316 GFP_NOFS);
363cad7f
JK		 1317 if (!do_replace) {
1318 kvfree(src_blkaddr);
1319 return -ENOMEM;
1320 }
0a2aa8fb 1321
363cad7f
JK		 1322 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1323 do_replace, src, olen);
1324 if (ret)
1325 goto roll_back;
0a2aa8fb 1326
363cad7f
JK		 1327 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1328 do_replace, src, dst, olen, full);
1329 if (ret)
1330 goto roll_back;
1331
1332 src += olen;
1333 dst += olen;
1334 len -= olen;
1335
1336 kvfree(src_blkaddr);
1337 kvfree(do_replace);
1338 }
0a2aa8fb
JK		 1339 return 0;
1340
1341roll_back:
9fd62605 1342 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
0a2aa8fb
JK		 1343 kvfree(src_blkaddr);
1344 kvfree(do_replace);
6e2c64ad
JK		 1345 return ret;
1346}
b4ace337 1347
6f8d4455 1348static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
6e2c64ad
JK		 1349{
1350 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f91108b8 1351 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
6f8d4455
JK		 1352 pgoff_t start = offset >> PAGE_SHIFT;
1353 pgoff_t end = (offset + len) >> PAGE_SHIFT;
0a2aa8fb 1354 int ret;
6e2c64ad 1355
0a2aa8fb 1356 f2fs_balance_fs(sbi, true);
5f281fab 1357
6f8d4455 1358 /* avoid gc operation during block exchange */
e4544b63 1359 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 1360 filemap_invalidate_lock(inode->i_mapping);
5f281fab 1361
6f8d4455
JK		 1362 f2fs_lock_op(sbi);
1363 f2fs_drop_extent_tree(inode);
1364 truncate_pagecache(inode, offset);
0a2aa8fb
JK		 1365 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1366 f2fs_unlock_op(sbi);
6f8d4455 1367
edc6d01b 1368 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 1369 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
b4ace337
CY		 1370 return ret;
1371}
1372
1373static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1374{
b4ace337
CY		 1375 loff_t new_size;
1376 int ret;
1377
b4ace337
CY		 1378 if (offset + len >= i_size_read(inode))
1379 return -EINVAL;
1380
1381 /* collapse range should be aligned to block size of f2fs. */
1382 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1383 return -EINVAL;
1384
b9d777b8
JK		 1385 ret = f2fs_convert_inline_inode(inode);
1386 if (ret)
1387 return ret;
97a7b2c2 1388
b4ace337
CY		 1389 /* write out all dirty pages from offset */
1390 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1391 if (ret)
6f8d4455 1392 return ret;
b4ace337 1393
6f8d4455 1394 ret = f2fs_do_collapse(inode, offset, len);
b4ace337 1395 if (ret)
6f8d4455 1396 return ret;
b4ace337 1397
6e2c64ad 1398 /* write out all moved pages, if possible */
edc6d01b 1399 filemap_invalidate_lock(inode->i_mapping);
6e2c64ad
JK		 1400 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1401 truncate_pagecache(inode, offset);
1402
b4ace337 1403 new_size = i_size_read(inode) - len;
c42d28ce 1404 ret = f2fs_truncate_blocks(inode, new_size, true);
edc6d01b 1405 filemap_invalidate_unlock(inode->i_mapping);
b4ace337 1406 if (!ret)
fc9581c8 1407 f2fs_i_size_write(inode, new_size);
b4ace337
CY		 1408 return ret;
1409}
1410
6e961949
CY		 1411static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1412 pgoff_t end)
1413{
1414 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1415 pgoff_t index = start;
1416 unsigned int ofs_in_node = dn->ofs_in_node;
1417 blkcnt_t count = 0;
1418 int ret;
1419
1420 for (; index < end; index++, dn->ofs_in_node++) {
a2ced1ce 1421 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
6e961949
CY		 1422 count++;
1423 }
1424
1425 dn->ofs_in_node = ofs_in_node;
4d57b86d 1426 ret = f2fs_reserve_new_blocks(dn, count);
6e961949
CY		 1427 if (ret)
1428 return ret;
1429
1430 dn->ofs_in_node = ofs_in_node;
1431 for (index = start; index < end; index++, dn->ofs_in_node++) {
a2ced1ce 1432 dn->data_blkaddr = f2fs_data_blkaddr(dn);
6e961949 1433 /*
4d57b86d 1434 * f2fs_reserve_new_blocks will not guarantee entire block
6e961949
CY		 1435 * allocation.
1436 */
1437 if (dn->data_blkaddr == NULL_ADDR) {
1438 ret = -ENOSPC;
1439 break;
1440 }
25f82362
CY		 1441
1442 if (dn->data_blkaddr == NEW_ADDR)
1443 continue;
1444
1445 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1446 DATA_GENERIC_ENHANCE)) {
1447 ret = -EFSCORRUPTED;
1448 break;
6e961949 1449 }
25f82362
CY		 1450
1451 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1452 dn->data_blkaddr = NEW_ADDR;
1453 f2fs_set_data_blkaddr(dn);
6e961949
CY		 1454 }
1455
1456 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1457
1458 return ret;
1459}
1460
75cd4e09
CY		 1461static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1462 int mode)
1463{
1464 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1465 struct address_space *mapping = inode->i_mapping;
1466 pgoff_t index, pg_start, pg_end;
1467 loff_t new_size = i_size_read(inode);
1468 loff_t off_start, off_end;
1469 int ret = 0;
1470
75cd4e09
CY		 1471 ret = inode_newsize_ok(inode, (len + offset));
1472 if (ret)
1473 return ret;
1474
b9d777b8
JK		 1475 ret = f2fs_convert_inline_inode(inode);
1476 if (ret)
1477 return ret;
75cd4e09
CY		 1478
1479 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1480 if (ret)
6f8d4455 1481 return ret;
75cd4e09 1482
09cbfeaf
KS		 1483 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1484 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
75cd4e09 1485
09cbfeaf
KS		 1486 off_start = offset & (PAGE_SIZE - 1);
1487 off_end = (offset + len) & (PAGE_SIZE - 1);
75cd4e09
CY		 1488
1489 if (pg_start == pg_end) {
6394328a
CY		 1490 ret = fill_zero(inode, pg_start, off_start,
1491 off_end - off_start);
1492 if (ret)
6f8d4455 1493 return ret;
6394328a 1494
75cd4e09
CY		 1495 new_size = max_t(loff_t, new_size, offset + len);
1496 } else {
1497 if (off_start) {
6394328a 1498 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 1499 PAGE_SIZE - off_start);
6394328a 1500 if (ret)
6f8d4455 1501 return ret;
6394328a 1502
75cd4e09 1503 new_size = max_t(loff_t, new_size,
09cbfeaf 1504 (loff_t)pg_start << PAGE_SHIFT);
75cd4e09
CY		 1505 }
1506
6e961949 1507 for (index = pg_start; index < pg_end;) {
75cd4e09 1508 struct dnode_of_data dn;
6e961949
CY		 1509 unsigned int end_offset;
1510 pgoff_t end;
75cd4e09 1511
e4544b63 1512 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 1513 filemap_invalidate_lock(mapping);
c7079853
CY		 1514
1515 truncate_pagecache_range(inode,
1516 (loff_t)index << PAGE_SHIFT,
1517 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1518
75cd4e09
CY		 1519 f2fs_lock_op(sbi);
1520
6e961949 1521 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1522 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
75cd4e09
CY		 1523 if (ret) {
1524 f2fs_unlock_op(sbi);
edc6d01b 1525 filemap_invalidate_unlock(mapping);
e4544b63 1526 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
75cd4e09
CY		 1527 goto out;
1528 }
1529
6e961949
CY		 1530 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1531 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1532
1533 ret = f2fs_do_zero_range(&dn, index, end);
75cd4e09 1534 f2fs_put_dnode(&dn);
c7079853 1535
75cd4e09 1536 f2fs_unlock_op(sbi);
edc6d01b 1537 filemap_invalidate_unlock(mapping);
e4544b63 1538 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
9434fcde
CY		 1539
1540 f2fs_balance_fs(sbi, dn.node_changed);
1541
6e961949
CY		 1542 if (ret)
1543 goto out;
75cd4e09 1544
6e961949 1545 index = end;
75cd4e09 1546 new_size = max_t(loff_t, new_size,
6e961949 1547 (loff_t)index << PAGE_SHIFT);
75cd4e09
CY		 1548 }
1549
1550 if (off_end) {
6394328a
CY		 1551 ret = fill_zero(inode, pg_end, 0, off_end);
1552 if (ret)
1553 goto out;
1554
75cd4e09
CY		 1555 new_size = max_t(loff_t, new_size, offset + len);
1556 }
1557 }
1558
1559out:
17cd07ae
CY		 1560 if (new_size > i_size_read(inode)) {
1561 if (mode & FALLOC_FL_KEEP_SIZE)
1562 file_set_keep_isize(inode);
1563 else
1564 f2fs_i_size_write(inode, new_size);
1565 }
75cd4e09
CY		 1566 return ret;
1567}
1568
f62185d0
CY		 1569static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1570{
1571 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
edc6d01b 1572 struct address_space *mapping = inode->i_mapping;
0a2aa8fb 1573 pgoff_t nr, pg_start, pg_end, delta, idx;
f62185d0 1574 loff_t new_size;
6e2c64ad 1575 int ret = 0;
f62185d0 1576
f62185d0 1577 new_size = i_size_read(inode) + len;
46e82fb1
KM		 1578 ret = inode_newsize_ok(inode, new_size);
1579 if (ret)
1580 return ret;
f62185d0
CY		 1581
1582 if (offset >= i_size_read(inode))
1583 return -EINVAL;
1584
1585 /* insert range should be aligned to block size of f2fs. */
1586 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1587 return -EINVAL;
1588
b9d777b8
JK		 1589 ret = f2fs_convert_inline_inode(inode);
1590 if (ret)
1591 return ret;
97a7b2c2 1592
2c4db1a6 1593 f2fs_balance_fs(sbi, true);
2a340760 1594
edc6d01b 1595 filemap_invalidate_lock(mapping);
c42d28ce 1596 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
edc6d01b 1597 filemap_invalidate_unlock(mapping);
f62185d0 1598 if (ret)
6f8d4455 1599 return ret;
f62185d0
CY		 1600
1601 /* write out all dirty pages from offset */
edc6d01b 1602 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
f62185d0 1603 if (ret)
6f8d4455 1604 return ret;
f62185d0 1605
09cbfeaf
KS		 1606 pg_start = offset >> PAGE_SHIFT;
1607 pg_end = (offset + len) >> PAGE_SHIFT;
f62185d0 1608 delta = pg_end - pg_start;
f91108b8 1609 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
0a2aa8fb 1610
6f8d4455 1611 /* avoid gc operation during block exchange */
e4544b63 1612 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 1613 filemap_invalidate_lock(mapping);
6f8d4455
JK		 1614 truncate_pagecache(inode, offset);
1615
0a2aa8fb
JK		 1616 while (!ret && idx > pg_start) {
1617 nr = idx - pg_start;
1618 if (nr > delta)
1619 nr = delta;
1620 idx -= nr;
f62185d0 1621
f62185d0 1622 f2fs_lock_op(sbi);
5f281fab
JK		 1623 f2fs_drop_extent_tree(inode);
1624
0a2aa8fb
JK		 1625 ret = __exchange_data_block(inode, inode, idx,
1626 idx + delta, nr, false);
f62185d0
CY		 1627 f2fs_unlock_op(sbi);
1628 }
edc6d01b 1629 filemap_invalidate_unlock(mapping);
e4544b63 1630 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
f62185d0 1631
6e2c64ad 1632 /* write out all moved pages, if possible */
edc6d01b
JK		 1633 filemap_invalidate_lock(mapping);
1634 filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
6e2c64ad 1635 truncate_pagecache(inode, offset);
edc6d01b 1636 filemap_invalidate_unlock(mapping);
6e2c64ad
JK		 1637
1638 if (!ret)
fc9581c8 1639 f2fs_i_size_write(inode, new_size);
f62185d0
CY		 1640 return ret;
1641}
1642
fbfa2cc5
JK		 1643static int expand_inode_data(struct inode *inode, loff_t offset,
1644 loff_t len, int mode)
1645{
4081363f 1646 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d5097be5 1647 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
f9d6d059
CY		 1648 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1649 .m_may_create = true };
88f2cfc5 1650 pgoff_t pg_start, pg_end;
fbfa2cc5 1651 loff_t new_size = i_size_read(inode);
e12dd7bd 1652 loff_t off_end;
88f2cfc5 1653 block_t expanded = 0;
a7de6086 1654 int err;
fbfa2cc5 1655
a7de6086
JK		 1656 err = inode_newsize_ok(inode, (len + offset));
1657 if (err)
1658 return err;
fbfa2cc5 1659
a7de6086
JK		 1660 err = f2fs_convert_inline_inode(inode);
1661 if (err)
1662 return err;
9e09fc85 1663
2c4db1a6 1664 f2fs_balance_fs(sbi, true);
2a340760 1665
88f2cfc5 1666 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
e12dd7bd 1667 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
09cbfeaf 1668 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5 1669
88f2cfc5
CY		 1670 map.m_lblk = pg_start;
1671 map.m_len = pg_end - pg_start;
e12dd7bd
JK		 1672 if (off_end)
1673 map.m_len++;
ead43275 1674
f5a53edc
JK		 1675 if (!map.m_len)
1676 return 0;
1677
1678 if (f2fs_is_pinned_file(inode)) {
e1175f02
CY		 1679 block_t sec_blks = BLKS_PER_SEC(sbi);
1680 block_t sec_len = roundup(map.m_len, sec_blks);
f5a53edc 1681
e1175f02 1682 map.m_len = sec_blks;
f5a53edc
JK		 1683next_alloc:
1684 if (has_not_enough_free_secs(sbi, 0,
1685 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
e4544b63 1686 f2fs_down_write(&sbi->gc_lock);
7dede886 1687 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2e42b7f8 1688 if (err && err != -ENODATA)
f5a53edc
JK		 1689 goto out_err;
1690 }
1691
e4544b63 1692 f2fs_down_write(&sbi->pin_sem);
fd612648
DJ		 1693
1694 f2fs_lock_op(sbi);
509f1010 1695 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
fd612648
DJ		 1696 f2fs_unlock_op(sbi);
1697
d0b9e42a 1698 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
f5a53edc 1699 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
d4dd19ec 1700 file_dont_truncate(inode);
d0b9e42a 1701
e4544b63 1702 f2fs_up_write(&sbi->pin_sem);
cad3836f 1703
88f2cfc5 1704 expanded += map.m_len;
e1175f02 1705 sec_len -= map.m_len;
f5a53edc 1706 map.m_lblk += map.m_len;
e1175f02 1707 if (!err && sec_len)
f5a53edc
JK		 1708 goto next_alloc;
1709
88f2cfc5 1710 map.m_len = expanded;
f5a53edc
JK		 1711 } else {
1712 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
88f2cfc5 1713 expanded = map.m_len;
f5a53edc
JK		 1714 }
1715out_err:
a7de6086 1716 if (err) {
e12dd7bd 1717 pgoff_t last_off;
fbfa2cc5 1718
88f2cfc5 1719 if (!expanded)
a7de6086 1720 return err;
98397ff3 1721
88f2cfc5 1722 last_off = pg_start + expanded - 1;
e12dd7bd
JK		 1723
1724 /* update new size to the failed position */
1061fd48 1725 new_size = (last_off == pg_end) ? offset + len :
e12dd7bd
JK		 1726 (loff_t)(last_off + 1) << PAGE_SHIFT;
1727 } else {
1728 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
fbfa2cc5
JK		 1729 }
1730
e8ed90a6
CY		 1731 if (new_size > i_size_read(inode)) {
1732 if (mode & FALLOC_FL_KEEP_SIZE)
1733 file_set_keep_isize(inode);
1734 else
1735 f2fs_i_size_write(inode, new_size);
1736 }
fbfa2cc5 1737
a7de6086 1738 return err;
fbfa2cc5
JK		 1739}
1740
1741static long f2fs_fallocate(struct file *file, int mode,
1742 loff_t offset, loff_t len)
1743{
6131ffaa 1744 struct inode *inode = file_inode(file);
587c0a42 1745 long ret = 0;
fbfa2cc5 1746
1f227a3e
JK		 1747 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1748 return -EIO;
00e09c0b
CY		 1749 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1750 return -ENOSPC;
4c8ff709
CY		 1751 if (!f2fs_is_compress_backend_ready(inode))
1752 return -EOPNOTSUPP;
1f227a3e 1753
c998012b
CY		 1754 /* f2fs only support ->fallocate for regular file */
1755 if (!S_ISREG(inode->i_mode))
1756 return -EINVAL;
1757
62230e0d 1758 if (IS_ENCRYPTED(inode) &&
f62185d0 1759 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
fcc85a4d
JK		 1760 return -EOPNOTSUPP;
1761
5fed0be8
JK		 1762 /*
1763 * Pinned file should not support partial trucation since the block
1764 * can be used by applications.
1765 */
1766 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
4c8ff709
CY		 1767 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1768 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1769 return -EOPNOTSUPP;
1770
b4ace337 1771 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
f62185d0
CY		 1772 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1773 FALLOC_FL_INSERT_RANGE))
fbfa2cc5
JK		 1774 return -EOPNOTSUPP;
1775
5955102c 1776 inode_lock(inode);
3375f696 1777
587c0a42
TY		 1778 if (mode & FALLOC_FL_PUNCH_HOLE) {
1779 if (offset >= inode->i_size)
1780 goto out;
1781
a66c7b2f 1782 ret = punch_hole(inode, offset, len);
b4ace337
CY		 1783 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1784 ret = f2fs_collapse_range(inode, offset, len);
75cd4e09
CY		 1785 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1786 ret = f2fs_zero_range(inode, offset, len, mode);
f62185d0
CY		 1787 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1788 ret = f2fs_insert_range(inode, offset, len);
b4ace337 1789 } else {
fbfa2cc5 1790 ret = expand_inode_data(inode, offset, len, mode);
b4ace337 1791 }
fbfa2cc5 1792
3af60a49 1793 if (!ret) {
078cd827 1794 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 1795 f2fs_mark_inode_dirty_sync(inode, false);
d0239e1b 1796 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3af60a49 1797 }
3375f696 1798
587c0a42 1799out:
5955102c 1800 inode_unlock(inode);
3375f696 1801
c01e2853 1802 trace_f2fs_fallocate(inode, mode, offset, len, ret);
fbfa2cc5
JK		 1803 return ret;
1804}
1805
1e84371f
JK		 1806static int f2fs_release_file(struct inode *inode, struct file *filp)
1807{
de5307e4
JK		 1808 /*
1809 * f2fs_relase_file is called at every close calls. So we should
1810 * not drop any inmemory pages by close called by other process.
1811 */
1812 if (!(filp->f_mode & FMODE_WRITE) ||
1813 atomic_read(&inode->i_writecount) != 1)
1814 return 0;
1815
1e84371f 1816 if (f2fs_is_atomic_file(inode))
3db1de0e 1817 f2fs_abort_atomic_write(inode, true);
1e84371f
JK		 1818 return 0;
1819}
1820
7a10f017 1821static int f2fs_file_flush(struct file *file, fl_owner_t id)
fbfa2cc5 1822{
7a10f017
JK		 1823 struct inode *inode = file_inode(file);
1824
1825 /*
1826 * If the process doing a transaction is crashed, we should do
1827 * roll-back. Otherwise, other reader/write can see corrupted database
1828 * until all the writers close its file. Since this should be done
1829 * before dropping file lock, it needs to do in ->flush.
1830 */
1831 if (f2fs_is_atomic_file(inode) &&
3db1de0e
DJ		 1832 F2FS_I(inode)->atomic_write_task == current)
1833 f2fs_abort_atomic_write(inode, true);
7a10f017 1834 return 0;
fbfa2cc5
JK		 1835}
1836
36098557 1837static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
2c1d0305
CY		 1838{
1839 struct f2fs_inode_info *fi = F2FS_I(inode);
99eabb91
JK		 1840 u32 masked_flags = fi->i_flags & mask;
1841
a7531039
JK		 1842 /* mask can be shrunk by flags_valid selector */
1843 iflags &= mask;
2c1d0305
CY		 1844
1845 /* Is it quota file? Do not allow user to mess with it */
1846 if (IS_NOQUOTA(inode))
1847 return -EPERM;
1848
99eabb91 1849 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
2c2eb7a3
DR		 1850 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1851 return -EOPNOTSUPP;
1852 if (!f2fs_empty_dir(inode))
1853 return -ENOTEMPTY;
1854 }
1855
4c8ff709
CY		 1856 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1857 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1858 return -EOPNOTSUPP;
1859 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1860 return -EINVAL;
1861 }
1862
99eabb91 1863 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
aa576970 1864 if (masked_flags & F2FS_COMPR_FL) {
78134d03 1865 if (!f2fs_disable_compressed_file(inode))
2536ac68
CY		 1866 return -EINVAL;
1867 }
4c8ff709
CY		 1868 if (iflags & F2FS_NOCOMP_FL)
1869 return -EINVAL;
1870 if (iflags & F2FS_COMPR_FL) {
4c8ff709
CY		 1871 if (!f2fs_may_compress(inode))
1872 return -EINVAL;
519a5a2f
CY		 1873 if (S_ISREG(inode->i_mode) && inode->i_size)
1874 return -EINVAL;
4c8ff709
CY		 1875
1876 set_compress_context(inode);
1877 }
1878 }
99eabb91
JK		 1879 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1880 if (masked_flags & F2FS_COMPR_FL)
4c8ff709
CY		 1881 return -EINVAL;
1882 }
1883
d5e5efa2 1884 fi->i_flags = iflags | (fi->i_flags & ~mask);
4c8ff709
CY		 1885 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1886 (fi->i_flags & F2FS_NOCOMP_FL));
2c1d0305 1887
59c84408 1888 if (fi->i_flags & F2FS_PROJINHERIT_FL)
2c1d0305
CY		 1889 set_inode_flag(inode, FI_PROJ_INHERIT);
1890 else
1891 clear_inode_flag(inode, FI_PROJ_INHERIT);
1892
1893 inode->i_ctime = current_time(inode);
1894 f2fs_set_inode_flags(inode);
b32e0190 1895 f2fs_mark_inode_dirty_sync(inode, true);
2c1d0305
CY		 1896 return 0;
1897}
1898
9b1bb01c 1899/* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
36098557
EB		 1900
1901/*
1902 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1903 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1904 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1905 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
9b1bb01c
MS		 1906 *
1907 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1908 * FS_IOC_FSSETXATTR is done by the VFS.
36098557
EB		 1909 */
1910
1911static const struct {
1912 u32 iflag;
1913 u32 fsflag;
1914} f2fs_fsflags_map[] = {
4c8ff709 1915 { F2FS_COMPR_FL, FS_COMPR_FL },
36098557
EB		 1916 { F2FS_SYNC_FL, FS_SYNC_FL },
1917 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1918 { F2FS_APPEND_FL, FS_APPEND_FL },
1919 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1920 { F2FS_NOATIME_FL, FS_NOATIME_FL },
4c8ff709 1921 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
36098557
EB		 1922 { F2FS_INDEX_FL, FS_INDEX_FL },
1923 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1924 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
2c2eb7a3 1925 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
36098557
EB		 1926};
1927
1928#define F2FS_GETTABLE_FS_FL ( \
4c8ff709 1929 FS_COMPR_FL | \
36098557
EB		 1930 FS_SYNC_FL | \
1931 FS_IMMUTABLE_FL | \
1932 FS_APPEND_FL | \
1933 FS_NODUMP_FL | \
1934 FS_NOATIME_FL | \
4c8ff709 1935 FS_NOCOMP_FL | \
36098557
EB		 1936 FS_INDEX_FL | \
1937 FS_DIRSYNC_FL | \
1938 FS_PROJINHERIT_FL | \
1939 FS_ENCRYPT_FL | \
1940 FS_INLINE_DATA_FL | \
95ae251f 1941 FS_NOCOW_FL | \
fbc246a1 1942 FS_VERITY_FL | \
2c2eb7a3 1943 FS_CASEFOLD_FL)
36098557
EB		 1944
1945#define F2FS_SETTABLE_FS_FL ( \
4c8ff709 1946 FS_COMPR_FL | \
36098557
EB		 1947 FS_SYNC_FL | \
1948 FS_IMMUTABLE_FL | \
1949 FS_APPEND_FL | \
1950 FS_NODUMP_FL | \
1951 FS_NOATIME_FL | \
4c8ff709 1952 FS_NOCOMP_FL | \
36098557 1953 FS_DIRSYNC_FL | \
2c2eb7a3
DR		 1954 FS_PROJINHERIT_FL | \
1955 FS_CASEFOLD_FL)
36098557
EB		 1956
1957/* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1958static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1959{
1960 u32 fsflags = 0;
1961 int i;
1962
1963 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1964 if (iflags & f2fs_fsflags_map[i].iflag)
1965 fsflags |= f2fs_fsflags_map[i].fsflag;
1966
1967 return fsflags;
1968}
1969
1970/* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1971static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1972{
1973 u32 iflags = 0;
1974 int i;
1975
1976 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1977 if (fsflags & f2fs_fsflags_map[i].fsflag)
1978 iflags |= f2fs_fsflags_map[i].iflag;
1979
1980 return iflags;
1981}
1982
d49f3e89
CY		 1983static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1984{
1985 struct inode *inode = file_inode(filp);
1986
1987 return put_user(inode->i_generation, (int __user *)arg);
1988}
1989
88b88a66
JK		 1990static int f2fs_ioc_start_atomic_write(struct file *filp)
1991{
1992 struct inode *inode = file_inode(filp);
984fc4e7 1993 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
743b620c
JK		 1994 struct f2fs_inode_info *fi = F2FS_I(inode);
1995 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3db1de0e 1996 struct inode *pinode;
f4c9c743 1997 int ret;
88b88a66 1998
984fc4e7 1999 if (!inode_owner_or_capable(mnt_userns, inode))
88b88a66
JK		 2000 return -EACCES;
2001
e811898c
JK		 2002 if (!S_ISREG(inode->i_mode))
2003 return -EINVAL;
2004
038d0698
CY		 2005 if (filp->f_flags & O_DIRECT)
2006 return -EINVAL;
2007
7fb17fe4
CY		 2008 ret = mnt_want_write_file(filp);
2009 if (ret)
2010 return ret;
2011
0fac558b
CY		 2012 inode_lock(inode);
2013
9b56adcf
FC		 2014 if (!f2fs_disable_compressed_file(inode)) {
2015 ret = -EINVAL;
2016 goto out;
2017 }
4c8ff709 2018
3db1de0e 2019 if (f2fs_is_atomic_file(inode))
7fb17fe4 2020 goto out;
88b88a66 2021
f4c9c743
CY		 2022 ret = f2fs_convert_inline_inode(inode);
2023 if (ret)
7fb17fe4 2024 goto out;
88b88a66 2025
e4544b63 2026 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 2027
31867b23
JK		 2028 /*
2029 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2030 * f2fs_is_atomic_file.
2031 */
2032 if (get_dirty_pages(inode))
dcbb4c10
JP		 2033 f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2034 inode->i_ino, get_dirty_pages(inode));
c27753d6 2035 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
6f8d4455 2036 if (ret) {
e4544b63 2037 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
684ca7e5 2038 goto out;
6f8d4455 2039 }
31867b23 2040
3db1de0e
DJ		 2041 /* Create a COW inode for atomic write */
2042 pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2043 if (IS_ERR(pinode)) {
2044 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2045 ret = PTR_ERR(pinode);
2046 goto out;
2047 }
2048
2049 ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2050 iput(pinode);
2051 if (ret) {
2052 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2053 goto out;
2054 }
2055 f2fs_i_size_write(fi->cow_inode, i_size_read(inode));
2056
743b620c 2057 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
677017d1 2058 sbi->atomic_files++;
743b620c
JK		 2059 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2060
054afda9 2061 set_inode_flag(inode, FI_ATOMIC_FILE);
3db1de0e
DJ		 2062 set_inode_flag(fi->cow_inode, FI_ATOMIC_FILE);
2063 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
e4544b63 2064 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
684ca7e5 2065
6f8d4455 2066 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3db1de0e 2067 F2FS_I(inode)->atomic_write_task = current;
26a28a0c 2068 stat_update_max_atomic_write(inode);
684ca7e5 2069out:
0fac558b 2070 inode_unlock(inode);
7fb17fe4 2071 mnt_drop_write_file(filp);
c27753d6 2072 return ret;
88b88a66
JK		 2073}
2074
2075static int f2fs_ioc_commit_atomic_write(struct file *filp)
2076{
2077 struct inode *inode = file_inode(filp);
984fc4e7 2078 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
88b88a66
JK		 2079 int ret;
2080
984fc4e7 2081 if (!inode_owner_or_capable(mnt_userns, inode))
88b88a66
JK		 2082 return -EACCES;
2083
2084 ret = mnt_want_write_file(filp);
2085 if (ret)
2086 return ret;
2087
6f8d4455 2088 f2fs_balance_fs(F2FS_I_SB(inode), true);
0fac558b 2089
6f8d4455 2090 inode_lock(inode);
1dc0f899 2091
6282adbf 2092 if (f2fs_is_atomic_file(inode)) {
3db1de0e 2093 ret = f2fs_commit_atomic_write(inode);
5fe45743 2094 if (ret)
7bc155fe 2095 goto unlock_out;
5fe45743 2096
26a28a0c 2097 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
743b620c 2098 if (!ret)
3db1de0e 2099 f2fs_abort_atomic_write(inode, false);
26a28a0c 2100 } else {
774e1b78 2101 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
6282adbf 2102 }
7bc155fe 2103unlock_out:
0fac558b 2104 inode_unlock(inode);
88b88a66
JK		 2105 mnt_drop_write_file(filp);
2106 return ret;
2107}
2108
1abff93d
JK		 2109static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2110{
2111 struct inode *inode = file_inode(filp);
2112 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2113 struct super_block *sb = sbi->sb;
2114 __u32 in;
2a96d8ad 2115 int ret = 0;
1abff93d
JK		 2116
2117 if (!capable(CAP_SYS_ADMIN))
2118 return -EPERM;
2119
2120 if (get_user(in, (__u32 __user *)arg))
2121 return -EFAULT;
2122
60b2b4ee
ST		 2123 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2124 ret = mnt_want_write_file(filp);
8626441f
CY		 2125 if (ret) {
2126 if (ret == -EROFS) {
2127 ret = 0;
2128 f2fs_stop_checkpoint(sbi, false);
2129 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2130 trace_f2fs_shutdown(sbi, in, ret);
2131 }
60b2b4ee 2132 return ret;
8626441f 2133 }
60b2b4ee 2134 }
7fb17fe4 2135
1abff93d
JK		 2136 switch (in) {
2137 case F2FS_GOING_DOWN_FULLSYNC:
040f04bd
CH		 2138 ret = freeze_bdev(sb->s_bdev);
2139 if (ret)
d027c484 2140 goto out;
040f04bd
CH		 2141 f2fs_stop_checkpoint(sbi, false);
2142 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2143 thaw_bdev(sb->s_bdev);
1abff93d
JK		 2144 break;
2145 case F2FS_GOING_DOWN_METASYNC:
2146 /* do checkpoint only */
d027c484
CY		 2147 ret = f2fs_sync_fs(sb, 1);
2148 if (ret)
2149 goto out;
38f91ca8 2150 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 2151 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1abff93d
JK		 2152 break;
2153 case F2FS_GOING_DOWN_NOSYNC:
38f91ca8 2154 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 2155 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1abff93d 2156 break;
c912a829 2157 case F2FS_GOING_DOWN_METAFLUSH:
4d57b86d 2158 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
38f91ca8 2159 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 2160 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
c912a829 2161 break;
0cd6d9b0
JK		 2162 case F2FS_GOING_DOWN_NEED_FSCK:
2163 set_sbi_flag(sbi, SBI_NEED_FSCK);
db610a64
JK		 2164 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2165 set_sbi_flag(sbi, SBI_IS_DIRTY);
0cd6d9b0
JK		 2166 /* do checkpoint only */
2167 ret = f2fs_sync_fs(sb, 1);
db610a64 2168 goto out;
1abff93d 2169 default:
7fb17fe4
CY		 2170 ret = -EINVAL;
2171 goto out;
1abff93d 2172 }
7950e9ac 2173
4d57b86d
CY		 2174 f2fs_stop_gc_thread(sbi);
2175 f2fs_stop_discard_thread(sbi);
7950e9ac 2176
4d57b86d 2177 f2fs_drop_discard_cmd(sbi);
7950e9ac
CY		 2178 clear_opt(sbi, DISCARD);
2179
d0239e1b 2180 f2fs_update_time(sbi, REQ_TIME);
7fb17fe4 2181out:
60b2b4ee
ST		 2182 if (in != F2FS_GOING_DOWN_FULLSYNC)
2183 mnt_drop_write_file(filp);
559e87c4
CY		 2184
2185 trace_f2fs_shutdown(sbi, in, ret);
2186
7fb17fe4 2187 return ret;
1abff93d
JK		 2188}
2189
52656e6c
JK		 2190static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2191{
2192 struct inode *inode = file_inode(filp);
2193 struct super_block *sb = inode->i_sb;
2194 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2195 struct fstrim_range range;
2196 int ret;
4b2fecc8 2197
52656e6c
JK		 2198 if (!capable(CAP_SYS_ADMIN))
2199 return -EPERM;
4b2fecc8 2200
7d20c8ab 2201 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
52656e6c 2202 return -EOPNOTSUPP;
4b2fecc8 2203
52656e6c
JK		 2204 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2205 sizeof(range)))
2206 return -EFAULT;
4b2fecc8 2207
7fb17fe4
CY		 2208 ret = mnt_want_write_file(filp);
2209 if (ret)
2210 return ret;
2211
52656e6c
JK		 2212 range.minlen = max((unsigned int)range.minlen,
2213 q->limits.discard_granularity);
2214 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
7fb17fe4 2215 mnt_drop_write_file(filp);
52656e6c
JK		 2216 if (ret < 0)
2217 return ret;
4b2fecc8 2218
52656e6c
JK		 2219 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2220 sizeof(range)))
2221 return -EFAULT;
d0239e1b 2222 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
52656e6c
JK		 2223 return 0;
2224}
2225
f424f664
JK		 2226static bool uuid_is_nonzero(__u8 u[16])
2227{
2228 int i;
2229
2230 for (i = 0; i < 16; i++)
2231 if (u[i])
2232 return true;
2233 return false;
2234}
2235
2236static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2237{
f424f664
JK		 2238 struct inode *inode = file_inode(filp);
2239
7beb01f7 2240 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
ead710b7
CY		 2241 return -EOPNOTSUPP;
2242
d0239e1b 2243 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 2244
db717d8e 2245 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
f424f664
JK		 2246}
2247
2248static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2249{
7beb01f7 2250 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
ead710b7 2251 return -EOPNOTSUPP;
db717d8e 2252 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
f424f664
JK		 2253}
2254
2255static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2256{
2257 struct inode *inode = file_inode(filp);
2258 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2259 int err;
2260
7beb01f7 2261 if (!f2fs_sb_has_encrypt(sbi))
f424f664
JK		 2262 return -EOPNOTSUPP;
2263
f424f664
JK		 2264 err = mnt_want_write_file(filp);
2265 if (err)
2266 return err;
2267
e4544b63 2268 f2fs_down_write(&sbi->sb_lock);
d0d3f1b3
CY		 2269
2270 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2271 goto got_it;
2272
f424f664
JK		 2273 /* update superblock with uuid */
2274 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2275
c5bda1c8 2276 err = f2fs_commit_super(sbi, false);
f424f664
JK		 2277 if (err) {
2278 /* undo new data */
2279 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
d0d3f1b3 2280 goto out_err;
f424f664
JK		 2281 }
2282got_it:
2283 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2284 16))
d0d3f1b3
CY		 2285 err = -EFAULT;
2286out_err:
e4544b63 2287 f2fs_up_write(&sbi->sb_lock);
d0d3f1b3
CY		 2288 mnt_drop_write_file(filp);
2289 return err;
f424f664
JK		 2290}
2291
8ce589c7
EB		 2292static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2293 unsigned long arg)
2294{
2295 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2296 return -EOPNOTSUPP;
2297
2298 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2299}
2300
2301static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2302{
2303 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2304 return -EOPNOTSUPP;
2305
2306 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2307}
2308
2309static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2310{
2311 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2312 return -EOPNOTSUPP;
2313
2314 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2315}
2316
2317static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2318 unsigned long arg)
2319{
2320 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2321 return -EOPNOTSUPP;
2322
2323 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2324}
2325
2326static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2327 unsigned long arg)
2328{
2329 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2330 return -EOPNOTSUPP;
2331
2332 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2333}
2334
ee446e1a
EB		 2335static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2336{
2337 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2338 return -EOPNOTSUPP;
2339
2340 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2341}
2342
c1c1b583
CY		 2343static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2344{
2345 struct inode *inode = file_inode(filp);
2346 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d530d4d8 2347 __u32 sync;
7fb17fe4 2348 int ret;
c1c1b583
CY		 2349
2350 if (!capable(CAP_SYS_ADMIN))
2351 return -EPERM;
2352
d530d4d8 2353 if (get_user(sync, (__u32 __user *)arg))
c1c1b583
CY		 2354 return -EFAULT;
2355
d530d4d8
CY		 2356 if (f2fs_readonly(sbi->sb))
2357 return -EROFS;
c1c1b583 2358
7fb17fe4
CY		 2359 ret = mnt_want_write_file(filp);
2360 if (ret)
2361 return ret;
2362
d530d4d8 2363 if (!sync) {
e4544b63 2364 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
7fb17fe4
CY		 2365 ret = -EBUSY;
2366 goto out;
2367 }
d530d4d8 2368 } else {
e4544b63 2369 f2fs_down_write(&sbi->gc_lock);
c1c1b583
CY		 2370 }
2371
7dede886 2372 ret = f2fs_gc(sbi, sync, true, false, NULL_SEGNO);
7fb17fe4
CY		 2373out:
2374 mnt_drop_write_file(filp);
2375 return ret;
c1c1b583
CY		 2376}
2377
34178b1b 2378static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
34dc77ad 2379{
34178b1b 2380 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
34dc77ad
JK		 2381 u64 end;
2382 int ret;
2383
2384 if (!capable(CAP_SYS_ADMIN))
2385 return -EPERM;
34dc77ad
JK		 2386 if (f2fs_readonly(sbi->sb))
2387 return -EROFS;
2388
34178b1b
CY		 2389 end = range->start + range->len;
2390 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
fbbf7799 2391 end >= MAX_BLKADDR(sbi))
b82f6e34 2392 return -EINVAL;
b82f6e34 2393
34dc77ad
JK		 2394 ret = mnt_want_write_file(filp);
2395 if (ret)
2396 return ret;
2397
34dc77ad 2398do_more:
34178b1b 2399 if (!range->sync) {
e4544b63 2400 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
34dc77ad
JK		 2401 ret = -EBUSY;
2402 goto out;
2403 }
2404 } else {
e4544b63 2405 f2fs_down_write(&sbi->gc_lock);
34dc77ad
JK		 2406 }
2407
7dede886
CY		 2408 ret = f2fs_gc(sbi, range->sync, true, false,
2409 GET_SEGNO(sbi, range->start));
97767500
QZ		 2410 if (ret) {
2411 if (ret == -EBUSY)
2412 ret = -EAGAIN;
2413 goto out;
2414 }
34178b1b
CY		 2415 range->start += BLKS_PER_SEC(sbi);
2416 if (range->start <= end)
34dc77ad
JK		 2417 goto do_more;
2418out:
2419 mnt_drop_write_file(filp);
2420 return ret;
2421}
2422
34178b1b
CY		 2423static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2424{
2425 struct f2fs_gc_range range;
2426
2427 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2428 sizeof(range)))
2429 return -EFAULT;
2430 return __f2fs_ioc_gc_range(filp, &range);
2431}
2432
059c0648 2433static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
456b88e4
CY		 2434{
2435 struct inode *inode = file_inode(filp);
2436 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
7fb17fe4 2437 int ret;
456b88e4
CY		 2438
2439 if (!capable(CAP_SYS_ADMIN))
2440 return -EPERM;
2441
2442 if (f2fs_readonly(sbi->sb))
2443 return -EROFS;
2444
4354994f 2445 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
dcbb4c10 2446 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
4354994f
DR		 2447 return -EINVAL;
2448 }
2449
7fb17fe4
CY		 2450 ret = mnt_want_write_file(filp);
2451 if (ret)
2452 return ret;
2453
2454 ret = f2fs_sync_fs(sbi->sb, 1);
2455
2456 mnt_drop_write_file(filp);
2457 return ret;
456b88e4
CY		 2458}
2459
d323d005
CY		 2460static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2461 struct file *filp,
2462 struct f2fs_defragment *range)
2463{
2464 struct inode *inode = file_inode(filp);
f3d98e74 2465 struct f2fs_map_blocks map = { .m_next_extent = NULL,
5f029c04 2466 .m_seg_type = NO_CHECK_TYPE,
f4f0b677 2467 .m_may_create = false };
1061fd48 2468 struct extent_info ei = {0, 0, 0};
f3d98e74 2469 pgoff_t pg_start, pg_end, next_pgofs;
3519e3f9 2470 unsigned int blk_per_seg = sbi->blocks_per_seg;
d323d005 2471 unsigned int total = 0, sec_num;
d323d005
CY		 2472 block_t blk_end = 0;
2473 bool fragmented = false;
2474 int err;
2475
09cbfeaf
KS		 2476 pg_start = range->start >> PAGE_SHIFT;
2477 pg_end = (range->start + range->len) >> PAGE_SHIFT;
d323d005 2478
2c4db1a6 2479 f2fs_balance_fs(sbi, true);
d323d005 2480
5955102c 2481 inode_lock(inode);
d323d005 2482
1018a546
CY		 2483 /* if in-place-update policy is enabled, don't waste time here */
2484 set_inode_flag(inode, FI_OPU_WRITE);
2485 if (f2fs_should_update_inplace(inode, NULL)) {
2486 err = -EINVAL;
2487 goto out;
2488 }
2489
d323d005
CY		 2490 /* writeback all dirty pages in the range */
2491 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
d8fe4f0e 2492 range->start + range->len - 1);
d323d005
CY		 2493 if (err)
2494 goto out;
2495
2496 /*
2497 * lookup mapping info in extent cache, skip defragmenting if physical
2498 * block addresses are continuous.
2499 */
2500 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2501 if (ei.fofs + ei.len >= pg_end)
2502 goto out;
2503 }
2504
2505 map.m_lblk = pg_start;
f3d98e74 2506 map.m_next_pgofs = &next_pgofs;
d323d005
CY		 2507
2508 /*
2509 * lookup mapping info in dnode page cache, skip defragmenting if all
2510 * physical block addresses are continuous even if there are hole(s)
2511 * in logical blocks.
2512 */
2513 while (map.m_lblk < pg_end) {
a1c1e9b7 2514 map.m_len = pg_end - map.m_lblk;
f2220c7f 2515 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2516 if (err)
2517 goto out;
2518
2519 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2520 map.m_lblk = next_pgofs;
d323d005
CY		 2521 continue;
2522 }
2523
25a912e5 2524 if (blk_end && blk_end != map.m_pblk)
d323d005 2525 fragmented = true;
25a912e5
CY		 2526
2527 /* record total count of block that we're going to move */
2528 total += map.m_len;
2529
d323d005
CY		 2530 blk_end = map.m_pblk + map.m_len;
2531
2532 map.m_lblk += map.m_len;
d323d005
CY		 2533 }
2534
d3a1a0e1
CY		 2535 if (!fragmented) {
2536 total = 0;
d323d005 2537 goto out;
d3a1a0e1 2538 }
d323d005 2539
f91108b8 2540 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
d323d005
CY		 2541
2542 /*
2543 * make sure there are enough free section for LFS allocation, this can
2544 * avoid defragment running in SSR mode when free section are allocated
2545 * intensively
2546 */
7f3037a5 2547 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
d323d005
CY		 2548 err = -EAGAIN;
2549 goto out;
2550 }
2551
25a912e5
CY		 2552 map.m_lblk = pg_start;
2553 map.m_len = pg_end - pg_start;
2554 total = 0;
2555
d323d005
CY		 2556 while (map.m_lblk < pg_end) {
2557 pgoff_t idx;
2558 int cnt = 0;
2559
2560do_map:
a1c1e9b7 2561 map.m_len = pg_end - map.m_lblk;
f2220c7f 2562 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2563 if (err)
2564 goto clear_out;
2565
2566 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2567 map.m_lblk = next_pgofs;
d3a1a0e1 2568 goto check;
d323d005
CY		 2569 }
2570
1018a546 2571 set_inode_flag(inode, FI_SKIP_WRITES);
d323d005
CY		 2572
2573 idx = map.m_lblk;
2574 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2575 struct page *page;
2576
4d57b86d 2577 page = f2fs_get_lock_data_page(inode, idx, true);
d323d005
CY		 2578 if (IS_ERR(page)) {
2579 err = PTR_ERR(page);
2580 goto clear_out;
2581 }
2582
2583 set_page_dirty(page);
2584 f2fs_put_page(page, 1);
2585
2586 idx++;
2587 cnt++;
2588 total++;
2589 }
2590
2591 map.m_lblk = idx;
d3a1a0e1
CY		 2592check:
2593 if (map.m_lblk < pg_end && cnt < blk_per_seg)
d323d005
CY		 2594 goto do_map;
2595
1018a546 2596 clear_inode_flag(inode, FI_SKIP_WRITES);
d323d005
CY		 2597
2598 err = filemap_fdatawrite(inode->i_mapping);
2599 if (err)
2600 goto out;
2601 }
2602clear_out:
1018a546 2603 clear_inode_flag(inode, FI_SKIP_WRITES);
d323d005 2604out:
1018a546 2605 clear_inode_flag(inode, FI_OPU_WRITE);
5955102c 2606 inode_unlock(inode);
d323d005 2607 if (!err)
09cbfeaf 2608 range->len = (u64)total << PAGE_SHIFT;
d323d005
CY		 2609 return err;
2610}
2611
2612static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2613{
2614 struct inode *inode = file_inode(filp);
2615 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2616 struct f2fs_defragment range;
2617 int err;
2618
2619 if (!capable(CAP_SYS_ADMIN))
2620 return -EPERM;
2621
7eab0c0d 2622 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
d323d005
CY		 2623 return -EINVAL;
2624
d7563861
KM		 2625 if (f2fs_readonly(sbi->sb))
2626 return -EROFS;
d323d005
CY		 2627
2628 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
d7563861
KM		 2629 sizeof(range)))
2630 return -EFAULT;
d323d005
CY		 2631
2632 /* verify alignment of offset & size */
d7563861
KM		 2633 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2634 return -EINVAL;
d323d005 2635
1941d7bc 2636 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
6d1451bf 2637 max_file_blocks(inode)))
d7563861
KM		 2638 return -EINVAL;
2639
2640 err = mnt_want_write_file(filp);
2641 if (err)
2642 return err;
1941d7bc 2643
d323d005 2644 err = f2fs_defragment_range(sbi, filp, &range);
d7563861
KM		 2645 mnt_drop_write_file(filp);
2646
d0239e1b 2647 f2fs_update_time(sbi, REQ_TIME);
d323d005 2648 if (err < 0)
d7563861 2649 return err;
d323d005
CY		 2650
2651 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2652 sizeof(range)))
d7563861
KM		 2653 return -EFAULT;
2654
2655 return 0;
d323d005
CY		 2656}
2657
4dd6f977
JK		 2658static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2659 struct file *file_out, loff_t pos_out, size_t len)
2660{
2661 struct inode *src = file_inode(file_in);
2662 struct inode *dst = file_inode(file_out);
2663 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2664 size_t olen = len, dst_max_i_size = 0;
2665 size_t dst_osize;
2666 int ret;
2667
2668 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2669 src->i_sb != dst->i_sb)
2670 return -EXDEV;
2671
2672 if (unlikely(f2fs_readonly(src->i_sb)))
2673 return -EROFS;
2674
fe8494bf
CY		 2675 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2676 return -EINVAL;
4dd6f977 2677
62230e0d 2678 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
4dd6f977
JK		 2679 return -EOPNOTSUPP;
2680
aad1383c
DR		 2681 if (pos_out < 0 || pos_in < 0)
2682 return -EINVAL;
2683
d95fd91c
FL		 2684 if (src == dst) {
2685 if (pos_in == pos_out)
2686 return 0;
2687 if (pos_out > pos_in && pos_out < pos_in + len)
2688 return -EINVAL;
2689 }
2690
4dd6f977 2691 inode_lock(src);
20a3d61d 2692 if (src != dst) {
bb06664a
CY		 2693 ret = -EBUSY;
2694 if (!inode_trylock(dst))
2695 goto out;
20a3d61d 2696 }
4dd6f977
JK		 2697
2698 ret = -EINVAL;
2699 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2700 goto out_unlock;
2701 if (len == 0)
2702 olen = len = src->i_size - pos_in;
2703 if (pos_in + len == src->i_size)
2704 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2705 if (len == 0) {
2706 ret = 0;
2707 goto out_unlock;
2708 }
2709
2710 dst_osize = dst->i_size;
2711 if (pos_out + olen > dst->i_size)
2712 dst_max_i_size = pos_out + olen;
2713
2714 /* verify the end result is block aligned */
2715 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2716 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2717 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2718 goto out_unlock;
2719
2720 ret = f2fs_convert_inline_inode(src);
2721 if (ret)
2722 goto out_unlock;
2723
2724 ret = f2fs_convert_inline_inode(dst);
2725 if (ret)
2726 goto out_unlock;
2727
2728 /* write out all dirty pages from offset */
2729 ret = filemap_write_and_wait_range(src->i_mapping,
2730 pos_in, pos_in + len);
2731 if (ret)
2732 goto out_unlock;
2733
2734 ret = filemap_write_and_wait_range(dst->i_mapping,
2735 pos_out, pos_out + len);
2736 if (ret)
2737 goto out_unlock;
2738
2739 f2fs_balance_fs(sbi, true);
6f8d4455 2740
e4544b63 2741 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
6f8d4455
JK		 2742 if (src != dst) {
2743 ret = -EBUSY;
e4544b63 2744 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
6f8d4455
JK		 2745 goto out_src;
2746 }
2747
4dd6f977 2748 f2fs_lock_op(sbi);
61e4da11
FL		 2749 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2750 pos_out >> F2FS_BLKSIZE_BITS,
2751 len >> F2FS_BLKSIZE_BITS, false);
4dd6f977
JK		 2752
2753 if (!ret) {
2754 if (dst_max_i_size)
2755 f2fs_i_size_write(dst, dst_max_i_size);
2756 else if (dst_osize != dst->i_size)
2757 f2fs_i_size_write(dst, dst_osize);
2758 }
2759 f2fs_unlock_op(sbi);
6f8d4455
JK		 2760
2761 if (src != dst)
e4544b63 2762 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
6f8d4455 2763out_src:
e4544b63 2764 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
6f8d4455
JK		 2765out_unlock:
2766 if (src != dst)
4dd6f977 2767 inode_unlock(dst);
20a3d61d 2768out:
4dd6f977
JK		 2769 inode_unlock(src);
2770 return ret;
2771}
2772
34178b1b
CY		 2773static int __f2fs_ioc_move_range(struct file *filp,
2774 struct f2fs_move_range *range)
4dd6f977 2775{
4dd6f977
JK		 2776 struct fd dst;
2777 int err;
2778
2779 if (!(filp->f_mode & FMODE_READ) ||
2780 !(filp->f_mode & FMODE_WRITE))
2781 return -EBADF;
2782
34178b1b 2783 dst = fdget(range->dst_fd);
4dd6f977
JK		 2784 if (!dst.file)
2785 return -EBADF;
2786
2787 if (!(dst.file->f_mode & FMODE_WRITE)) {
2788 err = -EBADF;
2789 goto err_out;
2790 }
2791
2792 err = mnt_want_write_file(filp);
2793 if (err)
2794 goto err_out;
2795
34178b1b
CY		 2796 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2797 range->pos_out, range->len);
4dd6f977
JK		 2798
2799 mnt_drop_write_file(filp);
4dd6f977
JK		 2800err_out:
2801 fdput(dst);
2802 return err;
2803}
2804
34178b1b
CY		 2805static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2806{
2807 struct f2fs_move_range range;
2808
2809 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2810 sizeof(range)))
2811 return -EFAULT;
2812 return __f2fs_ioc_move_range(filp, &range);
2813}
2814
e066b83c
JK		 2815static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2816{
2817 struct inode *inode = file_inode(filp);
2818 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2819 struct sit_info *sm = SIT_I(sbi);
2820 unsigned int start_segno = 0, end_segno = 0;
2821 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2822 struct f2fs_flush_device range;
2823 int ret;
2824
2825 if (!capable(CAP_SYS_ADMIN))
2826 return -EPERM;
2827
2828 if (f2fs_readonly(sbi->sb))
2829 return -EROFS;
2830
4354994f
DR		 2831 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2832 return -EINVAL;
2833
e066b83c
JK		 2834 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2835 sizeof(range)))
2836 return -EFAULT;
2837
0916878d 2838 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2c70c5e3 2839 __is_large_section(sbi)) {
dcbb4c10
JP		 2840 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2841 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
e066b83c
JK		 2842 return -EINVAL;
2843 }
2844
2845 ret = mnt_want_write_file(filp);
2846 if (ret)
2847 return ret;
2848
2849 if (range.dev_num != 0)
2850 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2851 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2852
2853 start_segno = sm->last_victim[FLUSH_DEVICE];
2854 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2855 start_segno = dev_start_segno;
2856 end_segno = min(start_segno + range.segments, dev_end_segno);
2857
2858 while (start_segno < end_segno) {
e4544b63 2859 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
e066b83c
JK		 2860 ret = -EBUSY;
2861 goto out;
2862 }
2863 sm->last_victim[GC_CB] = end_segno + 1;
2864 sm->last_victim[GC_GREEDY] = end_segno + 1;
2865 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
7dede886 2866 ret = f2fs_gc(sbi, true, true, true, start_segno);
e066b83c
JK		 2867 if (ret == -EAGAIN)
2868 ret = 0;
2869 else if (ret < 0)
2870 break;
2871 start_segno++;
2872 }
2873out:
2874 mnt_drop_write_file(filp);
2875 return ret;
2876}
2877
e65ef207
JK		 2878static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2879{
2880 struct inode *inode = file_inode(filp);
2881 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2882
2883 /* Must validate to set it with SQLite behavior in Android. */
2884 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2885
2886 return put_user(sb_feature, (u32 __user *)arg);
2887}
e066b83c 2888
2c1d0305 2889#ifdef CONFIG_QUOTA
78130819
CY		 2890int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2891{
2892 struct dquot *transfer_to[MAXQUOTAS] = {};
2893 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2894 struct super_block *sb = sbi->sb;
2895 int err = 0;
2896
2897 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2898 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2899 err = __dquot_transfer(inode, transfer_to);
2900 if (err)
2901 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2902 dqput(transfer_to[PRJQUOTA]);
2903 }
2904 return err;
2905}
2906
9b1bb01c 2907static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2c1d0305 2908{
2c1d0305
CY		 2909 struct f2fs_inode_info *fi = F2FS_I(inode);
2910 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d13732cc 2911 struct f2fs_inode *ri = NULL;
2c1d0305
CY		 2912 kprojid_t kprojid;
2913 int err;
2914
7beb01f7 2915 if (!f2fs_sb_has_project_quota(sbi)) {
2c1d0305
CY		 2916 if (projid != F2FS_DEF_PROJID)
2917 return -EOPNOTSUPP;
2918 else
2919 return 0;
2920 }
2921
2922 if (!f2fs_has_extra_attr(inode))
2923 return -EOPNOTSUPP;
2924
2925 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2926
2927 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2928 return 0;
2929
2c1d0305 2930 err = -EPERM;
2c1d0305
CY		 2931 /* Is it quota file? Do not allow user to mess with it */
2932 if (IS_NOQUOTA(inode))
c8e92757 2933 return err;
2c1d0305 2934
d13732cc
JY		 2935 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
2936 return -EOVERFLOW;
2c1d0305 2937
10a26878 2938 err = f2fs_dquot_initialize(inode);
c22aecd7 2939 if (err)
c8e92757 2940 return err;
2c1d0305 2941
78130819
CY		 2942 f2fs_lock_op(sbi);
2943 err = f2fs_transfer_project_quota(inode, kprojid);
2944 if (err)
2945 goto out_unlock;
2c1d0305
CY		 2946
2947 F2FS_I(inode)->i_projid = kprojid;
2948 inode->i_ctime = current_time(inode);
2c1d0305 2949 f2fs_mark_inode_dirty_sync(inode, true);
78130819
CY		 2950out_unlock:
2951 f2fs_unlock_op(sbi);
2c1d0305
CY		 2952 return err;
2953}
2954#else
78130819
CY		 2955int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2956{
2957 return 0;
2958}
2959
9b1bb01c 2960static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2c1d0305
CY		 2961{
2962 if (projid != F2FS_DEF_PROJID)
2963 return -EOPNOTSUPP;
2964 return 0;
2965}
2966#endif
2967
9b1bb01c 2968int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
2c1d0305 2969{
9b1bb01c 2970 struct inode *inode = d_inode(dentry);
2c1d0305 2971 struct f2fs_inode_info *fi = F2FS_I(inode);
9b1bb01c 2972 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
2c1d0305 2973
9b1bb01c
MS		 2974 if (IS_ENCRYPTED(inode))
2975 fsflags |= FS_ENCRYPT_FL;
2976 if (IS_VERITY(inode))
2977 fsflags |= FS_VERITY_FL;
2978 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
2979 fsflags |= FS_INLINE_DATA_FL;
2980 if (is_inode_flag_set(inode, FI_PIN_FILE))
2981 fsflags |= FS_NOCOW_FL;
2982
2983 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
2c1d0305 2984
7beb01f7 2985 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
6fc93c4e 2986 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2c1d0305 2987
c8e92757
WS		 2988 return 0;
2989}
2990
9b1bb01c
MS		 2991int f2fs_fileattr_set(struct user_namespace *mnt_userns,
2992 struct dentry *dentry, struct fileattr *fa)
2c1d0305 2993{
9b1bb01c
MS		 2994 struct inode *inode = d_inode(dentry);
2995 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
36098557 2996 u32 iflags;
2c1d0305
CY		 2997 int err;
2998
9b1bb01c
MS		 2999 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3000 return -EIO;
3001 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3002 return -ENOSPC;
3003 if (fsflags & ~F2FS_GETTABLE_FS_FL)
2c1d0305 3004 return -EOPNOTSUPP;
9b1bb01c
MS		 3005 fsflags &= F2FS_SETTABLE_FS_FL;
3006 if (!fa->flags_valid)
3007 mask &= FS_COMMON_FL;
2c1d0305 3008
9b1bb01c 3009 iflags = f2fs_fsflags_to_iflags(fsflags);
36098557 3010 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2c1d0305
CY		 3011 return -EOPNOTSUPP;
3012
9b1bb01c
MS		 3013 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3014 if (!err)
3015 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
2c1d0305 3016
c8e92757 3017 return err;
2c1d0305 3018}
e066b83c 3019
1ad71a27
JK		 3020int f2fs_pin_file_control(struct inode *inode, bool inc)
3021{
3022 struct f2fs_inode_info *fi = F2FS_I(inode);
3023 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3024
3025 /* Use i_gc_failures for normal file as a risk signal. */
3026 if (inc)
2ef79ecb
CY		 3027 f2fs_i_gc_failures_write(inode,
3028 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
1ad71a27 3029
2ef79ecb 3030 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
dcbb4c10
JP		 3031 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3032 __func__, inode->i_ino,
3033 fi->i_gc_failures[GC_FAILURE_PIN]);
1ad71a27
JK		 3034 clear_inode_flag(inode, FI_PIN_FILE);
3035 return -EAGAIN;
3036 }
3037 return 0;
3038}
3039
3040static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3041{
3042 struct inode *inode = file_inode(filp);
3043 __u32 pin;
3044 int ret = 0;
3045
1ad71a27
JK		 3046 if (get_user(pin, (__u32 __user *)arg))
3047 return -EFAULT;
3048
3049 if (!S_ISREG(inode->i_mode))
3050 return -EINVAL;
3051
3052 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3053 return -EROFS;
3054
3055 ret = mnt_want_write_file(filp);
3056 if (ret)
3057 return ret;
3058
3059 inode_lock(inode);
3060
3061 if (!pin) {
3062 clear_inode_flag(inode, FI_PIN_FILE);
30933364 3063 f2fs_i_gc_failures_write(inode, 0);
1ad71a27
JK		 3064 goto done;
3065 }
3066
19bdba52
JK		 3067 if (f2fs_should_update_outplace(inode, NULL)) {
3068 ret = -EINVAL;
3069 goto out;
3070 }
3071
1ad71a27
JK		 3072 if (f2fs_pin_file_control(inode, false)) {
3073 ret = -EAGAIN;
3074 goto out;
3075 }
4c8ff709 3076
1ad71a27
JK		 3077 ret = f2fs_convert_inline_inode(inode);
3078 if (ret)
3079 goto out;
3080
78134d03 3081 if (!f2fs_disable_compressed_file(inode)) {
4c8ff709
CY		 3082 ret = -EOPNOTSUPP;
3083 goto out;
3084 }
3085
1ad71a27 3086 set_inode_flag(inode, FI_PIN_FILE);
2ef79ecb 3087 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
1ad71a27
JK		 3088done:
3089 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3090out:
3091 inode_unlock(inode);
3092 mnt_drop_write_file(filp);
3093 return ret;
3094}
3095
3096static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3097{
3098 struct inode *inode = file_inode(filp);
3099 __u32 pin = 0;
3100
3101 if (is_inode_flag_set(inode, FI_PIN_FILE))
2ef79ecb 3102 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
1ad71a27
JK		 3103 return put_user(pin, (u32 __user *)arg);
3104}
3105
c4020b2d
CY		 3106int f2fs_precache_extents(struct inode *inode)
3107{
3108 struct f2fs_inode_info *fi = F2FS_I(inode);
3109 struct f2fs_map_blocks map;
3110 pgoff_t m_next_extent;
3111 loff_t end;
3112 int err;
3113
3114 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3115 return -EOPNOTSUPP;
3116
3117 map.m_lblk = 0;
3118 map.m_next_pgofs = NULL;
3119 map.m_next_extent = &m_next_extent;
3120 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 3121 map.m_may_create = false;
6d1451bf 3122 end = max_file_blocks(inode);
c4020b2d
CY		 3123
3124 while (map.m_lblk < end) {
3125 map.m_len = end - map.m_lblk;
3126
e4544b63 3127 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
c4020b2d 3128 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
e4544b63 3129 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
c4020b2d
CY		 3130 if (err)
3131 return err;
3132
3133 map.m_lblk = m_next_extent;
3134 }
3135
4f55dc2a 3136 return 0;
c4020b2d
CY		 3137}
3138
3139static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3140{
3141 return f2fs_precache_extents(file_inode(filp));
3142}
3143
04f0b2ea
QS		 3144static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3145{
3146 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3147 __u64 block_count;
04f0b2ea
QS		 3148
3149 if (!capable(CAP_SYS_ADMIN))
3150 return -EPERM;
3151
3152 if (f2fs_readonly(sbi->sb))
3153 return -EROFS;
3154
3155 if (copy_from_user(&block_count, (void __user *)arg,
3156 sizeof(block_count)))
3157 return -EFAULT;
3158
b4b10061 3159 return f2fs_resize_fs(sbi, block_count);
04f0b2ea
QS		 3160}
3161
95ae251f
EB		 3162static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3163{
3164 struct inode *inode = file_inode(filp);
3165
3166 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3167
3168 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3169 f2fs_warn(F2FS_I_SB(inode),
833dcd35 3170 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
95ae251f
EB		 3171 inode->i_ino);
3172 return -EOPNOTSUPP;
3173 }
3174
3175 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3176}
3177
3178static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3179{
3180 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3181 return -EOPNOTSUPP;
3182
3183 return fsverity_ioctl_measure(filp, (void __user *)arg);
3184}
3185
e17fe657
EB		 3186static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3187{
3188 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3189 return -EOPNOTSUPP;
3190
3191 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3192}
3193
3357af8f 3194static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
4507847c
CY		 3195{
3196 struct inode *inode = file_inode(filp);
3197 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3198 char *vbuf;
3199 int count;
3200 int err = 0;
3201
3202 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3203 if (!vbuf)
3204 return -ENOMEM;
3205
e4544b63 3206 f2fs_down_read(&sbi->sb_lock);
4507847c
CY		 3207 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3208 ARRAY_SIZE(sbi->raw_super->volume_name),
3209 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
e4544b63 3210 f2fs_up_read(&sbi->sb_lock);
4507847c
CY		 3211
3212 if (copy_to_user((char __user *)arg, vbuf,
3213 min(FSLABEL_MAX, count)))
3214 err = -EFAULT;
3215
c8eb7024 3216 kfree(vbuf);
4507847c
CY		 3217 return err;
3218}
3219
3357af8f 3220static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
4507847c
CY		 3221{
3222 struct inode *inode = file_inode(filp);
3223 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3224 char *vbuf;
3225 int err = 0;
3226
3227 if (!capable(CAP_SYS_ADMIN))
3228 return -EPERM;
3229
3230 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3231 if (IS_ERR(vbuf))
3232 return PTR_ERR(vbuf);
3233
3234 err = mnt_want_write_file(filp);
3235 if (err)
3236 goto out;
3237
e4544b63 3238 f2fs_down_write(&sbi->sb_lock);
4507847c
CY		 3239
3240 memset(sbi->raw_super->volume_name, 0,
3241 sizeof(sbi->raw_super->volume_name));
3242 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3243 sbi->raw_super->volume_name,
3244 ARRAY_SIZE(sbi->raw_super->volume_name));
3245
3246 err = f2fs_commit_super(sbi, false);
3247
e4544b63 3248 f2fs_up_write(&sbi->sb_lock);
4507847c
CY		 3249
3250 mnt_drop_write_file(filp);
3251out:
3252 kfree(vbuf);
3253 return err;
3254}
3255
439dfb10
CY		 3256static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3257{
3258 struct inode *inode = file_inode(filp);
3259 __u64 blocks;
3260
3261 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3262 return -EOPNOTSUPP;
3263
3264 if (!f2fs_compressed_file(inode))
3265 return -EINVAL;
3266
c2759eba 3267 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
439dfb10
CY		 3268 return put_user(blocks, (u64 __user *)arg);
3269}
3270
ef8d563f
CY		 3271static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3272{
3273 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3274 unsigned int released_blocks = 0;
3275 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3276 block_t blkaddr;
3277 int i;
3278
3279 for (i = 0; i < count; i++) {
3280 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3281 dn->ofs_in_node + i);
3282
3283 if (!__is_valid_data_blkaddr(blkaddr))
3284 continue;
3285 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3286 DATA_GENERIC_ENHANCE)))
3287 return -EFSCORRUPTED;
3288 }
3289
3290 while (count) {
3291 int compr_blocks = 0;
3292
3293 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3294 blkaddr = f2fs_data_blkaddr(dn);
3295
3296 if (i == 0) {
3297 if (blkaddr == COMPRESS_ADDR)
3298 continue;
3299 dn->ofs_in_node += cluster_size;
3300 goto next;
3301 }
3302
3303 if (__is_valid_data_blkaddr(blkaddr))
3304 compr_blocks++;
3305
3306 if (blkaddr != NEW_ADDR)
3307 continue;
3308
3309 dn->data_blkaddr = NULL_ADDR;
3310 f2fs_set_data_blkaddr(dn);
3311 }
3312
3313 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3314 dec_valid_block_count(sbi, dn->inode,
3315 cluster_size - compr_blocks);
3316
3317 released_blocks += cluster_size - compr_blocks;
3318next:
3319 count -= cluster_size;
3320 }
3321
3322 return released_blocks;
3323}
3324
3325static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3326{
3327 struct inode *inode = file_inode(filp);
3328 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3329 pgoff_t page_idx = 0, last_idx;
3330 unsigned int released_blocks = 0;
3331 int ret;
3332 int writecount;
3333
3334 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3335 return -EOPNOTSUPP;
3336
3337 if (!f2fs_compressed_file(inode))
3338 return -EINVAL;
3339
3340 if (f2fs_readonly(sbi->sb))
3341 return -EROFS;
3342
3343 ret = mnt_want_write_file(filp);
3344 if (ret)
3345 return ret;
3346
3347 f2fs_balance_fs(F2FS_I_SB(inode), true);
3348
3349 inode_lock(inode);
3350
3351 writecount = atomic_read(&inode->i_writecount);
8c8cf26a
DJ		 3352 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3353 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
ef8d563f
CY		 3354 ret = -EBUSY;
3355 goto out;
3356 }
3357
c6140415 3358 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
ef8d563f
CY		 3359 ret = -EINVAL;
3360 goto out;
3361 }
3362
3363 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3364 if (ret)
3365 goto out;
3366
c6140415 3367 set_inode_flag(inode, FI_COMPRESS_RELEASED);
ef8d563f
CY		 3368 inode->i_ctime = current_time(inode);
3369 f2fs_mark_inode_dirty_sync(inode, true);
3370
c2759eba 3371 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
567c4bf5
DJ		 3372 goto out;
3373
e4544b63 3374 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 3375 filemap_invalidate_lock(inode->i_mapping);
ef8d563f
CY		 3376
3377 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3378
3379 while (page_idx < last_idx) {
3380 struct dnode_of_data dn;
3381 pgoff_t end_offset, count;
3382
3383 set_new_dnode(&dn, inode, NULL, NULL, 0);
3384 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3385 if (ret) {
3386 if (ret == -ENOENT) {
3387 page_idx = f2fs_get_next_page_offset(&dn,
3388 page_idx);
3389 ret = 0;
3390 continue;
3391 }
3392 break;
3393 }
3394
3395 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3396 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4fec3fc0 3397 count = round_up(count, F2FS_I(inode)->i_cluster_size);
ef8d563f
CY		 3398
3399 ret = release_compress_blocks(&dn, count);
3400
3401 f2fs_put_dnode(&dn);
3402
3403 if (ret < 0)
3404 break;
3405
3406 page_idx += count;
3407 released_blocks += ret;
3408 }
3409
edc6d01b 3410 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 3411 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
ef8d563f
CY		 3412out:
3413 inode_unlock(inode);
3414
3415 mnt_drop_write_file(filp);
3416
3417 if (ret >= 0) {
3418 ret = put_user(released_blocks, (u64 __user *)arg);
c2759eba
DJ		 3419 } else if (released_blocks &&
3420 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
ef8d563f
CY		 3421 set_sbi_flag(sbi, SBI_NEED_FSCK);
3422 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
c2759eba 3423 "iblocks=%llu, released=%u, compr_blocks=%u, "
ef8d563f
CY		 3424 "run fsck to fix.",
3425 __func__, inode->i_ino, inode->i_blocks,
3426 released_blocks,
c2759eba 3427 atomic_read(&F2FS_I(inode)->i_compr_blocks));
ef8d563f
CY		 3428 }
3429
3430 return ret;
3431}
3432
c75488fb
CY		 3433static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3434{
3435 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3436 unsigned int reserved_blocks = 0;
3437 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3438 block_t blkaddr;
3439 int i;
3440
3441 for (i = 0; i < count; i++) {
3442 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3443 dn->ofs_in_node + i);
3444
3445 if (!__is_valid_data_blkaddr(blkaddr))
3446 continue;
3447 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3448 DATA_GENERIC_ENHANCE)))
3449 return -EFSCORRUPTED;
3450 }
3451
3452 while (count) {
3453 int compr_blocks = 0;
3454 blkcnt_t reserved;
3455 int ret;
3456
3457 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3458 blkaddr = f2fs_data_blkaddr(dn);
3459
3460 if (i == 0) {
3461 if (blkaddr == COMPRESS_ADDR)
3462 continue;
3463 dn->ofs_in_node += cluster_size;
3464 goto next;
3465 }
3466
3467 if (__is_valid_data_blkaddr(blkaddr)) {
3468 compr_blocks++;
3469 continue;
3470 }
3471
3472 dn->data_blkaddr = NEW_ADDR;
3473 f2fs_set_data_blkaddr(dn);
3474 }
3475
3476 reserved = cluster_size - compr_blocks;
3477 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3478 if (ret)
3479 return ret;
3480
3481 if (reserved != cluster_size - compr_blocks)
3482 return -ENOSPC;
3483
3484 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3485
3486 reserved_blocks += reserved;
3487next:
3488 count -= cluster_size;
3489 }
3490
3491 return reserved_blocks;
3492}
3493
3494static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3495{
3496 struct inode *inode = file_inode(filp);
3497 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3498 pgoff_t page_idx = 0, last_idx;
3499 unsigned int reserved_blocks = 0;
3500 int ret;
3501
3502 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3503 return -EOPNOTSUPP;
3504
3505 if (!f2fs_compressed_file(inode))
3506 return -EINVAL;
3507
3508 if (f2fs_readonly(sbi->sb))
3509 return -EROFS;
3510
3511 ret = mnt_want_write_file(filp);
3512 if (ret)
3513 return ret;
3514
c2759eba 3515 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
c75488fb
CY		 3516 goto out;
3517
3518 f2fs_balance_fs(F2FS_I_SB(inode), true);
3519
3520 inode_lock(inode);
3521
c6140415 3522 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
c75488fb
CY		 3523 ret = -EINVAL;
3524 goto unlock_inode;
3525 }
3526
e4544b63 3527 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 3528 filemap_invalidate_lock(inode->i_mapping);
c75488fb
CY		 3529
3530 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3531
3532 while (page_idx < last_idx) {
3533 struct dnode_of_data dn;
3534 pgoff_t end_offset, count;
3535
3536 set_new_dnode(&dn, inode, NULL, NULL, 0);
3537 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3538 if (ret) {
3539 if (ret == -ENOENT) {
3540 page_idx = f2fs_get_next_page_offset(&dn,
3541 page_idx);
3542 ret = 0;
3543 continue;
3544 }
3545 break;
3546 }
3547
3548 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3549 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4fec3fc0 3550 count = round_up(count, F2FS_I(inode)->i_cluster_size);
c75488fb
CY		 3551
3552 ret = reserve_compress_blocks(&dn, count);
3553
3554 f2fs_put_dnode(&dn);
3555
3556 if (ret < 0)
3557 break;
3558
3559 page_idx += count;
3560 reserved_blocks += ret;
3561 }
3562
edc6d01b 3563 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 3564 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
c75488fb
CY		 3565
3566 if (ret >= 0) {
c6140415 3567 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
c75488fb
CY		 3568 inode->i_ctime = current_time(inode);
3569 f2fs_mark_inode_dirty_sync(inode, true);
3570 }
3571unlock_inode:
3572 inode_unlock(inode);
3573out:
3574 mnt_drop_write_file(filp);
3575
3576 if (ret >= 0) {
3577 ret = put_user(reserved_blocks, (u64 __user *)arg);
c2759eba
DJ		 3578 } else if (reserved_blocks &&
3579 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
c75488fb
CY		 3580 set_sbi_flag(sbi, SBI_NEED_FSCK);
3581 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
c2759eba 3582 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
c75488fb
CY		 3583 "run fsck to fix.",
3584 __func__, inode->i_ino, inode->i_blocks,
3585 reserved_blocks,
c2759eba 3586 atomic_read(&F2FS_I(inode)->i_compr_blocks));
c75488fb
CY		 3587 }
3588
3589 return ret;
3590}
3591
9af84648
DJ		 3592static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3593 pgoff_t off, block_t block, block_t len, u32 flags)
3594{
3595 struct request_queue *q = bdev_get_queue(bdev);
3596 sector_t sector = SECTOR_FROM_BLOCK(block);
3597 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3598 int ret = 0;
3599
3600 if (!q)
3601 return -ENXIO;
3602
3603 if (flags & F2FS_TRIM_FILE_DISCARD)
3604 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3605 blk_queue_secure_erase(q) ?
3606 BLKDEV_DISCARD_SECURE : 0);
3607
3608 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3609 if (IS_ENCRYPTED(inode))
3610 ret = fscrypt_zeroout_range(inode, off, block, len);
3611 else
3612 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3613 GFP_NOFS, 0);
3614 }
3615
3616 return ret;
3617}
3618
3619static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3620{
3621 struct inode *inode = file_inode(filp);
3622 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3623 struct address_space *mapping = inode->i_mapping;
3624 struct block_device *prev_bdev = NULL;
3625 struct f2fs_sectrim_range range;
3626 pgoff_t index, pg_end, prev_index = 0;
3627 block_t prev_block = 0, len = 0;
3628 loff_t end_addr;
3629 bool to_end = false;
3630 int ret = 0;
3631
3632 if (!(filp->f_mode & FMODE_WRITE))
3633 return -EBADF;
3634
3635 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3636 sizeof(range)))
3637 return -EFAULT;
3638
3639 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3640 !S_ISREG(inode->i_mode))
3641 return -EINVAL;
3642
3643 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3644 !f2fs_hw_support_discard(sbi)) ||
3645 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3646 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3647 return -EOPNOTSUPP;
3648
3649 file_start_write(filp);
3650 inode_lock(inode);
3651
3652 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3653 range.start >= inode->i_size) {
3654 ret = -EINVAL;
3655 goto err;
3656 }
3657
3658 if (range.len == 0)
3659 goto err;
3660
3661 if (inode->i_size - range.start > range.len) {
3662 end_addr = range.start + range.len;
3663 } else {
3664 end_addr = range.len == (u64)-1 ?
3665 sbi->sb->s_maxbytes : inode->i_size;
3666 to_end = true;
3667 }
3668
3669 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3670 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3671 ret = -EINVAL;
3672 goto err;
3673 }
3674
3675 index = F2FS_BYTES_TO_BLK(range.start);
3676 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3677
3678 ret = f2fs_convert_inline_inode(inode);
3679 if (ret)
3680 goto err;
3681
e4544b63 3682 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 3683 filemap_invalidate_lock(mapping);
9af84648
DJ		 3684
3685 ret = filemap_write_and_wait_range(mapping, range.start,
3686 to_end ? LLONG_MAX : end_addr - 1);
3687 if (ret)
3688 goto out;
3689
3690 truncate_inode_pages_range(mapping, range.start,
3691 to_end ? -1 : end_addr - 1);
3692
3693 while (index < pg_end) {
3694 struct dnode_of_data dn;
3695 pgoff_t end_offset, count;
3696 int i;
3697
3698 set_new_dnode(&dn, inode, NULL, NULL, 0);
3699 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3700 if (ret) {
3701 if (ret == -ENOENT) {
3702 index = f2fs_get_next_page_offset(&dn, index);
3703 continue;
3704 }
3705 goto out;
3706 }
3707
3708 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3709 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3710 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3711 struct block_device *cur_bdev;
3712 block_t blkaddr = f2fs_data_blkaddr(&dn);
3713
3714 if (!__is_valid_data_blkaddr(blkaddr))
3715 continue;
3716
3717 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3718 DATA_GENERIC_ENHANCE)) {
3719 ret = -EFSCORRUPTED;
3720 f2fs_put_dnode(&dn);
3721 goto out;
3722 }
3723
3724 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3725 if (f2fs_is_multi_device(sbi)) {
3726 int di = f2fs_target_device_index(sbi, blkaddr);
3727
3728 blkaddr -= FDEV(di).start_blk;
3729 }
3730
3731 if (len) {
3732 if (prev_bdev == cur_bdev &&
3733 index == prev_index + len &&
3734 blkaddr == prev_block + len) {
3735 len++;
3736 } else {
3737 ret = f2fs_secure_erase(prev_bdev,
3738 inode, prev_index, prev_block,
3739 len, range.flags);
3740 if (ret) {
3741 f2fs_put_dnode(&dn);
3742 goto out;
3743 }
3744
3745 len = 0;
3746 }
3747 }
3748
3749 if (!len) {
3750 prev_bdev = cur_bdev;
3751 prev_index = index;
3752 prev_block = blkaddr;
3753 len = 1;
3754 }
3755 }
3756
3757 f2fs_put_dnode(&dn);
3758
3759 if (fatal_signal_pending(current)) {
3760 ret = -EINTR;
3761 goto out;
3762 }
3763 cond_resched();
3764 }
3765
3766 if (len)
3767 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3768 prev_block, len, range.flags);
3769out:
edc6d01b 3770 filemap_invalidate_unlock(mapping);
e4544b63 3771 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
9af84648
DJ		 3772err:
3773 inode_unlock(inode);
3774 file_end_write(filp);
3775
3776 return ret;
3777}
3778
9e2a5f8c 3779static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
52656e6c 3780{
9e2a5f8c
DJ		 3781 struct inode *inode = file_inode(filp);
3782 struct f2fs_comp_option option;
3783
3784 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3785 return -EOPNOTSUPP;
3786
3787 inode_lock_shared(inode);
3788
3789 if (!f2fs_compressed_file(inode)) {
3790 inode_unlock_shared(inode);
3791 return -ENODATA;
3792 }
3793
3794 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3795 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3796
3797 inode_unlock_shared(inode);
3798
3799 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3800 sizeof(option)))
3801 return -EFAULT;
3802
3803 return 0;
3804}
3805
e1e8debe
DJ		 3806static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3807{
3808 struct inode *inode = file_inode(filp);
3809 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3810 struct f2fs_comp_option option;
3811 int ret = 0;
1f227a3e 3812
e1e8debe
DJ		 3813 if (!f2fs_sb_has_compression(sbi))
3814 return -EOPNOTSUPP;
3815
3816 if (!(filp->f_mode & FMODE_WRITE))
3817 return -EBADF;
3818
3819 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3820 sizeof(option)))
3821 return -EFAULT;
3822
3823 if (!f2fs_compressed_file(inode) ||
3824 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3825 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3826 option.algorithm >= COMPRESS_MAX)
3827 return -EINVAL;
3828
3829 file_start_write(filp);
3830 inode_lock(inode);
3831
3832 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3833 ret = -EBUSY;
3834 goto out;
3835 }
3836
3837 if (inode->i_size != 0) {
3838 ret = -EFBIG;
3839 goto out;
3840 }
3841
3842 F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3843 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3844 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3845 f2fs_mark_inode_dirty_sync(inode, true);
3846
3847 if (!f2fs_is_compress_backend_ready(inode))
3848 f2fs_warn(sbi, "compression algorithm is successfully set, "
3849 "but current kernel doesn't support this algorithm.");
3850out:
3851 inode_unlock(inode);
3852 file_end_write(filp);
3853
3854 return ret;
3855}
3856
5fdb322f
DJ		 3857static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3858{
fcd9ae4f 3859 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
5fdb322f
DJ		 3860 struct address_space *mapping = inode->i_mapping;
3861 struct page *page;
3862 pgoff_t redirty_idx = page_idx;
3863 int i, page_len = 0, ret = 0;
3864
3865 page_cache_ra_unbounded(&ractl, len, 0);
3866
3867 for (i = 0; i < len; i++, page_idx++) {
3868 page = read_cache_page(mapping, page_idx, NULL, NULL);
3869 if (IS_ERR(page)) {
3870 ret = PTR_ERR(page);
3871 break;
3872 }
3873 page_len++;
3874 }
3875
3876 for (i = 0; i < page_len; i++, redirty_idx++) {
3877 page = find_lock_page(mapping, redirty_idx);
df0736d7
DJ		 3878 if (!page) {
3879 ret = -ENOMEM;
3880 break;
3881 }
5fdb322f
DJ		 3882 set_page_dirty(page);
3883 f2fs_put_page(page, 1);
3884 f2fs_put_page(page, 0);
3885 }
3886
3887 return ret;
3888}
3889
3890static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3891{
3892 struct inode *inode = file_inode(filp);
3893 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3894 struct f2fs_inode_info *fi = F2FS_I(inode);
3895 pgoff_t page_idx = 0, last_idx;
3896 unsigned int blk_per_seg = sbi->blocks_per_seg;
3897 int cluster_size = F2FS_I(inode)->i_cluster_size;
3898 int count, ret;
3899
3900 if (!f2fs_sb_has_compression(sbi) ||
3901 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3902 return -EOPNOTSUPP;
3903
3904 if (!(filp->f_mode & FMODE_WRITE))
3905 return -EBADF;
3906
3907 if (!f2fs_compressed_file(inode))
3908 return -EINVAL;
3909
3910 f2fs_balance_fs(F2FS_I_SB(inode), true);
3911
3912 file_start_write(filp);
3913 inode_lock(inode);
3914
3915 if (!f2fs_is_compress_backend_ready(inode)) {
3916 ret = -EOPNOTSUPP;
3917 goto out;
3918 }
3919
3920 if (f2fs_is_mmap_file(inode)) {
3921 ret = -EBUSY;
3922 goto out;
3923 }
3924
3925 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3926 if (ret)
3927 goto out;
3928
3929 if (!atomic_read(&fi->i_compr_blocks))
3930 goto out;
3931
3932 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3933
3934 count = last_idx - page_idx;
3935 while (count) {
3936 int len = min(cluster_size, count);
3937
3938 ret = redirty_blocks(inode, page_idx, len);
3939 if (ret < 0)
3940 break;
3941
3942 if (get_dirty_pages(inode) >= blk_per_seg)
3943 filemap_fdatawrite(inode->i_mapping);
3944
3945 count -= len;
3946 page_idx += len;
3947 }
3948
3949 if (!ret)
3950 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
3951 LLONG_MAX);
3952
3953 if (ret)
833dcd35
JP		 3954 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
3955 __func__, ret);
5fdb322f
DJ		 3956out:
3957 inode_unlock(inode);
3958 file_end_write(filp);
3959
3960 return ret;
3961}
3962
3963static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
3964{
3965 struct inode *inode = file_inode(filp);
3966 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3967 pgoff_t page_idx = 0, last_idx;
3968 unsigned int blk_per_seg = sbi->blocks_per_seg;
3969 int cluster_size = F2FS_I(inode)->i_cluster_size;
3970 int count, ret;
3971
3972 if (!f2fs_sb_has_compression(sbi) ||
3973 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3974 return -EOPNOTSUPP;
3975
3976 if (!(filp->f_mode & FMODE_WRITE))
3977 return -EBADF;
3978
3979 if (!f2fs_compressed_file(inode))
3980 return -EINVAL;
3981
3982 f2fs_balance_fs(F2FS_I_SB(inode), true);
3983
3984 file_start_write(filp);
3985 inode_lock(inode);
3986
3987 if (!f2fs_is_compress_backend_ready(inode)) {
3988 ret = -EOPNOTSUPP;
3989 goto out;
3990 }
3991
3992 if (f2fs_is_mmap_file(inode)) {
3993 ret = -EBUSY;
3994 goto out;
3995 }
3996
3997 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3998 if (ret)
3999 goto out;
4000
4001 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4002
4003 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4004
4005 count = last_idx - page_idx;
4006 while (count) {
4007 int len = min(cluster_size, count);
4008
4009 ret = redirty_blocks(inode, page_idx, len);
4010 if (ret < 0)
4011 break;
4012
4013 if (get_dirty_pages(inode) >= blk_per_seg)
4014 filemap_fdatawrite(inode->i_mapping);
4015
4016 count -= len;
4017 page_idx += len;
4018 }
4019
4020 if (!ret)
4021 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4022 LLONG_MAX);
4023
4024 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4025
4026 if (ret)
833dcd35
JP		 4027 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4028 __func__, ret);
5fdb322f
DJ		 4029out:
4030 inode_unlock(inode);
4031 file_end_write(filp);
4032
4033 return ret;
4034}
4035
34178b1b 4036static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
52656e6c
JK		 4037{
4038 switch (cmd) {
3357af8f 4039 case FS_IOC_GETVERSION:
d49f3e89 4040 return f2fs_ioc_getversion(filp, arg);
88b88a66
JK		 4041 case F2FS_IOC_START_ATOMIC_WRITE:
4042 return f2fs_ioc_start_atomic_write(filp);
4043 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4044 return f2fs_ioc_commit_atomic_write(filp);
02a1335f 4045 case F2FS_IOC_START_VOLATILE_WRITE:
1e84371f 4046 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1e84371f 4047 case F2FS_IOC_ABORT_VOLATILE_WRITE:
7bc155fe 4048 return -EOPNOTSUPP;
1abff93d
JK		 4049 case F2FS_IOC_SHUTDOWN:
4050 return f2fs_ioc_shutdown(filp, arg);
52656e6c
JK		 4051 case FITRIM:
4052 return f2fs_ioc_fitrim(filp, arg);
3357af8f 4053 case FS_IOC_SET_ENCRYPTION_POLICY:
f424f664 4054 return f2fs_ioc_set_encryption_policy(filp, arg);
3357af8f 4055 case FS_IOC_GET_ENCRYPTION_POLICY:
f424f664 4056 return f2fs_ioc_get_encryption_policy(filp, arg);
3357af8f 4057 case FS_IOC_GET_ENCRYPTION_PWSALT:
f424f664 4058 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
8ce589c7
EB		 4059 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4060 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4061 case FS_IOC_ADD_ENCRYPTION_KEY:
4062 return f2fs_ioc_add_encryption_key(filp, arg);
4063 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4064 return f2fs_ioc_remove_encryption_key(filp, arg);
4065 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4066 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4067 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4068 return f2fs_ioc_get_encryption_key_status(filp, arg);
ee446e1a
EB		 4069 case FS_IOC_GET_ENCRYPTION_NONCE:
4070 return f2fs_ioc_get_encryption_nonce(filp, arg);
c1c1b583
CY		 4071 case F2FS_IOC_GARBAGE_COLLECT:
4072 return f2fs_ioc_gc(filp, arg);
34dc77ad
JK		 4073 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4074 return f2fs_ioc_gc_range(filp, arg);
456b88e4 4075 case F2FS_IOC_WRITE_CHECKPOINT:
059c0648 4076 return f2fs_ioc_write_checkpoint(filp, arg);
d323d005
CY		 4077 case F2FS_IOC_DEFRAGMENT:
4078 return f2fs_ioc_defragment(filp, arg);
4dd6f977
JK		 4079 case F2FS_IOC_MOVE_RANGE:
4080 return f2fs_ioc_move_range(filp, arg);
e066b83c
JK		 4081 case F2FS_IOC_FLUSH_DEVICE:
4082 return f2fs_ioc_flush_device(filp, arg);
e65ef207
JK		 4083 case F2FS_IOC_GET_FEATURES:
4084 return f2fs_ioc_get_features(filp, arg);
1ad71a27
JK		 4085 case F2FS_IOC_GET_PIN_FILE:
4086 return f2fs_ioc_get_pin_file(filp, arg);
4087 case F2FS_IOC_SET_PIN_FILE:
4088 return f2fs_ioc_set_pin_file(filp, arg);
c4020b2d
CY		 4089 case F2FS_IOC_PRECACHE_EXTENTS:
4090 return f2fs_ioc_precache_extents(filp, arg);
04f0b2ea
QS		 4091 case F2FS_IOC_RESIZE_FS:
4092 return f2fs_ioc_resize_fs(filp, arg);
95ae251f
EB		 4093 case FS_IOC_ENABLE_VERITY:
4094 return f2fs_ioc_enable_verity(filp, arg);
4095 case FS_IOC_MEASURE_VERITY:
4096 return f2fs_ioc_measure_verity(filp, arg);
e17fe657
EB		 4097 case FS_IOC_READ_VERITY_METADATA:
4098 return f2fs_ioc_read_verity_metadata(filp, arg);
3357af8f
EB		 4099 case FS_IOC_GETFSLABEL:
4100 return f2fs_ioc_getfslabel(filp, arg);
4101 case FS_IOC_SETFSLABEL:
4102 return f2fs_ioc_setfslabel(filp, arg);
439dfb10
CY		 4103 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4104 return f2fs_get_compress_blocks(filp, arg);
ef8d563f
CY		 4105 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4106 return f2fs_release_compress_blocks(filp, arg);
c75488fb
CY		 4107 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4108 return f2fs_reserve_compress_blocks(filp, arg);
9af84648
DJ		 4109 case F2FS_IOC_SEC_TRIM_FILE:
4110 return f2fs_sec_trim_file(filp, arg);
9e2a5f8c
DJ		 4111 case F2FS_IOC_GET_COMPRESS_OPTION:
4112 return f2fs_ioc_get_compress_option(filp, arg);
e1e8debe
DJ		 4113 case F2FS_IOC_SET_COMPRESS_OPTION:
4114 return f2fs_ioc_set_compress_option(filp, arg);
5fdb322f
DJ		 4115 case F2FS_IOC_DECOMPRESS_FILE:
4116 return f2fs_ioc_decompress_file(filp, arg);
4117 case F2FS_IOC_COMPRESS_FILE:
4118 return f2fs_ioc_compress_file(filp, arg);
fbfa2cc5
JK		 4119 default:
4120 return -ENOTTY;
4121 }
4122}
4123
34178b1b
CY		 4124long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4125{
4126 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4127 return -EIO;
4128 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4129 return -ENOSPC;
4130
4131 return __f2fs_ioctl(filp, cmd, arg);
4132}
4133
a1e09b03
EB		 4134/*
4135 * Return %true if the given read or write request should use direct I/O, or
4136 * %false if it should use buffered I/O.
4137 */
4138static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4139 struct iov_iter *iter)
4140{
4141 unsigned int align;
4142
4143 if (!(iocb->ki_flags & IOCB_DIRECT))
4144 return false;
4145
4146 if (f2fs_force_buffered_io(inode, iocb, iter))
4147 return false;
4148
4149 /*
4150 * Direct I/O not aligned to the disk's logical_block_size will be
4151 * attempted, but will fail with -EINVAL.
4152 *
4153 * f2fs additionally requires that direct I/O be aligned to the
4154 * filesystem block size, which is often a stricter requirement.
4155 * However, f2fs traditionally falls back to buffered I/O on requests
4156 * that are logical_block_size-aligned but not fs-block aligned.
4157 *
4158 * The below logic implements this behavior.
4159 */
4160 align = iocb->ki_pos | iov_iter_alignment(iter);
4161 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4162 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4163 return false;
4164
4165 return true;
4166}
4167
4168static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4169 unsigned int flags)
4170{
4171 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4172
4173 dec_page_count(sbi, F2FS_DIO_READ);
4174 if (error)
4175 return error;
4176 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4177 return 0;
4178}
4179
4180static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4181 .end_io = f2fs_dio_read_end_io,
4182};
4183
4184static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4c8ff709
CY		 4185{
4186 struct file *file = iocb->ki_filp;
4187 struct inode *inode = file_inode(file);
a1e09b03
EB		 4188 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4189 struct f2fs_inode_info *fi = F2FS_I(inode);
4190 const loff_t pos = iocb->ki_pos;
4191 const size_t count = iov_iter_count(to);
4192 struct iomap_dio *dio;
4193 ssize_t ret;
4194
4195 if (count == 0)
4196 return 0; /* skip atime update */
4197
bd984c03 4198 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
a1e09b03
EB		 4199
4200 if (iocb->ki_flags & IOCB_NOWAIT) {
e4544b63 4201 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
a1e09b03
EB		 4202 ret = -EAGAIN;
4203 goto out;
4204 }
4205 } else {
e4544b63 4206 f2fs_down_read(&fi->i_gc_rwsem[READ]);
a1e09b03
EB		 4207 }
4208
4209 /*
4210 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4211 * the higher-level function iomap_dio_rw() in order to ensure that the
4212 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4213 */
4214 inc_page_count(sbi, F2FS_DIO_READ);
4215 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4216 &f2fs_iomap_dio_read_ops, 0, 0);
4217 if (IS_ERR_OR_NULL(dio)) {
4218 ret = PTR_ERR_OR_ZERO(dio);
4219 if (ret != -EIOCBQUEUED)
4220 dec_page_count(sbi, F2FS_DIO_READ);
4221 } else {
4222 ret = iomap_dio_complete(dio);
4223 }
4224
e4544b63 4225 f2fs_up_read(&fi->i_gc_rwsem[READ]);
a1e09b03
EB		 4226
4227 file_accessed(file);
4228out:
4229 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4230 return ret;
4231}
4232
4233static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4234{
4235 struct inode *inode = file_inode(iocb->ki_filp);
c277f141 4236 const loff_t pos = iocb->ki_pos;
a1e09b03 4237 ssize_t ret;
4c8ff709
CY		 4238
4239 if (!f2fs_is_compress_backend_ready(inode))
4240 return -EOPNOTSUPP;
4241
c277f141
JK		 4242 if (trace_f2fs_dataread_start_enabled()) {
4243 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4244 char *path;
4245
4246 if (!p)
4247 goto skip_read_trace;
4248
4249 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4250 if (IS_ERR(path)) {
4251 kfree(p);
4252 goto skip_read_trace;
4253 }
8b83ac81 4254
c277f141
JK		 4255 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4256 current->pid, path, current->comm);
4257 kfree(p);
4258 }
4259skip_read_trace:
4260 if (f2fs_should_use_dio(inode, iocb, to)) {
4261 ret = f2fs_dio_read_iter(iocb, to);
4262 } else {
4263 ret = filemap_read(iocb, to, 0);
4264 if (ret > 0)
4265 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4266 }
4267 if (trace_f2fs_dataread_end_enabled())
4268 trace_f2fs_dataread_end(inode, pos, ret);
8b83ac81 4269 return ret;
4c8ff709
CY		 4270}
4271
a1e09b03
EB		 4272static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4273{
4274 struct file *file = iocb->ki_filp;
4275 struct inode *inode = file_inode(file);
4276 ssize_t count;
4277 int err;
4278
4279 if (IS_IMMUTABLE(inode))
4280 return -EPERM;
4281
4282 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4283 return -EPERM;
4284
4285 count = generic_write_checks(iocb, from);
4286 if (count <= 0)
4287 return count;
4288
4289 err = file_modified(file);
4290 if (err)
4291 return err;
4292 return count;
4293}
4294
3d697a4a
EB		 4295/*
4296 * Preallocate blocks for a write request, if it is possible and helpful to do
4297 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4298 * blocks were preallocated, or a negative errno value if something went
4299 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4300 * requested blocks (not just some of them) have been allocated.
4301 */
a1e09b03
EB		 4302static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4303 bool dio)
3d697a4a
EB		 4304{
4305 struct inode *inode = file_inode(iocb->ki_filp);
4306 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4307 const loff_t pos = iocb->ki_pos;
4308 const size_t count = iov_iter_count(iter);
4309 struct f2fs_map_blocks map = {};
3d697a4a
EB		 4310 int flag;
4311 int ret;
4312
4313 /* If it will be an out-of-place direct write, don't bother. */
4314 if (dio && f2fs_lfs_mode(sbi))
4315 return 0;
d4dd19ec
JK		 4316 /*
4317 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4318 * buffered IO, if DIO meets any holes.
4319 */
4320 if (dio && i_size_read(inode) &&
4321 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4322 return 0;
3d697a4a
EB		 4323
4324 /* No-wait I/O can't allocate blocks. */
4325 if (iocb->ki_flags & IOCB_NOWAIT)
4326 return 0;
4327
4328 /* If it will be a short write, don't bother. */
4329 if (fault_in_iov_iter_readable(iter, count))
4330 return 0;
4331
4332 if (f2fs_has_inline_data(inode)) {
4333 /* If the data will fit inline, don't bother. */
4334 if (pos + count <= MAX_INLINE_DATA(inode))
4335 return 0;
4336 ret = f2fs_convert_inline_inode(inode);
4337 if (ret)
4338 return ret;
4339 }
4340
4341 /* Do not preallocate blocks that will be written partially in 4KB. */
4342 map.m_lblk = F2FS_BLK_ALIGN(pos);
4343 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4344 if (map.m_len > map.m_lblk)
4345 map.m_len -= map.m_lblk;
4346 else
4347 map.m_len = 0;
4348 map.m_may_create = true;
4349 if (dio) {
4350 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4351 flag = F2FS_GET_BLOCK_PRE_DIO;
4352 } else {
4353 map.m_seg_type = NO_CHECK_TYPE;
4354 flag = F2FS_GET_BLOCK_PRE_AIO;
4355 }
4356
4357 ret = f2fs_map_blocks(inode, &map, 1, flag);
d4dd19ec
JK		 4358 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4359 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
3d697a4a
EB		 4360 return ret;
4361 if (ret == 0)
4362 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4363 return map.m_len;
4364}
4365
a1e09b03
EB		 4366static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4367 struct iov_iter *from)
fcc85a4d 4368{
b439b103
JK		 4369 struct file *file = iocb->ki_filp;
4370 struct inode *inode = file_inode(file);
a1e09b03
EB		 4371 ssize_t ret;
4372
4373 if (iocb->ki_flags & IOCB_NOWAIT)
4374 return -EOPNOTSUPP;
4375
4376 current->backing_dev_info = inode_to_bdi(inode);
800ba295 4377 ret = generic_perform_write(iocb, from);
a1e09b03
EB		 4378 current->backing_dev_info = NULL;
4379
4380 if (ret > 0) {
4381 iocb->ki_pos += ret;
4382 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4383 }
4384 return ret;
4385}
4386
4387static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4388 unsigned int flags)
4389{
4390 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4391
4392 dec_page_count(sbi, F2FS_DIO_WRITE);
4393 if (error)
4394 return error;
4395 f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4396 return 0;
4397}
4398
4399static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4400 .end_io = f2fs_dio_write_end_io,
4401};
4402
4403static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4404 bool *may_need_sync)
4405{
4406 struct file *file = iocb->ki_filp;
4407 struct inode *inode = file_inode(file);
4408 struct f2fs_inode_info *fi = F2FS_I(inode);
4409 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4410 const bool do_opu = f2fs_lfs_mode(sbi);
a1e09b03
EB		 4411 const loff_t pos = iocb->ki_pos;
4412 const ssize_t count = iov_iter_count(from);
a1e09b03
EB		 4413 unsigned int dio_flags;
4414 struct iomap_dio *dio;
4415 ssize_t ret;
4416
bd984c03 4417 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
a1e09b03
EB		 4418
4419 if (iocb->ki_flags & IOCB_NOWAIT) {
4420 /* f2fs_convert_inline_inode() and block allocation can block */
4421 if (f2fs_has_inline_data(inode) ||
4422 !f2fs_overwrite_io(inode, pos, count)) {
4423 ret = -EAGAIN;
4424 goto out;
4425 }
4426
e4544b63 4427 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
a1e09b03
EB		 4428 ret = -EAGAIN;
4429 goto out;
4430 }
e4544b63
TM		 4431 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4432 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03
EB		 4433 ret = -EAGAIN;
4434 goto out;
4435 }
4436 } else {
4437 ret = f2fs_convert_inline_inode(inode);
4438 if (ret)
4439 goto out;
4440
e4544b63 4441 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03 4442 if (do_opu)
e4544b63 4443 f2fs_down_read(&fi->i_gc_rwsem[READ]);
a1e09b03 4444 }
a1e09b03
EB		 4445
4446 /*
4447 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4448 * the higher-level function iomap_dio_rw() in order to ensure that the
4449 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4450 */
4451 inc_page_count(sbi, F2FS_DIO_WRITE);
4452 dio_flags = 0;
4453 if (pos + count > inode->i_size)
4454 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4455 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4456 &f2fs_iomap_dio_write_ops, dio_flags, 0);
4457 if (IS_ERR_OR_NULL(dio)) {
4458 ret = PTR_ERR_OR_ZERO(dio);
4459 if (ret == -ENOTBLK)
4460 ret = 0;
4461 if (ret != -EIOCBQUEUED)
4462 dec_page_count(sbi, F2FS_DIO_WRITE);
4463 } else {
4464 ret = iomap_dio_complete(dio);
4465 }
4466
a1e09b03 4467 if (do_opu)
e4544b63
TM		 4468 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4469 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03
EB		 4470
4471 if (ret < 0)
4472 goto out;
4473 if (pos + ret > inode->i_size)
4474 f2fs_i_size_write(inode, pos + ret);
4475 if (!do_opu)
4476 set_inode_flag(inode, FI_UPDATE_WRITE);
4477
4478 if (iov_iter_count(from)) {
4479 ssize_t ret2;
4480 loff_t bufio_start_pos = iocb->ki_pos;
4481
4482 /*
4483 * The direct write was partial, so we need to fall back to a
4484 * buffered write for the remainder.
4485 */
4486
4487 ret2 = f2fs_buffered_write_iter(iocb, from);
4488 if (iov_iter_count(from))
4489 f2fs_write_failed(inode, iocb->ki_pos);
4490 if (ret2 < 0)
4491 goto out;
4492
4493 /*
4494 * Ensure that the pagecache pages are written to disk and
4495 * invalidated to preserve the expected O_DIRECT semantics.
4496 */
4497 if (ret2 > 0) {
4498 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4499
4500 ret += ret2;
4501
4502 ret2 = filemap_write_and_wait_range(file->f_mapping,
4503 bufio_start_pos,
4504 bufio_end_pos);
4505 if (ret2 < 0)
4506 goto out;
4507 invalidate_mapping_pages(file->f_mapping,
4508 bufio_start_pos >> PAGE_SHIFT,
4509 bufio_end_pos >> PAGE_SHIFT);
4510 }
4511 } else {
4512 /* iomap_dio_rw() already handled the generic_write_sync(). */
4513 *may_need_sync = false;
4514 }
4515out:
4516 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4517 return ret;
4518}
4519
4520static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4521{
4522 struct inode *inode = file_inode(iocb->ki_filp);
ccf7cf92
EB		 4523 const loff_t orig_pos = iocb->ki_pos;
4524 const size_t orig_count = iov_iter_count(from);
3d697a4a 4525 loff_t target_size;
a1e09b03
EB		 4526 bool dio;
4527 bool may_need_sync = true;
3d697a4a 4528 int preallocated;
b439b103 4529 ssize_t ret;
fcc85a4d 4530
126ce721
CY		 4531 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4532 ret = -EIO;
4533 goto out;
4534 }
1f227a3e 4535
7bd29358
CY		 4536 if (!f2fs_is_compress_backend_ready(inode)) {
4537 ret = -EOPNOTSUPP;
4538 goto out;
4539 }
4c8ff709 4540
cb8434f1
GR		 4541 if (iocb->ki_flags & IOCB_NOWAIT) {
4542 if (!inode_trylock(inode)) {
126ce721
CY		 4543 ret = -EAGAIN;
4544 goto out;
4545 }
cb8434f1 4546 } else {
b91050a8
HL		 4547 inode_lock(inode);
4548 }
4549
a1e09b03 4550 ret = f2fs_write_checks(iocb, from);
b31bf0f9
EB		 4551 if (ret <= 0)
4552 goto out_unlock;
4553
a1e09b03
EB		 4554 /* Determine whether we will do a direct write or a buffered write. */
4555 dio = f2fs_should_use_dio(inode, iocb, from);
3d697a4a 4556
b31bf0f9
EB		 4557 /* Possibly preallocate the blocks for the write. */
4558 target_size = iocb->ki_pos + iov_iter_count(from);
a1e09b03 4559 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
c277f141 4560 if (preallocated < 0) {
b31bf0f9 4561 ret = preallocated;
c277f141
JK		 4562 } else {
4563 if (trace_f2fs_datawrite_start_enabled()) {
4564 char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4565 PATH_MAX, GFP_KERNEL);
4566 char *path;
4567
4568 if (!p)
4569 goto skip_write_trace;
4570 path = dentry_path_raw(file_dentry(iocb->ki_filp),
4571 p, PATH_MAX);
4572 if (IS_ERR(path)) {
4573 kfree(p);
4574 goto skip_write_trace;
4575 }
4576 trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4577 current->pid, path, current->comm);
4578 kfree(p);
4579 }
4580skip_write_trace:
a1e09b03
EB		 4581 /* Do the actual write. */
4582 ret = dio ?
4583 f2fs_dio_write_iter(iocb, from, &may_need_sync):
4584 f2fs_buffered_write_iter(iocb, from);
dc7a10dd 4585
c277f141
JK		 4586 if (trace_f2fs_datawrite_end_enabled())
4587 trace_f2fs_datawrite_end(inode, orig_pos, ret);
4588 }
4589
b31bf0f9 4590 /* Don't leave any preallocated blocks around past i_size. */
d4dd19ec 4591 if (preallocated && i_size_read(inode) < target_size) {
e4544b63 4592 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
b31bf0f9 4593 filemap_invalidate_lock(inode->i_mapping);
d4dd19ec
JK		 4594 if (!f2fs_truncate(inode))
4595 file_dont_truncate(inode);
b31bf0f9 4596 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 4597 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
d4dd19ec
JK		 4598 } else {
4599 file_dont_truncate(inode);
b439b103 4600 }
b31bf0f9
EB		 4601
4602 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
3d697a4a 4603out_unlock:
b439b103 4604 inode_unlock(inode);
126ce721 4605out:
ccf7cf92 4606 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
a1e09b03 4607 if (ret > 0 && may_need_sync)
e2592217 4608 ret = generic_write_sync(iocb, ret);
b439b103 4609 return ret;
fcc85a4d
JK		 4610}
4611
0f6b56ec
DJ		 4612static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4613 int advice)
4614{
0f6b56ec
DJ		 4615 struct address_space *mapping;
4616 struct backing_dev_info *bdi;
e64347ae
FC		 4617 struct inode *inode = file_inode(filp);
4618 int err;
0f6b56ec
DJ		 4619
4620 if (advice == POSIX_FADV_SEQUENTIAL) {
0f6b56ec
DJ		 4621 if (S_ISFIFO(inode->i_mode))
4622 return -ESPIPE;
4623
4624 mapping = filp->f_mapping;
4625 if (!mapping || len < 0)
4626 return -EINVAL;
4627
4628 bdi = inode_to_bdi(mapping->host);
4629 filp->f_ra.ra_pages = bdi->ra_pages *
4630 F2FS_I_SB(inode)->seq_file_ra_mul;
4631 spin_lock(&filp->f_lock);
4632 filp->f_mode &= ~FMODE_RANDOM;
4633 spin_unlock(&filp->f_lock);
4634 return 0;
4635 }
4636
e64347ae
FC		 4637 err = generic_fadvise(filp, offset, len, advice);
4638 if (!err && advice == POSIX_FADV_DONTNEED &&
4639 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4640 f2fs_compressed_file(inode))
4641 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4642
4643 return err;
0f6b56ec
DJ		 4644}
4645
e9750824 4646#ifdef CONFIG_COMPAT
34178b1b
CY		 4647struct compat_f2fs_gc_range {
4648 u32 sync;
4649 compat_u64 start;
4650 compat_u64 len;
4651};
4652#define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4653 struct compat_f2fs_gc_range)
4654
4655static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4656{
4657 struct compat_f2fs_gc_range __user *urange;
4658 struct f2fs_gc_range range;
4659 int err;
4660
4661 urange = compat_ptr(arg);
4662 err = get_user(range.sync, &urange->sync);
4663 err |= get_user(range.start, &urange->start);
4664 err |= get_user(range.len, &urange->len);
4665 if (err)
4666 return -EFAULT;
4667
4668 return __f2fs_ioc_gc_range(file, &range);
4669}
4670
4671struct compat_f2fs_move_range {
4672 u32 dst_fd;
4673 compat_u64 pos_in;
4674 compat_u64 pos_out;
4675 compat_u64 len;
4676};
4677#define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4678 struct compat_f2fs_move_range)
4679
4680static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4681{
4682 struct compat_f2fs_move_range __user *urange;
4683 struct f2fs_move_range range;
4684 int err;
4685
4686 urange = compat_ptr(arg);
4687 err = get_user(range.dst_fd, &urange->dst_fd);
4688 err |= get_user(range.pos_in, &urange->pos_in);
4689 err |= get_user(range.pos_out, &urange->pos_out);
4690 err |= get_user(range.len, &urange->len);
4691 if (err)
4692 return -EFAULT;
4693
4694 return __f2fs_ioc_move_range(file, &range);
4695}
4696
e9750824
NJ		 4697long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4698{
34178b1b
CY		 4699 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4700 return -EIO;
4701 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4702 return -ENOSPC;
4703
e9750824 4704 switch (cmd) {
3357af8f
EB		 4705 case FS_IOC32_GETVERSION:
4706 cmd = FS_IOC_GETVERSION;
04ef4b62 4707 break;
34178b1b
CY		 4708 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4709 return f2fs_compat_ioc_gc_range(file, arg);
4710 case F2FS_IOC32_MOVE_RANGE:
4711 return f2fs_compat_ioc_move_range(file, arg);
04ef4b62
CY		 4712 case F2FS_IOC_START_ATOMIC_WRITE:
4713 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4714 case F2FS_IOC_START_VOLATILE_WRITE:
4715 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4716 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4717 case F2FS_IOC_SHUTDOWN:
314999dc 4718 case FITRIM:
3357af8f
EB		 4719 case FS_IOC_SET_ENCRYPTION_POLICY:
4720 case FS_IOC_GET_ENCRYPTION_PWSALT:
4721 case FS_IOC_GET_ENCRYPTION_POLICY:
8ce589c7
EB		 4722 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4723 case FS_IOC_ADD_ENCRYPTION_KEY:
4724 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4725 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4726 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
ee446e1a 4727 case FS_IOC_GET_ENCRYPTION_NONCE:
04ef4b62
CY		 4728 case F2FS_IOC_GARBAGE_COLLECT:
4729 case F2FS_IOC_WRITE_CHECKPOINT:
4730 case F2FS_IOC_DEFRAGMENT:
e066b83c 4731 case F2FS_IOC_FLUSH_DEVICE:
e65ef207 4732 case F2FS_IOC_GET_FEATURES:
1ad71a27
JK		 4733 case F2FS_IOC_GET_PIN_FILE:
4734 case F2FS_IOC_SET_PIN_FILE:
c4020b2d 4735 case F2FS_IOC_PRECACHE_EXTENTS:
04f0b2ea 4736 case F2FS_IOC_RESIZE_FS:
95ae251f
EB		 4737 case FS_IOC_ENABLE_VERITY:
4738 case FS_IOC_MEASURE_VERITY:
e17fe657 4739 case FS_IOC_READ_VERITY_METADATA:
3357af8f
EB		 4740 case FS_IOC_GETFSLABEL:
4741 case FS_IOC_SETFSLABEL:
439dfb10 4742 case F2FS_IOC_GET_COMPRESS_BLOCKS:
ef8d563f 4743 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
c75488fb 4744 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
9af84648 4745 case F2FS_IOC_SEC_TRIM_FILE:
9e2a5f8c 4746 case F2FS_IOC_GET_COMPRESS_OPTION:
e1e8debe 4747 case F2FS_IOC_SET_COMPRESS_OPTION:
5fdb322f
DJ		 4748 case F2FS_IOC_DECOMPRESS_FILE:
4749 case F2FS_IOC_COMPRESS_FILE:
4dd6f977 4750 break;
e9750824
NJ		 4751 default:
4752 return -ENOIOCTLCMD;
4753 }
34178b1b 4754 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
e9750824
NJ		 4755}
4756#endif
4757
fbfa2cc5 4758const struct file_operations f2fs_file_operations = {
267378d4 4759 .llseek = f2fs_llseek,
4c8ff709 4760 .read_iter = f2fs_file_read_iter,
fcc85a4d
JK		 4761 .write_iter = f2fs_file_write_iter,
4762 .open = f2fs_file_open,
12662234 4763 .release = f2fs_release_file,
fbfa2cc5 4764 .mmap = f2fs_file_mmap,
7a10f017 4765 .flush = f2fs_file_flush,
fbfa2cc5
JK		 4766 .fsync = f2fs_sync_file,
4767 .fallocate = f2fs_fallocate,
4768 .unlocked_ioctl = f2fs_ioctl,
e9750824
NJ		 4769#ifdef CONFIG_COMPAT
4770 .compat_ioctl = f2fs_compat_ioctl,
4771#endif
fbfa2cc5 4772 .splice_read = generic_file_splice_read,
8d020765 4773 .splice_write = iter_file_splice_write,
0f6b56ec 4774 .fadvise = f2fs_file_fadvise,
fbfa2cc5 4775};
Linux 6.x block layer and io_uring tree(s)