projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: replace F2FS_I(inode) and sbi by the local variable
[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 };
d147ea4a
JK		 1650 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1651 .init_gc_type = FG_GC,
1652 .should_migrate_blocks = false,
c81d5bae
JK		 1653 .err_gc_skipped = true,
1654 .nr_free_secs = 0 };
88f2cfc5 1655 pgoff_t pg_start, pg_end;
fbfa2cc5 1656 loff_t new_size = i_size_read(inode);
e12dd7bd 1657 loff_t off_end;
88f2cfc5 1658 block_t expanded = 0;
a7de6086 1659 int err;
fbfa2cc5 1660
a7de6086
JK		 1661 err = inode_newsize_ok(inode, (len + offset));
1662 if (err)
1663 return err;
fbfa2cc5 1664
a7de6086
JK		 1665 err = f2fs_convert_inline_inode(inode);
1666 if (err)
1667 return err;
9e09fc85 1668
2c4db1a6 1669 f2fs_balance_fs(sbi, true);
2a340760 1670
88f2cfc5 1671 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
e12dd7bd 1672 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
09cbfeaf 1673 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5 1674
88f2cfc5
CY		 1675 map.m_lblk = pg_start;
1676 map.m_len = pg_end - pg_start;
e12dd7bd
JK		 1677 if (off_end)
1678 map.m_len++;
ead43275 1679
f5a53edc
JK		 1680 if (!map.m_len)
1681 return 0;
1682
1683 if (f2fs_is_pinned_file(inode)) {
e1175f02
CY		 1684 block_t sec_blks = BLKS_PER_SEC(sbi);
1685 block_t sec_len = roundup(map.m_len, sec_blks);
f5a53edc 1686
e1175f02 1687 map.m_len = sec_blks;
f5a53edc
JK		 1688next_alloc:
1689 if (has_not_enough_free_secs(sbi, 0,
1690 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
e4544b63 1691 f2fs_down_write(&sbi->gc_lock);
d147ea4a 1692 err = f2fs_gc(sbi, &gc_control);
2e42b7f8 1693 if (err && err != -ENODATA)
f5a53edc
JK		 1694 goto out_err;
1695 }
1696
e4544b63 1697 f2fs_down_write(&sbi->pin_sem);
fd612648
DJ		 1698
1699 f2fs_lock_op(sbi);
509f1010 1700 f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
fd612648
DJ		 1701 f2fs_unlock_op(sbi);
1702
d0b9e42a 1703 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
f5a53edc 1704 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
d4dd19ec 1705 file_dont_truncate(inode);
d0b9e42a 1706
e4544b63 1707 f2fs_up_write(&sbi->pin_sem);
cad3836f 1708
88f2cfc5 1709 expanded += map.m_len;
e1175f02 1710 sec_len -= map.m_len;
f5a53edc 1711 map.m_lblk += map.m_len;
e1175f02 1712 if (!err && sec_len)
f5a53edc
JK		 1713 goto next_alloc;
1714
88f2cfc5 1715 map.m_len = expanded;
f5a53edc
JK		 1716 } else {
1717 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
88f2cfc5 1718 expanded = map.m_len;
f5a53edc
JK		 1719 }
1720out_err:
a7de6086 1721 if (err) {
e12dd7bd 1722 pgoff_t last_off;
fbfa2cc5 1723
88f2cfc5 1724 if (!expanded)
a7de6086 1725 return err;
98397ff3 1726
88f2cfc5 1727 last_off = pg_start + expanded - 1;
e12dd7bd
JK		 1728
1729 /* update new size to the failed position */
1061fd48 1730 new_size = (last_off == pg_end) ? offset + len :
e12dd7bd
JK		 1731 (loff_t)(last_off + 1) << PAGE_SHIFT;
1732 } else {
1733 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
fbfa2cc5
JK		 1734 }
1735
e8ed90a6
CY		 1736 if (new_size > i_size_read(inode)) {
1737 if (mode & FALLOC_FL_KEEP_SIZE)
1738 file_set_keep_isize(inode);
1739 else
1740 f2fs_i_size_write(inode, new_size);
1741 }
fbfa2cc5 1742
a7de6086 1743 return err;
fbfa2cc5
JK		 1744}
1745
1746static long f2fs_fallocate(struct file *file, int mode,
1747 loff_t offset, loff_t len)
1748{
6131ffaa 1749 struct inode *inode = file_inode(file);
587c0a42 1750 long ret = 0;
fbfa2cc5 1751
1f227a3e
JK		 1752 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1753 return -EIO;
00e09c0b
CY		 1754 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1755 return -ENOSPC;
4c8ff709
CY		 1756 if (!f2fs_is_compress_backend_ready(inode))
1757 return -EOPNOTSUPP;
1f227a3e 1758
c998012b
CY		 1759 /* f2fs only support ->fallocate for regular file */
1760 if (!S_ISREG(inode->i_mode))
1761 return -EINVAL;
1762
62230e0d 1763 if (IS_ENCRYPTED(inode) &&
f62185d0 1764 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
fcc85a4d
JK		 1765 return -EOPNOTSUPP;
1766
5fed0be8
JK		 1767 /*
1768 * Pinned file should not support partial trucation since the block
1769 * can be used by applications.
1770 */
1771 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
4c8ff709
CY		 1772 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1773 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1774 return -EOPNOTSUPP;
1775
b4ace337 1776 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
f62185d0
CY		 1777 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1778 FALLOC_FL_INSERT_RANGE))
fbfa2cc5
JK		 1779 return -EOPNOTSUPP;
1780
5955102c 1781 inode_lock(inode);
3375f696 1782
958ed929
CY		 1783 ret = file_modified(file);
1784 if (ret)
1785 goto out;
1786
587c0a42
TY		 1787 if (mode & FALLOC_FL_PUNCH_HOLE) {
1788 if (offset >= inode->i_size)
1789 goto out;
1790
a66c7b2f 1791 ret = punch_hole(inode, offset, len);
b4ace337
CY		 1792 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1793 ret = f2fs_collapse_range(inode, offset, len);
75cd4e09
CY		 1794 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1795 ret = f2fs_zero_range(inode, offset, len, mode);
f62185d0
CY		 1796 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1797 ret = f2fs_insert_range(inode, offset, len);
b4ace337 1798 } else {
fbfa2cc5 1799 ret = expand_inode_data(inode, offset, len, mode);
b4ace337 1800 }
fbfa2cc5 1801
3af60a49 1802 if (!ret) {
078cd827 1803 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 1804 f2fs_mark_inode_dirty_sync(inode, false);
d0239e1b 1805 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3af60a49 1806 }
3375f696 1807
587c0a42 1808out:
5955102c 1809 inode_unlock(inode);
3375f696 1810
c01e2853 1811 trace_f2fs_fallocate(inode, mode, offset, len, ret);
fbfa2cc5
JK		 1812 return ret;
1813}
1814
1e84371f
JK		 1815static int f2fs_release_file(struct inode *inode, struct file *filp)
1816{
de5307e4
JK		 1817 /*
1818 * f2fs_relase_file is called at every close calls. So we should
1819 * not drop any inmemory pages by close called by other process.
1820 */
1821 if (!(filp->f_mode & FMODE_WRITE) ||
1822 atomic_read(&inode->i_writecount) != 1)
1823 return 0;
1824
1e84371f 1825 if (f2fs_is_atomic_file(inode))
3db1de0e 1826 f2fs_abort_atomic_write(inode, true);
1e84371f
JK		 1827 return 0;
1828}
1829
7a10f017 1830static int f2fs_file_flush(struct file *file, fl_owner_t id)
fbfa2cc5 1831{
7a10f017
JK		 1832 struct inode *inode = file_inode(file);
1833
1834 /*
1835 * If the process doing a transaction is crashed, we should do
1836 * roll-back. Otherwise, other reader/write can see corrupted database
1837 * until all the writers close its file. Since this should be done
1838 * before dropping file lock, it needs to do in ->flush.
1839 */
1840 if (f2fs_is_atomic_file(inode) &&
3db1de0e
DJ		 1841 F2FS_I(inode)->atomic_write_task == current)
1842 f2fs_abort_atomic_write(inode, true);
7a10f017 1843 return 0;
fbfa2cc5
JK		 1844}
1845
36098557 1846static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
2c1d0305
CY		 1847{
1848 struct f2fs_inode_info *fi = F2FS_I(inode);
99eabb91
JK		 1849 u32 masked_flags = fi->i_flags & mask;
1850
a7531039
JK		 1851 /* mask can be shrunk by flags_valid selector */
1852 iflags &= mask;
2c1d0305
CY		 1853
1854 /* Is it quota file? Do not allow user to mess with it */
1855 if (IS_NOQUOTA(inode))
1856 return -EPERM;
1857
99eabb91 1858 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
2c2eb7a3
DR		 1859 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1860 return -EOPNOTSUPP;
1861 if (!f2fs_empty_dir(inode))
1862 return -ENOTEMPTY;
1863 }
1864
4c8ff709
CY		 1865 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1866 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1867 return -EOPNOTSUPP;
1868 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1869 return -EINVAL;
1870 }
1871
99eabb91 1872 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
aa576970 1873 if (masked_flags & F2FS_COMPR_FL) {
78134d03 1874 if (!f2fs_disable_compressed_file(inode))
2536ac68
CY		 1875 return -EINVAL;
1876 }
4c8ff709
CY		 1877 if (iflags & F2FS_NOCOMP_FL)
1878 return -EINVAL;
1879 if (iflags & F2FS_COMPR_FL) {
4c8ff709
CY		 1880 if (!f2fs_may_compress(inode))
1881 return -EINVAL;
519a5a2f
CY		 1882 if (S_ISREG(inode->i_mode) && inode->i_size)
1883 return -EINVAL;
4c8ff709
CY		 1884
1885 set_compress_context(inode);
1886 }
1887 }
99eabb91
JK		 1888 if ((iflags ^ masked_flags) & F2FS_NOCOMP_FL) {
1889 if (masked_flags & F2FS_COMPR_FL)
4c8ff709
CY		 1890 return -EINVAL;
1891 }
1892
d5e5efa2 1893 fi->i_flags = iflags | (fi->i_flags & ~mask);
4c8ff709
CY		 1894 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1895 (fi->i_flags & F2FS_NOCOMP_FL));
2c1d0305 1896
59c84408 1897 if (fi->i_flags & F2FS_PROJINHERIT_FL)
2c1d0305
CY		 1898 set_inode_flag(inode, FI_PROJ_INHERIT);
1899 else
1900 clear_inode_flag(inode, FI_PROJ_INHERIT);
1901
1902 inode->i_ctime = current_time(inode);
1903 f2fs_set_inode_flags(inode);
b32e0190 1904 f2fs_mark_inode_dirty_sync(inode, true);
2c1d0305
CY		 1905 return 0;
1906}
1907
9b1bb01c 1908/* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
36098557
EB		 1909
1910/*
1911 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1912 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1913 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
1914 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
9b1bb01c
MS		 1915 *
1916 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1917 * FS_IOC_FSSETXATTR is done by the VFS.
36098557
EB		 1918 */
1919
1920static const struct {
1921 u32 iflag;
1922 u32 fsflag;
1923} f2fs_fsflags_map[] = {
4c8ff709 1924 { F2FS_COMPR_FL, FS_COMPR_FL },
36098557
EB		 1925 { F2FS_SYNC_FL, FS_SYNC_FL },
1926 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
1927 { F2FS_APPEND_FL, FS_APPEND_FL },
1928 { F2FS_NODUMP_FL, FS_NODUMP_FL },
1929 { F2FS_NOATIME_FL, FS_NOATIME_FL },
4c8ff709 1930 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
36098557
EB		 1931 { F2FS_INDEX_FL, FS_INDEX_FL },
1932 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
1933 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
2c2eb7a3 1934 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
36098557
EB		 1935};
1936
1937#define F2FS_GETTABLE_FS_FL ( \
4c8ff709 1938 FS_COMPR_FL | \
36098557
EB		 1939 FS_SYNC_FL | \
1940 FS_IMMUTABLE_FL | \
1941 FS_APPEND_FL | \
1942 FS_NODUMP_FL | \
1943 FS_NOATIME_FL | \
4c8ff709 1944 FS_NOCOMP_FL | \
36098557
EB		 1945 FS_INDEX_FL | \
1946 FS_DIRSYNC_FL | \
1947 FS_PROJINHERIT_FL | \
1948 FS_ENCRYPT_FL | \
1949 FS_INLINE_DATA_FL | \
95ae251f 1950 FS_NOCOW_FL | \
fbc246a1 1951 FS_VERITY_FL | \
2c2eb7a3 1952 FS_CASEFOLD_FL)
36098557
EB		 1953
1954#define F2FS_SETTABLE_FS_FL ( \
4c8ff709 1955 FS_COMPR_FL | \
36098557
EB		 1956 FS_SYNC_FL | \
1957 FS_IMMUTABLE_FL | \
1958 FS_APPEND_FL | \
1959 FS_NODUMP_FL | \
1960 FS_NOATIME_FL | \
4c8ff709 1961 FS_NOCOMP_FL | \
36098557 1962 FS_DIRSYNC_FL | \
2c2eb7a3
DR		 1963 FS_PROJINHERIT_FL | \
1964 FS_CASEFOLD_FL)
36098557
EB		 1965
1966/* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1967static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1968{
1969 u32 fsflags = 0;
1970 int i;
1971
1972 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1973 if (iflags & f2fs_fsflags_map[i].iflag)
1974 fsflags |= f2fs_fsflags_map[i].fsflag;
1975
1976 return fsflags;
1977}
1978
1979/* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1980static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1981{
1982 u32 iflags = 0;
1983 int i;
1984
1985 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1986 if (fsflags & f2fs_fsflags_map[i].fsflag)
1987 iflags |= f2fs_fsflags_map[i].iflag;
1988
1989 return iflags;
1990}
1991
d49f3e89
CY		 1992static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1993{
1994 struct inode *inode = file_inode(filp);
1995
1996 return put_user(inode->i_generation, (int __user *)arg);
1997}
1998
88b88a66
JK		 1999static int f2fs_ioc_start_atomic_write(struct file *filp)
2000{
2001 struct inode *inode = file_inode(filp);
984fc4e7 2002 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
743b620c
JK		 2003 struct f2fs_inode_info *fi = F2FS_I(inode);
2004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3db1de0e 2005 struct inode *pinode;
f4c9c743 2006 int ret;
88b88a66 2007
984fc4e7 2008 if (!inode_owner_or_capable(mnt_userns, inode))
88b88a66
JK		 2009 return -EACCES;
2010
e811898c
JK		 2011 if (!S_ISREG(inode->i_mode))
2012 return -EINVAL;
2013
038d0698
CY		 2014 if (filp->f_flags & O_DIRECT)
2015 return -EINVAL;
2016
7fb17fe4
CY		 2017 ret = mnt_want_write_file(filp);
2018 if (ret)
2019 return ret;
2020
0fac558b
CY		 2021 inode_lock(inode);
2022
9b56adcf
FC		 2023 if (!f2fs_disable_compressed_file(inode)) {
2024 ret = -EINVAL;
2025 goto out;
2026 }
4c8ff709 2027
3db1de0e 2028 if (f2fs_is_atomic_file(inode))
7fb17fe4 2029 goto out;
88b88a66 2030
f4c9c743
CY		 2031 ret = f2fs_convert_inline_inode(inode);
2032 if (ret)
7fb17fe4 2033 goto out;
88b88a66 2034
054cb289 2035 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
6f8d4455 2036
31867b23
JK		 2037 /*
2038 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2039 * f2fs_is_atomic_file.
2040 */
2041 if (get_dirty_pages(inode))
054cb289 2042 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
dcbb4c10 2043 inode->i_ino, get_dirty_pages(inode));
c27753d6 2044 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
6f8d4455 2045 if (ret) {
054cb289 2046 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
684ca7e5 2047 goto out;
6f8d4455 2048 }
31867b23 2049
3db1de0e
DJ		 2050 /* Create a COW inode for atomic write */
2051 pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2052 if (IS_ERR(pinode)) {
054cb289 2053 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3db1de0e
DJ		 2054 ret = PTR_ERR(pinode);
2055 goto out;
2056 }
2057
2058 ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2059 iput(pinode);
2060 if (ret) {
054cb289 2061 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3db1de0e
DJ		 2062 goto out;
2063 }
2064 f2fs_i_size_write(fi->cow_inode, i_size_read(inode));
2065
743b620c 2066 spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
677017d1 2067 sbi->atomic_files++;
743b620c
JK		 2068 spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2069
054afda9 2070 set_inode_flag(inode, FI_ATOMIC_FILE);
3db1de0e
DJ		 2071 set_inode_flag(fi->cow_inode, FI_ATOMIC_FILE);
2072 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
054cb289 2073 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
684ca7e5 2074
054cb289
YY		 2075 f2fs_update_time(sbi, REQ_TIME);
2076 fi->atomic_write_task = current;
26a28a0c 2077 stat_update_max_atomic_write(inode);
684ca7e5 2078out:
0fac558b 2079 inode_unlock(inode);
7fb17fe4 2080 mnt_drop_write_file(filp);
c27753d6 2081 return ret;
88b88a66
JK		 2082}
2083
2084static int f2fs_ioc_commit_atomic_write(struct file *filp)
2085{
2086 struct inode *inode = file_inode(filp);
984fc4e7 2087 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
88b88a66
JK		 2088 int ret;
2089
984fc4e7 2090 if (!inode_owner_or_capable(mnt_userns, inode))
88b88a66
JK		 2091 return -EACCES;
2092
2093 ret = mnt_want_write_file(filp);
2094 if (ret)
2095 return ret;
2096
6f8d4455 2097 f2fs_balance_fs(F2FS_I_SB(inode), true);
0fac558b 2098
6f8d4455 2099 inode_lock(inode);
1dc0f899 2100
6282adbf 2101 if (f2fs_is_atomic_file(inode)) {
3db1de0e 2102 ret = f2fs_commit_atomic_write(inode);
5fe45743 2103 if (ret)
7bc155fe 2104 goto unlock_out;
5fe45743 2105
26a28a0c 2106 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
743b620c 2107 if (!ret)
3db1de0e 2108 f2fs_abort_atomic_write(inode, false);
26a28a0c 2109 } else {
774e1b78 2110 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
6282adbf 2111 }
7bc155fe 2112unlock_out:
0fac558b 2113 inode_unlock(inode);
88b88a66
JK		 2114 mnt_drop_write_file(filp);
2115 return ret;
2116}
2117
1abff93d
JK		 2118static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2119{
2120 struct inode *inode = file_inode(filp);
2121 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2122 struct super_block *sb = sbi->sb;
2123 __u32 in;
2a96d8ad 2124 int ret = 0;
1abff93d
JK		 2125
2126 if (!capable(CAP_SYS_ADMIN))
2127 return -EPERM;
2128
2129 if (get_user(in, (__u32 __user *)arg))
2130 return -EFAULT;
2131
60b2b4ee
ST		 2132 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2133 ret = mnt_want_write_file(filp);
8626441f
CY		 2134 if (ret) {
2135 if (ret == -EROFS) {
2136 ret = 0;
2137 f2fs_stop_checkpoint(sbi, false);
2138 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2139 trace_f2fs_shutdown(sbi, in, ret);
2140 }
60b2b4ee 2141 return ret;
8626441f 2142 }
60b2b4ee 2143 }
7fb17fe4 2144
1abff93d
JK		 2145 switch (in) {
2146 case F2FS_GOING_DOWN_FULLSYNC:
040f04bd
CH		 2147 ret = freeze_bdev(sb->s_bdev);
2148 if (ret)
d027c484 2149 goto out;
040f04bd
CH		 2150 f2fs_stop_checkpoint(sbi, false);
2151 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2152 thaw_bdev(sb->s_bdev);
1abff93d
JK		 2153 break;
2154 case F2FS_GOING_DOWN_METASYNC:
2155 /* do checkpoint only */
d027c484
CY		 2156 ret = f2fs_sync_fs(sb, 1);
2157 if (ret)
2158 goto out;
38f91ca8 2159 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 2160 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1abff93d
JK		 2161 break;
2162 case F2FS_GOING_DOWN_NOSYNC:
38f91ca8 2163 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 2164 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1abff93d 2165 break;
c912a829 2166 case F2FS_GOING_DOWN_METAFLUSH:
4d57b86d 2167 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
38f91ca8 2168 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 2169 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
c912a829 2170 break;
0cd6d9b0
JK		 2171 case F2FS_GOING_DOWN_NEED_FSCK:
2172 set_sbi_flag(sbi, SBI_NEED_FSCK);
db610a64
JK		 2173 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2174 set_sbi_flag(sbi, SBI_IS_DIRTY);
0cd6d9b0
JK		 2175 /* do checkpoint only */
2176 ret = f2fs_sync_fs(sb, 1);
db610a64 2177 goto out;
1abff93d 2178 default:
7fb17fe4
CY		 2179 ret = -EINVAL;
2180 goto out;
1abff93d 2181 }
7950e9ac 2182
4d57b86d
CY		 2183 f2fs_stop_gc_thread(sbi);
2184 f2fs_stop_discard_thread(sbi);
7950e9ac 2185
4d57b86d 2186 f2fs_drop_discard_cmd(sbi);
7950e9ac
CY		 2187 clear_opt(sbi, DISCARD);
2188
d0239e1b 2189 f2fs_update_time(sbi, REQ_TIME);
7fb17fe4 2190out:
60b2b4ee
ST		 2191 if (in != F2FS_GOING_DOWN_FULLSYNC)
2192 mnt_drop_write_file(filp);
559e87c4
CY		 2193
2194 trace_f2fs_shutdown(sbi, in, ret);
2195
7fb17fe4 2196 return ret;
1abff93d
JK		 2197}
2198
52656e6c
JK		 2199static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2200{
2201 struct inode *inode = file_inode(filp);
2202 struct super_block *sb = inode->i_sb;
2203 struct request_queue *q = bdev_get_queue(sb->s_bdev);
2204 struct fstrim_range range;
2205 int ret;
4b2fecc8 2206
52656e6c
JK		 2207 if (!capable(CAP_SYS_ADMIN))
2208 return -EPERM;
4b2fecc8 2209
7d20c8ab 2210 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
52656e6c 2211 return -EOPNOTSUPP;
4b2fecc8 2212
52656e6c
JK		 2213 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2214 sizeof(range)))
2215 return -EFAULT;
4b2fecc8 2216
7fb17fe4
CY		 2217 ret = mnt_want_write_file(filp);
2218 if (ret)
2219 return ret;
2220
52656e6c
JK		 2221 range.minlen = max((unsigned int)range.minlen,
2222 q->limits.discard_granularity);
2223 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
7fb17fe4 2224 mnt_drop_write_file(filp);
52656e6c
JK		 2225 if (ret < 0)
2226 return ret;
4b2fecc8 2227
52656e6c
JK		 2228 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2229 sizeof(range)))
2230 return -EFAULT;
d0239e1b 2231 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
52656e6c
JK		 2232 return 0;
2233}
2234
f424f664
JK		 2235static bool uuid_is_nonzero(__u8 u[16])
2236{
2237 int i;
2238
2239 for (i = 0; i < 16; i++)
2240 if (u[i])
2241 return true;
2242 return false;
2243}
2244
2245static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2246{
f424f664
JK		 2247 struct inode *inode = file_inode(filp);
2248
7beb01f7 2249 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
ead710b7
CY		 2250 return -EOPNOTSUPP;
2251
d0239e1b 2252 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 2253
db717d8e 2254 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
f424f664
JK		 2255}
2256
2257static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2258{
7beb01f7 2259 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
ead710b7 2260 return -EOPNOTSUPP;
db717d8e 2261 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
f424f664
JK		 2262}
2263
2264static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2265{
2266 struct inode *inode = file_inode(filp);
2267 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2268 int err;
2269
7beb01f7 2270 if (!f2fs_sb_has_encrypt(sbi))
f424f664
JK		 2271 return -EOPNOTSUPP;
2272
f424f664
JK		 2273 err = mnt_want_write_file(filp);
2274 if (err)
2275 return err;
2276
e4544b63 2277 f2fs_down_write(&sbi->sb_lock);
d0d3f1b3
CY		 2278
2279 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2280 goto got_it;
2281
f424f664
JK		 2282 /* update superblock with uuid */
2283 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2284
c5bda1c8 2285 err = f2fs_commit_super(sbi, false);
f424f664
JK		 2286 if (err) {
2287 /* undo new data */
2288 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
d0d3f1b3 2289 goto out_err;
f424f664
JK		 2290 }
2291got_it:
2292 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2293 16))
d0d3f1b3
CY		 2294 err = -EFAULT;
2295out_err:
e4544b63 2296 f2fs_up_write(&sbi->sb_lock);
d0d3f1b3
CY		 2297 mnt_drop_write_file(filp);
2298 return err;
f424f664
JK		 2299}
2300
8ce589c7
EB		 2301static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2302 unsigned long arg)
2303{
2304 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2305 return -EOPNOTSUPP;
2306
2307 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2308}
2309
2310static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2311{
2312 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2313 return -EOPNOTSUPP;
2314
2315 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2316}
2317
2318static int f2fs_ioc_remove_encryption_key(struct file *filp, 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(filp, (void __user *)arg);
2324}
2325
2326static int f2fs_ioc_remove_encryption_key_all_users(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_remove_key_all_users(filp, (void __user *)arg);
2333}
2334
2335static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2336 unsigned long arg)
2337{
2338 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2339 return -EOPNOTSUPP;
2340
2341 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2342}
2343
ee446e1a
EB		 2344static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2345{
2346 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2347 return -EOPNOTSUPP;
2348
2349 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2350}
2351
c1c1b583
CY		 2352static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2353{
2354 struct inode *inode = file_inode(filp);
2355 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d147ea4a
JK		 2356 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2357 .no_bg_gc = false,
c81d5bae
JK		 2358 .should_migrate_blocks = false,
2359 .nr_free_secs = 0 };
d530d4d8 2360 __u32 sync;
7fb17fe4 2361 int ret;
c1c1b583
CY		 2362
2363 if (!capable(CAP_SYS_ADMIN))
2364 return -EPERM;
2365
d530d4d8 2366 if (get_user(sync, (__u32 __user *)arg))
c1c1b583
CY		 2367 return -EFAULT;
2368
d530d4d8
CY		 2369 if (f2fs_readonly(sbi->sb))
2370 return -EROFS;
c1c1b583 2371
7fb17fe4
CY		 2372 ret = mnt_want_write_file(filp);
2373 if (ret)
2374 return ret;
2375
d530d4d8 2376 if (!sync) {
e4544b63 2377 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
7fb17fe4
CY		 2378 ret = -EBUSY;
2379 goto out;
2380 }
d530d4d8 2381 } else {
e4544b63 2382 f2fs_down_write(&sbi->gc_lock);
c1c1b583
CY		 2383 }
2384
d147ea4a
JK		 2385 gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2386 gc_control.err_gc_skipped = sync;
2387 ret = f2fs_gc(sbi, &gc_control);
7fb17fe4
CY		 2388out:
2389 mnt_drop_write_file(filp);
2390 return ret;
c1c1b583
CY		 2391}
2392
34178b1b 2393static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
34dc77ad 2394{
34178b1b 2395 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
d147ea4a
JK		 2396 struct f2fs_gc_control gc_control = {
2397 .init_gc_type = range->sync ? FG_GC : BG_GC,
2398 .no_bg_gc = false,
2399 .should_migrate_blocks = false,
c81d5bae
JK		 2400 .err_gc_skipped = range->sync,
2401 .nr_free_secs = 0 };
34dc77ad
JK		 2402 u64 end;
2403 int ret;
2404
2405 if (!capable(CAP_SYS_ADMIN))
2406 return -EPERM;
34dc77ad
JK		 2407 if (f2fs_readonly(sbi->sb))
2408 return -EROFS;
2409
34178b1b
CY		 2410 end = range->start + range->len;
2411 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
fbbf7799 2412 end >= MAX_BLKADDR(sbi))
b82f6e34 2413 return -EINVAL;
b82f6e34 2414
34dc77ad
JK		 2415 ret = mnt_want_write_file(filp);
2416 if (ret)
2417 return ret;
2418
34dc77ad 2419do_more:
34178b1b 2420 if (!range->sync) {
e4544b63 2421 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
34dc77ad
JK		 2422 ret = -EBUSY;
2423 goto out;
2424 }
2425 } else {
e4544b63 2426 f2fs_down_write(&sbi->gc_lock);
34dc77ad
JK		 2427 }
2428
d147ea4a
JK		 2429 gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2430 ret = f2fs_gc(sbi, &gc_control);
97767500
QZ		 2431 if (ret) {
2432 if (ret == -EBUSY)
2433 ret = -EAGAIN;
2434 goto out;
2435 }
34178b1b
CY		 2436 range->start += BLKS_PER_SEC(sbi);
2437 if (range->start <= end)
34dc77ad
JK		 2438 goto do_more;
2439out:
2440 mnt_drop_write_file(filp);
2441 return ret;
2442}
2443
34178b1b
CY		 2444static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2445{
2446 struct f2fs_gc_range range;
2447
2448 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2449 sizeof(range)))
2450 return -EFAULT;
2451 return __f2fs_ioc_gc_range(filp, &range);
2452}
2453
059c0648 2454static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
456b88e4
CY		 2455{
2456 struct inode *inode = file_inode(filp);
2457 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
7fb17fe4 2458 int ret;
456b88e4
CY		 2459
2460 if (!capable(CAP_SYS_ADMIN))
2461 return -EPERM;
2462
2463 if (f2fs_readonly(sbi->sb))
2464 return -EROFS;
2465
4354994f 2466 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
dcbb4c10 2467 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
4354994f
DR		 2468 return -EINVAL;
2469 }
2470
7fb17fe4
CY		 2471 ret = mnt_want_write_file(filp);
2472 if (ret)
2473 return ret;
2474
2475 ret = f2fs_sync_fs(sbi->sb, 1);
2476
2477 mnt_drop_write_file(filp);
2478 return ret;
456b88e4
CY		 2479}
2480
d323d005
CY		 2481static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2482 struct file *filp,
2483 struct f2fs_defragment *range)
2484{
2485 struct inode *inode = file_inode(filp);
f3d98e74 2486 struct f2fs_map_blocks map = { .m_next_extent = NULL,
5f029c04 2487 .m_seg_type = NO_CHECK_TYPE,
f4f0b677 2488 .m_may_create = false };
1061fd48 2489 struct extent_info ei = {0, 0, 0};
f3d98e74 2490 pgoff_t pg_start, pg_end, next_pgofs;
3519e3f9 2491 unsigned int blk_per_seg = sbi->blocks_per_seg;
d323d005 2492 unsigned int total = 0, sec_num;
d323d005
CY		 2493 block_t blk_end = 0;
2494 bool fragmented = false;
2495 int err;
2496
09cbfeaf
KS		 2497 pg_start = range->start >> PAGE_SHIFT;
2498 pg_end = (range->start + range->len) >> PAGE_SHIFT;
d323d005 2499
2c4db1a6 2500 f2fs_balance_fs(sbi, true);
d323d005 2501
5955102c 2502 inode_lock(inode);
d323d005 2503
1018a546
CY		 2504 /* if in-place-update policy is enabled, don't waste time here */
2505 set_inode_flag(inode, FI_OPU_WRITE);
2506 if (f2fs_should_update_inplace(inode, NULL)) {
2507 err = -EINVAL;
2508 goto out;
2509 }
2510
d323d005
CY		 2511 /* writeback all dirty pages in the range */
2512 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
d8fe4f0e 2513 range->start + range->len - 1);
d323d005
CY		 2514 if (err)
2515 goto out;
2516
2517 /*
2518 * lookup mapping info in extent cache, skip defragmenting if physical
2519 * block addresses are continuous.
2520 */
2521 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2522 if (ei.fofs + ei.len >= pg_end)
2523 goto out;
2524 }
2525
2526 map.m_lblk = pg_start;
f3d98e74 2527 map.m_next_pgofs = &next_pgofs;
d323d005
CY		 2528
2529 /*
2530 * lookup mapping info in dnode page cache, skip defragmenting if all
2531 * physical block addresses are continuous even if there are hole(s)
2532 * in logical blocks.
2533 */
2534 while (map.m_lblk < pg_end) {
a1c1e9b7 2535 map.m_len = pg_end - map.m_lblk;
f2220c7f 2536 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2537 if (err)
2538 goto out;
2539
2540 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2541 map.m_lblk = next_pgofs;
d323d005
CY		 2542 continue;
2543 }
2544
25a912e5 2545 if (blk_end && blk_end != map.m_pblk)
d323d005 2546 fragmented = true;
25a912e5
CY		 2547
2548 /* record total count of block that we're going to move */
2549 total += map.m_len;
2550
d323d005
CY		 2551 blk_end = map.m_pblk + map.m_len;
2552
2553 map.m_lblk += map.m_len;
d323d005
CY		 2554 }
2555
d3a1a0e1
CY		 2556 if (!fragmented) {
2557 total = 0;
d323d005 2558 goto out;
d3a1a0e1 2559 }
d323d005 2560
f91108b8 2561 sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
d323d005
CY		 2562
2563 /*
2564 * make sure there are enough free section for LFS allocation, this can
2565 * avoid defragment running in SSR mode when free section are allocated
2566 * intensively
2567 */
7f3037a5 2568 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
d323d005
CY		 2569 err = -EAGAIN;
2570 goto out;
2571 }
2572
25a912e5
CY		 2573 map.m_lblk = pg_start;
2574 map.m_len = pg_end - pg_start;
2575 total = 0;
2576
d323d005
CY		 2577 while (map.m_lblk < pg_end) {
2578 pgoff_t idx;
2579 int cnt = 0;
2580
2581do_map:
a1c1e9b7 2582 map.m_len = pg_end - map.m_lblk;
f2220c7f 2583 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2584 if (err)
2585 goto clear_out;
2586
2587 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2588 map.m_lblk = next_pgofs;
d3a1a0e1 2589 goto check;
d323d005
CY		 2590 }
2591
1018a546 2592 set_inode_flag(inode, FI_SKIP_WRITES);
d323d005
CY		 2593
2594 idx = map.m_lblk;
2595 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2596 struct page *page;
2597
4d57b86d 2598 page = f2fs_get_lock_data_page(inode, idx, true);
d323d005
CY		 2599 if (IS_ERR(page)) {
2600 err = PTR_ERR(page);
2601 goto clear_out;
2602 }
2603
2604 set_page_dirty(page);
2605 f2fs_put_page(page, 1);
2606
2607 idx++;
2608 cnt++;
2609 total++;
2610 }
2611
2612 map.m_lblk = idx;
d3a1a0e1
CY		 2613check:
2614 if (map.m_lblk < pg_end && cnt < blk_per_seg)
d323d005
CY		 2615 goto do_map;
2616
1018a546 2617 clear_inode_flag(inode, FI_SKIP_WRITES);
d323d005
CY		 2618
2619 err = filemap_fdatawrite(inode->i_mapping);
2620 if (err)
2621 goto out;
2622 }
2623clear_out:
1018a546 2624 clear_inode_flag(inode, FI_SKIP_WRITES);
d323d005 2625out:
1018a546 2626 clear_inode_flag(inode, FI_OPU_WRITE);
5955102c 2627 inode_unlock(inode);
d323d005 2628 if (!err)
09cbfeaf 2629 range->len = (u64)total << PAGE_SHIFT;
d323d005
CY		 2630 return err;
2631}
2632
2633static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2634{
2635 struct inode *inode = file_inode(filp);
2636 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2637 struct f2fs_defragment range;
2638 int err;
2639
2640 if (!capable(CAP_SYS_ADMIN))
2641 return -EPERM;
2642
7eab0c0d 2643 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
d323d005
CY		 2644 return -EINVAL;
2645
d7563861
KM		 2646 if (f2fs_readonly(sbi->sb))
2647 return -EROFS;
d323d005
CY		 2648
2649 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
d7563861
KM		 2650 sizeof(range)))
2651 return -EFAULT;
d323d005
CY		 2652
2653 /* verify alignment of offset & size */
d7563861
KM		 2654 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2655 return -EINVAL;
d323d005 2656
1941d7bc 2657 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
6d1451bf 2658 max_file_blocks(inode)))
d7563861
KM		 2659 return -EINVAL;
2660
2661 err = mnt_want_write_file(filp);
2662 if (err)
2663 return err;
1941d7bc 2664
d323d005 2665 err = f2fs_defragment_range(sbi, filp, &range);
d7563861
KM		 2666 mnt_drop_write_file(filp);
2667
d0239e1b 2668 f2fs_update_time(sbi, REQ_TIME);
d323d005 2669 if (err < 0)
d7563861 2670 return err;
d323d005
CY		 2671
2672 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2673 sizeof(range)))
d7563861
KM		 2674 return -EFAULT;
2675
2676 return 0;
d323d005
CY		 2677}
2678
4dd6f977
JK		 2679static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2680 struct file *file_out, loff_t pos_out, size_t len)
2681{
2682 struct inode *src = file_inode(file_in);
2683 struct inode *dst = file_inode(file_out);
2684 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2685 size_t olen = len, dst_max_i_size = 0;
2686 size_t dst_osize;
2687 int ret;
2688
2689 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2690 src->i_sb != dst->i_sb)
2691 return -EXDEV;
2692
2693 if (unlikely(f2fs_readonly(src->i_sb)))
2694 return -EROFS;
2695
fe8494bf
CY		 2696 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2697 return -EINVAL;
4dd6f977 2698
62230e0d 2699 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
4dd6f977
JK		 2700 return -EOPNOTSUPP;
2701
aad1383c
DR		 2702 if (pos_out < 0 || pos_in < 0)
2703 return -EINVAL;
2704
d95fd91c
FL		 2705 if (src == dst) {
2706 if (pos_in == pos_out)
2707 return 0;
2708 if (pos_out > pos_in && pos_out < pos_in + len)
2709 return -EINVAL;
2710 }
2711
4dd6f977 2712 inode_lock(src);
20a3d61d 2713 if (src != dst) {
bb06664a
CY		 2714 ret = -EBUSY;
2715 if (!inode_trylock(dst))
2716 goto out;
20a3d61d 2717 }
4dd6f977
JK		 2718
2719 ret = -EINVAL;
2720 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2721 goto out_unlock;
2722 if (len == 0)
2723 olen = len = src->i_size - pos_in;
2724 if (pos_in + len == src->i_size)
2725 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2726 if (len == 0) {
2727 ret = 0;
2728 goto out_unlock;
2729 }
2730
2731 dst_osize = dst->i_size;
2732 if (pos_out + olen > dst->i_size)
2733 dst_max_i_size = pos_out + olen;
2734
2735 /* verify the end result is block aligned */
2736 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2737 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2738 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2739 goto out_unlock;
2740
2741 ret = f2fs_convert_inline_inode(src);
2742 if (ret)
2743 goto out_unlock;
2744
2745 ret = f2fs_convert_inline_inode(dst);
2746 if (ret)
2747 goto out_unlock;
2748
2749 /* write out all dirty pages from offset */
2750 ret = filemap_write_and_wait_range(src->i_mapping,
2751 pos_in, pos_in + len);
2752 if (ret)
2753 goto out_unlock;
2754
2755 ret = filemap_write_and_wait_range(dst->i_mapping,
2756 pos_out, pos_out + len);
2757 if (ret)
2758 goto out_unlock;
2759
2760 f2fs_balance_fs(sbi, true);
6f8d4455 2761
e4544b63 2762 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
6f8d4455
JK		 2763 if (src != dst) {
2764 ret = -EBUSY;
e4544b63 2765 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
6f8d4455
JK		 2766 goto out_src;
2767 }
2768
4dd6f977 2769 f2fs_lock_op(sbi);
61e4da11
FL		 2770 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2771 pos_out >> F2FS_BLKSIZE_BITS,
2772 len >> F2FS_BLKSIZE_BITS, false);
4dd6f977
JK		 2773
2774 if (!ret) {
2775 if (dst_max_i_size)
2776 f2fs_i_size_write(dst, dst_max_i_size);
2777 else if (dst_osize != dst->i_size)
2778 f2fs_i_size_write(dst, dst_osize);
2779 }
2780 f2fs_unlock_op(sbi);
6f8d4455
JK		 2781
2782 if (src != dst)
e4544b63 2783 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
6f8d4455 2784out_src:
e4544b63 2785 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
6f8d4455
JK		 2786out_unlock:
2787 if (src != dst)
4dd6f977 2788 inode_unlock(dst);
20a3d61d 2789out:
4dd6f977
JK		 2790 inode_unlock(src);
2791 return ret;
2792}
2793
34178b1b
CY		 2794static int __f2fs_ioc_move_range(struct file *filp,
2795 struct f2fs_move_range *range)
4dd6f977 2796{
4dd6f977
JK		 2797 struct fd dst;
2798 int err;
2799
2800 if (!(filp->f_mode & FMODE_READ) ||
2801 !(filp->f_mode & FMODE_WRITE))
2802 return -EBADF;
2803
34178b1b 2804 dst = fdget(range->dst_fd);
4dd6f977
JK		 2805 if (!dst.file)
2806 return -EBADF;
2807
2808 if (!(dst.file->f_mode & FMODE_WRITE)) {
2809 err = -EBADF;
2810 goto err_out;
2811 }
2812
2813 err = mnt_want_write_file(filp);
2814 if (err)
2815 goto err_out;
2816
34178b1b
CY		 2817 err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2818 range->pos_out, range->len);
4dd6f977
JK		 2819
2820 mnt_drop_write_file(filp);
4dd6f977
JK		 2821err_out:
2822 fdput(dst);
2823 return err;
2824}
2825
34178b1b
CY		 2826static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2827{
2828 struct f2fs_move_range range;
2829
2830 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2831 sizeof(range)))
2832 return -EFAULT;
2833 return __f2fs_ioc_move_range(filp, &range);
2834}
2835
e066b83c
JK		 2836static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2837{
2838 struct inode *inode = file_inode(filp);
2839 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2840 struct sit_info *sm = SIT_I(sbi);
2841 unsigned int start_segno = 0, end_segno = 0;
2842 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2843 struct f2fs_flush_device range;
d147ea4a
JK		 2844 struct f2fs_gc_control gc_control = {
2845 .init_gc_type = FG_GC,
2846 .should_migrate_blocks = true,
c81d5bae
JK		 2847 .err_gc_skipped = true,
2848 .nr_free_secs = 0 };
e066b83c
JK		 2849 int ret;
2850
2851 if (!capable(CAP_SYS_ADMIN))
2852 return -EPERM;
2853
2854 if (f2fs_readonly(sbi->sb))
2855 return -EROFS;
2856
4354994f
DR		 2857 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2858 return -EINVAL;
2859
e066b83c
JK		 2860 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2861 sizeof(range)))
2862 return -EFAULT;
2863
0916878d 2864 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2c70c5e3 2865 __is_large_section(sbi)) {
dcbb4c10
JP		 2866 f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2867 range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
e066b83c
JK		 2868 return -EINVAL;
2869 }
2870
2871 ret = mnt_want_write_file(filp);
2872 if (ret)
2873 return ret;
2874
2875 if (range.dev_num != 0)
2876 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2877 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2878
2879 start_segno = sm->last_victim[FLUSH_DEVICE];
2880 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2881 start_segno = dev_start_segno;
2882 end_segno = min(start_segno + range.segments, dev_end_segno);
2883
2884 while (start_segno < end_segno) {
e4544b63 2885 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
e066b83c
JK		 2886 ret = -EBUSY;
2887 goto out;
2888 }
2889 sm->last_victim[GC_CB] = end_segno + 1;
2890 sm->last_victim[GC_GREEDY] = end_segno + 1;
2891 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
d147ea4a
JK		 2892
2893 gc_control.victim_segno = start_segno;
2894 ret = f2fs_gc(sbi, &gc_control);
e066b83c
JK		 2895 if (ret == -EAGAIN)
2896 ret = 0;
2897 else if (ret < 0)
2898 break;
2899 start_segno++;
2900 }
2901out:
2902 mnt_drop_write_file(filp);
2903 return ret;
2904}
2905
e65ef207
JK		 2906static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2907{
2908 struct inode *inode = file_inode(filp);
2909 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2910
2911 /* Must validate to set it with SQLite behavior in Android. */
2912 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2913
2914 return put_user(sb_feature, (u32 __user *)arg);
2915}
e066b83c 2916
2c1d0305 2917#ifdef CONFIG_QUOTA
78130819
CY		 2918int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2919{
2920 struct dquot *transfer_to[MAXQUOTAS] = {};
2921 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2922 struct super_block *sb = sbi->sb;
2923 int err = 0;
2924
2925 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2926 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2927 err = __dquot_transfer(inode, transfer_to);
2928 if (err)
2929 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2930 dqput(transfer_to[PRJQUOTA]);
2931 }
2932 return err;
2933}
2934
9b1bb01c 2935static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2c1d0305 2936{
2c1d0305
CY		 2937 struct f2fs_inode_info *fi = F2FS_I(inode);
2938 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d13732cc 2939 struct f2fs_inode *ri = NULL;
2c1d0305
CY		 2940 kprojid_t kprojid;
2941 int err;
2942
7beb01f7 2943 if (!f2fs_sb_has_project_quota(sbi)) {
2c1d0305
CY		 2944 if (projid != F2FS_DEF_PROJID)
2945 return -EOPNOTSUPP;
2946 else
2947 return 0;
2948 }
2949
2950 if (!f2fs_has_extra_attr(inode))
2951 return -EOPNOTSUPP;
2952
2953 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2954
054cb289 2955 if (projid_eq(kprojid, fi->i_projid))
2c1d0305
CY		 2956 return 0;
2957
2c1d0305 2958 err = -EPERM;
2c1d0305
CY		 2959 /* Is it quota file? Do not allow user to mess with it */
2960 if (IS_NOQUOTA(inode))
c8e92757 2961 return err;
2c1d0305 2962
d13732cc
JY		 2963 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
2964 return -EOVERFLOW;
2c1d0305 2965
10a26878 2966 err = f2fs_dquot_initialize(inode);
c22aecd7 2967 if (err)
c8e92757 2968 return err;
2c1d0305 2969
78130819
CY		 2970 f2fs_lock_op(sbi);
2971 err = f2fs_transfer_project_quota(inode, kprojid);
2972 if (err)
2973 goto out_unlock;
2c1d0305 2974
054cb289 2975 fi->i_projid = kprojid;
2c1d0305 2976 inode->i_ctime = current_time(inode);
2c1d0305 2977 f2fs_mark_inode_dirty_sync(inode, true);
78130819
CY		 2978out_unlock:
2979 f2fs_unlock_op(sbi);
2c1d0305
CY		 2980 return err;
2981}
2982#else
78130819
CY		 2983int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2984{
2985 return 0;
2986}
2987
9b1bb01c 2988static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2c1d0305
CY		 2989{
2990 if (projid != F2FS_DEF_PROJID)
2991 return -EOPNOTSUPP;
2992 return 0;
2993}
2994#endif
2995
9b1bb01c 2996int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
2c1d0305 2997{
9b1bb01c 2998 struct inode *inode = d_inode(dentry);
2c1d0305 2999 struct f2fs_inode_info *fi = F2FS_I(inode);
9b1bb01c 3000 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
2c1d0305 3001
9b1bb01c
MS		 3002 if (IS_ENCRYPTED(inode))
3003 fsflags |= FS_ENCRYPT_FL;
3004 if (IS_VERITY(inode))
3005 fsflags |= FS_VERITY_FL;
3006 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3007 fsflags |= FS_INLINE_DATA_FL;
3008 if (is_inode_flag_set(inode, FI_PIN_FILE))
3009 fsflags |= FS_NOCOW_FL;
3010
3011 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
2c1d0305 3012
7beb01f7 3013 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
6fc93c4e 3014 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2c1d0305 3015
c8e92757
WS		 3016 return 0;
3017}
3018
9b1bb01c
MS		 3019int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3020 struct dentry *dentry, struct fileattr *fa)
2c1d0305 3021{
9b1bb01c
MS		 3022 struct inode *inode = d_inode(dentry);
3023 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
36098557 3024 u32 iflags;
2c1d0305
CY		 3025 int err;
3026
9b1bb01c
MS		 3027 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3028 return -EIO;
3029 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3030 return -ENOSPC;
3031 if (fsflags & ~F2FS_GETTABLE_FS_FL)
2c1d0305 3032 return -EOPNOTSUPP;
9b1bb01c
MS		 3033 fsflags &= F2FS_SETTABLE_FS_FL;
3034 if (!fa->flags_valid)
3035 mask &= FS_COMMON_FL;
2c1d0305 3036
9b1bb01c 3037 iflags = f2fs_fsflags_to_iflags(fsflags);
36098557 3038 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2c1d0305
CY		 3039 return -EOPNOTSUPP;
3040
9b1bb01c
MS		 3041 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3042 if (!err)
3043 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
2c1d0305 3044
c8e92757 3045 return err;
2c1d0305 3046}
e066b83c 3047
1ad71a27
JK		 3048int f2fs_pin_file_control(struct inode *inode, bool inc)
3049{
3050 struct f2fs_inode_info *fi = F2FS_I(inode);
3051 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3052
3053 /* Use i_gc_failures for normal file as a risk signal. */
3054 if (inc)
2ef79ecb
CY		 3055 f2fs_i_gc_failures_write(inode,
3056 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
1ad71a27 3057
2ef79ecb 3058 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
dcbb4c10
JP		 3059 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3060 __func__, inode->i_ino,
3061 fi->i_gc_failures[GC_FAILURE_PIN]);
1ad71a27
JK		 3062 clear_inode_flag(inode, FI_PIN_FILE);
3063 return -EAGAIN;
3064 }
3065 return 0;
3066}
3067
3068static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3069{
3070 struct inode *inode = file_inode(filp);
3071 __u32 pin;
3072 int ret = 0;
3073
1ad71a27
JK		 3074 if (get_user(pin, (__u32 __user *)arg))
3075 return -EFAULT;
3076
3077 if (!S_ISREG(inode->i_mode))
3078 return -EINVAL;
3079
3080 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3081 return -EROFS;
3082
3083 ret = mnt_want_write_file(filp);
3084 if (ret)
3085 return ret;
3086
3087 inode_lock(inode);
3088
3089 if (!pin) {
3090 clear_inode_flag(inode, FI_PIN_FILE);
30933364 3091 f2fs_i_gc_failures_write(inode, 0);
1ad71a27
JK		 3092 goto done;
3093 }
3094
19bdba52
JK		 3095 if (f2fs_should_update_outplace(inode, NULL)) {
3096 ret = -EINVAL;
3097 goto out;
3098 }
3099
1ad71a27
JK		 3100 if (f2fs_pin_file_control(inode, false)) {
3101 ret = -EAGAIN;
3102 goto out;
3103 }
4c8ff709 3104
1ad71a27
JK		 3105 ret = f2fs_convert_inline_inode(inode);
3106 if (ret)
3107 goto out;
3108
78134d03 3109 if (!f2fs_disable_compressed_file(inode)) {
4c8ff709
CY		 3110 ret = -EOPNOTSUPP;
3111 goto out;
3112 }
3113
1ad71a27 3114 set_inode_flag(inode, FI_PIN_FILE);
2ef79ecb 3115 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
1ad71a27
JK		 3116done:
3117 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3118out:
3119 inode_unlock(inode);
3120 mnt_drop_write_file(filp);
3121 return ret;
3122}
3123
3124static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3125{
3126 struct inode *inode = file_inode(filp);
3127 __u32 pin = 0;
3128
3129 if (is_inode_flag_set(inode, FI_PIN_FILE))
2ef79ecb 3130 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
1ad71a27
JK		 3131 return put_user(pin, (u32 __user *)arg);
3132}
3133
c4020b2d
CY		 3134int f2fs_precache_extents(struct inode *inode)
3135{
3136 struct f2fs_inode_info *fi = F2FS_I(inode);
3137 struct f2fs_map_blocks map;
3138 pgoff_t m_next_extent;
3139 loff_t end;
3140 int err;
3141
3142 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3143 return -EOPNOTSUPP;
3144
3145 map.m_lblk = 0;
3146 map.m_next_pgofs = NULL;
3147 map.m_next_extent = &m_next_extent;
3148 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 3149 map.m_may_create = false;
6d1451bf 3150 end = max_file_blocks(inode);
c4020b2d
CY		 3151
3152 while (map.m_lblk < end) {
3153 map.m_len = end - map.m_lblk;
3154
e4544b63 3155 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
c4020b2d 3156 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
e4544b63 3157 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
c4020b2d
CY		 3158 if (err)
3159 return err;
3160
3161 map.m_lblk = m_next_extent;
3162 }
3163
4f55dc2a 3164 return 0;
c4020b2d
CY		 3165}
3166
3167static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3168{
3169 return f2fs_precache_extents(file_inode(filp));
3170}
3171
04f0b2ea
QS		 3172static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3173{
3174 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3175 __u64 block_count;
04f0b2ea
QS		 3176
3177 if (!capable(CAP_SYS_ADMIN))
3178 return -EPERM;
3179
3180 if (f2fs_readonly(sbi->sb))
3181 return -EROFS;
3182
3183 if (copy_from_user(&block_count, (void __user *)arg,
3184 sizeof(block_count)))
3185 return -EFAULT;
3186
b4b10061 3187 return f2fs_resize_fs(sbi, block_count);
04f0b2ea
QS		 3188}
3189
95ae251f
EB		 3190static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3191{
3192 struct inode *inode = file_inode(filp);
3193
3194 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3195
3196 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3197 f2fs_warn(F2FS_I_SB(inode),
833dcd35 3198 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
95ae251f
EB		 3199 inode->i_ino);
3200 return -EOPNOTSUPP;
3201 }
3202
3203 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3204}
3205
3206static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3207{
3208 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3209 return -EOPNOTSUPP;
3210
3211 return fsverity_ioctl_measure(filp, (void __user *)arg);
3212}
3213
e17fe657
EB		 3214static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3215{
3216 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3217 return -EOPNOTSUPP;
3218
3219 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3220}
3221
3357af8f 3222static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
4507847c
CY		 3223{
3224 struct inode *inode = file_inode(filp);
3225 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3226 char *vbuf;
3227 int count;
3228 int err = 0;
3229
3230 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3231 if (!vbuf)
3232 return -ENOMEM;
3233
e4544b63 3234 f2fs_down_read(&sbi->sb_lock);
4507847c
CY		 3235 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3236 ARRAY_SIZE(sbi->raw_super->volume_name),
3237 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
e4544b63 3238 f2fs_up_read(&sbi->sb_lock);
4507847c
CY		 3239
3240 if (copy_to_user((char __user *)arg, vbuf,
3241 min(FSLABEL_MAX, count)))
3242 err = -EFAULT;
3243
c8eb7024 3244 kfree(vbuf);
4507847c
CY		 3245 return err;
3246}
3247
3357af8f 3248static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
4507847c
CY		 3249{
3250 struct inode *inode = file_inode(filp);
3251 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3252 char *vbuf;
3253 int err = 0;
3254
3255 if (!capable(CAP_SYS_ADMIN))
3256 return -EPERM;
3257
3258 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3259 if (IS_ERR(vbuf))
3260 return PTR_ERR(vbuf);
3261
3262 err = mnt_want_write_file(filp);
3263 if (err)
3264 goto out;
3265
e4544b63 3266 f2fs_down_write(&sbi->sb_lock);
4507847c
CY		 3267
3268 memset(sbi->raw_super->volume_name, 0,
3269 sizeof(sbi->raw_super->volume_name));
3270 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3271 sbi->raw_super->volume_name,
3272 ARRAY_SIZE(sbi->raw_super->volume_name));
3273
3274 err = f2fs_commit_super(sbi, false);
3275
e4544b63 3276 f2fs_up_write(&sbi->sb_lock);
4507847c
CY		 3277
3278 mnt_drop_write_file(filp);
3279out:
3280 kfree(vbuf);
3281 return err;
3282}
3283
439dfb10
CY		 3284static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3285{
3286 struct inode *inode = file_inode(filp);
3287 __u64 blocks;
3288
3289 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3290 return -EOPNOTSUPP;
3291
3292 if (!f2fs_compressed_file(inode))
3293 return -EINVAL;
3294
c2759eba 3295 blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
439dfb10
CY		 3296 return put_user(blocks, (u64 __user *)arg);
3297}
3298
ef8d563f
CY		 3299static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3300{
3301 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3302 unsigned int released_blocks = 0;
3303 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3304 block_t blkaddr;
3305 int i;
3306
3307 for (i = 0; i < count; i++) {
3308 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3309 dn->ofs_in_node + i);
3310
3311 if (!__is_valid_data_blkaddr(blkaddr))
3312 continue;
3313 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3314 DATA_GENERIC_ENHANCE)))
3315 return -EFSCORRUPTED;
3316 }
3317
3318 while (count) {
3319 int compr_blocks = 0;
3320
3321 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3322 blkaddr = f2fs_data_blkaddr(dn);
3323
3324 if (i == 0) {
3325 if (blkaddr == COMPRESS_ADDR)
3326 continue;
3327 dn->ofs_in_node += cluster_size;
3328 goto next;
3329 }
3330
3331 if (__is_valid_data_blkaddr(blkaddr))
3332 compr_blocks++;
3333
3334 if (blkaddr != NEW_ADDR)
3335 continue;
3336
3337 dn->data_blkaddr = NULL_ADDR;
3338 f2fs_set_data_blkaddr(dn);
3339 }
3340
3341 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3342 dec_valid_block_count(sbi, dn->inode,
3343 cluster_size - compr_blocks);
3344
3345 released_blocks += cluster_size - compr_blocks;
3346next:
3347 count -= cluster_size;
3348 }
3349
3350 return released_blocks;
3351}
3352
3353static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3354{
3355 struct inode *inode = file_inode(filp);
3356 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3357 pgoff_t page_idx = 0, last_idx;
3358 unsigned int released_blocks = 0;
3359 int ret;
3360 int writecount;
3361
3362 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3363 return -EOPNOTSUPP;
3364
3365 if (!f2fs_compressed_file(inode))
3366 return -EINVAL;
3367
3368 if (f2fs_readonly(sbi->sb))
3369 return -EROFS;
3370
3371 ret = mnt_want_write_file(filp);
3372 if (ret)
3373 return ret;
3374
3375 f2fs_balance_fs(F2FS_I_SB(inode), true);
3376
3377 inode_lock(inode);
3378
3379 writecount = atomic_read(&inode->i_writecount);
8c8cf26a
DJ		 3380 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3381 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
ef8d563f
CY		 3382 ret = -EBUSY;
3383 goto out;
3384 }
3385
c6140415 3386 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
ef8d563f
CY		 3387 ret = -EINVAL;
3388 goto out;
3389 }
3390
3391 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3392 if (ret)
3393 goto out;
3394
c6140415 3395 set_inode_flag(inode, FI_COMPRESS_RELEASED);
ef8d563f
CY		 3396 inode->i_ctime = current_time(inode);
3397 f2fs_mark_inode_dirty_sync(inode, true);
3398
c2759eba 3399 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
567c4bf5
DJ		 3400 goto out;
3401
e4544b63 3402 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 3403 filemap_invalidate_lock(inode->i_mapping);
ef8d563f
CY		 3404
3405 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3406
3407 while (page_idx < last_idx) {
3408 struct dnode_of_data dn;
3409 pgoff_t end_offset, count;
3410
3411 set_new_dnode(&dn, inode, NULL, NULL, 0);
3412 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3413 if (ret) {
3414 if (ret == -ENOENT) {
3415 page_idx = f2fs_get_next_page_offset(&dn,
3416 page_idx);
3417 ret = 0;
3418 continue;
3419 }
3420 break;
3421 }
3422
3423 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3424 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4fec3fc0 3425 count = round_up(count, F2FS_I(inode)->i_cluster_size);
ef8d563f
CY		 3426
3427 ret = release_compress_blocks(&dn, count);
3428
3429 f2fs_put_dnode(&dn);
3430
3431 if (ret < 0)
3432 break;
3433
3434 page_idx += count;
3435 released_blocks += ret;
3436 }
3437
edc6d01b 3438 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 3439 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
ef8d563f
CY		 3440out:
3441 inode_unlock(inode);
3442
3443 mnt_drop_write_file(filp);
3444
3445 if (ret >= 0) {
3446 ret = put_user(released_blocks, (u64 __user *)arg);
c2759eba
DJ		 3447 } else if (released_blocks &&
3448 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
ef8d563f
CY		 3449 set_sbi_flag(sbi, SBI_NEED_FSCK);
3450 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
c2759eba 3451 "iblocks=%llu, released=%u, compr_blocks=%u, "
ef8d563f
CY		 3452 "run fsck to fix.",
3453 __func__, inode->i_ino, inode->i_blocks,
3454 released_blocks,
c2759eba 3455 atomic_read(&F2FS_I(inode)->i_compr_blocks));
ef8d563f
CY		 3456 }
3457
3458 return ret;
3459}
3460
c75488fb
CY		 3461static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3462{
3463 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3464 unsigned int reserved_blocks = 0;
3465 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3466 block_t blkaddr;
3467 int i;
3468
3469 for (i = 0; i < count; i++) {
3470 blkaddr = data_blkaddr(dn->inode, dn->node_page,
3471 dn->ofs_in_node + i);
3472
3473 if (!__is_valid_data_blkaddr(blkaddr))
3474 continue;
3475 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3476 DATA_GENERIC_ENHANCE)))
3477 return -EFSCORRUPTED;
3478 }
3479
3480 while (count) {
3481 int compr_blocks = 0;
3482 blkcnt_t reserved;
3483 int ret;
3484
3485 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3486 blkaddr = f2fs_data_blkaddr(dn);
3487
3488 if (i == 0) {
3489 if (blkaddr == COMPRESS_ADDR)
3490 continue;
3491 dn->ofs_in_node += cluster_size;
3492 goto next;
3493 }
3494
3495 if (__is_valid_data_blkaddr(blkaddr)) {
3496 compr_blocks++;
3497 continue;
3498 }
3499
3500 dn->data_blkaddr = NEW_ADDR;
3501 f2fs_set_data_blkaddr(dn);
3502 }
3503
3504 reserved = cluster_size - compr_blocks;
3505 ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3506 if (ret)
3507 return ret;
3508
3509 if (reserved != cluster_size - compr_blocks)
3510 return -ENOSPC;
3511
3512 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3513
3514 reserved_blocks += reserved;
3515next:
3516 count -= cluster_size;
3517 }
3518
3519 return reserved_blocks;
3520}
3521
3522static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3523{
3524 struct inode *inode = file_inode(filp);
3525 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3526 pgoff_t page_idx = 0, last_idx;
3527 unsigned int reserved_blocks = 0;
3528 int ret;
3529
3530 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3531 return -EOPNOTSUPP;
3532
3533 if (!f2fs_compressed_file(inode))
3534 return -EINVAL;
3535
3536 if (f2fs_readonly(sbi->sb))
3537 return -EROFS;
3538
3539 ret = mnt_want_write_file(filp);
3540 if (ret)
3541 return ret;
3542
c2759eba 3543 if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
c75488fb
CY		 3544 goto out;
3545
3546 f2fs_balance_fs(F2FS_I_SB(inode), true);
3547
3548 inode_lock(inode);
3549
c6140415 3550 if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
c75488fb
CY		 3551 ret = -EINVAL;
3552 goto unlock_inode;
3553 }
3554
e4544b63 3555 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 3556 filemap_invalidate_lock(inode->i_mapping);
c75488fb
CY		 3557
3558 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3559
3560 while (page_idx < last_idx) {
3561 struct dnode_of_data dn;
3562 pgoff_t end_offset, count;
3563
3564 set_new_dnode(&dn, inode, NULL, NULL, 0);
3565 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3566 if (ret) {
3567 if (ret == -ENOENT) {
3568 page_idx = f2fs_get_next_page_offset(&dn,
3569 page_idx);
3570 ret = 0;
3571 continue;
3572 }
3573 break;
3574 }
3575
3576 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3577 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
4fec3fc0 3578 count = round_up(count, F2FS_I(inode)->i_cluster_size);
c75488fb
CY		 3579
3580 ret = reserve_compress_blocks(&dn, count);
3581
3582 f2fs_put_dnode(&dn);
3583
3584 if (ret < 0)
3585 break;
3586
3587 page_idx += count;
3588 reserved_blocks += ret;
3589 }
3590
edc6d01b 3591 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 3592 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
c75488fb
CY		 3593
3594 if (ret >= 0) {
c6140415 3595 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
c75488fb
CY		 3596 inode->i_ctime = current_time(inode);
3597 f2fs_mark_inode_dirty_sync(inode, true);
3598 }
3599unlock_inode:
3600 inode_unlock(inode);
3601out:
3602 mnt_drop_write_file(filp);
3603
3604 if (ret >= 0) {
3605 ret = put_user(reserved_blocks, (u64 __user *)arg);
c2759eba
DJ		 3606 } else if (reserved_blocks &&
3607 atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
c75488fb
CY		 3608 set_sbi_flag(sbi, SBI_NEED_FSCK);
3609 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
c2759eba 3610 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
c75488fb
CY		 3611 "run fsck to fix.",
3612 __func__, inode->i_ino, inode->i_blocks,
3613 reserved_blocks,
c2759eba 3614 atomic_read(&F2FS_I(inode)->i_compr_blocks));
c75488fb
CY		 3615 }
3616
3617 return ret;
3618}
3619
9af84648
DJ		 3620static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3621 pgoff_t off, block_t block, block_t len, u32 flags)
3622{
3623 struct request_queue *q = bdev_get_queue(bdev);
3624 sector_t sector = SECTOR_FROM_BLOCK(block);
3625 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3626 int ret = 0;
3627
3628 if (!q)
3629 return -ENXIO;
3630
3631 if (flags & F2FS_TRIM_FILE_DISCARD)
3632 ret = blkdev_issue_discard(bdev, sector, nr_sects, GFP_NOFS,
3633 blk_queue_secure_erase(q) ?
3634 BLKDEV_DISCARD_SECURE : 0);
3635
3636 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3637 if (IS_ENCRYPTED(inode))
3638 ret = fscrypt_zeroout_range(inode, off, block, len);
3639 else
3640 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3641 GFP_NOFS, 0);
3642 }
3643
3644 return ret;
3645}
3646
3647static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3648{
3649 struct inode *inode = file_inode(filp);
3650 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3651 struct address_space *mapping = inode->i_mapping;
3652 struct block_device *prev_bdev = NULL;
3653 struct f2fs_sectrim_range range;
3654 pgoff_t index, pg_end, prev_index = 0;
3655 block_t prev_block = 0, len = 0;
3656 loff_t end_addr;
3657 bool to_end = false;
3658 int ret = 0;
3659
3660 if (!(filp->f_mode & FMODE_WRITE))
3661 return -EBADF;
3662
3663 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3664 sizeof(range)))
3665 return -EFAULT;
3666
3667 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3668 !S_ISREG(inode->i_mode))
3669 return -EINVAL;
3670
3671 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3672 !f2fs_hw_support_discard(sbi)) ||
3673 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3674 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3675 return -EOPNOTSUPP;
3676
3677 file_start_write(filp);
3678 inode_lock(inode);
3679
3680 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3681 range.start >= inode->i_size) {
3682 ret = -EINVAL;
3683 goto err;
3684 }
3685
3686 if (range.len == 0)
3687 goto err;
3688
3689 if (inode->i_size - range.start > range.len) {
3690 end_addr = range.start + range.len;
3691 } else {
3692 end_addr = range.len == (u64)-1 ?
3693 sbi->sb->s_maxbytes : inode->i_size;
3694 to_end = true;
3695 }
3696
3697 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3698 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3699 ret = -EINVAL;
3700 goto err;
3701 }
3702
3703 index = F2FS_BYTES_TO_BLK(range.start);
3704 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3705
3706 ret = f2fs_convert_inline_inode(inode);
3707 if (ret)
3708 goto err;
3709
e4544b63 3710 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 3711 filemap_invalidate_lock(mapping);
9af84648
DJ		 3712
3713 ret = filemap_write_and_wait_range(mapping, range.start,
3714 to_end ? LLONG_MAX : end_addr - 1);
3715 if (ret)
3716 goto out;
3717
3718 truncate_inode_pages_range(mapping, range.start,
3719 to_end ? -1 : end_addr - 1);
3720
3721 while (index < pg_end) {
3722 struct dnode_of_data dn;
3723 pgoff_t end_offset, count;
3724 int i;
3725
3726 set_new_dnode(&dn, inode, NULL, NULL, 0);
3727 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3728 if (ret) {
3729 if (ret == -ENOENT) {
3730 index = f2fs_get_next_page_offset(&dn, index);
3731 continue;
3732 }
3733 goto out;
3734 }
3735
3736 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3737 count = min(end_offset - dn.ofs_in_node, pg_end - index);
3738 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3739 struct block_device *cur_bdev;
3740 block_t blkaddr = f2fs_data_blkaddr(&dn);
3741
3742 if (!__is_valid_data_blkaddr(blkaddr))
3743 continue;
3744
3745 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3746 DATA_GENERIC_ENHANCE)) {
3747 ret = -EFSCORRUPTED;
3748 f2fs_put_dnode(&dn);
3749 goto out;
3750 }
3751
3752 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3753 if (f2fs_is_multi_device(sbi)) {
3754 int di = f2fs_target_device_index(sbi, blkaddr);
3755
3756 blkaddr -= FDEV(di).start_blk;
3757 }
3758
3759 if (len) {
3760 if (prev_bdev == cur_bdev &&
3761 index == prev_index + len &&
3762 blkaddr == prev_block + len) {
3763 len++;
3764 } else {
3765 ret = f2fs_secure_erase(prev_bdev,
3766 inode, prev_index, prev_block,
3767 len, range.flags);
3768 if (ret) {
3769 f2fs_put_dnode(&dn);
3770 goto out;
3771 }
3772
3773 len = 0;
3774 }
3775 }
3776
3777 if (!len) {
3778 prev_bdev = cur_bdev;
3779 prev_index = index;
3780 prev_block = blkaddr;
3781 len = 1;
3782 }
3783 }
3784
3785 f2fs_put_dnode(&dn);
3786
3787 if (fatal_signal_pending(current)) {
3788 ret = -EINTR;
3789 goto out;
3790 }
3791 cond_resched();
3792 }
3793
3794 if (len)
3795 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3796 prev_block, len, range.flags);
3797out:
edc6d01b 3798 filemap_invalidate_unlock(mapping);
e4544b63 3799 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
9af84648
DJ		 3800err:
3801 inode_unlock(inode);
3802 file_end_write(filp);
3803
3804 return ret;
3805}
3806
9e2a5f8c 3807static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
52656e6c 3808{
9e2a5f8c
DJ		 3809 struct inode *inode = file_inode(filp);
3810 struct f2fs_comp_option option;
3811
3812 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3813 return -EOPNOTSUPP;
3814
3815 inode_lock_shared(inode);
3816
3817 if (!f2fs_compressed_file(inode)) {
3818 inode_unlock_shared(inode);
3819 return -ENODATA;
3820 }
3821
3822 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3823 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3824
3825 inode_unlock_shared(inode);
3826
3827 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3828 sizeof(option)))
3829 return -EFAULT;
3830
3831 return 0;
3832}
3833
e1e8debe
DJ		 3834static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3835{
3836 struct inode *inode = file_inode(filp);
3837 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3838 struct f2fs_comp_option option;
3839 int ret = 0;
1f227a3e 3840
e1e8debe
DJ		 3841 if (!f2fs_sb_has_compression(sbi))
3842 return -EOPNOTSUPP;
3843
3844 if (!(filp->f_mode & FMODE_WRITE))
3845 return -EBADF;
3846
3847 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3848 sizeof(option)))
3849 return -EFAULT;
3850
3851 if (!f2fs_compressed_file(inode) ||
3852 option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3853 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3854 option.algorithm >= COMPRESS_MAX)
3855 return -EINVAL;
3856
3857 file_start_write(filp);
3858 inode_lock(inode);
3859
3860 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3861 ret = -EBUSY;
3862 goto out;
3863 }
3864
3865 if (inode->i_size != 0) {
3866 ret = -EFBIG;
3867 goto out;
3868 }
3869
3870 F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3871 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3872 F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3873 f2fs_mark_inode_dirty_sync(inode, true);
3874
3875 if (!f2fs_is_compress_backend_ready(inode))
3876 f2fs_warn(sbi, "compression algorithm is successfully set, "
3877 "but current kernel doesn't support this algorithm.");
3878out:
3879 inode_unlock(inode);
3880 file_end_write(filp);
3881
3882 return ret;
3883}
3884
5fdb322f
DJ		 3885static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3886{
fcd9ae4f 3887 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
5fdb322f
DJ		 3888 struct address_space *mapping = inode->i_mapping;
3889 struct page *page;
3890 pgoff_t redirty_idx = page_idx;
3891 int i, page_len = 0, ret = 0;
3892
3893 page_cache_ra_unbounded(&ractl, len, 0);
3894
3895 for (i = 0; i < len; i++, page_idx++) {
3896 page = read_cache_page(mapping, page_idx, NULL, NULL);
3897 if (IS_ERR(page)) {
3898 ret = PTR_ERR(page);
3899 break;
3900 }
3901 page_len++;
3902 }
3903
3904 for (i = 0; i < page_len; i++, redirty_idx++) {
3905 page = find_lock_page(mapping, redirty_idx);
df0736d7
DJ		 3906 if (!page) {
3907 ret = -ENOMEM;
3908 break;
3909 }
5fdb322f
DJ		 3910 set_page_dirty(page);
3911 f2fs_put_page(page, 1);
3912 f2fs_put_page(page, 0);
3913 }
3914
3915 return ret;
3916}
3917
3918static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3919{
3920 struct inode *inode = file_inode(filp);
3921 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3922 struct f2fs_inode_info *fi = F2FS_I(inode);
3923 pgoff_t page_idx = 0, last_idx;
3924 unsigned int blk_per_seg = sbi->blocks_per_seg;
054cb289 3925 int cluster_size = fi->i_cluster_size;
5fdb322f
DJ		 3926 int count, ret;
3927
3928 if (!f2fs_sb_has_compression(sbi) ||
3929 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3930 return -EOPNOTSUPP;
3931
3932 if (!(filp->f_mode & FMODE_WRITE))
3933 return -EBADF;
3934
3935 if (!f2fs_compressed_file(inode))
3936 return -EINVAL;
3937
3938 f2fs_balance_fs(F2FS_I_SB(inode), true);
3939
3940 file_start_write(filp);
3941 inode_lock(inode);
3942
3943 if (!f2fs_is_compress_backend_ready(inode)) {
3944 ret = -EOPNOTSUPP;
3945 goto out;
3946 }
3947
5fdb322f
DJ		 3948 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3949 if (ret)
3950 goto out;
3951
3952 if (!atomic_read(&fi->i_compr_blocks))
3953 goto out;
3954
3955 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3956
3957 count = last_idx - page_idx;
3958 while (count) {
3959 int len = min(cluster_size, count);
3960
3961 ret = redirty_blocks(inode, page_idx, len);
3962 if (ret < 0)
3963 break;
3964
3965 if (get_dirty_pages(inode) >= blk_per_seg)
3966 filemap_fdatawrite(inode->i_mapping);
3967
3968 count -= len;
3969 page_idx += len;
3970 }
3971
3972 if (!ret)
3973 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
3974 LLONG_MAX);
3975
3976 if (ret)
833dcd35
JP		 3977 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
3978 __func__, ret);
5fdb322f
DJ		 3979out:
3980 inode_unlock(inode);
3981 file_end_write(filp);
3982
3983 return ret;
3984}
3985
3986static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
3987{
3988 struct inode *inode = file_inode(filp);
3989 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3990 pgoff_t page_idx = 0, last_idx;
3991 unsigned int blk_per_seg = sbi->blocks_per_seg;
3992 int cluster_size = F2FS_I(inode)->i_cluster_size;
3993 int count, ret;
3994
3995 if (!f2fs_sb_has_compression(sbi) ||
3996 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3997 return -EOPNOTSUPP;
3998
3999 if (!(filp->f_mode & FMODE_WRITE))
4000 return -EBADF;
4001
4002 if (!f2fs_compressed_file(inode))
4003 return -EINVAL;
4004
4005 f2fs_balance_fs(F2FS_I_SB(inode), true);
4006
4007 file_start_write(filp);
4008 inode_lock(inode);
4009
4010 if (!f2fs_is_compress_backend_ready(inode)) {
4011 ret = -EOPNOTSUPP;
4012 goto out;
4013 }
4014
5fdb322f
DJ		 4015 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4016 if (ret)
4017 goto out;
4018
4019 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4020
4021 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4022
4023 count = last_idx - page_idx;
4024 while (count) {
4025 int len = min(cluster_size, count);
4026
4027 ret = redirty_blocks(inode, page_idx, len);
4028 if (ret < 0)
4029 break;
4030
4031 if (get_dirty_pages(inode) >= blk_per_seg)
4032 filemap_fdatawrite(inode->i_mapping);
4033
4034 count -= len;
4035 page_idx += len;
4036 }
4037
4038 if (!ret)
4039 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4040 LLONG_MAX);
4041
4042 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4043
4044 if (ret)
833dcd35
JP		 4045 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4046 __func__, ret);
5fdb322f
DJ		 4047out:
4048 inode_unlock(inode);
4049 file_end_write(filp);
4050
4051 return ret;
4052}
4053
34178b1b 4054static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
52656e6c
JK		 4055{
4056 switch (cmd) {
3357af8f 4057 case FS_IOC_GETVERSION:
d49f3e89 4058 return f2fs_ioc_getversion(filp, arg);
88b88a66
JK		 4059 case F2FS_IOC_START_ATOMIC_WRITE:
4060 return f2fs_ioc_start_atomic_write(filp);
4061 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4062 return f2fs_ioc_commit_atomic_write(filp);
02a1335f 4063 case F2FS_IOC_START_VOLATILE_WRITE:
1e84371f 4064 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
1e84371f 4065 case F2FS_IOC_ABORT_VOLATILE_WRITE:
7bc155fe 4066 return -EOPNOTSUPP;
1abff93d
JK		 4067 case F2FS_IOC_SHUTDOWN:
4068 return f2fs_ioc_shutdown(filp, arg);
52656e6c
JK		 4069 case FITRIM:
4070 return f2fs_ioc_fitrim(filp, arg);
3357af8f 4071 case FS_IOC_SET_ENCRYPTION_POLICY:
f424f664 4072 return f2fs_ioc_set_encryption_policy(filp, arg);
3357af8f 4073 case FS_IOC_GET_ENCRYPTION_POLICY:
f424f664 4074 return f2fs_ioc_get_encryption_policy(filp, arg);
3357af8f 4075 case FS_IOC_GET_ENCRYPTION_PWSALT:
f424f664 4076 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
8ce589c7
EB		 4077 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4078 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4079 case FS_IOC_ADD_ENCRYPTION_KEY:
4080 return f2fs_ioc_add_encryption_key(filp, arg);
4081 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4082 return f2fs_ioc_remove_encryption_key(filp, arg);
4083 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4084 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4085 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4086 return f2fs_ioc_get_encryption_key_status(filp, arg);
ee446e1a
EB		 4087 case FS_IOC_GET_ENCRYPTION_NONCE:
4088 return f2fs_ioc_get_encryption_nonce(filp, arg);
c1c1b583
CY		 4089 case F2FS_IOC_GARBAGE_COLLECT:
4090 return f2fs_ioc_gc(filp, arg);
34dc77ad
JK		 4091 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4092 return f2fs_ioc_gc_range(filp, arg);
456b88e4 4093 case F2FS_IOC_WRITE_CHECKPOINT:
059c0648 4094 return f2fs_ioc_write_checkpoint(filp, arg);
d323d005
CY		 4095 case F2FS_IOC_DEFRAGMENT:
4096 return f2fs_ioc_defragment(filp, arg);
4dd6f977
JK		 4097 case F2FS_IOC_MOVE_RANGE:
4098 return f2fs_ioc_move_range(filp, arg);
e066b83c
JK		 4099 case F2FS_IOC_FLUSH_DEVICE:
4100 return f2fs_ioc_flush_device(filp, arg);
e65ef207
JK		 4101 case F2FS_IOC_GET_FEATURES:
4102 return f2fs_ioc_get_features(filp, arg);
1ad71a27
JK		 4103 case F2FS_IOC_GET_PIN_FILE:
4104 return f2fs_ioc_get_pin_file(filp, arg);
4105 case F2FS_IOC_SET_PIN_FILE:
4106 return f2fs_ioc_set_pin_file(filp, arg);
c4020b2d
CY		 4107 case F2FS_IOC_PRECACHE_EXTENTS:
4108 return f2fs_ioc_precache_extents(filp, arg);
04f0b2ea
QS		 4109 case F2FS_IOC_RESIZE_FS:
4110 return f2fs_ioc_resize_fs(filp, arg);
95ae251f
EB		 4111 case FS_IOC_ENABLE_VERITY:
4112 return f2fs_ioc_enable_verity(filp, arg);
4113 case FS_IOC_MEASURE_VERITY:
4114 return f2fs_ioc_measure_verity(filp, arg);
e17fe657
EB		 4115 case FS_IOC_READ_VERITY_METADATA:
4116 return f2fs_ioc_read_verity_metadata(filp, arg);
3357af8f
EB		 4117 case FS_IOC_GETFSLABEL:
4118 return f2fs_ioc_getfslabel(filp, arg);
4119 case FS_IOC_SETFSLABEL:
4120 return f2fs_ioc_setfslabel(filp, arg);
439dfb10
CY		 4121 case F2FS_IOC_GET_COMPRESS_BLOCKS:
4122 return f2fs_get_compress_blocks(filp, arg);
ef8d563f
CY		 4123 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4124 return f2fs_release_compress_blocks(filp, arg);
c75488fb
CY		 4125 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4126 return f2fs_reserve_compress_blocks(filp, arg);
9af84648
DJ		 4127 case F2FS_IOC_SEC_TRIM_FILE:
4128 return f2fs_sec_trim_file(filp, arg);
9e2a5f8c
DJ		 4129 case F2FS_IOC_GET_COMPRESS_OPTION:
4130 return f2fs_ioc_get_compress_option(filp, arg);
e1e8debe
DJ		 4131 case F2FS_IOC_SET_COMPRESS_OPTION:
4132 return f2fs_ioc_set_compress_option(filp, arg);
5fdb322f
DJ		 4133 case F2FS_IOC_DECOMPRESS_FILE:
4134 return f2fs_ioc_decompress_file(filp, arg);
4135 case F2FS_IOC_COMPRESS_FILE:
4136 return f2fs_ioc_compress_file(filp, arg);
fbfa2cc5
JK		 4137 default:
4138 return -ENOTTY;
4139 }
4140}
4141
34178b1b
CY		 4142long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4143{
4144 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4145 return -EIO;
4146 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4147 return -ENOSPC;
4148
4149 return __f2fs_ioctl(filp, cmd, arg);
4150}
4151
a1e09b03
EB		 4152/*
4153 * Return %true if the given read or write request should use direct I/O, or
4154 * %false if it should use buffered I/O.
4155 */
4156static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4157 struct iov_iter *iter)
4158{
4159 unsigned int align;
4160
4161 if (!(iocb->ki_flags & IOCB_DIRECT))
4162 return false;
4163
4164 if (f2fs_force_buffered_io(inode, iocb, iter))
4165 return false;
4166
4167 /*
4168 * Direct I/O not aligned to the disk's logical_block_size will be
4169 * attempted, but will fail with -EINVAL.
4170 *
4171 * f2fs additionally requires that direct I/O be aligned to the
4172 * filesystem block size, which is often a stricter requirement.
4173 * However, f2fs traditionally falls back to buffered I/O on requests
4174 * that are logical_block_size-aligned but not fs-block aligned.
4175 *
4176 * The below logic implements this behavior.
4177 */
4178 align = iocb->ki_pos | iov_iter_alignment(iter);
4179 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4180 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4181 return false;
4182
4183 return true;
4184}
4185
4186static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4187 unsigned int flags)
4188{
4189 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4190
4191 dec_page_count(sbi, F2FS_DIO_READ);
4192 if (error)
4193 return error;
4194 f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4195 return 0;
4196}
4197
4198static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4199 .end_io = f2fs_dio_read_end_io,
4200};
4201
4202static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4c8ff709
CY		 4203{
4204 struct file *file = iocb->ki_filp;
4205 struct inode *inode = file_inode(file);
a1e09b03
EB		 4206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4207 struct f2fs_inode_info *fi = F2FS_I(inode);
4208 const loff_t pos = iocb->ki_pos;
4209 const size_t count = iov_iter_count(to);
4210 struct iomap_dio *dio;
4211 ssize_t ret;
4212
4213 if (count == 0)
4214 return 0; /* skip atime update */
4215
bd984c03 4216 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
a1e09b03
EB		 4217
4218 if (iocb->ki_flags & IOCB_NOWAIT) {
e4544b63 4219 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
a1e09b03
EB		 4220 ret = -EAGAIN;
4221 goto out;
4222 }
4223 } else {
e4544b63 4224 f2fs_down_read(&fi->i_gc_rwsem[READ]);
a1e09b03
EB		 4225 }
4226
4227 /*
4228 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4229 * the higher-level function iomap_dio_rw() in order to ensure that the
4230 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4231 */
4232 inc_page_count(sbi, F2FS_DIO_READ);
4233 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4234 &f2fs_iomap_dio_read_ops, 0, 0);
4235 if (IS_ERR_OR_NULL(dio)) {
4236 ret = PTR_ERR_OR_ZERO(dio);
4237 if (ret != -EIOCBQUEUED)
4238 dec_page_count(sbi, F2FS_DIO_READ);
4239 } else {
4240 ret = iomap_dio_complete(dio);
4241 }
4242
e4544b63 4243 f2fs_up_read(&fi->i_gc_rwsem[READ]);
a1e09b03
EB		 4244
4245 file_accessed(file);
4246out:
4247 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4248 return ret;
4249}
4250
4251static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4252{
4253 struct inode *inode = file_inode(iocb->ki_filp);
c277f141 4254 const loff_t pos = iocb->ki_pos;
a1e09b03 4255 ssize_t ret;
4c8ff709
CY		 4256
4257 if (!f2fs_is_compress_backend_ready(inode))
4258 return -EOPNOTSUPP;
4259
c277f141
JK		 4260 if (trace_f2fs_dataread_start_enabled()) {
4261 char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4262 char *path;
4263
4264 if (!p)
4265 goto skip_read_trace;
4266
4267 path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4268 if (IS_ERR(path)) {
4269 kfree(p);
4270 goto skip_read_trace;
4271 }
8b83ac81 4272
c277f141
JK		 4273 trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4274 current->pid, path, current->comm);
4275 kfree(p);
4276 }
4277skip_read_trace:
4278 if (f2fs_should_use_dio(inode, iocb, to)) {
4279 ret = f2fs_dio_read_iter(iocb, to);
4280 } else {
4281 ret = filemap_read(iocb, to, 0);
4282 if (ret > 0)
4283 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4284 }
4285 if (trace_f2fs_dataread_end_enabled())
4286 trace_f2fs_dataread_end(inode, pos, ret);
8b83ac81 4287 return ret;
4c8ff709
CY		 4288}
4289
a1e09b03
EB		 4290static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4291{
4292 struct file *file = iocb->ki_filp;
4293 struct inode *inode = file_inode(file);
4294 ssize_t count;
4295 int err;
4296
4297 if (IS_IMMUTABLE(inode))
4298 return -EPERM;
4299
4300 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4301 return -EPERM;
4302
4303 count = generic_write_checks(iocb, from);
4304 if (count <= 0)
4305 return count;
4306
4307 err = file_modified(file);
4308 if (err)
4309 return err;
4310 return count;
4311}
4312
3d697a4a
EB		 4313/*
4314 * Preallocate blocks for a write request, if it is possible and helpful to do
4315 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4316 * blocks were preallocated, or a negative errno value if something went
4317 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4318 * requested blocks (not just some of them) have been allocated.
4319 */
a1e09b03
EB		 4320static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4321 bool dio)
3d697a4a
EB		 4322{
4323 struct inode *inode = file_inode(iocb->ki_filp);
4324 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4325 const loff_t pos = iocb->ki_pos;
4326 const size_t count = iov_iter_count(iter);
4327 struct f2fs_map_blocks map = {};
3d697a4a
EB		 4328 int flag;
4329 int ret;
4330
4331 /* If it will be an out-of-place direct write, don't bother. */
4332 if (dio && f2fs_lfs_mode(sbi))
4333 return 0;
d4dd19ec
JK		 4334 /*
4335 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4336 * buffered IO, if DIO meets any holes.
4337 */
4338 if (dio && i_size_read(inode) &&
4339 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4340 return 0;
3d697a4a
EB		 4341
4342 /* No-wait I/O can't allocate blocks. */
4343 if (iocb->ki_flags & IOCB_NOWAIT)
4344 return 0;
4345
4346 /* If it will be a short write, don't bother. */
4347 if (fault_in_iov_iter_readable(iter, count))
4348 return 0;
4349
4350 if (f2fs_has_inline_data(inode)) {
4351 /* If the data will fit inline, don't bother. */
4352 if (pos + count <= MAX_INLINE_DATA(inode))
4353 return 0;
4354 ret = f2fs_convert_inline_inode(inode);
4355 if (ret)
4356 return ret;
4357 }
4358
4359 /* Do not preallocate blocks that will be written partially in 4KB. */
4360 map.m_lblk = F2FS_BLK_ALIGN(pos);
4361 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4362 if (map.m_len > map.m_lblk)
4363 map.m_len -= map.m_lblk;
4364 else
4365 map.m_len = 0;
4366 map.m_may_create = true;
4367 if (dio) {
4368 map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4369 flag = F2FS_GET_BLOCK_PRE_DIO;
4370 } else {
4371 map.m_seg_type = NO_CHECK_TYPE;
4372 flag = F2FS_GET_BLOCK_PRE_AIO;
4373 }
4374
4375 ret = f2fs_map_blocks(inode, &map, 1, flag);
d4dd19ec
JK		 4376 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4377 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
3d697a4a
EB		 4378 return ret;
4379 if (ret == 0)
4380 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4381 return map.m_len;
4382}
4383
a1e09b03
EB		 4384static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4385 struct iov_iter *from)
fcc85a4d 4386{
b439b103
JK		 4387 struct file *file = iocb->ki_filp;
4388 struct inode *inode = file_inode(file);
a1e09b03
EB		 4389 ssize_t ret;
4390
4391 if (iocb->ki_flags & IOCB_NOWAIT)
4392 return -EOPNOTSUPP;
4393
4394 current->backing_dev_info = inode_to_bdi(inode);
800ba295 4395 ret = generic_perform_write(iocb, from);
a1e09b03
EB		 4396 current->backing_dev_info = NULL;
4397
4398 if (ret > 0) {
4399 iocb->ki_pos += ret;
4400 f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4401 }
4402 return ret;
4403}
4404
4405static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4406 unsigned int flags)
4407{
4408 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4409
4410 dec_page_count(sbi, F2FS_DIO_WRITE);
4411 if (error)
4412 return error;
4413 f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4414 return 0;
4415}
4416
4417static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4418 .end_io = f2fs_dio_write_end_io,
4419};
4420
4421static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4422 bool *may_need_sync)
4423{
4424 struct file *file = iocb->ki_filp;
4425 struct inode *inode = file_inode(file);
4426 struct f2fs_inode_info *fi = F2FS_I(inode);
4427 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4428 const bool do_opu = f2fs_lfs_mode(sbi);
a1e09b03
EB		 4429 const loff_t pos = iocb->ki_pos;
4430 const ssize_t count = iov_iter_count(from);
a1e09b03
EB		 4431 unsigned int dio_flags;
4432 struct iomap_dio *dio;
4433 ssize_t ret;
4434
bd984c03 4435 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
a1e09b03
EB		 4436
4437 if (iocb->ki_flags & IOCB_NOWAIT) {
4438 /* f2fs_convert_inline_inode() and block allocation can block */
4439 if (f2fs_has_inline_data(inode) ||
4440 !f2fs_overwrite_io(inode, pos, count)) {
4441 ret = -EAGAIN;
4442 goto out;
4443 }
4444
e4544b63 4445 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
a1e09b03
EB		 4446 ret = -EAGAIN;
4447 goto out;
4448 }
e4544b63
TM		 4449 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4450 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03
EB		 4451 ret = -EAGAIN;
4452 goto out;
4453 }
4454 } else {
4455 ret = f2fs_convert_inline_inode(inode);
4456 if (ret)
4457 goto out;
4458
e4544b63 4459 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03 4460 if (do_opu)
e4544b63 4461 f2fs_down_read(&fi->i_gc_rwsem[READ]);
a1e09b03 4462 }
a1e09b03
EB		 4463
4464 /*
4465 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4466 * the higher-level function iomap_dio_rw() in order to ensure that the
4467 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4468 */
4469 inc_page_count(sbi, F2FS_DIO_WRITE);
4470 dio_flags = 0;
4471 if (pos + count > inode->i_size)
4472 dio_flags |= IOMAP_DIO_FORCE_WAIT;
4473 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4474 &f2fs_iomap_dio_write_ops, dio_flags, 0);
4475 if (IS_ERR_OR_NULL(dio)) {
4476 ret = PTR_ERR_OR_ZERO(dio);
4477 if (ret == -ENOTBLK)
4478 ret = 0;
4479 if (ret != -EIOCBQUEUED)
4480 dec_page_count(sbi, F2FS_DIO_WRITE);
4481 } else {
4482 ret = iomap_dio_complete(dio);
4483 }
4484
a1e09b03 4485 if (do_opu)
e4544b63
TM		 4486 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4487 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03
EB		 4488
4489 if (ret < 0)
4490 goto out;
4491 if (pos + ret > inode->i_size)
4492 f2fs_i_size_write(inode, pos + ret);
4493 if (!do_opu)
4494 set_inode_flag(inode, FI_UPDATE_WRITE);
4495
4496 if (iov_iter_count(from)) {
4497 ssize_t ret2;
4498 loff_t bufio_start_pos = iocb->ki_pos;
4499
4500 /*
4501 * The direct write was partial, so we need to fall back to a
4502 * buffered write for the remainder.
4503 */
4504
4505 ret2 = f2fs_buffered_write_iter(iocb, from);
4506 if (iov_iter_count(from))
4507 f2fs_write_failed(inode, iocb->ki_pos);
4508 if (ret2 < 0)
4509 goto out;
4510
4511 /*
4512 * Ensure that the pagecache pages are written to disk and
4513 * invalidated to preserve the expected O_DIRECT semantics.
4514 */
4515 if (ret2 > 0) {
4516 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4517
4518 ret += ret2;
4519
4520 ret2 = filemap_write_and_wait_range(file->f_mapping,
4521 bufio_start_pos,
4522 bufio_end_pos);
4523 if (ret2 < 0)
4524 goto out;
4525 invalidate_mapping_pages(file->f_mapping,
4526 bufio_start_pos >> PAGE_SHIFT,
4527 bufio_end_pos >> PAGE_SHIFT);
4528 }
4529 } else {
4530 /* iomap_dio_rw() already handled the generic_write_sync(). */
4531 *may_need_sync = false;
4532 }
4533out:
4534 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4535 return ret;
4536}
4537
4538static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4539{
4540 struct inode *inode = file_inode(iocb->ki_filp);
ccf7cf92
EB		 4541 const loff_t orig_pos = iocb->ki_pos;
4542 const size_t orig_count = iov_iter_count(from);
3d697a4a 4543 loff_t target_size;
a1e09b03
EB		 4544 bool dio;
4545 bool may_need_sync = true;
3d697a4a 4546 int preallocated;
b439b103 4547 ssize_t ret;
fcc85a4d 4548
126ce721
CY		 4549 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4550 ret = -EIO;
4551 goto out;
4552 }
1f227a3e 4553
7bd29358
CY		 4554 if (!f2fs_is_compress_backend_ready(inode)) {
4555 ret = -EOPNOTSUPP;
4556 goto out;
4557 }
4c8ff709 4558
cb8434f1
GR		 4559 if (iocb->ki_flags & IOCB_NOWAIT) {
4560 if (!inode_trylock(inode)) {
126ce721
CY		 4561 ret = -EAGAIN;
4562 goto out;
4563 }
cb8434f1 4564 } else {
b91050a8
HL		 4565 inode_lock(inode);
4566 }
4567
a1e09b03 4568 ret = f2fs_write_checks(iocb, from);
b31bf0f9
EB		 4569 if (ret <= 0)
4570 goto out_unlock;
4571
a1e09b03
EB		 4572 /* Determine whether we will do a direct write or a buffered write. */
4573 dio = f2fs_should_use_dio(inode, iocb, from);
3d697a4a 4574
b31bf0f9
EB		 4575 /* Possibly preallocate the blocks for the write. */
4576 target_size = iocb->ki_pos + iov_iter_count(from);
a1e09b03 4577 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
c277f141 4578 if (preallocated < 0) {
b31bf0f9 4579 ret = preallocated;
c277f141
JK		 4580 } else {
4581 if (trace_f2fs_datawrite_start_enabled()) {
4582 char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4583 PATH_MAX, GFP_KERNEL);
4584 char *path;
4585
4586 if (!p)
4587 goto skip_write_trace;
4588 path = dentry_path_raw(file_dentry(iocb->ki_filp),
4589 p, PATH_MAX);
4590 if (IS_ERR(path)) {
4591 kfree(p);
4592 goto skip_write_trace;
4593 }
4594 trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4595 current->pid, path, current->comm);
4596 kfree(p);
4597 }
4598skip_write_trace:
a1e09b03
EB		 4599 /* Do the actual write. */
4600 ret = dio ?
4601 f2fs_dio_write_iter(iocb, from, &may_need_sync):
4602 f2fs_buffered_write_iter(iocb, from);
dc7a10dd 4603
c277f141
JK		 4604 if (trace_f2fs_datawrite_end_enabled())
4605 trace_f2fs_datawrite_end(inode, orig_pos, ret);
4606 }
4607
b31bf0f9 4608 /* Don't leave any preallocated blocks around past i_size. */
d4dd19ec 4609 if (preallocated && i_size_read(inode) < target_size) {
e4544b63 4610 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
b31bf0f9 4611 filemap_invalidate_lock(inode->i_mapping);
d4dd19ec
JK		 4612 if (!f2fs_truncate(inode))
4613 file_dont_truncate(inode);
b31bf0f9 4614 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 4615 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
d4dd19ec
JK		 4616 } else {
4617 file_dont_truncate(inode);
b439b103 4618 }
b31bf0f9
EB		 4619
4620 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
3d697a4a 4621out_unlock:
b439b103 4622 inode_unlock(inode);
126ce721 4623out:
ccf7cf92 4624 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
a1e09b03 4625 if (ret > 0 && may_need_sync)
e2592217 4626 ret = generic_write_sync(iocb, ret);
b439b103 4627 return ret;
fcc85a4d
JK		 4628}
4629
0f6b56ec
DJ		 4630static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4631 int advice)
4632{
0f6b56ec
DJ		 4633 struct address_space *mapping;
4634 struct backing_dev_info *bdi;
e64347ae
FC		 4635 struct inode *inode = file_inode(filp);
4636 int err;
0f6b56ec
DJ		 4637
4638 if (advice == POSIX_FADV_SEQUENTIAL) {
0f6b56ec
DJ		 4639 if (S_ISFIFO(inode->i_mode))
4640 return -ESPIPE;
4641
4642 mapping = filp->f_mapping;
4643 if (!mapping || len < 0)
4644 return -EINVAL;
4645
4646 bdi = inode_to_bdi(mapping->host);
4647 filp->f_ra.ra_pages = bdi->ra_pages *
4648 F2FS_I_SB(inode)->seq_file_ra_mul;
4649 spin_lock(&filp->f_lock);
4650 filp->f_mode &= ~FMODE_RANDOM;
4651 spin_unlock(&filp->f_lock);
4652 return 0;
4653 }
4654
e64347ae
FC		 4655 err = generic_fadvise(filp, offset, len, advice);
4656 if (!err && advice == POSIX_FADV_DONTNEED &&
4657 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4658 f2fs_compressed_file(inode))
4659 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4660
4661 return err;
0f6b56ec
DJ		 4662}
4663
e9750824 4664#ifdef CONFIG_COMPAT
34178b1b
CY		 4665struct compat_f2fs_gc_range {
4666 u32 sync;
4667 compat_u64 start;
4668 compat_u64 len;
4669};
4670#define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
4671 struct compat_f2fs_gc_range)
4672
4673static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4674{
4675 struct compat_f2fs_gc_range __user *urange;
4676 struct f2fs_gc_range range;
4677 int err;
4678
4679 urange = compat_ptr(arg);
4680 err = get_user(range.sync, &urange->sync);
4681 err |= get_user(range.start, &urange->start);
4682 err |= get_user(range.len, &urange->len);
4683 if (err)
4684 return -EFAULT;
4685
4686 return __f2fs_ioc_gc_range(file, &range);
4687}
4688
4689struct compat_f2fs_move_range {
4690 u32 dst_fd;
4691 compat_u64 pos_in;
4692 compat_u64 pos_out;
4693 compat_u64 len;
4694};
4695#define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
4696 struct compat_f2fs_move_range)
4697
4698static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4699{
4700 struct compat_f2fs_move_range __user *urange;
4701 struct f2fs_move_range range;
4702 int err;
4703
4704 urange = compat_ptr(arg);
4705 err = get_user(range.dst_fd, &urange->dst_fd);
4706 err |= get_user(range.pos_in, &urange->pos_in);
4707 err |= get_user(range.pos_out, &urange->pos_out);
4708 err |= get_user(range.len, &urange->len);
4709 if (err)
4710 return -EFAULT;
4711
4712 return __f2fs_ioc_move_range(file, &range);
4713}
4714
e9750824
NJ		 4715long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4716{
34178b1b
CY		 4717 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4718 return -EIO;
4719 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4720 return -ENOSPC;
4721
e9750824 4722 switch (cmd) {
3357af8f
EB		 4723 case FS_IOC32_GETVERSION:
4724 cmd = FS_IOC_GETVERSION;
04ef4b62 4725 break;
34178b1b
CY		 4726 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4727 return f2fs_compat_ioc_gc_range(file, arg);
4728 case F2FS_IOC32_MOVE_RANGE:
4729 return f2fs_compat_ioc_move_range(file, arg);
04ef4b62
CY		 4730 case F2FS_IOC_START_ATOMIC_WRITE:
4731 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4732 case F2FS_IOC_START_VOLATILE_WRITE:
4733 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4734 case F2FS_IOC_ABORT_VOLATILE_WRITE:
4735 case F2FS_IOC_SHUTDOWN:
314999dc 4736 case FITRIM:
3357af8f
EB		 4737 case FS_IOC_SET_ENCRYPTION_POLICY:
4738 case FS_IOC_GET_ENCRYPTION_PWSALT:
4739 case FS_IOC_GET_ENCRYPTION_POLICY:
8ce589c7
EB		 4740 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4741 case FS_IOC_ADD_ENCRYPTION_KEY:
4742 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4743 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4744 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
ee446e1a 4745 case FS_IOC_GET_ENCRYPTION_NONCE:
04ef4b62
CY		 4746 case F2FS_IOC_GARBAGE_COLLECT:
4747 case F2FS_IOC_WRITE_CHECKPOINT:
4748 case F2FS_IOC_DEFRAGMENT:
e066b83c 4749 case F2FS_IOC_FLUSH_DEVICE:
e65ef207 4750 case F2FS_IOC_GET_FEATURES:
1ad71a27
JK		 4751 case F2FS_IOC_GET_PIN_FILE:
4752 case F2FS_IOC_SET_PIN_FILE:
c4020b2d 4753 case F2FS_IOC_PRECACHE_EXTENTS:
04f0b2ea 4754 case F2FS_IOC_RESIZE_FS:
95ae251f
EB		 4755 case FS_IOC_ENABLE_VERITY:
4756 case FS_IOC_MEASURE_VERITY:
e17fe657 4757 case FS_IOC_READ_VERITY_METADATA:
3357af8f
EB		 4758 case FS_IOC_GETFSLABEL:
4759 case FS_IOC_SETFSLABEL:
439dfb10 4760 case F2FS_IOC_GET_COMPRESS_BLOCKS:
ef8d563f 4761 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
c75488fb 4762 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
9af84648 4763 case F2FS_IOC_SEC_TRIM_FILE:
9e2a5f8c 4764 case F2FS_IOC_GET_COMPRESS_OPTION:
e1e8debe 4765 case F2FS_IOC_SET_COMPRESS_OPTION:
5fdb322f
DJ		 4766 case F2FS_IOC_DECOMPRESS_FILE:
4767 case F2FS_IOC_COMPRESS_FILE:
4dd6f977 4768 break;
e9750824
NJ		 4769 default:
4770 return -ENOIOCTLCMD;
4771 }
34178b1b 4772 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
e9750824
NJ		 4773}
4774#endif
4775
fbfa2cc5 4776const struct file_operations f2fs_file_operations = {
267378d4 4777 .llseek = f2fs_llseek,
4c8ff709 4778 .read_iter = f2fs_file_read_iter,
fcc85a4d
JK		 4779 .write_iter = f2fs_file_write_iter,
4780 .open = f2fs_file_open,
12662234 4781 .release = f2fs_release_file,
fbfa2cc5 4782 .mmap = f2fs_file_mmap,
7a10f017 4783 .flush = f2fs_file_flush,
fbfa2cc5
JK		 4784 .fsync = f2fs_sync_file,
4785 .fallocate = f2fs_fallocate,
4786 .unlocked_ioctl = f2fs_ioctl,
e9750824
NJ		 4787#ifdef CONFIG_COMPAT
4788 .compat_ioctl = f2fs_compat_ioctl,
4789#endif
fbfa2cc5 4790 .splice_read = generic_file_splice_read,
8d020765 4791 .splice_write = iter_file_splice_write,
0f6b56ec 4792 .fadvise = f2fs_file_fadvise,
fbfa2cc5 4793};
Linux 6.x block layer and io_uring tree(s)