projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'probes-fixes-v6.16-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6-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>
fbfa2cc5 11#include <linux/writeback.h>
ae51fb31 12#include <linux/blkdev.h>
fbfa2cc5
JK		 13#include <linux/falloc.h>
14#include <linux/types.h>
e9750824 15#include <linux/compat.h>
fbfa2cc5
JK		 16#include <linux/uaccess.h>
17#include <linux/mount.h>
7f7670fe 18#include <linux/pagevec.h>
dc91de78 19#include <linux/uio.h>
8da4b8c4 20#include <linux/uuid.h>
4dd6f977 21#include <linux/file.h>
4507847c 22#include <linux/nls.h>
9af84648 23#include <linux/sched/signal.h>
9b1bb01c 24#include <linux/fileattr.h>
0f6b56ec 25#include <linux/fadvise.h>
a1e09b03 26#include <linux/iomap.h>
fbfa2cc5
JK		 27
28#include "f2fs.h"
29#include "node.h"
30#include "segment.h"
31#include "xattr.h"
32#include "acl.h"
c1c1b583 33#include "gc.h"
52118743 34#include "iostat.h"
a2a4a7e4 35#include <trace/events/f2fs.h>
fa4320ce 36#include <uapi/linux/f2fs.h>
fbfa2cc5 37
ba8dac35
CY		 38static void f2fs_zero_post_eof_page(struct inode *inode, loff_t new_size)
39{
40 loff_t old_size = i_size_read(inode);
41
42 if (old_size >= new_size)
43 return;
44
45 /* zero or drop pages only in range of [old_size, new_size] */
46 truncate_pagecache(inode, old_size);
47}
48
ea4d479b 49static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
5a3a2d83
QS		 50{
51 struct inode *inode = file_inode(vmf->vma->vm_file);
eb70d5a6 52 vm_flags_t flags = vmf->vma->vm_flags;
ea4d479b 53 vm_fault_t ret;
5a3a2d83 54
ea4d479b 55 ret = filemap_fault(vmf);
bb34cc6c 56 if (ret & VM_FAULT_LOCKED)
34a23525
CY		 57 f2fs_update_iostat(F2FS_I_SB(inode), inode,
58 APP_MAPPED_READ_IO, F2FS_BLKSIZE);
8b83ac81 59
eb70d5a6 60 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret);
d7648343 61
ea4d479b 62 return ret;
5a3a2d83
QS		 63}
64
ea4d479b 65static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
fbfa2cc5 66{
aec57559 67 struct folio *folio = page_folio(vmf->page);
11bac800 68 struct inode *inode = file_inode(vmf->vma->vm_file);
4081363f 69 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
bdf03299 70 struct dnode_of_data dn;
3fdd89b4 71 bool need_alloc = !f2fs_is_pinned_file(inode);
4c8ff709 72 int err = 0;
87f3afd3 73 vm_fault_t ret;
fbfa2cc5 74
e0fcd015
CY		 75 if (unlikely(IS_IMMUTABLE(inode)))
76 return VM_FAULT_SIGBUS;
77
87f3afd3
CY		 78 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
79 err = -EIO;
80 goto out;
81 }
c6140415 82
1f227a3e
JK		 83 if (unlikely(f2fs_cp_error(sbi))) {
84 err = -EIO;
87f3afd3 85 goto out;
1f227a3e
JK		 86 }
87
00e09c0b
CY		 88 if (!f2fs_is_checkpoint_ready(sbi)) {
89 err = -ENOSPC;
87f3afd3 90 goto out;
00e09c0b 91 }
1f227a3e 92
c8e43d55
CY		 93 err = f2fs_convert_inline_inode(inode);
94 if (err)
87f3afd3 95 goto out;
c8e43d55 96
4c8ff709
CY		 97#ifdef CONFIG_F2FS_FS_COMPRESSION
98 if (f2fs_compressed_file(inode)) {
aec57559 99 int ret = f2fs_is_compressed_cluster(inode, folio->index);
4c8ff709
CY		 100
101 if (ret < 0) {
102 err = ret;
87f3afd3 103 goto out;
4c8ff709 104 } else if (ret) {
4c8ff709
CY		 105 need_alloc = false;
106 }
107 }
108#endif
bdf03299 109 /* should do out of any locked page */
4c8ff709
CY		 110 if (need_alloc)
111 f2fs_balance_fs(sbi, true);
bdf03299 112
fbfa2cc5 113 sb_start_pagefault(inode->i_sb);
b3d208f9
JK		 114
115 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
b067ba1f 116
ba8dac35
CY		 117 filemap_invalidate_lock(inode->i_mapping);
118 f2fs_zero_post_eof_page(inode, (folio->index + 1) << PAGE_SHIFT);
119 filemap_invalidate_unlock(inode->i_mapping);
120
11bac800 121 file_update_time(vmf->vma->vm_file);
edc6d01b 122 filemap_invalidate_lock_shared(inode->i_mapping);
ba8dac35 123
aec57559
CY		 124 folio_lock(folio);
125 if (unlikely(folio->mapping != inode->i_mapping ||
126 folio_pos(folio) > i_size_read(inode) ||
127 !folio_test_uptodate(folio))) {
128 folio_unlock(folio);
fbfa2cc5 129 err = -EFAULT;
5a3a2d83 130 goto out_sem;
fbfa2cc5
JK		 131 }
132
3fdd89b4 133 set_new_dnode(&dn, inode, NULL, NULL, 0);
4c8ff709
CY		 134 if (need_alloc) {
135 /* block allocation */
aec57559 136 err = f2fs_get_block_locked(&dn, folio->index);
3fdd89b4 137 } else {
aec57559 138 err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE);
06c7540f 139 f2fs_put_dnode(&dn);
3fdd89b4
DJ		 140 if (f2fs_is_pinned_file(inode) &&
141 !__is_valid_data_blkaddr(dn.data_blkaddr))
142 err = -EIO;
06c7540f 143 }
3fdd89b4 144
06c7540f 145 if (err) {
aec57559 146 folio_unlock(folio);
06c7540f 147 goto out_sem;
39a86958
CY		 148 }
149
46fd261c 150 f2fs_folio_wait_writeback(folio, DATA, false, true);
39a86958
CY		 151
152 /* wait for GCed page writeback via META_MAPPING */
153 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
154
fbfa2cc5
JK		 155 /*
156 * check to see if the page is mapped already (no holes)
157 */
aec57559 158 if (folio_test_mappedtodisk(folio))
39a86958 159 goto out_sem;
fbfa2cc5
JK		 160
161 /* page is wholly or partially inside EOF */
aec57559 162 if (((loff_t)(folio->index + 1) << PAGE_SHIFT) >
9edcdabf 163 i_size_read(inode)) {
193bea1d 164 loff_t offset;
f11e98bd 165
09cbfeaf 166 offset = i_size_read(inode) & ~PAGE_MASK;
aec57559 167 folio_zero_segment(folio, offset, folio_size(folio));
fbfa2cc5 168 }
aec57559 169 folio_mark_dirty(folio);
fbfa2cc5 170
34a23525 171 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE);
c75f2feb 172 f2fs_update_time(sbi, REQ_TIME);
b0af6d49 173
5a3a2d83 174out_sem:
edc6d01b 175 filemap_invalidate_unlock_shared(inode->i_mapping);
39a86958 176
fbfa2cc5 177 sb_end_pagefault(inode->i_sb);
87f3afd3
CY		 178out:
179 ret = vmf_fs_error(err);
180
aec57559 181 trace_f2fs_vm_page_mkwrite(inode, folio->index, vmf->vma->vm_flags, ret);
87f3afd3 182 return ret;
fbfa2cc5
JK		 183}
184
185static const struct vm_operations_struct f2fs_file_vm_ops = {
5a3a2d83 186 .fault = f2fs_filemap_fault,
f1820361 187 .map_pages = filemap_map_pages,
692bb55d 188 .page_mkwrite = f2fs_vm_page_mkwrite,
fbfa2cc5
JK		 189};
190
354a3399
JK		 191static int get_parent_ino(struct inode *inode, nid_t *pino)
192{
193 struct dentry *dentry;
194
84c9c2de
EB		 195 /*
196 * Make sure to get the non-deleted alias. The alias associated with
197 * the open file descriptor being fsync()'ed may be deleted already.
198 */
199 dentry = d_find_alias(inode);
354a3399
JK		 200 if (!dentry)
201 return 0;
202
f378ec4e 203 *pino = d_parent_ino(dentry);
f0947e5c 204 dput(dentry);
354a3399
JK		 205 return 1;
206}
207
a5fd5050 208static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
9d1589ef 209{
4081363f 210 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a5fd5050 211 enum cp_reason_type cp_reason = CP_NO_NEEDED;
9d1589ef 212
a5fd5050
CY		 213 if (!S_ISREG(inode->i_mode))
214 cp_reason = CP_NON_REGULAR;
4c8ff709
CY		 215 else if (f2fs_compressed_file(inode))
216 cp_reason = CP_COMPRESSED;
a5fd5050
CY		 217 else if (inode->i_nlink != 1)
218 cp_reason = CP_HARDLINK;
bbf156f7 219 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
a5fd5050 220 cp_reason = CP_SB_NEED_CP;
9d1589ef 221 else if (file_wrong_pino(inode))
a5fd5050 222 cp_reason = CP_WRONG_PINO;
4d57b86d 223 else if (!f2fs_space_for_roll_forward(sbi))
a5fd5050 224 cp_reason = CP_NO_SPC_ROLL;
4d57b86d 225 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
a5fd5050 226 cp_reason = CP_NODE_NEED_CP;
d5053a34 227 else if (test_opt(sbi, FASTBOOT))
a5fd5050 228 cp_reason = CP_FASTBOOT_MODE;
63189b78 229 else if (F2FS_OPTION(sbi).active_logs == 2)
a5fd5050 230 cp_reason = CP_SPEC_LOG_NUM;
63189b78 231 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
4d57b86d
CY		 232 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
233 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
234 TRANS_DIR_INO))
0a007b97 235 cp_reason = CP_RECOVER_DIR;
aaf8c0b9
CY		 236 else if (f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
237 XATTR_DIR_INO))
238 cp_reason = CP_XATTR_DIR;
9d1589ef 239
a5fd5050 240 return cp_reason;
9d1589ef
CY		 241}
242
9c7bb702
CL		 243static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
244{
ba13af45 245 struct folio *i = filemap_get_folio(NODE_MAPPING(sbi), ino);
9c7bb702
CL		 246 bool ret = false;
247 /* But we need to avoid that there are some inode updates */
ba13af45
MWO		 248 if ((!IS_ERR(i) && folio_test_dirty(i)) ||
249 f2fs_need_inode_block_update(sbi, ino))
9c7bb702 250 ret = true;
ba13af45 251 f2fs_folio_put(i, false);
9c7bb702
CL		 252 return ret;
253}
254
51455b19
CL		 255static void try_to_fix_pino(struct inode *inode)
256{
257 struct f2fs_inode_info *fi = F2FS_I(inode);
258 nid_t pino;
259
e4544b63 260 f2fs_down_write(&fi->i_sem);
51455b19
CL		 261 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
262 get_parent_ino(inode, &pino)) {
205b9822 263 f2fs_i_pino_write(inode, pino);
51455b19 264 file_got_pino(inode);
51455b19 265 }
e4544b63 266 f2fs_up_write(&fi->i_sem);
51455b19
CL		 267}
268
608514de
JK		 269static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
270 int datasync, bool atomic)
fbfa2cc5
JK		 271{
272 struct inode *inode = file->f_mapping->host;
4081363f 273 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2403c155 274 nid_t ino = inode->i_ino;
fbfa2cc5 275 int ret = 0;
a5fd5050 276 enum cp_reason_type cp_reason = 0;
fbfa2cc5 277 struct writeback_control wbc = {
c81bf1c8 278 .sync_mode = WB_SYNC_ALL,
fbfa2cc5 279 .nr_to_write = LONG_MAX,
fbfa2cc5 280 };
50fa53ec 281 unsigned int seq_id = 0;
fbfa2cc5 282
dddd3d65 283 if (unlikely(f2fs_readonly(inode->i_sb)))
1fa95b0b
NJ		 284 return 0;
285
a2a4a7e4 286 trace_f2fs_sync_file_enter(inode);
ea1aa12c 287
b61ac5b7
YH		 288 if (S_ISDIR(inode->i_mode))
289 goto go_write;
290
ea1aa12c 291 /* if fdatasync is triggered, let's do in-place-update */
c46a155b 292 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
91942321 293 set_inode_flag(inode, FI_NEED_IPU);
3b49c9a1 294 ret = file_write_and_wait_range(file, start, end);
91942321 295 clear_inode_flag(inode, FI_NEED_IPU);
c1ce1b02 296
dddd3d65 297 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
a5fd5050 298 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
fbfa2cc5 299 return ret;
a2a4a7e4 300 }
fbfa2cc5 301
9c7bb702 302 /* if the inode is dirty, let's recover all the time */
281518c6 303 if (!f2fs_skip_inode_update(inode, datasync)) {
2286c020 304 f2fs_write_inode(inode, NULL);
9c7bb702
CL		 305 goto go_write;
306 }
307
6d99ba41
JK		 308 /*
309 * if there is no written data, don't waste time to write recovery info.
310 */
91942321 311 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
4d57b86d 312 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
19c9c466 313
9c7bb702
CL		 314 /* it may call write_inode just prior to fsync */
315 if (need_inode_page_update(sbi, ino))
19c9c466 316 goto go_write;
19c9c466 317
91942321 318 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
4d57b86d 319 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
6d99ba41
JK		 320 goto flush_out;
321 goto out;
27879915
CY		 322 } else {
323 /*
324 * for OPU case, during fsync(), node can be persisted before
325 * data when lower device doesn't support write barrier, result
326 * in data corruption after SPO.
146949de 327 * So for strict fsync mode, force to use atomic write semantics
27879915
CY		 328 * to keep write order in between data/node and last node to
329 * avoid potential data corruption.
330 */
331 if (F2FS_OPTION(sbi).fsync_mode ==
332 FSYNC_MODE_STRICT && !atomic)
333 atomic = true;
6d99ba41 334 }
19c9c466 335go_write:
e5d2385e
JK		 336 /*
337 * Both of fdatasync() and fsync() are able to be recovered from
338 * sudden-power-off.
339 */
e4544b63 340 f2fs_down_read(&F2FS_I(inode)->i_sem);
a5fd5050 341 cp_reason = need_do_checkpoint(inode);
e4544b63 342 f2fs_up_read(&F2FS_I(inode)->i_sem);
d928bfbf 343
a5fd5050 344 if (cp_reason) {
fbfa2cc5
JK		 345 /* all the dirty node pages should be flushed for POR */
346 ret = f2fs_sync_fs(inode->i_sb, 1);
d928bfbf 347
51455b19
CL		 348 /*
349 * We've secured consistency through sync_fs. Following pino
350 * will be used only for fsynced inodes after checkpoint.
351 */
352 try_to_fix_pino(inode);
91942321
JK		 353 clear_inode_flag(inode, FI_APPEND_WRITE);
354 clear_inode_flag(inode, FI_UPDATE_WRITE);
51455b19
CL		 355 goto out;
356 }
88bd02c9 357sync_nodes:
c29fd0c0 358 atomic_inc(&sbi->wb_sync_req[NODE]);
50fa53ec 359 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
c29fd0c0 360 atomic_dec(&sbi->wb_sync_req[NODE]);
c267ec15
JK		 361 if (ret)
362 goto out;
51455b19 363
871f599f 364 /* if cp_error was enabled, we should avoid infinite loop */
6d5a1495
CY		 365 if (unlikely(f2fs_cp_error(sbi))) {
366 ret = -EIO;
871f599f 367 goto out;
6d5a1495 368 }
871f599f 369
4d57b86d 370 if (f2fs_need_inode_block_update(sbi, ino)) {
7c45729a 371 f2fs_mark_inode_dirty_sync(inode, true);
51455b19
CL		 372 f2fs_write_inode(inode, NULL);
373 goto sync_nodes;
fbfa2cc5 374 }
51455b19 375
b6a245eb
JK		 376 /*
377 * If it's atomic_write, it's just fine to keep write ordering. So
378 * here we don't need to wait for node write completion, since we use
379 * node chain which serializes node blocks. If one of node writes are
380 * reordered, we can see simply broken chain, resulting in stopping
381 * roll-forward recovery. It means we'll recover all or none node blocks
382 * given fsync mark.
383 */
384 if (!atomic) {
50fa53ec 385 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
b6a245eb
JK		 386 if (ret)
387 goto out;
388 }
51455b19
CL		 389
390 /* once recovery info is written, don't need to tack this */
4d57b86d 391 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
91942321 392 clear_inode_flag(inode, FI_APPEND_WRITE);
51455b19 393flush_out:
b722ff8a 394 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
39d787be 395 ret = f2fs_issue_flush(sbi, inode->i_ino);
3f06252f 396 if (!ret) {
4d57b86d 397 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
3f06252f 398 clear_inode_flag(inode, FI_UPDATE_WRITE);
4d57b86d 399 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
3f06252f 400 }
d0239e1b 401 f2fs_update_time(sbi, REQ_TIME);
fbfa2cc5 402out:
a5fd5050 403 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
fbfa2cc5
JK		 404 return ret;
405}
406
608514de
JK		 407int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
408{
1f227a3e
JK		 409 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
410 return -EIO;
608514de
JK		 411 return f2fs_do_sync_file(file, start, end, datasync, false);
412}
413
a94c7fde
DJ		 414static bool __found_offset(struct address_space *mapping,
415 struct dnode_of_data *dn, pgoff_t index, int whence)
7f7670fe 416{
a94c7fde
DJ		 417 block_t blkaddr = f2fs_data_blkaddr(dn);
418 struct inode *inode = mapping->host;
419 bool compressed_cluster = false;
420
421 if (f2fs_compressed_file(inode)) {
6f7ec661 422 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio,
a94c7fde
DJ		 423 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size));
424
425 compressed_cluster = first_blkaddr == COMPRESS_ADDR;
426 }
427
7f7670fe
JK		 428 switch (whence) {
429 case SEEK_DATA:
4cb03fec
MWO		 430 if (__is_valid_data_blkaddr(blkaddr))
431 return true;
432 if (blkaddr == NEW_ADDR &&
433 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
7f7670fe 434 return true;
a94c7fde
DJ		 435 if (compressed_cluster)
436 return true;
7f7670fe
JK		 437 break;
438 case SEEK_HOLE:
a94c7fde
DJ		 439 if (compressed_cluster)
440 return false;
7f7670fe
JK		 441 if (blkaddr == NULL_ADDR)
442 return true;
443 break;
444 }
445 return false;
446}
447
267378d4
CY		 448static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
449{
450 struct inode *inode = file->f_mapping->host;
d33ebd57 451 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
267378d4 452 struct dnode_of_data dn;
4cb03fec 453 pgoff_t pgofs, end_offset;
7f7670fe
JK		 454 loff_t data_ofs = offset;
455 loff_t isize;
267378d4
CY		 456 int err = 0;
457
36062b91 458 inode_lock_shared(inode);
267378d4
CY		 459
460 isize = i_size_read(inode);
461 if (offset >= isize)
462 goto fail;
463
464 /* handle inline data case */
7a6e59d7
CY		 465 if (f2fs_has_inline_data(inode)) {
466 if (whence == SEEK_HOLE) {
467 data_ofs = isize;
468 goto found;
469 } else if (whence == SEEK_DATA) {
470 data_ofs = offset;
471 goto found;
472 }
267378d4
CY		 473 }
474
09cbfeaf 475 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
267378d4 476
09cbfeaf 477 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 478 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 479 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
267378d4
CY		 480 if (err && err != -ENOENT) {
481 goto fail;
482 } else if (err == -ENOENT) {
e1c42045 483 /* direct node does not exists */
267378d4 484 if (whence == SEEK_DATA) {
4d57b86d 485 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
267378d4
CY		 486 continue;
487 } else {
488 goto found;
489 }
490 }
491
6f7ec661 492 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
267378d4
CY		 493
494 /* find data/hole in dnode block */
495 for (; dn.ofs_in_node < end_offset;
496 dn.ofs_in_node++, pgofs++,
09cbfeaf 497 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 498 block_t blkaddr;
f11e98bd 499
a2ced1ce 500 blkaddr = f2fs_data_blkaddr(&dn);
267378d4 501
c9b60788
CY		 502 if (__is_valid_data_blkaddr(blkaddr) &&
503 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
93770ab7 504 blkaddr, DATA_GENERIC_ENHANCE)) {
c9b60788
CY		 505 f2fs_put_dnode(&dn);
506 goto fail;
507 }
508
a94c7fde 509 if (__found_offset(file->f_mapping, &dn,
e1da7872 510 pgofs, whence)) {
267378d4
CY		 511 f2fs_put_dnode(&dn);
512 goto found;
513 }
514 }
515 f2fs_put_dnode(&dn);
516 }
517
518 if (whence == SEEK_DATA)
519 goto fail;
267378d4 520found:
fe369bc8
JK		 521 if (whence == SEEK_HOLE && data_ofs > isize)
522 data_ofs = isize;
36062b91 523 inode_unlock_shared(inode);
267378d4
CY		 524 return vfs_setpos(file, data_ofs, maxbytes);
525fail:
36062b91 526 inode_unlock_shared(inode);
267378d4
CY		 527 return -ENXIO;
528}
529
530static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
531{
532 struct inode *inode = file->f_mapping->host;
d33ebd57 533 loff_t maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
6d1451bf 534
267378d4
CY		 535 switch (whence) {
536 case SEEK_SET:
537 case SEEK_CUR:
538 case SEEK_END:
539 return generic_file_llseek_size(file, offset, whence,
540 maxbytes, i_size_read(inode));
541 case SEEK_DATA:
542 case SEEK_HOLE:
0b4c5afd
JK		 543 if (offset < 0)
544 return -ENXIO;
267378d4
CY		 545 return f2fs_seek_block(file, offset, whence);
546 }
547
548 return -EINVAL;
549}
550
fbfa2cc5
JK		 551static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
552{
b3d208f9
JK		 553 struct inode *inode = file_inode(file);
554
1f227a3e
JK		 555 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
556 return -EIO;
557
4c8ff709
CY		 558 if (!f2fs_is_compress_backend_ready(inode))
559 return -EOPNOTSUPP;
560
fbfa2cc5
JK		 561 file_accessed(file);
562 vma->vm_ops = &f2fs_file_vm_ops;
b5ab3276
CY		 563
564 f2fs_down_read(&F2FS_I(inode)->i_sem);
4c8ff709 565 set_inode_flag(inode, FI_MMAP_FILE);
b5ab3276
CY		 566 f2fs_up_read(&F2FS_I(inode)->i_sem);
567
fbfa2cc5
JK		 568 return 0;
569}
570
298b1e41
CY		 571static int finish_preallocate_blocks(struct inode *inode)
572{
ecf68ffe
CY		 573 int ret = 0;
574 bool opened;
298b1e41 575
ecf68ffe
CY		 576 f2fs_down_read(&F2FS_I(inode)->i_sem);
577 opened = is_inode_flag_set(inode, FI_OPENED_FILE);
578 f2fs_up_read(&F2FS_I(inode)->i_sem);
579 if (opened)
298b1e41 580 return 0;
298b1e41 581
ecf68ffe
CY		 582 inode_lock(inode);
583 if (is_inode_flag_set(inode, FI_OPENED_FILE))
584 goto out_unlock;
585
586 if (!file_should_truncate(inode))
587 goto out_update;
298b1e41
CY		 588
589 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
590 filemap_invalidate_lock(inode->i_mapping);
591
592 truncate_setsize(inode, i_size_read(inode));
593 ret = f2fs_truncate(inode);
594
595 filemap_invalidate_unlock(inode->i_mapping);
596 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
298b1e41 597 if (ret)
ecf68ffe 598 goto out_unlock;
298b1e41
CY		 599
600 file_dont_truncate(inode);
ecf68ffe
CY		 601out_update:
602 f2fs_down_write(&F2FS_I(inode)->i_sem);
603 set_inode_flag(inode, FI_OPENED_FILE);
604 f2fs_up_write(&F2FS_I(inode)->i_sem);
605out_unlock:
606 inode_unlock(inode);
607 return ret;
298b1e41
CY		 608}
609
fcc85a4d
JK		 610static int f2fs_file_open(struct inode *inode, struct file *filp)
611{
2e168c82 612 int err = fscrypt_file_open(inode, filp);
fcc85a4d 613
95ae251f
EB		 614 if (err)
615 return err;
616
4c8ff709
CY		 617 if (!f2fs_is_compress_backend_ready(inode))
618 return -EOPNOTSUPP;
619
95ae251f 620 err = fsverity_file_open(inode, filp);
2e168c82
EB		 621 if (err)
622 return err;
b91050a8 623
210a03c9 624 filp->f_mode |= FMODE_NOWAIT;
94c8431f 625 filp->f_mode |= FMODE_CAN_ODIRECT;
b91050a8 626
298b1e41
CY		 627 err = dquot_file_open(inode, filp);
628 if (err)
629 return err;
630
631 return finish_preallocate_blocks(inode);
fcc85a4d
JK		 632}
633
4d57b86d 634void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
fbfa2cc5 635{
4081363f 636 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
19b2c30d 637 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
fbfa2cc5 638 __le32 *addr;
4c8ff709
CY		 639 bool compressed_cluster = false;
640 int cluster_index = 0, valid_blocks = 0;
641 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
c2759eba 642 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
120ac1dc
YS		 643 block_t blkstart;
644 int blklen = 0;
7a2af766 645
7d28f13c 646 addr = get_dnode_addr(dn->inode, dn->node_folio) + ofs;
120ac1dc 647 blkstart = le32_to_cpu(*addr);
fbfa2cc5 648
a5029a57 649 /* Assumption: truncation starts with cluster */
4c8ff709 650 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
fbfa2cc5 651 block_t blkaddr = le32_to_cpu(*addr);
f11e98bd 652
4c8ff709
CY		 653 if (f2fs_compressed_file(dn->inode) &&
654 !(cluster_index & (cluster_size - 1))) {
655 if (compressed_cluster)
656 f2fs_i_compr_blocks_update(dn->inode,
657 valid_blocks, false);
658 compressed_cluster = (blkaddr == COMPRESS_ADDR);
659 valid_blocks = 0;
660 }
661
fbfa2cc5 662 if (blkaddr == NULL_ADDR)
120ac1dc 663 goto next;
fbfa2cc5 664
59d0d4c3 665 f2fs_set_data_blkaddr(dn, NULL_ADDR);
c9b60788 666
4c8ff709 667 if (__is_valid_data_blkaddr(blkaddr)) {
c7115e09 668 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE))
120ac1dc 669 goto next;
c7115e09 670 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr,
31f85ccc 671 DATA_GENERIC_ENHANCE))
120ac1dc 672 goto next;
4c8ff709
CY		 673 if (compressed_cluster)
674 valid_blocks++;
675 }
c9b60788 676
120ac1dc
YS		 677 if (blkstart + blklen == blkaddr) {
678 blklen++;
679 } else {
680 f2fs_invalidate_blocks(sbi, blkstart, blklen);
681 blkstart = blkaddr;
682 blklen = 1;
683 }
ef8d563f
CY		 684
685 if (!released || blkaddr != COMPRESS_ADDR)
686 nr_free++;
120ac1dc
YS		 687
688 continue;
689
690next:
691 if (blklen)
692 f2fs_invalidate_blocks(sbi, blkstart, blklen);
693
694 blkstart = le32_to_cpu(*(addr + 1));
695 blklen = 0;
fbfa2cc5 696 }
19b2c30d 697
120ac1dc
YS		 698 if (blklen)
699 f2fs_invalidate_blocks(sbi, blkstart, blklen);
700
4c8ff709
CY		 701 if (compressed_cluster)
702 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
703
fbfa2cc5 704 if (nr_free) {
19b2c30d
CY		 705 pgoff_t fofs;
706 /*
707 * once we invalidate valid blkaddr in range [ofs, ofs + count],
708 * we will invalidate all blkaddr in the whole range.
709 */
6f7ec661 710 fofs = f2fs_start_bidx_of_node(ofs_of_node(&dn->node_folio->page),
81ca7350 711 dn->inode) + ofs;
e7547dac 712 f2fs_update_read_extent_cache_range(dn, fofs, 0, len);
8c0ed062 713 f2fs_update_age_extent_cache_range(dn, fofs, len);
d7cc950b 714 dec_valid_block_count(sbi, dn->inode, nr_free);
fbfa2cc5
JK		 715 }
716 dn->ofs_in_node = ofs;
51dd6249 717
d0239e1b 718 f2fs_update_time(sbi, REQ_TIME);
51dd6249
NJ		 719 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
720 dn->ofs_in_node, nr_free);
fbfa2cc5
JK		 721}
722
0bfcfcca 723static int truncate_partial_data_page(struct inode *inode, u64 from,
43f3eae1 724 bool cache_only)
fbfa2cc5 725{
193bea1d 726 loff_t offset = from & (PAGE_SIZE - 1);
09cbfeaf 727 pgoff_t index = from >> PAGE_SHIFT;
43f3eae1 728 struct address_space *mapping = inode->i_mapping;
ab907aa2 729 struct folio *folio;
fbfa2cc5 730
43f3eae1 731 if (!offset && !cache_only)
b3d208f9 732 return 0;
fbfa2cc5 733
43f3eae1 734 if (cache_only) {
ab907aa2
MWO		 735 folio = filemap_lock_folio(mapping, index);
736 if (IS_ERR(folio))
737 return 0;
738 if (folio_test_uptodate(folio))
43f3eae1 739 goto truncate_out;
ab907aa2 740 f2fs_folio_put(folio, true);
b3d208f9 741 return 0;
43f3eae1 742 }
fbfa2cc5 743
ab907aa2
MWO		 744 folio = f2fs_get_lock_data_folio(inode, index, true);
745 if (IS_ERR(folio))
746 return PTR_ERR(folio) == -ENOENT ? 0 : PTR_ERR(folio);
43f3eae1 747truncate_out:
ab907aa2
MWO		 748 f2fs_folio_wait_writeback(folio, DATA, true, true);
749 folio_zero_segment(folio, offset, folio_size(folio));
a9bcf9bc
JK		 750
751 /* An encrypted inode should have a key and truncate the last page. */
62230e0d 752 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
a9bcf9bc 753 if (!cache_only)
ab907aa2
MWO		 754 folio_mark_dirty(folio);
755 f2fs_folio_put(folio, true);
b3d208f9 756 return 0;
fbfa2cc5
JK		 757}
758
3265d3db 759int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
fbfa2cc5 760{
4081363f 761 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 762 struct dnode_of_data dn;
763 pgoff_t free_from;
9ffe0fb5 764 int count = 0, err = 0;
d9554488 765 struct folio *ifolio;
0bfcfcca 766 bool truncate_page = false;
fbfa2cc5 767
51dd6249
NJ		 768 trace_f2fs_truncate_blocks_enter(inode, from);
769
128d333f
DJ		 770 if (IS_DEVICE_ALIASING(inode) && from) {
771 err = -EINVAL;
772 goto out_err;
773 }
774
df033caf 775 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
fbfa2cc5 776
6d1451bf 777 if (free_from >= max_file_blocks(inode))
09210c97
CY		 778 goto free_partial;
779
764aa3e9 780 if (lock)
c42d28ce 781 f2fs_lock_op(sbi);
9ffe0fb5 782
d9554488
MWO		 783 ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
784 if (IS_ERR(ifolio)) {
785 err = PTR_ERR(ifolio);
b3d208f9
JK		 786 goto out;
787 }
788
128d333f
DJ		 789 if (IS_DEVICE_ALIASING(inode)) {
790 struct extent_tree *et = F2FS_I(inode)->extent_tree[EX_READ];
791 struct extent_info ei = et->largest;
128d333f 792
6d4008dc 793 f2fs_invalidate_blocks(sbi, ei.blk, ei.len);
128d333f
DJ		 794
795 dec_valid_block_count(sbi, inode, ei.len);
796 f2fs_update_time(sbi, REQ_TIME);
797
d9554488 798 f2fs_folio_put(ifolio, true);
128d333f
DJ		 799 goto out;
800 }
801
b3d208f9 802 if (f2fs_has_inline_data(inode)) {
b3720382 803 f2fs_truncate_inline_inode(inode, ifolio, from);
d9554488 804 f2fs_folio_put(ifolio, true);
0bfcfcca 805 truncate_page = true;
b3d208f9
JK		 806 goto out;
807 }
808
214235c2 809 set_new_dnode(&dn, inode, ifolio, NULL, 0);
4d57b86d 810 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
fbfa2cc5
JK		 811 if (err) {
812 if (err == -ENOENT)
813 goto free_next;
b3d208f9 814 goto out;
1ce86bf6
JK		 815 }
816
6f7ec661 817 count = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
fbfa2cc5
JK		 818
819 count -= dn.ofs_in_node;
9850cf4a 820 f2fs_bug_on(sbi, count < 0);
39936837 821
6f7ec661 822 if (dn.ofs_in_node || IS_INODE(&dn.node_folio->page)) {
4d57b86d 823 f2fs_truncate_data_blocks_range(&dn, count);
fbfa2cc5
JK		 824 free_from += count;
825 }
826
827 f2fs_put_dnode(&dn);
828free_next:
4d57b86d 829 err = f2fs_truncate_inode_blocks(inode, free_from);
764d2c80
JK		 830out:
831 if (lock)
c42d28ce 832 f2fs_unlock_op(sbi);
09210c97 833free_partial:
b3d208f9
JK		 834 /* lastly zero out the first data page */
835 if (!err)
0bfcfcca 836 err = truncate_partial_data_page(inode, from, truncate_page);
128d333f 837out_err:
51dd6249 838 trace_f2fs_truncate_blocks_exit(inode, err);
fbfa2cc5
JK		 839 return err;
840}
841
4c8ff709
CY		 842int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
843{
844 u64 free_from = from;
3265d3db 845 int err;
4c8ff709 846
3265d3db 847#ifdef CONFIG_F2FS_FS_COMPRESSION
4c8ff709
CY		 848 /*
849 * for compressed file, only support cluster size
850 * aligned truncation.
851 */
4fec3fc0
CY		 852 if (f2fs_compressed_file(inode))
853 free_from = round_up(from,
854 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
3265d3db
CY		 855#endif
856
857 err = f2fs_do_truncate_blocks(inode, free_from, lock);
858 if (err)
859 return err;
860
861#ifdef CONFIG_F2FS_FS_COMPRESSION
4a4fc043
FC		 862 /*
863 * For compressed file, after release compress blocks, don't allow write
864 * direct, but we should allow write direct after truncate to zero.
865 */
866 if (f2fs_compressed_file(inode) && !free_from
867 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
868 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
869
17d7648d 870 if (from != free_from) {
3265d3db 871 err = f2fs_truncate_partial_cluster(inode, from, lock);
17d7648d
CY		 872 if (err)
873 return err;
874 }
3265d3db 875#endif
4c8ff709 876
17d7648d 877 return 0;
4c8ff709
CY		 878}
879
9a449e9c 880int f2fs_truncate(struct inode *inode)
fbfa2cc5 881{
b0154891
CY		 882 int err;
883
1f227a3e
JK		 884 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
885 return -EIO;
886
fbfa2cc5
JK		 887 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
888 S_ISLNK(inode->i_mode)))
b0154891 889 return 0;
fbfa2cc5 890
51dd6249
NJ		 891 trace_f2fs_truncate(inode);
892
c40e15a9 893 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE))
14b44d23 894 return -EIO;
7fa750a1 895
10a26878 896 err = f2fs_dquot_initialize(inode);
25fb04db
YC		 897 if (err)
898 return err;
899
92dffd01 900 /* we should check inline_data size */
b9d777b8 901 if (!f2fs_may_inline_data(inode)) {
b0154891
CY		 902 err = f2fs_convert_inline_inode(inode);
903 if (err)
904 return err;
92dffd01
JK		 905 }
906
c42d28ce 907 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
b0154891
CY		 908 if (err)
909 return err;
910
11cc6426 911 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
7c45729a 912 f2fs_mark_inode_dirty_sync(inode, false);
b0154891 913 return 0;
fbfa2cc5
JK		 914}
915
bd367329 916static bool f2fs_force_buffered_io(struct inode *inode, int rw)
2db0487f
EB		 917{
918 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2db0487f
EB		 919
920 if (!fscrypt_dio_supported(inode))
921 return true;
922 if (fsverity_active(inode))
923 return true;
924 if (f2fs_compressed_file(inode))
925 return true;
26e6f59d
CY		 926 /*
927 * only force direct read to use buffered IO, for direct write,
928 * it expects inline data conversion before committing IO.
929 */
930 if (f2fs_has_inline_data(inode) && rw == READ)
5c8764f8 931 return true;
2db0487f
EB		 932
933 /* disallow direct IO if any of devices has unaligned blksize */
934 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize)
935 return true;
5d170fe4
LT		 936 /*
937 * for blkzoned device, fallback direct IO to buffered IO, so
938 * all IOs can be serialized by log-structured write.
939 */
b2cf5a1f
DJ		 940 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) &&
941 !f2fs_is_pinned_file(inode))
5d170fe4 942 return true;
bd367329 943 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
2db0487f
EB		 944 return true;
945
946 return false;
947}
948
b74d24f7 949int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path,
549c7297 950 struct kstat *stat, u32 request_mask, unsigned int query_flags)
fbfa2cc5 951{
a528d35e 952 struct inode *inode = d_inode(path->dentry);
1c6d8ee4 953 struct f2fs_inode_info *fi = F2FS_I(inode);
d13732cc 954 struct f2fs_inode *ri = NULL;
1c6d8ee4
CY		 955 unsigned int flags;
956
1c1d35df 957 if (f2fs_has_extra_attr(inode) &&
7beb01f7 958 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
1c1d35df
CY		 959 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
960 stat->result_mask |= STATX_BTIME;
961 stat->btime.tv_sec = fi->i_crtime.tv_sec;
962 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
963 }
964
c8c02272
EB		 965 /*
966 * Return the DIO alignment restrictions if requested. We only return
967 * this information when requested, since on encrypted files it might
968 * take a fair bit of work to get if the file wasn't opened recently.
969 *
970 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN
971 * cannot represent that, so in that case we report no DIO support.
972 */
973 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
974 unsigned int bsize = i_blocksize(inode);
975
976 stat->result_mask |= STATX_DIOALIGN;
977 if (!f2fs_force_buffered_io(inode, WRITE)) {
978 stat->dio_mem_align = bsize;
979 stat->dio_offset_align = bsize;
980 }
981 }
982
36098557 983 flags = fi->i_flags;
fd26725f
CY		 984 if (flags & F2FS_COMPR_FL)
985 stat->attributes |= STATX_ATTR_COMPRESSED;
59c84408 986 if (flags & F2FS_APPEND_FL)
1c6d8ee4 987 stat->attributes |= STATX_ATTR_APPEND;
62230e0d 988 if (IS_ENCRYPTED(inode))
1c6d8ee4 989 stat->attributes |= STATX_ATTR_ENCRYPTED;
59c84408 990 if (flags & F2FS_IMMUTABLE_FL)
1c6d8ee4 991 stat->attributes |= STATX_ATTR_IMMUTABLE;
59c84408 992 if (flags & F2FS_NODUMP_FL)
1c6d8ee4 993 stat->attributes |= STATX_ATTR_NODUMP;
924e3194
EB		 994 if (IS_VERITY(inode))
995 stat->attributes |= STATX_ATTR_VERITY;
1c6d8ee4 996
fd26725f
CY		 997 stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
998 STATX_ATTR_APPEND |
1c6d8ee4
CY		 999 STATX_ATTR_ENCRYPTED |
1000 STATX_ATTR_IMMUTABLE |
924e3194
EB		 1001 STATX_ATTR_NODUMP |
1002 STATX_ATTR_VERITY);
1c6d8ee4 1003
0d72b928 1004 generic_fillattr(idmap, request_mask, inode, stat);
5b4267d1
JK		 1005
1006 /* we need to show initial sectors used for inline_data/dentries */
1007 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
1008 f2fs_has_inline_dentry(inode))
1009 stat->blocks += (stat->size + 511) >> 9;
1010
fbfa2cc5
JK		 1011 return 0;
1012}
1013
1014#ifdef CONFIG_F2FS_FS_POSIX_ACL
c1632a0f 1015static void __setattr_copy(struct mnt_idmap *idmap,
e65ce2a5 1016 struct inode *inode, const struct iattr *attr)
fbfa2cc5 1017{
fbfa2cc5
JK		 1018 unsigned int ia_valid = attr->ia_valid;
1019
0dbe12f2
CB		 1020 i_uid_update(idmap, attr, inode);
1021 i_gid_update(idmap, attr, inode);
eb31e2f6 1022 if (ia_valid & ATTR_ATIME)
11cc6426 1023 inode_set_atime_to_ts(inode, attr->ia_atime);
eb31e2f6 1024 if (ia_valid & ATTR_MTIME)
11cc6426 1025 inode_set_mtime_to_ts(inode, attr->ia_mtime);
eb31e2f6 1026 if (ia_valid & ATTR_CTIME)
c62ebd35 1027 inode_set_ctime_to_ts(inode, attr->ia_ctime);
fbfa2cc5
JK		 1028 if (ia_valid & ATTR_MODE) {
1029 umode_t mode = attr->ia_mode;
1030
8444ee22 1031 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode)))
fbfa2cc5 1032 mode &= ~S_ISGID;
91942321 1033 set_acl_inode(inode, mode);
fbfa2cc5
JK		 1034 }
1035}
1036#else
1037#define __setattr_copy setattr_copy
1038#endif
1039
c1632a0f 1040int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
549c7297 1041 struct iattr *attr)
fbfa2cc5 1042{
2b0143b5 1043 struct inode *inode = d_inode(dentry);
7309871c 1044 struct f2fs_inode_info *fi = F2FS_I(inode);
fbfa2cc5
JK		 1045 int err;
1046
1f227a3e
JK		 1047 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1048 return -EIO;
1049
e0fcd015
CY		 1050 if (unlikely(IS_IMMUTABLE(inode)))
1051 return -EPERM;
1052
1053 if (unlikely(IS_APPEND(inode) &&
1054 (attr->ia_valid & (ATTR_MODE | ATTR_UID |
1055 ATTR_GID | ATTR_TIMES_SET))))
1056 return -EPERM;
1057
29ed2b5d 1058 if ((attr->ia_valid & ATTR_SIZE)) {
128d333f
DJ		 1059 if (!f2fs_is_compress_backend_ready(inode) ||
1060 IS_DEVICE_ALIASING(inode))
29ed2b5d
CY		 1061 return -EOPNOTSUPP;
1062 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
1063 !IS_ALIGNED(attr->ia_size,
7309871c 1064 F2FS_BLK_TO_BYTES(fi->i_cluster_size)))
29ed2b5d
CY		 1065 return -EINVAL;
1066 }
4c8ff709 1067
c1632a0f 1068 err = setattr_prepare(idmap, dentry, attr);
fbfa2cc5
JK		 1069 if (err)
1070 return err;
1071
20bb2479
EB		 1072 err = fscrypt_prepare_setattr(dentry, attr);
1073 if (err)
1074 return err;
1075
95ae251f
EB		 1076 err = fsverity_prepare_setattr(dentry, attr);
1077 if (err)
1078 return err;
1079
f861646a 1080 if (is_quota_modification(idmap, inode, attr)) {
10a26878 1081 err = f2fs_dquot_initialize(inode);
0abd675e
CY		 1082 if (err)
1083 return err;
1084 }
0dbe12f2
CB		 1085 if (i_uid_needs_update(idmap, attr, inode) ||
1086 i_gid_needs_update(idmap, attr, inode)) {
af033b2a 1087 f2fs_lock_op(F2FS_I_SB(inode));
f861646a 1088 err = dquot_transfer(idmap, inode, attr);
af033b2a
CY		 1089 if (err) {
1090 set_sbi_flag(F2FS_I_SB(inode),
1091 SBI_QUOTA_NEED_REPAIR);
1092 f2fs_unlock_op(F2FS_I_SB(inode));
0abd675e 1093 return err;
af033b2a
CY		 1094 }
1095 /*
1096 * update uid/gid under lock_op(), so that dquot and inode can
1097 * be updated atomically.
1098 */
0dbe12f2
CB		 1099 i_uid_update(idmap, attr, inode);
1100 i_gid_update(idmap, attr, inode);
af033b2a
CY		 1101 f2fs_mark_inode_dirty_sync(inode, true);
1102 f2fs_unlock_op(F2FS_I_SB(inode));
0abd675e
CY		 1103 }
1104
09db6a2e 1105 if (attr->ia_valid & ATTR_SIZE) {
cfb9a34d
JK		 1106 loff_t old_size = i_size_read(inode);
1107
1108 if (attr->ia_size > MAX_INLINE_DATA(inode)) {
1109 /*
1110 * should convert inline inode before i_size_write to
1111 * keep smaller than inline_data size with inline flag.
1112 */
1113 err = f2fs_convert_inline_inode(inode);
1114 if (err)
1115 return err;
1116 }
a33c1502 1117
96cfeb03
CY		 1118 /*
1119 * wait for inflight dio, blocks should be removed after
1120 * IO completion.
1121 */
1122 if (attr->ia_size < old_size)
1123 inode_dio_wait(inode);
1124
7309871c 1125 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
edc6d01b 1126 filemap_invalidate_lock(inode->i_mapping);
a33c1502 1127
ba8dac35
CY		 1128 if (attr->ia_size > old_size)
1129 f2fs_zero_post_eof_page(inode, attr->ia_size);
a33c1502
CY		 1130 truncate_setsize(inode, attr->ia_size);
1131
cfb9a34d 1132 if (attr->ia_size <= old_size)
9a449e9c 1133 err = f2fs_truncate(inode);
a33c1502
CY		 1134 /*
1135 * do not trim all blocks after i_size if target size is
1136 * larger than i_size.
1137 */
edc6d01b 1138 filemap_invalidate_unlock(inode->i_mapping);
7309871c 1139 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
a33c1502
CY		 1140 if (err)
1141 return err;
0cab80ee 1142
7309871c 1143 spin_lock(&fi->i_size_lock);
11cc6426 1144 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
7309871c
CY		 1145 fi->last_disk_size = i_size_read(inode);
1146 spin_unlock(&fi->i_size_lock);
fbfa2cc5
JK		 1147 }
1148
c1632a0f 1149 __setattr_copy(idmap, inode, attr);
fbfa2cc5
JK		 1150
1151 if (attr->ia_valid & ATTR_MODE) {
13e83a49 1152 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode));
17232e83
CY		 1153
1154 if (is_inode_flag_set(inode, FI_ACL_MODE)) {
1155 if (!err)
7309871c 1156 inode->i_mode = fi->i_acl_mode;
91942321 1157 clear_inode_flag(inode, FI_ACL_MODE);
fbfa2cc5
JK		 1158 }
1159 }
1160
c0ed4405 1161 /* file size may changed here */
ca597bdd 1162 f2fs_mark_inode_dirty_sync(inode, true);
15d04354
JK		 1163
1164 /* inode change will produce dirty node pages flushed by checkpoint */
1165 f2fs_balance_fs(F2FS_I_SB(inode), true);
1166
fbfa2cc5
JK		 1167 return err;
1168}
1169
1170const struct inode_operations f2fs_file_inode_operations = {
1171 .getattr = f2fs_getattr,
1172 .setattr = f2fs_setattr,
cac2f8b8 1173 .get_inode_acl = f2fs_get_acl,
a6dda0e6 1174 .set_acl = f2fs_set_acl,
fbfa2cc5 1175 .listxattr = f2fs_listxattr,
9ab70134 1176 .fiemap = f2fs_fiemap,
9b1bb01c
MS		 1177 .fileattr_get = f2fs_fileattr_get,
1178 .fileattr_set = f2fs_fileattr_set,
fbfa2cc5
JK		 1179};
1180
6394328a 1181static int fill_zero(struct inode *inode, pgoff_t index,
fbfa2cc5
JK		 1182 loff_t start, loff_t len)
1183{
4081363f 1184 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
c35cc972 1185 struct folio *folio;
fbfa2cc5
JK		 1186
1187 if (!len)
6394328a 1188 return 0;
fbfa2cc5 1189
2c4db1a6 1190 f2fs_balance_fs(sbi, true);
bd43df02 1191
e479556b 1192 f2fs_lock_op(sbi);
c35cc972 1193 folio = f2fs_get_new_data_folio(inode, NULL, index, false);
e479556b 1194 f2fs_unlock_op(sbi);
fbfa2cc5 1195
c35cc972
MWO		 1196 if (IS_ERR(folio))
1197 return PTR_ERR(folio);
6394328a 1198
c35cc972
MWO		 1199 f2fs_folio_wait_writeback(folio, DATA, true, true);
1200 folio_zero_range(folio, start, len);
1201 folio_mark_dirty(folio);
1202 f2fs_folio_put(folio, true);
6394328a 1203 return 0;
fbfa2cc5
JK		 1204}
1205
4d57b86d 1206int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
fbfa2cc5 1207{
fbfa2cc5
JK		 1208 int err;
1209
ea58711e 1210 while (pg_start < pg_end) {
fbfa2cc5 1211 struct dnode_of_data dn;
ea58711e 1212 pgoff_t end_offset, count;
9eaeba70 1213
fbfa2cc5 1214 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1215 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
fbfa2cc5 1216 if (err) {
ea58711e 1217 if (err == -ENOENT) {
4d57b86d
CY		 1218 pg_start = f2fs_get_next_page_offset(&dn,
1219 pg_start);
fbfa2cc5 1220 continue;
ea58711e 1221 }
fbfa2cc5
JK		 1222 return err;
1223 }
1224
6f7ec661 1225 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
ea58711e
CY		 1226 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1227
1228 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1229
4d57b86d 1230 f2fs_truncate_data_blocks_range(&dn, count);
fbfa2cc5 1231 f2fs_put_dnode(&dn);
ea58711e
CY		 1232
1233 pg_start += count;
fbfa2cc5
JK		 1234 }
1235 return 0;
1236}
1237
1cd75654 1238static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
fbfa2cc5
JK		 1239{
1240 pgoff_t pg_start, pg_end;
1241 loff_t off_start, off_end;
b9d777b8 1242 int ret;
fbfa2cc5 1243
b9d777b8
JK		 1244 ret = f2fs_convert_inline_inode(inode);
1245 if (ret)
1246 return ret;
9ffe0fb5 1247
ba8dac35
CY		 1248 filemap_invalidate_lock(inode->i_mapping);
1249 f2fs_zero_post_eof_page(inode, offset + len);
1250 filemap_invalidate_unlock(inode->i_mapping);
1251
09cbfeaf
KS		 1252 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1253 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
fbfa2cc5 1254
09cbfeaf
KS		 1255 off_start = offset & (PAGE_SIZE - 1);
1256 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5
JK		 1257
1258 if (pg_start == pg_end) {
6394328a 1259 ret = fill_zero(inode, pg_start, off_start,
fbfa2cc5 1260 off_end - off_start);
6394328a
CY		 1261 if (ret)
1262 return ret;
fbfa2cc5 1263 } else {
6394328a
CY		 1264 if (off_start) {
1265 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 1266 PAGE_SIZE - off_start);
6394328a
CY		 1267 if (ret)
1268 return ret;
1269 }
1270 if (off_end) {
1271 ret = fill_zero(inode, pg_end, 0, off_end);
1272 if (ret)
1273 return ret;
1274 }
fbfa2cc5
JK		 1275
1276 if (pg_start < pg_end) {
fbfa2cc5 1277 loff_t blk_start, blk_end;
4081363f 1278 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127a3d4 1279
2c4db1a6 1280 f2fs_balance_fs(sbi, true);
fbfa2cc5 1281
09cbfeaf
KS		 1282 blk_start = (loff_t)pg_start << PAGE_SHIFT;
1283 blk_end = (loff_t)pg_end << PAGE_SHIFT;
a33c1502 1284
e4544b63 1285 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6abaa83c 1286 filemap_invalidate_lock(inode->i_mapping);
a33c1502 1287
c8dc3047 1288 truncate_pagecache_range(inode, blk_start, blk_end - 1);
39936837 1289
e479556b 1290 f2fs_lock_op(sbi);
4d57b86d 1291 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
e479556b 1292 f2fs_unlock_op(sbi);
a33c1502 1293
6abaa83c 1294 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 1295 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
fbfa2cc5
JK		 1296 }
1297 }
1298
fbfa2cc5
JK		 1299 return ret;
1300}
1301
0a2aa8fb
JK		 1302static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1303 int *do_replace, pgoff_t off, pgoff_t len)
b4ace337
CY		 1304{
1305 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1306 struct dnode_of_data dn;
0a2aa8fb 1307 int ret, done, i;
ecbaa406 1308
0a2aa8fb 1309next_dnode:
6e2c64ad 1310 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1311 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
6e2c64ad
JK		 1312 if (ret && ret != -ENOENT) {
1313 return ret;
1314 } else if (ret == -ENOENT) {
0a2aa8fb
JK		 1315 if (dn.max_level == 0)
1316 return -ENOENT;
4c8ff709
CY		 1317 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1318 dn.ofs_in_node, len);
0a2aa8fb
JK		 1319 blkaddr += done;
1320 do_replace += done;
1321 goto next;
1322 }
1323
6f7ec661 1324 done = min((pgoff_t)ADDRS_PER_PAGE(&dn.node_folio->page, inode) -
0a2aa8fb
JK		 1325 dn.ofs_in_node, len);
1326 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
a2ced1ce 1327 *blkaddr = f2fs_data_blkaddr(&dn);
93770ab7
CY		 1328
1329 if (__is_valid_data_blkaddr(*blkaddr) &&
1330 !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1331 DATA_GENERIC_ENHANCE)) {
1332 f2fs_put_dnode(&dn);
10f966bb 1333 return -EFSCORRUPTED;
93770ab7
CY		 1334 }
1335
4d57b86d 1336 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
0a2aa8fb 1337
b0332a0f 1338 if (f2fs_lfs_mode(sbi)) {
0a2aa8fb 1339 f2fs_put_dnode(&dn);
fd114ab2 1340 return -EOPNOTSUPP;
0a2aa8fb
JK		 1341 }
1342
6e2c64ad 1343 /* do not invalidate this block address */
f28b3434 1344 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
0a2aa8fb 1345 *do_replace = 1;
b4ace337 1346 }
6e2c64ad 1347 }
0a2aa8fb
JK		 1348 f2fs_put_dnode(&dn);
1349next:
1350 len -= done;
1351 off += done;
1352 if (len)
1353 goto next_dnode;
1354 return 0;
1355}
b4ace337 1356
0a2aa8fb
JK		 1357static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1358 int *do_replace, pgoff_t off, int len)
1359{
1360 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1361 struct dnode_of_data dn;
1362 int ret, i;
b4ace337 1363
0a2aa8fb
JK		 1364 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1365 if (*do_replace == 0)
1366 continue;
b4ace337 1367
0a2aa8fb 1368 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1369 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
0a2aa8fb
JK		 1370 if (ret) {
1371 dec_valid_block_count(sbi, inode, 1);
e53c568f 1372 f2fs_invalidate_blocks(sbi, *blkaddr, 1);
0a2aa8fb
JK		 1373 } else {
1374 f2fs_update_data_blkaddr(&dn, *blkaddr);
36abef4e 1375 }
0a2aa8fb
JK		 1376 f2fs_put_dnode(&dn);
1377 }
1378 return 0;
1379}
1380
1381static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1382 block_t *blkaddr, int *do_replace,
1383 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1384{
1385 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1386 pgoff_t i = 0;
1387 int ret;
36abef4e 1388
0a2aa8fb
JK		 1389 while (i < len) {
1390 if (blkaddr[i] == NULL_ADDR && !full) {
1391 i++;
1392 continue;
6e2c64ad 1393 }
b4ace337 1394
0a2aa8fb
JK		 1395 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1396 struct dnode_of_data dn;
1397 struct node_info ni;
1398 size_t new_size;
1399 pgoff_t ilen;
b4ace337 1400
0a2aa8fb 1401 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
4d57b86d 1402 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
0a2aa8fb
JK		 1403 if (ret)
1404 return ret;
b4ace337 1405
a9419b63 1406 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
7735730d
CY		 1407 if (ret) {
1408 f2fs_put_dnode(&dn);
1409 return ret;
1410 }
1411
0a2aa8fb 1412 ilen = min((pgoff_t)
6f7ec661 1413 ADDRS_PER_PAGE(&dn.node_folio->page, dst_inode) -
0a2aa8fb
JK		 1414 dn.ofs_in_node, len - i);
1415 do {
a2ced1ce 1416 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
4d57b86d 1417 f2fs_truncate_data_blocks_range(&dn, 1);
0a2aa8fb
JK		 1418
1419 if (do_replace[i]) {
1420 f2fs_i_blocks_write(src_inode,
0abd675e 1421 1, false, false);
0a2aa8fb 1422 f2fs_i_blocks_write(dst_inode,
0abd675e 1423 1, true, false);
0a2aa8fb
JK		 1424 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1425 blkaddr[i], ni.version, true, false);
1426
1427 do_replace[i] = 0;
1428 }
1429 dn.ofs_in_node++;
1430 i++;
1f0d5c91 1431 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
0a2aa8fb
JK		 1432 if (dst_inode->i_size < new_size)
1433 f2fs_i_size_write(dst_inode, new_size);
e87f7329 1434 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
6e2c64ad 1435
0a2aa8fb
JK		 1436 f2fs_put_dnode(&dn);
1437 } else {
3dfcb2b4 1438 struct folio *fsrc, *fdst;
0a2aa8fb 1439
3dfcb2b4 1440 fsrc = f2fs_get_lock_data_folio(src_inode,
4d57b86d 1441 src + i, true);
3dfcb2b4
MWO		 1442 if (IS_ERR(fsrc))
1443 return PTR_ERR(fsrc);
1444 fdst = f2fs_get_new_data_folio(dst_inode, NULL, dst + i,
0a2aa8fb 1445 true);
3dfcb2b4
MWO		 1446 if (IS_ERR(fdst)) {
1447 f2fs_folio_put(fsrc, true);
1448 return PTR_ERR(fdst);
0a2aa8fb 1449 }
d3876e34 1450
3dfcb2b4 1451 f2fs_folio_wait_writeback(fdst, DATA, true, true);
d3876e34 1452
3dfcb2b4
MWO		 1453 memcpy_folio(fdst, 0, fsrc, 0, PAGE_SIZE);
1454 folio_mark_dirty(fdst);
1455 set_page_private_gcing(&fdst->page);
1456 f2fs_folio_put(fdst, true);
1457 f2fs_folio_put(fsrc, true);
b4ace337 1458
4d57b86d
CY		 1459 ret = f2fs_truncate_hole(src_inode,
1460 src + i, src + i + 1);
0a2aa8fb
JK		 1461 if (ret)
1462 return ret;
1463 i++;
1464 }
6e2c64ad
JK		 1465 }
1466 return 0;
0a2aa8fb 1467}
b4ace337 1468
0a2aa8fb
JK		 1469static int __exchange_data_block(struct inode *src_inode,
1470 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
363cad7f 1471 pgoff_t len, bool full)
0a2aa8fb
JK		 1472{
1473 block_t *src_blkaddr;
1474 int *do_replace;
363cad7f 1475 pgoff_t olen;
0a2aa8fb
JK		 1476 int ret;
1477
363cad7f 1478 while (len) {
d02a6e61 1479 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
0a2aa8fb 1480
628b3d14 1481 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
9d2a789c 1482 array_size(olen, sizeof(block_t)),
4f4460c0 1483 GFP_NOFS);
363cad7f
JK		 1484 if (!src_blkaddr)
1485 return -ENOMEM;
0a2aa8fb 1486
628b3d14 1487 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
9d2a789c 1488 array_size(olen, sizeof(int)),
4f4460c0 1489 GFP_NOFS);
363cad7f
JK		 1490 if (!do_replace) {
1491 kvfree(src_blkaddr);
1492 return -ENOMEM;
1493 }
0a2aa8fb 1494
363cad7f
JK		 1495 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1496 do_replace, src, olen);
1497 if (ret)
1498 goto roll_back;
0a2aa8fb 1499
363cad7f
JK		 1500 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1501 do_replace, src, dst, olen, full);
1502 if (ret)
1503 goto roll_back;
1504
1505 src += olen;
1506 dst += olen;
1507 len -= olen;
1508
1509 kvfree(src_blkaddr);
1510 kvfree(do_replace);
1511 }
0a2aa8fb
JK		 1512 return 0;
1513
1514roll_back:
9fd62605 1515 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
0a2aa8fb
JK		 1516 kvfree(src_blkaddr);
1517 kvfree(do_replace);
6e2c64ad
JK		 1518 return ret;
1519}
b4ace337 1520
6f8d4455 1521static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
6e2c64ad
JK		 1522{
1523 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f91108b8 1524 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
6f8d4455
JK		 1525 pgoff_t start = offset >> PAGE_SHIFT;
1526 pgoff_t end = (offset + len) >> PAGE_SHIFT;
0a2aa8fb 1527 int ret;
6e2c64ad 1528
0a2aa8fb 1529 f2fs_balance_fs(sbi, true);
5f281fab 1530
6f8d4455 1531 /* avoid gc operation during block exchange */
e4544b63 1532 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 1533 filemap_invalidate_lock(inode->i_mapping);
5f281fab 1534
ba8dac35
CY		 1535 f2fs_zero_post_eof_page(inode, offset + len);
1536
6f8d4455
JK		 1537 f2fs_lock_op(sbi);
1538 f2fs_drop_extent_tree(inode);
1539 truncate_pagecache(inode, offset);
0a2aa8fb
JK		 1540 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1541 f2fs_unlock_op(sbi);
6f8d4455 1542
edc6d01b 1543 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 1544 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
b4ace337
CY		 1545 return ret;
1546}
1547
1548static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1549{
b4ace337
CY		 1550 loff_t new_size;
1551 int ret;
1552
b4ace337
CY		 1553 if (offset + len >= i_size_read(inode))
1554 return -EINVAL;
1555
1556 /* collapse range should be aligned to block size of f2fs. */
1557 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1558 return -EINVAL;
1559
b9d777b8
JK		 1560 ret = f2fs_convert_inline_inode(inode);
1561 if (ret)
1562 return ret;
97a7b2c2 1563
b4ace337
CY		 1564 /* write out all dirty pages from offset */
1565 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1566 if (ret)
6f8d4455 1567 return ret;
b4ace337 1568
6f8d4455 1569 ret = f2fs_do_collapse(inode, offset, len);
b4ace337 1570 if (ret)
6f8d4455 1571 return ret;
b4ace337 1572
6e2c64ad 1573 /* write out all moved pages, if possible */
edc6d01b 1574 filemap_invalidate_lock(inode->i_mapping);
6e2c64ad
JK		 1575 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1576 truncate_pagecache(inode, offset);
1577
b4ace337 1578 new_size = i_size_read(inode) - len;
c42d28ce 1579 ret = f2fs_truncate_blocks(inode, new_size, true);
edc6d01b 1580 filemap_invalidate_unlock(inode->i_mapping);
b4ace337 1581 if (!ret)
fc9581c8 1582 f2fs_i_size_write(inode, new_size);
b4ace337
CY		 1583 return ret;
1584}
1585
6e961949
CY		 1586static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1587 pgoff_t end)
1588{
1589 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1590 pgoff_t index = start;
1591 unsigned int ofs_in_node = dn->ofs_in_node;
1592 blkcnt_t count = 0;
1593 int ret;
1594
1595 for (; index < end; index++, dn->ofs_in_node++) {
a2ced1ce 1596 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
6e961949
CY		 1597 count++;
1598 }
1599
1600 dn->ofs_in_node = ofs_in_node;
4d57b86d 1601 ret = f2fs_reserve_new_blocks(dn, count);
6e961949
CY		 1602 if (ret)
1603 return ret;
1604
1605 dn->ofs_in_node = ofs_in_node;
1606 for (index = start; index < end; index++, dn->ofs_in_node++) {
a2ced1ce 1607 dn->data_blkaddr = f2fs_data_blkaddr(dn);
6e961949 1608 /*
4d57b86d 1609 * f2fs_reserve_new_blocks will not guarantee entire block
6e961949
CY		 1610 * allocation.
1611 */
1612 if (dn->data_blkaddr == NULL_ADDR) {
1613 ret = -ENOSPC;
1614 break;
1615 }
25f82362
CY		 1616
1617 if (dn->data_blkaddr == NEW_ADDR)
1618 continue;
1619
1620 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1621 DATA_GENERIC_ENHANCE)) {
1622 ret = -EFSCORRUPTED;
1623 break;
6e961949 1624 }
25f82362 1625
e53c568f 1626 f2fs_invalidate_blocks(sbi, dn->data_blkaddr, 1);
59d0d4c3 1627 f2fs_set_data_blkaddr(dn, NEW_ADDR);
6e961949
CY		 1628 }
1629
e7547dac 1630 f2fs_update_read_extent_cache_range(dn, start, 0, index - start);
a84153f9 1631 f2fs_update_age_extent_cache_range(dn, start, index - start);
6e961949
CY		 1632
1633 return ret;
1634}
1635
75cd4e09
CY		 1636static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1637 int mode)
1638{
1639 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1640 struct address_space *mapping = inode->i_mapping;
1641 pgoff_t index, pg_start, pg_end;
1642 loff_t new_size = i_size_read(inode);
1643 loff_t off_start, off_end;
1644 int ret = 0;
1645
75cd4e09
CY		 1646 ret = inode_newsize_ok(inode, (len + offset));
1647 if (ret)
1648 return ret;
1649
b9d777b8
JK		 1650 ret = f2fs_convert_inline_inode(inode);
1651 if (ret)
1652 return ret;
75cd4e09
CY		 1653
1654 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1655 if (ret)
6f8d4455 1656 return ret;
75cd4e09 1657
ba8dac35
CY		 1658 filemap_invalidate_lock(mapping);
1659 f2fs_zero_post_eof_page(inode, offset + len);
1660 filemap_invalidate_unlock(mapping);
1661
09cbfeaf
KS		 1662 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1663 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
75cd4e09 1664
09cbfeaf
KS		 1665 off_start = offset & (PAGE_SIZE - 1);
1666 off_end = (offset + len) & (PAGE_SIZE - 1);
75cd4e09
CY		 1667
1668 if (pg_start == pg_end) {
6394328a
CY		 1669 ret = fill_zero(inode, pg_start, off_start,
1670 off_end - off_start);
1671 if (ret)
6f8d4455 1672 return ret;
6394328a 1673
75cd4e09
CY		 1674 new_size = max_t(loff_t, new_size, offset + len);
1675 } else {
1676 if (off_start) {
6394328a 1677 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 1678 PAGE_SIZE - off_start);
6394328a 1679 if (ret)
6f8d4455 1680 return ret;
6394328a 1681
75cd4e09 1682 new_size = max_t(loff_t, new_size,
09cbfeaf 1683 (loff_t)pg_start << PAGE_SHIFT);
75cd4e09
CY		 1684 }
1685
6e961949 1686 for (index = pg_start; index < pg_end;) {
75cd4e09 1687 struct dnode_of_data dn;
6e961949
CY		 1688 unsigned int end_offset;
1689 pgoff_t end;
75cd4e09 1690
e4544b63 1691 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 1692 filemap_invalidate_lock(mapping);
c7079853
CY		 1693
1694 truncate_pagecache_range(inode,
1695 (loff_t)index << PAGE_SHIFT,
1696 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1697
75cd4e09
CY		 1698 f2fs_lock_op(sbi);
1699
6e961949 1700 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1701 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
75cd4e09
CY		 1702 if (ret) {
1703 f2fs_unlock_op(sbi);
edc6d01b 1704 filemap_invalidate_unlock(mapping);
e4544b63 1705 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
75cd4e09
CY		 1706 goto out;
1707 }
1708
6f7ec661 1709 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
6e961949
CY		 1710 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1711
1712 ret = f2fs_do_zero_range(&dn, index, end);
75cd4e09 1713 f2fs_put_dnode(&dn);
c7079853 1714
75cd4e09 1715 f2fs_unlock_op(sbi);
edc6d01b 1716 filemap_invalidate_unlock(mapping);
e4544b63 1717 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
9434fcde
CY		 1718
1719 f2fs_balance_fs(sbi, dn.node_changed);
1720
6e961949
CY		 1721 if (ret)
1722 goto out;
75cd4e09 1723
6e961949 1724 index = end;
75cd4e09 1725 new_size = max_t(loff_t, new_size,
6e961949 1726 (loff_t)index << PAGE_SHIFT);
75cd4e09
CY		 1727 }
1728
1729 if (off_end) {
6394328a
CY		 1730 ret = fill_zero(inode, pg_end, 0, off_end);
1731 if (ret)
1732 goto out;
1733
75cd4e09
CY		 1734 new_size = max_t(loff_t, new_size, offset + len);
1735 }
1736 }
1737
1738out:
17cd07ae
CY		 1739 if (new_size > i_size_read(inode)) {
1740 if (mode & FALLOC_FL_KEEP_SIZE)
1741 file_set_keep_isize(inode);
1742 else
1743 f2fs_i_size_write(inode, new_size);
1744 }
75cd4e09
CY		 1745 return ret;
1746}
1747
f62185d0
CY		 1748static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1749{
1750 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
edc6d01b 1751 struct address_space *mapping = inode->i_mapping;
0a2aa8fb 1752 pgoff_t nr, pg_start, pg_end, delta, idx;
f62185d0 1753 loff_t new_size;
6e2c64ad 1754 int ret = 0;
f62185d0 1755
f62185d0 1756 new_size = i_size_read(inode) + len;
46e82fb1
KM		 1757 ret = inode_newsize_ok(inode, new_size);
1758 if (ret)
1759 return ret;
f62185d0
CY		 1760
1761 if (offset >= i_size_read(inode))
1762 return -EINVAL;
1763
1764 /* insert range should be aligned to block size of f2fs. */
1765 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1766 return -EINVAL;
1767
b9d777b8
JK		 1768 ret = f2fs_convert_inline_inode(inode);
1769 if (ret)
1770 return ret;
97a7b2c2 1771
2c4db1a6 1772 f2fs_balance_fs(sbi, true);
2a340760 1773
edc6d01b 1774 filemap_invalidate_lock(mapping);
c42d28ce 1775 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
edc6d01b 1776 filemap_invalidate_unlock(mapping);
f62185d0 1777 if (ret)
6f8d4455 1778 return ret;
f62185d0
CY		 1779
1780 /* write out all dirty pages from offset */
edc6d01b 1781 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
f62185d0 1782 if (ret)
6f8d4455 1783 return ret;
f62185d0 1784
09cbfeaf
KS		 1785 pg_start = offset >> PAGE_SHIFT;
1786 pg_end = (offset + len) >> PAGE_SHIFT;
f62185d0 1787 delta = pg_end - pg_start;
f91108b8 1788 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
0a2aa8fb 1789
6f8d4455 1790 /* avoid gc operation during block exchange */
e4544b63 1791 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 1792 filemap_invalidate_lock(mapping);
ba8dac35
CY		 1793
1794 f2fs_zero_post_eof_page(inode, offset + len);
6f8d4455
JK		 1795 truncate_pagecache(inode, offset);
1796
0a2aa8fb
JK		 1797 while (!ret && idx > pg_start) {
1798 nr = idx - pg_start;
1799 if (nr > delta)
1800 nr = delta;
1801 idx -= nr;
f62185d0 1802
f62185d0 1803 f2fs_lock_op(sbi);
5f281fab
JK		 1804 f2fs_drop_extent_tree(inode);
1805
0a2aa8fb
JK		 1806 ret = __exchange_data_block(inode, inode, idx,
1807 idx + delta, nr, false);
f62185d0
CY		 1808 f2fs_unlock_op(sbi);
1809 }
edc6d01b 1810 filemap_invalidate_unlock(mapping);
e4544b63 1811 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2fc2bcc8
CY		 1812 if (ret)
1813 return ret;
f62185d0 1814
6e2c64ad 1815 /* write out all moved pages, if possible */
edc6d01b 1816 filemap_invalidate_lock(mapping);
2fc2bcc8 1817 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
6e2c64ad 1818 truncate_pagecache(inode, offset);
edc6d01b 1819 filemap_invalidate_unlock(mapping);
6e2c64ad
JK		 1820
1821 if (!ret)
fc9581c8 1822 f2fs_i_size_write(inode, new_size);
f62185d0
CY		 1823 return ret;
1824}
1825
1cd75654 1826static int f2fs_expand_inode_data(struct inode *inode, loff_t offset,
fbfa2cc5
JK		 1827 loff_t len, int mode)
1828{
4081363f 1829 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d5097be5 1830 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
f9d6d059
CY		 1831 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1832 .m_may_create = true };
d147ea4a
JK		 1833 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1834 .init_gc_type = FG_GC,
1835 .should_migrate_blocks = false,
c81d5bae
JK		 1836 .err_gc_skipped = true,
1837 .nr_free_secs = 0 };
88f2cfc5 1838 pgoff_t pg_start, pg_end;
39bee2e6 1839 loff_t new_size;
e12dd7bd 1840 loff_t off_end;
88f2cfc5 1841 block_t expanded = 0;
a7de6086 1842 int err;
fbfa2cc5 1843
a7de6086
JK		 1844 err = inode_newsize_ok(inode, (len + offset));
1845 if (err)
1846 return err;
fbfa2cc5 1847
a7de6086
JK		 1848 err = f2fs_convert_inline_inode(inode);
1849 if (err)
1850 return err;
9e09fc85 1851
ba8dac35
CY		 1852 filemap_invalidate_lock(inode->i_mapping);
1853 f2fs_zero_post_eof_page(inode, offset + len);
1854 filemap_invalidate_unlock(inode->i_mapping);
1855
2c4db1a6 1856 f2fs_balance_fs(sbi, true);
2a340760 1857
88f2cfc5 1858 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
e12dd7bd 1859 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
09cbfeaf 1860 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5 1861
88f2cfc5
CY		 1862 map.m_lblk = pg_start;
1863 map.m_len = pg_end - pg_start;
e12dd7bd
JK		 1864 if (off_end)
1865 map.m_len++;
ead43275 1866
f5a53edc
JK		 1867 if (!map.m_len)
1868 return 0;
1869
1870 if (f2fs_is_pinned_file(inode)) {
074b5ea2 1871 block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
e1175f02 1872 block_t sec_len = roundup(map.m_len, sec_blks);
f5a53edc 1873
e1175f02 1874 map.m_len = sec_blks;
f5a53edc 1875next_alloc:
48ea8b20
CY		 1876 f2fs_down_write(&sbi->pin_sem);
1877
f7f8932c
CY		 1878 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1879 if (has_not_enough_free_secs(sbi, 0, 0)) {
1880 f2fs_up_write(&sbi->pin_sem);
1881 err = -ENOSPC;
1882 f2fs_warn_ratelimited(sbi,
1883 "ino:%lu, start:%lu, end:%lu, need to trigger GC to "
1884 "reclaim enough free segment when checkpoint is enabled",
1885 inode->i_ino, pg_start, pg_end);
1886 goto out_err;
1887 }
1888 }
1889
fa08972b
DJ		 1890 if (has_not_enough_free_secs(sbi, 0, f2fs_sb_has_blkzoned(sbi) ?
1891 ZONED_PIN_SEC_REQUIRED_COUNT :
f5a53edc 1892 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
e4544b63 1893 f2fs_down_write(&sbi->gc_lock);
9bf1dcbd 1894 stat_inc_gc_call_count(sbi, FOREGROUND);
d147ea4a 1895 err = f2fs_gc(sbi, &gc_control);
48ea8b20
CY		 1896 if (err && err != -ENODATA) {
1897 f2fs_up_write(&sbi->pin_sem);
f5a53edc 1898 goto out_err;
48ea8b20 1899 }
f5a53edc
JK		 1900 }
1901
9703d69d
DJ		 1902 err = f2fs_allocate_pinning_section(sbi);
1903 if (err) {
1904 f2fs_up_write(&sbi->pin_sem);
1905 goto out_err;
1906 }
fd612648 1907
d0b9e42a 1908 map.m_seg_type = CURSEG_COLD_DATA_PINNED;
cd8fc522 1909 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO);
d4dd19ec 1910 file_dont_truncate(inode);
d0b9e42a 1911
e4544b63 1912 f2fs_up_write(&sbi->pin_sem);
cad3836f 1913
88f2cfc5 1914 expanded += map.m_len;
e1175f02 1915 sec_len -= map.m_len;
f5a53edc 1916 map.m_lblk += map.m_len;
e1175f02 1917 if (!err && sec_len)
f5a53edc
JK		 1918 goto next_alloc;
1919
88f2cfc5 1920 map.m_len = expanded;
f5a53edc 1921 } else {
cd8fc522 1922 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO);
88f2cfc5 1923 expanded = map.m_len;
f5a53edc
JK		 1924 }
1925out_err:
a7de6086 1926 if (err) {
e12dd7bd 1927 pgoff_t last_off;
fbfa2cc5 1928
88f2cfc5 1929 if (!expanded)
a7de6086 1930 return err;
98397ff3 1931
88f2cfc5 1932 last_off = pg_start + expanded - 1;
e12dd7bd
JK		 1933
1934 /* update new size to the failed position */
1061fd48 1935 new_size = (last_off == pg_end) ? offset + len :
e12dd7bd
JK		 1936 (loff_t)(last_off + 1) << PAGE_SHIFT;
1937 } else {
1938 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
fbfa2cc5
JK		 1939 }
1940
e8ed90a6
CY		 1941 if (new_size > i_size_read(inode)) {
1942 if (mode & FALLOC_FL_KEEP_SIZE)
1943 file_set_keep_isize(inode);
1944 else
1945 f2fs_i_size_write(inode, new_size);
1946 }
fbfa2cc5 1947
a7de6086 1948 return err;
fbfa2cc5
JK		 1949}
1950
1951static long f2fs_fallocate(struct file *file, int mode,
1952 loff_t offset, loff_t len)
1953{
6131ffaa 1954 struct inode *inode = file_inode(file);
587c0a42 1955 long ret = 0;
fbfa2cc5 1956
1f227a3e
JK		 1957 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1958 return -EIO;
00e09c0b
CY		 1959 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1960 return -ENOSPC;
128d333f 1961 if (!f2fs_is_compress_backend_ready(inode) || IS_DEVICE_ALIASING(inode))
4c8ff709 1962 return -EOPNOTSUPP;
1f227a3e 1963
c998012b
CY		 1964 /* f2fs only support ->fallocate for regular file */
1965 if (!S_ISREG(inode->i_mode))
1966 return -EINVAL;
1967
62230e0d 1968 if (IS_ENCRYPTED(inode) &&
f62185d0 1969 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
fcc85a4d
JK		 1970 return -EOPNOTSUPP;
1971
b4ace337 1972 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
f62185d0
CY		 1973 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1974 FALLOC_FL_INSERT_RANGE))
fbfa2cc5
JK		 1975 return -EOPNOTSUPP;
1976
5955102c 1977 inode_lock(inode);
3375f696 1978
278a6253
CY		 1979 /*
1980 * Pinned file should not support partial truncation since the block
1981 * can be used by applications.
1982 */
1983 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1984 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1985 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) {
1986 ret = -EOPNOTSUPP;
1987 goto out;
1988 }
1989
958ed929
CY		 1990 ret = file_modified(file);
1991 if (ret)
1992 goto out;
1993
96cfeb03
CY		 1994 /*
1995 * wait for inflight dio, blocks should be removed after IO
1996 * completion.
1997 */
1998 inode_dio_wait(inode);
1999
587c0a42
TY		 2000 if (mode & FALLOC_FL_PUNCH_HOLE) {
2001 if (offset >= inode->i_size)
2002 goto out;
2003
1cd75654 2004 ret = f2fs_punch_hole(inode, offset, len);
b4ace337
CY		 2005 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
2006 ret = f2fs_collapse_range(inode, offset, len);
75cd4e09
CY		 2007 } else if (mode & FALLOC_FL_ZERO_RANGE) {
2008 ret = f2fs_zero_range(inode, offset, len, mode);
f62185d0
CY		 2009 } else if (mode & FALLOC_FL_INSERT_RANGE) {
2010 ret = f2fs_insert_range(inode, offset, len);
b4ace337 2011 } else {
1cd75654 2012 ret = f2fs_expand_inode_data(inode, offset, len, mode);
b4ace337 2013 }
fbfa2cc5 2014
3af60a49 2015 if (!ret) {
11cc6426 2016 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
7c45729a 2017 f2fs_mark_inode_dirty_sync(inode, false);
d0239e1b 2018 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3af60a49 2019 }
3375f696 2020
587c0a42 2021out:
5955102c 2022 inode_unlock(inode);
3375f696 2023
c01e2853 2024 trace_f2fs_fallocate(inode, mode, offset, len, ret);
fbfa2cc5
JK		 2025 return ret;
2026}
2027
1e84371f
JK		 2028static int f2fs_release_file(struct inode *inode, struct file *filp)
2029{
de5307e4 2030 /*
146949de 2031 * f2fs_release_file is called at every close calls. So we should
de5307e4
JK		 2032 * not drop any inmemory pages by close called by other process.
2033 */
2034 if (!(filp->f_mode & FMODE_WRITE) ||
2035 atomic_read(&inode->i_writecount) != 1)
2036 return 0;
2037
a46bebd5 2038 inode_lock(inode);
e53f8643 2039 f2fs_abort_atomic_write(inode, true);
a46bebd5
DJ		 2040 inode_unlock(inode);
2041
1e84371f
JK		 2042 return 0;
2043}
2044
7a10f017 2045static int f2fs_file_flush(struct file *file, fl_owner_t id)
fbfa2cc5 2046{
7a10f017
JK		 2047 struct inode *inode = file_inode(file);
2048
2049 /*
2050 * If the process doing a transaction is crashed, we should do
2051 * roll-back. Otherwise, other reader/write can see corrupted database
2052 * until all the writers close its file. Since this should be done
2053 * before dropping file lock, it needs to do in ->flush.
2054 */
ae267fc1 2055 if (F2FS_I(inode)->atomic_write_task == current &&
a46bebd5
DJ		 2056 (current->flags & PF_EXITING)) {
2057 inode_lock(inode);
3db1de0e 2058 f2fs_abort_atomic_write(inode, true);
a46bebd5
DJ		 2059 inode_unlock(inode);
2060 }
2061
7a10f017 2062 return 0;
fbfa2cc5
JK		 2063}
2064
36098557 2065static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
2c1d0305
CY		 2066{
2067 struct f2fs_inode_info *fi = F2FS_I(inode);
99eabb91
JK		 2068 u32 masked_flags = fi->i_flags & mask;
2069
a7531039
JK		 2070 /* mask can be shrunk by flags_valid selector */
2071 iflags &= mask;
2c1d0305
CY		 2072
2073 /* Is it quota file? Do not allow user to mess with it */
2074 if (IS_NOQUOTA(inode))
2075 return -EPERM;
2076
99eabb91 2077 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
2c2eb7a3
DR		 2078 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
2079 return -EOPNOTSUPP;
2080 if (!f2fs_empty_dir(inode))
2081 return -ENOTEMPTY;
2082 }
2083
4c8ff709
CY		 2084 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
2085 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
2086 return -EOPNOTSUPP;
2087 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
2088 return -EINVAL;
2089 }
2090
99eabb91 2091 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
aa576970 2092 if (masked_flags & F2FS_COMPR_FL) {
78134d03 2093 if (!f2fs_disable_compressed_file(inode))
2536ac68 2094 return -EINVAL;
8ee236dc 2095 } else {
a995627e
JK		 2096 /* try to convert inline_data to support compression */
2097 int err = f2fs_convert_inline_inode(inode);
2098 if (err)
2099 return err;
b5ab3276 2100
7309871c 2101 f2fs_down_write(&fi->i_sem);
b5ab3276
CY		 2102 if (!f2fs_may_compress(inode) ||
2103 (S_ISREG(inode->i_mode) &&
2104 F2FS_HAS_BLOCKS(inode))) {
7309871c 2105 f2fs_up_write(&fi->i_sem);
519a5a2f 2106 return -EINVAL;
b5ab3276
CY		 2107 }
2108 err = set_compress_context(inode);
7309871c 2109 f2fs_up_write(&fi->i_sem);
b5ab3276
CY		 2110
2111 if (err)
2112 return err;
4c8ff709
CY		 2113 }
2114 }
4c8ff709 2115
d5e5efa2 2116 fi->i_flags = iflags | (fi->i_flags & ~mask);
4c8ff709
CY		 2117 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
2118 (fi->i_flags & F2FS_NOCOMP_FL));
2c1d0305 2119
59c84408 2120 if (fi->i_flags & F2FS_PROJINHERIT_FL)
2c1d0305
CY		 2121 set_inode_flag(inode, FI_PROJ_INHERIT);
2122 else
2123 clear_inode_flag(inode, FI_PROJ_INHERIT);
2124
c62ebd35 2125 inode_set_ctime_current(inode);
2c1d0305 2126 f2fs_set_inode_flags(inode);
b32e0190 2127 f2fs_mark_inode_dirty_sync(inode, true);
2c1d0305
CY		 2128 return 0;
2129}
2130
9b1bb01c 2131/* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
36098557
EB		 2132
2133/*
2134 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
2135 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
2136 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add
2137 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
9b1bb01c
MS		 2138 *
2139 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
2140 * FS_IOC_FSSETXATTR is done by the VFS.
36098557
EB		 2141 */
2142
2143static const struct {
2144 u32 iflag;
2145 u32 fsflag;
2146} f2fs_fsflags_map[] = {
4c8ff709 2147 { F2FS_COMPR_FL, FS_COMPR_FL },
36098557
EB		 2148 { F2FS_SYNC_FL, FS_SYNC_FL },
2149 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL },
2150 { F2FS_APPEND_FL, FS_APPEND_FL },
2151 { F2FS_NODUMP_FL, FS_NODUMP_FL },
2152 { F2FS_NOATIME_FL, FS_NOATIME_FL },
4c8ff709 2153 { F2FS_NOCOMP_FL, FS_NOCOMP_FL },
36098557
EB		 2154 { F2FS_INDEX_FL, FS_INDEX_FL },
2155 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL },
2156 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL },
2c2eb7a3 2157 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
36098557
EB		 2158};
2159
2160#define F2FS_GETTABLE_FS_FL ( \
4c8ff709 2161 FS_COMPR_FL | \
36098557
EB		 2162 FS_SYNC_FL | \
2163 FS_IMMUTABLE_FL | \
2164 FS_APPEND_FL | \
2165 FS_NODUMP_FL | \
2166 FS_NOATIME_FL | \
4c8ff709 2167 FS_NOCOMP_FL | \
36098557
EB		 2168 FS_INDEX_FL | \
2169 FS_DIRSYNC_FL | \
2170 FS_PROJINHERIT_FL | \
2171 FS_ENCRYPT_FL | \
2172 FS_INLINE_DATA_FL | \
95ae251f 2173 FS_NOCOW_FL | \
fbc246a1 2174 FS_VERITY_FL | \
2c2eb7a3 2175 FS_CASEFOLD_FL)
36098557
EB		 2176
2177#define F2FS_SETTABLE_FS_FL ( \
4c8ff709 2178 FS_COMPR_FL | \
36098557
EB		 2179 FS_SYNC_FL | \
2180 FS_IMMUTABLE_FL | \
2181 FS_APPEND_FL | \
2182 FS_NODUMP_FL | \
2183 FS_NOATIME_FL | \
4c8ff709 2184 FS_NOCOMP_FL | \
36098557 2185 FS_DIRSYNC_FL | \
2c2eb7a3
DR		 2186 FS_PROJINHERIT_FL | \
2187 FS_CASEFOLD_FL)
36098557
EB		 2188
2189/* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
2190static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
2191{
2192 u32 fsflags = 0;
2193 int i;
2194
2195 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2196 if (iflags & f2fs_fsflags_map[i].iflag)
2197 fsflags |= f2fs_fsflags_map[i].fsflag;
2198
2199 return fsflags;
2200}
2201
2202/* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
2203static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
2204{
2205 u32 iflags = 0;
2206 int i;
2207
2208 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
2209 if (fsflags & f2fs_fsflags_map[i].fsflag)
2210 iflags |= f2fs_fsflags_map[i].iflag;
2211
2212 return iflags;
2213}
2214
d49f3e89
CY		 2215static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
2216{
2217 struct inode *inode = file_inode(filp);
2218
2219 return put_user(inode->i_generation, (int __user *)arg);
2220}
2221
41e8f85a 2222static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
88b88a66
JK		 2223{
2224 struct inode *inode = file_inode(filp);
f2d40141 2225 struct mnt_idmap *idmap = file_mnt_idmap(filp);
743b620c
JK		 2226 struct f2fs_inode_info *fi = F2FS_I(inode);
2227 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4d8d45df 2228 loff_t isize;
f4c9c743 2229 int ret;
88b88a66 2230
4f5a100f
JH		 2231 if (!(filp->f_mode & FMODE_WRITE))
2232 return -EBADF;
2233
01beba79 2234 if (!inode_owner_or_capable(idmap, inode))
88b88a66
JK		 2235 return -EACCES;
2236
e811898c
JK		 2237 if (!S_ISREG(inode->i_mode))
2238 return -EINVAL;
2239
038d0698
CY		 2240 if (filp->f_flags & O_DIRECT)
2241 return -EINVAL;
2242
7fb17fe4
CY		 2243 ret = mnt_want_write_file(filp);
2244 if (ret)
2245 return ret;
2246
0fac558b
CY		 2247 inode_lock(inode);
2248
c644af13
DJ		 2249 if (!f2fs_disable_compressed_file(inode) ||
2250 f2fs_is_pinned_file(inode)) {
9b56adcf
FC		 2251 ret = -EINVAL;
2252 goto out;
2253 }
4c8ff709 2254
3db1de0e 2255 if (f2fs_is_atomic_file(inode))
7fb17fe4 2256 goto out;
88b88a66 2257
f4c9c743
CY		 2258 ret = f2fs_convert_inline_inode(inode);
2259 if (ret)
7fb17fe4 2260 goto out;
88b88a66 2261
054cb289 2262 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
b2c160f4 2263 f2fs_down_write(&fi->i_gc_rwsem[READ]);
6f8d4455 2264
31867b23
JK		 2265 /*
2266 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2267 * f2fs_is_atomic_file.
2268 */
2269 if (get_dirty_pages(inode))
054cb289 2270 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
dcbb4c10 2271 inode->i_ino, get_dirty_pages(inode));
c27753d6 2272 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
b2c160f4
CY		 2273 if (ret)
2274 goto out_unlock;
31867b23 2275
a46bebd5
DJ		 2276 /* Check if the inode already has a COW inode */
2277 if (fi->cow_inode == NULL) {
2278 /* Create a COW inode for atomic write */
8c1b7879
YG		 2279 struct dentry *dentry = file_dentry(filp);
2280 struct inode *dir = d_inode(dentry->d_parent);
3db1de0e 2281
8c1b7879 2282 ret = f2fs_get_tmpfile(idmap, dir, &fi->cow_inode);
b2c160f4
CY		 2283 if (ret)
2284 goto out_unlock;
a46bebd5
DJ		 2285
2286 set_inode_flag(fi->cow_inode, FI_COW_FILE);
2287 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
f18d0076
SJ		 2288
2289 /* Set the COW inode's atomic_inode to the atomic inode */
2290 F2FS_I(fi->cow_inode)->atomic_inode = inode;
a46bebd5
DJ		 2291 } else {
2292 /* Reuse the already created COW inode */
ebd3309a
CY		 2293 f2fs_bug_on(sbi, get_dirty_pages(fi->cow_inode));
2294
2295 invalidate_mapping_pages(fi->cow_inode->i_mapping, 0, -1);
2296
b851ee6b 2297 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
b2c160f4
CY		 2298 if (ret)
2299 goto out_unlock;
6f8d4455 2300 }
4d8d45df
DJ		 2301
2302 f2fs_write_inode(inode, NULL);
31867b23 2303
b4dac120 2304 stat_inc_atomic_inode(inode);
743b620c 2305
054afda9 2306 set_inode_flag(inode, FI_ATOMIC_FILE);
41e8f85a
DJ		 2307
2308 isize = i_size_read(inode);
2309 fi->original_i_size = isize;
2310 if (truncate) {
2311 set_inode_flag(inode, FI_ATOMIC_REPLACE);
2312 truncate_inode_pages_final(inode->i_mapping);
2313 f2fs_i_size_write(inode, 0);
2314 isize = 0;
2315 }
2316 f2fs_i_size_write(fi->cow_inode, isize);
2317
b2c160f4
CY		 2318out_unlock:
2319 f2fs_up_write(&fi->i_gc_rwsem[READ]);
054cb289 2320 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
b2c160f4
CY		 2321 if (ret)
2322 goto out;
684ca7e5 2323
054cb289
YY		 2324 f2fs_update_time(sbi, REQ_TIME);
2325 fi->atomic_write_task = current;
26a28a0c 2326 stat_update_max_atomic_write(inode);
f8e2f32b 2327 fi->atomic_write_cnt = 0;
684ca7e5 2328out:
0fac558b 2329 inode_unlock(inode);
7fb17fe4 2330 mnt_drop_write_file(filp);
c27753d6 2331 return ret;
88b88a66
JK		 2332}
2333
2334static int f2fs_ioc_commit_atomic_write(struct file *filp)
2335{
2336 struct inode *inode = file_inode(filp);
01beba79 2337 struct mnt_idmap *idmap = file_mnt_idmap(filp);
88b88a66
JK		 2338 int ret;
2339
4f5a100f
JH		 2340 if (!(filp->f_mode & FMODE_WRITE))
2341 return -EBADF;
2342
01beba79 2343 if (!inode_owner_or_capable(idmap, inode))
88b88a66
JK		 2344 return -EACCES;
2345
2346 ret = mnt_want_write_file(filp);
2347 if (ret)
2348 return ret;
2349
6f8d4455 2350 f2fs_balance_fs(F2FS_I_SB(inode), true);
0fac558b 2351
6f8d4455 2352 inode_lock(inode);
1dc0f899 2353
6282adbf 2354 if (f2fs_is_atomic_file(inode)) {
3db1de0e 2355 ret = f2fs_commit_atomic_write(inode);
743b620c 2356 if (!ret)
4d8d45df
DJ		 2357 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2358
2359 f2fs_abort_atomic_write(inode, ret);
26a28a0c 2360 } else {
774e1b78 2361 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
6282adbf 2362 }
4d8d45df 2363
0fac558b 2364 inode_unlock(inode);
1e84371f
JK		 2365 mnt_drop_write_file(filp);
2366 return ret;
2367}
2368
23339e57
DJ		 2369static int f2fs_ioc_abort_atomic_write(struct file *filp)
2370{
2371 struct inode *inode = file_inode(filp);
01beba79 2372 struct mnt_idmap *idmap = file_mnt_idmap(filp);
23339e57
DJ		 2373 int ret;
2374
4f5a100f
JH		 2375 if (!(filp->f_mode & FMODE_WRITE))
2376 return -EBADF;
2377
01beba79 2378 if (!inode_owner_or_capable(idmap, inode))
23339e57
DJ		 2379 return -EACCES;
2380
2381 ret = mnt_want_write_file(filp);
2382 if (ret)
2383 return ret;
2384
2385 inode_lock(inode);
2386
e53f8643 2387 f2fs_abort_atomic_write(inode, true);
23339e57
DJ		 2388
2389 inode_unlock(inode);
2390
2391 mnt_drop_write_file(filp);
2392 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2393 return ret;
2394}
2395
ee745e47 2396int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
c7f114d8 2397 bool readonly, bool need_lock)
1abff93d 2398{
1abff93d 2399 struct super_block *sb = sbi->sb;
2a96d8ad 2400 int ret = 0;
1abff93d 2401
ee745e47 2402 switch (flag) {
1abff93d 2403 case F2FS_GOING_DOWN_FULLSYNC:
982c3b30 2404 ret = bdev_freeze(sb->s_bdev);
040f04bd 2405 if (ret)
d027c484 2406 goto out;
a9cfee0e 2407 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
982c3b30 2408 bdev_thaw(sb->s_bdev);
1abff93d
JK		 2409 break;
2410 case F2FS_GOING_DOWN_METASYNC:
2411 /* do checkpoint only */
d027c484 2412 ret = f2fs_sync_fs(sb, 1);
7d009e04
CY		 2413 if (ret) {
2414 if (ret == -EIO)
2415 ret = 0;
d027c484 2416 goto out;
7d009e04 2417 }
a9cfee0e 2418 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
1abff93d
JK		 2419 break;
2420 case F2FS_GOING_DOWN_NOSYNC:
a9cfee0e 2421 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
1abff93d 2422 break;
c912a829 2423 case F2FS_GOING_DOWN_METAFLUSH:
4d57b86d 2424 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
a9cfee0e 2425 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
c912a829 2426 break;
0cd6d9b0
JK		 2427 case F2FS_GOING_DOWN_NEED_FSCK:
2428 set_sbi_flag(sbi, SBI_NEED_FSCK);
db610a64
JK		 2429 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2430 set_sbi_flag(sbi, SBI_IS_DIRTY);
0cd6d9b0
JK		 2431 /* do checkpoint only */
2432 ret = f2fs_sync_fs(sb, 1);
7d009e04
CY		 2433 if (ret == -EIO)
2434 ret = 0;
db610a64 2435 goto out;
1abff93d 2436 default:
7fb17fe4
CY		 2437 ret = -EINVAL;
2438 goto out;
1abff93d 2439 }
7950e9ac 2440
ee745e47
CY		 2441 if (readonly)
2442 goto out;
2443
7b0033db
LL		 2444 /*
2445 * grab sb->s_umount to avoid racing w/ remount() and other shutdown
2446 * paths.
2447 */
c7f114d8 2448 if (need_lock)
7b0033db 2449 down_write(&sbi->sb->s_umount);
c7f114d8 2450
4d57b86d
CY		 2451 f2fs_stop_gc_thread(sbi);
2452 f2fs_stop_discard_thread(sbi);
7950e9ac 2453
4d57b86d 2454 f2fs_drop_discard_cmd(sbi);
7950e9ac
CY		 2455 clear_opt(sbi, DISCARD);
2456
c7f114d8 2457 if (need_lock)
7b0033db 2458 up_write(&sbi->sb->s_umount);
c7f114d8 2459
d0239e1b 2460 f2fs_update_time(sbi, REQ_TIME);
7fb17fe4 2461out:
559e87c4 2462
ee745e47
CY		 2463 trace_f2fs_shutdown(sbi, flag, ret);
2464
2465 return ret;
2466}
2467
2468static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2469{
2470 struct inode *inode = file_inode(filp);
2471 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2472 __u32 in;
2473 int ret;
2474 bool need_drop = false, readonly = false;
2475
2476 if (!capable(CAP_SYS_ADMIN))
2477 return -EPERM;
2478
2479 if (get_user(in, (__u32 __user *)arg))
2480 return -EFAULT;
2481
2482 if (in != F2FS_GOING_DOWN_FULLSYNC) {
2483 ret = mnt_want_write_file(filp);
2484 if (ret) {
2485 if (ret != -EROFS)
2486 return ret;
2487
2488 /* fallback to nosync shutdown for readonly fs */
2489 in = F2FS_GOING_DOWN_NOSYNC;
2490 readonly = true;
2491 } else {
2492 need_drop = true;
2493 }
2494 }
2495
c7f114d8 2496 ret = f2fs_do_shutdown(sbi, in, readonly, true);
ee745e47
CY		 2497
2498 if (need_drop)
2499 mnt_drop_write_file(filp);
559e87c4 2500
7fb17fe4 2501 return ret;
1abff93d
JK		 2502}
2503
cf7cd17c 2504static int f2fs_keep_noreuse_range(struct inode *inode,
ef0c333c
JK		 2505 loff_t offset, loff_t len)
2506{
2507 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2508 u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
2509 u64 start, end;
cf7cd17c 2510 int ret = 0;
ef0c333c
JK		 2511
2512 if (!S_ISREG(inode->i_mode))
cf7cd17c 2513 return 0;
ef0c333c
JK		 2514
2515 if (offset >= max_bytes || len > max_bytes ||
2516 (offset + len) > max_bytes)
cf7cd17c 2517 return 0;
ef0c333c
JK		 2518
2519 start = offset >> PAGE_SHIFT;
2520 end = DIV_ROUND_UP(offset + len, PAGE_SIZE);
2521
2522 inode_lock(inode);
2523 if (f2fs_is_atomic_file(inode)) {
2524 inode_unlock(inode);
cf7cd17c 2525 return 0;
ef0c333c
JK		 2526 }
2527
2528 spin_lock(&sbi->inode_lock[DONATE_INODE]);
2529 /* let's remove the range, if len = 0 */
2530 if (!len) {
2531 if (!list_empty(&F2FS_I(inode)->gdonate_list)) {
2532 list_del_init(&F2FS_I(inode)->gdonate_list);
2533 sbi->donate_files--;
cf7cd17c
DJ		 2534 if (is_inode_flag_set(inode, FI_DONATE_FINISHED))
2535 ret = -EALREADY;
2536 else
2537 set_inode_flag(inode, FI_DONATE_FINISHED);
2538 } else
2539 ret = -ENOENT;
ef0c333c
JK		 2540 } else {
2541 if (list_empty(&F2FS_I(inode)->gdonate_list)) {
2542 list_add_tail(&F2FS_I(inode)->gdonate_list,
2543 &sbi->inode_list[DONATE_INODE]);
2544 sbi->donate_files++;
2545 } else {
2546 list_move_tail(&F2FS_I(inode)->gdonate_list,
2547 &sbi->inode_list[DONATE_INODE]);
2548 }
2549 F2FS_I(inode)->donate_start = start;
2550 F2FS_I(inode)->donate_end = end - 1;
cf7cd17c 2551 clear_inode_flag(inode, FI_DONATE_FINISHED);
ef0c333c
JK		 2552 }
2553 spin_unlock(&sbi->inode_lock[DONATE_INODE]);
2554 inode_unlock(inode);
cf7cd17c
DJ		 2555
2556 return ret;
ef0c333c
JK		 2557}
2558
52656e6c
JK		 2559static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2560{
2561 struct inode *inode = file_inode(filp);
2562 struct super_block *sb = inode->i_sb;
52656e6c
JK		 2563 struct fstrim_range range;
2564 int ret;
4b2fecc8 2565
52656e6c
JK		 2566 if (!capable(CAP_SYS_ADMIN))
2567 return -EPERM;
4b2fecc8 2568
7d20c8ab 2569 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
52656e6c 2570 return -EOPNOTSUPP;
4b2fecc8 2571
52656e6c
JK		 2572 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2573 sizeof(range)))
2574 return -EFAULT;
4b2fecc8 2575
7fb17fe4
CY		 2576 ret = mnt_want_write_file(filp);
2577 if (ret)
2578 return ret;
2579
52656e6c 2580 range.minlen = max((unsigned int)range.minlen,
7b47ef52 2581 bdev_discard_granularity(sb->s_bdev));
52656e6c 2582 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
7fb17fe4 2583 mnt_drop_write_file(filp);
52656e6c
JK		 2584 if (ret < 0)
2585 return ret;
4b2fecc8 2586
52656e6c
JK		 2587 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2588 sizeof(range)))
2589 return -EFAULT;
d0239e1b 2590 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
52656e6c
JK		 2591 return 0;
2592}
2593
f424f664
JK		 2594static bool uuid_is_nonzero(__u8 u[16])
2595{
2596 int i;
2597
2598 for (i = 0; i < 16; i++)
2599 if (u[i])
2600 return true;
2601 return false;
2602}
2603
2604static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2605{
f424f664 2606 struct inode *inode = file_inode(filp);
fa18d87c 2607 int ret;
f424f664 2608
7beb01f7 2609 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
ead710b7
CY		 2610 return -EOPNOTSUPP;
2611
fa18d87c 2612 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
d0239e1b 2613 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
fa18d87c 2614 return ret;
f424f664
JK		 2615}
2616
2617static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2618{
7beb01f7 2619 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
ead710b7 2620 return -EOPNOTSUPP;
db717d8e 2621 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
f424f664
JK		 2622}
2623
2624static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2625{
2626 struct inode *inode = file_inode(filp);
2627 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5eaac835 2628 u8 encrypt_pw_salt[16];
f424f664
JK		 2629 int err;
2630
7beb01f7 2631 if (!f2fs_sb_has_encrypt(sbi))
f424f664
JK		 2632 return -EOPNOTSUPP;
2633
f424f664
JK		 2634 err = mnt_want_write_file(filp);
2635 if (err)
2636 return err;
2637
e4544b63 2638 f2fs_down_write(&sbi->sb_lock);
d0d3f1b3
CY		 2639
2640 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2641 goto got_it;
2642
f424f664
JK		 2643 /* update superblock with uuid */
2644 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2645
c5bda1c8 2646 err = f2fs_commit_super(sbi, false);
f424f664
JK		 2647 if (err) {
2648 /* undo new data */
2649 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
d0d3f1b3 2650 goto out_err;
f424f664
JK		 2651 }
2652got_it:
5eaac835 2653 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16);
d0d3f1b3 2654out_err:
e4544b63 2655 f2fs_up_write(&sbi->sb_lock);
d0d3f1b3 2656 mnt_drop_write_file(filp);
5eaac835
CY		 2657
2658 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16))
2659 err = -EFAULT;
2660
d0d3f1b3 2661 return err;
f424f664
JK		 2662}
2663
8ce589c7
EB		 2664static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2665 unsigned long arg)
2666{
2667 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2668 return -EOPNOTSUPP;
2669
2670 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2671}
2672
2673static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2674{
2675 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2676 return -EOPNOTSUPP;
2677
2678 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2679}
2680
2681static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2682{
2683 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2684 return -EOPNOTSUPP;
2685
2686 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2687}
2688
2689static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2690 unsigned long arg)
2691{
2692 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2693 return -EOPNOTSUPP;
2694
2695 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2696}
2697
2698static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2699 unsigned long arg)
2700{
2701 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2702 return -EOPNOTSUPP;
2703
2704 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2705}
2706
ee446e1a
EB		 2707static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2708{
2709 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2710 return -EOPNOTSUPP;
2711
2712 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2713}
2714
c1c1b583
CY		 2715static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2716{
2717 struct inode *inode = file_inode(filp);
2718 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d147ea4a
JK		 2719 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2720 .no_bg_gc = false,
c81d5bae
JK		 2721 .should_migrate_blocks = false,
2722 .nr_free_secs = 0 };
d530d4d8 2723 __u32 sync;
7fb17fe4 2724 int ret;
c1c1b583
CY		 2725
2726 if (!capable(CAP_SYS_ADMIN))
2727 return -EPERM;
2728
d530d4d8 2729 if (get_user(sync, (__u32 __user *)arg))
c1c1b583
CY		 2730 return -EFAULT;
2731
d530d4d8
CY		 2732 if (f2fs_readonly(sbi->sb))
2733 return -EROFS;
c1c1b583 2734
7fb17fe4
CY		 2735 ret = mnt_want_write_file(filp);
2736 if (ret)
2737 return ret;
2738
d530d4d8 2739 if (!sync) {
e4544b63 2740 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
7fb17fe4
CY		 2741 ret = -EBUSY;
2742 goto out;
2743 }
d530d4d8 2744 } else {
e4544b63 2745 f2fs_down_write(&sbi->gc_lock);
c1c1b583
CY		 2746 }
2747
d147ea4a
JK		 2748 gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2749 gc_control.err_gc_skipped = sync;
9bf1dcbd 2750 stat_inc_gc_call_count(sbi, FOREGROUND);
d147ea4a 2751 ret = f2fs_gc(sbi, &gc_control);
7fb17fe4
CY		 2752out:
2753 mnt_drop_write_file(filp);
2754 return ret;
c1c1b583
CY		 2755}
2756
34178b1b 2757static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
34dc77ad 2758{
34178b1b 2759 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
d147ea4a
JK		 2760 struct f2fs_gc_control gc_control = {
2761 .init_gc_type = range->sync ? FG_GC : BG_GC,
2762 .no_bg_gc = false,
2763 .should_migrate_blocks = false,
c81d5bae
JK		 2764 .err_gc_skipped = range->sync,
2765 .nr_free_secs = 0 };
34dc77ad
JK		 2766 u64 end;
2767 int ret;
2768
2769 if (!capable(CAP_SYS_ADMIN))
2770 return -EPERM;
34dc77ad
JK		 2771 if (f2fs_readonly(sbi->sb))
2772 return -EROFS;
2773
34178b1b
CY		 2774 end = range->start + range->len;
2775 if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
fbbf7799 2776 end >= MAX_BLKADDR(sbi))
b82f6e34 2777 return -EINVAL;
b82f6e34 2778
34dc77ad
JK		 2779 ret = mnt_want_write_file(filp);
2780 if (ret)
2781 return ret;
2782
34dc77ad 2783do_more:
34178b1b 2784 if (!range->sync) {
e4544b63 2785 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
34dc77ad
JK		 2786 ret = -EBUSY;
2787 goto out;
2788 }
2789 } else {
e4544b63 2790 f2fs_down_write(&sbi->gc_lock);
34dc77ad
JK		 2791 }
2792
d147ea4a 2793 gc_control.victim_segno = GET_SEGNO(sbi, range->start);
9bf1dcbd 2794 stat_inc_gc_call_count(sbi, FOREGROUND);
d147ea4a 2795 ret = f2fs_gc(sbi, &gc_control);
97767500
QZ		 2796 if (ret) {
2797 if (ret == -EBUSY)
2798 ret = -EAGAIN;
2799 goto out;
2800 }
074b5ea2 2801 range->start += CAP_BLKS_PER_SEC(sbi);
34178b1b 2802 if (range->start <= end)
34dc77ad
JK		 2803 goto do_more;
2804out:
2805 mnt_drop_write_file(filp);
2806 return ret;
2807}
2808
34178b1b
CY		 2809static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2810{
2811 struct f2fs_gc_range range;
2812
2813 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2814 sizeof(range)))
2815 return -EFAULT;
2816 return __f2fs_ioc_gc_range(filp, &range);
2817}
2818
ddf1eca4 2819static int f2fs_ioc_write_checkpoint(struct file *filp)
456b88e4
CY		 2820{
2821 struct inode *inode = file_inode(filp);
2822 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
7fb17fe4 2823 int ret;
456b88e4
CY		 2824
2825 if (!capable(CAP_SYS_ADMIN))
2826 return -EPERM;
2827
2828 if (f2fs_readonly(sbi->sb))
2829 return -EROFS;
2830
4354994f 2831 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
dcbb4c10 2832 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
4354994f
DR		 2833 return -EINVAL;
2834 }
2835
7fb17fe4
CY		 2836 ret = mnt_want_write_file(filp);
2837 if (ret)
2838 return ret;
2839
2840 ret = f2fs_sync_fs(sbi->sb, 1);
2841
2842 mnt_drop_write_file(filp);
2843 return ret;
456b88e4
CY		 2844}
2845
d323d005
CY		 2846static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2847 struct file *filp,
2848 struct f2fs_defragment *range)
2849{
2850 struct inode *inode = file_inode(filp);
f3d98e74 2851 struct f2fs_map_blocks map = { .m_next_extent = NULL,
5f029c04 2852 .m_seg_type = NO_CHECK_TYPE,
f4f0b677 2853 .m_may_create = false };
fe59109a 2854 struct extent_info ei = {};
f3d98e74 2855 pgoff_t pg_start, pg_end, next_pgofs;
d323d005 2856 unsigned int total = 0, sec_num;
d323d005
CY		 2857 block_t blk_end = 0;
2858 bool fragmented = false;
2859 int err;
2860
2c4db1a6 2861 f2fs_balance_fs(sbi, true);
d323d005 2862
5955102c 2863 inode_lock(inode);
19708015
ZN		 2864 pg_start = range->start >> PAGE_SHIFT;
2865 pg_end = min_t(pgoff_t,
2866 (range->start + range->len) >> PAGE_SHIFT,
2867 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
d323d005 2868
bfe5c026
CY		 2869 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) ||
2870 f2fs_is_atomic_file(inode)) {
7cd2e5f7
YL		 2871 err = -EINVAL;
2872 goto unlock_out;
2873 }
2874
1018a546
CY		 2875 /* if in-place-update policy is enabled, don't waste time here */
2876 set_inode_flag(inode, FI_OPU_WRITE);
2877 if (f2fs_should_update_inplace(inode, NULL)) {
2878 err = -EINVAL;
2879 goto out;
2880 }
2881
d323d005 2882 /* writeback all dirty pages in the range */
19708015
ZN		 2883 err = filemap_write_and_wait_range(inode->i_mapping,
2884 pg_start << PAGE_SHIFT,
2885 (pg_end << PAGE_SHIFT) - 1);
d323d005
CY		 2886 if (err)
2887 goto out;
2888
2889 /*
2890 * lookup mapping info in extent cache, skip defragmenting if physical
2891 * block addresses are continuous.
2892 */
e7547dac 2893 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) {
1cade98c 2894 if ((pgoff_t)ei.fofs + ei.len >= pg_end)
d323d005
CY		 2895 goto out;
2896 }
2897
2898 map.m_lblk = pg_start;
f3d98e74 2899 map.m_next_pgofs = &next_pgofs;
d323d005
CY		 2900
2901 /*
2902 * lookup mapping info in dnode page cache, skip defragmenting if all
2903 * physical block addresses are continuous even if there are hole(s)
2904 * in logical blocks.
2905 */
2906 while (map.m_lblk < pg_end) {
a1c1e9b7 2907 map.m_len = pg_end - map.m_lblk;
cd8fc522 2908 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2909 if (err)
2910 goto out;
2911
2912 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2913 map.m_lblk = next_pgofs;
d323d005
CY		 2914 continue;
2915 }
2916
25a912e5 2917 if (blk_end && blk_end != map.m_pblk)
d323d005 2918 fragmented = true;
25a912e5
CY		 2919
2920 /* record total count of block that we're going to move */
2921 total += map.m_len;
2922
d323d005
CY		 2923 blk_end = map.m_pblk + map.m_len;
2924
2925 map.m_lblk += map.m_len;
d323d005
CY		 2926 }
2927
d3a1a0e1
CY		 2928 if (!fragmented) {
2929 total = 0;
d323d005 2930 goto out;
d3a1a0e1 2931 }
d323d005 2932
074b5ea2 2933 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
d323d005
CY		 2934
2935 /*
2936 * make sure there are enough free section for LFS allocation, this can
2937 * avoid defragment running in SSR mode when free section are allocated
2938 * intensively
2939 */
7f3037a5 2940 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
d323d005
CY		 2941 err = -EAGAIN;
2942 goto out;
2943 }
2944
25a912e5
CY		 2945 map.m_lblk = pg_start;
2946 map.m_len = pg_end - pg_start;
2947 total = 0;
2948
d323d005
CY		 2949 while (map.m_lblk < pg_end) {
2950 pgoff_t idx;
2951 int cnt = 0;
2952
2953do_map:
a1c1e9b7 2954 map.m_len = pg_end - map.m_lblk;
cd8fc522 2955 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2956 if (err)
2957 goto clear_out;
2958
2959 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2960 map.m_lblk = next_pgofs;
d3a1a0e1 2961 goto check;
d323d005
CY		 2962 }
2963
1018a546 2964 set_inode_flag(inode, FI_SKIP_WRITES);
d323d005
CY		 2965
2966 idx = map.m_lblk;
a60108f7
JK		 2967 while (idx < map.m_lblk + map.m_len &&
2968 cnt < BLKS_PER_SEG(sbi)) {
c0a4bb8a 2969 struct folio *folio;
d323d005 2970
c0a4bb8a
MWO		 2971 folio = f2fs_get_lock_data_folio(inode, idx, true);
2972 if (IS_ERR(folio)) {
2973 err = PTR_ERR(folio);
d323d005
CY		 2974 goto clear_out;
2975 }
2976
c0a4bb8a 2977 f2fs_folio_wait_writeback(folio, DATA, true, true);
a4d7f2b3 2978
c0a4bb8a
MWO		 2979 folio_mark_dirty(folio);
2980 set_page_private_gcing(&folio->page);
2981 f2fs_folio_put(folio, true);
d323d005
CY		 2982
2983 idx++;
2984 cnt++;
2985 total++;
2986 }
2987
2988 map.m_lblk = idx;
d3a1a0e1 2989check:
a60108f7 2990 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi))
d323d005
CY		 2991 goto do_map;
2992
1018a546 2993 clear_inode_flag(inode, FI_SKIP_WRITES);
d323d005
CY		 2994
2995 err = filemap_fdatawrite(inode->i_mapping);
2996 if (err)
2997 goto out;
2998 }
2999clear_out:
1018a546 3000 clear_inode_flag(inode, FI_SKIP_WRITES);
d323d005 3001out:
1018a546 3002 clear_inode_flag(inode, FI_OPU_WRITE);
7cd2e5f7 3003unlock_out:
5955102c 3004 inode_unlock(inode);
d323d005 3005 if (!err)
09cbfeaf 3006 range->len = (u64)total << PAGE_SHIFT;
d323d005
CY		 3007 return err;
3008}
3009
3010static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
3011{
3012 struct inode *inode = file_inode(filp);
3013 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3014 struct f2fs_defragment range;
3015 int err;
3016
3017 if (!capable(CAP_SYS_ADMIN))
3018 return -EPERM;
3019
744e66cb 3020 if (!S_ISREG(inode->i_mode))
d323d005
CY		 3021 return -EINVAL;
3022
d7563861
KM		 3023 if (f2fs_readonly(sbi->sb))
3024 return -EROFS;
d323d005
CY		 3025
3026 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
d7563861
KM		 3027 sizeof(range)))
3028 return -EFAULT;
d323d005
CY		 3029
3030 /* verify alignment of offset & size */
d7563861
KM		 3031 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
3032 return -EINVAL;
d323d005 3033
1941d7bc 3034 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
6d1451bf 3035 max_file_blocks(inode)))
d7563861
KM		 3036 return -EINVAL;
3037
3038 err = mnt_want_write_file(filp);
3039 if (err)
3040 return err;
1941d7bc 3041
d323d005 3042 err = f2fs_defragment_range(sbi, filp, &range);
d7563861
KM		 3043 mnt_drop_write_file(filp);
3044
fa18d87c
ZN		 3045 if (range.len)
3046 f2fs_update_time(sbi, REQ_TIME);
d323d005 3047 if (err < 0)
d7563861 3048 return err;
d323d005
CY		 3049
3050 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
3051 sizeof(range)))
d7563861
KM		 3052 return -EFAULT;
3053
3054 return 0;
d323d005
CY		 3055}
3056
4dd6f977
JK		 3057static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
3058 struct file *file_out, loff_t pos_out, size_t len)
3059{
3060 struct inode *src = file_inode(file_in);
3061 struct inode *dst = file_inode(file_out);
3062 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
3063 size_t olen = len, dst_max_i_size = 0;
3064 size_t dst_osize;
3065 int ret;
3066
3067 if (file_in->f_path.mnt != file_out->f_path.mnt ||
3068 src->i_sb != dst->i_sb)
3069 return -EXDEV;
3070
3071 if (unlikely(f2fs_readonly(src->i_sb)))
3072 return -EROFS;
3073
fe8494bf
CY		 3074 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
3075 return -EINVAL;
4dd6f977 3076
62230e0d 3077 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
4dd6f977
JK		 3078 return -EOPNOTSUPP;
3079
aad1383c
DR		 3080 if (pos_out < 0 || pos_in < 0)
3081 return -EINVAL;
3082
d95fd91c
FL		 3083 if (src == dst) {
3084 if (pos_in == pos_out)
3085 return 0;
3086 if (pos_out > pos_in && pos_out < pos_in + len)
3087 return -EINVAL;
3088 }
3089
4dd6f977 3090 inode_lock(src);
20a3d61d 3091 if (src != dst) {
bb06664a
CY		 3092 ret = -EBUSY;
3093 if (!inode_trylock(dst))
3094 goto out;
20a3d61d 3095 }
4dd6f977 3096
e07230da
CY		 3097 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) ||
3098 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) {
fb9b6534
CY		 3099 ret = -EOPNOTSUPP;
3100 goto out_unlock;
3101 }
3102
bfe5c026
CY		 3103 if (f2fs_is_atomic_file(src) || f2fs_is_atomic_file(dst)) {
3104 ret = -EINVAL;
3105 goto out_unlock;
3106 }
3107
4dd6f977
JK		 3108 ret = -EINVAL;
3109 if (pos_in + len > src->i_size || pos_in + len < pos_in)
3110 goto out_unlock;
3111 if (len == 0)
3112 olen = len = src->i_size - pos_in;
3113 if (pos_in + len == src->i_size)
3114 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
3115 if (len == 0) {
3116 ret = 0;
3117 goto out_unlock;
3118 }
3119
3120 dst_osize = dst->i_size;
3121 if (pos_out + olen > dst->i_size)
3122 dst_max_i_size = pos_out + olen;
3123
3124 /* verify the end result is block aligned */
3125 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
3126 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
3127 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
3128 goto out_unlock;
3129
3130 ret = f2fs_convert_inline_inode(src);
3131 if (ret)
3132 goto out_unlock;
3133
3134 ret = f2fs_convert_inline_inode(dst);
3135 if (ret)
3136 goto out_unlock;
3137
3138 /* write out all dirty pages from offset */
3139 ret = filemap_write_and_wait_range(src->i_mapping,
3140 pos_in, pos_in + len);
3141 if (ret)
3142 goto out_unlock;
3143
3144 ret = filemap_write_and_wait_range(dst->i_mapping,
3145 pos_out, pos_out + len);
3146 if (ret)
3147 goto out_unlock;
3148
3149 f2fs_balance_fs(sbi, true);
6f8d4455 3150
e4544b63 3151 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
6f8d4455
JK		 3152 if (src != dst) {
3153 ret = -EBUSY;
e4544b63 3154 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
6f8d4455
JK		 3155 goto out_src;
3156 }
3157
4dd6f977 3158 f2fs_lock_op(sbi);
8fb9f319
ZN		 3159 ret = __exchange_data_block(src, dst, F2FS_BYTES_TO_BLK(pos_in),
3160 F2FS_BYTES_TO_BLK(pos_out),
3161 F2FS_BYTES_TO_BLK(len), false);
4dd6f977
JK		 3162
3163 if (!ret) {
3164 if (dst_max_i_size)
3165 f2fs_i_size_write(dst, dst_max_i_size);
3166 else if (dst_osize != dst->i_size)
3167 f2fs_i_size_write(dst, dst_osize);
3168 }
3169 f2fs_unlock_op(sbi);
6f8d4455
JK		 3170
3171 if (src != dst)
e4544b63 3172 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
6f8d4455 3173out_src:
e4544b63 3174 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
396d0a28
YH		 3175 if (ret)
3176 goto out_unlock;
3177
11cc6426 3178 inode_set_mtime_to_ts(src, inode_set_ctime_current(src));
396d0a28
YH		 3179 f2fs_mark_inode_dirty_sync(src, false);
3180 if (src != dst) {
11cc6426 3181 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst));
396d0a28
YH		 3182 f2fs_mark_inode_dirty_sync(dst, false);
3183 }
3184 f2fs_update_time(sbi, REQ_TIME);
3185
6f8d4455
JK		 3186out_unlock:
3187 if (src != dst)
4dd6f977 3188 inode_unlock(dst);
20a3d61d 3189out:
4dd6f977
JK		 3190 inode_unlock(src);
3191 return ret;
3192}
3193
34178b1b
CY		 3194static int __f2fs_ioc_move_range(struct file *filp,
3195 struct f2fs_move_range *range)
4dd6f977 3196{
4dd6f977
JK		 3197 int err;
3198
3199 if (!(filp->f_mode & FMODE_READ) ||
3200 !(filp->f_mode & FMODE_WRITE))
3201 return -EBADF;
3202
8152f820
AV		 3203 CLASS(fd, dst)(range->dst_fd);
3204 if (fd_empty(dst))
4dd6f977
JK		 3205 return -EBADF;
3206
8152f820
AV		 3207 if (!(fd_file(dst)->f_mode & FMODE_WRITE))
3208 return -EBADF;
4dd6f977
JK		 3209
3210 err = mnt_want_write_file(filp);
3211 if (err)
8152f820 3212 return err;
4dd6f977 3213
1da91ea8 3214 err = f2fs_move_file_range(filp, range->pos_in, fd_file(dst),
34178b1b 3215 range->pos_out, range->len);
4dd6f977
JK		 3216
3217 mnt_drop_write_file(filp);
4dd6f977
JK		 3218 return err;
3219}
3220
34178b1b
CY		 3221static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
3222{
3223 struct f2fs_move_range range;
3224
3225 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
3226 sizeof(range)))
3227 return -EFAULT;
3228 return __f2fs_ioc_move_range(filp, &range);
3229}
3230
e066b83c
JK		 3231static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
3232{
3233 struct inode *inode = file_inode(filp);
3234 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3235 struct sit_info *sm = SIT_I(sbi);
3236 unsigned int start_segno = 0, end_segno = 0;
3237 unsigned int dev_start_segno = 0, dev_end_segno = 0;
3238 struct f2fs_flush_device range;
d147ea4a
JK		 3239 struct f2fs_gc_control gc_control = {
3240 .init_gc_type = FG_GC,
3241 .should_migrate_blocks = true,
c81d5bae
JK		 3242 .err_gc_skipped = true,
3243 .nr_free_secs = 0 };
e066b83c
JK		 3244 int ret;
3245
3246 if (!capable(CAP_SYS_ADMIN))
3247 return -EPERM;
3248
3249 if (f2fs_readonly(sbi->sb))
3250 return -EROFS;
3251
4354994f
DR		 3252 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
3253 return -EINVAL;
3254
e066b83c
JK		 3255 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
3256 sizeof(range)))
3257 return -EFAULT;
3258
0916878d 3259 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2c70c5e3 3260 __is_large_section(sbi)) {
a60108f7
JK		 3261 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1",
3262 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi));
e066b83c
JK		 3263 return -EINVAL;
3264 }
3265
3266 ret = mnt_want_write_file(filp);
3267 if (ret)
3268 return ret;
3269
3270 if (range.dev_num != 0)
3271 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
3272 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
3273
3274 start_segno = sm->last_victim[FLUSH_DEVICE];
3275 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
3276 start_segno = dev_start_segno;
3277 end_segno = min(start_segno + range.segments, dev_end_segno);
3278
3279 while (start_segno < end_segno) {
e4544b63 3280 if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
e066b83c
JK		 3281 ret = -EBUSY;
3282 goto out;
3283 }
3284 sm->last_victim[GC_CB] = end_segno + 1;
3285 sm->last_victim[GC_GREEDY] = end_segno + 1;
3286 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
d147ea4a
JK		 3287
3288 gc_control.victim_segno = start_segno;
9bf1dcbd 3289 stat_inc_gc_call_count(sbi, FOREGROUND);
d147ea4a 3290 ret = f2fs_gc(sbi, &gc_control);
e066b83c
JK		 3291 if (ret == -EAGAIN)
3292 ret = 0;
3293 else if (ret < 0)
3294 break;
3295 start_segno++;
3296 }
3297out:
3298 mnt_drop_write_file(filp);
3299 return ret;
3300}
3301
e65ef207
JK		 3302static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
3303{
3304 struct inode *inode = file_inode(filp);
3305 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
3306
3307 /* Must validate to set it with SQLite behavior in Android. */
3308 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
3309
3310 return put_user(sb_feature, (u32 __user *)arg);
3311}
e066b83c 3312
2c1d0305 3313#ifdef CONFIG_QUOTA
78130819
CY		 3314int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3315{
3316 struct dquot *transfer_to[MAXQUOTAS] = {};
3317 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3318 struct super_block *sb = sbi->sb;
8051692f 3319 int err;
78130819
CY		 3320
3321 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
8051692f
YL		 3322 if (IS_ERR(transfer_to[PRJQUOTA]))
3323 return PTR_ERR(transfer_to[PRJQUOTA]);
3324
3325 err = __dquot_transfer(inode, transfer_to);
3326 if (err)
3327 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3328 dqput(transfer_to[PRJQUOTA]);
78130819
CY		 3329 return err;
3330}
3331
9b1bb01c 3332static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2c1d0305 3333{
2c1d0305
CY		 3334 struct f2fs_inode_info *fi = F2FS_I(inode);
3335 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d13732cc 3336 struct f2fs_inode *ri = NULL;
2c1d0305
CY		 3337 kprojid_t kprojid;
3338 int err;
3339
7beb01f7 3340 if (!f2fs_sb_has_project_quota(sbi)) {
2c1d0305
CY		 3341 if (projid != F2FS_DEF_PROJID)
3342 return -EOPNOTSUPP;
3343 else
3344 return 0;
3345 }
3346
3347 if (!f2fs_has_extra_attr(inode))
3348 return -EOPNOTSUPP;
3349
3350 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
3351
054cb289 3352 if (projid_eq(kprojid, fi->i_projid))
2c1d0305
CY		 3353 return 0;
3354
2c1d0305 3355 err = -EPERM;
2c1d0305
CY		 3356 /* Is it quota file? Do not allow user to mess with it */
3357 if (IS_NOQUOTA(inode))
c8e92757 3358 return err;
2c1d0305 3359
d13732cc
JY		 3360 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
3361 return -EOVERFLOW;
2c1d0305 3362
10a26878 3363 err = f2fs_dquot_initialize(inode);
c22aecd7 3364 if (err)
c8e92757 3365 return err;
2c1d0305 3366
78130819
CY		 3367 f2fs_lock_op(sbi);
3368 err = f2fs_transfer_project_quota(inode, kprojid);
3369 if (err)
3370 goto out_unlock;
2c1d0305 3371
054cb289 3372 fi->i_projid = kprojid;
c62ebd35 3373 inode_set_ctime_current(inode);
2c1d0305 3374 f2fs_mark_inode_dirty_sync(inode, true);
78130819
CY		 3375out_unlock:
3376 f2fs_unlock_op(sbi);
2c1d0305
CY		 3377 return err;
3378}
3379#else
78130819
CY		 3380int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3381{
3382 return 0;
3383}
3384
9b1bb01c 3385static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2c1d0305
CY		 3386{
3387 if (projid != F2FS_DEF_PROJID)
3388 return -EOPNOTSUPP;
3389 return 0;
3390}
3391#endif
3392
9b1bb01c 3393int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
2c1d0305 3394{
9b1bb01c 3395 struct inode *inode = d_inode(dentry);
2c1d0305 3396 struct f2fs_inode_info *fi = F2FS_I(inode);
9b1bb01c 3397 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
2c1d0305 3398
9b1bb01c
MS		 3399 if (IS_ENCRYPTED(inode))
3400 fsflags |= FS_ENCRYPT_FL;
3401 if (IS_VERITY(inode))
3402 fsflags |= FS_VERITY_FL;
3403 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3404 fsflags |= FS_INLINE_DATA_FL;
3405 if (is_inode_flag_set(inode, FI_PIN_FILE))
3406 fsflags |= FS_NOCOW_FL;
3407
3408 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
2c1d0305 3409
7beb01f7 3410 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
6fc93c4e 3411 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
2c1d0305 3412
c8e92757
WS		 3413 return 0;
3414}
3415
8782a9ae 3416int f2fs_fileattr_set(struct mnt_idmap *idmap,
9b1bb01c 3417 struct dentry *dentry, struct fileattr *fa)
2c1d0305 3418{
9b1bb01c
MS		 3419 struct inode *inode = d_inode(dentry);
3420 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
36098557 3421 u32 iflags;
2c1d0305
CY		 3422 int err;
3423
9b1bb01c
MS		 3424 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3425 return -EIO;
3426 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3427 return -ENOSPC;
3428 if (fsflags & ~F2FS_GETTABLE_FS_FL)
2c1d0305 3429 return -EOPNOTSUPP;
9b1bb01c
MS		 3430 fsflags &= F2FS_SETTABLE_FS_FL;
3431 if (!fa->flags_valid)
3432 mask &= FS_COMMON_FL;
2c1d0305 3433
9b1bb01c 3434 iflags = f2fs_fsflags_to_iflags(fsflags);
36098557 3435 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
2c1d0305
CY		 3436 return -EOPNOTSUPP;
3437
9b1bb01c
MS		 3438 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3439 if (!err)
3440 err = f2fs_ioc_setproject(inode, fa->fsx_projid);
2c1d0305 3441
c8e92757 3442 return err;
2c1d0305 3443}
e066b83c 3444
1ad71a27
JK		 3445int f2fs_pin_file_control(struct inode *inode, bool inc)
3446{
3447 struct f2fs_inode_info *fi = F2FS_I(inode);
3448 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3449
128d333f
DJ		 3450 if (IS_DEVICE_ALIASING(inode))
3451 return -EINVAL;
3452
c521a6ab 3453 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) {
dcbb4c10 3454 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
968c4f72 3455 __func__, inode->i_ino, fi->i_gc_failures);
1ad71a27
JK		 3456 clear_inode_flag(inode, FI_PIN_FILE);
3457 return -EAGAIN;
3458 }
c521a6ab
CY		 3459
3460 /* Use i_gc_failures for normal file as a risk signal. */
3461 if (inc)
3462 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1);
3463
1ad71a27
JK		 3464 return 0;
3465}
3466
3467static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3468{
3469 struct inode *inode = file_inode(filp);
9703d69d 3470 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1ad71a27
JK		 3471 __u32 pin;
3472 int ret = 0;
3473
1ad71a27
JK		 3474 if (get_user(pin, (__u32 __user *)arg))
3475 return -EFAULT;
3476
3477 if (!S_ISREG(inode->i_mode))
3478 return -EINVAL;
3479
9703d69d 3480 if (f2fs_readonly(sbi->sb))
1ad71a27
JK		 3481 return -EROFS;
3482
128d333f
DJ		 3483 if (!pin && IS_DEVICE_ALIASING(inode))
3484 return -EOPNOTSUPP;
3485
1ad71a27
JK		 3486 ret = mnt_want_write_file(filp);
3487 if (ret)
3488 return ret;
3489
3490 inode_lock(inode);
3491
bfe5c026
CY		 3492 if (f2fs_is_atomic_file(inode)) {
3493 ret = -EINVAL;
3494 goto out;
3495 }
3496
1ad71a27
JK		 3497 if (!pin) {
3498 clear_inode_flag(inode, FI_PIN_FILE);
30933364 3499 f2fs_i_gc_failures_write(inode, 0);
1ad71a27 3500 goto done;
9703d69d
DJ		 3501 } else if (f2fs_is_pinned_file(inode)) {
3502 goto done;
3503 }
3504
3fdd89b4 3505 if (F2FS_HAS_BLOCKS(inode)) {
9703d69d
DJ		 3506 ret = -EFBIG;
3507 goto out;
1ad71a27
JK		 3508 }
3509
9703d69d
DJ		 3510 /* Let's allow file pinning on zoned device. */
3511 if (!f2fs_sb_has_blkzoned(sbi) &&
3512 f2fs_should_update_outplace(inode, NULL)) {
19bdba52
JK		 3513 ret = -EINVAL;
3514 goto out;
3515 }
3516
1ad71a27
JK		 3517 if (f2fs_pin_file_control(inode, false)) {
3518 ret = -EAGAIN;
3519 goto out;
3520 }
4c8ff709 3521
1ad71a27
JK		 3522 ret = f2fs_convert_inline_inode(inode);
3523 if (ret)
3524 goto out;
3525
78134d03 3526 if (!f2fs_disable_compressed_file(inode)) {
4c8ff709
CY		 3527 ret = -EOPNOTSUPP;
3528 goto out;
3529 }
3530
1ad71a27 3531 set_inode_flag(inode, FI_PIN_FILE);
968c4f72 3532 ret = F2FS_I(inode)->i_gc_failures;
1ad71a27 3533done:
9703d69d 3534 f2fs_update_time(sbi, REQ_TIME);
1ad71a27
JK		 3535out:
3536 inode_unlock(inode);
3537 mnt_drop_write_file(filp);
3538 return ret;
3539}
3540
3541static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3542{
3543 struct inode *inode = file_inode(filp);
3544 __u32 pin = 0;
3545
3546 if (is_inode_flag_set(inode, FI_PIN_FILE))
968c4f72 3547 pin = F2FS_I(inode)->i_gc_failures;
1ad71a27
JK		 3548 return put_user(pin, (u32 __user *)arg);
3549}
3550
128d333f
DJ		 3551static int f2fs_ioc_get_dev_alias_file(struct file *filp, unsigned long arg)
3552{
3553 return put_user(IS_DEVICE_ALIASING(file_inode(filp)) ? 1 : 0,
3554 (u32 __user *)arg);
3555}
3556
5f95c181
JK		 3557static int f2fs_ioc_io_prio(struct file *filp, unsigned long arg)
3558{
3559 struct inode *inode = file_inode(filp);
3560 __u32 level;
3561
3562 if (get_user(level, (__u32 __user *)arg))
3563 return -EFAULT;
3564
3565 if (!S_ISREG(inode->i_mode) || level >= F2FS_IOPRIO_MAX)
3566 return -EINVAL;
3567
3568 inode_lock(inode);
3569 F2FS_I(inode)->ioprio_hint = level;
3570 inode_unlock(inode);
3571 return 0;
3572}
3573
c4020b2d
CY		 3574int f2fs_precache_extents(struct inode *inode)
3575{
3576 struct f2fs_inode_info *fi = F2FS_I(inode);
3577 struct f2fs_map_blocks map;
3578 pgoff_t m_next_extent;
3579 loff_t end;
3580 int err;
3581
3582 if (is_inode_flag_set(inode, FI_NO_EXTENT))
3583 return -EOPNOTSUPP;
3584
3585 map.m_lblk = 0;
8b07c1fb 3586 map.m_pblk = 0;
c4020b2d
CY		 3587 map.m_next_pgofs = NULL;
3588 map.m_next_extent = &m_next_extent;
3589 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 3590 map.m_may_create = false;
6f560c0f 3591 end = F2FS_BLK_ALIGN(i_size_read(inode));
c4020b2d
CY		 3592
3593 while (map.m_lblk < end) {
3594 map.m_len = end - map.m_lblk;
3595
e4544b63 3596 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
cd8fc522 3597 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE);
e4544b63 3598 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
4ed33e69 3599 if (err || !map.m_len)
c4020b2d
CY		 3600 return err;
3601
3602 map.m_lblk = m_next_extent;
3603 }
3604
4f55dc2a 3605 return 0;
c4020b2d
CY		 3606}
3607
ddf1eca4 3608static int f2fs_ioc_precache_extents(struct file *filp)
c4020b2d
CY		 3609{
3610 return f2fs_precache_extents(file_inode(filp));
3611}
3612
04f0b2ea
QS		 3613static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3614{
3615 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3616 __u64 block_count;
04f0b2ea
QS		 3617
3618 if (!capable(CAP_SYS_ADMIN))
3619 return -EPERM;
3620
3621 if (f2fs_readonly(sbi->sb))
3622 return -EROFS;
3623
3624 if (copy_from_user(&block_count, (void __user *)arg,
3625 sizeof(block_count)))
3626 return -EFAULT;
3627
d8189834 3628 return f2fs_resize_fs(filp, block_count);
04f0b2ea
QS		 3629}
3630
95ae251f
EB		 3631static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3632{
3633 struct inode *inode = file_inode(filp);
3634
3635 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3636
3637 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3638 f2fs_warn(F2FS_I_SB(inode),
833dcd35 3639 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
95ae251f
EB		 3640 inode->i_ino);
3641 return -EOPNOTSUPP;
3642 }
3643
3644 return fsverity_ioctl_enable(filp, (const void __user *)arg);
3645}
3646
3647static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3648{
3649 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3650 return -EOPNOTSUPP;
3651
3652 return fsverity_ioctl_measure(filp, (void __user *)arg);
3653}
3654
e17fe657
EB		 3655static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3656{
3657 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3658 return -EOPNOTSUPP;
3659
3660 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3661}
3662
3357af8f 3663static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
4507847c
CY		 3664{
3665 struct inode *inode = file_inode(filp);
3666 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3667 char *vbuf;
3668 int count;
3669 int err = 0;
3670
3671 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3672 if (!vbuf)
3673 return -ENOMEM;
3674
e4544b63 3675 f2fs_down_read(&sbi->sb_lock);
4507847c
CY		 3676 count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3677 ARRAY_SIZE(sbi->raw_super->volume_name),
3678 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
e4544b63 3679 f2fs_up_read(&sbi->sb_lock);
4507847c
CY		 3680
3681 if (copy_to_user((char __user *)arg, vbuf,
3682 min(FSLABEL_MAX, count)))
3683 err = -EFAULT;
3684
c8eb7024 3685 kfree(vbuf);
4507847c
CY		 3686 return err;
3687}
3688
3357af8f 3689static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
4507847c
CY		 3690{
3691 struct inode *inode = file_inode(filp);
3692 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3693 char *vbuf;
3694 int err = 0;
3695
3696 if (!capable(CAP_SYS_ADMIN))
3697 return -EPERM;
3698
3699 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3700 if (IS_ERR(vbuf))
3701 return PTR_ERR(vbuf);
3702
3703 err = mnt_want_write_file(filp);
3704 if (err)
3705 goto out;
3706
e4544b63 3707 f2fs_down_write(&sbi->sb_lock);
4507847c
CY		 3708
3709 memset(sbi->raw_super->volume_name, 0,
3710 sizeof(sbi->raw_super->volume_name));
3711 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3712 sbi->raw_super->volume_name,
3713 ARRAY_SIZE(sbi->raw_super->volume_name));
3714
3715 err = f2fs_commit_super(sbi, false);
3716
e4544b63 3717 f2fs_up_write(&sbi->sb_lock);
4507847c
CY		 3718
3719 mnt_drop_write_file(filp);
3720out:
3721 kfree(vbuf);
3722 return err;
3723}
3724
ac1ee161 3725static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks)
439dfb10 3726{
439dfb10
CY		 3727 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3728 return -EOPNOTSUPP;
3729
3730 if (!f2fs_compressed_file(inode))
3731 return -EINVAL;
3732
ac1ee161
SY		 3733 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3734
3735 return 0;
3736}
3737
3738static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg)
3739{
3740 struct inode *inode = file_inode(filp);
3741 __u64 blocks;
3742 int ret;
3743
3744 ret = f2fs_get_compress_blocks(inode, &blocks);
3745 if (ret < 0)
3746 return ret;
3747
439dfb10
CY		 3748 return put_user(blocks, (u64 __user *)arg);
3749}
3750
ef8d563f
CY		 3751static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3752{
3753 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3754 unsigned int released_blocks = 0;
3755 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3756 block_t blkaddr;
3757 int i;
3758
3759 for (i = 0; i < count; i++) {
6f7ec661 3760 blkaddr = data_blkaddr(dn->inode, dn->node_folio,
ef8d563f
CY		 3761 dn->ofs_in_node + i);
3762
3763 if (!__is_valid_data_blkaddr(blkaddr))
3764 continue;
3765 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
31f85ccc 3766 DATA_GENERIC_ENHANCE)))
ef8d563f
CY		 3767 return -EFSCORRUPTED;
3768 }
3769
3770 while (count) {
3771 int compr_blocks = 0;
3772
3773 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3774 blkaddr = f2fs_data_blkaddr(dn);
3775
3776 if (i == 0) {
3777 if (blkaddr == COMPRESS_ADDR)
3778 continue;
3779 dn->ofs_in_node += cluster_size;
3780 goto next;
3781 }
3782
3783 if (__is_valid_data_blkaddr(blkaddr))
3784 compr_blocks++;
3785
3786 if (blkaddr != NEW_ADDR)
3787 continue;
3788
59d0d4c3 3789 f2fs_set_data_blkaddr(dn, NULL_ADDR);
ef8d563f
CY		 3790 }
3791
3792 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3793 dec_valid_block_count(sbi, dn->inode,
3794 cluster_size - compr_blocks);
3795
3796 released_blocks += cluster_size - compr_blocks;
3797next:
3798 count -= cluster_size;
3799 }
3800
3801 return released_blocks;
3802}
3803
3804static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3805{
3806 struct inode *inode = file_inode(filp);
7309871c 3807 struct f2fs_inode_info *fi = F2FS_I(inode);
ef8d563f
CY		 3808 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3809 pgoff_t page_idx = 0, last_idx;
3810 unsigned int released_blocks = 0;
3811 int ret;
3812 int writecount;
3813
c3355ea9 3814 if (!f2fs_sb_has_compression(sbi))
ef8d563f
CY		 3815 return -EOPNOTSUPP;
3816
ef8d563f
CY		 3817 if (f2fs_readonly(sbi->sb))
3818 return -EROFS;
3819
3820 ret = mnt_want_write_file(filp);
3821 if (ret)
3822 return ret;
3823
c3355ea9 3824 f2fs_balance_fs(sbi, true);
ef8d563f
CY		 3825
3826 inode_lock(inode);
3827
3828 writecount = atomic_read(&inode->i_writecount);
8c8cf26a
DJ		 3829 if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3830 (!(filp->f_mode & FMODE_WRITE) && writecount)) {
ef8d563f
CY		 3831 ret = -EBUSY;
3832 goto out;
3833 }
3834
7c5dffb3
CY		 3835 if (!f2fs_compressed_file(inode) ||
3836 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
ef8d563f
CY		 3837 ret = -EINVAL;
3838 goto out;
3839 }
3840
3841 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3842 if (ret)
3843 goto out;
3844
7309871c 3845 if (!atomic_read(&fi->i_compr_blocks)) {
87a91a15
SY		 3846 ret = -EPERM;
3847 goto out;
3848 }
3849
c6140415 3850 set_inode_flag(inode, FI_COMPRESS_RELEASED);
c62ebd35 3851 inode_set_ctime_current(inode);
ef8d563f
CY		 3852 f2fs_mark_inode_dirty_sync(inode, true);
3853
7309871c 3854 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
edc6d01b 3855 filemap_invalidate_lock(inode->i_mapping);
ef8d563f
CY		 3856
3857 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3858
3859 while (page_idx < last_idx) {
3860 struct dnode_of_data dn;
3861 pgoff_t end_offset, count;
3862
0a4ed2d9
CY		 3863 f2fs_lock_op(sbi);
3864
ef8d563f
CY		 3865 set_new_dnode(&dn, inode, NULL, NULL, 0);
3866 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3867 if (ret) {
0a4ed2d9 3868 f2fs_unlock_op(sbi);
ef8d563f
CY		 3869 if (ret == -ENOENT) {
3870 page_idx = f2fs_get_next_page_offset(&dn,
3871 page_idx);
3872 ret = 0;
3873 continue;
3874 }
3875 break;
3876 }
3877
6f7ec661 3878 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
ef8d563f 3879 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
7309871c 3880 count = round_up(count, fi->i_cluster_size);
ef8d563f
CY		 3881
3882 ret = release_compress_blocks(&dn, count);
3883
3884 f2fs_put_dnode(&dn);
3885
0a4ed2d9
CY		 3886 f2fs_unlock_op(sbi);
3887
ef8d563f
CY		 3888 if (ret < 0)
3889 break;
3890
3891 page_idx += count;
3892 released_blocks += ret;
3893 }
3894
edc6d01b 3895 filemap_invalidate_unlock(inode->i_mapping);
7309871c 3896 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
ef8d563f 3897out:
fa18d87c
ZN		 3898 if (released_blocks)
3899 f2fs_update_time(sbi, REQ_TIME);
ef8d563f
CY		 3900 inode_unlock(inode);
3901
3902 mnt_drop_write_file(filp);
3903
3904 if (ret >= 0) {
3905 ret = put_user(released_blocks, (u64 __user *)arg);
c2759eba 3906 } else if (released_blocks &&
7309871c 3907 atomic_read(&fi->i_compr_blocks)) {
ef8d563f
CY		 3908 set_sbi_flag(sbi, SBI_NEED_FSCK);
3909 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
c2759eba 3910 "iblocks=%llu, released=%u, compr_blocks=%u, "
ef8d563f
CY		 3911 "run fsck to fix.",
3912 __func__, inode->i_ino, inode->i_blocks,
3913 released_blocks,
7309871c 3914 atomic_read(&fi->i_compr_blocks));
ef8d563f
CY		 3915 }
3916
3917 return ret;
3918}
3919
2f6d721e
XW		 3920static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
3921 unsigned int *reserved_blocks)
c75488fb
CY		 3922{
3923 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
c75488fb
CY		 3924 int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3925 block_t blkaddr;
3926 int i;
3927
3928 for (i = 0; i < count; i++) {
6f7ec661 3929 blkaddr = data_blkaddr(dn->inode, dn->node_folio,
c75488fb
CY		 3930 dn->ofs_in_node + i);
3931
3932 if (!__is_valid_data_blkaddr(blkaddr))
3933 continue;
3934 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
31f85ccc 3935 DATA_GENERIC_ENHANCE)))
c75488fb
CY		 3936 return -EFSCORRUPTED;
3937 }
3938
3939 while (count) {
3940 int compr_blocks = 0;
186e7d71
CY		 3941 blkcnt_t reserved = 0;
3942 blkcnt_t to_reserved;
c75488fb
CY		 3943 int ret;
3944
54607494 3945 for (i = 0; i < cluster_size; i++) {
6f7ec661 3946 blkaddr = data_blkaddr(dn->inode, dn->node_folio,
54607494 3947 dn->ofs_in_node + i);
c75488fb
CY		 3948
3949 if (i == 0) {
54607494
CY		 3950 if (blkaddr != COMPRESS_ADDR) {
3951 dn->ofs_in_node += cluster_size;
3952 goto next;
3953 }
3954 continue;
c75488fb
CY		 3955 }
3956
eb8fbaa5
SY		 3957 /*
3958 * compressed cluster was not released due to it
3959 * fails in release_compress_blocks(), so NEW_ADDR
3960 * is a possible case.
3961 */
186e7d71
CY		 3962 if (blkaddr == NEW_ADDR) {
3963 reserved++;
3964 continue;
3965 }
3966 if (__is_valid_data_blkaddr(blkaddr)) {
c75488fb
CY		 3967 compr_blocks++;
3968 continue;
3969 }
c75488fb
CY		 3970 }
3971
186e7d71 3972 to_reserved = cluster_size - compr_blocks - reserved;
eb8fbaa5
SY		 3973
3974 /* for the case all blocks in cluster were reserved */
26413ce1 3975 if (reserved && to_reserved == 1) {
186e7d71 3976 dn->ofs_in_node += cluster_size;
eb8fbaa5 3977 goto next;
186e7d71 3978 }
eb8fbaa5 3979
186e7d71
CY		 3980 ret = inc_valid_block_count(sbi, dn->inode,
3981 &to_reserved, false);
54607494 3982 if (unlikely(ret))
c75488fb
CY		 3983 return ret;
3984
54607494
CY		 3985 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3986 if (f2fs_data_blkaddr(dn) == NULL_ADDR)
3987 f2fs_set_data_blkaddr(dn, NEW_ADDR);
3988 }
c75488fb
CY		 3989
3990 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3991
186e7d71 3992 *reserved_blocks += to_reserved;
c75488fb
CY		 3993next:
3994 count -= cluster_size;
3995 }
3996
2f6d721e 3997 return 0;
c75488fb
CY		 3998}
3999
4000static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
4001{
4002 struct inode *inode = file_inode(filp);
7309871c 4003 struct f2fs_inode_info *fi = F2FS_I(inode);
c75488fb
CY		 4004 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4005 pgoff_t page_idx = 0, last_idx;
4006 unsigned int reserved_blocks = 0;
4007 int ret;
4008
c3355ea9 4009 if (!f2fs_sb_has_compression(sbi))
c75488fb
CY		 4010 return -EOPNOTSUPP;
4011
c75488fb
CY		 4012 if (f2fs_readonly(sbi->sb))
4013 return -EROFS;
4014
4015 ret = mnt_want_write_file(filp);
4016 if (ret)
4017 return ret;
4018
c3355ea9 4019 f2fs_balance_fs(sbi, true);
c75488fb
CY		 4020
4021 inode_lock(inode);
4022
7c5dffb3
CY		 4023 if (!f2fs_compressed_file(inode) ||
4024 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
c75488fb
CY		 4025 ret = -EINVAL;
4026 goto unlock_inode;
4027 }
4028
7309871c 4029 if (atomic_read(&fi->i_compr_blocks))
b7d797d2
XW		 4030 goto unlock_inode;
4031
7309871c 4032 f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
edc6d01b 4033 filemap_invalidate_lock(inode->i_mapping);
c75488fb
CY		 4034
4035 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4036
4037 while (page_idx < last_idx) {
4038 struct dnode_of_data dn;
4039 pgoff_t end_offset, count;
4040
0a4ed2d9
CY		 4041 f2fs_lock_op(sbi);
4042
c75488fb
CY		 4043 set_new_dnode(&dn, inode, NULL, NULL, 0);
4044 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
4045 if (ret) {
0a4ed2d9 4046 f2fs_unlock_op(sbi);
c75488fb
CY		 4047 if (ret == -ENOENT) {
4048 page_idx = f2fs_get_next_page_offset(&dn,
4049 page_idx);
4050 ret = 0;
4051 continue;
4052 }
4053 break;
4054 }
4055
6f7ec661 4056 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
c75488fb 4057 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
7309871c 4058 count = round_up(count, fi->i_cluster_size);
c75488fb 4059
2f6d721e 4060 ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
c75488fb
CY		 4061
4062 f2fs_put_dnode(&dn);
4063
0a4ed2d9
CY		 4064 f2fs_unlock_op(sbi);
4065
c75488fb
CY		 4066 if (ret < 0)
4067 break;
4068
4069 page_idx += count;
c75488fb
CY		 4070 }
4071
edc6d01b 4072 filemap_invalidate_unlock(inode->i_mapping);
7309871c 4073 f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
c75488fb 4074
2f6d721e 4075 if (!ret) {
c6140415 4076 clear_inode_flag(inode, FI_COMPRESS_RELEASED);
c62ebd35 4077 inode_set_ctime_current(inode);
c75488fb
CY		 4078 f2fs_mark_inode_dirty_sync(inode, true);
4079 }
4080unlock_inode:
fa18d87c
ZN		 4081 if (reserved_blocks)
4082 f2fs_update_time(sbi, REQ_TIME);
c75488fb 4083 inode_unlock(inode);
c75488fb
CY		 4084 mnt_drop_write_file(filp);
4085
2f6d721e 4086 if (!ret) {
c75488fb 4087 ret = put_user(reserved_blocks, (u64 __user *)arg);
c2759eba 4088 } else if (reserved_blocks &&
7309871c 4089 atomic_read(&fi->i_compr_blocks)) {
c75488fb 4090 set_sbi_flag(sbi, SBI_NEED_FSCK);
a3a0bc6c 4091 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx "
c2759eba 4092 "iblocks=%llu, reserved=%u, compr_blocks=%u, "
c75488fb
CY		 4093 "run fsck to fix.",
4094 __func__, inode->i_ino, inode->i_blocks,
4095 reserved_blocks,
7309871c 4096 atomic_read(&fi->i_compr_blocks));
c75488fb
CY		 4097 }
4098
4099 return ret;
4100}
4101
9af84648
DJ		 4102static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
4103 pgoff_t off, block_t block, block_t len, u32 flags)
4104{
9af84648
DJ		 4105 sector_t sector = SECTOR_FROM_BLOCK(block);
4106 sector_t nr_sects = SECTOR_FROM_BLOCK(len);
4107 int ret = 0;
4108
44abff2c
CH		 4109 if (flags & F2FS_TRIM_FILE_DISCARD) {
4110 if (bdev_max_secure_erase_sectors(bdev))
4111 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
4112 GFP_NOFS);
4113 else
4114 ret = blkdev_issue_discard(bdev, sector, nr_sects,
4115 GFP_NOFS);
4116 }
9af84648
DJ		 4117
4118 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
4119 if (IS_ENCRYPTED(inode))
4120 ret = fscrypt_zeroout_range(inode, off, block, len);
4121 else
4122 ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
4123 GFP_NOFS, 0);
4124 }
4125
4126 return ret;
4127}
4128
4129static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
4130{
4131 struct inode *inode = file_inode(filp);
4132 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4133 struct address_space *mapping = inode->i_mapping;
4134 struct block_device *prev_bdev = NULL;
4135 struct f2fs_sectrim_range range;
4136 pgoff_t index, pg_end, prev_index = 0;
4137 block_t prev_block = 0, len = 0;
4138 loff_t end_addr;
4139 bool to_end = false;
4140 int ret = 0;
4141
4142 if (!(filp->f_mode & FMODE_WRITE))
4143 return -EBADF;
4144
4145 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
4146 sizeof(range)))
4147 return -EFAULT;
4148
4149 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
4150 !S_ISREG(inode->i_mode))
4151 return -EINVAL;
4152
4153 if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
4154 !f2fs_hw_support_discard(sbi)) ||
4155 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
4156 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
4157 return -EOPNOTSUPP;
4158
1efb7c8f
ZN		 4159 ret = mnt_want_write_file(filp);
4160 if (ret)
4161 return ret;
9af84648
DJ		 4162 inode_lock(inode);
4163
4164 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
4165 range.start >= inode->i_size) {
4166 ret = -EINVAL;
4167 goto err;
4168 }
4169
4170 if (range.len == 0)
4171 goto err;
4172
4173 if (inode->i_size - range.start > range.len) {
4174 end_addr = range.start + range.len;
4175 } else {
4176 end_addr = range.len == (u64)-1 ?
4177 sbi->sb->s_maxbytes : inode->i_size;
4178 to_end = true;
4179 }
4180
4181 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
4182 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
4183 ret = -EINVAL;
4184 goto err;
4185 }
4186
4187 index = F2FS_BYTES_TO_BLK(range.start);
4188 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
4189
4190 ret = f2fs_convert_inline_inode(inode);
4191 if (ret)
4192 goto err;
4193
e4544b63 4194 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
edc6d01b 4195 filemap_invalidate_lock(mapping);
9af84648
DJ		 4196
4197 ret = filemap_write_and_wait_range(mapping, range.start,
4198 to_end ? LLONG_MAX : end_addr - 1);
4199 if (ret)
4200 goto out;
4201
4202 truncate_inode_pages_range(mapping, range.start,
4203 to_end ? -1 : end_addr - 1);
4204
4205 while (index < pg_end) {
4206 struct dnode_of_data dn;
4207 pgoff_t end_offset, count;
4208 int i;
4209
4210 set_new_dnode(&dn, inode, NULL, NULL, 0);
4211 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
4212 if (ret) {
4213 if (ret == -ENOENT) {
4214 index = f2fs_get_next_page_offset(&dn, index);
4215 continue;
4216 }
4217 goto out;
4218 }
4219
6f7ec661 4220 end_offset = ADDRS_PER_PAGE(&dn.node_folio->page, inode);
9af84648
DJ		 4221 count = min(end_offset - dn.ofs_in_node, pg_end - index);
4222 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
4223 struct block_device *cur_bdev;
4224 block_t blkaddr = f2fs_data_blkaddr(&dn);
4225
4226 if (!__is_valid_data_blkaddr(blkaddr))
4227 continue;
4228
4229 if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
4230 DATA_GENERIC_ENHANCE)) {
4231 ret = -EFSCORRUPTED;
4232 f2fs_put_dnode(&dn);
4233 goto out;
4234 }
4235
4236 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
4237 if (f2fs_is_multi_device(sbi)) {
4238 int di = f2fs_target_device_index(sbi, blkaddr);
4239
4240 blkaddr -= FDEV(di).start_blk;
4241 }
4242
4243 if (len) {
4244 if (prev_bdev == cur_bdev &&
4245 index == prev_index + len &&
4246 blkaddr == prev_block + len) {
4247 len++;
4248 } else {
4249 ret = f2fs_secure_erase(prev_bdev,
4250 inode, prev_index, prev_block,
4251 len, range.flags);
4252 if (ret) {
4253 f2fs_put_dnode(&dn);
4254 goto out;
4255 }
4256
4257 len = 0;
4258 }
4259 }
4260
4261 if (!len) {
4262 prev_bdev = cur_bdev;
4263 prev_index = index;
4264 prev_block = blkaddr;
4265 len = 1;
4266 }
4267 }
4268
4269 f2fs_put_dnode(&dn);
4270
4271 if (fatal_signal_pending(current)) {
4272 ret = -EINTR;
4273 goto out;
4274 }
4275 cond_resched();
4276 }
4277
4278 if (len)
4279 ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
4280 prev_block, len, range.flags);
fa18d87c 4281 f2fs_update_time(sbi, REQ_TIME);
9af84648 4282out:
edc6d01b 4283 filemap_invalidate_unlock(mapping);
e4544b63 4284 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
9af84648
DJ		 4285err:
4286 inode_unlock(inode);
1efb7c8f 4287 mnt_drop_write_file(filp);
9af84648
DJ		 4288
4289 return ret;
4290}
4291
9e2a5f8c 4292static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
52656e6c 4293{
9e2a5f8c
DJ		 4294 struct inode *inode = file_inode(filp);
4295 struct f2fs_comp_option option;
4296
4297 if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
4298 return -EOPNOTSUPP;
4299
4300 inode_lock_shared(inode);
4301
4302 if (!f2fs_compressed_file(inode)) {
4303 inode_unlock_shared(inode);
4304 return -ENODATA;
4305 }
4306
4307 option.algorithm = F2FS_I(inode)->i_compress_algorithm;
4308 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
4309
4310 inode_unlock_shared(inode);
4311
4312 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
4313 sizeof(option)))
4314 return -EFAULT;
4315
4316 return 0;
4317}
4318
e1e8debe
DJ		 4319static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
4320{
4321 struct inode *inode = file_inode(filp);
7309871c 4322 struct f2fs_inode_info *fi = F2FS_I(inode);
e1e8debe
DJ		 4323 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4324 struct f2fs_comp_option option;
4325 int ret = 0;
1f227a3e 4326
e1e8debe
DJ		 4327 if (!f2fs_sb_has_compression(sbi))
4328 return -EOPNOTSUPP;
4329
4330 if (!(filp->f_mode & FMODE_WRITE))
4331 return -EBADF;
4332
4333 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
4334 sizeof(option)))
4335 return -EFAULT;
4336
ea59b12a
CY		 4337 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
4338 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
4339 option.algorithm >= COMPRESS_MAX)
e1e8debe
DJ		 4340 return -EINVAL;
4341
1efb7c8f
ZN		 4342 ret = mnt_want_write_file(filp);
4343 if (ret)
4344 return ret;
e1e8debe
DJ		 4345 inode_lock(inode);
4346
b5ab3276 4347 f2fs_down_write(&F2FS_I(inode)->i_sem);
ea59b12a
CY		 4348 if (!f2fs_compressed_file(inode)) {
4349 ret = -EINVAL;
4350 goto out;
4351 }
4352
e1e8debe
DJ		 4353 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
4354 ret = -EBUSY;
4355 goto out;
4356 }
4357
e6261beb 4358 if (F2FS_HAS_BLOCKS(inode)) {
e1e8debe
DJ		 4359 ret = -EFBIG;
4360 goto out;
4361 }
4362
7309871c
CY		 4363 fi->i_compress_algorithm = option.algorithm;
4364 fi->i_log_cluster_size = option.log_cluster_size;
4365 fi->i_cluster_size = BIT(option.log_cluster_size);
f5f3bd90 4366 /* Set default level */
7309871c
CY		 4367 if (fi->i_compress_algorithm == COMPRESS_ZSTD)
4368 fi->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
f5f3bd90 4369 else
7309871c 4370 fi->i_compress_level = 0;
f5f3bd90
JK		 4371 /* Adjust mount option level */
4372 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm &&
4373 F2FS_OPTION(sbi).compress_level)
7309871c 4374 fi->i_compress_level = F2FS_OPTION(sbi).compress_level;
e1e8debe
DJ		 4375 f2fs_mark_inode_dirty_sync(inode, true);
4376
4377 if (!f2fs_is_compress_backend_ready(inode))
4378 f2fs_warn(sbi, "compression algorithm is successfully set, "
4379 "but current kernel doesn't support this algorithm.");
4380out:
7309871c 4381 f2fs_up_write(&fi->i_sem);
e1e8debe 4382 inode_unlock(inode);
1efb7c8f 4383 mnt_drop_write_file(filp);
e1e8debe
DJ		 4384
4385 return ret;
4386}
4387
5fdb322f
DJ		 4388static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
4389{
fcd9ae4f 4390 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
5fdb322f 4391 struct address_space *mapping = inode->i_mapping;
5951fee4 4392 struct folio *folio;
5fdb322f 4393 pgoff_t redirty_idx = page_idx;
5951fee4 4394 int page_len = 0, ret = 0;
5fdb322f
DJ		 4395
4396 page_cache_ra_unbounded(&ractl, len, 0);
4397
5951fee4
MWO		 4398 do {
4399 folio = read_cache_folio(mapping, page_idx, NULL, NULL);
4400 if (IS_ERR(folio)) {
4401 ret = PTR_ERR(folio);
5fdb322f
DJ		 4402 break;
4403 }
5951fee4
MWO		 4404 page_len += folio_nr_pages(folio) - (page_idx - folio->index);
4405 page_idx = folio_next_index(folio);
4406 } while (page_len < len);
5fdb322f 4407
5951fee4
MWO		 4408 do {
4409 folio = filemap_lock_folio(mapping, redirty_idx);
a4a0e16d 4410
5951fee4
MWO		 4411 /* It will never fail, when folio has pinned above */
4412 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(folio));
a4a0e16d 4413
5951fee4 4414 f2fs_folio_wait_writeback(folio, DATA, true, true);
a4d7f2b3 4415
5951fee4
MWO		 4416 folio_mark_dirty(folio);
4417 set_page_private_gcing(&folio->page);
4418 redirty_idx = folio_next_index(folio);
4419 folio_unlock(folio);
4420 folio_put_refs(folio, 2);
4421 } while (redirty_idx < page_idx);
5fdb322f
DJ		 4422
4423 return ret;
4424}
4425
ddf1eca4 4426static int f2fs_ioc_decompress_file(struct file *filp)
5fdb322f
DJ		 4427{
4428 struct inode *inode = file_inode(filp);
4429 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4430 struct f2fs_inode_info *fi = F2FS_I(inode);
f785cec2
YG		 4431 pgoff_t page_idx = 0, last_idx, cluster_idx;
4432 int ret;
5fdb322f
DJ		 4433
4434 if (!f2fs_sb_has_compression(sbi) ||
4435 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4436 return -EOPNOTSUPP;
4437
4438 if (!(filp->f_mode & FMODE_WRITE))
4439 return -EBADF;
4440
c3355ea9 4441 f2fs_balance_fs(sbi, true);
5fdb322f 4442
1efb7c8f
ZN		 4443 ret = mnt_want_write_file(filp);
4444 if (ret)
4445 return ret;
5fdb322f
DJ		 4446 inode_lock(inode);
4447
4448 if (!f2fs_is_compress_backend_ready(inode)) {
4449 ret = -EOPNOTSUPP;
4450 goto out;
4451 }
4452
bd9ae4ae
CY		 4453 if (!f2fs_compressed_file(inode) ||
4454 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
90be48bd
JK		 4455 ret = -EINVAL;
4456 goto out;
4457 }
4458
5fdb322f
DJ		 4459 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4460 if (ret)
4461 goto out;
4462
4463 if (!atomic_read(&fi->i_compr_blocks))
4464 goto out;
4465
4466 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
f785cec2 4467 last_idx >>= fi->i_log_cluster_size;
5fdb322f 4468
f785cec2
YG		 4469 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4470 page_idx = cluster_idx << fi->i_log_cluster_size;
4471
4472 if (!f2fs_is_compressed_cluster(inode, page_idx))
4473 continue;
5fdb322f 4474
f785cec2 4475 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
5fdb322f
DJ		 4476 if (ret < 0)
4477 break;
4478
a60108f7 4479 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
b822dc91
YL		 4480 ret = filemap_fdatawrite(inode->i_mapping);
4481 if (ret < 0)
4482 break;
4483 }
5fdb322f 4484
3a2c0e55
CY		 4485 cond_resched();
4486 if (fatal_signal_pending(current)) {
4487 ret = -EINTR;
4488 break;
4489 }
5fdb322f
DJ		 4490 }
4491
4492 if (!ret)
4493 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4494 LLONG_MAX);
4495
4496 if (ret)
833dcd35
JP		 4497 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
4498 __func__, ret);
fa18d87c 4499 f2fs_update_time(sbi, REQ_TIME);
5fdb322f
DJ		 4500out:
4501 inode_unlock(inode);
1efb7c8f 4502 mnt_drop_write_file(filp);
5fdb322f
DJ		 4503
4504 return ret;
4505}
4506
ddf1eca4 4507static int f2fs_ioc_compress_file(struct file *filp)
5fdb322f
DJ		 4508{
4509 struct inode *inode = file_inode(filp);
4510 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
f785cec2
YG		 4511 struct f2fs_inode_info *fi = F2FS_I(inode);
4512 pgoff_t page_idx = 0, last_idx, cluster_idx;
4513 int ret;
5fdb322f
DJ		 4514
4515 if (!f2fs_sb_has_compression(sbi) ||
4516 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4517 return -EOPNOTSUPP;
4518
4519 if (!(filp->f_mode & FMODE_WRITE))
4520 return -EBADF;
4521
c3355ea9 4522 f2fs_balance_fs(sbi, true);
5fdb322f 4523
1efb7c8f
ZN		 4524 ret = mnt_want_write_file(filp);
4525 if (ret)
4526 return ret;
5fdb322f
DJ		 4527 inode_lock(inode);
4528
4529 if (!f2fs_is_compress_backend_ready(inode)) {
4530 ret = -EOPNOTSUPP;
4531 goto out;
4532 }
4533
bd9ae4ae
CY		 4534 if (!f2fs_compressed_file(inode) ||
4535 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
90be48bd
JK		 4536 ret = -EINVAL;
4537 goto out;
4538 }
4539
5fdb322f
DJ		 4540 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4541 if (ret)
4542 goto out;
4543
4544 set_inode_flag(inode, FI_ENABLE_COMPRESS);
4545
4546 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
f785cec2 4547 last_idx >>= fi->i_log_cluster_size;
5fdb322f 4548
f785cec2
YG		 4549 for (cluster_idx = 0; cluster_idx < last_idx; cluster_idx++) {
4550 page_idx = cluster_idx << fi->i_log_cluster_size;
4551
4552 if (f2fs_is_sparse_cluster(inode, page_idx))
4553 continue;
5fdb322f 4554
f785cec2 4555 ret = redirty_blocks(inode, page_idx, fi->i_cluster_size);
5fdb322f
DJ		 4556 if (ret < 0)
4557 break;
4558
a60108f7 4559 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) {
b822dc91
YL		 4560 ret = filemap_fdatawrite(inode->i_mapping);
4561 if (ret < 0)
4562 break;
4563 }
5fdb322f 4564
3a2c0e55
CY		 4565 cond_resched();
4566 if (fatal_signal_pending(current)) {
4567 ret = -EINTR;
4568 break;
4569 }
5fdb322f
DJ		 4570 }
4571
4572 if (!ret)
4573 ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4574 LLONG_MAX);
4575
4576 clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4577
4578 if (ret)
833dcd35
JP		 4579 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4580 __func__, ret);
fa18d87c 4581 f2fs_update_time(sbi, REQ_TIME);
5fdb322f
DJ		 4582out:
4583 inode_unlock(inode);
1efb7c8f 4584 mnt_drop_write_file(filp);
5fdb322f
DJ		 4585
4586 return ret;
4587}
4588
34178b1b 4589static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
52656e6c
JK		 4590{
4591 switch (cmd) {
3357af8f 4592 case FS_IOC_GETVERSION:
d49f3e89 4593 return f2fs_ioc_getversion(filp, arg);
88b88a66 4594 case F2FS_IOC_START_ATOMIC_WRITE:
41e8f85a
DJ		 4595 return f2fs_ioc_start_atomic_write(filp, false);
4596 case F2FS_IOC_START_ATOMIC_REPLACE:
4597 return f2fs_ioc_start_atomic_write(filp, true);
88b88a66
JK		 4598 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4599 return f2fs_ioc_commit_atomic_write(filp);
23339e57
DJ		 4600 case F2FS_IOC_ABORT_ATOMIC_WRITE:
4601 return f2fs_ioc_abort_atomic_write(filp);
02a1335f 4602 case F2FS_IOC_START_VOLATILE_WRITE:
1e84371f 4603 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
7bc155fe 4604 return -EOPNOTSUPP;
1abff93d
JK		 4605 case F2FS_IOC_SHUTDOWN:
4606 return f2fs_ioc_shutdown(filp, arg);
52656e6c
JK		 4607 case FITRIM:
4608 return f2fs_ioc_fitrim(filp, arg);
3357af8f 4609 case FS_IOC_SET_ENCRYPTION_POLICY:
f424f664 4610 return f2fs_ioc_set_encryption_policy(filp, arg);
3357af8f 4611 case FS_IOC_GET_ENCRYPTION_POLICY:
f424f664 4612 return f2fs_ioc_get_encryption_policy(filp, arg);
3357af8f 4613 case FS_IOC_GET_ENCRYPTION_PWSALT:
f424f664 4614 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
8ce589c7
EB		 4615 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4616 return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4617 case FS_IOC_ADD_ENCRYPTION_KEY:
4618 return f2fs_ioc_add_encryption_key(filp, arg);
4619 case FS_IOC_REMOVE_ENCRYPTION_KEY:
4620 return f2fs_ioc_remove_encryption_key(filp, arg);
4621 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4622 return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4623 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4624 return f2fs_ioc_get_encryption_key_status(filp, arg);
ee446e1a
EB		 4625 case FS_IOC_GET_ENCRYPTION_NONCE:
4626 return f2fs_ioc_get_encryption_nonce(filp, arg);
c1c1b583
CY		 4627 case F2FS_IOC_GARBAGE_COLLECT:
4628 return f2fs_ioc_gc(filp, arg);
34dc77ad
JK		 4629 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4630 return f2fs_ioc_gc_range(filp, arg);
456b88e4 4631 case F2FS_IOC_WRITE_CHECKPOINT:
ddf1eca4 4632 return f2fs_ioc_write_checkpoint(filp);
d323d005
CY		 4633 case F2FS_IOC_DEFRAGMENT:
4634 return f2fs_ioc_defragment(filp, arg);
4dd6f977
JK		 4635 case F2FS_IOC_MOVE_RANGE:
4636 return f2fs_ioc_move_range(filp, arg);
e066b83c
JK		 4637 case F2FS_IOC_FLUSH_DEVICE:
4638 return f2fs_ioc_flush_device(filp, arg);
e65ef207
JK		 4639 case F2FS_IOC_GET_FEATURES:
4640 return f2fs_ioc_get_features(filp, arg);
1ad71a27
JK		 4641 case F2FS_IOC_GET_PIN_FILE:
4642 return f2fs_ioc_get_pin_file(filp, arg);
4643 case F2FS_IOC_SET_PIN_FILE:
4644 return f2fs_ioc_set_pin_file(filp, arg);
c4020b2d 4645 case F2FS_IOC_PRECACHE_EXTENTS:
ddf1eca4 4646 return f2fs_ioc_precache_extents(filp);
04f0b2ea
QS		 4647 case F2FS_IOC_RESIZE_FS:
4648 return f2fs_ioc_resize_fs(filp, arg);
95ae251f
EB		 4649 case FS_IOC_ENABLE_VERITY:
4650 return f2fs_ioc_enable_verity(filp, arg);
4651 case FS_IOC_MEASURE_VERITY:
4652 return f2fs_ioc_measure_verity(filp, arg);
e17fe657
EB		 4653 case FS_IOC_READ_VERITY_METADATA:
4654 return f2fs_ioc_read_verity_metadata(filp, arg);
3357af8f
EB		 4655 case FS_IOC_GETFSLABEL:
4656 return f2fs_ioc_getfslabel(filp, arg);
4657 case FS_IOC_SETFSLABEL:
4658 return f2fs_ioc_setfslabel(filp, arg);
439dfb10 4659 case F2FS_IOC_GET_COMPRESS_BLOCKS:
ac1ee161 4660 return f2fs_ioc_get_compress_blocks(filp, arg);
ef8d563f
CY		 4661 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4662 return f2fs_release_compress_blocks(filp, arg);
c75488fb
CY		 4663 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4664 return f2fs_reserve_compress_blocks(filp, arg);
9af84648
DJ		 4665 case F2FS_IOC_SEC_TRIM_FILE:
4666 return f2fs_sec_trim_file(filp, arg);
9e2a5f8c
DJ		 4667 case F2FS_IOC_GET_COMPRESS_OPTION:
4668 return f2fs_ioc_get_compress_option(filp, arg);
e1e8debe
DJ		 4669 case F2FS_IOC_SET_COMPRESS_OPTION:
4670 return f2fs_ioc_set_compress_option(filp, arg);
5fdb322f 4671 case F2FS_IOC_DECOMPRESS_FILE:
ddf1eca4 4672 return f2fs_ioc_decompress_file(filp);
5fdb322f 4673 case F2FS_IOC_COMPRESS_FILE:
ddf1eca4 4674 return f2fs_ioc_compress_file(filp);
128d333f
DJ		 4675 case F2FS_IOC_GET_DEV_ALIAS_FILE:
4676 return f2fs_ioc_get_dev_alias_file(filp, arg);
5f95c181
JK		 4677 case F2FS_IOC_IO_PRIO:
4678 return f2fs_ioc_io_prio(filp, arg);
fbfa2cc5
JK		 4679 default:
4680 return -ENOTTY;
4681 }
4682}
4683
34178b1b
CY		 4684long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4685{
4686 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4687 return -EIO;
4688 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4689 return -ENOSPC;
4690
4691 return __f2fs_ioctl(filp, cmd, arg);
4692}
4693
a1e09b03
EB		 4694/*
4695 * Return %true if the given read or write request should use direct I/O, or
4696 * %false if it should use buffered I/O.
4697 */
4698static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4699 struct iov_iter *iter)
4700{
4701 unsigned int align;
4702
4703 if (!(iocb->ki_flags & IOCB_DIRECT))
4704 return false;
4705
bd367329 4706 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter)))
a1e09b03
EB		 4707 return false;
4708
4709 /*
4710 * Direct I/O not aligned to the disk's logical_block_size will be
4711 * attempted, but will fail with -EINVAL.
4712 *
4713 * f2fs additionally requires that direct I/O be aligned to the
4714 * filesystem block size, which is often a stricter requirement.
4715 * However, f2fs traditionally falls back to buffered I/O on requests
4716 * that are logical_block_size-aligned but not fs-block aligned.
4717 *
4718 * The below logic implements this behavior.
4719 */
4720 align = iocb->ki_pos | iov_iter_alignment(iter);
4721 if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4722 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4723 return false;
4724
4725 return true;
4726}
4727
4728static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4729 unsigned int flags)
4730{
4731 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4732
4733 dec_page_count(sbi, F2FS_DIO_READ);
4734 if (error)
4735 return error;
34a23525 4736 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
a1e09b03
EB		 4737 return 0;
4738}
4739
4740static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4741 .end_io = f2fs_dio_read_end_io,
4742};
4743
4744static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4c8ff709
CY		 4745{
4746 struct file *file = iocb->ki_filp;
4747 struct inode *inode = file_inode(file);
a1e09b03
EB		 4748 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4749 struct f2fs_inode_info *fi = F2FS_I(inode);
4750 const loff_t pos = iocb->ki_pos;
4751 const size_t count = iov_iter_count(to);
4752 struct iomap_dio *dio;
4753 ssize_t ret;
4754
4755 if (count == 0)
4756 return 0; /* skip atime update */
4757
bd984c03 4758 trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
a1e09b03
EB		 4759
4760 if (iocb->ki_flags & IOCB_NOWAIT) {
e4544b63 4761 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
a1e09b03
EB		 4762 ret = -EAGAIN;
4763 goto out;
4764 }
4765 } else {
e4544b63 4766 f2fs_down_read(&fi->i_gc_rwsem[READ]);
a1e09b03
EB		 4767 }
4768
b2c160f4
CY		 4769 /* dio is not compatible w/ atomic file */
4770 if (f2fs_is_atomic_file(inode)) {
4771 f2fs_up_read(&fi->i_gc_rwsem[READ]);
4772 ret = -EOPNOTSUPP;
4773 goto out;
4774 }
4775
a1e09b03
EB		 4776 /*
4777 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4778 * the higher-level function iomap_dio_rw() in order to ensure that the
4779 * F2FS_DIO_READ counter will be decremented correctly in all cases.
4780 */
4781 inc_page_count(sbi, F2FS_DIO_READ);
4782 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
786f847f 4783 &f2fs_iomap_dio_read_ops, 0, NULL, 0);
a1e09b03
EB		 4784 if (IS_ERR_OR_NULL(dio)) {
4785 ret = PTR_ERR_OR_ZERO(dio);
4786 if (ret != -EIOCBQUEUED)
4787 dec_page_count(sbi, F2FS_DIO_READ);
4788 } else {
4789 ret = iomap_dio_complete(dio);
4790 }
4791
e4544b63 4792 f2fs_up_read(&fi->i_gc_rwsem[READ]);
a1e09b03
EB		 4793
4794 file_accessed(file);
4795out:
4796 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4797 return ret;
4798}
4799
ceb11d0e
DH		 4800static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count,
4801 int rw)
a28bca0f 4802{
ceb11d0e 4803 struct inode *inode = file_inode(file);
a28bca0f
CH		 4804 char *buf, *path;
4805
55847850 4806 buf = f2fs_getname(F2FS_I_SB(inode));
a28bca0f
CH		 4807 if (!buf)
4808 return;
ceb11d0e 4809 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX);
a28bca0f
CH		 4810 if (IS_ERR(path))
4811 goto free_buf;
4812 if (rw == WRITE)
ceb11d0e 4813 trace_f2fs_datawrite_start(inode, pos, count,
a28bca0f
CH		 4814 current->pid, path, current->comm);
4815 else
ceb11d0e 4816 trace_f2fs_dataread_start(inode, pos, count,
a28bca0f
CH		 4817 current->pid, path, current->comm);
4818free_buf:
55847850 4819 f2fs_putname(buf);
a28bca0f
CH		 4820}
4821
a1e09b03
EB		 4822static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4823{
4824 struct inode *inode = file_inode(iocb->ki_filp);
c277f141 4825 const loff_t pos = iocb->ki_pos;
a1e09b03 4826 ssize_t ret;
4c8ff709
CY		 4827
4828 if (!f2fs_is_compress_backend_ready(inode))
4829 return -EOPNOTSUPP;
4830
a28bca0f 4831 if (trace_f2fs_dataread_start_enabled())
ceb11d0e
DH		 4832 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
4833 iov_iter_count(to), READ);
c277f141 4834
0cac5118 4835 /* In LFS mode, if there is inflight dio, wait for its completion */
332fade7
JK		 4836 if (f2fs_lfs_mode(F2FS_I_SB(inode)) &&
4837 get_pages(F2FS_I_SB(inode), F2FS_DIO_WRITE))
0cac5118
CY		 4838 inode_dio_wait(inode);
4839
c277f141
JK		 4840 if (f2fs_should_use_dio(inode, iocb, to)) {
4841 ret = f2fs_dio_read_iter(iocb, to);
4842 } else {
4843 ret = filemap_read(iocb, to, 0);
4844 if (ret > 0)
34a23525
CY		 4845 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4846 APP_BUFFERED_READ_IO, ret);
c277f141
JK		 4847 }
4848 if (trace_f2fs_dataread_end_enabled())
4849 trace_f2fs_dataread_end(inode, pos, ret);
8b83ac81 4850 return ret;
4c8ff709
CY		 4851}
4852
ceb11d0e
DH		 4853static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos,
4854 struct pipe_inode_info *pipe,
4855 size_t len, unsigned int flags)
4856{
4857 struct inode *inode = file_inode(in);
4858 const loff_t pos = *ppos;
4859 ssize_t ret;
4860
4861 if (!f2fs_is_compress_backend_ready(inode))
4862 return -EOPNOTSUPP;
4863
4864 if (trace_f2fs_dataread_start_enabled())
4865 f2fs_trace_rw_file_path(in, pos, len, READ);
4866
4867 ret = filemap_splice_read(in, ppos, pipe, len, flags);
4868 if (ret > 0)
4869 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4870 APP_BUFFERED_READ_IO, ret);
4871
4872 if (trace_f2fs_dataread_end_enabled())
4873 trace_f2fs_dataread_end(inode, pos, ret);
4874 return ret;
4875}
4876
a1e09b03
EB		 4877static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4878{
4879 struct file *file = iocb->ki_filp;
4880 struct inode *inode = file_inode(file);
4881 ssize_t count;
4882 int err;
4883
4884 if (IS_IMMUTABLE(inode))
4885 return -EPERM;
4886
4887 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4888 return -EPERM;
4889
4890 count = generic_write_checks(iocb, from);
4891 if (count <= 0)
4892 return count;
4893
4894 err = file_modified(file);
4895 if (err)
4896 return err;
ba8dac35
CY		 4897
4898 filemap_invalidate_lock(inode->i_mapping);
4899 f2fs_zero_post_eof_page(inode, iocb->ki_pos + iov_iter_count(from));
4900 filemap_invalidate_unlock(inode->i_mapping);
a1e09b03
EB		 4901 return count;
4902}
4903
3d697a4a
EB		 4904/*
4905 * Preallocate blocks for a write request, if it is possible and helpful to do
4906 * so. Returns a positive number if blocks may have been preallocated, 0 if no
4907 * blocks were preallocated, or a negative errno value if something went
4908 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4909 * requested blocks (not just some of them) have been allocated.
4910 */
a1e09b03
EB		 4911static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4912 bool dio)
3d697a4a
EB		 4913{
4914 struct inode *inode = file_inode(iocb->ki_filp);
4915 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4916 const loff_t pos = iocb->ki_pos;
4917 const size_t count = iov_iter_count(iter);
4918 struct f2fs_map_blocks map = {};
3d697a4a
EB		 4919 int flag;
4920 int ret;
4921
4922 /* If it will be an out-of-place direct write, don't bother. */
4923 if (dio && f2fs_lfs_mode(sbi))
4924 return 0;
d4dd19ec
JK		 4925 /*
4926 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4927 * buffered IO, if DIO meets any holes.
4928 */
4929 if (dio && i_size_read(inode) &&
4930 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4931 return 0;
3d697a4a
EB		 4932
4933 /* No-wait I/O can't allocate blocks. */
4934 if (iocb->ki_flags & IOCB_NOWAIT)
4935 return 0;
4936
4937 /* If it will be a short write, don't bother. */
4938 if (fault_in_iov_iter_readable(iter, count))
4939 return 0;
4940
4941 if (f2fs_has_inline_data(inode)) {
4942 /* If the data will fit inline, don't bother. */
4943 if (pos + count <= MAX_INLINE_DATA(inode))
4944 return 0;
4945 ret = f2fs_convert_inline_inode(inode);
4946 if (ret)
4947 return ret;
4948 }
4949
4950 /* Do not preallocate blocks that will be written partially in 4KB. */
4951 map.m_lblk = F2FS_BLK_ALIGN(pos);
4952 map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4953 if (map.m_len > map.m_lblk)
4954 map.m_len -= map.m_lblk;
4955 else
394e7f4d
CY		 4956 return 0;
4957
128d333f
DJ		 4958 if (!IS_DEVICE_ALIASING(inode))
4959 map.m_may_create = true;
3d697a4a 4960 if (dio) {
a320b2f0
CY		 4961 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi,
4962 inode->i_write_hint);
3d697a4a
EB		 4963 flag = F2FS_GET_BLOCK_PRE_DIO;
4964 } else {
4965 map.m_seg_type = NO_CHECK_TYPE;
4966 flag = F2FS_GET_BLOCK_PRE_AIO;
4967 }
4968
cd8fc522 4969 ret = f2fs_map_blocks(inode, &map, flag);
d4dd19ec
JK		 4970 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4971 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
3d697a4a
EB		 4972 return ret;
4973 if (ret == 0)
4974 set_inode_flag(inode, FI_PREALLOCATED_ALL);
4975 return map.m_len;
4976}
4977
a1e09b03
EB		 4978static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4979 struct iov_iter *from)
fcc85a4d 4980{
b439b103
JK		 4981 struct file *file = iocb->ki_filp;
4982 struct inode *inode = file_inode(file);
a1e09b03
EB		 4983 ssize_t ret;
4984
4985 if (iocb->ki_flags & IOCB_NOWAIT)
4986 return -EOPNOTSUPP;
4987
800ba295 4988 ret = generic_perform_write(iocb, from);
a1e09b03
EB		 4989
4990 if (ret > 0) {
34a23525
CY		 4991 f2fs_update_iostat(F2FS_I_SB(inode), inode,
4992 APP_BUFFERED_IO, ret);
a1e09b03
EB		 4993 }
4994 return ret;
4995}
4996
4997static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4998 unsigned int flags)
4999{
5000 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
5001
5002 dec_page_count(sbi, F2FS_DIO_WRITE);
5003 if (error)
5004 return error;
0cc81b1a 5005 f2fs_update_time(sbi, REQ_TIME);
34a23525 5006 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
a1e09b03
EB		 5007 return 0;
5008}
5009
7643f3fe
JK		 5010static void f2fs_dio_write_submit_io(const struct iomap_iter *iter,
5011 struct bio *bio, loff_t file_offset)
5012{
5013 struct inode *inode = iter->inode;
5014 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
9395fb09
CY		 5015 enum log_type type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint);
5016 enum temp_type temp = f2fs_get_segment_temp(sbi, type);
7643f3fe
JK		 5017
5018 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp);
5019 submit_bio(bio);
5020}
5021
a1e09b03 5022static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
7643f3fe
JK		 5023 .end_io = f2fs_dio_write_end_io,
5024 .submit_io = f2fs_dio_write_submit_io,
a1e09b03
EB		 5025};
5026
92318f20
HH		 5027static void f2fs_flush_buffered_write(struct address_space *mapping,
5028 loff_t start_pos, loff_t end_pos)
5029{
5030 int ret;
5031
5032 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos);
5033 if (ret < 0)
5034 return;
5035 invalidate_mapping_pages(mapping,
5036 start_pos >> PAGE_SHIFT,
5037 end_pos >> PAGE_SHIFT);
5038}
5039
a1e09b03
EB		 5040static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
5041 bool *may_need_sync)
5042{
5043 struct file *file = iocb->ki_filp;
5044 struct inode *inode = file_inode(file);
5045 struct f2fs_inode_info *fi = F2FS_I(inode);
5046 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
5047 const bool do_opu = f2fs_lfs_mode(sbi);
a1e09b03
EB		 5048 const loff_t pos = iocb->ki_pos;
5049 const ssize_t count = iov_iter_count(from);
a1e09b03
EB		 5050 unsigned int dio_flags;
5051 struct iomap_dio *dio;
5052 ssize_t ret;
5053
bd984c03 5054 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
a1e09b03
EB		 5055
5056 if (iocb->ki_flags & IOCB_NOWAIT) {
5057 /* f2fs_convert_inline_inode() and block allocation can block */
5058 if (f2fs_has_inline_data(inode) ||
5059 !f2fs_overwrite_io(inode, pos, count)) {
5060 ret = -EAGAIN;
5061 goto out;
5062 }
5063
e4544b63 5064 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
a1e09b03
EB		 5065 ret = -EAGAIN;
5066 goto out;
5067 }
e4544b63
TM		 5068 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
5069 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03
EB		 5070 ret = -EAGAIN;
5071 goto out;
5072 }
5073 } else {
5074 ret = f2fs_convert_inline_inode(inode);
5075 if (ret)
5076 goto out;
5077
e4544b63 5078 f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03 5079 if (do_opu)
e4544b63 5080 f2fs_down_read(&fi->i_gc_rwsem[READ]);
a1e09b03 5081 }
a1e09b03
EB		 5082
5083 /*
5084 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
5085 * the higher-level function iomap_dio_rw() in order to ensure that the
5086 * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
5087 */
5088 inc_page_count(sbi, F2FS_DIO_WRITE);
5089 dio_flags = 0;
5090 if (pos + count > inode->i_size)
5091 dio_flags |= IOMAP_DIO_FORCE_WAIT;
5092 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
786f847f 5093 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
a1e09b03
EB		 5094 if (IS_ERR_OR_NULL(dio)) {
5095 ret = PTR_ERR_OR_ZERO(dio);
5096 if (ret == -ENOTBLK)
5097 ret = 0;
5098 if (ret != -EIOCBQUEUED)
5099 dec_page_count(sbi, F2FS_DIO_WRITE);
5100 } else {
5101 ret = iomap_dio_complete(dio);
5102 }
5103
a1e09b03 5104 if (do_opu)
e4544b63
TM		 5105 f2fs_up_read(&fi->i_gc_rwsem[READ]);
5106 f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
a1e09b03
EB		 5107
5108 if (ret < 0)
5109 goto out;
5110 if (pos + ret > inode->i_size)
5111 f2fs_i_size_write(inode, pos + ret);
5112 if (!do_opu)
5113 set_inode_flag(inode, FI_UPDATE_WRITE);
5114
5115 if (iov_iter_count(from)) {
5116 ssize_t ret2;
5117 loff_t bufio_start_pos = iocb->ki_pos;
5118
5119 /*
5120 * The direct write was partial, so we need to fall back to a
5121 * buffered write for the remainder.
5122 */
5123
5124 ret2 = f2fs_buffered_write_iter(iocb, from);
5125 if (iov_iter_count(from))
5126 f2fs_write_failed(inode, iocb->ki_pos);
5127 if (ret2 < 0)
5128 goto out;
5129
5130 /*
5131 * Ensure that the pagecache pages are written to disk and
5132 * invalidated to preserve the expected O_DIRECT semantics.
5133 */
5134 if (ret2 > 0) {
5135 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
5136
5137 ret += ret2;
5138
92318f20
HH		 5139 f2fs_flush_buffered_write(file->f_mapping,
5140 bufio_start_pos,
5141 bufio_end_pos);
a1e09b03
EB		 5142 }
5143 } else {
5144 /* iomap_dio_rw() already handled the generic_write_sync(). */
5145 *may_need_sync = false;
5146 }
5147out:
5148 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
5149 return ret;
5150}
5151
5152static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
5153{
5154 struct inode *inode = file_inode(iocb->ki_filp);
ccf7cf92
EB		 5155 const loff_t orig_pos = iocb->ki_pos;
5156 const size_t orig_count = iov_iter_count(from);
3d697a4a 5157 loff_t target_size;
a1e09b03
EB		 5158 bool dio;
5159 bool may_need_sync = true;
3d697a4a 5160 int preallocated;
3fdd89b4
DJ		 5161 const loff_t pos = iocb->ki_pos;
5162 const ssize_t count = iov_iter_count(from);
b439b103 5163 ssize_t ret;
fcc85a4d 5164
126ce721
CY		 5165 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
5166 ret = -EIO;
5167 goto out;
5168 }
1f227a3e 5169
7bd29358
CY		 5170 if (!f2fs_is_compress_backend_ready(inode)) {
5171 ret = -EOPNOTSUPP;
5172 goto out;
5173 }
4c8ff709 5174
cb8434f1
GR		 5175 if (iocb->ki_flags & IOCB_NOWAIT) {
5176 if (!inode_trylock(inode)) {
126ce721
CY		 5177 ret = -EAGAIN;
5178 goto out;
5179 }
cb8434f1 5180 } else {
b91050a8
HL		 5181 inode_lock(inode);
5182 }
5183
3fdd89b4
DJ		 5184 if (f2fs_is_pinned_file(inode) &&
5185 !f2fs_overwrite_io(inode, pos, count)) {
5186 ret = -EIO;
5187 goto out_unlock;
5188 }
5189
a1e09b03 5190 ret = f2fs_write_checks(iocb, from);
b31bf0f9
EB		 5191 if (ret <= 0)
5192 goto out_unlock;
5193
a1e09b03
EB		 5194 /* Determine whether we will do a direct write or a buffered write. */
5195 dio = f2fs_should_use_dio(inode, iocb, from);
3d697a4a 5196
b2c160f4
CY		 5197 /* dio is not compatible w/ atomic write */
5198 if (dio && f2fs_is_atomic_file(inode)) {
5199 ret = -EOPNOTSUPP;
5200 goto out_unlock;
5201 }
5202
b31bf0f9
EB		 5203 /* Possibly preallocate the blocks for the write. */
5204 target_size = iocb->ki_pos + iov_iter_count(from);
a1e09b03 5205 preallocated = f2fs_preallocate_blocks(iocb, from, dio);
c277f141 5206 if (preallocated < 0) {
b31bf0f9 5207 ret = preallocated;
c277f141 5208 } else {
a28bca0f 5209 if (trace_f2fs_datawrite_start_enabled())
ceb11d0e
DH		 5210 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos,
5211 orig_count, WRITE);
a28bca0f 5212
a1e09b03
EB		 5213 /* Do the actual write. */
5214 ret = dio ?
544b53da 5215 f2fs_dio_write_iter(iocb, from, &may_need_sync) :
a1e09b03 5216 f2fs_buffered_write_iter(iocb, from);
dc7a10dd 5217
c277f141
JK		 5218 if (trace_f2fs_datawrite_end_enabled())
5219 trace_f2fs_datawrite_end(inode, orig_pos, ret);
5220 }
5221
b31bf0f9 5222 /* Don't leave any preallocated blocks around past i_size. */
d4dd19ec 5223 if (preallocated && i_size_read(inode) < target_size) {
e4544b63 5224 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
b31bf0f9 5225 filemap_invalidate_lock(inode->i_mapping);
d4dd19ec
JK		 5226 if (!f2fs_truncate(inode))
5227 file_dont_truncate(inode);
b31bf0f9 5228 filemap_invalidate_unlock(inode->i_mapping);
e4544b63 5229 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
d4dd19ec
JK		 5230 } else {
5231 file_dont_truncate(inode);
b439b103 5232 }
b31bf0f9
EB		 5233
5234 clear_inode_flag(inode, FI_PREALLOCATED_ALL);
3d697a4a 5235out_unlock:
b439b103 5236 inode_unlock(inode);
126ce721 5237out:
ccf7cf92 5238 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
92318f20 5239
a1e09b03 5240 if (ret > 0 && may_need_sync)
e2592217 5241 ret = generic_write_sync(iocb, ret);
92318f20
HH		 5242
5243 /* If buffered IO was forced, flush and drop the data from
5244 * the page cache to preserve O_DIRECT semantics
5245 */
5246 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
5247 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
5248 orig_pos,
5249 orig_pos + ret - 1);
5250
b439b103 5251 return ret;
fcc85a4d
JK		 5252}
5253
0f6b56ec
DJ		 5254static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
5255 int advice)
5256{
0f6b56ec
DJ		 5257 struct address_space *mapping;
5258 struct backing_dev_info *bdi;
e64347ae
FC		 5259 struct inode *inode = file_inode(filp);
5260 int err;
0f6b56ec
DJ		 5261
5262 if (advice == POSIX_FADV_SEQUENTIAL) {
0f6b56ec
DJ		 5263 if (S_ISFIFO(inode->i_mode))
5264 return -ESPIPE;
5265
5266 mapping = filp->f_mapping;
5267 if (!mapping || len < 0)
5268 return -EINVAL;
5269
5270 bdi = inode_to_bdi(mapping->host);
5271 filp->f_ra.ra_pages = bdi->ra_pages *
5272 F2FS_I_SB(inode)->seq_file_ra_mul;
5273 spin_lock(&filp->f_lock);
5274 filp->f_mode &= ~FMODE_RANDOM;
5275 spin_unlock(&filp->f_lock);
5276 return 0;
3e729e50
JK		 5277 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) {
5278 /* Load extent cache at the first readahead. */
5279 f2fs_precache_extents(inode);
0f6b56ec
DJ		 5280 }
5281
e64347ae 5282 err = generic_fadvise(filp, offset, len, advice);
ef0c333c
JK		 5283 if (err)
5284 return err;
e64347ae 5285
ef0c333c
JK		 5286 if (advice == POSIX_FADV_DONTNEED &&
5287 (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
5288 f2fs_compressed_file(inode)))
5289 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
5290 else if (advice == POSIX_FADV_NOREUSE)
cf7cd17c
DJ		 5291 err = f2fs_keep_noreuse_range(inode, offset, len);
5292 return err;
0f6b56ec
DJ		 5293}
5294
e9750824 5295#ifdef CONFIG_COMPAT
34178b1b
CY		 5296struct compat_f2fs_gc_range {
5297 u32 sync;
5298 compat_u64 start;
5299 compat_u64 len;
5300};
5301#define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\
5302 struct compat_f2fs_gc_range)
5303
5304static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
5305{
5306 struct compat_f2fs_gc_range __user *urange;
5307 struct f2fs_gc_range range;
5308 int err;
5309
5310 urange = compat_ptr(arg);
5311 err = get_user(range.sync, &urange->sync);
5312 err |= get_user(range.start, &urange->start);
5313 err |= get_user(range.len, &urange->len);
5314 if (err)
5315 return -EFAULT;
5316
5317 return __f2fs_ioc_gc_range(file, &range);
5318}
5319
5320struct compat_f2fs_move_range {
5321 u32 dst_fd;
5322 compat_u64 pos_in;
5323 compat_u64 pos_out;
5324 compat_u64 len;
5325};
5326#define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \
5327 struct compat_f2fs_move_range)
5328
5329static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
5330{
5331 struct compat_f2fs_move_range __user *urange;
5332 struct f2fs_move_range range;
5333 int err;
5334
5335 urange = compat_ptr(arg);
5336 err = get_user(range.dst_fd, &urange->dst_fd);
5337 err |= get_user(range.pos_in, &urange->pos_in);
5338 err |= get_user(range.pos_out, &urange->pos_out);
5339 err |= get_user(range.len, &urange->len);
5340 if (err)
5341 return -EFAULT;
5342
5343 return __f2fs_ioc_move_range(file, &range);
5344}
5345
e9750824
NJ		 5346long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5347{
34178b1b
CY		 5348 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
5349 return -EIO;
5350 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
5351 return -ENOSPC;
5352
e9750824 5353 switch (cmd) {
3357af8f
EB		 5354 case FS_IOC32_GETVERSION:
5355 cmd = FS_IOC_GETVERSION;
04ef4b62 5356 break;
34178b1b
CY		 5357 case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
5358 return f2fs_compat_ioc_gc_range(file, arg);
5359 case F2FS_IOC32_MOVE_RANGE:
5360 return f2fs_compat_ioc_move_range(file, arg);
04ef4b62 5361 case F2FS_IOC_START_ATOMIC_WRITE:
933141e4 5362 case F2FS_IOC_START_ATOMIC_REPLACE:
04ef4b62
CY		 5363 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
5364 case F2FS_IOC_START_VOLATILE_WRITE:
5365 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
23339e57 5366 case F2FS_IOC_ABORT_ATOMIC_WRITE:
04ef4b62 5367 case F2FS_IOC_SHUTDOWN:
314999dc 5368 case FITRIM:
3357af8f
EB		 5369 case FS_IOC_SET_ENCRYPTION_POLICY:
5370 case FS_IOC_GET_ENCRYPTION_PWSALT:
5371 case FS_IOC_GET_ENCRYPTION_POLICY:
8ce589c7
EB		 5372 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
5373 case FS_IOC_ADD_ENCRYPTION_KEY:
5374 case FS_IOC_REMOVE_ENCRYPTION_KEY:
5375 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
5376 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
ee446e1a 5377 case FS_IOC_GET_ENCRYPTION_NONCE:
04ef4b62
CY		 5378 case F2FS_IOC_GARBAGE_COLLECT:
5379 case F2FS_IOC_WRITE_CHECKPOINT:
5380 case F2FS_IOC_DEFRAGMENT:
e066b83c 5381 case F2FS_IOC_FLUSH_DEVICE:
e65ef207 5382 case F2FS_IOC_GET_FEATURES:
1ad71a27
JK		 5383 case F2FS_IOC_GET_PIN_FILE:
5384 case F2FS_IOC_SET_PIN_FILE:
c4020b2d 5385 case F2FS_IOC_PRECACHE_EXTENTS:
04f0b2ea 5386 case F2FS_IOC_RESIZE_FS:
95ae251f
EB		 5387 case FS_IOC_ENABLE_VERITY:
5388 case FS_IOC_MEASURE_VERITY:
e17fe657 5389 case FS_IOC_READ_VERITY_METADATA:
3357af8f
EB		 5390 case FS_IOC_GETFSLABEL:
5391 case FS_IOC_SETFSLABEL:
439dfb10 5392 case F2FS_IOC_GET_COMPRESS_BLOCKS:
ef8d563f 5393 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
c75488fb 5394 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
9af84648 5395 case F2FS_IOC_SEC_TRIM_FILE:
9e2a5f8c 5396 case F2FS_IOC_GET_COMPRESS_OPTION:
e1e8debe 5397 case F2FS_IOC_SET_COMPRESS_OPTION:
5fdb322f
DJ		 5398 case F2FS_IOC_DECOMPRESS_FILE:
5399 case F2FS_IOC_COMPRESS_FILE:
128d333f 5400 case F2FS_IOC_GET_DEV_ALIAS_FILE:
5f95c181 5401 case F2FS_IOC_IO_PRIO:
4dd6f977 5402 break;
e9750824
NJ		 5403 default:
5404 return -ENOIOCTLCMD;
5405 }
34178b1b 5406 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
e9750824
NJ		 5407}
5408#endif
5409
fbfa2cc5 5410const struct file_operations f2fs_file_operations = {
267378d4 5411 .llseek = f2fs_llseek,
4c8ff709 5412 .read_iter = f2fs_file_read_iter,
fcc85a4d 5413 .write_iter = f2fs_file_write_iter,
50aa6f44 5414 .iopoll = iocb_bio_iopoll,
fcc85a4d 5415 .open = f2fs_file_open,
12662234 5416 .release = f2fs_release_file,
fbfa2cc5 5417 .mmap = f2fs_file_mmap,
7a10f017 5418 .flush = f2fs_file_flush,
fbfa2cc5
JK		 5419 .fsync = f2fs_sync_file,
5420 .fallocate = f2fs_fallocate,
5421 .unlocked_ioctl = f2fs_ioctl,
e9750824
NJ		 5422#ifdef CONFIG_COMPAT
5423 .compat_ioctl = f2fs_compat_ioctl,
5424#endif
ceb11d0e 5425 .splice_read = f2fs_file_splice_read,
8d020765 5426 .splice_write = iter_file_splice_write,
0f6b56ec 5427 .fadvise = f2fs_file_fadvise,
210a03c9 5428 .fop_flags = FOP_BUFFER_RASYNC,
fbfa2cc5 5429};
Linux 6.x block layer and io_uring tree(s)