projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: avoid frequent costly fsck triggers
[linux-block.git] / fs / f2fs / file.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
fbfa2cc5
JK		 3 * fs/f2fs/file.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
fbfa2cc5
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/stat.h>
11#include <linux/buffer_head.h>
12#include <linux/writeback.h>
ae51fb31 13#include <linux/blkdev.h>
fbfa2cc5
JK		 14#include <linux/falloc.h>
15#include <linux/types.h>
e9750824 16#include <linux/compat.h>
fbfa2cc5
JK		 17#include <linux/uaccess.h>
18#include <linux/mount.h>
7f7670fe 19#include <linux/pagevec.h>
dc91de78 20#include <linux/uio.h>
8da4b8c4 21#include <linux/uuid.h>
4dd6f977 22#include <linux/file.h>
fbfa2cc5
JK		 23
24#include "f2fs.h"
25#include "node.h"
26#include "segment.h"
27#include "xattr.h"
28#include "acl.h"
c1c1b583 29#include "gc.h"
db9f7c1a 30#include "trace.h"
a2a4a7e4 31#include <trace/events/f2fs.h>
fbfa2cc5 32
ea4d479b 33static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
5a3a2d83
QS		 34{
35 struct inode *inode = file_inode(vmf->vma->vm_file);
ea4d479b 36 vm_fault_t ret;
5a3a2d83
QS		 37
38 down_read(&F2FS_I(inode)->i_mmap_sem);
ea4d479b 39 ret = filemap_fault(vmf);
5a3a2d83
QS		 40 up_read(&F2FS_I(inode)->i_mmap_sem);
41
ea4d479b 42 return ret;
5a3a2d83
QS		 43}
44
ea4d479b 45static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
fbfa2cc5
JK		 46{
47 struct page *page = vmf->page;
11bac800 48 struct inode *inode = file_inode(vmf->vma->vm_file);
4081363f 49 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
39a86958 50 struct dnode_of_data dn = { .node_changed = false };
e479556b 51 int err;
fbfa2cc5 52
1f227a3e
JK		 53 if (unlikely(f2fs_cp_error(sbi))) {
54 err = -EIO;
55 goto err;
56 }
57
fbfa2cc5 58 sb_start_pagefault(inode->i_sb);
b3d208f9
JK		 59
60 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
b067ba1f 61
11bac800 62 file_update_time(vmf->vma->vm_file);
5a3a2d83 63 down_read(&F2FS_I(inode)->i_mmap_sem);
fbfa2cc5 64 lock_page(page);
6bacf52f 65 if (unlikely(page->mapping != inode->i_mapping ||
9851e6e1 66 page_offset(page) > i_size_read(inode) ||
6bacf52f 67 !PageUptodate(page))) {
fbfa2cc5
JK		 68 unlock_page(page);
69 err = -EFAULT;
5a3a2d83 70 goto out_sem;
fbfa2cc5
JK		 71 }
72
39a86958
CY		 73 /* block allocation */
74 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
75 set_new_dnode(&dn, inode, NULL, NULL, 0);
76 err = f2fs_get_block(&dn, page->index);
77 f2fs_put_dnode(&dn);
78 __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
79 if (err) {
80 unlock_page(page);
81 goto out_sem;
82 }
83
84 /* fill the page */
85 f2fs_wait_on_page_writeback(page, DATA, false);
86
87 /* wait for GCed page writeback via META_MAPPING */
88 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
89
fbfa2cc5
JK		 90 /*
91 * check to see if the page is mapped already (no holes)
92 */
93 if (PageMappedToDisk(page))
39a86958 94 goto out_sem;
fbfa2cc5
JK		 95
96 /* page is wholly or partially inside EOF */
09cbfeaf 97 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
9edcdabf 98 i_size_read(inode)) {
193bea1d 99 loff_t offset;
f11e98bd 100
09cbfeaf
KS		 101 offset = i_size_read(inode) & ~PAGE_MASK;
102 zero_user_segment(page, offset, PAGE_SIZE);
fbfa2cc5
JK		 103 }
104 set_page_dirty(page);
237c0790
JK		 105 if (!PageUptodate(page))
106 SetPageUptodate(page);
fbfa2cc5 107
b0af6d49 108 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
c75f2feb 109 f2fs_update_time(sbi, REQ_TIME);
b0af6d49 110
e943a10d 111 trace_f2fs_vm_page_mkwrite(page, DATA);
5a3a2d83
QS		 112out_sem:
113 up_read(&F2FS_I(inode)->i_mmap_sem);
39a86958
CY		 114
115 f2fs_balance_fs(sbi, dn.node_changed);
116
fbfa2cc5 117 sb_end_pagefault(inode->i_sb);
1f227a3e 118err:
fbfa2cc5
JK		 119 return block_page_mkwrite_return(err);
120}
121
122static const struct vm_operations_struct f2fs_file_vm_ops = {
5a3a2d83 123 .fault = f2fs_filemap_fault,
f1820361 124 .map_pages = filemap_map_pages,
692bb55d 125 .page_mkwrite = f2fs_vm_page_mkwrite,
fbfa2cc5
JK		 126};
127
354a3399
JK		 128static int get_parent_ino(struct inode *inode, nid_t *pino)
129{
130 struct dentry *dentry;
131
132 inode = igrab(inode);
133 dentry = d_find_any_alias(inode);
134 iput(inode);
135 if (!dentry)
136 return 0;
137
f0947e5c
JK		 138 *pino = parent_ino(dentry);
139 dput(dentry);
354a3399
JK		 140 return 1;
141}
142
a5fd5050 143static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
9d1589ef 144{
4081363f 145 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
a5fd5050 146 enum cp_reason_type cp_reason = CP_NO_NEEDED;
9d1589ef 147
a5fd5050
CY		 148 if (!S_ISREG(inode->i_mode))
149 cp_reason = CP_NON_REGULAR;
150 else if (inode->i_nlink != 1)
151 cp_reason = CP_HARDLINK;
bbf156f7 152 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
a5fd5050 153 cp_reason = CP_SB_NEED_CP;
9d1589ef 154 else if (file_wrong_pino(inode))
a5fd5050 155 cp_reason = CP_WRONG_PINO;
4d57b86d 156 else if (!f2fs_space_for_roll_forward(sbi))
a5fd5050 157 cp_reason = CP_NO_SPC_ROLL;
4d57b86d 158 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
a5fd5050 159 cp_reason = CP_NODE_NEED_CP;
d5053a34 160 else if (test_opt(sbi, FASTBOOT))
a5fd5050 161 cp_reason = CP_FASTBOOT_MODE;
63189b78 162 else if (F2FS_OPTION(sbi).active_logs == 2)
a5fd5050 163 cp_reason = CP_SPEC_LOG_NUM;
63189b78 164 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
4d57b86d
CY		 165 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
166 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
167 TRANS_DIR_INO))
0a007b97 168 cp_reason = CP_RECOVER_DIR;
9d1589ef 169
a5fd5050 170 return cp_reason;
9d1589ef
CY		 171}
172
9c7bb702
CL		 173static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
174{
175 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
176 bool ret = false;
177 /* But we need to avoid that there are some inode updates */
4d57b86d 178 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
9c7bb702
CL		 179 ret = true;
180 f2fs_put_page(i, 0);
181 return ret;
182}
183
51455b19
CL		 184static void try_to_fix_pino(struct inode *inode)
185{
186 struct f2fs_inode_info *fi = F2FS_I(inode);
187 nid_t pino;
188
189 down_write(&fi->i_sem);
51455b19
CL		 190 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
191 get_parent_ino(inode, &pino)) {
205b9822 192 f2fs_i_pino_write(inode, pino);
51455b19 193 file_got_pino(inode);
51455b19 194 }
ee6d182f 195 up_write(&fi->i_sem);
51455b19
CL		 196}
197
608514de
JK		 198static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
199 int datasync, bool atomic)
fbfa2cc5
JK		 200{
201 struct inode *inode = file->f_mapping->host;
4081363f 202 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2403c155 203 nid_t ino = inode->i_ino;
fbfa2cc5 204 int ret = 0;
a5fd5050 205 enum cp_reason_type cp_reason = 0;
fbfa2cc5 206 struct writeback_control wbc = {
c81bf1c8 207 .sync_mode = WB_SYNC_ALL,
fbfa2cc5
JK		 208 .nr_to_write = LONG_MAX,
209 .for_reclaim = 0,
210 };
50fa53ec 211 unsigned int seq_id = 0;
fbfa2cc5 212
4354994f
DR		 213 if (unlikely(f2fs_readonly(inode->i_sb) ||
214 is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1fa95b0b
NJ		 215 return 0;
216
a2a4a7e4 217 trace_f2fs_sync_file_enter(inode);
ea1aa12c 218
b61ac5b7
YH		 219 if (S_ISDIR(inode->i_mode))
220 goto go_write;
221
ea1aa12c 222 /* if fdatasync is triggered, let's do in-place-update */
c46a155b 223 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
91942321 224 set_inode_flag(inode, FI_NEED_IPU);
3b49c9a1 225 ret = file_write_and_wait_range(file, start, end);
91942321 226 clear_inode_flag(inode, FI_NEED_IPU);
c1ce1b02 227
a2a4a7e4 228 if (ret) {
a5fd5050 229 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
fbfa2cc5 230 return ret;
a2a4a7e4 231 }
fbfa2cc5 232
9c7bb702 233 /* if the inode is dirty, let's recover all the time */
281518c6 234 if (!f2fs_skip_inode_update(inode, datasync)) {
2286c020 235 f2fs_write_inode(inode, NULL);
9c7bb702
CL		 236 goto go_write;
237 }
238
6d99ba41
JK		 239 /*
240 * if there is no written data, don't waste time to write recovery info.
241 */
91942321 242 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
4d57b86d 243 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
19c9c466 244
9c7bb702
CL		 245 /* it may call write_inode just prior to fsync */
246 if (need_inode_page_update(sbi, ino))
19c9c466 247 goto go_write;
19c9c466 248
91942321 249 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
4d57b86d 250 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
6d99ba41
JK		 251 goto flush_out;
252 goto out;
253 }
19c9c466 254go_write:
e5d2385e
JK		 255 /*
256 * Both of fdatasync() and fsync() are able to be recovered from
257 * sudden-power-off.
258 */
91942321 259 down_read(&F2FS_I(inode)->i_sem);
a5fd5050 260 cp_reason = need_do_checkpoint(inode);
91942321 261 up_read(&F2FS_I(inode)->i_sem);
d928bfbf 262
a5fd5050 263 if (cp_reason) {
fbfa2cc5
JK		 264 /* all the dirty node pages should be flushed for POR */
265 ret = f2fs_sync_fs(inode->i_sb, 1);
d928bfbf 266
51455b19
CL		 267 /*
268 * We've secured consistency through sync_fs. Following pino
269 * will be used only for fsynced inodes after checkpoint.
270 */
271 try_to_fix_pino(inode);
91942321
JK		 272 clear_inode_flag(inode, FI_APPEND_WRITE);
273 clear_inode_flag(inode, FI_UPDATE_WRITE);
51455b19
CL		 274 goto out;
275 }
88bd02c9 276sync_nodes:
c29fd0c0 277 atomic_inc(&sbi->wb_sync_req[NODE]);
50fa53ec 278 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
c29fd0c0 279 atomic_dec(&sbi->wb_sync_req[NODE]);
c267ec15
JK		 280 if (ret)
281 goto out;
51455b19 282
871f599f 283 /* if cp_error was enabled, we should avoid infinite loop */
6d5a1495
CY		 284 if (unlikely(f2fs_cp_error(sbi))) {
285 ret = -EIO;
871f599f 286 goto out;
6d5a1495 287 }
871f599f 288
4d57b86d 289 if (f2fs_need_inode_block_update(sbi, ino)) {
7c45729a 290 f2fs_mark_inode_dirty_sync(inode, true);
51455b19
CL		 291 f2fs_write_inode(inode, NULL);
292 goto sync_nodes;
fbfa2cc5 293 }
51455b19 294
b6a245eb
JK		 295 /*
296 * If it's atomic_write, it's just fine to keep write ordering. So
297 * here we don't need to wait for node write completion, since we use
298 * node chain which serializes node blocks. If one of node writes are
299 * reordered, we can see simply broken chain, resulting in stopping
300 * roll-forward recovery. It means we'll recover all or none node blocks
301 * given fsync mark.
302 */
303 if (!atomic) {
50fa53ec 304 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
b6a245eb
JK		 305 if (ret)
306 goto out;
307 }
51455b19
CL		 308
309 /* once recovery info is written, don't need to tack this */
4d57b86d 310 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
91942321 311 clear_inode_flag(inode, FI_APPEND_WRITE);
51455b19 312flush_out:
d6290814 313 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
39d787be 314 ret = f2fs_issue_flush(sbi, inode->i_ino);
3f06252f 315 if (!ret) {
4d57b86d 316 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
3f06252f 317 clear_inode_flag(inode, FI_UPDATE_WRITE);
4d57b86d 318 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
3f06252f 319 }
d0239e1b 320 f2fs_update_time(sbi, REQ_TIME);
fbfa2cc5 321out:
a5fd5050 322 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
05ca3632 323 f2fs_trace_ios(NULL, 1);
fbfa2cc5
JK		 324 return ret;
325}
326
608514de
JK		 327int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
328{
1f227a3e
JK		 329 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
330 return -EIO;
608514de
JK		 331 return f2fs_do_sync_file(file, start, end, datasync, false);
332}
333
7f7670fe
JK		 334static pgoff_t __get_first_dirty_index(struct address_space *mapping,
335 pgoff_t pgofs, int whence)
336{
8faab642 337 struct page *page;
7f7670fe
JK		 338 int nr_pages;
339
340 if (whence != SEEK_DATA)
341 return 0;
342
343 /* find first dirty page index */
8faab642
JK		 344 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
345 1, &page);
346 if (!nr_pages)
347 return ULONG_MAX;
348 pgofs = page->index;
349 put_page(page);
7f7670fe
JK		 350 return pgofs;
351}
352
e1da7872
CY		 353static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
354 pgoff_t dirty, pgoff_t pgofs, int whence)
7f7670fe
JK		 355{
356 switch (whence) {
357 case SEEK_DATA:
358 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
e1da7872 359 is_valid_data_blkaddr(sbi, blkaddr))
7f7670fe
JK		 360 return true;
361 break;
362 case SEEK_HOLE:
363 if (blkaddr == NULL_ADDR)
364 return true;
365 break;
366 }
367 return false;
368}
369
267378d4
CY		 370static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
371{
372 struct inode *inode = file->f_mapping->host;
373 loff_t maxbytes = inode->i_sb->s_maxbytes;
374 struct dnode_of_data dn;
7f7670fe
JK		 375 pgoff_t pgofs, end_offset, dirty;
376 loff_t data_ofs = offset;
377 loff_t isize;
267378d4
CY		 378 int err = 0;
379
5955102c 380 inode_lock(inode);
267378d4
CY		 381
382 isize = i_size_read(inode);
383 if (offset >= isize)
384 goto fail;
385
386 /* handle inline data case */
622f28ae 387 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
267378d4
CY		 388 if (whence == SEEK_HOLE)
389 data_ofs = isize;
390 goto found;
391 }
392
09cbfeaf 393 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
267378d4 394
7f7670fe
JK		 395 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
396
09cbfeaf 397 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 398 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 399 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
267378d4
CY		 400 if (err && err != -ENOENT) {
401 goto fail;
402 } else if (err == -ENOENT) {
e1c42045 403 /* direct node does not exists */
267378d4 404 if (whence == SEEK_DATA) {
4d57b86d 405 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
267378d4
CY		 406 continue;
407 } else {
408 goto found;
409 }
410 }
411
81ca7350 412 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
267378d4
CY		 413
414 /* find data/hole in dnode block */
415 for (; dn.ofs_in_node < end_offset;
416 dn.ofs_in_node++, pgofs++,
09cbfeaf 417 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
267378d4 418 block_t blkaddr;
f11e98bd 419
7a2af766
CY		 420 blkaddr = datablock_addr(dn.inode,
421 dn.node_page, dn.ofs_in_node);
267378d4 422
c9b60788
CY		 423 if (__is_valid_data_blkaddr(blkaddr) &&
424 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
425 blkaddr, DATA_GENERIC)) {
426 f2fs_put_dnode(&dn);
427 goto fail;
428 }
429
e1da7872
CY		 430 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
431 pgofs, whence)) {
267378d4
CY		 432 f2fs_put_dnode(&dn);
433 goto found;
434 }
435 }
436 f2fs_put_dnode(&dn);
437 }
438
439 if (whence == SEEK_DATA)
440 goto fail;
267378d4 441found:
fe369bc8
JK		 442 if (whence == SEEK_HOLE && data_ofs > isize)
443 data_ofs = isize;
5955102c 444 inode_unlock(inode);
267378d4
CY		 445 return vfs_setpos(file, data_ofs, maxbytes);
446fail:
5955102c 447 inode_unlock(inode);
267378d4
CY		 448 return -ENXIO;
449}
450
451static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
452{
453 struct inode *inode = file->f_mapping->host;
454 loff_t maxbytes = inode->i_sb->s_maxbytes;
455
456 switch (whence) {
457 case SEEK_SET:
458 case SEEK_CUR:
459 case SEEK_END:
460 return generic_file_llseek_size(file, offset, whence,
461 maxbytes, i_size_read(inode));
462 case SEEK_DATA:
463 case SEEK_HOLE:
0b4c5afd
JK		 464 if (offset < 0)
465 return -ENXIO;
267378d4
CY		 466 return f2fs_seek_block(file, offset, whence);
467 }
468
469 return -EINVAL;
470}
471
fbfa2cc5
JK		 472static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
473{
b3d208f9 474 struct inode *inode = file_inode(file);
b9d777b8 475 int err;
b3d208f9 476
1f227a3e
JK		 477 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
478 return -EIO;
479
b3d208f9 480 /* we don't need to use inline_data strictly */
b9d777b8
JK		 481 err = f2fs_convert_inline_inode(inode);
482 if (err)
483 return err;
b3d208f9 484
fbfa2cc5
JK		 485 file_accessed(file);
486 vma->vm_ops = &f2fs_file_vm_ops;
487 return 0;
488}
489
fcc85a4d
JK		 490static int f2fs_file_open(struct inode *inode, struct file *filp)
491{
2e168c82 492 int err = fscrypt_file_open(inode, filp);
fcc85a4d 493
2e168c82
EB		 494 if (err)
495 return err;
b91050a8
HL		 496
497 filp->f_mode |= FMODE_NOWAIT;
498
0abd675e 499 return dquot_file_open(inode, filp);
fcc85a4d
JK		 500}
501
4d57b86d 502void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
fbfa2cc5 503{
4081363f 504 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
fbfa2cc5 505 struct f2fs_node *raw_node;
19b2c30d 506 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
fbfa2cc5 507 __le32 *addr;
7a2af766
CY		 508 int base = 0;
509
510 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
511 base = get_extra_isize(dn->inode);
fbfa2cc5 512
45590710 513 raw_node = F2FS_NODE(dn->node_page);
7a2af766 514 addr = blkaddr_in_node(raw_node) + base + ofs;
fbfa2cc5 515
6c311ec6 516 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
fbfa2cc5 517 block_t blkaddr = le32_to_cpu(*addr);
f11e98bd 518
fbfa2cc5
JK		 519 if (blkaddr == NULL_ADDR)
520 continue;
521
e1509cf2 522 dn->data_blkaddr = NULL_ADDR;
4d57b86d 523 f2fs_set_data_blkaddr(dn);
c9b60788
CY		 524
525 if (__is_valid_data_blkaddr(blkaddr) &&
526 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
527 continue;
528
4d57b86d 529 f2fs_invalidate_blocks(sbi, blkaddr);
3c6c2beb 530 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
91942321 531 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
fbfa2cc5
JK		 532 nr_free++;
533 }
19b2c30d 534
fbfa2cc5 535 if (nr_free) {
19b2c30d
CY		 536 pgoff_t fofs;
537 /*
538 * once we invalidate valid blkaddr in range [ofs, ofs + count],
539 * we will invalidate all blkaddr in the whole range.
540 */
4d57b86d 541 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
81ca7350 542 dn->inode) + ofs;
19b2c30d 543 f2fs_update_extent_cache_range(dn, fofs, 0, len);
d7cc950b 544 dec_valid_block_count(sbi, dn->inode, nr_free);
fbfa2cc5
JK		 545 }
546 dn->ofs_in_node = ofs;
51dd6249 547
d0239e1b 548 f2fs_update_time(sbi, REQ_TIME);
51dd6249
NJ		 549 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
550 dn->ofs_in_node, nr_free);
fbfa2cc5
JK		 551}
552
4d57b86d 553void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
fbfa2cc5 554{
4d57b86d 555 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
fbfa2cc5
JK		 556}
557
0bfcfcca 558static int truncate_partial_data_page(struct inode *inode, u64 from,
43f3eae1 559 bool cache_only)
fbfa2cc5 560{
193bea1d 561 loff_t offset = from & (PAGE_SIZE - 1);
09cbfeaf 562 pgoff_t index = from >> PAGE_SHIFT;
43f3eae1 563 struct address_space *mapping = inode->i_mapping;
fbfa2cc5
JK		 564 struct page *page;
565
43f3eae1 566 if (!offset && !cache_only)
b3d208f9 567 return 0;
fbfa2cc5 568
43f3eae1 569 if (cache_only) {
34b5d5c2 570 page = find_lock_page(mapping, index);
43f3eae1
JK		 571 if (page && PageUptodate(page))
572 goto truncate_out;
573 f2fs_put_page(page, 1);
b3d208f9 574 return 0;
43f3eae1 575 }
fbfa2cc5 576
4d57b86d 577 page = f2fs_get_lock_data_page(inode, index, true);
43f3eae1 578 if (IS_ERR(page))
a78aaa2c 579 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
43f3eae1 580truncate_out:
fec1d657 581 f2fs_wait_on_page_writeback(page, DATA, true);
09cbfeaf 582 zero_user(page, offset, PAGE_SIZE - offset);
a9bcf9bc
JK		 583
584 /* An encrypted inode should have a key and truncate the last page. */
585 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
586 if (!cache_only)
0bfcfcca 587 set_page_dirty(page);
fbfa2cc5 588 f2fs_put_page(page, 1);
b3d208f9 589 return 0;
fbfa2cc5
JK		 590}
591
af033b2a
CY		 592int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock,
593 bool buf_write)
fbfa2cc5 594{
4081363f 595 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 596 struct dnode_of_data dn;
597 pgoff_t free_from;
9ffe0fb5 598 int count = 0, err = 0;
b3d208f9 599 struct page *ipage;
0bfcfcca 600 bool truncate_page = false;
af033b2a 601 int flag = buf_write ? F2FS_GET_BLOCK_PRE_AIO : F2FS_GET_BLOCK_PRE_DIO;
fbfa2cc5 602
51dd6249
NJ		 603 trace_f2fs_truncate_blocks_enter(inode, from);
604
df033caf 605 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
fbfa2cc5 606
09210c97
CY		 607 if (free_from >= sbi->max_file_blocks)
608 goto free_partial;
609
764aa3e9 610 if (lock)
af033b2a 611 __do_map_lock(sbi, flag, true);
9ffe0fb5 612
4d57b86d 613 ipage = f2fs_get_node_page(sbi, inode->i_ino);
b3d208f9
JK		 614 if (IS_ERR(ipage)) {
615 err = PTR_ERR(ipage);
616 goto out;
617 }
618
619 if (f2fs_has_inline_data(inode)) {
4d57b86d 620 f2fs_truncate_inline_inode(inode, ipage, from);
b3d208f9 621 f2fs_put_page(ipage, 1);
0bfcfcca 622 truncate_page = true;
b3d208f9
JK		 623 goto out;
624 }
625
626 set_new_dnode(&dn, inode, ipage, NULL, 0);
4d57b86d 627 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
fbfa2cc5
JK		 628 if (err) {
629 if (err == -ENOENT)
630 goto free_next;
b3d208f9 631 goto out;
1ce86bf6
JK		 632 }
633
81ca7350 634 count = ADDRS_PER_PAGE(dn.node_page, inode);
fbfa2cc5
JK		 635
636 count -= dn.ofs_in_node;
9850cf4a 637 f2fs_bug_on(sbi, count < 0);
39936837 638
fbfa2cc5 639 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
4d57b86d 640 f2fs_truncate_data_blocks_range(&dn, count);
fbfa2cc5
JK		 641 free_from += count;
642 }
643
644 f2fs_put_dnode(&dn);
645free_next:
4d57b86d 646 err = f2fs_truncate_inode_blocks(inode, free_from);
764d2c80
JK		 647out:
648 if (lock)
af033b2a 649 __do_map_lock(sbi, flag, false);
09210c97 650free_partial:
b3d208f9
JK		 651 /* lastly zero out the first data page */
652 if (!err)
0bfcfcca 653 err = truncate_partial_data_page(inode, from, truncate_page);
fbfa2cc5 654
51dd6249 655 trace_f2fs_truncate_blocks_exit(inode, err);
fbfa2cc5
JK		 656 return err;
657}
658
9a449e9c 659int f2fs_truncate(struct inode *inode)
fbfa2cc5 660{
b0154891
CY		 661 int err;
662
1f227a3e
JK		 663 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
664 return -EIO;
665
fbfa2cc5
JK		 666 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
667 S_ISLNK(inode->i_mode)))
b0154891 668 return 0;
fbfa2cc5 669
51dd6249
NJ		 670 trace_f2fs_truncate(inode);
671
14b44d23
JK		 672 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
673 f2fs_show_injection_info(FAULT_TRUNCATE);
674 return -EIO;
675 }
7fa750a1 676
92dffd01 677 /* we should check inline_data size */
b9d777b8 678 if (!f2fs_may_inline_data(inode)) {
b0154891
CY		 679 err = f2fs_convert_inline_inode(inode);
680 if (err)
681 return err;
92dffd01
JK		 682 }
683
af033b2a 684 err = f2fs_truncate_blocks(inode, i_size_read(inode), true, false);
b0154891
CY		 685 if (err)
686 return err;
687
078cd827 688 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 689 f2fs_mark_inode_dirty_sync(inode, false);
b0154891 690 return 0;
fbfa2cc5
JK		 691}
692
a528d35e 693int f2fs_getattr(const struct path *path, struct kstat *stat,
1c6d8ee4 694 u32 request_mask, unsigned int query_flags)
fbfa2cc5 695{
a528d35e 696 struct inode *inode = d_inode(path->dentry);
1c6d8ee4 697 struct f2fs_inode_info *fi = F2FS_I(inode);
1c1d35df 698 struct f2fs_inode *ri;
1c6d8ee4
CY		 699 unsigned int flags;
700
1c1d35df 701 if (f2fs_has_extra_attr(inode) &&
7beb01f7 702 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
1c1d35df
CY		 703 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
704 stat->result_mask |= STATX_BTIME;
705 stat->btime.tv_sec = fi->i_crtime.tv_sec;
706 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
707 }
708
c807a7cb 709 flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
59c84408 710 if (flags & F2FS_APPEND_FL)
1c6d8ee4 711 stat->attributes |= STATX_ATTR_APPEND;
59c84408 712 if (flags & F2FS_COMPR_FL)
1c6d8ee4
CY		 713 stat->attributes |= STATX_ATTR_COMPRESSED;
714 if (f2fs_encrypted_inode(inode))
715 stat->attributes |= STATX_ATTR_ENCRYPTED;
59c84408 716 if (flags & F2FS_IMMUTABLE_FL)
1c6d8ee4 717 stat->attributes |= STATX_ATTR_IMMUTABLE;
59c84408 718 if (flags & F2FS_NODUMP_FL)
1c6d8ee4
CY		 719 stat->attributes |= STATX_ATTR_NODUMP;
720
721 stat->attributes_mask |= (STATX_ATTR_APPEND |
722 STATX_ATTR_COMPRESSED |
723 STATX_ATTR_ENCRYPTED |
724 STATX_ATTR_IMMUTABLE |
725 STATX_ATTR_NODUMP);
726
fbfa2cc5 727 generic_fillattr(inode, stat);
5b4267d1
JK		 728
729 /* we need to show initial sectors used for inline_data/dentries */
730 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
731 f2fs_has_inline_dentry(inode))
732 stat->blocks += (stat->size + 511) >> 9;
733
fbfa2cc5
JK		 734 return 0;
735}
736
737#ifdef CONFIG_F2FS_FS_POSIX_ACL
738static void __setattr_copy(struct inode *inode, const struct iattr *attr)
739{
fbfa2cc5
JK		 740 unsigned int ia_valid = attr->ia_valid;
741
742 if (ia_valid & ATTR_UID)
743 inode->i_uid = attr->ia_uid;
744 if (ia_valid & ATTR_GID)
745 inode->i_gid = attr->ia_gid;
746 if (ia_valid & ATTR_ATIME)
95582b00
DD		 747 inode->i_atime = timespec64_trunc(attr->ia_atime,
748 inode->i_sb->s_time_gran);
fbfa2cc5 749 if (ia_valid & ATTR_MTIME)
95582b00
DD		 750 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
751 inode->i_sb->s_time_gran);
fbfa2cc5 752 if (ia_valid & ATTR_CTIME)
95582b00
DD		 753 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
754 inode->i_sb->s_time_gran);
fbfa2cc5
JK		 755 if (ia_valid & ATTR_MODE) {
756 umode_t mode = attr->ia_mode;
757
758 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
759 mode &= ~S_ISGID;
91942321 760 set_acl_inode(inode, mode);
fbfa2cc5
JK		 761 }
762}
763#else
764#define __setattr_copy setattr_copy
765#endif
766
767int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
768{
2b0143b5 769 struct inode *inode = d_inode(dentry);
fbfa2cc5 770 int err;
c0ed4405 771 bool size_changed = false;
fbfa2cc5 772
1f227a3e
JK		 773 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
774 return -EIO;
775
31051c85 776 err = setattr_prepare(dentry, attr);
fbfa2cc5
JK		 777 if (err)
778 return err;
779
20bb2479
EB		 780 err = fscrypt_prepare_setattr(dentry, attr);
781 if (err)
782 return err;
783
0abd675e
CY		 784 if (is_quota_modification(inode, attr)) {
785 err = dquot_initialize(inode);
786 if (err)
787 return err;
788 }
789 if ((attr->ia_valid & ATTR_UID &&
790 !uid_eq(attr->ia_uid, inode->i_uid)) ||
791 (attr->ia_valid & ATTR_GID &&
792 !gid_eq(attr->ia_gid, inode->i_gid))) {
af033b2a 793 f2fs_lock_op(F2FS_I_SB(inode));
0abd675e 794 err = dquot_transfer(inode, attr);
af033b2a
CY		 795 if (err) {
796 set_sbi_flag(F2FS_I_SB(inode),
797 SBI_QUOTA_NEED_REPAIR);
798 f2fs_unlock_op(F2FS_I_SB(inode));
0abd675e 799 return err;
af033b2a
CY		 800 }
801 /*
802 * update uid/gid under lock_op(), so that dquot and inode can
803 * be updated atomically.
804 */
805 if (attr->ia_valid & ATTR_UID)
806 inode->i_uid = attr->ia_uid;
807 if (attr->ia_valid & ATTR_GID)
808 inode->i_gid = attr->ia_gid;
809 f2fs_mark_inode_dirty_sync(inode, true);
810 f2fs_unlock_op(F2FS_I_SB(inode));
0abd675e
CY		 811 }
812
09db6a2e 813 if (attr->ia_valid & ATTR_SIZE) {
a33c1502
CY		 814 bool to_smaller = (attr->ia_size <= i_size_read(inode));
815
a33c1502 816 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 817 down_write(&F2FS_I(inode)->i_mmap_sem);
a33c1502
CY		 818
819 truncate_setsize(inode, attr->ia_size);
820
821 if (to_smaller)
9a449e9c 822 err = f2fs_truncate(inode);
a33c1502
CY		 823 /*
824 * do not trim all blocks after i_size if target size is
825 * larger than i_size.
826 */
a33c1502 827 up_write(&F2FS_I(inode)->i_mmap_sem);
6f8d4455 828 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
a33c1502
CY		 829
830 if (err)
831 return err;
0cab80ee 832
a33c1502 833 if (!to_smaller) {
0cab80ee 834 /* should convert inline inode here */
b9d777b8 835 if (!f2fs_may_inline_data(inode)) {
0cab80ee
CY		 836 err = f2fs_convert_inline_inode(inode);
837 if (err)
838 return err;
839 }
078cd827 840 inode->i_mtime = inode->i_ctime = current_time(inode);
09db6a2e 841 }
c0ed4405 842
a0d00fad
CY		 843 down_write(&F2FS_I(inode)->i_sem);
844 F2FS_I(inode)->last_disk_size = i_size_read(inode);
845 up_write(&F2FS_I(inode)->i_sem);
846
c0ed4405 847 size_changed = true;
fbfa2cc5
JK		 848 }
849
850 __setattr_copy(inode, attr);
851
852 if (attr->ia_valid & ATTR_MODE) {
4d57b86d 853 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
91942321
JK		 854 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
855 inode->i_mode = F2FS_I(inode)->i_acl_mode;
856 clear_inode_flag(inode, FI_ACL_MODE);
fbfa2cc5
JK		 857 }
858 }
859
c0ed4405
YH		 860 /* file size may changed here */
861 f2fs_mark_inode_dirty_sync(inode, size_changed);
15d04354
JK		 862
863 /* inode change will produce dirty node pages flushed by checkpoint */
864 f2fs_balance_fs(F2FS_I_SB(inode), true);
865
fbfa2cc5
JK		 866 return err;
867}
868
869const struct inode_operations f2fs_file_inode_operations = {
870 .getattr = f2fs_getattr,
871 .setattr = f2fs_setattr,
872 .get_acl = f2fs_get_acl,
a6dda0e6 873 .set_acl = f2fs_set_acl,
fbfa2cc5 874#ifdef CONFIG_F2FS_FS_XATTR
fbfa2cc5 875 .listxattr = f2fs_listxattr,
fbfa2cc5 876#endif
9ab70134 877 .fiemap = f2fs_fiemap,
fbfa2cc5
JK		 878};
879
6394328a 880static int fill_zero(struct inode *inode, pgoff_t index,
fbfa2cc5
JK		 881 loff_t start, loff_t len)
882{
4081363f 883 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
fbfa2cc5
JK		 884 struct page *page;
885
886 if (!len)
6394328a 887 return 0;
fbfa2cc5 888
2c4db1a6 889 f2fs_balance_fs(sbi, true);
bd43df02 890
e479556b 891 f2fs_lock_op(sbi);
4d57b86d 892 page = f2fs_get_new_data_page(inode, NULL, index, false);
e479556b 893 f2fs_unlock_op(sbi);
fbfa2cc5 894
6394328a
CY		 895 if (IS_ERR(page))
896 return PTR_ERR(page);
897
fec1d657 898 f2fs_wait_on_page_writeback(page, DATA, true);
6394328a
CY		 899 zero_user(page, start, len);
900 set_page_dirty(page);
901 f2fs_put_page(page, 1);
902 return 0;
fbfa2cc5
JK		 903}
904
4d57b86d 905int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
fbfa2cc5 906{
fbfa2cc5
JK		 907 int err;
908
ea58711e 909 while (pg_start < pg_end) {
fbfa2cc5 910 struct dnode_of_data dn;
ea58711e 911 pgoff_t end_offset, count;
9eaeba70 912
fbfa2cc5 913 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 914 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
fbfa2cc5 915 if (err) {
ea58711e 916 if (err == -ENOENT) {
4d57b86d
CY		 917 pg_start = f2fs_get_next_page_offset(&dn,
918 pg_start);
fbfa2cc5 919 continue;
ea58711e 920 }
fbfa2cc5
JK		 921 return err;
922 }
923
81ca7350 924 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
ea58711e
CY		 925 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
926
927 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
928
4d57b86d 929 f2fs_truncate_data_blocks_range(&dn, count);
fbfa2cc5 930 f2fs_put_dnode(&dn);
ea58711e
CY		 931
932 pg_start += count;
fbfa2cc5
JK		 933 }
934 return 0;
935}
936
a66c7b2f 937static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
fbfa2cc5
JK		 938{
939 pgoff_t pg_start, pg_end;
940 loff_t off_start, off_end;
b9d777b8 941 int ret;
fbfa2cc5 942
b9d777b8
JK		 943 ret = f2fs_convert_inline_inode(inode);
944 if (ret)
945 return ret;
9ffe0fb5 946
09cbfeaf
KS		 947 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
948 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
fbfa2cc5 949
09cbfeaf
KS		 950 off_start = offset & (PAGE_SIZE - 1);
951 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5
JK		 952
953 if (pg_start == pg_end) {
6394328a 954 ret = fill_zero(inode, pg_start, off_start,
fbfa2cc5 955 off_end - off_start);
6394328a
CY		 956 if (ret)
957 return ret;
fbfa2cc5 958 } else {
6394328a
CY		 959 if (off_start) {
960 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 961 PAGE_SIZE - off_start);
6394328a
CY		 962 if (ret)
963 return ret;
964 }
965 if (off_end) {
966 ret = fill_zero(inode, pg_end, 0, off_end);
967 if (ret)
968 return ret;
969 }
fbfa2cc5
JK		 970
971 if (pg_start < pg_end) {
972 struct address_space *mapping = inode->i_mapping;
973 loff_t blk_start, blk_end;
4081363f 974 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127a3d4 975
2c4db1a6 976 f2fs_balance_fs(sbi, true);
fbfa2cc5 977
09cbfeaf
KS		 978 blk_start = (loff_t)pg_start << PAGE_SHIFT;
979 blk_end = (loff_t)pg_end << PAGE_SHIFT;
a33c1502 980
a33c1502 981 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 982 down_write(&F2FS_I(inode)->i_mmap_sem);
a33c1502 983
fbfa2cc5
JK		 984 truncate_inode_pages_range(mapping, blk_start,
985 blk_end - 1);
39936837 986
e479556b 987 f2fs_lock_op(sbi);
4d57b86d 988 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
e479556b 989 f2fs_unlock_op(sbi);
a33c1502 990
5a3a2d83 991 up_write(&F2FS_I(inode)->i_mmap_sem);
6f8d4455 992 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
fbfa2cc5
JK		 993 }
994 }
995
fbfa2cc5
JK		 996 return ret;
997}
998
0a2aa8fb
JK		 999static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1000 int *do_replace, pgoff_t off, pgoff_t len)
b4ace337
CY		 1001{
1002 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1003 struct dnode_of_data dn;
0a2aa8fb 1004 int ret, done, i;
ecbaa406 1005
0a2aa8fb 1006next_dnode:
6e2c64ad 1007 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1008 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
6e2c64ad
JK		 1009 if (ret && ret != -ENOENT) {
1010 return ret;
1011 } else if (ret == -ENOENT) {
0a2aa8fb
JK		 1012 if (dn.max_level == 0)
1013 return -ENOENT;
1014 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
1015 blkaddr += done;
1016 do_replace += done;
1017 goto next;
1018 }
1019
1020 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1021 dn.ofs_in_node, len);
1022 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
7a2af766
CY		 1023 *blkaddr = datablock_addr(dn.inode,
1024 dn.node_page, dn.ofs_in_node);
4d57b86d 1025 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
0a2aa8fb
JK		 1026
1027 if (test_opt(sbi, LFS)) {
1028 f2fs_put_dnode(&dn);
1029 return -ENOTSUPP;
1030 }
1031
6e2c64ad 1032 /* do not invalidate this block address */
f28b3434 1033 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
0a2aa8fb 1034 *do_replace = 1;
b4ace337 1035 }
6e2c64ad 1036 }
0a2aa8fb
JK		 1037 f2fs_put_dnode(&dn);
1038next:
1039 len -= done;
1040 off += done;
1041 if (len)
1042 goto next_dnode;
1043 return 0;
1044}
b4ace337 1045
0a2aa8fb
JK		 1046static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1047 int *do_replace, pgoff_t off, int len)
1048{
1049 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1050 struct dnode_of_data dn;
1051 int ret, i;
b4ace337 1052
0a2aa8fb
JK		 1053 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1054 if (*do_replace == 0)
1055 continue;
b4ace337 1056
0a2aa8fb 1057 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1058 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
0a2aa8fb
JK		 1059 if (ret) {
1060 dec_valid_block_count(sbi, inode, 1);
4d57b86d 1061 f2fs_invalidate_blocks(sbi, *blkaddr);
0a2aa8fb
JK		 1062 } else {
1063 f2fs_update_data_blkaddr(&dn, *blkaddr);
36abef4e 1064 }
0a2aa8fb
JK		 1065 f2fs_put_dnode(&dn);
1066 }
1067 return 0;
1068}
1069
1070static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1071 block_t *blkaddr, int *do_replace,
1072 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1073{
1074 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1075 pgoff_t i = 0;
1076 int ret;
36abef4e 1077
0a2aa8fb
JK		 1078 while (i < len) {
1079 if (blkaddr[i] == NULL_ADDR && !full) {
1080 i++;
1081 continue;
6e2c64ad 1082 }
b4ace337 1083
0a2aa8fb
JK		 1084 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1085 struct dnode_of_data dn;
1086 struct node_info ni;
1087 size_t new_size;
1088 pgoff_t ilen;
b4ace337 1089
0a2aa8fb 1090 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
4d57b86d 1091 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
0a2aa8fb
JK		 1092 if (ret)
1093 return ret;
b4ace337 1094
7735730d
CY		 1095 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1096 if (ret) {
1097 f2fs_put_dnode(&dn);
1098 return ret;
1099 }
1100
0a2aa8fb
JK		 1101 ilen = min((pgoff_t)
1102 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1103 dn.ofs_in_node, len - i);
1104 do {
7a2af766
CY		 1105 dn.data_blkaddr = datablock_addr(dn.inode,
1106 dn.node_page, dn.ofs_in_node);
4d57b86d 1107 f2fs_truncate_data_blocks_range(&dn, 1);
0a2aa8fb
JK		 1108
1109 if (do_replace[i]) {
1110 f2fs_i_blocks_write(src_inode,
0abd675e 1111 1, false, false);
0a2aa8fb 1112 f2fs_i_blocks_write(dst_inode,
0abd675e 1113 1, true, false);
0a2aa8fb
JK		 1114 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1115 blkaddr[i], ni.version, true, false);
1116
1117 do_replace[i] = 0;
1118 }
1119 dn.ofs_in_node++;
1120 i++;
1121 new_size = (dst + i) << PAGE_SHIFT;
1122 if (dst_inode->i_size < new_size)
1123 f2fs_i_size_write(dst_inode, new_size);
e87f7329 1124 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
6e2c64ad 1125
0a2aa8fb
JK		 1126 f2fs_put_dnode(&dn);
1127 } else {
1128 struct page *psrc, *pdst;
1129
4d57b86d
CY		 1130 psrc = f2fs_get_lock_data_page(src_inode,
1131 src + i, true);
0a2aa8fb
JK		 1132 if (IS_ERR(psrc))
1133 return PTR_ERR(psrc);
4d57b86d 1134 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
0a2aa8fb
JK		 1135 true);
1136 if (IS_ERR(pdst)) {
1137 f2fs_put_page(psrc, 1);
1138 return PTR_ERR(pdst);
1139 }
1140 f2fs_copy_page(psrc, pdst);
1141 set_page_dirty(pdst);
1142 f2fs_put_page(pdst, 1);
6e2c64ad 1143 f2fs_put_page(psrc, 1);
b4ace337 1144
4d57b86d
CY		 1145 ret = f2fs_truncate_hole(src_inode,
1146 src + i, src + i + 1);
0a2aa8fb
JK		 1147 if (ret)
1148 return ret;
1149 i++;
1150 }
6e2c64ad
JK		 1151 }
1152 return 0;
0a2aa8fb 1153}
b4ace337 1154
0a2aa8fb
JK		 1155static int __exchange_data_block(struct inode *src_inode,
1156 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
363cad7f 1157 pgoff_t len, bool full)
0a2aa8fb
JK		 1158{
1159 block_t *src_blkaddr;
1160 int *do_replace;
363cad7f 1161 pgoff_t olen;
0a2aa8fb
JK		 1162 int ret;
1163
363cad7f
JK		 1164 while (len) {
1165 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
0a2aa8fb 1166
628b3d14 1167 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
9d2a789c
KC		 1168 array_size(olen, sizeof(block_t)),
1169 GFP_KERNEL);
363cad7f
JK		 1170 if (!src_blkaddr)
1171 return -ENOMEM;
0a2aa8fb 1172
628b3d14 1173 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
9d2a789c
KC		 1174 array_size(olen, sizeof(int)),
1175 GFP_KERNEL);
363cad7f
JK		 1176 if (!do_replace) {
1177 kvfree(src_blkaddr);
1178 return -ENOMEM;
1179 }
0a2aa8fb 1180
363cad7f
JK		 1181 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1182 do_replace, src, olen);
1183 if (ret)
1184 goto roll_back;
0a2aa8fb 1185
363cad7f
JK		 1186 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1187 do_replace, src, dst, olen, full);
1188 if (ret)
1189 goto roll_back;
1190
1191 src += olen;
1192 dst += olen;
1193 len -= olen;
1194
1195 kvfree(src_blkaddr);
1196 kvfree(do_replace);
1197 }
0a2aa8fb
JK		 1198 return 0;
1199
1200roll_back:
9fd62605 1201 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
0a2aa8fb
JK		 1202 kvfree(src_blkaddr);
1203 kvfree(do_replace);
6e2c64ad
JK		 1204 return ret;
1205}
b4ace337 1206
6f8d4455 1207static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
6e2c64ad
JK		 1208{
1209 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1210 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
6f8d4455
JK		 1211 pgoff_t start = offset >> PAGE_SHIFT;
1212 pgoff_t end = (offset + len) >> PAGE_SHIFT;
0a2aa8fb 1213 int ret;
6e2c64ad 1214
0a2aa8fb 1215 f2fs_balance_fs(sbi, true);
5f281fab 1216
6f8d4455
JK		 1217 /* avoid gc operation during block exchange */
1218 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1219 down_write(&F2FS_I(inode)->i_mmap_sem);
5f281fab 1220
6f8d4455
JK		 1221 f2fs_lock_op(sbi);
1222 f2fs_drop_extent_tree(inode);
1223 truncate_pagecache(inode, offset);
0a2aa8fb
JK		 1224 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1225 f2fs_unlock_op(sbi);
6f8d4455
JK		 1226
1227 up_write(&F2FS_I(inode)->i_mmap_sem);
1228 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
b4ace337
CY		 1229 return ret;
1230}
1231
1232static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1233{
b4ace337
CY		 1234 loff_t new_size;
1235 int ret;
1236
b4ace337
CY		 1237 if (offset + len >= i_size_read(inode))
1238 return -EINVAL;
1239
1240 /* collapse range should be aligned to block size of f2fs. */
1241 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1242 return -EINVAL;
1243
b9d777b8
JK		 1244 ret = f2fs_convert_inline_inode(inode);
1245 if (ret)
1246 return ret;
97a7b2c2 1247
b4ace337
CY		 1248 /* write out all dirty pages from offset */
1249 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1250 if (ret)
6f8d4455 1251 return ret;
b4ace337 1252
6f8d4455 1253 ret = f2fs_do_collapse(inode, offset, len);
b4ace337 1254 if (ret)
6f8d4455 1255 return ret;
b4ace337 1256
6e2c64ad 1257 /* write out all moved pages, if possible */
6f8d4455 1258 down_write(&F2FS_I(inode)->i_mmap_sem);
6e2c64ad
JK		 1259 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1260 truncate_pagecache(inode, offset);
1261
b4ace337 1262 new_size = i_size_read(inode) - len;
6e2c64ad 1263 truncate_pagecache(inode, new_size);
b4ace337 1264
af033b2a 1265 ret = f2fs_truncate_blocks(inode, new_size, true, false);
6f8d4455 1266 up_write(&F2FS_I(inode)->i_mmap_sem);
b4ace337 1267 if (!ret)
fc9581c8 1268 f2fs_i_size_write(inode, new_size);
b4ace337
CY		 1269 return ret;
1270}
1271
6e961949
CY		 1272static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1273 pgoff_t end)
1274{
1275 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1276 pgoff_t index = start;
1277 unsigned int ofs_in_node = dn->ofs_in_node;
1278 blkcnt_t count = 0;
1279 int ret;
1280
1281 for (; index < end; index++, dn->ofs_in_node++) {
7a2af766
CY		 1282 if (datablock_addr(dn->inode, dn->node_page,
1283 dn->ofs_in_node) == NULL_ADDR)
6e961949
CY		 1284 count++;
1285 }
1286
1287 dn->ofs_in_node = ofs_in_node;
4d57b86d 1288 ret = f2fs_reserve_new_blocks(dn, count);
6e961949
CY		 1289 if (ret)
1290 return ret;
1291
1292 dn->ofs_in_node = ofs_in_node;
1293 for (index = start; index < end; index++, dn->ofs_in_node++) {
7a2af766
CY		 1294 dn->data_blkaddr = datablock_addr(dn->inode,
1295 dn->node_page, dn->ofs_in_node);
6e961949 1296 /*
4d57b86d 1297 * f2fs_reserve_new_blocks will not guarantee entire block
6e961949
CY		 1298 * allocation.
1299 */
1300 if (dn->data_blkaddr == NULL_ADDR) {
1301 ret = -ENOSPC;
1302 break;
1303 }
1304 if (dn->data_blkaddr != NEW_ADDR) {
4d57b86d 1305 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
6e961949 1306 dn->data_blkaddr = NEW_ADDR;
4d57b86d 1307 f2fs_set_data_blkaddr(dn);
6e961949
CY		 1308 }
1309 }
1310
1311 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1312
1313 return ret;
1314}
1315
75cd4e09
CY		 1316static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1317 int mode)
1318{
1319 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1320 struct address_space *mapping = inode->i_mapping;
1321 pgoff_t index, pg_start, pg_end;
1322 loff_t new_size = i_size_read(inode);
1323 loff_t off_start, off_end;
1324 int ret = 0;
1325
75cd4e09
CY		 1326 ret = inode_newsize_ok(inode, (len + offset));
1327 if (ret)
1328 return ret;
1329
b9d777b8
JK		 1330 ret = f2fs_convert_inline_inode(inode);
1331 if (ret)
1332 return ret;
75cd4e09
CY		 1333
1334 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1335 if (ret)
6f8d4455 1336 return ret;
75cd4e09 1337
09cbfeaf
KS		 1338 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1339 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
75cd4e09 1340
09cbfeaf
KS		 1341 off_start = offset & (PAGE_SIZE - 1);
1342 off_end = (offset + len) & (PAGE_SIZE - 1);
75cd4e09
CY		 1343
1344 if (pg_start == pg_end) {
6394328a
CY		 1345 ret = fill_zero(inode, pg_start, off_start,
1346 off_end - off_start);
1347 if (ret)
6f8d4455 1348 return ret;
6394328a 1349
75cd4e09
CY		 1350 new_size = max_t(loff_t, new_size, offset + len);
1351 } else {
1352 if (off_start) {
6394328a 1353 ret = fill_zero(inode, pg_start++, off_start,
09cbfeaf 1354 PAGE_SIZE - off_start);
6394328a 1355 if (ret)
6f8d4455 1356 return ret;
6394328a 1357
75cd4e09 1358 new_size = max_t(loff_t, new_size,
09cbfeaf 1359 (loff_t)pg_start << PAGE_SHIFT);
75cd4e09
CY		 1360 }
1361
6e961949 1362 for (index = pg_start; index < pg_end;) {
75cd4e09 1363 struct dnode_of_data dn;
6e961949
CY		 1364 unsigned int end_offset;
1365 pgoff_t end;
75cd4e09 1366
c7079853 1367 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
6f8d4455 1368 down_write(&F2FS_I(inode)->i_mmap_sem);
c7079853
CY		 1369
1370 truncate_pagecache_range(inode,
1371 (loff_t)index << PAGE_SHIFT,
1372 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1373
75cd4e09
CY		 1374 f2fs_lock_op(sbi);
1375
6e961949 1376 set_new_dnode(&dn, inode, NULL, NULL, 0);
4d57b86d 1377 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
75cd4e09
CY		 1378 if (ret) {
1379 f2fs_unlock_op(sbi);
6f8d4455 1380 up_write(&F2FS_I(inode)->i_mmap_sem);
c7079853 1381 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
75cd4e09
CY		 1382 goto out;
1383 }
1384
6e961949
CY		 1385 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1386 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1387
1388 ret = f2fs_do_zero_range(&dn, index, end);
75cd4e09 1389 f2fs_put_dnode(&dn);
c7079853 1390
75cd4e09 1391 f2fs_unlock_op(sbi);
6f8d4455 1392 up_write(&F2FS_I(inode)->i_mmap_sem);
c7079853 1393 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
9434fcde
CY		 1394
1395 f2fs_balance_fs(sbi, dn.node_changed);
1396
6e961949
CY		 1397 if (ret)
1398 goto out;
75cd4e09 1399
6e961949 1400 index = end;
75cd4e09 1401 new_size = max_t(loff_t, new_size,
6e961949 1402 (loff_t)index << PAGE_SHIFT);
75cd4e09
CY		 1403 }
1404
1405 if (off_end) {
6394328a
CY		 1406 ret = fill_zero(inode, pg_end, 0, off_end);
1407 if (ret)
1408 goto out;
1409
75cd4e09
CY		 1410 new_size = max_t(loff_t, new_size, offset + len);
1411 }
1412 }
1413
1414out:
17cd07ae
CY		 1415 if (new_size > i_size_read(inode)) {
1416 if (mode & FALLOC_FL_KEEP_SIZE)
1417 file_set_keep_isize(inode);
1418 else
1419 f2fs_i_size_write(inode, new_size);
1420 }
75cd4e09
CY		 1421 return ret;
1422}
1423
f62185d0
CY		 1424static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1425{
1426 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
0a2aa8fb 1427 pgoff_t nr, pg_start, pg_end, delta, idx;
f62185d0 1428 loff_t new_size;
6e2c64ad 1429 int ret = 0;
f62185d0 1430
f62185d0 1431 new_size = i_size_read(inode) + len;
46e82fb1
KM		 1432 ret = inode_newsize_ok(inode, new_size);
1433 if (ret)
1434 return ret;
f62185d0
CY		 1435
1436 if (offset >= i_size_read(inode))
1437 return -EINVAL;
1438
1439 /* insert range should be aligned to block size of f2fs. */
1440 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1441 return -EINVAL;
1442
b9d777b8
JK		 1443 ret = f2fs_convert_inline_inode(inode);
1444 if (ret)
1445 return ret;
97a7b2c2 1446
2c4db1a6 1447 f2fs_balance_fs(sbi, true);
2a340760 1448
5a3a2d83 1449 down_write(&F2FS_I(inode)->i_mmap_sem);
af033b2a 1450 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true, false);
6f8d4455 1451 up_write(&F2FS_I(inode)->i_mmap_sem);
f62185d0 1452 if (ret)
6f8d4455 1453 return ret;
f62185d0
CY		 1454
1455 /* write out all dirty pages from offset */
1456 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1457 if (ret)
6f8d4455 1458 return ret;
f62185d0 1459
09cbfeaf
KS		 1460 pg_start = offset >> PAGE_SHIFT;
1461 pg_end = (offset + len) >> PAGE_SHIFT;
f62185d0 1462 delta = pg_end - pg_start;
0a2aa8fb
JK		 1463 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1464
6f8d4455
JK		 1465 /* avoid gc operation during block exchange */
1466 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1467 down_write(&F2FS_I(inode)->i_mmap_sem);
1468 truncate_pagecache(inode, offset);
1469
0a2aa8fb
JK		 1470 while (!ret && idx > pg_start) {
1471 nr = idx - pg_start;
1472 if (nr > delta)
1473 nr = delta;
1474 idx -= nr;
f62185d0 1475
f62185d0 1476 f2fs_lock_op(sbi);
5f281fab
JK		 1477 f2fs_drop_extent_tree(inode);
1478
0a2aa8fb
JK		 1479 ret = __exchange_data_block(inode, inode, idx,
1480 idx + delta, nr, false);
f62185d0
CY		 1481 f2fs_unlock_op(sbi);
1482 }
6f8d4455
JK		 1483 up_write(&F2FS_I(inode)->i_mmap_sem);
1484 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
f62185d0 1485
6e2c64ad 1486 /* write out all moved pages, if possible */
6f8d4455 1487 down_write(&F2FS_I(inode)->i_mmap_sem);
6e2c64ad
JK		 1488 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1489 truncate_pagecache(inode, offset);
6f8d4455 1490 up_write(&F2FS_I(inode)->i_mmap_sem);
6e2c64ad
JK		 1491
1492 if (!ret)
fc9581c8 1493 f2fs_i_size_write(inode, new_size);
f62185d0
CY		 1494 return ret;
1495}
1496
fbfa2cc5
JK		 1497static int expand_inode_data(struct inode *inode, loff_t offset,
1498 loff_t len, int mode)
1499{
4081363f 1500 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d5097be5 1501 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
f9d6d059
CY		 1502 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1503 .m_may_create = true };
e12dd7bd 1504 pgoff_t pg_end;
fbfa2cc5 1505 loff_t new_size = i_size_read(inode);
e12dd7bd 1506 loff_t off_end;
a7de6086 1507 int err;
fbfa2cc5 1508
a7de6086
JK		 1509 err = inode_newsize_ok(inode, (len + offset));
1510 if (err)
1511 return err;
fbfa2cc5 1512
a7de6086
JK		 1513 err = f2fs_convert_inline_inode(inode);
1514 if (err)
1515 return err;
9e09fc85 1516
2c4db1a6 1517 f2fs_balance_fs(sbi, true);
2a340760 1518
e12dd7bd 1519 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
09cbfeaf 1520 off_end = (offset + len) & (PAGE_SIZE - 1);
fbfa2cc5 1521
e12dd7bd
JK		 1522 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1523 map.m_len = pg_end - map.m_lblk;
1524 if (off_end)
1525 map.m_len++;
ead43275 1526
a7de6086
JK		 1527 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1528 if (err) {
e12dd7bd 1529 pgoff_t last_off;
fbfa2cc5 1530
e12dd7bd 1531 if (!map.m_len)
a7de6086 1532 return err;
98397ff3 1533
e12dd7bd
JK		 1534 last_off = map.m_lblk + map.m_len - 1;
1535
1536 /* update new size to the failed position */
1061fd48 1537 new_size = (last_off == pg_end) ? offset + len :
e12dd7bd
JK		 1538 (loff_t)(last_off + 1) << PAGE_SHIFT;
1539 } else {
1540 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
fbfa2cc5
JK		 1541 }
1542
e8ed90a6
CY		 1543 if (new_size > i_size_read(inode)) {
1544 if (mode & FALLOC_FL_KEEP_SIZE)
1545 file_set_keep_isize(inode);
1546 else
1547 f2fs_i_size_write(inode, new_size);
1548 }
fbfa2cc5 1549
a7de6086 1550 return err;
fbfa2cc5
JK		 1551}
1552
1553static long f2fs_fallocate(struct file *file, int mode,
1554 loff_t offset, loff_t len)
1555{
6131ffaa 1556 struct inode *inode = file_inode(file);
587c0a42 1557 long ret = 0;
fbfa2cc5 1558
1f227a3e
JK		 1559 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1560 return -EIO;
1561
c998012b
CY		 1562 /* f2fs only support ->fallocate for regular file */
1563 if (!S_ISREG(inode->i_mode))
1564 return -EINVAL;
1565
f62185d0
CY		 1566 if (f2fs_encrypted_inode(inode) &&
1567 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
fcc85a4d
JK		 1568 return -EOPNOTSUPP;
1569
b4ace337 1570 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
f62185d0
CY		 1571 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1572 FALLOC_FL_INSERT_RANGE))
fbfa2cc5
JK		 1573 return -EOPNOTSUPP;
1574
5955102c 1575 inode_lock(inode);
3375f696 1576
587c0a42
TY		 1577 if (mode & FALLOC_FL_PUNCH_HOLE) {
1578 if (offset >= inode->i_size)
1579 goto out;
1580
a66c7b2f 1581 ret = punch_hole(inode, offset, len);
b4ace337
CY		 1582 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1583 ret = f2fs_collapse_range(inode, offset, len);
75cd4e09
CY		 1584 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1585 ret = f2fs_zero_range(inode, offset, len, mode);
f62185d0
CY		 1586 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1587 ret = f2fs_insert_range(inode, offset, len);
b4ace337 1588 } else {
fbfa2cc5 1589 ret = expand_inode_data(inode, offset, len, mode);
b4ace337 1590 }
fbfa2cc5 1591
3af60a49 1592 if (!ret) {
078cd827 1593 inode->i_mtime = inode->i_ctime = current_time(inode);
7c45729a 1594 f2fs_mark_inode_dirty_sync(inode, false);
d0239e1b 1595 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3af60a49 1596 }
3375f696 1597
587c0a42 1598out:
5955102c 1599 inode_unlock(inode);
3375f696 1600
c01e2853 1601 trace_f2fs_fallocate(inode, mode, offset, len, ret);
fbfa2cc5
JK		 1602 return ret;
1603}
1604
1e84371f
JK		 1605static int f2fs_release_file(struct inode *inode, struct file *filp)
1606{
de5307e4
JK		 1607 /*
1608 * f2fs_relase_file is called at every close calls. So we should
1609 * not drop any inmemory pages by close called by other process.
1610 */
1611 if (!(filp->f_mode & FMODE_WRITE) ||
1612 atomic_read(&inode->i_writecount) != 1)
1613 return 0;
1614
1e84371f
JK		 1615 /* some remained atomic pages should discarded */
1616 if (f2fs_is_atomic_file(inode))
4d57b86d 1617 f2fs_drop_inmem_pages(inode);
1e84371f 1618 if (f2fs_is_volatile_file(inode)) {
91942321 1619 set_inode_flag(inode, FI_DROP_CACHE);
1e84371f 1620 filemap_fdatawrite(inode->i_mapping);
91942321 1621 clear_inode_flag(inode, FI_DROP_CACHE);
dfa74280
CY		 1622 clear_inode_flag(inode, FI_VOLATILE_FILE);
1623 stat_dec_volatile_write(inode);
1e84371f
JK		 1624 }
1625 return 0;
1626}
1627
7a10f017 1628static int f2fs_file_flush(struct file *file, fl_owner_t id)
fbfa2cc5 1629{
7a10f017
JK		 1630 struct inode *inode = file_inode(file);
1631
1632 /*
1633 * If the process doing a transaction is crashed, we should do
1634 * roll-back. Otherwise, other reader/write can see corrupted database
1635 * until all the writers close its file. Since this should be done
1636 * before dropping file lock, it needs to do in ->flush.
1637 */
1638 if (f2fs_is_atomic_file(inode) &&
1639 F2FS_I(inode)->inmem_task == current)
4d57b86d 1640 f2fs_drop_inmem_pages(inode);
7a10f017 1641 return 0;
fbfa2cc5
JK		 1642}
1643
52656e6c 1644static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
fbfa2cc5 1645{
6131ffaa 1646 struct inode *inode = file_inode(filp);
fbfa2cc5 1647 struct f2fs_inode_info *fi = F2FS_I(inode);
76481158
CY		 1648 unsigned int flags = fi->i_flags;
1649
5b72d5e0 1650 if (f2fs_encrypted_inode(inode))
76481158
CY		 1651 flags |= F2FS_ENCRYPT_FL;
1652 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1653 flags |= F2FS_INLINE_DATA_FL;
1654
1655 flags &= F2FS_FL_USER_VISIBLE;
1656
52656e6c
JK		 1657 return put_user(flags, (int __user *)arg);
1658}
fbfa2cc5 1659
2c1d0305
CY		 1660static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1661{
1662 struct f2fs_inode_info *fi = F2FS_I(inode);
1663 unsigned int oldflags;
1664
1665 /* Is it quota file? Do not allow user to mess with it */
1666 if (IS_NOQUOTA(inode))
1667 return -EPERM;
1668
1669 flags = f2fs_mask_flags(inode->i_mode, flags);
1670
1671 oldflags = fi->i_flags;
1672
59c84408 1673 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
2c1d0305
CY		 1674 if (!capable(CAP_LINUX_IMMUTABLE))
1675 return -EPERM;
1676
c807a7cb
CY		 1677 flags = flags & F2FS_FL_USER_MODIFIABLE;
1678 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
2c1d0305
CY		 1679 fi->i_flags = flags;
1680
59c84408 1681 if (fi->i_flags & F2FS_PROJINHERIT_FL)
2c1d0305
CY		 1682 set_inode_flag(inode, FI_PROJ_INHERIT);
1683 else
1684 clear_inode_flag(inode, FI_PROJ_INHERIT);
1685
1686 inode->i_ctime = current_time(inode);
1687 f2fs_set_inode_flags(inode);
1688 f2fs_mark_inode_dirty_sync(inode, false);
1689 return 0;
1690}
1691
52656e6c
JK		 1692static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1693{
1694 struct inode *inode = file_inode(filp);
69494229 1695 unsigned int flags;
52656e6c 1696 int ret;
fbfa2cc5 1697
7fb17fe4
CY		 1698 if (!inode_owner_or_capable(inode))
1699 return -EACCES;
1700
1701 if (get_user(flags, (int __user *)arg))
1702 return -EFAULT;
1703
52656e6c
JK		 1704 ret = mnt_want_write_file(filp);
1705 if (ret)
1706 return ret;
fbfa2cc5 1707
5955102c 1708 inode_lock(inode);
fbfa2cc5 1709
2c1d0305 1710 ret = __f2fs_ioc_setflags(inode, flags);
fbfa2cc5 1711
a72d4b97 1712 inode_unlock(inode);
52656e6c
JK		 1713 mnt_drop_write_file(filp);
1714 return ret;
1715}
4b2fecc8 1716
d49f3e89
CY		 1717static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1718{
1719 struct inode *inode = file_inode(filp);
1720
1721 return put_user(inode->i_generation, (int __user *)arg);
1722}
1723
88b88a66
JK		 1724static int f2fs_ioc_start_atomic_write(struct file *filp)
1725{
1726 struct inode *inode = file_inode(filp);
f4c9c743 1727 int ret;
88b88a66
JK		 1728
1729 if (!inode_owner_or_capable(inode))
1730 return -EACCES;
1731
e811898c
JK		 1732 if (!S_ISREG(inode->i_mode))
1733 return -EINVAL;
1734
7fb17fe4
CY		 1735 ret = mnt_want_write_file(filp);
1736 if (ret)
1737 return ret;
1738
0fac558b
CY		 1739 inode_lock(inode);
1740
455e3a58
JK		 1741 if (f2fs_is_atomic_file(inode)) {
1742 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1743 ret = -EINVAL;
7fb17fe4 1744 goto out;
455e3a58 1745 }
88b88a66 1746
f4c9c743
CY		 1747 ret = f2fs_convert_inline_inode(inode);
1748 if (ret)
7fb17fe4 1749 goto out;
88b88a66 1750
6f8d4455
JK		 1751 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1752
c27753d6 1753 if (!get_dirty_pages(inode))
054afda9 1754 goto skip_flush;
c27753d6
JK		 1755
1756 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
204706c7 1757 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
c27753d6
JK		 1758 inode->i_ino, get_dirty_pages(inode));
1759 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
6f8d4455
JK		 1760 if (ret) {
1761 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
684ca7e5 1762 goto out;
6f8d4455 1763 }
054afda9 1764skip_flush:
054afda9 1765 set_inode_flag(inode, FI_ATOMIC_FILE);
2ef79ecb 1766 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
6f8d4455 1767 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
684ca7e5 1768
6f8d4455 1769 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7a10f017 1770 F2FS_I(inode)->inmem_task = current;
26a28a0c
JK		 1771 stat_inc_atomic_write(inode);
1772 stat_update_max_atomic_write(inode);
684ca7e5 1773out:
0fac558b 1774 inode_unlock(inode);
7fb17fe4 1775 mnt_drop_write_file(filp);
c27753d6 1776 return ret;
88b88a66
JK		 1777}
1778
1779static int f2fs_ioc_commit_atomic_write(struct file *filp)
1780{
1781 struct inode *inode = file_inode(filp);
1782 int ret;
1783
1784 if (!inode_owner_or_capable(inode))
1785 return -EACCES;
1786
1787 ret = mnt_want_write_file(filp);
1788 if (ret)
1789 return ret;
1790
6f8d4455 1791 f2fs_balance_fs(F2FS_I_SB(inode), true);
0fac558b 1792
6f8d4455 1793 inode_lock(inode);
1dc0f899 1794
b169c3c5
CY		 1795 if (f2fs_is_volatile_file(inode)) {
1796 ret = -EINVAL;
7fb17fe4 1797 goto err_out;
b169c3c5 1798 }
7fb17fe4 1799
6282adbf 1800 if (f2fs_is_atomic_file(inode)) {
4d57b86d 1801 ret = f2fs_commit_inmem_pages(inode);
5fe45743 1802 if (ret)
edb27dee 1803 goto err_out;
5fe45743 1804
26a28a0c 1805 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
5fe45743
CY		 1806 if (!ret) {
1807 clear_inode_flag(inode, FI_ATOMIC_FILE);
2ef79ecb 1808 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
5fe45743 1809 stat_dec_atomic_write(inode);
447135a8 1810 }
26a28a0c 1811 } else {
774e1b78 1812 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
6282adbf 1813 }
edb27dee 1814err_out:
2ef79ecb
CY		 1815 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1816 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1817 ret = -EINVAL;
1818 }
0fac558b 1819 inode_unlock(inode);
88b88a66
JK		 1820 mnt_drop_write_file(filp);
1821 return ret;
1822}
1823
02a1335f
JK		 1824static int f2fs_ioc_start_volatile_write(struct file *filp)
1825{
1826 struct inode *inode = file_inode(filp);
f4c9c743 1827 int ret;
02a1335f
JK		 1828
1829 if (!inode_owner_or_capable(inode))
1830 return -EACCES;
1831
8ff0971f
CY		 1832 if (!S_ISREG(inode->i_mode))
1833 return -EINVAL;
1834
7fb17fe4
CY		 1835 ret = mnt_want_write_file(filp);
1836 if (ret)
1837 return ret;
1838
0fac558b
CY		 1839 inode_lock(inode);
1840
1e84371f 1841 if (f2fs_is_volatile_file(inode))
7fb17fe4 1842 goto out;
1e84371f 1843
f4c9c743
CY		 1844 ret = f2fs_convert_inline_inode(inode);
1845 if (ret)
7fb17fe4 1846 goto out;
b3d208f9 1847
648d50ba
CY		 1848 stat_inc_volatile_write(inode);
1849 stat_update_max_volatile_write(inode);
1850
91942321 1851 set_inode_flag(inode, FI_VOLATILE_FILE);
d0239e1b 1852 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 1853out:
0fac558b 1854 inode_unlock(inode);
7fb17fe4
CY		 1855 mnt_drop_write_file(filp);
1856 return ret;
02a1335f
JK		 1857}
1858
1e84371f
JK		 1859static int f2fs_ioc_release_volatile_write(struct file *filp)
1860{
1861 struct inode *inode = file_inode(filp);
7fb17fe4 1862 int ret;
1e84371f
JK		 1863
1864 if (!inode_owner_or_capable(inode))
1865 return -EACCES;
1866
7fb17fe4
CY		 1867 ret = mnt_want_write_file(filp);
1868 if (ret)
1869 return ret;
1870
0fac558b
CY		 1871 inode_lock(inode);
1872
1e84371f 1873 if (!f2fs_is_volatile_file(inode))
7fb17fe4 1874 goto out;
1e84371f 1875
7fb17fe4
CY		 1876 if (!f2fs_is_first_block_written(inode)) {
1877 ret = truncate_partial_data_page(inode, 0, true);
1878 goto out;
1879 }
3c6c2beb 1880
7fb17fe4
CY		 1881 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1882out:
0fac558b 1883 inode_unlock(inode);
7fb17fe4
CY		 1884 mnt_drop_write_file(filp);
1885 return ret;
1e84371f
JK		 1886}
1887
1888static int f2fs_ioc_abort_volatile_write(struct file *filp)
1889{
1890 struct inode *inode = file_inode(filp);
1891 int ret;
1892
1893 if (!inode_owner_or_capable(inode))
1894 return -EACCES;
1895
1896 ret = mnt_want_write_file(filp);
1897 if (ret)
1898 return ret;
1899
0fac558b
CY		 1900 inode_lock(inode);
1901
26dc3d44 1902 if (f2fs_is_atomic_file(inode))
4d57b86d 1903 f2fs_drop_inmem_pages(inode);
732d5648 1904 if (f2fs_is_volatile_file(inode)) {
91942321 1905 clear_inode_flag(inode, FI_VOLATILE_FILE);
648d50ba 1906 stat_dec_volatile_write(inode);
608514de 1907 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
732d5648 1908 }
de6a8ec9 1909
455e3a58
JK		 1910 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1911
0fac558b
CY		 1912 inode_unlock(inode);
1913
1e84371f 1914 mnt_drop_write_file(filp);
d0239e1b 1915 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1e84371f
JK		 1916 return ret;
1917}
1918
1abff93d
JK		 1919static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1920{
1921 struct inode *inode = file_inode(filp);
1922 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1923 struct super_block *sb = sbi->sb;
1924 __u32 in;
2a96d8ad 1925 int ret = 0;
1abff93d
JK		 1926
1927 if (!capable(CAP_SYS_ADMIN))
1928 return -EPERM;
1929
1930 if (get_user(in, (__u32 __user *)arg))
1931 return -EFAULT;
1932
60b2b4ee
ST		 1933 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1934 ret = mnt_want_write_file(filp);
1935 if (ret)
1936 return ret;
1937 }
7fb17fe4 1938
1abff93d
JK		 1939 switch (in) {
1940 case F2FS_GOING_DOWN_FULLSYNC:
1941 sb = freeze_bdev(sb->s_bdev);
d027c484
CY		 1942 if (IS_ERR(sb)) {
1943 ret = PTR_ERR(sb);
1944 goto out;
1945 }
1946 if (sb) {
38f91ca8 1947 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 1948 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1abff93d
JK		 1949 thaw_bdev(sb->s_bdev, sb);
1950 }
1951 break;
1952 case F2FS_GOING_DOWN_METASYNC:
1953 /* do checkpoint only */
d027c484
CY		 1954 ret = f2fs_sync_fs(sb, 1);
1955 if (ret)
1956 goto out;
38f91ca8 1957 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 1958 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1abff93d
JK		 1959 break;
1960 case F2FS_GOING_DOWN_NOSYNC:
38f91ca8 1961 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 1962 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1abff93d 1963 break;
c912a829 1964 case F2FS_GOING_DOWN_METAFLUSH:
4d57b86d 1965 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
38f91ca8 1966 f2fs_stop_checkpoint(sbi, false);
83a3bfdb 1967 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
c912a829 1968 break;
1abff93d 1969 default:
7fb17fe4
CY		 1970 ret = -EINVAL;
1971 goto out;
1abff93d 1972 }
7950e9ac 1973
4d57b86d
CY		 1974 f2fs_stop_gc_thread(sbi);
1975 f2fs_stop_discard_thread(sbi);
7950e9ac 1976
4d57b86d 1977 f2fs_drop_discard_cmd(sbi);
7950e9ac
CY		 1978 clear_opt(sbi, DISCARD);
1979
d0239e1b 1980 f2fs_update_time(sbi, REQ_TIME);
7fb17fe4 1981out:
60b2b4ee
ST		 1982 if (in != F2FS_GOING_DOWN_FULLSYNC)
1983 mnt_drop_write_file(filp);
7fb17fe4 1984 return ret;
1abff93d
JK		 1985}
1986
52656e6c
JK		 1987static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1988{
1989 struct inode *inode = file_inode(filp);
1990 struct super_block *sb = inode->i_sb;
1991 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1992 struct fstrim_range range;
1993 int ret;
4b2fecc8 1994
52656e6c
JK		 1995 if (!capable(CAP_SYS_ADMIN))
1996 return -EPERM;
4b2fecc8 1997
7d20c8ab 1998 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
52656e6c 1999 return -EOPNOTSUPP;
4b2fecc8 2000
52656e6c
JK		 2001 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2002 sizeof(range)))
2003 return -EFAULT;
4b2fecc8 2004
7fb17fe4
CY		 2005 ret = mnt_want_write_file(filp);
2006 if (ret)
2007 return ret;
2008
52656e6c
JK		 2009 range.minlen = max((unsigned int)range.minlen,
2010 q->limits.discard_granularity);
2011 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
7fb17fe4 2012 mnt_drop_write_file(filp);
52656e6c
JK		 2013 if (ret < 0)
2014 return ret;
4b2fecc8 2015
52656e6c
JK		 2016 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2017 sizeof(range)))
2018 return -EFAULT;
d0239e1b 2019 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
52656e6c
JK		 2020 return 0;
2021}
2022
f424f664
JK		 2023static bool uuid_is_nonzero(__u8 u[16])
2024{
2025 int i;
2026
2027 for (i = 0; i < 16; i++)
2028 if (u[i])
2029 return true;
2030 return false;
2031}
2032
2033static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2034{
f424f664
JK		 2035 struct inode *inode = file_inode(filp);
2036
7beb01f7 2037 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
ead710b7
CY		 2038 return -EOPNOTSUPP;
2039
d0239e1b 2040 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
7fb17fe4 2041
db717d8e 2042 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
f424f664
JK		 2043}
2044
2045static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2046{
7beb01f7 2047 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
ead710b7 2048 return -EOPNOTSUPP;
db717d8e 2049 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
f424f664
JK		 2050}
2051
2052static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2053{
2054 struct inode *inode = file_inode(filp);
2055 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2056 int err;
2057
7beb01f7 2058 if (!f2fs_sb_has_encrypt(sbi))
f424f664
JK		 2059 return -EOPNOTSUPP;
2060
f424f664
JK		 2061 err = mnt_want_write_file(filp);
2062 if (err)
2063 return err;
2064
846ae671 2065 down_write(&sbi->sb_lock);
d0d3f1b3
CY		 2066
2067 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2068 goto got_it;
2069
f424f664
JK		 2070 /* update superblock with uuid */
2071 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2072
c5bda1c8 2073 err = f2fs_commit_super(sbi, false);
f424f664
JK		 2074 if (err) {
2075 /* undo new data */
2076 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
d0d3f1b3 2077 goto out_err;
f424f664
JK		 2078 }
2079got_it:
2080 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2081 16))
d0d3f1b3
CY		 2082 err = -EFAULT;
2083out_err:
846ae671 2084 up_write(&sbi->sb_lock);
d0d3f1b3
CY		 2085 mnt_drop_write_file(filp);
2086 return err;
f424f664
JK		 2087}
2088
c1c1b583
CY		 2089static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2090{
2091 struct inode *inode = file_inode(filp);
2092 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
d530d4d8 2093 __u32 sync;
7fb17fe4 2094 int ret;
c1c1b583
CY		 2095
2096 if (!capable(CAP_SYS_ADMIN))
2097 return -EPERM;
2098
d530d4d8 2099 if (get_user(sync, (__u32 __user *)arg))
c1c1b583
CY		 2100 return -EFAULT;
2101
d530d4d8
CY		 2102 if (f2fs_readonly(sbi->sb))
2103 return -EROFS;
c1c1b583 2104
7fb17fe4
CY		 2105 ret = mnt_want_write_file(filp);
2106 if (ret)
2107 return ret;
2108
d530d4d8 2109 if (!sync) {
7fb17fe4
CY		 2110 if (!mutex_trylock(&sbi->gc_mutex)) {
2111 ret = -EBUSY;
2112 goto out;
2113 }
d530d4d8
CY		 2114 } else {
2115 mutex_lock(&sbi->gc_mutex);
c1c1b583
CY		 2116 }
2117
e066b83c 2118 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
7fb17fe4
CY		 2119out:
2120 mnt_drop_write_file(filp);
2121 return ret;
c1c1b583
CY		 2122}
2123
34dc77ad
JK		 2124static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2125{
2126 struct inode *inode = file_inode(filp);
2127 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2128 struct f2fs_gc_range range;
2129 u64 end;
2130 int ret;
2131
2132 if (!capable(CAP_SYS_ADMIN))
2133 return -EPERM;
2134
2135 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2136 sizeof(range)))
2137 return -EFAULT;
2138
2139 if (f2fs_readonly(sbi->sb))
2140 return -EROFS;
2141
b82f6e34
YH		 2142 end = range.start + range.len;
2143 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2144 return -EINVAL;
2145 }
2146
34dc77ad
JK		 2147 ret = mnt_want_write_file(filp);
2148 if (ret)
2149 return ret;
2150
34dc77ad
JK		 2151do_more:
2152 if (!range.sync) {
2153 if (!mutex_trylock(&sbi->gc_mutex)) {
2154 ret = -EBUSY;
2155 goto out;
2156 }
2157 } else {
2158 mutex_lock(&sbi->gc_mutex);
2159 }
2160
2161 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
67b0e42b 2162 range.start += BLKS_PER_SEC(sbi);
34dc77ad
JK		 2163 if (range.start <= end)
2164 goto do_more;
2165out:
2166 mnt_drop_write_file(filp);
2167 return ret;
2168}
2169
059c0648 2170static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
456b88e4
CY		 2171{
2172 struct inode *inode = file_inode(filp);
2173 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
7fb17fe4 2174 int ret;
456b88e4
CY		 2175
2176 if (!capable(CAP_SYS_ADMIN))
2177 return -EPERM;
2178
2179 if (f2fs_readonly(sbi->sb))
2180 return -EROFS;
2181
4354994f
DR		 2182 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2183 f2fs_msg(sbi->sb, KERN_INFO,
2184 "Skipping Checkpoint. Checkpoints currently disabled.");
2185 return -EINVAL;
2186 }
2187
7fb17fe4
CY		 2188 ret = mnt_want_write_file(filp);
2189 if (ret)
2190 return ret;
2191
2192 ret = f2fs_sync_fs(sbi->sb, 1);
2193
2194 mnt_drop_write_file(filp);
2195 return ret;
456b88e4
CY		 2196}
2197
d323d005
CY		 2198static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2199 struct file *filp,
2200 struct f2fs_defragment *range)
2201{
2202 struct inode *inode = file_inode(filp);
f3d98e74 2203 struct f2fs_map_blocks map = { .m_next_extent = NULL,
f4f0b677
JZ		 2204 .m_seg_type = NO_CHECK_TYPE ,
2205 .m_may_create = false };
1061fd48 2206 struct extent_info ei = {0, 0, 0};
f3d98e74 2207 pgoff_t pg_start, pg_end, next_pgofs;
3519e3f9 2208 unsigned int blk_per_seg = sbi->blocks_per_seg;
d323d005 2209 unsigned int total = 0, sec_num;
d323d005
CY		 2210 block_t blk_end = 0;
2211 bool fragmented = false;
2212 int err;
2213
2214 /* if in-place-update policy is enabled, don't waste time here */
4d57b86d 2215 if (f2fs_should_update_inplace(inode, NULL))
d323d005
CY		 2216 return -EINVAL;
2217
09cbfeaf
KS		 2218 pg_start = range->start >> PAGE_SHIFT;
2219 pg_end = (range->start + range->len) >> PAGE_SHIFT;
d323d005 2220
2c4db1a6 2221 f2fs_balance_fs(sbi, true);
d323d005 2222
5955102c 2223 inode_lock(inode);
d323d005
CY		 2224
2225 /* writeback all dirty pages in the range */
2226 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
d8fe4f0e 2227 range->start + range->len - 1);
d323d005
CY		 2228 if (err)
2229 goto out;
2230
2231 /*
2232 * lookup mapping info in extent cache, skip defragmenting if physical
2233 * block addresses are continuous.
2234 */
2235 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2236 if (ei.fofs + ei.len >= pg_end)
2237 goto out;
2238 }
2239
2240 map.m_lblk = pg_start;
f3d98e74 2241 map.m_next_pgofs = &next_pgofs;
d323d005
CY		 2242
2243 /*
2244 * lookup mapping info in dnode page cache, skip defragmenting if all
2245 * physical block addresses are continuous even if there are hole(s)
2246 * in logical blocks.
2247 */
2248 while (map.m_lblk < pg_end) {
a1c1e9b7 2249 map.m_len = pg_end - map.m_lblk;
f2220c7f 2250 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2251 if (err)
2252 goto out;
2253
2254 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2255 map.m_lblk = next_pgofs;
d323d005
CY		 2256 continue;
2257 }
2258
25a912e5 2259 if (blk_end && blk_end != map.m_pblk)
d323d005 2260 fragmented = true;
25a912e5
CY		 2261
2262 /* record total count of block that we're going to move */
2263 total += map.m_len;
2264
d323d005
CY		 2265 blk_end = map.m_pblk + map.m_len;
2266
2267 map.m_lblk += map.m_len;
d323d005
CY		 2268 }
2269
2270 if (!fragmented)
2271 goto out;
2272
25a912e5 2273 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
d323d005
CY		 2274
2275 /*
2276 * make sure there are enough free section for LFS allocation, this can
2277 * avoid defragment running in SSR mode when free section are allocated
2278 * intensively
2279 */
7f3037a5 2280 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
d323d005
CY		 2281 err = -EAGAIN;
2282 goto out;
2283 }
2284
25a912e5
CY		 2285 map.m_lblk = pg_start;
2286 map.m_len = pg_end - pg_start;
2287 total = 0;
2288
d323d005
CY		 2289 while (map.m_lblk < pg_end) {
2290 pgoff_t idx;
2291 int cnt = 0;
2292
2293do_map:
a1c1e9b7 2294 map.m_len = pg_end - map.m_lblk;
f2220c7f 2295 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
d323d005
CY		 2296 if (err)
2297 goto clear_out;
2298
2299 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
f3d98e74 2300 map.m_lblk = next_pgofs;
d323d005
CY		 2301 continue;
2302 }
2303
91942321 2304 set_inode_flag(inode, FI_DO_DEFRAG);
d323d005
CY		 2305
2306 idx = map.m_lblk;
2307 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2308 struct page *page;
2309
4d57b86d 2310 page = f2fs_get_lock_data_page(inode, idx, true);
d323d005
CY		 2311 if (IS_ERR(page)) {
2312 err = PTR_ERR(page);
2313 goto clear_out;
2314 }
2315
2316 set_page_dirty(page);
2317 f2fs_put_page(page, 1);
2318
2319 idx++;
2320 cnt++;
2321 total++;
2322 }
2323
2324 map.m_lblk = idx;
d323d005
CY		 2325
2326 if (idx < pg_end && cnt < blk_per_seg)
2327 goto do_map;
2328
91942321 2329 clear_inode_flag(inode, FI_DO_DEFRAG);
d323d005
CY		 2330
2331 err = filemap_fdatawrite(inode->i_mapping);
2332 if (err)
2333 goto out;
2334 }
2335clear_out:
91942321 2336 clear_inode_flag(inode, FI_DO_DEFRAG);
d323d005 2337out:
5955102c 2338 inode_unlock(inode);
d323d005 2339 if (!err)
09cbfeaf 2340 range->len = (u64)total << PAGE_SHIFT;
d323d005
CY		 2341 return err;
2342}
2343
2344static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2345{
2346 struct inode *inode = file_inode(filp);
2347 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2348 struct f2fs_defragment range;
2349 int err;
2350
2351 if (!capable(CAP_SYS_ADMIN))
2352 return -EPERM;
2353
7eab0c0d 2354 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
d323d005
CY		 2355 return -EINVAL;
2356
d7563861
KM		 2357 if (f2fs_readonly(sbi->sb))
2358 return -EROFS;
d323d005
CY		 2359
2360 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
d7563861
KM		 2361 sizeof(range)))
2362 return -EFAULT;
d323d005
CY		 2363
2364 /* verify alignment of offset & size */
d7563861
KM		 2365 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2366 return -EINVAL;
d323d005 2367
1941d7bc 2368 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
d7563861
KM		 2369 sbi->max_file_blocks))
2370 return -EINVAL;
2371
2372 err = mnt_want_write_file(filp);
2373 if (err)
2374 return err;
1941d7bc 2375
d323d005 2376 err = f2fs_defragment_range(sbi, filp, &range);
d7563861
KM		 2377 mnt_drop_write_file(filp);
2378
d0239e1b 2379 f2fs_update_time(sbi, REQ_TIME);
d323d005 2380 if (err < 0)
d7563861 2381 return err;
d323d005
CY		 2382
2383 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2384 sizeof(range)))
d7563861
KM		 2385 return -EFAULT;
2386
2387 return 0;
d323d005
CY		 2388}
2389
4dd6f977
JK		 2390static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2391 struct file *file_out, loff_t pos_out, size_t len)
2392{
2393 struct inode *src = file_inode(file_in);
2394 struct inode *dst = file_inode(file_out);
2395 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2396 size_t olen = len, dst_max_i_size = 0;
2397 size_t dst_osize;
2398 int ret;
2399
2400 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2401 src->i_sb != dst->i_sb)
2402 return -EXDEV;
2403
2404 if (unlikely(f2fs_readonly(src->i_sb)))
2405 return -EROFS;
2406
fe8494bf
CY		 2407 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2408 return -EINVAL;
4dd6f977
JK		 2409
2410 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2411 return -EOPNOTSUPP;
2412
d95fd91c
FL		 2413 if (src == dst) {
2414 if (pos_in == pos_out)
2415 return 0;
2416 if (pos_out > pos_in && pos_out < pos_in + len)
2417 return -EINVAL;
2418 }
2419
4dd6f977 2420 inode_lock(src);
20a3d61d 2421 if (src != dst) {
bb06664a
CY		 2422 ret = -EBUSY;
2423 if (!inode_trylock(dst))
2424 goto out;
20a3d61d 2425 }
4dd6f977
JK		 2426
2427 ret = -EINVAL;
2428 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2429 goto out_unlock;
2430 if (len == 0)
2431 olen = len = src->i_size - pos_in;
2432 if (pos_in + len == src->i_size)
2433 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2434 if (len == 0) {
2435 ret = 0;
2436 goto out_unlock;
2437 }
2438
2439 dst_osize = dst->i_size;
2440 if (pos_out + olen > dst->i_size)
2441 dst_max_i_size = pos_out + olen;
2442
2443 /* verify the end result is block aligned */
2444 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2445 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2446 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2447 goto out_unlock;
2448
2449 ret = f2fs_convert_inline_inode(src);
2450 if (ret)
2451 goto out_unlock;
2452
2453 ret = f2fs_convert_inline_inode(dst);
2454 if (ret)
2455 goto out_unlock;
2456
2457 /* write out all dirty pages from offset */
2458 ret = filemap_write_and_wait_range(src->i_mapping,
2459 pos_in, pos_in + len);
2460 if (ret)
2461 goto out_unlock;
2462
2463 ret = filemap_write_and_wait_range(dst->i_mapping,
2464 pos_out, pos_out + len);
2465 if (ret)
2466 goto out_unlock;
2467
2468 f2fs_balance_fs(sbi, true);
6f8d4455
JK		 2469
2470 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2471 if (src != dst) {
2472 ret = -EBUSY;
2473 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2474 goto out_src;
2475 }
2476
4dd6f977 2477 f2fs_lock_op(sbi);
61e4da11
FL		 2478 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2479 pos_out >> F2FS_BLKSIZE_BITS,
2480 len >> F2FS_BLKSIZE_BITS, false);
4dd6f977
JK		 2481
2482 if (!ret) {
2483 if (dst_max_i_size)
2484 f2fs_i_size_write(dst, dst_max_i_size);
2485 else if (dst_osize != dst->i_size)
2486 f2fs_i_size_write(dst, dst_osize);
2487 }
2488 f2fs_unlock_op(sbi);
6f8d4455
JK		 2489
2490 if (src != dst)
b2532c69 2491 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
6f8d4455
JK		 2492out_src:
2493 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2494out_unlock:
2495 if (src != dst)
4dd6f977 2496 inode_unlock(dst);
20a3d61d 2497out:
4dd6f977
JK		 2498 inode_unlock(src);
2499 return ret;
2500}
2501
2502static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2503{
2504 struct f2fs_move_range range;
2505 struct fd dst;
2506 int err;
2507
2508 if (!(filp->f_mode & FMODE_READ) ||
2509 !(filp->f_mode & FMODE_WRITE))
2510 return -EBADF;
2511
2512 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2513 sizeof(range)))
2514 return -EFAULT;
2515
2516 dst = fdget(range.dst_fd);
2517 if (!dst.file)
2518 return -EBADF;
2519
2520 if (!(dst.file->f_mode & FMODE_WRITE)) {
2521 err = -EBADF;
2522 goto err_out;
2523 }
2524
2525 err = mnt_want_write_file(filp);
2526 if (err)
2527 goto err_out;
2528
2529 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2530 range.pos_out, range.len);
2531
2532 mnt_drop_write_file(filp);
3cecfa5f
KM		 2533 if (err)
2534 goto err_out;
4dd6f977
JK		 2535
2536 if (copy_to_user((struct f2fs_move_range __user *)arg,
2537 &range, sizeof(range)))
2538 err = -EFAULT;
2539err_out:
2540 fdput(dst);
2541 return err;
2542}
2543
e066b83c
JK		 2544static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2545{
2546 struct inode *inode = file_inode(filp);
2547 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2548 struct sit_info *sm = SIT_I(sbi);
2549 unsigned int start_segno = 0, end_segno = 0;
2550 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2551 struct f2fs_flush_device range;
2552 int ret;
2553
2554 if (!capable(CAP_SYS_ADMIN))
2555 return -EPERM;
2556
2557 if (f2fs_readonly(sbi->sb))
2558 return -EROFS;
2559
4354994f
DR		 2560 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2561 return -EINVAL;
2562
e066b83c
JK		 2563 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2564 sizeof(range)))
2565 return -EFAULT;
2566
2567 if (sbi->s_ndevs <= 1 || sbi->s_ndevs - 1 <= range.dev_num ||
2c70c5e3 2568 __is_large_section(sbi)) {
e066b83c
JK		 2569 f2fs_msg(sbi->sb, KERN_WARNING,
2570 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2571 range.dev_num, sbi->s_ndevs,
2572 sbi->segs_per_sec);
2573 return -EINVAL;
2574 }
2575
2576 ret = mnt_want_write_file(filp);
2577 if (ret)
2578 return ret;
2579
2580 if (range.dev_num != 0)
2581 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2582 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2583
2584 start_segno = sm->last_victim[FLUSH_DEVICE];
2585 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2586 start_segno = dev_start_segno;
2587 end_segno = min(start_segno + range.segments, dev_end_segno);
2588
2589 while (start_segno < end_segno) {
2590 if (!mutex_trylock(&sbi->gc_mutex)) {
2591 ret = -EBUSY;
2592 goto out;
2593 }
2594 sm->last_victim[GC_CB] = end_segno + 1;
2595 sm->last_victim[GC_GREEDY] = end_segno + 1;
2596 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2597 ret = f2fs_gc(sbi, true, true, start_segno);
2598 if (ret == -EAGAIN)
2599 ret = 0;
2600 else if (ret < 0)
2601 break;
2602 start_segno++;
2603 }
2604out:
2605 mnt_drop_write_file(filp);
2606 return ret;
2607}
2608
e65ef207
JK		 2609static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2610{
2611 struct inode *inode = file_inode(filp);
2612 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2613
2614 /* Must validate to set it with SQLite behavior in Android. */
2615 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2616
2617 return put_user(sb_feature, (u32 __user *)arg);
2618}
e066b83c 2619
2c1d0305 2620#ifdef CONFIG_QUOTA
78130819
CY		 2621int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2622{
2623 struct dquot *transfer_to[MAXQUOTAS] = {};
2624 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2625 struct super_block *sb = sbi->sb;
2626 int err = 0;
2627
2628 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2629 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2630 err = __dquot_transfer(inode, transfer_to);
2631 if (err)
2632 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2633 dqput(transfer_to[PRJQUOTA]);
2634 }
2635 return err;
2636}
2637
2c1d0305
CY		 2638static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2639{
2640 struct inode *inode = file_inode(filp);
2641 struct f2fs_inode_info *fi = F2FS_I(inode);
2642 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2c1d0305
CY		 2643 struct page *ipage;
2644 kprojid_t kprojid;
2645 int err;
2646
7beb01f7 2647 if (!f2fs_sb_has_project_quota(sbi)) {
2c1d0305
CY		 2648 if (projid != F2FS_DEF_PROJID)
2649 return -EOPNOTSUPP;
2650 else
2651 return 0;
2652 }
2653
2654 if (!f2fs_has_extra_attr(inode))
2655 return -EOPNOTSUPP;
2656
2657 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2658
2659 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2660 return 0;
2661
2c1d0305 2662 err = -EPERM;
2c1d0305
CY		 2663 /* Is it quota file? Do not allow user to mess with it */
2664 if (IS_NOQUOTA(inode))
c8e92757 2665 return err;
2c1d0305 2666
4d57b86d 2667 ipage = f2fs_get_node_page(sbi, inode->i_ino);
c8e92757
WS		 2668 if (IS_ERR(ipage))
2669 return PTR_ERR(ipage);
2c1d0305
CY		 2670
2671 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2672 i_projid)) {
2673 err = -EOVERFLOW;
2674 f2fs_put_page(ipage, 1);
c8e92757 2675 return err;
2c1d0305
CY		 2676 }
2677 f2fs_put_page(ipage, 1);
2678
c22aecd7
CY		 2679 err = dquot_initialize(inode);
2680 if (err)
c8e92757 2681 return err;
2c1d0305 2682
78130819
CY		 2683 f2fs_lock_op(sbi);
2684 err = f2fs_transfer_project_quota(inode, kprojid);
2685 if (err)
2686 goto out_unlock;
2c1d0305
CY		 2687
2688 F2FS_I(inode)->i_projid = kprojid;
2689 inode->i_ctime = current_time(inode);
2c1d0305 2690 f2fs_mark_inode_dirty_sync(inode, true);
78130819
CY		 2691out_unlock:
2692 f2fs_unlock_op(sbi);
2c1d0305
CY		 2693 return err;
2694}
2695#else
78130819
CY		 2696int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2697{
2698 return 0;
2699}
2700
2c1d0305
CY		 2701static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2702{
2703 if (projid != F2FS_DEF_PROJID)
2704 return -EOPNOTSUPP;
2705 return 0;
2706}
2707#endif
2708
2709/* Transfer internal flags to xflags */
2710static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2711{
2712 __u32 xflags = 0;
2713
59c84408 2714 if (iflags & F2FS_SYNC_FL)
2c1d0305 2715 xflags |= FS_XFLAG_SYNC;
59c84408 2716 if (iflags & F2FS_IMMUTABLE_FL)
2c1d0305 2717 xflags |= FS_XFLAG_IMMUTABLE;
59c84408 2718 if (iflags & F2FS_APPEND_FL)
2c1d0305 2719 xflags |= FS_XFLAG_APPEND;
59c84408 2720 if (iflags & F2FS_NODUMP_FL)
2c1d0305 2721 xflags |= FS_XFLAG_NODUMP;
59c84408 2722 if (iflags & F2FS_NOATIME_FL)
2c1d0305 2723 xflags |= FS_XFLAG_NOATIME;
59c84408 2724 if (iflags & F2FS_PROJINHERIT_FL)
2c1d0305
CY		 2725 xflags |= FS_XFLAG_PROJINHERIT;
2726 return xflags;
2727}
2728
2729#define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2730 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2731 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2732
2c1d0305
CY		 2733/* Transfer xflags flags to internal */
2734static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2735{
2736 unsigned long iflags = 0;
2737
2738 if (xflags & FS_XFLAG_SYNC)
59c84408 2739 iflags |= F2FS_SYNC_FL;
2c1d0305 2740 if (xflags & FS_XFLAG_IMMUTABLE)
59c84408 2741 iflags |= F2FS_IMMUTABLE_FL;
2c1d0305 2742 if (xflags & FS_XFLAG_APPEND)
59c84408 2743 iflags |= F2FS_APPEND_FL;
2c1d0305 2744 if (xflags & FS_XFLAG_NODUMP)
59c84408 2745 iflags |= F2FS_NODUMP_FL;
2c1d0305 2746 if (xflags & FS_XFLAG_NOATIME)
59c84408 2747 iflags |= F2FS_NOATIME_FL;
2c1d0305 2748 if (xflags & FS_XFLAG_PROJINHERIT)
59c84408 2749 iflags |= F2FS_PROJINHERIT_FL;
2c1d0305
CY		 2750
2751 return iflags;
2752}
2753
2754static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2755{
2756 struct inode *inode = file_inode(filp);
2757 struct f2fs_inode_info *fi = F2FS_I(inode);
2758 struct fsxattr fa;
2759
2760 memset(&fa, 0, sizeof(struct fsxattr));
2761 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
c807a7cb 2762 F2FS_FL_USER_VISIBLE);
2c1d0305 2763
7beb01f7 2764 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
2c1d0305
CY		 2765 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2766 fi->i_projid);
2767
2768 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2769 return -EFAULT;
2770 return 0;
2771}
2772
c8e92757
WS		 2773static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
2774{
2775 /*
2776 * Project Quota ID state is only allowed to change from within the init
2777 * namespace. Enforce that restriction only if we are trying to change
2778 * the quota ID state. Everything else is allowed in user namespaces.
2779 */
2780 if (current_user_ns() == &init_user_ns)
2781 return 0;
2782
2783 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid)
2784 return -EINVAL;
2785
2786 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) {
2787 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
2788 return -EINVAL;
2789 } else {
2790 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
2791 return -EINVAL;
2792 }
2793
2794 return 0;
2795}
2796
2c1d0305
CY		 2797static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2798{
2799 struct inode *inode = file_inode(filp);
2800 struct f2fs_inode_info *fi = F2FS_I(inode);
2801 struct fsxattr fa;
2802 unsigned int flags;
2803 int err;
2804
2805 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2806 return -EFAULT;
2807
2808 /* Make sure caller has proper permission */
2809 if (!inode_owner_or_capable(inode))
2810 return -EACCES;
2811
2812 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2813 return -EOPNOTSUPP;
2814
2815 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2816 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2817 return -EOPNOTSUPP;
2818
2819 err = mnt_want_write_file(filp);
2820 if (err)
2821 return err;
2822
2823 inode_lock(inode);
c8e92757
WS		 2824 err = f2fs_ioctl_check_project(inode, &fa);
2825 if (err)
2826 goto out;
2c1d0305
CY		 2827 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2828 (flags & F2FS_FL_XFLAG_VISIBLE);
2829 err = __f2fs_ioc_setflags(inode, flags);
2c1d0305 2830 if (err)
c8e92757 2831 goto out;
2c1d0305
CY		 2832
2833 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
c8e92757
WS		 2834out:
2835 inode_unlock(inode);
2836 mnt_drop_write_file(filp);
2837 return err;
2c1d0305 2838}
e066b83c 2839
1ad71a27
JK		 2840int f2fs_pin_file_control(struct inode *inode, bool inc)
2841{
2842 struct f2fs_inode_info *fi = F2FS_I(inode);
2843 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2844
2845 /* Use i_gc_failures for normal file as a risk signal. */
2846 if (inc)
2ef79ecb
CY		 2847 f2fs_i_gc_failures_write(inode,
2848 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
1ad71a27 2849
2ef79ecb 2850 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
1ad71a27
JK		 2851 f2fs_msg(sbi->sb, KERN_WARNING,
2852 "%s: Enable GC = ino %lx after %x GC trials\n",
2ef79ecb
CY		 2853 __func__, inode->i_ino,
2854 fi->i_gc_failures[GC_FAILURE_PIN]);
1ad71a27
JK		 2855 clear_inode_flag(inode, FI_PIN_FILE);
2856 return -EAGAIN;
2857 }
2858 return 0;
2859}
2860
2861static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2862{
2863 struct inode *inode = file_inode(filp);
2864 __u32 pin;
2865 int ret = 0;
2866
2867 if (!inode_owner_or_capable(inode))
2868 return -EACCES;
2869
2870 if (get_user(pin, (__u32 __user *)arg))
2871 return -EFAULT;
2872
2873 if (!S_ISREG(inode->i_mode))
2874 return -EINVAL;
2875
2876 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2877 return -EROFS;
2878
2879 ret = mnt_want_write_file(filp);
2880 if (ret)
2881 return ret;
2882
2883 inode_lock(inode);
2884
4d57b86d 2885 if (f2fs_should_update_outplace(inode, NULL)) {
bb9e3bb8
CY		 2886 ret = -EINVAL;
2887 goto out;
2888 }
2889
1ad71a27
JK		 2890 if (!pin) {
2891 clear_inode_flag(inode, FI_PIN_FILE);
30933364 2892 f2fs_i_gc_failures_write(inode, 0);
1ad71a27
JK		 2893 goto done;
2894 }
2895
2896 if (f2fs_pin_file_control(inode, false)) {
2897 ret = -EAGAIN;
2898 goto out;
2899 }
2900 ret = f2fs_convert_inline_inode(inode);
2901 if (ret)
2902 goto out;
2903
2904 set_inode_flag(inode, FI_PIN_FILE);
2ef79ecb 2905 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
1ad71a27
JK		 2906done:
2907 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2908out:
2909 inode_unlock(inode);
2910 mnt_drop_write_file(filp);
2911 return ret;
2912}
2913
2914static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2915{
2916 struct inode *inode = file_inode(filp);
2917 __u32 pin = 0;
2918
2919 if (is_inode_flag_set(inode, FI_PIN_FILE))
2ef79ecb 2920 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
1ad71a27
JK		 2921 return put_user(pin, (u32 __user *)arg);
2922}
2923
c4020b2d
CY		 2924int f2fs_precache_extents(struct inode *inode)
2925{
2926 struct f2fs_inode_info *fi = F2FS_I(inode);
2927 struct f2fs_map_blocks map;
2928 pgoff_t m_next_extent;
2929 loff_t end;
2930 int err;
2931
2932 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2933 return -EOPNOTSUPP;
2934
2935 map.m_lblk = 0;
2936 map.m_next_pgofs = NULL;
2937 map.m_next_extent = &m_next_extent;
2938 map.m_seg_type = NO_CHECK_TYPE;
f4f0b677 2939 map.m_may_create = false;
c4020b2d
CY		 2940 end = F2FS_I_SB(inode)->max_file_blocks;
2941
2942 while (map.m_lblk < end) {
2943 map.m_len = end - map.m_lblk;
2944
b2532c69 2945 down_write(&fi->i_gc_rwsem[WRITE]);
c4020b2d 2946 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
b2532c69 2947 up_write(&fi->i_gc_rwsem[WRITE]);
c4020b2d
CY		 2948 if (err)
2949 return err;
2950
2951 map.m_lblk = m_next_extent;
2952 }
2953
2954 return err;
2955}
2956
2957static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2958{
2959 return f2fs_precache_extents(file_inode(filp));
2960}
2961
52656e6c
JK		 2962long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2963{
1f227a3e
JK		 2964 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2965 return -EIO;
2966
52656e6c
JK		 2967 switch (cmd) {
2968 case F2FS_IOC_GETFLAGS:
2969 return f2fs_ioc_getflags(filp, arg);
2970 case F2FS_IOC_SETFLAGS:
2971 return f2fs_ioc_setflags(filp, arg);
d49f3e89
CY		 2972 case F2FS_IOC_GETVERSION:
2973 return f2fs_ioc_getversion(filp, arg);
88b88a66
JK		 2974 case F2FS_IOC_START_ATOMIC_WRITE:
2975 return f2fs_ioc_start_atomic_write(filp);
2976 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2977 return f2fs_ioc_commit_atomic_write(filp);
02a1335f
JK		 2978 case F2FS_IOC_START_VOLATILE_WRITE:
2979 return f2fs_ioc_start_volatile_write(filp);
1e84371f
JK		 2980 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2981 return f2fs_ioc_release_volatile_write(filp);
2982 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2983 return f2fs_ioc_abort_volatile_write(filp);
1abff93d
JK		 2984 case F2FS_IOC_SHUTDOWN:
2985 return f2fs_ioc_shutdown(filp, arg);
52656e6c
JK		 2986 case FITRIM:
2987 return f2fs_ioc_fitrim(filp, arg);
f424f664
JK		 2988 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2989 return f2fs_ioc_set_encryption_policy(filp, arg);
2990 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2991 return f2fs_ioc_get_encryption_policy(filp, arg);
2992 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2993 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
c1c1b583
CY		 2994 case F2FS_IOC_GARBAGE_COLLECT:
2995 return f2fs_ioc_gc(filp, arg);
34dc77ad
JK		 2996 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2997 return f2fs_ioc_gc_range(filp, arg);
456b88e4 2998 case F2FS_IOC_WRITE_CHECKPOINT:
059c0648 2999 return f2fs_ioc_write_checkpoint(filp, arg);
d323d005
CY		 3000 case F2FS_IOC_DEFRAGMENT:
3001 return f2fs_ioc_defragment(filp, arg);
4dd6f977
JK		 3002 case F2FS_IOC_MOVE_RANGE:
3003 return f2fs_ioc_move_range(filp, arg);
e066b83c
JK		 3004 case F2FS_IOC_FLUSH_DEVICE:
3005 return f2fs_ioc_flush_device(filp, arg);
e65ef207
JK		 3006 case F2FS_IOC_GET_FEATURES:
3007 return f2fs_ioc_get_features(filp, arg);
2c1d0305
CY		 3008 case F2FS_IOC_FSGETXATTR:
3009 return f2fs_ioc_fsgetxattr(filp, arg);
3010 case F2FS_IOC_FSSETXATTR:
3011 return f2fs_ioc_fssetxattr(filp, arg);
1ad71a27
JK		 3012 case F2FS_IOC_GET_PIN_FILE:
3013 return f2fs_ioc_get_pin_file(filp, arg);
3014 case F2FS_IOC_SET_PIN_FILE:
3015 return f2fs_ioc_set_pin_file(filp, arg);
c4020b2d
CY		 3016 case F2FS_IOC_PRECACHE_EXTENTS:
3017 return f2fs_ioc_precache_extents(filp, arg);
fbfa2cc5
JK		 3018 default:
3019 return -ENOTTY;
3020 }
3021}
3022
fcc85a4d
JK		 3023static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3024{
b439b103
JK		 3025 struct file *file = iocb->ki_filp;
3026 struct inode *inode = file_inode(file);
3027 ssize_t ret;
fcc85a4d 3028
1f227a3e
JK		 3029 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3030 return -EIO;
3031
b91050a8
HL		 3032 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
3033 return -EINVAL;
3034
3035 if (!inode_trylock(inode)) {
3036 if (iocb->ki_flags & IOCB_NOWAIT)
3037 return -EAGAIN;
3038 inode_lock(inode);
3039 }
3040
b439b103
JK		 3041 ret = generic_write_checks(iocb, from);
3042 if (ret > 0) {
dc7a10dd
JK		 3043 bool preallocated = false;
3044 size_t target_size = 0;
dc91de78 3045 int err;
a7de6086 3046
dc91de78
JK		 3047 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3048 set_inode_flag(inode, FI_NO_PREALLOC);
a7de6086 3049
b91050a8
HL		 3050 if ((iocb->ki_flags & IOCB_NOWAIT) &&
3051 (iocb->ki_flags & IOCB_DIRECT)) {
3052 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3053 iov_iter_count(from)) ||
3054 f2fs_has_inline_data(inode) ||
f847c699
CY		 3055 f2fs_force_buffered_io(inode,
3056 iocb, from)) {
cba41be0
YH		 3057 clear_inode_flag(inode,
3058 FI_NO_PREALLOC);
b91050a8
HL		 3059 inode_unlock(inode);
3060 return -EAGAIN;
3061 }
3062
3063 } else {
dc7a10dd
JK		 3064 preallocated = true;
3065 target_size = iocb->ki_pos + iov_iter_count(from);
3066
b91050a8
HL		 3067 err = f2fs_preallocate_blocks(iocb, from);
3068 if (err) {
3069 clear_inode_flag(inode, FI_NO_PREALLOC);
3070 inode_unlock(inode);
3071 return err;
3072 }
9dfa1baf 3073 }
a7de6086 3074 ret = __generic_file_write_iter(iocb, from);
dc91de78 3075 clear_inode_flag(inode, FI_NO_PREALLOC);
b0af6d49 3076
dc7a10dd
JK		 3077 /* if we couldn't write data, we should deallocate blocks. */
3078 if (preallocated && i_size_read(inode) < target_size)
3079 f2fs_truncate(inode);
3080
b0af6d49
CY		 3081 if (ret > 0)
3082 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
b439b103
JK		 3083 }
3084 inode_unlock(inode);
3085
e2592217
CH		 3086 if (ret > 0)
3087 ret = generic_write_sync(iocb, ret);
b439b103 3088 return ret;
fcc85a4d
JK		 3089}
3090
e9750824
NJ		 3091#ifdef CONFIG_COMPAT
3092long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3093{
3094 switch (cmd) {
3095 case F2FS_IOC32_GETFLAGS:
3096 cmd = F2FS_IOC_GETFLAGS;
3097 break;
3098 case F2FS_IOC32_SETFLAGS:
3099 cmd = F2FS_IOC_SETFLAGS;
3100 break;
04ef4b62
CY		 3101 case F2FS_IOC32_GETVERSION:
3102 cmd = F2FS_IOC_GETVERSION;
3103 break;
3104 case F2FS_IOC_START_ATOMIC_WRITE:
3105 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3106 case F2FS_IOC_START_VOLATILE_WRITE:
3107 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3108 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3109 case F2FS_IOC_SHUTDOWN:
3110 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3111 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3112 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3113 case F2FS_IOC_GARBAGE_COLLECT:
34dc77ad 3114 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
04ef4b62
CY		 3115 case F2FS_IOC_WRITE_CHECKPOINT:
3116 case F2FS_IOC_DEFRAGMENT:
4dd6f977 3117 case F2FS_IOC_MOVE_RANGE:
e066b83c 3118 case F2FS_IOC_FLUSH_DEVICE:
e65ef207 3119 case F2FS_IOC_GET_FEATURES:
2c1d0305
CY		 3120 case F2FS_IOC_FSGETXATTR:
3121 case F2FS_IOC_FSSETXATTR:
1ad71a27
JK		 3122 case F2FS_IOC_GET_PIN_FILE:
3123 case F2FS_IOC_SET_PIN_FILE:
c4020b2d 3124 case F2FS_IOC_PRECACHE_EXTENTS:
4dd6f977 3125 break;
e9750824
NJ		 3126 default:
3127 return -ENOIOCTLCMD;
3128 }
3129 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3130}
3131#endif
3132
fbfa2cc5 3133const struct file_operations f2fs_file_operations = {
267378d4 3134 .llseek = f2fs_llseek,
aad4f8bb 3135 .read_iter = generic_file_read_iter,
fcc85a4d
JK		 3136 .write_iter = f2fs_file_write_iter,
3137 .open = f2fs_file_open,
12662234 3138 .release = f2fs_release_file,
fbfa2cc5 3139 .mmap = f2fs_file_mmap,
7a10f017 3140 .flush = f2fs_file_flush,
fbfa2cc5
JK		 3141 .fsync = f2fs_sync_file,
3142 .fallocate = f2fs_fallocate,
3143 .unlocked_ioctl = f2fs_ioctl,
e9750824
NJ		 3144#ifdef CONFIG_COMPAT
3145 .compat_ioctl = f2fs_compat_ioctl,
3146#endif
fbfa2cc5 3147 .splice_read = generic_file_splice_read,
8d020765 3148 .splice_write = iter_file_splice_write,
fbfa2cc5 3149};
Linux 6.x block layer and io_uring tree(s)