1 /*
2 * fs/f2fs/file.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11 #include <linux/fs.h>
12 #include <linux/f2fs_fs.h>
13 #include <linux/stat.h>
14 #include <linux/buffer_head.h>
15 #include <linux/writeback.h>
16 #include <linux/blkdev.h>
17 #include <linux/falloc.h>
18 #include <linux/types.h>
19 #include <linux/compat.h>
20 #include <linux/uaccess.h>
21 #include <linux/mount.h>
22 #include <linux/pagevec.h>
23 #include <linux/uio.h>
24 #include <linux/uuid.h>
25 #include <linux/file.h>
26
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "acl.h"
32 #include "gc.h"
33 #include "trace.h"
34 #include <trace/events/f2fs.h>
35
f2fs_filemap_fault(struct vm_fault * vmf)36 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
37 {
38 struct inode *inode = file_inode(vmf->vma->vm_file);
39 vm_fault_t ret;
40
41 down_read(&F2FS_I(inode)->i_mmap_sem);
42 ret = filemap_fault(vmf);
43 up_read(&F2FS_I(inode)->i_mmap_sem);
44
45 return ret;
46 }
47
f2fs_vm_page_mkwrite(struct vm_fault * vmf)48 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 struct page *page = vmf->page;
51 struct inode *inode = file_inode(vmf->vma->vm_file);
52 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 struct dnode_of_data dn;
54 int err;
55
56 if (unlikely(f2fs_cp_error(sbi))) {
57 err = -EIO;
58 goto err;
59 }
60
61 sb_start_pagefault(inode->i_sb);
62
63 f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
64
65 /* block allocation */
66 f2fs_lock_op(sbi);
67 set_new_dnode(&dn, inode, NULL, NULL, 0);
68 err = f2fs_reserve_block(&dn, page->index);
69 if (err) {
70 f2fs_unlock_op(sbi);
71 goto out;
72 }
73 f2fs_put_dnode(&dn);
74 f2fs_unlock_op(sbi);
75
76 f2fs_balance_fs(sbi, dn.node_changed);
77
78 file_update_time(vmf->vma->vm_file);
79 down_read(&F2FS_I(inode)->i_mmap_sem);
80 lock_page(page);
81 if (unlikely(page->mapping != inode->i_mapping ||
82 page_offset(page) > i_size_read(inode) ||
83 !PageUptodate(page))) {
84 unlock_page(page);
85 err = -EFAULT;
86 goto out_sem;
87 }
88
89 /*
90 * check to see if the page is mapped already (no holes)
91 */
92 if (PageMappedToDisk(page))
93 goto mapped;
94
95 /* page is wholly or partially inside EOF */
96 if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
97 i_size_read(inode)) {
98 loff_t offset;
99
100 offset = i_size_read(inode) & ~PAGE_MASK;
101 zero_user_segment(page, offset, PAGE_SIZE);
102 }
103 set_page_dirty(page);
104 if (!PageUptodate(page))
105 SetPageUptodate(page);
106
107 f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
108
109 trace_f2fs_vm_page_mkwrite(page, DATA);
110 mapped:
111 /* fill the page */
112 f2fs_wait_on_page_writeback(page, DATA, false);
113
114 /* wait for GCed page writeback via META_MAPPING */
115 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
116
117 out_sem:
118 up_read(&F2FS_I(inode)->i_mmap_sem);
119 out:
120 sb_end_pagefault(inode->i_sb);
121 f2fs_update_time(sbi, REQ_TIME);
122 err:
123 return block_page_mkwrite_return(err);
124 }
125
126 static const struct vm_operations_struct f2fs_file_vm_ops = {
127 .fault = f2fs_filemap_fault,
128 .map_pages = filemap_map_pages,
129 .page_mkwrite = f2fs_vm_page_mkwrite,
130 };
131
get_parent_ino(struct inode * inode,nid_t * pino)132 static int get_parent_ino(struct inode *inode, nid_t *pino)
133 {
134 struct dentry *dentry;
135
136 inode = igrab(inode);
137 dentry = d_find_any_alias(inode);
138 iput(inode);
139 if (!dentry)
140 return 0;
141
142 *pino = parent_ino(dentry);
143 dput(dentry);
144 return 1;
145 }
146
need_do_checkpoint(struct inode * inode)147 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
148 {
149 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
150 enum cp_reason_type cp_reason = CP_NO_NEEDED;
151
152 if (!S_ISREG(inode->i_mode))
153 cp_reason = CP_NON_REGULAR;
154 else if (inode->i_nlink != 1)
155 cp_reason = CP_HARDLINK;
156 else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
157 cp_reason = CP_SB_NEED_CP;
158 else if (file_wrong_pino(inode))
159 cp_reason = CP_WRONG_PINO;
160 else if (!f2fs_space_for_roll_forward(sbi))
161 cp_reason = CP_NO_SPC_ROLL;
162 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
163 cp_reason = CP_NODE_NEED_CP;
164 else if (test_opt(sbi, FASTBOOT))
165 cp_reason = CP_FASTBOOT_MODE;
166 else if (F2FS_OPTION(sbi).active_logs == 2)
167 cp_reason = CP_SPEC_LOG_NUM;
168 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
169 f2fs_need_dentry_mark(sbi, inode->i_ino) &&
170 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
171 TRANS_DIR_INO))
172 cp_reason = CP_RECOVER_DIR;
173
174 return cp_reason;
175 }
176
need_inode_page_update(struct f2fs_sb_info * sbi,nid_t ino)177 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
178 {
179 struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
180 bool ret = false;
181 /* But we need to avoid that there are some inode updates */
182 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
183 ret = true;
184 f2fs_put_page(i, 0);
185 return ret;
186 }
187
try_to_fix_pino(struct inode * inode)188 static void try_to_fix_pino(struct inode *inode)
189 {
190 struct f2fs_inode_info *fi = F2FS_I(inode);
191 nid_t pino;
192
193 down_write(&fi->i_sem);
194 if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
195 get_parent_ino(inode, &pino)) {
196 f2fs_i_pino_write(inode, pino);
197 file_got_pino(inode);
198 }
199 up_write(&fi->i_sem);
200 }
201
f2fs_do_sync_file(struct file * file,loff_t start,loff_t end,int datasync,bool atomic)202 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
203 int datasync, bool atomic)
204 {
205 struct inode *inode = file->f_mapping->host;
206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
207 nid_t ino = inode->i_ino;
208 int ret = 0;
209 enum cp_reason_type cp_reason = 0;
210 struct writeback_control wbc = {
211 .sync_mode = WB_SYNC_ALL,
212 .nr_to_write = LONG_MAX,
213 .for_reclaim = 0,
214 };
215 unsigned int seq_id = 0;
216
217 if (unlikely(f2fs_readonly(inode->i_sb)))
218 return 0;
219
220 trace_f2fs_sync_file_enter(inode);
221
222 if (S_ISDIR(inode->i_mode))
223 goto go_write;
224
225 /* if fdatasync is triggered, let's do in-place-update */
226 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
227 set_inode_flag(inode, FI_NEED_IPU);
228 ret = file_write_and_wait_range(file, start, end);
229 clear_inode_flag(inode, FI_NEED_IPU);
230
231 if (ret) {
232 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
233 return ret;
234 }
235
236 /* if the inode is dirty, let's recover all the time */
237 if (!f2fs_skip_inode_update(inode, datasync)) {
238 f2fs_write_inode(inode, NULL);
239 goto go_write;
240 }
241
242 /*
243 * if there is no written data, don't waste time to write recovery info.
244 */
245 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
246 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
247
248 /* it may call write_inode just prior to fsync */
249 if (need_inode_page_update(sbi, ino))
250 goto go_write;
251
252 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
253 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
254 goto flush_out;
255 goto out;
256 }
257 go_write:
258 /*
259 * Both of fdatasync() and fsync() are able to be recovered from
260 * sudden-power-off.
261 */
262 down_read(&F2FS_I(inode)->i_sem);
263 cp_reason = need_do_checkpoint(inode);
264 up_read(&F2FS_I(inode)->i_sem);
265
266 if (cp_reason) {
267 /* all the dirty node pages should be flushed for POR */
268 ret = f2fs_sync_fs(inode->i_sb, 1);
269
270 /*
271 * We've secured consistency through sync_fs. Following pino
272 * will be used only for fsynced inodes after checkpoint.
273 */
274 try_to_fix_pino(inode);
275 clear_inode_flag(inode, FI_APPEND_WRITE);
276 clear_inode_flag(inode, FI_UPDATE_WRITE);
277 goto out;
278 }
279 sync_nodes:
280 atomic_inc(&sbi->wb_sync_req[NODE]);
281 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
282 atomic_dec(&sbi->wb_sync_req[NODE]);
283 if (ret)
284 goto out;
285
286 /* if cp_error was enabled, we should avoid infinite loop */
287 if (unlikely(f2fs_cp_error(sbi))) {
288 ret = -EIO;
289 goto out;
290 }
291
292 if (f2fs_need_inode_block_update(sbi, ino)) {
293 f2fs_mark_inode_dirty_sync(inode, true);
294 f2fs_write_inode(inode, NULL);
295 goto sync_nodes;
296 }
297
298 /*
299 * If it's atomic_write, it's just fine to keep write ordering. So
300 * here we don't need to wait for node write completion, since we use
301 * node chain which serializes node blocks. If one of node writes are
302 * reordered, we can see simply broken chain, resulting in stopping
303 * roll-forward recovery. It means we'll recover all or none node blocks
304 * given fsync mark.
305 */
306 if (!atomic) {
307 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
308 if (ret)
309 goto out;
310 }
311
312 /* once recovery info is written, don't need to tack this */
313 f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
314 clear_inode_flag(inode, FI_APPEND_WRITE);
315 flush_out:
316 if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
317 ret = f2fs_issue_flush(sbi, inode->i_ino);
318 if (!ret) {
319 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
320 clear_inode_flag(inode, FI_UPDATE_WRITE);
321 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
322 }
323 f2fs_update_time(sbi, REQ_TIME);
324 out:
325 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
326 f2fs_trace_ios(NULL, 1);
327 return ret;
328 }
329
f2fs_sync_file(struct file * file,loff_t start,loff_t end,int datasync)330 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
331 {
332 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
333 return -EIO;
334 return f2fs_do_sync_file(file, start, end, datasync, false);
335 }
336
__get_first_dirty_index(struct address_space * mapping,pgoff_t pgofs,int whence)337 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
338 pgoff_t pgofs, int whence)
339 {
340 struct page *page;
341 int nr_pages;
342
343 if (whence != SEEK_DATA)
344 return 0;
345
346 /* find first dirty page index */
347 nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
348 1, &page);
349 if (!nr_pages)
350 return ULONG_MAX;
351 pgofs = page->index;
352 put_page(page);
353 return pgofs;
354 }
355
__found_offset(struct f2fs_sb_info * sbi,block_t blkaddr,pgoff_t dirty,pgoff_t pgofs,int whence)356 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
357 pgoff_t dirty, pgoff_t pgofs, int whence)
358 {
359 switch (whence) {
360 case SEEK_DATA:
361 if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
362 is_valid_data_blkaddr(sbi, blkaddr))
363 return true;
364 break;
365 case SEEK_HOLE:
366 if (blkaddr == NULL_ADDR)
367 return true;
368 break;
369 }
370 return false;
371 }
372
f2fs_seek_block(struct file * file,loff_t offset,int whence)373 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
374 {
375 struct inode *inode = file->f_mapping->host;
376 loff_t maxbytes = inode->i_sb->s_maxbytes;
377 struct dnode_of_data dn;
378 pgoff_t pgofs, end_offset, dirty;
379 loff_t data_ofs = offset;
380 loff_t isize;
381 int err = 0;
382
383 inode_lock(inode);
384
385 isize = i_size_read(inode);
386 if (offset >= isize)
387 goto fail;
388
389 /* handle inline data case */
390 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
391 if (whence == SEEK_HOLE)
392 data_ofs = isize;
393 goto found;
394 }
395
396 pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
397
398 dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
399
400 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
401 set_new_dnode(&dn, inode, NULL, NULL, 0);
402 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
403 if (err && err != -ENOENT) {
404 goto fail;
405 } else if (err == -ENOENT) {
406 /* direct node does not exists */
407 if (whence == SEEK_DATA) {
408 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
409 continue;
410 } else {
411 goto found;
412 }
413 }
414
415 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
416
417 /* find data/hole in dnode block */
418 for (; dn.ofs_in_node < end_offset;
419 dn.ofs_in_node++, pgofs++,
420 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
421 block_t blkaddr;
422
423 blkaddr = datablock_addr(dn.inode,
424 dn.node_page, dn.ofs_in_node);
425
426 if (__is_valid_data_blkaddr(blkaddr) &&
427 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
428 blkaddr, DATA_GENERIC)) {
429 f2fs_put_dnode(&dn);
430 goto fail;
431 }
432
433 if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
434 pgofs, whence)) {
435 f2fs_put_dnode(&dn);
436 goto found;
437 }
438 }
439 f2fs_put_dnode(&dn);
440 }
441
442 if (whence == SEEK_DATA)
443 goto fail;
444 found:
445 if (whence == SEEK_HOLE && data_ofs > isize)
446 data_ofs = isize;
447 inode_unlock(inode);
448 return vfs_setpos(file, data_ofs, maxbytes);
449 fail:
450 inode_unlock(inode);
451 return -ENXIO;
452 }
453
f2fs_llseek(struct file * file,loff_t offset,int whence)454 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
455 {
456 struct inode *inode = file->f_mapping->host;
457 loff_t maxbytes = inode->i_sb->s_maxbytes;
458
459 switch (whence) {
460 case SEEK_SET:
461 case SEEK_CUR:
462 case SEEK_END:
463 return generic_file_llseek_size(file, offset, whence,
464 maxbytes, i_size_read(inode));
465 case SEEK_DATA:
466 case SEEK_HOLE:
467 if (offset < 0)
468 return -ENXIO;
469 return f2fs_seek_block(file, offset, whence);
470 }
471
472 return -EINVAL;
473 }
474
f2fs_file_mmap(struct file * file,struct vm_area_struct * vma)475 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
476 {
477 struct inode *inode = file_inode(file);
478 int err;
479
480 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
481 return -EIO;
482
483 /* we don't need to use inline_data strictly */
484 err = f2fs_convert_inline_inode(inode);
485 if (err)
486 return err;
487
488 file_accessed(file);
489 vma->vm_ops = &f2fs_file_vm_ops;
490 return 0;
491 }
492
f2fs_file_open(struct inode * inode,struct file * filp)493 static int f2fs_file_open(struct inode *inode, struct file *filp)
494 {
495 int err = fscrypt_file_open(inode, filp);
496
497 if (err)
498 return err;
499
500 filp->f_mode |= FMODE_NOWAIT;
501
502 return dquot_file_open(inode, filp);
503 }
504
f2fs_truncate_data_blocks_range(struct dnode_of_data * dn,int count)505 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
506 {
507 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
508 struct f2fs_node *raw_node;
509 int nr_free = 0, ofs = dn->ofs_in_node, len = count;
510 __le32 *addr;
511 int base = 0;
512
513 if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
514 base = get_extra_isize(dn->inode);
515
516 raw_node = F2FS_NODE(dn->node_page);
517 addr = blkaddr_in_node(raw_node) + base + ofs;
518
519 for (; count > 0; count--, addr++, dn->ofs_in_node++) {
520 block_t blkaddr = le32_to_cpu(*addr);
521
522 if (blkaddr == NULL_ADDR)
523 continue;
524
525 dn->data_blkaddr = NULL_ADDR;
526 f2fs_set_data_blkaddr(dn);
527
528 if (__is_valid_data_blkaddr(blkaddr) &&
529 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC))
530 continue;
531
532 f2fs_invalidate_blocks(sbi, blkaddr);
533 if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
534 clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
535 nr_free++;
536 }
537
538 if (nr_free) {
539 pgoff_t fofs;
540 /*
541 * once we invalidate valid blkaddr in range [ofs, ofs + count],
542 * we will invalidate all blkaddr in the whole range.
543 */
544 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
545 dn->inode) + ofs;
546 f2fs_update_extent_cache_range(dn, fofs, 0, len);
547 dec_valid_block_count(sbi, dn->inode, nr_free);
548 }
549 dn->ofs_in_node = ofs;
550
551 f2fs_update_time(sbi, REQ_TIME);
552 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
553 dn->ofs_in_node, nr_free);
554 }
555
f2fs_truncate_data_blocks(struct dnode_of_data * dn)556 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
557 {
558 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
559 }
560
truncate_partial_data_page(struct inode * inode,u64 from,bool cache_only)561 static int truncate_partial_data_page(struct inode *inode, u64 from,
562 bool cache_only)
563 {
564 loff_t offset = from & (PAGE_SIZE - 1);
565 pgoff_t index = from >> PAGE_SHIFT;
566 struct address_space *mapping = inode->i_mapping;
567 struct page *page;
568
569 if (!offset && !cache_only)
570 return 0;
571
572 if (cache_only) {
573 page = find_lock_page(mapping, index);
574 if (page && PageUptodate(page))
575 goto truncate_out;
576 f2fs_put_page(page, 1);
577 return 0;
578 }
579
580 page = f2fs_get_lock_data_page(inode, index, true);
581 if (IS_ERR(page))
582 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
583 truncate_out:
584 f2fs_wait_on_page_writeback(page, DATA, true);
585 zero_user(page, offset, PAGE_SIZE - offset);
586
587 /* An encrypted inode should have a key and truncate the last page. */
588 f2fs_bug_on(F2FS_I_SB(inode), cache_only && f2fs_encrypted_inode(inode));
589 if (!cache_only)
590 set_page_dirty(page);
591 f2fs_put_page(page, 1);
592 return 0;
593 }
594
f2fs_truncate_blocks(struct inode * inode,u64 from,bool lock)595 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
596 {
597 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
598 struct dnode_of_data dn;
599 pgoff_t free_from;
600 int count = 0, err = 0;
601 struct page *ipage;
602 bool truncate_page = false;
603
604 trace_f2fs_truncate_blocks_enter(inode, from);
605
606 free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
607
608 if (free_from >= sbi->max_file_blocks)
609 goto free_partial;
610
611 if (lock)
612 f2fs_lock_op(sbi);
613
614 ipage = f2fs_get_node_page(sbi, inode->i_ino);
615 if (IS_ERR(ipage)) {
616 err = PTR_ERR(ipage);
617 goto out;
618 }
619
620 if (f2fs_has_inline_data(inode)) {
621 f2fs_truncate_inline_inode(inode, ipage, from);
622 f2fs_put_page(ipage, 1);
623 truncate_page = true;
624 goto out;
625 }
626
627 set_new_dnode(&dn, inode, ipage, NULL, 0);
628 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
629 if (err) {
630 if (err == -ENOENT)
631 goto free_next;
632 goto out;
633 }
634
635 count = ADDRS_PER_PAGE(dn.node_page, inode);
636
637 count -= dn.ofs_in_node;
638 f2fs_bug_on(sbi, count < 0);
639
640 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
641 f2fs_truncate_data_blocks_range(&dn, count);
642 free_from += count;
643 }
644
645 f2fs_put_dnode(&dn);
646 free_next:
647 err = f2fs_truncate_inode_blocks(inode, free_from);
648 out:
649 if (lock)
650 f2fs_unlock_op(sbi);
651 free_partial:
652 /* lastly zero out the first data page */
653 if (!err)
654 err = truncate_partial_data_page(inode, from, truncate_page);
655
656 trace_f2fs_truncate_blocks_exit(inode, err);
657 return err;
658 }
659
f2fs_truncate(struct inode * inode)660 int f2fs_truncate(struct inode *inode)
661 {
662 int err;
663
664 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
665 return -EIO;
666
667 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
668 S_ISLNK(inode->i_mode)))
669 return 0;
670
671 trace_f2fs_truncate(inode);
672
673 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
674 f2fs_show_injection_info(FAULT_TRUNCATE);
675 return -EIO;
676 }
677
678 /* we should check inline_data size */
679 if (!f2fs_may_inline_data(inode)) {
680 err = f2fs_convert_inline_inode(inode);
681 if (err)
682 return err;
683 }
684
685 err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
686 if (err)
687 return err;
688
689 inode->i_mtime = inode->i_ctime = current_time(inode);
690 f2fs_mark_inode_dirty_sync(inode, false);
691 return 0;
692 }
693
f2fs_getattr(const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)694 int f2fs_getattr(const struct path *path, struct kstat *stat,
695 u32 request_mask, unsigned int query_flags)
696 {
697 struct inode *inode = d_inode(path->dentry);
698 struct f2fs_inode_info *fi = F2FS_I(inode);
699 struct f2fs_inode *ri;
700 unsigned int flags;
701
702 if (f2fs_has_extra_attr(inode) &&
703 f2fs_sb_has_inode_crtime(inode->i_sb) &&
704 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
705 stat->result_mask |= STATX_BTIME;
706 stat->btime.tv_sec = fi->i_crtime.tv_sec;
707 stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
708 }
709
710 flags = fi->i_flags & F2FS_FL_USER_VISIBLE;
711 if (flags & F2FS_APPEND_FL)
712 stat->attributes |= STATX_ATTR_APPEND;
713 if (flags & F2FS_COMPR_FL)
714 stat->attributes |= STATX_ATTR_COMPRESSED;
715 if (f2fs_encrypted_inode(inode))
716 stat->attributes |= STATX_ATTR_ENCRYPTED;
717 if (flags & F2FS_IMMUTABLE_FL)
718 stat->attributes |= STATX_ATTR_IMMUTABLE;
719 if (flags & F2FS_NODUMP_FL)
720 stat->attributes |= STATX_ATTR_NODUMP;
721
722 stat->attributes_mask |= (STATX_ATTR_APPEND |
723 STATX_ATTR_COMPRESSED |
724 STATX_ATTR_ENCRYPTED |
725 STATX_ATTR_IMMUTABLE |
726 STATX_ATTR_NODUMP);
727
728 generic_fillattr(inode, stat);
729
730 /* we need to show initial sectors used for inline_data/dentries */
731 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
732 f2fs_has_inline_dentry(inode))
733 stat->blocks += (stat->size + 511) >> 9;
734
735 return 0;
736 }
737
738 #ifdef CONFIG_F2FS_FS_POSIX_ACL
__setattr_copy(struct inode * inode,const struct iattr * attr)739 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
740 {
741 unsigned int ia_valid = attr->ia_valid;
742
743 if (ia_valid & ATTR_UID)
744 inode->i_uid = attr->ia_uid;
745 if (ia_valid & ATTR_GID)
746 inode->i_gid = attr->ia_gid;
747 if (ia_valid & ATTR_ATIME)
748 inode->i_atime = timespec64_trunc(attr->ia_atime,
749 inode->i_sb->s_time_gran);
750 if (ia_valid & ATTR_MTIME)
751 inode->i_mtime = timespec64_trunc(attr->ia_mtime,
752 inode->i_sb->s_time_gran);
753 if (ia_valid & ATTR_CTIME)
754 inode->i_ctime = timespec64_trunc(attr->ia_ctime,
755 inode->i_sb->s_time_gran);
756 if (ia_valid & ATTR_MODE) {
757 umode_t mode = attr->ia_mode;
758
759 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
760 mode &= ~S_ISGID;
761 set_acl_inode(inode, mode);
762 }
763 }
764 #else
765 #define __setattr_copy setattr_copy
766 #endif
767
f2fs_setattr(struct dentry * dentry,struct iattr * attr)768 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
769 {
770 struct inode *inode = d_inode(dentry);
771 int err;
772
773 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
774 return -EIO;
775
776 err = setattr_prepare(dentry, attr);
777 if (err)
778 return err;
779
780 err = fscrypt_prepare_setattr(dentry, attr);
781 if (err)
782 return err;
783
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))) {
793 err = dquot_transfer(inode, attr);
794 if (err)
795 return err;
796 }
797
798 if (attr->ia_valid & ATTR_SIZE) {
799 bool to_smaller = (attr->ia_size <= i_size_read(inode));
800
801 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
802 down_write(&F2FS_I(inode)->i_mmap_sem);
803
804 truncate_setsize(inode, attr->ia_size);
805
806 if (to_smaller)
807 err = f2fs_truncate(inode);
808 /*
809 * do not trim all blocks after i_size if target size is
810 * larger than i_size.
811 */
812 up_write(&F2FS_I(inode)->i_mmap_sem);
813 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
814
815 if (err)
816 return err;
817
818 if (!to_smaller) {
819 /* should convert inline inode here */
820 if (!f2fs_may_inline_data(inode)) {
821 err = f2fs_convert_inline_inode(inode);
822 if (err)
823 return err;
824 }
825 inode->i_mtime = inode->i_ctime = current_time(inode);
826 }
827
828 down_write(&F2FS_I(inode)->i_sem);
829 F2FS_I(inode)->last_disk_size = i_size_read(inode);
830 up_write(&F2FS_I(inode)->i_sem);
831 }
832
833 __setattr_copy(inode, attr);
834
835 if (attr->ia_valid & ATTR_MODE) {
836 err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
837 if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
838 inode->i_mode = F2FS_I(inode)->i_acl_mode;
839 clear_inode_flag(inode, FI_ACL_MODE);
840 }
841 }
842
843 /* file size may changed here */
844 f2fs_mark_inode_dirty_sync(inode, true);
845
846 /* inode change will produce dirty node pages flushed by checkpoint */
847 f2fs_balance_fs(F2FS_I_SB(inode), true);
848
849 return err;
850 }
851
852 const struct inode_operations f2fs_file_inode_operations = {
853 .getattr = f2fs_getattr,
854 .setattr = f2fs_setattr,
855 .get_acl = f2fs_get_acl,
856 .set_acl = f2fs_set_acl,
857 #ifdef CONFIG_F2FS_FS_XATTR
858 .listxattr = f2fs_listxattr,
859 #endif
860 .fiemap = f2fs_fiemap,
861 };
862
fill_zero(struct inode * inode,pgoff_t index,loff_t start,loff_t len)863 static int fill_zero(struct inode *inode, pgoff_t index,
864 loff_t start, loff_t len)
865 {
866 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
867 struct page *page;
868
869 if (!len)
870 return 0;
871
872 f2fs_balance_fs(sbi, true);
873
874 f2fs_lock_op(sbi);
875 page = f2fs_get_new_data_page(inode, NULL, index, false);
876 f2fs_unlock_op(sbi);
877
878 if (IS_ERR(page))
879 return PTR_ERR(page);
880
881 f2fs_wait_on_page_writeback(page, DATA, true);
882 zero_user(page, start, len);
883 set_page_dirty(page);
884 f2fs_put_page(page, 1);
885 return 0;
886 }
887
f2fs_truncate_hole(struct inode * inode,pgoff_t pg_start,pgoff_t pg_end)888 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
889 {
890 int err;
891
892 while (pg_start < pg_end) {
893 struct dnode_of_data dn;
894 pgoff_t end_offset, count;
895
896 set_new_dnode(&dn, inode, NULL, NULL, 0);
897 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
898 if (err) {
899 if (err == -ENOENT) {
900 pg_start = f2fs_get_next_page_offset(&dn,
901 pg_start);
902 continue;
903 }
904 return err;
905 }
906
907 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
908 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
909
910 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
911
912 f2fs_truncate_data_blocks_range(&dn, count);
913 f2fs_put_dnode(&dn);
914
915 pg_start += count;
916 }
917 return 0;
918 }
919
punch_hole(struct inode * inode,loff_t offset,loff_t len)920 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
921 {
922 pgoff_t pg_start, pg_end;
923 loff_t off_start, off_end;
924 int ret;
925
926 ret = f2fs_convert_inline_inode(inode);
927 if (ret)
928 return ret;
929
930 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
931 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
932
933 off_start = offset & (PAGE_SIZE - 1);
934 off_end = (offset + len) & (PAGE_SIZE - 1);
935
936 if (pg_start == pg_end) {
937 ret = fill_zero(inode, pg_start, off_start,
938 off_end - off_start);
939 if (ret)
940 return ret;
941 } else {
942 if (off_start) {
943 ret = fill_zero(inode, pg_start++, off_start,
944 PAGE_SIZE - off_start);
945 if (ret)
946 return ret;
947 }
948 if (off_end) {
949 ret = fill_zero(inode, pg_end, 0, off_end);
950 if (ret)
951 return ret;
952 }
953
954 if (pg_start < pg_end) {
955 struct address_space *mapping = inode->i_mapping;
956 loff_t blk_start, blk_end;
957 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
958
959 f2fs_balance_fs(sbi, true);
960
961 blk_start = (loff_t)pg_start << PAGE_SHIFT;
962 blk_end = (loff_t)pg_end << PAGE_SHIFT;
963
964 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
965 down_write(&F2FS_I(inode)->i_mmap_sem);
966
967 truncate_inode_pages_range(mapping, blk_start,
968 blk_end - 1);
969
970 f2fs_lock_op(sbi);
971 ret = f2fs_truncate_hole(inode, pg_start, pg_end);
972 f2fs_unlock_op(sbi);
973
974 up_write(&F2FS_I(inode)->i_mmap_sem);
975 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
976 }
977 }
978
979 return ret;
980 }
981
__read_out_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,pgoff_t len)982 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
983 int *do_replace, pgoff_t off, pgoff_t len)
984 {
985 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
986 struct dnode_of_data dn;
987 int ret, done, i;
988
989 next_dnode:
990 set_new_dnode(&dn, inode, NULL, NULL, 0);
991 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
992 if (ret && ret != -ENOENT) {
993 return ret;
994 } else if (ret == -ENOENT) {
995 if (dn.max_level == 0)
996 return -ENOENT;
997 done = min((pgoff_t)ADDRS_PER_BLOCK - dn.ofs_in_node, len);
998 blkaddr += done;
999 do_replace += done;
1000 goto next;
1001 }
1002
1003 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1004 dn.ofs_in_node, len);
1005 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1006 *blkaddr = datablock_addr(dn.inode,
1007 dn.node_page, dn.ofs_in_node);
1008 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1009
1010 if (test_opt(sbi, LFS)) {
1011 f2fs_put_dnode(&dn);
1012 return -ENOTSUPP;
1013 }
1014
1015 /* do not invalidate this block address */
1016 f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1017 *do_replace = 1;
1018 }
1019 }
1020 f2fs_put_dnode(&dn);
1021 next:
1022 len -= done;
1023 off += done;
1024 if (len)
1025 goto next_dnode;
1026 return 0;
1027 }
1028
__roll_back_blkaddrs(struct inode * inode,block_t * blkaddr,int * do_replace,pgoff_t off,int len)1029 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1030 int *do_replace, pgoff_t off, int len)
1031 {
1032 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1033 struct dnode_of_data dn;
1034 int ret, i;
1035
1036 for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1037 if (*do_replace == 0)
1038 continue;
1039
1040 set_new_dnode(&dn, inode, NULL, NULL, 0);
1041 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1042 if (ret) {
1043 dec_valid_block_count(sbi, inode, 1);
1044 f2fs_invalidate_blocks(sbi, *blkaddr);
1045 } else {
1046 f2fs_update_data_blkaddr(&dn, *blkaddr);
1047 }
1048 f2fs_put_dnode(&dn);
1049 }
1050 return 0;
1051 }
1052
__clone_blkaddrs(struct inode * src_inode,struct inode * dst_inode,block_t * blkaddr,int * do_replace,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1053 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1054 block_t *blkaddr, int *do_replace,
1055 pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1056 {
1057 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1058 pgoff_t i = 0;
1059 int ret;
1060
1061 while (i < len) {
1062 if (blkaddr[i] == NULL_ADDR && !full) {
1063 i++;
1064 continue;
1065 }
1066
1067 if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1068 struct dnode_of_data dn;
1069 struct node_info ni;
1070 size_t new_size;
1071 pgoff_t ilen;
1072
1073 set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1074 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1075 if (ret)
1076 return ret;
1077
1078 ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1079 if (ret) {
1080 f2fs_put_dnode(&dn);
1081 return ret;
1082 }
1083
1084 ilen = min((pgoff_t)
1085 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1086 dn.ofs_in_node, len - i);
1087 do {
1088 dn.data_blkaddr = datablock_addr(dn.inode,
1089 dn.node_page, dn.ofs_in_node);
1090 f2fs_truncate_data_blocks_range(&dn, 1);
1091
1092 if (do_replace[i]) {
1093 f2fs_i_blocks_write(src_inode,
1094 1, false, false);
1095 f2fs_i_blocks_write(dst_inode,
1096 1, true, false);
1097 f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1098 blkaddr[i], ni.version, true, false);
1099
1100 do_replace[i] = 0;
1101 }
1102 dn.ofs_in_node++;
1103 i++;
1104 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1105 if (dst_inode->i_size < new_size)
1106 f2fs_i_size_write(dst_inode, new_size);
1107 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1108
1109 f2fs_put_dnode(&dn);
1110 } else {
1111 struct page *psrc, *pdst;
1112
1113 psrc = f2fs_get_lock_data_page(src_inode,
1114 src + i, true);
1115 if (IS_ERR(psrc))
1116 return PTR_ERR(psrc);
1117 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1118 true);
1119 if (IS_ERR(pdst)) {
1120 f2fs_put_page(psrc, 1);
1121 return PTR_ERR(pdst);
1122 }
1123 f2fs_copy_page(psrc, pdst);
1124 set_page_dirty(pdst);
1125 f2fs_put_page(pdst, 1);
1126 f2fs_put_page(psrc, 1);
1127
1128 ret = f2fs_truncate_hole(src_inode,
1129 src + i, src + i + 1);
1130 if (ret)
1131 return ret;
1132 i++;
1133 }
1134 }
1135 return 0;
1136 }
1137
__exchange_data_block(struct inode * src_inode,struct inode * dst_inode,pgoff_t src,pgoff_t dst,pgoff_t len,bool full)1138 static int __exchange_data_block(struct inode *src_inode,
1139 struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1140 pgoff_t len, bool full)
1141 {
1142 block_t *src_blkaddr;
1143 int *do_replace;
1144 pgoff_t olen;
1145 int ret;
1146
1147 while (len) {
1148 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
1149
1150 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1151 array_size(olen, sizeof(block_t)),
1152 GFP_KERNEL);
1153 if (!src_blkaddr)
1154 return -ENOMEM;
1155
1156 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1157 array_size(olen, sizeof(int)),
1158 GFP_KERNEL);
1159 if (!do_replace) {
1160 kvfree(src_blkaddr);
1161 return -ENOMEM;
1162 }
1163
1164 ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1165 do_replace, src, olen);
1166 if (ret)
1167 goto roll_back;
1168
1169 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1170 do_replace, src, dst, olen, full);
1171 if (ret)
1172 goto roll_back;
1173
1174 src += olen;
1175 dst += olen;
1176 len -= olen;
1177
1178 kvfree(src_blkaddr);
1179 kvfree(do_replace);
1180 }
1181 return 0;
1182
1183 roll_back:
1184 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1185 kvfree(src_blkaddr);
1186 kvfree(do_replace);
1187 return ret;
1188 }
1189
f2fs_do_collapse(struct inode * inode,loff_t offset,loff_t len)1190 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1191 {
1192 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1193 pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1194 pgoff_t start = offset >> PAGE_SHIFT;
1195 pgoff_t end = (offset + len) >> PAGE_SHIFT;
1196 int ret;
1197
1198 f2fs_balance_fs(sbi, true);
1199
1200 /* avoid gc operation during block exchange */
1201 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1202 down_write(&F2FS_I(inode)->i_mmap_sem);
1203
1204 f2fs_lock_op(sbi);
1205 f2fs_drop_extent_tree(inode);
1206 truncate_pagecache(inode, offset);
1207 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1208 f2fs_unlock_op(sbi);
1209
1210 up_write(&F2FS_I(inode)->i_mmap_sem);
1211 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1212 return ret;
1213 }
1214
f2fs_collapse_range(struct inode * inode,loff_t offset,loff_t len)1215 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1216 {
1217 loff_t new_size;
1218 int ret;
1219
1220 if (offset + len >= i_size_read(inode))
1221 return -EINVAL;
1222
1223 /* collapse range should be aligned to block size of f2fs. */
1224 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1225 return -EINVAL;
1226
1227 ret = f2fs_convert_inline_inode(inode);
1228 if (ret)
1229 return ret;
1230
1231 /* write out all dirty pages from offset */
1232 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1233 if (ret)
1234 return ret;
1235
1236 ret = f2fs_do_collapse(inode, offset, len);
1237 if (ret)
1238 return ret;
1239
1240 /* write out all moved pages, if possible */
1241 down_write(&F2FS_I(inode)->i_mmap_sem);
1242 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1243 truncate_pagecache(inode, offset);
1244
1245 new_size = i_size_read(inode) - len;
1246 truncate_pagecache(inode, new_size);
1247
1248 ret = f2fs_truncate_blocks(inode, new_size, true);
1249 up_write(&F2FS_I(inode)->i_mmap_sem);
1250 if (!ret)
1251 f2fs_i_size_write(inode, new_size);
1252 return ret;
1253 }
1254
f2fs_do_zero_range(struct dnode_of_data * dn,pgoff_t start,pgoff_t end)1255 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1256 pgoff_t end)
1257 {
1258 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1259 pgoff_t index = start;
1260 unsigned int ofs_in_node = dn->ofs_in_node;
1261 blkcnt_t count = 0;
1262 int ret;
1263
1264 for (; index < end; index++, dn->ofs_in_node++) {
1265 if (datablock_addr(dn->inode, dn->node_page,
1266 dn->ofs_in_node) == NULL_ADDR)
1267 count++;
1268 }
1269
1270 dn->ofs_in_node = ofs_in_node;
1271 ret = f2fs_reserve_new_blocks(dn, count);
1272 if (ret)
1273 return ret;
1274
1275 dn->ofs_in_node = ofs_in_node;
1276 for (index = start; index < end; index++, dn->ofs_in_node++) {
1277 dn->data_blkaddr = datablock_addr(dn->inode,
1278 dn->node_page, dn->ofs_in_node);
1279 /*
1280 * f2fs_reserve_new_blocks will not guarantee entire block
1281 * allocation.
1282 */
1283 if (dn->data_blkaddr == NULL_ADDR) {
1284 ret = -ENOSPC;
1285 break;
1286 }
1287 if (dn->data_blkaddr != NEW_ADDR) {
1288 f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1289 dn->data_blkaddr = NEW_ADDR;
1290 f2fs_set_data_blkaddr(dn);
1291 }
1292 }
1293
1294 f2fs_update_extent_cache_range(dn, start, 0, index - start);
1295
1296 return ret;
1297 }
1298
f2fs_zero_range(struct inode * inode,loff_t offset,loff_t len,int mode)1299 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1300 int mode)
1301 {
1302 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1303 struct address_space *mapping = inode->i_mapping;
1304 pgoff_t index, pg_start, pg_end;
1305 loff_t new_size = i_size_read(inode);
1306 loff_t off_start, off_end;
1307 int ret = 0;
1308
1309 ret = inode_newsize_ok(inode, (len + offset));
1310 if (ret)
1311 return ret;
1312
1313 ret = f2fs_convert_inline_inode(inode);
1314 if (ret)
1315 return ret;
1316
1317 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1318 if (ret)
1319 return ret;
1320
1321 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1322 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1323
1324 off_start = offset & (PAGE_SIZE - 1);
1325 off_end = (offset + len) & (PAGE_SIZE - 1);
1326
1327 if (pg_start == pg_end) {
1328 ret = fill_zero(inode, pg_start, off_start,
1329 off_end - off_start);
1330 if (ret)
1331 return ret;
1332
1333 new_size = max_t(loff_t, new_size, offset + len);
1334 } else {
1335 if (off_start) {
1336 ret = fill_zero(inode, pg_start++, off_start,
1337 PAGE_SIZE - off_start);
1338 if (ret)
1339 return ret;
1340
1341 new_size = max_t(loff_t, new_size,
1342 (loff_t)pg_start << PAGE_SHIFT);
1343 }
1344
1345 for (index = pg_start; index < pg_end;) {
1346 struct dnode_of_data dn;
1347 unsigned int end_offset;
1348 pgoff_t end;
1349
1350 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1351 down_write(&F2FS_I(inode)->i_mmap_sem);
1352
1353 truncate_pagecache_range(inode,
1354 (loff_t)index << PAGE_SHIFT,
1355 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1356
1357 f2fs_lock_op(sbi);
1358
1359 set_new_dnode(&dn, inode, NULL, NULL, 0);
1360 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1361 if (ret) {
1362 f2fs_unlock_op(sbi);
1363 up_write(&F2FS_I(inode)->i_mmap_sem);
1364 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1365 goto out;
1366 }
1367
1368 end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1369 end = min(pg_end, end_offset - dn.ofs_in_node + index);
1370
1371 ret = f2fs_do_zero_range(&dn, index, end);
1372 f2fs_put_dnode(&dn);
1373
1374 f2fs_unlock_op(sbi);
1375 up_write(&F2FS_I(inode)->i_mmap_sem);
1376 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1377
1378 f2fs_balance_fs(sbi, dn.node_changed);
1379
1380 if (ret)
1381 goto out;
1382
1383 index = end;
1384 new_size = max_t(loff_t, new_size,
1385 (loff_t)index << PAGE_SHIFT);
1386 }
1387
1388 if (off_end) {
1389 ret = fill_zero(inode, pg_end, 0, off_end);
1390 if (ret)
1391 goto out;
1392
1393 new_size = max_t(loff_t, new_size, offset + len);
1394 }
1395 }
1396
1397 out:
1398 if (new_size > i_size_read(inode)) {
1399 if (mode & FALLOC_FL_KEEP_SIZE)
1400 file_set_keep_isize(inode);
1401 else
1402 f2fs_i_size_write(inode, new_size);
1403 }
1404 return ret;
1405 }
1406
f2fs_insert_range(struct inode * inode,loff_t offset,loff_t len)1407 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1408 {
1409 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1410 pgoff_t nr, pg_start, pg_end, delta, idx;
1411 loff_t new_size;
1412 int ret = 0;
1413
1414 new_size = i_size_read(inode) + len;
1415 ret = inode_newsize_ok(inode, new_size);
1416 if (ret)
1417 return ret;
1418
1419 if (offset >= i_size_read(inode))
1420 return -EINVAL;
1421
1422 /* insert range should be aligned to block size of f2fs. */
1423 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1424 return -EINVAL;
1425
1426 ret = f2fs_convert_inline_inode(inode);
1427 if (ret)
1428 return ret;
1429
1430 f2fs_balance_fs(sbi, true);
1431
1432 down_write(&F2FS_I(inode)->i_mmap_sem);
1433 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1434 up_write(&F2FS_I(inode)->i_mmap_sem);
1435 if (ret)
1436 return ret;
1437
1438 /* write out all dirty pages from offset */
1439 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1440 if (ret)
1441 return ret;
1442
1443 pg_start = offset >> PAGE_SHIFT;
1444 pg_end = (offset + len) >> PAGE_SHIFT;
1445 delta = pg_end - pg_start;
1446 idx = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
1447
1448 /* avoid gc operation during block exchange */
1449 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1450 down_write(&F2FS_I(inode)->i_mmap_sem);
1451 truncate_pagecache(inode, offset);
1452
1453 while (!ret && idx > pg_start) {
1454 nr = idx - pg_start;
1455 if (nr > delta)
1456 nr = delta;
1457 idx -= nr;
1458
1459 f2fs_lock_op(sbi);
1460 f2fs_drop_extent_tree(inode);
1461
1462 ret = __exchange_data_block(inode, inode, idx,
1463 idx + delta, nr, false);
1464 f2fs_unlock_op(sbi);
1465 }
1466 up_write(&F2FS_I(inode)->i_mmap_sem);
1467 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1468
1469 /* write out all moved pages, if possible */
1470 down_write(&F2FS_I(inode)->i_mmap_sem);
1471 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1472 truncate_pagecache(inode, offset);
1473 up_write(&F2FS_I(inode)->i_mmap_sem);
1474
1475 if (!ret)
1476 f2fs_i_size_write(inode, new_size);
1477 return ret;
1478 }
1479
expand_inode_data(struct inode * inode,loff_t offset,loff_t len,int mode)1480 static int expand_inode_data(struct inode *inode, loff_t offset,
1481 loff_t len, int mode)
1482 {
1483 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1484 struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1485 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE };
1486 pgoff_t pg_end;
1487 loff_t new_size = i_size_read(inode);
1488 loff_t off_end;
1489 int err;
1490
1491 err = inode_newsize_ok(inode, (len + offset));
1492 if (err)
1493 return err;
1494
1495 err = f2fs_convert_inline_inode(inode);
1496 if (err)
1497 return err;
1498
1499 f2fs_balance_fs(sbi, true);
1500
1501 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1502 off_end = (offset + len) & (PAGE_SIZE - 1);
1503
1504 map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1505 map.m_len = pg_end - map.m_lblk;
1506 if (off_end)
1507 map.m_len++;
1508
1509 err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1510 if (err) {
1511 pgoff_t last_off;
1512
1513 if (!map.m_len)
1514 return err;
1515
1516 last_off = map.m_lblk + map.m_len - 1;
1517
1518 /* update new size to the failed position */
1519 new_size = (last_off == pg_end) ? offset + len :
1520 (loff_t)(last_off + 1) << PAGE_SHIFT;
1521 } else {
1522 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1523 }
1524
1525 if (new_size > i_size_read(inode)) {
1526 if (mode & FALLOC_FL_KEEP_SIZE)
1527 file_set_keep_isize(inode);
1528 else
1529 f2fs_i_size_write(inode, new_size);
1530 }
1531
1532 return err;
1533 }
1534
f2fs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)1535 static long f2fs_fallocate(struct file *file, int mode,
1536 loff_t offset, loff_t len)
1537 {
1538 struct inode *inode = file_inode(file);
1539 long ret = 0;
1540
1541 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1542 return -EIO;
1543
1544 /* f2fs only support ->fallocate for regular file */
1545 if (!S_ISREG(inode->i_mode))
1546 return -EINVAL;
1547
1548 if (f2fs_encrypted_inode(inode) &&
1549 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1550 return -EOPNOTSUPP;
1551
1552 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1553 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1554 FALLOC_FL_INSERT_RANGE))
1555 return -EOPNOTSUPP;
1556
1557 inode_lock(inode);
1558
1559 if (mode & FALLOC_FL_PUNCH_HOLE) {
1560 if (offset >= inode->i_size)
1561 goto out;
1562
1563 ret = punch_hole(inode, offset, len);
1564 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1565 ret = f2fs_collapse_range(inode, offset, len);
1566 } else if (mode & FALLOC_FL_ZERO_RANGE) {
1567 ret = f2fs_zero_range(inode, offset, len, mode);
1568 } else if (mode & FALLOC_FL_INSERT_RANGE) {
1569 ret = f2fs_insert_range(inode, offset, len);
1570 } else {
1571 ret = expand_inode_data(inode, offset, len, mode);
1572 }
1573
1574 if (!ret) {
1575 inode->i_mtime = inode->i_ctime = current_time(inode);
1576 f2fs_mark_inode_dirty_sync(inode, false);
1577 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1578 }
1579
1580 out:
1581 inode_unlock(inode);
1582
1583 trace_f2fs_fallocate(inode, mode, offset, len, ret);
1584 return ret;
1585 }
1586
f2fs_release_file(struct inode * inode,struct file * filp)1587 static int f2fs_release_file(struct inode *inode, struct file *filp)
1588 {
1589 /*
1590 * f2fs_relase_file is called at every close calls. So we should
1591 * not drop any inmemory pages by close called by other process.
1592 */
1593 if (!(filp->f_mode & FMODE_WRITE) ||
1594 atomic_read(&inode->i_writecount) != 1)
1595 return 0;
1596
1597 /* some remained atomic pages should discarded */
1598 if (f2fs_is_atomic_file(inode))
1599 f2fs_drop_inmem_pages(inode);
1600 if (f2fs_is_volatile_file(inode)) {
1601 set_inode_flag(inode, FI_DROP_CACHE);
1602 filemap_fdatawrite(inode->i_mapping);
1603 clear_inode_flag(inode, FI_DROP_CACHE);
1604 clear_inode_flag(inode, FI_VOLATILE_FILE);
1605 stat_dec_volatile_write(inode);
1606 }
1607 return 0;
1608 }
1609
f2fs_file_flush(struct file * file,fl_owner_t id)1610 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1611 {
1612 struct inode *inode = file_inode(file);
1613
1614 /*
1615 * If the process doing a transaction is crashed, we should do
1616 * roll-back. Otherwise, other reader/write can see corrupted database
1617 * until all the writers close its file. Since this should be done
1618 * before dropping file lock, it needs to do in ->flush.
1619 */
1620 if (f2fs_is_atomic_file(inode) &&
1621 F2FS_I(inode)->inmem_task == current)
1622 f2fs_drop_inmem_pages(inode);
1623 return 0;
1624 }
1625
f2fs_ioc_getflags(struct file * filp,unsigned long arg)1626 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1627 {
1628 struct inode *inode = file_inode(filp);
1629 struct f2fs_inode_info *fi = F2FS_I(inode);
1630 unsigned int flags = fi->i_flags;
1631
1632 if (f2fs_encrypted_inode(inode))
1633 flags |= F2FS_ENCRYPT_FL;
1634 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1635 flags |= F2FS_INLINE_DATA_FL;
1636
1637 flags &= F2FS_FL_USER_VISIBLE;
1638
1639 return put_user(flags, (int __user *)arg);
1640 }
1641
__f2fs_ioc_setflags(struct inode * inode,unsigned int flags)1642 static int __f2fs_ioc_setflags(struct inode *inode, unsigned int flags)
1643 {
1644 struct f2fs_inode_info *fi = F2FS_I(inode);
1645 unsigned int oldflags;
1646
1647 /* Is it quota file? Do not allow user to mess with it */
1648 if (IS_NOQUOTA(inode))
1649 return -EPERM;
1650
1651 flags = f2fs_mask_flags(inode->i_mode, flags);
1652
1653 oldflags = fi->i_flags;
1654
1655 if ((flags ^ oldflags) & (F2FS_APPEND_FL | F2FS_IMMUTABLE_FL))
1656 if (!capable(CAP_LINUX_IMMUTABLE))
1657 return -EPERM;
1658
1659 flags = flags & F2FS_FL_USER_MODIFIABLE;
1660 flags |= oldflags & ~F2FS_FL_USER_MODIFIABLE;
1661 fi->i_flags = flags;
1662
1663 if (fi->i_flags & F2FS_PROJINHERIT_FL)
1664 set_inode_flag(inode, FI_PROJ_INHERIT);
1665 else
1666 clear_inode_flag(inode, FI_PROJ_INHERIT);
1667
1668 inode->i_ctime = current_time(inode);
1669 f2fs_set_inode_flags(inode);
1670 f2fs_mark_inode_dirty_sync(inode, true);
1671 return 0;
1672 }
1673
f2fs_ioc_setflags(struct file * filp,unsigned long arg)1674 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1675 {
1676 struct inode *inode = file_inode(filp);
1677 unsigned int flags;
1678 int ret;
1679
1680 if (!inode_owner_or_capable(inode))
1681 return -EACCES;
1682
1683 if (get_user(flags, (int __user *)arg))
1684 return -EFAULT;
1685
1686 ret = mnt_want_write_file(filp);
1687 if (ret)
1688 return ret;
1689
1690 inode_lock(inode);
1691
1692 ret = __f2fs_ioc_setflags(inode, flags);
1693
1694 inode_unlock(inode);
1695 mnt_drop_write_file(filp);
1696 return ret;
1697 }
1698
f2fs_ioc_getversion(struct file * filp,unsigned long arg)1699 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1700 {
1701 struct inode *inode = file_inode(filp);
1702
1703 return put_user(inode->i_generation, (int __user *)arg);
1704 }
1705
f2fs_ioc_start_atomic_write(struct file * filp)1706 static int f2fs_ioc_start_atomic_write(struct file *filp)
1707 {
1708 struct inode *inode = file_inode(filp);
1709 int ret;
1710
1711 if (!inode_owner_or_capable(inode))
1712 return -EACCES;
1713
1714 if (!S_ISREG(inode->i_mode))
1715 return -EINVAL;
1716
1717 ret = mnt_want_write_file(filp);
1718 if (ret)
1719 return ret;
1720
1721 inode_lock(inode);
1722
1723 if (f2fs_is_atomic_file(inode)) {
1724 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
1725 ret = -EINVAL;
1726 goto out;
1727 }
1728
1729 ret = f2fs_convert_inline_inode(inode);
1730 if (ret)
1731 goto out;
1732
1733 down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1734
1735 /*
1736 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
1737 * f2fs_is_atomic_file.
1738 */
1739 if (get_dirty_pages(inode))
1740 f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
1741 "Unexpected flush for atomic writes: ino=%lu, npages=%u",
1742 inode->i_ino, get_dirty_pages(inode));
1743 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
1744 if (ret) {
1745 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1746 goto out;
1747 }
1748
1749 set_inode_flag(inode, FI_ATOMIC_FILE);
1750 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1751 up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1752
1753 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1754 F2FS_I(inode)->inmem_task = current;
1755 stat_inc_atomic_write(inode);
1756 stat_update_max_atomic_write(inode);
1757 out:
1758 inode_unlock(inode);
1759 mnt_drop_write_file(filp);
1760 return ret;
1761 }
1762
f2fs_ioc_commit_atomic_write(struct file * filp)1763 static int f2fs_ioc_commit_atomic_write(struct file *filp)
1764 {
1765 struct inode *inode = file_inode(filp);
1766 int ret;
1767
1768 if (!inode_owner_or_capable(inode))
1769 return -EACCES;
1770
1771 ret = mnt_want_write_file(filp);
1772 if (ret)
1773 return ret;
1774
1775 f2fs_balance_fs(F2FS_I_SB(inode), true);
1776
1777 inode_lock(inode);
1778
1779 if (f2fs_is_volatile_file(inode)) {
1780 ret = -EINVAL;
1781 goto err_out;
1782 }
1783
1784 if (f2fs_is_atomic_file(inode)) {
1785 ret = f2fs_commit_inmem_pages(inode);
1786 if (ret)
1787 goto err_out;
1788
1789 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1790 if (!ret) {
1791 clear_inode_flag(inode, FI_ATOMIC_FILE);
1792 F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC] = 0;
1793 stat_dec_atomic_write(inode);
1794 }
1795 } else {
1796 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
1797 }
1798 err_out:
1799 if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
1800 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1801 ret = -EINVAL;
1802 }
1803 inode_unlock(inode);
1804 mnt_drop_write_file(filp);
1805 return ret;
1806 }
1807
f2fs_ioc_start_volatile_write(struct file * filp)1808 static int f2fs_ioc_start_volatile_write(struct file *filp)
1809 {
1810 struct inode *inode = file_inode(filp);
1811 int ret;
1812
1813 if (!inode_owner_or_capable(inode))
1814 return -EACCES;
1815
1816 if (!S_ISREG(inode->i_mode))
1817 return -EINVAL;
1818
1819 ret = mnt_want_write_file(filp);
1820 if (ret)
1821 return ret;
1822
1823 inode_lock(inode);
1824
1825 if (f2fs_is_volatile_file(inode))
1826 goto out;
1827
1828 ret = f2fs_convert_inline_inode(inode);
1829 if (ret)
1830 goto out;
1831
1832 stat_inc_volatile_write(inode);
1833 stat_update_max_volatile_write(inode);
1834
1835 set_inode_flag(inode, FI_VOLATILE_FILE);
1836 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1837 out:
1838 inode_unlock(inode);
1839 mnt_drop_write_file(filp);
1840 return ret;
1841 }
1842
f2fs_ioc_release_volatile_write(struct file * filp)1843 static int f2fs_ioc_release_volatile_write(struct file *filp)
1844 {
1845 struct inode *inode = file_inode(filp);
1846 int ret;
1847
1848 if (!inode_owner_or_capable(inode))
1849 return -EACCES;
1850
1851 ret = mnt_want_write_file(filp);
1852 if (ret)
1853 return ret;
1854
1855 inode_lock(inode);
1856
1857 if (!f2fs_is_volatile_file(inode))
1858 goto out;
1859
1860 if (!f2fs_is_first_block_written(inode)) {
1861 ret = truncate_partial_data_page(inode, 0, true);
1862 goto out;
1863 }
1864
1865 ret = punch_hole(inode, 0, F2FS_BLKSIZE);
1866 out:
1867 inode_unlock(inode);
1868 mnt_drop_write_file(filp);
1869 return ret;
1870 }
1871
f2fs_ioc_abort_volatile_write(struct file * filp)1872 static int f2fs_ioc_abort_volatile_write(struct file *filp)
1873 {
1874 struct inode *inode = file_inode(filp);
1875 int ret;
1876
1877 if (!inode_owner_or_capable(inode))
1878 return -EACCES;
1879
1880 ret = mnt_want_write_file(filp);
1881 if (ret)
1882 return ret;
1883
1884 inode_lock(inode);
1885
1886 if (f2fs_is_atomic_file(inode))
1887 f2fs_drop_inmem_pages(inode);
1888 if (f2fs_is_volatile_file(inode)) {
1889 clear_inode_flag(inode, FI_VOLATILE_FILE);
1890 stat_dec_volatile_write(inode);
1891 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
1892 }
1893
1894 clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
1895
1896 inode_unlock(inode);
1897
1898 mnt_drop_write_file(filp);
1899 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1900 return ret;
1901 }
1902
f2fs_ioc_shutdown(struct file * filp,unsigned long arg)1903 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
1904 {
1905 struct inode *inode = file_inode(filp);
1906 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1907 struct super_block *sb = sbi->sb;
1908 __u32 in;
1909 int ret = 0;
1910
1911 if (!capable(CAP_SYS_ADMIN))
1912 return -EPERM;
1913
1914 if (get_user(in, (__u32 __user *)arg))
1915 return -EFAULT;
1916
1917 if (in != F2FS_GOING_DOWN_FULLSYNC) {
1918 ret = mnt_want_write_file(filp);
1919 if (ret)
1920 return ret;
1921 }
1922
1923 switch (in) {
1924 case F2FS_GOING_DOWN_FULLSYNC:
1925 sb = freeze_bdev(sb->s_bdev);
1926 if (IS_ERR(sb)) {
1927 ret = PTR_ERR(sb);
1928 goto out;
1929 }
1930 if (sb) {
1931 f2fs_stop_checkpoint(sbi, false);
1932 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1933 thaw_bdev(sb->s_bdev, sb);
1934 }
1935 break;
1936 case F2FS_GOING_DOWN_METASYNC:
1937 /* do checkpoint only */
1938 ret = f2fs_sync_fs(sb, 1);
1939 if (ret)
1940 goto out;
1941 f2fs_stop_checkpoint(sbi, false);
1942 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1943 break;
1944 case F2FS_GOING_DOWN_NOSYNC:
1945 f2fs_stop_checkpoint(sbi, false);
1946 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1947 break;
1948 case F2FS_GOING_DOWN_METAFLUSH:
1949 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
1950 f2fs_stop_checkpoint(sbi, false);
1951 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
1952 break;
1953 default:
1954 ret = -EINVAL;
1955 goto out;
1956 }
1957
1958 f2fs_stop_gc_thread(sbi);
1959 f2fs_stop_discard_thread(sbi);
1960
1961 f2fs_drop_discard_cmd(sbi);
1962 clear_opt(sbi, DISCARD);
1963
1964 f2fs_update_time(sbi, REQ_TIME);
1965 out:
1966 if (in != F2FS_GOING_DOWN_FULLSYNC)
1967 mnt_drop_write_file(filp);
1968 return ret;
1969 }
1970
f2fs_ioc_fitrim(struct file * filp,unsigned long arg)1971 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
1972 {
1973 struct inode *inode = file_inode(filp);
1974 struct super_block *sb = inode->i_sb;
1975 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1976 struct fstrim_range range;
1977 int ret;
1978
1979 if (!capable(CAP_SYS_ADMIN))
1980 return -EPERM;
1981
1982 if (!f2fs_hw_support_discard(F2FS_SB(sb)))
1983 return -EOPNOTSUPP;
1984
1985 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1986 sizeof(range)))
1987 return -EFAULT;
1988
1989 ret = mnt_want_write_file(filp);
1990 if (ret)
1991 return ret;
1992
1993 range.minlen = max((unsigned int)range.minlen,
1994 q->limits.discard_granularity);
1995 ret = f2fs_trim_fs(F2FS_SB(sb), &range);
1996 mnt_drop_write_file(filp);
1997 if (ret < 0)
1998 return ret;
1999
2000 if (copy_to_user((struct fstrim_range __user *)arg, &range,
2001 sizeof(range)))
2002 return -EFAULT;
2003 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2004 return 0;
2005 }
2006
uuid_is_nonzero(__u8 u[16])2007 static bool uuid_is_nonzero(__u8 u[16])
2008 {
2009 int i;
2010
2011 for (i = 0; i < 16; i++)
2012 if (u[i])
2013 return true;
2014 return false;
2015 }
2016
f2fs_ioc_set_encryption_policy(struct file * filp,unsigned long arg)2017 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2018 {
2019 struct inode *inode = file_inode(filp);
2020
2021 if (!f2fs_sb_has_encrypt(inode->i_sb))
2022 return -EOPNOTSUPP;
2023
2024 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2025
2026 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2027 }
2028
f2fs_ioc_get_encryption_policy(struct file * filp,unsigned long arg)2029 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2030 {
2031 if (!f2fs_sb_has_encrypt(file_inode(filp)->i_sb))
2032 return -EOPNOTSUPP;
2033 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2034 }
2035
f2fs_ioc_get_encryption_pwsalt(struct file * filp,unsigned long arg)2036 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2037 {
2038 struct inode *inode = file_inode(filp);
2039 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2040 int err;
2041
2042 if (!f2fs_sb_has_encrypt(inode->i_sb))
2043 return -EOPNOTSUPP;
2044
2045 err = mnt_want_write_file(filp);
2046 if (err)
2047 return err;
2048
2049 down_write(&sbi->sb_lock);
2050
2051 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2052 goto got_it;
2053
2054 /* update superblock with uuid */
2055 generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2056
2057 err = f2fs_commit_super(sbi, false);
2058 if (err) {
2059 /* undo new data */
2060 memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2061 goto out_err;
2062 }
2063 got_it:
2064 if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2065 16))
2066 err = -EFAULT;
2067 out_err:
2068 up_write(&sbi->sb_lock);
2069 mnt_drop_write_file(filp);
2070 return err;
2071 }
2072
f2fs_ioc_gc(struct file * filp,unsigned long arg)2073 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2074 {
2075 struct inode *inode = file_inode(filp);
2076 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2077 __u32 sync;
2078 int ret;
2079
2080 if (!capable(CAP_SYS_ADMIN))
2081 return -EPERM;
2082
2083 if (get_user(sync, (__u32 __user *)arg))
2084 return -EFAULT;
2085
2086 if (f2fs_readonly(sbi->sb))
2087 return -EROFS;
2088
2089 ret = mnt_want_write_file(filp);
2090 if (ret)
2091 return ret;
2092
2093 if (!sync) {
2094 if (!mutex_trylock(&sbi->gc_mutex)) {
2095 ret = -EBUSY;
2096 goto out;
2097 }
2098 } else {
2099 mutex_lock(&sbi->gc_mutex);
2100 }
2101
2102 ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2103 out:
2104 mnt_drop_write_file(filp);
2105 return ret;
2106 }
2107
f2fs_ioc_gc_range(struct file * filp,unsigned long arg)2108 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2109 {
2110 struct inode *inode = file_inode(filp);
2111 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2112 struct f2fs_gc_range range;
2113 u64 end;
2114 int ret;
2115
2116 if (!capable(CAP_SYS_ADMIN))
2117 return -EPERM;
2118
2119 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2120 sizeof(range)))
2121 return -EFAULT;
2122
2123 if (f2fs_readonly(sbi->sb))
2124 return -EROFS;
2125
2126 end = range.start + range.len;
2127 if (range.start < MAIN_BLKADDR(sbi) || end >= MAX_BLKADDR(sbi)) {
2128 return -EINVAL;
2129 }
2130
2131 ret = mnt_want_write_file(filp);
2132 if (ret)
2133 return ret;
2134
2135 do_more:
2136 if (!range.sync) {
2137 if (!mutex_trylock(&sbi->gc_mutex)) {
2138 ret = -EBUSY;
2139 goto out;
2140 }
2141 } else {
2142 mutex_lock(&sbi->gc_mutex);
2143 }
2144
2145 ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2146 range.start += BLKS_PER_SEC(sbi);
2147 if (range.start <= end)
2148 goto do_more;
2149 out:
2150 mnt_drop_write_file(filp);
2151 return ret;
2152 }
2153
f2fs_ioc_write_checkpoint(struct file * filp,unsigned long arg)2154 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2155 {
2156 struct inode *inode = file_inode(filp);
2157 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2158 int ret;
2159
2160 if (!capable(CAP_SYS_ADMIN))
2161 return -EPERM;
2162
2163 if (f2fs_readonly(sbi->sb))
2164 return -EROFS;
2165
2166 ret = mnt_want_write_file(filp);
2167 if (ret)
2168 return ret;
2169
2170 ret = f2fs_sync_fs(sbi->sb, 1);
2171
2172 mnt_drop_write_file(filp);
2173 return ret;
2174 }
2175
f2fs_defragment_range(struct f2fs_sb_info * sbi,struct file * filp,struct f2fs_defragment * range)2176 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2177 struct file *filp,
2178 struct f2fs_defragment *range)
2179 {
2180 struct inode *inode = file_inode(filp);
2181 struct f2fs_map_blocks map = { .m_next_extent = NULL,
2182 .m_seg_type = NO_CHECK_TYPE };
2183 struct extent_info ei = {0, 0, 0};
2184 pgoff_t pg_start, pg_end, next_pgofs;
2185 unsigned int blk_per_seg = sbi->blocks_per_seg;
2186 unsigned int total = 0, sec_num;
2187 block_t blk_end = 0;
2188 bool fragmented = false;
2189 int err;
2190
2191 /* if in-place-update policy is enabled, don't waste time here */
2192 if (f2fs_should_update_inplace(inode, NULL))
2193 return -EINVAL;
2194
2195 pg_start = range->start >> PAGE_SHIFT;
2196 pg_end = (range->start + range->len) >> PAGE_SHIFT;
2197
2198 f2fs_balance_fs(sbi, true);
2199
2200 inode_lock(inode);
2201
2202 /* writeback all dirty pages in the range */
2203 err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2204 range->start + range->len - 1);
2205 if (err)
2206 goto out;
2207
2208 /*
2209 * lookup mapping info in extent cache, skip defragmenting if physical
2210 * block addresses are continuous.
2211 */
2212 if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2213 if (ei.fofs + ei.len >= pg_end)
2214 goto out;
2215 }
2216
2217 map.m_lblk = pg_start;
2218 map.m_next_pgofs = &next_pgofs;
2219
2220 /*
2221 * lookup mapping info in dnode page cache, skip defragmenting if all
2222 * physical block addresses are continuous even if there are hole(s)
2223 * in logical blocks.
2224 */
2225 while (map.m_lblk < pg_end) {
2226 map.m_len = pg_end - map.m_lblk;
2227 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2228 if (err)
2229 goto out;
2230
2231 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2232 map.m_lblk = next_pgofs;
2233 continue;
2234 }
2235
2236 if (blk_end && blk_end != map.m_pblk)
2237 fragmented = true;
2238
2239 /* record total count of block that we're going to move */
2240 total += map.m_len;
2241
2242 blk_end = map.m_pblk + map.m_len;
2243
2244 map.m_lblk += map.m_len;
2245 }
2246
2247 if (!fragmented)
2248 goto out;
2249
2250 sec_num = (total + BLKS_PER_SEC(sbi) - 1) / BLKS_PER_SEC(sbi);
2251
2252 /*
2253 * make sure there are enough free section for LFS allocation, this can
2254 * avoid defragment running in SSR mode when free section are allocated
2255 * intensively
2256 */
2257 if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2258 err = -EAGAIN;
2259 goto out;
2260 }
2261
2262 map.m_lblk = pg_start;
2263 map.m_len = pg_end - pg_start;
2264 total = 0;
2265
2266 while (map.m_lblk < pg_end) {
2267 pgoff_t idx;
2268 int cnt = 0;
2269
2270 do_map:
2271 map.m_len = pg_end - map.m_lblk;
2272 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2273 if (err)
2274 goto clear_out;
2275
2276 if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2277 map.m_lblk = next_pgofs;
2278 continue;
2279 }
2280
2281 set_inode_flag(inode, FI_DO_DEFRAG);
2282
2283 idx = map.m_lblk;
2284 while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2285 struct page *page;
2286
2287 page = f2fs_get_lock_data_page(inode, idx, true);
2288 if (IS_ERR(page)) {
2289 err = PTR_ERR(page);
2290 goto clear_out;
2291 }
2292
2293 set_page_dirty(page);
2294 f2fs_put_page(page, 1);
2295
2296 idx++;
2297 cnt++;
2298 total++;
2299 }
2300
2301 map.m_lblk = idx;
2302
2303 if (idx < pg_end && cnt < blk_per_seg)
2304 goto do_map;
2305
2306 clear_inode_flag(inode, FI_DO_DEFRAG);
2307
2308 err = filemap_fdatawrite(inode->i_mapping);
2309 if (err)
2310 goto out;
2311 }
2312 clear_out:
2313 clear_inode_flag(inode, FI_DO_DEFRAG);
2314 out:
2315 inode_unlock(inode);
2316 if (!err)
2317 range->len = (u64)total << PAGE_SHIFT;
2318 return err;
2319 }
2320
f2fs_ioc_defragment(struct file * filp,unsigned long arg)2321 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2322 {
2323 struct inode *inode = file_inode(filp);
2324 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2325 struct f2fs_defragment range;
2326 int err;
2327
2328 if (!capable(CAP_SYS_ADMIN))
2329 return -EPERM;
2330
2331 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2332 return -EINVAL;
2333
2334 if (f2fs_readonly(sbi->sb))
2335 return -EROFS;
2336
2337 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2338 sizeof(range)))
2339 return -EFAULT;
2340
2341 /* verify alignment of offset & size */
2342 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2343 return -EINVAL;
2344
2345 if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2346 sbi->max_file_blocks))
2347 return -EINVAL;
2348
2349 err = mnt_want_write_file(filp);
2350 if (err)
2351 return err;
2352
2353 err = f2fs_defragment_range(sbi, filp, &range);
2354 mnt_drop_write_file(filp);
2355
2356 f2fs_update_time(sbi, REQ_TIME);
2357 if (err < 0)
2358 return err;
2359
2360 if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2361 sizeof(range)))
2362 return -EFAULT;
2363
2364 return 0;
2365 }
2366
f2fs_move_file_range(struct file * file_in,loff_t pos_in,struct file * file_out,loff_t pos_out,size_t len)2367 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2368 struct file *file_out, loff_t pos_out, size_t len)
2369 {
2370 struct inode *src = file_inode(file_in);
2371 struct inode *dst = file_inode(file_out);
2372 struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2373 size_t olen = len, dst_max_i_size = 0;
2374 size_t dst_osize;
2375 int ret;
2376
2377 if (file_in->f_path.mnt != file_out->f_path.mnt ||
2378 src->i_sb != dst->i_sb)
2379 return -EXDEV;
2380
2381 if (unlikely(f2fs_readonly(src->i_sb)))
2382 return -EROFS;
2383
2384 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2385 return -EINVAL;
2386
2387 if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
2388 return -EOPNOTSUPP;
2389
2390 if (src == dst) {
2391 if (pos_in == pos_out)
2392 return 0;
2393 if (pos_out > pos_in && pos_out < pos_in + len)
2394 return -EINVAL;
2395 }
2396
2397 inode_lock(src);
2398 if (src != dst) {
2399 ret = -EBUSY;
2400 if (!inode_trylock(dst))
2401 goto out;
2402 }
2403
2404 ret = -EINVAL;
2405 if (pos_in + len > src->i_size || pos_in + len < pos_in)
2406 goto out_unlock;
2407 if (len == 0)
2408 olen = len = src->i_size - pos_in;
2409 if (pos_in + len == src->i_size)
2410 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2411 if (len == 0) {
2412 ret = 0;
2413 goto out_unlock;
2414 }
2415
2416 dst_osize = dst->i_size;
2417 if (pos_out + olen > dst->i_size)
2418 dst_max_i_size = pos_out + olen;
2419
2420 /* verify the end result is block aligned */
2421 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2422 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2423 !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2424 goto out_unlock;
2425
2426 ret = f2fs_convert_inline_inode(src);
2427 if (ret)
2428 goto out_unlock;
2429
2430 ret = f2fs_convert_inline_inode(dst);
2431 if (ret)
2432 goto out_unlock;
2433
2434 /* write out all dirty pages from offset */
2435 ret = filemap_write_and_wait_range(src->i_mapping,
2436 pos_in, pos_in + len);
2437 if (ret)
2438 goto out_unlock;
2439
2440 ret = filemap_write_and_wait_range(dst->i_mapping,
2441 pos_out, pos_out + len);
2442 if (ret)
2443 goto out_unlock;
2444
2445 f2fs_balance_fs(sbi, true);
2446
2447 down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2448 if (src != dst) {
2449 ret = -EBUSY;
2450 if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2451 goto out_src;
2452 }
2453
2454 f2fs_lock_op(sbi);
2455 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2456 pos_out >> F2FS_BLKSIZE_BITS,
2457 len >> F2FS_BLKSIZE_BITS, false);
2458
2459 if (!ret) {
2460 if (dst_max_i_size)
2461 f2fs_i_size_write(dst, dst_max_i_size);
2462 else if (dst_osize != dst->i_size)
2463 f2fs_i_size_write(dst, dst_osize);
2464 }
2465 f2fs_unlock_op(sbi);
2466
2467 if (src != dst)
2468 up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2469 out_src:
2470 up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2471 out_unlock:
2472 if (src != dst)
2473 inode_unlock(dst);
2474 out:
2475 inode_unlock(src);
2476 return ret;
2477 }
2478
f2fs_ioc_move_range(struct file * filp,unsigned long arg)2479 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2480 {
2481 struct f2fs_move_range range;
2482 struct fd dst;
2483 int err;
2484
2485 if (!(filp->f_mode & FMODE_READ) ||
2486 !(filp->f_mode & FMODE_WRITE))
2487 return -EBADF;
2488
2489 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2490 sizeof(range)))
2491 return -EFAULT;
2492
2493 dst = fdget(range.dst_fd);
2494 if (!dst.file)
2495 return -EBADF;
2496
2497 if (!(dst.file->f_mode & FMODE_WRITE)) {
2498 err = -EBADF;
2499 goto err_out;
2500 }
2501
2502 err = mnt_want_write_file(filp);
2503 if (err)
2504 goto err_out;
2505
2506 err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2507 range.pos_out, range.len);
2508
2509 mnt_drop_write_file(filp);
2510 if (err)
2511 goto err_out;
2512
2513 if (copy_to_user((struct f2fs_move_range __user *)arg,
2514 &range, sizeof(range)))
2515 err = -EFAULT;
2516 err_out:
2517 fdput(dst);
2518 return err;
2519 }
2520
f2fs_ioc_flush_device(struct file * filp,unsigned long arg)2521 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2522 {
2523 struct inode *inode = file_inode(filp);
2524 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2525 struct sit_info *sm = SIT_I(sbi);
2526 unsigned int start_segno = 0, end_segno = 0;
2527 unsigned int dev_start_segno = 0, dev_end_segno = 0;
2528 struct f2fs_flush_device range;
2529 int ret;
2530
2531 if (!capable(CAP_SYS_ADMIN))
2532 return -EPERM;
2533
2534 if (f2fs_readonly(sbi->sb))
2535 return -EROFS;
2536
2537 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2538 sizeof(range)))
2539 return -EFAULT;
2540
2541 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2542 sbi->segs_per_sec != 1) {
2543 f2fs_msg(sbi->sb, KERN_WARNING,
2544 "Can't flush %u in %d for segs_per_sec %u != 1\n",
2545 range.dev_num, sbi->s_ndevs,
2546 sbi->segs_per_sec);
2547 return -EINVAL;
2548 }
2549
2550 ret = mnt_want_write_file(filp);
2551 if (ret)
2552 return ret;
2553
2554 if (range.dev_num != 0)
2555 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2556 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2557
2558 start_segno = sm->last_victim[FLUSH_DEVICE];
2559 if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2560 start_segno = dev_start_segno;
2561 end_segno = min(start_segno + range.segments, dev_end_segno);
2562
2563 while (start_segno < end_segno) {
2564 if (!mutex_trylock(&sbi->gc_mutex)) {
2565 ret = -EBUSY;
2566 goto out;
2567 }
2568 sm->last_victim[GC_CB] = end_segno + 1;
2569 sm->last_victim[GC_GREEDY] = end_segno + 1;
2570 sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2571 ret = f2fs_gc(sbi, true, true, start_segno);
2572 if (ret == -EAGAIN)
2573 ret = 0;
2574 else if (ret < 0)
2575 break;
2576 start_segno++;
2577 }
2578 out:
2579 mnt_drop_write_file(filp);
2580 return ret;
2581 }
2582
f2fs_ioc_get_features(struct file * filp,unsigned long arg)2583 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2584 {
2585 struct inode *inode = file_inode(filp);
2586 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2587
2588 /* Must validate to set it with SQLite behavior in Android. */
2589 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2590
2591 return put_user(sb_feature, (u32 __user *)arg);
2592 }
2593
2594 #ifdef CONFIG_QUOTA
f2fs_ioc_setproject(struct file * filp,__u32 projid)2595 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2596 {
2597 struct inode *inode = file_inode(filp);
2598 struct f2fs_inode_info *fi = F2FS_I(inode);
2599 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2600 struct super_block *sb = sbi->sb;
2601 struct dquot *transfer_to[MAXQUOTAS] = {};
2602 struct page *ipage;
2603 kprojid_t kprojid;
2604 int err;
2605
2606 if (!f2fs_sb_has_project_quota(sb)) {
2607 if (projid != F2FS_DEF_PROJID)
2608 return -EOPNOTSUPP;
2609 else
2610 return 0;
2611 }
2612
2613 if (!f2fs_has_extra_attr(inode))
2614 return -EOPNOTSUPP;
2615
2616 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2617
2618 if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2619 return 0;
2620
2621 err = -EPERM;
2622 /* Is it quota file? Do not allow user to mess with it */
2623 if (IS_NOQUOTA(inode))
2624 return err;
2625
2626 ipage = f2fs_get_node_page(sbi, inode->i_ino);
2627 if (IS_ERR(ipage))
2628 return PTR_ERR(ipage);
2629
2630 if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
2631 i_projid)) {
2632 err = -EOVERFLOW;
2633 f2fs_put_page(ipage, 1);
2634 return err;
2635 }
2636 f2fs_put_page(ipage, 1);
2637
2638 err = dquot_initialize(inode);
2639 if (err)
2640 return err;
2641
2642 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2643 if (!IS_ERR(transfer_to[PRJQUOTA])) {
2644 err = __dquot_transfer(inode, transfer_to);
2645 dqput(transfer_to[PRJQUOTA]);
2646 if (err)
2647 goto out_dirty;
2648 }
2649
2650 F2FS_I(inode)->i_projid = kprojid;
2651 inode->i_ctime = current_time(inode);
2652 out_dirty:
2653 f2fs_mark_inode_dirty_sync(inode, true);
2654 return err;
2655 }
2656 #else
f2fs_ioc_setproject(struct file * filp,__u32 projid)2657 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2658 {
2659 if (projid != F2FS_DEF_PROJID)
2660 return -EOPNOTSUPP;
2661 return 0;
2662 }
2663 #endif
2664
2665 /* Transfer internal flags to xflags */
f2fs_iflags_to_xflags(unsigned long iflags)2666 static inline __u32 f2fs_iflags_to_xflags(unsigned long iflags)
2667 {
2668 __u32 xflags = 0;
2669
2670 if (iflags & F2FS_SYNC_FL)
2671 xflags |= FS_XFLAG_SYNC;
2672 if (iflags & F2FS_IMMUTABLE_FL)
2673 xflags |= FS_XFLAG_IMMUTABLE;
2674 if (iflags & F2FS_APPEND_FL)
2675 xflags |= FS_XFLAG_APPEND;
2676 if (iflags & F2FS_NODUMP_FL)
2677 xflags |= FS_XFLAG_NODUMP;
2678 if (iflags & F2FS_NOATIME_FL)
2679 xflags |= FS_XFLAG_NOATIME;
2680 if (iflags & F2FS_PROJINHERIT_FL)
2681 xflags |= FS_XFLAG_PROJINHERIT;
2682 return xflags;
2683 }
2684
2685 #define F2FS_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
2686 FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
2687 FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
2688
2689 /* Transfer xflags flags to internal */
f2fs_xflags_to_iflags(__u32 xflags)2690 static inline unsigned long f2fs_xflags_to_iflags(__u32 xflags)
2691 {
2692 unsigned long iflags = 0;
2693
2694 if (xflags & FS_XFLAG_SYNC)
2695 iflags |= F2FS_SYNC_FL;
2696 if (xflags & FS_XFLAG_IMMUTABLE)
2697 iflags |= F2FS_IMMUTABLE_FL;
2698 if (xflags & FS_XFLAG_APPEND)
2699 iflags |= F2FS_APPEND_FL;
2700 if (xflags & FS_XFLAG_NODUMP)
2701 iflags |= F2FS_NODUMP_FL;
2702 if (xflags & FS_XFLAG_NOATIME)
2703 iflags |= F2FS_NOATIME_FL;
2704 if (xflags & FS_XFLAG_PROJINHERIT)
2705 iflags |= F2FS_PROJINHERIT_FL;
2706
2707 return iflags;
2708 }
2709
f2fs_ioc_fsgetxattr(struct file * filp,unsigned long arg)2710 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
2711 {
2712 struct inode *inode = file_inode(filp);
2713 struct f2fs_inode_info *fi = F2FS_I(inode);
2714 struct fsxattr fa;
2715
2716 memset(&fa, 0, sizeof(struct fsxattr));
2717 fa.fsx_xflags = f2fs_iflags_to_xflags(fi->i_flags &
2718 F2FS_FL_USER_VISIBLE);
2719
2720 if (f2fs_sb_has_project_quota(inode->i_sb))
2721 fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
2722 fi->i_projid);
2723
2724 if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
2725 return -EFAULT;
2726 return 0;
2727 }
2728
f2fs_ioctl_check_project(struct inode * inode,struct fsxattr * fa)2729 static int f2fs_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
2730 {
2731 /*
2732 * Project Quota ID state is only allowed to change from within the init
2733 * namespace. Enforce that restriction only if we are trying to change
2734 * the quota ID state. Everything else is allowed in user namespaces.
2735 */
2736 if (current_user_ns() == &init_user_ns)
2737 return 0;
2738
2739 if (__kprojid_val(F2FS_I(inode)->i_projid) != fa->fsx_projid)
2740 return -EINVAL;
2741
2742 if (F2FS_I(inode)->i_flags & F2FS_PROJINHERIT_FL) {
2743 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
2744 return -EINVAL;
2745 } else {
2746 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
2747 return -EINVAL;
2748 }
2749
2750 return 0;
2751 }
2752
f2fs_ioc_fssetxattr(struct file * filp,unsigned long arg)2753 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
2754 {
2755 struct inode *inode = file_inode(filp);
2756 struct f2fs_inode_info *fi = F2FS_I(inode);
2757 struct fsxattr fa;
2758 unsigned int flags;
2759 int err;
2760
2761 if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
2762 return -EFAULT;
2763
2764 /* Make sure caller has proper permission */
2765 if (!inode_owner_or_capable(inode))
2766 return -EACCES;
2767
2768 if (fa.fsx_xflags & ~F2FS_SUPPORTED_FS_XFLAGS)
2769 return -EOPNOTSUPP;
2770
2771 flags = f2fs_xflags_to_iflags(fa.fsx_xflags);
2772 if (f2fs_mask_flags(inode->i_mode, flags) != flags)
2773 return -EOPNOTSUPP;
2774
2775 err = mnt_want_write_file(filp);
2776 if (err)
2777 return err;
2778
2779 inode_lock(inode);
2780 err = f2fs_ioctl_check_project(inode, &fa);
2781 if (err)
2782 goto out;
2783 flags = (fi->i_flags & ~F2FS_FL_XFLAG_VISIBLE) |
2784 (flags & F2FS_FL_XFLAG_VISIBLE);
2785 err = __f2fs_ioc_setflags(inode, flags);
2786 if (err)
2787 goto out;
2788
2789 err = f2fs_ioc_setproject(filp, fa.fsx_projid);
2790 out:
2791 inode_unlock(inode);
2792 mnt_drop_write_file(filp);
2793 return err;
2794 }
2795
f2fs_pin_file_control(struct inode * inode,bool inc)2796 int f2fs_pin_file_control(struct inode *inode, bool inc)
2797 {
2798 struct f2fs_inode_info *fi = F2FS_I(inode);
2799 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2800
2801 /* Use i_gc_failures for normal file as a risk signal. */
2802 if (inc)
2803 f2fs_i_gc_failures_write(inode,
2804 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
2805
2806 if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
2807 f2fs_msg(sbi->sb, KERN_WARNING,
2808 "%s: Enable GC = ino %lx after %x GC trials\n",
2809 __func__, inode->i_ino,
2810 fi->i_gc_failures[GC_FAILURE_PIN]);
2811 clear_inode_flag(inode, FI_PIN_FILE);
2812 return -EAGAIN;
2813 }
2814 return 0;
2815 }
2816
f2fs_ioc_set_pin_file(struct file * filp,unsigned long arg)2817 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
2818 {
2819 struct inode *inode = file_inode(filp);
2820 __u32 pin;
2821 int ret = 0;
2822
2823 if (!inode_owner_or_capable(inode))
2824 return -EACCES;
2825
2826 if (get_user(pin, (__u32 __user *)arg))
2827 return -EFAULT;
2828
2829 if (!S_ISREG(inode->i_mode))
2830 return -EINVAL;
2831
2832 if (f2fs_readonly(F2FS_I_SB(inode)->sb))
2833 return -EROFS;
2834
2835 ret = mnt_want_write_file(filp);
2836 if (ret)
2837 return ret;
2838
2839 inode_lock(inode);
2840
2841 if (f2fs_should_update_outplace(inode, NULL)) {
2842 ret = -EINVAL;
2843 goto out;
2844 }
2845
2846 if (!pin) {
2847 clear_inode_flag(inode, FI_PIN_FILE);
2848 f2fs_i_gc_failures_write(inode, 0);
2849 goto done;
2850 }
2851
2852 if (f2fs_pin_file_control(inode, false)) {
2853 ret = -EAGAIN;
2854 goto out;
2855 }
2856 ret = f2fs_convert_inline_inode(inode);
2857 if (ret)
2858 goto out;
2859
2860 set_inode_flag(inode, FI_PIN_FILE);
2861 ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2862 done:
2863 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2864 out:
2865 inode_unlock(inode);
2866 mnt_drop_write_file(filp);
2867 return ret;
2868 }
2869
f2fs_ioc_get_pin_file(struct file * filp,unsigned long arg)2870 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
2871 {
2872 struct inode *inode = file_inode(filp);
2873 __u32 pin = 0;
2874
2875 if (is_inode_flag_set(inode, FI_PIN_FILE))
2876 pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
2877 return put_user(pin, (u32 __user *)arg);
2878 }
2879
f2fs_precache_extents(struct inode * inode)2880 int f2fs_precache_extents(struct inode *inode)
2881 {
2882 struct f2fs_inode_info *fi = F2FS_I(inode);
2883 struct f2fs_map_blocks map;
2884 pgoff_t m_next_extent;
2885 loff_t end;
2886 int err;
2887
2888 if (is_inode_flag_set(inode, FI_NO_EXTENT))
2889 return -EOPNOTSUPP;
2890
2891 map.m_lblk = 0;
2892 map.m_next_pgofs = NULL;
2893 map.m_next_extent = &m_next_extent;
2894 map.m_seg_type = NO_CHECK_TYPE;
2895 end = F2FS_I_SB(inode)->max_file_blocks;
2896
2897 while (map.m_lblk < end) {
2898 map.m_len = end - map.m_lblk;
2899
2900 down_write(&fi->i_gc_rwsem[WRITE]);
2901 err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
2902 up_write(&fi->i_gc_rwsem[WRITE]);
2903 if (err)
2904 return err;
2905
2906 map.m_lblk = m_next_extent;
2907 }
2908
2909 return err;
2910 }
2911
f2fs_ioc_precache_extents(struct file * filp,unsigned long arg)2912 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
2913 {
2914 return f2fs_precache_extents(file_inode(filp));
2915 }
2916
f2fs_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)2917 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
2918 {
2919 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
2920 return -EIO;
2921
2922 switch (cmd) {
2923 case F2FS_IOC_GETFLAGS:
2924 return f2fs_ioc_getflags(filp, arg);
2925 case F2FS_IOC_SETFLAGS:
2926 return f2fs_ioc_setflags(filp, arg);
2927 case F2FS_IOC_GETVERSION:
2928 return f2fs_ioc_getversion(filp, arg);
2929 case F2FS_IOC_START_ATOMIC_WRITE:
2930 return f2fs_ioc_start_atomic_write(filp);
2931 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
2932 return f2fs_ioc_commit_atomic_write(filp);
2933 case F2FS_IOC_START_VOLATILE_WRITE:
2934 return f2fs_ioc_start_volatile_write(filp);
2935 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
2936 return f2fs_ioc_release_volatile_write(filp);
2937 case F2FS_IOC_ABORT_VOLATILE_WRITE:
2938 return f2fs_ioc_abort_volatile_write(filp);
2939 case F2FS_IOC_SHUTDOWN:
2940 return f2fs_ioc_shutdown(filp, arg);
2941 case FITRIM:
2942 return f2fs_ioc_fitrim(filp, arg);
2943 case F2FS_IOC_SET_ENCRYPTION_POLICY:
2944 return f2fs_ioc_set_encryption_policy(filp, arg);
2945 case F2FS_IOC_GET_ENCRYPTION_POLICY:
2946 return f2fs_ioc_get_encryption_policy(filp, arg);
2947 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
2948 return f2fs_ioc_get_encryption_pwsalt(filp, arg);
2949 case F2FS_IOC_GARBAGE_COLLECT:
2950 return f2fs_ioc_gc(filp, arg);
2951 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
2952 return f2fs_ioc_gc_range(filp, arg);
2953 case F2FS_IOC_WRITE_CHECKPOINT:
2954 return f2fs_ioc_write_checkpoint(filp, arg);
2955 case F2FS_IOC_DEFRAGMENT:
2956 return f2fs_ioc_defragment(filp, arg);
2957 case F2FS_IOC_MOVE_RANGE:
2958 return f2fs_ioc_move_range(filp, arg);
2959 case F2FS_IOC_FLUSH_DEVICE:
2960 return f2fs_ioc_flush_device(filp, arg);
2961 case F2FS_IOC_GET_FEATURES:
2962 return f2fs_ioc_get_features(filp, arg);
2963 case F2FS_IOC_FSGETXATTR:
2964 return f2fs_ioc_fsgetxattr(filp, arg);
2965 case F2FS_IOC_FSSETXATTR:
2966 return f2fs_ioc_fssetxattr(filp, arg);
2967 case F2FS_IOC_GET_PIN_FILE:
2968 return f2fs_ioc_get_pin_file(filp, arg);
2969 case F2FS_IOC_SET_PIN_FILE:
2970 return f2fs_ioc_set_pin_file(filp, arg);
2971 case F2FS_IOC_PRECACHE_EXTENTS:
2972 return f2fs_ioc_precache_extents(filp, arg);
2973 default:
2974 return -ENOTTY;
2975 }
2976 }
2977
f2fs_file_write_iter(struct kiocb * iocb,struct iov_iter * from)2978 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
2979 {
2980 struct file *file = iocb->ki_filp;
2981 struct inode *inode = file_inode(file);
2982 ssize_t ret;
2983
2984 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
2985 return -EIO;
2986
2987 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
2988 return -EINVAL;
2989
2990 if (!inode_trylock(inode)) {
2991 if (iocb->ki_flags & IOCB_NOWAIT)
2992 return -EAGAIN;
2993 inode_lock(inode);
2994 }
2995
2996 ret = generic_write_checks(iocb, from);
2997 if (ret > 0) {
2998 bool preallocated = false;
2999 size_t target_size = 0;
3000 int err;
3001
3002 if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3003 set_inode_flag(inode, FI_NO_PREALLOC);
3004
3005 if ((iocb->ki_flags & IOCB_NOWAIT) &&
3006 (iocb->ki_flags & IOCB_DIRECT)) {
3007 if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3008 iov_iter_count(from)) ||
3009 f2fs_has_inline_data(inode) ||
3010 f2fs_force_buffered_io(inode, WRITE)) {
3011 clear_inode_flag(inode,
3012 FI_NO_PREALLOC);
3013 inode_unlock(inode);
3014 return -EAGAIN;
3015 }
3016
3017 } else {
3018 preallocated = true;
3019 target_size = iocb->ki_pos + iov_iter_count(from);
3020
3021 err = f2fs_preallocate_blocks(iocb, from);
3022 if (err) {
3023 clear_inode_flag(inode, FI_NO_PREALLOC);
3024 inode_unlock(inode);
3025 return err;
3026 }
3027 }
3028 ret = __generic_file_write_iter(iocb, from);
3029 clear_inode_flag(inode, FI_NO_PREALLOC);
3030
3031 /* if we couldn't write data, we should deallocate blocks. */
3032 if (preallocated && i_size_read(inode) < target_size)
3033 f2fs_truncate(inode);
3034
3035 if (ret > 0)
3036 f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3037 }
3038 inode_unlock(inode);
3039
3040 if (ret > 0)
3041 ret = generic_write_sync(iocb, ret);
3042 return ret;
3043 }
3044
3045 #ifdef CONFIG_COMPAT
f2fs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)3046 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3047 {
3048 switch (cmd) {
3049 case F2FS_IOC32_GETFLAGS:
3050 cmd = F2FS_IOC_GETFLAGS;
3051 break;
3052 case F2FS_IOC32_SETFLAGS:
3053 cmd = F2FS_IOC_SETFLAGS;
3054 break;
3055 case F2FS_IOC32_GETVERSION:
3056 cmd = F2FS_IOC_GETVERSION;
3057 break;
3058 case F2FS_IOC_START_ATOMIC_WRITE:
3059 case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3060 case F2FS_IOC_START_VOLATILE_WRITE:
3061 case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3062 case F2FS_IOC_ABORT_VOLATILE_WRITE:
3063 case F2FS_IOC_SHUTDOWN:
3064 case F2FS_IOC_SET_ENCRYPTION_POLICY:
3065 case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3066 case F2FS_IOC_GET_ENCRYPTION_POLICY:
3067 case F2FS_IOC_GARBAGE_COLLECT:
3068 case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3069 case F2FS_IOC_WRITE_CHECKPOINT:
3070 case F2FS_IOC_DEFRAGMENT:
3071 case F2FS_IOC_MOVE_RANGE:
3072 case F2FS_IOC_FLUSH_DEVICE:
3073 case F2FS_IOC_GET_FEATURES:
3074 case F2FS_IOC_FSGETXATTR:
3075 case F2FS_IOC_FSSETXATTR:
3076 case F2FS_IOC_GET_PIN_FILE:
3077 case F2FS_IOC_SET_PIN_FILE:
3078 case F2FS_IOC_PRECACHE_EXTENTS:
3079 break;
3080 default:
3081 return -ENOIOCTLCMD;
3082 }
3083 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3084 }
3085 #endif
3086
3087 const struct file_operations f2fs_file_operations = {
3088 .llseek = f2fs_llseek,
3089 .read_iter = generic_file_read_iter,
3090 .write_iter = f2fs_file_write_iter,
3091 .open = f2fs_file_open,
3092 .release = f2fs_release_file,
3093 .mmap = f2fs_file_mmap,
3094 .flush = f2fs_file_flush,
3095 .fsync = f2fs_sync_file,
3096 .fallocate = f2fs_fallocate,
3097 .unlocked_ioctl = f2fs_ioctl,
3098 #ifdef CONFIG_COMPAT
3099 .compat_ioctl = f2fs_compat_ioctl,
3100 #endif
3101 .splice_read = generic_file_splice_read,
3102 .splice_write = iter_file_splice_write,
3103 };
3104