]>
Commit | Line | Data |
---|---|---|
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 | ||
24 | #include "f2fs.h" | |
25 | #include "node.h" | |
26 | #include "segment.h" | |
27 | #include "xattr.h" | |
28 | #include "acl.h" | |
29 | #include "trace.h" | |
30 | #include <trace/events/f2fs.h> | |
31 | ||
32 | static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma, | |
33 | struct vm_fault *vmf) | |
34 | { | |
35 | struct page *page = vmf->page; | |
36 | struct inode *inode = file_inode(vma->vm_file); | |
37 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
38 | struct dnode_of_data dn; | |
39 | int err; | |
40 | ||
41 | f2fs_balance_fs(sbi); | |
42 | ||
43 | sb_start_pagefault(inode->i_sb); | |
44 | ||
45 | f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); | |
46 | ||
47 | /* block allocation */ | |
48 | f2fs_lock_op(sbi); | |
49 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
50 | err = f2fs_reserve_block(&dn, page->index); | |
51 | if (err) { | |
52 | f2fs_unlock_op(sbi); | |
53 | goto out; | |
54 | } | |
55 | f2fs_put_dnode(&dn); | |
56 | f2fs_unlock_op(sbi); | |
57 | ||
58 | file_update_time(vma->vm_file); | |
59 | lock_page(page); | |
60 | if (unlikely(page->mapping != inode->i_mapping || | |
61 | page_offset(page) > i_size_read(inode) || | |
62 | !PageUptodate(page))) { | |
63 | unlock_page(page); | |
64 | err = -EFAULT; | |
65 | goto out; | |
66 | } | |
67 | ||
68 | /* | |
69 | * check to see if the page is mapped already (no holes) | |
70 | */ | |
71 | if (PageMappedToDisk(page)) | |
72 | goto mapped; | |
73 | ||
74 | /* page is wholly or partially inside EOF */ | |
75 | if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) { | |
76 | unsigned offset; | |
77 | offset = i_size_read(inode) & ~PAGE_CACHE_MASK; | |
78 | zero_user_segment(page, offset, PAGE_CACHE_SIZE); | |
79 | } | |
80 | set_page_dirty(page); | |
81 | SetPageUptodate(page); | |
82 | ||
83 | trace_f2fs_vm_page_mkwrite(page, DATA); | |
84 | mapped: | |
85 | /* fill the page */ | |
86 | f2fs_wait_on_page_writeback(page, DATA); | |
87 | out: | |
88 | sb_end_pagefault(inode->i_sb); | |
89 | return block_page_mkwrite_return(err); | |
90 | } | |
91 | ||
92 | static const struct vm_operations_struct f2fs_file_vm_ops = { | |
93 | .fault = filemap_fault, | |
94 | .map_pages = filemap_map_pages, | |
95 | .page_mkwrite = f2fs_vm_page_mkwrite, | |
96 | .remap_pages = generic_file_remap_pages, | |
97 | }; | |
98 | ||
99 | static int get_parent_ino(struct inode *inode, nid_t *pino) | |
100 | { | |
101 | struct dentry *dentry; | |
102 | ||
103 | inode = igrab(inode); | |
104 | dentry = d_find_any_alias(inode); | |
105 | iput(inode); | |
106 | if (!dentry) | |
107 | return 0; | |
108 | ||
109 | if (update_dent_inode(inode, &dentry->d_name)) { | |
110 | dput(dentry); | |
111 | return 0; | |
112 | } | |
113 | ||
114 | *pino = parent_ino(dentry); | |
115 | dput(dentry); | |
116 | return 1; | |
117 | } | |
118 | ||
119 | static inline bool need_do_checkpoint(struct inode *inode) | |
120 | { | |
121 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
122 | bool need_cp = false; | |
123 | ||
124 | if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1) | |
125 | need_cp = true; | |
126 | else if (file_wrong_pino(inode)) | |
127 | need_cp = true; | |
128 | else if (!space_for_roll_forward(sbi)) | |
129 | need_cp = true; | |
130 | else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) | |
131 | need_cp = true; | |
132 | else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi))) | |
133 | need_cp = true; | |
134 | else if (test_opt(sbi, FASTBOOT)) | |
135 | need_cp = true; | |
136 | else if (sbi->active_logs == 2) | |
137 | need_cp = true; | |
138 | ||
139 | return need_cp; | |
140 | } | |
141 | ||
142 | static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) | |
143 | { | |
144 | struct page *i = find_get_page(NODE_MAPPING(sbi), ino); | |
145 | bool ret = false; | |
146 | /* But we need to avoid that there are some inode updates */ | |
147 | if ((i && PageDirty(i)) || need_inode_block_update(sbi, ino)) | |
148 | ret = true; | |
149 | f2fs_put_page(i, 0); | |
150 | return ret; | |
151 | } | |
152 | ||
153 | static void try_to_fix_pino(struct inode *inode) | |
154 | { | |
155 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
156 | nid_t pino; | |
157 | ||
158 | down_write(&fi->i_sem); | |
159 | fi->xattr_ver = 0; | |
160 | if (file_wrong_pino(inode) && inode->i_nlink == 1 && | |
161 | get_parent_ino(inode, &pino)) { | |
162 | fi->i_pino = pino; | |
163 | file_got_pino(inode); | |
164 | up_write(&fi->i_sem); | |
165 | ||
166 | mark_inode_dirty_sync(inode); | |
167 | f2fs_write_inode(inode, NULL); | |
168 | } else { | |
169 | up_write(&fi->i_sem); | |
170 | } | |
171 | } | |
172 | ||
173 | int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) | |
174 | { | |
175 | struct inode *inode = file->f_mapping->host; | |
176 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
177 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
178 | nid_t ino = inode->i_ino; | |
179 | int ret = 0; | |
180 | bool need_cp = false; | |
181 | struct writeback_control wbc = { | |
182 | .sync_mode = WB_SYNC_ALL, | |
183 | .nr_to_write = LONG_MAX, | |
184 | .for_reclaim = 0, | |
185 | }; | |
186 | ||
187 | if (unlikely(f2fs_readonly(inode->i_sb))) | |
188 | return 0; | |
189 | ||
190 | trace_f2fs_sync_file_enter(inode); | |
191 | ||
192 | /* if fdatasync is triggered, let's do in-place-update */ | |
193 | if (get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) | |
194 | set_inode_flag(fi, FI_NEED_IPU); | |
195 | ret = filemap_write_and_wait_range(inode->i_mapping, start, end); | |
196 | clear_inode_flag(fi, FI_NEED_IPU); | |
197 | ||
198 | if (ret) { | |
199 | trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret); | |
200 | return ret; | |
201 | } | |
202 | ||
203 | /* if the inode is dirty, let's recover all the time */ | |
204 | if (!datasync && is_inode_flag_set(fi, FI_DIRTY_INODE)) { | |
205 | update_inode_page(inode); | |
206 | goto go_write; | |
207 | } | |
208 | ||
209 | /* | |
210 | * if there is no written data, don't waste time to write recovery info. | |
211 | */ | |
212 | if (!is_inode_flag_set(fi, FI_APPEND_WRITE) && | |
213 | !exist_written_data(sbi, ino, APPEND_INO)) { | |
214 | ||
215 | /* it may call write_inode just prior to fsync */ | |
216 | if (need_inode_page_update(sbi, ino)) | |
217 | goto go_write; | |
218 | ||
219 | if (is_inode_flag_set(fi, FI_UPDATE_WRITE) || | |
220 | exist_written_data(sbi, ino, UPDATE_INO)) | |
221 | goto flush_out; | |
222 | goto out; | |
223 | } | |
224 | go_write: | |
225 | /* guarantee free sections for fsync */ | |
226 | f2fs_balance_fs(sbi); | |
227 | ||
228 | /* | |
229 | * Both of fdatasync() and fsync() are able to be recovered from | |
230 | * sudden-power-off. | |
231 | */ | |
232 | down_read(&fi->i_sem); | |
233 | need_cp = need_do_checkpoint(inode); | |
234 | up_read(&fi->i_sem); | |
235 | ||
236 | if (need_cp) { | |
237 | /* all the dirty node pages should be flushed for POR */ | |
238 | ret = f2fs_sync_fs(inode->i_sb, 1); | |
239 | ||
240 | /* | |
241 | * We've secured consistency through sync_fs. Following pino | |
242 | * will be used only for fsynced inodes after checkpoint. | |
243 | */ | |
244 | try_to_fix_pino(inode); | |
245 | goto out; | |
246 | } | |
247 | sync_nodes: | |
248 | sync_node_pages(sbi, ino, &wbc); | |
249 | ||
250 | /* if cp_error was enabled, we should avoid infinite loop */ | |
251 | if (unlikely(f2fs_cp_error(sbi))) | |
252 | goto out; | |
253 | ||
254 | if (need_inode_block_update(sbi, ino)) { | |
255 | mark_inode_dirty_sync(inode); | |
256 | f2fs_write_inode(inode, NULL); | |
257 | goto sync_nodes; | |
258 | } | |
259 | ||
260 | ret = wait_on_node_pages_writeback(sbi, ino); | |
261 | if (ret) | |
262 | goto out; | |
263 | ||
264 | /* once recovery info is written, don't need to tack this */ | |
265 | remove_dirty_inode(sbi, ino, APPEND_INO); | |
266 | clear_inode_flag(fi, FI_APPEND_WRITE); | |
267 | flush_out: | |
268 | remove_dirty_inode(sbi, ino, UPDATE_INO); | |
269 | clear_inode_flag(fi, FI_UPDATE_WRITE); | |
270 | ret = f2fs_issue_flush(sbi); | |
271 | out: | |
272 | trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret); | |
273 | f2fs_trace_ios(NULL, NULL, 1); | |
274 | return ret; | |
275 | } | |
276 | ||
277 | static pgoff_t __get_first_dirty_index(struct address_space *mapping, | |
278 | pgoff_t pgofs, int whence) | |
279 | { | |
280 | struct pagevec pvec; | |
281 | int nr_pages; | |
282 | ||
283 | if (whence != SEEK_DATA) | |
284 | return 0; | |
285 | ||
286 | /* find first dirty page index */ | |
287 | pagevec_init(&pvec, 0); | |
288 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs, | |
289 | PAGECACHE_TAG_DIRTY, 1); | |
290 | pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX; | |
291 | pagevec_release(&pvec); | |
292 | return pgofs; | |
293 | } | |
294 | ||
295 | static bool __found_offset(block_t blkaddr, pgoff_t dirty, pgoff_t pgofs, | |
296 | int whence) | |
297 | { | |
298 | switch (whence) { | |
299 | case SEEK_DATA: | |
300 | if ((blkaddr == NEW_ADDR && dirty == pgofs) || | |
301 | (blkaddr != NEW_ADDR && blkaddr != NULL_ADDR)) | |
302 | return true; | |
303 | break; | |
304 | case SEEK_HOLE: | |
305 | if (blkaddr == NULL_ADDR) | |
306 | return true; | |
307 | break; | |
308 | } | |
309 | return false; | |
310 | } | |
311 | ||
312 | static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) | |
313 | { | |
314 | struct inode *inode = file->f_mapping->host; | |
315 | loff_t maxbytes = inode->i_sb->s_maxbytes; | |
316 | struct dnode_of_data dn; | |
317 | pgoff_t pgofs, end_offset, dirty; | |
318 | loff_t data_ofs = offset; | |
319 | loff_t isize; | |
320 | int err = 0; | |
321 | ||
322 | mutex_lock(&inode->i_mutex); | |
323 | ||
324 | isize = i_size_read(inode); | |
325 | if (offset >= isize) | |
326 | goto fail; | |
327 | ||
328 | /* handle inline data case */ | |
329 | if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) { | |
330 | if (whence == SEEK_HOLE) | |
331 | data_ofs = isize; | |
332 | goto found; | |
333 | } | |
334 | ||
335 | pgofs = (pgoff_t)(offset >> PAGE_CACHE_SHIFT); | |
336 | ||
337 | dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence); | |
338 | ||
339 | for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) { | |
340 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
341 | err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA); | |
342 | if (err && err != -ENOENT) { | |
343 | goto fail; | |
344 | } else if (err == -ENOENT) { | |
345 | /* direct node does not exists */ | |
346 | if (whence == SEEK_DATA) { | |
347 | pgofs = PGOFS_OF_NEXT_DNODE(pgofs, | |
348 | F2FS_I(inode)); | |
349 | continue; | |
350 | } else { | |
351 | goto found; | |
352 | } | |
353 | } | |
354 | ||
355 | end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode)); | |
356 | ||
357 | /* find data/hole in dnode block */ | |
358 | for (; dn.ofs_in_node < end_offset; | |
359 | dn.ofs_in_node++, pgofs++, | |
360 | data_ofs = pgofs << PAGE_CACHE_SHIFT) { | |
361 | block_t blkaddr; | |
362 | blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
363 | ||
364 | if (__found_offset(blkaddr, dirty, pgofs, whence)) { | |
365 | f2fs_put_dnode(&dn); | |
366 | goto found; | |
367 | } | |
368 | } | |
369 | f2fs_put_dnode(&dn); | |
370 | } | |
371 | ||
372 | if (whence == SEEK_DATA) | |
373 | goto fail; | |
374 | found: | |
375 | if (whence == SEEK_HOLE && data_ofs > isize) | |
376 | data_ofs = isize; | |
377 | mutex_unlock(&inode->i_mutex); | |
378 | return vfs_setpos(file, data_ofs, maxbytes); | |
379 | fail: | |
380 | mutex_unlock(&inode->i_mutex); | |
381 | return -ENXIO; | |
382 | } | |
383 | ||
384 | static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) | |
385 | { | |
386 | struct inode *inode = file->f_mapping->host; | |
387 | loff_t maxbytes = inode->i_sb->s_maxbytes; | |
388 | ||
389 | switch (whence) { | |
390 | case SEEK_SET: | |
391 | case SEEK_CUR: | |
392 | case SEEK_END: | |
393 | return generic_file_llseek_size(file, offset, whence, | |
394 | maxbytes, i_size_read(inode)); | |
395 | case SEEK_DATA: | |
396 | case SEEK_HOLE: | |
397 | if (offset < 0) | |
398 | return -ENXIO; | |
399 | return f2fs_seek_block(file, offset, whence); | |
400 | } | |
401 | ||
402 | return -EINVAL; | |
403 | } | |
404 | ||
405 | static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) | |
406 | { | |
407 | struct inode *inode = file_inode(file); | |
408 | ||
409 | /* we don't need to use inline_data strictly */ | |
410 | if (f2fs_has_inline_data(inode)) { | |
411 | int err = f2fs_convert_inline_inode(inode); | |
412 | if (err) | |
413 | return err; | |
414 | } | |
415 | ||
416 | file_accessed(file); | |
417 | vma->vm_ops = &f2fs_file_vm_ops; | |
418 | return 0; | |
419 | } | |
420 | ||
421 | int truncate_data_blocks_range(struct dnode_of_data *dn, int count) | |
422 | { | |
423 | int nr_free = 0, ofs = dn->ofs_in_node; | |
424 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); | |
425 | struct f2fs_node *raw_node; | |
426 | __le32 *addr; | |
427 | ||
428 | raw_node = F2FS_NODE(dn->node_page); | |
429 | addr = blkaddr_in_node(raw_node) + ofs; | |
430 | ||
431 | for (; count > 0; count--, addr++, dn->ofs_in_node++) { | |
432 | block_t blkaddr = le32_to_cpu(*addr); | |
433 | if (blkaddr == NULL_ADDR) | |
434 | continue; | |
435 | ||
436 | dn->data_blkaddr = NULL_ADDR; | |
437 | update_extent_cache(dn); | |
438 | invalidate_blocks(sbi, blkaddr); | |
439 | nr_free++; | |
440 | } | |
441 | if (nr_free) { | |
442 | dec_valid_block_count(sbi, dn->inode, nr_free); | |
443 | set_page_dirty(dn->node_page); | |
444 | sync_inode_page(dn); | |
445 | } | |
446 | dn->ofs_in_node = ofs; | |
447 | ||
448 | trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, | |
449 | dn->ofs_in_node, nr_free); | |
450 | return nr_free; | |
451 | } | |
452 | ||
453 | void truncate_data_blocks(struct dnode_of_data *dn) | |
454 | { | |
455 | truncate_data_blocks_range(dn, ADDRS_PER_BLOCK); | |
456 | } | |
457 | ||
458 | static int truncate_partial_data_page(struct inode *inode, u64 from) | |
459 | { | |
460 | unsigned offset = from & (PAGE_CACHE_SIZE - 1); | |
461 | struct page *page; | |
462 | ||
463 | if (!offset) | |
464 | return 0; | |
465 | ||
466 | page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, false); | |
467 | if (IS_ERR(page)) | |
468 | return 0; | |
469 | ||
470 | lock_page(page); | |
471 | if (unlikely(!PageUptodate(page) || | |
472 | page->mapping != inode->i_mapping)) | |
473 | goto out; | |
474 | ||
475 | f2fs_wait_on_page_writeback(page, DATA); | |
476 | zero_user(page, offset, PAGE_CACHE_SIZE - offset); | |
477 | set_page_dirty(page); | |
478 | out: | |
479 | f2fs_put_page(page, 1); | |
480 | return 0; | |
481 | } | |
482 | ||
483 | int truncate_blocks(struct inode *inode, u64 from, bool lock) | |
484 | { | |
485 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
486 | unsigned int blocksize = inode->i_sb->s_blocksize; | |
487 | struct dnode_of_data dn; | |
488 | pgoff_t free_from; | |
489 | int count = 0, err = 0; | |
490 | struct page *ipage; | |
491 | ||
492 | trace_f2fs_truncate_blocks_enter(inode, from); | |
493 | ||
494 | free_from = (pgoff_t) | |
495 | ((from + blocksize - 1) >> (sbi->log_blocksize)); | |
496 | ||
497 | if (lock) | |
498 | f2fs_lock_op(sbi); | |
499 | ||
500 | ipage = get_node_page(sbi, inode->i_ino); | |
501 | if (IS_ERR(ipage)) { | |
502 | err = PTR_ERR(ipage); | |
503 | goto out; | |
504 | } | |
505 | ||
506 | if (f2fs_has_inline_data(inode)) { | |
507 | f2fs_put_page(ipage, 1); | |
508 | goto out; | |
509 | } | |
510 | ||
511 | set_new_dnode(&dn, inode, ipage, NULL, 0); | |
512 | err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE); | |
513 | if (err) { | |
514 | if (err == -ENOENT) | |
515 | goto free_next; | |
516 | goto out; | |
517 | } | |
518 | ||
519 | count = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode)); | |
520 | ||
521 | count -= dn.ofs_in_node; | |
522 | f2fs_bug_on(sbi, count < 0); | |
523 | ||
524 | if (dn.ofs_in_node || IS_INODE(dn.node_page)) { | |
525 | truncate_data_blocks_range(&dn, count); | |
526 | free_from += count; | |
527 | } | |
528 | ||
529 | f2fs_put_dnode(&dn); | |
530 | free_next: | |
531 | err = truncate_inode_blocks(inode, free_from); | |
532 | out: | |
533 | if (lock) | |
534 | f2fs_unlock_op(sbi); | |
535 | ||
536 | /* lastly zero out the first data page */ | |
537 | if (!err) | |
538 | err = truncate_partial_data_page(inode, from); | |
539 | ||
540 | trace_f2fs_truncate_blocks_exit(inode, err); | |
541 | return err; | |
542 | } | |
543 | ||
544 | void f2fs_truncate(struct inode *inode) | |
545 | { | |
546 | if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
547 | S_ISLNK(inode->i_mode))) | |
548 | return; | |
549 | ||
550 | trace_f2fs_truncate(inode); | |
551 | ||
552 | /* we should check inline_data size */ | |
553 | if (f2fs_has_inline_data(inode) && !f2fs_may_inline(inode)) { | |
554 | if (f2fs_convert_inline_inode(inode)) | |
555 | return; | |
556 | } | |
557 | ||
558 | if (!truncate_blocks(inode, i_size_read(inode), true)) { | |
559 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
560 | mark_inode_dirty(inode); | |
561 | } | |
562 | } | |
563 | ||
564 | int f2fs_getattr(struct vfsmount *mnt, | |
565 | struct dentry *dentry, struct kstat *stat) | |
566 | { | |
567 | struct inode *inode = dentry->d_inode; | |
568 | generic_fillattr(inode, stat); | |
569 | stat->blocks <<= 3; | |
570 | return 0; | |
571 | } | |
572 | ||
573 | #ifdef CONFIG_F2FS_FS_POSIX_ACL | |
574 | static void __setattr_copy(struct inode *inode, const struct iattr *attr) | |
575 | { | |
576 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
577 | unsigned int ia_valid = attr->ia_valid; | |
578 | ||
579 | if (ia_valid & ATTR_UID) | |
580 | inode->i_uid = attr->ia_uid; | |
581 | if (ia_valid & ATTR_GID) | |
582 | inode->i_gid = attr->ia_gid; | |
583 | if (ia_valid & ATTR_ATIME) | |
584 | inode->i_atime = timespec_trunc(attr->ia_atime, | |
585 | inode->i_sb->s_time_gran); | |
586 | if (ia_valid & ATTR_MTIME) | |
587 | inode->i_mtime = timespec_trunc(attr->ia_mtime, | |
588 | inode->i_sb->s_time_gran); | |
589 | if (ia_valid & ATTR_CTIME) | |
590 | inode->i_ctime = timespec_trunc(attr->ia_ctime, | |
591 | inode->i_sb->s_time_gran); | |
592 | if (ia_valid & ATTR_MODE) { | |
593 | umode_t mode = attr->ia_mode; | |
594 | ||
595 | if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID)) | |
596 | mode &= ~S_ISGID; | |
597 | set_acl_inode(fi, mode); | |
598 | } | |
599 | } | |
600 | #else | |
601 | #define __setattr_copy setattr_copy | |
602 | #endif | |
603 | ||
604 | int f2fs_setattr(struct dentry *dentry, struct iattr *attr) | |
605 | { | |
606 | struct inode *inode = dentry->d_inode; | |
607 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
608 | int err; | |
609 | ||
610 | err = inode_change_ok(inode, attr); | |
611 | if (err) | |
612 | return err; | |
613 | ||
614 | if (attr->ia_valid & ATTR_SIZE) { | |
615 | if (attr->ia_size != i_size_read(inode)) { | |
616 | truncate_setsize(inode, attr->ia_size); | |
617 | f2fs_truncate(inode); | |
618 | f2fs_balance_fs(F2FS_I_SB(inode)); | |
619 | } else { | |
620 | /* | |
621 | * giving a chance to truncate blocks past EOF which | |
622 | * are fallocated with FALLOC_FL_KEEP_SIZE. | |
623 | */ | |
624 | f2fs_truncate(inode); | |
625 | } | |
626 | } | |
627 | ||
628 | __setattr_copy(inode, attr); | |
629 | ||
630 | if (attr->ia_valid & ATTR_MODE) { | |
631 | err = posix_acl_chmod(inode, get_inode_mode(inode)); | |
632 | if (err || is_inode_flag_set(fi, FI_ACL_MODE)) { | |
633 | inode->i_mode = fi->i_acl_mode; | |
634 | clear_inode_flag(fi, FI_ACL_MODE); | |
635 | } | |
636 | } | |
637 | ||
638 | mark_inode_dirty(inode); | |
639 | return err; | |
640 | } | |
641 | ||
642 | const struct inode_operations f2fs_file_inode_operations = { | |
643 | .getattr = f2fs_getattr, | |
644 | .setattr = f2fs_setattr, | |
645 | .get_acl = f2fs_get_acl, | |
646 | .set_acl = f2fs_set_acl, | |
647 | #ifdef CONFIG_F2FS_FS_XATTR | |
648 | .setxattr = generic_setxattr, | |
649 | .getxattr = generic_getxattr, | |
650 | .listxattr = f2fs_listxattr, | |
651 | .removexattr = generic_removexattr, | |
652 | #endif | |
653 | .fiemap = f2fs_fiemap, | |
654 | }; | |
655 | ||
656 | static void fill_zero(struct inode *inode, pgoff_t index, | |
657 | loff_t start, loff_t len) | |
658 | { | |
659 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
660 | struct page *page; | |
661 | ||
662 | if (!len) | |
663 | return; | |
664 | ||
665 | f2fs_balance_fs(sbi); | |
666 | ||
667 | f2fs_lock_op(sbi); | |
668 | page = get_new_data_page(inode, NULL, index, false); | |
669 | f2fs_unlock_op(sbi); | |
670 | ||
671 | if (!IS_ERR(page)) { | |
672 | f2fs_wait_on_page_writeback(page, DATA); | |
673 | zero_user(page, start, len); | |
674 | set_page_dirty(page); | |
675 | f2fs_put_page(page, 1); | |
676 | } | |
677 | } | |
678 | ||
679 | int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) | |
680 | { | |
681 | pgoff_t index; | |
682 | int err; | |
683 | ||
684 | for (index = pg_start; index < pg_end; index++) { | |
685 | struct dnode_of_data dn; | |
686 | ||
687 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
688 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); | |
689 | if (err) { | |
690 | if (err == -ENOENT) | |
691 | continue; | |
692 | return err; | |
693 | } | |
694 | ||
695 | if (dn.data_blkaddr != NULL_ADDR) | |
696 | truncate_data_blocks_range(&dn, 1); | |
697 | f2fs_put_dnode(&dn); | |
698 | } | |
699 | return 0; | |
700 | } | |
701 | ||
702 | static int punch_hole(struct inode *inode, loff_t offset, loff_t len) | |
703 | { | |
704 | pgoff_t pg_start, pg_end; | |
705 | loff_t off_start, off_end; | |
706 | int ret = 0; | |
707 | ||
708 | if (!S_ISREG(inode->i_mode)) | |
709 | return -EOPNOTSUPP; | |
710 | ||
711 | /* skip punching hole beyond i_size */ | |
712 | if (offset >= inode->i_size) | |
713 | return ret; | |
714 | ||
715 | if (f2fs_has_inline_data(inode)) { | |
716 | ret = f2fs_convert_inline_inode(inode); | |
717 | if (ret) | |
718 | return ret; | |
719 | } | |
720 | ||
721 | pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT; | |
722 | pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT; | |
723 | ||
724 | off_start = offset & (PAGE_CACHE_SIZE - 1); | |
725 | off_end = (offset + len) & (PAGE_CACHE_SIZE - 1); | |
726 | ||
727 | if (pg_start == pg_end) { | |
728 | fill_zero(inode, pg_start, off_start, | |
729 | off_end - off_start); | |
730 | } else { | |
731 | if (off_start) | |
732 | fill_zero(inode, pg_start++, off_start, | |
733 | PAGE_CACHE_SIZE - off_start); | |
734 | if (off_end) | |
735 | fill_zero(inode, pg_end, 0, off_end); | |
736 | ||
737 | if (pg_start < pg_end) { | |
738 | struct address_space *mapping = inode->i_mapping; | |
739 | loff_t blk_start, blk_end; | |
740 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
741 | ||
742 | f2fs_balance_fs(sbi); | |
743 | ||
744 | blk_start = pg_start << PAGE_CACHE_SHIFT; | |
745 | blk_end = pg_end << PAGE_CACHE_SHIFT; | |
746 | truncate_inode_pages_range(mapping, blk_start, | |
747 | blk_end - 1); | |
748 | ||
749 | f2fs_lock_op(sbi); | |
750 | ret = truncate_hole(inode, pg_start, pg_end); | |
751 | f2fs_unlock_op(sbi); | |
752 | } | |
753 | } | |
754 | ||
755 | return ret; | |
756 | } | |
757 | ||
758 | static int expand_inode_data(struct inode *inode, loff_t offset, | |
759 | loff_t len, int mode) | |
760 | { | |
761 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
762 | pgoff_t index, pg_start, pg_end; | |
763 | loff_t new_size = i_size_read(inode); | |
764 | loff_t off_start, off_end; | |
765 | int ret = 0; | |
766 | ||
767 | f2fs_balance_fs(sbi); | |
768 | ||
769 | ret = inode_newsize_ok(inode, (len + offset)); | |
770 | if (ret) | |
771 | return ret; | |
772 | ||
773 | if (f2fs_has_inline_data(inode)) { | |
774 | ret = f2fs_convert_inline_inode(inode); | |
775 | if (ret) | |
776 | return ret; | |
777 | } | |
778 | ||
779 | pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT; | |
780 | pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT; | |
781 | ||
782 | off_start = offset & (PAGE_CACHE_SIZE - 1); | |
783 | off_end = (offset + len) & (PAGE_CACHE_SIZE - 1); | |
784 | ||
785 | f2fs_lock_op(sbi); | |
786 | ||
787 | for (index = pg_start; index <= pg_end; index++) { | |
788 | struct dnode_of_data dn; | |
789 | ||
790 | if (index == pg_end && !off_end) | |
791 | goto noalloc; | |
792 | ||
793 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
794 | ret = f2fs_reserve_block(&dn, index); | |
795 | if (ret) | |
796 | break; | |
797 | noalloc: | |
798 | if (pg_start == pg_end) | |
799 | new_size = offset + len; | |
800 | else if (index == pg_start && off_start) | |
801 | new_size = (index + 1) << PAGE_CACHE_SHIFT; | |
802 | else if (index == pg_end) | |
803 | new_size = (index << PAGE_CACHE_SHIFT) + off_end; | |
804 | else | |
805 | new_size += PAGE_CACHE_SIZE; | |
806 | } | |
807 | ||
808 | if (!(mode & FALLOC_FL_KEEP_SIZE) && | |
809 | i_size_read(inode) < new_size) { | |
810 | i_size_write(inode, new_size); | |
811 | mark_inode_dirty(inode); | |
812 | update_inode_page(inode); | |
813 | } | |
814 | f2fs_unlock_op(sbi); | |
815 | ||
816 | return ret; | |
817 | } | |
818 | ||
819 | static long f2fs_fallocate(struct file *file, int mode, | |
820 | loff_t offset, loff_t len) | |
821 | { | |
822 | struct inode *inode = file_inode(file); | |
823 | long ret; | |
824 | ||
825 | if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) | |
826 | return -EOPNOTSUPP; | |
827 | ||
828 | mutex_lock(&inode->i_mutex); | |
829 | ||
830 | if (mode & FALLOC_FL_PUNCH_HOLE) | |
831 | ret = punch_hole(inode, offset, len); | |
832 | else | |
833 | ret = expand_inode_data(inode, offset, len, mode); | |
834 | ||
835 | if (!ret) { | |
836 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
837 | mark_inode_dirty(inode); | |
838 | } | |
839 | ||
840 | mutex_unlock(&inode->i_mutex); | |
841 | ||
842 | trace_f2fs_fallocate(inode, mode, offset, len, ret); | |
843 | return ret; | |
844 | } | |
845 | ||
846 | static int f2fs_release_file(struct inode *inode, struct file *filp) | |
847 | { | |
848 | /* some remained atomic pages should discarded */ | |
849 | if (f2fs_is_atomic_file(inode)) | |
850 | commit_inmem_pages(inode, true); | |
851 | if (f2fs_is_volatile_file(inode)) { | |
852 | set_inode_flag(F2FS_I(inode), FI_DROP_CACHE); | |
853 | filemap_fdatawrite(inode->i_mapping); | |
854 | clear_inode_flag(F2FS_I(inode), FI_DROP_CACHE); | |
855 | } | |
856 | return 0; | |
857 | } | |
858 | ||
859 | #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL)) | |
860 | #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL) | |
861 | ||
862 | static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags) | |
863 | { | |
864 | if (S_ISDIR(mode)) | |
865 | return flags; | |
866 | else if (S_ISREG(mode)) | |
867 | return flags & F2FS_REG_FLMASK; | |
868 | else | |
869 | return flags & F2FS_OTHER_FLMASK; | |
870 | } | |
871 | ||
872 | static int f2fs_ioc_getflags(struct file *filp, unsigned long arg) | |
873 | { | |
874 | struct inode *inode = file_inode(filp); | |
875 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
876 | unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE; | |
877 | return put_user(flags, (int __user *)arg); | |
878 | } | |
879 | ||
880 | static int f2fs_ioc_setflags(struct file *filp, unsigned long arg) | |
881 | { | |
882 | struct inode *inode = file_inode(filp); | |
883 | struct f2fs_inode_info *fi = F2FS_I(inode); | |
884 | unsigned int flags = fi->i_flags & FS_FL_USER_VISIBLE; | |
885 | unsigned int oldflags; | |
886 | int ret; | |
887 | ||
888 | ret = mnt_want_write_file(filp); | |
889 | if (ret) | |
890 | return ret; | |
891 | ||
892 | if (!inode_owner_or_capable(inode)) { | |
893 | ret = -EACCES; | |
894 | goto out; | |
895 | } | |
896 | ||
897 | if (get_user(flags, (int __user *)arg)) { | |
898 | ret = -EFAULT; | |
899 | goto out; | |
900 | } | |
901 | ||
902 | flags = f2fs_mask_flags(inode->i_mode, flags); | |
903 | ||
904 | mutex_lock(&inode->i_mutex); | |
905 | ||
906 | oldflags = fi->i_flags; | |
907 | ||
908 | if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { | |
909 | if (!capable(CAP_LINUX_IMMUTABLE)) { | |
910 | mutex_unlock(&inode->i_mutex); | |
911 | ret = -EPERM; | |
912 | goto out; | |
913 | } | |
914 | } | |
915 | ||
916 | flags = flags & FS_FL_USER_MODIFIABLE; | |
917 | flags |= oldflags & ~FS_FL_USER_MODIFIABLE; | |
918 | fi->i_flags = flags; | |
919 | mutex_unlock(&inode->i_mutex); | |
920 | ||
921 | f2fs_set_inode_flags(inode); | |
922 | inode->i_ctime = CURRENT_TIME; | |
923 | mark_inode_dirty(inode); | |
924 | out: | |
925 | mnt_drop_write_file(filp); | |
926 | return ret; | |
927 | } | |
928 | ||
929 | static int f2fs_ioc_start_atomic_write(struct file *filp) | |
930 | { | |
931 | struct inode *inode = file_inode(filp); | |
932 | ||
933 | if (!inode_owner_or_capable(inode)) | |
934 | return -EACCES; | |
935 | ||
936 | f2fs_balance_fs(F2FS_I_SB(inode)); | |
937 | ||
938 | if (f2fs_is_atomic_file(inode)) | |
939 | return 0; | |
940 | ||
941 | set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE); | |
942 | ||
943 | return f2fs_convert_inline_inode(inode); | |
944 | } | |
945 | ||
946 | static int f2fs_ioc_commit_atomic_write(struct file *filp) | |
947 | { | |
948 | struct inode *inode = file_inode(filp); | |
949 | int ret; | |
950 | ||
951 | if (!inode_owner_or_capable(inode)) | |
952 | return -EACCES; | |
953 | ||
954 | if (f2fs_is_volatile_file(inode)) | |
955 | return 0; | |
956 | ||
957 | ret = mnt_want_write_file(filp); | |
958 | if (ret) | |
959 | return ret; | |
960 | ||
961 | if (f2fs_is_atomic_file(inode)) | |
962 | commit_inmem_pages(inode, false); | |
963 | ||
964 | ret = f2fs_sync_file(filp, 0, LONG_MAX, 0); | |
965 | mnt_drop_write_file(filp); | |
966 | clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE); | |
967 | return ret; | |
968 | } | |
969 | ||
970 | static int f2fs_ioc_start_volatile_write(struct file *filp) | |
971 | { | |
972 | struct inode *inode = file_inode(filp); | |
973 | ||
974 | if (!inode_owner_or_capable(inode)) | |
975 | return -EACCES; | |
976 | ||
977 | if (f2fs_is_volatile_file(inode)) | |
978 | return 0; | |
979 | ||
980 | set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE); | |
981 | ||
982 | return f2fs_convert_inline_inode(inode); | |
983 | } | |
984 | ||
985 | static int f2fs_ioc_release_volatile_write(struct file *filp) | |
986 | { | |
987 | struct inode *inode = file_inode(filp); | |
988 | ||
989 | if (!inode_owner_or_capable(inode)) | |
990 | return -EACCES; | |
991 | ||
992 | if (!f2fs_is_volatile_file(inode)) | |
993 | return 0; | |
994 | ||
995 | punch_hole(inode, 0, F2FS_BLKSIZE); | |
996 | return 0; | |
997 | } | |
998 | ||
999 | static int f2fs_ioc_abort_volatile_write(struct file *filp) | |
1000 | { | |
1001 | struct inode *inode = file_inode(filp); | |
1002 | int ret; | |
1003 | ||
1004 | if (!inode_owner_or_capable(inode)) | |
1005 | return -EACCES; | |
1006 | ||
1007 | ret = mnt_want_write_file(filp); | |
1008 | if (ret) | |
1009 | return ret; | |
1010 | ||
1011 | f2fs_balance_fs(F2FS_I_SB(inode)); | |
1012 | ||
1013 | if (f2fs_is_atomic_file(inode)) { | |
1014 | commit_inmem_pages(inode, false); | |
1015 | clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE); | |
1016 | } | |
1017 | ||
1018 | if (f2fs_is_volatile_file(inode)) { | |
1019 | clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE); | |
1020 | filemap_fdatawrite(inode->i_mapping); | |
1021 | set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE); | |
1022 | } | |
1023 | mnt_drop_write_file(filp); | |
1024 | return ret; | |
1025 | } | |
1026 | ||
1027 | static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) | |
1028 | { | |
1029 | struct inode *inode = file_inode(filp); | |
1030 | struct super_block *sb = inode->i_sb; | |
1031 | struct request_queue *q = bdev_get_queue(sb->s_bdev); | |
1032 | struct fstrim_range range; | |
1033 | int ret; | |
1034 | ||
1035 | if (!capable(CAP_SYS_ADMIN)) | |
1036 | return -EPERM; | |
1037 | ||
1038 | if (!blk_queue_discard(q)) | |
1039 | return -EOPNOTSUPP; | |
1040 | ||
1041 | if (copy_from_user(&range, (struct fstrim_range __user *)arg, | |
1042 | sizeof(range))) | |
1043 | return -EFAULT; | |
1044 | ||
1045 | range.minlen = max((unsigned int)range.minlen, | |
1046 | q->limits.discard_granularity); | |
1047 | ret = f2fs_trim_fs(F2FS_SB(sb), &range); | |
1048 | if (ret < 0) | |
1049 | return ret; | |
1050 | ||
1051 | if (copy_to_user((struct fstrim_range __user *)arg, &range, | |
1052 | sizeof(range))) | |
1053 | return -EFAULT; | |
1054 | return 0; | |
1055 | } | |
1056 | ||
1057 | long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) | |
1058 | { | |
1059 | switch (cmd) { | |
1060 | case F2FS_IOC_GETFLAGS: | |
1061 | return f2fs_ioc_getflags(filp, arg); | |
1062 | case F2FS_IOC_SETFLAGS: | |
1063 | return f2fs_ioc_setflags(filp, arg); | |
1064 | case F2FS_IOC_START_ATOMIC_WRITE: | |
1065 | return f2fs_ioc_start_atomic_write(filp); | |
1066 | case F2FS_IOC_COMMIT_ATOMIC_WRITE: | |
1067 | return f2fs_ioc_commit_atomic_write(filp); | |
1068 | case F2FS_IOC_START_VOLATILE_WRITE: | |
1069 | return f2fs_ioc_start_volatile_write(filp); | |
1070 | case F2FS_IOC_RELEASE_VOLATILE_WRITE: | |
1071 | return f2fs_ioc_release_volatile_write(filp); | |
1072 | case F2FS_IOC_ABORT_VOLATILE_WRITE: | |
1073 | return f2fs_ioc_abort_volatile_write(filp); | |
1074 | case FITRIM: | |
1075 | return f2fs_ioc_fitrim(filp, arg); | |
1076 | default: | |
1077 | return -ENOTTY; | |
1078 | } | |
1079 | } | |
1080 | ||
1081 | #ifdef CONFIG_COMPAT | |
1082 | long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
1083 | { | |
1084 | switch (cmd) { | |
1085 | case F2FS_IOC32_GETFLAGS: | |
1086 | cmd = F2FS_IOC_GETFLAGS; | |
1087 | break; | |
1088 | case F2FS_IOC32_SETFLAGS: | |
1089 | cmd = F2FS_IOC_SETFLAGS; | |
1090 | break; | |
1091 | default: | |
1092 | return -ENOIOCTLCMD; | |
1093 | } | |
1094 | return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); | |
1095 | } | |
1096 | #endif | |
1097 | ||
1098 | const struct file_operations f2fs_file_operations = { | |
1099 | .llseek = f2fs_llseek, | |
1100 | .read = new_sync_read, | |
1101 | .write = new_sync_write, | |
1102 | .read_iter = generic_file_read_iter, | |
1103 | .write_iter = generic_file_write_iter, | |
1104 | .open = generic_file_open, | |
1105 | .release = f2fs_release_file, | |
1106 | .mmap = f2fs_file_mmap, | |
1107 | .fsync = f2fs_sync_file, | |
1108 | .fallocate = f2fs_fallocate, | |
1109 | .unlocked_ioctl = f2fs_ioctl, | |
1110 | #ifdef CONFIG_COMPAT | |
1111 | .compat_ioctl = f2fs_compat_ioctl, | |
1112 | #endif | |
1113 | .splice_read = generic_file_splice_read, | |
1114 | .splice_write = iter_file_splice_write, | |
1115 | }; |