]>
Commit | Line | Data |
---|---|---|
05fe58fd RK |
1 | /* |
2 | * inode.c - NILFS inode operations. | |
3 | * | |
4 | * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
19 | * | |
20 | * Written by Ryusuke Konishi <ryusuke@osrg.net> | |
21 | * | |
22 | */ | |
23 | ||
24 | #include <linux/buffer_head.h> | |
5a0e3ad6 | 25 | #include <linux/gfp.h> |
05fe58fd RK |
26 | #include <linux/mpage.h> |
27 | #include <linux/writeback.h> | |
f30bf3e4 | 28 | #include <linux/uio.h> |
05fe58fd | 29 | #include "nilfs.h" |
6fd1e5c9 | 30 | #include "btnode.h" |
05fe58fd RK |
31 | #include "segment.h" |
32 | #include "page.h" | |
33 | #include "mdt.h" | |
34 | #include "cpfile.h" | |
35 | #include "ifile.h" | |
36 | ||
0e14a359 RK |
37 | struct nilfs_iget_args { |
38 | u64 ino; | |
39 | __u64 cno; | |
4d8d9293 | 40 | struct nilfs_root *root; |
0e14a359 RK |
41 | int for_gc; |
42 | }; | |
05fe58fd RK |
43 | |
44 | /** | |
45 | * nilfs_get_block() - get a file block on the filesystem (callback function) | |
46 | * @inode - inode struct of the target file | |
47 | * @blkoff - file block number | |
48 | * @bh_result - buffer head to be mapped on | |
49 | * @create - indicate whether allocating the block or not when it has not | |
50 | * been allocated yet. | |
51 | * | |
52 | * This function does not issue actual read request of the specified data | |
53 | * block. It is done by VFS. | |
05fe58fd RK |
54 | */ |
55 | int nilfs_get_block(struct inode *inode, sector_t blkoff, | |
56 | struct buffer_head *bh_result, int create) | |
57 | { | |
58 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
c3a7abf0 | 59 | __u64 blknum = 0; |
05fe58fd RK |
60 | int err = 0, ret; |
61 | struct inode *dat = nilfs_dat_inode(NILFS_I_NILFS(inode)); | |
c3a7abf0 | 62 | unsigned maxblocks = bh_result->b_size >> inode->i_blkbits; |
05fe58fd | 63 | |
c3a7abf0 RK |
64 | down_read(&NILFS_MDT(dat)->mi_sem); |
65 | ret = nilfs_bmap_lookup_contig(ii->i_bmap, blkoff, &blknum, maxblocks); | |
66 | up_read(&NILFS_MDT(dat)->mi_sem); | |
67 | if (ret >= 0) { /* found */ | |
05fe58fd | 68 | map_bh(bh_result, inode->i_sb, blknum); |
c3a7abf0 RK |
69 | if (ret > 0) |
70 | bh_result->b_size = (ret << inode->i_blkbits); | |
05fe58fd RK |
71 | goto out; |
72 | } | |
05fe58fd RK |
73 | /* data block was not found */ |
74 | if (ret == -ENOENT && create) { | |
75 | struct nilfs_transaction_info ti; | |
76 | ||
77 | bh_result->b_blocknr = 0; | |
78 | err = nilfs_transaction_begin(inode->i_sb, &ti, 1); | |
79 | if (unlikely(err)) | |
80 | goto out; | |
81 | err = nilfs_bmap_insert(ii->i_bmap, (unsigned long)blkoff, | |
82 | (unsigned long)bh_result); | |
05fe58fd RK |
83 | if (unlikely(err != 0)) { |
84 | if (err == -EEXIST) { | |
85 | /* | |
86 | * The get_block() function could be called | |
87 | * from multiple callers for an inode. | |
88 | * However, the page having this block must | |
89 | * be locked in this case. | |
90 | */ | |
1f5abe7e | 91 | printk(KERN_WARNING |
05fe58fd RK |
92 | "nilfs_get_block: a race condition " |
93 | "while inserting a data block. " | |
94 | "(inode number=%lu, file block " | |
95 | "offset=%llu)\n", | |
96 | inode->i_ino, | |
97 | (unsigned long long)blkoff); | |
1f5abe7e | 98 | err = 0; |
05fe58fd RK |
99 | } else if (err == -EINVAL) { |
100 | nilfs_error(inode->i_sb, __func__, | |
101 | "broken bmap (inode=%lu)\n", | |
102 | inode->i_ino); | |
103 | err = -EIO; | |
104 | } | |
47420c79 | 105 | nilfs_transaction_abort(inode->i_sb); |
05fe58fd RK |
106 | goto out; |
107 | } | |
abdb318b | 108 | nilfs_mark_inode_dirty(inode); |
47420c79 | 109 | nilfs_transaction_commit(inode->i_sb); /* never fails */ |
05fe58fd RK |
110 | /* Error handling should be detailed */ |
111 | set_buffer_new(bh_result); | |
112 | map_bh(bh_result, inode->i_sb, 0); /* dbn must be changed | |
113 | to proper value */ | |
114 | } else if (ret == -ENOENT) { | |
115 | /* not found is not error (e.g. hole); must return without | |
116 | the mapped state flag. */ | |
117 | ; | |
118 | } else { | |
119 | err = ret; | |
120 | } | |
121 | ||
122 | out: | |
123 | return err; | |
124 | } | |
125 | ||
126 | /** | |
127 | * nilfs_readpage() - implement readpage() method of nilfs_aops {} | |
128 | * address_space_operations. | |
129 | * @file - file struct of the file to be read | |
130 | * @page - the page to be read | |
131 | */ | |
132 | static int nilfs_readpage(struct file *file, struct page *page) | |
133 | { | |
134 | return mpage_readpage(page, nilfs_get_block); | |
135 | } | |
136 | ||
137 | /** | |
138 | * nilfs_readpages() - implement readpages() method of nilfs_aops {} | |
139 | * address_space_operations. | |
140 | * @file - file struct of the file to be read | |
141 | * @mapping - address_space struct used for reading multiple pages | |
142 | * @pages - the pages to be read | |
143 | * @nr_pages - number of pages to be read | |
144 | */ | |
145 | static int nilfs_readpages(struct file *file, struct address_space *mapping, | |
146 | struct list_head *pages, unsigned nr_pages) | |
147 | { | |
148 | return mpage_readpages(mapping, pages, nr_pages, nilfs_get_block); | |
149 | } | |
150 | ||
151 | static int nilfs_writepages(struct address_space *mapping, | |
152 | struct writeback_control *wbc) | |
153 | { | |
f30bf3e4 RK |
154 | struct inode *inode = mapping->host; |
155 | int err = 0; | |
156 | ||
157 | if (wbc->sync_mode == WB_SYNC_ALL) | |
158 | err = nilfs_construct_dsync_segment(inode->i_sb, inode, | |
159 | wbc->range_start, | |
160 | wbc->range_end); | |
161 | return err; | |
05fe58fd RK |
162 | } |
163 | ||
164 | static int nilfs_writepage(struct page *page, struct writeback_control *wbc) | |
165 | { | |
166 | struct inode *inode = page->mapping->host; | |
167 | int err; | |
168 | ||
169 | redirty_page_for_writepage(wbc, page); | |
170 | unlock_page(page); | |
171 | ||
172 | if (wbc->sync_mode == WB_SYNC_ALL) { | |
173 | err = nilfs_construct_segment(inode->i_sb); | |
174 | if (unlikely(err)) | |
175 | return err; | |
176 | } else if (wbc->for_reclaim) | |
177 | nilfs_flush_segment(inode->i_sb, inode->i_ino); | |
178 | ||
179 | return 0; | |
180 | } | |
181 | ||
182 | static int nilfs_set_page_dirty(struct page *page) | |
183 | { | |
184 | int ret = __set_page_dirty_buffers(page); | |
185 | ||
186 | if (ret) { | |
187 | struct inode *inode = page->mapping->host; | |
188 | struct nilfs_sb_info *sbi = NILFS_SB(inode->i_sb); | |
189 | unsigned nr_dirty = 1 << (PAGE_SHIFT - inode->i_blkbits); | |
190 | ||
191 | nilfs_set_file_dirty(sbi, inode, nr_dirty); | |
192 | } | |
193 | return ret; | |
194 | } | |
195 | ||
196 | static int nilfs_write_begin(struct file *file, struct address_space *mapping, | |
197 | loff_t pos, unsigned len, unsigned flags, | |
198 | struct page **pagep, void **fsdata) | |
199 | ||
200 | { | |
201 | struct inode *inode = mapping->host; | |
202 | int err = nilfs_transaction_begin(inode->i_sb, NULL, 1); | |
203 | ||
204 | if (unlikely(err)) | |
205 | return err; | |
206 | ||
155130a4 CH |
207 | err = block_write_begin(mapping, pos, len, flags, pagep, |
208 | nilfs_get_block); | |
209 | if (unlikely(err)) { | |
210 | loff_t isize = mapping->host->i_size; | |
211 | if (pos + len > isize) | |
212 | vmtruncate(mapping->host, isize); | |
213 | ||
47420c79 | 214 | nilfs_transaction_abort(inode->i_sb); |
155130a4 | 215 | } |
05fe58fd RK |
216 | return err; |
217 | } | |
218 | ||
219 | static int nilfs_write_end(struct file *file, struct address_space *mapping, | |
220 | loff_t pos, unsigned len, unsigned copied, | |
221 | struct page *page, void *fsdata) | |
222 | { | |
223 | struct inode *inode = mapping->host; | |
224 | unsigned start = pos & (PAGE_CACHE_SIZE - 1); | |
225 | unsigned nr_dirty; | |
226 | int err; | |
227 | ||
228 | nr_dirty = nilfs_page_count_clean_buffers(page, start, | |
229 | start + copied); | |
230 | copied = generic_write_end(file, mapping, pos, len, copied, page, | |
231 | fsdata); | |
232 | nilfs_set_file_dirty(NILFS_SB(inode->i_sb), inode, nr_dirty); | |
47420c79 | 233 | err = nilfs_transaction_commit(inode->i_sb); |
05fe58fd RK |
234 | return err ? : copied; |
235 | } | |
236 | ||
237 | static ssize_t | |
238 | nilfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, | |
239 | loff_t offset, unsigned long nr_segs) | |
240 | { | |
241 | struct file *file = iocb->ki_filp; | |
242 | struct inode *inode = file->f_mapping->host; | |
243 | ssize_t size; | |
05fe58fd RK |
244 | |
245 | if (rw == WRITE) | |
246 | return 0; | |
247 | ||
248 | /* Needs synchronization with the cleaner */ | |
249 | size = blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, | |
250 | offset, nr_segs, nilfs_get_block, NULL); | |
eafdc7d1 CH |
251 | |
252 | /* | |
253 | * In case of error extending write may have instantiated a few | |
254 | * blocks outside i_size. Trim these off again. | |
255 | */ | |
256 | if (unlikely((rw & WRITE) && size < 0)) { | |
257 | loff_t isize = i_size_read(inode); | |
258 | loff_t end = offset + iov_length(iov, nr_segs); | |
259 | ||
260 | if (end > isize) | |
261 | vmtruncate(inode, isize); | |
262 | } | |
263 | ||
05fe58fd RK |
264 | return size; |
265 | } | |
266 | ||
7f09410b | 267 | const struct address_space_operations nilfs_aops = { |
05fe58fd RK |
268 | .writepage = nilfs_writepage, |
269 | .readpage = nilfs_readpage, | |
e85dc1d5 | 270 | .sync_page = block_sync_page, |
05fe58fd RK |
271 | .writepages = nilfs_writepages, |
272 | .set_page_dirty = nilfs_set_page_dirty, | |
273 | .readpages = nilfs_readpages, | |
274 | .write_begin = nilfs_write_begin, | |
275 | .write_end = nilfs_write_end, | |
276 | /* .releasepage = nilfs_releasepage, */ | |
277 | .invalidatepage = block_invalidatepage, | |
278 | .direct_IO = nilfs_direct_IO, | |
258ef67e | 279 | .is_partially_uptodate = block_is_partially_uptodate, |
05fe58fd RK |
280 | }; |
281 | ||
282 | struct inode *nilfs_new_inode(struct inode *dir, int mode) | |
283 | { | |
284 | struct super_block *sb = dir->i_sb; | |
285 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
286 | struct inode *inode; | |
287 | struct nilfs_inode_info *ii; | |
4d8d9293 | 288 | struct nilfs_root *root; |
05fe58fd RK |
289 | int err = -ENOMEM; |
290 | ino_t ino; | |
291 | ||
292 | inode = new_inode(sb); | |
293 | if (unlikely(!inode)) | |
294 | goto failed; | |
295 | ||
296 | mapping_set_gfp_mask(inode->i_mapping, | |
297 | mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); | |
298 | ||
4d8d9293 | 299 | root = NILFS_I(dir)->i_root; |
05fe58fd RK |
300 | ii = NILFS_I(inode); |
301 | ii->i_state = 1 << NILFS_I_NEW; | |
4d8d9293 | 302 | ii->i_root = root; |
05fe58fd RK |
303 | |
304 | err = nilfs_ifile_create_inode(sbi->s_ifile, &ino, &ii->i_bh); | |
305 | if (unlikely(err)) | |
306 | goto failed_ifile_create_inode; | |
307 | /* reference count of i_bh inherits from nilfs_mdt_read_block() */ | |
308 | ||
309 | atomic_inc(&sbi->s_inodes_count); | |
73459dcc | 310 | inode_init_owner(inode, dir, mode); |
05fe58fd RK |
311 | inode->i_ino = ino; |
312 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
313 | ||
314 | if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)) { | |
315 | err = nilfs_bmap_read(ii->i_bmap, NULL); | |
316 | if (err < 0) | |
317 | goto failed_bmap; | |
318 | ||
319 | set_bit(NILFS_I_BMAP, &ii->i_state); | |
320 | /* No lock is needed; iget() ensures it. */ | |
321 | } | |
322 | ||
323 | ii->i_flags = NILFS_I(dir)->i_flags; | |
324 | if (S_ISLNK(mode)) | |
325 | ii->i_flags &= ~(NILFS_IMMUTABLE_FL | NILFS_APPEND_FL); | |
326 | if (!S_ISDIR(mode)) | |
327 | ii->i_flags &= ~NILFS_DIRSYNC_FL; | |
328 | ||
329 | /* ii->i_file_acl = 0; */ | |
330 | /* ii->i_dir_acl = 0; */ | |
05fe58fd | 331 | ii->i_dir_start_lookup = 0; |
05fe58fd RK |
332 | nilfs_set_inode_flags(inode); |
333 | spin_lock(&sbi->s_next_gen_lock); | |
334 | inode->i_generation = sbi->s_next_generation++; | |
335 | spin_unlock(&sbi->s_next_gen_lock); | |
336 | insert_inode_hash(inode); | |
337 | ||
338 | err = nilfs_init_acl(inode, dir); | |
339 | if (unlikely(err)) | |
340 | goto failed_acl; /* never occur. When supporting | |
341 | nilfs_init_acl(), proper cancellation of | |
342 | above jobs should be considered */ | |
343 | ||
05fe58fd RK |
344 | return inode; |
345 | ||
346 | failed_acl: | |
347 | failed_bmap: | |
348 | inode->i_nlink = 0; | |
349 | iput(inode); /* raw_inode will be deleted through | |
350 | generic_delete_inode() */ | |
351 | goto failed; | |
352 | ||
353 | failed_ifile_create_inode: | |
354 | make_bad_inode(inode); | |
355 | iput(inode); /* if i_nlink == 1, generic_forget_inode() will be | |
356 | called */ | |
357 | failed: | |
358 | return ERR_PTR(err); | |
359 | } | |
360 | ||
361 | void nilfs_free_inode(struct inode *inode) | |
362 | { | |
363 | struct super_block *sb = inode->i_sb; | |
364 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
365 | ||
05fe58fd RK |
366 | /* XXX: check error code? Is there any thing I can do? */ |
367 | (void) nilfs_ifile_delete_inode(sbi->s_ifile, inode->i_ino); | |
368 | atomic_dec(&sbi->s_inodes_count); | |
369 | } | |
370 | ||
371 | void nilfs_set_inode_flags(struct inode *inode) | |
372 | { | |
373 | unsigned int flags = NILFS_I(inode)->i_flags; | |
374 | ||
375 | inode->i_flags &= ~(S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | | |
376 | S_DIRSYNC); | |
377 | if (flags & NILFS_SYNC_FL) | |
378 | inode->i_flags |= S_SYNC; | |
379 | if (flags & NILFS_APPEND_FL) | |
380 | inode->i_flags |= S_APPEND; | |
381 | if (flags & NILFS_IMMUTABLE_FL) | |
382 | inode->i_flags |= S_IMMUTABLE; | |
383 | #ifndef NILFS_ATIME_DISABLE | |
384 | if (flags & NILFS_NOATIME_FL) | |
385 | #endif | |
386 | inode->i_flags |= S_NOATIME; | |
387 | if (flags & NILFS_DIRSYNC_FL) | |
388 | inode->i_flags |= S_DIRSYNC; | |
389 | mapping_set_gfp_mask(inode->i_mapping, | |
390 | mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); | |
391 | } | |
392 | ||
393 | int nilfs_read_inode_common(struct inode *inode, | |
394 | struct nilfs_inode *raw_inode) | |
395 | { | |
396 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
397 | int err; | |
398 | ||
399 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); | |
400 | inode->i_uid = (uid_t)le32_to_cpu(raw_inode->i_uid); | |
401 | inode->i_gid = (gid_t)le32_to_cpu(raw_inode->i_gid); | |
402 | inode->i_nlink = le16_to_cpu(raw_inode->i_links_count); | |
403 | inode->i_size = le64_to_cpu(raw_inode->i_size); | |
404 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
405 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); | |
406 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); | |
61239230 RK |
407 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); |
408 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); | |
409 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); | |
410 | if (inode->i_nlink == 0 && inode->i_mode == 0) | |
05fe58fd RK |
411 | return -EINVAL; /* this inode is deleted */ |
412 | ||
413 | inode->i_blocks = le64_to_cpu(raw_inode->i_blocks); | |
414 | ii->i_flags = le32_to_cpu(raw_inode->i_flags); | |
415 | #if 0 | |
416 | ii->i_file_acl = le32_to_cpu(raw_inode->i_file_acl); | |
417 | ii->i_dir_acl = S_ISREG(inode->i_mode) ? | |
418 | 0 : le32_to_cpu(raw_inode->i_dir_acl); | |
419 | #endif | |
3cc811bf | 420 | ii->i_dir_start_lookup = 0; |
05fe58fd RK |
421 | inode->i_generation = le32_to_cpu(raw_inode->i_generation); |
422 | ||
423 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
424 | S_ISLNK(inode->i_mode)) { | |
425 | err = nilfs_bmap_read(ii->i_bmap, raw_inode); | |
426 | if (err < 0) | |
427 | return err; | |
428 | set_bit(NILFS_I_BMAP, &ii->i_state); | |
429 | /* No lock is needed; iget() ensures it. */ | |
430 | } | |
431 | return 0; | |
432 | } | |
433 | ||
05fe58fd RK |
434 | static int __nilfs_read_inode(struct super_block *sb, unsigned long ino, |
435 | struct inode *inode) | |
436 | { | |
437 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
438 | struct inode *dat = nilfs_dat_inode(sbi->s_nilfs); | |
439 | struct buffer_head *bh; | |
440 | struct nilfs_inode *raw_inode; | |
441 | int err; | |
442 | ||
443 | down_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ | |
444 | err = nilfs_ifile_get_inode_block(sbi->s_ifile, ino, &bh); | |
445 | if (unlikely(err)) | |
446 | goto bad_inode; | |
447 | ||
448 | raw_inode = nilfs_ifile_map_inode(sbi->s_ifile, ino, bh); | |
449 | ||
1b2f5a64 RK |
450 | err = nilfs_read_inode_common(inode, raw_inode); |
451 | if (err) | |
05fe58fd RK |
452 | goto failed_unmap; |
453 | ||
454 | if (S_ISREG(inode->i_mode)) { | |
455 | inode->i_op = &nilfs_file_inode_operations; | |
456 | inode->i_fop = &nilfs_file_operations; | |
457 | inode->i_mapping->a_ops = &nilfs_aops; | |
05fe58fd RK |
458 | } else if (S_ISDIR(inode->i_mode)) { |
459 | inode->i_op = &nilfs_dir_inode_operations; | |
460 | inode->i_fop = &nilfs_dir_operations; | |
461 | inode->i_mapping->a_ops = &nilfs_aops; | |
462 | } else if (S_ISLNK(inode->i_mode)) { | |
463 | inode->i_op = &nilfs_symlink_inode_operations; | |
464 | inode->i_mapping->a_ops = &nilfs_aops; | |
465 | } else { | |
466 | inode->i_op = &nilfs_special_inode_operations; | |
467 | init_special_inode( | |
468 | inode, inode->i_mode, | |
cdce214e | 469 | huge_decode_dev(le64_to_cpu(raw_inode->i_device_code))); |
05fe58fd RK |
470 | } |
471 | nilfs_ifile_unmap_inode(sbi->s_ifile, ino, bh); | |
472 | brelse(bh); | |
473 | up_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ | |
474 | nilfs_set_inode_flags(inode); | |
475 | return 0; | |
476 | ||
477 | failed_unmap: | |
478 | nilfs_ifile_unmap_inode(sbi->s_ifile, ino, bh); | |
479 | brelse(bh); | |
480 | ||
481 | bad_inode: | |
482 | up_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ | |
483 | return err; | |
484 | } | |
485 | ||
0e14a359 RK |
486 | static int nilfs_iget_test(struct inode *inode, void *opaque) |
487 | { | |
488 | struct nilfs_iget_args *args = opaque; | |
489 | struct nilfs_inode_info *ii; | |
490 | ||
4d8d9293 | 491 | if (args->ino != inode->i_ino || args->root != NILFS_I(inode)->i_root) |
0e14a359 RK |
492 | return 0; |
493 | ||
494 | ii = NILFS_I(inode); | |
495 | if (!test_bit(NILFS_I_GCINODE, &ii->i_state)) | |
496 | return !args->for_gc; | |
497 | ||
498 | return args->for_gc && args->cno == ii->i_cno; | |
499 | } | |
500 | ||
501 | static int nilfs_iget_set(struct inode *inode, void *opaque) | |
502 | { | |
503 | struct nilfs_iget_args *args = opaque; | |
504 | ||
505 | inode->i_ino = args->ino; | |
506 | if (args->for_gc) { | |
507 | NILFS_I(inode)->i_state = 1 << NILFS_I_GCINODE; | |
508 | NILFS_I(inode)->i_cno = args->cno; | |
4d8d9293 RK |
509 | NILFS_I(inode)->i_root = NULL; |
510 | } else { | |
511 | if (args->root && args->ino == NILFS_ROOT_INO) | |
512 | nilfs_get_root(args->root); | |
513 | NILFS_I(inode)->i_root = args->root; | |
0e14a359 RK |
514 | } |
515 | return 0; | |
516 | } | |
517 | ||
4d8d9293 RK |
518 | struct inode *nilfs_iget(struct super_block *sb, struct nilfs_root *root, |
519 | unsigned long ino) | |
05fe58fd | 520 | { |
4d8d9293 RK |
521 | struct nilfs_iget_args args = { |
522 | .ino = ino, .root = root, .cno = 0, .for_gc = 0 | |
523 | }; | |
05fe58fd RK |
524 | struct inode *inode; |
525 | int err; | |
526 | ||
0e14a359 | 527 | inode = iget5_locked(sb, ino, nilfs_iget_test, nilfs_iget_set, &args); |
05fe58fd RK |
528 | if (unlikely(!inode)) |
529 | return ERR_PTR(-ENOMEM); | |
530 | if (!(inode->i_state & I_NEW)) | |
531 | return inode; | |
532 | ||
533 | err = __nilfs_read_inode(sb, ino, inode); | |
534 | if (unlikely(err)) { | |
535 | iget_failed(inode); | |
536 | return ERR_PTR(err); | |
537 | } | |
538 | unlock_new_inode(inode); | |
539 | return inode; | |
540 | } | |
541 | ||
263d90ce RK |
542 | struct inode *nilfs_iget_for_gc(struct super_block *sb, unsigned long ino, |
543 | __u64 cno) | |
544 | { | |
4d8d9293 RK |
545 | struct nilfs_iget_args args = { |
546 | .ino = ino, .root = NULL, .cno = cno, .for_gc = 1 | |
547 | }; | |
263d90ce RK |
548 | struct inode *inode; |
549 | int err; | |
550 | ||
551 | inode = iget5_locked(sb, ino, nilfs_iget_test, nilfs_iget_set, &args); | |
552 | if (unlikely(!inode)) | |
553 | return ERR_PTR(-ENOMEM); | |
554 | if (!(inode->i_state & I_NEW)) | |
555 | return inode; | |
556 | ||
557 | err = nilfs_init_gcinode(inode); | |
558 | if (unlikely(err)) { | |
559 | iget_failed(inode); | |
560 | return ERR_PTR(err); | |
561 | } | |
562 | unlock_new_inode(inode); | |
563 | return inode; | |
564 | } | |
565 | ||
05fe58fd RK |
566 | void nilfs_write_inode_common(struct inode *inode, |
567 | struct nilfs_inode *raw_inode, int has_bmap) | |
568 | { | |
569 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
570 | ||
571 | raw_inode->i_mode = cpu_to_le16(inode->i_mode); | |
572 | raw_inode->i_uid = cpu_to_le32(inode->i_uid); | |
573 | raw_inode->i_gid = cpu_to_le32(inode->i_gid); | |
574 | raw_inode->i_links_count = cpu_to_le16(inode->i_nlink); | |
575 | raw_inode->i_size = cpu_to_le64(inode->i_size); | |
576 | raw_inode->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); | |
577 | raw_inode->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec); | |
61239230 RK |
578 | raw_inode->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); |
579 | raw_inode->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); | |
05fe58fd RK |
580 | raw_inode->i_blocks = cpu_to_le64(inode->i_blocks); |
581 | ||
05fe58fd RK |
582 | raw_inode->i_flags = cpu_to_le32(ii->i_flags); |
583 | raw_inode->i_generation = cpu_to_le32(inode->i_generation); | |
584 | ||
585 | if (has_bmap) | |
586 | nilfs_bmap_write(ii->i_bmap, raw_inode); | |
587 | else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) | |
588 | raw_inode->i_device_code = | |
cdce214e | 589 | cpu_to_le64(huge_encode_dev(inode->i_rdev)); |
05fe58fd RK |
590 | /* When extending inode, nilfs->ns_inode_size should be checked |
591 | for substitutions of appended fields */ | |
592 | } | |
593 | ||
594 | void nilfs_update_inode(struct inode *inode, struct buffer_head *ibh) | |
595 | { | |
596 | ino_t ino = inode->i_ino; | |
597 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
598 | struct super_block *sb = inode->i_sb; | |
599 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
600 | struct nilfs_inode *raw_inode; | |
601 | ||
602 | raw_inode = nilfs_ifile_map_inode(sbi->s_ifile, ino, ibh); | |
603 | ||
05fe58fd RK |
604 | if (test_and_clear_bit(NILFS_I_NEW, &ii->i_state)) |
605 | memset(raw_inode, 0, NILFS_MDT(sbi->s_ifile)->mi_entry_size); | |
606 | set_bit(NILFS_I_INODE_DIRTY, &ii->i_state); | |
607 | ||
608 | nilfs_write_inode_common(inode, raw_inode, 0); | |
609 | /* XXX: call with has_bmap = 0 is a workaround to avoid | |
610 | deadlock of bmap. This delays update of i_bmap to just | |
611 | before writing */ | |
612 | nilfs_ifile_unmap_inode(sbi->s_ifile, ino, ibh); | |
613 | } | |
614 | ||
615 | #define NILFS_MAX_TRUNCATE_BLOCKS 16384 /* 64MB for 4KB block */ | |
616 | ||
617 | static void nilfs_truncate_bmap(struct nilfs_inode_info *ii, | |
618 | unsigned long from) | |
619 | { | |
620 | unsigned long b; | |
621 | int ret; | |
622 | ||
623 | if (!test_bit(NILFS_I_BMAP, &ii->i_state)) | |
624 | return; | |
625 | repeat: | |
626 | ret = nilfs_bmap_last_key(ii->i_bmap, &b); | |
627 | if (ret == -ENOENT) | |
628 | return; | |
629 | else if (ret < 0) | |
630 | goto failed; | |
631 | ||
632 | if (b < from) | |
633 | return; | |
634 | ||
635 | b -= min_t(unsigned long, NILFS_MAX_TRUNCATE_BLOCKS, b - from); | |
636 | ret = nilfs_bmap_truncate(ii->i_bmap, b); | |
637 | nilfs_relax_pressure_in_lock(ii->vfs_inode.i_sb); | |
638 | if (!ret || (ret == -ENOMEM && | |
639 | nilfs_bmap_truncate(ii->i_bmap, b) == 0)) | |
640 | goto repeat; | |
641 | ||
642 | failed: | |
643 | if (ret == -EINVAL) | |
644 | nilfs_error(ii->vfs_inode.i_sb, __func__, | |
645 | "bmap is broken (ino=%lu)", ii->vfs_inode.i_ino); | |
646 | else | |
647 | nilfs_warning(ii->vfs_inode.i_sb, __func__, | |
648 | "failed to truncate bmap (ino=%lu, err=%d)", | |
649 | ii->vfs_inode.i_ino, ret); | |
650 | } | |
651 | ||
652 | void nilfs_truncate(struct inode *inode) | |
653 | { | |
654 | unsigned long blkoff; | |
655 | unsigned int blocksize; | |
656 | struct nilfs_transaction_info ti; | |
657 | struct super_block *sb = inode->i_sb; | |
658 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
05fe58fd RK |
659 | |
660 | if (!test_bit(NILFS_I_BMAP, &ii->i_state)) | |
661 | return; | |
662 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
663 | return; | |
664 | ||
665 | blocksize = sb->s_blocksize; | |
666 | blkoff = (inode->i_size + blocksize - 1) >> sb->s_blocksize_bits; | |
1f5abe7e | 667 | nilfs_transaction_begin(sb, &ti, 0); /* never fails */ |
05fe58fd RK |
668 | |
669 | block_truncate_page(inode->i_mapping, inode->i_size, nilfs_get_block); | |
670 | ||
671 | nilfs_truncate_bmap(ii, blkoff); | |
672 | ||
673 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
674 | if (IS_SYNC(inode)) | |
675 | nilfs_set_transaction_flag(NILFS_TI_SYNC); | |
676 | ||
abdb318b | 677 | nilfs_mark_inode_dirty(inode); |
05fe58fd | 678 | nilfs_set_file_dirty(NILFS_SB(sb), inode, 0); |
47420c79 | 679 | nilfs_transaction_commit(sb); |
05fe58fd RK |
680 | /* May construct a logical segment and may fail in sync mode. |
681 | But truncate has no return value. */ | |
682 | } | |
683 | ||
6fd1e5c9 AV |
684 | static void nilfs_clear_inode(struct inode *inode) |
685 | { | |
686 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
687 | ||
688 | /* | |
689 | * Free resources allocated in nilfs_read_inode(), here. | |
690 | */ | |
691 | BUG_ON(!list_empty(&ii->i_dirty)); | |
692 | brelse(ii->i_bh); | |
693 | ii->i_bh = NULL; | |
694 | ||
695 | if (test_bit(NILFS_I_BMAP, &ii->i_state)) | |
696 | nilfs_bmap_clear(ii->i_bmap); | |
697 | ||
698 | nilfs_btnode_cache_clear(&ii->i_btnode_cache); | |
4d8d9293 RK |
699 | |
700 | if (ii->i_root && inode->i_ino == NILFS_ROOT_INO) | |
701 | nilfs_put_root(ii->i_root); | |
6fd1e5c9 AV |
702 | } |
703 | ||
704 | void nilfs_evict_inode(struct inode *inode) | |
05fe58fd RK |
705 | { |
706 | struct nilfs_transaction_info ti; | |
707 | struct super_block *sb = inode->i_sb; | |
708 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
05fe58fd | 709 | |
4d8d9293 | 710 | if (inode->i_nlink || !ii->i_root || unlikely(is_bad_inode(inode))) { |
05fe58fd RK |
711 | if (inode->i_data.nrpages) |
712 | truncate_inode_pages(&inode->i_data, 0); | |
6fd1e5c9 AV |
713 | end_writeback(inode); |
714 | nilfs_clear_inode(inode); | |
05fe58fd RK |
715 | return; |
716 | } | |
1f5abe7e RK |
717 | nilfs_transaction_begin(sb, &ti, 0); /* never fails */ |
718 | ||
05fe58fd RK |
719 | if (inode->i_data.nrpages) |
720 | truncate_inode_pages(&inode->i_data, 0); | |
721 | ||
722 | nilfs_truncate_bmap(ii, 0); | |
abdb318b | 723 | nilfs_mark_inode_dirty(inode); |
6fd1e5c9 AV |
724 | end_writeback(inode); |
725 | nilfs_clear_inode(inode); | |
05fe58fd RK |
726 | nilfs_free_inode(inode); |
727 | /* nilfs_free_inode() marks inode buffer dirty */ | |
728 | if (IS_SYNC(inode)) | |
729 | nilfs_set_transaction_flag(NILFS_TI_SYNC); | |
47420c79 | 730 | nilfs_transaction_commit(sb); |
05fe58fd RK |
731 | /* May construct a logical segment and may fail in sync mode. |
732 | But delete_inode has no return value. */ | |
733 | } | |
734 | ||
735 | int nilfs_setattr(struct dentry *dentry, struct iattr *iattr) | |
736 | { | |
737 | struct nilfs_transaction_info ti; | |
738 | struct inode *inode = dentry->d_inode; | |
739 | struct super_block *sb = inode->i_sb; | |
47420c79 | 740 | int err; |
05fe58fd RK |
741 | |
742 | err = inode_change_ok(inode, iattr); | |
743 | if (err) | |
744 | return err; | |
745 | ||
746 | err = nilfs_transaction_begin(sb, &ti, 0); | |
747 | if (unlikely(err)) | |
748 | return err; | |
1025774c CH |
749 | |
750 | if ((iattr->ia_valid & ATTR_SIZE) && | |
751 | iattr->ia_size != i_size_read(inode)) { | |
752 | err = vmtruncate(inode, iattr->ia_size); | |
753 | if (unlikely(err)) | |
754 | goto out_err; | |
755 | } | |
756 | ||
757 | setattr_copy(inode, iattr); | |
758 | mark_inode_dirty(inode); | |
759 | ||
760 | if (iattr->ia_valid & ATTR_MODE) { | |
05fe58fd | 761 | err = nilfs_acl_chmod(inode); |
1025774c CH |
762 | if (unlikely(err)) |
763 | goto out_err; | |
764 | } | |
765 | ||
766 | return nilfs_transaction_commit(sb); | |
47420c79 | 767 | |
1025774c CH |
768 | out_err: |
769 | nilfs_transaction_abort(sb); | |
47420c79 | 770 | return err; |
05fe58fd RK |
771 | } |
772 | ||
773 | int nilfs_load_inode_block(struct nilfs_sb_info *sbi, struct inode *inode, | |
774 | struct buffer_head **pbh) | |
775 | { | |
776 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
777 | int err; | |
778 | ||
779 | spin_lock(&sbi->s_inode_lock); | |
05fe58fd RK |
780 | if (ii->i_bh == NULL) { |
781 | spin_unlock(&sbi->s_inode_lock); | |
782 | err = nilfs_ifile_get_inode_block(sbi->s_ifile, inode->i_ino, | |
783 | pbh); | |
784 | if (unlikely(err)) | |
785 | return err; | |
786 | spin_lock(&sbi->s_inode_lock); | |
787 | if (ii->i_bh == NULL) | |
788 | ii->i_bh = *pbh; | |
789 | else { | |
790 | brelse(*pbh); | |
791 | *pbh = ii->i_bh; | |
792 | } | |
793 | } else | |
794 | *pbh = ii->i_bh; | |
795 | ||
796 | get_bh(*pbh); | |
797 | spin_unlock(&sbi->s_inode_lock); | |
798 | return 0; | |
799 | } | |
800 | ||
801 | int nilfs_inode_dirty(struct inode *inode) | |
802 | { | |
803 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
804 | struct nilfs_sb_info *sbi = NILFS_SB(inode->i_sb); | |
805 | int ret = 0; | |
806 | ||
807 | if (!list_empty(&ii->i_dirty)) { | |
808 | spin_lock(&sbi->s_inode_lock); | |
809 | ret = test_bit(NILFS_I_DIRTY, &ii->i_state) || | |
810 | test_bit(NILFS_I_BUSY, &ii->i_state); | |
811 | spin_unlock(&sbi->s_inode_lock); | |
812 | } | |
813 | return ret; | |
814 | } | |
815 | ||
816 | int nilfs_set_file_dirty(struct nilfs_sb_info *sbi, struct inode *inode, | |
817 | unsigned nr_dirty) | |
818 | { | |
819 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
820 | ||
821 | atomic_add(nr_dirty, &sbi->s_nilfs->ns_ndirtyblks); | |
822 | ||
458c5b08 | 823 | if (test_and_set_bit(NILFS_I_DIRTY, &ii->i_state)) |
05fe58fd RK |
824 | return 0; |
825 | ||
826 | spin_lock(&sbi->s_inode_lock); | |
827 | if (!test_bit(NILFS_I_QUEUED, &ii->i_state) && | |
828 | !test_bit(NILFS_I_BUSY, &ii->i_state)) { | |
829 | /* Because this routine may race with nilfs_dispose_list(), | |
830 | we have to check NILFS_I_QUEUED here, too. */ | |
831 | if (list_empty(&ii->i_dirty) && igrab(inode) == NULL) { | |
832 | /* This will happen when somebody is freeing | |
833 | this inode. */ | |
834 | nilfs_warning(sbi->s_super, __func__, | |
835 | "cannot get inode (ino=%lu)\n", | |
836 | inode->i_ino); | |
837 | spin_unlock(&sbi->s_inode_lock); | |
838 | return -EINVAL; /* NILFS_I_DIRTY may remain for | |
839 | freeing inode */ | |
840 | } | |
841 | list_del(&ii->i_dirty); | |
842 | list_add_tail(&ii->i_dirty, &sbi->s_dirty_files); | |
843 | set_bit(NILFS_I_QUEUED, &ii->i_state); | |
844 | } | |
845 | spin_unlock(&sbi->s_inode_lock); | |
846 | return 0; | |
847 | } | |
848 | ||
849 | int nilfs_mark_inode_dirty(struct inode *inode) | |
850 | { | |
851 | struct nilfs_sb_info *sbi = NILFS_SB(inode->i_sb); | |
852 | struct buffer_head *ibh; | |
853 | int err; | |
854 | ||
855 | err = nilfs_load_inode_block(sbi, inode, &ibh); | |
856 | if (unlikely(err)) { | |
857 | nilfs_warning(inode->i_sb, __func__, | |
858 | "failed to reget inode block.\n"); | |
859 | return err; | |
860 | } | |
05fe58fd | 861 | nilfs_update_inode(inode, ibh); |
05fe58fd RK |
862 | nilfs_mdt_mark_buffer_dirty(ibh); |
863 | nilfs_mdt_mark_dirty(sbi->s_ifile); | |
864 | brelse(ibh); | |
865 | return 0; | |
866 | } | |
867 | ||
868 | /** | |
869 | * nilfs_dirty_inode - reflect changes on given inode to an inode block. | |
870 | * @inode: inode of the file to be registered. | |
871 | * | |
872 | * nilfs_dirty_inode() loads a inode block containing the specified | |
873 | * @inode and copies data from a nilfs_inode to a corresponding inode | |
874 | * entry in the inode block. This operation is excluded from the segment | |
875 | * construction. This function can be called both as a single operation | |
876 | * and as a part of indivisible file operations. | |
877 | */ | |
878 | void nilfs_dirty_inode(struct inode *inode) | |
879 | { | |
880 | struct nilfs_transaction_info ti; | |
7d6cd92f | 881 | struct nilfs_mdt_info *mdi = NILFS_MDT(inode); |
05fe58fd RK |
882 | |
883 | if (is_bad_inode(inode)) { | |
884 | nilfs_warning(inode->i_sb, __func__, | |
885 | "tried to mark bad_inode dirty. ignored.\n"); | |
886 | dump_stack(); | |
887 | return; | |
888 | } | |
7d6cd92f RK |
889 | if (mdi) { |
890 | nilfs_mdt_mark_dirty(inode); | |
891 | return; | |
892 | } | |
05fe58fd | 893 | nilfs_transaction_begin(inode->i_sb, &ti, 0); |
458c5b08 | 894 | nilfs_mark_inode_dirty(inode); |
47420c79 | 895 | nilfs_transaction_commit(inode->i_sb); /* never fails */ |
05fe58fd | 896 | } |