Commit | Line | Data |
---|---|---|
ae98043f | 1 | // SPDX-License-Identifier: GPL-2.0+ |
783f6184 | 2 | /* |
94ee1d91 | 3 | * NILFS module and super block management. |
783f6184 RK |
4 | * |
5 | * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation. | |
6 | * | |
4b420ab4 | 7 | * Written by Ryusuke Konishi. |
783f6184 RK |
8 | */ |
9 | /* | |
10 | * linux/fs/ext2/super.c | |
11 | * | |
12 | * Copyright (C) 1992, 1993, 1994, 1995 | |
13 | * Remy Card (card@masi.ibp.fr) | |
14 | * Laboratoire MASI - Institut Blaise Pascal | |
15 | * Universite Pierre et Marie Curie (Paris VI) | |
16 | * | |
17 | * from | |
18 | * | |
19 | * linux/fs/minix/inode.c | |
20 | * | |
21 | * Copyright (C) 1991, 1992 Linus Torvalds | |
22 | * | |
23 | * Big-endian to little-endian byte-swapping/bitmaps by | |
24 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
25 | */ | |
26 | ||
27 | #include <linux/module.h> | |
28 | #include <linux/string.h> | |
29 | #include <linux/slab.h> | |
30 | #include <linux/init.h> | |
31 | #include <linux/blkdev.h> | |
783f6184 | 32 | #include <linux/crc32.h> |
783f6184 RK |
33 | #include <linux/vfs.h> |
34 | #include <linux/writeback.h> | |
b58a285b JS |
35 | #include <linux/seq_file.h> |
36 | #include <linux/mount.h> | |
c1e012ea | 37 | #include <linux/fs_context.h> |
36defdd9 | 38 | #include <linux/fs_parser.h> |
783f6184 | 39 | #include "nilfs.h" |
8e656fd5 | 40 | #include "export.h" |
783f6184 RK |
41 | #include "mdt.h" |
42 | #include "alloc.h" | |
05d0e94b RK |
43 | #include "btree.h" |
44 | #include "btnode.h" | |
783f6184 RK |
45 | #include "page.h" |
46 | #include "cpfile.h" | |
4e33f9ea | 47 | #include "sufile.h" /* nilfs_sufile_resize(), nilfs_sufile_set_alloc_range() */ |
783f6184 RK |
48 | #include "ifile.h" |
49 | #include "dat.h" | |
50 | #include "segment.h" | |
51 | #include "segbuf.h" | |
52 | ||
53 | MODULE_AUTHOR("NTT Corp."); | |
54 | MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem " | |
55 | "(NILFS)"); | |
783f6184 RK |
56 | MODULE_LICENSE("GPL"); |
57 | ||
abc0b50b | 58 | static struct kmem_cache *nilfs_inode_cachep; |
41c88bd7 LH |
59 | struct kmem_cache *nilfs_transaction_cachep; |
60 | struct kmem_cache *nilfs_segbuf_cachep; | |
61 | struct kmem_cache *nilfs_btree_path_cache; | |
62 | ||
f7545144 | 63 | static int nilfs_setup_super(struct super_block *sb, int is_mount); |
783f6184 | 64 | |
2987a4cf | 65 | void __nilfs_msg(struct super_block *sb, const char *fmt, ...) |
a66dfb0a RK |
66 | { |
67 | struct va_format vaf; | |
68 | va_list args; | |
2987a4cf | 69 | int level; |
a66dfb0a RK |
70 | |
71 | va_start(args, fmt); | |
2987a4cf JP |
72 | |
73 | level = printk_get_level(fmt); | |
74 | vaf.fmt = printk_skip_level(fmt); | |
a66dfb0a | 75 | vaf.va = &args; |
2987a4cf | 76 | |
a66dfb0a | 77 | if (sb) |
2987a4cf JP |
78 | printk("%c%cNILFS (%s): %pV\n", |
79 | KERN_SOH_ASCII, level, sb->s_id, &vaf); | |
a66dfb0a | 80 | else |
2987a4cf JP |
81 | printk("%c%cNILFS: %pV\n", |
82 | KERN_SOH_ASCII, level, &vaf); | |
83 | ||
a66dfb0a RK |
84 | va_end(args); |
85 | } | |
86 | ||
f7545144 | 87 | static void nilfs_set_error(struct super_block *sb) |
c8a11c8a | 88 | { |
e3154e97 | 89 | struct the_nilfs *nilfs = sb->s_fs_info; |
d26493b6 | 90 | struct nilfs_super_block **sbp; |
c8a11c8a RK |
91 | |
92 | down_write(&nilfs->ns_sem); | |
93 | if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) { | |
94 | nilfs->ns_mount_state |= NILFS_ERROR_FS; | |
f7545144 | 95 | sbp = nilfs_prepare_super(sb, 0); |
d26493b6 JS |
96 | if (likely(sbp)) { |
97 | sbp[0]->s_state |= cpu_to_le16(NILFS_ERROR_FS); | |
b2ac86e1 JS |
98 | if (sbp[1]) |
99 | sbp[1]->s_state |= cpu_to_le16(NILFS_ERROR_FS); | |
f7545144 | 100 | nilfs_commit_super(sb, NILFS_SB_COMMIT_ALL); |
d26493b6 | 101 | } |
c8a11c8a RK |
102 | } |
103 | up_write(&nilfs->ns_sem); | |
104 | } | |
105 | ||
783f6184 | 106 | /** |
cae3d4ca | 107 | * __nilfs_error() - report failure condition on a filesystem |
b79bdfdd RK |
108 | * @sb: super block instance |
109 | * @function: name of calling function | |
110 | * @fmt: format string for message to be output | |
111 | * @...: optional arguments to @fmt | |
783f6184 | 112 | * |
cae3d4ca RK |
113 | * __nilfs_error() sets an ERROR_FS flag on the superblock as well as |
114 | * reporting an error message. This function should be called when | |
115 | * NILFS detects incoherences or defects of meta data on disk. | |
783f6184 | 116 | * |
cae3d4ca RK |
117 | * This implements the body of nilfs_error() macro. Normally, |
118 | * nilfs_error() should be used. As for sustainable errors such as a | |
a1d0747a | 119 | * single-shot I/O error, nilfs_err() should be used instead. |
cae3d4ca RK |
120 | * |
121 | * Callers should not add a trailing newline since this will do it. | |
783f6184 | 122 | */ |
cae3d4ca RK |
123 | void __nilfs_error(struct super_block *sb, const char *function, |
124 | const char *fmt, ...) | |
783f6184 | 125 | { |
e3154e97 | 126 | struct the_nilfs *nilfs = sb->s_fs_info; |
b004a5eb | 127 | struct va_format vaf; |
783f6184 RK |
128 | va_list args; |
129 | ||
130 | va_start(args, fmt); | |
b004a5eb JP |
131 | |
132 | vaf.fmt = fmt; | |
133 | vaf.va = &args; | |
134 | ||
135 | printk(KERN_CRIT "NILFS error (device %s): %s: %pV\n", | |
136 | sb->s_id, function, &vaf); | |
137 | ||
783f6184 RK |
138 | va_end(args); |
139 | ||
bc98a42c | 140 | if (!sb_rdonly(sb)) { |
f7545144 | 141 | nilfs_set_error(sb); |
783f6184 | 142 | |
3b2ce58b | 143 | if (nilfs_test_opt(nilfs, ERRORS_RO)) { |
783f6184 | 144 | printk(KERN_CRIT "Remounting filesystem read-only\n"); |
1751e8a6 | 145 | sb->s_flags |= SB_RDONLY; |
783f6184 RK |
146 | } |
147 | } | |
148 | ||
3b2ce58b | 149 | if (nilfs_test_opt(nilfs, ERRORS_PANIC)) |
783f6184 RK |
150 | panic("NILFS (device %s): panic forced after error\n", |
151 | sb->s_id); | |
152 | } | |
153 | ||
2879ed66 | 154 | struct inode *nilfs_alloc_inode(struct super_block *sb) |
783f6184 RK |
155 | { |
156 | struct nilfs_inode_info *ii; | |
157 | ||
fd60b288 | 158 | ii = alloc_inode_sb(sb, nilfs_inode_cachep, GFP_NOFS); |
783f6184 RK |
159 | if (!ii) |
160 | return NULL; | |
161 | ii->i_bh = NULL; | |
162 | ii->i_state = 0; | |
d7cee0b3 | 163 | ii->i_type = 0; |
0e14a359 | 164 | ii->i_cno = 0; |
e897be17 RK |
165 | ii->i_assoc_inode = NULL; |
166 | ii->i_bmap = &ii->i_bmap_data; | |
783f6184 RK |
167 | return &ii->vfs_inode; |
168 | } | |
169 | ||
977c3d18 | 170 | static void nilfs_free_inode(struct inode *inode) |
783f6184 | 171 | { |
2d19961d RK |
172 | if (nilfs_is_metadata_file_inode(inode)) |
173 | nilfs_mdt_destroy(inode); | |
174 | ||
783f6184 RK |
175 | kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode)); |
176 | } | |
177 | ||
f7545144 | 178 | static int nilfs_sync_super(struct super_block *sb, int flag) |
783f6184 | 179 | { |
e3154e97 | 180 | struct the_nilfs *nilfs = sb->s_fs_info; |
783f6184 | 181 | int err; |
783f6184 | 182 | |
783f6184 | 183 | retry: |
e339ad31 | 184 | set_buffer_dirty(nilfs->ns_sbh[0]); |
3b2ce58b | 185 | if (nilfs_test_opt(nilfs, BARRIER)) { |
87e99511 | 186 | err = __sync_dirty_buffer(nilfs->ns_sbh[0], |
70fd7614 | 187 | REQ_SYNC | REQ_PREFLUSH | REQ_FUA); |
87e99511 CH |
188 | } else { |
189 | err = sync_dirty_buffer(nilfs->ns_sbh[0]); | |
783f6184 | 190 | } |
87e99511 | 191 | |
e339ad31 | 192 | if (unlikely(err)) { |
a1d0747a | 193 | nilfs_err(sb, "unable to write superblock: err=%d", err); |
e339ad31 | 194 | if (err == -EIO && nilfs->ns_sbh[1]) { |
b2ac86e1 JS |
195 | /* |
196 | * sbp[0] points to newer log than sbp[1], | |
197 | * so copy sbp[0] to sbp[1] to take over sbp[0]. | |
198 | */ | |
199 | memcpy(nilfs->ns_sbp[1], nilfs->ns_sbp[0], | |
200 | nilfs->ns_sbsize); | |
e339ad31 RK |
201 | nilfs_fall_back_super_block(nilfs); |
202 | goto retry; | |
203 | } | |
204 | } else { | |
205 | struct nilfs_super_block *sbp = nilfs->ns_sbp[0]; | |
206 | ||
b2ac86e1 JS |
207 | nilfs->ns_sbwcount++; |
208 | ||
e339ad31 RK |
209 | /* |
210 | * The latest segment becomes trailable from the position | |
211 | * written in superblock. | |
212 | */ | |
783f6184 | 213 | clear_nilfs_discontinued(nilfs); |
e339ad31 RK |
214 | |
215 | /* update GC protection for recent segments */ | |
216 | if (nilfs->ns_sbh[1]) { | |
b2ac86e1 | 217 | if (flag == NILFS_SB_COMMIT_ALL) { |
e339ad31 | 218 | set_buffer_dirty(nilfs->ns_sbh[1]); |
b2ac86e1 JS |
219 | if (sync_dirty_buffer(nilfs->ns_sbh[1]) < 0) |
220 | goto out; | |
e339ad31 | 221 | } |
b2ac86e1 JS |
222 | if (le64_to_cpu(nilfs->ns_sbp[1]->s_last_cno) < |
223 | le64_to_cpu(nilfs->ns_sbp[0]->s_last_cno)) | |
224 | sbp = nilfs->ns_sbp[1]; | |
e339ad31 | 225 | } |
783f6184 | 226 | |
b2ac86e1 JS |
227 | spin_lock(&nilfs->ns_last_segment_lock); |
228 | nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq); | |
229 | spin_unlock(&nilfs->ns_last_segment_lock); | |
230 | } | |
231 | out: | |
783f6184 RK |
232 | return err; |
233 | } | |
234 | ||
60f46b7e RK |
235 | void nilfs_set_log_cursor(struct nilfs_super_block *sbp, |
236 | struct the_nilfs *nilfs) | |
237 | { | |
238 | sector_t nfreeblocks; | |
239 | ||
240 | /* nilfs->ns_sem must be locked by the caller. */ | |
241 | nilfs_count_free_blocks(nilfs, &nfreeblocks); | |
242 | sbp->s_free_blocks_count = cpu_to_le64(nfreeblocks); | |
243 | ||
244 | spin_lock(&nilfs->ns_last_segment_lock); | |
245 | sbp->s_last_seq = cpu_to_le64(nilfs->ns_last_seq); | |
246 | sbp->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg); | |
247 | sbp->s_last_cno = cpu_to_le64(nilfs->ns_last_cno); | |
248 | spin_unlock(&nilfs->ns_last_segment_lock); | |
249 | } | |
250 | ||
f7545144 | 251 | struct nilfs_super_block **nilfs_prepare_super(struct super_block *sb, |
b2ac86e1 | 252 | int flip) |
783f6184 | 253 | { |
e3154e97 | 254 | struct the_nilfs *nilfs = sb->s_fs_info; |
e339ad31 | 255 | struct nilfs_super_block **sbp = nilfs->ns_sbp; |
783f6184 | 256 | |
d26493b6 | 257 | /* nilfs->ns_sem must be locked by the caller. */ |
34cb9b5c | 258 | if (sbp[0]->s_magic != cpu_to_le16(NILFS_SUPER_MAGIC)) { |
d26493b6 JS |
259 | if (sbp[1] && |
260 | sbp[1]->s_magic == cpu_to_le16(NILFS_SUPER_MAGIC)) { | |
b2ac86e1 | 261 | memcpy(sbp[0], sbp[1], nilfs->ns_sbsize); |
d26493b6 | 262 | } else { |
a1d0747a | 263 | nilfs_crit(sb, "superblock broke"); |
d26493b6 | 264 | return NULL; |
e339ad31 | 265 | } |
b2ac86e1 JS |
266 | } else if (sbp[1] && |
267 | sbp[1]->s_magic != cpu_to_le16(NILFS_SUPER_MAGIC)) { | |
7592ecde | 268 | memcpy(sbp[1], sbp[0], nilfs->ns_sbsize); |
e339ad31 | 269 | } |
b2ac86e1 JS |
270 | |
271 | if (flip && sbp[1]) | |
272 | nilfs_swap_super_block(nilfs); | |
273 | ||
d26493b6 JS |
274 | return sbp; |
275 | } | |
276 | ||
f7545144 | 277 | int nilfs_commit_super(struct super_block *sb, int flag) |
d26493b6 | 278 | { |
e3154e97 | 279 | struct the_nilfs *nilfs = sb->s_fs_info; |
d26493b6 | 280 | struct nilfs_super_block **sbp = nilfs->ns_sbp; |
fb04b91b | 281 | time64_t t; |
d26493b6 JS |
282 | |
283 | /* nilfs->ns_sem must be locked by the caller. */ | |
fb04b91b | 284 | t = ktime_get_real_seconds(); |
b2ac86e1 | 285 | nilfs->ns_sbwtime = t; |
e339ad31 RK |
286 | sbp[0]->s_wtime = cpu_to_le64(t); |
287 | sbp[0]->s_sum = 0; | |
288 | sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed, | |
289 | (unsigned char *)sbp[0], | |
290 | nilfs->ns_sbsize)); | |
b2ac86e1 JS |
291 | if (flag == NILFS_SB_COMMIT_ALL && sbp[1]) { |
292 | sbp[1]->s_wtime = sbp[0]->s_wtime; | |
293 | sbp[1]->s_sum = 0; | |
294 | sbp[1]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed, | |
295 | (unsigned char *)sbp[1], | |
296 | nilfs->ns_sbsize)); | |
e339ad31 | 297 | } |
e605f0a7 | 298 | clear_nilfs_sb_dirty(nilfs); |
e2c7617a AR |
299 | nilfs->ns_flushed_device = 1; |
300 | /* make sure store to ns_flushed_device cannot be reordered */ | |
301 | smp_wmb(); | |
f7545144 | 302 | return nilfs_sync_super(sb, flag); |
783f6184 RK |
303 | } |
304 | ||
7ecaa46c RK |
305 | /** |
306 | * nilfs_cleanup_super() - write filesystem state for cleanup | |
f7545144 | 307 | * @sb: super block instance to be unmounted or degraded to read-only |
7ecaa46c RK |
308 | * |
309 | * This function restores state flags in the on-disk super block. | |
310 | * This will set "clean" flag (i.e. NILFS_VALID_FS) unless the | |
311 | * filesystem was not clean previously. | |
312 | */ | |
f7545144 | 313 | int nilfs_cleanup_super(struct super_block *sb) |
7ecaa46c | 314 | { |
e3154e97 | 315 | struct the_nilfs *nilfs = sb->s_fs_info; |
d26493b6 | 316 | struct nilfs_super_block **sbp; |
b2ac86e1 | 317 | int flag = NILFS_SB_COMMIT; |
d26493b6 | 318 | int ret = -EIO; |
7ecaa46c | 319 | |
f7545144 | 320 | sbp = nilfs_prepare_super(sb, 0); |
d26493b6 | 321 | if (sbp) { |
f7545144 RK |
322 | sbp[0]->s_state = cpu_to_le16(nilfs->ns_mount_state); |
323 | nilfs_set_log_cursor(sbp[0], nilfs); | |
b2ac86e1 JS |
324 | if (sbp[1] && sbp[0]->s_last_cno == sbp[1]->s_last_cno) { |
325 | /* | |
326 | * make the "clean" flag also to the opposite | |
327 | * super block if both super blocks point to | |
328 | * the same checkpoint. | |
329 | */ | |
330 | sbp[1]->s_state = sbp[0]->s_state; | |
331 | flag = NILFS_SB_COMMIT_ALL; | |
332 | } | |
f7545144 | 333 | ret = nilfs_commit_super(sb, flag); |
d26493b6 | 334 | } |
7ecaa46c RK |
335 | return ret; |
336 | } | |
337 | ||
cfb0a4bf RK |
338 | /** |
339 | * nilfs_move_2nd_super - relocate secondary super block | |
340 | * @sb: super block instance | |
341 | * @sb2off: new offset of the secondary super block (in bytes) | |
342 | */ | |
343 | static int nilfs_move_2nd_super(struct super_block *sb, loff_t sb2off) | |
344 | { | |
345 | struct the_nilfs *nilfs = sb->s_fs_info; | |
346 | struct buffer_head *nsbh; | |
347 | struct nilfs_super_block *nsbp; | |
348 | sector_t blocknr, newblocknr; | |
349 | unsigned long offset; | |
4f05028f | 350 | int sb2i; /* array index of the secondary superblock */ |
cfb0a4bf RK |
351 | int ret = 0; |
352 | ||
353 | /* nilfs->ns_sem must be locked by the caller. */ | |
354 | if (nilfs->ns_sbh[1] && | |
355 | nilfs->ns_sbh[1]->b_blocknr > nilfs->ns_first_data_block) { | |
356 | sb2i = 1; | |
357 | blocknr = nilfs->ns_sbh[1]->b_blocknr; | |
358 | } else if (nilfs->ns_sbh[0]->b_blocknr > nilfs->ns_first_data_block) { | |
359 | sb2i = 0; | |
360 | blocknr = nilfs->ns_sbh[0]->b_blocknr; | |
4f05028f RK |
361 | } else { |
362 | sb2i = -1; | |
363 | blocknr = 0; | |
cfb0a4bf RK |
364 | } |
365 | if (sb2i >= 0 && (u64)blocknr << nilfs->ns_blocksize_bits == sb2off) | |
366 | goto out; /* super block location is unchanged */ | |
367 | ||
368 | /* Get new super block buffer */ | |
369 | newblocknr = sb2off >> nilfs->ns_blocksize_bits; | |
370 | offset = sb2off & (nilfs->ns_blocksize - 1); | |
371 | nsbh = sb_getblk(sb, newblocknr); | |
372 | if (!nsbh) { | |
a1d0747a JP |
373 | nilfs_warn(sb, |
374 | "unable to move secondary superblock to block %llu", | |
375 | (unsigned long long)newblocknr); | |
cfb0a4bf RK |
376 | ret = -EIO; |
377 | goto out; | |
378 | } | |
379 | nsbp = (void *)nsbh->b_data + offset; | |
cfb0a4bf | 380 | |
679bd7eb | 381 | lock_buffer(nsbh); |
cfb0a4bf | 382 | if (sb2i >= 0) { |
679bd7eb RK |
383 | /* |
384 | * The position of the second superblock only changes by 4KiB, | |
385 | * which is larger than the maximum superblock data size | |
386 | * (= 1KiB), so there is no need to use memmove() to allow | |
387 | * overlap between source and destination. | |
388 | */ | |
cfb0a4bf | 389 | memcpy(nsbp, nilfs->ns_sbp[sb2i], nilfs->ns_sbsize); |
679bd7eb RK |
390 | |
391 | /* | |
392 | * Zero fill after copy to avoid overwriting in case of move | |
393 | * within the same block. | |
394 | */ | |
395 | memset(nsbh->b_data, 0, offset); | |
396 | memset((void *)nsbp + nilfs->ns_sbsize, 0, | |
397 | nsbh->b_size - offset - nilfs->ns_sbsize); | |
398 | } else { | |
399 | memset(nsbh->b_data, 0, nsbh->b_size); | |
400 | } | |
401 | set_buffer_uptodate(nsbh); | |
402 | unlock_buffer(nsbh); | |
403 | ||
404 | if (sb2i >= 0) { | |
cfb0a4bf RK |
405 | brelse(nilfs->ns_sbh[sb2i]); |
406 | nilfs->ns_sbh[sb2i] = nsbh; | |
407 | nilfs->ns_sbp[sb2i] = nsbp; | |
408 | } else if (nilfs->ns_sbh[0]->b_blocknr < nilfs->ns_first_data_block) { | |
409 | /* secondary super block will be restored to index 1 */ | |
410 | nilfs->ns_sbh[1] = nsbh; | |
411 | nilfs->ns_sbp[1] = nsbp; | |
412 | } else { | |
413 | brelse(nsbh); | |
414 | } | |
415 | out: | |
416 | return ret; | |
417 | } | |
418 | ||
4e33f9ea RK |
419 | /** |
420 | * nilfs_resize_fs - resize the filesystem | |
421 | * @sb: super block instance | |
422 | * @newsize: new size of the filesystem (in bytes) | |
423 | */ | |
424 | int nilfs_resize_fs(struct super_block *sb, __u64 newsize) | |
425 | { | |
426 | struct the_nilfs *nilfs = sb->s_fs_info; | |
427 | struct nilfs_super_block **sbp; | |
428 | __u64 devsize, newnsegs; | |
429 | loff_t sb2off; | |
430 | int ret; | |
431 | ||
432 | ret = -ERANGE; | |
4fcd6979 | 433 | devsize = bdev_nr_bytes(sb->s_bdev); |
4e33f9ea RK |
434 | if (newsize > devsize) |
435 | goto out; | |
436 | ||
99b9402a RK |
437 | /* |
438 | * Prevent underflow in second superblock position calculation. | |
439 | * The exact minimum size check is done in nilfs_sufile_resize(). | |
440 | */ | |
441 | if (newsize < 4096) { | |
442 | ret = -ENOSPC; | |
443 | goto out; | |
444 | } | |
445 | ||
4e33f9ea RK |
446 | /* |
447 | * Write lock is required to protect some functions depending | |
448 | * on the number of segments, the number of reserved segments, | |
449 | * and so forth. | |
450 | */ | |
451 | down_write(&nilfs->ns_segctor_sem); | |
452 | ||
453 | sb2off = NILFS_SB2_OFFSET_BYTES(newsize); | |
454 | newnsegs = sb2off >> nilfs->ns_blocksize_bits; | |
adc2c8d0 | 455 | newnsegs = div64_ul(newnsegs, nilfs->ns_blocks_per_segment); |
4e33f9ea RK |
456 | |
457 | ret = nilfs_sufile_resize(nilfs->ns_sufile, newnsegs); | |
458 | up_write(&nilfs->ns_segctor_sem); | |
459 | if (ret < 0) | |
460 | goto out; | |
461 | ||
462 | ret = nilfs_construct_segment(sb); | |
463 | if (ret < 0) | |
464 | goto out; | |
465 | ||
466 | down_write(&nilfs->ns_sem); | |
467 | nilfs_move_2nd_super(sb, sb2off); | |
468 | ret = -EIO; | |
469 | sbp = nilfs_prepare_super(sb, 0); | |
470 | if (likely(sbp)) { | |
471 | nilfs_set_log_cursor(sbp[0], nilfs); | |
472 | /* | |
473 | * Drop NILFS_RESIZE_FS flag for compatibility with | |
474 | * mount-time resize which may be implemented in a | |
475 | * future release. | |
476 | */ | |
477 | sbp[0]->s_state = cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & | |
478 | ~NILFS_RESIZE_FS); | |
479 | sbp[0]->s_dev_size = cpu_to_le64(newsize); | |
480 | sbp[0]->s_nsegments = cpu_to_le64(nilfs->ns_nsegments); | |
481 | if (sbp[1]) | |
482 | memcpy(sbp[1], sbp[0], nilfs->ns_sbsize); | |
483 | ret = nilfs_commit_super(sb, NILFS_SB_COMMIT_ALL); | |
484 | } | |
485 | up_write(&nilfs->ns_sem); | |
486 | ||
487 | /* | |
488 | * Reset the range of allocatable segments last. This order | |
489 | * is important in the case of expansion because the secondary | |
490 | * superblock must be protected from log write until migration | |
491 | * completes. | |
492 | */ | |
493 | if (!ret) | |
494 | nilfs_sufile_set_alloc_range(nilfs->ns_sufile, 0, newnsegs - 1); | |
495 | out: | |
496 | return ret; | |
497 | } | |
498 | ||
783f6184 RK |
499 | static void nilfs_put_super(struct super_block *sb) |
500 | { | |
e3154e97 | 501 | struct the_nilfs *nilfs = sb->s_fs_info; |
783f6184 | 502 | |
f7545144 | 503 | nilfs_detach_log_writer(sb); |
783f6184 | 504 | |
bc98a42c | 505 | if (!sb_rdonly(sb)) { |
783f6184 | 506 | down_write(&nilfs->ns_sem); |
f7545144 | 507 | nilfs_cleanup_super(sb); |
783f6184 RK |
508 | up_write(&nilfs->ns_sem); |
509 | } | |
510 | ||
42560f9c | 511 | nilfs_sysfs_delete_device_group(nilfs); |
f1e89c86 RK |
512 | iput(nilfs->ns_sufile); |
513 | iput(nilfs->ns_cpfile); | |
514 | iput(nilfs->ns_dat); | |
515 | ||
348fe8da | 516 | destroy_nilfs(nilfs); |
783f6184 | 517 | sb->s_fs_info = NULL; |
783f6184 RK |
518 | } |
519 | ||
783f6184 RK |
520 | static int nilfs_sync_fs(struct super_block *sb, int wait) |
521 | { | |
e3154e97 | 522 | struct the_nilfs *nilfs = sb->s_fs_info; |
d26493b6 | 523 | struct nilfs_super_block **sbp; |
783f6184 RK |
524 | int err = 0; |
525 | ||
526 | /* This function is called when super block should be written back */ | |
527 | if (wait) | |
528 | err = nilfs_construct_segment(sb); | |
6233caa9 JS |
529 | |
530 | down_write(&nilfs->ns_sem); | |
d26493b6 | 531 | if (nilfs_sb_dirty(nilfs)) { |
f7545144 | 532 | sbp = nilfs_prepare_super(sb, nilfs_sb_will_flip(nilfs)); |
b2ac86e1 JS |
533 | if (likely(sbp)) { |
534 | nilfs_set_log_cursor(sbp[0], nilfs); | |
f7545144 | 535 | nilfs_commit_super(sb, NILFS_SB_COMMIT); |
b2ac86e1 | 536 | } |
d26493b6 | 537 | } |
6233caa9 JS |
538 | up_write(&nilfs->ns_sem); |
539 | ||
e2c7617a AR |
540 | if (!err) |
541 | err = nilfs_flush_device(nilfs); | |
542 | ||
783f6184 RK |
543 | return err; |
544 | } | |
545 | ||
f7545144 | 546 | int nilfs_attach_checkpoint(struct super_block *sb, __u64 cno, int curr_mnt, |
4d8d9293 | 547 | struct nilfs_root **rootp) |
783f6184 | 548 | { |
e3154e97 | 549 | struct the_nilfs *nilfs = sb->s_fs_info; |
4d8d9293 | 550 | struct nilfs_root *root; |
4d8d9293 | 551 | int err = -ENOMEM; |
783f6184 | 552 | |
4d8d9293 RK |
553 | root = nilfs_find_or_create_root( |
554 | nilfs, curr_mnt ? NILFS_CPTREE_CURRENT_CNO : cno); | |
555 | if (!root) | |
556 | return err; | |
783f6184 | 557 | |
e912a5b6 RK |
558 | if (root->ifile) |
559 | goto reuse; /* already attached checkpoint */ | |
783f6184 | 560 | |
1154ecbd | 561 | down_read(&nilfs->ns_segctor_sem); |
1244a6d7 | 562 | err = nilfs_ifile_read(sb, root, cno, nilfs->ns_inode_size); |
1154ecbd | 563 | up_read(&nilfs->ns_segctor_sem); |
1244a6d7 | 564 | if (unlikely(err)) |
783f6184 | 565 | goto failed; |
4d8d9293 | 566 | |
e912a5b6 | 567 | reuse: |
4d8d9293 | 568 | *rootp = root; |
783f6184 RK |
569 | return 0; |
570 | ||
783f6184 | 571 | failed: |
1244a6d7 RK |
572 | if (err == -EINVAL) |
573 | nilfs_err(sb, "Invalid checkpoint (checkpoint number=%llu)", | |
574 | (unsigned long long)cno); | |
4d8d9293 | 575 | nilfs_put_root(root); |
783f6184 RK |
576 | |
577 | return err; | |
578 | } | |
783f6184 | 579 | |
5beb6e0b RK |
580 | static int nilfs_freeze(struct super_block *sb) |
581 | { | |
e3154e97 | 582 | struct the_nilfs *nilfs = sb->s_fs_info; |
5beb6e0b RK |
583 | int err; |
584 | ||
bc98a42c | 585 | if (sb_rdonly(sb)) |
5beb6e0b | 586 | return 0; |
783f6184 | 587 | |
5beb6e0b RK |
588 | /* Mark super block clean */ |
589 | down_write(&nilfs->ns_sem); | |
f7545144 | 590 | err = nilfs_cleanup_super(sb); |
5beb6e0b | 591 | up_write(&nilfs->ns_sem); |
783f6184 RK |
592 | return err; |
593 | } | |
594 | ||
5beb6e0b | 595 | static int nilfs_unfreeze(struct super_block *sb) |
783f6184 | 596 | { |
e3154e97 | 597 | struct the_nilfs *nilfs = sb->s_fs_info; |
783f6184 | 598 | |
bc98a42c | 599 | if (sb_rdonly(sb)) |
5beb6e0b RK |
600 | return 0; |
601 | ||
602 | down_write(&nilfs->ns_sem); | |
f7545144 | 603 | nilfs_setup_super(sb, false); |
5beb6e0b RK |
604 | up_write(&nilfs->ns_sem); |
605 | return 0; | |
783f6184 RK |
606 | } |
607 | ||
783f6184 RK |
608 | static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
609 | { | |
610 | struct super_block *sb = dentry->d_sb; | |
2b0143b5 | 611 | struct nilfs_root *root = NILFS_I(d_inode(dentry))->i_root; |
b7c06342 | 612 | struct the_nilfs *nilfs = root->nilfs; |
c306af23 | 613 | u64 id = huge_encode_dev(sb->s_bdev->bd_dev); |
783f6184 RK |
614 | unsigned long long blocks; |
615 | unsigned long overhead; | |
616 | unsigned long nrsvblocks; | |
617 | sector_t nfreeblocks; | |
c7ef972c | 618 | u64 nmaxinodes, nfreeinodes; |
783f6184 RK |
619 | int err; |
620 | ||
621 | /* | |
622 | * Compute all of the segment blocks | |
623 | * | |
624 | * The blocks before first segment and after last segment | |
625 | * are excluded. | |
626 | */ | |
627 | blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments | |
628 | - nilfs->ns_first_data_block; | |
629 | nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment; | |
630 | ||
631 | /* | |
632 | * Compute the overhead | |
633 | * | |
7a65004b | 634 | * When distributing meta data blocks outside segment structure, |
783f6184 RK |
635 | * We must count them as the overhead. |
636 | */ | |
637 | overhead = 0; | |
638 | ||
639 | err = nilfs_count_free_blocks(nilfs, &nfreeblocks); | |
640 | if (unlikely(err)) | |
641 | return err; | |
642 | ||
c7ef972c VD |
643 | err = nilfs_ifile_count_free_inodes(root->ifile, |
644 | &nmaxinodes, &nfreeinodes); | |
645 | if (unlikely(err)) { | |
a1d0747a | 646 | nilfs_warn(sb, "failed to count free inodes: err=%d", err); |
c7ef972c VD |
647 | if (err == -ERANGE) { |
648 | /* | |
649 | * If nilfs_palloc_count_max_entries() returns | |
650 | * -ERANGE error code then we simply treat | |
651 | * curent inodes count as maximum possible and | |
652 | * zero as free inodes value. | |
653 | */ | |
e5f7f848 | 654 | nmaxinodes = atomic64_read(&root->inodes_count); |
c7ef972c VD |
655 | nfreeinodes = 0; |
656 | err = 0; | |
657 | } else | |
658 | return err; | |
659 | } | |
660 | ||
783f6184 RK |
661 | buf->f_type = NILFS_SUPER_MAGIC; |
662 | buf->f_bsize = sb->s_blocksize; | |
663 | buf->f_blocks = blocks - overhead; | |
664 | buf->f_bfree = nfreeblocks; | |
665 | buf->f_bavail = (buf->f_bfree >= nrsvblocks) ? | |
666 | (buf->f_bfree - nrsvblocks) : 0; | |
c7ef972c VD |
667 | buf->f_files = nmaxinodes; |
668 | buf->f_ffree = nfreeinodes; | |
783f6184 | 669 | buf->f_namelen = NILFS_NAME_LEN; |
6d1349c7 | 670 | buf->f_fsid = u64_to_fsid(id); |
c306af23 | 671 | |
783f6184 RK |
672 | return 0; |
673 | } | |
674 | ||
34c80b1d | 675 | static int nilfs_show_options(struct seq_file *seq, struct dentry *dentry) |
b58a285b | 676 | { |
34c80b1d | 677 | struct super_block *sb = dentry->d_sb; |
e3154e97 | 678 | struct the_nilfs *nilfs = sb->s_fs_info; |
2b0143b5 | 679 | struct nilfs_root *root = NILFS_I(d_inode(dentry))->i_root; |
b58a285b | 680 | |
3b2ce58b | 681 | if (!nilfs_test_opt(nilfs, BARRIER)) |
c6b4d57d | 682 | seq_puts(seq, ",nobarrier"); |
f11459ad RK |
683 | if (root->cno != NILFS_CPTREE_CURRENT_CNO) |
684 | seq_printf(seq, ",cp=%llu", (unsigned long long)root->cno); | |
3b2ce58b | 685 | if (nilfs_test_opt(nilfs, ERRORS_PANIC)) |
c6b4d57d | 686 | seq_puts(seq, ",errors=panic"); |
3b2ce58b | 687 | if (nilfs_test_opt(nilfs, ERRORS_CONT)) |
c6b4d57d | 688 | seq_puts(seq, ",errors=continue"); |
3b2ce58b | 689 | if (nilfs_test_opt(nilfs, STRICT_ORDER)) |
c6b4d57d | 690 | seq_puts(seq, ",order=strict"); |
3b2ce58b | 691 | if (nilfs_test_opt(nilfs, NORECOVERY)) |
c6b4d57d | 692 | seq_puts(seq, ",norecovery"); |
3b2ce58b | 693 | if (nilfs_test_opt(nilfs, DISCARD)) |
c6b4d57d | 694 | seq_puts(seq, ",discard"); |
b58a285b JS |
695 | |
696 | return 0; | |
697 | } | |
698 | ||
b87221de | 699 | static const struct super_operations nilfs_sops = { |
783f6184 | 700 | .alloc_inode = nilfs_alloc_inode, |
977c3d18 | 701 | .free_inode = nilfs_free_inode, |
783f6184 | 702 | .dirty_inode = nilfs_dirty_inode, |
6fd1e5c9 | 703 | .evict_inode = nilfs_evict_inode, |
783f6184 | 704 | .put_super = nilfs_put_super, |
783f6184 | 705 | .sync_fs = nilfs_sync_fs, |
5beb6e0b RK |
706 | .freeze_fs = nilfs_freeze, |
707 | .unfreeze_fs = nilfs_unfreeze, | |
783f6184 | 708 | .statfs = nilfs_statfs, |
b58a285b | 709 | .show_options = nilfs_show_options |
783f6184 RK |
710 | }; |
711 | ||
783f6184 | 712 | enum { |
36defdd9 ES |
713 | Opt_err, Opt_barrier, Opt_snapshot, Opt_order, Opt_norecovery, |
714 | Opt_discard, | |
783f6184 RK |
715 | }; |
716 | ||
36defdd9 ES |
717 | static const struct constant_table nilfs_param_err[] = { |
718 | {"continue", NILFS_MOUNT_ERRORS_CONT}, | |
719 | {"panic", NILFS_MOUNT_ERRORS_PANIC}, | |
720 | {"remount-ro", NILFS_MOUNT_ERRORS_RO}, | |
721 | {} | |
783f6184 RK |
722 | }; |
723 | ||
36defdd9 ES |
724 | static const struct fs_parameter_spec nilfs_param_spec[] = { |
725 | fsparam_enum ("errors", Opt_err, nilfs_param_err), | |
726 | fsparam_flag_no ("barrier", Opt_barrier), | |
727 | fsparam_u64 ("cp", Opt_snapshot), | |
728 | fsparam_string ("order", Opt_order), | |
729 | fsparam_flag ("norecovery", Opt_norecovery), | |
730 | fsparam_flag_no ("discard", Opt_discard), | |
731 | {} | |
732 | }; | |
4ad364ca | 733 | |
36defdd9 ES |
734 | struct nilfs_fs_context { |
735 | unsigned long ns_mount_opt; | |
736 | __u64 cno; | |
737 | }; | |
783f6184 | 738 | |
36defdd9 ES |
739 | static int nilfs_parse_param(struct fs_context *fc, struct fs_parameter *param) |
740 | { | |
741 | struct nilfs_fs_context *nilfs = fc->fs_private; | |
742 | int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; | |
743 | struct fs_parse_result result; | |
744 | int opt; | |
745 | ||
746 | opt = fs_parse(fc, nilfs_param_spec, param, &result); | |
747 | if (opt < 0) | |
748 | return opt; | |
749 | ||
750 | switch (opt) { | |
751 | case Opt_barrier: | |
752 | if (result.negated) | |
3b2ce58b | 753 | nilfs_clear_opt(nilfs, BARRIER); |
36defdd9 ES |
754 | else |
755 | nilfs_set_opt(nilfs, BARRIER); | |
756 | break; | |
757 | case Opt_order: | |
758 | if (strcmp(param->string, "relaxed") == 0) | |
759 | /* Ordered data semantics */ | |
760 | nilfs_clear_opt(nilfs, STRICT_ORDER); | |
761 | else if (strcmp(param->string, "strict") == 0) | |
762 | /* Strict in-order semantics */ | |
763 | nilfs_set_opt(nilfs, STRICT_ORDER); | |
764 | else | |
765 | return -EINVAL; | |
766 | break; | |
767 | case Opt_err: | |
768 | nilfs->ns_mount_opt &= ~NILFS_MOUNT_ERROR_MODE; | |
769 | nilfs->ns_mount_opt |= result.uint_32; | |
770 | break; | |
771 | case Opt_snapshot: | |
772 | if (is_remount) { | |
773 | struct super_block *sb = fc->root->d_sb; | |
774 | ||
775 | nilfs_err(sb, | |
776 | "\"%s\" option is invalid for remount", | |
777 | param->key); | |
778 | return -EINVAL; | |
779 | } | |
780 | if (result.uint_64 == 0) { | |
781 | nilfs_err(NULL, | |
782 | "invalid option \"cp=0\": invalid checkpoint number 0"); | |
783 | return -EINVAL; | |
783f6184 | 784 | } |
36defdd9 ES |
785 | nilfs->cno = result.uint_64; |
786 | break; | |
787 | case Opt_norecovery: | |
788 | nilfs_set_opt(nilfs, NORECOVERY); | |
789 | break; | |
790 | case Opt_discard: | |
791 | if (result.negated) | |
792 | nilfs_clear_opt(nilfs, DISCARD); | |
793 | else | |
794 | nilfs_set_opt(nilfs, DISCARD); | |
795 | break; | |
796 | default: | |
797 | return -EINVAL; | |
783f6184 | 798 | } |
3b2ce58b | 799 | |
36defdd9 | 800 | return 0; |
783f6184 RK |
801 | } |
802 | ||
f7545144 | 803 | static int nilfs_setup_super(struct super_block *sb, int is_mount) |
783f6184 | 804 | { |
e3154e97 | 805 | struct the_nilfs *nilfs = sb->s_fs_info; |
d26493b6 JS |
806 | struct nilfs_super_block **sbp; |
807 | int max_mnt_count; | |
808 | int mnt_count; | |
809 | ||
810 | /* nilfs->ns_sem must be locked by the caller. */ | |
f7545144 | 811 | sbp = nilfs_prepare_super(sb, 0); |
d26493b6 JS |
812 | if (!sbp) |
813 | return -EIO; | |
814 | ||
5beb6e0b RK |
815 | if (!is_mount) |
816 | goto skip_mount_setup; | |
817 | ||
d26493b6 JS |
818 | max_mnt_count = le16_to_cpu(sbp[0]->s_max_mnt_count); |
819 | mnt_count = le16_to_cpu(sbp[0]->s_mnt_count); | |
783f6184 | 820 | |
f50a4c81 | 821 | if (nilfs->ns_mount_state & NILFS_ERROR_FS) { |
a1d0747a | 822 | nilfs_warn(sb, "mounting fs with errors"); |
783f6184 RK |
823 | #if 0 |
824 | } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) { | |
a1d0747a | 825 | nilfs_warn(sb, "maximal mount count reached"); |
783f6184 RK |
826 | #endif |
827 | } | |
828 | if (!max_mnt_count) | |
d26493b6 | 829 | sbp[0]->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT); |
783f6184 | 830 | |
d26493b6 | 831 | sbp[0]->s_mnt_count = cpu_to_le16(mnt_count + 1); |
21a1a52d | 832 | sbp[0]->s_mtime = cpu_to_le64(ktime_get_real_seconds()); |
5beb6e0b RK |
833 | |
834 | skip_mount_setup: | |
d26493b6 JS |
835 | sbp[0]->s_state = |
836 | cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & ~NILFS_VALID_FS); | |
b2ac86e1 | 837 | /* synchronize sbp[1] with sbp[0] */ |
0ca7a5b9 RK |
838 | if (sbp[1]) |
839 | memcpy(sbp[1], sbp[0], nilfs->ns_sbsize); | |
f7545144 | 840 | return nilfs_commit_super(sb, NILFS_SB_COMMIT_ALL); |
783f6184 RK |
841 | } |
842 | ||
e339ad31 RK |
843 | struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb, |
844 | u64 pos, int blocksize, | |
845 | struct buffer_head **pbh) | |
783f6184 | 846 | { |
e339ad31 RK |
847 | unsigned long long sb_index = pos; |
848 | unsigned long offset; | |
783f6184 | 849 | |
e339ad31 | 850 | offset = do_div(sb_index, blocksize); |
783f6184 | 851 | *pbh = sb_bread(sb, sb_index); |
e339ad31 | 852 | if (!*pbh) |
783f6184 | 853 | return NULL; |
783f6184 RK |
854 | return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset); |
855 | } | |
856 | ||
36defdd9 ES |
857 | int nilfs_store_magic(struct super_block *sb, |
858 | struct nilfs_super_block *sbp) | |
783f6184 | 859 | { |
e3154e97 | 860 | struct the_nilfs *nilfs = sb->s_fs_info; |
783f6184 | 861 | |
783f6184 RK |
862 | sb->s_magic = le16_to_cpu(sbp->s_magic); |
863 | ||
864 | /* FS independent flags */ | |
865 | #ifdef NILFS_ATIME_DISABLE | |
1751e8a6 | 866 | sb->s_flags |= SB_NOATIME; |
783f6184 RK |
867 | #endif |
868 | ||
574e6c31 RK |
869 | nilfs->ns_resuid = le16_to_cpu(sbp->s_def_resuid); |
870 | nilfs->ns_resgid = le16_to_cpu(sbp->s_def_resgid); | |
871 | nilfs->ns_interval = le32_to_cpu(sbp->s_c_interval); | |
872 | nilfs->ns_watermark = le32_to_cpu(sbp->s_c_block_max); | |
783f6184 | 873 | |
36defdd9 | 874 | return 0; |
783f6184 RK |
875 | } |
876 | ||
c5ca48aa RK |
877 | int nilfs_check_feature_compatibility(struct super_block *sb, |
878 | struct nilfs_super_block *sbp) | |
879 | { | |
880 | __u64 features; | |
881 | ||
882 | features = le64_to_cpu(sbp->s_feature_incompat) & | |
883 | ~NILFS_FEATURE_INCOMPAT_SUPP; | |
884 | if (features) { | |
a1d0747a | 885 | nilfs_err(sb, |
feee880f RK |
886 | "couldn't mount because of unsupported optional features (%llx)", |
887 | (unsigned long long)features); | |
c5ca48aa RK |
888 | return -EINVAL; |
889 | } | |
890 | features = le64_to_cpu(sbp->s_feature_compat_ro) & | |
891 | ~NILFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 892 | if (!sb_rdonly(sb) && features) { |
a1d0747a | 893 | nilfs_err(sb, |
feee880f RK |
894 | "couldn't mount RDWR because of unsupported optional features (%llx)", |
895 | (unsigned long long)features); | |
c5ca48aa RK |
896 | return -EINVAL; |
897 | } | |
898 | return 0; | |
899 | } | |
900 | ||
367ea334 RK |
901 | static int nilfs_get_root_dentry(struct super_block *sb, |
902 | struct nilfs_root *root, | |
903 | struct dentry **root_dentry) | |
904 | { | |
905 | struct inode *inode; | |
906 | struct dentry *dentry; | |
907 | int ret = 0; | |
908 | ||
909 | inode = nilfs_iget(sb, root, NILFS_ROOT_INO); | |
910 | if (IS_ERR(inode)) { | |
367ea334 | 911 | ret = PTR_ERR(inode); |
a1d0747a | 912 | nilfs_err(sb, "error %d getting root inode", ret); |
367ea334 RK |
913 | goto out; |
914 | } | |
915 | if (!S_ISDIR(inode->i_mode) || !inode->i_blocks || !inode->i_size) { | |
916 | iput(inode); | |
a1d0747a | 917 | nilfs_err(sb, "corrupt root inode"); |
367ea334 RK |
918 | ret = -EINVAL; |
919 | goto out; | |
920 | } | |
921 | ||
f11459ad RK |
922 | if (root->cno == NILFS_CPTREE_CURRENT_CNO) { |
923 | dentry = d_find_alias(inode); | |
924 | if (!dentry) { | |
48fde701 | 925 | dentry = d_make_root(inode); |
f11459ad | 926 | if (!dentry) { |
f11459ad RK |
927 | ret = -ENOMEM; |
928 | goto failed_dentry; | |
929 | } | |
930 | } else { | |
931 | iput(inode); | |
932 | } | |
933 | } else { | |
1a0a397e | 934 | dentry = d_obtain_root(inode); |
f11459ad RK |
935 | if (IS_ERR(dentry)) { |
936 | ret = PTR_ERR(dentry); | |
937 | goto failed_dentry; | |
938 | } | |
367ea334 RK |
939 | } |
940 | *root_dentry = dentry; | |
941 | out: | |
942 | return ret; | |
f11459ad RK |
943 | |
944 | failed_dentry: | |
a1d0747a | 945 | nilfs_err(sb, "error %d getting root dentry", ret); |
f11459ad | 946 | goto out; |
367ea334 RK |
947 | } |
948 | ||
ab4d8f7e RK |
949 | static int nilfs_attach_snapshot(struct super_block *s, __u64 cno, |
950 | struct dentry **root_dentry) | |
951 | { | |
e3154e97 | 952 | struct the_nilfs *nilfs = s->s_fs_info; |
ab4d8f7e RK |
953 | struct nilfs_root *root; |
954 | int ret; | |
955 | ||
572d8b39 RK |
956 | mutex_lock(&nilfs->ns_snapshot_mount_mutex); |
957 | ||
ab4d8f7e RK |
958 | down_read(&nilfs->ns_segctor_sem); |
959 | ret = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile, cno); | |
960 | up_read(&nilfs->ns_segctor_sem); | |
961 | if (ret < 0) { | |
962 | ret = (ret == -ENOENT) ? -EINVAL : ret; | |
963 | goto out; | |
964 | } else if (!ret) { | |
a1d0747a | 965 | nilfs_err(s, |
feee880f RK |
966 | "The specified checkpoint is not a snapshot (checkpoint number=%llu)", |
967 | (unsigned long long)cno); | |
ab4d8f7e RK |
968 | ret = -EINVAL; |
969 | goto out; | |
970 | } | |
971 | ||
f7545144 | 972 | ret = nilfs_attach_checkpoint(s, cno, false, &root); |
ab4d8f7e | 973 | if (ret) { |
a1d0747a | 974 | nilfs_err(s, |
feee880f RK |
975 | "error %d while loading snapshot (checkpoint number=%llu)", |
976 | ret, (unsigned long long)cno); | |
ab4d8f7e RK |
977 | goto out; |
978 | } | |
979 | ret = nilfs_get_root_dentry(s, root, root_dentry); | |
980 | nilfs_put_root(root); | |
981 | out: | |
572d8b39 | 982 | mutex_unlock(&nilfs->ns_snapshot_mount_mutex); |
ab4d8f7e RK |
983 | return ret; |
984 | } | |
985 | ||
f11459ad | 986 | /** |
e95c311e | 987 | * nilfs_tree_is_busy() - try to shrink dentries of a checkpoint |
f11459ad RK |
988 | * @root_dentry: root dentry of the tree to be shrunk |
989 | * | |
990 | * This function returns true if the tree was in-use. | |
991 | */ | |
e95c311e | 992 | static bool nilfs_tree_is_busy(struct dentry *root_dentry) |
f11459ad | 993 | { |
f11459ad | 994 | shrink_dcache_parent(root_dentry); |
e95c311e | 995 | return d_count(root_dentry) > 1; |
f11459ad RK |
996 | } |
997 | ||
032dbb3b RK |
998 | int nilfs_checkpoint_is_mounted(struct super_block *sb, __u64 cno) |
999 | { | |
e3154e97 | 1000 | struct the_nilfs *nilfs = sb->s_fs_info; |
032dbb3b RK |
1001 | struct nilfs_root *root; |
1002 | struct inode *inode; | |
1003 | struct dentry *dentry; | |
1004 | int ret; | |
1005 | ||
3377f843 | 1006 | if (cno > nilfs->ns_cno) |
032dbb3b RK |
1007 | return false; |
1008 | ||
1009 | if (cno >= nilfs_last_cno(nilfs)) | |
1010 | return true; /* protect recent checkpoints */ | |
1011 | ||
1012 | ret = false; | |
e3154e97 | 1013 | root = nilfs_lookup_root(nilfs, cno); |
032dbb3b RK |
1014 | if (root) { |
1015 | inode = nilfs_ilookup(sb, root, NILFS_ROOT_INO); | |
1016 | if (inode) { | |
1017 | dentry = d_find_alias(inode); | |
1018 | if (dentry) { | |
e95c311e | 1019 | ret = nilfs_tree_is_busy(dentry); |
032dbb3b RK |
1020 | dput(dentry); |
1021 | } | |
1022 | iput(inode); | |
1023 | } | |
1024 | nilfs_put_root(root); | |
1025 | } | |
1026 | return ret; | |
1027 | } | |
1028 | ||
783f6184 RK |
1029 | /** |
1030 | * nilfs_fill_super() - initialize a super block instance | |
1031 | * @sb: super_block | |
36defdd9 | 1032 | * @fc: filesystem context |
783f6184 | 1033 | * |
aa7dfb89 | 1034 | * This function is called exclusively by nilfs->ns_mount_mutex. |
783f6184 RK |
1035 | * So, the recovery process is protected from other simultaneous mounts. |
1036 | */ | |
1037 | static int | |
36defdd9 | 1038 | nilfs_fill_super(struct super_block *sb, struct fs_context *fc) |
783f6184 | 1039 | { |
348fe8da | 1040 | struct the_nilfs *nilfs; |
4d8d9293 | 1041 | struct nilfs_root *fsroot; |
36defdd9 | 1042 | struct nilfs_fs_context *ctx = fc->fs_private; |
783f6184 RK |
1043 | __u64 cno; |
1044 | int err; | |
1045 | ||
6625689e | 1046 | nilfs = alloc_nilfs(sb); |
e3154e97 | 1047 | if (!nilfs) |
783f6184 RK |
1048 | return -ENOMEM; |
1049 | ||
e3154e97 | 1050 | sb->s_fs_info = nilfs; |
783f6184 | 1051 | |
36defdd9 | 1052 | err = init_nilfs(nilfs, sb); |
783f6184 | 1053 | if (err) |
348fe8da | 1054 | goto failed_nilfs; |
783f6184 | 1055 | |
36defdd9 ES |
1056 | /* Copy in parsed mount options */ |
1057 | nilfs->ns_mount_opt = ctx->ns_mount_opt; | |
1058 | ||
783f6184 RK |
1059 | sb->s_op = &nilfs_sops; |
1060 | sb->s_export_op = &nilfs_export_ops; | |
1061 | sb->s_root = NULL; | |
61239230 | 1062 | sb->s_time_gran = 1; |
8de52778 | 1063 | sb->s_max_links = NILFS_LINK_MAX; |
026a7d63 | 1064 | |
a11d7fc2 | 1065 | sb->s_bdi = bdi_get(sb->s_bdev->bd_disk->bdi); |
783f6184 | 1066 | |
f7545144 | 1067 | err = load_nilfs(nilfs, sb); |
f50a4c81 | 1068 | if (err) |
348fe8da | 1069 | goto failed_nilfs; |
f50a4c81 | 1070 | |
ef851d44 RK |
1071 | super_set_uuid(sb, nilfs->ns_sbp[0]->s_uuid, |
1072 | sizeof(nilfs->ns_sbp[0]->s_uuid)); | |
8d1dba2e | 1073 | super_set_sysfs_name_bdev(sb); |
ef851d44 | 1074 | |
783f6184 | 1075 | cno = nilfs_last_cno(nilfs); |
f7545144 | 1076 | err = nilfs_attach_checkpoint(sb, cno, true, &fsroot); |
783f6184 | 1077 | if (err) { |
a1d0747a | 1078 | nilfs_err(sb, |
feee880f RK |
1079 | "error %d while loading last checkpoint (checkpoint number=%llu)", |
1080 | err, (unsigned long long)cno); | |
f1e89c86 | 1081 | goto failed_unload; |
783f6184 RK |
1082 | } |
1083 | ||
bc98a42c | 1084 | if (!sb_rdonly(sb)) { |
f7545144 | 1085 | err = nilfs_attach_log_writer(sb, fsroot); |
783f6184 RK |
1086 | if (err) |
1087 | goto failed_checkpoint; | |
1088 | } | |
1089 | ||
367ea334 RK |
1090 | err = nilfs_get_root_dentry(sb, fsroot, &sb->s_root); |
1091 | if (err) | |
783f6184 | 1092 | goto failed_segctor; |
783f6184 | 1093 | |
4d8d9293 | 1094 | nilfs_put_root(fsroot); |
783f6184 | 1095 | |
bc98a42c | 1096 | if (!sb_rdonly(sb)) { |
783f6184 | 1097 | down_write(&nilfs->ns_sem); |
f7545144 | 1098 | nilfs_setup_super(sb, true); |
783f6184 RK |
1099 | up_write(&nilfs->ns_sem); |
1100 | } | |
1101 | ||
783f6184 RK |
1102 | return 0; |
1103 | ||
783f6184 | 1104 | failed_segctor: |
f7545144 | 1105 | nilfs_detach_log_writer(sb); |
783f6184 RK |
1106 | |
1107 | failed_checkpoint: | |
4d8d9293 | 1108 | nilfs_put_root(fsroot); |
783f6184 | 1109 | |
f1e89c86 | 1110 | failed_unload: |
42560f9c | 1111 | nilfs_sysfs_delete_device_group(nilfs); |
f1e89c86 RK |
1112 | iput(nilfs->ns_sufile); |
1113 | iput(nilfs->ns_cpfile); | |
1114 | iput(nilfs->ns_dat); | |
1115 | ||
348fe8da RK |
1116 | failed_nilfs: |
1117 | destroy_nilfs(nilfs); | |
783f6184 RK |
1118 | return err; |
1119 | } | |
1120 | ||
36defdd9 | 1121 | static int nilfs_reconfigure(struct fs_context *fc) |
783f6184 | 1122 | { |
36defdd9 ES |
1123 | struct nilfs_fs_context *ctx = fc->fs_private; |
1124 | struct super_block *sb = fc->root->d_sb; | |
e3154e97 | 1125 | struct the_nilfs *nilfs = sb->s_fs_info; |
f11459ad | 1126 | int err; |
783f6184 | 1127 | |
02b9984d | 1128 | sync_filesystem(sb); |
783f6184 | 1129 | |
d240e067 | 1130 | err = -EINVAL; |
783f6184 | 1131 | |
0234576d | 1132 | if (!nilfs_valid_fs(nilfs)) { |
a1d0747a JP |
1133 | nilfs_warn(sb, |
1134 | "couldn't remount because the filesystem is in an incomplete recovery state"); | |
36defdd9 | 1135 | goto ignore_opts; |
0234576d | 1136 | } |
36defdd9 | 1137 | if ((bool)(fc->sb_flags & SB_RDONLY) == sb_rdonly(sb)) |
783f6184 | 1138 | goto out; |
36defdd9 | 1139 | if (fc->sb_flags & SB_RDONLY) { |
1751e8a6 | 1140 | sb->s_flags |= SB_RDONLY; |
783f6184 | 1141 | |
783f6184 RK |
1142 | /* |
1143 | * Remounting a valid RW partition RDONLY, so set | |
1144 | * the RDONLY flag and then mark the partition as valid again. | |
1145 | */ | |
1146 | down_write(&nilfs->ns_sem); | |
f7545144 | 1147 | nilfs_cleanup_super(sb); |
783f6184 RK |
1148 | up_write(&nilfs->ns_sem); |
1149 | } else { | |
c5ca48aa | 1150 | __u64 features; |
e912a5b6 | 1151 | struct nilfs_root *root; |
c5ca48aa | 1152 | |
783f6184 RK |
1153 | /* |
1154 | * Mounting a RDONLY partition read-write, so reread and | |
1155 | * store the current valid flag. (It may have been changed | |
1156 | * by fsck since we originally mounted the partition.) | |
1157 | */ | |
c5ca48aa RK |
1158 | down_read(&nilfs->ns_sem); |
1159 | features = le64_to_cpu(nilfs->ns_sbp[0]->s_feature_compat_ro) & | |
1160 | ~NILFS_FEATURE_COMPAT_RO_SUPP; | |
1161 | up_read(&nilfs->ns_sem); | |
1162 | if (features) { | |
a1d0747a JP |
1163 | nilfs_warn(sb, |
1164 | "couldn't remount RDWR because of unsupported optional features (%llx)", | |
1165 | (unsigned long long)features); | |
c5ca48aa | 1166 | err = -EROFS; |
36defdd9 | 1167 | goto ignore_opts; |
c5ca48aa RK |
1168 | } |
1169 | ||
1751e8a6 | 1170 | sb->s_flags &= ~SB_RDONLY; |
783f6184 | 1171 | |
2b0143b5 | 1172 | root = NILFS_I(d_inode(sb->s_root))->i_root; |
f7545144 | 1173 | err = nilfs_attach_log_writer(sb, root); |
36defdd9 ES |
1174 | if (err) { |
1175 | sb->s_flags |= SB_RDONLY; | |
1176 | goto ignore_opts; | |
1177 | } | |
783f6184 RK |
1178 | |
1179 | down_write(&nilfs->ns_sem); | |
f7545144 | 1180 | nilfs_setup_super(sb, true); |
783f6184 | 1181 | up_write(&nilfs->ns_sem); |
783f6184 RK |
1182 | } |
1183 | out: | |
36defdd9 ES |
1184 | sb->s_flags = (sb->s_flags & ~SB_POSIXACL); |
1185 | /* Copy over parsed remount options */ | |
1186 | nilfs->ns_mount_opt = ctx->ns_mount_opt; | |
783f6184 | 1187 | |
783f6184 | 1188 | return 0; |
783f6184 | 1189 | |
36defdd9 ES |
1190 | ignore_opts: |
1191 | return err; | |
783f6184 RK |
1192 | } |
1193 | ||
36defdd9 ES |
1194 | static int |
1195 | nilfs_get_tree(struct fs_context *fc) | |
783f6184 | 1196 | { |
36defdd9 | 1197 | struct nilfs_fs_context *ctx = fc->fs_private; |
33c8e57c | 1198 | struct super_block *s; |
c1e012ea CH |
1199 | dev_t dev; |
1200 | int err; | |
783f6184 | 1201 | |
36defdd9 ES |
1202 | if (ctx->cno && !(fc->sb_flags & SB_RDONLY)) { |
1203 | nilfs_err(NULL, | |
1204 | "invalid option \"cp=%llu\": read-only option is not specified", | |
1205 | ctx->cno); | |
1206 | return -EINVAL; | |
1207 | } | |
783f6184 | 1208 | |
36defdd9 | 1209 | err = lookup_bdev(fc->source, &dev); |
c1e012ea | 1210 | if (err) |
36defdd9 | 1211 | return err; |
783f6184 | 1212 | |
36defdd9 | 1213 | s = sget_dev(fc, dev); |
c1e012ea | 1214 | if (IS_ERR(s)) |
36defdd9 | 1215 | return PTR_ERR(s); |
783f6184 RK |
1216 | |
1217 | if (!s->s_root) { | |
36defdd9 | 1218 | err = setup_bdev_super(s, fc->sb_flags, fc); |
c1e012ea | 1219 | if (!err) |
36defdd9 | 1220 | err = nilfs_fill_super(s, fc); |
783f6184 | 1221 | if (err) |
348fe8da | 1222 | goto failed_super; |
783f6184 | 1223 | |
1751e8a6 | 1224 | s->s_flags |= SB_ACTIVE; |
36defdd9 | 1225 | } else if (!ctx->cno) { |
e95c311e | 1226 | if (nilfs_tree_is_busy(s->s_root)) { |
36defdd9 | 1227 | if ((fc->sb_flags ^ s->s_flags) & SB_RDONLY) { |
a1d0747a | 1228 | nilfs_err(s, |
feee880f | 1229 | "the device already has a %s mount.", |
bc98a42c | 1230 | sb_rdonly(s) ? "read-only" : "read/write"); |
f11459ad RK |
1231 | err = -EBUSY; |
1232 | goto failed_super; | |
1233 | } | |
e95c311e | 1234 | } else { |
f11459ad | 1235 | /* |
36defdd9 | 1236 | * Try reconfigure to setup mount states if the current |
f11459ad | 1237 | * tree is not mounted and only snapshots use this sb. |
36defdd9 ES |
1238 | * |
1239 | * Since nilfs_reconfigure() requires fc->root to be | |
1240 | * set, set it first and release it on failure. | |
f11459ad | 1241 | */ |
36defdd9 ES |
1242 | fc->root = dget(s->s_root); |
1243 | err = nilfs_reconfigure(fc); | |
1244 | if (err) { | |
1245 | dput(fc->root); | |
1246 | fc->root = NULL; /* prevent double release */ | |
f11459ad | 1247 | goto failed_super; |
36defdd9 ES |
1248 | } |
1249 | return 0; | |
f11459ad | 1250 | } |
783f6184 RK |
1251 | } |
1252 | ||
36defdd9 | 1253 | if (ctx->cno) { |
c1e012ea CH |
1254 | struct dentry *root_dentry; |
1255 | ||
36defdd9 | 1256 | err = nilfs_attach_snapshot(s, ctx->cno, &root_dentry); |
348fe8da | 1257 | if (err) |
f11459ad | 1258 | goto failed_super; |
36defdd9 ES |
1259 | fc->root = root_dentry; |
1260 | return 0; | |
783f6184 RK |
1261 | } |
1262 | ||
36defdd9 ES |
1263 | fc->root = dget(s->s_root); |
1264 | return 0; | |
783f6184 | 1265 | |
f11459ad | 1266 | failed_super: |
a95161aa | 1267 | deactivate_locked_super(s); |
36defdd9 ES |
1268 | return err; |
1269 | } | |
1270 | ||
1271 | static void nilfs_free_fc(struct fs_context *fc) | |
1272 | { | |
1273 | kfree(fc->fs_private); | |
1274 | } | |
1275 | ||
1276 | static const struct fs_context_operations nilfs_context_ops = { | |
1277 | .parse_param = nilfs_parse_param, | |
1278 | .get_tree = nilfs_get_tree, | |
1279 | .reconfigure = nilfs_reconfigure, | |
1280 | .free = nilfs_free_fc, | |
1281 | }; | |
1282 | ||
1283 | static int nilfs_init_fs_context(struct fs_context *fc) | |
1284 | { | |
1285 | struct nilfs_fs_context *ctx; | |
1286 | ||
1287 | ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); | |
1288 | if (!ctx) | |
1289 | return -ENOMEM; | |
1290 | ||
1291 | ctx->ns_mount_opt = NILFS_MOUNT_ERRORS_RO | NILFS_MOUNT_BARRIER; | |
1292 | fc->fs_private = ctx; | |
1293 | fc->ops = &nilfs_context_ops; | |
1294 | ||
1295 | return 0; | |
783f6184 RK |
1296 | } |
1297 | ||
783f6184 RK |
1298 | struct file_system_type nilfs_fs_type = { |
1299 | .owner = THIS_MODULE, | |
1300 | .name = "nilfs2", | |
783f6184 RK |
1301 | .kill_sb = kill_block_super, |
1302 | .fs_flags = FS_REQUIRES_DEV, | |
36defdd9 ES |
1303 | .init_fs_context = nilfs_init_fs_context, |
1304 | .parameters = nilfs_param_spec, | |
783f6184 | 1305 | }; |
7f78e035 | 1306 | MODULE_ALIAS_FS("nilfs2"); |
783f6184 | 1307 | |
41c88bd7 | 1308 | static void nilfs_inode_init_once(void *obj) |
783f6184 | 1309 | { |
41c88bd7 | 1310 | struct nilfs_inode_info *ii = obj; |
783f6184 | 1311 | |
41c88bd7 LH |
1312 | INIT_LIST_HEAD(&ii->i_dirty); |
1313 | #ifdef CONFIG_NILFS_XATTR | |
1314 | init_rwsem(&ii->xattr_sem); | |
1315 | #endif | |
41c88bd7 LH |
1316 | inode_init_once(&ii->vfs_inode); |
1317 | } | |
783f6184 | 1318 | |
41c88bd7 LH |
1319 | static void nilfs_segbuf_init_once(void *obj) |
1320 | { | |
1321 | memset(obj, 0, sizeof(struct nilfs_segment_buffer)); | |
1322 | } | |
783f6184 | 1323 | |
41c88bd7 LH |
1324 | static void nilfs_destroy_cachep(void) |
1325 | { | |
8c0a8537 KS |
1326 | /* |
1327 | * Make sure all delayed rcu free inodes are flushed before we | |
1328 | * destroy cache. | |
1329 | */ | |
1330 | rcu_barrier(); | |
1331 | ||
da80a39f JL |
1332 | kmem_cache_destroy(nilfs_inode_cachep); |
1333 | kmem_cache_destroy(nilfs_transaction_cachep); | |
1334 | kmem_cache_destroy(nilfs_segbuf_cachep); | |
1335 | kmem_cache_destroy(nilfs_btree_path_cache); | |
41c88bd7 | 1336 | } |
783f6184 | 1337 | |
41c88bd7 LH |
1338 | static int __init nilfs_init_cachep(void) |
1339 | { | |
1340 | nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache", | |
1341 | sizeof(struct nilfs_inode_info), 0, | |
5d097056 VD |
1342 | SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, |
1343 | nilfs_inode_init_once); | |
41c88bd7 LH |
1344 | if (!nilfs_inode_cachep) |
1345 | goto fail; | |
1346 | ||
1347 | nilfs_transaction_cachep = kmem_cache_create("nilfs2_transaction_cache", | |
1348 | sizeof(struct nilfs_transaction_info), 0, | |
1349 | SLAB_RECLAIM_ACCOUNT, NULL); | |
1350 | if (!nilfs_transaction_cachep) | |
1351 | goto fail; | |
1352 | ||
1353 | nilfs_segbuf_cachep = kmem_cache_create("nilfs2_segbuf_cache", | |
1354 | sizeof(struct nilfs_segment_buffer), 0, | |
1355 | SLAB_RECLAIM_ACCOUNT, nilfs_segbuf_init_once); | |
1356 | if (!nilfs_segbuf_cachep) | |
1357 | goto fail; | |
1358 | ||
1359 | nilfs_btree_path_cache = kmem_cache_create("nilfs2_btree_path_cache", | |
1360 | sizeof(struct nilfs_btree_path) * NILFS_BTREE_LEVEL_MAX, | |
1361 | 0, 0, NULL); | |
1362 | if (!nilfs_btree_path_cache) | |
1363 | goto fail; | |
783f6184 RK |
1364 | |
1365 | return 0; | |
1366 | ||
41c88bd7 LH |
1367 | fail: |
1368 | nilfs_destroy_cachep(); | |
1369 | return -ENOMEM; | |
1370 | } | |
1371 | ||
1372 | static int __init init_nilfs_fs(void) | |
1373 | { | |
1374 | int err; | |
783f6184 | 1375 | |
41c88bd7 LH |
1376 | err = nilfs_init_cachep(); |
1377 | if (err) | |
1378 | goto fail; | |
783f6184 | 1379 | |
dd70edbd | 1380 | err = nilfs_sysfs_init(); |
41c88bd7 LH |
1381 | if (err) |
1382 | goto free_cachep; | |
783f6184 | 1383 | |
dd70edbd VD |
1384 | err = register_filesystem(&nilfs_fs_type); |
1385 | if (err) | |
1386 | goto deinit_sysfs_entry; | |
1387 | ||
9f130263 | 1388 | printk(KERN_INFO "NILFS version 2 loaded\n"); |
41c88bd7 | 1389 | return 0; |
783f6184 | 1390 | |
dd70edbd VD |
1391 | deinit_sysfs_entry: |
1392 | nilfs_sysfs_exit(); | |
41c88bd7 LH |
1393 | free_cachep: |
1394 | nilfs_destroy_cachep(); | |
1395 | fail: | |
783f6184 RK |
1396 | return err; |
1397 | } | |
1398 | ||
1399 | static void __exit exit_nilfs_fs(void) | |
1400 | { | |
41c88bd7 | 1401 | nilfs_destroy_cachep(); |
dd70edbd | 1402 | nilfs_sysfs_exit(); |
783f6184 RK |
1403 | unregister_filesystem(&nilfs_fs_type); |
1404 | } | |
1405 | ||
1406 | module_init(init_nilfs_fs) | |
1407 | module_exit(exit_nilfs_fs) |