projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nilfs2: get rid of sget use for checking if current mount is present
[deliverable/linux.git] / fs / nilfs2 / super.c
CommitLineData
783f6184
RK		 1/*
2 * super.c - NILFS module and super block management.
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 * linux/fs/ext2/super.c
24 *
25 * Copyright (C) 1992, 1993, 1994, 1995
26 * Remy Card (card@masi.ibp.fr)
27 * Laboratoire MASI - Institut Blaise Pascal
28 * Universite Pierre et Marie Curie (Paris VI)
29 *
30 * from
31 *
32 * linux/fs/minix/inode.c
33 *
34 * Copyright (C) 1991, 1992 Linus Torvalds
35 *
36 * Big-endian to little-endian byte-swapping/bitmaps by
37 * David S. Miller (davem@caip.rutgers.edu), 1995
38 */
39
40#include <linux/module.h>
41#include <linux/string.h>
42#include <linux/slab.h>
43#include <linux/init.h>
44#include <linux/blkdev.h>
45#include <linux/parser.h>
46#include <linux/random.h>
47#include <linux/crc32.h>
48#include <linux/smp_lock.h>
49#include <linux/vfs.h>
50#include <linux/writeback.h>
51#include <linux/kobject.h>
52#include <linux/exportfs.h>
53#include "nilfs.h"
54#include "mdt.h"
55#include "alloc.h"
56#include "page.h"
57#include "cpfile.h"
58#include "ifile.h"
59#include "dat.h"
60#include "segment.h"
61#include "segbuf.h"
62
63MODULE_AUTHOR("NTT Corp.");
64MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem "
65 "(NILFS)");
783f6184
RK		 66MODULE_LICENSE("GPL");
67
8c85e125 68static void nilfs_write_super(struct super_block *sb);
783f6184 69static int nilfs_remount(struct super_block *sb, int *flags, char *data);
783f6184
RK		 70
71/**
72 * nilfs_error() - report failure condition on a filesystem
73 *
74 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
75 * reporting an error message. It should be called when NILFS detects
76 * incoherences or defects of meta data on disk. As for sustainable
77 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
78 * function should be used instead.
79 *
80 * The segment constructor must not call this function because it can
81 * kill itself.
82 */
83void nilfs_error(struct super_block *sb, const char *function,
84 const char *fmt, ...)
85{
86 struct nilfs_sb_info *sbi = NILFS_SB(sb);
87 va_list args;
88
89 va_start(args, fmt);
90 printk(KERN_CRIT "NILFS error (device %s): %s: ", sb->s_id, function);
91 vprintk(fmt, args);
92 printk("\n");
93 va_end(args);
94
95 if (!(sb->s_flags & MS_RDONLY)) {
96 struct the_nilfs *nilfs = sbi->s_nilfs;
97
98 if (!nilfs_test_opt(sbi, ERRORS_CONT))
99 nilfs_detach_segment_constructor(sbi);
100
101 down_write(&nilfs->ns_sem);
102 if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) {
103 nilfs->ns_mount_state |= NILFS_ERROR_FS;
e339ad31
RK		 104 nilfs->ns_sbp[0]->s_state |=
105 cpu_to_le16(NILFS_ERROR_FS);
106 nilfs_commit_super(sbi, 1);
783f6184
RK		 107 }
108 up_write(&nilfs->ns_sem);
109
110 if (nilfs_test_opt(sbi, ERRORS_RO)) {
111 printk(KERN_CRIT "Remounting filesystem read-only\n");
112 sb->s_flags |= MS_RDONLY;
113 }
114 }
115
116 if (nilfs_test_opt(sbi, ERRORS_PANIC))
117 panic("NILFS (device %s): panic forced after error\n",
118 sb->s_id);
119}
120
121void nilfs_warning(struct super_block *sb, const char *function,
122 const char *fmt, ...)
123{
124 va_list args;
125
126 va_start(args, fmt);
127 printk(KERN_WARNING "NILFS warning (device %s): %s: ",
128 sb->s_id, function);
129 vprintk(fmt, args);
130 printk("\n");
131 va_end(args);
132}
133
134static struct kmem_cache *nilfs_inode_cachep;
135
136struct inode *nilfs_alloc_inode(struct super_block *sb)
137{
138 struct nilfs_inode_info *ii;
139
140 ii = kmem_cache_alloc(nilfs_inode_cachep, GFP_NOFS);
141 if (!ii)
142 return NULL;
143 ii->i_bh = NULL;
144 ii->i_state = 0;
145 ii->vfs_inode.i_version = 1;
146 nilfs_btnode_cache_init(&ii->i_btnode_cache);
147 return &ii->vfs_inode;
148}
149
150void nilfs_destroy_inode(struct inode *inode)
151{
152 kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
153}
154
155static void init_once(void *obj)
156{
157 struct nilfs_inode_info *ii = obj;
158
159 INIT_LIST_HEAD(&ii->i_dirty);
160#ifdef CONFIG_NILFS_XATTR
161 init_rwsem(&ii->xattr_sem);
162#endif
163 nilfs_btnode_cache_init_once(&ii->i_btnode_cache);
164 ii->i_bmap = (struct nilfs_bmap *)&ii->i_bmap_union;
165 inode_init_once(&ii->vfs_inode);
166}
167
168static int nilfs_init_inode_cache(void)
169{
170 nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache",
171 sizeof(struct nilfs_inode_info),
172 0, SLAB_RECLAIM_ACCOUNT,
173 init_once);
174
175 return (nilfs_inode_cachep == NULL) ? -ENOMEM : 0;
176}
177
178static inline void nilfs_destroy_inode_cache(void)
179{
180 kmem_cache_destroy(nilfs_inode_cachep);
181}
182
183static void nilfs_clear_inode(struct inode *inode)
184{
185 struct nilfs_inode_info *ii = NILFS_I(inode);
783f6184
RK		 186
187#ifdef CONFIG_NILFS_POSIX_ACL
188 if (ii->i_acl && ii->i_acl != NILFS_ACL_NOT_CACHED) {
189 posix_acl_release(ii->i_acl);
190 ii->i_acl = NILFS_ACL_NOT_CACHED;
191 }
192 if (ii->i_default_acl && ii->i_default_acl != NILFS_ACL_NOT_CACHED) {
193 posix_acl_release(ii->i_default_acl);
194 ii->i_default_acl = NILFS_ACL_NOT_CACHED;
195 }
196#endif
197 /*
198 * Free resources allocated in nilfs_read_inode(), here.
199 */
a2e7d2df 200 BUG_ON(!list_empty(&ii->i_dirty));
783f6184
RK		 201 brelse(ii->i_bh);
202 ii->i_bh = NULL;
783f6184
RK		 203
204 if (test_bit(NILFS_I_BMAP, &ii->i_state))
205 nilfs_bmap_clear(ii->i_bmap);
206
207 nilfs_btnode_cache_clear(&ii->i_btnode_cache);
783f6184
RK		 208}
209
e339ad31 210static int nilfs_sync_super(struct nilfs_sb_info *sbi, int dupsb)
783f6184
RK		 211{
212 struct the_nilfs *nilfs = sbi->s_nilfs;
213 int err;
214 int barrier_done = 0;
215
216 if (nilfs_test_opt(sbi, BARRIER)) {
e339ad31 217 set_buffer_ordered(nilfs->ns_sbh[0]);
783f6184
RK		 218 barrier_done = 1;
219 }
220 retry:
e339ad31
RK		 221 set_buffer_dirty(nilfs->ns_sbh[0]);
222 err = sync_dirty_buffer(nilfs->ns_sbh[0]);
783f6184
RK		 223 if (err == -EOPNOTSUPP && barrier_done) {
224 nilfs_warning(sbi->s_super, __func__,
225 "barrier-based sync failed. "
226 "disabling barriers\n");
227 nilfs_clear_opt(sbi, BARRIER);
228 barrier_done = 0;
e339ad31 229 clear_buffer_ordered(nilfs->ns_sbh[0]);
783f6184
RK		 230 goto retry;
231 }
e339ad31 232 if (unlikely(err)) {
783f6184
RK		 233 printk(KERN_ERR
234 "NILFS: unable to write superblock (err=%d)\n", err);
e339ad31
RK		 235 if (err == -EIO && nilfs->ns_sbh[1]) {
236 nilfs_fall_back_super_block(nilfs);
237 goto retry;
238 }
239 } else {
240 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
241
242 /*
243 * The latest segment becomes trailable from the position
244 * written in superblock.
245 */
783f6184 246 clear_nilfs_discontinued(nilfs);
e339ad31
RK		 247
248 /* update GC protection for recent segments */
249 if (nilfs->ns_sbh[1]) {
250 sbp = NULL;
251 if (dupsb) {
252 set_buffer_dirty(nilfs->ns_sbh[1]);
253 if (!sync_dirty_buffer(nilfs->ns_sbh[1]))
254 sbp = nilfs->ns_sbp[1];
255 }
256 }
257 if (sbp) {
258 spin_lock(&nilfs->ns_last_segment_lock);
259 nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq);
260 spin_unlock(&nilfs->ns_last_segment_lock);
261 }
783f6184
RK		 262 }
263
264 return err;
265}
266
e339ad31 267int nilfs_commit_super(struct nilfs_sb_info *sbi, int dupsb)
783f6184
RK		 268{
269 struct the_nilfs *nilfs = sbi->s_nilfs;
e339ad31 270 struct nilfs_super_block **sbp = nilfs->ns_sbp;
783f6184 271 sector_t nfreeblocks;
e339ad31 272 time_t t;
783f6184
RK		 273 int err;
274
275 /* nilfs->sem must be locked by the caller. */
e339ad31
RK		 276 if (sbp[0]->s_magic != NILFS_SUPER_MAGIC) {
277 if (sbp[1] && sbp[1]->s_magic == NILFS_SUPER_MAGIC)
278 nilfs_swap_super_block(nilfs);
279 else {
280 printk(KERN_CRIT "NILFS: superblock broke on dev %s\n",
281 sbi->s_super->s_id);
282 return -EIO;
283 }
284 }
783f6184
RK		 285 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
286 if (unlikely(err)) {
287 printk(KERN_ERR "NILFS: failed to count free blocks\n");
288 return err;
289 }
e339ad31
RK		 290 spin_lock(&nilfs->ns_last_segment_lock);
291 sbp[0]->s_last_seq = cpu_to_le64(nilfs->ns_last_seq);
292 sbp[0]->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg);
293 sbp[0]->s_last_cno = cpu_to_le64(nilfs->ns_last_cno);
294 spin_unlock(&nilfs->ns_last_segment_lock);
295
296 t = get_seconds();
297 nilfs->ns_sbwtime[0] = t;
298 sbp[0]->s_free_blocks_count = cpu_to_le64(nfreeblocks);
299 sbp[0]->s_wtime = cpu_to_le64(t);
300 sbp[0]->s_sum = 0;
301 sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed,
302 (unsigned char *)sbp[0],
303 nilfs->ns_sbsize));
304 if (dupsb && sbp[1]) {
305 memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
306 nilfs->ns_sbwtime[1] = t;
307 }
783f6184 308 sbi->s_super->s_dirt = 0;
e339ad31 309 return nilfs_sync_super(sbi, dupsb);
783f6184
RK		 310}
311
312static void nilfs_put_super(struct super_block *sb)
313{
314 struct nilfs_sb_info *sbi = NILFS_SB(sb);
315 struct the_nilfs *nilfs = sbi->s_nilfs;
316
6cfd0148
CH		 317 lock_kernel();
318
8c85e125
CH		 319 if (sb->s_dirt)
320 nilfs_write_super(sb);
321
783f6184
RK		 322 nilfs_detach_segment_constructor(sbi);
323
324 if (!(sb->s_flags & MS_RDONLY)) {
325 down_write(&nilfs->ns_sem);
e339ad31
RK		 326 nilfs->ns_sbp[0]->s_state = cpu_to_le16(nilfs->ns_mount_state);
327 nilfs_commit_super(sbi, 1);
783f6184
RK		 328 up_write(&nilfs->ns_sem);
329 }
3f82ff55
RK		 330 down_write(&nilfs->ns_sem);
331 if (nilfs->ns_current == sbi)
332 nilfs->ns_current = NULL;
333 up_write(&nilfs->ns_sem);
783f6184
RK		 334
335 nilfs_detach_checkpoint(sbi);
336 put_nilfs(sbi->s_nilfs);
337 sbi->s_super = NULL;
338 sb->s_fs_info = NULL;
339 kfree(sbi);
6cfd0148
CH		 340
341 unlock_kernel();
783f6184
RK		 342}
343
344/**
345 * nilfs_write_super - write super block(s) of NILFS
346 * @sb: super_block
347 *
348 * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and
349 * clears s_dirt. This function is called in the section protected by
350 * lock_super().
351 *
352 * The s_dirt flag is managed by each filesystem and we protect it by ns_sem
353 * of the struct the_nilfs. Lock order must be as follows:
354 *
355 * 1. lock_super()
356 * 2. down_write(&nilfs->ns_sem)
357 *
358 * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer
e339ad31 359 * of the super block (nilfs->ns_sbp[]).
783f6184
RK		 360 *
361 * In most cases, VFS functions call lock_super() before calling these
362 * methods. So we must be careful not to bring on deadlocks when using
363 * lock_super(); see generic_shutdown_super(), write_super(), and so on.
364 *
365 * Note that order of lock_kernel() and lock_super() depends on contexts
366 * of VFS. We should also note that lock_kernel() can be used in its
367 * protective section and only the outermost one has an effect.
368 */
369static void nilfs_write_super(struct super_block *sb)
370{
371 struct nilfs_sb_info *sbi = NILFS_SB(sb);
372 struct the_nilfs *nilfs = sbi->s_nilfs;
373
374 down_write(&nilfs->ns_sem);
e339ad31
RK		 375 if (!(sb->s_flags & MS_RDONLY)) {
376 struct nilfs_super_block **sbp = nilfs->ns_sbp;
377 u64 t = get_seconds();
378 int dupsb;
379
380 if (!nilfs_discontinued(nilfs) && t >= nilfs->ns_sbwtime[0] &&
381 t < nilfs->ns_sbwtime[0] + NILFS_SB_FREQ) {
382 up_write(&nilfs->ns_sem);
383 return;
384 }
385 dupsb = sbp[1] && t > nilfs->ns_sbwtime[1] + NILFS_ALTSB_FREQ;
386 nilfs_commit_super(sbi, dupsb);
387 }
783f6184
RK		 388 sb->s_dirt = 0;
389 up_write(&nilfs->ns_sem);
390}
391
392static int nilfs_sync_fs(struct super_block *sb, int wait)
393{
394 int err = 0;
395
d731e063
CH		 396 nilfs_write_super(sb);
397
783f6184
RK		 398 /* This function is called when super block should be written back */
399 if (wait)
400 err = nilfs_construct_segment(sb);
401 return err;
402}
403
404int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
405{
406 struct the_nilfs *nilfs = sbi->s_nilfs;
407 struct nilfs_checkpoint *raw_cp;
408 struct buffer_head *bh_cp;
409 int err;
410
411 down_write(&nilfs->ns_sem);
412 list_add(&sbi->s_list, &nilfs->ns_supers);
413 up_write(&nilfs->ns_sem);
414
415 sbi->s_ifile = nilfs_mdt_new(
416 nilfs, sbi->s_super, NILFS_IFILE_INO, NILFS_IFILE_GFP);
417 if (!sbi->s_ifile)
418 return -ENOMEM;
419
420 err = nilfs_palloc_init_blockgroup(sbi->s_ifile, nilfs->ns_inode_size);
421 if (unlikely(err))
422 goto failed;
423
424 err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp,
425 &bh_cp);
426 if (unlikely(err)) {
427 if (err == -ENOENT || err == -EINVAL) {
428 printk(KERN_ERR
429 "NILFS: Invalid checkpoint "
430 "(checkpoint number=%llu)\n",
431 (unsigned long long)cno);
432 err = -EINVAL;
433 }
434 goto failed;
435 }
436 err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode);
437 if (unlikely(err))
438 goto failed_bh;
439 atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
440 atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
441
442 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
443 return 0;
444
445 failed_bh:
446 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
447 failed:
448 nilfs_mdt_destroy(sbi->s_ifile);
449 sbi->s_ifile = NULL;
450
451 down_write(&nilfs->ns_sem);
452 list_del_init(&sbi->s_list);
453 up_write(&nilfs->ns_sem);
454
455 return err;
456}
457
458void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi)
459{
460 struct the_nilfs *nilfs = sbi->s_nilfs;
461
462 nilfs_mdt_clear(sbi->s_ifile);
463 nilfs_mdt_destroy(sbi->s_ifile);
464 sbi->s_ifile = NULL;
465 down_write(&nilfs->ns_sem);
466 list_del_init(&sbi->s_list);
467 up_write(&nilfs->ns_sem);
468}
469
470static int nilfs_mark_recovery_complete(struct nilfs_sb_info *sbi)
471{
472 struct the_nilfs *nilfs = sbi->s_nilfs;
473 int err = 0;
474
475 down_write(&nilfs->ns_sem);
476 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
477 nilfs->ns_mount_state |= NILFS_VALID_FS;
e339ad31 478 err = nilfs_commit_super(sbi, 1);
783f6184
RK		 479 if (likely(!err))
480 printk(KERN_INFO "NILFS: recovery complete.\n");
481 }
482 up_write(&nilfs->ns_sem);
483 return err;
484}
485
486static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
487{
488 struct super_block *sb = dentry->d_sb;
489 struct nilfs_sb_info *sbi = NILFS_SB(sb);
c306af23
RK		 490 struct the_nilfs *nilfs = sbi->s_nilfs;
491 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
783f6184
RK		 492 unsigned long long blocks;
493 unsigned long overhead;
494 unsigned long nrsvblocks;
495 sector_t nfreeblocks;
783f6184
RK		 496 int err;
497
498 /*
499 * Compute all of the segment blocks
500 *
501 * The blocks before first segment and after last segment
502 * are excluded.
503 */
504 blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments
505 - nilfs->ns_first_data_block;
506 nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment;
507
508 /*
509 * Compute the overhead
510 *
511 * When distributing meta data blocks outside semgent structure,
512 * We must count them as the overhead.
513 */
514 overhead = 0;
515
516 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
517 if (unlikely(err))
518 return err;
519
520 buf->f_type = NILFS_SUPER_MAGIC;
521 buf->f_bsize = sb->s_blocksize;
522 buf->f_blocks = blocks - overhead;
523 buf->f_bfree = nfreeblocks;
524 buf->f_bavail = (buf->f_bfree >= nrsvblocks) ?
525 (buf->f_bfree - nrsvblocks) : 0;
526 buf->f_files = atomic_read(&sbi->s_inodes_count);
527 buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */
528 buf->f_namelen = NILFS_NAME_LEN;
c306af23
RK		 529 buf->f_fsid.val[0] = (u32)id;
530 buf->f_fsid.val[1] = (u32)(id >> 32);
531
783f6184
RK		 532 return 0;
533}
534
535static struct super_operations nilfs_sops = {
536 .alloc_inode = nilfs_alloc_inode,
537 .destroy_inode = nilfs_destroy_inode,
538 .dirty_inode = nilfs_dirty_inode,
539 /* .write_inode = nilfs_write_inode, */
540 /* .put_inode = nilfs_put_inode, */
541 /* .drop_inode = nilfs_drop_inode, */
542 .delete_inode = nilfs_delete_inode,
543 .put_super = nilfs_put_super,
544 .write_super = nilfs_write_super,
545 .sync_fs = nilfs_sync_fs,
546 /* .write_super_lockfs */
547 /* .unlockfs */
548 .statfs = nilfs_statfs,
549 .remount_fs = nilfs_remount,
550 .clear_inode = nilfs_clear_inode,
551 /* .umount_begin */
552 /* .show_options */
553};
554
555static struct inode *
556nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation)
557{
558 struct inode *inode;
559
560 if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO &&
561 ino != NILFS_SKETCH_INO)
562 return ERR_PTR(-ESTALE);
563
564 inode = nilfs_iget(sb, ino);
565 if (IS_ERR(inode))
566 return ERR_CAST(inode);
567 if (generation && inode->i_generation != generation) {
568 iput(inode);
569 return ERR_PTR(-ESTALE);
570 }
571
572 return inode;
573}
574
575static struct dentry *
576nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
577 int fh_type)
578{
579 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
580 nilfs_nfs_get_inode);
581}
582
583static struct dentry *
584nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len,
585 int fh_type)
586{
587 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
588 nilfs_nfs_get_inode);
589}
590
591static struct export_operations nilfs_export_ops = {
592 .fh_to_dentry = nilfs_fh_to_dentry,
593 .fh_to_parent = nilfs_fh_to_parent,
594 .get_parent = nilfs_get_parent,
595};
596
597enum {
598 Opt_err_cont, Opt_err_panic, Opt_err_ro,
599 Opt_barrier, Opt_snapshot, Opt_order,
600 Opt_err,
601};
602
603static match_table_t tokens = {
604 {Opt_err_cont, "errors=continue"},
605 {Opt_err_panic, "errors=panic"},
606 {Opt_err_ro, "errors=remount-ro"},
607 {Opt_barrier, "barrier=%s"},
608 {Opt_snapshot, "cp=%u"},
609 {Opt_order, "order=%s"},
610 {Opt_err, NULL}
611};
612
613static int match_bool(substring_t *s, int *result)
614{
615 int len = s->to - s->from;
616
617 if (strncmp(s->from, "on", len) == 0)
618 *result = 1;
619 else if (strncmp(s->from, "off", len) == 0)
620 *result = 0;
621 else
622 return 1;
623 return 0;
624}
625
626static int parse_options(char *options, struct super_block *sb)
627{
628 struct nilfs_sb_info *sbi = NILFS_SB(sb);
629 char *p;
630 substring_t args[MAX_OPT_ARGS];
631 int option;
632
633 if (!options)
634 return 1;
635
636 while ((p = strsep(&options, ",")) != NULL) {
637 int token;
638 if (!*p)
639 continue;
640
641 token = match_token(p, tokens, args);
642 switch (token) {
643 case Opt_barrier:
644 if (match_bool(&args[0], &option))
645 return 0;
646 if (option)
647 nilfs_set_opt(sbi, BARRIER);
648 else
649 nilfs_clear_opt(sbi, BARRIER);
650 break;
651 case Opt_order:
652 if (strcmp(args[0].from, "relaxed") == 0)
653 /* Ordered data semantics */
654 nilfs_clear_opt(sbi, STRICT_ORDER);
655 else if (strcmp(args[0].from, "strict") == 0)
656 /* Strict in-order semantics */
657 nilfs_set_opt(sbi, STRICT_ORDER);
658 else
659 return 0;
660 break;
661 case Opt_err_panic:
662 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC);
663 break;
664 case Opt_err_ro:
665 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO);
666 break;
667 case Opt_err_cont:
668 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT);
669 break;
670 case Opt_snapshot:
671 if (match_int(&args[0], &option) || option <= 0)
672 return 0;
673 if (!(sb->s_flags & MS_RDONLY))
674 return 0;
675 sbi->s_snapshot_cno = option;
676 nilfs_set_opt(sbi, SNAPSHOT);
677 break;
678 default:
679 printk(KERN_ERR
680 "NILFS: Unrecognized mount option \"%s\"\n", p);
681 return 0;
682 }
683 }
684 return 1;
685}
686
687static inline void
688nilfs_set_default_options(struct nilfs_sb_info *sbi,
689 struct nilfs_super_block *sbp)
690{
691 sbi->s_mount_opt =
692 NILFS_MOUNT_ERRORS_CONT | NILFS_MOUNT_BARRIER;
693}
694
695static int nilfs_setup_super(struct nilfs_sb_info *sbi)
696{
697 struct the_nilfs *nilfs = sbi->s_nilfs;
e339ad31 698 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
783f6184
RK		 699 int max_mnt_count = le16_to_cpu(sbp->s_max_mnt_count);
700 int mnt_count = le16_to_cpu(sbp->s_mnt_count);
701
702 /* nilfs->sem must be locked by the caller. */
703 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
704 printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n");
705 } else if (nilfs->ns_mount_state & NILFS_ERROR_FS) {
706 printk(KERN_WARNING
707 "NILFS warning: mounting fs with errors\n");
708#if 0
709 } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) {
710 printk(KERN_WARNING
711 "NILFS warning: maximal mount count reached\n");
712#endif
713 }
714 if (!max_mnt_count)
715 sbp->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT);
716
717 sbp->s_mnt_count = cpu_to_le16(mnt_count + 1);
718 sbp->s_state = cpu_to_le16(le16_to_cpu(sbp->s_state) & ~NILFS_VALID_FS);
719 sbp->s_mtime = cpu_to_le64(get_seconds());
e339ad31 720 return nilfs_commit_super(sbi, 1);
783f6184
RK		 721}
722
e339ad31
RK		 723struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb,
724 u64 pos, int blocksize,
725 struct buffer_head **pbh)
783f6184 726{
e339ad31
RK		 727 unsigned long long sb_index = pos;
728 unsigned long offset;
783f6184 729
e339ad31 730 offset = do_div(sb_index, blocksize);
783f6184 731 *pbh = sb_bread(sb, sb_index);
e339ad31 732 if (!*pbh)
783f6184 733 return NULL;
783f6184
RK		 734 return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset);
735}
736
783f6184
RK		 737int nilfs_store_magic_and_option(struct super_block *sb,
738 struct nilfs_super_block *sbp,
739 char *data)
740{
741 struct nilfs_sb_info *sbi = NILFS_SB(sb);
742
783f6184
RK		 743 sb->s_magic = le16_to_cpu(sbp->s_magic);
744
745 /* FS independent flags */
746#ifdef NILFS_ATIME_DISABLE
747 sb->s_flags |= MS_NOATIME;
748#endif
749
783f6184
RK		 750 nilfs_set_default_options(sbi, sbp);
751
752 sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid);
753 sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid);
754 sbi->s_interval = le32_to_cpu(sbp->s_c_interval);
755 sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max);
756
e339ad31 757 return !parse_options(data, sb) ? -EINVAL : 0 ;
783f6184
RK		 758}
759
760/**
761 * nilfs_fill_super() - initialize a super block instance
762 * @sb: super_block
763 * @data: mount options
764 * @silent: silent mode flag
765 * @nilfs: the_nilfs struct
766 *
767 * This function is called exclusively by bd_mount_mutex.
768 * So, the recovery process is protected from other simultaneous mounts.
769 */
770static int
771nilfs_fill_super(struct super_block *sb, void *data, int silent,
772 struct the_nilfs *nilfs)
773{
774 struct nilfs_sb_info *sbi;
775 struct inode *root;
776 __u64 cno;
777 int err;
778
779 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
780 if (!sbi)
781 return -ENOMEM;
782
783 sb->s_fs_info = sbi;
784
785 get_nilfs(nilfs);
786 sbi->s_nilfs = nilfs;
787 sbi->s_super = sb;
788
789 err = init_nilfs(nilfs, sbi, (char *)data);
790 if (err)
791 goto failed_sbi;
792
793 spin_lock_init(&sbi->s_inode_lock);
794 INIT_LIST_HEAD(&sbi->s_dirty_files);
795 INIT_LIST_HEAD(&sbi->s_list);
796
797 /*
798 * Following initialization is overlapped because
799 * nilfs_sb_info structure has been cleared at the beginning.
800 * But we reserve them to keep our interest and make ready
801 * for the future change.
802 */
803 get_random_bytes(&sbi->s_next_generation,
804 sizeof(sbi->s_next_generation));
805 spin_lock_init(&sbi->s_next_gen_lock);
806
807 sb->s_op = &nilfs_sops;
808 sb->s_export_op = &nilfs_export_ops;
809 sb->s_root = NULL;
61239230 810 sb->s_time_gran = 1;
783f6184
RK		 811
812 if (!nilfs_loaded(nilfs)) {
813 err = load_nilfs(nilfs, sbi);
814 if (err)
815 goto failed_sbi;
816 }
817 cno = nilfs_last_cno(nilfs);
818
819 if (sb->s_flags & MS_RDONLY) {
820 if (nilfs_test_opt(sbi, SNAPSHOT)) {
1f5abe7e
RK		 821 err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile,
822 sbi->s_snapshot_cno);
823 if (err < 0)
824 goto failed_sbi;
825 if (!err) {
783f6184
RK		 826 printk(KERN_ERR
827 "NILFS: The specified checkpoint is "
828 "not a snapshot "
829 "(checkpoint number=%llu).\n",
830 (unsigned long long)sbi->s_snapshot_cno);
831 err = -EINVAL;
832 goto failed_sbi;
833 }
834 cno = sbi->s_snapshot_cno;
835 } else
836 /* Read-only mount */
837 sbi->s_snapshot_cno = cno;
838 }
839
840 err = nilfs_attach_checkpoint(sbi, cno);
841 if (err) {
842 printk(KERN_ERR "NILFS: error loading a checkpoint"
843 " (checkpoint number=%llu).\n", (unsigned long long)cno);
844 goto failed_sbi;
845 }
846
847 if (!(sb->s_flags & MS_RDONLY)) {
cece5520 848 err = nilfs_attach_segment_constructor(sbi);
783f6184
RK		 849 if (err)
850 goto failed_checkpoint;
851 }
852
853 root = nilfs_iget(sb, NILFS_ROOT_INO);
854 if (IS_ERR(root)) {
855 printk(KERN_ERR "NILFS: get root inode failed\n");
856 err = PTR_ERR(root);
857 goto failed_segctor;
858 }
859 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
860 iput(root);
861 printk(KERN_ERR "NILFS: corrupt root inode.\n");
862 err = -EINVAL;
863 goto failed_segctor;
864 }
865 sb->s_root = d_alloc_root(root);
866 if (!sb->s_root) {
867 iput(root);
868 printk(KERN_ERR "NILFS: get root dentry failed\n");
869 err = -ENOMEM;
870 goto failed_segctor;
871 }
872
873 if (!(sb->s_flags & MS_RDONLY)) {
874 down_write(&nilfs->ns_sem);
875 nilfs_setup_super(sbi);
876 up_write(&nilfs->ns_sem);
877 }
878
879 err = nilfs_mark_recovery_complete(sbi);
880 if (unlikely(err)) {
881 printk(KERN_ERR "NILFS: recovery failed.\n");
882 goto failed_root;
883 }
884
3f82ff55
RK		 885 down_write(&nilfs->ns_sem);
886 if (!nilfs_test_opt(sbi, SNAPSHOT))
887 nilfs->ns_current = sbi;
888 up_write(&nilfs->ns_sem);
889
783f6184
RK		 890 return 0;
891
892 failed_root:
893 dput(sb->s_root);
894 sb->s_root = NULL;
895
896 failed_segctor:
897 nilfs_detach_segment_constructor(sbi);
898
899 failed_checkpoint:
900 nilfs_detach_checkpoint(sbi);
901
902 failed_sbi:
903 put_nilfs(nilfs);
904 sb->s_fs_info = NULL;
905 kfree(sbi);
906 return err;
907}
908
909static int nilfs_remount(struct super_block *sb, int *flags, char *data)
910{
911 struct nilfs_sb_info *sbi = NILFS_SB(sb);
912 struct nilfs_super_block *sbp;
913 struct the_nilfs *nilfs = sbi->s_nilfs;
914 unsigned long old_sb_flags;
915 struct nilfs_mount_options old_opts;
916 int err;
917
337eb00a
AIB		 918 lock_kernel();
919
783f6184
RK		 920 old_sb_flags = sb->s_flags;
921 old_opts.mount_opt = sbi->s_mount_opt;
922 old_opts.snapshot_cno = sbi->s_snapshot_cno;
923
924 if (!parse_options(data, sb)) {
925 err = -EINVAL;
926 goto restore_opts;
927 }
928 sb->s_flags = (sb->s_flags & ~MS_POSIXACL);
929
930 if ((*flags & MS_RDONLY) &&
931 sbi->s_snapshot_cno != old_opts.snapshot_cno) {
932 printk(KERN_WARNING "NILFS (device %s): couldn't "
933 "remount to a different snapshot. \n",
934 sb->s_id);
935 err = -EINVAL;
936 goto restore_opts;
937 }
938
939 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
940 goto out;
941 if (*flags & MS_RDONLY) {
942 /* Shutting down the segment constructor */
943 nilfs_detach_segment_constructor(sbi);
944 sb->s_flags |= MS_RDONLY;
945
946 sbi->s_snapshot_cno = nilfs_last_cno(nilfs);
947 /* nilfs_set_opt(sbi, SNAPSHOT); */
948
949 /*
950 * Remounting a valid RW partition RDONLY, so set
951 * the RDONLY flag and then mark the partition as valid again.
952 */
953 down_write(&nilfs->ns_sem);
e339ad31 954 sbp = nilfs->ns_sbp[0];
783f6184
RK		 955 if (!(sbp->s_state & le16_to_cpu(NILFS_VALID_FS)) &&
956 (nilfs->ns_mount_state & NILFS_VALID_FS))
957 sbp->s_state = cpu_to_le16(nilfs->ns_mount_state);
958 sbp->s_mtime = cpu_to_le64(get_seconds());
e339ad31 959 nilfs_commit_super(sbi, 1);
783f6184
RK		 960 up_write(&nilfs->ns_sem);
961 } else {
962 /*
963 * Mounting a RDONLY partition read-write, so reread and
964 * store the current valid flag. (It may have been changed
965 * by fsck since we originally mounted the partition.)
966 */
967 down(&sb->s_bdev->bd_mount_sem);
3f82ff55
RK		 968 down_read(&nilfs->ns_sem);
969 if (nilfs->ns_current && nilfs->ns_current != sbi) {
783f6184 970 printk(KERN_WARNING "NILFS (device %s): couldn't "
3f82ff55 971 "remount because an RW-mount exists.\n",
783f6184 972 sb->s_id);
3f82ff55 973 up_read(&nilfs->ns_sem);
783f6184
RK		 974 err = -EBUSY;
975 goto rw_remount_failed;
976 }
3f82ff55 977 up_read(&nilfs->ns_sem);
783f6184
RK		 978 if (sbi->s_snapshot_cno != nilfs_last_cno(nilfs)) {
979 printk(KERN_WARNING "NILFS (device %s): couldn't "
980 "remount because the current RO-mount is not "
981 "the latest one.\n",
982 sb->s_id);
983 err = -EINVAL;
984 goto rw_remount_failed;
985 }
986 sb->s_flags &= ~MS_RDONLY;
987 nilfs_clear_opt(sbi, SNAPSHOT);
988 sbi->s_snapshot_cno = 0;
989
cece5520 990 err = nilfs_attach_segment_constructor(sbi);
783f6184
RK		 991 if (err)
992 goto rw_remount_failed;
993
994 down_write(&nilfs->ns_sem);
995 nilfs_setup_super(sbi);
3f82ff55 996 nilfs->ns_current = sbi;
783f6184
RK		 997 up_write(&nilfs->ns_sem);
998
999 up(&sb->s_bdev->bd_mount_sem);
1000 }
1001 out:
337eb00a 1002 unlock_kernel();
783f6184
RK		 1003 return 0;
1004
1005 rw_remount_failed:
1006 up(&sb->s_bdev->bd_mount_sem);
1007 restore_opts:
1008 sb->s_flags = old_sb_flags;
1009 sbi->s_mount_opt = old_opts.mount_opt;
1010 sbi->s_snapshot_cno = old_opts.snapshot_cno;
337eb00a 1011 unlock_kernel();
783f6184
RK		 1012 return err;
1013}
1014
1015struct nilfs_super_data {
1016 struct block_device *bdev;
1017 __u64 cno;
1018 int flags;
1019};
1020
1021/**
1022 * nilfs_identify - pre-read mount options needed to identify mount instance
1023 * @data: mount options
1024 * @sd: nilfs_super_data
1025 */
1026static int nilfs_identify(char *data, struct nilfs_super_data *sd)
1027{
1028 char *p, *options = data;
1029 substring_t args[MAX_OPT_ARGS];
1030 int option, token;
1031 int ret = 0;
1032
1033 do {
1034 p = strsep(&options, ",");
1035 if (p != NULL && *p) {
1036 token = match_token(p, tokens, args);
1037 if (token == Opt_snapshot) {
1038 if (!(sd->flags & MS_RDONLY))
1039 ret++;
1040 else {
1041 ret = match_int(&args[0], &option);
1042 if (!ret) {
1043 if (option > 0)
1044 sd->cno = option;
1045 else
1046 ret++;
1047 }
1048 }
1049 }
1050 if (ret)
1051 printk(KERN_ERR
1052 "NILFS: invalid mount option: %s\n", p);
1053 }
1054 if (!options)
1055 break;
1056 BUG_ON(options == data);
1057 *(options - 1) = ',';
1058 } while (!ret);
1059 return ret;
1060}
1061
1062static int nilfs_set_bdev_super(struct super_block *s, void *data)
1063{
1064 struct nilfs_super_data *sd = data;
1065
1066 s->s_bdev = sd->bdev;
1067 s->s_dev = s->s_bdev->bd_dev;
1068 return 0;
1069}
1070
1071static int nilfs_test_bdev_super(struct super_block *s, void *data)
783f6184
RK		 1072{
1073 struct nilfs_super_data *sd = data;
1074 int ret;
1075
1076 if (s->s_bdev != sd->bdev)
1077 return 0;
1078
1079 if (!((s->s_flags | sd->flags) & MS_RDONLY))
1080 return 1; /* Reuse an old R/W-mode super_block */
1081
1082 if (s->s_flags & sd->flags & MS_RDONLY) {
1083 if (down_read_trylock(&s->s_umount)) {
1084 ret = s->s_root &&
1085 (sd->cno == NILFS_SB(s)->s_snapshot_cno);
1086 up_read(&s->s_umount);
1087 /*
1088 * This path is locked with sb_lock by sget().
1089 * So, drop_super() causes deadlock.
1090 */
1091 return ret;
1092 }
1093 }
1094 return 0;
1095}
1096
1097static int
1098nilfs_get_sb(struct file_system_type *fs_type, int flags,
1099 const char *dev_name, void *data, struct vfsmount *mnt)
1100{
1101 struct nilfs_super_data sd;
33c8e57c
RK		 1102 struct super_block *s;
1103 struct the_nilfs *nilfs;
783f6184
RK		 1104 int err, need_to_close = 1;
1105
1106 sd.bdev = open_bdev_exclusive(dev_name, flags, fs_type);
1107 if (IS_ERR(sd.bdev))
1108 return PTR_ERR(sd.bdev);
1109
1110 /*
1111 * To get mount instance using sget() vfs-routine, NILFS needs
1112 * much more information than normal filesystems to identify mount
1113 * instance. For snapshot mounts, not only a mount type (ro-mount
1114 * or rw-mount) but also a checkpoint number is required.
1115 * The results are passed in sget() using nilfs_super_data.
1116 */
1117 sd.cno = 0;
1118 sd.flags = flags;
1119 if (nilfs_identify((char *)data, &sd)) {
1120 err = -EINVAL;
1121 goto failed;
1122 }
1123
33c8e57c
RK		 1124 nilfs = find_or_create_nilfs(sd.bdev);
1125 if (!nilfs) {
1126 err = -ENOMEM;
1127 goto failed;
1128 }
1129
783f6184 1130 down(&sd.bdev->bd_mount_sem);
3f82ff55
RK		 1131
1132 if (!sd.cno) {
1133 /*
1134 * Check if an exclusive mount exists or not.
1135 * Snapshot mounts coexist with a current mount
1136 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1137 * ro-mount are mutually exclusive.
1138 */
1139 down_read(&nilfs->ns_sem);
1140 if (nilfs->ns_current &&
1141 ((nilfs->ns_current->s_super->s_flags ^ flags)
1142 & MS_RDONLY)) {
1143 up_read(&nilfs->ns_sem);
1144 err = -EBUSY;
1145 goto failed_unlock;
1146 }
1147 up_read(&nilfs->ns_sem);
783f6184
RK		 1148 }
1149
1150 /*
33c8e57c 1151 * Search specified snapshot or R/W mode super_block
783f6184 1152 */
33c8e57c
RK		 1153 if (!sd.cno)
1154 /* trying to get the latest checkpoint. */
1155 sd.cno = nilfs_last_cno(nilfs);
783f6184 1156
33c8e57c
RK		 1157 s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd);
1158 if (IS_ERR(s)) {
1159 err = PTR_ERR(s);
1160 goto failed_unlock;
783f6184
RK		 1161 }
1162
1163 if (!s->s_root) {
1164 char b[BDEVNAME_SIZE];
1165
33c8e57c 1166 /* New superblock instance created */
783f6184
RK		 1167 s->s_flags = flags;
1168 strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id));
1169 sb_set_blocksize(s, block_size(sd.bdev));
1170
1171 err = nilfs_fill_super(s, data, flags & MS_VERBOSE, nilfs);
1172 if (err)
1173 goto cancel_new;
1174
1175 s->s_flags |= MS_ACTIVE;
1176 need_to_close = 0;
783f6184
RK		 1177 }
1178
1179 up(&sd.bdev->bd_mount_sem);
1180 put_nilfs(nilfs);
1181 if (need_to_close)
1182 close_bdev_exclusive(sd.bdev, flags);
1183 simple_set_mnt(mnt, s);
1184 return 0;
1185
783f6184
RK		 1186 failed_unlock:
1187 up(&sd.bdev->bd_mount_sem);
33c8e57c 1188 put_nilfs(nilfs);
783f6184
RK		 1189 failed:
1190 close_bdev_exclusive(sd.bdev, flags);
1191
1192 return err;
1193
1194 cancel_new:
1195 /* Abandoning the newly allocated superblock */
1196 up(&sd.bdev->bd_mount_sem);
33c8e57c 1197 put_nilfs(nilfs);
783f6184
RK		 1198 up_write(&s->s_umount);
1199 deactivate_super(s);
1200 /*
1201 * deactivate_super() invokes close_bdev_exclusive().
1202 * We must finish all post-cleaning before this call;
1203 * put_nilfs() and unlocking bd_mount_sem need the block device.
1204 */
1205 return err;
1206}
1207
783f6184
RK		 1208struct file_system_type nilfs_fs_type = {
1209 .owner = THIS_MODULE,
1210 .name = "nilfs2",
1211 .get_sb = nilfs_get_sb,
1212 .kill_sb = kill_block_super,
1213 .fs_flags = FS_REQUIRES_DEV,
1214};
1215
1216static int __init init_nilfs_fs(void)
1217{
1218 int err;
1219
1220 err = nilfs_init_inode_cache();
1221 if (err)
1222 goto failed;
1223
1224 err = nilfs_init_transaction_cache();
1225 if (err)
1226 goto failed_inode_cache;
1227
1228 err = nilfs_init_segbuf_cache();
1229 if (err)
1230 goto failed_transaction_cache;
1231
1232 err = nilfs_btree_path_cache_init();
1233 if (err)
1234 goto failed_segbuf_cache;
1235
1236 err = register_filesystem(&nilfs_fs_type);
1237 if (err)
1238 goto failed_btree_path_cache;
1239
1240 return 0;
1241
1242 failed_btree_path_cache:
1243 nilfs_btree_path_cache_destroy();
1244
1245 failed_segbuf_cache:
1246 nilfs_destroy_segbuf_cache();
1247
1248 failed_transaction_cache:
1249 nilfs_destroy_transaction_cache();
1250
1251 failed_inode_cache:
1252 nilfs_destroy_inode_cache();
1253
1254 failed:
1255 return err;
1256}
1257
1258static void __exit exit_nilfs_fs(void)
1259{
1260 nilfs_destroy_segbuf_cache();
1261 nilfs_destroy_transaction_cache();
1262 nilfs_destroy_inode_cache();
1263 nilfs_btree_path_cache_destroy();
1264 unregister_filesystem(&nilfs_fs_type);
1265}
1266
1267module_init(init_nilfs_fs)
1268module_exit(exit_nilfs_fs)
Linux kernel - Deliverables
This page took 0.091965 seconds and 5 git commands to generate.