Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
[linux-2.6-block.git] / fs / jfs / super.c
1 /*
2  *   Copyright (C) International Business Machines Corp., 2000-2004
3  *   Portions Copyright (C) Christoph Hellwig, 2001-2002
4  *
5  *   This program is free software;  you can redistribute it and/or modify
6  *   it under the terms of the GNU General Public License as published by
7  *   the Free Software Foundation; either version 2 of the License, or
8  *   (at your option) any later version.
9  *
10  *   This program is distributed in the hope that it will be useful,
11  *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
12  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
13  *   the GNU General Public License for more details.
14  *
15  *   You should have received a copy of the GNU General Public License
16  *   along with this program;  if not, write to the Free Software
17  *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18  */
19
20 #include <linux/fs.h>
21 #include <linux/module.h>
22 #include <linux/parser.h>
23 #include <linux/completion.h>
24 #include <linux/vfs.h>
25 #include <linux/quotaops.h>
26 #include <linux/mount.h>
27 #include <linux/moduleparam.h>
28 #include <linux/kthread.h>
29 #include <linux/posix_acl.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/crc32.h>
33 #include <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/seq_file.h>
36
37 #include "jfs_incore.h"
38 #include "jfs_filsys.h"
39 #include "jfs_inode.h"
40 #include "jfs_metapage.h"
41 #include "jfs_superblock.h"
42 #include "jfs_dmap.h"
43 #include "jfs_imap.h"
44 #include "jfs_acl.h"
45 #include "jfs_debug.h"
46
47 MODULE_DESCRIPTION("The Journaled Filesystem (JFS)");
48 MODULE_AUTHOR("Steve Best/Dave Kleikamp/Barry Arndt, IBM");
49 MODULE_LICENSE("GPL");
50
51 static struct kmem_cache * jfs_inode_cachep;
52
53 static const struct super_operations jfs_super_operations;
54 static const struct export_operations jfs_export_operations;
55 static struct file_system_type jfs_fs_type;
56
57 #define MAX_COMMIT_THREADS 64
58 static int commit_threads = 0;
59 module_param(commit_threads, int, 0);
60 MODULE_PARM_DESC(commit_threads, "Number of commit threads");
61
62 static struct task_struct *jfsCommitThread[MAX_COMMIT_THREADS];
63 struct task_struct *jfsIOthread;
64 struct task_struct *jfsSyncThread;
65
66 #ifdef CONFIG_JFS_DEBUG
67 int jfsloglevel = JFS_LOGLEVEL_WARN;
68 module_param(jfsloglevel, int, 0644);
69 MODULE_PARM_DESC(jfsloglevel, "Specify JFS loglevel (0, 1 or 2)");
70 #endif
71
72 static void jfs_handle_error(struct super_block *sb)
73 {
74         struct jfs_sb_info *sbi = JFS_SBI(sb);
75
76         if (sb->s_flags & MS_RDONLY)
77                 return;
78
79         updateSuper(sb, FM_DIRTY);
80
81         if (sbi->flag & JFS_ERR_PANIC)
82                 panic("JFS (device %s): panic forced after error\n",
83                         sb->s_id);
84         else if (sbi->flag & JFS_ERR_REMOUNT_RO) {
85                 jfs_err("ERROR: (device %s): remounting filesystem "
86                         "as read-only\n",
87                         sb->s_id);
88                 sb->s_flags |= MS_RDONLY;
89         }
90
91         /* nothing is done for continue beyond marking the superblock dirty */
92 }
93
94 void jfs_error(struct super_block *sb, const char * function, ...)
95 {
96         static char error_buf[256];
97         va_list args;
98
99         va_start(args, function);
100         vsnprintf(error_buf, sizeof(error_buf), function, args);
101         va_end(args);
102
103         printk(KERN_ERR "ERROR: (device %s): %s\n", sb->s_id, error_buf);
104
105         jfs_handle_error(sb);
106 }
107
108 static struct inode *jfs_alloc_inode(struct super_block *sb)
109 {
110         struct jfs_inode_info *jfs_inode;
111
112         jfs_inode = kmem_cache_alloc(jfs_inode_cachep, GFP_NOFS);
113         if (!jfs_inode)
114                 return NULL;
115         return &jfs_inode->vfs_inode;
116 }
117
118 static void jfs_i_callback(struct rcu_head *head)
119 {
120         struct inode *inode = container_of(head, struct inode, i_rcu);
121         struct jfs_inode_info *ji = JFS_IP(inode);
122         kmem_cache_free(jfs_inode_cachep, ji);
123 }
124
125 static void jfs_destroy_inode(struct inode *inode)
126 {
127         struct jfs_inode_info *ji = JFS_IP(inode);
128
129         BUG_ON(!list_empty(&ji->anon_inode_list));
130
131         spin_lock_irq(&ji->ag_lock);
132         if (ji->active_ag != -1) {
133                 struct bmap *bmap = JFS_SBI(inode->i_sb)->bmap;
134                 atomic_dec(&bmap->db_active[ji->active_ag]);
135                 ji->active_ag = -1;
136         }
137         spin_unlock_irq(&ji->ag_lock);
138         call_rcu(&inode->i_rcu, jfs_i_callback);
139 }
140
141 static int jfs_statfs(struct dentry *dentry, struct kstatfs *buf)
142 {
143         struct jfs_sb_info *sbi = JFS_SBI(dentry->d_sb);
144         s64 maxinodes;
145         struct inomap *imap = JFS_IP(sbi->ipimap)->i_imap;
146
147         jfs_info("In jfs_statfs");
148         buf->f_type = JFS_SUPER_MAGIC;
149         buf->f_bsize = sbi->bsize;
150         buf->f_blocks = sbi->bmap->db_mapsize;
151         buf->f_bfree = sbi->bmap->db_nfree;
152         buf->f_bavail = sbi->bmap->db_nfree;
153         /*
154          * If we really return the number of allocated & free inodes, some
155          * applications will fail because they won't see enough free inodes.
156          * We'll try to calculate some guess as to how may inodes we can
157          * really allocate
158          *
159          * buf->f_files = atomic_read(&imap->im_numinos);
160          * buf->f_ffree = atomic_read(&imap->im_numfree);
161          */
162         maxinodes = min((s64) atomic_read(&imap->im_numinos) +
163                         ((sbi->bmap->db_nfree >> imap->im_l2nbperiext)
164                          << L2INOSPEREXT), (s64) 0xffffffffLL);
165         buf->f_files = maxinodes;
166         buf->f_ffree = maxinodes - (atomic_read(&imap->im_numinos) -
167                                     atomic_read(&imap->im_numfree));
168         buf->f_fsid.val[0] = (u32)crc32_le(0, sbi->uuid, sizeof(sbi->uuid)/2);
169         buf->f_fsid.val[1] = (u32)crc32_le(0, sbi->uuid + sizeof(sbi->uuid)/2,
170                                         sizeof(sbi->uuid)/2);
171
172         buf->f_namelen = JFS_NAME_MAX;
173         return 0;
174 }
175
176 static void jfs_put_super(struct super_block *sb)
177 {
178         struct jfs_sb_info *sbi = JFS_SBI(sb);
179         int rc;
180
181         jfs_info("In jfs_put_super");
182
183         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
184
185         rc = jfs_umount(sb);
186         if (rc)
187                 jfs_err("jfs_umount failed with return code %d", rc);
188
189         unload_nls(sbi->nls_tab);
190
191         truncate_inode_pages(sbi->direct_inode->i_mapping, 0);
192         iput(sbi->direct_inode);
193
194         kfree(sbi);
195 }
196
197 enum {
198         Opt_integrity, Opt_nointegrity, Opt_iocharset, Opt_resize,
199         Opt_resize_nosize, Opt_errors, Opt_ignore, Opt_err, Opt_quota,
200         Opt_usrquota, Opt_grpquota, Opt_uid, Opt_gid, Opt_umask
201 };
202
203 static const match_table_t tokens = {
204         {Opt_integrity, "integrity"},
205         {Opt_nointegrity, "nointegrity"},
206         {Opt_iocharset, "iocharset=%s"},
207         {Opt_resize, "resize=%u"},
208         {Opt_resize_nosize, "resize"},
209         {Opt_errors, "errors=%s"},
210         {Opt_ignore, "noquota"},
211         {Opt_ignore, "quota"},
212         {Opt_usrquota, "usrquota"},
213         {Opt_grpquota, "grpquota"},
214         {Opt_uid, "uid=%u"},
215         {Opt_gid, "gid=%u"},
216         {Opt_umask, "umask=%u"},
217         {Opt_err, NULL}
218 };
219
220 static int parse_options(char *options, struct super_block *sb, s64 *newLVSize,
221                          int *flag)
222 {
223         void *nls_map = (void *)-1;     /* -1: no change;  NULL: none */
224         char *p;
225         struct jfs_sb_info *sbi = JFS_SBI(sb);
226
227         *newLVSize = 0;
228
229         if (!options)
230                 return 1;
231
232         while ((p = strsep(&options, ",")) != NULL) {
233                 substring_t args[MAX_OPT_ARGS];
234                 int token;
235                 if (!*p)
236                         continue;
237
238                 token = match_token(p, tokens, args);
239                 switch (token) {
240                 case Opt_integrity:
241                         *flag &= ~JFS_NOINTEGRITY;
242                         break;
243                 case Opt_nointegrity:
244                         *flag |= JFS_NOINTEGRITY;
245                         break;
246                 case Opt_ignore:
247                         /* Silently ignore the quota options */
248                         /* Don't do anything ;-) */
249                         break;
250                 case Opt_iocharset:
251                         if (nls_map && nls_map != (void *) -1)
252                                 unload_nls(nls_map);
253                         if (!strcmp(args[0].from, "none"))
254                                 nls_map = NULL;
255                         else {
256                                 nls_map = load_nls(args[0].from);
257                                 if (!nls_map) {
258                                         printk(KERN_ERR
259                                                "JFS: charset not found\n");
260                                         goto cleanup;
261                                 }
262                         }
263                         break;
264                 case Opt_resize:
265                 {
266                         char *resize = args[0].from;
267                         *newLVSize = simple_strtoull(resize, &resize, 0);
268                         break;
269                 }
270                 case Opt_resize_nosize:
271                 {
272                         *newLVSize = sb->s_bdev->bd_inode->i_size >>
273                                 sb->s_blocksize_bits;
274                         if (*newLVSize == 0)
275                                 printk(KERN_ERR
276                                        "JFS: Cannot determine volume size\n");
277                         break;
278                 }
279                 case Opt_errors:
280                 {
281                         char *errors = args[0].from;
282                         if (!errors || !*errors)
283                                 goto cleanup;
284                         if (!strcmp(errors, "continue")) {
285                                 *flag &= ~JFS_ERR_REMOUNT_RO;
286                                 *flag &= ~JFS_ERR_PANIC;
287                                 *flag |= JFS_ERR_CONTINUE;
288                         } else if (!strcmp(errors, "remount-ro")) {
289                                 *flag &= ~JFS_ERR_CONTINUE;
290                                 *flag &= ~JFS_ERR_PANIC;
291                                 *flag |= JFS_ERR_REMOUNT_RO;
292                         } else if (!strcmp(errors, "panic")) {
293                                 *flag &= ~JFS_ERR_CONTINUE;
294                                 *flag &= ~JFS_ERR_REMOUNT_RO;
295                                 *flag |= JFS_ERR_PANIC;
296                         } else {
297                                 printk(KERN_ERR
298                                        "JFS: %s is an invalid error handler\n",
299                                        errors);
300                                 goto cleanup;
301                         }
302                         break;
303                 }
304
305 #ifdef CONFIG_QUOTA
306                 case Opt_quota:
307                 case Opt_usrquota:
308                         *flag |= JFS_USRQUOTA;
309                         break;
310                 case Opt_grpquota:
311                         *flag |= JFS_GRPQUOTA;
312                         break;
313 #else
314                 case Opt_usrquota:
315                 case Opt_grpquota:
316                 case Opt_quota:
317                         printk(KERN_ERR
318                                "JFS: quota operations not supported\n");
319                         break;
320 #endif
321                 case Opt_uid:
322                 {
323                         char *uid = args[0].from;
324                         uid_t val = simple_strtoul(uid, &uid, 0);
325                         sbi->uid = make_kuid(current_user_ns(), val);
326                         if (!uid_valid(sbi->uid))
327                                 goto cleanup;
328                         break;
329                 }
330                 case Opt_gid:
331                 {
332                         char *gid = args[0].from;
333                         gid_t val = simple_strtoul(gid, &gid, 0);
334                         sbi->gid = make_kgid(current_user_ns(), val);
335                         if (!gid_valid(sbi->gid))
336                                 goto cleanup;
337                         break;
338                 }
339                 case Opt_umask:
340                 {
341                         char *umask = args[0].from;
342                         sbi->umask = simple_strtoul(umask, &umask, 8);
343                         if (sbi->umask & ~0777) {
344                                 printk(KERN_ERR
345                                        "JFS: Invalid value of umask\n");
346                                 goto cleanup;
347                         }
348                         break;
349                 }
350                 default:
351                         printk("jfs: Unrecognized mount option \"%s\" "
352                                         " or missing value\n", p);
353                         goto cleanup;
354                 }
355         }
356
357         if (nls_map != (void *) -1) {
358                 /* Discard old (if remount) */
359                 unload_nls(sbi->nls_tab);
360                 sbi->nls_tab = nls_map;
361         }
362         return 1;
363
364 cleanup:
365         if (nls_map && nls_map != (void *) -1)
366                 unload_nls(nls_map);
367         return 0;
368 }
369
370 static int jfs_remount(struct super_block *sb, int *flags, char *data)
371 {
372         s64 newLVSize = 0;
373         int rc = 0;
374         int flag = JFS_SBI(sb)->flag;
375         int ret;
376
377         if (!parse_options(data, sb, &newLVSize, &flag)) {
378                 return -EINVAL;
379         }
380
381         if (newLVSize) {
382                 if (sb->s_flags & MS_RDONLY) {
383                         printk(KERN_ERR
384                   "JFS: resize requires volume to be mounted read-write\n");
385                         return -EROFS;
386                 }
387                 rc = jfs_extendfs(sb, newLVSize, 0);
388                 if (rc)
389                         return rc;
390         }
391
392         if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
393                 /*
394                  * Invalidate any previously read metadata.  fsck may have
395                  * changed the on-disk data since we mounted r/o
396                  */
397                 truncate_inode_pages(JFS_SBI(sb)->direct_inode->i_mapping, 0);
398
399                 JFS_SBI(sb)->flag = flag;
400                 ret = jfs_mount_rw(sb, 1);
401
402                 /* mark the fs r/w for quota activity */
403                 sb->s_flags &= ~MS_RDONLY;
404
405                 dquot_resume(sb, -1);
406                 return ret;
407         }
408         if ((!(sb->s_flags & MS_RDONLY)) && (*flags & MS_RDONLY)) {
409                 rc = dquot_suspend(sb, -1);
410                 if (rc < 0) {
411                         return rc;
412                 }
413                 rc = jfs_umount_rw(sb);
414                 JFS_SBI(sb)->flag = flag;
415                 return rc;
416         }
417         if ((JFS_SBI(sb)->flag & JFS_NOINTEGRITY) != (flag & JFS_NOINTEGRITY))
418                 if (!(sb->s_flags & MS_RDONLY)) {
419                         rc = jfs_umount_rw(sb);
420                         if (rc)
421                                 return rc;
422
423                         JFS_SBI(sb)->flag = flag;
424                         ret = jfs_mount_rw(sb, 1);
425                         return ret;
426                 }
427         JFS_SBI(sb)->flag = flag;
428
429         return 0;
430 }
431
432 static int jfs_fill_super(struct super_block *sb, void *data, int silent)
433 {
434         struct jfs_sb_info *sbi;
435         struct inode *inode;
436         int rc;
437         s64 newLVSize = 0;
438         int flag, ret = -EINVAL;
439
440         jfs_info("In jfs_read_super: s_flags=0x%lx", sb->s_flags);
441
442         if (!new_valid_dev(sb->s_bdev->bd_dev))
443                 return -EOVERFLOW;
444
445         sbi = kzalloc(sizeof (struct jfs_sb_info), GFP_KERNEL);
446         if (!sbi)
447                 return -ENOMEM;
448
449         sb->s_fs_info = sbi;
450         sb->s_max_links = JFS_LINK_MAX;
451         sbi->sb = sb;
452         sbi->uid = INVALID_UID;
453         sbi->gid = INVALID_GID;
454         sbi->umask = -1;
455
456         /* initialize the mount flag and determine the default error handler */
457         flag = JFS_ERR_REMOUNT_RO;
458
459         if (!parse_options((char *) data, sb, &newLVSize, &flag))
460                 goto out_kfree;
461         sbi->flag = flag;
462
463 #ifdef CONFIG_JFS_POSIX_ACL
464         sb->s_flags |= MS_POSIXACL;
465 #endif
466
467         if (newLVSize) {
468                 printk(KERN_ERR "resize option for remount only\n");
469                 goto out_kfree;
470         }
471
472         /*
473          * Initialize blocksize to 4K.
474          */
475         sb_set_blocksize(sb, PSIZE);
476
477         /*
478          * Set method vectors.
479          */
480         sb->s_op = &jfs_super_operations;
481         sb->s_export_op = &jfs_export_operations;
482 #ifdef CONFIG_QUOTA
483         sb->dq_op = &dquot_operations;
484         sb->s_qcop = &dquot_quotactl_ops;
485 #endif
486
487         /*
488          * Initialize direct-mapping inode/address-space
489          */
490         inode = new_inode(sb);
491         if (inode == NULL) {
492                 ret = -ENOMEM;
493                 goto out_unload;
494         }
495         inode->i_ino = 0;
496         inode->i_size = sb->s_bdev->bd_inode->i_size;
497         inode->i_mapping->a_ops = &jfs_metapage_aops;
498         insert_inode_hash(inode);
499         mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
500
501         sbi->direct_inode = inode;
502
503         rc = jfs_mount(sb);
504         if (rc) {
505                 if (!silent) {
506                         jfs_err("jfs_mount failed w/return code = %d", rc);
507                 }
508                 goto out_mount_failed;
509         }
510         if (sb->s_flags & MS_RDONLY)
511                 sbi->log = NULL;
512         else {
513                 rc = jfs_mount_rw(sb, 0);
514                 if (rc) {
515                         if (!silent) {
516                                 jfs_err("jfs_mount_rw failed, return code = %d",
517                                         rc);
518                         }
519                         goto out_no_rw;
520                 }
521         }
522
523         sb->s_magic = JFS_SUPER_MAGIC;
524
525         if (sbi->mntflag & JFS_OS2)
526                 sb->s_d_op = &jfs_ci_dentry_operations;
527
528         inode = jfs_iget(sb, ROOT_I);
529         if (IS_ERR(inode)) {
530                 ret = PTR_ERR(inode);
531                 goto out_no_rw;
532         }
533         sb->s_root = d_make_root(inode);
534         if (!sb->s_root)
535                 goto out_no_root;
536
537         /* logical blocks are represented by 40 bits in pxd_t, etc. */
538         sb->s_maxbytes = ((u64) sb->s_blocksize) << 40;
539 #if BITS_PER_LONG == 32
540         /*
541          * Page cache is indexed by long.
542          * I would use MAX_LFS_FILESIZE, but it's only half as big
543          */
544         sb->s_maxbytes = min(((u64) PAGE_CACHE_SIZE << 32) - 1, (u64)sb->s_maxbytes);
545 #endif
546         sb->s_time_gran = 1;
547         return 0;
548
549 out_no_root:
550         jfs_err("jfs_read_super: get root dentry failed");
551
552 out_no_rw:
553         rc = jfs_umount(sb);
554         if (rc) {
555                 jfs_err("jfs_umount failed with return code %d", rc);
556         }
557 out_mount_failed:
558         filemap_write_and_wait(sbi->direct_inode->i_mapping);
559         truncate_inode_pages(sbi->direct_inode->i_mapping, 0);
560         make_bad_inode(sbi->direct_inode);
561         iput(sbi->direct_inode);
562         sbi->direct_inode = NULL;
563 out_unload:
564         if (sbi->nls_tab)
565                 unload_nls(sbi->nls_tab);
566 out_kfree:
567         kfree(sbi);
568         return ret;
569 }
570
571 static int jfs_freeze(struct super_block *sb)
572 {
573         struct jfs_sb_info *sbi = JFS_SBI(sb);
574         struct jfs_log *log = sbi->log;
575
576         if (!(sb->s_flags & MS_RDONLY)) {
577                 txQuiesce(sb);
578                 lmLogShutdown(log);
579                 updateSuper(sb, FM_CLEAN);
580         }
581         return 0;
582 }
583
584 static int jfs_unfreeze(struct super_block *sb)
585 {
586         struct jfs_sb_info *sbi = JFS_SBI(sb);
587         struct jfs_log *log = sbi->log;
588         int rc = 0;
589
590         if (!(sb->s_flags & MS_RDONLY)) {
591                 updateSuper(sb, FM_MOUNT);
592                 if ((rc = lmLogInit(log)))
593                         jfs_err("jfs_unlock failed with return code %d", rc);
594                 else
595                         txResume(sb);
596         }
597         return 0;
598 }
599
600 static struct dentry *jfs_do_mount(struct file_system_type *fs_type,
601         int flags, const char *dev_name, void *data)
602 {
603         return mount_bdev(fs_type, flags, dev_name, data, jfs_fill_super);
604 }
605
606 static int jfs_sync_fs(struct super_block *sb, int wait)
607 {
608         struct jfs_log *log = JFS_SBI(sb)->log;
609
610         /* log == NULL indicates read-only mount */
611         if (log) {
612                 /*
613                  * Write quota structures to quota file, sync_blockdev() will
614                  * write them to disk later
615                  */
616                 dquot_writeback_dquots(sb, -1);
617                 jfs_flush_journal(log, wait);
618                 jfs_syncpt(log, 0);
619         }
620
621         return 0;
622 }
623
624 static int jfs_show_options(struct seq_file *seq, struct dentry *root)
625 {
626         struct jfs_sb_info *sbi = JFS_SBI(root->d_sb);
627
628         if (uid_valid(sbi->uid))
629                 seq_printf(seq, ",uid=%d", from_kuid(&init_user_ns, sbi->uid));
630         if (gid_valid(sbi->gid))
631                 seq_printf(seq, ",gid=%d", from_kgid(&init_user_ns, sbi->gid));
632         if (sbi->umask != -1)
633                 seq_printf(seq, ",umask=%03o", sbi->umask);
634         if (sbi->flag & JFS_NOINTEGRITY)
635                 seq_puts(seq, ",nointegrity");
636         if (sbi->nls_tab)
637                 seq_printf(seq, ",iocharset=%s", sbi->nls_tab->charset);
638         if (sbi->flag & JFS_ERR_CONTINUE)
639                 seq_printf(seq, ",errors=continue");
640         if (sbi->flag & JFS_ERR_PANIC)
641                 seq_printf(seq, ",errors=panic");
642
643 #ifdef CONFIG_QUOTA
644         if (sbi->flag & JFS_USRQUOTA)
645                 seq_puts(seq, ",usrquota");
646
647         if (sbi->flag & JFS_GRPQUOTA)
648                 seq_puts(seq, ",grpquota");
649 #endif
650
651         return 0;
652 }
653
654 #ifdef CONFIG_QUOTA
655
656 /* Read data from quotafile - avoid pagecache and such because we cannot afford
657  * acquiring the locks... As quota files are never truncated and quota code
658  * itself serializes the operations (and no one else should touch the files)
659  * we don't have to be afraid of races */
660 static ssize_t jfs_quota_read(struct super_block *sb, int type, char *data,
661                               size_t len, loff_t off)
662 {
663         struct inode *inode = sb_dqopt(sb)->files[type];
664         sector_t blk = off >> sb->s_blocksize_bits;
665         int err = 0;
666         int offset = off & (sb->s_blocksize - 1);
667         int tocopy;
668         size_t toread;
669         struct buffer_head tmp_bh;
670         struct buffer_head *bh;
671         loff_t i_size = i_size_read(inode);
672
673         if (off > i_size)
674                 return 0;
675         if (off+len > i_size)
676                 len = i_size-off;
677         toread = len;
678         while (toread > 0) {
679                 tocopy = sb->s_blocksize - offset < toread ?
680                                 sb->s_blocksize - offset : toread;
681
682                 tmp_bh.b_state = 0;
683                 tmp_bh.b_size = 1 << inode->i_blkbits;
684                 err = jfs_get_block(inode, blk, &tmp_bh, 0);
685                 if (err)
686                         return err;
687                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
688                         memset(data, 0, tocopy);
689                 else {
690                         bh = sb_bread(sb, tmp_bh.b_blocknr);
691                         if (!bh)
692                                 return -EIO;
693                         memcpy(data, bh->b_data+offset, tocopy);
694                         brelse(bh);
695                 }
696                 offset = 0;
697                 toread -= tocopy;
698                 data += tocopy;
699                 blk++;
700         }
701         return len;
702 }
703
704 /* Write to quotafile */
705 static ssize_t jfs_quota_write(struct super_block *sb, int type,
706                                const char *data, size_t len, loff_t off)
707 {
708         struct inode *inode = sb_dqopt(sb)->files[type];
709         sector_t blk = off >> sb->s_blocksize_bits;
710         int err = 0;
711         int offset = off & (sb->s_blocksize - 1);
712         int tocopy;
713         size_t towrite = len;
714         struct buffer_head tmp_bh;
715         struct buffer_head *bh;
716
717         mutex_lock(&inode->i_mutex);
718         while (towrite > 0) {
719                 tocopy = sb->s_blocksize - offset < towrite ?
720                                 sb->s_blocksize - offset : towrite;
721
722                 tmp_bh.b_state = 0;
723                 tmp_bh.b_size = 1 << inode->i_blkbits;
724                 err = jfs_get_block(inode, blk, &tmp_bh, 1);
725                 if (err)
726                         goto out;
727                 if (offset || tocopy != sb->s_blocksize)
728                         bh = sb_bread(sb, tmp_bh.b_blocknr);
729                 else
730                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
731                 if (!bh) {
732                         err = -EIO;
733                         goto out;
734                 }
735                 lock_buffer(bh);
736                 memcpy(bh->b_data+offset, data, tocopy);
737                 flush_dcache_page(bh->b_page);
738                 set_buffer_uptodate(bh);
739                 mark_buffer_dirty(bh);
740                 unlock_buffer(bh);
741                 brelse(bh);
742                 offset = 0;
743                 towrite -= tocopy;
744                 data += tocopy;
745                 blk++;
746         }
747 out:
748         if (len == towrite) {
749                 mutex_unlock(&inode->i_mutex);
750                 return err;
751         }
752         if (inode->i_size < off+len-towrite)
753                 i_size_write(inode, off+len-towrite);
754         inode->i_version++;
755         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
756         mark_inode_dirty(inode);
757         mutex_unlock(&inode->i_mutex);
758         return len - towrite;
759 }
760
761 #endif
762
763 static const struct super_operations jfs_super_operations = {
764         .alloc_inode    = jfs_alloc_inode,
765         .destroy_inode  = jfs_destroy_inode,
766         .dirty_inode    = jfs_dirty_inode,
767         .write_inode    = jfs_write_inode,
768         .evict_inode    = jfs_evict_inode,
769         .put_super      = jfs_put_super,
770         .sync_fs        = jfs_sync_fs,
771         .freeze_fs      = jfs_freeze,
772         .unfreeze_fs    = jfs_unfreeze,
773         .statfs         = jfs_statfs,
774         .remount_fs     = jfs_remount,
775         .show_options   = jfs_show_options,
776 #ifdef CONFIG_QUOTA
777         .quota_read     = jfs_quota_read,
778         .quota_write    = jfs_quota_write,
779 #endif
780 };
781
782 static const struct export_operations jfs_export_operations = {
783         .fh_to_dentry   = jfs_fh_to_dentry,
784         .fh_to_parent   = jfs_fh_to_parent,
785         .get_parent     = jfs_get_parent,
786 };
787
788 static struct file_system_type jfs_fs_type = {
789         .owner          = THIS_MODULE,
790         .name           = "jfs",
791         .mount          = jfs_do_mount,
792         .kill_sb        = kill_block_super,
793         .fs_flags       = FS_REQUIRES_DEV,
794 };
795
796 static void init_once(void *foo)
797 {
798         struct jfs_inode_info *jfs_ip = (struct jfs_inode_info *) foo;
799
800         memset(jfs_ip, 0, sizeof(struct jfs_inode_info));
801         INIT_LIST_HEAD(&jfs_ip->anon_inode_list);
802         init_rwsem(&jfs_ip->rdwrlock);
803         mutex_init(&jfs_ip->commit_mutex);
804         init_rwsem(&jfs_ip->xattr_sem);
805         spin_lock_init(&jfs_ip->ag_lock);
806         jfs_ip->active_ag = -1;
807         inode_init_once(&jfs_ip->vfs_inode);
808 }
809
810 static int __init init_jfs_fs(void)
811 {
812         int i;
813         int rc;
814
815         jfs_inode_cachep =
816             kmem_cache_create("jfs_ip", sizeof(struct jfs_inode_info), 0,
817                             SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD,
818                             init_once);
819         if (jfs_inode_cachep == NULL)
820                 return -ENOMEM;
821
822         /*
823          * Metapage initialization
824          */
825         rc = metapage_init();
826         if (rc) {
827                 jfs_err("metapage_init failed w/rc = %d", rc);
828                 goto free_slab;
829         }
830
831         /*
832          * Transaction Manager initialization
833          */
834         rc = txInit();
835         if (rc) {
836                 jfs_err("txInit failed w/rc = %d", rc);
837                 goto free_metapage;
838         }
839
840         /*
841          * I/O completion thread (endio)
842          */
843         jfsIOthread = kthread_run(jfsIOWait, NULL, "jfsIO");
844         if (IS_ERR(jfsIOthread)) {
845                 rc = PTR_ERR(jfsIOthread);
846                 jfs_err("init_jfs_fs: fork failed w/rc = %d", rc);
847                 goto end_txmngr;
848         }
849
850         if (commit_threads < 1)
851                 commit_threads = num_online_cpus();
852         if (commit_threads > MAX_COMMIT_THREADS)
853                 commit_threads = MAX_COMMIT_THREADS;
854
855         for (i = 0; i < commit_threads; i++) {
856                 jfsCommitThread[i] = kthread_run(jfs_lazycommit, NULL, "jfsCommit");
857                 if (IS_ERR(jfsCommitThread[i])) {
858                         rc = PTR_ERR(jfsCommitThread[i]);
859                         jfs_err("init_jfs_fs: fork failed w/rc = %d", rc);
860                         commit_threads = i;
861                         goto kill_committask;
862                 }
863         }
864
865         jfsSyncThread = kthread_run(jfs_sync, NULL, "jfsSync");
866         if (IS_ERR(jfsSyncThread)) {
867                 rc = PTR_ERR(jfsSyncThread);
868                 jfs_err("init_jfs_fs: fork failed w/rc = %d", rc);
869                 goto kill_committask;
870         }
871
872 #ifdef PROC_FS_JFS
873         jfs_proc_init();
874 #endif
875
876         rc = register_filesystem(&jfs_fs_type);
877         if (!rc)
878                 return 0;
879
880 #ifdef PROC_FS_JFS
881         jfs_proc_clean();
882 #endif
883         kthread_stop(jfsSyncThread);
884 kill_committask:
885         for (i = 0; i < commit_threads; i++)
886                 kthread_stop(jfsCommitThread[i]);
887         kthread_stop(jfsIOthread);
888 end_txmngr:
889         txExit();
890 free_metapage:
891         metapage_exit();
892 free_slab:
893         kmem_cache_destroy(jfs_inode_cachep);
894         return rc;
895 }
896
897 static void __exit exit_jfs_fs(void)
898 {
899         int i;
900
901         jfs_info("exit_jfs_fs called");
902
903         txExit();
904         metapage_exit();
905
906         kthread_stop(jfsIOthread);
907         for (i = 0; i < commit_threads; i++)
908                 kthread_stop(jfsCommitThread[i]);
909         kthread_stop(jfsSyncThread);
910 #ifdef PROC_FS_JFS
911         jfs_proc_clean();
912 #endif
913         unregister_filesystem(&jfs_fs_type);
914
915         /*
916          * Make sure all delayed rcu free inodes are flushed before we
917          * destroy cache.
918          */
919         rcu_barrier();
920         kmem_cache_destroy(jfs_inode_cachep);
921 }
922
923 module_init(init_jfs_fs)
924 module_exit(exit_jfs_fs)