2 * Copyright (C) International Business Machines Corp., 2000-2004
3 * Portions Copyright (C) Christoph Hellwig, 2001-2002
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.
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.
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
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 #include <linux/smp_lock.h>
38 #include "jfs_incore.h"
39 #include "jfs_filsys.h"
40 #include "jfs_inode.h"
41 #include "jfs_metapage.h"
42 #include "jfs_superblock.h"
46 #include "jfs_debug.h"
48 MODULE_DESCRIPTION("The Journaled Filesystem (JFS)");
49 MODULE_AUTHOR("Steve Best/Dave Kleikamp/Barry Arndt, IBM");
50 MODULE_LICENSE("GPL");
52 static struct kmem_cache * jfs_inode_cachep;
54 static const struct super_operations jfs_super_operations;
55 static const struct export_operations jfs_export_operations;
56 static struct file_system_type jfs_fs_type;
58 #define MAX_COMMIT_THREADS 64
59 static int commit_threads = 0;
60 module_param(commit_threads, int, 0);
61 MODULE_PARM_DESC(commit_threads, "Number of commit threads");
63 static struct task_struct *jfsCommitThread[MAX_COMMIT_THREADS];
64 struct task_struct *jfsIOthread;
65 struct task_struct *jfsSyncThread;
67 #ifdef CONFIG_JFS_DEBUG
68 int jfsloglevel = JFS_LOGLEVEL_WARN;
69 module_param(jfsloglevel, int, 0644);
70 MODULE_PARM_DESC(jfsloglevel, "Specify JFS loglevel (0, 1 or 2)");
73 static void jfs_handle_error(struct super_block *sb)
75 struct jfs_sb_info *sbi = JFS_SBI(sb);
77 if (sb->s_flags & MS_RDONLY)
80 updateSuper(sb, FM_DIRTY);
82 if (sbi->flag & JFS_ERR_PANIC)
83 panic("JFS (device %s): panic forced after error\n",
85 else if (sbi->flag & JFS_ERR_REMOUNT_RO) {
86 jfs_err("ERROR: (device %s): remounting filesystem "
89 sb->s_flags |= MS_RDONLY;
92 /* nothing is done for continue beyond marking the superblock dirty */
95 void jfs_error(struct super_block *sb, const char * function, ...)
97 static char error_buf[256];
100 va_start(args, function);
101 vsnprintf(error_buf, sizeof(error_buf), function, args);
104 printk(KERN_ERR "ERROR: (device %s): %s\n", sb->s_id, error_buf);
106 jfs_handle_error(sb);
109 static struct inode *jfs_alloc_inode(struct super_block *sb)
111 struct jfs_inode_info *jfs_inode;
113 jfs_inode = kmem_cache_alloc(jfs_inode_cachep, GFP_NOFS);
116 return &jfs_inode->vfs_inode;
119 static void jfs_destroy_inode(struct inode *inode)
121 struct jfs_inode_info *ji = JFS_IP(inode);
123 BUG_ON(!list_empty(&ji->anon_inode_list));
125 spin_lock_irq(&ji->ag_lock);
126 if (ji->active_ag != -1) {
127 struct bmap *bmap = JFS_SBI(inode->i_sb)->bmap;
128 atomic_dec(&bmap->db_active[ji->active_ag]);
131 spin_unlock_irq(&ji->ag_lock);
132 kmem_cache_free(jfs_inode_cachep, ji);
135 static void jfs_clear_inode(struct inode *inode)
140 static int jfs_statfs(struct dentry *dentry, struct kstatfs *buf)
142 struct jfs_sb_info *sbi = JFS_SBI(dentry->d_sb);
144 struct inomap *imap = JFS_IP(sbi->ipimap)->i_imap;
146 jfs_info("In jfs_statfs");
147 buf->f_type = JFS_SUPER_MAGIC;
148 buf->f_bsize = sbi->bsize;
149 buf->f_blocks = sbi->bmap->db_mapsize;
150 buf->f_bfree = sbi->bmap->db_nfree;
151 buf->f_bavail = sbi->bmap->db_nfree;
153 * If we really return the number of allocated & free inodes, some
154 * applications will fail because they won't see enough free inodes.
155 * We'll try to calculate some guess as to how may inodes we can
158 * buf->f_files = atomic_read(&imap->im_numinos);
159 * buf->f_ffree = atomic_read(&imap->im_numfree);
161 maxinodes = min((s64) atomic_read(&imap->im_numinos) +
162 ((sbi->bmap->db_nfree >> imap->im_l2nbperiext)
163 << L2INOSPEREXT), (s64) 0xffffffffLL);
164 buf->f_files = maxinodes;
165 buf->f_ffree = maxinodes - (atomic_read(&imap->im_numinos) -
166 atomic_read(&imap->im_numfree));
167 buf->f_fsid.val[0] = (u32)crc32_le(0, sbi->uuid, sizeof(sbi->uuid)/2);
168 buf->f_fsid.val[1] = (u32)crc32_le(0, sbi->uuid + sizeof(sbi->uuid)/2,
169 sizeof(sbi->uuid)/2);
171 buf->f_namelen = JFS_NAME_MAX;
175 static void jfs_put_super(struct super_block *sb)
177 struct jfs_sb_info *sbi = JFS_SBI(sb);
180 jfs_info("In jfs_put_super");
182 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
188 jfs_err("jfs_umount failed with return code %d", rc);
190 unload_nls(sbi->nls_tab);
192 truncate_inode_pages(sbi->direct_inode->i_mapping, 0);
193 iput(sbi->direct_inode);
201 Opt_integrity, Opt_nointegrity, Opt_iocharset, Opt_resize,
202 Opt_resize_nosize, Opt_errors, Opt_ignore, Opt_err, Opt_quota,
203 Opt_usrquota, Opt_grpquota, Opt_uid, Opt_gid, Opt_umask
206 static const match_table_t tokens = {
207 {Opt_integrity, "integrity"},
208 {Opt_nointegrity, "nointegrity"},
209 {Opt_iocharset, "iocharset=%s"},
210 {Opt_resize, "resize=%u"},
211 {Opt_resize_nosize, "resize"},
212 {Opt_errors, "errors=%s"},
213 {Opt_ignore, "noquota"},
214 {Opt_ignore, "quota"},
215 {Opt_usrquota, "usrquota"},
216 {Opt_grpquota, "grpquota"},
219 {Opt_umask, "umask=%u"},
223 static int parse_options(char *options, struct super_block *sb, s64 *newLVSize,
226 void *nls_map = (void *)-1; /* -1: no change; NULL: none */
228 struct jfs_sb_info *sbi = JFS_SBI(sb);
235 while ((p = strsep(&options, ",")) != NULL) {
236 substring_t args[MAX_OPT_ARGS];
241 token = match_token(p, tokens, args);
244 *flag &= ~JFS_NOINTEGRITY;
246 case Opt_nointegrity:
247 *flag |= JFS_NOINTEGRITY;
250 /* Silently ignore the quota options */
251 /* Don't do anything ;-) */
254 if (nls_map && nls_map != (void *) -1)
256 if (!strcmp(args[0].from, "none"))
259 nls_map = load_nls(args[0].from);
262 "JFS: charset not found\n");
269 char *resize = args[0].from;
270 *newLVSize = simple_strtoull(resize, &resize, 0);
273 case Opt_resize_nosize:
275 *newLVSize = sb->s_bdev->bd_inode->i_size >>
276 sb->s_blocksize_bits;
279 "JFS: Cannot determine volume size\n");
284 char *errors = args[0].from;
285 if (!errors || !*errors)
287 if (!strcmp(errors, "continue")) {
288 *flag &= ~JFS_ERR_REMOUNT_RO;
289 *flag &= ~JFS_ERR_PANIC;
290 *flag |= JFS_ERR_CONTINUE;
291 } else if (!strcmp(errors, "remount-ro")) {
292 *flag &= ~JFS_ERR_CONTINUE;
293 *flag &= ~JFS_ERR_PANIC;
294 *flag |= JFS_ERR_REMOUNT_RO;
295 } else if (!strcmp(errors, "panic")) {
296 *flag &= ~JFS_ERR_CONTINUE;
297 *flag &= ~JFS_ERR_REMOUNT_RO;
298 *flag |= JFS_ERR_PANIC;
301 "JFS: %s is an invalid error handler\n",
311 *flag |= JFS_USRQUOTA;
314 *flag |= JFS_GRPQUOTA;
321 "JFS: quota operations not supported\n");
326 char *uid = args[0].from;
327 sbi->uid = simple_strtoul(uid, &uid, 0);
332 char *gid = args[0].from;
333 sbi->gid = simple_strtoul(gid, &gid, 0);
338 char *umask = args[0].from;
339 sbi->umask = simple_strtoul(umask, &umask, 8);
340 if (sbi->umask & ~0777) {
342 "JFS: Invalid value of umask\n");
348 printk("jfs: Unrecognized mount option \"%s\" "
349 " or missing value\n", p);
354 if (nls_map != (void *) -1) {
355 /* Discard old (if remount) */
356 unload_nls(sbi->nls_tab);
357 sbi->nls_tab = nls_map;
362 if (nls_map && nls_map != (void *) -1)
367 static int jfs_remount(struct super_block *sb, int *flags, char *data)
371 int flag = JFS_SBI(sb)->flag;
374 if (!parse_options(data, sb, &newLVSize, &flag)) {
379 if (sb->s_flags & MS_RDONLY) {
381 "JFS: resize requires volume to be mounted read-write\n");
385 rc = jfs_extendfs(sb, newLVSize, 0);
392 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
394 * Invalidate any previously read metadata. fsck may have
395 * changed the on-disk data since we mounted r/o
397 truncate_inode_pages(JFS_SBI(sb)->direct_inode->i_mapping, 0);
399 JFS_SBI(sb)->flag = flag;
400 ret = jfs_mount_rw(sb, 1);
402 /* mark the fs r/w for quota activity */
403 sb->s_flags &= ~MS_RDONLY;
406 dquot_resume(sb, -1);
409 if ((!(sb->s_flags & MS_RDONLY)) && (*flags & MS_RDONLY)) {
410 rc = dquot_suspend(sb, -1);
415 rc = jfs_umount_rw(sb);
416 JFS_SBI(sb)->flag = flag;
420 if ((JFS_SBI(sb)->flag & JFS_NOINTEGRITY) != (flag & JFS_NOINTEGRITY))
421 if (!(sb->s_flags & MS_RDONLY)) {
422 rc = jfs_umount_rw(sb);
427 JFS_SBI(sb)->flag = flag;
428 ret = jfs_mount_rw(sb, 1);
432 JFS_SBI(sb)->flag = flag;
438 static int jfs_fill_super(struct super_block *sb, void *data, int silent)
440 struct jfs_sb_info *sbi;
444 int flag, ret = -EINVAL;
446 jfs_info("In jfs_read_super: s_flags=0x%lx", sb->s_flags);
448 if (!new_valid_dev(sb->s_bdev->bd_dev))
451 sbi = kzalloc(sizeof (struct jfs_sb_info), GFP_KERNEL);
456 sbi->uid = sbi->gid = sbi->umask = -1;
458 /* initialize the mount flag and determine the default error handler */
459 flag = JFS_ERR_REMOUNT_RO;
461 if (!parse_options((char *) data, sb, &newLVSize, &flag))
465 #ifdef CONFIG_JFS_POSIX_ACL
466 sb->s_flags |= MS_POSIXACL;
470 printk(KERN_ERR "resize option for remount only\n");
475 * Initialize blocksize to 4K.
477 sb_set_blocksize(sb, PSIZE);
480 * Set method vectors.
482 sb->s_op = &jfs_super_operations;
483 sb->s_export_op = &jfs_export_operations;
485 sb->dq_op = &dquot_operations;
486 sb->s_qcop = &dquot_quotactl_ops;
490 * Initialize direct-mapping inode/address-space
492 inode = new_inode(sb);
499 inode->i_size = sb->s_bdev->bd_inode->i_size;
500 inode->i_mapping->a_ops = &jfs_metapage_aops;
501 insert_inode_hash(inode);
502 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
504 sbi->direct_inode = inode;
509 jfs_err("jfs_mount failed w/return code = %d", rc);
511 goto out_mount_failed;
513 if (sb->s_flags & MS_RDONLY)
516 rc = jfs_mount_rw(sb, 0);
519 jfs_err("jfs_mount_rw failed, return code = %d",
526 sb->s_magic = JFS_SUPER_MAGIC;
528 inode = jfs_iget(sb, ROOT_I);
530 ret = PTR_ERR(inode);
533 sb->s_root = d_alloc_root(inode);
537 if (sbi->mntflag & JFS_OS2)
538 sb->s_root->d_op = &jfs_ci_dentry_operations;
540 /* logical blocks are represented by 40 bits in pxd_t, etc. */
541 sb->s_maxbytes = ((u64) sb->s_blocksize) << 40;
542 #if BITS_PER_LONG == 32
544 * Page cache is indexed by long.
545 * I would use MAX_LFS_FILESIZE, but it's only half as big
547 sb->s_maxbytes = min(((u64) PAGE_CACHE_SIZE << 32) - 1, (u64)sb->s_maxbytes);
553 jfs_err("jfs_read_super: get root dentry failed");
559 jfs_err("jfs_umount failed with return code %d", rc);
562 filemap_write_and_wait(sbi->direct_inode->i_mapping);
563 truncate_inode_pages(sbi->direct_inode->i_mapping, 0);
564 make_bad_inode(sbi->direct_inode);
565 iput(sbi->direct_inode);
566 sbi->direct_inode = NULL;
569 unload_nls(sbi->nls_tab);
575 static int jfs_freeze(struct super_block *sb)
577 struct jfs_sb_info *sbi = JFS_SBI(sb);
578 struct jfs_log *log = sbi->log;
580 if (!(sb->s_flags & MS_RDONLY)) {
583 updateSuper(sb, FM_CLEAN);
588 static int jfs_unfreeze(struct super_block *sb)
590 struct jfs_sb_info *sbi = JFS_SBI(sb);
591 struct jfs_log *log = sbi->log;
594 if (!(sb->s_flags & MS_RDONLY)) {
595 updateSuper(sb, FM_MOUNT);
596 if ((rc = lmLogInit(log)))
597 jfs_err("jfs_unlock failed with return code %d", rc);
604 static int jfs_get_sb(struct file_system_type *fs_type,
605 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
607 return get_sb_bdev(fs_type, flags, dev_name, data, jfs_fill_super,
611 static int jfs_sync_fs(struct super_block *sb, int wait)
613 struct jfs_log *log = JFS_SBI(sb)->log;
615 /* log == NULL indicates read-only mount */
617 jfs_flush_journal(log, wait);
624 static int jfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
626 struct jfs_sb_info *sbi = JFS_SBI(vfs->mnt_sb);
629 seq_printf(seq, ",uid=%d", sbi->uid);
631 seq_printf(seq, ",gid=%d", 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");
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");
644 if (sbi->flag & JFS_USRQUOTA)
645 seq_puts(seq, ",usrquota");
647 if (sbi->flag & JFS_GRPQUOTA)
648 seq_puts(seq, ",grpquota");
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 noone 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)
663 struct inode *inode = sb_dqopt(sb)->files[type];
664 sector_t blk = off >> sb->s_blocksize_bits;
666 int offset = off & (sb->s_blocksize - 1);
669 struct buffer_head tmp_bh;
670 struct buffer_head *bh;
671 loff_t i_size = i_size_read(inode);
675 if (off+len > i_size)
679 tocopy = sb->s_blocksize - offset < toread ?
680 sb->s_blocksize - offset : toread;
683 tmp_bh.b_size = 1 << inode->i_blkbits;
684 err = jfs_get_block(inode, blk, &tmp_bh, 0);
687 if (!buffer_mapped(&tmp_bh)) /* A hole? */
688 memset(data, 0, tocopy);
690 bh = sb_bread(sb, tmp_bh.b_blocknr);
693 memcpy(data, bh->b_data+offset, tocopy);
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)
708 struct inode *inode = sb_dqopt(sb)->files[type];
709 sector_t blk = off >> sb->s_blocksize_bits;
711 int offset = off & (sb->s_blocksize - 1);
713 size_t towrite = len;
714 struct buffer_head tmp_bh;
715 struct buffer_head *bh;
717 mutex_lock(&inode->i_mutex);
718 while (towrite > 0) {
719 tocopy = sb->s_blocksize - offset < towrite ?
720 sb->s_blocksize - offset : towrite;
723 tmp_bh.b_size = 1 << inode->i_blkbits;
724 err = jfs_get_block(inode, blk, &tmp_bh, 1);
727 if (offset || tocopy != sb->s_blocksize)
728 bh = sb_bread(sb, tmp_bh.b_blocknr);
730 bh = sb_getblk(sb, tmp_bh.b_blocknr);
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);
748 if (len == towrite) {
749 mutex_unlock(&inode->i_mutex);
752 if (inode->i_size < off+len-towrite)
753 i_size_write(inode, off+len-towrite);
755 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
756 mark_inode_dirty(inode);
757 mutex_unlock(&inode->i_mutex);
758 return len - towrite;
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 .delete_inode = jfs_delete_inode,
769 .clear_inode = jfs_clear_inode,
770 .put_super = jfs_put_super,
771 .sync_fs = jfs_sync_fs,
772 .freeze_fs = jfs_freeze,
773 .unfreeze_fs = jfs_unfreeze,
774 .statfs = jfs_statfs,
775 .remount_fs = jfs_remount,
776 .show_options = jfs_show_options,
778 .quota_read = jfs_quota_read,
779 .quota_write = jfs_quota_write,
783 static const struct export_operations jfs_export_operations = {
784 .fh_to_dentry = jfs_fh_to_dentry,
785 .fh_to_parent = jfs_fh_to_parent,
786 .get_parent = jfs_get_parent,
789 static struct file_system_type jfs_fs_type = {
790 .owner = THIS_MODULE,
792 .get_sb = jfs_get_sb,
793 .kill_sb = kill_block_super,
794 .fs_flags = FS_REQUIRES_DEV,
797 static void init_once(void *foo)
799 struct jfs_inode_info *jfs_ip = (struct jfs_inode_info *) foo;
801 memset(jfs_ip, 0, sizeof(struct jfs_inode_info));
802 INIT_LIST_HEAD(&jfs_ip->anon_inode_list);
803 init_rwsem(&jfs_ip->rdwrlock);
804 mutex_init(&jfs_ip->commit_mutex);
805 init_rwsem(&jfs_ip->xattr_sem);
806 spin_lock_init(&jfs_ip->ag_lock);
807 jfs_ip->active_ag = -1;
808 inode_init_once(&jfs_ip->vfs_inode);
811 static int __init init_jfs_fs(void)
817 kmem_cache_create("jfs_ip", sizeof(struct jfs_inode_info), 0,
818 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD,
820 if (jfs_inode_cachep == NULL)
824 * Metapage initialization
826 rc = metapage_init();
828 jfs_err("metapage_init failed w/rc = %d", rc);
833 * Transaction Manager initialization
837 jfs_err("txInit failed w/rc = %d", rc);
842 * I/O completion thread (endio)
844 jfsIOthread = kthread_run(jfsIOWait, NULL, "jfsIO");
845 if (IS_ERR(jfsIOthread)) {
846 rc = PTR_ERR(jfsIOthread);
847 jfs_err("init_jfs_fs: fork failed w/rc = %d", rc);
851 if (commit_threads < 1)
852 commit_threads = num_online_cpus();
853 if (commit_threads > MAX_COMMIT_THREADS)
854 commit_threads = MAX_COMMIT_THREADS;
856 for (i = 0; i < commit_threads; i++) {
857 jfsCommitThread[i] = kthread_run(jfs_lazycommit, NULL, "jfsCommit");
858 if (IS_ERR(jfsCommitThread[i])) {
859 rc = PTR_ERR(jfsCommitThread[i]);
860 jfs_err("init_jfs_fs: fork failed w/rc = %d", rc);
862 goto kill_committask;
866 jfsSyncThread = kthread_run(jfs_sync, NULL, "jfsSync");
867 if (IS_ERR(jfsSyncThread)) {
868 rc = PTR_ERR(jfsSyncThread);
869 jfs_err("init_jfs_fs: fork failed w/rc = %d", rc);
870 goto kill_committask;
877 return register_filesystem(&jfs_fs_type);
880 for (i = 0; i < commit_threads; i++)
881 kthread_stop(jfsCommitThread[i]);
882 kthread_stop(jfsIOthread);
888 kmem_cache_destroy(jfs_inode_cachep);
892 static void __exit exit_jfs_fs(void)
896 jfs_info("exit_jfs_fs called");
901 kthread_stop(jfsIOthread);
902 for (i = 0; i < commit_threads; i++)
903 kthread_stop(jfsCommitThread[i]);
904 kthread_stop(jfsSyncThread);
908 unregister_filesystem(&jfs_fs_type);
909 kmem_cache_destroy(jfs_inode_cachep);
912 module_init(init_jfs_fs)
913 module_exit(exit_jfs_fs)