ocfs2: Add xattr bucket iteration for large numbers of EAs
[linux-2.6] / fs / bfs / inode.c
1 /*
2  *      fs/bfs/inode.c
3  *      BFS superblock and inode operations.
4  *      Copyright (C) 1999-2006 Tigran Aivazian <tigran@aivazian.fsnet.co.uk>
5  *      From fs/minix, Copyright (C) 1991, 1992 Linus Torvalds.
6  *
7  *      Made endianness-clean by Andrew Stribblehill <ads@wompom.org>, 2005.
8  */
9
10 #include <linux/module.h>
11 #include <linux/mm.h>
12 #include <linux/slab.h>
13 #include <linux/init.h>
14 #include <linux/fs.h>
15 #include <linux/smp_lock.h>
16 #include <linux/buffer_head.h>
17 #include <linux/vfs.h>
18 #include <asm/uaccess.h>
19 #include "bfs.h"
20
21 MODULE_AUTHOR("Tigran Aivazian <tigran@aivazian.fsnet.co.uk>");
22 MODULE_DESCRIPTION("SCO UnixWare BFS filesystem for Linux");
23 MODULE_LICENSE("GPL");
24
25 #undef DEBUG
26
27 #ifdef DEBUG
28 #define dprintf(x...)   printf(x)
29 #else
30 #define dprintf(x...)
31 #endif
32
33 void dump_imap(const char *prefix, struct super_block *s);
34
35 struct inode *bfs_iget(struct super_block *sb, unsigned long ino)
36 {
37         struct bfs_inode *di;
38         struct inode *inode;
39         struct buffer_head *bh;
40         int block, off;
41
42         inode = iget_locked(sb, ino);
43         if (IS_ERR(inode))
44                 return ERR_PTR(-ENOMEM);
45         if (!(inode->i_state & I_NEW))
46                 return inode;
47
48         if ((ino < BFS_ROOT_INO) || (ino > BFS_SB(inode->i_sb)->si_lasti)) {
49                 printf("Bad inode number %s:%08lx\n", inode->i_sb->s_id, ino);
50                 goto error;
51         }
52
53         block = (ino - BFS_ROOT_INO) / BFS_INODES_PER_BLOCK + 1;
54         bh = sb_bread(inode->i_sb, block);
55         if (!bh) {
56                 printf("Unable to read inode %s:%08lx\n", inode->i_sb->s_id,
57                                                                         ino);
58                 goto error;
59         }
60
61         off = (ino - BFS_ROOT_INO) % BFS_INODES_PER_BLOCK;
62         di = (struct bfs_inode *)bh->b_data + off;
63
64         inode->i_mode = 0x0000FFFF & le32_to_cpu(di->i_mode);
65         if (le32_to_cpu(di->i_vtype) == BFS_VDIR) {
66                 inode->i_mode |= S_IFDIR;
67                 inode->i_op = &bfs_dir_inops;
68                 inode->i_fop = &bfs_dir_operations;
69         } else if (le32_to_cpu(di->i_vtype) == BFS_VREG) {
70                 inode->i_mode |= S_IFREG;
71                 inode->i_op = &bfs_file_inops;
72                 inode->i_fop = &bfs_file_operations;
73                 inode->i_mapping->a_ops = &bfs_aops;
74         }
75
76         BFS_I(inode)->i_sblock =  le32_to_cpu(di->i_sblock);
77         BFS_I(inode)->i_eblock =  le32_to_cpu(di->i_eblock);
78         BFS_I(inode)->i_dsk_ino = le16_to_cpu(di->i_ino);
79         inode->i_uid =  le32_to_cpu(di->i_uid);
80         inode->i_gid =  le32_to_cpu(di->i_gid);
81         inode->i_nlink =  le32_to_cpu(di->i_nlink);
82         inode->i_size = BFS_FILESIZE(di);
83         inode->i_blocks = BFS_FILEBLOCKS(di);
84         inode->i_atime.tv_sec =  le32_to_cpu(di->i_atime);
85         inode->i_mtime.tv_sec =  le32_to_cpu(di->i_mtime);
86         inode->i_ctime.tv_sec =  le32_to_cpu(di->i_ctime);
87         inode->i_atime.tv_nsec = 0;
88         inode->i_mtime.tv_nsec = 0;
89         inode->i_ctime.tv_nsec = 0;
90
91         brelse(bh);
92         unlock_new_inode(inode);
93         return inode;
94
95 error:
96         iget_failed(inode);
97         return ERR_PTR(-EIO);
98 }
99
100 static int bfs_write_inode(struct inode *inode, int unused)
101 {
102         unsigned int ino = (u16)inode->i_ino;
103         unsigned long i_sblock;
104         struct bfs_inode *di;
105         struct buffer_head *bh;
106         int block, off;
107         struct bfs_sb_info *info = BFS_SB(inode->i_sb);
108
109         dprintf("ino=%08x\n", ino);
110
111         if ((ino < BFS_ROOT_INO) || (ino > BFS_SB(inode->i_sb)->si_lasti)) {
112                 printf("Bad inode number %s:%08x\n", inode->i_sb->s_id, ino);
113                 return -EIO;
114         }
115
116         mutex_lock(&info->bfs_lock);
117         block = (ino - BFS_ROOT_INO) / BFS_INODES_PER_BLOCK + 1;
118         bh = sb_bread(inode->i_sb, block);
119         if (!bh) {
120                 printf("Unable to read inode %s:%08x\n",
121                                 inode->i_sb->s_id, ino);
122                 mutex_unlock(&info->bfs_lock);
123                 return -EIO;
124         }
125
126         off = (ino - BFS_ROOT_INO) % BFS_INODES_PER_BLOCK;
127         di = (struct bfs_inode *)bh->b_data + off;
128
129         if (ino == BFS_ROOT_INO)
130                 di->i_vtype = cpu_to_le32(BFS_VDIR);
131         else
132                 di->i_vtype = cpu_to_le32(BFS_VREG);
133
134         di->i_ino = cpu_to_le16(ino);
135         di->i_mode = cpu_to_le32(inode->i_mode);
136         di->i_uid = cpu_to_le32(inode->i_uid);
137         di->i_gid = cpu_to_le32(inode->i_gid);
138         di->i_nlink = cpu_to_le32(inode->i_nlink);
139         di->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
140         di->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
141         di->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
142         i_sblock = BFS_I(inode)->i_sblock;
143         di->i_sblock = cpu_to_le32(i_sblock);
144         di->i_eblock = cpu_to_le32(BFS_I(inode)->i_eblock);
145         di->i_eoffset = cpu_to_le32(i_sblock * BFS_BSIZE + inode->i_size - 1);
146
147         mark_buffer_dirty(bh);
148         brelse(bh);
149         mutex_unlock(&info->bfs_lock);
150         return 0;
151 }
152
153 static void bfs_delete_inode(struct inode *inode)
154 {
155         unsigned long ino = inode->i_ino;
156         struct bfs_inode *di;
157         struct buffer_head *bh;
158         int block, off;
159         struct super_block *s = inode->i_sb;
160         struct bfs_sb_info *info = BFS_SB(s);
161         struct bfs_inode_info *bi = BFS_I(inode);
162
163         dprintf("ino=%08lx\n", ino);
164
165         truncate_inode_pages(&inode->i_data, 0);
166
167         if ((ino < BFS_ROOT_INO) || (ino > info->si_lasti)) {
168                 printf("invalid ino=%08lx\n", ino);
169                 return;
170         }
171         
172         inode->i_size = 0;
173         inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
174         mutex_lock(&info->bfs_lock);
175         mark_inode_dirty(inode);
176
177         block = (ino - BFS_ROOT_INO) / BFS_INODES_PER_BLOCK + 1;
178         bh = sb_bread(s, block);
179         if (!bh) {
180                 printf("Unable to read inode %s:%08lx\n",
181                                         inode->i_sb->s_id, ino);
182                 mutex_unlock(&info->bfs_lock);
183                 return;
184         }
185         off = (ino - BFS_ROOT_INO) % BFS_INODES_PER_BLOCK;
186         di = (struct bfs_inode *)bh->b_data + off;
187         memset((void *)di, 0, sizeof(struct bfs_inode));
188         mark_buffer_dirty(bh);
189         brelse(bh);
190
191         if (bi->i_dsk_ino) {
192                 if (bi->i_sblock)
193                         info->si_freeb += bi->i_eblock + 1 - bi->i_sblock;
194                 info->si_freei++;
195                 clear_bit(ino, info->si_imap);
196                 dump_imap("delete_inode", s);
197         }
198
199         /*
200          * If this was the last file, make the previous block
201          * "last block of the last file" even if there is no
202          * real file there, saves us 1 gap.
203          */
204         if (info->si_lf_eblk == bi->i_eblock) {
205                 info->si_lf_eblk = bi->i_sblock - 1;
206                 mark_buffer_dirty(info->si_sbh);
207         }
208         mutex_unlock(&info->bfs_lock);
209         clear_inode(inode);
210 }
211
212 static void bfs_put_super(struct super_block *s)
213 {
214         struct bfs_sb_info *info = BFS_SB(s);
215
216         brelse(info->si_sbh);
217         mutex_destroy(&info->bfs_lock);
218         kfree(info->si_imap);
219         kfree(info);
220         s->s_fs_info = NULL;
221 }
222
223 static int bfs_statfs(struct dentry *dentry, struct kstatfs *buf)
224 {
225         struct super_block *s = dentry->d_sb;
226         struct bfs_sb_info *info = BFS_SB(s);
227         u64 id = huge_encode_dev(s->s_bdev->bd_dev);
228         buf->f_type = BFS_MAGIC;
229         buf->f_bsize = s->s_blocksize;
230         buf->f_blocks = info->si_blocks;
231         buf->f_bfree = buf->f_bavail = info->si_freeb;
232         buf->f_files = info->si_lasti + 1 - BFS_ROOT_INO;
233         buf->f_ffree = info->si_freei;
234         buf->f_fsid.val[0] = (u32)id;
235         buf->f_fsid.val[1] = (u32)(id >> 32);
236         buf->f_namelen = BFS_NAMELEN;
237         return 0;
238 }
239
240 static void bfs_write_super(struct super_block *s)
241 {
242         struct bfs_sb_info *info = BFS_SB(s);
243
244         mutex_lock(&info->bfs_lock);
245         if (!(s->s_flags & MS_RDONLY))
246                 mark_buffer_dirty(info->si_sbh);
247         s->s_dirt = 0;
248         mutex_unlock(&info->bfs_lock);
249 }
250
251 static struct kmem_cache *bfs_inode_cachep;
252
253 static struct inode *bfs_alloc_inode(struct super_block *sb)
254 {
255         struct bfs_inode_info *bi;
256         bi = kmem_cache_alloc(bfs_inode_cachep, GFP_KERNEL);
257         if (!bi)
258                 return NULL;
259         return &bi->vfs_inode;
260 }
261
262 static void bfs_destroy_inode(struct inode *inode)
263 {
264         kmem_cache_free(bfs_inode_cachep, BFS_I(inode));
265 }
266
267 static void init_once(void *foo)
268 {
269         struct bfs_inode_info *bi = foo;
270
271         inode_init_once(&bi->vfs_inode);
272 }
273
274 static int init_inodecache(void)
275 {
276         bfs_inode_cachep = kmem_cache_create("bfs_inode_cache",
277                                              sizeof(struct bfs_inode_info),
278                                              0, (SLAB_RECLAIM_ACCOUNT|
279                                                 SLAB_MEM_SPREAD),
280                                              init_once);
281         if (bfs_inode_cachep == NULL)
282                 return -ENOMEM;
283         return 0;
284 }
285
286 static void destroy_inodecache(void)
287 {
288         kmem_cache_destroy(bfs_inode_cachep);
289 }
290
291 static const struct super_operations bfs_sops = {
292         .alloc_inode    = bfs_alloc_inode,
293         .destroy_inode  = bfs_destroy_inode,
294         .write_inode    = bfs_write_inode,
295         .delete_inode   = bfs_delete_inode,
296         .put_super      = bfs_put_super,
297         .write_super    = bfs_write_super,
298         .statfs         = bfs_statfs,
299 };
300
301 void dump_imap(const char *prefix, struct super_block *s)
302 {
303 #ifdef DEBUG
304         int i;
305         char *tmpbuf = (char *)get_zeroed_page(GFP_KERNEL);
306
307         if (!tmpbuf)
308                 return;
309         for (i = BFS_SB(s)->si_lasti; i >= 0; i--) {
310                 if (i > PAGE_SIZE - 100) break;
311                 if (test_bit(i, BFS_SB(s)->si_imap))
312                         strcat(tmpbuf, "1");
313                 else
314                         strcat(tmpbuf, "0");
315         }
316         printf("BFS-fs: %s: lasti=%08lx <%s>\n",
317                                 prefix, BFS_SB(s)->si_lasti, tmpbuf);
318         free_page((unsigned long)tmpbuf);
319 #endif
320 }
321
322 static int bfs_fill_super(struct super_block *s, void *data, int silent)
323 {
324         struct buffer_head *bh;
325         struct bfs_super_block *bfs_sb;
326         struct inode *inode;
327         unsigned i, imap_len;
328         struct bfs_sb_info *info;
329         long ret = -EINVAL;
330
331         info = kzalloc(sizeof(*info), GFP_KERNEL);
332         if (!info)
333                 return -ENOMEM;
334         s->s_fs_info = info;
335
336         sb_set_blocksize(s, BFS_BSIZE);
337
338         bh = sb_bread(s, 0);
339         if(!bh)
340                 goto out;
341         bfs_sb = (struct bfs_super_block *)bh->b_data;
342         if (le32_to_cpu(bfs_sb->s_magic) != BFS_MAGIC) {
343                 if (!silent)
344                         printf("No BFS filesystem on %s (magic=%08x)\n", 
345                                 s->s_id,  le32_to_cpu(bfs_sb->s_magic));
346                 goto out;
347         }
348         if (BFS_UNCLEAN(bfs_sb, s) && !silent)
349                 printf("%s is unclean, continuing\n", s->s_id);
350
351         s->s_magic = BFS_MAGIC;
352         info->si_sbh = bh;
353         info->si_lasti = (le32_to_cpu(bfs_sb->s_start) - BFS_BSIZE) /
354                                         sizeof(struct bfs_inode)
355                                         + BFS_ROOT_INO - 1;
356         imap_len = (info->si_lasti / 8) + 1;
357         info->si_imap = kzalloc(imap_len, GFP_KERNEL);
358         if (!info->si_imap)
359                 goto out;
360         for (i = 0; i < BFS_ROOT_INO; i++)
361                 set_bit(i, info->si_imap);
362
363         s->s_op = &bfs_sops;
364         inode = bfs_iget(s, BFS_ROOT_INO);
365         if (IS_ERR(inode)) {
366                 ret = PTR_ERR(inode);
367                 kfree(info->si_imap);
368                 goto out;
369         }
370         s->s_root = d_alloc_root(inode);
371         if (!s->s_root) {
372                 iput(inode);
373                 ret = -ENOMEM;
374                 kfree(info->si_imap);
375                 goto out;
376         }
377
378         info->si_blocks = (le32_to_cpu(bfs_sb->s_end) + 1) >> BFS_BSIZE_BITS;
379         info->si_freeb = (le32_to_cpu(bfs_sb->s_end) + 1
380                         - le32_to_cpu(bfs_sb->s_start)) >> BFS_BSIZE_BITS;
381         info->si_freei = 0;
382         info->si_lf_eblk = 0;
383         bh = NULL;
384         for (i = BFS_ROOT_INO; i <= info->si_lasti; i++) {
385                 struct bfs_inode *di;
386                 int block = (i - BFS_ROOT_INO) / BFS_INODES_PER_BLOCK + 1;
387                 int off = (i - BFS_ROOT_INO) % BFS_INODES_PER_BLOCK;
388                 unsigned long eblock;
389
390                 if (!off) {
391                         brelse(bh);
392                         bh = sb_bread(s, block);
393                 }
394
395                 if (!bh)
396                         continue;
397
398                 di = (struct bfs_inode *)bh->b_data + off;
399
400                 if (!di->i_ino) {
401                         info->si_freei++;
402                         continue;
403                 }
404                 set_bit(i, info->si_imap);
405                 info->si_freeb -= BFS_FILEBLOCKS(di);
406
407                 eblock =  le32_to_cpu(di->i_eblock);
408                 if (eblock > info->si_lf_eblk)
409                         info->si_lf_eblk = eblock;
410         }
411         brelse(bh);
412         if (!(s->s_flags & MS_RDONLY)) {
413                 mark_buffer_dirty(info->si_sbh);
414                 s->s_dirt = 1;
415         } 
416         dump_imap("read_super", s);
417         mutex_init(&info->bfs_lock);
418         return 0;
419
420 out:
421         brelse(bh);
422         kfree(info);
423         s->s_fs_info = NULL;
424         return ret;
425 }
426
427 static int bfs_get_sb(struct file_system_type *fs_type,
428         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
429 {
430         return get_sb_bdev(fs_type, flags, dev_name, data, bfs_fill_super, mnt);
431 }
432
433 static struct file_system_type bfs_fs_type = {
434         .owner          = THIS_MODULE,
435         .name           = "bfs",
436         .get_sb         = bfs_get_sb,
437         .kill_sb        = kill_block_super,
438         .fs_flags       = FS_REQUIRES_DEV,
439 };
440
441 static int __init init_bfs_fs(void)
442 {
443         int err = init_inodecache();
444         if (err)
445                 goto out1;
446         err = register_filesystem(&bfs_fs_type);
447         if (err)
448                 goto out;
449         return 0;
450 out:
451         destroy_inodecache();
452 out1:
453         return err;
454 }
455
456 static void __exit exit_bfs_fs(void)
457 {
458         unregister_filesystem(&bfs_fs_type);
459         destroy_inodecache();
460 }
461
462 module_init(init_bfs_fs)
463 module_exit(exit_bfs_fs)