1 /*
2 * linux/fs/ext2/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/blkdev.h>
25 #include <linux/parser.h>
26 #include <linux/random.h>
27 #include <linux/buffer_head.h>
28 #include <linux/exportfs.h>
29 #include <linux/vfs.h>
30 #include <linux/seq_file.h>
31 #include <linux/mount.h>
32 #include <linux/log2.h>
33 #include <linux/quotaops.h>
34 #include <linux/uaccess.h>
35 #include <linux/dax.h>
36 #include <linux/iversion.h>
37 #include "ext2.h"
38 #include "xattr.h"
39 #include "acl.h"
40
41 static void ext2_write_super(struct super_block *sb);
42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
44 static int ext2_sync_fs(struct super_block *sb, int wait);
45 static int ext2_freeze(struct super_block *sb);
46 static int ext2_unfreeze(struct super_block *sb);
47
ext2_error(struct super_block * sb,const char * function,const char * fmt,...)48 void ext2_error(struct super_block *sb, const char *function,
49 const char *fmt, ...)
50 {
51 struct va_format vaf;
52 va_list args;
53 struct ext2_sb_info *sbi = EXT2_SB(sb);
54 struct ext2_super_block *es = sbi->s_es;
55
56 if (!sb_rdonly(sb)) {
57 spin_lock(&sbi->s_lock);
58 sbi->s_mount_state |= EXT2_ERROR_FS;
59 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
60 spin_unlock(&sbi->s_lock);
61 ext2_sync_super(sb, es, 1);
62 }
63
64 va_start(args, fmt);
65
66 vaf.fmt = fmt;
67 vaf.va = &args;
68
69 printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
70 sb->s_id, function, &vaf);
71
72 va_end(args);
73
74 if (test_opt(sb, ERRORS_PANIC))
75 panic("EXT2-fs: panic from previous error\n");
76 if (test_opt(sb, ERRORS_RO)) {
77 ext2_msg(sb, KERN_CRIT,
78 "error: remounting filesystem read-only");
79 sb->s_flags |= SB_RDONLY;
80 }
81 }
82
ext2_msg(struct super_block * sb,const char * prefix,const char * fmt,...)83 void ext2_msg(struct super_block *sb, const char *prefix,
84 const char *fmt, ...)
85 {
86 struct va_format vaf;
87 va_list args;
88
89 va_start(args, fmt);
90
91 vaf.fmt = fmt;
92 vaf.va = &args;
93
94 printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
95
96 va_end(args);
97 }
98
99 /*
100 * This must be called with sbi->s_lock held.
101 */
ext2_update_dynamic_rev(struct super_block * sb)102 void ext2_update_dynamic_rev(struct super_block *sb)
103 {
104 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
105
106 if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
107 return;
108
109 ext2_msg(sb, KERN_WARNING,
110 "warning: updating to rev %d because of "
111 "new feature flag, running e2fsck is recommended",
112 EXT2_DYNAMIC_REV);
113
114 es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
115 es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
116 es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
117 /* leave es->s_feature_*compat flags alone */
118 /* es->s_uuid will be set by e2fsck if empty */
119
120 /*
121 * The rest of the superblock fields should be zero, and if not it
122 * means they are likely already in use, so leave them alone. We
123 * can leave it up to e2fsck to clean up any inconsistencies there.
124 */
125 }
126
127 #ifdef CONFIG_QUOTA
128 static int ext2_quota_off(struct super_block *sb, int type);
129
ext2_quota_off_umount(struct super_block * sb)130 static void ext2_quota_off_umount(struct super_block *sb)
131 {
132 int type;
133
134 for (type = 0; type < MAXQUOTAS; type++)
135 ext2_quota_off(sb, type);
136 }
137 #else
ext2_quota_off_umount(struct super_block * sb)138 static inline void ext2_quota_off_umount(struct super_block *sb)
139 {
140 }
141 #endif
142
ext2_put_super(struct super_block * sb)143 static void ext2_put_super (struct super_block * sb)
144 {
145 int db_count;
146 int i;
147 struct ext2_sb_info *sbi = EXT2_SB(sb);
148
149 ext2_quota_off_umount(sb);
150
151 if (sbi->s_ea_block_cache) {
152 ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
153 sbi->s_ea_block_cache = NULL;
154 }
155 if (!sb_rdonly(sb)) {
156 struct ext2_super_block *es = sbi->s_es;
157
158 spin_lock(&sbi->s_lock);
159 es->s_state = cpu_to_le16(sbi->s_mount_state);
160 spin_unlock(&sbi->s_lock);
161 ext2_sync_super(sb, es, 1);
162 }
163 db_count = sbi->s_gdb_count;
164 for (i = 0; i < db_count; i++)
165 if (sbi->s_group_desc[i])
166 brelse (sbi->s_group_desc[i]);
167 kfree(sbi->s_group_desc);
168 kfree(sbi->s_debts);
169 percpu_counter_destroy(&sbi->s_freeblocks_counter);
170 percpu_counter_destroy(&sbi->s_freeinodes_counter);
171 percpu_counter_destroy(&sbi->s_dirs_counter);
172 brelse (sbi->s_sbh);
173 sb->s_fs_info = NULL;
174 kfree(sbi->s_blockgroup_lock);
175 fs_put_dax(sbi->s_daxdev);
176 kfree(sbi);
177 }
178
179 static struct kmem_cache * ext2_inode_cachep;
180
ext2_alloc_inode(struct super_block * sb)181 static struct inode *ext2_alloc_inode(struct super_block *sb)
182 {
183 struct ext2_inode_info *ei;
184 ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
185 if (!ei)
186 return NULL;
187 ei->i_block_alloc_info = NULL;
188 inode_set_iversion(&ei->vfs_inode, 1);
189 #ifdef CONFIG_QUOTA
190 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
191 #endif
192
193 return &ei->vfs_inode;
194 }
195
ext2_i_callback(struct rcu_head * head)196 static void ext2_i_callback(struct rcu_head *head)
197 {
198 struct inode *inode = container_of(head, struct inode, i_rcu);
199 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
200 }
201
ext2_destroy_inode(struct inode * inode)202 static void ext2_destroy_inode(struct inode *inode)
203 {
204 call_rcu(&inode->i_rcu, ext2_i_callback);
205 }
206
init_once(void * foo)207 static void init_once(void *foo)
208 {
209 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
210
211 rwlock_init(&ei->i_meta_lock);
212 #ifdef CONFIG_EXT2_FS_XATTR
213 init_rwsem(&ei->xattr_sem);
214 #endif
215 mutex_init(&ei->truncate_mutex);
216 #ifdef CONFIG_FS_DAX
217 init_rwsem(&ei->dax_sem);
218 #endif
219 inode_init_once(&ei->vfs_inode);
220 }
221
init_inodecache(void)222 static int __init init_inodecache(void)
223 {
224 ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
225 sizeof(struct ext2_inode_info), 0,
226 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
227 SLAB_ACCOUNT),
228 offsetof(struct ext2_inode_info, i_data),
229 sizeof_field(struct ext2_inode_info, i_data),
230 init_once);
231 if (ext2_inode_cachep == NULL)
232 return -ENOMEM;
233 return 0;
234 }
235
destroy_inodecache(void)236 static void destroy_inodecache(void)
237 {
238 /*
239 * Make sure all delayed rcu free inodes are flushed before we
240 * destroy cache.
241 */
242 rcu_barrier();
243 kmem_cache_destroy(ext2_inode_cachep);
244 }
245
ext2_show_options(struct seq_file * seq,struct dentry * root)246 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
247 {
248 struct super_block *sb = root->d_sb;
249 struct ext2_sb_info *sbi = EXT2_SB(sb);
250 struct ext2_super_block *es = sbi->s_es;
251 unsigned long def_mount_opts;
252
253 spin_lock(&sbi->s_lock);
254 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
255
256 if (sbi->s_sb_block != 1)
257 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
258 if (test_opt(sb, MINIX_DF))
259 seq_puts(seq, ",minixdf");
260 if (test_opt(sb, GRPID))
261 seq_puts(seq, ",grpid");
262 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
263 seq_puts(seq, ",nogrpid");
264 if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
265 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
266 seq_printf(seq, ",resuid=%u",
267 from_kuid_munged(&init_user_ns, sbi->s_resuid));
268 }
269 if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
270 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
271 seq_printf(seq, ",resgid=%u",
272 from_kgid_munged(&init_user_ns, sbi->s_resgid));
273 }
274 if (test_opt(sb, ERRORS_RO)) {
275 int def_errors = le16_to_cpu(es->s_errors);
276
277 if (def_errors == EXT2_ERRORS_PANIC ||
278 def_errors == EXT2_ERRORS_CONTINUE) {
279 seq_puts(seq, ",errors=remount-ro");
280 }
281 }
282 if (test_opt(sb, ERRORS_CONT))
283 seq_puts(seq, ",errors=continue");
284 if (test_opt(sb, ERRORS_PANIC))
285 seq_puts(seq, ",errors=panic");
286 if (test_opt(sb, NO_UID32))
287 seq_puts(seq, ",nouid32");
288 if (test_opt(sb, DEBUG))
289 seq_puts(seq, ",debug");
290 if (test_opt(sb, OLDALLOC))
291 seq_puts(seq, ",oldalloc");
292
293 #ifdef CONFIG_EXT2_FS_XATTR
294 if (test_opt(sb, XATTR_USER))
295 seq_puts(seq, ",user_xattr");
296 if (!test_opt(sb, XATTR_USER) &&
297 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
298 seq_puts(seq, ",nouser_xattr");
299 }
300 #endif
301
302 #ifdef CONFIG_EXT2_FS_POSIX_ACL
303 if (test_opt(sb, POSIX_ACL))
304 seq_puts(seq, ",acl");
305 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
306 seq_puts(seq, ",noacl");
307 #endif
308
309 if (test_opt(sb, NOBH))
310 seq_puts(seq, ",nobh");
311
312 #if defined(CONFIG_QUOTA)
313 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
314 seq_puts(seq, ",usrquota");
315
316 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
317 seq_puts(seq, ",grpquota");
318 #endif
319
320 #ifdef CONFIG_FS_DAX
321 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
322 seq_puts(seq, ",xip");
323 if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
324 seq_puts(seq, ",dax");
325 #endif
326
327 if (!test_opt(sb, RESERVATION))
328 seq_puts(seq, ",noreservation");
329
330 spin_unlock(&sbi->s_lock);
331 return 0;
332 }
333
334 #ifdef CONFIG_QUOTA
335 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
336 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
337 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
338 const struct path *path);
ext2_get_dquots(struct inode * inode)339 static struct dquot **ext2_get_dquots(struct inode *inode)
340 {
341 return EXT2_I(inode)->i_dquot;
342 }
343
344 static const struct quotactl_ops ext2_quotactl_ops = {
345 .quota_on = ext2_quota_on,
346 .quota_off = ext2_quota_off,
347 .quota_sync = dquot_quota_sync,
348 .get_state = dquot_get_state,
349 .set_info = dquot_set_dqinfo,
350 .get_dqblk = dquot_get_dqblk,
351 .set_dqblk = dquot_set_dqblk,
352 .get_nextdqblk = dquot_get_next_dqblk,
353 };
354 #endif
355
356 static const struct super_operations ext2_sops = {
357 .alloc_inode = ext2_alloc_inode,
358 .destroy_inode = ext2_destroy_inode,
359 .write_inode = ext2_write_inode,
360 .evict_inode = ext2_evict_inode,
361 .put_super = ext2_put_super,
362 .sync_fs = ext2_sync_fs,
363 .freeze_fs = ext2_freeze,
364 .unfreeze_fs = ext2_unfreeze,
365 .statfs = ext2_statfs,
366 .remount_fs = ext2_remount,
367 .show_options = ext2_show_options,
368 #ifdef CONFIG_QUOTA
369 .quota_read = ext2_quota_read,
370 .quota_write = ext2_quota_write,
371 .get_dquots = ext2_get_dquots,
372 #endif
373 };
374
ext2_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)375 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
376 u64 ino, u32 generation)
377 {
378 struct inode *inode;
379
380 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
381 return ERR_PTR(-ESTALE);
382 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
383 return ERR_PTR(-ESTALE);
384
385 /*
386 * ext2_iget isn't quite right if the inode is currently unallocated!
387 * However ext2_iget currently does appropriate checks to handle stale
388 * inodes so everything is OK.
389 */
390 inode = ext2_iget(sb, ino);
391 if (IS_ERR(inode))
392 return ERR_CAST(inode);
393 if (generation && inode->i_generation != generation) {
394 /* we didn't find the right inode.. */
395 iput(inode);
396 return ERR_PTR(-ESTALE);
397 }
398 return inode;
399 }
400
ext2_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)401 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
402 int fh_len, int fh_type)
403 {
404 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
405 ext2_nfs_get_inode);
406 }
407
ext2_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)408 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
409 int fh_len, int fh_type)
410 {
411 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
412 ext2_nfs_get_inode);
413 }
414
415 static const struct export_operations ext2_export_ops = {
416 .fh_to_dentry = ext2_fh_to_dentry,
417 .fh_to_parent = ext2_fh_to_parent,
418 .get_parent = ext2_get_parent,
419 };
420
get_sb_block(void ** data)421 static unsigned long get_sb_block(void **data)
422 {
423 unsigned long sb_block;
424 char *options = (char *) *data;
425
426 if (!options || strncmp(options, "sb=", 3) != 0)
427 return 1; /* Default location */
428 options += 3;
429 sb_block = simple_strtoul(options, &options, 0);
430 if (*options && *options != ',') {
431 printk("EXT2-fs: Invalid sb specification: %s\n",
432 (char *) *data);
433 return 1;
434 }
435 if (*options == ',')
436 options++;
437 *data = (void *) options;
438 return sb_block;
439 }
440
441 enum {
442 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
443 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
444 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
445 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
446 Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
447 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
448 };
449
450 static const match_table_t tokens = {
451 {Opt_bsd_df, "bsddf"},
452 {Opt_minix_df, "minixdf"},
453 {Opt_grpid, "grpid"},
454 {Opt_grpid, "bsdgroups"},
455 {Opt_nogrpid, "nogrpid"},
456 {Opt_nogrpid, "sysvgroups"},
457 {Opt_resgid, "resgid=%u"},
458 {Opt_resuid, "resuid=%u"},
459 {Opt_sb, "sb=%u"},
460 {Opt_err_cont, "errors=continue"},
461 {Opt_err_panic, "errors=panic"},
462 {Opt_err_ro, "errors=remount-ro"},
463 {Opt_nouid32, "nouid32"},
464 {Opt_nocheck, "check=none"},
465 {Opt_nocheck, "nocheck"},
466 {Opt_debug, "debug"},
467 {Opt_oldalloc, "oldalloc"},
468 {Opt_orlov, "orlov"},
469 {Opt_nobh, "nobh"},
470 {Opt_user_xattr, "user_xattr"},
471 {Opt_nouser_xattr, "nouser_xattr"},
472 {Opt_acl, "acl"},
473 {Opt_noacl, "noacl"},
474 {Opt_xip, "xip"},
475 {Opt_dax, "dax"},
476 {Opt_grpquota, "grpquota"},
477 {Opt_ignore, "noquota"},
478 {Opt_quota, "quota"},
479 {Opt_usrquota, "usrquota"},
480 {Opt_reservation, "reservation"},
481 {Opt_noreservation, "noreservation"},
482 {Opt_err, NULL}
483 };
484
parse_options(char * options,struct super_block * sb,struct ext2_mount_options * opts)485 static int parse_options(char *options, struct super_block *sb,
486 struct ext2_mount_options *opts)
487 {
488 char *p;
489 substring_t args[MAX_OPT_ARGS];
490 int option;
491 kuid_t uid;
492 kgid_t gid;
493
494 if (!options)
495 return 1;
496
497 while ((p = strsep (&options, ",")) != NULL) {
498 int token;
499 if (!*p)
500 continue;
501
502 token = match_token(p, tokens, args);
503 switch (token) {
504 case Opt_bsd_df:
505 clear_opt (opts->s_mount_opt, MINIX_DF);
506 break;
507 case Opt_minix_df:
508 set_opt (opts->s_mount_opt, MINIX_DF);
509 break;
510 case Opt_grpid:
511 set_opt (opts->s_mount_opt, GRPID);
512 break;
513 case Opt_nogrpid:
514 clear_opt (opts->s_mount_opt, GRPID);
515 break;
516 case Opt_resuid:
517 if (match_int(&args[0], &option))
518 return 0;
519 uid = make_kuid(current_user_ns(), option);
520 if (!uid_valid(uid)) {
521 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
522 return 0;
523
524 }
525 opts->s_resuid = uid;
526 break;
527 case Opt_resgid:
528 if (match_int(&args[0], &option))
529 return 0;
530 gid = make_kgid(current_user_ns(), option);
531 if (!gid_valid(gid)) {
532 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
533 return 0;
534 }
535 opts->s_resgid = gid;
536 break;
537 case Opt_sb:
538 /* handled by get_sb_block() instead of here */
539 /* *sb_block = match_int(&args[0]); */
540 break;
541 case Opt_err_panic:
542 clear_opt (opts->s_mount_opt, ERRORS_CONT);
543 clear_opt (opts->s_mount_opt, ERRORS_RO);
544 set_opt (opts->s_mount_opt, ERRORS_PANIC);
545 break;
546 case Opt_err_ro:
547 clear_opt (opts->s_mount_opt, ERRORS_CONT);
548 clear_opt (opts->s_mount_opt, ERRORS_PANIC);
549 set_opt (opts->s_mount_opt, ERRORS_RO);
550 break;
551 case Opt_err_cont:
552 clear_opt (opts->s_mount_opt, ERRORS_RO);
553 clear_opt (opts->s_mount_opt, ERRORS_PANIC);
554 set_opt (opts->s_mount_opt, ERRORS_CONT);
555 break;
556 case Opt_nouid32:
557 set_opt (opts->s_mount_opt, NO_UID32);
558 break;
559 case Opt_nocheck:
560 ext2_msg(sb, KERN_WARNING,
561 "Option nocheck/check=none is deprecated and"
562 " will be removed in June 2020.");
563 clear_opt (opts->s_mount_opt, CHECK);
564 break;
565 case Opt_debug:
566 set_opt (opts->s_mount_opt, DEBUG);
567 break;
568 case Opt_oldalloc:
569 set_opt (opts->s_mount_opt, OLDALLOC);
570 break;
571 case Opt_orlov:
572 clear_opt (opts->s_mount_opt, OLDALLOC);
573 break;
574 case Opt_nobh:
575 set_opt (opts->s_mount_opt, NOBH);
576 break;
577 #ifdef CONFIG_EXT2_FS_XATTR
578 case Opt_user_xattr:
579 set_opt (opts->s_mount_opt, XATTR_USER);
580 break;
581 case Opt_nouser_xattr:
582 clear_opt (opts->s_mount_opt, XATTR_USER);
583 break;
584 #else
585 case Opt_user_xattr:
586 case Opt_nouser_xattr:
587 ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
588 "not supported");
589 break;
590 #endif
591 #ifdef CONFIG_EXT2_FS_POSIX_ACL
592 case Opt_acl:
593 set_opt(opts->s_mount_opt, POSIX_ACL);
594 break;
595 case Opt_noacl:
596 clear_opt(opts->s_mount_opt, POSIX_ACL);
597 break;
598 #else
599 case Opt_acl:
600 case Opt_noacl:
601 ext2_msg(sb, KERN_INFO,
602 "(no)acl options not supported");
603 break;
604 #endif
605 case Opt_xip:
606 ext2_msg(sb, KERN_INFO, "use dax instead of xip");
607 set_opt(opts->s_mount_opt, XIP);
608 /* Fall through */
609 case Opt_dax:
610 #ifdef CONFIG_FS_DAX
611 ext2_msg(sb, KERN_WARNING,
612 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
613 set_opt(opts->s_mount_opt, DAX);
614 #else
615 ext2_msg(sb, KERN_INFO, "dax option not supported");
616 #endif
617 break;
618
619 #if defined(CONFIG_QUOTA)
620 case Opt_quota:
621 case Opt_usrquota:
622 set_opt(opts->s_mount_opt, USRQUOTA);
623 break;
624
625 case Opt_grpquota:
626 set_opt(opts->s_mount_opt, GRPQUOTA);
627 break;
628 #else
629 case Opt_quota:
630 case Opt_usrquota:
631 case Opt_grpquota:
632 ext2_msg(sb, KERN_INFO,
633 "quota operations not supported");
634 break;
635 #endif
636
637 case Opt_reservation:
638 set_opt(opts->s_mount_opt, RESERVATION);
639 ext2_msg(sb, KERN_INFO, "reservations ON");
640 break;
641 case Opt_noreservation:
642 clear_opt(opts->s_mount_opt, RESERVATION);
643 ext2_msg(sb, KERN_INFO, "reservations OFF");
644 break;
645 case Opt_ignore:
646 break;
647 default:
648 return 0;
649 }
650 }
651 return 1;
652 }
653
ext2_setup_super(struct super_block * sb,struct ext2_super_block * es,int read_only)654 static int ext2_setup_super (struct super_block * sb,
655 struct ext2_super_block * es,
656 int read_only)
657 {
658 int res = 0;
659 struct ext2_sb_info *sbi = EXT2_SB(sb);
660
661 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
662 ext2_msg(sb, KERN_ERR,
663 "error: revision level too high, "
664 "forcing read-only mode");
665 res = SB_RDONLY;
666 }
667 if (read_only)
668 return res;
669 if (!(sbi->s_mount_state & EXT2_VALID_FS))
670 ext2_msg(sb, KERN_WARNING,
671 "warning: mounting unchecked fs, "
672 "running e2fsck is recommended");
673 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
674 ext2_msg(sb, KERN_WARNING,
675 "warning: mounting fs with errors, "
676 "running e2fsck is recommended");
677 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
678 le16_to_cpu(es->s_mnt_count) >=
679 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
680 ext2_msg(sb, KERN_WARNING,
681 "warning: maximal mount count reached, "
682 "running e2fsck is recommended");
683 else if (le32_to_cpu(es->s_checkinterval) &&
684 (le32_to_cpu(es->s_lastcheck) +
685 le32_to_cpu(es->s_checkinterval) <=
686 ktime_get_real_seconds()))
687 ext2_msg(sb, KERN_WARNING,
688 "warning: checktime reached, "
689 "running e2fsck is recommended");
690 if (!le16_to_cpu(es->s_max_mnt_count))
691 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
692 le16_add_cpu(&es->s_mnt_count, 1);
693 if (test_opt (sb, DEBUG))
694 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, gc=%lu, "
695 "bpg=%lu, ipg=%lu, mo=%04lx]",
696 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
697 sbi->s_groups_count,
698 EXT2_BLOCKS_PER_GROUP(sb),
699 EXT2_INODES_PER_GROUP(sb),
700 sbi->s_mount_opt);
701 return res;
702 }
703
ext2_check_descriptors(struct super_block * sb)704 static int ext2_check_descriptors(struct super_block *sb)
705 {
706 int i;
707 struct ext2_sb_info *sbi = EXT2_SB(sb);
708
709 ext2_debug ("Checking group descriptors");
710
711 for (i = 0; i < sbi->s_groups_count; i++) {
712 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
713 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
714 ext2_fsblk_t last_block;
715
716 if (i == sbi->s_groups_count - 1)
717 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
718 else
719 last_block = first_block +
720 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
721
722 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
723 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
724 {
725 ext2_error (sb, "ext2_check_descriptors",
726 "Block bitmap for group %d"
727 " not in group (block %lu)!",
728 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
729 return 0;
730 }
731 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
732 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
733 {
734 ext2_error (sb, "ext2_check_descriptors",
735 "Inode bitmap for group %d"
736 " not in group (block %lu)!",
737 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
738 return 0;
739 }
740 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
741 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
742 last_block)
743 {
744 ext2_error (sb, "ext2_check_descriptors",
745 "Inode table for group %d"
746 " not in group (block %lu)!",
747 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
748 return 0;
749 }
750 }
751 return 1;
752 }
753
754 /*
755 * Maximal file size. There is a direct, and {,double-,triple-}indirect
756 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
757 * We need to be 1 filesystem block less than the 2^32 sector limit.
758 */
ext2_max_size(int bits)759 static loff_t ext2_max_size(int bits)
760 {
761 loff_t res = EXT2_NDIR_BLOCKS;
762 int meta_blocks;
763 unsigned int upper_limit;
764 unsigned int ppb = 1 << (bits-2);
765
766 /* This is calculated to be the largest file size for a
767 * dense, file such that the total number of
768 * sectors in the file, including data and all indirect blocks,
769 * does not exceed 2^32 -1
770 * __u32 i_blocks representing the total number of
771 * 512 bytes blocks of the file
772 */
773 upper_limit = (1LL << 32) - 1;
774
775 /* total blocks in file system block size */
776 upper_limit >>= (bits - 9);
777
778 /* Compute how many blocks we can address by block tree */
779 res += 1LL << (bits-2);
780 res += 1LL << (2*(bits-2));
781 res += 1LL << (3*(bits-2));
782 /* Compute how many metadata blocks are needed */
783 meta_blocks = 1;
784 meta_blocks += 1 + ppb;
785 meta_blocks += 1 + ppb + ppb * ppb;
786 /* Does block tree limit file size? */
787 if (res + meta_blocks <= upper_limit)
788 goto check_lfs;
789
790 res = upper_limit;
791 /* How many metadata blocks are needed for addressing upper_limit? */
792 upper_limit -= EXT2_NDIR_BLOCKS;
793 /* indirect blocks */
794 meta_blocks = 1;
795 upper_limit -= ppb;
796 /* double indirect blocks */
797 if (upper_limit < ppb * ppb) {
798 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
799 res -= meta_blocks;
800 goto check_lfs;
801 }
802 meta_blocks += 1 + ppb;
803 upper_limit -= ppb * ppb;
804 /* tripple indirect blocks for the rest */
805 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
806 DIV_ROUND_UP(upper_limit, ppb*ppb);
807 res -= meta_blocks;
808 check_lfs:
809 res <<= bits;
810 if (res > MAX_LFS_FILESIZE)
811 res = MAX_LFS_FILESIZE;
812
813 return res;
814 }
815
descriptor_loc(struct super_block * sb,unsigned long logic_sb_block,int nr)816 static unsigned long descriptor_loc(struct super_block *sb,
817 unsigned long logic_sb_block,
818 int nr)
819 {
820 struct ext2_sb_info *sbi = EXT2_SB(sb);
821 unsigned long bg, first_meta_bg;
822 int has_super = 0;
823
824 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
825
826 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
827 nr < first_meta_bg)
828 return (logic_sb_block + nr + 1);
829 bg = sbi->s_desc_per_block * nr;
830 if (ext2_bg_has_super(sb, bg))
831 has_super = 1;
832
833 return ext2_group_first_block_no(sb, bg) + has_super;
834 }
835
ext2_fill_super(struct super_block * sb,void * data,int silent)836 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
837 {
838 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
839 struct buffer_head * bh;
840 struct ext2_sb_info * sbi;
841 struct ext2_super_block * es;
842 struct inode *root;
843 unsigned long block;
844 unsigned long sb_block = get_sb_block(&data);
845 unsigned long logic_sb_block;
846 unsigned long offset = 0;
847 unsigned long def_mount_opts;
848 long ret = -ENOMEM;
849 int blocksize = BLOCK_SIZE;
850 int db_count;
851 int i, j;
852 __le32 features;
853 int err;
854 struct ext2_mount_options opts;
855
856 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
857 if (!sbi)
858 goto failed;
859
860 sbi->s_blockgroup_lock =
861 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
862 if (!sbi->s_blockgroup_lock) {
863 kfree(sbi);
864 goto failed;
865 }
866 sb->s_fs_info = sbi;
867 sbi->s_sb_block = sb_block;
868 sbi->s_daxdev = dax_dev;
869
870 spin_lock_init(&sbi->s_lock);
871 ret = -EINVAL;
872
873 /*
874 * See what the current blocksize for the device is, and
875 * use that as the blocksize. Otherwise (or if the blocksize
876 * is smaller than the default) use the default.
877 * This is important for devices that have a hardware
878 * sectorsize that is larger than the default.
879 */
880 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
881 if (!blocksize) {
882 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
883 goto failed_sbi;
884 }
885
886 /*
887 * If the superblock doesn't start on a hardware sector boundary,
888 * calculate the offset.
889 */
890 if (blocksize != BLOCK_SIZE) {
891 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
892 offset = (sb_block*BLOCK_SIZE) % blocksize;
893 } else {
894 logic_sb_block = sb_block;
895 }
896
897 if (!(bh = sb_bread(sb, logic_sb_block))) {
898 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
899 goto failed_sbi;
900 }
901 /*
902 * Note: s_es must be initialized as soon as possible because
903 * some ext2 macro-instructions depend on its value
904 */
905 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
906 sbi->s_es = es;
907 sb->s_magic = le16_to_cpu(es->s_magic);
908
909 if (sb->s_magic != EXT2_SUPER_MAGIC)
910 goto cantfind_ext2;
911
912 opts.s_mount_opt = 0;
913 /* Set defaults before we parse the mount options */
914 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
915 if (def_mount_opts & EXT2_DEFM_DEBUG)
916 set_opt(opts.s_mount_opt, DEBUG);
917 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
918 set_opt(opts.s_mount_opt, GRPID);
919 if (def_mount_opts & EXT2_DEFM_UID16)
920 set_opt(opts.s_mount_opt, NO_UID32);
921 #ifdef CONFIG_EXT2_FS_XATTR
922 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
923 set_opt(opts.s_mount_opt, XATTR_USER);
924 #endif
925 #ifdef CONFIG_EXT2_FS_POSIX_ACL
926 if (def_mount_opts & EXT2_DEFM_ACL)
927 set_opt(opts.s_mount_opt, POSIX_ACL);
928 #endif
929
930 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
931 set_opt(opts.s_mount_opt, ERRORS_PANIC);
932 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
933 set_opt(opts.s_mount_opt, ERRORS_CONT);
934 else
935 set_opt(opts.s_mount_opt, ERRORS_RO);
936
937 opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
938 opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
939
940 set_opt(opts.s_mount_opt, RESERVATION);
941
942 if (!parse_options((char *) data, sb, &opts))
943 goto failed_mount;
944
945 sbi->s_mount_opt = opts.s_mount_opt;
946 sbi->s_resuid = opts.s_resuid;
947 sbi->s_resgid = opts.s_resgid;
948
949 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
950 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
951 SB_POSIXACL : 0);
952 sb->s_iflags |= SB_I_CGROUPWB;
953
954 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
955 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
956 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
957 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
958 ext2_msg(sb, KERN_WARNING,
959 "warning: feature flags set on rev 0 fs, "
960 "running e2fsck is recommended");
961 /*
962 * Check feature flags regardless of the revision level, since we
963 * previously didn't change the revision level when setting the flags,
964 * so there is a chance incompat flags are set on a rev 0 filesystem.
965 */
966 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
967 if (features) {
968 ext2_msg(sb, KERN_ERR, "error: couldn't mount because of "
969 "unsupported optional features (%x)",
970 le32_to_cpu(features));
971 goto failed_mount;
972 }
973 if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
974 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
975 "unsupported optional features (%x)",
976 le32_to_cpu(features));
977 goto failed_mount;
978 }
979
980 if (le32_to_cpu(es->s_log_block_size) >
981 (EXT2_MAX_BLOCK_LOG_SIZE - BLOCK_SIZE_BITS)) {
982 ext2_msg(sb, KERN_ERR,
983 "Invalid log block size: %u",
984 le32_to_cpu(es->s_log_block_size));
985 goto failed_mount;
986 }
987 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
988
989 if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
990 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
991 ext2_msg(sb, KERN_ERR,
992 "DAX unsupported by block device. Turning off DAX.");
993 sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
994 }
995 }
996
997 /* If the blocksize doesn't match, re-read the thing.. */
998 if (sb->s_blocksize != blocksize) {
999 brelse(bh);
1000
1001 if (!sb_set_blocksize(sb, blocksize)) {
1002 ext2_msg(sb, KERN_ERR,
1003 "error: bad blocksize %d", blocksize);
1004 goto failed_sbi;
1005 }
1006
1007 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
1008 offset = (sb_block*BLOCK_SIZE) % blocksize;
1009 bh = sb_bread(sb, logic_sb_block);
1010 if(!bh) {
1011 ext2_msg(sb, KERN_ERR, "error: couldn't read"
1012 "superblock on 2nd try");
1013 goto failed_sbi;
1014 }
1015 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
1016 sbi->s_es = es;
1017 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
1018 ext2_msg(sb, KERN_ERR, "error: magic mismatch");
1019 goto failed_mount;
1020 }
1021 }
1022
1023 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1024 sb->s_max_links = EXT2_LINK_MAX;
1025
1026 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1027 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1028 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1029 } else {
1030 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1031 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1032 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1033 !is_power_of_2(sbi->s_inode_size) ||
1034 (sbi->s_inode_size > blocksize)) {
1035 ext2_msg(sb, KERN_ERR,
1036 "error: unsupported inode size: %d",
1037 sbi->s_inode_size);
1038 goto failed_mount;
1039 }
1040 }
1041
1042 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1043 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1044
1045 if (EXT2_INODE_SIZE(sb) == 0)
1046 goto cantfind_ext2;
1047 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1048 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1049 goto cantfind_ext2;
1050 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1051 sbi->s_inodes_per_block;
1052 sbi->s_desc_per_block = sb->s_blocksize /
1053 sizeof (struct ext2_group_desc);
1054 sbi->s_sbh = bh;
1055 sbi->s_mount_state = le16_to_cpu(es->s_state);
1056 sbi->s_addr_per_block_bits =
1057 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1058 sbi->s_desc_per_block_bits =
1059 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1060
1061 if (sb->s_magic != EXT2_SUPER_MAGIC)
1062 goto cantfind_ext2;
1063
1064 if (sb->s_blocksize != bh->b_size) {
1065 if (!silent)
1066 ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1067 goto failed_mount;
1068 }
1069
1070 if (es->s_log_frag_size != es->s_log_block_size) {
1071 ext2_msg(sb, KERN_ERR,
1072 "error: fragsize log %u != blocksize log %u",
1073 le32_to_cpu(es->s_log_frag_size), sb->s_blocksize_bits);
1074 goto failed_mount;
1075 }
1076
1077 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1078 ext2_msg(sb, KERN_ERR,
1079 "error: #blocks per group too big: %lu",
1080 sbi->s_blocks_per_group);
1081 goto failed_mount;
1082 }
1083 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
1084 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1085 ext2_msg(sb, KERN_ERR,
1086 "error: invalid #inodes per group: %lu",
1087 sbi->s_inodes_per_group);
1088 goto failed_mount;
1089 }
1090
1091 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1092 goto cantfind_ext2;
1093 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1094 le32_to_cpu(es->s_first_data_block) - 1)
1095 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1096 if ((u64)sbi->s_groups_count * sbi->s_inodes_per_group !=
1097 le32_to_cpu(es->s_inodes_count)) {
1098 ext2_msg(sb, KERN_ERR, "error: invalid #inodes: %u vs computed %llu",
1099 le32_to_cpu(es->s_inodes_count),
1100 (u64)sbi->s_groups_count * sbi->s_inodes_per_group);
1101 goto failed_mount;
1102 }
1103 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1104 EXT2_DESC_PER_BLOCK(sb);
1105 sbi->s_group_desc = kmalloc_array (db_count,
1106 sizeof(struct buffer_head *),
1107 GFP_KERNEL);
1108 if (sbi->s_group_desc == NULL) {
1109 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1110 goto failed_mount;
1111 }
1112 bgl_lock_init(sbi->s_blockgroup_lock);
1113 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1114 if (!sbi->s_debts) {
1115 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1116 goto failed_mount_group_desc;
1117 }
1118 for (i = 0; i < db_count; i++) {
1119 block = descriptor_loc(sb, logic_sb_block, i);
1120 sbi->s_group_desc[i] = sb_bread(sb, block);
1121 if (!sbi->s_group_desc[i]) {
1122 for (j = 0; j < i; j++)
1123 brelse (sbi->s_group_desc[j]);
1124 ext2_msg(sb, KERN_ERR,
1125 "error: unable to read group descriptors");
1126 goto failed_mount_group_desc;
1127 }
1128 }
1129 if (!ext2_check_descriptors (sb)) {
1130 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1131 goto failed_mount2;
1132 }
1133 sbi->s_gdb_count = db_count;
1134 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1135 spin_lock_init(&sbi->s_next_gen_lock);
1136
1137 /* per fileystem reservation list head & lock */
1138 spin_lock_init(&sbi->s_rsv_window_lock);
1139 sbi->s_rsv_window_root = RB_ROOT;
1140 /*
1141 * Add a single, static dummy reservation to the start of the
1142 * reservation window list --- it gives us a placeholder for
1143 * append-at-start-of-list which makes the allocation logic
1144 * _much_ simpler.
1145 */
1146 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1147 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1148 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1149 sbi->s_rsv_window_head.rsv_goal_size = 0;
1150 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1151
1152 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1153 ext2_count_free_blocks(sb), GFP_KERNEL);
1154 if (!err) {
1155 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1156 ext2_count_free_inodes(sb), GFP_KERNEL);
1157 }
1158 if (!err) {
1159 err = percpu_counter_init(&sbi->s_dirs_counter,
1160 ext2_count_dirs(sb), GFP_KERNEL);
1161 }
1162 if (err) {
1163 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1164 goto failed_mount3;
1165 }
1166
1167 #ifdef CONFIG_EXT2_FS_XATTR
1168 sbi->s_ea_block_cache = ext2_xattr_create_cache();
1169 if (!sbi->s_ea_block_cache) {
1170 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1171 goto failed_mount3;
1172 }
1173 #endif
1174 /*
1175 * set up enough so that it can read an inode
1176 */
1177 sb->s_op = &ext2_sops;
1178 sb->s_export_op = &ext2_export_ops;
1179 sb->s_xattr = ext2_xattr_handlers;
1180
1181 #ifdef CONFIG_QUOTA
1182 sb->dq_op = &dquot_operations;
1183 sb->s_qcop = &ext2_quotactl_ops;
1184 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1185 #endif
1186
1187 root = ext2_iget(sb, EXT2_ROOT_INO);
1188 if (IS_ERR(root)) {
1189 ret = PTR_ERR(root);
1190 goto failed_mount3;
1191 }
1192 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1193 iput(root);
1194 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1195 goto failed_mount3;
1196 }
1197
1198 sb->s_root = d_make_root(root);
1199 if (!sb->s_root) {
1200 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1201 ret = -ENOMEM;
1202 goto failed_mount3;
1203 }
1204 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1205 ext2_msg(sb, KERN_WARNING,
1206 "warning: mounting ext3 filesystem as ext2");
1207 if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1208 sb->s_flags |= SB_RDONLY;
1209 ext2_write_super(sb);
1210 return 0;
1211
1212 cantfind_ext2:
1213 if (!silent)
1214 ext2_msg(sb, KERN_ERR,
1215 "error: can't find an ext2 filesystem on dev %s.",
1216 sb->s_id);
1217 goto failed_mount;
1218 failed_mount3:
1219 if (sbi->s_ea_block_cache)
1220 ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1221 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1222 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1223 percpu_counter_destroy(&sbi->s_dirs_counter);
1224 failed_mount2:
1225 for (i = 0; i < db_count; i++)
1226 brelse(sbi->s_group_desc[i]);
1227 failed_mount_group_desc:
1228 kfree(sbi->s_group_desc);
1229 kfree(sbi->s_debts);
1230 failed_mount:
1231 brelse(bh);
1232 failed_sbi:
1233 sb->s_fs_info = NULL;
1234 kfree(sbi->s_blockgroup_lock);
1235 kfree(sbi);
1236 failed:
1237 fs_put_dax(dax_dev);
1238 return ret;
1239 }
1240
ext2_clear_super_error(struct super_block * sb)1241 static void ext2_clear_super_error(struct super_block *sb)
1242 {
1243 struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1244
1245 if (buffer_write_io_error(sbh)) {
1246 /*
1247 * Oh, dear. A previous attempt to write the
1248 * superblock failed. This could happen because the
1249 * USB device was yanked out. Or it could happen to
1250 * be a transient write error and maybe the block will
1251 * be remapped. Nothing we can do but to retry the
1252 * write and hope for the best.
1253 */
1254 ext2_msg(sb, KERN_ERR,
1255 "previous I/O error to superblock detected");
1256 clear_buffer_write_io_error(sbh);
1257 set_buffer_uptodate(sbh);
1258 }
1259 }
1260
ext2_sync_super(struct super_block * sb,struct ext2_super_block * es,int wait)1261 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1262 int wait)
1263 {
1264 ext2_clear_super_error(sb);
1265 spin_lock(&EXT2_SB(sb)->s_lock);
1266 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1267 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1268 es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1269 /* unlock before we do IO */
1270 spin_unlock(&EXT2_SB(sb)->s_lock);
1271 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1272 if (wait)
1273 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1274 }
1275
1276 /*
1277 * In the second extended file system, it is not necessary to
1278 * write the super block since we use a mapping of the
1279 * disk super block in a buffer.
1280 *
1281 * However, this function is still used to set the fs valid
1282 * flags to 0. We need to set this flag to 0 since the fs
1283 * may have been checked while mounted and e2fsck may have
1284 * set s_state to EXT2_VALID_FS after some corrections.
1285 */
ext2_sync_fs(struct super_block * sb,int wait)1286 static int ext2_sync_fs(struct super_block *sb, int wait)
1287 {
1288 struct ext2_sb_info *sbi = EXT2_SB(sb);
1289 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1290
1291 /*
1292 * Write quota structures to quota file, sync_blockdev() will write
1293 * them to disk later
1294 */
1295 dquot_writeback_dquots(sb, -1);
1296
1297 spin_lock(&sbi->s_lock);
1298 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1299 ext2_debug("setting valid to 0\n");
1300 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1301 }
1302 spin_unlock(&sbi->s_lock);
1303 ext2_sync_super(sb, es, wait);
1304 return 0;
1305 }
1306
ext2_freeze(struct super_block * sb)1307 static int ext2_freeze(struct super_block *sb)
1308 {
1309 struct ext2_sb_info *sbi = EXT2_SB(sb);
1310
1311 /*
1312 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1313 * because we have unattached inodes and thus filesystem is not fully
1314 * consistent.
1315 */
1316 if (atomic_long_read(&sb->s_remove_count)) {
1317 ext2_sync_fs(sb, 1);
1318 return 0;
1319 }
1320 /* Set EXT2_FS_VALID flag */
1321 spin_lock(&sbi->s_lock);
1322 sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1323 spin_unlock(&sbi->s_lock);
1324 ext2_sync_super(sb, sbi->s_es, 1);
1325
1326 return 0;
1327 }
1328
ext2_unfreeze(struct super_block * sb)1329 static int ext2_unfreeze(struct super_block *sb)
1330 {
1331 /* Just write sb to clear EXT2_VALID_FS flag */
1332 ext2_write_super(sb);
1333
1334 return 0;
1335 }
1336
ext2_write_super(struct super_block * sb)1337 static void ext2_write_super(struct super_block *sb)
1338 {
1339 if (!sb_rdonly(sb))
1340 ext2_sync_fs(sb, 1);
1341 }
1342
ext2_remount(struct super_block * sb,int * flags,char * data)1343 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1344 {
1345 struct ext2_sb_info * sbi = EXT2_SB(sb);
1346 struct ext2_super_block * es;
1347 struct ext2_mount_options new_opts;
1348 int err;
1349
1350 sync_filesystem(sb);
1351
1352 spin_lock(&sbi->s_lock);
1353 new_opts.s_mount_opt = sbi->s_mount_opt;
1354 new_opts.s_resuid = sbi->s_resuid;
1355 new_opts.s_resgid = sbi->s_resgid;
1356 spin_unlock(&sbi->s_lock);
1357
1358 if (!parse_options(data, sb, &new_opts))
1359 return -EINVAL;
1360
1361 spin_lock(&sbi->s_lock);
1362 es = sbi->s_es;
1363 if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1364 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1365 "dax flag with busy inodes while remounting");
1366 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1367 }
1368 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1369 goto out_set;
1370 if (*flags & SB_RDONLY) {
1371 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1372 !(sbi->s_mount_state & EXT2_VALID_FS))
1373 goto out_set;
1374
1375 /*
1376 * OK, we are remounting a valid rw partition rdonly, so set
1377 * the rdonly flag and then mark the partition as valid again.
1378 */
1379 es->s_state = cpu_to_le16(sbi->s_mount_state);
1380 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1381 spin_unlock(&sbi->s_lock);
1382
1383 err = dquot_suspend(sb, -1);
1384 if (err < 0)
1385 return err;
1386
1387 ext2_sync_super(sb, es, 1);
1388 } else {
1389 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1390 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1391 if (ret) {
1392 spin_unlock(&sbi->s_lock);
1393 ext2_msg(sb, KERN_WARNING,
1394 "warning: couldn't remount RDWR because of "
1395 "unsupported optional features (%x).",
1396 le32_to_cpu(ret));
1397 return -EROFS;
1398 }
1399 /*
1400 * Mounting a RDONLY partition read-write, so reread and
1401 * store the current valid flag. (It may have been changed
1402 * by e2fsck since we originally mounted the partition.)
1403 */
1404 sbi->s_mount_state = le16_to_cpu(es->s_state);
1405 if (!ext2_setup_super (sb, es, 0))
1406 sb->s_flags &= ~SB_RDONLY;
1407 spin_unlock(&sbi->s_lock);
1408
1409 ext2_write_super(sb);
1410
1411 dquot_resume(sb, -1);
1412 }
1413
1414 spin_lock(&sbi->s_lock);
1415 out_set:
1416 sbi->s_mount_opt = new_opts.s_mount_opt;
1417 sbi->s_resuid = new_opts.s_resuid;
1418 sbi->s_resgid = new_opts.s_resgid;
1419 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1420 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1421 spin_unlock(&sbi->s_lock);
1422
1423 return 0;
1424 }
1425
ext2_statfs(struct dentry * dentry,struct kstatfs * buf)1426 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1427 {
1428 struct super_block *sb = dentry->d_sb;
1429 struct ext2_sb_info *sbi = EXT2_SB(sb);
1430 struct ext2_super_block *es = sbi->s_es;
1431 u64 fsid;
1432
1433 spin_lock(&sbi->s_lock);
1434
1435 if (test_opt (sb, MINIX_DF))
1436 sbi->s_overhead_last = 0;
1437 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1438 unsigned long i, overhead = 0;
1439 smp_rmb();
1440
1441 /*
1442 * Compute the overhead (FS structures). This is constant
1443 * for a given filesystem unless the number of block groups
1444 * changes so we cache the previous value until it does.
1445 */
1446
1447 /*
1448 * All of the blocks before first_data_block are
1449 * overhead
1450 */
1451 overhead = le32_to_cpu(es->s_first_data_block);
1452
1453 /*
1454 * Add the overhead attributed to the superblock and
1455 * block group descriptors. If the sparse superblocks
1456 * feature is turned on, then not all groups have this.
1457 */
1458 for (i = 0; i < sbi->s_groups_count; i++)
1459 overhead += ext2_bg_has_super(sb, i) +
1460 ext2_bg_num_gdb(sb, i);
1461
1462 /*
1463 * Every block group has an inode bitmap, a block
1464 * bitmap, and an inode table.
1465 */
1466 overhead += (sbi->s_groups_count *
1467 (2 + sbi->s_itb_per_group));
1468 sbi->s_overhead_last = overhead;
1469 smp_wmb();
1470 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1471 }
1472
1473 buf->f_type = EXT2_SUPER_MAGIC;
1474 buf->f_bsize = sb->s_blocksize;
1475 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1476 buf->f_bfree = ext2_count_free_blocks(sb);
1477 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1478 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1479 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1480 buf->f_bavail = 0;
1481 buf->f_files = le32_to_cpu(es->s_inodes_count);
1482 buf->f_ffree = ext2_count_free_inodes(sb);
1483 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1484 buf->f_namelen = EXT2_NAME_LEN;
1485 fsid = le64_to_cpup((void *)es->s_uuid) ^
1486 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1487 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1488 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1489 spin_unlock(&sbi->s_lock);
1490 return 0;
1491 }
1492
ext2_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)1493 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1494 int flags, const char *dev_name, void *data)
1495 {
1496 return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1497 }
1498
1499 #ifdef CONFIG_QUOTA
1500
1501 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1502 * acquiring the locks... As quota files are never truncated and quota code
1503 * itself serializes the operations (and no one else should touch the files)
1504 * we don't have to be afraid of races */
ext2_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)1505 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1506 size_t len, loff_t off)
1507 {
1508 struct inode *inode = sb_dqopt(sb)->files[type];
1509 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1510 int err = 0;
1511 int offset = off & (sb->s_blocksize - 1);
1512 int tocopy;
1513 size_t toread;
1514 struct buffer_head tmp_bh;
1515 struct buffer_head *bh;
1516 loff_t i_size = i_size_read(inode);
1517
1518 if (off > i_size)
1519 return 0;
1520 if (off+len > i_size)
1521 len = i_size-off;
1522 toread = len;
1523 while (toread > 0) {
1524 tocopy = sb->s_blocksize - offset < toread ?
1525 sb->s_blocksize - offset : toread;
1526
1527 tmp_bh.b_state = 0;
1528 tmp_bh.b_size = sb->s_blocksize;
1529 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1530 if (err < 0)
1531 return err;
1532 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1533 memset(data, 0, tocopy);
1534 else {
1535 bh = sb_bread(sb, tmp_bh.b_blocknr);
1536 if (!bh)
1537 return -EIO;
1538 memcpy(data, bh->b_data+offset, tocopy);
1539 brelse(bh);
1540 }
1541 offset = 0;
1542 toread -= tocopy;
1543 data += tocopy;
1544 blk++;
1545 }
1546 return len;
1547 }
1548
1549 /* Write to quotafile */
ext2_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)1550 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1551 const char *data, size_t len, loff_t off)
1552 {
1553 struct inode *inode = sb_dqopt(sb)->files[type];
1554 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1555 int err = 0;
1556 int offset = off & (sb->s_blocksize - 1);
1557 int tocopy;
1558 size_t towrite = len;
1559 struct buffer_head tmp_bh;
1560 struct buffer_head *bh;
1561
1562 while (towrite > 0) {
1563 tocopy = sb->s_blocksize - offset < towrite ?
1564 sb->s_blocksize - offset : towrite;
1565
1566 tmp_bh.b_state = 0;
1567 tmp_bh.b_size = sb->s_blocksize;
1568 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1569 if (err < 0)
1570 goto out;
1571 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1572 bh = sb_bread(sb, tmp_bh.b_blocknr);
1573 else
1574 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1575 if (unlikely(!bh)) {
1576 err = -EIO;
1577 goto out;
1578 }
1579 lock_buffer(bh);
1580 memcpy(bh->b_data+offset, data, tocopy);
1581 flush_dcache_page(bh->b_page);
1582 set_buffer_uptodate(bh);
1583 mark_buffer_dirty(bh);
1584 unlock_buffer(bh);
1585 brelse(bh);
1586 offset = 0;
1587 towrite -= tocopy;
1588 data += tocopy;
1589 blk++;
1590 }
1591 out:
1592 if (len == towrite)
1593 return err;
1594 if (inode->i_size < off+len-towrite)
1595 i_size_write(inode, off+len-towrite);
1596 inode_inc_iversion(inode);
1597 inode->i_mtime = inode->i_ctime = current_time(inode);
1598 mark_inode_dirty(inode);
1599 return len - towrite;
1600 }
1601
ext2_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)1602 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1603 const struct path *path)
1604 {
1605 int err;
1606 struct inode *inode;
1607
1608 err = dquot_quota_on(sb, type, format_id, path);
1609 if (err)
1610 return err;
1611
1612 inode = d_inode(path->dentry);
1613 inode_lock(inode);
1614 EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1615 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1616 S_NOATIME | S_IMMUTABLE);
1617 inode_unlock(inode);
1618 mark_inode_dirty(inode);
1619
1620 return 0;
1621 }
1622
ext2_quota_off(struct super_block * sb,int type)1623 static int ext2_quota_off(struct super_block *sb, int type)
1624 {
1625 struct inode *inode = sb_dqopt(sb)->files[type];
1626 int err;
1627
1628 if (!inode || !igrab(inode))
1629 goto out;
1630
1631 err = dquot_quota_off(sb, type);
1632 if (err)
1633 goto out_put;
1634
1635 inode_lock(inode);
1636 EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1637 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1638 inode_unlock(inode);
1639 mark_inode_dirty(inode);
1640 out_put:
1641 iput(inode);
1642 return err;
1643 out:
1644 return dquot_quota_off(sb, type);
1645 }
1646
1647 #endif
1648
1649 static struct file_system_type ext2_fs_type = {
1650 .owner = THIS_MODULE,
1651 .name = "ext2",
1652 .mount = ext2_mount,
1653 .kill_sb = kill_block_super,
1654 .fs_flags = FS_REQUIRES_DEV,
1655 };
1656 MODULE_ALIAS_FS("ext2");
1657
init_ext2_fs(void)1658 static int __init init_ext2_fs(void)
1659 {
1660 int err;
1661
1662 err = init_inodecache();
1663 if (err)
1664 return err;
1665 err = register_filesystem(&ext2_fs_type);
1666 if (err)
1667 goto out;
1668 return 0;
1669 out:
1670 destroy_inodecache();
1671 return err;
1672 }
1673
exit_ext2_fs(void)1674 static void __exit exit_ext2_fs(void)
1675 {
1676 unregister_filesystem(&ext2_fs_type);
1677 destroy_inodecache();
1678 }
1679
1680 MODULE_AUTHOR("Remy Card and others");
1681 MODULE_DESCRIPTION("Second Extended Filesystem");
1682 MODULE_LICENSE("GPL");
1683 module_init(init_ext2_fs)
1684 module_exit(exit_ext2_fs)
1685