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, fs=%lu, gc=%lu, "
695 "bpg=%lu, ipg=%lu, mo=%04lx]",
696 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
697 sbi->s_frag_size,
698 sbi->s_groups_count,
699 EXT2_BLOCKS_PER_GROUP(sb),
700 EXT2_INODES_PER_GROUP(sb),
701 sbi->s_mount_opt);
702 return res;
703 }
704
ext2_check_descriptors(struct super_block * sb)705 static int ext2_check_descriptors(struct super_block *sb)
706 {
707 int i;
708 struct ext2_sb_info *sbi = EXT2_SB(sb);
709
710 ext2_debug ("Checking group descriptors");
711
712 for (i = 0; i < sbi->s_groups_count; i++) {
713 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
714 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
715 ext2_fsblk_t last_block;
716
717 if (i == sbi->s_groups_count - 1)
718 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
719 else
720 last_block = first_block +
721 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
722
723 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
724 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
725 {
726 ext2_error (sb, "ext2_check_descriptors",
727 "Block bitmap for group %d"
728 " not in group (block %lu)!",
729 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
730 return 0;
731 }
732 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
733 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
734 {
735 ext2_error (sb, "ext2_check_descriptors",
736 "Inode bitmap for group %d"
737 " not in group (block %lu)!",
738 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
739 return 0;
740 }
741 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
742 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
743 last_block)
744 {
745 ext2_error (sb, "ext2_check_descriptors",
746 "Inode table for group %d"
747 " not in group (block %lu)!",
748 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
749 return 0;
750 }
751 }
752 return 1;
753 }
754
755 /*
756 * Maximal file size. There is a direct, and {,double-,triple-}indirect
757 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
758 * We need to be 1 filesystem block less than the 2^32 sector limit.
759 */
ext2_max_size(int bits)760 static loff_t ext2_max_size(int bits)
761 {
762 loff_t res = EXT2_NDIR_BLOCKS;
763 int meta_blocks;
764 unsigned int upper_limit;
765 unsigned int ppb = 1 << (bits-2);
766
767 /* This is calculated to be the largest file size for a
768 * dense, file such that the total number of
769 * sectors in the file, including data and all indirect blocks,
770 * does not exceed 2^32 -1
771 * __u32 i_blocks representing the total number of
772 * 512 bytes blocks of the file
773 */
774 upper_limit = (1LL << 32) - 1;
775
776 /* total blocks in file system block size */
777 upper_limit >>= (bits - 9);
778
779 /* Compute how many blocks we can address by block tree */
780 res += 1LL << (bits-2);
781 res += 1LL << (2*(bits-2));
782 res += 1LL << (3*(bits-2));
783 /* Does block tree limit file size? */
784 if (res < upper_limit)
785 goto check_lfs;
786
787 res = upper_limit;
788 /* How many metadata blocks are needed for addressing upper_limit? */
789 upper_limit -= EXT2_NDIR_BLOCKS;
790 /* indirect blocks */
791 meta_blocks = 1;
792 upper_limit -= ppb;
793 /* double indirect blocks */
794 if (upper_limit < ppb * ppb) {
795 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
796 res -= meta_blocks;
797 goto check_lfs;
798 }
799 meta_blocks += 1 + ppb;
800 upper_limit -= ppb * ppb;
801 /* tripple indirect blocks for the rest */
802 meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
803 DIV_ROUND_UP(upper_limit, ppb*ppb);
804 res -= meta_blocks;
805 check_lfs:
806 res <<= bits;
807 if (res > MAX_LFS_FILESIZE)
808 res = MAX_LFS_FILESIZE;
809
810 return res;
811 }
812
descriptor_loc(struct super_block * sb,unsigned long logic_sb_block,int nr)813 static unsigned long descriptor_loc(struct super_block *sb,
814 unsigned long logic_sb_block,
815 int nr)
816 {
817 struct ext2_sb_info *sbi = EXT2_SB(sb);
818 unsigned long bg, first_meta_bg;
819 int has_super = 0;
820
821 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
822
823 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
824 nr < first_meta_bg)
825 return (logic_sb_block + nr + 1);
826 bg = sbi->s_desc_per_block * nr;
827 if (ext2_bg_has_super(sb, bg))
828 has_super = 1;
829
830 return ext2_group_first_block_no(sb, bg) + has_super;
831 }
832
ext2_fill_super(struct super_block * sb,void * data,int silent)833 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
834 {
835 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
836 struct buffer_head * bh;
837 struct ext2_sb_info * sbi;
838 struct ext2_super_block * es;
839 struct inode *root;
840 unsigned long block;
841 unsigned long sb_block = get_sb_block(&data);
842 unsigned long logic_sb_block;
843 unsigned long offset = 0;
844 unsigned long def_mount_opts;
845 long ret = -ENOMEM;
846 int blocksize = BLOCK_SIZE;
847 int db_count;
848 int i, j;
849 __le32 features;
850 int err;
851 struct ext2_mount_options opts;
852
853 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
854 if (!sbi)
855 goto failed;
856
857 sbi->s_blockgroup_lock =
858 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
859 if (!sbi->s_blockgroup_lock) {
860 kfree(sbi);
861 goto failed;
862 }
863 sb->s_fs_info = sbi;
864 sbi->s_sb_block = sb_block;
865 sbi->s_daxdev = dax_dev;
866
867 spin_lock_init(&sbi->s_lock);
868 ret = -EINVAL;
869
870 /*
871 * See what the current blocksize for the device is, and
872 * use that as the blocksize. Otherwise (or if the blocksize
873 * is smaller than the default) use the default.
874 * This is important for devices that have a hardware
875 * sectorsize that is larger than the default.
876 */
877 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
878 if (!blocksize) {
879 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
880 goto failed_sbi;
881 }
882
883 /*
884 * If the superblock doesn't start on a hardware sector boundary,
885 * calculate the offset.
886 */
887 if (blocksize != BLOCK_SIZE) {
888 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
889 offset = (sb_block*BLOCK_SIZE) % blocksize;
890 } else {
891 logic_sb_block = sb_block;
892 }
893
894 if (!(bh = sb_bread(sb, logic_sb_block))) {
895 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
896 goto failed_sbi;
897 }
898 /*
899 * Note: s_es must be initialized as soon as possible because
900 * some ext2 macro-instructions depend on its value
901 */
902 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
903 sbi->s_es = es;
904 sb->s_magic = le16_to_cpu(es->s_magic);
905
906 if (sb->s_magic != EXT2_SUPER_MAGIC)
907 goto cantfind_ext2;
908
909 opts.s_mount_opt = 0;
910 /* Set defaults before we parse the mount options */
911 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
912 if (def_mount_opts & EXT2_DEFM_DEBUG)
913 set_opt(opts.s_mount_opt, DEBUG);
914 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
915 set_opt(opts.s_mount_opt, GRPID);
916 if (def_mount_opts & EXT2_DEFM_UID16)
917 set_opt(opts.s_mount_opt, NO_UID32);
918 #ifdef CONFIG_EXT2_FS_XATTR
919 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
920 set_opt(opts.s_mount_opt, XATTR_USER);
921 #endif
922 #ifdef CONFIG_EXT2_FS_POSIX_ACL
923 if (def_mount_opts & EXT2_DEFM_ACL)
924 set_opt(opts.s_mount_opt, POSIX_ACL);
925 #endif
926
927 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
928 set_opt(opts.s_mount_opt, ERRORS_PANIC);
929 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
930 set_opt(opts.s_mount_opt, ERRORS_CONT);
931 else
932 set_opt(opts.s_mount_opt, ERRORS_RO);
933
934 opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
935 opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
936
937 set_opt(opts.s_mount_opt, RESERVATION);
938
939 if (!parse_options((char *) data, sb, &opts))
940 goto failed_mount;
941
942 sbi->s_mount_opt = opts.s_mount_opt;
943 sbi->s_resuid = opts.s_resuid;
944 sbi->s_resgid = opts.s_resgid;
945
946 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
947 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
948 SB_POSIXACL : 0);
949 sb->s_iflags |= SB_I_CGROUPWB;
950
951 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
952 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
953 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
954 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
955 ext2_msg(sb, KERN_WARNING,
956 "warning: feature flags set on rev 0 fs, "
957 "running e2fsck is recommended");
958 /*
959 * Check feature flags regardless of the revision level, since we
960 * previously didn't change the revision level when setting the flags,
961 * so there is a chance incompat flags are set on a rev 0 filesystem.
962 */
963 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
964 if (features) {
965 ext2_msg(sb, KERN_ERR, "error: couldn't mount because of "
966 "unsupported optional features (%x)",
967 le32_to_cpu(features));
968 goto failed_mount;
969 }
970 if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
971 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
972 "unsupported optional features (%x)",
973 le32_to_cpu(features));
974 goto failed_mount;
975 }
976
977 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
978
979 if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
980 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
981 ext2_msg(sb, KERN_ERR,
982 "DAX unsupported by block device. Turning off DAX.");
983 sbi->s_mount_opt &= ~EXT2_MOUNT_DAX;
984 }
985 }
986
987 /* If the blocksize doesn't match, re-read the thing.. */
988 if (sb->s_blocksize != blocksize) {
989 brelse(bh);
990
991 if (!sb_set_blocksize(sb, blocksize)) {
992 ext2_msg(sb, KERN_ERR,
993 "error: bad blocksize %d", blocksize);
994 goto failed_sbi;
995 }
996
997 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
998 offset = (sb_block*BLOCK_SIZE) % blocksize;
999 bh = sb_bread(sb, logic_sb_block);
1000 if(!bh) {
1001 ext2_msg(sb, KERN_ERR, "error: couldn't read"
1002 "superblock on 2nd try");
1003 goto failed_sbi;
1004 }
1005 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
1006 sbi->s_es = es;
1007 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
1008 ext2_msg(sb, KERN_ERR, "error: magic mismatch");
1009 goto failed_mount;
1010 }
1011 }
1012
1013 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
1014 sb->s_max_links = EXT2_LINK_MAX;
1015
1016 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
1017 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
1018 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
1019 } else {
1020 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1021 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1022 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
1023 !is_power_of_2(sbi->s_inode_size) ||
1024 (sbi->s_inode_size > blocksize)) {
1025 ext2_msg(sb, KERN_ERR,
1026 "error: unsupported inode size: %d",
1027 sbi->s_inode_size);
1028 goto failed_mount;
1029 }
1030 }
1031
1032 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1033 le32_to_cpu(es->s_log_frag_size);
1034 if (sbi->s_frag_size == 0)
1035 goto cantfind_ext2;
1036 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1037
1038 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1039 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1040 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1041
1042 if (EXT2_INODE_SIZE(sb) == 0)
1043 goto cantfind_ext2;
1044 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1045 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1046 goto cantfind_ext2;
1047 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1048 sbi->s_inodes_per_block;
1049 sbi->s_desc_per_block = sb->s_blocksize /
1050 sizeof (struct ext2_group_desc);
1051 sbi->s_sbh = bh;
1052 sbi->s_mount_state = le16_to_cpu(es->s_state);
1053 sbi->s_addr_per_block_bits =
1054 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1055 sbi->s_desc_per_block_bits =
1056 ilog2 (EXT2_DESC_PER_BLOCK(sb));
1057
1058 if (sb->s_magic != EXT2_SUPER_MAGIC)
1059 goto cantfind_ext2;
1060
1061 if (sb->s_blocksize != bh->b_size) {
1062 if (!silent)
1063 ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1064 goto failed_mount;
1065 }
1066
1067 if (sb->s_blocksize != sbi->s_frag_size) {
1068 ext2_msg(sb, KERN_ERR,
1069 "error: fragsize %lu != blocksize %lu"
1070 "(not supported yet)",
1071 sbi->s_frag_size, sb->s_blocksize);
1072 goto failed_mount;
1073 }
1074
1075 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1076 ext2_msg(sb, KERN_ERR,
1077 "error: #blocks per group too big: %lu",
1078 sbi->s_blocks_per_group);
1079 goto failed_mount;
1080 }
1081 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1082 ext2_msg(sb, KERN_ERR,
1083 "error: #fragments per group too big: %lu",
1084 sbi->s_frags_per_group);
1085 goto failed_mount;
1086 }
1087 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1088 ext2_msg(sb, KERN_ERR,
1089 "error: #inodes per group too big: %lu",
1090 sbi->s_inodes_per_group);
1091 goto failed_mount;
1092 }
1093
1094 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1095 goto cantfind_ext2;
1096 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1097 le32_to_cpu(es->s_first_data_block) - 1)
1098 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1099 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1100 EXT2_DESC_PER_BLOCK(sb);
1101 sbi->s_group_desc = kmalloc_array (db_count,
1102 sizeof(struct buffer_head *),
1103 GFP_KERNEL);
1104 if (sbi->s_group_desc == NULL) {
1105 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1106 goto failed_mount;
1107 }
1108 bgl_lock_init(sbi->s_blockgroup_lock);
1109 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1110 if (!sbi->s_debts) {
1111 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1112 goto failed_mount_group_desc;
1113 }
1114 for (i = 0; i < db_count; i++) {
1115 block = descriptor_loc(sb, logic_sb_block, i);
1116 sbi->s_group_desc[i] = sb_bread(sb, block);
1117 if (!sbi->s_group_desc[i]) {
1118 for (j = 0; j < i; j++)
1119 brelse (sbi->s_group_desc[j]);
1120 ext2_msg(sb, KERN_ERR,
1121 "error: unable to read group descriptors");
1122 goto failed_mount_group_desc;
1123 }
1124 }
1125 if (!ext2_check_descriptors (sb)) {
1126 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1127 goto failed_mount2;
1128 }
1129 sbi->s_gdb_count = db_count;
1130 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1131 spin_lock_init(&sbi->s_next_gen_lock);
1132
1133 /* per fileystem reservation list head & lock */
1134 spin_lock_init(&sbi->s_rsv_window_lock);
1135 sbi->s_rsv_window_root = RB_ROOT;
1136 /*
1137 * Add a single, static dummy reservation to the start of the
1138 * reservation window list --- it gives us a placeholder for
1139 * append-at-start-of-list which makes the allocation logic
1140 * _much_ simpler.
1141 */
1142 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1143 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1144 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1145 sbi->s_rsv_window_head.rsv_goal_size = 0;
1146 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1147
1148 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1149 ext2_count_free_blocks(sb), GFP_KERNEL);
1150 if (!err) {
1151 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1152 ext2_count_free_inodes(sb), GFP_KERNEL);
1153 }
1154 if (!err) {
1155 err = percpu_counter_init(&sbi->s_dirs_counter,
1156 ext2_count_dirs(sb), GFP_KERNEL);
1157 }
1158 if (err) {
1159 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1160 goto failed_mount3;
1161 }
1162
1163 #ifdef CONFIG_EXT2_FS_XATTR
1164 sbi->s_ea_block_cache = ext2_xattr_create_cache();
1165 if (!sbi->s_ea_block_cache) {
1166 ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1167 goto failed_mount3;
1168 }
1169 #endif
1170 /*
1171 * set up enough so that it can read an inode
1172 */
1173 sb->s_op = &ext2_sops;
1174 sb->s_export_op = &ext2_export_ops;
1175 sb->s_xattr = ext2_xattr_handlers;
1176
1177 #ifdef CONFIG_QUOTA
1178 sb->dq_op = &dquot_operations;
1179 sb->s_qcop = &ext2_quotactl_ops;
1180 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1181 #endif
1182
1183 root = ext2_iget(sb, EXT2_ROOT_INO);
1184 if (IS_ERR(root)) {
1185 ret = PTR_ERR(root);
1186 goto failed_mount3;
1187 }
1188 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1189 iput(root);
1190 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1191 goto failed_mount3;
1192 }
1193
1194 sb->s_root = d_make_root(root);
1195 if (!sb->s_root) {
1196 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1197 ret = -ENOMEM;
1198 goto failed_mount3;
1199 }
1200 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1201 ext2_msg(sb, KERN_WARNING,
1202 "warning: mounting ext3 filesystem as ext2");
1203 if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1204 sb->s_flags |= SB_RDONLY;
1205 ext2_write_super(sb);
1206 return 0;
1207
1208 cantfind_ext2:
1209 if (!silent)
1210 ext2_msg(sb, KERN_ERR,
1211 "error: can't find an ext2 filesystem on dev %s.",
1212 sb->s_id);
1213 goto failed_mount;
1214 failed_mount3:
1215 if (sbi->s_ea_block_cache)
1216 ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1217 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1218 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1219 percpu_counter_destroy(&sbi->s_dirs_counter);
1220 failed_mount2:
1221 for (i = 0; i < db_count; i++)
1222 brelse(sbi->s_group_desc[i]);
1223 failed_mount_group_desc:
1224 kfree(sbi->s_group_desc);
1225 kfree(sbi->s_debts);
1226 failed_mount:
1227 brelse(bh);
1228 failed_sbi:
1229 sb->s_fs_info = NULL;
1230 kfree(sbi->s_blockgroup_lock);
1231 kfree(sbi);
1232 failed:
1233 fs_put_dax(dax_dev);
1234 return ret;
1235 }
1236
ext2_clear_super_error(struct super_block * sb)1237 static void ext2_clear_super_error(struct super_block *sb)
1238 {
1239 struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1240
1241 if (buffer_write_io_error(sbh)) {
1242 /*
1243 * Oh, dear. A previous attempt to write the
1244 * superblock failed. This could happen because the
1245 * USB device was yanked out. Or it could happen to
1246 * be a transient write error and maybe the block will
1247 * be remapped. Nothing we can do but to retry the
1248 * write and hope for the best.
1249 */
1250 ext2_msg(sb, KERN_ERR,
1251 "previous I/O error to superblock detected");
1252 clear_buffer_write_io_error(sbh);
1253 set_buffer_uptodate(sbh);
1254 }
1255 }
1256
ext2_sync_super(struct super_block * sb,struct ext2_super_block * es,int wait)1257 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1258 int wait)
1259 {
1260 ext2_clear_super_error(sb);
1261 spin_lock(&EXT2_SB(sb)->s_lock);
1262 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1263 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1264 es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1265 /* unlock before we do IO */
1266 spin_unlock(&EXT2_SB(sb)->s_lock);
1267 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1268 if (wait)
1269 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1270 }
1271
1272 /*
1273 * In the second extended file system, it is not necessary to
1274 * write the super block since we use a mapping of the
1275 * disk super block in a buffer.
1276 *
1277 * However, this function is still used to set the fs valid
1278 * flags to 0. We need to set this flag to 0 since the fs
1279 * may have been checked while mounted and e2fsck may have
1280 * set s_state to EXT2_VALID_FS after some corrections.
1281 */
ext2_sync_fs(struct super_block * sb,int wait)1282 static int ext2_sync_fs(struct super_block *sb, int wait)
1283 {
1284 struct ext2_sb_info *sbi = EXT2_SB(sb);
1285 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1286
1287 /*
1288 * Write quota structures to quota file, sync_blockdev() will write
1289 * them to disk later
1290 */
1291 dquot_writeback_dquots(sb, -1);
1292
1293 spin_lock(&sbi->s_lock);
1294 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1295 ext2_debug("setting valid to 0\n");
1296 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1297 }
1298 spin_unlock(&sbi->s_lock);
1299 ext2_sync_super(sb, es, wait);
1300 return 0;
1301 }
1302
ext2_freeze(struct super_block * sb)1303 static int ext2_freeze(struct super_block *sb)
1304 {
1305 struct ext2_sb_info *sbi = EXT2_SB(sb);
1306
1307 /*
1308 * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1309 * because we have unattached inodes and thus filesystem is not fully
1310 * consistent.
1311 */
1312 if (atomic_long_read(&sb->s_remove_count)) {
1313 ext2_sync_fs(sb, 1);
1314 return 0;
1315 }
1316 /* Set EXT2_FS_VALID flag */
1317 spin_lock(&sbi->s_lock);
1318 sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1319 spin_unlock(&sbi->s_lock);
1320 ext2_sync_super(sb, sbi->s_es, 1);
1321
1322 return 0;
1323 }
1324
ext2_unfreeze(struct super_block * sb)1325 static int ext2_unfreeze(struct super_block *sb)
1326 {
1327 /* Just write sb to clear EXT2_VALID_FS flag */
1328 ext2_write_super(sb);
1329
1330 return 0;
1331 }
1332
ext2_write_super(struct super_block * sb)1333 static void ext2_write_super(struct super_block *sb)
1334 {
1335 if (!sb_rdonly(sb))
1336 ext2_sync_fs(sb, 1);
1337 }
1338
ext2_remount(struct super_block * sb,int * flags,char * data)1339 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1340 {
1341 struct ext2_sb_info * sbi = EXT2_SB(sb);
1342 struct ext2_super_block * es;
1343 struct ext2_mount_options new_opts;
1344 int err;
1345
1346 sync_filesystem(sb);
1347
1348 spin_lock(&sbi->s_lock);
1349 new_opts.s_mount_opt = sbi->s_mount_opt;
1350 new_opts.s_resuid = sbi->s_resuid;
1351 new_opts.s_resgid = sbi->s_resgid;
1352 spin_unlock(&sbi->s_lock);
1353
1354 if (!parse_options(data, sb, &new_opts))
1355 return -EINVAL;
1356
1357 spin_lock(&sbi->s_lock);
1358 es = sbi->s_es;
1359 if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1360 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1361 "dax flag with busy inodes while remounting");
1362 new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1363 }
1364 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1365 goto out_set;
1366 if (*flags & SB_RDONLY) {
1367 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1368 !(sbi->s_mount_state & EXT2_VALID_FS))
1369 goto out_set;
1370
1371 /*
1372 * OK, we are remounting a valid rw partition rdonly, so set
1373 * the rdonly flag and then mark the partition as valid again.
1374 */
1375 es->s_state = cpu_to_le16(sbi->s_mount_state);
1376 es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1377 spin_unlock(&sbi->s_lock);
1378
1379 err = dquot_suspend(sb, -1);
1380 if (err < 0)
1381 return err;
1382
1383 ext2_sync_super(sb, es, 1);
1384 } else {
1385 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1386 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1387 if (ret) {
1388 spin_unlock(&sbi->s_lock);
1389 ext2_msg(sb, KERN_WARNING,
1390 "warning: couldn't remount RDWR because of "
1391 "unsupported optional features (%x).",
1392 le32_to_cpu(ret));
1393 return -EROFS;
1394 }
1395 /*
1396 * Mounting a RDONLY partition read-write, so reread and
1397 * store the current valid flag. (It may have been changed
1398 * by e2fsck since we originally mounted the partition.)
1399 */
1400 sbi->s_mount_state = le16_to_cpu(es->s_state);
1401 if (!ext2_setup_super (sb, es, 0))
1402 sb->s_flags &= ~SB_RDONLY;
1403 spin_unlock(&sbi->s_lock);
1404
1405 ext2_write_super(sb);
1406
1407 dquot_resume(sb, -1);
1408 }
1409
1410 spin_lock(&sbi->s_lock);
1411 out_set:
1412 sbi->s_mount_opt = new_opts.s_mount_opt;
1413 sbi->s_resuid = new_opts.s_resuid;
1414 sbi->s_resgid = new_opts.s_resgid;
1415 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1416 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1417 spin_unlock(&sbi->s_lock);
1418
1419 return 0;
1420 }
1421
ext2_statfs(struct dentry * dentry,struct kstatfs * buf)1422 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1423 {
1424 struct super_block *sb = dentry->d_sb;
1425 struct ext2_sb_info *sbi = EXT2_SB(sb);
1426 struct ext2_super_block *es = sbi->s_es;
1427 u64 fsid;
1428
1429 spin_lock(&sbi->s_lock);
1430
1431 if (test_opt (sb, MINIX_DF))
1432 sbi->s_overhead_last = 0;
1433 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1434 unsigned long i, overhead = 0;
1435 smp_rmb();
1436
1437 /*
1438 * Compute the overhead (FS structures). This is constant
1439 * for a given filesystem unless the number of block groups
1440 * changes so we cache the previous value until it does.
1441 */
1442
1443 /*
1444 * All of the blocks before first_data_block are
1445 * overhead
1446 */
1447 overhead = le32_to_cpu(es->s_first_data_block);
1448
1449 /*
1450 * Add the overhead attributed to the superblock and
1451 * block group descriptors. If the sparse superblocks
1452 * feature is turned on, then not all groups have this.
1453 */
1454 for (i = 0; i < sbi->s_groups_count; i++)
1455 overhead += ext2_bg_has_super(sb, i) +
1456 ext2_bg_num_gdb(sb, i);
1457
1458 /*
1459 * Every block group has an inode bitmap, a block
1460 * bitmap, and an inode table.
1461 */
1462 overhead += (sbi->s_groups_count *
1463 (2 + sbi->s_itb_per_group));
1464 sbi->s_overhead_last = overhead;
1465 smp_wmb();
1466 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1467 }
1468
1469 buf->f_type = EXT2_SUPER_MAGIC;
1470 buf->f_bsize = sb->s_blocksize;
1471 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1472 buf->f_bfree = ext2_count_free_blocks(sb);
1473 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1474 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1475 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1476 buf->f_bavail = 0;
1477 buf->f_files = le32_to_cpu(es->s_inodes_count);
1478 buf->f_ffree = ext2_count_free_inodes(sb);
1479 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1480 buf->f_namelen = EXT2_NAME_LEN;
1481 fsid = le64_to_cpup((void *)es->s_uuid) ^
1482 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1483 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1484 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1485 spin_unlock(&sbi->s_lock);
1486 return 0;
1487 }
1488
ext2_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)1489 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1490 int flags, const char *dev_name, void *data)
1491 {
1492 return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1493 }
1494
1495 #ifdef CONFIG_QUOTA
1496
1497 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1498 * acquiring the locks... As quota files are never truncated and quota code
1499 * itself serializes the operations (and no one else should touch the files)
1500 * 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)1501 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1502 size_t len, loff_t off)
1503 {
1504 struct inode *inode = sb_dqopt(sb)->files[type];
1505 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1506 int err = 0;
1507 int offset = off & (sb->s_blocksize - 1);
1508 int tocopy;
1509 size_t toread;
1510 struct buffer_head tmp_bh;
1511 struct buffer_head *bh;
1512 loff_t i_size = i_size_read(inode);
1513
1514 if (off > i_size)
1515 return 0;
1516 if (off+len > i_size)
1517 len = i_size-off;
1518 toread = len;
1519 while (toread > 0) {
1520 tocopy = sb->s_blocksize - offset < toread ?
1521 sb->s_blocksize - offset : toread;
1522
1523 tmp_bh.b_state = 0;
1524 tmp_bh.b_size = sb->s_blocksize;
1525 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1526 if (err < 0)
1527 return err;
1528 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1529 memset(data, 0, tocopy);
1530 else {
1531 bh = sb_bread(sb, tmp_bh.b_blocknr);
1532 if (!bh)
1533 return -EIO;
1534 memcpy(data, bh->b_data+offset, tocopy);
1535 brelse(bh);
1536 }
1537 offset = 0;
1538 toread -= tocopy;
1539 data += tocopy;
1540 blk++;
1541 }
1542 return len;
1543 }
1544
1545 /* Write to quotafile */
ext2_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)1546 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1547 const char *data, size_t len, loff_t off)
1548 {
1549 struct inode *inode = sb_dqopt(sb)->files[type];
1550 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1551 int err = 0;
1552 int offset = off & (sb->s_blocksize - 1);
1553 int tocopy;
1554 size_t towrite = len;
1555 struct buffer_head tmp_bh;
1556 struct buffer_head *bh;
1557
1558 while (towrite > 0) {
1559 tocopy = sb->s_blocksize - offset < towrite ?
1560 sb->s_blocksize - offset : towrite;
1561
1562 tmp_bh.b_state = 0;
1563 tmp_bh.b_size = sb->s_blocksize;
1564 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1565 if (err < 0)
1566 goto out;
1567 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1568 bh = sb_bread(sb, tmp_bh.b_blocknr);
1569 else
1570 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1571 if (unlikely(!bh)) {
1572 err = -EIO;
1573 goto out;
1574 }
1575 lock_buffer(bh);
1576 memcpy(bh->b_data+offset, data, tocopy);
1577 flush_dcache_page(bh->b_page);
1578 set_buffer_uptodate(bh);
1579 mark_buffer_dirty(bh);
1580 unlock_buffer(bh);
1581 brelse(bh);
1582 offset = 0;
1583 towrite -= tocopy;
1584 data += tocopy;
1585 blk++;
1586 }
1587 out:
1588 if (len == towrite)
1589 return err;
1590 if (inode->i_size < off+len-towrite)
1591 i_size_write(inode, off+len-towrite);
1592 inode_inc_iversion(inode);
1593 inode->i_mtime = inode->i_ctime = current_time(inode);
1594 mark_inode_dirty(inode);
1595 return len - towrite;
1596 }
1597
ext2_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)1598 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1599 const struct path *path)
1600 {
1601 int err;
1602 struct inode *inode;
1603
1604 err = dquot_quota_on(sb, type, format_id, path);
1605 if (err)
1606 return err;
1607
1608 inode = d_inode(path->dentry);
1609 inode_lock(inode);
1610 EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1611 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1612 S_NOATIME | S_IMMUTABLE);
1613 inode_unlock(inode);
1614 mark_inode_dirty(inode);
1615
1616 return 0;
1617 }
1618
ext2_quota_off(struct super_block * sb,int type)1619 static int ext2_quota_off(struct super_block *sb, int type)
1620 {
1621 struct inode *inode = sb_dqopt(sb)->files[type];
1622 int err;
1623
1624 if (!inode || !igrab(inode))
1625 goto out;
1626
1627 err = dquot_quota_off(sb, type);
1628 if (err)
1629 goto out_put;
1630
1631 inode_lock(inode);
1632 EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1633 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1634 inode_unlock(inode);
1635 mark_inode_dirty(inode);
1636 out_put:
1637 iput(inode);
1638 return err;
1639 out:
1640 return dquot_quota_off(sb, type);
1641 }
1642
1643 #endif
1644
1645 static struct file_system_type ext2_fs_type = {
1646 .owner = THIS_MODULE,
1647 .name = "ext2",
1648 .mount = ext2_mount,
1649 .kill_sb = kill_block_super,
1650 .fs_flags = FS_REQUIRES_DEV,
1651 };
1652 MODULE_ALIAS_FS("ext2");
1653
init_ext2_fs(void)1654 static int __init init_ext2_fs(void)
1655 {
1656 int err;
1657
1658 err = init_inodecache();
1659 if (err)
1660 return err;
1661 err = register_filesystem(&ext2_fs_type);
1662 if (err)
1663 goto out;
1664 return 0;
1665 out:
1666 destroy_inodecache();
1667 return err;
1668 }
1669
exit_ext2_fs(void)1670 static void __exit exit_ext2_fs(void)
1671 {
1672 unregister_filesystem(&ext2_fs_type);
1673 destroy_inodecache();
1674 }
1675
1676 MODULE_AUTHOR("Remy Card and others");
1677 MODULE_DESCRIPTION("Second Extended Filesystem");
1678 MODULE_LICENSE("GPL");
1679 module_init(init_ext2_fs)
1680 module_exit(exit_ext2_fs)
1681