]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: export iov_shorten from kernel for ext4's use
[mirror_ubuntu-bionic-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 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/time.h>
dab291af 23#include <linux/jbd2.h>
617ba13b 24#include <linux/ext4_fs.h>
dab291af 25#include <linux/ext4_jbd2.h>
ac27a0ec
DK		 26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
30#include <linux/smp_lock.h>
31#include <linux/buffer_head.h>
a5694255 32#include <linux/exportfs.h>
ac27a0ec
DK		 33#include <linux/vfs.h>
34#include <linux/random.h>
35#include <linux/mount.h>
36#include <linux/namei.h>
37#include <linux/quotaops.h>
38#include <linux/seq_file.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ac27a0ec
DK		 41
42#include <asm/uaccess.h>
43
44#include "xattr.h"
45#include "acl.h"
46#include "namei.h"
717d50e4 47#include "group.h"
ac27a0ec 48
617ba13b 49static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 50 unsigned long journal_devnum);
617ba13b 51static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 52 unsigned int);
617ba13b
MC		 53static void ext4_commit_super (struct super_block * sb,
54 struct ext4_super_block * es,
ac27a0ec 55 int sync);
617ba13b
MC		 56static void ext4_mark_recovery_complete(struct super_block * sb,
57 struct ext4_super_block * es);
58static void ext4_clear_journal_err(struct super_block * sb,
59 struct ext4_super_block * es);
60static int ext4_sync_fs(struct super_block *sb, int wait);
61static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec 62 char nbuf[16]);
617ba13b
MC		 63static int ext4_remount (struct super_block * sb, int * flags, char * data);
64static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
65static void ext4_unlockfs(struct super_block *sb);
66static void ext4_write_super (struct super_block * sb);
67static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 68
bd81d8ee 69
8fadc143
AR		 70ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 struct ext4_group_desc *bg)
bd81d8ee 72{
3a14589c 73 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 74 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 75 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 76}
77
8fadc143
AR		 78ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 struct ext4_group_desc *bg)
bd81d8ee 80{
5272f837 81 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 82 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 83 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 84}
85
8fadc143
AR		 86ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 struct ext4_group_desc *bg)
bd81d8ee 88{
5272f837 89 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 90 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 91 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 92}
93
8fadc143
AR		 94void ext4_block_bitmap_set(struct super_block *sb,
95 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 96{
3a14589c 97 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 98 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
99 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 100}
101
8fadc143
AR		 102void ext4_inode_bitmap_set(struct super_block *sb,
103 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 104{
5272f837 105 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 106 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
107 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 108}
109
8fadc143
AR		 110void ext4_inode_table_set(struct super_block *sb,
111 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 112{
5272f837 113 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 114 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
115 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 116}
117
ac27a0ec 118/*
dab291af 119 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 120 *
121 * The only special thing we need to do here is to make sure that all
122 * journal_end calls result in the superblock being marked dirty, so
123 * that sync() will call the filesystem's write_super callback if
124 * appropriate.
125 */
617ba13b 126handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 127{
128 journal_t *journal;
129
130 if (sb->s_flags & MS_RDONLY)
131 return ERR_PTR(-EROFS);
132
133 /* Special case here: if the journal has aborted behind our
134 * backs (eg. EIO in the commit thread), then we still need to
135 * take the FS itself readonly cleanly. */
617ba13b 136 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 137 if (is_journal_aborted(journal)) {
617ba13b 138 ext4_abort(sb, __FUNCTION__,
ac27a0ec
DK		 139 "Detected aborted journal");
140 return ERR_PTR(-EROFS);
141 }
142
dab291af 143 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 144}
145
146/*
147 * The only special thing we need to do here is to make sure that all
dab291af 148 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 149 * that sync() will call the filesystem's write_super callback if
150 * appropriate.
151 */
617ba13b 152int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 153{
154 struct super_block *sb;
155 int err;
156 int rc;
157
158 sb = handle->h_transaction->t_journal->j_private;
159 err = handle->h_err;
dab291af 160 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 161
162 if (!err)
163 err = rc;
164 if (err)
617ba13b 165 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 166 return err;
167}
168
617ba13b 169void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 170 struct buffer_head *bh, handle_t *handle, int err)
171{
172 char nbuf[16];
617ba13b 173 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 174
175 if (bh)
176 BUFFER_TRACE(bh, "abort");
177
178 if (!handle->h_err)
179 handle->h_err = err;
180
181 if (is_handle_aborted(handle))
182 return;
183
184 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
185 caller, errstr, err_fn);
186
dab291af 187 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 188}
189
190/* Deal with the reporting of failure conditions on a filesystem such as
191 * inconsistencies detected or read IO failures.
192 *
193 * On ext2, we can store the error state of the filesystem in the
617ba13b 194 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 195 * write ordering constraints on the superblock which prevent us from
196 * writing it out straight away; and given that the journal is about to
197 * be aborted, we can't rely on the current, or future, transactions to
198 * write out the superblock safely.
199 *
dab291af 200 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 201 * the journal instead. On recovery, the journal will compain about
202 * that error until we've noted it down and cleared it.
203 */
204
617ba13b 205static void ext4_handle_error(struct super_block *sb)
ac27a0ec 206{
617ba13b 207 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 208
617ba13b
MC		 209 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
210 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 211
212 if (sb->s_flags & MS_RDONLY)
213 return;
214
215 if (!test_opt (sb, ERRORS_CONT)) {
617ba13b 216 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 217
617ba13b 218 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 219 if (journal)
dab291af 220 jbd2_journal_abort(journal, -EIO);
ac27a0ec
DK		 221 }
222 if (test_opt (sb, ERRORS_RO)) {
223 printk (KERN_CRIT "Remounting filesystem read-only\n");
224 sb->s_flags |= MS_RDONLY;
225 }
617ba13b 226 ext4_commit_super(sb, es, 1);
ac27a0ec 227 if (test_opt(sb, ERRORS_PANIC))
617ba13b 228 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 229 sb->s_id);
230}
231
617ba13b 232void ext4_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 233 const char * fmt, ...)
234{
235 va_list args;
236
237 va_start(args, fmt);
617ba13b 238 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 239 vprintk(fmt, args);
240 printk("\n");
241 va_end(args);
242
617ba13b 243 ext4_handle_error(sb);
ac27a0ec
DK		 244}
245
617ba13b 246static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec
DK		 247 char nbuf[16])
248{
249 char *errstr = NULL;
250
251 switch (errno) {
252 case -EIO:
253 errstr = "IO failure";
254 break;
255 case -ENOMEM:
256 errstr = "Out of memory";
257 break;
258 case -EROFS:
dab291af 259 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 260 errstr = "Journal has aborted";
261 else
262 errstr = "Readonly filesystem";
263 break;
264 default:
265 /* If the caller passed in an extra buffer for unknown
266 * errors, textualise them now. Else we just return
267 * NULL. */
268 if (nbuf) {
269 /* Check for truncated error codes... */
270 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
271 errstr = nbuf;
272 }
273 break;
274 }
275
276 return errstr;
277}
278
617ba13b 279/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 280 * automatically and invokes the appropriate error response. */
281
617ba13b 282void __ext4_std_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 283 int errno)
284{
285 char nbuf[16];
286 const char *errstr;
287
288 /* Special case: if the error is EROFS, and we're not already
289 * inside a transaction, then there's really no point in logging
290 * an error. */
291 if (errno == -EROFS && journal_current_handle() == NULL &&
292 (sb->s_flags & MS_RDONLY))
293 return;
294
617ba13b
MC		 295 errstr = ext4_decode_error(sb, errno, nbuf);
296 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
ac27a0ec
DK		 297 sb->s_id, function, errstr);
298
617ba13b 299 ext4_handle_error(sb);
ac27a0ec
DK		 300}
301
302/*
617ba13b 303 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 304 * abort function may be used to deal with unrecoverable failures such
305 * as journal IO errors or ENOMEM at a critical moment in log management.
306 *
307 * We unconditionally force the filesystem into an ABORT|READONLY state,
308 * unless the error response on the fs has been set to panic in which
309 * case we take the easy way out and panic immediately.
310 */
311
617ba13b 312void ext4_abort (struct super_block * sb, const char * function,
ac27a0ec
DK		 313 const char * fmt, ...)
314{
315 va_list args;
316
617ba13b 317 printk (KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 318
319 va_start(args, fmt);
617ba13b 320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 321 vprintk(fmt, args);
322 printk("\n");
323 va_end(args);
324
325 if (test_opt(sb, ERRORS_PANIC))
617ba13b 326 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 327
328 if (sb->s_flags & MS_RDONLY)
329 return;
330
331 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 332 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 333 sb->s_flags |= MS_RDONLY;
617ba13b 334 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 335 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 336}
337
617ba13b 338void ext4_warning (struct super_block * sb, const char * function,
ac27a0ec
DK		 339 const char * fmt, ...)
340{
341 va_list args;
342
343 va_start(args, fmt);
617ba13b 344 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 345 sb->s_id, function);
346 vprintk(fmt, args);
347 printk("\n");
348 va_end(args);
349}
350
617ba13b 351void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 352{
617ba13b 353 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 354
617ba13b 355 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 356 return;
357
617ba13b 358 ext4_warning(sb, __FUNCTION__,
ac27a0ec
DK		 359 "updating to rev %d because of new feature flag, "
360 "running e2fsck is recommended",
617ba13b 361 EXT4_DYNAMIC_REV);
ac27a0ec 362
617ba13b
MC		 363 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
364 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
365 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 366 /* leave es->s_feature_*compat flags alone */
367 /* es->s_uuid will be set by e2fsck if empty */
368
369 /*
370 * The rest of the superblock fields should be zero, and if not it
371 * means they are likely already in use, so leave them alone. We
372 * can leave it up to e2fsck to clean up any inconsistencies there.
373 */
374}
375
99e6f829
AK		 376int ext4_update_compat_feature(handle_t *handle,
377 struct super_block *sb, __u32 compat)
378{
379 int err = 0;
380 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
381 err = ext4_journal_get_write_access(handle,
382 EXT4_SB(sb)->s_sbh);
383 if (err)
384 return err;
385 EXT4_SET_COMPAT_FEATURE(sb, compat);
386 sb->s_dirt = 1;
387 handle->h_sync = 1;
388 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
389 "call ext4_journal_dirty_met adata");
390 err = ext4_journal_dirty_metadata(handle,
391 EXT4_SB(sb)->s_sbh);
392 }
393 return err;
394}
395
396int ext4_update_rocompat_feature(handle_t *handle,
397 struct super_block *sb, __u32 rocompat)
398{
399 int err = 0;
400 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
401 err = ext4_journal_get_write_access(handle,
402 EXT4_SB(sb)->s_sbh);
403 if (err)
404 return err;
405 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
406 sb->s_dirt = 1;
407 handle->h_sync = 1;
408 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
409 "call ext4_journal_dirty_met adata");
410 err = ext4_journal_dirty_metadata(handle,
411 EXT4_SB(sb)->s_sbh);
412 }
413 return err;
414}
415
416int ext4_update_incompat_feature(handle_t *handle,
417 struct super_block *sb, __u32 incompat)
418{
419 int err = 0;
420 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
421 err = ext4_journal_get_write_access(handle,
422 EXT4_SB(sb)->s_sbh);
423 if (err)
424 return err;
425 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
426 sb->s_dirt = 1;
427 handle->h_sync = 1;
428 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
429 "call ext4_journal_dirty_met adata");
430 err = ext4_journal_dirty_metadata(handle,
431 EXT4_SB(sb)->s_sbh);
432 }
433 return err;
434}
435
ac27a0ec
DK		 436/*
437 * Open the external journal device
438 */
617ba13b 439static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 440{
441 struct block_device *bdev;
442 char b[BDEVNAME_SIZE];
443
444 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
445 if (IS_ERR(bdev))
446 goto fail;
447 return bdev;
448
449fail:
617ba13b 450 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 451 __bdevname(dev, b), PTR_ERR(bdev));
452 return NULL;
453}
454
455/*
456 * Release the journal device
457 */
617ba13b 458static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 459{
460 bd_release(bdev);
461 return blkdev_put(bdev);
462}
463
617ba13b 464static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 465{
466 struct block_device *bdev;
467 int ret = -ENODEV;
468
469 bdev = sbi->journal_bdev;
470 if (bdev) {
617ba13b 471 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 472 sbi->journal_bdev = NULL;
473 }
474 return ret;
475}
476
477static inline struct inode *orphan_list_entry(struct list_head *l)
478{
617ba13b 479 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 480}
481
617ba13b 482static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 483{
484 struct list_head *l;
485
486 printk(KERN_ERR "sb orphan head is %d\n",
487 le32_to_cpu(sbi->s_es->s_last_orphan));
488
489 printk(KERN_ERR "sb_info orphan list:\n");
490 list_for_each(l, &sbi->s_orphan) {
491 struct inode *inode = orphan_list_entry(l);
492 printk(KERN_ERR " "
493 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
494 inode->i_sb->s_id, inode->i_ino, inode,
495 inode->i_mode, inode->i_nlink,
496 NEXT_ORPHAN(inode));
497 }
498}
499
617ba13b 500static void ext4_put_super (struct super_block * sb)
ac27a0ec 501{
617ba13b
MC		 502 struct ext4_sb_info *sbi = EXT4_SB(sb);
503 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 504 int i;
505
a86c6181 506 ext4_ext_release(sb);
617ba13b 507 ext4_xattr_put_super(sb);
dab291af 508 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec 509 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 510 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 511 es->s_state = cpu_to_le16(sbi->s_mount_state);
512 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 mark_buffer_dirty(sbi->s_sbh);
617ba13b 514 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 515 }
516
517 for (i = 0; i < sbi->s_gdb_count; i++)
518 brelse(sbi->s_group_desc[i]);
519 kfree(sbi->s_group_desc);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
523 brelse(sbi->s_sbh);
524#ifdef CONFIG_QUOTA
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
527#endif
528
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
536
f98393a6 537 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
539 /*
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
543 */
544 sync_blockdev(sbi->journal_bdev);
f98393a6 545 invalidate_bdev(sbi->journal_bdev);
617ba13b 546 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 547 }
548 sb->s_fs_info = NULL;
549 kfree(sbi);
550 return;
551}
552
e18b890b 553static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 554
555/*
556 * Called inside transaction, so use GFP_NOFS
557 */
617ba13b 558static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 559{
617ba13b 560 struct ext4_inode_info *ei;
ac27a0ec 561
e6b4f8da 562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 563 if (!ei)
564 return NULL;
617ba13b
MC		 565#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 568#endif
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
a86c6181 571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
ac27a0ec
DK		 572 return &ei->vfs_inode;
573}
574
617ba13b 575static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 576{
9f7dd93d
VA		 577 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
578 printk("EXT4 Inode %p: orphan list check failed!\n",
579 EXT4_I(inode));
580 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
581 EXT4_I(inode), sizeof(struct ext4_inode_info),
582 true);
583 dump_stack();
584 }
617ba13b 585 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 586}
587
4ba9b9d0 588static void init_once(struct kmem_cache *cachep, void *foo)
ac27a0ec 589{
617ba13b 590 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 591
a35afb83 592 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 593#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 594 init_rwsem(&ei->xattr_sem);
ac27a0ec 595#endif
a35afb83
CL		 596 mutex_init(&ei->truncate_mutex);
597 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 598}
599
600static int init_inodecache(void)
601{
617ba13b
MC		 602 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
603 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 604 0, (SLAB_RECLAIM_ACCOUNT|
605 SLAB_MEM_SPREAD),
20c2df83 606 init_once);
617ba13b 607 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 608 return -ENOMEM;
609 return 0;
610}
611
612static void destroy_inodecache(void)
613{
617ba13b 614 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 615}
616
617ba13b 617static void ext4_clear_inode(struct inode *inode)
ac27a0ec 618{
617ba13b
MC		 619 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
620#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
621 if (EXT4_I(inode)->i_acl &&
622 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
623 posix_acl_release(EXT4_I(inode)->i_acl);
624 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
625 }
626 if (EXT4_I(inode)->i_default_acl &&
627 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
628 posix_acl_release(EXT4_I(inode)->i_default_acl);
629 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 630 }
631#endif
617ba13b
MC		 632 ext4_discard_reservation(inode);
633 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 634 if (unlikely(rsv))
635 kfree(rsv);
636}
637
617ba13b 638static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
ac27a0ec
DK		 639{
640#if defined(CONFIG_QUOTA)
617ba13b 641 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 642
643 if (sbi->s_jquota_fmt)
644 seq_printf(seq, ",jqfmt=%s",
645 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
646
647 if (sbi->s_qf_names[USRQUOTA])
648 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
649
650 if (sbi->s_qf_names[GRPQUOTA])
651 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
652
617ba13b 653 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 654 seq_puts(seq, ",usrquota");
655
617ba13b 656 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 657 seq_puts(seq, ",grpquota");
658#endif
659}
660
d9c9bef1
MS		 661/*
662 * Show an option if
663 * - it's set to a non-default value OR
664 * - if the per-sb default is different from the global default
665 */
617ba13b 666static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec
DK		 667{
668 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 669 struct ext4_sb_info *sbi = EXT4_SB(sb);
670 struct ext4_super_block *es = sbi->s_es;
671 unsigned long def_mount_opts;
672
673 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
674
675 if (sbi->s_sb_block != 1)
676 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
677 if (test_opt(sb, MINIX_DF))
678 seq_puts(seq, ",minixdf");
679 if (test_opt(sb, GRPID))
680 seq_puts(seq, ",grpid");
681 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
682 seq_puts(seq, ",nogrpid");
683 if (sbi->s_resuid != EXT4_DEF_RESUID ||
684 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
685 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
686 }
687 if (sbi->s_resgid != EXT4_DEF_RESGID ||
688 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
689 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
690 }
691 if (test_opt(sb, ERRORS_CONT)) {
692 int def_errors = le16_to_cpu(es->s_errors);
693
694 if (def_errors == EXT4_ERRORS_PANIC ||
695 def_errors == EXT4_ERRORS_RO) {
696 seq_puts(seq, ",errors=continue");
697 }
698 }
699 if (test_opt(sb, ERRORS_RO))
700 seq_puts(seq, ",errors=remount-ro");
701 if (test_opt(sb, ERRORS_PANIC))
702 seq_puts(seq, ",errors=panic");
703 if (test_opt(sb, NO_UID32))
704 seq_puts(seq, ",nouid32");
705 if (test_opt(sb, DEBUG))
706 seq_puts(seq, ",debug");
707 if (test_opt(sb, OLDALLOC))
708 seq_puts(seq, ",oldalloc");
709#ifdef CONFIG_EXT4_FS_XATTR
710 if (test_opt(sb, XATTR_USER))
711 seq_puts(seq, ",user_xattr");
712 if (!test_opt(sb, XATTR_USER) &&
713 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
714 seq_puts(seq, ",nouser_xattr");
715 }
716#endif
717#ifdef CONFIG_EXT4_FS_POSIX_ACL
718 if (test_opt(sb, POSIX_ACL))
719 seq_puts(seq, ",acl");
720 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
721 seq_puts(seq, ",noacl");
722#endif
723 if (!test_opt(sb, RESERVATION))
724 seq_puts(seq, ",noreservation");
725 if (sbi->s_commit_interval) {
726 seq_printf(seq, ",commit=%u",
727 (unsigned) (sbi->s_commit_interval / HZ));
728 }
729 if (test_opt(sb, BARRIER))
730 seq_puts(seq, ",barrier=1");
731 if (test_opt(sb, NOBH))
732 seq_puts(seq, ",nobh");
733 if (!test_opt(sb, EXTENTS))
734 seq_puts(seq, ",noextents");
ac27a0ec 735
617ba13b 736 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 737 seq_puts(seq, ",data=journal");
617ba13b 738 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 739 seq_puts(seq, ",data=ordered");
617ba13b 740 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 741 seq_puts(seq, ",data=writeback");
742
617ba13b 743 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 744
745 return 0;
746}
747
748
1b961ac0
CH		 749static struct inode *ext4_nfs_get_inode(struct super_block *sb,
750 u64 ino, u32 generation)
ac27a0ec 751{
ac27a0ec 752 struct inode *inode;
ac27a0ec 753
617ba13b 754 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 755 return ERR_PTR(-ESTALE);
617ba13b 756 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 757 return ERR_PTR(-ESTALE);
758
759 /* iget isn't really right if the inode is currently unallocated!!
760 *
617ba13b 761 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 762 * deleted, so we should be safe.
763 *
764 * Currently we don't know the generation for parent directory, so
765 * a generation of 0 means "accept any"
766 */
767 inode = iget(sb, ino);
768 if (inode == NULL)
769 return ERR_PTR(-ENOMEM);
770 if (is_bad_inode(inode) ||
771 (generation && inode->i_generation != generation)) {
772 iput(inode);
773 return ERR_PTR(-ESTALE);
774 }
1b961ac0
CH		 775
776 return inode;
777}
778
779static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
780 int fh_len, int fh_type)
781{
782 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
783 ext4_nfs_get_inode);
784}
785
786static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
787 int fh_len, int fh_type)
788{
789 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
790 ext4_nfs_get_inode);
ac27a0ec
DK		 791}
792
793#ifdef CONFIG_QUOTA
794#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
795#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
796
617ba13b
MC		 797static int ext4_dquot_initialize(struct inode *inode, int type);
798static int ext4_dquot_drop(struct inode *inode);
799static int ext4_write_dquot(struct dquot *dquot);
800static int ext4_acquire_dquot(struct dquot *dquot);
801static int ext4_release_dquot(struct dquot *dquot);
802static int ext4_mark_dquot_dirty(struct dquot *dquot);
803static int ext4_write_info(struct super_block *sb, int type);
804static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
805static int ext4_quota_on_mount(struct super_block *sb, int type);
806static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 807 size_t len, loff_t off);
617ba13b 808static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 809 const char *data, size_t len, loff_t off);
810
617ba13b
MC		 811static struct dquot_operations ext4_quota_operations = {
812 .initialize = ext4_dquot_initialize,
813 .drop = ext4_dquot_drop,
ac27a0ec
DK		 814 .alloc_space = dquot_alloc_space,
815 .alloc_inode = dquot_alloc_inode,
816 .free_space = dquot_free_space,
817 .free_inode = dquot_free_inode,
818 .transfer = dquot_transfer,
617ba13b
MC		 819 .write_dquot = ext4_write_dquot,
820 .acquire_dquot = ext4_acquire_dquot,
821 .release_dquot = ext4_release_dquot,
822 .mark_dirty = ext4_mark_dquot_dirty,
823 .write_info = ext4_write_info
ac27a0ec
DK		 824};
825
617ba13b
MC		 826static struct quotactl_ops ext4_qctl_operations = {
827 .quota_on = ext4_quota_on,
ac27a0ec
DK		 828 .quota_off = vfs_quota_off,
829 .quota_sync = vfs_quota_sync,
830 .get_info = vfs_get_dqinfo,
831 .set_info = vfs_set_dqinfo,
832 .get_dqblk = vfs_get_dqblk,
833 .set_dqblk = vfs_set_dqblk
834};
835#endif
836
ee9b6d61 837static const struct super_operations ext4_sops = {
617ba13b
MC		 838 .alloc_inode = ext4_alloc_inode,
839 .destroy_inode = ext4_destroy_inode,
840 .read_inode = ext4_read_inode,
841 .write_inode = ext4_write_inode,
842 .dirty_inode = ext4_dirty_inode,
843 .delete_inode = ext4_delete_inode,
844 .put_super = ext4_put_super,
845 .write_super = ext4_write_super,
846 .sync_fs = ext4_sync_fs,
847 .write_super_lockfs = ext4_write_super_lockfs,
848 .unlockfs = ext4_unlockfs,
849 .statfs = ext4_statfs,
850 .remount_fs = ext4_remount,
851 .clear_inode = ext4_clear_inode,
852 .show_options = ext4_show_options,
ac27a0ec 853#ifdef CONFIG_QUOTA
617ba13b
MC		 854 .quota_read = ext4_quota_read,
855 .quota_write = ext4_quota_write,
ac27a0ec
DK		 856#endif
857};
858
39655164 859static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 860 .fh_to_dentry = ext4_fh_to_dentry,
861 .fh_to_parent = ext4_fh_to_parent,
617ba13b 862 .get_parent = ext4_get_parent,
ac27a0ec
DK		 863};
864
865enum {
866 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
867 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
868 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
869 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
870 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
871 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
872 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
873 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
874 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
875 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
1e2462f9 876 Opt_grpquota, Opt_extents, Opt_noextents,
ac27a0ec
DK		 877};
878
879static match_table_t tokens = {
880 {Opt_bsd_df, "bsddf"},
881 {Opt_minix_df, "minixdf"},
882 {Opt_grpid, "grpid"},
883 {Opt_grpid, "bsdgroups"},
884 {Opt_nogrpid, "nogrpid"},
885 {Opt_nogrpid, "sysvgroups"},
886 {Opt_resgid, "resgid=%u"},
887 {Opt_resuid, "resuid=%u"},
888 {Opt_sb, "sb=%u"},
889 {Opt_err_cont, "errors=continue"},
890 {Opt_err_panic, "errors=panic"},
891 {Opt_err_ro, "errors=remount-ro"},
892 {Opt_nouid32, "nouid32"},
893 {Opt_nocheck, "nocheck"},
894 {Opt_nocheck, "check=none"},
895 {Opt_debug, "debug"},
896 {Opt_oldalloc, "oldalloc"},
897 {Opt_orlov, "orlov"},
898 {Opt_user_xattr, "user_xattr"},
899 {Opt_nouser_xattr, "nouser_xattr"},
900 {Opt_acl, "acl"},
901 {Opt_noacl, "noacl"},
902 {Opt_reservation, "reservation"},
903 {Opt_noreservation, "noreservation"},
904 {Opt_noload, "noload"},
905 {Opt_nobh, "nobh"},
906 {Opt_bh, "bh"},
907 {Opt_commit, "commit=%u"},
908 {Opt_journal_update, "journal=update"},
909 {Opt_journal_inum, "journal=%u"},
910 {Opt_journal_dev, "journal_dev=%u"},
911 {Opt_abort, "abort"},
912 {Opt_data_journal, "data=journal"},
913 {Opt_data_ordered, "data=ordered"},
914 {Opt_data_writeback, "data=writeback"},
915 {Opt_offusrjquota, "usrjquota="},
916 {Opt_usrjquota, "usrjquota=%s"},
917 {Opt_offgrpjquota, "grpjquota="},
918 {Opt_grpjquota, "grpjquota=%s"},
919 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
920 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
921 {Opt_grpquota, "grpquota"},
922 {Opt_noquota, "noquota"},
923 {Opt_quota, "quota"},
924 {Opt_usrquota, "usrquota"},
925 {Opt_barrier, "barrier=%u"},
a86c6181 926 {Opt_extents, "extents"},
1e2462f9 927 {Opt_noextents, "noextents"},
ac27a0ec
DK		 928 {Opt_err, NULL},
929 {Opt_resize, "resize"},
930};
931
617ba13b 932static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 933{
617ba13b 934 ext4_fsblk_t sb_block;
ac27a0ec
DK		 935 char *options = (char *) *data;
936
937 if (!options || strncmp(options, "sb=", 3) != 0)
938 return 1; /* Default location */
939 options += 3;
617ba13b 940 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 941 sb_block = simple_strtoul(options, &options, 0);
942 if (*options && *options != ',') {
617ba13b 943 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 944 (char *) *data);
945 return 1;
946 }
947 if (*options == ',')
948 options++;
949 *data = (void *) options;
950 return sb_block;
951}
952
953static int parse_options (char *options, struct super_block *sb,
954 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 955 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 956{
617ba13b 957 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 958 char * p;
959 substring_t args[MAX_OPT_ARGS];
960 int data_opt = 0;
961 int option;
962#ifdef CONFIG_QUOTA
963 int qtype;
964 char *qname;
965#endif
966
967 if (!options)
968 return 1;
969
970 while ((p = strsep (&options, ",")) != NULL) {
971 int token;
972 if (!*p)
973 continue;
974
975 token = match_token(p, tokens, args);
976 switch (token) {
977 case Opt_bsd_df:
978 clear_opt (sbi->s_mount_opt, MINIX_DF);
979 break;
980 case Opt_minix_df:
981 set_opt (sbi->s_mount_opt, MINIX_DF);
982 break;
983 case Opt_grpid:
984 set_opt (sbi->s_mount_opt, GRPID);
985 break;
986 case Opt_nogrpid:
987 clear_opt (sbi->s_mount_opt, GRPID);
988 break;
989 case Opt_resuid:
990 if (match_int(&args[0], &option))
991 return 0;
992 sbi->s_resuid = option;
993 break;
994 case Opt_resgid:
995 if (match_int(&args[0], &option))
996 return 0;
997 sbi->s_resgid = option;
998 break;
999 case Opt_sb:
1000 /* handled by get_sb_block() instead of here */
1001 /* *sb_block = match_int(&args[0]); */
1002 break;
1003 case Opt_err_panic:
1004 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1005 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1006 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1007 break;
1008 case Opt_err_ro:
1009 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1010 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1011 set_opt (sbi->s_mount_opt, ERRORS_RO);
1012 break;
1013 case Opt_err_cont:
1014 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1015 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1016 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1017 break;
1018 case Opt_nouid32:
1019 set_opt (sbi->s_mount_opt, NO_UID32);
1020 break;
1021 case Opt_nocheck:
1022 clear_opt (sbi->s_mount_opt, CHECK);
1023 break;
1024 case Opt_debug:
1025 set_opt (sbi->s_mount_opt, DEBUG);
1026 break;
1027 case Opt_oldalloc:
1028 set_opt (sbi->s_mount_opt, OLDALLOC);
1029 break;
1030 case Opt_orlov:
1031 clear_opt (sbi->s_mount_opt, OLDALLOC);
1032 break;
617ba13b 1033#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1034 case Opt_user_xattr:
1035 set_opt (sbi->s_mount_opt, XATTR_USER);
1036 break;
1037 case Opt_nouser_xattr:
1038 clear_opt (sbi->s_mount_opt, XATTR_USER);
1039 break;
1040#else
1041 case Opt_user_xattr:
1042 case Opt_nouser_xattr:
617ba13b 1043 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1044 break;
1045#endif
617ba13b 1046#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1047 case Opt_acl:
1048 set_opt(sbi->s_mount_opt, POSIX_ACL);
1049 break;
1050 case Opt_noacl:
1051 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1052 break;
1053#else
1054 case Opt_acl:
1055 case Opt_noacl:
617ba13b 1056 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1057 break;
1058#endif
1059 case Opt_reservation:
1060 set_opt(sbi->s_mount_opt, RESERVATION);
1061 break;
1062 case Opt_noreservation:
1063 clear_opt(sbi->s_mount_opt, RESERVATION);
1064 break;
1065 case Opt_journal_update:
1066 /* @@@ FIXME */
1067 /* Eventually we will want to be able to create
1068 a journal file here. For now, only allow the
1069 user to specify an existing inode to be the
1070 journal file. */
1071 if (is_remount) {
617ba13b 1072 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1073 "journal on remount\n");
1074 return 0;
1075 }
1076 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1077 break;
1078 case Opt_journal_inum:
1079 if (is_remount) {
617ba13b 1080 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1081 "journal on remount\n");
1082 return 0;
1083 }
1084 if (match_int(&args[0], &option))
1085 return 0;
1086 *inum = option;
1087 break;
1088 case Opt_journal_dev:
1089 if (is_remount) {
617ba13b 1090 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1091 "journal on remount\n");
1092 return 0;
1093 }
1094 if (match_int(&args[0], &option))
1095 return 0;
1096 *journal_devnum = option;
1097 break;
1098 case Opt_noload:
1099 set_opt (sbi->s_mount_opt, NOLOAD);
1100 break;
1101 case Opt_commit:
1102 if (match_int(&args[0], &option))
1103 return 0;
1104 if (option < 0)
1105 return 0;
1106 if (option == 0)
cd02ff0b 1107 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1108 sbi->s_commit_interval = HZ * option;
1109 break;
1110 case Opt_data_journal:
617ba13b 1111 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1112 goto datacheck;
1113 case Opt_data_ordered:
617ba13b 1114 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1115 goto datacheck;
1116 case Opt_data_writeback:
617ba13b 1117 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1118 datacheck:
1119 if (is_remount) {
617ba13b 1120 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1121 != data_opt) {
1122 printk(KERN_ERR
617ba13b 1123 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1124 "mode on remount\n");
1125 return 0;
1126 }
1127 } else {
617ba13b 1128 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1129 sbi->s_mount_opt |= data_opt;
1130 }
1131 break;
1132#ifdef CONFIG_QUOTA
1133 case Opt_usrjquota:
1134 qtype = USRQUOTA;
1135 goto set_qf_name;
1136 case Opt_grpjquota:
1137 qtype = GRPQUOTA;
1138set_qf_name:
1139 if (sb_any_quota_enabled(sb)) {
1140 printk(KERN_ERR
617ba13b 1141 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1142 "quota options when quota turned on.\n");
1143 return 0;
1144 }
1145 qname = match_strdup(&args[0]);
1146 if (!qname) {
1147 printk(KERN_ERR
617ba13b 1148 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1149 "storing quotafile name.\n");
1150 return 0;
1151 }
1152 if (sbi->s_qf_names[qtype] &&
1153 strcmp(sbi->s_qf_names[qtype], qname)) {
1154 printk(KERN_ERR
617ba13b 1155 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1156 "specified.\n", QTYPE2NAME(qtype));
1157 kfree(qname);
1158 return 0;
1159 }
1160 sbi->s_qf_names[qtype] = qname;
1161 if (strchr(sbi->s_qf_names[qtype], '/')) {
1162 printk(KERN_ERR
617ba13b 1163 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1164 "filesystem root.\n");
1165 kfree(sbi->s_qf_names[qtype]);
1166 sbi->s_qf_names[qtype] = NULL;
1167 return 0;
1168 }
1169 set_opt(sbi->s_mount_opt, QUOTA);
1170 break;
1171 case Opt_offusrjquota:
1172 qtype = USRQUOTA;
1173 goto clear_qf_name;
1174 case Opt_offgrpjquota:
1175 qtype = GRPQUOTA;
1176clear_qf_name:
1177 if (sb_any_quota_enabled(sb)) {
617ba13b 1178 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1179 "journalled quota options when "
1180 "quota turned on.\n");
1181 return 0;
1182 }
1183 /*
1184 * The space will be released later when all options
1185 * are confirmed to be correct
1186 */
1187 sbi->s_qf_names[qtype] = NULL;
1188 break;
1189 case Opt_jqfmt_vfsold:
1190 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1191 break;
1192 case Opt_jqfmt_vfsv0:
1193 sbi->s_jquota_fmt = QFMT_VFS_V0;
1194 break;
1195 case Opt_quota:
1196 case Opt_usrquota:
1197 set_opt(sbi->s_mount_opt, QUOTA);
1198 set_opt(sbi->s_mount_opt, USRQUOTA);
1199 break;
1200 case Opt_grpquota:
1201 set_opt(sbi->s_mount_opt, QUOTA);
1202 set_opt(sbi->s_mount_opt, GRPQUOTA);
1203 break;
1204 case Opt_noquota:
1205 if (sb_any_quota_enabled(sb)) {
617ba13b 1206 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1207 "options when quota turned on.\n");
1208 return 0;
1209 }
1210 clear_opt(sbi->s_mount_opt, QUOTA);
1211 clear_opt(sbi->s_mount_opt, USRQUOTA);
1212 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1213 break;
1214#else
1215 case Opt_quota:
1216 case Opt_usrquota:
1217 case Opt_grpquota:
1218 case Opt_usrjquota:
1219 case Opt_grpjquota:
1220 case Opt_offusrjquota:
1221 case Opt_offgrpjquota:
1222 case Opt_jqfmt_vfsold:
1223 case Opt_jqfmt_vfsv0:
1224 printk(KERN_ERR
617ba13b 1225 "EXT4-fs: journalled quota options not "
ac27a0ec
DK		 1226 "supported.\n");
1227 break;
1228 case Opt_noquota:
1229 break;
1230#endif
1231 case Opt_abort:
1232 set_opt(sbi->s_mount_opt, ABORT);
1233 break;
1234 case Opt_barrier:
1235 if (match_int(&args[0], &option))
1236 return 0;
1237 if (option)
1238 set_opt(sbi->s_mount_opt, BARRIER);
1239 else
1240 clear_opt(sbi->s_mount_opt, BARRIER);
1241 break;
1242 case Opt_ignore:
1243 break;
1244 case Opt_resize:
1245 if (!is_remount) {
617ba13b 1246 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1247 "for remount\n");
1248 return 0;
1249 }
1250 if (match_int(&args[0], &option) != 0)
1251 return 0;
1252 *n_blocks_count = option;
1253 break;
1254 case Opt_nobh:
1255 set_opt(sbi->s_mount_opt, NOBH);
1256 break;
1257 case Opt_bh:
1258 clear_opt(sbi->s_mount_opt, NOBH);
1259 break;
a86c6181
AT		 1260 case Opt_extents:
1261 set_opt (sbi->s_mount_opt, EXTENTS);
1262 break;
1e2462f9
MC		 1263 case Opt_noextents:
1264 clear_opt (sbi->s_mount_opt, EXTENTS);
1265 break;
ac27a0ec
DK		 1266 default:
1267 printk (KERN_ERR
617ba13b 1268 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1269 "or missing value\n", p);
1270 return 0;
1271 }
1272 }
1273#ifdef CONFIG_QUOTA
1274 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1275 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1276 sbi->s_qf_names[USRQUOTA])
1277 clear_opt(sbi->s_mount_opt, USRQUOTA);
1278
617ba13b 1279 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1280 sbi->s_qf_names[GRPQUOTA])
1281 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1282
1283 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1284 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1285 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1286 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1287 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1288 "format mixing.\n");
1289 return 0;
1290 }
1291
1292 if (!sbi->s_jquota_fmt) {
617ba13b 1293 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1294 "not specified.\n");
1295 return 0;
1296 }
1297 } else {
1298 if (sbi->s_jquota_fmt) {
617ba13b 1299 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1300 "specified with no journalling "
1301 "enabled.\n");
1302 return 0;
1303 }
1304 }
1305#endif
1306 return 1;
1307}
1308
617ba13b 1309static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1310 int read_only)
1311{
617ba13b 1312 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1313 int res = 0;
1314
617ba13b
MC		 1315 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1316 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1317 "forcing read-only mode\n");
1318 res = MS_RDONLY;
1319 }
1320 if (read_only)
1321 return res;
617ba13b
MC		 1322 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1323 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1324 "running e2fsck is recommended\n");
617ba13b 1325 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1326 printk (KERN_WARNING
617ba13b 1327 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1328 "running e2fsck is recommended\n");
1329 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1330 le16_to_cpu(es->s_mnt_count) >=
1331 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1332 printk (KERN_WARNING
617ba13b 1333 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1334 "running e2fsck is recommended\n");
1335 else if (le32_to_cpu(es->s_checkinterval) &&
1336 (le32_to_cpu(es->s_lastcheck) +
1337 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1338 printk (KERN_WARNING
617ba13b 1339 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1340 "running e2fsck is recommended\n");
1341#if 0
1342 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1343 * inconsistencies, to force a fsck at reboot. But for
1344 * a plain journaled filesystem we can keep it set as
1345 * valid forever! :)
1346 */
617ba13b 1347 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
ac27a0ec
DK		 1348#endif
1349 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1350 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
ac27a0ec
DK		 1351 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1352 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1353 ext4_update_dynamic_rev(sb);
1354 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1355
617ba13b 1356 ext4_commit_super(sb, es, 1);
ac27a0ec 1357 if (test_opt(sb, DEBUG))
617ba13b 1358 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1359 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1360 sb->s_blocksize,
1361 sbi->s_groups_count,
617ba13b
MC		 1362 EXT4_BLOCKS_PER_GROUP(sb),
1363 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1364 sbi->s_mount_opt);
1365
617ba13b
MC		 1366 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1367 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1368 char b[BDEVNAME_SIZE];
1369
1370 printk("external journal on %s\n",
617ba13b 1371 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1372 } else {
1373 printk("internal journal\n");
1374 }
1375 return res;
1376}
1377
717d50e4
AD		 1378__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1379 struct ext4_group_desc *gdp)
1380{
1381 __u16 crc = 0;
1382
1383 if (sbi->s_es->s_feature_ro_compat &
1384 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1385 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1386 __le32 le_group = cpu_to_le32(block_group);
1387
1388 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1389 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1390 crc = crc16(crc, (__u8 *)gdp, offset);
1391 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1392 /* for checksum of struct ext4_group_desc do the rest...*/
1393 if ((sbi->s_es->s_feature_incompat &
1394 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1395 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1396 crc = crc16(crc, (__u8 *)gdp + offset,
1397 le16_to_cpu(sbi->s_es->s_desc_size) -
1398 offset);
1399 }
1400
1401 return cpu_to_le16(crc);
1402}
1403
1404int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1405 struct ext4_group_desc *gdp)
1406{
1407 if ((sbi->s_es->s_feature_ro_compat &
1408 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1409 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1410 return 0;
1411
1412 return 1;
1413}
1414
ac27a0ec 1415/* Called at mount-time, super-block is locked */
617ba13b 1416static int ext4_check_descriptors (struct super_block * sb)
ac27a0ec 1417{
617ba13b
MC		 1418 struct ext4_sb_info *sbi = EXT4_SB(sb);
1419 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1420 ext4_fsblk_t last_block;
bd81d8ee
LV		 1421 ext4_fsblk_t block_bitmap;
1422 ext4_fsblk_t inode_bitmap;
1423 ext4_fsblk_t inode_table;
617ba13b 1424 struct ext4_group_desc * gdp = NULL;
ac27a0ec 1425 int desc_block = 0;
ce421581 1426 int flexbg_flag = 0;
fd2d4291 1427 ext4_group_t i;
ac27a0ec 1428
ce421581
JS		 1429 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1430 flexbg_flag = 1;
1431
617ba13b 1432 ext4_debug ("Checking group descriptors");
ac27a0ec
DK		 1433
1434 for (i = 0; i < sbi->s_groups_count; i++)
1435 {
ce421581 1436 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1437 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1438 else
1439 last_block = first_block +
617ba13b 1440 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1441
617ba13b
MC		 1442 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1443 gdp = (struct ext4_group_desc *)
ac27a0ec 1444 sbi->s_group_desc[desc_block++]->b_data;
8fadc143 1445 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1446 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1447 {
617ba13b 1448 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1449 "Block bitmap for group %lu"
2ae02107 1450 " not in group (block %llu)!",
bd81d8ee 1451 i, block_bitmap);
ac27a0ec
DK		 1452 return 0;
1453 }
8fadc143 1454 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1455 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1456 {
617ba13b 1457 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1458 "Inode bitmap for group %lu"
2ae02107 1459 " not in group (block %llu)!",
bd81d8ee 1460 i, inode_bitmap);
ac27a0ec
DK		 1461 return 0;
1462 }
8fadc143 1463 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1464 if (inode_table < first_block ||
780dcdb2 1465 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1466 {
617ba13b 1467 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1468 "Inode table for group %lu"
2ae02107 1469 " not in group (block %llu)!",
bd81d8ee 1470 i, inode_table);
ac27a0ec
DK		 1471 return 0;
1472 }
717d50e4
AD		 1473 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1474 ext4_error(sb, __FUNCTION__,
fd2d4291
AM		 1475 "Checksum for group %lu failed (%u!=%u)\n",
1476 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1477 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1478 return 0;
1479 }
ce421581
JS		 1480 if (!flexbg_flag)
1481 first_block += EXT4_BLOCKS_PER_GROUP(sb);
0d1ee42f
AR		 1482 gdp = (struct ext4_group_desc *)
1483 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
ac27a0ec
DK		 1484 }
1485
bd81d8ee 1486 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1487 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1488 return 1;
1489}
1490
617ba13b 1491/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1492 * the superblock) which were deleted from all directories, but held open by
1493 * a process at the time of a crash. We walk the list and try to delete these
1494 * inodes at recovery time (only with a read-write filesystem).
1495 *
1496 * In order to keep the orphan inode chain consistent during traversal (in
1497 * case of crash during recovery), we link each inode into the superblock
1498 * orphan list_head and handle it the same way as an inode deletion during
1499 * normal operation (which journals the operations for us).
1500 *
1501 * We only do an iget() and an iput() on each inode, which is very safe if we
1502 * accidentally point at an in-use or already deleted inode. The worst that
1503 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1504 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1505 * e2fsck was run on this filesystem, and it must have already done the orphan
1506 * inode cleanup for us, so we can safely abort without any further action.
1507 */
617ba13b
MC		 1508static void ext4_orphan_cleanup (struct super_block * sb,
1509 struct ext4_super_block * es)
ac27a0ec
DK		 1510{
1511 unsigned int s_flags = sb->s_flags;
1512 int nr_orphans = 0, nr_truncates = 0;
1513#ifdef CONFIG_QUOTA
1514 int i;
1515#endif
1516 if (!es->s_last_orphan) {
1517 jbd_debug(4, "no orphan inodes to clean up\n");
1518 return;
1519 }
1520
a8f48a95
ES		 1521 if (bdev_read_only(sb->s_bdev)) {
1522 printk(KERN_ERR "EXT4-fs: write access "
1523 "unavailable, skipping orphan cleanup.\n");
1524 return;
1525 }
1526
617ba13b 1527 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1528 if (es->s_last_orphan)
1529 jbd_debug(1, "Errors on filesystem, "
1530 "clearing orphan list.\n");
1531 es->s_last_orphan = 0;
1532 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1533 return;
1534 }
1535
1536 if (s_flags & MS_RDONLY) {
617ba13b 1537 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1538 sb->s_id);
1539 sb->s_flags &= ~MS_RDONLY;
1540 }
1541#ifdef CONFIG_QUOTA
1542 /* Needed for iput() to work correctly and not trash data */
1543 sb->s_flags |= MS_ACTIVE;
1544 /* Turn on quotas so that they are updated correctly */
1545 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1546 if (EXT4_SB(sb)->s_qf_names[i]) {
1547 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1548 if (ret < 0)
1549 printk(KERN_ERR
617ba13b 1550 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1551 "quota: error %d\n", ret);
1552 }
1553 }
1554#endif
1555
1556 while (es->s_last_orphan) {
1557 struct inode *inode;
1558
1559 if (!(inode =
617ba13b 1560 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
ac27a0ec
DK		 1561 es->s_last_orphan = 0;
1562 break;
1563 }
1564
617ba13b 1565 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1566 DQUOT_INIT(inode);
1567 if (inode->i_nlink) {
1568 printk(KERN_DEBUG
1569 "%s: truncating inode %lu to %Ld bytes\n",
1570 __FUNCTION__, inode->i_ino, inode->i_size);
1571 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1572 inode->i_ino, inode->i_size);
617ba13b 1573 ext4_truncate(inode);
ac27a0ec
DK		 1574 nr_truncates++;
1575 } else {
1576 printk(KERN_DEBUG
1577 "%s: deleting unreferenced inode %lu\n",
1578 __FUNCTION__, inode->i_ino);
1579 jbd_debug(2, "deleting unreferenced inode %lu\n",
1580 inode->i_ino);
1581 nr_orphans++;
1582 }
1583 iput(inode); /* The delete magic happens here! */
1584 }
1585
1586#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1587
1588 if (nr_orphans)
617ba13b 1589 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1590 sb->s_id, PLURAL(nr_orphans));
1591 if (nr_truncates)
617ba13b 1592 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1593 sb->s_id, PLURAL(nr_truncates));
1594#ifdef CONFIG_QUOTA
1595 /* Turn quotas off */
1596 for (i = 0; i < MAXQUOTAS; i++) {
1597 if (sb_dqopt(sb)->files[i])
1598 vfs_quota_off(sb, i);
1599 }
1600#endif
1601 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1602}
cd2291a4
ES		 1603/*
1604 * Maximal extent format file size.
1605 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1606 * extent format containers, within a sector_t, and within i_blocks
1607 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1608 * so that won't be a limiting factor.
1609 *
1610 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1611 */
1612static loff_t ext4_max_size(int blkbits)
1613{
1614 loff_t res;
1615 loff_t upper_limit = MAX_LFS_FILESIZE;
1616
1617 /* small i_blocks in vfs inode? */
1618 if (sizeof(blkcnt_t) < sizeof(u64)) {
1619 /*
1620 * CONFIG_LSF is not enabled implies the inode
1621 * i_block represent total blocks in 512 bytes
1622 * 32 == size of vfs inode i_blocks * 8
1623 */
1624 upper_limit = (1LL << 32) - 1;
1625
1626 /* total blocks in file system block size */
1627 upper_limit >>= (blkbits - 9);
1628 upper_limit <<= blkbits;
1629 }
1630
1631 /* 32-bit extent-start container, ee_block */
1632 res = 1LL << 32;
1633 res <<= blkbits;
1634 res -= 1;
1635
1636 /* Sanity check against vm- & vfs- imposed limits */
1637 if (res > upper_limit)
1638 res = upper_limit;
1639
1640 return res;
1641}
ac27a0ec 1642
ac27a0ec 1643/*
cd2291a4 1644 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1645 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1646 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1647 */
cd2291a4 1648static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1649{
617ba13b 1650 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1651 int meta_blocks;
1652 loff_t upper_limit;
1653 /* This is calculated to be the largest file size for a
cd2291a4 1654 * dense, bitmapped file such that the total number of
ac27a0ec 1655 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1656 * does not exceed 2^48 -1
1657 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1658 * total number of 512 bytes blocks of the file
1659 */
1660
1661 if (sizeof(blkcnt_t) < sizeof(u64)) {
1662 /*
1663 * CONFIG_LSF is not enabled implies the inode
1664 * i_block represent total blocks in 512 bytes
1665 * 32 == size of vfs inode i_blocks * 8
1666 */
1667 upper_limit = (1LL << 32) - 1;
1668
1669 /* total blocks in file system block size */
1670 upper_limit >>= (bits - 9);
1671
1672 } else {
8180a562
AK		 1673 /*
1674 * We use 48 bit ext4_inode i_blocks
1675 * With EXT4_HUGE_FILE_FL set the i_blocks
1676 * represent total number of blocks in
1677 * file system block size
1678 */
0fc1b451
AK		 1679 upper_limit = (1LL << 48) - 1;
1680
0fc1b451
AK		 1681 }
1682
1683 /* indirect blocks */
1684 meta_blocks = 1;
1685 /* double indirect blocks */
1686 meta_blocks += 1 + (1LL << (bits-2));
1687 /* tripple indirect blocks */
1688 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1689
1690 upper_limit -= meta_blocks;
1691 upper_limit <<= bits;
ac27a0ec
DK		 1692
1693 res += 1LL << (bits-2);
1694 res += 1LL << (2*(bits-2));
1695 res += 1LL << (3*(bits-2));
1696 res <<= bits;
1697 if (res > upper_limit)
1698 res = upper_limit;
0fc1b451
AK		 1699
1700 if (res > MAX_LFS_FILESIZE)
1701 res = MAX_LFS_FILESIZE;
1702
ac27a0ec
DK		 1703 return res;
1704}
1705
617ba13b 1706static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1707 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1708{
617ba13b 1709 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1710 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1711 int has_super = 0;
1712
1713 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1714
617ba13b 1715 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1716 nr < first_meta_bg)
70bbb3e0 1717 return logical_sb_block + nr + 1;
ac27a0ec 1718 bg = sbi->s_desc_per_block * nr;
617ba13b 1719 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1720 has_super = 1;
617ba13b 1721 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1722}
1723
1724
617ba13b 1725static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1d03ec98
AK		 1726 __releases(kernel_sem)
1727 __acquires(kernel_sem)
1728
ac27a0ec
DK		 1729{
1730 struct buffer_head * bh;
617ba13b
MC		 1731 struct ext4_super_block *es = NULL;
1732 struct ext4_sb_info *sbi;
1733 ext4_fsblk_t block;
1734 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1735 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1736 unsigned long offset = 0;
1737 unsigned int journal_inum = 0;
1738 unsigned long journal_devnum = 0;
1739 unsigned long def_mount_opts;
1740 struct inode *root;
1741 int blocksize;
1742 int hblock;
1743 int db_count;
1744 int i;
1745 int needs_recovery;
1746 __le32 features;
bd81d8ee 1747 __u64 blocks_count;
833f4077 1748 int err;
ac27a0ec
DK		 1749
1750 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1751 if (!sbi)
1752 return -ENOMEM;
1753 sb->s_fs_info = sbi;
1754 sbi->s_mount_opt = 0;
617ba13b
MC		 1755 sbi->s_resuid = EXT4_DEF_RESUID;
1756 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1757 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1758
1759 unlock_kernel();
1760
617ba13b 1761 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1762 if (!blocksize) {
617ba13b 1763 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1764 goto out_fail;
1765 }
1766
afc7cbca
TS		 1767 if (!sb_set_blocksize(sb, blocksize)) {
1768 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1769 goto out_fail;
1770 }
1771
ac27a0ec 1772 /*
617ba13b 1773 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1774 * block sizes. We need to calculate the offset from buffer start.
1775 */
617ba13b 1776 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1777 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1778 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1779 } else {
70bbb3e0 1780 logical_sb_block = sb_block;
ac27a0ec
DK		 1781 }
1782
70bbb3e0 1783 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1784 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1785 goto out_fail;
1786 }
1787 /*
1788 * Note: s_es must be initialized as soon as possible because
617ba13b 1789 * some ext4 macro-instructions depend on its value
ac27a0ec 1790 */
617ba13b 1791 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1792 sbi->s_es = es;
1793 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1794 if (sb->s_magic != EXT4_SUPER_MAGIC)
1795 goto cantfind_ext4;
ac27a0ec
DK		 1796
1797 /* Set defaults before we parse the mount options */
1798 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1799 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1800 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1801 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1802 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1803 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1804 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1805#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1806 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1807 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1808#endif
1809#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1810 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1811 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1812#endif
617ba13b
MC		 1813 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1814 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1815 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1816 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1817 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1818 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1819
1820 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1821 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
617ba13b 1822 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
ac27a0ec 1823 set_opt(sbi->s_mount_opt, ERRORS_RO);
ceea16bf
DM		 1824 else
1825 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1826
1827 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1828 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1829
1830 set_opt(sbi->s_mount_opt, RESERVATION);
1831
1e2462f9
MC		 1832 /*
1833 * turn on extents feature by default in ext4 filesystem
1834 * User -o noextents to turn it off
1835 */
1836 set_opt(sbi->s_mount_opt, EXTENTS);
1837
ac27a0ec
DK		 1838 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1839 NULL, 0))
1840 goto failed_mount;
1841
1842 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1843 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1844
617ba13b
MC		 1845 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1846 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1847 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1848 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1849 printk(KERN_WARNING
617ba13b 1850 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec
DK		 1851 "running e2fsck is recommended\n");
1852 /*
1853 * Check feature flags regardless of the revision level, since we
1854 * previously didn't change the revision level when setting the flags,
1855 * so there is a chance incompat flags are set on a rev 0 filesystem.
1856 */
617ba13b 1857 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1858 if (features) {
617ba13b 1859 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1860 "unsupported optional features (%x).\n",
1861 sb->s_id, le32_to_cpu(features));
1862 goto failed_mount;
1863 }
617ba13b 1864 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1865 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1866 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1867 "unsupported optional features (%x).\n",
1868 sb->s_id, le32_to_cpu(features));
1869 goto failed_mount;
1870 }
0fc1b451
AK		 1871 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1872 /*
1873 * Large file size enabled file system can only be
1874 * mount if kernel is build with CONFIG_LSF
1875 */
1876 if (sizeof(root->i_blocks) < sizeof(u64) &&
1877 !(sb->s_flags & MS_RDONLY)) {
1878 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1879 "files cannot be mounted read-write "
1880 "without CONFIG_LSF.\n", sb->s_id);
1881 goto failed_mount;
1882 }
1883 }
ac27a0ec
DK		 1884 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1885
617ba13b
MC		 1886 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1887 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1888 printk(KERN_ERR
617ba13b 1889 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1890 blocksize, sb->s_id);
1891 goto failed_mount;
1892 }
1893
1894 hblock = bdev_hardsect_size(sb->s_bdev);
1895 if (sb->s_blocksize != blocksize) {
1896 /*
1897 * Make sure the blocksize for the filesystem is larger
1898 * than the hardware sectorsize for the machine.
1899 */
1900 if (blocksize < hblock) {
617ba13b 1901 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
ac27a0ec
DK		 1902 "device blocksize %d.\n", blocksize, hblock);
1903 goto failed_mount;
1904 }
1905
1906 brelse (bh);
1907 sb_set_blocksize(sb, blocksize);
70bbb3e0
AM		 1908 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1909 offset = do_div(logical_sb_block, blocksize);
1910 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 1911 if (!bh) {
1912 printk(KERN_ERR
617ba13b 1913 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 1914 goto failed_mount;
1915 }
617ba13b 1916 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 1917 sbi->s_es = es;
617ba13b 1918 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 1919 printk (KERN_ERR
617ba13b 1920 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 1921 goto failed_mount;
1922 }
1923 }
1924
617ba13b 1925 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 1926
617ba13b
MC		 1927 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1928 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1929 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 1930 } else {
1931 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1932 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 1933 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 1934 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 1935 (sbi->s_inode_size > blocksize)) {
1936 printk (KERN_ERR
617ba13b 1937 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 1938 sbi->s_inode_size);
1939 goto failed_mount;
1940 }
ef7f3835
KS		 1941 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1942 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 1943 }
0d1ee42f
AR		 1944 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1945 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 1946 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 1947 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 1948 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 1949 printk(KERN_ERR
8fadc143 1950 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 1951 sbi->s_desc_size);
1952 goto failed_mount;
1953 }
1954 } else
1955 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 1956 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 1957 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1958 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 1959 goto cantfind_ext4;
1960 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 1961 if (sbi->s_inodes_per_block == 0)
617ba13b 1962 goto cantfind_ext4;
ac27a0ec
DK		 1963 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1964 sbi->s_inodes_per_block;
0d1ee42f 1965 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 1966 sbi->s_sbh = bh;
1967 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 1968 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1969 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 1970 for (i=0; i < 4; i++)
1971 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1972 sbi->s_def_hash_version = es->s_def_hash_version;
1973
1974 if (sbi->s_blocks_per_group > blocksize * 8) {
1975 printk (KERN_ERR
617ba13b 1976 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 1977 sbi->s_blocks_per_group);
1978 goto failed_mount;
1979 }
ac27a0ec
DK		 1980 if (sbi->s_inodes_per_group > blocksize * 8) {
1981 printk (KERN_ERR
617ba13b 1982 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 1983 sbi->s_inodes_per_group);
1984 goto failed_mount;
1985 }
1986
bd81d8ee 1987 if (ext4_blocks_count(es) >
ac27a0ec 1988 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 1989 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 1990 " too large to mount safely\n", sb->s_id);
1991 if (sizeof(sector_t) < 8)
617ba13b 1992 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 1993 "enabled\n");
1994 goto failed_mount;
1995 }
1996
617ba13b
MC		 1997 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1998 goto cantfind_ext4;
bd81d8ee
LV		 1999 blocks_count = (ext4_blocks_count(es) -
2000 le32_to_cpu(es->s_first_data_block) +
2001 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2002 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2003 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2004 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2005 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2006 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2007 GFP_KERNEL);
2008 if (sbi->s_group_desc == NULL) {
617ba13b 2009 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2010 goto failed_mount;
2011 }
2012
2013 bgl_lock_init(&sbi->s_blockgroup_lock);
2014
2015 for (i = 0; i < db_count; i++) {
70bbb3e0 2016 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2017 sbi->s_group_desc[i] = sb_bread(sb, block);
2018 if (!sbi->s_group_desc[i]) {
617ba13b 2019 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2020 "can't read group descriptor %d\n", i);
2021 db_count = i;
2022 goto failed_mount2;
2023 }
2024 }
617ba13b
MC		 2025 if (!ext4_check_descriptors (sb)) {
2026 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2027 goto failed_mount2;
2028 }
2029 sbi->s_gdb_count = db_count;
2030 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2031 spin_lock_init(&sbi->s_next_gen_lock);
2032
833f4077
PZ		 2033 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2034 ext4_count_free_blocks(sb));
2035 if (!err) {
2036 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2037 ext4_count_free_inodes(sb));
2038 }
2039 if (!err) {
2040 err = percpu_counter_init(&sbi->s_dirs_counter,
2041 ext4_count_dirs(sb));
2042 }
2043 if (err) {
2044 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2045 goto failed_mount3;
2046 }
ac27a0ec
DK		 2047
2048 /* per fileystem reservation list head & lock */
2049 spin_lock_init(&sbi->s_rsv_window_lock);
2050 sbi->s_rsv_window_root = RB_ROOT;
2051 /* Add a single, static dummy reservation to the start of the
2052 * reservation window list --- it gives us a placeholder for
2053 * append-at-start-of-list which makes the allocation logic
2054 * _much_ simpler. */
617ba13b
MC		 2055 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2056 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2057 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2058 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2059 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec
DK		 2060
2061 /*
2062 * set up enough so that it can read an inode
2063 */
617ba13b
MC		 2064 sb->s_op = &ext4_sops;
2065 sb->s_export_op = &ext4_export_ops;
2066 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2067#ifdef CONFIG_QUOTA
617ba13b
MC		 2068 sb->s_qcop = &ext4_qctl_operations;
2069 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2070#endif
2071 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2072
2073 sb->s_root = NULL;
2074
2075 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2076 EXT4_HAS_INCOMPAT_FEATURE(sb,
2077 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2078
2079 /*
2080 * The first inode we look at is the journal inode. Don't try
2081 * root first: it may be modified in the journal!
2082 */
2083 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2084 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2085 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 2086 goto failed_mount3;
2087 } else if (journal_inum) {
617ba13b 2088 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2089 goto failed_mount3;
2090 } else {
2091 if (!silent)
2092 printk (KERN_ERR
617ba13b 2093 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2094 sb->s_id);
2095 goto failed_mount3;
2096 }
2097
eb40a09c
JS		 2098 if (ext4_blocks_count(es) > 0xffffffffULL &&
2099 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2100 JBD2_FEATURE_INCOMPAT_64BIT)) {
2101 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2102 goto failed_mount4;
2103 }
2104
ac27a0ec
DK		 2105 /* We have now updated the journal if required, so we can
2106 * validate the data journaling mode. */
2107 switch (test_opt(sb, DATA_FLAGS)) {
2108 case 0:
2109 /* No mode set, assume a default based on the journal
63f57933
AM		 2110 * capabilities: ORDERED_DATA if the journal can
2111 * cope, else JOURNAL_DATA
2112 */
dab291af
MC		 2113 if (jbd2_journal_check_available_features
2114 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2115 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2116 else
2117 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2118 break;
2119
617ba13b
MC		 2120 case EXT4_MOUNT_ORDERED_DATA:
2121 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2122 if (!jbd2_journal_check_available_features
2123 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2124 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2125 "requested data journaling mode\n");
2126 goto failed_mount4;
2127 }
2128 default:
2129 break;
2130 }
2131
2132 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2133 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2134 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2135 "its supported only with writeback mode\n");
2136 clear_opt(sbi->s_mount_opt, NOBH);
2137 }
2138 }
2139 /*
dab291af 2140 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2141 * so we can safely mount the rest of the filesystem now.
2142 */
2143
617ba13b 2144 root = iget(sb, EXT4_ROOT_INO);
ac27a0ec
DK		 2145 sb->s_root = d_alloc_root(root);
2146 if (!sb->s_root) {
617ba13b 2147 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
ac27a0ec
DK		 2148 iput(root);
2149 goto failed_mount4;
2150 }
2151 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2152 dput(sb->s_root);
2153 sb->s_root = NULL;
617ba13b 2154 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2155 goto failed_mount4;
2156 }
2157
617ba13b 2158 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2159
2160 /* determine the minimum size of new large inodes, if present */
2161 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2162 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2163 EXT4_GOOD_OLD_INODE_SIZE;
2164 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2165 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2166 if (sbi->s_want_extra_isize <
2167 le16_to_cpu(es->s_want_extra_isize))
2168 sbi->s_want_extra_isize =
2169 le16_to_cpu(es->s_want_extra_isize);
2170 if (sbi->s_want_extra_isize <
2171 le16_to_cpu(es->s_min_extra_isize))
2172 sbi->s_want_extra_isize =
2173 le16_to_cpu(es->s_min_extra_isize);
2174 }
2175 }
2176 /* Check if enough inode space is available */
2177 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2178 sbi->s_inode_size) {
2179 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2180 EXT4_GOOD_OLD_INODE_SIZE;
2181 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2182 "available.\n");
2183 }
2184
ac27a0ec
DK		 2185 /*
2186 * akpm: core read_super() calls in here with the superblock locked.
2187 * That deadlocks, because orphan cleanup needs to lock the superblock
2188 * in numerous places. Here we just pop the lock - it's relatively
2189 * harmless, because we are now ready to accept write_super() requests,
2190 * and aviro says that's the only reason for hanging onto the
2191 * superblock lock.
2192 */
617ba13b
MC		 2193 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2194 ext4_orphan_cleanup(sb, es);
2195 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2196 if (needs_recovery)
617ba13b
MC		 2197 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2198 ext4_mark_recovery_complete(sb, es);
2199 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2200 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2201 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2202 "writeback");
2203
a86c6181
AT		 2204 ext4_ext_init(sb);
2205
ac27a0ec
DK		 2206 lock_kernel();
2207 return 0;
2208
617ba13b 2209cantfind_ext4:
ac27a0ec 2210 if (!silent)
617ba13b 2211 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2212 sb->s_id);
2213 goto failed_mount;
2214
2215failed_mount4:
dab291af 2216 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2217failed_mount3:
2218 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2219 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2220 percpu_counter_destroy(&sbi->s_dirs_counter);
2221failed_mount2:
2222 for (i = 0; i < db_count; i++)
2223 brelse(sbi->s_group_desc[i]);
2224 kfree(sbi->s_group_desc);
2225failed_mount:
2226#ifdef CONFIG_QUOTA
2227 for (i = 0; i < MAXQUOTAS; i++)
2228 kfree(sbi->s_qf_names[i]);
2229#endif
617ba13b 2230 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2231 brelse(bh);
2232out_fail:
2233 sb->s_fs_info = NULL;
2234 kfree(sbi);
2235 lock_kernel();
2236 return -EINVAL;
2237}
2238
2239/*
2240 * Setup any per-fs journal parameters now. We'll do this both on
2241 * initial mount, once the journal has been initialised but before we've
2242 * done any recovery; and again on any subsequent remount.
2243 */
617ba13b 2244static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2245{
617ba13b 2246 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2247
2248 if (sbi->s_commit_interval)
2249 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2250 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2251 * interval here, but for now we'll just fall back to the jbd
2252 * default. */
2253
2254 spin_lock(&journal->j_state_lock);
2255 if (test_opt(sb, BARRIER))
dab291af 2256 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2257 else
dab291af 2258 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2259 spin_unlock(&journal->j_state_lock);
2260}
2261
617ba13b 2262static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2263 unsigned int journal_inum)
2264{
2265 struct inode *journal_inode;
2266 journal_t *journal;
2267
2268 /* First, test for the existence of a valid inode on disk. Bad
2269 * things happen if we iget() an unused inode, as the subsequent
2270 * iput() will try to delete it. */
2271
2272 journal_inode = iget(sb, journal_inum);
2273 if (!journal_inode) {
617ba13b 2274 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2275 return NULL;
2276 }
2277 if (!journal_inode->i_nlink) {
2278 make_bad_inode(journal_inode);
2279 iput(journal_inode);
617ba13b 2280 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2281 return NULL;
2282 }
2283
2284 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2285 journal_inode, journal_inode->i_size);
2286 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
617ba13b 2287 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2288 iput(journal_inode);
2289 return NULL;
2290 }
2291
dab291af 2292 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2293 if (!journal) {
617ba13b 2294 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2295 iput(journal_inode);
2296 return NULL;
2297 }
2298 journal->j_private = sb;
617ba13b 2299 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2300 return journal;
2301}
2302
617ba13b 2303static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2304 dev_t j_dev)
2305{
2306 struct buffer_head * bh;
2307 journal_t *journal;
617ba13b
MC		 2308 ext4_fsblk_t start;
2309 ext4_fsblk_t len;
ac27a0ec 2310 int hblock, blocksize;
617ba13b 2311 ext4_fsblk_t sb_block;
ac27a0ec 2312 unsigned long offset;
617ba13b 2313 struct ext4_super_block * es;
ac27a0ec
DK		 2314 struct block_device *bdev;
2315
617ba13b 2316 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2317 if (bdev == NULL)
2318 return NULL;
2319
2320 if (bd_claim(bdev, sb)) {
2321 printk(KERN_ERR
8c55e204 2322 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2323 blkdev_put(bdev);
2324 return NULL;
2325 }
2326
2327 blocksize = sb->s_blocksize;
2328 hblock = bdev_hardsect_size(bdev);
2329 if (blocksize < hblock) {
2330 printk(KERN_ERR
617ba13b 2331 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2332 goto out_bdev;
2333 }
2334
617ba13b
MC		 2335 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2336 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2337 set_blocksize(bdev, blocksize);
2338 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2339 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2340 "external journal\n");
2341 goto out_bdev;
2342 }
2343
617ba13b
MC		 2344 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2345 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2346 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2347 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2348 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2349 "bad superblock\n");
2350 brelse(bh);
2351 goto out_bdev;
2352 }
2353
617ba13b
MC		 2354 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2355 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2356 brelse(bh);
2357 goto out_bdev;
2358 }
2359
bd81d8ee 2360 len = ext4_blocks_count(es);
ac27a0ec
DK		 2361 start = sb_block + 1;
2362 brelse(bh); /* we're done with the superblock */
2363
dab291af 2364 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2365 start, len, blocksize);
2366 if (!journal) {
617ba13b 2367 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2368 goto out_bdev;
2369 }
2370 journal->j_private = sb;
2371 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2372 wait_on_buffer(journal->j_sb_buffer);
2373 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2374 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2375 goto out_journal;
2376 }
2377 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2378 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2379 "user (unsupported) - %d\n",
2380 be32_to_cpu(journal->j_superblock->s_nr_users));
2381 goto out_journal;
2382 }
617ba13b
MC		 2383 EXT4_SB(sb)->journal_bdev = bdev;
2384 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2385 return journal;
2386out_journal:
dab291af 2387 jbd2_journal_destroy(journal);
ac27a0ec 2388out_bdev:
617ba13b 2389 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2390 return NULL;
2391}
2392
617ba13b
MC		 2393static int ext4_load_journal(struct super_block *sb,
2394 struct ext4_super_block *es,
ac27a0ec
DK		 2395 unsigned long journal_devnum)
2396{
2397 journal_t *journal;
2398 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2399 dev_t journal_dev;
2400 int err = 0;
2401 int really_read_only;
2402
2403 if (journal_devnum &&
2404 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2405 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2406 "numbers have changed\n");
2407 journal_dev = new_decode_dev(journal_devnum);
2408 } else
2409 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2410
2411 really_read_only = bdev_read_only(sb->s_bdev);
2412
2413 /*
2414 * Are we loading a blank journal or performing recovery after a
2415 * crash? For recovery, we need to check in advance whether we
2416 * can get read-write access to the device.
2417 */
2418
617ba13b 2419 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2420 if (sb->s_flags & MS_RDONLY) {
617ba13b 2421 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2422 "required on readonly filesystem.\n");
2423 if (really_read_only) {
617ba13b 2424 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2425 "unavailable, cannot proceed.\n");
2426 return -EROFS;
2427 }
617ba13b 2428 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2429 "be enabled during recovery.\n");
2430 }
2431 }
2432
2433 if (journal_inum && journal_dev) {
617ba13b 2434 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2435 "and inode journals!\n");
2436 return -EINVAL;
2437 }
2438
2439 if (journal_inum) {
617ba13b 2440 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2441 return -EINVAL;
2442 } else {
617ba13b 2443 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2444 return -EINVAL;
2445 }
2446
2447 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2448 err = jbd2_journal_update_format(journal);
ac27a0ec 2449 if (err) {
617ba13b 2450 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2451 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2452 return err;
2453 }
2454 }
2455
617ba13b 2456 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2457 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2458 if (!err)
dab291af 2459 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2460
2461 if (err) {
617ba13b 2462 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2463 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2464 return err;
2465 }
2466
617ba13b
MC		 2467 EXT4_SB(sb)->s_journal = journal;
2468 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2469
2470 if (journal_devnum &&
2471 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2472 es->s_journal_dev = cpu_to_le32(journal_devnum);
2473 sb->s_dirt = 1;
2474
2475 /* Make sure we flush the recovery flag to disk. */
617ba13b 2476 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2477 }
2478
2479 return 0;
2480}
2481
617ba13b
MC		 2482static int ext4_create_journal(struct super_block * sb,
2483 struct ext4_super_block * es,
ac27a0ec
DK		 2484 unsigned int journal_inum)
2485{
2486 journal_t *journal;
6c675bd4 2487 int err;
ac27a0ec
DK		 2488
2489 if (sb->s_flags & MS_RDONLY) {
617ba13b 2490 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2491 "create journal.\n");
2492 return -EROFS;
2493 }
2494
6c675bd4
BP		 2495 journal = ext4_get_journal(sb, journal_inum);
2496 if (!journal)
ac27a0ec
DK		 2497 return -EINVAL;
2498
617ba13b 2499 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2500 journal_inum);
2501
6c675bd4
BP		 2502 err = jbd2_journal_create(journal);
2503 if (err) {
617ba13b 2504 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2505 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2506 return -EIO;
2507 }
2508
617ba13b 2509 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2510
617ba13b
MC		 2511 ext4_update_dynamic_rev(sb);
2512 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2513 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2514
2515 es->s_journal_inum = cpu_to_le32(journal_inum);
2516 sb->s_dirt = 1;
2517
2518 /* Make sure we flush the recovery flag to disk. */
617ba13b 2519 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2520
2521 return 0;
2522}
2523
617ba13b
MC		 2524static void ext4_commit_super (struct super_block * sb,
2525 struct ext4_super_block * es,
ac27a0ec
DK		 2526 int sync)
2527{
617ba13b 2528 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2529
2530 if (!sbh)
2531 return;
2532 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2533 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2534 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2535 BUFFER_TRACE(sbh, "marking dirty");
2536 mark_buffer_dirty(sbh);
2537 if (sync)
2538 sync_dirty_buffer(sbh);
2539}
2540
2541
2542/*
2543 * Have we just finished recovery? If so, and if we are mounting (or
2544 * remounting) the filesystem readonly, then we will end up with a
2545 * consistent fs on disk. Record that fact.
2546 */
617ba13b
MC		 2547static void ext4_mark_recovery_complete(struct super_block * sb,
2548 struct ext4_super_block * es)
ac27a0ec 2549{
617ba13b 2550 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2551
dab291af
MC		 2552 jbd2_journal_lock_updates(journal);
2553 jbd2_journal_flush(journal);
32c37730 2554 lock_super(sb);
617ba13b 2555 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2556 sb->s_flags & MS_RDONLY) {
617ba13b 2557 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2558 sb->s_dirt = 0;
617ba13b 2559 ext4_commit_super(sb, es, 1);
ac27a0ec 2560 }
32c37730 2561 unlock_super(sb);
dab291af 2562 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2563}
2564
2565/*
2566 * If we are mounting (or read-write remounting) a filesystem whose journal
2567 * has recorded an error from a previous lifetime, move that error to the
2568 * main filesystem now.
2569 */
617ba13b
MC		 2570static void ext4_clear_journal_err(struct super_block * sb,
2571 struct ext4_super_block * es)
ac27a0ec
DK		 2572{
2573 journal_t *journal;
2574 int j_errno;
2575 const char *errstr;
2576
617ba13b 2577 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2578
2579 /*
2580 * Now check for any error status which may have been recorded in the
617ba13b 2581 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2582 */
2583
dab291af 2584 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2585 if (j_errno) {
2586 char nbuf[16];
2587
617ba13b
MC		 2588 errstr = ext4_decode_error(sb, j_errno, nbuf);
2589 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
ac27a0ec 2590 "from previous mount: %s", errstr);
617ba13b 2591 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
ac27a0ec
DK		 2592 "filesystem check.");
2593
617ba13b
MC		 2594 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2595 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2596 ext4_commit_super (sb, es, 1);
ac27a0ec 2597
dab291af 2598 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2599 }
2600}
2601
2602/*
2603 * Force the running and committing transactions to commit,
2604 * and wait on the commit.
2605 */
617ba13b 2606int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2607{
2608 journal_t *journal;
2609 int ret;
2610
2611 if (sb->s_flags & MS_RDONLY)
2612 return 0;
2613
617ba13b 2614 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2615 sb->s_dirt = 0;
617ba13b 2616 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2617 return ret;
2618}
2619
2620/*
617ba13b 2621 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2622 * have to propagate any other updates to the superblock on disk at this
2623 * point. Just start an async writeback to get the buffers on their way
2624 * to the disk.
2625 *
2626 * This implicitly triggers the writebehind on sync().
2627 */
2628
617ba13b 2629static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2630{
2631 if (mutex_trylock(&sb->s_lock) != 0)
2632 BUG();
2633 sb->s_dirt = 0;
2634}
2635
617ba13b 2636static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2637{
2638 tid_t target;
2639
2640 sb->s_dirt = 0;
dab291af 2641 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2642 if (wait)
dab291af 2643 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2644 }
2645 return 0;
2646}
2647
2648/*
2649 * LVM calls this function before a (read-only) snapshot is created. This
2650 * gives us a chance to flush the journal completely and mark the fs clean.
2651 */
617ba13b 2652static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2653{
2654 sb->s_dirt = 0;
2655
2656 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2657 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2658
2659 /* Now we set up the journal barrier. */
dab291af
MC		 2660 jbd2_journal_lock_updates(journal);
2661 jbd2_journal_flush(journal);
ac27a0ec
DK		 2662
2663 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2664 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2665 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2666 }
2667}
2668
2669/*
2670 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2671 * flag here, even though the filesystem is not technically dirty yet.
2672 */
617ba13b 2673static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2674{
2675 if (!(sb->s_flags & MS_RDONLY)) {
2676 lock_super(sb);
2677 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2678 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2679 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2680 unlock_super(sb);
dab291af 2681 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2682 }
2683}
2684
617ba13b 2685static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2686{
617ba13b
MC		 2687 struct ext4_super_block * es;
2688 struct ext4_sb_info *sbi = EXT4_SB(sb);
2689 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2690 unsigned long old_sb_flags;
617ba13b 2691 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2692 int err;
2693#ifdef CONFIG_QUOTA
2694 int i;
2695#endif
2696
2697 /* Store the original options */
2698 old_sb_flags = sb->s_flags;
2699 old_opts.s_mount_opt = sbi->s_mount_opt;
2700 old_opts.s_resuid = sbi->s_resuid;
2701 old_opts.s_resgid = sbi->s_resgid;
2702 old_opts.s_commit_interval = sbi->s_commit_interval;
2703#ifdef CONFIG_QUOTA
2704 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2705 for (i = 0; i < MAXQUOTAS; i++)
2706 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2707#endif
2708
2709 /*
2710 * Allow the "check" option to be passed as a remount option.
2711 */
2712 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2713 err = -EINVAL;
2714 goto restore_opts;
2715 }
2716
617ba13b
MC		 2717 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2718 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
ac27a0ec
DK		 2719
2720 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2721 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2722
2723 es = sbi->s_es;
2724
617ba13b 2725 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2726
2727 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2728 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2729 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2730 err = -EROFS;
2731 goto restore_opts;
2732 }
2733
2734 if (*flags & MS_RDONLY) {
2735 /*
2736 * First of all, the unconditional stuff we have to do
2737 * to disable replay of the journal when we next remount
2738 */
2739 sb->s_flags |= MS_RDONLY;
2740
2741 /*
2742 * OK, test if we are remounting a valid rw partition
2743 * readonly, and if so set the rdonly flag and then
2744 * mark the partition as valid again.
2745 */
617ba13b
MC		 2746 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2747 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2748 es->s_state = cpu_to_le16(sbi->s_mount_state);
2749
32c37730
JK		 2750 /*
2751 * We have to unlock super so that we can wait for
2752 * transactions.
2753 */
2754 unlock_super(sb);
617ba13b 2755 ext4_mark_recovery_complete(sb, es);
32c37730 2756 lock_super(sb);
ac27a0ec
DK		 2757 } else {
2758 __le32 ret;
617ba13b
MC		 2759 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2760 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2761 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2762 "remount RDWR because of unsupported "
2763 "optional features (%x).\n",
2764 sb->s_id, le32_to_cpu(ret));
2765 err = -EROFS;
2766 goto restore_opts;
2767 }
ead6596b
ES		 2768
2769 /*
2770 * If we have an unprocessed orphan list hanging
2771 * around from a previously readonly bdev mount,
2772 * require a full umount/remount for now.
2773 */
2774 if (es->s_last_orphan) {
2775 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2776 "remount RDWR because of unprocessed "
2777 "orphan inode list. Please "
2778 "umount/remount instead.\n",
2779 sb->s_id);
2780 err = -EINVAL;
2781 goto restore_opts;
2782 }
2783
ac27a0ec
DK		 2784 /*
2785 * Mounting a RDONLY partition read-write, so reread
2786 * and store the current valid flag. (It may have
2787 * been changed by e2fsck since we originally mounted
2788 * the partition.)
2789 */
617ba13b 2790 ext4_clear_journal_err(sb, es);
ac27a0ec 2791 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2792 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2793 goto restore_opts;
617ba13b 2794 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2795 sb->s_flags &= ~MS_RDONLY;
2796 }
2797 }
2798#ifdef CONFIG_QUOTA
2799 /* Release old quota file names */
2800 for (i = 0; i < MAXQUOTAS; i++)
2801 if (old_opts.s_qf_names[i] &&
2802 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2803 kfree(old_opts.s_qf_names[i]);
2804#endif
2805 return 0;
2806restore_opts:
2807 sb->s_flags = old_sb_flags;
2808 sbi->s_mount_opt = old_opts.s_mount_opt;
2809 sbi->s_resuid = old_opts.s_resuid;
2810 sbi->s_resgid = old_opts.s_resgid;
2811 sbi->s_commit_interval = old_opts.s_commit_interval;
2812#ifdef CONFIG_QUOTA
2813 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2814 for (i = 0; i < MAXQUOTAS; i++) {
2815 if (sbi->s_qf_names[i] &&
2816 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2817 kfree(sbi->s_qf_names[i]);
2818 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2819 }
2820#endif
2821 return err;
2822}
2823
617ba13b 2824static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2825{
2826 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2827 struct ext4_sb_info *sbi = EXT4_SB(sb);
2828 struct ext4_super_block *es = sbi->s_es;
960cc398 2829 u64 fsid;
ac27a0ec 2830
5e70030d
BP		 2831 if (test_opt(sb, MINIX_DF)) {
2832 sbi->s_overhead_last = 0;
6bc9feff 2833 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 2834 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 2835 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2836 smp_rmb();
2837
2838 /*
5e70030d
BP		 2839 * Compute the overhead (FS structures). This is constant
2840 * for a given filesystem unless the number of block groups
2841 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2842 */
2843
2844 /*
2845 * All of the blocks before first_data_block are
2846 * overhead
2847 */
2848 overhead = le32_to_cpu(es->s_first_data_block);
2849
2850 /*
2851 * Add the overhead attributed to the superblock and
2852 * block group descriptors. If the sparse superblocks
2853 * feature is turned on, then not all groups have this.
2854 */
2855 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2856 overhead += ext4_bg_has_super(sb, i) +
2857 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2858 cond_resched();
2859 }
2860
2861 /*
2862 * Every block group has an inode bitmap, a block
2863 * bitmap, and an inode table.
2864 */
5e70030d
BP		 2865 overhead += ngroups * (2 + sbi->s_itb_per_group);
2866 sbi->s_overhead_last = overhead;
2867 smp_wmb();
6bc9feff 2868 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 2869 }
2870
617ba13b 2871 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2872 buf->f_bsize = sb->s_blocksize;
5e70030d 2873 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 2874 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 2875 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 2876 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2877 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 2878 buf->f_bavail = 0;
2879 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2880 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 2881 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 2882 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 2883 fsid = le64_to_cpup((void *)es->s_uuid) ^
2884 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2885 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2886 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 2887 return 0;
2888}
2889
2890/* Helper function for writing quotas on sync - we need to start transaction before quota file
2891 * is locked for write. Otherwise the are possible deadlocks:
2892 * Process 1 Process 2
617ba13b 2893 * ext4_create() quota_sync()
dab291af 2894 * jbd2_journal_start() write_dquot()
ac27a0ec 2895 * DQUOT_INIT() down(dqio_mutex)
dab291af 2896 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 2897 *
2898 */
2899
2900#ifdef CONFIG_QUOTA
2901
2902static inline struct inode *dquot_to_inode(struct dquot *dquot)
2903{
2904 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2905}
2906
617ba13b 2907static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 2908{
2909 handle_t *handle;
2910 int ret, err;
2911
2912 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 2913 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2914 if (IS_ERR(handle))
2915 return PTR_ERR(handle);
2916 ret = dquot_initialize(inode, type);
617ba13b 2917 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2918 if (!ret)
2919 ret = err;
2920 return ret;
2921}
2922
617ba13b 2923static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 2924{
2925 handle_t *handle;
2926 int ret, err;
2927
2928 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 2929 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2930 if (IS_ERR(handle))
2931 return PTR_ERR(handle);
2932 ret = dquot_drop(inode);
617ba13b 2933 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2934 if (!ret)
2935 ret = err;
2936 return ret;
2937}
2938
617ba13b 2939static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 2940{
2941 int ret, err;
2942 handle_t *handle;
2943 struct inode *inode;
2944
2945 inode = dquot_to_inode(dquot);
617ba13b
MC		 2946 handle = ext4_journal_start(inode,
2947 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2948 if (IS_ERR(handle))
2949 return PTR_ERR(handle);
2950 ret = dquot_commit(dquot);
617ba13b 2951 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2952 if (!ret)
2953 ret = err;
2954 return ret;
2955}
2956
617ba13b 2957static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 2958{
2959 int ret, err;
2960 handle_t *handle;
2961
617ba13b
MC		 2962 handle = ext4_journal_start(dquot_to_inode(dquot),
2963 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2964 if (IS_ERR(handle))
2965 return PTR_ERR(handle);
2966 ret = dquot_acquire(dquot);
617ba13b 2967 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2968 if (!ret)
2969 ret = err;
2970 return ret;
2971}
2972
617ba13b 2973static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 2974{
2975 int ret, err;
2976 handle_t *handle;
2977
617ba13b
MC		 2978 handle = ext4_journal_start(dquot_to_inode(dquot),
2979 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2980 if (IS_ERR(handle)) {
2981 /* Release dquot anyway to avoid endless cycle in dqput() */
2982 dquot_release(dquot);
ac27a0ec 2983 return PTR_ERR(handle);
9c3013e9 2984 }
ac27a0ec 2985 ret = dquot_release(dquot);
617ba13b 2986 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2987 if (!ret)
2988 ret = err;
2989 return ret;
2990}
2991
617ba13b 2992static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 2993{
2994 /* Are we journalling quotas? */
617ba13b
MC		 2995 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2996 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 2997 dquot_mark_dquot_dirty(dquot);
617ba13b 2998 return ext4_write_dquot(dquot);
ac27a0ec
DK		 2999 } else {
3000 return dquot_mark_dquot_dirty(dquot);
3001 }
3002}
3003
617ba13b 3004static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3005{
3006 int ret, err;
3007 handle_t *handle;
3008
3009 /* Data block + inode block */
617ba13b 3010 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3011 if (IS_ERR(handle))
3012 return PTR_ERR(handle);
3013 ret = dquot_commit_info(sb, type);
617ba13b 3014 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3015 if (!ret)
3016 ret = err;
3017 return ret;
3018}
3019
3020/*
3021 * Turn on quotas during mount time - we need to find
3022 * the quota file and such...
3023 */
617ba13b 3024static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3025{
617ba13b
MC		 3026 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3027 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3028}
3029
3030/*
3031 * Standard function to be called on quota_on
3032 */
617ba13b 3033static int ext4_quota_on(struct super_block *sb, int type, int format_id,
ac27a0ec
DK		 3034 char *path)
3035{
3036 int err;
3037 struct nameidata nd;
3038
3039 if (!test_opt(sb, QUOTA))
3040 return -EINVAL;
3041 /* Not journalling quota? */
617ba13b
MC		 3042 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3043 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 3044 return vfs_quota_on(sb, type, format_id, path);
3045 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3046 if (err)
3047 return err;
3048 /* Quotafile not on the same filesystem? */
3049 if (nd.mnt->mnt_sb != sb) {
3050 path_release(&nd);
3051 return -EXDEV;
3052 }
3053 /* Quotafile not of fs root? */
3054 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
3055 printk(KERN_WARNING
617ba13b 3056 "EXT4-fs: Quota file not on filesystem root. "
ac27a0ec
DK		 3057 "Journalled quota will not work.\n");
3058 path_release(&nd);
3059 return vfs_quota_on(sb, type, format_id, path);
3060}
3061
3062/* Read data from quotafile - avoid pagecache and such because we cannot afford
3063 * acquiring the locks... As quota files are never truncated and quota code
3064 * itself serializes the operations (and noone else should touch the files)
3065 * we don't have to be afraid of races */
617ba13b 3066static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3067 size_t len, loff_t off)
3068{
3069 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3070 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3071 int err = 0;
3072 int offset = off & (sb->s_blocksize - 1);
3073 int tocopy;
3074 size_t toread;
3075 struct buffer_head *bh;
3076 loff_t i_size = i_size_read(inode);
3077
3078 if (off > i_size)
3079 return 0;
3080 if (off+len > i_size)
3081 len = i_size-off;
3082 toread = len;
3083 while (toread > 0) {
3084 tocopy = sb->s_blocksize - offset < toread ?
3085 sb->s_blocksize - offset : toread;
617ba13b 3086 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3087 if (err)
3088 return err;
3089 if (!bh) /* A hole? */
3090 memset(data, 0, tocopy);
3091 else
3092 memcpy(data, bh->b_data+offset, tocopy);
3093 brelse(bh);
3094 offset = 0;
3095 toread -= tocopy;
3096 data += tocopy;
3097 blk++;
3098 }
3099 return len;
3100}
3101
3102/* Write to quotafile (we know the transaction is already started and has
3103 * enough credits) */
617ba13b 3104static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3105 const char *data, size_t len, loff_t off)
3106{
3107 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3108 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3109 int err = 0;
3110 int offset = off & (sb->s_blocksize - 1);
3111 int tocopy;
617ba13b 3112 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3113 size_t towrite = len;
3114 struct buffer_head *bh;
3115 handle_t *handle = journal_current_handle();
3116
9c3013e9
JK		 3117 if (!handle) {
3118 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3119 " cancelled because transaction is not started.\n",
3120 (unsigned long long)off, (unsigned long long)len);
3121 return -EIO;
3122 }
ac27a0ec
DK		 3123 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3124 while (towrite > 0) {
3125 tocopy = sb->s_blocksize - offset < towrite ?
3126 sb->s_blocksize - offset : towrite;
617ba13b 3127 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3128 if (!bh)
3129 goto out;
3130 if (journal_quota) {
617ba13b 3131 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3132 if (err) {
3133 brelse(bh);
3134 goto out;
3135 }
3136 }
3137 lock_buffer(bh);
3138 memcpy(bh->b_data+offset, data, tocopy);
3139 flush_dcache_page(bh->b_page);
3140 unlock_buffer(bh);
3141 if (journal_quota)
617ba13b 3142 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3143 else {
3144 /* Always do at least ordered writes for quotas */
617ba13b 3145 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 3146 mark_buffer_dirty(bh);
3147 }
3148 brelse(bh);
3149 if (err)
3150 goto out;
3151 offset = 0;
3152 towrite -= tocopy;
3153 data += tocopy;
3154 blk++;
3155 }
3156out:
3157 if (len == towrite)
3158 return err;
3159 if (inode->i_size < off+len-towrite) {
3160 i_size_write(inode, off+len-towrite);
617ba13b 3161 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec
DK		 3162 }
3163 inode->i_version++;
3164 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3165 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3166 mutex_unlock(&inode->i_mutex);
3167 return len - towrite;
3168}
3169
3170#endif
3171
617ba13b 3172static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3173 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3174{
617ba13b 3175 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3176}
3177
617ba13b 3178static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3179 .owner = THIS_MODULE,
617ba13b
MC		 3180 .name = "ext4dev",
3181 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3182 .kill_sb = kill_block_super,
3183 .fs_flags = FS_REQUIRES_DEV,
3184};
3185
617ba13b 3186static int __init init_ext4_fs(void)
ac27a0ec 3187{
617ba13b 3188 int err = init_ext4_xattr();
ac27a0ec
DK		 3189 if (err)
3190 return err;
3191 err = init_inodecache();
3192 if (err)
3193 goto out1;
63f57933 3194 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3195 if (err)
3196 goto out;
3197 return 0;
3198out:
3199 destroy_inodecache();
3200out1:
617ba13b 3201 exit_ext4_xattr();
ac27a0ec
DK		 3202 return err;
3203}
3204
617ba13b 3205static void __exit exit_ext4_fs(void)
ac27a0ec 3206{
617ba13b 3207 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3208 destroy_inodecache();
617ba13b 3209 exit_ext4_xattr();
ac27a0ec
DK		 3210}
3211
3212MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3213MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3214MODULE_LICENSE("GPL");
617ba13b
MC		 3215module_init(init_ext4_fs)
3216module_exit(exit_ext4_fs)
ubuntu-bionic-kernel mirror