blob: 425c2f2cf1700a3f5d9db8fc4e93dcde0450d998 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Uwe Zeisbergerf30c2262006-10-03 23:01:26 +02002 * linux/fs/jbd/journal.c
Linus Torvalds1da177e2005-04-16 15:20:36 -07003 *
4 * Written by Stephen C. Tweedie <sct@redhat.com>, 1998
5 *
6 * Copyright 1998 Red Hat corp --- All Rights Reserved
7 *
8 * This file is part of the Linux kernel and is made available under
9 * the terms of the GNU General Public License, version 2, or at your
10 * option, any later version, incorporated herein by reference.
11 *
12 * Generic filesystem journal-writing code; part of the ext2fs
13 * journaling system.
14 *
15 * This file manages journals: areas of disk reserved for logging
16 * transactional updates. This includes the kernel journaling thread
17 * which is responsible for scheduling updates to the log.
18 *
19 * We do not actually manage the physical storage of the journal in this
20 * file: that is left to a per-journal policy function, which allows us
21 * to store the journal within a filesystem-specified area for ext2
22 * journaling (ext2 can use a reserved inode for storing the log).
23 */
24
25#include <linux/module.h>
26#include <linux/time.h>
27#include <linux/fs.h>
28#include <linux/jbd.h>
29#include <linux/errno.h>
30#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/init.h>
32#include <linux/mm.h>
Nigel Cunningham7dfb7102006-12-06 20:34:23 -080033#include <linux/freezer.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/pagemap.h>
Andrew Morton8d8c8512006-03-25 03:06:53 -080035#include <linux/kthread.h>
Randy Dunlapc9cf5522006-06-27 02:53:52 -070036#include <linux/poison.h>
Andrew Morton8d8c8512006-03-25 03:06:53 -080037#include <linux/proc_fs.h>
Jose R. Santosc2a91592007-10-18 23:39:22 -070038#include <linux/debugfs.h>
Namhyung Kimf81e3d42010-10-04 19:12:13 +090039#include <linux/ratelimit.h>
Andrew Morton8d8c8512006-03-25 03:06:53 -080040
Lukas Czerner99cb1a32011-05-23 18:33:02 +020041#define CREATE_TRACE_POINTS
42#include <trace/events/jbd.h>
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <asm/uaccess.h>
45#include <asm/page.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
47EXPORT_SYMBOL(journal_start);
48EXPORT_SYMBOL(journal_restart);
49EXPORT_SYMBOL(journal_extend);
50EXPORT_SYMBOL(journal_stop);
51EXPORT_SYMBOL(journal_lock_updates);
52EXPORT_SYMBOL(journal_unlock_updates);
53EXPORT_SYMBOL(journal_get_write_access);
54EXPORT_SYMBOL(journal_get_create_access);
55EXPORT_SYMBOL(journal_get_undo_access);
56EXPORT_SYMBOL(journal_dirty_data);
57EXPORT_SYMBOL(journal_dirty_metadata);
58EXPORT_SYMBOL(journal_release_buffer);
59EXPORT_SYMBOL(journal_forget);
60#if 0
61EXPORT_SYMBOL(journal_sync_buffer);
62#endif
63EXPORT_SYMBOL(journal_flush);
64EXPORT_SYMBOL(journal_revoke);
65
66EXPORT_SYMBOL(journal_init_dev);
67EXPORT_SYMBOL(journal_init_inode);
68EXPORT_SYMBOL(journal_update_format);
69EXPORT_SYMBOL(journal_check_used_features);
70EXPORT_SYMBOL(journal_check_available_features);
71EXPORT_SYMBOL(journal_set_features);
72EXPORT_SYMBOL(journal_create);
73EXPORT_SYMBOL(journal_load);
74EXPORT_SYMBOL(journal_destroy);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075EXPORT_SYMBOL(journal_abort);
76EXPORT_SYMBOL(journal_errno);
77EXPORT_SYMBOL(journal_ack_err);
78EXPORT_SYMBOL(journal_clear_err);
79EXPORT_SYMBOL(log_wait_commit);
Stefan Schmidtff5e4b52009-11-12 09:53:50 +010080EXPORT_SYMBOL(log_start_commit);
Linus Torvalds1da177e2005-04-16 15:20:36 -070081EXPORT_SYMBOL(journal_start_commit);
82EXPORT_SYMBOL(journal_force_commit_nested);
83EXPORT_SYMBOL(journal_wipe);
84EXPORT_SYMBOL(journal_blocks_per_page);
85EXPORT_SYMBOL(journal_invalidatepage);
86EXPORT_SYMBOL(journal_try_to_free_buffers);
87EXPORT_SYMBOL(journal_force_commit);
88
89static int journal_convert_superblock_v1(journal_t *, journal_superblock_t *);
Adrian Bunk022a4a72005-09-06 15:16:41 -070090static void __journal_abort_soft (journal_t *journal, int errno);
Darrick J. Wongdff68252010-10-04 12:35:05 -070091static const char *journal_dev_name(journal_t *journal, char *buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93/*
94 * Helper function used to manage commit timeouts
95 */
96
97static void commit_timeout(unsigned long __data)
98{
99 struct task_struct * p = (struct task_struct *) __data;
100
101 wake_up_process(p);
102}
103
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104/*
105 * kjournald: The main thread function used to manage a logging device
106 * journal.
107 *
108 * This kernel thread is responsible for two things:
109 *
110 * 1) COMMIT: Every so often we need to commit the current state of the
111 * filesystem to disk. The journal thread is responsible for writing
112 * all of the metadata buffers to disk.
113 *
114 * 2) CHECKPOINT: We cannot reuse a used section of the log file until all
115 * of the data in that part of the log has been rewritten elsewhere on
116 * the disk. Flushing these old buffers to reclaim space in the log is
117 * known as checkpointing, and this thread is responsible for that job.
118 */
119
Adrian Bunk022a4a72005-09-06 15:16:41 -0700120static int kjournald(void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121{
Andrew Mortone3df1892006-03-25 03:06:53 -0800122 journal_t *journal = arg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 transaction_t *transaction;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
Andrew Mortone3df1892006-03-25 03:06:53 -0800125 /*
126 * Set up an interval timer which can be used to trigger a commit wakeup
127 * after the commit interval expires
128 */
129 setup_timer(&journal->j_commit_timer, commit_timeout,
130 (unsigned long)current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
Nigel Cunningham35c80422012-02-03 19:59:41 +1100132 set_freezable();
133
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 /* Record that the journal thread is running */
135 journal->j_task = current;
136 wake_up(&journal->j_wait_done_commit);
137
138 printk(KERN_INFO "kjournald starting. Commit interval %ld seconds\n",
139 journal->j_commit_interval / HZ);
140
141 /*
142 * And now, wait forever for commit wakeup events.
143 */
144 spin_lock(&journal->j_state_lock);
145
146loop:
147 if (journal->j_flags & JFS_UNMOUNT)
148 goto end_loop;
149
150 jbd_debug(1, "commit_sequence=%d, commit_request=%d\n",
151 journal->j_commit_sequence, journal->j_commit_request);
152
153 if (journal->j_commit_sequence != journal->j_commit_request) {
154 jbd_debug(1, "OK, requests differ\n");
155 spin_unlock(&journal->j_state_lock);
Andrew Mortone3df1892006-03-25 03:06:53 -0800156 del_timer_sync(&journal->j_commit_timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 journal_commit_transaction(journal);
158 spin_lock(&journal->j_state_lock);
159 goto loop;
160 }
161
162 wake_up(&journal->j_wait_done_commit);
Christoph Lameter3e1d1d22005-06-24 23:13:50 -0700163 if (freezing(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 /*
165 * The simpler the better. Flushing journal isn't a
166 * good idea, because that depends on threads that may
167 * be already stopped.
168 */
169 jbd_debug(1, "Now suspending kjournald\n");
170 spin_unlock(&journal->j_state_lock);
Tejun Heoa0acae02011-11-21 12:32:22 -0800171 try_to_freeze();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 spin_lock(&journal->j_state_lock);
173 } else {
174 /*
175 * We assume on resume that commits are already there,
176 * so we don't sleep
177 */
178 DEFINE_WAIT(wait);
179 int should_sleep = 1;
180
181 prepare_to_wait(&journal->j_wait_commit, &wait,
182 TASK_INTERRUPTIBLE);
183 if (journal->j_commit_sequence != journal->j_commit_request)
184 should_sleep = 0;
185 transaction = journal->j_running_transaction;
186 if (transaction && time_after_eq(jiffies,
187 transaction->t_expires))
188 should_sleep = 0;
Mark Fashehcbf0d272005-09-06 15:19:08 -0700189 if (journal->j_flags & JFS_UNMOUNT)
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700190 should_sleep = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 if (should_sleep) {
192 spin_unlock(&journal->j_state_lock);
193 schedule();
194 spin_lock(&journal->j_state_lock);
195 }
196 finish_wait(&journal->j_wait_commit, &wait);
197 }
198
199 jbd_debug(1, "kjournald wakes\n");
200
201 /*
202 * Were we woken up by a commit wakeup event?
203 */
204 transaction = journal->j_running_transaction;
205 if (transaction && time_after_eq(jiffies, transaction->t_expires)) {
206 journal->j_commit_request = transaction->t_tid;
207 jbd_debug(1, "woke because of timeout\n");
208 }
209 goto loop;
210
211end_loop:
212 spin_unlock(&journal->j_state_lock);
Andrew Mortone3df1892006-03-25 03:06:53 -0800213 del_timer_sync(&journal->j_commit_timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 journal->j_task = NULL;
215 wake_up(&journal->j_wait_done_commit);
216 jbd_debug(1, "Journal thread exiting.\n");
217 return 0;
218}
219
Pavel Emelianov97f06782007-05-08 00:30:42 -0700220static int journal_start_thread(journal_t *journal)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221{
Pavel Emelianov97f06782007-05-08 00:30:42 -0700222 struct task_struct *t;
223
224 t = kthread_run(kjournald, journal, "kjournald");
225 if (IS_ERR(t))
226 return PTR_ERR(t);
227
Stephen Hemmingerc80544d2007-10-18 03:07:05 -0700228 wait_event(journal->j_wait_done_commit, journal->j_task != NULL);
Pavel Emelianov97f06782007-05-08 00:30:42 -0700229 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230}
231
232static void journal_kill_thread(journal_t *journal)
233{
234 spin_lock(&journal->j_state_lock);
235 journal->j_flags |= JFS_UNMOUNT;
236
237 while (journal->j_task) {
238 wake_up(&journal->j_wait_commit);
239 spin_unlock(&journal->j_state_lock);
Stephen Hemmingerc80544d2007-10-18 03:07:05 -0700240 wait_event(journal->j_wait_done_commit,
241 journal->j_task == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 spin_lock(&journal->j_state_lock);
243 }
244 spin_unlock(&journal->j_state_lock);
245}
246
247/*
248 * journal_write_metadata_buffer: write a metadata buffer to the journal.
249 *
250 * Writes a metadata buffer to a given disk block. The actual IO is not
251 * performed but a new buffer_head is constructed which labels the data
252 * to be written with the correct destination disk block.
253 *
254 * Any magic-number escaping which needs to be done will cause a
255 * copy-out here. If the buffer happens to start with the
256 * JFS_MAGIC_NUMBER, then we can't write it to the log directly: the
257 * magic number is only written to the log for descripter blocks. In
258 * this case, we copy the data and replace the first word with 0, and we
259 * return a result code which indicates that this buffer needs to be
260 * marked as an escaped buffer in the corresponding log descriptor
261 * block. The missing word can then be restored when the block is read
262 * during recovery.
263 *
264 * If the source buffer has already been modified by a new transaction
265 * since we took the last commit snapshot, we use the frozen copy of
266 * that data for IO. If we end up using the existing buffer_head's data
267 * for the write, then we *have* to lock the buffer to prevent anyone
268 * else from using and possibly modifying it while the IO is in
269 * progress.
270 *
271 * The function returns a pointer to the buffer_heads to be used for IO.
272 *
273 * We assume that the journal has already been locked in this function.
274 *
275 * Return value:
276 * <0: Error
277 * >=0: Finished OK
278 *
279 * On success:
280 * Bit 0 set == escape performed on the data
281 * Bit 1 set == buffer copy-out performed (kfree the data after IO)
282 */
283
284int journal_write_metadata_buffer(transaction_t *transaction,
285 struct journal_head *jh_in,
286 struct journal_head **jh_out,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200287 unsigned int blocknr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288{
289 int need_copy_out = 0;
290 int done_copy_out = 0;
291 int do_escape = 0;
292 char *mapped_data;
293 struct buffer_head *new_bh;
294 struct journal_head *new_jh;
295 struct page *new_page;
296 unsigned int new_offset;
297 struct buffer_head *bh_in = jh2bh(jh_in);
dingdinghuaf1015c42009-07-15 21:42:05 +0200298 journal_t *journal = transaction->t_journal;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
300 /*
301 * The buffer really shouldn't be locked: only the current committing
302 * transaction is allowed to write it, so nobody else is allowed
303 * to do any IO.
304 *
305 * akpm: except if we're journalling data, and write() output is
306 * also part of a shared mapping, and another thread has
307 * decided to launch a writepage() against this buffer.
308 */
309 J_ASSERT_BH(bh_in, buffer_jbddirty(bh_in));
310
311 new_bh = alloc_buffer_head(GFP_NOFS|__GFP_NOFAIL);
dingdinghuaf1015c42009-07-15 21:42:05 +0200312 /* keep subsequent assertions sane */
313 new_bh->b_state = 0;
314 init_buffer(new_bh, NULL, NULL);
315 atomic_set(&new_bh->b_count, 1);
316 new_jh = journal_add_journal_head(new_bh); /* This sleeps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
318 /*
319 * If a new transaction has already done a buffer copy-out, then
320 * we use that version of the data for the commit.
321 */
322 jbd_lock_bh_state(bh_in);
323repeat:
324 if (jh_in->b_frozen_data) {
325 done_copy_out = 1;
326 new_page = virt_to_page(jh_in->b_frozen_data);
327 new_offset = offset_in_page(jh_in->b_frozen_data);
328 } else {
329 new_page = jh2bh(jh_in)->b_page;
330 new_offset = offset_in_page(jh2bh(jh_in)->b_data);
331 }
332
Cong Wang8fb53c42011-11-25 23:14:31 +0800333 mapped_data = kmap_atomic(new_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 /*
335 * Check for escaping
336 */
337 if (*((__be32 *)(mapped_data + new_offset)) ==
338 cpu_to_be32(JFS_MAGIC_NUMBER)) {
339 need_copy_out = 1;
340 do_escape = 1;
341 }
Cong Wang8fb53c42011-11-25 23:14:31 +0800342 kunmap_atomic(mapped_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
344 /*
345 * Do we need to do a data copy?
346 */
347 if (need_copy_out && !done_copy_out) {
348 char *tmp;
349
350 jbd_unlock_bh_state(bh_in);
Mingming Caoc089d492007-10-16 18:38:25 -0400351 tmp = jbd_alloc(bh_in->b_size, GFP_NOFS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 jbd_lock_bh_state(bh_in);
353 if (jh_in->b_frozen_data) {
Mingming Caoc089d492007-10-16 18:38:25 -0400354 jbd_free(tmp, bh_in->b_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 goto repeat;
356 }
357
358 jh_in->b_frozen_data = tmp;
Cong Wang8fb53c42011-11-25 23:14:31 +0800359 mapped_data = kmap_atomic(new_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 memcpy(tmp, mapped_data + new_offset, jh2bh(jh_in)->b_size);
Cong Wang8fb53c42011-11-25 23:14:31 +0800361 kunmap_atomic(mapped_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
363 new_page = virt_to_page(tmp);
364 new_offset = offset_in_page(tmp);
365 done_copy_out = 1;
366 }
367
368 /*
369 * Did we need to do an escaping? Now we've done all the
370 * copying, we can finally do so.
371 */
372 if (do_escape) {
Cong Wang8fb53c42011-11-25 23:14:31 +0800373 mapped_data = kmap_atomic(new_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 *((unsigned int *)(mapped_data + new_offset)) = 0;
Cong Wang8fb53c42011-11-25 23:14:31 +0800375 kunmap_atomic(mapped_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 }
377
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 set_bh_page(new_bh, new_page, new_offset);
379 new_jh->b_transaction = NULL;
380 new_bh->b_size = jh2bh(jh_in)->b_size;
381 new_bh->b_bdev = transaction->t_journal->j_dev;
382 new_bh->b_blocknr = blocknr;
383 set_buffer_mapped(new_bh);
384 set_buffer_dirty(new_bh);
385
386 *jh_out = new_jh;
387
388 /*
389 * The to-be-written buffer needs to get moved to the io queue,
390 * and the original buffer whose contents we are shadowing or
391 * copying is moved to the transaction's shadow queue.
392 */
393 JBUFFER_TRACE(jh_in, "file as BJ_Shadow");
dingdinghuaf1015c42009-07-15 21:42:05 +0200394 spin_lock(&journal->j_list_lock);
395 __journal_file_buffer(jh_in, transaction, BJ_Shadow);
396 spin_unlock(&journal->j_list_lock);
397 jbd_unlock_bh_state(bh_in);
398
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 JBUFFER_TRACE(new_jh, "file as BJ_IO");
400 journal_file_buffer(new_jh, transaction, BJ_IO);
401
402 return do_escape | (done_copy_out << 1);
403}
404
405/*
406 * Allocation code for the journal file. Manage the space left in the
407 * journal, so that we can begin checkpointing when appropriate.
408 */
409
410/*
411 * __log_space_left: Return the number of free blocks left in the journal.
412 *
413 * Called with the journal already locked.
414 *
415 * Called under j_state_lock
416 */
417
418int __log_space_left(journal_t *journal)
419{
420 int left = journal->j_free;
421
422 assert_spin_locked(&journal->j_state_lock);
423
424 /*
425 * Be pessimistic here about the number of those free blocks which
426 * might be required for log descriptor control blocks.
427 */
428
429#define MIN_LOG_RESERVED_BLOCKS 32 /* Allow for rounding errors */
430
431 left -= MIN_LOG_RESERVED_BLOCKS;
432
433 if (left <= 0)
434 return 0;
435 left -= (left >> 3);
436 return left;
437}
438
439/*
Jan Kara8fe4cd02009-02-11 13:04:25 -0800440 * Called under j_state_lock. Returns true if a transaction commit was started.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 */
442int __log_start_commit(journal_t *journal, tid_t target)
443{
444 /*
Ted Ts'od9b01932011-04-30 13:17:11 -0400445 * The only transaction we can possibly wait upon is the
446 * currently running transaction (if it exists). Otherwise,
447 * the target tid must be an old one.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 */
Ted Ts'od9b01932011-04-30 13:17:11 -0400449 if (journal->j_running_transaction &&
450 journal->j_running_transaction->t_tid == target) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 /*
Andrea Gelminibcf3d0b2010-10-16 15:19:14 +0200452 * We want a new commit: OK, mark the request and wakeup the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 * commit thread. We do _not_ do the commit ourselves.
454 */
455
456 journal->j_commit_request = target;
457 jbd_debug(1, "JBD: requesting commit %d/%d\n",
458 journal->j_commit_request,
459 journal->j_commit_sequence);
460 wake_up(&journal->j_wait_commit);
461 return 1;
Ted Ts'od9b01932011-04-30 13:17:11 -0400462 } else if (!tid_geq(journal->j_commit_request, target))
463 /* This should never happen, but if it does, preserve
464 the evidence before kjournald goes into a loop and
465 increments j_commit_sequence beyond all recognition. */
466 WARN_ONCE(1, "jbd: bad log_start_commit: %u %u %u %u\n",
467 journal->j_commit_request, journal->j_commit_sequence,
468 target, journal->j_running_transaction ?
469 journal->j_running_transaction->t_tid : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 return 0;
471}
472
473int log_start_commit(journal_t *journal, tid_t tid)
474{
475 int ret;
476
477 spin_lock(&journal->j_state_lock);
478 ret = __log_start_commit(journal, tid);
479 spin_unlock(&journal->j_state_lock);
480 return ret;
481}
482
483/*
484 * Force and wait upon a commit if the calling process is not within
485 * transaction. This is used for forcing out undo-protected data which contains
486 * bitmaps, when the fs is running out of space.
487 *
488 * We can only force the running transaction if we don't have an active handle;
489 * otherwise, we will deadlock.
490 *
491 * Returns true if a transaction was started.
492 */
493int journal_force_commit_nested(journal_t *journal)
494{
495 transaction_t *transaction = NULL;
496 tid_t tid;
497
498 spin_lock(&journal->j_state_lock);
499 if (journal->j_running_transaction && !current->journal_info) {
500 transaction = journal->j_running_transaction;
501 __log_start_commit(journal, transaction->t_tid);
502 } else if (journal->j_committing_transaction)
503 transaction = journal->j_committing_transaction;
504
505 if (!transaction) {
506 spin_unlock(&journal->j_state_lock);
507 return 0; /* Nothing to retry */
508 }
509
510 tid = transaction->t_tid;
511 spin_unlock(&journal->j_state_lock);
512 log_wait_commit(journal, tid);
513 return 1;
514}
515
516/*
517 * Start a commit of the current running transaction (if any). Returns true
Jan Kara8fe4cd02009-02-11 13:04:25 -0800518 * if a transaction is going to be committed (or is currently already
519 * committing), and fills its tid in at *ptid
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 */
521int journal_start_commit(journal_t *journal, tid_t *ptid)
522{
523 int ret = 0;
524
525 spin_lock(&journal->j_state_lock);
526 if (journal->j_running_transaction) {
527 tid_t tid = journal->j_running_transaction->t_tid;
528
Jan Kara8fe4cd02009-02-11 13:04:25 -0800529 __log_start_commit(journal, tid);
530 /* There's a running transaction and we've just made sure
531 * it's commit has been scheduled. */
532 if (ptid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 *ptid = tid;
Jan Kara8fe4cd02009-02-11 13:04:25 -0800534 ret = 1;
535 } else if (journal->j_committing_transaction) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 /*
537 * If ext3_write_super() recently started a commit, then we
538 * have to wait for completion of that transaction
539 */
Jan Kara8fe4cd02009-02-11 13:04:25 -0800540 if (ptid)
541 *ptid = journal->j_committing_transaction->t_tid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 ret = 1;
543 }
544 spin_unlock(&journal->j_state_lock);
545 return ret;
546}
547
548/*
549 * Wait for a specified commit to complete.
550 * The caller may not hold the journal lock.
551 */
552int log_wait_commit(journal_t *journal, tid_t tid)
553{
554 int err = 0;
555
556#ifdef CONFIG_JBD_DEBUG
557 spin_lock(&journal->j_state_lock);
558 if (!tid_geq(journal->j_commit_request, tid)) {
559 printk(KERN_EMERG
560 "%s: error: j_commit_request=%d, tid=%d\n",
Harvey Harrison08fc99b2008-04-28 02:16:16 -0700561 __func__, journal->j_commit_request, tid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 }
563 spin_unlock(&journal->j_state_lock);
564#endif
565 spin_lock(&journal->j_state_lock);
Jan Kara2db938b2011-02-21 17:25:37 +0100566 if (!tid_geq(journal->j_commit_waited, tid))
567 journal->j_commit_waited = tid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 while (tid_gt(tid, journal->j_commit_sequence)) {
569 jbd_debug(1, "JBD: want %d, j_commit_sequence=%d\n",
570 tid, journal->j_commit_sequence);
571 wake_up(&journal->j_wait_commit);
572 spin_unlock(&journal->j_state_lock);
573 wait_event(journal->j_wait_done_commit,
574 !tid_gt(tid, journal->j_commit_sequence));
575 spin_lock(&journal->j_state_lock);
576 }
577 spin_unlock(&journal->j_state_lock);
578
579 if (unlikely(is_journal_aborted(journal))) {
580 printk(KERN_EMERG "journal commit I/O error\n");
581 err = -EIO;
582 }
583 return err;
584}
585
586/*
Jan Kara03f4d802010-04-15 22:16:24 +0200587 * Return 1 if a given transaction has not yet sent barrier request
588 * connected with a transaction commit. If 0 is returned, transaction
589 * may or may not have sent the barrier. Used to avoid sending barrier
590 * twice in common cases.
591 */
592int journal_trans_will_send_data_barrier(journal_t *journal, tid_t tid)
593{
594 int ret = 0;
595 transaction_t *commit_trans;
596
597 if (!(journal->j_flags & JFS_BARRIER))
598 return 0;
599 spin_lock(&journal->j_state_lock);
600 /* Transaction already committed? */
601 if (tid_geq(journal->j_commit_sequence, tid))
602 goto out;
603 /*
604 * Transaction is being committed and we already proceeded to
605 * writing commit record?
606 */
607 commit_trans = journal->j_committing_transaction;
608 if (commit_trans && commit_trans->t_tid == tid &&
609 commit_trans->t_state >= T_COMMIT_RECORD)
610 goto out;
611 ret = 1;
612out:
613 spin_unlock(&journal->j_state_lock);
614 return ret;
615}
Jan Kara52779702010-04-15 22:24:26 +0200616EXPORT_SYMBOL(journal_trans_will_send_data_barrier);
Jan Kara03f4d802010-04-15 22:16:24 +0200617
618/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 * Log buffer allocation routines:
620 */
621
Jan Kara9c28cbc2009-08-03 19:21:00 +0200622int journal_next_log_block(journal_t *journal, unsigned int *retp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623{
Jan Kara9c28cbc2009-08-03 19:21:00 +0200624 unsigned int blocknr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
626 spin_lock(&journal->j_state_lock);
627 J_ASSERT(journal->j_free > 1);
628
629 blocknr = journal->j_head;
630 journal->j_head++;
631 journal->j_free--;
632 if (journal->j_head == journal->j_last)
633 journal->j_head = journal->j_first;
634 spin_unlock(&journal->j_state_lock);
635 return journal_bmap(journal, blocknr, retp);
636}
637
638/*
639 * Conversion of logical to physical block numbers for the journal
640 *
641 * On external journals the journal blocks are identity-mapped, so
642 * this is a no-op. If needed, we can use j_blk_offset - everything is
643 * ready.
644 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200645int journal_bmap(journal_t *journal, unsigned int blocknr,
646 unsigned int *retp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647{
648 int err = 0;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200649 unsigned int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
651 if (journal->j_inode) {
652 ret = bmap(journal->j_inode, blocknr);
653 if (ret)
654 *retp = ret;
655 else {
656 char b[BDEVNAME_SIZE];
657
658 printk(KERN_ALERT "%s: journal block not found "
Jan Kara9c28cbc2009-08-03 19:21:00 +0200659 "at offset %u on %s\n",
Harvey Harrison08fc99b2008-04-28 02:16:16 -0700660 __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 blocknr,
662 bdevname(journal->j_dev, b));
663 err = -EIO;
664 __journal_abort_soft(journal, err);
665 }
666 } else {
667 *retp = blocknr; /* +journal->j_blk_offset */
668 }
669 return err;
670}
671
672/*
673 * We play buffer_head aliasing tricks to write data/metadata blocks to
674 * the journal without copying their contents, but for journal
675 * descriptor blocks we do need to generate bona fide buffers.
676 *
677 * After the caller of journal_get_descriptor_buffer() has finished modifying
678 * the buffer's contents they really should run flush_dcache_page(bh->b_page).
679 * But we don't bother doing that, so there will be coherency problems with
680 * mmaps of blockdevs which hold live JBD-controlled filesystems.
681 */
682struct journal_head *journal_get_descriptor_buffer(journal_t *journal)
683{
684 struct buffer_head *bh;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200685 unsigned int blocknr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 int err;
687
688 err = journal_next_log_block(journal, &blocknr);
689
690 if (err)
691 return NULL;
692
693 bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
Jan Karaecca9af2009-04-02 16:57:13 -0700694 if (!bh)
695 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 lock_buffer(bh);
697 memset(bh->b_data, 0, journal->j_blocksize);
698 set_buffer_uptodate(bh);
699 unlock_buffer(bh);
700 BUFFER_TRACE(bh, "return this buffer");
701 return journal_add_journal_head(bh);
702}
703
704/*
705 * Management for journal control blocks: functions to create and
706 * destroy journal_t structures, and to initialise and read existing
707 * journal blocks from disk. */
708
709/* First: create and setup a journal_t object in memory. We initialise
710 * very few fields yet: that has to wait until we have created the
711 * journal structures from from scratch, or loaded them from disk. */
712
713static journal_t * journal_init_common (void)
714{
715 journal_t *journal;
716 int err;
717
Mingming Cao8c3478a2007-10-18 23:39:20 -0700718 journal = kzalloc(sizeof(*journal), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 if (!journal)
720 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
722 init_waitqueue_head(&journal->j_wait_transaction_locked);
723 init_waitqueue_head(&journal->j_wait_logspace);
724 init_waitqueue_head(&journal->j_wait_done_commit);
725 init_waitqueue_head(&journal->j_wait_checkpoint);
726 init_waitqueue_head(&journal->j_wait_commit);
727 init_waitqueue_head(&journal->j_wait_updates);
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800728 mutex_init(&journal->j_checkpoint_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 spin_lock_init(&journal->j_revoke_lock);
730 spin_lock_init(&journal->j_list_lock);
731 spin_lock_init(&journal->j_state_lock);
732
733 journal->j_commit_interval = (HZ * JBD_DEFAULT_MAX_COMMIT_AGE);
734
735 /* The journal is marked for error until we succeed with recovery! */
736 journal->j_flags = JFS_ABORT;
737
738 /* Set up a default-sized revoke table for the new mount. */
739 err = journal_init_revoke(journal, JOURNAL_REVOKE_DEFAULT_HASH);
740 if (err) {
741 kfree(journal);
742 goto fail;
743 }
744 return journal;
745fail:
746 return NULL;
747}
748
749/* journal_init_dev and journal_init_inode:
750 *
751 * Create a journal structure assigned some fixed set of disk blocks to
752 * the journal. We don't actually touch those disk blocks yet, but we
753 * need to set up all of the mapping information to tell the journaling
754 * system where the journal blocks are.
755 *
756 */
757
758/**
Randy Dunlap0cf01f62008-03-19 17:00:44 -0700759 * journal_t * journal_init_dev() - creates and initialises a journal structure
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 * @bdev: Block device on which to create the journal
761 * @fs_dev: Device which hold journalled filesystem for this journal.
762 * @start: Block nr Start of journal.
Eric Sandeen37ed3222006-09-27 01:49:31 -0700763 * @len: Length of the journal in blocks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 * @blocksize: blocksize of journalling device
Randy Dunlap0cf01f62008-03-19 17:00:44 -0700765 *
766 * Returns: a newly created journal_t *
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700767 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 * journal_init_dev creates a journal which maps a fixed contiguous
769 * range of blocks on an arbitrary block device.
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700770 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 */
772journal_t * journal_init_dev(struct block_device *bdev,
773 struct block_device *fs_dev,
774 int start, int len, int blocksize)
775{
776 journal_t *journal = journal_init_common();
777 struct buffer_head *bh;
778 int n;
779
780 if (!journal)
781 return NULL;
782
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 /* journal descriptor can store up to n blocks -bzzz */
Zoltan Menyhartd1807792006-09-29 01:58:40 -0700784 journal->j_blocksize = blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 n = journal->j_blocksize / sizeof(journal_block_tag_t);
786 journal->j_wbufsize = n;
787 journal->j_wbuf = kmalloc(n * sizeof(struct buffer_head*), GFP_KERNEL);
788 if (!journal->j_wbuf) {
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300789 printk(KERN_ERR "%s: Can't allocate bhs for commit thread\n",
Harvey Harrison08fc99b2008-04-28 02:16:16 -0700790 __func__);
Jan Karaecca9af2009-04-02 16:57:13 -0700791 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 }
Zoltan Menyhartd1807792006-09-29 01:58:40 -0700793 journal->j_dev = bdev;
794 journal->j_fs_dev = fs_dev;
795 journal->j_blk_offset = start;
796 journal->j_maxlen = len;
797
798 bh = __getblk(journal->j_dev, start, journal->j_blocksize);
Jan Karaecca9af2009-04-02 16:57:13 -0700799 if (!bh) {
800 printk(KERN_ERR
801 "%s: Cannot get buffer for journal superblock\n",
802 __func__);
803 goto out_err;
804 }
Zoltan Menyhartd1807792006-09-29 01:58:40 -0700805 journal->j_sb_buffer = bh;
806 journal->j_superblock = (journal_superblock_t *)bh->b_data;
Jan Karaecca9af2009-04-02 16:57:13 -0700807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 return journal;
Jan Karaecca9af2009-04-02 16:57:13 -0700809out_err:
Tao Ma7b02bec2009-11-10 17:13:22 +0800810 kfree(journal->j_wbuf);
Jan Karaecca9af2009-04-02 16:57:13 -0700811 kfree(journal);
812 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813}
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700814
815/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 * journal_t * journal_init_inode () - creates a journal which maps to a inode.
817 * @inode: An inode to create the journal in
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700818 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 * journal_init_inode creates a journal which maps an on-disk inode as
820 * the journal. The inode must exist already, must support bmap() and
821 * must have all data blocks preallocated.
822 */
823journal_t * journal_init_inode (struct inode *inode)
824{
825 struct buffer_head *bh;
826 journal_t *journal = journal_init_common();
827 int err;
828 int n;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200829 unsigned int blocknr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830
831 if (!journal)
832 return NULL;
833
834 journal->j_dev = journal->j_fs_dev = inode->i_sb->s_bdev;
835 journal->j_inode = inode;
836 jbd_debug(1,
837 "journal %p: inode %s/%ld, size %Ld, bits %d, blksize %ld\n",
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700838 journal, inode->i_sb->s_id, inode->i_ino,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 (long long) inode->i_size,
840 inode->i_sb->s_blocksize_bits, inode->i_sb->s_blocksize);
841
842 journal->j_maxlen = inode->i_size >> inode->i_sb->s_blocksize_bits;
843 journal->j_blocksize = inode->i_sb->s_blocksize;
844
845 /* journal descriptor can store up to n blocks -bzzz */
846 n = journal->j_blocksize / sizeof(journal_block_tag_t);
847 journal->j_wbufsize = n;
848 journal->j_wbuf = kmalloc(n * sizeof(struct buffer_head*), GFP_KERNEL);
849 if (!journal->j_wbuf) {
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300850 printk(KERN_ERR "%s: Can't allocate bhs for commit thread\n",
Harvey Harrison08fc99b2008-04-28 02:16:16 -0700851 __func__);
Jan Karaecca9af2009-04-02 16:57:13 -0700852 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 }
854
855 err = journal_bmap(journal, 0, &blocknr);
856 /* If that failed, give up */
857 if (err) {
Justin P. Mattock3c26bdb2011-02-26 20:34:05 -0800858 printk(KERN_ERR "%s: Cannot locate journal superblock\n",
Harvey Harrison08fc99b2008-04-28 02:16:16 -0700859 __func__);
Jan Karaecca9af2009-04-02 16:57:13 -0700860 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 }
862
863 bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
Jan Karaecca9af2009-04-02 16:57:13 -0700864 if (!bh) {
865 printk(KERN_ERR
866 "%s: Cannot get buffer for journal superblock\n",
867 __func__);
868 goto out_err;
869 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 journal->j_sb_buffer = bh;
871 journal->j_superblock = (journal_superblock_t *)bh->b_data;
872
873 return journal;
Jan Karaecca9af2009-04-02 16:57:13 -0700874out_err:
Tao Ma7b02bec2009-11-10 17:13:22 +0800875 kfree(journal->j_wbuf);
Jan Karaecca9af2009-04-02 16:57:13 -0700876 kfree(journal);
877 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878}
879
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700880/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 * If the journal init or create aborts, we need to mark the journal
882 * superblock as being NULL to prevent the journal destroy from writing
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700883 * back a bogus superblock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 */
885static void journal_fail_superblock (journal_t *journal)
886{
887 struct buffer_head *bh = journal->j_sb_buffer;
888 brelse(bh);
889 journal->j_sb_buffer = NULL;
890}
891
892/*
893 * Given a journal_t structure, initialise the various fields for
894 * startup of a new journaling session. We use this both when creating
895 * a journal, and after recovering an old journal to reset it for
896 * subsequent use.
897 */
898
899static int journal_reset(journal_t *journal)
900{
901 journal_superblock_t *sb = journal->j_superblock;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200902 unsigned int first, last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903
904 first = be32_to_cpu(sb->s_first);
905 last = be32_to_cpu(sb->s_maxlen);
Jan Kara7447a662009-07-15 20:36:08 +0200906 if (first + JFS_MIN_JOURNAL_BLOCKS > last + 1) {
Jan Kara9c28cbc2009-08-03 19:21:00 +0200907 printk(KERN_ERR "JBD: Journal too short (blocks %u-%u).\n",
Jan Kara7447a662009-07-15 20:36:08 +0200908 first, last);
909 journal_fail_superblock(journal);
910 return -EINVAL;
911 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
913 journal->j_first = first;
914 journal->j_last = last;
915
916 journal->j_head = first;
917 journal->j_tail = first;
918 journal->j_free = last - first;
919
920 journal->j_tail_sequence = journal->j_transaction_sequence;
921 journal->j_commit_sequence = journal->j_transaction_sequence - 1;
922 journal->j_commit_request = journal->j_commit_sequence;
923
924 journal->j_max_transaction_buffers = journal->j_maxlen / 4;
925
Jan Kara9754e392012-04-07 12:33:03 +0200926 /*
927 * As a special case, if the on-disk copy is already marked as needing
928 * no recovery (s_start == 0), then we can safely defer the superblock
929 * update until the next commit by setting JFS_FLUSHED. This avoids
930 * attempting a write to a potential-readonly device.
931 */
932 if (sb->s_start == 0) {
933 jbd_debug(1,"JBD: Skipping superblock update on recovered sb "
934 "(start %u, seq %d, errno %d)\n",
935 journal->j_tail, journal->j_tail_sequence,
936 journal->j_errno);
937 journal->j_flags |= JFS_FLUSHED;
938 } else {
Jan Kara1ce84862012-04-07 12:50:13 +0200939 /* Lock here to make assertions happy... */
940 mutex_lock(&journal->j_checkpoint_mutex);
Jan Karafd2cbd42012-04-07 11:05:19 +0200941 /*
942 * Update log tail information. We use WRITE_FUA since new
943 * transaction will start reusing journal space and so we
944 * must make sure information about current log tail is on
945 * disk before that.
946 */
947 journal_update_sb_log_tail(journal,
948 journal->j_tail_sequence,
949 journal->j_tail,
950 WRITE_FUA);
Jan Kara1ce84862012-04-07 12:50:13 +0200951 mutex_unlock(&journal->j_checkpoint_mutex);
Jan Kara9754e392012-04-07 12:33:03 +0200952 }
Pavel Emelianov97f06782007-05-08 00:30:42 -0700953 return journal_start_thread(journal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954}
955
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700956/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 * int journal_create() - Initialise the new journal file
958 * @journal: Journal to create. This structure must have been initialised
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700959 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 * Given a journal_t structure which tells us which disk blocks we can
961 * use, create a new journal superblock and initialise all of the
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700962 * journal fields from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 **/
964int journal_create(journal_t *journal)
965{
Jan Kara9c28cbc2009-08-03 19:21:00 +0200966 unsigned int blocknr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 struct buffer_head *bh;
968 journal_superblock_t *sb;
969 int i, err;
970
971 if (journal->j_maxlen < JFS_MIN_JOURNAL_BLOCKS) {
972 printk (KERN_ERR "Journal length (%d blocks) too short.\n",
973 journal->j_maxlen);
974 journal_fail_superblock(journal);
975 return -EINVAL;
976 }
977
978 if (journal->j_inode == NULL) {
979 /*
980 * We don't know what block to start at!
981 */
982 printk(KERN_EMERG
983 "%s: creation of journal on external device!\n",
Harvey Harrison08fc99b2008-04-28 02:16:16 -0700984 __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 BUG();
986 }
987
988 /* Zero out the entire journal on disk. We cannot afford to
989 have any blocks on disk beginning with JFS_MAGIC_NUMBER. */
990 jbd_debug(1, "JBD: Zeroing out journal blocks...\n");
991 for (i = 0; i < journal->j_maxlen; i++) {
992 err = journal_bmap(journal, i, &blocknr);
993 if (err)
994 return err;
995 bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
Namhyung Kim2a0e3382010-10-13 17:17:18 +0900996 if (unlikely(!bh))
997 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 lock_buffer(bh);
999 memset (bh->b_data, 0, journal->j_blocksize);
1000 BUFFER_TRACE(bh, "marking dirty");
1001 mark_buffer_dirty(bh);
1002 BUFFER_TRACE(bh, "marking uptodate");
1003 set_buffer_uptodate(bh);
1004 unlock_buffer(bh);
1005 __brelse(bh);
1006 }
1007
1008 sync_blockdev(journal->j_dev);
1009 jbd_debug(1, "JBD: journal cleared.\n");
1010
1011 /* OK, fill in the initial static fields in the new superblock */
1012 sb = journal->j_superblock;
1013
1014 sb->s_header.h_magic = cpu_to_be32(JFS_MAGIC_NUMBER);
1015 sb->s_header.h_blocktype = cpu_to_be32(JFS_SUPERBLOCK_V2);
1016
1017 sb->s_blocksize = cpu_to_be32(journal->j_blocksize);
1018 sb->s_maxlen = cpu_to_be32(journal->j_maxlen);
1019 sb->s_first = cpu_to_be32(1);
1020
1021 journal->j_transaction_sequence = 1;
1022
1023 journal->j_flags &= ~JFS_ABORT;
1024 journal->j_format_version = 2;
1025
1026 return journal_reset(journal);
1027}
1028
Jan Karafd2cbd42012-04-07 11:05:19 +02001029static void journal_write_superblock(journal_t *journal, int write_op)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 struct buffer_head *bh = journal->j_sb_buffer;
Jan Karafd2cbd42012-04-07 11:05:19 +02001032 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
Jan Karafd2cbd42012-04-07 11:05:19 +02001034 trace_journal_write_superblock(journal, write_op);
1035 if (!(journal->j_flags & JFS_BARRIER))
1036 write_op &= ~(REQ_FUA | REQ_FLUSH);
1037 lock_buffer(bh);
Darrick J. Wongdff68252010-10-04 12:35:05 -07001038 if (buffer_write_io_error(bh)) {
1039 char b[BDEVNAME_SIZE];
1040 /*
1041 * Oh, dear. A previous attempt to write the journal
1042 * superblock failed. This could happen because the
1043 * USB device was yanked out. Or it could happen to
1044 * be a transient write error and maybe the block will
1045 * be remapped. Nothing we can do but to retry the
1046 * write and hope for the best.
1047 */
1048 printk(KERN_ERR "JBD: previous I/O error detected "
1049 "for journal superblock update for %s.\n",
1050 journal_dev_name(journal, b));
1051 clear_buffer_write_io_error(bh);
1052 set_buffer_uptodate(bh);
1053 }
1054
Jan Karafd2cbd42012-04-07 11:05:19 +02001055 get_bh(bh);
1056 bh->b_end_io = end_buffer_write_sync;
1057 ret = submit_bh(write_op, bh);
1058 wait_on_buffer(bh);
Jan Kara9754e392012-04-07 12:33:03 +02001059 if (buffer_write_io_error(bh)) {
Jan Kara9754e392012-04-07 12:33:03 +02001060 clear_buffer_write_io_error(bh);
1061 set_buffer_uptodate(bh);
Jan Karafd2cbd42012-04-07 11:05:19 +02001062 ret = -EIO;
1063 }
1064 if (ret) {
1065 char b[BDEVNAME_SIZE];
1066 printk(KERN_ERR "JBD: Error %d detected "
1067 "when updating journal superblock for %s.\n",
1068 ret, journal_dev_name(journal, b));
Jan Kara9754e392012-04-07 12:33:03 +02001069 }
1070}
1071
1072/**
1073 * journal_update_sb_log_tail() - Update log tail in journal sb on disk.
1074 * @journal: The journal to update.
Jan Karafd2cbd42012-04-07 11:05:19 +02001075 * @tail_tid: TID of the new transaction at the tail of the log
1076 * @tail_block: The first block of the transaction at the tail of the log
1077 * @write_op: With which operation should we write the journal sb
Jan Kara9754e392012-04-07 12:33:03 +02001078 *
1079 * Update a journal's superblock information about log tail and write it to
1080 * disk, waiting for the IO to complete.
1081 */
Jan Karafd2cbd42012-04-07 11:05:19 +02001082void journal_update_sb_log_tail(journal_t *journal, tid_t tail_tid,
1083 unsigned int tail_block, int write_op)
Jan Kara9754e392012-04-07 12:33:03 +02001084{
1085 journal_superblock_t *sb = journal->j_superblock;
1086
Jan Kara1ce84862012-04-07 12:50:13 +02001087 BUG_ON(!mutex_is_locked(&journal->j_checkpoint_mutex));
Jan Karafd2cbd42012-04-07 11:05:19 +02001088 jbd_debug(1,"JBD: updating superblock (start %u, seq %u)\n",
1089 tail_block, tail_tid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090
Jan Karafd2cbd42012-04-07 11:05:19 +02001091 sb->s_sequence = cpu_to_be32(tail_tid);
1092 sb->s_start = cpu_to_be32(tail_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093
Jan Karafd2cbd42012-04-07 11:05:19 +02001094 journal_write_superblock(journal, write_op);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095
Jan Kara9754e392012-04-07 12:33:03 +02001096 /* Log is no longer empty */
1097 spin_lock(&journal->j_state_lock);
1098 WARN_ON(!sb->s_sequence);
1099 journal->j_flags &= ~JFS_FLUSHED;
1100 spin_unlock(&journal->j_state_lock);
1101}
1102
1103/**
1104 * mark_journal_empty() - Mark on disk journal as empty.
1105 * @journal: The journal to update.
1106 *
1107 * Update a journal's dynamic superblock fields to show that journal is empty.
1108 * Write updated superblock to disk waiting for IO to complete.
1109 */
1110static void mark_journal_empty(journal_t *journal)
1111{
1112 journal_superblock_t *sb = journal->j_superblock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113
Jan Kara1ce84862012-04-07 12:50:13 +02001114 BUG_ON(!mutex_is_locked(&journal->j_checkpoint_mutex));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 spin_lock(&journal->j_state_lock);
Jan Kara9754e392012-04-07 12:33:03 +02001116 jbd_debug(1, "JBD: Marking journal as empty (seq %d)\n",
1117 journal->j_tail_sequence);
1118
1119 sb->s_sequence = cpu_to_be32(journal->j_tail_sequence);
1120 sb->s_start = cpu_to_be32(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 spin_unlock(&journal->j_state_lock);
Jan Kara9754e392012-04-07 12:33:03 +02001122
Jan Karafd2cbd42012-04-07 11:05:19 +02001123 journal_write_superblock(journal, WRITE_FUA);
Jan Kara9754e392012-04-07 12:33:03 +02001124
1125 spin_lock(&journal->j_state_lock);
1126 /* Log is empty */
1127 journal->j_flags |= JFS_FLUSHED;
1128 spin_unlock(&journal->j_state_lock);
1129}
1130
1131/**
1132 * journal_update_sb_errno() - Update error in the journal.
1133 * @journal: The journal to update.
1134 *
1135 * Update a journal's errno. Write updated superblock to disk waiting for IO
1136 * to complete.
1137 */
1138static void journal_update_sb_errno(journal_t *journal)
1139{
1140 journal_superblock_t *sb = journal->j_superblock;
1141
1142 spin_lock(&journal->j_state_lock);
1143 jbd_debug(1, "JBD: updating superblock error (errno %d)\n",
1144 journal->j_errno);
1145 sb->s_errno = cpu_to_be32(journal->j_errno);
1146 spin_unlock(&journal->j_state_lock);
1147
Jan Karafd2cbd42012-04-07 11:05:19 +02001148 journal_write_superblock(journal, WRITE_SYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149}
1150
1151/*
1152 * Read the superblock for a given journal, performing initial
1153 * validation of the format.
1154 */
1155
1156static int journal_get_superblock(journal_t *journal)
1157{
1158 struct buffer_head *bh;
1159 journal_superblock_t *sb;
1160 int err = -EIO;
1161
1162 bh = journal->j_sb_buffer;
1163
1164 J_ASSERT(bh != NULL);
1165 if (!buffer_uptodate(bh)) {
1166 ll_rw_block(READ, 1, &bh);
1167 wait_on_buffer(bh);
1168 if (!buffer_uptodate(bh)) {
1169 printk (KERN_ERR
1170 "JBD: IO error reading journal superblock\n");
1171 goto out;
1172 }
1173 }
1174
1175 sb = journal->j_superblock;
1176
1177 err = -EINVAL;
1178
1179 if (sb->s_header.h_magic != cpu_to_be32(JFS_MAGIC_NUMBER) ||
1180 sb->s_blocksize != cpu_to_be32(journal->j_blocksize)) {
1181 printk(KERN_WARNING "JBD: no valid journal superblock found\n");
1182 goto out;
1183 }
1184
1185 switch(be32_to_cpu(sb->s_header.h_blocktype)) {
1186 case JFS_SUPERBLOCK_V1:
1187 journal->j_format_version = 1;
1188 break;
1189 case JFS_SUPERBLOCK_V2:
1190 journal->j_format_version = 2;
1191 break;
1192 default:
1193 printk(KERN_WARNING "JBD: unrecognised superblock format ID\n");
1194 goto out;
1195 }
1196
1197 if (be32_to_cpu(sb->s_maxlen) < journal->j_maxlen)
1198 journal->j_maxlen = be32_to_cpu(sb->s_maxlen);
1199 else if (be32_to_cpu(sb->s_maxlen) > journal->j_maxlen) {
1200 printk (KERN_WARNING "JBD: journal file too short\n");
1201 goto out;
1202 }
1203
Eryu Guan87622022011-11-01 19:04:59 -04001204 if (be32_to_cpu(sb->s_first) == 0 ||
1205 be32_to_cpu(sb->s_first) >= journal->j_maxlen) {
1206 printk(KERN_WARNING
1207 "JBD: Invalid start block of journal: %u\n",
1208 be32_to_cpu(sb->s_first));
1209 goto out;
1210 }
1211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 return 0;
1213
1214out:
1215 journal_fail_superblock(journal);
1216 return err;
1217}
1218
1219/*
1220 * Load the on-disk journal superblock and read the key fields into the
1221 * journal_t.
1222 */
1223
1224static int load_superblock(journal_t *journal)
1225{
1226 int err;
1227 journal_superblock_t *sb;
1228
1229 err = journal_get_superblock(journal);
1230 if (err)
1231 return err;
1232
1233 sb = journal->j_superblock;
1234
1235 journal->j_tail_sequence = be32_to_cpu(sb->s_sequence);
1236 journal->j_tail = be32_to_cpu(sb->s_start);
1237 journal->j_first = be32_to_cpu(sb->s_first);
1238 journal->j_last = be32_to_cpu(sb->s_maxlen);
1239 journal->j_errno = be32_to_cpu(sb->s_errno);
1240
1241 return 0;
1242}
1243
1244
1245/**
1246 * int journal_load() - Read journal from disk.
1247 * @journal: Journal to act on.
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001248 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 * Given a journal_t structure which tells us which disk blocks contain
1250 * a journal, read the journal from disk to initialise the in-memory
1251 * structures.
1252 */
1253int journal_load(journal_t *journal)
1254{
1255 int err;
Badari Pulavartyea817392006-08-27 01:23:52 -07001256 journal_superblock_t *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257
1258 err = load_superblock(journal);
1259 if (err)
1260 return err;
1261
Badari Pulavartyea817392006-08-27 01:23:52 -07001262 sb = journal->j_superblock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 /* If this is a V2 superblock, then we have to check the
1264 * features flags on it. */
1265
1266 if (journal->j_format_version >= 2) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 if ((sb->s_feature_ro_compat &
1268 ~cpu_to_be32(JFS_KNOWN_ROCOMPAT_FEATURES)) ||
1269 (sb->s_feature_incompat &
1270 ~cpu_to_be32(JFS_KNOWN_INCOMPAT_FEATURES))) {
1271 printk (KERN_WARNING
1272 "JBD: Unrecognised features on journal\n");
1273 return -EINVAL;
1274 }
1275 }
1276
1277 /* Let the recovery code check whether it needs to recover any
1278 * data from the journal. */
1279 if (journal_recover(journal))
1280 goto recovery_error;
1281
1282 /* OK, we've finished with the dynamic journal bits:
1283 * reinitialise the dynamic contents of the superblock in memory
1284 * and reset them on disk. */
1285 if (journal_reset(journal))
1286 goto recovery_error;
1287
1288 journal->j_flags &= ~JFS_ABORT;
1289 journal->j_flags |= JFS_LOADED;
1290 return 0;
1291
1292recovery_error:
1293 printk (KERN_WARNING "JBD: recovery failed\n");
1294 return -EIO;
1295}
1296
1297/**
1298 * void journal_destroy() - Release a journal_t structure.
1299 * @journal: Journal to act on.
1300 *
1301 * Release a journal_t structure once it is no longer in use by the
1302 * journaled object.
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001303 * Return <0 if we couldn't clean up the journal.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 */
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001305int journal_destroy(journal_t *journal)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306{
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001307 int err = 0;
1308
Jan Kara03f4d802010-04-15 22:16:24 +02001309
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310 /* Wait for the commit thread to wake up and die. */
1311 journal_kill_thread(journal);
1312
1313 /* Force a final log commit */
1314 if (journal->j_running_transaction)
1315 journal_commit_transaction(journal);
1316
1317 /* Force any old transactions to disk */
1318
Jan Kara1ce84862012-04-07 12:50:13 +02001319 /* We cannot race with anybody but must keep assertions happy */
1320 mutex_lock(&journal->j_checkpoint_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 /* Totally anal locking here... */
1322 spin_lock(&journal->j_list_lock);
1323 while (journal->j_checkpoint_transactions != NULL) {
1324 spin_unlock(&journal->j_list_lock);
1325 log_do_checkpoint(journal);
1326 spin_lock(&journal->j_list_lock);
1327 }
1328
1329 J_ASSERT(journal->j_running_transaction == NULL);
1330 J_ASSERT(journal->j_committing_transaction == NULL);
1331 J_ASSERT(journal->j_checkpoint_transactions == NULL);
1332 spin_unlock(&journal->j_list_lock);
1333
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 if (journal->j_sb_buffer) {
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001335 if (!is_journal_aborted(journal)) {
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001336 journal->j_tail_sequence =
1337 ++journal->j_transaction_sequence;
Jan Kara9754e392012-04-07 12:33:03 +02001338 mark_journal_empty(journal);
1339 } else
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001340 err = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 brelse(journal->j_sb_buffer);
1342 }
Jan Kara1ce84862012-04-07 12:50:13 +02001343 mutex_unlock(&journal->j_checkpoint_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344
1345 if (journal->j_inode)
1346 iput(journal->j_inode);
1347 if (journal->j_revoke)
1348 journal_destroy_revoke(journal);
1349 kfree(journal->j_wbuf);
1350 kfree(journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001351
1352 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353}
1354
1355
1356/**
1357 *int journal_check_used_features () - Check if features specified are used.
1358 * @journal: Journal to check.
1359 * @compat: bitmask of compatible features
1360 * @ro: bitmask of features that force read-only mount
1361 * @incompat: bitmask of incompatible features
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001362 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 * Check whether the journal uses all of a given set of
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001364 * features. Return true (non-zero) if it does.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 **/
1366
1367int journal_check_used_features (journal_t *journal, unsigned long compat,
1368 unsigned long ro, unsigned long incompat)
1369{
1370 journal_superblock_t *sb;
1371
1372 if (!compat && !ro && !incompat)
1373 return 1;
1374 if (journal->j_format_version == 1)
1375 return 0;
1376
1377 sb = journal->j_superblock;
1378
1379 if (((be32_to_cpu(sb->s_feature_compat) & compat) == compat) &&
1380 ((be32_to_cpu(sb->s_feature_ro_compat) & ro) == ro) &&
1381 ((be32_to_cpu(sb->s_feature_incompat) & incompat) == incompat))
1382 return 1;
1383
1384 return 0;
1385}
1386
1387/**
1388 * int journal_check_available_features() - Check feature set in journalling layer
1389 * @journal: Journal to check.
1390 * @compat: bitmask of compatible features
1391 * @ro: bitmask of features that force read-only mount
1392 * @incompat: bitmask of incompatible features
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001393 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 * Check whether the journaling code supports the use of
1395 * all of a given set of features on this journal. Return true
1396 * (non-zero) if it can. */
1397
1398int journal_check_available_features (journal_t *journal, unsigned long compat,
1399 unsigned long ro, unsigned long incompat)
1400{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 if (!compat && !ro && !incompat)
1402 return 1;
1403
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 /* We can support any known requested features iff the
1405 * superblock is in version 2. Otherwise we fail to support any
1406 * extended sb features. */
1407
1408 if (journal->j_format_version != 2)
1409 return 0;
1410
1411 if ((compat & JFS_KNOWN_COMPAT_FEATURES) == compat &&
1412 (ro & JFS_KNOWN_ROCOMPAT_FEATURES) == ro &&
1413 (incompat & JFS_KNOWN_INCOMPAT_FEATURES) == incompat)
1414 return 1;
1415
1416 return 0;
1417}
1418
1419/**
1420 * int journal_set_features () - Mark a given journal feature in the superblock
1421 * @journal: Journal to act on.
1422 * @compat: bitmask of compatible features
1423 * @ro: bitmask of features that force read-only mount
1424 * @incompat: bitmask of incompatible features
1425 *
1426 * Mark a given journal feature as present on the
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001427 * superblock. Returns true if the requested features could be set.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 *
1429 */
1430
1431int journal_set_features (journal_t *journal, unsigned long compat,
1432 unsigned long ro, unsigned long incompat)
1433{
1434 journal_superblock_t *sb;
1435
1436 if (journal_check_used_features(journal, compat, ro, incompat))
1437 return 1;
1438
1439 if (!journal_check_available_features(journal, compat, ro, incompat))
1440 return 0;
1441
1442 jbd_debug(1, "Setting new features 0x%lx/0x%lx/0x%lx\n",
1443 compat, ro, incompat);
1444
1445 sb = journal->j_superblock;
1446
1447 sb->s_feature_compat |= cpu_to_be32(compat);
1448 sb->s_feature_ro_compat |= cpu_to_be32(ro);
1449 sb->s_feature_incompat |= cpu_to_be32(incompat);
1450
1451 return 1;
1452}
1453
1454
1455/**
1456 * int journal_update_format () - Update on-disk journal structure.
1457 * @journal: Journal to act on.
1458 *
1459 * Given an initialised but unloaded journal struct, poke about in the
1460 * on-disk structure to update it to the most recent supported version.
1461 */
1462int journal_update_format (journal_t *journal)
1463{
1464 journal_superblock_t *sb;
1465 int err;
1466
1467 err = journal_get_superblock(journal);
1468 if (err)
1469 return err;
1470
1471 sb = journal->j_superblock;
1472
1473 switch (be32_to_cpu(sb->s_header.h_blocktype)) {
1474 case JFS_SUPERBLOCK_V2:
1475 return 0;
1476 case JFS_SUPERBLOCK_V1:
1477 return journal_convert_superblock_v1(journal, sb);
1478 default:
1479 break;
1480 }
1481 return -EINVAL;
1482}
1483
1484static int journal_convert_superblock_v1(journal_t *journal,
1485 journal_superblock_t *sb)
1486{
1487 int offset, blocksize;
1488 struct buffer_head *bh;
1489
1490 printk(KERN_WARNING
1491 "JBD: Converting superblock from version 1 to 2.\n");
1492
1493 /* Pre-initialise new fields to zero */
1494 offset = ((char *) &(sb->s_feature_compat)) - ((char *) sb);
1495 blocksize = be32_to_cpu(sb->s_blocksize);
1496 memset(&sb->s_feature_compat, 0, blocksize-offset);
1497
1498 sb->s_nr_users = cpu_to_be32(1);
1499 sb->s_header.h_blocktype = cpu_to_be32(JFS_SUPERBLOCK_V2);
1500 journal->j_format_version = 2;
1501
1502 bh = journal->j_sb_buffer;
1503 BUFFER_TRACE(bh, "marking dirty");
1504 mark_buffer_dirty(bh);
1505 sync_dirty_buffer(bh);
1506 return 0;
1507}
1508
1509
1510/**
1511 * int journal_flush () - Flush journal
1512 * @journal: Journal to act on.
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001513 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 * Flush all data for a given journal to disk and empty the journal.
1515 * Filesystems can use this when remounting readonly to ensure that
1516 * recovery does not need to happen on remount.
1517 */
1518
1519int journal_flush(journal_t *journal)
1520{
1521 int err = 0;
1522 transaction_t *transaction = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523
1524 spin_lock(&journal->j_state_lock);
1525
1526 /* Force everything buffered to the log... */
1527 if (journal->j_running_transaction) {
1528 transaction = journal->j_running_transaction;
1529 __log_start_commit(journal, transaction->t_tid);
1530 } else if (journal->j_committing_transaction)
1531 transaction = journal->j_committing_transaction;
1532
1533 /* Wait for the log commit to complete... */
1534 if (transaction) {
1535 tid_t tid = transaction->t_tid;
1536
1537 spin_unlock(&journal->j_state_lock);
1538 log_wait_commit(journal, tid);
1539 } else {
1540 spin_unlock(&journal->j_state_lock);
1541 }
1542
1543 /* ...and flush everything in the log out to disk. */
1544 spin_lock(&journal->j_list_lock);
1545 while (!err && journal->j_checkpoint_transactions != NULL) {
1546 spin_unlock(&journal->j_list_lock);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001547 mutex_lock(&journal->j_checkpoint_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 err = log_do_checkpoint(journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001549 mutex_unlock(&journal->j_checkpoint_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 spin_lock(&journal->j_list_lock);
1551 }
1552 spin_unlock(&journal->j_list_lock);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001553
1554 if (is_journal_aborted(journal))
1555 return -EIO;
1556
Jan Kara1ce84862012-04-07 12:50:13 +02001557 mutex_lock(&journal->j_checkpoint_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 cleanup_journal_tail(journal);
1559
1560 /* Finally, mark the journal as really needing no recovery.
1561 * This sets s_start==0 in the underlying superblock, which is
1562 * the magic code for a fully-recovered superblock. Any future
1563 * commits of data to the journal will restore the current
1564 * s_start value. */
Jan Kara9754e392012-04-07 12:33:03 +02001565 mark_journal_empty(journal);
Jan Kara1ce84862012-04-07 12:50:13 +02001566 mutex_unlock(&journal->j_checkpoint_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 spin_lock(&journal->j_state_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 J_ASSERT(!journal->j_running_transaction);
1569 J_ASSERT(!journal->j_committing_transaction);
1570 J_ASSERT(!journal->j_checkpoint_transactions);
1571 J_ASSERT(journal->j_head == journal->j_tail);
1572 J_ASSERT(journal->j_tail_sequence == journal->j_transaction_sequence);
1573 spin_unlock(&journal->j_state_lock);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -07001574 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575}
1576
1577/**
1578 * int journal_wipe() - Wipe journal contents
1579 * @journal: Journal to act on.
1580 * @write: flag (see below)
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001581 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 * Wipe out all of the contents of a journal, safely. This will produce
1583 * a warning if the journal contains any valid recovery information.
1584 * Must be called between journal_init_*() and journal_load().
1585 *
1586 * If 'write' is non-zero, then we wipe out the journal on disk; otherwise
1587 * we merely suppress recovery.
1588 */
1589
1590int journal_wipe(journal_t *journal, int write)
1591{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 int err = 0;
1593
1594 J_ASSERT (!(journal->j_flags & JFS_LOADED));
1595
1596 err = load_superblock(journal);
1597 if (err)
1598 return err;
1599
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 if (!journal->j_tail)
1601 goto no_recovery;
1602
1603 printk (KERN_WARNING "JBD: %s recovery information on journal\n",
1604 write ? "Clearing" : "Ignoring");
1605
1606 err = journal_skip_recovery(journal);
Jan Kara1ce84862012-04-07 12:50:13 +02001607 if (write) {
1608 /* Lock to make assertions happy... */
1609 mutex_lock(&journal->j_checkpoint_mutex);
Jan Kara9754e392012-04-07 12:33:03 +02001610 mark_journal_empty(journal);
Jan Kara1ce84862012-04-07 12:50:13 +02001611 mutex_unlock(&journal->j_checkpoint_mutex);
1612 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613
1614 no_recovery:
1615 return err;
1616}
1617
1618/*
1619 * journal_dev_name: format a character string to describe on what
1620 * device this journal is present.
1621 */
1622
Adrian Bunk022a4a72005-09-06 15:16:41 -07001623static const char *journal_dev_name(journal_t *journal, char *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624{
1625 struct block_device *bdev;
1626
1627 if (journal->j_inode)
1628 bdev = journal->j_inode->i_sb->s_bdev;
1629 else
1630 bdev = journal->j_dev;
1631
1632 return bdevname(bdev, buffer);
1633}
1634
1635/*
1636 * Journal abort has very specific semantics, which we describe
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001637 * for journal abort.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 *
1639 * Two internal function, which provide abort to te jbd layer
1640 * itself are here.
1641 */
1642
1643/*
1644 * Quick version for internal journal use (doesn't lock the journal).
1645 * Aborts hard --- we mark the abort as occurred, but do _nothing_ else,
1646 * and don't attempt to make any other journal updates.
1647 */
Adrian Bunk53308382008-02-06 01:40:12 -08001648static void __journal_abort_hard(journal_t *journal)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649{
1650 transaction_t *transaction;
1651 char b[BDEVNAME_SIZE];
1652
1653 if (journal->j_flags & JFS_ABORT)
1654 return;
1655
1656 printk(KERN_ERR "Aborting journal on device %s.\n",
1657 journal_dev_name(journal, b));
1658
1659 spin_lock(&journal->j_state_lock);
1660 journal->j_flags |= JFS_ABORT;
1661 transaction = journal->j_running_transaction;
1662 if (transaction)
1663 __log_start_commit(journal, transaction->t_tid);
1664 spin_unlock(&journal->j_state_lock);
1665}
1666
1667/* Soft abort: record the abort error status in the journal superblock,
1668 * but don't do any other IO. */
Adrian Bunk022a4a72005-09-06 15:16:41 -07001669static void __journal_abort_soft (journal_t *journal, int errno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670{
1671 if (journal->j_flags & JFS_ABORT)
1672 return;
1673
1674 if (!journal->j_errno)
1675 journal->j_errno = errno;
1676
1677 __journal_abort_hard(journal);
1678
1679 if (errno)
Jan Kara9754e392012-04-07 12:33:03 +02001680 journal_update_sb_errno(journal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681}
1682
1683/**
1684 * void journal_abort () - Shutdown the journal immediately.
1685 * @journal: the journal to shutdown.
1686 * @errno: an error number to record in the journal indicating
1687 * the reason for the shutdown.
1688 *
1689 * Perform a complete, immediate shutdown of the ENTIRE
1690 * journal (not of a single transaction). This operation cannot be
1691 * undone without closing and reopening the journal.
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001692 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 * The journal_abort function is intended to support higher level error
1694 * recovery mechanisms such as the ext2/ext3 remount-readonly error
1695 * mode.
1696 *
1697 * Journal abort has very specific semantics. Any existing dirty,
1698 * unjournaled buffers in the main filesystem will still be written to
1699 * disk by bdflush, but the journaling mechanism will be suspended
1700 * immediately and no further transaction commits will be honoured.
1701 *
1702 * Any dirty, journaled buffers will be written back to disk without
1703 * hitting the journal. Atomicity cannot be guaranteed on an aborted
1704 * filesystem, but we _do_ attempt to leave as much data as possible
1705 * behind for fsck to use for cleanup.
1706 *
1707 * Any attempt to get a new transaction handle on a journal which is in
1708 * ABORT state will just result in an -EROFS error return. A
1709 * journal_stop on an existing handle will return -EIO if we have
1710 * entered abort state during the update.
1711 *
1712 * Recursive transactions are not disturbed by journal abort until the
1713 * final journal_stop, which will receive the -EIO error.
1714 *
1715 * Finally, the journal_abort call allows the caller to supply an errno
1716 * which will be recorded (if possible) in the journal superblock. This
1717 * allows a client to record failure conditions in the middle of a
1718 * transaction without having to complete the transaction to record the
1719 * failure to disk. ext3_error, for example, now uses this
1720 * functionality.
1721 *
1722 * Errors which originate from within the journaling layer will NOT
1723 * supply an errno; a null errno implies that absolutely no further
1724 * writes are done to the journal (unless there are any already in
1725 * progress).
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001726 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 */
1728
1729void journal_abort(journal_t *journal, int errno)
1730{
1731 __journal_abort_soft(journal, errno);
1732}
1733
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001734/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 * int journal_errno () - returns the journal's error state.
1736 * @journal: journal to examine.
1737 *
1738 * This is the errno numbet set with journal_abort(), the last
1739 * time the journal was mounted - if the journal was stopped
1740 * without calling abort this will be 0.
1741 *
1742 * If the journal has been aborted on this mount time -EROFS will
1743 * be returned.
1744 */
1745int journal_errno(journal_t *journal)
1746{
1747 int err;
1748
1749 spin_lock(&journal->j_state_lock);
1750 if (journal->j_flags & JFS_ABORT)
1751 err = -EROFS;
1752 else
1753 err = journal->j_errno;
1754 spin_unlock(&journal->j_state_lock);
1755 return err;
1756}
1757
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001758/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 * int journal_clear_err () - clears the journal's error state
1760 * @journal: journal to act on.
1761 *
1762 * An error must be cleared or Acked to take a FS out of readonly
1763 * mode.
1764 */
1765int journal_clear_err(journal_t *journal)
1766{
1767 int err = 0;
1768
1769 spin_lock(&journal->j_state_lock);
1770 if (journal->j_flags & JFS_ABORT)
1771 err = -EROFS;
1772 else
1773 journal->j_errno = 0;
1774 spin_unlock(&journal->j_state_lock);
1775 return err;
1776}
1777
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001778/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 * void journal_ack_err() - Ack journal err.
1780 * @journal: journal to act on.
1781 *
1782 * An error must be cleared or Acked to take a FS out of readonly
1783 * mode.
1784 */
1785void journal_ack_err(journal_t *journal)
1786{
1787 spin_lock(&journal->j_state_lock);
1788 if (journal->j_errno)
1789 journal->j_flags |= JFS_ACK_ERR;
1790 spin_unlock(&journal->j_state_lock);
1791}
1792
1793int journal_blocks_per_page(struct inode *inode)
1794{
1795 return 1 << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);
1796}
1797
1798/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 * Journal_head storage management
1800 */
Christoph Lametere18b8902006-12-06 20:33:20 -08001801static struct kmem_cache *journal_head_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802#ifdef CONFIG_JBD_DEBUG
1803static atomic_t nr_journal_heads = ATOMIC_INIT(0);
1804#endif
1805
1806static int journal_init_journal_head_cache(void)
1807{
1808 int retval;
1809
Al Viro1076d172008-03-29 03:07:18 +00001810 J_ASSERT(journal_head_cache == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 journal_head_cache = kmem_cache_create("journal_head",
1812 sizeof(struct journal_head),
1813 0, /* offset */
Mel Gormane12ba742007-10-16 01:25:52 -07001814 SLAB_TEMPORARY, /* flags */
Paul Mundt20c2df82007-07-20 10:11:58 +09001815 NULL); /* ctor */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 retval = 0;
Al Viro1076d172008-03-29 03:07:18 +00001817 if (!journal_head_cache) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 retval = -ENOMEM;
1819 printk(KERN_EMERG "JBD: no memory for journal_head cache\n");
1820 }
1821 return retval;
1822}
1823
1824static void journal_destroy_journal_head_cache(void)
1825{
Duane Griffin3850f7a2008-07-25 01:46:19 -07001826 if (journal_head_cache) {
1827 kmem_cache_destroy(journal_head_cache);
1828 journal_head_cache = NULL;
1829 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830}
1831
1832/*
1833 * journal_head splicing and dicing
1834 */
1835static struct journal_head *journal_alloc_journal_head(void)
1836{
1837 struct journal_head *ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838
1839#ifdef CONFIG_JBD_DEBUG
1840 atomic_inc(&nr_journal_heads);
1841#endif
1842 ret = kmem_cache_alloc(journal_head_cache, GFP_NOFS);
Stephen Hemmingerc80544d2007-10-18 03:07:05 -07001843 if (ret == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 jbd_debug(1, "out of memory for journal_head\n");
Namhyung Kimf81e3d42010-10-04 19:12:13 +09001845 printk_ratelimited(KERN_NOTICE "ENOMEM in %s, retrying.\n",
1846 __func__);
1847
Stephen Hemmingerc80544d2007-10-18 03:07:05 -07001848 while (ret == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 yield();
1850 ret = kmem_cache_alloc(journal_head_cache, GFP_NOFS);
1851 }
1852 }
1853 return ret;
1854}
1855
1856static void journal_free_journal_head(struct journal_head *jh)
1857{
1858#ifdef CONFIG_JBD_DEBUG
1859 atomic_dec(&nr_journal_heads);
Randy Dunlapc9cf5522006-06-27 02:53:52 -07001860 memset(jh, JBD_POISON_FREE, sizeof(*jh));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861#endif
1862 kmem_cache_free(journal_head_cache, jh);
1863}
1864
1865/*
1866 * A journal_head is attached to a buffer_head whenever JBD has an
1867 * interest in the buffer.
1868 *
1869 * Whenever a buffer has an attached journal_head, its ->b_state:BH_JBD bit
1870 * is set. This bit is tested in core kernel code where we need to take
1871 * JBD-specific actions. Testing the zeroness of ->b_private is not reliable
1872 * there.
1873 *
1874 * When a buffer has its BH_JBD bit set, its ->b_count is elevated by one.
1875 *
1876 * When a buffer has its BH_JBD bit set it is immune from being released by
1877 * core kernel code, mainly via ->b_count.
1878 *
Jan Karabb189242011-06-24 23:11:59 +02001879 * A journal_head is detached from its buffer_head when the journal_head's
1880 * b_jcount reaches zero. Running transaction (b_transaction) and checkpoint
1881 * transaction (b_cp_transaction) hold their references to b_jcount.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 *
1883 * Various places in the kernel want to attach a journal_head to a buffer_head
1884 * _before_ attaching the journal_head to a transaction. To protect the
1885 * journal_head in this situation, journal_add_journal_head elevates the
1886 * journal_head's b_jcount refcount by one. The caller must call
1887 * journal_put_journal_head() to undo this.
1888 *
1889 * So the typical usage would be:
1890 *
1891 * (Attach a journal_head if needed. Increments b_jcount)
1892 * struct journal_head *jh = journal_add_journal_head(bh);
1893 * ...
Jan Karabb189242011-06-24 23:11:59 +02001894 * (Get another reference for transaction)
1895 * journal_grab_journal_head(bh);
1896 * jh->b_transaction = xxx;
1897 * (Put original reference)
1898 * journal_put_journal_head(jh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 */
1900
1901/*
1902 * Give a buffer_head a journal_head.
1903 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 * May sleep.
1905 */
1906struct journal_head *journal_add_journal_head(struct buffer_head *bh)
1907{
1908 struct journal_head *jh;
1909 struct journal_head *new_jh = NULL;
1910
1911repeat:
1912 if (!buffer_jbd(bh)) {
1913 new_jh = journal_alloc_journal_head();
1914 memset(new_jh, 0, sizeof(*new_jh));
1915 }
1916
1917 jbd_lock_bh_journal_head(bh);
1918 if (buffer_jbd(bh)) {
1919 jh = bh2jh(bh);
1920 } else {
1921 J_ASSERT_BH(bh,
1922 (atomic_read(&bh->b_count) > 0) ||
1923 (bh->b_page && bh->b_page->mapping));
1924
1925 if (!new_jh) {
1926 jbd_unlock_bh_journal_head(bh);
1927 goto repeat;
1928 }
1929
1930 jh = new_jh;
1931 new_jh = NULL; /* We consumed it */
1932 set_buffer_jbd(bh);
1933 bh->b_private = jh;
1934 jh->b_bh = bh;
1935 get_bh(bh);
1936 BUFFER_TRACE(bh, "added journal_head");
1937 }
1938 jh->b_jcount++;
1939 jbd_unlock_bh_journal_head(bh);
1940 if (new_jh)
1941 journal_free_journal_head(new_jh);
1942 return bh->b_private;
1943}
1944
1945/*
1946 * Grab a ref against this buffer_head's journal_head. If it ended up not
1947 * having a journal_head, return NULL
1948 */
1949struct journal_head *journal_grab_journal_head(struct buffer_head *bh)
1950{
1951 struct journal_head *jh = NULL;
1952
1953 jbd_lock_bh_journal_head(bh);
1954 if (buffer_jbd(bh)) {
1955 jh = bh2jh(bh);
1956 jh->b_jcount++;
1957 }
1958 jbd_unlock_bh_journal_head(bh);
1959 return jh;
1960}
1961
1962static void __journal_remove_journal_head(struct buffer_head *bh)
1963{
1964 struct journal_head *jh = bh2jh(bh);
1965
1966 J_ASSERT_JH(jh, jh->b_jcount >= 0);
Jan Karabb189242011-06-24 23:11:59 +02001967 J_ASSERT_JH(jh, jh->b_transaction == NULL);
1968 J_ASSERT_JH(jh, jh->b_next_transaction == NULL);
1969 J_ASSERT_JH(jh, jh->b_cp_transaction == NULL);
1970 J_ASSERT_JH(jh, jh->b_jlist == BJ_None);
1971 J_ASSERT_BH(bh, buffer_jbd(bh));
1972 J_ASSERT_BH(bh, jh2bh(jh) == bh);
1973 BUFFER_TRACE(bh, "remove journal_head");
1974 if (jh->b_frozen_data) {
1975 printk(KERN_WARNING "%s: freeing b_frozen_data\n", __func__);
1976 jbd_free(jh->b_frozen_data, bh->b_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 }
Jan Karabb189242011-06-24 23:11:59 +02001978 if (jh->b_committed_data) {
1979 printk(KERN_WARNING "%s: freeing b_committed_data\n", __func__);
1980 jbd_free(jh->b_committed_data, bh->b_size);
1981 }
1982 bh->b_private = NULL;
1983 jh->b_bh = NULL; /* debug, really */
1984 clear_buffer_jbd(bh);
1985 journal_free_journal_head(jh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986}
1987
1988/*
Jan Karabb189242011-06-24 23:11:59 +02001989 * Drop a reference on the passed journal_head. If it fell to zero then
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 * release the journal_head from the buffer_head.
1991 */
1992void journal_put_journal_head(struct journal_head *jh)
1993{
1994 struct buffer_head *bh = jh2bh(jh);
1995
1996 jbd_lock_bh_journal_head(bh);
1997 J_ASSERT_JH(jh, jh->b_jcount > 0);
1998 --jh->b_jcount;
Jan Karabb189242011-06-24 23:11:59 +02001999 if (!jh->b_jcount) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 __journal_remove_journal_head(bh);
Jan Karabb189242011-06-24 23:11:59 +02002001 jbd_unlock_bh_journal_head(bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 __brelse(bh);
Jan Karabb189242011-06-24 23:11:59 +02002003 } else
2004 jbd_unlock_bh_journal_head(bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005}
2006
2007/*
Jose R. Santosc2a91592007-10-18 23:39:22 -07002008 * debugfs tunables
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 */
Jose R. Santosc2a91592007-10-18 23:39:22 -07002010#ifdef CONFIG_JBD_DEBUG
2011
2012u8 journal_enable_debug __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013EXPORT_SYMBOL(journal_enable_debug);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014
Jose R. Santosc2a91592007-10-18 23:39:22 -07002015static struct dentry *jbd_debugfs_dir;
2016static struct dentry *jbd_debug;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017
Jose R. Santosc2a91592007-10-18 23:39:22 -07002018static void __init jbd_create_debugfs_entry(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019{
Jose R. Santosc2a91592007-10-18 23:39:22 -07002020 jbd_debugfs_dir = debugfs_create_dir("jbd", NULL);
2021 if (jbd_debugfs_dir)
Yin Kangkai765f8362009-12-15 14:48:25 -08002022 jbd_debug = debugfs_create_u8("jbd-debug", S_IRUGO | S_IWUSR,
Jose R. Santosc2a91592007-10-18 23:39:22 -07002023 jbd_debugfs_dir,
2024 &journal_enable_debug);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025}
2026
Jose R. Santosc2a91592007-10-18 23:39:22 -07002027static void __exit jbd_remove_debugfs_entry(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028{
Jose R. Santosc2a91592007-10-18 23:39:22 -07002029 debugfs_remove(jbd_debug);
2030 debugfs_remove(jbd_debugfs_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031}
2032
2033#else
2034
Jose R. Santosc2a91592007-10-18 23:39:22 -07002035static inline void jbd_create_debugfs_entry(void)
2036{
2037}
2038
2039static inline void jbd_remove_debugfs_entry(void)
2040{
2041}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042
2043#endif
2044
Christoph Lametere18b8902006-12-06 20:33:20 -08002045struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046
2047static int __init journal_init_handle_cache(void)
2048{
2049 jbd_handle_cache = kmem_cache_create("journal_handle",
2050 sizeof(handle_t),
2051 0, /* offset */
Mel Gormane12ba742007-10-16 01:25:52 -07002052 SLAB_TEMPORARY, /* flags */
Paul Mundt20c2df82007-07-20 10:11:58 +09002053 NULL); /* ctor */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054 if (jbd_handle_cache == NULL) {
2055 printk(KERN_EMERG "JBD: failed to create handle cache\n");
2056 return -ENOMEM;
2057 }
2058 return 0;
2059}
2060
2061static void journal_destroy_handle_cache(void)
2062{
2063 if (jbd_handle_cache)
2064 kmem_cache_destroy(jbd_handle_cache);
2065}
2066
2067/*
2068 * Module startup and shutdown
2069 */
2070
2071static int __init journal_init_caches(void)
2072{
2073 int ret;
2074
2075 ret = journal_init_revoke_caches();
2076 if (ret == 0)
2077 ret = journal_init_journal_head_cache();
2078 if (ret == 0)
2079 ret = journal_init_handle_cache();
2080 return ret;
2081}
2082
2083static void journal_destroy_caches(void)
2084{
2085 journal_destroy_revoke_caches();
2086 journal_destroy_journal_head_cache();
2087 journal_destroy_handle_cache();
2088}
2089
2090static int __init journal_init(void)
2091{
2092 int ret;
2093
Alexey Dobriyan2aed3482006-09-27 01:49:28 -07002094 BUILD_BUG_ON(sizeof(struct journal_superblock_s) != 1024);
Adrian Bunk022a4a72005-09-06 15:16:41 -07002095
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 ret = journal_init_caches();
2097 if (ret != 0)
2098 journal_destroy_caches();
Jose R. Santosc2a91592007-10-18 23:39:22 -07002099 jbd_create_debugfs_entry();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100 return ret;
2101}
2102
2103static void __exit journal_exit(void)
2104{
2105#ifdef CONFIG_JBD_DEBUG
2106 int n = atomic_read(&nr_journal_heads);
2107 if (n)
2108 printk(KERN_EMERG "JBD: leaked %d journal_heads!\n", n);
2109#endif
Jose R. Santosc2a91592007-10-18 23:39:22 -07002110 jbd_remove_debugfs_entry();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 journal_destroy_caches();
2112}
2113
2114MODULE_LICENSE("GPL");
2115module_init(journal_init);
2116module_exit(journal_exit);
2117