2 * linux/fs/jbd2/journal.c
4 * Written by Stephen C. Tweedie <sct@redhat.com>, 1998
6 * Copyright 1998 Red Hat corp --- All Rights Reserved
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.
12 * Generic filesystem journal-writing code; part of the ext2fs
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.
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).
25 #include <linux/module.h>
26 #include <linux/time.h>
28 #include <linux/jbd2.h>
29 #include <linux/errno.h>
30 #include <linux/slab.h>
31 #include <linux/init.h>
33 #include <linux/freezer.h>
34 #include <linux/pagemap.h>
35 #include <linux/kthread.h>
36 #include <linux/poison.h>
37 #include <linux/proc_fs.h>
38 #include <linux/debugfs.h>
39 #include <linux/seq_file.h>
41 #include <asm/uaccess.h>
43 #include <asm/div64.h>
45 EXPORT_SYMBOL(jbd2_journal_start
);
46 EXPORT_SYMBOL(jbd2_journal_restart
);
47 EXPORT_SYMBOL(jbd2_journal_extend
);
48 EXPORT_SYMBOL(jbd2_journal_stop
);
49 EXPORT_SYMBOL(jbd2_journal_lock_updates
);
50 EXPORT_SYMBOL(jbd2_journal_unlock_updates
);
51 EXPORT_SYMBOL(jbd2_journal_get_write_access
);
52 EXPORT_SYMBOL(jbd2_journal_get_create_access
);
53 EXPORT_SYMBOL(jbd2_journal_get_undo_access
);
54 EXPORT_SYMBOL(jbd2_journal_dirty_metadata
);
55 EXPORT_SYMBOL(jbd2_journal_release_buffer
);
56 EXPORT_SYMBOL(jbd2_journal_forget
);
58 EXPORT_SYMBOL(journal_sync_buffer
);
60 EXPORT_SYMBOL(jbd2_journal_flush
);
61 EXPORT_SYMBOL(jbd2_journal_revoke
);
63 EXPORT_SYMBOL(jbd2_journal_init_dev
);
64 EXPORT_SYMBOL(jbd2_journal_init_inode
);
65 EXPORT_SYMBOL(jbd2_journal_update_format
);
66 EXPORT_SYMBOL(jbd2_journal_check_used_features
);
67 EXPORT_SYMBOL(jbd2_journal_check_available_features
);
68 EXPORT_SYMBOL(jbd2_journal_set_features
);
69 EXPORT_SYMBOL(jbd2_journal_load
);
70 EXPORT_SYMBOL(jbd2_journal_destroy
);
71 EXPORT_SYMBOL(jbd2_journal_abort
);
72 EXPORT_SYMBOL(jbd2_journal_errno
);
73 EXPORT_SYMBOL(jbd2_journal_ack_err
);
74 EXPORT_SYMBOL(jbd2_journal_clear_err
);
75 EXPORT_SYMBOL(jbd2_log_wait_commit
);
76 EXPORT_SYMBOL(jbd2_journal_start_commit
);
77 EXPORT_SYMBOL(jbd2_journal_force_commit_nested
);
78 EXPORT_SYMBOL(jbd2_journal_wipe
);
79 EXPORT_SYMBOL(jbd2_journal_blocks_per_page
);
80 EXPORT_SYMBOL(jbd2_journal_invalidatepage
);
81 EXPORT_SYMBOL(jbd2_journal_try_to_free_buffers
);
82 EXPORT_SYMBOL(jbd2_journal_force_commit
);
83 EXPORT_SYMBOL(jbd2_journal_file_inode
);
84 EXPORT_SYMBOL(jbd2_journal_init_jbd_inode
);
85 EXPORT_SYMBOL(jbd2_journal_release_jbd_inode
);
86 EXPORT_SYMBOL(jbd2_journal_begin_ordered_truncate
);
88 static int journal_convert_superblock_v1(journal_t
*, journal_superblock_t
*);
89 static void __journal_abort_soft (journal_t
*journal
, int errno
);
92 * Helper function used to manage commit timeouts
95 static void commit_timeout(unsigned long __data
)
97 struct task_struct
* p
= (struct task_struct
*) __data
;
103 * kjournald2: The main thread function used to manage a logging device
106 * This kernel thread is responsible for two things:
108 * 1) COMMIT: Every so often we need to commit the current state of the
109 * filesystem to disk. The journal thread is responsible for writing
110 * all of the metadata buffers to disk.
112 * 2) CHECKPOINT: We cannot reuse a used section of the log file until all
113 * of the data in that part of the log has been rewritten elsewhere on
114 * the disk. Flushing these old buffers to reclaim space in the log is
115 * known as checkpointing, and this thread is responsible for that job.
118 static int kjournald2(void *arg
)
120 journal_t
*journal
= arg
;
121 transaction_t
*transaction
;
124 * Set up an interval timer which can be used to trigger a commit wakeup
125 * after the commit interval expires
127 setup_timer(&journal
->j_commit_timer
, commit_timeout
,
128 (unsigned long)current
);
130 /* Record that the journal thread is running */
131 journal
->j_task
= current
;
132 wake_up(&journal
->j_wait_done_commit
);
134 printk(KERN_INFO
"kjournald2 starting. Commit interval %ld seconds\n",
135 journal
->j_commit_interval
/ HZ
);
138 * And now, wait forever for commit wakeup events.
140 spin_lock(&journal
->j_state_lock
);
143 if (journal
->j_flags
& JBD2_UNMOUNT
)
146 jbd_debug(1, "commit_sequence=%d, commit_request=%d\n",
147 journal
->j_commit_sequence
, journal
->j_commit_request
);
149 if (journal
->j_commit_sequence
!= journal
->j_commit_request
) {
150 jbd_debug(1, "OK, requests differ\n");
151 spin_unlock(&journal
->j_state_lock
);
152 del_timer_sync(&journal
->j_commit_timer
);
153 jbd2_journal_commit_transaction(journal
);
154 spin_lock(&journal
->j_state_lock
);
158 wake_up(&journal
->j_wait_done_commit
);
159 if (freezing(current
)) {
161 * The simpler the better. Flushing journal isn't a
162 * good idea, because that depends on threads that may
163 * be already stopped.
165 jbd_debug(1, "Now suspending kjournald2\n");
166 spin_unlock(&journal
->j_state_lock
);
168 spin_lock(&journal
->j_state_lock
);
171 * We assume on resume that commits are already there,
175 int should_sleep
= 1;
177 prepare_to_wait(&journal
->j_wait_commit
, &wait
,
179 if (journal
->j_commit_sequence
!= journal
->j_commit_request
)
181 transaction
= journal
->j_running_transaction
;
182 if (transaction
&& time_after_eq(jiffies
,
183 transaction
->t_expires
))
185 if (journal
->j_flags
& JBD2_UNMOUNT
)
188 spin_unlock(&journal
->j_state_lock
);
190 spin_lock(&journal
->j_state_lock
);
192 finish_wait(&journal
->j_wait_commit
, &wait
);
195 jbd_debug(1, "kjournald2 wakes\n");
198 * Were we woken up by a commit wakeup event?
200 transaction
= journal
->j_running_transaction
;
201 if (transaction
&& time_after_eq(jiffies
, transaction
->t_expires
)) {
202 journal
->j_commit_request
= transaction
->t_tid
;
203 jbd_debug(1, "woke because of timeout\n");
208 spin_unlock(&journal
->j_state_lock
);
209 del_timer_sync(&journal
->j_commit_timer
);
210 journal
->j_task
= NULL
;
211 wake_up(&journal
->j_wait_done_commit
);
212 jbd_debug(1, "Journal thread exiting.\n");
216 static int jbd2_journal_start_thread(journal_t
*journal
)
218 struct task_struct
*t
;
220 t
= kthread_run(kjournald2
, journal
, "kjournald2");
224 wait_event(journal
->j_wait_done_commit
, journal
->j_task
!= NULL
);
228 static void journal_kill_thread(journal_t
*journal
)
230 spin_lock(&journal
->j_state_lock
);
231 journal
->j_flags
|= JBD2_UNMOUNT
;
233 while (journal
->j_task
) {
234 wake_up(&journal
->j_wait_commit
);
235 spin_unlock(&journal
->j_state_lock
);
236 wait_event(journal
->j_wait_done_commit
, journal
->j_task
== NULL
);
237 spin_lock(&journal
->j_state_lock
);
239 spin_unlock(&journal
->j_state_lock
);
243 * jbd2_journal_write_metadata_buffer: write a metadata buffer to the journal.
245 * Writes a metadata buffer to a given disk block. The actual IO is not
246 * performed but a new buffer_head is constructed which labels the data
247 * to be written with the correct destination disk block.
249 * Any magic-number escaping which needs to be done will cause a
250 * copy-out here. If the buffer happens to start with the
251 * JBD2_MAGIC_NUMBER, then we can't write it to the log directly: the
252 * magic number is only written to the log for descripter blocks. In
253 * this case, we copy the data and replace the first word with 0, and we
254 * return a result code which indicates that this buffer needs to be
255 * marked as an escaped buffer in the corresponding log descriptor
256 * block. The missing word can then be restored when the block is read
259 * If the source buffer has already been modified by a new transaction
260 * since we took the last commit snapshot, we use the frozen copy of
261 * that data for IO. If we end up using the existing buffer_head's data
262 * for the write, then we *have* to lock the buffer to prevent anyone
263 * else from using and possibly modifying it while the IO is in
266 * The function returns a pointer to the buffer_heads to be used for IO.
268 * We assume that the journal has already been locked in this function.
275 * Bit 0 set == escape performed on the data
276 * Bit 1 set == buffer copy-out performed (kfree the data after IO)
279 int jbd2_journal_write_metadata_buffer(transaction_t
*transaction
,
280 struct journal_head
*jh_in
,
281 struct journal_head
**jh_out
,
282 unsigned long long blocknr
)
284 int need_copy_out
= 0;
285 int done_copy_out
= 0;
288 struct buffer_head
*new_bh
;
289 struct journal_head
*new_jh
;
290 struct page
*new_page
;
291 unsigned int new_offset
;
292 struct buffer_head
*bh_in
= jh2bh(jh_in
);
295 * The buffer really shouldn't be locked: only the current committing
296 * transaction is allowed to write it, so nobody else is allowed
299 * akpm: except if we're journalling data, and write() output is
300 * also part of a shared mapping, and another thread has
301 * decided to launch a writepage() against this buffer.
303 J_ASSERT_BH(bh_in
, buffer_jbddirty(bh_in
));
305 new_bh
= alloc_buffer_head(GFP_NOFS
|__GFP_NOFAIL
);
308 * If a new transaction has already done a buffer copy-out, then
309 * we use that version of the data for the commit.
311 jbd_lock_bh_state(bh_in
);
313 if (jh_in
->b_frozen_data
) {
315 new_page
= virt_to_page(jh_in
->b_frozen_data
);
316 new_offset
= offset_in_page(jh_in
->b_frozen_data
);
318 new_page
= jh2bh(jh_in
)->b_page
;
319 new_offset
= offset_in_page(jh2bh(jh_in
)->b_data
);
322 mapped_data
= kmap_atomic(new_page
, KM_USER0
);
326 if (*((__be32
*)(mapped_data
+ new_offset
)) ==
327 cpu_to_be32(JBD2_MAGIC_NUMBER
)) {
331 kunmap_atomic(mapped_data
, KM_USER0
);
334 * Do we need to do a data copy?
336 if (need_copy_out
&& !done_copy_out
) {
339 jbd_unlock_bh_state(bh_in
);
340 tmp
= jbd2_alloc(bh_in
->b_size
, GFP_NOFS
);
341 jbd_lock_bh_state(bh_in
);
342 if (jh_in
->b_frozen_data
) {
343 jbd2_free(tmp
, bh_in
->b_size
);
347 jh_in
->b_frozen_data
= tmp
;
348 mapped_data
= kmap_atomic(new_page
, KM_USER0
);
349 memcpy(tmp
, mapped_data
+ new_offset
, jh2bh(jh_in
)->b_size
);
350 kunmap_atomic(mapped_data
, KM_USER0
);
352 new_page
= virt_to_page(tmp
);
353 new_offset
= offset_in_page(tmp
);
358 * Did we need to do an escaping? Now we've done all the
359 * copying, we can finally do so.
362 mapped_data
= kmap_atomic(new_page
, KM_USER0
);
363 *((unsigned int *)(mapped_data
+ new_offset
)) = 0;
364 kunmap_atomic(mapped_data
, KM_USER0
);
367 /* keep subsequent assertions sane */
369 init_buffer(new_bh
, NULL
, NULL
);
370 atomic_set(&new_bh
->b_count
, 1);
371 jbd_unlock_bh_state(bh_in
);
373 new_jh
= jbd2_journal_add_journal_head(new_bh
); /* This sleeps */
375 set_bh_page(new_bh
, new_page
, new_offset
);
376 new_jh
->b_transaction
= NULL
;
377 new_bh
->b_size
= jh2bh(jh_in
)->b_size
;
378 new_bh
->b_bdev
= transaction
->t_journal
->j_dev
;
379 new_bh
->b_blocknr
= blocknr
;
380 set_buffer_mapped(new_bh
);
381 set_buffer_dirty(new_bh
);
386 * The to-be-written buffer needs to get moved to the io queue,
387 * and the original buffer whose contents we are shadowing or
388 * copying is moved to the transaction's shadow queue.
390 JBUFFER_TRACE(jh_in
, "file as BJ_Shadow");
391 jbd2_journal_file_buffer(jh_in
, transaction
, BJ_Shadow
);
392 JBUFFER_TRACE(new_jh
, "file as BJ_IO");
393 jbd2_journal_file_buffer(new_jh
, transaction
, BJ_IO
);
395 return do_escape
| (done_copy_out
<< 1);
399 * Allocation code for the journal file. Manage the space left in the
400 * journal, so that we can begin checkpointing when appropriate.
404 * __jbd2_log_space_left: Return the number of free blocks left in the journal.
406 * Called with the journal already locked.
408 * Called under j_state_lock
411 int __jbd2_log_space_left(journal_t
*journal
)
413 int left
= journal
->j_free
;
415 assert_spin_locked(&journal
->j_state_lock
);
418 * Be pessimistic here about the number of those free blocks which
419 * might be required for log descriptor control blocks.
422 #define MIN_LOG_RESERVED_BLOCKS 32 /* Allow for rounding errors */
424 left
-= MIN_LOG_RESERVED_BLOCKS
;
433 * Called under j_state_lock. Returns true if a transaction was started.
435 int __jbd2_log_start_commit(journal_t
*journal
, tid_t target
)
438 * Are we already doing a recent enough commit?
440 if (!tid_geq(journal
->j_commit_request
, target
)) {
442 * We want a new commit: OK, mark the request and wakup the
443 * commit thread. We do _not_ do the commit ourselves.
446 journal
->j_commit_request
= target
;
447 jbd_debug(1, "JBD: requesting commit %d/%d\n",
448 journal
->j_commit_request
,
449 journal
->j_commit_sequence
);
450 wake_up(&journal
->j_wait_commit
);
456 int jbd2_log_start_commit(journal_t
*journal
, tid_t tid
)
460 spin_lock(&journal
->j_state_lock
);
461 ret
= __jbd2_log_start_commit(journal
, tid
);
462 spin_unlock(&journal
->j_state_lock
);
467 * Force and wait upon a commit if the calling process is not within
468 * transaction. This is used for forcing out undo-protected data which contains
469 * bitmaps, when the fs is running out of space.
471 * We can only force the running transaction if we don't have an active handle;
472 * otherwise, we will deadlock.
474 * Returns true if a transaction was started.
476 int jbd2_journal_force_commit_nested(journal_t
*journal
)
478 transaction_t
*transaction
= NULL
;
481 spin_lock(&journal
->j_state_lock
);
482 if (journal
->j_running_transaction
&& !current
->journal_info
) {
483 transaction
= journal
->j_running_transaction
;
484 __jbd2_log_start_commit(journal
, transaction
->t_tid
);
485 } else if (journal
->j_committing_transaction
)
486 transaction
= journal
->j_committing_transaction
;
489 spin_unlock(&journal
->j_state_lock
);
490 return 0; /* Nothing to retry */
493 tid
= transaction
->t_tid
;
494 spin_unlock(&journal
->j_state_lock
);
495 jbd2_log_wait_commit(journal
, tid
);
500 * Start a commit of the current running transaction (if any). Returns true
501 * if a transaction was started, and fills its tid in at *ptid
503 int jbd2_journal_start_commit(journal_t
*journal
, tid_t
*ptid
)
507 spin_lock(&journal
->j_state_lock
);
508 if (journal
->j_running_transaction
) {
509 tid_t tid
= journal
->j_running_transaction
->t_tid
;
511 ret
= __jbd2_log_start_commit(journal
, tid
);
514 } else if (journal
->j_committing_transaction
&& ptid
) {
516 * If ext3_write_super() recently started a commit, then we
517 * have to wait for completion of that transaction
519 *ptid
= journal
->j_committing_transaction
->t_tid
;
522 spin_unlock(&journal
->j_state_lock
);
527 * Wait for a specified commit to complete.
528 * The caller may not hold the journal lock.
530 int jbd2_log_wait_commit(journal_t
*journal
, tid_t tid
)
534 #ifdef CONFIG_JBD2_DEBUG
535 spin_lock(&journal
->j_state_lock
);
536 if (!tid_geq(journal
->j_commit_request
, tid
)) {
538 "%s: error: j_commit_request=%d, tid=%d\n",
539 __func__
, journal
->j_commit_request
, tid
);
541 spin_unlock(&journal
->j_state_lock
);
543 spin_lock(&journal
->j_state_lock
);
544 while (tid_gt(tid
, journal
->j_commit_sequence
)) {
545 jbd_debug(1, "JBD: want %d, j_commit_sequence=%d\n",
546 tid
, journal
->j_commit_sequence
);
547 wake_up(&journal
->j_wait_commit
);
548 spin_unlock(&journal
->j_state_lock
);
549 wait_event(journal
->j_wait_done_commit
,
550 !tid_gt(tid
, journal
->j_commit_sequence
));
551 spin_lock(&journal
->j_state_lock
);
553 spin_unlock(&journal
->j_state_lock
);
555 if (unlikely(is_journal_aborted(journal
))) {
556 printk(KERN_EMERG
"journal commit I/O error\n");
563 * Log buffer allocation routines:
566 int jbd2_journal_next_log_block(journal_t
*journal
, unsigned long long *retp
)
568 unsigned long blocknr
;
570 spin_lock(&journal
->j_state_lock
);
571 J_ASSERT(journal
->j_free
> 1);
573 blocknr
= journal
->j_head
;
576 if (journal
->j_head
== journal
->j_last
)
577 journal
->j_head
= journal
->j_first
;
578 spin_unlock(&journal
->j_state_lock
);
579 return jbd2_journal_bmap(journal
, blocknr
, retp
);
583 * Conversion of logical to physical block numbers for the journal
585 * On external journals the journal blocks are identity-mapped, so
586 * this is a no-op. If needed, we can use j_blk_offset - everything is
589 int jbd2_journal_bmap(journal_t
*journal
, unsigned long blocknr
,
590 unsigned long long *retp
)
593 unsigned long long ret
;
595 if (journal
->j_inode
) {
596 ret
= bmap(journal
->j_inode
, blocknr
);
600 printk(KERN_ALERT
"%s: journal block not found "
601 "at offset %lu on %s\n",
602 __func__
, blocknr
, journal
->j_devname
);
604 __journal_abort_soft(journal
, err
);
607 *retp
= blocknr
; /* +journal->j_blk_offset */
613 * We play buffer_head aliasing tricks to write data/metadata blocks to
614 * the journal without copying their contents, but for journal
615 * descriptor blocks we do need to generate bona fide buffers.
617 * After the caller of jbd2_journal_get_descriptor_buffer() has finished modifying
618 * the buffer's contents they really should run flush_dcache_page(bh->b_page).
619 * But we don't bother doing that, so there will be coherency problems with
620 * mmaps of blockdevs which hold live JBD-controlled filesystems.
622 struct journal_head
*jbd2_journal_get_descriptor_buffer(journal_t
*journal
)
624 struct buffer_head
*bh
;
625 unsigned long long blocknr
;
628 err
= jbd2_journal_next_log_block(journal
, &blocknr
);
633 bh
= __getblk(journal
->j_dev
, blocknr
, journal
->j_blocksize
);
635 memset(bh
->b_data
, 0, journal
->j_blocksize
);
636 set_buffer_uptodate(bh
);
638 BUFFER_TRACE(bh
, "return this buffer");
639 return jbd2_journal_add_journal_head(bh
);
642 struct jbd2_stats_proc_session
{
644 struct transaction_stats_s
*stats
;
649 static void *jbd2_history_skip_empty(struct jbd2_stats_proc_session
*s
,
650 struct transaction_stats_s
*ts
,
653 if (ts
== s
->stats
+ s
->max
)
655 if (!first
&& ts
== s
->stats
+ s
->start
)
657 while (ts
->ts_type
== 0) {
659 if (ts
== s
->stats
+ s
->max
)
661 if (ts
== s
->stats
+ s
->start
)
668 static void *jbd2_seq_history_start(struct seq_file
*seq
, loff_t
*pos
)
670 struct jbd2_stats_proc_session
*s
= seq
->private;
671 struct transaction_stats_s
*ts
;
675 return SEQ_START_TOKEN
;
676 ts
= jbd2_history_skip_empty(s
, s
->stats
+ s
->start
, 1);
681 ts
= jbd2_history_skip_empty(s
, ++ts
, 0);
689 static void *jbd2_seq_history_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
691 struct jbd2_stats_proc_session
*s
= seq
->private;
692 struct transaction_stats_s
*ts
= v
;
695 if (v
== SEQ_START_TOKEN
)
696 return jbd2_history_skip_empty(s
, s
->stats
+ s
->start
, 1);
698 return jbd2_history_skip_empty(s
, ++ts
, 0);
701 static int jbd2_seq_history_show(struct seq_file
*seq
, void *v
)
703 struct transaction_stats_s
*ts
= v
;
704 if (v
== SEQ_START_TOKEN
) {
705 seq_printf(seq
, "%-4s %-5s %-5s %-5s %-5s %-5s %-5s %-6s %-5s "
706 "%-5s %-5s %-5s %-5s %-5s\n", "R/C", "tid",
707 "wait", "run", "lock", "flush", "log", "hndls",
708 "block", "inlog", "ctime", "write", "drop",
712 if (ts
->ts_type
== JBD2_STATS_RUN
)
713 seq_printf(seq
, "%-4s %-5lu %-5u %-5u %-5u %-5u %-5u "
714 "%-6lu %-5lu %-5lu\n", "R", ts
->ts_tid
,
715 jiffies_to_msecs(ts
->u
.run
.rs_wait
),
716 jiffies_to_msecs(ts
->u
.run
.rs_running
),
717 jiffies_to_msecs(ts
->u
.run
.rs_locked
),
718 jiffies_to_msecs(ts
->u
.run
.rs_flushing
),
719 jiffies_to_msecs(ts
->u
.run
.rs_logging
),
720 ts
->u
.run
.rs_handle_count
,
722 ts
->u
.run
.rs_blocks_logged
);
723 else if (ts
->ts_type
== JBD2_STATS_CHECKPOINT
)
724 seq_printf(seq
, "%-4s %-5lu %48s %-5u %-5lu %-5lu %-5lu\n",
725 "C", ts
->ts_tid
, " ",
726 jiffies_to_msecs(ts
->u
.chp
.cs_chp_time
),
727 ts
->u
.chp
.cs_written
, ts
->u
.chp
.cs_dropped
,
728 ts
->u
.chp
.cs_forced_to_close
);
734 static void jbd2_seq_history_stop(struct seq_file
*seq
, void *v
)
738 static struct seq_operations jbd2_seq_history_ops
= {
739 .start
= jbd2_seq_history_start
,
740 .next
= jbd2_seq_history_next
,
741 .stop
= jbd2_seq_history_stop
,
742 .show
= jbd2_seq_history_show
,
745 static int jbd2_seq_history_open(struct inode
*inode
, struct file
*file
)
747 journal_t
*journal
= PDE(inode
)->data
;
748 struct jbd2_stats_proc_session
*s
;
751 s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
754 size
= sizeof(struct transaction_stats_s
) * journal
->j_history_max
;
755 s
->stats
= kmalloc(size
, GFP_KERNEL
);
756 if (s
->stats
== NULL
) {
760 spin_lock(&journal
->j_history_lock
);
761 memcpy(s
->stats
, journal
->j_history
, size
);
762 s
->max
= journal
->j_history_max
;
763 s
->start
= journal
->j_history_cur
% s
->max
;
764 spin_unlock(&journal
->j_history_lock
);
766 rc
= seq_open(file
, &jbd2_seq_history_ops
);
768 struct seq_file
*m
= file
->private_data
;
778 static int jbd2_seq_history_release(struct inode
*inode
, struct file
*file
)
780 struct seq_file
*seq
= file
->private_data
;
781 struct jbd2_stats_proc_session
*s
= seq
->private;
785 return seq_release(inode
, file
);
788 static struct file_operations jbd2_seq_history_fops
= {
789 .owner
= THIS_MODULE
,
790 .open
= jbd2_seq_history_open
,
793 .release
= jbd2_seq_history_release
,
796 static void *jbd2_seq_info_start(struct seq_file
*seq
, loff_t
*pos
)
798 return *pos
? NULL
: SEQ_START_TOKEN
;
801 static void *jbd2_seq_info_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
806 static int jbd2_seq_info_show(struct seq_file
*seq
, void *v
)
808 struct jbd2_stats_proc_session
*s
= seq
->private;
810 if (v
!= SEQ_START_TOKEN
)
812 seq_printf(seq
, "%lu transaction, each upto %u blocks\n",
814 s
->journal
->j_max_transaction_buffers
);
815 if (s
->stats
->ts_tid
== 0)
817 seq_printf(seq
, "average: \n %ums waiting for transaction\n",
818 jiffies_to_msecs(s
->stats
->u
.run
.rs_wait
/ s
->stats
->ts_tid
));
819 seq_printf(seq
, " %ums running transaction\n",
820 jiffies_to_msecs(s
->stats
->u
.run
.rs_running
/ s
->stats
->ts_tid
));
821 seq_printf(seq
, " %ums transaction was being locked\n",
822 jiffies_to_msecs(s
->stats
->u
.run
.rs_locked
/ s
->stats
->ts_tid
));
823 seq_printf(seq
, " %ums flushing data (in ordered mode)\n",
824 jiffies_to_msecs(s
->stats
->u
.run
.rs_flushing
/ s
->stats
->ts_tid
));
825 seq_printf(seq
, " %ums logging transaction\n",
826 jiffies_to_msecs(s
->stats
->u
.run
.rs_logging
/ s
->stats
->ts_tid
));
827 seq_printf(seq
, " %luus average transaction commit time\n",
828 do_div(s
->journal
->j_average_commit_time
, 1000));
829 seq_printf(seq
, " %lu handles per transaction\n",
830 s
->stats
->u
.run
.rs_handle_count
/ s
->stats
->ts_tid
);
831 seq_printf(seq
, " %lu blocks per transaction\n",
832 s
->stats
->u
.run
.rs_blocks
/ s
->stats
->ts_tid
);
833 seq_printf(seq
, " %lu logged blocks per transaction\n",
834 s
->stats
->u
.run
.rs_blocks_logged
/ s
->stats
->ts_tid
);
838 static void jbd2_seq_info_stop(struct seq_file
*seq
, void *v
)
842 static struct seq_operations jbd2_seq_info_ops
= {
843 .start
= jbd2_seq_info_start
,
844 .next
= jbd2_seq_info_next
,
845 .stop
= jbd2_seq_info_stop
,
846 .show
= jbd2_seq_info_show
,
849 static int jbd2_seq_info_open(struct inode
*inode
, struct file
*file
)
851 journal_t
*journal
= PDE(inode
)->data
;
852 struct jbd2_stats_proc_session
*s
;
855 s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
858 size
= sizeof(struct transaction_stats_s
);
859 s
->stats
= kmalloc(size
, GFP_KERNEL
);
860 if (s
->stats
== NULL
) {
864 spin_lock(&journal
->j_history_lock
);
865 memcpy(s
->stats
, &journal
->j_stats
, size
);
866 s
->journal
= journal
;
867 spin_unlock(&journal
->j_history_lock
);
869 rc
= seq_open(file
, &jbd2_seq_info_ops
);
871 struct seq_file
*m
= file
->private_data
;
881 static int jbd2_seq_info_release(struct inode
*inode
, struct file
*file
)
883 struct seq_file
*seq
= file
->private_data
;
884 struct jbd2_stats_proc_session
*s
= seq
->private;
887 return seq_release(inode
, file
);
890 static struct file_operations jbd2_seq_info_fops
= {
891 .owner
= THIS_MODULE
,
892 .open
= jbd2_seq_info_open
,
895 .release
= jbd2_seq_info_release
,
898 static struct proc_dir_entry
*proc_jbd2_stats
;
900 static void jbd2_stats_proc_init(journal_t
*journal
)
902 journal
->j_proc_entry
= proc_mkdir(journal
->j_devname
, proc_jbd2_stats
);
903 if (journal
->j_proc_entry
) {
904 proc_create_data("history", S_IRUGO
, journal
->j_proc_entry
,
905 &jbd2_seq_history_fops
, journal
);
906 proc_create_data("info", S_IRUGO
, journal
->j_proc_entry
,
907 &jbd2_seq_info_fops
, journal
);
911 static void jbd2_stats_proc_exit(journal_t
*journal
)
913 remove_proc_entry("info", journal
->j_proc_entry
);
914 remove_proc_entry("history", journal
->j_proc_entry
);
915 remove_proc_entry(journal
->j_devname
, proc_jbd2_stats
);
918 static void journal_init_stats(journal_t
*journal
)
922 if (!proc_jbd2_stats
)
925 journal
->j_history_max
= 100;
926 size
= sizeof(struct transaction_stats_s
) * journal
->j_history_max
;
927 journal
->j_history
= kzalloc(size
, GFP_KERNEL
);
928 if (!journal
->j_history
) {
929 journal
->j_history_max
= 0;
932 spin_lock_init(&journal
->j_history_lock
);
936 * Management for journal control blocks: functions to create and
937 * destroy journal_t structures, and to initialise and read existing
938 * journal blocks from disk. */
940 /* First: create and setup a journal_t object in memory. We initialise
941 * very few fields yet: that has to wait until we have created the
942 * journal structures from from scratch, or loaded them from disk. */
944 static journal_t
* journal_init_common (void)
949 journal
= kzalloc(sizeof(*journal
), GFP_KERNEL
|__GFP_NOFAIL
);
953 init_waitqueue_head(&journal
->j_wait_transaction_locked
);
954 init_waitqueue_head(&journal
->j_wait_logspace
);
955 init_waitqueue_head(&journal
->j_wait_done_commit
);
956 init_waitqueue_head(&journal
->j_wait_checkpoint
);
957 init_waitqueue_head(&journal
->j_wait_commit
);
958 init_waitqueue_head(&journal
->j_wait_updates
);
959 mutex_init(&journal
->j_barrier
);
960 mutex_init(&journal
->j_checkpoint_mutex
);
961 spin_lock_init(&journal
->j_revoke_lock
);
962 spin_lock_init(&journal
->j_list_lock
);
963 spin_lock_init(&journal
->j_state_lock
);
965 journal
->j_commit_interval
= (HZ
* JBD2_DEFAULT_MAX_COMMIT_AGE
);
966 journal
->j_min_batch_time
= 0;
967 journal
->j_max_batch_time
= 15000; /* 15ms */
969 /* The journal is marked for error until we succeed with recovery! */
970 journal
->j_flags
= JBD2_ABORT
;
972 /* Set up a default-sized revoke table for the new mount. */
973 err
= jbd2_journal_init_revoke(journal
, JOURNAL_REVOKE_DEFAULT_HASH
);
979 journal_init_stats(journal
);
986 /* jbd2_journal_init_dev and jbd2_journal_init_inode:
988 * Create a journal structure assigned some fixed set of disk blocks to
989 * the journal. We don't actually touch those disk blocks yet, but we
990 * need to set up all of the mapping information to tell the journaling
991 * system where the journal blocks are.
996 * journal_t * jbd2_journal_init_dev() - creates and initialises a journal structure
997 * @bdev: Block device on which to create the journal
998 * @fs_dev: Device which hold journalled filesystem for this journal.
999 * @start: Block nr Start of journal.
1000 * @len: Length of the journal in blocks.
1001 * @blocksize: blocksize of journalling device
1003 * Returns: a newly created journal_t *
1005 * jbd2_journal_init_dev creates a journal which maps a fixed contiguous
1006 * range of blocks on an arbitrary block device.
1009 journal_t
* jbd2_journal_init_dev(struct block_device
*bdev
,
1010 struct block_device
*fs_dev
,
1011 unsigned long long start
, int len
, int blocksize
)
1013 journal_t
*journal
= journal_init_common();
1014 struct buffer_head
*bh
;
1021 /* journal descriptor can store up to n blocks -bzzz */
1022 journal
->j_blocksize
= blocksize
;
1023 n
= journal
->j_blocksize
/ sizeof(journal_block_tag_t
);
1024 journal
->j_wbufsize
= n
;
1025 journal
->j_wbuf
= kmalloc(n
* sizeof(struct buffer_head
*), GFP_KERNEL
);
1026 if (!journal
->j_wbuf
) {
1027 printk(KERN_ERR
"%s: Cant allocate bhs for commit thread\n",
1033 journal
->j_dev
= bdev
;
1034 journal
->j_fs_dev
= fs_dev
;
1035 journal
->j_blk_offset
= start
;
1036 journal
->j_maxlen
= len
;
1037 bdevname(journal
->j_dev
, journal
->j_devname
);
1038 p
= journal
->j_devname
;
1039 while ((p
= strchr(p
, '/')))
1041 jbd2_stats_proc_init(journal
);
1043 bh
= __getblk(journal
->j_dev
, start
, journal
->j_blocksize
);
1044 J_ASSERT(bh
!= NULL
);
1045 journal
->j_sb_buffer
= bh
;
1046 journal
->j_superblock
= (journal_superblock_t
*)bh
->b_data
;
1052 * journal_t * jbd2_journal_init_inode () - creates a journal which maps to a inode.
1053 * @inode: An inode to create the journal in
1055 * jbd2_journal_init_inode creates a journal which maps an on-disk inode as
1056 * the journal. The inode must exist already, must support bmap() and
1057 * must have all data blocks preallocated.
1059 journal_t
* jbd2_journal_init_inode (struct inode
*inode
)
1061 struct buffer_head
*bh
;
1062 journal_t
*journal
= journal_init_common();
1066 unsigned long long blocknr
;
1071 journal
->j_dev
= journal
->j_fs_dev
= inode
->i_sb
->s_bdev
;
1072 journal
->j_inode
= inode
;
1073 bdevname(journal
->j_dev
, journal
->j_devname
);
1074 p
= journal
->j_devname
;
1075 while ((p
= strchr(p
, '/')))
1077 p
= journal
->j_devname
+ strlen(journal
->j_devname
);
1078 sprintf(p
, ":%lu", journal
->j_inode
->i_ino
);
1080 "journal %p: inode %s/%ld, size %Ld, bits %d, blksize %ld\n",
1081 journal
, inode
->i_sb
->s_id
, inode
->i_ino
,
1082 (long long) inode
->i_size
,
1083 inode
->i_sb
->s_blocksize_bits
, inode
->i_sb
->s_blocksize
);
1085 journal
->j_maxlen
= inode
->i_size
>> inode
->i_sb
->s_blocksize_bits
;
1086 journal
->j_blocksize
= inode
->i_sb
->s_blocksize
;
1087 jbd2_stats_proc_init(journal
);
1089 /* journal descriptor can store up to n blocks -bzzz */
1090 n
= journal
->j_blocksize
/ sizeof(journal_block_tag_t
);
1091 journal
->j_wbufsize
= n
;
1092 journal
->j_wbuf
= kmalloc(n
* sizeof(struct buffer_head
*), GFP_KERNEL
);
1093 if (!journal
->j_wbuf
) {
1094 printk(KERN_ERR
"%s: Cant allocate bhs for commit thread\n",
1096 jbd2_stats_proc_exit(journal
);
1101 err
= jbd2_journal_bmap(journal
, 0, &blocknr
);
1102 /* If that failed, give up */
1104 printk(KERN_ERR
"%s: Cannnot locate journal superblock\n",
1106 jbd2_stats_proc_exit(journal
);
1111 bh
= __getblk(journal
->j_dev
, blocknr
, journal
->j_blocksize
);
1112 J_ASSERT(bh
!= NULL
);
1113 journal
->j_sb_buffer
= bh
;
1114 journal
->j_superblock
= (journal_superblock_t
*)bh
->b_data
;
1120 * If the journal init or create aborts, we need to mark the journal
1121 * superblock as being NULL to prevent the journal destroy from writing
1122 * back a bogus superblock.
1124 static void journal_fail_superblock (journal_t
*journal
)
1126 struct buffer_head
*bh
= journal
->j_sb_buffer
;
1128 journal
->j_sb_buffer
= NULL
;
1132 * Given a journal_t structure, initialise the various fields for
1133 * startup of a new journaling session. We use this both when creating
1134 * a journal, and after recovering an old journal to reset it for
1138 static int journal_reset(journal_t
*journal
)
1140 journal_superblock_t
*sb
= journal
->j_superblock
;
1141 unsigned long long first
, last
;
1143 first
= be32_to_cpu(sb
->s_first
);
1144 last
= be32_to_cpu(sb
->s_maxlen
);
1146 journal
->j_first
= first
;
1147 journal
->j_last
= last
;
1149 journal
->j_head
= first
;
1150 journal
->j_tail
= first
;
1151 journal
->j_free
= last
- first
;
1153 journal
->j_tail_sequence
= journal
->j_transaction_sequence
;
1154 journal
->j_commit_sequence
= journal
->j_transaction_sequence
- 1;
1155 journal
->j_commit_request
= journal
->j_commit_sequence
;
1157 journal
->j_max_transaction_buffers
= journal
->j_maxlen
/ 4;
1159 /* Add the dynamic fields and write it to disk. */
1160 jbd2_journal_update_superblock(journal
, 1);
1161 return jbd2_journal_start_thread(journal
);
1165 * void jbd2_journal_update_superblock() - Update journal sb on disk.
1166 * @journal: The journal to update.
1167 * @wait: Set to '0' if you don't want to wait for IO completion.
1169 * Update a journal's dynamic superblock fields and write it to disk,
1170 * optionally waiting for the IO to complete.
1172 void jbd2_journal_update_superblock(journal_t
*journal
, int wait
)
1174 journal_superblock_t
*sb
= journal
->j_superblock
;
1175 struct buffer_head
*bh
= journal
->j_sb_buffer
;
1178 * As a special case, if the on-disk copy is already marked as needing
1179 * no recovery (s_start == 0) and there are no outstanding transactions
1180 * in the filesystem, then we can safely defer the superblock update
1181 * until the next commit by setting JBD2_FLUSHED. This avoids
1182 * attempting a write to a potential-readonly device.
1184 if (sb
->s_start
== 0 && journal
->j_tail_sequence
==
1185 journal
->j_transaction_sequence
) {
1186 jbd_debug(1,"JBD: Skipping superblock update on recovered sb "
1187 "(start %ld, seq %d, errno %d)\n",
1188 journal
->j_tail
, journal
->j_tail_sequence
,
1193 if (buffer_write_io_error(bh
)) {
1195 * Oh, dear. A previous attempt to write the journal
1196 * superblock failed. This could happen because the
1197 * USB device was yanked out. Or it could happen to
1198 * be a transient write error and maybe the block will
1199 * be remapped. Nothing we can do but to retry the
1200 * write and hope for the best.
1202 printk(KERN_ERR
"JBD2: previous I/O error detected "
1203 "for journal superblock update for %s.\n",
1204 journal
->j_devname
);
1205 clear_buffer_write_io_error(bh
);
1206 set_buffer_uptodate(bh
);
1209 spin_lock(&journal
->j_state_lock
);
1210 jbd_debug(1,"JBD: updating superblock (start %ld, seq %d, errno %d)\n",
1211 journal
->j_tail
, journal
->j_tail_sequence
, journal
->j_errno
);
1213 sb
->s_sequence
= cpu_to_be32(journal
->j_tail_sequence
);
1214 sb
->s_start
= cpu_to_be32(journal
->j_tail
);
1215 sb
->s_errno
= cpu_to_be32(journal
->j_errno
);
1216 spin_unlock(&journal
->j_state_lock
);
1218 BUFFER_TRACE(bh
, "marking dirty");
1219 mark_buffer_dirty(bh
);
1221 sync_dirty_buffer(bh
);
1222 if (buffer_write_io_error(bh
)) {
1223 printk(KERN_ERR
"JBD2: I/O error detected "
1224 "when updating journal superblock for %s.\n",
1225 journal
->j_devname
);
1226 clear_buffer_write_io_error(bh
);
1227 set_buffer_uptodate(bh
);
1230 ll_rw_block(SWRITE
, 1, &bh
);
1233 /* If we have just flushed the log (by marking s_start==0), then
1234 * any future commit will have to be careful to update the
1235 * superblock again to re-record the true start of the log. */
1237 spin_lock(&journal
->j_state_lock
);
1239 journal
->j_flags
&= ~JBD2_FLUSHED
;
1241 journal
->j_flags
|= JBD2_FLUSHED
;
1242 spin_unlock(&journal
->j_state_lock
);
1246 * Read the superblock for a given journal, performing initial
1247 * validation of the format.
1250 static int journal_get_superblock(journal_t
*journal
)
1252 struct buffer_head
*bh
;
1253 journal_superblock_t
*sb
;
1256 bh
= journal
->j_sb_buffer
;
1258 J_ASSERT(bh
!= NULL
);
1259 if (!buffer_uptodate(bh
)) {
1260 ll_rw_block(READ
, 1, &bh
);
1262 if (!buffer_uptodate(bh
)) {
1264 "JBD: IO error reading journal superblock\n");
1269 sb
= journal
->j_superblock
;
1273 if (sb
->s_header
.h_magic
!= cpu_to_be32(JBD2_MAGIC_NUMBER
) ||
1274 sb
->s_blocksize
!= cpu_to_be32(journal
->j_blocksize
)) {
1275 printk(KERN_WARNING
"JBD: no valid journal superblock found\n");
1279 switch(be32_to_cpu(sb
->s_header
.h_blocktype
)) {
1280 case JBD2_SUPERBLOCK_V1
:
1281 journal
->j_format_version
= 1;
1283 case JBD2_SUPERBLOCK_V2
:
1284 journal
->j_format_version
= 2;
1287 printk(KERN_WARNING
"JBD: unrecognised superblock format ID\n");
1291 if (be32_to_cpu(sb
->s_maxlen
) < journal
->j_maxlen
)
1292 journal
->j_maxlen
= be32_to_cpu(sb
->s_maxlen
);
1293 else if (be32_to_cpu(sb
->s_maxlen
) > journal
->j_maxlen
) {
1294 printk (KERN_WARNING
"JBD: journal file too short\n");
1301 journal_fail_superblock(journal
);
1306 * Load the on-disk journal superblock and read the key fields into the
1310 static int load_superblock(journal_t
*journal
)
1313 journal_superblock_t
*sb
;
1315 err
= journal_get_superblock(journal
);
1319 sb
= journal
->j_superblock
;
1321 journal
->j_tail_sequence
= be32_to_cpu(sb
->s_sequence
);
1322 journal
->j_tail
= be32_to_cpu(sb
->s_start
);
1323 journal
->j_first
= be32_to_cpu(sb
->s_first
);
1324 journal
->j_last
= be32_to_cpu(sb
->s_maxlen
);
1325 journal
->j_errno
= be32_to_cpu(sb
->s_errno
);
1332 * int jbd2_journal_load() - Read journal from disk.
1333 * @journal: Journal to act on.
1335 * Given a journal_t structure which tells us which disk blocks contain
1336 * a journal, read the journal from disk to initialise the in-memory
1339 int jbd2_journal_load(journal_t
*journal
)
1342 journal_superblock_t
*sb
;
1344 err
= load_superblock(journal
);
1348 sb
= journal
->j_superblock
;
1349 /* If this is a V2 superblock, then we have to check the
1350 * features flags on it. */
1352 if (journal
->j_format_version
>= 2) {
1353 if ((sb
->s_feature_ro_compat
&
1354 ~cpu_to_be32(JBD2_KNOWN_ROCOMPAT_FEATURES
)) ||
1355 (sb
->s_feature_incompat
&
1356 ~cpu_to_be32(JBD2_KNOWN_INCOMPAT_FEATURES
))) {
1357 printk (KERN_WARNING
1358 "JBD: Unrecognised features on journal\n");
1363 /* Let the recovery code check whether it needs to recover any
1364 * data from the journal. */
1365 if (jbd2_journal_recover(journal
))
1366 goto recovery_error
;
1368 /* OK, we've finished with the dynamic journal bits:
1369 * reinitialise the dynamic contents of the superblock in memory
1370 * and reset them on disk. */
1371 if (journal_reset(journal
))
1372 goto recovery_error
;
1374 journal
->j_flags
&= ~JBD2_ABORT
;
1375 journal
->j_flags
|= JBD2_LOADED
;
1379 printk (KERN_WARNING
"JBD: recovery failed\n");
1384 * void jbd2_journal_destroy() - Release a journal_t structure.
1385 * @journal: Journal to act on.
1387 * Release a journal_t structure once it is no longer in use by the
1389 * Return <0 if we couldn't clean up the journal.
1391 int jbd2_journal_destroy(journal_t
*journal
)
1395 /* Wait for the commit thread to wake up and die. */
1396 journal_kill_thread(journal
);
1398 /* Force a final log commit */
1399 if (journal
->j_running_transaction
)
1400 jbd2_journal_commit_transaction(journal
);
1402 /* Force any old transactions to disk */
1404 /* Totally anal locking here... */
1405 spin_lock(&journal
->j_list_lock
);
1406 while (journal
->j_checkpoint_transactions
!= NULL
) {
1407 spin_unlock(&journal
->j_list_lock
);
1408 mutex_lock(&journal
->j_checkpoint_mutex
);
1409 jbd2_log_do_checkpoint(journal
);
1410 mutex_unlock(&journal
->j_checkpoint_mutex
);
1411 spin_lock(&journal
->j_list_lock
);
1414 J_ASSERT(journal
->j_running_transaction
== NULL
);
1415 J_ASSERT(journal
->j_committing_transaction
== NULL
);
1416 J_ASSERT(journal
->j_checkpoint_transactions
== NULL
);
1417 spin_unlock(&journal
->j_list_lock
);
1419 if (journal
->j_sb_buffer
) {
1420 if (!is_journal_aborted(journal
)) {
1421 /* We can now mark the journal as empty. */
1422 journal
->j_tail
= 0;
1423 journal
->j_tail_sequence
=
1424 ++journal
->j_transaction_sequence
;
1425 jbd2_journal_update_superblock(journal
, 1);
1429 brelse(journal
->j_sb_buffer
);
1432 if (journal
->j_proc_entry
)
1433 jbd2_stats_proc_exit(journal
);
1434 if (journal
->j_inode
)
1435 iput(journal
->j_inode
);
1436 if (journal
->j_revoke
)
1437 jbd2_journal_destroy_revoke(journal
);
1438 kfree(journal
->j_wbuf
);
1446 *int jbd2_journal_check_used_features () - Check if features specified are used.
1447 * @journal: Journal to check.
1448 * @compat: bitmask of compatible features
1449 * @ro: bitmask of features that force read-only mount
1450 * @incompat: bitmask of incompatible features
1452 * Check whether the journal uses all of a given set of
1453 * features. Return true (non-zero) if it does.
1456 int jbd2_journal_check_used_features (journal_t
*journal
, unsigned long compat
,
1457 unsigned long ro
, unsigned long incompat
)
1459 journal_superblock_t
*sb
;
1461 if (!compat
&& !ro
&& !incompat
)
1463 if (journal
->j_format_version
== 1)
1466 sb
= journal
->j_superblock
;
1468 if (((be32_to_cpu(sb
->s_feature_compat
) & compat
) == compat
) &&
1469 ((be32_to_cpu(sb
->s_feature_ro_compat
) & ro
) == ro
) &&
1470 ((be32_to_cpu(sb
->s_feature_incompat
) & incompat
) == incompat
))
1477 * int jbd2_journal_check_available_features() - Check feature set in journalling layer
1478 * @journal: Journal to check.
1479 * @compat: bitmask of compatible features
1480 * @ro: bitmask of features that force read-only mount
1481 * @incompat: bitmask of incompatible features
1483 * Check whether the journaling code supports the use of
1484 * all of a given set of features on this journal. Return true
1485 * (non-zero) if it can. */
1487 int jbd2_journal_check_available_features (journal_t
*journal
, unsigned long compat
,
1488 unsigned long ro
, unsigned long incompat
)
1490 journal_superblock_t
*sb
;
1492 if (!compat
&& !ro
&& !incompat
)
1495 sb
= journal
->j_superblock
;
1497 /* We can support any known requested features iff the
1498 * superblock is in version 2. Otherwise we fail to support any
1499 * extended sb features. */
1501 if (journal
->j_format_version
!= 2)
1504 if ((compat
& JBD2_KNOWN_COMPAT_FEATURES
) == compat
&&
1505 (ro
& JBD2_KNOWN_ROCOMPAT_FEATURES
) == ro
&&
1506 (incompat
& JBD2_KNOWN_INCOMPAT_FEATURES
) == incompat
)
1513 * int jbd2_journal_set_features () - Mark a given journal feature in the superblock
1514 * @journal: Journal to act on.
1515 * @compat: bitmask of compatible features
1516 * @ro: bitmask of features that force read-only mount
1517 * @incompat: bitmask of incompatible features
1519 * Mark a given journal feature as present on the
1520 * superblock. Returns true if the requested features could be set.
1524 int jbd2_journal_set_features (journal_t
*journal
, unsigned long compat
,
1525 unsigned long ro
, unsigned long incompat
)
1527 journal_superblock_t
*sb
;
1529 if (jbd2_journal_check_used_features(journal
, compat
, ro
, incompat
))
1532 if (!jbd2_journal_check_available_features(journal
, compat
, ro
, incompat
))
1535 jbd_debug(1, "Setting new features 0x%lx/0x%lx/0x%lx\n",
1536 compat
, ro
, incompat
);
1538 sb
= journal
->j_superblock
;
1540 sb
->s_feature_compat
|= cpu_to_be32(compat
);
1541 sb
->s_feature_ro_compat
|= cpu_to_be32(ro
);
1542 sb
->s_feature_incompat
|= cpu_to_be32(incompat
);
1548 * jbd2_journal_clear_features () - Clear a given journal feature in the
1550 * @journal: Journal to act on.
1551 * @compat: bitmask of compatible features
1552 * @ro: bitmask of features that force read-only mount
1553 * @incompat: bitmask of incompatible features
1555 * Clear a given journal feature as present on the
1558 void jbd2_journal_clear_features(journal_t
*journal
, unsigned long compat
,
1559 unsigned long ro
, unsigned long incompat
)
1561 journal_superblock_t
*sb
;
1563 jbd_debug(1, "Clear features 0x%lx/0x%lx/0x%lx\n",
1564 compat
, ro
, incompat
);
1566 sb
= journal
->j_superblock
;
1568 sb
->s_feature_compat
&= ~cpu_to_be32(compat
);
1569 sb
->s_feature_ro_compat
&= ~cpu_to_be32(ro
);
1570 sb
->s_feature_incompat
&= ~cpu_to_be32(incompat
);
1572 EXPORT_SYMBOL(jbd2_journal_clear_features
);
1575 * int jbd2_journal_update_format () - Update on-disk journal structure.
1576 * @journal: Journal to act on.
1578 * Given an initialised but unloaded journal struct, poke about in the
1579 * on-disk structure to update it to the most recent supported version.
1581 int jbd2_journal_update_format (journal_t
*journal
)
1583 journal_superblock_t
*sb
;
1586 err
= journal_get_superblock(journal
);
1590 sb
= journal
->j_superblock
;
1592 switch (be32_to_cpu(sb
->s_header
.h_blocktype
)) {
1593 case JBD2_SUPERBLOCK_V2
:
1595 case JBD2_SUPERBLOCK_V1
:
1596 return journal_convert_superblock_v1(journal
, sb
);
1603 static int journal_convert_superblock_v1(journal_t
*journal
,
1604 journal_superblock_t
*sb
)
1606 int offset
, blocksize
;
1607 struct buffer_head
*bh
;
1610 "JBD: Converting superblock from version 1 to 2.\n");
1612 /* Pre-initialise new fields to zero */
1613 offset
= ((char *) &(sb
->s_feature_compat
)) - ((char *) sb
);
1614 blocksize
= be32_to_cpu(sb
->s_blocksize
);
1615 memset(&sb
->s_feature_compat
, 0, blocksize
-offset
);
1617 sb
->s_nr_users
= cpu_to_be32(1);
1618 sb
->s_header
.h_blocktype
= cpu_to_be32(JBD2_SUPERBLOCK_V2
);
1619 journal
->j_format_version
= 2;
1621 bh
= journal
->j_sb_buffer
;
1622 BUFFER_TRACE(bh
, "marking dirty");
1623 mark_buffer_dirty(bh
);
1624 sync_dirty_buffer(bh
);
1630 * int jbd2_journal_flush () - Flush journal
1631 * @journal: Journal to act on.
1633 * Flush all data for a given journal to disk and empty the journal.
1634 * Filesystems can use this when remounting readonly to ensure that
1635 * recovery does not need to happen on remount.
1638 int jbd2_journal_flush(journal_t
*journal
)
1641 transaction_t
*transaction
= NULL
;
1642 unsigned long old_tail
;
1644 spin_lock(&journal
->j_state_lock
);
1646 /* Force everything buffered to the log... */
1647 if (journal
->j_running_transaction
) {
1648 transaction
= journal
->j_running_transaction
;
1649 __jbd2_log_start_commit(journal
, transaction
->t_tid
);
1650 } else if (journal
->j_committing_transaction
)
1651 transaction
= journal
->j_committing_transaction
;
1653 /* Wait for the log commit to complete... */
1655 tid_t tid
= transaction
->t_tid
;
1657 spin_unlock(&journal
->j_state_lock
);
1658 jbd2_log_wait_commit(journal
, tid
);
1660 spin_unlock(&journal
->j_state_lock
);
1663 /* ...and flush everything in the log out to disk. */
1664 spin_lock(&journal
->j_list_lock
);
1665 while (!err
&& journal
->j_checkpoint_transactions
!= NULL
) {
1666 spin_unlock(&journal
->j_list_lock
);
1667 mutex_lock(&journal
->j_checkpoint_mutex
);
1668 err
= jbd2_log_do_checkpoint(journal
);
1669 mutex_unlock(&journal
->j_checkpoint_mutex
);
1670 spin_lock(&journal
->j_list_lock
);
1672 spin_unlock(&journal
->j_list_lock
);
1674 if (is_journal_aborted(journal
))
1677 jbd2_cleanup_journal_tail(journal
);
1679 /* Finally, mark the journal as really needing no recovery.
1680 * This sets s_start==0 in the underlying superblock, which is
1681 * the magic code for a fully-recovered superblock. Any future
1682 * commits of data to the journal will restore the current
1684 spin_lock(&journal
->j_state_lock
);
1685 old_tail
= journal
->j_tail
;
1686 journal
->j_tail
= 0;
1687 spin_unlock(&journal
->j_state_lock
);
1688 jbd2_journal_update_superblock(journal
, 1);
1689 spin_lock(&journal
->j_state_lock
);
1690 journal
->j_tail
= old_tail
;
1692 J_ASSERT(!journal
->j_running_transaction
);
1693 J_ASSERT(!journal
->j_committing_transaction
);
1694 J_ASSERT(!journal
->j_checkpoint_transactions
);
1695 J_ASSERT(journal
->j_head
== journal
->j_tail
);
1696 J_ASSERT(journal
->j_tail_sequence
== journal
->j_transaction_sequence
);
1697 spin_unlock(&journal
->j_state_lock
);
1702 * int jbd2_journal_wipe() - Wipe journal contents
1703 * @journal: Journal to act on.
1704 * @write: flag (see below)
1706 * Wipe out all of the contents of a journal, safely. This will produce
1707 * a warning if the journal contains any valid recovery information.
1708 * Must be called between journal_init_*() and jbd2_journal_load().
1710 * If 'write' is non-zero, then we wipe out the journal on disk; otherwise
1711 * we merely suppress recovery.
1714 int jbd2_journal_wipe(journal_t
*journal
, int write
)
1716 journal_superblock_t
*sb
;
1719 J_ASSERT (!(journal
->j_flags
& JBD2_LOADED
));
1721 err
= load_superblock(journal
);
1725 sb
= journal
->j_superblock
;
1727 if (!journal
->j_tail
)
1730 printk (KERN_WARNING
"JBD: %s recovery information on journal\n",
1731 write
? "Clearing" : "Ignoring");
1733 err
= jbd2_journal_skip_recovery(journal
);
1735 jbd2_journal_update_superblock(journal
, 1);
1742 * Journal abort has very specific semantics, which we describe
1743 * for journal abort.
1745 * Two internal function, which provide abort to te jbd layer
1750 * Quick version for internal journal use (doesn't lock the journal).
1751 * Aborts hard --- we mark the abort as occurred, but do _nothing_ else,
1752 * and don't attempt to make any other journal updates.
1754 void __jbd2_journal_abort_hard(journal_t
*journal
)
1756 transaction_t
*transaction
;
1758 if (journal
->j_flags
& JBD2_ABORT
)
1761 printk(KERN_ERR
"Aborting journal on device %s.\n",
1762 journal
->j_devname
);
1764 spin_lock(&journal
->j_state_lock
);
1765 journal
->j_flags
|= JBD2_ABORT
;
1766 transaction
= journal
->j_running_transaction
;
1768 __jbd2_log_start_commit(journal
, transaction
->t_tid
);
1769 spin_unlock(&journal
->j_state_lock
);
1772 /* Soft abort: record the abort error status in the journal superblock,
1773 * but don't do any other IO. */
1774 static void __journal_abort_soft (journal_t
*journal
, int errno
)
1776 if (journal
->j_flags
& JBD2_ABORT
)
1779 if (!journal
->j_errno
)
1780 journal
->j_errno
= errno
;
1782 __jbd2_journal_abort_hard(journal
);
1785 jbd2_journal_update_superblock(journal
, 1);
1789 * void jbd2_journal_abort () - Shutdown the journal immediately.
1790 * @journal: the journal to shutdown.
1791 * @errno: an error number to record in the journal indicating
1792 * the reason for the shutdown.
1794 * Perform a complete, immediate shutdown of the ENTIRE
1795 * journal (not of a single transaction). This operation cannot be
1796 * undone without closing and reopening the journal.
1798 * The jbd2_journal_abort function is intended to support higher level error
1799 * recovery mechanisms such as the ext2/ext3 remount-readonly error
1802 * Journal abort has very specific semantics. Any existing dirty,
1803 * unjournaled buffers in the main filesystem will still be written to
1804 * disk by bdflush, but the journaling mechanism will be suspended
1805 * immediately and no further transaction commits will be honoured.
1807 * Any dirty, journaled buffers will be written back to disk without
1808 * hitting the journal. Atomicity cannot be guaranteed on an aborted
1809 * filesystem, but we _do_ attempt to leave as much data as possible
1810 * behind for fsck to use for cleanup.
1812 * Any attempt to get a new transaction handle on a journal which is in
1813 * ABORT state will just result in an -EROFS error return. A
1814 * jbd2_journal_stop on an existing handle will return -EIO if we have
1815 * entered abort state during the update.
1817 * Recursive transactions are not disturbed by journal abort until the
1818 * final jbd2_journal_stop, which will receive the -EIO error.
1820 * Finally, the jbd2_journal_abort call allows the caller to supply an errno
1821 * which will be recorded (if possible) in the journal superblock. This
1822 * allows a client to record failure conditions in the middle of a
1823 * transaction without having to complete the transaction to record the
1824 * failure to disk. ext3_error, for example, now uses this
1827 * Errors which originate from within the journaling layer will NOT
1828 * supply an errno; a null errno implies that absolutely no further
1829 * writes are done to the journal (unless there are any already in
1834 void jbd2_journal_abort(journal_t
*journal
, int errno
)
1836 __journal_abort_soft(journal
, errno
);
1840 * int jbd2_journal_errno () - returns the journal's error state.
1841 * @journal: journal to examine.
1843 * This is the errno numbet set with jbd2_journal_abort(), the last
1844 * time the journal was mounted - if the journal was stopped
1845 * without calling abort this will be 0.
1847 * If the journal has been aborted on this mount time -EROFS will
1850 int jbd2_journal_errno(journal_t
*journal
)
1854 spin_lock(&journal
->j_state_lock
);
1855 if (journal
->j_flags
& JBD2_ABORT
)
1858 err
= journal
->j_errno
;
1859 spin_unlock(&journal
->j_state_lock
);
1864 * int jbd2_journal_clear_err () - clears the journal's error state
1865 * @journal: journal to act on.
1867 * An error must be cleared or Acked to take a FS out of readonly
1870 int jbd2_journal_clear_err(journal_t
*journal
)
1874 spin_lock(&journal
->j_state_lock
);
1875 if (journal
->j_flags
& JBD2_ABORT
)
1878 journal
->j_errno
= 0;
1879 spin_unlock(&journal
->j_state_lock
);
1884 * void jbd2_journal_ack_err() - Ack journal err.
1885 * @journal: journal to act on.
1887 * An error must be cleared or Acked to take a FS out of readonly
1890 void jbd2_journal_ack_err(journal_t
*journal
)
1892 spin_lock(&journal
->j_state_lock
);
1893 if (journal
->j_errno
)
1894 journal
->j_flags
|= JBD2_ACK_ERR
;
1895 spin_unlock(&journal
->j_state_lock
);
1898 int jbd2_journal_blocks_per_page(struct inode
*inode
)
1900 return 1 << (PAGE_CACHE_SHIFT
- inode
->i_sb
->s_blocksize_bits
);
1904 * helper functions to deal with 32 or 64bit block numbers.
1906 size_t journal_tag_bytes(journal_t
*journal
)
1908 if (JBD2_HAS_INCOMPAT_FEATURE(journal
, JBD2_FEATURE_INCOMPAT_64BIT
))
1909 return JBD2_TAG_SIZE64
;
1911 return JBD2_TAG_SIZE32
;
1915 * Journal_head storage management
1917 static struct kmem_cache
*jbd2_journal_head_cache
;
1918 #ifdef CONFIG_JBD2_DEBUG
1919 static atomic_t nr_journal_heads
= ATOMIC_INIT(0);
1922 static int journal_init_jbd2_journal_head_cache(void)
1926 J_ASSERT(jbd2_journal_head_cache
== NULL
);
1927 jbd2_journal_head_cache
= kmem_cache_create("jbd2_journal_head",
1928 sizeof(struct journal_head
),
1930 SLAB_TEMPORARY
, /* flags */
1933 if (!jbd2_journal_head_cache
) {
1935 printk(KERN_EMERG
"JBD: no memory for journal_head cache\n");
1940 static void jbd2_journal_destroy_jbd2_journal_head_cache(void)
1942 if (jbd2_journal_head_cache
) {
1943 kmem_cache_destroy(jbd2_journal_head_cache
);
1944 jbd2_journal_head_cache
= NULL
;
1949 * journal_head splicing and dicing
1951 static struct journal_head
*journal_alloc_journal_head(void)
1953 struct journal_head
*ret
;
1954 static unsigned long last_warning
;
1956 #ifdef CONFIG_JBD2_DEBUG
1957 atomic_inc(&nr_journal_heads
);
1959 ret
= kmem_cache_alloc(jbd2_journal_head_cache
, GFP_NOFS
);
1961 jbd_debug(1, "out of memory for journal_head\n");
1962 if (time_after(jiffies
, last_warning
+ 5*HZ
)) {
1963 printk(KERN_NOTICE
"ENOMEM in %s, retrying.\n",
1965 last_warning
= jiffies
;
1969 ret
= kmem_cache_alloc(jbd2_journal_head_cache
, GFP_NOFS
);
1975 static void journal_free_journal_head(struct journal_head
*jh
)
1977 #ifdef CONFIG_JBD2_DEBUG
1978 atomic_dec(&nr_journal_heads
);
1979 memset(jh
, JBD2_POISON_FREE
, sizeof(*jh
));
1981 kmem_cache_free(jbd2_journal_head_cache
, jh
);
1985 * A journal_head is attached to a buffer_head whenever JBD has an
1986 * interest in the buffer.
1988 * Whenever a buffer has an attached journal_head, its ->b_state:BH_JBD bit
1989 * is set. This bit is tested in core kernel code where we need to take
1990 * JBD-specific actions. Testing the zeroness of ->b_private is not reliable
1993 * When a buffer has its BH_JBD bit set, its ->b_count is elevated by one.
1995 * When a buffer has its BH_JBD bit set it is immune from being released by
1996 * core kernel code, mainly via ->b_count.
1998 * A journal_head may be detached from its buffer_head when the journal_head's
1999 * b_transaction, b_cp_transaction and b_next_transaction pointers are NULL.
2000 * Various places in JBD call jbd2_journal_remove_journal_head() to indicate that the
2001 * journal_head can be dropped if needed.
2003 * Various places in the kernel want to attach a journal_head to a buffer_head
2004 * _before_ attaching the journal_head to a transaction. To protect the
2005 * journal_head in this situation, jbd2_journal_add_journal_head elevates the
2006 * journal_head's b_jcount refcount by one. The caller must call
2007 * jbd2_journal_put_journal_head() to undo this.
2009 * So the typical usage would be:
2011 * (Attach a journal_head if needed. Increments b_jcount)
2012 * struct journal_head *jh = jbd2_journal_add_journal_head(bh);
2014 * jh->b_transaction = xxx;
2015 * jbd2_journal_put_journal_head(jh);
2017 * Now, the journal_head's b_jcount is zero, but it is safe from being released
2018 * because it has a non-zero b_transaction.
2022 * Give a buffer_head a journal_head.
2024 * Doesn't need the journal lock.
2027 struct journal_head
*jbd2_journal_add_journal_head(struct buffer_head
*bh
)
2029 struct journal_head
*jh
;
2030 struct journal_head
*new_jh
= NULL
;
2033 if (!buffer_jbd(bh
)) {
2034 new_jh
= journal_alloc_journal_head();
2035 memset(new_jh
, 0, sizeof(*new_jh
));
2038 jbd_lock_bh_journal_head(bh
);
2039 if (buffer_jbd(bh
)) {
2043 (atomic_read(&bh
->b_count
) > 0) ||
2044 (bh
->b_page
&& bh
->b_page
->mapping
));
2047 jbd_unlock_bh_journal_head(bh
);
2052 new_jh
= NULL
; /* We consumed it */
2057 BUFFER_TRACE(bh
, "added journal_head");
2060 jbd_unlock_bh_journal_head(bh
);
2062 journal_free_journal_head(new_jh
);
2063 return bh
->b_private
;
2067 * Grab a ref against this buffer_head's journal_head. If it ended up not
2068 * having a journal_head, return NULL
2070 struct journal_head
*jbd2_journal_grab_journal_head(struct buffer_head
*bh
)
2072 struct journal_head
*jh
= NULL
;
2074 jbd_lock_bh_journal_head(bh
);
2075 if (buffer_jbd(bh
)) {
2079 jbd_unlock_bh_journal_head(bh
);
2083 static void __journal_remove_journal_head(struct buffer_head
*bh
)
2085 struct journal_head
*jh
= bh2jh(bh
);
2087 J_ASSERT_JH(jh
, jh
->b_jcount
>= 0);
2090 if (jh
->b_jcount
== 0) {
2091 if (jh
->b_transaction
== NULL
&&
2092 jh
->b_next_transaction
== NULL
&&
2093 jh
->b_cp_transaction
== NULL
) {
2094 J_ASSERT_JH(jh
, jh
->b_jlist
== BJ_None
);
2095 J_ASSERT_BH(bh
, buffer_jbd(bh
));
2096 J_ASSERT_BH(bh
, jh2bh(jh
) == bh
);
2097 BUFFER_TRACE(bh
, "remove journal_head");
2098 if (jh
->b_frozen_data
) {
2099 printk(KERN_WARNING
"%s: freeing "
2102 jbd2_free(jh
->b_frozen_data
, bh
->b_size
);
2104 if (jh
->b_committed_data
) {
2105 printk(KERN_WARNING
"%s: freeing "
2106 "b_committed_data\n",
2108 jbd2_free(jh
->b_committed_data
, bh
->b_size
);
2110 bh
->b_private
= NULL
;
2111 jh
->b_bh
= NULL
; /* debug, really */
2112 clear_buffer_jbd(bh
);
2114 journal_free_journal_head(jh
);
2116 BUFFER_TRACE(bh
, "journal_head was locked");
2122 * jbd2_journal_remove_journal_head(): if the buffer isn't attached to a transaction
2123 * and has a zero b_jcount then remove and release its journal_head. If we did
2124 * see that the buffer is not used by any transaction we also "logically"
2125 * decrement ->b_count.
2127 * We in fact take an additional increment on ->b_count as a convenience,
2128 * because the caller usually wants to do additional things with the bh
2129 * after calling here.
2130 * The caller of jbd2_journal_remove_journal_head() *must* run __brelse(bh) at some
2131 * time. Once the caller has run __brelse(), the buffer is eligible for
2132 * reaping by try_to_free_buffers().
2134 void jbd2_journal_remove_journal_head(struct buffer_head
*bh
)
2136 jbd_lock_bh_journal_head(bh
);
2137 __journal_remove_journal_head(bh
);
2138 jbd_unlock_bh_journal_head(bh
);
2142 * Drop a reference on the passed journal_head. If it fell to zero then try to
2143 * release the journal_head from the buffer_head.
2145 void jbd2_journal_put_journal_head(struct journal_head
*jh
)
2147 struct buffer_head
*bh
= jh2bh(jh
);
2149 jbd_lock_bh_journal_head(bh
);
2150 J_ASSERT_JH(jh
, jh
->b_jcount
> 0);
2152 if (!jh
->b_jcount
&& !jh
->b_transaction
) {
2153 __journal_remove_journal_head(bh
);
2156 jbd_unlock_bh_journal_head(bh
);
2160 * Initialize jbd inode head
2162 void jbd2_journal_init_jbd_inode(struct jbd2_inode
*jinode
, struct inode
*inode
)
2164 jinode
->i_transaction
= NULL
;
2165 jinode
->i_next_transaction
= NULL
;
2166 jinode
->i_vfs_inode
= inode
;
2167 jinode
->i_flags
= 0;
2168 INIT_LIST_HEAD(&jinode
->i_list
);
2172 * Function to be called before we start removing inode from memory (i.e.,
2173 * clear_inode() is a fine place to be called from). It removes inode from
2174 * transaction's lists.
2176 void jbd2_journal_release_jbd_inode(journal_t
*journal
,
2177 struct jbd2_inode
*jinode
)
2184 spin_lock(&journal
->j_list_lock
);
2185 /* Is commit writing out inode - we have to wait */
2186 if (jinode
->i_flags
& JI_COMMIT_RUNNING
) {
2187 wait_queue_head_t
*wq
;
2188 DEFINE_WAIT_BIT(wait
, &jinode
->i_flags
, __JI_COMMIT_RUNNING
);
2189 wq
= bit_waitqueue(&jinode
->i_flags
, __JI_COMMIT_RUNNING
);
2190 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
2191 spin_unlock(&journal
->j_list_lock
);
2193 finish_wait(wq
, &wait
.wait
);
2197 /* Do we need to wait for data writeback? */
2198 if (journal
->j_committing_transaction
== jinode
->i_transaction
)
2200 if (jinode
->i_transaction
) {
2201 list_del(&jinode
->i_list
);
2202 jinode
->i_transaction
= NULL
;
2204 spin_unlock(&journal
->j_list_lock
);
2210 #ifdef CONFIG_JBD2_DEBUG
2211 u8 jbd2_journal_enable_debug __read_mostly
;
2212 EXPORT_SYMBOL(jbd2_journal_enable_debug
);
2214 #define JBD2_DEBUG_NAME "jbd2-debug"
2216 static struct dentry
*jbd2_debugfs_dir
;
2217 static struct dentry
*jbd2_debug
;
2219 static void __init
jbd2_create_debugfs_entry(void)
2221 jbd2_debugfs_dir
= debugfs_create_dir("jbd2", NULL
);
2222 if (jbd2_debugfs_dir
)
2223 jbd2_debug
= debugfs_create_u8(JBD2_DEBUG_NAME
, S_IRUGO
,
2225 &jbd2_journal_enable_debug
);
2228 static void __exit
jbd2_remove_debugfs_entry(void)
2230 debugfs_remove(jbd2_debug
);
2231 debugfs_remove(jbd2_debugfs_dir
);
2236 static void __init
jbd2_create_debugfs_entry(void)
2240 static void __exit
jbd2_remove_debugfs_entry(void)
2246 #ifdef CONFIG_PROC_FS
2248 #define JBD2_STATS_PROC_NAME "fs/jbd2"
2250 static void __init
jbd2_create_jbd_stats_proc_entry(void)
2252 proc_jbd2_stats
= proc_mkdir(JBD2_STATS_PROC_NAME
, NULL
);
2255 static void __exit
jbd2_remove_jbd_stats_proc_entry(void)
2257 if (proc_jbd2_stats
)
2258 remove_proc_entry(JBD2_STATS_PROC_NAME
, NULL
);
2263 #define jbd2_create_jbd_stats_proc_entry() do {} while (0)
2264 #define jbd2_remove_jbd_stats_proc_entry() do {} while (0)
2268 struct kmem_cache
*jbd2_handle_cache
;
2270 static int __init
journal_init_handle_cache(void)
2272 jbd2_handle_cache
= kmem_cache_create("jbd2_journal_handle",
2275 SLAB_TEMPORARY
, /* flags */
2277 if (jbd2_handle_cache
== NULL
) {
2278 printk(KERN_EMERG
"JBD: failed to create handle cache\n");
2284 static void jbd2_journal_destroy_handle_cache(void)
2286 if (jbd2_handle_cache
)
2287 kmem_cache_destroy(jbd2_handle_cache
);
2291 * Module startup and shutdown
2294 static int __init
journal_init_caches(void)
2298 ret
= jbd2_journal_init_revoke_caches();
2300 ret
= journal_init_jbd2_journal_head_cache();
2302 ret
= journal_init_handle_cache();
2306 static void jbd2_journal_destroy_caches(void)
2308 jbd2_journal_destroy_revoke_caches();
2309 jbd2_journal_destroy_jbd2_journal_head_cache();
2310 jbd2_journal_destroy_handle_cache();
2313 static int __init
journal_init(void)
2317 BUILD_BUG_ON(sizeof(struct journal_superblock_s
) != 1024);
2319 ret
= journal_init_caches();
2321 jbd2_create_debugfs_entry();
2322 jbd2_create_jbd_stats_proc_entry();
2324 jbd2_journal_destroy_caches();
2329 static void __exit
journal_exit(void)
2331 #ifdef CONFIG_JBD2_DEBUG
2332 int n
= atomic_read(&nr_journal_heads
);
2334 printk(KERN_EMERG
"JBD: leaked %d journal_heads!\n", n
);
2336 jbd2_remove_debugfs_entry();
2337 jbd2_remove_jbd_stats_proc_entry();
2338 jbd2_journal_destroy_caches();
2341 MODULE_LICENSE("GPL");
2342 module_init(journal_init
);
2343 module_exit(journal_exit
);