2006-10-11 08:20:57 +00:00
|
|
|
/*
|
2007-05-09 05:51:49 +00:00
|
|
|
* linux/fs/jbd2/checkpoint.c
|
2006-10-11 08:20:57 +00:00
|
|
|
*
|
|
|
|
* Written by Stephen C. Tweedie <sct@redhat.com>, 1999
|
|
|
|
*
|
|
|
|
* Copyright 1999 Red Hat Software --- All Rights Reserved
|
|
|
|
*
|
|
|
|
* This file is part of the Linux kernel and is made available under
|
|
|
|
* the terms of the GNU General Public License, version 2, or at your
|
|
|
|
* option, any later version, incorporated herein by reference.
|
|
|
|
*
|
|
|
|
* Checkpoint routines for the generic filesystem journaling code.
|
|
|
|
* Part of the ext2fs journaling system.
|
|
|
|
*
|
|
|
|
* Checkpointing is the process of ensuring that a section of the log is
|
|
|
|
* committed fully to disk, so that that portion of the log can be
|
|
|
|
* reused.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/time.h>
|
|
|
|
#include <linux/fs.h>
|
2006-10-11 08:20:59 +00:00
|
|
|
#include <linux/jbd2.h>
|
2006-10-11 08:20:57 +00:00
|
|
|
#include <linux/errno.h>
|
|
|
|
#include <linux/slab.h>
|
2009-12-23 11:52:08 +00:00
|
|
|
#include <linux/blkdev.h>
|
2009-06-17 15:47:48 +00:00
|
|
|
#include <trace/events/jbd2.h>
|
2006-10-11 08:20:57 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Unlink a buffer from a transaction checkpoint list.
|
|
|
|
*
|
|
|
|
* Called with j_list_lock held.
|
|
|
|
*/
|
|
|
|
static inline void __buffer_unlink_first(struct journal_head *jh)
|
|
|
|
{
|
|
|
|
transaction_t *transaction = jh->b_cp_transaction;
|
|
|
|
|
|
|
|
jh->b_cpnext->b_cpprev = jh->b_cpprev;
|
|
|
|
jh->b_cpprev->b_cpnext = jh->b_cpnext;
|
|
|
|
if (transaction->t_checkpoint_list == jh) {
|
|
|
|
transaction->t_checkpoint_list = jh->b_cpnext;
|
|
|
|
if (transaction->t_checkpoint_list == jh)
|
|
|
|
transaction->t_checkpoint_list = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Unlink a buffer from a transaction checkpoint(io) list.
|
|
|
|
*
|
|
|
|
* Called with j_list_lock held.
|
|
|
|
*/
|
|
|
|
static inline void __buffer_unlink(struct journal_head *jh)
|
|
|
|
{
|
|
|
|
transaction_t *transaction = jh->b_cp_transaction;
|
|
|
|
|
|
|
|
__buffer_unlink_first(jh);
|
|
|
|
if (transaction->t_checkpoint_io_list == jh) {
|
|
|
|
transaction->t_checkpoint_io_list = jh->b_cpnext;
|
|
|
|
if (transaction->t_checkpoint_io_list == jh)
|
|
|
|
transaction->t_checkpoint_io_list = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Move a buffer from the checkpoint list to the checkpoint io list
|
|
|
|
*
|
|
|
|
* Called with j_list_lock held
|
|
|
|
*/
|
|
|
|
static inline void __buffer_relink_io(struct journal_head *jh)
|
|
|
|
{
|
|
|
|
transaction_t *transaction = jh->b_cp_transaction;
|
|
|
|
|
|
|
|
__buffer_unlink_first(jh);
|
|
|
|
|
|
|
|
if (!transaction->t_checkpoint_io_list) {
|
|
|
|
jh->b_cpnext = jh->b_cpprev = jh;
|
|
|
|
} else {
|
|
|
|
jh->b_cpnext = transaction->t_checkpoint_io_list;
|
|
|
|
jh->b_cpprev = transaction->t_checkpoint_io_list->b_cpprev;
|
|
|
|
jh->b_cpprev->b_cpnext = jh;
|
|
|
|
jh->b_cpnext->b_cpprev = jh;
|
|
|
|
}
|
|
|
|
transaction->t_checkpoint_io_list = jh;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to release a checkpointed buffer from its transaction.
|
|
|
|
* Returns 1 if we released it and 2 if we also released the
|
|
|
|
* whole transaction.
|
|
|
|
*
|
|
|
|
* Requires j_list_lock
|
|
|
|
*/
|
|
|
|
static int __try_to_free_cp_buf(struct journal_head *jh)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
struct buffer_head *bh = jh2bh(jh);
|
|
|
|
|
2012-03-14 02:45:25 +00:00
|
|
|
if (jh->b_transaction == NULL && !buffer_locked(bh) &&
|
2008-10-11 00:29:13 +00:00
|
|
|
!buffer_dirty(bh) && !buffer_write_io_error(bh)) {
|
2006-10-11 08:20:57 +00:00
|
|
|
JBUFFER_TRACE(jh, "remove from checkpoint list");
|
2006-10-11 08:20:59 +00:00
|
|
|
ret = __jbd2_journal_remove_checkpoint(jh) + 1;
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-10-11 08:20:59 +00:00
|
|
|
* __jbd2_log_wait_for_space: wait until there is space in the journal.
|
2006-10-11 08:20:57 +00:00
|
|
|
*
|
|
|
|
* Called under j-state_lock *only*. It will be unlocked if we have to wait
|
|
|
|
* for a checkpoint to free up some space in the log.
|
|
|
|
*/
|
2006-10-11 08:20:59 +00:00
|
|
|
void __jbd2_log_wait_for_space(journal_t *journal)
|
2006-10-11 08:20:57 +00:00
|
|
|
{
|
2008-11-07 03:38:07 +00:00
|
|
|
int nblocks, space_left;
|
2010-08-04 01:35:12 +00:00
|
|
|
/* assert_spin_locked(&journal->j_state_lock); */
|
2006-10-11 08:20:57 +00:00
|
|
|
|
2013-06-04 16:12:57 +00:00
|
|
|
nblocks = jbd2_space_needed(journal);
|
|
|
|
while (jbd2_log_space_left(journal) < nblocks) {
|
2010-08-04 01:35:12 +00:00
|
|
|
write_unlock(&journal->j_state_lock);
|
2006-10-11 08:20:57 +00:00
|
|
|
mutex_lock(&journal->j_checkpoint_mutex);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Test again, another process may have checkpointed while we
|
2008-10-09 03:28:31 +00:00
|
|
|
* were waiting for the checkpoint lock. If there are no
|
2008-11-07 03:38:07 +00:00
|
|
|
* transactions ready to be checkpointed, try to recover
|
|
|
|
* journal space by calling cleanup_journal_tail(), and if
|
|
|
|
* that doesn't work, by waiting for the currently committing
|
|
|
|
* transaction to complete. If there is absolutely no way
|
|
|
|
* to make progress, this is either a BUG or corrupted
|
|
|
|
* filesystem, so abort the journal and leave a stack
|
|
|
|
* trace for forensic evidence.
|
2006-10-11 08:20:57 +00:00
|
|
|
*/
|
2010-08-04 01:35:12 +00:00
|
|
|
write_lock(&journal->j_state_lock);
|
2014-09-16 18:50:50 +00:00
|
|
|
if (journal->j_flags & JBD2_ABORT) {
|
|
|
|
mutex_unlock(&journal->j_checkpoint_mutex);
|
|
|
|
return;
|
|
|
|
}
|
2008-10-09 03:28:31 +00:00
|
|
|
spin_lock(&journal->j_list_lock);
|
2013-06-04 16:12:57 +00:00
|
|
|
nblocks = jbd2_space_needed(journal);
|
|
|
|
space_left = jbd2_log_space_left(journal);
|
2008-11-07 03:38:07 +00:00
|
|
|
if (space_left < nblocks) {
|
2008-10-09 03:28:31 +00:00
|
|
|
int chkpt = journal->j_checkpoint_transactions != NULL;
|
2008-11-07 03:38:07 +00:00
|
|
|
tid_t tid = 0;
|
2008-10-09 03:28:31 +00:00
|
|
|
|
2008-11-07 03:38:07 +00:00
|
|
|
if (journal->j_committing_transaction)
|
|
|
|
tid = journal->j_committing_transaction->t_tid;
|
2008-10-09 03:28:31 +00:00
|
|
|
spin_unlock(&journal->j_list_lock);
|
2010-08-04 01:35:12 +00:00
|
|
|
write_unlock(&journal->j_state_lock);
|
2008-10-09 03:28:31 +00:00
|
|
|
if (chkpt) {
|
|
|
|
jbd2_log_do_checkpoint(journal);
|
2008-11-07 03:38:07 +00:00
|
|
|
} else if (jbd2_cleanup_journal_tail(journal) == 0) {
|
|
|
|
/* We were able to recover space; yay! */
|
|
|
|
;
|
|
|
|
} else if (tid) {
|
jbd2: drop checkpoint mutex when waiting in __jbd2_log_wait_for_space()
While trying to debug an an issue under extreme I/O loading
on preempt-rt kernels, the following backtrace was observed
via SysRQ output:
rm D ffff8802203afbc0 4600 4878 4748 0x00000000
ffff8802217bfb78 0000000000000082 ffff88021fc2bb80 ffff88021fc2bb80
ffff88021fc2bb80 ffff8802217bffd8 ffff8802217bffd8 ffff8802217bffd8
ffff88021f1d4c80 ffff88021fc2bb80 ffff8802217bfb88 ffff88022437b000
Call Trace:
[<ffffffff8172dc34>] schedule+0x24/0x70
[<ffffffff81225b5d>] jbd2_log_wait_commit+0xbd/0x140
[<ffffffff81060390>] ? __init_waitqueue_head+0x50/0x50
[<ffffffff81223635>] jbd2_log_do_checkpoint+0xf5/0x520
[<ffffffff81223b09>] __jbd2_log_wait_for_space+0xa9/0x1f0
[<ffffffff8121dc40>] start_this_handle.isra.10+0x2e0/0x530
[<ffffffff81060390>] ? __init_waitqueue_head+0x50/0x50
[<ffffffff8121e0a3>] jbd2__journal_start+0xc3/0x110
[<ffffffff811de7ce>] ? ext4_rmdir+0x6e/0x230
[<ffffffff8121e0fe>] jbd2_journal_start+0xe/0x10
[<ffffffff811f308b>] ext4_journal_start_sb+0x5b/0x160
[<ffffffff811de7ce>] ext4_rmdir+0x6e/0x230
[<ffffffff811435c5>] vfs_rmdir+0xd5/0x140
[<ffffffff8114370f>] do_rmdir+0xdf/0x120
[<ffffffff8105c6b4>] ? task_work_run+0x44/0x80
[<ffffffff81002889>] ? do_notify_resume+0x89/0x100
[<ffffffff817361ae>] ? int_signal+0x12/0x17
[<ffffffff81145d85>] sys_unlinkat+0x25/0x40
[<ffffffff81735f22>] system_call_fastpath+0x16/0x1b
What is interesting here, is that we call log_wait_commit, from
within wait_for_space, but we are still holding the checkpoint_mutex
as it surrounds mostly the whole of wait_for_space. And then, as we
are waiting, journal_commit_transaction can run, and if the JBD2_FLUSHED
bit is set, then we will also try to take the same checkpoint_mutex.
It seems that we need to drop the checkpoint_mutex while sitting in
jbd2_log_wait_commit, if we want to guarantee that progress can be made
by jbd2_journal_commit_transaction(). There does not seem to be
anything preempt-rt specific about this, other then perhaps increasing
the odds of it happening.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-06-13 02:47:35 +00:00
|
|
|
/*
|
|
|
|
* jbd2_journal_commit_transaction() may want
|
|
|
|
* to take the checkpoint_mutex if JBD2_FLUSHED
|
|
|
|
* is set. So we need to temporarily drop it.
|
|
|
|
*/
|
|
|
|
mutex_unlock(&journal->j_checkpoint_mutex);
|
2008-11-07 03:38:07 +00:00
|
|
|
jbd2_log_wait_commit(journal, tid);
|
jbd2: drop checkpoint mutex when waiting in __jbd2_log_wait_for_space()
While trying to debug an an issue under extreme I/O loading
on preempt-rt kernels, the following backtrace was observed
via SysRQ output:
rm D ffff8802203afbc0 4600 4878 4748 0x00000000
ffff8802217bfb78 0000000000000082 ffff88021fc2bb80 ffff88021fc2bb80
ffff88021fc2bb80 ffff8802217bffd8 ffff8802217bffd8 ffff8802217bffd8
ffff88021f1d4c80 ffff88021fc2bb80 ffff8802217bfb88 ffff88022437b000
Call Trace:
[<ffffffff8172dc34>] schedule+0x24/0x70
[<ffffffff81225b5d>] jbd2_log_wait_commit+0xbd/0x140
[<ffffffff81060390>] ? __init_waitqueue_head+0x50/0x50
[<ffffffff81223635>] jbd2_log_do_checkpoint+0xf5/0x520
[<ffffffff81223b09>] __jbd2_log_wait_for_space+0xa9/0x1f0
[<ffffffff8121dc40>] start_this_handle.isra.10+0x2e0/0x530
[<ffffffff81060390>] ? __init_waitqueue_head+0x50/0x50
[<ffffffff8121e0a3>] jbd2__journal_start+0xc3/0x110
[<ffffffff811de7ce>] ? ext4_rmdir+0x6e/0x230
[<ffffffff8121e0fe>] jbd2_journal_start+0xe/0x10
[<ffffffff811f308b>] ext4_journal_start_sb+0x5b/0x160
[<ffffffff811de7ce>] ext4_rmdir+0x6e/0x230
[<ffffffff811435c5>] vfs_rmdir+0xd5/0x140
[<ffffffff8114370f>] do_rmdir+0xdf/0x120
[<ffffffff8105c6b4>] ? task_work_run+0x44/0x80
[<ffffffff81002889>] ? do_notify_resume+0x89/0x100
[<ffffffff817361ae>] ? int_signal+0x12/0x17
[<ffffffff81145d85>] sys_unlinkat+0x25/0x40
[<ffffffff81735f22>] system_call_fastpath+0x16/0x1b
What is interesting here, is that we call log_wait_commit, from
within wait_for_space, but we are still holding the checkpoint_mutex
as it surrounds mostly the whole of wait_for_space. And then, as we
are waiting, journal_commit_transaction can run, and if the JBD2_FLUSHED
bit is set, then we will also try to take the same checkpoint_mutex.
It seems that we need to drop the checkpoint_mutex while sitting in
jbd2_log_wait_commit, if we want to guarantee that progress can be made
by jbd2_journal_commit_transaction(). There does not seem to be
anything preempt-rt specific about this, other then perhaps increasing
the odds of it happening.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-06-13 02:47:35 +00:00
|
|
|
write_lock(&journal->j_state_lock);
|
|
|
|
continue;
|
2008-10-09 03:28:31 +00:00
|
|
|
} else {
|
2008-11-07 03:38:07 +00:00
|
|
|
printk(KERN_ERR "%s: needed %d blocks and "
|
|
|
|
"only had %d space available\n",
|
|
|
|
__func__, nblocks, space_left);
|
|
|
|
printk(KERN_ERR "%s: no way to get more "
|
|
|
|
"journal space in %s\n", __func__,
|
|
|
|
journal->j_devname);
|
|
|
|
WARN_ON(1);
|
2008-10-09 03:28:31 +00:00
|
|
|
jbd2_journal_abort(journal, 0);
|
|
|
|
}
|
2010-08-04 01:35:12 +00:00
|
|
|
write_lock(&journal->j_state_lock);
|
2008-10-09 03:28:31 +00:00
|
|
|
} else {
|
|
|
|
spin_unlock(&journal->j_list_lock);
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
|
|
|
mutex_unlock(&journal->j_checkpoint_mutex);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2008-11-05 05:09:22 +00:00
|
|
|
__flush_batch(journal_t *journal, int *batch_count)
|
2006-10-11 08:20:57 +00:00
|
|
|
{
|
|
|
|
int i;
|
2011-06-27 16:36:29 +00:00
|
|
|
struct blk_plug plug;
|
2006-10-11 08:20:57 +00:00
|
|
|
|
2011-06-27 16:36:29 +00:00
|
|
|
blk_start_plug(&plug);
|
2010-08-11 15:06:24 +00:00
|
|
|
for (i = 0; i < *batch_count; i++)
|
2011-06-27 16:36:29 +00:00
|
|
|
write_dirty_buffer(journal->j_chkpt_bhs[i], WRITE_SYNC);
|
|
|
|
blk_finish_plug(&plug);
|
2010-08-11 15:06:24 +00:00
|
|
|
|
2006-10-11 08:20:57 +00:00
|
|
|
for (i = 0; i < *batch_count; i++) {
|
2008-11-05 05:09:22 +00:00
|
|
|
struct buffer_head *bh = journal->j_chkpt_bhs[i];
|
2006-10-11 08:20:57 +00:00
|
|
|
BUFFER_TRACE(bh, "brelse");
|
|
|
|
__brelse(bh);
|
|
|
|
}
|
|
|
|
*batch_count = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Perform an actual checkpoint. We take the first transaction on the
|
|
|
|
* list of transactions to be checkpointed and send all its buffers
|
|
|
|
* to disk. We submit larger chunks of data at once.
|
|
|
|
*
|
|
|
|
* The journal should be locked before calling this function.
|
2008-10-11 00:29:13 +00:00
|
|
|
* Called with j_checkpoint_mutex held.
|
2006-10-11 08:20:57 +00:00
|
|
|
*/
|
2006-10-11 08:20:59 +00:00
|
|
|
int jbd2_log_do_checkpoint(journal_t *journal)
|
2006-10-11 08:20:57 +00:00
|
|
|
{
|
2014-09-02 01:19:01 +00:00
|
|
|
struct journal_head *jh;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
transaction_t *transaction;
|
|
|
|
tid_t this_tid;
|
2014-09-04 22:09:22 +00:00
|
|
|
int result, batch_count = 0;
|
2006-10-11 08:20:57 +00:00
|
|
|
|
|
|
|
jbd_debug(1, "Start checkpoint\n");
|
|
|
|
|
|
|
|
/*
|
|
|
|
* First thing: if there are any transactions in the log which
|
|
|
|
* don't need checkpointing, just eliminate them from the
|
|
|
|
* journal straight away.
|
|
|
|
*/
|
2006-10-11 08:20:59 +00:00
|
|
|
result = jbd2_cleanup_journal_tail(journal);
|
2009-06-17 15:47:48 +00:00
|
|
|
trace_jbd2_checkpoint(journal, result);
|
2006-10-11 08:20:57 +00:00
|
|
|
jbd_debug(1, "cleanup_journal_tail returned %d\n", result);
|
|
|
|
if (result <= 0)
|
|
|
|
return result;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* OK, we need to start writing disk blocks. Take one transaction
|
|
|
|
* and write it.
|
|
|
|
*/
|
2008-10-11 00:29:13 +00:00
|
|
|
result = 0;
|
2006-10-11 08:20:57 +00:00
|
|
|
spin_lock(&journal->j_list_lock);
|
|
|
|
if (!journal->j_checkpoint_transactions)
|
|
|
|
goto out;
|
|
|
|
transaction = journal->j_checkpoint_transactions;
|
2008-01-29 04:58:27 +00:00
|
|
|
if (transaction->t_chp_stats.cs_chp_time == 0)
|
|
|
|
transaction->t_chp_stats.cs_chp_time = jiffies;
|
2006-10-11 08:20:57 +00:00
|
|
|
this_tid = transaction->t_tid;
|
|
|
|
restart:
|
|
|
|
/*
|
|
|
|
* If someone cleaned up this transaction while we slept, we're
|
|
|
|
* done (maybe it's a new transaction, but it fell at the same
|
|
|
|
* address).
|
|
|
|
*/
|
2014-09-02 01:19:01 +00:00
|
|
|
if (journal->j_checkpoint_transactions != transaction ||
|
|
|
|
transaction->t_tid != this_tid)
|
|
|
|
goto out;
|
2006-10-11 08:20:57 +00:00
|
|
|
|
2014-09-02 01:19:01 +00:00
|
|
|
/* checkpoint all of the transaction's buffers */
|
|
|
|
while (transaction->t_checkpoint_list) {
|
|
|
|
jh = transaction->t_checkpoint_list;
|
|
|
|
bh = jh2bh(jh);
|
|
|
|
|
|
|
|
if (buffer_locked(bh)) {
|
|
|
|
spin_unlock(&journal->j_list_lock);
|
|
|
|
get_bh(bh);
|
|
|
|
wait_on_buffer(bh);
|
|
|
|
/* the journal_head may have gone by now */
|
|
|
|
BUFFER_TRACE(bh, "brelse");
|
|
|
|
__brelse(bh);
|
|
|
|
goto retry;
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
2014-09-02 01:19:01 +00:00
|
|
|
if (jh->b_transaction != NULL) {
|
|
|
|
transaction_t *t = jh->b_transaction;
|
|
|
|
tid_t tid = t->t_tid;
|
2006-10-11 08:20:57 +00:00
|
|
|
|
2014-09-02 01:19:01 +00:00
|
|
|
transaction->t_chp_stats.cs_forced_to_close++;
|
|
|
|
spin_unlock(&journal->j_list_lock);
|
|
|
|
if (unlikely(journal->j_flags & JBD2_UNMOUNT))
|
|
|
|
/*
|
|
|
|
* The journal thread is dead; so
|
|
|
|
* starting and waiting for a commit
|
|
|
|
* to finish will cause us to wait for
|
|
|
|
* a _very_ long time.
|
|
|
|
*/
|
|
|
|
printk(KERN_ERR
|
|
|
|
"JBD2: %s: Waiting for Godot: block %llu\n",
|
|
|
|
journal->j_devname, (unsigned long long) bh->b_blocknr);
|
|
|
|
|
|
|
|
jbd2_log_start_commit(journal, tid);
|
|
|
|
jbd2_log_wait_commit(journal, tid);
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
if (!buffer_dirty(bh)) {
|
|
|
|
if (unlikely(buffer_write_io_error(bh)) && !result)
|
|
|
|
result = -EIO;
|
|
|
|
BUFFER_TRACE(bh, "remove from checkpoint");
|
2014-09-04 22:09:29 +00:00
|
|
|
if (__jbd2_journal_remove_checkpoint(jh))
|
|
|
|
/* The transaction was released; we're done */
|
|
|
|
goto out;
|
|
|
|
continue;
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
|
|
|
/*
|
2014-09-02 01:19:01 +00:00
|
|
|
* Important: we are about to write the buffer, and
|
|
|
|
* possibly block, while still holding the journal
|
|
|
|
* lock. We cannot afford to let the transaction
|
|
|
|
* logic start messing around with this buffer before
|
|
|
|
* we write it to disk, as that would break
|
|
|
|
* recoverability.
|
2006-10-11 08:20:57 +00:00
|
|
|
*/
|
2014-09-02 01:19:01 +00:00
|
|
|
BUFFER_TRACE(bh, "queue");
|
|
|
|
get_bh(bh);
|
|
|
|
J_ASSERT_BH(bh, !buffer_jwrite(bh));
|
|
|
|
journal->j_chkpt_bhs[batch_count++] = bh;
|
|
|
|
__buffer_relink_io(jh);
|
|
|
|
transaction->t_chp_stats.cs_written++;
|
|
|
|
if ((batch_count == JBD2_NR_BATCH) ||
|
|
|
|
need_resched() ||
|
|
|
|
spin_needbreak(&journal->j_list_lock))
|
|
|
|
goto unlock_and_flush;
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
2014-09-02 01:19:01 +00:00
|
|
|
|
|
|
|
if (batch_count) {
|
|
|
|
unlock_and_flush:
|
|
|
|
spin_unlock(&journal->j_list_lock);
|
|
|
|
retry:
|
|
|
|
if (batch_count)
|
|
|
|
__flush_batch(journal, &batch_count);
|
|
|
|
spin_lock(&journal->j_list_lock);
|
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Now we issued all of the transaction's buffers, let's deal
|
|
|
|
* with the buffers that are out for I/O.
|
|
|
|
*/
|
2014-09-02 01:26:09 +00:00
|
|
|
restart2:
|
|
|
|
/* Did somebody clean up the transaction in the meanwhile? */
|
|
|
|
if (journal->j_checkpoint_transactions != transaction ||
|
|
|
|
transaction->t_tid != this_tid)
|
|
|
|
goto out;
|
|
|
|
|
2014-09-04 22:09:22 +00:00
|
|
|
while (transaction->t_checkpoint_io_list) {
|
2014-09-02 01:26:09 +00:00
|
|
|
jh = transaction->t_checkpoint_io_list;
|
|
|
|
bh = jh2bh(jh);
|
|
|
|
if (buffer_locked(bh)) {
|
|
|
|
spin_unlock(&journal->j_list_lock);
|
2014-09-04 22:09:22 +00:00
|
|
|
get_bh(bh);
|
2014-09-02 01:26:09 +00:00
|
|
|
wait_on_buffer(bh);
|
|
|
|
/* the journal_head may have gone by now */
|
|
|
|
BUFFER_TRACE(bh, "brelse");
|
|
|
|
__brelse(bh);
|
|
|
|
spin_lock(&journal->j_list_lock);
|
|
|
|
goto restart2;
|
|
|
|
}
|
|
|
|
if (unlikely(buffer_write_io_error(bh)) && !result)
|
|
|
|
result = -EIO;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Now in whatever state the buffer currently is, we
|
|
|
|
* know that it has been written out and so we can
|
|
|
|
* drop it from the list
|
|
|
|
*/
|
2014-09-04 22:09:22 +00:00
|
|
|
if (__jbd2_journal_remove_checkpoint(jh))
|
|
|
|
break;
|
2014-09-02 01:26:09 +00:00
|
|
|
}
|
2006-10-11 08:20:57 +00:00
|
|
|
out:
|
|
|
|
spin_unlock(&journal->j_list_lock);
|
|
|
|
if (result < 0)
|
2008-10-11 00:29:13 +00:00
|
|
|
jbd2_journal_abort(journal, result);
|
|
|
|
else
|
|
|
|
result = jbd2_cleanup_journal_tail(journal);
|
|
|
|
|
|
|
|
return (result < 0) ? result : 0;
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check the list of checkpoint transactions for the journal to see if
|
|
|
|
* we have already got rid of any since the last update of the log tail
|
|
|
|
* in the journal superblock. If so, we can instantly roll the
|
|
|
|
* superblock forward to remove those transactions from the log.
|
|
|
|
*
|
|
|
|
* Return <0 on error, 0 on success, 1 if there was nothing to clean up.
|
|
|
|
*
|
|
|
|
* Called with the journal lock held.
|
|
|
|
*
|
|
|
|
* This is the only part of the journaling code which really needs to be
|
|
|
|
* aware of transaction aborts. Checkpointing involves writing to the
|
|
|
|
* main filesystem area rather than to the journal, so it can proceed
|
2008-10-11 00:29:13 +00:00
|
|
|
* even in abort state, but we must not update the super block if
|
|
|
|
* checkpointing may have failed. Otherwise, we would lose some metadata
|
|
|
|
* buffers which should be written-back to the filesystem.
|
2006-10-11 08:20:57 +00:00
|
|
|
*/
|
|
|
|
|
2006-10-11 08:20:59 +00:00
|
|
|
int jbd2_cleanup_journal_tail(journal_t *journal)
|
2006-10-11 08:20:57 +00:00
|
|
|
{
|
|
|
|
tid_t first_tid;
|
2012-03-14 02:22:54 +00:00
|
|
|
unsigned long blocknr;
|
2006-10-11 08:20:57 +00:00
|
|
|
|
2008-10-11 00:29:13 +00:00
|
|
|
if (is_journal_aborted(journal))
|
|
|
|
return 1;
|
|
|
|
|
2012-03-14 02:22:54 +00:00
|
|
|
if (!jbd2_journal_get_log_tail(journal, &first_tid, &blocknr))
|
2006-10-11 08:20:57 +00:00
|
|
|
return 1;
|
2012-03-14 02:22:54 +00:00
|
|
|
J_ASSERT(blocknr != 0);
|
2009-12-23 11:52:08 +00:00
|
|
|
|
|
|
|
/*
|
2012-03-14 02:22:54 +00:00
|
|
|
* We need to make sure that any blocks that were recently written out
|
|
|
|
* --- perhaps by jbd2_log_do_checkpoint() --- are flushed out before
|
|
|
|
* we drop the transactions from the journal. It's unlikely this will
|
|
|
|
* be necessary, especially with an appropriately sized journal, but we
|
|
|
|
* need this to guarantee correctness. Fortunately
|
|
|
|
* jbd2_cleanup_journal_tail() doesn't get called all that often.
|
2009-12-23 11:52:08 +00:00
|
|
|
*/
|
2012-03-14 02:22:54 +00:00
|
|
|
if (journal->j_flags & JBD2_BARRIER)
|
2010-09-16 18:51:46 +00:00
|
|
|
blkdev_issue_flush(journal->j_fs_dev, GFP_KERNEL, NULL);
|
2012-03-14 02:22:54 +00:00
|
|
|
|
|
|
|
__jbd2_update_log_tail(journal, first_tid, blocknr);
|
2006-10-11 08:20:57 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* Checkpoint list management */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* journal_clean_one_cp_list
|
|
|
|
*
|
2011-06-13 19:38:22 +00:00
|
|
|
* Find all the written-back checkpoint buffers in the given list and
|
|
|
|
* release them.
|
2006-10-11 08:20:57 +00:00
|
|
|
*
|
|
|
|
* Called with j_list_lock held.
|
2011-12-05 12:00:34 +00:00
|
|
|
* Returns number of buffers reaped (for debug)
|
2006-10-11 08:20:57 +00:00
|
|
|
*/
|
|
|
|
|
|
|
|
static int journal_clean_one_cp_list(struct journal_head *jh, int *released)
|
|
|
|
{
|
|
|
|
struct journal_head *last_jh;
|
|
|
|
struct journal_head *next_jh = jh;
|
|
|
|
int ret, freed = 0;
|
|
|
|
|
|
|
|
*released = 0;
|
|
|
|
if (!jh)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
last_jh = jh->b_cpprev;
|
|
|
|
do {
|
|
|
|
jh = next_jh;
|
|
|
|
next_jh = jh->b_cpnext;
|
2012-03-14 02:45:25 +00:00
|
|
|
ret = __try_to_free_cp_buf(jh);
|
2014-09-18 04:42:16 +00:00
|
|
|
if (!ret)
|
|
|
|
return freed;
|
|
|
|
freed++;
|
|
|
|
if (ret == 2) {
|
|
|
|
*released = 1;
|
|
|
|
return freed;
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
|
|
|
/*
|
|
|
|
* This function only frees up some memory
|
|
|
|
* if possible so we dont have an obligation
|
|
|
|
* to finish processing. Bail out if preemption
|
|
|
|
* requested:
|
|
|
|
*/
|
|
|
|
if (need_resched())
|
|
|
|
return freed;
|
|
|
|
} while (jh != last_jh);
|
|
|
|
|
|
|
|
return freed;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* journal_clean_checkpoint_list
|
|
|
|
*
|
|
|
|
* Find all the written-back checkpoint buffers in the journal and release them.
|
|
|
|
*
|
|
|
|
* Called with j_list_lock held.
|
|
|
|
* Returns number of buffers reaped (for debug)
|
|
|
|
*/
|
|
|
|
|
2006-10-11 08:20:59 +00:00
|
|
|
int __jbd2_journal_clean_checkpoint_list(journal_t *journal)
|
2006-10-11 08:20:57 +00:00
|
|
|
{
|
|
|
|
transaction_t *transaction, *last_transaction, *next_transaction;
|
2014-09-18 04:42:16 +00:00
|
|
|
int ret;
|
|
|
|
int freed = 0;
|
2006-10-11 08:20:57 +00:00
|
|
|
int released;
|
|
|
|
|
|
|
|
transaction = journal->j_checkpoint_transactions;
|
|
|
|
if (!transaction)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
last_transaction = transaction->t_cpprev;
|
|
|
|
next_transaction = transaction;
|
|
|
|
do {
|
|
|
|
transaction = next_transaction;
|
|
|
|
next_transaction = transaction->t_cpnext;
|
2014-09-18 04:42:16 +00:00
|
|
|
ret = journal_clean_one_cp_list(transaction->
|
2006-10-11 08:20:57 +00:00
|
|
|
t_checkpoint_list, &released);
|
|
|
|
/*
|
|
|
|
* This function only frees up some memory if possible so we
|
|
|
|
* dont have an obligation to finish processing. Bail out if
|
|
|
|
* preemption requested:
|
|
|
|
*/
|
2014-09-18 04:42:16 +00:00
|
|
|
if (need_resched()) {
|
|
|
|
freed += ret;
|
2006-10-11 08:20:57 +00:00
|
|
|
goto out;
|
2014-09-18 04:42:16 +00:00
|
|
|
}
|
|
|
|
if (released) {
|
|
|
|
freed += ret;
|
2006-10-11 08:20:57 +00:00
|
|
|
continue;
|
2014-09-18 04:42:16 +00:00
|
|
|
}
|
2006-10-11 08:20:57 +00:00
|
|
|
/*
|
|
|
|
* It is essential that we are as careful as in the case of
|
|
|
|
* t_checkpoint_list with removing the buffer from the list as
|
|
|
|
* we can possibly see not yet submitted buffers on io_list
|
|
|
|
*/
|
|
|
|
ret += journal_clean_one_cp_list(transaction->
|
|
|
|
t_checkpoint_io_list, &released);
|
2014-09-18 04:42:16 +00:00
|
|
|
freed += ret;
|
|
|
|
if (need_resched() || !ret)
|
2006-10-11 08:20:57 +00:00
|
|
|
goto out;
|
|
|
|
} while (transaction != last_transaction);
|
|
|
|
out:
|
2014-09-18 04:42:16 +00:00
|
|
|
return freed;
|
2006-10-11 08:20:57 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* journal_remove_checkpoint: called after a buffer has been committed
|
|
|
|
* to disk (either by being write-back flushed to disk, or being
|
|
|
|
* committed to the log).
|
|
|
|
*
|
|
|
|
* We cannot safely clean a transaction out of the log until all of the
|
|
|
|
* buffer updates committed in that transaction have safely been stored
|
|
|
|
* elsewhere on disk. To achieve this, all of the buffers in a
|
|
|
|
* transaction need to be maintained on the transaction's checkpoint
|
|
|
|
* lists until they have been rewritten, at which point this function is
|
|
|
|
* called to remove the buffer from the existing transaction's
|
|
|
|
* checkpoint lists.
|
|
|
|
*
|
|
|
|
* The function returns 1 if it frees the transaction, 0 otherwise.
|
2011-06-13 19:38:22 +00:00
|
|
|
* The function can free jh and bh.
|
2006-10-11 08:20:57 +00:00
|
|
|
*
|
|
|
|
* This function is called with j_list_lock held.
|
|
|
|
*/
|
2006-10-11 08:20:59 +00:00
|
|
|
int __jbd2_journal_remove_checkpoint(struct journal_head *jh)
|
2006-10-11 08:20:57 +00:00
|
|
|
{
|
2009-09-30 04:32:06 +00:00
|
|
|
struct transaction_chp_stats_s *stats;
|
2006-10-11 08:20:57 +00:00
|
|
|
transaction_t *transaction;
|
|
|
|
journal_t *journal;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
JBUFFER_TRACE(jh, "entry");
|
|
|
|
|
|
|
|
if ((transaction = jh->b_cp_transaction) == NULL) {
|
|
|
|
JBUFFER_TRACE(jh, "not on transaction");
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
journal = transaction->t_journal;
|
|
|
|
|
2011-06-13 19:38:22 +00:00
|
|
|
JBUFFER_TRACE(jh, "removing from transaction");
|
2006-10-11 08:20:57 +00:00
|
|
|
__buffer_unlink(jh);
|
|
|
|
jh->b_cp_transaction = NULL;
|
2011-06-13 19:38:22 +00:00
|
|
|
jbd2_journal_put_journal_head(jh);
|
2006-10-11 08:20:57 +00:00
|
|
|
|
|
|
|
if (transaction->t_checkpoint_list != NULL ||
|
|
|
|
transaction->t_checkpoint_io_list != NULL)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* There is one special case to worry about: if we have just pulled the
|
2008-01-29 04:58:27 +00:00
|
|
|
* buffer off a running or committing transaction's checkpoing list,
|
|
|
|
* then even if the checkpoint list is empty, the transaction obviously
|
|
|
|
* cannot be dropped!
|
2006-10-11 08:20:57 +00:00
|
|
|
*
|
2008-01-29 04:58:27 +00:00
|
|
|
* The locking here around t_state is a bit sleazy.
|
2006-10-11 08:20:59 +00:00
|
|
|
* See the comment at the end of jbd2_journal_commit_transaction().
|
2006-10-11 08:20:57 +00:00
|
|
|
*/
|
2011-06-13 19:38:22 +00:00
|
|
|
if (transaction->t_state != T_FINISHED)
|
2006-10-11 08:20:57 +00:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* OK, that was the last buffer for the transaction: we can now
|
|
|
|
safely remove this transaction from the log */
|
2009-09-30 04:32:06 +00:00
|
|
|
stats = &transaction->t_chp_stats;
|
|
|
|
if (stats->cs_chp_time)
|
|
|
|
stats->cs_chp_time = jbd2_time_diff(stats->cs_chp_time,
|
|
|
|
jiffies);
|
|
|
|
trace_jbd2_checkpoint_stats(journal->j_fs_dev->bd_dev,
|
|
|
|
transaction->t_tid, stats);
|
2006-10-11 08:20:57 +00:00
|
|
|
|
2006-10-11 08:20:59 +00:00
|
|
|
__jbd2_journal_drop_transaction(journal, transaction);
|
2012-02-20 22:53:02 +00:00
|
|
|
jbd2_journal_free_transaction(transaction);
|
2006-10-11 08:20:57 +00:00
|
|
|
ret = 1;
|
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* journal_insert_checkpoint: put a committed buffer onto a checkpoint
|
|
|
|
* list so that we know when it is safe to clean the transaction out of
|
|
|
|
* the log.
|
|
|
|
*
|
|
|
|
* Called with the journal locked.
|
|
|
|
* Called with j_list_lock held.
|
|
|
|
*/
|
2006-10-11 08:20:59 +00:00
|
|
|
void __jbd2_journal_insert_checkpoint(struct journal_head *jh,
|
2006-10-11 08:20:57 +00:00
|
|
|
transaction_t *transaction)
|
|
|
|
{
|
|
|
|
JBUFFER_TRACE(jh, "entry");
|
|
|
|
J_ASSERT_JH(jh, buffer_dirty(jh2bh(jh)) || buffer_jbddirty(jh2bh(jh)));
|
|
|
|
J_ASSERT_JH(jh, jh->b_cp_transaction == NULL);
|
|
|
|
|
2011-06-13 19:38:22 +00:00
|
|
|
/* Get reference for checkpointing transaction */
|
|
|
|
jbd2_journal_grab_journal_head(jh2bh(jh));
|
2006-10-11 08:20:57 +00:00
|
|
|
jh->b_cp_transaction = transaction;
|
|
|
|
|
|
|
|
if (!transaction->t_checkpoint_list) {
|
|
|
|
jh->b_cpnext = jh->b_cpprev = jh;
|
|
|
|
} else {
|
|
|
|
jh->b_cpnext = transaction->t_checkpoint_list;
|
|
|
|
jh->b_cpprev = transaction->t_checkpoint_list->b_cpprev;
|
|
|
|
jh->b_cpprev->b_cpnext = jh;
|
|
|
|
jh->b_cpnext->b_cpprev = jh;
|
|
|
|
}
|
|
|
|
transaction->t_checkpoint_list = jh;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We've finished with this transaction structure: adios...
|
|
|
|
*
|
|
|
|
* The transaction must have no links except for the checkpoint by this
|
|
|
|
* point.
|
|
|
|
*
|
|
|
|
* Called with the journal locked.
|
|
|
|
* Called with j_list_lock held.
|
|
|
|
*/
|
|
|
|
|
2006-10-11 08:20:59 +00:00
|
|
|
void __jbd2_journal_drop_transaction(journal_t *journal, transaction_t *transaction)
|
2006-10-11 08:20:57 +00:00
|
|
|
{
|
|
|
|
assert_spin_locked(&journal->j_list_lock);
|
|
|
|
if (transaction->t_cpnext) {
|
|
|
|
transaction->t_cpnext->t_cpprev = transaction->t_cpprev;
|
|
|
|
transaction->t_cpprev->t_cpnext = transaction->t_cpnext;
|
|
|
|
if (journal->j_checkpoint_transactions == transaction)
|
|
|
|
journal->j_checkpoint_transactions =
|
|
|
|
transaction->t_cpnext;
|
|
|
|
if (journal->j_checkpoint_transactions == transaction)
|
|
|
|
journal->j_checkpoint_transactions = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
J_ASSERT(transaction->t_state == T_FINISHED);
|
|
|
|
J_ASSERT(transaction->t_buffers == NULL);
|
|
|
|
J_ASSERT(transaction->t_forget == NULL);
|
|
|
|
J_ASSERT(transaction->t_shadow_list == NULL);
|
|
|
|
J_ASSERT(transaction->t_checkpoint_list == NULL);
|
|
|
|
J_ASSERT(transaction->t_checkpoint_io_list == NULL);
|
2010-08-02 12:43:25 +00:00
|
|
|
J_ASSERT(atomic_read(&transaction->t_updates) == 0);
|
2006-10-11 08:20:57 +00:00
|
|
|
J_ASSERT(journal->j_committing_transaction != transaction);
|
|
|
|
J_ASSERT(journal->j_running_transaction != transaction);
|
|
|
|
|
2012-02-20 22:53:01 +00:00
|
|
|
trace_jbd2_drop_transaction(journal, transaction);
|
|
|
|
|
2006-10-11 08:20:57 +00:00
|
|
|
jbd_debug(1, "Dropping transaction %d, all done\n", transaction->t_tid);
|
|
|
|
}
|