/*
* __log_wait_for_space: wait until there is space in the journal.
*
* 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.
*/
void __log_wait_for_space(journal_t *journal)
{
int nblocks, space_left;
assert_spin_locked(&journal->j_state_lock);
nblocks = jbd_space_needed(journal);
while (__log_space_left(journal) < nblocks) {
if (journal->j_flags & JFS_ABORT)
return;
spin_unlock(&journal->j_state_lock);
mutex_lock(&journal->j_checkpoint_mutex);
/*
* Test again, another process may have checkpointed while we
* were waiting for the checkpoint lock. If there are no
* 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.
*/
spin_lock(&journal->j_state_lock);
spin_lock(&journal->j_list_lock);
nblocks = jbd_space_needed(journal);
space_left = __log_space_left(journal);
if (space_left < nblocks) {
int chkpt = journal->j_checkpoint_transactions != NULL;
tid_t tid = 0;
if (journal->j_committing_transaction)
tid = journal->j_committing_transaction->t_tid;
spin_unlock(&journal->j_list_lock);
spin_unlock(&journal->j_state_lock);
if (chkpt) {
log_do_checkpoint(journal);
} else if (cleanup_journal_tail(journal) == 0) {
/* We were able to recover space; yay! */
;
} else if (tid) {
log_wait_commit(journal, tid);
} else {
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\n", __func__);
WARN_ON(1);
journal_abort(journal, 0);
}
spin_lock(&journal->j_state_lock);
} else {
spin_unlock(&journal->j_list_lock);
}
mutex_unlock(&journal->j_checkpoint_mutex);
}
}
void __log_wait_for_space(journal_t *journal)
{
int nblocks, space_left;
assert_spin_locked(&journal->j_state_lock);
nblocks = jbd_space_needed(journal);
while (__log_space_left(journal) < nblocks) {
if (journal->j_flags & JFS_ABORT)
return;
spin_unlock(&journal->j_state_lock);
mutex_lock(&journal->j_checkpoint_mutex);
/*
*/
spin_lock(&journal->j_state_lock);
spin_lock(&journal->j_list_lock);
nblocks = jbd_space_needed(journal);
space_left = __log_space_left(journal);
if (space_left < nblocks) {
int chkpt = journal->j_checkpoint_transactions != NULL;
tid_t tid = 0;
if (journal->j_committing_transaction)
tid = journal->j_committing_transaction->t_tid;
spin_unlock(&journal->j_list_lock);
spin_unlock(&journal->j_state_lock);
if (chkpt) {
log_do_checkpoint(journal);
} else if (cleanup_journal_tail(journal) == 0) {
/* */
;
} else if (tid) {
log_wait_commit(journal, tid);
} else {
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\n", __func__);
WARN_ON(1);
journal_abort(journal, 0);
}
spin_lock(&journal->j_state_lock);
} else {
spin_unlock(&journal->j_list_lock);
}
mutex_unlock(&journal->j_checkpoint_mutex);
}
}
/* @@@ `nblocks' is unused. Should it be used? */
int log_do_checkpoint (journal_t *journal, int nblocks)
{
transaction_t *transaction, *last_transaction, *next_transaction;
int result;
int target;
int batch_count = 0;
struct buffer_head *bhs[NR_BATCH];
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.
*/
result = cleanup_journal_tail(journal);
jbd_debug(1, "cleanup_journal_tail returned %d\n", result);
if (result <= 0)
return result;
/*
* OK, we need to start writing disk blocks. Try to free up a
* quarter of the log in a single checkpoint if we can.
*/
/*
* AKPM: check this code. I had a feeling a while back that it
* degenerates into a busy loop at unmount time.
*/
target = (journal->j_last - journal->j_first) / 4;
spin_lock(&journal_datalist_lock);
repeat:
transaction = journal->j_checkpoint_transactions;
if (transaction == NULL)
goto done;
last_transaction = transaction->t_cpprev;
next_transaction = transaction;
do {
struct journal_head *jh, *last_jh, *next_jh;
int drop_count = 0;
int cleanup_ret, retry = 0;
transaction = next_transaction;
next_transaction = transaction->t_cpnext;
jh = transaction->t_checkpoint_list;
last_jh = jh->b_cpprev;
next_jh = jh;
do {
jh = next_jh;
next_jh = jh->b_cpnext;
retry = __flush_buffer(journal, jh, bhs, &batch_count,
&drop_count);
} while (jh != last_jh && !retry);
if (batch_count) {
__flush_batch(bhs, &batch_count);
goto repeat;
}
if (retry)
goto repeat;
/*
* We have walked the whole transaction list without
* finding anything to write to disk. We had better be
* able to make some progress or we are in trouble.
*/
cleanup_ret = __cleanup_transaction(journal, transaction);
J_ASSERT(drop_count != 0 || cleanup_ret != 0);
goto repeat; /* __cleanup may have dropped lock */
} while (transaction != last_transaction);
done:
spin_unlock(&journal_datalist_lock);
result = cleanup_journal_tail(journal);
if (result < 0)
return result;
return 0;
}
/*
* Perform an actual checkpoint. We don't write out only enough to
* satisfy the current blocked requests: rather we submit a reasonably
* sized chunk of the outstanding data to disk at once for
* efficiency. __log_wait_for_space() will retry if we didn't free enough.
*
* However, we _do_ take into account the amount requested so that once
* the IO has been queued, we can return as soon as enough of it has
* completed to disk.
*
* The journal should be locked before calling this function.
*/
int log_do_checkpoint(journal_t *journal)
{
int result;
int batch_count = 0;
struct buffer_head *bhs[NR_BATCH];
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.
*/
result = cleanup_journal_tail(journal);
jbd_debug(1, "cleanup_journal_tail returned %d\n", result);
if (result <= 0)
return result;
/*
* OK, we need to start writing disk blocks. Try to free up a
* quarter of the log in a single checkpoint if we can.
*/
/*
* AKPM: check this code. I had a feeling a while back that it
* degenerates into a busy loop at unmount time.
*/
spin_lock(&journal->j_list_lock);
while (journal->j_checkpoint_transactions) {
transaction_t *transaction;
struct journal_head *jh, *last_jh, *next_jh;
int drop_count = 0;
int cleanup_ret, retry = 0;
tid_t this_tid;
transaction = journal->j_checkpoint_transactions;
this_tid = transaction->t_tid;
jh = transaction->t_checkpoint_list;
last_jh = jh->b_cpprev;
next_jh = jh;
do {
struct buffer_head *bh;
jh = next_jh;
next_jh = jh->b_cpnext;
bh = jh2bh(jh);
if (!jbd_trylock_bh_state(bh)) {
jbd_sync_bh(journal, bh);
spin_lock(&journal->j_list_lock);
retry = 1;
break;
}
retry = __flush_buffer(journal, jh, bhs, &batch_count, &drop_count);
if (cond_resched_lock(&journal->j_list_lock)) {
retry = 1;
break;
}
} while (jh != last_jh && !retry);
if (batch_count) {
__flush_batch(journal, bhs, &batch_count);
retry = 1;
}
/*
* If someone cleaned up this transaction while we slept, we're
* done
*/
if (journal->j_checkpoint_transactions != transaction)
break;
if (retry)
continue;
/*
* Maybe it's a new transaction, but it fell at the same
* address
*/
if (transaction->t_tid != this_tid)
continue;
/*
* We have walked the whole transaction list without
* finding anything to write to disk. We had better be
* able to make some progress or we are in trouble.
*/
cleanup_ret = __cleanup_transaction(journal, transaction);
J_ASSERT(drop_count != 0 || cleanup_ret != 0);
if (journal->j_checkpoint_transactions != transaction)
break;
}
spin_unlock(&journal->j_list_lock);
result = cleanup_journal_tail(journal);
if (result < 0)
return result;
return 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.
*/
int log_do_checkpoint(journal_t *journal)
{
transaction_t *transaction;
tid_t this_tid;
int result;
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.
*/
result = cleanup_journal_tail(journal);
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.
*/
spin_lock(&journal->j_list_lock);
if (!journal->j_checkpoint_transactions)
goto out;
transaction = journal->j_checkpoint_transactions;
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).
*/
if (journal->j_checkpoint_transactions == transaction &&
transaction->t_tid == this_tid) {
int batch_count = 0;
struct buffer_head *bhs[NR_BATCH];
struct journal_head *jh;
int retry = 0;
while (!retry && transaction->t_checkpoint_list) {
struct buffer_head *bh;
jh = transaction->t_checkpoint_list;
bh = jh2bh(jh);
if (!jbd_trylock_bh_state(bh)) {
jbd_sync_bh(journal, bh);
retry = 1;
break;
}
retry = __process_buffer(journal, jh, bhs,&batch_count);
if (!retry && lock_need_resched(&journal->j_list_lock)){
spin_unlock(&journal->j_list_lock);
retry = 1;
break;
}
}
if (batch_count) {
if (!retry) {
spin_unlock(&journal->j_list_lock);
retry = 1;
}
__flush_batch(journal, bhs, &batch_count);
}
if (retry) {
spin_lock(&journal->j_list_lock);
goto restart;
}
/*
* Now we have cleaned up the first transaction's checkpoint
* list. Let's clean up the second one
*/
__wait_cp_io(journal, transaction);
}
out:
spin_unlock(&journal->j_list_lock);
result = cleanup_journal_tail(journal);
if (result < 0)
return result;
return 0;
}