summaryrefslogtreecommitdiffstats
path: root/fs/jbd
diff options
context:
space:
mode:
Diffstat (limited to 'fs/jbd')
-rw-r--r--fs/jbd/checkpoint.c27
-rw-r--r--fs/jbd/commit.c6
-rw-r--r--fs/jbd/journal.c1
-rw-r--r--fs/jbd/recovery.c4
-rw-r--r--fs/jbd/revoke.c34
-rw-r--r--fs/jbd/transaction.c38
6 files changed, 88 insertions, 22 deletions
diff --git a/fs/jbd/checkpoint.c b/fs/jbd/checkpoint.c
index 5d1a00a5041b..05f0754f2b46 100644
--- a/fs/jbd/checkpoint.c
+++ b/fs/jbd/checkpoint.c
@@ -453,8 +453,6 @@ out:
*
* 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
@@ -472,13 +470,14 @@ int cleanup_journal_tail(journal_t *journal)
if (is_journal_aborted(journal))
return 1;
- /* OK, work out the oldest transaction remaining in the log, and
+ /*
+ * OK, work out the oldest transaction remaining in the log, and
* the log block it starts at.
*
* If the log is now empty, we need to work out which is the
* next transaction ID we will write, and where it will
- * start. */
-
+ * start.
+ */
spin_lock(&journal->j_state_lock);
spin_lock(&journal->j_list_lock);
transaction = journal->j_checkpoint_transactions;
@@ -504,7 +503,25 @@ int cleanup_journal_tail(journal_t *journal)
spin_unlock(&journal->j_state_lock);
return 1;
}
+ spin_unlock(&journal->j_state_lock);
+
+ /*
+ * We need to make sure that any blocks that were recently written out
+ * --- perhaps by 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
+ * cleanup_journal_tail() doesn't get called all that often.
+ */
+ if (journal->j_flags & JFS_BARRIER)
+ blkdev_issue_flush(journal->j_fs_dev, GFP_KERNEL, NULL);
+ spin_lock(&journal->j_state_lock);
+ if (!tid_gt(first_tid, journal->j_tail_sequence)) {
+ spin_unlock(&journal->j_state_lock);
+ /* Someone else cleaned up journal so return 0 */
+ return 0;
+ }
/* OK, update the superblock to recover the freed space.
* Physical blocks come first: have we wrapped beyond the end of
* the log? */
diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c
index 8799207df058..f2b9a571f4cf 100644
--- a/fs/jbd/commit.c
+++ b/fs/jbd/commit.c
@@ -392,6 +392,12 @@ void journal_commit_transaction(journal_t *journal)
jbd_debug (3, "JBD: commit phase 1\n");
/*
+ * Clear revoked flag to reflect there is no revoked buffers
+ * in the next transaction which is going to be started.
+ */
+ journal_clear_buffer_revoked_flags(journal);
+
+ /*
* Switch to a new revoke table.
*/
journal_switch_revoke_table(journal);
diff --git a/fs/jbd/journal.c b/fs/jbd/journal.c
index a96cff0c5f1d..59c09f9541b5 100644
--- a/fs/jbd/journal.c
+++ b/fs/jbd/journal.c
@@ -721,7 +721,6 @@ static journal_t * journal_init_common (void)
init_waitqueue_head(&journal->j_wait_checkpoint);
init_waitqueue_head(&journal->j_wait_commit);
init_waitqueue_head(&journal->j_wait_updates);
- mutex_init(&journal->j_barrier);
mutex_init(&journal->j_checkpoint_mutex);
spin_lock_init(&journal->j_revoke_lock);
spin_lock_init(&journal->j_list_lock);
diff --git a/fs/jbd/recovery.c b/fs/jbd/recovery.c
index 5b43e96788e6..008bf062fd26 100644
--- a/fs/jbd/recovery.c
+++ b/fs/jbd/recovery.c
@@ -20,6 +20,7 @@
#include <linux/fs.h>
#include <linux/jbd.h>
#include <linux/errno.h>
+#include <linux/blkdev.h>
#endif
/*
@@ -263,6 +264,9 @@ int journal_recover(journal_t *journal)
err2 = sync_blockdev(journal->j_fs_dev);
if (!err)
err = err2;
+ /* Flush disk caches to get replayed data on the permanent storage */
+ if (journal->j_flags & JFS_BARRIER)
+ blkdev_issue_flush(journal->j_fs_dev, GFP_KERNEL, NULL);
return err;
}
diff --git a/fs/jbd/revoke.c b/fs/jbd/revoke.c
index 305a90763154..25c713e7071c 100644
--- a/fs/jbd/revoke.c
+++ b/fs/jbd/revoke.c
@@ -47,6 +47,10 @@
* overwriting the new data. We don't even need to clear the revoke
* bit here.
*
+ * We cache revoke status of a buffer in the current transaction in b_states
+ * bits. As the name says, revokevalid flag indicates that the cached revoke
+ * status of a buffer is valid and we can rely on the cached status.
+ *
* Revoke information on buffers is a tri-state value:
*
* RevokeValid clear: no cached revoke status, need to look it up
@@ -479,6 +483,36 @@ int journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
return did_revoke;
}
+/*
+ * journal_clear_revoked_flags clears revoked flag of buffers in
+ * revoke table to reflect there is no revoked buffer in the next
+ * transaction which is going to be started.
+ */
+void journal_clear_buffer_revoked_flags(journal_t *journal)
+{
+ struct jbd_revoke_table_s *revoke = journal->j_revoke;
+ int i = 0;
+
+ for (i = 0; i < revoke->hash_size; i++) {
+ struct list_head *hash_list;
+ struct list_head *list_entry;
+ hash_list = &revoke->hash_table[i];
+
+ list_for_each(list_entry, hash_list) {
+ struct jbd_revoke_record_s *record;
+ struct buffer_head *bh;
+ record = (struct jbd_revoke_record_s *)list_entry;
+ bh = __find_get_block(journal->j_fs_dev,
+ record->blocknr,
+ journal->j_blocksize);
+ if (bh) {
+ clear_buffer_revoked(bh);
+ __brelse(bh);
+ }
+ }
+ }
+}
+
/* journal_switch_revoke table select j_revoke for next transaction
* we do not want to suspend any processing until all revokes are
* written -bzzz
diff --git a/fs/jbd/transaction.c b/fs/jbd/transaction.c
index 7e59c6e66f9b..7fce94b04bc3 100644
--- a/fs/jbd/transaction.c
+++ b/fs/jbd/transaction.c
@@ -426,17 +426,34 @@ int journal_restart(handle_t *handle, int nblocks)
* void journal_lock_updates () - establish a transaction barrier.
* @journal: Journal to establish a barrier on.
*
- * This locks out any further updates from being started, and blocks
- * until all existing updates have completed, returning only once the
- * journal is in a quiescent state with no updates running.
- *
- * The journal lock should not be held on entry.
+ * This locks out any further updates from being started, and blocks until all
+ * existing updates have completed, returning only once the journal is in a
+ * quiescent state with no updates running.
+ *
+ * We do not use simple mutex for synchronization as there are syscalls which
+ * want to return with filesystem locked and that trips up lockdep. Also
+ * hibernate needs to lock filesystem but locked mutex then blocks hibernation.
+ * Since locking filesystem is rare operation, we use simple counter and
+ * waitqueue for locking.
*/
void journal_lock_updates(journal_t *journal)
{
DEFINE_WAIT(wait);
+wait:
+ /* Wait for previous locked operation to finish */
+ wait_event(journal->j_wait_transaction_locked,
+ journal->j_barrier_count == 0);
+
spin_lock(&journal->j_state_lock);
+ /*
+ * Check reliably under the lock whether we are the ones winning the race
+ * and locking the journal
+ */
+ if (journal->j_barrier_count > 0) {
+ spin_unlock(&journal->j_state_lock);
+ goto wait;
+ }
++journal->j_barrier_count;
/* Wait until there are no running updates */
@@ -460,14 +477,6 @@ void journal_lock_updates(journal_t *journal)
spin_lock(&journal->j_state_lock);
}
spin_unlock(&journal->j_state_lock);
-
- /*
- * We have now established a barrier against other normal updates, but
- * we also need to barrier against other journal_lock_updates() calls
- * to make sure that we serialise special journal-locked operations
- * too.
- */
- mutex_lock(&journal->j_barrier);
}
/**
@@ -475,14 +484,11 @@ void journal_lock_updates(journal_t *journal)
* @journal: Journal to release the barrier on.
*
* Release a transaction barrier obtained with journal_lock_updates().
- *
- * Should be called without the journal lock held.
*/
void journal_unlock_updates (journal_t *journal)
{
J_ASSERT(journal->j_barrier_count != 0);
- mutex_unlock(&journal->j_barrier);
spin_lock(&journal->j_state_lock);
--journal->j_barrier_count;
spin_unlock(&journal->j_state_lock);