jbd: don't wait (forever) for stale tid caused by wraparound

In the case where an inode has a very stale transaction id (tid) in
i_datasync_tid or i_sync_tid, it's possible that after a very large
(2**31) number of transactions, that the tid number space might wrap,
causing tid_geq()'s calculations to fail.

Commit d9b0193 "jbd: fix fsync() tid wraparound bug" attempted to fix
this problem, but it only avoided kjournald spinning forever by fixing
the logic in jbd_log_start_commit().

Signed-off-by: Jan Kara <jack@suse.cz>

1 files changed, 11 insertions, 0 deletions

diff --git a/fs/jbd/journal.c b/fs/jbd/journal.c
index 81cc7eaff863..81880c6d6b5e 100644
--- a/fs/jbd/journal.c
+++ b/fs/jbd/journal.c
@@ -564,6 +564,16 @@ int log_wait_commit(journal_t *journal, tid_t tid)
         spin_unlock(&journal->j_state_lock);
 #endif
         spin_lock(&journal->j_state_lock);
+        /*
+         * Not running or committing trans? Must be already committed. This
+         * saves us from waiting for a *long* time when tid overflows.
+         */
+        if (!((journal->j_running_transaction &&
+               journal->j_running_transaction->t_tid == tid) ||
+              (journal->j_committing_transaction &&
+               journal->j_committing_transaction->t_tid == tid)))
+                goto out_unlock;
+
         if (!tid_geq(journal->j_commit_waited, tid))
                 journal->j_commit_waited = tid;
         while (tid_gt(tid, journal->j_commit_sequence)) {
@@ -575,6 +585,7 @@ int log_wait_commit(journal_t *journal, tid_t tid)
                                 !tid_gt(tid, journal->j_commit_sequence));
                 spin_lock(&journal->j_state_lock);
         }
+out_unlock:
         spin_unlock(&journal->j_state_lock);
 
         if (unlikely(is_journal_aborted(journal))) {