jbd2: avoid infinite loop when destroying aborted journal

Commit 6f6a6fda2945 "jbd2: fix ocfs2 corrupt when updating journal
superblock fails" changed jbd2_cleanup_journal_tail() to return EIO
when the journal is aborted. That makes logic in
jbd2_log_do_checkpoint() bail out which is fine, except that
jbd2_journal_destroy() expects jbd2_log_do_checkpoint() to always make
a progress in cleaning the journal. Without it jbd2_journal_destroy()
just loops in an infinite loop.

Fix jbd2_journal_destroy() to cleanup journal checkpoint lists of
jbd2_log_do_checkpoint() fails with error.

Reported-by: Eryu Guan <guaneryu@gmail.com>
Tested-by: Eryu Guan <guaneryu@gmail.com>
Fixes: 6f6a6fda294506dfe0e3e0a253bb2d2923f28f0a
Signed-off-by: Jan Kara <jack@suse.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c
index 4227dc4..8c44654 100644 (file)
--- a/fs/jbd2/checkpoint.c
+++ b/fs/jbd2/checkpoint.c
@@ -417,12 +417,12 @@ int jbd2_cleanup_journal_tail(journal_t *journal)
  * journal_clean_one_cp_list
  *
  * Find all the written-back checkpoint buffers in the given list and
- * release them.
+ * release them. If 'destroy' is set, clean all buffers unconditionally.
  *
  * Called with j_list_lock held.
  * Returns 1 if we freed the transaction, 0 otherwise.
  */
-static int journal_clean_one_cp_list(struct journal_head *jh)
+static int journal_clean_one_cp_list(struct journal_head *jh, bool destroy)
 {
        struct journal_head *last_jh;
        struct journal_head *next_jh = jh;
@@ -436,7 +436,10 @@ static int journal_clean_one_cp_list(struct journal_head *jh)
        do {
                jh = next_jh;
                next_jh = jh->b_cpnext;
-               ret = __try_to_free_cp_buf(jh);
+               if (!destroy)
+                       ret = __try_to_free_cp_buf(jh);
+               else
+                       ret = __jbd2_journal_remove_checkpoint(jh) + 1;
                if (!ret)
                        return freed;
                if (ret == 2)
@@ -459,10 +462,11 @@ static int journal_clean_one_cp_list(struct journal_head *jh)
  * journal_clean_checkpoint_list
  *
  * Find all the written-back checkpoint buffers in the journal and release them.
+ * If 'destroy' is set, release all buffers unconditionally.
  *
  * Called with j_list_lock held.
  */
-void __jbd2_journal_clean_checkpoint_list(journal_t *journal)
+void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy)
 {
        transaction_t *transaction, *last_transaction, *next_transaction;
        int ret;
@@ -476,7 +480,8 @@ void __jbd2_journal_clean_checkpoint_list(journal_t *journal)
        do {
                transaction = next_transaction;
                next_transaction = transaction->t_cpnext;
-               ret = journal_clean_one_cp_list(transaction->t_checkpoint_list);
+               ret = journal_clean_one_cp_list(transaction->t_checkpoint_list,
+                                               destroy);
                /*
                 * This function only frees up some memory if possible so we
                 * dont have an obligation to finish processing. Bail out if
@@ -492,7 +497,7 @@ void __jbd2_journal_clean_checkpoint_list(journal_t *journal)
                 * we can possibly see not yet submitted buffers on io_list
                 */
                ret = journal_clean_one_cp_list(transaction->
-                               t_checkpoint_io_list);
+                               t_checkpoint_io_list, destroy);
                if (need_resched())
                        return;
                /*
@@ -505,6 +510,28 @@ void __jbd2_journal_clean_checkpoint_list(journal_t *journal)
        } while (transaction != last_transaction);
 }
 
+/*
+ * Remove buffers from all checkpoint lists as journal is aborted and we just
+ * need to free memory
+ */
+void jbd2_journal_destroy_checkpoint(journal_t *journal)
+{
+       /*
+        * We loop because __jbd2_journal_clean_checkpoint_list() may abort
+        * early due to a need of rescheduling.
+        */
+       while (1) {
+               spin_lock(&journal->j_list_lock);
+               if (!journal->j_checkpoint_transactions) {
+                       spin_unlock(&journal->j_list_lock);
+                       break;
+               }
+               __jbd2_journal_clean_checkpoint_list(journal, true);
+               spin_unlock(&journal->j_list_lock);
+               cond_resched();
+       }
+}
+
 /*
  * journal_remove_checkpoint: called after a buffer has been committed
  * to disk (either by being write-back flushed to disk, or being

diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index b73e021..362e5f6 100644 (file)
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c
@@ -510,7 +510,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
         * frees some memory
         */
        spin_lock(&journal->j_list_lock);
-       __jbd2_journal_clean_checkpoint_list(journal);
+       __jbd2_journal_clean_checkpoint_list(journal, false);
        spin_unlock(&journal->j_list_lock);
 
        jbd_debug(3, "JBD2: commit phase 1\n");

diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index fe1b4bd..8270fe9 100644 (file)
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -1693,8 +1693,17 @@ int jbd2_journal_destroy(journal_t *journal)
        while (journal->j_checkpoint_transactions != NULL) {
                spin_unlock(&journal->j_list_lock);
                mutex_lock(&journal->j_checkpoint_mutex);
-               jbd2_log_do_checkpoint(journal);
+               err = jbd2_log_do_checkpoint(journal);
                mutex_unlock(&journal->j_checkpoint_mutex);
+               /*
+                * If checkpointing failed, just free the buffers to avoid
+                * looping forever
+                */
+               if (err) {
+                       jbd2_journal_destroy_checkpoint(journal);
+                       spin_lock(&journal->j_list_lock);
+                       break;
+               }
                spin_lock(&journal->j_list_lock);
        }
 

diff --git a/include/linux/jbd2.h b/include/linux/jbd2.h
index edb640a..eb1cebe 100644 (file)
--- a/include/linux/jbd2.h
+++ b/include/linux/jbd2.h
@@ -1042,8 +1042,9 @@ void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
 extern void jbd2_journal_commit_transaction(journal_t *);
 
 /* Checkpoint list management */
-void __jbd2_journal_clean_checkpoint_list(journal_t *journal);
+void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
 int __jbd2_journal_remove_checkpoint(struct journal_head *);
+void jbd2_journal_destroy_checkpoint(journal_t *journal);
 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);