jbd2,ext4: add a shrinker to release checkpointed buffers

Current metadata buffer release logic in bdev_try_to_free_page() have
a lot of use-after-free issues when umount filesystem concurrently, and
it is difficult to fix directly because ext4 is the only user of
s_op->bdev_try_to_free_page callback and we may have to add more special
refcount or lock that is only used by ext4 into the common vfs layer,
which is unacceptable.

One better solution is remove the bdev_try_to_free_page callback, but
the real problem is we cannot easily release journal_head on the
checkpointed buffer, so try_to_free_buffers() cannot release buffers and
page under memory pressure, which is more likely to trigger
out-of-memory. So we cannot remove the callback directly before we find
another way to release journal_head.

This patch introduce a shrinker to free journal_head on the checkpointed
transaction. After the journal_head got freed, try_to_free_buffers()
could free buffer properly.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Suggested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20210610112440.3438139-6-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

1 files changed, 87 insertions, 0 deletions

diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index 8b3f5bbd65f9..7c52feb6f753 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -2051,6 +2051,91 @@ recovery_error:
 }
 
 /**
+ * jbd2_journal_shrink_scan()
+ *
+ * Scan the checkpointed buffer on the checkpoint list and release the
+ * journal_head.
+ */
+static unsigned long jbd2_journal_shrink_scan(struct shrinker *shrink,
+					      struct shrink_control *sc)
+{
+	journal_t *journal = container_of(shrink, journal_t, j_shrinker);
+	unsigned long nr_to_scan = sc->nr_to_scan;
+	unsigned long nr_shrunk;
+	unsigned long count;
+
+	count = percpu_counter_read_positive(&journal->j_jh_shrink_count);
+	trace_jbd2_shrink_scan_enter(journal, sc->nr_to_scan, count);
+
+	nr_shrunk = jbd2_journal_shrink_checkpoint_list(journal, &nr_to_scan);
+
+	count = percpu_counter_read_positive(&journal->j_jh_shrink_count);
+	trace_jbd2_shrink_scan_exit(journal, nr_to_scan, nr_shrunk, count);
+
+	return nr_shrunk;
+}
+
+/**
+ * jbd2_journal_shrink_count()
+ *
+ * Count the number of checkpoint buffers on the checkpoint list.
+ */
+static unsigned long jbd2_journal_shrink_count(struct shrinker *shrink,
+					       struct shrink_control *sc)
+{
+	journal_t *journal = container_of(shrink, journal_t, j_shrinker);
+	unsigned long count;
+
+	count = percpu_counter_read_positive(&journal->j_jh_shrink_count);
+	trace_jbd2_shrink_count(journal, sc->nr_to_scan, count);
+
+	return count;
+}
+
+/**
+ * jbd2_journal_register_shrinker()
+ * @journal: Journal to act on.
+ *
+ * Init a percpu counter to record the checkpointed buffers on the checkpoint
+ * list and register a shrinker to release their journal_head.
+ */
+int jbd2_journal_register_shrinker(journal_t *journal)
+{
+	int err;
+
+	journal->j_shrink_transaction = NULL;
+
+	err = percpu_counter_init(&journal->j_jh_shrink_count, 0, GFP_KERNEL);
+	if (err)
+		return err;
+
+	journal->j_shrinker.scan_objects = jbd2_journal_shrink_scan;
+	journal->j_shrinker.count_objects = jbd2_journal_shrink_count;
+	journal->j_shrinker.seeks = DEFAULT_SEEKS;
+	journal->j_shrinker.batch = journal->j_max_transaction_buffers;
+
+	err = register_shrinker(&journal->j_shrinker);
+	if (err) {
+		percpu_counter_destroy(&journal->j_jh_shrink_count);
+		return err;
+	}
+
+	return 0;
+}
+
+/**
+ * jbd2_journal_unregister_shrinker()
+ * @journal: Journal to act on.
+ *
+ * Unregister the checkpointed buffer shrinker and destroy the percpu counter.
+ */
+void jbd2_journal_unregister_shrinker(journal_t *journal)
+{
+	percpu_counter_destroy(&journal->j_jh_shrink_count);
+	unregister_shrinker(&journal->j_shrinker);
+}
+
+/**
  * jbd2_journal_destroy() - Release a journal_t structure.
  * @journal: Journal to act on.
  *
@@ -2122,6 +2207,8 @@ int jbd2_journal_destroy(journal_t *journal)
 		brelse(journal->j_sb_buffer);
 	}
 
+	jbd2_journal_unregister_shrinker(journal);
+
 	if (journal->j_proc_entry)
 		jbd2_stats_proc_exit(journal);
 	iput(journal->j_inode);