aboutsummaryrefslogtreecommitdiff
path: root/fs/ocfs2/journal.c
diff options
context:
space:
mode:
authorSrinivas Eeda <srinivas.eeda@oracle.com>2009-06-03 17:02:55 -0700
committerJoel Becker <joel.becker@oracle.com>2009-06-03 19:14:31 -0700
commit83273932fbefb6ceef9c0b82ac4d23900728f4d9 (patch)
tree7ece8471af3d2a22543542e990369aca47aa3d25 /fs/ocfs2/journal.c
parentedd45c08499a3e9d4c25431cd2b6a9ce5f692c92 (diff)
ocfs2: timer to queue scan of all orphan slots
When a dentry is unlinked, the unlinking node takes an EX on the dentry lock before moving the dentry to the orphan directory. Other nodes that have this dentry in cache have a PR on the same dentry lock. When the EX is requested, the other nodes flag the corresponding inode as MAYBE_ORPHANED during downconvert. The inode is finally deleted when the last node to iput the inode sees that i_nlink==0 and the MAYBE_ORPHANED flag is set. A problem arises if a node is forced to free dentry locks because of memory pressure. If this happens, the node will no longer get downconvert notifications for the dentries that have been unlinked on another node. If it also happens that node is actively using the corresponding inode and happens to be the one performing the last iput on that inode, it will fail to delete the inode as it will not have the MAYBE_ORPHANED flag set. This patch fixes this shortcoming by introducing a periodic scan of the orphan directories to delete such inodes. Care has been taken to distribute the workload across the cluster so that no one node has to perform the task all the time. Signed-off-by: Srinivas Eeda <srinivas.eeda@oracle.com> Signed-off-by: Joel Becker <joel.becker@oracle.com>
Diffstat (limited to 'fs/ocfs2/journal.c')
-rw-r--r--fs/ocfs2/journal.c107
1 files changed, 107 insertions, 0 deletions
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index a20a0f1e37f..44ed768782e 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -28,6 +28,8 @@
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/kthread.h>
+#include <linux/time.h>
+#include <linux/random.h>
#define MLOG_MASK_PREFIX ML_JOURNAL
#include <cluster/masklog.h>
@@ -52,6 +54,8 @@
DEFINE_SPINLOCK(trans_inc_lock);
+#define ORPHAN_SCAN_SCHEDULE_TIMEOUT 300000
+
static int ocfs2_force_read_journal(struct inode *inode);
static int ocfs2_recover_node(struct ocfs2_super *osb,
int node_num, int slot_num);
@@ -1841,6 +1845,109 @@ bail:
return status;
}
+/*
+ * Scan timer should get fired every ORPHAN_SCAN_SCHEDULE_TIMEOUT. Add some
+ * randomness to the timeout to minimize multple nodes firing the timer at the
+ * same time.
+ */
+static inline unsigned long ocfs2_orphan_scan_timeout(void)
+{
+ unsigned long time;
+
+ get_random_bytes(&time, sizeof(time));
+ time = ORPHAN_SCAN_SCHEDULE_TIMEOUT + (time % 5000);
+ return msecs_to_jiffies(time);
+}
+
+/*
+ * ocfs2_queue_orphan_scan calls ocfs2_queue_recovery_completion for
+ * every slot, queuing a recovery of the slot on the ocfs2_wq thread. This
+ * is done to catch any orphans that are left over in orphan directories.
+ *
+ * ocfs2_queue_orphan_scan gets called every ORPHAN_SCAN_SCHEDULE_TIMEOUT
+ * seconds. It gets an EX lock on os_lockres and checks sequence number
+ * stored in LVB. If the sequence number has changed, it means some other
+ * node has done the scan. This node skips the scan and tracks the
+ * sequence number. If the sequence number didn't change, it means a scan
+ * hasn't happened. The node queues a scan and increments the
+ * sequence number in the LVB.
+ */
+void ocfs2_queue_orphan_scan(struct ocfs2_super *osb)
+{
+ struct ocfs2_orphan_scan *os;
+ int status, i;
+ u32 seqno = 0;
+
+ os = &osb->osb_orphan_scan;
+
+ status = ocfs2_orphan_scan_lock(osb, &seqno, DLM_LOCK_EX);
+ if (status < 0) {
+ if (status != -EAGAIN)
+ mlog_errno(status);
+ goto out;
+ }
+
+ if (os->os_seqno != seqno) {
+ os->os_seqno = seqno;
+ goto unlock;
+ }
+
+ for (i = 0; i < osb->max_slots; i++)
+ ocfs2_queue_recovery_completion(osb->journal, i, NULL, NULL,
+ NULL);
+ /*
+ * We queued a recovery on orphan slots, increment the sequence
+ * number and update LVB so other node will skip the scan for a while
+ */
+ seqno++;
+unlock:
+ ocfs2_orphan_scan_unlock(osb, seqno, DLM_LOCK_EX);
+out:
+ return;
+}
+
+/* Worker task that gets fired every ORPHAN_SCAN_SCHEDULE_TIMEOUT millsec */
+void ocfs2_orphan_scan_work(struct work_struct *work)
+{
+ struct ocfs2_orphan_scan *os;
+ struct ocfs2_super *osb;
+
+ os = container_of(work, struct ocfs2_orphan_scan,
+ os_orphan_scan_work.work);
+ osb = os->os_osb;
+
+ mutex_lock(&os->os_lock);
+ ocfs2_queue_orphan_scan(osb);
+ schedule_delayed_work(&os->os_orphan_scan_work,
+ ocfs2_orphan_scan_timeout());
+ mutex_unlock(&os->os_lock);
+}
+
+void ocfs2_orphan_scan_stop(struct ocfs2_super *osb)
+{
+ struct ocfs2_orphan_scan *os;
+
+ os = &osb->osb_orphan_scan;
+ mutex_lock(&os->os_lock);
+ cancel_delayed_work(&os->os_orphan_scan_work);
+ mutex_unlock(&os->os_lock);
+}
+
+int ocfs2_orphan_scan_init(struct ocfs2_super *osb)
+{
+ struct ocfs2_orphan_scan *os;
+
+ os = &osb->osb_orphan_scan;
+ os->os_osb = osb;
+ mutex_init(&os->os_lock);
+
+ INIT_DELAYED_WORK(&os->os_orphan_scan_work,
+ ocfs2_orphan_scan_work);
+ schedule_delayed_work(&os->os_orphan_scan_work,
+ ocfs2_orphan_scan_timeout());
+ return 0;
+}
+
struct ocfs2_orphan_filldir_priv {
struct inode *head;
struct ocfs2_super *osb;