ocfs2: add orphan recovery types in ocfs2_recover_orphans
authorJoseph Qi <joseph.qi@huawei.com>
Mon, 16 Feb 2015 23:59:57 +0000 (15:59 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 17 Feb 2015 01:56:04 +0000 (17:56 -0800)
Define two orphan recovery types, which indicates if need truncate file or
not.

Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Weiwei Wang <wangww631@huawei.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Xuejiufei <xuejiufei@huawei.com>
Cc: alex chen <alex.chen@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/journal.c
fs/ocfs2/ocfs2.h

index d10860f..9730f53 100644 (file)
@@ -50,6 +50,8 @@
 #include "sysfile.h"
 #include "uptodate.h"
 #include "quota.h"
+#include "file.h"
+#include "namei.h"
 
 #include "buffer_head_io.h"
 #include "ocfs2_trace.h"
@@ -69,13 +71,15 @@ static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb,
 static int ocfs2_trylock_journal(struct ocfs2_super *osb,
                                 int slot_num);
 static int ocfs2_recover_orphans(struct ocfs2_super *osb,
-                                int slot);
+                                int slot,
+                                enum ocfs2_orphan_reco_type orphan_reco_type);
 static int ocfs2_commit_thread(void *arg);
 static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal,
                                            int slot_num,
                                            struct ocfs2_dinode *la_dinode,
                                            struct ocfs2_dinode *tl_dinode,
-                                           struct ocfs2_quota_recovery *qrec);
+                                           struct ocfs2_quota_recovery *qrec,
+                                           enum ocfs2_orphan_reco_type orphan_reco_type);
 
 static inline int ocfs2_wait_on_mount(struct ocfs2_super *osb)
 {
@@ -149,7 +153,8 @@ int ocfs2_compute_replay_slots(struct ocfs2_super *osb)
        return 0;
 }
 
-void ocfs2_queue_replay_slots(struct ocfs2_super *osb)
+void ocfs2_queue_replay_slots(struct ocfs2_super *osb,
+               enum ocfs2_orphan_reco_type orphan_reco_type)
 {
        struct ocfs2_replay_map *replay_map = osb->replay_map;
        int i;
@@ -163,7 +168,8 @@ void ocfs2_queue_replay_slots(struct ocfs2_super *osb)
        for (i = 0; i < replay_map->rm_slots; i++)
                if (replay_map->rm_replay_slots[i])
                        ocfs2_queue_recovery_completion(osb->journal, i, NULL,
-                                                       NULL, NULL);
+                                                       NULL, NULL,
+                                                       orphan_reco_type);
        replay_map->rm_state = REPLAY_DONE;
 }
 
@@ -1174,6 +1180,7 @@ struct ocfs2_la_recovery_item {
        struct ocfs2_dinode     *lri_la_dinode;
        struct ocfs2_dinode     *lri_tl_dinode;
        struct ocfs2_quota_recovery *lri_qrec;
+       enum ocfs2_orphan_reco_type  lri_orphan_reco_type;
 };
 
 /* Does the second half of the recovery process. By this point, the
@@ -1195,6 +1202,7 @@ void ocfs2_complete_recovery(struct work_struct *work)
        struct ocfs2_dinode *la_dinode, *tl_dinode;
        struct ocfs2_la_recovery_item *item, *n;
        struct ocfs2_quota_recovery *qrec;
+       enum ocfs2_orphan_reco_type orphan_reco_type;
        LIST_HEAD(tmp_la_list);
 
        trace_ocfs2_complete_recovery(
@@ -1212,6 +1220,7 @@ void ocfs2_complete_recovery(struct work_struct *work)
                la_dinode = item->lri_la_dinode;
                tl_dinode = item->lri_tl_dinode;
                qrec = item->lri_qrec;
+               orphan_reco_type = item->lri_orphan_reco_type;
 
                trace_ocfs2_complete_recovery_slot(item->lri_slot,
                        la_dinode ? le64_to_cpu(la_dinode->i_blkno) : 0,
@@ -1236,7 +1245,8 @@ void ocfs2_complete_recovery(struct work_struct *work)
                        kfree(tl_dinode);
                }
 
-               ret = ocfs2_recover_orphans(osb, item->lri_slot);
+               ret = ocfs2_recover_orphans(osb, item->lri_slot,
+                               orphan_reco_type);
                if (ret < 0)
                        mlog_errno(ret);
 
@@ -1261,7 +1271,8 @@ static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal,
                                            int slot_num,
                                            struct ocfs2_dinode *la_dinode,
                                            struct ocfs2_dinode *tl_dinode,
-                                           struct ocfs2_quota_recovery *qrec)
+                                           struct ocfs2_quota_recovery *qrec,
+                                           enum ocfs2_orphan_reco_type orphan_reco_type)
 {
        struct ocfs2_la_recovery_item *item;
 
@@ -1285,6 +1296,7 @@ static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal,
        item->lri_slot = slot_num;
        item->lri_tl_dinode = tl_dinode;
        item->lri_qrec = qrec;
+       item->lri_orphan_reco_type = orphan_reco_type;
 
        spin_lock(&journal->j_lock);
        list_add_tail(&item->lri_list, &journal->j_la_cleanups);
@@ -1304,7 +1316,8 @@ void ocfs2_complete_mount_recovery(struct ocfs2_super *osb)
        /* No need to queue up our truncate_log as regular cleanup will catch
         * that */
        ocfs2_queue_recovery_completion(journal, osb->slot_num,
-                                       osb->local_alloc_copy, NULL, NULL);
+                                       osb->local_alloc_copy, NULL, NULL,
+                                       ORPHAN_NEED_TRUNCATE);
        ocfs2_schedule_truncate_log_flush(osb, 0);
 
        osb->local_alloc_copy = NULL;
@@ -1312,7 +1325,7 @@ void ocfs2_complete_mount_recovery(struct ocfs2_super *osb)
 
        /* queue to recover orphan slots for all offline slots */
        ocfs2_replay_map_set_state(osb, REPLAY_NEEDED);
-       ocfs2_queue_replay_slots(osb);
+       ocfs2_queue_replay_slots(osb, ORPHAN_NEED_TRUNCATE);
        ocfs2_free_replay_slots(osb);
 }
 
@@ -1323,7 +1336,8 @@ void ocfs2_complete_quota_recovery(struct ocfs2_super *osb)
                                                osb->slot_num,
                                                NULL,
                                                NULL,
-                                               osb->quota_rec);
+                                               osb->quota_rec,
+                                               ORPHAN_NEED_TRUNCATE);
                osb->quota_rec = NULL;
        }
 }
@@ -1360,7 +1374,7 @@ restart:
 
        /* queue recovery for our own slot */
        ocfs2_queue_recovery_completion(osb->journal, osb->slot_num, NULL,
-                                       NULL, NULL);
+                                       NULL, NULL, ORPHAN_NO_NEED_TRUNCATE);
 
        spin_lock(&osb->osb_lock);
        while (rm->rm_used) {
@@ -1419,13 +1433,14 @@ skip_recovery:
                        continue;
                }
                ocfs2_queue_recovery_completion(osb->journal, rm_quota[i],
-                                               NULL, NULL, qrec);
+                                               NULL, NULL, qrec,
+                                               ORPHAN_NEED_TRUNCATE);
        }
 
        ocfs2_super_unlock(osb, 1);
 
        /* queue recovery for offline slots */
-       ocfs2_queue_replay_slots(osb);
+       ocfs2_queue_replay_slots(osb, ORPHAN_NEED_TRUNCATE);
 
 bail:
        mutex_lock(&osb->recovery_lock);
@@ -1711,7 +1726,7 @@ static int ocfs2_recover_node(struct ocfs2_super *osb,
 
        /* This will kfree the memory pointed to by la_copy and tl_copy */
        ocfs2_queue_recovery_completion(osb->journal, slot_num, la_copy,
-                                       tl_copy, NULL);
+                                       tl_copy, NULL, ORPHAN_NEED_TRUNCATE);
 
        status = 0;
 done:
@@ -1901,7 +1916,7 @@ void ocfs2_queue_orphan_scan(struct ocfs2_super *osb)
 
        for (i = 0; i < osb->max_slots; i++)
                ocfs2_queue_recovery_completion(osb->journal, i, NULL, NULL,
-                                               NULL);
+                                               NULL, ORPHAN_NO_NEED_TRUNCATE);
        /*
         * We queued a recovery on orphan slots, increment the sequence
         * number and update LVB so other node will skip the scan for a while
@@ -2000,6 +2015,13 @@ static int ocfs2_orphan_filldir(struct dir_context *ctx, const char *name,
        if (IS_ERR(iter))
                return 0;
 
+       /* Skip inodes which are already added to recover list, since dio may
+        * happen concurrently with unlink/rename */
+       if (OCFS2_I(iter)->ip_next_orphan) {
+               iput(iter);
+               return 0;
+       }
+
        trace_ocfs2_orphan_filldir((unsigned long long)OCFS2_I(iter)->ip_blkno);
        /* No locking is required for the next_orphan queue as there
         * is only ever a single process doing orphan recovery. */
@@ -2108,7 +2130,8 @@ static void ocfs2_clear_recovering_orphan_dir(struct ocfs2_super *osb,
  *   advertising our state to ocfs2_delete_inode().
  */
 static int ocfs2_recover_orphans(struct ocfs2_super *osb,
-                                int slot)
+                                int slot,
+                                enum ocfs2_orphan_reco_type orphan_reco_type)
 {
        int ret = 0;
        struct inode *inode = NULL;
@@ -2132,13 +2155,58 @@ static int ocfs2_recover_orphans(struct ocfs2_super *osb,
                                        (unsigned long long)oi->ip_blkno);
 
                iter = oi->ip_next_orphan;
+               oi->ip_next_orphan = NULL;
+
+               /*
+                * We need to take and drop the inode lock to
+                * force read inode from disk.
+                */
+               ret = ocfs2_inode_lock(inode, NULL, 0);
+               if (ret) {
+                       mlog_errno(ret);
+                       goto next;
+               }
+               ocfs2_inode_unlock(inode, 0);
+
+               if (inode->i_nlink == 0) {
+                       spin_lock(&oi->ip_lock);
+                       /* Set the proper information to get us going into
+                        * ocfs2_delete_inode. */
+                       oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED;
+                       spin_unlock(&oi->ip_lock);
+               } else if (orphan_reco_type == ORPHAN_NEED_TRUNCATE) {
+                       struct buffer_head *di_bh = NULL;
+
+                       ret = ocfs2_rw_lock(inode, 1);
+                       if (ret) {
+                               mlog_errno(ret);
+                               goto next;
+                       }
 
-               spin_lock(&oi->ip_lock);
-               /* Set the proper information to get us going into
-                * ocfs2_delete_inode. */
-               oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED;
-               spin_unlock(&oi->ip_lock);
+                       ret = ocfs2_inode_lock(inode, &di_bh, 1);
+                       if (ret < 0) {
+                               ocfs2_rw_unlock(inode, 1);
+                               mlog_errno(ret);
+                               goto next;
+                       }
+
+                       ret = ocfs2_truncate_file(inode, di_bh,
+                                       i_size_read(inode));
+                       ocfs2_inode_unlock(inode, 1);
+                       ocfs2_rw_unlock(inode, 1);
+                       brelse(di_bh);
+                       if (ret < 0) {
+                               if (ret != -ENOSPC)
+                                       mlog_errno(ret);
+                               goto next;
+                       }
+
+                       ret = ocfs2_del_inode_from_orphan(osb, inode, 0, 0);
+                       if (ret)
+                               mlog_errno(ret);
+               } /* else if ORPHAN_NO_NEED_TRUNCATE, do nothing */
 
+next:
                iput(inode);
 
                inode = iter;
index fdbcbfe..df9a95c 100644 (file)
@@ -209,6 +209,11 @@ struct ocfs2_lock_res {
 #endif
 };
 
+enum ocfs2_orphan_reco_type {
+       ORPHAN_NO_NEED_TRUNCATE = 0,
+       ORPHAN_NEED_TRUNCATE,
+};
+
 enum ocfs2_orphan_scan_state {
        ORPHAN_SCAN_ACTIVE,
        ORPHAN_SCAN_INACTIVE