OSDN Git Service

ocfs2: Use s_umount for quota recovery protection
authorJan Kara <jack@suse.cz>
Wed, 23 Nov 2016 13:35:26 +0000 (14:35 +0100)
committerJan Kara <jack@suse.cz>
Wed, 30 Nov 2016 07:37:21 +0000 (08:37 +0100)
Currently we use dqonoff_mutex to serialize quota recovery protection
and turning of quotas on / off. Use s_umount semaphore instead.

Tested-by: Eric Ren <zren@suse.com>
Signed-off-by: Jan Kara <jack@suse.cz>
fs/ocfs2/quota_local.c

index 8a54fd8..32c5a40 100644 (file)
@@ -454,7 +454,7 @@ out:
 /* Sync changes in local quota file into global quota file and
  * reinitialize local quota file.
  * The function expects local quota file to be already locked and
- * dqonoff_mutex locked. */
+ * s_umount locked in shared mode. */
 static int ocfs2_recover_local_quota_file(struct inode *lqinode,
                                          int type,
                                          struct ocfs2_quota_recovery *rec)
@@ -597,7 +597,7 @@ int ocfs2_finish_quota_recovery(struct ocfs2_super *osb,
        printk(KERN_NOTICE "ocfs2: Finishing quota recovery on device (%s) for "
               "slot %u\n", osb->dev_str, slot_num);
 
-       mutex_lock(&sb_dqopt(sb)->dqonoff_mutex);
+       down_read(&sb->s_umount);
        for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
                if (list_empty(&(rec->r_list[type])))
                        continue;
@@ -674,7 +674,7 @@ out_put:
                        break;
        }
 out:
-       mutex_unlock(&sb_dqopt(sb)->dqonoff_mutex);
+       up_read(&sb->s_umount);
        kfree(rec);
        return status;
 }
@@ -840,7 +840,10 @@ static int ocfs2_local_free_info(struct super_block *sb, int type)
        }
        ocfs2_release_local_quota_bitmaps(&oinfo->dqi_chunk);
 
-       /* dqonoff_mutex protects us against racing with recovery thread... */
+       /*
+        * s_umount held in exclusive mode protects us against racing with
+        * recovery thread...
+        */
        if (oinfo->dqi_rec) {
                ocfs2_free_quota_recovery(oinfo->dqi_rec);
                mark_clean = 0;