LU-3573 osd-zfs: Only advance zap cursor as needed

Only advance the zap cursor when ozi_pos is not advanced, otherwise
occasionally the a file could get "lost" because the zap_cursor would
advance over it before the retrieve happened.

Signed-off-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
Change-Id: Iad560e2ffb4cfe2c74a1cf9197be7c2537538822
Reviewed-on: http://review.whamcloud.com/12582
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Isaac Huang <he.huang@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/osd-zfs/osd_index.c b/lustre/osd-zfs/osd_index.c
index 210a00e..8765df1 100644 (file)
--- a/lustre/osd-zfs/osd_index.c
+++ b/lustre/osd-zfs/osd_index.c
@@ -842,6 +842,8 @@ static int osd_dir_it_next(const struct lu_env *env, struct dt_it *di)
        zap_attribute_t   *za = &osd_oti_get(env)->oti_za;
        int                rc;
 
+       ENTRY;
+
        /* temp. storage should be enough for any key supported by ZFS */
        CLASSERT(sizeof(za->za_name) <= sizeof(it->ozi_name));
 
@@ -854,9 +856,10 @@ static int osd_dir_it_next(const struct lu_env *env, struct dt_it *di)
                it->ozi_pos++;
                if (it->ozi_pos <=2)
                        RETURN(0);
-       }
 
-       zap_cursor_advance(it->ozi_zc);
+       } else {
+               zap_cursor_advance(it->ozi_zc);
+       }
 
        /*
         * According to current API we need to return error if its last entry.