LU-3915 osd-ldiskfs: don't assert on possible upgrade

There are some cases possible during upgrade to the Lustre 2.5
from older versions which need workaround instead assertion.
The osd_object_ref_del() shouldn't assert on nlink == 0 if
object is local file but produce error message.

The underlying ext4_unlink() code doesn't treat decref on an
inode with nlink == 0 as a fatal error either, so only print
an error in the non-upgrade case for regular objects.

Signed-off-by: Mikhail Pershin <mike.pershin@intel.com>
Change-Id: I5aa1cc3d9a6d6210e8ce9f55193a4f25507b0986
Reviewed-on: http://review.whamcloud.com/7673
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: James Simmons <uja.ornl@gmail.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/osd-ldiskfs/osd_handler.c b/lustre/osd-ldiskfs/osd_handler.c
index d657021..7a6815a 100644 (file)
--- a/lustre/osd-ldiskfs/osd_handler.c
+++ b/lustre/osd-ldiskfs/osd_handler.c
@@ -2654,18 +2654,29 @@ static int osd_declare_object_ref_del(const struct lu_env *env,
 static int osd_object_ref_del(const struct lu_env *env, struct dt_object *dt,
                               struct thandle *th)
 {
-        struct osd_object *obj = osd_dt_obj(dt);
-        struct inode      *inode = obj->oo_inode;
+       struct osd_object       *obj = osd_dt_obj(dt);
+       struct inode            *inode = obj->oo_inode;
+       struct osd_device       *osd = osd_dev(dt->do_lu.lo_dev);
 
-        LINVRNT(osd_invariant(obj));
+       LINVRNT(osd_invariant(obj));
        LASSERT(dt_object_exists(dt) && !dt_object_remote(dt));
-        LASSERT(osd_write_locked(env, obj));
-        LASSERT(th != NULL);
+       LASSERT(osd_write_locked(env, obj));
+       LASSERT(th != NULL);
 
        osd_trans_exec_op(env, th, OSD_OT_REF_DEL);
 
        spin_lock(&obj->oo_guard);
-       LASSERT(inode->i_nlink > 0);
+       /* That can be result of upgrade from old Lustre version and
+        * applied only to local files.  Just skip this ref_del call.
+        * ext4_unlink() only treats this as a warning, don't LASSERT here.*/
+       if (inode->i_nlink == 0) {
+               CDEBUG_LIMIT(fid_is_norm(lu_object_fid(&dt->do_lu)) ?
+                            D_ERROR : D_INODE, "%s: nlink == 0 on "DFID
+                            ", maybe an upgraded file? (LU-3915)\n",
+                            osd_name(osd), PFID(lu_object_fid(&dt->do_lu)));
+               spin_unlock(&obj->oo_guard);
+               return 0;
+       }
 
        /* This based on ldiskfs_dec_count(), which is not exported.
         *
@@ -2674,7 +2685,7 @@ static int osd_object_ref_del(const struct lu_env *env, struct dt_object *dt,
         * threads not holding oo_guard seeing i_nlink == 0 in rare cases.
         *
         * nlink == 1 means the directory has/had > EXT4_LINK_MAX subdirs.
-        * */
+        */
        if (!S_ISDIR(inode->i_mode) || inode->i_nlink > 1) {
                drop_nlink(inode);