Description of problem:
Created a replicate volume and mounted it via fuse and nfs clients. Started running some tests (sanity script) om the fuse mount and added 2 more bricks (thus making the volume 2x2 distributed replicate), and gave rebalance. gluster-rebalance process crashed with the following backtrace.


Program terminated with signal 6, Aborted.
#0  0x00000034c2232905 in raise () from /lib64/libc.so.6
Missing separate debuginfos, use: debuginfo-install glibc-2.12-1.25.el6_1.3.x86_64 libgcc-4.4.5-6.el6.x86_64
(gdb) bt
#0  0x00000034c2232905 in raise () from /lib64/libc.so.6
#1  0x00000034c22340e5 in abort () from /lib64/libc.so.6
#2  0x00000034c222b9be in __assert_fail_base () from /lib64/libc.so.6
#3  0x00000034c222ba80 in __assert_fail () from /lib64/libc.so.6
#4  0x00007fab81cafa21 in client3_1_entrylk (frame=0x7fab84e37c58, this=0x10fd200, data=0x7fffa77095e0)
    at ../../../../../xlators/protocol/client/src/client3_1-fops.c:4763
#5  0x00007fab81c95b79 in client_entrylk (frame=0x7fab84e37c58, this=0x10fd200, volume=0x10fffc0 "mirror-replicate-1", loc=0x7fab801f2e50, 
    basename=0x11ba2d0 "coverage", cmd=ENTRYLK_LOCK_NB, type=ENTRYLK_WRLCK) at ../../../../../xlators/protocol/client/src/client.c:1665
#6  0x00007fab81a539a7 in afr_nonblocking_entrylk (frame=0x7fab84c2a028, this=0x1100820)
    at ../../../../../xlators/cluster/afr/src/afr-lk-common.c:1300
#7  0x00007fab81a31d59 in afr_lock_rec (frame=0x7fab84c2a028, this=0x1100820)
    at ../../../../../xlators/cluster/afr/src/afr-transaction.c:1156
#8  0x00007fab81a31dc6 in afr_lock (frame=0x7fab84c2a028, this=0x1100820) at ../../../../../xlators/cluster/afr/src/afr-transaction.c:1175
#9  0x00007fab81a3208a in afr_transaction (frame=0x7fab84c2a028, this=0x1100820, type=AFR_ENTRY_TRANSACTION)
    at ../../../../../xlators/cluster/afr/src/afr-transaction.c:1266
#10 0x00007fab81a10b37 in afr_mkdir (frame=0x7fab84e37bac, this=0x1100820, loc=0x7fab80713054, mode=16877, params=0x10f56e0)
    at ../../../../../xlators/cluster/afr/src/afr-dir-write.c:770
#11 0x00007fab817b3b7b in dht_selfheal_dir_mkdir (frame=0x7fab84e370ec, loc=0x7fab80713054, layout=0x11ba040, force=0)
    at ../../../../../xlators/cluster/dht/src/dht-selfheal.c:434
#12 0x00007fab817b4fbc in dht_selfheal_directory (frame=0x7fab84e370ec, dir_cbk=0x7fab817bddbb <dht_lookup_selfheal_cbk>, 
    loc=0x7fab80713054, layout=0x11ba040) at ../../../../../xlators/cluster/dht/src/dht-selfheal.c:855
#13 0x00007fab817c00d6 in dht_lookup_dir_cbk (frame=0x7fab84e370ec, cookie=0x7fab84e379a8, this=0x11014a0, op_ret=-1, op_errno=2, inode=0x0, 
    stbuf=0x7fab801ee6a4, xattr=0x0, postparent=0x7fab801ee714) at ../../../../../xlators/cluster/dht/src/dht-common.c:504
#14 0x00007fab81a6161e in afr_lookup_done (frame=0x7fab84e379a8, this=0x1100820) at ../../../../../xlators/cluster/afr/src/afr-common.c:1743
#15 0x00007fab81a61d20 in afr_lookup_cbk (frame=0x7fab84e379a8, cookie=0x1, this=0x1100820, op_ret=-1, op_errno=2, inode=0x7fab792ee174, 
    buf=0x7fffa7709cd0, xattr=0x0, postparent=0x7fffa7709c60) at ../../../../../xlators/cluster/afr/src/afr-common.c:1906
#16 0x00007fab81ca4367 in client3_1_lookup_cbk (req=0x7fab800d604c, iov=0x7fab800d608c, count=1, myframe=0x7fab84e37b00)
    at ../../../../../xlators/protocol/client/src/client3_1-fops.c:2185
#17 0x00007fab85ddb919 in rpc_clnt_handle_reply (clnt=0x11383d0, pollin=0x10f3980) at ../../../../rpc/rpc-lib/src/rpc-clnt.c:796
#18 0x00007fab85ddbcb6 in rpc_clnt_notify (trans=0x1147eb0, mydata=0x1138400, event=RPC_TRANSPORT_MSG_RECEIVED, data=0x10f3980)
    at ../../../../rpc/rpc-lib/src/rpc-clnt.c:915
#19 0x00007fab85dd7da8 in rpc_transport_notify (this=0x1147eb0, event=RPC_TRANSPORT_MSG_RECEIVED, data=0x10f3980)
    at ../../../../rpc/rpc-lib/src/rpc-transport.c:498
#20 0x00007fab82ae8270 in socket_event_poll_in (this=0x1147eb0) at ../../../../../rpc/rpc-transport/socket/src/socket.c:1686
#21 0x00007fab82ae87f4 in socket_event_handler (fd=11, idx=4, data=0x1147eb0, poll_in=1, poll_out=0, poll_err=0)
    at ../../../../../rpc/rpc-transport/socket/src/socket.c:1801
#22 0x00007fab86032030 in event_dispatch_epoll_handler (event_pool=0x10dac20, events=0x10f4870, i=0) at ../../../libglusterfs/src/event.c:794
#23 0x00007fab86032253 in event_dispatch_epoll (event_pool=0x10dac20) at ../../../libglusterfs/src/event.c:856
#24 0x00007fab860325de in event_dispatch (event_pool=0x10dac20) at ../../../libglusterfs/src/event.c:956
#25 0x0000000000407dcc in main (argc=19, argv=0x7fffa770a3c8) at ../../../glusterfsd/src/glusterfsd.c:1612
(gdb) f 4
#4  0x00007fab81cafa21 in client3_1_entrylk (frame=0x7fab84e37c58, this=0x10fd200, data=0x7fffa77095e0)
    at ../../../../../xlators/protocol/client/src/client3_1-fops.c:4763
4763            GF_ASSERT_AND_GOTO_WITH_ERROR (this->name,
(gdb) l
4758            if (!uuid_is_null (args->loc->inode->gfid))
4759                    memcpy (req.gfid,  args->loc->inode->gfid, 16);
4760            else
4761                    memcpy (req.gfid, args->loc->gfid, 16);
4762
4763            GF_ASSERT_AND_GOTO_WITH_ERROR (this->name,
4764                                           !uuid_is_null (*((uuid_t*)req.gfid)),
4765                                           unwind, op_errno, EINVAL);
4766            req.cmd = args->cmd_entrylk;
4767            req.type = args->type;
(gdb) p req.gfid
$1 = '\000' <repeats 15 times>
(gdb) p args->loc->gfid
$2 = '\000' <repeats 15 times>
(gdb) p args->loc->inode->gfid
$3 = '\000' <repeats 15 times>
(gdb) p *args->loc
$4 = {path = 0x11ba7b0 "/run19400", name = 0x0, inode = 0x7fab792ee0e0, parent = 0x0, gfid = '\000' <repeats 15 times>, 
  pargfid = '\000' <repeats 15 times>}
(gdb) f 10
#10 0x00007fab81a10b37 in afr_mkdir (frame=0x7fab84e37bac, this=0x1100820, loc=0x7fab80713054, mode=16877, params=0x10f56e0)
    at ../../../../../xlators/cluster/afr/src/afr-dir-write.c:770
770             afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
(gdb) p *loc
$5 = {path = 0x7fab6c01c570 "/run19400/coverage", name = 0x7fab6c01c57a "coverage", inode = 0x7fab792ee174, parent = 0x7fab792ee0e0, 
  gfid = "\347\327\323\377gfHۑ'U\256\232\216\264t", pargfid = "\320\066g\353\241\324A3\267\\G\375 f+~"}
(gdb) l afr_mkdir
712
713
714     int
715     afr_mkdir (call_frame_t *frame, xlator_t *this,
716                loc_t *loc, mode_t mode, dict_t *params)
717     {
718             afr_private_t * priv  = NULL;
719             afr_local_t   * local = NULL;
720             call_frame_t  * transaction_frame = NULL;
721             int ret = -1;
(gdb) 
722             int op_errno = 0;
723
724             VALIDATE_OR_GOTO (frame, out);
725             VALIDATE_OR_GOTO (this, out);
726             VALIDATE_OR_GOTO (this->private, out);
727
728             priv = this->private;
729
730             QUORUM_CHECK(mkdir,out);
731
(gdb) 
732             transaction_frame = copy_frame (frame);
733             if (!transaction_frame) {
734                     op_errno = ENOMEM;
735                     goto out;
736             }
737
738             AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
739             local = transaction_frame->local;
740
741             ret = afr_local_init (local, priv, &op_errno);
(gdb) 
742             if (ret < 0)
743                     goto out;
744
745             loc_copy (&local->loc, loc);
746
747             LOCK (&priv->read_child_lock);
748             {
749                     local->read_child_index = (++priv->read_child_rr)
750                             % (priv->child_count);
751             }
(gdb) 752             UNLOCK (&priv->read_child_lock);
753
754             local->cont.mkdir.mode  = mode;
755             if (params)
756                     local->cont.mkdir.params = dict_ref (params);
757
758             local->transaction.fop    = afr_mkdir_wind;
759             local->transaction.done   = afr_mkdir_done;
760             local->transaction.unwind = afr_mkdir_unwind;
761
(gdb) 
762             ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
763                                         &op_errno);
764             if (ret)
765                     goto out;
766
767             local->transaction.main_frame = frame;
768             local->transaction.basename = AFR_BASENAME (loc->path);
769
770             afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
 l afr_build_parent_loc
47      #include "afr.h"
48      #include "afr-transaction.h"
49
50      int
51      afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno)
52      {
53              int     ret = -1;
54              char    *child_path = NULL;
55
56              if (!child->parent) {
(gdb) 
57                      if (op_errno)
58                              *op_errno = EINVAL;
59                      goto out;
60              }
61
62              child_path = gf_strdup (child->path);
63              if (!child_path) {
64                      if (op_errno)
65                              *op_errno = ENOMEM;
66                      goto out;
(gdb) 
67              }
68              parent->path = dirname (child_path);
69              parent->inode  = inode_ref (child->parent);
70              ret = 0;
71      out:
72              return ret;
73      }
74
75      /* {{{ create */
76
(gdb) f 10
#10 0x00007fab81a10b37 in afr_mkdir (frame=0x7fab84e37bac, this=0x1100820, loc=0x7fab80713054, mode=16877, params=0x10f56e0)
    at ../../../../../xlators/cluster/afr/src/afr-dir-write.c:770
770             afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
(gdb) p local->transaction->parent_loc
$6 = {path = 0x11ba7b0 "/run19400", name = 0x0, inode = 0x7fab792ee0e0, parent = 0x0, gfid = '\000' <repeats 15 times>, 
  pargfid = '\000' <repeats 15 times>}
(gdb) 








Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1. create replicate volume, start it and mount it
2. start some tests on the mount point
3. add more bricks and give rebalance
  
Actual results:

rebalance process crashed

Expected results:

rebalance process should not crash

Additional info:

In rebalance after the lookup we are not linking the inode to inode table (since its done by either protocol/server for bricks and fuse for client and here the inode is created for rebalance purpose only), thus loc->inode and loc->parent will have their gfids NULL. 

Now afr_mkdir while building the parent_loc will not consider the loc->pargfid present in the child loc, hence keeping parent_loc->gfid NULL. Thus in protocol/client it looks for gfid either in loc->gfid or loc->inode->gfid and assert if it cannot find gfid in either of them.