Comment 2 for bug 1233301

 lock_mutex_enter is held in lock_clust_rec_read_check_and_lock

after which it calls lock_rec_lock (as per call trace), eventually leading to innobase_kill_connection where it again calls lock_mutex_enter leading to a deadlock.

Quick bt:

pthread_cond_wait@@GLIBC_2.3.2
os_cond_wait
os_event_wait_low
sync_array_wait_event
mutex_spin_wait
mutex_enter_func
pfs_mutex_enter_func
innobase_kill_connection --------------------------------------------------------------------------> lock_mutex_enter deadlocks here again.
kill_handlerton
plugin_foreach_with_mask
ha_kill_connection
THD::awake
wsrep_thd_awake
wsrep_innobase_kill_one_trx
wsrep_kill_victim
lock_rec_other_has_conflicting
lock_rec_lock_slow
lock_rec_lock
lock_clust_rec_read_check_and_lock -------------------------------------------------> lock_mutex_enter
sel_set_rec_lock
row_search_for_mysql
ha_innobase::index_read
handler::index_read_map
handler::ha_index_read_map
handler::read_range_first
handler::multi_range_read_next
DsMrr_impl::dsmrr_next
ha_innobase::multi_range_read_next
QUICK_RANGE_SELECT::get_next
rr_quick
mysql_update
mysql_execute_command
mysql_parse
Query_log_event::do_apply_event
Query_log_event::do_apply_event
Log_event::apply_event
wsrep_apply_rbr
wsrep_apply_cb
galera::TrxHandle::apply
apply_trx_ws
galera::ReplicatorSMM::apply_trx
galera::ReplicatorSMM::process_trx
galera::GcsActionSource::dispatch
galera::GcsActionSource::process
galera::ReplicatorSMM::async_recv
galera_recv
wsrep_replication_process
start_wsrep_THD
start_thread
clone