Comment 14 for bug 1128848

a)There seems to be something wrong with backtrace in #8

#5 0x00000000008c02da in fil_delete_tablespace (id=484, evict_all=1) at /home/jenkins/workspace/percona-server-5.5-nigthly-binaries/label_exp/centos5-64/Percona-Server-5.5.28-rel29.3/storage/innobase/fil/fil0fil.c:2415
#6 0x00000000008c036d in fil_discard_tablespace (id=29462) at /home/jenkins/workspace/percona-server-5.5-nigthly-binaries/label_exp/centos5-64/Percona-Server-5.5.28-rel29.3/storage/innobase/fil/fil0fil.c:2500

The space id is 29462 in fil_discard_tablespace but in
fil_delete_tablespace shows 484.

b) In fil_delete_tablespace, it already checks for
node->n_pending in a loop but inside a fil_system->mutex which it
later drops before reacquiring. Is it not possible that during
this gap, node receives additional I/O? (from fil_node_prepare_for_io)

c) I also see fil_flush_file_spaces waiting on fil_system->mutex
(with purpose of FIL_TABLESPACE), can this be the pending I/O
waiting on the node? (hence blocked by fil_system->mutex acquired
again)