In failover test, it reports memory leak about tqpair->stats when
detaching a tcp controller and it failover to the other controller.
Because during resetting the controller, we disconnect the controller
at first and then reconnect. when disconnecting, the adminq is not
freed which means the corresponding tqpair and tqpair->stats are not
freed. But when reconnecting, nvme_tcp_ctrlr_connect_qpair will
allocate memory for tqpair->stats again which causes memory leak.
So this patch fix the bug by not reallocating memory for tqpair->stats
if it's not NULL. We keep the old stats because from user perspective,
the spdk_nvme_qpair is the same one.
Besides, when destroying a qpair, the qpair->poll_group is set as
NULL which means if qpair->poll_group is not NULL, it should be a
new qpair. So there's no need to check if stats is NULL or not if
qpair->poll_group is not NULL. So adjusting the if...else... in
_nvme_pcie_ctrlr_create_io_qpair.
Change-Id: I4108a980aeffe4797e5bca5b1a8ea89f7457162b
Signed-off-by: Richael Zhuang <richael.zhuang@arm.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/17718
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
df4600f4c9