RDMA/hns: Fix bug during CMDQ initialization

mainline inclusion
from mainline-v5.13
commit af06b628
category: bugfix
bugzilla: NA
CVE: NA

When re-insmod hns-roce-hw-v2.ko and init CMDQ, we need to reinit PI and CI of CMDQ.
But, after last rmmod hns-roce-hw-v2.ko, the CMDQ is NOT reset, firmware still is waiting software to send cmd..
So,  if we re-init PI first, the firmware would thought that the software sent new cmds. then it start to process CMDQ.
It may be a long time for driver. If driver start to init hardware and really send cmds, the status of PI and CI are unexpected.
Signed-off-by: NShunfeng Yang <yangshunfeng2@huawei.com>
Signed-off-by: NYangyang Li <liyangyang20@huawei.com>
Reviewed-by: Nchunzhi hu <huchunzhi@huawei.com>
Signed-off-by: NYang Yingliang <yangyingliang@huawei.com>

drivers/infiniband/hw/hns/hns_roce_hw_v2.c

@@ -1039,8 +1039,10 @@ static void hns_roce_cmq_init_regs(struct hns_roce_dev *hr_dev, bool ring_type)
 			   upper_32_bits(dma));
 		roce_write(hr_dev, ROCEE_TX_CMQ_DEPTH_REG,
 			   ring->desc_num >> HNS_ROCE_CMQ_DESC_NUM_S);
-		roce_write(hr_dev, ROCEE_TX_CMQ_HEAD_REG, 0);
+
+		/* Make sure to write tail first and then head */
 		roce_write(hr_dev, ROCEE_TX_CMQ_TAIL_REG, 0);
+		roce_write(hr_dev, ROCEE_TX_CMQ_HEAD_REG, 0);
 	} else {
 		roce_write(hr_dev, ROCEE_RX_CMQ_BASEADDR_L_REG, (u32)dma);
 		roce_write(hr_dev, ROCEE_RX_CMQ_BASEADDR_H_REG,