在 2018/12/9 21:28, Yuval Shaia 写道:
> Sorry, please ignore my last comment but pls chk question and comment here.
>
> On Sun, Dec 09, 2018 at 03:18:27PM +0200, Yuval Shaia wrote:
>> On Sat, Dec 08, 2018 at 06:40:08PM +0800, Lijun Ou wrote:
>>> This patch move the codes of qp state transition into the
>>> new function as well as simplify the logic for other
>>> qp states transition.
> I don't see code that do "state transition" in this patch.
The legal qp state jump. for example, the transition from a state of qp to either state.
>>> Signed-off-by: Lijun Ou <oulijun@xxxxxxxxxx>
>>> ---
>>> V1->V2:
>>> 1. Use inline function instead of macro definition
>>> 2. Simplify the qp state transition logic
>>> ---
>>> drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 31 +++++++++++++++---------------
>>> 1 file changed, 16 insertions(+), 15 deletions(-)
>>>
>>> diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
>>> index 1cffe74..ca71e10 100644
>>> --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
>>> +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
>>> @@ -3670,6 +3670,21 @@ static int modify_qp_rtr_to_rts(struct ib_qp *ibqp,
>>> return 0;
>>> }
>>>
>>> +static inline bool hns_roce_v2_check_qp_stat(enum ib_qp_state cur_state,
>>> + enum ib_qp_state new_state)
>>> +{
>>> +
>>> + if ((cur_state != IB_QPS_RESET &&
>>> + (new_state == IB_QPS_ERR || new_state == IB_QPS_RESET)) ||
>>> + ((cur_state == IB_QPS_RTS || cur_state == IB_QPS_SQD) &&
>>> + (new_state == IB_QPS_RTS || new_state == IB_QPS_SQD)) ||
>>> + (cur_state == IB_QPS_SQE && new_state == IB_QPS_RTS))
>>> + return true;
>>> +
>>> + return false;
>>> +
>>> +}
>>> +
>>> static int hns_roce_v2_modify_qp(struct ib_qp *ibqp,
>>> const struct ib_qp_attr *attr,
>>> int attr_mask, enum ib_qp_state cur_state,
>>> @@ -3711,21 +3726,7 @@ static int hns_roce_v2_modify_qp(struct ib_qp *ibqp,
>>> qpc_mask);
>>> if (ret)
>>> goto out;
>>> - } else if ((cur_state == IB_QPS_RTS && new_state == IB_QPS_RTS) ||
>>> - (cur_state == IB_QPS_SQE && new_state == IB_QPS_RTS) ||
>>> - (cur_state == IB_QPS_RTS && new_state == IB_QPS_SQD) ||
>>> - (cur_state == IB_QPS_SQD && new_state == IB_QPS_SQD) ||
>>> - (cur_state == IB_QPS_SQD && new_state == IB_QPS_RTS) ||
>>> - (cur_state == IB_QPS_INIT && new_state == IB_QPS_RESET) ||
>>> - (cur_state == IB_QPS_RTR && new_state == IB_QPS_RESET) ||
>>> - (cur_state == IB_QPS_RTS && new_state == IB_QPS_RESET) ||
>>> - (cur_state == IB_QPS_ERR && new_state == IB_QPS_RESET) ||
>>> - (cur_state == IB_QPS_INIT && new_state == IB_QPS_ERR) ||
>>> - (cur_state == IB_QPS_RTR && new_state == IB_QPS_ERR) ||
>>> - (cur_state == IB_QPS_RTS && new_state == IB_QPS_ERR) ||
>>> - (cur_state == IB_QPS_SQD && new_state == IB_QPS_ERR) ||
>>> - (cur_state == IB_QPS_SQE && new_state == IB_QPS_ERR) ||
>>> - (cur_state == IB_QPS_ERR && new_state == IB_QPS_ERR)) {
>>> + } else if (hns_roce_v2_check_qp_stat(cur_state, new_state)) {
>>> /* Nothing */
>> If doing nothing then suggesting to make hns_roce_v2_check_qp_stat return
>> void.
> Ignore the above but, since the return value from hns_roce_v2_check_qp_stat
> is ignored then suggesting to remove the 'if'.
>
>>> ;
>>> } else {
>>> --
>>> 1.9.1
>>>
> .
>
