Re: [PATCH v1 for-rc] RDMA/vmw_pvrdma: Return the correct opcode when creating WR

[Adit Ranadive <aditr@vmware.com>]

On 1/7/19 3:28 PM, Jason Gunthorpe wrote:
> On Mon, Jan 07, 2019 at 11:23:05PM +0000, Adit Ranadive wrote:
>> From: Adit Ranadive <aditr@vmware.com>
>>
>> Since the IB_WR_REG_MR opcode value changed, let's set the PVRDMA
>> device opcodes explicitly.
>>
>> Reported-by: Ruishuang Wang <ruishuangw@vmware.com>
>> Fixes: 9a59739bd01f ("IB/rxe: Revise the ib_wr_opcode enum")
>> Cc: stable@vger.kernel.org
>> Reviewed-by: Bryan Tan <bryantan@vmware.com>
>> Reviewed-by: Ruishuang Wang <ruishuangw@vmware.com>
>> Reviewed-by: Vishnu Dasa <vdasa@vmware.com>
>> Signed-off-by: Adit Ranadive <aditr@vmware.com>
>>  drivers/infiniband/hw/vmw_pvrdma/pvrdma.h | 43 +++++++++++++++++++++++++++++++
>>  1 file changed, 43 insertions(+)
>>
>> diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h
>> index 42b8685c997e..a85d7ba96d55 100644
>> +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h
>> @@ -427,6 +427,49 @@ static inline enum ib_qp_state pvrdma_qp_state_to_ib(enum pvrdma_qp_state state)
>>  
>>  static inline enum pvrdma_wr_opcode ib_wr_opcode_to_pvrdma(enum ib_wr_opcode op)
>>  {
>> +	switch (op) {
>> +	case IB_WR_RDMA_WRITE:
>> +		return PVRDMA_WR_RDMA_WRITE;
>> +	case IB_WR_RDMA_WRITE_WITH_IMM:
>> +		return PVRDMA_WR_RDMA_WRITE_WITH_IMM;
>> +	case IB_WR_SEND:
>> +		return PVRDMA_WR_SEND;
>> +	case IB_WR_SEND_WITH_IMM:
>> +		return PVRDMA_WR_SEND_WITH_IMM;
>> +	case IB_WR_RDMA_READ:
>> +		return PVRDMA_WR_RDMA_READ;
>> +	case IB_WR_ATOMIC_CMP_AND_SWP:
>> +		return PVRDMA_WR_ATOMIC_CMP_AND_SWP;
>> +	case IB_WR_ATOMIC_FETCH_AND_ADD:
>> +		return PVRDMA_WR_ATOMIC_FETCH_AND_ADD;
>> +	case IB_WR_LSO:
>> +		return PVRDMA_WR_LSO;
>> +	case IB_WR_SEND_WITH_INV:
>> +		return PVRDMA_WR_SEND_WITH_INV;
>> +	case IB_WR_RDMA_READ_WITH_INV:
>> +		return PVRDMA_WR_RDMA_READ_WITH_INV;
>> +	case IB_WR_LOCAL_INV:
>> +		return PVRDMA_WR_LOCAL_INV;
>> +	case IB_WR_REG_MR:
>> +		return PVRDMA_WR_FAST_REG_MR;
>> +	case IB_WR_MASKED_ATOMIC_CMP_AND_SWP:
>> +		return PVRDMA_WR_MASKED_ATOMIC_CMP_AND_SWP;
>> +	case IB_WR_MASKED_ATOMIC_FETCH_AND_ADD:
>> +		return PVRDMA_WR_MASKED_ATOMIC_FETCH_AND_ADD;
>> +	case IB_WR_REG_SIG_MR:
>> +		return PVRDMA_WR_REG_SIG_MR;
>> +	case IB_WR_RESERVED1:
>> +	case IB_WR_RESERVED2:
>> +	case IB_WR_RESERVED3:
>> +	case IB_WR_RESERVED4:
>> +	case IB_WR_RESERVED5:
>> +	case IB_WR_RESERVED6:
>> +	case IB_WR_RESERVED7:
>> +	case IB_WR_RESERVED8:
>> +	case IB_WR_RESERVED9:
>> +	case IB_WR_RESERVED10:
>> +		return (enum pvrdma_wr_opcode)op;
>> +	}
>>  	return (enum pvrdma_wr_opcode)op;
> 
> I ment it should always translate, not cast like this - if no
> translation is available then it should be a posting failure.
> 
> Jason
> 

Not sure if I follow you. We can post a failure but that would be 
in pvrdma_qp.c when it looks at the opcode returned from here. We 
would have to handle all the IB enums here though if you don't want 
a default case (you would get compiler warnings otherwise).  

diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
index 3acf74cbe266..eeb9e9ef4655 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
@@ -721,6 +721,12 @@ int pvrdma_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
                    wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM)
                        wqe_hdr->ex.imm_data = wr->ex.imm_data;

+               if (wqe_hdr->opcode > PVRDMA_WR_REG_SIG_MR) {
+                       *bad_wr = wr;
+                       ret = -EINVAL;
+                       goto out;
+               }
+
                switch (qp->ibqp.qp_type) {
                case IB_QPT_GSI:
                case IB_QPT_UD: