On Tue, 2018-01-09 at 09:04 +0800, oulijun wrote:
> 在 2018/1/9 5:20, Doug Ledford 写道:
> > On Thu, 2018-01-04 at 12:19 +0800, Lijun Ou wrote:
> > > @@ -2342,7 +2366,7 @@ static void modify_qp_init_to_init(struct ib_qp *ibqp,
> > > V2_QPC_BYTE_80_RX_CQN_S, 0);
> > >
> > > roce_set_field(context->byte_252_err_txcqn, V2_QPC_BYTE_252_TX_CQN_M,
> > > - V2_QPC_BYTE_252_TX_CQN_S, to_hr_cq(ibqp->recv_cq)->cqn);
> > > + V2_QPC_BYTE_252_TX_CQN_S, to_hr_cq(ibqp->send_cq)->cqn);
> >
> > This looks like a bugfix unrelated to the rest of the patch.
> >
>
> Sure, This is found for debugging CM and The other modification of qp context in
> this patch is unified for CM. As a result, I put it into the patch-set of CM.
>
> Do I need to send PATCHv2?
That depends. What's the effect of this bug? Is it something that
should be sent to stable? If the common case is that the send and recv
cq sizes are the same, and this bug is mostly never an issue, then no,
no v2 is necessary. If this is something we should send to stable, then
yes, pull out the bugfix, tag it for stable, and submit v2.
--
Doug Ledford <dledford@xxxxxxxxxx>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDDAttachment:
signature.asc
Description: This is a digitally signed message part
