| Message ID | 20220105080727.2143737-5-yanjun.zhu@linux.dev (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | Generate UDP src port with flow label or lqpn/rqpn | expand |
> Subject: [PATCH 4/4] RDMA/rxe: Use the standard method to produce udp source > port > > From: Zhu Yanjun <yanjun.zhu@linux.dev> > > Use the standard method to produce udp source port. > > Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev> > --- > drivers/infiniband/sw/rxe/rxe_verbs.c | 10 ++++++++++ > 1 file changed, 10 insertions(+) > > diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c > b/drivers/infiniband/sw/rxe/rxe_verbs.c > index 0aa0d7e52773..f30d98ad13cd 100644 > --- a/drivers/infiniband/sw/rxe/rxe_verbs.c > +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c > @@ -469,6 +469,16 @@ static int rxe_modify_qp(struct ib_qp *ibqp, struct > ib_qp_attr *attr, > if (err) > goto err1; > > + if (mask & IB_QP_AV) { > + if (attr->ah_attr.ah_flags & IB_AH_GRH) { > + u32 fl = attr->ah_attr.grh.flow_label; > + u32 lqp = qp->ibqp.qp_num; > + u32 rqp = qp->attr.dest_qp_num; > + Isn't the randomization for src_port done in rxe_qp_init_req redundant then? https://elixir.bootlin.com/linux/v5.16-rc8/source/drivers/infiniband/sw/rxe/rxe_qp.c#L220 Can we remove it? > + qp->src_port = rdma_get_udp_sport(fl, lqp, rqp); > + } > + } > + > return 0; > > err1: > -- > 2.27.0
在 2022/1/5 1:17, Saleem, Shiraz 写道: >> Subject: [PATCH 4/4] RDMA/rxe: Use the standard method to produce udp source >> port >> >> From: Zhu Yanjun <yanjun.zhu@linux.dev> >> >> Use the standard method to produce udp source port. >> >> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev> >> --- >> drivers/infiniband/sw/rxe/rxe_verbs.c | 10 ++++++++++ >> 1 file changed, 10 insertions(+) >> >> diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c >> b/drivers/infiniband/sw/rxe/rxe_verbs.c >> index 0aa0d7e52773..f30d98ad13cd 100644 >> --- a/drivers/infiniband/sw/rxe/rxe_verbs.c >> +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c >> @@ -469,6 +469,16 @@ static int rxe_modify_qp(struct ib_qp *ibqp, struct >> ib_qp_attr *attr, >> if (err) >> goto err1; >> >> + if (mask & IB_QP_AV) { >> + if (attr->ah_attr.ah_flags & IB_AH_GRH) { >> + u32 fl = attr->ah_attr.grh.flow_label; >> + u32 lqp = qp->ibqp.qp_num; >> + u32 rqp = qp->attr.dest_qp_num; >> + > Isn't the randomization for src_port done in rxe_qp_init_req redundant then? > > https://elixir.bootlin.com/linux/v5.16-rc8/source/drivers/infiniband/sw/rxe/rxe_qp.c#L220 > > Can we remove it? Yes. We can remove it. Because this "randomization for src_port done in rxe_qp_init_req" is replaced by rdma_get_udp_sport in rxe_modify_qp, I do not remove it. I will remove it in the latest commits soon. Zhu Yanjun > >> + qp->src_port = rdma_get_udp_sport(fl, lqp, rqp); >> + } >> + } >> + >> return 0; >> >> err1: >> -- >> 2.27.0 >
在 2022/1/5 8:33, Yanjun Zhu 写道: > 在 2022/1/5 1:17, Saleem, Shiraz 写道: >>> Subject: [PATCH 4/4] RDMA/rxe: Use the standard method to produce >>> udp source >>> port >>> >>> From: Zhu Yanjun <yanjun.zhu@linux.dev> >>> >>> Use the standard method to produce udp source port. >>> >>> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev> >>> --- >>> drivers/infiniband/sw/rxe/rxe_verbs.c | 10 ++++++++++ >>> 1 file changed, 10 insertions(+) >>> >>> diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c >>> b/drivers/infiniband/sw/rxe/rxe_verbs.c >>> index 0aa0d7e52773..f30d98ad13cd 100644 >>> --- a/drivers/infiniband/sw/rxe/rxe_verbs.c >>> +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c >>> @@ -469,6 +469,16 @@ static int rxe_modify_qp(struct ib_qp *ibqp, >>> struct >>> ib_qp_attr *attr, >>> if (err) >>> goto err1; >>> >>> + if (mask & IB_QP_AV) { >>> + if (attr->ah_attr.ah_flags & IB_AH_GRH) { >>> + u32 fl = attr->ah_attr.grh.flow_label; >>> + u32 lqp = qp->ibqp.qp_num; >>> + u32 rqp = qp->attr.dest_qp_num; >>> + >> Isn't the randomization for src_port done in rxe_qp_init_req >> redundant then? >> >> https://elixir.bootlin.com/linux/v5.16-rc8/source/drivers/infiniband/sw/rxe/rxe_qp.c#L220 >> >> >> Can we remove it? Based on the discussion with Leon Romanovsky, when in-subnet communications, GRH is optional. So It is possible that GRH is not set. Without the randomization for src_port done in rxe_qp_init_req, udp source port will be 0xC000 in that case. So it is better to keep this randomization for src_port done in rxe_qp_init_req. Zhu Yanjun > > Yes. We can remove it. > Because this "randomization for src_port done in rxe_qp_init_req" is > replaced by rdma_get_udp_sport in rxe_modify_qp, I do not remove it. > > I will remove it in the latest commits soon. > > Zhu Yanjun > >> >>> + qp->src_port = rdma_get_udp_sport(fl, lqp, rqp); >>> + } >>> + } >>> + >>> return 0; >>> >>> err1: >>> -- >>> 2.27.0 >> >
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c index 0aa0d7e52773..f30d98ad13cd 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -469,6 +469,16 @@ static int rxe_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, if (err) goto err1; + if (mask & IB_QP_AV) { + if (attr->ah_attr.ah_flags & IB_AH_GRH) { + u32 fl = attr->ah_attr.grh.flow_label; + u32 lqp = qp->ibqp.qp_num; + u32 rqp = qp->attr.dest_qp_num; + + qp->src_port = rdma_get_udp_sport(fl, lqp, rqp); + } + } + return 0; err1: