| Message ID | 1548146357-11880-1-git-send-email-oulijun@huawei.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [V2,for-next] RDMA/hns: Update the kernel header file of hns | expand |
On Tue, Jan 22, 2019 at 04:39:17PM +0800, Lijun Ou wrote: > The hns_roce_ib_create_srq_resp is used to interact with the > user for data. As a result, it is added to tranfer the srqn > to the user. > > Signed-off-by: Lijun Ou <oulijun@huawei.com> > --- > V1->V2: > - Use hns_roce_ib_create_srq_resp to transfer srqn > --- > drivers/infiniband/hw/hns/hns_roce_srq.c | 4 +++- > include/uapi/rdma/hns-abi.h | 5 +++++ > 2 files changed, 8 insertions(+), 1 deletion(-) Yes, okay, this is better > diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c > index 8975f85..8492153 100644 > --- a/drivers/infiniband/hw/hns/hns_roce_srq.c > +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c > @@ -210,6 +210,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, > struct ib_udata *udata) > { > struct hns_roce_dev *hr_dev = to_hr_dev(pd->device); > + struct hns_roce_ib_create_srq_resp resp = {}; > struct hns_roce_srq *srq; > int srq_desc_size; > int srq_buf_size; > @@ -377,9 +378,10 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, > > srq->event = hns_roce_ib_srq_event; > srq->ibsrq.ext.xrc.srq_num = srq->srqn; > + resp.srqn = srq->srqn; > > if (udata) { > - if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) { > + if (ib_copy_to_udata(udata, &resp, sizeof(resp))) { min(udata->outlen, sizeof(resp)) Jason
On Tue, Jan 22, 2019 at 09:05:26AM -0700, Jason Gunthorpe wrote: > On Tue, Jan 22, 2019 at 04:39:17PM +0800, Lijun Ou wrote: > > The hns_roce_ib_create_srq_resp is used to interact with the > > user for data. As a result, it is added to tranfer the srqn > > to the user. > > > > Signed-off-by: Lijun Ou <oulijun@huawei.com> > > --- > > V1->V2: > > - Use hns_roce_ib_create_srq_resp to transfer srqn > > --- > > drivers/infiniband/hw/hns/hns_roce_srq.c | 4 +++- > > include/uapi/rdma/hns-abi.h | 5 +++++ > > 2 files changed, 8 insertions(+), 1 deletion(-) > > Yes, okay, this is better > > > diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c > > index 8975f85..8492153 100644 > > --- a/drivers/infiniband/hw/hns/hns_roce_srq.c > > +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c > > @@ -210,6 +210,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, > > struct ib_udata *udata) > > { > > struct hns_roce_dev *hr_dev = to_hr_dev(pd->device); > > + struct hns_roce_ib_create_srq_resp resp = {}; > > struct hns_roce_srq *srq; > > int srq_desc_size; > > int srq_buf_size; > > @@ -377,9 +378,10 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, > > > > srq->event = hns_roce_ib_srq_event; > > srq->ibsrq.ext.xrc.srq_num = srq->srqn; > > + resp.srqn = srq->srqn; > > > > if (udata) { > > - if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) { > > + if (ib_copy_to_udata(udata, &resp, sizeof(resp))) { > > min(udata->outlen, sizeof(resp)) IMHO, we should update ib_copy_to_udata() to have 4 parameters instead of 3 and do min() internally. Thanks > > Jason
在 2019/1/23 0:05, Jason Gunthorpe 写道: > On Tue, Jan 22, 2019 at 04:39:17PM +0800, Lijun Ou wrote: >> The hns_roce_ib_create_srq_resp is used to interact with the >> user for data. As a result, it is added to tranfer the srqn >> to the user. >> >> Signed-off-by: Lijun Ou <oulijun@huawei.com> >> --- >> V1->V2: >> - Use hns_roce_ib_create_srq_resp to transfer srqn >> --- >> drivers/infiniband/hw/hns/hns_roce_srq.c | 4 +++- >> include/uapi/rdma/hns-abi.h | 5 +++++ >> 2 files changed, 8 insertions(+), 1 deletion(-) > Yes, okay, this is better > >> diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c >> index 8975f85..8492153 100644 >> --- a/drivers/infiniband/hw/hns/hns_roce_srq.c >> +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c >> @@ -210,6 +210,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, >> struct ib_udata *udata) >> { >> struct hns_roce_dev *hr_dev = to_hr_dev(pd->device); >> + struct hns_roce_ib_create_srq_resp resp = {}; >> struct hns_roce_srq *srq; >> int srq_desc_size; >> int srq_buf_size; >> @@ -377,9 +378,10 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, >> >> srq->event = hns_roce_ib_srq_event; >> srq->ibsrq.ext.xrc.srq_num = srq->srqn; >> + resp.srqn = srq->srqn; >> >> if (udata) { >> - if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) { >> + if (ib_copy_to_udata(udata, &resp, sizeof(resp))) { > min(udata->outlen, sizeof(resp)) > > Jason > Thanks. I will send the v3.
On Tue, Jan 22, 2019 at 07:48:11PM +0200, Leon Romanovsky wrote: > On Tue, Jan 22, 2019 at 09:05:26AM -0700, Jason Gunthorpe wrote: > > On Tue, Jan 22, 2019 at 04:39:17PM +0800, Lijun Ou wrote: > > > The hns_roce_ib_create_srq_resp is used to interact with the > > > user for data. As a result, it is added to tranfer the srqn > > > to the user. > > > > > > Signed-off-by: Lijun Ou <oulijun@huawei.com> > > > V1->V2: > > > - Use hns_roce_ib_create_srq_resp to transfer srqn > > > drivers/infiniband/hw/hns/hns_roce_srq.c | 4 +++- > > > include/uapi/rdma/hns-abi.h | 5 +++++ > > > 2 files changed, 8 insertions(+), 1 deletion(-) > > > > Yes, okay, this is better > > > > > diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c > > > index 8975f85..8492153 100644 > > > +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c > > > @@ -210,6 +210,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, > > > struct ib_udata *udata) > > > { > > > struct hns_roce_dev *hr_dev = to_hr_dev(pd->device); > > > + struct hns_roce_ib_create_srq_resp resp = {}; > > > struct hns_roce_srq *srq; > > > int srq_desc_size; > > > int srq_buf_size; > > > @@ -377,9 +378,10 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, > > > > > > srq->event = hns_roce_ib_srq_event; > > > srq->ibsrq.ext.xrc.srq_num = srq->srqn; > > > + resp.srqn = srq->srqn; > > > > > > if (udata) { > > > - if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) { > > > + if (ib_copy_to_udata(udata, &resp, sizeof(resp))) { > > > > min(udata->outlen, sizeof(resp)) > > IMHO, we should update ib_copy_to_udata() to have 4 parameters instead > of 3 and do min() internally. Yes.. I fixed all the core code, the drivers need to have the same API. Jason
diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c index 8975f85..8492153 100644 --- a/drivers/infiniband/hw/hns/hns_roce_srq.c +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c @@ -210,6 +210,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, struct ib_udata *udata) { struct hns_roce_dev *hr_dev = to_hr_dev(pd->device); + struct hns_roce_ib_create_srq_resp resp = {}; struct hns_roce_srq *srq; int srq_desc_size; int srq_buf_size; @@ -377,9 +378,10 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd, srq->event = hns_roce_ib_srq_event; srq->ibsrq.ext.xrc.srq_num = srq->srqn; + resp.srqn = srq->srqn; if (udata) { - if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) { + if (ib_copy_to_udata(udata, &resp, sizeof(resp))) { ret = -EFAULT; goto err_wrid; } diff --git a/include/uapi/rdma/hns-abi.h b/include/uapi/rdma/hns-abi.h index ef3c7ec..eb76b38 100644 --- a/include/uapi/rdma/hns-abi.h +++ b/include/uapi/rdma/hns-abi.h @@ -52,6 +52,11 @@ struct hns_roce_ib_create_srq { __aligned_u64 que_addr; }; +struct hns_roce_ib_create_srq_resp { + __u32 srqn; + __u32 reserved; +}; + struct hns_roce_ib_create_qp { __aligned_u64 buf_addr; __aligned_u64 db_addr;
The hns_roce_ib_create_srq_resp is used to interact with the user for data. As a result, it is added to tranfer the srqn to the user. Signed-off-by: Lijun Ou <oulijun@huawei.com> --- V1->V2: - Use hns_roce_ib_create_srq_resp to transfer srqn --- drivers/infiniband/hw/hns/hns_roce_srq.c | 4 +++- include/uapi/rdma/hns-abi.h | 5 +++++ 2 files changed, 8 insertions(+), 1 deletion(-)