Linux RDMA and InfiniBand development
 help / color / mirror / Atom feed
From: Lang Cheng <chenglang@huawei.com>
To: Weihang Li <liweihang@huawei.com>, <dledford@redhat.com>, <jgg@ziepe.ca>
Cc: <leon@kernel.org>, <linux-rdma@vger.kernel.org>, <linuxarm@huawei.com>
Subject: Re: [PATCH for-next 03/11] RDMA/hns: Do shift on traffic class of UD SQ WQE
Date: Fri, 4 Dec 2020 18:03:59 +0800	[thread overview]
Message-ID: <13aca8fb-8171-b5ca-ec58-540bb36a8512@huawei.com> (raw)
In-Reply-To: <1606899553-54592-4-git-send-email-liweihang@huawei.com>



On 2020/12/2 16:59, Weihang Li wrote:
> The high 6 bits of traffic class in GRH is DSCP (Differentiated Services
> Codepoint), the driver should shift it before the hardware gets it. In
> addition, there is no need to check whether it's RoCEv2 when filling TC
> into QPC because the DSCP field is meaningless for RoCEv1.

I think the high 6 bits of DSCP are valid for ROCEv2, and the entire 8 
bits are valid for ROCEv1.

> 
> Fixes: 606bf89e98ef ("RDMA/hns: Refactor for hns_roce_v2_modify_qp function")
> Fixes: d6a3627e311c ("RDMA/hns: Optimize wqe buffer set flow for post send")
> Signed-off-by: Weihang Li <liweihang@huawei.com>
> ---
>   drivers/infiniband/hw/hns/hns_roce_device.h |  2 ++
>   drivers/infiniband/hw/hns/hns_roce_hw_v2.c  | 13 +++++--------
>   2 files changed, 7 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
> index a5c6bb0..1981501 100644
> --- a/drivers/infiniband/hw/hns/hns_roce_device.h
> +++ b/drivers/infiniband/hw/hns/hns_roce_device.h
> @@ -119,6 +119,8 @@
>   
>   #define HNS_ROCE_QP_BANK_NUM 8
>   
> +#define DSCP_SHIFT 2
> +
>   /* The chip implementation of the consumer index is calculated
>    * according to twice the actual EQ depth
>    */
> diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
> index 8d37067..13c8a2c 100644
> --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
> +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
> @@ -430,7 +430,8 @@ static int fill_ud_av(struct hns_roce_v2_ud_send_wqe *ud_sq_wqe,
>   	roce_set_field(ud_sq_wqe->byte_36, V2_UD_SEND_WQE_BYTE_36_HOPLIMIT_M,
>   		       V2_UD_SEND_WQE_BYTE_36_HOPLIMIT_S, ah->av.hop_limit);
>   	roce_set_field(ud_sq_wqe->byte_36, V2_UD_SEND_WQE_BYTE_36_TCLASS_M,
> -		       V2_UD_SEND_WQE_BYTE_36_TCLASS_S, ah->av.tclass);
> +		       V2_UD_SEND_WQE_BYTE_36_TCLASS_S,
> +		       ah->av.tclass >> DSCP_SHIFT);
>   	roce_set_field(ud_sq_wqe->byte_40, V2_UD_SEND_WQE_BYTE_40_FLOW_LABEL_M,
>   		       V2_UD_SEND_WQE_BYTE_40_FLOW_LABEL_S, ah->av.flowlabel);
>   	roce_set_field(ud_sq_wqe->byte_40, V2_UD_SEND_WQE_BYTE_40_SL_M,
> @@ -4596,15 +4597,11 @@ static int hns_roce_v2_set_path(struct ib_qp *ibqp,
>   	roce_set_field(qpc_mask->byte_24_mtu_tc, V2_QPC_BYTE_24_HOP_LIMIT_M,
>   		       V2_QPC_BYTE_24_HOP_LIMIT_S, 0);
>   
> -	if (is_udp)
> -		roce_set_field(context->byte_24_mtu_tc, V2_QPC_BYTE_24_TC_M,
> -			       V2_QPC_BYTE_24_TC_S, grh->traffic_class >> 2);
> -	else
> -		roce_set_field(context->byte_24_mtu_tc, V2_QPC_BYTE_24_TC_M,
> -			       V2_QPC_BYTE_24_TC_S, grh->traffic_class);
> -
> +	roce_set_field(context->byte_24_mtu_tc, V2_QPC_BYTE_24_TC_M,
> +		       V2_QPC_BYTE_24_TC_S, grh->traffic_class >> DSCP_SHIFT);
>   	roce_set_field(qpc_mask->byte_24_mtu_tc, V2_QPC_BYTE_24_TC_M,
>   		       V2_QPC_BYTE_24_TC_S, 0);
> +
>   	roce_set_field(context->byte_28_at_fl, V2_QPC_BYTE_28_FL_M,
>   		       V2_QPC_BYTE_28_FL_S, grh->flow_label);
>   	roce_set_field(qpc_mask->byte_28_at_fl, V2_QPC_BYTE_28_FL_M,
> 

  reply	other threads:[~2020-12-04 10:05 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-12-02  8:59 [PATCH for-next 00/11] RDMA/hns: Updates for 5.11 Weihang Li
2020-12-02  8:59 ` [PATCH for-next 01/11] RDMA/hns: Limit the length of data copied between kernel and userspace Weihang Li
2020-12-02  8:59 ` [PATCH for-next 02/11] RDMA/hns: Normalization the judgment of some features Weihang Li
2020-12-02  8:59 ` [PATCH for-next 03/11] RDMA/hns: Do shift on traffic class of UD SQ WQE Weihang Li
2020-12-04 10:03   ` Lang Cheng [this message]
2020-12-04 10:10     ` liweihang
2020-12-02  8:59 ` [PATCH for-next 04/11] RDMA/hns: Avoid filling sl in high 3 bits of vlan_id Weihang Li
2020-12-02  8:59 ` [PATCH for-next 05/11] RDMA/hns: WARN_ON if get a reserved sl from users Weihang Li
2020-12-02  8:59 ` [PATCH for-next 06/11] RDMA/hns: Remove unnecessary access right set during INIT2INIT Weihang Li
2020-12-02  8:59 ` [PATCH for-next 07/11] RDMA/hns: Fix coding style issues Weihang Li
2020-12-02  8:59 ` [PATCH for-next 08/11] RDMA/hns: Clear redundant variable initialization Weihang Li
2020-12-02  8:59 ` [PATCH for-next 09/11] RDMA/hns: Fix incorrect symbol types Weihang Li
2020-12-02  8:59 ` [PATCH for-next 10/11] RDMA/hns: Fix inaccurate prints Weihang Li
2020-12-02  8:59 ` [PATCH for-next 11/11] RDMA/hns: Simplify AEQE process for different types of queue Weihang Li

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=13aca8fb-8171-b5ca-ec58-540bb36a8512@huawei.com \
    --to=chenglang@huawei.com \
    --cc=dledford@redhat.com \
    --cc=jgg@ziepe.ca \
    --cc=leon@kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    --cc=linuxarm@huawei.com \
    --cc=liweihang@huawei.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox