From mboxrd@z Thu Jan 1 00:00:00 1970 From: Amir Vadai Subject: [PATCH net-next V7 0/8] net/mlx4_en: DCB QoS support Date: Thu, 5 Apr 2012 10:33:23 +0300 Message-ID: <1333611211-1032-1-git-send-email-amirv@mellanox.com> Cc: netdev@vger.kernel.org, Roland Dreier , Yevgeny Petrilin , Oren Duer , Amir Vadai , Amir Vadai , John Fastabend , Eric Dumazet To: "David S. Miller" Return-path: Received: from eu1sys200aog120.obsmtp.com ([207.126.144.149]:58034 "HELO eu1sys200aog120.obsmtp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1754312Ab2DEHdq (ORCPT ); Thu, 5 Apr 2012 03:33:46 -0400 Sender: netdev-owner@vger.kernel.org List-ID: DCBX version 802.1qaz is supported. User Priority (UP) is set in QP context instead of in WQE (QP Work Queue Element), which means that all traffic from a queue will have the same UP. UP is also set for untagged traffic to be able to classify such traffic too. Mapping from sk_prio to User Priority is done by sch_mqprio mapping. Although confusingly sch_mqprio maps sk_prio to something called TC, it is not related to DCBX's TC, and is interpreted by mlx4_en driver as UP. User can set maximal BW for an ETS TC. This could be done by a new optional attribute added to DCB_ATTR_IEEE called DCBNL DCB_ATTR_IEEE_MAXRATE. It accept an array of 8 64 bits values, 1 for every ETS TC. Units are in Kbps. CC: John Fastabend CC: Eric Dumazet --- Changes from V6: - Patch 7/8 Add an optional max rate attribute to DCBNL - Added a comment for ieee_maxrate structure - code convention fixes according to John F. comments. Changes from V5: - Patch 7/8 Add an optional max rate attribute to DCBNL - Do not use NLA_PUT macro Changes from V4: - Patch 2/8 - set port QoS attributes - HW commands are using HW ratelimit units (100Mbps) - Patch 3/8 - DCB QoS support - ets attributes are stored inline mlx4_en_priv and not alloced separately - Changed variable name: priv->mlx4_en_ieee_ets into priv->ets - Patch 8/8 - Set max rate-limit for a TC in mlx4 driver using the - ratelimit is stored inline mlx4_en_priv and not allocated separately - ratelimit values are converted from Kbps into HW units (100Mbps) in the netlink callback instead of in the HW access function. - ratelimit values are rounded up. 0 is used for unlimited. - moving ratelimit to be inline mlx4_en_priv also addresses Eric D. comments Changes from V3: After having a discussion about the API to set/get rate limit of an ETS TC, decided to extend DCBNL with an optional attribute for that: - new patch 7/8 Add an optional max rate attribute to DCBNL - new patch 8/8 Set max rate-limit for a TC in mlx4 driver using the infrastructure from patch 7/8 Changes from V2: - Removed patch 4 who deal with setting ratelimit through sysfs - Still under discussion - Patch 2/7 - set port QoS attributes - ratelimit is given in Kbps instead of Mbps units. Changes from V1: - Patch 1/7 - Force user priority by QP attribute - removed unneeded comment - Patch 2/7 - set port QoS attributes - ratelimit is given in Mbps instead of 100Mbps units. - Patch 3/7 - DCB QoS support - Split mlx4_en_config_port_scheduler from mlx4_en_dcbnl_ieee_setets - enable setting ratelimit only, without setting up to tc mapping - Patch 4/7 - Set max rate-limit for a TC - Rewritten patch according to comment from Eric D. - setting ratelimit is through sysfs files: /sys/class/net//ratelimit/tc0 /sys/class/net//ratelimit/tc1 ... /sys/class/net//ratelimit/tc7 - Patch 5/7 - sk_prio <=> UP for untagged traffic - Using skb_tx_hash for queue selection of untagged traffic - Added Patch 6/7 - export symbol ip_tos2prio - Fixed patch according to comment from Eric D. - Patch 7/7 - TOS <=> UP mapping for IBoE - No need to clone code from route.c anymore - removed it Changes from V0: - Removed patches 6 and 7 who deal with the interaction between the kernel HW QoS constructs to the queue selection logic are still under discussion with John and some changes might be needed there. Amir Vadai (8): net/mlx4_en: Force user priority by QP attribute net/mlx4_core: set port QoS attributes net/mlx4_en: DCB QoS support net/mlx4_en: sk_prio <=> UP for untagged traffic net/route: export symbol ip_tos2prio IB/rdma_cm: TOS <=> UP mapping for IBoE net/dcb: Add an optional max rate attribute net/mlx4_en: Set max rate-limit for a TC drivers/infiniband/core/cma.c | 6 +- drivers/net/ethernet/mellanox/mlx4/Kconfig | 12 + drivers/net/ethernet/mellanox/mlx4/Makefile | 1 + drivers/net/ethernet/mellanox/mlx4/en_dcb_nl.c | 254 +++++++++++++++++++++ drivers/net/ethernet/mellanox/mlx4/en_main.c | 2 +- drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 26 ++- drivers/net/ethernet/mellanox/mlx4/en_port.h | 2 + drivers/net/ethernet/mellanox/mlx4/en_resources.c | 6 +- drivers/net/ethernet/mellanox/mlx4/en_rx.c | 4 +- drivers/net/ethernet/mellanox/mlx4/en_tx.c | 10 +- drivers/net/ethernet/mellanox/mlx4/mlx4.h | 20 ++ drivers/net/ethernet/mellanox/mlx4/mlx4_en.h | 28 ++- drivers/net/ethernet/mellanox/mlx4/port.c | 62 +++++ include/linux/dcbnl.h | 12 + include/linux/mlx4/cmd.h | 4 + include/linux/mlx4/device.h | 3 + include/linux/mlx4/qp.h | 3 +- include/net/dcbnl.h | 2 + net/dcb/dcbnl.c | 20 ++ net/ipv4/route.c | 2 +- 20 files changed, 461 insertions(+), 18 deletions(-) create mode 100644 drivers/net/ethernet/mellanox/mlx4/en_dcb_nl.c -- 1.7.8.2