From owner-svn-src-all@freebsd.org Fri Feb 10 15:28:20 2017 Return-Path: Delivered-To: svn-src-all@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id CA379CD93D7; Fri, 10 Feb 2017 15:28:20 +0000 (UTC) (envelope-from hselasky@FreeBSD.org) Received: from repo.freebsd.org (repo.freebsd.org [IPv6:2610:1c1:1:6068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 8069F114F; Fri, 10 Feb 2017 15:28:20 +0000 (UTC) (envelope-from hselasky@FreeBSD.org) Received: from repo.freebsd.org ([127.0.1.37]) by repo.freebsd.org (8.15.2/8.15.2) with ESMTP id v1AFSJ1p022567; Fri, 10 Feb 2017 15:28:19 GMT (envelope-from hselasky@FreeBSD.org) Received: (from hselasky@localhost) by repo.freebsd.org (8.15.2/8.15.2/Submit) id v1AFSJUm022561; Fri, 10 Feb 2017 15:28:19 GMT (envelope-from hselasky@FreeBSD.org) Message-Id: <201702101528.v1AFSJUm022561@repo.freebsd.org> X-Authentication-Warning: repo.freebsd.org: hselasky set sender to hselasky@FreeBSD.org using -f From: Hans Petter Selasky Date: Fri, 10 Feb 2017 15:28:19 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r313556 - in head/sys/dev/mlx4: . mlx4_core X-SVN-Group: head MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 10 Feb 2017 15:28:20 -0000 Author: hselasky Date: Fri Feb 10 15:28:18 2017 New Revision: 313556 URL: https://svnweb.freebsd.org/changeset/base/313556 Log: Change mlx4 QP allocation scheme. When using Blue-Flame, BF, the QPN overrides the VLAN, CV, and SV fields in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset. The current ethernet driver code reserves a TX QP range with 256b alignment. This is wrong because if there are more than 64 TX QPs in use, QPNs >= base + 65 will have bits 6/7 set. This problem is not specific for the Ethernet driver, any entity that tries to reserve more than 64 BF-enabled QPs should fail. Also, using ranges is not necessary here and is wasteful. The new mechanism introduced here will support reservation for "Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs (when hypervisors support WC in VMs). The flow we use is: 1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation, and request "BF enabled QPs" if BF is supported for the function 2. In the ALLOC_RES FW command, change param1 to: a. param1[23:0] - number of QPs b. param1[31-24] - flags controlling QPs reservation Bit 31 refers to Eth blueflame supported QPs. Those QPs must have bits 6 and 7 unset in order to be used in Ethernet. Bits 24-30 of the flags are currently reserved. When a function tries to allocate a QP, it states the required attributes for this QP. Those attributes are considered "best-effort". If an attribute, such as Ethernet BF enabled QP, is a must-have attribute, the function has to check that attribute is supported before trying to do the allocation. In a lower layer of the code, mlx4_qp_reserve_range masks out the bits which are unsupported. If SRIOV is used, the PF validates those attributes and masks out unsupported attributes as well. In order to notify VFs which attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's mailbox is filled by the PF, which notifies which QP allocation attributes it supports. Obtained from: Linux (dual BSD/GPLv2 licensed) Submitted by: Dexuan Cui @ microsoft . com Differential Revision: https://reviews.freebsd.org/D8868 MFC after: 2 weeks Sponsored by: Mellanox Technologies Modified: head/sys/dev/mlx4/device.h head/sys/dev/mlx4/mlx4_core/fw.h head/sys/dev/mlx4/mlx4_core/mlx4_fw.c head/sys/dev/mlx4/mlx4_core/mlx4_main.c head/sys/dev/mlx4/mlx4_core/mlx4_qp.c head/sys/dev/mlx4/mlx4_core/mlx4_resource_tracker.c Modified: head/sys/dev/mlx4/device.h ============================================================================== --- head/sys/dev/mlx4/device.h Fri Feb 10 15:22:21 2017 (r313555) +++ head/sys/dev/mlx4/device.h Fri Feb 10 15:28:18 2017 (r313556) @@ -219,6 +219,23 @@ enum { }; enum { + MLX4_QUERY_FUNC_FLAGS_BF_RES_QP = 1LL << 0 +}; + +/* bit enums for an 8-bit flags field indicating special use + * QPs which require special handling in qp_reserve_range. + * Currently, this only includes QPs used by the ETH interface, + * where we expect to use blueflame. These QPs must not have + * bits 6 and 7 set in their qp number. + * + * This enum may use only bits 0..7. + */ +enum { + MLX4_RESERVE_ETH_BF_QP = 1 << 7, +}; + + +enum { MLX4_DEV_CAP_64B_EQE_ENABLED = 1LL << 0, MLX4_DEV_CAP_64B_CQE_ENABLED = 1LL << 1 }; @@ -533,6 +550,7 @@ struct mlx4_caps { u32 max_basic_counters; u32 max_extended_counters; u8 def_counter_index[MLX4_MAX_PORTS + 1]; + u8 alloc_res_qp_mask; }; struct mlx4_buf_list { Modified: head/sys/dev/mlx4/mlx4_core/fw.h ============================================================================== --- head/sys/dev/mlx4/mlx4_core/fw.h Fri Feb 10 15:22:21 2017 (r313555) +++ head/sys/dev/mlx4/mlx4_core/fw.h Fri Feb 10 15:28:18 2017 (r313556) @@ -145,6 +145,7 @@ struct mlx4_func_cap { u8 physical_port; u8 port_flags; u8 def_counter_index; + u8 extra_flags; }; struct mlx4_func { Modified: head/sys/dev/mlx4/mlx4_core/mlx4_fw.c ============================================================================== --- head/sys/dev/mlx4/mlx4_core/mlx4_fw.c Fri Feb 10 15:22:21 2017 (r313555) +++ head/sys/dev/mlx4/mlx4_core/mlx4_fw.c Fri Feb 10 15:28:18 2017 (r313556) @@ -265,10 +265,15 @@ int mlx4_QUERY_FUNC_CAP_wrapper(struct m #define QUERY_FUNC_CAP_MTT_QUOTA_OFFSET 0x64 #define QUERY_FUNC_CAP_MCG_QUOTA_OFFSET 0x68 +#define QUERY_FUNC_CAP_EXTRA_FLAGS_OFFSET 0x6c + #define QUERY_FUNC_CAP_FMR_FLAG 0x80 #define QUERY_FUNC_CAP_FLAG_RDMA 0x40 #define QUERY_FUNC_CAP_FLAG_ETH 0x80 #define QUERY_FUNC_CAP_FLAG_QUOTAS 0x10 +#define QUERY_FUNC_CAP_FLAG_VALID_MAILBOX 0x04 + +#define QUERY_FUNC_CAP_EXTRA_FLAGS_BF_QP_ALLOC_FLAG (1UL << 31) /* when opcode modifier = 1 */ #define QUERY_FUNC_CAP_PHYS_PORT_OFFSET 0x3 @@ -322,7 +327,7 @@ int mlx4_QUERY_FUNC_CAP_wrapper(struct m } else if (vhcr->op_modifier == 0) { /* enable rdma and ethernet interfaces, and new quota locations */ field = (QUERY_FUNC_CAP_FLAG_ETH | QUERY_FUNC_CAP_FLAG_RDMA | - QUERY_FUNC_CAP_FLAG_QUOTAS); + QUERY_FUNC_CAP_FLAG_QUOTAS | QUERY_FUNC_CAP_FLAG_VALID_MAILBOX); MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_FLAGS_OFFSET); field = dev->caps.num_ports; @@ -382,6 +387,8 @@ int mlx4_QUERY_FUNC_CAP_wrapper(struct m MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET); MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET_DEP); + size = QUERY_FUNC_CAP_EXTRA_FLAGS_BF_QP_ALLOC_FLAG; + MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_EXTRA_FLAGS_OFFSET); } else err = -EINVAL; @@ -474,6 +481,17 @@ int mlx4_QUERY_FUNC_CAP(struct mlx4_dev MLX4_GET(size, outbox, QUERY_FUNC_CAP_RESERVED_EQ_OFFSET); func_cap->reserved_eq = size & 0xFFFFFF; + func_cap->extra_flags = 0; + + /* Mailbox data from 0x6c and onward should only be treated if + * QUERY_FUNC_CAP_FLAG_VALID_MAILBOX is set in func_cap->flags + */ + if (func_cap->flags & QUERY_FUNC_CAP_FLAG_VALID_MAILBOX) { + MLX4_GET(size, outbox, QUERY_FUNC_CAP_EXTRA_FLAGS_OFFSET); + if (size & QUERY_FUNC_CAP_EXTRA_FLAGS_BF_QP_ALLOC_FLAG) + func_cap->extra_flags |= MLX4_QUERY_FUNC_FLAGS_BF_RES_QP; + } + goto out; } Modified: head/sys/dev/mlx4/mlx4_core/mlx4_main.c ============================================================================== --- head/sys/dev/mlx4/mlx4_core/mlx4_main.c Fri Feb 10 15:22:21 2017 (r313555) +++ head/sys/dev/mlx4/mlx4_core/mlx4_main.c Fri Feb 10 15:28:18 2017 (r313556) @@ -849,6 +849,11 @@ static int mlx4_dev_cap(struct mlx4_dev if (!mlx4_is_slave(dev)) { for (i = 0; i < dev->caps.num_ports; ++i) dev->caps.def_counter_index[i] = i << 1; + + dev->caps.alloc_res_qp_mask = + (dev->caps.bf_reg_size ? MLX4_RESERVE_ETH_BF_QP : 0); + } else { + dev->caps.alloc_res_qp_mask = 0; } return 0; @@ -1114,6 +1119,10 @@ static int mlx4_slave_cap(struct mlx4_de slave_adjust_steering_mode(dev, &dev_cap, &hca_param); + if (func_cap.extra_flags & MLX4_QUERY_FUNC_FLAGS_BF_RES_QP && + dev->caps.bf_reg_size) + dev->caps.alloc_res_qp_mask |= MLX4_RESERVE_ETH_BF_QP; + return 0; err_mem: Modified: head/sys/dev/mlx4/mlx4_core/mlx4_qp.c ============================================================================== --- head/sys/dev/mlx4/mlx4_core/mlx4_qp.c Fri Feb 10 15:22:21 2017 (r313555) +++ head/sys/dev/mlx4/mlx4_core/mlx4_qp.c Fri Feb 10 15:28:18 2017 (r313556) @@ -242,6 +242,9 @@ int mlx4_qp_reserve_range(struct mlx4_de u64 out_param; int err; + /* Turn off all unsupported QP allocation flags */ + flags &= dev->caps.alloc_res_qp_mask; + if (mlx4_is_mfunc(dev)) { set_param_l(&in_param, (((u32) flags) << 24) | (u32) cnt); set_param_h(&in_param, align); Modified: head/sys/dev/mlx4/mlx4_core/mlx4_resource_tracker.c ============================================================================== --- head/sys/dev/mlx4/mlx4_core/mlx4_resource_tracker.c Fri Feb 10 15:22:21 2017 (r313555) +++ head/sys/dev/mlx4/mlx4_core/mlx4_resource_tracker.c Fri Feb 10 15:28:18 2017 (r313556) @@ -1544,7 +1544,10 @@ static int qp_alloc_res(struct mlx4_dev switch (op) { case RES_OP_RESERVE: count = get_param_l(&in_param) & 0xffffff; - flags = get_param_l(&in_param) >> 24; + /* Turn off all unsupported QP allocation flags that the + * slave tries to set. + */ + flags = (get_param_l(&in_param) >> 24) & dev->caps.alloc_res_qp_mask; align = get_param_h(&in_param); err = mlx4_grant_resource(dev, slave, RES_QP, count, 0); if (err)