From cd207ccaf0803f4ec9501eed4bb04e24db8a4517 Mon Sep 17 00:00:00 2001 From: hselasky Date: Wed, 29 Jun 2016 10:33:11 +0000 Subject: [PATCH] MFC r301545: Add SR-IOV guest support to the mlx5en driver. This patch adds the missing pieces needed for device setup using the mlx5en driver inside a virtual machine which is providing hardware access through SR-IOV. Sponsored by: Mellanox Technologies git-svn-id: svn://svn.freebsd.org/base/stable/10@302270 ccf9f872-aa2e-dd11-9fc8-001c23d0bc1f --- sys/dev/mlx5/driver.h | 1 + sys/dev/mlx5/mlx5_core/mlx5_vport.c | 358 ++++++++++++++++++++++ sys/dev/mlx5/mlx5_en/mlx5_en_flow_table.c | 132 ++++++++ sys/dev/mlx5/mlx5_en/mlx5_en_main.c | 7 + sys/dev/mlx5/vport.h | 28 ++ 5 files changed, 526 insertions(+) diff --git a/sys/dev/mlx5/driver.h b/sys/dev/mlx5/driver.h index 8136e5751..7f648501f 100644 --- a/sys/dev/mlx5/driver.h +++ b/sys/dev/mlx5/driver.h @@ -33,6 +33,7 @@ #include #include #include +#include #include #include #include diff --git a/sys/dev/mlx5/mlx5_core/mlx5_vport.c b/sys/dev/mlx5/mlx5_core/mlx5_vport.c index 83d8258bb..a3e1751b8 100644 --- a/sys/dev/mlx5/mlx5_core/mlx5_vport.c +++ b/sys/dev/mlx5/mlx5_core/mlx5_vport.c @@ -471,6 +471,241 @@ int mlx5_set_nic_vport_promisc(struct mlx5_core_dev *mdev, int vport, return mlx5_modify_nic_vport_context(mdev, in, sizeof(in)); } EXPORT_SYMBOL_GPL(mlx5_set_nic_vport_promisc); + +int mlx5_query_nic_vport_mac_list(struct mlx5_core_dev *dev, + u32 vport, + enum mlx5_list_type list_type, + u8 addr_list[][ETH_ALEN], + int *list_size) +{ + u32 in[MLX5_ST_SZ_DW(query_nic_vport_context_in)]; + void *nic_vport_ctx; + int max_list_size; + int req_list_size; + u8 *mac_addr; + int out_sz; + void *out; + int err; + int i; + + req_list_size = *list_size; + + max_list_size = (list_type == MLX5_NIC_VPORT_LIST_TYPE_UC) ? + 1 << MLX5_CAP_GEN_MAX(dev, log_max_current_uc_list) : + 1 << MLX5_CAP_GEN_MAX(dev, log_max_current_mc_list); + + if (req_list_size > max_list_size) { + mlx5_core_warn(dev, "Requested list size (%d) > (%d) max_list_size\n", + req_list_size, max_list_size); + req_list_size = max_list_size; + } + + out_sz = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in) + + req_list_size * MLX5_ST_SZ_BYTES(mac_address_layout); + + memset(in, 0, sizeof(in)); + out = kzalloc(out_sz, GFP_KERNEL); + if (!out) + return -ENOMEM; + + MLX5_SET(query_nic_vport_context_in, in, opcode, + MLX5_CMD_OP_QUERY_NIC_VPORT_CONTEXT); + MLX5_SET(query_nic_vport_context_in, in, allowed_list_type, list_type); + MLX5_SET(query_nic_vport_context_in, in, vport_number, vport); + + if (vport) + MLX5_SET(query_nic_vport_context_in, in, other_vport, 1); + + err = mlx5_cmd_exec_check_status(dev, in, sizeof(in), out, out_sz); + if (err) + goto out; + + nic_vport_ctx = MLX5_ADDR_OF(query_nic_vport_context_out, out, + nic_vport_context); + req_list_size = MLX5_GET(nic_vport_context, nic_vport_ctx, + allowed_list_size); + + *list_size = req_list_size; + for (i = 0; i < req_list_size; i++) { + mac_addr = MLX5_ADDR_OF(nic_vport_context, + nic_vport_ctx, + current_uc_mac_address[i]) + 2; + ether_addr_copy(addr_list[i], mac_addr); + } +out: + kfree(out); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_query_nic_vport_mac_list); + +int mlx5_modify_nic_vport_mac_list(struct mlx5_core_dev *dev, + enum mlx5_list_type list_type, + u8 addr_list[][ETH_ALEN], + int list_size) +{ + u32 out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)]; + void *nic_vport_ctx; + int max_list_size; + int in_sz; + void *in; + int err; + int i; + + max_list_size = list_type == MLX5_NIC_VPORT_LIST_TYPE_UC ? + 1 << MLX5_CAP_GEN(dev, log_max_current_uc_list) : + 1 << MLX5_CAP_GEN(dev, log_max_current_mc_list); + + if (list_size > max_list_size) + return -ENOSPC; + + in_sz = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in) + + list_size * MLX5_ST_SZ_BYTES(mac_address_layout); + + memset(out, 0, sizeof(out)); + in = kzalloc(in_sz, GFP_KERNEL); + if (!in) + return -ENOMEM; + + MLX5_SET(modify_nic_vport_context_in, in, opcode, + MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT); + MLX5_SET(modify_nic_vport_context_in, in, + field_select.addresses_list, 1); + + nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in, in, + nic_vport_context); + + MLX5_SET(nic_vport_context, nic_vport_ctx, + allowed_list_type, list_type); + MLX5_SET(nic_vport_context, nic_vport_ctx, + allowed_list_size, list_size); + + for (i = 0; i < list_size; i++) { + u8 *curr_mac = MLX5_ADDR_OF(nic_vport_context, + nic_vport_ctx, + current_uc_mac_address[i]) + 2; + ether_addr_copy(curr_mac, addr_list[i]); + } + + err = mlx5_cmd_exec_check_status(dev, in, in_sz, out, sizeof(out)); + kfree(in); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_modify_nic_vport_mac_list); + +int mlx5_query_nic_vport_vlan_list(struct mlx5_core_dev *dev, + u32 vport, + u16 *vlan_list, + int *list_size) +{ + u32 in[MLX5_ST_SZ_DW(query_nic_vport_context_in)]; + void *nic_vport_ctx; + int max_list_size; + int req_list_size; + int out_sz; + void *out; + void *vlan_addr; + int err; + int i; + + req_list_size = *list_size; + + max_list_size = 1 << MLX5_CAP_GEN_MAX(dev, log_max_vlan_list); + + if (req_list_size > max_list_size) { + mlx5_core_warn(dev, "Requested list size (%d) > (%d) max_list_size\n", + req_list_size, max_list_size); + req_list_size = max_list_size; + } + + out_sz = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in) + + req_list_size * MLX5_ST_SZ_BYTES(vlan_layout); + + memset(in, 0, sizeof(in)); + out = kzalloc(out_sz, GFP_KERNEL); + if (!out) + return -ENOMEM; + + MLX5_SET(query_nic_vport_context_in, in, opcode, + MLX5_CMD_OP_QUERY_NIC_VPORT_CONTEXT); + MLX5_SET(query_nic_vport_context_in, in, allowed_list_type, + MLX5_NIC_VPORT_CONTEXT_ALLOWED_LIST_TYPE_VLAN_LIST); + MLX5_SET(query_nic_vport_context_in, in, vport_number, vport); + + if (vport) + MLX5_SET(query_nic_vport_context_in, in, other_vport, 1); + + err = mlx5_cmd_exec_check_status(dev, in, sizeof(in), out, out_sz); + if (err) + goto out; + + nic_vport_ctx = MLX5_ADDR_OF(query_nic_vport_context_out, out, + nic_vport_context); + req_list_size = MLX5_GET(nic_vport_context, nic_vport_ctx, + allowed_list_size); + + *list_size = req_list_size; + for (i = 0; i < req_list_size; i++) { + vlan_addr = MLX5_ADDR_OF(nic_vport_context, nic_vport_ctx, + current_uc_mac_address[i]); + vlan_list[i] = MLX5_GET(vlan_layout, vlan_addr, vlan); + } +out: + kfree(out); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_query_nic_vport_vlan_list); + +int mlx5_modify_nic_vport_vlans(struct mlx5_core_dev *dev, + u16 vlans[], + int list_size) +{ + u32 out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)]; + void *nic_vport_ctx; + int max_list_size; + int in_sz; + void *in; + int err; + int i; + + max_list_size = 1 << MLX5_CAP_GEN(dev, log_max_vlan_list); + + if (list_size > max_list_size) + return -ENOSPC; + + in_sz = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in) + + list_size * MLX5_ST_SZ_BYTES(vlan_layout); + + memset(out, 0, sizeof(out)); + in = kzalloc(in_sz, GFP_KERNEL); + if (!in) + return -ENOMEM; + + MLX5_SET(modify_nic_vport_context_in, in, opcode, + MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT); + MLX5_SET(modify_nic_vport_context_in, in, + field_select.addresses_list, 1); + + nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in, in, + nic_vport_context); + + MLX5_SET(nic_vport_context, nic_vport_ctx, + allowed_list_type, MLX5_NIC_VPORT_LIST_TYPE_VLAN); + MLX5_SET(nic_vport_context, nic_vport_ctx, + allowed_list_size, list_size); + + for (i = 0; i < list_size; i++) { + void *vlan_addr = MLX5_ADDR_OF(nic_vport_context, + nic_vport_ctx, + current_uc_mac_address[i]); + MLX5_SET(vlan_layout, vlan_addr, vlan, vlans[i]); + } + + err = mlx5_cmd_exec_check_status(dev, in, in_sz, out, sizeof(out)); + kfree(in); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_modify_nic_vport_vlans); + int mlx5_set_nic_vport_permanent_mac(struct mlx5_core_dev *mdev, int vport, u8 *addr) { @@ -785,6 +1020,129 @@ int mlx5_set_eswitch_cvlan_info(struct mlx5_core_dev *mdev, u8 vport, } EXPORT_SYMBOL_GPL(mlx5_set_eswitch_cvlan_info); +int mlx5_arm_vport_context_events(struct mlx5_core_dev *mdev, + u8 vport, + u32 events_mask) +{ + u32 *in; + u32 inlen = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in); + void *nic_vport_ctx; + int err; + + in = mlx5_vzalloc(inlen); + if (!in) + return -ENOMEM; + + MLX5_SET(modify_nic_vport_context_in, + in, + opcode, + MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT); + MLX5_SET(modify_nic_vport_context_in, + in, + field_select.change_event, + 1); + MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport); + if (vport) + MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1); + nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in, + in, + nic_vport_context); + + MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1); + + if (events_mask & MLX5_UC_ADDR_CHANGE) + MLX5_SET(nic_vport_context, + nic_vport_ctx, + event_on_uc_address_change, + 1); + if (events_mask & MLX5_MC_ADDR_CHANGE) + MLX5_SET(nic_vport_context, + nic_vport_ctx, + event_on_mc_address_change, + 1); + if (events_mask & MLX5_VLAN_CHANGE) + MLX5_SET(nic_vport_context, + nic_vport_ctx, + event_on_vlan_change, + 1); + if (events_mask & MLX5_PROMISC_CHANGE) + MLX5_SET(nic_vport_context, + nic_vport_ctx, + event_on_promisc_change, + 1); + if (events_mask & MLX5_MTU_CHANGE) + MLX5_SET(nic_vport_context, + nic_vport_ctx, + event_on_mtu, + 1); + + err = mlx5_modify_nic_vport_context(mdev, in, inlen); + + kvfree(in); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_arm_vport_context_events); + +int mlx5_query_vport_promisc(struct mlx5_core_dev *mdev, + u32 vport, + u8 *promisc_uc, + u8 *promisc_mc, + u8 *promisc_all) +{ + u32 *out; + int outlen = MLX5_ST_SZ_BYTES(query_nic_vport_context_out); + int err; + + out = kzalloc(outlen, GFP_KERNEL); + if (!out) + return -ENOMEM; + + err = mlx5_query_nic_vport_context(mdev, vport, out, outlen); + if (err) + goto out; + + *promisc_uc = MLX5_GET(query_nic_vport_context_out, out, + nic_vport_context.promisc_uc); + *promisc_mc = MLX5_GET(query_nic_vport_context_out, out, + nic_vport_context.promisc_mc); + *promisc_all = MLX5_GET(query_nic_vport_context_out, out, + nic_vport_context.promisc_all); + +out: + kfree(out); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_query_nic_vport_promisc); + +int mlx5_modify_nic_vport_promisc(struct mlx5_core_dev *mdev, + int promisc_uc, + int promisc_mc, + int promisc_all) +{ + void *in; + int inlen = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in); + int err; + + in = mlx5_vzalloc(inlen); + if (!in) { + mlx5_core_err(mdev, "failed to allocate inbox\n"); + return -ENOMEM; + } + + MLX5_SET(modify_nic_vport_context_in, in, field_select.promisc, 1); + MLX5_SET(modify_nic_vport_context_in, in, + nic_vport_context.promisc_uc, promisc_uc); + MLX5_SET(modify_nic_vport_context_in, in, + nic_vport_context.promisc_mc, promisc_mc); + MLX5_SET(modify_nic_vport_context_in, in, + nic_vport_context.promisc_all, promisc_all); + + err = mlx5_modify_nic_vport_context(mdev, in, inlen); + kvfree(in); + return err; +} +EXPORT_SYMBOL_GPL(mlx5_modify_nic_vport_promisc); + int mlx5_query_vport_counter(struct mlx5_core_dev *dev, u8 port_num, u16 vport_num, void *out, int out_size) diff --git a/sys/dev/mlx5/mlx5_en/mlx5_en_flow_table.c b/sys/dev/mlx5/mlx5_en/mlx5_en_flow_table.c index c618c9268..a0dcd2d0d 100644 --- a/sys/dev/mlx5/mlx5_en/mlx5_en_flow_table.c +++ b/sys/dev/mlx5/mlx5_en/mlx5_en_flow_table.c @@ -390,6 +390,49 @@ mlx5e_add_eth_addr_rule(struct mlx5e_priv *priv, return (err); } +static int mlx5e_vport_context_update_vlans(struct mlx5e_priv *priv) +{ + struct ifnet *ifp = priv->ifp; + int max_list_size; + int list_size; + u16 *vlans; + int vlan; + int err; + int i; + + list_size = 0; + for_each_set_bit(vlan, priv->vlan.active_vlans, VLAN_N_VID) + list_size++; + + max_list_size = 1 << MLX5_CAP_GEN(priv->mdev, log_max_vlan_list); + + if (list_size > max_list_size) { + if_printf(ifp, + "ifnet vlans list size (%d) > (%d) max vport list size, some vlans will be dropped\n", + list_size, max_list_size); + list_size = max_list_size; + } + + vlans = kcalloc(list_size, sizeof(*vlans), GFP_KERNEL); + if (!vlans) + return -ENOMEM; + + i = 0; + for_each_set_bit(vlan, priv->vlan.active_vlans, VLAN_N_VID) { + if (i >= list_size) + break; + vlans[i++] = vlan; + } + + err = mlx5_modify_nic_vport_vlans(priv->mdev, vlans, list_size); + if (err) + if_printf(ifp, "Failed to modify vport vlans list err(%d)\n", + err); + + kfree(vlans); + return err; +} + enum mlx5e_vlan_rule_type { MLX5E_VLAN_RULE_TYPE_UNTAGGED, MLX5E_VLAN_RULE_TYPE_ANY_VID, @@ -448,6 +491,7 @@ mlx5e_add_vlan_rule(struct mlx5e_priv *priv, outer_headers.first_vid); MLX5_SET(fte_match_param, match_value, outer_headers.first_vid, vid); + mlx5e_vport_context_update_vlans(priv); break; } @@ -478,6 +522,7 @@ mlx5e_del_vlan_rule(struct mlx5e_priv *priv, case MLX5E_VLAN_RULE_TYPE_MATCH_VID: mlx5_del_flow_table_entry(priv->ft.vlan, priv->vlan.active_vlans_ft_ix[vid]); + mlx5e_vport_context_update_vlans(priv); break; } } @@ -628,6 +673,91 @@ mlx5e_sync_ifp_addr(struct mlx5e_priv *priv) if_maddr_runlock(ifp); } +static void mlx5e_fill_addr_array(struct mlx5e_priv *priv, int list_type, + u8 addr_array[][ETH_ALEN], int size) +{ + bool is_uc = (list_type == MLX5_NIC_VPORT_LIST_TYPE_UC); + struct ifnet *ifp = priv->ifp; + struct mlx5e_eth_addr_hash_node *hn; + struct mlx5e_eth_addr_hash_head *addr_list; + struct mlx5e_eth_addr_hash_node *tmp; + int i = 0; + int hi; + + addr_list = is_uc ? priv->eth_addr.if_uc : priv->eth_addr.if_mc; + + if (is_uc) /* Make sure our own address is pushed first */ + ether_addr_copy(addr_array[i++], IF_LLADDR(ifp)); + else if (priv->eth_addr.broadcast_enabled) + ether_addr_copy(addr_array[i++], ifp->if_broadcastaddr); + + mlx5e_for_each_hash_node(hn, tmp, addr_list, hi) { + if (ether_addr_equal(IF_LLADDR(ifp), hn->ai.addr)) + continue; + if (i >= size) + break; + ether_addr_copy(addr_array[i++], hn->ai.addr); + } +} + +static void mlx5e_vport_context_update_addr_list(struct mlx5e_priv *priv, + int list_type) +{ + bool is_uc = (list_type == MLX5_NIC_VPORT_LIST_TYPE_UC); + struct mlx5e_eth_addr_hash_node *hn; + u8 (*addr_array)[ETH_ALEN] = NULL; + struct mlx5e_eth_addr_hash_head *addr_list; + struct mlx5e_eth_addr_hash_node *tmp; + int max_size; + int size; + int err; + int hi; + + size = is_uc ? 0 : (priv->eth_addr.broadcast_enabled ? 1 : 0); + max_size = is_uc ? + 1 << MLX5_CAP_GEN(priv->mdev, log_max_current_uc_list) : + 1 << MLX5_CAP_GEN(priv->mdev, log_max_current_mc_list); + + addr_list = is_uc ? priv->eth_addr.if_uc : priv->eth_addr.if_mc; + mlx5e_for_each_hash_node(hn, tmp, addr_list, hi) + size++; + + if (size > max_size) { + if_printf(priv->ifp, + "ifp %s list size (%d) > (%d) max vport list size, some addresses will be dropped\n", + is_uc ? "UC" : "MC", size, max_size); + size = max_size; + } + + if (size) { + addr_array = kcalloc(size, ETH_ALEN, GFP_KERNEL); + if (!addr_array) { + err = -ENOMEM; + goto out; + } + mlx5e_fill_addr_array(priv, list_type, addr_array, size); + } + + err = mlx5_modify_nic_vport_mac_list(priv->mdev, list_type, addr_array, size); +out: + if (err) + if_printf(priv->ifp, + "Failed to modify vport %s list err(%d)\n", + is_uc ? "UC" : "MC", err); + kfree(addr_array); +} + +static void mlx5e_vport_context_update(struct mlx5e_priv *priv) +{ + struct mlx5e_eth_addr_db *ea = &priv->eth_addr; + + mlx5e_vport_context_update_addr_list(priv, MLX5_NIC_VPORT_LIST_TYPE_UC); + mlx5e_vport_context_update_addr_list(priv, MLX5_NIC_VPORT_LIST_TYPE_MC); + mlx5_modify_nic_vport_promisc(priv->mdev, 0, + ea->allmulti_enabled, + ea->promisc_enabled); +} + static void mlx5e_apply_ifp_addr(struct mlx5e_priv *priv) { @@ -701,6 +831,8 @@ mlx5e_set_rx_mode_core(struct mlx5e_priv *priv) ea->promisc_enabled = promisc_enabled; ea->allmulti_enabled = allmulti_enabled; ea->broadcast_enabled = broadcast_enabled; + + mlx5e_vport_context_update(priv); } void diff --git a/sys/dev/mlx5/mlx5_en/mlx5_en_main.c b/sys/dev/mlx5/mlx5_en/mlx5_en_main.c index d71cbb36b..4a8029bd9 100644 --- a/sys/dev/mlx5/mlx5_en/mlx5_en_main.c +++ b/sys/dev/mlx5/mlx5_en/mlx5_en_main.c @@ -3001,6 +3001,13 @@ mlx5e_create_ifp(struct mlx5_core_dev *mdev) } mlx5_query_nic_vport_mac_address(priv->mdev, 0, dev_addr); + /* check if we should generate a random MAC address */ + if (MLX5_CAP_GEN(priv->mdev, vport_group_manager) == 0 && + is_zero_ether_addr(dev_addr)) { + random_ether_addr(dev_addr); + if_printf(ifp, "Assigned random MAC address\n"); + } + /* set default MTU */ mlx5e_set_dev_port_mtu(ifp, ifp->if_mtu); diff --git a/sys/dev/mlx5/vport.h b/sys/dev/mlx5/vport.h index c5948e78e..cf52785e8 100644 --- a/sys/dev/mlx5/vport.h +++ b/sys/dev/mlx5/vport.h @@ -38,6 +38,18 @@ int mlx5_vport_query_out_of_rx_buffer(struct mlx5_core_dev *mdev, u32 *out_of_rx_buffer); u8 mlx5_query_vport_state(struct mlx5_core_dev *mdev, u8 opmod); +int mlx5_arm_vport_context_events(struct mlx5_core_dev *mdev, + u8 vport, + u32 events_mask); +int mlx5_query_vport_promisc(struct mlx5_core_dev *mdev, + u32 vport, + u8 *promisc_uc, + u8 *promisc_mc, + u8 *promisc_all); +int mlx5_modify_nic_vport_promisc(struct mlx5_core_dev *mdev, + int promisc_uc, + int promisc_mc, + int promisc_all); int mlx5_query_nic_vport_mac_address(struct mlx5_core_dev *mdev, u32 vport, u8 *addr); int mlx5_set_nic_vport_current_mac(struct mlx5_core_dev *mdev, int vport, @@ -49,6 +61,22 @@ int mlx5_set_nic_vport_mc_list(struct mlx5_core_dev *mdev, int vport, int mlx5_set_nic_vport_promisc(struct mlx5_core_dev *mdev, int vport, bool promisc_mc, bool promisc_uc, bool promisc_all); +int mlx5_query_nic_vport_mac_list(struct mlx5_core_dev *dev, + u32 vport, + enum mlx5_list_type list_type, + u8 addr_list[][ETH_ALEN], + int *list_size); +int mlx5_modify_nic_vport_mac_list(struct mlx5_core_dev *dev, + enum mlx5_list_type list_type, + u8 addr_list[][ETH_ALEN], + int list_size); +int mlx5_query_nic_vport_vlan_list(struct mlx5_core_dev *dev, + u32 vport, + u16 *vlan_list, + int *list_size); +int mlx5_modify_nic_vport_vlans(struct mlx5_core_dev *dev, + u16 vlans[], + int list_size); int mlx5_set_nic_vport_permanent_mac(struct mlx5_core_dev *mdev, int vport, u8 *addr); int mlx5_nic_vport_enable_roce(struct mlx5_core_dev *mdev); -- 2.45.0