The requires: 1. Replacing the paravirtualized pkey index (inserted by the guest) with the real pkey index
2. For UD qp's, placing the guest's true source gid index in the address path structure mgid field, and setting the ud_force_mgid bit so that the mgid is taken from the qp context and not from the WQE when posting sends. 3. For UC and RC QPs, placing the guest's true source gid index in the address path structure mgid field. 4. For tunnel and proxy QPs, setting the QKEY value reserved for that proxy/tunnel pair. Since not all the above adjustments occur in all the QP transitions, the QP transitions require separate wrapper functions. Finally, initialize the pkey virtualization table to its default values: Master virtualized table is 1-1 with the real pkey table, guest virtualized table has pkey index 0 mapped to the real pkey index 0, and all the other pkey indices mapped to the reserved (invalid) pkey at index 127. Signed-off-by: Jack Morgenstein <[email protected]> --- drivers/infiniband/hw/mlx4/main.c | 35 +++ drivers/net/ethernet/mellanox/mlx4/cmd.c | 12 +- drivers/net/ethernet/mellanox/mlx4/main.c | 11 + drivers/net/ethernet/mellanox/mlx4/mlx4.h | 47 +++++ .../net/ethernet/mellanox/mlx4/resource_tracker.c | 217 +++++++++++++++++++- include/linux/mlx4/device.h | 3 + 6 files changed, 313 insertions(+), 12 deletions(-) diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 66f28b6..213d459 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -1084,6 +1084,38 @@ static int mlx4_ib_netdev_event(struct notifier_block *this, unsigned long event return NOTIFY_DONE; } +static void init_pkeys(struct mlx4_ib_dev *ibdev) +{ + int port; + int slave; + int i; + + if (mlx4_is_master(ibdev->dev)) { + for (slave = 0; slave <= ibdev->dev->num_vfs; ++slave) { + for (port = 1; port <= ibdev->dev->caps.num_ports; ++port) { + for (i = 0; + i < ibdev->dev->phys_caps.pkey_phys_table_len[port]; + ++i) { + ibdev->pkeys.virt2phys_pkey[slave][port - 1][i] = + /* master has the identity virt2phys pkey mapping */ + (slave == ibdev->dev->caps.function || !i) ? i : + ibdev->dev->phys_caps.pkey_phys_table_len[port] - 1; + mlx4_sync_pkey_table(ibdev->dev, slave, port, i, + ibdev->pkeys.virt2phys_pkey[slave][port - 1][i]); + } + } + } + /* initialize pkey cache */ + for (port = 1; port <= ibdev->dev->caps.num_ports; ++port) { + for (i = 0; + i < ibdev->dev->phys_caps.pkey_phys_table_len[port]; + ++i) + ibdev->pkeys.phys_pkey_cache[port-1][i] = + (i) ? 0 : 0xFFFF; + } + } +} + static void mlx4_ib_alloc_eqs(struct mlx4_dev *dev, struct mlx4_ib_dev *ibdev) { char name[32]; @@ -1337,6 +1369,9 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) ibdev->ib_active = true; + if (mlx4_is_mfunc(ibdev->dev)) + init_pkeys(ibdev); + return ibdev; err_notif: diff --git a/drivers/net/ethernet/mellanox/mlx4/cmd.c b/drivers/net/ethernet/mellanox/mlx4/cmd.c index fc5e701..bf55be6 100644 --- a/drivers/net/ethernet/mellanox/mlx4/cmd.c +++ b/drivers/net/ethernet/mellanox/mlx4/cmd.c @@ -905,7 +905,7 @@ static struct mlx4_cmd_info cmd_info[] = { .out_is_imm = false, .encode_slave_id = false, .verify = NULL, - .wrapper = mlx4_GEN_QP_wrapper + .wrapper = mlx4_INIT2INIT_QP_wrapper }, { .opcode = MLX4_CMD_INIT2RTR_QP, @@ -923,7 +923,7 @@ static struct mlx4_cmd_info cmd_info[] = { .out_is_imm = false, .encode_slave_id = false, .verify = NULL, - .wrapper = mlx4_GEN_QP_wrapper + .wrapper = mlx4_RTR2RTS_QP_wrapper }, { .opcode = MLX4_CMD_RTS2RTS_QP, @@ -932,7 +932,7 @@ static struct mlx4_cmd_info cmd_info[] = { .out_is_imm = false, .encode_slave_id = false, .verify = NULL, - .wrapper = mlx4_GEN_QP_wrapper + .wrapper = mlx4_RTS2RTS_QP_wrapper }, { .opcode = MLX4_CMD_SQERR2RTS_QP, @@ -941,7 +941,7 @@ static struct mlx4_cmd_info cmd_info[] = { .out_is_imm = false, .encode_slave_id = false, .verify = NULL, - .wrapper = mlx4_GEN_QP_wrapper + .wrapper = mlx4_SQERR2RTS_QP_wrapper }, { .opcode = MLX4_CMD_2ERR_QP, @@ -968,7 +968,7 @@ static struct mlx4_cmd_info cmd_info[] = { .out_is_imm = false, .encode_slave_id = false, .verify = NULL, - .wrapper = mlx4_GEN_QP_wrapper + .wrapper = mlx4_SQD2SQD_QP_wrapper }, { .opcode = MLX4_CMD_SQD2RTS_QP, @@ -977,7 +977,7 @@ static struct mlx4_cmd_info cmd_info[] = { .out_is_imm = false, .encode_slave_id = false, .verify = NULL, - .wrapper = mlx4_GEN_QP_wrapper + .wrapper = mlx4_SQD2RTS_QP_wrapper }, { .opcode = MLX4_CMD_2RST_QP, diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c index 4e046d4..79dd3be 100644 --- a/drivers/net/ethernet/mellanox/mlx4/main.c +++ b/drivers/net/ethernet/mellanox/mlx4/main.c @@ -403,6 +403,17 @@ int mlx4_get_parav_qkey(struct mlx4_dev *dev, u32 qpn, u32 *qkey) } EXPORT_SYMBOL(mlx4_get_parav_qkey); +void mlx4_sync_pkey_table(struct mlx4_dev *dev, int slave, int port, int i, int val) +{ + struct mlx4_priv *priv = container_of(dev, struct mlx4_priv, dev); + + if (!mlx4_is_master(dev)) + return; + + priv->virt2phys_pkey[slave][port - 1][i] = val; +} +EXPORT_SYMBOL(mlx4_sync_pkey_table); + int mlx4_is_slave_active(struct mlx4_dev *dev, int slave) { struct mlx4_priv *priv = mlx4_priv(dev); diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4.h b/drivers/net/ethernet/mellanox/mlx4/mlx4.h index cde6e51..87cc0d8 100644 --- a/drivers/net/ethernet/mellanox/mlx4/mlx4.h +++ b/drivers/net/ethernet/mellanox/mlx4/mlx4.h @@ -716,6 +716,8 @@ struct mlx4_priv { struct mutex bf_mutex; struct io_mapping *bf_mapping; int reserved_mtts; + u8 virt2phys_pkey[MLX4_MFUNC_MAX][MLX4_MAX_PORTS][MLX4_MAX_PORT_PKEYS]; + }; static inline struct mlx4_priv *mlx4_priv(struct mlx4_dev *dev) @@ -920,16 +922,61 @@ int mlx4_RST2INIT_QP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_cmd_mailbox *inbox, struct mlx4_cmd_mailbox *outbox, struct mlx4_cmd_info *cmd); +int mlx4_INIT2INIT_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); int mlx4_INIT2RTR_QP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr, struct mlx4_cmd_mailbox *inbox, struct mlx4_cmd_mailbox *outbox, struct mlx4_cmd_info *cmd); +int mlx4_RTR2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); +int mlx4_RTS2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); +int mlx4_SQERR2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); +int mlx4_2ERR_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); +int mlx4_RTS2SQD_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); +int mlx4_SQD2SQD_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); +int mlx4_SQD2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); int mlx4_2RST_QP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr, struct mlx4_cmd_mailbox *inbox, struct mlx4_cmd_mailbox *outbox, struct mlx4_cmd_info *cmd); +int mlx4_QUERY_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd); int mlx4_GEN_EQE(struct mlx4_dev *dev, int slave, struct mlx4_eqe *eqe); diff --git a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c index e3a3b24..637603e 100644 --- a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c +++ b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c @@ -189,6 +189,15 @@ struct res_xrcdn { int port; }; +enum qp_transition { + QP_TRANS_INIT2RTR, + QP_TRANS_RTR2RTS, + QP_TRANS_RTS2RTS, + QP_TRANS_SQERR2RTS, + QP_TRANS_SQD2SQD, + QP_TRANS_SQD2RTS +}; + /* For Debug uses */ static const char *ResourceType(enum mlx4_resource rt) { @@ -255,14 +264,41 @@ void mlx4_free_resource_tracker(struct mlx4_dev *dev, } } -static void update_ud_gid(struct mlx4_dev *dev, - struct mlx4_qp_context *qp_ctx, u8 slave) +static void update_pkey_index(struct mlx4_dev *dev, int slave, + struct mlx4_cmd_mailbox *inbox) { - u32 ts = (be32_to_cpu(qp_ctx->flags) >> 16) & 0xff; + u8 sched = *(u8 *)(inbox->buf + 64); + u8 orig_index = *(u8 *)(inbox->buf + 35); + u8 new_index; + struct mlx4_priv *priv = mlx4_priv(dev); + int port; + + port = (sched >> 6 & 1) + 1; + + new_index = priv->virt2phys_pkey[slave][port - 1][orig_index]; + *(u8 *)(inbox->buf + 35) = new_index; + + mlx4_dbg(dev, "port = %d, orig pkey index = %d, " + "new pkey index = %d\n", port, orig_index, new_index); +} + +static void update_gid(struct mlx4_dev *dev, struct mlx4_cmd_mailbox *inbox, + u8 slave) +{ + struct mlx4_qp_context *qp_ctx = inbox->buf + 8; + enum mlx4_qp_optpar optpar = be32_to_cpu(*(__be32 *) inbox->buf); + u32 ts = (be32_to_cpu(qp_ctx->flags) >> 16) & 0xff; if (MLX4_QP_ST_UD == ts) qp_ctx->pri_path.mgid_index = 0x80 | slave; + if (MLX4_QP_ST_RC == ts || MLX4_QP_ST_UC == ts) { + if (optpar & MLX4_QP_OPTPAR_PRIMARY_ADDR_PATH) + qp_ctx->pri_path.mgid_index = slave & 0x7F; + if (optpar & MLX4_QP_OPTPAR_ALT_ADDR_PATH) + qp_ctx->alt_path.mgid_index = slave & 0x7F; + } + mlx4_dbg(dev, "slave %d, new gid index: 0x%x ", slave, qp_ctx->pri_path.mgid_index); } @@ -1027,6 +1063,11 @@ static int valid_reserved(struct mlx4_dev *dev, int slave, int qpn) (mlx4_is_master(dev) || mlx4_is_guest_proxy(dev, slave, qpn)); } +static int fw_reserved(struct mlx4_dev *dev, int qpn) +{ + return qpn < dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW]; +} + static int qp_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd, u64 in_param, u64 *out_param) { @@ -1064,7 +1105,7 @@ static int qp_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd, if (err) return err; - if (!valid_reserved(dev, slave, qpn)) { + if (!fw_reserved(dev, qpn)) { err = __mlx4_qp_alloc_icm(dev, qpn); if (err) { res_abort_move(dev, slave, RES_QP, qpn); @@ -1417,7 +1458,7 @@ static int qp_free_res(struct mlx4_dev *dev, int slave, int op, int cmd, if (err) return err; - if (!valid_reserved(dev, slave, qpn)) + if (!fw_reserved(dev, qpn)) __mlx4_qp_free_icm(dev, qpn); res_end_move(dev, slave, RES_QP, qpn); @@ -1857,6 +1898,19 @@ static u32 qp_get_srqn(struct mlx4_qp_context *qpc) return be32_to_cpu(qpc->srqn) & 0x1ffffff; } +static void adjust_proxy_tun_qkey(struct mlx4_dev *dev, struct mlx4_vhcr *vhcr, + struct mlx4_qp_context *context) +{ + u32 qpn = vhcr->in_modifier & 0xffffff; + u32 qkey = 0; + + if (mlx4_get_parav_qkey(dev, qpn, &qkey)) + return; + + /* adjust qkey in qp context */ + context->qkey = cpu_to_be32(qkey); +} + int mlx4_RST2INIT_QP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr, struct mlx4_cmd_mailbox *inbox, @@ -1909,6 +1963,8 @@ int mlx4_RST2INIT_QP_wrapper(struct mlx4_dev *dev, int slave, goto ex_put_scq; } + adjust_proxy_tun_qkey(dev, vhcr, qpc); + update_pkey_index(dev, slave, inbox); err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd); if (err) goto ex_put_srq; @@ -2054,6 +2110,48 @@ static int get_containing_mtt(struct mlx4_dev *dev, int slave, int start, return err; } +static int verify_qp_parameters(struct mlx4_dev *dev, + struct mlx4_cmd_mailbox *inbox, + enum qp_transition transition, u8 slave) +{ + u32 qp_type; + struct mlx4_qp_context *qp_ctx; + enum mlx4_qp_optpar optpar; + + qp_ctx = inbox->buf + 8; + qp_type = (be32_to_cpu(qp_ctx->flags) >> 16) & 0xff; + optpar = be32_to_cpu(*(__be32 *) inbox->buf); + + switch (qp_type) { + case MLX4_QP_ST_RC: + case MLX4_QP_ST_UC: + switch (transition) { + case QP_TRANS_INIT2RTR: + case QP_TRANS_RTR2RTS: + case QP_TRANS_RTS2RTS: + case QP_TRANS_SQD2SQD: + case QP_TRANS_SQD2RTS: + if (dev->caps.function != slave) + /* slaves have only gid index 0 */ + if (optpar & MLX4_QP_OPTPAR_PRIMARY_ADDR_PATH) + if (qp_ctx->pri_path.mgid_index) + return -EINVAL; + if (optpar & MLX4_QP_OPTPAR_ALT_ADDR_PATH) + if (qp_ctx->alt_path.mgid_index) + return -EINVAL; + break; + default: + break; + } + + break; + default: + break; + } + + return 0; +} + int mlx4_WRITE_MTT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr, struct mlx4_cmd_mailbox *inbox, @@ -2541,16 +2639,123 @@ out: return err; } +int mlx4_INIT2INIT_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd) +{ + struct mlx4_qp_context *context = inbox->buf + 8; + adjust_proxy_tun_qkey(dev, vhcr, context); + update_pkey_index(dev, slave, inbox); + return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd); +} + int mlx4_INIT2RTR_QP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr, struct mlx4_cmd_mailbox *inbox, struct mlx4_cmd_mailbox *outbox, struct mlx4_cmd_info *cmd) { + int err; struct mlx4_qp_context *qpc = inbox->buf + 8; - update_ud_gid(dev, qpc, (u8)slave); + err = verify_qp_parameters(dev, inbox, QP_TRANS_INIT2RTR, slave); + if (err) + return err; + + update_pkey_index(dev, slave, inbox); + update_gid(dev, inbox, (u8)slave); + adjust_proxy_tun_qkey(dev, vhcr, qpc); + + return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd); +} + +int mlx4_RTR2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd) +{ + int err; + struct mlx4_qp_context *context = inbox->buf + 8; + + err = verify_qp_parameters(dev, inbox, QP_TRANS_RTR2RTS, slave); + if (err) + return err; + + update_pkey_index(dev, slave, inbox); + update_gid(dev, inbox, (u8)slave); + adjust_proxy_tun_qkey(dev, vhcr, context); + return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd); +} + +int mlx4_RTS2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd) +{ + int err; + struct mlx4_qp_context *context = inbox->buf + 8; + + err = verify_qp_parameters(dev, inbox, QP_TRANS_RTS2RTS, slave); + if (err) + return err; + + update_pkey_index(dev, slave, inbox); + update_gid(dev, inbox, (u8)slave); + adjust_proxy_tun_qkey(dev, vhcr, context); + return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd); +} + + +int mlx4_SQERR2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd) +{ + struct mlx4_qp_context *context = inbox->buf + 8; + adjust_proxy_tun_qkey(dev, vhcr, context); + return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd); +} + +int mlx4_SQD2SQD_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd) +{ + int err; + struct mlx4_qp_context *context = inbox->buf + 8; + + err = verify_qp_parameters(dev, inbox, QP_TRANS_SQD2SQD, slave); + if (err) + return err; + + adjust_proxy_tun_qkey(dev, vhcr, context); + update_gid(dev, inbox, (u8)slave); + update_pkey_index(dev, slave, inbox); + return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd); +} + +int mlx4_SQD2RTS_QP_wrapper(struct mlx4_dev *dev, int slave, + struct mlx4_vhcr *vhcr, + struct mlx4_cmd_mailbox *inbox, + struct mlx4_cmd_mailbox *outbox, + struct mlx4_cmd_info *cmd) +{ + int err; + struct mlx4_qp_context *context = inbox->buf + 8; + + err = verify_qp_parameters(dev, inbox, QP_TRANS_SQD2RTS, slave); + if (err) + return err; + adjust_proxy_tun_qkey(dev, vhcr, context); + update_gid(dev, inbox, (u8)slave); + update_pkey_index(dev, slave, inbox); return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd); } diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h index fd53a26..a714048 100644 --- a/include/linux/mlx4/device.h +++ b/include/linux/mlx4/device.h @@ -792,6 +792,9 @@ int mlx4_wol_write(struct mlx4_dev *dev, u64 config, int port); int mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx); void mlx4_counter_free(struct mlx4_dev *dev, u32 idx); +void mlx4_sync_pkey_table(struct mlx4_dev *dev, int slave, int port, + int i, int val); + int mlx4_get_parav_qkey(struct mlx4_dev *dev, u32 qpn, u32 *qkey); #endif /* MLX4_DEVICE_H */ -- 1.7.1 -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
