From 587443e7773e150ae29e643ee8f41a1eed226565 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Wed, 6 Mar 2019 19:17:56 +0200 Subject: [PATCH 1/5] IB/mlx4: Fix race condition between catas error reset and aliasguid flows Code review revealed a race condition which could allow the catas error flow to interrupt the alias guid query post mechanism at random points. Thiis is fixed by doing cancel_delayed_work_sync() instead of cancel_delayed_work() during the alias guid mechanism destroy flow. Fixes: a0c64a17aba8 ("mlx4: Add alias_guid mechanism") Signed-off-by: Jack Morgenstein Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx4/alias_GUID.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/mlx4/alias_GUID.c b/drivers/infiniband/hw/mlx4/alias_GUID.c index 782499abcd98..2a0b59a4b6eb 100644 --- a/drivers/infiniband/hw/mlx4/alias_GUID.c +++ b/drivers/infiniband/hw/mlx4/alias_GUID.c @@ -804,8 +804,8 @@ void mlx4_ib_destroy_alias_guid_service(struct mlx4_ib_dev *dev) unsigned long flags; for (i = 0 ; i < dev->num_ports; i++) { - cancel_delayed_work(&dev->sriov.alias_guid.ports_guid[i].alias_guid_work); det = &sriov->alias_guid.ports_guid[i]; + cancel_delayed_work_sync(&det->alias_guid_work); spin_lock_irqsave(&sriov->alias_guid.ag_work_lock, flags); while (!list_empty(&det->cb_list)) { cb_ctx = list_entry(det->cb_list.next, From f84b66b9cce78e8f9d38204fdaa75f07c75f4911 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Wed, 6 Mar 2019 19:20:50 +0200 Subject: [PATCH 2/5] net/mlx5: Fix DCT creation bad flow In case the DCT creation command has succeeded a DRAIN must be issued before calling DESTROY. In addition, the original code used the wrong parameter for the DESTROY command, 'in' instead of 'din', which caused another creation try instead of destroying. Cc: # 4.15 Fixes: 57cda166bbe0 ("net/mlx5: Add DCT command interface") Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/net/ethernet/mellanox/mlx5/core/qp.c | 66 +++++++++++--------- 1 file changed, 36 insertions(+), 30 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/qp.c b/drivers/net/ethernet/mellanox/mlx5/core/qp.c index 370ca94b6775..c7c2920c05c4 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/qp.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/qp.c @@ -40,6 +40,9 @@ #include "mlx5_core.h" #include "lib/eq.h" +static int mlx5_core_drain_dct(struct mlx5_core_dev *dev, + struct mlx5_core_dct *dct); + static struct mlx5_core_rsc_common * mlx5_get_rsc(struct mlx5_qp_table *table, u32 rsn) { @@ -227,13 +230,42 @@ static void destroy_resource_common(struct mlx5_core_dev *dev, wait_for_completion(&qp->common.free); } +static int _mlx5_core_destroy_dct(struct mlx5_core_dev *dev, + struct mlx5_core_dct *dct, bool need_cleanup) +{ + u32 out[MLX5_ST_SZ_DW(destroy_dct_out)] = {0}; + u32 in[MLX5_ST_SZ_DW(destroy_dct_in)] = {0}; + struct mlx5_core_qp *qp = &dct->mqp; + int err; + + err = mlx5_core_drain_dct(dev, dct); + if (err) { + if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) { + goto destroy; + } else { + mlx5_core_warn( + dev, "failed drain DCT 0x%x with error 0x%x\n", + qp->qpn, err); + return err; + } + } + wait_for_completion(&dct->drained); +destroy: + if (need_cleanup) + destroy_resource_common(dev, &dct->mqp); + MLX5_SET(destroy_dct_in, in, opcode, MLX5_CMD_OP_DESTROY_DCT); + MLX5_SET(destroy_dct_in, in, dctn, qp->qpn); + MLX5_SET(destroy_dct_in, in, uid, qp->uid); + err = mlx5_cmd_exec(dev, (void *)&in, sizeof(in), + (void *)&out, sizeof(out)); + return err; +} + int mlx5_core_create_dct(struct mlx5_core_dev *dev, struct mlx5_core_dct *dct, u32 *in, int inlen) { u32 out[MLX5_ST_SZ_DW(create_dct_out)] = {0}; - u32 din[MLX5_ST_SZ_DW(destroy_dct_in)] = {0}; - u32 dout[MLX5_ST_SZ_DW(destroy_dct_out)] = {0}; struct mlx5_core_qp *qp = &dct->mqp; int err; @@ -254,11 +286,7 @@ int mlx5_core_create_dct(struct mlx5_core_dev *dev, return 0; err_cmd: - MLX5_SET(destroy_dct_in, din, opcode, MLX5_CMD_OP_DESTROY_DCT); - MLX5_SET(destroy_dct_in, din, dctn, qp->qpn); - MLX5_SET(destroy_dct_in, din, uid, qp->uid); - mlx5_cmd_exec(dev, (void *)&in, sizeof(din), - (void *)&out, sizeof(dout)); + _mlx5_core_destroy_dct(dev, dct, false); return err; } EXPORT_SYMBOL_GPL(mlx5_core_create_dct); @@ -323,29 +351,7 @@ static int mlx5_core_drain_dct(struct mlx5_core_dev *dev, int mlx5_core_destroy_dct(struct mlx5_core_dev *dev, struct mlx5_core_dct *dct) { - u32 out[MLX5_ST_SZ_DW(destroy_dct_out)] = {0}; - u32 in[MLX5_ST_SZ_DW(destroy_dct_in)] = {0}; - struct mlx5_core_qp *qp = &dct->mqp; - int err; - - err = mlx5_core_drain_dct(dev, dct); - if (err) { - if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) { - goto destroy; - } else { - mlx5_core_warn(dev, "failed drain DCT 0x%x with error 0x%x\n", qp->qpn, err); - return err; - } - } - wait_for_completion(&dct->drained); -destroy: - destroy_resource_common(dev, &dct->mqp); - MLX5_SET(destroy_dct_in, in, opcode, MLX5_CMD_OP_DESTROY_DCT); - MLX5_SET(destroy_dct_in, in, dctn, qp->qpn); - MLX5_SET(destroy_dct_in, in, uid, qp->uid); - err = mlx5_cmd_exec(dev, (void *)&in, sizeof(in), - (void *)&out, sizeof(out)); - return err; + return _mlx5_core_destroy_dct(dev, dct, true); } EXPORT_SYMBOL_GPL(mlx5_core_destroy_dct); From c5ae1954c47d3fd8815bd5a592aba18702c93f33 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Wed, 6 Mar 2019 19:21:42 +0200 Subject: [PATCH 3/5] IB/mlx5: Use mlx5 core to create/destroy a DEVX DCT To prevent a hardware memory leak when a DEVX DCT object is destroyed without calling DRAIN DCT before, (e.g. under cleanup flow), need to manage its creation and destruction via mlx5 core. In that case the DRAIN DCT command will be called and only once that it will be completed the DESTROY DCT command will be called. Otherwise, the DESTROY DCT may fail and a hardware leak may occur. As of that change the DRAIN DCT command should not be exposed any more from DEVX, it's managed internally by the driver to work as expected by the device specification. Fixes: 7efce3691d33 ("IB/mlx5: Add obj create and destroy functionality") Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/devx.c | 34 +++++++++++++++----- drivers/infiniband/hw/mlx5/qp.c | 4 ++- drivers/net/ethernet/mellanox/mlx5/core/qp.c | 6 ++-- include/linux/mlx5/qp.h | 3 +- 4 files changed, 34 insertions(+), 13 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index eaa055007f28..9e08df7914aa 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -20,6 +20,7 @@ enum devx_obj_flags { DEVX_OBJ_FLAGS_INDIRECT_MKEY = 1 << 0, + DEVX_OBJ_FLAGS_DCT = 1 << 1, }; struct devx_async_data { @@ -39,7 +40,10 @@ struct devx_obj { u32 dinlen; /* destroy inbox length */ u32 dinbox[MLX5_MAX_DESTROY_INBOX_SIZE_DW]; u32 flags; - struct mlx5_ib_devx_mr devx_mr; + union { + struct mlx5_ib_devx_mr devx_mr; + struct mlx5_core_dct core_dct; + }; }; struct devx_umem { @@ -347,7 +351,6 @@ static u64 devx_get_obj_id(const void *in) obj_id = get_enc_obj_id(MLX5_CMD_OP_CREATE_RQ, MLX5_GET(arm_rq_in, in, srq_number)); break; - case MLX5_CMD_OP_DRAIN_DCT: case MLX5_CMD_OP_ARM_DCT_FOR_KEY_VIOLATION: obj_id = get_enc_obj_id(MLX5_CMD_OP_CREATE_DCT, MLX5_GET(drain_dct_in, in, dctn)); @@ -618,7 +621,6 @@ static bool devx_is_obj_modify_cmd(const void *in) case MLX5_CMD_OP_2RST_QP: case MLX5_CMD_OP_ARM_XRC_SRQ: case MLX5_CMD_OP_ARM_RQ: - case MLX5_CMD_OP_DRAIN_DCT: case MLX5_CMD_OP_ARM_DCT_FOR_KEY_VIOLATION: case MLX5_CMD_OP_ARM_XRQ: case MLX5_CMD_OP_SET_XRQ_DC_PARAMS_ENTRY: @@ -1124,7 +1126,11 @@ static int devx_obj_cleanup(struct ib_uobject *uobject, if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) devx_cleanup_mkey(obj); - ret = mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, sizeof(out)); + if (obj->flags & DEVX_OBJ_FLAGS_DCT) + ret = mlx5_core_destroy_dct(obj->mdev, &obj->core_dct); + else + ret = mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, + sizeof(out)); if (ib_is_destroy_retryable(ret, why, uobject)) return ret; @@ -1185,9 +1191,17 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)( devx_set_umem_valid(cmd_in); } - err = mlx5_cmd_exec(dev->mdev, cmd_in, - cmd_in_len, - cmd_out, cmd_out_len); + if (opcode == MLX5_CMD_OP_CREATE_DCT) { + obj->flags |= DEVX_OBJ_FLAGS_DCT; + err = mlx5_core_create_dct(dev->mdev, &obj->core_dct, + cmd_in, cmd_in_len, + cmd_out, cmd_out_len); + } else { + err = mlx5_cmd_exec(dev->mdev, cmd_in, + cmd_in_len, + cmd_out, cmd_out_len); + } + if (err) goto obj_free; @@ -1214,7 +1228,11 @@ err_copy: if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) devx_cleanup_mkey(obj); obj_destroy: - mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, sizeof(out)); + if (obj->flags & DEVX_OBJ_FLAGS_DCT) + mlx5_core_destroy_dct(obj->mdev, &obj->core_dct); + else + mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out, + sizeof(out)); obj_free: kfree(obj); return err; diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 6b1f0e76900b..7cd006da1dae 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -3729,6 +3729,7 @@ static int mlx5_ib_modify_dct(struct ib_qp *ibqp, struct ib_qp_attr *attr, } else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) { struct mlx5_ib_modify_qp_resp resp = {}; + u32 out[MLX5_ST_SZ_DW(create_dct_out)] = {0}; u32 min_resp_len = offsetof(typeof(resp), dctn) + sizeof(resp.dctn); @@ -3747,7 +3748,8 @@ static int mlx5_ib_modify_dct(struct ib_qp *ibqp, struct ib_qp_attr *attr, MLX5_SET(dctc, dctc, hop_limit, attr->ah_attr.grh.hop_limit); err = mlx5_core_create_dct(dev->mdev, &qp->dct.mdct, qp->dct.in, - MLX5_ST_SZ_BYTES(create_dct_in)); + MLX5_ST_SZ_BYTES(create_dct_in), out, + sizeof(out)); if (err) return err; resp.dctn = qp->dct.mdct.mqp.qpn; diff --git a/drivers/net/ethernet/mellanox/mlx5/core/qp.c b/drivers/net/ethernet/mellanox/mlx5/core/qp.c index c7c2920c05c4..b8ba74de9555 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/qp.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/qp.c @@ -263,16 +263,16 @@ destroy: int mlx5_core_create_dct(struct mlx5_core_dev *dev, struct mlx5_core_dct *dct, - u32 *in, int inlen) + u32 *in, int inlen, + u32 *out, int outlen) { - u32 out[MLX5_ST_SZ_DW(create_dct_out)] = {0}; struct mlx5_core_qp *qp = &dct->mqp; int err; init_completion(&dct->drained); MLX5_SET(create_dct_in, in, opcode, MLX5_CMD_OP_CREATE_DCT); - err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out)); + err = mlx5_cmd_exec(dev, in, inlen, out, outlen); if (err) { mlx5_core_warn(dev, "create DCT failed, ret %d\n", err); return err; diff --git a/include/linux/mlx5/qp.h b/include/linux/mlx5/qp.h index b26ea9077384..0343c81d4c5f 100644 --- a/include/linux/mlx5/qp.h +++ b/include/linux/mlx5/qp.h @@ -557,7 +557,8 @@ static inline struct mlx5_core_mkey *__mlx5_mr_lookup(struct mlx5_core_dev *dev, int mlx5_core_create_dct(struct mlx5_core_dev *dev, struct mlx5_core_dct *qp, - u32 *in, int inlen); + u32 *in, int inlen, + u32 *out, int outlen); int mlx5_core_create_qp(struct mlx5_core_dev *dev, struct mlx5_core_qp *qp, u32 *in, From cd27287562d69629c5f007c6f64c27b6ff15f3e5 Mon Sep 17 00:00:00 2001 From: Aya Levin Date: Mon, 11 Mar 2019 14:35:58 +0200 Subject: [PATCH 4/5] IB/mlx5: Fix mapping of link-mode to IB width and speed Add mapping of link mode: CAUI4 100Gbps CR4/KR4 with 4 lines and 25Gbps. Fix mapping of link mode: GAUI2 50Gbps CR2/KR2 to be 2 lines with 25Gbps. Fixes: 08e8676f1607 ("IB/mlx5: Add support for 50Gbps per lane link modes") Signed-off-by: Aya Levin Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 994c19d01211..531ff20b32ad 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -415,10 +415,17 @@ static int translate_eth_ext_proto_oper(u32 eth_proto_oper, u8 *active_speed, *active_speed = IB_SPEED_EDR; break; case MLX5E_PROT_MASK(MLX5E_50GAUI_2_LAUI_2_50GBASE_CR2_KR2): + *active_width = IB_WIDTH_2X; + *active_speed = IB_SPEED_EDR; + break; case MLX5E_PROT_MASK(MLX5E_50GAUI_1_LAUI_1_50GBASE_CR_KR): *active_width = IB_WIDTH_1X; *active_speed = IB_SPEED_HDR; break; + case MLX5E_PROT_MASK(MLX5E_CAUI_4_100GBASE_CR4_KR4): + *active_width = IB_WIDTH_4X; + *active_speed = IB_SPEED_EDR; + break; case MLX5E_PROT_MASK(MLX5E_100GAUI_2_100GBASE_CR2_KR2): *active_width = IB_WIDTH_2X; *active_speed = IB_SPEED_HDR; From ec4fe4bcc584b55e24e8d1768f5510a62c0fd619 Mon Sep 17 00:00:00 2001 From: Feng Tang Date: Thu, 14 Mar 2019 18:37:29 +0800 Subject: [PATCH 5/5] i40iw: Avoid panic when handling the inetdev event There is a panic reported that on a system with x722 ethernet, when doing the operations like: # ip link add br0 type bridge # ip link set eno1 master br0 # systemctl restart systemd-networkd The system will panic "BUG: unable to handle kernel null pointer dereference at 0000000000000034", with call chain: i40iw_inetaddr_event notifier_call_chain blocking_notifier_call_chain notifier_call_chain __inet_del_ifa inet_rtm_deladdr rtnetlink_rcv_msg netlink_rcv_skb rtnetlink_rcv netlink_unicast netlink_sendmsg sock_sendmsg __sys_sendto It is caused by "local_ipaddr = ntohl(in->ifa_list->ifa_address)", while the in->ifa_list is NULL. So add a check for the "in->ifa_list == NULL" case, and skip the ARP operation accordingly. Signed-off-by: Feng Tang Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/i40iw/i40iw_utils.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/i40iw/i40iw_utils.c b/drivers/infiniband/hw/i40iw/i40iw_utils.c index c5a881172524..337410f40860 100644 --- a/drivers/infiniband/hw/i40iw/i40iw_utils.c +++ b/drivers/infiniband/hw/i40iw/i40iw_utils.c @@ -173,7 +173,12 @@ int i40iw_inetaddr_event(struct notifier_block *notifier, rcu_read_lock(); in = __in_dev_get_rcu(upper_dev); - local_ipaddr = ntohl(in->ifa_list->ifa_address); + + if (!in->ifa_list) + local_ipaddr = 0; + else + local_ipaddr = ntohl(in->ifa_list->ifa_address); + rcu_read_unlock(); } else { local_ipaddr = ntohl(ifa->ifa_address); @@ -185,6 +190,11 @@ int i40iw_inetaddr_event(struct notifier_block *notifier, case NETDEV_UP: /* Fall through */ case NETDEV_CHANGEADDR: + + /* Just skip if no need to handle ARP cache */ + if (!local_ipaddr) + break; + i40iw_manage_arp_cache(iwdev, netdev->dev_addr, &local_ipaddr,