From 314fdf4473bc05fa9a1f33b27b98eb0f0c836f71 Mon Sep 17 00:00:00 2001 From: Selvin Xavier Date: Tue, 19 May 2015 11:32:32 +0530 Subject: RDMA/ocrdma: Fix EQ destroy failure during driver unload Changing the destroy sequence of mailbox queue and event queues. FW expects mailbox queue to be destroyed before desroying the EQs. Signed-off-by: Selvin Xavier Signed-off-by: Devesh Sharma Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma_hw.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index 0c9e95909a64..3a5ea5afc2a2 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -3147,9 +3147,9 @@ void ocrdma_cleanup_hw(struct ocrdma_dev *dev) ocrdma_free_pd_pool(dev); ocrdma_mbx_delete_ah_tbl(dev); - /* cleanup the eqs */ - ocrdma_destroy_eqs(dev); - /* cleanup the control path */ ocrdma_destroy_mq(dev); + + /* cleanup the eqs */ + ocrdma_destroy_eqs(dev); } -- cgit From 5e6f9237f8e676e8a9110b4dafed36b1dd0b5d84 Mon Sep 17 00:00:00 2001 From: Devesh Sharma Date: Tue, 19 May 2015 11:32:33 +0530 Subject: RDMA/ocrdma: Report EQ full fatal error Detect when Event Queue (EQ) becomes full and print a warning message. Signed-off-by: Devesh Sharma Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma_hw.c | 6 ++++++ drivers/infiniband/hw/ocrdma/ocrdma_sli.h | 7 +++++++ 2 files changed, 13 insertions(+) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index 3a5ea5afc2a2..65759acb12dc 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -933,12 +933,18 @@ static irqreturn_t ocrdma_irq_handler(int irq, void *handle) struct ocrdma_eqe eqe; struct ocrdma_eqe *ptr; u16 cq_id; + u8 mcode; int budget = eq->cq_cnt; do { ptr = ocrdma_get_eqe(eq); eqe = *ptr; ocrdma_le32_to_cpu(&eqe, sizeof(eqe)); + mcode = (eqe.id_valid & OCRDMA_EQE_MAJOR_CODE_MASK) + >> OCRDMA_EQE_MAJOR_CODE_SHIFT; + if (mcode == OCRDMA_MAJOR_CODE_SENTINAL) + pr_err("EQ full on eqid = 0x%x, eqe = 0x%x\n", + eq->q.id, eqe.id_valid); if ((eqe.id_valid & OCRDMA_EQE_VALID_MASK) == 0) break; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h index 243c87c8bd65..baf9b8a0f278 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h @@ -1624,12 +1624,19 @@ struct ocrdma_delete_ah_tbl_rsp { enum { OCRDMA_EQE_VALID_SHIFT = 0, OCRDMA_EQE_VALID_MASK = BIT(0), + OCRDMA_EQE_MAJOR_CODE_MASK = 0x0E, + OCRDMA_EQE_MAJOR_CODE_SHIFT = 0x01, OCRDMA_EQE_FOR_CQE_MASK = 0xFFFE, OCRDMA_EQE_RESOURCE_ID_SHIFT = 16, OCRDMA_EQE_RESOURCE_ID_MASK = 0xFFFF << OCRDMA_EQE_RESOURCE_ID_SHIFT, }; +enum major_code { + OCRDMA_MAJOR_CODE_COMPLETION = 0x00, + OCRDMA_MAJOR_CODE_SENTINAL = 0x01 +}; + struct ocrdma_eqe { u32 id_valid; }; -- cgit From fe48822bc6d5b455aab963038a9c776d641645cc Mon Sep 17 00:00:00 2001 From: Devesh Sharma Date: Tue, 19 May 2015 11:32:34 +0530 Subject: RDMA/ocrdma: Fix QP state transition in destroy_qp Don't move QP to error state, if QP is in reset state during QP destroy operation. Signed-off-by: Devesh Sharma Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c index 877175563634..06e8ab7825b9 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c @@ -1721,18 +1721,20 @@ int ocrdma_destroy_qp(struct ib_qp *ibqp) struct ocrdma_qp *qp; struct ocrdma_dev *dev; struct ib_qp_attr attrs; - int attr_mask = IB_QP_STATE; + int attr_mask; unsigned long flags; qp = get_ocrdma_qp(ibqp); dev = get_ocrdma_dev(ibqp->device); - attrs.qp_state = IB_QPS_ERR; pd = qp->pd; /* change the QP state to ERROR */ - _ocrdma_modify_qp(ibqp, &attrs, attr_mask); - + if (qp->state != OCRDMA_QPS_RST) { + attrs.qp_state = IB_QPS_ERR; + attr_mask = IB_QP_STATE; + _ocrdma_modify_qp(ibqp, &attrs, attr_mask); + } /* ensure that CQEs for newly created QP (whose id may be same with * one which just getting destroyed are same), dont get * discarded until the old CQEs are discarded. -- cgit From 6f5deab0bed6bcfad0dbafcb40a1e269a01ab01d Mon Sep 17 00:00:00 2001 From: Devesh Sharma Date: Tue, 19 May 2015 11:32:35 +0530 Subject: RDMA/ocrdma: Use VID 0 if PFC is enabled and vlan is not configured If the adapter ports are in PFC mode and VLAN is not configured, use vlan tag 0 for RoCE traffic. Also, log an advisory message in system logs. Signed-off-by: Padmanabh Ratnakar Signed-off-by: Devesh Sharma Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma_ah.c | 8 +++++++- drivers/infiniband/hw/ocrdma/ocrdma_hw.c | 12 +++++++++++- 2 files changed, 18 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c index d812904f3984..2a5993b75651 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c @@ -56,7 +56,13 @@ static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, vlan_tag = attr->vlan_id; if (!vlan_tag || (vlan_tag > 0xFFF)) vlan_tag = dev->pvid; - if (vlan_tag && (vlan_tag < 0x1000)) { + if (vlan_tag || dev->pfc_state) { + if (!vlan_tag) { + pr_err("ocrdma%d:Using VLAN with PFC is recommended\n", + dev->id); + pr_err("ocrdma%d:Using VLAN 0 for this connection\n", + dev->id); + } eth.eth_type = cpu_to_be16(0x8100); eth.roce_eth_type = cpu_to_be16(OCRDMA_ROCE_ETH_TYPE); vlan_tag |= (dev->sl & 0x07) << OCRDMA_VID_PCP_SHIFT; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index 65759acb12dc..da688d78fc02 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -2434,7 +2434,7 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, int status; struct ib_ah_attr *ah_attr = &attrs->ah_attr; union ib_gid sgid, zgid; - u32 vlan_id; + u32 vlan_id = 0xFFFF; u8 mac_addr[6]; struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); @@ -2474,12 +2474,22 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, cmd->params.vlan_dmac_b4_to_b5 = mac_addr[4] | (mac_addr[5] << 8); if (attr_mask & IB_QP_VID) { vlan_id = attrs->vlan_id; + } else if (dev->pfc_state) { + vlan_id = 0; + pr_err("ocrdma%d:Using VLAN with PFC is recommended\n", + dev->id); + pr_err("ocrdma%d:Using VLAN 0 for this connection\n", + dev->id); + } + + if (vlan_id < 0x1000) { cmd->params.vlan_dmac_b4_to_b5 |= vlan_id << OCRDMA_QP_PARAMS_VLAN_SHIFT; cmd->flags |= OCRDMA_QP_PARA_VLAN_EN_VALID; cmd->params.rnt_rc_sl_fl |= (dev->sl & 0x07) << OCRDMA_QP_PARAMS_SL_SHIFT; } + return 0; } -- cgit From 038ab8b7432c0280f064a47173dc5b6412243674 Mon Sep 17 00:00:00 2001 From: Mitesh Ahuja Date: Tue, 19 May 2015 11:32:36 +0530 Subject: RDMA/ocrdma: Fix the request length for RDMA_QUERY_QP mailbox command to FW. Fix ocrdma_query_qp to pass correct mailbox request length to FW. Signed-off-by: Mitesh Ahuja Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma_hw.c | 2 +- drivers/infiniband/hw/ocrdma/ocrdma_sli.h | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index da688d78fc02..5636eb6f9307 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -2412,7 +2412,7 @@ int ocrdma_mbx_query_qp(struct ocrdma_dev *dev, struct ocrdma_qp *qp, struct ocrdma_query_qp *cmd; struct ocrdma_query_qp_rsp *rsp; - cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_QUERY_QP, sizeof(*cmd)); + cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_QUERY_QP, sizeof(*rsp)); if (!cmd) return status; cmd->qp_id = qp->id; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h index baf9b8a0f278..02ad0aee99af 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h @@ -1176,6 +1176,8 @@ struct ocrdma_query_qp_rsp { struct ocrdma_mqe_hdr hdr; struct ocrdma_mbx_rsp rsp; struct ocrdma_qp_params params; + u32 dpp_credits_cqid; + u32 rbq_id; }; enum { -- cgit From 59582d86e23f30466413a04cc4db678b09f937ea Mon Sep 17 00:00:00 2001 From: Mitesh Ahuja Date: Tue, 19 May 2015 11:32:37 +0530 Subject: RDMA/ocrdma: Prevent allocation of DPP PDs if FW doesnt support it If DPP PDs are not supported by the FW, allocate only normal PDs. Signed-off-by: Mitesh Ahuja Signed-off-by: Padmanabh Ratnakar Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma_hw.c | 53 ++++++++++++++--------------- drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 2 +- 2 files changed, 27 insertions(+), 28 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index 5636eb6f9307..ac7347ad90be 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -1440,27 +1440,30 @@ static int ocrdma_mbx_alloc_pd_range(struct ocrdma_dev *dev) struct ocrdma_alloc_pd_range_rsp *rsp; /* Pre allocate the DPP PDs */ - cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_ALLOC_PD_RANGE, sizeof(*cmd)); - if (!cmd) - return -ENOMEM; - cmd->pd_count = dev->attr.max_dpp_pds; - cmd->enable_dpp_rsvd |= OCRDMA_ALLOC_PD_ENABLE_DPP; - status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); - if (status) - goto mbx_err; - rsp = (struct ocrdma_alloc_pd_range_rsp *)cmd; - - if ((rsp->dpp_page_pdid & OCRDMA_ALLOC_PD_RSP_DPP) && rsp->pd_count) { - dev->pd_mgr->dpp_page_index = rsp->dpp_page_pdid >> - OCRDMA_ALLOC_PD_RSP_DPP_PAGE_SHIFT; - dev->pd_mgr->pd_dpp_start = rsp->dpp_page_pdid & - OCRDMA_ALLOC_PD_RNG_RSP_START_PDID_MASK; - dev->pd_mgr->max_dpp_pd = rsp->pd_count; - pd_bitmap_size = BITS_TO_LONGS(rsp->pd_count) * sizeof(long); - dev->pd_mgr->pd_dpp_bitmap = kzalloc(pd_bitmap_size, - GFP_KERNEL); + if (dev->attr.max_dpp_pds) { + cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_ALLOC_PD_RANGE, + sizeof(*cmd)); + if (!cmd) + return -ENOMEM; + cmd->pd_count = dev->attr.max_dpp_pds; + cmd->enable_dpp_rsvd |= OCRDMA_ALLOC_PD_ENABLE_DPP; + status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); + rsp = (struct ocrdma_alloc_pd_range_rsp *)cmd; + + if (!status && (rsp->dpp_page_pdid & OCRDMA_ALLOC_PD_RSP_DPP) && + rsp->pd_count) { + dev->pd_mgr->dpp_page_index = rsp->dpp_page_pdid >> + OCRDMA_ALLOC_PD_RSP_DPP_PAGE_SHIFT; + dev->pd_mgr->pd_dpp_start = rsp->dpp_page_pdid & + OCRDMA_ALLOC_PD_RNG_RSP_START_PDID_MASK; + dev->pd_mgr->max_dpp_pd = rsp->pd_count; + pd_bitmap_size = + BITS_TO_LONGS(rsp->pd_count) * sizeof(long); + dev->pd_mgr->pd_dpp_bitmap = kzalloc(pd_bitmap_size, + GFP_KERNEL); + } + kfree(cmd); } - kfree(cmd); cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_ALLOC_PD_RANGE, sizeof(*cmd)); if (!cmd) @@ -1468,10 +1471,8 @@ static int ocrdma_mbx_alloc_pd_range(struct ocrdma_dev *dev) cmd->pd_count = dev->attr.max_pd - dev->attr.max_dpp_pds; status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); - if (status) - goto mbx_err; rsp = (struct ocrdma_alloc_pd_range_rsp *)cmd; - if (rsp->pd_count) { + if (!status && rsp->pd_count) { dev->pd_mgr->pd_norm_start = rsp->dpp_page_pdid & OCRDMA_ALLOC_PD_RNG_RSP_START_PDID_MASK; dev->pd_mgr->max_normal_pd = rsp->pd_count; @@ -1479,15 +1480,13 @@ static int ocrdma_mbx_alloc_pd_range(struct ocrdma_dev *dev) dev->pd_mgr->pd_norm_bitmap = kzalloc(pd_bitmap_size, GFP_KERNEL); } + kfree(cmd); if (dev->pd_mgr->pd_norm_bitmap || dev->pd_mgr->pd_dpp_bitmap) { /* Enable PD resource manager */ dev->pd_mgr->pd_prealloc_valid = true; - } else { - return -ENOMEM; + return 0; } -mbx_err: - kfree(cmd); return status; } diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c index 06e8ab7825b9..9dcb66077d6c 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c @@ -365,7 +365,7 @@ static struct ocrdma_pd *_ocrdma_alloc_pd(struct ocrdma_dev *dev, if (!pd) return ERR_PTR(-ENOMEM); - if (udata && uctx) { + if (udata && uctx && dev->attr.max_dpp_pds) { pd->dpp_enabled = ocrdma_get_asic_type(dev) == OCRDMA_ASIC_GEN_SKH_R; pd->num_dpp_qp = -- cgit From d27b2f15eb535b4f02a75070108c08cefb540257 Mon Sep 17 00:00:00 2001 From: Mitesh Ahuja Date: Tue, 19 May 2015 11:32:38 +0530 Subject: RDMA/ocrdma: Fix dmac resolution for link local address rdma_addr_find_dmac_by_grh fails to resolve dmac for link local address. Use rdma_get_ll_mac to resolve the link local address. Signed-off-by: Padmanabh Ratnakar Signed-off-by: Mitesh Ahuja Signed-off-by: Devesh Sharma Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma.h | 2 ++ drivers/infiniband/hw/ocrdma/ocrdma_ah.c | 4 +++- 2 files changed, 5 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma.h b/drivers/infiniband/hw/ocrdma/ocrdma.h index c9780d919769..ee9e3351a64c 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma.h @@ -515,6 +515,8 @@ static inline int ocrdma_resolve_dmac(struct ocrdma_dev *dev, memcpy(&in6, ah_attr->grh.dgid.raw, sizeof(in6)); if (rdma_is_multicast_addr(&in6)) rdma_get_mcast_mac(&in6, mac_addr); + else if (rdma_link_local_addr(&in6)) + rdma_get_ll_mac(&in6, mac_addr); else memcpy(mac_addr, ah_attr->dmac, ETH_ALEN); return 0; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c index 2a5993b75651..f5a5ea836dbd 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c @@ -127,7 +127,9 @@ struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr) goto av_conf_err; } - if (pd->uctx) { + if ((pd->uctx) && + (!rdma_is_multicast_addr((struct in6_addr *)attr->grh.dgid.raw)) && + (!rdma_link_local_addr((struct in6_addr *)attr->grh.dgid.raw))) { status = rdma_addr_find_dmac_by_grh(&sgid, &attr->grh.dgid, attr->dmac, &attr->vlan_id); if (status) { -- cgit From 72d8a013d75fc845a34451bdce3d3a3dd97488f1 Mon Sep 17 00:00:00 2001 From: Naga Irrinki Date: Tue, 19 May 2015 11:32:39 +0530 Subject: RDMA/ocrdma: Fail connection for MTU lesser than 512 HW currently restricts the IB MTU range between 512 and 4096. Fail connection for MTUs lesser than 512. Signed-off-by: Naga Irrinki Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma_hw.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index ac7347ad90be..47615ff33bc6 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -2534,8 +2534,10 @@ static int ocrdma_set_qp_params(struct ocrdma_qp *qp, cmd->flags |= OCRDMA_QP_PARA_DST_QPN_VALID; } if (attr_mask & IB_QP_PATH_MTU) { - if (attrs->path_mtu < IB_MTU_256 || + if (attrs->path_mtu < IB_MTU_512 || attrs->path_mtu > IB_MTU_4096) { + pr_err("ocrdma%d: IB MTU %d is not supported\n", + dev->id, ib_mtu_enum_to_int(attrs->path_mtu)); status = -EINVAL; goto pmtu_err; } -- cgit From 235dfcd47e4442b2d8766bed3ebaf655b36c7f1c Mon Sep 17 00:00:00 2001 From: Selvin Xavier Date: Tue, 19 May 2015 11:32:40 +0530 Subject: RDMA/ocrdma: Update ocrdma version number Updating the driver version to 10.6.0.0 Signed-off-by: Selvin Xavier Signed-off-by: Doug Ledford --- drivers/infiniband/hw/ocrdma/ocrdma.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/ocrdma/ocrdma.h b/drivers/infiniband/hw/ocrdma/ocrdma.h index ee9e3351a64c..b396344fae16 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma.h @@ -40,7 +40,7 @@ #include #include "ocrdma_sli.h" -#define OCRDMA_ROCE_DRV_VERSION "10.4.205.0u" +#define OCRDMA_ROCE_DRV_VERSION "10.6.0.0" #define OCRDMA_ROCE_DRV_DESC "Emulex OneConnect RoCE Driver" #define OCRDMA_NODE_DESC "Emulex OneConnect RoCE HCA" -- cgit From c29ed5a4566fc7e0c5d06324d62974c6163d1e06 Mon Sep 17 00:00:00 2001 From: Ted Kim Date: Thu, 14 May 2015 12:49:01 -0700 Subject: ib/cm: Change reject message type when destroying cm_id Problem reported by: Ted Kim : We have a case where a Linux system and a non-Linux system are trying to interoperate. The Linux host is the active side and starts the connection establishment, but later decides to not go through with the connection setup and does rdma_destroy_id(). The rdma_destroy_id() eventually works its way down to cm_destroy_id() in core/cm.c, where a REJ is sent. The non-Linux system has some trouble recognizing the REJ because of: A. CM states which can't receive the REJ B. Some issues about REJ formatting (missing comm ID) ISSUE A: That part of the spec says, a Consumer Reject REJ can be sent for a connection abort, but it goes further and says: can send a REJ message with a "Consumer Reject" Reason code if they are in a CM state (i.e. REP Rcvd, MRA(REP) Sent, REQ Rcvd, MRA Sent) that allows a REJ to be sent (lines 35-38). Of the states listed there in that sentence, it would seem to limit the active side to using the Consumer Reject (for the abort case) in just the REP-Rcvd and MRA-REP-Sent states. That is basically only after the active side sees a REP (or alternatively goes down the state transitions to timeout in which case a Timeout REJ is sent). As a fix, in cm-destroy-id() move the IB-CM-MRA-REQ-RCVD case to the same as REQ-SENT. Essentially, make a REJ sent after getting an MRA on active side a timeout rather than Consumer- Reject, which is arguably more correct with the CM state diagrams previous to getting a REP. Signed-off-by: Ted Kim Signed-off-by: Sean Hefty --- drivers/infiniband/core/cm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index 0c1419105ff0..0271608a51c4 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -861,6 +861,7 @@ retest: cm_reject_sidr_req(cm_id_priv, IB_SIDR_REJECT); break; case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); spin_unlock_irq(&cm_id_priv->lock); ib_send_cm_rej(cm_id, IB_CM_REJ_TIMEOUT, @@ -879,7 +880,6 @@ retest: NULL, 0, NULL, 0); } break; - case IB_CM_MRA_REQ_RCVD: case IB_CM_REP_SENT: case IB_CM_MRA_REP_RCVD: ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); -- cgit From c07678bb01374c510b0f6d4a3832c28ba33e9613 Mon Sep 17 00:00:00 2001 From: Matthew Finlay Date: Tue, 19 May 2015 00:11:48 -0700 Subject: IB/cma: Fix broken AF_IB UD support Support for using UD and AF_IB is currently broken. The IB_CM_SIDR_REQ_RECEIVED message is not handled properly in cma_save_net_info() and we end up falling into code that will try and process the request as ipv4/ipv6, which will end up failing. The resolution is to add a check for the SIDR_REQ and call cma_save_ib_info() with a NULL path record. Change cma_save_ib_info() to copy the src sib info from the listen_id when the path record is NULL. Reported-by: Hari Shankar Signed-off-by: Matt Finlay Acked-by: Sean Hefty Signed-off-by: Doug Ledford --- drivers/infiniband/core/cma.c | 32 +++++++++++++++++++++----------- 1 file changed, 21 insertions(+), 11 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 06441a43c3aa..38ffe0981503 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -845,18 +845,26 @@ static void cma_save_ib_info(struct rdma_cm_id *id, struct rdma_cm_id *listen_id listen_ib = (struct sockaddr_ib *) &listen_id->route.addr.src_addr; ib = (struct sockaddr_ib *) &id->route.addr.src_addr; ib->sib_family = listen_ib->sib_family; - ib->sib_pkey = path->pkey; - ib->sib_flowinfo = path->flow_label; - memcpy(&ib->sib_addr, &path->sgid, 16); + if (path) { + ib->sib_pkey = path->pkey; + ib->sib_flowinfo = path->flow_label; + memcpy(&ib->sib_addr, &path->sgid, 16); + } else { + ib->sib_pkey = listen_ib->sib_pkey; + ib->sib_flowinfo = listen_ib->sib_flowinfo; + ib->sib_addr = listen_ib->sib_addr; + } ib->sib_sid = listen_ib->sib_sid; ib->sib_sid_mask = cpu_to_be64(0xffffffffffffffffULL); ib->sib_scope_id = listen_ib->sib_scope_id; - ib = (struct sockaddr_ib *) &id->route.addr.dst_addr; - ib->sib_family = listen_ib->sib_family; - ib->sib_pkey = path->pkey; - ib->sib_flowinfo = path->flow_label; - memcpy(&ib->sib_addr, &path->dgid, 16); + if (path) { + ib = (struct sockaddr_ib *) &id->route.addr.dst_addr; + ib->sib_family = listen_ib->sib_family; + ib->sib_pkey = path->pkey; + ib->sib_flowinfo = path->flow_label; + memcpy(&ib->sib_addr, &path->dgid, 16); + } } static __be16 ss_get_port(const struct sockaddr_storage *ss) @@ -905,9 +913,11 @@ static int cma_save_net_info(struct rdma_cm_id *id, struct rdma_cm_id *listen_id { struct cma_hdr *hdr; - if ((listen_id->route.addr.src_addr.ss_family == AF_IB) && - (ib_event->event == IB_CM_REQ_RECEIVED)) { - cma_save_ib_info(id, listen_id, ib_event->param.req_rcvd.primary_path); + if (listen_id->route.addr.src_addr.ss_family == AF_IB) { + if (ib_event->event == IB_CM_REQ_RECEIVED) + cma_save_ib_info(id, listen_id, ib_event->param.req_rcvd.primary_path); + else if (ib_event->event == IB_CM_SIDR_REQ_RECEIVED) + cma_save_ib_info(id, listen_id, NULL); return 0; } -- cgit