diff options
Diffstat (limited to 'drivers/infiniband')
46 files changed, 1212 insertions, 484 deletions
diff --git a/drivers/infiniband/core/iwcm.c b/drivers/infiniband/core/iwcm.c index 8f9509e1ebf..55d093a36ae 100644 --- a/drivers/infiniband/core/iwcm.c +++ b/drivers/infiniband/core/iwcm.c @@ -362,6 +362,7 @@ static void destroy_cm_id(struct iw_cm_id *cm_id) * In either case, must tell the provider to reject. */ cm_id_priv->state = IW_CM_STATE_DESTROYING; + cm_id->device->iwcm->reject(cm_id, NULL, 0); break; case IW_CM_STATE_CONN_SENT: case IW_CM_STATE_DESTROYING: diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index de922a04ca2..7522008fda8 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -2,6 +2,7 @@ * Copyright (c) 2004-2007 Voltaire, Inc. All rights reserved. * Copyright (c) 2005 Intel Corporation. All rights reserved. * Copyright (c) 2005 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2009 HNR Consulting. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU @@ -45,14 +46,21 @@ MODULE_DESCRIPTION("kernel IB MAD API"); MODULE_AUTHOR("Hal Rosenstock"); MODULE_AUTHOR("Sean Hefty"); +int mad_sendq_size = IB_MAD_QP_SEND_SIZE; +int mad_recvq_size = IB_MAD_QP_RECV_SIZE; + +module_param_named(send_queue_size, mad_sendq_size, int, 0444); +MODULE_PARM_DESC(send_queue_size, "Size of send queue in number of work requests"); +module_param_named(recv_queue_size, mad_recvq_size, int, 0444); +MODULE_PARM_DESC(recv_queue_size, "Size of receive queue in number of work requests"); + static struct kmem_cache *ib_mad_cache; static struct list_head ib_mad_port_list; static u32 ib_mad_client_id = 0; /* Port list lock */ -static spinlock_t ib_mad_port_list_lock; - +static DEFINE_SPINLOCK(ib_mad_port_list_lock); /* Forward declarations */ static int method_in_use(struct ib_mad_mgmt_method_table **method, @@ -1974,7 +1982,7 @@ static void adjust_timeout(struct ib_mad_agent_private *mad_agent_priv) unsigned long delay; if (list_empty(&mad_agent_priv->wait_list)) { - cancel_delayed_work(&mad_agent_priv->timed_work); + __cancel_delayed_work(&mad_agent_priv->timed_work); } else { mad_send_wr = list_entry(mad_agent_priv->wait_list.next, struct ib_mad_send_wr_private, @@ -1983,7 +1991,7 @@ static void adjust_timeout(struct ib_mad_agent_private *mad_agent_priv) if (time_after(mad_agent_priv->timeout, mad_send_wr->timeout)) { mad_agent_priv->timeout = mad_send_wr->timeout; - cancel_delayed_work(&mad_agent_priv->timed_work); + __cancel_delayed_work(&mad_agent_priv->timed_work); delay = mad_send_wr->timeout - jiffies; if ((long)delay <= 0) delay = 1; @@ -2023,7 +2031,7 @@ static void wait_for_response(struct ib_mad_send_wr_private *mad_send_wr) /* Reschedule a work item if we have a shorter timeout */ if (mad_agent_priv->wait_list.next == &mad_send_wr->agent_list) { - cancel_delayed_work(&mad_agent_priv->timed_work); + __cancel_delayed_work(&mad_agent_priv->timed_work); queue_delayed_work(mad_agent_priv->qp_info->port_priv->wq, &mad_agent_priv->timed_work, delay); } @@ -2736,8 +2744,8 @@ static int create_mad_qp(struct ib_mad_qp_info *qp_info, qp_init_attr.send_cq = qp_info->port_priv->cq; qp_init_attr.recv_cq = qp_info->port_priv->cq; qp_init_attr.sq_sig_type = IB_SIGNAL_ALL_WR; - qp_init_attr.cap.max_send_wr = IB_MAD_QP_SEND_SIZE; - qp_init_attr.cap.max_recv_wr = IB_MAD_QP_RECV_SIZE; + qp_init_attr.cap.max_send_wr = mad_sendq_size; + qp_init_attr.cap.max_recv_wr = mad_recvq_size; qp_init_attr.cap.max_send_sge = IB_MAD_SEND_REQ_MAX_SG; qp_init_attr.cap.max_recv_sge = IB_MAD_RECV_REQ_MAX_SG; qp_init_attr.qp_type = qp_type; @@ -2752,8 +2760,8 @@ static int create_mad_qp(struct ib_mad_qp_info *qp_info, goto error; } /* Use minimum queue sizes unless the CQ is resized */ - qp_info->send_queue.max_active = IB_MAD_QP_SEND_SIZE; - qp_info->recv_queue.max_active = IB_MAD_QP_RECV_SIZE; + qp_info->send_queue.max_active = mad_sendq_size; + qp_info->recv_queue.max_active = mad_recvq_size; return 0; error: @@ -2792,7 +2800,7 @@ static int ib_mad_port_open(struct ib_device *device, init_mad_qp(port_priv, &port_priv->qp_info[0]); init_mad_qp(port_priv, &port_priv->qp_info[1]); - cq_size = (IB_MAD_QP_SEND_SIZE + IB_MAD_QP_RECV_SIZE) * 2; + cq_size = (mad_sendq_size + mad_recvq_size) * 2; port_priv->cq = ib_create_cq(port_priv->device, ib_mad_thread_completion_handler, NULL, port_priv, cq_size, 0); @@ -2984,7 +2992,11 @@ static int __init ib_mad_init_module(void) { int ret; - spin_lock_init(&ib_mad_port_list_lock); + mad_recvq_size = min(mad_recvq_size, IB_MAD_QP_MAX_SIZE); + mad_recvq_size = max(mad_recvq_size, IB_MAD_QP_MIN_SIZE); + + mad_sendq_size = min(mad_sendq_size, IB_MAD_QP_MAX_SIZE); + mad_sendq_size = max(mad_sendq_size, IB_MAD_QP_MIN_SIZE); ib_mad_cache = kmem_cache_create("ib_mad", sizeof(struct ib_mad_private), @@ -3021,4 +3033,3 @@ static void __exit ib_mad_cleanup_module(void) module_init(ib_mad_init_module); module_exit(ib_mad_cleanup_module); - diff --git a/drivers/infiniband/core/mad_priv.h b/drivers/infiniband/core/mad_priv.h index 05ce331733b..9430ab4969c 100644 --- a/drivers/infiniband/core/mad_priv.h +++ b/drivers/infiniband/core/mad_priv.h @@ -2,6 +2,7 @@ * Copyright (c) 2004, 2005, Voltaire, Inc. All rights reserved. * Copyright (c) 2005 Intel Corporation. All rights reserved. * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2009 HNR Consulting. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU @@ -49,6 +50,8 @@ /* QP and CQ parameters */ #define IB_MAD_QP_SEND_SIZE 128 #define IB_MAD_QP_RECV_SIZE 512 +#define IB_MAD_QP_MIN_SIZE 64 +#define IB_MAD_QP_MAX_SIZE 8192 #define IB_MAD_SEND_REQ_MAX_SG 2 #define IB_MAD_RECV_REQ_MAX_SG 1 diff --git a/drivers/infiniband/core/multicast.c b/drivers/infiniband/core/multicast.c index 107f170c57c..8d82ba17135 100644 --- a/drivers/infiniband/core/multicast.c +++ b/drivers/infiniband/core/multicast.c @@ -106,6 +106,8 @@ struct mcast_group { struct ib_sa_query *query; int query_id; u16 pkey_index; + u8 leave_state; + int retries; }; struct mcast_member { @@ -350,6 +352,7 @@ static int send_leave(struct mcast_group *group, u8 leave_state) rec = group->rec; rec.join_state = leave_state; + group->leave_state = leave_state; ret = ib_sa_mcmember_rec_query(&sa_client, port->dev->device, port->port_num, IB_SA_METHOD_DELETE, &rec, @@ -542,7 +545,11 @@ static void leave_handler(int status, struct ib_sa_mcmember_rec *rec, { struct mcast_group *group = context; - mcast_work_handler(&group->work); + if (status && group->retries > 0 && + !send_leave(group, group->leave_state)) + group->retries--; + else + mcast_work_handler(&group->work); } static struct mcast_group *acquire_group(struct mcast_port *port, @@ -565,6 +572,7 @@ static struct mcast_group *acquire_group(struct mcast_port *port, if (!group) return NULL; + group->retries = 3; group->port = port; group->rec.mgid = *mgid; group->pkey_index = MCAST_INVALID_PKEY_INDEX; diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index 1865049e80f..82543716d59 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -109,10 +109,10 @@ static struct ib_client sa_client = { .remove = ib_sa_remove_one }; -static spinlock_t idr_lock; +static DEFINE_SPINLOCK(idr_lock); static DEFINE_IDR(query_idr); -static spinlock_t tid_lock; +static DEFINE_SPINLOCK(tid_lock); static u32 tid; #define PATH_REC_FIELD(field) \ @@ -1077,9 +1077,6 @@ static int __init ib_sa_init(void) { int ret; - spin_lock_init(&idr_lock); - spin_lock_init(&tid_lock); - get_random_bytes(&tid, sizeof tid); ret = ib_register_client(&sa_client); diff --git a/drivers/infiniband/core/smi.c b/drivers/infiniband/core/smi.c index 87236753bce..5855e4405d9 100644 --- a/drivers/infiniband/core/smi.c +++ b/drivers/infiniband/core/smi.c @@ -52,6 +52,10 @@ enum smi_action smi_handle_dr_smp_send(struct ib_smp *smp, hop_cnt = smp->hop_cnt; /* See section 14.2.2.2, Vol 1 IB spec */ + /* C14-6 -- valid hop_cnt values are from 0 to 63 */ + if (hop_cnt >= IB_SMP_MAX_PATH_HOPS) + return IB_SMI_DISCARD; + if (!ib_get_smp_direction(smp)) { /* C14-9:1 */ if (hop_cnt && hop_ptr == 0) { @@ -133,6 +137,10 @@ enum smi_action smi_handle_dr_smp_recv(struct ib_smp *smp, u8 node_type, hop_cnt = smp->hop_cnt; /* See section 14.2.2.2, Vol 1 IB spec */ + /* C14-6 -- valid hop_cnt values are from 0 to 63 */ + if (hop_cnt >= IB_SMP_MAX_PATH_HOPS) + return IB_SMI_DISCARD; + if (!ib_get_smp_direction(smp)) { /* C14-9:1 -- sender should have incremented hop_ptr */ if (hop_cnt && hop_ptr == 0) diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index eb36a81dd09..d3fff9e008a 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -73,7 +73,7 @@ DEFINE_IDR(ib_uverbs_cq_idr); DEFINE_IDR(ib_uverbs_qp_idr); DEFINE_IDR(ib_uverbs_srq_idr); -static spinlock_t map_lock; +static DEFINE_SPINLOCK(map_lock); static struct ib_uverbs_device *dev_table[IB_UVERBS_MAX_DEVICES]; static DECLARE_BITMAP(dev_map, IB_UVERBS_MAX_DEVICES); @@ -584,14 +584,16 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, if (hdr.command < 0 || hdr.command >= ARRAY_SIZE(uverbs_cmd_table) || - !uverbs_cmd_table[hdr.command] || - !(file->device->ib_dev->uverbs_cmd_mask & (1ull << hdr.command))) + !uverbs_cmd_table[hdr.command]) return -EINVAL; if (!file->ucontext && hdr.command != IB_USER_VERBS_CMD_GET_CONTEXT) return -EINVAL; + if (!(file->device->ib_dev->uverbs_cmd_mask & (1ull << hdr.command))) + return -ENOSYS; + return uverbs_cmd_table[hdr.command](file, buf + sizeof hdr, hdr.in_words * 4, hdr.out_words * 4); } @@ -836,8 +838,6 @@ static int __init ib_uverbs_init(void) { int ret; - spin_lock_init(&map_lock); - ret = register_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES, "infiniband_verbs"); if (ret) { diff --git a/drivers/infiniband/hw/amso1100/c2.c b/drivers/infiniband/hw/amso1100/c2.c index 0cfbb6d2f76..8250740c94b 100644 --- a/drivers/infiniband/hw/amso1100/c2.c +++ b/drivers/infiniband/hw/amso1100/c2.c @@ -86,11 +86,7 @@ MODULE_DEVICE_TABLE(pci, c2_pci_table); static void c2_print_macaddr(struct net_device *netdev) { - pr_debug("%s: MAC %02X:%02X:%02X:%02X:%02X:%02X, " - "IRQ %u\n", netdev->name, - netdev->dev_addr[0], netdev->dev_addr[1], netdev->dev_addr[2], - netdev->dev_addr[3], netdev->dev_addr[4], netdev->dev_addr[5], - netdev->irq); + pr_debug("%s: MAC %pM, IRQ %u\n", netdev->name, netdev->dev_addr, netdev->irq); } static void c2_set_rxbufsize(struct c2_port *c2_port) diff --git a/drivers/infiniband/hw/amso1100/c2_provider.c b/drivers/infiniband/hw/amso1100/c2_provider.c index f1948fad85d..ad723bd8bf4 100644 --- a/drivers/infiniband/hw/amso1100/c2_provider.c +++ b/drivers/infiniband/hw/amso1100/c2_provider.c @@ -780,11 +780,11 @@ int c2_register_device(struct c2_dev *dev) /* Register pseudo network device */ dev->pseudo_netdev = c2_pseudo_netdev_init(dev); if (!dev->pseudo_netdev) - goto out3; + goto out; ret = register_netdev(dev->pseudo_netdev); if (ret) - goto out2; + goto out_free_netdev; pr_debug("%s:%u\n", __func__, __LINE__); strlcpy(dev->ibdev.name, "amso%d", IB_DEVICE_NAME_MAX); @@ -851,6 +851,10 @@ int c2_register_device(struct c2_dev *dev) dev->ibdev.post_recv = c2_post_receive; dev->ibdev.iwcm = kmalloc(sizeof(*dev->ibdev.iwcm), GFP_KERNEL); + if (dev->ibdev.iwcm == NULL) { + ret = -ENOMEM; + goto out_unregister_netdev; + } dev->ibdev.iwcm->add_ref = c2_add_ref; dev->ibdev.iwcm->rem_ref = c2_rem_ref; dev->ibdev.iwcm->get_qp = c2_get_qp; @@ -862,23 +866,25 @@ int c2_register_device(struct c2_dev *dev) ret = ib_register_device(&dev->ibdev); if (ret) - goto out1; + goto out_free_iwcm; for (i = 0; i < ARRAY_SIZE(c2_dev_attributes); ++i) { ret = device_create_file(&dev->ibdev.dev, c2_dev_attributes[i]); if (ret) - goto out0; + goto out_unregister_ibdev; } - goto out3; + goto out; -out0: +out_unregister_ibdev: ib_unregister_device(&dev->ibdev); -out1: +out_free_iwcm: + kfree(dev->ibdev.iwcm); +out_unregister_netdev: unregister_netdev(dev->pseudo_netdev); -out2: +out_free_netdev: free_netdev(dev->pseudo_netdev); -out3: +out: pr_debug("%s:%u ret=%d\n", __func__, __LINE__, ret); return ret; } diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c index 62f9cf2f94e..72ed3396b72 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_hal.c +++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c @@ -852,7 +852,9 @@ int cxio_rdma_init(struct cxio_rdev *rdev_p, struct t3_rdma_init_attr *attr) wqe->qpcaps = attr->qpcaps; wqe->ulpdu_size = cpu_to_be16(attr->tcp_emss); wqe->rqe_count = cpu_to_be16(attr->rqe_count); - wqe->flags_rtr_type = cpu_to_be16(attr->flags|V_RTR_TYPE(attr->rtr_type)); + wqe->flags_rtr_type = cpu_to_be16(attr->flags | + V_RTR_TYPE(attr->rtr_type) | + V_CHAN(attr->chan)); wqe->ord = cpu_to_be32(attr->ord); wqe->ird = cpu_to_be32(attr->ird); wqe->qp_dma_addr = cpu_to_be64(attr->qp_dma_addr); @@ -1032,6 +1034,7 @@ err3: err2: cxio_hal_destroy_ctrl_qp(rdev_p); err1: + rdev_p->t3cdev_p->ulp = NULL; list_del(&rdev_p->entry); return err; } diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h index 32e3b1461d8..a197a5b7ac7 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_wr.h +++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h @@ -327,6 +327,11 @@ enum rdma_init_rtr_types { #define V_RTR_TYPE(x) ((x) << S_RTR_TYPE) #define G_RTR_TYPE(x) ((((x) >> S_RTR_TYPE)) & M_RTR_TYPE) +#define S_CHAN 4 +#define M_CHAN 0x3 +#define V_CHAN(x) ((x) << S_CHAN) +#define G_CHAN(x) ((((x) >> S_CHAN)) & M_CHAN) + struct t3_rdma_init_attr { u32 tid; u32 qpid; @@ -346,6 +351,7 @@ struct t3_rdma_init_attr { u16 flags; u16 rqe_count; u32 irs; + u32 chan; }; struct t3_rdma_init_wr { diff --git a/drivers/infiniband/hw/cxgb3/iwch.c b/drivers/infiniband/hw/cxgb3/iwch.c index 26fc0a4eaa7..b0ea0105ddf 100644 --- a/drivers/infiniband/hw/cxgb3/iwch.c +++ b/drivers/infiniband/hw/cxgb3/iwch.c @@ -51,7 +51,7 @@ cxgb3_cpl_handler_func t3c_handlers[NUM_CPL_CMDS]; static void open_rnic_dev(struct t3cdev *); static void close_rnic_dev(struct t3cdev *); -static void iwch_err_handler(struct t3cdev *, u32, u32); +static void iwch_event_handler(struct t3cdev *, u32, u32); struct cxgb3_client t3c_client = { .name = "iw_cxgb3", @@ -59,7 +59,7 @@ struct cxgb3_client t3c_client = { .remove = close_rnic_dev, .handlers = t3c_handlers, .redirect = iwch_ep_redirect, - .err_handler = iwch_err_handler + .event_handler = iwch_event_handler }; static LIST_HEAD(dev_list); @@ -105,11 +105,9 @@ static void rnic_init(struct iwch_dev *rnicp) static void open_rnic_dev(struct t3cdev *tdev) { struct iwch_dev *rnicp; - static int vers_printed; PDBG("%s t3cdev %p\n", __func__, tdev); - if (!vers_printed++) - printk(KERN_INFO MOD "Chelsio T3 RDMA Driver - version %s\n", + printk_once(KERN_INFO MOD "Chelsio T3 RDMA Driver - version %s\n", DRV_VERSION); rnicp = (struct iwch_dev *)ib_alloc_device(sizeof(*rnicp)); if (!rnicp) { @@ -162,21 +160,36 @@ static void close_rnic_dev(struct t3cdev *tdev) mutex_unlock(&dev_mutex); } -static void iwch_err_handler(struct t3cdev *tdev, u32 status, u32 error) +static void iwch_event_handler(struct t3cdev *tdev, u32 evt, u32 port_id) { struct cxio_rdev *rdev = tdev->ulp; - struct iwch_dev *rnicp = rdev_to_iwch_dev(rdev); + struct iwch_dev *rnicp; struct ib_event event; + u32 portnum = port_id + 1; - if (status == OFFLOAD_STATUS_DOWN) { + if (!rdev) + return; + rnicp = rdev_to_iwch_dev(rdev); + switch (evt) { + case OFFLOAD_STATUS_DOWN: { rdev->flags = CXIO_ERROR_FATAL; - - event.device = &rnicp->ibdev; event.event = IB_EVENT_DEVICE_FATAL; - event.element.port_num = 0; - ib_dispatch_event(&event); + break; + } + case OFFLOAD_PORT_DOWN: { + event.event = IB_EVENT_PORT_ERR; + break; + } + case OFFLOAD_PORT_UP: { + event.event = IB_EVENT_PORT_ACTIVE; + break; + } } + event.device = &rnicp->ibdev; + event.element.port_num = portnum; + ib_dispatch_event(&event); + return; } diff --git a/drivers/infiniband/hw/cxgb3/iwch_cm.c b/drivers/infiniband/hw/cxgb3/iwch_cm.c index 52d7bb0c2a1..66b41351910 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_cm.c +++ b/drivers/infiniband/hw/cxgb3/iwch_cm.c @@ -286,7 +286,7 @@ void __free_ep(struct kref *kref) ep = container_of(container_of(kref, struct iwch_ep_common, kref), struct iwch_ep, com); PDBG("%s ep %p state %s\n", __func__, ep, states[state_read(&ep->com)]); - if (ep->com.flags & RELEASE_RESOURCES) { + if (test_bit(RELEASE_RESOURCES, &ep->com.flags)) { cxgb3_remove_tid(ep->com.tdev, (void *)ep, ep->hwtid); dst_release(ep->dst); l2t_release(L2DATA(ep->com.tdev), ep->l2t); @@ -297,7 +297,7 @@ void __free_ep(struct kref *kref) static void release_ep_resources(struct iwch_ep *ep) { PDBG("%s ep %p tid %d\n", __func__, ep, ep->hwtid); - ep->com.flags |= RELEASE_RESOURCES; + set_bit(RELEASE_RESOURCES, &ep->com.flags); put_ep(&ep->com); } @@ -786,10 +786,12 @@ static void connect_request_upcall(struct iwch_ep *ep) event.private_data_len = ep->plen; event.private_data = ep->mpa_pkt + sizeof(struct mpa_message); event.provider_data = ep; - if (state_read(&ep->parent_ep->com) != DEAD) + if (state_read(&ep->parent_ep->com) != DEAD) { + get_ep(&ep->com); ep->parent_ep->com.cm_id->event_handler( ep->parent_ep->com.cm_id, &event); + } put_ep(&ep->parent_ep->com); ep->parent_ep = NULL; } @@ -1156,8 +1158,7 @@ static int abort_rpl(struct t3cdev *tdev, struct sk_buff *skb, void *ctx) * We get 2 abort replies from the HW. The first one must * be ignored except for scribbling that we need one more. */ - if (!(ep->com.flags & ABORT_REQ_IN_PROGRESS)) { - ep->com.flags |= ABORT_REQ_IN_PROGRESS; + if (!test_and_set_bit(ABORT_REQ_IN_PROGRESS, &ep->com.flags)) { return CPL_RET_BUF_DONE; } @@ -1477,10 +1478,14 @@ static int peer_close(struct t3cdev *tdev, struct sk_buff *skb, void *ctx) /* * We're gonna mark this puppy DEAD, but keep * the reference on it until the ULP accepts or - * rejects the CR. + * rejects the CR. Also wake up anyone waiting + * in rdma connection migration (see iwch_accept_cr()). */ __state_set(&ep->com, CLOSING); - get_ep(&ep->com); + ep->com.rpl_done = 1; + ep->com.rpl_err = -ECONNRESET; + PDBG("waking up ep %p\n", ep); + wake_up(&ep->com.waitq); break; case MPA_REP_SENT: __state_set(&ep->com, CLOSING); @@ -1561,8 +1566,7 @@ static int peer_abort(struct t3cdev *tdev, struct sk_buff *skb, void *ctx) * We get 2 peer aborts from the HW. The first one must * be ignored except for scribbling that we need one more. */ - if (!(ep->com.flags & PEER_ABORT_IN_PROGRESS)) { - ep->com.flags |= PEER_ABORT_IN_PROGRESS; + if (!test_and_set_bit(PEER_ABORT_IN_PROGRESS, &ep->com.flags)) { return CPL_RET_BUF_DONE; } @@ -1589,9 +1593,13 @@ static int peer_abort(struct t3cdev *tdev, struct sk_buff *skb, void *ctx) /* * We're gonna mark this puppy DEAD, but keep * the reference on it until the ULP accepts or - * rejects the CR. + * rejects the CR. Also wake up anyone waiting + * in rdma connection migration (see iwch_accept_cr()). */ - get_ep(&ep->com); + ep->com.rpl_done = 1; + ep->com.rpl_err = -ECONNRESET; + PDBG("waking up ep %p\n", ep); + wake_up(&ep->com.waitq); break; case MORIBUND: case CLOSING: @@ -1797,6 +1805,7 @@ int iwch_reject_cr(struct iw_cm_id *cm_id, const void *pdata, u8 pdata_len) err = send_mpa_reject(ep, pdata, pdata_len); err = iwch_ep_disconnect(ep, 0, GFP_KERNEL); } + put_ep(&ep->com); return 0; } @@ -1810,8 +1819,10 @@ int iwch_accept_cr(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param) struct iwch_qp *qp = get_qhp(h, conn_param->qpn); PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid); - if (state_read(&ep->com) == DEAD) - return -ECONNRESET; + if (state_read(&ep->com) == DEAD) { + err = -ECONNRESET; + goto err; + } BUG_ON(state_read(&ep->com) != MPA_REQ_RCVD); BUG_ON(!qp); @@ -1819,15 +1830,14 @@ int iwch_accept_cr(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param) if ((conn_param->ord > qp->rhp->attr.max_rdma_read_qp_depth) || (conn_param->ird > qp->rhp->attr.max_rdma_reads_per_qp)) { abort_connection(ep, NULL, GFP_KERNEL); - return -EINVAL; + err = -EINVAL; + goto err; } cm_id->add_ref(cm_id); ep->com.cm_id = cm_id; ep->com.qp = qp; - ep->com.rpl_done = 0; - ep->com.rpl_err = 0; ep->ird = conn_param->ird; ep->ord = conn_param->ord; @@ -1836,8 +1846,6 @@ int iwch_accept_cr(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param) PDBG("%s %d ird %d ord %d\n", __func__, __LINE__, ep->ird, ep->ord); - get_ep(&ep->com); - /* bind QP to EP and move to RTS */ attrs.mpa_attr = ep->mpa_attr; attrs.max_ird = ep->ird; @@ -1855,30 +1863,31 @@ int iwch_accept_cr(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param) err = iwch_modify_qp(ep->com.qp->rhp, ep->com.qp, mask, &attrs, 1); if (err) - goto err; + goto err1; /* if needed, wait for wr_ack */ if (iwch_rqes_posted(qp)) { wait_event(ep->com.waitq, ep->com.rpl_done); err = ep->com.rpl_err; if (err) - goto err; + goto err1; } err = send_mpa_reply(ep, conn_param->private_data, conn_param->private_data_len); if (err) - goto err; + goto err1; state_set(&ep->com, FPDU_MODE); established_upcall(ep); put_ep(&ep->com); return 0; -err: +err1: ep->com.cm_id = NULL; ep->com.qp = NULL; cm_id->rem_ref(cm_id); +err: put_ep(&ep->com); return err; } @@ -2097,14 +2106,17 @@ int iwch_ep_disconnect(struct iwch_ep *ep, int abrupt, gfp_t gfp) ep->com.state = CLOSING; start_ep_timer(ep); } + set_bit(CLOSE_SENT, &ep->com.flags); break; case CLOSING: - close = 1; - if (abrupt) { - stop_ep_timer(ep); - ep->com.state = ABORTING; - } else - ep->com.state = MORIBUND; + if (!test_and_set_bit(CLOSE_SENT, &ep->com.flags)) { + close = 1; + if (abrupt) { + stop_ep_timer(ep); + ep->com.state = ABORTING; + } else + ep->com.state = MORIBUND; + } break; case MORIBUND: case ABORTING: diff --git a/drivers/infiniband/hw/cxgb3/iwch_cm.h b/drivers/infiniband/hw/cxgb3/iwch_cm.h index 43c0aea7ead..b9efadfffb4 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_cm.h +++ b/drivers/infiniband/hw/cxgb3/iwch_cm.h @@ -145,9 +145,10 @@ enum iwch_ep_state { }; enum iwch_ep_flags { - PEER_ABORT_IN_PROGRESS = (1 << 0), - ABORT_REQ_IN_PROGRESS = (1 << 1), - RELEASE_RESOURCES = (1 << 2), + PEER_ABORT_IN_PROGRESS = 0, + ABORT_REQ_IN_PROGRESS = 1, + RELEASE_RESOURCES = 2, + CLOSE_SENT = 3, }; struct iwch_ep_common { @@ -162,7 +163,7 @@ struct iwch_ep_common { wait_queue_head_t waitq; int rpl_done; int rpl_err; - u32 flags; + unsigned long flags; }; struct iwch_listen_ep { diff --git a/drivers/infiniband/hw/cxgb3/iwch_mem.c b/drivers/infiniband/hw/cxgb3/iwch_mem.c index ec49a5cbdeb..e1ec65ebb01 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_mem.c +++ b/drivers/infiniband/hw/cxgb3/iwch_mem.c @@ -39,7 +39,7 @@ #include "iwch.h" #include "iwch_provider.h" -static void iwch_finish_mem_reg(struct iwch_mr *mhp, u32 stag) +static int iwch_finish_mem_reg(struct iwch_mr *mhp, u32 stag) { u32 mmid; @@ -47,14 +47,15 @@ static void iwch_finish_mem_reg(struct iwch_mr *mhp, u32 stag) mhp->attr.stag = stag; mmid = stag >> 8; mhp->ibmr.rkey = mhp->ibmr.lkey = stag; - insert_handle(mhp->rhp, &mhp->rhp->mmidr, mhp, mmid); PDBG("%s mmid 0x%x mhp %p\n", __func__, mmid, mhp); + return insert_handle(mhp->rhp, &mhp->rhp->mmidr, mhp, mmid); } int iwch_register_mem(struct iwch_dev *rhp, struct iwch_pd *php, struct iwch_mr *mhp, int shift) { u32 stag; + int ret; if (cxio_register_phys_mem(&rhp->rdev, &stag, mhp->attr.pdid, @@ -66,9 +67,11 @@ int iwch_register_mem(struct iwch_dev *rhp, struct iwch_pd *php, mhp->attr.pbl_size, mhp->attr.pbl_addr)) return -ENOMEM; - iwch_finish_mem_reg(mhp, stag); - - return 0; + ret = iwch_finish_mem_reg(mhp, stag); + if (ret) + cxio_dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size, + mhp->attr.pbl_addr); + return ret; } int iwch_reregister_mem(struct iwch_dev *rhp, struct iwch_pd *php, @@ -77,6 +80,7 @@ int iwch_reregister_mem(struct iwch_dev *rhp, struct iwch_pd *php, int npages) { u32 stag; + int ret; /* We could support this... */ if (npages > mhp->attr.pbl_size) @@ -93,9 +97,12 @@ int iwch_reregister_mem(struct iwch_dev *rhp, struct iwch_pd *php, mhp->attr.pbl_size, mhp->attr.pbl_addr)) return -ENOMEM; - iwch_finish_mem_reg(mhp, stag); + ret = iwch_finish_mem_reg(mhp, stag); + if (ret) + cxio_dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size, + mhp->attr.pbl_addr); - return 0; + return ret; } int iwch_alloc_pbl(struct iwch_mr *mhp, int npages) diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index e2a63214008..6895523779d 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -195,7 +195,11 @@ static struct ib_cq *iwch_create_cq(struct ib_device *ibdev, int entries, int ve spin_lock_init(&chp->lock); atomic_set(&chp->refcnt, 1); init_waitqueue_head(&chp->wait); - insert_handle(rhp, &rhp->cqidr, chp, chp->cq.cqid); + if (insert_handle(rhp, &rhp->cqidr, chp, chp->cq.cqid)) { + cxio_destroy_cq(&chp->rhp->rdev, &chp->cq); + kfree(chp); + return ERR_PTR(-ENOMEM); + } if (ucontext) { struct iwch_mm_entry *mm; @@ -750,7 +754,11 @@ static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd) mhp->attr.stag = stag; mmid = (stag) >> 8; mhp->ibmw.rkey = stag; - insert_handle(rhp, &rhp->mmidr, mhp, mmid); + if (insert_handle(rhp, &rhp->mmidr, mhp, mmid)) { + cxio_deallocate_window(&rhp->rdev, mhp->attr.stag); + kfree(mhp); + return ERR_PTR(-ENOMEM); + } PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag); return &(mhp->ibmw); } @@ -778,37 +786,43 @@ static struct ib_mr *iwch_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth) struct iwch_mr *mhp; u32 mmid; u32 stag = 0; - int ret; + int ret = 0; php = to_iwch_pd(pd); rhp = php->rhp; mhp = kzalloc(sizeof(*mhp), GFP_KERNEL); if (!mhp) - return ERR_PTR(-ENOMEM); + goto err; mhp->rhp = rhp; ret = iwch_alloc_pbl(mhp, pbl_depth); - if (ret) { - kfree(mhp); - return ERR_PTR(ret); - } + if (ret) + goto err1; mhp->attr.pbl_size = pbl_depth; ret = cxio_allocate_stag(&rhp->rdev, &stag, php->pdid, mhp->attr.pbl_size, mhp->attr.pbl_addr); - if (ret) { - iwch_free_pbl(mhp); - kfree(mhp); - return ERR_PTR(ret); - } + if (ret) + goto err2; mhp->attr.pdid = php->pdid; mhp->attr.type = TPT_NON_SHARED_MR; mhp->attr.stag = stag; mhp->attr.state = 1; mmid = (stag) >> 8; mhp->ibmr.rkey = mhp->ibmr.lkey = stag; - insert_handle(rhp, &rhp->mmidr, mhp, mmid); + if (insert_handle(rhp, &rhp->mmidr, mhp, mmid)) + goto err3; + PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag); return &(mhp->ibmr); +err3: + cxio_dereg_mem(&rhp->rdev, stag, mhp->attr.pbl_size, + mhp->attr.pbl_addr); +err2: + iwch_free_pbl(mhp); +err1: + kfree(mhp); +err: + return ERR_PTR(ret); } static struct ib_fast_reg_page_list *iwch_alloc_fastreg_pbl( @@ -961,7 +975,13 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd, spin_lock_init(&qhp->lock); init_waitqueue_head(&qhp->wait); atomic_set(&qhp->refcnt, 1); - insert_handle(rhp, &rhp->qpidr, qhp, qhp->wq.qpid); + + if (insert_handle(rhp, &rhp->qpidr, qhp, qhp->wq.qpid)) { + cxio_destroy_qp(&rhp->rdev, &qhp->wq, + ucontext ? &ucontext->uctx : &rhp->rdev.uctx); + kfree(qhp); + return ERR_PTR(-ENOMEM); + } if (udata) { @@ -1418,6 +1438,7 @@ int iwch_register_device(struct iwch_dev *dev) bail2: ib_unregister_device(&dev->ibdev); bail1: + kfree(dev->ibdev.iwcm); return ret; } @@ -1430,5 +1451,6 @@ void iwch_unregister_device(struct iwch_dev *dev) device_remove_file(&dev->ibdev.dev, iwch_class_attributes[i]); ib_unregister_device(&dev->ibdev); + kfree(dev->ibdev.iwcm); return; } diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c index 27bbdc8e773..6e865347194 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_qp.c +++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c @@ -889,6 +889,7 @@ static int rdma_init(struct iwch_dev *rhp, struct iwch_qp *qhp, init_attr.qp_dma_size = (1UL << qhp->wq.size_log2); init_attr.rqe_count = iwch_rqes_posted(qhp); init_attr.flags = qhp->attr.mpa_attr.initiator ? MPA_INITIATOR : 0; + init_attr.chan = qhp->ep->l2t->smt_idx; if (peer2peer) { init_attr.rtr_type = RTR_READ; if (init_attr.ord == 0 && qhp->attr.mpa_attr.initiator) diff --git a/drivers/infiniband/hw/ehca/ehca_main.c b/drivers/infiniband/hw/ehca/ehca_main.c index fab18a2c74a..5b635aa5947 100644 --- a/drivers/infiniband/hw/ehca/ehca_main.c +++ b/drivers/infiniband/hw/ehca/ehca_main.c @@ -52,7 +52,7 @@ #include "ehca_tools.h" #include "hcp_if.h" -#define HCAD_VERSION "0028" +#define HCAD_VERSION "0029" MODULE_LICENSE("Dual BSD/GPL"); MODULE_AUTHOR("Christoph Raisch <raisch@de.ibm.com>"); @@ -64,7 +64,7 @@ static int ehca_hw_level = 0; static int ehca_poll_all_eqs = 1; int ehca_debug_level = 0; -int ehca_nr_ports = 2; +int ehca_nr_ports = -1; int ehca_use_hp_mr = 0; int ehca_port_act_time = 30; int ehca_static_rate = -1; @@ -95,8 +95,8 @@ MODULE_PARM_DESC(hw_level, "Hardware level (0: autosensing (default), " "0x10..0x14: eHCA, 0x20..0x23: eHCA2)"); MODULE_PARM_DESC(nr_ports, - "number of connected ports (-1: autodetect, 1: port one only, " - "2: two ports (default)"); + "number of connected ports (-1: autodetect (default), " + "1: port one only, 2: two ports)"); MODULE_PARM_DESC(use_hp_mr, "Use high performance MRs (default: no)"); MODULE_PARM_DESC(port_act_time, diff --git a/drivers/infiniband/hw/ehca/ehca_reqs.c b/drivers/infiniband/hw/ehca/ehca_reqs.c index 5a3d96f84c7..8fd88cd828f 100644 --- a/drivers/infiniband/hw/ehca/ehca_reqs.c +++ b/drivers/infiniband/hw/ehca/ehca_reqs.c @@ -786,7 +786,11 @@ repoll: wc->slid = cqe->rlid; wc->dlid_path_bits = cqe->dlid; wc->src_qp = cqe->remote_qp_number; - wc->wc_flags = cqe->w_completion_flags; + /* + * HW has "Immed data present" and "GRH present" in bits 6 and 5. + * SW defines those in bits 1 and 0, so we can just shift and mask. + */ + wc->wc_flags = (cqe->w_completion_flags >> 5) & 3; wc->ex.imm_data = cpu_to_be32(cqe->immediate_data); wc->sl = cqe->service_level; diff --git a/drivers/infiniband/hw/ehca/ehca_sqp.c b/drivers/infiniband/hw/ehca/ehca_sqp.c index c568b28f4e2..8c1213f8916 100644 --- a/drivers/infiniband/hw/ehca/ehca_sqp.c +++ b/drivers/infiniband/hw/ehca/ehca_sqp.c @@ -125,14 +125,30 @@ struct ib_perf { u8 data[192]; } __attribute__ ((packed)); +/* TC/SL/FL packed into 32 bits, as in ClassPortInfo */ +struct tcslfl { + u32 tc:8; + u32 sl:4; + u32 fl:20; +} __attribute__ ((packed)); + +/* IP Version/TC/FL packed into 32 bits, as in GRH */ +struct vertcfl { + u32 ver:4; + u32 tc:8; + u32 fl:20; +} __attribute__ ((packed)); static int ehca_process_perf(struct ib_device *ibdev, u8 port_num, + struct ib_wc *in_wc, struct ib_grh *in_grh, struct ib_mad *in_mad, struct ib_mad *out_mad) { struct ib_perf *in_perf = (struct ib_perf *)in_mad; struct ib_perf *out_perf = (struct ib_perf *)out_mad; struct ib_class_port_info *poi = (struct ib_class_port_info *)out_perf->data; + struct tcslfl *tcslfl = + (struct tcslfl *)&poi->redirect_tcslfl; struct ehca_shca *shca = container_of(ibdev, struct ehca_shca, ib_device); struct ehca_sport *sport = &shca->sport[port_num - 1]; @@ -158,10 +174,29 @@ static int ehca_process_perf(struct ib_device *ibdev, u8 port_num, poi->base_version = 1; poi->class_version = 1; poi->resp_time_value = 18; - poi->redirect_lid = sport->saved_attr.lid; - poi->redirect_qp = sport->pma_qp_nr; + + /* copy local routing information from WC where applicable */ + tcslfl->sl = in_wc->sl; + poi->redirect_lid = + sport->saved_attr.lid | in_wc->dlid_path_bits; + poi->redirect_qp = sport->pma_qp_nr; poi->redirect_qkey = IB_QP1_QKEY; - poi->redirect_pkey = IB_DEFAULT_PKEY_FULL; + + ehca_query_pkey(ibdev, port_num, in_wc->pkey_index, + &poi->redirect_pkey); + + /* if request was globally routed, copy route info */ + if (in_grh) { + struct vertcfl *vertcfl = + (struct vertcfl *)&in_grh->version_tclass_flow; + memcpy(poi->redirect_gid, in_grh->dgid.raw, + sizeof(poi->redirect_gid)); + tcslfl->tc = vertcfl->tc; + tcslfl->fl = vertcfl->fl; + } else + /* else only fill in default GID */ + ehca_query_gid(ibdev, port_num, 0, + (union ib_gid *)&poi->redirect_gid); ehca_dbg(ibdev, "ehca_pma_lid=%x ehca_pma_qp=%x", sport->saved_attr.lid, sport->pma_qp_nr); @@ -183,8 +218,7 @@ perf_reply: int ehca_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num, struct ib_wc *in_wc, struct ib_grh *in_grh, - struct ib_mad *in_mad, - struct ib_mad *out_mad) + struct ib_mad *in_mad, struct ib_mad *out_mad) { int ret; @@ -196,7 +230,8 @@ int ehca_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num, return IB_MAD_RESULT_SUCCESS; ehca_dbg(ibdev, "port_num=%x src_qp=%x", port_num, in_wc->src_qp); - ret = ehca_process_perf(ibdev, port_num, in_mad, out_mad); + ret = ehca_process_perf(ibdev, port_num, in_wc, in_grh, + in_mad, out_mad); return ret; } diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c index 23173982b32..38a28700661 100644 --- a/drivers/infiniband/hw/ipath/ipath_file_ops.c +++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c @@ -1616,7 +1616,7 @@ static int try_alloc_port(struct ipath_devdata *dd, int port, pd->port_cnt = 1; port_fp(fp) = pd; pd->port_pid = get_pid(task_pid(current)); - strncpy(pd->port_comm, current->comm, sizeof(pd->port_comm)); + strlcpy(pd->port_comm, current->comm, sizeof(pd->port_comm)); ipath_stats.sps_ports++; ret = 0; } else diff --git a/drivers/infiniband/hw/ipath/ipath_mad.c b/drivers/infiniband/hw/ipath/ipath_mad.c index 16a702d4601..ceb98ee7866 100644 --- a/drivers/infiniband/hw/ipath/ipath_mad.c +++ b/drivers/infiniband/hw/ipath/ipath_mad.c @@ -60,7 +60,7 @@ static int recv_subn_get_nodedescription(struct ib_smp *smp, if (smp->attr_mod) smp->status |= IB_SMP_INVALID_FIELD; - strncpy(smp->data, ibdev->node_desc, sizeof(smp->data)); + memcpy(smp->data, ibdev->node_desc, sizeof(smp->data)); return reply(smp); } diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index ae3d7590346..3cb3f47a10b 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -342,6 +342,9 @@ static struct ib_ucontext *mlx4_ib_alloc_ucontext(struct ib_device *ibdev, struct mlx4_ib_alloc_ucontext_resp resp; int err; + if (!dev->ib_active) + return ERR_PTR(-EAGAIN); + resp.qp_tab_size = dev->dev->caps.num_qps; resp.bf_reg_size = dev->dev->caps.bf_reg_size; resp.bf_regs_per_page = dev->dev->caps.bf_regs_per_page; @@ -540,15 +543,11 @@ static struct device_attribute *mlx4_class_attributes[] = { static void *mlx4_ib_add(struct mlx4_dev *dev) { - static int mlx4_ib_version_printed; struct mlx4_ib_dev *ibdev; int num_ports = 0; int i; - if (!mlx4_ib_version_printed) { - printk(KERN_INFO "%s", mlx4_ib_version); - ++mlx4_ib_version_printed; - } + printk_once(KERN_INFO "%s", mlx4_ib_version); mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_IB) num_ports++; @@ -673,6 +672,8 @@ static void *mlx4_ib_add(struct mlx4_dev *dev) goto err_reg; } + ibdev->ib_active = true; + return ibdev; err_reg: @@ -729,6 +730,7 @@ static void mlx4_ib_event(struct mlx4_dev *dev, void *ibdev_ptr, break; case MLX4_DEV_EVENT_CATASTROPHIC_ERROR: + ibdev->ib_active = false; ibev.event = IB_EVENT_DEVICE_FATAL; break; diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 8a7dd6795fa..3486d7675e5 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -175,6 +175,7 @@ struct mlx4_ib_dev { spinlock_t sm_lock; struct mutex cap_mask_mutex; + bool ib_active; }; static inline struct mlx4_ib_dev *to_mdev(struct ib_device *ibdev) diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index c4a02648c8a..219b10397b4 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -615,10 +615,12 @@ static enum mlx4_qp_state to_mlx4_state(enum ib_qp_state state) } static void mlx4_ib_lock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq) + __acquires(&send_cq->lock) __acquires(&recv_cq->lock) { - if (send_cq == recv_cq) + if (send_cq == recv_cq) { spin_lock_irq(&send_cq->lock); - else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) { + __acquire(&recv_cq->lock); + } else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) { spin_lock_irq(&send_cq->lock); spin_lock_nested(&recv_cq->lock, SINGLE_DEPTH_NESTING); } else { @@ -628,10 +630,12 @@ static void mlx4_ib_lock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv } static void mlx4_ib_unlock_cqs(struct mlx4_ib_cq *send_cq, struct mlx4_ib_cq *recv_cq) + __releases(&send_cq->lock) __releases(&recv_cq->lock) { - if (send_cq == recv_cq) + if (send_cq == recv_cq) { + __release(&recv_cq->lock); spin_unlock_irq(&send_cq->lock); - else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) { + } else if (send_cq->mcq.cqn < recv_cq->mcq.cqn) { spin_unlock(&recv_cq->lock); spin_unlock_irq(&send_cq->lock); } else { diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c index 65ad359fdf1..056b2a4c697 100644 --- a/drivers/infiniband/hw/mthca/mthca_catas.c +++ b/drivers/infiniband/hw/mthca/mthca_catas.c @@ -88,6 +88,7 @@ static void handle_catas(struct mthca_dev *dev) event.device = &dev->ib_dev; event.event = IB_EVENT_DEVICE_FATAL; event.element.port_num = 0; + dev->active = false; ib_dispatch_event(&event); diff --git a/drivers/infiniband/hw/mthca/mthca_config_reg.h b/drivers/infiniband/hw/mthca/mthca_config_reg.h index 75671f75cac..155bc66395b 100644 --- a/drivers/infiniband/hw/mthca/mthca_config_reg.h +++ b/drivers/infiniband/hw/mthca/mthca_config_reg.h @@ -34,8 +34,6 @@ #ifndef MTHCA_CONFIG_REG_H #define MTHCA_CONFIG_REG_H -#include <asm/page.h> - #define MTHCA_HCR_BASE 0x80680 #define MTHCA_HCR_SIZE 0x0001c #define MTHCA_ECR_BASE 0x80700 diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h index 9ef611f6dd3..7e6a6d64ad4 100644 --- a/drivers/infiniband/hw/mthca/mthca_dev.h +++ b/drivers/infiniband/hw/mthca/mthca_dev.h @@ -357,6 +357,7 @@ struct mthca_dev { struct ib_ah *sm_ah[MTHCA_MAX_PORTS]; spinlock_t sm_lock; u8 rate[MTHCA_MAX_PORTS]; + bool active; }; #ifdef CONFIG_INFINIBAND_MTHCA_DEBUG diff --git a/drivers/infiniband/hw/mthca/mthca_eq.c b/drivers/infiniband/hw/mthca/mthca_eq.c index 90e4e450a12..8c31fa36e95 100644 --- a/drivers/infiniband/hw/mthca/mthca_eq.c +++ b/drivers/infiniband/hw/mthca/mthca_eq.c @@ -829,27 +829,34 @@ int mthca_init_eq_table(struct mthca_dev *dev) if (dev->mthca_flags & MTHCA_FLAG_MSI_X) { static const char *eq_name[] = { - [MTHCA_EQ_COMP] = DRV_NAME " (comp)", - [MTHCA_EQ_ASYNC] = DRV_NAME " (async)", - [MTHCA_EQ_CMD] = DRV_NAME " (cmd)" + [MTHCA_EQ_COMP] = DRV_NAME "-comp", + [MTHCA_EQ_ASYNC] = DRV_NAME "-async", + [MTHCA_EQ_CMD] = DRV_NAME "-cmd" }; for (i = 0; i < MTHCA_NUM_EQ; ++i) { + snprintf(dev->eq_table.eq[i].irq_name, + IB_DEVICE_NAME_MAX, + "%s@pci:%s", eq_name[i], + pci_name(dev->pdev)); err = request_irq(dev->eq_table.eq[i].msi_x_vector, mthca_is_memfree(dev) ? mthca_arbel_msi_x_interrupt : mthca_tavor_msi_x_interrupt, - 0, eq_name[i], dev->eq_table.eq + i); + 0, dev->eq_table.eq[i].irq_name, + dev->eq_table.eq + i); if (err) goto err_out_cmd; dev->eq_table.eq[i].have_irq = 1; } } else { + snprintf(dev->eq_table.eq[0].irq_name, IB_DEVICE_NAME_MAX, + DRV_NAME "@pci:%s", pci_name(dev->pdev)); err = request_irq(dev->pdev->irq, mthca_is_memfree(dev) ? mthca_arbel_interrupt : mthca_tavor_interrupt, - IRQF_SHARED, DRV_NAME, dev); + IRQF_SHARED, dev->eq_table.eq[0].irq_name, dev); if (err) goto err_out_cmd; dev->eq_table.have_irq = 1; diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c index 13da9f1d24c..b01b2898787 100644 --- a/drivers/infiniband/hw/mthca/mthca_main.c +++ b/drivers/infiniband/hw/mthca/mthca_main.c @@ -1116,6 +1116,8 @@ static int __mthca_init_one(struct pci_dev *pdev, int hca_type) pci_set_drvdata(pdev, mdev); mdev->hca_type = hca_type; + mdev->active = true; + return 0; err_unregister: @@ -1215,15 +1217,11 @@ int __mthca_restart_one(struct pci_dev *pdev) static int __devinit mthca_init_one(struct pci_dev *pdev, const struct pci_device_id *id) { - static int mthca_version_printed = 0; int ret; mutex_lock(&mthca_device_mutex); - if (!mthca_version_printed) { - printk(KERN_INFO "%s", mthca_version); - ++mthca_version_printed; - } + printk_once(KERN_INFO "%s", mthca_version); if (id->driver_data >= ARRAY_SIZE(mthca_hca_table)) { printk(KERN_ERR PFX "%s has invalid driver data %lx\n", diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index 87ad889e367..bcf7a401482 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -334,6 +334,9 @@ static struct ib_ucontext *mthca_alloc_ucontext(struct ib_device *ibdev, struct mthca_ucontext *context; int err; + if (!(to_mdev(ibdev)->active)) + return ERR_PTR(-EAGAIN); + memset(&uresp, 0, sizeof uresp); uresp.qp_tab_size = to_mdev(ibdev)->limits.num_qps; diff --git a/drivers/infiniband/hw/mthca/mthca_provider.h b/drivers/infiniband/hw/mthca/mthca_provider.h index c621f8794b8..90f4c4d2e98 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.h +++ b/drivers/infiniband/hw/mthca/mthca_provider.h @@ -113,6 +113,7 @@ struct mthca_eq { int nent; struct mthca_buf_list *page_list; struct mthca_mr mr; + char irq_name[IB_DEVICE_NAME_MAX]; }; struct mthca_av; diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index f5081bfde6d..c10576fa60c 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -1319,10 +1319,12 @@ int mthca_alloc_qp(struct mthca_dev *dev, } static void mthca_lock_cqs(struct mthca_cq *send_cq, struct mthca_cq *recv_cq) + __acquires(&send_cq->lock) __acquires(&recv_cq->lock) { - if (send_cq == recv_cq) + if (send_cq == recv_cq) { spin_lock_irq(&send_cq->lock); - else if (send_cq->cqn < recv_cq->cqn) { + __acquire(&recv_cq->lock); + } else if (send_cq->cqn < recv_cq->cqn) { spin_lock_irq(&send_cq->lock); spin_lock_nested(&recv_cq->lock, SINGLE_DEPTH_NESTING); } else { @@ -1332,10 +1334,12 @@ static void mthca_lock_cqs(struct mthca_cq *send_cq, struct mthca_cq *recv_cq) } static void mthca_unlock_cqs(struct mthca_cq *send_cq, struct mthca_cq *recv_cq) + __releases(&send_cq->lock) __releases(&recv_cq->lock) { - if (send_cq == recv_cq) + if (send_cq == recv_cq) { + __release(&recv_cq->lock); spin_unlock_irq(&send_cq->lock); - else if (send_cq->cqn < recv_cq->cqn) { + } else if (send_cq->cqn < recv_cq->cqn) { spin_unlock(&recv_cq->lock); spin_unlock_irq(&send_cq->lock); } else { diff --git a/drivers/infiniband/hw/mthca/mthca_reset.c b/drivers/infiniband/hw/mthca/mthca_reset.c index acb6817f606..2a13a163d33 100644 --- a/drivers/infiniband/hw/mthca/mthca_reset.c +++ b/drivers/infiniband/hw/mthca/mthca_reset.c @@ -30,7 +30,6 @@ * SOFTWARE. */ -#include <linux/init.h> #include <linux/errno.h> #include <linux/pci.h> #include <linux/delay.h> diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h index bf1720f7f35..bcc6abc4faf 100644 --- a/drivers/infiniband/hw/nes/nes.h +++ b/drivers/infiniband/hw/nes/nes.h @@ -523,7 +523,7 @@ int nes_cm_disconn(struct nes_qp *); void nes_cm_disconn_worker(void *); /* nes_verbs.c */ -int nes_hw_modify_qp(struct nes_device *, struct nes_qp *, u32, u32); +int nes_hw_modify_qp(struct nes_device *, struct nes_qp *, u32, u32, u32); int nes_modify_qp(struct ib_qp *, struct ib_qp_attr *, int, struct ib_udata *); struct nes_ib_device *nes_init_ofa_device(struct net_device *); void nes_destroy_ofa_device(struct nes_ib_device *); diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c index 114b802771a..73473db1986 100644 --- a/drivers/infiniband/hw/nes/nes_cm.c +++ b/drivers/infiniband/hw/nes/nes_cm.c @@ -2450,19 +2450,16 @@ static int nes_cm_init_tsa_conn(struct nes_qp *nesqp, struct nes_cm_node *cm_nod */ int nes_cm_disconn(struct nes_qp *nesqp) { - unsigned long flags; - - spin_lock_irqsave(&nesqp->lock, flags); - if (nesqp->disconn_pending == 0) { - nesqp->disconn_pending++; - spin_unlock_irqrestore(&nesqp->lock, flags); - /* init our disconnect work element, to */ - INIT_WORK(&nesqp->disconn_work, nes_disconnect_worker); + struct disconn_work *work; - queue_work(g_cm_core->disconn_wq, &nesqp->disconn_work); - } else - spin_unlock_irqrestore(&nesqp->lock, flags); + work = kzalloc(sizeof *work, GFP_ATOMIC); + if (!work) + return -ENOMEM; /* Timer will clean up */ + nes_add_ref(&nesqp->ibqp); + work->nesqp = nesqp; + INIT_WORK(&work->work, nes_disconnect_worker); + queue_work(g_cm_core->disconn_wq, &work->work); return 0; } @@ -2472,11 +2469,14 @@ int nes_cm_disconn(struct nes_qp *nesqp) */ static void nes_disconnect_worker(struct work_struct *work) { - struct nes_qp *nesqp = container_of(work, struct nes_qp, disconn_work); + struct disconn_work *dwork = container_of(work, struct disconn_work, work); + struct nes_qp *nesqp = dwork->nesqp; + kfree(dwork); nes_debug(NES_DBG_CM, "processing AEQE id 0x%04X for QP%u.\n", nesqp->last_aeq, nesqp->hwqp.qp_id); nes_cm_disconn_true(nesqp); + nes_rem_ref(&nesqp->ibqp); } @@ -2493,7 +2493,12 @@ static int nes_cm_disconn_true(struct nes_qp *nesqp) u16 last_ae; u8 original_hw_tcp_state; u8 original_ibqp_state; - u8 issued_disconnect_reset = 0; + enum iw_cm_event_type disconn_status = IW_CM_EVENT_STATUS_OK; + int issue_disconn = 0; + int issue_close = 0; + int issue_flush = 0; + u32 flush_q = NES_CQP_FLUSH_RQ; + struct ib_event ibevent; if (!nesqp) { nes_debug(NES_DBG_CM, "disconnect_worker nesqp is NULL\n"); @@ -2517,24 +2522,55 @@ static int nes_cm_disconn_true(struct nes_qp *nesqp) original_ibqp_state = nesqp->ibqp_state; last_ae = nesqp->last_aeq; + if (nesqp->term_flags) { + issue_disconn = 1; + issue_close = 1; + nesqp->cm_id = NULL; + if (nesqp->flush_issued == 0) { + nesqp->flush_issued = 1; + issue_flush = 1; + } + } else if ((original_hw_tcp_state == NES_AEQE_TCP_STATE_CLOSE_WAIT) || + ((original_ibqp_state == IB_QPS_RTS) && + (last_ae == NES_AEQE_AEID_LLP_CONNECTION_RESET))) { + issue_disconn = 1; + if (last_ae == NES_AEQE_AEID_LLP_CONNECTION_RESET) + disconn_status = IW_CM_EVENT_STATUS_RESET; + } + + if (((original_hw_tcp_state == NES_AEQE_TCP_STATE_CLOSED) || + (original_hw_tcp_state == NES_AEQE_TCP_STATE_TIME_WAIT) || + (last_ae == NES_AEQE_AEID_RDMAP_ROE_BAD_LLP_CLOSE) || + (last_ae == NES_AEQE_AEID_LLP_CONNECTION_RESET))) { + issue_close = 1; + nesqp->cm_id = NULL; + if (nesqp->flush_issued == 0) { + nesqp->flush_issued = 1; + issue_flush = 1; + } + } + + spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_CM, "set ibqp_state=%u\n", nesqp->ibqp_state); + if ((issue_flush) && (nesqp->destroyed == 0)) { + /* Flush the queue(s) */ + if (nesqp->hw_iwarp_state >= NES_AEQE_IWARP_STATE_TERMINATE) + flush_q |= NES_CQP_FLUSH_SQ; + flush_wqes(nesvnic->nesdev, nesqp, flush_q, 1); - if ((nesqp->cm_id) && (cm_id->event_handler)) { - if ((original_hw_tcp_state == NES_AEQE_TCP_STATE_CLOSE_WAIT) || - ((original_ibqp_state == IB_QPS_RTS) && - (last_ae == NES_AEQE_AEID_LLP_CONNECTION_RESET))) { + if (nesqp->term_flags) { + ibevent.device = nesqp->ibqp.device; + ibevent.event = nesqp->terminate_eventtype; + ibevent.element.qp = &nesqp->ibqp; + nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); + } + } + + if ((cm_id) && (cm_id->event_handler)) { + if (issue_disconn) { atomic_inc(&cm_disconnects); cm_event.event = IW_CM_EVENT_DISCONNECT; - if (last_ae == NES_AEQE_AEID_LLP_CONNECTION_RESET) { - cm_event.status = IW_CM_EVENT_STATUS_RESET; - nes_debug(NES_DBG_CM, "Generating a CM " - "Disconnect Event (status reset) for " - "QP%u, cm_id = %p. \n", - nesqp->hwqp.qp_id, cm_id); - } else - cm_event.status = IW_CM_EVENT_STATUS_OK; - + cm_event.status = disconn_status; cm_event.local_addr = cm_id->local_addr; cm_event.remote_addr = cm_id->remote_addr; cm_event.private_data = NULL; @@ -2547,29 +2583,14 @@ static int nes_cm_disconn_true(struct nes_qp *nesqp) nesqp->hwqp.sq_tail, cm_id, atomic_read(&nesqp->refcount)); - spin_unlock_irqrestore(&nesqp->lock, flags); ret = cm_id->event_handler(cm_id, &cm_event); if (ret) nes_debug(NES_DBG_CM, "OFA CM event_handler " "returned, ret=%d\n", ret); - spin_lock_irqsave(&nesqp->lock, flags); } - nesqp->disconn_pending = 0; - /* There might have been another AE while the lock was released */ - original_hw_tcp_state = nesqp->hw_tcp_state; - original_ibqp_state = nesqp->ibqp_state; - last_ae = nesqp->last_aeq; - - if ((issued_disconnect_reset == 0) && (nesqp->cm_id) && - ((original_hw_tcp_state == NES_AEQE_TCP_STATE_CLOSED) || - (original_hw_tcp_state == NES_AEQE_TCP_STATE_TIME_WAIT) || - (last_ae == NES_AEQE_AEID_RDMAP_ROE_BAD_LLP_CLOSE) || - (last_ae == NES_AEQE_AEID_LLP_CONNECTION_RESET))) { + if (issue_close) { atomic_inc(&cm_closes); - nesqp->cm_id = NULL; - nesqp->in_disconnect = 0; - spin_unlock_irqrestore(&nesqp->lock, flags); nes_disconnect(nesqp, 1); cm_id->provider_data = nesqp; @@ -2588,28 +2609,7 @@ static int nes_cm_disconn_true(struct nes_qp *nesqp) } cm_id->rem_ref(cm_id); - - spin_lock_irqsave(&nesqp->lock, flags); - if (nesqp->flush_issued == 0) { - nesqp->flush_issued = 1; - spin_unlock_irqrestore(&nesqp->lock, flags); - flush_wqes(nesvnic->nesdev, nesqp, - NES_CQP_FLUSH_RQ, 1); - } else - spin_unlock_irqrestore(&nesqp->lock, flags); - } else { - cm_id = nesqp->cm_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - /* check to see if the inbound reset beat the outbound reset */ - if ((!cm_id) && (last_ae==NES_AEQE_AEID_RESET_SENT)) { - nes_debug(NES_DBG_CM, "QP%u: Decing refcount " - "due to inbound reset beating the " - "outbound reset.\n", nesqp->hwqp.qp_id); - } } - } else { - nesqp->disconn_pending = 0; - spin_unlock_irqrestore(&nesqp->lock, flags); } return 0; diff --git a/drivers/infiniband/hw/nes/nes_cm.h b/drivers/infiniband/hw/nes/nes_cm.h index 8b7e7c0e496..90e8e4d8a5c 100644 --- a/drivers/infiniband/hw/nes/nes_cm.h +++ b/drivers/infiniband/hw/nes/nes_cm.h @@ -410,8 +410,6 @@ struct nes_cm_ops { int schedule_nes_timer(struct nes_cm_node *, struct sk_buff *, enum nes_timer_type, int, int); -int nes_cm_disconn(struct nes_qp *); - int nes_accept(struct iw_cm_id *, struct iw_cm_conn_param *); int nes_reject(struct iw_cm_id *, const void *, u8); int nes_connect(struct iw_cm_id *, struct iw_cm_conn_param *); diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index 4a84d02ece0..63a1a8e1e8a 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -74,6 +74,8 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, static void process_critical_error(struct nes_device *nesdev); static void nes_process_mac_intr(struct nes_device *nesdev, u32 mac_number); static unsigned int nes_reset_adapter_ne020(struct nes_device *nesdev, u8 *OneG_Mode); +static void nes_terminate_timeout(unsigned long context); +static void nes_terminate_start_timer(struct nes_qp *nesqp); #ifdef CONFIG_INFINIBAND_NES_DEBUG static unsigned char *nes_iwarp_state_str[] = { @@ -2903,6 +2905,417 @@ static void nes_cqp_ce_handler(struct nes_device *nesdev, struct nes_hw_cq *cq) } +static u8 *locate_mpa(u8 *pkt, u32 aeq_info) +{ + u16 pkt_len; + + if (aeq_info & NES_AEQE_Q2_DATA_ETHERNET) { + /* skip over ethernet header */ + pkt_len = be16_to_cpu(*(u16 *)(pkt + ETH_HLEN - 2)); + pkt += ETH_HLEN; + + /* Skip over IP and TCP headers */ + pkt += 4 * (pkt[0] & 0x0f); + pkt += 4 * ((pkt[12] >> 4) & 0x0f); + } + return pkt; +} + +/* Determine if incoming error pkt is rdma layer */ +static u32 iwarp_opcode(struct nes_qp *nesqp, u32 aeq_info) +{ + u8 *pkt; + u16 *mpa; + u32 opcode = 0xffffffff; + + if (aeq_info & NES_AEQE_Q2_DATA_WRITTEN) { + pkt = nesqp->hwqp.q2_vbase + BAD_FRAME_OFFSET; + mpa = (u16 *)locate_mpa(pkt, aeq_info); + opcode = be16_to_cpu(mpa[1]) & 0xf; + } + + return opcode; +} + +/* Build iWARP terminate header */ +static int nes_bld_terminate_hdr(struct nes_qp *nesqp, u16 async_event_id, u32 aeq_info) +{ + u8 *pkt = nesqp->hwqp.q2_vbase + BAD_FRAME_OFFSET; + u16 ddp_seg_len; + int copy_len = 0; + u8 is_tagged = 0; + u8 flush_code = 0; + struct nes_terminate_hdr *termhdr; + + termhdr = (struct nes_terminate_hdr *)nesqp->hwqp.q2_vbase; + memset(termhdr, 0, 64); + + if (aeq_info & NES_AEQE_Q2_DATA_WRITTEN) { + + /* Use data from offending packet to fill in ddp & rdma hdrs */ + pkt = locate_mpa(pkt, aeq_info); + ddp_seg_len = be16_to_cpu(*(u16 *)pkt); + if (ddp_seg_len) { + copy_len = 2; + termhdr->hdrct = DDP_LEN_FLAG; + if (pkt[2] & 0x80) { + is_tagged = 1; + if (ddp_seg_len >= TERM_DDP_LEN_TAGGED) { + copy_len += TERM_DDP_LEN_TAGGED; + termhdr->hdrct |= DDP_HDR_FLAG; + } + } else { + if (ddp_seg_len >= TERM_DDP_LEN_UNTAGGED) { + copy_len += TERM_DDP_LEN_UNTAGGED; + termhdr->hdrct |= DDP_HDR_FLAG; + } + + if (ddp_seg_len >= (TERM_DDP_LEN_UNTAGGED + TERM_RDMA_LEN)) { + if ((pkt[3] & RDMA_OPCODE_MASK) == RDMA_READ_REQ_OPCODE) { + copy_len += TERM_RDMA_LEN; + termhdr->hdrct |= RDMA_HDR_FLAG; + } + } + } + } + } + + switch (async_event_id) { + case NES_AEQE_AEID_AMP_UNALLOCATED_STAG: + switch (iwarp_opcode(nesqp, aeq_info)) { + case IWARP_OPCODE_WRITE: + flush_code = IB_WC_LOC_PROT_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_TAGGED_BUFFER; + termhdr->error_code = DDP_TAGGED_INV_STAG; + break; + default: + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_INV_STAG; + } + break; + case NES_AEQE_AEID_AMP_INVALID_STAG: + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_INV_STAG; + break; + case NES_AEQE_AEID_AMP_BAD_QP: + flush_code = IB_WC_LOC_QP_OP_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER; + termhdr->error_code = DDP_UNTAGGED_INV_QN; + break; + case NES_AEQE_AEID_AMP_BAD_STAG_KEY: + case NES_AEQE_AEID_AMP_BAD_STAG_INDEX: + switch (iwarp_opcode(nesqp, aeq_info)) { + case IWARP_OPCODE_SEND_INV: + case IWARP_OPCODE_SEND_SE_INV: + flush_code = IB_WC_REM_OP_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_OP; + termhdr->error_code = RDMAP_CANT_INV_STAG; + break; + default: + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_INV_STAG; + } + break; + case NES_AEQE_AEID_AMP_BOUNDS_VIOLATION: + if (aeq_info & (NES_AEQE_Q2_DATA_ETHERNET | NES_AEQE_Q2_DATA_MPA)) { + flush_code = IB_WC_LOC_PROT_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_TAGGED_BUFFER; + termhdr->error_code = DDP_TAGGED_BOUNDS; + } else { + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_INV_BOUNDS; + } + break; + case NES_AEQE_AEID_AMP_RIGHTS_VIOLATION: + case NES_AEQE_AEID_AMP_INVALIDATE_NO_REMOTE_ACCESS_RIGHTS: + case NES_AEQE_AEID_PRIV_OPERATION_DENIED: + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_ACCESS; + break; + case NES_AEQE_AEID_AMP_TO_WRAP: + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_TO_WRAP; + break; + case NES_AEQE_AEID_AMP_BAD_PD: + switch (iwarp_opcode(nesqp, aeq_info)) { + case IWARP_OPCODE_WRITE: + flush_code = IB_WC_LOC_PROT_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_TAGGED_BUFFER; + termhdr->error_code = DDP_TAGGED_UNASSOC_STAG; + break; + case IWARP_OPCODE_SEND_INV: + case IWARP_OPCODE_SEND_SE_INV: + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_CANT_INV_STAG; + break; + default: + flush_code = IB_WC_REM_ACCESS_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_PROT; + termhdr->error_code = RDMAP_UNASSOC_STAG; + } + break; + case NES_AEQE_AEID_LLP_RECEIVED_MARKER_AND_LENGTH_FIELDS_DONT_MATCH: + flush_code = IB_WC_LOC_LEN_ERR; + termhdr->layer_etype = (LAYER_MPA << 4) | DDP_LLP; + termhdr->error_code = MPA_MARKER; + break; + case NES_AEQE_AEID_LLP_RECEIVED_MPA_CRC_ERROR: + flush_code = IB_WC_GENERAL_ERR; + termhdr->layer_etype = (LAYER_MPA << 4) | DDP_LLP; + termhdr->error_code = MPA_CRC; + break; + case NES_AEQE_AEID_LLP_SEGMENT_TOO_LARGE: + case NES_AEQE_AEID_LLP_SEGMENT_TOO_SMALL: + flush_code = IB_WC_LOC_LEN_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_CATASTROPHIC; + termhdr->error_code = DDP_CATASTROPHIC_LOCAL; + break; + case NES_AEQE_AEID_DDP_LCE_LOCAL_CATASTROPHIC: + case NES_AEQE_AEID_DDP_NO_L_BIT: + flush_code = IB_WC_FATAL_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_CATASTROPHIC; + termhdr->error_code = DDP_CATASTROPHIC_LOCAL; + break; + case NES_AEQE_AEID_DDP_INVALID_MSN_GAP_IN_MSN: + case NES_AEQE_AEID_DDP_INVALID_MSN_RANGE_IS_NOT_VALID: + flush_code = IB_WC_GENERAL_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER; + termhdr->error_code = DDP_UNTAGGED_INV_MSN_RANGE; + break; + case NES_AEQE_AEID_DDP_UBE_DDP_MESSAGE_TOO_LONG_FOR_AVAILABLE_BUFFER: + flush_code = IB_WC_LOC_LEN_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER; + termhdr->error_code = DDP_UNTAGGED_INV_TOO_LONG; + break; + case NES_AEQE_AEID_DDP_UBE_INVALID_DDP_VERSION: + flush_code = IB_WC_GENERAL_ERR; + if (is_tagged) { + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_TAGGED_BUFFER; + termhdr->error_code = DDP_TAGGED_INV_DDP_VER; + } else { + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER; + termhdr->error_code = DDP_UNTAGGED_INV_DDP_VER; + } + break; + case NES_AEQE_AEID_DDP_UBE_INVALID_MO: + flush_code = IB_WC_GENERAL_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER; + termhdr->error_code = DDP_UNTAGGED_INV_MO; + break; + case NES_AEQE_AEID_DDP_UBE_INVALID_MSN_NO_BUFFER_AVAILABLE: + flush_code = IB_WC_REM_OP_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER; + termhdr->error_code = DDP_UNTAGGED_INV_MSN_NO_BUF; + break; + case NES_AEQE_AEID_DDP_UBE_INVALID_QN: + flush_code = IB_WC_GENERAL_ERR; + termhdr->layer_etype = (LAYER_DDP << 4) | DDP_UNTAGGED_BUFFER; + termhdr->error_code = DDP_UNTAGGED_INV_QN; + break; + case NES_AEQE_AEID_RDMAP_ROE_INVALID_RDMAP_VERSION: + flush_code = IB_WC_GENERAL_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_OP; + termhdr->error_code = RDMAP_INV_RDMAP_VER; + break; + case NES_AEQE_AEID_RDMAP_ROE_UNEXPECTED_OPCODE: + flush_code = IB_WC_LOC_QP_OP_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_OP; + termhdr->error_code = RDMAP_UNEXPECTED_OP; + break; + default: + flush_code = IB_WC_FATAL_ERR; + termhdr->layer_etype = (LAYER_RDMA << 4) | RDMAP_REMOTE_OP; + termhdr->error_code = RDMAP_UNSPECIFIED; + break; + } + + if (copy_len) + memcpy(termhdr + 1, pkt, copy_len); + + if ((flush_code) && ((NES_AEQE_INBOUND_RDMA & aeq_info) == 0)) { + if (aeq_info & NES_AEQE_SQ) + nesqp->term_sq_flush_code = flush_code; + else + nesqp->term_rq_flush_code = flush_code; + } + + return sizeof(struct nes_terminate_hdr) + copy_len; +} + +static void nes_terminate_connection(struct nes_device *nesdev, struct nes_qp *nesqp, + struct nes_hw_aeqe *aeqe, enum ib_event_type eventtype) +{ + u64 context; + unsigned long flags; + u32 aeq_info; + u16 async_event_id; + u8 tcp_state; + u8 iwarp_state; + u32 termlen = 0; + u32 mod_qp_flags = NES_CQP_QP_IWARP_STATE_TERMINATE | + NES_CQP_QP_TERM_DONT_SEND_FIN; + struct nes_adapter *nesadapter = nesdev->nesadapter; + + if (nesqp->term_flags & NES_TERM_SENT) + return; /* Sanity check */ + + aeq_info = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_MISC_IDX]); + tcp_state = (aeq_info & NES_AEQE_TCP_STATE_MASK) >> NES_AEQE_TCP_STATE_SHIFT; + iwarp_state = (aeq_info & NES_AEQE_IWARP_STATE_MASK) >> NES_AEQE_IWARP_STATE_SHIFT; + async_event_id = (u16)aeq_info; + + context = (unsigned long)nesadapter->qp_table[le32_to_cpu( + aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX]) - NES_FIRST_QPN]; + if (!context) { + WARN_ON(!context); + return; + } + + nesqp = (struct nes_qp *)(unsigned long)context; + spin_lock_irqsave(&nesqp->lock, flags); + nesqp->hw_iwarp_state = iwarp_state; + nesqp->hw_tcp_state = tcp_state; + nesqp->last_aeq = async_event_id; + nesqp->terminate_eventtype = eventtype; + spin_unlock_irqrestore(&nesqp->lock, flags); + + if (nesadapter->send_term_ok) + termlen = nes_bld_terminate_hdr(nesqp, async_event_id, aeq_info); + else + mod_qp_flags |= NES_CQP_QP_TERM_DONT_SEND_TERM_MSG; + + nes_terminate_start_timer(nesqp); + nesqp->term_flags |= NES_TERM_SENT; + nes_hw_modify_qp(nesdev, nesqp, mod_qp_flags, termlen, 0); +} + +static void nes_terminate_send_fin(struct nes_device *nesdev, + struct nes_qp *nesqp, struct nes_hw_aeqe *aeqe) +{ + u32 aeq_info; + u16 async_event_id; + u8 tcp_state; + u8 iwarp_state; + unsigned long flags; + + aeq_info = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_MISC_IDX]); + tcp_state = (aeq_info & NES_AEQE_TCP_STATE_MASK) >> NES_AEQE_TCP_STATE_SHIFT; + iwarp_state = (aeq_info & NES_AEQE_IWARP_STATE_MASK) >> NES_AEQE_IWARP_STATE_SHIFT; + async_event_id = (u16)aeq_info; + + spin_lock_irqsave(&nesqp->lock, flags); + nesqp->hw_iwarp_state = iwarp_state; + nesqp->hw_tcp_state = tcp_state; + nesqp->last_aeq = async_event_id; + spin_unlock_irqrestore(&nesqp->lock, flags); + + /* Send the fin only */ + nes_hw_modify_qp(nesdev, nesqp, NES_CQP_QP_IWARP_STATE_TERMINATE | + NES_CQP_QP_TERM_DONT_SEND_TERM_MSG, 0, 0); +} + +/* Cleanup after a terminate sent or received */ +static void nes_terminate_done(struct nes_qp *nesqp, int timeout_occurred) +{ + u32 next_iwarp_state = NES_CQP_QP_IWARP_STATE_ERROR; + unsigned long flags; + struct nes_vnic *nesvnic = to_nesvnic(nesqp->ibqp.device); + struct nes_device *nesdev = nesvnic->nesdev; + u8 first_time = 0; + + spin_lock_irqsave(&nesqp->lock, flags); + if (nesqp->hte_added) { + nesqp->hte_added = 0; + next_iwarp_state |= NES_CQP_QP_DEL_HTE; + } + + first_time = (nesqp->term_flags & NES_TERM_DONE) == 0; + nesqp->term_flags |= NES_TERM_DONE; + spin_unlock_irqrestore(&nesqp->lock, flags); + + /* Make sure we go through this only once */ + if (first_time) { + if (timeout_occurred == 0) + del_timer(&nesqp->terminate_timer); + else + next_iwarp_state |= NES_CQP_QP_RESET; + + nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0, 0); + nes_cm_disconn(nesqp); + } +} + +static void nes_terminate_received(struct nes_device *nesdev, + struct nes_qp *nesqp, struct nes_hw_aeqe *aeqe) +{ + u32 aeq_info; + u8 *pkt; + u32 *mpa; + u8 ddp_ctl; + u8 rdma_ctl; + u16 aeq_id = 0; + + aeq_info = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_MISC_IDX]); + if (aeq_info & NES_AEQE_Q2_DATA_WRITTEN) { + /* Terminate is not a performance path so the silicon */ + /* did not validate the frame - do it now */ + pkt = nesqp->hwqp.q2_vbase + BAD_FRAME_OFFSET; + mpa = (u32 *)locate_mpa(pkt, aeq_info); + ddp_ctl = (be32_to_cpu(mpa[0]) >> 8) & 0xff; + rdma_ctl = be32_to_cpu(mpa[0]) & 0xff; + if ((ddp_ctl & 0xc0) != 0x40) + aeq_id = NES_AEQE_AEID_DDP_LCE_LOCAL_CATASTROPHIC; + else if ((ddp_ctl & 0x03) != 1) + aeq_id = NES_AEQE_AEID_DDP_UBE_INVALID_DDP_VERSION; + else if (be32_to_cpu(mpa[2]) != 2) + aeq_id = NES_AEQE_AEID_DDP_UBE_INVALID_QN; + else if (be32_to_cpu(mpa[3]) != 1) + aeq_id = NES_AEQE_AEID_DDP_INVALID_MSN_GAP_IN_MSN; + else if (be32_to_cpu(mpa[4]) != 0) + aeq_id = NES_AEQE_AEID_DDP_UBE_INVALID_MO; + else if ((rdma_ctl & 0xc0) != 0x40) + aeq_id = NES_AEQE_AEID_RDMAP_ROE_INVALID_RDMAP_VERSION; + + if (aeq_id) { + /* Bad terminate recvd - send back a terminate */ + aeq_info = (aeq_info & 0xffff0000) | aeq_id; + aeqe->aeqe_words[NES_AEQE_MISC_IDX] = cpu_to_le32(aeq_info); + nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_FATAL); + return; + } + } + + nesqp->term_flags |= NES_TERM_RCVD; + nesqp->terminate_eventtype = IB_EVENT_QP_FATAL; + nes_terminate_start_timer(nesqp); + nes_terminate_send_fin(nesdev, nesqp, aeqe); +} + +/* Timeout routine in case terminate fails to complete */ +static void nes_terminate_timeout(unsigned long context) +{ + struct nes_qp *nesqp = (struct nes_qp *)(unsigned long)context; + + nes_terminate_done(nesqp, 1); +} + +/* Set a timer in case hw cannot complete the terminate sequence */ +static void nes_terminate_start_timer(struct nes_qp *nesqp) +{ + init_timer(&nesqp->terminate_timer); + nesqp->terminate_timer.function = nes_terminate_timeout; + nesqp->terminate_timer.expires = jiffies + HZ; + nesqp->terminate_timer.data = (unsigned long)nesqp; + add_timer(&nesqp->terminate_timer); +} + /** * nes_process_iwarp_aeqe */ @@ -2910,28 +3323,27 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, struct nes_hw_aeqe *aeqe) { u64 context; - u64 aeqe_context = 0; unsigned long flags; struct nes_qp *nesqp; + struct nes_hw_cq *hw_cq; + struct nes_cq *nescq; int resource_allocated; - /* struct iw_cm_id *cm_id; */ struct nes_adapter *nesadapter = nesdev->nesadapter; - struct ib_event ibevent; - /* struct iw_cm_event cm_event; */ u32 aeq_info; u32 next_iwarp_state = 0; u16 async_event_id; u8 tcp_state; u8 iwarp_state; + int must_disconn = 1; + int must_terminate = 0; + struct ib_event ibevent; nes_debug(NES_DBG_AEQ, "\n"); aeq_info = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_MISC_IDX]); - if ((NES_AEQE_INBOUND_RDMA&aeq_info) || (!(NES_AEQE_QP&aeq_info))) { + if ((NES_AEQE_INBOUND_RDMA & aeq_info) || (!(NES_AEQE_QP & aeq_info))) { context = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_CTXT_LOW_IDX]); context += ((u64)le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_CTXT_HIGH_IDX])) << 32; } else { - aeqe_context = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_CTXT_LOW_IDX]); - aeqe_context += ((u64)le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_CTXT_HIGH_IDX])) << 32; context = (unsigned long)nesadapter->qp_table[le32_to_cpu( aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX]) - NES_FIRST_QPN]; BUG_ON(!context); @@ -2948,7 +3360,11 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, switch (async_event_id) { case NES_AEQE_AEID_LLP_FIN_RECEIVED: - nesqp = *((struct nes_qp **)&context); + nesqp = (struct nes_qp *)(unsigned long)context; + + if (nesqp->term_flags) + return; /* Ignore it, wait for close complete */ + if (atomic_inc_return(&nesqp->close_timer_started) == 1) { nesqp->cm_id->add_ref(nesqp->cm_id); schedule_nes_timer(nesqp->cm_node, (struct sk_buff *)nesqp, @@ -2959,18 +3375,24 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, nesqp->hwqp.qp_id, atomic_read(&nesqp->refcount), async_event_id, nesqp->last_aeq, tcp_state); } + if ((tcp_state != NES_AEQE_TCP_STATE_CLOSE_WAIT) || (nesqp->ibqp_state != IB_QPS_RTS)) { /* FIN Received but tcp state or IB state moved on, should expect a close complete */ return; } + case NES_AEQE_AEID_LLP_CLOSE_COMPLETE: + nesqp = (struct nes_qp *)(unsigned long)context; + if (nesqp->term_flags) { + nes_terminate_done(nesqp, 0); + return; + } + case NES_AEQE_AEID_LLP_CONNECTION_RESET: - case NES_AEQE_AEID_TERMINATE_SENT: - case NES_AEQE_AEID_RDMAP_ROE_BAD_LLP_CLOSE: case NES_AEQE_AEID_RESET_SENT: - nesqp = *((struct nes_qp **)&context); + nesqp = (struct nes_qp *)(unsigned long)context; if (async_event_id == NES_AEQE_AEID_RESET_SENT) { tcp_state = NES_AEQE_TCP_STATE_CLOSED; } @@ -2982,12 +3404,7 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, if ((tcp_state == NES_AEQE_TCP_STATE_CLOSED) || (tcp_state == NES_AEQE_TCP_STATE_TIME_WAIT)) { nesqp->hte_added = 0; - spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "issuing hw modifyqp for QP%u to remove hte\n", - nesqp->hwqp.qp_id); - nes_hw_modify_qp(nesdev, nesqp, - NES_CQP_QP_IWARP_STATE_ERROR | NES_CQP_QP_DEL_HTE, 0); - spin_lock_irqsave(&nesqp->lock, flags); + next_iwarp_state = NES_CQP_QP_IWARP_STATE_ERROR | NES_CQP_QP_DEL_HTE; } if ((nesqp->ibqp_state == IB_QPS_RTS) && @@ -2999,151 +3416,106 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, nesqp->hw_iwarp_state = NES_AEQE_IWARP_STATE_CLOSING; break; case NES_AEQE_IWARP_STATE_TERMINATE: - next_iwarp_state = NES_CQP_QP_IWARP_STATE_TERMINATE; - nesqp->hw_iwarp_state = NES_AEQE_IWARP_STATE_TERMINATE; - if (async_event_id == NES_AEQE_AEID_RDMAP_ROE_BAD_LLP_CLOSE) { - next_iwarp_state |= 0x02000000; - nesqp->hw_tcp_state = NES_AEQE_TCP_STATE_CLOSED; - } + must_disconn = 0; /* terminate path takes care of disconn */ + if (nesqp->term_flags == 0) + must_terminate = 1; break; - default: - next_iwarp_state = 0; - } - spin_unlock_irqrestore(&nesqp->lock, flags); - if (next_iwarp_state) { - nes_debug(NES_DBG_AEQ, "issuing hw modifyqp for QP%u. next state = 0x%08X," - " also added another reference\n", - nesqp->hwqp.qp_id, next_iwarp_state); - nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0); } - nes_cm_disconn(nesqp); } else { if (async_event_id == NES_AEQE_AEID_LLP_FIN_RECEIVED) { /* FIN Received but ib state not RTS, close complete will be on its way */ - spin_unlock_irqrestore(&nesqp->lock, flags); - return; - } - spin_unlock_irqrestore(&nesqp->lock, flags); - if (async_event_id == NES_AEQE_AEID_RDMAP_ROE_BAD_LLP_CLOSE) { - next_iwarp_state = NES_CQP_QP_IWARP_STATE_TERMINATE | 0x02000000; - nesqp->hw_tcp_state = NES_AEQE_TCP_STATE_CLOSED; - nes_debug(NES_DBG_AEQ, "issuing hw modifyqp for QP%u. next state = 0x%08X," - " also added another reference\n", - nesqp->hwqp.qp_id, next_iwarp_state); - nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0); + must_disconn = 0; } - nes_cm_disconn(nesqp); } - break; - case NES_AEQE_AEID_LLP_TERMINATE_RECEIVED: - nesqp = *((struct nes_qp **)&context); - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_LLP_TERMINATE_RECEIVED" - " event on QP%u \n Q2 Data:\n", - nesqp->hwqp.qp_id); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_FATAL; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); - } - if ((tcp_state == NES_AEQE_TCP_STATE_CLOSE_WAIT) || - ((nesqp->ibqp_state == IB_QPS_RTS)&& - (async_event_id == NES_AEQE_AEID_LLP_CONNECTION_RESET))) { + + if (must_terminate) + nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_FATAL); + else if (must_disconn) { + if (next_iwarp_state) { + nes_debug(NES_DBG_AEQ, "issuing hw modifyqp for QP%u. next state = 0x%08X\n", + nesqp->hwqp.qp_id, next_iwarp_state); + nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0, 0); + } nes_cm_disconn(nesqp); - } else { - nesqp->in_disconnect = 0; - wake_up(&nesqp->kick_waitq); } break; - case NES_AEQE_AEID_LLP_TOO_MANY_RETRIES: - nesqp = *((struct nes_qp **)&context); - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = NES_AEQE_IWARP_STATE_ERROR; - nesqp->hw_tcp_state = NES_AEQE_TCP_STATE_CLOSED; - nesqp->last_aeq = async_event_id; - if (nesqp->cm_id) { - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_LLP_TOO_MANY_RETRIES" - " event on QP%u, remote IP = 0x%08X \n", - nesqp->hwqp.qp_id, - ntohl(nesqp->cm_id->remote_addr.sin_addr.s_addr)); - } else { - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_LLP_TOO_MANY_RETRIES" - " event on QP%u \n", - nesqp->hwqp.qp_id); - } - spin_unlock_irqrestore(&nesqp->lock, flags); - next_iwarp_state = NES_CQP_QP_IWARP_STATE_ERROR | NES_CQP_QP_RESET; - nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_FATAL; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); - } + + case NES_AEQE_AEID_TERMINATE_SENT: + nesqp = (struct nes_qp *)(unsigned long)context; + nes_terminate_send_fin(nesdev, nesqp, aeqe); break; - case NES_AEQE_AEID_AMP_BAD_STAG_INDEX: - if (NES_AEQE_INBOUND_RDMA&aeq_info) { - nesqp = nesadapter->qp_table[le32_to_cpu( - aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX])-NES_FIRST_QPN]; - } else { - /* TODO: get the actual WQE and mask off wqe index */ - context &= ~((u64)511); - nesqp = *((struct nes_qp **)&context); - } - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_AMP_BAD_STAG_INDEX event on QP%u\n", - nesqp->hwqp.qp_id); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_ACCESS_ERR; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); - } + + case NES_AEQE_AEID_LLP_TERMINATE_RECEIVED: + nesqp = (struct nes_qp *)(unsigned long)context; + nes_terminate_received(nesdev, nesqp, aeqe); break; + + case NES_AEQE_AEID_AMP_BAD_STAG_KEY: + case NES_AEQE_AEID_AMP_BAD_STAG_INDEX: case NES_AEQE_AEID_AMP_UNALLOCATED_STAG: - nesqp = *((struct nes_qp **)&context); - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_AMP_UNALLOCATED_STAG event on QP%u\n", - nesqp->hwqp.qp_id); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_ACCESS_ERR; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); - } - break; + case NES_AEQE_AEID_AMP_INVALID_STAG: + case NES_AEQE_AEID_AMP_RIGHTS_VIOLATION: + case NES_AEQE_AEID_AMP_INVALIDATE_NO_REMOTE_ACCESS_RIGHTS: case NES_AEQE_AEID_PRIV_OPERATION_DENIED: - nesqp = nesadapter->qp_table[le32_to_cpu(aeqe->aeqe_words - [NES_AEQE_COMP_QP_CQ_ID_IDX])-NES_FIRST_QPN]; - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_PRIV_OPERATION_DENIED event on QP%u," - " nesqp = %p, AE reported %p\n", - nesqp->hwqp.qp_id, nesqp, *((struct nes_qp **)&context)); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_ACCESS_ERR; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); + case NES_AEQE_AEID_DDP_UBE_DDP_MESSAGE_TOO_LONG_FOR_AVAILABLE_BUFFER: + case NES_AEQE_AEID_AMP_BOUNDS_VIOLATION: + case NES_AEQE_AEID_AMP_TO_WRAP: + nesqp = (struct nes_qp *)(unsigned long)context; + nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_ACCESS_ERR); + break; + + case NES_AEQE_AEID_LLP_SEGMENT_TOO_LARGE: + case NES_AEQE_AEID_LLP_SEGMENT_TOO_SMALL: + case NES_AEQE_AEID_DDP_UBE_INVALID_MO: + case NES_AEQE_AEID_DDP_UBE_INVALID_QN: + nesqp = (struct nes_qp *)(unsigned long)context; + if (iwarp_opcode(nesqp, aeq_info) > IWARP_OPCODE_TERM) { + aeq_info &= 0xffff0000; + aeq_info |= NES_AEQE_AEID_RDMAP_ROE_UNEXPECTED_OPCODE; + aeqe->aeqe_words[NES_AEQE_MISC_IDX] = cpu_to_le32(aeq_info); } + + case NES_AEQE_AEID_RDMAP_ROE_BAD_LLP_CLOSE: + case NES_AEQE_AEID_LLP_TOO_MANY_RETRIES: + case NES_AEQE_AEID_DDP_UBE_INVALID_MSN_NO_BUFFER_AVAILABLE: + case NES_AEQE_AEID_LLP_RECEIVED_MPA_CRC_ERROR: + case NES_AEQE_AEID_AMP_BAD_QP: + case NES_AEQE_AEID_LLP_RECEIVED_MARKER_AND_LENGTH_FIELDS_DONT_MATCH: + case NES_AEQE_AEID_DDP_LCE_LOCAL_CATASTROPHIC: + case NES_AEQE_AEID_DDP_NO_L_BIT: + case NES_AEQE_AEID_DDP_INVALID_MSN_GAP_IN_MSN: + case NES_AEQE_AEID_DDP_INVALID_MSN_RANGE_IS_NOT_VALID: + case NES_AEQE_AEID_DDP_UBE_INVALID_DDP_VERSION: + case NES_AEQE_AEID_RDMAP_ROE_INVALID_RDMAP_VERSION: + case NES_AEQE_AEID_RDMAP_ROE_UNEXPECTED_OPCODE: + case NES_AEQE_AEID_AMP_BAD_PD: + case NES_AEQE_AEID_AMP_FASTREG_SHARED: + case NES_AEQE_AEID_AMP_FASTREG_VALID_STAG: + case NES_AEQE_AEID_AMP_FASTREG_MW_STAG: + case NES_AEQE_AEID_AMP_FASTREG_INVALID_RIGHTS: + case NES_AEQE_AEID_AMP_FASTREG_PBL_TABLE_OVERFLOW: + case NES_AEQE_AEID_AMP_FASTREG_INVALID_LENGTH: + case NES_AEQE_AEID_AMP_INVALIDATE_SHARED: + case NES_AEQE_AEID_AMP_INVALIDATE_MR_WITH_BOUND_WINDOWS: + case NES_AEQE_AEID_AMP_MWBIND_VALID_STAG: + case NES_AEQE_AEID_AMP_MWBIND_OF_MR_STAG: + case NES_AEQE_AEID_AMP_MWBIND_TO_ZERO_BASED_STAG: + case NES_AEQE_AEID_AMP_MWBIND_TO_MW_STAG: + case NES_AEQE_AEID_AMP_MWBIND_INVALID_RIGHTS: + case NES_AEQE_AEID_AMP_MWBIND_INVALID_BOUNDS: + case NES_AEQE_AEID_AMP_MWBIND_TO_INVALID_PARENT: + case NES_AEQE_AEID_AMP_MWBIND_BIND_DISABLED: + case NES_AEQE_AEID_BAD_CLOSE: + case NES_AEQE_AEID_RDMA_READ_WHILE_ORD_ZERO: + case NES_AEQE_AEID_STAG_ZERO_INVALID: + case NES_AEQE_AEID_ROE_INVALID_RDMA_READ_REQUEST: + case NES_AEQE_AEID_ROE_INVALID_RDMA_WRITE_OR_READ_RESP: + nesqp = (struct nes_qp *)(unsigned long)context; + nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_FATAL); break; + case NES_AEQE_AEID_CQ_OPERATION_ERROR: context <<= 1; nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_CQ_OPERATION_ERROR event on CQ%u, %p\n", @@ -3153,83 +3525,19 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, if (resource_allocated) { printk(KERN_ERR PFX "%s: Processing an NES_AEQE_AEID_CQ_OPERATION_ERROR event on CQ%u\n", __func__, le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX])); + hw_cq = (struct nes_hw_cq *)(unsigned long)context; + if (hw_cq) { + nescq = container_of(hw_cq, struct nes_cq, hw_cq); + if (nescq->ibcq.event_handler) { + ibevent.device = nescq->ibcq.device; + ibevent.event = IB_EVENT_CQ_ERR; + ibevent.element.cq = &nescq->ibcq; + nescq->ibcq.event_handler(&ibevent, nescq->ibcq.cq_context); + } + } } break; - case NES_AEQE_AEID_DDP_UBE_DDP_MESSAGE_TOO_LONG_FOR_AVAILABLE_BUFFER: - nesqp = nesadapter->qp_table[le32_to_cpu( - aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX])-NES_FIRST_QPN]; - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_DDP_UBE_DDP_MESSAGE_TOO_LONG" - "_FOR_AVAILABLE_BUFFER event on QP%u\n", - nesqp->hwqp.qp_id); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_ACCESS_ERR; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); - } - /* tell cm to disconnect, cm will queue work to thread */ - nes_cm_disconn(nesqp); - break; - case NES_AEQE_AEID_DDP_UBE_INVALID_MSN_NO_BUFFER_AVAILABLE: - nesqp = *((struct nes_qp **)&context); - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_DDP_UBE_INVALID_MSN" - "_NO_BUFFER_AVAILABLE event on QP%u\n", - nesqp->hwqp.qp_id); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_FATAL; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); - } - /* tell cm to disconnect, cm will queue work to thread */ - nes_cm_disconn(nesqp); - break; - case NES_AEQE_AEID_LLP_RECEIVED_MPA_CRC_ERROR: - nesqp = *((struct nes_qp **)&context); - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - nes_debug(NES_DBG_AEQ, "Processing an NES_AEQE_AEID_LLP_RECEIVED_MPA_CRC_ERROR" - " event on QP%u \n Q2 Data:\n", - nesqp->hwqp.qp_id); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_FATAL; - nesqp->ibqp.event_handler(&ibevent, nesqp->ibqp.qp_context); - } - /* tell cm to disconnect, cm will queue work to thread */ - nes_cm_disconn(nesqp); - break; - /* TODO: additional AEs need to be here */ - case NES_AEQE_AEID_AMP_BOUNDS_VIOLATION: - nesqp = *((struct nes_qp **)&context); - spin_lock_irqsave(&nesqp->lock, flags); - nesqp->hw_iwarp_state = iwarp_state; - nesqp->hw_tcp_state = tcp_state; - nesqp->last_aeq = async_event_id; - spin_unlock_irqrestore(&nesqp->lock, flags); - if (nesqp->ibqp.event_handler) { - ibevent.device = nesqp->ibqp.device; - ibevent.element.qp = &nesqp->ibqp; - ibevent.event = IB_EVENT_QP_ACCESS_ERR; - nesqp->ibqp.event_handler(&ibevent, - nesqp->ibqp.qp_context); - } - nes_cm_disconn(nesqp); - break; + default: nes_debug(NES_DBG_AEQ, "Processing an iWARP related AE for QP, misc = 0x%04X\n", async_event_id); @@ -3238,7 +3546,6 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev, } - /** * nes_iwarp_ce_handler */ @@ -3373,6 +3680,8 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp, { struct nes_cqp_request *cqp_request; struct nes_hw_cqp_wqe *cqp_wqe; + u32 sq_code = (NES_IWARP_CQE_MAJOR_FLUSH << 16) | NES_IWARP_CQE_MINOR_FLUSH; + u32 rq_code = (NES_IWARP_CQE_MAJOR_FLUSH << 16) | NES_IWARP_CQE_MINOR_FLUSH; int ret; cqp_request = nes_get_cqp_request(nesdev); @@ -3389,6 +3698,24 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp, cqp_wqe = &cqp_request->cqp_wqe; nes_fill_init_cqp_wqe(cqp_wqe, nesdev); + /* If wqe in error was identified, set code to be put into cqe */ + if ((nesqp->term_sq_flush_code) && (which_wq & NES_CQP_FLUSH_SQ)) { + which_wq |= NES_CQP_FLUSH_MAJ_MIN; + sq_code = (CQE_MAJOR_DRV << 16) | nesqp->term_sq_flush_code; + nesqp->term_sq_flush_code = 0; + } + + if ((nesqp->term_rq_flush_code) && (which_wq & NES_CQP_FLUSH_RQ)) { + which_wq |= NES_CQP_FLUSH_MAJ_MIN; + rq_code = (CQE_MAJOR_DRV << 16) | nesqp->term_rq_flush_code; + nesqp->term_rq_flush_code = 0; + } + + if (which_wq & NES_CQP_FLUSH_MAJ_MIN) { + cqp_wqe->wqe_words[NES_CQP_QP_WQE_FLUSH_SQ_CODE] = cpu_to_le32(sq_code); + cqp_wqe->wqe_words[NES_CQP_QP_WQE_FLUSH_RQ_CODE] = cpu_to_le32(rq_code); + } + cqp_wqe->wqe_words[NES_CQP_WQE_OPCODE_IDX] = cpu_to_le32(NES_CQP_FLUSH_WQES | which_wq); cqp_wqe->wqe_words[NES_CQP_WQE_ID_IDX] = cpu_to_le32(nesqp->hwqp.qp_id); diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h index c3654c6383f..f28a41ba9fa 100644 --- a/drivers/infiniband/hw/nes/nes_hw.h +++ b/drivers/infiniband/hw/nes/nes_hw.h @@ -241,6 +241,7 @@ enum nes_cqp_stag_wqeword_idx { }; #define NES_CQP_OP_IWARP_STATE_SHIFT 28 +#define NES_CQP_OP_TERMLEN_SHIFT 28 enum nes_cqp_qp_bits { NES_CQP_QP_ARP_VALID = (1<<8), @@ -265,12 +266,16 @@ enum nes_cqp_qp_bits { NES_CQP_QP_IWARP_STATE_TERMINATE = (5<<NES_CQP_OP_IWARP_STATE_SHIFT), NES_CQP_QP_IWARP_STATE_ERROR = (6<<NES_CQP_OP_IWARP_STATE_SHIFT), NES_CQP_QP_IWARP_STATE_MASK = (7<<NES_CQP_OP_IWARP_STATE_SHIFT), + NES_CQP_QP_TERM_DONT_SEND_FIN = (1<<24), + NES_CQP_QP_TERM_DONT_SEND_TERM_MSG = (1<<25), NES_CQP_QP_RESET = (1<<31), }; enum nes_cqp_qp_wqe_word_idx { NES_CQP_QP_WQE_CONTEXT_LOW_IDX = 6, NES_CQP_QP_WQE_CONTEXT_HIGH_IDX = 7, + NES_CQP_QP_WQE_FLUSH_SQ_CODE = 8, + NES_CQP_QP_WQE_FLUSH_RQ_CODE = 9, NES_CQP_QP_WQE_NEW_MSS_IDX = 15, }; @@ -361,6 +366,7 @@ enum nes_cqp_arp_bits { enum nes_cqp_flush_bits { NES_CQP_FLUSH_SQ = (1<<30), NES_CQP_FLUSH_RQ = (1<<31), + NES_CQP_FLUSH_MAJ_MIN = (1<<28), }; enum nes_cqe_opcode_bits { @@ -633,11 +639,14 @@ enum nes_aeqe_bits { NES_AEQE_INBOUND_RDMA = (1<<19), NES_AEQE_IWARP_STATE_MASK = (7<<20), NES_AEQE_TCP_STATE_MASK = (0xf<<24), + NES_AEQE_Q2_DATA_WRITTEN = (0x3<<28), NES_AEQE_VALID = (1<<31), }; #define NES_AEQE_IWARP_STATE_SHIFT 20 #define NES_AEQE_TCP_STATE_SHIFT 24 +#define NES_AEQE_Q2_DATA_ETHERNET (1<<28) +#define NES_AEQE_Q2_DATA_MPA (1<<29) enum nes_aeqe_iwarp_state { NES_AEQE_IWARP_STATE_NON_EXISTANT = 0, @@ -751,6 +760,15 @@ enum nes_iwarp_sq_wqe_bits { NES_IWARP_SQ_OP_NOP = 12, }; +enum nes_iwarp_cqe_major_code { + NES_IWARP_CQE_MAJOR_FLUSH = 1, + NES_IWARP_CQE_MAJOR_DRV = 0x8000 +}; + +enum nes_iwarp_cqe_minor_code { + NES_IWARP_CQE_MINOR_FLUSH = 1 +}; + #define NES_EEPROM_READ_REQUEST (1<<16) #define NES_MAC_ADDR_VALID (1<<20) @@ -1119,6 +1137,7 @@ struct nes_adapter { u8 netdev_max; /* from host nic address count in EEPROM */ u8 port_count; u8 virtwq; + u8 send_term_ok; u8 et_use_adaptive_rx_coalesce; u8 adapter_fcn_count; u8 pft_mcast_map[NES_PFT_SIZE]; @@ -1217,6 +1236,90 @@ struct nes_ib_device { u32 num_pd; }; +enum nes_hdrct_flags { + DDP_LEN_FLAG = 0x80, + DDP_HDR_FLAG = 0x40, + RDMA_HDR_FLAG = 0x20 +}; + +enum nes_term_layers { + LAYER_RDMA = 0, + LAYER_DDP = 1, + LAYER_MPA = 2 +}; + +enum nes_term_error_types { + RDMAP_CATASTROPHIC = 0, + RDMAP_REMOTE_PROT = 1, + RDMAP_REMOTE_OP = 2, + DDP_CATASTROPHIC = 0, + DDP_TAGGED_BUFFER = 1, + DDP_UNTAGGED_BUFFER = 2, + DDP_LLP = 3 +}; + +enum nes_term_rdma_errors { + RDMAP_INV_STAG = 0x00, + RDMAP_INV_BOUNDS = 0x01, + RDMAP_ACCESS = 0x02, + RDMAP_UNASSOC_STAG = 0x03, + RDMAP_TO_WRAP = 0x04, + RDMAP_INV_RDMAP_VER = 0x05, + RDMAP_UNEXPECTED_OP = 0x06, + RDMAP_CATASTROPHIC_LOCAL = 0x07, + RDMAP_CATASTROPHIC_GLOBAL = 0x08, + RDMAP_CANT_INV_STAG = 0x09, + RDMAP_UNSPECIFIED = 0xff +}; + +enum nes_term_ddp_errors { + DDP_CATASTROPHIC_LOCAL = 0x00, + DDP_TAGGED_INV_STAG = 0x00, + DDP_TAGGED_BOUNDS = 0x01, + DDP_TAGGED_UNASSOC_STAG = 0x02, + DDP_TAGGED_TO_WRAP = 0x03, + DDP_TAGGED_INV_DDP_VER = 0x04, + DDP_UNTAGGED_INV_QN = 0x01, + DDP_UNTAGGED_INV_MSN_NO_BUF = 0x02, + DDP_UNTAGGED_INV_MSN_RANGE = 0x03, + DDP_UNTAGGED_INV_MO = 0x04, + DDP_UNTAGGED_INV_TOO_LONG = 0x05, + DDP_UNTAGGED_INV_DDP_VER = 0x06 +}; + +enum nes_term_mpa_errors { + MPA_CLOSED = 0x01, + MPA_CRC = 0x02, + MPA_MARKER = 0x03, + MPA_REQ_RSP = 0x04, +}; + +struct nes_terminate_hdr { + u8 layer_etype; + u8 error_code; + u8 hdrct; + u8 rsvd; +}; + +/* Used to determine how to fill in terminate error codes */ +#define IWARP_OPCODE_WRITE 0 +#define IWARP_OPCODE_READREQ 1 +#define IWARP_OPCODE_READRSP 2 +#define IWARP_OPCODE_SEND 3 +#define IWARP_OPCODE_SEND_INV 4 +#define IWARP_OPCODE_SEND_SE 5 +#define IWARP_OPCODE_SEND_SE_INV 6 +#define IWARP_OPCODE_TERM 7 + +/* These values are used only during terminate processing */ +#define TERM_DDP_LEN_TAGGED 14 +#define TERM_DDP_LEN_UNTAGGED 18 +#define TERM_RDMA_LEN 28 +#define RDMA_OPCODE_MASK 0x0f +#define RDMA_READ_REQ_OPCODE 1 +#define BAD_FRAME_OFFSET 64 +#define CQE_MAJOR_DRV 0x8000 + #define nes_vlan_rx vlan_hwaccel_receive_skb #define nes_netif_rx netif_receive_skb diff --git a/drivers/infiniband/hw/nes/nes_utils.c b/drivers/infiniband/hw/nes/nes_utils.c index a282031d15c..9687c397ce1 100644 --- a/drivers/infiniband/hw/nes/nes_utils.c +++ b/drivers/infiniband/hw/nes/nes_utils.c @@ -183,6 +183,9 @@ int nes_read_eeprom_values(struct nes_device *nesdev, struct nes_adapter *nesada } else if (((major_ver == 2) && (minor_ver > 21)) || ((major_ver > 2) && (major_ver != 255))) { nesadapter->virtwq = 1; } + if (((major_ver == 3) && (minor_ver >= 16)) || (major_ver > 3)) + nesadapter->send_term_ok = 1; + nesadapter->firmware_version = (((u32)(u8)(eeprom_data>>8)) << 16) + (u32)((u8)eeprom_data); @@ -548,7 +551,7 @@ struct nes_cqp_request *nes_get_cqp_request(struct nes_device *nesdev) spin_unlock_irqrestore(&nesdev->cqp.lock, flags); } if (cqp_request == NULL) { - cqp_request = kzalloc(sizeof(struct nes_cqp_request), GFP_KERNEL); + cqp_request = kzalloc(sizeof(struct nes_cqp_request), GFP_ATOMIC); if (cqp_request) { cqp_request->dynamic = 1; INIT_LIST_HEAD(&cqp_request->list); diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index 21e0fd336cf..a680c42d6e8 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -667,15 +667,32 @@ static int nes_query_device(struct ib_device *ibdev, struct ib_device_attr *prop */ static int nes_query_port(struct ib_device *ibdev, u8 port, struct ib_port_attr *props) { + struct nes_vnic *nesvnic = to_nesvnic(ibdev); + struct net_device *netdev = nesvnic->netdev; + memset(props, 0, sizeof(*props)); - props->max_mtu = IB_MTU_2048; - props->active_mtu = IB_MTU_2048; + props->max_mtu = IB_MTU_4096; + + if (netdev->mtu >= 4096) + props->active_mtu = IB_MTU_4096; + else if (netdev->mtu >= 2048) + props->active_mtu = IB_MTU_2048; + else if (netdev->mtu >= 1024) + props->active_mtu = IB_MTU_1024; + else if (netdev->mtu >= 512) + props->active_mtu = IB_MTU_512; + else + props->active_mtu = IB_MTU_256; + props->lid = 1; props->lmc = 0; props->sm_lid = 0; props->sm_sl = 0; - props->state = IB_PORT_ACTIVE; + if (nesvnic->linkup) + props->state = IB_PORT_ACTIVE; + else + props->state = IB_PORT_DOWN; props->phys_state = 0; props->port_cap_flags = IB_PORT_CM_SUP | IB_PORT_REINIT_SUP | IB_PORT_VENDOR_CLASS_SUP | IB_PORT_BOOT_MGMT_SUP; @@ -1506,12 +1523,45 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd, /** + * nes_clean_cq + */ +static void nes_clean_cq(struct nes_qp *nesqp, struct nes_cq *nescq) +{ + u32 cq_head; + u32 lo; + u32 hi; + u64 u64temp; + unsigned long flags = 0; + + spin_lock_irqsave(&nescq->lock, flags); + + cq_head = nescq->hw_cq.cq_head; + while (le32_to_cpu(nescq->hw_cq.cq_vbase[cq_head].cqe_words[NES_CQE_OPCODE_IDX]) & NES_CQE_VALID) { + rmb(); + lo = le32_to_cpu(nescq->hw_cq.cq_vbase[cq_head].cqe_words[NES_CQE_COMP_COMP_CTX_LOW_IDX]); + hi = le32_to_cpu(nescq->hw_cq.cq_vbase[cq_head].cqe_words[NES_CQE_COMP_COMP_CTX_HIGH_IDX]); + u64temp = (((u64)hi) << 32) | ((u64)lo); + u64temp &= ~(NES_SW_CONTEXT_ALIGN-1); + if (u64temp == (u64)(unsigned long)nesqp) { + /* Zero the context value so cqe will be ignored */ + nescq->hw_cq.cq_vbase[cq_head].cqe_words[NES_CQE_COMP_COMP_CTX_LOW_IDX] = 0; + nescq->hw_cq.cq_vbase[cq_head].cqe_words[NES_CQE_COMP_COMP_CTX_HIGH_IDX] = 0; + } + + if (++cq_head >= nescq->hw_cq.cq_size) + cq_head = 0; + } + + spin_unlock_irqrestore(&nescq->lock, flags); +} + + +/** * nes_destroy_qp */ static int nes_destroy_qp(struct ib_qp *ibqp) { struct nes_qp *nesqp = to_nesqp(ibqp); - /* struct nes_vnic *nesvnic = to_nesvnic(ibqp->device); */ struct nes_ucontext *nes_ucontext; struct ib_qp_attr attr; struct iw_cm_id *cm_id; @@ -1548,7 +1598,6 @@ static int nes_destroy_qp(struct ib_qp *ibqp) nes_debug(NES_DBG_QP, "OFA CM event_handler returned, ret=%d\n", ret); } - if (nesqp->user_mode) { if ((ibqp->uobject)&&(ibqp->uobject->context)) { nes_ucontext = to_nesucontext(ibqp->uobject->context); @@ -1560,6 +1609,13 @@ static int nes_destroy_qp(struct ib_qp *ibqp) } if (nesqp->pbl_pbase) kunmap(nesqp->page); + } else { + /* Clean any pending completions from the cq(s) */ + if (nesqp->nesscq) + nes_clean_cq(nesqp, nesqp->nesscq); + + if ((nesqp->nesrcq) && (nesqp->nesrcq != nesqp->nesscq)) + nes_clean_cq(nesqp, nesqp->nesrcq); } nes_rem_ref(&nesqp->ibqp); @@ -2884,7 +2940,7 @@ static int nes_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, * nes_hw_modify_qp */ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp, - u32 next_iwarp_state, u32 wait_completion) + u32 next_iwarp_state, u32 termlen, u32 wait_completion) { struct nes_hw_cqp_wqe *cqp_wqe; /* struct iw_cm_id *cm_id = nesqp->cm_id; */ @@ -2916,6 +2972,13 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp, set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_WQE_ID_IDX, nesqp->hwqp.qp_id); set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, (u64)nesqp->nesqp_context_pbase); + /* If sending a terminate message, fill in the length (in words) */ + if (((next_iwarp_state & NES_CQP_QP_IWARP_STATE_MASK) == NES_CQP_QP_IWARP_STATE_TERMINATE) && + !(next_iwarp_state & NES_CQP_QP_TERM_DONT_SEND_TERM_MSG)) { + termlen = ((termlen + 3) >> 2) << NES_CQP_OP_TERMLEN_SHIFT; + set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_NEW_MSS_IDX, termlen); + } + atomic_set(&cqp_request->refcount, 2); nes_post_cqp_request(nesdev, cqp_request); @@ -3086,6 +3149,9 @@ int nes_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, } nes_debug(NES_DBG_MOD_QP, "QP%u: new state = error\n", nesqp->hwqp.qp_id); + if (nesqp->term_flags) + del_timer(&nesqp->terminate_timer); + next_iwarp_state = NES_CQP_QP_IWARP_STATE_ERROR; /* next_iwarp_state = (NES_CQP_QP_IWARP_STATE_TERMINATE | 0x02000000); */ if (nesqp->hte_added) { @@ -3163,7 +3229,7 @@ int nes_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, if (issue_modify_qp) { nes_debug(NES_DBG_MOD_QP, "call nes_hw_modify_qp\n"); - ret = nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 1); + ret = nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0, 1); if (ret) nes_debug(NES_DBG_MOD_QP, "nes_hw_modify_qp (next_iwarp_state = 0x%08X)" " failed for QP%u.\n", @@ -3328,6 +3394,12 @@ static int nes_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr, head = nesqp->hwqp.sq_head; while (ib_wr) { + /* Check for QP error */ + if (nesqp->term_flags) { + err = -EINVAL; + break; + } + /* Check for SQ overflow */ if (((head + (2 * qsize) - nesqp->hwqp.sq_tail) % qsize) == (qsize - 1)) { err = -EINVAL; @@ -3484,6 +3556,12 @@ static int nes_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *ib_wr, head = nesqp->hwqp.rq_head; while (ib_wr) { + /* Check for QP error */ + if (nesqp->term_flags) { + err = -EINVAL; + break; + } + if (ib_wr->num_sge > nesdev->nesadapter->max_sge) { err = -EINVAL; break; @@ -3547,7 +3625,6 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) { u64 u64temp; u64 wrid; - /* u64 u64temp; */ unsigned long flags = 0; struct nes_vnic *nesvnic = to_nesvnic(ibcq->device); struct nes_device *nesdev = nesvnic->nesdev; @@ -3555,12 +3632,13 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) struct nes_qp *nesqp; struct nes_hw_cqe cqe; u32 head; - u32 wq_tail; + u32 wq_tail = 0; u32 cq_size; u32 cqe_count = 0; u32 wqe_index; u32 u32temp; - /* u32 counter; */ + u32 move_cq_head = 1; + u32 err_code; nes_debug(NES_DBG_CQ, "\n"); @@ -3570,29 +3648,40 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) cq_size = nescq->hw_cq.cq_size; while (cqe_count < num_entries) { - if (le32_to_cpu(nescq->hw_cq.cq_vbase[head].cqe_words[NES_CQE_OPCODE_IDX]) & - NES_CQE_VALID) { - /* - * Make sure we read CQ entry contents *after* - * we've checked the valid bit. - */ - rmb(); - - cqe = nescq->hw_cq.cq_vbase[head]; - nescq->hw_cq.cq_vbase[head].cqe_words[NES_CQE_OPCODE_IDX] = 0; - u32temp = le32_to_cpu(cqe.cqe_words[NES_CQE_COMP_COMP_CTX_LOW_IDX]); - wqe_index = u32temp & - (nesdev->nesadapter->max_qp_wr - 1); - u32temp &= ~(NES_SW_CONTEXT_ALIGN-1); - /* parse CQE, get completion context from WQE (either rq or sq */ - u64temp = (((u64)(le32_to_cpu(cqe.cqe_words[NES_CQE_COMP_COMP_CTX_HIGH_IDX])))<<32) | - ((u64)u32temp); - nesqp = *((struct nes_qp **)&u64temp); + if ((le32_to_cpu(nescq->hw_cq.cq_vbase[head].cqe_words[NES_CQE_OPCODE_IDX]) & + NES_CQE_VALID) == 0) + break; + + /* + * Make sure we read CQ entry contents *after* + * we've checked the valid bit. + */ + rmb(); + + cqe = nescq->hw_cq.cq_vbase[head]; + u32temp = le32_to_cpu(cqe.cqe_words[NES_CQE_COMP_COMP_CTX_LOW_IDX]); + wqe_index = u32temp & (nesdev->nesadapter->max_qp_wr - 1); + u32temp &= ~(NES_SW_CONTEXT_ALIGN-1); + /* parse CQE, get completion context from WQE (either rq or sq) */ + u64temp = (((u64)(le32_to_cpu(cqe.cqe_words[NES_CQE_COMP_COMP_CTX_HIGH_IDX])))<<32) | + ((u64)u32temp); + + if (u64temp) { + nesqp = (struct nes_qp *)(unsigned long)u64temp; memset(entry, 0, sizeof *entry); if (cqe.cqe_words[NES_CQE_ERROR_CODE_IDX] == 0) { entry->status = IB_WC_SUCCESS; } else { - entry->status = IB_WC_WR_FLUSH_ERR; + err_code = le32_to_cpu(cqe.cqe_words[NES_CQE_ERROR_CODE_IDX]); + if (NES_IWARP_CQE_MAJOR_DRV == (err_code >> 16)) { + entry->status = err_code & 0x0000ffff; + + /* The rest of the cqe's will be marked as flushed */ + nescq->hw_cq.cq_vbase[head].cqe_words[NES_CQE_ERROR_CODE_IDX] = + cpu_to_le32((NES_IWARP_CQE_MAJOR_FLUSH << 16) | + NES_IWARP_CQE_MINOR_FLUSH); + } else + entry->status = IB_WC_WR_FLUSH_ERR; } entry->qp = &nesqp->ibqp; @@ -3601,20 +3690,18 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) if (le32_to_cpu(cqe.cqe_words[NES_CQE_OPCODE_IDX]) & NES_CQE_SQ) { if (nesqp->skip_lsmm) { nesqp->skip_lsmm = 0; - wq_tail = nesqp->hwqp.sq_tail++; + nesqp->hwqp.sq_tail++; } /* Working on a SQ Completion*/ - wq_tail = wqe_index; - nesqp->hwqp.sq_tail = (wqe_index+1)&(nesqp->hwqp.sq_size - 1); - wrid = (((u64)(cpu_to_le32((u32)nesqp->hwqp.sq_vbase[wq_tail]. + wrid = (((u64)(cpu_to_le32((u32)nesqp->hwqp.sq_vbase[wqe_index]. wqe_words[NES_IWARP_SQ_WQE_COMP_SCRATCH_HIGH_IDX]))) << 32) | - ((u64)(cpu_to_le32((u32)nesqp->hwqp.sq_vbase[wq_tail]. + ((u64)(cpu_to_le32((u32)nesqp->hwqp.sq_vbase[wqe_index]. wqe_words[NES_IWARP_SQ_WQE_COMP_SCRATCH_LOW_IDX]))); - entry->byte_len = le32_to_cpu(nesqp->hwqp.sq_vbase[wq_tail]. + entry->byte_len = le32_to_cpu(nesqp->hwqp.sq_vbase[wqe_index]. wqe_words[NES_IWARP_SQ_WQE_TOTAL_PAYLOAD_IDX]); - switch (le32_to_cpu(nesqp->hwqp.sq_vbase[wq_tail]. + switch (le32_to_cpu(nesqp->hwqp.sq_vbase[wqe_index]. wqe_words[NES_IWARP_SQ_WQE_MISC_IDX]) & 0x3f) { case NES_IWARP_SQ_OP_RDMAW: nes_debug(NES_DBG_CQ, "Operation = RDMA WRITE.\n"); @@ -3623,7 +3710,7 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) case NES_IWARP_SQ_OP_RDMAR: nes_debug(NES_DBG_CQ, "Operation = RDMA READ.\n"); entry->opcode = IB_WC_RDMA_READ; - entry->byte_len = le32_to_cpu(nesqp->hwqp.sq_vbase[wq_tail]. + entry->byte_len = le32_to_cpu(nesqp->hwqp.sq_vbase[wqe_index]. wqe_words[NES_IWARP_SQ_WQE_RDMA_LENGTH_IDX]); break; case NES_IWARP_SQ_OP_SENDINV: @@ -3634,33 +3721,54 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) entry->opcode = IB_WC_SEND; break; } + + nesqp->hwqp.sq_tail = (wqe_index+1)&(nesqp->hwqp.sq_size - 1); + if ((entry->status != IB_WC_SUCCESS) && (nesqp->hwqp.sq_tail != nesqp->hwqp.sq_head)) { + move_cq_head = 0; + wq_tail = nesqp->hwqp.sq_tail; + } } else { /* Working on a RQ Completion*/ - wq_tail = wqe_index; - nesqp->hwqp.rq_tail = (wqe_index+1)&(nesqp->hwqp.rq_size - 1); entry->byte_len = le32_to_cpu(cqe.cqe_words[NES_CQE_PAYLOAD_LENGTH_IDX]); - wrid = ((u64)(le32_to_cpu(nesqp->hwqp.rq_vbase[wq_tail].wqe_words[NES_IWARP_RQ_WQE_COMP_SCRATCH_LOW_IDX]))) | - ((u64)(le32_to_cpu(nesqp->hwqp.rq_vbase[wq_tail].wqe_words[NES_IWARP_RQ_WQE_COMP_SCRATCH_HIGH_IDX]))<<32); + wrid = ((u64)(le32_to_cpu(nesqp->hwqp.rq_vbase[wqe_index].wqe_words[NES_IWARP_RQ_WQE_COMP_SCRATCH_LOW_IDX]))) | + ((u64)(le32_to_cpu(nesqp->hwqp.rq_vbase[wqe_index].wqe_words[NES_IWARP_RQ_WQE_COMP_SCRATCH_HIGH_IDX]))<<32); entry->opcode = IB_WC_RECV; + + nesqp->hwqp.rq_tail = (wqe_index+1)&(nesqp->hwqp.rq_size - 1); + if ((entry->status != IB_WC_SUCCESS) && (nesqp->hwqp.rq_tail != nesqp->hwqp.rq_head)) { + move_cq_head = 0; + wq_tail = nesqp->hwqp.rq_tail; + } } + entry->wr_id = wrid; + entry++; + cqe_count++; + } + if (move_cq_head) { + nescq->hw_cq.cq_vbase[head].cqe_words[NES_CQE_OPCODE_IDX] = 0; if (++head >= cq_size) head = 0; - cqe_count++; nescq->polled_completions++; + if ((nescq->polled_completions > (cq_size / 2)) || (nescq->polled_completions == 255)) { nes_debug(NES_DBG_CQ, "CQ%u Issuing CQE Allocate since more than half of cqes" - " are pending %u of %u.\n", - nescq->hw_cq.cq_number, nescq->polled_completions, cq_size); + " are pending %u of %u.\n", + nescq->hw_cq.cq_number, nescq->polled_completions, cq_size); nes_write32(nesdev->regs+NES_CQE_ALLOC, - nescq->hw_cq.cq_number | (nescq->polled_completions << 16)); + nescq->hw_cq.cq_number | (nescq->polled_completions << 16)); nescq->polled_completions = 0; } - entry++; - } else - break; + } else { + /* Update the wqe index and set status to flush */ + wqe_index = le32_to_cpu(cqe.cqe_words[NES_CQE_COMP_COMP_CTX_LOW_IDX]); + wqe_index = (wqe_index & (~(nesdev->nesadapter->max_qp_wr - 1))) | wq_tail; + nescq->hw_cq.cq_vbase[head].cqe_words[NES_CQE_COMP_COMP_CTX_LOW_IDX] = + cpu_to_le32(wqe_index); + move_cq_head = 1; /* ready for next pass */ + } } if (nescq->polled_completions) { diff --git a/drivers/infiniband/hw/nes/nes_verbs.h b/drivers/infiniband/hw/nes/nes_verbs.h index 41c07f29f7c..89822d75f82 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.h +++ b/drivers/infiniband/hw/nes/nes_verbs.h @@ -40,6 +40,10 @@ struct nes_device; #define NES_MAX_USER_DB_REGIONS 4096 #define NES_MAX_USER_WQ_REGIONS 4096 +#define NES_TERM_SENT 0x01 +#define NES_TERM_RCVD 0x02 +#define NES_TERM_DONE 0x04 + struct nes_ucontext { struct ib_ucontext ibucontext; struct nes_device *nesdev; @@ -119,6 +123,11 @@ struct nes_wq { spinlock_t lock; }; +struct disconn_work { + struct work_struct work; + struct nes_qp *nesqp; +}; + struct iw_cm_id; struct ietf_mpa_frame; @@ -127,7 +136,6 @@ struct nes_qp { void *allocated_buffer; struct iw_cm_id *cm_id; struct workqueue_struct *wq; - struct work_struct disconn_work; struct nes_cq *nesscq; struct nes_cq *nesrcq; struct nes_pd *nespd; @@ -155,9 +163,13 @@ struct nes_qp { void *pbl_vbase; dma_addr_t pbl_pbase; struct page *page; + struct timer_list terminate_timer; + enum ib_event_type terminate_eventtype; wait_queue_head_t kick_waitq; u16 in_disconnect; u16 private_data_len; + u16 term_sq_flush_code; + u16 term_rq_flush_code; u8 active_conn; u8 skip_lsmm; u8 user_mode; @@ -165,7 +177,7 @@ struct nes_qp { u8 hw_iwarp_state; u8 flush_issued; u8 hw_tcp_state; - u8 disconn_pending; + u8 term_flags; u8 destroyed; }; #endif /* NES_VERBS_H */ diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 181b1f32325..8f4b4fca2a1 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -31,7 +31,6 @@ */ #include <rdma/ib_cm.h> -#include <rdma/ib_cache.h> #include <net/dst.h> #include <net/icmp.h> #include <linux/icmpv6.h> diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index e7e5adf84e8..e35f4a0ea9d 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -36,7 +36,6 @@ #include <linux/delay.h> #include <linux/dma-mapping.h> -#include <rdma/ib_cache.h> #include <linux/ip.h> #include <linux/tcp.h> diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index e319d91f60a..2bf5116deec 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -604,8 +604,11 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev) skb_queue_len(&neigh->queue)); goto err_drop; } - } else + } else { + spin_unlock_irqrestore(&priv->lock, flags); ipoib_send(dev, skb, path->ah, IPOIB_QPN(skb_dst(skb)->neighbour->ha)); + return; + } } else { neigh->ah = NULL; @@ -688,7 +691,9 @@ static void unicast_arp_send(struct sk_buff *skb, struct net_device *dev, ipoib_dbg(priv, "Send unicast ARP to %04x\n", be16_to_cpu(path->pathrec.dlid)); + spin_unlock_irqrestore(&priv->lock, flags); ipoib_send(dev, skb, path->ah, IPOIB_QPN(phdr->hwaddr)); + return; } else if ((path->query || !path_rec_start(dev, path)) && skb_queue_len(&path->queue) < IPOIB_MAX_PATH_REC_QUEUE) { /* put pseudoheader back on for next time */ diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index a0e97532e71..25874fc680c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -720,7 +720,9 @@ out: } } + spin_unlock_irqrestore(&priv->lock, flags); ipoib_send(dev, skb, mcast->ah, IB_MULTICAST_QPN); + return; } unlock: @@ -758,6 +760,20 @@ void ipoib_mcast_dev_flush(struct net_device *dev) } } +static int ipoib_mcast_addr_is_valid(const u8 *addr, unsigned int addrlen, + const u8 *broadcast) +{ + if (addrlen != INFINIBAND_ALEN) + return 0; + /* reserved QPN, prefix, scope */ + if (memcmp(addr, broadcast, 6)) + return 0; + /* signature lower, pkey */ + if (memcmp(addr + 7, broadcast + 7, 3)) + return 0; + return 1; +} + void ipoib_mcast_restart_task(struct work_struct *work) { struct ipoib_dev_priv *priv = @@ -791,6 +807,11 @@ void ipoib_mcast_restart_task(struct work_struct *work) for (mclist = dev->mc_list; mclist; mclist = mclist->next) { union ib_gid mgid; + if (!ipoib_mcast_addr_is_valid(mclist->dmi_addr, + mclist->dmi_addrlen, + dev->broadcast)) + continue; + memcpy(mgid.raw, mclist->dmi_addr + 4, sizeof mgid); mcast = __ipoib_mcast_find(dev, &mgid); |