diff options
Diffstat (limited to 'drivers/infiniband/core/mad.c')
| -rw-r--r-- | drivers/infiniband/core/mad.c | 450 |
1 files changed, 275 insertions, 175 deletions
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index 493f4c65c7a..ab31f136d04 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -1,7 +1,8 @@ /* - * Copyright (c) 2004, 2005 Voltaire, Inc. All rights reserved. + * 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 @@ -31,9 +32,10 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. * - * $Id: mad.c 5596 2006-03-03 01:00:07Z sean.hefty $ */ #include <linux/dma-mapping.h> +#include <linux/slab.h> +#include <linux/module.h> #include <rdma/ib_cache.h> #include "mad_priv.h" @@ -46,14 +48,21 @@ MODULE_DESCRIPTION("kernel IB MAD API"); MODULE_AUTHOR("Hal Rosenstock"); MODULE_AUTHOR("Sean Hefty"); -static kmem_cache_t *ib_mad_cache; +static int mad_sendq_size = IB_MAD_QP_SEND_SIZE; +static 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, @@ -65,8 +74,8 @@ static struct ib_mad_agent_private *find_mad_agent( static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info, struct ib_mad_private *mad); static void cancel_mads(struct ib_mad_agent_private *mad_agent_priv); -static void timeout_sends(void *data); -static void local_completions(void *data); +static void timeout_sends(struct work_struct *work); +static void local_completions(struct work_struct *work); static int add_nonoui_reg_req(struct ib_mad_reg_req *mad_reg_req, struct ib_mad_agent_private *agent_priv, u8 mgmt_class); @@ -268,6 +277,13 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, goto error1; } + /* Verify the QP requested is supported. For example, Ethernet devices + * will not have QP0 */ + if (!port_priv->qp_info[qpn].qp) { + ret = ERR_PTR(-EPROTONOSUPPORT); + goto error1; + } + /* Allocate structures */ mad_agent_priv = kzalloc(sizeof *mad_agent_priv, GFP_KERNEL); if (!mad_agent_priv) { @@ -283,13 +299,11 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, } if (mad_reg_req) { - reg_req = kmalloc(sizeof *reg_req, GFP_KERNEL); + reg_req = kmemdup(mad_reg_req, sizeof *reg_req, GFP_KERNEL); if (!reg_req) { ret = ERR_PTR(-ENOMEM); goto error3; } - /* Make a copy of the MAD registration request */ - memcpy(reg_req, mad_reg_req, sizeof *reg_req); } /* Now, fill in the various structures */ @@ -302,6 +316,16 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, mad_agent_priv->agent.context = context; mad_agent_priv->agent.qp = port_priv->qp_info[qpn].qp; mad_agent_priv->agent.port_num = port_num; + spin_lock_init(&mad_agent_priv->lock); + INIT_LIST_HEAD(&mad_agent_priv->send_list); + INIT_LIST_HEAD(&mad_agent_priv->wait_list); + INIT_LIST_HEAD(&mad_agent_priv->done_list); + INIT_LIST_HEAD(&mad_agent_priv->rmpp_list); + INIT_DELAYED_WORK(&mad_agent_priv->timed_work, timeout_sends); + INIT_LIST_HEAD(&mad_agent_priv->local_list); + INIT_WORK(&mad_agent_priv->local_work, local_completions); + atomic_set(&mad_agent_priv->refcount, 1); + init_completion(&mad_agent_priv->comp); spin_lock_irqsave(&port_priv->reg_lock, flags); mad_agent_priv->agent.hi_tid = ++ib_mad_client_id; @@ -351,18 +375,6 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, list_add_tail(&mad_agent_priv->agent_list, &port_priv->agent_list); spin_unlock_irqrestore(&port_priv->reg_lock, flags); - spin_lock_init(&mad_agent_priv->lock); - INIT_LIST_HEAD(&mad_agent_priv->send_list); - INIT_LIST_HEAD(&mad_agent_priv->wait_list); - INIT_LIST_HEAD(&mad_agent_priv->done_list); - INIT_LIST_HEAD(&mad_agent_priv->rmpp_list); - INIT_WORK(&mad_agent_priv->timed_work, timeout_sends, mad_agent_priv); - INIT_LIST_HEAD(&mad_agent_priv->local_list); - INIT_WORK(&mad_agent_priv->local_work, local_completions, - mad_agent_priv); - atomic_set(&mad_agent_priv->refcount, 1); - init_completion(&mad_agent_priv->comp); - return &mad_agent_priv->agent; error4: @@ -408,19 +420,15 @@ static int register_snoop_agent(struct ib_mad_qp_info *qp_info, if (i == qp_info->snoop_table_size) { /* Grow table. */ - new_snoop_table = kmalloc(sizeof mad_snoop_priv * - qp_info->snoop_table_size + 1, - GFP_ATOMIC); + new_snoop_table = krealloc(qp_info->snoop_table, + sizeof mad_snoop_priv * + (qp_info->snoop_table_size + 1), + GFP_ATOMIC); if (!new_snoop_table) { i = -ENOMEM; goto out; } - if (qp_info->snoop_table) { - memcpy(new_snoop_table, qp_info->snoop_table, - sizeof mad_snoop_priv * - qp_info->snoop_table_size); - kfree(qp_info->snoop_table); - } + qp_info->snoop_table = new_snoop_table; qp_info->snoop_table_size++; } @@ -643,7 +651,8 @@ static void snoop_recv(struct ib_mad_qp_info *qp_info, spin_unlock_irqrestore(&qp_info->snoop_lock, flags); } -static void build_smp_wc(u64 wr_id, u16 slid, u16 pkey_index, u8 port_num, +static void build_smp_wc(struct ib_qp *qp, + u64 wr_id, u16 slid, u16 pkey_index, u8 port_num, struct ib_wc *wc) { memset(wc, 0, sizeof *wc); @@ -653,7 +662,7 @@ static void build_smp_wc(u64 wr_id, u16 slid, u16 pkey_index, u8 port_num, wc->pkey_index = pkey_index; wc->byte_len = sizeof(struct ib_mad) + sizeof(struct ib_grh); wc->src_qp = IB_QP0; - wc->qp_num = IB_QP0; + wc->qp = qp; wc->slid = slid; wc->sl = 0; wc->dlid_path_bits = 0; @@ -668,7 +677,7 @@ static void build_smp_wc(u64 wr_id, u16 slid, u16 pkey_index, u8 port_num, static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, struct ib_mad_send_wr_private *mad_send_wr) { - int ret; + int ret = 0; struct ib_smp *smp = mad_send_wr->send_buf.mad; unsigned long flags; struct ib_mad_local_private *local; @@ -676,10 +685,16 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, struct ib_mad_port_private *port_priv; struct ib_mad_agent_private *recv_mad_agent = NULL; struct ib_device *device = mad_agent_priv->agent.device; - u8 port_num = mad_agent_priv->agent.port_num; + u8 port_num; struct ib_wc mad_wc; struct ib_send_wr *send_wr = &mad_send_wr->send_wr; + if (device->node_type == RDMA_NODE_IB_SWITCH && + smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) + port_num = send_wr->wr.ud.port_num; + else + port_num = mad_agent_priv->agent.port_num; + /* * Directed route handling starts if the initial LID routed part of * a request or the ending LID routed part of a response is empty. @@ -688,14 +703,16 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, */ if ((ib_get_smp_direction(smp) ? smp->dr_dlid : smp->dr_slid) == IB_LID_PERMISSIVE && - !smi_handle_dr_smp_send(smp, device->node_type, port_num)) { + smi_handle_dr_smp_send(smp, device->node_type, port_num) == + IB_SMI_DISCARD) { ret = -EINVAL; printk(KERN_ERR PFX "Invalid directed route\n"); goto out; } + /* Check to post send on QP or process locally */ - ret = smi_check_local_smp(smp, device); - if (!ret) + if (smi_check_local_smp(smp, device) == IB_SMI_DISCARD && + smi_check_local_returning_smp(smp, device) == IB_SMI_DISCARD) goto out; local = kmalloc(sizeof *local, GFP_ATOMIC); @@ -714,7 +731,8 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, goto out; } - build_smp_wc(send_wr->wr_id, be16_to_cpu(smp->dr_slid), + build_smp_wc(mad_agent_priv->agent.qp, + send_wr->wr_id, be16_to_cpu(smp->dr_slid), send_wr->wr.ud.pkey_index, send_wr->wr.ud.port_num, &mad_wc); @@ -745,16 +763,17 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, port_priv = ib_get_mad_port(mad_agent_priv->agent.device, mad_agent_priv->agent.port_num); if (port_priv) { - mad_priv->mad.mad.mad_hdr.tid = - ((struct ib_mad *)smp)->mad_hdr.tid; + memcpy(&mad_priv->mad.mad, smp, sizeof(struct ib_mad)); recv_mad_agent = find_mad_agent(port_priv, &mad_priv->mad.mad); } if (!port_priv || !recv_mad_agent) { + /* + * No receiving agent so drop packet and + * generate send completion. + */ kmem_cache_free(ib_mad_cache, mad_priv); - kfree(local); - ret = 0; - goto out; + break; } local->mad_priv = mad_priv; local->recv_mad_agent = recv_mad_agent; @@ -999,17 +1018,26 @@ int ib_send_mad(struct ib_mad_send_wr_private *mad_send_wr) mad_agent = mad_send_wr->send_buf.mad_agent; sge = mad_send_wr->sg_list; - sge[0].addr = dma_map_single(mad_agent->device->dma_device, - mad_send_wr->send_buf.mad, - sge[0].length, - DMA_TO_DEVICE); - pci_unmap_addr_set(mad_send_wr, header_mapping, sge[0].addr); - - sge[1].addr = dma_map_single(mad_agent->device->dma_device, - ib_get_payload(mad_send_wr), - sge[1].length, - DMA_TO_DEVICE); - pci_unmap_addr_set(mad_send_wr, payload_mapping, sge[1].addr); + sge[0].addr = ib_dma_map_single(mad_agent->device, + mad_send_wr->send_buf.mad, + sge[0].length, + DMA_TO_DEVICE); + if (unlikely(ib_dma_mapping_error(mad_agent->device, sge[0].addr))) + return -ENOMEM; + + mad_send_wr->header_mapping = sge[0].addr; + + sge[1].addr = ib_dma_map_single(mad_agent->device, + ib_get_payload(mad_send_wr), + sge[1].length, + DMA_TO_DEVICE); + if (unlikely(ib_dma_mapping_error(mad_agent->device, sge[1].addr))) { + ib_dma_unmap_single(mad_agent->device, + mad_send_wr->header_mapping, + sge[0].length, DMA_TO_DEVICE); + return -ENOMEM; + } + mad_send_wr->payload_mapping = sge[1].addr; spin_lock_irqsave(&qp_info->send_queue.lock, flags); if (qp_info->send_queue.count < qp_info->send_queue.max_active) { @@ -1027,12 +1055,12 @@ int ib_send_mad(struct ib_mad_send_wr_private *mad_send_wr) } spin_unlock_irqrestore(&qp_info->send_queue.lock, flags); if (ret) { - dma_unmap_single(mad_agent->device->dma_device, - pci_unmap_addr(mad_send_wr, header_mapping), - sge[0].length, DMA_TO_DEVICE); - dma_unmap_single(mad_agent->device->dma_device, - pci_unmap_addr(mad_send_wr, payload_mapping), - sge[1].length, DMA_TO_DEVICE); + ib_dma_unmap_single(mad_agent->device, + mad_send_wr->header_mapping, + sge[0].length, DMA_TO_DEVICE); + ib_dma_unmap_single(mad_agent->device, + mad_send_wr->payload_mapping, + sge[1].length, DMA_TO_DEVICE); } return ret; } @@ -1093,7 +1121,9 @@ int ib_post_send_mad(struct ib_mad_send_buf *send_buf, mad_send_wr->tid = ((struct ib_mad_hdr *) send_buf->mad)->tid; /* Timeout will be updated after send completes */ mad_send_wr->timeout = msecs_to_jiffies(send_buf->timeout_ms); - mad_send_wr->retries = send_buf->retries; + mad_send_wr->max_retries = send_buf->retries; + mad_send_wr->retries_left = send_buf->retries; + send_buf->retries = 0; /* Reference for work request to QP + response */ mad_send_wr->refcount = 1 + (mad_send_wr->timeout > 0); mad_send_wr->status = IB_WC_SUCCESS; @@ -1179,10 +1209,7 @@ static int method_in_use(struct ib_mad_mgmt_method_table **method, { int i; - for (i = find_first_bit(mad_reg_req->method_mask, IB_MGMT_MAX_METHODS); - i < IB_MGMT_MAX_METHODS; - i = find_next_bit(mad_reg_req->method_mask, IB_MGMT_MAX_METHODS, - 1+i)) { + for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) { if ((*method)->agent[i]) { printk(KERN_ERR PFX "Method %d already in use\n", i); return -EINVAL; @@ -1316,13 +1343,9 @@ static int add_nonoui_reg_req(struct ib_mad_reg_req *mad_reg_req, goto error3; /* Finally, add in methods being registered */ - for (i = find_first_bit(mad_reg_req->method_mask, - IB_MGMT_MAX_METHODS); - i < IB_MGMT_MAX_METHODS; - i = find_next_bit(mad_reg_req->method_mask, IB_MGMT_MAX_METHODS, - 1+i)) { + for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) (*method)->agent[i] = agent_priv; - } + return 0; error3: @@ -1415,13 +1438,9 @@ check_in_use: goto error4; /* Finally, add in methods being registered */ - for (i = find_first_bit(mad_reg_req->method_mask, - IB_MGMT_MAX_METHODS); - i < IB_MGMT_MAX_METHODS; - i = find_next_bit(mad_reg_req->method_mask, IB_MGMT_MAX_METHODS, - 1+i)) { + for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) (*method)->agent[i] = agent_priv; - } + return 0; error4: @@ -1587,6 +1606,9 @@ find_mad_agent(struct ib_mad_port_private *port_priv, mad->mad_hdr.class_version].class; if (!class) goto out; + if (convert_mgmt_class(mad->mad_hdr.mgmt_class) >= + IB_MGMT_MAX_METHODS) + goto out; method = class->method_table[convert_mgmt_class( mad->mad_hdr.mgmt_class)]; if (method) @@ -1686,9 +1708,8 @@ static inline int rcv_has_same_gid(struct ib_mad_agent_private *mad_agent_priv, u8 port_num = mad_agent_priv->agent.port_num; u8 lmc; - send_resp = ((struct ib_mad *)(wr->send_buf.mad))-> - mad_hdr.method & IB_MGMT_METHOD_RESP; - rcv_resp = rwc->recv_buf.mad->mad_hdr.method & IB_MGMT_METHOD_RESP; + send_resp = ib_response_mad((struct ib_mad *)wr->send_buf.mad); + rcv_resp = ib_response_mad(rwc->recv_buf.mad); if (send_resp == rcv_resp) /* both requests, or both responses. GIDs different */ @@ -1750,7 +1771,7 @@ ib_find_send_mad(struct ib_mad_agent_private *mad_agent_priv, */ (is_direct(wc->recv_buf.mad->mad_hdr.mgmt_class) || rcv_has_same_gid(mad_agent_priv, wr, wc))) - return wr; + return (wr->status == IB_WC_SUCCESS) ? wr : NULL; } /* @@ -1830,19 +1851,36 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv, } } +static bool generate_unmatched_resp(struct ib_mad_private *recv, + struct ib_mad_private *response) +{ + if (recv->mad.mad.mad_hdr.method == IB_MGMT_METHOD_GET || + recv->mad.mad.mad_hdr.method == IB_MGMT_METHOD_SET) { + memcpy(response, recv, sizeof *response); + response->header.recv_wc.wc = &response->header.wc; + response->header.recv_wc.recv_buf.mad = &response->mad.mad; + response->header.recv_wc.recv_buf.grh = &response->grh; + response->mad.mad.mad_hdr.method = IB_MGMT_METHOD_GET_RESP; + response->mad.mad.mad_hdr.status = + cpu_to_be16(IB_MGMT_MAD_STATUS_UNSUPPORTED_METHOD_ATTRIB); + if (recv->mad.mad.mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) + response->mad.mad.mad_hdr.status |= IB_SMP_DIRECTION; + + return true; + } else { + return false; + } +} static void ib_mad_recv_done_handler(struct ib_mad_port_private *port_priv, struct ib_wc *wc) { struct ib_mad_qp_info *qp_info; struct ib_mad_private_header *mad_priv_hdr; - struct ib_mad_private *recv, *response; + struct ib_mad_private *recv, *response = NULL; struct ib_mad_list_head *mad_list; struct ib_mad_agent_private *mad_agent; - - response = kmem_cache_alloc(ib_mad_cache, GFP_KERNEL); - if (!response) - printk(KERN_ERR PFX "ib_mad_recv_done_handler no memory " - "for response buffer\n"); + int port_num; + int ret = IB_MAD_RESULT_SUCCESS; mad_list = (struct ib_mad_list_head *)(unsigned long)wc->wr_id; qp_info = mad_list->mad_queue->qp_info; @@ -1851,11 +1889,11 @@ static void ib_mad_recv_done_handler(struct ib_mad_port_private *port_priv, mad_priv_hdr = container_of(mad_list, struct ib_mad_private_header, mad_list); recv = container_of(mad_priv_hdr, struct ib_mad_private, header); - dma_unmap_single(port_priv->device->dma_device, - pci_unmap_addr(&recv->header, mapping), - sizeof(struct ib_mad_private) - - sizeof(struct ib_mad_private_header), - DMA_FROM_DEVICE); + ib_dma_unmap_single(port_priv->device, + recv->header.mapping, + sizeof(struct ib_mad_private) - + sizeof(struct ib_mad_private_header), + DMA_FROM_DEVICE); /* Setup MAD receive work completion from "normal" work completion */ recv->header.wc = *wc; @@ -1871,37 +1909,61 @@ static void ib_mad_recv_done_handler(struct ib_mad_port_private *port_priv, if (!validate_mad(&recv->mad.mad, qp_info->qp->qp_num)) goto out; + response = kmem_cache_alloc(ib_mad_cache, GFP_KERNEL); + if (!response) { + printk(KERN_ERR PFX "ib_mad_recv_done_handler no memory " + "for response buffer\n"); + goto out; + } + + if (port_priv->device->node_type == RDMA_NODE_IB_SWITCH) + port_num = wc->port_num; + else + port_num = port_priv->port_num; + if (recv->mad.mad.mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) { - if (!smi_handle_dr_smp_recv(&recv->mad.smp, - port_priv->device->node_type, - port_priv->port_num, - port_priv->device->phys_port_cnt)) + enum smi_forward_action retsmi; + + if (smi_handle_dr_smp_recv(&recv->mad.smp, + port_priv->device->node_type, + port_num, + port_priv->device->phys_port_cnt) == + IB_SMI_DISCARD) goto out; - if (!smi_check_forward_dr_smp(&recv->mad.smp)) + + retsmi = smi_check_forward_dr_smp(&recv->mad.smp); + if (retsmi == IB_SMI_LOCAL) goto local; - if (!smi_handle_dr_smp_send(&recv->mad.smp, - port_priv->device->node_type, - port_priv->port_num)) - goto out; - if (!smi_check_local_smp(&recv->mad.smp, port_priv->device)) + + if (retsmi == IB_SMI_SEND) { /* don't forward */ + if (smi_handle_dr_smp_send(&recv->mad.smp, + port_priv->device->node_type, + port_num) == IB_SMI_DISCARD) + goto out; + + if (smi_check_local_smp(&recv->mad.smp, port_priv->device) == IB_SMI_DISCARD) + goto out; + } else if (port_priv->device->node_type == RDMA_NODE_IB_SWITCH) { + /* forward case for switches */ + memcpy(response, recv, sizeof(*response)); + response->header.recv_wc.wc = &response->header.wc; + response->header.recv_wc.recv_buf.mad = &response->mad.mad; + response->header.recv_wc.recv_buf.grh = &response->grh; + + agent_send_response(&response->mad.mad, + &response->grh, wc, + port_priv->device, + smi_get_fwd_port(&recv->mad.smp), + qp_info->qp->qp_num); + goto out; + } } local: /* Give driver "right of first refusal" on incoming MAD */ if (port_priv->device->process_mad) { - int ret; - - if (!response) { - printk(KERN_ERR PFX "No memory for response MAD\n"); - /* - * Is it better to assume that - * it wouldn't be processed ? - */ - goto out; - } - ret = port_priv->device->process_mad(port_priv->device, 0, port_priv->port_num, wc, &recv->grh, @@ -1914,7 +1976,7 @@ local: agent_send_response(&response->mad.mad, &recv->grh, wc, port_priv->device, - port_priv->port_num, + port_num, qp_info->qp->qp_num); goto out; } @@ -1929,6 +1991,10 @@ local: * or via recv_handler in ib_mad_complete_recv() */ recv = NULL; + } else if ((ret & IB_MAD_RESULT_SUCCESS) && + generate_unmatched_resp(recv, response)) { + agent_send_response(&response->mad.mad, &recv->grh, wc, + port_priv->device, port_num, qp_info->qp->qp_num); } out: @@ -1956,13 +2022,11 @@ 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); delay = mad_send_wr->timeout - jiffies; if ((long)delay <= 0) delay = 1; - queue_delayed_work(mad_agent_priv->qp_info-> - port_priv->wq, - &mad_agent_priv->timed_work, delay); + mod_delayed_work(mad_agent_priv->qp_info->port_priv->wq, + &mad_agent_priv->timed_work, delay); } } } @@ -1995,11 +2059,9 @@ static void wait_for_response(struct ib_mad_send_wr_private *mad_send_wr) list_add(&mad_send_wr->agent_list, list_item); /* 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); - queue_delayed_work(mad_agent_priv->qp_info->port_priv->wq, - &mad_agent_priv->timed_work, delay); - } + if (mad_agent_priv->wait_list.next == &mad_send_wr->agent_list) + mod_delayed_work(mad_agent_priv->qp_info->port_priv->wq, + &mad_agent_priv->timed_work, delay); } void ib_reset_mad_timeout(struct ib_mad_send_wr_private *mad_send_wr, @@ -2081,12 +2143,12 @@ static void ib_mad_send_done_handler(struct ib_mad_port_private *port_priv, qp_info = send_queue->qp_info; retry: - dma_unmap_single(mad_send_wr->send_buf.mad_agent->device->dma_device, - pci_unmap_addr(mad_send_wr, header_mapping), - mad_send_wr->sg_list[0].length, DMA_TO_DEVICE); - dma_unmap_single(mad_send_wr->send_buf.mad_agent->device->dma_device, - pci_unmap_addr(mad_send_wr, payload_mapping), - mad_send_wr->sg_list[1].length, DMA_TO_DEVICE); + ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device, + mad_send_wr->header_mapping, + mad_send_wr->sg_list[0].length, DMA_TO_DEVICE); + ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device, + mad_send_wr->payload_mapping, + mad_send_wr->sg_list[1].length, DMA_TO_DEVICE); queued_send_wr = NULL; spin_lock_irqsave(&send_queue->lock, flags); list_del(&mad_list->list); @@ -2198,12 +2260,12 @@ static void mad_error_handler(struct ib_mad_port_private *port_priv, /* * IB MAD completion callback */ -static void ib_mad_completion_handler(void *data) +static void ib_mad_completion_handler(struct work_struct *work) { struct ib_mad_port_private *port_priv; struct ib_wc wc; - port_priv = (struct ib_mad_port_private *)data; + port_priv = container_of(work, struct ib_mad_port_private, work); ib_req_notify_cq(port_priv->cq, IB_CQ_NEXT_COMP); while (ib_poll_cq(port_priv->cq, 1, &wc) == 1) { @@ -2244,8 +2306,6 @@ static void cancel_mads(struct ib_mad_agent_private *mad_agent_priv) /* Empty wait list to prevent receives from finding a request */ list_splice_init(&mad_agent_priv->wait_list, &cancel_list); - /* Empty local completion list as well */ - list_splice_init(&mad_agent_priv->local_list, &cancel_list); spin_unlock_irqrestore(&mad_agent_priv->lock, flags); /* Report all cancelled requests */ @@ -2324,17 +2384,18 @@ void ib_cancel_mad(struct ib_mad_agent *mad_agent, } EXPORT_SYMBOL(ib_cancel_mad); -static void local_completions(void *data) +static void local_completions(struct work_struct *work) { struct ib_mad_agent_private *mad_agent_priv; struct ib_mad_local_private *local; struct ib_mad_agent_private *recv_mad_agent; unsigned long flags; - int recv = 0; + int free_mad; struct ib_wc wc; struct ib_mad_send_wc mad_send_wc; - mad_agent_priv = (struct ib_mad_agent_private *)data; + mad_agent_priv = + container_of(work, struct ib_mad_agent_private, local_work); spin_lock_irqsave(&mad_agent_priv->lock, flags); while (!list_empty(&mad_agent_priv->local_list)) { @@ -2343,19 +2404,21 @@ static void local_completions(void *data) completion_list); list_del(&local->completion_list); spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + free_mad = 0; if (local->mad_priv) { recv_mad_agent = local->recv_mad_agent; if (!recv_mad_agent) { printk(KERN_ERR PFX "No receive MAD agent for local completion\n"); + free_mad = 1; goto local_send_completion; } - recv = 1; /* * Defined behavior is to complete response * before request */ - build_smp_wc((unsigned long) local->mad_send_wr, + build_smp_wc(recv_mad_agent->agent.qp, + (unsigned long) local->mad_send_wr, be16_to_cpu(IB_LID_PERMISSIVE), 0, recv_mad_agent->agent.port_num, &wc); @@ -2394,7 +2457,7 @@ local_send_completion: spin_lock_irqsave(&mad_agent_priv->lock, flags); atomic_dec(&mad_agent_priv->refcount); - if (!recv) + if (free_mad) kmem_cache_free(ib_mad_cache, local->mad_priv); kfree(local); } @@ -2405,9 +2468,12 @@ static int retry_send(struct ib_mad_send_wr_private *mad_send_wr) { int ret; - if (!mad_send_wr->retries--) + if (!mad_send_wr->retries_left) return -ETIMEDOUT; + mad_send_wr->retries_left--; + mad_send_wr->send_buf.retries++; + mad_send_wr->timeout = msecs_to_jiffies(mad_send_wr->send_buf.timeout_ms); if (mad_send_wr->mad_agent_priv->agent.rmpp_version) { @@ -2434,14 +2500,15 @@ static int retry_send(struct ib_mad_send_wr_private *mad_send_wr) return ret; } -static void timeout_sends(void *data) +static void timeout_sends(struct work_struct *work) { struct ib_mad_agent_private *mad_agent_priv; struct ib_mad_send_wr_private *mad_send_wr; struct ib_mad_send_wc mad_send_wc; unsigned long flags, delay; - mad_agent_priv = (struct ib_mad_agent_private *)data; + mad_agent_priv = container_of(work, struct ib_mad_agent_private, + timed_work.work); mad_send_wc.vendor_err = 0; spin_lock_irqsave(&mad_agent_priv->lock, flags); @@ -2527,13 +2594,17 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info, break; } } - sg_list.addr = dma_map_single(qp_info->port_priv-> - device->dma_device, - &mad_priv->grh, - sizeof *mad_priv - - sizeof mad_priv->header, - DMA_FROM_DEVICE); - pci_unmap_addr_set(&mad_priv->header, mapping, sg_list.addr); + sg_list.addr = ib_dma_map_single(qp_info->port_priv->device, + &mad_priv->grh, + sizeof *mad_priv - + sizeof mad_priv->header, + DMA_FROM_DEVICE); + if (unlikely(ib_dma_mapping_error(qp_info->port_priv->device, + sg_list.addr))) { + ret = -ENOMEM; + break; + } + mad_priv->header.mapping = sg_list.addr; recv_wr.wr_id = (unsigned long)&mad_priv->header.mad_list; mad_priv->header.mad_list.mad_queue = recv_queue; @@ -2548,12 +2619,11 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info, list_del(&mad_priv->header.mad_list.list); recv_queue->count--; spin_unlock_irqrestore(&recv_queue->lock, flags); - dma_unmap_single(qp_info->port_priv->device->dma_device, - pci_unmap_addr(&mad_priv->header, - mapping), - sizeof *mad_priv - - sizeof mad_priv->header, - DMA_FROM_DEVICE); + ib_dma_unmap_single(qp_info->port_priv->device, + mad_priv->header.mapping, + sizeof *mad_priv - + sizeof mad_priv->header, + DMA_FROM_DEVICE); kmem_cache_free(ib_mad_cache, mad_priv); printk(KERN_ERR PFX "ib_post_recv failed: %d\n", ret); break; @@ -2572,6 +2642,9 @@ static void cleanup_recv_queue(struct ib_mad_qp_info *qp_info) struct ib_mad_private *recv; struct ib_mad_list_head *mad_list; + if (!qp_info->qp) + return; + while (!list_empty(&qp_info->recv_queue.list)) { mad_list = list_entry(qp_info->recv_queue.list.next, @@ -2585,11 +2658,11 @@ static void cleanup_recv_queue(struct ib_mad_qp_info *qp_info) /* Remove from posted receive MAD list */ list_del(&mad_list->list); - dma_unmap_single(qp_info->port_priv->device->dma_device, - pci_unmap_addr(&recv->header, mapping), - sizeof(struct ib_mad_private) - - sizeof(struct ib_mad_private_header), - DMA_FROM_DEVICE); + ib_dma_unmap_single(qp_info->port_priv->device, + recv->header.mapping, + sizeof(struct ib_mad_private) - + sizeof(struct ib_mad_private_header), + DMA_FROM_DEVICE); kmem_cache_free(ib_mad_cache, recv); } @@ -2604,6 +2677,7 @@ static int ib_mad_port_start(struct ib_mad_port_private *port_priv) int ret, i; struct ib_qp_attr *attr; struct ib_qp *qp; + u16 pkey_index; attr = kmalloc(sizeof *attr, GFP_KERNEL); if (!attr) { @@ -2611,14 +2685,22 @@ static int ib_mad_port_start(struct ib_mad_port_private *port_priv) return -ENOMEM; } + ret = ib_find_pkey(port_priv->device, port_priv->port_num, + IB_DEFAULT_PKEY_FULL, &pkey_index); + if (ret) + pkey_index = 0; + for (i = 0; i < IB_MAD_QPS_CORE; i++) { qp = port_priv->qp_info[i].qp; + if (!qp) + continue; + /* * PKey index for QP1 is irrelevant but * one is needed for the Reset to Init transition */ attr->qp_state = IB_QPS_INIT; - attr->pkey_index = 0; + attr->pkey_index = pkey_index; attr->qkey = (qp->qp_num == 0) ? 0 : IB_QP1_QKEY; ret = ib_modify_qp(qp, attr, IB_QP_STATE | IB_QP_PKEY_INDEX | IB_QP_QKEY); @@ -2654,6 +2736,9 @@ static int ib_mad_port_start(struct ib_mad_port_private *port_priv) } for (i = 0; i < IB_MAD_QPS_CORE; i++) { + if (!port_priv->qp_info[i].qp) + continue; + ret = ib_mad_post_receive_mads(&port_priv->qp_info[i], NULL); if (ret) { printk(KERN_ERR PFX "Couldn't post receive WRs\n"); @@ -2706,8 +2791,8 @@ 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; @@ -2722,8 +2807,8 @@ 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: @@ -2732,6 +2817,9 @@ error: static void destroy_mad_qp(struct ib_mad_qp_info *qp_info) { + if (!qp_info->qp) + return; + ib_destroy_qp(qp_info->qp); kfree(qp_info->snoop_table); } @@ -2747,6 +2835,7 @@ static int ib_mad_port_open(struct ib_device *device, struct ib_mad_port_private *port_priv; unsigned long flags; char name[sizeof "ib_mad123"]; + int has_smi; /* Create new device info */ port_priv = kzalloc(sizeof *port_priv, GFP_KERNEL); @@ -2762,10 +2851,14 @@ 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; + has_smi = rdma_port_get_link_layer(device, port_num) == IB_LINK_LAYER_INFINIBAND; + if (has_smi) + cq_size *= 2; + port_priv->cq = ib_create_cq(port_priv->device, ib_mad_thread_completion_handler, - NULL, port_priv, cq_size); + NULL, port_priv, cq_size, 0); if (IS_ERR(port_priv->cq)) { printk(KERN_ERR PFX "Couldn't create ib_mad CQ\n"); ret = PTR_ERR(port_priv->cq); @@ -2786,9 +2879,11 @@ static int ib_mad_port_open(struct ib_device *device, goto error5; } - ret = create_mad_qp(&port_priv->qp_info[0], IB_QPT_SMI); - if (ret) - goto error6; + if (has_smi) { + ret = create_mad_qp(&port_priv->qp_info[0], IB_QPT_SMI); + if (ret) + goto error6; + } ret = create_mad_qp(&port_priv->qp_info[1], IB_QPT_GSI); if (ret) goto error7; @@ -2799,7 +2894,7 @@ static int ib_mad_port_open(struct ib_device *device, ret = -ENOMEM; goto error8; } - INIT_WORK(&port_priv->work, ib_mad_completion_handler, port_priv); + INIT_WORK(&port_priv->work, ib_mad_completion_handler); spin_lock_irqsave(&ib_mad_port_list_lock, flags); list_add_tail(&port_priv->port_list, &ib_mad_port_list); @@ -2926,6 +3021,9 @@ static void ib_mad_remove_device(struct ib_device *device) { int i, num_ports, cur_port; + if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB) + return; + if (device->node_type == RDMA_NODE_IB_SWITCH) { num_ports = 1; cur_port = 0; @@ -2954,13 +3052,16 @@ 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), 0, SLAB_HWCACHE_ALIGN, - NULL, NULL); if (!ib_mad_cache) { printk(KERN_ERR PFX "Couldn't create ib_mad cache\n"); @@ -2992,4 +3093,3 @@ static void __exit ib_mad_cleanup_module(void) module_init(ib_mad_init_module); module_exit(ib_mad_cleanup_module); - |