mad.c

h内核

		mad_send_wr->retry = 1;
	}
	spin_unlock_irqrestore(&qp_info->send_queue.lock, flags);
}

static void mad_error_handler(struct ib_mad_port_private *port_priv,
			      struct ib_wc *wc)
{
	struct ib_mad_list_head *mad_list;
	struct ib_mad_qp_info *qp_info;
	struct ib_mad_send_wr_private *mad_send_wr;
	int ret;

	/* Determine if failure was a send or receive */
	mad_list = (struct ib_mad_list_head *)(unsigned long)wc->wr_id;
	qp_info = mad_list->mad_queue->qp_info;
	if (mad_list->mad_queue == &qp_info->recv_queue)
		/*
		 * Receive errors indicate that the QP has entered the error
		 * state - error handling/shutdown code will cleanup
		 */
		return;

	/*
	 * Send errors will transition the QP to SQE - move
	 * QP to RTS and repost flushed work requests
	 */
	mad_send_wr = container_of(mad_list, struct ib_mad_send_wr_private,
				   mad_list);
	if (wc->status == IB_WC_WR_FLUSH_ERR) {
		if (mad_send_wr->retry) {
			/* Repost send */
			struct ib_send_wr *bad_send_wr;

			mad_send_wr->retry = 0;
			ret = ib_post_send(qp_info->qp, &mad_send_wr->send_wr,
					&bad_send_wr);
			if (ret)
				ib_mad_send_done_handler(port_priv, wc);
		} else
			ib_mad_send_done_handler(port_priv, wc);
	} else {
		struct ib_qp_attr *attr;

		/* Transition QP to RTS and fail offending send */
		attr = kmalloc(sizeof *attr, GFP_KERNEL);
		if (attr) {
			attr->qp_state = IB_QPS_RTS;
			attr->cur_qp_state = IB_QPS_SQE;
			ret = ib_modify_qp(qp_info->qp, attr,
					   IB_QP_STATE | IB_QP_CUR_STATE);
			kfree(attr);
			if (ret)
				printk(KERN_ERR PFX "mad_error_handler - "
				       "ib_modify_qp to RTS : %d\n", ret);
			else
				mark_sends_for_retry(qp_info);
		}
		ib_mad_send_done_handler(port_priv, wc);
	}
}

/*
 * IB MAD completion callback
 */
static void ib_mad_completion_handler(void *data)
{
	struct ib_mad_port_private *port_priv;
	struct ib_wc wc;

	port_priv = (struct ib_mad_port_private *)data;
	ib_req_notify_cq(port_priv->cq, IB_CQ_NEXT_COMP);

	while (ib_poll_cq(port_priv->cq, 1, &wc) == 1) {
		if (wc.status == IB_WC_SUCCESS) {
			switch (wc.opcode) {
			case IB_WC_SEND:
				ib_mad_send_done_handler(port_priv, &wc);
				break;
			case IB_WC_RECV:
				ib_mad_recv_done_handler(port_priv, &wc);
				break;
			default:
				BUG_ON(1);
				break;
			}
		} else
			mad_error_handler(port_priv, &wc);
	}
}

static void cancel_mads(struct ib_mad_agent_private *mad_agent_priv)
{
	unsigned long flags;
	struct ib_mad_send_wr_private *mad_send_wr, *temp_mad_send_wr;
	struct ib_mad_send_wc mad_send_wc;
	struct list_head cancel_list;

	INIT_LIST_HEAD(&cancel_list);

	spin_lock_irqsave(&mad_agent_priv->lock, flags);
	list_for_each_entry_safe(mad_send_wr, temp_mad_send_wr,
				 &mad_agent_priv->send_list, agent_list) {
		if (mad_send_wr->status == IB_WC_SUCCESS) {
 			mad_send_wr->status = IB_WC_WR_FLUSH_ERR;
			mad_send_wr->refcount -= (mad_send_wr->timeout > 0);
		}
	}

	/* Empty wait list to prevent receives from finding a request */
	list_splice_init(&mad_agent_priv->wait_list, &cancel_list);
	spin_unlock_irqrestore(&mad_agent_priv->lock, flags);

	/* Report all cancelled requests */
	mad_send_wc.status = IB_WC_WR_FLUSH_ERR;
	mad_send_wc.vendor_err = 0;

	list_for_each_entry_safe(mad_send_wr, temp_mad_send_wr,
				 &cancel_list, agent_list) {
		mad_send_wc.wr_id = mad_send_wr->wr_id;
		mad_agent_priv->agent.send_handler(&mad_agent_priv->agent,
						   &mad_send_wc);

		list_del(&mad_send_wr->agent_list);
		kfree(mad_send_wr);
		atomic_dec(&mad_agent_priv->refcount);
	}
}

static struct ib_mad_send_wr_private*
find_send_by_wr_id(struct ib_mad_agent_private *mad_agent_priv,
		   u64 wr_id)
{
	struct ib_mad_send_wr_private *mad_send_wr;

	list_for_each_entry(mad_send_wr, &mad_agent_priv->wait_list,
			    agent_list) {
		if (mad_send_wr->wr_id == wr_id)
			return mad_send_wr;
	}

	list_for_each_entry(mad_send_wr, &mad_agent_priv->send_list,
			    agent_list) {
		if (mad_send_wr->wr_id == wr_id)
			return mad_send_wr;
	}
	return NULL;
}

void ib_cancel_mad(struct ib_mad_agent *mad_agent,
		  u64 wr_id)
{
	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;

	mad_agent_priv = container_of(mad_agent, struct ib_mad_agent_private,
				      agent);
	spin_lock_irqsave(&mad_agent_priv->lock, flags);
	mad_send_wr = find_send_by_wr_id(mad_agent_priv, wr_id);
	if (!mad_send_wr) {
		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
		goto out;
	}

	if (mad_send_wr->status == IB_WC_SUCCESS)
		mad_send_wr->refcount -= (mad_send_wr->timeout > 0);

	if (mad_send_wr->refcount != 0) {
		mad_send_wr->status = IB_WC_WR_FLUSH_ERR;
		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
		goto out;
	}

	list_del(&mad_send_wr->agent_list);
	adjust_timeout(mad_agent_priv);
	spin_unlock_irqrestore(&mad_agent_priv->lock, flags);

	mad_send_wc.status = IB_WC_WR_FLUSH_ERR;
	mad_send_wc.vendor_err = 0;
	mad_send_wc.wr_id = mad_send_wr->wr_id;
	mad_agent_priv->agent.send_handler(&mad_agent_priv->agent,
					   &mad_send_wc);

	kfree(mad_send_wr);
	if (atomic_dec_and_test(&mad_agent_priv->refcount))
		wake_up(&mad_agent_priv->wait);

out:
	return;
}
EXPORT_SYMBOL(ib_cancel_mad);

static void local_completions(void *data)
{
	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;
	struct ib_wc wc;
	struct ib_mad_send_wc mad_send_wc;

	mad_agent_priv = (struct ib_mad_agent_private *)data;

	spin_lock_irqsave(&mad_agent_priv->lock, flags);
	while (!list_empty(&mad_agent_priv->local_list)) {
		local = list_entry(mad_agent_priv->local_list.next,
				   struct ib_mad_local_private,
				   completion_list);
		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
		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");
				kmem_cache_free(ib_mad_cache, local->mad_priv);
				goto local_send_completion;
			}

			/*
			 * Defined behavior is to complete response
			 * before request
			 */
			build_smp_wc(local->wr_id, IB_LID_PERMISSIVE,
				     0 /* pkey index */,
				     recv_mad_agent->agent.port_num, &wc);

			local->mad_priv->header.recv_wc.wc = &wc;
			local->mad_priv->header.recv_wc.mad_len =
						sizeof(struct ib_mad);
			INIT_LIST_HEAD(&local->mad_priv->header.recv_wc.recv_buf.list);
			local->mad_priv->header.recv_wc.recv_buf.grh = NULL;
			local->mad_priv->header.recv_wc.recv_buf.mad =
						&local->mad_priv->mad.mad;
			if (atomic_read(&recv_mad_agent->qp_info->snoop_count))
				snoop_recv(recv_mad_agent->qp_info,
					  &local->mad_priv->header.recv_wc,
					   IB_MAD_SNOOP_RECVS);
			recv_mad_agent->agent.recv_handler(
						&recv_mad_agent->agent,
						&local->mad_priv->header.recv_wc);
			spin_lock_irqsave(&recv_mad_agent->lock, flags);
			atomic_dec(&recv_mad_agent->refcount);
			spin_unlock_irqrestore(&recv_mad_agent->lock, flags);
		}

local_send_completion:
		/* Complete send */
		mad_send_wc.status = IB_WC_SUCCESS;
		mad_send_wc.vendor_err = 0;
		mad_send_wc.wr_id = local->wr_id;
		if (atomic_read(&mad_agent_priv->qp_info->snoop_count))
			snoop_send(mad_agent_priv->qp_info, &local->send_wr,
				  &mad_send_wc,
				   IB_MAD_SNOOP_SEND_COMPLETIONS);
		mad_agent_priv->agent.send_handler(&mad_agent_priv->agent,
						   &mad_send_wc);

		spin_lock_irqsave(&mad_agent_priv->lock, flags);
		list_del(&local->completion_list);
		atomic_dec(&mad_agent_priv->refcount);
		kfree(local);
	}
	spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
}

static void timeout_sends(void *data)
{
	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_send_wc.status = IB_WC_RESP_TIMEOUT_ERR;
	mad_send_wc.vendor_err = 0;

	spin_lock_irqsave(&mad_agent_priv->lock, flags);
	while (!list_empty(&mad_agent_priv->wait_list)) {
		mad_send_wr = list_entry(mad_agent_priv->wait_list.next,
					 struct ib_mad_send_wr_private,
					 agent_list);

		if (time_after(mad_send_wr->timeout, jiffies)) {
			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);
			break;
		}

		list_del(&mad_send_wr->agent_list);
		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);

		mad_send_wc.wr_id = mad_send_wr->wr_id;
		mad_agent_priv->agent.send_handler(&mad_agent_priv->agent,
						   &mad_send_wc);

		kfree(mad_send_wr);
		atomic_dec(&mad_agent_priv->refcount);
		spin_lock_irqsave(&mad_agent_priv->lock, flags);
	}
	spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
}

static void ib_mad_thread_completion_handler(struct ib_cq *cq)
{
	struct ib_mad_port_private *port_priv = cq->cq_context;

	queue_work(port_priv->wq, &port_priv->work);
}

/*
 * Allocate receive MADs and post receive WRs for them
 */
static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
				    struct ib_mad_private *mad)
{
	unsigned long flags;
	int post, ret;
	struct ib_mad_private *mad_priv;
	struct ib_sge sg_list;
	struct ib_recv_wr recv_wr, *bad_recv_wr;
	struct ib_mad_queue *recv_queue = &qp_info->recv_queue;

	/* Initialize common scatter list fields */
	sg_list.length = sizeof *mad_priv - sizeof mad_priv->header;
	sg_list.lkey = (*qp_info->port_priv->mr).lkey;

	/* Initialize common receive WR fields */
	recv_wr.next = NULL;
	recv_wr.sg_list = &sg_list;
	recv_wr.num_sge = 1;
	recv_wr.recv_flags = IB_RECV_SIGNALED;

	do {
		/* Allocate and map receive buffer */
		if (mad) {
			mad_priv = mad;
			mad = NULL;
		} else {
			mad_priv = kmem_cache_alloc(ib_mad_cache, GFP_KERNEL);
			if (!mad_priv) {
				printk(KERN_ERR PFX "No memory for receive buffer\n");
				ret = -ENOMEM;
				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);
		recv_wr.wr_id = (unsigned long)&mad_priv->header.mad_list;
		mad_priv->header.mad_list.mad_queue = recv_queue;

		/* Post receive WR */
		spin_lock_irqsave(&recv_queue->lock, flags);
		post = (++recv_queue->count < recv_queue->max_active);
		list_add_tail(&mad_priv->header.mad_list.list, &recv_queue->list);
		spin_unlock_irqrestore(&recv_queue->lock, flags);
		ret = ib_post_recv(qp_info->qp, &recv_wr, &bad_recv_wr);
		if (ret) {
			spin_lock_irqsave(&recv_queue->lock, flags);
			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);
			kmem_cache_free(ib_mad_cache, mad_priv);
			printk(KERN_ERR PFX "ib_post_recv failed: %d\n", ret);
			break;
		}
	} while (post);

	return ret;
}

/*
 * Return all the posted receive MADs
 */
static void cleanup_recv_queue(struct ib_mad_qp_info *qp_info)
{
	struct ib_mad_private_header *mad_priv_hdr;
	struct ib_mad_private *recv;
	struct ib_mad_list_head *mad_list;

	while (!list_empty(&qp_info->recv_queue.list)) {

		mad_list = list_entry(qp_info->recv_queue.list.next,
				      struct ib_mad_list_head, list);
		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);

		/* Remove from posted receive MAD list */
		list_del(&mad_list->list);

		/* Undo PCI mapping */
		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);
		kmem_cache_free(ib_mad_cache, recv);
	}

	qp_info->recv_queue.count = 0;
}

/*
 * Start the port
 */
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;

	attr = kmalloc(sizeof *attr, GFP_KERNEL);
 	if (!attr) {
		printk(KERN_ERR PFX "Couldn't kmalloc ib_qp_attr\n");
		return -ENOMEM;
	}

	for (i = 0; i < IB_MAD_QPS_CORE; i++) {
		qp = port_priv->qp_info[i].qp;
		/*
		 * 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->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);
		if (ret) {
			printk(KERN_ERR PFX "Couldn't change QP%d state to "
			       "INIT: %d\n", i, ret);
			goto out;
		}

		attr->qp_state = IB_QPS_RTR;
		ret = ib_modify_qp(qp, attr, IB_QP_STATE);
		if (ret) {
			printk(KERN_ERR PFX "Couldn't change QP%d state to "
			       "RTR: %d\n", i, ret);
			goto out;
		}

		attr->qp_sta