iwcm.c

来自「linux 内核源代码」· C语言 代码 · 共 1,023 行 · 第 1/2 页

/*
 * Copyright (c) 2004, 2005 Intel Corporation.  All rights reserved.
 * Copyright (c) 2004 Topspin Corporation.  All rights reserved.
 * Copyright (c) 2004, 2005 Voltaire Corporation.  All rights reserved.
 * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
 * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved.
 * Copyright (c) 2005 Network Appliance, Inc. 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
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */
#include <linux/dma-mapping.h>
#include <linux/err.h>
#include <linux/idr.h>
#include <linux/interrupt.h>
#include <linux/rbtree.h>
#include <linux/spinlock.h>
#include <linux/workqueue.h>
#include <linux/completion.h>

#include <rdma/iw_cm.h>
#include <rdma/ib_addr.h>

#include "iwcm.h"

MODULE_AUTHOR("Tom Tucker");
MODULE_DESCRIPTION("iWARP CM");
MODULE_LICENSE("Dual BSD/GPL");

static struct workqueue_struct *iwcm_wq;
struct iwcm_work {
	struct work_struct work;
	struct iwcm_id_private *cm_id;
	struct list_head list;
	struct iw_cm_event event;
	struct list_head free_list;
};

/*
 * The following services provide a mechanism for pre-allocating iwcm_work
 * elements.  The design pre-allocates them  based on the cm_id type:
 *	LISTENING IDS: 	Get enough elements preallocated to handle the
 *			listen backlog.
 *	ACTIVE IDS:	4: CONNECT_REPLY, ESTABLISHED, DISCONNECT, CLOSE
 *	PASSIVE IDS:	3: ESTABLISHED, DISCONNECT, CLOSE
 *
 * Allocating them in connect and listen avoids having to deal
 * with allocation failures on the event upcall from the provider (which
 * is called in the interrupt context).
 *
 * One exception is when creating the cm_id for incoming connection requests.
 * There are two cases:
 * 1) in the event upcall, cm_event_handler(), for a listening cm_id.  If
 *    the backlog is exceeded, then no more connection request events will
 *    be processed.  cm_event_handler() returns -ENOMEM in this case.  Its up
 *    to the provider to reject the connection request.
 * 2) in the connection request workqueue handler, cm_conn_req_handler().
 *    If work elements cannot be allocated for the new connect request cm_id,
 *    then IWCM will call the provider reject method.  This is ok since
 *    cm_conn_req_handler() runs in the workqueue thread context.
 */

static struct iwcm_work *get_work(struct iwcm_id_private *cm_id_priv)
{
	struct iwcm_work *work;

	if (list_empty(&cm_id_priv->work_free_list))
		return NULL;
	work = list_entry(cm_id_priv->work_free_list.next, struct iwcm_work,
			  free_list);
	list_del_init(&work->free_list);
	return work;
}

static void put_work(struct iwcm_work *work)
{
	list_add(&work->free_list, &work->cm_id->work_free_list);
}

static void dealloc_work_entries(struct iwcm_id_private *cm_id_priv)
{
	struct list_head *e, *tmp;

	list_for_each_safe(e, tmp, &cm_id_priv->work_free_list)
		kfree(list_entry(e, struct iwcm_work, free_list));
}

static int alloc_work_entries(struct iwcm_id_private *cm_id_priv, int count)
{
	struct iwcm_work *work;

	BUG_ON(!list_empty(&cm_id_priv->work_free_list));
	while (count--) {
		work = kmalloc(sizeof(struct iwcm_work), GFP_KERNEL);
		if (!work) {
			dealloc_work_entries(cm_id_priv);
			return -ENOMEM;
		}
		work->cm_id = cm_id_priv;
		INIT_LIST_HEAD(&work->list);
		put_work(work);
	}
	return 0;
}

/*
 * Save private data from incoming connection requests to
 * iw_cm_event, so the low level driver doesn't have to. Adjust
 * the event ptr to point to the local copy.
 */
static int copy_private_data(struct iw_cm_event *event)
{
	void *p;

	p = kmemdup(event->private_data, event->private_data_len, GFP_ATOMIC);
	if (!p)
		return -ENOMEM;
	event->private_data = p;
	return 0;
}

static void free_cm_id(struct iwcm_id_private *cm_id_priv)
{
	dealloc_work_entries(cm_id_priv);
	kfree(cm_id_priv);
}

/*
 * Release a reference on cm_id. If the last reference is being
 * released, enable the waiting thread (in iw_destroy_cm_id) to
 * get woken up, and return 1 if a thread is already waiting.
 */
static int iwcm_deref_id(struct iwcm_id_private *cm_id_priv)
{
	BUG_ON(atomic_read(&cm_id_priv->refcount)==0);
	if (atomic_dec_and_test(&cm_id_priv->refcount)) {
		BUG_ON(!list_empty(&cm_id_priv->work_list));
		complete(&cm_id_priv->destroy_comp);
		return 1;
	}

	return 0;
}

static void add_ref(struct iw_cm_id *cm_id)
{
	struct iwcm_id_private *cm_id_priv;
	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
	atomic_inc(&cm_id_priv->refcount);
}

static void rem_ref(struct iw_cm_id *cm_id)
{
	struct iwcm_id_private *cm_id_priv;
	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
	if (iwcm_deref_id(cm_id_priv) &&
	    test_bit(IWCM_F_CALLBACK_DESTROY, &cm_id_priv->flags)) {
		BUG_ON(!list_empty(&cm_id_priv->work_list));
		free_cm_id(cm_id_priv);
	}
}

static int cm_event_handler(struct iw_cm_id *cm_id, struct iw_cm_event *event);

struct iw_cm_id *iw_create_cm_id(struct ib_device *device,
				 iw_cm_handler cm_handler,
				 void *context)
{
	struct iwcm_id_private *cm_id_priv;

	cm_id_priv = kzalloc(sizeof(*cm_id_priv), GFP_KERNEL);
	if (!cm_id_priv)
		return ERR_PTR(-ENOMEM);

	cm_id_priv->state = IW_CM_STATE_IDLE;
	cm_id_priv->id.device = device;
	cm_id_priv->id.cm_handler = cm_handler;
	cm_id_priv->id.context = context;
	cm_id_priv->id.event_handler = cm_event_handler;
	cm_id_priv->id.add_ref = add_ref;
	cm_id_priv->id.rem_ref = rem_ref;
	spin_lock_init(&cm_id_priv->lock);
	atomic_set(&cm_id_priv->refcount, 1);
	init_waitqueue_head(&cm_id_priv->connect_wait);
	init_completion(&cm_id_priv->destroy_comp);
	INIT_LIST_HEAD(&cm_id_priv->work_list);
	INIT_LIST_HEAD(&cm_id_priv->work_free_list);

	return &cm_id_priv->id;
}
EXPORT_SYMBOL(iw_create_cm_id);


static int iwcm_modify_qp_err(struct ib_qp *qp)
{
	struct ib_qp_attr qp_attr;

	if (!qp)
		return -EINVAL;

	qp_attr.qp_state = IB_QPS_ERR;
	return ib_modify_qp(qp, &qp_attr, IB_QP_STATE);
}

/*
 * This is really the RDMAC CLOSING state. It is most similar to the
 * IB SQD QP state.
 */
static int iwcm_modify_qp_sqd(struct ib_qp *qp)
{
	struct ib_qp_attr qp_attr;

	BUG_ON(qp == NULL);
	qp_attr.qp_state = IB_QPS_SQD;
	return ib_modify_qp(qp, &qp_attr, IB_QP_STATE);
}

/*
 * CM_ID <-- CLOSING
 *
 * Block if a passive or active connection is currently being processed. Then
 * process the event as follows:
 * - If we are ESTABLISHED, move to CLOSING and modify the QP state
 *   based on the abrupt flag
 * - If the connection is already in the CLOSING or IDLE state, the peer is
 *   disconnecting concurrently with us and we've already seen the
 *   DISCONNECT event -- ignore the request and return 0
 * - Disconnect on a listening endpoint returns -EINVAL
 */
int iw_cm_disconnect(struct iw_cm_id *cm_id, int abrupt)
{
	struct iwcm_id_private *cm_id_priv;
	unsigned long flags;
	int ret = 0;
	struct ib_qp *qp = NULL;

	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
	/* Wait if we're currently in a connect or accept downcall */
	wait_event(cm_id_priv->connect_wait,
		   !test_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags));

	spin_lock_irqsave(&cm_id_priv->lock, flags);
	switch (cm_id_priv->state) {
	case IW_CM_STATE_ESTABLISHED:
		cm_id_priv->state = IW_CM_STATE_CLOSING;

		/* QP could be <nul> for user-mode client */
		if (cm_id_priv->qp)
			qp = cm_id_priv->qp;
		else
			ret = -EINVAL;
		break;
	case IW_CM_STATE_LISTEN:
		ret = -EINVAL;
		break;
	case IW_CM_STATE_CLOSING:
		/* remote peer closed first */
	case IW_CM_STATE_IDLE:
		/* accept or connect returned !0 */
		break;
	case IW_CM_STATE_CONN_RECV:
		/*
		 * App called disconnect before/without calling accept after
		 * connect_request event delivered.
		 */
		break;
	case IW_CM_STATE_CONN_SENT:
		/* Can only get here if wait above fails */
	default:
		BUG();
	}
	spin_unlock_irqrestore(&cm_id_priv->lock, flags);

	if (qp) {
		if (abrupt)
			ret = iwcm_modify_qp_err(qp);
		else
			ret = iwcm_modify_qp_sqd(qp);

		/*
		 * If both sides are disconnecting the QP could
		 * already be in ERR or SQD states
		 */
		ret = 0;
	}

	return ret;
}
EXPORT_SYMBOL(iw_cm_disconnect);

/*
 * CM_ID <-- DESTROYING
 *
 * Clean up all resources associated with the connection and release
 * the initial reference taken by iw_create_cm_id.
 */
static void destroy_cm_id(struct iw_cm_id *cm_id)
{
	struct iwcm_id_private *cm_id_priv;
	unsigned long flags;
	int ret;

	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
	/*
	 * Wait if we're currently in a connect or accept downcall. A
	 * listening endpoint should never block here.
	 */
	wait_event(cm_id_priv->connect_wait,
		   !test_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags));

	spin_lock_irqsave(&cm_id_priv->lock, flags);
	switch (cm_id_priv->state) {
	case IW_CM_STATE_LISTEN:
		cm_id_priv->state = IW_CM_STATE_DESTROYING;
		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
		/* destroy the listening endpoint */
		ret = cm_id->device->iwcm->destroy_listen(cm_id);
		spin_lock_irqsave(&cm_id_priv->lock, flags);
		break;
	case IW_CM_STATE_ESTABLISHED:
		cm_id_priv->state = IW_CM_STATE_DESTROYING;
		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
		/* Abrupt close of the connection */
		(void)iwcm_modify_qp_err(cm_id_priv->qp);
		spin_lock_irqsave(&cm_id_priv->lock, flags);
		break;
	case IW_CM_STATE_IDLE:
	case IW_CM_STATE_CLOSING:
		cm_id_priv->state = IW_CM_STATE_DESTROYING;
		break;
	case IW_CM_STATE_CONN_RECV:
		/*
		 * App called destroy before/without calling accept after
		 * receiving connection request event notification or
		 * returned non zero from the event callback function.
		 * In either case, must tell the provider to reject.
		 */
		cm_id_priv->state = IW_CM_STATE_DESTROYING;
		break;
	case IW_CM_STATE_CONN_SENT:
	case IW_CM_STATE_DESTROYING:
	default:
		BUG();
		break;
	}
	if (cm_id_priv->qp) {
		cm_id_priv->id.device->iwcm->rem_ref(cm_id_priv->qp);
		cm_id_priv->qp = NULL;
	}
	spin_unlock_irqrestore(&cm_id_priv->lock, flags);

	(void)iwcm_deref_id(cm_id_priv);
}

/*
 * This function is only called by the application thread and cannot
 * be called by the event thread. The function will wait for all
 * references to be released on the cm_id and then kfree the cm_id
 * object.
 */
void iw_destroy_cm_id(struct iw_cm_id *cm_id)
{
	struct iwcm_id_private *cm_id_priv;

	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
	BUG_ON(test_bit(IWCM_F_CALLBACK_DESTROY, &cm_id_priv->flags));

	destroy_cm_id(cm_id);

	wait_for_completion(&cm_id_priv->destroy_comp);

	free_cm_id(cm_id_priv);
}
EXPORT_SYMBOL(iw_destroy_cm_id);

/*
 * CM_ID <-- LISTEN
 *
 * Start listening for connect requests. Generates one CONNECT_REQUEST
 * event for each inbound connect request.
 */
int iw_cm_listen(struct iw_cm_id *cm_id, int backlog)
{
	struct iwcm_id_private *cm_id_priv;
	unsigned long flags;
	int ret;

	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);

	ret = alloc_work_entries(cm_id_priv, backlog);
	if (ret)
		return ret;

	spin_lock_irqsave(&cm_id_priv->lock, flags);
	switch (cm_id_priv->state) {
	case IW_CM_STATE_IDLE:
		cm_id_priv->state = IW_CM_STATE_LISTEN;
		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
		ret = cm_id->device->iwcm->create_listen(cm_id, backlog);
		if (ret)
			cm_id_priv->state = IW_CM_STATE_IDLE;
		spin_lock_irqsave(&cm_id_priv->lock, flags);
		break;
	default:
		ret = -EINVAL;
	}
	spin_unlock_irqrestore(&cm_id_priv->lock, flags);

	return ret;
}
EXPORT_SYMBOL(iw_cm_listen);

/*
 * CM_ID <-- IDLE
 *
 * Rejects an inbound connection request. No events are generated.
 */
int iw_cm_reject(struct iw_cm_id *cm_id,
		 const void *private_data,
		 u8 private_data_len)
{
	struct iwcm_id_private *cm_id_priv;
	unsigned long flags;
	int ret;

	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
	set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);

	spin_lock_irqsave(&cm_id_priv->lock, flags);
	if (cm_id_priv->state != IW_CM_STATE_CONN_RECV) {
		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
		clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
		wake_up_all(&cm_id_priv->connect_wait);
		return -EINVAL;
	}
	cm_id_priv->state = IW_CM_STATE_IDLE;
	spin_unlock_irqrestore(&cm_id_priv->lock, flags);

	ret = cm_id->device->iwcm->reject(cm_id, private_data,
					  private_data_len);

	clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
	wake_up_all(&cm_id_priv->connect_wait);

	return ret;
}
EXPORT_SYMBOL(iw_cm_reject);

/*
 * CM_ID <-- ESTABLISHED
 *
 * Accepts an inbound connection request and generates an ESTABLISHED
 * event. Callers of iw_cm_disconnect and iw_destroy_cm_id will block
 * until the ESTABLISHED event is received from the provider.
 */
int iw_cm_accept(struct iw_cm_id *cm_id,
		 struct iw_cm_conn_param *iw_param)
{
	struct iwcm_id_private *cm_id_priv;
	struct ib_qp *qp;
	unsigned long flags;
	int ret;

	cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
	set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);

	spin_lock_irqsave(&cm_id_priv->lock, flags);
	if (cm_id_priv->state != IW_CM_STATE_CONN_RECV) {
		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
		clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
		wake_up_all(&cm_id_priv->connect_wait);
		return -EINVAL;
	}
	/* Get the ib_qp given the QPN */
	qp = cm_id->device->iwcm->get_qp(cm_id->device, iw_param->qpn);
	if (!qp) {
		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
		return -EINVAL;
	}
	cm_id->device->iwcm->add_ref(qp);
	cm_id_priv->qp = qp;
	spin_unlock_irqrestore(&cm_id_priv->lock, flags);

	ret = cm_id->device->iwcm->accept(cm_id, iw_param);
	if (ret) {
		/* An error on accept precludes provider events */