[deliverable/linux.git] / net / vmw_vsock / virtio_transport.c

/*
 * virtio transport for vsock
 *
 * Copyright (C) 2013-2015 Red Hat, Inc.
 * Author: Asias He <asias@redhat.com>
 *         Stefan Hajnoczi <stefanha@redhat.com>
 *
 * Some of the code is take from Gerd Hoffmann <kraxel@redhat.com>'s
 * early virtio-vsock proof-of-concept bits.
 *
 * This work is licensed under the terms of the GNU GPL, version 2.
 */
#include <linux/spinlock.h>
#include <linux/module.h>
#include <linux/list.h>
#include <linux/atomic.h>
#include <linux/virtio.h>
#include <linux/virtio_ids.h>
#include <linux/virtio_config.h>
#include <linux/virtio_vsock.h>
#include <net/sock.h>
#include <linux/mutex.h>
#include <net/af_vsock.h>

static struct workqueue_struct *virtio_vsock_workqueue;
static struct virtio_vsock *the_virtio_vsock;
static DEFINE_MUTEX(the_virtio_vsock_mutex); /* protects the_virtio_vsock */

struct virtio_vsock {
	struct virtio_device *vdev;
	struct virtqueue *vqs[VSOCK_VQ_MAX];

	/* Virtqueue processing is deferred to a workqueue */
	struct work_struct tx_work;
	struct work_struct rx_work;
	struct work_struct event_work;

	/* The following fields are protected by tx_lock.  vqs[VSOCK_VQ_TX]
	 * must be accessed with tx_lock held.
	 */
	struct mutex tx_lock;

	struct work_struct send_pkt_work;
	spinlock_t send_pkt_list_lock;
	struct list_head send_pkt_list;

	atomic_t queued_replies;

	/* The following fields are protected by rx_lock.  vqs[VSOCK_VQ_RX]
	 * must be accessed with rx_lock held.
	 */
	struct mutex rx_lock;
	int rx_buf_nr;
	int rx_buf_max_nr;

	/* The following fields are protected by event_lock.
	 * vqs[VSOCK_VQ_EVENT] must be accessed with event_lock held.
	 */
	struct mutex event_lock;
	struct virtio_vsock_event event_list[8];

	u32 guest_cid;
};

static struct virtio_vsock *virtio_vsock_get(void)
{
	return the_virtio_vsock;
}

static u32 virtio_transport_get_local_cid(void)
{
	struct virtio_vsock *vsock = virtio_vsock_get();

	return vsock->guest_cid;
}

static void
virtio_transport_send_pkt_work(struct work_struct *work)
{
	struct virtio_vsock *vsock =
		container_of(work, struct virtio_vsock, send_pkt_work);
	struct virtqueue *vq;
	bool added = false;
	bool restart_rx = false;

	mutex_lock(&vsock->tx_lock);

	vq = vsock->vqs[VSOCK_VQ_TX];

	/* Avoid unnecessary interrupts while we're processing the ring */
	virtqueue_disable_cb(vq);

	for (;;) {
		struct virtio_vsock_pkt *pkt;
		struct scatterlist hdr, buf, *sgs[2];
		int ret, in_sg = 0, out_sg = 0;
		bool reply;

		spin_lock_bh(&vsock->send_pkt_list_lock);
		if (list_empty(&vsock->send_pkt_list)) {
			spin_unlock_bh(&vsock->send_pkt_list_lock);
			virtqueue_enable_cb(vq);
			break;
		}

		pkt = list_first_entry(&vsock->send_pkt_list,
				       struct virtio_vsock_pkt, list);
		list_del_init(&pkt->list);
		spin_unlock_bh(&vsock->send_pkt_list_lock);

		reply = pkt->reply;

		sg_init_one(&hdr, &pkt->hdr, sizeof(pkt->hdr));
		sgs[out_sg++] = &hdr;
		if (pkt->buf) {
			sg_init_one(&buf, pkt->buf, pkt->len);
			sgs[out_sg++] = &buf;
		}

		ret = virtqueue_add_sgs(vq, sgs, out_sg, in_sg, pkt, GFP_KERNEL);
		if (ret < 0) {
			spin_lock_bh(&vsock->send_pkt_list_lock);
			list_add(&pkt->list, &vsock->send_pkt_list);
			spin_unlock_bh(&vsock->send_pkt_list_lock);

			if (!virtqueue_enable_cb(vq) && ret == -ENOSPC)
				continue; /* retry now that we have more space */
			break;
		}

		if (reply) {
			struct virtqueue *rx_vq = vsock->vqs[VSOCK_VQ_RX];
			int val;

			val = atomic_dec_return(&vsock->queued_replies);

			/* Do we now have resources to resume rx processing? */
			if (val + 1 == virtqueue_get_vring_size(rx_vq))
				restart_rx = true;
		}

		added = true;
	}

	if (added)
		virtqueue_kick(vq);

	mutex_unlock(&vsock->tx_lock);

	if (restart_rx)
		queue_work(virtio_vsock_workqueue, &vsock->rx_work);
}

static int
virtio_transport_send_pkt(struct virtio_vsock_pkt *pkt)
{
	struct virtio_vsock *vsock;
	int len = pkt->len;

	vsock = virtio_vsock_get();
	if (!vsock) {
		virtio_transport_free_pkt(pkt);
		return -ENODEV;
	}

	if (pkt->reply)
		atomic_inc(&vsock->queued_replies);

	spin_lock_bh(&vsock->send_pkt_list_lock);
	list_add_tail(&pkt->list, &vsock->send_pkt_list);
	spin_unlock_bh(&vsock->send_pkt_list_lock);

	queue_work(virtio_vsock_workqueue, &vsock->send_pkt_work);
	return len;
}

static void virtio_vsock_rx_fill(struct virtio_vsock *vsock)
{
	int buf_len = VIRTIO_VSOCK_DEFAULT_RX_BUF_SIZE;
	struct virtio_vsock_pkt *pkt;
	struct scatterlist hdr, buf, *sgs[2];
	struct virtqueue *vq;
	int ret;

	vq = vsock->vqs[VSOCK_VQ_RX];

	do {
		pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
		if (!pkt)
			break;

		pkt->buf = kmalloc(buf_len, GFP_KERNEL);
		if (!pkt->buf) {
			virtio_transport_free_pkt(pkt);
			break;
		}

		pkt->len = buf_len;

		sg_init_one(&hdr, &pkt->hdr, sizeof(pkt->hdr));
		sgs[0] = &hdr;

		sg_init_one(&buf, pkt->buf, buf_len);
		sgs[1] = &buf;
		ret = virtqueue_add_sgs(vq, sgs, 0, 2, pkt, GFP_KERNEL);
		if (ret) {
			virtio_transport_free_pkt(pkt);
			break;
		}
		vsock->rx_buf_nr++;
	} while (vq->num_free);
	if (vsock->rx_buf_nr > vsock->rx_buf_max_nr)
		vsock->rx_buf_max_nr = vsock->rx_buf_nr;
	virtqueue_kick(vq);
}

static void virtio_transport_tx_work(struct work_struct *work)
{
	struct virtio_vsock *vsock =
		container_of(work, struct virtio_vsock, tx_work);
	struct virtqueue *vq;
	bool added = false;

	vq = vsock->vqs[VSOCK_VQ_TX];
	mutex_lock(&vsock->tx_lock);
	do {
		struct virtio_vsock_pkt *pkt;
		unsigned int len;

		virtqueue_disable_cb(vq);
		while ((pkt = virtqueue_get_buf(vq, &len)) != NULL) {
			virtio_transport_free_pkt(pkt);
			added = true;
		}
	} while (!virtqueue_enable_cb(vq));
	mutex_unlock(&vsock->tx_lock);

	if (added)
		queue_work(virtio_vsock_workqueue, &vsock->send_pkt_work);
}

/* Is there space left for replies to rx packets? */
static bool virtio_transport_more_replies(struct virtio_vsock *vsock)
{
	struct virtqueue *vq = vsock->vqs[VSOCK_VQ_RX];
	int val;

	smp_rmb(); /* paired with atomic_inc() and atomic_dec_return() */
	val = atomic_read(&vsock->queued_replies);

	return val < virtqueue_get_vring_size(vq);
}

static void virtio_transport_rx_work(struct work_struct *work)
{
	struct virtio_vsock *vsock =
		container_of(work, struct virtio_vsock, rx_work);
	struct virtqueue *vq;

	vq = vsock->vqs[VSOCK_VQ_RX];

	mutex_lock(&vsock->rx_lock);

	do {
		virtqueue_disable_cb(vq);
		for (;;) {
			struct virtio_vsock_pkt *pkt;
			unsigned int len;

			if (!virtio_transport_more_replies(vsock)) {
				/* Stop rx until the device processes already
				 * pending replies.  Leave rx virtqueue
				 * callbacks disabled.
				 */
				goto out;
			}

			pkt = virtqueue_get_buf(vq, &len);
			if (!pkt) {
				break;
			}

			vsock->rx_buf_nr--;

			/* Drop short/long packets */
			if (unlikely(len < sizeof(pkt->hdr) ||
				     len > sizeof(pkt->hdr) + pkt->len)) {
				virtio_transport_free_pkt(pkt);
				continue;
			}

			pkt->len = len - sizeof(pkt->hdr);
			virtio_transport_recv_pkt(pkt);
		}
	} while (!virtqueue_enable_cb(vq));

out:
	if (vsock->rx_buf_nr < vsock->rx_buf_max_nr / 2)
		virtio_vsock_rx_fill(vsock);
	mutex_unlock(&vsock->rx_lock);
}

/* event_lock must be held */
static int virtio_vsock_event_fill_one(struct virtio_vsock *vsock,
				       struct virtio_vsock_event *event)
{
	struct scatterlist sg;
	struct virtqueue *vq;

	vq = vsock->vqs[VSOCK_VQ_EVENT];

	sg_init_one(&sg, event, sizeof(*event));

	return virtqueue_add_inbuf(vq, &sg, 1, event, GFP_KERNEL);
}

/* event_lock must be held */
static void virtio_vsock_event_fill(struct virtio_vsock *vsock)
{
	size_t i;

	for (i = 0; i < ARRAY_SIZE(vsock->event_list); i++) {
		struct virtio_vsock_event *event = &vsock->event_list[i];

		virtio_vsock_event_fill_one(vsock, event);
	}

	virtqueue_kick(vsock->vqs[VSOCK_VQ_EVENT]);
}

static void virtio_vsock_reset_sock(struct sock *sk)
{
	lock_sock(sk);
	sk->sk_state = SS_UNCONNECTED;
	sk->sk_err = ECONNRESET;
	sk->sk_error_report(sk);
	release_sock(sk);
}

static void virtio_vsock_update_guest_cid(struct virtio_vsock *vsock)
{
	struct virtio_device *vdev = vsock->vdev;
	u64 guest_cid;

	vdev->config->get(vdev, offsetof(struct virtio_vsock_config, guest_cid),
			  &guest_cid, sizeof(guest_cid));
	vsock->guest_cid = le64_to_cpu(guest_cid);
}

/* event_lock must be held */
static void virtio_vsock_event_handle(struct virtio_vsock *vsock,
				      struct virtio_vsock_event *event)
{
	switch (le32_to_cpu(event->id)) {
	case VIRTIO_VSOCK_EVENT_TRANSPORT_RESET:
		virtio_vsock_update_guest_cid(vsock);
		vsock_for_each_connected_socket(virtio_vsock_reset_sock);
		break;
	}
}

static void virtio_transport_event_work(struct work_struct *work)
{
	struct virtio_vsock *vsock =
		container_of(work, struct virtio_vsock, event_work);
	struct virtqueue *vq;

	vq = vsock->vqs[VSOCK_VQ_EVENT];

	mutex_lock(&vsock->event_lock);

	do {
		struct virtio_vsock_event *event;
		unsigned int len;

		virtqueue_disable_cb(vq);
		while ((event = virtqueue_get_buf(vq, &len)) != NULL) {
			if (len == sizeof(*event))
				virtio_vsock_event_handle(vsock, event);

			virtio_vsock_event_fill_one(vsock, event);
		}
	} while (!virtqueue_enable_cb(vq));

	virtqueue_kick(vsock->vqs[VSOCK_VQ_EVENT]);

	mutex_unlock(&vsock->event_lock);
}

static void virtio_vsock_event_done(struct virtqueue *vq)
{
	struct virtio_vsock *vsock = vq->vdev->priv;

	if (!vsock)
		return;
	queue_work(virtio_vsock_workqueue, &vsock->event_work);
}

static void virtio_vsock_tx_done(struct virtqueue *vq)
{
	struct virtio_vsock *vsock = vq->vdev->priv;

	if (!vsock)
		return;
	queue_work(virtio_vsock_workqueue, &vsock->tx_work);
}

static void virtio_vsock_rx_done(struct virtqueue *vq)
{
	struct virtio_vsock *vsock = vq->vdev->priv;

	if (!vsock)
		return;
	queue_work(virtio_vsock_workqueue, &vsock->rx_work);
}

static struct virtio_transport virtio_transport = {
	.transport = {
		.get_local_cid            = virtio_transport_get_local_cid,

		.init                     = virtio_transport_do_socket_init,
		.destruct                 = virtio_transport_destruct,
		.release                  = virtio_transport_release,
		.connect                  = virtio_transport_connect,
		.shutdown                 = virtio_transport_shutdown,

		.dgram_bind               = virtio_transport_dgram_bind,
		.dgram_dequeue            = virtio_transport_dgram_dequeue,
		.dgram_enqueue            = virtio_transport_dgram_enqueue,
		.dgram_allow              = virtio_transport_dgram_allow,

		.stream_dequeue           = virtio_transport_stream_dequeue,
		.stream_enqueue           = virtio_transport_stream_enqueue,
		.stream_has_data          = virtio_transport_stream_has_data,
		.stream_has_space         = virtio_transport_stream_has_space,
		.stream_rcvhiwat          = virtio_transport_stream_rcvhiwat,
		.stream_is_active         = virtio_transport_stream_is_active,
		.stream_allow             = virtio_transport_stream_allow,

		.notify_poll_in           = virtio_transport_notify_poll_in,
		.notify_poll_out          = virtio_transport_notify_poll_out,
		.notify_recv_init         = virtio_transport_notify_recv_init,
		.notify_recv_pre_block    = virtio_transport_notify_recv_pre_block,
		.notify_recv_pre_dequeue  = virtio_transport_notify_recv_pre_dequeue,
		.notify_recv_post_dequeue = virtio_transport_notify_recv_post_dequeue,
		.notify_send_init         = virtio_transport_notify_send_init,
		.notify_send_pre_block    = virtio_transport_notify_send_pre_block,
		.notify_send_pre_enqueue  = virtio_transport_notify_send_pre_enqueue,
		.notify_send_post_enqueue = virtio_transport_notify_send_post_enqueue,

		.set_buffer_size          = virtio_transport_set_buffer_size,
		.set_min_buffer_size      = virtio_transport_set_min_buffer_size,
		.set_max_buffer_size      = virtio_transport_set_max_buffer_size,
		.get_buffer_size          = virtio_transport_get_buffer_size,
		.get_min_buffer_size      = virtio_transport_get_min_buffer_size,
		.get_max_buffer_size      = virtio_transport_get_max_buffer_size,
	},

	.send_pkt = virtio_transport_send_pkt,
};

static int virtio_vsock_probe(struct virtio_device *vdev)
{
	vq_callback_t *callbacks[] = {
		virtio_vsock_rx_done,
		virtio_vsock_tx_done,
		virtio_vsock_event_done,
	};
	static const char * const names[] = {
		"rx",
		"tx",
		"event",
	};
	struct virtio_vsock *vsock = NULL;
	int ret;

	ret = mutex_lock_interruptible(&the_virtio_vsock_mutex);
	if (ret)
		return ret;

	/* Only one virtio-vsock device per guest is supported */
	if (the_virtio_vsock) {
		ret = -EBUSY;
		goto out;
	}

	vsock = kzalloc(sizeof(*vsock), GFP_KERNEL);
	if (!vsock) {
		ret = -ENOMEM;
		goto out;
	}

	vsock->vdev = vdev;

	ret = vsock->vdev->config->find_vqs(vsock->vdev, VSOCK_VQ_MAX,
					    vsock->vqs, callbacks, names);
	if (ret < 0)
		goto out;

	virtio_vsock_update_guest_cid(vsock);

	ret = vsock_core_init(&virtio_transport.transport);
	if (ret < 0)
		goto out_vqs;

	vsock->rx_buf_nr = 0;
	vsock->rx_buf_max_nr = 0;
	atomic_set(&vsock->queued_replies, 0);

	vdev->priv = vsock;
	the_virtio_vsock = vsock;
	mutex_init(&vsock->tx_lock);
	mutex_init(&vsock->rx_lock);
	mutex_init(&vsock->event_lock);
	spin_lock_init(&vsock->send_pkt_list_lock);
	INIT_LIST_HEAD(&vsock->send_pkt_list);
	INIT_WORK(&vsock->rx_work, virtio_transport_rx_work);
	INIT_WORK(&vsock->tx_work, virtio_transport_tx_work);
	INIT_WORK(&vsock->event_work, virtio_transport_event_work);
	INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work);

	mutex_lock(&vsock->rx_lock);
	virtio_vsock_rx_fill(vsock);
	mutex_unlock(&vsock->rx_lock);

	mutex_lock(&vsock->event_lock);
	virtio_vsock_event_fill(vsock);
	mutex_unlock(&vsock->event_lock);

	mutex_unlock(&the_virtio_vsock_mutex);
	return 0;

out_vqs:
	vsock->vdev->config->del_vqs(vsock->vdev);
out:
	kfree(vsock);
	mutex_unlock(&the_virtio_vsock_mutex);
	return ret;
}

static void virtio_vsock_remove(struct virtio_device *vdev)
{
	struct virtio_vsock *vsock = vdev->priv;
	struct virtio_vsock_pkt *pkt;

	flush_work(&vsock->rx_work);
	flush_work(&vsock->tx_work);
	flush_work(&vsock->event_work);
	flush_work(&vsock->send_pkt_work);

	vdev->config->reset(vdev);

	mutex_lock(&vsock->rx_lock);
	while ((pkt = virtqueue_detach_unused_buf(vsock->vqs[VSOCK_VQ_RX])))
		virtio_transport_free_pkt(pkt);
	mutex_unlock(&vsock->rx_lock);

	mutex_lock(&vsock->tx_lock);
	while ((pkt = virtqueue_detach_unused_buf(vsock->vqs[VSOCK_VQ_TX])))
		virtio_transport_free_pkt(pkt);
	mutex_unlock(&vsock->tx_lock);

	spin_lock_bh(&vsock->send_pkt_list_lock);
	while (!list_empty(&vsock->send_pkt_list)) {
		pkt = list_first_entry(&vsock->send_pkt_list,
				       struct virtio_vsock_pkt, list);
		list_del(&pkt->list);
		virtio_transport_free_pkt(pkt);
	}
	spin_unlock_bh(&vsock->send_pkt_list_lock);

	mutex_lock(&the_virtio_vsock_mutex);
	the_virtio_vsock = NULL;
	vsock_core_exit();
	mutex_unlock(&the_virtio_vsock_mutex);

	vdev->config->del_vqs(vdev);

	kfree(vsock);
}

static struct virtio_device_id id_table[] = {
	{ VIRTIO_ID_VSOCK, VIRTIO_DEV_ANY_ID },
	{ 0 },
};

static unsigned int features[] = {
};

static struct virtio_driver virtio_vsock_driver = {
	.feature_table = features,
	.feature_table_size = ARRAY_SIZE(features),
	.driver.name = KBUILD_MODNAME,
	.driver.owner = THIS_MODULE,
	.id_table = id_table,
	.probe = virtio_vsock_probe,
	.remove = virtio_vsock_remove,
};

static int __init virtio_vsock_init(void)
{
	int ret;

	virtio_vsock_workqueue = alloc_workqueue("virtio_vsock", 0, 0);
	if (!virtio_vsock_workqueue)
		return -ENOMEM;
	ret = register_virtio_driver(&virtio_vsock_driver);
	if (ret)
		destroy_workqueue(virtio_vsock_workqueue);
	return ret;
}

static void __exit virtio_vsock_exit(void)
{
	unregister_virtio_driver(&virtio_vsock_driver);
	destroy_workqueue(virtio_vsock_workqueue);
}

module_init(virtio_vsock_init);
module_exit(virtio_vsock_exit);
MODULE_LICENSE("GPL v2");
MODULE_AUTHOR("Asias He");
MODULE_DESCRIPTION("virtio transport for vsock");
MODULE_DEVICE_TABLE(virtio, id_table);
Commit	Line	Data
0ea9e1d3 AH	1	/*
	2	* virtio transport for vsock
	3	*
	4	* Copyright (C) 2013-2015 Red Hat, Inc.
	5	* Author: Asias He <asias@redhat.com>
	6	* Stefan Hajnoczi <stefanha@redhat.com>
	7	*
	8	* Some of the code is take from Gerd Hoffmann <kraxel@redhat.com>'s
	9	* early virtio-vsock proof-of-concept bits.
	10	*
	11	* This work is licensed under the terms of the GNU GPL, version 2.
	12	*/
	13	#include <linux/spinlock.h>
	14	#include <linux/module.h>
	15	#include <linux/list.h>
	16	#include <linux/atomic.h>
	17	#include <linux/virtio.h>
	18	#include <linux/virtio_ids.h>
	19	#include <linux/virtio_config.h>
	20	#include <linux/virtio_vsock.h>
	21	#include <net/sock.h>
	22	#include <linux/mutex.h>
	23	#include <net/af_vsock.h>
	24
	25	static struct workqueue_struct *virtio_vsock_workqueue;
	26	static struct virtio_vsock *the_virtio_vsock;
	27	static DEFINE_MUTEX(the_virtio_vsock_mutex); /* protects the_virtio_vsock */
	28
	29	struct virtio_vsock {
	30	struct virtio_device *vdev;
	31	struct virtqueue *vqs[VSOCK_VQ_MAX];
	32
	33	/* Virtqueue processing is deferred to a workqueue */
	34	struct work_struct tx_work;
	35	struct work_struct rx_work;
	36	struct work_struct event_work;
	37
	38	/* The following fields are protected by tx_lock. vqs[VSOCK_VQ_TX]
	39	* must be accessed with tx_lock held.
	40	*/
	41	struct mutex tx_lock;
	42
	43	struct work_struct send_pkt_work;
	44	spinlock_t send_pkt_list_lock;
	45	struct list_head send_pkt_list;
	46
	47	atomic_t queued_replies;
	48
	49	/* The following fields are protected by rx_lock. vqs[VSOCK_VQ_RX]
	50	* must be accessed with rx_lock held.
	51	*/
	52	struct mutex rx_lock;
	53	int rx_buf_nr;
	54	int rx_buf_max_nr;
	55
	56	/* The following fields are protected by event_lock.
	57	* vqs[VSOCK_VQ_EVENT] must be accessed with event_lock held.
	58	*/
	59	struct mutex event_lock;
	60	struct virtio_vsock_event event_list[8];
	61
	62	u32 guest_cid;
	63	};
	64
65	static struct virtio_vsock *virtio_vsock_get(void)
66	{
67	return the_virtio_vsock;
68	}
69
70	static u32 virtio_transport_get_local_cid(void)
71	{
72	struct virtio_vsock *vsock = virtio_vsock_get();
73
74	return vsock->guest_cid;
75	}
76
77	static void
78	virtio_transport_send_pkt_work(struct work_struct *work)
79	{
80	struct virtio_vsock *vsock =
81	container_of(work, struct virtio_vsock, send_pkt_work);
82	struct virtqueue *vq;
83	bool added = false;
84	bool restart_rx = false;
85
86	mutex_lock(&vsock->tx_lock);
87
88	vq = vsock->vqs[VSOCK_VQ_TX];
89
90	/* Avoid unnecessary interrupts while we're processing the ring */
91	virtqueue_disable_cb(vq);
92
93	for (;;) {
94	struct virtio_vsock_pkt *pkt;
95	struct scatterlist hdr, buf, *sgs[2];
96	int ret, in_sg = 0, out_sg = 0;
97	bool reply;
98
99	spin_lock_bh(&vsock->send_pkt_list_lock);
100	if (list_empty(&vsock->send_pkt_list)) {
101	spin_unlock_bh(&vsock->send_pkt_list_lock);
102	virtqueue_enable_cb(vq);
103	break;
104	}
105
106	pkt = list_first_entry(&vsock->send_pkt_list,
107	struct virtio_vsock_pkt, list);
108	list_del_init(&pkt->list);
109	spin_unlock_bh(&vsock->send_pkt_list_lock);
110
111	reply = pkt->reply;
112
113	sg_init_one(&hdr, &pkt->hdr, sizeof(pkt->hdr));
114	sgs[out_sg++] = &hdr;
115	if (pkt->buf) {
116	sg_init_one(&buf, pkt->buf, pkt->len);
117	sgs[out_sg++] = &buf;
118	}
119
120	ret = virtqueue_add_sgs(vq, sgs, out_sg, in_sg, pkt, GFP_KERNEL);
121	if (ret < 0) {
122	spin_lock_bh(&vsock->send_pkt_list_lock);
123	list_add(&pkt->list, &vsock->send_pkt_list);
124	spin_unlock_bh(&vsock->send_pkt_list_lock);
125
126	if (!virtqueue_enable_cb(vq) && ret == -ENOSPC)
127	continue; /* retry now that we have more space */
128	break;
129	}
130
131	if (reply) {
132	struct virtqueue *rx_vq = vsock->vqs[VSOCK_VQ_RX];
133	int val;
134
135	val = atomic_dec_return(&vsock->queued_replies);
136
137	/* Do we now have resources to resume rx processing? */
138	if (val + 1 == virtqueue_get_vring_size(rx_vq))
139	restart_rx = true;
140	}
141
142	added = true;
143	}
144
145	if (added)
146	virtqueue_kick(vq);
147
148	mutex_unlock(&vsock->tx_lock);
149
150	if (restart_rx)
151	queue_work(virtio_vsock_workqueue, &vsock->rx_work);
152	}
153
154	static int
155	virtio_transport_send_pkt(struct virtio_vsock_pkt *pkt)
156	{
157	struct virtio_vsock *vsock;
158	int len = pkt->len;
159
160	vsock = virtio_vsock_get();
161	if (!vsock) {
162	virtio_transport_free_pkt(pkt);
163	return -ENODEV;
164	}
165
166	if (pkt->reply)
167	atomic_inc(&vsock->queued_replies);
168
169	spin_lock_bh(&vsock->send_pkt_list_lock);
170	list_add_tail(&pkt->list, &vsock->send_pkt_list);
171	spin_unlock_bh(&vsock->send_pkt_list_lock);
172
173	queue_work(virtio_vsock_workqueue, &vsock->send_pkt_work);
174	return len;
175	}
176
177	static void virtio_vsock_rx_fill(struct virtio_vsock *vsock)
178	{
179	int buf_len = VIRTIO_VSOCK_DEFAULT_RX_BUF_SIZE;
180	struct virtio_vsock_pkt *pkt;
181	struct scatterlist hdr, buf, *sgs[2];
182	struct virtqueue *vq;
183	int ret;
184
185	vq = vsock->vqs[VSOCK_VQ_RX];
186
187	do {
188	pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
189	if (!pkt)
190	break;
191
192	pkt->buf = kmalloc(buf_len, GFP_KERNEL);
193	if (!pkt->buf) {
194	virtio_transport_free_pkt(pkt);
195	break;
196	}
197
198	pkt->len = buf_len;
199
200	sg_init_one(&hdr, &pkt->hdr, sizeof(pkt->hdr));
201	sgs[0] = &hdr;
202
203	sg_init_one(&buf, pkt->buf, buf_len);
204	sgs[1] = &buf;
205	ret = virtqueue_add_sgs(vq, sgs, 0, 2, pkt, GFP_KERNEL);
206	if (ret) {
207	virtio_transport_free_pkt(pkt);
208	break;
209	}
210	vsock->rx_buf_nr++;
211	} while (vq->num_free);
212	if (vsock->rx_buf_nr > vsock->rx_buf_max_nr)
213	vsock->rx_buf_max_nr = vsock->rx_buf_nr;
214	virtqueue_kick(vq);
215	}
216
217	static void virtio_transport_tx_work(struct work_struct *work)
218	{
219	struct virtio_vsock *vsock =
220	container_of(work, struct virtio_vsock, tx_work);
221	struct virtqueue *vq;
222	bool added = false;
223
224	vq = vsock->vqs[VSOCK_VQ_TX];
225	mutex_lock(&vsock->tx_lock);
226	do {
227	struct virtio_vsock_pkt *pkt;
228	unsigned int len;
229
230	virtqueue_disable_cb(vq);
231	while ((pkt = virtqueue_get_buf(vq, &len)) != NULL) {
232	virtio_transport_free_pkt(pkt);
233	added = true;
234	}
235	} while (!virtqueue_enable_cb(vq));
236	mutex_unlock(&vsock->tx_lock);
237
238	if (added)
239	queue_work(virtio_vsock_workqueue, &vsock->send_pkt_work);
240	}
241
242	/* Is there space left for replies to rx packets? */
243	static bool virtio_transport_more_replies(struct virtio_vsock *vsock)
244	{
245	struct virtqueue *vq = vsock->vqs[VSOCK_VQ_RX];
246	int val;
247
248	smp_rmb(); /* paired with atomic_inc() and atomic_dec_return() */
249	val = atomic_read(&vsock->queued_replies);
250
251	return val < virtqueue_get_vring_size(vq);
252	}
253
254	static void virtio_transport_rx_work(struct work_struct *work)
255	{
256	struct virtio_vsock *vsock =
257	container_of(work, struct virtio_vsock, rx_work);
258	struct virtqueue *vq;
259
260	vq = vsock->vqs[VSOCK_VQ_RX];
261
262	mutex_lock(&vsock->rx_lock);
263
264	do {
265	virtqueue_disable_cb(vq);
266	for (;;) {
267	struct virtio_vsock_pkt *pkt;
268	unsigned int len;
269
270	if (!virtio_transport_more_replies(vsock)) {
271	/* Stop rx until the device processes already
272	* pending replies. Leave rx virtqueue
273	* callbacks disabled.
274	*/
275	goto out;
276	}
277
278	pkt = virtqueue_get_buf(vq, &len);
279	if (!pkt) {
280	break;
281	}
282
283	vsock->rx_buf_nr--;
284
285	/* Drop short/long packets */
286	if (unlikely(len < sizeof(pkt->hdr) \|\|
287	len > sizeof(pkt->hdr) + pkt->len)) {
288	virtio_transport_free_pkt(pkt);
289	continue;
290	}
291
292	pkt->len = len - sizeof(pkt->hdr);
293	virtio_transport_recv_pkt(pkt);
294	}
295	} while (!virtqueue_enable_cb(vq));
296
297	out:
298	if (vsock->rx_buf_nr < vsock->rx_buf_max_nr / 2)
299	virtio_vsock_rx_fill(vsock);
300	mutex_unlock(&vsock->rx_lock);
301	}
302
303	/* event_lock must be held */
304	static int virtio_vsock_event_fill_one(struct virtio_vsock *vsock,
305	struct virtio_vsock_event *event)
306	{
307	struct scatterlist sg;
308	struct virtqueue *vq;
309
310	vq = vsock->vqs[VSOCK_VQ_EVENT];
311
312	sg_init_one(&sg, event, sizeof(*event));
313
314	return virtqueue_add_inbuf(vq, &sg, 1, event, GFP_KERNEL);
315	}
316
317	/* event_lock must be held */
318	static void virtio_vsock_event_fill(struct virtio_vsock *vsock)
319	{
320	size_t i;
321
322	for (i = 0; i < ARRAY_SIZE(vsock->event_list); i++) {
323	struct virtio_vsock_event *event = &vsock->event_list[i];
324
325	virtio_vsock_event_fill_one(vsock, event);
326	}
327
328	virtqueue_kick(vsock->vqs[VSOCK_VQ_EVENT]);
329	}
330
331	static void virtio_vsock_reset_sock(struct sock *sk)
332	{
333	lock_sock(sk);
334	sk->sk_state = SS_UNCONNECTED;
335	sk->sk_err = ECONNRESET;
336	sk->sk_error_report(sk);
337	release_sock(sk);
338	}
339
340	static void virtio_vsock_update_guest_cid(struct virtio_vsock *vsock)
341	{
342	struct virtio_device *vdev = vsock->vdev;
343	u64 guest_cid;
344
345	vdev->config->get(vdev, offsetof(struct virtio_vsock_config, guest_cid),
346	&guest_cid, sizeof(guest_cid));
347	vsock->guest_cid = le64_to_cpu(guest_cid);
348	}
349
350	/* event_lock must be held */
351	static void virtio_vsock_event_handle(struct virtio_vsock *vsock,
352	struct virtio_vsock_event *event)
353	{
354	switch (le32_to_cpu(event->id)) {
355	case VIRTIO_VSOCK_EVENT_TRANSPORT_RESET:
356	virtio_vsock_update_guest_cid(vsock);
357	vsock_for_each_connected_socket(virtio_vsock_reset_sock);
358	break;
359	}
360	}
361
362	static void virtio_transport_event_work(struct work_struct *work)
363	{
364	struct virtio_vsock *vsock =
365	container_of(work, struct virtio_vsock, event_work);
366	struct virtqueue *vq;
367
368	vq = vsock->vqs[VSOCK_VQ_EVENT];
369
370	mutex_lock(&vsock->event_lock);
371
372	do {
373	struct virtio_vsock_event *event;
374	unsigned int len;
375
376	virtqueue_disable_cb(vq);
377	while ((event = virtqueue_get_buf(vq, &len)) != NULL) {
378	if (len == sizeof(*event))
379	virtio_vsock_event_handle(vsock, event);
380
381	virtio_vsock_event_fill_one(vsock, event);
382	}
383	} while (!virtqueue_enable_cb(vq));
384
385	virtqueue_kick(vsock->vqs[VSOCK_VQ_EVENT]);
386
387	mutex_unlock(&vsock->event_lock);
388	}
389
390	static void virtio_vsock_event_done(struct virtqueue *vq)
391	{
392	struct virtio_vsock *vsock = vq->vdev->priv;
393
394	if (!vsock)
395	return;
396	queue_work(virtio_vsock_workqueue, &vsock->event_work);
397	}
398
399	static void virtio_vsock_tx_done(struct virtqueue *vq)
400	{
401	struct virtio_vsock *vsock = vq->vdev->priv;
402
403	if (!vsock)
404	return;
405	queue_work(virtio_vsock_workqueue, &vsock->tx_work);
406	}
407
408	static void virtio_vsock_rx_done(struct virtqueue *vq)
409	{
410	struct virtio_vsock *vsock = vq->vdev->priv;
411
412	if (!vsock)
413	return;
414	queue_work(virtio_vsock_workqueue, &vsock->rx_work);
415	}
416
417	static struct virtio_transport virtio_transport = {
418	.transport = {
419	.get_local_cid = virtio_transport_get_local_cid,
420
421	.init = virtio_transport_do_socket_init,
422	.destruct = virtio_transport_destruct,
423	.release = virtio_transport_release,
424	.connect = virtio_transport_connect,
425	.shutdown = virtio_transport_shutdown,
426
427	.dgram_bind = virtio_transport_dgram_bind,
428	.dgram_dequeue = virtio_transport_dgram_dequeue,
429	.dgram_enqueue = virtio_transport_dgram_enqueue,
430	.dgram_allow = virtio_transport_dgram_allow,
431
432	.stream_dequeue = virtio_transport_stream_dequeue,
433	.stream_enqueue = virtio_transport_stream_enqueue,
434	.stream_has_data = virtio_transport_stream_has_data,
435	.stream_has_space = virtio_transport_stream_has_space,
436	.stream_rcvhiwat = virtio_transport_stream_rcvhiwat,
437	.stream_is_active = virtio_transport_stream_is_active,
438	.stream_allow = virtio_transport_stream_allow,
439
440	.notify_poll_in = virtio_transport_notify_poll_in,
441	.notify_poll_out = virtio_transport_notify_poll_out,
442	.notify_recv_init = virtio_transport_notify_recv_init,
443	.notify_recv_pre_block = virtio_transport_notify_recv_pre_block,
444	.notify_recv_pre_dequeue = virtio_transport_notify_recv_pre_dequeue,
445	.notify_recv_post_dequeue = virtio_transport_notify_recv_post_dequeue,
446	.notify_send_init = virtio_transport_notify_send_init,
447	.notify_send_pre_block = virtio_transport_notify_send_pre_block,
448	.notify_send_pre_enqueue = virtio_transport_notify_send_pre_enqueue,
449	.notify_send_post_enqueue = virtio_transport_notify_send_post_enqueue,
450
451	.set_buffer_size = virtio_transport_set_buffer_size,
452	.set_min_buffer_size = virtio_transport_set_min_buffer_size,
453	.set_max_buffer_size = virtio_transport_set_max_buffer_size,
454	.get_buffer_size = virtio_transport_get_buffer_size,
455	.get_min_buffer_size = virtio_transport_get_min_buffer_size,
456	.get_max_buffer_size = virtio_transport_get_max_buffer_size,
457	},
458
459	.send_pkt = virtio_transport_send_pkt,
460	};
461
462	static int virtio_vsock_probe(struct virtio_device *vdev)
463	{
464	vq_callback_t *callbacks[] = {
465	virtio_vsock_rx_done,
466	virtio_vsock_tx_done,
467	virtio_vsock_event_done,
468	};
469	static const char * const names[] = {
470	"rx",
471	"tx",
472	"event",
473	};
474	struct virtio_vsock *vsock = NULL;
475	int ret;
476
477	ret = mutex_lock_interruptible(&the_virtio_vsock_mutex);
478	if (ret)
479	return ret;
480
481	/* Only one virtio-vsock device per guest is supported */
482	if (the_virtio_vsock) {
483	ret = -EBUSY;
484	goto out;
485	}
486
487	vsock = kzalloc(sizeof(*vsock), GFP_KERNEL);
488	if (!vsock) {
489	ret = -ENOMEM;
490	goto out;
491	}
492
493	vsock->vdev = vdev;
494
495	ret = vsock->vdev->config->find_vqs(vsock->vdev, VSOCK_VQ_MAX,
496	vsock->vqs, callbacks, names);
497	if (ret < 0)
498	goto out;
499
500	virtio_vsock_update_guest_cid(vsock);
501
502	ret = vsock_core_init(&virtio_transport.transport);
503	if (ret < 0)
504	goto out_vqs;
505
506	vsock->rx_buf_nr = 0;
507	vsock->rx_buf_max_nr = 0;
508	atomic_set(&vsock->queued_replies, 0);
509
510	vdev->priv = vsock;
511	the_virtio_vsock = vsock;
512	mutex_init(&vsock->tx_lock);
513	mutex_init(&vsock->rx_lock);
514	mutex_init(&vsock->event_lock);
515	spin_lock_init(&vsock->send_pkt_list_lock);
516	INIT_LIST_HEAD(&vsock->send_pkt_list);
517	INIT_WORK(&vsock->rx_work, virtio_transport_rx_work);
518	INIT_WORK(&vsock->tx_work, virtio_transport_tx_work);
519	INIT_WORK(&vsock->event_work, virtio_transport_event_work);
520	INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work);
521
522	mutex_lock(&vsock->rx_lock);
523	virtio_vsock_rx_fill(vsock);
524	mutex_unlock(&vsock->rx_lock);
525
526	mutex_lock(&vsock->event_lock);
527	virtio_vsock_event_fill(vsock);
528	mutex_unlock(&vsock->event_lock);
529
530	mutex_unlock(&the_virtio_vsock_mutex);
531	return 0;
532
533	out_vqs:
534	vsock->vdev->config->del_vqs(vsock->vdev);
535	out:
536	kfree(vsock);
537	mutex_unlock(&the_virtio_vsock_mutex);
538	return ret;
539	}
540
541	static void virtio_vsock_remove(struct virtio_device *vdev)
542	{
543	struct virtio_vsock *vsock = vdev->priv;
544	struct virtio_vsock_pkt *pkt;
545
546	flush_work(&vsock->rx_work);
547	flush_work(&vsock->tx_work);
548	flush_work(&vsock->event_work);
549	flush_work(&vsock->send_pkt_work);
550
551	vdev->config->reset(vdev);
552
553	mutex_lock(&vsock->rx_lock);
554	while ((pkt = virtqueue_detach_unused_buf(vsock->vqs[VSOCK_VQ_RX])))
555	virtio_transport_free_pkt(pkt);
556	mutex_unlock(&vsock->rx_lock);
557
558	mutex_lock(&vsock->tx_lock);
559	while ((pkt = virtqueue_detach_unused_buf(vsock->vqs[VSOCK_VQ_TX])))
560	virtio_transport_free_pkt(pkt);
561	mutex_unlock(&vsock->tx_lock);
562
563	spin_lock_bh(&vsock->send_pkt_list_lock);
564	while (!list_empty(&vsock->send_pkt_list)) {
565	pkt = list_first_entry(&vsock->send_pkt_list,
566	struct virtio_vsock_pkt, list);
567	list_del(&pkt->list);
568	virtio_transport_free_pkt(pkt);
569	}
570	spin_unlock_bh(&vsock->send_pkt_list_lock);
571
572	mutex_lock(&the_virtio_vsock_mutex);
573	the_virtio_vsock = NULL;
574	vsock_core_exit();
575	mutex_unlock(&the_virtio_vsock_mutex);
576
577	vdev->config->del_vqs(vdev);
578
579	kfree(vsock);
580	}
581
582	static struct virtio_device_id id_table[] = {
583	{ VIRTIO_ID_VSOCK, VIRTIO_DEV_ANY_ID },
584	{ 0 },
585	};
586
587	static unsigned int features[] = {
588	};
589
590	static struct virtio_driver virtio_vsock_driver = {
591	.feature_table = features,
592	.feature_table_size = ARRAY_SIZE(features),
593	.driver.name = KBUILD_MODNAME,
594	.driver.owner = THIS_MODULE,
595	.id_table = id_table,
596	.probe = virtio_vsock_probe,
597	.remove = virtio_vsock_remove,
598	};
599
600	static int __init virtio_vsock_init(void)
601	{
602	int ret;
603
604	virtio_vsock_workqueue = alloc_workqueue("virtio_vsock", 0, 0);
605	if (!virtio_vsock_workqueue)
606	return -ENOMEM;
607	ret = register_virtio_driver(&virtio_vsock_driver);
608	if (ret)
609	destroy_workqueue(virtio_vsock_workqueue);
610	return ret;
611	}
612
613	static void __exit virtio_vsock_exit(void)
614	{
615	unregister_virtio_driver(&virtio_vsock_driver);
616	destroy_workqueue(virtio_vsock_workqueue);
617	}
618
619	module_init(virtio_vsock_init);
620	module_exit(virtio_vsock_exit);
621	MODULE_LICENSE("GPL v2");
622	MODULE_AUTHOR("Asias He");
623	MODULE_DESCRIPTION("virtio transport for vsock");
624	MODULE_DEVICE_TABLE(virtio, id_table);