[linux-2.6-block.git] / drivers / staging / hv / netvsc_drv.c

/*
 * Copyright (c) 2009, Microsoft Corporation.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
 * Place - Suite 330, Boston, MA 02111-1307 USA.
 *
 * Authors:
 *   Haiyang Zhang <haiyangz@microsoft.com>
 *   Hank Janssen  <hjanssen@microsoft.com>
 */
#include <linux/init.h>
#include <linux/module.h>
#include <linux/highmem.h>
#include <linux/device.h>
#include <linux/io.h>
#include <linux/delay.h>
#include <linux/netdevice.h>
#include <linux/inetdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/in.h>
#include <linux/slab.h>
#include <net/arp.h>
#include <net/route.h>
#include <net/sock.h>
#include <net/pkt_sched.h>
#include "osd.h"
#include "logging.h"
#include "VersionInfo.h"
#include "vmbus.h"
#include "NetVscApi.h"

struct net_device_context {
	/* point back to our device context */
	struct vm_device *device_ctx;
};

struct netvsc_driver_context {
	/* !! These must be the first 2 fields !! */
	/* Which is a bug FIXME! */
	struct driver_context drv_ctx;
	struct netvsc_driver drv_obj;
};

static int netvsc_ringbuffer_size = NETVSC_DEVICE_RING_BUFFER_SIZE;

/* The one and only one */
static struct netvsc_driver_context g_netvsc_drv;

static void netvsc_set_multicast_list(struct net_device *net)
{
}

static int netvsc_open(struct net_device *net)
{
	struct net_device_context *net_device_ctx = netdev_priv(net);
	struct hv_device *device_obj = &net_device_ctx->device_ctx->device_obj;
	int ret = 0;

	DPRINT_ENTER(NETVSC_DRV);

	if (netif_carrier_ok(net)) {
		/* Open up the device */
		ret = RndisFilterOnOpen(device_obj);
		if (ret != 0) {
			DPRINT_ERR(NETVSC_DRV,
				   "unable to open device (ret %d).", ret);
			return ret;
		}

		netif_start_queue(net);
	} else {
		DPRINT_ERR(NETVSC_DRV, "unable to open device...link is down.");
	}

	DPRINT_EXIT(NETVSC_DRV);
	return ret;
}

static int netvsc_close(struct net_device *net)
{
	struct net_device_context *net_device_ctx = netdev_priv(net);
	struct hv_device *device_obj = &net_device_ctx->device_ctx->device_obj;
	int ret;

	DPRINT_ENTER(NETVSC_DRV);

	netif_stop_queue(net);

	ret = RndisFilterOnClose(device_obj);
	if (ret != 0)
		DPRINT_ERR(NETVSC_DRV, "unable to close device (ret %d).", ret);

	DPRINT_EXIT(NETVSC_DRV);

	return ret;
}

static void netvsc_xmit_completion(void *context)
{
	struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
	struct sk_buff *skb = (struct sk_buff *)
		(unsigned long)packet->Completion.Send.SendCompletionTid;

	DPRINT_ENTER(NETVSC_DRV);

	kfree(packet);

	if (skb) {
		struct net_device *net = skb->dev;
		dev_kfree_skb_any(skb);

		if (netif_queue_stopped(net)) {
			DPRINT_INFO(NETVSC_DRV, "net device (%p) waking up...",
				    net);

			netif_wake_queue(net);
		}
	}

	DPRINT_EXIT(NETVSC_DRV);
}

static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
{
	struct net_device_context *net_device_ctx = netdev_priv(net);
	struct driver_context *driver_ctx =
	    driver_to_driver_context(net_device_ctx->device_ctx->device.driver);
	struct netvsc_driver_context *net_drv_ctx =
		(struct netvsc_driver_context *)driver_ctx;
	struct netvsc_driver *net_drv_obj = &net_drv_ctx->drv_obj;
	struct hv_netvsc_packet *packet;
	int i;
	int ret;
	int num_frags;
	int retries = 0;

	DPRINT_ENTER(NETVSC_DRV);

	/* Support only 1 chain of frags */
	ASSERT(skb_shinfo(skb)->frag_list == NULL);
	ASSERT(skb->dev == net);

	DPRINT_DBG(NETVSC_DRV, "xmit packet - len %d data_len %d",
		   skb->len, skb->data_len);

	/* Add 1 for skb->data and any additional ones requested */
	num_frags = skb_shinfo(skb)->nr_frags + 1 +
		    net_drv_obj->AdditionalRequestPageBufferCount;

	/* Allocate a netvsc packet based on # of frags. */
	packet = kzalloc(sizeof(struct hv_netvsc_packet) +
			 (num_frags * sizeof(struct hv_page_buffer)) +
			 net_drv_obj->RequestExtSize, GFP_ATOMIC);
	if (!packet) {
		DPRINT_ERR(NETVSC_DRV, "unable to allocate hv_netvsc_packet");
		return -1;
	}

	packet->Extension = (void *)(unsigned long)packet +
				sizeof(struct hv_netvsc_packet) +
				    (num_frags * sizeof(struct hv_page_buffer));

	/* Setup the rndis header */
	packet->PageBufferCount = num_frags;

	/* TODO: Flush all write buffers/ memory fence ??? */
	/* wmb(); */

	/* Initialize it from the skb */
	ASSERT(skb->data);
	packet->TotalDataBufferLength	= skb->len;

	/*
	 * Start filling in the page buffers starting at
	 * AdditionalRequestPageBufferCount offset
	 */
	packet->PageBuffers[net_drv_obj->AdditionalRequestPageBufferCount].Pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
	packet->PageBuffers[net_drv_obj->AdditionalRequestPageBufferCount].Offset = (unsigned long)skb->data & (PAGE_SIZE - 1);
	packet->PageBuffers[net_drv_obj->AdditionalRequestPageBufferCount].Length = skb->len - skb->data_len;

	ASSERT((skb->len - skb->data_len) <= PAGE_SIZE);

	for (i = net_drv_obj->AdditionalRequestPageBufferCount + 1;
	     i < num_frags; i++) {
		packet->PageBuffers[i].Pfn =
			page_to_pfn(skb_shinfo(skb)->frags[i-(net_drv_obj->AdditionalRequestPageBufferCount+1)].page);
		packet->PageBuffers[i].Offset =
			skb_shinfo(skb)->frags[i-(net_drv_obj->AdditionalRequestPageBufferCount+1)].page_offset;
		packet->PageBuffers[i].Length =
			skb_shinfo(skb)->frags[i-(net_drv_obj->AdditionalRequestPageBufferCount+1)].size;
	}

	/* Set the completion routine */
	packet->Completion.Send.OnSendCompletion = netvsc_xmit_completion;
	packet->Completion.Send.SendCompletionContext = packet;
	packet->Completion.Send.SendCompletionTid = (unsigned long)skb;

retry_send:
	ret = net_drv_obj->OnSend(&net_device_ctx->device_ctx->device_obj,
				  packet);

	if (ret == 0) {
		ret = NETDEV_TX_OK;
		net->stats.tx_bytes += skb->len;
		net->stats.tx_packets++;
	} else {
		retries++;
		if (retries < 4) {
			DPRINT_ERR(NETVSC_DRV, "unable to send..."
					"retrying %d...", retries);
			udelay(100);
			goto retry_send;
		}

		/* no more room or we are shutting down */
		DPRINT_ERR(NETVSC_DRV, "unable to send (%d)..."
			   "marking net device (%p) busy", ret, net);
		DPRINT_INFO(NETVSC_DRV, "net device (%p) stopping", net);

		ret = NETDEV_TX_BUSY;
		net->stats.tx_dropped++;

		netif_stop_queue(net);

		/*
		 * Null it since the caller will free it instead of the
		 * completion routine
		 */
		packet->Completion.Send.SendCompletionTid = 0;

		/*
		 * Release the resources since we will not get any send
		 * completion
		 */
		netvsc_xmit_completion((void *)packet);
	}

	DPRINT_DBG(NETVSC_DRV, "# of xmits %lu total size %lu",
		   net->stats.tx_packets,
		   net->stats.tx_bytes);

	DPRINT_EXIT(NETVSC_DRV);
	return ret;
}

/*
 * netvsc_linkstatus_callback - Link up/down notification
 */
static void netvsc_linkstatus_callback(struct hv_device *device_obj,
				       unsigned int status)
{
	struct vm_device *device_ctx = to_vm_device(device_obj);
	struct net_device *net = dev_get_drvdata(&device_ctx->device);

	DPRINT_ENTER(NETVSC_DRV);

	if (!net) {
		DPRINT_ERR(NETVSC_DRV, "got link status but net device "
				"not initialized yet");
		return;
	}

	if (status == 1) {
		netif_carrier_on(net);
		netif_wake_queue(net);
	} else {
		netif_carrier_off(net);
		netif_stop_queue(net);
	}
	DPRINT_EXIT(NETVSC_DRV);
}

/*
 * netvsc_recv_callback -  Callback when we receive a packet from the
 * "wire" on the specified device.
 */
static int netvsc_recv_callback(struct hv_device *device_obj,
				struct hv_netvsc_packet *packet)
{
	struct vm_device *device_ctx = to_vm_device(device_obj);
	struct net_device *net = dev_get_drvdata(&device_ctx->device);
	struct sk_buff *skb;
	void *data;
	int i;
	unsigned long flags;

	DPRINT_ENTER(NETVSC_DRV);

	if (!net) {
		DPRINT_ERR(NETVSC_DRV, "got receive callback but net device "
				"not initialized yet");
		return 0;
	}

	/* Allocate a skb - TODO direct I/O to pages? */
	skb = netdev_alloc_skb_ip_align(net, packet->TotalDataBufferLength);
	if (unlikely(!skb)) {
		++net->stats.rx_dropped;
		return 0;
	}

	/* for kmap_atomic */
	local_irq_save(flags);

	/*
	 * Copy to skb. This copy is needed here since the memory pointed by
	 * hv_netvsc_packet cannot be deallocated
	 */
	for (i = 0; i < packet->PageBufferCount; i++) {
		data = kmap_atomic(pfn_to_page(packet->PageBuffers[i].Pfn),
					       KM_IRQ1);
		data = (void *)(unsigned long)data +
				packet->PageBuffers[i].Offset;

		memcpy(skb_put(skb, packet->PageBuffers[i].Length), data,
		       packet->PageBuffers[i].Length);

		kunmap_atomic((void *)((unsigned long)data -
				       packet->PageBuffers[i].Offset), KM_IRQ1);
	}

	local_irq_restore(flags);

	skb->protocol = eth_type_trans(skb, net);
	skb->ip_summed = CHECKSUM_NONE;

	net->stats.rx_packets++;
	net->stats.rx_bytes += skb->len;

	/*
	 * Pass the skb back up. Network stack will deallocate the skb when it
	 * is done.
	 * TODO - use NAPI?
	 */
	netif_rx(skb);

	DPRINT_DBG(NETVSC_DRV, "# of recvs %lu total size %lu",
		   net->stats.rx_packets, net->stats.rx_bytes);

	DPRINT_EXIT(NETVSC_DRV);

	return 0;
}

static const struct net_device_ops device_ops = {
	.ndo_open =			netvsc_open,
	.ndo_stop =			netvsc_close,
	.ndo_start_xmit =		netvsc_start_xmit,
	.ndo_set_multicast_list =	netvsc_set_multicast_list,
};

static int netvsc_probe(struct device *device)
{
	struct driver_context *driver_ctx =
		driver_to_driver_context(device->driver);
	struct netvsc_driver_context *net_drv_ctx =
		(struct netvsc_driver_context *)driver_ctx;
	struct netvsc_driver *net_drv_obj = &net_drv_ctx->drv_obj;
	struct vm_device *device_ctx = device_to_vm_device(device);
	struct hv_device *device_obj = &device_ctx->device_obj;
	struct net_device *net = NULL;
	struct net_device_context *net_device_ctx;
	struct netvsc_device_info device_info;
	int ret;

	DPRINT_ENTER(NETVSC_DRV);

	if (!net_drv_obj->Base.OnDeviceAdd)
		return -1;

	net = alloc_netdev(sizeof(struct net_device_context), "seth%d",
			   ether_setup);
	if (!net)
		return -1;

	/* Set initial state */
	netif_carrier_off(net);
	netif_stop_queue(net);

	net_device_ctx = netdev_priv(net);
	net_device_ctx->device_ctx = device_ctx;
	dev_set_drvdata(device, net);

	/* Notify the netvsc driver of the new device */
	ret = net_drv_obj->Base.OnDeviceAdd(device_obj, &device_info);
	if (ret != 0) {
		free_netdev(net);
		dev_set_drvdata(device, NULL);

		DPRINT_ERR(NETVSC_DRV, "unable to add netvsc device (ret %d)",
			   ret);
		return ret;
	}

	/*
	 * If carrier is still off ie we did not get a link status callback,
	 * update it if necessary
	 */
	/*
	 * FIXME: We should use a atomic or test/set instead to avoid getting
	 * out of sync with the device's link status
	 */
	if (!netif_carrier_ok(net))
		if (!device_info.LinkState)
			netif_carrier_on(net);

	memcpy(net->dev_addr, device_info.MacAddr, ETH_ALEN);

	net->netdev_ops = &device_ops;

	SET_NETDEV_DEV(net, device);

	ret = register_netdev(net);
	if (ret != 0) {
		/* Remove the device and release the resource */
		net_drv_obj->Base.OnDeviceRemove(device_obj);
		free_netdev(net);
	}

	DPRINT_EXIT(NETVSC_DRV);
	return ret;
}

static int netvsc_remove(struct device *device)
{
	struct driver_context *driver_ctx =
		driver_to_driver_context(device->driver);
	struct netvsc_driver_context *net_drv_ctx =
		(struct netvsc_driver_context *)driver_ctx;
	struct netvsc_driver *net_drv_obj = &net_drv_ctx->drv_obj;
	struct vm_device *device_ctx = device_to_vm_device(device);
	struct net_device *net = dev_get_drvdata(&device_ctx->device);
	struct hv_device *device_obj = &device_ctx->device_obj;
	int ret;

	DPRINT_ENTER(NETVSC_DRV);

	if (net == NULL) {
		DPRINT_INFO(NETVSC, "no net device to remove");
		DPRINT_EXIT(NETVSC_DRV);
		return 0;
	}

	if (!net_drv_obj->Base.OnDeviceRemove) {
		DPRINT_EXIT(NETVSC_DRV);
		return -1;
	}

	/* Stop outbound asap */
	netif_stop_queue(net);
	/* netif_carrier_off(net); */

	unregister_netdev(net);

	/*
	 * Call to the vsc driver to let it know that the device is being
	 * removed
	 */
	ret = net_drv_obj->Base.OnDeviceRemove(device_obj);
	if (ret != 0) {
		/* TODO: */
		DPRINT_ERR(NETVSC, "unable to remove vsc device (ret %d)", ret);
	}

	free_netdev(net);
	DPRINT_EXIT(NETVSC_DRV);
	return ret;
}

static int netvsc_drv_exit_cb(struct device *dev, void *data)
{
	struct device **curr = (struct device **)data;

	*curr = dev;
	/* stop iterating */
	return 1;
}

static void netvsc_drv_exit(void)
{
	struct netvsc_driver *netvsc_drv_obj = &g_netvsc_drv.drv_obj;
	struct driver_context *drv_ctx = &g_netvsc_drv.drv_ctx;
	struct device *current_dev;
	int ret;

	DPRINT_ENTER(NETVSC_DRV);

	while (1) {
		current_dev = NULL;

		/* Get the device */
		ret = driver_for_each_device(&drv_ctx->driver, NULL,
					     &current_dev, netvsc_drv_exit_cb);
		if (ret)
			DPRINT_WARN(NETVSC_DRV,
				    "driver_for_each_device returned %d", ret);

		if (current_dev == NULL)
			break;

		/* Initiate removal from the top-down */
		DPRINT_INFO(NETVSC_DRV, "unregistering device (%p)...",
			    current_dev);

		device_unregister(current_dev);
	}

	if (netvsc_drv_obj->Base.OnCleanup)
		netvsc_drv_obj->Base.OnCleanup(&netvsc_drv_obj->Base);

	vmbus_child_driver_unregister(drv_ctx);

	DPRINT_EXIT(NETVSC_DRV);

	return;
}

static int netvsc_drv_init(int (*drv_init)(struct hv_driver *drv))
{
	struct netvsc_driver *net_drv_obj = &g_netvsc_drv.drv_obj;
	struct driver_context *drv_ctx = &g_netvsc_drv.drv_ctx;
	int ret;

	DPRINT_ENTER(NETVSC_DRV);

	vmbus_get_interface(&net_drv_obj->Base.VmbusChannelInterface);

	net_drv_obj->RingBufferSize = netvsc_ringbuffer_size;
	net_drv_obj->OnReceiveCallback = netvsc_recv_callback;
	net_drv_obj->OnLinkStatusChanged = netvsc_linkstatus_callback;

	/* Callback to client driver to complete the initialization */
	drv_init(&net_drv_obj->Base);

	drv_ctx->driver.name = net_drv_obj->Base.name;
	memcpy(&drv_ctx->class_id, &net_drv_obj->Base.deviceType,
	       sizeof(struct hv_guid));

	drv_ctx->probe = netvsc_probe;
	drv_ctx->remove = netvsc_remove;

	/* The driver belongs to vmbus */
	ret = vmbus_child_driver_register(drv_ctx);

	DPRINT_EXIT(NETVSC_DRV);

	return ret;
}

static int __init netvsc_init(void)
{
	int ret;

	DPRINT_ENTER(NETVSC_DRV);
	DPRINT_INFO(NETVSC_DRV, "Netvsc initializing....");

	ret = netvsc_drv_init(NetVscInitialize);

	DPRINT_EXIT(NETVSC_DRV);

	return ret;
}

static void __exit netvsc_exit(void)
{
	DPRINT_ENTER(NETVSC_DRV);
	netvsc_drv_exit();
	DPRINT_EXIT(NETVSC_DRV);
}

MODULE_LICENSE("GPL");
MODULE_VERSION(HV_DRV_VERSION);
MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
module_param(netvsc_ringbuffer_size, int, S_IRUGO);

module_init(netvsc_init);
module_exit(netvsc_exit);
Commit	Line	Data
fceaf24a	1	/*
fceaf24a HJ	2	* Copyright (c) 2009, Microsoft Corporation.
	3	*
	4	* This program is free software; you can redistribute it and/or modify it
	5	* under the terms and conditions of the GNU General Public License,
	6	* version 2, as published by the Free Software Foundation.
	7	*
	8	* This program is distributed in the hope it will be useful, but WITHOUT
	9	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	10	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	11	* more details.
	12	*
	13	* You should have received a copy of the GNU General Public License along with
	14	* this program; if not, write to the Free Software Foundation, Inc., 59 Temple
	15	* Place - Suite 330, Boston, MA 02111-1307 USA.
	16	*
	17	* Authors:
d0e94d17	18	* Haiyang Zhang <haiyangz@microsoft.com>
fceaf24a	19	* Hank Janssen <hjanssen@microsoft.com>
fceaf24a	20	*/
fceaf24a HJ	21	#include <linux/init.h>
	22	#include <linux/module.h>
	23	#include <linux/highmem.h>
	24	#include <linux/device.h>
fceaf24a	25	#include <linux/io.h>
fceaf24a HJ	26	#include <linux/delay.h>
	27	#include <linux/netdevice.h>
	28	#include <linux/inetdevice.h>
	29	#include <linux/etherdevice.h>
	30	#include <linux/skbuff.h>
	31	#include <linux/in.h>
5a0e3ad6	32	#include <linux/slab.h>
fceaf24a HJ	33	#include <net/arp.h>
	34	#include <net/route.h>
	35	#include <net/sock.h>
	36	#include <net/pkt_sched.h>
4983b39a	37	#include "osd.h"
645954c5	38	#include "logging.h"
26c14cc1	39	#include "VersionInfo.h"
870cde80	40	#include "vmbus.h"
cc211812	41	#include "NetVscApi.h"
fceaf24a	42
fceaf24a	43	struct net_device_context {
02fafbc6	44	/* point back to our device context */
f916a34d	45	struct vm_device *device_ctx;
fceaf24a HJ	46	};
	47
	48	struct netvsc_driver_context {
454f18a9	49	/* !! These must be the first 2 fields !! */
7e23a6e9	50	/* Which is a bug FIXME! */
02fafbc6	51	struct driver_context drv_ctx;
7e23a6e9	52	struct netvsc_driver drv_obj;
fceaf24a HJ	53	};
fceaf24a HJ	54
fceaf24a HJ	55	static int netvsc_ringbuffer_size = NETVSC_DEVICE_RING_BUFFER_SIZE;
fceaf24a HJ	56
454f18a9	57	/* The one and only one */
fceaf24a HJ	58	static struct netvsc_driver_context g_netvsc_drv;
fceaf24a HJ	59
4e9bfefa	60	static void netvsc_set_multicast_list(struct net_device *net)
fceaf24a HJ	61	{
	62	}
	63
fceaf24a HJ	64	static int netvsc_open(struct net_device *net)
fceaf24a HJ	65	{
fceaf24a	66	struct net_device_context *net_device_ctx = netdev_priv(net);
3d3b5518	67	struct hv_device *device_obj = &net_device_ctx->device_ctx->device_obj;
02fafbc6	68	int ret = 0;
fceaf24a HJ	69
	70	DPRINT_ENTER(NETVSC_DRV);
	71
02fafbc6	72	if (netif_carrier_ok(net)) {
454f18a9	73	/* Open up the device */
2d075346	74	ret = RndisFilterOnOpen(device_obj);
02fafbc6 GKH	75	if (ret != 0) {
	76	DPRINT_ERR(NETVSC_DRV,
	77	"unable to open device (ret %d).", ret);
fceaf24a HJ	78	return ret;
	79	}
	80
	81	netif_start_queue(net);
02fafbc6	82	} else {
fceaf24a HJ	83	DPRINT_ERR(NETVSC_DRV, "unable to open device...link is down.");
	84	}
	85
	86	DPRINT_EXIT(NETVSC_DRV);
	87	return ret;
	88	}
	89
fceaf24a HJ	90	static int netvsc_close(struct net_device *net)
fceaf24a HJ	91	{
fceaf24a	92	struct net_device_context *net_device_ctx = netdev_priv(net);
3d3b5518	93	struct hv_device *device_obj = &net_device_ctx->device_ctx->device_obj;
02fafbc6	94	int ret;
fceaf24a HJ	95
	96	DPRINT_ENTER(NETVSC_DRV);
	97
	98	netif_stop_queue(net);
	99
4f28900b	100	ret = RndisFilterOnClose(device_obj);
fceaf24a	101	if (ret != 0)
fceaf24a	102	DPRINT_ERR(NETVSC_DRV, "unable to close device (ret %d).", ret);
fceaf24a HJ	103
	104	DPRINT_EXIT(NETVSC_DRV);
	105
	106	return ret;
	107	}
	108
fceaf24a HJ	109	static void netvsc_xmit_completion(void *context)
fceaf24a HJ	110	{
4193d4f4	111	struct hv_netvsc_packet packet = (struct hv_netvsc_packet )context;
02fafbc6 GKH	112	struct sk_buff skb = (struct sk_buff )
02fafbc6 GKH	113	(unsigned long)packet->Completion.Send.SendCompletionTid;
fceaf24a HJ	114
	115	DPRINT_ENTER(NETVSC_DRV);
	116
	117	kfree(packet);
	118
02fafbc6	119	if (skb) {
7880fc54	120	struct net_device *net = skb->dev;
fceaf24a HJ	121	dev_kfree_skb_any(skb);
fceaf24a HJ	122
02fafbc6 GKH	123	if (netif_queue_stopped(net)) {
	124	DPRINT_INFO(NETVSC_DRV, "net device (%p) waking up...",
	125	net);
fceaf24a HJ	126
	127	netif_wake_queue(net);
	128	}
	129	}
	130
	131	DPRINT_EXIT(NETVSC_DRV);
	132	}
	133
02fafbc6	134	static int netvsc_start_xmit(struct sk_buff skb, struct net_device net)
fceaf24a	135	{
fceaf24a	136	struct net_device_context *net_device_ctx = netdev_priv(net);
02fafbc6 GKH	137	struct driver_context *driver_ctx =
	138	driver_to_driver_context(net_device_ctx->device_ctx->device.driver);
	139	struct netvsc_driver_context *net_drv_ctx =
	140	(struct netvsc_driver_context *)driver_ctx;
7e23a6e9	141	struct netvsc_driver *net_drv_obj = &net_drv_ctx->drv_obj;
4193d4f4	142	struct hv_netvsc_packet *packet;
02fafbc6 GKH	143	int i;
02fafbc6 GKH	144	int ret;
fceaf24a	145	int num_frags;
02fafbc6	146	int retries = 0;
fceaf24a HJ	147
	148	DPRINT_ENTER(NETVSC_DRV);
	149
454f18a9	150	/* Support only 1 chain of frags */
fceaf24a HJ	151	ASSERT(skb_shinfo(skb)->frag_list == NULL);
	152	ASSERT(skb->dev == net);
	153
02fafbc6 GKH	154	DPRINT_DBG(NETVSC_DRV, "xmit packet - len %d data_len %d",
02fafbc6 GKH	155	skb->len, skb->data_len);
fceaf24a	156
454f18a9	157	/* Add 1 for skb->data and any additional ones requested */
02fafbc6 GKH	158	num_frags = skb_shinfo(skb)->nr_frags + 1 +
02fafbc6 GKH	159	net_drv_obj->AdditionalRequestPageBufferCount;
fceaf24a	160
454f18a9	161	/* Allocate a netvsc packet based on # of frags. */
02fafbc6 GKH	162	packet = kzalloc(sizeof(struct hv_netvsc_packet) +
	163	(num_frags * sizeof(struct hv_page_buffer)) +
	164	net_drv_obj->RequestExtSize, GFP_ATOMIC);
	165	if (!packet) {
4193d4f4	166	DPRINT_ERR(NETVSC_DRV, "unable to allocate hv_netvsc_packet");
fceaf24a HJ	167	return -1;
	168	}
	169
02fafbc6 GKH	170	packet->Extension = (void *)(unsigned long)packet +
	171	sizeof(struct hv_netvsc_packet) +
	172	(num_frags * sizeof(struct hv_page_buffer));
fceaf24a	173
454f18a9	174	/* Setup the rndis header */
fceaf24a HJ	175	packet->PageBufferCount = num_frags;
fceaf24a HJ	176
454f18a9 BP	177	/* TODO: Flush all write buffers/ memory fence ??? */
454f18a9 BP	178	/* wmb(); */
fceaf24a	179
454f18a9	180	/* Initialize it from the skb */
fceaf24a HJ	181	ASSERT(skb->data);
	182	packet->TotalDataBufferLength = skb->len;
	183
02fafbc6 GKH	184	/*
	185	* Start filling in the page buffers starting at
	186	* AdditionalRequestPageBufferCount offset
	187	*/
	188	packet->PageBuffers[net_drv_obj->AdditionalRequestPageBufferCount].Pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
	189	packet->PageBuffers[net_drv_obj->AdditionalRequestPageBufferCount].Offset = (unsigned long)skb->data & (PAGE_SIZE - 1);
	190	packet->PageBuffers[net_drv_obj->AdditionalRequestPageBufferCount].Length = skb->len - skb->data_len;
fceaf24a HJ	191
	192	ASSERT((skb->len - skb->data_len) <= PAGE_SIZE);
	193
02fafbc6 GKH	194	for (i = net_drv_obj->AdditionalRequestPageBufferCount + 1;
	195	i < num_frags; i++) {
	196	packet->PageBuffers[i].Pfn =
	197	page_to_pfn(skb_shinfo(skb)->frags[i-(net_drv_obj->AdditionalRequestPageBufferCount+1)].page);
	198	packet->PageBuffers[i].Offset =
	199	skb_shinfo(skb)->frags[i-(net_drv_obj->AdditionalRequestPageBufferCount+1)].page_offset;
	200	packet->PageBuffers[i].Length =
	201	skb_shinfo(skb)->frags[i-(net_drv_obj->AdditionalRequestPageBufferCount+1)].size;
fceaf24a HJ	202	}
fceaf24a HJ	203
454f18a9	204	/* Set the completion routine */
fceaf24a HJ	205	packet->Completion.Send.OnSendCompletion = netvsc_xmit_completion;
fceaf24a HJ	206	packet->Completion.Send.SendCompletionContext = packet;
c4b0bc94	207	packet->Completion.Send.SendCompletionTid = (unsigned long)skb;
fceaf24a HJ	208
fceaf24a HJ	209	retry_send:
02fafbc6 GKH	210	ret = net_drv_obj->OnSend(&net_device_ctx->device_ctx->device_obj,
02fafbc6 GKH	211	packet);
fceaf24a	212
02fafbc6	213	if (ret == 0) {
fceaf24a	214	ret = NETDEV_TX_OK;
b852fdce SH	215	net->stats.tx_bytes += skb->len;
b852fdce SH	216	net->stats.tx_packets++;
02fafbc6	217	} else {
fceaf24a	218	retries++;
02fafbc6 GKH	219	if (retries < 4) {
	220	DPRINT_ERR(NETVSC_DRV, "unable to send..."
	221	"retrying %d...", retries);
fceaf24a HJ	222	udelay(100);
	223	goto retry_send;
	224	}
	225
454f18a9	226	/* no more room or we are shutting down */
02fafbc6 GKH	227	DPRINT_ERR(NETVSC_DRV, "unable to send (%d)..."
02fafbc6 GKH	228	"marking net device (%p) busy", ret, net);
fceaf24a HJ	229	DPRINT_INFO(NETVSC_DRV, "net device (%p) stopping", net);
	230
	231	ret = NETDEV_TX_BUSY;
b852fdce	232	net->stats.tx_dropped++;
fceaf24a HJ	233
	234	netif_stop_queue(net);
	235
02fafbc6 GKH	236	/*
	237	* Null it since the caller will free it instead of the
	238	* completion routine
	239	*/
fceaf24a HJ	240	packet->Completion.Send.SendCompletionTid = 0;
fceaf24a HJ	241
02fafbc6 GKH	242	/*
	243	* Release the resources since we will not get any send
	244	* completion
	245	*/
	246	netvsc_xmit_completion((void *)packet);
fceaf24a HJ	247	}
fceaf24a HJ	248
02fafbc6	249	DPRINT_DBG(NETVSC_DRV, "# of xmits %lu total size %lu",
b852fdce SH	250	net->stats.tx_packets,
b852fdce SH	251	net->stats.tx_bytes);
fceaf24a HJ	252
	253	DPRINT_EXIT(NETVSC_DRV);
	254	return ret;
	255	}
	256
3e189519	257	/*
02fafbc6 GKH	258	* netvsc_linkstatus_callback - Link up/down notification
	259	*/
	260	static void netvsc_linkstatus_callback(struct hv_device *device_obj,
	261	unsigned int status)
fceaf24a	262	{
f916a34d	263	struct vm_device *device_ctx = to_vm_device(device_obj);
02fafbc6	264	struct net_device *net = dev_get_drvdata(&device_ctx->device);
fceaf24a HJ	265
	266	DPRINT_ENTER(NETVSC_DRV);
	267
02fafbc6 GKH	268	if (!net) {
	269	DPRINT_ERR(NETVSC_DRV, "got link status but net device "
	270	"not initialized yet");
fceaf24a HJ	271	return;
	272	}
	273
02fafbc6	274	if (status == 1) {
fceaf24a HJ	275	netif_carrier_on(net);
fceaf24a HJ	276	netif_wake_queue(net);
02fafbc6	277	} else {
fceaf24a HJ	278	netif_carrier_off(net);
	279	netif_stop_queue(net);
	280	}
	281	DPRINT_EXIT(NETVSC_DRV);
	282	}
	283
3e189519 HJ	284	/*
	285	* netvsc_recv_callback - Callback when we receive a packet from the
	286	* "wire" on the specified device.
02fafbc6 GKH	287	*/
	288	static int netvsc_recv_callback(struct hv_device *device_obj,
	289	struct hv_netvsc_packet *packet)
fceaf24a	290	{
f916a34d	291	struct vm_device *device_ctx = to_vm_device(device_obj);
621d7fb7	292	struct net_device *net = dev_get_drvdata(&device_ctx->device);
fceaf24a HJ	293	struct sk_buff *skb;
fceaf24a HJ	294	void *data;
02fafbc6	295	int i;
fceaf24a HJ	296	unsigned long flags;
	297
	298	DPRINT_ENTER(NETVSC_DRV);
	299
02fafbc6 GKH	300	if (!net) {
	301	DPRINT_ERR(NETVSC_DRV, "got receive callback but net device "
	302	"not initialized yet");
fceaf24a HJ	303	return 0;
	304	}
	305
9495c282 SH	306	/* Allocate a skb - TODO direct I/O to pages? */
	307	skb = netdev_alloc_skb_ip_align(net, packet->TotalDataBufferLength);
	308	if (unlikely(!skb)) {
	309	++net->stats.rx_dropped;
	310	return 0;
	311	}
fceaf24a	312
454f18a9	313	/* for kmap_atomic */
fceaf24a HJ	314	local_irq_save(flags);
fceaf24a HJ	315
02fafbc6 GKH	316	/*
	317	* Copy to skb. This copy is needed here since the memory pointed by
	318	* hv_netvsc_packet cannot be deallocated
	319	*/
	320	for (i = 0; i < packet->PageBufferCount; i++) {
	321	data = kmap_atomic(pfn_to_page(packet->PageBuffers[i].Pfn),
	322	KM_IRQ1);
	323	data = (void *)(unsigned long)data +
	324	packet->PageBuffers[i].Offset;
	325
	326	memcpy(skb_put(skb, packet->PageBuffers[i].Length), data,
	327	packet->PageBuffers[i].Length);
	328
	329	kunmap_atomic((void *)((unsigned long)data -
	330	packet->PageBuffers[i].Offset), KM_IRQ1);
fceaf24a HJ	331	}
	332
	333	local_irq_restore(flags);
	334
	335	skb->protocol = eth_type_trans(skb, net);
fceaf24a HJ	336	skb->ip_summed = CHECKSUM_NONE;
fceaf24a HJ	337
9495c282 SH	338	net->stats.rx_packets++;
	339	net->stats.rx_bytes += skb->len;
	340
02fafbc6 GKH	341	/*
02fafbc6 GKH	342	* Pass the skb back up. Network stack will deallocate the skb when it
9495c282 SH	343	* is done.
9495c282 SH	344	* TODO - use NAPI?
02fafbc6	345	*/
9495c282	346	netif_rx(skb);
fceaf24a	347
02fafbc6	348	DPRINT_DBG(NETVSC_DRV, "# of recvs %lu total size %lu",
b852fdce	349	net->stats.rx_packets, net->stats.rx_bytes);
fceaf24a HJ	350
	351	DPRINT_EXIT(NETVSC_DRV);
	352
	353	return 0;
	354	}
	355
df2fff28 GKH	356	static const struct net_device_ops device_ops = {
	357	.ndo_open = netvsc_open,
	358	.ndo_stop = netvsc_close,
	359	.ndo_start_xmit = netvsc_start_xmit,
df2fff28 GKH	360	.ndo_set_multicast_list = netvsc_set_multicast_list,
	361	};
	362
	363	static int netvsc_probe(struct device *device)
	364	{
	365	struct driver_context *driver_ctx =
	366	driver_to_driver_context(device->driver);
	367	struct netvsc_driver_context *net_drv_ctx =
	368	(struct netvsc_driver_context *)driver_ctx;
	369	struct netvsc_driver *net_drv_obj = &net_drv_ctx->drv_obj;
f916a34d	370	struct vm_device *device_ctx = device_to_vm_device(device);
df2fff28 GKH	371	struct hv_device *device_obj = &device_ctx->device_obj;
	372	struct net_device *net = NULL;
	373	struct net_device_context *net_device_ctx;
	374	struct netvsc_device_info device_info;
	375	int ret;
	376
	377	DPRINT_ENTER(NETVSC_DRV);
	378
	379	if (!net_drv_obj->Base.OnDeviceAdd)
	380	return -1;
	381
	382	net = alloc_netdev(sizeof(struct net_device_context), "seth%d",
	383	ether_setup);
	384	if (!net)
	385	return -1;
	386
	387	/* Set initial state */
	388	netif_carrier_off(net);
	389	netif_stop_queue(net);
	390
	391	net_device_ctx = netdev_priv(net);
	392	net_device_ctx->device_ctx = device_ctx;
	393	dev_set_drvdata(device, net);
	394
	395	/* Notify the netvsc driver of the new device */
	396	ret = net_drv_obj->Base.OnDeviceAdd(device_obj, &device_info);
	397	if (ret != 0) {
	398	free_netdev(net);
	399	dev_set_drvdata(device, NULL);
	400
	401	DPRINT_ERR(NETVSC_DRV, "unable to add netvsc device (ret %d)",
	402	ret);
	403	return ret;
	404	}
	405
	406	/*
	407	* If carrier is still off ie we did not get a link status callback,
	408	* update it if necessary
	409	*/
	410	/*
	411	* FIXME: We should use a atomic or test/set instead to avoid getting
	412	* out of sync with the device's link status
	413	*/
	414	if (!netif_carrier_ok(net))
	415	if (!device_info.LinkState)
	416	netif_carrier_on(net);
	417
	418	memcpy(net->dev_addr, device_info.MacAddr, ETH_ALEN);
	419
	420	net->netdev_ops = &device_ops;
	421
	422	SET_NETDEV_DEV(net, device);
	423
	424	ret = register_netdev(net);
	425	if (ret != 0) {
	426	/* Remove the device and release the resource */
	427	net_drv_obj->Base.OnDeviceRemove(device_obj);
	428	free_netdev(net);
	429	}
	430
	431	DPRINT_EXIT(NETVSC_DRV);
	432	return ret;
	433	}
	434
435	static int netvsc_remove(struct device *device)
436	{
437	struct driver_context *driver_ctx =
438	driver_to_driver_context(device->driver);
439	struct netvsc_driver_context *net_drv_ctx =
440	(struct netvsc_driver_context *)driver_ctx;
441	struct netvsc_driver *net_drv_obj = &net_drv_ctx->drv_obj;
f916a34d	442	struct vm_device *device_ctx = device_to_vm_device(device);
df2fff28 GKH	443	struct net_device *net = dev_get_drvdata(&device_ctx->device);
	444	struct hv_device *device_obj = &device_ctx->device_obj;
	445	int ret;
	446
	447	DPRINT_ENTER(NETVSC_DRV);
	448
	449	if (net == NULL) {
	450	DPRINT_INFO(NETVSC, "no net device to remove");
	451	DPRINT_EXIT(NETVSC_DRV);
	452	return 0;
	453	}
	454
	455	if (!net_drv_obj->Base.OnDeviceRemove) {
	456	DPRINT_EXIT(NETVSC_DRV);
	457	return -1;
	458	}
	459
	460	/* Stop outbound asap */
	461	netif_stop_queue(net);
	462	/* netif_carrier_off(net); */
	463
	464	unregister_netdev(net);
	465
	466	/*
	467	* Call to the vsc driver to let it know that the device is being
	468	* removed
	469	*/
	470	ret = net_drv_obj->Base.OnDeviceRemove(device_obj);
	471	if (ret != 0) {
	472	/* TODO: */
	473	DPRINT_ERR(NETVSC, "unable to remove vsc device (ret %d)", ret);
	474	}
	475
	476	free_netdev(net);
	477	DPRINT_EXIT(NETVSC_DRV);
	478	return ret;
	479	}
	480
fceaf24a HJ	481	static int netvsc_drv_exit_cb(struct device dev, void data)
	482	{
	483	struct device curr = (struct device )data;
02fafbc6	484
fceaf24a	485	*curr = dev;
02fafbc6 GKH	486	/* stop iterating */
02fafbc6 GKH	487	return 1;
fceaf24a HJ	488	}
fceaf24a HJ	489
bd1de709	490	static void netvsc_drv_exit(void)
fceaf24a	491	{
02fafbc6 GKH	492	struct netvsc_driver *netvsc_drv_obj = &g_netvsc_drv.drv_obj;
	493	struct driver_context *drv_ctx = &g_netvsc_drv.drv_ctx;
	494	struct device *current_dev;
2295ba2e	495	int ret;
fceaf24a HJ	496
	497	DPRINT_ENTER(NETVSC_DRV);
	498
02fafbc6	499	while (1) {
fceaf24a HJ	500	current_dev = NULL;
fceaf24a HJ	501
454f18a9	502	/* Get the device */
2295ba2e	503	ret = driver_for_each_device(&drv_ctx->driver, NULL,
02fafbc6	504	&current_dev, netvsc_drv_exit_cb);
2295ba2e BP	505	if (ret)
	506	DPRINT_WARN(NETVSC_DRV,
	507	"driver_for_each_device returned %d", ret);
	508
fceaf24a HJ	509	if (current_dev == NULL)
	510	break;
	511
454f18a9	512	/* Initiate removal from the top-down */
02fafbc6 GKH	513	DPRINT_INFO(NETVSC_DRV, "unregistering device (%p)...",
02fafbc6 GKH	514	current_dev);
fceaf24a HJ	515
	516	device_unregister(current_dev);
	517	}
	518
	519	if (netvsc_drv_obj->Base.OnCleanup)
	520	netvsc_drv_obj->Base.OnCleanup(&netvsc_drv_obj->Base);
	521
	522	vmbus_child_driver_unregister(drv_ctx);
	523
	524	DPRINT_EXIT(NETVSC_DRV);
	525
	526	return;
	527	}
	528
21707bed	529	static int netvsc_drv_init(int (drv_init)(struct hv_driver drv))
df2fff28 GKH	530	{
	531	struct netvsc_driver *net_drv_obj = &g_netvsc_drv.drv_obj;
	532	struct driver_context *drv_ctx = &g_netvsc_drv.drv_ctx;
	533	int ret;
	534
	535	DPRINT_ENTER(NETVSC_DRV);
	536
	537	vmbus_get_interface(&net_drv_obj->Base.VmbusChannelInterface);
	538
	539	net_drv_obj->RingBufferSize = netvsc_ringbuffer_size;
	540	net_drv_obj->OnReceiveCallback = netvsc_recv_callback;
	541	net_drv_obj->OnLinkStatusChanged = netvsc_linkstatus_callback;
	542
	543	/* Callback to client driver to complete the initialization */
21707bed	544	drv_init(&net_drv_obj->Base);
df2fff28 GKH	545
	546	drv_ctx->driver.name = net_drv_obj->Base.name;
	547	memcpy(&drv_ctx->class_id, &net_drv_obj->Base.deviceType,
	548	sizeof(struct hv_guid));
	549
	550	drv_ctx->probe = netvsc_probe;
	551	drv_ctx->remove = netvsc_remove;
	552
	553	/* The driver belongs to vmbus */
	554	ret = vmbus_child_driver_register(drv_ctx);
	555
	556	DPRINT_EXIT(NETVSC_DRV);
	557
	558	return ret;
	559	}
	560
fceaf24a HJ	561	static int __init netvsc_init(void)
	562	{
	563	int ret;
	564
	565	DPRINT_ENTER(NETVSC_DRV);
	566	DPRINT_INFO(NETVSC_DRV, "Netvsc initializing....");
	567
	568	ret = netvsc_drv_init(NetVscInitialize);
	569
	570	DPRINT_EXIT(NETVSC_DRV);
	571
	572	return ret;
	573	}
	574
	575	static void __exit netvsc_exit(void)
	576	{
	577	DPRINT_ENTER(NETVSC_DRV);
fceaf24a	578	netvsc_drv_exit();
fceaf24a HJ	579	DPRINT_EXIT(NETVSC_DRV);
	580	}
	581
26c14cc1 HJ	582	MODULE_LICENSE("GPL");
26c14cc1 HJ	583	MODULE_VERSION(HV_DRV_VERSION);
7880fc54	584	MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
fceaf24a HJ	585	module_param(netvsc_ringbuffer_size, int, S_IRUGO);
	586
	587	module_init(netvsc_init);
	588	module_exit(netvsc_exit);