netvsc_drv.c 13.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
/*
 * 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:
18
 *   Haiyang Zhang <haiyangz@microsoft.com>
19 20
 *   Hank Janssen  <hjanssen@microsoft.com>
 */
21 22
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

23
#include <linux/init.h>
24
#include <linux/atomic.h>
25 26 27 28 29 30 31 32 33
#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>
34
#include <linux/if_vlan.h>
35
#include <linux/in.h>
36
#include <linux/slab.h>
37 38 39 40
#include <net/arp.h>
#include <net/route.h>
#include <net/sock.h>
#include <net/pkt_sched.h>
41

42
#include "hyperv_net.h"
43 44

struct net_device_context {
45
	/* point back to our device context */
46
	struct hv_device *device_ctx;
47
	struct delayed_work dwork;
48
	struct work_struct work;
49 50
};

51
#define RING_SIZE_MIN 64
52
static int ring_size = 128;
S
Stephen Hemminger 已提交
53 54
module_param(ring_size, int, S_IRUGO);
MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)");
55

56 57
static void do_set_multicast(struct work_struct *w)
{
58 59
	struct net_device_context *ndevctx =
		container_of(w, struct net_device_context, work);
60 61 62 63
	struct netvsc_device *nvdev;
	struct rndis_device *rdev;

	nvdev = hv_get_drvdata(ndevctx->device_ctx);
64 65
	if (nvdev == NULL || nvdev->ndev == NULL)
		return;
66 67 68

	rdev = nvdev->extension;
	if (rdev == NULL)
69
		return;
70

71
	if (nvdev->ndev->flags & IFF_PROMISC)
72 73 74 75 76 77 78 79 80
		rndis_filter_set_packet_filter(rdev,
			NDIS_PACKET_TYPE_PROMISCUOUS);
	else
		rndis_filter_set_packet_filter(rdev,
			NDIS_PACKET_TYPE_BROADCAST |
			NDIS_PACKET_TYPE_ALL_MULTICAST |
			NDIS_PACKET_TYPE_DIRECTED);
}

81
static void netvsc_set_multicast_list(struct net_device *net)
82
{
83
	struct net_device_context *net_device_ctx = netdev_priv(net);
84

85
	schedule_work(&net_device_ctx->work);
86 87 88 89 90
}

static int netvsc_open(struct net_device *net)
{
	struct net_device_context *net_device_ctx = netdev_priv(net);
91
	struct hv_device *device_obj = net_device_ctx->device_ctx;
92
	int ret = 0;
93

94 95 96 97 98
	/* Open up the device */
	ret = rndis_filter_open(device_obj);
	if (ret != 0) {
		netdev_err(net, "unable to open device (ret %d).\n", ret);
		return ret;
99 100
	}

101 102
	netif_start_queue(net);

103 104 105 106 107 108
	return ret;
}

static int netvsc_close(struct net_device *net)
{
	struct net_device_context *net_device_ctx = netdev_priv(net);
109
	struct hv_device *device_obj = net_device_ctx->device_ctx;
110
	int ret;
111

112
	netif_tx_disable(net);
113

114 115
	/* Make sure netvsc_set_multicast_list doesn't re-enable filter! */
	cancel_work_sync(&net_device_ctx->work);
116
	ret = rndis_filter_close(device_obj);
117
	if (ret != 0)
118
		netdev_err(net, "unable to close device (ret %d).\n", ret);
119 120 121 122 123 124

	return ret;
}

static void netvsc_xmit_completion(void *context)
{
125
	struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
126
	struct sk_buff *skb = (struct sk_buff *)
127
		(unsigned long)packet->completion.send.send_completion_tid;
128 129 130

	kfree(packet);

131
	if (skb)
132
		dev_kfree_skb_any(skb);
133 134
}

135
static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
136 137
{
	struct net_device_context *net_device_ctx = netdev_priv(net);
138
	struct hv_netvsc_packet *packet;
139
	int ret;
140
	unsigned int i, num_pages, npg_data;
141

142
	/* Add multipages for skb->data and additional 2 for RNDIS */
143 144
	npg_data = (((unsigned long)skb->data + skb_headlen(skb) - 1)
		>> PAGE_SHIFT) - ((unsigned long)skb->data >> PAGE_SHIFT) + 1;
145
	num_pages = skb_shinfo(skb)->nr_frags + npg_data + 2;
146

147
	/* Allocate a netvsc packet based on # of frags. */
148
	packet = kzalloc(sizeof(struct hv_netvsc_packet) +
149
			 (num_pages * sizeof(struct hv_page_buffer)) +
150 151
			 sizeof(struct rndis_filter_packet) +
			 NDIS_VLAN_PPI_SIZE, GFP_ATOMIC);
152
	if (!packet) {
153
		/* out of memory, drop packet */
154
		netdev_err(net, "unable to allocate hv_netvsc_packet\n");
155 156 157

		dev_kfree_skb(skb);
		net->stats.tx_dropped++;
158
		return NETDEV_TX_OK;
159 160
	}

161 162
	packet->vlan_tci = skb->vlan_tci;

163
	packet->extension = (void *)(unsigned long)packet +
164
				sizeof(struct hv_netvsc_packet) +
165
				    (num_pages * sizeof(struct hv_page_buffer));
166

167 168
	/* If the rndis msg goes beyond 1 page, we will add 1 later */
	packet->page_buf_cnt = num_pages - 1;
169

170
	/* Initialize it from the skb */
171
	packet->total_data_buflen = skb->len;
172

173
	/* Start filling in the page buffers starting after RNDIS buffer. */
174 175
	packet->page_buf[1].pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
	packet->page_buf[1].offset
176
		= (unsigned long)skb->data & (PAGE_SIZE - 1);
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
	if (npg_data == 1)
		packet->page_buf[1].len = skb_headlen(skb);
	else
		packet->page_buf[1].len = PAGE_SIZE
			- packet->page_buf[1].offset;

	for (i = 2; i <= npg_data; i++) {
		packet->page_buf[i].pfn = virt_to_phys(skb->data
			+ PAGE_SIZE * (i-1)) >> PAGE_SHIFT;
		packet->page_buf[i].offset = 0;
		packet->page_buf[i].len = PAGE_SIZE;
	}
	if (npg_data > 1)
		packet->page_buf[npg_data].len = (((unsigned long)skb->data
			+ skb_headlen(skb) - 1) & (PAGE_SIZE - 1)) + 1;
192 193 194

	/* Additional fragments are after SKB data */
	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
E
Eric Dumazet 已提交
195
		const skb_frag_t *f = &skb_shinfo(skb)->frags[i];
196

197 198 199 200
		packet->page_buf[i+npg_data+1].pfn =
			page_to_pfn(skb_frag_page(f));
		packet->page_buf[i+npg_data+1].offset = f->page_offset;
		packet->page_buf[i+npg_data+1].len = skb_frag_size(f);
201 202
	}

203
	/* Set the completion routine */
204 205 206
	packet->completion.send.send_completion = netvsc_xmit_completion;
	packet->completion.send.send_completion_ctx = packet;
	packet->completion.send.send_completion_tid = (unsigned long)skb;
207

208
	ret = rndis_filter_send(net_device_ctx->device_ctx,
209 210
				  packet);
	if (ret == 0) {
211 212
		net->stats.tx_bytes += skb->len;
		net->stats.tx_packets++;
213
	} else {
214
		kfree(packet);
215 216 217 218
		if (ret != -EAGAIN) {
			dev_kfree_skb_any(skb);
			net->stats.tx_dropped++;
		}
219 220
	}

221
	return (ret == -EAGAIN) ? NETDEV_TX_BUSY : NETDEV_TX_OK;
222 223
}

224
/*
225 226
 * netvsc_linkstatus_callback - Link up/down notification
 */
227
void netvsc_linkstatus_callback(struct hv_device *device_obj,
228
				       unsigned int status)
229
{
230
	struct net_device *net;
231
	struct net_device_context *ndev_ctx;
232 233 234 235
	struct netvsc_device *net_device;

	net_device = hv_get_drvdata(device_obj);
	net = net_device->ndev;
236

237
	if (!net) {
238 239
		netdev_err(net, "got link status but net device "
				"not initialized yet\n");
240 241 242
		return;
	}

243
	if (status == 1) {
244
		netif_carrier_on(net);
245
		ndev_ctx = netdev_priv(net);
246
		schedule_delayed_work(&ndev_ctx->dwork, 0);
247
		schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20));
248
	} else {
249 250 251 252
		netif_carrier_off(net);
	}
}

253 254 255
/*
 * netvsc_recv_callback -  Callback when we receive a packet from the
 * "wire" on the specified device.
256
 */
257
int netvsc_recv_callback(struct hv_device *device_obj,
258
				struct hv_netvsc_packet *packet)
259
{
260
	struct net_device *net;
261 262
	struct sk_buff *skb;

263
	net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
264
	if (!net) {
265 266
		netdev_err(net, "got receive callback but net device"
			" not initialized yet\n");
267
		packet->status = NVSP_STAT_FAIL;
268 269 270
		return 0;
	}

271
	/* Allocate a skb - TODO direct I/O to pages? */
272
	skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
273 274
	if (unlikely(!skb)) {
		++net->stats.rx_dropped;
275
		packet->status = NVSP_STAT_FAIL;
276 277
		return 0;
	}
278

279 280 281 282
	/*
	 * Copy to skb. This copy is needed here since the memory pointed by
	 * hv_netvsc_packet cannot be deallocated
	 */
283 284
	memcpy(skb_put(skb, packet->total_data_buflen), packet->data,
		packet->total_data_buflen);
285 286 287

	skb->protocol = eth_type_trans(skb, net);
	skb->ip_summed = CHECKSUM_NONE;
288 289 290
	if (packet->vlan_tci & VLAN_TAG_PRESENT)
		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
				       packet->vlan_tci);
291

292
	net->stats.rx_packets++;
293
	net->stats.rx_bytes += packet->total_data_buflen;
294

295 296
	/*
	 * Pass the skb back up. Network stack will deallocate the skb when it
297 298
	 * is done.
	 * TODO - use NAPI?
299
	 */
300
	netif_rx(skb);
301 302 303 304

	return 0;
}

305 306 307
static void netvsc_get_drvinfo(struct net_device *net,
			       struct ethtool_drvinfo *info)
{
308 309
	strlcpy(info->driver, KBUILD_MODNAME, sizeof(info->driver));
	strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
310 311
}

312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
static int netvsc_change_mtu(struct net_device *ndev, int mtu)
{
	struct net_device_context *ndevctx = netdev_priv(ndev);
	struct hv_device *hdev =  ndevctx->device_ctx;
	struct netvsc_device *nvdev = hv_get_drvdata(hdev);
	struct netvsc_device_info device_info;
	int limit = ETH_DATA_LEN;

	if (nvdev == NULL || nvdev->destroy)
		return -ENODEV;

	if (nvdev->nvsp_version == NVSP_PROTOCOL_VERSION_2)
		limit = NETVSC_MTU;

	if (mtu < 68 || mtu > limit)
		return -EINVAL;

	nvdev->start_remove = true;
	cancel_delayed_work_sync(&ndevctx->dwork);
331
	cancel_work_sync(&ndevctx->work);
332
	netif_tx_disable(ndev);
333 334 335 336 337 338 339 340 341 342 343 344 345
	rndis_filter_device_remove(hdev);

	ndev->mtu = mtu;

	ndevctx->device_ctx = hdev;
	hv_set_drvdata(hdev, ndev);
	device_info.ring_size = ring_size;
	rndis_filter_device_add(hdev, &device_info);
	netif_wake_queue(ndev);

	return 0;
}

346 347 348 349 350 351

static int netvsc_set_mac_addr(struct net_device *ndev, void *p)
{
	struct net_device_context *ndevctx = netdev_priv(ndev);
	struct hv_device *hdev =  ndevctx->device_ctx;
	struct sockaddr *addr = p;
352
	char save_adr[ETH_ALEN];
353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373
	unsigned char save_aatype;
	int err;

	memcpy(save_adr, ndev->dev_addr, ETH_ALEN);
	save_aatype = ndev->addr_assign_type;

	err = eth_mac_addr(ndev, p);
	if (err != 0)
		return err;

	err = rndis_filter_set_device_mac(hdev, addr->sa_data);
	if (err != 0) {
		/* roll back to saved MAC */
		memcpy(ndev->dev_addr, save_adr, ETH_ALEN);
		ndev->addr_assign_type = save_aatype;
	}

	return err;
}


374 375 376 377 378
static const struct ethtool_ops ethtool_ops = {
	.get_drvinfo	= netvsc_get_drvinfo,
	.get_link	= ethtool_op_get_link,
};

379 380 381 382
static const struct net_device_ops device_ops = {
	.ndo_open =			netvsc_open,
	.ndo_stop =			netvsc_close,
	.ndo_start_xmit =		netvsc_start_xmit,
383
	.ndo_set_rx_mode =		netvsc_set_multicast_list,
384
	.ndo_change_mtu =		netvsc_change_mtu,
385
	.ndo_validate_addr =		eth_validate_addr,
386
	.ndo_set_mac_address =		netvsc_set_mac_addr,
387 388
};

389 390 391 392
/*
 * Send GARP packet to network peers after migrations.
 * After Quick Migration, the network is not immediately operational in the
 * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
393
 * another netif_notify_peers() into a delayed work, otherwise GARP packet
394 395 396 397 398 399
 * will not be sent after quick migration, and cause network disconnection.
 */
static void netvsc_send_garp(struct work_struct *w)
{
	struct net_device_context *ndev_ctx;
	struct net_device *net;
400
	struct netvsc_device *net_device;
401

402
	ndev_ctx = container_of(w, struct net_device_context, dwork.work);
403 404
	net_device = hv_get_drvdata(ndev_ctx->device_ctx);
	net = net_device->ndev;
405
	netdev_notify_peers(net);
406 407 408
}


409 410
static int netvsc_probe(struct hv_device *dev,
			const struct hv_vmbus_device_id *dev_id)
411 412 413 414 415 416
{
	struct net_device *net = NULL;
	struct net_device_context *net_device_ctx;
	struct netvsc_device_info device_info;
	int ret;

417
	net = alloc_etherdev(sizeof(struct net_device_context));
418
	if (!net)
419
		return -ENOMEM;
420 421 422 423 424

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

	net_device_ctx = netdev_priv(net);
425
	net_device_ctx->device_ctx = dev;
426
	hv_set_drvdata(dev, net);
427
	INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_send_garp);
428
	INIT_WORK(&net_device_ctx->work, do_set_multicast);
429 430 431

	net->netdev_ops = &device_ops;

432
	/* TODO: Add GSO and Checksum offload */
433 434
	net->hw_features = 0;
	net->features = NETIF_F_HW_VLAN_CTAG_TX;
435

436
	SET_ETHTOOL_OPS(net, &ethtool_ops);
437
	SET_NETDEV_DEV(net, &dev->device);
438 439 440

	ret = register_netdev(net);
	if (ret != 0) {
441
		pr_err("Unable to register netdev.\n");
442
		free_netdev(net);
443
		goto out;
444 445
	}

446 447 448 449 450 451
	/* Notify the netvsc driver of the new device */
	device_info.ring_size = ring_size;
	ret = rndis_filter_device_add(dev, &device_info);
	if (ret != 0) {
		netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
		unregister_netdev(net);
452
		free_netdev(net);
453
		hv_set_drvdata(dev, NULL);
454
		return ret;
455
	}
456 457 458
	memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);

	netif_carrier_on(net);
459

460
out:
461 462 463
	return ret;
}

464
static int netvsc_remove(struct hv_device *dev)
465
{
466
	struct net_device *net;
467
	struct net_device_context *ndev_ctx;
468 469 470 471
	struct netvsc_device *net_device;

	net_device = hv_get_drvdata(dev);
	net = net_device->ndev;
472 473

	if (net == NULL) {
474
		dev_err(&dev->device, "No net device to remove\n");
475 476 477
		return 0;
	}

478 479
	net_device->start_remove = true;

480 481
	ndev_ctx = netdev_priv(net);
	cancel_delayed_work_sync(&ndev_ctx->dwork);
482
	cancel_work_sync(&ndev_ctx->work);
483

484
	/* Stop outbound asap */
485
	netif_tx_disable(net);
486 487 488 489 490 491 492

	unregister_netdev(net);

	/*
	 * Call to the vsc driver to let it know that the device is being
	 * removed
	 */
493
	rndis_filter_device_remove(dev);
494 495

	free_netdev(net);
496
	return 0;
497 498
}

499
static const struct hv_vmbus_device_id id_table[] = {
500
	/* Network guid */
501
	{ HV_NIC_GUID, },
502
	{ },
503 504 505 506
};

MODULE_DEVICE_TABLE(vmbus, id_table);

507
/* The one and only one */
508
static struct  hv_driver netvsc_drv = {
509
	.name = KBUILD_MODNAME,
510
	.id_table = id_table,
511 512
	.probe = netvsc_probe,
	.remove = netvsc_remove,
513
};
514

515
static void __exit netvsc_drv_exit(void)
516
{
517
	vmbus_driver_unregister(&netvsc_drv);
518 519
}

520
static int __init netvsc_drv_init(void)
521
{
522 523 524 525 526
	if (ring_size < RING_SIZE_MIN) {
		ring_size = RING_SIZE_MIN;
		pr_info("Increased ring_size to %d (min allowed)\n",
			ring_size);
	}
527
	return vmbus_driver_register(&netvsc_drv);
528 529
}

530
MODULE_LICENSE("GPL");
531
MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
532

533
module_init(netvsc_drv_init);
534
module_exit(netvsc_drv_exit);