1/*
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:
18 *   Haiyang Zhang <haiyangz@microsoft.com>
19 *   Hank Janssen  <hjanssen@microsoft.com>
20 */
21#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22
23#include <linux/init.h>
24#include <linux/atomic.h>
25#include <linux/module.h>
26#include <linux/highmem.h>
27#include <linux/device.h>
28#include <linux/io.h>
29#include <linux/delay.h>
30#include <linux/netdevice.h>
31#include <linux/inetdevice.h>
32#include <linux/etherdevice.h>
33#include <linux/skbuff.h>
34#include <linux/in.h>
35#include <linux/slab.h>
36#include <net/arp.h>
37#include <net/route.h>
38#include <net/sock.h>
39#include <net/pkt_sched.h>
40
41#include "hyperv_net.h"
42
43struct net_device_context {
44	/* point back to our device context */
45	struct hv_device *device_ctx;
46	struct delayed_work dwork;
47	struct work_struct work;
48};
49
50
51static int ring_size = 128;
52module_param(ring_size, int, S_IRUGO);
53MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)");
54
55static void do_set_multicast(struct work_struct *w)
56{
57	struct net_device_context *ndevctx =
58		container_of(w, struct net_device_context, work);
59	struct netvsc_device *nvdev;
60	struct rndis_device *rdev;
61
62	nvdev = hv_get_drvdata(ndevctx->device_ctx);
63	if (nvdev == NULL || nvdev->ndev == NULL)
64		return;
65
66	rdev = nvdev->extension;
67	if (rdev == NULL)
68		return;
69
70	if (nvdev->ndev->flags & IFF_PROMISC)
71		rndis_filter_set_packet_filter(rdev,
72			NDIS_PACKET_TYPE_PROMISCUOUS);
73	else
74		rndis_filter_set_packet_filter(rdev,
75			NDIS_PACKET_TYPE_BROADCAST |
76			NDIS_PACKET_TYPE_ALL_MULTICAST |
77			NDIS_PACKET_TYPE_DIRECTED);
78}
79
80static void netvsc_set_multicast_list(struct net_device *net)
81{
82	struct net_device_context *net_device_ctx = netdev_priv(net);
83
84	schedule_work(&net_device_ctx->work);
85}
86
87static int netvsc_open(struct net_device *net)
88{
89	struct net_device_context *net_device_ctx = netdev_priv(net);
90	struct hv_device *device_obj = net_device_ctx->device_ctx;
91	int ret = 0;
92
93	/* Open up the device */
94	ret = rndis_filter_open(device_obj);
95	if (ret != 0) {
96		netdev_err(net, "unable to open device (ret %d).\n", ret);
97		return ret;
98	}
99
100	netif_start_queue(net);
101
102	return ret;
103}
104
105static int netvsc_close(struct net_device *net)
106{
107	struct net_device_context *net_device_ctx = netdev_priv(net);
108	struct hv_device *device_obj = net_device_ctx->device_ctx;
109	int ret;
110
111	netif_tx_disable(net);
112
113	/* Make sure netvsc_set_multicast_list doesn't re-enable filter! */
114	cancel_work_sync(&net_device_ctx->work);
115	ret = rndis_filter_close(device_obj);
116	if (ret != 0)
117		netdev_err(net, "unable to close device (ret %d).\n", ret);
118
119	return ret;
120}
121
122static void netvsc_xmit_completion(void *context)
123{
124	struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
125	struct sk_buff *skb = (struct sk_buff *)
126		(unsigned long)packet->completion.send.send_completion_tid;
127
128	kfree(packet);
129
130	if (skb)
131		dev_kfree_skb_any(skb);
132}
133
134static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
135{
136	struct net_device_context *net_device_ctx = netdev_priv(net);
137	struct hv_netvsc_packet *packet;
138	int ret;
139	unsigned int i, num_pages, npg_data;
140
141	/* Add multipages for skb->data and additional 2 for RNDIS */
142	npg_data = (((unsigned long)skb->data + skb_headlen(skb) - 1)
143		>> PAGE_SHIFT) - ((unsigned long)skb->data >> PAGE_SHIFT) + 1;
144	num_pages = skb_shinfo(skb)->nr_frags + npg_data + 2;
145
146	/* Allocate a netvsc packet based on # of frags. */
147	packet = kzalloc(sizeof(struct hv_netvsc_packet) +
148			 (num_pages * sizeof(struct hv_page_buffer)) +
149			 sizeof(struct rndis_filter_packet) +
150			 NDIS_VLAN_PPI_SIZE, GFP_ATOMIC);
151	if (!packet) {
152		/* out of memory, drop packet */
153		netdev_err(net, "unable to allocate hv_netvsc_packet\n");
154
155		dev_kfree_skb(skb);
156		net->stats.tx_dropped++;
157		return NETDEV_TX_OK;
158	}
159
160	packet->vlan_tci = skb->vlan_tci;
161
162	packet->extension = (void *)(unsigned long)packet +
163				sizeof(struct hv_netvsc_packet) +
164				    (num_pages * sizeof(struct hv_page_buffer));
165
166	/* If the rndis msg goes beyond 1 page, we will add 1 later */
167	packet->page_buf_cnt = num_pages - 1;
168
169	/* Initialize it from the skb */
170	packet->total_data_buflen = skb->len;
171
172	/* Start filling in the page buffers starting after RNDIS buffer. */
173	packet->page_buf[1].pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
174	packet->page_buf[1].offset
175		= (unsigned long)skb->data & (PAGE_SIZE - 1);
176	if (npg_data == 1)
177		packet->page_buf[1].len = skb_headlen(skb);
178	else
179		packet->page_buf[1].len = PAGE_SIZE
180			- packet->page_buf[1].offset;
181
182	for (i = 2; i <= npg_data; i++) {
183		packet->page_buf[i].pfn = virt_to_phys(skb->data
184			+ PAGE_SIZE * (i-1)) >> PAGE_SHIFT;
185		packet->page_buf[i].offset = 0;
186		packet->page_buf[i].len = PAGE_SIZE;
187	}
188	if (npg_data > 1)
189		packet->page_buf[npg_data].len = (((unsigned long)skb->data
190			+ skb_headlen(skb) - 1) & (PAGE_SIZE - 1)) + 1;
191
192	/* Additional fragments are after SKB data */
193	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
194		const skb_frag_t *f = &skb_shinfo(skb)->frags[i];
195
196		packet->page_buf[i+npg_data+1].pfn =
197			page_to_pfn(skb_frag_page(f));
198		packet->page_buf[i+npg_data+1].offset = f->page_offset;
199		packet->page_buf[i+npg_data+1].len = skb_frag_size(f);
200	}
201
202	/* Set the completion routine */
203	packet->completion.send.send_completion = netvsc_xmit_completion;
204	packet->completion.send.send_completion_ctx = packet;
205	packet->completion.send.send_completion_tid = (unsigned long)skb;
206
207	ret = rndis_filter_send(net_device_ctx->device_ctx,
208				  packet);
209	if (ret == 0) {
210		net->stats.tx_bytes += skb->len;
211		net->stats.tx_packets++;
212	} else {
213		kfree(packet);
214	}
215
216	return ret ? NETDEV_TX_BUSY : NETDEV_TX_OK;
217}
218
219/*
220 * netvsc_linkstatus_callback - Link up/down notification
221 */
222void netvsc_linkstatus_callback(struct hv_device *device_obj,
223				       unsigned int status)
224{
225	struct net_device *net;
226	struct net_device_context *ndev_ctx;
227	struct netvsc_device *net_device;
228
229	net_device = hv_get_drvdata(device_obj);
230	net = net_device->ndev;
231
232	if (!net) {
233		netdev_err(net, "got link status but net device "
234				"not initialized yet\n");
235		return;
236	}
237
238	if (status == 1) {
239		netif_carrier_on(net);
240		netif_wake_queue(net);
241		ndev_ctx = netdev_priv(net);
242		schedule_delayed_work(&ndev_ctx->dwork, 0);
243		schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20));
244	} else {
245		netif_carrier_off(net);
246		netif_tx_disable(net);
247	}
248}
249
250/*
251 * netvsc_recv_callback -  Callback when we receive a packet from the
252 * "wire" on the specified device.
253 */
254int netvsc_recv_callback(struct hv_device *device_obj,
255				struct hv_netvsc_packet *packet)
256{
257	struct net_device *net;
258	struct sk_buff *skb;
259
260	net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
261	if (!net) {
262		netdev_err(net, "got receive callback but net device"
263			" not initialized yet\n");
264		return 0;
265	}
266
267	/* Allocate a skb - TODO direct I/O to pages? */
268	skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
269	if (unlikely(!skb)) {
270		++net->stats.rx_dropped;
271		return 0;
272	}
273
274	/*
275	 * Copy to skb. This copy is needed here since the memory pointed by
276	 * hv_netvsc_packet cannot be deallocated
277	 */
278	memcpy(skb_put(skb, packet->total_data_buflen), packet->data,
279		packet->total_data_buflen);
280
281	skb->protocol = eth_type_trans(skb, net);
282	skb->ip_summed = CHECKSUM_NONE;
283	skb->vlan_tci = packet->vlan_tci;
284
285	net->stats.rx_packets++;
286	net->stats.rx_bytes += packet->total_data_buflen;
287
288	/*
289	 * Pass the skb back up. Network stack will deallocate the skb when it
290	 * is done.
291	 * TODO - use NAPI?
292	 */
293	netif_rx(skb);
294
295	return 0;
296}
297
298static void netvsc_get_drvinfo(struct net_device *net,
299			       struct ethtool_drvinfo *info)
300{
301	strcpy(info->driver, KBUILD_MODNAME);
302	strcpy(info->version, HV_DRV_VERSION);
303	strcpy(info->fw_version, "N/A");
304}
305
306static int netvsc_change_mtu(struct net_device *ndev, int mtu)
307{
308	struct net_device_context *ndevctx = netdev_priv(ndev);
309	struct hv_device *hdev =  ndevctx->device_ctx;
310	struct netvsc_device *nvdev = hv_get_drvdata(hdev);
311	struct netvsc_device_info device_info;
312	int limit = ETH_DATA_LEN;
313
314	if (nvdev == NULL || nvdev->destroy)
315		return -ENODEV;
316
317	if (nvdev->nvsp_version == NVSP_PROTOCOL_VERSION_2)
318		limit = NETVSC_MTU;
319
320	if (mtu < 68 || mtu > limit)
321		return -EINVAL;
322
323	nvdev->start_remove = true;
324	cancel_delayed_work_sync(&ndevctx->dwork);
325	cancel_work_sync(&ndevctx->work);
326	netif_tx_disable(ndev);
327	rndis_filter_device_remove(hdev);
328
329	ndev->mtu = mtu;
330
331	ndevctx->device_ctx = hdev;
332	hv_set_drvdata(hdev, ndev);
333	device_info.ring_size = ring_size;
334	rndis_filter_device_add(hdev, &device_info);
335	netif_wake_queue(ndev);
336
337	return 0;
338}
339
340static const struct ethtool_ops ethtool_ops = {
341	.get_drvinfo	= netvsc_get_drvinfo,
342	.get_link	= ethtool_op_get_link,
343};
344
345static const struct net_device_ops device_ops = {
346	.ndo_open =			netvsc_open,
347	.ndo_stop =			netvsc_close,
348	.ndo_start_xmit =		netvsc_start_xmit,
349	.ndo_set_rx_mode =		netvsc_set_multicast_list,
350	.ndo_change_mtu =		netvsc_change_mtu,
351	.ndo_validate_addr =		eth_validate_addr,
352	.ndo_set_mac_address =		eth_mac_addr,
353};
354
355/*
356 * Send GARP packet to network peers after migrations.
357 * After Quick Migration, the network is not immediately operational in the
358 * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
359 * another netif_notify_peers() into a delayed work, otherwise GARP packet
360 * will not be sent after quick migration, and cause network disconnection.
361 */
362static void netvsc_send_garp(struct work_struct *w)
363{
364	struct net_device_context *ndev_ctx;
365	struct net_device *net;
366	struct netvsc_device *net_device;
367
368	ndev_ctx = container_of(w, struct net_device_context, dwork.work);
369	net_device = hv_get_drvdata(ndev_ctx->device_ctx);
370	net = net_device->ndev;
371	netif_notify_peers(net);
372}
373
374
375static int netvsc_probe(struct hv_device *dev,
376			const struct hv_vmbus_device_id *dev_id)
377{
378	struct net_device *net = NULL;
379	struct net_device_context *net_device_ctx;
380	struct netvsc_device_info device_info;
381	int ret;
382
383	net = alloc_etherdev(sizeof(struct net_device_context));
384	if (!net)
385		return -ENOMEM;
386
387	/* Set initial state */
388	netif_carrier_off(net);
389
390	net_device_ctx = netdev_priv(net);
391	net_device_ctx->device_ctx = dev;
392	hv_set_drvdata(dev, net);
393	INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_send_garp);
394	INIT_WORK(&net_device_ctx->work, do_set_multicast);
395
396	net->netdev_ops = &device_ops;
397
398	/* TODO: Add GSO and Checksum offload */
399	net->hw_features = NETIF_F_SG;
400	net->features = NETIF_F_SG | NETIF_F_HW_VLAN_TX;
401
402	SET_ETHTOOL_OPS(net, &ethtool_ops);
403	SET_NETDEV_DEV(net, &dev->device);
404
405	ret = register_netdev(net);
406	if (ret != 0) {
407		pr_err("Unable to register netdev.\n");
408		free_netdev(net);
409		goto out;
410	}
411
412	/* Notify the netvsc driver of the new device */
413	device_info.ring_size = ring_size;
414	ret = rndis_filter_device_add(dev, &device_info);
415	if (ret != 0) {
416		netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
417		unregister_netdev(net);
418		free_netdev(net);
419		hv_set_drvdata(dev, NULL);
420		return ret;
421	}
422	memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);
423
424	netif_carrier_on(net);
425
426out:
427	return ret;
428}
429
430static int netvsc_remove(struct hv_device *dev)
431{
432	struct net_device *net;
433	struct net_device_context *ndev_ctx;
434	struct netvsc_device *net_device;
435
436	net_device = hv_get_drvdata(dev);
437	net = net_device->ndev;
438
439	if (net == NULL) {
440		dev_err(&dev->device, "No net device to remove\n");
441		return 0;
442	}
443
444	net_device->start_remove = true;
445
446	ndev_ctx = netdev_priv(net);
447	cancel_delayed_work_sync(&ndev_ctx->dwork);
448	cancel_work_sync(&ndev_ctx->work);
449
450	/* Stop outbound asap */
451	netif_tx_disable(net);
452
453	unregister_netdev(net);
454
455	/*
456	 * Call to the vsc driver to let it know that the device is being
457	 * removed
458	 */
459	rndis_filter_device_remove(dev);
460
461	free_netdev(net);
462	return 0;
463}
464
465static const struct hv_vmbus_device_id id_table[] = {
466	/* Network guid */
467	{ VMBUS_DEVICE(0x63, 0x51, 0x61, 0xF8, 0x3E, 0xDF, 0xc5, 0x46,
468		       0x91, 0x3F, 0xF2, 0xD2, 0xF9, 0x65, 0xED, 0x0E) },
469	{ },
470};
471
472MODULE_DEVICE_TABLE(vmbus, id_table);
473
474/* The one and only one */
475static struct  hv_driver netvsc_drv = {
476	.name = KBUILD_MODNAME,
477	.id_table = id_table,
478	.probe = netvsc_probe,
479	.remove = netvsc_remove,
480};
481
482static void __exit netvsc_drv_exit(void)
483{
484	vmbus_driver_unregister(&netvsc_drv);
485}
486
487static int __init netvsc_drv_init(void)
488{
489	return vmbus_driver_register(&netvsc_drv);
490}
491
492MODULE_LICENSE("GPL");
493MODULE_VERSION(HV_DRV_VERSION);
494MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
495
496module_init(netvsc_drv_init);
497module_exit(netvsc_drv_exit);
498