2 * Copyright (c) 2009, Microsoft Corporation.
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.
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
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.
18 * Haiyang Zhang <haiyangz@microsoft.com>
19 * Hank Janssen <hjanssen@microsoft.com>
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
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>
35 #include <linux/slab.h>
37 #include <net/route.h>
39 #include <net/pkt_sched.h>
41 #include "hyperv_net.h"
43 struct net_device_context
{
44 /* point back to our device context */
45 struct hv_device
*device_ctx
;
46 struct delayed_work dwork
;
50 static int ring_size
= 128;
51 module_param(ring_size
, int, S_IRUGO
);
52 MODULE_PARM_DESC(ring_size
, "Ring buffer size (# of pages)");
54 struct set_multicast_work
{
55 struct work_struct work
;
56 struct net_device
*net
;
59 static void do_set_multicast(struct work_struct
*w
)
61 struct set_multicast_work
*swk
=
62 container_of(w
, struct set_multicast_work
, work
);
63 struct net_device
*net
= swk
->net
;
65 struct net_device_context
*ndevctx
= netdev_priv(net
);
66 struct netvsc_device
*nvdev
;
67 struct rndis_device
*rdev
;
69 nvdev
= hv_get_drvdata(ndevctx
->device_ctx
);
73 rdev
= nvdev
->extension
;
77 if (net
->flags
& IFF_PROMISC
)
78 rndis_filter_set_packet_filter(rdev
,
79 NDIS_PACKET_TYPE_PROMISCUOUS
);
81 rndis_filter_set_packet_filter(rdev
,
82 NDIS_PACKET_TYPE_BROADCAST
|
83 NDIS_PACKET_TYPE_ALL_MULTICAST
|
84 NDIS_PACKET_TYPE_DIRECTED
);
90 static void netvsc_set_multicast_list(struct net_device
*net
)
92 struct set_multicast_work
*swk
=
93 kmalloc(sizeof(struct set_multicast_work
), GFP_ATOMIC
);
98 INIT_WORK(&swk
->work
, do_set_multicast
);
99 schedule_work(&swk
->work
);
102 static int netvsc_open(struct net_device
*net
)
104 struct net_device_context
*net_device_ctx
= netdev_priv(net
);
105 struct hv_device
*device_obj
= net_device_ctx
->device_ctx
;
108 /* Open up the device */
109 ret
= rndis_filter_open(device_obj
);
111 netdev_err(net
, "unable to open device (ret %d).\n", ret
);
115 netif_start_queue(net
);
120 static int netvsc_close(struct net_device
*net
)
122 struct net_device_context
*net_device_ctx
= netdev_priv(net
);
123 struct hv_device
*device_obj
= net_device_ctx
->device_ctx
;
126 netif_tx_disable(net
);
128 ret
= rndis_filter_close(device_obj
);
130 netdev_err(net
, "unable to close device (ret %d).\n", ret
);
135 static void netvsc_xmit_completion(void *context
)
137 struct hv_netvsc_packet
*packet
= (struct hv_netvsc_packet
*)context
;
138 struct sk_buff
*skb
= (struct sk_buff
*)
139 (unsigned long)packet
->completion
.send
.send_completion_tid
;
144 dev_kfree_skb_any(skb
);
147 static int netvsc_start_xmit(struct sk_buff
*skb
, struct net_device
*net
)
149 struct net_device_context
*net_device_ctx
= netdev_priv(net
);
150 struct hv_netvsc_packet
*packet
;
152 unsigned int i
, num_pages
, npg_data
;
154 /* Add multipages for skb->data and additional 2 for RNDIS */
155 npg_data
= (((unsigned long)skb
->data
+ skb_headlen(skb
) - 1)
156 >> PAGE_SHIFT
) - ((unsigned long)skb
->data
>> PAGE_SHIFT
) + 1;
157 num_pages
= skb_shinfo(skb
)->nr_frags
+ npg_data
+ 2;
159 /* Allocate a netvsc packet based on # of frags. */
160 packet
= kzalloc(sizeof(struct hv_netvsc_packet
) +
161 (num_pages
* sizeof(struct hv_page_buffer
)) +
162 sizeof(struct rndis_filter_packet
) +
163 NDIS_VLAN_PPI_SIZE
, GFP_ATOMIC
);
165 /* out of memory, drop packet */
166 netdev_err(net
, "unable to allocate hv_netvsc_packet\n");
169 net
->stats
.tx_dropped
++;
170 return NETDEV_TX_BUSY
;
173 packet
->vlan_tci
= skb
->vlan_tci
;
175 packet
->extension
= (void *)(unsigned long)packet
+
176 sizeof(struct hv_netvsc_packet
) +
177 (num_pages
* sizeof(struct hv_page_buffer
));
179 /* If the rndis msg goes beyond 1 page, we will add 1 later */
180 packet
->page_buf_cnt
= num_pages
- 1;
182 /* Initialize it from the skb */
183 packet
->total_data_buflen
= skb
->len
;
185 /* Start filling in the page buffers starting after RNDIS buffer. */
186 packet
->page_buf
[1].pfn
= virt_to_phys(skb
->data
) >> PAGE_SHIFT
;
187 packet
->page_buf
[1].offset
188 = (unsigned long)skb
->data
& (PAGE_SIZE
- 1);
190 packet
->page_buf
[1].len
= skb_headlen(skb
);
192 packet
->page_buf
[1].len
= PAGE_SIZE
193 - packet
->page_buf
[1].offset
;
195 for (i
= 2; i
<= npg_data
; i
++) {
196 packet
->page_buf
[i
].pfn
= virt_to_phys(skb
->data
197 + PAGE_SIZE
* (i
-1)) >> PAGE_SHIFT
;
198 packet
->page_buf
[i
].offset
= 0;
199 packet
->page_buf
[i
].len
= PAGE_SIZE
;
202 packet
->page_buf
[npg_data
].len
= (((unsigned long)skb
->data
203 + skb_headlen(skb
) - 1) & (PAGE_SIZE
- 1)) + 1;
205 /* Additional fragments are after SKB data */
206 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
207 const skb_frag_t
*f
= &skb_shinfo(skb
)->frags
[i
];
209 packet
->page_buf
[i
+npg_data
+1].pfn
=
210 page_to_pfn(skb_frag_page(f
));
211 packet
->page_buf
[i
+npg_data
+1].offset
= f
->page_offset
;
212 packet
->page_buf
[i
+npg_data
+1].len
= skb_frag_size(f
);
215 /* Set the completion routine */
216 packet
->completion
.send
.send_completion
= netvsc_xmit_completion
;
217 packet
->completion
.send
.send_completion_ctx
= packet
;
218 packet
->completion
.send
.send_completion_tid
= (unsigned long)skb
;
220 ret
= rndis_filter_send(net_device_ctx
->device_ctx
,
223 net
->stats
.tx_bytes
+= skb
->len
;
224 net
->stats
.tx_packets
++;
226 /* we are shutting down or bus overloaded, just drop packet */
227 net
->stats
.tx_dropped
++;
229 dev_kfree_skb_any(skb
);
232 return ret
? NETDEV_TX_BUSY
: NETDEV_TX_OK
;
236 * netvsc_linkstatus_callback - Link up/down notification
238 void netvsc_linkstatus_callback(struct hv_device
*device_obj
,
241 struct net_device
*net
;
242 struct net_device_context
*ndev_ctx
;
243 struct netvsc_device
*net_device
;
245 net_device
= hv_get_drvdata(device_obj
);
246 net
= net_device
->ndev
;
249 netdev_err(net
, "got link status but net device "
250 "not initialized yet\n");
255 netif_carrier_on(net
);
256 netif_wake_queue(net
);
257 ndev_ctx
= netdev_priv(net
);
258 schedule_delayed_work(&ndev_ctx
->dwork
, 0);
259 schedule_delayed_work(&ndev_ctx
->dwork
, msecs_to_jiffies(20));
261 netif_carrier_off(net
);
262 netif_tx_disable(net
);
267 * netvsc_recv_callback - Callback when we receive a packet from the
268 * "wire" on the specified device.
270 int netvsc_recv_callback(struct hv_device
*device_obj
,
271 struct hv_netvsc_packet
*packet
)
273 struct net_device
*net
;
276 net
= ((struct netvsc_device
*)hv_get_drvdata(device_obj
))->ndev
;
278 netdev_err(net
, "got receive callback but net device"
279 " not initialized yet\n");
283 /* Allocate a skb - TODO direct I/O to pages? */
284 skb
= netdev_alloc_skb_ip_align(net
, packet
->total_data_buflen
);
285 if (unlikely(!skb
)) {
286 ++net
->stats
.rx_dropped
;
291 * Copy to skb. This copy is needed here since the memory pointed by
292 * hv_netvsc_packet cannot be deallocated
294 memcpy(skb_put(skb
, packet
->total_data_buflen
), packet
->data
,
295 packet
->total_data_buflen
);
297 skb
->protocol
= eth_type_trans(skb
, net
);
298 skb
->ip_summed
= CHECKSUM_NONE
;
299 skb
->vlan_tci
= packet
->vlan_tci
;
301 net
->stats
.rx_packets
++;
302 net
->stats
.rx_bytes
+= packet
->total_data_buflen
;
305 * Pass the skb back up. Network stack will deallocate the skb when it
314 static void netvsc_get_drvinfo(struct net_device
*net
,
315 struct ethtool_drvinfo
*info
)
317 strcpy(info
->driver
, KBUILD_MODNAME
);
318 strcpy(info
->version
, HV_DRV_VERSION
);
319 strcpy(info
->fw_version
, "N/A");
322 static int netvsc_change_mtu(struct net_device
*ndev
, int mtu
)
324 struct net_device_context
*ndevctx
= netdev_priv(ndev
);
325 struct hv_device
*hdev
= ndevctx
->device_ctx
;
326 struct netvsc_device
*nvdev
= hv_get_drvdata(hdev
);
327 struct netvsc_device_info device_info
;
328 int limit
= ETH_DATA_LEN
;
330 if (nvdev
== NULL
|| nvdev
->destroy
)
333 if (nvdev
->nvsp_version
== NVSP_PROTOCOL_VERSION_2
)
336 if (mtu
< 68 || mtu
> limit
)
339 nvdev
->start_remove
= true;
340 cancel_delayed_work_sync(&ndevctx
->dwork
);
341 netif_tx_disable(ndev
);
342 rndis_filter_device_remove(hdev
);
346 ndevctx
->device_ctx
= hdev
;
347 hv_set_drvdata(hdev
, ndev
);
348 device_info
.ring_size
= ring_size
;
349 rndis_filter_device_add(hdev
, &device_info
);
350 netif_wake_queue(ndev
);
355 static const struct ethtool_ops ethtool_ops
= {
356 .get_drvinfo
= netvsc_get_drvinfo
,
357 .get_link
= ethtool_op_get_link
,
360 static const struct net_device_ops device_ops
= {
361 .ndo_open
= netvsc_open
,
362 .ndo_stop
= netvsc_close
,
363 .ndo_start_xmit
= netvsc_start_xmit
,
364 .ndo_set_rx_mode
= netvsc_set_multicast_list
,
365 .ndo_change_mtu
= netvsc_change_mtu
,
366 .ndo_validate_addr
= eth_validate_addr
,
367 .ndo_set_mac_address
= eth_mac_addr
,
371 * Send GARP packet to network peers after migrations.
372 * After Quick Migration, the network is not immediately operational in the
373 * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
374 * another netif_notify_peers() into a delayed work, otherwise GARP packet
375 * will not be sent after quick migration, and cause network disconnection.
377 static void netvsc_send_garp(struct work_struct
*w
)
379 struct net_device_context
*ndev_ctx
;
380 struct net_device
*net
;
381 struct netvsc_device
*net_device
;
383 ndev_ctx
= container_of(w
, struct net_device_context
, dwork
.work
);
384 net_device
= hv_get_drvdata(ndev_ctx
->device_ctx
);
385 net
= net_device
->ndev
;
386 netif_notify_peers(net
);
390 static int netvsc_probe(struct hv_device
*dev
,
391 const struct hv_vmbus_device_id
*dev_id
)
393 struct net_device
*net
= NULL
;
394 struct net_device_context
*net_device_ctx
;
395 struct netvsc_device_info device_info
;
398 net
= alloc_etherdev(sizeof(struct net_device_context
));
402 /* Set initial state */
403 netif_carrier_off(net
);
405 net_device_ctx
= netdev_priv(net
);
406 net_device_ctx
->device_ctx
= dev
;
407 hv_set_drvdata(dev
, net
);
408 INIT_DELAYED_WORK(&net_device_ctx
->dwork
, netvsc_send_garp
);
410 net
->netdev_ops
= &device_ops
;
412 /* TODO: Add GSO and Checksum offload */
413 net
->hw_features
= NETIF_F_SG
;
414 net
->features
= NETIF_F_SG
| NETIF_F_HW_VLAN_TX
;
416 SET_ETHTOOL_OPS(net
, ðtool_ops
);
417 SET_NETDEV_DEV(net
, &dev
->device
);
419 ret
= register_netdev(net
);
421 pr_err("Unable to register netdev.\n");
426 /* Notify the netvsc driver of the new device */
427 device_info
.ring_size
= ring_size
;
428 ret
= rndis_filter_device_add(dev
, &device_info
);
430 netdev_err(net
, "unable to add netvsc device (ret %d)\n", ret
);
431 unregister_netdev(net
);
433 hv_set_drvdata(dev
, NULL
);
436 memcpy(net
->dev_addr
, device_info
.mac_adr
, ETH_ALEN
);
438 netif_carrier_on(net
);
444 static int netvsc_remove(struct hv_device
*dev
)
446 struct net_device
*net
;
447 struct net_device_context
*ndev_ctx
;
448 struct netvsc_device
*net_device
;
450 net_device
= hv_get_drvdata(dev
);
451 net
= net_device
->ndev
;
454 dev_err(&dev
->device
, "No net device to remove\n");
458 net_device
->start_remove
= true;
460 ndev_ctx
= netdev_priv(net
);
461 cancel_delayed_work_sync(&ndev_ctx
->dwork
);
463 /* Stop outbound asap */
464 netif_tx_disable(net
);
466 unregister_netdev(net
);
469 * Call to the vsc driver to let it know that the device is being
472 rndis_filter_device_remove(dev
);
478 static const struct hv_vmbus_device_id id_table
[] = {
480 { VMBUS_DEVICE(0x63, 0x51, 0x61, 0xF8, 0x3E, 0xDF, 0xc5, 0x46,
481 0x91, 0x3F, 0xF2, 0xD2, 0xF9, 0x65, 0xED, 0x0E) },
485 MODULE_DEVICE_TABLE(vmbus
, id_table
);
487 /* The one and only one */
488 static struct hv_driver netvsc_drv
= {
489 .name
= KBUILD_MODNAME
,
490 .id_table
= id_table
,
491 .probe
= netvsc_probe
,
492 .remove
= netvsc_remove
,
495 static void __exit
netvsc_drv_exit(void)
497 vmbus_driver_unregister(&netvsc_drv
);
500 static int __init
netvsc_drv_init(void)
502 return vmbus_driver_register(&netvsc_drv
);
505 MODULE_LICENSE("GPL");
506 MODULE_VERSION(HV_DRV_VERSION
);
507 MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
509 module_init(netvsc_drv_init
);
510 module_exit(netvsc_drv_exit
);