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/module.h>
25 #include <linux/highmem.h>
26 #include <linux/device.h>
28 #include <linux/delay.h>
29 #include <linux/netdevice.h>
30 #include <linux/inetdevice.h>
31 #include <linux/etherdevice.h>
32 #include <linux/skbuff.h>
34 #include <linux/slab.h>
35 #include <linux/dmi.h>
36 #include <linux/pci.h>
38 #include <net/route.h>
40 #include <net/pkt_sched.h>
44 #include "netvsc_api.h"
46 struct net_device_context
{
47 /* point back to our device context */
48 struct hv_device
*device_ctx
;
50 struct work_struct work
;
54 #define PACKET_PAGES_LOWATER 8
55 /* Need this many pages to handle worst case fragmented packet */
56 #define PACKET_PAGES_HIWATER (MAX_SKB_FRAGS + 2)
58 static int ring_size
= 128;
59 module_param(ring_size
, int, S_IRUGO
);
60 MODULE_PARM_DESC(ring_size
, "Ring buffer size (# of pages)");
62 /* no-op so the netdev core doesn't return -EINVAL when modifying the the
63 * multicast address list in SIOCADDMULTI. hv is setup to get all multicast
64 * when it calls RndisFilterOnOpen() */
65 static void netvsc_set_multicast_list(struct net_device
*net
)
69 static int netvsc_open(struct net_device
*net
)
71 struct net_device_context
*net_device_ctx
= netdev_priv(net
);
72 struct hv_device
*device_obj
= net_device_ctx
->device_ctx
;
75 if (netif_carrier_ok(net
)) {
76 /* Open up the device */
77 ret
= rndis_filter_open(device_obj
);
79 netdev_err(net
, "unable to open device (ret %d).\n",
84 netif_start_queue(net
);
86 netdev_err(net
, "unable to open device...link is down.\n");
92 static int netvsc_close(struct net_device
*net
)
94 struct net_device_context
*net_device_ctx
= netdev_priv(net
);
95 struct hv_device
*device_obj
= net_device_ctx
->device_ctx
;
98 netif_stop_queue(net
);
100 ret
= rndis_filter_close(device_obj
);
102 netdev_err(net
, "unable to close device (ret %d).\n", ret
);
107 static void netvsc_xmit_completion(void *context
)
109 struct hv_netvsc_packet
*packet
= (struct hv_netvsc_packet
*)context
;
110 struct sk_buff
*skb
= (struct sk_buff
*)
111 (unsigned long)packet
->completion
.send
.send_completion_tid
;
116 struct net_device
*net
= skb
->dev
;
117 struct net_device_context
*net_device_ctx
= netdev_priv(net
);
118 unsigned int num_pages
= skb_shinfo(skb
)->nr_frags
+ 2;
120 dev_kfree_skb_any(skb
);
122 net_device_ctx
->avail
+= num_pages
;
123 if (net_device_ctx
->avail
>= PACKET_PAGES_HIWATER
)
124 netif_wake_queue(net
);
128 static int netvsc_start_xmit(struct sk_buff
*skb
, struct net_device
*net
)
130 struct net_device_context
*net_device_ctx
= netdev_priv(net
);
131 struct netvsc_driver
*net_drv_obj
=
132 drv_to_netvscdrv(net_device_ctx
->device_ctx
->device
.driver
);
133 struct hv_netvsc_packet
*packet
;
135 unsigned int i
, num_pages
;
137 /* Add 1 for skb->data and additional one for RNDIS */
138 num_pages
= skb_shinfo(skb
)->nr_frags
+ 1 + 1;
139 if (num_pages
> net_device_ctx
->avail
)
140 return NETDEV_TX_BUSY
;
142 /* Allocate a netvsc packet based on # of frags. */
143 packet
= kzalloc(sizeof(struct hv_netvsc_packet
) +
144 (num_pages
* sizeof(struct hv_page_buffer
)) +
145 net_drv_obj
->req_ext_size
, GFP_ATOMIC
);
147 /* out of memory, silently drop packet */
148 netdev_err(net
, "unable to allocate hv_netvsc_packet\n");
151 net
->stats
.tx_dropped
++;
155 packet
->extension
= (void *)(unsigned long)packet
+
156 sizeof(struct hv_netvsc_packet
) +
157 (num_pages
* sizeof(struct hv_page_buffer
));
159 /* Setup the rndis header */
160 packet
->page_buf_cnt
= num_pages
;
162 /* TODO: Flush all write buffers/ memory fence ??? */
165 /* Initialize it from the skb */
166 packet
->total_data_buflen
= skb
->len
;
168 /* Start filling in the page buffers starting after RNDIS buffer. */
169 packet
->page_buf
[1].pfn
= virt_to_phys(skb
->data
) >> PAGE_SHIFT
;
170 packet
->page_buf
[1].offset
171 = (unsigned long)skb
->data
& (PAGE_SIZE
- 1);
172 packet
->page_buf
[1].len
= skb_headlen(skb
);
174 /* Additional fragments are after SKB data */
175 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
176 skb_frag_t
*f
= &skb_shinfo(skb
)->frags
[i
];
178 packet
->page_buf
[i
+2].pfn
= page_to_pfn(f
->page
);
179 packet
->page_buf
[i
+2].offset
= f
->page_offset
;
180 packet
->page_buf
[i
+2].len
= f
->size
;
183 /* Set the completion routine */
184 packet
->completion
.send
.send_completion
= netvsc_xmit_completion
;
185 packet
->completion
.send
.send_completion_ctx
= packet
;
186 packet
->completion
.send
.send_completion_tid
= (unsigned long)skb
;
188 ret
= net_drv_obj
->send(net_device_ctx
->device_ctx
,
191 net
->stats
.tx_bytes
+= skb
->len
;
192 net
->stats
.tx_packets
++;
194 net_device_ctx
->avail
-= num_pages
;
195 if (net_device_ctx
->avail
< PACKET_PAGES_LOWATER
)
196 netif_stop_queue(net
);
198 /* we are shutting down or bus overloaded, just drop packet */
199 net
->stats
.tx_dropped
++;
200 netvsc_xmit_completion(packet
);
207 * netvsc_linkstatus_callback - Link up/down notification
209 static void netvsc_linkstatus_callback(struct hv_device
*device_obj
,
212 struct net_device
*net
= dev_get_drvdata(&device_obj
->device
);
213 struct net_device_context
*ndev_ctx
;
216 netdev_err(net
, "got link status but net device "
217 "not initialized yet\n");
222 netif_carrier_on(net
);
223 netif_wake_queue(net
);
224 netif_notify_peers(net
);
225 ndev_ctx
= netdev_priv(net
);
226 schedule_work(&ndev_ctx
->work
);
228 netif_carrier_off(net
);
229 netif_stop_queue(net
);
234 * netvsc_recv_callback - Callback when we receive a packet from the
235 * "wire" on the specified device.
237 static int netvsc_recv_callback(struct hv_device
*device_obj
,
238 struct hv_netvsc_packet
*packet
)
240 struct net_device
*net
= dev_get_drvdata(&device_obj
->device
);
247 netdev_err(net
, "got receive callback but net device"
248 " not initialized yet\n");
252 /* Allocate a skb - TODO direct I/O to pages? */
253 skb
= netdev_alloc_skb_ip_align(net
, packet
->total_data_buflen
);
254 if (unlikely(!skb
)) {
255 ++net
->stats
.rx_dropped
;
259 /* for kmap_atomic */
260 local_irq_save(flags
);
263 * Copy to skb. This copy is needed here since the memory pointed by
264 * hv_netvsc_packet cannot be deallocated
266 for (i
= 0; i
< packet
->page_buf_cnt
; i
++) {
267 data
= kmap_atomic(pfn_to_page(packet
->page_buf
[i
].pfn
),
269 data
= (void *)(unsigned long)data
+
270 packet
->page_buf
[i
].offset
;
272 memcpy(skb_put(skb
, packet
->page_buf
[i
].len
), data
,
273 packet
->page_buf
[i
].len
);
275 kunmap_atomic((void *)((unsigned long)data
-
276 packet
->page_buf
[i
].offset
), KM_IRQ1
);
279 local_irq_restore(flags
);
281 skb
->protocol
= eth_type_trans(skb
, net
);
282 skb
->ip_summed
= CHECKSUM_NONE
;
284 net
->stats
.rx_packets
++;
285 net
->stats
.rx_bytes
+= skb
->len
;
288 * Pass the skb back up. Network stack will deallocate the skb when it
297 static void netvsc_get_drvinfo(struct net_device
*net
,
298 struct ethtool_drvinfo
*info
)
300 strcpy(info
->driver
, "hv_netvsc");
301 strcpy(info
->version
, HV_DRV_VERSION
);
302 strcpy(info
->fw_version
, "N/A");
305 static const struct ethtool_ops ethtool_ops
= {
306 .get_drvinfo
= netvsc_get_drvinfo
,
307 .get_link
= ethtool_op_get_link
,
310 static const struct net_device_ops device_ops
= {
311 .ndo_open
= netvsc_open
,
312 .ndo_stop
= netvsc_close
,
313 .ndo_start_xmit
= netvsc_start_xmit
,
314 .ndo_set_multicast_list
= netvsc_set_multicast_list
,
315 .ndo_change_mtu
= eth_change_mtu
,
316 .ndo_validate_addr
= eth_validate_addr
,
317 .ndo_set_mac_address
= eth_mac_addr
,
321 * Send GARP packet to network peers after migrations.
322 * After Quick Migration, the network is not immediately operational in the
323 * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
324 * another netif_notify_peers() into a scheduled work, otherwise GARP packet
325 * will not be sent after quick migration, and cause network disconnection.
327 static void netvsc_send_garp(struct work_struct
*w
)
329 struct net_device_context
*ndev_ctx
;
330 struct net_device
*net
;
333 ndev_ctx
= container_of(w
, struct net_device_context
, work
);
334 net
= dev_get_drvdata(&ndev_ctx
->device_ctx
->device
);
335 netif_notify_peers(net
);
339 static int netvsc_probe(struct hv_device
*dev
)
341 struct net_device
*net
= NULL
;
342 struct net_device_context
*net_device_ctx
;
343 struct netvsc_device_info device_info
;
346 net
= alloc_etherdev(sizeof(struct net_device_context
));
350 /* Set initial state */
351 netif_carrier_off(net
);
353 net_device_ctx
= netdev_priv(net
);
354 net_device_ctx
->device_ctx
= dev
;
355 net_device_ctx
->avail
= ring_size
;
356 dev_set_drvdata(&dev
->device
, net
);
357 INIT_WORK(&net_device_ctx
->work
, netvsc_send_garp
);
359 /* Notify the netvsc driver of the new device */
360 ret
= rndis_filte_device_add(dev
, &device_info
);
363 dev_set_drvdata(&dev
->device
, NULL
);
365 netdev_err(net
, "unable to add netvsc device (ret %d)\n", ret
);
370 * If carrier is still off ie we did not get a link status callback,
371 * update it if necessary
374 * FIXME: We should use a atomic or test/set instead to avoid getting
375 * out of sync with the device's link status
377 if (!netif_carrier_ok(net
))
378 if (!device_info
.link_state
)
379 netif_carrier_on(net
);
381 memcpy(net
->dev_addr
, device_info
.mac_adr
, ETH_ALEN
);
383 net
->netdev_ops
= &device_ops
;
385 /* TODO: Add GSO and Checksum offload */
386 net
->hw_features
= NETIF_F_SG
;
387 net
->features
= NETIF_F_SG
;
389 SET_ETHTOOL_OPS(net
, ðtool_ops
);
390 SET_NETDEV_DEV(net
, &dev
->device
);
392 ret
= register_netdev(net
);
394 /* Remove the device and release the resource */
395 rndis_filter_device_remove(dev
);
402 static int netvsc_remove(struct hv_device
*dev
)
404 struct net_device
*net
= dev_get_drvdata(&dev
->device
);
408 dev_err(&dev
->device
, "No net device to remove\n");
412 /* Stop outbound asap */
413 netif_stop_queue(net
);
414 /* netif_carrier_off(net); */
416 unregister_netdev(net
);
419 * Call to the vsc driver to let it know that the device is being
422 ret
= rndis_filter_device_remove(dev
);
425 netdev_err(net
, "unable to remove vsc device (ret %d)\n", ret
);
432 static int netvsc_drv_exit_cb(struct device
*dev
, void *data
)
434 struct device
**curr
= (struct device
**)data
;
441 /* The one and only one */
442 static struct netvsc_driver netvsc_drv
= {
443 .base
.probe
= netvsc_probe
,
444 .base
.remove
= netvsc_remove
,
447 static void netvsc_drv_exit(void)
449 struct hv_driver
*drv
= &netvsc_drv
.base
;
450 struct device
*current_dev
;
457 ret
= driver_for_each_device(&drv
->driver
, NULL
,
458 ¤t_dev
, netvsc_drv_exit_cb
);
460 if (current_dev
== NULL
)
463 /* Initiate removal from the top-down */
464 dev_err(current_dev
, "unregistering device (%s)...\n",
465 dev_name(current_dev
));
467 device_unregister(current_dev
);
470 vmbus_child_driver_unregister(&drv
->driver
);
475 static int netvsc_drv_init(int (*drv_init
)(struct hv_driver
*drv
))
477 struct netvsc_driver
*net_drv_obj
= &netvsc_drv
;
478 struct hv_driver
*drv
= &netvsc_drv
.base
;
481 net_drv_obj
->ring_buf_size
= ring_size
* PAGE_SIZE
;
482 net_drv_obj
->recv_cb
= netvsc_recv_callback
;
483 net_drv_obj
->link_status_change
= netvsc_linkstatus_callback
;
485 /* Callback to client driver to complete the initialization */
486 drv_init(&net_drv_obj
->base
);
488 drv
->driver
.name
= net_drv_obj
->base
.name
;
490 /* The driver belongs to vmbus */
491 ret
= vmbus_child_driver_register(&drv
->driver
);
496 static const struct dmi_system_id __initconst
497 hv_netvsc_dmi_table
[] __maybe_unused
= {
501 DMI_MATCH(DMI_SYS_VENDOR
, "Microsoft Corporation"),
502 DMI_MATCH(DMI_PRODUCT_NAME
, "Virtual Machine"),
503 DMI_MATCH(DMI_BOARD_NAME
, "Virtual Machine"),
508 MODULE_DEVICE_TABLE(dmi
, hv_netvsc_dmi_table
);
510 static int __init
netvsc_init(void)
512 pr_info("initializing....");
514 if (!dmi_check_system(hv_netvsc_dmi_table
))
517 return netvsc_drv_init(netvsc_initialize
);
520 static void __exit
netvsc_exit(void)
525 static const struct pci_device_id __initconst
526 hv_netvsc_pci_table
[] __maybe_unused
= {
527 { PCI_DEVICE(0x1414, 0x5353) }, /* VGA compatible controller */
530 MODULE_DEVICE_TABLE(pci
, hv_netvsc_pci_table
);
532 MODULE_LICENSE("GPL");
533 MODULE_VERSION(HV_DRV_VERSION
);
534 MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
536 module_init(netvsc_init
);
537 module_exit(netvsc_exit
);