VBoxNetFlt-linux.c@ 38050

Last change on this file since 38050 was 38050, checked in by vboxsync, 13 years ago
netflt: RHEL5 fix
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 87.8 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 38050 2011-07-19 08:18:52Z vboxsync $ */
2	/** @file
3	* VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4	*/
5
6	/*
7	* Copyright (C) 2006-2008 Oracle Corporation
8	*
9	* This file is part of VirtualBox Open Source Edition (OSE), as
10	* available from http://www.virtualbox.org. This file is free software;
11	* you can redistribute it and/or modify it under the terms of the GNU
12	* General Public License (GPL) as published by the Free Software
13	* Foundation, in version 2 as it comes in the "COPYING" file of the
14	* VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15	* hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16	*/
17
18	/*******************************************************************************
19	* Header Files *
20	*******************************************************************************/
21	#define LOG_GROUP LOG_GROUP_NET_FLT_DRV
22	#define VBOXNETFLT_LINUX_NO_XMIT_QUEUE
23	#include "the-linux-kernel.h"
24	#include "version-generated.h"
25	#include "product-generated.h"
26	#include <linux/netdevice.h>
27	#include <linux/etherdevice.h>
28	#include <linux/rtnetlink.h>
29	#include <linux/miscdevice.h>
30	#include <linux/ip.h>
31
32	#include <VBox/log.h>
33	#include <VBox/err.h>
34	#include <VBox/intnetinline.h>
35	#include <VBox/vmm/pdmnetinline.h>
36	#include <VBox/param.h>
37	#include <iprt/alloca.h>
38	#include <iprt/assert.h>
39	#include <iprt/spinlock.h>
40	#include <iprt/semaphore.h>
41	#include <iprt/initterm.h>
42	#include <iprt/process.h>
43	#include <iprt/mem.h>
44	#include <iprt/net.h>
45	#include <iprt/log.h>
46	#include <iprt/mp.h>
47	#include <iprt/mem.h>
48	#include <iprt/time.h>
49
50	#define VBOXNETFLT_OS_SPECFIC 1
51	#include "../VBoxNetFltInternal.h"
52
53	#define VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
54	#ifdef CONFIG_NET_SCHED
55	/# define VBOXNETFLT_WITH_QDISC Comment this out to disable qdisc support /
56	# ifdef VBOXNETFLT_WITH_QDISC
57	# include <net/pkt_sched.h>
58	# endif /* VBOXNETFLT_WITH_QDISC */
59	#endif
60
61
62	/*******************************************************************************
63	* Defined Constants And Macros *
64	*******************************************************************************/
65	#define VBOX_FLT_NB_TO_INST(pNB) RT_FROM_MEMBER(pNB, VBOXNETFLTINS, u.s.Notifier)
66	#define VBOX_FLT_PT_TO_INST(pPT) RT_FROM_MEMBER(pPT, VBOXNETFLTINS, u.s.PacketType)
67	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
68	# define VBOX_FLT_XT_TO_INST(pXT) RT_FROM_MEMBER(pXT, VBOXNETFLTINS, u.s.XmitTask)
69	#endif
70
71	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
72	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
73	# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
74	#else
75	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
76	# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
77	#endif
78
79	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 19)
80	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
81	#else
82	# define CHECKSUM_PARTIAL CHECKSUM_HW
83	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 10)
84	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
85	# else
86	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 7)
87	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
88	# else
89	# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
90	# endif
91	/* Versions prior 2.6.10 use stats for both bstats and qstats */
92	# define bstats stats
93	# define qstats stats
94	# endif
95	#endif
96
97	#ifdef VBOXNETFLT_WITH_QDISC
98	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 13)
99	static inline int qdisc_drop(struct sk_buff skb, struct Qdisc sch)
100	{
101	kfree_skb(skb);
102	sch->stats.drops++;
103
104	return NET_XMIT_DROP;
105	}
106	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 13) */
107	#endif /* VBOXNETFLT_WITH_QDISC */
108
109	#ifndef NET_IP_ALIGN
110	# define NET_IP_ALIGN 2
111	#endif
112
113	#if 0
114	/** Create scatter / gather segments for fragments. When not used, we will
115	* linearize the socket buffer before creating the internal networking SG. */
116	# define VBOXNETFLT_SG_SUPPORT 1
117	#endif
118
119	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
120	/** Indicates that the linux kernel may send us GSO frames. */
121	# define VBOXNETFLT_WITH_GSO 1
122
123	/** This enables or disables the transmitting of GSO frame from the internal
124	* network and to the host. */
125	# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
126
127	# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
128	/** This enables or disables the transmitting of GSO frame from the internal
129	* network and to the wire. */
130	# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
131	# endif
132
133	/** This enables or disables the forwarding/flooding of GSO frame from the host
134	* to the internal network. */
135	# define VBOXNETFLT_WITH_GSO_RECV 1
136
137	#endif
138
139	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
140	/** This enables or disables handling of GSO frames coming from the wire (GRO). */
141	# define VBOXNETFLT_WITH_GRO 1
142	#endif
143	/*
144	* GRO support was backported to RHEL 5.4
145	*/
146	#ifdef RHEL_RELEASE_CODE
147	# if RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(5, 4)
148	# define VBOXNETFLT_WITH_GRO 1
149	# endif
150	#endif
151
152	/*******************************************************************************
153	* Internal Functions *
154	*******************************************************************************/
155	static int VBoxNetFltLinuxInit(void);
156	static void VBoxNetFltLinuxUnload(void);
157	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf);
158
159
160	/*******************************************************************************
161	* Global Variables *
162	*******************************************************************************/
163	/**
164	* The (common) global data.
165	*/
166	static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
167
168	module_init(VBoxNetFltLinuxInit);
169	module_exit(VBoxNetFltLinuxUnload);
170
171	MODULE_AUTHOR(VBOX_VENDOR);
172	MODULE_DESCRIPTION(VBOX_PRODUCT " Network Filter Driver");
173	MODULE_LICENSE("GPL");
174	#ifdef MODULE_VERSION
175	MODULE_VERSION(VBOX_VERSION_STRING " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
176	#endif
177
178
179	#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) && defined(LOG_ENABLED)
180	unsigned dev_get_flags(const struct net_device *dev)
181	{
182	unsigned flags;
183
184	flags = (dev->flags & ~(IFF_PROMISC \|
185	IFF_ALLMULTI \|
186	IFF_RUNNING)) \|
187	(dev->gflags & (IFF_PROMISC \|
188	IFF_ALLMULTI));
189
190	if (netif_running(dev) && netif_carrier_ok(dev))
191	flags \|= IFF_RUNNING;
192
193	return flags;
194	}
195	#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) */
196
197
198	#ifdef VBOXNETFLT_WITH_QDISC
199	//#define QDISC_LOG(x) printk x
200	# define QDISC_LOG(x) do { } while (0)
201
202	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 20)
203	# define QDISC_CREATE(dev, queue, ops, parent) qdisc_create_dflt(dev, ops)
204	# elif LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 27)
205	# define QDISC_CREATE(dev, queue, ops, parent) qdisc_create_dflt(dev, ops, parent)
206	# elif LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 37)
207	# define QDISC_CREATE(dev, queue, ops, parent) qdisc_create_dflt(dev, queue, ops, parent)
208	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37) */
209	# define QDISC_CREATE(dev, queue, ops, parent) qdisc_create_dflt(queue, ops, parent)
210	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37) */
211
212	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 27)
213	# define qdisc_dev(qdisc) (qdisc->dev)
214	# define qdisc_pkt_len(skb) (skb->len)
215	# define QDISC_GET(dev) (dev->qdisc_sleeping)
216	# else
217	# define QDISC_GET(dev) (netdev_get_tx_queue(dev, 0)->qdisc_sleeping)
218	# endif
219
220	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 27)
221	# define QDISC_SAVED_NUM(dev) 1
222	# elif LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 32)
223	# define QDISC_SAVED_NUM(dev) dev->num_tx_queues
224	# else
225	# define QDISC_SAVED_NUM(dev) dev->num_tx_queues+1
226	# endif
227
228	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 27)
229	# define QDISC_IS_BUSY(dev, qdisc) test_bit(__LINK_STATE_SCHED, &dev->state)
230	# elif LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36)
231	# define QDISC_IS_BUSY(dev, qdisc) (test_bit(__QDISC_STATE_RUNNING, &qdisc->state) \|\| \
232	test_bit(__QDISC_STATE_SCHED, &qdisc->state))
233	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 36) */
234	# define QDISC_IS_BUSY(dev, qdisc) (qdisc_is_running(qdisc) \|\| \
235	test_bit(__QDISC_STATE_SCHED, &qdisc->state))
236	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 36) */
237
238	struct VBoxNetQDiscPriv
239	{
240	/** Pointer to the single child qdisc. */
241	struct Qdisc *pChild;
242	/*
243	* Technically it is possible to have different qdiscs for different TX
244	* queues so we have to save them all.
245	*/
246	/** Pointer to the array of saved qdiscs. */
247	struct Qdisc **ppSaved;
248	/** Pointer to the net filter instance. */
249	PVBOXNETFLTINS pVBoxNetFlt;
250	};
251	typedef struct VBoxNetQDiscPriv *PVBOXNETQDISCPRIV;
252
253	//#define VBOXNETFLT_QDISC_ENQUEUE
254	static int vboxNetFltQdiscEnqueue(struct sk_buff skb, struct Qdisc sch)
255	{
256	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
257	int rc;
258
259	# ifdef VBOXNETFLT_QDISC_ENQUEUE
260	if (VALID_PTR(pPriv->pVBoxNetFlt))
261	{
262	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
263	PCRTNETETHERHDR pEtherHdr;
264	PINTNETTRUNKSWPORT pSwitchPort;
265	uint32_t cbHdrs = skb_headlen(skb);
266
267	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
268	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(skb, 0, cbHdrs, &abHdrBuf[0]);
269	if ( pEtherHdr
270	&& (pSwitchPort = pPriv->pVBoxNetFlt->pSwitchPort) != NULL
271	&& VALID_PTR(pSwitchPort)
272	&& cbHdrs >= 6)
273	{
274	/** @todo consider reference counting, etc. */
275	INTNETSWDECISION enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
276	if (enmDecision == INTNETSWDECISION_INTNET)
277	{
278	struct sk_buff *pBuf = skb_copy(skb, GFP_ATOMIC);
279	pBuf->pkt_type = PACKET_OUTGOING;
280	vboxNetFltLinuxForwardToIntNet(pPriv->pVBoxNetFlt, pBuf);
281	qdisc_drop(skb, sch);
282	++sch->bstats.packets;
283	sch->bstats.bytes += qdisc_pkt_len(skb);
284	return NET_XMIT_SUCCESS;
285	}
286	}
287	}
288	# endif /* VBOXNETFLT_QDISC_ENQUEUE */
289	rc = pPriv->pChild->enqueue(skb, pPriv->pChild);
290	if (rc == NET_XMIT_SUCCESS)
291	{
292	++sch->q.qlen;
293	++sch->bstats.packets;
294	sch->bstats.bytes += qdisc_pkt_len(skb);
295	}
296	else
297	++sch->qstats.drops;
298	return rc;
299	}
300
301	static struct sk_buff vboxNetFltQdiscDequeue(struct Qdisc sch)
302	{
303	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
304	# ifdef VBOXNETFLT_QDISC_ENQUEUE
305	--sch->q.qlen;
306	return pPriv->pChild->dequeue(pPriv->pChild);
307	# else /* VBOXNETFLT_QDISC_ENQUEUE */
308	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
309	PCRTNETETHERHDR pEtherHdr;
310	PINTNETTRUNKSWPORT pSwitchPort;
311	struct sk_buff *pSkb;
312
313	QDISC_LOG(("vboxNetFltDequeue: Enter pThis=%p\n", pPriv->pVBoxNetFlt));
314
315	while ((pSkb = pPriv->pChild->dequeue(pPriv->pChild)) != NULL)
316	{
317	struct sk_buff *pBuf;
318	INTNETSWDECISION enmDecision;
319	uint32_t cbHdrs;
320
321	--sch->q.qlen;
322
323	if (!VALID_PTR(pPriv->pVBoxNetFlt))
324	break;
325
326	cbHdrs = skb_headlen(pSkb);
327	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
328	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
329	if ( !pEtherHdr
330	\|\| (pSwitchPort = pPriv->pVBoxNetFlt->pSwitchPort) == NULL
331	\|\| !VALID_PTR(pSwitchPort)
332	\|\| cbHdrs < 6)
333	break;
334
335	/** @todo consider reference counting, etc. */
336	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
337	if (enmDecision != INTNETSWDECISION_INTNET)
338	break;
339
340	pBuf = skb_copy(pSkb, GFP_ATOMIC);
341	pBuf->pkt_type = PACKET_OUTGOING;
342	QDISC_LOG(("vboxNetFltDequeue: pThis=%p\n", pPriv->pVBoxNetFlt));
343	vboxNetFltLinuxForwardToIntNet(pPriv->pVBoxNetFlt, pBuf);
344	qdisc_drop(pSkb, sch);
345	QDISC_LOG(("VBoxNetFlt: Packet for %02x:%02x:%02x:%02x:%02x:%02x dropped\n",
346	pSkb->data[0], pSkb->data[1], pSkb->data[2],
347	pSkb->data[3], pSkb->data[4], pSkb->data[5]));
348	}
349
350	return pSkb;
351	# endif /* VBOXNETFLT_QDISC_ENQUEUE */
352	}
353
354	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
355	static int vboxNetFltQdiscRequeue(struct sk_buff skb, struct Qdisc sch)
356	{
357	int rc;
358	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
359
360	rc = pPriv->pChild->ops->requeue(skb, pPriv->pChild);
361	if (rc == 0)
362	{
363	sch->q.qlen++;
364	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 10)
365	sch->qstats.requeues++;
366	# endif
367	}
368
369	return rc;
370	}
371	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
372
373	static unsigned int vboxNetFltQdiscDrop(struct Qdisc *sch)
374	{
375	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
376	unsigned int cbLen;
377
378	if (pPriv->pChild->ops->drop)
379	{
380	cbLen = pPriv->pChild->ops->drop(pPriv->pChild);
381	if (cbLen != 0)
382	{
383	++sch->qstats.drops;
384	--sch->q.qlen;
385	return cbLen;
386	}
387	}
388
389	return 0;
390	}
391
392	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 25)
393	static int vboxNetFltQdiscInit(struct Qdisc sch, struct rtattr opt)
394	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 25) */
395	static int vboxNetFltQdiscInit(struct Qdisc sch, struct nlattr opt)
396	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 25) */
397	{
398	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
399	struct net_device *pDev = qdisc_dev(sch);
400
401	pPriv->pVBoxNetFlt = NULL;
402
403	pPriv->ppSaved = kcalloc(QDISC_SAVED_NUM(pDev), sizeof(pPriv->ppSaved[0]),
404	GFP_KERNEL);
405	if (!pPriv->ppSaved)
406	return -ENOMEM;
407
408	pPriv->pChild = QDISC_CREATE(pDev, netdev_get_tx_queue(pDev, 0),
409	&pfifo_qdisc_ops,
410	TC_H_MAKE(TC_H_MAJ(sch->handle),
411	TC_H_MIN(1)));
412	if (!pPriv->pChild)
413	{
414	kfree(pPriv->ppSaved);
415	pPriv->ppSaved = NULL;
416	return -ENOMEM;
417	}
418
419	return 0;
420	}
421
422	static void vboxNetFltQdiscReset(struct Qdisc *sch)
423	{
424	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
425
426	qdisc_reset(pPriv->pChild);
427	sch->q.qlen = 0;
428	sch->qstats.backlog = 0;
429	}
430
431	static void vboxNetFltQdiscDestroy(struct Qdisc* sch)
432	{
433	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
434	struct net_device *pDev = qdisc_dev(sch);
435
436	qdisc_destroy(pPriv->pChild);
437	pPriv->pChild = NULL;
438
439	if (pPriv->ppSaved)
440	{
441	int i;
442	for (i = 0; i < QDISC_SAVED_NUM(pDev); i++)
443	if (pPriv->ppSaved[i])
444	qdisc_destroy(pPriv->ppSaved[i]);
445	kfree(pPriv->ppSaved);
446	pPriv->ppSaved = NULL;
447	}
448	}
449
450	static int vboxNetFltClassGraft(struct Qdisc sch, unsigned long arg, struct Qdisc pNew,
451	struct Qdisc **ppOld)
452	{
453	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
454
455	if (pNew == NULL)
456	pNew = &noop_qdisc;
457
458	sch_tree_lock(sch);
459	*ppOld = pPriv->pChild;
460	pPriv->pChild = pNew;
461	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 20)
462	sch->q.qlen = 0;
463	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20) */
464	qdisc_tree_decrease_qlen(ppOld, (ppOld)->q.qlen);
465	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20) */
466	qdisc_reset(*ppOld);
467	sch_tree_unlock(sch);
468
469	return 0;
470	}
471
472	static struct Qdisc vboxNetFltClassLeaf(struct Qdisc sch, unsigned long arg)
473	{
474	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
475	return pPriv->pChild;
476	}
477
478	static unsigned long vboxNetFltClassGet(struct Qdisc *sch, u32 classid)
479	{
480	return 1;
481	}
482
483	static void vboxNetFltClassPut(struct Qdisc *sch, unsigned long arg)
484	{
485	}
486
487	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 25)
488	static int vboxNetFltClassChange(struct Qdisc *sch, u32 classid, u32 parentid,
489	struct rtattr *tca, unsigned long arg)
490	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 25) */
491	static int vboxNetFltClassChange(struct Qdisc *sch, u32 classid, u32 parentid,
492	struct nlattr *tca, unsigned long arg)
493	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 25) */
494	{
495	return -ENOSYS;
496	}
497
498	static int vboxNetFltClassDelete(struct Qdisc *sch, unsigned long arg)
499	{
500	return -ENOSYS;
501	}
502
503	static void vboxNetFltClassWalk(struct Qdisc sch, struct qdisc_walker walker)
504	{
505	if (!walker->stop) {
506	if (walker->count >= walker->skip)
507	if (walker->fn(sch, 1, walker) < 0) {
508	walker->stop = 1;
509	return;
510	}
511	walker->count++;
512	}
513	}
514
515	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 32)
516	static struct tcf_proto *vboxNetFltClassFindTcf(struct Qdisc sch, unsigned long cl)
517	{
518	return NULL;
519	}
520	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 32) */
521
522	static int vboxNetFltClassDump(struct Qdisc *sch, unsigned long cl,
523	struct sk_buff skb, struct tcmsg tcm)
524	{
525	PVBOXNETQDISCPRIV pPriv = qdisc_priv(sch);
526
527	if (cl != 1)
528	return -ENOENT;
529
530	tcm->tcm_handle \|= TC_H_MIN(1);
531	tcm->tcm_info = pPriv->pChild->handle;
532
533	return 0;
534	}
535
536
537	static struct Qdisc_class_ops g_VBoxNetFltClassOps =
538	{
539	.graft = vboxNetFltClassGraft,
540	.leaf = vboxNetFltClassLeaf,
541	.get = vboxNetFltClassGet,
542	.put = vboxNetFltClassPut,
543	.change = vboxNetFltClassChange,
544	.delete = vboxNetFltClassDelete,
545	.walk = vboxNetFltClassWalk,
546	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 32)
547	.tcf_chain = vboxNetFltClassFindTcf,
548	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 32) */
549	.dump = vboxNetFltClassDump,
550	};
551
552
553	static struct Qdisc_ops g_VBoxNetFltQDiscOps = {
554	.cl_ops = &g_VBoxNetFltClassOps,
555	.id = "vboxnetflt",
556	.priv_size = sizeof(struct VBoxNetQDiscPriv),
557	.enqueue = vboxNetFltQdiscEnqueue,
558	.dequeue = vboxNetFltQdiscDequeue,
559	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
560	.requeue = vboxNetFltQdiscRequeue,
561	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
562	.peek = qdisc_peek_dequeued,
563	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
564	.drop = vboxNetFltQdiscDrop,
565	.init = vboxNetFltQdiscInit,
566	.reset = vboxNetFltQdiscReset,
567	.destroy = vboxNetFltQdiscDestroy,
568	.owner = THIS_MODULE
569	};
570
571	/*
572	* If our qdisc is already attached to the device (that means the user
573	* installed it from command line with 'tc' command) we simply update
574	* the pointer to vboxnetflt instance in qdisc's private structure.
575	* Otherwise we need to take some additional steps:
576	* - Create our qdisc;
577	* - Save all references to qdiscs;
578	* - Replace our child with the first qdisc reference;
579	* - Replace all references so they point to our qdisc.
580	*/
581	static void vboxNetFltLinuxQdiscInstall(PVBOXNETFLTINS pThis, struct net_device *pDev)
582	{
583	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)
584	int i;
585	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
586	PVBOXNETQDISCPRIV pPriv;
587
588	struct Qdisc *pExisting = QDISC_GET(pDev);
589	/* Do not install our qdisc for devices with no TX queues */
590	if (!pExisting->enqueue)
591	return;
592	if (strcmp(pExisting->ops->id, "vboxnetflt"))
593	{
594	/* The existing qdisc is different from ours, let's create new one. */
595	struct Qdisc *pNew = QDISC_CREATE(pDev, netdev_get_tx_queue(pDev, 0),
596	&g_VBoxNetFltQDiscOps, TC_H_ROOT);
597	if (!pNew)
598	return; // TODO: Error?
599
600	if (!try_module_get(THIS_MODULE))
601	{
602	/*
603	* This may cause a memory leak but calling qdisc_destroy()
604	* is not an option as it will call module_put().
605	*/
606	return;
607	}
608	pPriv = qdisc_priv(pNew);
609
610	qdisc_destroy(pPriv->pChild);
611	pPriv->pChild = QDISC_GET(pDev);
612	atomic_inc(&pPriv->pChild->refcnt);
613	/*
614	* There is no need in deactivating the device or acquiring any locks
615	* prior changing qdiscs since we do not destroy the old qdisc.
616	* Atomic replacement of pointers is enough.
617	*/
618	/*
619	* No need to change reference counters here as we merely move
620	* the pointer and the reference counter of the newly allocated
621	* qdisc is already 1.
622	*/
623	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 27)
624	pPriv->ppSaved[0] = pDev->qdisc_sleeping;
625	ASMAtomicWritePtr(&pDev->qdisc_sleeping, pNew);
626	ASMAtomicWritePtr(&pDev->qdisc, pNew);
627	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
628	for (i = 0; i < pDev->num_tx_queues; i++)
629	{
630	struct netdev_queue *pQueue = netdev_get_tx_queue(pDev, i);
631
632	pPriv->ppSaved[i] = pQueue->qdisc_sleeping;
633	ASMAtomicWritePtr(&pQueue->qdisc_sleeping, pNew);
634	ASMAtomicWritePtr(&pQueue->qdisc, pNew);
635	if (i)
636	atomic_inc(&pNew->refcnt);
637	}
638	/* Newer kernels store root qdisc in netdev structure as well. */
639	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 32)
640	pPriv->ppSaved[pDev->num_tx_queues] = pDev->qdisc;
641	ASMAtomicWritePtr(&pDev->qdisc, pNew);
642	atomic_inc(&pNew->refcnt);
643	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 32) */
644	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
645	/* Sync the queue len with our child */
646	pNew->q.qlen = pPriv->pChild->q.qlen;
647	}
648	else
649	{
650	/* We already have vboxnetflt qdisc, let's use it. */
651	pPriv = qdisc_priv(pExisting);
652	}
653	ASMAtomicWritePtr(&pPriv->pVBoxNetFlt, pThis);
654	QDISC_LOG(("vboxNetFltLinuxInstallQdisc: pThis=%p\n", pPriv->pVBoxNetFlt));
655	}
656
657	static void vboxNetFltLinuxQdiscRemove(PVBOXNETFLTINS pThis, struct net_device *pDev)
658	{
659	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)
660	int i;
661	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
662	PVBOXNETQDISCPRIV pPriv;
663	struct Qdisc pQdisc, pChild;
664	if (!pDev)
665	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
666	if (!VALID_PTR(pDev))
667	{
668	printk("VBoxNetFlt: Failed to detach qdisc, invalid device pointer: %p\n",
669	pDev);
670	return; // TODO: Consider returing an error
671	}
672
673
674	pQdisc = QDISC_GET(pDev);
675	if (strcmp(pQdisc->ops->id, "vboxnetflt"))
676	{
677	if (pQdisc->enqueue)
678	{
679	/* Looks like the user has replaced our qdisc manually. */
680	printk("VBoxNetFlt: Failed to detach qdisc, wrong qdisc: %s\n",
681	pQdisc->ops->id);
682	}
683	return; // TODO: Consider returing an error
684	}
685
686	pPriv = qdisc_priv(pQdisc);
687	Assert(pPriv->pVBoxNetFlt == pThis);
688	ASMAtomicWriteNullPtr(&pPriv->pVBoxNetFlt);
689	pChild = ASMAtomicXchgPtrT(&pPriv->pChild, &noop_qdisc, struct Qdisc *);
690	qdisc_destroy(pChild); /* It won't be the last reference. */
691
692	QDISC_LOG(("vboxNetFltLinuxQdiscRemove: refcnt=%d num_tx_queues=%d\n",
693	atomic_read(&pQdisc->refcnt), pDev->num_tx_queues));
694	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 27)
695	/* Play it safe, make sure the qdisc is not being used. */
696	if (pPriv->ppSaved[0])
697	{
698	ASMAtomicWritePtr(&pDev->qdisc_sleeping, pPriv->ppSaved[0]);
699	ASMAtomicWritePtr(&pDev->qdisc, pPriv->ppSaved[0]);
700	pPriv->ppSaved[0] = NULL;
701	while (QDISC_IS_BUSY(pDev, pQdisc))
702	yield();
703	qdisc_destroy(pQdisc); /* Destroy reference */
704	}
705	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
706	for (i = 0; i < pDev->num_tx_queues; i++)
707	{
708	struct netdev_queue *pQueue = netdev_get_tx_queue(pDev, i);
709	if (pPriv->ppSaved[i])
710	{
711	Assert(pQueue->qdisc_sleeping == pQdisc);
712	ASMAtomicWritePtr(&pQueue->qdisc_sleeping, pPriv->ppSaved[i]);
713	ASMAtomicWritePtr(&pQueue->qdisc, pPriv->ppSaved[i]);
714	pPriv->ppSaved[i] = NULL;
715	while (QDISC_IS_BUSY(pDev, pQdisc))
716	yield();
717	qdisc_destroy(pQdisc); /* Destroy reference */
718	}
719	}
720	/* Newer kernels store root qdisc in netdev structure as well. */
721	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 32)
722	ASMAtomicWritePtr(&pDev->qdisc, pPriv->ppSaved[pDev->num_tx_queues]);
723	pPriv->ppSaved[pDev->num_tx_queues] = NULL;
724	while (QDISC_IS_BUSY(pDev, pQdisc))
725	yield();
726	qdisc_destroy(pQdisc); /* Destroy reference */
727	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 32) */
728	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
729
730	/*
731	* At this point all references to our qdisc should be gone
732	* unless the user had installed it manually.
733	*/
734	QDISC_LOG(("vboxNetFltLinuxRemoveQdisc: pThis=%p\n", pPriv->pVBoxNetFlt));
735	}
736
737	#endif /* VBOXNETFLT_WITH_QDISC */
738
739
740	/**
741	* Initialize module.
742	*
743	* @returns appropriate status code.
744	*/
745	static int __init VBoxNetFltLinuxInit(void)
746	{
747	int rc;
748	/*
749	* Initialize IPRT.
750	*/
751	rc = RTR0Init(0);
752	if (RT_SUCCESS(rc))
753	{
754	Log(("VBoxNetFltLinuxInit\n"));
755
756	/*
757	* Initialize the globals and connect to the support driver.
758	*
759	* This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
760	* for establishing the connect to the support driver.
761	*/
762	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
763	rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
764	if (RT_SUCCESS(rc))
765	{
766	#ifdef VBOXNETFLT_WITH_QDISC
767	/*memcpy(&g_VBoxNetFltQDiscOps, &pfifo_qdisc_ops, sizeof(g_VBoxNetFltQDiscOps));
768	strcpy(g_VBoxNetFltQDiscOps.id, "vboxnetflt");
769	g_VBoxNetFltQDiscOps.owner = THIS_MODULE;*/
770	rc = register_qdisc(&g_VBoxNetFltQDiscOps);
771	if (rc)
772	{
773	LogRel(("VBoxNetFlt: Failed to registered qdisc: %d\n", rc));
774	return rc;
775	}
776	#endif /* VBOXNETFLT_WITH_QDISC */
777	LogRel(("VBoxNetFlt: Successfully started.\n"));
778	return 0;
779	}
780
781	LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
782	RTR0Term();
783	}
784	else
785	LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
786
787	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
788	return -RTErrConvertToErrno(rc);
789	}
790
791
792	/**
793	* Unload the module.
794	*
795	* @todo We have to prevent this if we're busy!
796	*/
797	static void __exit VBoxNetFltLinuxUnload(void)
798	{
799	int rc;
800	Log(("VBoxNetFltLinuxUnload\n"));
801	Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
802
803	#ifdef VBOXNETFLT_WITH_QDISC
804	unregister_qdisc(&g_VBoxNetFltQDiscOps);
805	#endif /* VBOXNETFLT_WITH_QDISC */
806	/*
807	* Undo the work done during start (in reverse order).
808	*/
809	rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
810	AssertRC(rc); NOREF(rc);
811
812	RTR0Term();
813
814	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
815
816	Log(("VBoxNetFltLinuxUnload - done\n"));
817	}
818
819
820	/**
821	* Experiment where we filter traffic from the host to the internal network
822	* before it reaches the NIC driver.
823	*
824	* The current code uses a very ugly hack and only works on kernels using the
825	* net_device_ops (>= 2.6.29). It has been shown to give us a
826	* performance boost of 60-100% though. So, we have to find some less hacky way
827	* of getting this job done eventually.
828	*
829	* #define VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
830	*/
831	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
832
833	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
834
835	# include <linux/ethtool.h>
836
837	typedef struct ethtool_ops OVR_OPSTYPE;
838	# define OVR_OPS ethtool_ops
839	# define OVR_XMIT pfnStartXmit
840
841	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
842
843	typedef struct net_device_ops OVR_OPSTYPE;
844	# define OVR_OPS netdev_ops
845	# define OVR_XMIT pOrgOps->ndo_start_xmit
846
847	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
848
849	/**
850	* The overridden net_device_ops of the device we're attached to.
851	*
852	* As there is no net_device_ops structure in pre-2.6.29 kernels we override
853	* ethtool_ops instead along with hard_start_xmit callback in net_device
854	* structure.
855	*
856	* This is a very dirty hack that was created to explore how much we can improve
857	* the host to guest transfers by not CC'ing the NIC. It turns out to be
858	* the only way to filter outgoing packets for devices without TX queue.
859	*/
860	typedef struct VBoxNetDeviceOpsOverride
861	{
862	/** Our overridden ops. */
863	OVR_OPSTYPE Ops;
864	/** Magic word. */
865	uint32_t u32Magic;
866	/** Pointer to the original ops. */
867	OVR_OPSTYPE const *pOrgOps;
868	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
869	/** Pointer to the original hard_start_xmit function. */
870	int (pfnStartXmit)(struct sk_buff pSkb, struct net_device *pDev);
871	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
872	/** Pointer to the net filter instance. */
873	PVBOXNETFLTINS pVBoxNetFlt;
874	/** The number of filtered packages. */
875	uint64_t cFiltered;
876	/** The total number of packets */
877	uint64_t cTotal;
878	} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
879	/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
880	#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
881
882	/**
883	* ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
884	* because they belong on the internal network.
885	*
886	* @returns NETDEV_TX_XXX.
887	* @param pSkb The socket buffer to transmit.
888	* @param pDev The net device.
889	*/
890	static int vboxNetFltLinuxStartXmitFilter(struct sk_buff pSkb, struct net_device pDev)
891	{
892	PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
893	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
894	PCRTNETETHERHDR pEtherHdr;
895	PINTNETTRUNKSWPORT pSwitchPort;
896	uint32_t cbHdrs;
897
898
899	/*
900	* Validate the override structure.
901	*
902	* Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
903	* to be production quality code, we would have to be much more
904	* careful here and avoid the race.
905	*/
906	if ( !VALID_PTR(pOverride)
907	\|\| pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
908	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) \|\| defined(VBOXNETFLT_WITH_GRO)
909	\|\| !VALID_PTR(pOverride->pOrgOps)
910	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) \|\| defined(VBOXNETFLT_WITH_GRO) */
911	)
912	{
913	printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
914	dev_kfree_skb(pSkb);
915	return NETDEV_TX_OK;
916	}
917	pOverride->cTotal++;
918
919	/*
920	* Do the filtering base on the default OUI of our virtual NICs
921	*
922	* Note! In a real solution, we would ask the switch whether the
923	* destination MAC is 100% to be on the internal network and then
924	* drop it.
925	*/
926	cbHdrs = skb_headlen(pSkb);
927	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
928	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
929	if ( pEtherHdr
930	&& VALID_PTR(pOverride->pVBoxNetFlt)
931	&& (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
932	&& VALID_PTR(pSwitchPort)
933	&& cbHdrs >= 6)
934	{
935	INTNETSWDECISION enmDecision;
936
937	/** @todo consider reference counting, etc. */
938	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
939	if (enmDecision == INTNETSWDECISION_INTNET)
940	{
941	dev_kfree_skb(pSkb);
942	pOverride->cFiltered++;
943	return NETDEV_TX_OK;
944	}
945	}
946
947	return pOverride->OVR_XMIT(pSkb, pDev);
948	}
949
950	/**
951	* Hooks the device ndo_start_xmit operation of the device.
952	*
953	* @param pThis The net filter instance.
954	* @param pDev The net device.
955	*/
956	static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
957	{
958	PVBOXNETDEVICEOPSOVERRIDE pOverride;
959	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
960
961	pOverride = RTMemAlloc(sizeof(*pOverride));
962	if (!pOverride)
963	return;
964	pOverride->pOrgOps = pDev->OVR_OPS;
965	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) && !defined(VBOXNETFLT_WITH_GRO)
966	/**
967	* There is no need to save ethtool_ops structure since we only modify
968	* the pointer itself and the structure is optional (#5712).
969	*/
970	pOverride->pfnStartXmit = pDev->hard_start_xmit;
971	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) \|\| defined(VBOXNETFLT_WITH_GRO) */
972	pOverride->Ops = *pDev->OVR_OPS;
973	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
974	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
975	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
976	pOverride->cTotal = 0;
977	pOverride->cFiltered = 0;
978	pOverride->pVBoxNetFlt = pThis;
979
980	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp); /* (this isn't necessary, but so what) */
981	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
982	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
983	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
984	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
985	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
986	}
987
988	/**
989	* Undos what vboxNetFltLinuxHookDev did.
990	*
991	* @param pThis The net filter instance.
992	* @param pDev The net device. Can be NULL, in which case
993	* we'll try retrieve it from @a pThis.
994	*/
995	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
996	{
997	PVBOXNETDEVICEOPSOVERRIDE pOverride;
998	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
999
1000	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
1001	if (!pDev)
1002	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1003	if (VALID_PTR(pDev))
1004	{
1005	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
1006	if ( VALID_PTR(pOverride)
1007	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
1008	&& VALID_PTR(pOverride->pOrgOps)
1009	)
1010	{
1011	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
1012	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
1013	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
1014	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
1015	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
1016	}
1017	else
1018	pOverride = NULL;
1019	}
1020	else
1021	pOverride = NULL;
1022	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
1023
1024	if (pOverride)
1025	{
1026	printk("vboxnetflt: dropped %llu out of %llu packets\n", pOverride->cFiltered, pOverride->cTotal);
1027	RTMemFree(pOverride);
1028	}
1029	}
1030
1031	#endif /* VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT */
1032
1033
1034	/**
1035	* Reads and retains the host interface handle.
1036	*
1037	* @returns The handle, NULL if detached.
1038	* @param pThis
1039	*/
1040	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
1041	{
1042	#if 0
1043	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1044	struct net_device *pDev = NULL;
1045
1046	Log(("vboxNetFltLinuxRetainNetDev\n"));
1047	/*
1048	* Be careful here to avoid problems racing the detached callback.
1049	*/
1050	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1051	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
1052	{
1053	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
1054	if (pDev)
1055	{
1056	dev_hold(pDev);
1057	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
1058	pDev, pDev->name,
1059	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1060	netdev_refcnt_read(pDev)
1061	#else
1062	atomic_read(&pDev->refcnt)
1063	#endif
1064	));
1065	}
1066	}
1067	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1068
1069	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
1070	return pDev;
1071	#else
1072	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1073	#endif
1074	}
1075
1076
1077	/**
1078	* Release the host interface handle previously retained
1079	* by vboxNetFltLinuxRetainNetDev.
1080	*
1081	* @param pThis The instance.
1082	* @param pDev The vboxNetFltLinuxRetainNetDev
1083	* return value, NULL is fine.
1084	*/
1085	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
1086	{
1087	#if 0
1088	Log(("vboxNetFltLinuxReleaseNetDev\n"));
1089	NOREF(pThis);
1090	if (pDev)
1091	{
1092	dev_put(pDev);
1093	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
1094	pDev, pDev->name,
1095	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1096	netdev_refcnt_read(pDev)
1097	#else
1098	atomic_read(&pDev->refcnt)
1099	#endif
1100	));
1101	}
1102	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
1103	#endif
1104	}
1105
1106	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
1107	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
1108
1109	/**
1110	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
1111	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
1112	*
1113	* @returns true / false accordingly.
1114	* @param pBuf The sk_buff.
1115	*/
1116	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
1117	{
1118	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
1119	}
1120
1121
1122	/**
1123	* Internal worker that create a linux sk_buff for a
1124	* (scatter/)gather list.
1125	*
1126	* @returns Pointer to the sk_buff.
1127	* @param pThis The instance.
1128	* @param pSG The (scatter/)gather list.
1129	* @param fDstWire Set if the destination is the wire.
1130	*/
1131	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
1132	{
1133	struct sk_buff *pPkt;
1134	struct net_device *pDev;
1135	unsigned fGsoType = 0;
1136
1137	if (pSG->cbTotal == 0)
1138	{
1139	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
1140	return NULL;
1141	}
1142
1143	/** @todo We should use fragments mapping the SG buffers with large packets.
1144	* 256 bytes seems to be the a threshold used a lot for this. It
1145	* requires some nasty work on the intnet side though... */
1146	/*
1147	* Allocate a packet and copy over the data.
1148	*/
1149	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1150	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
1151	if (RT_UNLIKELY(!pPkt))
1152	{
1153	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
1154	pSG->pvUserData = NULL;
1155	return NULL;
1156	}
1157	pPkt->dev = pDev;
1158	pPkt->ip_summed = CHECKSUM_NONE;
1159
1160	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
1161	skb_reserve(pPkt, NET_IP_ALIGN);
1162
1163	/* Copy the segments. */
1164	skb_put(pPkt, pSG->cbTotal);
1165	IntNetSgRead(pSG, pPkt->data);
1166
1167	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1168	/*
1169	* Setup GSO if used by this packet.
1170	*/
1171	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
1172	{
1173	default:
1174	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
1175	/* fall thru */
1176	case PDMNETWORKGSOTYPE_INVALID:
1177	fGsoType = 0;
1178	break;
1179	case PDMNETWORKGSOTYPE_IPV4_TCP:
1180	fGsoType = SKB_GSO_TCPV4;
1181	break;
1182	case PDMNETWORKGSOTYPE_IPV4_UDP:
1183	fGsoType = SKB_GSO_UDP;
1184	break;
1185	case PDMNETWORKGSOTYPE_IPV6_TCP:
1186	fGsoType = SKB_GSO_TCPV6;
1187	break;
1188	}
1189	if (fGsoType)
1190	{
1191	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
1192
1193	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
1194	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
1195	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
1196
1197	/*
1198	* We need to set checksum fields even if the packet goes to the host
1199	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
1200	*/
1201	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrs);
1202	pPkt->ip_summed = CHECKSUM_PARTIAL;
1203	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1204	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
1205	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
1206	pPkt->csum_offset = RT_OFFSETOF(RTNETTCP, th_sum);
1207	else
1208	pPkt->csum_offset = RT_OFFSETOF(RTNETUDP, uh_sum);
1209	# else
1210	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
1211	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
1212	pPkt->csum = RT_OFFSETOF(RTNETTCP, th_sum);
1213	else
1214	pPkt->csum = RT_OFFSETOF(RTNETUDP, uh_sum);
1215	# endif
1216	if (!fDstWire)
1217	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
1218	}
1219	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
1220
1221	/*
1222	* Finish up the socket buffer.
1223	*/
1224	pPkt->protocol = eth_type_trans(pPkt, pDev);
1225	if (fDstWire)
1226	{
1227	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
1228
1229	/* Restore ethernet header back. */
1230	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
1231	VBOX_SKB_RESET_MAC_HDR(pPkt);
1232	}
1233	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
1234
1235	return pPkt;
1236	}
1237
1238
1239	/**
1240	* Initializes a SG list from an sk_buff.
1241	*
1242	* @returns Number of segments.
1243	* @param pThis The instance.
1244	* @param pBuf The sk_buff.
1245	* @param pSG The SG.
1246	* @param pvFrame The frame pointer, optional.
1247	* @param cSegs The number of segments allocated for the SG.
1248	* This should match the number in the mbuf exactly!
1249	* @param fSrc The source of the frame.
1250	* @param pGso Pointer to the GSO context if it's a GSO
1251	* internal network frame. NULL if regular frame.
1252	*/
1253	DECLINLINE(void) vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
1254	unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1255	{
1256	int i;
1257	NOREF(pThis);
1258
1259	Assert(!skb_shinfo(pBuf)->frag_list);
1260
1261	if (!pGsoCtx)
1262	IntNetSgInitTempSegs(pSG, pBuf->len, cSegs, 0 /cSegsUsed/);
1263	else
1264	IntNetSgInitTempSegsGso(pSG, pBuf->len, cSegs, 0 /cSegsUsed/, pGsoCtx);
1265
1266	#ifdef VBOXNETFLT_SG_SUPPORT
1267	pSG->aSegs[0].cb = skb_headlen(pBuf);
1268	pSG->aSegs[0].pv = pBuf->data;
1269	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
1270
1271	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1272	{
1273	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
1274	pSG->aSegs[i+1].cb = pFrag->size;
1275	pSG->aSegs[i+1].pv = kmap(pFrag->page);
1276	printk("%p = kmap()\n", pSG->aSegs[i+1].pv);
1277	pSG->aSegs[i+1].Phys = NIL_RTHCPHYS;
1278	}
1279	++i;
1280
1281	#else
1282	pSG->aSegs[0].cb = pBuf->len;
1283	pSG->aSegs[0].pv = pBuf->data;
1284	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
1285	i = 1;
1286	#endif
1287
1288	pSG->cSegsUsed = i;
1289
1290	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1291	/*
1292	* Add a trailer if the frame is too small.
1293	*
1294	* Since we're getting to the packet before it is framed, it has not
1295	* yet been padded. The current solution is to add a segment pointing
1296	* to a buffer containing all zeros and pray that works for all frames...
1297	*/
1298	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
1299	{
1300	static uint8_t const s_abZero[128] = {0};
1301
1302	AssertReturnVoid(i < cSegs);
1303
1304	pSG->aSegs[i].Phys = NIL_RTHCPHYS;
1305	pSG->aSegs[i].pv = (void *)&s_abZero[0];
1306	pSG->aSegs[i].cb = 60 - pSG->cbTotal;
1307	pSG->cbTotal = 60;
1308	pSG->cSegsUsed++;
1309	Assert(i + 1 <= pSG->cSegsAlloc)
1310	}
1311	#endif
1312
1313	Log4(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1314	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1315	for (i = 0; i < pSG->cSegsUsed; i++)
1316	Log4(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1317	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1318	}
1319
1320	/**
1321	* Packet handler,
1322	*
1323	* @returns 0 or EJUSTRETURN.
1324	* @param pThis The instance.
1325	* @param pMBuf The mbuf.
1326	* @param pvFrame The start of the frame, optional.
1327	* @param fSrc Where the packet (allegedly) comes from, one INTNETTRUNKDIR_* value.
1328	* @param eProtocol The protocol.
1329	*/
1330	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 14)
1331	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1332	struct net_device *pSkbDev,
1333	struct packet_type *pPacketType,
1334	struct net_device *pOrigDev)
1335	#else
1336	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1337	struct net_device *pSkbDev,
1338	struct packet_type *pPacketType)
1339	#endif
1340	{
1341	PVBOXNETFLTINS pThis;
1342	struct net_device *pDev;
1343	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1344	pBuf, pSkbDev, pPacketType));
1345	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1346	Log3(("vboxNetFltLinuxPacketHandler: skb len=%u data_len=%u truesize=%u next=%p nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1347	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size, skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1348	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1349	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1350	# endif
1351	#else
1352	Log3(("vboxNetFltLinuxPacketHandler: skb len=%u data_len=%u truesize=%u next=%p nr_frags=%u tso_size=%u tso_seqs=%u frag_list=%p pkt_type=%x\n",
1353	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->tso_size, skb_shinfo(pBuf)->tso_segs, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1354	#endif
1355	/*
1356	* Drop it immediately?
1357	*/
1358	if (!pBuf)
1359	return 0;
1360
1361	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1362	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1363	if (pDev != pSkbDev)
1364	{
1365	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1366	return 0;
1367	}
1368
1369	Log4(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1370	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1371	{
1372	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1373	dev_kfree_skb(pBuf);
1374	return 0;
1375	}
1376
1377	#ifndef VBOXNETFLT_SG_SUPPORT
1378	{
1379	/*
1380	* Get rid of fragmented packets, they cause too much trouble.
1381	*/
1382	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1383	kfree_skb(pBuf);
1384	if (!pCopy)
1385	{
1386	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1387	return 0;
1388	}
1389	pBuf = pCopy;
1390	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1391	Log3(("vboxNetFltLinuxPacketHandler: skb copy len=%u data_len=%u truesize=%u next=%p nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1392	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size, skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1393	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1394	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1395	# endif
1396	# else
1397	Log3(("vboxNetFltLinuxPacketHandler: skb copy len=%u data_len=%u truesize=%u next=%p nr_frags=%u tso_size=%u tso_seqs=%u frag_list=%p pkt_type=%x\n",
1398	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->tso_size, skb_shinfo(pBuf)->tso_segs, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1399	# endif
1400	}
1401	#endif
1402
1403	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1404	/* Forward it to the internal network. */
1405	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1406	#else
1407	/* Add the packet to transmit queue and schedule the bottom half. */
1408	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1409	schedule_work(&pThis->u.s.XmitTask);
1410	Log4(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1411	&pThis->u.s.XmitTask, pBuf));
1412	#endif
1413
1414	/* It does not really matter what we return, it is ignored by the kernel. */
1415	return 0;
1416	}
1417
1418	/**
1419	* Calculate the number of INTNETSEG segments the socket buffer will need.
1420	*
1421	* @returns Segment count.
1422	* @param pBuf The socket buffer.
1423	*/
1424	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff *pBuf)
1425	{
1426	#ifdef VBOXNETFLT_SG_SUPPORT
1427	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1428	#else
1429	unsigned cSegs = 1;
1430	#endif
1431	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1432	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1433	if (pBuf->len < 60)
1434	cSegs++;
1435	#endif
1436	return cSegs;
1437	}
1438
1439	/**
1440	* Destroy the intnet scatter / gather buffer created by
1441	* vboxNetFltLinuxSkBufToSG.
1442	*/
1443	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG)
1444	{
1445	#ifdef VBOXNETFLT_SG_SUPPORT
1446	int i;
1447
1448	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1449	{
1450	printk("kunmap(%p)\n", pSG->aSegs[i+1].pv);
1451	kunmap(pSG->aSegs[i+1].pv);
1452	}
1453	#endif
1454	NOREF(pSG);
1455	}
1456
1457	#ifdef LOG_ENABLED
1458	/**
1459	* Logging helper.
1460	*/
1461	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1462	{
1463	uint8_t pInt, pExt;
1464	static int iPacketNo = 1;
1465	iPacketNo += iIncrement;
1466	if (fEgress)
1467	{
1468	pExt = pSG->aSegs[0].pv;
1469	pInt = pExt + 6;
1470	}
1471	else
1472	{
1473	pInt = pSG->aSegs[0].pv;
1474	pExt = pInt + 6;
1475	}
1476	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1477	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1478	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1479	fEgress ? "-->" : "<--", pszWhere,
1480	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1481	pSG->cbTotal, iPacketNo));
1482	Log3(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1483	}
1484	#else
1485	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1486	#endif
1487
1488	#ifdef VBOXNETFLT_WITH_GSO_RECV
1489
1490	/**
1491	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1492	* GSO socket buffer without having to segment it.
1493	*
1494	* @returns true on success, false if needs segmenting.
1495	* @param pThis The net filter instance.
1496	* @param pSkb The GSO socket buffer.
1497	* @param fSrc The source.
1498	* @param pGsoCtx Where to return the GSO context on success.
1499	*/
1500	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1501	PPDMNETWORKGSO pGsoCtx)
1502	{
1503	PDMNETWORKGSOTYPE enmGsoType;
1504	uint16_t uEtherType;
1505	unsigned int cbTransport;
1506	unsigned int offTransport;
1507	unsigned int cbTransportHdr;
1508	unsigned uProtocol;
1509	union
1510	{
1511	RTNETIPV4 IPv4;
1512	RTNETIPV6 IPv6;
1513	RTNETTCP Tcp;
1514	uint8_t ab[40];
1515	uint16_t au16[40/2];
1516	uint32_t au32[40/4];
1517	} Buf;
1518
1519	/*
1520	* Check the GSO properties of the socket buffer and make sure it fits.
1521	*/
1522	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1523	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_UDP \| SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1524	{
1525	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1526	return false;
1527	}
1528	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1529	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1530	{
1531	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1532	return false;
1533	}
1534	/*
1535	* It is possible to receive GSO packets from wire if GRO is enabled.
1536	*/
1537	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1538	{
1539	Log5(("vboxNetFltLinuxCanForwardAsGso: fSrc=wire\n"));
1540	#ifdef VBOXNETFLT_WITH_GRO
1541	/*
1542	* The packet came from the wire and the driver has already consumed
1543	* mac header. We need to restore it back.
1544	*/
1545	pSkb->mac_len = skb_network_header(pSkb) - skb_mac_header(pSkb);
1546	skb_push(pSkb, pSkb->mac_len);
1547	Log5(("vboxNetFltLinuxCanForwardAsGso: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1548	pSkb->mac_len, pSkb->data, skb_mac_header(pSkb), skb_network_header(pSkb)));
1549	#else /* !VBOXNETFLT_WITH_GRO */
1550	/* Older kernels didn't have GRO. */
1551	return false;
1552	#endif /* !VBOXNETFLT_WITH_GRO */
1553	}
1554	else
1555	{
1556	/*
1557	* skb_gso_segment does the following. Do we need to do it as well?
1558	*/
1559	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1560	skb_reset_mac_header(pSkb);
1561	pSkb->mac_len = pSkb->network_header - pSkb->mac_header;
1562	#else
1563	pSkb->mac.raw = pSkb->data;
1564	pSkb->mac_len = pSkb->nh.raw - pSkb->data;
1565	#endif
1566	}
1567
1568	/*
1569	* Switch on the ethertype.
1570	*/
1571	uEtherType = pSkb->protocol;
1572	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1573	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1574	{
1575	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1576	if (puEtherType)
1577	uEtherType = *puEtherType;
1578	}
1579	switch (uEtherType)
1580	{
1581	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1582	{
1583	unsigned int cbHdr;
1584	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1585	if (RT_UNLIKELY(!pIPv4))
1586	{
1587	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1588	return false;
1589	}
1590
1591	cbHdr = pIPv4->ip_hl * 4;
1592	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1593	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1594	\|\| cbHdr > cbTransport ))
1595	{
1596	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1597	return false;
1598	}
1599	cbTransport -= cbHdr;
1600	offTransport = pSkb->mac_len + cbHdr;
1601	uProtocol = pIPv4->ip_p;
1602	if (uProtocol == RTNETIPV4_PROT_TCP)
1603	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1604	else if (uProtocol == RTNETIPV4_PROT_UDP)
1605	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1606	else /** @todo IPv6: 4to6 tunneling */
1607	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1608	break;
1609	}
1610
1611	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1612	{
1613	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1614	if (RT_UNLIKELY(!pIPv6))
1615	{
1616	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1617	return false;
1618	}
1619
1620	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1621	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1622	uProtocol = pIPv6->ip6_nxt;
1623	/** @todo IPv6: Dig our way out of the other headers. */
1624	if (uProtocol == RTNETIPV4_PROT_TCP)
1625	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1626	else if (uProtocol == RTNETIPV4_PROT_UDP)
1627	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1628	else
1629	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1630	break;
1631	}
1632
1633	default:
1634	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1635	return false;
1636	}
1637
1638	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1639	{
1640	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1641	return false;
1642	}
1643
1644	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1645	\|\| offTransport + cbTransport > pSkb->len
1646	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1647	{
1648	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1649	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1650	return false;
1651	}
1652
1653	/*
1654	* Check the TCP/UDP bits.
1655	*/
1656	if (uProtocol == RTNETIPV4_PROT_TCP)
1657	{
1658	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1659	if (RT_UNLIKELY(!pTcp))
1660	{
1661	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1662	return false;
1663	}
1664
1665	cbTransportHdr = pTcp->th_off * 4;
1666	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1667	\|\| cbTransportHdr > cbTransport
1668	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1669	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1670	{
1671	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1672	return false;
1673	}
1674
1675	}
1676	else
1677	{
1678	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1679	cbTransportHdr = sizeof(RTNETUDP);
1680	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1681	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1682	{
1683	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1684	return false;
1685	}
1686	}
1687
1688	/*
1689	* We're good, init the GSO context.
1690	*/
1691	pGsoCtx->u8Type = enmGsoType;
1692	pGsoCtx->cbHdrs = offTransport + cbTransportHdr;
1693	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1694	pGsoCtx->offHdr1 = pSkb->mac_len;
1695	pGsoCtx->offHdr2 = offTransport;
1696	pGsoCtx->au8Unused[0] = 0;
1697	pGsoCtx->au8Unused[1] = 0;
1698
1699	return true;
1700	}
1701
1702	/**
1703	* Forward the socket buffer as a GSO internal network frame.
1704	*
1705	* @returns IPRT status code.
1706	* @param pThis The net filter instance.
1707	* @param pSkb The GSO socket buffer.
1708	* @param fSrc The source.
1709	* @param pGsoCtx Where to return the GSO context on success.
1710	*/
1711	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1712	{
1713	int rc;
1714	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb);
1715	if (RT_LIKELY(cSegs <= MAX_SKB_FRAGS + 1))
1716	{
1717	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1718	if (RT_LIKELY(pSG))
1719	{
1720	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cSegs, fSrc, pGsoCtx);
1721
1722	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1723	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1724
1725	vboxNetFltLinuxDestroySG(pSG);
1726	rc = VINF_SUCCESS;
1727	}
1728	else
1729	{
1730	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1731	rc = VERR_NO_MEMORY;
1732	}
1733	}
1734	else
1735	{
1736	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1737	rc = VERR_INTERNAL_ERROR_3;
1738	}
1739
1740	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1741	dev_kfree_skb(pSkb);
1742	return rc;
1743	}
1744
1745	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1746
1747	/**
1748	* Worker for vboxNetFltLinuxForwardToIntNet.
1749	*
1750	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1751	* @param pThis The net filter instance.
1752	* @param pBuf The socket buffer.
1753	* @param fSrc The source.
1754	*/
1755	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1756	{
1757	int rc;
1758	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf);
1759	if (cSegs <= MAX_SKB_FRAGS + 1)
1760	{
1761	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1762	if (RT_LIKELY(pSG))
1763	{
1764	if (fSrc & INTNETTRUNKDIR_WIRE)
1765	{
1766	/*
1767	* The packet came from wire, ethernet header was removed by device driver.
1768	* Restore it.
1769	*/
1770	skb_push(pBuf, ETH_HLEN);
1771	}
1772
1773	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cSegs, fSrc, NULL /pGsoCtx/);
1774
1775	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1776	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1777
1778	vboxNetFltLinuxDestroySG(pSG);
1779	rc = VINF_SUCCESS;
1780	}
1781	else
1782	{
1783	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1784	rc = VERR_NO_MEMORY;
1785	}
1786	}
1787	else
1788	{
1789	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1790	rc = VERR_INTERNAL_ERROR_3;
1791	}
1792
1793	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1794	dev_kfree_skb(pBuf);
1795	return rc;
1796	}
1797
1798	/**
1799	*
1800	* @param pBuf The socket buffer. This is consumed by this function.
1801	*/
1802	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1803	{
1804	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1805
1806	#ifdef VBOXNETFLT_WITH_GSO
1807	if (skb_is_gso(pBuf))
1808	{
1809	PDMNETWORKGSO GsoCtx;
1810	Log3(("vboxNetFltLinuxForwardToIntNet: skb len=%u data_len=%u truesize=%u next=%p nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x ip_summed=%d\n",
1811	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size, skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, pBuf->ip_summed));
1812	# ifdef VBOXNETFLT_WITH_GSO_RECV
1813	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_UDP \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1814	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1815	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1816	else
1817	# endif
1818	{
1819	/* Need to segment the packet */
1820	struct sk_buff *pNext;
1821	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1822	if (IS_ERR(pSegment))
1823	{
1824	dev_kfree_skb(pBuf);
1825	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1826	return;
1827	}
1828
1829	for (; pSegment; pSegment = pNext)
1830	{
1831	Log3(("vboxNetFltLinuxForwardToIntNet: segment len=%u data_len=%u truesize=%u next=%p nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1832	pSegment->len, pSegment->data_len, pSegment->truesize, pSegment->next, skb_shinfo(pSegment)->nr_frags, skb_shinfo(pSegment)->gso_size, skb_shinfo(pSegment)->gso_segs, skb_shinfo(pSegment)->gso_type, skb_shinfo(pSegment)->frag_list, pSegment->pkt_type));
1833	pNext = pSegment->next;
1834	pSegment->next = 0;
1835	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1836	}
1837	dev_kfree_skb(pBuf);
1838	}
1839	}
1840	else
1841	#endif /* VBOXNETFLT_WITH_GSO */
1842	{
1843	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1844	{
1845	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1846	/*
1847	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1848	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1849	* header length from the header itself and reconstruct 'h' pointer
1850	* to TCP (or whatever) header.
1851	*/
1852	unsigned char *tmp = pBuf->h.raw;
1853	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1854	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1855	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1856	if (VBOX_SKB_CHECKSUM_HELP(pBuf))
1857	{
1858	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1859	dev_kfree_skb(pBuf);
1860	return;
1861	}
1862	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1863	/* Restore the original (wrong) pointer. */
1864	pBuf->h.raw = tmp;
1865	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1866	}
1867	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1868	}
1869	}
1870
1871	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1872	/**
1873	* Work queue handler that forwards the socket buffers queued by
1874	* vboxNetFltLinuxPacketHandler to the internal network.
1875	*
1876	* @param pWork The work queue.
1877	*/
1878	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
1879	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1880	# else
1881	static void vboxNetFltLinuxXmitTask(void *pWork)
1882	# endif
1883	{
1884	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1885	struct sk_buff *pBuf;
1886
1887	Log4(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1888
1889	/*
1890	* Active? Retain the instance and increment the busy counter.
1891	*/
1892	if (vboxNetFltTryRetainBusyActive(pThis))
1893	{
1894	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1895	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1896
1897	vboxNetFltRelease(pThis, true /* fBusy */);
1898	}
1899	else
1900	{
1901	/** @todo Shouldn't we just drop the packets here? There is little point in
1902	* making them accumulate when the VM is paused and it'll only waste
1903	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1904	* before start draining the packets (goes for the intnet ring buf
1905	* too)? */
1906	}
1907	}
1908	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1909
1910	/**
1911	* Reports the GSO capabilities of the hardware NIC.
1912	*
1913	* @param pThis The net filter instance. The caller hold a
1914	* reference to this.
1915	*/
1916	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1917	{
1918	#ifdef VBOXNETFLT_WITH_GSO_XMIT_WIRE
1919	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1920	{
1921	struct net_device *pDev;
1922	PINTNETTRUNKSWPORT pSwitchPort;
1923	unsigned int fFeatures;
1924	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1925
1926	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
1927
1928	pSwitchPort = pThis->pSwitchPort; /* this doesn't need to be here, but it doesn't harm. */
1929	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1930	if (pDev)
1931	fFeatures = pDev->features;
1932	else
1933	fFeatures = 0;
1934
1935	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
1936
1937	if (pThis->pSwitchPort)
1938	{
1939	/* Set/update the GSO capabilities of the NIC. */
1940	uint32_t fGsoCapabilites = 0;
1941	if (fFeatures & NETIF_F_TSO)
1942	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1943	if (fFeatures & NETIF_F_TSO6)
1944	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1945	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
1946	if (fFeatures & NETIF_F_UFO)
1947	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP);
1948	if (fFeatures & NETIF_F_UFO)
1949	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP);
1950	# endif
1951	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1952	}
1953
1954	vboxNetFltRelease(pThis, true /fBusy/);
1955	}
1956	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE */
1957	}
1958
1959	/**
1960	* Helper that determines whether the host (ignoreing us) is operating the
1961	* interface in promiscuous mode or not.
1962	*/
1963	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1964	{
1965	bool fRc = false;
1966	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1967	if (pDev)
1968	{
1969	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1970	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1971	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1972	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1973	}
1974	return fRc;
1975	}
1976
1977	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 36)
1978	/**
1979	* Helper for detecting TAP devices.
1980	*/
1981	static bool vboxNetFltIsTapDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1982	{
1983	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1984	{
1985	struct ethtool_drvinfo Info;
1986
1987	memset(&Info, 0, sizeof(Info));
1988	Info.cmd = ETHTOOL_GDRVINFO;
1989	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1990	Log3(("vboxNetFltIsTapDevice: driver=%s version=%s bus_info=%s\n",
1991	Info.driver, Info.version, Info.bus_info));
1992
1993	return !strncmp(Info.driver, "tun", 4)
1994	&& !strncmp(Info.bus_info, "tap", 4);
1995	}
1996
1997	return false;
1998	}
1999
2000	/**
2001	* Helper for updating the link state of TAP devices.
2002	* Only TAP devices are affected.
2003	*/
2004	static void vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
2005	{
2006	if (vboxNetFltIsTapDevice(pThis, pDev))
2007	{
2008	Log3(("vboxNetFltSetTapLinkState: bringing %s tap device link state\n",
2009	fLinkUp ? "up" : "down"));
2010	netif_tx_lock_bh(pDev);
2011	if (fLinkUp)
2012	netif_carrier_on(pDev);
2013	else
2014	netif_carrier_off(pDev);
2015	netif_tx_unlock_bh(pDev);
2016	}
2017	}
2018	#else /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
2019	DECLINLINE(void) vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
2020	{
2021	/* Nothing to do for pre-2.6.36 kernels. */
2022	}
2023	#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
2024
2025	/**
2026	* Internal worker for vboxNetFltLinuxNotifierCallback.
2027	*
2028	* @returns VBox status code.
2029	* @param pThis The instance.
2030	* @param fRediscovery If set we're doing a rediscovery attempt, so, don't
2031	* flood the release log.
2032	*/
2033	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
2034	{
2035	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2036	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
2037
2038	/*
2039	* Retain and store the device.
2040	*/
2041	dev_hold(pDev);
2042
2043	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2044	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
2045	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2046
2047	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
2048	pDev, pDev->name,
2049	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2050	netdev_refcnt_read(pDev)
2051	#else
2052	atomic_read(&pDev->refcnt)
2053	#endif
2054	));
2055	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2056	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
2057
2058	/* Get the mac address while we still have a valid net_device reference. */
2059	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
2060
2061	/*
2062	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
2063	*/
2064	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
2065	pThis->u.s.PacketType.dev = pDev;
2066	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
2067	dev_add_pack(&pThis->u.s.PacketType);
2068
2069	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2070	vboxNetFltLinuxHookDev(pThis, pDev);
2071	#endif
2072	#ifdef VBOXNETFLT_WITH_QDISC
2073	vboxNetFltLinuxQdiscInstall(pThis, pDev);
2074	#endif /* VBOXNETFLT_WITH_QDISC */
2075
2076	/*
2077	* If attaching to TAP interface we need to bring the link state up
2078	* starting from 2.6.36 kernel.
2079	*/
2080	vboxNetFltSetTapLinkState(pThis, pDev, true);
2081
2082	/*
2083	* Set indicators that require the spinlock. Be abit paranoid about racing
2084	* the device notification handle.
2085	*/
2086	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2087	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2088	if (pDev)
2089	{
2090	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
2091	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
2092	pDev = NULL; /* don't dereference it */
2093	}
2094	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2095	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
2096
2097	/*
2098	* If the above succeeded report GSO capabilities, if not undo and
2099	* release the device.
2100	*/
2101	if (!pDev)
2102	{
2103	Assert(pThis->pSwitchPort);
2104	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
2105	{
2106	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2107	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
2108	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
2109	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
2110	vboxNetFltRelease(pThis, true /fBusy/);
2111	}
2112	}
2113	else
2114	{
2115	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2116	vboxNetFltLinuxUnhookDev(pThis, pDev);
2117	#endif
2118	#ifdef VBOXNETFLT_WITH_QDISC
2119	vboxNetFltLinuxQdiscRemove(pThis, pDev);
2120	#endif /* VBOXNETFLT_WITH_QDISC */
2121	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2122	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2123	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2124	dev_put(pDev);
2125	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) released. ref=%d\n",
2126	pDev, pDev->name,
2127	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2128	netdev_refcnt_read(pDev)
2129	#else
2130	atomic_read(&pDev->refcnt)
2131	#endif
2132	));
2133	}
2134
2135	LogRel(("VBoxNetFlt: attached to '%s' / %.*Rhxs\n", pThis->szName, sizeof(pThis->u.s.MacAddr), &pThis->u.s.MacAddr));
2136	return VINF_SUCCESS;
2137	}
2138
2139
2140	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
2141	{
2142	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2143
2144	Assert(!pThis->fDisconnectedFromHost);
2145
2146	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2147	vboxNetFltLinuxUnhookDev(pThis, pDev);
2148	#endif
2149	#ifdef VBOXNETFLT_WITH_QDISC
2150	vboxNetFltLinuxQdiscRemove(pThis, pDev);
2151	#endif /* VBOXNETFLT_WITH_QDISC */
2152
2153	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2154	ASMAtomicWriteBool(&pThis->u.s.fRegistered, false);
2155	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
2156	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2157	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2158
2159	dev_remove_pack(&pThis->u.s.PacketType);
2160	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2161	skb_queue_purge(&pThis->u.s.XmitQueue);
2162	#endif
2163	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
2164	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
2165	pDev, pDev->name,
2166	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2167	netdev_refcnt_read(pDev)
2168	#else
2169	atomic_read(&pDev->refcnt)
2170	#endif
2171	));
2172	dev_put(pDev);
2173
2174	return NOTIFY_OK;
2175	}
2176
2177	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
2178	{
2179	/* Check if we are not suspended and promiscuous mode has not been set. */
2180	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
2181	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
2182	{
2183	/* Note that there is no need for locking as the kernel got hold of the lock already. */
2184	dev_set_promiscuity(pDev, 1);
2185	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
2186	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2187	}
2188	else
2189	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2190	return NOTIFY_OK;
2191	}
2192
2193	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
2194	{
2195	/* Undo promiscuous mode if we has set it. */
2196	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
2197	{
2198	/* Note that there is no need for locking as the kernel got hold of the lock already. */
2199	dev_set_promiscuity(pDev, -1);
2200	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
2201	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2202	}
2203	else
2204	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2205	return NOTIFY_OK;
2206	}
2207
2208	#ifdef LOG_ENABLED
2209	/** Stringify the NETDEV_XXX constants. */
2210	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2211	{
2212	const char *pszEvent = "NETDRV_<unknown>";
2213	switch (ulEventType)
2214	{
2215	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2216	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2217	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2218	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2219	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2220	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2221	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2222	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2223	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2224	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2225	# ifdef NETDEV_FEAT_CHANGE
2226	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2227	# endif
2228	}
2229	return pszEvent;
2230	}
2231	#endif /* LOG_ENABLED */
2232
2233	/**
2234	* Callback for listening to netdevice events.
2235	*
2236	* This works the rediscovery, clean up on unregistration, promiscuity on
2237	* up/down, and GSO feature changes from ethtool.
2238	*
2239	* @returns NOTIFY_OK
2240	* @param self Pointer to our notifier registration block.
2241	* @param ulEventType The event.
2242	* @param ptr Event specific, but it is usually the device it
2243	* relates to.
2244	*/
2245	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2246
2247	{
2248	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2249	struct net_device pDev = (struct net_device )ptr;
2250	int rc = NOTIFY_OK;
2251
2252	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2253	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
2254	if ( ulEventType == NETDEV_REGISTER
2255	&& !strcmp(pDev->name, pThis->szName))
2256	{
2257	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2258	}
2259	else
2260	{
2261	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2262	if (pDev == ptr)
2263	{
2264	switch (ulEventType)
2265	{
2266	case NETDEV_UNREGISTER:
2267	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2268	break;
2269	case NETDEV_UP:
2270	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2271	break;
2272	case NETDEV_GOING_DOWN:
2273	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2274	break;
2275	case NETDEV_CHANGENAME:
2276	break;
2277	#ifdef NETDEV_FEAT_CHANGE
2278	case NETDEV_FEAT_CHANGE:
2279	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2280	break;
2281	#endif
2282	}
2283	}
2284	}
2285
2286	return rc;
2287	}
2288
2289	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2290	{
2291	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2292	}
2293
2294	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2295	{
2296	struct net_device * pDev;
2297	int err;
2298	int rc = VINF_SUCCESS;
2299	NOREF(pvIfData);
2300
2301	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2302
2303	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2304	if (pDev)
2305	{
2306	/*
2307	* Create a sk_buff for the gather list and push it onto the wire.
2308	*/
2309	if (fDst & INTNETTRUNKDIR_WIRE)
2310	{
2311	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2312	if (pBuf)
2313	{
2314	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2315	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2316	Log4(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2317	err = dev_queue_xmit(pBuf);
2318	if (err)
2319	rc = RTErrConvertFromErrno(err);
2320	}
2321	else
2322	rc = VERR_NO_MEMORY;
2323	}
2324
2325	/*
2326	* Create a sk_buff for the gather list and push it onto the host stack.
2327	*/
2328	if (fDst & INTNETTRUNKDIR_HOST)
2329	{
2330	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2331	if (pBuf)
2332	{
2333	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2334	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2335	Log4(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2336	err = netif_rx_ni(pBuf);
2337	if (err)
2338	rc = RTErrConvertFromErrno(err);
2339	}
2340	else
2341	rc = VERR_NO_MEMORY;
2342	}
2343
2344	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2345	}
2346
2347	return rc;
2348	}
2349
2350
2351	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2352	{
2353	struct net_device * pDev;
2354
2355	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%s, fDisablePromiscuous=%s\n",
2356	pThis, pThis->szName, fActive?"true":"false",
2357	pThis->fDisablePromiscuous?"true":"false"));
2358
2359	if (pThis->fDisablePromiscuous)
2360	return;
2361
2362	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2363	if (pDev)
2364	{
2365	/*
2366	* This api is a bit weird, the best reference is the code.
2367	*
2368	* Also, we have a bit or race conditions wrt the maintenance of
2369	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2370	*/
2371	#ifdef LOG_ENABLED
2372	u_int16_t fIf;
2373	unsigned const cPromiscBefore = pDev->promiscuity;
2374	#endif
2375	if (fActive)
2376	{
2377	Assert(!pThis->u.s.fPromiscuousSet);
2378
2379	rtnl_lock();
2380	dev_set_promiscuity(pDev, 1);
2381	rtnl_unlock();
2382	pThis->u.s.fPromiscuousSet = true;
2383	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2384	}
2385	else
2386	{
2387	if (pThis->u.s.fPromiscuousSet)
2388	{
2389	rtnl_lock();
2390	dev_set_promiscuity(pDev, -1);
2391	rtnl_unlock();
2392	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2393	}
2394	pThis->u.s.fPromiscuousSet = false;
2395
2396	#ifdef LOG_ENABLED
2397	fIf = dev_get_flags(pDev);
2398	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2399	#endif
2400	}
2401
2402	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2403	}
2404	}
2405
2406
2407	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2408	{
2409	#ifdef VBOXNETFLT_WITH_QDISC
2410	vboxNetFltLinuxQdiscRemove(pThis, NULL);
2411	#endif /* VBOXNETFLT_WITH_QDISC */
2412	/*
2413	* Remove packet handler when we get disconnected from internal switch as
2414	* we don't want the handler to forward packets to disconnected switch.
2415	*/
2416	dev_remove_pack(&pThis->u.s.PacketType);
2417	return VINF_SUCCESS;
2418	}
2419
2420
2421	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2422	{
2423	/*
2424	* Report the GSO capabilities of the host and device (if connected).
2425	* Note! No need to mark ourselves busy here.
2426	*/
2427	/** @todo duplicate work here now? Attach */
2428	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2429	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2430	0
2431	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2432	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2433	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
2434	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP)
2435	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP)
2436	# endif
2437	, INTNETTRUNKDIR_HOST);
2438
2439	#endif
2440	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2441
2442	return VINF_SUCCESS;
2443	}
2444
2445
2446	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2447	{
2448	struct net_device *pDev;
2449	bool fRegistered;
2450	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2451
2452	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2453	vboxNetFltLinuxUnhookDev(pThis, NULL);
2454	#endif
2455
2456	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2457	* unlikely, but none the less). Since it doesn't actually update the
2458	* state (just reads it), it is likely to panic in some interesting
2459	* ways. */
2460
2461	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2462	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2463	fRegistered = ASMAtomicUoReadBool(&pThis->u.s.fRegistered);
2464	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2465
2466	if (fRegistered)
2467	{
2468	vboxNetFltSetTapLinkState(pThis, pDev, false);
2469
2470	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2471	skb_queue_purge(&pThis->u.s.XmitQueue);
2472	#endif
2473	Log(("vboxNetFltOsDeleteInstance: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
2474	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2475	pDev, pDev->name,
2476	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2477	netdev_refcnt_read(pDev)
2478	#else
2479	atomic_read(&pDev->refcnt)
2480	#endif
2481	));
2482	dev_put(pDev);
2483	}
2484	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2485	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2486	module_put(THIS_MODULE);
2487	}
2488
2489
2490	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2491	{
2492	int err;
2493	NOREF(pvContext);
2494
2495	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2496	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2497	if (err)
2498	return VERR_INTNET_FLT_IF_FAILED;
2499	if (!pThis->u.s.fRegistered)
2500	{
2501	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2502	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2503	return VERR_INTNET_FLT_IF_NOT_FOUND;
2504	}
2505
2506	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2507	if ( pThis->fDisconnectedFromHost
2508	\|\| !try_module_get(THIS_MODULE))
2509	return VERR_INTNET_FLT_IF_FAILED;
2510
2511	return VINF_SUCCESS;
2512	}
2513
2514	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2515	{
2516	/*
2517	* Init the linux specific members.
2518	*/
2519	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2520	pThis->u.s.fRegistered = false;
2521	pThis->u.s.fPromiscuousSet = false;
2522	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2523	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2524	skb_queue_head_init(&pThis->u.s.XmitQueue);
2525	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
2526	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2527	# else
2528	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2529	# endif
2530	#endif
2531
2532	return VINF_SUCCESS;
2533	}
2534
2535
2536	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2537	{
2538	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2539	}
2540
2541
2542	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2543	{
2544	/* Nothing to do */
2545	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2546	return VINF_SUCCESS;
2547	}
2548
2549
2550	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2551	{
2552	/* Nothing to do */
2553	NOREF(pThis); NOREF(pvIfData);
2554	return VINF_SUCCESS;
2555	}
2556

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 38050

Download in other formats: