VBoxNetFlt-linux.c@ 38054

Last change on this file since 38054 was 38054, checked in by vboxsync, 14 years ago
netflt: fix for regression introduced in r72926
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 87.7 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 38054 2011-07-19 08:50:19Z 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)
909	\|\| !VALID_PTR(pOverride->pOrgOps)
910	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
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	/* We only need to save ethtool_ops structure if it is present (#5712) */
966	if (VALID_PTR(pDev->OVR_OPS))
967	pOverride->Ops = *pDev->OVR_OPS;
968	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
969	pOverride->pfnStartXmit = pDev->hard_start_xmit;
970	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
971	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
972	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
973	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
974	pOverride->cTotal = 0;
975	pOverride->cFiltered = 0;
976	pOverride->pVBoxNetFlt = pThis;
977
978	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp); /* (this isn't necessary, but so what) */
979	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
980	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
981	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
982	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
983	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
984	}
985
986	/**
987	* Undos what vboxNetFltLinuxHookDev did.
988	*
989	* @param pThis The net filter instance.
990	* @param pDev The net device. Can be NULL, in which case
991	* we'll try retrieve it from @a pThis.
992	*/
993	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
994	{
995	PVBOXNETDEVICEOPSOVERRIDE pOverride;
996	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
997
998	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
999	if (!pDev)
1000	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1001	if (VALID_PTR(pDev))
1002	{
1003	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
1004	if ( VALID_PTR(pOverride)
1005	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
1006	&& VALID_PTR(pOverride->pOrgOps)
1007	)
1008	{
1009	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
1010	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
1011	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
1012	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
1013	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
1014	}
1015	else
1016	pOverride = NULL;
1017	}
1018	else
1019	pOverride = NULL;
1020	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
1021
1022	if (pOverride)
1023	{
1024	printk("vboxnetflt: dropped %llu out of %llu packets\n", pOverride->cFiltered, pOverride->cTotal);
1025	RTMemFree(pOverride);
1026	}
1027	}
1028
1029	#endif /* VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT */
1030
1031
1032	/**
1033	* Reads and retains the host interface handle.
1034	*
1035	* @returns The handle, NULL if detached.
1036	* @param pThis
1037	*/
1038	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
1039	{
1040	#if 0
1041	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1042	struct net_device *pDev = NULL;
1043
1044	Log(("vboxNetFltLinuxRetainNetDev\n"));
1045	/*
1046	* Be careful here to avoid problems racing the detached callback.
1047	*/
1048	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1049	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
1050	{
1051	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
1052	if (pDev)
1053	{
1054	dev_hold(pDev);
1055	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
1056	pDev, pDev->name,
1057	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1058	netdev_refcnt_read(pDev)
1059	#else
1060	atomic_read(&pDev->refcnt)
1061	#endif
1062	));
1063	}
1064	}
1065	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1066
1067	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
1068	return pDev;
1069	#else
1070	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1071	#endif
1072	}
1073
1074
1075	/**
1076	* Release the host interface handle previously retained
1077	* by vboxNetFltLinuxRetainNetDev.
1078	*
1079	* @param pThis The instance.
1080	* @param pDev The vboxNetFltLinuxRetainNetDev
1081	* return value, NULL is fine.
1082	*/
1083	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
1084	{
1085	#if 0
1086	Log(("vboxNetFltLinuxReleaseNetDev\n"));
1087	NOREF(pThis);
1088	if (pDev)
1089	{
1090	dev_put(pDev);
1091	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
1092	pDev, pDev->name,
1093	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1094	netdev_refcnt_read(pDev)
1095	#else
1096	atomic_read(&pDev->refcnt)
1097	#endif
1098	));
1099	}
1100	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
1101	#endif
1102	}
1103
1104	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
1105	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
1106
1107	/**
1108	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
1109	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
1110	*
1111	* @returns true / false accordingly.
1112	* @param pBuf The sk_buff.
1113	*/
1114	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
1115	{
1116	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
1117	}
1118
1119
1120	/**
1121	* Internal worker that create a linux sk_buff for a
1122	* (scatter/)gather list.
1123	*
1124	* @returns Pointer to the sk_buff.
1125	* @param pThis The instance.
1126	* @param pSG The (scatter/)gather list.
1127	* @param fDstWire Set if the destination is the wire.
1128	*/
1129	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
1130	{
1131	struct sk_buff *pPkt;
1132	struct net_device *pDev;
1133	unsigned fGsoType = 0;
1134
1135	if (pSG->cbTotal == 0)
1136	{
1137	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
1138	return NULL;
1139	}
1140
1141	/** @todo We should use fragments mapping the SG buffers with large packets.
1142	* 256 bytes seems to be the a threshold used a lot for this. It
1143	* requires some nasty work on the intnet side though... */
1144	/*
1145	* Allocate a packet and copy over the data.
1146	*/
1147	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1148	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
1149	if (RT_UNLIKELY(!pPkt))
1150	{
1151	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
1152	pSG->pvUserData = NULL;
1153	return NULL;
1154	}
1155	pPkt->dev = pDev;
1156	pPkt->ip_summed = CHECKSUM_NONE;
1157
1158	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
1159	skb_reserve(pPkt, NET_IP_ALIGN);
1160
1161	/* Copy the segments. */
1162	skb_put(pPkt, pSG->cbTotal);
1163	IntNetSgRead(pSG, pPkt->data);
1164
1165	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1166	/*
1167	* Setup GSO if used by this packet.
1168	*/
1169	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
1170	{
1171	default:
1172	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
1173	/* fall thru */
1174	case PDMNETWORKGSOTYPE_INVALID:
1175	fGsoType = 0;
1176	break;
1177	case PDMNETWORKGSOTYPE_IPV4_TCP:
1178	fGsoType = SKB_GSO_TCPV4;
1179	break;
1180	case PDMNETWORKGSOTYPE_IPV4_UDP:
1181	fGsoType = SKB_GSO_UDP;
1182	break;
1183	case PDMNETWORKGSOTYPE_IPV6_TCP:
1184	fGsoType = SKB_GSO_TCPV6;
1185	break;
1186	}
1187	if (fGsoType)
1188	{
1189	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
1190
1191	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
1192	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
1193	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
1194
1195	/*
1196	* We need to set checksum fields even if the packet goes to the host
1197	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
1198	*/
1199	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrs);
1200	pPkt->ip_summed = CHECKSUM_PARTIAL;
1201	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1202	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
1203	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
1204	pPkt->csum_offset = RT_OFFSETOF(RTNETTCP, th_sum);
1205	else
1206	pPkt->csum_offset = RT_OFFSETOF(RTNETUDP, uh_sum);
1207	# else
1208	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
1209	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
1210	pPkt->csum = RT_OFFSETOF(RTNETTCP, th_sum);
1211	else
1212	pPkt->csum = RT_OFFSETOF(RTNETUDP, uh_sum);
1213	# endif
1214	if (!fDstWire)
1215	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
1216	}
1217	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
1218
1219	/*
1220	* Finish up the socket buffer.
1221	*/
1222	pPkt->protocol = eth_type_trans(pPkt, pDev);
1223	if (fDstWire)
1224	{
1225	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
1226
1227	/* Restore ethernet header back. */
1228	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
1229	VBOX_SKB_RESET_MAC_HDR(pPkt);
1230	}
1231	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
1232
1233	return pPkt;
1234	}
1235
1236
1237	/**
1238	* Initializes a SG list from an sk_buff.
1239	*
1240	* @returns Number of segments.
1241	* @param pThis The instance.
1242	* @param pBuf The sk_buff.
1243	* @param pSG The SG.
1244	* @param pvFrame The frame pointer, optional.
1245	* @param cSegs The number of segments allocated for the SG.
1246	* This should match the number in the mbuf exactly!
1247	* @param fSrc The source of the frame.
1248	* @param pGso Pointer to the GSO context if it's a GSO
1249	* internal network frame. NULL if regular frame.
1250	*/
1251	DECLINLINE(void) vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
1252	unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1253	{
1254	int i;
1255	NOREF(pThis);
1256
1257	Assert(!skb_shinfo(pBuf)->frag_list);
1258
1259	if (!pGsoCtx)
1260	IntNetSgInitTempSegs(pSG, pBuf->len, cSegs, 0 /cSegsUsed/);
1261	else
1262	IntNetSgInitTempSegsGso(pSG, pBuf->len, cSegs, 0 /cSegsUsed/, pGsoCtx);
1263
1264	#ifdef VBOXNETFLT_SG_SUPPORT
1265	pSG->aSegs[0].cb = skb_headlen(pBuf);
1266	pSG->aSegs[0].pv = pBuf->data;
1267	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
1268
1269	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1270	{
1271	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
1272	pSG->aSegs[i+1].cb = pFrag->size;
1273	pSG->aSegs[i+1].pv = kmap(pFrag->page);
1274	printk("%p = kmap()\n", pSG->aSegs[i+1].pv);
1275	pSG->aSegs[i+1].Phys = NIL_RTHCPHYS;
1276	}
1277	++i;
1278
1279	#else
1280	pSG->aSegs[0].cb = pBuf->len;
1281	pSG->aSegs[0].pv = pBuf->data;
1282	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
1283	i = 1;
1284	#endif
1285
1286	pSG->cSegsUsed = i;
1287
1288	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1289	/*
1290	* Add a trailer if the frame is too small.
1291	*
1292	* Since we're getting to the packet before it is framed, it has not
1293	* yet been padded. The current solution is to add a segment pointing
1294	* to a buffer containing all zeros and pray that works for all frames...
1295	*/
1296	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
1297	{
1298	static uint8_t const s_abZero[128] = {0};
1299
1300	AssertReturnVoid(i < cSegs);
1301
1302	pSG->aSegs[i].Phys = NIL_RTHCPHYS;
1303	pSG->aSegs[i].pv = (void *)&s_abZero[0];
1304	pSG->aSegs[i].cb = 60 - pSG->cbTotal;
1305	pSG->cbTotal = 60;
1306	pSG->cSegsUsed++;
1307	Assert(i + 1 <= pSG->cSegsAlloc)
1308	}
1309	#endif
1310
1311	Log4(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1312	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1313	for (i = 0; i < pSG->cSegsUsed; i++)
1314	Log4(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1315	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1316	}
1317
1318	/**
1319	* Packet handler,
1320	*
1321	* @returns 0 or EJUSTRETURN.
1322	* @param pThis The instance.
1323	* @param pMBuf The mbuf.
1324	* @param pvFrame The start of the frame, optional.
1325	* @param fSrc Where the packet (allegedly) comes from, one INTNETTRUNKDIR_* value.
1326	* @param eProtocol The protocol.
1327	*/
1328	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 14)
1329	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1330	struct net_device *pSkbDev,
1331	struct packet_type *pPacketType,
1332	struct net_device *pOrigDev)
1333	#else
1334	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1335	struct net_device *pSkbDev,
1336	struct packet_type *pPacketType)
1337	#endif
1338	{
1339	PVBOXNETFLTINS pThis;
1340	struct net_device *pDev;
1341	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1342	pBuf, pSkbDev, pPacketType));
1343	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1344	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",
1345	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));
1346	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1347	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1348	# endif
1349	#else
1350	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",
1351	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));
1352	#endif
1353	/*
1354	* Drop it immediately?
1355	*/
1356	if (!pBuf)
1357	return 0;
1358
1359	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1360	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1361	if (pDev != pSkbDev)
1362	{
1363	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1364	return 0;
1365	}
1366
1367	Log4(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1368	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1369	{
1370	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1371	dev_kfree_skb(pBuf);
1372	return 0;
1373	}
1374
1375	#ifndef VBOXNETFLT_SG_SUPPORT
1376	{
1377	/*
1378	* Get rid of fragmented packets, they cause too much trouble.
1379	*/
1380	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1381	kfree_skb(pBuf);
1382	if (!pCopy)
1383	{
1384	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1385	return 0;
1386	}
1387	pBuf = pCopy;
1388	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1389	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",
1390	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));
1391	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1392	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1393	# endif
1394	# else
1395	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",
1396	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));
1397	# endif
1398	}
1399	#endif
1400
1401	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1402	/* Forward it to the internal network. */
1403	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1404	#else
1405	/* Add the packet to transmit queue and schedule the bottom half. */
1406	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1407	schedule_work(&pThis->u.s.XmitTask);
1408	Log4(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1409	&pThis->u.s.XmitTask, pBuf));
1410	#endif
1411
1412	/* It does not really matter what we return, it is ignored by the kernel. */
1413	return 0;
1414	}
1415
1416	/**
1417	* Calculate the number of INTNETSEG segments the socket buffer will need.
1418	*
1419	* @returns Segment count.
1420	* @param pBuf The socket buffer.
1421	*/
1422	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff *pBuf)
1423	{
1424	#ifdef VBOXNETFLT_SG_SUPPORT
1425	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1426	#else
1427	unsigned cSegs = 1;
1428	#endif
1429	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1430	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1431	if (pBuf->len < 60)
1432	cSegs++;
1433	#endif
1434	return cSegs;
1435	}
1436
1437	/**
1438	* Destroy the intnet scatter / gather buffer created by
1439	* vboxNetFltLinuxSkBufToSG.
1440	*/
1441	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG)
1442	{
1443	#ifdef VBOXNETFLT_SG_SUPPORT
1444	int i;
1445
1446	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1447	{
1448	printk("kunmap(%p)\n", pSG->aSegs[i+1].pv);
1449	kunmap(pSG->aSegs[i+1].pv);
1450	}
1451	#endif
1452	NOREF(pSG);
1453	}
1454
1455	#ifdef LOG_ENABLED
1456	/**
1457	* Logging helper.
1458	*/
1459	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1460	{
1461	uint8_t pInt, pExt;
1462	static int iPacketNo = 1;
1463	iPacketNo += iIncrement;
1464	if (fEgress)
1465	{
1466	pExt = pSG->aSegs[0].pv;
1467	pInt = pExt + 6;
1468	}
1469	else
1470	{
1471	pInt = pSG->aSegs[0].pv;
1472	pExt = pInt + 6;
1473	}
1474	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1475	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1476	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1477	fEgress ? "-->" : "<--", pszWhere,
1478	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1479	pSG->cbTotal, iPacketNo));
1480	Log3(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1481	}
1482	#else
1483	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1484	#endif
1485
1486	#ifdef VBOXNETFLT_WITH_GSO_RECV
1487
1488	/**
1489	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1490	* GSO socket buffer without having to segment it.
1491	*
1492	* @returns true on success, false if needs segmenting.
1493	* @param pThis The net filter instance.
1494	* @param pSkb The GSO socket buffer.
1495	* @param fSrc The source.
1496	* @param pGsoCtx Where to return the GSO context on success.
1497	*/
1498	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1499	PPDMNETWORKGSO pGsoCtx)
1500	{
1501	PDMNETWORKGSOTYPE enmGsoType;
1502	uint16_t uEtherType;
1503	unsigned int cbTransport;
1504	unsigned int offTransport;
1505	unsigned int cbTransportHdr;
1506	unsigned uProtocol;
1507	union
1508	{
1509	RTNETIPV4 IPv4;
1510	RTNETIPV6 IPv6;
1511	RTNETTCP Tcp;
1512	uint8_t ab[40];
1513	uint16_t au16[40/2];
1514	uint32_t au32[40/4];
1515	} Buf;
1516
1517	/*
1518	* Check the GSO properties of the socket buffer and make sure it fits.
1519	*/
1520	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1521	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_UDP \| SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1522	{
1523	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1524	return false;
1525	}
1526	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1527	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1528	{
1529	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1530	return false;
1531	}
1532	/*
1533	* It is possible to receive GSO packets from wire if GRO is enabled.
1534	*/
1535	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1536	{
1537	Log5(("vboxNetFltLinuxCanForwardAsGso: fSrc=wire\n"));
1538	#ifdef VBOXNETFLT_WITH_GRO
1539	/*
1540	* The packet came from the wire and the driver has already consumed
1541	* mac header. We need to restore it back.
1542	*/
1543	pSkb->mac_len = skb_network_header(pSkb) - skb_mac_header(pSkb);
1544	skb_push(pSkb, pSkb->mac_len);
1545	Log5(("vboxNetFltLinuxCanForwardAsGso: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1546	pSkb->mac_len, pSkb->data, skb_mac_header(pSkb), skb_network_header(pSkb)));
1547	#else /* !VBOXNETFLT_WITH_GRO */
1548	/* Older kernels didn't have GRO. */
1549	return false;
1550	#endif /* !VBOXNETFLT_WITH_GRO */
1551	}
1552	else
1553	{
1554	/*
1555	* skb_gso_segment does the following. Do we need to do it as well?
1556	*/
1557	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1558	skb_reset_mac_header(pSkb);
1559	pSkb->mac_len = pSkb->network_header - pSkb->mac_header;
1560	#else
1561	pSkb->mac.raw = pSkb->data;
1562	pSkb->mac_len = pSkb->nh.raw - pSkb->data;
1563	#endif
1564	}
1565
1566	/*
1567	* Switch on the ethertype.
1568	*/
1569	uEtherType = pSkb->protocol;
1570	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1571	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1572	{
1573	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1574	if (puEtherType)
1575	uEtherType = *puEtherType;
1576	}
1577	switch (uEtherType)
1578	{
1579	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1580	{
1581	unsigned int cbHdr;
1582	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1583	if (RT_UNLIKELY(!pIPv4))
1584	{
1585	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1586	return false;
1587	}
1588
1589	cbHdr = pIPv4->ip_hl * 4;
1590	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1591	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1592	\|\| cbHdr > cbTransport ))
1593	{
1594	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1595	return false;
1596	}
1597	cbTransport -= cbHdr;
1598	offTransport = pSkb->mac_len + cbHdr;
1599	uProtocol = pIPv4->ip_p;
1600	if (uProtocol == RTNETIPV4_PROT_TCP)
1601	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1602	else if (uProtocol == RTNETIPV4_PROT_UDP)
1603	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1604	else /** @todo IPv6: 4to6 tunneling */
1605	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1606	break;
1607	}
1608
1609	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1610	{
1611	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1612	if (RT_UNLIKELY(!pIPv6))
1613	{
1614	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1615	return false;
1616	}
1617
1618	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1619	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1620	uProtocol = pIPv6->ip6_nxt;
1621	/** @todo IPv6: Dig our way out of the other headers. */
1622	if (uProtocol == RTNETIPV4_PROT_TCP)
1623	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1624	else if (uProtocol == RTNETIPV4_PROT_UDP)
1625	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1626	else
1627	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1628	break;
1629	}
1630
1631	default:
1632	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1633	return false;
1634	}
1635
1636	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1637	{
1638	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1639	return false;
1640	}
1641
1642	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1643	\|\| offTransport + cbTransport > pSkb->len
1644	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1645	{
1646	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1647	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1648	return false;
1649	}
1650
1651	/*
1652	* Check the TCP/UDP bits.
1653	*/
1654	if (uProtocol == RTNETIPV4_PROT_TCP)
1655	{
1656	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1657	if (RT_UNLIKELY(!pTcp))
1658	{
1659	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1660	return false;
1661	}
1662
1663	cbTransportHdr = pTcp->th_off * 4;
1664	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1665	\|\| cbTransportHdr > cbTransport
1666	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1667	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1668	{
1669	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1670	return false;
1671	}
1672
1673	}
1674	else
1675	{
1676	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1677	cbTransportHdr = sizeof(RTNETUDP);
1678	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1679	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1680	{
1681	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1682	return false;
1683	}
1684	}
1685
1686	/*
1687	* We're good, init the GSO context.
1688	*/
1689	pGsoCtx->u8Type = enmGsoType;
1690	pGsoCtx->cbHdrs = offTransport + cbTransportHdr;
1691	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1692	pGsoCtx->offHdr1 = pSkb->mac_len;
1693	pGsoCtx->offHdr2 = offTransport;
1694	pGsoCtx->au8Unused[0] = 0;
1695	pGsoCtx->au8Unused[1] = 0;
1696
1697	return true;
1698	}
1699
1700	/**
1701	* Forward the socket buffer as a GSO internal network frame.
1702	*
1703	* @returns IPRT status code.
1704	* @param pThis The net filter instance.
1705	* @param pSkb The GSO socket buffer.
1706	* @param fSrc The source.
1707	* @param pGsoCtx Where to return the GSO context on success.
1708	*/
1709	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1710	{
1711	int rc;
1712	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb);
1713	if (RT_LIKELY(cSegs <= MAX_SKB_FRAGS + 1))
1714	{
1715	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1716	if (RT_LIKELY(pSG))
1717	{
1718	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cSegs, fSrc, pGsoCtx);
1719
1720	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1721	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1722
1723	vboxNetFltLinuxDestroySG(pSG);
1724	rc = VINF_SUCCESS;
1725	}
1726	else
1727	{
1728	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1729	rc = VERR_NO_MEMORY;
1730	}
1731	}
1732	else
1733	{
1734	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1735	rc = VERR_INTERNAL_ERROR_3;
1736	}
1737
1738	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1739	dev_kfree_skb(pSkb);
1740	return rc;
1741	}
1742
1743	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1744
1745	/**
1746	* Worker for vboxNetFltLinuxForwardToIntNet.
1747	*
1748	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1749	* @param pThis The net filter instance.
1750	* @param pBuf The socket buffer.
1751	* @param fSrc The source.
1752	*/
1753	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1754	{
1755	int rc;
1756	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf);
1757	if (cSegs <= MAX_SKB_FRAGS + 1)
1758	{
1759	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1760	if (RT_LIKELY(pSG))
1761	{
1762	if (fSrc & INTNETTRUNKDIR_WIRE)
1763	{
1764	/*
1765	* The packet came from wire, ethernet header was removed by device driver.
1766	* Restore it.
1767	*/
1768	skb_push(pBuf, ETH_HLEN);
1769	}
1770
1771	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cSegs, fSrc, NULL /pGsoCtx/);
1772
1773	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1774	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1775
1776	vboxNetFltLinuxDestroySG(pSG);
1777	rc = VINF_SUCCESS;
1778	}
1779	else
1780	{
1781	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1782	rc = VERR_NO_MEMORY;
1783	}
1784	}
1785	else
1786	{
1787	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1788	rc = VERR_INTERNAL_ERROR_3;
1789	}
1790
1791	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1792	dev_kfree_skb(pBuf);
1793	return rc;
1794	}
1795
1796	/**
1797	*
1798	* @param pBuf The socket buffer. This is consumed by this function.
1799	*/
1800	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1801	{
1802	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1803
1804	#ifdef VBOXNETFLT_WITH_GSO
1805	if (skb_is_gso(pBuf))
1806	{
1807	PDMNETWORKGSO GsoCtx;
1808	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",
1809	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));
1810	# ifdef VBOXNETFLT_WITH_GSO_RECV
1811	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_UDP \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1812	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1813	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1814	else
1815	# endif
1816	{
1817	/* Need to segment the packet */
1818	struct sk_buff *pNext;
1819	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1820	if (IS_ERR(pSegment))
1821	{
1822	dev_kfree_skb(pBuf);
1823	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1824	return;
1825	}
1826
1827	for (; pSegment; pSegment = pNext)
1828	{
1829	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",
1830	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));
1831	pNext = pSegment->next;
1832	pSegment->next = 0;
1833	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1834	}
1835	dev_kfree_skb(pBuf);
1836	}
1837	}
1838	else
1839	#endif /* VBOXNETFLT_WITH_GSO */
1840	{
1841	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1842	{
1843	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1844	/*
1845	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1846	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1847	* header length from the header itself and reconstruct 'h' pointer
1848	* to TCP (or whatever) header.
1849	*/
1850	unsigned char *tmp = pBuf->h.raw;
1851	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1852	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1853	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1854	if (VBOX_SKB_CHECKSUM_HELP(pBuf))
1855	{
1856	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1857	dev_kfree_skb(pBuf);
1858	return;
1859	}
1860	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1861	/* Restore the original (wrong) pointer. */
1862	pBuf->h.raw = tmp;
1863	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1864	}
1865	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1866	}
1867	}
1868
1869	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1870	/**
1871	* Work queue handler that forwards the socket buffers queued by
1872	* vboxNetFltLinuxPacketHandler to the internal network.
1873	*
1874	* @param pWork The work queue.
1875	*/
1876	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
1877	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1878	# else
1879	static void vboxNetFltLinuxXmitTask(void *pWork)
1880	# endif
1881	{
1882	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1883	struct sk_buff *pBuf;
1884
1885	Log4(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1886
1887	/*
1888	* Active? Retain the instance and increment the busy counter.
1889	*/
1890	if (vboxNetFltTryRetainBusyActive(pThis))
1891	{
1892	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1893	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1894
1895	vboxNetFltRelease(pThis, true /* fBusy */);
1896	}
1897	else
1898	{
1899	/** @todo Shouldn't we just drop the packets here? There is little point in
1900	* making them accumulate when the VM is paused and it'll only waste
1901	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1902	* before start draining the packets (goes for the intnet ring buf
1903	* too)? */
1904	}
1905	}
1906	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1907
1908	/**
1909	* Reports the GSO capabilities of the hardware NIC.
1910	*
1911	* @param pThis The net filter instance. The caller hold a
1912	* reference to this.
1913	*/
1914	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1915	{
1916	#ifdef VBOXNETFLT_WITH_GSO_XMIT_WIRE
1917	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1918	{
1919	struct net_device *pDev;
1920	PINTNETTRUNKSWPORT pSwitchPort;
1921	unsigned int fFeatures;
1922	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1923
1924	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
1925
1926	pSwitchPort = pThis->pSwitchPort; /* this doesn't need to be here, but it doesn't harm. */
1927	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1928	if (pDev)
1929	fFeatures = pDev->features;
1930	else
1931	fFeatures = 0;
1932
1933	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
1934
1935	if (pThis->pSwitchPort)
1936	{
1937	/* Set/update the GSO capabilities of the NIC. */
1938	uint32_t fGsoCapabilites = 0;
1939	if (fFeatures & NETIF_F_TSO)
1940	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1941	if (fFeatures & NETIF_F_TSO6)
1942	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1943	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
1944	if (fFeatures & NETIF_F_UFO)
1945	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP);
1946	if (fFeatures & NETIF_F_UFO)
1947	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP);
1948	# endif
1949	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1950	}
1951
1952	vboxNetFltRelease(pThis, true /fBusy/);
1953	}
1954	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE */
1955	}
1956
1957	/**
1958	* Helper that determines whether the host (ignoreing us) is operating the
1959	* interface in promiscuous mode or not.
1960	*/
1961	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1962	{
1963	bool fRc = false;
1964	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1965	if (pDev)
1966	{
1967	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1968	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1969	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1970	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1971	}
1972	return fRc;
1973	}
1974
1975	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 36)
1976	/**
1977	* Helper for detecting TAP devices.
1978	*/
1979	static bool vboxNetFltIsTapDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1980	{
1981	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1982	{
1983	struct ethtool_drvinfo Info;
1984
1985	memset(&Info, 0, sizeof(Info));
1986	Info.cmd = ETHTOOL_GDRVINFO;
1987	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1988	Log3(("vboxNetFltIsTapDevice: driver=%s version=%s bus_info=%s\n",
1989	Info.driver, Info.version, Info.bus_info));
1990
1991	return !strncmp(Info.driver, "tun", 4)
1992	&& !strncmp(Info.bus_info, "tap", 4);
1993	}
1994
1995	return false;
1996	}
1997
1998	/**
1999	* Helper for updating the link state of TAP devices.
2000	* Only TAP devices are affected.
2001	*/
2002	static void vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
2003	{
2004	if (vboxNetFltIsTapDevice(pThis, pDev))
2005	{
2006	Log3(("vboxNetFltSetTapLinkState: bringing %s tap device link state\n",
2007	fLinkUp ? "up" : "down"));
2008	netif_tx_lock_bh(pDev);
2009	if (fLinkUp)
2010	netif_carrier_on(pDev);
2011	else
2012	netif_carrier_off(pDev);
2013	netif_tx_unlock_bh(pDev);
2014	}
2015	}
2016	#else /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
2017	DECLINLINE(void) vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
2018	{
2019	/* Nothing to do for pre-2.6.36 kernels. */
2020	}
2021	#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
2022
2023	/**
2024	* Internal worker for vboxNetFltLinuxNotifierCallback.
2025	*
2026	* @returns VBox status code.
2027	* @param pThis The instance.
2028	* @param fRediscovery If set we're doing a rediscovery attempt, so, don't
2029	* flood the release log.
2030	*/
2031	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
2032	{
2033	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2034	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
2035
2036	/*
2037	* Retain and store the device.
2038	*/
2039	dev_hold(pDev);
2040
2041	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2042	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
2043	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2044
2045	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
2046	pDev, pDev->name,
2047	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2048	netdev_refcnt_read(pDev)
2049	#else
2050	atomic_read(&pDev->refcnt)
2051	#endif
2052	));
2053	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2054	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
2055
2056	/* Get the mac address while we still have a valid net_device reference. */
2057	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
2058
2059	/*
2060	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
2061	*/
2062	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
2063	pThis->u.s.PacketType.dev = pDev;
2064	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
2065	dev_add_pack(&pThis->u.s.PacketType);
2066
2067	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2068	vboxNetFltLinuxHookDev(pThis, pDev);
2069	#endif
2070	#ifdef VBOXNETFLT_WITH_QDISC
2071	vboxNetFltLinuxQdiscInstall(pThis, pDev);
2072	#endif /* VBOXNETFLT_WITH_QDISC */
2073
2074	/*
2075	* If attaching to TAP interface we need to bring the link state up
2076	* starting from 2.6.36 kernel.
2077	*/
2078	vboxNetFltSetTapLinkState(pThis, pDev, true);
2079
2080	/*
2081	* Set indicators that require the spinlock. Be abit paranoid about racing
2082	* the device notification handle.
2083	*/
2084	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2085	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2086	if (pDev)
2087	{
2088	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
2089	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
2090	pDev = NULL; /* don't dereference it */
2091	}
2092	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2093	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
2094
2095	/*
2096	* If the above succeeded report GSO capabilities, if not undo and
2097	* release the device.
2098	*/
2099	if (!pDev)
2100	{
2101	Assert(pThis->pSwitchPort);
2102	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
2103	{
2104	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2105	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
2106	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
2107	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
2108	vboxNetFltRelease(pThis, true /fBusy/);
2109	}
2110	}
2111	else
2112	{
2113	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2114	vboxNetFltLinuxUnhookDev(pThis, pDev);
2115	#endif
2116	#ifdef VBOXNETFLT_WITH_QDISC
2117	vboxNetFltLinuxQdiscRemove(pThis, pDev);
2118	#endif /* VBOXNETFLT_WITH_QDISC */
2119	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2120	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2121	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2122	dev_put(pDev);
2123	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) released. ref=%d\n",
2124	pDev, pDev->name,
2125	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2126	netdev_refcnt_read(pDev)
2127	#else
2128	atomic_read(&pDev->refcnt)
2129	#endif
2130	));
2131	}
2132
2133	LogRel(("VBoxNetFlt: attached to '%s' / %.*Rhxs\n", pThis->szName, sizeof(pThis->u.s.MacAddr), &pThis->u.s.MacAddr));
2134	return VINF_SUCCESS;
2135	}
2136
2137
2138	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
2139	{
2140	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2141
2142	Assert(!pThis->fDisconnectedFromHost);
2143
2144	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2145	vboxNetFltLinuxUnhookDev(pThis, pDev);
2146	#endif
2147	#ifdef VBOXNETFLT_WITH_QDISC
2148	vboxNetFltLinuxQdiscRemove(pThis, pDev);
2149	#endif /* VBOXNETFLT_WITH_QDISC */
2150
2151	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2152	ASMAtomicWriteBool(&pThis->u.s.fRegistered, false);
2153	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
2154	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2155	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2156
2157	dev_remove_pack(&pThis->u.s.PacketType);
2158	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2159	skb_queue_purge(&pThis->u.s.XmitQueue);
2160	#endif
2161	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
2162	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
2163	pDev, pDev->name,
2164	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2165	netdev_refcnt_read(pDev)
2166	#else
2167	atomic_read(&pDev->refcnt)
2168	#endif
2169	));
2170	dev_put(pDev);
2171
2172	return NOTIFY_OK;
2173	}
2174
2175	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
2176	{
2177	/* Check if we are not suspended and promiscuous mode has not been set. */
2178	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
2179	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
2180	{
2181	/* Note that there is no need for locking as the kernel got hold of the lock already. */
2182	dev_set_promiscuity(pDev, 1);
2183	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
2184	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2185	}
2186	else
2187	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2188	return NOTIFY_OK;
2189	}
2190
2191	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
2192	{
2193	/* Undo promiscuous mode if we has set it. */
2194	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
2195	{
2196	/* Note that there is no need for locking as the kernel got hold of the lock already. */
2197	dev_set_promiscuity(pDev, -1);
2198	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
2199	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2200	}
2201	else
2202	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2203	return NOTIFY_OK;
2204	}
2205
2206	#ifdef LOG_ENABLED
2207	/** Stringify the NETDEV_XXX constants. */
2208	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2209	{
2210	const char *pszEvent = "NETDRV_<unknown>";
2211	switch (ulEventType)
2212	{
2213	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2214	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2215	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2216	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2217	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2218	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2219	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2220	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2221	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2222	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2223	# ifdef NETDEV_FEAT_CHANGE
2224	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2225	# endif
2226	}
2227	return pszEvent;
2228	}
2229	#endif /* LOG_ENABLED */
2230
2231	/**
2232	* Callback for listening to netdevice events.
2233	*
2234	* This works the rediscovery, clean up on unregistration, promiscuity on
2235	* up/down, and GSO feature changes from ethtool.
2236	*
2237	* @returns NOTIFY_OK
2238	* @param self Pointer to our notifier registration block.
2239	* @param ulEventType The event.
2240	* @param ptr Event specific, but it is usually the device it
2241	* relates to.
2242	*/
2243	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2244
2245	{
2246	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2247	struct net_device pDev = (struct net_device )ptr;
2248	int rc = NOTIFY_OK;
2249
2250	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2251	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
2252	if ( ulEventType == NETDEV_REGISTER
2253	&& !strcmp(pDev->name, pThis->szName))
2254	{
2255	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2256	}
2257	else
2258	{
2259	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2260	if (pDev == ptr)
2261	{
2262	switch (ulEventType)
2263	{
2264	case NETDEV_UNREGISTER:
2265	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2266	break;
2267	case NETDEV_UP:
2268	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2269	break;
2270	case NETDEV_GOING_DOWN:
2271	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2272	break;
2273	case NETDEV_CHANGENAME:
2274	break;
2275	#ifdef NETDEV_FEAT_CHANGE
2276	case NETDEV_FEAT_CHANGE:
2277	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2278	break;
2279	#endif
2280	}
2281	}
2282	}
2283
2284	return rc;
2285	}
2286
2287	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2288	{
2289	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2290	}
2291
2292	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2293	{
2294	struct net_device * pDev;
2295	int err;
2296	int rc = VINF_SUCCESS;
2297	NOREF(pvIfData);
2298
2299	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2300
2301	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2302	if (pDev)
2303	{
2304	/*
2305	* Create a sk_buff for the gather list and push it onto the wire.
2306	*/
2307	if (fDst & INTNETTRUNKDIR_WIRE)
2308	{
2309	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2310	if (pBuf)
2311	{
2312	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2313	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2314	Log4(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2315	err = dev_queue_xmit(pBuf);
2316	if (err)
2317	rc = RTErrConvertFromErrno(err);
2318	}
2319	else
2320	rc = VERR_NO_MEMORY;
2321	}
2322
2323	/*
2324	* Create a sk_buff for the gather list and push it onto the host stack.
2325	*/
2326	if (fDst & INTNETTRUNKDIR_HOST)
2327	{
2328	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2329	if (pBuf)
2330	{
2331	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2332	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2333	Log4(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2334	err = netif_rx_ni(pBuf);
2335	if (err)
2336	rc = RTErrConvertFromErrno(err);
2337	}
2338	else
2339	rc = VERR_NO_MEMORY;
2340	}
2341
2342	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2343	}
2344
2345	return rc;
2346	}
2347
2348
2349	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2350	{
2351	struct net_device * pDev;
2352
2353	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%s, fDisablePromiscuous=%s\n",
2354	pThis, pThis->szName, fActive?"true":"false",
2355	pThis->fDisablePromiscuous?"true":"false"));
2356
2357	if (pThis->fDisablePromiscuous)
2358	return;
2359
2360	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2361	if (pDev)
2362	{
2363	/*
2364	* This api is a bit weird, the best reference is the code.
2365	*
2366	* Also, we have a bit or race conditions wrt the maintenance of
2367	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2368	*/
2369	#ifdef LOG_ENABLED
2370	u_int16_t fIf;
2371	unsigned const cPromiscBefore = pDev->promiscuity;
2372	#endif
2373	if (fActive)
2374	{
2375	Assert(!pThis->u.s.fPromiscuousSet);
2376
2377	rtnl_lock();
2378	dev_set_promiscuity(pDev, 1);
2379	rtnl_unlock();
2380	pThis->u.s.fPromiscuousSet = true;
2381	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2382	}
2383	else
2384	{
2385	if (pThis->u.s.fPromiscuousSet)
2386	{
2387	rtnl_lock();
2388	dev_set_promiscuity(pDev, -1);
2389	rtnl_unlock();
2390	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2391	}
2392	pThis->u.s.fPromiscuousSet = false;
2393
2394	#ifdef LOG_ENABLED
2395	fIf = dev_get_flags(pDev);
2396	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2397	#endif
2398	}
2399
2400	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2401	}
2402	}
2403
2404
2405	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2406	{
2407	#ifdef VBOXNETFLT_WITH_QDISC
2408	vboxNetFltLinuxQdiscRemove(pThis, NULL);
2409	#endif /* VBOXNETFLT_WITH_QDISC */
2410	/*
2411	* Remove packet handler when we get disconnected from internal switch as
2412	* we don't want the handler to forward packets to disconnected switch.
2413	*/
2414	dev_remove_pack(&pThis->u.s.PacketType);
2415	return VINF_SUCCESS;
2416	}
2417
2418
2419	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2420	{
2421	/*
2422	* Report the GSO capabilities of the host and device (if connected).
2423	* Note! No need to mark ourselves busy here.
2424	*/
2425	/** @todo duplicate work here now? Attach */
2426	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2427	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2428	0
2429	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2430	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2431	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
2432	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP)
2433	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP)
2434	# endif
2435	, INTNETTRUNKDIR_HOST);
2436
2437	#endif
2438	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2439
2440	return VINF_SUCCESS;
2441	}
2442
2443
2444	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2445	{
2446	struct net_device *pDev;
2447	bool fRegistered;
2448	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2449
2450	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2451	vboxNetFltLinuxUnhookDev(pThis, NULL);
2452	#endif
2453
2454	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2455	* unlikely, but none the less). Since it doesn't actually update the
2456	* state (just reads it), it is likely to panic in some interesting
2457	* ways. */
2458
2459	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2460	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2461	fRegistered = ASMAtomicUoReadBool(&pThis->u.s.fRegistered);
2462	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2463
2464	if (fRegistered)
2465	{
2466	vboxNetFltSetTapLinkState(pThis, pDev, false);
2467
2468	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2469	skb_queue_purge(&pThis->u.s.XmitQueue);
2470	#endif
2471	Log(("vboxNetFltOsDeleteInstance: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
2472	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2473	pDev, pDev->name,
2474	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2475	netdev_refcnt_read(pDev)
2476	#else
2477	atomic_read(&pDev->refcnt)
2478	#endif
2479	));
2480	dev_put(pDev);
2481	}
2482	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2483	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2484	module_put(THIS_MODULE);
2485	}
2486
2487
2488	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2489	{
2490	int err;
2491	NOREF(pvContext);
2492
2493	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2494	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2495	if (err)
2496	return VERR_INTNET_FLT_IF_FAILED;
2497	if (!pThis->u.s.fRegistered)
2498	{
2499	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2500	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2501	return VERR_INTNET_FLT_IF_NOT_FOUND;
2502	}
2503
2504	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2505	if ( pThis->fDisconnectedFromHost
2506	\|\| !try_module_get(THIS_MODULE))
2507	return VERR_INTNET_FLT_IF_FAILED;
2508
2509	return VINF_SUCCESS;
2510	}
2511
2512	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2513	{
2514	/*
2515	* Init the linux specific members.
2516	*/
2517	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2518	pThis->u.s.fRegistered = false;
2519	pThis->u.s.fPromiscuousSet = false;
2520	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2521	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2522	skb_queue_head_init(&pThis->u.s.XmitQueue);
2523	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
2524	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2525	# else
2526	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2527	# endif
2528	#endif
2529
2530	return VINF_SUCCESS;
2531	}
2532
2533
2534	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2535	{
2536	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2537	}
2538
2539
2540	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2541	{
2542	/* Nothing to do */
2543	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2544	return VINF_SUCCESS;
2545	}
2546
2547
2548	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2549	{
2550	/* Nothing to do */
2551	NOREF(pThis); NOREF(pvIfData);
2552	return VINF_SUCCESS;
2553	}
2554

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

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 38054

Download in other formats: