VBoxNetFlt-linux.c@ 38549

Last change on this file since 38549 was 38549, checked in by vboxsync, 13 years ago
GSO: UDP fragmentation offloading (#5846)
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 88.7 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 38549 2011-08-26 13:26:07Z 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	/* Cancel override if ethtool_ops is missing (host-only case, #5712) */
962	if (!VALID_PTR(pDev->OVR_OPS))
963	return;
964	pOverride = RTMemAlloc(sizeof(*pOverride));
965	if (!pOverride)
966	return;
967	pOverride->pOrgOps = pDev->OVR_OPS;
968	pOverride->Ops = *pDev->OVR_OPS;
969	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
970	pOverride->pfnStartXmit = pDev->hard_start_xmit;
971	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
972	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
973	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
974	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
975	pOverride->cTotal = 0;
976	pOverride->cFiltered = 0;
977	pOverride->pVBoxNetFlt = pThis;
978
979	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp); /* (this isn't necessary, but so what) */
980	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
981	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
982	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
983	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
984	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
985	}
986
987	/**
988	* Undos what vboxNetFltLinuxHookDev did.
989	*
990	* @param pThis The net filter instance.
991	* @param pDev The net device. Can be NULL, in which case
992	* we'll try retrieve it from @a pThis.
993	*/
994	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
995	{
996	PVBOXNETDEVICEOPSOVERRIDE pOverride;
997	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
998
999	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
1000	if (!pDev)
1001	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1002	if (VALID_PTR(pDev))
1003	{
1004	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
1005	if ( VALID_PTR(pOverride)
1006	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
1007	&& VALID_PTR(pOverride->pOrgOps)
1008	)
1009	{
1010	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
1011	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
1012	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
1013	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
1014	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
1015	}
1016	else
1017	pOverride = NULL;
1018	}
1019	else
1020	pOverride = NULL;
1021	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
1022
1023	if (pOverride)
1024	{
1025	printk("vboxnetflt: dropped %llu out of %llu packets\n", pOverride->cFiltered, pOverride->cTotal);
1026	RTMemFree(pOverride);
1027	}
1028	}
1029
1030	#endif /* VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT */
1031
1032
1033	/**
1034	* Reads and retains the host interface handle.
1035	*
1036	* @returns The handle, NULL if detached.
1037	* @param pThis
1038	*/
1039	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
1040	{
1041	#if 0
1042	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1043	struct net_device *pDev = NULL;
1044
1045	Log(("vboxNetFltLinuxRetainNetDev\n"));
1046	/*
1047	* Be careful here to avoid problems racing the detached callback.
1048	*/
1049	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1050	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
1051	{
1052	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
1053	if (pDev)
1054	{
1055	dev_hold(pDev);
1056	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
1057	pDev, pDev->name,
1058	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1059	netdev_refcnt_read(pDev)
1060	#else
1061	atomic_read(&pDev->refcnt)
1062	#endif
1063	));
1064	}
1065	}
1066	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1067
1068	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
1069	return pDev;
1070	#else
1071	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1072	#endif
1073	}
1074
1075
1076	/**
1077	* Release the host interface handle previously retained
1078	* by vboxNetFltLinuxRetainNetDev.
1079	*
1080	* @param pThis The instance.
1081	* @param pDev The vboxNetFltLinuxRetainNetDev
1082	* return value, NULL is fine.
1083	*/
1084	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
1085	{
1086	#if 0
1087	Log(("vboxNetFltLinuxReleaseNetDev\n"));
1088	NOREF(pThis);
1089	if (pDev)
1090	{
1091	dev_put(pDev);
1092	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
1093	pDev, pDev->name,
1094	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1095	netdev_refcnt_read(pDev)
1096	#else
1097	atomic_read(&pDev->refcnt)
1098	#endif
1099	));
1100	}
1101	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
1102	#endif
1103	}
1104
1105	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
1106	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
1107
1108	/**
1109	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
1110	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
1111	*
1112	* @returns true / false accordingly.
1113	* @param pBuf The sk_buff.
1114	*/
1115	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
1116	{
1117	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
1118	}
1119
1120
1121	/**
1122	* Internal worker that create a linux sk_buff for a
1123	* (scatter/)gather list.
1124	*
1125	* @returns Pointer to the sk_buff.
1126	* @param pThis The instance.
1127	* @param pSG The (scatter/)gather list.
1128	* @param fDstWire Set if the destination is the wire.
1129	*/
1130	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
1131	{
1132	struct sk_buff *pPkt;
1133	struct net_device *pDev;
1134	unsigned fGsoType = 0;
1135
1136	if (pSG->cbTotal == 0)
1137	{
1138	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
1139	return NULL;
1140	}
1141
1142	/** @todo We should use fragments mapping the SG buffers with large packets.
1143	* 256 bytes seems to be the a threshold used a lot for this. It
1144	* requires some nasty work on the intnet side though... */
1145	/*
1146	* Allocate a packet and copy over the data.
1147	*/
1148	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1149	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
1150	if (RT_UNLIKELY(!pPkt))
1151	{
1152	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
1153	pSG->pvUserData = NULL;
1154	return NULL;
1155	}
1156	pPkt->dev = pDev;
1157	pPkt->ip_summed = CHECKSUM_NONE;
1158
1159	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
1160	skb_reserve(pPkt, NET_IP_ALIGN);
1161
1162	/* Copy the segments. */
1163	skb_put(pPkt, pSG->cbTotal);
1164	IntNetSgRead(pSG, pPkt->data);
1165
1166	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1167	/*
1168	* Setup GSO if used by this packet.
1169	*/
1170	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
1171	{
1172	default:
1173	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
1174	/* fall thru */
1175	case PDMNETWORKGSOTYPE_INVALID:
1176	fGsoType = 0;
1177	break;
1178	case PDMNETWORKGSOTYPE_IPV4_TCP:
1179	fGsoType = SKB_GSO_TCPV4;
1180	break;
1181	case PDMNETWORKGSOTYPE_IPV4_UDP:
1182	fGsoType = SKB_GSO_UDP;
1183	break;
1184	case PDMNETWORKGSOTYPE_IPV6_TCP:
1185	fGsoType = SKB_GSO_TCPV6;
1186	break;
1187	}
1188	if (fGsoType)
1189	{
1190	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
1191
1192	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
1193	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
1194	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
1195
1196	/*
1197	* We need to set checksum fields even if the packet goes to the host
1198	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
1199	*/
1200	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrsTotal);
1201	pPkt->ip_summed = CHECKSUM_PARTIAL;
1202	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1203	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
1204	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
1205	pPkt->csum_offset = RT_OFFSETOF(RTNETTCP, th_sum);
1206	else
1207	pPkt->csum_offset = RT_OFFSETOF(RTNETUDP, uh_sum);
1208	# else
1209	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
1210	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
1211	pPkt->csum = RT_OFFSETOF(RTNETTCP, th_sum);
1212	else
1213	pPkt->csum = RT_OFFSETOF(RTNETUDP, uh_sum);
1214	# endif
1215	if (!fDstWire)
1216	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
1217	}
1218	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
1219
1220	/*
1221	* Finish up the socket buffer.
1222	*/
1223	pPkt->protocol = eth_type_trans(pPkt, pDev);
1224	if (fDstWire)
1225	{
1226	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
1227
1228	/* Restore ethernet header back. */
1229	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
1230	VBOX_SKB_RESET_MAC_HDR(pPkt);
1231	}
1232	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
1233
1234	return pPkt;
1235	}
1236
1237
1238	/**
1239	* Initializes a SG list from an sk_buff.
1240	*
1241	* @returns Number of segments.
1242	* @param pThis The instance.
1243	* @param pBuf The sk_buff.
1244	* @param pSG The SG.
1245	* @param pvFrame The frame pointer, optional.
1246	* @param cSegs The number of segments allocated for the SG.
1247	* This should match the number in the mbuf exactly!
1248	* @param fSrc The source of the frame.
1249	* @param pGso Pointer to the GSO context if it's a GSO
1250	* internal network frame. NULL if regular frame.
1251	*/
1252	DECLINLINE(void) vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
1253	unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1254	{
1255	int i;
1256	NOREF(pThis);
1257
1258	Assert(!skb_shinfo(pBuf)->frag_list);
1259
1260	if (!pGsoCtx)
1261	IntNetSgInitTempSegs(pSG, pBuf->len, cSegs, 0 /cSegsUsed/);
1262	else
1263	IntNetSgInitTempSegsGso(pSG, pBuf->len, cSegs, 0 /cSegsUsed/, pGsoCtx);
1264
1265	#ifdef VBOXNETFLT_SG_SUPPORT
1266	pSG->aSegs[0].cb = skb_headlen(pBuf);
1267	pSG->aSegs[0].pv = pBuf->data;
1268	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
1269
1270	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1271	{
1272	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
1273	pSG->aSegs[i+1].cb = pFrag->size;
1274	pSG->aSegs[i+1].pv = kmap(pFrag->page);
1275	printk("%p = kmap()\n", pSG->aSegs[i+1].pv);
1276	pSG->aSegs[i+1].Phys = NIL_RTHCPHYS;
1277	}
1278	++i;
1279
1280	#else
1281	pSG->aSegs[0].cb = pBuf->len;
1282	pSG->aSegs[0].pv = pBuf->data;
1283	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
1284	i = 1;
1285	#endif
1286
1287	pSG->cSegsUsed = i;
1288
1289	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1290	/*
1291	* Add a trailer if the frame is too small.
1292	*
1293	* Since we're getting to the packet before it is framed, it has not
1294	* yet been padded. The current solution is to add a segment pointing
1295	* to a buffer containing all zeros and pray that works for all frames...
1296	*/
1297	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
1298	{
1299	static uint8_t const s_abZero[128] = {0};
1300
1301	AssertReturnVoid(i < cSegs);
1302
1303	pSG->aSegs[i].Phys = NIL_RTHCPHYS;
1304	pSG->aSegs[i].pv = (void *)&s_abZero[0];
1305	pSG->aSegs[i].cb = 60 - pSG->cbTotal;
1306	pSG->cbTotal = 60;
1307	pSG->cSegsUsed++;
1308	Assert(i + 1 <= pSG->cSegsAlloc)
1309	}
1310	#endif
1311
1312	Log4(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1313	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1314	for (i = 0; i < pSG->cSegsUsed; i++)
1315	Log4(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1316	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1317	}
1318
1319	/**
1320	* Packet handler,
1321	*
1322	* @returns 0 or EJUSTRETURN.
1323	* @param pThis The instance.
1324	* @param pMBuf The mbuf.
1325	* @param pvFrame The start of the frame, optional.
1326	* @param fSrc Where the packet (allegedly) comes from, one INTNETTRUNKDIR_* value.
1327	* @param eProtocol The protocol.
1328	*/
1329	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 14)
1330	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1331	struct net_device *pSkbDev,
1332	struct packet_type *pPacketType,
1333	struct net_device *pOrigDev)
1334	#else
1335	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1336	struct net_device *pSkbDev,
1337	struct packet_type *pPacketType)
1338	#endif
1339	{
1340	PVBOXNETFLTINS pThis;
1341	struct net_device *pDev;
1342	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1343	pBuf, pSkbDev, pPacketType));
1344	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1345	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",
1346	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));
1347	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1348	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1349	# endif
1350	#else
1351	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",
1352	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));
1353	#endif
1354	/*
1355	* Drop it immediately?
1356	*/
1357	if (!pBuf)
1358	return 0;
1359
1360	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1361	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1362	if (pDev != pSkbDev)
1363	{
1364	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1365	return 0;
1366	}
1367
1368	Log4(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1369	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1370	{
1371	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1372	dev_kfree_skb(pBuf);
1373	return 0;
1374	}
1375
1376	#ifndef VBOXNETFLT_SG_SUPPORT
1377	{
1378	/*
1379	* Get rid of fragmented packets, they cause too much trouble.
1380	*/
1381	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1382	kfree_skb(pBuf);
1383	if (!pCopy)
1384	{
1385	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1386	return 0;
1387	}
1388	pBuf = pCopy;
1389	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1390	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",
1391	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));
1392	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1393	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1394	# endif
1395	# else
1396	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",
1397	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));
1398	# endif
1399	}
1400	#endif
1401
1402	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1403	/* Forward it to the internal network. */
1404	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1405	#else
1406	/* Add the packet to transmit queue and schedule the bottom half. */
1407	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1408	schedule_work(&pThis->u.s.XmitTask);
1409	Log4(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1410	&pThis->u.s.XmitTask, pBuf));
1411	#endif
1412
1413	/* It does not really matter what we return, it is ignored by the kernel. */
1414	return 0;
1415	}
1416
1417	/**
1418	* Calculate the number of INTNETSEG segments the socket buffer will need.
1419	*
1420	* @returns Segment count.
1421	* @param pBuf The socket buffer.
1422	*/
1423	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff *pBuf)
1424	{
1425	#ifdef VBOXNETFLT_SG_SUPPORT
1426	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1427	#else
1428	unsigned cSegs = 1;
1429	#endif
1430	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1431	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1432	if (pBuf->len < 60)
1433	cSegs++;
1434	#endif
1435	return cSegs;
1436	}
1437
1438	/**
1439	* Destroy the intnet scatter / gather buffer created by
1440	* vboxNetFltLinuxSkBufToSG.
1441	*/
1442	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG)
1443	{
1444	#ifdef VBOXNETFLT_SG_SUPPORT
1445	int i;
1446
1447	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1448	{
1449	printk("kunmap(%p)\n", pSG->aSegs[i+1].pv);
1450	kunmap(pSG->aSegs[i+1].pv);
1451	}
1452	#endif
1453	NOREF(pSG);
1454	}
1455
1456	#ifdef LOG_ENABLED
1457	/**
1458	* Logging helper.
1459	*/
1460	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1461	{
1462	int i, offSeg;
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	if (pSG->cSegsUsed == 1)
1483	{
1484	Log3(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1485	}
1486	else
1487	{
1488	for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1489	{
1490	Log3(("-- segment %d at 0x%x (%d bytes) --\n%.*Rhxd\n",
1491	i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1492	offSeg += pSG->aSegs[i].cb;
1493	}
1494	}
1495
1496	}
1497	#else
1498	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1499	#endif
1500
1501	#ifdef VBOXNETFLT_WITH_GSO_RECV
1502
1503	/**
1504	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1505	* GSO socket buffer without having to segment it.
1506	*
1507	* @returns true on success, false if needs segmenting.
1508	* @param pThis The net filter instance.
1509	* @param pSkb The GSO socket buffer.
1510	* @param fSrc The source.
1511	* @param pGsoCtx Where to return the GSO context on success.
1512	*/
1513	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1514	PPDMNETWORKGSO pGsoCtx)
1515	{
1516	PDMNETWORKGSOTYPE enmGsoType;
1517	uint16_t uEtherType;
1518	unsigned int cbTransport;
1519	unsigned int offTransport;
1520	unsigned int cbTransportHdr;
1521	unsigned uProtocol;
1522	union
1523	{
1524	RTNETIPV4 IPv4;
1525	RTNETIPV6 IPv6;
1526	RTNETTCP Tcp;
1527	uint8_t ab[40];
1528	uint16_t au16[40/2];
1529	uint32_t au32[40/4];
1530	} Buf;
1531
1532	/*
1533	* Check the GSO properties of the socket buffer and make sure it fits.
1534	*/
1535	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1536	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_UDP \| SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1537	{
1538	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1539	return false;
1540	}
1541	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1542	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1543	{
1544	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1545	return false;
1546	}
1547	/*
1548	* It is possible to receive GSO packets from wire if GRO is enabled.
1549	*/
1550	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1551	{
1552	Log5(("vboxNetFltLinuxCanForwardAsGso: fSrc=wire\n"));
1553	#ifdef VBOXNETFLT_WITH_GRO
1554	/*
1555	* The packet came from the wire and the driver has already consumed
1556	* mac header. We need to restore it back.
1557	*/
1558	pSkb->mac_len = skb_network_header(pSkb) - skb_mac_header(pSkb);
1559	skb_push(pSkb, pSkb->mac_len);
1560	Log5(("vboxNetFltLinuxCanForwardAsGso: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1561	pSkb->mac_len, pSkb->data, skb_mac_header(pSkb), skb_network_header(pSkb)));
1562	#else /* !VBOXNETFLT_WITH_GRO */
1563	/* Older kernels didn't have GRO. */
1564	return false;
1565	#endif /* !VBOXNETFLT_WITH_GRO */
1566	}
1567	else
1568	{
1569	/*
1570	* skb_gso_segment does the following. Do we need to do it as well?
1571	*/
1572	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1573	skb_reset_mac_header(pSkb);
1574	pSkb->mac_len = pSkb->network_header - pSkb->mac_header;
1575	#else
1576	pSkb->mac.raw = pSkb->data;
1577	pSkb->mac_len = pSkb->nh.raw - pSkb->data;
1578	#endif
1579	}
1580
1581	/*
1582	* Switch on the ethertype.
1583	*/
1584	uEtherType = pSkb->protocol;
1585	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1586	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1587	{
1588	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1589	if (puEtherType)
1590	uEtherType = *puEtherType;
1591	}
1592	switch (uEtherType)
1593	{
1594	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1595	{
1596	unsigned int cbHdr;
1597	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1598	if (RT_UNLIKELY(!pIPv4))
1599	{
1600	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1601	return false;
1602	}
1603
1604	cbHdr = pIPv4->ip_hl * 4;
1605	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1606	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1607	\|\| cbHdr > cbTransport ))
1608	{
1609	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1610	return false;
1611	}
1612	cbTransport -= cbHdr;
1613	offTransport = pSkb->mac_len + cbHdr;
1614	uProtocol = pIPv4->ip_p;
1615	if (uProtocol == RTNETIPV4_PROT_TCP)
1616	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1617	else if (uProtocol == RTNETIPV4_PROT_UDP)
1618	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1619	else /** @todo IPv6: 4to6 tunneling */
1620	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1621	break;
1622	}
1623
1624	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1625	{
1626	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1627	if (RT_UNLIKELY(!pIPv6))
1628	{
1629	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1630	return false;
1631	}
1632
1633	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1634	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1635	uProtocol = pIPv6->ip6_nxt;
1636	/** @todo IPv6: Dig our way out of the other headers. */
1637	if (uProtocol == RTNETIPV4_PROT_TCP)
1638	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1639	else if (uProtocol == RTNETIPV4_PROT_UDP)
1640	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1641	else
1642	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1643	break;
1644	}
1645
1646	default:
1647	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1648	return false;
1649	}
1650
1651	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1652	{
1653	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1654	return false;
1655	}
1656
1657	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1658	\|\| offTransport + cbTransport > pSkb->len
1659	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1660	{
1661	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1662	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1663	return false;
1664	}
1665
1666	/*
1667	* Check the TCP/UDP bits.
1668	*/
1669	if (uProtocol == RTNETIPV4_PROT_TCP)
1670	{
1671	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1672	if (RT_UNLIKELY(!pTcp))
1673	{
1674	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1675	return false;
1676	}
1677
1678	cbTransportHdr = pTcp->th_off * 4;
1679	pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1680	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1681	\|\| cbTransportHdr > cbTransport
1682	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1683	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1684	{
1685	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1686	return false;
1687	}
1688
1689	}
1690	else
1691	{
1692	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1693	cbTransportHdr = sizeof(RTNETUDP);
1694	pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1695	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1696	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1697	{
1698	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1699	return false;
1700	}
1701	}
1702
1703	/*
1704	* We're good, init the GSO context.
1705	*/
1706	pGsoCtx->u8Type = enmGsoType;
1707	pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1708	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1709	pGsoCtx->offHdr1 = pSkb->mac_len;
1710	pGsoCtx->offHdr2 = offTransport;
1711	pGsoCtx->u8Unused = 0;
1712
1713	return true;
1714	}
1715
1716	/**
1717	* Forward the socket buffer as a GSO internal network frame.
1718	*
1719	* @returns IPRT status code.
1720	* @param pThis The net filter instance.
1721	* @param pSkb The GSO socket buffer.
1722	* @param fSrc The source.
1723	* @param pGsoCtx Where to return the GSO context on success.
1724	*/
1725	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1726	{
1727	int rc;
1728	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb);
1729	if (RT_LIKELY(cSegs <= MAX_SKB_FRAGS + 1))
1730	{
1731	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1732	if (RT_LIKELY(pSG))
1733	{
1734	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cSegs, fSrc, pGsoCtx);
1735
1736	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1737	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1738
1739	vboxNetFltLinuxDestroySG(pSG);
1740	rc = VINF_SUCCESS;
1741	}
1742	else
1743	{
1744	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1745	rc = VERR_NO_MEMORY;
1746	}
1747	}
1748	else
1749	{
1750	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1751	rc = VERR_INTERNAL_ERROR_3;
1752	}
1753
1754	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1755	dev_kfree_skb(pSkb);
1756	return rc;
1757	}
1758
1759	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1760
1761	/**
1762	* Worker for vboxNetFltLinuxForwardToIntNet.
1763	*
1764	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1765	* @param pThis The net filter instance.
1766	* @param pBuf The socket buffer.
1767	* @param fSrc The source.
1768	*/
1769	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1770	{
1771	int rc;
1772	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf);
1773	if (cSegs <= MAX_SKB_FRAGS + 1)
1774	{
1775	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1776	if (RT_LIKELY(pSG))
1777	{
1778	if (fSrc & INTNETTRUNKDIR_WIRE)
1779	{
1780	/*
1781	* The packet came from wire, ethernet header was removed by device driver.
1782	* Restore it.
1783	*/
1784	skb_push(pBuf, ETH_HLEN);
1785	}
1786
1787	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cSegs, fSrc, NULL /pGsoCtx/);
1788
1789	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1790	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1791
1792	vboxNetFltLinuxDestroySG(pSG);
1793	rc = VINF_SUCCESS;
1794	}
1795	else
1796	{
1797	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1798	rc = VERR_NO_MEMORY;
1799	}
1800	}
1801	else
1802	{
1803	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1804	rc = VERR_INTERNAL_ERROR_3;
1805	}
1806
1807	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1808	dev_kfree_skb(pBuf);
1809	return rc;
1810	}
1811
1812	/**
1813	*
1814	* @param pBuf The socket buffer. This is consumed by this function.
1815	*/
1816	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1817	{
1818	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1819
1820	#ifdef VBOXNETFLT_WITH_GSO
1821	if (skb_is_gso(pBuf))
1822	{
1823	PDMNETWORKGSO GsoCtx;
1824	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",
1825	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));
1826	# ifdef VBOXNETFLT_WITH_GSO_RECV
1827	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_UDP \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1828	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1829	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1830	else
1831	# endif
1832	{
1833	/* Need to segment the packet */
1834	struct sk_buff *pNext;
1835	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1836	if (IS_ERR(pSegment))
1837	{
1838	dev_kfree_skb(pBuf);
1839	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1840	return;
1841	}
1842
1843	for (; pSegment; pSegment = pNext)
1844	{
1845	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",
1846	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));
1847	pNext = pSegment->next;
1848	pSegment->next = 0;
1849	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1850	}
1851	dev_kfree_skb(pBuf);
1852	}
1853	}
1854	else
1855	#endif /* VBOXNETFLT_WITH_GSO */
1856	{
1857	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1858	{
1859	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1860	/*
1861	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1862	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1863	* header length from the header itself and reconstruct 'h' pointer
1864	* to TCP (or whatever) header.
1865	*/
1866	unsigned char *tmp = pBuf->h.raw;
1867	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1868	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1869	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1870	if (VBOX_SKB_CHECKSUM_HELP(pBuf))
1871	{
1872	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1873	dev_kfree_skb(pBuf);
1874	return;
1875	}
1876	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1877	/* Restore the original (wrong) pointer. */
1878	pBuf->h.raw = tmp;
1879	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1880	}
1881	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1882	}
1883	}
1884
1885	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1886	/**
1887	* Work queue handler that forwards the socket buffers queued by
1888	* vboxNetFltLinuxPacketHandler to the internal network.
1889	*
1890	* @param pWork The work queue.
1891	*/
1892	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
1893	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1894	# else
1895	static void vboxNetFltLinuxXmitTask(void *pWork)
1896	# endif
1897	{
1898	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1899	struct sk_buff *pBuf;
1900
1901	Log4(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1902
1903	/*
1904	* Active? Retain the instance and increment the busy counter.
1905	*/
1906	if (vboxNetFltTryRetainBusyActive(pThis))
1907	{
1908	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1909	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1910
1911	vboxNetFltRelease(pThis, true /* fBusy */);
1912	}
1913	else
1914	{
1915	/** @todo Shouldn't we just drop the packets here? There is little point in
1916	* making them accumulate when the VM is paused and it'll only waste
1917	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1918	* before start draining the packets (goes for the intnet ring buf
1919	* too)? */
1920	}
1921	}
1922	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1923
1924	/**
1925	* Reports the GSO capabilities of the hardware NIC.
1926	*
1927	* @param pThis The net filter instance. The caller hold a
1928	* reference to this.
1929	*/
1930	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1931	{
1932	#ifdef VBOXNETFLT_WITH_GSO_XMIT_WIRE
1933	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1934	{
1935	struct net_device *pDev;
1936	PINTNETTRUNKSWPORT pSwitchPort;
1937	unsigned int fFeatures;
1938	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1939
1940	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
1941
1942	pSwitchPort = pThis->pSwitchPort; /* this doesn't need to be here, but it doesn't harm. */
1943	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1944	if (pDev)
1945	fFeatures = pDev->features;
1946	else
1947	fFeatures = 0;
1948
1949	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
1950
1951	if (pThis->pSwitchPort)
1952	{
1953	/* Set/update the GSO capabilities of the NIC. */
1954	uint32_t fGsoCapabilites = 0;
1955	if (fFeatures & NETIF_F_TSO)
1956	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1957	if (fFeatures & NETIF_F_TSO6)
1958	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1959	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
1960	if (fFeatures & NETIF_F_UFO)
1961	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP);
1962	if (fFeatures & NETIF_F_UFO)
1963	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP);
1964	# endif
1965	Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s%s%s\n",
1966	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1967	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : "",
1968	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP)) ? "ufo " : "",
1969	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP)) ? "ufo6 " : ""));
1970	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1971	}
1972
1973	vboxNetFltRelease(pThis, true /fBusy/);
1974	}
1975	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE */
1976	}
1977
1978	/**
1979	* Helper that determines whether the host (ignoreing us) is operating the
1980	* interface in promiscuous mode or not.
1981	*/
1982	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1983	{
1984	bool fRc = false;
1985	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1986	if (pDev)
1987	{
1988	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1989	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1990	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1991	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1992	}
1993	return fRc;
1994	}
1995
1996	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 36)
1997	/**
1998	* Helper for detecting TAP devices.
1999	*/
2000	static bool vboxNetFltIsTapDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
2001	{
2002	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
2003	{
2004	struct ethtool_drvinfo Info;
2005
2006	memset(&Info, 0, sizeof(Info));
2007	Info.cmd = ETHTOOL_GDRVINFO;
2008	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
2009	Log3(("vboxNetFltIsTapDevice: driver=%s version=%s bus_info=%s\n",
2010	Info.driver, Info.version, Info.bus_info));
2011
2012	return !strncmp(Info.driver, "tun", 4)
2013	&& !strncmp(Info.bus_info, "tap", 4);
2014	}
2015
2016	return false;
2017	}
2018
2019	/**
2020	* Helper for updating the link state of TAP devices.
2021	* Only TAP devices are affected.
2022	*/
2023	static void vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
2024	{
2025	if (vboxNetFltIsTapDevice(pThis, pDev))
2026	{
2027	Log3(("vboxNetFltSetTapLinkState: bringing %s tap device link state\n",
2028	fLinkUp ? "up" : "down"));
2029	netif_tx_lock_bh(pDev);
2030	if (fLinkUp)
2031	netif_carrier_on(pDev);
2032	else
2033	netif_carrier_off(pDev);
2034	netif_tx_unlock_bh(pDev);
2035	}
2036	}
2037	#else /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
2038	DECLINLINE(void) vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
2039	{
2040	/* Nothing to do for pre-2.6.36 kernels. */
2041	}
2042	#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
2043
2044	/**
2045	* Internal worker for vboxNetFltLinuxNotifierCallback.
2046	*
2047	* @returns VBox status code.
2048	* @param pThis The instance.
2049	* @param fRediscovery If set we're doing a rediscovery attempt, so, don't
2050	* flood the release log.
2051	*/
2052	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
2053	{
2054	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2055	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
2056
2057	/*
2058	* Retain and store the device.
2059	*/
2060	dev_hold(pDev);
2061
2062	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2063	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
2064	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2065
2066	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
2067	pDev, pDev->name,
2068	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2069	netdev_refcnt_read(pDev)
2070	#else
2071	atomic_read(&pDev->refcnt)
2072	#endif
2073	));
2074	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2075	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
2076
2077	/* Get the mac address while we still have a valid net_device reference. */
2078	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
2079
2080	/*
2081	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
2082	*/
2083	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
2084	pThis->u.s.PacketType.dev = pDev;
2085	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
2086	dev_add_pack(&pThis->u.s.PacketType);
2087
2088	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2089	vboxNetFltLinuxHookDev(pThis, pDev);
2090	#endif
2091	#ifdef VBOXNETFLT_WITH_QDISC
2092	vboxNetFltLinuxQdiscInstall(pThis, pDev);
2093	#endif /* VBOXNETFLT_WITH_QDISC */
2094
2095	/*
2096	* If attaching to TAP interface we need to bring the link state up
2097	* starting from 2.6.36 kernel.
2098	*/
2099	vboxNetFltSetTapLinkState(pThis, pDev, true);
2100
2101	/*
2102	* Set indicators that require the spinlock. Be abit paranoid about racing
2103	* the device notification handle.
2104	*/
2105	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2106	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2107	if (pDev)
2108	{
2109	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
2110	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
2111	pDev = NULL; /* don't dereference it */
2112	}
2113	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2114	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
2115
2116	/*
2117	* If the above succeeded report GSO capabilities, if not undo and
2118	* release the device.
2119	*/
2120	if (!pDev)
2121	{
2122	Assert(pThis->pSwitchPort);
2123	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
2124	{
2125	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2126	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
2127	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
2128	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
2129	vboxNetFltRelease(pThis, true /fBusy/);
2130	}
2131	}
2132	else
2133	{
2134	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2135	vboxNetFltLinuxUnhookDev(pThis, pDev);
2136	#endif
2137	#ifdef VBOXNETFLT_WITH_QDISC
2138	vboxNetFltLinuxQdiscRemove(pThis, pDev);
2139	#endif /* VBOXNETFLT_WITH_QDISC */
2140	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2141	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2142	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2143	dev_put(pDev);
2144	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) released. ref=%d\n",
2145	pDev, pDev->name,
2146	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2147	netdev_refcnt_read(pDev)
2148	#else
2149	atomic_read(&pDev->refcnt)
2150	#endif
2151	));
2152	}
2153
2154	LogRel(("VBoxNetFlt: attached to '%s' / %.*Rhxs\n", pThis->szName, sizeof(pThis->u.s.MacAddr), &pThis->u.s.MacAddr));
2155	return VINF_SUCCESS;
2156	}
2157
2158
2159	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
2160	{
2161	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2162
2163	Assert(!pThis->fDisconnectedFromHost);
2164
2165	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2166	vboxNetFltLinuxUnhookDev(pThis, pDev);
2167	#endif
2168	#ifdef VBOXNETFLT_WITH_QDISC
2169	vboxNetFltLinuxQdiscRemove(pThis, pDev);
2170	#endif /* VBOXNETFLT_WITH_QDISC */
2171
2172	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2173	ASMAtomicWriteBool(&pThis->u.s.fRegistered, false);
2174	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
2175	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2176	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2177
2178	dev_remove_pack(&pThis->u.s.PacketType);
2179	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2180	skb_queue_purge(&pThis->u.s.XmitQueue);
2181	#endif
2182	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
2183	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
2184	pDev, pDev->name,
2185	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2186	netdev_refcnt_read(pDev)
2187	#else
2188	atomic_read(&pDev->refcnt)
2189	#endif
2190	));
2191	dev_put(pDev);
2192
2193	return NOTIFY_OK;
2194	}
2195
2196	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
2197	{
2198	/* Check if we are not suspended and promiscuous mode has not been set. */
2199	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
2200	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
2201	{
2202	/* Note that there is no need for locking as the kernel got hold of the lock already. */
2203	dev_set_promiscuity(pDev, 1);
2204	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
2205	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2206	}
2207	else
2208	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2209	return NOTIFY_OK;
2210	}
2211
2212	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
2213	{
2214	/* Undo promiscuous mode if we has set it. */
2215	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
2216	{
2217	/* Note that there is no need for locking as the kernel got hold of the lock already. */
2218	dev_set_promiscuity(pDev, -1);
2219	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
2220	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2221	}
2222	else
2223	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2224	return NOTIFY_OK;
2225	}
2226
2227	#ifdef LOG_ENABLED
2228	/** Stringify the NETDEV_XXX constants. */
2229	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2230	{
2231	const char *pszEvent = "NETDRV_<unknown>";
2232	switch (ulEventType)
2233	{
2234	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2235	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2236	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2237	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2238	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2239	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2240	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2241	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2242	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2243	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2244	# ifdef NETDEV_FEAT_CHANGE
2245	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2246	# endif
2247	}
2248	return pszEvent;
2249	}
2250	#endif /* LOG_ENABLED */
2251
2252	/**
2253	* Callback for listening to netdevice events.
2254	*
2255	* This works the rediscovery, clean up on unregistration, promiscuity on
2256	* up/down, and GSO feature changes from ethtool.
2257	*
2258	* @returns NOTIFY_OK
2259	* @param self Pointer to our notifier registration block.
2260	* @param ulEventType The event.
2261	* @param ptr Event specific, but it is usually the device it
2262	* relates to.
2263	*/
2264	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2265
2266	{
2267	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2268	struct net_device pDev = (struct net_device )ptr;
2269	int rc = NOTIFY_OK;
2270
2271	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2272	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
2273	if ( ulEventType == NETDEV_REGISTER
2274	&& !strcmp(pDev->name, pThis->szName))
2275	{
2276	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2277	}
2278	else
2279	{
2280	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2281	if (pDev == ptr)
2282	{
2283	switch (ulEventType)
2284	{
2285	case NETDEV_UNREGISTER:
2286	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2287	break;
2288	case NETDEV_UP:
2289	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2290	break;
2291	case NETDEV_GOING_DOWN:
2292	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2293	break;
2294	case NETDEV_CHANGENAME:
2295	break;
2296	#ifdef NETDEV_FEAT_CHANGE
2297	case NETDEV_FEAT_CHANGE:
2298	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2299	break;
2300	#endif
2301	}
2302	}
2303	}
2304
2305	return rc;
2306	}
2307
2308	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2309	{
2310	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2311	}
2312
2313	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2314	{
2315	struct net_device * pDev;
2316	int err;
2317	int rc = VINF_SUCCESS;
2318	NOREF(pvIfData);
2319
2320	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2321
2322	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2323	if (pDev)
2324	{
2325	/*
2326	* Create a sk_buff for the gather list and push it onto the wire.
2327	*/
2328	if (fDst & INTNETTRUNKDIR_WIRE)
2329	{
2330	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2331	if (pBuf)
2332	{
2333	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2334	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2335	Log4(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2336	err = dev_queue_xmit(pBuf);
2337	if (err)
2338	rc = RTErrConvertFromErrno(err);
2339	}
2340	else
2341	rc = VERR_NO_MEMORY;
2342	}
2343
2344	/*
2345	* Create a sk_buff for the gather list and push it onto the host stack.
2346	*/
2347	if (fDst & INTNETTRUNKDIR_HOST)
2348	{
2349	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2350	if (pBuf)
2351	{
2352	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2353	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2354	Log4(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2355	err = netif_rx_ni(pBuf);
2356	if (err)
2357	rc = RTErrConvertFromErrno(err);
2358	}
2359	else
2360	rc = VERR_NO_MEMORY;
2361	}
2362
2363	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2364	}
2365
2366	return rc;
2367	}
2368
2369
2370	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2371	{
2372	struct net_device * pDev;
2373
2374	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%s, fDisablePromiscuous=%s\n",
2375	pThis, pThis->szName, fActive?"true":"false",
2376	pThis->fDisablePromiscuous?"true":"false"));
2377
2378	if (pThis->fDisablePromiscuous)
2379	return;
2380
2381	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2382	if (pDev)
2383	{
2384	/*
2385	* This api is a bit weird, the best reference is the code.
2386	*
2387	* Also, we have a bit or race conditions wrt the maintenance of
2388	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2389	*/
2390	#ifdef LOG_ENABLED
2391	u_int16_t fIf;
2392	unsigned const cPromiscBefore = pDev->promiscuity;
2393	#endif
2394	if (fActive)
2395	{
2396	Assert(!pThis->u.s.fPromiscuousSet);
2397
2398	rtnl_lock();
2399	dev_set_promiscuity(pDev, 1);
2400	rtnl_unlock();
2401	pThis->u.s.fPromiscuousSet = true;
2402	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2403	}
2404	else
2405	{
2406	if (pThis->u.s.fPromiscuousSet)
2407	{
2408	rtnl_lock();
2409	dev_set_promiscuity(pDev, -1);
2410	rtnl_unlock();
2411	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2412	}
2413	pThis->u.s.fPromiscuousSet = false;
2414
2415	#ifdef LOG_ENABLED
2416	fIf = dev_get_flags(pDev);
2417	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2418	#endif
2419	}
2420
2421	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2422	}
2423	}
2424
2425
2426	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2427	{
2428	#ifdef VBOXNETFLT_WITH_QDISC
2429	vboxNetFltLinuxQdiscRemove(pThis, NULL);
2430	#endif /* VBOXNETFLT_WITH_QDISC */
2431	/*
2432	* Remove packet handler when we get disconnected from internal switch as
2433	* we don't want the handler to forward packets to disconnected switch.
2434	*/
2435	dev_remove_pack(&pThis->u.s.PacketType);
2436	return VINF_SUCCESS;
2437	}
2438
2439
2440	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2441	{
2442	/*
2443	* Report the GSO capabilities of the host and device (if connected).
2444	* Note! No need to mark ourselves busy here.
2445	*/
2446	/** @todo duplicate work here now? Attach */
2447	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2448	Log3(("vboxNetFltOsConnectIt: reporting host tso tso6 ufo\n"));
2449	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2450	0
2451	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2452	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2453	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP)
2454	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
2455	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP)
2456	# endif
2457	, INTNETTRUNKDIR_HOST);
2458
2459	#endif
2460	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2461
2462	return VINF_SUCCESS;
2463	}
2464
2465
2466	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2467	{
2468	struct net_device *pDev;
2469	bool fRegistered;
2470	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
2471
2472	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
2473	vboxNetFltLinuxUnhookDev(pThis, NULL);
2474	#endif
2475
2476	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2477	* unlikely, but none the less). Since it doesn't actually update the
2478	* state (just reads it), it is likely to panic in some interesting
2479	* ways. */
2480
2481	RTSpinlockAcquireNoInts(pThis->hSpinlock, &Tmp);
2482	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2483	fRegistered = ASMAtomicUoReadBool(&pThis->u.s.fRegistered);
2484	RTSpinlockReleaseNoInts(pThis->hSpinlock, &Tmp);
2485
2486	if (fRegistered)
2487	{
2488	vboxNetFltSetTapLinkState(pThis, pDev, false);
2489
2490	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2491	skb_queue_purge(&pThis->u.s.XmitQueue);
2492	#endif
2493	Log(("vboxNetFltOsDeleteInstance: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
2494	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2495	pDev, pDev->name,
2496	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2497	netdev_refcnt_read(pDev)
2498	#else
2499	atomic_read(&pDev->refcnt)
2500	#endif
2501	));
2502	dev_put(pDev);
2503	}
2504	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2505	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2506	module_put(THIS_MODULE);
2507	}
2508
2509
2510	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2511	{
2512	int err;
2513	NOREF(pvContext);
2514
2515	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2516	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2517	if (err)
2518	return VERR_INTNET_FLT_IF_FAILED;
2519	if (!pThis->u.s.fRegistered)
2520	{
2521	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2522	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2523	return VERR_INTNET_FLT_IF_NOT_FOUND;
2524	}
2525
2526	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2527	if ( pThis->fDisconnectedFromHost
2528	\|\| !try_module_get(THIS_MODULE))
2529	return VERR_INTNET_FLT_IF_FAILED;
2530
2531	return VINF_SUCCESS;
2532	}
2533
2534	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2535	{
2536	/*
2537	* Init the linux specific members.
2538	*/
2539	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2540	pThis->u.s.fRegistered = false;
2541	pThis->u.s.fPromiscuousSet = false;
2542	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2543	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2544	skb_queue_head_init(&pThis->u.s.XmitQueue);
2545	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
2546	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2547	# else
2548	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2549	# endif
2550	#endif
2551
2552	return VINF_SUCCESS;
2553	}
2554
2555
2556	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2557	{
2558	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2559	}
2560
2561
2562	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2563	{
2564	/* Nothing to do */
2565	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2566	return VINF_SUCCESS;
2567	}
2568
2569
2570	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2571	{
2572	/* Nothing to do */
2573	NOREF(pThis); NOREF(pvIfData);
2574	return VINF_SUCCESS;
2575	}
2576

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

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 38549

Download in other formats: