VBoxNetFlt-linux.c@ 28281

Last change on this file since 28281 was 28281, checked in by vboxsync, 15 years ago
VBoxNetFlt/linux: 2.6.22 that is
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 60.7 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 28281 2010-04-14 06:40:48Z vboxsync $ */
2	/** @file
3	* VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4	*/
5
6	/*
7	* Copyright (C) 2006-2008 Sun Microsystems, Inc.
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	* Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa
18	* Clara, CA 95054 USA or visit http://www.sun.com if you need
19	* additional information or have any questions.
20	*/
21
22	/*******************************************************************************
23	* Header Files *
24	*******************************************************************************/
25	#include "the-linux-kernel.h"
26	#include "version-generated.h"
27	#include <linux/netdevice.h>
28	#include <linux/etherdevice.h>
29	#include <linux/rtnetlink.h>
30	#include <linux/miscdevice.h>
31	#include <linux/ip.h>
32
33	#define LOG_GROUP LOG_GROUP_NET_FLT_DRV
34	#include <VBox/log.h>
35	#include <VBox/err.h>
36	#include <VBox/intnetinline.h>
37	#include <VBox/pdmnetinline.h>
38	#include <VBox/param.h>
39	#include <iprt/alloca.h>
40	#include <iprt/assert.h>
41	#include <iprt/spinlock.h>
42	#include <iprt/semaphore.h>
43	#include <iprt/initterm.h>
44	#include <iprt/process.h>
45	#include <iprt/mem.h>
46	#include <iprt/net.h>
47	#include <iprt/log.h>
48	#include <iprt/mp.h>
49	#include <iprt/mem.h>
50	#include <iprt/time.h>
51
52	#define VBOXNETFLT_OS_SPECFIC 1
53	#include "../VBoxNetFltInternal.h"
54
55
56	/*******************************************************************************
57	* Defined Constants And Macros *
58	*******************************************************************************/
59	#define VBOX_FLT_NB_TO_INST(pNB) ((PVBOXNETFLTINS)((uint8_t *)pNB - RT_OFFSETOF(VBOXNETFLTINS, u.s.Notifier)))
60	#define VBOX_FLT_PT_TO_INST(pPT) ((PVBOXNETFLTINS)((uint8_t *)pPT - RT_OFFSETOF(VBOXNETFLTINS, u.s.PacketType)))
61	#define VBOX_FLT_XT_TO_INST(pXT) ((PVBOXNETFLTINS)((uint8_t *)pXT - RT_OFFSETOF(VBOXNETFLTINS, u.s.XmitTask)))
62
63	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
64	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
65	# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
66	#else
67	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
68	# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
69	#endif
70
71	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 19)
72	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
73	#else
74	# define CHECKSUM_PARTIAL CHECKSUM_HW
75	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 10)
76	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
77	# else
78	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 7)
79	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
80	# else
81	# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
82	# endif
83	# endif
84	#endif
85
86	#ifndef NET_IP_ALIGN
87	# define NET_IP_ALIGN 2
88	#endif
89
90	#if 0
91	/** Create scatter / gather segments for fragments. When not used, we will
92	* linearize the socket buffer before creating the internal networking SG. */
93	# define VBOXNETFLT_SG_SUPPORT 1
94	#endif
95
96	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
97	/** Indicates that the linux kernel may send us GSO frames. */
98	# define VBOXNETFLT_WITH_GSO 1
99
100	/** This enables or disables the transmitting of GSO frame from the internal
101	* network and to the host. */
102	# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
103
104	# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
105	/** This enables or disables the transmitting of GSO frame from the internal
106	* network and to the wire. */
107	# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
108	# endif
109
110	/** This enables or disables the forwarding/flooding of GSO frame from the host
111	* to the internal network. */
112	# define VBOXNETFLT_WITH_GSO_RECV 1
113
114	#endif
115
116
117	/*******************************************************************************
118	* Internal Functions *
119	*******************************************************************************/
120	static int VBoxNetFltLinuxInit(void);
121	static void VBoxNetFltLinuxUnload(void);
122
123
124	/*******************************************************************************
125	* Global Variables *
126	*******************************************************************************/
127	/**
128	* The (common) global data.
129	*/
130	static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
131
132	module_init(VBoxNetFltLinuxInit);
133	module_exit(VBoxNetFltLinuxUnload);
134
135	MODULE_AUTHOR("Oracle Corporation");
136	MODULE_DESCRIPTION("VirtualBox Network Filter Driver");
137	MODULE_LICENSE("GPL");
138	#ifdef MODULE_VERSION
139	MODULE_VERSION(VBOX_VERSION_STRING " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
140	#endif
141
142
143	#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) && defined(LOG_ENABLED)
144	unsigned dev_get_flags(const struct net_device *dev)
145	{
146	unsigned flags;
147
148	flags = (dev->flags & ~(IFF_PROMISC \|
149	IFF_ALLMULTI \|
150	IFF_RUNNING)) \|
151	(dev->gflags & (IFF_PROMISC \|
152	IFF_ALLMULTI));
153
154	if (netif_running(dev) && netif_carrier_ok(dev))
155	flags \|= IFF_RUNNING;
156
157	return flags;
158	}
159	#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) */
160
161
162	/**
163	* Initialize module.
164	*
165	* @returns appropriate status code.
166	*/
167	static int __init VBoxNetFltLinuxInit(void)
168	{
169	int rc;
170	/*
171	* Initialize IPRT.
172	*/
173	rc = RTR0Init(0);
174	if (RT_SUCCESS(rc))
175	{
176	Log(("VBoxNetFltLinuxInit\n"));
177
178	/*
179	* Initialize the globals and connect to the support driver.
180	*
181	* This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
182	* for establishing the connect to the support driver.
183	*/
184	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
185	rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
186	if (RT_SUCCESS(rc))
187	{
188	LogRel(("VBoxNetFlt: Successfully started.\n"));
189	return 0;
190	}
191
192	LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
193	RTR0Term();
194	}
195	else
196	LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
197
198	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
199	return -RTErrConvertToErrno(rc);
200	}
201
202
203	/**
204	* Unload the module.
205	*
206	* @todo We have to prevent this if we're busy!
207	*/
208	static void __exit VBoxNetFltLinuxUnload(void)
209	{
210	int rc;
211	Log(("VBoxNetFltLinuxUnload\n"));
212	Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
213
214	/*
215	* Undo the work done during start (in reverse order).
216	*/
217	rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
218	AssertRC(rc); NOREF(rc);
219
220	RTR0Term();
221
222	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
223
224	Log(("VBoxNetFltLinuxUnload - done\n"));
225	}
226
227	/**
228	* Experiment where we filter trafic from the host to the internal network
229	* before it reaches the NIC driver.
230	*
231	* The current code uses a very ugly hack and only works on kernels using the
232	* net_device_ops (>= 2.6.29). It has been shown to give us a
233	* performance boost of 60-100% though. So, we have to find some less hacky way
234	* of getting this job done eventually.
235	*
236	* #define VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
237	*/
238	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
239
240	/**
241	* The overridden net_device_ops of the device we're attached to.
242	*
243	* Requires Linux 2.6.29 or later.
244	*
245	* This is a very dirty hack that was create to explore how much we can improve
246	* the host to guest transfers by not CC'ing the NIC.
247	*/
248	typedef struct VBoxNetDeviceOpsOverride
249	{
250	/** Our overridden ops. */
251	struct net_device_ops Ops;
252	/** Magic word. */
253	uint32_t u32Magic;
254	/** Pointer to the original ops. */
255	struct net_device_ops const *pOrgOps;
256	/** Pointer to the net filter instance. */
257	PVBOXNETFLTINS pVBoxNetFlt;
258	/** The number of filtered packages. */
259	uint64_t cFiltered;
260	/** The total number of packets */
261	uint64_t cTotal;
262	} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
263	/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
264	#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
265
266	/**
267	* ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
268	* because they belong on the internal network.
269	*
270	* @returns NETDEV_TX_XXX.
271	* @param pSkb The socket buffer to transmit.
272	* @param pDev The net device.
273	*/
274	static int vboxNetFltLinuxStartXmitFilter(struct sk_buff pSkb, struct net_device pDev)
275	{
276	PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->netdev_ops;
277	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
278	PCRTNETETHERHDR pEtherHdr;
279	PINTNETTRUNKSWPORT pSwitchPort;
280
281
282	/*
283	* Validate the override structure.
284	*
285	* Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
286	* to be production quality code, we would have to be much more
287	* careful here and avoid the race.
288	*/
289	if ( !VALID_PTR(pOverride)
290	\|\| pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
291	\|\| !VALID_PTR(pOverride->pOrgOps))
292	{
293	printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
294	dev_kfree_skb(pSkb);
295	return NETDEV_TX_OK;
296	}
297	pOverride->cTotal++;
298
299	/*
300	* Do the filtering base on the defaul OUI of our virtual NICs
301	*
302	* Note! In a real solution, we would ask the switch whether the
303	* destination MAC is 100% to be on the internal network and then
304	* drop it.
305	*/
306	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, sizeof(abHdrBuf), &abHdrBuf[0]);
307	if ( pEtherHdr
308	&& VALID_PTR(pOverride->pVBoxNetFlt)
309	&& (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
310	&& VALID_PTR(pSwitchPort)
311	)
312	{
313	INTNETSWDECISION enmDecision;
314	uint32_t cbHdrs = skb_headlen(pSkb);
315	cbHdrs = RT_MAX(cbHdrs, sizeof(abHdrBuf));
316
317	/** @todo consider reference counting, etc. */
318	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
319	if (enmDecision == INTNETSWDECISION_INTNET)
320	{
321	dev_kfree_skb(pSkb);
322	pOverride->cFiltered++;
323	return NETDEV_TX_OK;
324	}
325	}
326
327	return pOverride->pOrgOps->ndo_start_xmit(pSkb, pDev);
328	}
329
330	/**
331	* Hooks the device ndo_start_xmit operation of the device.
332	*
333	* @param pThis The net filter instance.
334	* @param pDev The net device.
335	*/
336	static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
337	{
338	PVBOXNETDEVICEOPSOVERRIDE pOverride;
339	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
340
341	pOverride = RTMemAlloc(sizeof(*pOverride));
342	if (!pOverride)
343	return;
344	pOverride->pOrgOps = pDev->netdev_ops;
345	pOverride->Ops = *pDev->netdev_ops;
346	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
347	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
348	pOverride->cTotal = 0;
349	pOverride->cFiltered = 0;
350	pOverride->pVBoxNetFlt = pThis;
351
352	RTSpinlockAcquire(pThis->hSpinlock, &Tmp); /* (this isn't necessary, but so what) */
353	ASMAtomicXchgPtr((void * volatile *)&pDev->netdev_ops, pOverride);
354	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
355	}
356
357	/**
358	* Undos what vboxNetFltLinuxHookDev did.
359	*
360	* @param pThis The net filter instance.
361	* @param pDev The net device. Can be NULL, in which case
362	* we'll try retrieve it from @a pThis.
363	*/
364	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
365	{
366	PVBOXNETDEVICEOPSOVERRIDE pOverride;
367	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
368
369	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
370	if (!pDev)
371	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
372	if (VALID_PTR(pDev))
373	{
374	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->netdev_ops;
375	if ( VALID_PTR(pOverride)
376	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
377	&& VALID_PTR(pOverride->pOrgOps)
378	)
379	{
380	ASMAtomicXchgPtr((void * volatile *)&pDev->netdev_ops, pOverride->pOrgOps);
381	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
382	}
383	else
384	pOverride = NULL;
385	}
386	else
387	pOverride = NULL;
388	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
389
390	if (pOverride)
391	{
392	printk("vboxnetflt: dropped %llu out of %llu packets\n", pOverride->cFiltered, pOverride->cTotal);
393	RTMemFree(pOverride);
394	}
395	}
396
397	#endif /* VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT */
398
399
400	/**
401	* Reads and retains the host interface handle.
402	*
403	* @returns The handle, NULL if detached.
404	* @param pThis
405	*/
406	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
407	{
408	#if 0
409	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
410	struct net_device *pDev = NULL;
411
412	Log(("vboxNetFltLinuxRetainNetDev\n"));
413	/*
414	* Be careful here to avoid problems racing the detached callback.
415	*/
416	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
417	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
418	{
419	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
420	if (pDev)
421	{
422	dev_hold(pDev);
423	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n", pDev, pDev->name, atomic_read(&pDev->refcnt)));
424	}
425	}
426	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
427
428	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
429	return pDev;
430	#else
431	return (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
432	#endif
433	}
434
435
436	/**
437	* Release the host interface handle previously retained
438	* by vboxNetFltLinuxRetainNetDev.
439	*
440	* @param pThis The instance.
441	* @param pDev The vboxNetFltLinuxRetainNetDev
442	* return value, NULL is fine.
443	*/
444	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
445	{
446	#if 0
447	Log(("vboxNetFltLinuxReleaseNetDev\n"));
448	NOREF(pThis);
449	if (pDev)
450	{
451	dev_put(pDev);
452	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n", pDev, pDev->name, atomic_read(&pDev->refcnt)));
453	}
454	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
455	#endif
456	}
457
458	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
459	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
460
461	/**
462	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
463	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
464	*
465	* @returns true / false accordingly.
466	* @param pBuf The sk_buff.
467	*/
468	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
469	{
470	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
471	}
472
473
474	/**
475	* Internal worker that create a linux sk_buff for a
476	* (scatter/)gather list.
477	*
478	* @returns Pointer to the sk_buff.
479	* @param pThis The instance.
480	* @param pSG The (scatter/)gather list.
481	* @param fDstWire Set if the destination is the wire.
482	*/
483	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
484	{
485	struct sk_buff *pPkt;
486	struct net_device *pDev;
487	unsigned fGsoType = 0;
488
489	if (pSG->cbTotal == 0)
490	{
491	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
492	return NULL;
493	}
494
495	/** @todo We should use fragments mapping the SG buffers with large packets.
496	* 256 bytes seems to be the a threshold used a lot for this. It
497	* requires some nasty work on the intnet side though... */
498	/*
499	* Allocate a packet and copy over the data.
500	*/
501	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
502	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
503	if (RT_UNLIKELY(!pPkt))
504	{
505	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
506	pSG->pvUserData = NULL;
507	return NULL;
508	}
509	pPkt->dev = pDev;
510	pPkt->ip_summed = CHECKSUM_NONE;
511
512	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
513	skb_reserve(pPkt, NET_IP_ALIGN);
514
515	/* Copy the segments. */
516	skb_put(pPkt, pSG->cbTotal);
517	INTNETSgRead(pSG, pPkt->data);
518
519	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
520	/*
521	* Setup GSO if used by this packet.
522	*/
523	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
524	{
525	default:
526	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
527	/* fall thru */
528	case PDMNETWORKGSOTYPE_INVALID:
529	fGsoType = 0;
530	break;
531	case PDMNETWORKGSOTYPE_IPV4_TCP:
532	fGsoType = SKB_GSO_TCPV4;
533	break;
534	case PDMNETWORKGSOTYPE_IPV4_UDP:
535	fGsoType = SKB_GSO_UDP;
536	break;
537	case PDMNETWORKGSOTYPE_IPV6_TCP:
538	fGsoType = SKB_GSO_TCPV6;
539	break;
540	}
541	if (fGsoType)
542	{
543	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
544
545	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
546	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
547	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
548
549	if (fDstWire)
550	{
551	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrs);
552	pPkt->ip_summed = CHECKSUM_PARTIAL;
553	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
554	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
555	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
556	pPkt->csum_offset = RT_OFFSETOF(RTNETTCP, th_sum);
557	else
558	pPkt->csum_offset = RT_OFFSETOF(RTNETUDP, uh_sum);
559	# else
560	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
561	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
562	pPkt->csum = RT_OFFSETOF(RTNETTCP, th_sum);
563	else
564	pPkt->csum = RT_OFFSETOF(RTNETUDP, uh_sum);
565	# endif
566	}
567	else
568	{
569	pPkt->ip_summed = CHECKSUM_UNNECESSARY;
570	pPkt->csum = 0;
571	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, false /fPayloadChecksum/);
572	}
573	}
574	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
575
576	/*
577	* Finish up the socket buffer.
578	*/
579	pPkt->protocol = eth_type_trans(pPkt, pDev);
580	if (fDstWire)
581	{
582	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
583
584	/* Restore ethernet header back. */
585	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
586	VBOX_SKB_RESET_MAC_HDR(pPkt);
587	}
588	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
589
590	return pPkt;
591	}
592
593
594	/**
595	* Initializes a SG list from an sk_buff.
596	*
597	* @returns Number of segments.
598	* @param pThis The instance.
599	* @param pBuf The sk_buff.
600	* @param pSG The SG.
601	* @param pvFrame The frame pointer, optional.
602	* @param cSegs The number of segments allocated for the SG.
603	* This should match the number in the mbuf exactly!
604	* @param fSrc The source of the frame.
605	* @param pGso Pointer to the GSO context if it's a GSO
606	* internal network frame. NULL if regular frame.
607	*/
608	DECLINLINE(void) vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
609	unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
610	{
611	int i;
612	NOREF(pThis);
613
614	Assert(!skb_shinfo(pBuf)->frag_list);
615
616	if (fSrc & INTNETTRUNKDIR_WIRE)
617	{
618	/*
619	* The packet came from wire, ethernet header was removed by device driver.
620	* Restore it.
621	*/
622	skb_push(pBuf, ETH_HLEN);
623	}
624
625	if (!pGsoCtx)
626	INTNETSgInitTempSegs(pSG, pBuf->len, cSegs, 0 /cSegsUsed/);
627	else
628	INTNETSgInitTempSegsGso(pSG, pBuf->len, cSegs, 0 /cSegsUsed/, pGsoCtx);
629
630	#ifdef VBOXNETFLT_SG_SUPPORT
631	pSG->aSegs[0].cb = skb_headlen(pBuf);
632	pSG->aSegs[0].pv = pBuf->data;
633	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
634
635	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
636	{
637	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
638	pSG->aSegs[i+1].cb = pFrag->size;
639	pSG->aSegs[i+1].pv = kmap(pFrag->page);
640	printk("%p = kmap()\n", pSG->aSegs[i+1].pv);
641	pSG->aSegs[i+1].Phys = NIL_RTHCPHYS;
642	}
643	++i;
644
645	#else
646	pSG->aSegs[0].cb = pBuf->len;
647	pSG->aSegs[0].pv = pBuf->data;
648	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
649	i = 1;
650	#endif
651
652	pSG->cSegsUsed = i;
653
654	#ifdef PADD_RUNT_FRAMES_FROM_HOST
655	/*
656	* Add a trailer if the frame is too small.
657	*
658	* Since we're getting to the packet before it is framed, it has not
659	* yet been padded. The current solution is to add a segment pointing
660	* to a buffer containing all zeros and pray that works for all frames...
661	*/
662	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
663	{
664	static uint8_t const s_abZero[128] = {0};
665
666	AssertReturnVoid(i < cSegs);
667
668	pSG->aSegs[i].Phys = NIL_RTHCPHYS;
669	pSG->aSegs[i].pv = (void *)&s_abZero[0];
670	pSG->aSegs[i].cb = 60 - pSG->cbTotal;
671	pSG->cbTotal = 60;
672	pSG->cSegsUsed++;
673	Assert(i + 1 <= pSG->cSegsAlloc)
674	}
675	#endif
676
677	Log4(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
678	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
679	for (i = 0; i < pSG->cSegsUsed; i++)
680	Log4(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
681	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
682	}
683
684	/**
685	* Packet handler,
686	*
687	* @returns 0 or EJUSTRETURN.
688	* @param pThis The instance.
689	* @param pMBuf The mbuf.
690	* @param pvFrame The start of the frame, optional.
691	* @param fSrc Where the packet (allegedly) comes from, one INTNETTRUNKDIR_* value.
692	* @param eProtocol The protocol.
693	*/
694	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 14)
695	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
696	struct net_device *pSkbDev,
697	struct packet_type *pPacketType,
698	struct net_device *pOrigDev)
699	#else
700	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
701	struct net_device *pSkbDev,
702	struct packet_type *pPacketType)
703	#endif
704	{
705	PVBOXNETFLTINS pThis;
706	struct net_device *pDev;
707	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
708	pBuf, pSkbDev, pPacketType));
709	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
710	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",
711	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));
712	#else
713	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",
714	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));
715	#endif
716	/*
717	* Drop it immediately?
718	*/
719	if (!pBuf)
720	return 0;
721
722	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
723	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
724	if (pThis->u.s.pDev != pSkbDev)
725	{
726	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
727	return 0;
728	}
729
730	Log4(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
731	if (vboxNetFltLinuxSkBufIsOur(pBuf))
732	{
733	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
734	dev_kfree_skb(pBuf);
735	return 0;
736	}
737
738	#ifndef VBOXNETFLT_SG_SUPPORT
739	{
740	/*
741	* Get rid of fragmented packets, they cause too much trouble.
742	*/
743	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
744	kfree_skb(pBuf);
745	if (!pCopy)
746	{
747	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
748	return 0;
749	}
750	pBuf = pCopy;
751	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
752	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",
753	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));
754	# else
755	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",
756	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));
757	# endif
758	}
759	#endif
760
761	/* Add the packet to transmit queue and schedule the bottom half. */
762	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
763	schedule_work(&pThis->u.s.XmitTask);
764	Log4(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
765	&pThis->u.s.XmitTask, pBuf));
766
767	/* It does not really matter what we return, it is ignored by the kernel. */
768	return 0;
769	}
770
771	/**
772	* Calculate the number of INTNETSEG segments the socket buffer will need.
773	*
774	* @returns Segment count.
775	* @param pBuf The socket buffer.
776	*/
777	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff *pBuf)
778	{
779	#ifdef VBOXNETFLT_SG_SUPPORT
780	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
781	#else
782	unsigned cSegs = 1;
783	#endif
784	#ifdef PADD_RUNT_FRAMES_FROM_HOST
785	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
786	if (pBuf->len < 60)
787	cSegs++;
788	#endif
789	return cSegs;
790	}
791
792	/**
793	* Destroy the intnet scatter / gather buffer created by
794	* vboxNetFltLinuxSkBufToSG.
795	*/
796	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG)
797	{
798	#ifdef VBOXNETFLT_SG_SUPPORT
799	int i;
800
801	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
802	{
803	printk("kunmap(%p)\n", pSG->aSegs[i+1].pv);
804	kunmap(pSG->aSegs[i+1].pv);
805	}
806	#endif
807	NOREF(pSG);
808	}
809
810	#ifdef LOG_ENABLED
811	/**
812	* Logging helper.
813	*/
814	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
815	{
816	uint8_t pInt, pExt;
817	static int iPacketNo = 1;
818	iPacketNo += iIncrement;
819	if (fEgress)
820	{
821	pExt = pSG->aSegs[0].pv;
822	pInt = pExt + 6;
823	}
824	else
825	{
826	pInt = pSG->aSegs[0].pv;
827	pExt = pInt + 6;
828	}
829	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
830	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
831	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
832	fEgress ? "-->" : "<--", pszWhere,
833	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
834	pSG->cbTotal, iPacketNo));
835	Log3(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
836	}
837	#else
838	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
839	#endif
840
841	#ifdef VBOXNETFLT_WITH_GSO_RECV
842
843	/**
844	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
845	* GSO socket buffer without having to segment it.
846	*
847	* @returns true on success, false if needs segmenting.
848	* @param pThis The net filter instance.
849	* @param pSkb The GSO socket buffer.
850	* @param fSrc The source.
851	* @param pGsoCtx Where to return the GSO context on success.
852	*/
853	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
854	PPDMNETWORKGSO pGsoCtx)
855	{
856	PDMNETWORKGSOTYPE enmGsoType;
857	uint16_t uEtherType;
858	unsigned int cbTransport;
859	unsigned int offTransport;
860	unsigned int cbTransportHdr;
861	unsigned uProtocol;
862	union
863	{
864	RTNETIPV4 IPv4;
865	RTNETIPV6 IPv6;
866	RTNETTCP Tcp;
867	uint8_t ab[40];
868	uint16_t au16[40/2];
869	uint32_t au32[40/4];
870	} Buf;
871
872	/*
873	* Check the GSO properties of the socket buffer and make sure it fits.
874	*/
875	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
876	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_UDP \| SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
877	{
878	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
879	return false;
880	}
881	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
882	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
883	{
884	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
885	return false;
886	}
887	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
888	{
889	Log5(("vboxNetFltLinuxCanForwardAsGso: fSrc=wire\n"));
890	return false;
891	}
892
893	/*
894	* skb_gso_segment does the following. Do we need to do it as well?
895	*/
896	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
897	skb_reset_mac_header(pSkb);
898	pSkb->mac_len = pSkb->network_header - pSkb->mac_header;
899	#else
900	pSkb->mac.raw = pSkb->data;
901	pSkb->mac_len = pSkb->nh.raw - pSkb->data;
902	#endif
903
904	/*
905	* Switch on the ethertype.
906	*/
907	uEtherType = pSkb->protocol;
908	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
909	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
910	{
911	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
912	if (puEtherType)
913	uEtherType = *puEtherType;
914	}
915	switch (uEtherType)
916	{
917	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
918	{
919	unsigned int cbHdr;
920	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
921	if (RT_UNLIKELY(!pIPv4))
922	{
923	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
924	return false;
925	}
926
927	cbHdr = pIPv4->ip_hl * 4;
928	cbTransport = RT_N2H_U16(pIPv4->ip_len);
929	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
930	\|\| cbHdr > cbTransport ))
931	{
932	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
933	return false;
934	}
935	cbTransport -= cbHdr;
936	offTransport = pSkb->mac_len + cbHdr;
937	uProtocol = pIPv4->ip_p;
938	if (uProtocol == RTNETIPV4_PROT_TCP)
939	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
940	else if (uProtocol == RTNETIPV4_PROT_UDP)
941	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
942	else /** @todo IPv6: 4to6 tunneling */
943	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
944	break;
945	}
946
947	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
948	{
949	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
950	if (RT_UNLIKELY(!pIPv6))
951	{
952	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
953	return false;
954	}
955
956	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
957	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
958	uProtocol = pIPv6->ip6_nxt;
959	/** @todo IPv6: Dig our way out of the other headers. */
960	if (uProtocol == RTNETIPV4_PROT_TCP)
961	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
962	else if (uProtocol == RTNETIPV4_PROT_UDP)
963	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
964	else
965	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
966	break;
967	}
968
969	default:
970	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
971	return false;
972	}
973
974	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
975	{
976	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
977	return false;
978	}
979
980	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
981	\|\| offTransport + cbTransport > pSkb->len
982	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
983	{
984	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
985	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
986	return false;
987	}
988
989	/*
990	* Check the TCP/UDP bits.
991	*/
992	if (uProtocol == RTNETIPV4_PROT_TCP)
993	{
994	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
995	if (RT_UNLIKELY(!pTcp))
996	{
997	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
998	return false;
999	}
1000
1001	cbTransportHdr = pTcp->th_off * 4;
1002	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1003	\|\| cbTransportHdr > cbTransport
1004	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1005	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1006	{
1007	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1008	return false;
1009	}
1010
1011	}
1012	else
1013	{
1014	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1015	cbTransportHdr = sizeof(RTNETUDP);
1016	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1017	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1018	{
1019	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1020	return false;
1021	}
1022	}
1023
1024	/*
1025	* We're good, init the GSO context.
1026	*/
1027	pGsoCtx->u8Type = enmGsoType;
1028	pGsoCtx->cbHdrs = offTransport + cbTransportHdr;
1029	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1030	pGsoCtx->offHdr1 = pSkb->mac_len;
1031	pGsoCtx->offHdr2 = offTransport;
1032	pGsoCtx->au8Unused[0] = 0;
1033	pGsoCtx->au8Unused[1] = 0;
1034
1035	return true;
1036	}
1037
1038	/**
1039	* Forward the socket buffer as a GSO internal network frame.
1040	*
1041	* @returns IPRT status code.
1042	* @param pThis The net filter instance.
1043	* @param pSkb The GSO socket buffer.
1044	* @param fSrc The source.
1045	* @param pGsoCtx Where to return the GSO context on success.
1046	*/
1047	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1048	{
1049	int rc;
1050	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb);
1051	if (RT_LIKELY(cSegs <= MAX_SKB_FRAGS + 1))
1052	{
1053	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1054	if (RT_LIKELY(pSG))
1055	{
1056	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cSegs, fSrc, pGsoCtx);
1057
1058	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1059	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, pSG, fSrc);
1060
1061	vboxNetFltLinuxDestroySG(pSG);
1062	rc = VINF_SUCCESS;
1063	}
1064	else
1065	{
1066	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1067	rc = VERR_NO_MEMORY;
1068	}
1069	}
1070	else
1071	{
1072	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1073	rc = VERR_INTERNAL_ERROR_3;
1074	}
1075
1076	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1077	dev_kfree_skb(pSkb);
1078	return rc;
1079	}
1080
1081	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1082
1083	/**
1084	* Worker for vboxNetFltLinuxForwardToIntNet.
1085	*
1086	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1087	* @param pThis The net filter instance.
1088	* @param pBuf The socket buffer.
1089	* @param fSrc The source.
1090	*/
1091	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1092	{
1093	int rc;
1094	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf);
1095	if (cSegs <= MAX_SKB_FRAGS + 1)
1096	{
1097	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1098	if (RT_LIKELY(pSG))
1099	{
1100	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cSegs, fSrc, NULL /pGsoCtx/);
1101
1102	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1103	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, pSG, fSrc);
1104
1105	vboxNetFltLinuxDestroySG(pSG);
1106	rc = VINF_SUCCESS;
1107	}
1108	else
1109	{
1110	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1111	rc = VERR_NO_MEMORY;
1112	}
1113	}
1114	else
1115	{
1116	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1117	rc = VERR_INTERNAL_ERROR_3;
1118	}
1119
1120	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1121	dev_kfree_skb(pBuf);
1122	return rc;
1123	}
1124
1125	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1126	{
1127	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1128
1129	#ifdef VBOXNETFLT_WITH_GSO
1130	if (skb_is_gso(pBuf))
1131	{
1132	PDMNETWORKGSO GsoCtx;
1133	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",
1134	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));
1135	# ifdef VBOXNETFLT_WITH_GSO_RECV
1136	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_UDP \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1137	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1138	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1139	else
1140	# endif
1141	{
1142	/* Need to segment the packet */
1143	struct sk_buff *pNext;
1144	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1145	if (IS_ERR(pSegment))
1146	{
1147	dev_kfree_skb(pBuf);
1148	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1149	return;
1150	}
1151
1152	for (; pSegment; pSegment = pNext)
1153	{
1154	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",
1155	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));
1156	pNext = pSegment->next;
1157	pSegment->next = 0;
1158	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1159	}
1160	dev_kfree_skb(pBuf);
1161	}
1162	}
1163	else
1164	#endif /* VBOXNETFLT_WITH_GSO */
1165	{
1166	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1167	{
1168	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1169	/*
1170	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1171	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1172	* header length from the header itself and reconstruct 'h' pointer
1173	* to TCP (or whatever) header.
1174	*/
1175	unsigned char *tmp = pBuf->h.raw;
1176	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1177	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1178	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1179	if (VBOX_SKB_CHECKSUM_HELP(pBuf))
1180	{
1181	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1182	dev_kfree_skb(pBuf);
1183	return;
1184	}
1185	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1186	/* Restore the original (wrong) pointer. */
1187	pBuf->h.raw = tmp;
1188	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1189	}
1190	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1191	}
1192	}
1193
1194	/**
1195	* Work queue handler that forwards the socket buffers queued by
1196	* vboxNetFltLinuxPacketHandler to the internal network.
1197	*
1198	* @param pWork The work queue.
1199	*/
1200	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
1201	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1202	#else
1203	static void vboxNetFltLinuxXmitTask(void *pWork)
1204	#endif
1205	{
1206	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1207	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1208	struct sk_buff *pBuf;
1209
1210	Log4(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1211
1212	/*
1213	* Active? Retain the instance and increment the busy counter.
1214	*/
1215	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1216	if (ASMAtomicUoReadBool(&pThis->fActive))
1217	{
1218	vboxNetFltRetain(pThis, true /* fBusy */);
1219	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1220
1221	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1222	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1223
1224	vboxNetFltRelease(pThis, true /* fBusy */);
1225	}
1226	else
1227	{
1228	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1229	/** @todo Shouldn't we just drop the packets here? There is little point in
1230	* making them accumulate when the VM is paused and it'll only waste
1231	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1232	* before start draining the packets (goes for the intnet ring buf
1233	* too)? */
1234	}
1235	}
1236
1237	/**
1238	* Reports the GSO capabilites of the hardware NIC.
1239	*
1240	* @param pThis The net filter instance. The caller hold a
1241	* reference to this.
1242	*/
1243	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1244	{
1245	#ifdef VBOXNETFLT_WITH_GSO_XMIT_WIRE
1246	struct net_device *pDev;
1247	PINTNETTRUNKSWPORT pSwitchPort;
1248	unsigned int fFeatures;
1249	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1250
1251	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1252
1253	pSwitchPort = pThis->pSwitchPort; /* this doesn't need to be here, but it doesn't harm. */
1254	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
1255	if (pDev)
1256	fFeatures = pDev->features;
1257	else
1258	fFeatures = 0;
1259
1260	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1261
1262	if (pThis->pSwitchPort)
1263	{
1264	/* Set/update the GSO capabilities of the NIC. */
1265	uint32_t fGsoCapabilites = 0;
1266	if (fFeatures & NETIF_F_TSO)
1267	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1268	if (fFeatures & NETIF_F_TSO6)
1269	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1270	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
1271	if (fFeatures & NETIF_F_UFO)
1272	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP);
1273	if (fFeatures & NETIF_F_UFO)
1274	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP);
1275	# endif
1276	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1277	}
1278	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE */
1279	}
1280
1281	/**
1282	* Internal worker for vboxNetFltLinuxNotifierCallback.
1283	*
1284	* @returns VBox status code.
1285	* @param pThis The instance.
1286	* @param fRediscovery If set we're doing a rediscovery attempt, so, don't
1287	* flood the release log.
1288	*/
1289	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1290	{
1291	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1292	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1293
1294	/*
1295	* Retain and store the device.
1296	*/
1297	dev_hold(pDev);
1298
1299	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1300	ASMAtomicUoWritePtr((void * volatile *)&pThis->u.s.pDev, pDev);
1301	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1302
1303	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n", pDev, pDev->name, atomic_read(&pDev->refcnt)));
1304	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n", pDev, pThis, ASMAtomicUoReadPtr((void * volatile *)&pThis->u.s.pDev)));
1305
1306	/* Get the mac address while we still have a valid net_device reference. */
1307	memcpy(&pThis->u.s.Mac, pDev->dev_addr, sizeof(pThis->u.s.Mac));
1308
1309	/*
1310	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1311	*/
1312	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1313	pThis->u.s.PacketType.dev = pDev;
1314	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1315	dev_add_pack(&pThis->u.s.PacketType);
1316
1317	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
1318	vboxNetFltLinuxHookDev(pThis, pDev);
1319	#endif
1320
1321	/*
1322	* Set indicators that require the spinlock. Be abit paranoid about racing
1323	* the device notification handle.
1324	*/
1325	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1326	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
1327	if (pDev)
1328	{
1329	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1330	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1331	pDev = NULL; /* don't dereference it */
1332	}
1333	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1334	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1335
1336	/*
1337	* If the above succeeded report GSO capabilites, if not undo and
1338	* release the device.
1339	*/
1340	if (!pDev)
1341	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1342	else
1343	{
1344	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
1345	vboxNetFltLinuxUnhookDev(pThis, pDev);
1346	#endif
1347	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1348	ASMAtomicUoWritePtr((void * volatile *)&pThis->u.s.pDev, NULL);
1349	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1350	dev_put(pDev);
1351	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) released. ref=%d\n", pDev, pDev->name, atomic_read(&pDev->refcnt)));
1352	}
1353
1354	LogRel(("VBoxNetFlt: attached to '%s' / %.*Rhxs\n", pThis->szName, sizeof(pThis->u.s.Mac), &pThis->u.s.Mac));
1355	return VINF_SUCCESS;
1356	}
1357
1358
1359	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1360	{
1361	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1362
1363	Assert(!pThis->fDisconnectedFromHost);
1364
1365	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
1366	vboxNetFltLinuxUnhookDev(pThis, pDev);
1367	#endif
1368
1369	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1370	ASMAtomicWriteBool(&pThis->u.s.fRegistered, false);
1371	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1372	ASMAtomicUoWritePtr((void * volatile *)&pThis->u.s.pDev, NULL);
1373	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1374
1375	dev_remove_pack(&pThis->u.s.PacketType);
1376	skb_queue_purge(&pThis->u.s.XmitQueue);
1377	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
1378	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n", pDev, pDev->name, atomic_read(&pDev->refcnt)));
1379	dev_put(pDev);
1380
1381	return NOTIFY_OK;
1382	}
1383
1384	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1385	{
1386	/* Check if we are not suspended and promiscuous mode has not been set. */
1387	if (ASMAtomicUoReadBool(&pThis->fActive) && !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1388	{
1389	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1390	dev_set_promiscuity(pDev, 1);
1391	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1392	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1393	}
1394	else
1395	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1396	return NOTIFY_OK;
1397	}
1398
1399	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1400	{
1401	/* Undo promiscuous mode if we has set it. */
1402	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1403	{
1404	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1405	dev_set_promiscuity(pDev, -1);
1406	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1407	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1408	}
1409	else
1410	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1411	return NOTIFY_OK;
1412	}
1413
1414	#ifdef LOG_ENABLED
1415	/** Stringify the NETDEV_XXX constants. */
1416	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
1417	{
1418	const char *pszEvent = "NETDRV_<unknown>";
1419	switch (ulEventType)
1420	{
1421	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
1422	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
1423	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
1424	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
1425	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
1426	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
1427	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
1428	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
1429	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
1430	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
1431	# ifdef NETDEV_FEAT_CHANGE
1432	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
1433	# endif
1434	}
1435	return pszEvent;
1436	}
1437	#endif /* LOG_ENABLED */
1438
1439	/**
1440	* Callback for listening to netdevice events.
1441	*
1442	* This works the rediscovery, clean up on unregistration, promiscuity on
1443	* up/down, and GSO feature changes from ethtool.
1444	*
1445	* @returns NOTIFY_OK
1446	* @param self Pointer to our notifier registration block.
1447	* @param ulEventType The event.
1448	* @param ptr Event specific, but it is usually the device it
1449	* relates to.
1450	*/
1451	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
1452
1453	{
1454	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
1455	struct net_device pDev = (struct net_device )ptr;
1456	int rc = NOTIFY_OK;
1457
1458	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1459	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, ASMAtomicUoReadPtr((void * volatile *)&pThis->u.s.pDev)));
1460	if ( ulEventType == NETDEV_REGISTER
1461	&& !strcmp(pDev->name, pThis->szName))
1462	{
1463	vboxNetFltLinuxAttachToInterface(pThis, pDev);
1464	}
1465	else
1466	{
1467	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
1468	if (pDev == ptr)
1469	{
1470	switch (ulEventType)
1471	{
1472	case NETDEV_UNREGISTER:
1473	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
1474	break;
1475	case NETDEV_UP:
1476	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
1477	break;
1478	case NETDEV_GOING_DOWN:
1479	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
1480	break;
1481	case NETDEV_CHANGENAME:
1482	break;
1483	#ifdef NETDEV_FEAT_CHANGE
1484	case NETDEV_FEAT_CHANGE:
1485	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1486	break;
1487	#endif
1488	}
1489	}
1490	}
1491
1492	return rc;
1493	}
1494
1495	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
1496	{
1497	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
1498	}
1499
1500	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, PINTNETSG pSG, uint32_t fDst)
1501	{
1502	struct net_device * pDev;
1503	int err;
1504	int rc = VINF_SUCCESS;
1505
1506	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
1507
1508	pDev = vboxNetFltLinuxRetainNetDev(pThis);
1509	if (pDev)
1510	{
1511	/*
1512	* Create a sk_buff for the gather list and push it onto the wire.
1513	*/
1514	if (fDst & INTNETTRUNKDIR_WIRE)
1515	{
1516	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
1517	if (pBuf)
1518	{
1519	vboxNetFltDumpPacket(pSG, true, "wire", 1);
1520	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1521	Log4(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
1522	err = dev_queue_xmit(pBuf);
1523	if (err)
1524	rc = RTErrConvertFromErrno(err);
1525	}
1526	else
1527	rc = VERR_NO_MEMORY;
1528	}
1529
1530	/*
1531	* Create a sk_buff for the gather list and push it onto the host stack.
1532	*/
1533	if (fDst & INTNETTRUNKDIR_HOST)
1534	{
1535	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
1536	if (pBuf)
1537	{
1538	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
1539	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1540	Log4(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
1541	err = netif_rx_ni(pBuf);
1542	if (err)
1543	rc = RTErrConvertFromErrno(err);
1544	}
1545	else
1546	rc = VERR_NO_MEMORY;
1547	}
1548
1549	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1550	}
1551
1552	return rc;
1553	}
1554
1555
1556	bool vboxNetFltPortOsIsPromiscuous(PVBOXNETFLTINS pThis)
1557	{
1558	bool fRc = false;
1559	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1560	if (pDev)
1561	{
1562	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1563	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1564	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1565	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1566	}
1567	return fRc;
1568	}
1569
1570
1571	void vboxNetFltPortOsGetMacAddress(PVBOXNETFLTINS pThis, PRTMAC pMac)
1572	{
1573	*pMac = pThis->u.s.Mac;
1574	}
1575
1576
1577	bool vboxNetFltPortOsIsHostMac(PVBOXNETFLTINS pThis, PCRTMAC pMac)
1578	{
1579	/* ASSUMES that the MAC address never changes. */
1580	return pThis->u.s.Mac.au16[0] == pMac->au16[0]
1581	&& pThis->u.s.Mac.au16[1] == pMac->au16[1]
1582	&& pThis->u.s.Mac.au16[2] == pMac->au16[2];
1583	}
1584
1585
1586	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
1587	{
1588	struct net_device * pDev;
1589
1590	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%s, fDisablePromiscuous=%s\n",
1591	pThis, pThis->szName, fActive?"true":"false",
1592	pThis->fDisablePromiscuous?"true":"false"));
1593
1594	if (pThis->fDisablePromiscuous)
1595	return;
1596
1597	pDev = vboxNetFltLinuxRetainNetDev(pThis);
1598	if (pDev)
1599	{
1600	/*
1601	* This api is a bit weird, the best reference is the code.
1602	*
1603	* Also, we have a bit or race conditions wrt the maintance of
1604	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
1605	*/
1606	#ifdef LOG_ENABLED
1607	u_int16_t fIf;
1608	unsigned const cPromiscBefore = pDev->promiscuity;
1609	#endif
1610	if (fActive)
1611	{
1612	Assert(!pThis->u.s.fPromiscuousSet);
1613
1614	rtnl_lock();
1615	dev_set_promiscuity(pDev, 1);
1616	rtnl_unlock();
1617	pThis->u.s.fPromiscuousSet = true;
1618	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1619	}
1620	else
1621	{
1622	if (pThis->u.s.fPromiscuousSet)
1623	{
1624	rtnl_lock();
1625	dev_set_promiscuity(pDev, -1);
1626	rtnl_unlock();
1627	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1628	}
1629	pThis->u.s.fPromiscuousSet = false;
1630
1631	#ifdef LOG_ENABLED
1632	fIf = dev_get_flags(pDev);
1633	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
1634	#endif
1635	}
1636
1637	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1638	}
1639	}
1640
1641
1642	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
1643	{
1644	/* Nothing to do here. */
1645	return VINF_SUCCESS;
1646	}
1647
1648
1649	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
1650	{
1651	/*
1652	* Report the GSO capabilities of the host and device (if connected).
1653	*/
1654	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1655	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
1656	0
1657	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
1658	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
1659	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
1660	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP)
1661	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP)
1662	# endif
1663	, INTNETTRUNKDIR_HOST);
1664
1665	#endif
1666	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1667
1668	return VINF_SUCCESS;
1669	}
1670
1671
1672	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
1673	{
1674	struct net_device *pDev;
1675	bool fRegistered;
1676	RTSPINLOCKTMP Tmp = RTSPINLOCKTMP_INITIALIZER;
1677
1678	#ifdef VBOXNETFLT_WITH_FILTER_HOST2GUEST_SKBS_EXPERIMENT
1679	vboxNetFltLinuxUnhookDev(pThis, NULL);
1680	#endif
1681
1682	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
1683	* unlikely, but none the less). Since it doesn't actually update the
1684	* state (just reads it), it is likely to panic in some interesting
1685	* ways. */
1686
1687	RTSpinlockAcquire(pThis->hSpinlock, &Tmp);
1688	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
1689	fRegistered = ASMAtomicUoReadBool(&pThis->u.s.fRegistered);
1690	RTSpinlockRelease(pThis->hSpinlock, &Tmp);
1691
1692	if (fRegistered)
1693	{
1694	dev_remove_pack(&pThis->u.s.PacketType);
1695	skb_queue_purge(&pThis->u.s.XmitQueue);
1696	Log(("vboxNetFltOsDeleteInstance: this=%p: Packet handler removed, xmit queue purged.\n", pThis));
1697	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n", pDev, pDev->name, atomic_read(&pDev->refcnt)));
1698	dev_put(pDev);
1699	}
1700	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
1701	unregister_netdevice_notifier(&pThis->u.s.Notifier);
1702	module_put(THIS_MODULE);
1703	}
1704
1705
1706	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
1707	{
1708	int err;
1709	NOREF(pvContext);
1710
1711	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
1712	err = register_netdevice_notifier(&pThis->u.s.Notifier);
1713	if (err)
1714	return VERR_INTNET_FLT_IF_FAILED;
1715	if (!pThis->u.s.fRegistered)
1716	{
1717	unregister_netdevice_notifier(&pThis->u.s.Notifier);
1718	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
1719	return VERR_INTNET_FLT_IF_NOT_FOUND;
1720	}
1721
1722	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
1723	if ( pThis->fDisconnectedFromHost
1724	\|\| !try_module_get(THIS_MODULE))
1725	return VERR_INTNET_FLT_IF_FAILED;
1726
1727	return VINF_SUCCESS;
1728	}
1729
1730	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
1731	{
1732	/*
1733	* Init the linux specific members.
1734	*/
1735	pThis->u.s.pDev = NULL;
1736	pThis->u.s.fRegistered = false;
1737	pThis->u.s.fPromiscuousSet = false;
1738	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
1739	skb_queue_head_init(&pThis->u.s.XmitQueue);
1740	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
1741	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
1742	#else
1743	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
1744	#endif
1745
1746	return VINF_SUCCESS;
1747	}
1748

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

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 28281

Download in other formats: