VBoxNetFlt-linux.c@ 55998

Last change on this file since 55998 was 55685, checked in by vboxsync, 10 years ago
VBoxNetFlt/linux: inet6_dev::addr_list became struct list_head in 2.6.35.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 79.6 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 55685 2015-05-06 03:32:48Z vboxsync $ */
2	/** @file
3	* VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4	*/
5
6	/*
7	* Copyright (C) 2006-2014 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/inetdevice.h>
31	#include <linux/ip.h>
32	#include <linux/if_vlan.h>
33	#include <net/if_inet6.h>
34	#include <net/addrconf.h>
35
36	#include <VBox/log.h>
37	#include <VBox/err.h>
38	#include <VBox/intnetinline.h>
39	#include <VBox/vmm/pdmnetinline.h>
40	#include <VBox/param.h>
41	#include <iprt/alloca.h>
42	#include <iprt/assert.h>
43	#include <iprt/spinlock.h>
44	#include <iprt/semaphore.h>
45	#include <iprt/initterm.h>
46	#include <iprt/process.h>
47	#include <iprt/mem.h>
48	#include <iprt/net.h>
49	#include <iprt/log.h>
50	#include <iprt/mp.h>
51	#include <iprt/mem.h>
52	#include <iprt/time.h>
53
54	#define VBOXNETFLT_OS_SPECFIC 1
55	#include "../VBoxNetFltInternal.h"
56
57	#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 20, 0)
58	# define vlan_tx_tag_get(skb) skb_vlan_tag_get(skb)
59	# define vlan_tx_tag_present(skb) skb_vlan_tag_present(skb)
60	#endif
61
62
63	/*******************************************************************************
64	* Defined Constants And Macros *
65	*******************************************************************************/
66	#define VBOX_FLT_NB_TO_INST(pNB) RT_FROM_MEMBER(pNB, VBOXNETFLTINS, u.s.Notifier)
67	#define VBOX_FLT_PT_TO_INST(pPT) RT_FROM_MEMBER(pPT, VBOXNETFLTINS, u.s.PacketType)
68	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
69	# define VBOX_FLT_XT_TO_INST(pXT) RT_FROM_MEMBER(pXT, VBOXNETFLTINS, u.s.XmitTask)
70	#endif
71
72	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 34)
73	# define VBOX_NETDEV_NAME(dev) netdev_name(dev)
74	#else
75	# define VBOX_NETDEV_NAME(dev) ((dev)->reg_state != NETREG_REGISTERED ? "(unregistered net_device)" : (dev)->name)
76	#endif
77
78	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 26)
79	# define VBOX_DEV_NET(dev) dev_net(dev)
80	#else
81	# define VBOX_DEV_NET(dev) ((dev)->nd_net)
82	#endif
83
84	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
85	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
86	# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
87	#else
88	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
89	# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
90	#endif
91
92	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 19)
93	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
94	#else
95	# define CHECKSUM_PARTIAL CHECKSUM_HW
96	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 10)
97	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
98	# else
99	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 7)
100	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
101	# else
102	# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
103	# endif
104	/* Versions prior 2.6.10 use stats for both bstats and qstats */
105	# define bstats stats
106	# define qstats stats
107	# endif
108	#endif
109
110	#ifndef NET_IP_ALIGN
111	# define NET_IP_ALIGN 2
112	#endif
113
114	#if 0
115	/** Create scatter / gather segments for fragments. When not used, we will
116	* linearize the socket buffer before creating the internal networking SG. */
117	# define VBOXNETFLT_SG_SUPPORT 1
118	#endif
119
120	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
121	/** Indicates that the linux kernel may send us GSO frames. */
122	# define VBOXNETFLT_WITH_GSO 1
123
124	/** This enables or disables the transmitting of GSO frame from the internal
125	* network and to the host. */
126	# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
127
128	# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
129	/** This enables or disables the transmitting of GSO frame from the internal
130	* network and to the wire. */
131	# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
132	# endif
133
134	/** This enables or disables the forwarding/flooding of GSO frame from the host
135	* to the internal network. */
136	# define VBOXNETFLT_WITH_GSO_RECV 1
137
138	#endif
139
140	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
141	/** This enables or disables handling of GSO frames coming from the wire (GRO). */
142	# define VBOXNETFLT_WITH_GRO 1
143	#endif
144	/*
145	* GRO support was backported to RHEL 5.4
146	*/
147	#ifdef RHEL_RELEASE_CODE
148	# if RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(5, 4)
149	# define VBOXNETFLT_WITH_GRO 1
150	# endif
151	#endif
152
153	/*******************************************************************************
154	* Internal Functions *
155	*******************************************************************************/
156	static int VBoxNetFltLinuxInit(void);
157	static void VBoxNetFltLinuxUnload(void);
158	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf);
159
160
161	/*******************************************************************************
162	* Global Variables *
163	*******************************************************************************/
164	/**
165	* The (common) global data.
166	*/
167	static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
168
169	module_init(VBoxNetFltLinuxInit);
170	module_exit(VBoxNetFltLinuxUnload);
171
172	MODULE_AUTHOR(VBOX_VENDOR);
173	MODULE_DESCRIPTION(VBOX_PRODUCT " Network Filter Driver");
174	MODULE_LICENSE("GPL");
175	#ifdef MODULE_VERSION
176	MODULE_VERSION(VBOX_VERSION_STRING " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
177	#endif
178
179
180	#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) && defined(LOG_ENABLED)
181	unsigned dev_get_flags(const struct net_device *dev)
182	{
183	unsigned flags;
184
185	flags = (dev->flags & ~(IFF_PROMISC \|
186	IFF_ALLMULTI \|
187	IFF_RUNNING)) \|
188	(dev->gflags & (IFF_PROMISC \|
189	IFF_ALLMULTI));
190
191	if (netif_running(dev) && netif_carrier_ok(dev))
192	flags \|= IFF_RUNNING;
193
194	return flags;
195	}
196	#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) */
197
198
199	/**
200	* Initialize module.
201	*
202	* @returns appropriate status code.
203	*/
204	static int __init VBoxNetFltLinuxInit(void)
205	{
206	int rc;
207	/*
208	* Initialize IPRT.
209	*/
210	rc = RTR0Init(0);
211	if (RT_SUCCESS(rc))
212	{
213	Log(("VBoxNetFltLinuxInit\n"));
214
215	/*
216	* Initialize the globals and connect to the support driver.
217	*
218	* This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
219	* for establishing the connect to the support driver.
220	*/
221	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
222	rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
223	if (RT_SUCCESS(rc))
224	{
225	LogRel(("VBoxNetFlt: Successfully started.\n"));
226	return 0;
227	}
228
229	LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
230	RTR0Term();
231	}
232	else
233	LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
234
235	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
236	return -RTErrConvertToErrno(rc);
237	}
238
239
240	/**
241	* Unload the module.
242	*
243	* @todo We have to prevent this if we're busy!
244	*/
245	static void __exit VBoxNetFltLinuxUnload(void)
246	{
247	int rc;
248	Log(("VBoxNetFltLinuxUnload\n"));
249	Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
250
251	/*
252	* Undo the work done during start (in reverse order).
253	*/
254	rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
255	AssertRC(rc); NOREF(rc);
256
257	RTR0Term();
258
259	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
260
261	Log(("VBoxNetFltLinuxUnload - done\n"));
262	}
263
264
265	/**
266	* We filter traffic from the host to the internal network
267	* before it reaches the NIC driver.
268	*
269	* The current code uses a very ugly hack overriding hard_start_xmit
270	* callback in the device structure, but it has been shown to give us a
271	* performance boost of 60-100% though. Eventually we have to find some
272	* less hacky way of getting this job done.
273	*/
274	#define VBOXNETFLT_WITH_HOST2WIRE_FILTER
275
276	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
277
278	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
279
280	# include <linux/ethtool.h>
281
282	typedef struct ethtool_ops OVR_OPSTYPE;
283	# define OVR_OPS ethtool_ops
284	# define OVR_XMIT pfnStartXmit
285
286	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
287
288	typedef struct net_device_ops OVR_OPSTYPE;
289	# define OVR_OPS netdev_ops
290	# define OVR_XMIT pOrgOps->ndo_start_xmit
291
292	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
293
294	/**
295	* The overridden net_device_ops of the device we're attached to.
296	*
297	* As there is no net_device_ops structure in pre-2.6.29 kernels we override
298	* ethtool_ops instead along with hard_start_xmit callback in net_device
299	* structure.
300	*
301	* This is a very dirty hack that was created to explore how much we can improve
302	* the host to guest transfers by not CC'ing the NIC. It turns out to be
303	* the only way to filter outgoing packets for devices without TX queue.
304	*/
305	typedef struct VBoxNetDeviceOpsOverride
306	{
307	/** Our overridden ops. */
308	OVR_OPSTYPE Ops;
309	/** Magic word. */
310	uint32_t u32Magic;
311	/** Pointer to the original ops. */
312	OVR_OPSTYPE const *pOrgOps;
313	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
314	/** Pointer to the original hard_start_xmit function. */
315	int (pfnStartXmit)(struct sk_buff pSkb, struct net_device *pDev);
316	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
317	/** Pointer to the net filter instance. */
318	PVBOXNETFLTINS pVBoxNetFlt;
319	/** The number of filtered packages. */
320	uint64_t cFiltered;
321	/** The total number of packets */
322	uint64_t cTotal;
323	} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
324	/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
325	#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
326
327	/**
328	* ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
329	* because they belong on the internal network.
330	*
331	* @returns NETDEV_TX_XXX.
332	* @param pSkb The socket buffer to transmit.
333	* @param pDev The net device.
334	*/
335	static int vboxNetFltLinuxStartXmitFilter(struct sk_buff pSkb, struct net_device pDev)
336	{
337	PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
338	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
339	PCRTNETETHERHDR pEtherHdr;
340	PINTNETTRUNKSWPORT pSwitchPort;
341	uint32_t cbHdrs;
342
343
344	/*
345	* Validate the override structure.
346	*
347	* Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
348	* to be production quality code, we would have to be much more
349	* careful here and avoid the race.
350	*/
351	if ( !VALID_PTR(pOverride)
352	\|\| pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
353	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
354	\|\| !VALID_PTR(pOverride->pOrgOps)
355	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
356	)
357	{
358	printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
359	dev_kfree_skb(pSkb);
360	return NETDEV_TX_OK;
361	}
362	pOverride->cTotal++;
363
364	/*
365	* Do the filtering base on the default OUI of our virtual NICs
366	*
367	* Note! In a real solution, we would ask the switch whether the
368	* destination MAC is 100% to be on the internal network and then
369	* drop it.
370	*/
371	cbHdrs = skb_headlen(pSkb);
372	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
373	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
374	if ( pEtherHdr
375	&& VALID_PTR(pOverride->pVBoxNetFlt)
376	&& (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
377	&& VALID_PTR(pSwitchPort)
378	&& cbHdrs >= 6)
379	{
380	INTNETSWDECISION enmDecision;
381
382	/** @todo consider reference counting, etc. */
383	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
384	if (enmDecision == INTNETSWDECISION_INTNET)
385	{
386	dev_kfree_skb(pSkb);
387	pOverride->cFiltered++;
388	return NETDEV_TX_OK;
389	}
390	}
391
392	return pOverride->OVR_XMIT(pSkb, pDev);
393	}
394
395	/**
396	* Hooks the device ndo_start_xmit operation of the device.
397	*
398	* @param pThis The net filter instance.
399	* @param pDev The net device.
400	*/
401	static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
402	{
403	PVBOXNETDEVICEOPSOVERRIDE pOverride;
404
405	/* Cancel override if ethtool_ops is missing (host-only case, @bugref{5712}) */
406	if (!VALID_PTR(pDev->OVR_OPS))
407	return;
408	pOverride = RTMemAlloc(sizeof(*pOverride));
409	if (!pOverride)
410	return;
411	pOverride->pOrgOps = pDev->OVR_OPS;
412	pOverride->Ops = *pDev->OVR_OPS;
413	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
414	pOverride->pfnStartXmit = pDev->hard_start_xmit;
415	# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
416	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
417	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
418	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
419	pOverride->cTotal = 0;
420	pOverride->cFiltered = 0;
421	pOverride->pVBoxNetFlt = pThis;
422
423	RTSpinlockAcquire(pThis->hSpinlock); /* (this isn't necessary, but so what) */
424	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
425	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
426	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
427	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
428	RTSpinlockRelease(pThis->hSpinlock);
429	}
430
431	/**
432	* Undos what vboxNetFltLinuxHookDev did.
433	*
434	* @param pThis The net filter instance.
435	* @param pDev The net device. Can be NULL, in which case
436	* we'll try retrieve it from @a pThis.
437	*/
438	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
439	{
440	PVBOXNETDEVICEOPSOVERRIDE pOverride;
441
442	RTSpinlockAcquire(pThis->hSpinlock);
443	if (!pDev)
444	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
445	if (VALID_PTR(pDev))
446	{
447	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
448	if ( VALID_PTR(pOverride)
449	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
450	&& VALID_PTR(pOverride->pOrgOps)
451	)
452	{
453	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
454	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
455	# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
456	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
457	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
458	}
459	else
460	pOverride = NULL;
461	}
462	else
463	pOverride = NULL;
464	RTSpinlockRelease(pThis->hSpinlock);
465
466	if (pOverride)
467	{
468	printk("vboxnetflt: %llu out of %llu packets were not sent (directed to host)\n", pOverride->cFiltered, pOverride->cTotal);
469	RTMemFree(pOverride);
470	}
471	}
472
473	#endif /* VBOXNETFLT_WITH_HOST2WIRE_FILTER */
474
475
476	/**
477	* Reads and retains the host interface handle.
478	*
479	* @returns The handle, NULL if detached.
480	* @param pThis
481	*/
482	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
483	{
484	#if 0
485	struct net_device *pDev = NULL;
486
487	Log(("vboxNetFltLinuxRetainNetDev\n"));
488	/*
489	* Be careful here to avoid problems racing the detached callback.
490	*/
491	RTSpinlockAcquire(pThis->hSpinlock);
492	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
493	{
494	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
495	if (pDev)
496	{
497	dev_hold(pDev);
498	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
499	pDev, pDev->name,
500	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
501	netdev_refcnt_read(pDev)
502	#else
503	atomic_read(&pDev->refcnt)
504	#endif
505	));
506	}
507	}
508	RTSpinlockRelease(pThis->hSpinlock);
509
510	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
511	return pDev;
512	#else
513	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
514	#endif
515	}
516
517
518	/**
519	* Release the host interface handle previously retained
520	* by vboxNetFltLinuxRetainNetDev.
521	*
522	* @param pThis The instance.
523	* @param pDev The vboxNetFltLinuxRetainNetDev
524	* return value, NULL is fine.
525	*/
526	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
527	{
528	#if 0
529	Log(("vboxNetFltLinuxReleaseNetDev\n"));
530	NOREF(pThis);
531	if (pDev)
532	{
533	dev_put(pDev);
534	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
535	pDev, pDev->name,
536	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
537	netdev_refcnt_read(pDev)
538	#else
539	atomic_read(&pDev->refcnt)
540	#endif
541	));
542	}
543	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
544	#endif
545	}
546
547	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
548	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
549
550	/**
551	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
552	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
553	*
554	* @returns true / false accordingly.
555	* @param pBuf The sk_buff.
556	*/
557	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
558	{
559	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
560	}
561
562
563	/**
564	* Checks whether this SG list contains a GSO packet.
565	*
566	* @returns true / false accordingly.
567	* @param pSG The (scatter/)gather list.
568	*/
569	DECLINLINE(bool) vboxNetFltLinuxIsGso(PINTNETSG pSG)
570	{
571	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
572	return !((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type == PDMNETWORKGSOTYPE_INVALID);
573	#else /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
574	return false;
575	#endif /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
576	}
577
578
579	/**
580	* Find out the frame size (of a single segment in case of GSO frames).
581	*
582	* @returns the frame size.
583	* @param pSG The (scatter/)gather list.
584	*/
585	DECLINLINE(uint32_t) vboxNetFltLinuxFrameSize(PINTNETSG pSG)
586	{
587	uint16_t u16Type = 0;
588	uint32_t cbVlanTag = 0;
589	if (pSG->aSegs[0].cb >= sizeof(RTNETETHERHDR))
590	u16Type = RT_BE2H_U16(((PCRTNETETHERHDR)pSG->aSegs[0].pv)->EtherType);
591	else if (pSG->cbTotal >= sizeof(RTNETETHERHDR))
592	{
593	uint32_t off = RT_OFFSETOF(RTNETETHERHDR, EtherType);
594	uint32_t i;
595	for (i = 0; i < pSG->cSegsUsed; ++i)
596	{
597	if (off <= pSG->aSegs[i].cb)
598	{
599	if (off + sizeof(uint16_t) <= pSG->aSegs[i].cb)
600	u16Type = RT_BE2H_U16((uint16_t )((uintptr_t)pSG->aSegs[i].pv + off));
601	else if (i + 1 < pSG->cSegsUsed)
602	u16Type = RT_BE2H_U16( ((uint16_t)( ((uint8_t *)pSG->aSegs[i].pv)[off] ) << 8)
603	+ (uint8_t )pSG->aSegs[i + 1].pv); /* ASSUMES no empty segments! */
604	/* else: frame is too short. */
605	break;
606	}
607	off -= pSG->aSegs[i].cb;
608	}
609	}
610	if (u16Type == RTNET_ETHERTYPE_VLAN)
611	cbVlanTag = 4;
612	return (vboxNetFltLinuxIsGso(pSG) ? (uint32_t)pSG->GsoCtx.cbMaxSeg + pSG->GsoCtx.cbHdrsTotal : pSG->cbTotal) - cbVlanTag;
613	}
614
615
616	/**
617	* Internal worker that create a linux sk_buff for a
618	* (scatter/)gather list.
619	*
620	* @returns Pointer to the sk_buff.
621	* @param pThis The instance.
622	* @param pSG The (scatter/)gather list.
623	* @param fDstWire Set if the destination is the wire.
624	*/
625	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
626	{
627	struct sk_buff *pPkt;
628	struct net_device *pDev;
629	unsigned fGsoType = 0;
630
631	if (pSG->cbTotal == 0)
632	{
633	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
634	return NULL;
635	}
636	Log5(("VBoxNetFlt: Packet to %s of %d bytes (frame=%d).\n", fDstWire?"wire":"host", pSG->cbTotal, vboxNetFltLinuxFrameSize(pSG)));
637	if (fDstWire && (vboxNetFltLinuxFrameSize(pSG) > ASMAtomicReadU32(&pThis->u.s.cbMtu) + 14))
638	{
639	static bool s_fOnce = true;
640	if (s_fOnce)
641	{
642	s_fOnce = false;
643	printk("VBoxNetFlt: Dropped over-sized packet (%d bytes) coming from internal network.\n", vboxNetFltLinuxFrameSize(pSG));
644	}
645	return NULL;
646	}
647
648	/** @todo We should use fragments mapping the SG buffers with large packets.
649	* 256 bytes seems to be the a threshold used a lot for this. It
650	* requires some nasty work on the intnet side though... */
651	/*
652	* Allocate a packet and copy over the data.
653	*/
654	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
655	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
656	if (RT_UNLIKELY(!pPkt))
657	{
658	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
659	pSG->pvUserData = NULL;
660	return NULL;
661	}
662	pPkt->dev = pDev;
663	pPkt->ip_summed = CHECKSUM_NONE;
664
665	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
666	skb_reserve(pPkt, NET_IP_ALIGN);
667
668	/* Copy the segments. */
669	skb_put(pPkt, pSG->cbTotal);
670	IntNetSgRead(pSG, pPkt->data);
671
672	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
673	/*
674	* Setup GSO if used by this packet.
675	*/
676	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
677	{
678	default:
679	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
680	/* fall thru */
681	case PDMNETWORKGSOTYPE_INVALID:
682	fGsoType = 0;
683	break;
684	case PDMNETWORKGSOTYPE_IPV4_TCP:
685	fGsoType = SKB_GSO_TCPV4;
686	break;
687	case PDMNETWORKGSOTYPE_IPV4_UDP:
688	fGsoType = SKB_GSO_UDP;
689	break;
690	case PDMNETWORKGSOTYPE_IPV6_TCP:
691	fGsoType = SKB_GSO_TCPV6;
692	break;
693	}
694	if (fGsoType)
695	{
696	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
697
698	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
699	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
700	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
701
702	/*
703	* We need to set checksum fields even if the packet goes to the host
704	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
705	*/
706	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrsTotal);
707	pPkt->ip_summed = CHECKSUM_PARTIAL;
708	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
709	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
710	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
711	pPkt->csum_offset = RT_OFFSETOF(RTNETTCP, th_sum);
712	else
713	pPkt->csum_offset = RT_OFFSETOF(RTNETUDP, uh_sum);
714	# else
715	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
716	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
717	pPkt->csum = RT_OFFSETOF(RTNETTCP, th_sum);
718	else
719	pPkt->csum = RT_OFFSETOF(RTNETUDP, uh_sum);
720	# endif
721	if (!fDstWire)
722	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
723	}
724	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
725
726	/*
727	* Finish up the socket buffer.
728	*/
729	pPkt->protocol = eth_type_trans(pPkt, pDev);
730	if (fDstWire)
731	{
732	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
733
734	/* Restore ethernet header back. */
735	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
736	VBOX_SKB_RESET_MAC_HDR(pPkt);
737	}
738	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
739
740	return pPkt;
741	}
742
743
744	/**
745	* Initializes a SG list from an sk_buff.
746	*
747	* @returns Number of segments.
748	* @param pThis The instance.
749	* @param pBuf The sk_buff.
750	* @param pSG The SG.
751	* @param pvFrame The frame pointer, optional.
752	* @param cSegs The number of segments allocated for the SG.
753	* This should match the number in the mbuf exactly!
754	* @param fSrc The source of the frame.
755	* @param pGso Pointer to the GSO context if it's a GSO
756	* internal network frame. NULL if regular frame.
757	*/
758	DECLINLINE(void) vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
759	unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
760	{
761	int i;
762	NOREF(pThis);
763
764	Assert(!skb_shinfo(pBuf)->frag_list);
765
766	if (!pGsoCtx)
767	IntNetSgInitTempSegs(pSG, pBuf->len, cSegs, 0 /cSegsUsed/);
768	else
769	IntNetSgInitTempSegsGso(pSG, pBuf->len, cSegs, 0 /cSegsUsed/, pGsoCtx);
770
771	#ifdef VBOXNETFLT_SG_SUPPORT
772	pSG->aSegs[0].cb = skb_headlen(pBuf);
773	pSG->aSegs[0].pv = pBuf->data;
774	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
775
776	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
777	{
778	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
779	pSG->aSegs[i+1].cb = pFrag->size;
780	pSG->aSegs[i+1].pv = kmap(pFrag->page);
781	printk("%p = kmap()\n", pSG->aSegs[i+1].pv);
782	pSG->aSegs[i+1].Phys = NIL_RTHCPHYS;
783	}
784	++i;
785
786	#else
787	pSG->aSegs[0].cb = pBuf->len;
788	pSG->aSegs[0].pv = pBuf->data;
789	pSG->aSegs[0].Phys = NIL_RTHCPHYS;
790	i = 1;
791	#endif
792
793	pSG->cSegsUsed = i;
794
795	#ifdef PADD_RUNT_FRAMES_FROM_HOST
796	/*
797	* Add a trailer if the frame is too small.
798	*
799	* Since we're getting to the packet before it is framed, it has not
800	* yet been padded. The current solution is to add a segment pointing
801	* to a buffer containing all zeros and pray that works for all frames...
802	*/
803	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
804	{
805	static uint8_t const s_abZero[128] = {0};
806
807	AssertReturnVoid(i < cSegs);
808
809	pSG->aSegs[i].Phys = NIL_RTHCPHYS;
810	pSG->aSegs[i].pv = (void *)&s_abZero[0];
811	pSG->aSegs[i].cb = 60 - pSG->cbTotal;
812	pSG->cbTotal = 60;
813	pSG->cSegsUsed++;
814	Assert(i + 1 <= pSG->cSegsAlloc)
815	}
816	#endif
817
818	Log4(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
819	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
820	for (i = 0; i < pSG->cSegsUsed; i++)
821	Log4(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
822	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
823	}
824
825	/**
826	* Packet handler,
827	*
828	* @returns 0 or EJUSTRETURN.
829	* @param pThis The instance.
830	* @param pMBuf The mbuf.
831	* @param pvFrame The start of the frame, optional.
832	* @param fSrc Where the packet (allegedly) comes from, one INTNETTRUNKDIR_* value.
833	* @param eProtocol The protocol.
834	*/
835	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 14)
836	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
837	struct net_device *pSkbDev,
838	struct packet_type *pPacketType,
839	struct net_device *pOrigDev)
840	#else
841	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
842	struct net_device *pSkbDev,
843	struct packet_type *pPacketType)
844	#endif
845	{
846	PVBOXNETFLTINS pThis;
847	struct net_device *pDev;
848	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
849	pBuf, pSkbDev, pPacketType));
850	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
851	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",
852	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));
853	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
854	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
855	# endif
856	#else
857	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",
858	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));
859	#endif
860	/*
861	* Drop it immediately?
862	*/
863	if (!pBuf)
864	return 0;
865
866	if (pBuf->pkt_type == PACKET_LOOPBACK)
867	{
868	/*
869	* We are not interested in loopbacked packets as they will always have
870	* another copy going to the wire.
871	*/
872	Log2(("vboxNetFltLinuxPacketHandler: dropped loopback packet (cb=%u)\n", pBuf->len));
873	dev_kfree_skb(pBuf); /* We must 'consume' all packets we get (@bugref{6539})! */
874	return 0;
875	}
876
877	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
878	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
879	if (pDev != pSkbDev)
880	{
881	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
882	kfree_skb(pBuf); /* This is a failure, so we use kfree_skb instead of dev_kfree_skb. */
883	return 0;
884	}
885
886	Log4(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
887	if (vboxNetFltLinuxSkBufIsOur(pBuf))
888	{
889	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
890	dev_kfree_skb(pBuf);
891	return 0;
892	}
893
894	#ifndef VBOXNETFLT_SG_SUPPORT
895	{
896	/*
897	* Get rid of fragmented packets, they cause too much trouble.
898	*/
899	unsigned int uMacLen = pBuf->mac_len;
900	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
901	dev_kfree_skb(pBuf);
902	if (!pCopy)
903	{
904	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
905	return 0;
906	}
907	pBuf = pCopy;
908	/* Somehow skb_copy ignores mac_len */
909	pBuf->mac_len = uMacLen;
910	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)
911	/* Restore VLAN tag stripped by host hardware */
912	if (vlan_tx_tag_present(pBuf) && skb_headroom(pBuf) >= VLAN_ETH_HLEN)
913	{
914	uint8_t pMac = (uint8_t)skb_mac_header(pBuf);
915	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )(pMac - VLAN_HLEN);
916	# if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 4, 0)
917	memmove(pVHdr, pMac, ETH_ALEN * 2);
918	# else
919	memmove(pVHdr, pMac, VLAN_ETH_ALEN * 2);
920	# endif
921	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
922	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
923	pBuf->mac_header -= VLAN_HLEN;
924	pBuf->mac_len += VLAN_HLEN;
925	}
926	# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
927
928	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
929	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",
930	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));
931	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
932	Log4(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
933	# endif
934	# else
935	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",
936	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));
937	# endif
938	}
939	#endif
940
941	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
942	/* Forward it to the internal network. */
943	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
944	#else
945	/* Add the packet to transmit queue and schedule the bottom half. */
946	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
947	schedule_work(&pThis->u.s.XmitTask);
948	Log4(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
949	&pThis->u.s.XmitTask, pBuf));
950	#endif
951
952	/* It does not really matter what we return, it is ignored by the kernel. */
953	return 0;
954	}
955
956	/**
957	* Calculate the number of INTNETSEG segments the socket buffer will need.
958	*
959	* @returns Segment count.
960	* @param pBuf The socket buffer.
961	*/
962	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff *pBuf)
963	{
964	#ifdef VBOXNETFLT_SG_SUPPORT
965	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
966	#else
967	unsigned cSegs = 1;
968	#endif
969	#ifdef PADD_RUNT_FRAMES_FROM_HOST
970	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
971	if (pBuf->len < 60)
972	cSegs++;
973	#endif
974	return cSegs;
975	}
976
977	/**
978	* Destroy the intnet scatter / gather buffer created by
979	* vboxNetFltLinuxSkBufToSG.
980	*/
981	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG)
982	{
983	#ifdef VBOXNETFLT_SG_SUPPORT
984	int i;
985
986	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
987	{
988	printk("kunmap(%p)\n", pSG->aSegs[i+1].pv);
989	kunmap(pSG->aSegs[i+1].pv);
990	}
991	#endif
992	NOREF(pSG);
993	}
994
995	#ifdef LOG_ENABLED
996	/**
997	* Logging helper.
998	*/
999	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1000	{
1001	int i, offSeg;
1002	uint8_t pInt, pExt;
1003	static int iPacketNo = 1;
1004	iPacketNo += iIncrement;
1005	if (fEgress)
1006	{
1007	pExt = pSG->aSegs[0].pv;
1008	pInt = pExt + 6;
1009	}
1010	else
1011	{
1012	pInt = pSG->aSegs[0].pv;
1013	pExt = pInt + 6;
1014	}
1015	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1016	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1017	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1018	fEgress ? "-->" : "<--", pszWhere,
1019	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1020	pSG->cbTotal, iPacketNo));
1021	if (pSG->cSegsUsed == 1)
1022	{
1023	Log3(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1024	}
1025	else
1026	{
1027	for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1028	{
1029	Log3(("-- segment %d at 0x%x (%d bytes) --\n%.*Rhxd\n",
1030	i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1031	offSeg += pSG->aSegs[i].cb;
1032	}
1033	}
1034
1035	}
1036	#else
1037	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1038	#endif
1039
1040	#ifdef VBOXNETFLT_WITH_GSO_RECV
1041
1042	/**
1043	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1044	* GSO socket buffer without having to segment it.
1045	*
1046	* @returns true on success, false if needs segmenting.
1047	* @param pThis The net filter instance.
1048	* @param pSkb The GSO socket buffer.
1049	* @param fSrc The source.
1050	* @param pGsoCtx Where to return the GSO context on success.
1051	*/
1052	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1053	PPDMNETWORKGSO pGsoCtx)
1054	{
1055	PDMNETWORKGSOTYPE enmGsoType;
1056	uint16_t uEtherType;
1057	unsigned int cbTransport;
1058	unsigned int offTransport;
1059	unsigned int cbTransportHdr;
1060	unsigned uProtocol;
1061	union
1062	{
1063	RTNETIPV4 IPv4;
1064	RTNETIPV6 IPv6;
1065	RTNETTCP Tcp;
1066	uint8_t ab[40];
1067	uint16_t au16[40/2];
1068	uint32_t au32[40/4];
1069	} Buf;
1070
1071	/*
1072	* Check the GSO properties of the socket buffer and make sure it fits.
1073	*/
1074	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1075	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_UDP \| SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1076	{
1077	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1078	return false;
1079	}
1080	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1081	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1082	{
1083	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1084	return false;
1085	}
1086	/*
1087	* It is possible to receive GSO packets from wire if GRO is enabled.
1088	*/
1089	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1090	{
1091	Log5(("vboxNetFltLinuxCanForwardAsGso: fSrc=wire\n"));
1092	#ifdef VBOXNETFLT_WITH_GRO
1093	/*
1094	* The packet came from the wire and the driver has already consumed
1095	* mac header. We need to restore it back.
1096	*/
1097	pSkb->mac_len = skb_network_header(pSkb) - skb_mac_header(pSkb);
1098	skb_push(pSkb, pSkb->mac_len);
1099	Log5(("vboxNetFltLinuxCanForwardAsGso: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1100	pSkb->mac_len, pSkb->data, skb_mac_header(pSkb), skb_network_header(pSkb)));
1101	#else /* !VBOXNETFLT_WITH_GRO */
1102	/* Older kernels didn't have GRO. */
1103	return false;
1104	#endif /* !VBOXNETFLT_WITH_GRO */
1105	}
1106	else
1107	{
1108	/*
1109	* skb_gso_segment does the following. Do we need to do it as well?
1110	*/
1111	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1112	skb_reset_mac_header(pSkb);
1113	pSkb->mac_len = pSkb->network_header - pSkb->mac_header;
1114	#else
1115	pSkb->mac.raw = pSkb->data;
1116	pSkb->mac_len = pSkb->nh.raw - pSkb->data;
1117	#endif
1118	}
1119
1120	/*
1121	* Switch on the ethertype.
1122	*/
1123	uEtherType = pSkb->protocol;
1124	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1125	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1126	{
1127	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1128	if (puEtherType)
1129	uEtherType = *puEtherType;
1130	}
1131	switch (uEtherType)
1132	{
1133	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1134	{
1135	unsigned int cbHdr;
1136	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1137	if (RT_UNLIKELY(!pIPv4))
1138	{
1139	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1140	return false;
1141	}
1142
1143	cbHdr = pIPv4->ip_hl * 4;
1144	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1145	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1146	\|\| cbHdr > cbTransport ))
1147	{
1148	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1149	return false;
1150	}
1151	cbTransport -= cbHdr;
1152	offTransport = pSkb->mac_len + cbHdr;
1153	uProtocol = pIPv4->ip_p;
1154	if (uProtocol == RTNETIPV4_PROT_TCP)
1155	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1156	else if (uProtocol == RTNETIPV4_PROT_UDP)
1157	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1158	else /** @todo IPv6: 4to6 tunneling */
1159	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1160	break;
1161	}
1162
1163	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1164	{
1165	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1166	if (RT_UNLIKELY(!pIPv6))
1167	{
1168	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1169	return false;
1170	}
1171
1172	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1173	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1174	uProtocol = pIPv6->ip6_nxt;
1175	/** @todo IPv6: Dig our way out of the other headers. */
1176	if (uProtocol == RTNETIPV4_PROT_TCP)
1177	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1178	else if (uProtocol == RTNETIPV4_PROT_UDP)
1179	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1180	else
1181	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1182	break;
1183	}
1184
1185	default:
1186	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1187	return false;
1188	}
1189
1190	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1191	{
1192	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1193	return false;
1194	}
1195
1196	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1197	\|\| offTransport + cbTransport > pSkb->len
1198	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1199	{
1200	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1201	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1202	return false;
1203	}
1204
1205	/*
1206	* Check the TCP/UDP bits.
1207	*/
1208	if (uProtocol == RTNETIPV4_PROT_TCP)
1209	{
1210	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1211	if (RT_UNLIKELY(!pTcp))
1212	{
1213	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1214	return false;
1215	}
1216
1217	cbTransportHdr = pTcp->th_off * 4;
1218	pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1219	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1220	\|\| cbTransportHdr > cbTransport
1221	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1222	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1223	{
1224	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1225	return false;
1226	}
1227
1228	}
1229	else
1230	{
1231	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1232	cbTransportHdr = sizeof(RTNETUDP);
1233	pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1234	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1235	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1236	{
1237	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1238	return false;
1239	}
1240	}
1241
1242	/*
1243	* We're good, init the GSO context.
1244	*/
1245	pGsoCtx->u8Type = enmGsoType;
1246	pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1247	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1248	pGsoCtx->offHdr1 = pSkb->mac_len;
1249	pGsoCtx->offHdr2 = offTransport;
1250	pGsoCtx->u8Unused = 0;
1251
1252	return true;
1253	}
1254
1255	/**
1256	* Forward the socket buffer as a GSO internal network frame.
1257	*
1258	* @returns IPRT status code.
1259	* @param pThis The net filter instance.
1260	* @param pSkb The GSO socket buffer.
1261	* @param fSrc The source.
1262	* @param pGsoCtx Where to return the GSO context on success.
1263	*/
1264	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1265	{
1266	int rc;
1267	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb);
1268	if (RT_LIKELY(cSegs <= MAX_SKB_FRAGS + 1))
1269	{
1270	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1271	if (RT_LIKELY(pSG))
1272	{
1273	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cSegs, fSrc, pGsoCtx);
1274
1275	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1276	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1277
1278	vboxNetFltLinuxDestroySG(pSG);
1279	rc = VINF_SUCCESS;
1280	}
1281	else
1282	{
1283	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1284	rc = VERR_NO_MEMORY;
1285	}
1286	}
1287	else
1288	{
1289	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1290	rc = VERR_INTERNAL_ERROR_3;
1291	}
1292
1293	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1294	dev_kfree_skb(pSkb);
1295	return rc;
1296	}
1297
1298	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1299
1300	/**
1301	* Worker for vboxNetFltLinuxForwardToIntNet.
1302	*
1303	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1304	* @param pThis The net filter instance.
1305	* @param pBuf The socket buffer.
1306	* @param fSrc The source.
1307	*/
1308	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1309	{
1310	int rc;
1311	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf);
1312	if (cSegs <= MAX_SKB_FRAGS + 1)
1313	{
1314	PINTNETSG pSG = (PINTNETSG)alloca(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
1315	if (RT_LIKELY(pSG))
1316	{
1317	if (fSrc & INTNETTRUNKDIR_WIRE)
1318	{
1319	/*
1320	* The packet came from wire, ethernet header was removed by device driver.
1321	* Restore it using mac_len field. This takes into account VLAN headers too.
1322	*/
1323	skb_push(pBuf, pBuf->mac_len);
1324	}
1325
1326	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cSegs, fSrc, NULL /pGsoCtx/);
1327
1328	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1329	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1330
1331	vboxNetFltLinuxDestroySG(pSG);
1332	rc = VINF_SUCCESS;
1333	}
1334	else
1335	{
1336	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1337	rc = VERR_NO_MEMORY;
1338	}
1339	}
1340	else
1341	{
1342	Log(("VBoxNetFlt: Bad sk_buff? cSegs=%#x.\n", cSegs));
1343	rc = VERR_INTERNAL_ERROR_3;
1344	}
1345
1346	Log4(("VBoxNetFlt: Dropping the sk_buff.\n"));
1347	dev_kfree_skb(pBuf);
1348	return rc;
1349	}
1350
1351	/**
1352	*
1353	* @param pBuf The socket buffer. This is consumed by this function.
1354	*/
1355	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1356	{
1357	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1358
1359	#ifdef VBOXNETFLT_WITH_GSO
1360	if (skb_is_gso(pBuf))
1361	{
1362	PDMNETWORKGSO GsoCtx;
1363	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",
1364	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));
1365	# ifdef VBOXNETFLT_WITH_GSO_RECV
1366	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_UDP \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1367	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1368	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1369	else
1370	# endif
1371	{
1372	/* Need to segment the packet */
1373	struct sk_buff *pNext;
1374	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1375	if (IS_ERR(pSegment))
1376	{
1377	dev_kfree_skb(pBuf);
1378	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1379	return;
1380	}
1381
1382	for (; pSegment; pSegment = pNext)
1383	{
1384	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",
1385	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));
1386	pNext = pSegment->next;
1387	pSegment->next = 0;
1388	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1389	}
1390	dev_kfree_skb(pBuf);
1391	}
1392	}
1393	else
1394	#endif /* VBOXNETFLT_WITH_GSO */
1395	{
1396	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1397	{
1398	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1399	/*
1400	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1401	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1402	* header length from the header itself and reconstruct 'h' pointer
1403	* to TCP (or whatever) header.
1404	*/
1405	unsigned char *tmp = pBuf->h.raw;
1406	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1407	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1408	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1409	if (VBOX_SKB_CHECKSUM_HELP(pBuf))
1410	{
1411	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1412	dev_kfree_skb(pBuf);
1413	return;
1414	}
1415	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1416	/* Restore the original (wrong) pointer. */
1417	pBuf->h.raw = tmp;
1418	#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1419	}
1420	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1421	}
1422	}
1423
1424	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1425	/**
1426	* Work queue handler that forwards the socket buffers queued by
1427	* vboxNetFltLinuxPacketHandler to the internal network.
1428	*
1429	* @param pWork The work queue.
1430	*/
1431	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
1432	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1433	# else
1434	static void vboxNetFltLinuxXmitTask(void *pWork)
1435	# endif
1436	{
1437	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1438	struct sk_buff *pBuf;
1439
1440	Log4(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1441
1442	/*
1443	* Active? Retain the instance and increment the busy counter.
1444	*/
1445	if (vboxNetFltTryRetainBusyActive(pThis))
1446	{
1447	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1448	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1449
1450	vboxNetFltRelease(pThis, true /* fBusy */);
1451	}
1452	else
1453	{
1454	/** @todo Shouldn't we just drop the packets here? There is little point in
1455	* making them accumulate when the VM is paused and it'll only waste
1456	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1457	* before start draining the packets (goes for the intnet ring buf
1458	* too)? */
1459	}
1460	}
1461	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1462
1463	/**
1464	* Reports the GSO capabilities of the hardware NIC.
1465	*
1466	* @param pThis The net filter instance. The caller hold a
1467	* reference to this.
1468	*/
1469	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1470	{
1471	#ifdef VBOXNETFLT_WITH_GSO_XMIT_WIRE
1472	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1473	{
1474	struct net_device *pDev;
1475	PINTNETTRUNKSWPORT pSwitchPort;
1476	unsigned int fFeatures;
1477
1478	RTSpinlockAcquire(pThis->hSpinlock);
1479
1480	pSwitchPort = pThis->pSwitchPort; /* this doesn't need to be here, but it doesn't harm. */
1481	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1482	if (pDev)
1483	fFeatures = pDev->features;
1484	else
1485	fFeatures = 0;
1486
1487	RTSpinlockRelease(pThis->hSpinlock);
1488
1489	if (pThis->pSwitchPort)
1490	{
1491	/* Set/update the GSO capabilities of the NIC. */
1492	uint32_t fGsoCapabilites = 0;
1493	if (fFeatures & NETIF_F_TSO)
1494	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1495	if (fFeatures & NETIF_F_TSO6)
1496	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1497	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
1498	if (fFeatures & NETIF_F_UFO)
1499	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP);
1500	if (fFeatures & NETIF_F_UFO)
1501	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP);
1502	# endif
1503	Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s%s%s\n",
1504	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1505	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : "",
1506	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP)) ? "ufo " : "",
1507	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP)) ? "ufo6 " : ""));
1508	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1509	}
1510
1511	vboxNetFltRelease(pThis, true /fBusy/);
1512	}
1513	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE */
1514	}
1515
1516	/**
1517	* Helper that determines whether the host (ignoreing us) is operating the
1518	* interface in promiscuous mode or not.
1519	*/
1520	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1521	{
1522	bool fRc = false;
1523	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1524	if (pDev)
1525	{
1526	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1527	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1528	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1529	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1530	}
1531	return fRc;
1532	}
1533
1534	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 36)
1535	/**
1536	* Helper for detecting TAP devices.
1537	*/
1538	static bool vboxNetFltIsTapDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1539	{
1540	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1541	{
1542	struct ethtool_drvinfo Info;
1543
1544	memset(&Info, 0, sizeof(Info));
1545	Info.cmd = ETHTOOL_GDRVINFO;
1546	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1547	Log3(("vboxNetFltIsTapDevice: driver=%s version=%s bus_info=%s\n",
1548	Info.driver, Info.version, Info.bus_info));
1549
1550	return !strncmp(Info.driver, "tun", 4)
1551	&& !strncmp(Info.bus_info, "tap", 4);
1552	}
1553
1554	return false;
1555	}
1556
1557	/**
1558	* Helper for updating the link state of TAP devices.
1559	* Only TAP devices are affected.
1560	*/
1561	static void vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1562	{
1563	if (vboxNetFltIsTapDevice(pThis, pDev))
1564	{
1565	Log3(("vboxNetFltSetTapLinkState: bringing %s tap device link state\n",
1566	fLinkUp ? "up" : "down"));
1567	netif_tx_lock_bh(pDev);
1568	if (fLinkUp)
1569	netif_carrier_on(pDev);
1570	else
1571	netif_carrier_off(pDev);
1572	netif_tx_unlock_bh(pDev);
1573	}
1574	}
1575	#else /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
1576	DECLINLINE(void) vboxNetFltSetTapLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1577	{
1578	/* Nothing to do for pre-2.6.36 kernels. */
1579	}
1580	#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36) */
1581
1582	/**
1583	* Internal worker for vboxNetFltLinuxNotifierCallback.
1584	*
1585	* @returns VBox status code.
1586	* @param pThis The instance.
1587	* @param fRediscovery If set we're doing a rediscovery attempt, so, don't
1588	* flood the release log.
1589	*/
1590	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1591	{
1592	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1593
1594	/*
1595	* Retain and store the device.
1596	*/
1597	dev_hold(pDev);
1598
1599	RTSpinlockAcquire(pThis->hSpinlock);
1600	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
1601	RTSpinlockRelease(pThis->hSpinlock);
1602
1603	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
1604	pDev, pDev->name,
1605	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1606	netdev_refcnt_read(pDev)
1607	#else
1608	atomic_read(&pDev->refcnt)
1609	#endif
1610	));
1611	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1612	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1613
1614	/* Get the mac address while we still have a valid net_device reference. */
1615	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
1616	/* Initialize MTU */
1617	pThis->u.s.cbMtu = pDev->mtu;
1618
1619	/*
1620	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1621	*/
1622	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1623	pThis->u.s.PacketType.dev = pDev;
1624	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1625	dev_add_pack(&pThis->u.s.PacketType);
1626	ASMAtomicUoWriteBool(&pThis->u.s.fPacketHandler, true);
1627	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1628
1629	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1630	vboxNetFltLinuxHookDev(pThis, pDev);
1631	#endif
1632
1633	/*
1634	* If attaching to TAP interface we need to bring the link state up
1635	* starting from 2.6.36 kernel.
1636	*/
1637	vboxNetFltSetTapLinkState(pThis, pDev, true);
1638
1639	/*
1640	* Set indicators that require the spinlock. Be abit paranoid about racing
1641	* the device notification handle.
1642	*/
1643	RTSpinlockAcquire(pThis->hSpinlock);
1644	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1645	if (pDev)
1646	{
1647	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1648	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1649	pDev = NULL; /* don't dereference it */
1650	}
1651	RTSpinlockRelease(pThis->hSpinlock);
1652
1653	/*
1654	* If the above succeeded report GSO capabilities, if not undo and
1655	* release the device.
1656	*/
1657	if (!pDev)
1658	{
1659	Assert(pThis->pSwitchPort);
1660	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1661	{
1662	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1663	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
1664	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
1665	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
1666	vboxNetFltRelease(pThis, true /fBusy/);
1667	}
1668	}
1669	else
1670	{
1671	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1672	vboxNetFltLinuxUnhookDev(pThis, pDev);
1673	#endif
1674	RTSpinlockAcquire(pThis->hSpinlock);
1675	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1676	RTSpinlockRelease(pThis->hSpinlock);
1677	dev_put(pDev);
1678	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) released. ref=%d\n",
1679	pDev, pDev->name,
1680	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1681	netdev_refcnt_read(pDev)
1682	#else
1683	atomic_read(&pDev->refcnt)
1684	#endif
1685	));
1686	}
1687
1688	LogRel(("VBoxNetFlt: attached to '%s' / %.*Rhxs\n", pThis->szName, sizeof(pThis->u.s.MacAddr), &pThis->u.s.MacAddr));
1689	return VINF_SUCCESS;
1690	}
1691
1692
1693	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1694	{
1695	bool fRegistered;
1696	Assert(!pThis->fDisconnectedFromHost);
1697
1698	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1699	vboxNetFltLinuxUnhookDev(pThis, pDev);
1700	#endif
1701
1702	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
1703	{
1704	dev_remove_pack(&pThis->u.s.PacketType);
1705	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: packet handler removed.\n", pThis));
1706	}
1707
1708	RTSpinlockAcquire(pThis->hSpinlock);
1709	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
1710	if (fRegistered)
1711	{
1712	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1713	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1714	}
1715	RTSpinlockRelease(pThis->hSpinlock);
1716
1717	if (fRegistered)
1718	{
1719	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1720	skb_queue_purge(&pThis->u.s.XmitQueue);
1721	#endif
1722	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: xmit queue purged.\n", pThis));
1723	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
1724	pDev, pDev->name,
1725	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1726	netdev_refcnt_read(pDev)
1727	#else
1728	atomic_read(&pDev->refcnt)
1729	#endif
1730	));
1731	dev_put(pDev);
1732	}
1733
1734	return NOTIFY_OK;
1735	}
1736
1737	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1738	{
1739	/* Check if we are not suspended and promiscuous mode has not been set. */
1740	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
1741	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1742	{
1743	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1744	dev_set_promiscuity(pDev, 1);
1745	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1746	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1747	}
1748	else
1749	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1750	return NOTIFY_OK;
1751	}
1752
1753	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1754	{
1755	/* Undo promiscuous mode if we has set it. */
1756	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1757	{
1758	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1759	dev_set_promiscuity(pDev, -1);
1760	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1761	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1762	}
1763	else
1764	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1765	return NOTIFY_OK;
1766	}
1767
1768	/**
1769	* Callback for listening to MTU change event.
1770	*
1771	* We need to track changes of host's inteface MTU to discard over-sized frames
1772	* coming from the internal network as they may hang the TX queue of host's
1773	* adapter.
1774	*
1775	* @returns NOTIFY_OK
1776	* @param pThis The netfilter instance.
1777	* @param pDev Pointer to device structure of host's interface.
1778	*/
1779	static int vboxNetFltLinuxDeviceMtuChange(PVBOXNETFLTINS pThis, struct net_device *pDev)
1780	{
1781	ASMAtomicWriteU32(&pThis->u.s.cbMtu, pDev->mtu);
1782	Log(("vboxNetFltLinuxDeviceMtuChange: set MTU for %s to %d\n", pThis->szName, pDev->mtu));
1783	return NOTIFY_OK;
1784	}
1785
1786	#ifdef LOG_ENABLED
1787	/** Stringify the NETDEV_XXX constants. */
1788	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
1789	{
1790	const char *pszEvent = "NETDRV_<unknown>";
1791	switch (ulEventType)
1792	{
1793	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
1794	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
1795	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
1796	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
1797	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
1798	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
1799	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
1800	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
1801	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
1802	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
1803	# ifdef NETDEV_FEAT_CHANGE
1804	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
1805	# endif
1806	}
1807	return pszEvent;
1808	}
1809	#endif /* LOG_ENABLED */
1810
1811	/**
1812	* Callback for listening to netdevice events.
1813	*
1814	* This works the rediscovery, clean up on unregistration, promiscuity on
1815	* up/down, and GSO feature changes from ethtool.
1816	*
1817	* @returns NOTIFY_OK
1818	* @param self Pointer to our notifier registration block.
1819	* @param ulEventType The event.
1820	* @param ptr Event specific, but it is usually the device it
1821	* relates to.
1822	*/
1823	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
1824
1825	{
1826	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
1827	#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0)
1828	struct net_device *pDev = netdev_notifier_info_to_dev(ptr);
1829	#else
1830	struct net_device pDev = (struct net_device )ptr;
1831	#endif
1832	int rc = NOTIFY_OK;
1833
1834	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1835	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1836	if ( ulEventType == NETDEV_REGISTER
1837	&& !strcmp(pDev->name, pThis->szName))
1838	{
1839	vboxNetFltLinuxAttachToInterface(pThis, pDev);
1840	}
1841	else
1842	{
1843	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1844	if (pDev == ptr)
1845	{
1846	switch (ulEventType)
1847	{
1848	case NETDEV_UNREGISTER:
1849	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
1850	break;
1851	case NETDEV_UP:
1852	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
1853	break;
1854	case NETDEV_GOING_DOWN:
1855	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
1856	break;
1857	case NETDEV_CHANGEMTU:
1858	rc = vboxNetFltLinuxDeviceMtuChange(pThis, pDev);
1859	break;
1860	case NETDEV_CHANGENAME:
1861	break;
1862	#ifdef NETDEV_FEAT_CHANGE
1863	case NETDEV_FEAT_CHANGE:
1864	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1865	break;
1866	#endif
1867	}
1868	}
1869	}
1870
1871	return rc;
1872	}
1873
1874	#if 0 /* XXX: temporarily disable */
1875	static int vboxNetFltLinuxNotifierIPv4Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
1876	{
1877	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv4);
1878	struct net_device *pDev;
1879	struct in_ifaddr ifa = (struct in_ifaddr )ptr;
1880	int rc = NOTIFY_OK;
1881
1882	pDev = vboxNetFltLinuxRetainNetDev(pThis);
1883	Log(("VBoxNetFlt: %s: IPv4 event %s(0x%lx): addr %RTnaipv4 mask %RTnaipv4\n",
1884	pDev ? VBOX_NETDEV_NAME(pDev) : "<???>",
1885	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
1886	ifa->ifa_address, ifa->ifa_mask));
1887
1888	if (pDev != NULL)
1889	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1890
1891	if (pThis->pSwitchPort->pfnNotifyHostAddress)
1892	{
1893	bool fAdded;
1894	if (ulEventType == NETDEV_UP)
1895	fAdded = true;
1896	else if (ulEventType == NETDEV_DOWN)
1897	fAdded = false;
1898	else
1899	return NOTIFY_OK;
1900
1901	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
1902	kIntNetAddrType_IPv4, &ifa->ifa_local);
1903	}
1904
1905	return rc;
1906	}
1907
1908
1909	static int vboxNetFltLinuxNotifierIPv6Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
1910	{
1911	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv6);
1912	struct net_device *pDev;
1913	struct inet6_ifaddr ifa = (struct inet6_ifaddr )ptr;
1914	int rc = NOTIFY_OK;
1915
1916	pDev = vboxNetFltLinuxRetainNetDev(pThis);
1917	Log(("VBoxNetFlt: %s: IPv6 event %s(0x%lx): %RTnaipv6\n",
1918	pDev ? VBOX_NETDEV_NAME(pDev) : "<???>",
1919	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
1920	&ifa->addr));
1921
1922	if (pDev != NULL)
1923	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1924
1925	if (pThis->pSwitchPort->pfnNotifyHostAddress)
1926	{
1927	bool fAdded;
1928	if (ulEventType == NETDEV_UP)
1929	fAdded = true;
1930	else if (ulEventType == NETDEV_DOWN)
1931	fAdded = false;
1932	else
1933	return NOTIFY_OK;
1934
1935	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
1936	kIntNetAddrType_IPv6, &ifa->addr);
1937	}
1938
1939	return rc;
1940	}
1941	#endif /* 0 */
1942
1943
1944	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
1945	{
1946	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
1947	}
1948
1949	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
1950	{
1951	struct net_device * pDev;
1952	int err;
1953	int rc = VINF_SUCCESS;
1954	NOREF(pvIfData);
1955
1956	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
1957
1958	pDev = vboxNetFltLinuxRetainNetDev(pThis);
1959	if (pDev)
1960	{
1961	/*
1962	* Create a sk_buff for the gather list and push it onto the wire.
1963	*/
1964	if (fDst & INTNETTRUNKDIR_WIRE)
1965	{
1966	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
1967	if (pBuf)
1968	{
1969	vboxNetFltDumpPacket(pSG, true, "wire", 1);
1970	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1971	Log4(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
1972	err = dev_queue_xmit(pBuf);
1973	if (err)
1974	rc = RTErrConvertFromErrno(err);
1975	}
1976	else
1977	rc = VERR_NO_MEMORY;
1978	}
1979
1980	/*
1981	* Create a sk_buff for the gather list and push it onto the host stack.
1982	*/
1983	if (fDst & INTNETTRUNKDIR_HOST)
1984	{
1985	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
1986	if (pBuf)
1987	{
1988	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
1989	Log4(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1990	Log4(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
1991	err = netif_rx_ni(pBuf);
1992	if (err)
1993	rc = RTErrConvertFromErrno(err);
1994	}
1995	else
1996	rc = VERR_NO_MEMORY;
1997	}
1998
1999	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2000	}
2001
2002	return rc;
2003	}
2004
2005
2006	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2007	{
2008	struct net_device * pDev;
2009
2010	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%s, fDisablePromiscuous=%s\n",
2011	pThis, pThis->szName, fActive?"true":"false",
2012	pThis->fDisablePromiscuous?"true":"false"));
2013
2014	if (pThis->fDisablePromiscuous)
2015	return;
2016
2017	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2018	if (pDev)
2019	{
2020	/*
2021	* This api is a bit weird, the best reference is the code.
2022	*
2023	* Also, we have a bit or race conditions wrt the maintenance of
2024	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2025	*/
2026	#ifdef LOG_ENABLED
2027	u_int16_t fIf;
2028	unsigned const cPromiscBefore = pDev->promiscuity;
2029	#endif
2030	if (fActive)
2031	{
2032	Assert(!pThis->u.s.fPromiscuousSet);
2033
2034	rtnl_lock();
2035	dev_set_promiscuity(pDev, 1);
2036	rtnl_unlock();
2037	pThis->u.s.fPromiscuousSet = true;
2038	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2039	}
2040	else
2041	{
2042	if (pThis->u.s.fPromiscuousSet)
2043	{
2044	rtnl_lock();
2045	dev_set_promiscuity(pDev, -1);
2046	rtnl_unlock();
2047	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2048	}
2049	pThis->u.s.fPromiscuousSet = false;
2050
2051	#ifdef LOG_ENABLED
2052	fIf = dev_get_flags(pDev);
2053	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2054	#endif
2055	}
2056
2057	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2058	}
2059	}
2060
2061
2062	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2063	{
2064	/*
2065	* Remove packet handler when we get disconnected from internal switch as
2066	* we don't want the handler to forward packets to disconnected switch.
2067	*/
2068	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
2069	{
2070	dev_remove_pack(&pThis->u.s.PacketType);
2071	Log(("vboxNetFltOsDisconnectIt: this=%p: Packet handler removed.\n", pThis));
2072	}
2073	return VINF_SUCCESS;
2074	}
2075
2076
2077	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2078	{
2079	/*
2080	* Report the GSO capabilities of the host and device (if connected).
2081	* Note! No need to mark ourselves busy here.
2082	*/
2083	/** @todo duplicate work here now? Attach */
2084	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2085	Log3(("vboxNetFltOsConnectIt: reporting host tso tso6 ufo\n"));
2086	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2087	0
2088	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2089	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2090	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_UDP)
2091	# if 0 /** @todo GSO: Test UDP offloading (UFO) on linux. */
2092	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_UDP)
2093	# endif
2094	, INTNETTRUNKDIR_HOST);
2095
2096	#endif
2097	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2098
2099	return VINF_SUCCESS;
2100	}
2101
2102
2103	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2104	{
2105	struct net_device *pDev;
2106	bool fRegistered;
2107
2108	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
2109	vboxNetFltLinuxUnhookDev(pThis, NULL);
2110	#endif
2111
2112	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2113	* unlikely, but none the less). Since it doesn't actually update the
2114	* state (just reads it), it is likely to panic in some interesting
2115	* ways. */
2116
2117	RTSpinlockAcquire(pThis->hSpinlock);
2118	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2119	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
2120	RTSpinlockRelease(pThis->hSpinlock);
2121
2122	if (fRegistered)
2123	{
2124	vboxNetFltSetTapLinkState(pThis, pDev, false);
2125
2126	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2127	skb_queue_purge(&pThis->u.s.XmitQueue);
2128	#endif
2129	Log(("vboxNetFltOsDeleteInstance: this=%p: xmit queue purged.\n", pThis));
2130	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2131	pDev, pDev->name,
2132	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2133	netdev_refcnt_read(pDev)
2134	#else
2135	atomic_read(&pDev->refcnt)
2136	#endif
2137	));
2138	dev_put(pDev);
2139	}
2140
2141	unregister_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2142	unregister_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2143
2144	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2145	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2146	module_put(THIS_MODULE);
2147	}
2148
2149
2150	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2151	{
2152	int err;
2153	NOREF(pvContext);
2154
2155	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2156	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2157	if (err)
2158	return VERR_INTNET_FLT_IF_FAILED;
2159	if (!pThis->u.s.fRegistered)
2160	{
2161	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2162	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2163	return VERR_INTNET_FLT_IF_NOT_FOUND;
2164	}
2165
2166	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2167	if ( pThis->fDisconnectedFromHost
2168	\|\| !try_module_get(THIS_MODULE))
2169	return VERR_INTNET_FLT_IF_FAILED;
2170
2171	#if 0 /* XXX: temporarily disable */
2172	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2173	{
2174	struct net *net = VBOX_DEV_NET(pThis->u.s.pDev);
2175	struct net_device *dev;
2176
2177	#if !defined(for_each_netdev_rcu) /* introduced in 2.6.33 */
2178	read_lock(&dev_base_lock);
2179	#endif
2180	rcu_read_lock();
2181
2182	#if !defined(for_each_netdev_rcu)
2183	for_each_netdev(net, dev)
2184	#else
2185	for_each_netdev_rcu(net, dev)
2186	#endif
2187	{
2188	struct in_device *in_dev;
2189	struct inet6_dev *in6_dev;
2190
2191	/*
2192	* IPv4
2193	*/
2194	in_dev = __in_dev_get_rcu(dev);
2195	if (in_dev != NULL)
2196	{
2197	for_ifa(in_dev) {
2198	if (ifa->ifa_address == htonl(INADDR_LOOPBACK))
2199	goto continue_netdev;
2200
2201	Log(("%s: %s: IPv4: addr %RTnaipv4 mask %RTnaipv4\n",
2202	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2203	ifa->ifa_address, ifa->ifa_mask));
2204
2205	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2206	/* :fAdded */ true, kIntNetAddrType_IPv4, &ifa->ifa_address);
2207	} endfor_ifa(in_dev);
2208	}
2209
2210	/*
2211	* IPv6
2212	*/
2213	in6_dev = __in6_dev_get(dev);
2214	if (in6_dev != NULL)
2215	{
2216	struct inet6_ifaddr *ifa;
2217
2218	read_lock_bh(&in6_dev->lock);
2219	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 35)
2220	list_for_each_entry(ifa, &in6_dev->addr_list, if_list)
2221	#else
2222	for (ifa = in6_dev->addr_list; ifa != NULL; ifa = ifa->if_next)
2223	#endif
2224	{
2225	Log(("%s: %s: IPv6: addr %RTnaipv6/%u\n",
2226	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2227	&ifa->addr, (unsigned)ifa->prefix_len));
2228
2229	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2230	/* :fAdded */ true, kIntNetAddrType_IPv6, &ifa->addr);
2231	}
2232	read_unlock_bh(&in6_dev->lock);
2233	}
2234
2235	continue_netdev:
2236	/* continue */;
2237	}
2238	rcu_read_unlock();
2239	#if !defined(for_each_netdev_rcu)
2240	read_unlock(&dev_base_lock);
2241	#endif
2242
2243	Log(("%s: pfnNotifyHostAddress is set, register notifiers\n", __FUNCTION__));
2244
2245	pThis->u.s.NotifierIPv4.notifier_call = vboxNetFltLinuxNotifierIPv4Callback;
2246	err = register_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2247	if (err)
2248	LogRel(("%s: failed to register IPv4 notifier: error %d\n",
2249	__FUNCTION__, err));
2250
2251	pThis->u.s.NotifierIPv6.notifier_call = vboxNetFltLinuxNotifierIPv6Callback;
2252	err = register_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2253	if (err)
2254	LogRel(("%s: failed to register IPv6 notifier: error %d\n",
2255	__FUNCTION__, err));
2256	}
2257	else
2258	Log(("%s: uwe: pfnNotifyHostAddress is NULL\n", __FUNCTION__));
2259	#endif
2260
2261	return VINF_SUCCESS;
2262	}
2263
2264	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2265	{
2266	/*
2267	* Init the linux specific members.
2268	*/
2269	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2270	pThis->u.s.fRegistered = false;
2271	pThis->u.s.fPromiscuousSet = false;
2272	pThis->u.s.fPacketHandler = false;
2273	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2274	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2275	skb_queue_head_init(&pThis->u.s.XmitQueue);
2276	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
2277	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2278	# else
2279	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2280	# endif
2281	#endif
2282
2283	return VINF_SUCCESS;
2284	}
2285
2286
2287	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2288	{
2289	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2290	}
2291
2292
2293	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2294	{
2295	/* Nothing to do */
2296	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2297	return VINF_SUCCESS;
2298	}
2299
2300
2301	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2302	{
2303	/* Nothing to do */
2304	NOREF(pThis); NOREF(pvIfData);
2305	return VINF_SUCCESS;
2306	}
2307

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

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 55998

Download in other formats: