VBoxNetFlt-linux.c@ 85701

Last change on this file since 85701 was 85701, checked in by vboxsync, 4 years ago
IPRT,lnx-kmods: Added VBOX_RHEL_MAJ_PREREQ and RTLNX_SUSE_MAJ_PREREQ to version.h. Started using them.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 90.1 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 85701 2020-08-11 17:46:17Z vboxsync $ */
2	/** @file
3	* VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4	*/
5
6	/*
7	* Copyright (C) 2006-2020 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	* The contents of this file may alternatively be used under the terms
18	* of the Common Development and Distribution License Version 1.0
19	* (CDDL) only, as it comes in the "COPYING.CDDL" file of the
20	* VirtualBox OSE distribution, in which case the provisions of the
21	* CDDL are applicable instead of those of the GPL.
22	*
23	* You may elect to license modified versions of this file under the
24	* terms and conditions of either the GPL or the CDDL or both.
25	*/
26
27
28	/*********************************************************************************************************************************
29	* Header Files *
30	*********************************************************************************************************************************/
31	#define LOG_GROUP LOG_GROUP_NET_FLT_DRV
32	#define VBOXNETFLT_LINUX_NO_XMIT_QUEUE
33	#include "the-linux-kernel.h"
34	#include "version-generated.h"
35	#include "revision-generated.h"
36	#include "product-generated.h"
37	#if RTLNX_VER_MIN(2,6,24)
38	#include <linux/nsproxy.h>
39	#endif
40	#include <linux/netdevice.h>
41	#include <linux/etherdevice.h>
42	#include <linux/rtnetlink.h>
43	#include <linux/miscdevice.h>
44	#include <linux/inetdevice.h>
45	#include <linux/in.h>
46	#include <linux/ip.h>
47	#include <linux/if_vlan.h>
48	#if RTLNX_VER_MIN(4,5,0)
49	#include <uapi/linux/pkt_cls.h>
50	#endif
51	#include <net/ipv6.h>
52	#include <net/if_inet6.h>
53	#include <net/addrconf.h>
54
55	#include <VBox/log.h>
56	#include <VBox/err.h>
57	#include <VBox/intnetinline.h>
58	#include <VBox/vmm/pdmnetinline.h>
59	#include <VBox/param.h>
60	#include <iprt/alloca.h>
61	#include <iprt/assert.h>
62	#include <iprt/spinlock.h>
63	#include <iprt/semaphore.h>
64	#include <iprt/initterm.h>
65	#include <iprt/process.h>
66	#include <iprt/mem.h>
67	#include <iprt/net.h>
68	#include <iprt/log.h>
69	#include <iprt/mp.h>
70	#include <iprt/mem.h>
71	#include <iprt/time.h>
72
73	#define VBOXNETFLT_OS_SPECFIC 1
74	#include "../VBoxNetFltInternal.h"
75
76	typedef struct VBOXNETFLTNOTIFIER {
77	struct notifier_block Notifier;
78	PVBOXNETFLTINS pThis;
79	} VBOXNETFLTNOTIFIER;
80	typedef struct VBOXNETFLTNOTIFIER *PVBOXNETFLTNOTIFIER;
81
82
83	/*********************************************************************************************************************************
84	* Defined Constants And Macros *
85	*********************************************************************************************************************************/
86	#define VBOX_FLT_NB_TO_INST(pNB) RT_FROM_MEMBER(pNB, VBOXNETFLTINS, u.s.Notifier)
87	#define VBOX_FLT_PT_TO_INST(pPT) RT_FROM_MEMBER(pPT, VBOXNETFLTINS, u.s.PacketType)
88	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
89	# define VBOX_FLT_XT_TO_INST(pXT) RT_FROM_MEMBER(pXT, VBOXNETFLTINS, u.s.XmitTask)
90	#endif
91
92	#if RTLNX_VER_MIN(3,11,0)
93	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) netdev_notifier_info_to_dev(ptr)
94	#else
95	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) ((struct net_device *)ptr)
96	#endif
97
98	#if RTLNX_VER_MIN(3,5,0)
99	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag))
100	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr)
101	#else
102	# if RTLNX_VER_MIN(3,2,0)
103	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag), KM_SKB_DATA_SOFTIRQ)
104	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
105	# else
106	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ)
107	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
108	# endif
109	#endif
110
111	#if RTLNX_VER_MIN(2,6,34)
112	# define VBOX_NETDEV_NAME(dev) netdev_name(dev)
113	#else
114	# define VBOX_NETDEV_NAME(dev) ((dev)->reg_state != NETREG_REGISTERED ? "(unregistered net_device)" : (dev)->name)
115	#endif
116
117	#if RTLNX_VER_MIN(2,6,25)
118	# define VBOX_IPV4_IS_LOOPBACK(addr) ipv4_is_loopback(addr)
119	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ipv4_is_linklocal_169(addr)
120	#else
121	# define VBOX_IPV4_IS_LOOPBACK(addr) ((addr & htonl(IN_CLASSA_NET)) == htonl(0x7f000000))
122	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ((addr & htonl(IN_CLASSB_NET)) == htonl(0xa9fe0000))
123	#endif
124
125	#if RTLNX_VER_MIN(2,6,22)
126	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
127	# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
128	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum_offset
129	#else
130	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
131	# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
132	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum
133	#endif
134
135	#if RTLNX_VER_MIN(2,6,19)
136	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
137	#else
138	# define CHECKSUM_PARTIAL CHECKSUM_HW
139	# if RTLNX_VER_MIN(2,6,10)
140	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
141	# else
142	# if RTLNX_VER_MIN(2,6,7)
143	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
144	# else
145	# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
146	# endif
147	/* Versions prior 2.6.10 use stats for both bstats and qstats */
148	# define bstats stats
149	# define qstats stats
150	# endif
151	#endif
152
153	#if RTLNX_VER_MIN(3,20,0)
154	# define VBOX_HAVE_SKB_VLAN
155	#else
156	# ifdef RHEL_RELEASE_CODE
157	# if (RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(7, 2) && RHEL_RELEASE_CODE < RHEL_RELEASE_VERSION(8, 0)) \|\| \
158	(RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(6, 8) && RHEL_RELEASE_CODE < RHEL_RELEASE_VERSION(7, 0))
159	# define VBOX_HAVE_SKB_VLAN
160	# endif
161	# endif
162	#endif
163
164	#ifdef VBOX_HAVE_SKB_VLAN
165	# define vlan_tx_tag_get(skb) skb_vlan_tag_get(skb)
166	# define vlan_tx_tag_present(skb) skb_vlan_tag_present(skb)
167	#endif
168
169	#ifndef NET_IP_ALIGN
170	# define NET_IP_ALIGN 2
171	#endif
172
173	#if 1
174	/** Create scatter / gather segments for fragments. When not used, we will
175	* linearize the socket buffer before creating the internal networking SG. */
176	# define VBOXNETFLT_SG_SUPPORT 1
177	#endif
178
179	#if RTLNX_VER_MIN(2,6,18)
180
181	/** Indicates that the linux kernel may send us GSO frames. */
182	# define VBOXNETFLT_WITH_GSO 1
183
184	/** This enables or disables the transmitting of GSO frame from the internal
185	* network and to the host. */
186	# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
187
188	# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
189	/** This enables or disables the transmitting of GSO frame from the internal
190	* network and to the wire. */
191	# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
192	# endif
193
194	/** This enables or disables the forwarding/flooding of GSO frame from the host
195	* to the internal network. */
196	# define VBOXNETFLT_WITH_GSO_RECV 1
197
198	#endif /* RTLNX_VER_MIN(2,6,18) */
199
200	#if RTLNX_VER_MIN(2,6,29)
201	/** This enables or disables handling of GSO frames coming from the wire (GRO). */
202	# define VBOXNETFLT_WITH_GRO 1
203	#endif
204
205	/*
206	* GRO support was backported to RHEL 5.4
207	*/
208	#if RTLNX_RHEL_MAJ_PREREQ(5, 4)
209	# define VBOXNETFLT_WITH_GRO 1
210	#endif
211
212
213	/*********************************************************************************************************************************
214	* Internal Functions *
215	*********************************************************************************************************************************/
216	static int __init VBoxNetFltLinuxInit(void);
217	static void __exit VBoxNetFltLinuxUnload(void);
218	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf);
219
220
221	/*********************************************************************************************************************************
222	* Global Variables *
223	*********************************************************************************************************************************/
224	/**
225	* The (common) global data.
226	*/
227	static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
228
229	module_init(VBoxNetFltLinuxInit);
230	module_exit(VBoxNetFltLinuxUnload);
231
232	MODULE_AUTHOR(VBOX_VENDOR);
233	MODULE_DESCRIPTION(VBOX_PRODUCT " Network Filter Driver");
234	MODULE_LICENSE("GPL");
235	#ifdef MODULE_VERSION
236	MODULE_VERSION(VBOX_VERSION_STRING " r" RT_XSTR(VBOX_SVN_REV) " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
237	#endif
238
239
240	#if RTLNX_VER_MAX(2,6,12) && defined(LOG_ENABLED)
241	unsigned dev_get_flags(const struct net_device *dev)
242	{
243	unsigned flags;
244
245	flags = (dev->flags & ~(IFF_PROMISC \|
246	IFF_ALLMULTI \|
247	IFF_RUNNING)) \|
248	(dev->gflags & (IFF_PROMISC \|
249	IFF_ALLMULTI));
250
251	if (netif_running(dev) && netif_carrier_ok(dev))
252	flags \|= IFF_RUNNING;
253
254	return flags;
255	}
256	#endif /* RTLNX_VER_MAX(2,6,12) */
257
258
259	/**
260	* Initialize module.
261	*
262	* @returns appropriate status code.
263	*/
264	static int __init VBoxNetFltLinuxInit(void)
265	{
266	int rc;
267	/*
268	* Initialize IPRT.
269	*/
270	rc = RTR0Init(0);
271	if (RT_SUCCESS(rc))
272	{
273	Log(("VBoxNetFltLinuxInit\n"));
274
275	/*
276	* Initialize the globals and connect to the support driver.
277	*
278	* This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
279	* for establishing the connect to the support driver.
280	*/
281	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
282	rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
283	if (RT_SUCCESS(rc))
284	{
285	LogRel(("VBoxNetFlt: Successfully started.\n"));
286	return 0;
287	}
288
289	LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
290	RTR0Term();
291	}
292	else
293	LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
294
295	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
296	return -RTErrConvertToErrno(rc);
297	}
298
299
300	/**
301	* Unload the module.
302	*
303	* @todo We have to prevent this if we're busy!
304	*/
305	static void __exit VBoxNetFltLinuxUnload(void)
306	{
307	int rc;
308	Log(("VBoxNetFltLinuxUnload\n"));
309	Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
310
311	/*
312	* Undo the work done during start (in reverse order).
313	*/
314	rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
315	AssertRC(rc); NOREF(rc);
316
317	RTR0Term();
318
319	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
320
321	Log(("VBoxNetFltLinuxUnload - done\n"));
322	}
323
324
325	/**
326	* We filter traffic from the host to the internal network
327	* before it reaches the NIC driver.
328	*
329	* The current code uses a very ugly hack overriding hard_start_xmit
330	* callback in the device structure, but it has been shown to give us a
331	* performance boost of 60-100% though. Eventually we have to find some
332	* less hacky way of getting this job done.
333	*/
334	#define VBOXNETFLT_WITH_HOST2WIRE_FILTER
335
336	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
337
338	# if RTLNX_VER_MAX(2,6,29)
339
340	# include <linux/ethtool.h>
341
342	typedef struct ethtool_ops OVR_OPSTYPE;
343	# define OVR_OPS ethtool_ops
344	# define OVR_XMIT pfnStartXmit
345
346	# else /* RTLNX_VER_MIN(2,6,29) */
347
348	typedef struct net_device_ops OVR_OPSTYPE;
349	# define OVR_OPS netdev_ops
350	# define OVR_XMIT pOrgOps->ndo_start_xmit
351
352	# endif /* RTLNX_VER_MIN(2,6,29) */
353
354	/**
355	* The overridden net_device_ops of the device we're attached to.
356	*
357	* As there is no net_device_ops structure in pre-2.6.29 kernels we override
358	* ethtool_ops instead along with hard_start_xmit callback in net_device
359	* structure.
360	*
361	* This is a very dirty hack that was created to explore how much we can improve
362	* the host to guest transfers by not CC'ing the NIC. It turns out to be
363	* the only way to filter outgoing packets for devices without TX queue.
364	*/
365	typedef struct VBoxNetDeviceOpsOverride
366	{
367	/** Our overridden ops. */
368	OVR_OPSTYPE Ops;
369	/** Magic word. */
370	uint32_t u32Magic;
371	/** Pointer to the original ops. */
372	OVR_OPSTYPE const *pOrgOps;
373	# if RTLNX_VER_MAX(2,6,29)
374	/** Pointer to the original hard_start_xmit function. */
375	int (pfnStartXmit)(struct sk_buff pSkb, struct net_device *pDev);
376	# endif /* RTLNX_VER_MAX(2,6,29) */
377	/** Pointer to the net filter instance. */
378	PVBOXNETFLTINS pVBoxNetFlt;
379	/** The number of filtered packages. */
380	uint64_t cFiltered;
381	/** The total number of packets */
382	uint64_t cTotal;
383	} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
384	/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
385	#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
386
387	/**
388	* ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
389	* because they belong on the internal network.
390	*
391	* @returns NETDEV_TX_XXX.
392	* @param pSkb The socket buffer to transmit.
393	* @param pDev The net device.
394	*/
395	static int vboxNetFltLinuxStartXmitFilter(struct sk_buff pSkb, struct net_device pDev)
396	{
397	PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
398	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
399	PCRTNETETHERHDR pEtherHdr;
400	PINTNETTRUNKSWPORT pSwitchPort;
401	uint32_t cbHdrs;
402
403
404	/*
405	* Validate the override structure.
406	*
407	* Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
408	* to be production quality code, we would have to be much more
409	* careful here and avoid the race.
410	*/
411	if ( !VALID_PTR(pOverride)
412	\|\| pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
413	# if RTLNX_VER_MIN(2,6,29)
414	\|\| !VALID_PTR(pOverride->pOrgOps)
415	# endif /* RTLNX_VER_MIN(2,6,29) */
416	)
417	{
418	printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
419	dev_kfree_skb(pSkb);
420	return NETDEV_TX_OK;
421	}
422	pOverride->cTotal++;
423
424	/*
425	* Do the filtering base on the default OUI of our virtual NICs
426	*
427	* Note! In a real solution, we would ask the switch whether the
428	* destination MAC is 100% to be on the internal network and then
429	* drop it.
430	*/
431	cbHdrs = skb_headlen(pSkb);
432	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
433	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
434	if ( pEtherHdr
435	&& VALID_PTR(pOverride->pVBoxNetFlt)
436	&& (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
437	&& VALID_PTR(pSwitchPort)
438	&& cbHdrs >= 6)
439	{
440	INTNETSWDECISION enmDecision;
441
442	/** @todo consider reference counting, etc. */
443	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
444	if (enmDecision == INTNETSWDECISION_INTNET)
445	{
446	dev_kfree_skb(pSkb);
447	pOverride->cFiltered++;
448	return NETDEV_TX_OK;
449	}
450	}
451
452	return pOverride->OVR_XMIT(pSkb, pDev);
453	}
454
455	/**
456	* Hooks the device ndo_start_xmit operation of the device.
457	*
458	* @param pThis The net filter instance.
459	* @param pDev The net device.
460	*/
461	static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
462	{
463	PVBOXNETDEVICEOPSOVERRIDE pOverride;
464
465	/* Cancel override if ethtool_ops is missing (host-only case, @bugref{5712}) */
466	if (!VALID_PTR(pDev->OVR_OPS))
467	return;
468	pOverride = RTMemAlloc(sizeof(*pOverride));
469	if (!pOverride)
470	return;
471	pOverride->pOrgOps = pDev->OVR_OPS;
472	pOverride->Ops = *pDev->OVR_OPS;
473	# if RTLNX_VER_MAX(2,6,29)
474	pOverride->pfnStartXmit = pDev->hard_start_xmit;
475	# else /* RTLNX_VER_MIN(2,6,29) */
476	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
477	# endif /* RTLNX_VER_MIN(2,6,29) */
478	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
479	pOverride->cTotal = 0;
480	pOverride->cFiltered = 0;
481	pOverride->pVBoxNetFlt = pThis;
482
483	RTSpinlockAcquire(pThis->hSpinlock); /* (this isn't necessary, but so what) */
484	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
485	# if RTLNX_VER_MAX(2,6,29)
486	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
487	# endif /* RTLNX_VER_MAX(2,6,29) */
488	RTSpinlockRelease(pThis->hSpinlock);
489	}
490
491	/**
492	* Undos what vboxNetFltLinuxHookDev did.
493	*
494	* @param pThis The net filter instance.
495	* @param pDev The net device. Can be NULL, in which case
496	* we'll try retrieve it from @a pThis.
497	*/
498	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
499	{
500	PVBOXNETDEVICEOPSOVERRIDE pOverride;
501
502	RTSpinlockAcquire(pThis->hSpinlock);
503	if (!pDev)
504	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
505	if (VALID_PTR(pDev))
506	{
507	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
508	if ( VALID_PTR(pOverride)
509	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
510	&& VALID_PTR(pOverride->pOrgOps)
511	)
512	{
513	# if RTLNX_VER_MAX(2,6,29)
514	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
515	# endif /* RTLNX_VER_MAX(2,6,29) */
516	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
517	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
518	}
519	else
520	pOverride = NULL;
521	}
522	else
523	pOverride = NULL;
524	RTSpinlockRelease(pThis->hSpinlock);
525
526	if (pOverride)
527	{
528	printk("vboxnetflt: %llu out of %llu packets were not sent (directed to host)\n", pOverride->cFiltered, pOverride->cTotal);
529	RTMemFree(pOverride);
530	}
531	}
532
533	#endif /* VBOXNETFLT_WITH_HOST2WIRE_FILTER */
534
535
536	/**
537	* Reads and retains the host interface handle.
538	*
539	* @returns The handle, NULL if detached.
540	* @param pThis
541	*/
542	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
543	{
544	#if 0
545	struct net_device *pDev = NULL;
546
547	Log(("vboxNetFltLinuxRetainNetDev\n"));
548	/*
549	* Be careful here to avoid problems racing the detached callback.
550	*/
551	RTSpinlockAcquire(pThis->hSpinlock);
552	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
553	{
554	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
555	if (pDev)
556	{
557	dev_hold(pDev);
558	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
559	pDev, pDev->name,
560	#if RTLNX_VER_MIN(2,6,37)
561	netdev_refcnt_read(pDev)
562	#else
563	atomic_read(&pDev->refcnt)
564	#endif
565	));
566	}
567	}
568	RTSpinlockRelease(pThis->hSpinlock);
569
570	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
571	return pDev;
572	#else
573	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
574	#endif
575	}
576
577
578	/**
579	* Release the host interface handle previously retained
580	* by vboxNetFltLinuxRetainNetDev.
581	*
582	* @param pThis The instance.
583	* @param pDev The vboxNetFltLinuxRetainNetDev
584	* return value, NULL is fine.
585	*/
586	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
587	{
588	#if 0
589	Log(("vboxNetFltLinuxReleaseNetDev\n"));
590	NOREF(pThis);
591	if (pDev)
592	{
593	dev_put(pDev);
594	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
595	pDev, pDev->name,
596	#if RTLNX_VER_MIN(2,6,37)
597	netdev_refcnt_read(pDev)
598	#else
599	atomic_read(&pDev->refcnt)
600	#endif
601	));
602	}
603	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
604	#endif
605	}
606
607	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
608	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
609
610	/**
611	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
612	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
613	*
614	* @returns true / false accordingly.
615	* @param pBuf The sk_buff.
616	*/
617	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
618	{
619	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
620	}
621
622
623	/**
624	* Checks whether this SG list contains a GSO packet.
625	*
626	* @returns true / false accordingly.
627	* @param pSG The (scatter/)gather list.
628	*/
629	DECLINLINE(bool) vboxNetFltLinuxIsGso(PINTNETSG pSG)
630	{
631	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
632	return !((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type == PDMNETWORKGSOTYPE_INVALID);
633	#else /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
634	return false;
635	#endif /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
636	}
637
638
639	/**
640	* Find out the frame size (of a single segment in case of GSO frames).
641	*
642	* @returns the frame size.
643	* @param pSG The (scatter/)gather list.
644	*/
645	DECLINLINE(uint32_t) vboxNetFltLinuxFrameSize(PINTNETSG pSG)
646	{
647	uint16_t u16Type = 0;
648	uint32_t cbVlanTag = 0;
649	if (pSG->aSegs[0].cb >= sizeof(RTNETETHERHDR))
650	u16Type = RT_BE2H_U16(((PCRTNETETHERHDR)pSG->aSegs[0].pv)->EtherType);
651	else if (pSG->cbTotal >= sizeof(RTNETETHERHDR))
652	{
653	uint32_t off = RT_UOFFSETOF(RTNETETHERHDR, EtherType);
654	uint32_t i;
655	for (i = 0; i < pSG->cSegsUsed; ++i)
656	{
657	if (off <= pSG->aSegs[i].cb)
658	{
659	if (off + sizeof(uint16_t) <= pSG->aSegs[i].cb)
660	u16Type = RT_BE2H_U16((uint16_t )((uintptr_t)pSG->aSegs[i].pv + off));
661	else if (i + 1 < pSG->cSegsUsed)
662	u16Type = RT_BE2H_U16( ((uint16_t)( ((uint8_t *)pSG->aSegs[i].pv)[off] ) << 8)
663	+ (uint8_t )pSG->aSegs[i + 1].pv); /* ASSUMES no empty segments! */
664	/* else: frame is too short. */
665	break;
666	}
667	off -= pSG->aSegs[i].cb;
668	}
669	}
670	if (u16Type == RTNET_ETHERTYPE_VLAN)
671	cbVlanTag = 4;
672	return (vboxNetFltLinuxIsGso(pSG) ? (uint32_t)pSG->GsoCtx.cbMaxSeg + pSG->GsoCtx.cbHdrsTotal : pSG->cbTotal) - cbVlanTag;
673	}
674
675
676	/**
677	* Internal worker that create a linux sk_buff for a
678	* (scatter/)gather list.
679	*
680	* @returns Pointer to the sk_buff.
681	* @param pThis The instance.
682	* @param pSG The (scatter/)gather list.
683	* @param fDstWire Set if the destination is the wire.
684	*/
685	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
686	{
687	struct sk_buff *pPkt;
688	struct net_device *pDev;
689	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
690	unsigned fGsoType = 0;
691	#endif
692
693	if (pSG->cbTotal == 0)
694	{
695	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
696	return NULL;
697	}
698	Log5(("VBoxNetFlt: Packet to %s of %d bytes (frame=%d).\n", fDstWire?"wire":"host", pSG->cbTotal, vboxNetFltLinuxFrameSize(pSG)));
699	if (fDstWire && (vboxNetFltLinuxFrameSize(pSG) > ASMAtomicReadU32(&pThis->u.s.cbMtu) + 14))
700	{
701	static bool s_fOnce = true;
702	if (s_fOnce)
703	{
704	s_fOnce = false;
705	printk("VBoxNetFlt: Dropped over-sized packet (%d bytes) coming from internal network.\n", vboxNetFltLinuxFrameSize(pSG));
706	}
707	return NULL;
708	}
709
710	/** @todo We should use fragments mapping the SG buffers with large packets.
711	* 256 bytes seems to be the a threshold used a lot for this. It
712	* requires some nasty work on the intnet side though... */
713	/*
714	* Allocate a packet and copy over the data.
715	*/
716	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
717	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
718	if (RT_UNLIKELY(!pPkt))
719	{
720	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
721	pSG->pvUserData = NULL;
722	return NULL;
723	}
724	pPkt->dev = pDev;
725	pPkt->ip_summed = CHECKSUM_NONE;
726
727	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
728	skb_reserve(pPkt, NET_IP_ALIGN);
729
730	/* Copy the segments. */
731	skb_put(pPkt, pSG->cbTotal);
732	IntNetSgRead(pSG, pPkt->data);
733
734	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
735	/*
736	* Setup GSO if used by this packet.
737	*/
738	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
739	{
740	default:
741	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
742	/* fall thru */
743	case PDMNETWORKGSOTYPE_INVALID:
744	fGsoType = 0;
745	break;
746	case PDMNETWORKGSOTYPE_IPV4_TCP:
747	fGsoType = SKB_GSO_TCPV4;
748	break;
749	case PDMNETWORKGSOTYPE_IPV6_TCP:
750	fGsoType = SKB_GSO_TCPV6;
751	break;
752	}
753	if (fGsoType)
754	{
755	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
756
757	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
758	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
759	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
760
761	/*
762	* We need to set checksum fields even if the packet goes to the host
763	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
764	*/
765	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrsTotal);
766	pPkt->ip_summed = CHECKSUM_PARTIAL;
767	# if RTLNX_VER_MIN(2,6,22)
768	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
769	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
770	pPkt->csum_offset = RT_UOFFSETOF(RTNETTCP, th_sum);
771	else
772	pPkt->csum_offset = RT_UOFFSETOF(RTNETUDP, uh_sum);
773	# else
774	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
775	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
776	pPkt->csum = RT_UOFFSETOF(RTNETTCP, th_sum);
777	else
778	pPkt->csum = RT_UOFFSETOF(RTNETUDP, uh_sum);
779	# endif
780	if (!fDstWire)
781	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
782	}
783	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
784
785	/*
786	* Finish up the socket buffer.
787	*/
788	pPkt->protocol = eth_type_trans(pPkt, pDev);
789	if (fDstWire)
790	{
791	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
792
793	/* Restore ethernet header back. */
794	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
795	VBOX_SKB_RESET_MAC_HDR(pPkt);
796	}
797	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
798
799	return pPkt;
800	}
801
802
803	/**
804	* Return the offset where to start checksum computation from.
805	*
806	* @returns the offset relative to pBuf->data.
807	* @param pBuf The socket buffer.
808	*/
809	DECLINLINE(unsigned) vboxNetFltLinuxGetChecksumStartOffset(struct sk_buff *pBuf)
810	{
811	#if RTLNX_VER_MIN(2,6,38)
812	return skb_checksum_start_offset(pBuf);
813	#elif RTLNX_VER_MIN(2,6,22)
814	return pBuf->csum_start - skb_headroom(pBuf);
815	#else
816	unsigned char *pTransportHdr = pBuf->h.raw;
817	# if RTLNX_VER_MAX(2,6,19)
818	/*
819	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
820	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
821	* header length from the header itself and reconstruct 'h' pointer
822	* to TCP (or whatever) header.
823	*/
824	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
825	pTransportHdr = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
826	# endif
827	return pTransportHdr - pBuf->data;
828	#endif
829	}
830
831
832	/**
833	* Initializes a SG list from an sk_buff.
834	*
835	* @returns Number of segments.
836	* @param pThis The instance.
837	* @param pBuf The sk_buff.
838	* @param pSG The SG.
839	* @param cbExtra The number of bytes of extra space allocated immediately after the SG.
840	* @param cSegs The number of segments allocated for the SG.
841	* This should match the number in the mbuf exactly!
842	* @param fSrc The source of the frame.
843	* @param pGsoCtx Pointer to the GSO context if it's a GSO
844	* internal network frame. NULL if regular frame.
845	*/
846	static void vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
847	unsigned cbExtra, unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
848	{
849	int i;
850	NOREF(pThis);
851
852	#ifndef VBOXNETFLT_SG_SUPPORT
853	Assert(!skb_shinfo(pBuf)->frag_list);
854	#else /* VBOXNETFLT_SG_SUPPORT */
855	uint8_t pExtra = (uint8_t )&pSG->aSegs[cSegs];
856	unsigned cbConsumed = 0;
857	unsigned cbProduced = 0;
858
859	# if RTLNX_VER_MIN(2,6,27)
860	/* Restore VLAN tag stripped by host hardware */
861	if (vlan_tx_tag_present(pBuf))
862	{
863	uint8_t *pMac = pBuf->data;
864	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )pExtra;
865	Assert(ETH_ALEN * 2 + VLAN_HLEN <= cbExtra);
866	memmove(pVHdr, pMac, ETH_ALEN * 2);
867	/* Consume whole Ethernet header: 2 addresses + EtherType (see @bugref{8599}) */
868	cbConsumed += ETH_ALEN * 2 + sizeof(uint16_t);
869	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
870	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
871	pVHdr->h_vlan_encapsulated_proto = (uint16_t)(pMac + ETH_ALEN * 2);
872	cbProduced += VLAN_ETH_HLEN;
873	}
874	# endif /* RTLNX_VER_MIN(2,6,27) */
875
876	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
877	{
878	unsigned uCsumStartOffset = vboxNetFltLinuxGetChecksumStartOffset(pBuf);
879	unsigned uCsumStoreOffset = uCsumStartOffset + VBOX_SKB_CSUM_OFFSET(pBuf) - cbConsumed;
880	Log3(("cbConsumed=%u cbProduced=%u uCsumStartOffset=%u uCsumStoreOffset=%u\n",
881	cbConsumed, cbProduced, uCsumStartOffset, uCsumStoreOffset));
882	Assert(cbProduced + uCsumStoreOffset + sizeof(uint16_t) <= cbExtra);
883	/*
884	* We assume that the checksum is stored at the very end of the transport header
885	* so we will have all headers in a single fragment. If our assumption is wrong
886	* we may see suboptimal performance.
887	*/
888	memmove(pExtra + cbProduced,
889	pBuf->data + cbConsumed,
890	uCsumStoreOffset);
891	unsigned uChecksum = skb_checksum(pBuf, uCsumStartOffset, pBuf->len - uCsumStartOffset, 0);
892	(uint16_t)(pExtra + cbProduced + uCsumStoreOffset) = csum_fold(uChecksum);
893	cbProduced += uCsumStoreOffset + sizeof(uint16_t);
894	cbConsumed += uCsumStoreOffset + sizeof(uint16_t);
895	}
896	#endif /* VBOXNETFLT_SG_SUPPORT */
897
898	if (!pGsoCtx)
899	IntNetSgInitTempSegs(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/);
900	else
901	IntNetSgInitTempSegsGso(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/, pGsoCtx);
902
903	int iSeg = 0;
904	#ifdef VBOXNETFLT_SG_SUPPORT
905	if (cbProduced)
906	{
907	pSG->aSegs[iSeg].cb = cbProduced;
908	pSG->aSegs[iSeg].pv = pExtra;
909	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
910	}
911	pSG->aSegs[iSeg].cb = skb_headlen(pBuf) - cbConsumed;
912	pSG->aSegs[iSeg].pv = pBuf->data + cbConsumed;
913	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
914	Assert(iSeg <= pSG->cSegsAlloc);
915
916	# ifdef LOG_ENABLED
917	if (pBuf->data_len)
918	Log6((" kmap_atomic:"));
919	# endif /* LOG_ENABLED */
920	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
921	{
922	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
923	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
924	pSG->aSegs[iSeg].cb = pFrag->bv_len;
925	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
926	# else /* < KERNEL_VERSION(5, 4, 0) */
927	pSG->aSegs[iSeg].cb = pFrag->size;
928	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
929	# endif /* >= KERNEL_VERSION(5, 4, 0) */
930	Log6((" %p", pSG->aSegs[iSeg].pv));
931	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
932	Assert(iSeg <= pSG->cSegsAlloc);
933	}
934	struct sk_buff *pFragBuf;
935	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
936	{
937	pSG->aSegs[iSeg].cb = skb_headlen(pFragBuf);
938	pSG->aSegs[iSeg].pv = pFragBuf->data;
939	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
940	Assert(iSeg <= pSG->cSegsAlloc);
941	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
942	{
943	skb_frag_t *pFrag = &skb_shinfo(pFragBuf)->frags[i];
944	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
945	pSG->aSegs[iSeg].cb = pFrag->bv_len;
946	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
947	# else /* < KERNEL_VERSION(5, 4, 0) */
948	pSG->aSegs[iSeg].cb = pFrag->size;
949	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
950	# endif /* >= KERNEL_VERSION(5, 4, 0) */
951	Log6((" %p", pSG->aSegs[iSeg].pv));
952	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
953	Assert(iSeg <= pSG->cSegsAlloc);
954	}
955	}
956	# ifdef LOG_ENABLED
957	if (pBuf->data_len)
958	Log6(("\n"));
959	# endif /* LOG_ENABLED */
960	#else
961	pSG->aSegs[iSeg].cb = pBuf->len;
962	pSG->aSegs[iSeg].pv = pBuf->data;
963	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
964	#endif
965
966	pSG->cSegsUsed = iSeg;
967
968	#if 0
969	if (cbProduced)
970	{
971	LogRel(("vboxNetFltLinuxSkBufToSG: original packet dump:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
972	LogRel(("vboxNetFltLinuxSkBufToSG: cbConsumed=%u cbProduced=%u cbExtra=%u\n", cbConsumed, cbProduced, cbExtra));
973	uint32_t offset = 0;
974	for (i = 0; i < pSG->cSegsUsed; ++i)
975	{
976	LogRel(("vboxNetFltLinuxSkBufToSG: seg#%d (%d bytes, starting at 0x%x):\n%.*Rhxd\n",
977	i, pSG->aSegs[i].cb, offset, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
978	offset += pSG->aSegs[i].cb;
979	}
980	}
981	#endif
982
983	#ifdef PADD_RUNT_FRAMES_FROM_HOST
984	/*
985	* Add a trailer if the frame is too small.
986	*
987	* Since we're getting to the packet before it is framed, it has not
988	* yet been padded. The current solution is to add a segment pointing
989	* to a buffer containing all zeros and pray that works for all frames...
990	*/
991	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
992	{
993	Assert(pBuf->data_len == 0); /* Packets with fragments are never small! */
994	static uint8_t const s_abZero[128] = {0};
995
996	AssertReturnVoid(iSeg < cSegs);
997
998	pSG->aSegs[iSeg].Phys = NIL_RTHCPHYS;
999	pSG->aSegs[iSeg].pv = (void *)&s_abZero[0];
1000	pSG->aSegs[iSeg++].cb = 60 - pSG->cbTotal;
1001	pSG->cbTotal = 60;
1002	pSG->cSegsUsed++;
1003	Assert(iSeg <= pSG->cSegsAlloc)
1004	}
1005	#endif
1006
1007	Log6(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1008	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1009	for (i = 0; i < pSG->cSegsUsed; i++)
1010	Log6(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1011	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1012	}
1013
1014	/**
1015	* Packet handler; not really documented - figure it out yourself.
1016	*
1017	* @returns 0 or EJUSTRETURN - this is probably copy & pastry and thus wrong.
1018	*/
1019	#if RTLNX_VER_MIN(2,6,14)
1020	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1021	struct net_device *pSkbDev,
1022	struct packet_type *pPacketType,
1023	struct net_device *pOrigDev)
1024	#else
1025	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1026	struct net_device *pSkbDev,
1027	struct packet_type *pPacketType)
1028	#endif
1029	{
1030	PVBOXNETFLTINS pThis;
1031	struct net_device *pDev;
1032	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1033	pBuf, pSkbDev, pPacketType));
1034	#if RTLNX_VER_MIN(2,6,18)
1035	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",
1036	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));
1037	# if RTLNX_VER_MIN(2,6,22)
1038	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1039	# endif
1040	#else
1041	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",
1042	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));
1043	#endif
1044	/*
1045	* Drop it immediately?
1046	*/
1047	if (!pBuf)
1048	return 0;
1049
1050	if (pBuf->pkt_type == PACKET_LOOPBACK)
1051	{
1052	/*
1053	* We are not interested in loopbacked packets as they will always have
1054	* another copy going to the wire.
1055	*/
1056	Log2(("vboxNetFltLinuxPacketHandler: dropped loopback packet (cb=%u)\n", pBuf->len));
1057	dev_kfree_skb(pBuf); /* We must 'consume' all packets we get (@bugref{6539})! */
1058	return 0;
1059	}
1060
1061	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1062	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1063	if (pDev != pSkbDev)
1064	{
1065	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1066	kfree_skb(pBuf); /* This is a failure, so we use kfree_skb instead of dev_kfree_skb. */
1067	return 0;
1068	}
1069
1070	Log6(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1071	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1072	{
1073	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1074	dev_kfree_skb(pBuf);
1075	return 0;
1076	}
1077
1078	#ifndef VBOXNETFLT_SG_SUPPORT
1079	{
1080	/*
1081	* Get rid of fragmented packets, they cause too much trouble.
1082	*/
1083	unsigned int uMacLen = pBuf->mac_len;
1084	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1085	dev_kfree_skb(pBuf);
1086	if (!pCopy)
1087	{
1088	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1089	return 0;
1090	}
1091	pBuf = pCopy;
1092	/* Somehow skb_copy ignores mac_len */
1093	pBuf->mac_len = uMacLen;
1094	# if RTLNX_VER_MIN(2,6,27)
1095	/* Restore VLAN tag stripped by host hardware */
1096	if (vlan_tx_tag_present(pBuf) && skb_headroom(pBuf) >= VLAN_ETH_HLEN)
1097	{
1098	uint8_t pMac = (uint8_t)skb_mac_header(pBuf);
1099	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )(pMac - VLAN_HLEN);
1100	memmove(pVHdr, pMac, ETH_ALEN * 2);
1101	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
1102	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
1103	pBuf->mac_header -= VLAN_HLEN;
1104	pBuf->mac_len += VLAN_HLEN;
1105	}
1106	# endif /* RTLNX_VER_MIN(2,6,27) */
1107
1108	# if RTLNX_VER_MIN(2,6,18)
1109	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",
1110	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));
1111	# if RTLNX_VER_MIN(2,6,22)
1112	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1113	# endif /* RTLNX_VER_MIN(2,6,22) */
1114	# else /* RTLNX_VER_MAX(2,6,18) */
1115	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",
1116	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));
1117	# endif /* RTLNX_VER_MAX(2,6,18) */
1118	}
1119	#endif /* !VBOXNETFLT_SG_SUPPORT */
1120
1121	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1122	/* Forward it to the internal network. */
1123	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1124	#else /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1125	/* Add the packet to transmit queue and schedule the bottom half. */
1126	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1127	schedule_work(&pThis->u.s.XmitTask);
1128	Log6(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1129	&pThis->u.s.XmitTask, pBuf));
1130	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1131
1132	/* It does not really matter what we return, it is ignored by the kernel. */
1133	return 0;
1134	}
1135
1136	/**
1137	* Calculate the number of INTNETSEG segments the socket buffer will need.
1138	*
1139	* @returns Segment count.
1140	* @param pBuf The socket buffer.
1141	* @param pcbTemp Where to store the number of bytes of the part
1142	* of the socket buffer that will be copied to
1143	* a temporary storage.
1144	*/
1145	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff pBuf, unsigned pcbTemp)
1146	{
1147	*pcbTemp = 0;
1148	#ifdef VBOXNETFLT_SG_SUPPORT
1149	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1150	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1151	{
1152	*pcbTemp = vboxNetFltLinuxGetChecksumStartOffset(pBuf) + VBOX_SKB_CSUM_OFFSET(pBuf) + sizeof(uint16_t);
1153	}
1154	# if RTLNX_VER_MIN(2,6,27)
1155	if (vlan_tx_tag_present(pBuf))
1156	{
1157	if (*pcbTemp)
1158	*pcbTemp += VLAN_HLEN;
1159	else
1160	*pcbTemp = VLAN_ETH_HLEN;
1161	}
1162	# endif /* RTLNX_VER_MIN(2,6,27) */
1163	if (*pcbTemp)
1164	++cSegs;
1165	struct sk_buff *pFrag;
1166	for (pFrag = skb_shinfo(pBuf)->frag_list; pFrag; pFrag = pFrag->next)
1167	{
1168	Log6(("vboxNetFltLinuxCalcSGSegments: frag=%p len=%d data_len=%d frags=%d frag_list=%p next=%p\n",
1169	pFrag, pFrag->len, pFrag->data_len, skb_shinfo(pFrag)->nr_frags, skb_shinfo(pFrag)->frag_list, pFrag->next));
1170	cSegs += 1 + skb_shinfo(pFrag)->nr_frags;
1171	}
1172	#else
1173	unsigned cSegs = 1;
1174	#endif
1175	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1176	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1177	if (pBuf->len < 60)
1178	cSegs++;
1179	#endif
1180	return cSegs;
1181	}
1182
1183
1184	/**
1185	* Destroy the intnet scatter / gather buffer created by
1186	* vboxNetFltLinuxSkBufToSG.
1187	*
1188	* @param pSG The (scatter/)gather list.
1189	* @param pBuf The original socket buffer that was used to create
1190	* the scatter/gather list.
1191	*/
1192	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG, struct sk_buff *pBuf)
1193	{
1194	#ifdef VBOXNETFLT_SG_SUPPORT
1195	int i, iSeg = 1; /* Skip non-paged part of SKB */
1196	/* Check if the extra buffer behind SG structure was used for modified packet header */
1197	if (pBuf->data != pSG->aSegs[0].pv)
1198	++iSeg; /* Skip it as well */
1199	# ifdef LOG_ENABLED
1200	if (pBuf->data_len)
1201	Log6(("kunmap_atomic:"));
1202	# endif /* LOG_ENABLED */
1203	/* iSeg now points to the first mapped fragment if there are any */
1204	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1205	{
1206	Log6((" %p", pSG->aSegs[iSeg].pv));
1207	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1208	}
1209	struct sk_buff *pFragBuf;
1210	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
1211	{
1212	++iSeg; /* Non-fragment (unmapped) portion of chained SKB */
1213	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
1214	{
1215	Log6((" %p", pSG->aSegs[iSeg].pv));
1216	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1217	}
1218	}
1219	# ifdef LOG_ENABLED
1220	if (pBuf->data_len)
1221	Log6(("\n"));
1222	# endif /* LOG_ENABLED */
1223	#endif
1224	NOREF(pSG);
1225	}
1226
1227	#ifdef LOG_ENABLED
1228	/**
1229	* Logging helper.
1230	*/
1231	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1232	{
1233	int i, offSeg;
1234	uint8_t pInt, pExt;
1235	static int iPacketNo = 1;
1236	iPacketNo += iIncrement;
1237	if (fEgress)
1238	{
1239	pExt = pSG->aSegs[0].pv;
1240	pInt = pExt + 6;
1241	}
1242	else
1243	{
1244	pInt = pSG->aSegs[0].pv;
1245	pExt = pInt + 6;
1246	}
1247	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1248	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1249	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1250	fEgress ? "-->" : "<--", pszWhere,
1251	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1252	pSG->cbTotal, iPacketNo));
1253	if (pSG->cSegsUsed == 1)
1254	{
1255	Log4(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1256	}
1257	else
1258	{
1259	for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1260	{
1261	Log4(("-- segment %d at 0x%x (%d bytes)\n --\n%.*Rhxd\n",
1262	i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1263	offSeg += pSG->aSegs[i].cb;
1264	}
1265	}
1266	}
1267	#else
1268	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1269	#endif
1270
1271	#ifdef VBOXNETFLT_WITH_GSO_RECV
1272
1273	/**
1274	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1275	* GSO socket buffer without having to segment it.
1276	*
1277	* @returns true on success, false if needs segmenting.
1278	* @param pThis The net filter instance.
1279	* @param pSkb The GSO socket buffer.
1280	* @param fSrc The source.
1281	* @param pGsoCtx Where to return the GSO context on success.
1282	*/
1283	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1284	PPDMNETWORKGSO pGsoCtx)
1285	{
1286	PDMNETWORKGSOTYPE enmGsoType;
1287	uint16_t uEtherType;
1288	unsigned int cbTransport;
1289	unsigned int offTransport;
1290	unsigned int cbTransportHdr;
1291	unsigned uProtocol;
1292	union
1293	{
1294	RTNETIPV4 IPv4;
1295	RTNETIPV6 IPv6;
1296	RTNETTCP Tcp;
1297	uint8_t ab[40];
1298	uint16_t au16[40/2];
1299	uint32_t au32[40/4];
1300	} Buf;
1301
1302	/*
1303	* Check the GSO properties of the socket buffer and make sure it fits.
1304	*/
1305	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1306	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1307	{
1308	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1309	return false;
1310	}
1311	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1312	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1313	{
1314	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1315	return false;
1316	}
1317
1318	/*
1319	* Switch on the ethertype.
1320	*/
1321	uEtherType = pSkb->protocol;
1322	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1323	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1324	{
1325	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1326	if (puEtherType)
1327	uEtherType = *puEtherType;
1328	}
1329	switch (uEtherType)
1330	{
1331	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1332	{
1333	unsigned int cbHdr;
1334	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1335	if (RT_UNLIKELY(!pIPv4))
1336	{
1337	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1338	return false;
1339	}
1340
1341	cbHdr = pIPv4->ip_hl * 4;
1342	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1343	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1344	\|\| cbHdr > cbTransport ))
1345	{
1346	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1347	return false;
1348	}
1349	cbTransport -= cbHdr;
1350	offTransport = pSkb->mac_len + cbHdr;
1351	uProtocol = pIPv4->ip_p;
1352	if (uProtocol == RTNETIPV4_PROT_TCP)
1353	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1354	else if (uProtocol == RTNETIPV4_PROT_UDP)
1355	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1356	else /** @todo IPv6: 4to6 tunneling */
1357	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1358	break;
1359	}
1360
1361	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1362	{
1363	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1364	if (RT_UNLIKELY(!pIPv6))
1365	{
1366	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1367	return false;
1368	}
1369
1370	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1371	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1372	uProtocol = pIPv6->ip6_nxt;
1373	/** @todo IPv6: Dig our way out of the other headers. */
1374	if (uProtocol == RTNETIPV4_PROT_TCP)
1375	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1376	else if (uProtocol == RTNETIPV4_PROT_UDP)
1377	enmGsoType = PDMNETWORKGSOTYPE_IPV6_UDP;
1378	else
1379	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1380	break;
1381	}
1382
1383	default:
1384	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1385	return false;
1386	}
1387
1388	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1389	{
1390	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1391	return false;
1392	}
1393
1394	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1395	\|\| offTransport + cbTransport > pSkb->len
1396	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1397	{
1398	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1399	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1400	return false;
1401	}
1402
1403	/*
1404	* Check the TCP/UDP bits.
1405	*/
1406	if (uProtocol == RTNETIPV4_PROT_TCP)
1407	{
1408	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1409	if (RT_UNLIKELY(!pTcp))
1410	{
1411	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1412	return false;
1413	}
1414
1415	cbTransportHdr = pTcp->th_off * 4;
1416	pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1417	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1418	\|\| cbTransportHdr > cbTransport
1419	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1420	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1421	{
1422	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1423	return false;
1424	}
1425
1426	}
1427	else
1428	{
1429	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1430	cbTransportHdr = sizeof(RTNETUDP);
1431	pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1432	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1433	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1434	{
1435	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1436	return false;
1437	}
1438	}
1439
1440	/*
1441	* We're good, init the GSO context.
1442	*/
1443	pGsoCtx->u8Type = enmGsoType;
1444	pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1445	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1446	pGsoCtx->offHdr1 = pSkb->mac_len;
1447	pGsoCtx->offHdr2 = offTransport;
1448	pGsoCtx->u8Unused = 0;
1449
1450	return true;
1451	}
1452
1453	/**
1454	* Forward the socket buffer as a GSO internal network frame.
1455	*
1456	* @returns IPRT status code.
1457	* @param pThis The net filter instance.
1458	* @param pSkb The GSO socket buffer.
1459	* @param fSrc The source.
1460	* @param pGsoCtx Where to return the GSO context on success.
1461	*/
1462	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1463	{
1464	int rc;
1465	unsigned cbExtra;
1466	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb, &cbExtra);
1467	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1468	if (RT_LIKELY(pSG))
1469	{
1470	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cbExtra, cSegs, fSrc, pGsoCtx);
1471
1472	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1473	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1474
1475	vboxNetFltLinuxDestroySG(pSG, pSkb);
1476	rc = VINF_SUCCESS;
1477	}
1478	else
1479	{
1480	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1481	rc = VERR_NO_MEMORY;
1482	}
1483	return rc;
1484	}
1485
1486	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1487
1488	/**
1489	* Worker for vboxNetFltLinuxForwardToIntNet.
1490	*
1491	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1492	* @param pThis The net filter instance.
1493	* @param pBuf The socket buffer.
1494	* @param fSrc The source.
1495	*/
1496	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1497	{
1498	int rc;
1499	unsigned cbExtra;
1500	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf, &cbExtra);
1501	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1502	if (RT_LIKELY(pSG))
1503	{
1504	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cbExtra, cSegs, fSrc, NULL /pGsoCtx/);
1505
1506	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1507	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1508
1509	vboxNetFltLinuxDestroySG(pSG, pBuf);
1510	rc = VINF_SUCCESS;
1511	}
1512	else
1513	{
1514	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1515	rc = VERR_NO_MEMORY;
1516	}
1517	return rc;
1518	}
1519
1520
1521	/**
1522	* I won't disclose what I do, figure it out yourself, including pThis referencing.
1523	*
1524	* @param pThis The net filter instance.
1525	* @param pBuf The socket buffer.
1526	* @param fSrc Where the packet comes from.
1527	*/
1528	static void vboxNetFltLinuxForwardToIntNetInner(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1529	{
1530	#ifdef VBOXNETFLT_WITH_GSO
1531	if (skb_is_gso(pBuf))
1532	{
1533	PDMNETWORKGSO GsoCtx;
1534	Log6(("vboxNetFltLinuxForwardToIntNetInner: skb len=%u data_len=%u truesize=%u next=%p"
1535	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x ip_summed=%d\n",
1536	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next,
1537	skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size,
1538	skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type,
1539	skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, pBuf->ip_summed));
1540	#ifndef VBOXNETFLT_SG_SUPPORT
1541	if (RT_LIKELY(fSrc & INTNETTRUNKDIR_HOST))
1542	{
1543	/*
1544	* skb_gso_segment does the following. Do we need to do it as well?
1545	*/
1546	# if RTLNX_VER_MIN(2,6,22)
1547	skb_reset_mac_header(pBuf);
1548	pBuf->mac_len = pBuf->network_header - pBuf->mac_header;
1549	# else
1550	pBuf->mac.raw = pBuf->data;
1551	pBuf->mac_len = pBuf->nh.raw - pBuf->data;
1552	# endif
1553	}
1554	#endif /* !VBOXNETFLT_SG_SUPPORT */
1555	# ifdef VBOXNETFLT_WITH_GSO_RECV
1556	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1557	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1558	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1559	else
1560	# endif /* VBOXNETFLT_WITH_GSO_RECV */
1561	{
1562	/* Need to segment the packet */
1563	struct sk_buff *pNext;
1564	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1565	if (IS_ERR(pSegment))
1566	{
1567	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1568	return;
1569	}
1570
1571	for (; pSegment; pSegment = pNext)
1572	{
1573	Log6(("vboxNetFltLinuxForwardToIntNetInner: segment len=%u data_len=%u truesize=%u next=%p"
1574	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1575	pSegment->len, pSegment->data_len, pSegment->truesize, pSegment->next,
1576	skb_shinfo(pSegment)->nr_frags, skb_shinfo(pSegment)->gso_size,
1577	skb_shinfo(pSegment)->gso_segs, skb_shinfo(pSegment)->gso_type,
1578	skb_shinfo(pSegment)->frag_list, pSegment->pkt_type));
1579	pNext = pSegment->next;
1580	pSegment->next = 0;
1581	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1582	dev_kfree_skb(pSegment);
1583	}
1584	}
1585	}
1586	else
1587	#endif /* VBOXNETFLT_WITH_GSO */
1588	{
1589	Log6(("vboxNetFltLinuxForwardToIntNetInner: ptk_type=%d ip_summed=%d len=%d"
1590	" data_len=%d headroom=%d hdr_len=%d csum_offset=%d\n",
1591	pBuf->pkt_type, pBuf->ip_summed, pBuf->len, pBuf->data_len, skb_headroom(pBuf),
1592	skb_headlen(pBuf), vboxNetFltLinuxGetChecksumStartOffset(pBuf)));
1593	#ifndef VBOXNETFLT_SG_SUPPORT
1594	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1595	{
1596	# if RTLNX_VER_MIN(2,6,19)
1597	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1598	# else
1599	/*
1600	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1601	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1602	* header length from the header itself and reconstruct 'h' pointer
1603	* to TCP (or whatever) header.
1604	*/
1605	unsigned char *tmp = pBuf->h.raw;
1606	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1607	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1608	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1609	/* Restore the original (wrong) pointer. */
1610	pBuf->h.raw = tmp;
1611	# endif
1612	if (rc)
1613	{
1614	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1615	return;
1616	}
1617	}
1618	#endif /* !VBOXNETFLT_SG_SUPPORT */
1619	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1620	}
1621	}
1622
1623
1624	/**
1625	* Temporarily adjust pBuf->data so it always points to the Ethernet header,
1626	* then forward it to the internal network.
1627	*
1628	* @param pThis The net filter instance.
1629	* @param pBuf The socket buffer. This is consumed by this function.
1630	*/
1631	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1632	{
1633	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1634
1635	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1636	{
1637	/*
1638	* The packet came from the wire and the driver has already consumed
1639	* mac header. We need to restore it back. Moreover, after we are
1640	* through with this skb we need to restore its original state!
1641	*/
1642	skb_push(pBuf, pBuf->mac_len);
1643	Log5(("vboxNetFltLinuxForwardToIntNet: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1644	pBuf->mac_len, pBuf->data, skb_mac_header(pBuf), skb_network_header(pBuf)));
1645	}
1646
1647	vboxNetFltLinuxForwardToIntNetInner(pThis, pBuf, fSrc);
1648
1649	/*
1650	* Restore the original state of skb as there are other handlers this skb
1651	* will be provided to.
1652	*/
1653	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1654	skb_pull(pBuf, pBuf->mac_len);
1655
1656	dev_kfree_skb(pBuf);
1657	}
1658
1659
1660	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1661	/**
1662	* Work queue handler that forwards the socket buffers queued by
1663	* vboxNetFltLinuxPacketHandler to the internal network.
1664	*
1665	* @param pWork The work queue.
1666	*/
1667	# if RTLNX_VER_MIN(2,6,20)
1668	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1669	# else
1670	static void vboxNetFltLinuxXmitTask(void *pWork)
1671	# endif
1672	{
1673	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1674	struct sk_buff *pBuf;
1675
1676	Log6(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1677
1678	/*
1679	* Active? Retain the instance and increment the busy counter.
1680	*/
1681	if (vboxNetFltTryRetainBusyActive(pThis))
1682	{
1683	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1684	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1685
1686	vboxNetFltRelease(pThis, true /* fBusy */);
1687	}
1688	else
1689	{
1690	/** @todo Shouldn't we just drop the packets here? There is little point in
1691	* making them accumulate when the VM is paused and it'll only waste
1692	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1693	* before start draining the packets (goes for the intnet ring buf
1694	* too)? */
1695	}
1696	}
1697	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1698
1699	/**
1700	* Reports the GSO capabilities of the hardware NIC.
1701	*
1702	* @param pThis The net filter instance. The caller hold a
1703	* reference to this.
1704	*/
1705	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1706	{
1707	#ifdef VBOXNETFLT_WITH_GSO_XMIT_WIRE
1708	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1709	{
1710	struct net_device *pDev;
1711	PINTNETTRUNKSWPORT pSwitchPort;
1712	unsigned int fFeatures;
1713
1714	RTSpinlockAcquire(pThis->hSpinlock);
1715
1716	pSwitchPort = pThis->pSwitchPort; /* this doesn't need to be here, but it doesn't harm. */
1717	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1718	if (pDev)
1719	fFeatures = pDev->features;
1720	else
1721	fFeatures = 0;
1722
1723	RTSpinlockRelease(pThis->hSpinlock);
1724
1725	if (pThis->pSwitchPort)
1726	{
1727	/* Set/update the GSO capabilities of the NIC. */
1728	uint32_t fGsoCapabilites = 0;
1729	if (fFeatures & NETIF_F_TSO)
1730	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1731	if (fFeatures & NETIF_F_TSO6)
1732	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1733	Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s%s%s\n",
1734	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1735	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : ""));
1736	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1737	}
1738
1739	vboxNetFltRelease(pThis, true /fBusy/);
1740	}
1741	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE */
1742	}
1743
1744	/**
1745	* Helper that determines whether the host (ignoreing us) is operating the
1746	* interface in promiscuous mode or not.
1747	*/
1748	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1749	{
1750	bool fRc = false;
1751	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1752	if (pDev)
1753	{
1754	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1755	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1756	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1757	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1758	}
1759	return fRc;
1760	}
1761
1762	/**
1763	* Does this device needs link state change signaled?
1764	* Currently we need it for our own VBoxNetAdp and TAP.
1765	*/
1766	static bool vboxNetFltNeedsLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev)
1767	{
1768	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1769	{
1770	struct ethtool_drvinfo Info;
1771
1772	memset(&Info, 0, sizeof(Info));
1773	Info.cmd = ETHTOOL_GDRVINFO;
1774	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1775	Log3(("%s: driver=%.s version=%.s bus_info=%.*s\n",
1776	__FUNCTION__,
1777	sizeof(Info.driver), Info.driver,
1778	sizeof(Info.version), Info.version,
1779	sizeof(Info.bus_info), Info.bus_info));
1780
1781	if (!strncmp(Info.driver, "vboxnet", sizeof(Info.driver)))
1782	return true;
1783
1784	#if RTLNX_VER_MIN(2,6,36) /* TAP started doing carrier */
1785	return !strncmp(Info.driver, "tun", 4)
1786	&& !strncmp(Info.bus_info, "tap", 4);
1787	#endif
1788	}
1789
1790	return false;
1791	}
1792
1793	#if RTLNX_VER_MAX(2,6,18)
1794	DECLINLINE(void) netif_tx_lock_bh(struct net_device *pDev)
1795	{
1796	spin_lock_bh(&pDev->xmit_lock);
1797	}
1798
1799	DECLINLINE(void) netif_tx_unlock_bh(struct net_device *pDev)
1800	{
1801	spin_unlock_bh(&pDev->xmit_lock);
1802	}
1803	#endif
1804
1805	/**
1806	* Some devices need link state change when filter attaches/detaches
1807	* since the filter is their link in a sense.
1808	*/
1809	static void vboxNetFltSetLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1810	{
1811	if (vboxNetFltNeedsLinkState(pThis, pDev))
1812	{
1813	Log3(("%s: bringing device link %s\n",
1814	__FUNCTION__, fLinkUp ? "up" : "down"));
1815	netif_tx_lock_bh(pDev);
1816	if (fLinkUp)
1817	netif_carrier_on(pDev);
1818	else
1819	netif_carrier_off(pDev);
1820	netif_tx_unlock_bh(pDev);
1821	}
1822	}
1823
1824	/**
1825	* Internal worker for vboxNetFltLinuxNotifierCallback.
1826	*
1827	* @returns VBox status code.
1828	* @param pThis The instance.
1829	* @param pDev The device to attach to.
1830	*/
1831	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1832	{
1833	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1834
1835	/*
1836	* Retain and store the device.
1837	*/
1838	dev_hold(pDev);
1839
1840	RTSpinlockAcquire(pThis->hSpinlock);
1841	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
1842	RTSpinlockRelease(pThis->hSpinlock);
1843
1844	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
1845	pDev, pDev->name,
1846	#if RTLNX_VER_MIN(2,6,37)
1847	netdev_refcnt_read(pDev)
1848	#else
1849	atomic_read(&pDev->refcnt)
1850	#endif
1851	));
1852	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1853	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1854
1855	/* Get the mac address while we still have a valid net_device reference. */
1856	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
1857	/* Initialize MTU */
1858	pThis->u.s.cbMtu = pDev->mtu;
1859
1860	/*
1861	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1862	*/
1863	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1864	pThis->u.s.PacketType.dev = pDev;
1865	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1866	dev_add_pack(&pThis->u.s.PacketType);
1867	ASMAtomicUoWriteBool(&pThis->u.s.fPacketHandler, true);
1868	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1869
1870	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1871	vboxNetFltLinuxHookDev(pThis, pDev);
1872	#endif
1873
1874	/*
1875	* Are we the "carrier" for this device (e.g. vboxnet or tap)?
1876	*/
1877	vboxNetFltSetLinkState(pThis, pDev, true);
1878
1879	/*
1880	* Set indicators that require the spinlock. Be abit paranoid about racing
1881	* the device notification handle.
1882	*/
1883	RTSpinlockAcquire(pThis->hSpinlock);
1884	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1885	if (pDev)
1886	{
1887	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1888	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1889	pDev = NULL; /* don't dereference it */
1890	}
1891	RTSpinlockRelease(pThis->hSpinlock);
1892
1893	/*
1894	* Report GSO capabilities
1895	*/
1896	Assert(pThis->pSwitchPort);
1897	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1898	{
1899	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1900	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
1901	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
1902	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
1903	vboxNetFltRelease(pThis, true /fBusy/);
1904	}
1905
1906	LogRel(("VBoxNetFlt: attached to '%s' / %RTmac\n", pThis->szName, &pThis->u.s.MacAddr));
1907	return VINF_SUCCESS;
1908	}
1909
1910
1911	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1912	{
1913	bool fRegistered;
1914	Assert(!pThis->fDisconnectedFromHost);
1915
1916	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1917	vboxNetFltLinuxUnhookDev(pThis, pDev);
1918	#endif
1919
1920	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
1921	{
1922	dev_remove_pack(&pThis->u.s.PacketType);
1923	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: packet handler removed.\n", pThis));
1924	}
1925
1926	RTSpinlockAcquire(pThis->hSpinlock);
1927	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
1928	if (fRegistered)
1929	{
1930	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1931	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1932	}
1933	RTSpinlockRelease(pThis->hSpinlock);
1934
1935	if (fRegistered)
1936	{
1937	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1938	skb_queue_purge(&pThis->u.s.XmitQueue);
1939	#endif
1940	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: xmit queue purged.\n", pThis));
1941	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
1942	pDev, pDev->name,
1943	#if RTLNX_VER_MIN(2,6,37)
1944	netdev_refcnt_read(pDev)
1945	#else
1946	atomic_read(&pDev->refcnt)
1947	#endif
1948	));
1949	dev_put(pDev);
1950	}
1951
1952	return NOTIFY_OK;
1953	}
1954
1955	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1956	{
1957	/* Check if we are not suspended and promiscuous mode has not been set. */
1958	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
1959	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1960	{
1961	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1962	dev_set_promiscuity(pDev, 1);
1963	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1964	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1965	}
1966	else
1967	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1968	return NOTIFY_OK;
1969	}
1970
1971	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1972	{
1973	/* Undo promiscuous mode if we has set it. */
1974	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1975	{
1976	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1977	dev_set_promiscuity(pDev, -1);
1978	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1979	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1980	}
1981	else
1982	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1983	return NOTIFY_OK;
1984	}
1985
1986	/**
1987	* Callback for listening to MTU change event.
1988	*
1989	* We need to track changes of host's inteface MTU to discard over-sized frames
1990	* coming from the internal network as they may hang the TX queue of host's
1991	* adapter.
1992	*
1993	* @returns NOTIFY_OK
1994	* @param pThis The netfilter instance.
1995	* @param pDev Pointer to device structure of host's interface.
1996	*/
1997	static int vboxNetFltLinuxDeviceMtuChange(PVBOXNETFLTINS pThis, struct net_device *pDev)
1998	{
1999	ASMAtomicWriteU32(&pThis->u.s.cbMtu, pDev->mtu);
2000	Log(("vboxNetFltLinuxDeviceMtuChange: set MTU for %s to %d\n", pThis->szName, pDev->mtu));
2001	return NOTIFY_OK;
2002	}
2003
2004	#ifdef LOG_ENABLED
2005	/** Stringify the NETDEV_XXX constants. */
2006	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2007	{
2008	const char *pszEvent = "NETDEV_<unknown>";
2009	switch (ulEventType)
2010	{
2011	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2012	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2013	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2014	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2015	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2016	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2017	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2018	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2019	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2020	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2021	# ifdef NETDEV_FEAT_CHANGE
2022	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2023	# endif
2024	}
2025	return pszEvent;
2026	}
2027	#endif /* LOG_ENABLED */
2028
2029	/**
2030	* Callback for listening to netdevice events.
2031	*
2032	* This works the rediscovery, clean up on unregistration, promiscuity on
2033	* up/down, and GSO feature changes from ethtool.
2034	*
2035	* @returns NOTIFY_OK
2036	* @param self Pointer to our notifier registration block.
2037	* @param ulEventType The event.
2038	* @param ptr Event specific, but it is usually the device it
2039	* relates to.
2040	*/
2041	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2042
2043	{
2044	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2045	struct net_device pMyDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device );
2046	struct net_device *pDev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2047	int rc = NOTIFY_OK;
2048
2049	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2050	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, pMyDev));
2051
2052	if (ulEventType == NETDEV_REGISTER)
2053	{
2054	#if RTLNX_VER_MIN(2,6,24) /* cgroups/namespaces introduced */
2055	# if RTLNX_VER_MIN(2,6,26)
2056	# define VBOX_DEV_NET(dev) dev_net(dev)
2057	# define VBOX_NET_EQ(n1, n2) net_eq((n1), (n2))
2058	# else
2059	# define VBOX_DEV_NET(dev) ((dev)->nd_net)
2060	# define VBOX_NET_EQ(n1, n2) ((n1) == (n2))
2061	# endif
2062	struct net *pMyNet = current->nsproxy->net_ns;
2063	struct net *pDevNet = VBOX_DEV_NET(pDev);
2064
2065	if (VBOX_NET_EQ(pDevNet, pMyNet))
2066	#endif /* namespaces */
2067	{
2068	if (strcmp(pDev->name, pThis->szName) == 0)
2069	{
2070	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2071	}
2072	}
2073	}
2074	else
2075	{
2076	if (pDev == pMyDev)
2077	{
2078	switch (ulEventType)
2079	{
2080	case NETDEV_UNREGISTER:
2081	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2082	break;
2083	case NETDEV_UP:
2084	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2085	break;
2086	case NETDEV_GOING_DOWN:
2087	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2088	break;
2089	case NETDEV_CHANGEMTU:
2090	rc = vboxNetFltLinuxDeviceMtuChange(pThis, pDev);
2091	break;
2092	case NETDEV_CHANGENAME:
2093	break;
2094	#ifdef NETDEV_FEAT_CHANGE
2095	case NETDEV_FEAT_CHANGE:
2096	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2097	break;
2098	#endif
2099	}
2100	}
2101	}
2102
2103	return rc;
2104	}
2105
2106	/*
2107	* Initial enumeration of netdevs. Called with NETDEV_REGISTER by
2108	* register_netdevice_notifier() under rtnl lock.
2109	*/
2110	static int vboxNetFltLinuxEnumeratorCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2111	{
2112	PVBOXNETFLTINS pThis = ((PVBOXNETFLTNOTIFIER)self)->pThis;
2113	struct net_device *dev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2114	struct in_device *in_dev;
2115	struct inet6_dev *in6_dev;
2116
2117	if (ulEventType != NETDEV_REGISTER)
2118	return NOTIFY_OK;
2119
2120	if (RT_UNLIKELY(pThis->pSwitchPort->pfnNotifyHostAddress == NULL))
2121	return NOTIFY_OK;
2122
2123	/*
2124	* IPv4
2125	*/
2126	#if RTLNX_VER_MIN(2,6,14)
2127	in_dev = __in_dev_get_rtnl(dev);
2128	#else
2129	in_dev = __in_dev_get(dev);
2130	#endif
2131	if (in_dev != NULL)
2132	{
2133	struct in_ifaddr *ifa;
2134
2135	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2136	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2137	return NOTIFY_OK;
2138
2139	if ( dev != pThis->u.s.pDev
2140	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2141	continue;
2142
2143	Log(("%s: %s: IPv4 addr %RTnaipv4 mask %RTnaipv4\n",
2144	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2145	ifa->ifa_address, ifa->ifa_mask));
2146
2147	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2148	/* :fAdded */ true, kIntNetAddrType_IPv4, &ifa->ifa_address);
2149	}
2150	}
2151
2152	/*
2153	* IPv6
2154	*/
2155	in6_dev = __in6_dev_get(dev);
2156	if (in6_dev != NULL)
2157	{
2158	struct inet6_ifaddr *ifa;
2159
2160	read_lock_bh(&in6_dev->lock);
2161	#if RTLNX_VER_MIN(2,6,35)
2162	list_for_each_entry(ifa, &in6_dev->addr_list, if_list)
2163	#else
2164	for (ifa = in6_dev->addr_list; ifa != NULL; ifa = ifa->if_next)
2165	#endif
2166	{
2167	if ( dev != pThis->u.s.pDev
2168	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2169	continue;
2170
2171	Log(("%s: %s: IPv6 addr %RTnaipv6/%u\n",
2172	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2173	&ifa->addr, (unsigned)ifa->prefix_len));
2174
2175	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2176	/* :fAdded */ true, kIntNetAddrType_IPv6, &ifa->addr);
2177	}
2178	read_unlock_bh(&in6_dev->lock);
2179	}
2180
2181	return NOTIFY_OK;
2182	}
2183
2184
2185	static int vboxNetFltLinuxNotifierIPv4Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2186	{
2187	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv4);
2188	struct net_device pDev, pEventDev;
2189	struct in_ifaddr ifa = (struct in_ifaddr )ptr;
2190	bool fMyDev;
2191	int rc = NOTIFY_OK;
2192
2193	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2194	pEventDev = ifa->ifa_dev->dev;
2195	fMyDev = (pDev == pEventDev);
2196	Log(("VBoxNetFlt: %s: IPv4 event %s(0x%lx) %s: addr %RTnaipv4 mask %RTnaipv4\n",
2197	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2198	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2199	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2200	ifa->ifa_address, ifa->ifa_mask));
2201
2202	if (pDev != NULL)
2203	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2204
2205	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2206	return NOTIFY_OK;
2207
2208	if ( !fMyDev
2209	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2210	return NOTIFY_OK;
2211
2212	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2213	{
2214	bool fAdded;
2215	if (ulEventType == NETDEV_UP)
2216	fAdded = true;
2217	else if (ulEventType == NETDEV_DOWN)
2218	fAdded = false;
2219	else
2220	return NOTIFY_OK;
2221
2222	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2223	kIntNetAddrType_IPv4, &ifa->ifa_local);
2224	}
2225
2226	return rc;
2227	}
2228
2229
2230	static int vboxNetFltLinuxNotifierIPv6Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2231	{
2232	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv6);
2233	struct net_device pDev, pEventDev;
2234	struct inet6_ifaddr ifa = (struct inet6_ifaddr )ptr;
2235	bool fMyDev;
2236	int rc = NOTIFY_OK;
2237
2238	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2239	pEventDev = ifa->idev->dev;
2240	fMyDev = (pDev == pEventDev);
2241	Log(("VBoxNetFlt: %s: IPv6 event %s(0x%lx) %s: %RTnaipv6\n",
2242	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2243	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2244	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2245	&ifa->addr));
2246
2247	if (pDev != NULL)
2248	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2249
2250	if ( !fMyDev
2251	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2252	return NOTIFY_OK;
2253
2254	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2255	{
2256	bool fAdded;
2257	if (ulEventType == NETDEV_UP)
2258	fAdded = true;
2259	else if (ulEventType == NETDEV_DOWN)
2260	fAdded = false;
2261	else
2262	return NOTIFY_OK;
2263
2264	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2265	kIntNetAddrType_IPv6, &ifa->addr);
2266	}
2267
2268	return rc;
2269	}
2270
2271
2272	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2273	{
2274	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2275	}
2276
2277	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2278	{
2279	struct net_device * pDev;
2280	int err;
2281	int rc = VINF_SUCCESS;
2282	IPRT_LINUX_SAVE_EFL_AC();
2283	NOREF(pvIfData);
2284
2285	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2286
2287	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2288	if (pDev)
2289	{
2290	/*
2291	* Create a sk_buff for the gather list and push it onto the wire.
2292	*/
2293	if (fDst & INTNETTRUNKDIR_WIRE)
2294	{
2295	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2296	if (pBuf)
2297	{
2298	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2299	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2300	Log6(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2301	err = dev_queue_xmit(pBuf);
2302	if (err)
2303	rc = RTErrConvertFromErrno(err);
2304	}
2305	else
2306	rc = VERR_NO_MEMORY;
2307	}
2308
2309	/*
2310	* Create a sk_buff for the gather list and push it onto the host stack.
2311	*/
2312	if (fDst & INTNETTRUNKDIR_HOST)
2313	{
2314	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2315	if (pBuf)
2316	{
2317	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2318	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2319	Log6(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2320	err = netif_rx_ni(pBuf);
2321	if (err)
2322	rc = RTErrConvertFromErrno(err);
2323	}
2324	else
2325	rc = VERR_NO_MEMORY;
2326	}
2327
2328	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2329	}
2330
2331	IPRT_LINUX_RESTORE_EFL_AC();
2332	return rc;
2333	}
2334
2335
2336	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2337	{
2338	struct net_device *pDev;
2339	IPRT_LINUX_SAVE_EFL_AC();
2340
2341	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%RTbool, fDisablePromiscuous=%RTbool\n",
2342	pThis, pThis->szName, fActive, pThis->fDisablePromiscuous));
2343
2344	if (pThis->fDisablePromiscuous)
2345	return;
2346
2347	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2348	if (pDev)
2349	{
2350	/*
2351	* This api is a bit weird, the best reference is the code.
2352	*
2353	* Also, we have a bit or race conditions wrt the maintenance of
2354	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2355	*/
2356	#ifdef LOG_ENABLED
2357	u_int16_t fIf;
2358	unsigned const cPromiscBefore = pDev->promiscuity;
2359	#endif
2360	if (fActive)
2361	{
2362	Assert(!pThis->u.s.fPromiscuousSet);
2363
2364	rtnl_lock();
2365	dev_set_promiscuity(pDev, 1);
2366	rtnl_unlock();
2367	pThis->u.s.fPromiscuousSet = true;
2368	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2369	}
2370	else
2371	{
2372	if (pThis->u.s.fPromiscuousSet)
2373	{
2374	rtnl_lock();
2375	dev_set_promiscuity(pDev, -1);
2376	rtnl_unlock();
2377	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2378	}
2379	pThis->u.s.fPromiscuousSet = false;
2380
2381	#ifdef LOG_ENABLED
2382	fIf = dev_get_flags(pDev);
2383	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2384	#endif
2385	}
2386
2387	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2388	}
2389	IPRT_LINUX_RESTORE_EFL_AC();
2390	}
2391
2392
2393	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2394	{
2395	/*
2396	* Remove packet handler when we get disconnected from internal switch as
2397	* we don't want the handler to forward packets to disconnected switch.
2398	*/
2399	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
2400	{
2401	IPRT_LINUX_SAVE_EFL_AC();
2402	dev_remove_pack(&pThis->u.s.PacketType);
2403	Log(("vboxNetFltOsDisconnectIt: this=%p: Packet handler removed.\n", pThis));
2404	IPRT_LINUX_RESTORE_EFL_AC();
2405	}
2406	return VINF_SUCCESS;
2407	}
2408
2409
2410	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2411	{
2412	IPRT_LINUX_SAVE_EFL_AC();
2413
2414	/*
2415	* Report the GSO capabilities of the host and device (if connected).
2416	* Note! No need to mark ourselves busy here.
2417	*/
2418	/** @todo duplicate work here now? Attach */
2419	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2420	Log3(("vboxNetFltOsConnectIt: reporting host tso tso6\n"));
2421	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2422	0
2423	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2424	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2425	, INTNETTRUNKDIR_HOST);
2426
2427	#endif
2428	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2429
2430	IPRT_LINUX_RESTORE_EFL_AC();
2431	return VINF_SUCCESS;
2432	}
2433
2434
2435	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2436	{
2437	struct net_device *pDev;
2438	bool fRegistered;
2439	IPRT_LINUX_SAVE_EFL_AC();
2440
2441	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
2442	vboxNetFltLinuxUnhookDev(pThis, NULL);
2443	#endif
2444
2445	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2446	* unlikely, but none the less). Since it doesn't actually update the
2447	* state (just reads it), it is likely to panic in some interesting
2448	* ways. */
2449
2450	RTSpinlockAcquire(pThis->hSpinlock);
2451	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2452	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
2453	RTSpinlockRelease(pThis->hSpinlock);
2454
2455	if (fRegistered)
2456	{
2457	vboxNetFltSetLinkState(pThis, pDev, false);
2458
2459	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2460	skb_queue_purge(&pThis->u.s.XmitQueue);
2461	#endif
2462	Log(("vboxNetFltOsDeleteInstance: this=%p: xmit queue purged.\n", pThis));
2463	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2464	pDev, pDev->name,
2465	#if RTLNX_VER_MIN(2,6,37)
2466	netdev_refcnt_read(pDev)
2467	#else
2468	atomic_read(&pDev->refcnt)
2469	#endif
2470	));
2471	dev_put(pDev);
2472	}
2473
2474	unregister_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2475	unregister_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2476
2477	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2478	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2479	module_put(THIS_MODULE);
2480
2481	IPRT_LINUX_RESTORE_EFL_AC();
2482	}
2483
2484
2485	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2486	{
2487	int err;
2488	IPRT_LINUX_SAVE_EFL_AC();
2489	NOREF(pvContext);
2490
2491	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2492	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2493	if (err)
2494	{
2495	IPRT_LINUX_RESTORE_EFL_AC();
2496	return VERR_INTNET_FLT_IF_FAILED;
2497	}
2498	if (!pThis->u.s.fRegistered)
2499	{
2500	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2501	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2502	IPRT_LINUX_RESTORE_EFL_AC();
2503	return VERR_INTNET_FLT_IF_NOT_FOUND;
2504	}
2505
2506	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2507	if ( pThis->fDisconnectedFromHost
2508	\|\| !try_module_get(THIS_MODULE))
2509	{
2510	IPRT_LINUX_RESTORE_EFL_AC();
2511	return VERR_INTNET_FLT_IF_FAILED;
2512	}
2513
2514	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2515	{
2516	VBOXNETFLTNOTIFIER Enumerator;
2517
2518	/*
2519	* register_inetaddr_notifier() and register_inet6addr_notifier()
2520	* do not call the callback for existing devices. Enumerating
2521	* all network devices explicitly is a bit of an ifdef mess,
2522	* so co-opt register_netdevice_notifier() to do that for us.
2523	*/
2524	RT_ZERO(Enumerator);
2525	Enumerator.Notifier.notifier_call = vboxNetFltLinuxEnumeratorCallback;
2526	Enumerator.pThis = pThis;
2527
2528	err = register_netdevice_notifier(&Enumerator.Notifier);
2529	if (err)
2530	{
2531	LogRel(("%s: failed to enumerate network devices: error %d\n", __FUNCTION__, err));
2532	IPRT_LINUX_RESTORE_EFL_AC();
2533	return VINF_SUCCESS;
2534	}
2535
2536	unregister_netdevice_notifier(&Enumerator.Notifier);
2537
2538	pThis->u.s.NotifierIPv4.notifier_call = vboxNetFltLinuxNotifierIPv4Callback;
2539	err = register_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2540	if (err)
2541	LogRel(("%s: failed to register IPv4 notifier: error %d\n", __FUNCTION__, err));
2542
2543	pThis->u.s.NotifierIPv6.notifier_call = vboxNetFltLinuxNotifierIPv6Callback;
2544	err = register_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2545	if (err)
2546	LogRel(("%s: failed to register IPv6 notifier: error %d\n", __FUNCTION__, err));
2547	}
2548
2549	IPRT_LINUX_RESTORE_EFL_AC();
2550	return VINF_SUCCESS;
2551	}
2552
2553	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2554	{
2555	IPRT_LINUX_SAVE_EFL_AC();
2556
2557	/*
2558	* Init the linux specific members.
2559	*/
2560	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2561	pThis->u.s.fRegistered = false;
2562	pThis->u.s.fPromiscuousSet = false;
2563	pThis->u.s.fPacketHandler = false;
2564	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2565	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2566	skb_queue_head_init(&pThis->u.s.XmitQueue);
2567	# if RTLNX_VER_MIN(2,6,20)
2568	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2569	# else
2570	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2571	# endif
2572	#endif
2573
2574	IPRT_LINUX_RESTORE_EFL_AC();
2575	return VINF_SUCCESS;
2576	}
2577
2578
2579	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2580	{
2581	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2582	}
2583
2584
2585	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2586	{
2587	/* Nothing to do */
2588	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2589	return VINF_SUCCESS;
2590	}
2591
2592
2593	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2594	{
2595	/* Nothing to do */
2596	NOREF(pThis); NOREF(pvIfData);
2597	return VINF_SUCCESS;
2598	}
2599

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

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 85701

Download in other formats: