VBoxNetFlt-linux.c@ 97405

Last change on this file since 97405 was 96407, checked in by vboxsync, 2 years ago
scm copyright and license note update
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 90.3 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 96407 2022-08-22 17:43:14Z vboxsync $ */
2	/** @file
3	* VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4	*/
5
6	/*
7	* Copyright (C) 2006-2022 Oracle and/or its affiliates.
8	*
9	* This file is part of VirtualBox base platform packages, as
10	* available from https://www.virtualbox.org.
11	*
12	* This program is free software; you can redistribute it and/or
13	* modify it under the terms of the GNU General Public License
14	* as published by the Free Software Foundation, in version 3 of the
15	* License.
16	*
17	* This program is distributed in the hope that it will be useful, but
18	* WITHOUT ANY WARRANTY; without even the implied warranty of
19	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20	* General Public License for more details.
21	*
22	* You should have received a copy of the GNU General Public License
23	* along with this program; if not, see <https://www.gnu.org/licenses>.
24	*
25	* The contents of this file may alternatively be used under the terms
26	* of the Common Development and Distribution License Version 1.0
27	* (CDDL), a copy of it is provided in the "COPYING.CDDL" file included
28	* in the VirtualBox distribution, in which case the provisions of the
29	* CDDL are applicable instead of those of the GPL.
30	*
31	* You may elect to license modified versions of this file under the
32	* terms and conditions of either the GPL or the CDDL or both.
33	*
34	* SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0
35	*/
36
37
38	/*********************************************************************************************************************************
39	* Header Files *
40	*********************************************************************************************************************************/
41	#define LOG_GROUP LOG_GROUP_NET_FLT_DRV
42	#define VBOXNETFLT_LINUX_NO_XMIT_QUEUE
43	#include "the-linux-kernel.h"
44	#include "version-generated.h"
45	#include "revision-generated.h"
46	#include "product-generated.h"
47	#if RTLNX_VER_MIN(2,6,24)
48	# include <linux/nsproxy.h>
49	#endif
50	#include <linux/netdevice.h>
51	#if RTLNX_VER_MAX(2,6,29) \|\| RTLNX_VER_MIN(5,11,0)
52	# include <linux/ethtool.h>
53	#endif
54	#include <linux/etherdevice.h>
55	#include <linux/rtnetlink.h>
56	#include <linux/miscdevice.h>
57	#include <linux/inetdevice.h>
58	#include <linux/in.h>
59	#include <linux/ip.h>
60	#include <linux/if_vlan.h>
61	#if RTLNX_VER_MIN(4,5,0)
62	# include <uapi/linux/pkt_cls.h>
63	#endif
64	#include <net/ipv6.h>
65	#include <net/if_inet6.h>
66	#include <net/addrconf.h>
67
68	#include <VBox/log.h>
69	#include <VBox/err.h>
70	#include <VBox/intnetinline.h>
71	#include <VBox/vmm/pdmnetinline.h>
72	#include <VBox/param.h>
73	#include <iprt/alloca.h>
74	#include <iprt/assert.h>
75	#include <iprt/spinlock.h>
76	#include <iprt/semaphore.h>
77	#include <iprt/initterm.h>
78	#include <iprt/process.h>
79	#include <iprt/mem.h>
80	#include <iprt/net.h>
81	#include <iprt/log.h>
82	#include <iprt/mp.h>
83	#include <iprt/mem.h>
84	#include <iprt/time.h>
85
86	#define VBOXNETFLT_OS_SPECFIC 1
87	#include "../VBoxNetFltInternal.h"
88
89	typedef struct VBOXNETFLTNOTIFIER {
90	struct notifier_block Notifier;
91	PVBOXNETFLTINS pThis;
92	} VBOXNETFLTNOTIFIER;
93	typedef struct VBOXNETFLTNOTIFIER *PVBOXNETFLTNOTIFIER;
94
95
96	/*********************************************************************************************************************************
97	* Defined Constants And Macros *
98	*********************************************************************************************************************************/
99	#define VBOX_FLT_NB_TO_INST(pNB) RT_FROM_MEMBER(pNB, VBOXNETFLTINS, u.s.Notifier)
100	#define VBOX_FLT_PT_TO_INST(pPT) RT_FROM_MEMBER(pPT, VBOXNETFLTINS, u.s.PacketType)
101	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
102	# define VBOX_FLT_XT_TO_INST(pXT) RT_FROM_MEMBER(pXT, VBOXNETFLTINS, u.s.XmitTask)
103	#endif
104
105	#if RTLNX_VER_MIN(3,11,0)
106	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) netdev_notifier_info_to_dev(ptr)
107	#else
108	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) ((struct net_device *)ptr)
109	#endif
110
111	#if RTLNX_VER_MIN(3,5,0)
112	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag))
113	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr)
114	#else
115	# if RTLNX_VER_MIN(3,2,0)
116	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag), KM_SKB_DATA_SOFTIRQ)
117	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
118	# else
119	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ)
120	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
121	# endif
122	#endif
123
124	#if RTLNX_VER_MIN(2,6,34)
125	# define VBOX_NETDEV_NAME(dev) netdev_name(dev)
126	#else
127	# define VBOX_NETDEV_NAME(dev) ((dev)->reg_state != NETREG_REGISTERED ? "(unregistered net_device)" : (dev)->name)
128	#endif
129
130	#if RTLNX_VER_MIN(2,6,25)
131	# define VBOX_IPV4_IS_LOOPBACK(addr) ipv4_is_loopback(addr)
132	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ipv4_is_linklocal_169(addr)
133	#else
134	# define VBOX_IPV4_IS_LOOPBACK(addr) ((addr & htonl(IN_CLASSA_NET)) == htonl(0x7f000000))
135	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ((addr & htonl(IN_CLASSB_NET)) == htonl(0xa9fe0000))
136	#endif
137
138	#if RTLNX_VER_MIN(2,6,22)
139	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
140	# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
141	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum_offset
142	#else
143	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
144	# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
145	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum
146	#endif
147
148	#if RTLNX_VER_MIN(2,6,19)
149	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
150	#else
151	# define CHECKSUM_PARTIAL CHECKSUM_HW
152	# if RTLNX_VER_MIN(2,6,10)
153	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
154	# else
155	# if RTLNX_VER_MIN(2,6,7)
156	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
157	# else
158	# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
159	# endif
160	/* Versions prior 2.6.10 use stats for both bstats and qstats */
161	# define bstats stats
162	# define qstats stats
163	# endif
164	#endif
165
166	#if RTLNX_VER_MIN(3,20,0) \|\| RTLNX_RHEL_RANGE(7,2, 8,0) \|\| RTLNX_RHEL_RANGE(6,8, 7,0)
167	# define VBOX_HAVE_SKB_VLAN
168	#endif
169
170	#ifdef VBOX_HAVE_SKB_VLAN
171	# define vlan_tx_tag_get(skb) skb_vlan_tag_get(skb)
172	# define vlan_tx_tag_present(skb) skb_vlan_tag_present(skb)
173	#endif
174
175	#ifndef NET_IP_ALIGN
176	# define NET_IP_ALIGN 2
177	#endif
178
179	#if 1
180	/** Create scatter / gather segments for fragments. When not used, we will
181	* linearize the socket buffer before creating the internal networking SG. */
182	# define VBOXNETFLT_SG_SUPPORT 1
183	#endif
184
185	#if RTLNX_VER_MIN(2,6,18)
186
187	/** Indicates that the linux kernel may send us GSO frames. */
188	# define VBOXNETFLT_WITH_GSO 1
189
190	/** This enables or disables the transmitting of GSO frame from the internal
191	* network and to the host. */
192	# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
193
194	# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
195	/** This enables or disables the transmitting of GSO frame from the internal
196	* network and to the wire. */
197	# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
198	# endif
199
200	/** This enables or disables the forwarding/flooding of GSO frame from the host
201	* to the internal network. */
202	# define VBOXNETFLT_WITH_GSO_RECV 1
203
204	#endif /* RTLNX_VER_MIN(2,6,18) */
205
206	#if RTLNX_VER_MIN(2,6,29)
207	/** This enables or disables handling of GSO frames coming from the wire (GRO). */
208	# define VBOXNETFLT_WITH_GRO 1
209	#endif
210
211	/*
212	* GRO support was backported to RHEL 5.4
213	*/
214	#if RTLNX_RHEL_MAJ_PREREQ(5, 4)
215	# define VBOXNETFLT_WITH_GRO 1
216	#endif
217
218
219	/*********************************************************************************************************************************
220	* Internal Functions *
221	*********************************************************************************************************************************/
222	static int __init VBoxNetFltLinuxInit(void);
223	static void __exit VBoxNetFltLinuxUnload(void);
224	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf);
225
226
227	/*********************************************************************************************************************************
228	* Global Variables *
229	*********************************************************************************************************************************/
230	/**
231	* The (common) global data.
232	*/
233	static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
234
235	module_init(VBoxNetFltLinuxInit);
236	module_exit(VBoxNetFltLinuxUnload);
237
238	MODULE_AUTHOR(VBOX_VENDOR);
239	MODULE_DESCRIPTION(VBOX_PRODUCT " Network Filter Driver");
240	MODULE_LICENSE("GPL");
241	#ifdef MODULE_VERSION
242	MODULE_VERSION(VBOX_VERSION_STRING " r" RT_XSTR(VBOX_SVN_REV) " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
243	#endif
244
245
246	#if RTLNX_VER_MAX(2,6,12) && defined(LOG_ENABLED)
247	unsigned dev_get_flags(const struct net_device *dev)
248	{
249	unsigned flags;
250
251	flags = (dev->flags & ~(IFF_PROMISC \|
252	IFF_ALLMULTI \|
253	IFF_RUNNING)) \|
254	(dev->gflags & (IFF_PROMISC \|
255	IFF_ALLMULTI));
256
257	if (netif_running(dev) && netif_carrier_ok(dev))
258	flags \|= IFF_RUNNING;
259
260	return flags;
261	}
262	#endif /* RTLNX_VER_MAX(2,6,12) */
263
264
265	/**
266	* Initialize module.
267	*
268	* @returns appropriate status code.
269	*/
270	static int __init VBoxNetFltLinuxInit(void)
271	{
272	int rc;
273	/*
274	* Initialize IPRT.
275	*/
276	rc = RTR0Init(0);
277	if (RT_SUCCESS(rc))
278	{
279	Log(("VBoxNetFltLinuxInit\n"));
280
281	/*
282	* Initialize the globals and connect to the support driver.
283	*
284	* This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
285	* for establishing the connect to the support driver.
286	*/
287	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
288	rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
289	if (RT_SUCCESS(rc))
290	{
291	LogRel(("VBoxNetFlt: Successfully started.\n"));
292	return 0;
293	}
294
295	LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
296	RTR0Term();
297	}
298	else
299	LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
300
301	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
302	return -RTErrConvertToErrno(rc);
303	}
304
305
306	/**
307	* Unload the module.
308	*
309	* @todo We have to prevent this if we're busy!
310	*/
311	static void __exit VBoxNetFltLinuxUnload(void)
312	{
313	int rc;
314	Log(("VBoxNetFltLinuxUnload\n"));
315	Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
316
317	/*
318	* Undo the work done during start (in reverse order).
319	*/
320	rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
321	AssertRC(rc); NOREF(rc);
322
323	RTR0Term();
324
325	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
326
327	Log(("VBoxNetFltLinuxUnload - done\n"));
328	}
329
330
331	/**
332	* We filter traffic from the host to the internal network
333	* before it reaches the NIC driver.
334	*
335	* The current code uses a very ugly hack overriding hard_start_xmit
336	* callback in the device structure, but it has been shown to give us a
337	* performance boost of 60-100% though. Eventually we have to find some
338	* less hacky way of getting this job done.
339	*/
340	#define VBOXNETFLT_WITH_HOST2WIRE_FILTER
341
342	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
343
344	# if RTLNX_VER_MAX(2,6,29)
345
346	typedef struct ethtool_ops OVR_OPSTYPE;
347	# define OVR_OPS ethtool_ops
348	# define OVR_XMIT pfnStartXmit
349
350	# else /* RTLNX_VER_MIN(2,6,29) */
351
352	typedef struct net_device_ops OVR_OPSTYPE;
353	# define OVR_OPS netdev_ops
354	# define OVR_XMIT pOrgOps->ndo_start_xmit
355
356	# endif /* RTLNX_VER_MIN(2,6,29) */
357
358	/**
359	* The overridden net_device_ops of the device we're attached to.
360	*
361	* As there is no net_device_ops structure in pre-2.6.29 kernels we override
362	* ethtool_ops instead along with hard_start_xmit callback in net_device
363	* structure.
364	*
365	* This is a very dirty hack that was created to explore how much we can improve
366	* the host to guest transfers by not CC'ing the NIC. It turns out to be
367	* the only way to filter outgoing packets for devices without TX queue.
368	*/
369	typedef struct VBoxNetDeviceOpsOverride
370	{
371	/** Our overridden ops. */
372	OVR_OPSTYPE Ops;
373	/** Magic word. */
374	uint32_t u32Magic;
375	/** Pointer to the original ops. */
376	OVR_OPSTYPE const *pOrgOps;
377	# if RTLNX_VER_MAX(2,6,29)
378	/** Pointer to the original hard_start_xmit function. */
379	int (pfnStartXmit)(struct sk_buff pSkb, struct net_device *pDev);
380	# endif /* RTLNX_VER_MAX(2,6,29) */
381	/** Pointer to the net filter instance. */
382	PVBOXNETFLTINS pVBoxNetFlt;
383	/** The number of filtered packages. */
384	uint64_t cFiltered;
385	/** The total number of packets */
386	uint64_t cTotal;
387	} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
388	/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
389	#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
390
391	/**
392	* ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
393	* because they belong on the internal network.
394	*
395	* @returns NETDEV_TX_XXX.
396	* @param pSkb The socket buffer to transmit.
397	* @param pDev The net device.
398	*/
399	static int vboxNetFltLinuxStartXmitFilter(struct sk_buff pSkb, struct net_device pDev)
400	{
401	PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
402	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
403	PCRTNETETHERHDR pEtherHdr;
404	PINTNETTRUNKSWPORT pSwitchPort;
405	uint32_t cbHdrs;
406
407
408	/*
409	* Validate the override structure.
410	*
411	* Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
412	* to be production quality code, we would have to be much more
413	* careful here and avoid the race.
414	*/
415	if ( !RT_VALID_PTR(pOverride)
416	\|\| pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
417	# if RTLNX_VER_MIN(2,6,29)
418	\|\| !RT_VALID_PTR(pOverride->pOrgOps)
419	# endif
420	)
421	{
422	printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
423	dev_kfree_skb(pSkb);
424	return NETDEV_TX_OK;
425	}
426	pOverride->cTotal++;
427
428	/*
429	* Do the filtering base on the default OUI of our virtual NICs
430	*
431	* Note! In a real solution, we would ask the switch whether the
432	* destination MAC is 100% to be on the internal network and then
433	* drop it.
434	*/
435	cbHdrs = skb_headlen(pSkb);
436	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
437	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
438	if ( pEtherHdr
439	&& RT_VALID_PTR(pOverride->pVBoxNetFlt)
440	&& (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
441	&& RT_VALID_PTR(pSwitchPort)
442	&& cbHdrs >= 6)
443	{
444	INTNETSWDECISION enmDecision;
445
446	/** @todo consider reference counting, etc. */
447	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
448	if (enmDecision == INTNETSWDECISION_INTNET)
449	{
450	dev_kfree_skb(pSkb);
451	pOverride->cFiltered++;
452	return NETDEV_TX_OK;
453	}
454	}
455
456	return pOverride->OVR_XMIT(pSkb, pDev);
457	}
458
459	/**
460	* Hooks the device ndo_start_xmit operation of the device.
461	*
462	* @param pThis The net filter instance.
463	* @param pDev The net device.
464	*/
465	static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
466	{
467	PVBOXNETDEVICEOPSOVERRIDE pOverride;
468
469	/* Cancel override if ethtool_ops is missing (host-only case, @bugref{5712}) */
470	if (!RT_VALID_PTR(pDev->OVR_OPS))
471	return;
472	pOverride = RTMemAlloc(sizeof(*pOverride));
473	if (!pOverride)
474	return;
475	pOverride->pOrgOps = pDev->OVR_OPS;
476	pOverride->Ops = *pDev->OVR_OPS;
477	# if RTLNX_VER_MAX(2,6,29)
478	pOverride->pfnStartXmit = pDev->hard_start_xmit;
479	# else /* RTLNX_VER_MIN(2,6,29) */
480	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
481	# endif /* RTLNX_VER_MIN(2,6,29) */
482	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
483	pOverride->cTotal = 0;
484	pOverride->cFiltered = 0;
485	pOverride->pVBoxNetFlt = pThis;
486
487	RTSpinlockAcquire(pThis->hSpinlock); /* (this isn't necessary, but so what) */
488	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
489	# if RTLNX_VER_MAX(2,6,29)
490	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
491	# endif /* RTLNX_VER_MAX(2,6,29) */
492	RTSpinlockRelease(pThis->hSpinlock);
493	}
494
495	/**
496	* Undos what vboxNetFltLinuxHookDev did.
497	*
498	* @param pThis The net filter instance.
499	* @param pDev The net device. Can be NULL, in which case
500	* we'll try retrieve it from @a pThis.
501	*/
502	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
503	{
504	PVBOXNETDEVICEOPSOVERRIDE pOverride;
505
506	RTSpinlockAcquire(pThis->hSpinlock);
507	if (!pDev)
508	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
509	if (RT_VALID_PTR(pDev))
510	{
511	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
512	if ( RT_VALID_PTR(pOverride)
513	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
514	&& RT_VALID_PTR(pOverride->pOrgOps)
515	)
516	{
517	# if RTLNX_VER_MAX(2,6,29)
518	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
519	# endif /* RTLNX_VER_MAX(2,6,29) */
520	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
521	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
522	}
523	else
524	pOverride = NULL;
525	}
526	else
527	pOverride = NULL;
528	RTSpinlockRelease(pThis->hSpinlock);
529
530	if (pOverride)
531	{
532	printk("vboxnetflt: %llu out of %llu packets were not sent (directed to host)\n", pOverride->cFiltered, pOverride->cTotal);
533	RTMemFree(pOverride);
534	}
535	}
536
537	#endif /* VBOXNETFLT_WITH_HOST2WIRE_FILTER */
538
539
540	/**
541	* Reads and retains the host interface handle.
542	*
543	* @returns The handle, NULL if detached.
544	* @param pThis
545	*/
546	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
547	{
548	#if 0
549	struct net_device *pDev = NULL;
550
551	Log(("vboxNetFltLinuxRetainNetDev\n"));
552	/*
553	* Be careful here to avoid problems racing the detached callback.
554	*/
555	RTSpinlockAcquire(pThis->hSpinlock);
556	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
557	{
558	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
559	if (pDev)
560	{
561	dev_hold(pDev);
562	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
563	pDev, pDev->name,
564	#if RTLNX_VER_MIN(2,6,37)
565	netdev_refcnt_read(pDev)
566	#else
567	atomic_read(&pDev->refcnt)
568	#endif
569	));
570	}
571	}
572	RTSpinlockRelease(pThis->hSpinlock);
573
574	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
575	return pDev;
576	#else
577	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
578	#endif
579	}
580
581
582	/**
583	* Release the host interface handle previously retained
584	* by vboxNetFltLinuxRetainNetDev.
585	*
586	* @param pThis The instance.
587	* @param pDev The vboxNetFltLinuxRetainNetDev
588	* return value, NULL is fine.
589	*/
590	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
591	{
592	#if 0
593	Log(("vboxNetFltLinuxReleaseNetDev\n"));
594	NOREF(pThis);
595	if (pDev)
596	{
597	dev_put(pDev);
598	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
599	pDev, pDev->name,
600	#if RTLNX_VER_MIN(2,6,37)
601	netdev_refcnt_read(pDev)
602	#else
603	atomic_read(&pDev->refcnt)
604	#endif
605	));
606	}
607	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
608	#endif
609	}
610
611	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
612	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
613
614	/**
615	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
616	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
617	*
618	* @returns true / false accordingly.
619	* @param pBuf The sk_buff.
620	*/
621	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
622	{
623	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
624	}
625
626
627	/**
628	* Checks whether this SG list contains a GSO packet.
629	*
630	* @returns true / false accordingly.
631	* @param pSG The (scatter/)gather list.
632	*/
633	DECLINLINE(bool) vboxNetFltLinuxIsGso(PINTNETSG pSG)
634	{
635	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
636	return !((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type == PDMNETWORKGSOTYPE_INVALID);
637	#else /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
638	return false;
639	#endif /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
640	}
641
642
643	/**
644	* Find out the frame size (of a single segment in case of GSO frames).
645	*
646	* @returns the frame size.
647	* @param pSG The (scatter/)gather list.
648	*/
649	DECLINLINE(uint32_t) vboxNetFltLinuxFrameSize(PINTNETSG pSG)
650	{
651	uint16_t u16Type = 0;
652	uint32_t cbVlanTag = 0;
653	if (pSG->aSegs[0].cb >= sizeof(RTNETETHERHDR))
654	u16Type = RT_BE2H_U16(((PCRTNETETHERHDR)pSG->aSegs[0].pv)->EtherType);
655	else if (pSG->cbTotal >= sizeof(RTNETETHERHDR))
656	{
657	uint32_t off = RT_UOFFSETOF(RTNETETHERHDR, EtherType);
658	uint32_t i;
659	for (i = 0; i < pSG->cSegsUsed; ++i)
660	{
661	if (off <= pSG->aSegs[i].cb)
662	{
663	if (off + sizeof(uint16_t) <= pSG->aSegs[i].cb)
664	u16Type = RT_BE2H_U16((uint16_t )((uintptr_t)pSG->aSegs[i].pv + off));
665	else if (i + 1 < pSG->cSegsUsed)
666	u16Type = RT_BE2H_U16( ((uint16_t)( ((uint8_t *)pSG->aSegs[i].pv)[off] ) << 8)
667	+ (uint8_t )pSG->aSegs[i + 1].pv); /* ASSUMES no empty segments! */
668	/* else: frame is too short. */
669	break;
670	}
671	off -= pSG->aSegs[i].cb;
672	}
673	}
674	if (u16Type == RTNET_ETHERTYPE_VLAN)
675	cbVlanTag = 4;
676	return (vboxNetFltLinuxIsGso(pSG) ? (uint32_t)pSG->GsoCtx.cbMaxSeg + pSG->GsoCtx.cbHdrsTotal : pSG->cbTotal) - cbVlanTag;
677	}
678
679
680	/**
681	* Internal worker that create a linux sk_buff for a
682	* (scatter/)gather list.
683	*
684	* @returns Pointer to the sk_buff.
685	* @param pThis The instance.
686	* @param pSG The (scatter/)gather list.
687	* @param fDstWire Set if the destination is the wire.
688	*/
689	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
690	{
691	struct sk_buff *pPkt;
692	struct net_device *pDev;
693	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
694	unsigned fGsoType = 0;
695	#endif
696
697	if (pSG->cbTotal == 0)
698	{
699	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
700	return NULL;
701	}
702	Log5(("VBoxNetFlt: Packet to %s of %d bytes (frame=%d).\n", fDstWire?"wire":"host", pSG->cbTotal, vboxNetFltLinuxFrameSize(pSG)));
703	if (fDstWire && (vboxNetFltLinuxFrameSize(pSG) > ASMAtomicReadU32(&pThis->u.s.cbMtu) + 14))
704	{
705	static bool s_fOnce = true;
706	if (s_fOnce)
707	{
708	s_fOnce = false;
709	printk("VBoxNetFlt: Dropped over-sized packet (%d bytes) coming from internal network.\n", vboxNetFltLinuxFrameSize(pSG));
710	}
711	return NULL;
712	}
713
714	/** @todo We should use fragments mapping the SG buffers with large packets.
715	* 256 bytes seems to be the a threshold used a lot for this. It
716	* requires some nasty work on the intnet side though... */
717	/*
718	* Allocate a packet and copy over the data.
719	*/
720	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
721	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
722	if (RT_UNLIKELY(!pPkt))
723	{
724	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
725	pSG->pvUserData = NULL;
726	return NULL;
727	}
728	pPkt->dev = pDev;
729	pPkt->ip_summed = CHECKSUM_NONE;
730
731	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
732	skb_reserve(pPkt, NET_IP_ALIGN);
733
734	/* Copy the segments. */
735	skb_put(pPkt, pSG->cbTotal);
736	IntNetSgRead(pSG, pPkt->data);
737
738	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
739	/*
740	* Setup GSO if used by this packet.
741	*/
742	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
743	{
744	default:
745	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
746	RT_FALL_THRU();
747	case PDMNETWORKGSOTYPE_INVALID:
748	fGsoType = 0;
749	break;
750	case PDMNETWORKGSOTYPE_IPV4_TCP:
751	fGsoType = SKB_GSO_TCPV4;
752	break;
753	case PDMNETWORKGSOTYPE_IPV6_TCP:
754	fGsoType = SKB_GSO_TCPV6;
755	break;
756	}
757	if (fGsoType)
758	{
759	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
760
761	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
762	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
763	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
764
765	/*
766	* We need to set checksum fields even if the packet goes to the host
767	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
768	*/
769	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrsTotal);
770	pPkt->ip_summed = CHECKSUM_PARTIAL;
771	# if RTLNX_VER_MIN(2,6,22)
772	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
773	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
774	pPkt->csum_offset = RT_UOFFSETOF(RTNETTCP, th_sum);
775	else
776	pPkt->csum_offset = RT_UOFFSETOF(RTNETUDP, uh_sum);
777	# else
778	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
779	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
780	pPkt->csum = RT_UOFFSETOF(RTNETTCP, th_sum);
781	else
782	pPkt->csum = RT_UOFFSETOF(RTNETUDP, uh_sum);
783	# endif
784	if (!fDstWire)
785	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
786	}
787	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
788
789	/*
790	* Finish up the socket buffer.
791	*/
792	pPkt->protocol = eth_type_trans(pPkt, pDev);
793	if (fDstWire)
794	{
795	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
796
797	/* Restore ethernet header back. */
798	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
799	VBOX_SKB_RESET_MAC_HDR(pPkt);
800	}
801	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
802
803	return pPkt;
804	}
805
806
807	/**
808	* Return the offset where to start checksum computation from.
809	*
810	* @returns the offset relative to pBuf->data.
811	* @param pBuf The socket buffer.
812	*/
813	DECLINLINE(unsigned) vboxNetFltLinuxGetChecksumStartOffset(struct sk_buff *pBuf)
814	{
815	#if RTLNX_VER_MIN(2,6,38)
816	return skb_checksum_start_offset(pBuf);
817	#elif RTLNX_VER_MIN(2,6,22)
818	return pBuf->csum_start - skb_headroom(pBuf);
819	#else
820	unsigned char *pTransportHdr = pBuf->h.raw;
821	# if RTLNX_VER_MAX(2,6,19)
822	/*
823	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
824	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
825	* header length from the header itself and reconstruct 'h' pointer
826	* to TCP (or whatever) header.
827	*/
828	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
829	pTransportHdr = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
830	# endif
831	return pTransportHdr - pBuf->data;
832	#endif
833	}
834
835
836	/**
837	* Initializes a SG list from an sk_buff.
838	*
839	* @returns Number of segments.
840	* @param pThis The instance.
841	* @param pBuf The sk_buff.
842	* @param pSG The SG.
843	* @param cbExtra The number of bytes of extra space allocated immediately after the SG.
844	* @param cSegs The number of segments allocated for the SG.
845	* This should match the number in the mbuf exactly!
846	* @param fSrc The source of the frame.
847	* @param pGsoCtx Pointer to the GSO context if it's a GSO
848	* internal network frame. NULL if regular frame.
849	*/
850	static void vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
851	unsigned cbExtra, unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
852	{
853	int i;
854	NOREF(pThis);
855
856	#ifndef VBOXNETFLT_SG_SUPPORT
857	Assert(!skb_shinfo(pBuf)->frag_list);
858	#else /* VBOXNETFLT_SG_SUPPORT */
859	uint8_t pExtra = (uint8_t )&pSG->aSegs[cSegs];
860	unsigned cbConsumed = 0;
861	unsigned cbProduced = 0;
862
863	# if RTLNX_VER_MIN(2,6,27)
864	/* Restore VLAN tag stripped by host hardware */
865	if (vlan_tx_tag_present(pBuf))
866	{
867	uint8_t *pMac = pBuf->data;
868	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )pExtra;
869	Assert(ETH_ALEN * 2 + VLAN_HLEN <= cbExtra);
870	memmove(pVHdr, pMac, ETH_ALEN * 2);
871	/* Consume whole Ethernet header: 2 addresses + EtherType (see @bugref{8599}) */
872	cbConsumed += ETH_ALEN * 2 + sizeof(uint16_t);
873	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
874	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
875	pVHdr->h_vlan_encapsulated_proto = (uint16_t)(pMac + ETH_ALEN * 2);
876	cbProduced += VLAN_ETH_HLEN;
877	}
878	# endif /* RTLNX_VER_MIN(2,6,27) */
879
880	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
881	{
882	unsigned uCsumStartOffset = vboxNetFltLinuxGetChecksumStartOffset(pBuf);
883	unsigned uCsumStoreOffset = uCsumStartOffset + VBOX_SKB_CSUM_OFFSET(pBuf) - cbConsumed;
884	Log3(("cbConsumed=%u cbProduced=%u uCsumStartOffset=%u uCsumStoreOffset=%u\n",
885	cbConsumed, cbProduced, uCsumStartOffset, uCsumStoreOffset));
886	Assert(cbProduced + uCsumStoreOffset + sizeof(uint16_t) <= cbExtra);
887	/*
888	* We assume that the checksum is stored at the very end of the transport header
889	* so we will have all headers in a single fragment. If our assumption is wrong
890	* we may see suboptimal performance.
891	*/
892	memmove(pExtra + cbProduced,
893	pBuf->data + cbConsumed,
894	uCsumStoreOffset);
895	unsigned uChecksum = skb_checksum(pBuf, uCsumStartOffset, pBuf->len - uCsumStartOffset, 0);
896	(uint16_t)(pExtra + cbProduced + uCsumStoreOffset) = csum_fold(uChecksum);
897	cbProduced += uCsumStoreOffset + sizeof(uint16_t);
898	cbConsumed += uCsumStoreOffset + sizeof(uint16_t);
899	}
900	#endif /* VBOXNETFLT_SG_SUPPORT */
901
902	if (!pGsoCtx)
903	IntNetSgInitTempSegs(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/);
904	else
905	IntNetSgInitTempSegsGso(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/, pGsoCtx);
906
907	int iSeg = 0;
908	#ifdef VBOXNETFLT_SG_SUPPORT
909	if (cbProduced)
910	{
911	pSG->aSegs[iSeg].cb = cbProduced;
912	pSG->aSegs[iSeg].pv = pExtra;
913	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
914	}
915	pSG->aSegs[iSeg].cb = skb_headlen(pBuf) - cbConsumed;
916	pSG->aSegs[iSeg].pv = pBuf->data + cbConsumed;
917	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
918	Assert(iSeg <= pSG->cSegsAlloc);
919
920	# ifdef LOG_ENABLED
921	if (pBuf->data_len)
922	Log6((" kmap_atomic:"));
923	# endif /* LOG_ENABLED */
924	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
925	{
926	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
927	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
928	pSG->aSegs[iSeg].cb = pFrag->bv_len;
929	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
930	# else /* < KERNEL_VERSION(5, 4, 0) */
931	pSG->aSegs[iSeg].cb = pFrag->size;
932	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
933	# endif /* >= KERNEL_VERSION(5, 4, 0) */
934	Log6((" %p", pSG->aSegs[iSeg].pv));
935	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
936	Assert(iSeg <= pSG->cSegsAlloc);
937	}
938	struct sk_buff *pFragBuf;
939	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
940	{
941	pSG->aSegs[iSeg].cb = skb_headlen(pFragBuf);
942	pSG->aSegs[iSeg].pv = pFragBuf->data;
943	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
944	Assert(iSeg <= pSG->cSegsAlloc);
945	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
946	{
947	skb_frag_t *pFrag = &skb_shinfo(pFragBuf)->frags[i];
948	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
949	pSG->aSegs[iSeg].cb = pFrag->bv_len;
950	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
951	# else /* < KERNEL_VERSION(5, 4, 0) */
952	pSG->aSegs[iSeg].cb = pFrag->size;
953	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
954	# endif /* >= KERNEL_VERSION(5, 4, 0) */
955	Log6((" %p", pSG->aSegs[iSeg].pv));
956	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
957	Assert(iSeg <= pSG->cSegsAlloc);
958	}
959	}
960	# ifdef LOG_ENABLED
961	if (pBuf->data_len)
962	Log6(("\n"));
963	# endif /* LOG_ENABLED */
964	#else
965	pSG->aSegs[iSeg].cb = pBuf->len;
966	pSG->aSegs[iSeg].pv = pBuf->data;
967	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
968	#endif
969
970	pSG->cSegsUsed = iSeg;
971
972	#if 0
973	if (cbProduced)
974	{
975	LogRel(("vboxNetFltLinuxSkBufToSG: original packet dump:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
976	LogRel(("vboxNetFltLinuxSkBufToSG: cbConsumed=%u cbProduced=%u cbExtra=%u\n", cbConsumed, cbProduced, cbExtra));
977	uint32_t offset = 0;
978	for (i = 0; i < pSG->cSegsUsed; ++i)
979	{
980	LogRel(("vboxNetFltLinuxSkBufToSG: seg#%d (%d bytes, starting at 0x%x):\n%.*Rhxd\n",
981	i, pSG->aSegs[i].cb, offset, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
982	offset += pSG->aSegs[i].cb;
983	}
984	}
985	#endif
986
987	#ifdef PADD_RUNT_FRAMES_FROM_HOST
988	/*
989	* Add a trailer if the frame is too small.
990	*
991	* Since we're getting to the packet before it is framed, it has not
992	* yet been padded. The current solution is to add a segment pointing
993	* to a buffer containing all zeros and pray that works for all frames...
994	*/
995	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
996	{
997	Assert(pBuf->data_len == 0); /* Packets with fragments are never small! */
998	static uint8_t const s_abZero[128] = {0};
999
1000	AssertReturnVoid(iSeg < cSegs);
1001
1002	pSG->aSegs[iSeg].Phys = NIL_RTHCPHYS;
1003	pSG->aSegs[iSeg].pv = (void *)&s_abZero[0];
1004	pSG->aSegs[iSeg++].cb = 60 - pSG->cbTotal;
1005	pSG->cbTotal = 60;
1006	pSG->cSegsUsed++;
1007	Assert(iSeg <= pSG->cSegsAlloc)
1008	}
1009	#endif
1010
1011	Log6(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1012	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1013	for (i = 0; i < pSG->cSegsUsed; i++)
1014	Log6(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1015	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1016	}
1017
1018	/**
1019	* Packet handler; not really documented - figure it out yourself.
1020	*
1021	* @returns 0 or EJUSTRETURN - this is probably copy & pastry and thus wrong.
1022	*/
1023	#if RTLNX_VER_MIN(2,6,14)
1024	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1025	struct net_device *pSkbDev,
1026	struct packet_type *pPacketType,
1027	struct net_device *pOrigDev)
1028	#else
1029	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1030	struct net_device *pSkbDev,
1031	struct packet_type *pPacketType)
1032	#endif
1033	{
1034	PVBOXNETFLTINS pThis;
1035	struct net_device *pDev;
1036	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1037	pBuf, pSkbDev, pPacketType));
1038	#if RTLNX_VER_MIN(2,6,18)
1039	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",
1040	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));
1041	# if RTLNX_VER_MIN(2,6,22)
1042	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1043	# endif
1044	#else
1045	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",
1046	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));
1047	#endif
1048	/*
1049	* Drop it immediately?
1050	*/
1051	if (!pBuf)
1052	return 0;
1053
1054	if (pBuf->pkt_type == PACKET_LOOPBACK)
1055	{
1056	/*
1057	* We are not interested in loopbacked packets as they will always have
1058	* another copy going to the wire.
1059	*/
1060	Log2(("vboxNetFltLinuxPacketHandler: dropped loopback packet (cb=%u)\n", pBuf->len));
1061	dev_kfree_skb(pBuf); /* We must 'consume' all packets we get (@bugref{6539})! */
1062	return 0;
1063	}
1064
1065	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1066	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1067	if (pDev != pSkbDev)
1068	{
1069	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1070	kfree_skb(pBuf); /* This is a failure, so we use kfree_skb instead of dev_kfree_skb. */
1071	return 0;
1072	}
1073
1074	Log6(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1075	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1076	{
1077	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1078	dev_kfree_skb(pBuf);
1079	return 0;
1080	}
1081
1082	#ifndef VBOXNETFLT_SG_SUPPORT
1083	{
1084	/*
1085	* Get rid of fragmented packets, they cause too much trouble.
1086	*/
1087	unsigned int uMacLen = pBuf->mac_len;
1088	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1089	dev_kfree_skb(pBuf);
1090	if (!pCopy)
1091	{
1092	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1093	return 0;
1094	}
1095	pBuf = pCopy;
1096	/* Somehow skb_copy ignores mac_len */
1097	pBuf->mac_len = uMacLen;
1098	# if RTLNX_VER_MIN(2,6,27)
1099	/* Restore VLAN tag stripped by host hardware */
1100	if (vlan_tx_tag_present(pBuf) && skb_headroom(pBuf) >= VLAN_ETH_HLEN)
1101	{
1102	uint8_t pMac = (uint8_t)skb_mac_header(pBuf);
1103	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )(pMac - VLAN_HLEN);
1104	memmove(pVHdr, pMac, ETH_ALEN * 2);
1105	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
1106	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
1107	pBuf->mac_header -= VLAN_HLEN;
1108	pBuf->mac_len += VLAN_HLEN;
1109	}
1110	# endif /* RTLNX_VER_MIN(2,6,27) */
1111
1112	# if RTLNX_VER_MIN(2,6,18)
1113	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",
1114	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));
1115	# if RTLNX_VER_MIN(2,6,22)
1116	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1117	# endif /* RTLNX_VER_MIN(2,6,22) */
1118	# else /* RTLNX_VER_MAX(2,6,18) */
1119	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",
1120	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));
1121	# endif /* RTLNX_VER_MAX(2,6,18) */
1122	}
1123	#endif /* !VBOXNETFLT_SG_SUPPORT */
1124
1125	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1126	/* Forward it to the internal network. */
1127	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1128	#else /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1129	/* Add the packet to transmit queue and schedule the bottom half. */
1130	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1131	schedule_work(&pThis->u.s.XmitTask);
1132	Log6(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1133	&pThis->u.s.XmitTask, pBuf));
1134	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1135
1136	/* It does not really matter what we return, it is ignored by the kernel. */
1137	return 0;
1138	}
1139
1140	/**
1141	* Calculate the number of INTNETSEG segments the socket buffer will need.
1142	*
1143	* @returns Segment count.
1144	* @param pBuf The socket buffer.
1145	* @param pcbTemp Where to store the number of bytes of the part
1146	* of the socket buffer that will be copied to
1147	* a temporary storage.
1148	*/
1149	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff pBuf, unsigned pcbTemp)
1150	{
1151	*pcbTemp = 0;
1152	#ifdef VBOXNETFLT_SG_SUPPORT
1153	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1154	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1155	{
1156	*pcbTemp = vboxNetFltLinuxGetChecksumStartOffset(pBuf) + VBOX_SKB_CSUM_OFFSET(pBuf) + sizeof(uint16_t);
1157	}
1158	# if RTLNX_VER_MIN(2,6,27)
1159	if (vlan_tx_tag_present(pBuf))
1160	{
1161	if (*pcbTemp)
1162	*pcbTemp += VLAN_HLEN;
1163	else
1164	*pcbTemp = VLAN_ETH_HLEN;
1165	}
1166	# endif /* RTLNX_VER_MIN(2,6,27) */
1167	if (*pcbTemp)
1168	++cSegs;
1169	struct sk_buff *pFrag;
1170	for (pFrag = skb_shinfo(pBuf)->frag_list; pFrag; pFrag = pFrag->next)
1171	{
1172	Log6(("vboxNetFltLinuxCalcSGSegments: frag=%p len=%d data_len=%d frags=%d frag_list=%p next=%p\n",
1173	pFrag, pFrag->len, pFrag->data_len, skb_shinfo(pFrag)->nr_frags, skb_shinfo(pFrag)->frag_list, pFrag->next));
1174	cSegs += 1 + skb_shinfo(pFrag)->nr_frags;
1175	}
1176	#else
1177	unsigned cSegs = 1;
1178	#endif
1179	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1180	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1181	if (pBuf->len < 60)
1182	cSegs++;
1183	#endif
1184	return cSegs;
1185	}
1186
1187
1188	/**
1189	* Destroy the intnet scatter / gather buffer created by
1190	* vboxNetFltLinuxSkBufToSG.
1191	*
1192	* @param pSG The (scatter/)gather list.
1193	* @param pBuf The original socket buffer that was used to create
1194	* the scatter/gather list.
1195	*/
1196	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG, struct sk_buff *pBuf)
1197	{
1198	#ifdef VBOXNETFLT_SG_SUPPORT
1199	int i, iSeg = 1; /* Skip non-paged part of SKB */
1200	/* Check if the extra buffer behind SG structure was used for modified packet header */
1201	if (pBuf->data != pSG->aSegs[0].pv)
1202	++iSeg; /* Skip it as well */
1203	# ifdef LOG_ENABLED
1204	if (pBuf->data_len)
1205	Log6(("kunmap_atomic:"));
1206	# endif /* LOG_ENABLED */
1207	/* iSeg now points to the first mapped fragment if there are any */
1208	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1209	{
1210	Log6((" %p", pSG->aSegs[iSeg].pv));
1211	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1212	}
1213	struct sk_buff *pFragBuf;
1214	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
1215	{
1216	++iSeg; /* Non-fragment (unmapped) portion of chained SKB */
1217	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
1218	{
1219	Log6((" %p", pSG->aSegs[iSeg].pv));
1220	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1221	}
1222	}
1223	# ifdef LOG_ENABLED
1224	if (pBuf->data_len)
1225	Log6(("\n"));
1226	# endif /* LOG_ENABLED */
1227	#endif
1228	NOREF(pSG);
1229	}
1230
1231	#ifdef LOG_ENABLED
1232	/**
1233	* Logging helper.
1234	*/
1235	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1236	{
1237	int i, offSeg;
1238	uint8_t pInt, pExt;
1239	static int iPacketNo = 1;
1240	iPacketNo += iIncrement;
1241	if (fEgress)
1242	{
1243	pExt = pSG->aSegs[0].pv;
1244	pInt = pExt + 6;
1245	}
1246	else
1247	{
1248	pInt = pSG->aSegs[0].pv;
1249	pExt = pInt + 6;
1250	}
1251	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1252	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1253	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1254	fEgress ? "-->" : "<--", pszWhere,
1255	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1256	pSG->cbTotal, iPacketNo));
1257	if (pSG->cSegsUsed == 1)
1258	{
1259	Log4(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1260	}
1261	else
1262	{
1263	for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1264	{
1265	Log4(("-- segment %d at 0x%x (%d bytes)\n --\n%.*Rhxd\n",
1266	i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1267	offSeg += pSG->aSegs[i].cb;
1268	}
1269	}
1270	}
1271	#else
1272	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1273	#endif
1274
1275	#ifdef VBOXNETFLT_WITH_GSO_RECV
1276
1277	/**
1278	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1279	* GSO socket buffer without having to segment it.
1280	*
1281	* @returns true on success, false if needs segmenting.
1282	* @param pThis The net filter instance.
1283	* @param pSkb The GSO socket buffer.
1284	* @param fSrc The source.
1285	* @param pGsoCtx Where to return the GSO context on success.
1286	*/
1287	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1288	PPDMNETWORKGSO pGsoCtx)
1289	{
1290	PDMNETWORKGSOTYPE enmGsoType;
1291	uint16_t uEtherType;
1292	unsigned int cbTransport;
1293	unsigned int offTransport;
1294	unsigned int cbTransportHdr;
1295	unsigned uProtocol;
1296	union
1297	{
1298	RTNETIPV4 IPv4;
1299	RTNETIPV6 IPv6;
1300	RTNETTCP Tcp;
1301	uint8_t ab[40];
1302	uint16_t au16[40/2];
1303	uint32_t au32[40/4];
1304	} Buf;
1305
1306	/*
1307	* Check the GSO properties of the socket buffer and make sure it fits.
1308	*/
1309	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1310	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1311	{
1312	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1313	return false;
1314	}
1315	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1316	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1317	{
1318	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1319	return false;
1320	}
1321
1322	/*
1323	* Switch on the ethertype.
1324	*/
1325	uEtherType = pSkb->protocol;
1326	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1327	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1328	{
1329	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1330	if (puEtherType)
1331	uEtherType = *puEtherType;
1332	}
1333	switch (uEtherType)
1334	{
1335	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1336	{
1337	unsigned int cbHdr;
1338	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1339	if (RT_UNLIKELY(!pIPv4))
1340	{
1341	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1342	return false;
1343	}
1344
1345	cbHdr = pIPv4->ip_hl * 4;
1346	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1347	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1348	\|\| cbHdr > cbTransport ))
1349	{
1350	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1351	return false;
1352	}
1353	cbTransport -= cbHdr;
1354	offTransport = pSkb->mac_len + cbHdr;
1355	uProtocol = pIPv4->ip_p;
1356	if (uProtocol == RTNETIPV4_PROT_TCP)
1357	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1358	else if (uProtocol == RTNETIPV4_PROT_UDP)
1359	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1360	else /** @todo IPv6: 4to6 tunneling */
1361	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1362	break;
1363	}
1364
1365	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1366	{
1367	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1368	if (RT_UNLIKELY(!pIPv6))
1369	{
1370	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1371	return false;
1372	}
1373
1374	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1375	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1376	uProtocol = pIPv6->ip6_nxt;
1377	/** @todo IPv6: Dig our way out of the other headers. */
1378	if (uProtocol == RTNETIPV4_PROT_TCP)
1379	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1380	else if (uProtocol == RTNETIPV4_PROT_UDP)
1381	enmGsoType = PDMNETWORKGSOTYPE_IPV6_UDP;
1382	else
1383	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1384	break;
1385	}
1386
1387	default:
1388	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1389	return false;
1390	}
1391
1392	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1393	{
1394	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1395	return false;
1396	}
1397
1398	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1399	\|\| offTransport + cbTransport > pSkb->len
1400	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1401	{
1402	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1403	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1404	return false;
1405	}
1406
1407	/*
1408	* Check the TCP/UDP bits.
1409	*/
1410	if (uProtocol == RTNETIPV4_PROT_TCP)
1411	{
1412	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1413	if (RT_UNLIKELY(!pTcp))
1414	{
1415	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1416	return false;
1417	}
1418
1419	cbTransportHdr = pTcp->th_off * 4;
1420	pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1421	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1422	\|\| cbTransportHdr > cbTransport
1423	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1424	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1425	{
1426	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1427	return false;
1428	}
1429
1430	}
1431	else
1432	{
1433	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1434	cbTransportHdr = sizeof(RTNETUDP);
1435	pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1436	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1437	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1438	{
1439	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1440	return false;
1441	}
1442	}
1443
1444	/*
1445	* We're good, init the GSO context.
1446	*/
1447	pGsoCtx->u8Type = enmGsoType;
1448	pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1449	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1450	pGsoCtx->offHdr1 = pSkb->mac_len;
1451	pGsoCtx->offHdr2 = offTransport;
1452	pGsoCtx->u8Unused = 0;
1453
1454	return true;
1455	}
1456
1457	/**
1458	* Forward the socket buffer as a GSO internal network frame.
1459	*
1460	* @returns IPRT status code.
1461	* @param pThis The net filter instance.
1462	* @param pSkb The GSO socket buffer.
1463	* @param fSrc The source.
1464	* @param pGsoCtx Where to return the GSO context on success.
1465	*/
1466	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1467	{
1468	int rc;
1469	unsigned cbExtra;
1470	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb, &cbExtra);
1471	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1472	if (RT_LIKELY(pSG))
1473	{
1474	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cbExtra, cSegs, fSrc, pGsoCtx);
1475
1476	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1477	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1478
1479	vboxNetFltLinuxDestroySG(pSG, pSkb);
1480	rc = VINF_SUCCESS;
1481	}
1482	else
1483	{
1484	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1485	rc = VERR_NO_MEMORY;
1486	}
1487	return rc;
1488	}
1489
1490	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1491
1492	/**
1493	* Worker for vboxNetFltLinuxForwardToIntNet.
1494	*
1495	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1496	* @param pThis The net filter instance.
1497	* @param pBuf The socket buffer.
1498	* @param fSrc The source.
1499	*/
1500	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1501	{
1502	int rc;
1503	unsigned cbExtra;
1504	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf, &cbExtra);
1505	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1506	if (RT_LIKELY(pSG))
1507	{
1508	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cbExtra, cSegs, fSrc, NULL /pGsoCtx/);
1509
1510	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1511	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1512
1513	vboxNetFltLinuxDestroySG(pSG, pBuf);
1514	rc = VINF_SUCCESS;
1515	}
1516	else
1517	{
1518	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1519	rc = VERR_NO_MEMORY;
1520	}
1521	return rc;
1522	}
1523
1524
1525	/**
1526	* I won't disclose what I do, figure it out yourself, including pThis referencing.
1527	*
1528	* @param pThis The net filter instance.
1529	* @param pBuf The socket buffer.
1530	* @param fSrc Where the packet comes from.
1531	*/
1532	static void vboxNetFltLinuxForwardToIntNetInner(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1533	{
1534	#ifdef VBOXNETFLT_WITH_GSO
1535	if (skb_is_gso(pBuf))
1536	{
1537	PDMNETWORKGSO GsoCtx;
1538	Log6(("vboxNetFltLinuxForwardToIntNetInner: skb len=%u data_len=%u truesize=%u next=%p"
1539	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x ip_summed=%d\n",
1540	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next,
1541	skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size,
1542	skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type,
1543	skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, pBuf->ip_summed));
1544
1545	if (RT_LIKELY(fSrc & INTNETTRUNKDIR_HOST))
1546	{
1547	/*
1548	* skb_gso_segment does the following. Do we need to do it as well?
1549	*/
1550	# if RTLNX_VER_MIN(2,6,22)
1551	skb_reset_mac_header(pBuf);
1552	pBuf->mac_len = pBuf->network_header - pBuf->mac_header;
1553	# else
1554	pBuf->mac.raw = pBuf->data;
1555	pBuf->mac_len = pBuf->nh.raw - pBuf->data;
1556	# endif
1557	}
1558
1559	# ifdef VBOXNETFLT_WITH_GSO_RECV
1560	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1561	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1562	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1563	else
1564	# endif /* VBOXNETFLT_WITH_GSO_RECV */
1565	{
1566	/* Need to segment the packet */
1567	struct sk_buff *pNext;
1568	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1569	if (IS_ERR(pSegment))
1570	{
1571	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1572	return;
1573	}
1574
1575	for (; pSegment; pSegment = pNext)
1576	{
1577	Log6(("vboxNetFltLinuxForwardToIntNetInner: segment len=%u data_len=%u truesize=%u next=%p"
1578	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1579	pSegment->len, pSegment->data_len, pSegment->truesize, pSegment->next,
1580	skb_shinfo(pSegment)->nr_frags, skb_shinfo(pSegment)->gso_size,
1581	skb_shinfo(pSegment)->gso_segs, skb_shinfo(pSegment)->gso_type,
1582	skb_shinfo(pSegment)->frag_list, pSegment->pkt_type));
1583	pNext = pSegment->next;
1584	pSegment->next = 0;
1585	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1586	dev_kfree_skb(pSegment);
1587	}
1588	}
1589	}
1590	else
1591	#endif /* VBOXNETFLT_WITH_GSO */
1592	{
1593	Log6(("vboxNetFltLinuxForwardToIntNetInner: ptk_type=%d ip_summed=%d len=%d"
1594	" data_len=%d headroom=%d hdr_len=%d csum_offset=%d\n",
1595	pBuf->pkt_type, pBuf->ip_summed, pBuf->len, pBuf->data_len, skb_headroom(pBuf),
1596	skb_headlen(pBuf), vboxNetFltLinuxGetChecksumStartOffset(pBuf)));
1597	#ifndef VBOXNETFLT_SG_SUPPORT
1598	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1599	{
1600	# if RTLNX_VER_MIN(2,6,19)
1601	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1602	# else
1603	/*
1604	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1605	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1606	* header length from the header itself and reconstruct 'h' pointer
1607	* to TCP (or whatever) header.
1608	*/
1609	unsigned char *tmp = pBuf->h.raw;
1610	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1611	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1612	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1613	/* Restore the original (wrong) pointer. */
1614	pBuf->h.raw = tmp;
1615	# endif
1616	if (rc)
1617	{
1618	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1619	return;
1620	}
1621	}
1622	#endif /* !VBOXNETFLT_SG_SUPPORT */
1623	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1624	}
1625	}
1626
1627
1628	/**
1629	* Temporarily adjust pBuf->data so it always points to the Ethernet header,
1630	* then forward it to the internal network.
1631	*
1632	* @param pThis The net filter instance.
1633	* @param pBuf The socket buffer. This is consumed by this function.
1634	*/
1635	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1636	{
1637	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1638
1639	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1640	{
1641	/*
1642	* The packet came from the wire and the driver has already consumed
1643	* mac header. We need to restore it back. Moreover, after we are
1644	* through with this skb we need to restore its original state!
1645	*/
1646	skb_push(pBuf, pBuf->mac_len);
1647	Log5(("vboxNetFltLinuxForwardToIntNet: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1648	pBuf->mac_len, pBuf->data, skb_mac_header(pBuf), skb_network_header(pBuf)));
1649	}
1650
1651	vboxNetFltLinuxForwardToIntNetInner(pThis, pBuf, fSrc);
1652
1653	/*
1654	* Restore the original state of skb as there are other handlers this skb
1655	* will be provided to.
1656	*/
1657	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1658	skb_pull(pBuf, pBuf->mac_len);
1659
1660	dev_kfree_skb(pBuf);
1661	}
1662
1663
1664	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1665	/**
1666	* Work queue handler that forwards the socket buffers queued by
1667	* vboxNetFltLinuxPacketHandler to the internal network.
1668	*
1669	* @param pWork The work queue.
1670	*/
1671	# if RTLNX_VER_MIN(2,6,20)
1672	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1673	# else
1674	static void vboxNetFltLinuxXmitTask(void *pWork)
1675	# endif
1676	{
1677	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1678	struct sk_buff *pBuf;
1679
1680	Log6(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1681
1682	/*
1683	* Active? Retain the instance and increment the busy counter.
1684	*/
1685	if (vboxNetFltTryRetainBusyActive(pThis))
1686	{
1687	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1688	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1689
1690	vboxNetFltRelease(pThis, true /* fBusy */);
1691	}
1692	else
1693	{
1694	/** @todo Shouldn't we just drop the packets here? There is little point in
1695	* making them accumulate when the VM is paused and it'll only waste
1696	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1697	* before start draining the packets (goes for the intnet ring buf
1698	* too)? */
1699	}
1700	}
1701	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1702
1703	/**
1704	* Reports the GSO capabilities of the hardware NIC.
1705	*
1706	* @param pThis The net filter instance. The caller hold a
1707	* reference to this.
1708	*/
1709	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1710	{
1711	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1712	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1713	{
1714	struct net_device *pDev;
1715	unsigned int fFeatures;
1716
1717	RTSpinlockAcquire(pThis->hSpinlock);
1718
1719	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1720	if (pDev)
1721	fFeatures = pDev->features;
1722	else
1723	fFeatures = 0;
1724
1725	RTSpinlockRelease(pThis->hSpinlock);
1726
1727	if (pThis->pSwitchPort)
1728	{
1729	/* Set/update the GSO capabilities of the NIC. */
1730	uint32_t fGsoCapabilites = 0;
1731	if (fFeatures & NETIF_F_TSO)
1732	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1733	if (fFeatures & NETIF_F_TSO6)
1734	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1735	Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s\n",
1736	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1737	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : ""));
1738	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1739	}
1740
1741	vboxNetFltRelease(pThis, true /fBusy/);
1742	}
1743	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
1744	}
1745
1746	/**
1747	* Helper that determines whether the host (ignoreing us) is operating the
1748	* interface in promiscuous mode or not.
1749	*/
1750	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1751	{
1752	bool fRc = false;
1753	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1754	if (pDev)
1755	{
1756	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1757	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1758	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1759	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1760	}
1761	return fRc;
1762	}
1763
1764	/**
1765	* Does this device needs link state change signaled?
1766	* Currently we need it for our own VBoxNetAdp and TAP.
1767	*/
1768	static bool vboxNetFltNeedsLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev)
1769	{
1770	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1771	{
1772	struct ethtool_drvinfo Info;
1773
1774	memset(&Info, 0, sizeof(Info));
1775	Info.cmd = ETHTOOL_GDRVINFO;
1776	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1777	Log3(("%s: driver=%.s version=%.s bus_info=%.*s\n",
1778	__FUNCTION__,
1779	sizeof(Info.driver), Info.driver,
1780	sizeof(Info.version), Info.version,
1781	sizeof(Info.bus_info), Info.bus_info));
1782
1783	if (!strncmp(Info.driver, "vboxnet", sizeof(Info.driver)))
1784	return true;
1785
1786	#if RTLNX_VER_MIN(2,6,36) /* TAP started doing carrier */
1787	return !strncmp(Info.driver, "tun", 4)
1788	&& !strncmp(Info.bus_info, "tap", 4);
1789	#endif
1790	}
1791
1792	return false;
1793	}
1794
1795	#if RTLNX_VER_MAX(2,6,18)
1796	DECLINLINE(void) netif_tx_lock_bh(struct net_device *pDev)
1797	{
1798	spin_lock_bh(&pDev->xmit_lock);
1799	}
1800
1801	DECLINLINE(void) netif_tx_unlock_bh(struct net_device *pDev)
1802	{
1803	spin_unlock_bh(&pDev->xmit_lock);
1804	}
1805	#endif
1806
1807	/**
1808	* Some devices need link state change when filter attaches/detaches
1809	* since the filter is their link in a sense.
1810	*/
1811	static void vboxNetFltSetLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1812	{
1813	if (vboxNetFltNeedsLinkState(pThis, pDev))
1814	{
1815	Log3(("%s: bringing device link %s\n",
1816	__FUNCTION__, fLinkUp ? "up" : "down"));
1817	netif_tx_lock_bh(pDev);
1818	if (fLinkUp)
1819	netif_carrier_on(pDev);
1820	else
1821	netif_carrier_off(pDev);
1822	netif_tx_unlock_bh(pDev);
1823	}
1824	}
1825
1826	/**
1827	* Internal worker for vboxNetFltLinuxNotifierCallback.
1828	*
1829	* @returns VBox status code.
1830	* @param pThis The instance.
1831	* @param pDev The device to attach to.
1832	*/
1833	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1834	{
1835	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1836
1837	/*
1838	* Retain and store the device.
1839	*/
1840	dev_hold(pDev);
1841
1842	RTSpinlockAcquire(pThis->hSpinlock);
1843	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
1844	RTSpinlockRelease(pThis->hSpinlock);
1845
1846	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
1847	pDev, pDev->name,
1848	#if RTLNX_VER_MIN(2,6,37)
1849	netdev_refcnt_read(pDev)
1850	#else
1851	atomic_read(&pDev->refcnt)
1852	#endif
1853	));
1854	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1855	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1856
1857	/* Get the mac address while we still have a valid net_device reference. */
1858	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
1859	/* Initialize MTU */
1860	pThis->u.s.cbMtu = pDev->mtu;
1861
1862	/*
1863	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1864	*/
1865	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1866	pThis->u.s.PacketType.dev = pDev;
1867	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1868	dev_add_pack(&pThis->u.s.PacketType);
1869	ASMAtomicUoWriteBool(&pThis->u.s.fPacketHandler, true);
1870	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1871
1872	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1873	vboxNetFltLinuxHookDev(pThis, pDev);
1874	#endif
1875
1876	/*
1877	* Are we the "carrier" for this device (e.g. vboxnet or tap)?
1878	*/
1879	vboxNetFltSetLinkState(pThis, pDev, true);
1880
1881	/*
1882	* Set indicators that require the spinlock. Be abit paranoid about racing
1883	* the device notification handle.
1884	*/
1885	RTSpinlockAcquire(pThis->hSpinlock);
1886	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1887	if (pDev)
1888	{
1889	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1890	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1891	pDev = NULL; /* don't dereference it */
1892	}
1893	RTSpinlockRelease(pThis->hSpinlock);
1894
1895	/*
1896	* Report GSO capabilities
1897	*/
1898	Assert(pThis->pSwitchPort);
1899	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1900	{
1901	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1902	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
1903	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
1904	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
1905	vboxNetFltRelease(pThis, true /fBusy/);
1906	}
1907
1908	LogRel(("VBoxNetFlt: attached to '%s' / %RTmac\n", pThis->szName, &pThis->u.s.MacAddr));
1909	return VINF_SUCCESS;
1910	}
1911
1912
1913	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1914	{
1915	bool fRegistered;
1916	Assert(!pThis->fDisconnectedFromHost);
1917
1918	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1919	vboxNetFltLinuxUnhookDev(pThis, pDev);
1920	#endif
1921
1922	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
1923	{
1924	dev_remove_pack(&pThis->u.s.PacketType);
1925	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: packet handler removed.\n", pThis));
1926	}
1927
1928	RTSpinlockAcquire(pThis->hSpinlock);
1929	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
1930	if (fRegistered)
1931	{
1932	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1933	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1934	}
1935	RTSpinlockRelease(pThis->hSpinlock);
1936
1937	if (fRegistered)
1938	{
1939	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1940	skb_queue_purge(&pThis->u.s.XmitQueue);
1941	#endif
1942	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: xmit queue purged.\n", pThis));
1943	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
1944	pDev, pDev->name,
1945	#if RTLNX_VER_MIN(2,6,37)
1946	netdev_refcnt_read(pDev)
1947	#else
1948	atomic_read(&pDev->refcnt)
1949	#endif
1950	));
1951	dev_put(pDev);
1952	}
1953
1954	return NOTIFY_OK;
1955	}
1956
1957	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1958	{
1959	/* Check if we are not suspended and promiscuous mode has not been set. */
1960	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
1961	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1962	{
1963	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1964	dev_set_promiscuity(pDev, 1);
1965	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1966	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1967	}
1968	else
1969	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1970	return NOTIFY_OK;
1971	}
1972
1973	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1974	{
1975	/* Undo promiscuous mode if we has set it. */
1976	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1977	{
1978	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1979	dev_set_promiscuity(pDev, -1);
1980	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1981	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1982	}
1983	else
1984	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1985	return NOTIFY_OK;
1986	}
1987
1988	/**
1989	* Callback for listening to MTU change event.
1990	*
1991	* We need to track changes of host's inteface MTU to discard over-sized frames
1992	* coming from the internal network as they may hang the TX queue of host's
1993	* adapter.
1994	*
1995	* @returns NOTIFY_OK
1996	* @param pThis The netfilter instance.
1997	* @param pDev Pointer to device structure of host's interface.
1998	*/
1999	static int vboxNetFltLinuxDeviceMtuChange(PVBOXNETFLTINS pThis, struct net_device *pDev)
2000	{
2001	ASMAtomicWriteU32(&pThis->u.s.cbMtu, pDev->mtu);
2002	Log(("vboxNetFltLinuxDeviceMtuChange: set MTU for %s to %d\n", pThis->szName, pDev->mtu));
2003	return NOTIFY_OK;
2004	}
2005
2006	#ifdef LOG_ENABLED
2007	/** Stringify the NETDEV_XXX constants. */
2008	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2009	{
2010	const char *pszEvent = "NETDEV_<unknown>";
2011	switch (ulEventType)
2012	{
2013	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2014	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2015	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2016	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2017	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2018	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2019	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2020	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2021	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2022	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2023	# ifdef NETDEV_FEAT_CHANGE
2024	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2025	# endif
2026	}
2027	return pszEvent;
2028	}
2029	#endif /* LOG_ENABLED */
2030
2031	/**
2032	* Callback for listening to netdevice events.
2033	*
2034	* This works the rediscovery, clean up on unregistration, promiscuity on
2035	* up/down, and GSO feature changes from ethtool.
2036	*
2037	* @returns NOTIFY_OK
2038	* @param self Pointer to our notifier registration block.
2039	* @param ulEventType The event.
2040	* @param ptr Event specific, but it is usually the device it
2041	* relates to.
2042	*/
2043	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2044
2045	{
2046	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2047	struct net_device pMyDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device );
2048	struct net_device *pDev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2049	int rc = NOTIFY_OK;
2050
2051	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2052	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, pMyDev));
2053
2054	if (ulEventType == NETDEV_REGISTER)
2055	{
2056	#if RTLNX_VER_MIN(2,6,24) /* cgroups/namespaces introduced */
2057	# if RTLNX_VER_MIN(2,6,26)
2058	# define VBOX_DEV_NET(dev) dev_net(dev)
2059	# define VBOX_NET_EQ(n1, n2) net_eq((n1), (n2))
2060	# else
2061	# define VBOX_DEV_NET(dev) ((dev)->nd_net)
2062	# define VBOX_NET_EQ(n1, n2) ((n1) == (n2))
2063	# endif
2064	struct net *pMyNet = current->nsproxy->net_ns;
2065	struct net *pDevNet = VBOX_DEV_NET(pDev);
2066
2067	if (VBOX_NET_EQ(pDevNet, pMyNet))
2068	#endif /* namespaces */
2069	{
2070	if (strcmp(pDev->name, pThis->szName) == 0)
2071	{
2072	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2073	}
2074	}
2075	}
2076	else
2077	{
2078	if (pDev == pMyDev)
2079	{
2080	switch (ulEventType)
2081	{
2082	case NETDEV_UNREGISTER:
2083	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2084	break;
2085	case NETDEV_UP:
2086	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2087	break;
2088	case NETDEV_GOING_DOWN:
2089	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2090	break;
2091	case NETDEV_CHANGEMTU:
2092	rc = vboxNetFltLinuxDeviceMtuChange(pThis, pDev);
2093	break;
2094	case NETDEV_CHANGENAME:
2095	break;
2096	#ifdef NETDEV_FEAT_CHANGE
2097	case NETDEV_FEAT_CHANGE:
2098	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2099	break;
2100	#endif
2101	}
2102	}
2103	}
2104
2105	return rc;
2106	}
2107
2108	/*
2109	* Initial enumeration of netdevs. Called with NETDEV_REGISTER by
2110	* register_netdevice_notifier() under rtnl lock.
2111	*/
2112	static int vboxNetFltLinuxEnumeratorCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2113	{
2114	PVBOXNETFLTINS pThis = ((PVBOXNETFLTNOTIFIER)self)->pThis;
2115	struct net_device *dev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2116	struct in_device *in_dev;
2117	struct inet6_dev *in6_dev;
2118
2119	if (ulEventType != NETDEV_REGISTER)
2120	return NOTIFY_OK;
2121
2122	if (RT_UNLIKELY(pThis->pSwitchPort->pfnNotifyHostAddress == NULL))
2123	return NOTIFY_OK;
2124
2125	/*
2126	* IPv4
2127	*/
2128	#if RTLNX_VER_MIN(2,6,14)
2129	in_dev = __in_dev_get_rtnl(dev);
2130	#else
2131	in_dev = __in_dev_get(dev);
2132	#endif
2133	if (in_dev != NULL)
2134	{
2135	struct in_ifaddr *ifa;
2136
2137	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2138	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2139	return NOTIFY_OK;
2140
2141	if ( dev != pThis->u.s.pDev
2142	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2143	continue;
2144
2145	Log(("%s: %s: IPv4 addr %RTnaipv4 mask %RTnaipv4\n",
2146	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2147	ifa->ifa_address, ifa->ifa_mask));
2148
2149	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2150	/* :fAdded */ true, kIntNetAddrType_IPv4, &ifa->ifa_address);
2151	}
2152	}
2153
2154	/*
2155	* IPv6
2156	*/
2157	in6_dev = __in6_dev_get(dev);
2158	if (in6_dev != NULL)
2159	{
2160	struct inet6_ifaddr *ifa;
2161
2162	read_lock_bh(&in6_dev->lock);
2163	#if RTLNX_VER_MIN(2,6,35)
2164	list_for_each_entry(ifa, &in6_dev->addr_list, if_list)
2165	#else
2166	for (ifa = in6_dev->addr_list; ifa != NULL; ifa = ifa->if_next)
2167	#endif
2168	{
2169	if ( dev != pThis->u.s.pDev
2170	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2171	continue;
2172
2173	Log(("%s: %s: IPv6 addr %RTnaipv6/%u\n",
2174	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2175	&ifa->addr, (unsigned)ifa->prefix_len));
2176
2177	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2178	/* :fAdded */ true, kIntNetAddrType_IPv6, &ifa->addr);
2179	}
2180	read_unlock_bh(&in6_dev->lock);
2181	}
2182
2183	return NOTIFY_OK;
2184	}
2185
2186
2187	static int vboxNetFltLinuxNotifierIPv4Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2188	{
2189	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv4);
2190	struct net_device pDev, pEventDev;
2191	struct in_ifaddr ifa = (struct in_ifaddr )ptr;
2192	bool fMyDev;
2193	int rc = NOTIFY_OK;
2194
2195	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2196	pEventDev = ifa->ifa_dev->dev;
2197	fMyDev = (pDev == pEventDev);
2198	Log(("VBoxNetFlt: %s: IPv4 event %s(0x%lx) %s: addr %RTnaipv4 mask %RTnaipv4\n",
2199	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2200	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2201	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2202	ifa->ifa_address, ifa->ifa_mask));
2203
2204	if (pDev != NULL)
2205	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2206
2207	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2208	return NOTIFY_OK;
2209
2210	if ( !fMyDev
2211	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2212	return NOTIFY_OK;
2213
2214	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2215	{
2216	bool fAdded;
2217	if (ulEventType == NETDEV_UP)
2218	fAdded = true;
2219	else if (ulEventType == NETDEV_DOWN)
2220	fAdded = false;
2221	else
2222	return NOTIFY_OK;
2223
2224	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2225	kIntNetAddrType_IPv4, &ifa->ifa_local);
2226	}
2227
2228	return rc;
2229	}
2230
2231
2232	static int vboxNetFltLinuxNotifierIPv6Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2233	{
2234	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv6);
2235	struct net_device pDev, pEventDev;
2236	struct inet6_ifaddr ifa = (struct inet6_ifaddr )ptr;
2237	bool fMyDev;
2238	int rc = NOTIFY_OK;
2239
2240	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2241	pEventDev = ifa->idev->dev;
2242	fMyDev = (pDev == pEventDev);
2243	Log(("VBoxNetFlt: %s: IPv6 event %s(0x%lx) %s: %RTnaipv6\n",
2244	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2245	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2246	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2247	&ifa->addr));
2248
2249	if (pDev != NULL)
2250	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2251
2252	if ( !fMyDev
2253	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2254	return NOTIFY_OK;
2255
2256	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2257	{
2258	bool fAdded;
2259	if (ulEventType == NETDEV_UP)
2260	fAdded = true;
2261	else if (ulEventType == NETDEV_DOWN)
2262	fAdded = false;
2263	else
2264	return NOTIFY_OK;
2265
2266	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2267	kIntNetAddrType_IPv6, &ifa->addr);
2268	}
2269
2270	return rc;
2271	}
2272
2273
2274	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2275	{
2276	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2277	}
2278
2279	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2280	{
2281	struct net_device * pDev;
2282	int err;
2283	int rc = VINF_SUCCESS;
2284	IPRT_LINUX_SAVE_EFL_AC();
2285	NOREF(pvIfData);
2286
2287	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2288
2289	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2290	if (pDev)
2291	{
2292	/*
2293	* Create a sk_buff for the gather list and push it onto the wire.
2294	*/
2295	if (fDst & INTNETTRUNKDIR_WIRE)
2296	{
2297	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2298	if (pBuf)
2299	{
2300	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2301	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2302	Log6(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2303	err = dev_queue_xmit(pBuf);
2304	if (err)
2305	rc = RTErrConvertFromErrno(err);
2306	}
2307	else
2308	rc = VERR_NO_MEMORY;
2309	}
2310
2311	/*
2312	* Create a sk_buff for the gather list and push it onto the host stack.
2313	*/
2314	if (fDst & INTNETTRUNKDIR_HOST)
2315	{
2316	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2317	if (pBuf)
2318	{
2319	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2320	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2321	Log6(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2322	#if RTLNX_VER_MIN(5,18,0)
2323	local_bh_disable();
2324	err = netif_rx(pBuf);
2325	local_bh_enable();
2326	#else
2327	err = netif_rx_ni(pBuf);
2328	#endif
2329	if (err)
2330	rc = RTErrConvertFromErrno(err);
2331	}
2332	else
2333	rc = VERR_NO_MEMORY;
2334	}
2335
2336	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2337	}
2338
2339	IPRT_LINUX_RESTORE_EFL_AC();
2340	return rc;
2341	}
2342
2343
2344	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2345	{
2346	struct net_device *pDev;
2347	IPRT_LINUX_SAVE_EFL_AC();
2348
2349	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%RTbool, fDisablePromiscuous=%RTbool\n",
2350	pThis, pThis->szName, fActive, pThis->fDisablePromiscuous));
2351
2352	if (pThis->fDisablePromiscuous)
2353	return;
2354
2355	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2356	if (pDev)
2357	{
2358	/*
2359	* This api is a bit weird, the best reference is the code.
2360	*
2361	* Also, we have a bit or race conditions wrt the maintenance of
2362	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2363	*/
2364	#ifdef LOG_ENABLED
2365	u_int16_t fIf;
2366	unsigned const cPromiscBefore = pDev->promiscuity;
2367	#endif
2368	if (fActive)
2369	{
2370	Assert(!pThis->u.s.fPromiscuousSet);
2371
2372	rtnl_lock();
2373	dev_set_promiscuity(pDev, 1);
2374	rtnl_unlock();
2375	pThis->u.s.fPromiscuousSet = true;
2376	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2377	}
2378	else
2379	{
2380	if (pThis->u.s.fPromiscuousSet)
2381	{
2382	rtnl_lock();
2383	dev_set_promiscuity(pDev, -1);
2384	rtnl_unlock();
2385	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2386	}
2387	pThis->u.s.fPromiscuousSet = false;
2388
2389	#ifdef LOG_ENABLED
2390	fIf = dev_get_flags(pDev);
2391	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2392	#endif
2393	}
2394
2395	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2396	}
2397	IPRT_LINUX_RESTORE_EFL_AC();
2398	}
2399
2400
2401	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2402	{
2403	/*
2404	* Remove packet handler when we get disconnected from internal switch as
2405	* we don't want the handler to forward packets to disconnected switch.
2406	*/
2407	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
2408	{
2409	IPRT_LINUX_SAVE_EFL_AC();
2410	dev_remove_pack(&pThis->u.s.PacketType);
2411	Log(("vboxNetFltOsDisconnectIt: this=%p: Packet handler removed.\n", pThis));
2412	IPRT_LINUX_RESTORE_EFL_AC();
2413	}
2414	return VINF_SUCCESS;
2415	}
2416
2417
2418	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2419	{
2420	IPRT_LINUX_SAVE_EFL_AC();
2421
2422	/*
2423	* Report the GSO capabilities of the host and device (if connected).
2424	* Note! No need to mark ourselves busy here.
2425	*/
2426	/** @todo duplicate work here now? Attach */
2427	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2428	Log3(("vboxNetFltOsConnectIt: reporting host tso tso6\n"));
2429	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2430	0
2431	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2432	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2433	, INTNETTRUNKDIR_HOST);
2434
2435	#endif
2436	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2437
2438	IPRT_LINUX_RESTORE_EFL_AC();
2439	return VINF_SUCCESS;
2440	}
2441
2442
2443	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2444	{
2445	struct net_device *pDev;
2446	bool fRegistered;
2447	IPRT_LINUX_SAVE_EFL_AC();
2448
2449	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
2450	vboxNetFltLinuxUnhookDev(pThis, NULL);
2451	#endif
2452
2453	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2454	* unlikely, but none the less). Since it doesn't actually update the
2455	* state (just reads it), it is likely to panic in some interesting
2456	* ways. */
2457
2458	RTSpinlockAcquire(pThis->hSpinlock);
2459	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2460	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
2461	RTSpinlockRelease(pThis->hSpinlock);
2462
2463	if (fRegistered)
2464	{
2465	vboxNetFltSetLinkState(pThis, pDev, false);
2466
2467	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2468	skb_queue_purge(&pThis->u.s.XmitQueue);
2469	#endif
2470	Log(("vboxNetFltOsDeleteInstance: this=%p: xmit queue purged.\n", pThis));
2471	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2472	pDev, pDev->name,
2473	#if RTLNX_VER_MIN(2,6,37)
2474	netdev_refcnt_read(pDev)
2475	#else
2476	atomic_read(&pDev->refcnt)
2477	#endif
2478	));
2479	dev_put(pDev);
2480	}
2481
2482	unregister_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2483	unregister_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2484
2485	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2486	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2487	module_put(THIS_MODULE);
2488
2489	IPRT_LINUX_RESTORE_EFL_AC();
2490	}
2491
2492
2493	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2494	{
2495	int err;
2496	IPRT_LINUX_SAVE_EFL_AC();
2497	NOREF(pvContext);
2498
2499	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2500	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2501	if (err)
2502	{
2503	IPRT_LINUX_RESTORE_EFL_AC();
2504	return VERR_INTNET_FLT_IF_FAILED;
2505	}
2506	if (!pThis->u.s.fRegistered)
2507	{
2508	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2509	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2510	IPRT_LINUX_RESTORE_EFL_AC();
2511	return VERR_INTNET_FLT_IF_NOT_FOUND;
2512	}
2513
2514	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2515	if ( pThis->fDisconnectedFromHost
2516	\|\| !try_module_get(THIS_MODULE))
2517	{
2518	IPRT_LINUX_RESTORE_EFL_AC();
2519	return VERR_INTNET_FLT_IF_FAILED;
2520	}
2521
2522	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2523	{
2524	VBOXNETFLTNOTIFIER Enumerator;
2525
2526	/*
2527	* register_inetaddr_notifier() and register_inet6addr_notifier()
2528	* do not call the callback for existing devices. Enumerating
2529	* all network devices explicitly is a bit of an ifdef mess,
2530	* so co-opt register_netdevice_notifier() to do that for us.
2531	*/
2532	RT_ZERO(Enumerator);
2533	Enumerator.Notifier.notifier_call = vboxNetFltLinuxEnumeratorCallback;
2534	Enumerator.pThis = pThis;
2535
2536	err = register_netdevice_notifier(&Enumerator.Notifier);
2537	if (err)
2538	{
2539	LogRel(("%s: failed to enumerate network devices: error %d\n", __FUNCTION__, err));
2540	IPRT_LINUX_RESTORE_EFL_AC();
2541	return VINF_SUCCESS;
2542	}
2543
2544	unregister_netdevice_notifier(&Enumerator.Notifier);
2545
2546	pThis->u.s.NotifierIPv4.notifier_call = vboxNetFltLinuxNotifierIPv4Callback;
2547	err = register_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2548	if (err)
2549	LogRel(("%s: failed to register IPv4 notifier: error %d\n", __FUNCTION__, err));
2550
2551	pThis->u.s.NotifierIPv6.notifier_call = vboxNetFltLinuxNotifierIPv6Callback;
2552	err = register_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2553	if (err)
2554	LogRel(("%s: failed to register IPv6 notifier: error %d\n", __FUNCTION__, err));
2555	}
2556
2557	IPRT_LINUX_RESTORE_EFL_AC();
2558	return VINF_SUCCESS;
2559	}
2560
2561	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2562	{
2563	IPRT_LINUX_SAVE_EFL_AC();
2564
2565	/*
2566	* Init the linux specific members.
2567	*/
2568	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2569	pThis->u.s.fRegistered = false;
2570	pThis->u.s.fPromiscuousSet = false;
2571	pThis->u.s.fPacketHandler = false;
2572	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2573	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2574	skb_queue_head_init(&pThis->u.s.XmitQueue);
2575	# if RTLNX_VER_MIN(2,6,20)
2576	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2577	# else
2578	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2579	# endif
2580	#endif
2581
2582	IPRT_LINUX_RESTORE_EFL_AC();
2583	return VINF_SUCCESS;
2584	}
2585
2586
2587	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2588	{
2589	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2590	}
2591
2592
2593	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2594	{
2595	/* Nothing to do */
2596	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2597	return VINF_SUCCESS;
2598	}
2599
2600
2601	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2602	{
2603	/* Nothing to do */
2604	NOREF(pThis); NOREF(pvIfData);
2605	return VINF_SUCCESS;
2606	}
2607

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

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 97405

Download in other formats: