VBoxNetFlt-linux.c@ 99941

最後變更在這個檔案從99941是 99739,由 vboxsync 提交於 23 月前
*: doxygen corrections (mostly about removing @returns from functions returning void).
屬性 svn:eol-style 設為 `native` 屬性 svn:keywords 設為 `Author Date Id Revision`
檔案大小: 90.3 KB

行
1	/* $Id: VBoxNetFlt-linux.c 99739 2023-05-11 01:01:08Z vboxsync $ */
2	/** @file
3	* VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4	*/
5
6	/*
7	* Copyright (C) 2006-2023 Oracle and/or its affiliates.
8	*
9	* This file is part of VirtualBox base platform packages, as
10	* available from https://www.alldomusa.eu.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	* @param pThis The instance.
840	* @param pBuf The sk_buff.
841	* @param pSG The SG.
842	* @param cbExtra The number of bytes of extra space allocated immediately after the SG.
843	* @param cSegs The number of segments allocated for the SG.
844	* This should match the number in the mbuf exactly!
845	* @param fSrc The source of the frame.
846	* @param pGsoCtx Pointer to the GSO context if it's a GSO
847	* internal network frame. NULL if regular frame.
848	*/
849	static void vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
850	unsigned cbExtra, unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
851	{
852	int i;
853	NOREF(pThis);
854
855	#ifndef VBOXNETFLT_SG_SUPPORT
856	Assert(!skb_shinfo(pBuf)->frag_list);
857	#else /* VBOXNETFLT_SG_SUPPORT */
858	uint8_t pExtra = (uint8_t )&pSG->aSegs[cSegs];
859	unsigned cbConsumed = 0;
860	unsigned cbProduced = 0;
861
862	# if RTLNX_VER_MIN(2,6,27)
863	/* Restore VLAN tag stripped by host hardware */
864	if (vlan_tx_tag_present(pBuf))
865	{
866	uint8_t *pMac = pBuf->data;
867	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )pExtra;
868	Assert(ETH_ALEN * 2 + VLAN_HLEN <= cbExtra);
869	memmove(pVHdr, pMac, ETH_ALEN * 2);
870	/* Consume whole Ethernet header: 2 addresses + EtherType (see @bugref{8599}) */
871	cbConsumed += ETH_ALEN * 2 + sizeof(uint16_t);
872	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
873	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
874	pVHdr->h_vlan_encapsulated_proto = (uint16_t)(pMac + ETH_ALEN * 2);
875	cbProduced += VLAN_ETH_HLEN;
876	}
877	# endif /* RTLNX_VER_MIN(2,6,27) */
878
879	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
880	{
881	unsigned uCsumStartOffset = vboxNetFltLinuxGetChecksumStartOffset(pBuf);
882	unsigned uCsumStoreOffset = uCsumStartOffset + VBOX_SKB_CSUM_OFFSET(pBuf) - cbConsumed;
883	Log3(("cbConsumed=%u cbProduced=%u uCsumStartOffset=%u uCsumStoreOffset=%u\n",
884	cbConsumed, cbProduced, uCsumStartOffset, uCsumStoreOffset));
885	Assert(cbProduced + uCsumStoreOffset + sizeof(uint16_t) <= cbExtra);
886	/*
887	* We assume that the checksum is stored at the very end of the transport header
888	* so we will have all headers in a single fragment. If our assumption is wrong
889	* we may see suboptimal performance.
890	*/
891	memmove(pExtra + cbProduced,
892	pBuf->data + cbConsumed,
893	uCsumStoreOffset);
894	unsigned uChecksum = skb_checksum(pBuf, uCsumStartOffset, pBuf->len - uCsumStartOffset, 0);
895	(uint16_t)(pExtra + cbProduced + uCsumStoreOffset) = csum_fold(uChecksum);
896	cbProduced += uCsumStoreOffset + sizeof(uint16_t);
897	cbConsumed += uCsumStoreOffset + sizeof(uint16_t);
898	}
899	#endif /* VBOXNETFLT_SG_SUPPORT */
900
901	if (!pGsoCtx)
902	IntNetSgInitTempSegs(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/);
903	else
904	IntNetSgInitTempSegsGso(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/, pGsoCtx);
905
906	int iSeg = 0;
907	#ifdef VBOXNETFLT_SG_SUPPORT
908	if (cbProduced)
909	{
910	pSG->aSegs[iSeg].cb = cbProduced;
911	pSG->aSegs[iSeg].pv = pExtra;
912	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
913	}
914	pSG->aSegs[iSeg].cb = skb_headlen(pBuf) - cbConsumed;
915	pSG->aSegs[iSeg].pv = pBuf->data + cbConsumed;
916	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
917	Assert(iSeg <= pSG->cSegsAlloc);
918
919	# ifdef LOG_ENABLED
920	if (pBuf->data_len)
921	Log6((" kmap_atomic:"));
922	# endif /* LOG_ENABLED */
923	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
924	{
925	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
926	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
927	pSG->aSegs[iSeg].cb = pFrag->bv_len;
928	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
929	# else /* < KERNEL_VERSION(5, 4, 0) */
930	pSG->aSegs[iSeg].cb = pFrag->size;
931	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
932	# endif /* >= KERNEL_VERSION(5, 4, 0) */
933	Log6((" %p", pSG->aSegs[iSeg].pv));
934	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
935	Assert(iSeg <= pSG->cSegsAlloc);
936	}
937	struct sk_buff *pFragBuf;
938	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
939	{
940	pSG->aSegs[iSeg].cb = skb_headlen(pFragBuf);
941	pSG->aSegs[iSeg].pv = pFragBuf->data;
942	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
943	Assert(iSeg <= pSG->cSegsAlloc);
944	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
945	{
946	skb_frag_t *pFrag = &skb_shinfo(pFragBuf)->frags[i];
947	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
948	pSG->aSegs[iSeg].cb = pFrag->bv_len;
949	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
950	# else /* < KERNEL_VERSION(5, 4, 0) */
951	pSG->aSegs[iSeg].cb = pFrag->size;
952	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
953	# endif /* >= KERNEL_VERSION(5, 4, 0) */
954	Log6((" %p", pSG->aSegs[iSeg].pv));
955	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
956	Assert(iSeg <= pSG->cSegsAlloc);
957	}
958	}
959	# ifdef LOG_ENABLED
960	if (pBuf->data_len)
961	Log6(("\n"));
962	# endif /* LOG_ENABLED */
963	#else
964	pSG->aSegs[iSeg].cb = pBuf->len;
965	pSG->aSegs[iSeg].pv = pBuf->data;
966	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
967	#endif
968
969	pSG->cSegsUsed = iSeg;
970
971	#if 0
972	if (cbProduced)
973	{
974	LogRel(("vboxNetFltLinuxSkBufToSG: original packet dump:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
975	LogRel(("vboxNetFltLinuxSkBufToSG: cbConsumed=%u cbProduced=%u cbExtra=%u\n", cbConsumed, cbProduced, cbExtra));
976	uint32_t offset = 0;
977	for (i = 0; i < pSG->cSegsUsed; ++i)
978	{
979	LogRel(("vboxNetFltLinuxSkBufToSG: seg#%d (%d bytes, starting at 0x%x):\n%.*Rhxd\n",
980	i, pSG->aSegs[i].cb, offset, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
981	offset += pSG->aSegs[i].cb;
982	}
983	}
984	#endif
985
986	#ifdef PADD_RUNT_FRAMES_FROM_HOST
987	/*
988	* Add a trailer if the frame is too small.
989	*
990	* Since we're getting to the packet before it is framed, it has not
991	* yet been padded. The current solution is to add a segment pointing
992	* to a buffer containing all zeros and pray that works for all frames...
993	*/
994	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
995	{
996	Assert(pBuf->data_len == 0); /* Packets with fragments are never small! */
997	static uint8_t const s_abZero[128] = {0};
998
999	AssertReturnVoid(iSeg < cSegs);
1000
1001	pSG->aSegs[iSeg].Phys = NIL_RTHCPHYS;
1002	pSG->aSegs[iSeg].pv = (void *)&s_abZero[0];
1003	pSG->aSegs[iSeg++].cb = 60 - pSG->cbTotal;
1004	pSG->cbTotal = 60;
1005	pSG->cSegsUsed++;
1006	Assert(iSeg <= pSG->cSegsAlloc)
1007	}
1008	#endif
1009
1010	Log6(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1011	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1012	for (i = 0; i < pSG->cSegsUsed; i++)
1013	Log6(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1014	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1015	}
1016
1017	/**
1018	* Packet handler; not really documented - figure it out yourself.
1019	*
1020	* @returns 0 or EJUSTRETURN - this is probably copy & pastry and thus wrong.
1021	*/
1022	#if RTLNX_VER_MIN(2,6,14)
1023	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1024	struct net_device *pSkbDev,
1025	struct packet_type *pPacketType,
1026	struct net_device *pOrigDev)
1027	#else
1028	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1029	struct net_device *pSkbDev,
1030	struct packet_type *pPacketType)
1031	#endif
1032	{
1033	PVBOXNETFLTINS pThis;
1034	struct net_device *pDev;
1035	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1036	pBuf, pSkbDev, pPacketType));
1037	#if RTLNX_VER_MIN(2,6,18)
1038	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",
1039	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));
1040	# if RTLNX_VER_MIN(2,6,22)
1041	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1042	# endif
1043	#else
1044	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",
1045	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));
1046	#endif
1047	/*
1048	* Drop it immediately?
1049	*/
1050	if (!pBuf)
1051	return 0;
1052
1053	if (pBuf->pkt_type == PACKET_LOOPBACK)
1054	{
1055	/*
1056	* We are not interested in loopbacked packets as they will always have
1057	* another copy going to the wire.
1058	*/
1059	Log2(("vboxNetFltLinuxPacketHandler: dropped loopback packet (cb=%u)\n", pBuf->len));
1060	dev_kfree_skb(pBuf); /* We must 'consume' all packets we get (@bugref{6539})! */
1061	return 0;
1062	}
1063
1064	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1065	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1066	if (pDev != pSkbDev)
1067	{
1068	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1069	kfree_skb(pBuf); /* This is a failure, so we use kfree_skb instead of dev_kfree_skb. */
1070	return 0;
1071	}
1072
1073	Log6(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1074	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1075	{
1076	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1077	dev_kfree_skb(pBuf);
1078	return 0;
1079	}
1080
1081	#ifndef VBOXNETFLT_SG_SUPPORT
1082	{
1083	/*
1084	* Get rid of fragmented packets, they cause too much trouble.
1085	*/
1086	unsigned int uMacLen = pBuf->mac_len;
1087	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1088	dev_kfree_skb(pBuf);
1089	if (!pCopy)
1090	{
1091	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1092	return 0;
1093	}
1094	pBuf = pCopy;
1095	/* Somehow skb_copy ignores mac_len */
1096	pBuf->mac_len = uMacLen;
1097	# if RTLNX_VER_MIN(2,6,27)
1098	/* Restore VLAN tag stripped by host hardware */
1099	if (vlan_tx_tag_present(pBuf) && skb_headroom(pBuf) >= VLAN_ETH_HLEN)
1100	{
1101	uint8_t pMac = (uint8_t)skb_mac_header(pBuf);
1102	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )(pMac - VLAN_HLEN);
1103	memmove(pVHdr, pMac, ETH_ALEN * 2);
1104	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
1105	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
1106	pBuf->mac_header -= VLAN_HLEN;
1107	pBuf->mac_len += VLAN_HLEN;
1108	}
1109	# endif /* RTLNX_VER_MIN(2,6,27) */
1110
1111	# if RTLNX_VER_MIN(2,6,18)
1112	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",
1113	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));
1114	# if RTLNX_VER_MIN(2,6,22)
1115	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1116	# endif /* RTLNX_VER_MIN(2,6,22) */
1117	# else /* RTLNX_VER_MAX(2,6,18) */
1118	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",
1119	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));
1120	# endif /* RTLNX_VER_MAX(2,6,18) */
1121	}
1122	#endif /* !VBOXNETFLT_SG_SUPPORT */
1123
1124	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1125	/* Forward it to the internal network. */
1126	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1127	#else /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1128	/* Add the packet to transmit queue and schedule the bottom half. */
1129	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1130	schedule_work(&pThis->u.s.XmitTask);
1131	Log6(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1132	&pThis->u.s.XmitTask, pBuf));
1133	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1134
1135	/* It does not really matter what we return, it is ignored by the kernel. */
1136	return 0;
1137	}
1138
1139	/**
1140	* Calculate the number of INTNETSEG segments the socket buffer will need.
1141	*
1142	* @returns Segment count.
1143	* @param pBuf The socket buffer.
1144	* @param pcbTemp Where to store the number of bytes of the part
1145	* of the socket buffer that will be copied to
1146	* a temporary storage.
1147	*/
1148	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff pBuf, unsigned pcbTemp)
1149	{
1150	*pcbTemp = 0;
1151	#ifdef VBOXNETFLT_SG_SUPPORT
1152	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1153	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1154	{
1155	*pcbTemp = vboxNetFltLinuxGetChecksumStartOffset(pBuf) + VBOX_SKB_CSUM_OFFSET(pBuf) + sizeof(uint16_t);
1156	}
1157	# if RTLNX_VER_MIN(2,6,27)
1158	if (vlan_tx_tag_present(pBuf))
1159	{
1160	if (*pcbTemp)
1161	*pcbTemp += VLAN_HLEN;
1162	else
1163	*pcbTemp = VLAN_ETH_HLEN;
1164	}
1165	# endif /* RTLNX_VER_MIN(2,6,27) */
1166	if (*pcbTemp)
1167	++cSegs;
1168	struct sk_buff *pFrag;
1169	for (pFrag = skb_shinfo(pBuf)->frag_list; pFrag; pFrag = pFrag->next)
1170	{
1171	Log6(("vboxNetFltLinuxCalcSGSegments: frag=%p len=%d data_len=%d frags=%d frag_list=%p next=%p\n",
1172	pFrag, pFrag->len, pFrag->data_len, skb_shinfo(pFrag)->nr_frags, skb_shinfo(pFrag)->frag_list, pFrag->next));
1173	cSegs += 1 + skb_shinfo(pFrag)->nr_frags;
1174	}
1175	#else
1176	unsigned cSegs = 1;
1177	#endif
1178	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1179	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1180	if (pBuf->len < 60)
1181	cSegs++;
1182	#endif
1183	return cSegs;
1184	}
1185
1186
1187	/**
1188	* Destroy the intnet scatter / gather buffer created by
1189	* vboxNetFltLinuxSkBufToSG.
1190	*
1191	* @param pSG The (scatter/)gather list.
1192	* @param pBuf The original socket buffer that was used to create
1193	* the scatter/gather list.
1194	*/
1195	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG, struct sk_buff *pBuf)
1196	{
1197	#ifdef VBOXNETFLT_SG_SUPPORT
1198	int i, iSeg = 1; /* Skip non-paged part of SKB */
1199	/* Check if the extra buffer behind SG structure was used for modified packet header */
1200	if (pBuf->data != pSG->aSegs[0].pv)
1201	++iSeg; /* Skip it as well */
1202	# ifdef LOG_ENABLED
1203	if (pBuf->data_len)
1204	Log6(("kunmap_atomic:"));
1205	# endif /* LOG_ENABLED */
1206	/* iSeg now points to the first mapped fragment if there are any */
1207	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1208	{
1209	Log6((" %p", pSG->aSegs[iSeg].pv));
1210	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1211	}
1212	struct sk_buff *pFragBuf;
1213	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
1214	{
1215	++iSeg; /* Non-fragment (unmapped) portion of chained SKB */
1216	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
1217	{
1218	Log6((" %p", pSG->aSegs[iSeg].pv));
1219	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1220	}
1221	}
1222	# ifdef LOG_ENABLED
1223	if (pBuf->data_len)
1224	Log6(("\n"));
1225	# endif /* LOG_ENABLED */
1226	#endif
1227	NOREF(pSG);
1228	}
1229
1230	#ifdef LOG_ENABLED
1231	/**
1232	* Logging helper.
1233	*/
1234	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1235	{
1236	int i, offSeg;
1237	uint8_t pInt, pExt;
1238	static int iPacketNo = 1;
1239	iPacketNo += iIncrement;
1240	if (fEgress)
1241	{
1242	pExt = pSG->aSegs[0].pv;
1243	pInt = pExt + 6;
1244	}
1245	else
1246	{
1247	pInt = pSG->aSegs[0].pv;
1248	pExt = pInt + 6;
1249	}
1250	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1251	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1252	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1253	fEgress ? "-->" : "<--", pszWhere,
1254	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1255	pSG->cbTotal, iPacketNo));
1256	if (pSG->cSegsUsed == 1)
1257	{
1258	Log4(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1259	}
1260	else
1261	{
1262	for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1263	{
1264	Log4(("-- segment %d at 0x%x (%d bytes)\n --\n%.*Rhxd\n",
1265	i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1266	offSeg += pSG->aSegs[i].cb;
1267	}
1268	}
1269	}
1270	#else
1271	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1272	#endif
1273
1274	#ifdef VBOXNETFLT_WITH_GSO_RECV
1275
1276	/**
1277	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1278	* GSO socket buffer without having to segment it.
1279	*
1280	* @returns true on success, false if needs segmenting.
1281	* @param pThis The net filter instance.
1282	* @param pSkb The GSO socket buffer.
1283	* @param fSrc The source.
1284	* @param pGsoCtx Where to return the GSO context on success.
1285	*/
1286	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1287	PPDMNETWORKGSO pGsoCtx)
1288	{
1289	PDMNETWORKGSOTYPE enmGsoType;
1290	uint16_t uEtherType;
1291	unsigned int cbTransport;
1292	unsigned int offTransport;
1293	unsigned int cbTransportHdr;
1294	unsigned uProtocol;
1295	union
1296	{
1297	RTNETIPV4 IPv4;
1298	RTNETIPV6 IPv6;
1299	RTNETTCP Tcp;
1300	uint8_t ab[40];
1301	uint16_t au16[40/2];
1302	uint32_t au32[40/4];
1303	} Buf;
1304
1305	/*
1306	* Check the GSO properties of the socket buffer and make sure it fits.
1307	*/
1308	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1309	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1310	{
1311	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1312	return false;
1313	}
1314	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1315	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1316	{
1317	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1318	return false;
1319	}
1320
1321	/*
1322	* Switch on the ethertype.
1323	*/
1324	uEtherType = pSkb->protocol;
1325	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1326	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1327	{
1328	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1329	if (puEtherType)
1330	uEtherType = *puEtherType;
1331	}
1332	switch (uEtherType)
1333	{
1334	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1335	{
1336	unsigned int cbHdr;
1337	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1338	if (RT_UNLIKELY(!pIPv4))
1339	{
1340	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1341	return false;
1342	}
1343
1344	cbHdr = pIPv4->ip_hl * 4;
1345	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1346	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1347	\|\| cbHdr > cbTransport ))
1348	{
1349	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1350	return false;
1351	}
1352	cbTransport -= cbHdr;
1353	offTransport = pSkb->mac_len + cbHdr;
1354	uProtocol = pIPv4->ip_p;
1355	if (uProtocol == RTNETIPV4_PROT_TCP)
1356	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1357	else if (uProtocol == RTNETIPV4_PROT_UDP)
1358	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1359	else /** @todo IPv6: 4to6 tunneling */
1360	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1361	break;
1362	}
1363
1364	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1365	{
1366	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1367	if (RT_UNLIKELY(!pIPv6))
1368	{
1369	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1370	return false;
1371	}
1372
1373	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1374	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1375	uProtocol = pIPv6->ip6_nxt;
1376	/** @todo IPv6: Dig our way out of the other headers. */
1377	if (uProtocol == RTNETIPV4_PROT_TCP)
1378	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1379	else if (uProtocol == RTNETIPV4_PROT_UDP)
1380	enmGsoType = PDMNETWORKGSOTYPE_IPV6_UDP;
1381	else
1382	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1383	break;
1384	}
1385
1386	default:
1387	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1388	return false;
1389	}
1390
1391	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1392	{
1393	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1394	return false;
1395	}
1396
1397	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1398	\|\| offTransport + cbTransport > pSkb->len
1399	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1400	{
1401	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1402	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1403	return false;
1404	}
1405
1406	/*
1407	* Check the TCP/UDP bits.
1408	*/
1409	if (uProtocol == RTNETIPV4_PROT_TCP)
1410	{
1411	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1412	if (RT_UNLIKELY(!pTcp))
1413	{
1414	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1415	return false;
1416	}
1417
1418	cbTransportHdr = pTcp->th_off * 4;
1419	pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1420	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1421	\|\| cbTransportHdr > cbTransport
1422	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1423	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1424	{
1425	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1426	return false;
1427	}
1428
1429	}
1430	else
1431	{
1432	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1433	cbTransportHdr = sizeof(RTNETUDP);
1434	pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1435	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1436	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1437	{
1438	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1439	return false;
1440	}
1441	}
1442
1443	/*
1444	* We're good, init the GSO context.
1445	*/
1446	pGsoCtx->u8Type = enmGsoType;
1447	pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1448	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1449	pGsoCtx->offHdr1 = pSkb->mac_len;
1450	pGsoCtx->offHdr2 = offTransport;
1451	pGsoCtx->u8Unused = 0;
1452
1453	return true;
1454	}
1455
1456	/**
1457	* Forward the socket buffer as a GSO internal network frame.
1458	*
1459	* @returns IPRT status code.
1460	* @param pThis The net filter instance.
1461	* @param pSkb The GSO socket buffer.
1462	* @param fSrc The source.
1463	* @param pGsoCtx Where to return the GSO context on success.
1464	*/
1465	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1466	{
1467	int rc;
1468	unsigned cbExtra;
1469	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb, &cbExtra);
1470	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1471	if (RT_LIKELY(pSG))
1472	{
1473	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cbExtra, cSegs, fSrc, pGsoCtx);
1474
1475	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1476	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1477
1478	vboxNetFltLinuxDestroySG(pSG, pSkb);
1479	rc = VINF_SUCCESS;
1480	}
1481	else
1482	{
1483	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1484	rc = VERR_NO_MEMORY;
1485	}
1486	return rc;
1487	}
1488
1489	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1490
1491	/**
1492	* Worker for vboxNetFltLinuxForwardToIntNet.
1493	*
1494	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1495	* @param pThis The net filter instance.
1496	* @param pBuf The socket buffer.
1497	* @param fSrc The source.
1498	*/
1499	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1500	{
1501	int rc;
1502	unsigned cbExtra;
1503	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf, &cbExtra);
1504	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1505	if (RT_LIKELY(pSG))
1506	{
1507	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cbExtra, cSegs, fSrc, NULL /pGsoCtx/);
1508
1509	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1510	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1511
1512	vboxNetFltLinuxDestroySG(pSG, pBuf);
1513	rc = VINF_SUCCESS;
1514	}
1515	else
1516	{
1517	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1518	rc = VERR_NO_MEMORY;
1519	}
1520	return rc;
1521	}
1522
1523
1524	/**
1525	* I won't disclose what I do, figure it out yourself, including pThis referencing.
1526	*
1527	* @param pThis The net filter instance.
1528	* @param pBuf The socket buffer.
1529	* @param fSrc Where the packet comes from.
1530	*/
1531	static void vboxNetFltLinuxForwardToIntNetInner(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1532	{
1533	#ifdef VBOXNETFLT_WITH_GSO
1534	if (skb_is_gso(pBuf))
1535	{
1536	PDMNETWORKGSO GsoCtx;
1537	Log6(("vboxNetFltLinuxForwardToIntNetInner: skb len=%u data_len=%u truesize=%u next=%p"
1538	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x ip_summed=%d\n",
1539	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next,
1540	skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size,
1541	skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type,
1542	skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, pBuf->ip_summed));
1543
1544	if (RT_LIKELY(fSrc & INTNETTRUNKDIR_HOST))
1545	{
1546	/*
1547	* skb_gso_segment does the following. Do we need to do it as well?
1548	*/
1549	# if RTLNX_VER_MIN(2,6,22)
1550	skb_reset_mac_header(pBuf);
1551	pBuf->mac_len = pBuf->network_header - pBuf->mac_header;
1552	# else
1553	pBuf->mac.raw = pBuf->data;
1554	pBuf->mac_len = pBuf->nh.raw - pBuf->data;
1555	# endif
1556	}
1557
1558	# ifdef VBOXNETFLT_WITH_GSO_RECV
1559	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1560	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1561	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1562	else
1563	# endif /* VBOXNETFLT_WITH_GSO_RECV */
1564	{
1565	/* Need to segment the packet */
1566	struct sk_buff *pNext;
1567	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1568	if (IS_ERR(pSegment))
1569	{
1570	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1571	return;
1572	}
1573
1574	for (; pSegment; pSegment = pNext)
1575	{
1576	Log6(("vboxNetFltLinuxForwardToIntNetInner: segment len=%u data_len=%u truesize=%u next=%p"
1577	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1578	pSegment->len, pSegment->data_len, pSegment->truesize, pSegment->next,
1579	skb_shinfo(pSegment)->nr_frags, skb_shinfo(pSegment)->gso_size,
1580	skb_shinfo(pSegment)->gso_segs, skb_shinfo(pSegment)->gso_type,
1581	skb_shinfo(pSegment)->frag_list, pSegment->pkt_type));
1582	pNext = pSegment->next;
1583	pSegment->next = 0;
1584	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1585	dev_kfree_skb(pSegment);
1586	}
1587	}
1588	}
1589	else
1590	#endif /* VBOXNETFLT_WITH_GSO */
1591	{
1592	Log6(("vboxNetFltLinuxForwardToIntNetInner: ptk_type=%d ip_summed=%d len=%d"
1593	" data_len=%d headroom=%d hdr_len=%d csum_offset=%d\n",
1594	pBuf->pkt_type, pBuf->ip_summed, pBuf->len, pBuf->data_len, skb_headroom(pBuf),
1595	skb_headlen(pBuf), vboxNetFltLinuxGetChecksumStartOffset(pBuf)));
1596	#ifndef VBOXNETFLT_SG_SUPPORT
1597	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1598	{
1599	# if RTLNX_VER_MIN(2,6,19)
1600	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1601	# else
1602	/*
1603	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1604	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1605	* header length from the header itself and reconstruct 'h' pointer
1606	* to TCP (or whatever) header.
1607	*/
1608	unsigned char *tmp = pBuf->h.raw;
1609	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1610	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1611	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1612	/* Restore the original (wrong) pointer. */
1613	pBuf->h.raw = tmp;
1614	# endif
1615	if (rc)
1616	{
1617	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1618	return;
1619	}
1620	}
1621	#endif /* !VBOXNETFLT_SG_SUPPORT */
1622	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1623	}
1624	}
1625
1626
1627	/**
1628	* Temporarily adjust pBuf->data so it always points to the Ethernet header,
1629	* then forward it to the internal network.
1630	*
1631	* @param pThis The net filter instance.
1632	* @param pBuf The socket buffer. This is consumed by this function.
1633	*/
1634	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1635	{
1636	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1637
1638	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1639	{
1640	/*
1641	* The packet came from the wire and the driver has already consumed
1642	* mac header. We need to restore it back. Moreover, after we are
1643	* through with this skb we need to restore its original state!
1644	*/
1645	skb_push(pBuf, pBuf->mac_len);
1646	Log5(("vboxNetFltLinuxForwardToIntNet: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1647	pBuf->mac_len, pBuf->data, skb_mac_header(pBuf), skb_network_header(pBuf)));
1648	}
1649
1650	vboxNetFltLinuxForwardToIntNetInner(pThis, pBuf, fSrc);
1651
1652	/*
1653	* Restore the original state of skb as there are other handlers this skb
1654	* will be provided to.
1655	*/
1656	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1657	skb_pull(pBuf, pBuf->mac_len);
1658
1659	dev_kfree_skb(pBuf);
1660	}
1661
1662
1663	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1664	/**
1665	* Work queue handler that forwards the socket buffers queued by
1666	* vboxNetFltLinuxPacketHandler to the internal network.
1667	*
1668	* @param pWork The work queue.
1669	*/
1670	# if RTLNX_VER_MIN(2,6,20)
1671	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1672	# else
1673	static void vboxNetFltLinuxXmitTask(void *pWork)
1674	# endif
1675	{
1676	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1677	struct sk_buff *pBuf;
1678
1679	Log6(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1680
1681	/*
1682	* Active? Retain the instance and increment the busy counter.
1683	*/
1684	if (vboxNetFltTryRetainBusyActive(pThis))
1685	{
1686	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1687	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1688
1689	vboxNetFltRelease(pThis, true /* fBusy */);
1690	}
1691	else
1692	{
1693	/** @todo Shouldn't we just drop the packets here? There is little point in
1694	* making them accumulate when the VM is paused and it'll only waste
1695	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1696	* before start draining the packets (goes for the intnet ring buf
1697	* too)? */
1698	}
1699	}
1700	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1701
1702	/**
1703	* Reports the GSO capabilities of the hardware NIC.
1704	*
1705	* @param pThis The net filter instance. The caller hold a
1706	* reference to this.
1707	*/
1708	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1709	{
1710	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1711	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1712	{
1713	struct net_device *pDev;
1714	unsigned int fFeatures;
1715
1716	RTSpinlockAcquire(pThis->hSpinlock);
1717
1718	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1719	if (pDev)
1720	fFeatures = pDev->features;
1721	else
1722	fFeatures = 0;
1723
1724	RTSpinlockRelease(pThis->hSpinlock);
1725
1726	if (pThis->pSwitchPort)
1727	{
1728	/* Set/update the GSO capabilities of the NIC. */
1729	uint32_t fGsoCapabilites = 0;
1730	if (fFeatures & NETIF_F_TSO)
1731	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1732	if (fFeatures & NETIF_F_TSO6)
1733	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1734	Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s\n",
1735	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1736	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : ""));
1737	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1738	}
1739
1740	vboxNetFltRelease(pThis, true /fBusy/);
1741	}
1742	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
1743	}
1744
1745	/**
1746	* Helper that determines whether the host (ignoreing us) is operating the
1747	* interface in promiscuous mode or not.
1748	*/
1749	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1750	{
1751	bool fRc = false;
1752	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1753	if (pDev)
1754	{
1755	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1756	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1757	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1758	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1759	}
1760	return fRc;
1761	}
1762
1763	/**
1764	* Does this device needs link state change signaled?
1765	* Currently we need it for our own VBoxNetAdp and TAP.
1766	*/
1767	static bool vboxNetFltNeedsLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev)
1768	{
1769	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1770	{
1771	struct ethtool_drvinfo Info;
1772
1773	memset(&Info, 0, sizeof(Info));
1774	Info.cmd = ETHTOOL_GDRVINFO;
1775	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1776	Log3(("%s: driver=%.s version=%.s bus_info=%.*s\n",
1777	__FUNCTION__,
1778	sizeof(Info.driver), Info.driver,
1779	sizeof(Info.version), Info.version,
1780	sizeof(Info.bus_info), Info.bus_info));
1781
1782	if (!strncmp(Info.driver, "vboxnet", sizeof(Info.driver)))
1783	return true;
1784
1785	#if RTLNX_VER_MIN(2,6,36) /* TAP started doing carrier */
1786	return !strncmp(Info.driver, "tun", 4)
1787	&& !strncmp(Info.bus_info, "tap", 4);
1788	#endif
1789	}
1790
1791	return false;
1792	}
1793
1794	#if RTLNX_VER_MAX(2,6,18)
1795	DECLINLINE(void) netif_tx_lock_bh(struct net_device *pDev)
1796	{
1797	spin_lock_bh(&pDev->xmit_lock);
1798	}
1799
1800	DECLINLINE(void) netif_tx_unlock_bh(struct net_device *pDev)
1801	{
1802	spin_unlock_bh(&pDev->xmit_lock);
1803	}
1804	#endif
1805
1806	/**
1807	* Some devices need link state change when filter attaches/detaches
1808	* since the filter is their link in a sense.
1809	*/
1810	static void vboxNetFltSetLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1811	{
1812	if (vboxNetFltNeedsLinkState(pThis, pDev))
1813	{
1814	Log3(("%s: bringing device link %s\n",
1815	__FUNCTION__, fLinkUp ? "up" : "down"));
1816	netif_tx_lock_bh(pDev);
1817	if (fLinkUp)
1818	netif_carrier_on(pDev);
1819	else
1820	netif_carrier_off(pDev);
1821	netif_tx_unlock_bh(pDev);
1822	}
1823	}
1824
1825	/**
1826	* Internal worker for vboxNetFltLinuxNotifierCallback.
1827	*
1828	* @returns VBox status code.
1829	* @param pThis The instance.
1830	* @param pDev The device to attach to.
1831	*/
1832	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1833	{
1834	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1835
1836	/*
1837	* Retain and store the device.
1838	*/
1839	dev_hold(pDev);
1840
1841	RTSpinlockAcquire(pThis->hSpinlock);
1842	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
1843	RTSpinlockRelease(pThis->hSpinlock);
1844
1845	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
1846	pDev, pDev->name,
1847	#if RTLNX_VER_MIN(2,6,37)
1848	netdev_refcnt_read(pDev)
1849	#else
1850	atomic_read(&pDev->refcnt)
1851	#endif
1852	));
1853	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1854	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1855
1856	/* Get the mac address while we still have a valid net_device reference. */
1857	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
1858	/* Initialize MTU */
1859	pThis->u.s.cbMtu = pDev->mtu;
1860
1861	/*
1862	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1863	*/
1864	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1865	pThis->u.s.PacketType.dev = pDev;
1866	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1867	dev_add_pack(&pThis->u.s.PacketType);
1868	ASMAtomicUoWriteBool(&pThis->u.s.fPacketHandler, true);
1869	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1870
1871	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1872	vboxNetFltLinuxHookDev(pThis, pDev);
1873	#endif
1874
1875	/*
1876	* Are we the "carrier" for this device (e.g. vboxnet or tap)?
1877	*/
1878	vboxNetFltSetLinkState(pThis, pDev, true);
1879
1880	/*
1881	* Set indicators that require the spinlock. Be abit paranoid about racing
1882	* the device notification handle.
1883	*/
1884	RTSpinlockAcquire(pThis->hSpinlock);
1885	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1886	if (pDev)
1887	{
1888	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1889	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1890	pDev = NULL; /* don't dereference it */
1891	}
1892	RTSpinlockRelease(pThis->hSpinlock);
1893
1894	/*
1895	* Report GSO capabilities
1896	*/
1897	Assert(pThis->pSwitchPort);
1898	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1899	{
1900	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1901	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
1902	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
1903	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
1904	vboxNetFltRelease(pThis, true /fBusy/);
1905	}
1906
1907	LogRel(("VBoxNetFlt: attached to '%s' / %RTmac\n", pThis->szName, &pThis->u.s.MacAddr));
1908	return VINF_SUCCESS;
1909	}
1910
1911
1912	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1913	{
1914	bool fRegistered;
1915	Assert(!pThis->fDisconnectedFromHost);
1916
1917	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1918	vboxNetFltLinuxUnhookDev(pThis, pDev);
1919	#endif
1920
1921	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
1922	{
1923	dev_remove_pack(&pThis->u.s.PacketType);
1924	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: packet handler removed.\n", pThis));
1925	}
1926
1927	RTSpinlockAcquire(pThis->hSpinlock);
1928	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
1929	if (fRegistered)
1930	{
1931	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1932	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1933	}
1934	RTSpinlockRelease(pThis->hSpinlock);
1935
1936	if (fRegistered)
1937	{
1938	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1939	skb_queue_purge(&pThis->u.s.XmitQueue);
1940	#endif
1941	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: xmit queue purged.\n", pThis));
1942	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
1943	pDev, pDev->name,
1944	#if RTLNX_VER_MIN(2,6,37)
1945	netdev_refcnt_read(pDev)
1946	#else
1947	atomic_read(&pDev->refcnt)
1948	#endif
1949	));
1950	dev_put(pDev);
1951	}
1952
1953	return NOTIFY_OK;
1954	}
1955
1956	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1957	{
1958	/* Check if we are not suspended and promiscuous mode has not been set. */
1959	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
1960	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1961	{
1962	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1963	dev_set_promiscuity(pDev, 1);
1964	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1965	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1966	}
1967	else
1968	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1969	return NOTIFY_OK;
1970	}
1971
1972	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1973	{
1974	/* Undo promiscuous mode if we has set it. */
1975	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1976	{
1977	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1978	dev_set_promiscuity(pDev, -1);
1979	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1980	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1981	}
1982	else
1983	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1984	return NOTIFY_OK;
1985	}
1986
1987	/**
1988	* Callback for listening to MTU change event.
1989	*
1990	* We need to track changes of host's inteface MTU to discard over-sized frames
1991	* coming from the internal network as they may hang the TX queue of host's
1992	* adapter.
1993	*
1994	* @returns NOTIFY_OK
1995	* @param pThis The netfilter instance.
1996	* @param pDev Pointer to device structure of host's interface.
1997	*/
1998	static int vboxNetFltLinuxDeviceMtuChange(PVBOXNETFLTINS pThis, struct net_device *pDev)
1999	{
2000	ASMAtomicWriteU32(&pThis->u.s.cbMtu, pDev->mtu);
2001	Log(("vboxNetFltLinuxDeviceMtuChange: set MTU for %s to %d\n", pThis->szName, pDev->mtu));
2002	return NOTIFY_OK;
2003	}
2004
2005	#ifdef LOG_ENABLED
2006	/** Stringify the NETDEV_XXX constants. */
2007	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2008	{
2009	const char *pszEvent = "NETDEV_<unknown>";
2010	switch (ulEventType)
2011	{
2012	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2013	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2014	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2015	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2016	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2017	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2018	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2019	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2020	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2021	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2022	# ifdef NETDEV_FEAT_CHANGE
2023	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2024	# endif
2025	}
2026	return pszEvent;
2027	}
2028	#endif /* LOG_ENABLED */
2029
2030	/**
2031	* Callback for listening to netdevice events.
2032	*
2033	* This works the rediscovery, clean up on unregistration, promiscuity on
2034	* up/down, and GSO feature changes from ethtool.
2035	*
2036	* @returns NOTIFY_OK
2037	* @param self Pointer to our notifier registration block.
2038	* @param ulEventType The event.
2039	* @param ptr Event specific, but it is usually the device it
2040	* relates to.
2041	*/
2042	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2043
2044	{
2045	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2046	struct net_device pMyDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device );
2047	struct net_device *pDev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2048	int rc = NOTIFY_OK;
2049
2050	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2051	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, pMyDev));
2052
2053	if (ulEventType == NETDEV_REGISTER)
2054	{
2055	#if RTLNX_VER_MIN(2,6,24) /* cgroups/namespaces introduced */
2056	# if RTLNX_VER_MIN(2,6,26)
2057	# define VBOX_DEV_NET(dev) dev_net(dev)
2058	# define VBOX_NET_EQ(n1, n2) net_eq((n1), (n2))
2059	# else
2060	# define VBOX_DEV_NET(dev) ((dev)->nd_net)
2061	# define VBOX_NET_EQ(n1, n2) ((n1) == (n2))
2062	# endif
2063	struct net *pMyNet = current->nsproxy->net_ns;
2064	struct net *pDevNet = VBOX_DEV_NET(pDev);
2065
2066	if (VBOX_NET_EQ(pDevNet, pMyNet))
2067	#endif /* namespaces */
2068	{
2069	if (strcmp(pDev->name, pThis->szName) == 0)
2070	{
2071	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2072	}
2073	}
2074	}
2075	else
2076	{
2077	if (pDev == pMyDev)
2078	{
2079	switch (ulEventType)
2080	{
2081	case NETDEV_UNREGISTER:
2082	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2083	break;
2084	case NETDEV_UP:
2085	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2086	break;
2087	case NETDEV_GOING_DOWN:
2088	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2089	break;
2090	case NETDEV_CHANGEMTU:
2091	rc = vboxNetFltLinuxDeviceMtuChange(pThis, pDev);
2092	break;
2093	case NETDEV_CHANGENAME:
2094	break;
2095	#ifdef NETDEV_FEAT_CHANGE
2096	case NETDEV_FEAT_CHANGE:
2097	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2098	break;
2099	#endif
2100	}
2101	}
2102	}
2103
2104	return rc;
2105	}
2106
2107	/*
2108	* Initial enumeration of netdevs. Called with NETDEV_REGISTER by
2109	* register_netdevice_notifier() under rtnl lock.
2110	*/
2111	static int vboxNetFltLinuxEnumeratorCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2112	{
2113	PVBOXNETFLTINS pThis = ((PVBOXNETFLTNOTIFIER)self)->pThis;
2114	struct net_device *dev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2115	struct in_device *in_dev;
2116	struct inet6_dev *in6_dev;
2117
2118	if (ulEventType != NETDEV_REGISTER)
2119	return NOTIFY_OK;
2120
2121	if (RT_UNLIKELY(pThis->pSwitchPort->pfnNotifyHostAddress == NULL))
2122	return NOTIFY_OK;
2123
2124	/*
2125	* IPv4
2126	*/
2127	#if RTLNX_VER_MIN(2,6,14)
2128	in_dev = __in_dev_get_rtnl(dev);
2129	#else
2130	in_dev = __in_dev_get(dev);
2131	#endif
2132	if (in_dev != NULL)
2133	{
2134	struct in_ifaddr *ifa;
2135
2136	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2137	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2138	return NOTIFY_OK;
2139
2140	if ( dev != pThis->u.s.pDev
2141	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2142	continue;
2143
2144	Log(("%s: %s: IPv4 addr %RTnaipv4 mask %RTnaipv4\n",
2145	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2146	ifa->ifa_address, ifa->ifa_mask));
2147
2148	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2149	/* :fAdded */ true, kIntNetAddrType_IPv4, &ifa->ifa_address);
2150	}
2151	}
2152
2153	/*
2154	* IPv6
2155	*/
2156	in6_dev = __in6_dev_get(dev);
2157	if (in6_dev != NULL)
2158	{
2159	struct inet6_ifaddr *ifa;
2160
2161	read_lock_bh(&in6_dev->lock);
2162	#if RTLNX_VER_MIN(2,6,35)
2163	list_for_each_entry(ifa, &in6_dev->addr_list, if_list)
2164	#else
2165	for (ifa = in6_dev->addr_list; ifa != NULL; ifa = ifa->if_next)
2166	#endif
2167	{
2168	if ( dev != pThis->u.s.pDev
2169	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2170	continue;
2171
2172	Log(("%s: %s: IPv6 addr %RTnaipv6/%u\n",
2173	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2174	&ifa->addr, (unsigned)ifa->prefix_len));
2175
2176	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2177	/* :fAdded */ true, kIntNetAddrType_IPv6, &ifa->addr);
2178	}
2179	read_unlock_bh(&in6_dev->lock);
2180	}
2181
2182	return NOTIFY_OK;
2183	}
2184
2185
2186	static int vboxNetFltLinuxNotifierIPv4Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2187	{
2188	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv4);
2189	struct net_device pDev, pEventDev;
2190	struct in_ifaddr ifa = (struct in_ifaddr )ptr;
2191	bool fMyDev;
2192	int rc = NOTIFY_OK;
2193
2194	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2195	pEventDev = ifa->ifa_dev->dev;
2196	fMyDev = (pDev == pEventDev);
2197	Log(("VBoxNetFlt: %s: IPv4 event %s(0x%lx) %s: addr %RTnaipv4 mask %RTnaipv4\n",
2198	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2199	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2200	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2201	ifa->ifa_address, ifa->ifa_mask));
2202
2203	if (pDev != NULL)
2204	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2205
2206	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2207	return NOTIFY_OK;
2208
2209	if ( !fMyDev
2210	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2211	return NOTIFY_OK;
2212
2213	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2214	{
2215	bool fAdded;
2216	if (ulEventType == NETDEV_UP)
2217	fAdded = true;
2218	else if (ulEventType == NETDEV_DOWN)
2219	fAdded = false;
2220	else
2221	return NOTIFY_OK;
2222
2223	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2224	kIntNetAddrType_IPv4, &ifa->ifa_local);
2225	}
2226
2227	return rc;
2228	}
2229
2230
2231	static int vboxNetFltLinuxNotifierIPv6Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2232	{
2233	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv6);
2234	struct net_device pDev, pEventDev;
2235	struct inet6_ifaddr ifa = (struct inet6_ifaddr )ptr;
2236	bool fMyDev;
2237	int rc = NOTIFY_OK;
2238
2239	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2240	pEventDev = ifa->idev->dev;
2241	fMyDev = (pDev == pEventDev);
2242	Log(("VBoxNetFlt: %s: IPv6 event %s(0x%lx) %s: %RTnaipv6\n",
2243	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2244	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2245	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2246	&ifa->addr));
2247
2248	if (pDev != NULL)
2249	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2250
2251	if ( !fMyDev
2252	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2253	return NOTIFY_OK;
2254
2255	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2256	{
2257	bool fAdded;
2258	if (ulEventType == NETDEV_UP)
2259	fAdded = true;
2260	else if (ulEventType == NETDEV_DOWN)
2261	fAdded = false;
2262	else
2263	return NOTIFY_OK;
2264
2265	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2266	kIntNetAddrType_IPv6, &ifa->addr);
2267	}
2268
2269	return rc;
2270	}
2271
2272
2273	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2274	{
2275	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2276	}
2277
2278	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2279	{
2280	struct net_device * pDev;
2281	int err;
2282	int rc = VINF_SUCCESS;
2283	IPRT_LINUX_SAVE_EFL_AC();
2284	NOREF(pvIfData);
2285
2286	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2287
2288	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2289	if (pDev)
2290	{
2291	/*
2292	* Create a sk_buff for the gather list and push it onto the wire.
2293	*/
2294	if (fDst & INTNETTRUNKDIR_WIRE)
2295	{
2296	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2297	if (pBuf)
2298	{
2299	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2300	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2301	Log6(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2302	err = dev_queue_xmit(pBuf);
2303	if (err)
2304	rc = RTErrConvertFromErrno(err);
2305	}
2306	else
2307	rc = VERR_NO_MEMORY;
2308	}
2309
2310	/*
2311	* Create a sk_buff for the gather list and push it onto the host stack.
2312	*/
2313	if (fDst & INTNETTRUNKDIR_HOST)
2314	{
2315	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2316	if (pBuf)
2317	{
2318	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2319	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2320	Log6(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2321	#if RTLNX_VER_MIN(5,18,0) \|\| RTLNX_RHEL_MIN(9,1)
2322	local_bh_disable();
2323	err = netif_rx(pBuf);
2324	local_bh_enable();
2325	#else
2326	err = netif_rx_ni(pBuf);
2327	#endif
2328	if (err)
2329	rc = RTErrConvertFromErrno(err);
2330	}
2331	else
2332	rc = VERR_NO_MEMORY;
2333	}
2334
2335	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2336	}
2337
2338	IPRT_LINUX_RESTORE_EFL_AC();
2339	return rc;
2340	}
2341
2342
2343	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2344	{
2345	struct net_device *pDev;
2346	IPRT_LINUX_SAVE_EFL_AC();
2347
2348	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%RTbool, fDisablePromiscuous=%RTbool\n",
2349	pThis, pThis->szName, fActive, pThis->fDisablePromiscuous));
2350
2351	if (pThis->fDisablePromiscuous)
2352	return;
2353
2354	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2355	if (pDev)
2356	{
2357	/*
2358	* This api is a bit weird, the best reference is the code.
2359	*
2360	* Also, we have a bit or race conditions wrt the maintenance of
2361	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2362	*/
2363	#ifdef LOG_ENABLED
2364	u_int16_t fIf;
2365	unsigned const cPromiscBefore = pDev->promiscuity;
2366	#endif
2367	if (fActive)
2368	{
2369	Assert(!pThis->u.s.fPromiscuousSet);
2370
2371	rtnl_lock();
2372	dev_set_promiscuity(pDev, 1);
2373	rtnl_unlock();
2374	pThis->u.s.fPromiscuousSet = true;
2375	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2376	}
2377	else
2378	{
2379	if (pThis->u.s.fPromiscuousSet)
2380	{
2381	rtnl_lock();
2382	dev_set_promiscuity(pDev, -1);
2383	rtnl_unlock();
2384	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2385	}
2386	pThis->u.s.fPromiscuousSet = false;
2387
2388	#ifdef LOG_ENABLED
2389	fIf = dev_get_flags(pDev);
2390	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2391	#endif
2392	}
2393
2394	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2395	}
2396	IPRT_LINUX_RESTORE_EFL_AC();
2397	}
2398
2399
2400	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2401	{
2402	/*
2403	* Remove packet handler when we get disconnected from internal switch as
2404	* we don't want the handler to forward packets to disconnected switch.
2405	*/
2406	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
2407	{
2408	IPRT_LINUX_SAVE_EFL_AC();
2409	dev_remove_pack(&pThis->u.s.PacketType);
2410	Log(("vboxNetFltOsDisconnectIt: this=%p: Packet handler removed.\n", pThis));
2411	IPRT_LINUX_RESTORE_EFL_AC();
2412	}
2413	return VINF_SUCCESS;
2414	}
2415
2416
2417	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2418	{
2419	IPRT_LINUX_SAVE_EFL_AC();
2420
2421	/*
2422	* Report the GSO capabilities of the host and device (if connected).
2423	* Note! No need to mark ourselves busy here.
2424	*/
2425	/** @todo duplicate work here now? Attach */
2426	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2427	Log3(("vboxNetFltOsConnectIt: reporting host tso tso6\n"));
2428	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2429	0
2430	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2431	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2432	, INTNETTRUNKDIR_HOST);
2433
2434	#endif
2435	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2436
2437	IPRT_LINUX_RESTORE_EFL_AC();
2438	return VINF_SUCCESS;
2439	}
2440
2441
2442	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2443	{
2444	struct net_device *pDev;
2445	bool fRegistered;
2446	IPRT_LINUX_SAVE_EFL_AC();
2447
2448	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
2449	vboxNetFltLinuxUnhookDev(pThis, NULL);
2450	#endif
2451
2452	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2453	* unlikely, but none the less). Since it doesn't actually update the
2454	* state (just reads it), it is likely to panic in some interesting
2455	* ways. */
2456
2457	RTSpinlockAcquire(pThis->hSpinlock);
2458	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2459	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
2460	RTSpinlockRelease(pThis->hSpinlock);
2461
2462	if (fRegistered)
2463	{
2464	vboxNetFltSetLinkState(pThis, pDev, false);
2465
2466	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2467	skb_queue_purge(&pThis->u.s.XmitQueue);
2468	#endif
2469	Log(("vboxNetFltOsDeleteInstance: this=%p: xmit queue purged.\n", pThis));
2470	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2471	pDev, pDev->name,
2472	#if RTLNX_VER_MIN(2,6,37)
2473	netdev_refcnt_read(pDev)
2474	#else
2475	atomic_read(&pDev->refcnt)
2476	#endif
2477	));
2478	dev_put(pDev);
2479	}
2480
2481	unregister_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2482	unregister_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2483
2484	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2485	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2486	module_put(THIS_MODULE);
2487
2488	IPRT_LINUX_RESTORE_EFL_AC();
2489	}
2490
2491
2492	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2493	{
2494	int err;
2495	IPRT_LINUX_SAVE_EFL_AC();
2496	NOREF(pvContext);
2497
2498	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2499	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2500	if (err)
2501	{
2502	IPRT_LINUX_RESTORE_EFL_AC();
2503	return VERR_INTNET_FLT_IF_FAILED;
2504	}
2505	if (!pThis->u.s.fRegistered)
2506	{
2507	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2508	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2509	IPRT_LINUX_RESTORE_EFL_AC();
2510	return VERR_INTNET_FLT_IF_NOT_FOUND;
2511	}
2512
2513	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2514	if ( pThis->fDisconnectedFromHost
2515	\|\| !try_module_get(THIS_MODULE))
2516	{
2517	IPRT_LINUX_RESTORE_EFL_AC();
2518	return VERR_INTNET_FLT_IF_FAILED;
2519	}
2520
2521	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2522	{
2523	VBOXNETFLTNOTIFIER Enumerator;
2524
2525	/*
2526	* register_inetaddr_notifier() and register_inet6addr_notifier()
2527	* do not call the callback for existing devices. Enumerating
2528	* all network devices explicitly is a bit of an ifdef mess,
2529	* so co-opt register_netdevice_notifier() to do that for us.
2530	*/
2531	RT_ZERO(Enumerator);
2532	Enumerator.Notifier.notifier_call = vboxNetFltLinuxEnumeratorCallback;
2533	Enumerator.pThis = pThis;
2534
2535	err = register_netdevice_notifier(&Enumerator.Notifier);
2536	if (err)
2537	{
2538	LogRel(("%s: failed to enumerate network devices: error %d\n", __FUNCTION__, err));
2539	IPRT_LINUX_RESTORE_EFL_AC();
2540	return VINF_SUCCESS;
2541	}
2542
2543	unregister_netdevice_notifier(&Enumerator.Notifier);
2544
2545	pThis->u.s.NotifierIPv4.notifier_call = vboxNetFltLinuxNotifierIPv4Callback;
2546	err = register_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2547	if (err)
2548	LogRel(("%s: failed to register IPv4 notifier: error %d\n", __FUNCTION__, err));
2549
2550	pThis->u.s.NotifierIPv6.notifier_call = vboxNetFltLinuxNotifierIPv6Callback;
2551	err = register_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2552	if (err)
2553	LogRel(("%s: failed to register IPv6 notifier: error %d\n", __FUNCTION__, err));
2554	}
2555
2556	IPRT_LINUX_RESTORE_EFL_AC();
2557	return VINF_SUCCESS;
2558	}
2559
2560	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2561	{
2562	IPRT_LINUX_SAVE_EFL_AC();
2563
2564	/*
2565	* Init the linux specific members.
2566	*/
2567	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2568	pThis->u.s.fRegistered = false;
2569	pThis->u.s.fPromiscuousSet = false;
2570	pThis->u.s.fPacketHandler = false;
2571	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2572	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2573	skb_queue_head_init(&pThis->u.s.XmitQueue);
2574	# if RTLNX_VER_MIN(2,6,20)
2575	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2576	# else
2577	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2578	# endif
2579	#endif
2580
2581	IPRT_LINUX_RESTORE_EFL_AC();
2582	return VINF_SUCCESS;
2583	}
2584
2585
2586	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2587	{
2588	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2589	}
2590
2591
2592	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2593	{
2594	/* Nothing to do */
2595	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2596	return VINF_SUCCESS;
2597	}
2598
2599
2600	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2601	{
2602	/* Nothing to do */
2603	NOREF(pThis); NOREF(pvIfData);
2604	return VINF_SUCCESS;
2605	}
2606

注意: 瀏覽 TracBrowser 來幫助您使用儲存庫瀏覽器

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 99941

以其他格式下載: