VBoxNetFlt-linux.c@ 100862

Last change on this file since 100862 was 100862, checked in by vboxsync, 16 months ago
Linux: Host drivers: VBoxNetFlt: Build fix for kernel 6.4.10, bugref:10441, bugref:10482.
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 100862 2023-08-11 17:16:38Z 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.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	#if RTLNX_VER_MIN(6,4,10)
51	# include <net/gso.h>
52	#endif
53	#include <linux/netdevice.h>
54	#if RTLNX_VER_MAX(2,6,29) \|\| RTLNX_VER_MIN(5,11,0)
55	# include <linux/ethtool.h>
56	#endif
57	#include <linux/etherdevice.h>
58	#include <linux/rtnetlink.h>
59	#include <linux/miscdevice.h>
60	#include <linux/inetdevice.h>
61	#include <linux/in.h>
62	#include <linux/ip.h>
63	#include <linux/if_vlan.h>
64	#if RTLNX_VER_MIN(4,5,0)
65	# include <uapi/linux/pkt_cls.h>
66	#endif
67	#include <net/ipv6.h>
68	#include <net/if_inet6.h>
69	#include <net/addrconf.h>
70
71	#include <VBox/log.h>
72	#include <VBox/err.h>
73	#include <VBox/intnetinline.h>
74	#include <VBox/vmm/pdmnetinline.h>
75	#include <VBox/param.h>
76	#include <iprt/alloca.h>
77	#include <iprt/assert.h>
78	#include <iprt/spinlock.h>
79	#include <iprt/semaphore.h>
80	#include <iprt/initterm.h>
81	#include <iprt/process.h>
82	#include <iprt/mem.h>
83	#include <iprt/net.h>
84	#include <iprt/log.h>
85	#include <iprt/mp.h>
86	#include <iprt/mem.h>
87	#include <iprt/time.h>
88
89	#define VBOXNETFLT_OS_SPECFIC 1
90	#include "../VBoxNetFltInternal.h"
91
92	typedef struct VBOXNETFLTNOTIFIER {
93	struct notifier_block Notifier;
94	PVBOXNETFLTINS pThis;
95	} VBOXNETFLTNOTIFIER;
96	typedef struct VBOXNETFLTNOTIFIER *PVBOXNETFLTNOTIFIER;
97
98
99	/*********************************************************************************************************************************
100	* Defined Constants And Macros *
101	*********************************************************************************************************************************/
102	#define VBOX_FLT_NB_TO_INST(pNB) RT_FROM_MEMBER(pNB, VBOXNETFLTINS, u.s.Notifier)
103	#define VBOX_FLT_PT_TO_INST(pPT) RT_FROM_MEMBER(pPT, VBOXNETFLTINS, u.s.PacketType)
104	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
105	# define VBOX_FLT_XT_TO_INST(pXT) RT_FROM_MEMBER(pXT, VBOXNETFLTINS, u.s.XmitTask)
106	#endif
107
108	#if RTLNX_VER_MIN(3,11,0)
109	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) netdev_notifier_info_to_dev(ptr)
110	#else
111	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) ((struct net_device *)ptr)
112	#endif
113
114	#if RTLNX_VER_MIN(3,5,0)
115	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag))
116	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr)
117	#else
118	# if RTLNX_VER_MIN(3,2,0)
119	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag), KM_SKB_DATA_SOFTIRQ)
120	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
121	# else
122	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ)
123	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
124	# endif
125	#endif
126
127	#if RTLNX_VER_MIN(2,6,34)
128	# define VBOX_NETDEV_NAME(dev) netdev_name(dev)
129	#else
130	# define VBOX_NETDEV_NAME(dev) ((dev)->reg_state != NETREG_REGISTERED ? "(unregistered net_device)" : (dev)->name)
131	#endif
132
133	#if RTLNX_VER_MIN(2,6,25)
134	# define VBOX_IPV4_IS_LOOPBACK(addr) ipv4_is_loopback(addr)
135	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ipv4_is_linklocal_169(addr)
136	#else
137	# define VBOX_IPV4_IS_LOOPBACK(addr) ((addr & htonl(IN_CLASSA_NET)) == htonl(0x7f000000))
138	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ((addr & htonl(IN_CLASSB_NET)) == htonl(0xa9fe0000))
139	#endif
140
141	#if RTLNX_VER_MIN(2,6,22)
142	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
143	# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
144	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum_offset
145	#else
146	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
147	# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
148	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum
149	#endif
150
151	#if RTLNX_VER_MIN(2,6,19)
152	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
153	#else
154	# define CHECKSUM_PARTIAL CHECKSUM_HW
155	# if RTLNX_VER_MIN(2,6,10)
156	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
157	# else
158	# if RTLNX_VER_MIN(2,6,7)
159	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
160	# else
161	# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
162	# endif
163	/* Versions prior 2.6.10 use stats for both bstats and qstats */
164	# define bstats stats
165	# define qstats stats
166	# endif
167	#endif
168
169	#if RTLNX_VER_MIN(3,20,0) \|\| RTLNX_RHEL_RANGE(7,2, 8,0) \|\| RTLNX_RHEL_RANGE(6,8, 7,0)
170	# define VBOX_HAVE_SKB_VLAN
171	#endif
172
173	#ifdef VBOX_HAVE_SKB_VLAN
174	# define vlan_tx_tag_get(skb) skb_vlan_tag_get(skb)
175	# define vlan_tx_tag_present(skb) skb_vlan_tag_present(skb)
176	#endif
177
178	#ifndef NET_IP_ALIGN
179	# define NET_IP_ALIGN 2
180	#endif
181
182	#if 1
183	/** Create scatter / gather segments for fragments. When not used, we will
184	* linearize the socket buffer before creating the internal networking SG. */
185	# define VBOXNETFLT_SG_SUPPORT 1
186	#endif
187
188	#if RTLNX_VER_MIN(2,6,18)
189
190	/** Indicates that the linux kernel may send us GSO frames. */
191	# define VBOXNETFLT_WITH_GSO 1
192
193	/** This enables or disables the transmitting of GSO frame from the internal
194	* network and to the host. */
195	# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
196
197	# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
198	/** This enables or disables the transmitting of GSO frame from the internal
199	* network and to the wire. */
200	# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
201	# endif
202
203	/** This enables or disables the forwarding/flooding of GSO frame from the host
204	* to the internal network. */
205	# define VBOXNETFLT_WITH_GSO_RECV 1
206
207	#endif /* RTLNX_VER_MIN(2,6,18) */
208
209	#if RTLNX_VER_MIN(2,6,29)
210	/** This enables or disables handling of GSO frames coming from the wire (GRO). */
211	# define VBOXNETFLT_WITH_GRO 1
212	#endif
213
214	/*
215	* GRO support was backported to RHEL 5.4
216	*/
217	#if RTLNX_RHEL_MAJ_PREREQ(5, 4)
218	# define VBOXNETFLT_WITH_GRO 1
219	#endif
220
221
222	/*********************************************************************************************************************************
223	* Internal Functions *
224	*********************************************************************************************************************************/
225	static int __init VBoxNetFltLinuxInit(void);
226	static void __exit VBoxNetFltLinuxUnload(void);
227	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf);
228
229
230	/*********************************************************************************************************************************
231	* Global Variables *
232	*********************************************************************************************************************************/
233	/**
234	* The (common) global data.
235	*/
236	static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
237
238	module_init(VBoxNetFltLinuxInit);
239	module_exit(VBoxNetFltLinuxUnload);
240
241	MODULE_AUTHOR(VBOX_VENDOR);
242	MODULE_DESCRIPTION(VBOX_PRODUCT " Network Filter Driver");
243	MODULE_LICENSE("GPL");
244	#ifdef MODULE_VERSION
245	MODULE_VERSION(VBOX_VERSION_STRING " r" RT_XSTR(VBOX_SVN_REV) " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
246	#endif
247
248
249	#if RTLNX_VER_MAX(2,6,12) && defined(LOG_ENABLED)
250	unsigned dev_get_flags(const struct net_device *dev)
251	{
252	unsigned flags;
253
254	flags = (dev->flags & ~(IFF_PROMISC \|
255	IFF_ALLMULTI \|
256	IFF_RUNNING)) \|
257	(dev->gflags & (IFF_PROMISC \|
258	IFF_ALLMULTI));
259
260	if (netif_running(dev) && netif_carrier_ok(dev))
261	flags \|= IFF_RUNNING;
262
263	return flags;
264	}
265	#endif /* RTLNX_VER_MAX(2,6,12) */
266
267
268	/**
269	* Initialize module.
270	*
271	* @returns appropriate status code.
272	*/
273	static int __init VBoxNetFltLinuxInit(void)
274	{
275	int rc;
276	/*
277	* Initialize IPRT.
278	*/
279	rc = RTR0Init(0);
280	if (RT_SUCCESS(rc))
281	{
282	Log(("VBoxNetFltLinuxInit\n"));
283
284	/*
285	* Initialize the globals and connect to the support driver.
286	*
287	* This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
288	* for establishing the connect to the support driver.
289	*/
290	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
291	rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
292	if (RT_SUCCESS(rc))
293	{
294	LogRel(("VBoxNetFlt: Successfully started.\n"));
295	return 0;
296	}
297
298	LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
299	RTR0Term();
300	}
301	else
302	LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
303
304	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
305	return -RTErrConvertToErrno(rc);
306	}
307
308
309	/**
310	* Unload the module.
311	*
312	* @todo We have to prevent this if we're busy!
313	*/
314	static void __exit VBoxNetFltLinuxUnload(void)
315	{
316	int rc;
317	Log(("VBoxNetFltLinuxUnload\n"));
318	Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
319
320	/*
321	* Undo the work done during start (in reverse order).
322	*/
323	rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
324	AssertRC(rc); NOREF(rc);
325
326	RTR0Term();
327
328	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
329
330	Log(("VBoxNetFltLinuxUnload - done\n"));
331	}
332
333
334	/**
335	* We filter traffic from the host to the internal network
336	* before it reaches the NIC driver.
337	*
338	* The current code uses a very ugly hack overriding hard_start_xmit
339	* callback in the device structure, but it has been shown to give us a
340	* performance boost of 60-100% though. Eventually we have to find some
341	* less hacky way of getting this job done.
342	*/
343	#define VBOXNETFLT_WITH_HOST2WIRE_FILTER
344
345	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
346
347	# if RTLNX_VER_MAX(2,6,29)
348
349	typedef struct ethtool_ops OVR_OPSTYPE;
350	# define OVR_OPS ethtool_ops
351	# define OVR_XMIT pfnStartXmit
352
353	# else /* RTLNX_VER_MIN(2,6,29) */
354
355	typedef struct net_device_ops OVR_OPSTYPE;
356	# define OVR_OPS netdev_ops
357	# define OVR_XMIT pOrgOps->ndo_start_xmit
358
359	# endif /* RTLNX_VER_MIN(2,6,29) */
360
361	/**
362	* The overridden net_device_ops of the device we're attached to.
363	*
364	* As there is no net_device_ops structure in pre-2.6.29 kernels we override
365	* ethtool_ops instead along with hard_start_xmit callback in net_device
366	* structure.
367	*
368	* This is a very dirty hack that was created to explore how much we can improve
369	* the host to guest transfers by not CC'ing the NIC. It turns out to be
370	* the only way to filter outgoing packets for devices without TX queue.
371	*/
372	typedef struct VBoxNetDeviceOpsOverride
373	{
374	/** Our overridden ops. */
375	OVR_OPSTYPE Ops;
376	/** Magic word. */
377	uint32_t u32Magic;
378	/** Pointer to the original ops. */
379	OVR_OPSTYPE const *pOrgOps;
380	# if RTLNX_VER_MAX(2,6,29)
381	/** Pointer to the original hard_start_xmit function. */
382	int (pfnStartXmit)(struct sk_buff pSkb, struct net_device *pDev);
383	# endif /* RTLNX_VER_MAX(2,6,29) */
384	/** Pointer to the net filter instance. */
385	PVBOXNETFLTINS pVBoxNetFlt;
386	/** The number of filtered packages. */
387	uint64_t cFiltered;
388	/** The total number of packets */
389	uint64_t cTotal;
390	} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
391	/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
392	#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
393
394	/**
395	* ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
396	* because they belong on the internal network.
397	*
398	* @returns NETDEV_TX_XXX.
399	* @param pSkb The socket buffer to transmit.
400	* @param pDev The net device.
401	*/
402	static int vboxNetFltLinuxStartXmitFilter(struct sk_buff pSkb, struct net_device pDev)
403	{
404	PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
405	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
406	PCRTNETETHERHDR pEtherHdr;
407	PINTNETTRUNKSWPORT pSwitchPort;
408	uint32_t cbHdrs;
409
410
411	/*
412	* Validate the override structure.
413	*
414	* Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
415	* to be production quality code, we would have to be much more
416	* careful here and avoid the race.
417	*/
418	if ( !RT_VALID_PTR(pOverride)
419	\|\| pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
420	# if RTLNX_VER_MIN(2,6,29)
421	\|\| !RT_VALID_PTR(pOverride->pOrgOps)
422	# endif
423	)
424	{
425	printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
426	dev_kfree_skb(pSkb);
427	return NETDEV_TX_OK;
428	}
429	pOverride->cTotal++;
430
431	/*
432	* Do the filtering base on the default OUI of our virtual NICs
433	*
434	* Note! In a real solution, we would ask the switch whether the
435	* destination MAC is 100% to be on the internal network and then
436	* drop it.
437	*/
438	cbHdrs = skb_headlen(pSkb);
439	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
440	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
441	if ( pEtherHdr
442	&& RT_VALID_PTR(pOverride->pVBoxNetFlt)
443	&& (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
444	&& RT_VALID_PTR(pSwitchPort)
445	&& cbHdrs >= 6)
446	{
447	INTNETSWDECISION enmDecision;
448
449	/** @todo consider reference counting, etc. */
450	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
451	if (enmDecision == INTNETSWDECISION_INTNET)
452	{
453	dev_kfree_skb(pSkb);
454	pOverride->cFiltered++;
455	return NETDEV_TX_OK;
456	}
457	}
458
459	return pOverride->OVR_XMIT(pSkb, pDev);
460	}
461
462	/**
463	* Hooks the device ndo_start_xmit operation of the device.
464	*
465	* @param pThis The net filter instance.
466	* @param pDev The net device.
467	*/
468	static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
469	{
470	PVBOXNETDEVICEOPSOVERRIDE pOverride;
471
472	/* Cancel override if ethtool_ops is missing (host-only case, @bugref{5712}) */
473	if (!RT_VALID_PTR(pDev->OVR_OPS))
474	return;
475	pOverride = RTMemAlloc(sizeof(*pOverride));
476	if (!pOverride)
477	return;
478	pOverride->pOrgOps = pDev->OVR_OPS;
479	pOverride->Ops = *pDev->OVR_OPS;
480	# if RTLNX_VER_MAX(2,6,29)
481	pOverride->pfnStartXmit = pDev->hard_start_xmit;
482	# else /* RTLNX_VER_MIN(2,6,29) */
483	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
484	# endif /* RTLNX_VER_MIN(2,6,29) */
485	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
486	pOverride->cTotal = 0;
487	pOverride->cFiltered = 0;
488	pOverride->pVBoxNetFlt = pThis;
489
490	RTSpinlockAcquire(pThis->hSpinlock); /* (this isn't necessary, but so what) */
491	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
492	# if RTLNX_VER_MAX(2,6,29)
493	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
494	# endif /* RTLNX_VER_MAX(2,6,29) */
495	RTSpinlockRelease(pThis->hSpinlock);
496	}
497
498	/**
499	* Undos what vboxNetFltLinuxHookDev did.
500	*
501	* @param pThis The net filter instance.
502	* @param pDev The net device. Can be NULL, in which case
503	* we'll try retrieve it from @a pThis.
504	*/
505	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
506	{
507	PVBOXNETDEVICEOPSOVERRIDE pOverride;
508
509	RTSpinlockAcquire(pThis->hSpinlock);
510	if (!pDev)
511	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
512	if (RT_VALID_PTR(pDev))
513	{
514	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
515	if ( RT_VALID_PTR(pOverride)
516	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
517	&& RT_VALID_PTR(pOverride->pOrgOps)
518	)
519	{
520	# if RTLNX_VER_MAX(2,6,29)
521	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
522	# endif /* RTLNX_VER_MAX(2,6,29) */
523	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
524	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
525	}
526	else
527	pOverride = NULL;
528	}
529	else
530	pOverride = NULL;
531	RTSpinlockRelease(pThis->hSpinlock);
532
533	if (pOverride)
534	{
535	printk("vboxnetflt: %llu out of %llu packets were not sent (directed to host)\n", pOverride->cFiltered, pOverride->cTotal);
536	RTMemFree(pOverride);
537	}
538	}
539
540	#endif /* VBOXNETFLT_WITH_HOST2WIRE_FILTER */
541
542
543	/**
544	* Reads and retains the host interface handle.
545	*
546	* @returns The handle, NULL if detached.
547	* @param pThis
548	*/
549	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
550	{
551	#if 0
552	struct net_device *pDev = NULL;
553
554	Log(("vboxNetFltLinuxRetainNetDev\n"));
555	/*
556	* Be careful here to avoid problems racing the detached callback.
557	*/
558	RTSpinlockAcquire(pThis->hSpinlock);
559	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
560	{
561	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
562	if (pDev)
563	{
564	dev_hold(pDev);
565	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
566	pDev, pDev->name,
567	#if RTLNX_VER_MIN(2,6,37)
568	netdev_refcnt_read(pDev)
569	#else
570	atomic_read(&pDev->refcnt)
571	#endif
572	));
573	}
574	}
575	RTSpinlockRelease(pThis->hSpinlock);
576
577	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
578	return pDev;
579	#else
580	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
581	#endif
582	}
583
584
585	/**
586	* Release the host interface handle previously retained
587	* by vboxNetFltLinuxRetainNetDev.
588	*
589	* @param pThis The instance.
590	* @param pDev The vboxNetFltLinuxRetainNetDev
591	* return value, NULL is fine.
592	*/
593	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
594	{
595	#if 0
596	Log(("vboxNetFltLinuxReleaseNetDev\n"));
597	NOREF(pThis);
598	if (pDev)
599	{
600	dev_put(pDev);
601	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
602	pDev, pDev->name,
603	#if RTLNX_VER_MIN(2,6,37)
604	netdev_refcnt_read(pDev)
605	#else
606	atomic_read(&pDev->refcnt)
607	#endif
608	));
609	}
610	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
611	#endif
612	}
613
614	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
615	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
616
617	/**
618	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
619	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
620	*
621	* @returns true / false accordingly.
622	* @param pBuf The sk_buff.
623	*/
624	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
625	{
626	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
627	}
628
629
630	/**
631	* Checks whether this SG list contains a GSO packet.
632	*
633	* @returns true / false accordingly.
634	* @param pSG The (scatter/)gather list.
635	*/
636	DECLINLINE(bool) vboxNetFltLinuxIsGso(PINTNETSG pSG)
637	{
638	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
639	return !((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type == PDMNETWORKGSOTYPE_INVALID);
640	#else /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
641	return false;
642	#endif /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
643	}
644
645
646	/**
647	* Find out the frame size (of a single segment in case of GSO frames).
648	*
649	* @returns the frame size.
650	* @param pSG The (scatter/)gather list.
651	*/
652	DECLINLINE(uint32_t) vboxNetFltLinuxFrameSize(PINTNETSG pSG)
653	{
654	uint16_t u16Type = 0;
655	uint32_t cbVlanTag = 0;
656	if (pSG->aSegs[0].cb >= sizeof(RTNETETHERHDR))
657	u16Type = RT_BE2H_U16(((PCRTNETETHERHDR)pSG->aSegs[0].pv)->EtherType);
658	else if (pSG->cbTotal >= sizeof(RTNETETHERHDR))
659	{
660	uint32_t off = RT_UOFFSETOF(RTNETETHERHDR, EtherType);
661	uint32_t i;
662	for (i = 0; i < pSG->cSegsUsed; ++i)
663	{
664	if (off <= pSG->aSegs[i].cb)
665	{
666	if (off + sizeof(uint16_t) <= pSG->aSegs[i].cb)
667	u16Type = RT_BE2H_U16((uint16_t )((uintptr_t)pSG->aSegs[i].pv + off));
668	else if (i + 1 < pSG->cSegsUsed)
669	u16Type = RT_BE2H_U16( ((uint16_t)( ((uint8_t *)pSG->aSegs[i].pv)[off] ) << 8)
670	+ (uint8_t )pSG->aSegs[i + 1].pv); /* ASSUMES no empty segments! */
671	/* else: frame is too short. */
672	break;
673	}
674	off -= pSG->aSegs[i].cb;
675	}
676	}
677	if (u16Type == RTNET_ETHERTYPE_VLAN)
678	cbVlanTag = 4;
679	return (vboxNetFltLinuxIsGso(pSG) ? (uint32_t)pSG->GsoCtx.cbMaxSeg + pSG->GsoCtx.cbHdrsTotal : pSG->cbTotal) - cbVlanTag;
680	}
681
682
683	/**
684	* Internal worker that create a linux sk_buff for a
685	* (scatter/)gather list.
686	*
687	* @returns Pointer to the sk_buff.
688	* @param pThis The instance.
689	* @param pSG The (scatter/)gather list.
690	* @param fDstWire Set if the destination is the wire.
691	*/
692	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
693	{
694	struct sk_buff *pPkt;
695	struct net_device *pDev;
696	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
697	unsigned fGsoType = 0;
698	#endif
699
700	if (pSG->cbTotal == 0)
701	{
702	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
703	return NULL;
704	}
705	Log5(("VBoxNetFlt: Packet to %s of %d bytes (frame=%d).\n", fDstWire?"wire":"host", pSG->cbTotal, vboxNetFltLinuxFrameSize(pSG)));
706	if (fDstWire && (vboxNetFltLinuxFrameSize(pSG) > ASMAtomicReadU32(&pThis->u.s.cbMtu) + 14))
707	{
708	static bool s_fOnce = true;
709	if (s_fOnce)
710	{
711	s_fOnce = false;
712	printk("VBoxNetFlt: Dropped over-sized packet (%d bytes) coming from internal network.\n", vboxNetFltLinuxFrameSize(pSG));
713	}
714	return NULL;
715	}
716
717	/** @todo We should use fragments mapping the SG buffers with large packets.
718	* 256 bytes seems to be the a threshold used a lot for this. It
719	* requires some nasty work on the intnet side though... */
720	/*
721	* Allocate a packet and copy over the data.
722	*/
723	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
724	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
725	if (RT_UNLIKELY(!pPkt))
726	{
727	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
728	pSG->pvUserData = NULL;
729	return NULL;
730	}
731	pPkt->dev = pDev;
732	pPkt->ip_summed = CHECKSUM_NONE;
733
734	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
735	skb_reserve(pPkt, NET_IP_ALIGN);
736
737	/* Copy the segments. */
738	skb_put(pPkt, pSG->cbTotal);
739	IntNetSgRead(pSG, pPkt->data);
740
741	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
742	/*
743	* Setup GSO if used by this packet.
744	*/
745	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
746	{
747	default:
748	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
749	RT_FALL_THRU();
750	case PDMNETWORKGSOTYPE_INVALID:
751	fGsoType = 0;
752	break;
753	case PDMNETWORKGSOTYPE_IPV4_TCP:
754	fGsoType = SKB_GSO_TCPV4;
755	break;
756	case PDMNETWORKGSOTYPE_IPV6_TCP:
757	fGsoType = SKB_GSO_TCPV6;
758	break;
759	}
760	if (fGsoType)
761	{
762	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
763
764	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
765	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
766	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
767
768	/*
769	* We need to set checksum fields even if the packet goes to the host
770	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
771	*/
772	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrsTotal);
773	pPkt->ip_summed = CHECKSUM_PARTIAL;
774	# if RTLNX_VER_MIN(2,6,22)
775	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
776	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
777	pPkt->csum_offset = RT_UOFFSETOF(RTNETTCP, th_sum);
778	else
779	pPkt->csum_offset = RT_UOFFSETOF(RTNETUDP, uh_sum);
780	# else
781	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
782	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
783	pPkt->csum = RT_UOFFSETOF(RTNETTCP, th_sum);
784	else
785	pPkt->csum = RT_UOFFSETOF(RTNETUDP, uh_sum);
786	# endif
787	if (!fDstWire)
788	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
789	}
790	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
791
792	/*
793	* Finish up the socket buffer.
794	*/
795	pPkt->protocol = eth_type_trans(pPkt, pDev);
796	if (fDstWire)
797	{
798	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
799
800	/* Restore ethernet header back. */
801	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
802	VBOX_SKB_RESET_MAC_HDR(pPkt);
803	}
804	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
805
806	return pPkt;
807	}
808
809
810	/**
811	* Return the offset where to start checksum computation from.
812	*
813	* @returns the offset relative to pBuf->data.
814	* @param pBuf The socket buffer.
815	*/
816	DECLINLINE(unsigned) vboxNetFltLinuxGetChecksumStartOffset(struct sk_buff *pBuf)
817	{
818	#if RTLNX_VER_MIN(2,6,38)
819	return skb_checksum_start_offset(pBuf);
820	#elif RTLNX_VER_MIN(2,6,22)
821	return pBuf->csum_start - skb_headroom(pBuf);
822	#else
823	unsigned char *pTransportHdr = pBuf->h.raw;
824	# if RTLNX_VER_MAX(2,6,19)
825	/*
826	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
827	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
828	* header length from the header itself and reconstruct 'h' pointer
829	* to TCP (or whatever) header.
830	*/
831	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
832	pTransportHdr = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
833	# endif
834	return pTransportHdr - pBuf->data;
835	#endif
836	}
837
838
839	/**
840	* Initializes a SG list from an sk_buff.
841	*
842	* @param pThis The instance.
843	* @param pBuf The sk_buff.
844	* @param pSG The SG.
845	* @param cbExtra The number of bytes of extra space allocated immediately after the SG.
846	* @param cSegs The number of segments allocated for the SG.
847	* This should match the number in the mbuf exactly!
848	* @param fSrc The source of the frame.
849	* @param pGsoCtx Pointer to the GSO context if it's a GSO
850	* internal network frame. NULL if regular frame.
851	*/
852	static void vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
853	unsigned cbExtra, unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
854	{
855	int i;
856	NOREF(pThis);
857
858	#ifndef VBOXNETFLT_SG_SUPPORT
859	Assert(!skb_shinfo(pBuf)->frag_list);
860	#else /* VBOXNETFLT_SG_SUPPORT */
861	uint8_t pExtra = (uint8_t )&pSG->aSegs[cSegs];
862	unsigned cbConsumed = 0;
863	unsigned cbProduced = 0;
864
865	# if RTLNX_VER_MIN(2,6,27)
866	/* Restore VLAN tag stripped by host hardware */
867	if (vlan_tx_tag_present(pBuf))
868	{
869	uint8_t *pMac = pBuf->data;
870	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )pExtra;
871	Assert(ETH_ALEN * 2 + VLAN_HLEN <= cbExtra);
872	memmove(pVHdr, pMac, ETH_ALEN * 2);
873	/* Consume whole Ethernet header: 2 addresses + EtherType (see @bugref{8599}) */
874	cbConsumed += ETH_ALEN * 2 + sizeof(uint16_t);
875	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
876	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
877	pVHdr->h_vlan_encapsulated_proto = (uint16_t)(pMac + ETH_ALEN * 2);
878	cbProduced += VLAN_ETH_HLEN;
879	}
880	# endif /* RTLNX_VER_MIN(2,6,27) */
881
882	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
883	{
884	unsigned uCsumStartOffset = vboxNetFltLinuxGetChecksumStartOffset(pBuf);
885	unsigned uCsumStoreOffset = uCsumStartOffset + VBOX_SKB_CSUM_OFFSET(pBuf) - cbConsumed;
886	Log3(("cbConsumed=%u cbProduced=%u uCsumStartOffset=%u uCsumStoreOffset=%u\n",
887	cbConsumed, cbProduced, uCsumStartOffset, uCsumStoreOffset));
888	Assert(cbProduced + uCsumStoreOffset + sizeof(uint16_t) <= cbExtra);
889	/*
890	* We assume that the checksum is stored at the very end of the transport header
891	* so we will have all headers in a single fragment. If our assumption is wrong
892	* we may see suboptimal performance.
893	*/
894	memmove(pExtra + cbProduced,
895	pBuf->data + cbConsumed,
896	uCsumStoreOffset);
897	unsigned uChecksum = skb_checksum(pBuf, uCsumStartOffset, pBuf->len - uCsumStartOffset, 0);
898	(uint16_t)(pExtra + cbProduced + uCsumStoreOffset) = csum_fold(uChecksum);
899	cbProduced += uCsumStoreOffset + sizeof(uint16_t);
900	cbConsumed += uCsumStoreOffset + sizeof(uint16_t);
901	}
902	#endif /* VBOXNETFLT_SG_SUPPORT */
903
904	if (!pGsoCtx)
905	IntNetSgInitTempSegs(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/);
906	else
907	IntNetSgInitTempSegsGso(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/, pGsoCtx);
908
909	int iSeg = 0;
910	#ifdef VBOXNETFLT_SG_SUPPORT
911	if (cbProduced)
912	{
913	pSG->aSegs[iSeg].cb = cbProduced;
914	pSG->aSegs[iSeg].pv = pExtra;
915	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
916	}
917	pSG->aSegs[iSeg].cb = skb_headlen(pBuf) - cbConsumed;
918	pSG->aSegs[iSeg].pv = pBuf->data + cbConsumed;
919	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
920	Assert(iSeg <= pSG->cSegsAlloc);
921
922	# ifdef LOG_ENABLED
923	if (pBuf->data_len)
924	Log6((" kmap_atomic:"));
925	# endif /* LOG_ENABLED */
926	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
927	{
928	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
929	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
930	pSG->aSegs[iSeg].cb = pFrag->bv_len;
931	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
932	# else /* < KERNEL_VERSION(5, 4, 0) */
933	pSG->aSegs[iSeg].cb = pFrag->size;
934	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
935	# endif /* >= KERNEL_VERSION(5, 4, 0) */
936	Log6((" %p", pSG->aSegs[iSeg].pv));
937	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
938	Assert(iSeg <= pSG->cSegsAlloc);
939	}
940	struct sk_buff *pFragBuf;
941	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
942	{
943	pSG->aSegs[iSeg].cb = skb_headlen(pFragBuf);
944	pSG->aSegs[iSeg].pv = pFragBuf->data;
945	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
946	Assert(iSeg <= pSG->cSegsAlloc);
947	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
948	{
949	skb_frag_t *pFrag = &skb_shinfo(pFragBuf)->frags[i];
950	# if RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
951	pSG->aSegs[iSeg].cb = pFrag->bv_len;
952	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->bv_offset;
953	# else /* < KERNEL_VERSION(5, 4, 0) */
954	pSG->aSegs[iSeg].cb = pFrag->size;
955	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
956	# endif /* >= KERNEL_VERSION(5, 4, 0) */
957	Log6((" %p", pSG->aSegs[iSeg].pv));
958	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
959	Assert(iSeg <= pSG->cSegsAlloc);
960	}
961	}
962	# ifdef LOG_ENABLED
963	if (pBuf->data_len)
964	Log6(("\n"));
965	# endif /* LOG_ENABLED */
966	#else
967	pSG->aSegs[iSeg].cb = pBuf->len;
968	pSG->aSegs[iSeg].pv = pBuf->data;
969	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
970	#endif
971
972	pSG->cSegsUsed = iSeg;
973
974	#if 0
975	if (cbProduced)
976	{
977	LogRel(("vboxNetFltLinuxSkBufToSG: original packet dump:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
978	LogRel(("vboxNetFltLinuxSkBufToSG: cbConsumed=%u cbProduced=%u cbExtra=%u\n", cbConsumed, cbProduced, cbExtra));
979	uint32_t offset = 0;
980	for (i = 0; i < pSG->cSegsUsed; ++i)
981	{
982	LogRel(("vboxNetFltLinuxSkBufToSG: seg#%d (%d bytes, starting at 0x%x):\n%.*Rhxd\n",
983	i, pSG->aSegs[i].cb, offset, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
984	offset += pSG->aSegs[i].cb;
985	}
986	}
987	#endif
988
989	#ifdef PADD_RUNT_FRAMES_FROM_HOST
990	/*
991	* Add a trailer if the frame is too small.
992	*
993	* Since we're getting to the packet before it is framed, it has not
994	* yet been padded. The current solution is to add a segment pointing
995	* to a buffer containing all zeros and pray that works for all frames...
996	*/
997	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
998	{
999	Assert(pBuf->data_len == 0); /* Packets with fragments are never small! */
1000	static uint8_t const s_abZero[128] = {0};
1001
1002	AssertReturnVoid(iSeg < cSegs);
1003
1004	pSG->aSegs[iSeg].Phys = NIL_RTHCPHYS;
1005	pSG->aSegs[iSeg].pv = (void *)&s_abZero[0];
1006	pSG->aSegs[iSeg++].cb = 60 - pSG->cbTotal;
1007	pSG->cbTotal = 60;
1008	pSG->cSegsUsed++;
1009	Assert(iSeg <= pSG->cSegsAlloc)
1010	}
1011	#endif
1012
1013	Log6(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1014	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1015	for (i = 0; i < pSG->cSegsUsed; i++)
1016	Log6(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1017	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1018	}
1019
1020	/**
1021	* Packet handler; not really documented - figure it out yourself.
1022	*
1023	* @returns 0 or EJUSTRETURN - this is probably copy & pastry and thus wrong.
1024	*/
1025	#if RTLNX_VER_MIN(2,6,14)
1026	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1027	struct net_device *pSkbDev,
1028	struct packet_type *pPacketType,
1029	struct net_device *pOrigDev)
1030	#else
1031	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1032	struct net_device *pSkbDev,
1033	struct packet_type *pPacketType)
1034	#endif
1035	{
1036	PVBOXNETFLTINS pThis;
1037	struct net_device *pDev;
1038	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1039	pBuf, pSkbDev, pPacketType));
1040	#if RTLNX_VER_MIN(2,6,18)
1041	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",
1042	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));
1043	# if RTLNX_VER_MIN(2,6,22)
1044	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1045	# endif
1046	#else
1047	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",
1048	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));
1049	#endif
1050	/*
1051	* Drop it immediately?
1052	*/
1053	if (!pBuf)
1054	return 0;
1055
1056	if (pBuf->pkt_type == PACKET_LOOPBACK)
1057	{
1058	/*
1059	* We are not interested in loopbacked packets as they will always have
1060	* another copy going to the wire.
1061	*/
1062	Log2(("vboxNetFltLinuxPacketHandler: dropped loopback packet (cb=%u)\n", pBuf->len));
1063	dev_kfree_skb(pBuf); /* We must 'consume' all packets we get (@bugref{6539})! */
1064	return 0;
1065	}
1066
1067	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1068	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1069	if (pDev != pSkbDev)
1070	{
1071	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1072	kfree_skb(pBuf); /* This is a failure, so we use kfree_skb instead of dev_kfree_skb. */
1073	return 0;
1074	}
1075
1076	Log6(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1077	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1078	{
1079	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1080	dev_kfree_skb(pBuf);
1081	return 0;
1082	}
1083
1084	#ifndef VBOXNETFLT_SG_SUPPORT
1085	{
1086	/*
1087	* Get rid of fragmented packets, they cause too much trouble.
1088	*/
1089	unsigned int uMacLen = pBuf->mac_len;
1090	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1091	dev_kfree_skb(pBuf);
1092	if (!pCopy)
1093	{
1094	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1095	return 0;
1096	}
1097	pBuf = pCopy;
1098	/* Somehow skb_copy ignores mac_len */
1099	pBuf->mac_len = uMacLen;
1100	# if RTLNX_VER_MIN(2,6,27)
1101	/* Restore VLAN tag stripped by host hardware */
1102	if (vlan_tx_tag_present(pBuf) && skb_headroom(pBuf) >= VLAN_ETH_HLEN)
1103	{
1104	uint8_t pMac = (uint8_t)skb_mac_header(pBuf);
1105	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )(pMac - VLAN_HLEN);
1106	memmove(pVHdr, pMac, ETH_ALEN * 2);
1107	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
1108	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
1109	pBuf->mac_header -= VLAN_HLEN;
1110	pBuf->mac_len += VLAN_HLEN;
1111	}
1112	# endif /* RTLNX_VER_MIN(2,6,27) */
1113
1114	# if RTLNX_VER_MIN(2,6,18)
1115	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",
1116	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));
1117	# if RTLNX_VER_MIN(2,6,22)
1118	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1119	# endif /* RTLNX_VER_MIN(2,6,22) */
1120	# else /* RTLNX_VER_MAX(2,6,18) */
1121	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",
1122	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));
1123	# endif /* RTLNX_VER_MAX(2,6,18) */
1124	}
1125	#endif /* !VBOXNETFLT_SG_SUPPORT */
1126
1127	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1128	/* Forward it to the internal network. */
1129	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1130	#else /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1131	/* Add the packet to transmit queue and schedule the bottom half. */
1132	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1133	schedule_work(&pThis->u.s.XmitTask);
1134	Log6(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1135	&pThis->u.s.XmitTask, pBuf));
1136	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1137
1138	/* It does not really matter what we return, it is ignored by the kernel. */
1139	return 0;
1140	}
1141
1142	/**
1143	* Calculate the number of INTNETSEG segments the socket buffer will need.
1144	*
1145	* @returns Segment count.
1146	* @param pBuf The socket buffer.
1147	* @param pcbTemp Where to store the number of bytes of the part
1148	* of the socket buffer that will be copied to
1149	* a temporary storage.
1150	*/
1151	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff pBuf, unsigned pcbTemp)
1152	{
1153	*pcbTemp = 0;
1154	#ifdef VBOXNETFLT_SG_SUPPORT
1155	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1156	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1157	{
1158	*pcbTemp = vboxNetFltLinuxGetChecksumStartOffset(pBuf) + VBOX_SKB_CSUM_OFFSET(pBuf) + sizeof(uint16_t);
1159	}
1160	# if RTLNX_VER_MIN(2,6,27)
1161	if (vlan_tx_tag_present(pBuf))
1162	{
1163	if (*pcbTemp)
1164	*pcbTemp += VLAN_HLEN;
1165	else
1166	*pcbTemp = VLAN_ETH_HLEN;
1167	}
1168	# endif /* RTLNX_VER_MIN(2,6,27) */
1169	if (*pcbTemp)
1170	++cSegs;
1171	struct sk_buff *pFrag;
1172	for (pFrag = skb_shinfo(pBuf)->frag_list; pFrag; pFrag = pFrag->next)
1173	{
1174	Log6(("vboxNetFltLinuxCalcSGSegments: frag=%p len=%d data_len=%d frags=%d frag_list=%p next=%p\n",
1175	pFrag, pFrag->len, pFrag->data_len, skb_shinfo(pFrag)->nr_frags, skb_shinfo(pFrag)->frag_list, pFrag->next));
1176	cSegs += 1 + skb_shinfo(pFrag)->nr_frags;
1177	}
1178	#else
1179	unsigned cSegs = 1;
1180	#endif
1181	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1182	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1183	if (pBuf->len < 60)
1184	cSegs++;
1185	#endif
1186	return cSegs;
1187	}
1188
1189
1190	/**
1191	* Destroy the intnet scatter / gather buffer created by
1192	* vboxNetFltLinuxSkBufToSG.
1193	*
1194	* @param pSG The (scatter/)gather list.
1195	* @param pBuf The original socket buffer that was used to create
1196	* the scatter/gather list.
1197	*/
1198	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG, struct sk_buff *pBuf)
1199	{
1200	#ifdef VBOXNETFLT_SG_SUPPORT
1201	int i, iSeg = 1; /* Skip non-paged part of SKB */
1202	/* Check if the extra buffer behind SG structure was used for modified packet header */
1203	if (pBuf->data != pSG->aSegs[0].pv)
1204	++iSeg; /* Skip it as well */
1205	# ifdef LOG_ENABLED
1206	if (pBuf->data_len)
1207	Log6(("kunmap_atomic:"));
1208	# endif /* LOG_ENABLED */
1209	/* iSeg now points to the first mapped fragment if there are any */
1210	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1211	{
1212	Log6((" %p", pSG->aSegs[iSeg].pv));
1213	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1214	}
1215	struct sk_buff *pFragBuf;
1216	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
1217	{
1218	++iSeg; /* Non-fragment (unmapped) portion of chained SKB */
1219	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
1220	{
1221	Log6((" %p", pSG->aSegs[iSeg].pv));
1222	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1223	}
1224	}
1225	# ifdef LOG_ENABLED
1226	if (pBuf->data_len)
1227	Log6(("\n"));
1228	# endif /* LOG_ENABLED */
1229	#endif
1230	NOREF(pSG);
1231	}
1232
1233	#ifdef LOG_ENABLED
1234	/**
1235	* Logging helper.
1236	*/
1237	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1238	{
1239	int i, offSeg;
1240	uint8_t pInt, pExt;
1241	static int iPacketNo = 1;
1242	iPacketNo += iIncrement;
1243	if (fEgress)
1244	{
1245	pExt = pSG->aSegs[0].pv;
1246	pInt = pExt + 6;
1247	}
1248	else
1249	{
1250	pInt = pSG->aSegs[0].pv;
1251	pExt = pInt + 6;
1252	}
1253	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1254	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1255	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1256	fEgress ? "-->" : "<--", pszWhere,
1257	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1258	pSG->cbTotal, iPacketNo));
1259	if (pSG->cSegsUsed == 1)
1260	{
1261	Log4(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1262	}
1263	else
1264	{
1265	for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1266	{
1267	Log4(("-- segment %d at 0x%x (%d bytes)\n --\n%.*Rhxd\n",
1268	i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1269	offSeg += pSG->aSegs[i].cb;
1270	}
1271	}
1272	}
1273	#else
1274	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1275	#endif
1276
1277	#ifdef VBOXNETFLT_WITH_GSO_RECV
1278
1279	/**
1280	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1281	* GSO socket buffer without having to segment it.
1282	*
1283	* @returns true on success, false if needs segmenting.
1284	* @param pThis The net filter instance.
1285	* @param pSkb The GSO socket buffer.
1286	* @param fSrc The source.
1287	* @param pGsoCtx Where to return the GSO context on success.
1288	*/
1289	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1290	PPDMNETWORKGSO pGsoCtx)
1291	{
1292	PDMNETWORKGSOTYPE enmGsoType;
1293	uint16_t uEtherType;
1294	unsigned int cbTransport;
1295	unsigned int offTransport;
1296	unsigned int cbTransportHdr;
1297	unsigned uProtocol;
1298	union
1299	{
1300	RTNETIPV4 IPv4;
1301	RTNETIPV6 IPv6;
1302	RTNETTCP Tcp;
1303	uint8_t ab[40];
1304	uint16_t au16[40/2];
1305	uint32_t au32[40/4];
1306	} Buf;
1307
1308	/*
1309	* Check the GSO properties of the socket buffer and make sure it fits.
1310	*/
1311	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1312	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1313	{
1314	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1315	return false;
1316	}
1317	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1318	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1319	{
1320	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1321	return false;
1322	}
1323
1324	/*
1325	* Switch on the ethertype.
1326	*/
1327	uEtherType = pSkb->protocol;
1328	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1329	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1330	{
1331	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1332	if (puEtherType)
1333	uEtherType = *puEtherType;
1334	}
1335	switch (uEtherType)
1336	{
1337	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1338	{
1339	unsigned int cbHdr;
1340	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1341	if (RT_UNLIKELY(!pIPv4))
1342	{
1343	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1344	return false;
1345	}
1346
1347	cbHdr = pIPv4->ip_hl * 4;
1348	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1349	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1350	\|\| cbHdr > cbTransport ))
1351	{
1352	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1353	return false;
1354	}
1355	cbTransport -= cbHdr;
1356	offTransport = pSkb->mac_len + cbHdr;
1357	uProtocol = pIPv4->ip_p;
1358	if (uProtocol == RTNETIPV4_PROT_TCP)
1359	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1360	else if (uProtocol == RTNETIPV4_PROT_UDP)
1361	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1362	else /** @todo IPv6: 4to6 tunneling */
1363	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1364	break;
1365	}
1366
1367	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1368	{
1369	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1370	if (RT_UNLIKELY(!pIPv6))
1371	{
1372	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1373	return false;
1374	}
1375
1376	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1377	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1378	uProtocol = pIPv6->ip6_nxt;
1379	/** @todo IPv6: Dig our way out of the other headers. */
1380	if (uProtocol == RTNETIPV4_PROT_TCP)
1381	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1382	else if (uProtocol == RTNETIPV4_PROT_UDP)
1383	enmGsoType = PDMNETWORKGSOTYPE_IPV6_UDP;
1384	else
1385	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1386	break;
1387	}
1388
1389	default:
1390	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1391	return false;
1392	}
1393
1394	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1395	{
1396	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1397	return false;
1398	}
1399
1400	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1401	\|\| offTransport + cbTransport > pSkb->len
1402	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1403	{
1404	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1405	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1406	return false;
1407	}
1408
1409	/*
1410	* Check the TCP/UDP bits.
1411	*/
1412	if (uProtocol == RTNETIPV4_PROT_TCP)
1413	{
1414	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1415	if (RT_UNLIKELY(!pTcp))
1416	{
1417	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1418	return false;
1419	}
1420
1421	cbTransportHdr = pTcp->th_off * 4;
1422	pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1423	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1424	\|\| cbTransportHdr > cbTransport
1425	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1426	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1427	{
1428	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1429	return false;
1430	}
1431
1432	}
1433	else
1434	{
1435	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1436	cbTransportHdr = sizeof(RTNETUDP);
1437	pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1438	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1439	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1440	{
1441	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1442	return false;
1443	}
1444	}
1445
1446	/*
1447	* We're good, init the GSO context.
1448	*/
1449	pGsoCtx->u8Type = enmGsoType;
1450	pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1451	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1452	pGsoCtx->offHdr1 = pSkb->mac_len;
1453	pGsoCtx->offHdr2 = offTransport;
1454	pGsoCtx->u8Unused = 0;
1455
1456	return true;
1457	}
1458
1459	/**
1460	* Forward the socket buffer as a GSO internal network frame.
1461	*
1462	* @returns IPRT status code.
1463	* @param pThis The net filter instance.
1464	* @param pSkb The GSO socket buffer.
1465	* @param fSrc The source.
1466	* @param pGsoCtx Where to return the GSO context on success.
1467	*/
1468	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1469	{
1470	int rc;
1471	unsigned cbExtra;
1472	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb, &cbExtra);
1473	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1474	if (RT_LIKELY(pSG))
1475	{
1476	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cbExtra, cSegs, fSrc, pGsoCtx);
1477
1478	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1479	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1480
1481	vboxNetFltLinuxDestroySG(pSG, pSkb);
1482	rc = VINF_SUCCESS;
1483	}
1484	else
1485	{
1486	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1487	rc = VERR_NO_MEMORY;
1488	}
1489	return rc;
1490	}
1491
1492	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1493
1494	/**
1495	* Worker for vboxNetFltLinuxForwardToIntNet.
1496	*
1497	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1498	* @param pThis The net filter instance.
1499	* @param pBuf The socket buffer.
1500	* @param fSrc The source.
1501	*/
1502	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1503	{
1504	int rc;
1505	unsigned cbExtra;
1506	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf, &cbExtra);
1507	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1508	if (RT_LIKELY(pSG))
1509	{
1510	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cbExtra, cSegs, fSrc, NULL /pGsoCtx/);
1511
1512	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1513	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1514
1515	vboxNetFltLinuxDestroySG(pSG, pBuf);
1516	rc = VINF_SUCCESS;
1517	}
1518	else
1519	{
1520	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1521	rc = VERR_NO_MEMORY;
1522	}
1523	return rc;
1524	}
1525
1526
1527	/**
1528	* I won't disclose what I do, figure it out yourself, including pThis referencing.
1529	*
1530	* @param pThis The net filter instance.
1531	* @param pBuf The socket buffer.
1532	* @param fSrc Where the packet comes from.
1533	*/
1534	static void vboxNetFltLinuxForwardToIntNetInner(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1535	{
1536	#ifdef VBOXNETFLT_WITH_GSO
1537	if (skb_is_gso(pBuf))
1538	{
1539	PDMNETWORKGSO GsoCtx;
1540	Log6(("vboxNetFltLinuxForwardToIntNetInner: skb len=%u data_len=%u truesize=%u next=%p"
1541	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x ip_summed=%d\n",
1542	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next,
1543	skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size,
1544	skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type,
1545	skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, pBuf->ip_summed));
1546
1547	if (RT_LIKELY(fSrc & INTNETTRUNKDIR_HOST))
1548	{
1549	/*
1550	* skb_gso_segment does the following. Do we need to do it as well?
1551	*/
1552	# if RTLNX_VER_MIN(2,6,22)
1553	skb_reset_mac_header(pBuf);
1554	pBuf->mac_len = pBuf->network_header - pBuf->mac_header;
1555	# else
1556	pBuf->mac.raw = pBuf->data;
1557	pBuf->mac_len = pBuf->nh.raw - pBuf->data;
1558	# endif
1559	}
1560
1561	# ifdef VBOXNETFLT_WITH_GSO_RECV
1562	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1563	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1564	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1565	else
1566	# endif /* VBOXNETFLT_WITH_GSO_RECV */
1567	{
1568	/* Need to segment the packet */
1569	struct sk_buff *pNext;
1570	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1571	if (IS_ERR(pSegment))
1572	{
1573	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1574	return;
1575	}
1576
1577	for (; pSegment; pSegment = pNext)
1578	{
1579	Log6(("vboxNetFltLinuxForwardToIntNetInner: segment len=%u data_len=%u truesize=%u next=%p"
1580	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1581	pSegment->len, pSegment->data_len, pSegment->truesize, pSegment->next,
1582	skb_shinfo(pSegment)->nr_frags, skb_shinfo(pSegment)->gso_size,
1583	skb_shinfo(pSegment)->gso_segs, skb_shinfo(pSegment)->gso_type,
1584	skb_shinfo(pSegment)->frag_list, pSegment->pkt_type));
1585	pNext = pSegment->next;
1586	pSegment->next = 0;
1587	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1588	dev_kfree_skb(pSegment);
1589	}
1590	}
1591	}
1592	else
1593	#endif /* VBOXNETFLT_WITH_GSO */
1594	{
1595	Log6(("vboxNetFltLinuxForwardToIntNetInner: ptk_type=%d ip_summed=%d len=%d"
1596	" data_len=%d headroom=%d hdr_len=%d csum_offset=%d\n",
1597	pBuf->pkt_type, pBuf->ip_summed, pBuf->len, pBuf->data_len, skb_headroom(pBuf),
1598	skb_headlen(pBuf), vboxNetFltLinuxGetChecksumStartOffset(pBuf)));
1599	#ifndef VBOXNETFLT_SG_SUPPORT
1600	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1601	{
1602	# if RTLNX_VER_MIN(2,6,19)
1603	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1604	# else
1605	/*
1606	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1607	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1608	* header length from the header itself and reconstruct 'h' pointer
1609	* to TCP (or whatever) header.
1610	*/
1611	unsigned char *tmp = pBuf->h.raw;
1612	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1613	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1614	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1615	/* Restore the original (wrong) pointer. */
1616	pBuf->h.raw = tmp;
1617	# endif
1618	if (rc)
1619	{
1620	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1621	return;
1622	}
1623	}
1624	#endif /* !VBOXNETFLT_SG_SUPPORT */
1625	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1626	}
1627	}
1628
1629
1630	/**
1631	* Temporarily adjust pBuf->data so it always points to the Ethernet header,
1632	* then forward it to the internal network.
1633	*
1634	* @param pThis The net filter instance.
1635	* @param pBuf The socket buffer. This is consumed by this function.
1636	*/
1637	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1638	{
1639	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1640
1641	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1642	{
1643	/*
1644	* The packet came from the wire and the driver has already consumed
1645	* mac header. We need to restore it back. Moreover, after we are
1646	* through with this skb we need to restore its original state!
1647	*/
1648	skb_push(pBuf, pBuf->mac_len);
1649	Log5(("vboxNetFltLinuxForwardToIntNet: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1650	pBuf->mac_len, pBuf->data, skb_mac_header(pBuf), skb_network_header(pBuf)));
1651	}
1652
1653	vboxNetFltLinuxForwardToIntNetInner(pThis, pBuf, fSrc);
1654
1655	/*
1656	* Restore the original state of skb as there are other handlers this skb
1657	* will be provided to.
1658	*/
1659	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1660	skb_pull(pBuf, pBuf->mac_len);
1661
1662	dev_kfree_skb(pBuf);
1663	}
1664
1665
1666	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1667	/**
1668	* Work queue handler that forwards the socket buffers queued by
1669	* vboxNetFltLinuxPacketHandler to the internal network.
1670	*
1671	* @param pWork The work queue.
1672	*/
1673	# if RTLNX_VER_MIN(2,6,20)
1674	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1675	# else
1676	static void vboxNetFltLinuxXmitTask(void *pWork)
1677	# endif
1678	{
1679	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1680	struct sk_buff *pBuf;
1681
1682	Log6(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1683
1684	/*
1685	* Active? Retain the instance and increment the busy counter.
1686	*/
1687	if (vboxNetFltTryRetainBusyActive(pThis))
1688	{
1689	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1690	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1691
1692	vboxNetFltRelease(pThis, true /* fBusy */);
1693	}
1694	else
1695	{
1696	/** @todo Shouldn't we just drop the packets here? There is little point in
1697	* making them accumulate when the VM is paused and it'll only waste
1698	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1699	* before start draining the packets (goes for the intnet ring buf
1700	* too)? */
1701	}
1702	}
1703	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1704
1705	/**
1706	* Reports the GSO capabilities of the hardware NIC.
1707	*
1708	* @param pThis The net filter instance. The caller hold a
1709	* reference to this.
1710	*/
1711	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1712	{
1713	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1714	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1715	{
1716	struct net_device *pDev;
1717	unsigned int fFeatures;
1718
1719	RTSpinlockAcquire(pThis->hSpinlock);
1720
1721	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1722	if (pDev)
1723	fFeatures = pDev->features;
1724	else
1725	fFeatures = 0;
1726
1727	RTSpinlockRelease(pThis->hSpinlock);
1728
1729	if (pThis->pSwitchPort)
1730	{
1731	/* Set/update the GSO capabilities of the NIC. */
1732	uint32_t fGsoCapabilites = 0;
1733	if (fFeatures & NETIF_F_TSO)
1734	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1735	if (fFeatures & NETIF_F_TSO6)
1736	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1737	Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s\n",
1738	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1739	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : ""));
1740	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1741	}
1742
1743	vboxNetFltRelease(pThis, true /fBusy/);
1744	}
1745	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
1746	}
1747
1748	/**
1749	* Helper that determines whether the host (ignoreing us) is operating the
1750	* interface in promiscuous mode or not.
1751	*/
1752	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1753	{
1754	bool fRc = false;
1755	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1756	if (pDev)
1757	{
1758	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1759	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1760	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1761	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1762	}
1763	return fRc;
1764	}
1765
1766	/**
1767	* Does this device needs link state change signaled?
1768	* Currently we need it for our own VBoxNetAdp and TAP.
1769	*/
1770	static bool vboxNetFltNeedsLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev)
1771	{
1772	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1773	{
1774	struct ethtool_drvinfo Info;
1775
1776	memset(&Info, 0, sizeof(Info));
1777	Info.cmd = ETHTOOL_GDRVINFO;
1778	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1779	Log3(("%s: driver=%.s version=%.s bus_info=%.*s\n",
1780	__FUNCTION__,
1781	sizeof(Info.driver), Info.driver,
1782	sizeof(Info.version), Info.version,
1783	sizeof(Info.bus_info), Info.bus_info));
1784
1785	if (!strncmp(Info.driver, "vboxnet", sizeof(Info.driver)))
1786	return true;
1787
1788	#if RTLNX_VER_MIN(2,6,36) /* TAP started doing carrier */
1789	return !strncmp(Info.driver, "tun", 4)
1790	&& !strncmp(Info.bus_info, "tap", 4);
1791	#endif
1792	}
1793
1794	return false;
1795	}
1796
1797	#if RTLNX_VER_MAX(2,6,18)
1798	DECLINLINE(void) netif_tx_lock_bh(struct net_device *pDev)
1799	{
1800	spin_lock_bh(&pDev->xmit_lock);
1801	}
1802
1803	DECLINLINE(void) netif_tx_unlock_bh(struct net_device *pDev)
1804	{
1805	spin_unlock_bh(&pDev->xmit_lock);
1806	}
1807	#endif
1808
1809	/**
1810	* Some devices need link state change when filter attaches/detaches
1811	* since the filter is their link in a sense.
1812	*/
1813	static void vboxNetFltSetLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1814	{
1815	if (vboxNetFltNeedsLinkState(pThis, pDev))
1816	{
1817	Log3(("%s: bringing device link %s\n",
1818	__FUNCTION__, fLinkUp ? "up" : "down"));
1819	netif_tx_lock_bh(pDev);
1820	if (fLinkUp)
1821	netif_carrier_on(pDev);
1822	else
1823	netif_carrier_off(pDev);
1824	netif_tx_unlock_bh(pDev);
1825	}
1826	}
1827
1828	/**
1829	* Internal worker for vboxNetFltLinuxNotifierCallback.
1830	*
1831	* @returns VBox status code.
1832	* @param pThis The instance.
1833	* @param pDev The device to attach to.
1834	*/
1835	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1836	{
1837	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1838
1839	/*
1840	* Retain and store the device.
1841	*/
1842	dev_hold(pDev);
1843
1844	RTSpinlockAcquire(pThis->hSpinlock);
1845	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
1846	RTSpinlockRelease(pThis->hSpinlock);
1847
1848	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
1849	pDev, pDev->name,
1850	#if RTLNX_VER_MIN(2,6,37)
1851	netdev_refcnt_read(pDev)
1852	#else
1853	atomic_read(&pDev->refcnt)
1854	#endif
1855	));
1856	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1857	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1858
1859	/* Get the mac address while we still have a valid net_device reference. */
1860	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
1861	/* Initialize MTU */
1862	pThis->u.s.cbMtu = pDev->mtu;
1863
1864	/*
1865	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1866	*/
1867	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1868	pThis->u.s.PacketType.dev = pDev;
1869	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1870	dev_add_pack(&pThis->u.s.PacketType);
1871	ASMAtomicUoWriteBool(&pThis->u.s.fPacketHandler, true);
1872	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1873
1874	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1875	vboxNetFltLinuxHookDev(pThis, pDev);
1876	#endif
1877
1878	/*
1879	* Are we the "carrier" for this device (e.g. vboxnet or tap)?
1880	*/
1881	vboxNetFltSetLinkState(pThis, pDev, true);
1882
1883	/*
1884	* Set indicators that require the spinlock. Be abit paranoid about racing
1885	* the device notification handle.
1886	*/
1887	RTSpinlockAcquire(pThis->hSpinlock);
1888	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1889	if (pDev)
1890	{
1891	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1892	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1893	pDev = NULL; /* don't dereference it */
1894	}
1895	RTSpinlockRelease(pThis->hSpinlock);
1896
1897	/*
1898	* Report GSO capabilities
1899	*/
1900	Assert(pThis->pSwitchPort);
1901	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1902	{
1903	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1904	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
1905	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
1906	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
1907	vboxNetFltRelease(pThis, true /fBusy/);
1908	}
1909
1910	LogRel(("VBoxNetFlt: attached to '%s' / %RTmac\n", pThis->szName, &pThis->u.s.MacAddr));
1911	return VINF_SUCCESS;
1912	}
1913
1914
1915	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1916	{
1917	bool fRegistered;
1918	Assert(!pThis->fDisconnectedFromHost);
1919
1920	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1921	vboxNetFltLinuxUnhookDev(pThis, pDev);
1922	#endif
1923
1924	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
1925	{
1926	dev_remove_pack(&pThis->u.s.PacketType);
1927	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: packet handler removed.\n", pThis));
1928	}
1929
1930	RTSpinlockAcquire(pThis->hSpinlock);
1931	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
1932	if (fRegistered)
1933	{
1934	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1935	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1936	}
1937	RTSpinlockRelease(pThis->hSpinlock);
1938
1939	if (fRegistered)
1940	{
1941	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1942	skb_queue_purge(&pThis->u.s.XmitQueue);
1943	#endif
1944	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: xmit queue purged.\n", pThis));
1945	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
1946	pDev, pDev->name,
1947	#if RTLNX_VER_MIN(2,6,37)
1948	netdev_refcnt_read(pDev)
1949	#else
1950	atomic_read(&pDev->refcnt)
1951	#endif
1952	));
1953	dev_put(pDev);
1954	}
1955
1956	return NOTIFY_OK;
1957	}
1958
1959	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1960	{
1961	/* Check if we are not suspended and promiscuous mode has not been set. */
1962	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
1963	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1964	{
1965	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1966	dev_set_promiscuity(pDev, 1);
1967	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1968	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1969	}
1970	else
1971	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1972	return NOTIFY_OK;
1973	}
1974
1975	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1976	{
1977	/* Undo promiscuous mode if we has set it. */
1978	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1979	{
1980	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1981	dev_set_promiscuity(pDev, -1);
1982	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1983	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1984	}
1985	else
1986	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1987	return NOTIFY_OK;
1988	}
1989
1990	/**
1991	* Callback for listening to MTU change event.
1992	*
1993	* We need to track changes of host's inteface MTU to discard over-sized frames
1994	* coming from the internal network as they may hang the TX queue of host's
1995	* adapter.
1996	*
1997	* @returns NOTIFY_OK
1998	* @param pThis The netfilter instance.
1999	* @param pDev Pointer to device structure of host's interface.
2000	*/
2001	static int vboxNetFltLinuxDeviceMtuChange(PVBOXNETFLTINS pThis, struct net_device *pDev)
2002	{
2003	ASMAtomicWriteU32(&pThis->u.s.cbMtu, pDev->mtu);
2004	Log(("vboxNetFltLinuxDeviceMtuChange: set MTU for %s to %d\n", pThis->szName, pDev->mtu));
2005	return NOTIFY_OK;
2006	}
2007
2008	#ifdef LOG_ENABLED
2009	/** Stringify the NETDEV_XXX constants. */
2010	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2011	{
2012	const char *pszEvent = "NETDEV_<unknown>";
2013	switch (ulEventType)
2014	{
2015	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2016	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2017	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2018	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2019	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2020	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2021	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2022	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2023	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2024	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2025	# ifdef NETDEV_FEAT_CHANGE
2026	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2027	# endif
2028	}
2029	return pszEvent;
2030	}
2031	#endif /* LOG_ENABLED */
2032
2033	/**
2034	* Callback for listening to netdevice events.
2035	*
2036	* This works the rediscovery, clean up on unregistration, promiscuity on
2037	* up/down, and GSO feature changes from ethtool.
2038	*
2039	* @returns NOTIFY_OK
2040	* @param self Pointer to our notifier registration block.
2041	* @param ulEventType The event.
2042	* @param ptr Event specific, but it is usually the device it
2043	* relates to.
2044	*/
2045	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2046
2047	{
2048	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2049	struct net_device pMyDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device );
2050	struct net_device *pDev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2051	int rc = NOTIFY_OK;
2052
2053	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2054	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, pMyDev));
2055
2056	if (ulEventType == NETDEV_REGISTER)
2057	{
2058	#if RTLNX_VER_MIN(2,6,24) /* cgroups/namespaces introduced */
2059	# if RTLNX_VER_MIN(2,6,26)
2060	# define VBOX_DEV_NET(dev) dev_net(dev)
2061	# define VBOX_NET_EQ(n1, n2) net_eq((n1), (n2))
2062	# else
2063	# define VBOX_DEV_NET(dev) ((dev)->nd_net)
2064	# define VBOX_NET_EQ(n1, n2) ((n1) == (n2))
2065	# endif
2066	struct net *pMyNet = current->nsproxy->net_ns;
2067	struct net *pDevNet = VBOX_DEV_NET(pDev);
2068
2069	if (VBOX_NET_EQ(pDevNet, pMyNet))
2070	#endif /* namespaces */
2071	{
2072	if (strcmp(pDev->name, pThis->szName) == 0)
2073	{
2074	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2075	}
2076	}
2077	}
2078	else
2079	{
2080	if (pDev == pMyDev)
2081	{
2082	switch (ulEventType)
2083	{
2084	case NETDEV_UNREGISTER:
2085	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2086	break;
2087	case NETDEV_UP:
2088	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2089	break;
2090	case NETDEV_GOING_DOWN:
2091	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2092	break;
2093	case NETDEV_CHANGEMTU:
2094	rc = vboxNetFltLinuxDeviceMtuChange(pThis, pDev);
2095	break;
2096	case NETDEV_CHANGENAME:
2097	break;
2098	#ifdef NETDEV_FEAT_CHANGE
2099	case NETDEV_FEAT_CHANGE:
2100	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2101	break;
2102	#endif
2103	}
2104	}
2105	}
2106
2107	return rc;
2108	}
2109
2110	/*
2111	* Initial enumeration of netdevs. Called with NETDEV_REGISTER by
2112	* register_netdevice_notifier() under rtnl lock.
2113	*/
2114	static int vboxNetFltLinuxEnumeratorCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2115	{
2116	PVBOXNETFLTINS pThis = ((PVBOXNETFLTNOTIFIER)self)->pThis;
2117	struct net_device *dev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2118	struct in_device *in_dev;
2119	struct inet6_dev *in6_dev;
2120
2121	if (ulEventType != NETDEV_REGISTER)
2122	return NOTIFY_OK;
2123
2124	if (RT_UNLIKELY(pThis->pSwitchPort->pfnNotifyHostAddress == NULL))
2125	return NOTIFY_OK;
2126
2127	/*
2128	* IPv4
2129	*/
2130	#if RTLNX_VER_MIN(2,6,14)
2131	in_dev = __in_dev_get_rtnl(dev);
2132	#else
2133	in_dev = __in_dev_get(dev);
2134	#endif
2135	if (in_dev != NULL)
2136	{
2137	struct in_ifaddr *ifa;
2138
2139	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2140	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2141	return NOTIFY_OK;
2142
2143	if ( dev != pThis->u.s.pDev
2144	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2145	continue;
2146
2147	Log(("%s: %s: IPv4 addr %RTnaipv4 mask %RTnaipv4\n",
2148	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2149	ifa->ifa_address, ifa->ifa_mask));
2150
2151	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2152	/* :fAdded */ true, kIntNetAddrType_IPv4, &ifa->ifa_address);
2153	}
2154	}
2155
2156	/*
2157	* IPv6
2158	*/
2159	in6_dev = __in6_dev_get(dev);
2160	if (in6_dev != NULL)
2161	{
2162	struct inet6_ifaddr *ifa;
2163
2164	read_lock_bh(&in6_dev->lock);
2165	#if RTLNX_VER_MIN(2,6,35)
2166	list_for_each_entry(ifa, &in6_dev->addr_list, if_list)
2167	#else
2168	for (ifa = in6_dev->addr_list; ifa != NULL; ifa = ifa->if_next)
2169	#endif
2170	{
2171	if ( dev != pThis->u.s.pDev
2172	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2173	continue;
2174
2175	Log(("%s: %s: IPv6 addr %RTnaipv6/%u\n",
2176	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2177	&ifa->addr, (unsigned)ifa->prefix_len));
2178
2179	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2180	/* :fAdded */ true, kIntNetAddrType_IPv6, &ifa->addr);
2181	}
2182	read_unlock_bh(&in6_dev->lock);
2183	}
2184
2185	return NOTIFY_OK;
2186	}
2187
2188
2189	static int vboxNetFltLinuxNotifierIPv4Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2190	{
2191	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv4);
2192	struct net_device pDev, pEventDev;
2193	struct in_ifaddr ifa = (struct in_ifaddr )ptr;
2194	bool fMyDev;
2195	int rc = NOTIFY_OK;
2196
2197	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2198	pEventDev = ifa->ifa_dev->dev;
2199	fMyDev = (pDev == pEventDev);
2200	Log(("VBoxNetFlt: %s: IPv4 event %s(0x%lx) %s: addr %RTnaipv4 mask %RTnaipv4\n",
2201	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2202	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2203	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2204	ifa->ifa_address, ifa->ifa_mask));
2205
2206	if (pDev != NULL)
2207	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2208
2209	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2210	return NOTIFY_OK;
2211
2212	if ( !fMyDev
2213	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2214	return NOTIFY_OK;
2215
2216	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2217	{
2218	bool fAdded;
2219	if (ulEventType == NETDEV_UP)
2220	fAdded = true;
2221	else if (ulEventType == NETDEV_DOWN)
2222	fAdded = false;
2223	else
2224	return NOTIFY_OK;
2225
2226	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2227	kIntNetAddrType_IPv4, &ifa->ifa_local);
2228	}
2229
2230	return rc;
2231	}
2232
2233
2234	static int vboxNetFltLinuxNotifierIPv6Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2235	{
2236	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv6);
2237	struct net_device pDev, pEventDev;
2238	struct inet6_ifaddr ifa = (struct inet6_ifaddr )ptr;
2239	bool fMyDev;
2240	int rc = NOTIFY_OK;
2241
2242	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2243	pEventDev = ifa->idev->dev;
2244	fMyDev = (pDev == pEventDev);
2245	Log(("VBoxNetFlt: %s: IPv6 event %s(0x%lx) %s: %RTnaipv6\n",
2246	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2247	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2248	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2249	&ifa->addr));
2250
2251	if (pDev != NULL)
2252	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2253
2254	if ( !fMyDev
2255	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2256	return NOTIFY_OK;
2257
2258	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2259	{
2260	bool fAdded;
2261	if (ulEventType == NETDEV_UP)
2262	fAdded = true;
2263	else if (ulEventType == NETDEV_DOWN)
2264	fAdded = false;
2265	else
2266	return NOTIFY_OK;
2267
2268	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2269	kIntNetAddrType_IPv6, &ifa->addr);
2270	}
2271
2272	return rc;
2273	}
2274
2275
2276	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2277	{
2278	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2279	}
2280
2281	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2282	{
2283	struct net_device * pDev;
2284	int err;
2285	int rc = VINF_SUCCESS;
2286	IPRT_LINUX_SAVE_EFL_AC();
2287	NOREF(pvIfData);
2288
2289	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2290
2291	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2292	if (pDev)
2293	{
2294	/*
2295	* Create a sk_buff for the gather list and push it onto the wire.
2296	*/
2297	if (fDst & INTNETTRUNKDIR_WIRE)
2298	{
2299	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2300	if (pBuf)
2301	{
2302	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2303	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2304	Log6(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2305	err = dev_queue_xmit(pBuf);
2306	if (err)
2307	rc = RTErrConvertFromErrno(err);
2308	}
2309	else
2310	rc = VERR_NO_MEMORY;
2311	}
2312
2313	/*
2314	* Create a sk_buff for the gather list and push it onto the host stack.
2315	*/
2316	if (fDst & INTNETTRUNKDIR_HOST)
2317	{
2318	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2319	if (pBuf)
2320	{
2321	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2322	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2323	Log6(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2324	#if RTLNX_VER_MIN(5,18,0) \|\| RTLNX_RHEL_MIN(9,1)
2325	local_bh_disable();
2326	err = netif_rx(pBuf);
2327	local_bh_enable();
2328	#else
2329	err = netif_rx_ni(pBuf);
2330	#endif
2331	if (err)
2332	rc = RTErrConvertFromErrno(err);
2333	}
2334	else
2335	rc = VERR_NO_MEMORY;
2336	}
2337
2338	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2339	}
2340
2341	IPRT_LINUX_RESTORE_EFL_AC();
2342	return rc;
2343	}
2344
2345
2346	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2347	{
2348	struct net_device *pDev;
2349	IPRT_LINUX_SAVE_EFL_AC();
2350
2351	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%RTbool, fDisablePromiscuous=%RTbool\n",
2352	pThis, pThis->szName, fActive, pThis->fDisablePromiscuous));
2353
2354	if (pThis->fDisablePromiscuous)
2355	return;
2356
2357	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2358	if (pDev)
2359	{
2360	/*
2361	* This api is a bit weird, the best reference is the code.
2362	*
2363	* Also, we have a bit or race conditions wrt the maintenance of
2364	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2365	*/
2366	#ifdef LOG_ENABLED
2367	u_int16_t fIf;
2368	unsigned const cPromiscBefore = pDev->promiscuity;
2369	#endif
2370	if (fActive)
2371	{
2372	Assert(!pThis->u.s.fPromiscuousSet);
2373
2374	rtnl_lock();
2375	dev_set_promiscuity(pDev, 1);
2376	rtnl_unlock();
2377	pThis->u.s.fPromiscuousSet = true;
2378	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2379	}
2380	else
2381	{
2382	if (pThis->u.s.fPromiscuousSet)
2383	{
2384	rtnl_lock();
2385	dev_set_promiscuity(pDev, -1);
2386	rtnl_unlock();
2387	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2388	}
2389	pThis->u.s.fPromiscuousSet = false;
2390
2391	#ifdef LOG_ENABLED
2392	fIf = dev_get_flags(pDev);
2393	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2394	#endif
2395	}
2396
2397	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2398	}
2399	IPRT_LINUX_RESTORE_EFL_AC();
2400	}
2401
2402
2403	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2404	{
2405	/*
2406	* Remove packet handler when we get disconnected from internal switch as
2407	* we don't want the handler to forward packets to disconnected switch.
2408	*/
2409	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
2410	{
2411	IPRT_LINUX_SAVE_EFL_AC();
2412	dev_remove_pack(&pThis->u.s.PacketType);
2413	Log(("vboxNetFltOsDisconnectIt: this=%p: Packet handler removed.\n", pThis));
2414	IPRT_LINUX_RESTORE_EFL_AC();
2415	}
2416	return VINF_SUCCESS;
2417	}
2418
2419
2420	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2421	{
2422	IPRT_LINUX_SAVE_EFL_AC();
2423
2424	/*
2425	* Report the GSO capabilities of the host and device (if connected).
2426	* Note! No need to mark ourselves busy here.
2427	*/
2428	/** @todo duplicate work here now? Attach */
2429	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2430	Log3(("vboxNetFltOsConnectIt: reporting host tso tso6\n"));
2431	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2432	0
2433	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2434	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2435	, INTNETTRUNKDIR_HOST);
2436
2437	#endif
2438	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2439
2440	IPRT_LINUX_RESTORE_EFL_AC();
2441	return VINF_SUCCESS;
2442	}
2443
2444
2445	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2446	{
2447	struct net_device *pDev;
2448	bool fRegistered;
2449	IPRT_LINUX_SAVE_EFL_AC();
2450
2451	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
2452	vboxNetFltLinuxUnhookDev(pThis, NULL);
2453	#endif
2454
2455	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2456	* unlikely, but none the less). Since it doesn't actually update the
2457	* state (just reads it), it is likely to panic in some interesting
2458	* ways. */
2459
2460	RTSpinlockAcquire(pThis->hSpinlock);
2461	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2462	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
2463	RTSpinlockRelease(pThis->hSpinlock);
2464
2465	if (fRegistered)
2466	{
2467	vboxNetFltSetLinkState(pThis, pDev, false);
2468
2469	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2470	skb_queue_purge(&pThis->u.s.XmitQueue);
2471	#endif
2472	Log(("vboxNetFltOsDeleteInstance: this=%p: xmit queue purged.\n", pThis));
2473	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2474	pDev, pDev->name,
2475	#if RTLNX_VER_MIN(2,6,37)
2476	netdev_refcnt_read(pDev)
2477	#else
2478	atomic_read(&pDev->refcnt)
2479	#endif
2480	));
2481	dev_put(pDev);
2482	}
2483
2484	unregister_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2485	unregister_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2486
2487	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2488	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2489	module_put(THIS_MODULE);
2490
2491	IPRT_LINUX_RESTORE_EFL_AC();
2492	}
2493
2494
2495	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2496	{
2497	int err;
2498	IPRT_LINUX_SAVE_EFL_AC();
2499	NOREF(pvContext);
2500
2501	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2502	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2503	if (err)
2504	{
2505	IPRT_LINUX_RESTORE_EFL_AC();
2506	return VERR_INTNET_FLT_IF_FAILED;
2507	}
2508	if (!pThis->u.s.fRegistered)
2509	{
2510	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2511	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2512	IPRT_LINUX_RESTORE_EFL_AC();
2513	return VERR_INTNET_FLT_IF_NOT_FOUND;
2514	}
2515
2516	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2517	if ( pThis->fDisconnectedFromHost
2518	\|\| !try_module_get(THIS_MODULE))
2519	{
2520	IPRT_LINUX_RESTORE_EFL_AC();
2521	return VERR_INTNET_FLT_IF_FAILED;
2522	}
2523
2524	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2525	{
2526	VBOXNETFLTNOTIFIER Enumerator;
2527
2528	/*
2529	* register_inetaddr_notifier() and register_inet6addr_notifier()
2530	* do not call the callback for existing devices. Enumerating
2531	* all network devices explicitly is a bit of an ifdef mess,
2532	* so co-opt register_netdevice_notifier() to do that for us.
2533	*/
2534	RT_ZERO(Enumerator);
2535	Enumerator.Notifier.notifier_call = vboxNetFltLinuxEnumeratorCallback;
2536	Enumerator.pThis = pThis;
2537
2538	err = register_netdevice_notifier(&Enumerator.Notifier);
2539	if (err)
2540	{
2541	LogRel(("%s: failed to enumerate network devices: error %d\n", __FUNCTION__, err));
2542	IPRT_LINUX_RESTORE_EFL_AC();
2543	return VINF_SUCCESS;
2544	}
2545
2546	unregister_netdevice_notifier(&Enumerator.Notifier);
2547
2548	pThis->u.s.NotifierIPv4.notifier_call = vboxNetFltLinuxNotifierIPv4Callback;
2549	err = register_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2550	if (err)
2551	LogRel(("%s: failed to register IPv4 notifier: error %d\n", __FUNCTION__, err));
2552
2553	pThis->u.s.NotifierIPv6.notifier_call = vboxNetFltLinuxNotifierIPv6Callback;
2554	err = register_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2555	if (err)
2556	LogRel(("%s: failed to register IPv6 notifier: error %d\n", __FUNCTION__, err));
2557	}
2558
2559	IPRT_LINUX_RESTORE_EFL_AC();
2560	return VINF_SUCCESS;
2561	}
2562
2563	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2564	{
2565	IPRT_LINUX_SAVE_EFL_AC();
2566
2567	/*
2568	* Init the linux specific members.
2569	*/
2570	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2571	pThis->u.s.fRegistered = false;
2572	pThis->u.s.fPromiscuousSet = false;
2573	pThis->u.s.fPacketHandler = false;
2574	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2575	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2576	skb_queue_head_init(&pThis->u.s.XmitQueue);
2577	# if RTLNX_VER_MIN(2,6,20)
2578	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2579	# else
2580	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2581	# endif
2582	#endif
2583
2584	IPRT_LINUX_RESTORE_EFL_AC();
2585	return VINF_SUCCESS;
2586	}
2587
2588
2589	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2590	{
2591	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2592	}
2593
2594
2595	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2596	{
2597	/* Nothing to do */
2598	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2599	return VINF_SUCCESS;
2600	}
2601
2602
2603	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2604	{
2605	/* Nothing to do */
2606	NOREF(pThis); NOREF(pvIfData);
2607	return VINF_SUCCESS;
2608	}
2609

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

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 100862

Download in other formats: