VirtualBox

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 73892

Last change on this file since 73892 was 73097, checked in by vboxsync, 7 years ago

*: Made RT_UOFFSETOF, RT_OFFSETOF, RT_UOFFSETOF_ADD and RT_OFFSETOF_ADD work like builtin_offsetof() and require compile time resolvable requests, adding RT_UOFFSETOF_DYN for the dynamic questions that can only be answered at runtime.
  • Property svn:eol-style set to native
  • Property svn:keywords set to Author Date Id Revision
File size: 91.2 KB
Line 
1/* $Id: VBoxNetFlt-linux.c 73097 2018-07-12 21:06:33Z vboxsync $ */
2/** @file
3 * VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4 */
5
6/*
7 * Copyright (C) 2006-2017 Oracle Corporation
8 *
9 * This file is part of VirtualBox Open Source Edition (OSE), as
10 * available from http://www.virtualbox.org. This file is free software;
11 * you can redistribute it and/or modify it under the terms of the GNU
12 * General Public License (GPL) as published by the Free Software
13 * Foundation, in version 2 as it comes in the "COPYING" file of the
14 * VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15 * hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16 *
17 * The contents of this file may alternatively be used under the terms
18 * of the Common Development and Distribution License Version 1.0
19 * (CDDL) only, as it comes in the "COPYING.CDDL" file of the
20 * VirtualBox OSE distribution, in which case the provisions of the
21 * CDDL are applicable instead of those of the GPL.
22 *
23 * You may elect to license modified versions of this file under the
24 * terms and conditions of either the GPL or the CDDL or both.
25 */
26
27
28/*********************************************************************************************************************************
29* Header Files *
30*********************************************************************************************************************************/
31#define LOG_GROUP LOG_GROUP_NET_FLT_DRV
32#define VBOXNETFLT_LINUX_NO_XMIT_QUEUE
33#include "the-linux-kernel.h"
34#include "version-generated.h"
35#include "revision-generated.h"
36#include "product-generated.h"
37#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 24)
38#include <linux/nsproxy.h>
39#endif
40#include <linux/netdevice.h>
41#include <linux/etherdevice.h>
42#include <linux/rtnetlink.h>
43#include <linux/miscdevice.h>
44#include <linux/inetdevice.h>
45#include <linux/in.h>
46#include <linux/ip.h>
47#include <linux/if_vlan.h>
48#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 5, 0)
49#include <uapi/linux/pkt_cls.h>
50#endif
51#include <net/ipv6.h>
52#include <net/if_inet6.h>
53#include <net/addrconf.h>
54
55#include <VBox/log.h>
56#include <VBox/err.h>
57#include <VBox/intnetinline.h>
58#include <VBox/vmm/pdmnetinline.h>
59#include <VBox/param.h>
60#include <iprt/alloca.h>
61#include <iprt/assert.h>
62#include <iprt/spinlock.h>
63#include <iprt/semaphore.h>
64#include <iprt/initterm.h>
65#include <iprt/process.h>
66#include <iprt/mem.h>
67#include <iprt/net.h>
68#include <iprt/log.h>
69#include <iprt/mp.h>
70#include <iprt/mem.h>
71#include <iprt/time.h>
72
73#define VBOXNETFLT_OS_SPECFIC 1
74#include "../VBoxNetFltInternal.h"
75
76typedef struct VBOXNETFLTNOTIFIER {
77 struct notifier_block Notifier;
78 PVBOXNETFLTINS pThis;
79} VBOXNETFLTNOTIFIER;
80typedef struct VBOXNETFLTNOTIFIER *PVBOXNETFLTNOTIFIER;
81
82
83/*********************************************************************************************************************************
84* Defined Constants And Macros *
85*********************************************************************************************************************************/
86#define VBOX_FLT_NB_TO_INST(pNB) RT_FROM_MEMBER(pNB, VBOXNETFLTINS, u.s.Notifier)
87#define VBOX_FLT_PT_TO_INST(pPT) RT_FROM_MEMBER(pPT, VBOXNETFLTINS, u.s.PacketType)
88#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
89# define VBOX_FLT_XT_TO_INST(pXT) RT_FROM_MEMBER(pXT, VBOXNETFLTINS, u.s.XmitTask)
90#endif
91
92#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0)
93# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) netdev_notifier_info_to_dev(ptr)
94#else
95# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) ((struct net_device *)ptr)
96#endif
97
98#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 5, 0)
99# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag))
100# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr)
101#else
102# if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 2, 0)
103# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag), KM_SKB_DATA_SOFTIRQ)
104# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
105# else
106# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ)
107# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
108# endif
109#endif
110
111#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 34)
112# define VBOX_NETDEV_NAME(dev) netdev_name(dev)
113#else
114# define VBOX_NETDEV_NAME(dev) ((dev)->reg_state != NETREG_REGISTERED ? "(unregistered net_device)" : (dev)->name)
115#endif
116
117#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 25)
118# define VBOX_IPV4_IS_LOOPBACK(addr) ipv4_is_loopback(addr)
119# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ipv4_is_linklocal_169(addr)
120#else
121# define VBOX_IPV4_IS_LOOPBACK(addr) ((addr & htonl(IN_CLASSA_NET)) == htonl(0x7f000000))
122# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ((addr & htonl(IN_CLASSB_NET)) == htonl(0xa9fe0000))
123#endif
124
125#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
126# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
127# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
128# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum_offset
129#else
130# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
131# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
132# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum
133#endif
134
135#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 19)
136# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
137#else
138# define CHECKSUM_PARTIAL CHECKSUM_HW
139# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 10)
140# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
141# else
142# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 7)
143# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
144# else
145# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
146# endif
147/* Versions prior 2.6.10 use stats for both bstats and qstats */
148# define bstats stats
149# define qstats stats
150# endif
151#endif
152
153#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 20, 0)
154# define VBOX_HAVE_SKB_VLAN
155#else
156# ifdef RHEL_RELEASE_CODE
157# if (RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(7, 2) && RHEL_RELEASE_CODE < RHEL_RELEASE_VERSION(8, 0)) || \
158 (RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(6, 8) && RHEL_RELEASE_CODE < RHEL_RELEASE_VERSION(7, 0))
159# define VBOX_HAVE_SKB_VLAN
160# endif
161# endif
162#endif
163
164#ifdef VBOX_HAVE_SKB_VLAN
165# define vlan_tx_tag_get(skb) skb_vlan_tag_get(skb)
166# define vlan_tx_tag_present(skb) skb_vlan_tag_present(skb)
167#endif
168
169#ifndef NET_IP_ALIGN
170# define NET_IP_ALIGN 2
171#endif
172
173#if 1
174/** Create scatter / gather segments for fragments. When not used, we will
175 * linearize the socket buffer before creating the internal networking SG. */
176# define VBOXNETFLT_SG_SUPPORT 1
177#endif
178
179#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
180
181/** Indicates that the linux kernel may send us GSO frames. */
182# define VBOXNETFLT_WITH_GSO 1
183
184/** This enables or disables the transmitting of GSO frame from the internal
185 * network and to the host. */
186# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
187
188# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
189/** This enables or disables the transmitting of GSO frame from the internal
190 * network and to the wire. */
191# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
192# endif
193
194/** This enables or disables the forwarding/flooding of GSO frame from the host
195 * to the internal network. */
196# define VBOXNETFLT_WITH_GSO_RECV 1
197
198#endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18) */
199
200#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
201/** This enables or disables handling of GSO frames coming from the wire (GRO). */
202# define VBOXNETFLT_WITH_GRO 1
203#endif
204
205/*
206 * GRO support was backported to RHEL 5.4
207 */
208#ifdef RHEL_RELEASE_CODE
209# if RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(5, 4)
210# define VBOXNETFLT_WITH_GRO 1
211# endif
212#endif
213
214
215/*********************************************************************************************************************************
216* Internal Functions *
217*********************************************************************************************************************************/
218static int VBoxNetFltLinuxInit(void);
219static void VBoxNetFltLinuxUnload(void);
220static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf);
221
222
223/*********************************************************************************************************************************
224* Global Variables *
225*********************************************************************************************************************************/
226/**
227 * The (common) global data.
228 */
229static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
230
231module_init(VBoxNetFltLinuxInit);
232module_exit(VBoxNetFltLinuxUnload);
233
234MODULE_AUTHOR(VBOX_VENDOR);
235MODULE_DESCRIPTION(VBOX_PRODUCT " Network Filter Driver");
236MODULE_LICENSE("GPL");
237#ifdef MODULE_VERSION
238MODULE_VERSION(VBOX_VERSION_STRING " r" RT_XSTR(VBOX_SVN_REV) " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
239#endif
240
241
242#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) && defined(LOG_ENABLED)
243unsigned dev_get_flags(const struct net_device *dev)
244{
245 unsigned flags;
246
247 flags = (dev->flags & ~(IFF_PROMISC |
248 IFF_ALLMULTI |
249 IFF_RUNNING)) |
250 (dev->gflags & (IFF_PROMISC |
251 IFF_ALLMULTI));
252
253 if (netif_running(dev) && netif_carrier_ok(dev))
254 flags |= IFF_RUNNING;
255
256 return flags;
257}
258#endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 12) */
259
260
261/**
262 * Initialize module.
263 *
264 * @returns appropriate status code.
265 */
266static int __init VBoxNetFltLinuxInit(void)
267{
268 int rc;
269 /*
270 * Initialize IPRT.
271 */
272 rc = RTR0Init(0);
273 if (RT_SUCCESS(rc))
274 {
275 Log(("VBoxNetFltLinuxInit\n"));
276
277 /*
278 * Initialize the globals and connect to the support driver.
279 *
280 * This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
281 * for establishing the connect to the support driver.
282 */
283 memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
284 rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
285 if (RT_SUCCESS(rc))
286 {
287 LogRel(("VBoxNetFlt: Successfully started.\n"));
288 return 0;
289 }
290
291 LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
292 RTR0Term();
293 }
294 else
295 LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
296
297 memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
298 return -RTErrConvertToErrno(rc);
299}
300
301
302/**
303 * Unload the module.
304 *
305 * @todo We have to prevent this if we're busy!
306 */
307static void __exit VBoxNetFltLinuxUnload(void)
308{
309 int rc;
310 Log(("VBoxNetFltLinuxUnload\n"));
311 Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
312
313 /*
314 * Undo the work done during start (in reverse order).
315 */
316 rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
317 AssertRC(rc); NOREF(rc);
318
319 RTR0Term();
320
321 memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
322
323 Log(("VBoxNetFltLinuxUnload - done\n"));
324}
325
326
327/**
328 * We filter traffic from the host to the internal network
329 * before it reaches the NIC driver.
330 *
331 * The current code uses a very ugly hack overriding hard_start_xmit
332 * callback in the device structure, but it has been shown to give us a
333 * performance boost of 60-100% though. Eventually we have to find some
334 * less hacky way of getting this job done.
335 */
336#define VBOXNETFLT_WITH_HOST2WIRE_FILTER
337
338#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
339
340# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
341
342# include <linux/ethtool.h>
343
344typedef struct ethtool_ops OVR_OPSTYPE;
345# define OVR_OPS ethtool_ops
346# define OVR_XMIT pfnStartXmit
347
348# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
349
350typedef struct net_device_ops OVR_OPSTYPE;
351# define OVR_OPS netdev_ops
352# define OVR_XMIT pOrgOps->ndo_start_xmit
353
354# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
355
356/**
357 * The overridden net_device_ops of the device we're attached to.
358 *
359 * As there is no net_device_ops structure in pre-2.6.29 kernels we override
360 * ethtool_ops instead along with hard_start_xmit callback in net_device
361 * structure.
362 *
363 * This is a very dirty hack that was created to explore how much we can improve
364 * the host to guest transfers by not CC'ing the NIC. It turns out to be
365 * the only way to filter outgoing packets for devices without TX queue.
366 */
367typedef struct VBoxNetDeviceOpsOverride
368{
369 /** Our overridden ops. */
370 OVR_OPSTYPE Ops;
371 /** Magic word. */
372 uint32_t u32Magic;
373 /** Pointer to the original ops. */
374 OVR_OPSTYPE const *pOrgOps;
375# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
376 /** Pointer to the original hard_start_xmit function. */
377 int (*pfnStartXmit)(struct sk_buff *pSkb, struct net_device *pDev);
378# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
379 /** Pointer to the net filter instance. */
380 PVBOXNETFLTINS pVBoxNetFlt;
381 /** The number of filtered packages. */
382 uint64_t cFiltered;
383 /** The total number of packets */
384 uint64_t cTotal;
385} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
386/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
387#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
388
389/**
390 * ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
391 * because they belong on the internal network.
392 *
393 * @returns NETDEV_TX_XXX.
394 * @param pSkb The socket buffer to transmit.
395 * @param pDev The net device.
396 */
397static int vboxNetFltLinuxStartXmitFilter(struct sk_buff *pSkb, struct net_device *pDev)
398{
399 PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
400 uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
401 PCRTNETETHERHDR pEtherHdr;
402 PINTNETTRUNKSWPORT pSwitchPort;
403 uint32_t cbHdrs;
404
405
406 /*
407 * Validate the override structure.
408 *
409 * Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
410 * to be production quality code, we would have to be much more
411 * careful here and avoid the race.
412 */
413 if ( !VALID_PTR(pOverride)
414 || pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
415# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
416 || !VALID_PTR(pOverride->pOrgOps)
417# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
418 )
419 {
420 printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
421 dev_kfree_skb(pSkb);
422 return NETDEV_TX_OK;
423 }
424 pOverride->cTotal++;
425
426 /*
427 * Do the filtering base on the default OUI of our virtual NICs
428 *
429 * Note! In a real solution, we would ask the switch whether the
430 * destination MAC is 100% to be on the internal network and then
431 * drop it.
432 */
433 cbHdrs = skb_headlen(pSkb);
434 cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
435 pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
436 if ( pEtherHdr
437 && VALID_PTR(pOverride->pVBoxNetFlt)
438 && (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
439 && VALID_PTR(pSwitchPort)
440 && cbHdrs >= 6)
441 {
442 INTNETSWDECISION enmDecision;
443
444 /** @todo consider reference counting, etc. */
445 enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
446 if (enmDecision == INTNETSWDECISION_INTNET)
447 {
448 dev_kfree_skb(pSkb);
449 pOverride->cFiltered++;
450 return NETDEV_TX_OK;
451 }
452 }
453
454 return pOverride->OVR_XMIT(pSkb, pDev);
455}
456
457/**
458 * Hooks the device ndo_start_xmit operation of the device.
459 *
460 * @param pThis The net filter instance.
461 * @param pDev The net device.
462 */
463static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
464{
465 PVBOXNETDEVICEOPSOVERRIDE pOverride;
466
467 /* Cancel override if ethtool_ops is missing (host-only case, @bugref{5712}) */
468 if (!VALID_PTR(pDev->OVR_OPS))
469 return;
470 pOverride = RTMemAlloc(sizeof(*pOverride));
471 if (!pOverride)
472 return;
473 pOverride->pOrgOps = pDev->OVR_OPS;
474 pOverride->Ops = *pDev->OVR_OPS;
475# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
476 pOverride->pfnStartXmit = pDev->hard_start_xmit;
477# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
478 pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
479# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29) */
480 pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
481 pOverride->cTotal = 0;
482 pOverride->cFiltered = 0;
483 pOverride->pVBoxNetFlt = pThis;
484
485 RTSpinlockAcquire(pThis->hSpinlock); /* (this isn't necessary, but so what) */
486 ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
487# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
488 ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
489# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
490 RTSpinlockRelease(pThis->hSpinlock);
491}
492
493/**
494 * Undos what vboxNetFltLinuxHookDev did.
495 *
496 * @param pThis The net filter instance.
497 * @param pDev The net device. Can be NULL, in which case
498 * we'll try retrieve it from @a pThis.
499 */
500static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
501{
502 PVBOXNETDEVICEOPSOVERRIDE pOverride;
503
504 RTSpinlockAcquire(pThis->hSpinlock);
505 if (!pDev)
506 pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
507 if (VALID_PTR(pDev))
508 {
509 pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
510 if ( VALID_PTR(pOverride)
511 && pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
512 && VALID_PTR(pOverride->pOrgOps)
513 )
514 {
515# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
516 ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
517# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29) */
518 ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
519 ASMAtomicWriteU32(&pOverride->u32Magic, 0);
520 }
521 else
522 pOverride = NULL;
523 }
524 else
525 pOverride = NULL;
526 RTSpinlockRelease(pThis->hSpinlock);
527
528 if (pOverride)
529 {
530 printk("vboxnetflt: %llu out of %llu packets were not sent (directed to host)\n", pOverride->cFiltered, pOverride->cTotal);
531 RTMemFree(pOverride);
532 }
533}
534
535#endif /* VBOXNETFLT_WITH_HOST2WIRE_FILTER */
536
537
538/**
539 * Reads and retains the host interface handle.
540 *
541 * @returns The handle, NULL if detached.
542 * @param pThis
543 */
544DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
545{
546#if 0
547 struct net_device *pDev = NULL;
548
549 Log(("vboxNetFltLinuxRetainNetDev\n"));
550 /*
551 * Be careful here to avoid problems racing the detached callback.
552 */
553 RTSpinlockAcquire(pThis->hSpinlock);
554 if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
555 {
556 pDev = (struct net_device *)ASMAtomicUoReadPtr((void * volatile *)&pThis->u.s.pDev);
557 if (pDev)
558 {
559 dev_hold(pDev);
560 Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
561 pDev, pDev->name,
562#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
563 netdev_refcnt_read(pDev)
564#else
565 atomic_read(&pDev->refcnt)
566#endif
567 ));
568 }
569 }
570 RTSpinlockRelease(pThis->hSpinlock);
571
572 Log(("vboxNetFltLinuxRetainNetDev - done\n"));
573 return pDev;
574#else
575 return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
576#endif
577}
578
579
580/**
581 * Release the host interface handle previously retained
582 * by vboxNetFltLinuxRetainNetDev.
583 *
584 * @param pThis The instance.
585 * @param pDev The vboxNetFltLinuxRetainNetDev
586 * return value, NULL is fine.
587 */
588DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
589{
590#if 0
591 Log(("vboxNetFltLinuxReleaseNetDev\n"));
592 NOREF(pThis);
593 if (pDev)
594 {
595 dev_put(pDev);
596 Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
597 pDev, pDev->name,
598#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
599 netdev_refcnt_read(pDev)
600#else
601 atomic_read(&pDev->refcnt)
602#endif
603 ));
604 }
605 Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
606#endif
607}
608
609#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 | (skb->dev->ifindex & 0xFFFF))
610#define VBOXNETFLT_SKB_TAG(skb) (*(uint32_t*)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
611
612/**
613 * Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
614 * i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
615 *
616 * @returns true / false accordingly.
617 * @param pBuf The sk_buff.
618 */
619DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
620{
621 return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
622}
623
624
625/**
626 * Checks whether this SG list contains a GSO packet.
627 *
628 * @returns true / false accordingly.
629 * @param pSG The (scatter/)gather list.
630 */
631DECLINLINE(bool) vboxNetFltLinuxIsGso(PINTNETSG pSG)
632{
633#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) || defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
634 return !((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type == PDMNETWORKGSOTYPE_INVALID);
635#else /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
636 return false;
637#endif /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
638}
639
640
641/**
642 * Find out the frame size (of a single segment in case of GSO frames).
643 *
644 * @returns the frame size.
645 * @param pSG The (scatter/)gather list.
646 */
647DECLINLINE(uint32_t) vboxNetFltLinuxFrameSize(PINTNETSG pSG)
648{
649 uint16_t u16Type = 0;
650 uint32_t cbVlanTag = 0;
651 if (pSG->aSegs[0].cb >= sizeof(RTNETETHERHDR))
652 u16Type = RT_BE2H_U16(((PCRTNETETHERHDR)pSG->aSegs[0].pv)->EtherType);
653 else if (pSG->cbTotal >= sizeof(RTNETETHERHDR))
654 {
655 uint32_t off = RT_UOFFSETOF(RTNETETHERHDR, EtherType);
656 uint32_t i;
657 for (i = 0; i < pSG->cSegsUsed; ++i)
658 {
659 if (off <= pSG->aSegs[i].cb)
660 {
661 if (off + sizeof(uint16_t) <= pSG->aSegs[i].cb)
662 u16Type = RT_BE2H_U16(*(uint16_t *)((uintptr_t)pSG->aSegs[i].pv + off));
663 else if (i + 1 < pSG->cSegsUsed)
664 u16Type = RT_BE2H_U16( ((uint16_t)( ((uint8_t *)pSG->aSegs[i].pv)[off] ) << 8)
665 + *(uint8_t *)pSG->aSegs[i + 1].pv); /* ASSUMES no empty segments! */
666 /* else: frame is too short. */
667 break;
668 }
669 off -= pSG->aSegs[i].cb;
670 }
671 }
672 if (u16Type == RTNET_ETHERTYPE_VLAN)
673 cbVlanTag = 4;
674 return (vboxNetFltLinuxIsGso(pSG) ? (uint32_t)pSG->GsoCtx.cbMaxSeg + pSG->GsoCtx.cbHdrsTotal : pSG->cbTotal) - cbVlanTag;
675}
676
677
678/**
679 * Internal worker that create a linux sk_buff for a
680 * (scatter/)gather list.
681 *
682 * @returns Pointer to the sk_buff.
683 * @param pThis The instance.
684 * @param pSG The (scatter/)gather list.
685 * @param fDstWire Set if the destination is the wire.
686 */
687static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
688{
689 struct sk_buff *pPkt;
690 struct net_device *pDev;
691 unsigned fGsoType = 0;
692
693 if (pSG->cbTotal == 0)
694 {
695 LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
696 return NULL;
697 }
698 Log5(("VBoxNetFlt: Packet to %s of %d bytes (frame=%d).\n", fDstWire?"wire":"host", pSG->cbTotal, vboxNetFltLinuxFrameSize(pSG)));
699 if (fDstWire && (vboxNetFltLinuxFrameSize(pSG) > ASMAtomicReadU32(&pThis->u.s.cbMtu) + 14))
700 {
701 static bool s_fOnce = true;
702 if (s_fOnce)
703 {
704 s_fOnce = false;
705 printk("VBoxNetFlt: Dropped over-sized packet (%d bytes) coming from internal network.\n", vboxNetFltLinuxFrameSize(pSG));
706 }
707 return NULL;
708 }
709
710 /** @todo We should use fragments mapping the SG buffers with large packets.
711 * 256 bytes seems to be the a threshold used a lot for this. It
712 * requires some nasty work on the intnet side though... */
713 /*
714 * Allocate a packet and copy over the data.
715 */
716 pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
717 pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
718 if (RT_UNLIKELY(!pPkt))
719 {
720 Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
721 pSG->pvUserData = NULL;
722 return NULL;
723 }
724 pPkt->dev = pDev;
725 pPkt->ip_summed = CHECKSUM_NONE;
726
727 /* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
728 skb_reserve(pPkt, NET_IP_ALIGN);
729
730 /* Copy the segments. */
731 skb_put(pPkt, pSG->cbTotal);
732 IntNetSgRead(pSG, pPkt->data);
733
734#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) || defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
735 /*
736 * Setup GSO if used by this packet.
737 */
738 switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
739 {
740 default:
741 AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
742 /* fall thru */
743 case PDMNETWORKGSOTYPE_INVALID:
744 fGsoType = 0;
745 break;
746 case PDMNETWORKGSOTYPE_IPV4_TCP:
747 fGsoType = SKB_GSO_TCPV4;
748 break;
749 case PDMNETWORKGSOTYPE_IPV6_TCP:
750 fGsoType = SKB_GSO_TCPV6;
751 break;
752 }
753 if (fGsoType)
754 {
755 struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
756
757 pShInfo->gso_type = fGsoType | SKB_GSO_DODGY;
758 pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
759 pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
760
761 /*
762 * We need to set checksum fields even if the packet goes to the host
763 * directly as it may be immediately forwarded by IP layer @bugref{5020}.
764 */
765 Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrsTotal);
766 pPkt->ip_summed = CHECKSUM_PARTIAL;
767# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
768 pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
769 if (fGsoType & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6))
770 pPkt->csum_offset = RT_UOFFSETOF(RTNETTCP, th_sum);
771 else
772 pPkt->csum_offset = RT_UOFFSETOF(RTNETUDP, uh_sum);
773# else
774 pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
775 if (fGsoType & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6))
776 pPkt->csum = RT_UOFFSETOF(RTNETTCP, th_sum);
777 else
778 pPkt->csum = RT_UOFFSETOF(RTNETUDP, uh_sum);
779# endif
780 if (!fDstWire)
781 PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
782 }
783#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE || VBOXNETFLT_WITH_GSO_XMIT_HOST */
784
785 /*
786 * Finish up the socket buffer.
787 */
788 pPkt->protocol = eth_type_trans(pPkt, pDev);
789 if (fDstWire)
790 {
791 VBOX_SKB_RESET_NETWORK_HDR(pPkt);
792
793 /* Restore ethernet header back. */
794 skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
795 VBOX_SKB_RESET_MAC_HDR(pPkt);
796 }
797 VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
798
799 return pPkt;
800}
801
802
803/**
804 * Return the offset where to start checksum computation from.
805 *
806 * @returns the offset relative to pBuf->data.
807 * @param pBuf The socket buffer.
808 */
809DECLINLINE(unsigned) vboxNetFltLinuxGetChecksumStartOffset(struct sk_buff *pBuf)
810{
811# if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 21)
812 unsigned char *pTransportHdr = pBuf->h.raw;
813# if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
814 /*
815 * Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
816 * and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
817 * header length from the header itself and reconstruct 'h' pointer
818 * to TCP (or whatever) header.
819 */
820 if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
821 pTransportHdr = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
822# endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
823 return pTransportHdr - pBuf->data;
824# else /* LINUX_VERSION_CODE > KERNEL_VERSION(2, 6, 21) */
825# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 38)
826 return pBuf->csum_start - skb_headroom(pBuf);
827# else /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 38) */
828 return skb_checksum_start_offset(pBuf);
829# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 38) */
830# endif /* LINUX_VERSION_CODE > KERNEL_VERSION(2, 6, 21) */
831}
832
833
834/**
835 * Initializes a SG list from an sk_buff.
836 *
837 * @returns Number of segments.
838 * @param pThis The instance.
839 * @param pBuf The sk_buff.
840 * @param pSG The SG.
841 * @param cbExtra The number of bytes of extra space allocated immediately after the SG.
842 * @param cSegs The number of segments allocated for the SG.
843 * This should match the number in the mbuf exactly!
844 * @param fSrc The source of the frame.
845 * @param pGsoCtx Pointer to the GSO context if it's a GSO
846 * internal network frame. NULL if regular frame.
847 */
848static void vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
849 unsigned cbExtra, unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
850{
851 int i;
852 NOREF(pThis);
853
854#ifndef VBOXNETFLT_SG_SUPPORT
855 Assert(!skb_shinfo(pBuf)->frag_list);
856#else /* VBOXNETFLT_SG_SUPPORT */
857 uint8_t *pExtra = (uint8_t *)&pSG->aSegs[cSegs];
858 unsigned cbConsumed = 0;
859 unsigned cbProduced = 0;
860
861# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)
862 /* Restore VLAN tag stripped by host hardware */
863 if (vlan_tx_tag_present(pBuf))
864 {
865 uint8_t *pMac = pBuf->data;
866 struct vlan_ethhdr *pVHdr = (struct vlan_ethhdr *)pExtra;
867 Assert(ETH_ALEN * 2 + VLAN_HLEN <= cbExtra);
868 memmove(pVHdr, pMac, ETH_ALEN * 2);
869 /* Consume whole Ethernet header: 2 addresses + EtherType (see @bugref{8599}) */
870 cbConsumed += ETH_ALEN * 2 + sizeof(uint16_t);
871 pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
872 pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
873 pVHdr->h_vlan_encapsulated_proto = *(uint16_t*)(pMac + ETH_ALEN * 2);
874 cbProduced += VLAN_ETH_HLEN;
875 }
876# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
877
878 if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
879 {
880 unsigned uCsumStartOffset = vboxNetFltLinuxGetChecksumStartOffset(pBuf);
881 unsigned uCsumStoreOffset = uCsumStartOffset + VBOX_SKB_CSUM_OFFSET(pBuf) - cbConsumed;
882 Log3(("cbConsumed=%u cbProduced=%u uCsumStartOffset=%u uCsumStoreOffset=%u\n",
883 cbConsumed, cbProduced, uCsumStartOffset, uCsumStoreOffset));
884 Assert(cbProduced + uCsumStoreOffset + sizeof(uint16_t) <= cbExtra);
885 /*
886 * We assume that the checksum is stored at the very end of the transport header
887 * so we will have all headers in a single fragment. If our assumption is wrong
888 * we may see suboptimal performance.
889 */
890 memmove(pExtra + cbProduced,
891 pBuf->data + cbConsumed,
892 uCsumStoreOffset);
893 unsigned uChecksum = skb_checksum(pBuf, uCsumStartOffset, pBuf->len - uCsumStartOffset, 0);
894 *(uint16_t*)(pExtra + cbProduced + uCsumStoreOffset) = csum_fold(uChecksum);
895 cbProduced += uCsumStoreOffset + sizeof(uint16_t);
896 cbConsumed += uCsumStoreOffset + sizeof(uint16_t);
897 }
898#endif /* VBOXNETFLT_SG_SUPPORT */
899
900 if (!pGsoCtx)
901 IntNetSgInitTempSegs(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /*cSegsUsed*/);
902 else
903 IntNetSgInitTempSegsGso(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /*cSegsUsed*/, pGsoCtx);
904
905 int iSeg = 0;
906#ifdef VBOXNETFLT_SG_SUPPORT
907 if (cbProduced)
908 {
909 pSG->aSegs[iSeg].cb = cbProduced;
910 pSG->aSegs[iSeg].pv = pExtra;
911 pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
912 }
913 pSG->aSegs[iSeg].cb = skb_headlen(pBuf) - cbConsumed;
914 pSG->aSegs[iSeg].pv = pBuf->data + cbConsumed;
915 pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
916 Assert(iSeg <= pSG->cSegsAlloc);
917
918# ifdef LOG_ENABLED
919 if (pBuf->data_len)
920 Log6((" kmap_atomic:"));
921# endif /* LOG_ENABLED */
922 for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
923 {
924 skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
925 pSG->aSegs[iSeg].cb = pFrag->size;
926 pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
927 Log6((" %p", pSG->aSegs[iSeg].pv));
928 pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
929 Assert(iSeg <= pSG->cSegsAlloc);
930 }
931 struct sk_buff *pFragBuf;
932 for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
933 {
934 pSG->aSegs[iSeg].cb = skb_headlen(pFragBuf);
935 pSG->aSegs[iSeg].pv = pFragBuf->data;
936 pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
937 Assert(iSeg <= pSG->cSegsAlloc);
938 for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
939 {
940 skb_frag_t *pFrag = &skb_shinfo(pFragBuf)->frags[i];
941 pSG->aSegs[iSeg].cb = pFrag->size;
942 pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + pFrag->page_offset;
943 Log6((" %p", pSG->aSegs[iSeg].pv));
944 pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
945 Assert(iSeg <= pSG->cSegsAlloc);
946 }
947 }
948# ifdef LOG_ENABLED
949 if (pBuf->data_len)
950 Log6(("\n"));
951# endif /* LOG_ENABLED */
952#else
953 pSG->aSegs[iSeg].cb = pBuf->len;
954 pSG->aSegs[iSeg].pv = pBuf->data;
955 pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
956#endif
957
958 pSG->cSegsUsed = iSeg;
959
960#if 0
961 if (cbProduced)
962 {
963 LogRel(("vboxNetFltLinuxSkBufToSG: original packet dump:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
964 LogRel(("vboxNetFltLinuxSkBufToSG: cbConsumed=%u cbProduced=%u cbExtra=%u\n", cbConsumed, cbProduced, cbExtra));
965 uint32_t offset = 0;
966 for (i = 0; i < pSG->cSegsUsed; ++i)
967 {
968 LogRel(("vboxNetFltLinuxSkBufToSG: seg#%d (%d bytes, starting at 0x%x):\n%.*Rhxd\n",
969 i, pSG->aSegs[i].cb, offset, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
970 offset += pSG->aSegs[i].cb;
971 }
972 }
973#endif
974
975#ifdef PADD_RUNT_FRAMES_FROM_HOST
976 /*
977 * Add a trailer if the frame is too small.
978 *
979 * Since we're getting to the packet before it is framed, it has not
980 * yet been padded. The current solution is to add a segment pointing
981 * to a buffer containing all zeros and pray that works for all frames...
982 */
983 if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
984 {
985 Assert(pBuf->data_len == 0); /* Packets with fragments are never small! */
986 static uint8_t const s_abZero[128] = {0};
987
988 AssertReturnVoid(iSeg < cSegs);
989
990 pSG->aSegs[iSeg].Phys = NIL_RTHCPHYS;
991 pSG->aSegs[iSeg].pv = (void *)&s_abZero[0];
992 pSG->aSegs[iSeg++].cb = 60 - pSG->cbTotal;
993 pSG->cbTotal = 60;
994 pSG->cSegsUsed++;
995 Assert(iSeg <= pSG->cSegsAlloc)
996 }
997#endif
998
999 Log6(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1000 pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1001 for (i = 0; i < pSG->cSegsUsed; i++)
1002 Log6(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1003 i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1004}
1005
1006/**
1007 * Packet handler; not really documented - figure it out yourself.
1008 *
1009 * @returns 0 or EJUSTRETURN - this is probably copy & pastry and thus wrong.
1010 */
1011#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 14)
1012static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1013 struct net_device *pSkbDev,
1014 struct packet_type *pPacketType,
1015 struct net_device *pOrigDev)
1016#else
1017static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1018 struct net_device *pSkbDev,
1019 struct packet_type *pPacketType)
1020#endif
1021{
1022 PVBOXNETFLTINS pThis;
1023 struct net_device *pDev;
1024 LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1025 pBuf, pSkbDev, pPacketType));
1026#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1027 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",
1028 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));
1029# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1030 Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1031# endif
1032#else
1033 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",
1034 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));
1035#endif
1036 /*
1037 * Drop it immediately?
1038 */
1039 if (!pBuf)
1040 return 0;
1041
1042 if (pBuf->pkt_type == PACKET_LOOPBACK)
1043 {
1044 /*
1045 * We are not interested in loopbacked packets as they will always have
1046 * another copy going to the wire.
1047 */
1048 Log2(("vboxNetFltLinuxPacketHandler: dropped loopback packet (cb=%u)\n", pBuf->len));
1049 dev_kfree_skb(pBuf); /* We must 'consume' all packets we get (@bugref{6539})! */
1050 return 0;
1051 }
1052
1053 pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1054 pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1055 if (pDev != pSkbDev)
1056 {
1057 Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1058 kfree_skb(pBuf); /* This is a failure, so we use kfree_skb instead of dev_kfree_skb. */
1059 return 0;
1060 }
1061
1062 Log6(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1063 if (vboxNetFltLinuxSkBufIsOur(pBuf))
1064 {
1065 Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1066 dev_kfree_skb(pBuf);
1067 return 0;
1068 }
1069
1070#ifndef VBOXNETFLT_SG_SUPPORT
1071 {
1072 /*
1073 * Get rid of fragmented packets, they cause too much trouble.
1074 */
1075 unsigned int uMacLen = pBuf->mac_len;
1076 struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1077 dev_kfree_skb(pBuf);
1078 if (!pCopy)
1079 {
1080 LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1081 return 0;
1082 }
1083 pBuf = pCopy;
1084 /* Somehow skb_copy ignores mac_len */
1085 pBuf->mac_len = uMacLen;
1086# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)
1087 /* Restore VLAN tag stripped by host hardware */
1088 if (vlan_tx_tag_present(pBuf) && skb_headroom(pBuf) >= VLAN_ETH_HLEN)
1089 {
1090 uint8_t *pMac = (uint8_t*)skb_mac_header(pBuf);
1091 struct vlan_ethhdr *pVHdr = (struct vlan_ethhdr *)(pMac - VLAN_HLEN);
1092 memmove(pVHdr, pMac, ETH_ALEN * 2);
1093 pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
1094 pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
1095 pBuf->mac_header -= VLAN_HLEN;
1096 pBuf->mac_len += VLAN_HLEN;
1097 }
1098# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
1099
1100# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
1101 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",
1102 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));
1103# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1104 Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1105# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22) */
1106# else /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 18) */
1107 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",
1108 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));
1109# endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 18) */
1110 }
1111#endif /* !VBOXNETFLT_SG_SUPPORT */
1112
1113#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1114 /* Forward it to the internal network. */
1115 vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1116#else /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1117 /* Add the packet to transmit queue and schedule the bottom half. */
1118 skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1119 schedule_work(&pThis->u.s.XmitTask);
1120 Log6(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1121 &pThis->u.s.XmitTask, pBuf));
1122#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1123
1124 /* It does not really matter what we return, it is ignored by the kernel. */
1125 return 0;
1126}
1127
1128/**
1129 * Calculate the number of INTNETSEG segments the socket buffer will need.
1130 *
1131 * @returns Segment count.
1132 * @param pBuf The socket buffer.
1133 * @param pcbTemp Where to store the number of bytes of the part
1134 * of the socket buffer that will be copied to
1135 * a temporary storage.
1136 */
1137DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff *pBuf, unsigned *pcbTemp)
1138{
1139 *pcbTemp = 0;
1140#ifdef VBOXNETFLT_SG_SUPPORT
1141 unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1142 if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1143 {
1144 *pcbTemp = vboxNetFltLinuxGetChecksumStartOffset(pBuf) + VBOX_SKB_CSUM_OFFSET(pBuf) + sizeof(uint16_t);
1145 }
1146# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27)
1147 if (vlan_tx_tag_present(pBuf))
1148 {
1149 if (*pcbTemp)
1150 *pcbTemp += VLAN_HLEN;
1151 else
1152 *pcbTemp = VLAN_ETH_HLEN;
1153 }
1154# endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 27) */
1155 if (*pcbTemp)
1156 ++cSegs;
1157 struct sk_buff *pFrag;
1158 for (pFrag = skb_shinfo(pBuf)->frag_list; pFrag; pFrag = pFrag->next)
1159 {
1160 Log6(("vboxNetFltLinuxCalcSGSegments: frag=%p len=%d data_len=%d frags=%d frag_list=%p next=%p\n",
1161 pFrag, pFrag->len, pFrag->data_len, skb_shinfo(pFrag)->nr_frags, skb_shinfo(pFrag)->frag_list, pFrag->next));
1162 cSegs += 1 + skb_shinfo(pFrag)->nr_frags;
1163 }
1164#else
1165 unsigned cSegs = 1;
1166#endif
1167#ifdef PADD_RUNT_FRAMES_FROM_HOST
1168 /* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1169 if (pBuf->len < 60)
1170 cSegs++;
1171#endif
1172 return cSegs;
1173}
1174
1175
1176/**
1177 * Destroy the intnet scatter / gather buffer created by
1178 * vboxNetFltLinuxSkBufToSG.
1179 *
1180 * @param pSG The (scatter/)gather list.
1181 * @param pBuf The original socket buffer that was used to create
1182 * the scatter/gather list.
1183 */
1184static void vboxNetFltLinuxDestroySG(PINTNETSG pSG, struct sk_buff *pBuf)
1185{
1186#ifdef VBOXNETFLT_SG_SUPPORT
1187 int i, iSeg = 1; /* Skip non-paged part of SKB */
1188 /* Check if the extra buffer behind SG structure was used for modified packet header */
1189 if (pBuf->data != pSG->aSegs[0].pv)
1190 ++iSeg; /* Skip it as well */
1191# ifdef LOG_ENABLED
1192 if (pBuf->data_len)
1193 Log6(("kunmap_atomic:"));
1194# endif /* LOG_ENABLED */
1195 /* iSeg now points to the first mapped fragment if there are any */
1196 for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1197 {
1198 Log6((" %p", pSG->aSegs[iSeg].pv));
1199 VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1200 }
1201 struct sk_buff *pFragBuf;
1202 for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
1203 {
1204 ++iSeg; /* Non-fragment (unmapped) portion of chained SKB */
1205 for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
1206 {
1207 Log6((" %p", pSG->aSegs[iSeg].pv));
1208 VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1209 }
1210 }
1211# ifdef LOG_ENABLED
1212 if (pBuf->data_len)
1213 Log6(("\n"));
1214# endif /* LOG_ENABLED */
1215#endif
1216 NOREF(pSG);
1217}
1218
1219#ifdef LOG_ENABLED
1220/**
1221 * Logging helper.
1222 */
1223static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1224{
1225 int i, offSeg;
1226 uint8_t *pInt, *pExt;
1227 static int iPacketNo = 1;
1228 iPacketNo += iIncrement;
1229 if (fEgress)
1230 {
1231 pExt = pSG->aSegs[0].pv;
1232 pInt = pExt + 6;
1233 }
1234 else
1235 {
1236 pInt = pSG->aSegs[0].pv;
1237 pExt = pInt + 6;
1238 }
1239 Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1240 " %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1241 pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1242 fEgress ? "-->" : "<--", pszWhere,
1243 pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1244 pSG->cbTotal, iPacketNo));
1245 if (pSG->cSegsUsed == 1)
1246 {
1247 Log4(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1248 }
1249 else
1250 {
1251 for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1252 {
1253 Log4(("-- segment %d at 0x%x (%d bytes)\n --\n%.*Rhxd\n",
1254 i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1255 offSeg += pSG->aSegs[i].cb;
1256 }
1257 }
1258}
1259#else
1260# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1261#endif
1262
1263#ifdef VBOXNETFLT_WITH_GSO_RECV
1264
1265/**
1266 * Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1267 * GSO socket buffer without having to segment it.
1268 *
1269 * @returns true on success, false if needs segmenting.
1270 * @param pThis The net filter instance.
1271 * @param pSkb The GSO socket buffer.
1272 * @param fSrc The source.
1273 * @param pGsoCtx Where to return the GSO context on success.
1274 */
1275static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1276 PPDMNETWORKGSO pGsoCtx)
1277{
1278 PDMNETWORKGSOTYPE enmGsoType;
1279 uint16_t uEtherType;
1280 unsigned int cbTransport;
1281 unsigned int offTransport;
1282 unsigned int cbTransportHdr;
1283 unsigned uProtocol;
1284 union
1285 {
1286 RTNETIPV4 IPv4;
1287 RTNETIPV6 IPv6;
1288 RTNETTCP Tcp;
1289 uint8_t ab[40];
1290 uint16_t au16[40/2];
1291 uint32_t au32[40/4];
1292 } Buf;
1293
1294 /*
1295 * Check the GSO properties of the socket buffer and make sure it fits.
1296 */
1297 /** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1298 if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_DODGY | SKB_GSO_TCPV6 | SKB_GSO_TCPV4) ))
1299 {
1300 Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1301 return false;
1302 }
1303 if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1304 || pSkb->len > VBOX_MAX_GSO_SIZE ))
1305 {
1306 Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1307 return false;
1308 }
1309
1310 /*
1311 * Switch on the ethertype.
1312 */
1313 uEtherType = pSkb->protocol;
1314 if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1315 && pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1316 {
1317 uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1318 if (puEtherType)
1319 uEtherType = *puEtherType;
1320 }
1321 switch (uEtherType)
1322 {
1323 case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1324 {
1325 unsigned int cbHdr;
1326 PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1327 if (RT_UNLIKELY(!pIPv4))
1328 {
1329 Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1330 return false;
1331 }
1332
1333 cbHdr = pIPv4->ip_hl * 4;
1334 cbTransport = RT_N2H_U16(pIPv4->ip_len);
1335 if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1336 || cbHdr > cbTransport ))
1337 {
1338 Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1339 return false;
1340 }
1341 cbTransport -= cbHdr;
1342 offTransport = pSkb->mac_len + cbHdr;
1343 uProtocol = pIPv4->ip_p;
1344 if (uProtocol == RTNETIPV4_PROT_TCP)
1345 enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1346 else if (uProtocol == RTNETIPV4_PROT_UDP)
1347 enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1348 else /** @todo IPv6: 4to6 tunneling */
1349 enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1350 break;
1351 }
1352
1353 case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1354 {
1355 PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1356 if (RT_UNLIKELY(!pIPv6))
1357 {
1358 Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1359 return false;
1360 }
1361
1362 cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1363 offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1364 uProtocol = pIPv6->ip6_nxt;
1365 /** @todo IPv6: Dig our way out of the other headers. */
1366 if (uProtocol == RTNETIPV4_PROT_TCP)
1367 enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1368 else if (uProtocol == RTNETIPV4_PROT_UDP)
1369 enmGsoType = PDMNETWORKGSOTYPE_IPV6_UDP;
1370 else
1371 enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1372 break;
1373 }
1374
1375 default:
1376 Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1377 return false;
1378 }
1379
1380 if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1381 {
1382 Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1383 return false;
1384 }
1385
1386 if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1387 || offTransport + cbTransport > pSkb->len
1388 || cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1389 {
1390 Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1391 offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1392 return false;
1393 }
1394
1395 /*
1396 * Check the TCP/UDP bits.
1397 */
1398 if (uProtocol == RTNETIPV4_PROT_TCP)
1399 {
1400 PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1401 if (RT_UNLIKELY(!pTcp))
1402 {
1403 Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1404 return false;
1405 }
1406
1407 cbTransportHdr = pTcp->th_off * 4;
1408 pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1409 if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1410 || cbTransportHdr > cbTransport
1411 || offTransport + cbTransportHdr >= UINT8_MAX
1412 || offTransport + cbTransportHdr >= pSkb->len ))
1413 {
1414 Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1415 return false;
1416 }
1417
1418 }
1419 else
1420 {
1421 Assert(uProtocol == RTNETIPV4_PROT_UDP);
1422 cbTransportHdr = sizeof(RTNETUDP);
1423 pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1424 if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1425 || offTransport + cbTransportHdr >= pSkb->len ))
1426 {
1427 Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1428 return false;
1429 }
1430 }
1431
1432 /*
1433 * We're good, init the GSO context.
1434 */
1435 pGsoCtx->u8Type = enmGsoType;
1436 pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1437 pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1438 pGsoCtx->offHdr1 = pSkb->mac_len;
1439 pGsoCtx->offHdr2 = offTransport;
1440 pGsoCtx->u8Unused = 0;
1441
1442 return true;
1443}
1444
1445/**
1446 * Forward the socket buffer as a GSO internal network frame.
1447 *
1448 * @returns IPRT status code.
1449 * @param pThis The net filter instance.
1450 * @param pSkb The GSO socket buffer.
1451 * @param fSrc The source.
1452 * @param pGsoCtx Where to return the GSO context on success.
1453 */
1454static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1455{
1456 int rc;
1457 unsigned cbExtra;
1458 unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb, &cbExtra);
1459 PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1460 if (RT_LIKELY(pSG))
1461 {
1462 vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cbExtra, cSegs, fSrc, pGsoCtx);
1463
1464 vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1465 pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1466
1467 vboxNetFltLinuxDestroySG(pSG, pSkb);
1468 rc = VINF_SUCCESS;
1469 }
1470 else
1471 {
1472 Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1473 rc = VERR_NO_MEMORY;
1474 }
1475 return rc;
1476}
1477
1478#endif /* VBOXNETFLT_WITH_GSO_RECV */
1479
1480/**
1481 * Worker for vboxNetFltLinuxForwardToIntNet.
1482 *
1483 * @returns VINF_SUCCESS or VERR_NO_MEMORY.
1484 * @param pThis The net filter instance.
1485 * @param pBuf The socket buffer.
1486 * @param fSrc The source.
1487 */
1488static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1489{
1490 int rc;
1491 unsigned cbExtra;
1492 unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf, &cbExtra);
1493 PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1494 if (RT_LIKELY(pSG))
1495 {
1496 vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cbExtra, cSegs, fSrc, NULL /*pGsoCtx*/);
1497
1498 vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1499 pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1500
1501 vboxNetFltLinuxDestroySG(pSG, pBuf);
1502 rc = VINF_SUCCESS;
1503 }
1504 else
1505 {
1506 Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1507 rc = VERR_NO_MEMORY;
1508 }
1509 return rc;
1510}
1511
1512
1513/**
1514 * I won't disclose what I do, figure it out yourself, including pThis referencing.
1515 *
1516 * @param pThis The net filter instance.
1517 * @param pBuf The socket buffer.
1518 * @param fSrc Where the packet comes from.
1519 */
1520static void vboxNetFltLinuxForwardToIntNetInner(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1521{
1522#ifdef VBOXNETFLT_WITH_GSO
1523 if (skb_is_gso(pBuf))
1524 {
1525 PDMNETWORKGSO GsoCtx;
1526 Log6(("vboxNetFltLinuxForwardToIntNetInner: skb len=%u data_len=%u truesize=%u next=%p"
1527 " nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x ip_summed=%d\n",
1528 pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next,
1529 skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size,
1530 skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type,
1531 skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, pBuf->ip_summed));
1532#ifndef VBOXNETFLT_SG_SUPPORT
1533 if (RT_LIKELY(fSrc & INTNETTRUNKDIR_HOST))
1534 {
1535 /*
1536 * skb_gso_segment does the following. Do we need to do it as well?
1537 */
1538# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 22)
1539 skb_reset_mac_header(pBuf);
1540 pBuf->mac_len = pBuf->network_header - pBuf->mac_header;
1541# else
1542 pBuf->mac.raw = pBuf->data;
1543 pBuf->mac_len = pBuf->nh.raw - pBuf->data;
1544# endif
1545 }
1546#endif /* !VBOXNETFLT_SG_SUPPORT */
1547# ifdef VBOXNETFLT_WITH_GSO_RECV
1548 if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_TCPV6 | SKB_GSO_TCPV4))
1549 && vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1550 vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1551 else
1552# endif /* VBOXNETFLT_WITH_GSO_RECV */
1553 {
1554 /* Need to segment the packet */
1555 struct sk_buff *pNext;
1556 struct sk_buff *pSegment = skb_gso_segment(pBuf, 0 /*supported features*/);
1557 if (IS_ERR(pSegment))
1558 {
1559 LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1560 return;
1561 }
1562
1563 for (; pSegment; pSegment = pNext)
1564 {
1565 Log6(("vboxNetFltLinuxForwardToIntNetInner: segment len=%u data_len=%u truesize=%u next=%p"
1566 " nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1567 pSegment->len, pSegment->data_len, pSegment->truesize, pSegment->next,
1568 skb_shinfo(pSegment)->nr_frags, skb_shinfo(pSegment)->gso_size,
1569 skb_shinfo(pSegment)->gso_segs, skb_shinfo(pSegment)->gso_type,
1570 skb_shinfo(pSegment)->frag_list, pSegment->pkt_type));
1571 pNext = pSegment->next;
1572 pSegment->next = 0;
1573 vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1574 dev_kfree_skb(pSegment);
1575 }
1576 }
1577 }
1578 else
1579#endif /* VBOXNETFLT_WITH_GSO */
1580 {
1581 Log6(("vboxNetFltLinuxForwardToIntNetInner: ptk_type=%d ip_summed=%d len=%d"
1582 " data_len=%d headroom=%d hdr_len=%d csum_offset=%d\n",
1583 pBuf->pkt_type, pBuf->ip_summed, pBuf->len, pBuf->data_len, skb_headroom(pBuf),
1584 skb_headlen(pBuf), vboxNetFltLinuxGetChecksumStartOffset(pBuf)));
1585#ifndef VBOXNETFLT_SG_SUPPORT
1586 if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1587 {
1588#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1589 /*
1590 * Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1591 * and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1592 * header length from the header itself and reconstruct 'h' pointer
1593 * to TCP (or whatever) header.
1594 */
1595 unsigned char *tmp = pBuf->h.raw;
1596 if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1597 pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1598#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1599 int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1600#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18)
1601 /* Restore the original (wrong) pointer. */
1602 pBuf->h.raw = tmp;
1603#endif /* LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 18) */
1604 if (rc) {
1605 LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1606 return;
1607 }
1608 }
1609#endif /* !VBOXNETFLT_SG_SUPPORT */
1610 vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1611 }
1612}
1613
1614
1615/**
1616 * Temporarily adjust pBuf->data so it always points to the Ethernet header,
1617 * then forward it to the internal network.
1618 *
1619 * @param pThis The net filter instance.
1620 * @param pBuf The socket buffer. This is consumed by this function.
1621 */
1622static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1623{
1624 uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1625
1626 if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1627 {
1628 /*
1629 * The packet came from the wire and the driver has already consumed
1630 * mac header. We need to restore it back. Moreover, after we are
1631 * through with this skb we need to restore its original state!
1632 */
1633 skb_push(pBuf, pBuf->mac_len);
1634 Log5(("vboxNetFltLinuxForwardToIntNet: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1635 pBuf->mac_len, pBuf->data, skb_mac_header(pBuf), skb_network_header(pBuf)));
1636 }
1637
1638 vboxNetFltLinuxForwardToIntNetInner(pThis, pBuf, fSrc);
1639
1640 /*
1641 * Restore the original state of skb as there are other handlers this skb
1642 * will be provided to.
1643 */
1644 if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1645 skb_pull(pBuf, pBuf->mac_len);
1646
1647 dev_kfree_skb(pBuf);
1648}
1649
1650
1651#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1652/**
1653 * Work queue handler that forwards the socket buffers queued by
1654 * vboxNetFltLinuxPacketHandler to the internal network.
1655 *
1656 * @param pWork The work queue.
1657 */
1658# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
1659static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1660# else
1661static void vboxNetFltLinuxXmitTask(void *pWork)
1662# endif
1663{
1664 PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1665 struct sk_buff *pBuf;
1666
1667 Log6(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1668
1669 /*
1670 * Active? Retain the instance and increment the busy counter.
1671 */
1672 if (vboxNetFltTryRetainBusyActive(pThis))
1673 {
1674 while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1675 vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1676
1677 vboxNetFltRelease(pThis, true /* fBusy */);
1678 }
1679 else
1680 {
1681 /** @todo Shouldn't we just drop the packets here? There is little point in
1682 * making them accumulate when the VM is paused and it'll only waste
1683 * kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1684 * before start draining the packets (goes for the intnet ring buf
1685 * too)? */
1686 }
1687}
1688#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1689
1690/**
1691 * Reports the GSO capabilities of the hardware NIC.
1692 *
1693 * @param pThis The net filter instance. The caller hold a
1694 * reference to this.
1695 */
1696static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1697{
1698#ifdef VBOXNETFLT_WITH_GSO_XMIT_WIRE
1699 if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1700 {
1701 struct net_device *pDev;
1702 PINTNETTRUNKSWPORT pSwitchPort;
1703 unsigned int fFeatures;
1704
1705 RTSpinlockAcquire(pThis->hSpinlock);
1706
1707 pSwitchPort = pThis->pSwitchPort; /* this doesn't need to be here, but it doesn't harm. */
1708 pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1709 if (pDev)
1710 fFeatures = pDev->features;
1711 else
1712 fFeatures = 0;
1713
1714 RTSpinlockRelease(pThis->hSpinlock);
1715
1716 if (pThis->pSwitchPort)
1717 {
1718 /* Set/update the GSO capabilities of the NIC. */
1719 uint32_t fGsoCapabilites = 0;
1720 if (fFeatures & NETIF_F_TSO)
1721 fGsoCapabilites |= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1722 if (fFeatures & NETIF_F_TSO6)
1723 fGsoCapabilites |= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1724 Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s%s%s\n",
1725 (fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1726 (fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : ""));
1727 pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1728 }
1729
1730 vboxNetFltRelease(pThis, true /*fBusy*/);
1731 }
1732#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE */
1733}
1734
1735/**
1736 * Helper that determines whether the host (ignoreing us) is operating the
1737 * interface in promiscuous mode or not.
1738 */
1739static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1740{
1741 bool fRc = false;
1742 struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1743 if (pDev)
1744 {
1745 fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1746 LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1747 fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1748 vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1749 }
1750 return fRc;
1751}
1752
1753/**
1754 * Does this device needs link state change signaled?
1755 * Currently we need it for our own VBoxNetAdp and TAP.
1756 */
1757static bool vboxNetFltNeedsLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev)
1758{
1759 if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1760 {
1761 struct ethtool_drvinfo Info;
1762
1763 memset(&Info, 0, sizeof(Info));
1764 Info.cmd = ETHTOOL_GDRVINFO;
1765 pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1766 Log3(("%s: driver=%.*s version=%.*s bus_info=%.*s\n",
1767 __FUNCTION__,
1768 sizeof(Info.driver), Info.driver,
1769 sizeof(Info.version), Info.version,
1770 sizeof(Info.bus_info), Info.bus_info));
1771
1772 if (!strncmp(Info.driver, "vboxnet", sizeof(Info.driver)))
1773 return true;
1774
1775#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 36) /* TAP started doing carrier */
1776 return !strncmp(Info.driver, "tun", 4)
1777 && !strncmp(Info.bus_info, "tap", 4);
1778#endif
1779 }
1780
1781 return false;
1782}
1783
1784/**
1785 * Some devices need link state change when filter attaches/detaches
1786 * since the filter is their link in a sense.
1787 */
1788static void vboxNetFltSetLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1789{
1790 if (vboxNetFltNeedsLinkState(pThis, pDev))
1791 {
1792 Log3(("%s: bringing device link %s\n",
1793 __FUNCTION__, fLinkUp ? "up" : "down"));
1794 netif_tx_lock_bh(pDev);
1795 if (fLinkUp)
1796 netif_carrier_on(pDev);
1797 else
1798 netif_carrier_off(pDev);
1799 netif_tx_unlock_bh(pDev);
1800 }
1801}
1802
1803/**
1804 * Internal worker for vboxNetFltLinuxNotifierCallback.
1805 *
1806 * @returns VBox status code.
1807 * @param pThis The instance.
1808 * @param pDev The device to attach to.
1809 */
1810static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1811{
1812 LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1813
1814 /*
1815 * Retain and store the device.
1816 */
1817 dev_hold(pDev);
1818
1819 RTSpinlockAcquire(pThis->hSpinlock);
1820 ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
1821 RTSpinlockRelease(pThis->hSpinlock);
1822
1823 Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
1824 pDev, pDev->name,
1825#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1826 netdev_refcnt_read(pDev)
1827#else
1828 atomic_read(&pDev->refcnt)
1829#endif
1830 ));
1831 Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1832 pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1833
1834 /* Get the mac address while we still have a valid net_device reference. */
1835 memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
1836 /* Initialize MTU */
1837 pThis->u.s.cbMtu = pDev->mtu;
1838
1839 /*
1840 * Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1841 */
1842 pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1843 pThis->u.s.PacketType.dev = pDev;
1844 pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1845 dev_add_pack(&pThis->u.s.PacketType);
1846 ASMAtomicUoWriteBool(&pThis->u.s.fPacketHandler, true);
1847 Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1848
1849#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1850 vboxNetFltLinuxHookDev(pThis, pDev);
1851#endif
1852
1853 /*
1854 * Are we the "carrier" for this device (e.g. vboxnet or tap)?
1855 */
1856 vboxNetFltSetLinkState(pThis, pDev, true);
1857
1858 /*
1859 * Set indicators that require the spinlock. Be abit paranoid about racing
1860 * the device notification handle.
1861 */
1862 RTSpinlockAcquire(pThis->hSpinlock);
1863 pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1864 if (pDev)
1865 {
1866 ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1867 ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1868 pDev = NULL; /* don't dereference it */
1869 }
1870 RTSpinlockRelease(pThis->hSpinlock);
1871
1872 /*
1873 * Report GSO capabilities
1874 */
1875 Assert(pThis->pSwitchPort);
1876 if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1877 {
1878 vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1879 pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
1880 pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
1881 pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE | INTNETTRUNKDIR_HOST);
1882 vboxNetFltRelease(pThis, true /*fBusy*/);
1883 }
1884
1885 LogRel(("VBoxNetFlt: attached to '%s' / %RTmac\n", pThis->szName, &pThis->u.s.MacAddr));
1886 return VINF_SUCCESS;
1887}
1888
1889
1890static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1891{
1892 bool fRegistered;
1893 Assert(!pThis->fDisconnectedFromHost);
1894
1895#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1896 vboxNetFltLinuxUnhookDev(pThis, pDev);
1897#endif
1898
1899 if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
1900 {
1901 dev_remove_pack(&pThis->u.s.PacketType);
1902 Log(("vboxNetFltLinuxUnregisterDevice: this=%p: packet handler removed.\n", pThis));
1903 }
1904
1905 RTSpinlockAcquire(pThis->hSpinlock);
1906 fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
1907 if (fRegistered)
1908 {
1909 ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1910 ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1911 }
1912 RTSpinlockRelease(pThis->hSpinlock);
1913
1914 if (fRegistered)
1915 {
1916#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1917 skb_queue_purge(&pThis->u.s.XmitQueue);
1918#endif
1919 Log(("vboxNetFltLinuxUnregisterDevice: this=%p: xmit queue purged.\n", pThis));
1920 Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
1921 pDev, pDev->name,
1922#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
1923 netdev_refcnt_read(pDev)
1924#else
1925 atomic_read(&pDev->refcnt)
1926#endif
1927 ));
1928 dev_put(pDev);
1929 }
1930
1931 return NOTIFY_OK;
1932}
1933
1934static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1935{
1936 /* Check if we are not suspended and promiscuous mode has not been set. */
1937 if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
1938 && !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1939 {
1940 /* Note that there is no need for locking as the kernel got hold of the lock already. */
1941 dev_set_promiscuity(pDev, 1);
1942 ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1943 Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1944 }
1945 else
1946 Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1947 return NOTIFY_OK;
1948}
1949
1950static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1951{
1952 /* Undo promiscuous mode if we has set it. */
1953 if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1954 {
1955 /* Note that there is no need for locking as the kernel got hold of the lock already. */
1956 dev_set_promiscuity(pDev, -1);
1957 ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1958 Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1959 }
1960 else
1961 Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1962 return NOTIFY_OK;
1963}
1964
1965/**
1966 * Callback for listening to MTU change event.
1967 *
1968 * We need to track changes of host's inteface MTU to discard over-sized frames
1969 * coming from the internal network as they may hang the TX queue of host's
1970 * adapter.
1971 *
1972 * @returns NOTIFY_OK
1973 * @param pThis The netfilter instance.
1974 * @param pDev Pointer to device structure of host's interface.
1975 */
1976static int vboxNetFltLinuxDeviceMtuChange(PVBOXNETFLTINS pThis, struct net_device *pDev)
1977{
1978 ASMAtomicWriteU32(&pThis->u.s.cbMtu, pDev->mtu);
1979 Log(("vboxNetFltLinuxDeviceMtuChange: set MTU for %s to %d\n", pThis->szName, pDev->mtu));
1980 return NOTIFY_OK;
1981}
1982
1983#ifdef LOG_ENABLED
1984/** Stringify the NETDEV_XXX constants. */
1985static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
1986{
1987 const char *pszEvent = "NETDEV_<unknown>";
1988 switch (ulEventType)
1989 {
1990 case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
1991 case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
1992 case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
1993 case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
1994 case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
1995 case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
1996 case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
1997 case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
1998 case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
1999 case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2000# ifdef NETDEV_FEAT_CHANGE
2001 case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2002# endif
2003 }
2004 return pszEvent;
2005}
2006#endif /* LOG_ENABLED */
2007
2008/**
2009 * Callback for listening to netdevice events.
2010 *
2011 * This works the rediscovery, clean up on unregistration, promiscuity on
2012 * up/down, and GSO feature changes from ethtool.
2013 *
2014 * @returns NOTIFY_OK
2015 * @param self Pointer to our notifier registration block.
2016 * @param ulEventType The event.
2017 * @param ptr Event specific, but it is usually the device it
2018 * relates to.
2019 */
2020static int vboxNetFltLinuxNotifierCallback(struct notifier_block *self, unsigned long ulEventType, void *ptr)
2021
2022{
2023 PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2024 struct net_device *pMyDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2025 struct net_device *pDev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2026 int rc = NOTIFY_OK;
2027
2028 Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2029 vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, pMyDev));
2030
2031 if (ulEventType == NETDEV_REGISTER)
2032 {
2033#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 24) /* cgroups/namespaces introduced */
2034# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 26)
2035# define VBOX_DEV_NET(dev) dev_net(dev)
2036# define VBOX_NET_EQ(n1, n2) net_eq((n1), (n2))
2037# else
2038# define VBOX_DEV_NET(dev) ((dev)->nd_net)
2039# define VBOX_NET_EQ(n1, n2) ((n1) == (n2))
2040# endif
2041 struct net *pMyNet = current->nsproxy->net_ns;
2042 struct net *pDevNet = VBOX_DEV_NET(pDev);
2043
2044 if (VBOX_NET_EQ(pDevNet, pMyNet))
2045#endif /* namespaces */
2046 {
2047 if (strcmp(pDev->name, pThis->szName) == 0)
2048 {
2049 vboxNetFltLinuxAttachToInterface(pThis, pDev);
2050 }
2051 }
2052 }
2053 else
2054 {
2055 if (pDev == pMyDev)
2056 {
2057 switch (ulEventType)
2058 {
2059 case NETDEV_UNREGISTER:
2060 rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2061 break;
2062 case NETDEV_UP:
2063 rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2064 break;
2065 case NETDEV_GOING_DOWN:
2066 rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2067 break;
2068 case NETDEV_CHANGEMTU:
2069 rc = vboxNetFltLinuxDeviceMtuChange(pThis, pDev);
2070 break;
2071 case NETDEV_CHANGENAME:
2072 break;
2073#ifdef NETDEV_FEAT_CHANGE
2074 case NETDEV_FEAT_CHANGE:
2075 vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2076 break;
2077#endif
2078 }
2079 }
2080 }
2081
2082 return rc;
2083}
2084
2085/*
2086 * Initial enumeration of netdevs. Called with NETDEV_REGISTER by
2087 * register_netdevice_notifier() under rtnl lock.
2088 */
2089static int vboxNetFltLinuxEnumeratorCallback(struct notifier_block *self, unsigned long ulEventType, void *ptr)
2090{
2091 PVBOXNETFLTINS pThis = ((PVBOXNETFLTNOTIFIER)self)->pThis;
2092 struct net_device *dev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2093 struct in_device *in_dev;
2094 struct inet6_dev *in6_dev;
2095
2096 if (ulEventType != NETDEV_REGISTER)
2097 return NOTIFY_OK;
2098
2099 if (RT_UNLIKELY(pThis->pSwitchPort->pfnNotifyHostAddress == NULL))
2100 return NOTIFY_OK;
2101
2102 /*
2103 * IPv4
2104 */
2105#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 14)
2106 in_dev = __in_dev_get_rtnl(dev);
2107#else
2108 in_dev = __in_dev_get(dev);
2109#endif
2110 if (in_dev != NULL)
2111 {
2112 for_ifa(in_dev) {
2113 if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2114 return NOTIFY_OK;
2115
2116 if ( dev != pThis->u.s.pDev
2117 && VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2118 continue;
2119
2120 Log(("%s: %s: IPv4 addr %RTnaipv4 mask %RTnaipv4\n",
2121 __FUNCTION__, VBOX_NETDEV_NAME(dev),
2122 ifa->ifa_address, ifa->ifa_mask));
2123
2124 pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2125 /* :fAdded */ true, kIntNetAddrType_IPv4, &ifa->ifa_address);
2126 } endfor_ifa(in_dev);
2127 }
2128
2129 /*
2130 * IPv6
2131 */
2132 in6_dev = __in6_dev_get(dev);
2133 if (in6_dev != NULL)
2134 {
2135 struct inet6_ifaddr *ifa;
2136
2137 read_lock_bh(&in6_dev->lock);
2138#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 35)
2139 list_for_each_entry(ifa, &in6_dev->addr_list, if_list)
2140#else
2141 for (ifa = in6_dev->addr_list; ifa != NULL; ifa = ifa->if_next)
2142#endif
2143 {
2144 if ( dev != pThis->u.s.pDev
2145 && ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK))
2146 continue;
2147
2148 Log(("%s: %s: IPv6 addr %RTnaipv6/%u\n",
2149 __FUNCTION__, VBOX_NETDEV_NAME(dev),
2150 &ifa->addr, (unsigned)ifa->prefix_len));
2151
2152 pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2153 /* :fAdded */ true, kIntNetAddrType_IPv6, &ifa->addr);
2154 }
2155 read_unlock_bh(&in6_dev->lock);
2156 }
2157
2158 return NOTIFY_OK;
2159}
2160
2161
2162static int vboxNetFltLinuxNotifierIPv4Callback(struct notifier_block *self, unsigned long ulEventType, void *ptr)
2163{
2164 PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv4);
2165 struct net_device *pDev, *pEventDev;
2166 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
2167 bool fMyDev;
2168 int rc = NOTIFY_OK;
2169
2170 pDev = vboxNetFltLinuxRetainNetDev(pThis);
2171 pEventDev = ifa->ifa_dev->dev;
2172 fMyDev = (pDev == pEventDev);
2173 Log(("VBoxNetFlt: %s: IPv4 event %s(0x%lx) %s: addr %RTnaipv4 mask %RTnaipv4\n",
2174 pDev ? VBOX_NETDEV_NAME(pDev) : "<???>",
2175 vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2176 pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<???>",
2177 ifa->ifa_address, ifa->ifa_mask));
2178
2179 if (pDev != NULL)
2180 vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2181
2182 if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2183 return NOTIFY_OK;
2184
2185 if ( !fMyDev
2186 && VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2187 return NOTIFY_OK;
2188
2189 if (pThis->pSwitchPort->pfnNotifyHostAddress)
2190 {
2191 bool fAdded;
2192 if (ulEventType == NETDEV_UP)
2193 fAdded = true;
2194 else if (ulEventType == NETDEV_DOWN)
2195 fAdded = false;
2196 else
2197 return NOTIFY_OK;
2198
2199 pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2200 kIntNetAddrType_IPv4, &ifa->ifa_local);
2201 }
2202
2203 return rc;
2204}
2205
2206
2207static int vboxNetFltLinuxNotifierIPv6Callback(struct notifier_block *self, unsigned long ulEventType, void *ptr)
2208{
2209 PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv6);
2210 struct net_device *pDev, *pEventDev;
2211 struct inet6_ifaddr *ifa = (struct inet6_ifaddr *)ptr;
2212 bool fMyDev;
2213 int rc = NOTIFY_OK;
2214
2215 pDev = vboxNetFltLinuxRetainNetDev(pThis);
2216 pEventDev = ifa->idev->dev;
2217 fMyDev = (pDev == pEventDev);
2218 Log(("VBoxNetFlt: %s: IPv6 event %s(0x%lx) %s: %RTnaipv6\n",
2219 pDev ? VBOX_NETDEV_NAME(pDev) : "<???>",
2220 vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2221 pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<???>",
2222 &ifa->addr));
2223
2224 if (pDev != NULL)
2225 vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2226
2227 if ( !fMyDev
2228 && ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK))
2229 return NOTIFY_OK;
2230
2231 if (pThis->pSwitchPort->pfnNotifyHostAddress)
2232 {
2233 bool fAdded;
2234 if (ulEventType == NETDEV_UP)
2235 fAdded = true;
2236 else if (ulEventType == NETDEV_DOWN)
2237 fAdded = false;
2238 else
2239 return NOTIFY_OK;
2240
2241 pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2242 kIntNetAddrType_IPv6, &ifa->addr);
2243 }
2244
2245 return rc;
2246}
2247
2248
2249bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2250{
2251 return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2252}
2253
2254int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2255{
2256 struct net_device * pDev;
2257 int err;
2258 int rc = VINF_SUCCESS;
2259 IPRT_LINUX_SAVE_EFL_AC();
2260 NOREF(pvIfData);
2261
2262 LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2263
2264 pDev = vboxNetFltLinuxRetainNetDev(pThis);
2265 if (pDev)
2266 {
2267 /*
2268 * Create a sk_buff for the gather list and push it onto the wire.
2269 */
2270 if (fDst & INTNETTRUNKDIR_WIRE)
2271 {
2272 struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2273 if (pBuf)
2274 {
2275 vboxNetFltDumpPacket(pSG, true, "wire", 1);
2276 Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2277 Log6(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2278 err = dev_queue_xmit(pBuf);
2279 if (err)
2280 rc = RTErrConvertFromErrno(err);
2281 }
2282 else
2283 rc = VERR_NO_MEMORY;
2284 }
2285
2286 /*
2287 * Create a sk_buff for the gather list and push it onto the host stack.
2288 */
2289 if (fDst & INTNETTRUNKDIR_HOST)
2290 {
2291 struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2292 if (pBuf)
2293 {
2294 vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2295 Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2296 Log6(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2297 err = netif_rx_ni(pBuf);
2298 if (err)
2299 rc = RTErrConvertFromErrno(err);
2300 }
2301 else
2302 rc = VERR_NO_MEMORY;
2303 }
2304
2305 vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2306 }
2307
2308 IPRT_LINUX_RESTORE_EFL_AC();
2309 return rc;
2310}
2311
2312
2313void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2314{
2315 struct net_device *pDev;
2316 IPRT_LINUX_SAVE_EFL_AC();
2317
2318 LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%RTbool, fDisablePromiscuous=%RTbool\n",
2319 pThis, pThis->szName, fActive, pThis->fDisablePromiscuous));
2320
2321 if (pThis->fDisablePromiscuous)
2322 return;
2323
2324 pDev = vboxNetFltLinuxRetainNetDev(pThis);
2325 if (pDev)
2326 {
2327 /*
2328 * This api is a bit weird, the best reference is the code.
2329 *
2330 * Also, we have a bit or race conditions wrt the maintenance of
2331 * host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2332 */
2333#ifdef LOG_ENABLED
2334 u_int16_t fIf;
2335 unsigned const cPromiscBefore = pDev->promiscuity;
2336#endif
2337 if (fActive)
2338 {
2339 Assert(!pThis->u.s.fPromiscuousSet);
2340
2341 rtnl_lock();
2342 dev_set_promiscuity(pDev, 1);
2343 rtnl_unlock();
2344 pThis->u.s.fPromiscuousSet = true;
2345 Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2346 }
2347 else
2348 {
2349 if (pThis->u.s.fPromiscuousSet)
2350 {
2351 rtnl_lock();
2352 dev_set_promiscuity(pDev, -1);
2353 rtnl_unlock();
2354 Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2355 }
2356 pThis->u.s.fPromiscuousSet = false;
2357
2358#ifdef LOG_ENABLED
2359 fIf = dev_get_flags(pDev);
2360 Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2361#endif
2362 }
2363
2364 vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2365 }
2366 IPRT_LINUX_RESTORE_EFL_AC();
2367}
2368
2369
2370int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2371{
2372 /*
2373 * Remove packet handler when we get disconnected from internal switch as
2374 * we don't want the handler to forward packets to disconnected switch.
2375 */
2376 if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
2377 {
2378 IPRT_LINUX_SAVE_EFL_AC();
2379 dev_remove_pack(&pThis->u.s.PacketType);
2380 Log(("vboxNetFltOsDisconnectIt: this=%p: Packet handler removed.\n", pThis));
2381 IPRT_LINUX_RESTORE_EFL_AC();
2382 }
2383 return VINF_SUCCESS;
2384}
2385
2386
2387int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2388{
2389 IPRT_LINUX_SAVE_EFL_AC();
2390
2391 /*
2392 * Report the GSO capabilities of the host and device (if connected).
2393 * Note! No need to mark ourselves busy here.
2394 */
2395 /** @todo duplicate work here now? Attach */
2396#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2397 Log3(("vboxNetFltOsConnectIt: reporting host tso tso6\n"));
2398 pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2399 0
2400 | RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2401 | RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2402 , INTNETTRUNKDIR_HOST);
2403
2404#endif
2405 vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2406
2407 IPRT_LINUX_RESTORE_EFL_AC();
2408 return VINF_SUCCESS;
2409}
2410
2411
2412void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2413{
2414 struct net_device *pDev;
2415 bool fRegistered;
2416 IPRT_LINUX_SAVE_EFL_AC();
2417
2418#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
2419 vboxNetFltLinuxUnhookDev(pThis, NULL);
2420#endif
2421
2422 /** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2423 * unlikely, but none the less). Since it doesn't actually update the
2424 * state (just reads it), it is likely to panic in some interesting
2425 * ways. */
2426
2427 RTSpinlockAcquire(pThis->hSpinlock);
2428 pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2429 fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
2430 RTSpinlockRelease(pThis->hSpinlock);
2431
2432 if (fRegistered)
2433 {
2434 vboxNetFltSetLinkState(pThis, pDev, false);
2435
2436#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2437 skb_queue_purge(&pThis->u.s.XmitQueue);
2438#endif
2439 Log(("vboxNetFltOsDeleteInstance: this=%p: xmit queue purged.\n", pThis));
2440 Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2441 pDev, pDev->name,
2442#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37)
2443 netdev_refcnt_read(pDev)
2444#else
2445 atomic_read(&pDev->refcnt)
2446#endif
2447 ));
2448 dev_put(pDev);
2449 }
2450
2451 unregister_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2452 unregister_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2453
2454 Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2455 unregister_netdevice_notifier(&pThis->u.s.Notifier);
2456 module_put(THIS_MODULE);
2457
2458 IPRT_LINUX_RESTORE_EFL_AC();
2459}
2460
2461
2462int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2463{
2464 int err;
2465 IPRT_LINUX_SAVE_EFL_AC();
2466 NOREF(pvContext);
2467
2468 pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2469 err = register_netdevice_notifier(&pThis->u.s.Notifier);
2470 if (err)
2471 {
2472 IPRT_LINUX_RESTORE_EFL_AC();
2473 return VERR_INTNET_FLT_IF_FAILED;
2474 }
2475 if (!pThis->u.s.fRegistered)
2476 {
2477 unregister_netdevice_notifier(&pThis->u.s.Notifier);
2478 LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2479 IPRT_LINUX_RESTORE_EFL_AC();
2480 return VERR_INTNET_FLT_IF_NOT_FOUND;
2481 }
2482
2483 Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2484 if ( pThis->fDisconnectedFromHost
2485 || !try_module_get(THIS_MODULE))
2486 {
2487 IPRT_LINUX_RESTORE_EFL_AC();
2488 return VERR_INTNET_FLT_IF_FAILED;
2489 }
2490
2491 if (pThis->pSwitchPort->pfnNotifyHostAddress)
2492 {
2493 VBOXNETFLTNOTIFIER Enumerator;
2494
2495 /*
2496 * register_inetaddr_notifier() and register_inet6addr_notifier()
2497 * do not call the callback for existing devices. Enumerating
2498 * all network devices explicitly is a bit of an ifdef mess,
2499 * so co-opt register_netdevice_notifier() to do that for us.
2500 */
2501 RT_ZERO(Enumerator);
2502 Enumerator.Notifier.notifier_call = vboxNetFltLinuxEnumeratorCallback;
2503 Enumerator.pThis = pThis;
2504
2505 err = register_netdevice_notifier(&Enumerator.Notifier);
2506 if (err)
2507 {
2508 LogRel(("%s: failed to enumerate network devices: error %d\n", __FUNCTION__, err));
2509 IPRT_LINUX_RESTORE_EFL_AC();
2510 return VINF_SUCCESS;
2511 }
2512
2513 unregister_netdevice_notifier(&Enumerator.Notifier);
2514
2515 pThis->u.s.NotifierIPv4.notifier_call = vboxNetFltLinuxNotifierIPv4Callback;
2516 err = register_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2517 if (err)
2518 LogRel(("%s: failed to register IPv4 notifier: error %d\n", __FUNCTION__, err));
2519
2520 pThis->u.s.NotifierIPv6.notifier_call = vboxNetFltLinuxNotifierIPv6Callback;
2521 err = register_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2522 if (err)
2523 LogRel(("%s: failed to register IPv6 notifier: error %d\n", __FUNCTION__, err));
2524 }
2525
2526 IPRT_LINUX_RESTORE_EFL_AC();
2527 return VINF_SUCCESS;
2528}
2529
2530int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2531{
2532 IPRT_LINUX_SAVE_EFL_AC();
2533
2534 /*
2535 * Init the linux specific members.
2536 */
2537 ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2538 pThis->u.s.fRegistered = false;
2539 pThis->u.s.fPromiscuousSet = false;
2540 pThis->u.s.fPacketHandler = false;
2541 memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2542#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2543 skb_queue_head_init(&pThis->u.s.XmitQueue);
2544# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 20)
2545 INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2546# else
2547 INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2548# endif
2549#endif
2550
2551 IPRT_LINUX_RESTORE_EFL_AC();
2552 return VINF_SUCCESS;
2553}
2554
2555
2556void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2557{
2558 NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2559}
2560
2561
2562int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void *pvIf, void **pvIfData)
2563{
2564 /* Nothing to do */
2565 NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2566 return VINF_SUCCESS;
2567}
2568
2569
2570int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2571{
2572 /* Nothing to do */
2573 NOREF(pThis); NOREF(pvIfData);
2574 return VINF_SUCCESS;
2575}
2576
Note: See TracBrowser for help on using the repository browser.

© 2024 Oracle Support Privacy / Do Not Sell My Info Terms of Use Trademark Policy Automated Access Etiquette