SUPDrv-linux.c@ 78351

Last change on this file since 78351 was 77722, checked in by vboxsync, 6 years ago
Drivers/Linux: add missing init and exit markers. bugref:4567: Linux kernel driver maintenance. Building against linux-next, we get some ugly warnings in vboxpci, vboxnetflt and vboxnetadp (...) Fortunately, the fix is simple - match the forward definition:s attributes to the target attributes. I am not sure why vboxdrv does not throw a warning, but we will swat it there as well while we are at it. Michael: do VBoxGuest as well. Thank you Valdis Klētnieks for the patch.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 45.2 KB

Line
1	/* $Rev: 77722 $ */
2	/** @file
3	* VBoxDrv - The VirtualBox Support Driver - Linux specifics.
4	*/
5
6	/*
7	* Copyright (C) 2006-2019 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_SUP_DRV
32	#include "../SUPDrvInternal.h"
33	#include "the-linux-kernel.h"
34	#include "version-generated.h"
35	#include "product-generated.h"
36	#include "revision-generated.h"
37
38	#include <iprt/assert.h>
39	#include <iprt/spinlock.h>
40	#include <iprt/semaphore.h>
41	#include <iprt/initterm.h>
42	#include <iprt/process.h>
43	#include <VBox/err.h>
44	#include <iprt/mem.h>
45	#include <VBox/log.h>
46	#include <iprt/mp.h>
47
48	/** @todo figure out the exact version number */
49	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 16)
50	# include <iprt/power.h>
51	# define VBOX_WITH_SUSPEND_NOTIFICATION
52	#endif
53
54	#include <linux/sched.h>
55	#include <linux/miscdevice.h>
56	#ifdef VBOX_WITH_SUSPEND_NOTIFICATION
57	# include <linux/platform_device.h>
58	#endif
59	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 28)) && defined(SUPDRV_WITH_MSR_PROBER)
60	# define SUPDRV_LINUX_HAS_SAFE_MSR_API
61	# include <asm/msr.h>
62	#endif
63
64	#include <asm/desc.h>
65
66	#include <iprt/asm-amd64-x86.h>
67
68
69	/*********************************************************************************************************************************
70	* Defined Constants And Macros *
71	*********************************************************************************************************************************/
72	/* check kernel version */
73	# ifndef SUPDRV_AGNOSTIC
74	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 0)
75	# error Unsupported kernel version!
76	# endif
77	# endif
78
79	#ifdef CONFIG_X86_HIGH_ENTRY
80	# error "CONFIG_X86_HIGH_ENTRY is not supported by VBoxDrv at this time."
81	#endif
82
83	/* We cannot include x86.h, so we copy the defines we need here: */
84	#define X86_EFL_IF RT_BIT(9)
85	#define X86_EFL_AC RT_BIT(18)
86	#define X86_EFL_DF RT_BIT(10)
87	#define X86_EFL_IOPL (RT_BIT(12) \| RT_BIT(13))
88
89	/* To include the version number of VirtualBox into kernel backtraces: */
90	#define VBoxDrvLinuxVersion RT_CONCAT3(RT_CONCAT(VBOX_VERSION_MAJOR, _), \
91	RT_CONCAT(VBOX_VERSION_MINOR, _), \
92	VBOX_VERSION_BUILD)
93	#define VBoxDrvLinuxIOCtl RT_CONCAT(VBoxDrvLinuxIOCtl_,VBoxDrvLinuxVersion)
94
95
96
97	/*********************************************************************************************************************************
98	* Internal Functions *
99	*********************************************************************************************************************************/
100	static int __init VBoxDrvLinuxInit(void);
101	static void __exit VBoxDrvLinuxUnload(void);
102	static int VBoxDrvLinuxCreateSys(struct inode pInode, struct file pFilp);
103	static int VBoxDrvLinuxCreateUsr(struct inode pInode, struct file pFilp);
104	static int VBoxDrvLinuxClose(struct inode pInode, struct file pFilp);
105	#ifdef HAVE_UNLOCKED_IOCTL
106	static long VBoxDrvLinuxIOCtl(struct file *pFilp, unsigned int uCmd, unsigned long ulArg);
107	#else
108	static int VBoxDrvLinuxIOCtl(struct inode pInode, struct file pFilp, unsigned int uCmd, unsigned long ulArg);
109	#endif
110	static int VBoxDrvLinuxIOCtlSlow(struct file *pFilp, unsigned int uCmd, unsigned long ulArg, PSUPDRVSESSION pSession);
111	static int VBoxDrvLinuxErr2LinuxErr(int);
112	#ifdef VBOX_WITH_SUSPEND_NOTIFICATION
113	static int VBoxDrvProbe(struct platform_device *pDev);
114	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 30)
115	static int VBoxDrvSuspend(struct device *pDev);
116	static int VBoxDrvResume(struct device *pDev);
117	# else
118	static int VBoxDrvSuspend(struct platform_device *pDev, pm_message_t State);
119	static int VBoxDrvResume(struct platform_device *pDev);
120	# endif
121	static void VBoxDevRelease(struct device *pDev);
122	#endif
123
124
125	/*********************************************************************************************************************************
126	* Global Variables *
127	*********************************************************************************************************************************/
128	/**
129	* Device extention & session data association structure.
130	*/
131	static SUPDRVDEVEXT g_DevExt;
132
133	/** Module parameter.
134	* Not prefixed because the name is used by macros and the end of this file. */
135	static int force_async_tsc = 0;
136
137	/** The system device name. */
138	#define DEVICE_NAME_SYS "vboxdrv"
139	/** The user device name. */
140	#define DEVICE_NAME_USR "vboxdrvu"
141
142	#if (defined(RT_ARCH_AMD64) && LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 23)) \|\| defined(VBOX_WITH_TEXT_MODMEM_HACK)
143	/**
144	* Memory for the executable memory heap (in IPRT).
145	*/
146	# ifdef DEBUG
147	# define EXEC_MEMORY_SIZE 8388608 /* 8 MB */
148	# else
149	# define EXEC_MEMORY_SIZE 2097152 /* 2 MB */
150	# endif
151	extern uint8_t g_abExecMemory[EXEC_MEMORY_SIZE];
152	# ifndef VBOX_WITH_TEXT_MODMEM_HACK
153	__asm__(".section execmemory, \"awx\", @progbits\n\t"
154	".align 32\n\t"
155	".globl g_abExecMemory\n"
156	"g_abExecMemory:\n\t"
157	".zero " RT_XSTR(EXEC_MEMORY_SIZE) "\n\t"
158	".type g_abExecMemory, @object\n\t"
159	".size g_abExecMemory, " RT_XSTR(EXEC_MEMORY_SIZE) "\n\t"
160	".text\n\t");
161	# else
162	__asm__(".text\n\t"
163	".align 4096\n\t"
164	".globl g_abExecMemory\n"
165	"g_abExecMemory:\n\t"
166	".zero " RT_XSTR(EXEC_MEMORY_SIZE) "\n\t"
167	".type g_abExecMemory, @object\n\t"
168	".size g_abExecMemory, " RT_XSTR(EXEC_MEMORY_SIZE) "\n\t"
169	".text\n\t");
170	# endif
171	#endif
172
173	/** The file_operations structure. */
174	static struct file_operations gFileOpsVBoxDrvSys =
175	{
176	owner: THIS_MODULE,
177	open: VBoxDrvLinuxCreateSys,
178	release: VBoxDrvLinuxClose,
179	#ifdef HAVE_UNLOCKED_IOCTL
180	unlocked_ioctl: VBoxDrvLinuxIOCtl,
181	#else
182	ioctl: VBoxDrvLinuxIOCtl,
183	#endif
184	};
185
186	/** The file_operations structure. */
187	static struct file_operations gFileOpsVBoxDrvUsr =
188	{
189	owner: THIS_MODULE,
190	open: VBoxDrvLinuxCreateUsr,
191	release: VBoxDrvLinuxClose,
192	#ifdef HAVE_UNLOCKED_IOCTL
193	unlocked_ioctl: VBoxDrvLinuxIOCtl,
194	#else
195	ioctl: VBoxDrvLinuxIOCtl,
196	#endif
197	};
198
199	/** The miscdevice structure for vboxdrv. */
200	static struct miscdevice gMiscDeviceSys =
201	{
202	minor: MISC_DYNAMIC_MINOR,
203	name: DEVICE_NAME_SYS,
204	fops: &gFileOpsVBoxDrvSys,
205	# if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 17)
206	devfs_name: DEVICE_NAME_SYS,
207	# endif
208	};
209	/** The miscdevice structure for vboxdrvu. */
210	static struct miscdevice gMiscDeviceUsr =
211	{
212	minor: MISC_DYNAMIC_MINOR,
213	name: DEVICE_NAME_USR,
214	fops: &gFileOpsVBoxDrvUsr,
215	# if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 17)
216	devfs_name: DEVICE_NAME_USR,
217	# endif
218	};
219
220
221	#ifdef VBOX_WITH_SUSPEND_NOTIFICATION
222	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 30)
223	static struct dev_pm_ops gPlatformPMOps =
224	{
225	.suspend = VBoxDrvSuspend, /* before entering deep sleep */
226	.resume = VBoxDrvResume, /* after wakeup from deep sleep */
227	.freeze = VBoxDrvSuspend, /* before creating hibernation image */
228	.restore = VBoxDrvResume, /* after waking up from hibernation */
229	};
230	# endif
231
232	static struct platform_driver gPlatformDriver =
233	{
234	.probe = VBoxDrvProbe,
235	# if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 30)
236	.suspend = VBoxDrvSuspend,
237	.resume = VBoxDrvResume,
238	# endif
239	/** @todo .shutdown? */
240	.driver =
241	{
242	.name = "vboxdrv",
243	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 30)
244	.pm = &gPlatformPMOps,
245	# endif
246	}
247	};
248
249	static struct platform_device gPlatformDevice =
250	{
251	.name = "vboxdrv",
252	.dev =
253	{
254	.release = VBoxDevRelease
255	}
256	};
257	#endif /* VBOX_WITH_SUSPEND_NOTIFICATION */
258
259
260	DECLINLINE(RTUID) vboxdrvLinuxUid(void)
261	{
262	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
263	# if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 5, 0)
264	return from_kuid(current_user_ns(), current->cred->uid);
265	# else
266	return current->cred->uid;
267	# endif
268	#else
269	return current->uid;
270	#endif
271	}
272
273	DECLINLINE(RTGID) vboxdrvLinuxGid(void)
274	{
275	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
276	# if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 5, 0)
277	return from_kgid(current_user_ns(), current->cred->gid);
278	# else
279	return current->cred->gid;
280	# endif
281	#else
282	return current->gid;
283	#endif
284	}
285
286	DECLINLINE(RTUID) vboxdrvLinuxEuid(void)
287	{
288	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 29)
289	# if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 5, 0)
290	return from_kuid(current_user_ns(), current->cred->euid);
291	# else
292	return current->cred->euid;
293	# endif
294	#else
295	return current->euid;
296	#endif
297	}
298
299	/**
300	* Initialize module.
301	*
302	* @returns appropriate status code.
303	*/
304	static int __init VBoxDrvLinuxInit(void)
305	{
306	int rc;
307
308	/*
309	* Check for synchronous/asynchronous TSC mode.
310	*/
311	printk(KERN_DEBUG "vboxdrv: Found %u processor cores\n", (unsigned)RTMpGetOnlineCount());
312	rc = misc_register(&gMiscDeviceSys);
313	if (rc)
314	{
315	printk(KERN_ERR "vboxdrv: Can't register system misc device! rc=%d\n", rc);
316	return rc;
317	}
318	rc = misc_register(&gMiscDeviceUsr);
319	if (rc)
320	{
321	printk(KERN_ERR "vboxdrv: Can't register user misc device! rc=%d\n", rc);
322	misc_deregister(&gMiscDeviceSys);
323	return rc;
324	}
325	if (!rc)
326	{
327	/*
328	* Initialize the runtime.
329	* On AMD64 we'll have to donate the high rwx memory block to the exec allocator.
330	*/
331	rc = RTR0Init(0);
332	if (RT_SUCCESS(rc))
333	{
334	#if (defined(RT_ARCH_AMD64) && LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 23)) \|\| defined(VBOX_WITH_TEXT_MODMEM_HACK)
335	# ifdef VBOX_WITH_TEXT_MODMEM_HACK
336	set_memory_x(&g_abExecMemory[0], sizeof(g_abExecMemory) / PAGE_SIZE);
337	set_memory_rw(&g_abExecMemory[0], sizeof(g_abExecMemory) / PAGE_SIZE);
338	# endif
339	rc = RTR0MemExecDonate(&g_abExecMemory[0], sizeof(g_abExecMemory));
340	printk(KERN_DEBUG "VBoxDrv: dbg - g_abExecMemory=%p\n", (void *)&g_abExecMemory[0]);
341	#endif
342	Log(("VBoxDrv::ModuleInit\n"));
343
344	/*
345	* Initialize the device extension.
346	*/
347	if (RT_SUCCESS(rc))
348	rc = supdrvInitDevExt(&g_DevExt, sizeof(SUPDRVSESSION));
349	if (RT_SUCCESS(rc))
350	{
351	#ifdef VBOX_WITH_SUSPEND_NOTIFICATION
352	rc = platform_driver_register(&gPlatformDriver);
353	if (rc == 0)
354	{
355	rc = platform_device_register(&gPlatformDevice);
356	if (rc == 0)
357	#endif
358	{
359	printk(KERN_INFO "vboxdrv: TSC mode is %s, tentative frequency %llu Hz\n",
360	SUPGetGIPModeName(g_DevExt.pGip), g_DevExt.pGip->u64CpuHz);
361	LogFlow(("VBoxDrv::ModuleInit returning %#x\n", rc));
362	printk(KERN_DEBUG "vboxdrv: Successfully loaded version "
363	VBOX_VERSION_STRING " (interface " RT_XSTR(SUPDRV_IOC_VERSION) ")\n");
364	return rc;
365	}
366	#ifdef VBOX_WITH_SUSPEND_NOTIFICATION
367	else
368	platform_driver_unregister(&gPlatformDriver);
369	}
370	#endif
371	}
372
373	rc = -EINVAL;
374	RTR0TermForced();
375	}
376	else
377	rc = -EINVAL;
378
379	/*
380	* Failed, cleanup and return the error code.
381	*/
382	}
383	misc_deregister(&gMiscDeviceSys);
384	misc_deregister(&gMiscDeviceUsr);
385	Log(("VBoxDrv::ModuleInit returning %#x (minor:%d & %d)\n", rc, gMiscDeviceSys.minor, gMiscDeviceUsr.minor));
386	return rc;
387	}
388
389
390	/**
391	* Unload the module.
392	*/
393	static void __exit VBoxDrvLinuxUnload(void)
394	{
395	Log(("VBoxDrvLinuxUnload\n"));
396
397	#ifdef VBOX_WITH_SUSPEND_NOTIFICATION
398	platform_device_unregister(&gPlatformDevice);
399	platform_driver_unregister(&gPlatformDriver);
400	#endif
401
402	/*
403	* I Don't think it's possible to unload a driver which processes have
404	* opened, at least we'll blindly assume that here.
405	*/
406	misc_deregister(&gMiscDeviceUsr);
407	misc_deregister(&gMiscDeviceSys);
408
409	/*
410	* Destroy GIP, delete the device extension and terminate IPRT.
411	*/
412	supdrvDeleteDevExt(&g_DevExt);
413	RTR0TermForced();
414	}
415
416
417	/**
418	* Common open code.
419	*
420	* @param pInode Pointer to inode info structure.
421	* @param pFilp Associated file pointer.
422	* @param fUnrestricted Indicates which device node which was opened.
423	*/
424	static int vboxdrvLinuxCreateCommon(struct inode pInode, struct file pFilp, bool fUnrestricted)
425	{
426	int rc;
427	PSUPDRVSESSION pSession;
428	Log(("VBoxDrvLinuxCreate: pFilp=%p pid=%d/%d %s\n", pFilp, RTProcSelf(), current->pid, current->comm));
429
430	#ifdef VBOX_WITH_HARDENING
431	/*
432	* Only root is allowed to access the unrestricted device, enforce it!
433	*/
434	if ( fUnrestricted
435	&& vboxdrvLinuxEuid() != 0 /* root */ )
436	{
437	Log(("VBoxDrvLinuxCreate: euid=%d, expected 0 (root)\n", vboxdrvLinuxEuid()));
438	return -EPERM;
439	}
440	#endif /* VBOX_WITH_HARDENING */
441
442	/*
443	* Call common code for the rest.
444	*/
445	rc = supdrvCreateSession(&g_DevExt, true /* fUser */, fUnrestricted, &pSession);
446	if (!rc)
447	{
448	pSession->Uid = vboxdrvLinuxUid();
449	pSession->Gid = vboxdrvLinuxGid();
450	}
451
452	pFilp->private_data = pSession;
453
454	Log(("VBoxDrvLinuxCreate: g_DevExt=%p pSession=%p rc=%d/%d (pid=%d/%d %s)\n",
455	&g_DevExt, pSession, rc, VBoxDrvLinuxErr2LinuxErr(rc),
456	RTProcSelf(), current->pid, current->comm));
457	return VBoxDrvLinuxErr2LinuxErr(rc);
458	}
459
460
461	/** /dev/vboxdrv. */
462	static int VBoxDrvLinuxCreateSys(struct inode pInode, struct file pFilp)
463	{
464	return vboxdrvLinuxCreateCommon(pInode, pFilp, true);
465	}
466
467
468	/** /dev/vboxdrvu. */
469	static int VBoxDrvLinuxCreateUsr(struct inode pInode, struct file pFilp)
470	{
471	return vboxdrvLinuxCreateCommon(pInode, pFilp, false);
472	}
473
474
475	/**
476	* Close device.
477	*
478	* @param pInode Pointer to inode info structure.
479	* @param pFilp Associated file pointer.
480	*/
481	static int VBoxDrvLinuxClose(struct inode pInode, struct file pFilp)
482	{
483	Log(("VBoxDrvLinuxClose: pFilp=%p pSession=%p pid=%d/%d %s\n",
484	pFilp, pFilp->private_data, RTProcSelf(), current->pid, current->comm));
485	supdrvSessionRelease((PSUPDRVSESSION)pFilp->private_data);
486	pFilp->private_data = NULL;
487	return 0;
488	}
489
490
491	#ifdef VBOX_WITH_SUSPEND_NOTIFICATION
492	/**
493	* Dummy device release function. We have to provide this function,
494	* otherwise the kernel will complain.
495	*
496	* @param pDev Pointer to the platform device.
497	*/
498	static void VBoxDevRelease(struct device *pDev)
499	{
500	}
501
502	/**
503	* Dummy probe function.
504	*
505	* @param pDev Pointer to the platform device.
506	*/
507	static int VBoxDrvProbe(struct platform_device *pDev)
508	{
509	return 0;
510	}
511
512	/**
513	* Suspend callback.
514	* @param pDev Pointer to the platform device.
515	* @param State Message type, see Documentation/power/devices.txt.
516	* Ignored.
517	*/
518	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 30) && !defined(DOXYGEN_RUNNING)
519	static int VBoxDrvSuspend(struct device *pDev)
520	# else
521	static int VBoxDrvSuspend(struct platform_device *pDev, pm_message_t State)
522	# endif
523	{
524	RTPowerSignalEvent(RTPOWEREVENT_SUSPEND);
525	return 0;
526	}
527
528	/**
529	* Resume callback.
530	*
531	* @param pDev Pointer to the platform device.
532	*/
533	# if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 30)
534	static int VBoxDrvResume(struct device *pDev)
535	# else
536	static int VBoxDrvResume(struct platform_device *pDev)
537	# endif
538	{
539	RTPowerSignalEvent(RTPOWEREVENT_RESUME);
540	return 0;
541	}
542	#endif /* VBOX_WITH_SUSPEND_NOTIFICATION */
543
544
545	/**
546	* Device I/O Control entry point.
547	*
548	* @param pFilp Associated file pointer.
549	* @param uCmd The function specified to ioctl().
550	* @param ulArg The argument specified to ioctl().
551	*/
552	#if defined(HAVE_UNLOCKED_IOCTL) \|\| defined(DOXYGEN_RUNNING)
553	static long VBoxDrvLinuxIOCtl(struct file *pFilp, unsigned int uCmd, unsigned long ulArg)
554	#else
555	static int VBoxDrvLinuxIOCtl(struct inode pInode, struct file pFilp, unsigned int uCmd, unsigned long ulArg)
556	#endif
557	{
558	PSUPDRVSESSION pSession = (PSUPDRVSESSION)pFilp->private_data;
559	int rc;
560	#if defined(VBOX_STRICT) \|\| defined(VBOX_WITH_EFLAGS_AC_SET_IN_VBOXDRV)
561	RTCCUINTREG fSavedEfl;
562
563	/*
564	* Refuse all I/O control calls if we've ever detected EFLAGS.AC being cleared.
565	*
566	* This isn't a problem, as there is absolutely nothing in the kernel context that
567	* depend on user context triggering cleanups. That would be pretty wild, right?
568	*/
569	if (RT_UNLIKELY(g_DevExt.cBadContextCalls > 0))
570	{
571	SUPR0Printf("VBoxDrvLinuxIOCtl: EFLAGS.AC=0 detected %u times, refusing all I/O controls!\n", g_DevExt.cBadContextCalls);
572	return ESPIPE;
573	}
574
575	fSavedEfl = ASMAddFlags(X86_EFL_AC);
576	# else
577	stac();
578	# endif
579
580	/*
581	* Deal with the two high-speed IOCtl that takes it's arguments from
582	* the session and iCmd, and only returns a VBox status code.
583	*/
584	AssertCompile(_IOC_NRSHIFT == 0 && _IOC_NRBITS == 8);
585	#ifdef HAVE_UNLOCKED_IOCTL
586	if (RT_LIKELY( (unsigned int)(uCmd - SUP_IOCTL_FAST_DO_FIRST) < (unsigned int)32
587	&& pSession->fUnrestricted))
588	rc = supdrvIOCtlFast(uCmd - SUP_IOCTL_FAST_DO_FIRST, ulArg, &g_DevExt, pSession);
589	else
590	rc = VBoxDrvLinuxIOCtlSlow(pFilp, uCmd, ulArg, pSession);
591	#else /* !HAVE_UNLOCKED_IOCTL */
592	unlock_kernel();
593	if (RT_LIKELY( (unsigned int)(uCmd - SUP_IOCTL_FAST_DO_FIRST) < (unsigned int)32
594	&& pSession->fUnrestricted))
595	rc = supdrvIOCtlFast(uCmd - SUP_IOCTL_FAST_DO_FIRST, ulArg, &g_DevExt, pSession);
596	else
597	rc = VBoxDrvLinuxIOCtlSlow(pFilp, uCmd, ulArg, pSession);
598	lock_kernel();
599	#endif /* !HAVE_UNLOCKED_IOCTL */
600
601	#if defined(VBOX_STRICT) \|\| defined(VBOX_WITH_EFLAGS_AC_SET_IN_VBOXDRV)
602	/*
603	* Before we restore AC and the rest of EFLAGS, check if the IOCtl handler code
604	* accidentially modified it or some other important flag.
605	*/
606	if (RT_UNLIKELY( (ASMGetFlags() & (X86_EFL_AC \| X86_EFL_IF \| X86_EFL_DF))
607	!= ((fSavedEfl & (X86_EFL_AC \| X86_EFL_IF \| X86_EFL_DF)) \| X86_EFL_AC) ))
608	{
609	char szTmp[48];
610	RTStrPrintf(szTmp, sizeof(szTmp), "uCmd=%#x: %#x->%#x!", _IOC_NR(uCmd), (uint32_t)fSavedEfl, (uint32_t)ASMGetFlags());
611	supdrvBadContext(&g_DevExt, "SUPDrv-linux.c", __LINE__, szTmp);
612	}
613	ASMSetFlags(fSavedEfl);
614	#else
615	clac();
616	#endif
617	return rc;
618	}
619
620
621	/**
622	* Device I/O Control entry point.
623	*
624	* @param pFilp Associated file pointer.
625	* @param uCmd The function specified to ioctl().
626	* @param ulArg The argument specified to ioctl().
627	* @param pSession The session instance.
628	*/
629	static int VBoxDrvLinuxIOCtlSlow(struct file *pFilp, unsigned int uCmd, unsigned long ulArg, PSUPDRVSESSION pSession)
630	{
631	int rc;
632	SUPREQHDR Hdr;
633	PSUPREQHDR pHdr;
634	uint32_t cbBuf;
635
636	Log6(("VBoxDrvLinuxIOCtl: pFilp=%p uCmd=%#x ulArg=%p pid=%d/%d\n", pFilp, uCmd, (void *)ulArg, RTProcSelf(), current->pid));
637
638	/*
639	* Read the header.
640	*/
641	if (RT_FAILURE(RTR0MemUserCopyFrom(&Hdr, ulArg, sizeof(Hdr))))
642	{
643	Log(("VBoxDrvLinuxIOCtl: copy_from_user(,%#lx,) failed; uCmd=%#x\n", ulArg, uCmd));
644	return -EFAULT;
645	}
646	if (RT_UNLIKELY((Hdr.fFlags & SUPREQHDR_FLAGS_MAGIC_MASK) != SUPREQHDR_FLAGS_MAGIC))
647	{
648	Log(("VBoxDrvLinuxIOCtl: bad header magic %#x; uCmd=%#x\n", Hdr.fFlags & SUPREQHDR_FLAGS_MAGIC_MASK, uCmd));
649	return -EINVAL;
650	}
651
652	/*
653	* Buffer the request.
654	*/
655	cbBuf = RT_MAX(Hdr.cbIn, Hdr.cbOut);
656	if (RT_UNLIKELY(cbBuf > _1M*16))
657	{
658	Log(("VBoxDrvLinuxIOCtl: too big cbBuf=%#x; uCmd=%#x\n", cbBuf, uCmd));
659	return -E2BIG;
660	}
661	if (RT_UNLIKELY(_IOC_SIZE(uCmd) ? cbBuf != _IOC_SIZE(uCmd) : Hdr.cbIn < sizeof(Hdr)))
662	{
663	Log(("VBoxDrvLinuxIOCtl: bad ioctl cbBuf=%#x _IOC_SIZE=%#x; uCmd=%#x\n", cbBuf, _IOC_SIZE(uCmd), uCmd));
664	return -EINVAL;
665	}
666	pHdr = RTMemAlloc(cbBuf);
667	if (RT_UNLIKELY(!pHdr))
668	{
669	OSDBGPRINT(("VBoxDrvLinuxIOCtl: failed to allocate buffer of %d bytes for uCmd=%#x\n", cbBuf, uCmd));
670	return -ENOMEM;
671	}
672	if (RT_FAILURE(RTR0MemUserCopyFrom(pHdr, ulArg, Hdr.cbIn)))
673	{
674	Log(("VBoxDrvLinuxIOCtl: copy_from_user(,%#lx, %#x) failed; uCmd=%#x\n", ulArg, Hdr.cbIn, uCmd));
675	RTMemFree(pHdr);
676	return -EFAULT;
677	}
678	if (Hdr.cbIn < cbBuf)
679	RT_BZERO((uint8_t *)pHdr + Hdr.cbIn, cbBuf - Hdr.cbIn);
680
681	/*
682	* Process the IOCtl.
683	*/
684	rc = supdrvIOCtl(uCmd, &g_DevExt, pSession, pHdr, cbBuf);
685
686	/*
687	* Copy ioctl data and output buffer back to user space.
688	*/
689	if (RT_LIKELY(!rc))
690	{
691	uint32_t cbOut = pHdr->cbOut;
692	if (RT_UNLIKELY(cbOut > cbBuf))
693	{
694	OSDBGPRINT(("VBoxDrvLinuxIOCtl: too much output! %#x > %#x; uCmd=%#x!\n", cbOut, cbBuf, uCmd));
695	cbOut = cbBuf;
696	}
697	if (RT_FAILURE(RTR0MemUserCopyTo(ulArg, pHdr, cbOut)))
698	{
699	/* this is really bad! */
700	OSDBGPRINT(("VBoxDrvLinuxIOCtl: copy_to_user(%#lx,,%#x); uCmd=%#x!\n", ulArg, cbOut, uCmd));
701	rc = -EFAULT;
702	}
703	}
704	else
705	{
706	Log(("VBoxDrvLinuxIOCtl: pFilp=%p uCmd=%#x ulArg=%p failed, rc=%d\n", pFilp, uCmd, (void *)ulArg, rc));
707	rc = -EINVAL;
708	}
709	RTMemFree(pHdr);
710
711	Log6(("VBoxDrvLinuxIOCtl: returns %d (pid=%d/%d)\n", rc, RTProcSelf(), current->pid));
712	return rc;
713	}
714
715
716	/**
717	* The SUPDRV IDC entry point.
718	*
719	* @returns VBox status code, see supdrvIDC.
720	* @param uReq The request code.
721	* @param pReq The request.
722	*/
723	int VBOXCALL SUPDrvLinuxIDC(uint32_t uReq, PSUPDRVIDCREQHDR pReq)
724	{
725	PSUPDRVSESSION pSession;
726
727	/*
728	* Some quick validations.
729	*/
730	if (RT_UNLIKELY(!VALID_PTR(pReq)))
731	return VERR_INVALID_POINTER;
732
733	pSession = pReq->pSession;
734	if (pSession)
735	{
736	if (RT_UNLIKELY(!VALID_PTR(pSession)))
737	return VERR_INVALID_PARAMETER;
738	if (RT_UNLIKELY(pSession->pDevExt != &g_DevExt))
739	return VERR_INVALID_PARAMETER;
740	}
741	else if (RT_UNLIKELY(uReq != SUPDRV_IDC_REQ_CONNECT))
742	return VERR_INVALID_PARAMETER;
743
744	/*
745	* Do the job.
746	*/
747	return supdrvIDC(uReq, &g_DevExt, pSession, pReq);
748	}
749
750	EXPORT_SYMBOL(SUPDrvLinuxIDC);
751
752
753	RTCCUINTREG VBOXCALL supdrvOSChangeCR4(RTCCUINTREG fOrMask, RTCCUINTREG fAndMask)
754	{
755	#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 20, 0)
756	RTCCUINTREG uOld = this_cpu_read(cpu_tlbstate.cr4);
757	RTCCUINTREG uNew = (uOld & fAndMask) \| fOrMask;
758	if (uNew != uOld)
759	{
760	this_cpu_write(cpu_tlbstate.cr4, uNew);
761	__write_cr4(uNew);
762	}
763	#else
764	RTCCUINTREG uOld = ASMGetCR4();
765	RTCCUINTREG uNew = (uOld & fAndMask) \| fOrMask;
766	if (uNew != uOld)
767	ASMSetCR4(uNew);
768	#endif
769	return uOld;
770	}
771
772
773	void VBOXCALL supdrvOSCleanupSession(PSUPDRVDEVEXT pDevExt, PSUPDRVSESSION pSession)
774	{
775	NOREF(pDevExt);
776	NOREF(pSession);
777	}
778
779
780	void VBOXCALL supdrvOSSessionHashTabInserted(PSUPDRVDEVEXT pDevExt, PSUPDRVSESSION pSession, void *pvUser)
781	{
782	NOREF(pDevExt); NOREF(pSession); NOREF(pvUser);
783	}
784
785
786	void VBOXCALL supdrvOSSessionHashTabRemoved(PSUPDRVDEVEXT pDevExt, PSUPDRVSESSION pSession, void *pvUser)
787	{
788	NOREF(pDevExt); NOREF(pSession); NOREF(pvUser);
789	}
790
791
792	/**
793	* Initializes any OS specific object creator fields.
794	*/
795	void VBOXCALL supdrvOSObjInitCreator(PSUPDRVOBJ pObj, PSUPDRVSESSION pSession)
796	{
797	NOREF(pObj);
798	NOREF(pSession);
799	}
800
801
802	/**
803	* Checks if the session can access the object.
804	*
805	* @returns true if a decision has been made.
806	* @returns false if the default access policy should be applied.
807	*
808	* @param pObj The object in question.
809	* @param pSession The session wanting to access the object.
810	* @param pszObjName The object name, can be NULL.
811	* @param prc Where to store the result when returning true.
812	*/
813	bool VBOXCALL supdrvOSObjCanAccess(PSUPDRVOBJ pObj, PSUPDRVSESSION pSession, const char pszObjName, int prc)
814	{
815	NOREF(pObj);
816	NOREF(pSession);
817	NOREF(pszObjName);
818	NOREF(prc);
819	return false;
820	}
821
822
823	bool VBOXCALL supdrvOSGetForcedAsyncTscMode(PSUPDRVDEVEXT pDevExt)
824	{
825	return force_async_tsc != 0;
826	}
827
828
829	bool VBOXCALL supdrvOSAreCpusOfflinedOnSuspend(void)
830	{
831	return true;
832	}
833
834
835	bool VBOXCALL supdrvOSAreTscDeltasInSync(void)
836	{
837	return false;
838	}
839
840
841	int VBOXCALL supdrvOSLdrOpen(PSUPDRVDEVEXT pDevExt, PSUPDRVLDRIMAGE pImage, const char *pszFilename)
842	{
843	NOREF(pDevExt); NOREF(pImage); NOREF(pszFilename);
844	return VERR_NOT_SUPPORTED;
845	}
846
847
848	int VBOXCALL supdrvOSLdrValidatePointer(PSUPDRVDEVEXT pDevExt, PSUPDRVLDRIMAGE pImage, void *pv,
849	const uint8_t pbImageBits, const char pszSymbol)
850	{
851	NOREF(pDevExt); NOREF(pImage); NOREF(pv); NOREF(pbImageBits); NOREF(pszSymbol);
852	return VERR_NOT_SUPPORTED;
853	}
854
855
856	int VBOXCALL supdrvOSLdrLoad(PSUPDRVDEVEXT pDevExt, PSUPDRVLDRIMAGE pImage, const uint8_t *pbImageBits, PSUPLDRLOAD pReq)
857	{
858	NOREF(pDevExt); NOREF(pImage); NOREF(pbImageBits); NOREF(pReq);
859	return VERR_NOT_SUPPORTED;
860	}
861
862
863	void VBOXCALL supdrvOSLdrUnload(PSUPDRVDEVEXT pDevExt, PSUPDRVLDRIMAGE pImage)
864	{
865	NOREF(pDevExt); NOREF(pImage);
866	}
867
868
869	/** @def VBOX_WITH_NON_PROD_HACK_FOR_PERF_STACKS
870	* A very crude hack for debugging using perf and dtrace.
871	*
872	* DO ABSOLUTELY NOT ENABLE IN PRODUCTION BUILDS! DEVELOPMENT ONLY!!
873	* DO ABSOLUTELY NOT ENABLE IN PRODUCTION BUILDS! DEVELOPMENT ONLY!!
874	* DO ABSOLUTELY NOT ENABLE IN PRODUCTION BUILDS! DEVELOPMENT ONLY!!
875	*
876	*/
877	#if 0 \|\| defined(DOXYGEN_RUNNING)
878	# define VBOX_WITH_NON_PROD_HACK_FOR_PERF_STACKS
879	#endif
880
881	#if defined(VBOX_WITH_NON_PROD_HACK_FOR_PERF_STACKS) && defined(CONFIG_MODULES_TREE_LOOKUP)
882	/** Whether g_pfnModTreeInsert and g_pfnModTreeRemove have been initialized.
883	* @remarks can still be NULL after init. */
884	static volatile bool g_fLookedForModTreeFunctions = false;
885	static void (g_pfnModTreeInsert)(struct mod_tree_node ) = NULL; /*< __mod_tree_insert /
886	static void (g_pfnModTreeRemove)(struct mod_tree_node ) = NULL; /*< __mod_tree_remove /
887	#endif
888
889
890	void VBOXCALL supdrvOSLdrNotifyOpened(PSUPDRVDEVEXT pDevExt, PSUPDRVLDRIMAGE pImage, const char *pszFilename)
891	{
892	#ifdef VBOX_WITH_NON_PROD_HACK_FOR_PERF_STACKS /* Not for production use!! Debugging only! */
893	/*
894	* This trick stops working with 4.2 when CONFIG_MODULES_TREE_LOOKUP is
895	* defined. The module lookups are done via a tree structure and we
896	* cannot get at the root of it. :-(
897	*/
898	# ifdef CONFIG_KALLSYMS
899	size_t const cchName = strlen(pImage->szName);
900	# endif
901	struct module pMyMod, pSelfMod, pTestMod, pTestModByName;
902	IPRT_LINUX_SAVE_EFL_AC();
903
904	pImage->pLnxModHack = NULL;
905
906	# ifdef CONFIG_MODULES_TREE_LOOKUP
907	/*
908	* This is pretty naive, but works for 4.2 on arch linux. I don't think we
909	* can count on finding __mod_tree_remove in all kernel builds as it's not
910	* marked noinline like __mod_tree_insert.
911	*/
912	if (!g_fLookedForModTreeFunctions)
913	{
914	unsigned long ulInsert = kallsyms_lookup_name("__mod_tree_insert");
915	unsigned long ulRemove = kallsyms_lookup_name("__mod_tree_remove");
916	if (!ulInsert \|\| !ulRemove)
917	{
918	g_fLookedForModTreeFunctions = true;
919	printk(KERN_ERR "vboxdrv: failed to locate __mod_tree_insert and __mod_tree_remove.\n");
920	IPRT_LINUX_RESTORE_EFL_AC();
921	return;
922	}
923	(unsigned long )&g_pfnModTreeInsert = ulInsert;
924	(unsigned long )&g_pfnModTreeRemove = ulRemove;
925	ASMCompilerBarrier();
926	g_fLookedForModTreeFunctions = true;
927	}
928	else if (!g_pfnModTreeInsert \|\| !g_pfnModTreeRemove)
929	return;
930	#endif
931
932	/*
933	* Make sure we've found our own module, otherwise we cannot access the linked list.
934	*/
935	mutex_lock(&module_mutex);
936	pSelfMod = find_module("vboxdrv");
937	mutex_unlock(&module_mutex);
938	if (!pSelfMod)
939	{
940	IPRT_LINUX_RESTORE_EFL_AC();
941	return;
942	}
943
944	/*
945	* Cook up a module structure for the image.
946	* We allocate symbol and string tables in the allocation and the module to keep things simple.
947	*/
948	# ifdef CONFIG_KALLSYMS
949	pMyMod = (struct module )RTMemAllocZ(sizeof(pMyMod)
950	+ sizeof(Elf_Sym) * 3
951	+ 1 + cchName * 2 + sizeof("_start") + sizeof("_end") + 4 );
952	# else
953	pMyMod = (struct module )RTMemAllocZ(sizeof(pMyMod));
954	# endif
955	if (pMyMod)
956	{
957	int rc = VINF_SUCCESS;
958	# ifdef CONFIG_KALLSYMS
959	Elf_Sym paSymbols = (Elf_Sym )(pMyMod + 1);
960	char pchStrTab = (char )(paSymbols + 3);
961	# endif
962
963	pMyMod->state = MODULE_STATE_LIVE;
964	INIT_LIST_HEAD(&pMyMod->list); /* just in case */
965
966	/* Perf only matches up files with a .ko extension (maybe .ko.gz),
967	so in order for this crap to work smoothly, we append .ko to the
968	module name and require the user to create symbolic links in
969	/lib/modules/`uname -r`:
970	for i in VMMR0.r0 VBoxDDR0.r0 VBoxDD2R0.r0; do
971	sudo ln -s /mnt/scratch/vbox/svn/trunk/out/linux.amd64/debug/bin/$i /lib/modules/`uname -r`/$i.ko;
972	done */
973	RTStrPrintf(pMyMod->name, sizeof(pMyMod->name), "%s", pImage->szName);
974
975	/* sysfs bits. */
976	INIT_LIST_HEAD(&pMyMod->mkobj.kobj.entry); /* rest of kobj is already zeroed, hopefully never accessed... */
977	pMyMod->mkobj.mod = pMyMod;
978	pMyMod->mkobj.drivers_dir = NULL;
979	pMyMod->mkobj.mp = NULL;
980	pMyMod->mkobj.kobj_completion = NULL;
981
982	pMyMod->modinfo_attrs = NULL; /* hopefully not accessed after setup. */
983	pMyMod->holders_dir = NULL; /* hopefully not accessed. */
984	pMyMod->version = "N/A";
985	pMyMod->srcversion = "N/A";
986
987	/* We export no symbols. */
988	pMyMod->num_syms = 0;
989	pMyMod->syms = NULL;
990	pMyMod->crcs = NULL;
991
992	pMyMod->num_gpl_syms = 0;
993	pMyMod->gpl_syms = NULL;
994	pMyMod->gpl_crcs = NULL;
995
996	pMyMod->num_gpl_future_syms = 0;
997	pMyMod->gpl_future_syms = NULL;
998	pMyMod->gpl_future_crcs = NULL;
999
1000	# if CONFIG_UNUSED_SYMBOLS
1001	pMyMod->num_unused_syms = 0;
1002	pMyMod->unused_syms = NULL;
1003	pMyMod->unused_crcs = NULL;
1004
1005	pMyMod->num_unused_gpl_syms = 0;
1006	pMyMod->unused_gpl_syms = NULL;
1007	pMyMod->unused_gpl_crcs = NULL;
1008	# endif
1009	/* No kernel parameters either. */
1010	pMyMod->kp = NULL;
1011	pMyMod->num_kp = 0;
1012
1013	# ifdef CONFIG_MODULE_SIG
1014	/* Pretend ok signature. */
1015	pMyMod->sig_ok = true;
1016	# endif
1017	/* No exception table. */
1018	pMyMod->num_exentries = 0;
1019	pMyMod->extable = NULL;
1020
1021	/* No init function */
1022	pMyMod->init = NULL;
1023	pMyMod->module_init = NULL;
1024	pMyMod->init_size = 0;
1025	pMyMod->init_ro_size = 0;
1026	pMyMod->init_text_size = 0;
1027
1028	/* The module address and size. It's all text. */
1029	pMyMod->module_core = pImage->pvImage;
1030	pMyMod->core_size = pImage->cbImageBits;
1031	pMyMod->core_text_size = pImage->cbImageBits;
1032	pMyMod->core_ro_size = pImage->cbImageBits;
1033
1034	#ifdef CONFIG_MODULES_TREE_LOOKUP
1035	/* Fill in the self pointers for the tree nodes. */
1036	pMyMod->mtn_core.mod = pMyMod;
1037	pMyMod->mtn_init.mod = pMyMod;
1038	#endif
1039	/* They invented the tained bit for us, didn't they? */
1040	pMyMod->taints = 1;
1041
1042	# ifdef CONFIG_GENERIC_BUGS
1043	/* No BUGs in our modules. */
1044	pMyMod->num_bugs = 0;
1045	INIT_LIST_HEAD(&pMyMod->bug_list);
1046	pMyMod->bug_table = NULL;
1047	# endif
1048
1049	# ifdef CONFIG_KALLSYMS
1050	/* The core stuff is documented as only used when loading. So just zero them. */
1051	pMyMod->core_num_syms = 0;
1052	pMyMod->core_symtab = NULL;
1053	pMyMod->core_strtab = NULL;
1054
1055	/* Construct a symbol table with start and end symbols.
1056	Note! We don't have our own symbol table at this point, image bit
1057	are not uploaded yet! */
1058	pMyMod->num_symtab = 3;
1059	pMyMod->symtab = paSymbols;
1060	pMyMod->strtab = pchStrTab;
1061	RT_ZERO(paSymbols[0]);
1062	pchStrTab[0] = '\0';
1063	paSymbols[1].st_name = 1;
1064	paSymbols[2].st_name = 2 + RTStrPrintf(&pchStrTab[paSymbols[1].st_name], cchName + sizeof("_start"),
1065	"%s_start", pImage->szName);
1066	RTStrPrintf(&pchStrTab[paSymbols[2].st_name], cchName + sizeof("_end"), "%s_end", pImage->szName);
1067	paSymbols[1].st_info = 't';
1068	paSymbols[2].st_info = 'b';
1069	paSymbols[1].st_other = 0;
1070	paSymbols[2].st_other = 0;
1071	paSymbols[1].st_shndx = 0;
1072	paSymbols[2].st_shndx = 0;
1073	paSymbols[1].st_value = (uintptr_t)pImage->pvImage;
1074	paSymbols[2].st_value = (uintptr_t)pImage->pvImage + pImage->cbImageBits - 1;
1075	paSymbols[1].st_size = pImage->cbImageBits - 1;
1076	paSymbols[2].st_size = 1;
1077	# endif
1078	/* No arguments, but seems its always non-NULL so put empty string there. */
1079	pMyMod->args = "";
1080
1081	# ifdef CONFIG_SMP
1082	/* No per CPU data. */
1083	pMyMod->percpu = NULL;
1084	pMyMod->percpu_size = 0;
1085	# endif
1086	# ifdef CONFIG_TRACEPOINTS
1087	/* No tracepoints we like to share. */
1088	pMyMod->num_tracepoints = 0;
1089	pMyMod->tracepoints_ptrs = NULL;
1090	#endif
1091	# ifdef HAVE_JUMP_LABEL
1092	/* No jump lable stuff either. */
1093	pMyMod->jump_entries = NULL;
1094	pMyMod->num_jump_entries = 0;
1095	# endif
1096	# ifdef CONFIG_TRACING
1097	pMyMod->num_trace_bprintk_fmt = 0;
1098	pMyMod->trace_bprintk_fmt_start = NULL;
1099	# endif
1100	# ifdef CONFIG_EVENT_TRACING
1101	pMyMod->trace_events = NULL;
1102	pMyMod->num_trace_events = 0;
1103	# endif
1104	# ifdef CONFIG_FTRACE_MCOUNT_RECORD
1105	pMyMod->num_ftrace_callsites = 0;
1106	pMyMod->ftrace_callsites = NULL;
1107	# endif
1108	# ifdef CONFIG_MODULE_UNLOAD
1109	/* Dependency lists, not worth sharing */
1110	INIT_LIST_HEAD(&pMyMod->source_list);
1111	INIT_LIST_HEAD(&pMyMod->target_list);
1112
1113	/* Nobody waiting and no exit function. */
1114	# if LINUX_VERSION_CODE < KERNEL_VERSION(3, 13, 0)
1115	pMyMod->waiter = NULL;
1116	# endif
1117	pMyMod->exit = NULL;
1118
1119	/* References, very important as we must not allow the module
1120	to be unloaded using rmmod. */
1121	# if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 19, 0)
1122	atomic_set(&pMyMod->refcnt, 42);
1123	# else
1124	pMyMod->refptr = alloc_percpu(struct module_ref);
1125	if (pMyMod->refptr)
1126	{
1127	int iCpu;
1128	for_each_possible_cpu(iCpu)
1129	{
1130	per_cpu_ptr(pMyMod->refptr, iCpu)->decs = 0;
1131	per_cpu_ptr(pMyMod->refptr, iCpu)->incs = 1;
1132	}
1133	}
1134	else
1135	rc = VERR_NO_MEMORY;
1136	# endif
1137	# endif
1138	# ifdef CONFIG_CONSTRUCTORS
1139	/* No constructors. */
1140	pMyMod->ctors = NULL;
1141	pMyMod->num_ctors = 0;
1142	# endif
1143	if (RT_SUCCESS(rc))
1144	{
1145	bool fIsModText;
1146
1147	/*
1148	* Add the module to the list.
1149	*/
1150	mutex_lock(&module_mutex);
1151	list_add_rcu(&pMyMod->list, &pSelfMod->list);
1152	pImage->pLnxModHack = pMyMod;
1153	# ifdef CONFIG_MODULES_TREE_LOOKUP
1154	g_pfnModTreeInsert(&pMyMod->mtn_core); /* __mod_tree_insert */
1155	# endif
1156	mutex_unlock(&module_mutex);
1157
1158	/*
1159	* Test it.
1160	*/
1161	mutex_lock(&module_mutex);
1162	pTestModByName = find_module(pMyMod->name);
1163	pTestMod = __module_address((uintptr_t)pImage->pvImage + pImage->cbImageBits / 4);
1164	fIsModText = __module_text_address((uintptr_t)pImage->pvImage + pImage->cbImageBits / 2);
1165	mutex_unlock(&module_mutex);
1166	if ( pTestMod == pMyMod
1167	&& pTestModByName == pMyMod
1168	&& fIsModText)
1169	printk(KERN_ERR "vboxdrv: fake module works for '%s' (%#lx to %#lx)\n",
1170	pMyMod->name, (unsigned long)paSymbols[1].st_value, (unsigned long)paSymbols[2].st_value);
1171	else
1172	printk(KERN_ERR "vboxdrv: failed to find fake module (pTestMod=%p, pTestModByName=%p, pMyMod=%p, fIsModText=%d)\n",
1173	pTestMod, pTestModByName, pMyMod, fIsModText);
1174	}
1175	else
1176	RTMemFree(pMyMod);
1177	}
1178
1179	IPRT_LINUX_RESTORE_EFL_AC();
1180	#else
1181	pImage->pLnxModHack = NULL;
1182	#endif
1183	NOREF(pDevExt); NOREF(pImage);
1184	}
1185
1186
1187	void VBOXCALL supdrvOSLdrNotifyUnloaded(PSUPDRVDEVEXT pDevExt, PSUPDRVLDRIMAGE pImage)
1188	{
1189	#ifdef VBOX_WITH_NON_PROD_HACK_FOR_PERF_STACKS /* Not for production use!! Debugging only! */
1190	struct module *pMyMod = pImage->pLnxModHack;
1191	pImage->pLnxModHack = NULL;
1192	if (pMyMod)
1193	{
1194	/*
1195	* Remove the fake module list entry and free it.
1196	*/
1197	IPRT_LINUX_SAVE_EFL_AC();
1198	mutex_lock(&module_mutex);
1199	list_del_rcu(&pMyMod->list);
1200	# ifdef CONFIG_MODULES_TREE_LOOKUP
1201	g_pfnModTreeRemove(&pMyMod->mtn_core);
1202	# endif
1203	synchronize_sched();
1204	mutex_unlock(&module_mutex);
1205
1206	# if LINUX_VERSION_CODE < KERNEL_VERSION(3, 19, 0)
1207	free_percpu(pMyMod->refptr);
1208	# endif
1209	RTMemFree(pMyMod);
1210	IPRT_LINUX_RESTORE_EFL_AC();
1211	}
1212
1213	#else
1214	Assert(pImage->pLnxModHack == NULL);
1215	#endif
1216	NOREF(pDevExt); NOREF(pImage);
1217	}
1218
1219
1220	int VBOXCALL supdrvOSLdrQuerySymbol(PSUPDRVDEVEXT pDevExt, PSUPDRVLDRIMAGE pImage,
1221	const char pszSymbol, size_t cchSymbol, void *ppvSymbol)
1222	{
1223	#ifdef VBOX_WITH_NON_PROD_HACK_FOR_PERF_STACKS
1224	# error "implement me!"
1225	#endif
1226	RT_NOREF(pDevExt, pImage, pszSymbol, cchSymbol, ppvSymbol);
1227	return VERR_WRONG_ORDER;
1228	}
1229
1230
1231	#ifdef SUPDRV_WITH_MSR_PROBER
1232
1233	int VBOXCALL supdrvOSMsrProberRead(uint32_t uMsr, RTCPUID idCpu, uint64_t *puValue)
1234	{
1235	# ifdef SUPDRV_LINUX_HAS_SAFE_MSR_API
1236	uint32_t u32Low, u32High;
1237	int rc;
1238
1239	IPRT_LINUX_SAVE_EFL_AC();
1240	if (idCpu == NIL_RTCPUID)
1241	rc = rdmsr_safe(uMsr, &u32Low, &u32High);
1242	else if (RTMpIsCpuOnline(idCpu))
1243	rc = rdmsr_safe_on_cpu(idCpu, uMsr, &u32Low, &u32High);
1244	else
1245	return VERR_CPU_OFFLINE;
1246	IPRT_LINUX_RESTORE_EFL_AC();
1247	if (rc == 0)
1248	{
1249	*puValue = RT_MAKE_U64(u32Low, u32High);
1250	return VINF_SUCCESS;
1251	}
1252	return VERR_ACCESS_DENIED;
1253	# else
1254	return VERR_NOT_SUPPORTED;
1255	# endif
1256	}
1257
1258
1259	int VBOXCALL supdrvOSMsrProberWrite(uint32_t uMsr, RTCPUID idCpu, uint64_t uValue)
1260	{
1261	# ifdef SUPDRV_LINUX_HAS_SAFE_MSR_API
1262	int rc;
1263
1264	IPRT_LINUX_SAVE_EFL_AC();
1265	if (idCpu == NIL_RTCPUID)
1266	rc = wrmsr_safe(uMsr, RT_LODWORD(uValue), RT_HIDWORD(uValue));
1267	else if (RTMpIsCpuOnline(idCpu))
1268	rc = wrmsr_safe_on_cpu(idCpu, uMsr, RT_LODWORD(uValue), RT_HIDWORD(uValue));
1269	else
1270	return VERR_CPU_OFFLINE;
1271	IPRT_LINUX_RESTORE_EFL_AC();
1272
1273	if (rc == 0)
1274	return VINF_SUCCESS;
1275	return VERR_ACCESS_DENIED;
1276	# else
1277	return VERR_NOT_SUPPORTED;
1278	# endif
1279	}
1280
1281	# ifdef SUPDRV_LINUX_HAS_SAFE_MSR_API
1282	/**
1283	* Worker for supdrvOSMsrProberModify.
1284	*/
1285	static DECLCALLBACK(void) supdrvLnxMsrProberModifyOnCpu(RTCPUID idCpu, void pvUser1, void pvUser2)
1286	{
1287	PSUPMSRPROBER pReq = (PSUPMSRPROBER)pvUser1;
1288	register uint32_t uMsr = pReq->u.In.uMsr;
1289	bool const fFaster = pReq->u.In.enmOp == SUPMSRPROBEROP_MODIFY_FASTER;
1290	uint64_t uBefore;
1291	uint64_t uWritten;
1292	uint64_t uAfter;
1293	int rcBefore, rcWrite, rcAfter, rcRestore;
1294	RTCCUINTREG fOldFlags;
1295
1296	/* Initialize result variables. */
1297	uBefore = uWritten = uAfter = 0;
1298	rcWrite = rcAfter = rcRestore = -EIO;
1299
1300	/*
1301	* Do the job.
1302	*/
1303	fOldFlags = ASMIntDisableFlags();
1304	ASMCompilerBarrier(); /* paranoia */
1305	if (!fFaster)
1306	ASMWriteBackAndInvalidateCaches();
1307
1308	rcBefore = rdmsrl_safe(uMsr, &uBefore);
1309	if (rcBefore >= 0)
1310	{
1311	register uint64_t uRestore = uBefore;
1312	uWritten = uRestore;
1313	uWritten &= pReq->u.In.uArgs.Modify.fAndMask;
1314	uWritten \|= pReq->u.In.uArgs.Modify.fOrMask;
1315
1316	rcWrite = wrmsr_safe(uMsr, RT_LODWORD(uWritten), RT_HIDWORD(uWritten));
1317	rcAfter = rdmsrl_safe(uMsr, &uAfter);
1318	rcRestore = wrmsr_safe(uMsr, RT_LODWORD(uRestore), RT_HIDWORD(uRestore));
1319
1320	if (!fFaster)
1321	{
1322	ASMWriteBackAndInvalidateCaches();
1323	ASMReloadCR3();
1324	ASMNopPause();
1325	}
1326	}
1327
1328	ASMCompilerBarrier(); /* paranoia */
1329	ASMSetFlags(fOldFlags);
1330
1331	/*
1332	* Write out the results.
1333	*/
1334	pReq->u.Out.uResults.Modify.uBefore = uBefore;
1335	pReq->u.Out.uResults.Modify.uWritten = uWritten;
1336	pReq->u.Out.uResults.Modify.uAfter = uAfter;
1337	pReq->u.Out.uResults.Modify.fBeforeGp = rcBefore != 0;
1338	pReq->u.Out.uResults.Modify.fModifyGp = rcWrite != 0;
1339	pReq->u.Out.uResults.Modify.fAfterGp = rcAfter != 0;
1340	pReq->u.Out.uResults.Modify.fRestoreGp = rcRestore != 0;
1341	RT_ZERO(pReq->u.Out.uResults.Modify.afReserved);
1342	}
1343	# endif
1344
1345
1346	int VBOXCALL supdrvOSMsrProberModify(RTCPUID idCpu, PSUPMSRPROBER pReq)
1347	{
1348	# ifdef SUPDRV_LINUX_HAS_SAFE_MSR_API
1349	if (idCpu == NIL_RTCPUID)
1350	{
1351	supdrvLnxMsrProberModifyOnCpu(idCpu, pReq, NULL);
1352	return VINF_SUCCESS;
1353	}
1354	return RTMpOnSpecific(idCpu, supdrvLnxMsrProberModifyOnCpu, pReq, NULL);
1355	# else
1356	return VERR_NOT_SUPPORTED;
1357	# endif
1358	}
1359
1360	#endif /* SUPDRV_WITH_MSR_PROBER */
1361
1362
1363	/**
1364	* Converts a supdrv error code to an linux error code.
1365	*
1366	* @returns corresponding linux error code.
1367	* @param rc IPRT status code.
1368	*/
1369	static int VBoxDrvLinuxErr2LinuxErr(int rc)
1370	{
1371	switch (rc)
1372	{
1373	case VINF_SUCCESS: return 0;
1374	case VERR_GENERAL_FAILURE: return -EACCES;
1375	case VERR_INVALID_PARAMETER: return -EINVAL;
1376	case VERR_INVALID_MAGIC: return -EILSEQ;
1377	case VERR_INVALID_HANDLE: return -ENXIO;
1378	case VERR_INVALID_POINTER: return -EFAULT;
1379	case VERR_LOCK_FAILED: return -ENOLCK;
1380	case VERR_ALREADY_LOADED: return -EEXIST;
1381	case VERR_PERMISSION_DENIED: return -EPERM;
1382	case VERR_VERSION_MISMATCH: return -ENOSYS;
1383	case VERR_IDT_FAILED: return -1000;
1384	}
1385
1386	return -EPERM;
1387	}
1388
1389
1390	RTDECL(int) SUPR0Printf(const char *pszFormat, ...)
1391	{
1392	va_list va;
1393	char szMsg[512];
1394	IPRT_LINUX_SAVE_EFL_AC();
1395
1396	va_start(va, pszFormat);
1397	RTStrPrintfV(szMsg, sizeof(szMsg) - 1, pszFormat, va);
1398	va_end(va);
1399	szMsg[sizeof(szMsg) - 1] = '\0';
1400
1401	printk("%s", szMsg);
1402
1403	IPRT_LINUX_RESTORE_EFL_AC();
1404	return 0;
1405	}
1406
1407
1408	SUPR0DECL(uint32_t) SUPR0GetKernelFeatures(void)
1409	{
1410	uint32_t fFlags = 0;
1411	#ifdef CONFIG_PAX_KERNEXEC
1412	fFlags \|= SUPKERNELFEATURES_GDT_READ_ONLY;
1413	#endif
1414	#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 12, 0)
1415	fFlags \|= SUPKERNELFEATURES_GDT_NEED_WRITABLE;
1416	#endif
1417	#if defined(VBOX_STRICT) \|\| defined(VBOX_WITH_EFLAGS_AC_SET_IN_VBOXDRV)
1418	fFlags \|= SUPKERNELFEATURES_SMAP;
1419	#elif defined(CONFIG_X86_SMAP)
1420	if (ASMGetCR4() & X86_CR4_SMAP)
1421	fFlags \|= SUPKERNELFEATURES_SMAP;
1422	#endif
1423	return fFlags;
1424	}
1425
1426
1427	int VBOXCALL supdrvOSGetCurrentGdtRw(RTHCUINTPTR *pGdtRw)
1428	{
1429	#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 12, 0)
1430	*pGdtRw = (RTHCUINTPTR)get_current_gdt_rw();
1431	return VINF_SUCCESS;
1432	#else
1433	return VERR_NOT_IMPLEMENTED;
1434	#endif
1435	}
1436
1437
1438	module_init(VBoxDrvLinuxInit);
1439	module_exit(VBoxDrvLinuxUnload);
1440
1441	MODULE_AUTHOR(VBOX_VENDOR);
1442	MODULE_DESCRIPTION(VBOX_PRODUCT " Support Driver");
1443	MODULE_LICENSE("GPL");
1444	#ifdef MODULE_VERSION
1445	MODULE_VERSION(VBOX_VERSION_STRING " r" RT_XSTR(VBOX_SVN_REV) " (" RT_XSTR(SUPDRV_IOC_VERSION) ")");
1446	#endif
1447
1448	module_param(force_async_tsc, int, 0444);
1449	MODULE_PARM_DESC(force_async_tsc, "force the asynchronous TSC mode");
1450

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

source: vbox/trunk/src/VBox/HostDrivers/Support/linux/SUPDrv-linux.c@ 78351

Download in other formats: