VirtualBox

source: vbox/trunk/src/VBox/VMM/include/PGMInternal.h@ 87933

Last change on this file since 87933 was 87485, checked in by vboxsync, 4 years ago

VMM/PGM: Corrected todo on aGCPhysGstPaePDs & aGstPaePdpeRegs.

  • Property svn:eol-style set to native
  • Property svn:keywords set to Id Revision
File size: 181.9 KB
Line 
1/* $Id: PGMInternal.h 87485 2021-01-29 17:38:47Z vboxsync $ */
2/** @file
3 * PGM - Internal header file.
4 */
5
6/*
7 * Copyright (C) 2006-2020 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
18#ifndef VMM_INCLUDED_SRC_include_PGMInternal_h
19#define VMM_INCLUDED_SRC_include_PGMInternal_h
20#ifndef RT_WITHOUT_PRAGMA_ONCE
21# pragma once
22#endif
23
24#include <VBox/cdefs.h>
25#include <VBox/types.h>
26#include <VBox/err.h>
27#include <VBox/dbg.h>
28#include <VBox/vmm/stam.h>
29#include <VBox/param.h>
30#include <VBox/vmm/vmm.h>
31#include <VBox/vmm/mm.h>
32#include <VBox/vmm/pdmcritsect.h>
33#include <VBox/vmm/pdmapi.h>
34#include <VBox/dis.h>
35#include <VBox/vmm/dbgf.h>
36#include <VBox/log.h>
37#include <VBox/vmm/gmm.h>
38#include <VBox/vmm/hm.h>
39#include <VBox/vmm/hm_vmx.h>
40#include <iprt/asm.h>
41#include <iprt/assert.h>
42#include <iprt/avl.h>
43#include <iprt/critsect.h>
44#include <iprt/list-off32.h>
45#include <iprt/sha.h>
46
47
48
49/** @defgroup grp_pgm_int Internals
50 * @ingroup grp_pgm
51 * @internal
52 * @{
53 */
54
55
56/** @name PGM Compile Time Config
57 * @{
58 */
59
60/**
61 * Indicates that there are no guest mappings in the shadow tables.
62 *
63 * Note! In ring-3 the macro is also used to exclude the managment of the
64 * intermediate context page tables. On 32-bit systems we use the intermediate
65 * context to support 64-bit guest execution. Thus, we cannot fully make it
66 * without mappings there even when VBOX_WITH_RAW_MODE is not defined.
67 *
68 * In raw-mode context there are by design always guest mappings (the code is
69 * executed from one), while in ring-0 there are none at all. Neither context
70 * manages the page tables for intermediate switcher context, that's all done in
71 * ring-3.
72 *
73 * Update 6.1: It is always defined now, in pgm.h
74 */
75#if defined(IN_RING0) \
76 || ( !defined(VBOX_WITH_RAW_MODE) \
77 && ( HC_ARCH_BITS != 32 \
78 || !defined(VBOX_WITH_64_BITS_GUESTS) \
79 ) \
80 )
81# undef PGM_WITHOUT_MAPPINGS
82# define PGM_WITHOUT_MAPPINGS
83#endif
84
85/**
86 * Check and skip global PDEs for non-global flushes
87 */
88#define PGM_SKIP_GLOBAL_PAGEDIRS_ON_NONGLOBAL_FLUSH
89
90/**
91 * Optimization for PAE page tables that are modified often
92 */
93//#if 0 /* disabled again while debugging */
94#define PGMPOOL_WITH_OPTIMIZED_DIRTY_PT
95//#endif
96
97/**
98 * Large page support enabled only on 64 bits hosts; applies to nested paging only.
99 */
100#define PGM_WITH_LARGE_PAGES
101
102/**
103 * Enables optimizations for MMIO handlers that exploits X86_TRAP_PF_RSVD and
104 * VMX_EXIT_EPT_MISCONFIG.
105 */
106#define PGM_WITH_MMIO_OPTIMIZATIONS
107
108/**
109 * Sync N pages instead of a whole page table
110 */
111#define PGM_SYNC_N_PAGES
112
113/**
114 * Number of pages to sync during a page fault
115 *
116 * When PGMPOOL_WITH_GCPHYS_TRACKING is enabled using high values here
117 * causes a lot of unnecessary extents and also is slower than taking more \#PFs.
118 *
119 * Note that \#PFs are much more expensive in the VT-x/AMD-V case due to
120 * world switch overhead, so let's sync more.
121 */
122# ifdef IN_RING0
123/* Chose 32 based on the compile test in @bugref{4219}; 64 shows worse stats.
124 * 32 again shows better results than 16; slightly more overhead in the \#PF handler,
125 * but ~5% fewer faults.
126 */
127# define PGM_SYNC_NR_PAGES 32
128#else
129# define PGM_SYNC_NR_PAGES 8
130#endif
131
132/**
133 * Number of PGMPhysRead/Write cache entries (must be <= sizeof(uint64_t))
134 */
135#define PGM_MAX_PHYSCACHE_ENTRIES 64
136#define PGM_MAX_PHYSCACHE_ENTRIES_MASK (PGM_MAX_PHYSCACHE_ENTRIES-1)
137
138
139/** @def PGMPOOL_CFG_MAX_GROW
140 * The maximum number of pages to add to the pool in one go.
141 */
142#define PGMPOOL_CFG_MAX_GROW (_2M >> PAGE_SHIFT)
143
144/** @def VBOX_STRICT_PGM_HANDLER_VIRTUAL
145 * Enables some extra assertions for virtual handlers (mainly phys2virt related).
146 */
147#ifdef VBOX_STRICT
148# define VBOX_STRICT_PGM_HANDLER_VIRTUAL
149#endif
150
151/** @def VBOX_WITH_NEW_LAZY_PAGE_ALLOC
152 * Enables the experimental lazy page allocation code. */
153#ifdef DOXYGEN_RUNNING
154# define VBOX_WITH_NEW_LAZY_PAGE_ALLOC
155#endif
156
157/** @def VBOX_WITH_REAL_WRITE_MONITORED_PAGES
158 * Enables real write monitoring of pages, i.e. mapping them read-only and
159 * only making them writable when getting a write access \#PF. */
160#define VBOX_WITH_REAL_WRITE_MONITORED_PAGES
161
162/** @} */
163
164
165/** @name PDPT and PML4 flags.
166 * These are placed in the three bits available for system programs in
167 * the PDPT and PML4 entries.
168 * @{ */
169/** The entry is a permanent one and it's must always be present.
170 * Never free such an entry. */
171#define PGM_PLXFLAGS_PERMANENT RT_BIT_64(10)
172#ifndef PGM_WITHOUT_MAPPINGS
173/** Mapping (hypervisor allocated pagetable). */
174# define PGM_PLXFLAGS_MAPPING RT_BIT_64(11)
175# define PGM_PML4_FLAGS RT_BIT_64(11)
176#endif
177/** PGM specific bits in PML4 entries. */
178#define PGM_PML4_FLAGS 0
179/** PGM specific bits in PDPT entries. */
180#ifndef PGM_WITHOUT_MAPPINGS
181# define PGM_PDPT_FLAGS (PGM_PLXFLAGS_PERMANENT | PGM_PLXFLAGS_MAPPING)
182#else
183# define PGM_PDPT_FLAGS (PGM_PLXFLAGS_PERMANENT)
184#endif
185/** @} */
186
187/** @name Page directory flags.
188 * These are placed in the three bits available for system programs in
189 * the page directory entries.
190 * @{ */
191/** Indicates the original entry was a big page.
192 * @remarks This is currently only used for statistics and can be recycled. */
193#define PGM_PDFLAGS_BIG_PAGE RT_BIT_64(9)
194#ifndef PGM_WITHOUT_MAPPINGS
195/** Mapping (hypervisor allocated pagetable). */
196# define PGM_PDFLAGS_MAPPING RT_BIT_64(10)
197#endif
198/** Made read-only to facilitate dirty bit tracking. */
199#define PGM_PDFLAGS_TRACK_DIRTY RT_BIT_64(11)
200/** @} */
201
202/** @name Page flags.
203 * These are placed in the three bits available for system programs in
204 * the page entries.
205 * @{ */
206/** Made read-only to facilitate dirty bit tracking. */
207#define PGM_PTFLAGS_TRACK_DIRTY RT_BIT_64(9)
208
209#ifndef PGM_PTFLAGS_CSAM_VALIDATED
210/** Scanned and approved by CSAM (tm).
211 * NOTE: Must be identical to the one defined in CSAMInternal.h!!
212 * @todo Move PGM_PTFLAGS_* and PGM_PDFLAGS_* to VBox/vmm/pgm.h. */
213#define PGM_PTFLAGS_CSAM_VALIDATED RT_BIT_64(11)
214#endif
215
216/** @} */
217
218/** @name Defines used to indicate the shadow and guest paging in the templates.
219 * @{ */
220#define PGM_TYPE_REAL 1
221#define PGM_TYPE_PROT 2
222#define PGM_TYPE_32BIT 3
223#define PGM_TYPE_PAE 4
224#define PGM_TYPE_AMD64 5
225#define PGM_TYPE_NESTED_32BIT 6
226#define PGM_TYPE_NESTED_PAE 7
227#define PGM_TYPE_NESTED_AMD64 8
228#define PGM_TYPE_EPT 9
229#define PGM_TYPE_NONE 10 /**< Dummy shadow paging mode for NEM. */
230#define PGM_TYPE_END (PGM_TYPE_NONE + 1)
231#define PGM_TYPE_FIRST_SHADOW PGM_TYPE_32BIT /**< The first type used by shadow paging. */
232/** @} */
233
234/** Macro for checking if the guest is using paging.
235 * @param uGstType PGM_TYPE_*
236 * @param uShwType PGM_TYPE_*
237 * @remark ASSUMES certain order of the PGM_TYPE_* values.
238 */
239#define PGM_WITH_PAGING(uGstType, uShwType) \
240 ( (uGstType) >= PGM_TYPE_32BIT \
241 && (uShwType) < PGM_TYPE_NESTED_32BIT)
242
243/** Macro for checking if the guest supports the NX bit.
244 * @param uGstType PGM_TYPE_*
245 * @param uShwType PGM_TYPE_*
246 * @remark ASSUMES certain order of the PGM_TYPE_* values.
247 */
248#define PGM_WITH_NX(uGstType, uShwType) \
249 ( (uGstType) >= PGM_TYPE_PAE \
250 && (uShwType) < PGM_TYPE_NESTED_32BIT)
251
252/** Macro for checking for nested or EPT.
253 * @param uType PGM_TYPE_*
254 */
255#define PGM_TYPE_IS_NESTED(uType) \
256 ( (uType) == PGM_TYPE_NESTED_32BIT \
257 || (uType) == PGM_TYPE_NESTED_PAE \
258 || (uType) == PGM_TYPE_NESTED_AMD64)
259
260/** Macro for checking for nested or EPT.
261 * @param uType PGM_TYPE_*
262 */
263#define PGM_TYPE_IS_NESTED_OR_EPT(uType) \
264 ( (uType) == PGM_TYPE_NESTED_32BIT \
265 || (uType) == PGM_TYPE_NESTED_PAE \
266 || (uType) == PGM_TYPE_NESTED_AMD64 \
267 || (uType) == PGM_TYPE_EPT)
268
269
270
271/** @def PGM_HCPHYS_2_PTR
272 * Maps a HC physical page pool address to a virtual address.
273 *
274 * @returns VBox status code.
275 * @param pVM The cross context VM structure.
276 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
277 * @param HCPhys The HC physical address to map to a virtual one.
278 * @param ppv Where to store the virtual address. No need to cast
279 * this.
280 *
281 * @remark There is no need to assert on the result.
282 */
283#define PGM_HCPHYS_2_PTR(pVM, pVCpu, HCPhys, ppv) pgmPoolHCPhys2Ptr(pVM, HCPhys, (void **)(ppv))
284
285/** @def PGM_GCPHYS_2_PTR_V2
286 * Maps a GC physical page address to a virtual address.
287 *
288 * @returns VBox status code.
289 * @param pVM The cross context VM structure.
290 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
291 * @param GCPhys The GC physical address to map to a virtual one.
292 * @param ppv Where to store the virtual address. No need to cast this.
293 *
294 * @remark Use with care as we don't have so much dynamic mapping space in
295 * ring-0 on 32-bit darwin and in RC.
296 * @remark There is no need to assert on the result.
297 */
298#define PGM_GCPHYS_2_PTR_V2(pVM, pVCpu, GCPhys, ppv) \
299 pgmPhysGCPhys2R3Ptr(pVM, GCPhys, (PRTR3PTR)(ppv)) /** @todo this isn't asserting! */
300
301/** @def PGM_GCPHYS_2_PTR
302 * Maps a GC physical page address to a virtual address.
303 *
304 * @returns VBox status code.
305 * @param pVM The cross context VM structure.
306 * @param GCPhys The GC physical address to map to a virtual one.
307 * @param ppv Where to store the virtual address. No need to cast this.
308 *
309 * @remark Use with care as we don't have so much dynamic mapping space in
310 * ring-0 on 32-bit darwin and in RC.
311 * @remark There is no need to assert on the result.
312 */
313#define PGM_GCPHYS_2_PTR(pVM, GCPhys, ppv) PGM_GCPHYS_2_PTR_V2(pVM, VMMGetCpu(pVM), GCPhys, ppv)
314
315/** @def PGM_GCPHYS_2_PTR_BY_VMCPU
316 * Maps a GC physical page address to a virtual address.
317 *
318 * @returns VBox status code.
319 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
320 * @param GCPhys The GC physical address to map to a virtual one.
321 * @param ppv Where to store the virtual address. No need to cast this.
322 *
323 * @remark Use with care as we don't have so much dynamic mapping space in
324 * ring-0 on 32-bit darwin and in RC.
325 * @remark There is no need to assert on the result.
326 */
327#define PGM_GCPHYS_2_PTR_BY_VMCPU(pVCpu, GCPhys, ppv) PGM_GCPHYS_2_PTR_V2((pVCpu)->CTX_SUFF(pVM), pVCpu, GCPhys, ppv)
328
329/** @def PGM_GCPHYS_2_PTR_EX
330 * Maps a unaligned GC physical page address to a virtual address.
331 *
332 * @returns VBox status code.
333 * @param pVM The cross context VM structure.
334 * @param GCPhys The GC physical address to map to a virtual one.
335 * @param ppv Where to store the virtual address. No need to cast this.
336 *
337 * @remark Use with care as we don't have so much dynamic mapping space in
338 * ring-0 on 32-bit darwin and in RC.
339 * @remark There is no need to assert on the result.
340 */
341#define PGM_GCPHYS_2_PTR_EX(pVM, GCPhys, ppv) \
342 pgmPhysGCPhys2R3Ptr(pVM, GCPhys, (PRTR3PTR)(ppv)) /** @todo this isn't asserting! */
343
344/** @def PGM_DYNMAP_UNUSED_HINT
345 * Hints to the dynamic mapping code in RC and R0/darwin that the specified page
346 * is no longer used.
347 *
348 * For best effect only apply this to the page that was mapped most recently.
349 *
350 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
351 * @param pvPage The pool page.
352 */
353#define PGM_DYNMAP_UNUSED_HINT(pVCpu, pvPage) do {} while (0)
354
355/** @def PGM_DYNMAP_UNUSED_HINT_VM
356 * Hints to the dynamic mapping code in RC and R0/darwin that the specified page
357 * is no longer used.
358 *
359 * For best effect only apply this to the page that was mapped most recently.
360 *
361 * @param pVM The cross context VM structure.
362 * @param pvPage The pool page.
363 */
364#define PGM_DYNMAP_UNUSED_HINT_VM(pVM, pvPage) PGM_DYNMAP_UNUSED_HINT(VMMGetCpu(pVM), pvPage)
365
366
367/** @def PGM_INVL_PG
368 * Invalidates a page.
369 *
370 * @param pVCpu The cross context virtual CPU structure.
371 * @param GCVirt The virtual address of the page to invalidate.
372 */
373#ifdef IN_RING0
374# define PGM_INVL_PG(pVCpu, GCVirt) HMInvalidatePage(pVCpu, (RTGCPTR)(GCVirt))
375#elif defined(IN_RING3)
376# define PGM_INVL_PG(pVCpu, GCVirt) HMInvalidatePage(pVCpu, (RTGCPTR)(GCVirt))
377#else
378# error "Not IN_RING0 or IN_RING3!"
379#endif
380
381/** @def PGM_INVL_PG_ALL_VCPU
382 * Invalidates a page on all VCPUs
383 *
384 * @param pVM The cross context VM structure.
385 * @param GCVirt The virtual address of the page to invalidate.
386 */
387#ifdef IN_RING0
388# define PGM_INVL_PG_ALL_VCPU(pVM, GCVirt) HMInvalidatePageOnAllVCpus(pVM, (RTGCPTR)(GCVirt))
389#else
390# define PGM_INVL_PG_ALL_VCPU(pVM, GCVirt) HMInvalidatePageOnAllVCpus(pVM, (RTGCPTR)(GCVirt))
391#endif
392
393/** @def PGM_INVL_BIG_PG
394 * Invalidates a 4MB page directory entry.
395 *
396 * @param pVCpu The cross context virtual CPU structure.
397 * @param GCVirt The virtual address within the page directory to invalidate.
398 */
399#ifdef IN_RING0
400# define PGM_INVL_BIG_PG(pVCpu, GCVirt) HMFlushTlb(pVCpu)
401#else
402# define PGM_INVL_BIG_PG(pVCpu, GCVirt) HMFlushTlb(pVCpu)
403#endif
404
405/** @def PGM_INVL_VCPU_TLBS()
406 * Invalidates the TLBs of the specified VCPU
407 *
408 * @param pVCpu The cross context virtual CPU structure.
409 */
410#ifdef IN_RING0
411# define PGM_INVL_VCPU_TLBS(pVCpu) HMFlushTlb(pVCpu)
412#else
413# define PGM_INVL_VCPU_TLBS(pVCpu) HMFlushTlb(pVCpu)
414#endif
415
416/** @def PGM_INVL_ALL_VCPU_TLBS()
417 * Invalidates the TLBs of all VCPUs
418 *
419 * @param pVM The cross context VM structure.
420 */
421#ifdef IN_RING0
422# define PGM_INVL_ALL_VCPU_TLBS(pVM) HMFlushTlbOnAllVCpus(pVM)
423#else
424# define PGM_INVL_ALL_VCPU_TLBS(pVM) HMFlushTlbOnAllVCpus(pVM)
425#endif
426
427
428/** @name Safer Shadow PAE PT/PTE
429 * For helping avoid misinterpreting invalid PAE/AMD64 page table entries as
430 * present.
431 *
432 * @{
433 */
434#if 1
435/**
436 * For making sure that u1Present and X86_PTE_P checks doesn't mistake
437 * invalid entries for present.
438 * @sa X86PTEPAE.
439 */
440typedef union PGMSHWPTEPAE
441{
442 /** Unsigned integer view */
443 X86PGPAEUINT uCareful;
444 /* Not other views. */
445} PGMSHWPTEPAE;
446
447# define PGMSHWPTEPAE_IS_P(Pte) ( ((Pte).uCareful & (X86_PTE_P | X86_PTE_PAE_MBZ_MASK_NX)) == X86_PTE_P )
448# define PGMSHWPTEPAE_IS_RW(Pte) ( !!((Pte).uCareful & X86_PTE_RW))
449# define PGMSHWPTEPAE_IS_US(Pte) ( !!((Pte).uCareful & X86_PTE_US))
450# define PGMSHWPTEPAE_IS_A(Pte) ( !!((Pte).uCareful & X86_PTE_A))
451# define PGMSHWPTEPAE_IS_D(Pte) ( !!((Pte).uCareful & X86_PTE_D))
452# define PGMSHWPTEPAE_IS_TRACK_DIRTY(Pte) ( !!((Pte).uCareful & PGM_PTFLAGS_TRACK_DIRTY) )
453# define PGMSHWPTEPAE_IS_P_RW(Pte) ( ((Pte).uCareful & (X86_PTE_P | X86_PTE_RW | X86_PTE_PAE_MBZ_MASK_NX)) == (X86_PTE_P | X86_PTE_RW) )
454# define PGMSHWPTEPAE_GET_LOG(Pte) ( (Pte).uCareful )
455# define PGMSHWPTEPAE_GET_HCPHYS(Pte) ( (Pte).uCareful & X86_PTE_PAE_PG_MASK )
456# define PGMSHWPTEPAE_GET_U(Pte) ( (Pte).uCareful ) /**< Use with care. */
457# define PGMSHWPTEPAE_SET(Pte, uVal) do { (Pte).uCareful = (uVal); } while (0)
458# define PGMSHWPTEPAE_SET2(Pte, Pte2) do { (Pte).uCareful = (Pte2).uCareful; } while (0)
459# define PGMSHWPTEPAE_ATOMIC_SET(Pte, uVal) do { ASMAtomicWriteU64(&(Pte).uCareful, (uVal)); } while (0)
460# define PGMSHWPTEPAE_ATOMIC_SET2(Pte, Pte2) do { ASMAtomicWriteU64(&(Pte).uCareful, (Pte2).uCareful); } while (0)
461# define PGMSHWPTEPAE_SET_RO(Pte) do { (Pte).uCareful &= ~(X86PGPAEUINT)X86_PTE_RW; } while (0)
462# define PGMSHWPTEPAE_SET_RW(Pte) do { (Pte).uCareful |= X86_PTE_RW; } while (0)
463
464/**
465 * For making sure that u1Present and X86_PTE_P checks doesn't mistake
466 * invalid entries for present.
467 * @sa X86PTPAE.
468 */
469typedef struct PGMSHWPTPAE
470{
471 PGMSHWPTEPAE a[X86_PG_PAE_ENTRIES];
472} PGMSHWPTPAE;
473
474#else
475typedef X86PTEPAE PGMSHWPTEPAE;
476typedef X86PTPAE PGMSHWPTPAE;
477# define PGMSHWPTEPAE_IS_P(Pte) ( (Pte).n.u1Present )
478# define PGMSHWPTEPAE_IS_RW(Pte) ( (Pte).n.u1Write )
479# define PGMSHWPTEPAE_IS_US(Pte) ( (Pte).n.u1User )
480# define PGMSHWPTEPAE_IS_A(Pte) ( (Pte).n.u1Accessed )
481# define PGMSHWPTEPAE_IS_D(Pte) ( (Pte).n.u1Dirty )
482# define PGMSHWPTEPAE_IS_TRACK_DIRTY(Pte) ( !!((Pte).u & PGM_PTFLAGS_TRACK_DIRTY) )
483# define PGMSHWPTEPAE_IS_P_RW(Pte) ( ((Pte).u & (X86_PTE_P | X86_PTE_RW)) == (X86_PTE_P | X86_PTE_RW) )
484# define PGMSHWPTEPAE_GET_LOG(Pte) ( (Pte).u )
485# define PGMSHWPTEPAE_GET_HCPHYS(Pte) ( (Pte).u & X86_PTE_PAE_PG_MASK )
486# define PGMSHWPTEPAE_GET_U(Pte) ( (Pte).u ) /**< Use with care. */
487# define PGMSHWPTEPAE_SET(Pte, uVal) do { (Pte).u = (uVal); } while (0)
488# define PGMSHWPTEPAE_SET2(Pte, Pte2) do { (Pte).u = (Pte2).u; } while (0)
489# define PGMSHWPTEPAE_ATOMIC_SET(Pte, uVal) do { ASMAtomicWriteU64(&(Pte).u, (uVal)); } while (0)
490# define PGMSHWPTEPAE_ATOMIC_SET2(Pte, Pte2) do { ASMAtomicWriteU64(&(Pte).u, (Pte2).u); } while (0)
491# define PGMSHWPTEPAE_SET_RO(Pte) do { (Pte).u &= ~(X86PGPAEUINT)X86_PTE_RW; } while (0)
492# define PGMSHWPTEPAE_SET_RW(Pte) do { (Pte).u |= X86_PTE_RW; } while (0)
493
494#endif
495
496/** Pointer to a shadow PAE PTE. */
497typedef PGMSHWPTEPAE *PPGMSHWPTEPAE;
498/** Pointer to a const shadow PAE PTE. */
499typedef PGMSHWPTEPAE const *PCPGMSHWPTEPAE;
500
501/** Pointer to a shadow PAE page table. */
502typedef PGMSHWPTPAE *PPGMSHWPTPAE;
503/** Pointer to a const shadow PAE page table. */
504typedef PGMSHWPTPAE const *PCPGMSHWPTPAE;
505/** @} */
506
507#ifndef PGM_WITHOUT_MAPPINGS
508
509/** Size of the GCPtrConflict array in PGMMAPPING.
510 * @remarks Must be a power of two. */
511# define PGMMAPPING_CONFLICT_MAX 8
512
513/**
514 * Structure for tracking GC Mappings.
515 *
516 * This structure is used by linked list in both GC and HC.
517 */
518typedef struct PGMMAPPING
519{
520 /** Pointer to next entry. */
521 R3PTRTYPE(struct PGMMAPPING *) pNextR3;
522 /** Pointer to next entry. */
523 R0PTRTYPE(struct PGMMAPPING *) pNextR0;
524 /** Indicate whether this entry is finalized. */
525 bool fFinalized;
526 bool afPadding[7];
527 /** Start Virtual address. */
528 RTGCPTR GCPtr;
529 /** Last Virtual address (inclusive). */
530 RTGCPTR GCPtrLast;
531 /** Range size (bytes). */
532 RTGCPTR cb;
533 /** Pointer to relocation callback function. */
534 R3PTRTYPE(PFNPGMRELOCATE) pfnRelocate;
535 /** User argument to the callback. */
536 R3PTRTYPE(void *) pvUser;
537 /** Mapping description / name. For easing debugging. */
538 R3PTRTYPE(const char *) pszDesc;
539 /** Last 8 addresses that caused conflicts. */
540 RTGCPTR aGCPtrConflicts[PGMMAPPING_CONFLICT_MAX];
541 /** Number of conflicts for this hypervisor mapping. */
542 uint32_t cConflicts;
543 /** Number of page tables. */
544 uint32_t cPTs;
545
546 /** Array of page table mapping data. Each entry
547 * describes one page table. The array can be longer
548 * than the declared length.
549 */
550 struct
551 {
552 /** The HC physical address of the page table. */
553 RTHCPHYS HCPhysPT;
554 /** The HC physical address of the first PAE page table. */
555 RTHCPHYS HCPhysPaePT0;
556 /** The HC physical address of the second PAE page table. */
557 RTHCPHYS HCPhysPaePT1;
558 /** The HC virtual address of the 32-bit page table. */
559 R3PTRTYPE(PX86PT) pPTR3;
560 /** The HC virtual address of the two PAE page table. (i.e 1024 entries instead of 512) */
561 R3PTRTYPE(PPGMSHWPTPAE) paPaePTsR3;
562 /** The R0 virtual address of the 32-bit page table. */
563 R0PTRTYPE(PX86PT) pPTR0;
564 /** The R0 virtual address of the two PAE page table. */
565 R0PTRTYPE(PPGMSHWPTPAE) paPaePTsR0;
566 } aPTs[1];
567} PGMMAPPING;
568/** Pointer to structure for tracking GC Mappings. */
569typedef struct PGMMAPPING *PPGMMAPPING;
570
571#endif /* !PGM_WITHOUT_MAPPINGS */
572
573
574/**
575 * Physical page access handler type registration.
576 */
577typedef struct PGMPHYSHANDLERTYPEINT
578{
579 /** Number of references. */
580 uint32_t volatile cRefs;
581 /** Magic number (PGMPHYSHANDLERTYPEINT_MAGIC). */
582 uint32_t u32Magic;
583 /** Link of handler types anchored in PGMTREES::HeadPhysHandlerTypes. */
584 RTLISTOFF32NODE ListNode;
585 /** The kind of accesses we're handling. */
586 PGMPHYSHANDLERKIND enmKind;
587 /** The PGM_PAGE_HNDL_PHYS_STATE_XXX value corresponding to enmKind. */
588 uint32_t uState;
589 /** Pointer to R3 callback function. */
590 R3PTRTYPE(PFNPGMPHYSHANDLER) pfnHandlerR3;
591 /** Pointer to R0 callback function. */
592 R0PTRTYPE(PFNPGMPHYSHANDLER) pfnHandlerR0;
593 /** Pointer to R0 callback function for \#PFs. */
594 R0PTRTYPE(PFNPGMRZPHYSPFHANDLER) pfnPfHandlerR0;
595 /** Description / Name. For easing debugging. */
596 R3PTRTYPE(const char *) pszDesc;
597} PGMPHYSHANDLERTYPEINT;
598/** Pointer to a physical access handler type registration. */
599typedef PGMPHYSHANDLERTYPEINT *PPGMPHYSHANDLERTYPEINT;
600/** Magic value for the physical handler callbacks (Robert A. Heinlein). */
601#define PGMPHYSHANDLERTYPEINT_MAGIC UINT32_C(0x19070707)
602/** Magic value for the physical handler callbacks. */
603#define PGMPHYSHANDLERTYPEINT_MAGIC_DEAD UINT32_C(0x19880508)
604
605/**
606 * Converts a handle to a pointer.
607 * @returns PPGMPHYSHANDLERTYPEINT
608 * @param a_pVM The cross context VM structure.
609 * @param a_hType Physical access handler type handle.
610 */
611#define PGMPHYSHANDLERTYPEINT_FROM_HANDLE(a_pVM, a_hType) ((PPGMPHYSHANDLERTYPEINT)MMHyperHeapOffsetToPtr(a_pVM, a_hType))
612
613
614/**
615 * Physical page access handler structure.
616 *
617 * This is used to keep track of physical address ranges
618 * which are being monitored in some kind of way.
619 */
620typedef struct PGMPHYSHANDLER
621{
622 AVLROGCPHYSNODECORE Core;
623 /** Number of pages to update. */
624 uint32_t cPages;
625 /** Set if we have pages that have been aliased. */
626 uint32_t cAliasedPages;
627 /** Set if we have pages that have temporarily been disabled. */
628 uint32_t cTmpOffPages;
629 /** Registered handler type handle (heap offset). */
630 PGMPHYSHANDLERTYPE hType;
631 /** User argument for R3 handlers. */
632 R3PTRTYPE(void *) pvUserR3;
633 /** User argument for R0 handlers. */
634 R0PTRTYPE(void *) pvUserR0;
635 /** Description / Name. For easing debugging. */
636 R3PTRTYPE(const char *) pszDesc;
637#ifdef VBOX_WITH_STATISTICS
638 /** Profiling of this handler. */
639 STAMPROFILE Stat;
640#endif
641} PGMPHYSHANDLER;
642/** Pointer to a physical page access handler structure. */
643typedef PGMPHYSHANDLER *PPGMPHYSHANDLER;
644
645/**
646 * Gets the type record for a physical handler (no reference added).
647 * @returns PPGMPHYSHANDLERTYPEINT
648 * @param a_pVM The cross context VM structure.
649 * @param a_pPhysHandler Pointer to the physical handler structure
650 * (PGMPHYSHANDLER).
651 */
652#define PGMPHYSHANDLER_GET_TYPE(a_pVM, a_pPhysHandler) PGMPHYSHANDLERTYPEINT_FROM_HANDLE(a_pVM, (a_pPhysHandler)->hType)
653
654
655/**
656 * A Physical Guest Page tracking structure.
657 *
658 * The format of this structure is complicated because we have to fit a lot
659 * of information into as few bits as possible. The format is also subject
660 * to change (there is one coming up soon). Which means that for we'll be
661 * using PGM_PAGE_GET_*, PGM_PAGE_IS_ and PGM_PAGE_SET_* macros for *all*
662 * accesses to the structure.
663 */
664typedef union PGMPAGE
665{
666 /** Structured view. */
667 struct
668 {
669 /** 1:0 - The physical handler state (PGM_PAGE_HNDL_PHYS_STATE_*). */
670 uint64_t u2HandlerPhysStateY : 2;
671 /** 3:2 - Paging structure needed to map the page
672 * (PGM_PAGE_PDE_TYPE_*). */
673 uint64_t u2PDETypeY : 2;
674 /** 4 - Unused (was used by FTE for dirty tracking). */
675 uint64_t fUnused1 : 1;
676 /** 5 - Flag indicating that a write monitored page was written to
677 * when set. */
678 uint64_t fWrittenToY : 1;
679 /** 7:6 - Unused. */
680 uint64_t u2Unused0 : 2;
681 /** 9:8 - Unused (was used by PGM_PAGE_HNDL_VIRT_STATE_*). */
682 uint64_t u2Unused1 : 2;
683 /** 11:10 - NEM state bits. */
684 uint64_t u2NemStateY : 2;
685 /** 12:48 - The host physical frame number (shift left to get the
686 * address). */
687 uint64_t HCPhysFN : 36;
688 /** 50:48 - The page state. */
689 uint64_t uStateY : 3;
690 /** 51:53 - The page type (PGMPAGETYPE). */
691 uint64_t uTypeY : 3;
692 /** 63:54 - PTE index for usage tracking (page pool). */
693 uint64_t u10PteIdx : 10;
694
695 /** The GMM page ID.
696 * @remarks In the current implementation, MMIO2 and pages aliased to
697 * MMIO2 pages will be exploiting this field to calculate the
698 * ring-3 mapping address corresponding to the page.
699 * Later we may consider including MMIO2 management into GMM. */
700 uint32_t idPage;
701 /** Usage tracking (page pool). */
702 uint16_t u16TrackingY;
703 /** The number of read locks on this page. */
704 uint8_t cReadLocksY;
705 /** The number of write locks on this page. */
706 uint8_t cWriteLocksY;
707 } s;
708
709 /** 64-bit integer view. */
710 uint64_t au64[2];
711 /** 16-bit view. */
712 uint32_t au32[4];
713 /** 16-bit view. */
714 uint16_t au16[8];
715 /** 8-bit view. */
716 uint8_t au8[16];
717} PGMPAGE;
718AssertCompileSize(PGMPAGE, 16);
719/** Pointer to a physical guest page. */
720typedef PGMPAGE *PPGMPAGE;
721/** Pointer to a const physical guest page. */
722typedef const PGMPAGE *PCPGMPAGE;
723/** Pointer to a physical guest page pointer. */
724typedef PPGMPAGE *PPPGMPAGE;
725
726
727/**
728 * Clears the page structure.
729 * @param a_pPage Pointer to the physical guest page tracking structure.
730 */
731#define PGM_PAGE_CLEAR(a_pPage) \
732 do { \
733 (a_pPage)->au64[0] = 0; \
734 (a_pPage)->au64[1] = 0; \
735 } while (0)
736
737/**
738 * Initializes the page structure.
739 * @param a_pPage Pointer to the physical guest page tracking structure.
740 * @param a_HCPhys The host physical address of the page.
741 * @param a_idPage The (GMM) page ID of the page.
742 * @param a_uType The page type (PGMPAGETYPE).
743 * @param a_uState The page state (PGM_PAGE_STATE_XXX).
744 */
745#define PGM_PAGE_INIT(a_pPage, a_HCPhys, a_idPage, a_uType, a_uState) \
746 do { \
747 RTHCPHYS SetHCPhysTmp = (a_HCPhys); \
748 AssertFatal(!(SetHCPhysTmp & ~UINT64_C(0x0000fffffffff000))); \
749 (a_pPage)->au64[0] = SetHCPhysTmp; \
750 (a_pPage)->au64[1] = 0; \
751 (a_pPage)->s.idPage = (a_idPage); \
752 (a_pPage)->s.uStateY = (a_uState); \
753 (a_pPage)->s.uTypeY = (a_uType); \
754 } while (0)
755
756/**
757 * Initializes the page structure of a ZERO page.
758 * @param a_pPage Pointer to the physical guest page tracking structure.
759 * @param a_pVM The VM handle (for getting the zero page address).
760 * @param a_uType The page type (PGMPAGETYPE).
761 */
762#define PGM_PAGE_INIT_ZERO(a_pPage, a_pVM, a_uType) \
763 PGM_PAGE_INIT((a_pPage), (a_pVM)->pgm.s.HCPhysZeroPg, NIL_GMM_PAGEID, (a_uType), PGM_PAGE_STATE_ZERO)
764
765
766/** @name The Page state, PGMPAGE::uStateY.
767 * @{ */
768/** The zero page.
769 * This is a per-VM page that's never ever mapped writable. */
770#define PGM_PAGE_STATE_ZERO 0U
771/** A allocated page.
772 * This is a per-VM page allocated from the page pool (or wherever
773 * we get MMIO2 pages from if the type is MMIO2).
774 */
775#define PGM_PAGE_STATE_ALLOCATED 1U
776/** A allocated page that's being monitored for writes.
777 * The shadow page table mappings are read-only. When a write occurs, the
778 * fWrittenTo member is set, the page remapped as read-write and the state
779 * moved back to allocated. */
780#define PGM_PAGE_STATE_WRITE_MONITORED 2U
781/** The page is shared, aka. copy-on-write.
782 * This is a page that's shared with other VMs. */
783#define PGM_PAGE_STATE_SHARED 3U
784/** The page is ballooned, so no longer available for this VM. */
785#define PGM_PAGE_STATE_BALLOONED 4U
786/** @} */
787
788
789/** Asserts lock ownership in some of the PGM_PAGE_XXX macros. */
790#if defined(VBOX_STRICT) && 0 /** @todo triggers in pgmRZDynMapGCPageV2Inlined */
791# define PGM_PAGE_ASSERT_LOCK(a_pVM) PGM_LOCK_ASSERT_OWNER(a_pVM)
792#else
793# define PGM_PAGE_ASSERT_LOCK(a_pVM) do { } while (0)
794#endif
795
796/**
797 * Gets the page state.
798 * @returns page state (PGM_PAGE_STATE_*).
799 * @param a_pPage Pointer to the physical guest page tracking structure.
800 *
801 * @remarks See PGM_PAGE_GET_HCPHYS_NA for remarks about GCC and strict
802 * builds.
803 */
804#define PGM_PAGE_GET_STATE_NA(a_pPage) ( (a_pPage)->s.uStateY )
805#if defined(__GNUC__) && defined(VBOX_STRICT)
806# define PGM_PAGE_GET_STATE(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_STATE_NA(a_pPage); })
807#else
808# define PGM_PAGE_GET_STATE PGM_PAGE_GET_STATE_NA
809#endif
810
811/**
812 * Sets the page state.
813 * @param a_pVM The VM handle, only used for lock ownership assertions.
814 * @param a_pPage Pointer to the physical guest page tracking structure.
815 * @param a_uState The new page state.
816 */
817#define PGM_PAGE_SET_STATE(a_pVM, a_pPage, a_uState) \
818 do { (a_pPage)->s.uStateY = (a_uState); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
819
820
821/**
822 * Gets the host physical address of the guest page.
823 * @returns host physical address (RTHCPHYS).
824 * @param a_pPage Pointer to the physical guest page tracking structure.
825 *
826 * @remarks In strict builds on gcc platforms, this macro will make some ugly
827 * assumption about a valid pVM variable/parameter being in the
828 * current context. It will use this pVM variable to assert that the
829 * PGM lock is held. Use the PGM_PAGE_GET_HCPHYS_NA in contexts where
830 * pVM is not around.
831 */
832#if 0
833# define PGM_PAGE_GET_HCPHYS_NA(a_pPage) ( (a_pPage)->s.HCPhysFN << 12 )
834# define PGM_PAGE_GET_HCPHYS PGM_PAGE_GET_HCPHYS_NA
835#else
836# define PGM_PAGE_GET_HCPHYS_NA(a_pPage) ( (a_pPage)->au64[0] & UINT64_C(0x0000fffffffff000) )
837# if defined(__GNUC__) && defined(VBOX_STRICT)
838# define PGM_PAGE_GET_HCPHYS(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_HCPHYS_NA(a_pPage); })
839# else
840# define PGM_PAGE_GET_HCPHYS PGM_PAGE_GET_HCPHYS_NA
841# endif
842#endif
843
844/**
845 * Sets the host physical address of the guest page.
846 *
847 * @param a_pVM The VM handle, only used for lock ownership assertions.
848 * @param a_pPage Pointer to the physical guest page tracking structure.
849 * @param a_HCPhys The new host physical address.
850 */
851#define PGM_PAGE_SET_HCPHYS(a_pVM, a_pPage, a_HCPhys) \
852 do { \
853 RTHCPHYS const SetHCPhysTmp = (a_HCPhys); \
854 AssertFatal(!(SetHCPhysTmp & ~UINT64_C(0x0000fffffffff000))); \
855 (a_pPage)->s.HCPhysFN = SetHCPhysTmp >> 12; \
856 PGM_PAGE_ASSERT_LOCK(a_pVM); \
857 } while (0)
858
859/**
860 * Get the Page ID.
861 * @returns The Page ID; NIL_GMM_PAGEID if it's a ZERO page.
862 * @param a_pPage Pointer to the physical guest page tracking structure.
863 */
864#define PGM_PAGE_GET_PAGEID(a_pPage) ( (uint32_t)(a_pPage)->s.idPage )
865
866/**
867 * Sets the Page ID.
868 * @param a_pVM The VM handle, only used for lock ownership assertions.
869 * @param a_pPage Pointer to the physical guest page tracking structure.
870 * @param a_idPage The new page ID.
871 */
872#define PGM_PAGE_SET_PAGEID(a_pVM, a_pPage, a_idPage) \
873 do { \
874 (a_pPage)->s.idPage = (a_idPage); \
875 PGM_PAGE_ASSERT_LOCK(a_pVM); \
876 } while (0)
877
878/**
879 * Get the Chunk ID.
880 * @returns The Chunk ID; NIL_GMM_CHUNKID if it's a ZERO page.
881 * @param a_pPage Pointer to the physical guest page tracking structure.
882 */
883#define PGM_PAGE_GET_CHUNKID(a_pPage) ( PGM_PAGE_GET_PAGEID(a_pPage) >> GMM_CHUNKID_SHIFT )
884
885/**
886 * Get the index of the page within the allocation chunk.
887 * @returns The page index.
888 * @param a_pPage Pointer to the physical guest page tracking structure.
889 */
890#define PGM_PAGE_GET_PAGE_IN_CHUNK(a_pPage) ( PGM_PAGE_GET_PAGEID(a_pPage) & GMM_PAGEID_IDX_MASK )
891
892/**
893 * Gets the page type.
894 * @returns The page type.
895 * @param a_pPage Pointer to the physical guest page tracking structure.
896 *
897 * @remarks See PGM_PAGE_GET_HCPHYS_NA for remarks about GCC and strict
898 * builds.
899 */
900#define PGM_PAGE_GET_TYPE_NA(a_pPage) ( (a_pPage)->s.uTypeY )
901#if defined(__GNUC__) && defined(VBOX_STRICT)
902# define PGM_PAGE_GET_TYPE(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_TYPE_NA(a_pPage); })
903#else
904# define PGM_PAGE_GET_TYPE PGM_PAGE_GET_TYPE_NA
905#endif
906
907/**
908 * Sets the page type.
909 *
910 * @param a_pVM The VM handle, only used for lock ownership assertions.
911 * @param a_pPage Pointer to the physical guest page tracking structure.
912 * @param a_enmType The new page type (PGMPAGETYPE).
913 */
914#define PGM_PAGE_SET_TYPE(a_pVM, a_pPage, a_enmType) \
915 do { (a_pPage)->s.uTypeY = (a_enmType); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
916
917/**
918 * Gets the page table index
919 * @returns The page table index.
920 * @param a_pPage Pointer to the physical guest page tracking structure.
921 */
922#define PGM_PAGE_GET_PTE_INDEX(a_pPage) ( (a_pPage)->s.u10PteIdx )
923
924/**
925 * Sets the page table index.
926 * @param a_pVM The VM handle, only used for lock ownership assertions.
927 * @param a_pPage Pointer to the physical guest page tracking structure.
928 * @param a_iPte New page table index.
929 */
930#define PGM_PAGE_SET_PTE_INDEX(a_pVM, a_pPage, a_iPte) \
931 do { (a_pPage)->s.u10PteIdx = (a_iPte); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
932
933/**
934 * Checks if the page is marked for MMIO, no MMIO2 aliasing.
935 * @returns true/false.
936 * @param a_pPage Pointer to the physical guest page tracking structure.
937 */
938#define PGM_PAGE_IS_MMIO(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO )
939
940/**
941 * Checks if the page is marked for MMIO, including both aliases.
942 * @returns true/false.
943 * @param a_pPage Pointer to the physical guest page tracking structure.
944 */
945#define PGM_PAGE_IS_MMIO_OR_ALIAS(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO \
946 || (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO2_ALIAS_MMIO \
947 || (a_pPage)->s.uTypeY == PGMPAGETYPE_SPECIAL_ALIAS_MMIO \
948 )
949
950/**
951 * Checks if the page is marked for MMIO, including special aliases.
952 * @returns true/false.
953 * @param a_pPage Pointer to the physical guest page tracking structure.
954 */
955#define PGM_PAGE_IS_MMIO_OR_SPECIAL_ALIAS(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO \
956 || (a_pPage)->s.uTypeY == PGMPAGETYPE_SPECIAL_ALIAS_MMIO )
957
958/**
959 * Checks if the page is a special aliased MMIO page.
960 * @returns true/false.
961 * @param a_pPage Pointer to the physical guest page tracking structure.
962 */
963#define PGM_PAGE_IS_SPECIAL_ALIAS_MMIO(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_SPECIAL_ALIAS_MMIO )
964
965/**
966 * Checks if the page is backed by the ZERO page.
967 * @returns true/false.
968 * @param a_pPage Pointer to the physical guest page tracking structure.
969 */
970#define PGM_PAGE_IS_ZERO(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_ZERO )
971
972/**
973 * Checks if the page is backed by a SHARED page.
974 * @returns true/false.
975 * @param a_pPage Pointer to the physical guest page tracking structure.
976 */
977#define PGM_PAGE_IS_SHARED(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_SHARED )
978
979/**
980 * Checks if the page is ballooned.
981 * @returns true/false.
982 * @param a_pPage Pointer to the physical guest page tracking structure.
983 */
984#define PGM_PAGE_IS_BALLOONED(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_BALLOONED )
985
986/**
987 * Checks if the page is allocated.
988 * @returns true/false.
989 * @param a_pPage Pointer to the physical guest page tracking structure.
990 */
991#define PGM_PAGE_IS_ALLOCATED(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_ALLOCATED )
992
993/**
994 * Marks the page as written to (for GMM change monitoring).
995 * @param a_pVM The VM handle, only used for lock ownership assertions.
996 * @param a_pPage Pointer to the physical guest page tracking structure.
997 */
998#define PGM_PAGE_SET_WRITTEN_TO(a_pVM, a_pPage) \
999 do { (a_pPage)->s.fWrittenToY = 1; PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1000
1001/**
1002 * Clears the written-to indicator.
1003 * @param a_pVM The VM handle, only used for lock ownership assertions.
1004 * @param a_pPage Pointer to the physical guest page tracking structure.
1005 */
1006#define PGM_PAGE_CLEAR_WRITTEN_TO(a_pVM, a_pPage) \
1007 do { (a_pPage)->s.fWrittenToY = 0; PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1008
1009/**
1010 * Checks if the page was marked as written-to.
1011 * @returns true/false.
1012 * @param a_pPage Pointer to the physical guest page tracking structure.
1013 */
1014#define PGM_PAGE_IS_WRITTEN_TO(a_pPage) ( (a_pPage)->s.fWrittenToY )
1015
1016
1017/** @name PT usage values (PGMPAGE::u2PDEType).
1018 *
1019 * @{ */
1020/** Either as a PT or PDE. */
1021#define PGM_PAGE_PDE_TYPE_DONTCARE 0
1022/** Must use a page table to map the range. */
1023#define PGM_PAGE_PDE_TYPE_PT 1
1024/** Can use a page directory entry to map the continuous range. */
1025#define PGM_PAGE_PDE_TYPE_PDE 2
1026/** Can use a page directory entry to map the continuous range - temporarily disabled (by page monitoring). */
1027#define PGM_PAGE_PDE_TYPE_PDE_DISABLED 3
1028/** @} */
1029
1030/**
1031 * Set the PDE type of the page
1032 * @param a_pVM The VM handle, only used for lock ownership assertions.
1033 * @param a_pPage Pointer to the physical guest page tracking structure.
1034 * @param a_uType PGM_PAGE_PDE_TYPE_*.
1035 */
1036#define PGM_PAGE_SET_PDE_TYPE(a_pVM, a_pPage, a_uType) \
1037 do { (a_pPage)->s.u2PDETypeY = (a_uType); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1038
1039/**
1040 * Checks if the page was marked being part of a large page
1041 * @returns true/false.
1042 * @param a_pPage Pointer to the physical guest page tracking structure.
1043 */
1044#define PGM_PAGE_GET_PDE_TYPE(a_pPage) ( (a_pPage)->s.u2PDETypeY )
1045
1046/** @name Physical Access Handler State values (PGMPAGE::u2HandlerPhysStateY).
1047 *
1048 * @remarks The values are assigned in order of priority, so we can calculate
1049 * the correct state for a page with different handlers installed.
1050 * @{ */
1051/** No handler installed. */
1052#define PGM_PAGE_HNDL_PHYS_STATE_NONE 0
1053/** Monitoring is temporarily disabled. */
1054#define PGM_PAGE_HNDL_PHYS_STATE_DISABLED 1
1055/** Write access is monitored. */
1056#define PGM_PAGE_HNDL_PHYS_STATE_WRITE 2
1057/** All access is monitored. */
1058#define PGM_PAGE_HNDL_PHYS_STATE_ALL 3
1059/** @} */
1060
1061/**
1062 * Gets the physical access handler state of a page.
1063 * @returns PGM_PAGE_HNDL_PHYS_STATE_* value.
1064 * @param a_pPage Pointer to the physical guest page tracking structure.
1065 */
1066#define PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) ( (a_pPage)->s.u2HandlerPhysStateY )
1067
1068/**
1069 * Sets the physical access handler state of a page.
1070 * @param a_pPage Pointer to the physical guest page tracking structure.
1071 * @param a_uState The new state value.
1072 */
1073#define PGM_PAGE_SET_HNDL_PHYS_STATE(a_pPage, a_uState) \
1074 do { (a_pPage)->s.u2HandlerPhysStateY = (a_uState); } while (0)
1075
1076/**
1077 * Checks if the page has any physical access handlers, including temporarily disabled ones.
1078 * @returns true/false
1079 * @param a_pPage Pointer to the physical guest page tracking structure.
1080 */
1081#define PGM_PAGE_HAS_ANY_PHYSICAL_HANDLERS(a_pPage) \
1082 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) != PGM_PAGE_HNDL_PHYS_STATE_NONE )
1083
1084/**
1085 * Checks if the page has any active physical access handlers.
1086 * @returns true/false
1087 * @param a_pPage Pointer to the physical guest page tracking structure.
1088 */
1089#define PGM_PAGE_HAS_ACTIVE_PHYSICAL_HANDLERS(a_pPage) \
1090 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) >= PGM_PAGE_HNDL_PHYS_STATE_WRITE )
1091
1092/**
1093 * Checks if the page has any access handlers, including temporarily disabled ones.
1094 * @returns true/false
1095 * @param a_pPage Pointer to the physical guest page tracking structure.
1096 */
1097#define PGM_PAGE_HAS_ANY_HANDLERS(a_pPage) \
1098 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) != PGM_PAGE_HNDL_PHYS_STATE_NONE )
1099
1100/**
1101 * Checks if the page has any active access handlers.
1102 * @returns true/false
1103 * @param a_pPage Pointer to the physical guest page tracking structure.
1104 */
1105#define PGM_PAGE_HAS_ACTIVE_HANDLERS(a_pPage) \
1106 (PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) >= PGM_PAGE_HNDL_PHYS_STATE_WRITE )
1107
1108/**
1109 * Checks if the page has any active access handlers catching all accesses.
1110 * @returns true/false
1111 * @param a_pPage Pointer to the physical guest page tracking structure.
1112 */
1113#define PGM_PAGE_HAS_ACTIVE_ALL_HANDLERS(a_pPage) \
1114 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) == PGM_PAGE_HNDL_PHYS_STATE_ALL )
1115
1116
1117/** @def PGM_PAGE_GET_TRACKING
1118 * Gets the packed shadow page pool tracking data associated with a guest page.
1119 * @returns uint16_t containing the data.
1120 * @param a_pPage Pointer to the physical guest page tracking structure.
1121 */
1122#define PGM_PAGE_GET_TRACKING_NA(a_pPage) ( (a_pPage)->s.u16TrackingY )
1123#if defined(__GNUC__) && defined(VBOX_STRICT)
1124# define PGM_PAGE_GET_TRACKING(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_TRACKING_NA(a_pPage); })
1125#else
1126# define PGM_PAGE_GET_TRACKING PGM_PAGE_GET_TRACKING_NA
1127#endif
1128
1129/** @def PGM_PAGE_SET_TRACKING
1130 * Sets the packed shadow page pool tracking data associated with a guest page.
1131 * @param a_pVM The VM handle, only used for lock ownership assertions.
1132 * @param a_pPage Pointer to the physical guest page tracking structure.
1133 * @param a_u16TrackingData The tracking data to store.
1134 */
1135#define PGM_PAGE_SET_TRACKING(a_pVM, a_pPage, a_u16TrackingData) \
1136 do { (a_pPage)->s.u16TrackingY = (a_u16TrackingData); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1137
1138/** @def PGM_PAGE_GET_TD_CREFS
1139 * Gets the @a cRefs tracking data member.
1140 * @returns cRefs.
1141 * @param a_pPage Pointer to the physical guest page tracking structure.
1142 */
1143#define PGM_PAGE_GET_TD_CREFS(a_pPage) \
1144 ((PGM_PAGE_GET_TRACKING(a_pPage) >> PGMPOOL_TD_CREFS_SHIFT) & PGMPOOL_TD_CREFS_MASK)
1145#define PGM_PAGE_GET_TD_CREFS_NA(a_pPage) \
1146 ((PGM_PAGE_GET_TRACKING_NA(a_pPage) >> PGMPOOL_TD_CREFS_SHIFT) & PGMPOOL_TD_CREFS_MASK)
1147
1148/** @def PGM_PAGE_GET_TD_IDX
1149 * Gets the @a idx tracking data member.
1150 * @returns idx.
1151 * @param a_pPage Pointer to the physical guest page tracking structure.
1152 */
1153#define PGM_PAGE_GET_TD_IDX(a_pPage) \
1154 ((PGM_PAGE_GET_TRACKING(a_pPage) >> PGMPOOL_TD_IDX_SHIFT) & PGMPOOL_TD_IDX_MASK)
1155#define PGM_PAGE_GET_TD_IDX_NA(a_pPage) \
1156 ((PGM_PAGE_GET_TRACKING_NA(a_pPage) >> PGMPOOL_TD_IDX_SHIFT) & PGMPOOL_TD_IDX_MASK)
1157
1158
1159/** Max number of locks on a page. */
1160#define PGM_PAGE_MAX_LOCKS UINT8_C(254)
1161
1162/** Get the read lock count.
1163 * @returns count.
1164 * @param a_pPage Pointer to the physical guest page tracking structure.
1165 */
1166#define PGM_PAGE_GET_READ_LOCKS(a_pPage) ( (a_pPage)->s.cReadLocksY )
1167
1168/** Get the write lock count.
1169 * @returns count.
1170 * @param a_pPage Pointer to the physical guest page tracking structure.
1171 */
1172#define PGM_PAGE_GET_WRITE_LOCKS(a_pPage) ( (a_pPage)->s.cWriteLocksY )
1173
1174/** Decrement the read lock counter.
1175 * @param a_pPage Pointer to the physical guest page tracking structure.
1176 */
1177#define PGM_PAGE_DEC_READ_LOCKS(a_pPage) do { --(a_pPage)->s.cReadLocksY; } while (0)
1178
1179/** Decrement the write lock counter.
1180 * @param a_pPage Pointer to the physical guest page tracking structure.
1181 */
1182#define PGM_PAGE_DEC_WRITE_LOCKS(a_pPage) do { --(a_pPage)->s.cWriteLocksY; } while (0)
1183
1184/** Increment the read lock counter.
1185 * @param a_pPage Pointer to the physical guest page tracking structure.
1186 */
1187#define PGM_PAGE_INC_READ_LOCKS(a_pPage) do { ++(a_pPage)->s.cReadLocksY; } while (0)
1188
1189/** Increment the write lock counter.
1190 * @param a_pPage Pointer to the physical guest page tracking structure.
1191 */
1192#define PGM_PAGE_INC_WRITE_LOCKS(a_pPage) do { ++(a_pPage)->s.cWriteLocksY; } while (0)
1193
1194
1195/** Gets the NEM state.
1196 * @returns NEM state value (two bits).
1197 * @param a_pPage Pointer to the physical guest page tracking structure.
1198 */
1199#define PGM_PAGE_GET_NEM_STATE(a_pPage) ((a_pPage)->s.u2NemStateY)
1200
1201/** Sets the NEM state.
1202 * @param a_pPage Pointer to the physical guest page tracking structure.
1203 * @param a_u2State The NEM state value (specific to NEM impl.).
1204 */
1205#define PGM_PAGE_SET_NEM_STATE(a_pPage, a_u2State) \
1206 do { Assert((a_u2State) < 4); (a_pPage)->s.u2NemStateY = (a_u2State); } while (0)
1207
1208
1209#if 0
1210/** Enables sanity checking of write monitoring using CRC-32. */
1211# define PGMLIVESAVERAMPAGE_WITH_CRC32
1212#endif
1213
1214/**
1215 * Per page live save tracking data.
1216 */
1217typedef struct PGMLIVESAVERAMPAGE
1218{
1219 /** Number of times it has been dirtied. */
1220 uint32_t cDirtied : 24;
1221 /** Whether it is currently dirty. */
1222 uint32_t fDirty : 1;
1223 /** Ignore the page.
1224 * This is used for pages that has been MMIO, MMIO2 or ROM pages once. We will
1225 * deal with these after pausing the VM and DevPCI have said it bit about
1226 * remappings. */
1227 uint32_t fIgnore : 1;
1228 /** Was a ZERO page last time around. */
1229 uint32_t fZero : 1;
1230 /** Was a SHARED page last time around. */
1231 uint32_t fShared : 1;
1232 /** Whether the page is/was write monitored in a previous pass. */
1233 uint32_t fWriteMonitored : 1;
1234 /** Whether the page is/was write monitored earlier in this pass. */
1235 uint32_t fWriteMonitoredJustNow : 1;
1236 /** Bits reserved for future use. */
1237 uint32_t u2Reserved : 2;
1238#ifdef PGMLIVESAVERAMPAGE_WITH_CRC32
1239 /** CRC-32 for the page. This is for internal consistency checks. */
1240 uint32_t u32Crc;
1241#endif
1242} PGMLIVESAVERAMPAGE;
1243#ifdef PGMLIVESAVERAMPAGE_WITH_CRC32
1244AssertCompileSize(PGMLIVESAVERAMPAGE, 8);
1245#else
1246AssertCompileSize(PGMLIVESAVERAMPAGE, 4);
1247#endif
1248/** Pointer to the per page live save tracking data. */
1249typedef PGMLIVESAVERAMPAGE *PPGMLIVESAVERAMPAGE;
1250
1251/** The max value of PGMLIVESAVERAMPAGE::cDirtied. */
1252#define PGMLIVSAVEPAGE_MAX_DIRTIED 0x00fffff0
1253
1254
1255/**
1256 * RAM range for GC Phys to HC Phys conversion.
1257 *
1258 * Can be used for HC Virt to GC Phys and HC Virt to HC Phys
1259 * conversions too, but we'll let MM handle that for now.
1260 *
1261 * This structure is used by linked lists in both GC and HC.
1262 */
1263typedef struct PGMRAMRANGE
1264{
1265 /** Start of the range. Page aligned. */
1266 RTGCPHYS GCPhys;
1267 /** Size of the range. (Page aligned of course). */
1268 RTGCPHYS cb;
1269 /** Pointer to the next RAM range - for R3. */
1270 R3PTRTYPE(struct PGMRAMRANGE *) pNextR3;
1271 /** Pointer to the next RAM range - for R0. */
1272 R0PTRTYPE(struct PGMRAMRANGE *) pNextR0;
1273 /** PGM_RAM_RANGE_FLAGS_* flags. */
1274 uint32_t fFlags;
1275 uint32_t fPadding1;
1276 /** Last address in the range (inclusive). Page aligned (-1). */
1277 RTGCPHYS GCPhysLast;
1278 /** Start of the HC mapping of the range. This is only used for MMIO2. */
1279 R3PTRTYPE(void *) pvR3;
1280 /** Live save per page tracking data. */
1281 R3PTRTYPE(PPGMLIVESAVERAMPAGE) paLSPages;
1282 /** The range description. */
1283 R3PTRTYPE(const char *) pszDesc;
1284 /** Pointer to self - R0 pointer. */
1285 R0PTRTYPE(struct PGMRAMRANGE *) pSelfR0;
1286
1287 /** Pointer to the left search three node - ring-3 context. */
1288 R3PTRTYPE(struct PGMRAMRANGE *) pLeftR3;
1289 /** Pointer to the right search three node - ring-3 context. */
1290 R3PTRTYPE(struct PGMRAMRANGE *) pRightR3;
1291 /** Pointer to the left search three node - ring-0 context. */
1292 R0PTRTYPE(struct PGMRAMRANGE *) pLeftR0;
1293 /** Pointer to the right search three node - ring-0 context. */
1294 R0PTRTYPE(struct PGMRAMRANGE *) pRightR0;
1295
1296 /** Padding to make aPage aligned on sizeof(PGMPAGE). */
1297#if HC_ARCH_BITS == 32
1298 uint32_t au32Alignment2[HC_ARCH_BITS == 32 ? 2 : 0];
1299#endif
1300 /** Array of physical guest page tracking structures. */
1301 PGMPAGE aPages[1];
1302} PGMRAMRANGE;
1303/** Pointer to RAM range for GC Phys to HC Phys conversion. */
1304typedef PGMRAMRANGE *PPGMRAMRANGE;
1305
1306/** @name PGMRAMRANGE::fFlags
1307 * @{ */
1308/** The RAM range is floating around as an independent guest mapping. */
1309#define PGM_RAM_RANGE_FLAGS_FLOATING RT_BIT(20)
1310/** Ad hoc RAM range for an ROM mapping. */
1311#define PGM_RAM_RANGE_FLAGS_AD_HOC_ROM RT_BIT(21)
1312/** Ad hoc RAM range for an MMIO mapping. */
1313#define PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO RT_BIT(22)
1314/** Ad hoc RAM range for an MMIO2 or pre-registered MMIO mapping. */
1315#define PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO_EX RT_BIT(23)
1316/** @} */
1317
1318/** Tests if a RAM range is an ad hoc one or not.
1319 * @returns true/false.
1320 * @param pRam The RAM range.
1321 */
1322#define PGM_RAM_RANGE_IS_AD_HOC(pRam) \
1323 (!!( (pRam)->fFlags & (PGM_RAM_RANGE_FLAGS_AD_HOC_ROM | PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO | PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO_EX) ) )
1324
1325/** The number of entries in the RAM range TLBs (there is one for each
1326 * context). Must be a power of two. */
1327#define PGM_RAMRANGE_TLB_ENTRIES 8
1328
1329/**
1330 * Calculates the RAM range TLB index for the physical address.
1331 *
1332 * @returns RAM range TLB index.
1333 * @param a_GCPhys The guest physical address.
1334 */
1335#define PGM_RAMRANGE_TLB_IDX(a_GCPhys) ( ((a_GCPhys) >> 20) & (PGM_RAMRANGE_TLB_ENTRIES - 1) )
1336
1337
1338
1339/**
1340 * Per page tracking structure for ROM image.
1341 *
1342 * A ROM image may have a shadow page, in which case we may have two pages
1343 * backing it. This structure contains the PGMPAGE for both while
1344 * PGMRAMRANGE have a copy of the active one. It is important that these
1345 * aren't out of sync in any regard other than page pool tracking data.
1346 */
1347typedef struct PGMROMPAGE
1348{
1349 /** The page structure for the virgin ROM page. */
1350 PGMPAGE Virgin;
1351 /** The page structure for the shadow RAM page. */
1352 PGMPAGE Shadow;
1353 /** The current protection setting. */
1354 PGMROMPROT enmProt;
1355 /** Live save status information. Makes use of unused alignment space. */
1356 struct
1357 {
1358 /** The previous protection value. */
1359 uint8_t u8Prot;
1360 /** Written to flag set by the handler. */
1361 bool fWrittenTo;
1362 /** Whether the shadow page is dirty or not. */
1363 bool fDirty;
1364 /** Whether it was dirtied in the recently. */
1365 bool fDirtiedRecently;
1366 } LiveSave;
1367} PGMROMPAGE;
1368AssertCompileSizeAlignment(PGMROMPAGE, 8);
1369/** Pointer to a ROM page tracking structure. */
1370typedef PGMROMPAGE *PPGMROMPAGE;
1371
1372
1373/**
1374 * A registered ROM image.
1375 *
1376 * This is needed to keep track of ROM image since they generally intrude
1377 * into a PGMRAMRANGE. It also keeps track of additional info like the
1378 * two page sets (read-only virgin and read-write shadow), the current
1379 * state of each page.
1380 *
1381 * Because access handlers cannot easily be executed in a different
1382 * context, the ROM ranges needs to be accessible and in all contexts.
1383 */
1384typedef struct PGMROMRANGE
1385{
1386 /** Pointer to the next range - R3. */
1387 R3PTRTYPE(struct PGMROMRANGE *) pNextR3;
1388 /** Pointer to the next range - R0. */
1389 R0PTRTYPE(struct PGMROMRANGE *) pNextR0;
1390 /** Address of the range. */
1391 RTGCPHYS GCPhys;
1392 /** Address of the last byte in the range. */
1393 RTGCPHYS GCPhysLast;
1394 /** Size of the range. */
1395 RTGCPHYS cb;
1396 /** The flags (PGMPHYS_ROM_FLAGS_*). */
1397 uint32_t fFlags;
1398 /** The saved state range ID. */
1399 uint8_t idSavedState;
1400 /** Alignment padding. */
1401 uint8_t au8Alignment[3];
1402 /** Alignment padding ensuring that aPages is sizeof(PGMROMPAGE) aligned. */
1403 uint32_t au32Alignemnt[HC_ARCH_BITS == 32 ? 5 : 1];
1404 /** The size bits pvOriginal points to. */
1405 uint32_t cbOriginal;
1406 /** Pointer to the original bits when PGMPHYS_ROM_FLAGS_PERMANENT_BINARY was specified.
1407 * This is used for strictness checks. */
1408 R3PTRTYPE(const void *) pvOriginal;
1409 /** The ROM description. */
1410 R3PTRTYPE(const char *) pszDesc;
1411 /** The per page tracking structures. */
1412 PGMROMPAGE aPages[1];
1413} PGMROMRANGE;
1414/** Pointer to a ROM range. */
1415typedef PGMROMRANGE *PPGMROMRANGE;
1416
1417
1418/**
1419 * Live save per page data for an MMIO2 page.
1420 *
1421 * Not using PGMLIVESAVERAMPAGE here because we cannot use normal write monitoring
1422 * of MMIO2 pages. The current approach is using some optimistic SHA-1 +
1423 * CRC-32 for detecting changes as well as special handling of zero pages. This
1424 * is a TEMPORARY measure which isn't perfect, but hopefully it is good enough
1425 * for speeding things up. (We're using SHA-1 and not SHA-256 or SHA-512
1426 * because of speed (2.5x and 6x slower).)
1427 *
1428 * @todo Implement dirty MMIO2 page reporting that can be enabled during live
1429 * save but normally is disabled. Since we can write monitor guest
1430 * accesses on our own, we only need this for host accesses. Shouldn't be
1431 * too difficult for DevVGA, VMMDev might be doable, the planned
1432 * networking fun will be fun since it involves ring-0.
1433 */
1434typedef struct PGMLIVESAVEMMIO2PAGE
1435{
1436 /** Set if the page is considered dirty. */
1437 bool fDirty;
1438 /** The number of scans this page has remained unchanged for.
1439 * Only updated for dirty pages. */
1440 uint8_t cUnchangedScans;
1441 /** Whether this page was zero at the last scan. */
1442 bool fZero;
1443 /** Alignment padding. */
1444 bool fReserved;
1445 /** CRC-32 for the first half of the page.
1446 * This is used together with u32CrcH2 to quickly detect changes in the page
1447 * during the non-final passes. */
1448 uint32_t u32CrcH1;
1449 /** CRC-32 for the second half of the page. */
1450 uint32_t u32CrcH2;
1451 /** SHA-1 for the saved page.
1452 * This is used in the final pass to skip pages without changes. */
1453 uint8_t abSha1Saved[RTSHA1_HASH_SIZE];
1454} PGMLIVESAVEMMIO2PAGE;
1455/** Pointer to a live save status data for an MMIO2 page. */
1456typedef PGMLIVESAVEMMIO2PAGE *PPGMLIVESAVEMMIO2PAGE;
1457
1458/**
1459 * A registered MMIO2 (= Device RAM) range.
1460 *
1461 * There are a few reason why we need to keep track of these registrations. One
1462 * of them is the deregistration & cleanup stuff, while another is that the
1463 * PGMRAMRANGE associated with such a region may have to be removed from the ram
1464 * range list.
1465 *
1466 * Overlapping with a RAM range has to be 100% or none at all. The pages in the
1467 * existing RAM range must not be ROM nor MMIO. A guru meditation will be
1468 * raised if a partial overlap or an overlap of ROM pages is encountered. On an
1469 * overlap we will free all the existing RAM pages and put in the ram range
1470 * pages instead.
1471 */
1472typedef struct PGMREGMMIO2RANGE
1473{
1474 /** The owner of the range. (a device) */
1475 PPDMDEVINSR3 pDevInsR3;
1476 /** Pointer to the ring-3 mapping of the allocation. */
1477 RTR3PTR pvR3;
1478#if defined(VBOX_WITH_RAM_IN_KERNEL) && !defined(VBOX_WITH_LINEAR_HOST_PHYS_MEM)
1479 /** Pointer to the ring-0 mapping of the allocation. */
1480 RTR0PTR pvR0;
1481#endif
1482 /** Pointer to the next range - R3. */
1483 R3PTRTYPE(struct PGMREGMMIO2RANGE *) pNextR3;
1484 /** Flags (PGMREGMMIO2RANGE_F_XXX). */
1485 uint16_t fFlags;
1486 /** The sub device number (internal PCI config (CFGM) number). */
1487 uint8_t iSubDev;
1488 /** The PCI region number. */
1489 uint8_t iRegion;
1490 /** The saved state range ID. */
1491 uint8_t idSavedState;
1492 /** MMIO2 range identifier, for page IDs (PGMPAGE::s.idPage). */
1493 uint8_t idMmio2;
1494 /** Alignment padding for putting the ram range on a PGMPAGE alignment boundary. */
1495#if defined(VBOX_WITH_RAM_IN_KERNEL) && !defined(VBOX_WITH_LINEAR_HOST_PHYS_MEM)
1496 uint8_t abAlignment[HC_ARCH_BITS == 32 ? 6 + 4 : 2];
1497#else
1498 uint8_t abAlignment[HC_ARCH_BITS == 32 ? 6 + 8 : 2 + 8];
1499#endif
1500 /** The real size.
1501 * This may be larger than indicated by RamRange.cb if the range has been
1502 * reduced during saved state loading. */
1503 RTGCPHYS cbReal;
1504 /** Pointer to the physical handler for MMIO. */
1505 R3PTRTYPE(PPGMPHYSHANDLER) pPhysHandlerR3;
1506 /** Live save per page tracking data for MMIO2. */
1507 R3PTRTYPE(PPGMLIVESAVEMMIO2PAGE) paLSPages;
1508 /** The associated RAM range. */
1509 PGMRAMRANGE RamRange;
1510} PGMREGMMIO2RANGE;
1511AssertCompileMemberAlignment(PGMREGMMIO2RANGE, RamRange, 16);
1512/** Pointer to a MMIO2 or pre-registered MMIO range. */
1513typedef PGMREGMMIO2RANGE *PPGMREGMMIO2RANGE;
1514
1515/** @name PGMREGMMIO2RANGE_F_XXX - Registered MMIO2 range flags.
1516 * @{ */
1517/** Set if it's an MMIO2 range.
1518 * @note Historical. For a while we did some of the MMIO this way too. */
1519#define PGMREGMMIO2RANGE_F_MMIO2 UINT16_C(0x0001)
1520/** Set if this is the first chunk in the MMIO2 range. */
1521#define PGMREGMMIO2RANGE_F_FIRST_CHUNK UINT16_C(0x0002)
1522/** Set if this is the last chunk in the MMIO2 range. */
1523#define PGMREGMMIO2RANGE_F_LAST_CHUNK UINT16_C(0x0004)
1524/** Set if the whole range is mapped. */
1525#define PGMREGMMIO2RANGE_F_MAPPED UINT16_C(0x0008)
1526/** Set if it's overlapping, clear if not. */
1527#define PGMREGMMIO2RANGE_F_OVERLAPPING UINT16_C(0x0010)
1528/** @} */
1529
1530
1531/** @name Internal MMIO2 constants.
1532 * @{ */
1533/** The maximum number of MMIO2 ranges. */
1534#define PGM_MMIO2_MAX_RANGES 32
1535/** The maximum number of pages in a MMIO2 range. */
1536#define PGM_MMIO2_MAX_PAGE_COUNT UINT32_C(0x01000000)
1537/** Makes a MMIO2 page ID out of a MMIO2 range ID and page index number. */
1538#define PGM_MMIO2_PAGEID_MAKE(a_idMmio2, a_iPage) ( ((uint32_t)(a_idMmio2) << 24) | (uint32_t)(a_iPage) )
1539/** Gets the MMIO2 range ID from an MMIO2 page ID. */
1540#define PGM_MMIO2_PAGEID_GET_MMIO2_ID(a_idPage) ( (uint8_t)((a_idPage) >> 24) )
1541/** Gets the MMIO2 page index from an MMIO2 page ID. */
1542#define PGM_MMIO2_PAGEID_GET_IDX(a_idPage) ( ((a_idPage) & UINT32_C(0x00ffffff)) )
1543/** @} */
1544
1545
1546
1547/**
1548 * PGMPhysRead/Write cache entry
1549 */
1550typedef struct PGMPHYSCACHEENTRY
1551{
1552 /** R3 pointer to physical page. */
1553 R3PTRTYPE(uint8_t *) pbR3;
1554 /** GC Physical address for cache entry */
1555 RTGCPHYS GCPhys;
1556#if HC_ARCH_BITS == 64 && GC_ARCH_BITS == 32
1557 RTGCPHYS u32Padding0; /**< alignment padding. */
1558#endif
1559} PGMPHYSCACHEENTRY;
1560
1561/**
1562 * PGMPhysRead/Write cache to reduce REM memory access overhead
1563 */
1564typedef struct PGMPHYSCACHE
1565{
1566 /** Bitmap of valid cache entries */
1567 uint64_t aEntries;
1568 /** Cache entries */
1569 PGMPHYSCACHEENTRY Entry[PGM_MAX_PHYSCACHE_ENTRIES];
1570} PGMPHYSCACHE;
1571
1572
1573/** @name Ring-3 page mapping TLBs
1574 * @{ */
1575
1576/** Pointer to an allocation chunk ring-3 mapping. */
1577typedef struct PGMCHUNKR3MAP *PPGMCHUNKR3MAP;
1578/** Pointer to an allocation chunk ring-3 mapping pointer. */
1579typedef PPGMCHUNKR3MAP *PPPGMCHUNKR3MAP;
1580
1581/**
1582 * Ring-3 tracking structure for an allocation chunk ring-3 mapping.
1583 *
1584 * The primary tree (Core) uses the chunk id as key.
1585 */
1586typedef struct PGMCHUNKR3MAP
1587{
1588 /** The key is the chunk id. */
1589 AVLU32NODECORE Core;
1590 /** The time (ChunkR3Map.iNow) this chunk was last used. Used for unmap
1591 * selection. */
1592 uint32_t iLastUsed;
1593 /** The current reference count. */
1594 uint32_t volatile cRefs;
1595 /** The current permanent reference count. */
1596 uint32_t volatile cPermRefs;
1597 /** The mapping address. */
1598 void *pv;
1599} PGMCHUNKR3MAP;
1600
1601/**
1602 * Allocation chunk ring-3 mapping TLB entry.
1603 */
1604typedef struct PGMCHUNKR3MAPTLBE
1605{
1606 /** The chunk id. */
1607 uint32_t volatile idChunk;
1608#if HC_ARCH_BITS == 64
1609 uint32_t u32Padding; /**< alignment padding. */
1610#endif
1611 /** The chunk map. */
1612#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1613 R3PTRTYPE(PPGMCHUNKR3MAP) volatile pChunk;
1614#else
1615 R3R0PTRTYPE(PPGMCHUNKR3MAP) volatile pChunk;
1616#endif
1617} PGMCHUNKR3MAPTLBE;
1618/** Pointer to the an allocation chunk ring-3 mapping TLB entry. */
1619typedef PGMCHUNKR3MAPTLBE *PPGMCHUNKR3MAPTLBE;
1620
1621/** The number of TLB entries in PGMCHUNKR3MAPTLB.
1622 * @remark Must be a power of two value. */
1623#define PGM_CHUNKR3MAPTLB_ENTRIES 64
1624
1625/**
1626 * Allocation chunk ring-3 mapping TLB.
1627 *
1628 * @remarks We use a TLB to speed up lookups by avoiding walking the AVL.
1629 * At first glance this might look kinda odd since AVL trees are
1630 * supposed to give the most optimal lookup times of all trees
1631 * due to their balancing. However, take a tree with 1023 nodes
1632 * in it, that's 10 levels, meaning that most searches has to go
1633 * down 9 levels before they find what they want. This isn't fast
1634 * compared to a TLB hit. There is the factor of cache misses,
1635 * and of course the problem with trees and branch prediction.
1636 * This is why we use TLBs in front of most of the trees.
1637 *
1638 * @todo Generalize this TLB + AVL stuff, shouldn't be all that
1639 * difficult when we switch to the new inlined AVL trees (from kStuff).
1640 */
1641typedef struct PGMCHUNKR3MAPTLB
1642{
1643 /** The TLB entries. */
1644 PGMCHUNKR3MAPTLBE aEntries[PGM_CHUNKR3MAPTLB_ENTRIES];
1645} PGMCHUNKR3MAPTLB;
1646
1647/**
1648 * Calculates the index of a guest page in the Ring-3 Chunk TLB.
1649 * @returns Chunk TLB index.
1650 * @param idChunk The Chunk ID.
1651 */
1652#define PGM_CHUNKR3MAPTLB_IDX(idChunk) ( (idChunk) & (PGM_CHUNKR3MAPTLB_ENTRIES - 1) )
1653
1654
1655/**
1656 * Ring-3 guest page mapping TLB entry.
1657 * @remarks used in ring-0 as well at the moment.
1658 */
1659typedef struct PGMPAGER3MAPTLBE
1660{
1661 /** Address of the page. */
1662 RTGCPHYS volatile GCPhys;
1663 /** The guest page. */
1664#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1665 R3PTRTYPE(PPGMPAGE) volatile pPage;
1666#else
1667 R3R0PTRTYPE(PPGMPAGE) volatile pPage;
1668#endif
1669 /** Pointer to the page mapping tracking structure, PGMCHUNKR3MAP. */
1670#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1671 R3PTRTYPE(PPGMCHUNKR3MAP) volatile pMap;
1672#else
1673 R3R0PTRTYPE(PPGMCHUNKR3MAP) volatile pMap;
1674#endif
1675 /** The address */
1676#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1677 R3PTRTYPE(void *) volatile pv;
1678#else
1679 R3R0PTRTYPE(void *) volatile pv;
1680#endif
1681#if HC_ARCH_BITS == 32
1682 uint32_t u32Padding; /**< alignment padding. */
1683#endif
1684} PGMPAGER3MAPTLBE;
1685/** Pointer to an entry in the HC physical TLB. */
1686typedef PGMPAGER3MAPTLBE *PPGMPAGER3MAPTLBE;
1687
1688
1689/** The number of entries in the ring-3 guest page mapping TLB.
1690 * @remarks The value must be a power of two. */
1691#define PGM_PAGER3MAPTLB_ENTRIES 256
1692
1693/**
1694 * Ring-3 guest page mapping TLB.
1695 * @remarks used in ring-0 as well at the moment.
1696 */
1697typedef struct PGMPAGER3MAPTLB
1698{
1699 /** The TLB entries. */
1700 PGMPAGER3MAPTLBE aEntries[PGM_PAGER3MAPTLB_ENTRIES];
1701} PGMPAGER3MAPTLB;
1702/** Pointer to the ring-3 guest page mapping TLB. */
1703typedef PGMPAGER3MAPTLB *PPGMPAGER3MAPTLB;
1704
1705/**
1706 * Calculates the index of the TLB entry for the specified guest page.
1707 * @returns Physical TLB index.
1708 * @param GCPhys The guest physical address.
1709 */
1710#define PGM_PAGER3MAPTLB_IDX(GCPhys) ( ((GCPhys) >> PAGE_SHIFT) & (PGM_PAGER3MAPTLB_ENTRIES - 1) )
1711
1712/** @} */
1713
1714#if defined(VBOX_WITH_RAM_IN_KERNEL) || defined(DOXYGEN_RUNNING)
1715/** @name Ring-0 page mapping TLB
1716 * @{ */
1717/**
1718 * Ring-0 guest page mapping TLB entry.
1719 */
1720typedef struct PGMPAGER0MAPTLBE
1721{
1722 /** Address of the page. */
1723 RTGCPHYS volatile GCPhys;
1724 /** The guest page. */
1725 R0PTRTYPE(PPGMPAGE) volatile pPage;
1726 /** The address */
1727 R0PTRTYPE(void *) volatile pv;
1728} PGMPAGER0MAPTLBE;
1729/** Pointer to an entry in the HC physical TLB. */
1730typedef PGMPAGER0MAPTLBE *PPGMPAGER0MAPTLBE;
1731
1732
1733/** The number of entries in the ring-3 guest page mapping TLB.
1734 * @remarks The value must be a power of two. */
1735#define PGM_PAGER0MAPTLB_ENTRIES 256
1736
1737/**
1738 * Ring-3 guest page mapping TLB.
1739 * @remarks used in ring-0 as well at the moment.
1740 */
1741typedef struct PGMPAGER0MAPTLB
1742{
1743 /** The TLB entries. */
1744 PGMPAGER0MAPTLBE aEntries[PGM_PAGER0MAPTLB_ENTRIES];
1745} PGMPAGER0MAPTLB;
1746/** Pointer to the ring-3 guest page mapping TLB. */
1747typedef PGMPAGER0MAPTLB *PPGMPAGER0MAPTLB;
1748
1749/**
1750 * Calculates the index of the TLB entry for the specified guest page.
1751 * @returns Physical TLB index.
1752 * @param GCPhys The guest physical address.
1753 */
1754#define PGM_PAGER0MAPTLB_IDX(GCPhys) ( ((GCPhys) >> PAGE_SHIFT) & (PGM_PAGER0MAPTLB_ENTRIES - 1) )
1755/** @} */
1756#endif /* VBOX_WITH_RAM_IN_KERNEL || DOXYGEN_RUNNING */
1757
1758/**
1759 * Raw-mode context dynamic mapping cache entry.
1760 *
1761 * Because of raw-mode context being reloctable and all relocations are applied
1762 * in ring-3, this has to be defined here and be RC specific.
1763 *
1764 * @sa PGMRZDYNMAPENTRY, PGMR0DYNMAPENTRY.
1765 */
1766typedef struct PGMRCDYNMAPENTRY
1767{
1768 /** The physical address of the currently mapped page.
1769 * This is duplicate for three reasons: cache locality, cache policy of the PT
1770 * mappings and sanity checks. */
1771 RTHCPHYS HCPhys;
1772 /** Pointer to the page. */
1773 RTRCPTR pvPage;
1774 /** The number of references. */
1775 int32_t volatile cRefs;
1776 /** PTE pointer union. */
1777 struct PGMRCDYNMAPENTRY_PPTE
1778 {
1779 /** PTE pointer, 32-bit legacy version. */
1780 RCPTRTYPE(PX86PTE) pLegacy;
1781 /** PTE pointer, PAE version. */
1782 RCPTRTYPE(PX86PTEPAE) pPae;
1783 } uPte;
1784} PGMRCDYNMAPENTRY;
1785/** Pointer to a dynamic mapping cache entry for the raw-mode context. */
1786typedef PGMRCDYNMAPENTRY *PPGMRCDYNMAPENTRY;
1787
1788
1789/**
1790 * Dynamic mapping cache for the raw-mode context.
1791 *
1792 * This is initialized during VMMRC init based upon the pbDynPageMapBaseGC and
1793 * paDynPageMap* PGM members. However, it has to be defined in PGMInternal.h
1794 * so that we can perform relocations from PGMR3Relocate. This has the
1795 * consequence that we must have separate ring-0 and raw-mode context versions
1796 * of this struct even if they share the basic elements.
1797 *
1798 * @sa PPGMRZDYNMAP, PGMR0DYNMAP.
1799 */
1800typedef struct PGMRCDYNMAP
1801{
1802 /** The usual magic number / eye catcher (PGMRZDYNMAP_MAGIC). */
1803 uint32_t u32Magic;
1804 /** Array for tracking and managing the pages. */
1805 RCPTRTYPE(PPGMRCDYNMAPENTRY) paPages;
1806 /** The cache size given as a number of pages. */
1807 uint32_t cPages;
1808 /** The current load.
1809 * This does not include guard pages. */
1810 uint32_t cLoad;
1811 /** The max load ever.
1812 * This is maintained to get trigger adding of more mapping space. */
1813 uint32_t cMaxLoad;
1814 /** The number of guard pages. */
1815 uint32_t cGuardPages;
1816 /** The number of users (protected by hInitLock). */
1817 uint32_t cUsers;
1818} PGMRCDYNMAP;
1819/** Pointer to the dynamic cache for the raw-mode context. */
1820typedef PGMRCDYNMAP *PPGMRCDYNMAP;
1821
1822
1823/**
1824 * Mapping cache usage set entry.
1825 *
1826 * @remarks 16-bit ints was chosen as the set is not expected to be used beyond
1827 * the dynamic ring-0 and (to some extent) raw-mode context mapping
1828 * cache. If it's extended to include ring-3, well, then something
1829 * will have be changed here...
1830 */
1831typedef struct PGMMAPSETENTRY
1832{
1833 /** Pointer to the page. */
1834 RTR0PTR pvPage;
1835 /** The mapping cache index. */
1836 uint16_t iPage;
1837 /** The number of references.
1838 * The max is UINT16_MAX - 1. */
1839 uint16_t cRefs;
1840 /** The number inlined references.
1841 * The max is UINT16_MAX - 1. */
1842 uint16_t cInlinedRefs;
1843 /** Unreferences. */
1844 uint16_t cUnrefs;
1845
1846#if HC_ARCH_BITS == 32
1847 uint32_t u32Alignment1;
1848#endif
1849 /** The physical address for this entry. */
1850 RTHCPHYS HCPhys;
1851} PGMMAPSETENTRY;
1852AssertCompileMemberOffset(PGMMAPSETENTRY, iPage, RT_MAX(sizeof(RTR0PTR), sizeof(RTRCPTR)));
1853AssertCompileMemberAlignment(PGMMAPSETENTRY, HCPhys, sizeof(RTHCPHYS));
1854/** Pointer to a mapping cache usage set entry. */
1855typedef PGMMAPSETENTRY *PPGMMAPSETENTRY;
1856
1857/**
1858 * Mapping cache usage set.
1859 *
1860 * This is used in ring-0 and the raw-mode context to track dynamic mappings
1861 * done during exits / traps. The set is
1862 */
1863typedef struct PGMMAPSET
1864{
1865 /** The number of occupied entries.
1866 * This is PGMMAPSET_CLOSED if the set is closed and we're not supposed to do
1867 * dynamic mappings. */
1868 uint32_t cEntries;
1869 /** The start of the current subset.
1870 * This is UINT32_MAX if no subset is currently open. */
1871 uint32_t iSubset;
1872 /** The index of the current CPU, only valid if the set is open. */
1873 int32_t iCpu;
1874 uint32_t alignment;
1875 /** The entries. */
1876 PGMMAPSETENTRY aEntries[64];
1877 /** HCPhys -> iEntry fast lookup table.
1878 * Use PGMMAPSET_HASH for hashing.
1879 * The entries may or may not be valid, check against cEntries. */
1880 uint8_t aiHashTable[128];
1881} PGMMAPSET;
1882AssertCompileSizeAlignment(PGMMAPSET, 8);
1883/** Pointer to the mapping cache set. */
1884typedef PGMMAPSET *PPGMMAPSET;
1885
1886/** PGMMAPSET::cEntries value for a closed set. */
1887#define PGMMAPSET_CLOSED UINT32_C(0xdeadc0fe)
1888
1889/** Hash function for aiHashTable. */
1890#define PGMMAPSET_HASH(HCPhys) (((HCPhys) >> PAGE_SHIFT) & 127)
1891
1892
1893/** @name Context neutral page mapper TLB.
1894 *
1895 * Hoping to avoid some code and bug duplication parts of the GCxxx->CCPtr
1896 * code is writting in a kind of context neutral way. Time will show whether
1897 * this actually makes sense or not...
1898 *
1899 * @todo this needs to be reconsidered and dropped/redone since the ring-0
1900 * context ends up using a global mapping cache on some platforms
1901 * (darwin).
1902 *
1903 * @{ */
1904/** @typedef PPGMPAGEMAPTLB
1905 * The page mapper TLB pointer type for the current context. */
1906/** @typedef PPGMPAGEMAPTLB
1907 * The page mapper TLB entry pointer type for the current context. */
1908/** @typedef PPGMPAGEMAPTLB
1909 * The page mapper TLB entry pointer pointer type for the current context. */
1910/** @def PGM_PAGEMAPTLB_ENTRIES
1911 * The number of TLB entries in the page mapper TLB for the current context. */
1912/** @def PGM_PAGEMAPTLB_IDX
1913 * Calculate the TLB index for a guest physical address.
1914 * @returns The TLB index.
1915 * @param GCPhys The guest physical address. */
1916/** @typedef PPGMPAGEMAP
1917 * Pointer to a page mapper unit for current context. */
1918/** @typedef PPPGMPAGEMAP
1919 * Pointer to a page mapper unit pointer for current context. */
1920#if defined(IN_RING0) && defined(VBOX_WITH_RAM_IN_KERNEL)
1921typedef PPGMPAGER0MAPTLB PPGMPAGEMAPTLB;
1922typedef PPGMPAGER0MAPTLBE PPGMPAGEMAPTLBE;
1923typedef PPGMPAGER0MAPTLBE *PPPGMPAGEMAPTLBE;
1924# define PGM_PAGEMAPTLB_ENTRIES PGM_PAGER0MAPTLB_ENTRIES
1925# define PGM_PAGEMAPTLB_IDX(GCPhys) PGM_PAGER0MAPTLB_IDX(GCPhys)
1926typedef struct PGMCHUNKR0MAP *PPGMPAGEMAP;
1927typedef struct PGMCHUNKR0MAP **PPPGMPAGEMAP;
1928#else
1929typedef PPGMPAGER3MAPTLB PPGMPAGEMAPTLB;
1930typedef PPGMPAGER3MAPTLBE PPGMPAGEMAPTLBE;
1931typedef PPGMPAGER3MAPTLBE *PPPGMPAGEMAPTLBE;
1932# define PGM_PAGEMAPTLB_ENTRIES PGM_PAGER3MAPTLB_ENTRIES
1933# define PGM_PAGEMAPTLB_IDX(GCPhys) PGM_PAGER3MAPTLB_IDX(GCPhys)
1934typedef PPGMCHUNKR3MAP PPGMPAGEMAP;
1935typedef PPPGMCHUNKR3MAP PPPGMPAGEMAP;
1936#endif
1937/** @} */
1938
1939
1940/** @name PGM Pool Indexes.
1941 * Aka. the unique shadow page identifier.
1942 * @{ */
1943/** NIL page pool IDX. */
1944#define NIL_PGMPOOL_IDX 0
1945/** The first normal index. There used to be 5 fictive pages up front, now
1946 * there is only the NIL page. */
1947#define PGMPOOL_IDX_FIRST 1
1948/** The last valid index. (inclusive, 14 bits) */
1949#define PGMPOOL_IDX_LAST 0x3fff
1950/** @} */
1951
1952/** The NIL index for the parent chain. */
1953#define NIL_PGMPOOL_USER_INDEX ((uint16_t)0xffff)
1954#define NIL_PGMPOOL_PRESENT_INDEX ((uint16_t)0xffff)
1955
1956/**
1957 * Node in the chain linking a shadowed page to it's parent (user).
1958 */
1959#pragma pack(1)
1960typedef struct PGMPOOLUSER
1961{
1962 /** The index to the next item in the chain. NIL_PGMPOOL_USER_INDEX is no next. */
1963 uint16_t iNext;
1964 /** The user page index. */
1965 uint16_t iUser;
1966 /** Index into the user table. */
1967 uint32_t iUserTable;
1968} PGMPOOLUSER, *PPGMPOOLUSER;
1969typedef const PGMPOOLUSER *PCPGMPOOLUSER;
1970#pragma pack()
1971
1972
1973/** The NIL index for the phys ext chain. */
1974#define NIL_PGMPOOL_PHYSEXT_INDEX ((uint16_t)0xffff)
1975/** The NIL pte index for a phys ext chain slot. */
1976#define NIL_PGMPOOL_PHYSEXT_IDX_PTE ((uint16_t)0xffff)
1977
1978/**
1979 * Node in the chain of physical cross reference extents.
1980 * @todo Calling this an 'extent' is not quite right, find a better name.
1981 * @todo find out the optimal size of the aidx array
1982 */
1983#pragma pack(1)
1984typedef struct PGMPOOLPHYSEXT
1985{
1986 /** The index to the next item in the chain. NIL_PGMPOOL_PHYSEXT_INDEX is no next. */
1987 uint16_t iNext;
1988 /** Alignment. */
1989 uint16_t u16Align;
1990 /** The user page index. */
1991 uint16_t aidx[3];
1992 /** The page table index or NIL_PGMPOOL_PHYSEXT_IDX_PTE if unknown. */
1993 uint16_t apte[3];
1994} PGMPOOLPHYSEXT, *PPGMPOOLPHYSEXT;
1995typedef const PGMPOOLPHYSEXT *PCPGMPOOLPHYSEXT;
1996#pragma pack()
1997
1998
1999/**
2000 * The kind of page that's being shadowed.
2001 */
2002typedef enum PGMPOOLKIND
2003{
2004 /** The virtual invalid 0 entry. */
2005 PGMPOOLKIND_INVALID = 0,
2006 /** The entry is free (=unused). */
2007 PGMPOOLKIND_FREE,
2008
2009 /** Shw: 32-bit page table; Gst: no paging. */
2010 PGMPOOLKIND_32BIT_PT_FOR_PHYS,
2011 /** Shw: 32-bit page table; Gst: 32-bit page table. */
2012 PGMPOOLKIND_32BIT_PT_FOR_32BIT_PT,
2013 /** Shw: 32-bit page table; Gst: 4MB page. */
2014 PGMPOOLKIND_32BIT_PT_FOR_32BIT_4MB,
2015 /** Shw: PAE page table; Gst: no paging. */
2016 PGMPOOLKIND_PAE_PT_FOR_PHYS,
2017 /** Shw: PAE page table; Gst: 32-bit page table. */
2018 PGMPOOLKIND_PAE_PT_FOR_32BIT_PT,
2019 /** Shw: PAE page table; Gst: Half of a 4MB page. */
2020 PGMPOOLKIND_PAE_PT_FOR_32BIT_4MB,
2021 /** Shw: PAE page table; Gst: PAE page table. */
2022 PGMPOOLKIND_PAE_PT_FOR_PAE_PT,
2023 /** Shw: PAE page table; Gst: 2MB page. */
2024 PGMPOOLKIND_PAE_PT_FOR_PAE_2MB,
2025
2026 /** Shw: 32-bit page directory. Gst: 32-bit page directory. */
2027 PGMPOOLKIND_32BIT_PD,
2028 /** Shw: 32-bit page directory. Gst: no paging. */
2029 PGMPOOLKIND_32BIT_PD_PHYS,
2030 /** Shw: PAE page directory 0; Gst: 32-bit page directory. */
2031 PGMPOOLKIND_PAE_PD0_FOR_32BIT_PD,
2032 /** Shw: PAE page directory 1; Gst: 32-bit page directory. */
2033 PGMPOOLKIND_PAE_PD1_FOR_32BIT_PD,
2034 /** Shw: PAE page directory 2; Gst: 32-bit page directory. */
2035 PGMPOOLKIND_PAE_PD2_FOR_32BIT_PD,
2036 /** Shw: PAE page directory 3; Gst: 32-bit page directory. */
2037 PGMPOOLKIND_PAE_PD3_FOR_32BIT_PD,
2038 /** Shw: PAE page directory; Gst: PAE page directory. */
2039 PGMPOOLKIND_PAE_PD_FOR_PAE_PD,
2040 /** Shw: PAE page directory; Gst: no paging. Note: +NP. */
2041 PGMPOOLKIND_PAE_PD_PHYS,
2042
2043 /** Shw: PAE page directory pointer table (legacy, 4 entries); Gst 32 bits paging. */
2044 PGMPOOLKIND_PAE_PDPT_FOR_32BIT,
2045 /** Shw: PAE page directory pointer table (legacy, 4 entries); Gst PAE PDPT. */
2046 PGMPOOLKIND_PAE_PDPT,
2047 /** Shw: PAE page directory pointer table (legacy, 4 entries); Gst: no paging. */
2048 PGMPOOLKIND_PAE_PDPT_PHYS,
2049
2050 /** Shw: 64-bit page directory pointer table; Gst: 64-bit page directory pointer table. */
2051 PGMPOOLKIND_64BIT_PDPT_FOR_64BIT_PDPT,
2052 /** Shw: 64-bit page directory pointer table; Gst: no paging. */
2053 PGMPOOLKIND_64BIT_PDPT_FOR_PHYS,
2054 /** Shw: 64-bit page directory table; Gst: 64-bit page directory table. */
2055 PGMPOOLKIND_64BIT_PD_FOR_64BIT_PD,
2056 /** Shw: 64-bit page directory table; Gst: no paging. */
2057 PGMPOOLKIND_64BIT_PD_FOR_PHYS, /* 24 */
2058
2059 /** Shw: 64-bit PML4; Gst: 64-bit PML4. */
2060 PGMPOOLKIND_64BIT_PML4,
2061
2062 /** Shw: EPT page directory pointer table; Gst: no paging. */
2063 PGMPOOLKIND_EPT_PDPT_FOR_PHYS,
2064 /** Shw: EPT page directory table; Gst: no paging. */
2065 PGMPOOLKIND_EPT_PD_FOR_PHYS,
2066 /** Shw: EPT page table; Gst: no paging. */
2067 PGMPOOLKIND_EPT_PT_FOR_PHYS,
2068
2069 /** Shw: Root Nested paging table. */
2070 PGMPOOLKIND_ROOT_NESTED,
2071
2072 /** The last valid entry. */
2073 PGMPOOLKIND_LAST = PGMPOOLKIND_ROOT_NESTED
2074} PGMPOOLKIND;
2075
2076/**
2077 * The access attributes of the page; only applies to big pages.
2078 */
2079typedef enum
2080{
2081 PGMPOOLACCESS_DONTCARE = 0,
2082 PGMPOOLACCESS_USER_RW,
2083 PGMPOOLACCESS_USER_R,
2084 PGMPOOLACCESS_USER_RW_NX,
2085 PGMPOOLACCESS_USER_R_NX,
2086 PGMPOOLACCESS_SUPERVISOR_RW,
2087 PGMPOOLACCESS_SUPERVISOR_R,
2088 PGMPOOLACCESS_SUPERVISOR_RW_NX,
2089 PGMPOOLACCESS_SUPERVISOR_R_NX
2090} PGMPOOLACCESS;
2091
2092/**
2093 * The tracking data for a page in the pool.
2094 */
2095typedef struct PGMPOOLPAGE
2096{
2097 /** AVL node code with the (HC) physical address of this page. */
2098 AVLOHCPHYSNODECORE Core;
2099 /** Pointer to the R3 mapping of the page. */
2100 R3PTRTYPE(void *) pvPageR3;
2101 /** Pointer to the R0 mapping of the page. */
2102 R0PTRTYPE(void *) pvPageR0;
2103 /** The guest physical address. */
2104 RTGCPHYS GCPhys;
2105 /** The kind of page we're shadowing. (This is really a PGMPOOLKIND enum.) */
2106 uint8_t enmKind;
2107 /** The subkind of page we're shadowing. (This is really a PGMPOOLACCESS enum.) */
2108 uint8_t enmAccess;
2109 /** This supplements enmKind and enmAccess */
2110 bool fA20Enabled : 1;
2111
2112 /** Used to indicate that the page is zeroed. */
2113 bool fZeroed : 1;
2114 /** Used to indicate that a PT has non-global entries. */
2115 bool fSeenNonGlobal : 1;
2116 /** Used to indicate that we're monitoring writes to the guest page. */
2117 bool fMonitored : 1;
2118 /** Used to indicate that the page is in the cache (e.g. in the GCPhys hash).
2119 * (All pages are in the age list.) */
2120 bool fCached : 1;
2121 /** This is used by the R3 access handlers when invoked by an async thread.
2122 * It's a hack required because of REMR3NotifyHandlerPhysicalDeregister. */
2123 bool volatile fReusedFlushPending : 1;
2124 /** Used to mark the page as dirty (write monitoring is temporarily
2125 * off). */
2126 bool fDirty : 1;
2127 bool fPadding1 : 1;
2128 bool fPadding2;
2129
2130 /** The index of this page. */
2131 uint16_t idx;
2132 /** The next entry in the list this page currently resides in.
2133 * It's either in the free list or in the GCPhys hash. */
2134 uint16_t iNext;
2135 /** Head of the user chain. NIL_PGMPOOL_USER_INDEX if not currently in use. */
2136 uint16_t iUserHead;
2137 /** The number of present entries. */
2138 uint16_t cPresent;
2139 /** The first entry in the table which is present. */
2140 uint16_t iFirstPresent;
2141 /** The number of modifications to the monitored page. */
2142 uint16_t cModifications;
2143 /** The next modified page. NIL_PGMPOOL_IDX if tail. */
2144 uint16_t iModifiedNext;
2145 /** The previous modified page. NIL_PGMPOOL_IDX if head. */
2146 uint16_t iModifiedPrev;
2147 /** The next page sharing access handler. NIL_PGMPOOL_IDX if tail. */
2148 uint16_t iMonitoredNext;
2149 /** The previous page sharing access handler. NIL_PGMPOOL_IDX if head. */
2150 uint16_t iMonitoredPrev;
2151 /** The next page in the age list. */
2152 uint16_t iAgeNext;
2153 /** The previous page in the age list. */
2154 uint16_t iAgePrev;
2155 /** Index into PGMPOOL::aDirtyPages if fDirty is set. */
2156 uint8_t idxDirtyEntry;
2157
2158 /** @name Access handler statistics to determine whether the guest is
2159 * (re)initializing a page table.
2160 * @{ */
2161 RTGCPTR GCPtrLastAccessHandlerRip;
2162 RTGCPTR GCPtrLastAccessHandlerFault;
2163 uint64_t cLastAccessHandler;
2164 /** @} */
2165 /** Used to indicate that this page can't be flushed. Important for cr3 root pages or shadow pae pd pages. */
2166 uint32_t volatile cLocked;
2167#if GC_ARCH_BITS == 64
2168 uint32_t u32Alignment3;
2169#endif
2170# ifdef VBOX_STRICT
2171 RTGCPTR GCPtrDirtyFault;
2172# endif
2173} PGMPOOLPAGE;
2174/** Pointer to a pool page. */
2175typedef PGMPOOLPAGE *PPGMPOOLPAGE;
2176/** Pointer to a const pool page. */
2177typedef PGMPOOLPAGE const *PCPGMPOOLPAGE;
2178/** Pointer to a pool page pointer. */
2179typedef PGMPOOLPAGE **PPPGMPOOLPAGE;
2180
2181
2182/** The hash table size. */
2183# define PGMPOOL_HASH_SIZE 0x40
2184/** The hash function. */
2185# define PGMPOOL_HASH(GCPhys) ( ((GCPhys) >> PAGE_SHIFT) & (PGMPOOL_HASH_SIZE - 1) )
2186
2187
2188/**
2189 * The shadow page pool instance data.
2190 *
2191 * It's all one big allocation made at init time, except for the
2192 * pages that is. The user nodes follows immediately after the
2193 * page structures.
2194 */
2195typedef struct PGMPOOL
2196{
2197 /** The VM handle - R3 Ptr. */
2198 PVMR3 pVMR3;
2199 /** The VM handle - R0 Ptr. */
2200 R0PTRTYPE(PVMCC) pVMR0;
2201 /** The max pool size. This includes the special IDs. */
2202 uint16_t cMaxPages;
2203 /** The current pool size. */
2204 uint16_t cCurPages;
2205 /** The head of the free page list. */
2206 uint16_t iFreeHead;
2207 /* Padding. */
2208 uint16_t u16Padding;
2209 /** Head of the chain of free user nodes. */
2210 uint16_t iUserFreeHead;
2211 /** The number of user nodes we've allocated. */
2212 uint16_t cMaxUsers;
2213 /** The number of present page table entries in the entire pool. */
2214 uint32_t cPresent;
2215 /** Pointer to the array of user nodes - R3 pointer. */
2216 R3PTRTYPE(PPGMPOOLUSER) paUsersR3;
2217 /** Pointer to the array of user nodes - R0 pointer. */
2218 R0PTRTYPE(PPGMPOOLUSER) paUsersR0;
2219 /** Head of the chain of free phys ext nodes. */
2220 uint16_t iPhysExtFreeHead;
2221 /** The number of user nodes we've allocated. */
2222 uint16_t cMaxPhysExts;
2223 uint32_t u32Padding0b;
2224 /** Pointer to the array of physical xref extent nodes - R3 pointer. */
2225 R3PTRTYPE(PPGMPOOLPHYSEXT) paPhysExtsR3;
2226 /** Pointer to the array of physical xref extent nodes - R0 pointer. */
2227 R0PTRTYPE(PPGMPOOLPHYSEXT) paPhysExtsR0;
2228 /** Hash table for GCPhys addresses. */
2229 uint16_t aiHash[PGMPOOL_HASH_SIZE];
2230 /** The head of the age list. */
2231 uint16_t iAgeHead;
2232 /** The tail of the age list. */
2233 uint16_t iAgeTail;
2234 /** Set if the cache is enabled. */
2235 bool fCacheEnabled;
2236 /** Alignment padding. */
2237 bool afPadding1[3];
2238 /** Head of the list of modified pages. */
2239 uint16_t iModifiedHead;
2240 /** The current number of modified pages. */
2241 uint16_t cModifiedPages;
2242 /** Physical access handler type registration handle. */
2243 PGMPHYSHANDLERTYPE hAccessHandlerType;
2244 /** Next available slot (in aDirtyPages). */
2245 uint32_t idxFreeDirtyPage;
2246 /** Number of active dirty pages. */
2247 uint32_t cDirtyPages;
2248 /** Array of current dirty pgm pool page indices. */
2249 uint16_t aidxDirtyPages[16];
2250 /** Array running in parallel to aidxDirtyPages with the page data. */
2251 struct
2252 {
2253 uint64_t aPage[512];
2254 } aDirtyPages[16];
2255
2256 /** The number of pages currently in use. */
2257 uint16_t cUsedPages;
2258#ifdef VBOX_WITH_STATISTICS
2259 /** The high water mark for cUsedPages. */
2260 uint16_t cUsedPagesHigh;
2261 uint32_t Alignment1; /**< Align the next member on a 64-bit boundary. */
2262 /** Profiling pgmPoolAlloc(). */
2263 STAMPROFILEADV StatAlloc;
2264 /** Profiling pgmR3PoolClearDoIt(). */
2265 STAMPROFILE StatClearAll;
2266 /** Profiling pgmR3PoolReset(). */
2267 STAMPROFILE StatR3Reset;
2268 /** Profiling pgmPoolFlushPage(). */
2269 STAMPROFILE StatFlushPage;
2270 /** Profiling pgmPoolFree(). */
2271 STAMPROFILE StatFree;
2272 /** Counting explicit flushes by PGMPoolFlushPage(). */
2273 STAMCOUNTER StatForceFlushPage;
2274 /** Counting explicit flushes of dirty pages by PGMPoolFlushPage(). */
2275 STAMCOUNTER StatForceFlushDirtyPage;
2276 /** Counting flushes for reused pages. */
2277 STAMCOUNTER StatForceFlushReused;
2278 /** Profiling time spent zeroing pages. */
2279 STAMPROFILE StatZeroPage;
2280 /** Profiling of pgmPoolTrackDeref. */
2281 STAMPROFILE StatTrackDeref;
2282 /** Profiling pgmTrackFlushGCPhysPT. */
2283 STAMPROFILE StatTrackFlushGCPhysPT;
2284 /** Profiling pgmTrackFlushGCPhysPTs. */
2285 STAMPROFILE StatTrackFlushGCPhysPTs;
2286 /** Profiling pgmTrackFlushGCPhysPTsSlow. */
2287 STAMPROFILE StatTrackFlushGCPhysPTsSlow;
2288 /** Number of times we've been out of user records. */
2289 STAMCOUNTER StatTrackFreeUpOneUser;
2290 /** Nr of flushed entries. */
2291 STAMCOUNTER StatTrackFlushEntry;
2292 /** Nr of updated entries. */
2293 STAMCOUNTER StatTrackFlushEntryKeep;
2294 /** Profiling deref activity related tracking GC physical pages. */
2295 STAMPROFILE StatTrackDerefGCPhys;
2296 /** Number of linear searches for a HCPhys in the ram ranges. */
2297 STAMCOUNTER StatTrackLinearRamSearches;
2298 /** The number of failing pgmPoolTrackPhysExtAlloc calls. */
2299 STAMCOUNTER StamTrackPhysExtAllocFailures;
2300
2301 /** Profiling the RC/R0 \#PF access handler. */
2302 STAMPROFILE StatMonitorPfRZ;
2303 /** Profiling the RC/R0 access we've handled (except REP STOSD). */
2304 STAMPROFILE StatMonitorPfRZHandled;
2305 /** Times we've failed interpreting the instruction. */
2306 STAMCOUNTER StatMonitorPfRZEmulateInstr;
2307 /** Profiling the pgmPoolFlushPage calls made from the RC/R0 access handler. */
2308 STAMPROFILE StatMonitorPfRZFlushPage;
2309 /** Times we've detected a page table reinit. */
2310 STAMCOUNTER StatMonitorPfRZFlushReinit;
2311 /** Counting flushes for pages that are modified too often. */
2312 STAMCOUNTER StatMonitorPfRZFlushModOverflow;
2313 /** Times we've detected fork(). */
2314 STAMCOUNTER StatMonitorPfRZFork;
2315 /** Times we've failed interpreting a patch code instruction. */
2316 STAMCOUNTER StatMonitorPfRZIntrFailPatch1;
2317 /** Times we've failed interpreting a patch code instruction during flushing. */
2318 STAMCOUNTER StatMonitorPfRZIntrFailPatch2;
2319 /** The number of times we've seen rep prefixes we can't handle. */
2320 STAMCOUNTER StatMonitorPfRZRepPrefix;
2321 /** Profiling the REP STOSD cases we've handled. */
2322 STAMPROFILE StatMonitorPfRZRepStosd;
2323
2324 /** Profiling the R0/RC regular access handler. */
2325 STAMPROFILE StatMonitorRZ;
2326 /** Profiling the pgmPoolFlushPage calls made from the regular access handler in R0/RC. */
2327 STAMPROFILE StatMonitorRZFlushPage;
2328 /** Per access size counts indexed by size minus 1, last for larger. */
2329 STAMCOUNTER aStatMonitorRZSizes[16+3];
2330 /** Missaligned access counts indexed by offset - 1. */
2331 STAMCOUNTER aStatMonitorRZMisaligned[7];
2332
2333 /** Nr of handled PT faults. */
2334 STAMCOUNTER StatMonitorRZFaultPT;
2335 /** Nr of handled PD faults. */
2336 STAMCOUNTER StatMonitorRZFaultPD;
2337 /** Nr of handled PDPT faults. */
2338 STAMCOUNTER StatMonitorRZFaultPDPT;
2339 /** Nr of handled PML4 faults. */
2340 STAMCOUNTER StatMonitorRZFaultPML4;
2341
2342 /** Profiling the R3 access handler. */
2343 STAMPROFILE StatMonitorR3;
2344 /** Profiling the pgmPoolFlushPage calls made from the R3 access handler. */
2345 STAMPROFILE StatMonitorR3FlushPage;
2346 /** Per access size counts indexed by size minus 1, last for larger. */
2347 STAMCOUNTER aStatMonitorR3Sizes[16+3];
2348 /** Missaligned access counts indexed by offset - 1. */
2349 STAMCOUNTER aStatMonitorR3Misaligned[7];
2350 /** Nr of handled PT faults. */
2351 STAMCOUNTER StatMonitorR3FaultPT;
2352 /** Nr of handled PD faults. */
2353 STAMCOUNTER StatMonitorR3FaultPD;
2354 /** Nr of handled PDPT faults. */
2355 STAMCOUNTER StatMonitorR3FaultPDPT;
2356 /** Nr of handled PML4 faults. */
2357 STAMCOUNTER StatMonitorR3FaultPML4;
2358
2359 /** Times we've called pgmPoolResetDirtyPages (and there were dirty page). */
2360 STAMCOUNTER StatResetDirtyPages;
2361 /** Times we've called pgmPoolAddDirtyPage. */
2362 STAMCOUNTER StatDirtyPage;
2363 /** Times we've had to flush duplicates for dirty page management. */
2364 STAMCOUNTER StatDirtyPageDupFlush;
2365 /** Times we've had to flush because of overflow. */
2366 STAMCOUNTER StatDirtyPageOverFlowFlush;
2367
2368 /** The high water mark for cModifiedPages. */
2369 uint16_t cModifiedPagesHigh;
2370 uint16_t Alignment2[3]; /**< Align the next member on a 64-bit boundary. */
2371
2372 /** The number of cache hits. */
2373 STAMCOUNTER StatCacheHits;
2374 /** The number of cache misses. */
2375 STAMCOUNTER StatCacheMisses;
2376 /** The number of times we've got a conflict of 'kind' in the cache. */
2377 STAMCOUNTER StatCacheKindMismatches;
2378 /** Number of times we've been out of pages. */
2379 STAMCOUNTER StatCacheFreeUpOne;
2380 /** The number of cacheable allocations. */
2381 STAMCOUNTER StatCacheCacheable;
2382 /** The number of uncacheable allocations. */
2383 STAMCOUNTER StatCacheUncacheable;
2384#else
2385 uint32_t Alignment3; /**< Align the next member on a 64-bit boundary. */
2386#endif
2387 /** Profiling PGMR0PoolGrow(). */
2388 STAMPROFILE StatGrow;
2389 /** The AVL tree for looking up a page by its HC physical address. */
2390 AVLOHCPHYSTREE HCPhysTree;
2391 uint32_t Alignment4; /**< Align the next member on a 64-bit boundary. */
2392 /** Array of pages. (cMaxPages in length)
2393 * The Id is the index into thist array.
2394 */
2395 PGMPOOLPAGE aPages[PGMPOOL_IDX_FIRST];
2396} PGMPOOL, *PPGMPOOL, **PPPGMPOOL;
2397AssertCompileMemberAlignment(PGMPOOL, iModifiedHead, 8);
2398AssertCompileMemberAlignment(PGMPOOL, aDirtyPages, 8);
2399AssertCompileMemberAlignment(PGMPOOL, cUsedPages, 8);
2400#ifdef VBOX_WITH_STATISTICS
2401AssertCompileMemberAlignment(PGMPOOL, StatAlloc, 8);
2402#endif
2403AssertCompileMemberAlignment(PGMPOOL, aPages, 8);
2404
2405
2406/** @def PGMPOOL_PAGE_2_PTR
2407 * Maps a pool page pool into the current context.
2408 *
2409 * @returns VBox status code.
2410 * @param a_pVM Pointer to the VM.
2411 * @param a_pPage The pool page.
2412 *
2413 * @remark In RC this uses PGMGCDynMapHCPage(), so it will consume of the
2414 * small page window employeed by that function. Be careful.
2415 * @remark There is no need to assert on the result.
2416 */
2417#if defined(VBOX_STRICT) || 1 /* temporarily going strict here */
2418# define PGMPOOL_PAGE_2_PTR(a_pVM, a_pPage) pgmPoolMapPageStrict(a_pPage, __FUNCTION__)
2419DECLINLINE(void *) pgmPoolMapPageStrict(PPGMPOOLPAGE a_pPage, const char *pszCaller)
2420{
2421 RT_NOREF(pszCaller);
2422 AssertPtr(a_pPage);
2423 AssertMsg(RT_VALID_PTR(a_pPage->CTX_SUFF(pvPage)),
2424 ("enmKind=%d idx=%#x HCPhys=%RHp GCPhys=%RGp pvPageR3=%p pvPageR0=%p caller=%s\n",
2425 a_pPage->enmKind, a_pPage->idx, a_pPage->Core.Key, a_pPage->GCPhys, a_pPage->pvPageR3, a_pPage->pvPageR0, pszCaller));
2426 return a_pPage->CTX_SUFF(pvPage);
2427}
2428#else
2429# define PGMPOOL_PAGE_2_PTR(pVM, a_pPage) ((a_pPage)->CTX_SUFF(pvPage))
2430#endif
2431
2432
2433/** @def PGMPOOL_PAGE_2_PTR_V2
2434 * Maps a pool page pool into the current context, taking both VM and VMCPU.
2435 *
2436 * @returns VBox status code.
2437 * @param a_pVM Pointer to the VM.
2438 * @param a_pVCpu The current CPU.
2439 * @param a_pPage The pool page.
2440 *
2441 * @remark In RC this uses PGMGCDynMapHCPage(), so it will consume of the
2442 * small page window employeed by that function. Be careful.
2443 * @remark There is no need to assert on the result.
2444 */
2445#define PGMPOOL_PAGE_2_PTR_V2(a_pVM, a_pVCpu, a_pPage) PGMPOOL_PAGE_2_PTR((a_pVM), (a_pPage))
2446
2447
2448/** @name Per guest page tracking data.
2449 * This is currently as a 16-bit word in the PGMPAGE structure, the idea though
2450 * is to use more bits for it and split it up later on. But for now we'll play
2451 * safe and change as little as possible.
2452 *
2453 * The 16-bit word has two parts:
2454 *
2455 * The first 14-bit forms the @a idx field. It is either the index of a page in
2456 * the shadow page pool, or and index into the extent list.
2457 *
2458 * The 2 topmost bits makes up the @a cRefs field, which counts the number of
2459 * shadow page pool references to the page. If cRefs equals
2460 * PGMPOOL_CREFS_PHYSEXT, then the @a idx field is an indext into the extent
2461 * (misnomer) table and not the shadow page pool.
2462 *
2463 * See PGM_PAGE_GET_TRACKING and PGM_PAGE_SET_TRACKING for how to get and set
2464 * the 16-bit word.
2465 *
2466 * @{ */
2467/** The shift count for getting to the cRefs part. */
2468#define PGMPOOL_TD_CREFS_SHIFT 14
2469/** The mask applied after shifting the tracking data down by
2470 * PGMPOOL_TD_CREFS_SHIFT. */
2471#define PGMPOOL_TD_CREFS_MASK 0x3
2472/** The cRefs value used to indicate that the idx is the head of a
2473 * physical cross reference list. */
2474#define PGMPOOL_TD_CREFS_PHYSEXT PGMPOOL_TD_CREFS_MASK
2475/** The shift used to get idx. */
2476#define PGMPOOL_TD_IDX_SHIFT 0
2477/** The mask applied to the idx after shifting down by PGMPOOL_TD_IDX_SHIFT. */
2478#define PGMPOOL_TD_IDX_MASK 0x3fff
2479/** The idx value when we're out of of PGMPOOLPHYSEXT entries or/and there are
2480 * simply too many mappings of this page. */
2481#define PGMPOOL_TD_IDX_OVERFLOWED PGMPOOL_TD_IDX_MASK
2482
2483/** @def PGMPOOL_TD_MAKE
2484 * Makes a 16-bit tracking data word.
2485 *
2486 * @returns tracking data.
2487 * @param cRefs The @a cRefs field. Must be within bounds!
2488 * @param idx The @a idx field. Must also be within bounds! */
2489#define PGMPOOL_TD_MAKE(cRefs, idx) ( ((cRefs) << PGMPOOL_TD_CREFS_SHIFT) | (idx) )
2490
2491/** @def PGMPOOL_TD_GET_CREFS
2492 * Get the @a cRefs field from a tracking data word.
2493 *
2494 * @returns The @a cRefs field
2495 * @param u16 The tracking data word.
2496 * @remarks This will only return 1 or PGMPOOL_TD_CREFS_PHYSEXT for a
2497 * non-zero @a u16. */
2498#define PGMPOOL_TD_GET_CREFS(u16) ( ((u16) >> PGMPOOL_TD_CREFS_SHIFT) & PGMPOOL_TD_CREFS_MASK )
2499
2500/** @def PGMPOOL_TD_GET_IDX
2501 * Get the @a idx field from a tracking data word.
2502 *
2503 * @returns The @a idx field
2504 * @param u16 The tracking data word. */
2505#define PGMPOOL_TD_GET_IDX(u16) ( ((u16) >> PGMPOOL_TD_IDX_SHIFT) & PGMPOOL_TD_IDX_MASK )
2506/** @} */
2507
2508
2509
2510/** @name A20 gate macros
2511 * @{ */
2512#define PGM_WITH_A20
2513#ifdef PGM_WITH_A20
2514# define PGM_A20_IS_ENABLED(a_pVCpu) ((a_pVCpu)->pgm.s.fA20Enabled)
2515# define PGM_A20_APPLY(a_pVCpu, a_GCPhys) ((a_GCPhys) & (a_pVCpu)->pgm.s.GCPhysA20Mask)
2516# define PGM_A20_APPLY_TO_VAR(a_pVCpu, a_GCPhysVar) \
2517 do { a_GCPhysVar &= (a_pVCpu)->pgm.s.GCPhysA20Mask; } while (0)
2518# define PGM_A20_ASSERT_MASKED(pVCpu, a_GCPhys) Assert(PGM_A20_APPLY(pVCpu, a_GCPhys) == (a_GCPhys))
2519#else
2520# define PGM_A20_IS_ENABLED(a_pVCpu) (true)
2521# define PGM_A20_APPLY(a_pVCpu, a_GCPhys) (a_GCPhys)
2522# define PGM_A20_APPLY_TO_VAR(a_pVCpu, a_GCPhysVar) do { } while (0)
2523# define PGM_A20_ASSERT_MASKED(pVCpu, a_GCPhys) do { } while (0)
2524#endif
2525/** @} */
2526
2527
2528/**
2529 * Roots and anchors for trees and list employing self relative offsets as
2530 * pointers.
2531 *
2532 * When using self-relative offsets instead of pointers, the offsets needs to be
2533 * the same in all offsets. Thus the roots and anchors needs to live on the
2534 * hyper heap just like the nodes.
2535 */
2536typedef struct PGMTREES
2537{
2538 /** List of physical access handler types (offset pointers) of type
2539 * PGMPHYSHANDLERTYPEINT. This is needed for relocations. */
2540 RTLISTOFF32ANCHOR HeadPhysHandlerTypes;
2541 /** Physical access handlers (AVL range+offsetptr tree). */
2542 AVLROGCPHYSTREE PhysHandlers;
2543} PGMTREES;
2544/** Pointer to PGM trees. */
2545typedef PGMTREES *PPGMTREES;
2546
2547
2548/**
2549 * Page fault guest state for the AMD64 paging mode.
2550 */
2551typedef struct PGMPTWALKCORE
2552{
2553 /** The guest virtual address that is being resolved by the walk
2554 * (input). */
2555 RTGCPTR GCPtr;
2556
2557 /** The guest physical address that is the result of the walk.
2558 * @remarks only valid if fSucceeded is set. */
2559 RTGCPHYS GCPhys;
2560
2561 /** Set if the walk succeeded, i.d. GCPhys is valid. */
2562 bool fSucceeded;
2563 /** The level problem arrised at.
2564 * PTE is level 1, PDE is level 2, PDPE is level 3, PML4 is level 4, CR3 is
2565 * level 8. This is 0 on success. */
2566 uint8_t uLevel;
2567 /** Set if the page isn't present. */
2568 bool fNotPresent;
2569 /** Encountered a bad physical address. */
2570 bool fBadPhysAddr;
2571 /** Set if there was reserved bit violations. */
2572 bool fRsvdError;
2573 /** Set if it involves a big page (2/4 MB). */
2574 bool fBigPage;
2575 /** Set if it involves a gigantic page (1 GB). */
2576 bool fGigantPage;
2577 /** The effective X86_PTE_US flag for the address. */
2578 bool fEffectiveUS;
2579 /** The effective X86_PTE_RW flag for the address. */
2580 bool fEffectiveRW;
2581 /** The effective X86_PTE_NX flag for the address. */
2582 bool fEffectiveNX;
2583 bool afPadding1[2];
2584 /** Effective flags thus far: RW, US, PWT, PCD, A, ~NX >> 63.
2585 * The NX bit is inverted and shifted down 63 places to bit 0. */
2586 uint32_t fEffective;
2587} PGMPTWALKCORE;
2588
2589/** @name PGMPTWALKCORE::fEffective bits.
2590 * @{ */
2591/** Effective execute bit (!NX). */
2592#define PGMPTWALK_EFF_X UINT32_C(1)
2593/** Effective write access bit. */
2594#define PGMPTWALK_EFF_RW X86_PTE_RW
2595/** Effective user-mode access bit. */
2596#define PGMPTWALK_EFF_US X86_PTE_US
2597/** Effective write through cache bit. */
2598#define PGMPTWALK_EFF_PWT X86_PTE_PWT
2599/** Effective cache disabled bit. */
2600#define PGMPTWALK_EFF_PCD X86_PTE_PCD
2601/** Effective accessed bit. */
2602#define PGMPTWALK_EFF_A X86_PTE_A
2603/** The dirty bit of the final entry. */
2604#define PGMPTWALK_EFF_D X86_PTE_D
2605/** The PAT bit of the final entry. */
2606#define PGMPTWALK_EFF_PAT X86_PTE_PAT
2607/** The global bit of the final entry. */
2608#define PGMPTWALK_EFF_G X86_PTE_G
2609/** @} */
2610
2611
2612/**
2613 * Guest page table walk for the AMD64 mode.
2614 */
2615typedef struct PGMPTWALKGSTAMD64
2616{
2617 /** The common core. */
2618 PGMPTWALKCORE Core;
2619
2620 PX86PML4 pPml4;
2621 PX86PML4E pPml4e;
2622 X86PML4E Pml4e;
2623
2624 PX86PDPT pPdpt;
2625 PX86PDPE pPdpe;
2626 X86PDPE Pdpe;
2627
2628 PX86PDPAE pPd;
2629 PX86PDEPAE pPde;
2630 X86PDEPAE Pde;
2631
2632 PX86PTPAE pPt;
2633 PX86PTEPAE pPte;
2634 X86PTEPAE Pte;
2635} PGMPTWALKGSTAMD64;
2636/** Pointer to a AMD64 guest page table walk. */
2637typedef PGMPTWALKGSTAMD64 *PPGMPTWALKGSTAMD64;
2638/** Pointer to a const AMD64 guest page table walk. */
2639typedef PGMPTWALKGSTAMD64 const *PCPGMPTWALKGSTAMD64;
2640
2641/**
2642 * Guest page table walk for the PAE mode.
2643 */
2644typedef struct PGMPTWALKGSTPAE
2645{
2646 /** The common core. */
2647 PGMPTWALKCORE Core;
2648
2649 PX86PDPT pPdpt;
2650 PX86PDPE pPdpe;
2651 X86PDPE Pdpe;
2652
2653 PX86PDPAE pPd;
2654 PX86PDEPAE pPde;
2655 X86PDEPAE Pde;
2656
2657 PX86PTPAE pPt;
2658 PX86PTEPAE pPte;
2659 X86PTEPAE Pte;
2660} PGMPTWALKGSTPAE;
2661/** Pointer to a PAE guest page table walk. */
2662typedef PGMPTWALKGSTPAE *PPGMPTWALKGSTPAE;
2663/** Pointer to a const AMD64 guest page table walk. */
2664typedef PGMPTWALKGSTPAE const *PCPGMPTWALKGSTPAE;
2665
2666/**
2667 * Guest page table walk for the 32-bit mode.
2668 */
2669typedef struct PGMPTWALKGST32BIT
2670{
2671 /** The common core. */
2672 PGMPTWALKCORE Core;
2673
2674 PX86PD pPd;
2675 PX86PDE pPde;
2676 X86PDE Pde;
2677
2678 PX86PT pPt;
2679 PX86PTE pPte;
2680 X86PTE Pte;
2681} PGMPTWALKGST32BIT;
2682/** Pointer to a 32-bit guest page table walk. */
2683typedef PGMPTWALKGST32BIT *PPGMPTWALKGST32BIT;
2684/** Pointer to a const 32-bit guest page table walk. */
2685typedef PGMPTWALKGST32BIT const *PCPGMPTWALKGST32BIT;
2686
2687/**
2688 * Which part of PGMPTWALKGST that is valid.
2689 */
2690typedef enum PGMPTWALKGSTTYPE
2691{
2692 /** Customary invalid 0 value. */
2693 PGMPTWALKGSTTYPE_INVALID = 0,
2694 /** PGMPTWALKGST::u.Amd64 is valid. */
2695 PGMPTWALKGSTTYPE_AMD64,
2696 /** PGMPTWALKGST::u.Pae is valid. */
2697 PGMPTWALKGSTTYPE_PAE,
2698 /** PGMPTWALKGST::u.Legacy is valid. */
2699 PGMPTWALKGSTTYPE_32BIT,
2700 /** Customary 32-bit type hack. */
2701 PGMPTWALKGSTTYPE_32BIT_HACK = 0x7fff0000
2702} PGMPTWALKGSTTYPE;
2703
2704/**
2705 * Combined guest page table walk result.
2706 */
2707typedef struct PGMPTWALKGST
2708{
2709 union
2710 {
2711 /** The page walker core - always valid. */
2712 PGMPTWALKCORE Core;
2713 /** The page walker for AMD64. */
2714 PGMPTWALKGSTAMD64 Amd64;
2715 /** The page walker for PAE (32-bit). */
2716 PGMPTWALKGSTPAE Pae;
2717 /** The page walker for 32-bit paging (called legacy due to C naming
2718 * convension). */
2719 PGMPTWALKGST32BIT Legacy;
2720 } u;
2721 /** Indicates which part of the union is valid. */
2722 PGMPTWALKGSTTYPE enmType;
2723} PGMPTWALKGST;
2724/** Pointer to a combined guest page table walk result. */
2725typedef PGMPTWALKGST *PPGMPTWALKGST;
2726/** Pointer to a read-only combined guest page table walk result. */
2727typedef PGMPTWALKGST const *PCPGMPTWALKGST;
2728
2729
2730/** @name Paging mode macros
2731 * @{
2732 */
2733#ifdef IN_RING3
2734# define PGM_CTX(a,b) a##R3##b
2735# define PGM_CTX_STR(a,b) a "R3" b
2736# define PGM_CTX_DECL(type) DECLCALLBACK(type)
2737#elif defined(IN_RING0)
2738# define PGM_CTX(a,b) a##R0##b
2739# define PGM_CTX_STR(a,b) a "R0" b
2740# define PGM_CTX_DECL(type) VMMDECL(type)
2741#else
2742# error "Not IN_RING3 or IN_RING0!"
2743#endif
2744
2745#define PGM_GST_NAME_REAL(name) PGM_CTX(pgm,GstReal##name)
2746#define PGM_GST_NAME_RC_REAL_STR(name) "pgmRCGstReal" #name
2747#define PGM_GST_NAME_R0_REAL_STR(name) "pgmR0GstReal" #name
2748#define PGM_GST_NAME_PROT(name) PGM_CTX(pgm,GstProt##name)
2749#define PGM_GST_NAME_RC_PROT_STR(name) "pgmRCGstProt" #name
2750#define PGM_GST_NAME_R0_PROT_STR(name) "pgmR0GstProt" #name
2751#define PGM_GST_NAME_32BIT(name) PGM_CTX(pgm,Gst32Bit##name)
2752#define PGM_GST_NAME_RC_32BIT_STR(name) "pgmRCGst32Bit" #name
2753#define PGM_GST_NAME_R0_32BIT_STR(name) "pgmR0Gst32Bit" #name
2754#define PGM_GST_NAME_PAE(name) PGM_CTX(pgm,GstPAE##name)
2755#define PGM_GST_NAME_RC_PAE_STR(name) "pgmRCGstPAE" #name
2756#define PGM_GST_NAME_R0_PAE_STR(name) "pgmR0GstPAE" #name
2757#define PGM_GST_NAME_AMD64(name) PGM_CTX(pgm,GstAMD64##name)
2758#define PGM_GST_NAME_RC_AMD64_STR(name) "pgmRCGstAMD64" #name
2759#define PGM_GST_NAME_R0_AMD64_STR(name) "pgmR0GstAMD64" #name
2760#define PGM_GST_DECL(type, name) PGM_CTX_DECL(type) PGM_GST_NAME(name)
2761
2762#define PGM_SHW_NAME_32BIT(name) PGM_CTX(pgm,Shw32Bit##name)
2763#define PGM_SHW_NAME_RC_32BIT_STR(name) "pgmRCShw32Bit" #name
2764#define PGM_SHW_NAME_R0_32BIT_STR(name) "pgmR0Shw32Bit" #name
2765#define PGM_SHW_NAME_PAE(name) PGM_CTX(pgm,ShwPAE##name)
2766#define PGM_SHW_NAME_RC_PAE_STR(name) "pgmRCShwPAE" #name
2767#define PGM_SHW_NAME_R0_PAE_STR(name) "pgmR0ShwPAE" #name
2768#define PGM_SHW_NAME_AMD64(name) PGM_CTX(pgm,ShwAMD64##name)
2769#define PGM_SHW_NAME_RC_AMD64_STR(name) "pgmRCShwAMD64" #name
2770#define PGM_SHW_NAME_R0_AMD64_STR(name) "pgmR0ShwAMD64" #name
2771#define PGM_SHW_NAME_NESTED_32BIT(name) PGM_CTX(pgm,ShwNested32Bit##name)
2772#define PGM_SHW_NAME_RC_NESTED_32BIT_STR(name) "pgmRCShwNested32Bit" #name
2773#define PGM_SHW_NAME_R0_NESTED_32BIT_STR(name) "pgmR0ShwNested32Bit" #name
2774#define PGM_SHW_NAME_NESTED_PAE(name) PGM_CTX(pgm,ShwNestedPAE##name)
2775#define PGM_SHW_NAME_RC_NESTED_PAE_STR(name) "pgmRCShwNestedPAE" #name
2776#define PGM_SHW_NAME_R0_NESTED_PAE_STR(name) "pgmR0ShwNestedPAE" #name
2777#define PGM_SHW_NAME_NESTED_AMD64(name) PGM_CTX(pgm,ShwNestedAMD64##name)
2778#define PGM_SHW_NAME_RC_NESTED_AMD64_STR(name) "pgmRCShwNestedAMD64" #name
2779#define PGM_SHW_NAME_R0_NESTED_AMD64_STR(name) "pgmR0ShwNestedAMD64" #name
2780#define PGM_SHW_NAME_EPT(name) PGM_CTX(pgm,ShwEPT##name)
2781#define PGM_SHW_NAME_RC_EPT_STR(name) "pgmRCShwEPT" #name
2782#define PGM_SHW_NAME_R0_EPT_STR(name) "pgmR0ShwEPT" #name
2783#define PGM_SHW_NAME_NONE(name) PGM_CTX(pgm,ShwNone##name)
2784#define PGM_SHW_NAME_RC_NONE_STR(name) "pgmRCShwNone" #name
2785#define PGM_SHW_NAME_R0_NONE_STR(name) "pgmR0ShwNone" #name
2786#define PGM_SHW_DECL(type, name) PGM_CTX_DECL(type) PGM_SHW_NAME(name)
2787
2788/* Shw_Gst */
2789#define PGM_BTH_NAME_32BIT_REAL(name) PGM_CTX(pgm,Bth32BitReal##name)
2790#define PGM_BTH_NAME_32BIT_PROT(name) PGM_CTX(pgm,Bth32BitProt##name)
2791#define PGM_BTH_NAME_32BIT_32BIT(name) PGM_CTX(pgm,Bth32Bit32Bit##name)
2792#define PGM_BTH_NAME_PAE_REAL(name) PGM_CTX(pgm,BthPAEReal##name)
2793#define PGM_BTH_NAME_PAE_PROT(name) PGM_CTX(pgm,BthPAEProt##name)
2794#define PGM_BTH_NAME_PAE_32BIT(name) PGM_CTX(pgm,BthPAE32Bit##name)
2795#define PGM_BTH_NAME_PAE_PAE(name) PGM_CTX(pgm,BthPAEPAE##name)
2796#define PGM_BTH_NAME_AMD64_PROT(name) PGM_CTX(pgm,BthAMD64Prot##name)
2797#define PGM_BTH_NAME_AMD64_AMD64(name) PGM_CTX(pgm,BthAMD64AMD64##name)
2798#define PGM_BTH_NAME_NESTED_32BIT_REAL(name) PGM_CTX(pgm,BthNested32BitReal##name)
2799#define PGM_BTH_NAME_NESTED_32BIT_PROT(name) PGM_CTX(pgm,BthNested32BitProt##name)
2800#define PGM_BTH_NAME_NESTED_32BIT_32BIT(name) PGM_CTX(pgm,BthNested32Bit32Bit##name)
2801#define PGM_BTH_NAME_NESTED_32BIT_PAE(name) PGM_CTX(pgm,BthNested32BitPAE##name)
2802#define PGM_BTH_NAME_NESTED_32BIT_AMD64(name) PGM_CTX(pgm,BthNested32BitAMD64##name)
2803#define PGM_BTH_NAME_NESTED_PAE_REAL(name) PGM_CTX(pgm,BthNestedPAEReal##name)
2804#define PGM_BTH_NAME_NESTED_PAE_PROT(name) PGM_CTX(pgm,BthNestedPAEProt##name)
2805#define PGM_BTH_NAME_NESTED_PAE_32BIT(name) PGM_CTX(pgm,BthNestedPAE32Bit##name)
2806#define PGM_BTH_NAME_NESTED_PAE_PAE(name) PGM_CTX(pgm,BthNestedPAEPAE##name)
2807#define PGM_BTH_NAME_NESTED_PAE_AMD64(name) PGM_CTX(pgm,BthNestedPAEAMD64##name)
2808#define PGM_BTH_NAME_NESTED_AMD64_REAL(name) PGM_CTX(pgm,BthNestedAMD64Real##name)
2809#define PGM_BTH_NAME_NESTED_AMD64_PROT(name) PGM_CTX(pgm,BthNestedAMD64Prot##name)
2810#define PGM_BTH_NAME_NESTED_AMD64_32BIT(name) PGM_CTX(pgm,BthNestedAMD6432Bit##name)
2811#define PGM_BTH_NAME_NESTED_AMD64_PAE(name) PGM_CTX(pgm,BthNestedAMD64PAE##name)
2812#define PGM_BTH_NAME_NESTED_AMD64_AMD64(name) PGM_CTX(pgm,BthNestedAMD64AMD64##name)
2813#define PGM_BTH_NAME_EPT_REAL(name) PGM_CTX(pgm,BthEPTReal##name)
2814#define PGM_BTH_NAME_EPT_PROT(name) PGM_CTX(pgm,BthEPTProt##name)
2815#define PGM_BTH_NAME_EPT_32BIT(name) PGM_CTX(pgm,BthEPT32Bit##name)
2816#define PGM_BTH_NAME_EPT_PAE(name) PGM_CTX(pgm,BthEPTPAE##name)
2817#define PGM_BTH_NAME_EPT_AMD64(name) PGM_CTX(pgm,BthEPTAMD64##name)
2818#define PGM_BTH_NAME_NONE_REAL(name) PGM_CTX(pgm,BthNoneReal##name)
2819#define PGM_BTH_NAME_NONE_PROT(name) PGM_CTX(pgm,BthNoneProt##name)
2820#define PGM_BTH_NAME_NONE_32BIT(name) PGM_CTX(pgm,BthNone32Bit##name)
2821#define PGM_BTH_NAME_NONE_PAE(name) PGM_CTX(pgm,BthNonePAE##name)
2822#define PGM_BTH_NAME_NONE_AMD64(name) PGM_CTX(pgm,BthNoneAMD64##name)
2823
2824#define PGM_BTH_NAME_RC_32BIT_REAL_STR(name) "pgmRCBth32BitReal" #name
2825#define PGM_BTH_NAME_RC_32BIT_PROT_STR(name) "pgmRCBth32BitProt" #name
2826#define PGM_BTH_NAME_RC_32BIT_32BIT_STR(name) "pgmRCBth32Bit32Bit" #name
2827#define PGM_BTH_NAME_RC_PAE_REAL_STR(name) "pgmRCBthPAEReal" #name
2828#define PGM_BTH_NAME_RC_PAE_PROT_STR(name) "pgmRCBthPAEProt" #name
2829#define PGM_BTH_NAME_RC_PAE_32BIT_STR(name) "pgmRCBthPAE32Bit" #name
2830#define PGM_BTH_NAME_RC_PAE_PAE_STR(name) "pgmRCBthPAEPAE" #name
2831#define PGM_BTH_NAME_RC_AMD64_AMD64_STR(name) "pgmRCBthAMD64AMD64" #name
2832#define PGM_BTH_NAME_RC_NESTED_32BIT_REAL_STR(name) "pgmRCBthNested32BitReal" #name
2833#define PGM_BTH_NAME_RC_NESTED_32BIT_PROT_STR(name) "pgmRCBthNested32BitProt" #name
2834#define PGM_BTH_NAME_RC_NESTED_32BIT_32BIT_STR(name) "pgmRCBthNested32Bit32Bit" #name
2835#define PGM_BTH_NAME_RC_NESTED_32BIT_PAE_STR(name) "pgmRCBthNested32BitPAE" #name
2836#define PGM_BTH_NAME_RC_NESTED_32BIT_AMD64_STR(name) "pgmRCBthNested32BitAMD64" #name
2837#define PGM_BTH_NAME_RC_NESTED_PAE_REAL_STR(name) "pgmRCBthNestedPAEReal" #name
2838#define PGM_BTH_NAME_RC_NESTED_PAE_PROT_STR(name) "pgmRCBthNestedPAEProt" #name
2839#define PGM_BTH_NAME_RC_NESTED_PAE_32BIT_STR(name) "pgmRCBthNestedPAE32Bit" #name
2840#define PGM_BTH_NAME_RC_NESTED_PAE_PAE_STR(name) "pgmRCBthNestedPAEPAE" #name
2841#define PGM_BTH_NAME_RC_NESTED_PAE_AMD64_STR(name) "pgmRCBthNestedPAEAMD64" #name
2842#define PGM_BTH_NAME_RC_NESTED_AMD64_REAL_STR(name) "pgmRCBthNestedAMD64Real" #name
2843#define PGM_BTH_NAME_RC_NESTED_AMD64_PROT_STR(name) "pgmRCBthNestedAMD64Prot" #name
2844#define PGM_BTH_NAME_RC_NESTED_AMD64_32BIT_STR(name) "pgmRCBthNestedAMD6432Bit" #name
2845#define PGM_BTH_NAME_RC_NESTED_AMD64_PAE_STR(name) "pgmRCBthNestedAMD64PAE" #name
2846#define PGM_BTH_NAME_RC_NESTED_AMD64_AMD64_STR(name) "pgmRCBthNestedAMD64AMD64" #name
2847#define PGM_BTH_NAME_RC_EPT_REAL_STR(name) "pgmRCBthEPTReal" #name
2848#define PGM_BTH_NAME_RC_EPT_PROT_STR(name) "pgmRCBthEPTProt" #name
2849#define PGM_BTH_NAME_RC_EPT_32BIT_STR(name) "pgmRCBthEPT32Bit" #name
2850#define PGM_BTH_NAME_RC_EPT_PAE_STR(name) "pgmRCBthEPTPAE" #name
2851#define PGM_BTH_NAME_RC_EPT_AMD64_STR(name) "pgmRCBthEPTAMD64" #name
2852
2853#define PGM_BTH_NAME_R0_32BIT_REAL_STR(name) "pgmR0Bth32BitReal" #name
2854#define PGM_BTH_NAME_R0_32BIT_PROT_STR(name) "pgmR0Bth32BitProt" #name
2855#define PGM_BTH_NAME_R0_32BIT_32BIT_STR(name) "pgmR0Bth32Bit32Bit" #name
2856#define PGM_BTH_NAME_R0_PAE_REAL_STR(name) "pgmR0BthPAEReal" #name
2857#define PGM_BTH_NAME_R0_PAE_PROT_STR(name) "pgmR0BthPAEProt" #name
2858#define PGM_BTH_NAME_R0_PAE_32BIT_STR(name) "pgmR0BthPAE32Bit" #name
2859#define PGM_BTH_NAME_R0_PAE_PAE_STR(name) "pgmR0BthPAEPAE" #name
2860#define PGM_BTH_NAME_R0_AMD64_PROT_STR(name) "pgmR0BthAMD64Prot" #name
2861#define PGM_BTH_NAME_R0_AMD64_AMD64_STR(name) "pgmR0BthAMD64AMD64" #name
2862#define PGM_BTH_NAME_R0_NESTED_32BIT_REAL_STR(name) "pgmR0BthNested32BitReal" #name
2863#define PGM_BTH_NAME_R0_NESTED_32BIT_PROT_STR(name) "pgmR0BthNested32BitProt" #name
2864#define PGM_BTH_NAME_R0_NESTED_32BIT_32BIT_STR(name) "pgmR0BthNested32Bit32Bit" #name
2865#define PGM_BTH_NAME_R0_NESTED_32BIT_PAE_STR(name) "pgmR0BthNested32BitPAE" #name
2866#define PGM_BTH_NAME_R0_NESTED_32BIT_AMD64_STR(name) "pgmR0BthNested32BitAMD64" #name
2867#define PGM_BTH_NAME_R0_NESTED_PAE_REAL_STR(name) "pgmR0BthNestedPAEReal" #name
2868#define PGM_BTH_NAME_R0_NESTED_PAE_PROT_STR(name) "pgmR0BthNestedPAEProt" #name
2869#define PGM_BTH_NAME_R0_NESTED_PAE_32BIT_STR(name) "pgmR0BthNestedPAE32Bit" #name
2870#define PGM_BTH_NAME_R0_NESTED_PAE_PAE_STR(name) "pgmR0BthNestedPAEPAE" #name
2871#define PGM_BTH_NAME_R0_NESTED_PAE_AMD64_STR(name) "pgmR0BthNestedPAEAMD64" #name
2872#define PGM_BTH_NAME_R0_NESTED_AMD64_REAL_STR(name) "pgmR0BthNestedAMD64Real" #name
2873#define PGM_BTH_NAME_R0_NESTED_AMD64_PROT_STR(name) "pgmR0BthNestedAMD64Prot" #name
2874#define PGM_BTH_NAME_R0_NESTED_AMD64_32BIT_STR(name) "pgmR0BthNestedAMD6432Bit" #name
2875#define PGM_BTH_NAME_R0_NESTED_AMD64_PAE_STR(name) "pgmR0BthNestedAMD64PAE" #name
2876#define PGM_BTH_NAME_R0_NESTED_AMD64_AMD64_STR(name) "pgmR0BthNestedAMD64AMD64" #name
2877#define PGM_BTH_NAME_R0_EPT_REAL_STR(name) "pgmR0BthEPTReal" #name
2878#define PGM_BTH_NAME_R0_EPT_PROT_STR(name) "pgmR0BthEPTProt" #name
2879#define PGM_BTH_NAME_R0_EPT_32BIT_STR(name) "pgmR0BthEPT32Bit" #name
2880#define PGM_BTH_NAME_R0_EPT_PAE_STR(name) "pgmR0BthEPTPAE" #name
2881#define PGM_BTH_NAME_R0_EPT_AMD64_STR(name) "pgmR0BthEPTAMD64" #name
2882
2883#define PGM_BTH_DECL(type, name) PGM_CTX_DECL(type) PGM_BTH_NAME(name)
2884/** @} */
2885
2886
2887/**
2888 * Function pointers for guest paging.
2889 */
2890typedef struct PGMMODEDATAGST
2891{
2892 /** The guest mode type. */
2893 uint32_t uType;
2894 DECLCALLBACKMEMBER(int, pfnGetPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, uint64_t *pfFlags, PRTGCPHYS pGCPhys));
2895 DECLCALLBACKMEMBER(int, pfnModifyPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, size_t cbPages, uint64_t fFlags, uint64_t fMask));
2896 DECLCALLBACKMEMBER(int, pfnGetPDE,(PVMCPUCC pVCpu, RTGCPTR GCPtr, PX86PDEPAE pPde));
2897 DECLCALLBACKMEMBER(int, pfnEnter,(PVMCPUCC pVCpu, RTGCPHYS GCPhysCR3));
2898 DECLCALLBACKMEMBER(int, pfnExit,(PVMCPUCC pVCpu));
2899#ifdef IN_RING3
2900 DECLCALLBACKMEMBER(int, pfnRelocate,(PVMCPUCC pVCpu, RTGCPTR offDelta)); /**< Only in ring-3. */
2901#endif
2902} PGMMODEDATAGST;
2903
2904/** The length of g_aPgmGuestModeData. */
2905#ifdef VBOX_WITH_64_BITS_GUESTS
2906# define PGM_GUEST_MODE_DATA_ARRAY_SIZE (PGM_TYPE_AMD64 + 1)
2907#else
2908# define PGM_GUEST_MODE_DATA_ARRAY_SIZE (PGM_TYPE_PAE + 1)
2909#endif
2910/** The guest mode data array. */
2911extern PGMMODEDATAGST const g_aPgmGuestModeData[PGM_GUEST_MODE_DATA_ARRAY_SIZE];
2912
2913
2914/**
2915 * Function pointers for shadow paging.
2916 */
2917typedef struct PGMMODEDATASHW
2918{
2919 /** The shadow mode type. */
2920 uint32_t uType;
2921 DECLCALLBACKMEMBER(int, pfnGetPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, uint64_t *pfFlags, PRTHCPHYS pHCPhys));
2922 DECLCALLBACKMEMBER(int, pfnModifyPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, size_t cbPages, uint64_t fFlags,
2923 uint64_t fMask, uint32_t fOpFlags));
2924 DECLCALLBACKMEMBER(int, pfnEnter,(PVMCPUCC pVCpu, bool fIs64BitsPagingMode));
2925 DECLCALLBACKMEMBER(int, pfnExit,(PVMCPUCC pVCpu));
2926#ifdef IN_RING3
2927 DECLCALLBACKMEMBER(int, pfnRelocate,(PVMCPUCC pVCpu, RTGCPTR offDelta)); /**< Only in ring-3. */
2928#endif
2929} PGMMODEDATASHW;
2930
2931/** The length of g_aPgmShadowModeData. */
2932#define PGM_SHADOW_MODE_DATA_ARRAY_SIZE PGM_TYPE_END
2933/** The shadow mode data array. */
2934extern PGMMODEDATASHW const g_aPgmShadowModeData[PGM_SHADOW_MODE_DATA_ARRAY_SIZE];
2935
2936
2937/**
2938 * Function pointers for guest+shadow paging.
2939 */
2940typedef struct PGMMODEDATABTH
2941{
2942 /** The shadow mode type. */
2943 uint32_t uShwType;
2944 /** The guest mode type. */
2945 uint32_t uGstType;
2946
2947 DECLCALLBACKMEMBER(int, pfnInvalidatePage,(PVMCPUCC pVCpu, RTGCPTR GCPtrPage));
2948 DECLCALLBACKMEMBER(int, pfnSyncCR3,(PVMCPUCC pVCpu, uint64_t cr0, uint64_t cr3, uint64_t cr4, bool fGlobal));
2949 DECLCALLBACKMEMBER(int, pfnPrefetchPage,(PVMCPUCC pVCpu, RTGCPTR GCPtrPage));
2950 DECLCALLBACKMEMBER(int, pfnVerifyAccessSyncPage,(PVMCPUCC pVCpu, RTGCPTR GCPtrPage, unsigned fFlags, unsigned uError));
2951 DECLCALLBACKMEMBER(int, pfnMapCR3,(PVMCPUCC pVCpu, RTGCPHYS GCPhysCR3));
2952 DECLCALLBACKMEMBER(int, pfnUnmapCR3,(PVMCPUCC pVCpu));
2953 DECLCALLBACKMEMBER(int, pfnEnter,(PVMCPUCC pVCpu, RTGCPHYS GCPhysCR3));
2954#ifndef IN_RING3
2955 DECLCALLBACKMEMBER(int, pfnTrap0eHandler,(PVMCPUCC pVCpu, RTGCUINT uErr, PCPUMCTXCORE pRegFrame, RTGCPTR pvFault, bool *pfLockTaken));
2956#endif
2957#ifdef VBOX_STRICT
2958 DECLCALLBACKMEMBER(unsigned, pfnAssertCR3,(PVMCPUCC pVCpu, uint64_t cr3, uint64_t cr4, RTGCPTR GCPtr, RTGCPTR cb));
2959#endif
2960} PGMMODEDATABTH;
2961
2962/** The length of g_aPgmBothModeData. */
2963#define PGM_BOTH_MODE_DATA_ARRAY_SIZE ((PGM_TYPE_END - PGM_TYPE_FIRST_SHADOW) * PGM_TYPE_END)
2964/** The guest+shadow mode data array. */
2965extern PGMMODEDATABTH const g_aPgmBothModeData[PGM_BOTH_MODE_DATA_ARRAY_SIZE];
2966
2967
2968#ifdef VBOX_WITH_STATISTICS
2969/**
2970 * PGM statistics.
2971 *
2972 * These lives on the heap when compiled in as they would otherwise waste
2973 * unnecessary space in release builds.
2974 */
2975typedef struct PGMSTATS
2976{
2977 /* R3 only: */
2978 STAMCOUNTER StatR3DetectedConflicts; /**< R3: Number of times PGMR3MapHasConflicts() detected a conflict. */
2979 STAMPROFILE StatR3ResolveConflict; /**< R3: pgmR3SyncPTResolveConflict() profiling (includes the entire relocation). */
2980
2981 /* R3+RZ */
2982 STAMCOUNTER StatRZChunkR3MapTlbHits; /**< RC/R0: Ring-3/0 chunk mapper TLB hits. */
2983 STAMCOUNTER StatRZChunkR3MapTlbMisses; /**< RC/R0: Ring-3/0 chunk mapper TLB misses. */
2984 STAMCOUNTER StatRZPageMapTlbHits; /**< RC/R0: Ring-3/0 page mapper TLB hits. */
2985 STAMCOUNTER StatRZPageMapTlbMisses; /**< RC/R0: Ring-3/0 page mapper TLB misses. */
2986 STAMCOUNTER StatPageMapTlbFlushes; /**< ALL: Ring-3/0 page mapper TLB flushes. */
2987 STAMCOUNTER StatPageMapTlbFlushEntry; /**< ALL: Ring-3/0 page mapper TLB flushes. */
2988 STAMCOUNTER StatR3ChunkR3MapTlbHits; /**< R3: Ring-3/0 chunk mapper TLB hits. */
2989 STAMCOUNTER StatR3ChunkR3MapTlbMisses; /**< R3: Ring-3/0 chunk mapper TLB misses. */
2990 STAMCOUNTER StatR3PageMapTlbHits; /**< R3: Ring-3/0 page mapper TLB hits. */
2991 STAMCOUNTER StatR3PageMapTlbMisses; /**< R3: Ring-3/0 page mapper TLB misses. */
2992 STAMCOUNTER StatRZRamRangeTlbHits; /**< RC/R0: RAM range TLB hits. */
2993 STAMCOUNTER StatRZRamRangeTlbMisses; /**< RC/R0: RAM range TLB misses. */
2994 STAMCOUNTER StatR3RamRangeTlbHits; /**< R3: RAM range TLB hits. */
2995 STAMCOUNTER StatR3RamRangeTlbMisses; /**< R3: RAM range TLB misses. */
2996 STAMCOUNTER StatR3PhysHandlerReset; /**< R3: The number of times PGMHandlerPhysicalReset is called. */
2997 STAMCOUNTER StatRZPhysHandlerReset; /**< RC/R0: The number of times PGMHandlerPhysicalReset is called. */
2998 STAMCOUNTER StatR3PhysHandlerLookupHits; /**< R3: Number of cache hits when looking up physical handlers. */
2999 STAMCOUNTER StatR3PhysHandlerLookupMisses; /**< R3: Number of cache misses when looking up physical handlers. */
3000 STAMCOUNTER StatRZPhysHandlerLookupHits; /**< RC/R0: Number of cache hits when lookup up physical handlers. */
3001 STAMCOUNTER StatRZPhysHandlerLookupMisses; /**< RC/R0: Number of cache misses when looking up physical handlers */
3002 STAMCOUNTER StatRZPageReplaceShared; /**< RC/R0: Times a shared page has been replaced by a private one. */
3003 STAMCOUNTER StatRZPageReplaceZero; /**< RC/R0: Times the zero page has been replaced by a private one. */
3004/// @todo STAMCOUNTER StatRZPageHandyAllocs; /**< RC/R0: The number of times we've executed GMMR3AllocateHandyPages. */
3005 STAMCOUNTER StatR3PageReplaceShared; /**< R3: Times a shared page has been replaced by a private one. */
3006 STAMCOUNTER StatR3PageReplaceZero; /**< R3: Times the zero page has been replaced by a private one. */
3007/// @todo STAMCOUNTER StatR3PageHandyAllocs; /**< R3: The number of times we've executed GMMR3AllocateHandyPages. */
3008
3009 /* RC only: */
3010 STAMCOUNTER StatRCInvlPgConflict; /**< RC: Number of times PGMInvalidatePage() detected a mapping conflict. */
3011 STAMCOUNTER StatRCInvlPgSyncMonCR3; /**< RC: Number of times PGMInvalidatePage() ran into PGM_SYNC_MONITOR_CR3. */
3012
3013 STAMCOUNTER StatRZPhysRead;
3014 STAMCOUNTER StatRZPhysReadBytes;
3015 STAMCOUNTER StatRZPhysWrite;
3016 STAMCOUNTER StatRZPhysWriteBytes;
3017 STAMCOUNTER StatR3PhysRead;
3018 STAMCOUNTER StatR3PhysReadBytes;
3019 STAMCOUNTER StatR3PhysWrite;
3020 STAMCOUNTER StatR3PhysWriteBytes;
3021 STAMCOUNTER StatRCPhysRead;
3022 STAMCOUNTER StatRCPhysReadBytes;
3023 STAMCOUNTER StatRCPhysWrite;
3024 STAMCOUNTER StatRCPhysWriteBytes;
3025
3026 STAMCOUNTER StatRZPhysSimpleRead;
3027 STAMCOUNTER StatRZPhysSimpleReadBytes;
3028 STAMCOUNTER StatRZPhysSimpleWrite;
3029 STAMCOUNTER StatRZPhysSimpleWriteBytes;
3030 STAMCOUNTER StatR3PhysSimpleRead;
3031 STAMCOUNTER StatR3PhysSimpleReadBytes;
3032 STAMCOUNTER StatR3PhysSimpleWrite;
3033 STAMCOUNTER StatR3PhysSimpleWriteBytes;
3034 STAMCOUNTER StatRCPhysSimpleRead;
3035 STAMCOUNTER StatRCPhysSimpleReadBytes;
3036 STAMCOUNTER StatRCPhysSimpleWrite;
3037 STAMCOUNTER StatRCPhysSimpleWriteBytes;
3038
3039 STAMCOUNTER StatTrackVirgin; /**< The number of first time shadowings. */
3040 STAMCOUNTER StatTrackAliased; /**< The number of times switching to cRef2, i.e. the page is being shadowed by two PTs. */
3041 STAMCOUNTER StatTrackAliasedMany; /**< The number of times we're tracking using cRef2. */
3042 STAMCOUNTER StatTrackAliasedLots; /**< The number of times we're hitting pages which has overflowed cRef2. */
3043 STAMCOUNTER StatTrackNoExtentsLeft; /**< The number of times the extent list was exhausted. */
3044 STAMCOUNTER StatTrackOverflows; /**< The number of times the extent list grows to long. */
3045 STAMPROFILE StatTrackDeref; /**< Profiling of SyncPageWorkerTrackDeref (expensive). */
3046
3047 /** Time spent by the host OS for large page allocation. */
3048 STAMPROFILE StatAllocLargePage;
3049 /** Time spent clearing the newly allocated large pages. */
3050 STAMPROFILE StatClearLargePage;
3051 /** The number of times allocating a large pages takes more than the allowed period. */
3052 STAMCOUNTER StatLargePageOverflow;
3053 /** pgmPhysIsValidLargePage profiling - R3 */
3054 STAMPROFILE StatR3IsValidLargePage;
3055 /** pgmPhysIsValidLargePage profiling - RZ*/
3056 STAMPROFILE StatRZIsValidLargePage;
3057
3058 STAMPROFILE StatChunkAging;
3059 STAMPROFILE StatChunkFindCandidate;
3060 STAMPROFILE StatChunkUnmap;
3061 STAMPROFILE StatChunkMap;
3062} PGMSTATS;
3063#endif /* VBOX_WITH_STATISTICS */
3064
3065
3066/**
3067 * Converts a PGM pointer into a VM pointer.
3068 * @returns Pointer to the VM structure the PGM is part of.
3069 * @param pPGM Pointer to PGM instance data.
3070 */
3071#define PGM2VM(pPGM) ( (PVM)((char*)pPGM - pPGM->offVM) )
3072
3073/**
3074 * PGM Data (part of VM)
3075 */
3076typedef struct PGM
3077{
3078 /** Offset to the VM structure. */
3079 int32_t offVM;
3080 /** Offset of the PGMCPU structure relative to VMCPU. */
3081 int32_t offVCpuPGM;
3082
3083 /** @cfgm{/RamPreAlloc, boolean, false}
3084 * Indicates whether the base RAM should all be allocated before starting
3085 * the VM (default), or if it should be allocated when first written to.
3086 */
3087 bool fRamPreAlloc;
3088 /** Indicates whether write monitoring is currently in use.
3089 * This is used to prevent conflicts between live saving and page sharing
3090 * detection. */
3091 bool fPhysWriteMonitoringEngaged;
3092 /** Set if the CPU has less than 52-bit physical address width.
3093 * This is used */
3094 bool fLessThan52PhysicalAddressBits;
3095 /** Set when nested paging is active.
3096 * This is meant to save calls to HMIsNestedPagingActive and let the
3097 * compilers optimize the code better. Whether we use nested paging or
3098 * not is something we find out during VMM initialization and we won't
3099 * change this later on. */
3100 bool fNestedPaging;
3101 /** The host paging mode. (This is what SUPLib reports.) */
3102 SUPPAGINGMODE enmHostMode;
3103 /** We're not in a state which permits writes to guest memory.
3104 * (Only used in strict builds.) */
3105 bool fNoMorePhysWrites;
3106 /** @cfgm{/PageFusionAllowed, boolean, false}
3107 * Whether page fusion is allowed. */
3108 bool fPageFusionAllowed;
3109 /** @cfgm{/PGM/PciPassThrough, boolean, false}
3110 * Whether PCI passthrough is enabled. */
3111 bool fPciPassthrough;
3112 /** The number of MMIO2 regions (serves as the next MMIO2 ID). */
3113 uint8_t cMmio2Regions;
3114 /** Restore original ROM page content when resetting after loading state.
3115 * The flag is set by pgmR3LoadRomRanges and cleared at reset. This
3116 * enables the VM to start using an updated ROM without requiring powering
3117 * down the VM, just rebooting or resetting it. */
3118 bool fRestoreRomPagesOnReset;
3119 /** Whether to automatically clear all RAM pages on reset. */
3120 bool fZeroRamPagesOnReset;
3121 /** Alignment padding. */
3122 bool afAlignment3[7];
3123
3124 /** Indicates that PGMR3FinalizeMappings has been called and that further
3125 * PGMR3MapIntermediate calls will be rejected. */
3126 bool fFinalizedMappings;
3127 /** If set no conflict checks are required. */
3128 bool fMappingsFixed;
3129 /** If set if restored as fixed but we were unable to re-fixate at the old
3130 * location because of room or address incompatibilities. */
3131 bool fMappingsFixedRestored;
3132 /** Size of fixed mapping.
3133 * This is valid if either fMappingsFixed or fMappingsFixedRestored is set. */
3134 uint32_t cbMappingFixed;
3135 /** Generation ID for the RAM ranges. This member is incremented everytime
3136 * a RAM range is linked or unlinked. */
3137 uint32_t volatile idRamRangesGen;
3138
3139 /** Base address (GC) of fixed mapping.
3140 * This is valid if either fMappingsFixed or fMappingsFixedRestored is set. */
3141 RTGCPTR GCPtrMappingFixed;
3142 /** The address of the previous RAM range mapping. */
3143 RTGCPTR GCPtrPrevRamRangeMapping;
3144
3145 /** Physical access handler type for ROM protection. */
3146 PGMPHYSHANDLERTYPE hRomPhysHandlerType;
3147 /** Alignment padding. */
3148 uint32_t u32Padding;
3149
3150 /** 4 MB page mask; 32 or 36 bits depending on PSE-36 (identical for all VCPUs) */
3151 RTGCPHYS GCPhys4MBPSEMask;
3152 /** Mask containing the invalid bits of a guest physical address.
3153 * @remarks this does not stop at bit 52. */
3154 RTGCPHYS GCPhysInvAddrMask;
3155
3156
3157 /** RAM range TLB for R3. */
3158 R3PTRTYPE(PPGMRAMRANGE) apRamRangesTlbR3[PGM_RAMRANGE_TLB_ENTRIES];
3159 /** Pointer to the list of RAM ranges (Phys GC -> Phys HC conversion) - for R3.
3160 * This is sorted by physical address and contains no overlapping ranges. */
3161 R3PTRTYPE(PPGMRAMRANGE) pRamRangesXR3;
3162 /** Root of the RAM range search tree for ring-3. */
3163 R3PTRTYPE(PPGMRAMRANGE) pRamRangeTreeR3;
3164 /** PGM offset based trees - R3 Ptr. */
3165 R3PTRTYPE(PPGMTREES) pTreesR3;
3166 /** Caching the last physical handler we looked up in R3. */
3167 R3PTRTYPE(PPGMPHYSHANDLER) pLastPhysHandlerR3;
3168 /** Shadow Page Pool - R3 Ptr. */
3169 R3PTRTYPE(PPGMPOOL) pPoolR3;
3170#ifndef PGM_WITHOUT_MAPPINGS
3171 /** Linked list of GC mappings - for HC.
3172 * The list is sorted ascending on address. */
3173 R3PTRTYPE(PPGMMAPPING) pMappingsR3;
3174#endif
3175 /** Pointer to the list of ROM ranges - for R3.
3176 * This is sorted by physical address and contains no overlapping ranges. */
3177 R3PTRTYPE(PPGMROMRANGE) pRomRangesR3;
3178 /** Pointer to the list of MMIO2 ranges - for R3.
3179 * Registration order. */
3180 R3PTRTYPE(PPGMREGMMIO2RANGE) pRegMmioRangesR3;
3181 /** MMIO2 lookup array for ring-3. Indexed by idMmio2 minus 1. */
3182 R3PTRTYPE(PPGMREGMMIO2RANGE) apMmio2RangesR3[PGM_MMIO2_MAX_RANGES];
3183
3184 /** RAM range TLB for R0. */
3185 R0PTRTYPE(PPGMRAMRANGE) apRamRangesTlbR0[PGM_RAMRANGE_TLB_ENTRIES];
3186 /** R0 pointer corresponding to PGM::pRamRangesXR3. */
3187 R0PTRTYPE(PPGMRAMRANGE) pRamRangesXR0;
3188 /** Root of the RAM range search tree for ring-0. */
3189 R0PTRTYPE(PPGMRAMRANGE) pRamRangeTreeR0;
3190 /** PGM offset based trees - R0 Ptr. */
3191 R0PTRTYPE(PPGMTREES) pTreesR0;
3192 /** Caching the last physical handler we looked up in R0. */
3193 R0PTRTYPE(PPGMPHYSHANDLER) pLastPhysHandlerR0;
3194 /** Shadow Page Pool - R0 Ptr. */
3195 R0PTRTYPE(PPGMPOOL) pPoolR0;
3196#ifndef PGM_WITHOUT_MAPPINGS
3197 /** Linked list of GC mappings - for R0.
3198 * The list is sorted ascending on address. */
3199 R0PTRTYPE(PPGMMAPPING) pMappingsR0;
3200 RTR0PTR R0PtrAlignment0;
3201#endif
3202 /** R0 pointer corresponding to PGM::pRomRangesR3. */
3203 R0PTRTYPE(PPGMROMRANGE) pRomRangesR0;
3204 /** MMIO2 lookup array for ring-0. Indexed by idMmio2 minus 1. */
3205 R0PTRTYPE(PPGMREGMMIO2RANGE) apMmio2RangesR0[PGM_MMIO2_MAX_RANGES];
3206
3207#ifndef PGM_WITHOUT_MAPPINGS
3208 /** Pointer to the 5 page CR3 content mapping.
3209 * The first page is always the CR3 (in some form) while the 4 other pages
3210 * are used for the PDs in PAE mode. */
3211 RTGCPTR GCPtrCR3Mapping;
3212
3213 /** @name Intermediate Context
3214 * @{ */
3215 /** Pointer to the intermediate page directory - Normal. */
3216 R3PTRTYPE(PX86PD) pInterPD;
3217 /** Pointer to the intermediate page tables - Normal.
3218 * There are two page tables, one for the identity mapping and one for
3219 * the host context mapping (of the core code). */
3220 R3PTRTYPE(PX86PT) apInterPTs[2];
3221 /** Pointer to the intermediate page tables - PAE. */
3222 R3PTRTYPE(PX86PTPAE) apInterPaePTs[2];
3223 /** Pointer to the intermediate page directory - PAE. */
3224 R3PTRTYPE(PX86PDPAE) apInterPaePDs[4];
3225 /** Pointer to the intermediate page directory - PAE. */
3226 R3PTRTYPE(PX86PDPT) pInterPaePDPT;
3227 /** Pointer to the intermediate page-map level 4 - AMD64. */
3228 R3PTRTYPE(PX86PML4) pInterPaePML4;
3229 /** Pointer to the intermediate page directory - AMD64. */
3230 R3PTRTYPE(PX86PDPT) pInterPaePDPT64;
3231 /** The Physical Address (HC) of the intermediate Page Directory - Normal. */
3232 RTHCPHYS HCPhysInterPD;
3233 /** The Physical Address (HC) of the intermediate Page Directory Pointer Table - PAE. */
3234 RTHCPHYS HCPhysInterPaePDPT;
3235 /** The Physical Address (HC) of the intermediate Page Map Level 4 table - AMD64. */
3236 RTHCPHYS HCPhysInterPaePML4;
3237 /** @} */
3238#endif
3239
3240#ifndef PGM_WITHOUT_MAPPINGS
3241 /** Base address of the dynamic page mapping area.
3242 * The array is MM_HYPER_DYNAMIC_SIZE bytes big.
3243 *
3244 * @todo The plan of keeping PGMRCDYNMAP private to PGMRZDynMap.cpp didn't
3245 * work out. Some cleaning up of the initialization that would
3246 * remove this memory is yet to be done...
3247 */
3248 RCPTRTYPE(uint8_t *) pbDynPageMapBaseGC;
3249 /** The address of the raw-mode context mapping cache. */
3250 RCPTRTYPE(PPGMRCDYNMAP) pRCDynMap;
3251 /** The address of the ring-0 mapping cache if we're making use of it. */
3252 RTR0PTR pvR0DynMapUsed;
3253#endif
3254
3255 /** Hack: Number of deprecated page mapping locks taken by the current lock
3256 * owner via pgmPhysGCPhys2CCPtrInternalDepr. */
3257 uint32_t cDeprecatedPageLocks;
3258 /** Alignment padding. */
3259 uint32_t au32Alignment2[1];
3260
3261
3262 /** PGM critical section.
3263 * This protects the physical, ram ranges, and the page flag updating (some of
3264 * it anyway).
3265 */
3266 PDMCRITSECT CritSectX;
3267
3268 /**
3269 * Data associated with managing the ring-3 mappings of the allocation chunks.
3270 */
3271 struct
3272 {
3273 /** The chunk mapping TLB. */
3274 PGMCHUNKR3MAPTLB Tlb;
3275 /** The chunk tree, ordered by chunk id. */
3276#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
3277 R3PTRTYPE(PAVLU32NODECORE) pTree;
3278#else
3279 R3R0PTRTYPE(PAVLU32NODECORE) pTree;
3280#endif
3281#if HC_ARCH_BITS == 32
3282 uint32_t u32Alignment0;
3283#endif
3284 /** The number of mapped chunks. */
3285 uint32_t c;
3286 /** @cfgm{/PGM/MaxRing3Chunks, uint32_t, host dependent}
3287 * The maximum number of mapped chunks. On 64-bit this is unlimited by default,
3288 * on 32-bit it defaults to 1 or 3 GB depending on the host. */
3289 uint32_t cMax;
3290 /** The current time. This is incremented whenever a chunk is inserted. */
3291 uint32_t iNow;
3292 /** Alignment padding. */
3293 uint32_t au32Alignment1[3];
3294 } ChunkR3Map;
3295
3296 /** The page mapping TLB for ring-3. */
3297 PGMPAGER3MAPTLB PhysTlbR3;
3298#ifdef VBOX_WITH_RAM_IN_KERNEL
3299 /** The page mapping TLB for ring-0. */
3300 PGMPAGER0MAPTLB PhysTlbR0;
3301#else
3302 /** The page mapping TLB for ring-0 (still using ring-3 mappings). */
3303 PGMPAGER3MAPTLB PhysTlbR0;
3304#endif
3305
3306 /** @name The zero page.
3307 * @{ */
3308 /** The host physical address of the zero page. */
3309 RTHCPHYS HCPhysZeroPg;
3310 /** The ring-3 mapping of the zero page. */
3311 RTR3PTR pvZeroPgR3;
3312 /** The ring-0 mapping of the zero page. */
3313 RTR0PTR pvZeroPgR0;
3314 /** The GC mapping of the zero page. */
3315 RTRCPTR pvZeroPgRC;
3316 RTRCPTR RCPtrAlignment3;
3317 /** @}*/
3318
3319 /** @name The Invalid MMIO page.
3320 * This page is filled with 0xfeedface.
3321 * @{ */
3322 /** The host physical address of the invalid MMIO page. */
3323 RTHCPHYS HCPhysMmioPg;
3324 /** The host pysical address of the invalid MMIO page plus all invalid
3325 * physical address bits set. This is used to trigger X86_TRAP_PF_RSVD.
3326 * @remarks Check fLessThan52PhysicalAddressBits before use. */
3327 RTHCPHYS HCPhysInvMmioPg;
3328 /** The ring-3 mapping of the invalid MMIO page. */
3329 RTR3PTR pvMmioPgR3;
3330#if HC_ARCH_BITS == 32
3331 RTR3PTR R3PtrAlignment4;
3332#endif
3333 /** @} */
3334
3335
3336 /** The number of handy pages. */
3337 uint32_t cHandyPages;
3338
3339 /** The number of large handy pages. */
3340 uint32_t cLargeHandyPages;
3341
3342 /**
3343 * Array of handy pages.
3344 *
3345 * This array is used in a two way communication between pgmPhysAllocPage
3346 * and GMMR0AllocateHandyPages, with PGMR3PhysAllocateHandyPages serving as
3347 * an intermediary.
3348 *
3349 * The size of this array is important, see pgmPhysEnsureHandyPage for details.
3350 * (The current size of 32 pages, means 128 KB of handy memory.)
3351 */
3352 GMMPAGEDESC aHandyPages[PGM_HANDY_PAGES];
3353
3354 /**
3355 * Array of large handy pages. (currently size 1)
3356 *
3357 * This array is used in a two way communication between pgmPhysAllocLargePage
3358 * and GMMR0AllocateLargePage, with PGMR3PhysAllocateLargePage serving as
3359 * an intermediary.
3360 */
3361 GMMPAGEDESC aLargeHandyPage[1];
3362
3363 /**
3364 * Live save data.
3365 */
3366 struct
3367 {
3368 /** Per type statistics. */
3369 struct
3370 {
3371 /** The number of ready pages. */
3372 uint32_t cReadyPages;
3373 /** The number of dirty pages. */
3374 uint32_t cDirtyPages;
3375 /** The number of ready zero pages. */
3376 uint32_t cZeroPages;
3377 /** The number of write monitored pages. */
3378 uint32_t cMonitoredPages;
3379 } Rom,
3380 Mmio2,
3381 Ram;
3382 /** The number of ignored pages in the RAM ranges (i.e. MMIO, MMIO2 and ROM). */
3383 uint32_t cIgnoredPages;
3384 /** Indicates that a live save operation is active. */
3385 bool fActive;
3386 /** Padding. */
3387 bool afReserved[2];
3388 /** The next history index. */
3389 uint8_t iDirtyPagesHistory;
3390 /** History of the total amount of dirty pages. */
3391 uint32_t acDirtyPagesHistory[64];
3392 /** Short term dirty page average. */
3393 uint32_t cDirtyPagesShort;
3394 /** Long term dirty page average. */
3395 uint32_t cDirtyPagesLong;
3396 /** The number of saved pages. This is used to get some kind of estimate of the
3397 * link speed so we can decide when we're done. It is reset after the first
3398 * 7 passes so the speed estimate doesn't get inflated by the initial set of
3399 * zero pages. */
3400 uint64_t cSavedPages;
3401 /** The nanosecond timestamp when cSavedPages was 0. */
3402 uint64_t uSaveStartNS;
3403 /** Pages per second (for statistics). */
3404 uint32_t cPagesPerSecond;
3405 uint32_t cAlignment;
3406 } LiveSave;
3407
3408 /** @name Error injection.
3409 * @{ */
3410 /** Inject handy page allocation errors pretending we're completely out of
3411 * memory. */
3412 bool volatile fErrInjHandyPages;
3413 /** Padding. */
3414 bool afReserved[3];
3415 /** @} */
3416
3417 /** @name Release Statistics
3418 * @{ */
3419 uint32_t cAllPages; /**< The total number of pages. (Should be Private + Shared + Zero + Pure MMIO.) */
3420 uint32_t cPrivatePages; /**< The number of private pages. */
3421 uint32_t cSharedPages; /**< The number of shared pages. */
3422 uint32_t cReusedSharedPages; /**< The number of reused shared pages. */
3423 uint32_t cZeroPages; /**< The number of zero backed pages. */
3424 uint32_t cPureMmioPages; /**< The number of pure MMIO pages. */
3425 uint32_t cMonitoredPages; /**< The number of write monitored pages. */
3426 uint32_t cWrittenToPages; /**< The number of previously write monitored pages. */
3427 uint32_t cWriteLockedPages; /**< The number of write locked pages. */
3428 uint32_t cReadLockedPages; /**< The number of read locked pages. */
3429 uint32_t cBalloonedPages; /**< The number of ballooned pages. */
3430 uint32_t cMappedChunks; /**< Number of times we mapped a chunk. */
3431 uint32_t cUnmappedChunks; /**< Number of times we unmapped a chunk. */
3432 uint32_t cLargePages; /**< The number of large pages. */
3433 uint32_t cLargePagesDisabled; /**< The number of disabled large pages. */
3434/* uint32_t aAlignment4[1]; */
3435
3436 /** The number of times we were forced to change the hypervisor region location. */
3437 STAMCOUNTER cRelocations;
3438
3439 STAMCOUNTER StatLargePageReused; /**< The number of large pages we've reused.*/
3440 STAMCOUNTER StatLargePageRefused; /**< The number of times we couldn't use a large page.*/
3441 STAMCOUNTER StatLargePageRecheck; /**< The number of times we rechecked a disabled large page.*/
3442
3443 STAMPROFILE StatShModCheck; /**< Profiles shared module checks. */
3444 /** @} */
3445
3446#ifdef VBOX_WITH_STATISTICS
3447 /** @name Statistics on the heap.
3448 * @{ */
3449 R3PTRTYPE(PGMSTATS *) pStatsR3;
3450 R0PTRTYPE(PGMSTATS *) pStatsR0;
3451 /** @} */
3452#endif
3453} PGM;
3454#ifndef IN_TSTVMSTRUCTGC /* HACK */
3455# ifndef PGM_WITHOUT_MAPPINGS
3456AssertCompileMemberAlignment(PGM, paDynPageMap32BitPTEsGC, 8);
3457# endif
3458AssertCompileMemberAlignment(PGM, GCPtrMappingFixed, sizeof(RTGCPTR));
3459# ifndef PGM_WITHOUT_MAPPINGS
3460AssertCompileMemberAlignment(PGM, HCPhysInterPD, 8);
3461# endif
3462AssertCompileMemberAlignment(PGM, CritSectX, 8);
3463AssertCompileMemberAlignment(PGM, ChunkR3Map, 16);
3464AssertCompileMemberAlignment(PGM, PhysTlbR3, 32); /** @todo 32 byte alignment! */
3465AssertCompileMemberAlignment(PGM, PhysTlbR0, 32);
3466AssertCompileMemberAlignment(PGM, HCPhysZeroPg, 8);
3467AssertCompileMemberAlignment(PGM, aHandyPages, 8);
3468AssertCompileMemberAlignment(PGM, cRelocations, 8);
3469#endif /* !IN_TSTVMSTRUCTGC */
3470/** Pointer to the PGM instance data. */
3471typedef PGM *PPGM;
3472
3473
3474
3475typedef struct PGMCPUSTATS
3476{
3477 /* Common */
3478 STAMCOUNTER StatSyncPtPD[X86_PG_ENTRIES]; /**< SyncPT - PD distribution. */
3479 STAMCOUNTER StatSyncPagePD[X86_PG_ENTRIES]; /**< SyncPage - PD distribution. */
3480
3481 /* R0 only: */
3482 STAMPROFILE StatR0NpMiscfg; /**< R0: PGMR0Trap0eHandlerNPMisconfig() profiling. */
3483 STAMCOUNTER StatR0NpMiscfgSyncPage; /**< R0: SyncPage calls from PGMR0Trap0eHandlerNPMisconfig(). */
3484
3485 /* RZ only: */
3486 STAMPROFILE StatRZTrap0e; /**< RC/R0: PGMTrap0eHandler() profiling. */
3487 STAMPROFILE StatRZTrap0eTime2Ballooned; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is read access to a ballooned page. */
3488 STAMPROFILE StatRZTrap0eTime2CSAM; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CSAM. */
3489 STAMPROFILE StatRZTrap0eTime2DirtyAndAccessed; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is dirty and/or accessed bit emulation. */
3490 STAMPROFILE StatRZTrap0eTime2GuestTrap; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is a guest trap. */
3491 STAMPROFILE StatRZTrap0eTime2HndPhys; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is a physical handler. */
3492 STAMPROFILE StatRZTrap0eTime2HndUnhandled; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is access outside the monitored areas of a monitored page. */
3493 STAMPROFILE StatRZTrap0eTime2InvalidPhys; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is access to an invalid physical guest address. */
3494 STAMPROFILE StatRZTrap0eTime2MakeWritable; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is a page that needed to be made writable. */
3495 STAMPROFILE StatRZTrap0eTime2Mapping; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is the guest mappings. */
3496 STAMPROFILE StatRZTrap0eTime2Misc; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is not known. */
3497 STAMPROFILE StatRZTrap0eTime2OutOfSync; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is an out-of-sync page. */
3498 STAMPROFILE StatRZTrap0eTime2OutOfSyncHndPhys; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is an out-of-sync physical handler page. */
3499 STAMPROFILE StatRZTrap0eTime2OutOfSyncHndObs; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is an obsolete handler page. */
3500 STAMPROFILE StatRZTrap0eTime2SyncPT; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is lazy syncing of a PT. */
3501 STAMPROFILE StatRZTrap0eTime2WPEmulation; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CR0.WP emulation. */
3502 STAMPROFILE StatRZTrap0eTime2Wp0RoUsHack; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CR0.WP and netware hack to be enabled. */
3503 STAMPROFILE StatRZTrap0eTime2Wp0RoUsUnhack; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CR0.WP and netware hack to be disabled. */
3504 STAMCOUNTER StatRZTrap0eConflicts; /**< RC/R0: The number of times \#PF was caused by an undetected conflict. */
3505 STAMCOUNTER StatRZTrap0eHandlersMapping; /**< RC/R0: Number of traps due to access handlers in mappings. */
3506 STAMCOUNTER StatRZTrap0eHandlersOutOfSync; /**< RC/R0: Number of out-of-sync handled pages. */
3507 STAMCOUNTER StatRZTrap0eHandlersPhysAll; /**< RC/R0: Number of traps due to physical all-access handlers. */
3508 STAMCOUNTER StatRZTrap0eHandlersPhysAllOpt; /**< RC/R0: Number of the physical all-access handler traps using the optimization. */
3509 STAMCOUNTER StatRZTrap0eHandlersPhysWrite; /**< RC/R0: Number of traps due to write-physical access handlers. */
3510 STAMCOUNTER StatRZTrap0eHandlersUnhandled; /**< RC/R0: Number of traps due to access outside range of monitored page(s). */
3511 STAMCOUNTER StatRZTrap0eHandlersInvalid; /**< RC/R0: Number of traps due to access to invalid physical memory. */
3512 STAMCOUNTER StatRZTrap0eUSNotPresentRead; /**< RC/R0: \#PF err kind */
3513 STAMCOUNTER StatRZTrap0eUSNotPresentWrite; /**< RC/R0: \#PF err kind */
3514 STAMCOUNTER StatRZTrap0eUSWrite; /**< RC/R0: \#PF err kind */
3515 STAMCOUNTER StatRZTrap0eUSReserved; /**< RC/R0: \#PF err kind */
3516 STAMCOUNTER StatRZTrap0eUSNXE; /**< RC/R0: \#PF err kind */
3517 STAMCOUNTER StatRZTrap0eUSRead; /**< RC/R0: \#PF err kind */
3518 STAMCOUNTER StatRZTrap0eSVNotPresentRead; /**< RC/R0: \#PF err kind */
3519 STAMCOUNTER StatRZTrap0eSVNotPresentWrite; /**< RC/R0: \#PF err kind */
3520 STAMCOUNTER StatRZTrap0eSVWrite; /**< RC/R0: \#PF err kind */
3521 STAMCOUNTER StatRZTrap0eSVReserved; /**< RC/R0: \#PF err kind */
3522 STAMCOUNTER StatRZTrap0eSNXE; /**< RC/R0: \#PF err kind */
3523 STAMCOUNTER StatRZTrap0eGuestPF; /**< RC/R0: Real guest \#PFs. */
3524 STAMCOUNTER StatRZTrap0eGuestPFMapping; /**< RC/R0: Real guest \#PF to HMA or other mapping. */
3525 STAMCOUNTER StatRZTrap0eWPEmulInRZ; /**< RC/R0: WP=0 virtualization trap, handled. */
3526 STAMCOUNTER StatRZTrap0eWPEmulToR3; /**< RC/R0: WP=0 virtualization trap, chickened out. */
3527 STAMCOUNTER StatRZTrap0ePD[X86_PG_ENTRIES]; /**< RC/R0: PD distribution of the \#PFs. */
3528 STAMCOUNTER StatRZGuestCR3WriteHandled; /**< RC/R0: The number of times WriteHandlerCR3() was successfully called. */
3529 STAMCOUNTER StatRZGuestCR3WriteUnhandled; /**< RC/R0: The number of times WriteHandlerCR3() was called and we had to fall back to the recompiler. */
3530 STAMCOUNTER StatRZGuestCR3WriteConflict; /**< RC/R0: The number of times WriteHandlerCR3() was called and a conflict was detected. */
3531 STAMCOUNTER StatRZGuestROMWriteHandled; /**< RC/R0: The number of times pgmPhysRomWriteHandler() was successfully called. */
3532 STAMCOUNTER StatRZGuestROMWriteUnhandled; /**< RC/R0: The number of times pgmPhysRomWriteHandler() was called and we had to fall back to the recompiler */
3533 STAMCOUNTER StatRZDynMapMigrateInvlPg; /**< RZ: invlpg in PGMR0DynMapMigrateAutoSet. */
3534 STAMPROFILE StatRZDynMapGCPageInl; /**< RZ: Calls to pgmRZDynMapGCPageInlined. */
3535 STAMCOUNTER StatRZDynMapGCPageInlHits; /**< RZ: Hash table lookup hits. */
3536 STAMCOUNTER StatRZDynMapGCPageInlMisses; /**< RZ: Misses that falls back to the code common. */
3537 STAMCOUNTER StatRZDynMapGCPageInlRamHits; /**< RZ: 1st ram range hits. */
3538 STAMCOUNTER StatRZDynMapGCPageInlRamMisses; /**< RZ: 1st ram range misses, takes slow path. */
3539 STAMPROFILE StatRZDynMapHCPageInl; /**< RZ: Calls to pgmRZDynMapHCPageInlined. */
3540 STAMCOUNTER StatRZDynMapHCPageInlHits; /**< RZ: Hash table lookup hits. */
3541 STAMCOUNTER StatRZDynMapHCPageInlMisses; /**< RZ: Misses that falls back to the code common. */
3542 STAMPROFILE StatRZDynMapHCPage; /**< RZ: Calls to pgmRZDynMapHCPageCommon. */
3543 STAMCOUNTER StatRZDynMapSetOptimize; /**< RZ: Calls to pgmRZDynMapOptimizeAutoSet. */
3544 STAMCOUNTER StatRZDynMapSetSearchFlushes; /**< RZ: Set search restoring to subset flushes. */
3545 STAMCOUNTER StatRZDynMapSetSearchHits; /**< RZ: Set search hits. */
3546 STAMCOUNTER StatRZDynMapSetSearchMisses; /**< RZ: Set search misses. */
3547 STAMCOUNTER StatRZDynMapPage; /**< RZ: Calls to pgmR0DynMapPage. */
3548 STAMCOUNTER StatRZDynMapPageHits0; /**< RZ: Hits at iPage+0. */
3549 STAMCOUNTER StatRZDynMapPageHits1; /**< RZ: Hits at iPage+1. */
3550 STAMCOUNTER StatRZDynMapPageHits2; /**< RZ: Hits at iPage+2. */
3551 STAMCOUNTER StatRZDynMapPageInvlPg; /**< RZ: invlpg. */
3552 STAMCOUNTER StatRZDynMapPageSlow; /**< RZ: Calls to pgmR0DynMapPageSlow. */
3553 STAMCOUNTER StatRZDynMapPageSlowLoopHits; /**< RZ: Hits in the pgmR0DynMapPageSlow search loop. */
3554 STAMCOUNTER StatRZDynMapPageSlowLoopMisses; /**< RZ: Misses in the pgmR0DynMapPageSlow search loop. */
3555 //STAMCOUNTER StatRZDynMapPageSlowLostHits; /**< RZ: Lost hits. */
3556 STAMCOUNTER StatRZDynMapSubsets; /**< RZ: Times PGMDynMapPushAutoSubset was called. */
3557 STAMCOUNTER StatRZDynMapPopFlushes; /**< RZ: Times PGMDynMapPopAutoSubset flushes the subset. */
3558 STAMCOUNTER aStatRZDynMapSetFilledPct[11]; /**< RZ: Set fill distribution, percent. */
3559
3560 /* HC - R3 and (maybe) R0: */
3561
3562 /* RZ & R3: */
3563 STAMPROFILE StatRZSyncCR3; /**< RC/R0: PGMSyncCR3() profiling. */
3564 STAMPROFILE StatRZSyncCR3Handlers; /**< RC/R0: Profiling of the PGMSyncCR3() update handler section. */
3565 STAMCOUNTER StatRZSyncCR3Global; /**< RC/R0: The number of global CR3 syncs. */
3566 STAMCOUNTER StatRZSyncCR3NotGlobal; /**< RC/R0: The number of non-global CR3 syncs. */
3567 STAMCOUNTER StatRZSyncCR3DstCacheHit; /**< RC/R0: The number of times we got some kind of cache hit on a page table. */
3568 STAMCOUNTER StatRZSyncCR3DstFreed; /**< RC/R0: The number of times we've had to free a shadow entry. */
3569 STAMCOUNTER StatRZSyncCR3DstFreedSrcNP; /**< RC/R0: The number of times we've had to free a shadow entry for which the source entry was not present. */
3570 STAMCOUNTER StatRZSyncCR3DstNotPresent; /**< RC/R0: The number of times we've encountered a not present shadow entry for a present guest entry. */
3571 STAMCOUNTER StatRZSyncCR3DstSkippedGlobalPD; /**< RC/R0: The number of times a global page directory wasn't flushed. */
3572 STAMCOUNTER StatRZSyncCR3DstSkippedGlobalPT; /**< RC/R0: The number of times a page table with only global entries wasn't flushed. */
3573 STAMPROFILE StatRZSyncPT; /**< RC/R0: PGMSyncPT() profiling. */
3574 STAMCOUNTER StatRZSyncPTFailed; /**< RC/R0: The number of times PGMSyncPT() failed. */
3575 STAMCOUNTER StatRZSyncPT4K; /**< RC/R0: Number of 4KB syncs. */
3576 STAMCOUNTER StatRZSyncPT4M; /**< RC/R0: Number of 4MB syncs. */
3577 STAMCOUNTER StatRZSyncPagePDNAs; /**< RC/R0: The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit. */
3578 STAMCOUNTER StatRZSyncPagePDOutOfSync; /**< RC/R0: The number of time we've encountered an out-of-sync PD in SyncPage. */
3579 STAMCOUNTER StatRZAccessedPage; /**< RC/R0: The number of pages marked not present for accessed bit emulation. */
3580 STAMPROFILE StatRZDirtyBitTracking; /**< RC/R0: Profiling the dirty bit tracking in CheckPageFault(). */
3581 STAMCOUNTER StatRZDirtyPage; /**< RC/R0: The number of pages marked read-only for dirty bit tracking. */
3582 STAMCOUNTER StatRZDirtyPageBig; /**< RC/R0: The number of pages marked read-only for dirty bit tracking. */
3583 STAMCOUNTER StatRZDirtyPageSkipped; /**< RC/R0: The number of pages already dirty or readonly. */
3584 STAMCOUNTER StatRZDirtyPageTrap; /**< RC/R0: The number of traps generated for dirty bit tracking. */
3585 STAMCOUNTER StatRZDirtyPageStale; /**< RC/R0: The number of traps generated for dirty bit tracking. (stale tlb entries) */
3586 STAMCOUNTER StatRZDirtyTrackRealPF; /**< RC/R0: The number of real pages faults during dirty bit tracking. */
3587 STAMCOUNTER StatRZDirtiedPage; /**< RC/R0: The number of pages marked dirty because of write accesses. */
3588 STAMCOUNTER StatRZPageAlreadyDirty; /**< RC/R0: The number of pages already marked dirty because of write accesses. */
3589 STAMPROFILE StatRZInvalidatePage; /**< RC/R0: PGMInvalidatePage() profiling. */
3590 STAMCOUNTER StatRZInvalidatePage4KBPages; /**< RC/R0: The number of times PGMInvalidatePage() was called for a 4KB page. */
3591 STAMCOUNTER StatRZInvalidatePage4MBPages; /**< RC/R0: The number of times PGMInvalidatePage() was called for a 4MB page. */
3592 STAMCOUNTER StatRZInvalidatePage4MBPagesSkip; /**< RC/R0: The number of times PGMInvalidatePage() skipped a 4MB page. */
3593 STAMCOUNTER StatRZInvalidatePagePDMappings; /**< RC/R0: The number of times PGMInvalidatePage() was called for a page directory containing mappings (no conflict). */
3594 STAMCOUNTER StatRZInvalidatePagePDNAs; /**< RC/R0: The number of times PGMInvalidatePage() was called for a not accessed page directory. */
3595 STAMCOUNTER StatRZInvalidatePagePDNPs; /**< RC/R0: The number of times PGMInvalidatePage() was called for a not present page directory. */
3596 STAMCOUNTER StatRZInvalidatePagePDOutOfSync; /**< RC/R0: The number of times PGMInvalidatePage() was called for an out of sync page directory. */
3597 STAMCOUNTER StatRZInvalidatePageSizeChanges ; /**< RC/R0: The number of times PGMInvalidatePage() was called on a page size change (4KB <-> 2/4MB). */
3598 STAMCOUNTER StatRZInvalidatePageSkipped; /**< RC/R0: The number of times PGMInvalidatePage() was skipped due to not present shw or pending pending SyncCR3. */
3599 STAMCOUNTER StatRZPageOutOfSyncUser; /**< RC/R0: The number of times user page is out of sync was detected in \#PF or VerifyAccessSyncPage. */
3600 STAMCOUNTER StatRZPageOutOfSyncSupervisor; /**< RC/R0: The number of times supervisor page is out of sync was detected in in \#PF or VerifyAccessSyncPage. */
3601 STAMCOUNTER StatRZPageOutOfSyncUserWrite; /**< RC/R0: The number of times user page is out of sync was detected in \#PF. */
3602 STAMCOUNTER StatRZPageOutOfSyncSupervisorWrite; /**< RC/R0: The number of times supervisor page is out of sync was detected in in \#PF. */
3603 STAMCOUNTER StatRZPageOutOfSyncBallloon; /**< RC/R0: The number of times a ballooned page was accessed (read). */
3604 STAMPROFILE StatRZPrefetch; /**< RC/R0: PGMPrefetchPage. */
3605 STAMPROFILE StatRZFlushTLB; /**< RC/R0: Profiling of the PGMFlushTLB() body. */
3606 STAMCOUNTER StatRZFlushTLBNewCR3; /**< RC/R0: The number of times PGMFlushTLB was called with a new CR3, non-global. (switch) */
3607 STAMCOUNTER StatRZFlushTLBNewCR3Global; /**< RC/R0: The number of times PGMFlushTLB was called with a new CR3, global. (switch) */
3608 STAMCOUNTER StatRZFlushTLBSameCR3; /**< RC/R0: The number of times PGMFlushTLB was called with the same CR3, non-global. (flush) */
3609 STAMCOUNTER StatRZFlushTLBSameCR3Global; /**< RC/R0: The number of times PGMFlushTLB was called with the same CR3, global. (flush) */
3610 STAMPROFILE StatRZGstModifyPage; /**< RC/R0: Profiling of the PGMGstModifyPage() body */
3611
3612 STAMPROFILE StatR3SyncCR3; /**< R3: PGMSyncCR3() profiling. */
3613 STAMPROFILE StatR3SyncCR3Handlers; /**< R3: Profiling of the PGMSyncCR3() update handler section. */
3614 STAMCOUNTER StatR3SyncCR3Global; /**< R3: The number of global CR3 syncs. */
3615 STAMCOUNTER StatR3SyncCR3NotGlobal; /**< R3: The number of non-global CR3 syncs. */
3616 STAMCOUNTER StatR3SyncCR3DstFreed; /**< R3: The number of times we've had to free a shadow entry. */
3617 STAMCOUNTER StatR3SyncCR3DstFreedSrcNP; /**< R3: The number of times we've had to free a shadow entry for which the source entry was not present. */
3618 STAMCOUNTER StatR3SyncCR3DstNotPresent; /**< R3: The number of times we've encountered a not present shadow entry for a present guest entry. */
3619 STAMCOUNTER StatR3SyncCR3DstSkippedGlobalPD; /**< R3: The number of times a global page directory wasn't flushed. */
3620 STAMCOUNTER StatR3SyncCR3DstSkippedGlobalPT; /**< R3: The number of times a page table with only global entries wasn't flushed. */
3621 STAMCOUNTER StatR3SyncCR3DstCacheHit; /**< R3: The number of times we got some kind of cache hit on a page table. */
3622 STAMPROFILE StatR3SyncPT; /**< R3: PGMSyncPT() profiling. */
3623 STAMCOUNTER StatR3SyncPTFailed; /**< R3: The number of times PGMSyncPT() failed. */
3624 STAMCOUNTER StatR3SyncPT4K; /**< R3: Number of 4KB syncs. */
3625 STAMCOUNTER StatR3SyncPT4M; /**< R3: Number of 4MB syncs. */
3626 STAMCOUNTER StatR3SyncPagePDNAs; /**< R3: The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit. */
3627 STAMCOUNTER StatR3SyncPagePDOutOfSync; /**< R3: The number of time we've encountered an out-of-sync PD in SyncPage. */
3628 STAMCOUNTER StatR3AccessedPage; /**< R3: The number of pages marked not present for accessed bit emulation. */
3629 STAMPROFILE StatR3DirtyBitTracking; /**< R3: Profiling the dirty bit tracking in CheckPageFault(). */
3630 STAMCOUNTER StatR3DirtyPage; /**< R3: The number of pages marked read-only for dirty bit tracking. */
3631 STAMCOUNTER StatR3DirtyPageBig; /**< R3: The number of pages marked read-only for dirty bit tracking. */
3632 STAMCOUNTER StatR3DirtyPageSkipped; /**< R3: The number of pages already dirty or readonly. */
3633 STAMCOUNTER StatR3DirtyPageTrap; /**< R3: The number of traps generated for dirty bit tracking. */
3634 STAMCOUNTER StatR3DirtyTrackRealPF; /**< R3: The number of real pages faults during dirty bit tracking. */
3635 STAMCOUNTER StatR3DirtiedPage; /**< R3: The number of pages marked dirty because of write accesses. */
3636 STAMCOUNTER StatR3PageAlreadyDirty; /**< R3: The number of pages already marked dirty because of write accesses. */
3637 STAMPROFILE StatR3InvalidatePage; /**< R3: PGMInvalidatePage() profiling. */
3638 STAMCOUNTER StatR3InvalidatePage4KBPages; /**< R3: The number of times PGMInvalidatePage() was called for a 4KB page. */
3639 STAMCOUNTER StatR3InvalidatePage4MBPages; /**< R3: The number of times PGMInvalidatePage() was called for a 4MB page. */
3640 STAMCOUNTER StatR3InvalidatePage4MBPagesSkip; /**< R3: The number of times PGMInvalidatePage() skipped a 4MB page. */
3641 STAMCOUNTER StatR3InvalidatePagePDNAs; /**< R3: The number of times PGMInvalidatePage() was called for a not accessed page directory. */
3642 STAMCOUNTER StatR3InvalidatePagePDNPs; /**< R3: The number of times PGMInvalidatePage() was called for a not present page directory. */
3643 STAMCOUNTER StatR3InvalidatePagePDMappings; /**< R3: The number of times PGMInvalidatePage() was called for a page directory containing mappings (no conflict). */
3644 STAMCOUNTER StatR3InvalidatePagePDOutOfSync; /**< R3: The number of times PGMInvalidatePage() was called for an out of sync page directory. */
3645 STAMCOUNTER StatR3InvalidatePageSizeChanges ; /**< R3: The number of times PGMInvalidatePage() was called on a page size change (4KB <-> 2/4MB). */
3646 STAMCOUNTER StatR3InvalidatePageSkipped; /**< R3: The number of times PGMInvalidatePage() was skipped due to not present shw or pending pending SyncCR3. */
3647 STAMCOUNTER StatR3PageOutOfSyncUser; /**< R3: The number of times user page is out of sync was detected in \#PF or VerifyAccessSyncPage. */
3648 STAMCOUNTER StatR3PageOutOfSyncSupervisor; /**< R3: The number of times supervisor page is out of sync was detected in in \#PF or VerifyAccessSyncPage. */
3649 STAMCOUNTER StatR3PageOutOfSyncUserWrite; /**< R3: The number of times user page is out of sync was detected in \#PF. */
3650 STAMCOUNTER StatR3PageOutOfSyncSupervisorWrite; /**< R3: The number of times supervisor page is out of sync was detected in in \#PF. */
3651 STAMCOUNTER StatR3PageOutOfSyncBallloon; /**< R3: The number of times a ballooned page was accessed (read). */
3652 STAMPROFILE StatR3Prefetch; /**< R3: PGMPrefetchPage. */
3653 STAMPROFILE StatR3FlushTLB; /**< R3: Profiling of the PGMFlushTLB() body. */
3654 STAMCOUNTER StatR3FlushTLBNewCR3; /**< R3: The number of times PGMFlushTLB was called with a new CR3, non-global. (switch) */
3655 STAMCOUNTER StatR3FlushTLBNewCR3Global; /**< R3: The number of times PGMFlushTLB was called with a new CR3, global. (switch) */
3656 STAMCOUNTER StatR3FlushTLBSameCR3; /**< R3: The number of times PGMFlushTLB was called with the same CR3, non-global. (flush) */
3657 STAMCOUNTER StatR3FlushTLBSameCR3Global; /**< R3: The number of times PGMFlushTLB was called with the same CR3, global. (flush) */
3658 STAMPROFILE StatR3GstModifyPage; /**< R3: Profiling of the PGMGstModifyPage() body */
3659} PGMCPUSTATS;
3660
3661
3662/**
3663 * Converts a PGMCPU pointer into a VM pointer.
3664 * @returns Pointer to the VM structure the PGM is part of.
3665 * @param pPGM Pointer to PGMCPU instance data.
3666 */
3667#define PGMCPU2VM(pPGM) ( (PVM)((char*)(pPGM) - (pPGM)->offVM) )
3668
3669/**
3670 * Converts a PGMCPU pointer into a PGM pointer.
3671 * @returns Pointer to the VM structure the PGM is part of.
3672 * @param pPGMCpu Pointer to PGMCPU instance data.
3673 */
3674#define PGMCPU2PGM(pPGMCpu) ( (PPGM)((char *)(pPGMCpu) - (pPGMCpu)->offPGM) )
3675
3676/**
3677 * PGMCPU Data (part of VMCPU).
3678 */
3679typedef struct PGMCPU
3680{
3681 /** Offset to the VM structure. */
3682 int32_t offVM;
3683 /** Offset to the VMCPU structure. */
3684 int32_t offVCpu;
3685 /** Offset of the PGM structure relative to VMCPU. */
3686 int32_t offPGM;
3687 uint32_t uPadding0; /**< structure size alignment. */
3688
3689#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE
3690 /** Automatically tracked physical memory mapping set.
3691 * Ring-0 and strict raw-mode builds. */
3692 PGMMAPSET AutoSet;
3693#endif
3694
3695 /** A20 gate mask.
3696 * Our current approach to A20 emulation is to let REM do it and don't bother
3697 * anywhere else. The interesting Guests will be operating with it enabled anyway.
3698 * But whould need arrise, we'll subject physical addresses to this mask. */
3699 RTGCPHYS GCPhysA20Mask;
3700 /** A20 gate state - boolean! */
3701 bool fA20Enabled;
3702 /** Mirror of the EFER.NXE bit. Managed by PGMNotifyNxeChanged. */
3703 bool fNoExecuteEnabled;
3704 /** Unused bits. */
3705 bool afUnused[2];
3706
3707 /** What needs syncing (PGM_SYNC_*).
3708 * This is used to queue operations for PGMSyncCR3, PGMInvalidatePage,
3709 * PGMFlushTLB, and PGMR3Load. */
3710 uint32_t fSyncFlags;
3711
3712 /** The shadow paging mode. */
3713 PGMMODE enmShadowMode;
3714 /** The guest paging mode. */
3715 PGMMODE enmGuestMode;
3716 /** Guest mode data table index (PGM_TYPE_XXX). */
3717 uint8_t volatile idxGuestModeData;
3718 /** Shadow mode data table index (PGM_TYPE_XXX). */
3719 uint8_t volatile idxShadowModeData;
3720 /** Both mode data table index (complicated). */
3721 uint8_t volatile idxBothModeData;
3722 /** Alignment padding. */
3723 uint8_t abPadding[5];
3724
3725 /** The current physical address represented in the guest CR3 register. */
3726 RTGCPHYS GCPhysCR3;
3727
3728 /** @name 32-bit Guest Paging.
3729 * @{ */
3730 /** The guest's page directory, R3 pointer. */
3731 R3PTRTYPE(PX86PD) pGst32BitPdR3;
3732#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3733 /** The guest's page directory, R0 pointer. */
3734 R0PTRTYPE(PX86PD) pGst32BitPdR0;
3735#endif
3736 /** Mask containing the MBZ bits of a big page PDE. */
3737 uint32_t fGst32BitMbzBigPdeMask;
3738 /** Set if the page size extension (PSE) is enabled. */
3739 bool fGst32BitPageSizeExtension;
3740 /** Alignment padding. */
3741 bool afAlignment2[3];
3742 /** @} */
3743
3744 /** @name PAE Guest Paging.
3745 * @{ */
3746 /** The guest's page directory pointer table, R3 pointer. */
3747 R3PTRTYPE(PX86PDPT) pGstPaePdptR3;
3748#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3749 /** The guest's page directory pointer table, R0 pointer. */
3750 R0PTRTYPE(PX86PDPT) pGstPaePdptR0;
3751#endif
3752
3753 /** The guest's page directories, R3 pointers.
3754 * These are individual pointers and don't have to be adjacent.
3755 * These don't have to be up-to-date - use pgmGstGetPaePD() to access them. */
3756 R3PTRTYPE(PX86PDPAE) apGstPaePDsR3[4];
3757 /** The guest's page directories, R0 pointers.
3758 * Same restrictions as apGstPaePDsR3. */
3759#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3760 R0PTRTYPE(PX86PDPAE) apGstPaePDsR0[4];
3761#endif
3762 /** The physical addresses of the guest page directories (PAE) pointed to by apGstPagePDsHC/GC. */
3763 RTGCPHYS aGCPhysGstPaePDs[4];
3764 /** The values of the 4 PDPE CPU registers (PAE).
3765 * @todo Not really maintained by PGM atm, only by VT-x in EPT mode. Should
3766 * load on cr3 load and use instead of guest memory version like real
3767 * HW. We probably should move these to the CPUMCTX and treat them
3768 * like the rest of the register wrt exporting to VT-x and import back. */
3769 X86PDPE aGstPaePdpeRegs[4];
3770 /** The physical addresses of the monitored guest page directories (PAE). */
3771 RTGCPHYS aGCPhysGstPaePDsMonitored[4];
3772 /** Mask containing the MBZ PTE bits. */
3773 uint64_t fGstPaeMbzPteMask;
3774 /** Mask containing the MBZ PDE bits. */
3775 uint64_t fGstPaeMbzPdeMask;
3776 /** Mask containing the MBZ big page PDE bits. */
3777 uint64_t fGstPaeMbzBigPdeMask;
3778 /** Mask containing the MBZ PDPE bits. */
3779 uint64_t fGstPaeMbzPdpeMask;
3780 /** @} */
3781
3782 /** @name AMD64 Guest Paging.
3783 * @{ */
3784 /** The guest's page directory pointer table, R3 pointer. */
3785 R3PTRTYPE(PX86PML4) pGstAmd64Pml4R3;
3786#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3787 /** The guest's page directory pointer table, R0 pointer. */
3788 R0PTRTYPE(PX86PML4) pGstAmd64Pml4R0;
3789#else
3790 RTR0PTR alignment6b; /**< alignment equalizer. */
3791#endif
3792 /** Mask containing the MBZ PTE bits. */
3793 uint64_t fGstAmd64MbzPteMask;
3794 /** Mask containing the MBZ PDE bits. */
3795 uint64_t fGstAmd64MbzPdeMask;
3796 /** Mask containing the MBZ big page PDE bits. */
3797 uint64_t fGstAmd64MbzBigPdeMask;
3798 /** Mask containing the MBZ PDPE bits. */
3799 uint64_t fGstAmd64MbzPdpeMask;
3800 /** Mask containing the MBZ big page PDPE bits. */
3801 uint64_t fGstAmd64MbzBigPdpeMask;
3802 /** Mask containing the MBZ PML4E bits. */
3803 uint64_t fGstAmd64MbzPml4eMask;
3804 /** Mask containing the PDPE bits that we shadow. */
3805 uint64_t fGstAmd64ShadowedPdpeMask;
3806 /** Mask containing the PML4E bits that we shadow. */
3807 uint64_t fGstAmd64ShadowedPml4eMask;
3808 /** @} */
3809
3810 /** @name PAE and AMD64 Guest Paging.
3811 * @{ */
3812 /** Mask containing the PTE bits that we shadow. */
3813 uint64_t fGst64ShadowedPteMask;
3814 /** Mask containing the PDE bits that we shadow. */
3815 uint64_t fGst64ShadowedPdeMask;
3816 /** Mask containing the big page PDE bits that we shadow in the PDE. */
3817 uint64_t fGst64ShadowedBigPdeMask;
3818 /** Mask containing the big page PDE bits that we shadow in the PTE. */
3819 uint64_t fGst64ShadowedBigPde4PteMask;
3820 /** @} */
3821
3822 /** Pointer to the page of the current active CR3 - R3 Ptr. */
3823 R3PTRTYPE(PPGMPOOLPAGE) pShwPageCR3R3;
3824 /** Pointer to the page of the current active CR3 - R0 Ptr. */
3825 R0PTRTYPE(PPGMPOOLPAGE) pShwPageCR3R0;
3826
3827 /** For saving stack space, the disassembler state is allocated here instead of
3828 * on the stack. */
3829 DISCPUSTATE DisState;
3830
3831 /** Counts the number of times the netware WP0+RO+US hack has been applied. */
3832 uint64_t cNetwareWp0Hacks;
3833
3834 /** Count the number of pgm pool access handler calls. */
3835 uint64_t cPoolAccessHandler;
3836
3837 /** @name Release Statistics
3838 * @{ */
3839 /** The number of times the guest has switched mode since last reset or statistics reset. */
3840 STAMCOUNTER cGuestModeChanges;
3841 /** The number of times the guest has switched mode since last reset or statistics reset. */
3842 STAMCOUNTER cA20Changes;
3843 /** @} */
3844
3845#ifdef VBOX_WITH_STATISTICS /** @todo move this chunk to the heap. */
3846 /** @name Statistics
3847 * @{ */
3848 /** R0: Pointer to the statistics. */
3849 R0PTRTYPE(PGMCPUSTATS *) pStatsR0;
3850 /** R0: Which statistic this \#PF should be attributed to. */
3851 R0PTRTYPE(PSTAMPROFILE) pStatTrap0eAttributionR0;
3852 /** R3: Pointer to the statistics. */
3853 R3PTRTYPE(PGMCPUSTATS *) pStatsR3;
3854 /** Alignment padding. */
3855 RTR3PTR pPaddingR3;
3856 /** @} */
3857#endif /* VBOX_WITH_STATISTICS */
3858} PGMCPU;
3859/** Pointer to the per-cpu PGM data. */
3860typedef PGMCPU *PPGMCPU;
3861
3862
3863/** @name PGM::fSyncFlags Flags
3864 * @note Was part of saved state a long time ago.
3865 * @{
3866 */
3867/* 0 used to be PGM_SYNC_UPDATE_PAGE_BIT_VIRTUAL */
3868/** Always sync CR3. */
3869#define PGM_SYNC_ALWAYS RT_BIT(1)
3870/** Check monitoring on next CR3 (re)load and invalidate page.
3871 * @todo This is obsolete now. Remove after 2.2.0 is branched off. */
3872#define PGM_SYNC_MONITOR_CR3 RT_BIT(2)
3873/** Check guest mapping in SyncCR3. */
3874#define PGM_SYNC_MAP_CR3 RT_BIT(3)
3875/** Clear the page pool (a light weight flush). */
3876#define PGM_SYNC_CLEAR_PGM_POOL_BIT 8
3877#define PGM_SYNC_CLEAR_PGM_POOL RT_BIT(PGM_SYNC_CLEAR_PGM_POOL_BIT)
3878/** @} */
3879
3880
3881/**
3882 * PGM GVM instance data.
3883 */
3884typedef struct PGMR0PERVM
3885{
3886 /** @name PGM Pool related stuff.
3887 * @{ */
3888 /** Critical section for serializing pool growth. */
3889 RTCRITSECT PoolGrowCritSect;
3890 /** The memory objects for the pool pages. */
3891 RTR0MEMOBJ ahPoolMemObjs[(PGMPOOL_IDX_LAST + PGMPOOL_CFG_MAX_GROW - 1) / PGMPOOL_CFG_MAX_GROW];
3892 /** The ring-3 mapping objects for the pool pages. */
3893 RTR0MEMOBJ ahPoolMapObjs[(PGMPOOL_IDX_LAST + PGMPOOL_CFG_MAX_GROW - 1) / PGMPOOL_CFG_MAX_GROW];
3894 /** @} */
3895} PGMR0PERVM;
3896
3897RT_C_DECLS_BEGIN
3898
3899#if defined(VBOX_STRICT) && defined(IN_RING3)
3900int pgmLockDebug(PVMCC pVM, RT_SRC_POS_DECL);
3901# define pgmLock(a_pVM) pgmLockDebug(a_pVM, RT_SRC_POS)
3902#else
3903int pgmLock(PVMCC pVM);
3904#endif
3905void pgmUnlock(PVM pVM);
3906/**
3907 * Asserts that the caller owns the PDM lock.
3908 * This is the internal variant of PGMIsLockOwner.
3909 * @param a_pVM Pointer to the VM.
3910 */
3911#define PGM_LOCK_ASSERT_OWNER(a_pVM) Assert(PDMCritSectIsOwner(&(a_pVM)->pgm.s.CritSectX))
3912/**
3913 * Asserts that the caller owns the PDM lock.
3914 * This is the internal variant of PGMIsLockOwner.
3915 * @param a_pVM Pointer to the VM.
3916 * @param a_pVCpu The current CPU handle.
3917 */
3918#define PGM_LOCK_ASSERT_OWNER_EX(a_pVM, a_pVCpu) Assert(PDMCritSectIsOwnerEx(&(a_pVM)->pgm.s.CritSectX, a_pVCpu))
3919
3920#ifndef PGM_WITHOUT_MAPPINGS
3921int pgmR3MappingsFixInternal(PVM pVM, RTGCPTR GCPtrBase, uint32_t cb);
3922int pgmR3SyncPTResolveConflict(PVM pVM, PPGMMAPPING pMapping, PX86PD pPDSrc, RTGCPTR GCPtrOldMapping);
3923int pgmR3SyncPTResolveConflictPAE(PVM pVM, PPGMMAPPING pMapping, RTGCPTR GCPtrOldMapping);
3924int pgmMapResolveConflicts(PVM pVM);
3925PPGMMAPPING pgmGetMapping(PVM pVM, RTGCPTR GCPtr);
3926DECLCALLBACK(void) pgmR3MapInfo(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
3927#endif /* !PGM_WITHOUT_MAPPINGS */
3928
3929int pgmHandlerPhysicalExCreate(PVMCC pVM, PGMPHYSHANDLERTYPE hType, RTR3PTR pvUserR3, RTR0PTR pvUserR0,
3930 RTRCPTR pvUserRC, R3PTRTYPE(const char *) pszDesc, PPGMPHYSHANDLER *ppPhysHandler);
3931int pgmHandlerPhysicalExDup(PVMCC pVM, PPGMPHYSHANDLER pPhysHandlerSrc, PPGMPHYSHANDLER *ppPhysHandler);
3932int pgmHandlerPhysicalExRegister(PVMCC pVM, PPGMPHYSHANDLER pPhysHandler, RTGCPHYS GCPhys, RTGCPHYS GCPhysLast);
3933int pgmHandlerPhysicalExDeregister(PVMCC pVM, PPGMPHYSHANDLER pPhysHandler, int fRestoreAsRAM);
3934int pgmHandlerPhysicalExDestroy(PVMCC pVM, PPGMPHYSHANDLER pHandler);
3935void pgmR3HandlerPhysicalUpdateAll(PVM pVM);
3936bool pgmHandlerPhysicalIsAll(PVMCC pVM, RTGCPHYS GCPhys);
3937void pgmHandlerPhysicalResetAliasedPage(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhysPage, bool fDoAccounting);
3938DECLCALLBACK(void) pgmR3InfoHandlers(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
3939int pgmR3InitSavedState(PVM pVM, uint64_t cbRam);
3940
3941int pgmPhysAllocPage(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3942int pgmPhysAllocLargePage(PVMCC pVM, RTGCPHYS GCPhys);
3943int pgmPhysRecheckLargePage(PVMCC pVM, RTGCPHYS GCPhys, PPGMPAGE pLargePage);
3944int pgmPhysPageLoadIntoTlb(PVMCC pVM, RTGCPHYS GCPhys);
3945int pgmPhysPageLoadIntoTlbWithPage(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3946void pgmPhysPageMakeWriteMonitoredWritable(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3947int pgmPhysPageMakeWritable(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3948int pgmPhysPageMakeWritableAndMap(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv);
3949int pgmPhysPageMap(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv);
3950int pgmPhysPageMapReadOnly(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void const **ppv);
3951int pgmPhysPageMapByPageID(PVMCC pVM, uint32_t idPage, RTHCPHYS HCPhys, void **ppv);
3952int pgmPhysGCPhys2R3Ptr(PVMCC pVM, RTGCPHYS GCPhys, PRTR3PTR pR3Ptr);
3953int pgmPhysCr3ToHCPtr(PVM pVM, RTGCPHYS GCPhys, PRTR3PTR pR3Ptr);
3954int pgmPhysGCPhys2CCPtrInternalDepr(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv);
3955int pgmPhysGCPhys2CCPtrInternal(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv, PPGMPAGEMAPLOCK pLock);
3956int pgmPhysGCPhys2CCPtrInternalReadOnly(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, const void **ppv, PPGMPAGEMAPLOCK pLock);
3957void pgmPhysReleaseInternalPageMappingLock(PVMCC pVM, PPGMPAGEMAPLOCK pLock);
3958PGM_ALL_CB2_PROTO(FNPGMPHYSHANDLER) pgmPhysRomWriteHandler;
3959#ifndef IN_RING3
3960DECLEXPORT(FNPGMPHYSHANDLER) pgmPhysHandlerRedirectToHC;
3961DECLEXPORT(FNPGMRZPHYSPFHANDLER) pgmPhysPfHandlerRedirectToHC;
3962DECLEXPORT(FNPGMRZPHYSPFHANDLER) pgmPhysRomWritePfHandler;
3963#endif
3964int pgmPhysFreePage(PVM pVM, PGMMFREEPAGESREQ pReq, uint32_t *pcPendingPages, PPGMPAGE pPage, RTGCPHYS GCPhys,
3965 PGMPAGETYPE enmNewType);
3966void pgmPhysInvalidRamRangeTlbs(PVMCC pVM);
3967void pgmPhysInvalidatePageMapTLB(PVMCC pVM);
3968void pgmPhysInvalidatePageMapTLBEntry(PVM pVM, RTGCPHYS GCPhys);
3969PPGMRAMRANGE pgmPhysGetRangeSlow(PVM pVM, RTGCPHYS GCPhys);
3970PPGMRAMRANGE pgmPhysGetRangeAtOrAboveSlow(PVM pVM, RTGCPHYS GCPhys);
3971PPGMPAGE pgmPhysGetPageSlow(PVM pVM, RTGCPHYS GCPhys);
3972int pgmPhysGetPageExSlow(PVM pVM, RTGCPHYS GCPhys, PPPGMPAGE ppPage);
3973int pgmPhysGetPageAndRangeExSlow(PVM pVM, RTGCPHYS GCPhys, PPPGMPAGE ppPage, PPGMRAMRANGE *ppRam);
3974
3975#ifdef IN_RING3
3976void pgmR3PhysRelinkRamRanges(PVM pVM);
3977int pgmR3PhysRamPreAllocate(PVM pVM);
3978int pgmR3PhysRamReset(PVM pVM);
3979int pgmR3PhysRomReset(PVM pVM);
3980int pgmR3PhysRamZeroAll(PVM pVM);
3981int pgmR3PhysChunkMap(PVM pVM, uint32_t idChunk, PPPGMCHUNKR3MAP ppChunk);
3982int pgmR3PhysRamTerm(PVM pVM);
3983void pgmR3PhysRomTerm(PVM pVM);
3984void pgmR3PhysAssertSharedPageChecksums(PVM pVM);
3985
3986int pgmR3PoolInit(PVM pVM);
3987void pgmR3PoolRelocate(PVM pVM);
3988void pgmR3PoolResetUnpluggedCpu(PVM pVM, PVMCPU pVCpu);
3989void pgmR3PoolReset(PVM pVM);
3990void pgmR3PoolClearAll(PVM pVM, bool fFlushRemTlb);
3991DECLCALLBACK(VBOXSTRICTRC) pgmR3PoolClearAllRendezvous(PVM pVM, PVMCPU pVCpu, void *fpvFlushRemTbl);
3992void pgmR3PoolWriteProtectPages(PVM pVM);
3993
3994#endif /* IN_RING3 */
3995int pgmPoolAlloc(PVMCC pVM, RTGCPHYS GCPhys, PGMPOOLKIND enmKind, PGMPOOLACCESS enmAccess, bool fA20Enabled,
3996 uint16_t iUser, uint32_t iUserTable, bool fLockPage, PPPGMPOOLPAGE ppPage);
3997void pgmPoolFree(PVM pVM, RTHCPHYS HCPhys, uint16_t iUser, uint32_t iUserTable);
3998void pgmPoolFreeByPage(PPGMPOOL pPool, PPGMPOOLPAGE pPage, uint16_t iUser, uint32_t iUserTable);
3999int pgmPoolFlushPage(PPGMPOOL pPool, PPGMPOOLPAGE pPage, bool fFlush = true /* DO NOT USE false UNLESS YOU KNOWN WHAT YOU'RE DOING!! */);
4000void pgmPoolFlushPageByGCPhys(PVM pVM, RTGCPHYS GCPhys);
4001PPGMPOOLPAGE pgmPoolGetPage(PPGMPOOL pPool, RTHCPHYS HCPhys);
4002PPGMPOOLPAGE pgmPoolQueryPageForDbg(PPGMPOOL pPool, RTHCPHYS HCPhys);
4003int pgmPoolHCPhys2Ptr(PVM pVM, RTHCPHYS HCPhys, void **ppv);
4004int pgmPoolSyncCR3(PVMCPUCC pVCpu);
4005bool pgmPoolIsDirtyPageSlow(PVM pVM, RTGCPHYS GCPhys);
4006void pgmPoolInvalidateDirtyPage(PVMCC pVM, RTGCPHYS GCPhysPT);
4007int pgmPoolTrackUpdateGCPhys(PVMCC pVM, RTGCPHYS GCPhysPage, PPGMPAGE pPhysPage, bool fFlushPTEs, bool *pfFlushTLBs);
4008void pgmPoolTracDerefGCPhysHint(PPGMPOOL pPool, PPGMPOOLPAGE pPage, RTHCPHYS HCPhys, RTGCPHYS GCPhysHint, uint16_t iPte);
4009uint16_t pgmPoolTrackPhysExtAddref(PVMCC pVM, PPGMPAGE pPhysPage, uint16_t u16, uint16_t iShwPT, uint16_t iPte);
4010void pgmPoolTrackPhysExtDerefGCPhys(PPGMPOOL pPool, PPGMPOOLPAGE pPoolPage, PPGMPAGE pPhysPage, uint16_t iPte);
4011void pgmPoolMonitorChainFlush(PPGMPOOL pPool, PPGMPOOLPAGE pPage);
4012void pgmPoolMonitorModifiedInsert(PPGMPOOL pPool, PPGMPOOLPAGE pPage);
4013PGM_ALL_CB2_PROTO(FNPGMPHYSHANDLER) pgmPoolAccessHandler;
4014#ifndef IN_RING3
4015DECLEXPORT(FNPGMRZPHYSPFHANDLER) pgmRZPoolAccessPfHandler;
4016#endif
4017
4018void pgmPoolAddDirtyPage(PVMCC pVM, PPGMPOOL pPool, PPGMPOOLPAGE pPage);
4019void pgmPoolResetDirtyPages(PVMCC pVM);
4020void pgmPoolResetDirtyPage(PVM pVM, RTGCPTR GCPtrPage);
4021
4022int pgmR3ExitShadowModeBeforePoolFlush(PVMCPU pVCpu);
4023int pgmR3ReEnterShadowModeAfterPoolFlush(PVM pVM, PVMCPU pVCpu);
4024void pgmR3RefreshShadowModeAfterA20Change(PVMCPU pVCpu);
4025
4026#ifndef PGM_WITHOUT_MAPPINGS
4027void pgmMapSetShadowPDEs(PVM pVM, PPGMMAPPING pMap, unsigned iNewPDE);
4028void pgmMapClearShadowPDEs(PVM pVM, PPGMPOOLPAGE pShwPageCR3, PPGMMAPPING pMap, unsigned iOldPDE, bool fDeactivateCR3);
4029int pgmMapActivateCR3(PVM pVM, PPGMPOOLPAGE pShwPageCR3);
4030int pgmMapDeactivateCR3(PVM pVM, PPGMPOOLPAGE pShwPageCR3);
4031#endif
4032
4033int pgmShwMakePageSupervisorAndWritable(PVMCPUCC pVCpu, RTGCPTR GCPtr, bool fBigPage, uint32_t fOpFlags);
4034int pgmShwSyncPaePDPtr(PVMCPUCC pVCpu, RTGCPTR GCPtr, X86PGPAEUINT uGstPdpe, PX86PDPAE *ppPD);
4035int pgmShwSyncNestedPageLocked(PVMCPUCC pVCpu, RTGCPHYS GCPhysFault, uint32_t cPages, PGMMODE enmShwPagingMode);
4036
4037int pgmGstLazyMap32BitPD(PVMCPUCC pVCpu, PX86PD *ppPd);
4038int pgmGstLazyMapPaePDPT(PVMCPUCC pVCpu, PX86PDPT *ppPdpt);
4039int pgmGstLazyMapPaePD(PVMCPUCC pVCpu, uint32_t iPdpt, PX86PDPAE *ppPd);
4040int pgmGstLazyMapPml4(PVMCPUCC pVCpu, PX86PML4 *ppPml4);
4041int pgmGstPtWalk(PVMCPUCC pVCpu, RTGCPTR GCPtr, PPGMPTWALKGST pWalk);
4042int pgmGstPtWalkNext(PVMCPUCC pVCpu, RTGCPTR GCPtr, PPGMPTWALKGST pWalk);
4043
4044# if defined(VBOX_STRICT) && HC_ARCH_BITS == 64 && defined(IN_RING3)
4045FNDBGCCMD pgmR3CmdCheckDuplicatePages;
4046FNDBGCCMD pgmR3CmdShowSharedModules;
4047# endif
4048
4049void pgmLogState(PVM pVM);
4050
4051RT_C_DECLS_END
4052
4053/** @} */
4054
4055#endif /* !VMM_INCLUDED_SRC_include_PGMInternal_h */
4056
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