VirtualBox

source: vbox/trunk/src/VBox/VMM/include/PGMInternal.h@ 86461

Last change on this file since 86461 was 86461, checked in by vboxsync, 4 years ago

VMM/PGM: Don't check for, or even define, PGM_PDFLAGS_MAPPING when PGM_WITHOUT_MAPPINGS is active (the default now).

  • Property svn:eol-style set to native
  • Property svn:keywords set to Id Revision
File size: 182.7 KB
Line 
1/* $Id: PGMInternal.h 86461 2020-10-06 16:40:59Z 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/** Mapping (hypervisor allocated pagetable). */
173#define PGM_PLXFLAGS_MAPPING RT_BIT_64(11)
174/** @} */
175
176/** @name Page directory flags.
177 * These are placed in the three bits available for system programs in
178 * the page directory entries.
179 * @{ */
180/** Indicates the original entry was a big page.
181 * @remarks This is currently only used for statistics and can be recycled. */
182#define PGM_PDFLAGS_BIG_PAGE RT_BIT_64(9)
183#ifndef PGM_WITHOUT_MAPPINGS
184/** Mapping (hypervisor allocated pagetable). */
185# define PGM_PDFLAGS_MAPPING RT_BIT_64(10)
186#endif
187/** Made read-only to facilitate dirty bit tracking. */
188#define PGM_PDFLAGS_TRACK_DIRTY RT_BIT_64(11)
189/** @} */
190
191/** @name Page flags.
192 * These are placed in the three bits available for system programs in
193 * the page entries.
194 * @{ */
195/** Made read-only to facilitate dirty bit tracking. */
196#define PGM_PTFLAGS_TRACK_DIRTY RT_BIT_64(9)
197
198#ifndef PGM_PTFLAGS_CSAM_VALIDATED
199/** Scanned and approved by CSAM (tm).
200 * NOTE: Must be identical to the one defined in CSAMInternal.h!!
201 * @todo Move PGM_PTFLAGS_* and PGM_PDFLAGS_* to VBox/vmm/pgm.h. */
202#define PGM_PTFLAGS_CSAM_VALIDATED RT_BIT_64(11)
203#endif
204
205/** @} */
206
207/** @name Defines used to indicate the shadow and guest paging in the templates.
208 * @{ */
209#define PGM_TYPE_REAL 1
210#define PGM_TYPE_PROT 2
211#define PGM_TYPE_32BIT 3
212#define PGM_TYPE_PAE 4
213#define PGM_TYPE_AMD64 5
214#define PGM_TYPE_NESTED_32BIT 6
215#define PGM_TYPE_NESTED_PAE 7
216#define PGM_TYPE_NESTED_AMD64 8
217#define PGM_TYPE_EPT 9
218#define PGM_TYPE_NONE 10 /**< Dummy shadow paging mode for NEM. */
219#define PGM_TYPE_END (PGM_TYPE_NONE + 1)
220#define PGM_TYPE_FIRST_SHADOW PGM_TYPE_32BIT /**< The first type used by shadow paging. */
221/** @} */
222
223/** Macro for checking if the guest is using paging.
224 * @param uGstType PGM_TYPE_*
225 * @param uShwType PGM_TYPE_*
226 * @remark ASSUMES certain order of the PGM_TYPE_* values.
227 */
228#define PGM_WITH_PAGING(uGstType, uShwType) \
229 ( (uGstType) >= PGM_TYPE_32BIT \
230 && (uShwType) < PGM_TYPE_NESTED_32BIT)
231
232/** Macro for checking if the guest supports the NX bit.
233 * @param uGstType PGM_TYPE_*
234 * @param uShwType PGM_TYPE_*
235 * @remark ASSUMES certain order of the PGM_TYPE_* values.
236 */
237#define PGM_WITH_NX(uGstType, uShwType) \
238 ( (uGstType) >= PGM_TYPE_PAE \
239 && (uShwType) < PGM_TYPE_NESTED_32BIT)
240
241/** Macro for checking for nested or EPT.
242 * @param uType PGM_TYPE_*
243 */
244#define PGM_TYPE_IS_NESTED(uType) \
245 ( (uType) == PGM_TYPE_NESTED_32BIT \
246 || (uType) == PGM_TYPE_NESTED_PAE \
247 || (uType) == PGM_TYPE_NESTED_AMD64)
248
249/** Macro for checking for nested or EPT.
250 * @param uType PGM_TYPE_*
251 */
252#define PGM_TYPE_IS_NESTED_OR_EPT(uType) \
253 ( (uType) == PGM_TYPE_NESTED_32BIT \
254 || (uType) == PGM_TYPE_NESTED_PAE \
255 || (uType) == PGM_TYPE_NESTED_AMD64 \
256 || (uType) == PGM_TYPE_EPT)
257
258
259
260/** @def PGM_HCPHYS_2_PTR
261 * Maps a HC physical page pool address to a virtual address.
262 *
263 * @returns VBox status code.
264 * @param pVM The cross context VM structure.
265 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
266 * @param HCPhys The HC physical address to map to a virtual one.
267 * @param ppv Where to store the virtual address. No need to cast
268 * this.
269 *
270 * @remark There is no need to assert on the result.
271 */
272#define PGM_HCPHYS_2_PTR(pVM, pVCpu, HCPhys, ppv) pgmPoolHCPhys2Ptr(pVM, HCPhys, (void **)(ppv))
273
274/** @def PGM_GCPHYS_2_PTR_V2
275 * Maps a GC physical page address to a virtual address.
276 *
277 * @returns VBox status code.
278 * @param pVM The cross context VM structure.
279 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
280 * @param GCPhys The GC physical address to map to a virtual one.
281 * @param ppv Where to store the virtual address. No need to cast this.
282 *
283 * @remark Use with care as we don't have so much dynamic mapping space in
284 * ring-0 on 32-bit darwin and in RC.
285 * @remark There is no need to assert on the result.
286 */
287#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE_IN_R0
288# define PGM_GCPHYS_2_PTR_V2(pVM, pVCpu, GCPhys, ppv) \
289 pgmRZDynMapGCPageV2Inlined(pVM, pVCpu, GCPhys, (void **)(ppv) RTLOG_COMMA_SRC_POS)
290#else
291# define PGM_GCPHYS_2_PTR_V2(pVM, pVCpu, GCPhys, ppv) \
292 pgmPhysGCPhys2R3Ptr(pVM, GCPhys, (PRTR3PTR)(ppv)) /** @todo this isn't asserting! */
293#endif
294
295/** @def PGM_GCPHYS_2_PTR
296 * Maps a GC physical page address to a virtual address.
297 *
298 * @returns VBox status code.
299 * @param pVM The cross context VM structure.
300 * @param GCPhys The GC physical address to map to a virtual one.
301 * @param ppv Where to store the virtual address. No need to cast this.
302 *
303 * @remark Use with care as we don't have so much dynamic mapping space in
304 * ring-0 on 32-bit darwin and in RC.
305 * @remark There is no need to assert on the result.
306 */
307#define PGM_GCPHYS_2_PTR(pVM, GCPhys, ppv) PGM_GCPHYS_2_PTR_V2(pVM, VMMGetCpu(pVM), GCPhys, ppv)
308
309/** @def PGM_GCPHYS_2_PTR_BY_VMCPU
310 * Maps a GC physical page address to a virtual address.
311 *
312 * @returns VBox status code.
313 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
314 * @param GCPhys The GC physical address to map to a virtual one.
315 * @param ppv Where to store the virtual address. No need to cast this.
316 *
317 * @remark Use with care as we don't have so much dynamic mapping space in
318 * ring-0 on 32-bit darwin and in RC.
319 * @remark There is no need to assert on the result.
320 */
321#define PGM_GCPHYS_2_PTR_BY_VMCPU(pVCpu, GCPhys, ppv) PGM_GCPHYS_2_PTR_V2((pVCpu)->CTX_SUFF(pVM), pVCpu, GCPhys, ppv)
322
323/** @def PGM_GCPHYS_2_PTR_EX
324 * Maps a unaligned GC physical page address to a virtual address.
325 *
326 * @returns VBox status code.
327 * @param pVM The cross context VM structure.
328 * @param GCPhys The GC physical address to map to a virtual one.
329 * @param ppv Where to store the virtual address. No need to cast this.
330 *
331 * @remark Use with care as we don't have so much dynamic mapping space in
332 * ring-0 on 32-bit darwin and in RC.
333 * @remark There is no need to assert on the result.
334 */
335#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE_IN_R0
336# define PGM_GCPHYS_2_PTR_EX(pVM, GCPhys, ppv) \
337 pgmRZDynMapGCPageOffInlined(VMMGetCpu(pVM), GCPhys, (void **)(ppv) RTLOG_COMMA_SRC_POS)
338#else
339# define PGM_GCPHYS_2_PTR_EX(pVM, GCPhys, ppv) \
340 pgmPhysGCPhys2R3Ptr(pVM, GCPhys, (PRTR3PTR)(ppv)) /** @todo this isn't asserting! */
341#endif
342
343/** @def PGM_DYNMAP_UNUSED_HINT
344 * Hints to the dynamic mapping code in RC and R0/darwin that the specified page
345 * is no longer used.
346 *
347 * For best effect only apply this to the page that was mapped most recently.
348 *
349 * @param pVCpu The cross context virtual CPU structure of the calling EMT.
350 * @param pvPage The pool page.
351 */
352#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE_IN_R0
353# ifdef LOG_ENABLED
354# define PGM_DYNMAP_UNUSED_HINT(pVCpu, pvPage) pgmRZDynMapUnusedHint(pVCpu, pvPage, RT_SRC_POS)
355# else
356# define PGM_DYNMAP_UNUSED_HINT(pVCpu, pvPage) pgmRZDynMapUnusedHint(pVCpu, pvPage)
357# endif
358#else
359# define PGM_DYNMAP_UNUSED_HINT(pVCpu, pvPage) do {} while (0)
360#endif
361
362/** @def PGM_DYNMAP_UNUSED_HINT_VM
363 * Hints to the dynamic mapping code in RC and R0/darwin that the specified page
364 * is no longer used.
365 *
366 * For best effect only apply this to the page that was mapped most recently.
367 *
368 * @param pVM The cross context VM structure.
369 * @param pvPage The pool page.
370 */
371#define PGM_DYNMAP_UNUSED_HINT_VM(pVM, pvPage) PGM_DYNMAP_UNUSED_HINT(VMMGetCpu(pVM), pvPage)
372
373
374/** @def PGM_INVL_PG
375 * Invalidates a page.
376 *
377 * @param pVCpu The cross context virtual CPU structure.
378 * @param GCVirt The virtual address of the page to invalidate.
379 */
380#ifdef IN_RING0
381# define PGM_INVL_PG(pVCpu, GCVirt) HMInvalidatePage(pVCpu, (RTGCPTR)(GCVirt))
382#elif defined(IN_RING3)
383# define PGM_INVL_PG(pVCpu, GCVirt) HMInvalidatePage(pVCpu, (RTGCPTR)(GCVirt))
384#else
385# error "Not IN_RING0 or IN_RING3!"
386#endif
387
388/** @def PGM_INVL_PG_ALL_VCPU
389 * Invalidates a page on all VCPUs
390 *
391 * @param pVM The cross context VM structure.
392 * @param GCVirt The virtual address of the page to invalidate.
393 */
394#ifdef IN_RING0
395# define PGM_INVL_PG_ALL_VCPU(pVM, GCVirt) HMInvalidatePageOnAllVCpus(pVM, (RTGCPTR)(GCVirt))
396#else
397# define PGM_INVL_PG_ALL_VCPU(pVM, GCVirt) HMInvalidatePageOnAllVCpus(pVM, (RTGCPTR)(GCVirt))
398#endif
399
400/** @def PGM_INVL_BIG_PG
401 * Invalidates a 4MB page directory entry.
402 *
403 * @param pVCpu The cross context virtual CPU structure.
404 * @param GCVirt The virtual address within the page directory to invalidate.
405 */
406#ifdef IN_RING0
407# define PGM_INVL_BIG_PG(pVCpu, GCVirt) HMFlushTlb(pVCpu)
408#else
409# define PGM_INVL_BIG_PG(pVCpu, GCVirt) HMFlushTlb(pVCpu)
410#endif
411
412/** @def PGM_INVL_VCPU_TLBS()
413 * Invalidates the TLBs of the specified VCPU
414 *
415 * @param pVCpu The cross context virtual CPU structure.
416 */
417#ifdef IN_RING0
418# define PGM_INVL_VCPU_TLBS(pVCpu) HMFlushTlb(pVCpu)
419#else
420# define PGM_INVL_VCPU_TLBS(pVCpu) HMFlushTlb(pVCpu)
421#endif
422
423/** @def PGM_INVL_ALL_VCPU_TLBS()
424 * Invalidates the TLBs of all VCPUs
425 *
426 * @param pVM The cross context VM structure.
427 */
428#ifdef IN_RING0
429# define PGM_INVL_ALL_VCPU_TLBS(pVM) HMFlushTlbOnAllVCpus(pVM)
430#else
431# define PGM_INVL_ALL_VCPU_TLBS(pVM) HMFlushTlbOnAllVCpus(pVM)
432#endif
433
434
435/** @name Safer Shadow PAE PT/PTE
436 * For helping avoid misinterpreting invalid PAE/AMD64 page table entries as
437 * present.
438 *
439 * @{
440 */
441#if 1
442/**
443 * For making sure that u1Present and X86_PTE_P checks doesn't mistake
444 * invalid entries for present.
445 * @sa X86PTEPAE.
446 */
447typedef union PGMSHWPTEPAE
448{
449 /** Unsigned integer view */
450 X86PGPAEUINT uCareful;
451 /* Not other views. */
452} PGMSHWPTEPAE;
453
454# define PGMSHWPTEPAE_IS_P(Pte) ( ((Pte).uCareful & (X86_PTE_P | X86_PTE_PAE_MBZ_MASK_NX)) == X86_PTE_P )
455# define PGMSHWPTEPAE_IS_RW(Pte) ( !!((Pte).uCareful & X86_PTE_RW))
456# define PGMSHWPTEPAE_IS_US(Pte) ( !!((Pte).uCareful & X86_PTE_US))
457# define PGMSHWPTEPAE_IS_A(Pte) ( !!((Pte).uCareful & X86_PTE_A))
458# define PGMSHWPTEPAE_IS_D(Pte) ( !!((Pte).uCareful & X86_PTE_D))
459# define PGMSHWPTEPAE_IS_TRACK_DIRTY(Pte) ( !!((Pte).uCareful & PGM_PTFLAGS_TRACK_DIRTY) )
460# 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) )
461# define PGMSHWPTEPAE_GET_LOG(Pte) ( (Pte).uCareful )
462# define PGMSHWPTEPAE_GET_HCPHYS(Pte) ( (Pte).uCareful & X86_PTE_PAE_PG_MASK )
463# define PGMSHWPTEPAE_GET_U(Pte) ( (Pte).uCareful ) /**< Use with care. */
464# define PGMSHWPTEPAE_SET(Pte, uVal) do { (Pte).uCareful = (uVal); } while (0)
465# define PGMSHWPTEPAE_SET2(Pte, Pte2) do { (Pte).uCareful = (Pte2).uCareful; } while (0)
466# define PGMSHWPTEPAE_ATOMIC_SET(Pte, uVal) do { ASMAtomicWriteU64(&(Pte).uCareful, (uVal)); } while (0)
467# define PGMSHWPTEPAE_ATOMIC_SET2(Pte, Pte2) do { ASMAtomicWriteU64(&(Pte).uCareful, (Pte2).uCareful); } while (0)
468# define PGMSHWPTEPAE_SET_RO(Pte) do { (Pte).uCareful &= ~(X86PGPAEUINT)X86_PTE_RW; } while (0)
469# define PGMSHWPTEPAE_SET_RW(Pte) do { (Pte).uCareful |= X86_PTE_RW; } while (0)
470
471/**
472 * For making sure that u1Present and X86_PTE_P checks doesn't mistake
473 * invalid entries for present.
474 * @sa X86PTPAE.
475 */
476typedef struct PGMSHWPTPAE
477{
478 PGMSHWPTEPAE a[X86_PG_PAE_ENTRIES];
479} PGMSHWPTPAE;
480
481#else
482typedef X86PTEPAE PGMSHWPTEPAE;
483typedef X86PTPAE PGMSHWPTPAE;
484# define PGMSHWPTEPAE_IS_P(Pte) ( (Pte).n.u1Present )
485# define PGMSHWPTEPAE_IS_RW(Pte) ( (Pte).n.u1Write )
486# define PGMSHWPTEPAE_IS_US(Pte) ( (Pte).n.u1User )
487# define PGMSHWPTEPAE_IS_A(Pte) ( (Pte).n.u1Accessed )
488# define PGMSHWPTEPAE_IS_D(Pte) ( (Pte).n.u1Dirty )
489# define PGMSHWPTEPAE_IS_TRACK_DIRTY(Pte) ( !!((Pte).u & PGM_PTFLAGS_TRACK_DIRTY) )
490# define PGMSHWPTEPAE_IS_P_RW(Pte) ( ((Pte).u & (X86_PTE_P | X86_PTE_RW)) == (X86_PTE_P | X86_PTE_RW) )
491# define PGMSHWPTEPAE_GET_LOG(Pte) ( (Pte).u )
492# define PGMSHWPTEPAE_GET_HCPHYS(Pte) ( (Pte).u & X86_PTE_PAE_PG_MASK )
493# define PGMSHWPTEPAE_GET_U(Pte) ( (Pte).u ) /**< Use with care. */
494# define PGMSHWPTEPAE_SET(Pte, uVal) do { (Pte).u = (uVal); } while (0)
495# define PGMSHWPTEPAE_SET2(Pte, Pte2) do { (Pte).u = (Pte2).u; } while (0)
496# define PGMSHWPTEPAE_ATOMIC_SET(Pte, uVal) do { ASMAtomicWriteU64(&(Pte).u, (uVal)); } while (0)
497# define PGMSHWPTEPAE_ATOMIC_SET2(Pte, Pte2) do { ASMAtomicWriteU64(&(Pte).u, (Pte2).u); } while (0)
498# define PGMSHWPTEPAE_SET_RO(Pte) do { (Pte).u &= ~(X86PGPAEUINT)X86_PTE_RW; } while (0)
499# define PGMSHWPTEPAE_SET_RW(Pte) do { (Pte).u |= X86_PTE_RW; } while (0)
500
501#endif
502
503/** Pointer to a shadow PAE PTE. */
504typedef PGMSHWPTEPAE *PPGMSHWPTEPAE;
505/** Pointer to a const shadow PAE PTE. */
506typedef PGMSHWPTEPAE const *PCPGMSHWPTEPAE;
507
508/** Pointer to a shadow PAE page table. */
509typedef PGMSHWPTPAE *PPGMSHWPTPAE;
510/** Pointer to a const shadow PAE page table. */
511typedef PGMSHWPTPAE const *PCPGMSHWPTPAE;
512/** @} */
513
514#ifndef PGM_WITHOUT_MAPPINGS
515
516/** Size of the GCPtrConflict array in PGMMAPPING.
517 * @remarks Must be a power of two. */
518# define PGMMAPPING_CONFLICT_MAX 8
519
520/**
521 * Structure for tracking GC Mappings.
522 *
523 * This structure is used by linked list in both GC and HC.
524 */
525typedef struct PGMMAPPING
526{
527 /** Pointer to next entry. */
528 R3PTRTYPE(struct PGMMAPPING *) pNextR3;
529 /** Pointer to next entry. */
530 R0PTRTYPE(struct PGMMAPPING *) pNextR0;
531 /** Indicate whether this entry is finalized. */
532 bool fFinalized;
533 bool afPadding[7];
534 /** Start Virtual address. */
535 RTGCPTR GCPtr;
536 /** Last Virtual address (inclusive). */
537 RTGCPTR GCPtrLast;
538 /** Range size (bytes). */
539 RTGCPTR cb;
540 /** Pointer to relocation callback function. */
541 R3PTRTYPE(PFNPGMRELOCATE) pfnRelocate;
542 /** User argument to the callback. */
543 R3PTRTYPE(void *) pvUser;
544 /** Mapping description / name. For easing debugging. */
545 R3PTRTYPE(const char *) pszDesc;
546 /** Last 8 addresses that caused conflicts. */
547 RTGCPTR aGCPtrConflicts[PGMMAPPING_CONFLICT_MAX];
548 /** Number of conflicts for this hypervisor mapping. */
549 uint32_t cConflicts;
550 /** Number of page tables. */
551 uint32_t cPTs;
552
553 /** Array of page table mapping data. Each entry
554 * describes one page table. The array can be longer
555 * than the declared length.
556 */
557 struct
558 {
559 /** The HC physical address of the page table. */
560 RTHCPHYS HCPhysPT;
561 /** The HC physical address of the first PAE page table. */
562 RTHCPHYS HCPhysPaePT0;
563 /** The HC physical address of the second PAE page table. */
564 RTHCPHYS HCPhysPaePT1;
565 /** The HC virtual address of the 32-bit page table. */
566 R3PTRTYPE(PX86PT) pPTR3;
567 /** The HC virtual address of the two PAE page table. (i.e 1024 entries instead of 512) */
568 R3PTRTYPE(PPGMSHWPTPAE) paPaePTsR3;
569 /** The R0 virtual address of the 32-bit page table. */
570 R0PTRTYPE(PX86PT) pPTR0;
571 /** The R0 virtual address of the two PAE page table. */
572 R0PTRTYPE(PPGMSHWPTPAE) paPaePTsR0;
573 } aPTs[1];
574} PGMMAPPING;
575/** Pointer to structure for tracking GC Mappings. */
576typedef struct PGMMAPPING *PPGMMAPPING;
577
578#endif /* !PGM_WITHOUT_MAPPINGS */
579
580
581/**
582 * Physical page access handler type registration.
583 */
584typedef struct PGMPHYSHANDLERTYPEINT
585{
586 /** Number of references. */
587 uint32_t volatile cRefs;
588 /** Magic number (PGMPHYSHANDLERTYPEINT_MAGIC). */
589 uint32_t u32Magic;
590 /** Link of handler types anchored in PGMTREES::HeadPhysHandlerTypes. */
591 RTLISTOFF32NODE ListNode;
592 /** The kind of accesses we're handling. */
593 PGMPHYSHANDLERKIND enmKind;
594 /** The PGM_PAGE_HNDL_PHYS_STATE_XXX value corresponding to enmKind. */
595 uint32_t uState;
596 /** Pointer to R3 callback function. */
597 R3PTRTYPE(PFNPGMPHYSHANDLER) pfnHandlerR3;
598 /** Pointer to R0 callback function. */
599 R0PTRTYPE(PFNPGMPHYSHANDLER) pfnHandlerR0;
600 /** Pointer to R0 callback function for \#PFs. */
601 R0PTRTYPE(PFNPGMRZPHYSPFHANDLER) pfnPfHandlerR0;
602 /** Description / Name. For easing debugging. */
603 R3PTRTYPE(const char *) pszDesc;
604} PGMPHYSHANDLERTYPEINT;
605/** Pointer to a physical access handler type registration. */
606typedef PGMPHYSHANDLERTYPEINT *PPGMPHYSHANDLERTYPEINT;
607/** Magic value for the physical handler callbacks (Robert A. Heinlein). */
608#define PGMPHYSHANDLERTYPEINT_MAGIC UINT32_C(0x19070707)
609/** Magic value for the physical handler callbacks. */
610#define PGMPHYSHANDLERTYPEINT_MAGIC_DEAD UINT32_C(0x19880508)
611
612/**
613 * Converts a handle to a pointer.
614 * @returns PPGMPHYSHANDLERTYPEINT
615 * @param a_pVM The cross context VM structure.
616 * @param a_hType Physical access handler type handle.
617 */
618#define PGMPHYSHANDLERTYPEINT_FROM_HANDLE(a_pVM, a_hType) ((PPGMPHYSHANDLERTYPEINT)MMHyperHeapOffsetToPtr(a_pVM, a_hType))
619
620
621/**
622 * Physical page access handler structure.
623 *
624 * This is used to keep track of physical address ranges
625 * which are being monitored in some kind of way.
626 */
627typedef struct PGMPHYSHANDLER
628{
629 AVLROGCPHYSNODECORE Core;
630 /** Number of pages to update. */
631 uint32_t cPages;
632 /** Set if we have pages that have been aliased. */
633 uint32_t cAliasedPages;
634 /** Set if we have pages that have temporarily been disabled. */
635 uint32_t cTmpOffPages;
636 /** Registered handler type handle (heap offset). */
637 PGMPHYSHANDLERTYPE hType;
638 /** User argument for R3 handlers. */
639 R3PTRTYPE(void *) pvUserR3;
640 /** User argument for R0 handlers. */
641 R0PTRTYPE(void *) pvUserR0;
642 /** Description / Name. For easing debugging. */
643 R3PTRTYPE(const char *) pszDesc;
644#ifdef VBOX_WITH_STATISTICS
645 /** Profiling of this handler. */
646 STAMPROFILE Stat;
647#endif
648} PGMPHYSHANDLER;
649/** Pointer to a physical page access handler structure. */
650typedef PGMPHYSHANDLER *PPGMPHYSHANDLER;
651
652/**
653 * Gets the type record for a physical handler (no reference added).
654 * @returns PPGMPHYSHANDLERTYPEINT
655 * @param a_pVM The cross context VM structure.
656 * @param a_pPhysHandler Pointer to the physical handler structure
657 * (PGMPHYSHANDLER).
658 */
659#define PGMPHYSHANDLER_GET_TYPE(a_pVM, a_pPhysHandler) PGMPHYSHANDLERTYPEINT_FROM_HANDLE(a_pVM, (a_pPhysHandler)->hType)
660
661
662/**
663 * A Physical Guest Page tracking structure.
664 *
665 * The format of this structure is complicated because we have to fit a lot
666 * of information into as few bits as possible. The format is also subject
667 * to change (there is one coming up soon). Which means that for we'll be
668 * using PGM_PAGE_GET_*, PGM_PAGE_IS_ and PGM_PAGE_SET_* macros for *all*
669 * accesses to the structure.
670 */
671typedef union PGMPAGE
672{
673 /** Structured view. */
674 struct
675 {
676 /** 1:0 - The physical handler state (PGM_PAGE_HNDL_PHYS_STATE_*). */
677 uint64_t u2HandlerPhysStateY : 2;
678 /** 3:2 - Paging structure needed to map the page
679 * (PGM_PAGE_PDE_TYPE_*). */
680 uint64_t u2PDETypeY : 2;
681 /** 4 - Unused (was used by FTE for dirty tracking). */
682 uint64_t fUnused1 : 1;
683 /** 5 - Flag indicating that a write monitored page was written to
684 * when set. */
685 uint64_t fWrittenToY : 1;
686 /** 7:6 - Unused. */
687 uint64_t u2Unused0 : 2;
688 /** 9:8 - Unused (was used by PGM_PAGE_HNDL_VIRT_STATE_*). */
689 uint64_t u2Unused1 : 2;
690 /** 11:10 - NEM state bits. */
691 uint64_t u2NemStateY : 2;
692 /** 12:48 - The host physical frame number (shift left to get the
693 * address). */
694 uint64_t HCPhysFN : 36;
695 /** 50:48 - The page state. */
696 uint64_t uStateY : 3;
697 /** 51:53 - The page type (PGMPAGETYPE). */
698 uint64_t uTypeY : 3;
699 /** 63:54 - PTE index for usage tracking (page pool). */
700 uint64_t u10PteIdx : 10;
701
702 /** The GMM page ID.
703 * @remarks In the current implementation, MMIO2 and pages aliased to
704 * MMIO2 pages will be exploiting this field to calculate the
705 * ring-3 mapping address corresponding to the page.
706 * Later we may consider including MMIO2 management into GMM. */
707 uint32_t idPage;
708 /** Usage tracking (page pool). */
709 uint16_t u16TrackingY;
710 /** The number of read locks on this page. */
711 uint8_t cReadLocksY;
712 /** The number of write locks on this page. */
713 uint8_t cWriteLocksY;
714 } s;
715
716 /** 64-bit integer view. */
717 uint64_t au64[2];
718 /** 16-bit view. */
719 uint32_t au32[4];
720 /** 16-bit view. */
721 uint16_t au16[8];
722 /** 8-bit view. */
723 uint8_t au8[16];
724} PGMPAGE;
725AssertCompileSize(PGMPAGE, 16);
726/** Pointer to a physical guest page. */
727typedef PGMPAGE *PPGMPAGE;
728/** Pointer to a const physical guest page. */
729typedef const PGMPAGE *PCPGMPAGE;
730/** Pointer to a physical guest page pointer. */
731typedef PPGMPAGE *PPPGMPAGE;
732
733
734/**
735 * Clears the page structure.
736 * @param a_pPage Pointer to the physical guest page tracking structure.
737 */
738#define PGM_PAGE_CLEAR(a_pPage) \
739 do { \
740 (a_pPage)->au64[0] = 0; \
741 (a_pPage)->au64[1] = 0; \
742 } while (0)
743
744/**
745 * Initializes the page structure.
746 * @param a_pPage Pointer to the physical guest page tracking structure.
747 * @param a_HCPhys The host physical address of the page.
748 * @param a_idPage The (GMM) page ID of the page.
749 * @param a_uType The page type (PGMPAGETYPE).
750 * @param a_uState The page state (PGM_PAGE_STATE_XXX).
751 */
752#define PGM_PAGE_INIT(a_pPage, a_HCPhys, a_idPage, a_uType, a_uState) \
753 do { \
754 RTHCPHYS SetHCPhysTmp = (a_HCPhys); \
755 AssertFatal(!(SetHCPhysTmp & ~UINT64_C(0x0000fffffffff000))); \
756 (a_pPage)->au64[0] = SetHCPhysTmp; \
757 (a_pPage)->au64[1] = 0; \
758 (a_pPage)->s.idPage = (a_idPage); \
759 (a_pPage)->s.uStateY = (a_uState); \
760 (a_pPage)->s.uTypeY = (a_uType); \
761 } while (0)
762
763/**
764 * Initializes the page structure of a ZERO page.
765 * @param a_pPage Pointer to the physical guest page tracking structure.
766 * @param a_pVM The VM handle (for getting the zero page address).
767 * @param a_uType The page type (PGMPAGETYPE).
768 */
769#define PGM_PAGE_INIT_ZERO(a_pPage, a_pVM, a_uType) \
770 PGM_PAGE_INIT((a_pPage), (a_pVM)->pgm.s.HCPhysZeroPg, NIL_GMM_PAGEID, (a_uType), PGM_PAGE_STATE_ZERO)
771
772
773/** @name The Page state, PGMPAGE::uStateY.
774 * @{ */
775/** The zero page.
776 * This is a per-VM page that's never ever mapped writable. */
777#define PGM_PAGE_STATE_ZERO 0U
778/** A allocated page.
779 * This is a per-VM page allocated from the page pool (or wherever
780 * we get MMIO2 pages from if the type is MMIO2).
781 */
782#define PGM_PAGE_STATE_ALLOCATED 1U
783/** A allocated page that's being monitored for writes.
784 * The shadow page table mappings are read-only. When a write occurs, the
785 * fWrittenTo member is set, the page remapped as read-write and the state
786 * moved back to allocated. */
787#define PGM_PAGE_STATE_WRITE_MONITORED 2U
788/** The page is shared, aka. copy-on-write.
789 * This is a page that's shared with other VMs. */
790#define PGM_PAGE_STATE_SHARED 3U
791/** The page is ballooned, so no longer available for this VM. */
792#define PGM_PAGE_STATE_BALLOONED 4U
793/** @} */
794
795
796/** Asserts lock ownership in some of the PGM_PAGE_XXX macros. */
797#if defined(VBOX_STRICT) && 0 /** @todo triggers in pgmRZDynMapGCPageV2Inlined */
798# define PGM_PAGE_ASSERT_LOCK(a_pVM) PGM_LOCK_ASSERT_OWNER(a_pVM)
799#else
800# define PGM_PAGE_ASSERT_LOCK(a_pVM) do { } while (0)
801#endif
802
803/**
804 * Gets the page state.
805 * @returns page state (PGM_PAGE_STATE_*).
806 * @param a_pPage Pointer to the physical guest page tracking structure.
807 *
808 * @remarks See PGM_PAGE_GET_HCPHYS_NA for remarks about GCC and strict
809 * builds.
810 */
811#define PGM_PAGE_GET_STATE_NA(a_pPage) ( (a_pPage)->s.uStateY )
812#if defined(__GNUC__) && defined(VBOX_STRICT)
813# define PGM_PAGE_GET_STATE(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_STATE_NA(a_pPage); })
814#else
815# define PGM_PAGE_GET_STATE PGM_PAGE_GET_STATE_NA
816#endif
817
818/**
819 * Sets the page state.
820 * @param a_pVM The VM handle, only used for lock ownership assertions.
821 * @param a_pPage Pointer to the physical guest page tracking structure.
822 * @param a_uState The new page state.
823 */
824#define PGM_PAGE_SET_STATE(a_pVM, a_pPage, a_uState) \
825 do { (a_pPage)->s.uStateY = (a_uState); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
826
827
828/**
829 * Gets the host physical address of the guest page.
830 * @returns host physical address (RTHCPHYS).
831 * @param a_pPage Pointer to the physical guest page tracking structure.
832 *
833 * @remarks In strict builds on gcc platforms, this macro will make some ugly
834 * assumption about a valid pVM variable/parameter being in the
835 * current context. It will use this pVM variable to assert that the
836 * PGM lock is held. Use the PGM_PAGE_GET_HCPHYS_NA in contexts where
837 * pVM is not around.
838 */
839#if 0
840# define PGM_PAGE_GET_HCPHYS_NA(a_pPage) ( (a_pPage)->s.HCPhysFN << 12 )
841# define PGM_PAGE_GET_HCPHYS PGM_PAGE_GET_HCPHYS_NA
842#else
843# define PGM_PAGE_GET_HCPHYS_NA(a_pPage) ( (a_pPage)->au64[0] & UINT64_C(0x0000fffffffff000) )
844# if defined(__GNUC__) && defined(VBOX_STRICT)
845# define PGM_PAGE_GET_HCPHYS(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_HCPHYS_NA(a_pPage); })
846# else
847# define PGM_PAGE_GET_HCPHYS PGM_PAGE_GET_HCPHYS_NA
848# endif
849#endif
850
851/**
852 * Sets the host physical address of the guest page.
853 *
854 * @param a_pVM The VM handle, only used for lock ownership assertions.
855 * @param a_pPage Pointer to the physical guest page tracking structure.
856 * @param a_HCPhys The new host physical address.
857 */
858#define PGM_PAGE_SET_HCPHYS(a_pVM, a_pPage, a_HCPhys) \
859 do { \
860 RTHCPHYS const SetHCPhysTmp = (a_HCPhys); \
861 AssertFatal(!(SetHCPhysTmp & ~UINT64_C(0x0000fffffffff000))); \
862 (a_pPage)->s.HCPhysFN = SetHCPhysTmp >> 12; \
863 PGM_PAGE_ASSERT_LOCK(a_pVM); \
864 } while (0)
865
866/**
867 * Get the Page ID.
868 * @returns The Page ID; NIL_GMM_PAGEID if it's a ZERO page.
869 * @param a_pPage Pointer to the physical guest page tracking structure.
870 */
871#define PGM_PAGE_GET_PAGEID(a_pPage) ( (uint32_t)(a_pPage)->s.idPage )
872
873/**
874 * Sets the Page ID.
875 * @param a_pVM The VM handle, only used for lock ownership assertions.
876 * @param a_pPage Pointer to the physical guest page tracking structure.
877 * @param a_idPage The new page ID.
878 */
879#define PGM_PAGE_SET_PAGEID(a_pVM, a_pPage, a_idPage) \
880 do { \
881 (a_pPage)->s.idPage = (a_idPage); \
882 PGM_PAGE_ASSERT_LOCK(a_pVM); \
883 } while (0)
884
885/**
886 * Get the Chunk ID.
887 * @returns The Chunk ID; NIL_GMM_CHUNKID if it's a ZERO page.
888 * @param a_pPage Pointer to the physical guest page tracking structure.
889 */
890#define PGM_PAGE_GET_CHUNKID(a_pPage) ( PGM_PAGE_GET_PAGEID(a_pPage) >> GMM_CHUNKID_SHIFT )
891
892/**
893 * Get the index of the page within the allocation chunk.
894 * @returns The page index.
895 * @param a_pPage Pointer to the physical guest page tracking structure.
896 */
897#define PGM_PAGE_GET_PAGE_IN_CHUNK(a_pPage) ( PGM_PAGE_GET_PAGEID(a_pPage) & GMM_PAGEID_IDX_MASK )
898
899/**
900 * Gets the page type.
901 * @returns The page type.
902 * @param a_pPage Pointer to the physical guest page tracking structure.
903 *
904 * @remarks See PGM_PAGE_GET_HCPHYS_NA for remarks about GCC and strict
905 * builds.
906 */
907#define PGM_PAGE_GET_TYPE_NA(a_pPage) ( (a_pPage)->s.uTypeY )
908#if defined(__GNUC__) && defined(VBOX_STRICT)
909# define PGM_PAGE_GET_TYPE(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_TYPE_NA(a_pPage); })
910#else
911# define PGM_PAGE_GET_TYPE PGM_PAGE_GET_TYPE_NA
912#endif
913
914/**
915 * Sets the page type.
916 *
917 * @param a_pVM The VM handle, only used for lock ownership assertions.
918 * @param a_pPage Pointer to the physical guest page tracking structure.
919 * @param a_enmType The new page type (PGMPAGETYPE).
920 */
921#define PGM_PAGE_SET_TYPE(a_pVM, a_pPage, a_enmType) \
922 do { (a_pPage)->s.uTypeY = (a_enmType); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
923
924/**
925 * Gets the page table index
926 * @returns The page table index.
927 * @param a_pPage Pointer to the physical guest page tracking structure.
928 */
929#define PGM_PAGE_GET_PTE_INDEX(a_pPage) ( (a_pPage)->s.u10PteIdx )
930
931/**
932 * Sets the page table index.
933 * @param a_pVM The VM handle, only used for lock ownership assertions.
934 * @param a_pPage Pointer to the physical guest page tracking structure.
935 * @param a_iPte New page table index.
936 */
937#define PGM_PAGE_SET_PTE_INDEX(a_pVM, a_pPage, a_iPte) \
938 do { (a_pPage)->s.u10PteIdx = (a_iPte); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
939
940/**
941 * Checks if the page is marked for MMIO, no MMIO2 aliasing.
942 * @returns true/false.
943 * @param a_pPage Pointer to the physical guest page tracking structure.
944 */
945#define PGM_PAGE_IS_MMIO(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO )
946
947/**
948 * Checks if the page is marked for MMIO, including both aliases.
949 * @returns true/false.
950 * @param a_pPage Pointer to the physical guest page tracking structure.
951 */
952#define PGM_PAGE_IS_MMIO_OR_ALIAS(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO \
953 || (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO2_ALIAS_MMIO \
954 || (a_pPage)->s.uTypeY == PGMPAGETYPE_SPECIAL_ALIAS_MMIO \
955 )
956
957/**
958 * Checks if the page is marked for MMIO, including special aliases.
959 * @returns true/false.
960 * @param a_pPage Pointer to the physical guest page tracking structure.
961 */
962#define PGM_PAGE_IS_MMIO_OR_SPECIAL_ALIAS(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_MMIO \
963 || (a_pPage)->s.uTypeY == PGMPAGETYPE_SPECIAL_ALIAS_MMIO )
964
965/**
966 * Checks if the page is a special aliased MMIO page.
967 * @returns true/false.
968 * @param a_pPage Pointer to the physical guest page tracking structure.
969 */
970#define PGM_PAGE_IS_SPECIAL_ALIAS_MMIO(a_pPage) ( (a_pPage)->s.uTypeY == PGMPAGETYPE_SPECIAL_ALIAS_MMIO )
971
972/**
973 * Checks if the page is backed by the ZERO page.
974 * @returns true/false.
975 * @param a_pPage Pointer to the physical guest page tracking structure.
976 */
977#define PGM_PAGE_IS_ZERO(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_ZERO )
978
979/**
980 * Checks if the page is backed by a SHARED page.
981 * @returns true/false.
982 * @param a_pPage Pointer to the physical guest page tracking structure.
983 */
984#define PGM_PAGE_IS_SHARED(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_SHARED )
985
986/**
987 * Checks if the page is ballooned.
988 * @returns true/false.
989 * @param a_pPage Pointer to the physical guest page tracking structure.
990 */
991#define PGM_PAGE_IS_BALLOONED(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_BALLOONED )
992
993/**
994 * Checks if the page is allocated.
995 * @returns true/false.
996 * @param a_pPage Pointer to the physical guest page tracking structure.
997 */
998#define PGM_PAGE_IS_ALLOCATED(a_pPage) ( (a_pPage)->s.uStateY == PGM_PAGE_STATE_ALLOCATED )
999
1000/**
1001 * Marks the page as written to (for GMM change monitoring).
1002 * @param a_pVM The VM handle, only used for lock ownership assertions.
1003 * @param a_pPage Pointer to the physical guest page tracking structure.
1004 */
1005#define PGM_PAGE_SET_WRITTEN_TO(a_pVM, a_pPage) \
1006 do { (a_pPage)->s.fWrittenToY = 1; PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1007
1008/**
1009 * Clears the written-to indicator.
1010 * @param a_pVM The VM handle, only used for lock ownership assertions.
1011 * @param a_pPage Pointer to the physical guest page tracking structure.
1012 */
1013#define PGM_PAGE_CLEAR_WRITTEN_TO(a_pVM, a_pPage) \
1014 do { (a_pPage)->s.fWrittenToY = 0; PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1015
1016/**
1017 * Checks if the page was marked as written-to.
1018 * @returns true/false.
1019 * @param a_pPage Pointer to the physical guest page tracking structure.
1020 */
1021#define PGM_PAGE_IS_WRITTEN_TO(a_pPage) ( (a_pPage)->s.fWrittenToY )
1022
1023
1024/** @name PT usage values (PGMPAGE::u2PDEType).
1025 *
1026 * @{ */
1027/** Either as a PT or PDE. */
1028#define PGM_PAGE_PDE_TYPE_DONTCARE 0
1029/** Must use a page table to map the range. */
1030#define PGM_PAGE_PDE_TYPE_PT 1
1031/** Can use a page directory entry to map the continuous range. */
1032#define PGM_PAGE_PDE_TYPE_PDE 2
1033/** Can use a page directory entry to map the continuous range - temporarily disabled (by page monitoring). */
1034#define PGM_PAGE_PDE_TYPE_PDE_DISABLED 3
1035/** @} */
1036
1037/**
1038 * Set the PDE type of the page
1039 * @param a_pVM The VM handle, only used for lock ownership assertions.
1040 * @param a_pPage Pointer to the physical guest page tracking structure.
1041 * @param a_uType PGM_PAGE_PDE_TYPE_*.
1042 */
1043#define PGM_PAGE_SET_PDE_TYPE(a_pVM, a_pPage, a_uType) \
1044 do { (a_pPage)->s.u2PDETypeY = (a_uType); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1045
1046/**
1047 * Checks if the page was marked being part of a large page
1048 * @returns true/false.
1049 * @param a_pPage Pointer to the physical guest page tracking structure.
1050 */
1051#define PGM_PAGE_GET_PDE_TYPE(a_pPage) ( (a_pPage)->s.u2PDETypeY )
1052
1053/** @name Physical Access Handler State values (PGMPAGE::u2HandlerPhysStateY).
1054 *
1055 * @remarks The values are assigned in order of priority, so we can calculate
1056 * the correct state for a page with different handlers installed.
1057 * @{ */
1058/** No handler installed. */
1059#define PGM_PAGE_HNDL_PHYS_STATE_NONE 0
1060/** Monitoring is temporarily disabled. */
1061#define PGM_PAGE_HNDL_PHYS_STATE_DISABLED 1
1062/** Write access is monitored. */
1063#define PGM_PAGE_HNDL_PHYS_STATE_WRITE 2
1064/** All access is monitored. */
1065#define PGM_PAGE_HNDL_PHYS_STATE_ALL 3
1066/** @} */
1067
1068/**
1069 * Gets the physical access handler state of a page.
1070 * @returns PGM_PAGE_HNDL_PHYS_STATE_* value.
1071 * @param a_pPage Pointer to the physical guest page tracking structure.
1072 */
1073#define PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) ( (a_pPage)->s.u2HandlerPhysStateY )
1074
1075/**
1076 * Sets the physical access handler state of a page.
1077 * @param a_pPage Pointer to the physical guest page tracking structure.
1078 * @param a_uState The new state value.
1079 */
1080#define PGM_PAGE_SET_HNDL_PHYS_STATE(a_pPage, a_uState) \
1081 do { (a_pPage)->s.u2HandlerPhysStateY = (a_uState); } while (0)
1082
1083/**
1084 * Checks if the page has any physical access handlers, including temporarily disabled ones.
1085 * @returns true/false
1086 * @param a_pPage Pointer to the physical guest page tracking structure.
1087 */
1088#define PGM_PAGE_HAS_ANY_PHYSICAL_HANDLERS(a_pPage) \
1089 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) != PGM_PAGE_HNDL_PHYS_STATE_NONE )
1090
1091/**
1092 * Checks if the page has any active physical access handlers.
1093 * @returns true/false
1094 * @param a_pPage Pointer to the physical guest page tracking structure.
1095 */
1096#define PGM_PAGE_HAS_ACTIVE_PHYSICAL_HANDLERS(a_pPage) \
1097 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) >= PGM_PAGE_HNDL_PHYS_STATE_WRITE )
1098
1099/**
1100 * Checks if the page has any access handlers, including temporarily disabled ones.
1101 * @returns true/false
1102 * @param a_pPage Pointer to the physical guest page tracking structure.
1103 */
1104#define PGM_PAGE_HAS_ANY_HANDLERS(a_pPage) \
1105 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) != PGM_PAGE_HNDL_PHYS_STATE_NONE )
1106
1107/**
1108 * Checks if the page has any active access handlers.
1109 * @returns true/false
1110 * @param a_pPage Pointer to the physical guest page tracking structure.
1111 */
1112#define PGM_PAGE_HAS_ACTIVE_HANDLERS(a_pPage) \
1113 (PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) >= PGM_PAGE_HNDL_PHYS_STATE_WRITE )
1114
1115/**
1116 * Checks if the page has any active access handlers catching all accesses.
1117 * @returns true/false
1118 * @param a_pPage Pointer to the physical guest page tracking structure.
1119 */
1120#define PGM_PAGE_HAS_ACTIVE_ALL_HANDLERS(a_pPage) \
1121 ( PGM_PAGE_GET_HNDL_PHYS_STATE(a_pPage) == PGM_PAGE_HNDL_PHYS_STATE_ALL )
1122
1123
1124/** @def PGM_PAGE_GET_TRACKING
1125 * Gets the packed shadow page pool tracking data associated with a guest page.
1126 * @returns uint16_t containing the data.
1127 * @param a_pPage Pointer to the physical guest page tracking structure.
1128 */
1129#define PGM_PAGE_GET_TRACKING_NA(a_pPage) ( (a_pPage)->s.u16TrackingY )
1130#if defined(__GNUC__) && defined(VBOX_STRICT)
1131# define PGM_PAGE_GET_TRACKING(a_pPage) __extension__ ({ PGM_PAGE_ASSERT_LOCK(pVM); PGM_PAGE_GET_TRACKING_NA(a_pPage); })
1132#else
1133# define PGM_PAGE_GET_TRACKING PGM_PAGE_GET_TRACKING_NA
1134#endif
1135
1136/** @def PGM_PAGE_SET_TRACKING
1137 * Sets the packed shadow page pool tracking data associated with a guest page.
1138 * @param a_pVM The VM handle, only used for lock ownership assertions.
1139 * @param a_pPage Pointer to the physical guest page tracking structure.
1140 * @param a_u16TrackingData The tracking data to store.
1141 */
1142#define PGM_PAGE_SET_TRACKING(a_pVM, a_pPage, a_u16TrackingData) \
1143 do { (a_pPage)->s.u16TrackingY = (a_u16TrackingData); PGM_PAGE_ASSERT_LOCK(a_pVM); } while (0)
1144
1145/** @def PGM_PAGE_GET_TD_CREFS
1146 * Gets the @a cRefs tracking data member.
1147 * @returns cRefs.
1148 * @param a_pPage Pointer to the physical guest page tracking structure.
1149 */
1150#define PGM_PAGE_GET_TD_CREFS(a_pPage) \
1151 ((PGM_PAGE_GET_TRACKING(a_pPage) >> PGMPOOL_TD_CREFS_SHIFT) & PGMPOOL_TD_CREFS_MASK)
1152#define PGM_PAGE_GET_TD_CREFS_NA(a_pPage) \
1153 ((PGM_PAGE_GET_TRACKING_NA(a_pPage) >> PGMPOOL_TD_CREFS_SHIFT) & PGMPOOL_TD_CREFS_MASK)
1154
1155/** @def PGM_PAGE_GET_TD_IDX
1156 * Gets the @a idx tracking data member.
1157 * @returns idx.
1158 * @param a_pPage Pointer to the physical guest page tracking structure.
1159 */
1160#define PGM_PAGE_GET_TD_IDX(a_pPage) \
1161 ((PGM_PAGE_GET_TRACKING(a_pPage) >> PGMPOOL_TD_IDX_SHIFT) & PGMPOOL_TD_IDX_MASK)
1162#define PGM_PAGE_GET_TD_IDX_NA(a_pPage) \
1163 ((PGM_PAGE_GET_TRACKING_NA(a_pPage) >> PGMPOOL_TD_IDX_SHIFT) & PGMPOOL_TD_IDX_MASK)
1164
1165
1166/** Max number of locks on a page. */
1167#define PGM_PAGE_MAX_LOCKS UINT8_C(254)
1168
1169/** Get the read lock count.
1170 * @returns count.
1171 * @param a_pPage Pointer to the physical guest page tracking structure.
1172 */
1173#define PGM_PAGE_GET_READ_LOCKS(a_pPage) ( (a_pPage)->s.cReadLocksY )
1174
1175/** Get the write lock count.
1176 * @returns count.
1177 * @param a_pPage Pointer to the physical guest page tracking structure.
1178 */
1179#define PGM_PAGE_GET_WRITE_LOCKS(a_pPage) ( (a_pPage)->s.cWriteLocksY )
1180
1181/** Decrement the read lock counter.
1182 * @param a_pPage Pointer to the physical guest page tracking structure.
1183 */
1184#define PGM_PAGE_DEC_READ_LOCKS(a_pPage) do { --(a_pPage)->s.cReadLocksY; } while (0)
1185
1186/** Decrement the write lock counter.
1187 * @param a_pPage Pointer to the physical guest page tracking structure.
1188 */
1189#define PGM_PAGE_DEC_WRITE_LOCKS(a_pPage) do { --(a_pPage)->s.cWriteLocksY; } while (0)
1190
1191/** Increment the read lock counter.
1192 * @param a_pPage Pointer to the physical guest page tracking structure.
1193 */
1194#define PGM_PAGE_INC_READ_LOCKS(a_pPage) do { ++(a_pPage)->s.cReadLocksY; } while (0)
1195
1196/** Increment the write lock counter.
1197 * @param a_pPage Pointer to the physical guest page tracking structure.
1198 */
1199#define PGM_PAGE_INC_WRITE_LOCKS(a_pPage) do { ++(a_pPage)->s.cWriteLocksY; } while (0)
1200
1201
1202/** Gets the NEM state.
1203 * @returns NEM state value (two bits).
1204 * @param a_pPage Pointer to the physical guest page tracking structure.
1205 */
1206#define PGM_PAGE_GET_NEM_STATE(a_pPage) ((a_pPage)->s.u2NemStateY)
1207
1208/** Sets the NEM state.
1209 * @param a_pPage Pointer to the physical guest page tracking structure.
1210 * @param a_u2State The NEM state value (specific to NEM impl.).
1211 */
1212#define PGM_PAGE_SET_NEM_STATE(a_pPage, a_u2State) \
1213 do { Assert((a_u2State) < 4); (a_pPage)->s.u2NemStateY = (a_u2State); } while (0)
1214
1215
1216#if 0
1217/** Enables sanity checking of write monitoring using CRC-32. */
1218# define PGMLIVESAVERAMPAGE_WITH_CRC32
1219#endif
1220
1221/**
1222 * Per page live save tracking data.
1223 */
1224typedef struct PGMLIVESAVERAMPAGE
1225{
1226 /** Number of times it has been dirtied. */
1227 uint32_t cDirtied : 24;
1228 /** Whether it is currently dirty. */
1229 uint32_t fDirty : 1;
1230 /** Ignore the page.
1231 * This is used for pages that has been MMIO, MMIO2 or ROM pages once. We will
1232 * deal with these after pausing the VM and DevPCI have said it bit about
1233 * remappings. */
1234 uint32_t fIgnore : 1;
1235 /** Was a ZERO page last time around. */
1236 uint32_t fZero : 1;
1237 /** Was a SHARED page last time around. */
1238 uint32_t fShared : 1;
1239 /** Whether the page is/was write monitored in a previous pass. */
1240 uint32_t fWriteMonitored : 1;
1241 /** Whether the page is/was write monitored earlier in this pass. */
1242 uint32_t fWriteMonitoredJustNow : 1;
1243 /** Bits reserved for future use. */
1244 uint32_t u2Reserved : 2;
1245#ifdef PGMLIVESAVERAMPAGE_WITH_CRC32
1246 /** CRC-32 for the page. This is for internal consistency checks. */
1247 uint32_t u32Crc;
1248#endif
1249} PGMLIVESAVERAMPAGE;
1250#ifdef PGMLIVESAVERAMPAGE_WITH_CRC32
1251AssertCompileSize(PGMLIVESAVERAMPAGE, 8);
1252#else
1253AssertCompileSize(PGMLIVESAVERAMPAGE, 4);
1254#endif
1255/** Pointer to the per page live save tracking data. */
1256typedef PGMLIVESAVERAMPAGE *PPGMLIVESAVERAMPAGE;
1257
1258/** The max value of PGMLIVESAVERAMPAGE::cDirtied. */
1259#define PGMLIVSAVEPAGE_MAX_DIRTIED 0x00fffff0
1260
1261
1262/**
1263 * RAM range for GC Phys to HC Phys conversion.
1264 *
1265 * Can be used for HC Virt to GC Phys and HC Virt to HC Phys
1266 * conversions too, but we'll let MM handle that for now.
1267 *
1268 * This structure is used by linked lists in both GC and HC.
1269 */
1270typedef struct PGMRAMRANGE
1271{
1272 /** Start of the range. Page aligned. */
1273 RTGCPHYS GCPhys;
1274 /** Size of the range. (Page aligned of course). */
1275 RTGCPHYS cb;
1276 /** Pointer to the next RAM range - for R3. */
1277 R3PTRTYPE(struct PGMRAMRANGE *) pNextR3;
1278 /** Pointer to the next RAM range - for R0. */
1279 R0PTRTYPE(struct PGMRAMRANGE *) pNextR0;
1280 /** PGM_RAM_RANGE_FLAGS_* flags. */
1281 uint32_t fFlags;
1282 uint32_t fPadding1;
1283 /** Last address in the range (inclusive). Page aligned (-1). */
1284 RTGCPHYS GCPhysLast;
1285 /** Start of the HC mapping of the range. This is only used for MMIO2. */
1286 R3PTRTYPE(void *) pvR3;
1287 /** Live save per page tracking data. */
1288 R3PTRTYPE(PPGMLIVESAVERAMPAGE) paLSPages;
1289 /** The range description. */
1290 R3PTRTYPE(const char *) pszDesc;
1291 /** Pointer to self - R0 pointer. */
1292 R0PTRTYPE(struct PGMRAMRANGE *) pSelfR0;
1293
1294 /** Pointer to the left search three node - ring-3 context. */
1295 R3PTRTYPE(struct PGMRAMRANGE *) pLeftR3;
1296 /** Pointer to the right search three node - ring-3 context. */
1297 R3PTRTYPE(struct PGMRAMRANGE *) pRightR3;
1298 /** Pointer to the left search three node - ring-0 context. */
1299 R0PTRTYPE(struct PGMRAMRANGE *) pLeftR0;
1300 /** Pointer to the right search three node - ring-0 context. */
1301 R0PTRTYPE(struct PGMRAMRANGE *) pRightR0;
1302
1303 /** Padding to make aPage aligned on sizeof(PGMPAGE). */
1304#if HC_ARCH_BITS == 32
1305 uint32_t au32Alignment2[HC_ARCH_BITS == 32 ? 2 : 0];
1306#endif
1307 /** Array of physical guest page tracking structures. */
1308 PGMPAGE aPages[1];
1309} PGMRAMRANGE;
1310/** Pointer to RAM range for GC Phys to HC Phys conversion. */
1311typedef PGMRAMRANGE *PPGMRAMRANGE;
1312
1313/** @name PGMRAMRANGE::fFlags
1314 * @{ */
1315/** The RAM range is floating around as an independent guest mapping. */
1316#define PGM_RAM_RANGE_FLAGS_FLOATING RT_BIT(20)
1317/** Ad hoc RAM range for an ROM mapping. */
1318#define PGM_RAM_RANGE_FLAGS_AD_HOC_ROM RT_BIT(21)
1319/** Ad hoc RAM range for an MMIO mapping. */
1320#define PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO RT_BIT(22)
1321/** Ad hoc RAM range for an MMIO2 or pre-registered MMIO mapping. */
1322#define PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO_EX RT_BIT(23)
1323/** @} */
1324
1325/** Tests if a RAM range is an ad hoc one or not.
1326 * @returns true/false.
1327 * @param pRam The RAM range.
1328 */
1329#define PGM_RAM_RANGE_IS_AD_HOC(pRam) \
1330 (!!( (pRam)->fFlags & (PGM_RAM_RANGE_FLAGS_AD_HOC_ROM | PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO | PGM_RAM_RANGE_FLAGS_AD_HOC_MMIO_EX) ) )
1331
1332/** The number of entries in the RAM range TLBs (there is one for each
1333 * context). Must be a power of two. */
1334#define PGM_RAMRANGE_TLB_ENTRIES 8
1335
1336/**
1337 * Calculates the RAM range TLB index for the physical address.
1338 *
1339 * @returns RAM range TLB index.
1340 * @param a_GCPhys The guest physical address.
1341 */
1342#define PGM_RAMRANGE_TLB_IDX(a_GCPhys) ( ((a_GCPhys) >> 20) & (PGM_RAMRANGE_TLB_ENTRIES - 1) )
1343
1344
1345
1346/**
1347 * Per page tracking structure for ROM image.
1348 *
1349 * A ROM image may have a shadow page, in which case we may have two pages
1350 * backing it. This structure contains the PGMPAGE for both while
1351 * PGMRAMRANGE have a copy of the active one. It is important that these
1352 * aren't out of sync in any regard other than page pool tracking data.
1353 */
1354typedef struct PGMROMPAGE
1355{
1356 /** The page structure for the virgin ROM page. */
1357 PGMPAGE Virgin;
1358 /** The page structure for the shadow RAM page. */
1359 PGMPAGE Shadow;
1360 /** The current protection setting. */
1361 PGMROMPROT enmProt;
1362 /** Live save status information. Makes use of unused alignment space. */
1363 struct
1364 {
1365 /** The previous protection value. */
1366 uint8_t u8Prot;
1367 /** Written to flag set by the handler. */
1368 bool fWrittenTo;
1369 /** Whether the shadow page is dirty or not. */
1370 bool fDirty;
1371 /** Whether it was dirtied in the recently. */
1372 bool fDirtiedRecently;
1373 } LiveSave;
1374} PGMROMPAGE;
1375AssertCompileSizeAlignment(PGMROMPAGE, 8);
1376/** Pointer to a ROM page tracking structure. */
1377typedef PGMROMPAGE *PPGMROMPAGE;
1378
1379
1380/**
1381 * A registered ROM image.
1382 *
1383 * This is needed to keep track of ROM image since they generally intrude
1384 * into a PGMRAMRANGE. It also keeps track of additional info like the
1385 * two page sets (read-only virgin and read-write shadow), the current
1386 * state of each page.
1387 *
1388 * Because access handlers cannot easily be executed in a different
1389 * context, the ROM ranges needs to be accessible and in all contexts.
1390 */
1391typedef struct PGMROMRANGE
1392{
1393 /** Pointer to the next range - R3. */
1394 R3PTRTYPE(struct PGMROMRANGE *) pNextR3;
1395 /** Pointer to the next range - R0. */
1396 R0PTRTYPE(struct PGMROMRANGE *) pNextR0;
1397 /** Address of the range. */
1398 RTGCPHYS GCPhys;
1399 /** Address of the last byte in the range. */
1400 RTGCPHYS GCPhysLast;
1401 /** Size of the range. */
1402 RTGCPHYS cb;
1403 /** The flags (PGMPHYS_ROM_FLAGS_*). */
1404 uint32_t fFlags;
1405 /** The saved state range ID. */
1406 uint8_t idSavedState;
1407 /** Alignment padding. */
1408 uint8_t au8Alignment[3];
1409 /** Alignment padding ensuring that aPages is sizeof(PGMROMPAGE) aligned. */
1410 uint32_t au32Alignemnt[HC_ARCH_BITS == 32 ? 5 : 1];
1411 /** The size bits pvOriginal points to. */
1412 uint32_t cbOriginal;
1413 /** Pointer to the original bits when PGMPHYS_ROM_FLAGS_PERMANENT_BINARY was specified.
1414 * This is used for strictness checks. */
1415 R3PTRTYPE(const void *) pvOriginal;
1416 /** The ROM description. */
1417 R3PTRTYPE(const char *) pszDesc;
1418 /** The per page tracking structures. */
1419 PGMROMPAGE aPages[1];
1420} PGMROMRANGE;
1421/** Pointer to a ROM range. */
1422typedef PGMROMRANGE *PPGMROMRANGE;
1423
1424
1425/**
1426 * Live save per page data for an MMIO2 page.
1427 *
1428 * Not using PGMLIVESAVERAMPAGE here because we cannot use normal write monitoring
1429 * of MMIO2 pages. The current approach is using some optimistic SHA-1 +
1430 * CRC-32 for detecting changes as well as special handling of zero pages. This
1431 * is a TEMPORARY measure which isn't perfect, but hopefully it is good enough
1432 * for speeding things up. (We're using SHA-1 and not SHA-256 or SHA-512
1433 * because of speed (2.5x and 6x slower).)
1434 *
1435 * @todo Implement dirty MMIO2 page reporting that can be enabled during live
1436 * save but normally is disabled. Since we can write monitor guest
1437 * accesses on our own, we only need this for host accesses. Shouldn't be
1438 * too difficult for DevVGA, VMMDev might be doable, the planned
1439 * networking fun will be fun since it involves ring-0.
1440 */
1441typedef struct PGMLIVESAVEMMIO2PAGE
1442{
1443 /** Set if the page is considered dirty. */
1444 bool fDirty;
1445 /** The number of scans this page has remained unchanged for.
1446 * Only updated for dirty pages. */
1447 uint8_t cUnchangedScans;
1448 /** Whether this page was zero at the last scan. */
1449 bool fZero;
1450 /** Alignment padding. */
1451 bool fReserved;
1452 /** CRC-32 for the first half of the page.
1453 * This is used together with u32CrcH2 to quickly detect changes in the page
1454 * during the non-final passes. */
1455 uint32_t u32CrcH1;
1456 /** CRC-32 for the second half of the page. */
1457 uint32_t u32CrcH2;
1458 /** SHA-1 for the saved page.
1459 * This is used in the final pass to skip pages without changes. */
1460 uint8_t abSha1Saved[RTSHA1_HASH_SIZE];
1461} PGMLIVESAVEMMIO2PAGE;
1462/** Pointer to a live save status data for an MMIO2 page. */
1463typedef PGMLIVESAVEMMIO2PAGE *PPGMLIVESAVEMMIO2PAGE;
1464
1465/**
1466 * A registered MMIO2 (= Device RAM) range.
1467 *
1468 * There are a few reason why we need to keep track of these registrations. One
1469 * of them is the deregistration & cleanup stuff, while another is that the
1470 * PGMRAMRANGE associated with such a region may have to be removed from the ram
1471 * range list.
1472 *
1473 * Overlapping with a RAM range has to be 100% or none at all. The pages in the
1474 * existing RAM range must not be ROM nor MMIO. A guru meditation will be
1475 * raised if a partial overlap or an overlap of ROM pages is encountered. On an
1476 * overlap we will free all the existing RAM pages and put in the ram range
1477 * pages instead.
1478 */
1479typedef struct PGMREGMMIO2RANGE
1480{
1481 /** The owner of the range. (a device) */
1482 PPDMDEVINSR3 pDevInsR3;
1483 /** Pointer to the ring-3 mapping of the allocation. */
1484 RTR3PTR pvR3;
1485#if defined(VBOX_WITH_RAM_IN_KERNEL) && !defined(VBOX_WITH_LINEAR_HOST_PHYS_MEM)
1486 /** Pointer to the ring-0 mapping of the allocation. */
1487 RTR0PTR pvR0;
1488#endif
1489 /** Pointer to the next range - R3. */
1490 R3PTRTYPE(struct PGMREGMMIO2RANGE *) pNextR3;
1491 /** Flags (PGMREGMMIO2RANGE_F_XXX). */
1492 uint16_t fFlags;
1493 /** The sub device number (internal PCI config (CFGM) number). */
1494 uint8_t iSubDev;
1495 /** The PCI region number. */
1496 uint8_t iRegion;
1497 /** The saved state range ID. */
1498 uint8_t idSavedState;
1499 /** MMIO2 range identifier, for page IDs (PGMPAGE::s.idPage). */
1500 uint8_t idMmio2;
1501 /** Alignment padding for putting the ram range on a PGMPAGE alignment boundary. */
1502#if defined(VBOX_WITH_RAM_IN_KERNEL) && !defined(VBOX_WITH_LINEAR_HOST_PHYS_MEM)
1503 uint8_t abAlignment[HC_ARCH_BITS == 32 ? 6 + 4 : 2];
1504#else
1505 uint8_t abAlignment[HC_ARCH_BITS == 32 ? 6 + 8 : 2 + 8];
1506#endif
1507 /** The real size.
1508 * This may be larger than indicated by RamRange.cb if the range has been
1509 * reduced during saved state loading. */
1510 RTGCPHYS cbReal;
1511 /** Pointer to the physical handler for MMIO. */
1512 R3PTRTYPE(PPGMPHYSHANDLER) pPhysHandlerR3;
1513 /** Live save per page tracking data for MMIO2. */
1514 R3PTRTYPE(PPGMLIVESAVEMMIO2PAGE) paLSPages;
1515 /** The associated RAM range. */
1516 PGMRAMRANGE RamRange;
1517} PGMREGMMIO2RANGE;
1518AssertCompileMemberAlignment(PGMREGMMIO2RANGE, RamRange, 16);
1519/** Pointer to a MMIO2 or pre-registered MMIO range. */
1520typedef PGMREGMMIO2RANGE *PPGMREGMMIO2RANGE;
1521
1522/** @name PGMREGMMIO2RANGE_F_XXX - Registered MMIO2 range flags.
1523 * @{ */
1524/** Set if it's an MMIO2 range.
1525 * @note Historical. For a while we did some of the MMIO this way too. */
1526#define PGMREGMMIO2RANGE_F_MMIO2 UINT16_C(0x0001)
1527/** Set if this is the first chunk in the MMIO2 range. */
1528#define PGMREGMMIO2RANGE_F_FIRST_CHUNK UINT16_C(0x0002)
1529/** Set if this is the last chunk in the MMIO2 range. */
1530#define PGMREGMMIO2RANGE_F_LAST_CHUNK UINT16_C(0x0004)
1531/** Set if the whole range is mapped. */
1532#define PGMREGMMIO2RANGE_F_MAPPED UINT16_C(0x0008)
1533/** Set if it's overlapping, clear if not. */
1534#define PGMREGMMIO2RANGE_F_OVERLAPPING UINT16_C(0x0010)
1535/** @} */
1536
1537
1538/** @name Internal MMIO2 constants.
1539 * @{ */
1540/** The maximum number of MMIO2 ranges. */
1541#define PGM_MMIO2_MAX_RANGES 32
1542/** The maximum number of pages in a MMIO2 range. */
1543#define PGM_MMIO2_MAX_PAGE_COUNT UINT32_C(0x01000000)
1544/** Makes a MMIO2 page ID out of a MMIO2 range ID and page index number. */
1545#define PGM_MMIO2_PAGEID_MAKE(a_idMmio2, a_iPage) ( ((uint32_t)(a_idMmio2) << 24) | (uint32_t)(a_iPage) )
1546/** Gets the MMIO2 range ID from an MMIO2 page ID. */
1547#define PGM_MMIO2_PAGEID_GET_MMIO2_ID(a_idPage) ( (uint8_t)((a_idPage) >> 24) )
1548/** Gets the MMIO2 page index from an MMIO2 page ID. */
1549#define PGM_MMIO2_PAGEID_GET_IDX(a_idPage) ( ((a_idPage) & UINT32_C(0x00ffffff)) )
1550/** @} */
1551
1552
1553
1554/**
1555 * PGMPhysRead/Write cache entry
1556 */
1557typedef struct PGMPHYSCACHEENTRY
1558{
1559 /** R3 pointer to physical page. */
1560 R3PTRTYPE(uint8_t *) pbR3;
1561 /** GC Physical address for cache entry */
1562 RTGCPHYS GCPhys;
1563#if HC_ARCH_BITS == 64 && GC_ARCH_BITS == 32
1564 RTGCPHYS u32Padding0; /**< alignment padding. */
1565#endif
1566} PGMPHYSCACHEENTRY;
1567
1568/**
1569 * PGMPhysRead/Write cache to reduce REM memory access overhead
1570 */
1571typedef struct PGMPHYSCACHE
1572{
1573 /** Bitmap of valid cache entries */
1574 uint64_t aEntries;
1575 /** Cache entries */
1576 PGMPHYSCACHEENTRY Entry[PGM_MAX_PHYSCACHE_ENTRIES];
1577} PGMPHYSCACHE;
1578
1579
1580/** @name Ring-3 page mapping TLBs
1581 * @{ */
1582
1583/** Pointer to an allocation chunk ring-3 mapping. */
1584typedef struct PGMCHUNKR3MAP *PPGMCHUNKR3MAP;
1585/** Pointer to an allocation chunk ring-3 mapping pointer. */
1586typedef PPGMCHUNKR3MAP *PPPGMCHUNKR3MAP;
1587
1588/**
1589 * Ring-3 tracking structure for an allocation chunk ring-3 mapping.
1590 *
1591 * The primary tree (Core) uses the chunk id as key.
1592 */
1593typedef struct PGMCHUNKR3MAP
1594{
1595 /** The key is the chunk id. */
1596 AVLU32NODECORE Core;
1597 /** The time (ChunkR3Map.iNow) this chunk was last used. Used for unmap
1598 * selection. */
1599 uint32_t iLastUsed;
1600 /** The current reference count. */
1601 uint32_t volatile cRefs;
1602 /** The current permanent reference count. */
1603 uint32_t volatile cPermRefs;
1604 /** The mapping address. */
1605 void *pv;
1606} PGMCHUNKR3MAP;
1607
1608/**
1609 * Allocation chunk ring-3 mapping TLB entry.
1610 */
1611typedef struct PGMCHUNKR3MAPTLBE
1612{
1613 /** The chunk id. */
1614 uint32_t volatile idChunk;
1615#if HC_ARCH_BITS == 64
1616 uint32_t u32Padding; /**< alignment padding. */
1617#endif
1618 /** The chunk map. */
1619#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1620 R3PTRTYPE(PPGMCHUNKR3MAP) volatile pChunk;
1621#else
1622 R3R0PTRTYPE(PPGMCHUNKR3MAP) volatile pChunk;
1623#endif
1624} PGMCHUNKR3MAPTLBE;
1625/** Pointer to the an allocation chunk ring-3 mapping TLB entry. */
1626typedef PGMCHUNKR3MAPTLBE *PPGMCHUNKR3MAPTLBE;
1627
1628/** The number of TLB entries in PGMCHUNKR3MAPTLB.
1629 * @remark Must be a power of two value. */
1630#define PGM_CHUNKR3MAPTLB_ENTRIES 64
1631
1632/**
1633 * Allocation chunk ring-3 mapping TLB.
1634 *
1635 * @remarks We use a TLB to speed up lookups by avoiding walking the AVL.
1636 * At first glance this might look kinda odd since AVL trees are
1637 * supposed to give the most optimal lookup times of all trees
1638 * due to their balancing. However, take a tree with 1023 nodes
1639 * in it, that's 10 levels, meaning that most searches has to go
1640 * down 9 levels before they find what they want. This isn't fast
1641 * compared to a TLB hit. There is the factor of cache misses,
1642 * and of course the problem with trees and branch prediction.
1643 * This is why we use TLBs in front of most of the trees.
1644 *
1645 * @todo Generalize this TLB + AVL stuff, shouldn't be all that
1646 * difficult when we switch to the new inlined AVL trees (from kStuff).
1647 */
1648typedef struct PGMCHUNKR3MAPTLB
1649{
1650 /** The TLB entries. */
1651 PGMCHUNKR3MAPTLBE aEntries[PGM_CHUNKR3MAPTLB_ENTRIES];
1652} PGMCHUNKR3MAPTLB;
1653
1654/**
1655 * Calculates the index of a guest page in the Ring-3 Chunk TLB.
1656 * @returns Chunk TLB index.
1657 * @param idChunk The Chunk ID.
1658 */
1659#define PGM_CHUNKR3MAPTLB_IDX(idChunk) ( (idChunk) & (PGM_CHUNKR3MAPTLB_ENTRIES - 1) )
1660
1661
1662/**
1663 * Ring-3 guest page mapping TLB entry.
1664 * @remarks used in ring-0 as well at the moment.
1665 */
1666typedef struct PGMPAGER3MAPTLBE
1667{
1668 /** Address of the page. */
1669 RTGCPHYS volatile GCPhys;
1670 /** The guest page. */
1671#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1672 R3PTRTYPE(PPGMPAGE) volatile pPage;
1673#else
1674 R3R0PTRTYPE(PPGMPAGE) volatile pPage;
1675#endif
1676 /** Pointer to the page mapping tracking structure, PGMCHUNKR3MAP. */
1677#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1678 R3PTRTYPE(PPGMCHUNKR3MAP) volatile pMap;
1679#else
1680 R3R0PTRTYPE(PPGMCHUNKR3MAP) volatile pMap;
1681#endif
1682 /** The address */
1683#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
1684 R3PTRTYPE(void *) volatile pv;
1685#else
1686 R3R0PTRTYPE(void *) volatile pv;
1687#endif
1688#if HC_ARCH_BITS == 32
1689 uint32_t u32Padding; /**< alignment padding. */
1690#endif
1691} PGMPAGER3MAPTLBE;
1692/** Pointer to an entry in the HC physical TLB. */
1693typedef PGMPAGER3MAPTLBE *PPGMPAGER3MAPTLBE;
1694
1695
1696/** The number of entries in the ring-3 guest page mapping TLB.
1697 * @remarks The value must be a power of two. */
1698#define PGM_PAGER3MAPTLB_ENTRIES 256
1699
1700/**
1701 * Ring-3 guest page mapping TLB.
1702 * @remarks used in ring-0 as well at the moment.
1703 */
1704typedef struct PGMPAGER3MAPTLB
1705{
1706 /** The TLB entries. */
1707 PGMPAGER3MAPTLBE aEntries[PGM_PAGER3MAPTLB_ENTRIES];
1708} PGMPAGER3MAPTLB;
1709/** Pointer to the ring-3 guest page mapping TLB. */
1710typedef PGMPAGER3MAPTLB *PPGMPAGER3MAPTLB;
1711
1712/**
1713 * Calculates the index of the TLB entry for the specified guest page.
1714 * @returns Physical TLB index.
1715 * @param GCPhys The guest physical address.
1716 */
1717#define PGM_PAGER3MAPTLB_IDX(GCPhys) ( ((GCPhys) >> PAGE_SHIFT) & (PGM_PAGER3MAPTLB_ENTRIES - 1) )
1718
1719/** @} */
1720
1721#if defined(VBOX_WITH_RAM_IN_KERNEL) || defined(DOXYGEN_RUNNING)
1722/** @name Ring-0 page mapping TLB
1723 * @{ */
1724/**
1725 * Ring-0 guest page mapping TLB entry.
1726 */
1727typedef struct PGMPAGER0MAPTLBE
1728{
1729 /** Address of the page. */
1730 RTGCPHYS volatile GCPhys;
1731 /** The guest page. */
1732 R0PTRTYPE(PPGMPAGE) volatile pPage;
1733 /** The address */
1734 R0PTRTYPE(void *) volatile pv;
1735} PGMPAGER0MAPTLBE;
1736/** Pointer to an entry in the HC physical TLB. */
1737typedef PGMPAGER0MAPTLBE *PPGMPAGER0MAPTLBE;
1738
1739
1740/** The number of entries in the ring-3 guest page mapping TLB.
1741 * @remarks The value must be a power of two. */
1742#define PGM_PAGER0MAPTLB_ENTRIES 256
1743
1744/**
1745 * Ring-3 guest page mapping TLB.
1746 * @remarks used in ring-0 as well at the moment.
1747 */
1748typedef struct PGMPAGER0MAPTLB
1749{
1750 /** The TLB entries. */
1751 PGMPAGER0MAPTLBE aEntries[PGM_PAGER0MAPTLB_ENTRIES];
1752} PGMPAGER0MAPTLB;
1753/** Pointer to the ring-3 guest page mapping TLB. */
1754typedef PGMPAGER0MAPTLB *PPGMPAGER0MAPTLB;
1755
1756/**
1757 * Calculates the index of the TLB entry for the specified guest page.
1758 * @returns Physical TLB index.
1759 * @param GCPhys The guest physical address.
1760 */
1761#define PGM_PAGER0MAPTLB_IDX(GCPhys) ( ((GCPhys) >> PAGE_SHIFT) & (PGM_PAGER0MAPTLB_ENTRIES - 1) )
1762/** @} */
1763#endif /* VBOX_WITH_RAM_IN_KERNEL || DOXYGEN_RUNNING */
1764
1765/**
1766 * Raw-mode context dynamic mapping cache entry.
1767 *
1768 * Because of raw-mode context being reloctable and all relocations are applied
1769 * in ring-3, this has to be defined here and be RC specific.
1770 *
1771 * @sa PGMRZDYNMAPENTRY, PGMR0DYNMAPENTRY.
1772 */
1773typedef struct PGMRCDYNMAPENTRY
1774{
1775 /** The physical address of the currently mapped page.
1776 * This is duplicate for three reasons: cache locality, cache policy of the PT
1777 * mappings and sanity checks. */
1778 RTHCPHYS HCPhys;
1779 /** Pointer to the page. */
1780 RTRCPTR pvPage;
1781 /** The number of references. */
1782 int32_t volatile cRefs;
1783 /** PTE pointer union. */
1784 struct PGMRCDYNMAPENTRY_PPTE
1785 {
1786 /** PTE pointer, 32-bit legacy version. */
1787 RCPTRTYPE(PX86PTE) pLegacy;
1788 /** PTE pointer, PAE version. */
1789 RCPTRTYPE(PX86PTEPAE) pPae;
1790 } uPte;
1791} PGMRCDYNMAPENTRY;
1792/** Pointer to a dynamic mapping cache entry for the raw-mode context. */
1793typedef PGMRCDYNMAPENTRY *PPGMRCDYNMAPENTRY;
1794
1795
1796/**
1797 * Dynamic mapping cache for the raw-mode context.
1798 *
1799 * This is initialized during VMMRC init based upon the pbDynPageMapBaseGC and
1800 * paDynPageMap* PGM members. However, it has to be defined in PGMInternal.h
1801 * so that we can perform relocations from PGMR3Relocate. This has the
1802 * consequence that we must have separate ring-0 and raw-mode context versions
1803 * of this struct even if they share the basic elements.
1804 *
1805 * @sa PPGMRZDYNMAP, PGMR0DYNMAP.
1806 */
1807typedef struct PGMRCDYNMAP
1808{
1809 /** The usual magic number / eye catcher (PGMRZDYNMAP_MAGIC). */
1810 uint32_t u32Magic;
1811 /** Array for tracking and managing the pages. */
1812 RCPTRTYPE(PPGMRCDYNMAPENTRY) paPages;
1813 /** The cache size given as a number of pages. */
1814 uint32_t cPages;
1815 /** The current load.
1816 * This does not include guard pages. */
1817 uint32_t cLoad;
1818 /** The max load ever.
1819 * This is maintained to get trigger adding of more mapping space. */
1820 uint32_t cMaxLoad;
1821 /** The number of guard pages. */
1822 uint32_t cGuardPages;
1823 /** The number of users (protected by hInitLock). */
1824 uint32_t cUsers;
1825} PGMRCDYNMAP;
1826/** Pointer to the dynamic cache for the raw-mode context. */
1827typedef PGMRCDYNMAP *PPGMRCDYNMAP;
1828
1829
1830/**
1831 * Mapping cache usage set entry.
1832 *
1833 * @remarks 16-bit ints was chosen as the set is not expected to be used beyond
1834 * the dynamic ring-0 and (to some extent) raw-mode context mapping
1835 * cache. If it's extended to include ring-3, well, then something
1836 * will have be changed here...
1837 */
1838typedef struct PGMMAPSETENTRY
1839{
1840 /** Pointer to the page. */
1841 RTR0PTR pvPage;
1842 /** The mapping cache index. */
1843 uint16_t iPage;
1844 /** The number of references.
1845 * The max is UINT16_MAX - 1. */
1846 uint16_t cRefs;
1847 /** The number inlined references.
1848 * The max is UINT16_MAX - 1. */
1849 uint16_t cInlinedRefs;
1850 /** Unreferences. */
1851 uint16_t cUnrefs;
1852
1853#if HC_ARCH_BITS == 32
1854 uint32_t u32Alignment1;
1855#endif
1856 /** The physical address for this entry. */
1857 RTHCPHYS HCPhys;
1858} PGMMAPSETENTRY;
1859AssertCompileMemberOffset(PGMMAPSETENTRY, iPage, RT_MAX(sizeof(RTR0PTR), sizeof(RTRCPTR)));
1860AssertCompileMemberAlignment(PGMMAPSETENTRY, HCPhys, sizeof(RTHCPHYS));
1861/** Pointer to a mapping cache usage set entry. */
1862typedef PGMMAPSETENTRY *PPGMMAPSETENTRY;
1863
1864/**
1865 * Mapping cache usage set.
1866 *
1867 * This is used in ring-0 and the raw-mode context to track dynamic mappings
1868 * done during exits / traps. The set is
1869 */
1870typedef struct PGMMAPSET
1871{
1872 /** The number of occupied entries.
1873 * This is PGMMAPSET_CLOSED if the set is closed and we're not supposed to do
1874 * dynamic mappings. */
1875 uint32_t cEntries;
1876 /** The start of the current subset.
1877 * This is UINT32_MAX if no subset is currently open. */
1878 uint32_t iSubset;
1879 /** The index of the current CPU, only valid if the set is open. */
1880 int32_t iCpu;
1881 uint32_t alignment;
1882 /** The entries. */
1883 PGMMAPSETENTRY aEntries[64];
1884 /** HCPhys -> iEntry fast lookup table.
1885 * Use PGMMAPSET_HASH for hashing.
1886 * The entries may or may not be valid, check against cEntries. */
1887 uint8_t aiHashTable[128];
1888} PGMMAPSET;
1889AssertCompileSizeAlignment(PGMMAPSET, 8);
1890/** Pointer to the mapping cache set. */
1891typedef PGMMAPSET *PPGMMAPSET;
1892
1893/** PGMMAPSET::cEntries value for a closed set. */
1894#define PGMMAPSET_CLOSED UINT32_C(0xdeadc0fe)
1895
1896/** Hash function for aiHashTable. */
1897#define PGMMAPSET_HASH(HCPhys) (((HCPhys) >> PAGE_SHIFT) & 127)
1898
1899
1900/** @name Context neutral page mapper TLB.
1901 *
1902 * Hoping to avoid some code and bug duplication parts of the GCxxx->CCPtr
1903 * code is writting in a kind of context neutral way. Time will show whether
1904 * this actually makes sense or not...
1905 *
1906 * @todo this needs to be reconsidered and dropped/redone since the ring-0
1907 * context ends up using a global mapping cache on some platforms
1908 * (darwin).
1909 *
1910 * @{ */
1911/** @typedef PPGMPAGEMAPTLB
1912 * The page mapper TLB pointer type for the current context. */
1913/** @typedef PPGMPAGEMAPTLB
1914 * The page mapper TLB entry pointer type for the current context. */
1915/** @typedef PPGMPAGEMAPTLB
1916 * The page mapper TLB entry pointer pointer type for the current context. */
1917/** @def PGM_PAGEMAPTLB_ENTRIES
1918 * The number of TLB entries in the page mapper TLB for the current context. */
1919/** @def PGM_PAGEMAPTLB_IDX
1920 * Calculate the TLB index for a guest physical address.
1921 * @returns The TLB index.
1922 * @param GCPhys The guest physical address. */
1923/** @typedef PPGMPAGEMAP
1924 * Pointer to a page mapper unit for current context. */
1925/** @typedef PPPGMPAGEMAP
1926 * Pointer to a page mapper unit pointer for current context. */
1927#if defined(IN_RING0) && defined(VBOX_WITH_RAM_IN_KERNEL)
1928typedef PPGMPAGER0MAPTLB PPGMPAGEMAPTLB;
1929typedef PPGMPAGER0MAPTLBE PPGMPAGEMAPTLBE;
1930typedef PPGMPAGER0MAPTLBE *PPPGMPAGEMAPTLBE;
1931# define PGM_PAGEMAPTLB_ENTRIES PGM_PAGER0MAPTLB_ENTRIES
1932# define PGM_PAGEMAPTLB_IDX(GCPhys) PGM_PAGER0MAPTLB_IDX(GCPhys)
1933typedef struct PGMCHUNKR0MAP *PPGMPAGEMAP;
1934typedef struct PGMCHUNKR0MAP **PPPGMPAGEMAP;
1935#else
1936typedef PPGMPAGER3MAPTLB PPGMPAGEMAPTLB;
1937typedef PPGMPAGER3MAPTLBE PPGMPAGEMAPTLBE;
1938typedef PPGMPAGER3MAPTLBE *PPPGMPAGEMAPTLBE;
1939# define PGM_PAGEMAPTLB_ENTRIES PGM_PAGER3MAPTLB_ENTRIES
1940# define PGM_PAGEMAPTLB_IDX(GCPhys) PGM_PAGER3MAPTLB_IDX(GCPhys)
1941typedef PPGMCHUNKR3MAP PPGMPAGEMAP;
1942typedef PPPGMCHUNKR3MAP PPPGMPAGEMAP;
1943#endif
1944/** @} */
1945
1946
1947/** @name PGM Pool Indexes.
1948 * Aka. the unique shadow page identifier.
1949 * @{ */
1950/** NIL page pool IDX. */
1951#define NIL_PGMPOOL_IDX 0
1952/** The first normal index. There used to be 5 fictive pages up front, now
1953 * there is only the NIL page. */
1954#define PGMPOOL_IDX_FIRST 1
1955/** The last valid index. (inclusive, 14 bits) */
1956#define PGMPOOL_IDX_LAST 0x3fff
1957/** @} */
1958
1959/** The NIL index for the parent chain. */
1960#define NIL_PGMPOOL_USER_INDEX ((uint16_t)0xffff)
1961#define NIL_PGMPOOL_PRESENT_INDEX ((uint16_t)0xffff)
1962
1963/**
1964 * Node in the chain linking a shadowed page to it's parent (user).
1965 */
1966#pragma pack(1)
1967typedef struct PGMPOOLUSER
1968{
1969 /** The index to the next item in the chain. NIL_PGMPOOL_USER_INDEX is no next. */
1970 uint16_t iNext;
1971 /** The user page index. */
1972 uint16_t iUser;
1973 /** Index into the user table. */
1974 uint32_t iUserTable;
1975} PGMPOOLUSER, *PPGMPOOLUSER;
1976typedef const PGMPOOLUSER *PCPGMPOOLUSER;
1977#pragma pack()
1978
1979
1980/** The NIL index for the phys ext chain. */
1981#define NIL_PGMPOOL_PHYSEXT_INDEX ((uint16_t)0xffff)
1982/** The NIL pte index for a phys ext chain slot. */
1983#define NIL_PGMPOOL_PHYSEXT_IDX_PTE ((uint16_t)0xffff)
1984
1985/**
1986 * Node in the chain of physical cross reference extents.
1987 * @todo Calling this an 'extent' is not quite right, find a better name.
1988 * @todo find out the optimal size of the aidx array
1989 */
1990#pragma pack(1)
1991typedef struct PGMPOOLPHYSEXT
1992{
1993 /** The index to the next item in the chain. NIL_PGMPOOL_PHYSEXT_INDEX is no next. */
1994 uint16_t iNext;
1995 /** Alignment. */
1996 uint16_t u16Align;
1997 /** The user page index. */
1998 uint16_t aidx[3];
1999 /** The page table index or NIL_PGMPOOL_PHYSEXT_IDX_PTE if unknown. */
2000 uint16_t apte[3];
2001} PGMPOOLPHYSEXT, *PPGMPOOLPHYSEXT;
2002typedef const PGMPOOLPHYSEXT *PCPGMPOOLPHYSEXT;
2003#pragma pack()
2004
2005
2006/**
2007 * The kind of page that's being shadowed.
2008 */
2009typedef enum PGMPOOLKIND
2010{
2011 /** The virtual invalid 0 entry. */
2012 PGMPOOLKIND_INVALID = 0,
2013 /** The entry is free (=unused). */
2014 PGMPOOLKIND_FREE,
2015
2016 /** Shw: 32-bit page table; Gst: no paging. */
2017 PGMPOOLKIND_32BIT_PT_FOR_PHYS,
2018 /** Shw: 32-bit page table; Gst: 32-bit page table. */
2019 PGMPOOLKIND_32BIT_PT_FOR_32BIT_PT,
2020 /** Shw: 32-bit page table; Gst: 4MB page. */
2021 PGMPOOLKIND_32BIT_PT_FOR_32BIT_4MB,
2022 /** Shw: PAE page table; Gst: no paging. */
2023 PGMPOOLKIND_PAE_PT_FOR_PHYS,
2024 /** Shw: PAE page table; Gst: 32-bit page table. */
2025 PGMPOOLKIND_PAE_PT_FOR_32BIT_PT,
2026 /** Shw: PAE page table; Gst: Half of a 4MB page. */
2027 PGMPOOLKIND_PAE_PT_FOR_32BIT_4MB,
2028 /** Shw: PAE page table; Gst: PAE page table. */
2029 PGMPOOLKIND_PAE_PT_FOR_PAE_PT,
2030 /** Shw: PAE page table; Gst: 2MB page. */
2031 PGMPOOLKIND_PAE_PT_FOR_PAE_2MB,
2032
2033 /** Shw: 32-bit page directory. Gst: 32-bit page directory. */
2034 PGMPOOLKIND_32BIT_PD,
2035 /** Shw: 32-bit page directory. Gst: no paging. */
2036 PGMPOOLKIND_32BIT_PD_PHYS,
2037 /** Shw: PAE page directory 0; Gst: 32-bit page directory. */
2038 PGMPOOLKIND_PAE_PD0_FOR_32BIT_PD,
2039 /** Shw: PAE page directory 1; Gst: 32-bit page directory. */
2040 PGMPOOLKIND_PAE_PD1_FOR_32BIT_PD,
2041 /** Shw: PAE page directory 2; Gst: 32-bit page directory. */
2042 PGMPOOLKIND_PAE_PD2_FOR_32BIT_PD,
2043 /** Shw: PAE page directory 3; Gst: 32-bit page directory. */
2044 PGMPOOLKIND_PAE_PD3_FOR_32BIT_PD,
2045 /** Shw: PAE page directory; Gst: PAE page directory. */
2046 PGMPOOLKIND_PAE_PD_FOR_PAE_PD,
2047 /** Shw: PAE page directory; Gst: no paging. Note: +NP. */
2048 PGMPOOLKIND_PAE_PD_PHYS,
2049
2050 /** Shw: PAE page directory pointer table (legacy, 4 entries); Gst 32 bits paging. */
2051 PGMPOOLKIND_PAE_PDPT_FOR_32BIT,
2052 /** Shw: PAE page directory pointer table (legacy, 4 entries); Gst PAE PDPT. */
2053 PGMPOOLKIND_PAE_PDPT,
2054 /** Shw: PAE page directory pointer table (legacy, 4 entries); Gst: no paging. */
2055 PGMPOOLKIND_PAE_PDPT_PHYS,
2056
2057 /** Shw: 64-bit page directory pointer table; Gst: 64-bit page directory pointer table. */
2058 PGMPOOLKIND_64BIT_PDPT_FOR_64BIT_PDPT,
2059 /** Shw: 64-bit page directory pointer table; Gst: no paging. */
2060 PGMPOOLKIND_64BIT_PDPT_FOR_PHYS,
2061 /** Shw: 64-bit page directory table; Gst: 64-bit page directory table. */
2062 PGMPOOLKIND_64BIT_PD_FOR_64BIT_PD,
2063 /** Shw: 64-bit page directory table; Gst: no paging. */
2064 PGMPOOLKIND_64BIT_PD_FOR_PHYS, /* 24 */
2065
2066 /** Shw: 64-bit PML4; Gst: 64-bit PML4. */
2067 PGMPOOLKIND_64BIT_PML4,
2068
2069 /** Shw: EPT page directory pointer table; Gst: no paging. */
2070 PGMPOOLKIND_EPT_PDPT_FOR_PHYS,
2071 /** Shw: EPT page directory table; Gst: no paging. */
2072 PGMPOOLKIND_EPT_PD_FOR_PHYS,
2073 /** Shw: EPT page table; Gst: no paging. */
2074 PGMPOOLKIND_EPT_PT_FOR_PHYS,
2075
2076 /** Shw: Root Nested paging table. */
2077 PGMPOOLKIND_ROOT_NESTED,
2078
2079 /** The last valid entry. */
2080 PGMPOOLKIND_LAST = PGMPOOLKIND_ROOT_NESTED
2081} PGMPOOLKIND;
2082
2083/**
2084 * The access attributes of the page; only applies to big pages.
2085 */
2086typedef enum
2087{
2088 PGMPOOLACCESS_DONTCARE = 0,
2089 PGMPOOLACCESS_USER_RW,
2090 PGMPOOLACCESS_USER_R,
2091 PGMPOOLACCESS_USER_RW_NX,
2092 PGMPOOLACCESS_USER_R_NX,
2093 PGMPOOLACCESS_SUPERVISOR_RW,
2094 PGMPOOLACCESS_SUPERVISOR_R,
2095 PGMPOOLACCESS_SUPERVISOR_RW_NX,
2096 PGMPOOLACCESS_SUPERVISOR_R_NX
2097} PGMPOOLACCESS;
2098
2099/**
2100 * The tracking data for a page in the pool.
2101 */
2102typedef struct PGMPOOLPAGE
2103{
2104 /** AVL node code with the (HC) physical address of this page. */
2105 AVLOHCPHYSNODECORE Core;
2106 /** Pointer to the R3 mapping of the page. */
2107 R3PTRTYPE(void *) pvPageR3;
2108 /** Pointer to the R0 mapping of the page. */
2109 R0PTRTYPE(void *) pvPageR0;
2110 /** The guest physical address. */
2111 RTGCPHYS GCPhys;
2112 /** The kind of page we're shadowing. (This is really a PGMPOOLKIND enum.) */
2113 uint8_t enmKind;
2114 /** The subkind of page we're shadowing. (This is really a PGMPOOLACCESS enum.) */
2115 uint8_t enmAccess;
2116 /** This supplements enmKind and enmAccess */
2117 bool fA20Enabled : 1;
2118
2119 /** Used to indicate that the page is zeroed. */
2120 bool fZeroed : 1;
2121 /** Used to indicate that a PT has non-global entries. */
2122 bool fSeenNonGlobal : 1;
2123 /** Used to indicate that we're monitoring writes to the guest page. */
2124 bool fMonitored : 1;
2125 /** Used to indicate that the page is in the cache (e.g. in the GCPhys hash).
2126 * (All pages are in the age list.) */
2127 bool fCached : 1;
2128 /** This is used by the R3 access handlers when invoked by an async thread.
2129 * It's a hack required because of REMR3NotifyHandlerPhysicalDeregister. */
2130 bool volatile fReusedFlushPending : 1;
2131 /** Used to mark the page as dirty (write monitoring is temporarily
2132 * off). */
2133 bool fDirty : 1;
2134 bool fPadding1 : 1;
2135 bool fPadding2;
2136
2137 /** The index of this page. */
2138 uint16_t idx;
2139 /** The next entry in the list this page currently resides in.
2140 * It's either in the free list or in the GCPhys hash. */
2141 uint16_t iNext;
2142 /** Head of the user chain. NIL_PGMPOOL_USER_INDEX if not currently in use. */
2143 uint16_t iUserHead;
2144 /** The number of present entries. */
2145 uint16_t cPresent;
2146 /** The first entry in the table which is present. */
2147 uint16_t iFirstPresent;
2148 /** The number of modifications to the monitored page. */
2149 uint16_t cModifications;
2150 /** The next modified page. NIL_PGMPOOL_IDX if tail. */
2151 uint16_t iModifiedNext;
2152 /** The previous modified page. NIL_PGMPOOL_IDX if head. */
2153 uint16_t iModifiedPrev;
2154 /** The next page sharing access handler. NIL_PGMPOOL_IDX if tail. */
2155 uint16_t iMonitoredNext;
2156 /** The previous page sharing access handler. NIL_PGMPOOL_IDX if head. */
2157 uint16_t iMonitoredPrev;
2158 /** The next page in the age list. */
2159 uint16_t iAgeNext;
2160 /** The previous page in the age list. */
2161 uint16_t iAgePrev;
2162 /** Index into PGMPOOL::aDirtyPages if fDirty is set. */
2163 uint8_t idxDirtyEntry;
2164
2165 /** @name Access handler statistics to determine whether the guest is
2166 * (re)initializing a page table.
2167 * @{ */
2168 RTGCPTR GCPtrLastAccessHandlerRip;
2169 RTGCPTR GCPtrLastAccessHandlerFault;
2170 uint64_t cLastAccessHandler;
2171 /** @} */
2172 /** Used to indicate that this page can't be flushed. Important for cr3 root pages or shadow pae pd pages. */
2173 uint32_t volatile cLocked;
2174#if GC_ARCH_BITS == 64
2175 uint32_t u32Alignment3;
2176#endif
2177# ifdef VBOX_STRICT
2178 RTGCPTR GCPtrDirtyFault;
2179# endif
2180} PGMPOOLPAGE;
2181/** Pointer to a pool page. */
2182typedef PGMPOOLPAGE *PPGMPOOLPAGE;
2183/** Pointer to a const pool page. */
2184typedef PGMPOOLPAGE const *PCPGMPOOLPAGE;
2185/** Pointer to a pool page pointer. */
2186typedef PGMPOOLPAGE **PPPGMPOOLPAGE;
2187
2188
2189/** The hash table size. */
2190# define PGMPOOL_HASH_SIZE 0x40
2191/** The hash function. */
2192# define PGMPOOL_HASH(GCPhys) ( ((GCPhys) >> PAGE_SHIFT) & (PGMPOOL_HASH_SIZE - 1) )
2193
2194
2195/**
2196 * The shadow page pool instance data.
2197 *
2198 * It's all one big allocation made at init time, except for the
2199 * pages that is. The user nodes follows immediately after the
2200 * page structures.
2201 */
2202typedef struct PGMPOOL
2203{
2204 /** The VM handle - R3 Ptr. */
2205 PVMR3 pVMR3;
2206 /** The VM handle - R0 Ptr. */
2207 R0PTRTYPE(PVMCC) pVMR0;
2208 /** The max pool size. This includes the special IDs. */
2209 uint16_t cMaxPages;
2210 /** The current pool size. */
2211 uint16_t cCurPages;
2212 /** The head of the free page list. */
2213 uint16_t iFreeHead;
2214 /* Padding. */
2215 uint16_t u16Padding;
2216 /** Head of the chain of free user nodes. */
2217 uint16_t iUserFreeHead;
2218 /** The number of user nodes we've allocated. */
2219 uint16_t cMaxUsers;
2220 /** The number of present page table entries in the entire pool. */
2221 uint32_t cPresent;
2222 /** Pointer to the array of user nodes - R3 pointer. */
2223 R3PTRTYPE(PPGMPOOLUSER) paUsersR3;
2224 /** Pointer to the array of user nodes - R0 pointer. */
2225 R0PTRTYPE(PPGMPOOLUSER) paUsersR0;
2226 /** Head of the chain of free phys ext nodes. */
2227 uint16_t iPhysExtFreeHead;
2228 /** The number of user nodes we've allocated. */
2229 uint16_t cMaxPhysExts;
2230 uint32_t u32Padding0b;
2231 /** Pointer to the array of physical xref extent nodes - R3 pointer. */
2232 R3PTRTYPE(PPGMPOOLPHYSEXT) paPhysExtsR3;
2233 /** Pointer to the array of physical xref extent nodes - R0 pointer. */
2234 R0PTRTYPE(PPGMPOOLPHYSEXT) paPhysExtsR0;
2235 /** Hash table for GCPhys addresses. */
2236 uint16_t aiHash[PGMPOOL_HASH_SIZE];
2237 /** The head of the age list. */
2238 uint16_t iAgeHead;
2239 /** The tail of the age list. */
2240 uint16_t iAgeTail;
2241 /** Set if the cache is enabled. */
2242 bool fCacheEnabled;
2243 /** Alignment padding. */
2244 bool afPadding1[3];
2245 /** Head of the list of modified pages. */
2246 uint16_t iModifiedHead;
2247 /** The current number of modified pages. */
2248 uint16_t cModifiedPages;
2249 /** Physical access handler type registration handle. */
2250 PGMPHYSHANDLERTYPE hAccessHandlerType;
2251 /** Next available slot (in aDirtyPages). */
2252 uint32_t idxFreeDirtyPage;
2253 /** Number of active dirty pages. */
2254 uint32_t cDirtyPages;
2255 /** Array of current dirty pgm pool page indices. */
2256 uint16_t aidxDirtyPages[16];
2257 /** Array running in parallel to aidxDirtyPages with the page data. */
2258 struct
2259 {
2260 uint64_t aPage[512];
2261 } aDirtyPages[16];
2262
2263 /** The number of pages currently in use. */
2264 uint16_t cUsedPages;
2265#ifdef VBOX_WITH_STATISTICS
2266 /** The high water mark for cUsedPages. */
2267 uint16_t cUsedPagesHigh;
2268 uint32_t Alignment1; /**< Align the next member on a 64-bit boundary. */
2269 /** Profiling pgmPoolAlloc(). */
2270 STAMPROFILEADV StatAlloc;
2271 /** Profiling pgmR3PoolClearDoIt(). */
2272 STAMPROFILE StatClearAll;
2273 /** Profiling pgmR3PoolReset(). */
2274 STAMPROFILE StatR3Reset;
2275 /** Profiling pgmPoolFlushPage(). */
2276 STAMPROFILE StatFlushPage;
2277 /** Profiling pgmPoolFree(). */
2278 STAMPROFILE StatFree;
2279 /** Counting explicit flushes by PGMPoolFlushPage(). */
2280 STAMCOUNTER StatForceFlushPage;
2281 /** Counting explicit flushes of dirty pages by PGMPoolFlushPage(). */
2282 STAMCOUNTER StatForceFlushDirtyPage;
2283 /** Counting flushes for reused pages. */
2284 STAMCOUNTER StatForceFlushReused;
2285 /** Profiling time spent zeroing pages. */
2286 STAMPROFILE StatZeroPage;
2287 /** Profiling of pgmPoolTrackDeref. */
2288 STAMPROFILE StatTrackDeref;
2289 /** Profiling pgmTrackFlushGCPhysPT. */
2290 STAMPROFILE StatTrackFlushGCPhysPT;
2291 /** Profiling pgmTrackFlushGCPhysPTs. */
2292 STAMPROFILE StatTrackFlushGCPhysPTs;
2293 /** Profiling pgmTrackFlushGCPhysPTsSlow. */
2294 STAMPROFILE StatTrackFlushGCPhysPTsSlow;
2295 /** Number of times we've been out of user records. */
2296 STAMCOUNTER StatTrackFreeUpOneUser;
2297 /** Nr of flushed entries. */
2298 STAMCOUNTER StatTrackFlushEntry;
2299 /** Nr of updated entries. */
2300 STAMCOUNTER StatTrackFlushEntryKeep;
2301 /** Profiling deref activity related tracking GC physical pages. */
2302 STAMPROFILE StatTrackDerefGCPhys;
2303 /** Number of linear searches for a HCPhys in the ram ranges. */
2304 STAMCOUNTER StatTrackLinearRamSearches;
2305 /** The number of failing pgmPoolTrackPhysExtAlloc calls. */
2306 STAMCOUNTER StamTrackPhysExtAllocFailures;
2307
2308 /** Profiling the RC/R0 \#PF access handler. */
2309 STAMPROFILE StatMonitorPfRZ;
2310 /** Profiling the RC/R0 access we've handled (except REP STOSD). */
2311 STAMPROFILE StatMonitorPfRZHandled;
2312 /** Times we've failed interpreting the instruction. */
2313 STAMCOUNTER StatMonitorPfRZEmulateInstr;
2314 /** Profiling the pgmPoolFlushPage calls made from the RC/R0 access handler. */
2315 STAMPROFILE StatMonitorPfRZFlushPage;
2316 /** Times we've detected a page table reinit. */
2317 STAMCOUNTER StatMonitorPfRZFlushReinit;
2318 /** Counting flushes for pages that are modified too often. */
2319 STAMCOUNTER StatMonitorPfRZFlushModOverflow;
2320 /** Times we've detected fork(). */
2321 STAMCOUNTER StatMonitorPfRZFork;
2322 /** Times we've failed interpreting a patch code instruction. */
2323 STAMCOUNTER StatMonitorPfRZIntrFailPatch1;
2324 /** Times we've failed interpreting a patch code instruction during flushing. */
2325 STAMCOUNTER StatMonitorPfRZIntrFailPatch2;
2326 /** The number of times we've seen rep prefixes we can't handle. */
2327 STAMCOUNTER StatMonitorPfRZRepPrefix;
2328 /** Profiling the REP STOSD cases we've handled. */
2329 STAMPROFILE StatMonitorPfRZRepStosd;
2330
2331 /** Profiling the R0/RC regular access handler. */
2332 STAMPROFILE StatMonitorRZ;
2333 /** Profiling the pgmPoolFlushPage calls made from the regular access handler in R0/RC. */
2334 STAMPROFILE StatMonitorRZFlushPage;
2335 /** Per access size counts indexed by size minus 1, last for larger. */
2336 STAMCOUNTER aStatMonitorRZSizes[16+3];
2337 /** Missaligned access counts indexed by offset - 1. */
2338 STAMCOUNTER aStatMonitorRZMisaligned[7];
2339
2340 /** Nr of handled PT faults. */
2341 STAMCOUNTER StatMonitorRZFaultPT;
2342 /** Nr of handled PD faults. */
2343 STAMCOUNTER StatMonitorRZFaultPD;
2344 /** Nr of handled PDPT faults. */
2345 STAMCOUNTER StatMonitorRZFaultPDPT;
2346 /** Nr of handled PML4 faults. */
2347 STAMCOUNTER StatMonitorRZFaultPML4;
2348
2349 /** Profiling the R3 access handler. */
2350 STAMPROFILE StatMonitorR3;
2351 /** Profiling the pgmPoolFlushPage calls made from the R3 access handler. */
2352 STAMPROFILE StatMonitorR3FlushPage;
2353 /** Per access size counts indexed by size minus 1, last for larger. */
2354 STAMCOUNTER aStatMonitorR3Sizes[16+3];
2355 /** Missaligned access counts indexed by offset - 1. */
2356 STAMCOUNTER aStatMonitorR3Misaligned[7];
2357 /** Nr of handled PT faults. */
2358 STAMCOUNTER StatMonitorR3FaultPT;
2359 /** Nr of handled PD faults. */
2360 STAMCOUNTER StatMonitorR3FaultPD;
2361 /** Nr of handled PDPT faults. */
2362 STAMCOUNTER StatMonitorR3FaultPDPT;
2363 /** Nr of handled PML4 faults. */
2364 STAMCOUNTER StatMonitorR3FaultPML4;
2365
2366 /** Times we've called pgmPoolResetDirtyPages (and there were dirty page). */
2367 STAMCOUNTER StatResetDirtyPages;
2368 /** Times we've called pgmPoolAddDirtyPage. */
2369 STAMCOUNTER StatDirtyPage;
2370 /** Times we've had to flush duplicates for dirty page management. */
2371 STAMCOUNTER StatDirtyPageDupFlush;
2372 /** Times we've had to flush because of overflow. */
2373 STAMCOUNTER StatDirtyPageOverFlowFlush;
2374
2375 /** The high water mark for cModifiedPages. */
2376 uint16_t cModifiedPagesHigh;
2377 uint16_t Alignment2[3]; /**< Align the next member on a 64-bit boundary. */
2378
2379 /** The number of cache hits. */
2380 STAMCOUNTER StatCacheHits;
2381 /** The number of cache misses. */
2382 STAMCOUNTER StatCacheMisses;
2383 /** The number of times we've got a conflict of 'kind' in the cache. */
2384 STAMCOUNTER StatCacheKindMismatches;
2385 /** Number of times we've been out of pages. */
2386 STAMCOUNTER StatCacheFreeUpOne;
2387 /** The number of cacheable allocations. */
2388 STAMCOUNTER StatCacheCacheable;
2389 /** The number of uncacheable allocations. */
2390 STAMCOUNTER StatCacheUncacheable;
2391#else
2392 uint32_t Alignment3; /**< Align the next member on a 64-bit boundary. */
2393#endif
2394 /** Profiling PGMR0PoolGrow(). */
2395 STAMPROFILE StatGrow;
2396 /** The AVL tree for looking up a page by its HC physical address. */
2397 AVLOHCPHYSTREE HCPhysTree;
2398 uint32_t Alignment4; /**< Align the next member on a 64-bit boundary. */
2399 /** Array of pages. (cMaxPages in length)
2400 * The Id is the index into thist array.
2401 */
2402 PGMPOOLPAGE aPages[PGMPOOL_IDX_FIRST];
2403} PGMPOOL, *PPGMPOOL, **PPPGMPOOL;
2404AssertCompileMemberAlignment(PGMPOOL, iModifiedHead, 8);
2405AssertCompileMemberAlignment(PGMPOOL, aDirtyPages, 8);
2406AssertCompileMemberAlignment(PGMPOOL, cUsedPages, 8);
2407#ifdef VBOX_WITH_STATISTICS
2408AssertCompileMemberAlignment(PGMPOOL, StatAlloc, 8);
2409#endif
2410AssertCompileMemberAlignment(PGMPOOL, aPages, 8);
2411
2412
2413/** @def PGMPOOL_PAGE_2_PTR
2414 * Maps a pool page pool into the current context.
2415 *
2416 * @returns VBox status code.
2417 * @param a_pVM Pointer to the VM.
2418 * @param a_pPage The pool page.
2419 *
2420 * @remark In RC this uses PGMGCDynMapHCPage(), so it will consume of the
2421 * small page window employeed by that function. Be careful.
2422 * @remark There is no need to assert on the result.
2423 */
2424#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE_IN_R0
2425# define PGMPOOL_PAGE_2_PTR(a_pVM, a_pPage) pgmPoolMapPageInlined((a_pVM), (a_pPage) RTLOG_COMMA_SRC_POS)
2426#elif defined(VBOX_STRICT) || 1 /* temporarily going strict here */
2427# define PGMPOOL_PAGE_2_PTR(a_pVM, a_pPage) pgmPoolMapPageStrict(a_pPage, __FUNCTION__)
2428DECLINLINE(void *) pgmPoolMapPageStrict(PPGMPOOLPAGE a_pPage, const char *pszCaller)
2429{
2430 RT_NOREF(pszCaller);
2431 AssertPtr(a_pPage);
2432 AssertMsg(RT_VALID_PTR(a_pPage->CTX_SUFF(pvPage)),
2433 ("enmKind=%d idx=%#x HCPhys=%RHp GCPhys=%RGp pvPageR3=%p pvPageR0=%p caller=%s\n",
2434 a_pPage->enmKind, a_pPage->idx, a_pPage->Core.Key, a_pPage->GCPhys, a_pPage->pvPageR3, a_pPage->pvPageR0, pszCaller));
2435 return a_pPage->CTX_SUFF(pvPage);
2436}
2437#else
2438# define PGMPOOL_PAGE_2_PTR(pVM, a_pPage) ((a_pPage)->CTX_SUFF(pvPage))
2439#endif
2440
2441
2442/** @def PGMPOOL_PAGE_2_PTR_V2
2443 * Maps a pool page pool into the current context, taking both VM and VMCPU.
2444 *
2445 * @returns VBox status code.
2446 * @param a_pVM Pointer to the VM.
2447 * @param a_pVCpu The current CPU.
2448 * @param a_pPage The pool page.
2449 *
2450 * @remark In RC this uses PGMGCDynMapHCPage(), so it will consume of the
2451 * small page window employeed by that function. Be careful.
2452 * @remark There is no need to assert on the result.
2453 */
2454#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE_IN_R0
2455# define PGMPOOL_PAGE_2_PTR_V2(a_pVM, a_pVCpu, a_pPage) pgmPoolMapPageV2Inlined((a_pVM), (a_pVCpu), (a_pPage) RTLOG_COMMA_SRC_POS)
2456#else
2457# define PGMPOOL_PAGE_2_PTR_V2(a_pVM, a_pVCpu, a_pPage) PGMPOOL_PAGE_2_PTR((a_pVM), (a_pPage))
2458#endif
2459
2460
2461/** @name Per guest page tracking data.
2462 * This is currently as a 16-bit word in the PGMPAGE structure, the idea though
2463 * is to use more bits for it and split it up later on. But for now we'll play
2464 * safe and change as little as possible.
2465 *
2466 * The 16-bit word has two parts:
2467 *
2468 * The first 14-bit forms the @a idx field. It is either the index of a page in
2469 * the shadow page pool, or and index into the extent list.
2470 *
2471 * The 2 topmost bits makes up the @a cRefs field, which counts the number of
2472 * shadow page pool references to the page. If cRefs equals
2473 * PGMPOOL_CREFS_PHYSEXT, then the @a idx field is an indext into the extent
2474 * (misnomer) table and not the shadow page pool.
2475 *
2476 * See PGM_PAGE_GET_TRACKING and PGM_PAGE_SET_TRACKING for how to get and set
2477 * the 16-bit word.
2478 *
2479 * @{ */
2480/** The shift count for getting to the cRefs part. */
2481#define PGMPOOL_TD_CREFS_SHIFT 14
2482/** The mask applied after shifting the tracking data down by
2483 * PGMPOOL_TD_CREFS_SHIFT. */
2484#define PGMPOOL_TD_CREFS_MASK 0x3
2485/** The cRefs value used to indicate that the idx is the head of a
2486 * physical cross reference list. */
2487#define PGMPOOL_TD_CREFS_PHYSEXT PGMPOOL_TD_CREFS_MASK
2488/** The shift used to get idx. */
2489#define PGMPOOL_TD_IDX_SHIFT 0
2490/** The mask applied to the idx after shifting down by PGMPOOL_TD_IDX_SHIFT. */
2491#define PGMPOOL_TD_IDX_MASK 0x3fff
2492/** The idx value when we're out of of PGMPOOLPHYSEXT entries or/and there are
2493 * simply too many mappings of this page. */
2494#define PGMPOOL_TD_IDX_OVERFLOWED PGMPOOL_TD_IDX_MASK
2495
2496/** @def PGMPOOL_TD_MAKE
2497 * Makes a 16-bit tracking data word.
2498 *
2499 * @returns tracking data.
2500 * @param cRefs The @a cRefs field. Must be within bounds!
2501 * @param idx The @a idx field. Must also be within bounds! */
2502#define PGMPOOL_TD_MAKE(cRefs, idx) ( ((cRefs) << PGMPOOL_TD_CREFS_SHIFT) | (idx) )
2503
2504/** @def PGMPOOL_TD_GET_CREFS
2505 * Get the @a cRefs field from a tracking data word.
2506 *
2507 * @returns The @a cRefs field
2508 * @param u16 The tracking data word.
2509 * @remarks This will only return 1 or PGMPOOL_TD_CREFS_PHYSEXT for a
2510 * non-zero @a u16. */
2511#define PGMPOOL_TD_GET_CREFS(u16) ( ((u16) >> PGMPOOL_TD_CREFS_SHIFT) & PGMPOOL_TD_CREFS_MASK )
2512
2513/** @def PGMPOOL_TD_GET_IDX
2514 * Get the @a idx field from a tracking data word.
2515 *
2516 * @returns The @a idx field
2517 * @param u16 The tracking data word. */
2518#define PGMPOOL_TD_GET_IDX(u16) ( ((u16) >> PGMPOOL_TD_IDX_SHIFT) & PGMPOOL_TD_IDX_MASK )
2519/** @} */
2520
2521
2522
2523/** @name A20 gate macros
2524 * @{ */
2525#define PGM_WITH_A20
2526#ifdef PGM_WITH_A20
2527# define PGM_A20_IS_ENABLED(a_pVCpu) ((a_pVCpu)->pgm.s.fA20Enabled)
2528# define PGM_A20_APPLY(a_pVCpu, a_GCPhys) ((a_GCPhys) & (a_pVCpu)->pgm.s.GCPhysA20Mask)
2529# define PGM_A20_APPLY_TO_VAR(a_pVCpu, a_GCPhysVar) \
2530 do { a_GCPhysVar &= (a_pVCpu)->pgm.s.GCPhysA20Mask; } while (0)
2531# define PGM_A20_ASSERT_MASKED(pVCpu, a_GCPhys) Assert(PGM_A20_APPLY(pVCpu, a_GCPhys) == (a_GCPhys))
2532#else
2533# define PGM_A20_IS_ENABLED(a_pVCpu) (true)
2534# define PGM_A20_APPLY(a_pVCpu, a_GCPhys) (a_GCPhys)
2535# define PGM_A20_APPLY_TO_VAR(a_pVCpu, a_GCPhysVar) do { } while (0)
2536# define PGM_A20_ASSERT_MASKED(pVCpu, a_GCPhys) do { } while (0)
2537#endif
2538/** @} */
2539
2540
2541/**
2542 * Roots and anchors for trees and list employing self relative offsets as
2543 * pointers.
2544 *
2545 * When using self-relative offsets instead of pointers, the offsets needs to be
2546 * the same in all offsets. Thus the roots and anchors needs to live on the
2547 * hyper heap just like the nodes.
2548 */
2549typedef struct PGMTREES
2550{
2551 /** List of physical access handler types (offset pointers) of type
2552 * PGMPHYSHANDLERTYPEINT. This is needed for relocations. */
2553 RTLISTOFF32ANCHOR HeadPhysHandlerTypes;
2554 /** Physical access handlers (AVL range+offsetptr tree). */
2555 AVLROGCPHYSTREE PhysHandlers;
2556} PGMTREES;
2557/** Pointer to PGM trees. */
2558typedef PGMTREES *PPGMTREES;
2559
2560
2561/**
2562 * Page fault guest state for the AMD64 paging mode.
2563 */
2564typedef struct PGMPTWALKCORE
2565{
2566 /** The guest virtual address that is being resolved by the walk
2567 * (input). */
2568 RTGCPTR GCPtr;
2569
2570 /** The guest physical address that is the result of the walk.
2571 * @remarks only valid if fSucceeded is set. */
2572 RTGCPHYS GCPhys;
2573
2574 /** Set if the walk succeeded, i.d. GCPhys is valid. */
2575 bool fSucceeded;
2576 /** The level problem arrised at.
2577 * PTE is level 1, PDE is level 2, PDPE is level 3, PML4 is level 4, CR3 is
2578 * level 8. This is 0 on success. */
2579 uint8_t uLevel;
2580 /** Set if the page isn't present. */
2581 bool fNotPresent;
2582 /** Encountered a bad physical address. */
2583 bool fBadPhysAddr;
2584 /** Set if there was reserved bit violations. */
2585 bool fRsvdError;
2586 /** Set if it involves a big page (2/4 MB). */
2587 bool fBigPage;
2588 /** Set if it involves a gigantic page (1 GB). */
2589 bool fGigantPage;
2590 /** The effective X86_PTE_US flag for the address. */
2591 bool fEffectiveUS;
2592 /** The effective X86_PTE_RW flag for the address. */
2593 bool fEffectiveRW;
2594 /** The effective X86_PTE_NX flag for the address. */
2595 bool fEffectiveNX;
2596 bool afPadding1[2];
2597 /** Effective flags thus far: RW, US, PWT, PCD, A, ~NX >> 63.
2598 * The NX bit is inverted and shifted down 63 places to bit 0. */
2599 uint32_t fEffective;
2600} PGMPTWALKCORE;
2601
2602/** @name PGMPTWALKCORE::fEffective bits.
2603 * @{ */
2604/** Effective execute bit (!NX). */
2605#define PGMPTWALK_EFF_X UINT32_C(1)
2606/** Effective write access bit. */
2607#define PGMPTWALK_EFF_RW X86_PTE_RW
2608/** Effective user-mode access bit. */
2609#define PGMPTWALK_EFF_US X86_PTE_US
2610/** Effective write through cache bit. */
2611#define PGMPTWALK_EFF_PWT X86_PTE_PWT
2612/** Effective cache disabled bit. */
2613#define PGMPTWALK_EFF_PCD X86_PTE_PCD
2614/** Effective accessed bit. */
2615#define PGMPTWALK_EFF_A X86_PTE_A
2616/** The dirty bit of the final entry. */
2617#define PGMPTWALK_EFF_D X86_PTE_D
2618/** The PAT bit of the final entry. */
2619#define PGMPTWALK_EFF_PAT X86_PTE_PAT
2620/** The global bit of the final entry. */
2621#define PGMPTWALK_EFF_G X86_PTE_G
2622/** @} */
2623
2624
2625/**
2626 * Guest page table walk for the AMD64 mode.
2627 */
2628typedef struct PGMPTWALKGSTAMD64
2629{
2630 /** The common core. */
2631 PGMPTWALKCORE Core;
2632
2633 PX86PML4 pPml4;
2634 PX86PML4E pPml4e;
2635 X86PML4E Pml4e;
2636
2637 PX86PDPT pPdpt;
2638 PX86PDPE pPdpe;
2639 X86PDPE Pdpe;
2640
2641 PX86PDPAE pPd;
2642 PX86PDEPAE pPde;
2643 X86PDEPAE Pde;
2644
2645 PX86PTPAE pPt;
2646 PX86PTEPAE pPte;
2647 X86PTEPAE Pte;
2648} PGMPTWALKGSTAMD64;
2649/** Pointer to a AMD64 guest page table walk. */
2650typedef PGMPTWALKGSTAMD64 *PPGMPTWALKGSTAMD64;
2651/** Pointer to a const AMD64 guest page table walk. */
2652typedef PGMPTWALKGSTAMD64 const *PCPGMPTWALKGSTAMD64;
2653
2654/**
2655 * Guest page table walk for the PAE mode.
2656 */
2657typedef struct PGMPTWALKGSTPAE
2658{
2659 /** The common core. */
2660 PGMPTWALKCORE Core;
2661
2662 PX86PDPT pPdpt;
2663 PX86PDPE pPdpe;
2664 X86PDPE Pdpe;
2665
2666 PX86PDPAE pPd;
2667 PX86PDEPAE pPde;
2668 X86PDEPAE Pde;
2669
2670 PX86PTPAE pPt;
2671 PX86PTEPAE pPte;
2672 X86PTEPAE Pte;
2673} PGMPTWALKGSTPAE;
2674/** Pointer to a PAE guest page table walk. */
2675typedef PGMPTWALKGSTPAE *PPGMPTWALKGSTPAE;
2676/** Pointer to a const AMD64 guest page table walk. */
2677typedef PGMPTWALKGSTPAE const *PCPGMPTWALKGSTPAE;
2678
2679/**
2680 * Guest page table walk for the 32-bit mode.
2681 */
2682typedef struct PGMPTWALKGST32BIT
2683{
2684 /** The common core. */
2685 PGMPTWALKCORE Core;
2686
2687 PX86PD pPd;
2688 PX86PDE pPde;
2689 X86PDE Pde;
2690
2691 PX86PT pPt;
2692 PX86PTE pPte;
2693 X86PTE Pte;
2694} PGMPTWALKGST32BIT;
2695/** Pointer to a 32-bit guest page table walk. */
2696typedef PGMPTWALKGST32BIT *PPGMPTWALKGST32BIT;
2697/** Pointer to a const 32-bit guest page table walk. */
2698typedef PGMPTWALKGST32BIT const *PCPGMPTWALKGST32BIT;
2699
2700/**
2701 * Which part of PGMPTWALKGST that is valid.
2702 */
2703typedef enum PGMPTWALKGSTTYPE
2704{
2705 /** Customary invalid 0 value. */
2706 PGMPTWALKGSTTYPE_INVALID = 0,
2707 /** PGMPTWALKGST::u.Amd64 is valid. */
2708 PGMPTWALKGSTTYPE_AMD64,
2709 /** PGMPTWALKGST::u.Pae is valid. */
2710 PGMPTWALKGSTTYPE_PAE,
2711 /** PGMPTWALKGST::u.Legacy is valid. */
2712 PGMPTWALKGSTTYPE_32BIT,
2713 /** Customary 32-bit type hack. */
2714 PGMPTWALKGSTTYPE_32BIT_HACK = 0x7fff0000
2715} PGMPTWALKGSTTYPE;
2716
2717/**
2718 * Combined guest page table walk result.
2719 */
2720typedef struct PGMPTWALKGST
2721{
2722 union
2723 {
2724 /** The page walker core - always valid. */
2725 PGMPTWALKCORE Core;
2726 /** The page walker for AMD64. */
2727 PGMPTWALKGSTAMD64 Amd64;
2728 /** The page walker for PAE (32-bit). */
2729 PGMPTWALKGSTPAE Pae;
2730 /** The page walker for 32-bit paging (called legacy due to C naming
2731 * convension). */
2732 PGMPTWALKGST32BIT Legacy;
2733 } u;
2734 /** Indicates which part of the union is valid. */
2735 PGMPTWALKGSTTYPE enmType;
2736} PGMPTWALKGST;
2737/** Pointer to a combined guest page table walk result. */
2738typedef PGMPTWALKGST *PPGMPTWALKGST;
2739/** Pointer to a read-only combined guest page table walk result. */
2740typedef PGMPTWALKGST const *PCPGMPTWALKGST;
2741
2742
2743/** @name Paging mode macros
2744 * @{
2745 */
2746#ifdef IN_RING3
2747# define PGM_CTX(a,b) a##R3##b
2748# define PGM_CTX_STR(a,b) a "R3" b
2749# define PGM_CTX_DECL(type) DECLCALLBACK(type)
2750#elif defined(IN_RING0)
2751# define PGM_CTX(a,b) a##R0##b
2752# define PGM_CTX_STR(a,b) a "R0" b
2753# define PGM_CTX_DECL(type) VMMDECL(type)
2754#else
2755# error "Not IN_RING3 or IN_RING0!"
2756#endif
2757
2758#define PGM_GST_NAME_REAL(name) PGM_CTX(pgm,GstReal##name)
2759#define PGM_GST_NAME_RC_REAL_STR(name) "pgmRCGstReal" #name
2760#define PGM_GST_NAME_R0_REAL_STR(name) "pgmR0GstReal" #name
2761#define PGM_GST_NAME_PROT(name) PGM_CTX(pgm,GstProt##name)
2762#define PGM_GST_NAME_RC_PROT_STR(name) "pgmRCGstProt" #name
2763#define PGM_GST_NAME_R0_PROT_STR(name) "pgmR0GstProt" #name
2764#define PGM_GST_NAME_32BIT(name) PGM_CTX(pgm,Gst32Bit##name)
2765#define PGM_GST_NAME_RC_32BIT_STR(name) "pgmRCGst32Bit" #name
2766#define PGM_GST_NAME_R0_32BIT_STR(name) "pgmR0Gst32Bit" #name
2767#define PGM_GST_NAME_PAE(name) PGM_CTX(pgm,GstPAE##name)
2768#define PGM_GST_NAME_RC_PAE_STR(name) "pgmRCGstPAE" #name
2769#define PGM_GST_NAME_R0_PAE_STR(name) "pgmR0GstPAE" #name
2770#define PGM_GST_NAME_AMD64(name) PGM_CTX(pgm,GstAMD64##name)
2771#define PGM_GST_NAME_RC_AMD64_STR(name) "pgmRCGstAMD64" #name
2772#define PGM_GST_NAME_R0_AMD64_STR(name) "pgmR0GstAMD64" #name
2773#define PGM_GST_DECL(type, name) PGM_CTX_DECL(type) PGM_GST_NAME(name)
2774
2775#define PGM_SHW_NAME_32BIT(name) PGM_CTX(pgm,Shw32Bit##name)
2776#define PGM_SHW_NAME_RC_32BIT_STR(name) "pgmRCShw32Bit" #name
2777#define PGM_SHW_NAME_R0_32BIT_STR(name) "pgmR0Shw32Bit" #name
2778#define PGM_SHW_NAME_PAE(name) PGM_CTX(pgm,ShwPAE##name)
2779#define PGM_SHW_NAME_RC_PAE_STR(name) "pgmRCShwPAE" #name
2780#define PGM_SHW_NAME_R0_PAE_STR(name) "pgmR0ShwPAE" #name
2781#define PGM_SHW_NAME_AMD64(name) PGM_CTX(pgm,ShwAMD64##name)
2782#define PGM_SHW_NAME_RC_AMD64_STR(name) "pgmRCShwAMD64" #name
2783#define PGM_SHW_NAME_R0_AMD64_STR(name) "pgmR0ShwAMD64" #name
2784#define PGM_SHW_NAME_NESTED_32BIT(name) PGM_CTX(pgm,ShwNested32Bit##name)
2785#define PGM_SHW_NAME_RC_NESTED_32BIT_STR(name) "pgmRCShwNested32Bit" #name
2786#define PGM_SHW_NAME_R0_NESTED_32BIT_STR(name) "pgmR0ShwNested32Bit" #name
2787#define PGM_SHW_NAME_NESTED_PAE(name) PGM_CTX(pgm,ShwNestedPAE##name)
2788#define PGM_SHW_NAME_RC_NESTED_PAE_STR(name) "pgmRCShwNestedPAE" #name
2789#define PGM_SHW_NAME_R0_NESTED_PAE_STR(name) "pgmR0ShwNestedPAE" #name
2790#define PGM_SHW_NAME_NESTED_AMD64(name) PGM_CTX(pgm,ShwNestedAMD64##name)
2791#define PGM_SHW_NAME_RC_NESTED_AMD64_STR(name) "pgmRCShwNestedAMD64" #name
2792#define PGM_SHW_NAME_R0_NESTED_AMD64_STR(name) "pgmR0ShwNestedAMD64" #name
2793#define PGM_SHW_NAME_EPT(name) PGM_CTX(pgm,ShwEPT##name)
2794#define PGM_SHW_NAME_RC_EPT_STR(name) "pgmRCShwEPT" #name
2795#define PGM_SHW_NAME_R0_EPT_STR(name) "pgmR0ShwEPT" #name
2796#define PGM_SHW_NAME_NONE(name) PGM_CTX(pgm,ShwNone##name)
2797#define PGM_SHW_NAME_RC_NONE_STR(name) "pgmRCShwNone" #name
2798#define PGM_SHW_NAME_R0_NONE_STR(name) "pgmR0ShwNone" #name
2799#define PGM_SHW_DECL(type, name) PGM_CTX_DECL(type) PGM_SHW_NAME(name)
2800
2801/* Shw_Gst */
2802#define PGM_BTH_NAME_32BIT_REAL(name) PGM_CTX(pgm,Bth32BitReal##name)
2803#define PGM_BTH_NAME_32BIT_PROT(name) PGM_CTX(pgm,Bth32BitProt##name)
2804#define PGM_BTH_NAME_32BIT_32BIT(name) PGM_CTX(pgm,Bth32Bit32Bit##name)
2805#define PGM_BTH_NAME_PAE_REAL(name) PGM_CTX(pgm,BthPAEReal##name)
2806#define PGM_BTH_NAME_PAE_PROT(name) PGM_CTX(pgm,BthPAEProt##name)
2807#define PGM_BTH_NAME_PAE_32BIT(name) PGM_CTX(pgm,BthPAE32Bit##name)
2808#define PGM_BTH_NAME_PAE_PAE(name) PGM_CTX(pgm,BthPAEPAE##name)
2809#define PGM_BTH_NAME_AMD64_PROT(name) PGM_CTX(pgm,BthAMD64Prot##name)
2810#define PGM_BTH_NAME_AMD64_AMD64(name) PGM_CTX(pgm,BthAMD64AMD64##name)
2811#define PGM_BTH_NAME_NESTED_32BIT_REAL(name) PGM_CTX(pgm,BthNested32BitReal##name)
2812#define PGM_BTH_NAME_NESTED_32BIT_PROT(name) PGM_CTX(pgm,BthNested32BitProt##name)
2813#define PGM_BTH_NAME_NESTED_32BIT_32BIT(name) PGM_CTX(pgm,BthNested32Bit32Bit##name)
2814#define PGM_BTH_NAME_NESTED_32BIT_PAE(name) PGM_CTX(pgm,BthNested32BitPAE##name)
2815#define PGM_BTH_NAME_NESTED_32BIT_AMD64(name) PGM_CTX(pgm,BthNested32BitAMD64##name)
2816#define PGM_BTH_NAME_NESTED_PAE_REAL(name) PGM_CTX(pgm,BthNestedPAEReal##name)
2817#define PGM_BTH_NAME_NESTED_PAE_PROT(name) PGM_CTX(pgm,BthNestedPAEProt##name)
2818#define PGM_BTH_NAME_NESTED_PAE_32BIT(name) PGM_CTX(pgm,BthNestedPAE32Bit##name)
2819#define PGM_BTH_NAME_NESTED_PAE_PAE(name) PGM_CTX(pgm,BthNestedPAEPAE##name)
2820#define PGM_BTH_NAME_NESTED_PAE_AMD64(name) PGM_CTX(pgm,BthNestedPAEAMD64##name)
2821#define PGM_BTH_NAME_NESTED_AMD64_REAL(name) PGM_CTX(pgm,BthNestedAMD64Real##name)
2822#define PGM_BTH_NAME_NESTED_AMD64_PROT(name) PGM_CTX(pgm,BthNestedAMD64Prot##name)
2823#define PGM_BTH_NAME_NESTED_AMD64_32BIT(name) PGM_CTX(pgm,BthNestedAMD6432Bit##name)
2824#define PGM_BTH_NAME_NESTED_AMD64_PAE(name) PGM_CTX(pgm,BthNestedAMD64PAE##name)
2825#define PGM_BTH_NAME_NESTED_AMD64_AMD64(name) PGM_CTX(pgm,BthNestedAMD64AMD64##name)
2826#define PGM_BTH_NAME_EPT_REAL(name) PGM_CTX(pgm,BthEPTReal##name)
2827#define PGM_BTH_NAME_EPT_PROT(name) PGM_CTX(pgm,BthEPTProt##name)
2828#define PGM_BTH_NAME_EPT_32BIT(name) PGM_CTX(pgm,BthEPT32Bit##name)
2829#define PGM_BTH_NAME_EPT_PAE(name) PGM_CTX(pgm,BthEPTPAE##name)
2830#define PGM_BTH_NAME_EPT_AMD64(name) PGM_CTX(pgm,BthEPTAMD64##name)
2831#define PGM_BTH_NAME_NONE_REAL(name) PGM_CTX(pgm,BthNoneReal##name)
2832#define PGM_BTH_NAME_NONE_PROT(name) PGM_CTX(pgm,BthNoneProt##name)
2833#define PGM_BTH_NAME_NONE_32BIT(name) PGM_CTX(pgm,BthNone32Bit##name)
2834#define PGM_BTH_NAME_NONE_PAE(name) PGM_CTX(pgm,BthNonePAE##name)
2835#define PGM_BTH_NAME_NONE_AMD64(name) PGM_CTX(pgm,BthNoneAMD64##name)
2836
2837#define PGM_BTH_NAME_RC_32BIT_REAL_STR(name) "pgmRCBth32BitReal" #name
2838#define PGM_BTH_NAME_RC_32BIT_PROT_STR(name) "pgmRCBth32BitProt" #name
2839#define PGM_BTH_NAME_RC_32BIT_32BIT_STR(name) "pgmRCBth32Bit32Bit" #name
2840#define PGM_BTH_NAME_RC_PAE_REAL_STR(name) "pgmRCBthPAEReal" #name
2841#define PGM_BTH_NAME_RC_PAE_PROT_STR(name) "pgmRCBthPAEProt" #name
2842#define PGM_BTH_NAME_RC_PAE_32BIT_STR(name) "pgmRCBthPAE32Bit" #name
2843#define PGM_BTH_NAME_RC_PAE_PAE_STR(name) "pgmRCBthPAEPAE" #name
2844#define PGM_BTH_NAME_RC_AMD64_AMD64_STR(name) "pgmRCBthAMD64AMD64" #name
2845#define PGM_BTH_NAME_RC_NESTED_32BIT_REAL_STR(name) "pgmRCBthNested32BitReal" #name
2846#define PGM_BTH_NAME_RC_NESTED_32BIT_PROT_STR(name) "pgmRCBthNested32BitProt" #name
2847#define PGM_BTH_NAME_RC_NESTED_32BIT_32BIT_STR(name) "pgmRCBthNested32Bit32Bit" #name
2848#define PGM_BTH_NAME_RC_NESTED_32BIT_PAE_STR(name) "pgmRCBthNested32BitPAE" #name
2849#define PGM_BTH_NAME_RC_NESTED_32BIT_AMD64_STR(name) "pgmRCBthNested32BitAMD64" #name
2850#define PGM_BTH_NAME_RC_NESTED_PAE_REAL_STR(name) "pgmRCBthNestedPAEReal" #name
2851#define PGM_BTH_NAME_RC_NESTED_PAE_PROT_STR(name) "pgmRCBthNestedPAEProt" #name
2852#define PGM_BTH_NAME_RC_NESTED_PAE_32BIT_STR(name) "pgmRCBthNestedPAE32Bit" #name
2853#define PGM_BTH_NAME_RC_NESTED_PAE_PAE_STR(name) "pgmRCBthNestedPAEPAE" #name
2854#define PGM_BTH_NAME_RC_NESTED_PAE_AMD64_STR(name) "pgmRCBthNestedPAEAMD64" #name
2855#define PGM_BTH_NAME_RC_NESTED_AMD64_REAL_STR(name) "pgmRCBthNestedAMD64Real" #name
2856#define PGM_BTH_NAME_RC_NESTED_AMD64_PROT_STR(name) "pgmRCBthNestedAMD64Prot" #name
2857#define PGM_BTH_NAME_RC_NESTED_AMD64_32BIT_STR(name) "pgmRCBthNestedAMD6432Bit" #name
2858#define PGM_BTH_NAME_RC_NESTED_AMD64_PAE_STR(name) "pgmRCBthNestedAMD64PAE" #name
2859#define PGM_BTH_NAME_RC_NESTED_AMD64_AMD64_STR(name) "pgmRCBthNestedAMD64AMD64" #name
2860#define PGM_BTH_NAME_RC_EPT_REAL_STR(name) "pgmRCBthEPTReal" #name
2861#define PGM_BTH_NAME_RC_EPT_PROT_STR(name) "pgmRCBthEPTProt" #name
2862#define PGM_BTH_NAME_RC_EPT_32BIT_STR(name) "pgmRCBthEPT32Bit" #name
2863#define PGM_BTH_NAME_RC_EPT_PAE_STR(name) "pgmRCBthEPTPAE" #name
2864#define PGM_BTH_NAME_RC_EPT_AMD64_STR(name) "pgmRCBthEPTAMD64" #name
2865
2866#define PGM_BTH_NAME_R0_32BIT_REAL_STR(name) "pgmR0Bth32BitReal" #name
2867#define PGM_BTH_NAME_R0_32BIT_PROT_STR(name) "pgmR0Bth32BitProt" #name
2868#define PGM_BTH_NAME_R0_32BIT_32BIT_STR(name) "pgmR0Bth32Bit32Bit" #name
2869#define PGM_BTH_NAME_R0_PAE_REAL_STR(name) "pgmR0BthPAEReal" #name
2870#define PGM_BTH_NAME_R0_PAE_PROT_STR(name) "pgmR0BthPAEProt" #name
2871#define PGM_BTH_NAME_R0_PAE_32BIT_STR(name) "pgmR0BthPAE32Bit" #name
2872#define PGM_BTH_NAME_R0_PAE_PAE_STR(name) "pgmR0BthPAEPAE" #name
2873#define PGM_BTH_NAME_R0_AMD64_PROT_STR(name) "pgmR0BthAMD64Prot" #name
2874#define PGM_BTH_NAME_R0_AMD64_AMD64_STR(name) "pgmR0BthAMD64AMD64" #name
2875#define PGM_BTH_NAME_R0_NESTED_32BIT_REAL_STR(name) "pgmR0BthNested32BitReal" #name
2876#define PGM_BTH_NAME_R0_NESTED_32BIT_PROT_STR(name) "pgmR0BthNested32BitProt" #name
2877#define PGM_BTH_NAME_R0_NESTED_32BIT_32BIT_STR(name) "pgmR0BthNested32Bit32Bit" #name
2878#define PGM_BTH_NAME_R0_NESTED_32BIT_PAE_STR(name) "pgmR0BthNested32BitPAE" #name
2879#define PGM_BTH_NAME_R0_NESTED_32BIT_AMD64_STR(name) "pgmR0BthNested32BitAMD64" #name
2880#define PGM_BTH_NAME_R0_NESTED_PAE_REAL_STR(name) "pgmR0BthNestedPAEReal" #name
2881#define PGM_BTH_NAME_R0_NESTED_PAE_PROT_STR(name) "pgmR0BthNestedPAEProt" #name
2882#define PGM_BTH_NAME_R0_NESTED_PAE_32BIT_STR(name) "pgmR0BthNestedPAE32Bit" #name
2883#define PGM_BTH_NAME_R0_NESTED_PAE_PAE_STR(name) "pgmR0BthNestedPAEPAE" #name
2884#define PGM_BTH_NAME_R0_NESTED_PAE_AMD64_STR(name) "pgmR0BthNestedPAEAMD64" #name
2885#define PGM_BTH_NAME_R0_NESTED_AMD64_REAL_STR(name) "pgmR0BthNestedAMD64Real" #name
2886#define PGM_BTH_NAME_R0_NESTED_AMD64_PROT_STR(name) "pgmR0BthNestedAMD64Prot" #name
2887#define PGM_BTH_NAME_R0_NESTED_AMD64_32BIT_STR(name) "pgmR0BthNestedAMD6432Bit" #name
2888#define PGM_BTH_NAME_R0_NESTED_AMD64_PAE_STR(name) "pgmR0BthNestedAMD64PAE" #name
2889#define PGM_BTH_NAME_R0_NESTED_AMD64_AMD64_STR(name) "pgmR0BthNestedAMD64AMD64" #name
2890#define PGM_BTH_NAME_R0_EPT_REAL_STR(name) "pgmR0BthEPTReal" #name
2891#define PGM_BTH_NAME_R0_EPT_PROT_STR(name) "pgmR0BthEPTProt" #name
2892#define PGM_BTH_NAME_R0_EPT_32BIT_STR(name) "pgmR0BthEPT32Bit" #name
2893#define PGM_BTH_NAME_R0_EPT_PAE_STR(name) "pgmR0BthEPTPAE" #name
2894#define PGM_BTH_NAME_R0_EPT_AMD64_STR(name) "pgmR0BthEPTAMD64" #name
2895
2896#define PGM_BTH_DECL(type, name) PGM_CTX_DECL(type) PGM_BTH_NAME(name)
2897/** @} */
2898
2899
2900/**
2901 * Function pointers for guest paging.
2902 */
2903typedef struct PGMMODEDATAGST
2904{
2905 /** The guest mode type. */
2906 uint32_t uType;
2907 DECLCALLBACKMEMBER(int, pfnGetPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, uint64_t *pfFlags, PRTGCPHYS pGCPhys));
2908 DECLCALLBACKMEMBER(int, pfnModifyPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, size_t cbPages, uint64_t fFlags, uint64_t fMask));
2909 DECLCALLBACKMEMBER(int, pfnGetPDE,(PVMCPUCC pVCpu, RTGCPTR GCPtr, PX86PDEPAE pPde));
2910 DECLCALLBACKMEMBER(int, pfnEnter,(PVMCPUCC pVCpu, RTGCPHYS GCPhysCR3));
2911 DECLCALLBACKMEMBER(int, pfnExit,(PVMCPUCC pVCpu));
2912#ifdef IN_RING3
2913 DECLCALLBACKMEMBER(int, pfnRelocate,(PVMCPUCC pVCpu, RTGCPTR offDelta)); /**< Only in ring-3. */
2914#endif
2915} PGMMODEDATAGST;
2916
2917/** The length of g_aPgmGuestModeData. */
2918#ifdef VBOX_WITH_64_BITS_GUESTS
2919# define PGM_GUEST_MODE_DATA_ARRAY_SIZE (PGM_TYPE_AMD64 + 1)
2920#else
2921# define PGM_GUEST_MODE_DATA_ARRAY_SIZE (PGM_TYPE_PAE + 1)
2922#endif
2923/** The guest mode data array. */
2924extern PGMMODEDATAGST const g_aPgmGuestModeData[PGM_GUEST_MODE_DATA_ARRAY_SIZE];
2925
2926
2927/**
2928 * Function pointers for shadow paging.
2929 */
2930typedef struct PGMMODEDATASHW
2931{
2932 /** The shadow mode type. */
2933 uint32_t uType;
2934 DECLCALLBACKMEMBER(int, pfnGetPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, uint64_t *pfFlags, PRTHCPHYS pHCPhys));
2935 DECLCALLBACKMEMBER(int, pfnModifyPage,(PVMCPUCC pVCpu, RTGCPTR GCPtr, size_t cbPages, uint64_t fFlags,
2936 uint64_t fMask, uint32_t fOpFlags));
2937 DECLCALLBACKMEMBER(int, pfnEnter,(PVMCPUCC pVCpu, bool fIs64BitsPagingMode));
2938 DECLCALLBACKMEMBER(int, pfnExit,(PVMCPUCC pVCpu));
2939#ifdef IN_RING3
2940 DECLCALLBACKMEMBER(int, pfnRelocate,(PVMCPUCC pVCpu, RTGCPTR offDelta)); /**< Only in ring-3. */
2941#endif
2942} PGMMODEDATASHW;
2943
2944/** The length of g_aPgmShadowModeData. */
2945#define PGM_SHADOW_MODE_DATA_ARRAY_SIZE PGM_TYPE_END
2946/** The shadow mode data array. */
2947extern PGMMODEDATASHW const g_aPgmShadowModeData[PGM_SHADOW_MODE_DATA_ARRAY_SIZE];
2948
2949
2950/**
2951 * Function pointers for guest+shadow paging.
2952 */
2953typedef struct PGMMODEDATABTH
2954{
2955 /** The shadow mode type. */
2956 uint32_t uShwType;
2957 /** The guest mode type. */
2958 uint32_t uGstType;
2959
2960 DECLCALLBACKMEMBER(int, pfnInvalidatePage,(PVMCPUCC pVCpu, RTGCPTR GCPtrPage));
2961 DECLCALLBACKMEMBER(int, pfnSyncCR3,(PVMCPUCC pVCpu, uint64_t cr0, uint64_t cr3, uint64_t cr4, bool fGlobal));
2962 DECLCALLBACKMEMBER(int, pfnPrefetchPage,(PVMCPUCC pVCpu, RTGCPTR GCPtrPage));
2963 DECLCALLBACKMEMBER(int, pfnVerifyAccessSyncPage,(PVMCPUCC pVCpu, RTGCPTR GCPtrPage, unsigned fFlags, unsigned uError));
2964 DECLCALLBACKMEMBER(int, pfnMapCR3,(PVMCPUCC pVCpu, RTGCPHYS GCPhysCR3));
2965 DECLCALLBACKMEMBER(int, pfnUnmapCR3,(PVMCPUCC pVCpu));
2966 DECLCALLBACKMEMBER(int, pfnEnter,(PVMCPUCC pVCpu, RTGCPHYS GCPhysCR3));
2967#ifndef IN_RING3
2968 DECLCALLBACKMEMBER(int, pfnTrap0eHandler,(PVMCPUCC pVCpu, RTGCUINT uErr, PCPUMCTXCORE pRegFrame, RTGCPTR pvFault, bool *pfLockTaken));
2969#endif
2970#ifdef VBOX_STRICT
2971 DECLCALLBACKMEMBER(unsigned, pfnAssertCR3,(PVMCPUCC pVCpu, uint64_t cr3, uint64_t cr4, RTGCPTR GCPtr, RTGCPTR cb));
2972#endif
2973} PGMMODEDATABTH;
2974
2975/** The length of g_aPgmBothModeData. */
2976#define PGM_BOTH_MODE_DATA_ARRAY_SIZE ((PGM_TYPE_END - PGM_TYPE_FIRST_SHADOW) * PGM_TYPE_END)
2977/** The guest+shadow mode data array. */
2978extern PGMMODEDATABTH const g_aPgmBothModeData[PGM_BOTH_MODE_DATA_ARRAY_SIZE];
2979
2980
2981#ifdef VBOX_WITH_STATISTICS
2982/**
2983 * PGM statistics.
2984 *
2985 * These lives on the heap when compiled in as they would otherwise waste
2986 * unnecessary space in release builds.
2987 */
2988typedef struct PGMSTATS
2989{
2990 /* R3 only: */
2991 STAMCOUNTER StatR3DetectedConflicts; /**< R3: Number of times PGMR3MapHasConflicts() detected a conflict. */
2992 STAMPROFILE StatR3ResolveConflict; /**< R3: pgmR3SyncPTResolveConflict() profiling (includes the entire relocation). */
2993
2994 /* R3+RZ */
2995 STAMCOUNTER StatRZChunkR3MapTlbHits; /**< RC/R0: Ring-3/0 chunk mapper TLB hits. */
2996 STAMCOUNTER StatRZChunkR3MapTlbMisses; /**< RC/R0: Ring-3/0 chunk mapper TLB misses. */
2997 STAMCOUNTER StatRZPageMapTlbHits; /**< RC/R0: Ring-3/0 page mapper TLB hits. */
2998 STAMCOUNTER StatRZPageMapTlbMisses; /**< RC/R0: Ring-3/0 page mapper TLB misses. */
2999 STAMCOUNTER StatPageMapTlbFlushes; /**< ALL: Ring-3/0 page mapper TLB flushes. */
3000 STAMCOUNTER StatPageMapTlbFlushEntry; /**< ALL: Ring-3/0 page mapper TLB flushes. */
3001 STAMCOUNTER StatR3ChunkR3MapTlbHits; /**< R3: Ring-3/0 chunk mapper TLB hits. */
3002 STAMCOUNTER StatR3ChunkR3MapTlbMisses; /**< R3: Ring-3/0 chunk mapper TLB misses. */
3003 STAMCOUNTER StatR3PageMapTlbHits; /**< R3: Ring-3/0 page mapper TLB hits. */
3004 STAMCOUNTER StatR3PageMapTlbMisses; /**< R3: Ring-3/0 page mapper TLB misses. */
3005 STAMCOUNTER StatRZRamRangeTlbHits; /**< RC/R0: RAM range TLB hits. */
3006 STAMCOUNTER StatRZRamRangeTlbMisses; /**< RC/R0: RAM range TLB misses. */
3007 STAMCOUNTER StatR3RamRangeTlbHits; /**< R3: RAM range TLB hits. */
3008 STAMCOUNTER StatR3RamRangeTlbMisses; /**< R3: RAM range TLB misses. */
3009 STAMCOUNTER StatR3PhysHandlerReset; /**< R3: The number of times PGMHandlerPhysicalReset is called. */
3010 STAMCOUNTER StatRZPhysHandlerReset; /**< RC/R0: The number of times PGMHandlerPhysicalReset is called. */
3011 STAMCOUNTER StatR3PhysHandlerLookupHits; /**< R3: Number of cache hits when looking up physical handlers. */
3012 STAMCOUNTER StatR3PhysHandlerLookupMisses; /**< R3: Number of cache misses when looking up physical handlers. */
3013 STAMCOUNTER StatRZPhysHandlerLookupHits; /**< RC/R0: Number of cache hits when lookup up physical handlers. */
3014 STAMCOUNTER StatRZPhysHandlerLookupMisses; /**< RC/R0: Number of cache misses when looking up physical handlers */
3015 STAMCOUNTER StatRZPageReplaceShared; /**< RC/R0: Times a shared page has been replaced by a private one. */
3016 STAMCOUNTER StatRZPageReplaceZero; /**< RC/R0: Times the zero page has been replaced by a private one. */
3017/// @todo STAMCOUNTER StatRZPageHandyAllocs; /**< RC/R0: The number of times we've executed GMMR3AllocateHandyPages. */
3018 STAMCOUNTER StatR3PageReplaceShared; /**< R3: Times a shared page has been replaced by a private one. */
3019 STAMCOUNTER StatR3PageReplaceZero; /**< R3: Times the zero page has been replaced by a private one. */
3020/// @todo STAMCOUNTER StatR3PageHandyAllocs; /**< R3: The number of times we've executed GMMR3AllocateHandyPages. */
3021
3022 /* RC only: */
3023 STAMCOUNTER StatRCInvlPgConflict; /**< RC: Number of times PGMInvalidatePage() detected a mapping conflict. */
3024 STAMCOUNTER StatRCInvlPgSyncMonCR3; /**< RC: Number of times PGMInvalidatePage() ran into PGM_SYNC_MONITOR_CR3. */
3025
3026 STAMCOUNTER StatRZPhysRead;
3027 STAMCOUNTER StatRZPhysReadBytes;
3028 STAMCOUNTER StatRZPhysWrite;
3029 STAMCOUNTER StatRZPhysWriteBytes;
3030 STAMCOUNTER StatR3PhysRead;
3031 STAMCOUNTER StatR3PhysReadBytes;
3032 STAMCOUNTER StatR3PhysWrite;
3033 STAMCOUNTER StatR3PhysWriteBytes;
3034 STAMCOUNTER StatRCPhysRead;
3035 STAMCOUNTER StatRCPhysReadBytes;
3036 STAMCOUNTER StatRCPhysWrite;
3037 STAMCOUNTER StatRCPhysWriteBytes;
3038
3039 STAMCOUNTER StatRZPhysSimpleRead;
3040 STAMCOUNTER StatRZPhysSimpleReadBytes;
3041 STAMCOUNTER StatRZPhysSimpleWrite;
3042 STAMCOUNTER StatRZPhysSimpleWriteBytes;
3043 STAMCOUNTER StatR3PhysSimpleRead;
3044 STAMCOUNTER StatR3PhysSimpleReadBytes;
3045 STAMCOUNTER StatR3PhysSimpleWrite;
3046 STAMCOUNTER StatR3PhysSimpleWriteBytes;
3047 STAMCOUNTER StatRCPhysSimpleRead;
3048 STAMCOUNTER StatRCPhysSimpleReadBytes;
3049 STAMCOUNTER StatRCPhysSimpleWrite;
3050 STAMCOUNTER StatRCPhysSimpleWriteBytes;
3051
3052 STAMCOUNTER StatTrackVirgin; /**< The number of first time shadowings. */
3053 STAMCOUNTER StatTrackAliased; /**< The number of times switching to cRef2, i.e. the page is being shadowed by two PTs. */
3054 STAMCOUNTER StatTrackAliasedMany; /**< The number of times we're tracking using cRef2. */
3055 STAMCOUNTER StatTrackAliasedLots; /**< The number of times we're hitting pages which has overflowed cRef2. */
3056 STAMCOUNTER StatTrackNoExtentsLeft; /**< The number of times the extent list was exhausted. */
3057 STAMCOUNTER StatTrackOverflows; /**< The number of times the extent list grows to long. */
3058 STAMPROFILE StatTrackDeref; /**< Profiling of SyncPageWorkerTrackDeref (expensive). */
3059
3060 /** Time spent by the host OS for large page allocation. */
3061 STAMPROFILE StatAllocLargePage;
3062 /** Time spent clearing the newly allocated large pages. */
3063 STAMPROFILE StatClearLargePage;
3064 /** The number of times allocating a large pages takes more than the allowed period. */
3065 STAMCOUNTER StatLargePageOverflow;
3066 /** pgmPhysIsValidLargePage profiling - R3 */
3067 STAMPROFILE StatR3IsValidLargePage;
3068 /** pgmPhysIsValidLargePage profiling - RZ*/
3069 STAMPROFILE StatRZIsValidLargePage;
3070
3071 STAMPROFILE StatChunkAging;
3072 STAMPROFILE StatChunkFindCandidate;
3073 STAMPROFILE StatChunkUnmap;
3074 STAMPROFILE StatChunkMap;
3075} PGMSTATS;
3076#endif /* VBOX_WITH_STATISTICS */
3077
3078
3079/**
3080 * Converts a PGM pointer into a VM pointer.
3081 * @returns Pointer to the VM structure the PGM is part of.
3082 * @param pPGM Pointer to PGM instance data.
3083 */
3084#define PGM2VM(pPGM) ( (PVM)((char*)pPGM - pPGM->offVM) )
3085
3086/**
3087 * PGM Data (part of VM)
3088 */
3089typedef struct PGM
3090{
3091 /** Offset to the VM structure. */
3092 int32_t offVM;
3093 /** Offset of the PGMCPU structure relative to VMCPU. */
3094 int32_t offVCpuPGM;
3095
3096 /** @cfgm{/RamPreAlloc, boolean, false}
3097 * Indicates whether the base RAM should all be allocated before starting
3098 * the VM (default), or if it should be allocated when first written to.
3099 */
3100 bool fRamPreAlloc;
3101 /** Indicates whether write monitoring is currently in use.
3102 * This is used to prevent conflicts between live saving and page sharing
3103 * detection. */
3104 bool fPhysWriteMonitoringEngaged;
3105 /** Set if the CPU has less than 52-bit physical address width.
3106 * This is used */
3107 bool fLessThan52PhysicalAddressBits;
3108 /** Set when nested paging is active.
3109 * This is meant to save calls to HMIsNestedPagingActive and let the
3110 * compilers optimize the code better. Whether we use nested paging or
3111 * not is something we find out during VMM initialization and we won't
3112 * change this later on. */
3113 bool fNestedPaging;
3114 /** The host paging mode. (This is what SUPLib reports.) */
3115 SUPPAGINGMODE enmHostMode;
3116 /** We're not in a state which permits writes to guest memory.
3117 * (Only used in strict builds.) */
3118 bool fNoMorePhysWrites;
3119 /** @cfgm{/PageFusionAllowed, boolean, false}
3120 * Whether page fusion is allowed. */
3121 bool fPageFusionAllowed;
3122 /** @cfgm{/PGM/PciPassThrough, boolean, false}
3123 * Whether PCI passthrough is enabled. */
3124 bool fPciPassthrough;
3125 /** The number of MMIO2 regions (serves as the next MMIO2 ID). */
3126 uint8_t cMmio2Regions;
3127 /** Restore original ROM page content when resetting after loading state.
3128 * The flag is set by pgmR3LoadRomRanges and cleared at reset. This
3129 * enables the VM to start using an updated ROM without requiring powering
3130 * down the VM, just rebooting or resetting it. */
3131 bool fRestoreRomPagesOnReset;
3132 /** Whether to automatically clear all RAM pages on reset. */
3133 bool fZeroRamPagesOnReset;
3134 /** Alignment padding. */
3135 bool afAlignment3[7];
3136
3137 /** Indicates that PGMR3FinalizeMappings has been called and that further
3138 * PGMR3MapIntermediate calls will be rejected. */
3139 bool fFinalizedMappings;
3140 /** If set no conflict checks are required. */
3141 bool fMappingsFixed;
3142 /** If set if restored as fixed but we were unable to re-fixate at the old
3143 * location because of room or address incompatibilities. */
3144 bool fMappingsFixedRestored;
3145 /** Size of fixed mapping.
3146 * This is valid if either fMappingsFixed or fMappingsFixedRestored is set. */
3147 uint32_t cbMappingFixed;
3148 /** Generation ID for the RAM ranges. This member is incremented everytime
3149 * a RAM range is linked or unlinked. */
3150 uint32_t volatile idRamRangesGen;
3151
3152 /** Base address (GC) of fixed mapping.
3153 * This is valid if either fMappingsFixed or fMappingsFixedRestored is set. */
3154 RTGCPTR GCPtrMappingFixed;
3155 /** The address of the previous RAM range mapping. */
3156 RTGCPTR GCPtrPrevRamRangeMapping;
3157
3158 /** Physical access handler type for ROM protection. */
3159 PGMPHYSHANDLERTYPE hRomPhysHandlerType;
3160 /** Alignment padding. */
3161 uint32_t u32Padding;
3162
3163 /** 4 MB page mask; 32 or 36 bits depending on PSE-36 (identical for all VCPUs) */
3164 RTGCPHYS GCPhys4MBPSEMask;
3165 /** Mask containing the invalid bits of a guest physical address.
3166 * @remarks this does not stop at bit 52. */
3167 RTGCPHYS GCPhysInvAddrMask;
3168
3169
3170 /** RAM range TLB for R3. */
3171 R3PTRTYPE(PPGMRAMRANGE) apRamRangesTlbR3[PGM_RAMRANGE_TLB_ENTRIES];
3172 /** Pointer to the list of RAM ranges (Phys GC -> Phys HC conversion) - for R3.
3173 * This is sorted by physical address and contains no overlapping ranges. */
3174 R3PTRTYPE(PPGMRAMRANGE) pRamRangesXR3;
3175 /** Root of the RAM range search tree for ring-3. */
3176 R3PTRTYPE(PPGMRAMRANGE) pRamRangeTreeR3;
3177 /** PGM offset based trees - R3 Ptr. */
3178 R3PTRTYPE(PPGMTREES) pTreesR3;
3179 /** Caching the last physical handler we looked up in R3. */
3180 R3PTRTYPE(PPGMPHYSHANDLER) pLastPhysHandlerR3;
3181 /** Shadow Page Pool - R3 Ptr. */
3182 R3PTRTYPE(PPGMPOOL) pPoolR3;
3183#ifndef PGM_WITHOUT_MAPPINGS
3184 /** Linked list of GC mappings - for HC.
3185 * The list is sorted ascending on address. */
3186 R3PTRTYPE(PPGMMAPPING) pMappingsR3;
3187#endif
3188 /** Pointer to the list of ROM ranges - for R3.
3189 * This is sorted by physical address and contains no overlapping ranges. */
3190 R3PTRTYPE(PPGMROMRANGE) pRomRangesR3;
3191 /** Pointer to the list of MMIO2 ranges - for R3.
3192 * Registration order. */
3193 R3PTRTYPE(PPGMREGMMIO2RANGE) pRegMmioRangesR3;
3194 /** MMIO2 lookup array for ring-3. Indexed by idMmio2 minus 1. */
3195 R3PTRTYPE(PPGMREGMMIO2RANGE) apMmio2RangesR3[PGM_MMIO2_MAX_RANGES];
3196
3197 /** RAM range TLB for R0. */
3198 R0PTRTYPE(PPGMRAMRANGE) apRamRangesTlbR0[PGM_RAMRANGE_TLB_ENTRIES];
3199 /** R0 pointer corresponding to PGM::pRamRangesXR3. */
3200 R0PTRTYPE(PPGMRAMRANGE) pRamRangesXR0;
3201 /** Root of the RAM range search tree for ring-0. */
3202 R0PTRTYPE(PPGMRAMRANGE) pRamRangeTreeR0;
3203 /** PGM offset based trees - R0 Ptr. */
3204 R0PTRTYPE(PPGMTREES) pTreesR0;
3205 /** Caching the last physical handler we looked up in R0. */
3206 R0PTRTYPE(PPGMPHYSHANDLER) pLastPhysHandlerR0;
3207 /** Shadow Page Pool - R0 Ptr. */
3208 R0PTRTYPE(PPGMPOOL) pPoolR0;
3209#ifndef PGM_WITHOUT_MAPPINGS
3210 /** Linked list of GC mappings - for R0.
3211 * The list is sorted ascending on address. */
3212 R0PTRTYPE(PPGMMAPPING) pMappingsR0;
3213 RTR0PTR R0PtrAlignment0;
3214#endif
3215 /** R0 pointer corresponding to PGM::pRomRangesR3. */
3216 R0PTRTYPE(PPGMROMRANGE) pRomRangesR0;
3217 /** MMIO2 lookup array for ring-0. Indexed by idMmio2 minus 1. */
3218 R0PTRTYPE(PPGMREGMMIO2RANGE) apMmio2RangesR0[PGM_MMIO2_MAX_RANGES];
3219
3220#ifndef PGM_WITHOUT_MAPPINGS
3221 /** Pointer to the 5 page CR3 content mapping.
3222 * The first page is always the CR3 (in some form) while the 4 other pages
3223 * are used for the PDs in PAE mode. */
3224 RTGCPTR GCPtrCR3Mapping;
3225
3226 /** @name Intermediate Context
3227 * @{ */
3228 /** Pointer to the intermediate page directory - Normal. */
3229 R3PTRTYPE(PX86PD) pInterPD;
3230 /** Pointer to the intermediate page tables - Normal.
3231 * There are two page tables, one for the identity mapping and one for
3232 * the host context mapping (of the core code). */
3233 R3PTRTYPE(PX86PT) apInterPTs[2];
3234 /** Pointer to the intermediate page tables - PAE. */
3235 R3PTRTYPE(PX86PTPAE) apInterPaePTs[2];
3236 /** Pointer to the intermediate page directory - PAE. */
3237 R3PTRTYPE(PX86PDPAE) apInterPaePDs[4];
3238 /** Pointer to the intermediate page directory - PAE. */
3239 R3PTRTYPE(PX86PDPT) pInterPaePDPT;
3240 /** Pointer to the intermediate page-map level 4 - AMD64. */
3241 R3PTRTYPE(PX86PML4) pInterPaePML4;
3242 /** Pointer to the intermediate page directory - AMD64. */
3243 R3PTRTYPE(PX86PDPT) pInterPaePDPT64;
3244 /** The Physical Address (HC) of the intermediate Page Directory - Normal. */
3245 RTHCPHYS HCPhysInterPD;
3246 /** The Physical Address (HC) of the intermediate Page Directory Pointer Table - PAE. */
3247 RTHCPHYS HCPhysInterPaePDPT;
3248 /** The Physical Address (HC) of the intermediate Page Map Level 4 table - AMD64. */
3249 RTHCPHYS HCPhysInterPaePML4;
3250 /** @} */
3251#endif
3252
3253#ifndef PGM_WITHOUT_MAPPINGS
3254 /** Base address of the dynamic page mapping area.
3255 * The array is MM_HYPER_DYNAMIC_SIZE bytes big.
3256 *
3257 * @todo The plan of keeping PGMRCDYNMAP private to PGMRZDynMap.cpp didn't
3258 * work out. Some cleaning up of the initialization that would
3259 * remove this memory is yet to be done...
3260 */
3261 RCPTRTYPE(uint8_t *) pbDynPageMapBaseGC;
3262 /** The address of the raw-mode context mapping cache. */
3263 RCPTRTYPE(PPGMRCDYNMAP) pRCDynMap;
3264 /** The address of the ring-0 mapping cache if we're making use of it. */
3265 RTR0PTR pvR0DynMapUsed;
3266#endif
3267
3268 /** Hack: Number of deprecated page mapping locks taken by the current lock
3269 * owner via pgmPhysGCPhys2CCPtrInternalDepr. */
3270 uint32_t cDeprecatedPageLocks;
3271 /** Alignment padding. */
3272 uint32_t au32Alignment2[1];
3273
3274
3275 /** PGM critical section.
3276 * This protects the physical, ram ranges, and the page flag updating (some of
3277 * it anyway).
3278 */
3279 PDMCRITSECT CritSectX;
3280
3281 /**
3282 * Data associated with managing the ring-3 mappings of the allocation chunks.
3283 */
3284 struct
3285 {
3286 /** The chunk mapping TLB. */
3287 PGMCHUNKR3MAPTLB Tlb;
3288 /** The chunk tree, ordered by chunk id. */
3289#if defined(VBOX_WITH_2X_4GB_ADDR_SPACE) || defined(VBOX_WITH_RAM_IN_KERNEL)
3290 R3PTRTYPE(PAVLU32NODECORE) pTree;
3291#else
3292 R3R0PTRTYPE(PAVLU32NODECORE) pTree;
3293#endif
3294#if HC_ARCH_BITS == 32
3295 uint32_t u32Alignment0;
3296#endif
3297 /** The number of mapped chunks. */
3298 uint32_t c;
3299 /** @cfgm{/PGM/MaxRing3Chunks, uint32_t, host dependent}
3300 * The maximum number of mapped chunks. On 64-bit this is unlimited by default,
3301 * on 32-bit it defaults to 1 or 3 GB depending on the host. */
3302 uint32_t cMax;
3303 /** The current time. This is incremented whenever a chunk is inserted. */
3304 uint32_t iNow;
3305 /** Alignment padding. */
3306 uint32_t au32Alignment1[3];
3307 } ChunkR3Map;
3308
3309 /** The page mapping TLB for ring-3. */
3310 PGMPAGER3MAPTLB PhysTlbR3;
3311#ifdef VBOX_WITH_RAM_IN_KERNEL
3312 /** The page mapping TLB for ring-0. */
3313 PGMPAGER0MAPTLB PhysTlbR0;
3314#else
3315 /** The page mapping TLB for ring-0 (still using ring-3 mappings). */
3316 PGMPAGER3MAPTLB PhysTlbR0;
3317#endif
3318
3319 /** @name The zero page.
3320 * @{ */
3321 /** The host physical address of the zero page. */
3322 RTHCPHYS HCPhysZeroPg;
3323 /** The ring-3 mapping of the zero page. */
3324 RTR3PTR pvZeroPgR3;
3325 /** The ring-0 mapping of the zero page. */
3326 RTR0PTR pvZeroPgR0;
3327 /** The GC mapping of the zero page. */
3328 RTRCPTR pvZeroPgRC;
3329 RTRCPTR RCPtrAlignment3;
3330 /** @}*/
3331
3332 /** @name The Invalid MMIO page.
3333 * This page is filled with 0xfeedface.
3334 * @{ */
3335 /** The host physical address of the invalid MMIO page. */
3336 RTHCPHYS HCPhysMmioPg;
3337 /** The host pysical address of the invalid MMIO page plus all invalid
3338 * physical address bits set. This is used to trigger X86_TRAP_PF_RSVD.
3339 * @remarks Check fLessThan52PhysicalAddressBits before use. */
3340 RTHCPHYS HCPhysInvMmioPg;
3341 /** The ring-3 mapping of the invalid MMIO page. */
3342 RTR3PTR pvMmioPgR3;
3343#if HC_ARCH_BITS == 32
3344 RTR3PTR R3PtrAlignment4;
3345#endif
3346 /** @} */
3347
3348
3349 /** The number of handy pages. */
3350 uint32_t cHandyPages;
3351
3352 /** The number of large handy pages. */
3353 uint32_t cLargeHandyPages;
3354
3355 /**
3356 * Array of handy pages.
3357 *
3358 * This array is used in a two way communication between pgmPhysAllocPage
3359 * and GMMR0AllocateHandyPages, with PGMR3PhysAllocateHandyPages serving as
3360 * an intermediary.
3361 *
3362 * The size of this array is important, see pgmPhysEnsureHandyPage for details.
3363 * (The current size of 32 pages, means 128 KB of handy memory.)
3364 */
3365 GMMPAGEDESC aHandyPages[PGM_HANDY_PAGES];
3366
3367 /**
3368 * Array of large handy pages. (currently size 1)
3369 *
3370 * This array is used in a two way communication between pgmPhysAllocLargePage
3371 * and GMMR0AllocateLargePage, with PGMR3PhysAllocateLargePage serving as
3372 * an intermediary.
3373 */
3374 GMMPAGEDESC aLargeHandyPage[1];
3375
3376 /**
3377 * Live save data.
3378 */
3379 struct
3380 {
3381 /** Per type statistics. */
3382 struct
3383 {
3384 /** The number of ready pages. */
3385 uint32_t cReadyPages;
3386 /** The number of dirty pages. */
3387 uint32_t cDirtyPages;
3388 /** The number of ready zero pages. */
3389 uint32_t cZeroPages;
3390 /** The number of write monitored pages. */
3391 uint32_t cMonitoredPages;
3392 } Rom,
3393 Mmio2,
3394 Ram;
3395 /** The number of ignored pages in the RAM ranges (i.e. MMIO, MMIO2 and ROM). */
3396 uint32_t cIgnoredPages;
3397 /** Indicates that a live save operation is active. */
3398 bool fActive;
3399 /** Padding. */
3400 bool afReserved[2];
3401 /** The next history index. */
3402 uint8_t iDirtyPagesHistory;
3403 /** History of the total amount of dirty pages. */
3404 uint32_t acDirtyPagesHistory[64];
3405 /** Short term dirty page average. */
3406 uint32_t cDirtyPagesShort;
3407 /** Long term dirty page average. */
3408 uint32_t cDirtyPagesLong;
3409 /** The number of saved pages. This is used to get some kind of estimate of the
3410 * link speed so we can decide when we're done. It is reset after the first
3411 * 7 passes so the speed estimate doesn't get inflated by the initial set of
3412 * zero pages. */
3413 uint64_t cSavedPages;
3414 /** The nanosecond timestamp when cSavedPages was 0. */
3415 uint64_t uSaveStartNS;
3416 /** Pages per second (for statistics). */
3417 uint32_t cPagesPerSecond;
3418 uint32_t cAlignment;
3419 } LiveSave;
3420
3421 /** @name Error injection.
3422 * @{ */
3423 /** Inject handy page allocation errors pretending we're completely out of
3424 * memory. */
3425 bool volatile fErrInjHandyPages;
3426 /** Padding. */
3427 bool afReserved[3];
3428 /** @} */
3429
3430 /** @name Release Statistics
3431 * @{ */
3432 uint32_t cAllPages; /**< The total number of pages. (Should be Private + Shared + Zero + Pure MMIO.) */
3433 uint32_t cPrivatePages; /**< The number of private pages. */
3434 uint32_t cSharedPages; /**< The number of shared pages. */
3435 uint32_t cReusedSharedPages; /**< The number of reused shared pages. */
3436 uint32_t cZeroPages; /**< The number of zero backed pages. */
3437 uint32_t cPureMmioPages; /**< The number of pure MMIO pages. */
3438 uint32_t cMonitoredPages; /**< The number of write monitored pages. */
3439 uint32_t cWrittenToPages; /**< The number of previously write monitored pages. */
3440 uint32_t cWriteLockedPages; /**< The number of write locked pages. */
3441 uint32_t cReadLockedPages; /**< The number of read locked pages. */
3442 uint32_t cBalloonedPages; /**< The number of ballooned pages. */
3443 uint32_t cMappedChunks; /**< Number of times we mapped a chunk. */
3444 uint32_t cUnmappedChunks; /**< Number of times we unmapped a chunk. */
3445 uint32_t cLargePages; /**< The number of large pages. */
3446 uint32_t cLargePagesDisabled; /**< The number of disabled large pages. */
3447/* uint32_t aAlignment4[1]; */
3448
3449 /** The number of times we were forced to change the hypervisor region location. */
3450 STAMCOUNTER cRelocations;
3451
3452 STAMCOUNTER StatLargePageReused; /**< The number of large pages we've reused.*/
3453 STAMCOUNTER StatLargePageRefused; /**< The number of times we couldn't use a large page.*/
3454 STAMCOUNTER StatLargePageRecheck; /**< The number of times we rechecked a disabled large page.*/
3455
3456 STAMPROFILE StatShModCheck; /**< Profiles shared module checks. */
3457 /** @} */
3458
3459#ifdef VBOX_WITH_STATISTICS
3460 /** @name Statistics on the heap.
3461 * @{ */
3462 R3PTRTYPE(PGMSTATS *) pStatsR3;
3463 R0PTRTYPE(PGMSTATS *) pStatsR0;
3464 /** @} */
3465#endif
3466} PGM;
3467#ifndef IN_TSTVMSTRUCTGC /* HACK */
3468# ifndef PGM_WITHOUT_MAPPINGS
3469AssertCompileMemberAlignment(PGM, paDynPageMap32BitPTEsGC, 8);
3470# endif
3471AssertCompileMemberAlignment(PGM, GCPtrMappingFixed, sizeof(RTGCPTR));
3472# ifndef PGM_WITHOUT_MAPPINGS
3473AssertCompileMemberAlignment(PGM, HCPhysInterPD, 8);
3474# endif
3475AssertCompileMemberAlignment(PGM, CritSectX, 8);
3476AssertCompileMemberAlignment(PGM, ChunkR3Map, 16);
3477AssertCompileMemberAlignment(PGM, PhysTlbR3, 32); /** @todo 32 byte alignment! */
3478AssertCompileMemberAlignment(PGM, PhysTlbR0, 32);
3479AssertCompileMemberAlignment(PGM, HCPhysZeroPg, 8);
3480AssertCompileMemberAlignment(PGM, aHandyPages, 8);
3481AssertCompileMemberAlignment(PGM, cRelocations, 8);
3482#endif /* !IN_TSTVMSTRUCTGC */
3483/** Pointer to the PGM instance data. */
3484typedef PGM *PPGM;
3485
3486
3487
3488typedef struct PGMCPUSTATS
3489{
3490 /* Common */
3491 STAMCOUNTER StatSyncPtPD[X86_PG_ENTRIES]; /**< SyncPT - PD distribution. */
3492 STAMCOUNTER StatSyncPagePD[X86_PG_ENTRIES]; /**< SyncPage - PD distribution. */
3493
3494 /* R0 only: */
3495 STAMPROFILE StatR0NpMiscfg; /**< R0: PGMR0Trap0eHandlerNPMisconfig() profiling. */
3496 STAMCOUNTER StatR0NpMiscfgSyncPage; /**< R0: SyncPage calls from PGMR0Trap0eHandlerNPMisconfig(). */
3497
3498 /* RZ only: */
3499 STAMPROFILE StatRZTrap0e; /**< RC/R0: PGMTrap0eHandler() profiling. */
3500 STAMPROFILE StatRZTrap0eTime2Ballooned; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is read access to a ballooned page. */
3501 STAMPROFILE StatRZTrap0eTime2CSAM; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CSAM. */
3502 STAMPROFILE StatRZTrap0eTime2DirtyAndAccessed; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is dirty and/or accessed bit emulation. */
3503 STAMPROFILE StatRZTrap0eTime2GuestTrap; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is a guest trap. */
3504 STAMPROFILE StatRZTrap0eTime2HndPhys; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is a physical handler. */
3505 STAMPROFILE StatRZTrap0eTime2HndUnhandled; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is access outside the monitored areas of a monitored page. */
3506 STAMPROFILE StatRZTrap0eTime2InvalidPhys; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is access to an invalid physical guest address. */
3507 STAMPROFILE StatRZTrap0eTime2MakeWritable; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is a page that needed to be made writable. */
3508 STAMPROFILE StatRZTrap0eTime2Mapping; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is the guest mappings. */
3509 STAMPROFILE StatRZTrap0eTime2Misc; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is not known. */
3510 STAMPROFILE StatRZTrap0eTime2OutOfSync; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is an out-of-sync page. */
3511 STAMPROFILE StatRZTrap0eTime2OutOfSyncHndPhys; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is an out-of-sync physical handler page. */
3512 STAMPROFILE StatRZTrap0eTime2OutOfSyncHndObs; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is an obsolete handler page. */
3513 STAMPROFILE StatRZTrap0eTime2SyncPT; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is lazy syncing of a PT. */
3514 STAMPROFILE StatRZTrap0eTime2WPEmulation; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CR0.WP emulation. */
3515 STAMPROFILE StatRZTrap0eTime2Wp0RoUsHack; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CR0.WP and netware hack to be enabled. */
3516 STAMPROFILE StatRZTrap0eTime2Wp0RoUsUnhack; /**< RC/R0: Profiling of the Trap0eHandler body when the cause is CR0.WP and netware hack to be disabled. */
3517 STAMCOUNTER StatRZTrap0eConflicts; /**< RC/R0: The number of times \#PF was caused by an undetected conflict. */
3518 STAMCOUNTER StatRZTrap0eHandlersMapping; /**< RC/R0: Number of traps due to access handlers in mappings. */
3519 STAMCOUNTER StatRZTrap0eHandlersOutOfSync; /**< RC/R0: Number of out-of-sync handled pages. */
3520 STAMCOUNTER StatRZTrap0eHandlersPhysAll; /**< RC/R0: Number of traps due to physical all-access handlers. */
3521 STAMCOUNTER StatRZTrap0eHandlersPhysAllOpt; /**< RC/R0: Number of the physical all-access handler traps using the optimization. */
3522 STAMCOUNTER StatRZTrap0eHandlersPhysWrite; /**< RC/R0: Number of traps due to write-physical access handlers. */
3523 STAMCOUNTER StatRZTrap0eHandlersUnhandled; /**< RC/R0: Number of traps due to access outside range of monitored page(s). */
3524 STAMCOUNTER StatRZTrap0eHandlersInvalid; /**< RC/R0: Number of traps due to access to invalid physical memory. */
3525 STAMCOUNTER StatRZTrap0eUSNotPresentRead; /**< RC/R0: \#PF err kind */
3526 STAMCOUNTER StatRZTrap0eUSNotPresentWrite; /**< RC/R0: \#PF err kind */
3527 STAMCOUNTER StatRZTrap0eUSWrite; /**< RC/R0: \#PF err kind */
3528 STAMCOUNTER StatRZTrap0eUSReserved; /**< RC/R0: \#PF err kind */
3529 STAMCOUNTER StatRZTrap0eUSNXE; /**< RC/R0: \#PF err kind */
3530 STAMCOUNTER StatRZTrap0eUSRead; /**< RC/R0: \#PF err kind */
3531 STAMCOUNTER StatRZTrap0eSVNotPresentRead; /**< RC/R0: \#PF err kind */
3532 STAMCOUNTER StatRZTrap0eSVNotPresentWrite; /**< RC/R0: \#PF err kind */
3533 STAMCOUNTER StatRZTrap0eSVWrite; /**< RC/R0: \#PF err kind */
3534 STAMCOUNTER StatRZTrap0eSVReserved; /**< RC/R0: \#PF err kind */
3535 STAMCOUNTER StatRZTrap0eSNXE; /**< RC/R0: \#PF err kind */
3536 STAMCOUNTER StatRZTrap0eGuestPF; /**< RC/R0: Real guest \#PFs. */
3537 STAMCOUNTER StatRZTrap0eGuestPFMapping; /**< RC/R0: Real guest \#PF to HMA or other mapping. */
3538 STAMCOUNTER StatRZTrap0eWPEmulInRZ; /**< RC/R0: WP=0 virtualization trap, handled. */
3539 STAMCOUNTER StatRZTrap0eWPEmulToR3; /**< RC/R0: WP=0 virtualization trap, chickened out. */
3540 STAMCOUNTER StatRZTrap0ePD[X86_PG_ENTRIES]; /**< RC/R0: PD distribution of the \#PFs. */
3541 STAMCOUNTER StatRZGuestCR3WriteHandled; /**< RC/R0: The number of times WriteHandlerCR3() was successfully called. */
3542 STAMCOUNTER StatRZGuestCR3WriteUnhandled; /**< RC/R0: The number of times WriteHandlerCR3() was called and we had to fall back to the recompiler. */
3543 STAMCOUNTER StatRZGuestCR3WriteConflict; /**< RC/R0: The number of times WriteHandlerCR3() was called and a conflict was detected. */
3544 STAMCOUNTER StatRZGuestROMWriteHandled; /**< RC/R0: The number of times pgmPhysRomWriteHandler() was successfully called. */
3545 STAMCOUNTER StatRZGuestROMWriteUnhandled; /**< RC/R0: The number of times pgmPhysRomWriteHandler() was called and we had to fall back to the recompiler */
3546 STAMCOUNTER StatRZDynMapMigrateInvlPg; /**< RZ: invlpg in PGMR0DynMapMigrateAutoSet. */
3547 STAMPROFILE StatRZDynMapGCPageInl; /**< RZ: Calls to pgmRZDynMapGCPageInlined. */
3548 STAMCOUNTER StatRZDynMapGCPageInlHits; /**< RZ: Hash table lookup hits. */
3549 STAMCOUNTER StatRZDynMapGCPageInlMisses; /**< RZ: Misses that falls back to the code common. */
3550 STAMCOUNTER StatRZDynMapGCPageInlRamHits; /**< RZ: 1st ram range hits. */
3551 STAMCOUNTER StatRZDynMapGCPageInlRamMisses; /**< RZ: 1st ram range misses, takes slow path. */
3552 STAMPROFILE StatRZDynMapHCPageInl; /**< RZ: Calls to pgmRZDynMapHCPageInlined. */
3553 STAMCOUNTER StatRZDynMapHCPageInlHits; /**< RZ: Hash table lookup hits. */
3554 STAMCOUNTER StatRZDynMapHCPageInlMisses; /**< RZ: Misses that falls back to the code common. */
3555 STAMPROFILE StatRZDynMapHCPage; /**< RZ: Calls to pgmRZDynMapHCPageCommon. */
3556 STAMCOUNTER StatRZDynMapSetOptimize; /**< RZ: Calls to pgmRZDynMapOptimizeAutoSet. */
3557 STAMCOUNTER StatRZDynMapSetSearchFlushes; /**< RZ: Set search restoring to subset flushes. */
3558 STAMCOUNTER StatRZDynMapSetSearchHits; /**< RZ: Set search hits. */
3559 STAMCOUNTER StatRZDynMapSetSearchMisses; /**< RZ: Set search misses. */
3560 STAMCOUNTER StatRZDynMapPage; /**< RZ: Calls to pgmR0DynMapPage. */
3561 STAMCOUNTER StatRZDynMapPageHits0; /**< RZ: Hits at iPage+0. */
3562 STAMCOUNTER StatRZDynMapPageHits1; /**< RZ: Hits at iPage+1. */
3563 STAMCOUNTER StatRZDynMapPageHits2; /**< RZ: Hits at iPage+2. */
3564 STAMCOUNTER StatRZDynMapPageInvlPg; /**< RZ: invlpg. */
3565 STAMCOUNTER StatRZDynMapPageSlow; /**< RZ: Calls to pgmR0DynMapPageSlow. */
3566 STAMCOUNTER StatRZDynMapPageSlowLoopHits; /**< RZ: Hits in the pgmR0DynMapPageSlow search loop. */
3567 STAMCOUNTER StatRZDynMapPageSlowLoopMisses; /**< RZ: Misses in the pgmR0DynMapPageSlow search loop. */
3568 //STAMCOUNTER StatRZDynMapPageSlowLostHits; /**< RZ: Lost hits. */
3569 STAMCOUNTER StatRZDynMapSubsets; /**< RZ: Times PGMDynMapPushAutoSubset was called. */
3570 STAMCOUNTER StatRZDynMapPopFlushes; /**< RZ: Times PGMDynMapPopAutoSubset flushes the subset. */
3571 STAMCOUNTER aStatRZDynMapSetFilledPct[11]; /**< RZ: Set fill distribution, percent. */
3572
3573 /* HC - R3 and (maybe) R0: */
3574
3575 /* RZ & R3: */
3576 STAMPROFILE StatRZSyncCR3; /**< RC/R0: PGMSyncCR3() profiling. */
3577 STAMPROFILE StatRZSyncCR3Handlers; /**< RC/R0: Profiling of the PGMSyncCR3() update handler section. */
3578 STAMCOUNTER StatRZSyncCR3Global; /**< RC/R0: The number of global CR3 syncs. */
3579 STAMCOUNTER StatRZSyncCR3NotGlobal; /**< RC/R0: The number of non-global CR3 syncs. */
3580 STAMCOUNTER StatRZSyncCR3DstCacheHit; /**< RC/R0: The number of times we got some kind of cache hit on a page table. */
3581 STAMCOUNTER StatRZSyncCR3DstFreed; /**< RC/R0: The number of times we've had to free a shadow entry. */
3582 STAMCOUNTER StatRZSyncCR3DstFreedSrcNP; /**< RC/R0: The number of times we've had to free a shadow entry for which the source entry was not present. */
3583 STAMCOUNTER StatRZSyncCR3DstNotPresent; /**< RC/R0: The number of times we've encountered a not present shadow entry for a present guest entry. */
3584 STAMCOUNTER StatRZSyncCR3DstSkippedGlobalPD; /**< RC/R0: The number of times a global page directory wasn't flushed. */
3585 STAMCOUNTER StatRZSyncCR3DstSkippedGlobalPT; /**< RC/R0: The number of times a page table with only global entries wasn't flushed. */
3586 STAMPROFILE StatRZSyncPT; /**< RC/R0: PGMSyncPT() profiling. */
3587 STAMCOUNTER StatRZSyncPTFailed; /**< RC/R0: The number of times PGMSyncPT() failed. */
3588 STAMCOUNTER StatRZSyncPT4K; /**< RC/R0: Number of 4KB syncs. */
3589 STAMCOUNTER StatRZSyncPT4M; /**< RC/R0: Number of 4MB syncs. */
3590 STAMCOUNTER StatRZSyncPagePDNAs; /**< RC/R0: The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit. */
3591 STAMCOUNTER StatRZSyncPagePDOutOfSync; /**< RC/R0: The number of time we've encountered an out-of-sync PD in SyncPage. */
3592 STAMCOUNTER StatRZAccessedPage; /**< RC/R0: The number of pages marked not present for accessed bit emulation. */
3593 STAMPROFILE StatRZDirtyBitTracking; /**< RC/R0: Profiling the dirty bit tracking in CheckPageFault(). */
3594 STAMCOUNTER StatRZDirtyPage; /**< RC/R0: The number of pages marked read-only for dirty bit tracking. */
3595 STAMCOUNTER StatRZDirtyPageBig; /**< RC/R0: The number of pages marked read-only for dirty bit tracking. */
3596 STAMCOUNTER StatRZDirtyPageSkipped; /**< RC/R0: The number of pages already dirty or readonly. */
3597 STAMCOUNTER StatRZDirtyPageTrap; /**< RC/R0: The number of traps generated for dirty bit tracking. */
3598 STAMCOUNTER StatRZDirtyPageStale; /**< RC/R0: The number of traps generated for dirty bit tracking. (stale tlb entries) */
3599 STAMCOUNTER StatRZDirtyTrackRealPF; /**< RC/R0: The number of real pages faults during dirty bit tracking. */
3600 STAMCOUNTER StatRZDirtiedPage; /**< RC/R0: The number of pages marked dirty because of write accesses. */
3601 STAMCOUNTER StatRZPageAlreadyDirty; /**< RC/R0: The number of pages already marked dirty because of write accesses. */
3602 STAMPROFILE StatRZInvalidatePage; /**< RC/R0: PGMInvalidatePage() profiling. */
3603 STAMCOUNTER StatRZInvalidatePage4KBPages; /**< RC/R0: The number of times PGMInvalidatePage() was called for a 4KB page. */
3604 STAMCOUNTER StatRZInvalidatePage4MBPages; /**< RC/R0: The number of times PGMInvalidatePage() was called for a 4MB page. */
3605 STAMCOUNTER StatRZInvalidatePage4MBPagesSkip; /**< RC/R0: The number of times PGMInvalidatePage() skipped a 4MB page. */
3606 STAMCOUNTER StatRZInvalidatePagePDMappings; /**< RC/R0: The number of times PGMInvalidatePage() was called for a page directory containing mappings (no conflict). */
3607 STAMCOUNTER StatRZInvalidatePagePDNAs; /**< RC/R0: The number of times PGMInvalidatePage() was called for a not accessed page directory. */
3608 STAMCOUNTER StatRZInvalidatePagePDNPs; /**< RC/R0: The number of times PGMInvalidatePage() was called for a not present page directory. */
3609 STAMCOUNTER StatRZInvalidatePagePDOutOfSync; /**< RC/R0: The number of times PGMInvalidatePage() was called for an out of sync page directory. */
3610 STAMCOUNTER StatRZInvalidatePageSizeChanges ; /**< RC/R0: The number of times PGMInvalidatePage() was called on a page size change (4KB <-> 2/4MB). */
3611 STAMCOUNTER StatRZInvalidatePageSkipped; /**< RC/R0: The number of times PGMInvalidatePage() was skipped due to not present shw or pending pending SyncCR3. */
3612 STAMCOUNTER StatRZPageOutOfSyncUser; /**< RC/R0: The number of times user page is out of sync was detected in \#PF or VerifyAccessSyncPage. */
3613 STAMCOUNTER StatRZPageOutOfSyncSupervisor; /**< RC/R0: The number of times supervisor page is out of sync was detected in in \#PF or VerifyAccessSyncPage. */
3614 STAMCOUNTER StatRZPageOutOfSyncUserWrite; /**< RC/R0: The number of times user page is out of sync was detected in \#PF. */
3615 STAMCOUNTER StatRZPageOutOfSyncSupervisorWrite; /**< RC/R0: The number of times supervisor page is out of sync was detected in in \#PF. */
3616 STAMCOUNTER StatRZPageOutOfSyncBallloon; /**< RC/R0: The number of times a ballooned page was accessed (read). */
3617 STAMPROFILE StatRZPrefetch; /**< RC/R0: PGMPrefetchPage. */
3618 STAMPROFILE StatRZFlushTLB; /**< RC/R0: Profiling of the PGMFlushTLB() body. */
3619 STAMCOUNTER StatRZFlushTLBNewCR3; /**< RC/R0: The number of times PGMFlushTLB was called with a new CR3, non-global. (switch) */
3620 STAMCOUNTER StatRZFlushTLBNewCR3Global; /**< RC/R0: The number of times PGMFlushTLB was called with a new CR3, global. (switch) */
3621 STAMCOUNTER StatRZFlushTLBSameCR3; /**< RC/R0: The number of times PGMFlushTLB was called with the same CR3, non-global. (flush) */
3622 STAMCOUNTER StatRZFlushTLBSameCR3Global; /**< RC/R0: The number of times PGMFlushTLB was called with the same CR3, global. (flush) */
3623 STAMPROFILE StatRZGstModifyPage; /**< RC/R0: Profiling of the PGMGstModifyPage() body */
3624
3625 STAMPROFILE StatR3SyncCR3; /**< R3: PGMSyncCR3() profiling. */
3626 STAMPROFILE StatR3SyncCR3Handlers; /**< R3: Profiling of the PGMSyncCR3() update handler section. */
3627 STAMCOUNTER StatR3SyncCR3Global; /**< R3: The number of global CR3 syncs. */
3628 STAMCOUNTER StatR3SyncCR3NotGlobal; /**< R3: The number of non-global CR3 syncs. */
3629 STAMCOUNTER StatR3SyncCR3DstFreed; /**< R3: The number of times we've had to free a shadow entry. */
3630 STAMCOUNTER StatR3SyncCR3DstFreedSrcNP; /**< R3: The number of times we've had to free a shadow entry for which the source entry was not present. */
3631 STAMCOUNTER StatR3SyncCR3DstNotPresent; /**< R3: The number of times we've encountered a not present shadow entry for a present guest entry. */
3632 STAMCOUNTER StatR3SyncCR3DstSkippedGlobalPD; /**< R3: The number of times a global page directory wasn't flushed. */
3633 STAMCOUNTER StatR3SyncCR3DstSkippedGlobalPT; /**< R3: The number of times a page table with only global entries wasn't flushed. */
3634 STAMCOUNTER StatR3SyncCR3DstCacheHit; /**< R3: The number of times we got some kind of cache hit on a page table. */
3635 STAMPROFILE StatR3SyncPT; /**< R3: PGMSyncPT() profiling. */
3636 STAMCOUNTER StatR3SyncPTFailed; /**< R3: The number of times PGMSyncPT() failed. */
3637 STAMCOUNTER StatR3SyncPT4K; /**< R3: Number of 4KB syncs. */
3638 STAMCOUNTER StatR3SyncPT4M; /**< R3: Number of 4MB syncs. */
3639 STAMCOUNTER StatR3SyncPagePDNAs; /**< R3: The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit. */
3640 STAMCOUNTER StatR3SyncPagePDOutOfSync; /**< R3: The number of time we've encountered an out-of-sync PD in SyncPage. */
3641 STAMCOUNTER StatR3AccessedPage; /**< R3: The number of pages marked not present for accessed bit emulation. */
3642 STAMPROFILE StatR3DirtyBitTracking; /**< R3: Profiling the dirty bit tracking in CheckPageFault(). */
3643 STAMCOUNTER StatR3DirtyPage; /**< R3: The number of pages marked read-only for dirty bit tracking. */
3644 STAMCOUNTER StatR3DirtyPageBig; /**< R3: The number of pages marked read-only for dirty bit tracking. */
3645 STAMCOUNTER StatR3DirtyPageSkipped; /**< R3: The number of pages already dirty or readonly. */
3646 STAMCOUNTER StatR3DirtyPageTrap; /**< R3: The number of traps generated for dirty bit tracking. */
3647 STAMCOUNTER StatR3DirtyTrackRealPF; /**< R3: The number of real pages faults during dirty bit tracking. */
3648 STAMCOUNTER StatR3DirtiedPage; /**< R3: The number of pages marked dirty because of write accesses. */
3649 STAMCOUNTER StatR3PageAlreadyDirty; /**< R3: The number of pages already marked dirty because of write accesses. */
3650 STAMPROFILE StatR3InvalidatePage; /**< R3: PGMInvalidatePage() profiling. */
3651 STAMCOUNTER StatR3InvalidatePage4KBPages; /**< R3: The number of times PGMInvalidatePage() was called for a 4KB page. */
3652 STAMCOUNTER StatR3InvalidatePage4MBPages; /**< R3: The number of times PGMInvalidatePage() was called for a 4MB page. */
3653 STAMCOUNTER StatR3InvalidatePage4MBPagesSkip; /**< R3: The number of times PGMInvalidatePage() skipped a 4MB page. */
3654 STAMCOUNTER StatR3InvalidatePagePDNAs; /**< R3: The number of times PGMInvalidatePage() was called for a not accessed page directory. */
3655 STAMCOUNTER StatR3InvalidatePagePDNPs; /**< R3: The number of times PGMInvalidatePage() was called for a not present page directory. */
3656 STAMCOUNTER StatR3InvalidatePagePDMappings; /**< R3: The number of times PGMInvalidatePage() was called for a page directory containing mappings (no conflict). */
3657 STAMCOUNTER StatR3InvalidatePagePDOutOfSync; /**< R3: The number of times PGMInvalidatePage() was called for an out of sync page directory. */
3658 STAMCOUNTER StatR3InvalidatePageSizeChanges ; /**< R3: The number of times PGMInvalidatePage() was called on a page size change (4KB <-> 2/4MB). */
3659 STAMCOUNTER StatR3InvalidatePageSkipped; /**< R3: The number of times PGMInvalidatePage() was skipped due to not present shw or pending pending SyncCR3. */
3660 STAMCOUNTER StatR3PageOutOfSyncUser; /**< R3: The number of times user page is out of sync was detected in \#PF or VerifyAccessSyncPage. */
3661 STAMCOUNTER StatR3PageOutOfSyncSupervisor; /**< R3: The number of times supervisor page is out of sync was detected in in \#PF or VerifyAccessSyncPage. */
3662 STAMCOUNTER StatR3PageOutOfSyncUserWrite; /**< R3: The number of times user page is out of sync was detected in \#PF. */
3663 STAMCOUNTER StatR3PageOutOfSyncSupervisorWrite; /**< R3: The number of times supervisor page is out of sync was detected in in \#PF. */
3664 STAMCOUNTER StatR3PageOutOfSyncBallloon; /**< R3: The number of times a ballooned page was accessed (read). */
3665 STAMPROFILE StatR3Prefetch; /**< R3: PGMPrefetchPage. */
3666 STAMPROFILE StatR3FlushTLB; /**< R3: Profiling of the PGMFlushTLB() body. */
3667 STAMCOUNTER StatR3FlushTLBNewCR3; /**< R3: The number of times PGMFlushTLB was called with a new CR3, non-global. (switch) */
3668 STAMCOUNTER StatR3FlushTLBNewCR3Global; /**< R3: The number of times PGMFlushTLB was called with a new CR3, global. (switch) */
3669 STAMCOUNTER StatR3FlushTLBSameCR3; /**< R3: The number of times PGMFlushTLB was called with the same CR3, non-global. (flush) */
3670 STAMCOUNTER StatR3FlushTLBSameCR3Global; /**< R3: The number of times PGMFlushTLB was called with the same CR3, global. (flush) */
3671 STAMPROFILE StatR3GstModifyPage; /**< R3: Profiling of the PGMGstModifyPage() body */
3672} PGMCPUSTATS;
3673
3674
3675/**
3676 * Converts a PGMCPU pointer into a VM pointer.
3677 * @returns Pointer to the VM structure the PGM is part of.
3678 * @param pPGM Pointer to PGMCPU instance data.
3679 */
3680#define PGMCPU2VM(pPGM) ( (PVM)((char*)(pPGM) - (pPGM)->offVM) )
3681
3682/**
3683 * Converts a PGMCPU pointer into a PGM pointer.
3684 * @returns Pointer to the VM structure the PGM is part of.
3685 * @param pPGMCpu Pointer to PGMCPU instance data.
3686 */
3687#define PGMCPU2PGM(pPGMCpu) ( (PPGM)((char *)(pPGMCpu) - (pPGMCpu)->offPGM) )
3688
3689/**
3690 * PGMCPU Data (part of VMCPU).
3691 */
3692typedef struct PGMCPU
3693{
3694 /** Offset to the VM structure. */
3695 int32_t offVM;
3696 /** Offset to the VMCPU structure. */
3697 int32_t offVCpu;
3698 /** Offset of the PGM structure relative to VMCPU. */
3699 int32_t offPGM;
3700 uint32_t uPadding0; /**< structure size alignment. */
3701
3702#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE
3703 /** Automatically tracked physical memory mapping set.
3704 * Ring-0 and strict raw-mode builds. */
3705 PGMMAPSET AutoSet;
3706#endif
3707
3708 /** A20 gate mask.
3709 * Our current approach to A20 emulation is to let REM do it and don't bother
3710 * anywhere else. The interesting Guests will be operating with it enabled anyway.
3711 * But whould need arrise, we'll subject physical addresses to this mask. */
3712 RTGCPHYS GCPhysA20Mask;
3713 /** A20 gate state - boolean! */
3714 bool fA20Enabled;
3715 /** Mirror of the EFER.NXE bit. Managed by PGMNotifyNxeChanged. */
3716 bool fNoExecuteEnabled;
3717 /** Unused bits. */
3718 bool afUnused[2];
3719
3720 /** What needs syncing (PGM_SYNC_*).
3721 * This is used to queue operations for PGMSyncCR3, PGMInvalidatePage,
3722 * PGMFlushTLB, and PGMR3Load. */
3723 uint32_t fSyncFlags;
3724
3725 /** The shadow paging mode. */
3726 PGMMODE enmShadowMode;
3727 /** The guest paging mode. */
3728 PGMMODE enmGuestMode;
3729 /** Guest mode data table index (PGM_TYPE_XXX). */
3730 uint8_t volatile idxGuestModeData;
3731 /** Shadow mode data table index (PGM_TYPE_XXX). */
3732 uint8_t volatile idxShadowModeData;
3733 /** Both mode data table index (complicated). */
3734 uint8_t volatile idxBothModeData;
3735 /** Alignment padding. */
3736 uint8_t abPadding[5];
3737
3738 /** The current physical address represented in the guest CR3 register. */
3739 RTGCPHYS GCPhysCR3;
3740
3741 /** @name 32-bit Guest Paging.
3742 * @{ */
3743 /** The guest's page directory, R3 pointer. */
3744 R3PTRTYPE(PX86PD) pGst32BitPdR3;
3745#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3746 /** The guest's page directory, R0 pointer. */
3747 R0PTRTYPE(PX86PD) pGst32BitPdR0;
3748#endif
3749 /** Mask containing the MBZ bits of a big page PDE. */
3750 uint32_t fGst32BitMbzBigPdeMask;
3751 /** Set if the page size extension (PSE) is enabled. */
3752 bool fGst32BitPageSizeExtension;
3753 /** Alignment padding. */
3754 bool afAlignment2[3];
3755 /** @} */
3756
3757 /** @name PAE Guest Paging.
3758 * @{ */
3759 /** The guest's page directory pointer table, R3 pointer. */
3760 R3PTRTYPE(PX86PDPT) pGstPaePdptR3;
3761#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3762 /** The guest's page directory pointer table, R0 pointer. */
3763 R0PTRTYPE(PX86PDPT) pGstPaePdptR0;
3764#endif
3765
3766 /** The guest's page directories, R3 pointers.
3767 * These are individual pointers and don't have to be adjacent.
3768 * These don't have to be up-to-date - use pgmGstGetPaePD() to access them. */
3769 R3PTRTYPE(PX86PDPAE) apGstPaePDsR3[4];
3770 /** The guest's page directories, R0 pointers.
3771 * Same restrictions as apGstPaePDsR3. */
3772#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3773 R0PTRTYPE(PX86PDPAE) apGstPaePDsR0[4];
3774#endif
3775 /** The physical addresses of the guest page directories (PAE) pointed to by apGstPagePDsHC/GC.
3776 * @todo Remove this and use aGstPaePdpeRegs instead? */
3777 RTGCPHYS aGCPhysGstPaePDs[4];
3778 /** The values of the 4 PDPE CPU registers (PAE). */
3779 X86PDPE aGstPaePdpeRegs[4];
3780 /** The physical addresses of the monitored guest page directories (PAE). */
3781 RTGCPHYS aGCPhysGstPaePDsMonitored[4];
3782 /** Mask containing the MBZ PTE bits. */
3783 uint64_t fGstPaeMbzPteMask;
3784 /** Mask containing the MBZ PDE bits. */
3785 uint64_t fGstPaeMbzPdeMask;
3786 /** Mask containing the MBZ big page PDE bits. */
3787 uint64_t fGstPaeMbzBigPdeMask;
3788 /** Mask containing the MBZ PDPE bits. */
3789 uint64_t fGstPaeMbzPdpeMask;
3790 /** @} */
3791
3792 /** @name AMD64 Guest Paging.
3793 * @{ */
3794 /** The guest's page directory pointer table, R3 pointer. */
3795 R3PTRTYPE(PX86PML4) pGstAmd64Pml4R3;
3796#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
3797 /** The guest's page directory pointer table, R0 pointer. */
3798 R0PTRTYPE(PX86PML4) pGstAmd64Pml4R0;
3799#else
3800 RTR0PTR alignment6b; /**< alignment equalizer. */
3801#endif
3802 /** Mask containing the MBZ PTE bits. */
3803 uint64_t fGstAmd64MbzPteMask;
3804 /** Mask containing the MBZ PDE bits. */
3805 uint64_t fGstAmd64MbzPdeMask;
3806 /** Mask containing the MBZ big page PDE bits. */
3807 uint64_t fGstAmd64MbzBigPdeMask;
3808 /** Mask containing the MBZ PDPE bits. */
3809 uint64_t fGstAmd64MbzPdpeMask;
3810 /** Mask containing the MBZ big page PDPE bits. */
3811 uint64_t fGstAmd64MbzBigPdpeMask;
3812 /** Mask containing the MBZ PML4E bits. */
3813 uint64_t fGstAmd64MbzPml4eMask;
3814 /** Mask containing the PDPE bits that we shadow. */
3815 uint64_t fGstAmd64ShadowedPdpeMask;
3816 /** Mask containing the PML4E bits that we shadow. */
3817 uint64_t fGstAmd64ShadowedPml4eMask;
3818 /** @} */
3819
3820 /** @name PAE and AMD64 Guest Paging.
3821 * @{ */
3822 /** Mask containing the PTE bits that we shadow. */
3823 uint64_t fGst64ShadowedPteMask;
3824 /** Mask containing the PDE bits that we shadow. */
3825 uint64_t fGst64ShadowedPdeMask;
3826 /** Mask containing the big page PDE bits that we shadow in the PDE. */
3827 uint64_t fGst64ShadowedBigPdeMask;
3828 /** Mask containing the big page PDE bits that we shadow in the PTE. */
3829 uint64_t fGst64ShadowedBigPde4PteMask;
3830 /** @} */
3831
3832 /** Pointer to the page of the current active CR3 - R3 Ptr. */
3833 R3PTRTYPE(PPGMPOOLPAGE) pShwPageCR3R3;
3834 /** Pointer to the page of the current active CR3 - R0 Ptr. */
3835 R0PTRTYPE(PPGMPOOLPAGE) pShwPageCR3R0;
3836
3837 /** For saving stack space, the disassembler state is allocated here instead of
3838 * on the stack. */
3839 DISCPUSTATE DisState;
3840
3841 /** Counts the number of times the netware WP0+RO+US hack has been applied. */
3842 uint64_t cNetwareWp0Hacks;
3843
3844 /** Count the number of pgm pool access handler calls. */
3845 uint64_t cPoolAccessHandler;
3846
3847 /** @name Release Statistics
3848 * @{ */
3849 /** The number of times the guest has switched mode since last reset or statistics reset. */
3850 STAMCOUNTER cGuestModeChanges;
3851 /** The number of times the guest has switched mode since last reset or statistics reset. */
3852 STAMCOUNTER cA20Changes;
3853 /** @} */
3854
3855#ifdef VBOX_WITH_STATISTICS /** @todo move this chunk to the heap. */
3856 /** @name Statistics
3857 * @{ */
3858 /** R0: Pointer to the statistics. */
3859 R0PTRTYPE(PGMCPUSTATS *) pStatsR0;
3860 /** R0: Which statistic this \#PF should be attributed to. */
3861 R0PTRTYPE(PSTAMPROFILE) pStatTrap0eAttributionR0;
3862 /** R3: Pointer to the statistics. */
3863 R3PTRTYPE(PGMCPUSTATS *) pStatsR3;
3864 /** Alignment padding. */
3865 RTR3PTR pPaddingR3;
3866 /** @} */
3867#endif /* VBOX_WITH_STATISTICS */
3868} PGMCPU;
3869/** Pointer to the per-cpu PGM data. */
3870typedef PGMCPU *PPGMCPU;
3871
3872
3873/** @name PGM::fSyncFlags Flags
3874 * @note Was part of saved state a long time ago.
3875 * @{
3876 */
3877/* 0 used to be PGM_SYNC_UPDATE_PAGE_BIT_VIRTUAL */
3878/** Always sync CR3. */
3879#define PGM_SYNC_ALWAYS RT_BIT(1)
3880/** Check monitoring on next CR3 (re)load and invalidate page.
3881 * @todo This is obsolete now. Remove after 2.2.0 is branched off. */
3882#define PGM_SYNC_MONITOR_CR3 RT_BIT(2)
3883/** Check guest mapping in SyncCR3. */
3884#define PGM_SYNC_MAP_CR3 RT_BIT(3)
3885/** Clear the page pool (a light weight flush). */
3886#define PGM_SYNC_CLEAR_PGM_POOL_BIT 8
3887#define PGM_SYNC_CLEAR_PGM_POOL RT_BIT(PGM_SYNC_CLEAR_PGM_POOL_BIT)
3888/** @} */
3889
3890
3891/**
3892 * PGM GVM instance data.
3893 */
3894typedef struct PGMR0PERVM
3895{
3896 /** @name PGM Pool related stuff.
3897 * @{ */
3898 /** Critical section for serializing pool growth. */
3899 RTCRITSECT PoolGrowCritSect;
3900 /** The memory objects for the pool pages. */
3901 RTR0MEMOBJ ahPoolMemObjs[(PGMPOOL_IDX_LAST + PGMPOOL_CFG_MAX_GROW - 1) / PGMPOOL_CFG_MAX_GROW];
3902 /** The ring-3 mapping objects for the pool pages. */
3903 RTR0MEMOBJ ahPoolMapObjs[(PGMPOOL_IDX_LAST + PGMPOOL_CFG_MAX_GROW - 1) / PGMPOOL_CFG_MAX_GROW];
3904 /** @} */
3905} PGMR0PERVM;
3906
3907RT_C_DECLS_BEGIN
3908
3909#if defined(VBOX_STRICT) && defined(IN_RING3)
3910int pgmLockDebug(PVMCC pVM, RT_SRC_POS_DECL);
3911# define pgmLock(a_pVM) pgmLockDebug(a_pVM, RT_SRC_POS)
3912#else
3913int pgmLock(PVMCC pVM);
3914#endif
3915void pgmUnlock(PVM pVM);
3916/**
3917 * Asserts that the caller owns the PDM lock.
3918 * This is the internal variant of PGMIsLockOwner.
3919 * @param a_pVM Pointer to the VM.
3920 */
3921#define PGM_LOCK_ASSERT_OWNER(a_pVM) Assert(PDMCritSectIsOwner(&(a_pVM)->pgm.s.CritSectX))
3922/**
3923 * Asserts that the caller owns the PDM lock.
3924 * This is the internal variant of PGMIsLockOwner.
3925 * @param a_pVM Pointer to the VM.
3926 * @param a_pVCpu The current CPU handle.
3927 */
3928#define PGM_LOCK_ASSERT_OWNER_EX(a_pVM, a_pVCpu) Assert(PDMCritSectIsOwnerEx(&(a_pVM)->pgm.s.CritSectX, a_pVCpu))
3929
3930#ifndef PGM_WITHOUT_MAPPINGS
3931int pgmR3MappingsFixInternal(PVM pVM, RTGCPTR GCPtrBase, uint32_t cb);
3932int pgmR3SyncPTResolveConflict(PVM pVM, PPGMMAPPING pMapping, PX86PD pPDSrc, RTGCPTR GCPtrOldMapping);
3933int pgmR3SyncPTResolveConflictPAE(PVM pVM, PPGMMAPPING pMapping, RTGCPTR GCPtrOldMapping);
3934int pgmMapResolveConflicts(PVM pVM);
3935PPGMMAPPING pgmGetMapping(PVM pVM, RTGCPTR GCPtr);
3936DECLCALLBACK(void) pgmR3MapInfo(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
3937#endif /* !PGM_WITHOUT_MAPPINGS */
3938
3939int pgmHandlerPhysicalExCreate(PVMCC pVM, PGMPHYSHANDLERTYPE hType, RTR3PTR pvUserR3, RTR0PTR pvUserR0,
3940 RTRCPTR pvUserRC, R3PTRTYPE(const char *) pszDesc, PPGMPHYSHANDLER *ppPhysHandler);
3941int pgmHandlerPhysicalExDup(PVMCC pVM, PPGMPHYSHANDLER pPhysHandlerSrc, PPGMPHYSHANDLER *ppPhysHandler);
3942int pgmHandlerPhysicalExRegister(PVMCC pVM, PPGMPHYSHANDLER pPhysHandler, RTGCPHYS GCPhys, RTGCPHYS GCPhysLast);
3943int pgmHandlerPhysicalExDeregister(PVMCC pVM, PPGMPHYSHANDLER pPhysHandler, int fRestoreAsRAM);
3944int pgmHandlerPhysicalExDestroy(PVMCC pVM, PPGMPHYSHANDLER pHandler);
3945void pgmR3HandlerPhysicalUpdateAll(PVM pVM);
3946bool pgmHandlerPhysicalIsAll(PVMCC pVM, RTGCPHYS GCPhys);
3947void pgmHandlerPhysicalResetAliasedPage(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhysPage, bool fDoAccounting);
3948DECLCALLBACK(void) pgmR3InfoHandlers(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
3949int pgmR3InitSavedState(PVM pVM, uint64_t cbRam);
3950
3951int pgmPhysAllocPage(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3952int pgmPhysAllocLargePage(PVMCC pVM, RTGCPHYS GCPhys);
3953int pgmPhysRecheckLargePage(PVMCC pVM, RTGCPHYS GCPhys, PPGMPAGE pLargePage);
3954int pgmPhysPageLoadIntoTlb(PVMCC pVM, RTGCPHYS GCPhys);
3955int pgmPhysPageLoadIntoTlbWithPage(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3956void pgmPhysPageMakeWriteMonitoredWritable(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3957int pgmPhysPageMakeWritable(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys);
3958int pgmPhysPageMakeWritableAndMap(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv);
3959int pgmPhysPageMap(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv);
3960int pgmPhysPageMapReadOnly(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void const **ppv);
3961int pgmPhysPageMapByPageID(PVMCC pVM, uint32_t idPage, RTHCPHYS HCPhys, void **ppv);
3962int pgmPhysGCPhys2R3Ptr(PVMCC pVM, RTGCPHYS GCPhys, PRTR3PTR pR3Ptr);
3963int pgmPhysCr3ToHCPtr(PVM pVM, RTGCPHYS GCPhys, PRTR3PTR pR3Ptr);
3964int pgmPhysGCPhys2CCPtrInternalDepr(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv);
3965int pgmPhysGCPhys2CCPtrInternal(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, void **ppv, PPGMPAGEMAPLOCK pLock);
3966int pgmPhysGCPhys2CCPtrInternalReadOnly(PVMCC pVM, PPGMPAGE pPage, RTGCPHYS GCPhys, const void **ppv, PPGMPAGEMAPLOCK pLock);
3967void pgmPhysReleaseInternalPageMappingLock(PVMCC pVM, PPGMPAGEMAPLOCK pLock);
3968PGM_ALL_CB2_PROTO(FNPGMPHYSHANDLER) pgmPhysRomWriteHandler;
3969#ifndef IN_RING3
3970DECLEXPORT(FNPGMPHYSHANDLER) pgmPhysHandlerRedirectToHC;
3971DECLEXPORT(FNPGMRZPHYSPFHANDLER) pgmPhysPfHandlerRedirectToHC;
3972DECLEXPORT(FNPGMRZPHYSPFHANDLER) pgmPhysRomWritePfHandler;
3973#endif
3974int pgmPhysFreePage(PVM pVM, PGMMFREEPAGESREQ pReq, uint32_t *pcPendingPages, PPGMPAGE pPage, RTGCPHYS GCPhys,
3975 PGMPAGETYPE enmNewType);
3976void pgmPhysInvalidRamRangeTlbs(PVMCC pVM);
3977void pgmPhysInvalidatePageMapTLB(PVMCC pVM);
3978void pgmPhysInvalidatePageMapTLBEntry(PVM pVM, RTGCPHYS GCPhys);
3979PPGMRAMRANGE pgmPhysGetRangeSlow(PVM pVM, RTGCPHYS GCPhys);
3980PPGMRAMRANGE pgmPhysGetRangeAtOrAboveSlow(PVM pVM, RTGCPHYS GCPhys);
3981PPGMPAGE pgmPhysGetPageSlow(PVM pVM, RTGCPHYS GCPhys);
3982int pgmPhysGetPageExSlow(PVM pVM, RTGCPHYS GCPhys, PPPGMPAGE ppPage);
3983int pgmPhysGetPageAndRangeExSlow(PVM pVM, RTGCPHYS GCPhys, PPPGMPAGE ppPage, PPGMRAMRANGE *ppRam);
3984
3985#ifdef IN_RING3
3986void pgmR3PhysRelinkRamRanges(PVM pVM);
3987int pgmR3PhysRamPreAllocate(PVM pVM);
3988int pgmR3PhysRamReset(PVM pVM);
3989int pgmR3PhysRomReset(PVM pVM);
3990int pgmR3PhysRamZeroAll(PVM pVM);
3991int pgmR3PhysChunkMap(PVM pVM, uint32_t idChunk, PPPGMCHUNKR3MAP ppChunk);
3992int pgmR3PhysRamTerm(PVM pVM);
3993void pgmR3PhysRomTerm(PVM pVM);
3994void pgmR3PhysAssertSharedPageChecksums(PVM pVM);
3995
3996int pgmR3PoolInit(PVM pVM);
3997void pgmR3PoolRelocate(PVM pVM);
3998void pgmR3PoolResetUnpluggedCpu(PVM pVM, PVMCPU pVCpu);
3999void pgmR3PoolReset(PVM pVM);
4000void pgmR3PoolClearAll(PVM pVM, bool fFlushRemTlb);
4001DECLCALLBACK(VBOXSTRICTRC) pgmR3PoolClearAllRendezvous(PVM pVM, PVMCPU pVCpu, void *fpvFlushRemTbl);
4002void pgmR3PoolWriteProtectPages(PVM pVM);
4003
4004#endif /* IN_RING3 */
4005#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE_IN_R0
4006int pgmRZDynMapHCPageCommon(PPGMMAPSET pSet, RTHCPHYS HCPhys, void **ppv RTLOG_COMMA_SRC_POS_DECL);
4007int pgmRZDynMapGCPageCommon(PVM pVM, PVMCPU pVCpu, RTGCPHYS GCPhys, void **ppv RTLOG_COMMA_SRC_POS_DECL);
4008# ifdef LOG_ENABLED
4009void pgmRZDynMapUnusedHint(PVMCPU pVCpu, void *pvHint, RT_SRC_POS_DECL);
4010# else
4011void pgmRZDynMapUnusedHint(PVMCPU pVCpu, void *pvHint);
4012# endif
4013#endif
4014int pgmPoolAlloc(PVMCC pVM, RTGCPHYS GCPhys, PGMPOOLKIND enmKind, PGMPOOLACCESS enmAccess, bool fA20Enabled,
4015 uint16_t iUser, uint32_t iUserTable, bool fLockPage, PPPGMPOOLPAGE ppPage);
4016void pgmPoolFree(PVM pVM, RTHCPHYS HCPhys, uint16_t iUser, uint32_t iUserTable);
4017void pgmPoolFreeByPage(PPGMPOOL pPool, PPGMPOOLPAGE pPage, uint16_t iUser, uint32_t iUserTable);
4018int pgmPoolFlushPage(PPGMPOOL pPool, PPGMPOOLPAGE pPage, bool fFlush = true /* DO NOT USE false UNLESS YOU KNOWN WHAT YOU'RE DOING!! */);
4019void pgmPoolFlushPageByGCPhys(PVM pVM, RTGCPHYS GCPhys);
4020PPGMPOOLPAGE pgmPoolGetPage(PPGMPOOL pPool, RTHCPHYS HCPhys);
4021PPGMPOOLPAGE pgmPoolQueryPageForDbg(PPGMPOOL pPool, RTHCPHYS HCPhys);
4022int pgmPoolHCPhys2Ptr(PVM pVM, RTHCPHYS HCPhys, void **ppv);
4023int pgmPoolSyncCR3(PVMCPUCC pVCpu);
4024bool pgmPoolIsDirtyPageSlow(PVM pVM, RTGCPHYS GCPhys);
4025void pgmPoolInvalidateDirtyPage(PVMCC pVM, RTGCPHYS GCPhysPT);
4026int pgmPoolTrackUpdateGCPhys(PVMCC pVM, RTGCPHYS GCPhysPage, PPGMPAGE pPhysPage, bool fFlushPTEs, bool *pfFlushTLBs);
4027void pgmPoolTracDerefGCPhysHint(PPGMPOOL pPool, PPGMPOOLPAGE pPage, RTHCPHYS HCPhys, RTGCPHYS GCPhysHint, uint16_t iPte);
4028uint16_t pgmPoolTrackPhysExtAddref(PVMCC pVM, PPGMPAGE pPhysPage, uint16_t u16, uint16_t iShwPT, uint16_t iPte);
4029void pgmPoolTrackPhysExtDerefGCPhys(PPGMPOOL pPool, PPGMPOOLPAGE pPoolPage, PPGMPAGE pPhysPage, uint16_t iPte);
4030void pgmPoolMonitorChainFlush(PPGMPOOL pPool, PPGMPOOLPAGE pPage);
4031void pgmPoolMonitorModifiedInsert(PPGMPOOL pPool, PPGMPOOLPAGE pPage);
4032PGM_ALL_CB2_PROTO(FNPGMPHYSHANDLER) pgmPoolAccessHandler;
4033#ifndef IN_RING3
4034DECLEXPORT(FNPGMRZPHYSPFHANDLER) pgmRZPoolAccessPfHandler;
4035#endif
4036
4037void pgmPoolAddDirtyPage(PVMCC pVM, PPGMPOOL pPool, PPGMPOOLPAGE pPage);
4038void pgmPoolResetDirtyPages(PVMCC pVM);
4039void pgmPoolResetDirtyPage(PVM pVM, RTGCPTR GCPtrPage);
4040
4041int pgmR3ExitShadowModeBeforePoolFlush(PVMCPU pVCpu);
4042int pgmR3ReEnterShadowModeAfterPoolFlush(PVM pVM, PVMCPU pVCpu);
4043void pgmR3RefreshShadowModeAfterA20Change(PVMCPU pVCpu);
4044
4045#ifndef PGM_WITHOUT_MAPPINGS
4046void pgmMapSetShadowPDEs(PVM pVM, PPGMMAPPING pMap, unsigned iNewPDE);
4047void pgmMapClearShadowPDEs(PVM pVM, PPGMPOOLPAGE pShwPageCR3, PPGMMAPPING pMap, unsigned iOldPDE, bool fDeactivateCR3);
4048int pgmMapActivateCR3(PVM pVM, PPGMPOOLPAGE pShwPageCR3);
4049int pgmMapDeactivateCR3(PVM pVM, PPGMPOOLPAGE pShwPageCR3);
4050#endif
4051
4052int pgmShwMakePageSupervisorAndWritable(PVMCPUCC pVCpu, RTGCPTR GCPtr, bool fBigPage, uint32_t fOpFlags);
4053int pgmShwSyncPaePDPtr(PVMCPUCC pVCpu, RTGCPTR GCPtr, X86PGPAEUINT uGstPdpe, PX86PDPAE *ppPD);
4054int pgmShwSyncNestedPageLocked(PVMCPUCC pVCpu, RTGCPHYS GCPhysFault, uint32_t cPages, PGMMODE enmShwPagingMode);
4055
4056int pgmGstLazyMap32BitPD(PVMCPUCC pVCpu, PX86PD *ppPd);
4057int pgmGstLazyMapPaePDPT(PVMCPUCC pVCpu, PX86PDPT *ppPdpt);
4058int pgmGstLazyMapPaePD(PVMCPUCC pVCpu, uint32_t iPdpt, PX86PDPAE *ppPd);
4059int pgmGstLazyMapPml4(PVMCPUCC pVCpu, PX86PML4 *ppPml4);
4060int pgmGstPtWalk(PVMCPUCC pVCpu, RTGCPTR GCPtr, PPGMPTWALKGST pWalk);
4061int pgmGstPtWalkNext(PVMCPUCC pVCpu, RTGCPTR GCPtr, PPGMPTWALKGST pWalk);
4062
4063# if defined(VBOX_STRICT) && HC_ARCH_BITS == 64 && defined(IN_RING3)
4064FNDBGCCMD pgmR3CmdCheckDuplicatePages;
4065FNDBGCCMD pgmR3CmdShowSharedModules;
4066# endif
4067
4068void pgmLogState(PVM pVM);
4069
4070RT_C_DECLS_END
4071
4072/** @} */
4073
4074#endif /* !VMM_INCLUDED_SRC_include_PGMInternal_h */
4075
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