VirtualBox

source: vbox/trunk/src/VBox/VMM/PGM.cpp@ 20062

Last change on this file since 20062 was 20012, checked in by vboxsync, 16 years ago

PGM: MMR3PageAlloc() / MMR3PageAllocLow() already set the error info in case of failure

  • Property svn:eol-style set to native
  • Property svn:keywords set to Id
File size: 230.9 KB
Line 
1/* $Id: PGM.cpp 20012 2009-05-25 19:38:47Z vboxsync $ */
2/** @file
3 * PGM - Page Manager and Monitor. (Mixing stuff here, not good?)
4 */
5
6/*
7 * Copyright (C) 2006-2007 Sun Microsystems, Inc.
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 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa
18 * Clara, CA 95054 USA or visit http://www.sun.com if you need
19 * additional information or have any questions.
20 */
21
22
23/** @page pg_pgm PGM - The Page Manager and Monitor
24 *
25 * @see grp_pgm,
26 * @ref pg_pgm_pool,
27 * @ref pg_pgm_phys.
28 *
29 *
30 * @section sec_pgm_modes Paging Modes
31 *
32 * There are three memory contexts: Host Context (HC), Guest Context (GC)
33 * and intermediate context. When talking about paging HC can also be refered to
34 * as "host paging", and GC refered to as "shadow paging".
35 *
36 * We define three basic paging modes: 32-bit, PAE and AMD64. The host paging mode
37 * is defined by the host operating system. The mode used in the shadow paging mode
38 * depends on the host paging mode and what the mode the guest is currently in. The
39 * following relation between the two is defined:
40 *
41 * @verbatim
42 Host > 32-bit | PAE | AMD64 |
43 Guest | | | |
44 ==v================================
45 32-bit 32-bit PAE PAE
46 -------|--------|--------|--------|
47 PAE PAE PAE PAE
48 -------|--------|--------|--------|
49 AMD64 AMD64 AMD64 AMD64
50 -------|--------|--------|--------| @endverbatim
51 *
52 * All configuration except those in the diagonal (upper left) are expected to
53 * require special effort from the switcher (i.e. a bit slower).
54 *
55 *
56 *
57 *
58 * @section sec_pgm_shw The Shadow Memory Context
59 *
60 *
61 * [..]
62 *
63 * Because of guest context mappings requires PDPT and PML4 entries to allow
64 * writing on AMD64, the two upper levels will have fixed flags whatever the
65 * guest is thinking of using there. So, when shadowing the PD level we will
66 * calculate the effective flags of PD and all the higher levels. In legacy
67 * PAE mode this only applies to the PWT and PCD bits (the rest are
68 * ignored/reserved/MBZ). We will ignore those bits for the present.
69 *
70 *
71 *
72 * @section sec_pgm_int The Intermediate Memory Context
73 *
74 * The world switch goes thru an intermediate memory context which purpose it is
75 * to provide different mappings of the switcher code. All guest mappings are also
76 * present in this context.
77 *
78 * The switcher code is mapped at the same location as on the host, at an
79 * identity mapped location (physical equals virtual address), and at the
80 * hypervisor location. The identity mapped location is for when the world
81 * switches that involves disabling paging.
82 *
83 * PGM maintain page tables for 32-bit, PAE and AMD64 paging modes. This
84 * simplifies switching guest CPU mode and consistency at the cost of more
85 * code to do the work. All memory use for those page tables is located below
86 * 4GB (this includes page tables for guest context mappings).
87 *
88 *
89 * @subsection subsec_pgm_int_gc Guest Context Mappings
90 *
91 * During assignment and relocation of a guest context mapping the intermediate
92 * memory context is used to verify the new location.
93 *
94 * Guest context mappings are currently restricted to below 4GB, for reasons
95 * of simplicity. This may change when we implement AMD64 support.
96 *
97 *
98 *
99 *
100 * @section sec_pgm_misc Misc
101 *
102 * @subsection subsec_pgm_misc_diff Differences Between Legacy PAE and Long Mode PAE
103 *
104 * The differences between legacy PAE and long mode PAE are:
105 * -# PDPE bits 1, 2, 5 and 6 are defined differently. In leagcy mode they are
106 * all marked down as must-be-zero, while in long mode 1, 2 and 5 have the
107 * usual meanings while 6 is ignored (AMD). This means that upon switching to
108 * legacy PAE mode we'll have to clear these bits and when going to long mode
109 * they must be set. This applies to both intermediate and shadow contexts,
110 * however we don't need to do it for the intermediate one since we're
111 * executing with CR0.WP at that time.
112 * -# CR3 allows a 32-byte aligned address in legacy mode, while in long mode
113 * a page aligned one is required.
114 *
115 *
116 * @section sec_pgm_handlers Access Handlers
117 *
118 * Placeholder.
119 *
120 *
121 * @subsection sec_pgm_handlers_virt Virtual Access Handlers
122 *
123 * Placeholder.
124 *
125 *
126 * @subsection sec_pgm_handlers_virt Virtual Access Handlers
127 *
128 * We currently implement three types of virtual access handlers: ALL, WRITE
129 * and HYPERVISOR (WRITE). See PGMVIRTHANDLERTYPE for some more details.
130 *
131 * The HYPERVISOR access handlers is kept in a separate tree since it doesn't apply
132 * to physical pages (PGMTREES::HyperVirtHandlers) and only needs to be consulted in
133 * a special \#PF case. The ALL and WRITE are in the PGMTREES::VirtHandlers tree, the
134 * rest of this section is going to be about these handlers.
135 *
136 * We'll go thru the life cycle of a handler and try make sense of it all, don't know
137 * how successfull this is gonna be...
138 *
139 * 1. A handler is registered thru the PGMR3HandlerVirtualRegister and
140 * PGMHandlerVirtualRegisterEx APIs. We check for conflicting virtual handlers
141 * and create a new node that is inserted into the AVL tree (range key). Then
142 * a full PGM resync is flagged (clear pool, sync cr3, update virtual bit of PGMPAGE).
143 *
144 * 2. The following PGMSyncCR3/SyncCR3 operation will first make invoke HandlerVirtualUpdate.
145 *
146 * 2a. HandlerVirtualUpdate will will lookup all the pages covered by virtual handlers
147 * via the current guest CR3 and update the physical page -> virtual handler
148 * translation. Needless to say, this doesn't exactly scale very well. If any changes
149 * are detected, it will flag a virtual bit update just like we did on registration.
150 * PGMPHYS pages with changes will have their virtual handler state reset to NONE.
151 *
152 * 2b. The virtual bit update process will iterate all the pages covered by all the
153 * virtual handlers and update the PGMPAGE virtual handler state to the max of all
154 * virtual handlers on that page.
155 *
156 * 2c. Back in SyncCR3 we will now flush the entire shadow page cache to make sure
157 * we don't miss any alias mappings of the monitored pages.
158 *
159 * 2d. SyncCR3 will then proceed with syncing the CR3 table.
160 *
161 * 3. \#PF(np,read) on a page in the range. This will cause it to be synced
162 * read-only and resumed if it's a WRITE handler. If it's an ALL handler we
163 * will call the handlers like in the next step. If the physical mapping has
164 * changed we will - some time in the future - perform a handler callback
165 * (optional) and update the physical -> virtual handler cache.
166 *
167 * 4. \#PF(,write) on a page in the range. This will cause the handler to
168 * be invoked.
169 *
170 * 5. The guest invalidates the page and changes the physical backing or
171 * unmaps it. This should cause the invalidation callback to be invoked
172 * (it might not yet be 100% perfect). Exactly what happens next... is
173 * this where we mess up and end up out of sync for a while?
174 *
175 * 6. The handler is deregistered by the client via PGMHandlerVirtualDeregister.
176 * We will then set all PGMPAGEs in the physical -> virtual handler cache for
177 * this handler to NONE and trigger a full PGM resync (basically the same
178 * as int step 1). Which means 2 is executed again.
179 *
180 *
181 * @subsubsection sub_sec_pgm_handler_virt_todo TODOs
182 *
183 * There is a bunch of things that needs to be done to make the virtual handlers
184 * work 100% correctly and work more efficiently.
185 *
186 * The first bit hasn't been implemented yet because it's going to slow the
187 * whole mess down even more, and besides it seems to be working reliably for
188 * our current uses. OTOH, some of the optimizations might end up more or less
189 * implementing the missing bits, so we'll see.
190 *
191 * On the optimization side, the first thing to do is to try avoid unnecessary
192 * cache flushing. Then try team up with the shadowing code to track changes
193 * in mappings by means of access to them (shadow in), updates to shadows pages,
194 * invlpg, and shadow PT discarding (perhaps).
195 *
196 * Some idea that have popped up for optimization for current and new features:
197 * - bitmap indicating where there are virtual handlers installed.
198 * (4KB => 2**20 pages, page 2**12 => covers 32-bit address space 1:1!)
199 * - Further optimize this by min/max (needs min/max avl getters).
200 * - Shadow page table entry bit (if any left)?
201 *
202 */
203
204
205/** @page pg_pgm_phys PGM Physical Guest Memory Management
206 *
207 *
208 * Objectives:
209 * - Guest RAM over-commitment using memory ballooning,
210 * zero pages and general page sharing.
211 * - Moving or mirroring a VM onto a different physical machine.
212 *
213 *
214 * @subsection subsec_pgmPhys_Definitions Definitions
215 *
216 * Allocation chunk - A RTR0MemObjAllocPhysNC object and the tracking
217 * machinery assoicated with it.
218 *
219 *
220 *
221 *
222 * @subsection subsec_pgmPhys_AllocPage Allocating a page.
223 *
224 * Initially we map *all* guest memory to the (per VM) zero page, which
225 * means that none of the read functions will cause pages to be allocated.
226 *
227 * Exception, access bit in page tables that have been shared. This must
228 * be handled, but we must also make sure PGMGst*Modify doesn't make
229 * unnecessary modifications.
230 *
231 * Allocation points:
232 * - PGMPhysSimpleWriteGCPhys and PGMPhysWrite.
233 * - Replacing a zero page mapping at \#PF.
234 * - Replacing a shared page mapping at \#PF.
235 * - ROM registration (currently MMR3RomRegister).
236 * - VM restore (pgmR3Load).
237 *
238 * For the first three it would make sense to keep a few pages handy
239 * until we've reached the max memory commitment for the VM.
240 *
241 * For the ROM registration, we know exactly how many pages we need
242 * and will request these from ring-0. For restore, we will save
243 * the number of non-zero pages in the saved state and allocate
244 * them up front. This would allow the ring-0 component to refuse
245 * the request if the isn't sufficient memory available for VM use.
246 *
247 * Btw. for both ROM and restore allocations we won't be requiring
248 * zeroed pages as they are going to be filled instantly.
249 *
250 *
251 * @subsection subsec_pgmPhys_FreePage Freeing a page
252 *
253 * There are a few points where a page can be freed:
254 * - After being replaced by the zero page.
255 * - After being replaced by a shared page.
256 * - After being ballooned by the guest additions.
257 * - At reset.
258 * - At restore.
259 *
260 * When freeing one or more pages they will be returned to the ring-0
261 * component and replaced by the zero page.
262 *
263 * The reasoning for clearing out all the pages on reset is that it will
264 * return us to the exact same state as on power on, and may thereby help
265 * us reduce the memory load on the system. Further it might have a
266 * (temporary) positive influence on memory fragmentation (@see subsec_pgmPhys_Fragmentation).
267 *
268 * On restore, as mention under the allocation topic, pages should be
269 * freed / allocated depending on how many is actually required by the
270 * new VM state. The simplest approach is to do like on reset, and free
271 * all non-ROM pages and then allocate what we need.
272 *
273 * A measure to prevent some fragmentation, would be to let each allocation
274 * chunk have some affinity towards the VM having allocated the most pages
275 * from it. Also, try make sure to allocate from allocation chunks that
276 * are almost full. Admittedly, both these measures might work counter to
277 * our intentions and its probably not worth putting a lot of effort,
278 * cpu time or memory into this.
279 *
280 *
281 * @subsection subsec_pgmPhys_SharePage Sharing a page
282 *
283 * The basic idea is that there there will be a idle priority kernel
284 * thread walking the non-shared VM pages hashing them and looking for
285 * pages with the same checksum. If such pages are found, it will compare
286 * them byte-by-byte to see if they actually are identical. If found to be
287 * identical it will allocate a shared page, copy the content, check that
288 * the page didn't change while doing this, and finally request both the
289 * VMs to use the shared page instead. If the page is all zeros (special
290 * checksum and byte-by-byte check) it will request the VM that owns it
291 * to replace it with the zero page.
292 *
293 * To make this efficient, we will have to make sure not to try share a page
294 * that will change its contents soon. This part requires the most work.
295 * A simple idea would be to request the VM to write monitor the page for
296 * a while to make sure it isn't modified any time soon. Also, it may
297 * make sense to skip pages that are being write monitored since this
298 * information is readily available to the thread if it works on the
299 * per-VM guest memory structures (presently called PGMRAMRANGE).
300 *
301 *
302 * @subsection subsec_pgmPhys_Fragmentation Fragmentation Concerns and Counter Measures
303 *
304 * The pages are organized in allocation chunks in ring-0, this is a necessity
305 * if we wish to have an OS agnostic approach to this whole thing. (On Linux we
306 * could easily work on a page-by-page basis if we liked. Whether this is possible
307 * or efficient on NT I don't quite know.) Fragmentation within these chunks may
308 * become a problem as part of the idea here is that we wish to return memory to
309 * the host system.
310 *
311 * For instance, starting two VMs at the same time, they will both allocate the
312 * guest memory on-demand and if permitted their page allocations will be
313 * intermixed. Shut down one of the two VMs and it will be difficult to return
314 * any memory to the host system because the page allocation for the two VMs are
315 * mixed up in the same allocation chunks.
316 *
317 * To further complicate matters, when pages are freed because they have been
318 * ballooned or become shared/zero the whole idea is that the page is supposed
319 * to be reused by another VM or returned to the host system. This will cause
320 * allocation chunks to contain pages belonging to different VMs and prevent
321 * returning memory to the host when one of those VM shuts down.
322 *
323 * The only way to really deal with this problem is to move pages. This can
324 * either be done at VM shutdown and or by the idle priority worker thread
325 * that will be responsible for finding sharable/zero pages. The mechanisms
326 * involved for coercing a VM to move a page (or to do it for it) will be
327 * the same as when telling it to share/zero a page.
328 *
329 *
330 * @subsection subsec_pgmPhys_Tracking Tracking Structures And Their Cost
331 *
332 * There's a difficult balance between keeping the per-page tracking structures
333 * (global and guest page) easy to use and keeping them from eating too much
334 * memory. We have limited virtual memory resources available when operating in
335 * 32-bit kernel space (on 64-bit there'll it's quite a different story). The
336 * tracking structures will be attemted designed such that we can deal with up
337 * to 32GB of memory on a 32-bit system and essentially unlimited on 64-bit ones.
338 *
339 *
340 * @subsubsection subsubsec_pgmPhys_Tracking_Kernel Kernel Space
341 *
342 * @see pg_GMM
343 *
344 * @subsubsection subsubsec_pgmPhys_Tracking_PerVM Per-VM
345 *
346 * Fixed info is the physical address of the page (HCPhys) and the page id
347 * (described above). Theoretically we'll need 48(-12) bits for the HCPhys part.
348 * Today we've restricting ourselves to 40(-12) bits because this is the current
349 * restrictions of all AMD64 implementations (I think Barcelona will up this
350 * to 48(-12) bits, not that it really matters) and I needed the bits for
351 * tracking mappings of a page. 48-12 = 36. That leaves 28 bits, which means a
352 * decent range for the page id: 2^(28+12) = 1024TB.
353 *
354 * In additions to these, we'll have to keep maintaining the page flags as we
355 * currently do. Although it wouldn't harm to optimize these quite a bit, like
356 * for instance the ROM shouldn't depend on having a write handler installed
357 * in order for it to become read-only. A RO/RW bit should be considered so
358 * that the page syncing code doesn't have to mess about checking multiple
359 * flag combinations (ROM || RW handler || write monitored) in order to
360 * figure out how to setup a shadow PTE. But this of course, is second
361 * priority at present. Current this requires 12 bits, but could probably
362 * be optimized to ~8.
363 *
364 * Then there's the 24 bits used to track which shadow page tables are
365 * currently mapping a page for the purpose of speeding up physical
366 * access handlers, and thereby the page pool cache. More bit for this
367 * purpose wouldn't hurt IIRC.
368 *
369 * Then there is a new bit in which we need to record what kind of page
370 * this is, shared, zero, normal or write-monitored-normal. This'll
371 * require 2 bits. One bit might be needed for indicating whether a
372 * write monitored page has been written to. And yet another one or
373 * two for tracking migration status. 3-4 bits total then.
374 *
375 * Whatever is left will can be used to record the sharabilitiy of a
376 * page. The page checksum will not be stored in the per-VM table as
377 * the idle thread will not be permitted to do modifications to it.
378 * It will instead have to keep its own working set of potentially
379 * shareable pages and their check sums and stuff.
380 *
381 * For the present we'll keep the current packing of the
382 * PGMRAMRANGE::aHCPhys to keep the changes simple, only of course,
383 * we'll have to change it to a struct with a total of 128-bits at
384 * our disposal.
385 *
386 * The initial layout will be like this:
387 * @verbatim
388 RTHCPHYS HCPhys; The current stuff.
389 63:40 Current shadow PT tracking stuff.
390 39:12 The physical page frame number.
391 11:0 The current flags.
392 uint32_t u28PageId : 28; The page id.
393 uint32_t u2State : 2; The page state { zero, shared, normal, write monitored }.
394 uint32_t fWrittenTo : 1; Whether a write monitored page was written to.
395 uint32_t u1Reserved : 1; Reserved for later.
396 uint32_t u32Reserved; Reserved for later, mostly sharing stats.
397 @endverbatim
398 *
399 * The final layout will be something like this:
400 * @verbatim
401 RTHCPHYS HCPhys; The current stuff.
402 63:48 High page id (12+).
403 47:12 The physical page frame number.
404 11:0 Low page id.
405 uint32_t fReadOnly : 1; Whether it's readonly page (rom or monitored in some way).
406 uint32_t u3Type : 3; The page type {RESERVED, MMIO, MMIO2, ROM, shadowed ROM, RAM}.
407 uint32_t u2PhysMon : 2; Physical access handler type {none, read, write, all}.
408 uint32_t u2VirtMon : 2; Virtual access handler type {none, read, write, all}..
409 uint32_t u2State : 2; The page state { zero, shared, normal, write monitored }.
410 uint32_t fWrittenTo : 1; Whether a write monitored page was written to.
411 uint32_t u20Reserved : 20; Reserved for later, mostly sharing stats.
412 uint32_t u32Tracking; The shadow PT tracking stuff, roughly.
413 @endverbatim
414 *
415 * Cost wise, this means we'll double the cost for guest memory. There isn't anyway
416 * around that I'm afraid. It means that the cost of dealing out 32GB of memory
417 * to one or more VMs is: (32GB >> PAGE_SHIFT) * 16 bytes, or 128MBs. Or another
418 * example, the VM heap cost when assigning 1GB to a VM will be: 4MB.
419 *
420 * A couple of cost examples for the total cost per-VM + kernel.
421 * 32-bit Windows and 32-bit linux:
422 * 1GB guest ram, 256K pages: 4MB + 2MB(+) = 6MB
423 * 4GB guest ram, 1M pages: 16MB + 8MB(+) = 24MB
424 * 32GB guest ram, 8M pages: 128MB + 64MB(+) = 192MB
425 * 64-bit Windows and 64-bit linux:
426 * 1GB guest ram, 256K pages: 4MB + 3MB(+) = 7MB
427 * 4GB guest ram, 1M pages: 16MB + 12MB(+) = 28MB
428 * 32GB guest ram, 8M pages: 128MB + 96MB(+) = 224MB
429 *
430 * UPDATE - 2007-09-27:
431 * Will need a ballooned flag/state too because we cannot
432 * trust the guest 100% and reporting the same page as ballooned more
433 * than once will put the GMM off balance.
434 *
435 *
436 * @subsection subsec_pgmPhys_Serializing Serializing Access
437 *
438 * Initially, we'll try a simple scheme:
439 *
440 * - The per-VM RAM tracking structures (PGMRAMRANGE) is only modified
441 * by the EMT thread of that VM while in the pgm critsect.
442 * - Other threads in the VM process that needs to make reliable use of
443 * the per-VM RAM tracking structures will enter the critsect.
444 * - No process external thread or kernel thread will ever try enter
445 * the pgm critical section, as that just won't work.
446 * - The idle thread (and similar threads) doesn't not need 100% reliable
447 * data when performing it tasks as the EMT thread will be the one to
448 * do the actual changes later anyway. So, as long as it only accesses
449 * the main ram range, it can do so by somehow preventing the VM from
450 * being destroyed while it works on it...
451 *
452 * - The over-commitment management, including the allocating/freeing
453 * chunks, is serialized by a ring-0 mutex lock (a fast one since the
454 * more mundane mutex implementation is broken on Linux).
455 * - A separeate mutex is protecting the set of allocation chunks so
456 * that pages can be shared or/and freed up while some other VM is
457 * allocating more chunks. This mutex can be take from under the other
458 * one, but not the otherway around.
459 *
460 *
461 * @subsection subsec_pgmPhys_Request VM Request interface
462 *
463 * When in ring-0 it will become necessary to send requests to a VM so it can
464 * for instance move a page while defragmenting during VM destroy. The idle
465 * thread will make use of this interface to request VMs to setup shared
466 * pages and to perform write monitoring of pages.
467 *
468 * I would propose an interface similar to the current VMReq interface, similar
469 * in that it doesn't require locking and that the one sending the request may
470 * wait for completion if it wishes to. This shouldn't be very difficult to
471 * realize.
472 *
473 * The requests themselves are also pretty simple. They are basically:
474 * -# Check that some precondition is still true.
475 * -# Do the update.
476 * -# Update all shadow page tables involved with the page.
477 *
478 * The 3rd step is identical to what we're already doing when updating a
479 * physical handler, see pgmHandlerPhysicalSetRamFlagsAndFlushShadowPTs.
480 *
481 *
482 *
483 * @section sec_pgmPhys_MappingCaches Mapping Caches
484 *
485 * In order to be able to map in and out memory and to be able to support
486 * guest with more RAM than we've got virtual address space, we'll employing
487 * a mapping cache. There is already a tiny one for GC (see PGMGCDynMapGCPageEx)
488 * and we'll create a similar one for ring-0 unless we decide to setup a dedicate
489 * memory context for the HWACCM execution.
490 *
491 *
492 * @subsection subsec_pgmPhys_MappingCaches_R3 Ring-3
493 *
494 * We've considered implementing the ring-3 mapping cache page based but found
495 * that this was bother some when one had to take into account TLBs+SMP and
496 * portability (missing the necessary APIs on several platforms). There were
497 * also some performance concerns with this approach which hadn't quite been
498 * worked out.
499 *
500 * Instead, we'll be mapping allocation chunks into the VM process. This simplifies
501 * matters greatly quite a bit since we don't need to invent any new ring-0 stuff,
502 * only some minor RTR0MEMOBJ mapping stuff. The main concern here is that mapping
503 * compared to the previous idea is that mapping or unmapping a 1MB chunk is more
504 * costly than a single page, although how much more costly is uncertain. We'll
505 * try address this by using a very big cache, preferably bigger than the actual
506 * VM RAM size if possible. The current VM RAM sizes should give some idea for
507 * 32-bit boxes, while on 64-bit we can probably get away with employing an
508 * unlimited cache.
509 *
510 * The cache have to parts, as already indicated, the ring-3 side and the
511 * ring-0 side.
512 *
513 * The ring-0 will be tied to the page allocator since it will operate on the
514 * memory objects it contains. It will therefore require the first ring-0 mutex
515 * discussed in @ref subsec_pgmPhys_Serializing. We
516 * some double house keeping wrt to who has mapped what I think, since both
517 * VMMR0.r0 and RTR0MemObj will keep track of mapping relataions
518 *
519 * The ring-3 part will be protected by the pgm critsect. For simplicity, we'll
520 * require anyone that desires to do changes to the mapping cache to do that
521 * from within this critsect. Alternatively, we could employ a separate critsect
522 * for serializing changes to the mapping cache as this would reduce potential
523 * contention with other threads accessing mappings unrelated to the changes
524 * that are in process. We can see about this later, contention will show
525 * up in the statistics anyway, so it'll be simple to tell.
526 *
527 * The organization of the ring-3 part will be very much like how the allocation
528 * chunks are organized in ring-0, that is in an AVL tree by chunk id. To avoid
529 * having to walk the tree all the time, we'll have a couple of lookaside entries
530 * like in we do for I/O ports and MMIO in IOM.
531 *
532 * The simplified flow of a PGMPhysRead/Write function:
533 * -# Enter the PGM critsect.
534 * -# Lookup GCPhys in the ram ranges and get the Page ID.
535 * -# Calc the Allocation Chunk ID from the Page ID.
536 * -# Check the lookaside entries and then the AVL tree for the Chunk ID.
537 * If not found in cache:
538 * -# Call ring-0 and request it to be mapped and supply
539 * a chunk to be unmapped if the cache is maxed out already.
540 * -# Insert the new mapping into the AVL tree (id + R3 address).
541 * -# Update the relevant lookaside entry and return the mapping address.
542 * -# Do the read/write according to monitoring flags and everything.
543 * -# Leave the critsect.
544 *
545 *
546 * @section sec_pgmPhys_Fallback Fallback
547 *
548 * Current all the "second tier" hosts will not support the RTR0MemObjAllocPhysNC
549 * API and thus require a fallback.
550 *
551 * So, when RTR0MemObjAllocPhysNC returns VERR_NOT_SUPPORTED the page allocator
552 * will return to the ring-3 caller (and later ring-0) and asking it to seed
553 * the page allocator with some fresh pages (VERR_GMM_SEED_ME). Ring-3 will
554 * then perform an SUPPageAlloc(cbChunk >> PAGE_SHIFT) call and make a
555 * "SeededAllocPages" call to ring-0.
556 *
557 * The first time ring-0 sees the VERR_NOT_SUPPORTED failure it will disable
558 * all page sharing (zero page detection will continue). It will also force
559 * all allocations to come from the VM which seeded the page. Both these
560 * measures are taken to make sure that there will never be any need for
561 * mapping anything into ring-3 - everything will be mapped already.
562 *
563 * Whether we'll continue to use the current MM locked memory management
564 * for this I don't quite know (I'd prefer not to and just ditch that all
565 * togther), we'll see what's simplest to do.
566 *
567 *
568 *
569 * @section sec_pgmPhys_Changes Changes
570 *
571 * Breakdown of the changes involved?
572 */
573
574/*******************************************************************************
575* Header Files *
576*******************************************************************************/
577#define LOG_GROUP LOG_GROUP_PGM
578#include <VBox/dbgf.h>
579#include <VBox/pgm.h>
580#include <VBox/cpum.h>
581#include <VBox/iom.h>
582#include <VBox/sup.h>
583#include <VBox/mm.h>
584#include <VBox/em.h>
585#include <VBox/stam.h>
586#include <VBox/rem.h>
587#include <VBox/dbgf.h>
588#include <VBox/rem.h>
589#include <VBox/selm.h>
590#include <VBox/ssm.h>
591#include "PGMInternal.h"
592#include <VBox/vm.h>
593#include <VBox/dbg.h>
594#include <VBox/hwaccm.h>
595
596#include <iprt/assert.h>
597#include <iprt/alloc.h>
598#include <iprt/asm.h>
599#include <iprt/thread.h>
600#include <iprt/string.h>
601#ifdef DEBUG_bird
602# include <iprt/env.h>
603#endif
604#include <VBox/param.h>
605#include <VBox/err.h>
606
607
608/*******************************************************************************
609* Defined Constants And Macros *
610*******************************************************************************/
611/** Saved state data unit version for 2.5.x and later. */
612#define PGM_SAVED_STATE_VERSION 9
613/** Saved state data unit version for 2.2.2 and later. */
614#define PGM_SAVED_STATE_VERSION_2_2_2 8
615/** Saved state data unit version for 2.2.0. */
616#define PGM_SAVED_STATE_VERSION_RR_DESC 7
617/** Saved state data unit version. */
618#define PGM_SAVED_STATE_VERSION_OLD_PHYS_CODE 6
619
620
621/*******************************************************************************
622* Internal Functions *
623*******************************************************************************/
624static int pgmR3InitPaging(PVM pVM);
625static void pgmR3InitStats(PVM pVM);
626static DECLCALLBACK(void) pgmR3PhysInfo(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
627static DECLCALLBACK(void) pgmR3InfoMode(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
628static DECLCALLBACK(void) pgmR3InfoCr3(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
629static DECLCALLBACK(int) pgmR3RelocatePhysHandler(PAVLROGCPHYSNODECORE pNode, void *pvUser);
630static DECLCALLBACK(int) pgmR3RelocateVirtHandler(PAVLROGCPTRNODECORE pNode, void *pvUser);
631static DECLCALLBACK(int) pgmR3RelocateHyperVirtHandler(PAVLROGCPTRNODECORE pNode, void *pvUser);
632#ifdef VBOX_STRICT
633static DECLCALLBACK(void) pgmR3ResetNoMorePhysWritesFlag(PVM pVM, VMSTATE enmState, VMSTATE enmOldState, void *pvUser);
634#endif
635static DECLCALLBACK(int) pgmR3Save(PVM pVM, PSSMHANDLE pSSM);
636static DECLCALLBACK(int) pgmR3Load(PVM pVM, PSSMHANDLE pSSM, uint32_t u32Version);
637static int pgmR3ModeDataInit(PVM pVM, bool fResolveGCAndR0);
638static void pgmR3ModeDataSwitch(PVM pVM, PVMCPU pVCpu, PGMMODE enmShw, PGMMODE enmGst);
639static PGMMODE pgmR3CalcShadowMode(PVM pVM, PGMMODE enmGuestMode, SUPPAGINGMODE enmHostMode, PGMMODE enmShadowMode, VMMSWITCHER *penmSwitcher);
640
641#ifdef VBOX_WITH_DEBUGGER
642/** @todo Convert the first two commands to 'info' items. */
643static DECLCALLBACK(int) pgmR3CmdRam(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
644static DECLCALLBACK(int) pgmR3CmdMap(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
645static DECLCALLBACK(int) pgmR3CmdError(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
646static DECLCALLBACK(int) pgmR3CmdSync(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
647static DECLCALLBACK(int) pgmR3CmdSyncAlways(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
648# ifdef VBOX_STRICT
649static DECLCALLBACK(int) pgmR3CmdAssertCR3(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
650# endif
651#endif
652
653
654/*******************************************************************************
655* Global Variables *
656*******************************************************************************/
657#ifdef VBOX_WITH_DEBUGGER
658/** Argument descriptors for '.pgmerror' and '.pgmerroroff'. */
659static const DBGCVARDESC g_aPgmErrorArgs[] =
660{
661 /* cTimesMin, cTimesMax, enmCategory, fFlags, pszName, pszDescription */
662 { 0, 1, DBGCVAR_CAT_STRING, 0, "where", "Error injection location." },
663};
664
665/** Command descriptors. */
666static const DBGCCMD g_aCmds[] =
667{
668 /* pszCmd, cArgsMin, cArgsMax, paArgDesc, cArgDescs, pResultDesc, fFlags, pfnHandler pszSyntax, ....pszDescription */
669 { "pgmram", 0, 0, NULL, 0, NULL, 0, pgmR3CmdRam, "", "Display the ram ranges." },
670 { "pgmmap", 0, 0, NULL, 0, NULL, 0, pgmR3CmdMap, "", "Display the mapping ranges." },
671 { "pgmsync", 0, 0, NULL, 0, NULL, 0, pgmR3CmdSync, "", "Sync the CR3 page." },
672 { "pgmerror", 0, 1, &g_aPgmErrorArgs[0],1, NULL, 0, pgmR3CmdError, "", "Enables inject runtime of errors into parts of PGM." },
673 { "pgmerroroff", 0, 1, &g_aPgmErrorArgs[0],1, NULL, 0, pgmR3CmdError, "", "Disables inject runtime errors into parts of PGM." },
674#ifdef VBOX_STRICT
675 { "pgmassertcr3", 0, 0, NULL, 0, NULL, 0, pgmR3CmdAssertCR3, "", "Check the shadow CR3 mapping." },
676#endif
677 { "pgmsyncalways", 0, 0, NULL, 0, NULL, 0, pgmR3CmdSyncAlways, "", "Toggle permanent CR3 syncing." },
678};
679#endif
680
681
682
683
684/*
685 * Shadow - 32-bit mode
686 */
687#define PGM_SHW_TYPE PGM_TYPE_32BIT
688#define PGM_SHW_NAME(name) PGM_SHW_NAME_32BIT(name)
689#define PGM_SHW_NAME_RC_STR(name) PGM_SHW_NAME_RC_32BIT_STR(name)
690#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_32BIT_STR(name)
691#include "PGMShw.h"
692
693/* Guest - real mode */
694#define PGM_GST_TYPE PGM_TYPE_REAL
695#define PGM_GST_NAME(name) PGM_GST_NAME_REAL(name)
696#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_REAL_STR(name)
697#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_REAL_STR(name)
698#define PGM_BTH_NAME(name) PGM_BTH_NAME_32BIT_REAL(name)
699#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_32BIT_REAL_STR(name)
700#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_32BIT_REAL_STR(name)
701#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_32BIT_PT_FOR_PHYS
702#define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_32BIT_PD_PHYS
703#include "PGMBth.h"
704#include "PGMGstDefs.h"
705#include "PGMGst.h"
706#undef BTH_PGMPOOLKIND_PT_FOR_PT
707#undef BTH_PGMPOOLKIND_ROOT
708#undef PGM_BTH_NAME
709#undef PGM_BTH_NAME_RC_STR
710#undef PGM_BTH_NAME_R0_STR
711#undef PGM_GST_TYPE
712#undef PGM_GST_NAME
713#undef PGM_GST_NAME_RC_STR
714#undef PGM_GST_NAME_R0_STR
715
716/* Guest - protected mode */
717#define PGM_GST_TYPE PGM_TYPE_PROT
718#define PGM_GST_NAME(name) PGM_GST_NAME_PROT(name)
719#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_PROT_STR(name)
720#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PROT_STR(name)
721#define PGM_BTH_NAME(name) PGM_BTH_NAME_32BIT_PROT(name)
722#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_32BIT_PROT_STR(name)
723#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_32BIT_PROT_STR(name)
724#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_32BIT_PT_FOR_PHYS
725#define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_32BIT_PD_PHYS
726#include "PGMBth.h"
727#include "PGMGstDefs.h"
728#include "PGMGst.h"
729#undef BTH_PGMPOOLKIND_PT_FOR_PT
730#undef BTH_PGMPOOLKIND_ROOT
731#undef PGM_BTH_NAME
732#undef PGM_BTH_NAME_RC_STR
733#undef PGM_BTH_NAME_R0_STR
734#undef PGM_GST_TYPE
735#undef PGM_GST_NAME
736#undef PGM_GST_NAME_RC_STR
737#undef PGM_GST_NAME_R0_STR
738
739/* Guest - 32-bit mode */
740#define PGM_GST_TYPE PGM_TYPE_32BIT
741#define PGM_GST_NAME(name) PGM_GST_NAME_32BIT(name)
742#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_32BIT_STR(name)
743#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_32BIT_STR(name)
744#define PGM_BTH_NAME(name) PGM_BTH_NAME_32BIT_32BIT(name)
745#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_32BIT_32BIT_STR(name)
746#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_32BIT_32BIT_STR(name)
747#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_32BIT_PT_FOR_32BIT_PT
748#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_32BIT_PT_FOR_32BIT_4MB
749#define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_32BIT_PD
750#include "PGMBth.h"
751#include "PGMGstDefs.h"
752#include "PGMGst.h"
753#undef BTH_PGMPOOLKIND_PT_FOR_BIG
754#undef BTH_PGMPOOLKIND_PT_FOR_PT
755#undef BTH_PGMPOOLKIND_ROOT
756#undef PGM_BTH_NAME
757#undef PGM_BTH_NAME_RC_STR
758#undef PGM_BTH_NAME_R0_STR
759#undef PGM_GST_TYPE
760#undef PGM_GST_NAME
761#undef PGM_GST_NAME_RC_STR
762#undef PGM_GST_NAME_R0_STR
763
764#undef PGM_SHW_TYPE
765#undef PGM_SHW_NAME
766#undef PGM_SHW_NAME_RC_STR
767#undef PGM_SHW_NAME_R0_STR
768
769
770/*
771 * Shadow - PAE mode
772 */
773#define PGM_SHW_TYPE PGM_TYPE_PAE
774#define PGM_SHW_NAME(name) PGM_SHW_NAME_PAE(name)
775#define PGM_SHW_NAME_RC_STR(name) PGM_SHW_NAME_RC_PAE_STR(name)
776#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_PAE_STR(name)
777#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_REAL(name)
778#include "PGMShw.h"
779
780/* Guest - real mode */
781#define PGM_GST_TYPE PGM_TYPE_REAL
782#define PGM_GST_NAME(name) PGM_GST_NAME_REAL(name)
783#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_REAL_STR(name)
784#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_REAL_STR(name)
785#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_REAL(name)
786#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_PAE_REAL_STR(name)
787#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_REAL_STR(name)
788#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
789#define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_PAE_PDPT_PHYS
790#include "PGMGstDefs.h"
791#include "PGMBth.h"
792#undef BTH_PGMPOOLKIND_PT_FOR_PT
793#undef BTH_PGMPOOLKIND_ROOT
794#undef PGM_BTH_NAME
795#undef PGM_BTH_NAME_RC_STR
796#undef PGM_BTH_NAME_R0_STR
797#undef PGM_GST_TYPE
798#undef PGM_GST_NAME
799#undef PGM_GST_NAME_RC_STR
800#undef PGM_GST_NAME_R0_STR
801
802/* Guest - protected mode */
803#define PGM_GST_TYPE PGM_TYPE_PROT
804#define PGM_GST_NAME(name) PGM_GST_NAME_PROT(name)
805#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_PROT_STR(name)
806#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PROT_STR(name)
807#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_PROT(name)
808#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_PAE_PROT_STR(name)
809#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_PROT_STR(name)
810#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
811#define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_PAE_PDPT_PHYS
812#include "PGMGstDefs.h"
813#include "PGMBth.h"
814#undef BTH_PGMPOOLKIND_PT_FOR_PT
815#undef BTH_PGMPOOLKIND_ROOT
816#undef PGM_BTH_NAME
817#undef PGM_BTH_NAME_RC_STR
818#undef PGM_BTH_NAME_R0_STR
819#undef PGM_GST_TYPE
820#undef PGM_GST_NAME
821#undef PGM_GST_NAME_RC_STR
822#undef PGM_GST_NAME_R0_STR
823
824/* Guest - 32-bit mode */
825#define PGM_GST_TYPE PGM_TYPE_32BIT
826#define PGM_GST_NAME(name) PGM_GST_NAME_32BIT(name)
827#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_32BIT_STR(name)
828#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_32BIT_STR(name)
829#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_32BIT(name)
830#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_PAE_32BIT_STR(name)
831#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_32BIT_STR(name)
832#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_32BIT_PT
833#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_32BIT_4MB
834#define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_PAE_PDPT_FOR_32BIT
835#include "PGMGstDefs.h"
836#include "PGMBth.h"
837#undef BTH_PGMPOOLKIND_PT_FOR_BIG
838#undef BTH_PGMPOOLKIND_PT_FOR_PT
839#undef BTH_PGMPOOLKIND_ROOT
840#undef PGM_BTH_NAME
841#undef PGM_BTH_NAME_RC_STR
842#undef PGM_BTH_NAME_R0_STR
843#undef PGM_GST_TYPE
844#undef PGM_GST_NAME
845#undef PGM_GST_NAME_RC_STR
846#undef PGM_GST_NAME_R0_STR
847
848/* Guest - PAE mode */
849#define PGM_GST_TYPE PGM_TYPE_PAE
850#define PGM_GST_NAME(name) PGM_GST_NAME_PAE(name)
851#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_PAE_STR(name)
852#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PAE_STR(name)
853#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_PAE(name)
854#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_PAE_PAE_STR(name)
855#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_PAE_STR(name)
856#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
857#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
858#define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_PAE_PDPT
859#include "PGMBth.h"
860#include "PGMGstDefs.h"
861#include "PGMGst.h"
862#undef BTH_PGMPOOLKIND_PT_FOR_BIG
863#undef BTH_PGMPOOLKIND_PT_FOR_PT
864#undef BTH_PGMPOOLKIND_ROOT
865#undef PGM_BTH_NAME
866#undef PGM_BTH_NAME_RC_STR
867#undef PGM_BTH_NAME_R0_STR
868#undef PGM_GST_TYPE
869#undef PGM_GST_NAME
870#undef PGM_GST_NAME_RC_STR
871#undef PGM_GST_NAME_R0_STR
872
873#undef PGM_SHW_TYPE
874#undef PGM_SHW_NAME
875#undef PGM_SHW_NAME_RC_STR
876#undef PGM_SHW_NAME_R0_STR
877
878
879/*
880 * Shadow - AMD64 mode
881 */
882#define PGM_SHW_TYPE PGM_TYPE_AMD64
883#define PGM_SHW_NAME(name) PGM_SHW_NAME_AMD64(name)
884#define PGM_SHW_NAME_RC_STR(name) PGM_SHW_NAME_RC_AMD64_STR(name)
885#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_AMD64_STR(name)
886#include "PGMShw.h"
887
888#ifdef VBOX_WITH_64_BITS_GUESTS
889/* Guest - AMD64 mode */
890# define PGM_GST_TYPE PGM_TYPE_AMD64
891# define PGM_GST_NAME(name) PGM_GST_NAME_AMD64(name)
892# define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_AMD64_STR(name)
893# define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_AMD64_STR(name)
894# define PGM_BTH_NAME(name) PGM_BTH_NAME_AMD64_AMD64(name)
895# define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_AMD64_AMD64_STR(name)
896# define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_AMD64_AMD64_STR(name)
897# define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
898# define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
899# define BTH_PGMPOOLKIND_ROOT PGMPOOLKIND_64BIT_PML4
900# include "PGMBth.h"
901# include "PGMGstDefs.h"
902# include "PGMGst.h"
903# undef BTH_PGMPOOLKIND_PT_FOR_BIG
904# undef BTH_PGMPOOLKIND_PT_FOR_PT
905# undef BTH_PGMPOOLKIND_ROOT
906# undef PGM_BTH_NAME
907# undef PGM_BTH_NAME_RC_STR
908# undef PGM_BTH_NAME_R0_STR
909# undef PGM_GST_TYPE
910# undef PGM_GST_NAME
911# undef PGM_GST_NAME_RC_STR
912# undef PGM_GST_NAME_R0_STR
913#endif /* VBOX_WITH_64_BITS_GUESTS */
914
915#undef PGM_SHW_TYPE
916#undef PGM_SHW_NAME
917#undef PGM_SHW_NAME_RC_STR
918#undef PGM_SHW_NAME_R0_STR
919
920
921/*
922 * Shadow - Nested paging mode
923 */
924#define PGM_SHW_TYPE PGM_TYPE_NESTED
925#define PGM_SHW_NAME(name) PGM_SHW_NAME_NESTED(name)
926#define PGM_SHW_NAME_RC_STR(name) PGM_SHW_NAME_RC_NESTED_STR(name)
927#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_NESTED_STR(name)
928#include "PGMShw.h"
929
930/* Guest - real mode */
931#define PGM_GST_TYPE PGM_TYPE_REAL
932#define PGM_GST_NAME(name) PGM_GST_NAME_REAL(name)
933#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_REAL_STR(name)
934#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_REAL_STR(name)
935#define PGM_BTH_NAME(name) PGM_BTH_NAME_NESTED_REAL(name)
936#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_NESTED_REAL_STR(name)
937#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_NESTED_REAL_STR(name)
938#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
939#include "PGMGstDefs.h"
940#include "PGMBth.h"
941#undef BTH_PGMPOOLKIND_PT_FOR_PT
942#undef PGM_BTH_NAME
943#undef PGM_BTH_NAME_RC_STR
944#undef PGM_BTH_NAME_R0_STR
945#undef PGM_GST_TYPE
946#undef PGM_GST_NAME
947#undef PGM_GST_NAME_RC_STR
948#undef PGM_GST_NAME_R0_STR
949
950/* Guest - protected mode */
951#define PGM_GST_TYPE PGM_TYPE_PROT
952#define PGM_GST_NAME(name) PGM_GST_NAME_PROT(name)
953#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_PROT_STR(name)
954#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PROT_STR(name)
955#define PGM_BTH_NAME(name) PGM_BTH_NAME_NESTED_PROT(name)
956#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_NESTED_PROT_STR(name)
957#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_NESTED_PROT_STR(name)
958#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
959#include "PGMGstDefs.h"
960#include "PGMBth.h"
961#undef BTH_PGMPOOLKIND_PT_FOR_PT
962#undef PGM_BTH_NAME
963#undef PGM_BTH_NAME_RC_STR
964#undef PGM_BTH_NAME_R0_STR
965#undef PGM_GST_TYPE
966#undef PGM_GST_NAME
967#undef PGM_GST_NAME_RC_STR
968#undef PGM_GST_NAME_R0_STR
969
970/* Guest - 32-bit mode */
971#define PGM_GST_TYPE PGM_TYPE_32BIT
972#define PGM_GST_NAME(name) PGM_GST_NAME_32BIT(name)
973#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_32BIT_STR(name)
974#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_32BIT_STR(name)
975#define PGM_BTH_NAME(name) PGM_BTH_NAME_NESTED_32BIT(name)
976#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_NESTED_32BIT_STR(name)
977#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_NESTED_32BIT_STR(name)
978#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_32BIT_PT
979#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_32BIT_4MB
980#include "PGMGstDefs.h"
981#include "PGMBth.h"
982#undef BTH_PGMPOOLKIND_PT_FOR_BIG
983#undef BTH_PGMPOOLKIND_PT_FOR_PT
984#undef PGM_BTH_NAME
985#undef PGM_BTH_NAME_RC_STR
986#undef PGM_BTH_NAME_R0_STR
987#undef PGM_GST_TYPE
988#undef PGM_GST_NAME
989#undef PGM_GST_NAME_RC_STR
990#undef PGM_GST_NAME_R0_STR
991
992/* Guest - PAE mode */
993#define PGM_GST_TYPE PGM_TYPE_PAE
994#define PGM_GST_NAME(name) PGM_GST_NAME_PAE(name)
995#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_PAE_STR(name)
996#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PAE_STR(name)
997#define PGM_BTH_NAME(name) PGM_BTH_NAME_NESTED_PAE(name)
998#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_NESTED_PAE_STR(name)
999#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_NESTED_PAE_STR(name)
1000#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
1001#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
1002#include "PGMGstDefs.h"
1003#include "PGMBth.h"
1004#undef BTH_PGMPOOLKIND_PT_FOR_BIG
1005#undef BTH_PGMPOOLKIND_PT_FOR_PT
1006#undef PGM_BTH_NAME
1007#undef PGM_BTH_NAME_RC_STR
1008#undef PGM_BTH_NAME_R0_STR
1009#undef PGM_GST_TYPE
1010#undef PGM_GST_NAME
1011#undef PGM_GST_NAME_RC_STR
1012#undef PGM_GST_NAME_R0_STR
1013
1014#ifdef VBOX_WITH_64_BITS_GUESTS
1015/* Guest - AMD64 mode */
1016# define PGM_GST_TYPE PGM_TYPE_AMD64
1017# define PGM_GST_NAME(name) PGM_GST_NAME_AMD64(name)
1018# define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_AMD64_STR(name)
1019# define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_AMD64_STR(name)
1020# define PGM_BTH_NAME(name) PGM_BTH_NAME_NESTED_AMD64(name)
1021# define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_NESTED_AMD64_STR(name)
1022# define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_NESTED_AMD64_STR(name)
1023# define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
1024# define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
1025# include "PGMGstDefs.h"
1026# include "PGMBth.h"
1027# undef BTH_PGMPOOLKIND_PT_FOR_BIG
1028# undef BTH_PGMPOOLKIND_PT_FOR_PT
1029# undef PGM_BTH_NAME
1030# undef PGM_BTH_NAME_RC_STR
1031# undef PGM_BTH_NAME_R0_STR
1032# undef PGM_GST_TYPE
1033# undef PGM_GST_NAME
1034# undef PGM_GST_NAME_RC_STR
1035# undef PGM_GST_NAME_R0_STR
1036#endif /* VBOX_WITH_64_BITS_GUESTS */
1037
1038#undef PGM_SHW_TYPE
1039#undef PGM_SHW_NAME
1040#undef PGM_SHW_NAME_RC_STR
1041#undef PGM_SHW_NAME_R0_STR
1042
1043
1044/*
1045 * Shadow - EPT
1046 */
1047#define PGM_SHW_TYPE PGM_TYPE_EPT
1048#define PGM_SHW_NAME(name) PGM_SHW_NAME_EPT(name)
1049#define PGM_SHW_NAME_RC_STR(name) PGM_SHW_NAME_RC_EPT_STR(name)
1050#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_EPT_STR(name)
1051#include "PGMShw.h"
1052
1053/* Guest - real mode */
1054#define PGM_GST_TYPE PGM_TYPE_REAL
1055#define PGM_GST_NAME(name) PGM_GST_NAME_REAL(name)
1056#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_REAL_STR(name)
1057#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_REAL_STR(name)
1058#define PGM_BTH_NAME(name) PGM_BTH_NAME_EPT_REAL(name)
1059#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_EPT_REAL_STR(name)
1060#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_EPT_REAL_STR(name)
1061#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
1062#include "PGMGstDefs.h"
1063#include "PGMBth.h"
1064#undef BTH_PGMPOOLKIND_PT_FOR_PT
1065#undef PGM_BTH_NAME
1066#undef PGM_BTH_NAME_RC_STR
1067#undef PGM_BTH_NAME_R0_STR
1068#undef PGM_GST_TYPE
1069#undef PGM_GST_NAME
1070#undef PGM_GST_NAME_RC_STR
1071#undef PGM_GST_NAME_R0_STR
1072
1073/* Guest - protected mode */
1074#define PGM_GST_TYPE PGM_TYPE_PROT
1075#define PGM_GST_NAME(name) PGM_GST_NAME_PROT(name)
1076#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_PROT_STR(name)
1077#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PROT_STR(name)
1078#define PGM_BTH_NAME(name) PGM_BTH_NAME_EPT_PROT(name)
1079#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_EPT_PROT_STR(name)
1080#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_EPT_PROT_STR(name)
1081#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
1082#include "PGMGstDefs.h"
1083#include "PGMBth.h"
1084#undef BTH_PGMPOOLKIND_PT_FOR_PT
1085#undef PGM_BTH_NAME
1086#undef PGM_BTH_NAME_RC_STR
1087#undef PGM_BTH_NAME_R0_STR
1088#undef PGM_GST_TYPE
1089#undef PGM_GST_NAME
1090#undef PGM_GST_NAME_RC_STR
1091#undef PGM_GST_NAME_R0_STR
1092
1093/* Guest - 32-bit mode */
1094#define PGM_GST_TYPE PGM_TYPE_32BIT
1095#define PGM_GST_NAME(name) PGM_GST_NAME_32BIT(name)
1096#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_32BIT_STR(name)
1097#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_32BIT_STR(name)
1098#define PGM_BTH_NAME(name) PGM_BTH_NAME_EPT_32BIT(name)
1099#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_EPT_32BIT_STR(name)
1100#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_EPT_32BIT_STR(name)
1101#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_32BIT_PT
1102#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_32BIT_4MB
1103#include "PGMGstDefs.h"
1104#include "PGMBth.h"
1105#undef BTH_PGMPOOLKIND_PT_FOR_BIG
1106#undef BTH_PGMPOOLKIND_PT_FOR_PT
1107#undef PGM_BTH_NAME
1108#undef PGM_BTH_NAME_RC_STR
1109#undef PGM_BTH_NAME_R0_STR
1110#undef PGM_GST_TYPE
1111#undef PGM_GST_NAME
1112#undef PGM_GST_NAME_RC_STR
1113#undef PGM_GST_NAME_R0_STR
1114
1115/* Guest - PAE mode */
1116#define PGM_GST_TYPE PGM_TYPE_PAE
1117#define PGM_GST_NAME(name) PGM_GST_NAME_PAE(name)
1118#define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_PAE_STR(name)
1119#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PAE_STR(name)
1120#define PGM_BTH_NAME(name) PGM_BTH_NAME_EPT_PAE(name)
1121#define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_EPT_PAE_STR(name)
1122#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_EPT_PAE_STR(name)
1123#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
1124#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
1125#include "PGMGstDefs.h"
1126#include "PGMBth.h"
1127#undef BTH_PGMPOOLKIND_PT_FOR_BIG
1128#undef BTH_PGMPOOLKIND_PT_FOR_PT
1129#undef PGM_BTH_NAME
1130#undef PGM_BTH_NAME_RC_STR
1131#undef PGM_BTH_NAME_R0_STR
1132#undef PGM_GST_TYPE
1133#undef PGM_GST_NAME
1134#undef PGM_GST_NAME_RC_STR
1135#undef PGM_GST_NAME_R0_STR
1136
1137#ifdef VBOX_WITH_64_BITS_GUESTS
1138/* Guest - AMD64 mode */
1139# define PGM_GST_TYPE PGM_TYPE_AMD64
1140# define PGM_GST_NAME(name) PGM_GST_NAME_AMD64(name)
1141# define PGM_GST_NAME_RC_STR(name) PGM_GST_NAME_RC_AMD64_STR(name)
1142# define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_AMD64_STR(name)
1143# define PGM_BTH_NAME(name) PGM_BTH_NAME_EPT_AMD64(name)
1144# define PGM_BTH_NAME_RC_STR(name) PGM_BTH_NAME_RC_EPT_AMD64_STR(name)
1145# define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_EPT_AMD64_STR(name)
1146# define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
1147# define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
1148# include "PGMGstDefs.h"
1149# include "PGMBth.h"
1150# undef BTH_PGMPOOLKIND_PT_FOR_BIG
1151# undef BTH_PGMPOOLKIND_PT_FOR_PT
1152# undef PGM_BTH_NAME
1153# undef PGM_BTH_NAME_RC_STR
1154# undef PGM_BTH_NAME_R0_STR
1155# undef PGM_GST_TYPE
1156# undef PGM_GST_NAME
1157# undef PGM_GST_NAME_RC_STR
1158# undef PGM_GST_NAME_R0_STR
1159#endif /* VBOX_WITH_64_BITS_GUESTS */
1160
1161#undef PGM_SHW_TYPE
1162#undef PGM_SHW_NAME
1163#undef PGM_SHW_NAME_RC_STR
1164#undef PGM_SHW_NAME_R0_STR
1165
1166
1167
1168/**
1169 * Initiates the paging of VM.
1170 *
1171 * @returns VBox status code.
1172 * @param pVM Pointer to VM structure.
1173 */
1174VMMR3DECL(int) PGMR3Init(PVM pVM)
1175{
1176 LogFlow(("PGMR3Init:\n"));
1177 PCFGMNODE pCfgPGM = CFGMR3GetChild(CFGMR3GetRoot(pVM), "/PGM");
1178 int rc;
1179
1180 /*
1181 * Assert alignment and sizes.
1182 */
1183 AssertCompile(sizeof(pVM->pgm.s) <= sizeof(pVM->pgm.padding));
1184 AssertCompileMemberAlignment(PGM, CritSect, sizeof(uintptr_t));
1185
1186 /*
1187 * Init the structure.
1188 */
1189 pVM->pgm.s.offVM = RT_OFFSETOF(VM, pgm.s);
1190 pVM->pgm.s.offVCpuPGM = RT_OFFSETOF(VMCPU, pgm.s);
1191
1192 /* Init the per-CPU part. */
1193 for (unsigned i=0;i<pVM->cCPUs;i++)
1194 {
1195 PVMCPU pVCpu = &pVM->aCpus[i];
1196 PPGMCPU pPGM = &pVCpu->pgm.s;
1197
1198 pPGM->offVM = (uintptr_t)&pVCpu->pgm.s - (uintptr_t)pVM;
1199 pPGM->offVCpu = RT_OFFSETOF(VMCPU, pgm.s);
1200 pPGM->offPGM = (uintptr_t)&pVCpu->pgm.s - (uintptr_t)&pVM->pgm.s;
1201
1202 pPGM->enmShadowMode = PGMMODE_INVALID;
1203 pPGM->enmGuestMode = PGMMODE_INVALID;
1204
1205 pPGM->GCPhysCR3 = NIL_RTGCPHYS;
1206
1207 pPGM->pGstPaePdptR3 = NULL;
1208#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
1209 pPGM->pGstPaePdptR0 = NIL_RTR0PTR;
1210#endif
1211 pPGM->pGstPaePdptRC = NIL_RTRCPTR;
1212 for (unsigned i = 0; i < RT_ELEMENTS(pVCpu->pgm.s.apGstPaePDsR3); i++)
1213 {
1214 pPGM->apGstPaePDsR3[i] = NULL;
1215#ifndef VBOX_WITH_2X_4GB_ADDR_SPACE
1216 pPGM->apGstPaePDsR0[i] = NIL_RTR0PTR;
1217#endif
1218 pPGM->apGstPaePDsRC[i] = NIL_RTRCPTR;
1219 pPGM->aGCPhysGstPaePDs[i] = NIL_RTGCPHYS;
1220 pPGM->aGCPhysGstPaePDsMonitored[i] = NIL_RTGCPHYS;
1221 }
1222
1223 pPGM->fA20Enabled = true;
1224 }
1225
1226 pVM->pgm.s.enmHostMode = SUPPAGINGMODE_INVALID;
1227 pVM->pgm.s.GCPhys4MBPSEMask = RT_BIT_64(32) - 1; /* default; checked later */
1228 pVM->pgm.s.GCPtrPrevRamRangeMapping = MM_HYPER_AREA_ADDRESS;
1229
1230 rc = CFGMR3QueryBoolDef(CFGMR3GetRoot(pVM), "RamPreAlloc", &pVM->pgm.s.fRamPreAlloc,
1231#ifdef VBOX_WITH_PREALLOC_RAM_BY_DEFAULT
1232 true
1233#else
1234 false
1235#endif
1236 );
1237 AssertLogRelRCReturn(rc, rc);
1238
1239#if HC_ARCH_BITS == 64 || 1 /** @todo 4GB/32-bit: remove || 1 later and adjust the limit. */
1240 rc = CFGMR3QueryU32Def(pCfgPGM, "MaxRing3Chunks", &pVM->pgm.s.ChunkR3Map.cMax, UINT32_MAX);
1241#else
1242 rc = CFGMR3QueryU32Def(pCfgPGM, "MaxRing3Chunks", &pVM->pgm.s.ChunkR3Map.cMax, _1G / GMM_CHUNK_SIZE);
1243#endif
1244 AssertLogRelRCReturn(rc, rc);
1245 for (uint32_t i = 0; i < RT_ELEMENTS(pVM->pgm.s.ChunkR3Map.Tlb.aEntries); i++)
1246 pVM->pgm.s.ChunkR3Map.Tlb.aEntries[i].idChunk = NIL_GMM_CHUNKID;
1247
1248 /*
1249 * Get the configured RAM size - to estimate saved state size.
1250 */
1251 uint64_t cbRam;
1252 rc = CFGMR3QueryU64(CFGMR3GetRoot(pVM), "RamSize", &cbRam);
1253 if (rc == VERR_CFGM_VALUE_NOT_FOUND)
1254 cbRam = 0;
1255 else if (RT_SUCCESS(rc))
1256 {
1257 if (cbRam < PAGE_SIZE)
1258 cbRam = 0;
1259 cbRam = RT_ALIGN_64(cbRam, PAGE_SIZE);
1260 }
1261 else
1262 {
1263 AssertMsgFailed(("Configuration error: Failed to query integer \"RamSize\", rc=%Rrc.\n", rc));
1264 return rc;
1265 }
1266
1267 /*
1268 * Register callbacks, string formatters and the saved state data unit.
1269 */
1270#ifdef VBOX_STRICT
1271 VMR3AtStateRegister(pVM, pgmR3ResetNoMorePhysWritesFlag, NULL);
1272#endif
1273 PGMRegisterStringFormatTypes();
1274
1275 rc = SSMR3RegisterInternal(pVM, "pgm", 1, PGM_SAVED_STATE_VERSION, (size_t)cbRam + sizeof(PGM),
1276 NULL, pgmR3Save, NULL,
1277 NULL, pgmR3Load, NULL);
1278 if (RT_FAILURE(rc))
1279 return rc;
1280
1281 /*
1282 * Initialize the PGM critical section and flush the phys TLBs
1283 */
1284 rc = PDMR3CritSectInit(pVM, &pVM->pgm.s.CritSect, "PGM");
1285 AssertRCReturn(rc, rc);
1286
1287 PGMR3PhysChunkInvalidateTLB(pVM);
1288 PGMPhysInvalidatePageR3MapTLB(pVM);
1289 PGMPhysInvalidatePageR0MapTLB(pVM);
1290 PGMPhysInvalidatePageGCMapTLB(pVM);
1291
1292 /*
1293 * For the time being we sport a full set of handy pages in addition to the base
1294 * memory to simplify things.
1295 */
1296 rc = MMR3ReserveHandyPages(pVM, RT_ELEMENTS(pVM->pgm.s.aHandyPages)); /** @todo this should be changed to PGM_HANDY_PAGES_MIN but this needs proper testing... */
1297 AssertRCReturn(rc, rc);
1298
1299 /*
1300 * Trees
1301 */
1302 rc = MMHyperAlloc(pVM, sizeof(PGMTREES), 0, MM_TAG_PGM, (void **)&pVM->pgm.s.pTreesR3);
1303 if (RT_SUCCESS(rc))
1304 {
1305 pVM->pgm.s.pTreesR0 = MMHyperR3ToR0(pVM, pVM->pgm.s.pTreesR3);
1306 pVM->pgm.s.pTreesRC = MMHyperR3ToRC(pVM, pVM->pgm.s.pTreesR3);
1307
1308 /*
1309 * Alocate the zero page.
1310 */
1311 rc = MMHyperAlloc(pVM, PAGE_SIZE, PAGE_SIZE, MM_TAG_PGM, &pVM->pgm.s.pvZeroPgR3);
1312 }
1313 if (RT_SUCCESS(rc))
1314 {
1315 pVM->pgm.s.pvZeroPgRC = MMHyperR3ToRC(pVM, pVM->pgm.s.pvZeroPgR3);
1316 pVM->pgm.s.pvZeroPgR0 = MMHyperR3ToR0(pVM, pVM->pgm.s.pvZeroPgR3);
1317 pVM->pgm.s.HCPhysZeroPg = MMR3HyperHCVirt2HCPhys(pVM, pVM->pgm.s.pvZeroPgR3);
1318 AssertRelease(pVM->pgm.s.HCPhysZeroPg != NIL_RTHCPHYS);
1319
1320 /*
1321 * Init the paging.
1322 */
1323 rc = pgmR3InitPaging(pVM);
1324 }
1325 if (RT_SUCCESS(rc))
1326 {
1327 /*
1328 * Init the page pool.
1329 */
1330 rc = pgmR3PoolInit(pVM);
1331 }
1332 if (RT_SUCCESS(rc))
1333 {
1334 for (unsigned i=0;i<pVM->cCPUs;i++)
1335 {
1336 PVMCPU pVCpu = &pVM->aCpus[i];
1337
1338 rc = PGMR3ChangeMode(pVM, pVCpu, PGMMODE_REAL);
1339 if (RT_FAILURE(rc))
1340 break;
1341 }
1342 }
1343
1344 if (RT_SUCCESS(rc))
1345 {
1346 /*
1347 * Info & statistics
1348 */
1349 DBGFR3InfoRegisterInternal(pVM, "mode",
1350 "Shows the current paging mode. "
1351 "Recognizes 'all', 'guest', 'shadow' and 'host' as arguments, defaulting to 'all' if nothing's given.",
1352 pgmR3InfoMode);
1353 DBGFR3InfoRegisterInternal(pVM, "pgmcr3",
1354 "Dumps all the entries in the top level paging table. No arguments.",
1355 pgmR3InfoCr3);
1356 DBGFR3InfoRegisterInternal(pVM, "phys",
1357 "Dumps all the physical address ranges. No arguments.",
1358 pgmR3PhysInfo);
1359 DBGFR3InfoRegisterInternal(pVM, "handlers",
1360 "Dumps physical, virtual and hyper virtual handlers. "
1361 "Pass 'phys', 'virt', 'hyper' as argument if only one kind is wanted."
1362 "Add 'nost' if the statistics are unwanted, use together with 'all' or explicit selection.",
1363 pgmR3InfoHandlers);
1364 DBGFR3InfoRegisterInternal(pVM, "mappings",
1365 "Dumps guest mappings.",
1366 pgmR3MapInfo);
1367
1368 pgmR3InitStats(pVM);
1369
1370#ifdef VBOX_WITH_DEBUGGER
1371 /*
1372 * Debugger commands.
1373 */
1374 static bool s_fRegisteredCmds = false;
1375 if (!s_fRegisteredCmds)
1376 {
1377 int rc = DBGCRegisterCommands(&g_aCmds[0], RT_ELEMENTS(g_aCmds));
1378 if (RT_SUCCESS(rc))
1379 s_fRegisteredCmds = true;
1380 }
1381#endif
1382 return VINF_SUCCESS;
1383 }
1384
1385 /* Almost no cleanup necessary, MM frees all memory. */
1386 PDMR3CritSectDelete(&pVM->pgm.s.CritSect);
1387
1388 return rc;
1389}
1390
1391
1392/**
1393 * Initializes the per-VCPU PGM.
1394 *
1395 * @returns VBox status code.
1396 * @param pVM The VM to operate on.
1397 */
1398VMMR3DECL(int) PGMR3InitCPU(PVM pVM)
1399{
1400 LogFlow(("PGMR3InitCPU\n"));
1401 return VINF_SUCCESS;
1402}
1403
1404
1405/**
1406 * Init paging.
1407 *
1408 * Since we need to check what mode the host is operating in before we can choose
1409 * the right paging functions for the host we have to delay this until R0 has
1410 * been initialized.
1411 *
1412 * @returns VBox status code.
1413 * @param pVM VM handle.
1414 */
1415static int pgmR3InitPaging(PVM pVM)
1416{
1417 /*
1418 * Force a recalculation of modes and switcher so everyone gets notified.
1419 */
1420 for (unsigned i=0;i<pVM->cCPUs;i++)
1421 {
1422 PVMCPU pVCpu = &pVM->aCpus[i];
1423
1424 pVCpu->pgm.s.enmShadowMode = PGMMODE_INVALID;
1425 pVCpu->pgm.s.enmGuestMode = PGMMODE_INVALID;
1426 }
1427
1428 pVM->pgm.s.enmHostMode = SUPPAGINGMODE_INVALID;
1429
1430 /*
1431 * Allocate static mapping space for whatever the cr3 register
1432 * points to and in the case of PAE mode to the 4 PDs.
1433 */
1434 int rc = MMR3HyperReserve(pVM, PAGE_SIZE * 5, "CR3 mapping", &pVM->pgm.s.GCPtrCR3Mapping);
1435 if (RT_FAILURE(rc))
1436 {
1437 AssertMsgFailed(("Failed to reserve two pages for cr mapping in HMA, rc=%Rrc\n", rc));
1438 return rc;
1439 }
1440 MMR3HyperReserve(pVM, PAGE_SIZE, "fence", NULL);
1441
1442 /*
1443 * Allocate pages for the three possible intermediate contexts
1444 * (AMD64, PAE and plain 32-Bit). We maintain all three contexts
1445 * for the sake of simplicity. The AMD64 uses the PAE for the
1446 * lower levels, making the total number of pages 11 (3 + 7 + 1).
1447 *
1448 * We assume that two page tables will be enought for the core code
1449 * mappings (HC virtual and identity).
1450 */
1451 pVM->pgm.s.pInterPD = (PX86PD)MMR3PageAllocLow(pVM); AssertReturn(pVM->pgm.s.pInterPD, VERR_NO_PAGE_MEMORY);
1452 pVM->pgm.s.apInterPTs[0] = (PX86PT)MMR3PageAllocLow(pVM); AssertReturn(pVM->pgm.s.apInterPTs[0], VERR_NO_PAGE_MEMORY);
1453 pVM->pgm.s.apInterPTs[1] = (PX86PT)MMR3PageAllocLow(pVM); AssertReturn(pVM->pgm.s.apInterPTs[1], VERR_NO_PAGE_MEMORY);
1454 pVM->pgm.s.apInterPaePTs[0] = (PX86PTPAE)MMR3PageAlloc(pVM); AssertReturn(pVM->pgm.s.apInterPaePTs[0], VERR_NO_PAGE_MEMORY);
1455 pVM->pgm.s.apInterPaePTs[1] = (PX86PTPAE)MMR3PageAlloc(pVM); AssertReturn(pVM->pgm.s.apInterPaePTs[1], VERR_NO_PAGE_MEMORY);
1456 pVM->pgm.s.apInterPaePDs[0] = (PX86PDPAE)MMR3PageAlloc(pVM); AssertReturn(pVM->pgm.s.apInterPaePDs[0], VERR_NO_PAGE_MEMORY);
1457 pVM->pgm.s.apInterPaePDs[1] = (PX86PDPAE)MMR3PageAlloc(pVM); AssertReturn(pVM->pgm.s.apInterPaePDs[1], VERR_NO_PAGE_MEMORY);
1458 pVM->pgm.s.apInterPaePDs[2] = (PX86PDPAE)MMR3PageAlloc(pVM); AssertReturn(pVM->pgm.s.apInterPaePDs[2], VERR_NO_PAGE_MEMORY);
1459 pVM->pgm.s.apInterPaePDs[3] = (PX86PDPAE)MMR3PageAlloc(pVM); AssertReturn(pVM->pgm.s.apInterPaePDs[3], VERR_NO_PAGE_MEMORY);
1460 pVM->pgm.s.pInterPaePDPT = (PX86PDPT)MMR3PageAllocLow(pVM); AssertReturn(pVM->pgm.s.pInterPaePDPT, VERR_NO_PAGE_MEMORY);
1461 pVM->pgm.s.pInterPaePDPT64 = (PX86PDPT)MMR3PageAllocLow(pVM); AssertReturn(pVM->pgm.s.pInterPaePDPT64, VERR_NO_PAGE_MEMORY);
1462 pVM->pgm.s.pInterPaePML4 = (PX86PML4)MMR3PageAllocLow(pVM); AssertReturn(pVM->pgm.s.pInterPaePML4, VERR_NO_PAGE_MEMORY);
1463
1464 pVM->pgm.s.HCPhysInterPD = MMPage2Phys(pVM, pVM->pgm.s.pInterPD);
1465 AssertRelease(pVM->pgm.s.HCPhysInterPD != NIL_RTHCPHYS && !(pVM->pgm.s.HCPhysInterPD & PAGE_OFFSET_MASK));
1466 pVM->pgm.s.HCPhysInterPaePDPT = MMPage2Phys(pVM, pVM->pgm.s.pInterPaePDPT);
1467 AssertRelease(pVM->pgm.s.HCPhysInterPaePDPT != NIL_RTHCPHYS && !(pVM->pgm.s.HCPhysInterPaePDPT & PAGE_OFFSET_MASK));
1468 pVM->pgm.s.HCPhysInterPaePML4 = MMPage2Phys(pVM, pVM->pgm.s.pInterPaePML4);
1469 AssertRelease(pVM->pgm.s.HCPhysInterPaePML4 != NIL_RTHCPHYS && !(pVM->pgm.s.HCPhysInterPaePML4 & PAGE_OFFSET_MASK) && pVM->pgm.s.HCPhysInterPaePML4 < 0xffffffff);
1470
1471 /*
1472 * Initialize the pages, setting up the PML4 and PDPT for repetitive 4GB action.
1473 */
1474 ASMMemZeroPage(pVM->pgm.s.pInterPD);
1475 ASMMemZeroPage(pVM->pgm.s.apInterPTs[0]);
1476 ASMMemZeroPage(pVM->pgm.s.apInterPTs[1]);
1477
1478 ASMMemZeroPage(pVM->pgm.s.apInterPaePTs[0]);
1479 ASMMemZeroPage(pVM->pgm.s.apInterPaePTs[1]);
1480
1481 ASMMemZeroPage(pVM->pgm.s.pInterPaePDPT);
1482 for (unsigned i = 0; i < RT_ELEMENTS(pVM->pgm.s.apInterPaePDs); i++)
1483 {
1484 ASMMemZeroPage(pVM->pgm.s.apInterPaePDs[i]);
1485 pVM->pgm.s.pInterPaePDPT->a[i].u = X86_PDPE_P | PGM_PLXFLAGS_PERMANENT
1486 | MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[i]);
1487 }
1488
1489 for (unsigned i = 0; i < RT_ELEMENTS(pVM->pgm.s.pInterPaePDPT64->a); i++)
1490 {
1491 const unsigned iPD = i % RT_ELEMENTS(pVM->pgm.s.apInterPaePDs);
1492 pVM->pgm.s.pInterPaePDPT64->a[i].u = X86_PDPE_P | X86_PDPE_RW | X86_PDPE_US | X86_PDPE_A | PGM_PLXFLAGS_PERMANENT
1493 | MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[iPD]);
1494 }
1495
1496 RTHCPHYS HCPhysInterPaePDPT64 = MMPage2Phys(pVM, pVM->pgm.s.pInterPaePDPT64);
1497 for (unsigned i = 0; i < RT_ELEMENTS(pVM->pgm.s.pInterPaePML4->a); i++)
1498 pVM->pgm.s.pInterPaePML4->a[i].u = X86_PML4E_P | X86_PML4E_RW | X86_PML4E_US | X86_PML4E_A | PGM_PLXFLAGS_PERMANENT
1499 | HCPhysInterPaePDPT64;
1500
1501 /*
1502 * Initialize paging workers and mode from current host mode
1503 * and the guest running in real mode.
1504 */
1505 pVM->pgm.s.enmHostMode = SUPGetPagingMode();
1506 switch (pVM->pgm.s.enmHostMode)
1507 {
1508 case SUPPAGINGMODE_32_BIT:
1509 case SUPPAGINGMODE_32_BIT_GLOBAL:
1510 case SUPPAGINGMODE_PAE:
1511 case SUPPAGINGMODE_PAE_GLOBAL:
1512 case SUPPAGINGMODE_PAE_NX:
1513 case SUPPAGINGMODE_PAE_GLOBAL_NX:
1514 break;
1515
1516 case SUPPAGINGMODE_AMD64:
1517 case SUPPAGINGMODE_AMD64_GLOBAL:
1518 case SUPPAGINGMODE_AMD64_NX:
1519 case SUPPAGINGMODE_AMD64_GLOBAL_NX:
1520#ifndef VBOX_WITH_HYBRID_32BIT_KERNEL
1521 if (ARCH_BITS != 64)
1522 {
1523 AssertMsgFailed(("Host mode %d (64-bit) is not supported by non-64bit builds\n", pVM->pgm.s.enmHostMode));
1524 LogRel(("Host mode %d (64-bit) is not supported by non-64bit builds\n", pVM->pgm.s.enmHostMode));
1525 return VERR_PGM_UNSUPPORTED_HOST_PAGING_MODE;
1526 }
1527#endif
1528 break;
1529 default:
1530 AssertMsgFailed(("Host mode %d is not supported\n", pVM->pgm.s.enmHostMode));
1531 return VERR_PGM_UNSUPPORTED_HOST_PAGING_MODE;
1532 }
1533 rc = pgmR3ModeDataInit(pVM, false /* don't resolve GC and R0 syms yet */);
1534 if (RT_SUCCESS(rc))
1535 {
1536 LogFlow(("pgmR3InitPaging: returns successfully\n"));
1537#if HC_ARCH_BITS == 64
1538 LogRel(("Debug: HCPhysInterPD=%RHp HCPhysInterPaePDPT=%RHp HCPhysInterPaePML4=%RHp\n",
1539 pVM->pgm.s.HCPhysInterPD, pVM->pgm.s.HCPhysInterPaePDPT, pVM->pgm.s.HCPhysInterPaePML4));
1540 LogRel(("Debug: apInterPTs={%RHp,%RHp} apInterPaePTs={%RHp,%RHp} apInterPaePDs={%RHp,%RHp,%RHp,%RHp} pInterPaePDPT64=%RHp\n",
1541 MMPage2Phys(pVM, pVM->pgm.s.apInterPTs[0]), MMPage2Phys(pVM, pVM->pgm.s.apInterPTs[1]),
1542 MMPage2Phys(pVM, pVM->pgm.s.apInterPaePTs[0]), MMPage2Phys(pVM, pVM->pgm.s.apInterPaePTs[1]),
1543 MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[0]), MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[1]), MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[2]), MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[3]),
1544 MMPage2Phys(pVM, pVM->pgm.s.pInterPaePDPT64)));
1545#endif
1546
1547 return VINF_SUCCESS;
1548 }
1549
1550 LogFlow(("pgmR3InitPaging: returns %Rrc\n", rc));
1551 return rc;
1552}
1553
1554
1555/**
1556 * Init statistics
1557 */
1558static void pgmR3InitStats(PVM pVM)
1559{
1560 PPGM pPGM = &pVM->pgm.s;
1561 int rc;
1562
1563 /* Common - misc variables */
1564 STAM_REL_REG(pVM, &pPGM->cAllPages, STAMTYPE_U32, "/PGM/Page/cAllPages", STAMUNIT_OCCURENCES, "The total number of pages.");
1565 STAM_REL_REG(pVM, &pPGM->cPrivatePages, STAMTYPE_U32, "/PGM/Page/cPrivatePages", STAMUNIT_OCCURENCES, "The number of private pages.");
1566 STAM_REL_REG(pVM, &pPGM->cSharedPages, STAMTYPE_U32, "/PGM/Page/cSharedPages", STAMUNIT_OCCURENCES, "The number of shared pages.");
1567 STAM_REL_REG(pVM, &pPGM->cZeroPages, STAMTYPE_U32, "/PGM/Page/cZeroPages", STAMUNIT_OCCURENCES, "The number of zero backed pages.");
1568 STAM_REL_REG(pVM, &pPGM->cHandyPages, STAMTYPE_U32, "/PGM/Page/cHandyPages", STAMUNIT_OCCURENCES, "The number of handy pages (not included in cAllPages).");
1569 STAM_REL_REG(pVM, &pPGM->cRelocations, STAMTYPE_COUNTER, "/PGM/cRelocations", STAMUNIT_OCCURENCES, "Number of hypervisor relocations.");
1570 STAM_REL_REG(pVM, &pPGM->ChunkR3Map.c, STAMTYPE_U32, "/PGM/ChunkR3Map/c", STAMUNIT_OCCURENCES, "Number of mapped chunks.");
1571 STAM_REL_REG(pVM, &pPGM->ChunkR3Map.cMax, STAMTYPE_U32, "/PGM/ChunkR3Map/cMax", STAMUNIT_OCCURENCES, "Maximum number of mapped chunks.");
1572
1573#ifdef VBOX_WITH_STATISTICS
1574
1575# define PGM_REG_COUNTER(a, b, c) \
1576 rc = STAMR3RegisterF(pVM, a, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS, STAMUNIT_OCCURENCES, c, b); \
1577 AssertRC(rc);
1578
1579# define PGM_REG_PROFILE(a, b, c) \
1580 rc = STAMR3RegisterF(pVM, a, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS, STAMUNIT_TICKS_PER_CALL, c, b); \
1581 AssertRC(rc);
1582
1583 PGM_REG_COUNTER(&pPGM->StatR3DetectedConflicts, "/PGM/R3/DetectedConflicts", "The number of times PGMR3CheckMappingConflicts() detected a conflict.");
1584 PGM_REG_PROFILE(&pPGM->StatR3ResolveConflict, "/PGM/R3/ResolveConflict", "pgmR3SyncPTResolveConflict() profiling (includes the entire relocation).");
1585
1586 PGM_REG_COUNTER(&pPGM->StatRZChunkR3MapTlbHits, "/PGM/ChunkR3Map/TlbHitsRZ", "TLB hits.");
1587 PGM_REG_COUNTER(&pPGM->StatRZChunkR3MapTlbMisses, "/PGM/ChunkR3Map/TlbMissesRZ", "TLB misses.");
1588 PGM_REG_COUNTER(&pPGM->StatRZPageMapTlbHits, "/PGM/RZ/Page/MapTlbHits", "TLB hits.");
1589 PGM_REG_COUNTER(&pPGM->StatRZPageMapTlbMisses, "/PGM/RZ/Page/MapTlbMisses", "TLB misses.");
1590 PGM_REG_COUNTER(&pPGM->StatR3ChunkR3MapTlbHits, "/PGM/ChunkR3Map/TlbHitsR3", "TLB hits.");
1591 PGM_REG_COUNTER(&pPGM->StatR3ChunkR3MapTlbMisses, "/PGM/ChunkR3Map/TlbMissesR3", "TLB misses.");
1592 PGM_REG_COUNTER(&pPGM->StatR3PageMapTlbHits, "/PGM/R3/Page/MapTlbHits", "TLB hits.");
1593 PGM_REG_COUNTER(&pPGM->StatR3PageMapTlbMisses, "/PGM/R3/Page/MapTlbMisses", "TLB misses.");
1594
1595 PGM_REG_PROFILE(&pPGM->StatRZSyncCR3HandlerVirtualUpdate, "/PGM/RZ/SyncCR3/Handlers/VirtualUpdate", "Profiling of the virtual handler updates.");
1596 PGM_REG_PROFILE(&pPGM->StatRZSyncCR3HandlerVirtualReset, "/PGM/RZ/SyncCR3/Handlers/VirtualReset", "Profiling of the virtual handler resets.");
1597 PGM_REG_PROFILE(&pPGM->StatR3SyncCR3HandlerVirtualUpdate, "/PGM/R3/SyncCR3/Handlers/VirtualUpdate", "Profiling of the virtual handler updates.");
1598 PGM_REG_PROFILE(&pPGM->StatR3SyncCR3HandlerVirtualReset, "/PGM/R3/SyncCR3/Handlers/VirtualReset", "Profiling of the virtual handler resets.");
1599
1600 PGM_REG_COUNTER(&pPGM->StatRZPhysHandlerReset, "/PGM/RZ/PhysHandlerReset", "The number of times PGMHandlerPhysicalReset is called.");
1601 PGM_REG_COUNTER(&pPGM->StatR3PhysHandlerReset, "/PGM/R3/PhysHandlerReset", "The number of times PGMHandlerPhysicalReset is called.");
1602 PGM_REG_PROFILE(&pPGM->StatRZVirtHandlerSearchByPhys, "/PGM/RZ/VirtHandlerSearchByPhys", "Profiling of pgmHandlerVirtualFindByPhysAddr.");
1603 PGM_REG_PROFILE(&pPGM->StatR3VirtHandlerSearchByPhys, "/PGM/R3/VirtHandlerSearchByPhys", "Profiling of pgmHandlerVirtualFindByPhysAddr.");
1604
1605 PGM_REG_COUNTER(&pPGM->StatRZPageReplaceShared, "/PGM/RZ/Page/ReplacedShared", "Times a shared page was replaced.");
1606 PGM_REG_COUNTER(&pPGM->StatRZPageReplaceZero, "/PGM/RZ/Page/ReplacedZero", "Times the zero page was replaced.");
1607/// @todo PGM_REG_COUNTER(&pPGM->StatRZPageHandyAllocs, "/PGM/RZ/Page/HandyAllocs", "Number of times we've allocated more handy pages.");
1608 PGM_REG_COUNTER(&pPGM->StatR3PageReplaceShared, "/PGM/R3/Page/ReplacedShared", "Times a shared page was replaced.");
1609 PGM_REG_COUNTER(&pPGM->StatR3PageReplaceZero, "/PGM/R3/Page/ReplacedZero", "Times the zero page was replaced.");
1610/// @todo PGM_REG_COUNTER(&pPGM->StatR3PageHandyAllocs, "/PGM/R3/Page/HandyAllocs", "Number of times we've allocated more handy pages.");
1611
1612 /* GC only: */
1613 PGM_REG_COUNTER(&pPGM->StatRCDynMapCacheHits, "/PGM/RC/DynMapCache/Hits" , "Number of dynamic page mapping cache hits.");
1614 PGM_REG_COUNTER(&pPGM->StatRCDynMapCacheMisses, "/PGM/RC/DynMapCache/Misses" , "Number of dynamic page mapping cache misses.");
1615 PGM_REG_COUNTER(&pPGM->StatRCInvlPgConflict, "/PGM/RC/InvlPgConflict", "Number of times PGMInvalidatePage() detected a mapping conflict.");
1616 PGM_REG_COUNTER(&pPGM->StatRCInvlPgSyncMonCR3, "/PGM/RC/InvlPgSyncMonitorCR3", "Number of times PGMInvalidatePage() ran into PGM_SYNC_MONITOR_CR3.");
1617
1618# ifdef PGMPOOL_WITH_GCPHYS_TRACKING
1619 PGM_REG_COUNTER(&pPGM->StatTrackVirgin, "/PGM/Track/Virgin", "The number of first time shadowings");
1620 PGM_REG_COUNTER(&pPGM->StatTrackAliased, "/PGM/Track/Aliased", "The number of times switching to cRef2, i.e. the page is being shadowed by two PTs.");
1621 PGM_REG_COUNTER(&pPGM->StatTrackAliasedMany, "/PGM/Track/AliasedMany", "The number of times we're tracking using cRef2.");
1622 PGM_REG_COUNTER(&pPGM->StatTrackAliasedLots, "/PGM/Track/AliasedLots", "The number of times we're hitting pages which has overflowed cRef2");
1623 PGM_REG_COUNTER(&pPGM->StatTrackOverflows, "/PGM/Track/Overflows", "The number of times the extent list grows too long.");
1624 PGM_REG_PROFILE(&pPGM->StatTrackDeref, "/PGM/Track/Deref", "Profiling of SyncPageWorkerTrackDeref (expensive).");
1625# endif
1626
1627# undef PGM_REG_COUNTER
1628# undef PGM_REG_PROFILE
1629#endif
1630
1631 /*
1632 * Note! The layout below matches the member layout exactly!
1633 */
1634
1635 /*
1636 * Common - stats
1637 */
1638 for (unsigned i=0;i<pVM->cCPUs;i++)
1639 {
1640 PVMCPU pVCpu = &pVM->aCpus[i];
1641 PPGMCPU pPGM = &pVCpu->pgm.s;
1642
1643#define PGM_REG_COUNTER(a, b, c) \
1644 rc = STAMR3RegisterF(pVM, a, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS, STAMUNIT_OCCURENCES, c, b, i); \
1645 AssertRC(rc);
1646#define PGM_REG_PROFILE(a, b, c) \
1647 rc = STAMR3RegisterF(pVM, a, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS, STAMUNIT_TICKS_PER_CALL, c, b, i); \
1648 AssertRC(rc);
1649
1650 PGM_REG_COUNTER(&pPGM->cGuestModeChanges, "/PGM/CPU%d/cGuestModeChanges", "Number of guest mode changes.");
1651
1652#ifdef VBOX_WITH_STATISTICS
1653 for (unsigned j = 0; j < RT_ELEMENTS(pPGM->StatSyncPtPD); j++)
1654 STAMR3RegisterF(pVM, &pPGM->StatSyncPtPD[i], STAMTYPE_COUNTER, STAMVISIBILITY_USED, STAMUNIT_OCCURENCES,
1655 "The number of SyncPT per PD n.", "/PGM/CPU%d/PDSyncPT/%04X", i, j);
1656 for (unsigned j = 0; j < RT_ELEMENTS(pPGM->StatSyncPagePD); j++)
1657 STAMR3RegisterF(pVM, &pPGM->StatSyncPagePD[i], STAMTYPE_COUNTER, STAMVISIBILITY_USED, STAMUNIT_OCCURENCES,
1658 "The number of SyncPage per PD n.", "/PGM/CPU%d/PDSyncPage/%04X", i, j);
1659
1660 /* R0 only: */
1661 PGM_REG_COUNTER(&pPGM->StatR0DynMapMigrateInvlPg, "/PGM/CPU%d/R0/DynMapMigrateInvlPg", "invlpg count in PGMDynMapMigrateAutoSet.");
1662 PGM_REG_PROFILE(&pPGM->StatR0DynMapGCPageInl, "/PGM/CPU%d/R0/DynMapPageGCPageInl", "Calls to pgmR0DynMapGCPageInlined.");
1663 PGM_REG_COUNTER(&pPGM->StatR0DynMapGCPageInlHits, "/PGM/CPU%d/R0/DynMapPageGCPageInl/Hits", "Hash table lookup hits.");
1664 PGM_REG_COUNTER(&pPGM->StatR0DynMapGCPageInlMisses, "/PGM/CPU%d/R0/DynMapPageGCPageInl/Misses", "Misses that falls back to code common with PGMDynMapHCPage.");
1665 PGM_REG_COUNTER(&pPGM->StatR0DynMapGCPageInlRamHits, "/PGM/CPU%d/R0/DynMapPageGCPageInl/RamHits", "1st ram range hits.");
1666 PGM_REG_COUNTER(&pPGM->StatR0DynMapGCPageInlRamMisses, "/PGM/CPU%d/R0/DynMapPageGCPageInl/RamMisses", "1st ram range misses, takes slow path.");
1667 PGM_REG_PROFILE(&pPGM->StatR0DynMapHCPageInl, "/PGM/CPU%d/R0/DynMapPageHCPageInl", "Calls to pgmR0DynMapHCPageInlined.");
1668 PGM_REG_COUNTER(&pPGM->StatR0DynMapHCPageInlHits, "/PGM/CPU%d/R0/DynMapPageHCPageInl/Hits", "Hash table lookup hits.");
1669 PGM_REG_COUNTER(&pPGM->StatR0DynMapHCPageInlMisses, "/PGM/CPU%d/R0/DynMapPageHCPageInl/Misses", "Misses that falls back to code common with PGMDynMapHCPage.");
1670 PGM_REG_COUNTER(&pPGM->StatR0DynMapPage, "/PGM/CPU%d/R0/DynMapPage", "Calls to pgmR0DynMapPage");
1671 PGM_REG_COUNTER(&pPGM->StatR0DynMapSetOptimize, "/PGM/CPU%d/R0/DynMapPage/SetOptimize", "Calls to pgmDynMapOptimizeAutoSet.");
1672 PGM_REG_COUNTER(&pPGM->StatR0DynMapSetSearchFlushes, "/PGM/CPU%d/R0/DynMapPage/SetSearchFlushes","Set search restorting to subset flushes.");
1673 PGM_REG_COUNTER(&pPGM->StatR0DynMapSetSearchHits, "/PGM/CPU%d/R0/DynMapPage/SetSearchHits", "Set search hits.");
1674 PGM_REG_COUNTER(&pPGM->StatR0DynMapSetSearchMisses, "/PGM/CPU%d/R0/DynMapPage/SetSearchMisses", "Set search misses.");
1675 PGM_REG_PROFILE(&pPGM->StatR0DynMapHCPage, "/PGM/CPU%d/R0/DynMapPage/HCPage", "Calls to PGMDynMapHCPage (ring-0).");
1676 PGM_REG_COUNTER(&pPGM->StatR0DynMapPageHits0, "/PGM/CPU%d/R0/DynMapPage/Hits0", "Hits at iPage+0");
1677 PGM_REG_COUNTER(&pPGM->StatR0DynMapPageHits1, "/PGM/CPU%d/R0/DynMapPage/Hits1", "Hits at iPage+1");
1678 PGM_REG_COUNTER(&pPGM->StatR0DynMapPageHits2, "/PGM/CPU%d/R0/DynMapPage/Hits2", "Hits at iPage+2");
1679 PGM_REG_COUNTER(&pPGM->StatR0DynMapPageInvlPg, "/PGM/CPU%d/R0/DynMapPage/InvlPg", "invlpg count in pgmR0DynMapPageSlow.");
1680 PGM_REG_COUNTER(&pPGM->StatR0DynMapPageSlow, "/PGM/CPU%d/R0/DynMapPage/Slow", "Calls to pgmR0DynMapPageSlow - subtract this from pgmR0DynMapPage to get 1st level hits.");
1681 PGM_REG_COUNTER(&pPGM->StatR0DynMapPageSlowLoopHits, "/PGM/CPU%d/R0/DynMapPage/SlowLoopHits" , "Hits in the loop path.");
1682 PGM_REG_COUNTER(&pPGM->StatR0DynMapPageSlowLoopMisses, "/PGM/CPU%d/R0/DynMapPage/SlowLoopMisses", "Misses in the loop path. NonLoopMisses = Slow - SlowLoopHit - SlowLoopMisses");
1683 //PGM_REG_COUNTER(&pPGM->StatR0DynMapPageSlowLostHits, "/PGM/CPU%d/R0/DynMapPage/SlowLostHits", "Lost hits.");
1684 PGM_REG_COUNTER(&pPGM->StatR0DynMapSubsets, "/PGM/CPU%d/R0/Subsets", "Times PGMDynMapPushAutoSubset was called.");
1685 PGM_REG_COUNTER(&pPGM->StatR0DynMapPopFlushes, "/PGM/CPU%d/R0/SubsetPopFlushes", "Times PGMDynMapPopAutoSubset flushes the subset.");
1686 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[0], "/PGM/CPU%d/R0/SetSize000..09", "00-09% filled");
1687 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[1], "/PGM/CPU%d/R0/SetSize010..19", "10-19% filled");
1688 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[2], "/PGM/CPU%d/R0/SetSize020..29", "20-29% filled");
1689 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[3], "/PGM/CPU%d/R0/SetSize030..39", "30-39% filled");
1690 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[4], "/PGM/CPU%d/R0/SetSize040..49", "40-49% filled");
1691 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[5], "/PGM/CPU%d/R0/SetSize050..59", "50-59% filled");
1692 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[6], "/PGM/CPU%d/R0/SetSize060..69", "60-69% filled");
1693 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[7], "/PGM/CPU%d/R0/SetSize070..79", "70-79% filled");
1694 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[8], "/PGM/CPU%d/R0/SetSize080..89", "80-89% filled");
1695 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[9], "/PGM/CPU%d/R0/SetSize090..99", "90-99% filled");
1696 PGM_REG_COUNTER(&pPGM->aStatR0DynMapSetSize[10], "/PGM/CPU%d/R0/SetSize100", "100% filled");
1697
1698 /* RZ only: */
1699 PGM_REG_PROFILE(&pPGM->StatRZTrap0e, "/PGM/CPU%d/RZ/Trap0e", "Profiling of the PGMTrap0eHandler() body.");
1700 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTimeCheckPageFault, "/PGM/CPU%d/RZ/Trap0e/Time/CheckPageFault", "Profiling of checking for dirty/access emulation faults.");
1701 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTimeSyncPT, "/PGM/CPU%d/RZ/Trap0e/Time/SyncPT", "Profiling of lazy page table syncing.");
1702 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTimeMapping, "/PGM/CPU%d/RZ/Trap0e/Time/Mapping", "Profiling of checking virtual mappings.");
1703 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTimeOutOfSync, "/PGM/CPU%d/RZ/Trap0e/Time/OutOfSync", "Profiling of out of sync page handling.");
1704 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTimeHandlers, "/PGM/CPU%d/RZ/Trap0e/Time/Handlers", "Profiling of checking handlers.");
1705 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2CSAM, "/PGM/CPU%d/RZ/Trap0e/Time2/CSAM", "Profiling of the Trap0eHandler body when the cause is CSAM.");
1706 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2DirtyAndAccessed, "/PGM/CPU%d/RZ/Trap0e/Time2/DirtyAndAccessedBits", "Profiling of the Trap0eHandler body when the cause is dirty and/or accessed bit emulation.");
1707 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2GuestTrap, "/PGM/CPU%d/RZ/Trap0e/Time2/GuestTrap", "Profiling of the Trap0eHandler body when the cause is a guest trap.");
1708 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2HndPhys, "/PGM/CPU%d/RZ/Trap0e/Time2/HandlerPhysical", "Profiling of the Trap0eHandler body when the cause is a physical handler.");
1709 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2HndVirt, "/PGM/CPU%d/RZ/Trap0e/Time2/HandlerVirtual", "Profiling of the Trap0eHandler body when the cause is a virtual handler.");
1710 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2HndUnhandled, "/PGM/CPU%d/RZ/Trap0e/Time2/HandlerUnhandled", "Profiling of the Trap0eHandler body when the cause is access outside the monitored areas of a monitored page.");
1711 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2Misc, "/PGM/CPU%d/RZ/Trap0e/Time2/Misc", "Profiling of the Trap0eHandler body when the cause is not known.");
1712 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2OutOfSync, "/PGM/CPU%d/RZ/Trap0e/Time2/OutOfSync", "Profiling of the Trap0eHandler body when the cause is an out-of-sync page.");
1713 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2OutOfSyncHndPhys, "/PGM/CPU%d/RZ/Trap0e/Time2/OutOfSyncHndPhys", "Profiling of the Trap0eHandler body when the cause is an out-of-sync physical handler page.");
1714 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2OutOfSyncHndVirt, "/PGM/CPU%d/RZ/Trap0e/Time2/OutOfSyncHndVirt", "Profiling of the Trap0eHandler body when the cause is an out-of-sync virtual handler page.");
1715 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2OutOfSyncHndObs, "/PGM/CPU%d/RZ/Trap0e/Time2/OutOfSyncObsHnd", "Profiling of the Trap0eHandler body when the cause is an obsolete handler page.");
1716 PGM_REG_PROFILE(&pPGM->StatRZTrap0eTime2SyncPT, "/PGM/CPU%d/RZ/Trap0e/Time2/SyncPT", "Profiling of the Trap0eHandler body when the cause is lazy syncing of a PT.");
1717 PGM_REG_COUNTER(&pPGM->StatRZTrap0eConflicts, "/PGM/CPU%d/RZ/Trap0e/Conflicts", "The number of times #PF was caused by an undetected conflict.");
1718 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersMapping, "/PGM/CPU%d/RZ/Trap0e/Handlers/Mapping", "Number of traps due to access handlers in mappings.");
1719 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersOutOfSync, "/PGM/CPU%d/RZ/Trap0e/Handlers/OutOfSync", "Number of traps due to out-of-sync handled pages.");
1720 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersPhysical, "/PGM/CPU%d/RZ/Trap0e/Handlers/Physical", "Number of traps due to physical access handlers.");
1721 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersVirtual, "/PGM/CPU%d/RZ/Trap0e/Handlers/Virtual", "Number of traps due to virtual access handlers.");
1722 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersVirtualByPhys, "/PGM/CPU%d/RZ/Trap0e/Handlers/VirtualByPhys", "Number of traps due to virtual access handlers by physical address.");
1723 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersVirtualUnmarked,"/PGM/CPU%d/RZ/Trap0e/Handlers/VirtualUnmarked","Number of traps due to virtual access handlers by virtual address (without proper physical flags).");
1724 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersUnhandled, "/PGM/CPU%d/RZ/Trap0e/Handlers/Unhandled", "Number of traps due to access outside range of monitored page(s).");
1725 PGM_REG_COUNTER(&pPGM->StatRZTrap0eHandlersInvalid, "/PGM/CPU%d/RZ/Trap0e/Handlers/Invalid", "Number of traps due to access to invalid physical memory.");
1726 PGM_REG_COUNTER(&pPGM->StatRZTrap0eUSNotPresentRead, "/PGM/CPU%d/RZ/Trap0e/Err/User/NPRead", "Number of user mode not present read page faults.");
1727 PGM_REG_COUNTER(&pPGM->StatRZTrap0eUSNotPresentWrite, "/PGM/CPU%d/RZ/Trap0e/Err/User/NPWrite", "Number of user mode not present write page faults.");
1728 PGM_REG_COUNTER(&pPGM->StatRZTrap0eUSWrite, "/PGM/CPU%d/RZ/Trap0e/Err/User/Write", "Number of user mode write page faults.");
1729 PGM_REG_COUNTER(&pPGM->StatRZTrap0eUSReserved, "/PGM/CPU%d/RZ/Trap0e/Err/User/Reserved", "Number of user mode reserved bit page faults.");
1730 PGM_REG_COUNTER(&pPGM->StatRZTrap0eUSNXE, "/PGM/CPU%d/RZ/Trap0e/Err/User/NXE", "Number of user mode NXE page faults.");
1731 PGM_REG_COUNTER(&pPGM->StatRZTrap0eUSRead, "/PGM/CPU%d/RZ/Trap0e/Err/User/Read", "Number of user mode read page faults.");
1732 PGM_REG_COUNTER(&pPGM->StatRZTrap0eSVNotPresentRead, "/PGM/CPU%d/RZ/Trap0e/Err/Supervisor/NPRead", "Number of supervisor mode not present read page faults.");
1733 PGM_REG_COUNTER(&pPGM->StatRZTrap0eSVNotPresentWrite, "/PGM/CPU%d/RZ/Trap0e/Err/Supervisor/NPWrite", "Number of supervisor mode not present write page faults.");
1734 PGM_REG_COUNTER(&pPGM->StatRZTrap0eSVWrite, "/PGM/CPU%d/RZ/Trap0e/Err/Supervisor/Write", "Number of supervisor mode write page faults.");
1735 PGM_REG_COUNTER(&pPGM->StatRZTrap0eSVReserved, "/PGM/CPU%d/RZ/Trap0e/Err/Supervisor/Reserved", "Number of supervisor mode reserved bit page faults.");
1736 PGM_REG_COUNTER(&pPGM->StatRZTrap0eSNXE, "/PGM/CPU%d/RZ/Trap0e/Err/Supervisor/NXE", "Number of supervisor mode NXE page faults.");
1737 PGM_REG_COUNTER(&pPGM->StatRZTrap0eGuestPF, "/PGM/CPU%d/RZ/Trap0e/GuestPF", "Number of real guest page faults.");
1738 PGM_REG_COUNTER(&pPGM->StatRZTrap0eGuestPFUnh, "/PGM/CPU%d/RZ/Trap0e/GuestPF/Unhandled", "Number of real guest page faults from the 'unhandled' case.");
1739 PGM_REG_COUNTER(&pPGM->StatRZTrap0eGuestPFMapping, "/PGM/CPU%d/RZ/Trap0e/GuestPF/InMapping", "Number of real guest page faults in a mapping.");
1740 PGM_REG_COUNTER(&pPGM->StatRZTrap0eWPEmulInRZ, "/PGM/CPU%d/RZ/Trap0e/WP/InRZ", "Number of guest page faults due to X86_CR0_WP emulation.");
1741 PGM_REG_COUNTER(&pPGM->StatRZTrap0eWPEmulToR3, "/PGM/CPU%d/RZ/Trap0e/WP/ToR3", "Number of guest page faults due to X86_CR0_WP emulation (forward to R3 for emulation).");
1742 for (unsigned j = 0; j < RT_ELEMENTS(pPGM->StatRZTrap0ePD); j++)
1743 STAMR3RegisterF(pVM, &pPGM->StatRZTrap0ePD[i], STAMTYPE_COUNTER, STAMVISIBILITY_USED, STAMUNIT_OCCURENCES,
1744 "The number of traps in page directory n.", "/PGM/CPU%d/RZ/Trap0e/PD/%04X", i, j);
1745
1746 PGM_REG_COUNTER(&pPGM->StatRZGuestCR3WriteHandled, "/PGM/CPU%d/RZ/CR3WriteHandled", "The number of times the Guest CR3 change was successfully handled.");
1747 PGM_REG_COUNTER(&pPGM->StatRZGuestCR3WriteUnhandled, "/PGM/CPU%d/RZ/CR3WriteUnhandled", "The number of times the Guest CR3 change was passed back to the recompiler.");
1748 PGM_REG_COUNTER(&pPGM->StatRZGuestCR3WriteConflict, "/PGM/CPU%d/RZ/CR3WriteConflict", "The number of times the Guest CR3 monitoring detected a conflict.");
1749 PGM_REG_COUNTER(&pPGM->StatRZGuestROMWriteHandled, "/PGM/CPU%d/RZ/ROMWriteHandled", "The number of times the Guest ROM change was successfully handled.");
1750 PGM_REG_COUNTER(&pPGM->StatRZGuestROMWriteUnhandled, "/PGM/CPU%d/RZ/ROMWriteUnhandled", "The number of times the Guest ROM change was passed back to the recompiler.");
1751
1752 /* HC only: */
1753
1754 /* RZ & R3: */
1755 PGM_REG_PROFILE(&pPGM->StatRZSyncCR3, "/PGM/CPU%d/RZ/SyncCR3", "Profiling of the PGMSyncCR3() body.");
1756 PGM_REG_PROFILE(&pPGM->StatRZSyncCR3Handlers, "/PGM/CPU%d/RZ/SyncCR3/Handlers", "Profiling of the PGMSyncCR3() update handler section.");
1757 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3Global, "/PGM/CPU%d/RZ/SyncCR3/Global", "The number of global CR3 syncs.");
1758 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3NotGlobal, "/PGM/CPU%d/RZ/SyncCR3/NotGlobal", "The number of non-global CR3 syncs.");
1759 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3DstCacheHit, "/PGM/CPU%d/RZ/SyncCR3/DstChacheHit", "The number of times we got some kind of a cache hit.");
1760 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3DstFreed, "/PGM/CPU%d/RZ/SyncCR3/DstFreed", "The number of times we've had to free a shadow entry.");
1761 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3DstFreedSrcNP, "/PGM/CPU%d/RZ/SyncCR3/DstFreedSrcNP", "The number of times we've had to free a shadow entry for which the source entry was not present.");
1762 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3DstNotPresent, "/PGM/CPU%d/RZ/SyncCR3/DstNotPresent", "The number of times we've encountered a not present shadow entry for a present guest entry.");
1763 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3DstSkippedGlobalPD, "/PGM/CPU%d/RZ/SyncCR3/DstSkippedGlobalPD", "The number of times a global page directory wasn't flushed.");
1764 PGM_REG_COUNTER(&pPGM->StatRZSyncCR3DstSkippedGlobalPT, "/PGM/CPU%d/RZ/SyncCR3/DstSkippedGlobalPT", "The number of times a page table with only global entries wasn't flushed.");
1765 PGM_REG_PROFILE(&pPGM->StatRZSyncPT, "/PGM/CPU%d/RZ/SyncPT", "Profiling of the pfnSyncPT() body.");
1766 PGM_REG_COUNTER(&pPGM->StatRZSyncPTFailed, "/PGM/CPU%d/RZ/SyncPT/Failed", "The number of times pfnSyncPT() failed.");
1767 PGM_REG_COUNTER(&pPGM->StatRZSyncPT4K, "/PGM/CPU%d/RZ/SyncPT/4K", "Nr of 4K PT syncs");
1768 PGM_REG_COUNTER(&pPGM->StatRZSyncPT4M, "/PGM/CPU%d/RZ/SyncPT/4M", "Nr of 4M PT syncs");
1769 PGM_REG_COUNTER(&pPGM->StatRZSyncPagePDNAs, "/PGM/CPU%d/RZ/SyncPagePDNAs", "The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit.");
1770 PGM_REG_COUNTER(&pPGM->StatRZSyncPagePDOutOfSync, "/PGM/CPU%d/RZ/SyncPagePDOutOfSync", "The number of time we've encountered an out-of-sync PD in SyncPage.");
1771 PGM_REG_COUNTER(&pPGM->StatRZAccessedPage, "/PGM/CPU%d/RZ/AccessedPage", "The number of pages marked not present for accessed bit emulation.");
1772 PGM_REG_PROFILE(&pPGM->StatRZDirtyBitTracking, "/PGM/CPU%d/RZ/DirtyPage", "Profiling the dirty bit tracking in CheckPageFault().");
1773 PGM_REG_COUNTER(&pPGM->StatRZDirtyPage, "/PGM/CPU%d/RZ/DirtyPage/Mark", "The number of pages marked read-only for dirty bit tracking.");
1774 PGM_REG_COUNTER(&pPGM->StatRZDirtyPageBig, "/PGM/CPU%d/RZ/DirtyPage/MarkBig", "The number of 4MB pages marked read-only for dirty bit tracking.");
1775 PGM_REG_COUNTER(&pPGM->StatRZDirtyPageSkipped, "/PGM/CPU%d/RZ/DirtyPage/Skipped", "The number of pages already dirty or readonly.");
1776 PGM_REG_COUNTER(&pPGM->StatRZDirtyPageTrap, "/PGM/CPU%d/RZ/DirtyPage/Trap", "The number of traps generated for dirty bit tracking.");
1777 PGM_REG_COUNTER(&pPGM->StatRZDirtyPageStale, "/PGM/CPU%d/RZ/DirtyPage/Stale", "The number of traps generated for dirty bit tracking (stale tlb entries).");
1778 PGM_REG_COUNTER(&pPGM->StatRZDirtiedPage, "/PGM/CPU%d/RZ/DirtyPage/SetDirty", "The number of pages marked dirty because of write accesses.");
1779 PGM_REG_COUNTER(&pPGM->StatRZDirtyTrackRealPF, "/PGM/CPU%d/RZ/DirtyPage/RealPF", "The number of real pages faults during dirty bit tracking.");
1780 PGM_REG_COUNTER(&pPGM->StatRZPageAlreadyDirty, "/PGM/CPU%d/RZ/DirtyPage/AlreadySet", "The number of pages already marked dirty because of write accesses.");
1781 PGM_REG_PROFILE(&pPGM->StatRZInvalidatePage, "/PGM/CPU%d/RZ/InvalidatePage", "PGMInvalidatePage() profiling.");
1782 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePage4KBPages, "/PGM/CPU%d/RZ/InvalidatePage/4KBPages", "The number of times PGMInvalidatePage() was called for a 4KB page.");
1783 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePage4MBPages, "/PGM/CPU%d/RZ/InvalidatePage/4MBPages", "The number of times PGMInvalidatePage() was called for a 4MB page.");
1784 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePage4MBPagesSkip, "/PGM/CPU%d/RZ/InvalidatePage/4MBPagesSkip","The number of times PGMInvalidatePage() skipped a 4MB page.");
1785 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePagePDMappings, "/PGM/CPU%d/RZ/InvalidatePage/PDMappings", "The number of times PGMInvalidatePage() was called for a page directory containing mappings (no conflict).");
1786 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePagePDNAs, "/PGM/CPU%d/RZ/InvalidatePage/PDNAs", "The number of times PGMInvalidatePage() was called for a not accessed page directory.");
1787 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePagePDNPs, "/PGM/CPU%d/RZ/InvalidatePage/PDNPs", "The number of times PGMInvalidatePage() was called for a not present page directory.");
1788 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePagePDOutOfSync, "/PGM/CPU%d/RZ/InvalidatePage/PDOutOfSync", "The number of times PGMInvalidatePage() was called for an out of sync page directory.");
1789 PGM_REG_COUNTER(&pPGM->StatRZInvalidatePageSkipped, "/PGM/CPU%d/RZ/InvalidatePage/Skipped", "The number of times PGMInvalidatePage() was skipped due to not present shw or pending pending SyncCR3.");
1790 PGM_REG_COUNTER(&pPGM->StatRZPageOutOfSyncSupervisor, "/PGM/CPU%d/RZ/OutOfSync/SuperVisor", "Number of traps due to pages out of sync and times VerifyAccessSyncPage calls SyncPage.");
1791 PGM_REG_COUNTER(&pPGM->StatRZPageOutOfSyncUser, "/PGM/CPU%d/RZ/OutOfSync/User", "Number of traps due to pages out of sync and times VerifyAccessSyncPage calls SyncPage.");
1792 PGM_REG_PROFILE(&pPGM->StatRZPrefetch, "/PGM/CPU%d/RZ/Prefetch", "PGMPrefetchPage profiling.");
1793 PGM_REG_PROFILE(&pPGM->StatRZFlushTLB, "/PGM/CPU%d/RZ/FlushTLB", "Profiling of the PGMFlushTLB() body.");
1794 PGM_REG_COUNTER(&pPGM->StatRZFlushTLBNewCR3, "/PGM/CPU%d/RZ/FlushTLB/NewCR3", "The number of times PGMFlushTLB was called with a new CR3, non-global. (switch)");
1795 PGM_REG_COUNTER(&pPGM->StatRZFlushTLBNewCR3Global, "/PGM/CPU%d/RZ/FlushTLB/NewCR3Global", "The number of times PGMFlushTLB was called with a new CR3, global. (switch)");
1796 PGM_REG_COUNTER(&pPGM->StatRZFlushTLBSameCR3, "/PGM/CPU%d/RZ/FlushTLB/SameCR3", "The number of times PGMFlushTLB was called with the same CR3, non-global. (flush)");
1797 PGM_REG_COUNTER(&pPGM->StatRZFlushTLBSameCR3Global, "/PGM/CPU%d/RZ/FlushTLB/SameCR3Global", "The number of times PGMFlushTLB was called with the same CR3, global. (flush)");
1798 PGM_REG_PROFILE(&pPGM->StatRZGstModifyPage, "/PGM/CPU%d/RZ/GstModifyPage", "Profiling of the PGMGstModifyPage() body.");
1799
1800 PGM_REG_PROFILE(&pPGM->StatR3SyncCR3, "/PGM/CPU%d/R3/SyncCR3", "Profiling of the PGMSyncCR3() body.");
1801 PGM_REG_PROFILE(&pPGM->StatR3SyncCR3Handlers, "/PGM/CPU%d/R3/SyncCR3/Handlers", "Profiling of the PGMSyncCR3() update handler section.");
1802 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3Global, "/PGM/CPU%d/R3/SyncCR3/Global", "The number of global CR3 syncs.");
1803 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3NotGlobal, "/PGM/CPU%d/R3/SyncCR3/NotGlobal", "The number of non-global CR3 syncs.");
1804 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3DstCacheHit, "/PGM/CPU%d/R3/SyncCR3/DstChacheHit", "The number of times we got some kind of a cache hit.");
1805 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3DstFreed, "/PGM/CPU%d/R3/SyncCR3/DstFreed", "The number of times we've had to free a shadow entry.");
1806 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3DstFreedSrcNP, "/PGM/CPU%d/R3/SyncCR3/DstFreedSrcNP", "The number of times we've had to free a shadow entry for which the source entry was not present.");
1807 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3DstNotPresent, "/PGM/CPU%d/R3/SyncCR3/DstNotPresent", "The number of times we've encountered a not present shadow entry for a present guest entry.");
1808 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3DstSkippedGlobalPD, "/PGM/CPU%d/R3/SyncCR3/DstSkippedGlobalPD", "The number of times a global page directory wasn't flushed.");
1809 PGM_REG_COUNTER(&pPGM->StatR3SyncCR3DstSkippedGlobalPT, "/PGM/CPU%d/R3/SyncCR3/DstSkippedGlobalPT", "The number of times a page table with only global entries wasn't flushed.");
1810 PGM_REG_PROFILE(&pPGM->StatR3SyncPT, "/PGM/CPU%d/R3/SyncPT", "Profiling of the pfnSyncPT() body.");
1811 PGM_REG_COUNTER(&pPGM->StatR3SyncPTFailed, "/PGM/CPU%d/R3/SyncPT/Failed", "The number of times pfnSyncPT() failed.");
1812 PGM_REG_COUNTER(&pPGM->StatR3SyncPT4K, "/PGM/CPU%d/R3/SyncPT/4K", "Nr of 4K PT syncs");
1813 PGM_REG_COUNTER(&pPGM->StatR3SyncPT4M, "/PGM/CPU%d/R3/SyncPT/4M", "Nr of 4M PT syncs");
1814 PGM_REG_COUNTER(&pPGM->StatR3SyncPagePDNAs, "/PGM/CPU%d/R3/SyncPagePDNAs", "The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit.");
1815 PGM_REG_COUNTER(&pPGM->StatR3SyncPagePDOutOfSync, "/PGM/CPU%d/R3/SyncPagePDOutOfSync", "The number of time we've encountered an out-of-sync PD in SyncPage.");
1816 PGM_REG_COUNTER(&pPGM->StatR3AccessedPage, "/PGM/CPU%d/R3/AccessedPage", "The number of pages marked not present for accessed bit emulation.");
1817 PGM_REG_PROFILE(&pPGM->StatR3DirtyBitTracking, "/PGM/CPU%d/R3/DirtyPage", "Profiling the dirty bit tracking in CheckPageFault().");
1818 PGM_REG_COUNTER(&pPGM->StatR3DirtyPage, "/PGM/CPU%d/R3/DirtyPage/Mark", "The number of pages marked read-only for dirty bit tracking.");
1819 PGM_REG_COUNTER(&pPGM->StatR3DirtyPageBig, "/PGM/CPU%d/R3/DirtyPage/MarkBig", "The number of 4MB pages marked read-only for dirty bit tracking.");
1820 PGM_REG_COUNTER(&pPGM->StatR3DirtyPageSkipped, "/PGM/CPU%d/R3/DirtyPage/Skipped", "The number of pages already dirty or readonly.");
1821 PGM_REG_COUNTER(&pPGM->StatR3DirtyPageTrap, "/PGM/CPU%d/R3/DirtyPage/Trap", "The number of traps generated for dirty bit tracking.");
1822 PGM_REG_COUNTER(&pPGM->StatR3DirtiedPage, "/PGM/CPU%d/R3/DirtyPage/SetDirty", "The number of pages marked dirty because of write accesses.");
1823 PGM_REG_COUNTER(&pPGM->StatR3DirtyTrackRealPF, "/PGM/CPU%d/R3/DirtyPage/RealPF", "The number of real pages faults during dirty bit tracking.");
1824 PGM_REG_COUNTER(&pPGM->StatR3PageAlreadyDirty, "/PGM/CPU%d/R3/DirtyPage/AlreadySet", "The number of pages already marked dirty because of write accesses.");
1825 PGM_REG_PROFILE(&pPGM->StatR3InvalidatePage, "/PGM/CPU%d/R3/InvalidatePage", "PGMInvalidatePage() profiling.");
1826 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePage4KBPages, "/PGM/CPU%d/R3/InvalidatePage/4KBPages", "The number of times PGMInvalidatePage() was called for a 4KB page.");
1827 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePage4MBPages, "/PGM/CPU%d/R3/InvalidatePage/4MBPages", "The number of times PGMInvalidatePage() was called for a 4MB page.");
1828 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePage4MBPagesSkip, "/PGM/CPU%d/R3/InvalidatePage/4MBPagesSkip","The number of times PGMInvalidatePage() skipped a 4MB page.");
1829 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePagePDMappings, "/PGM/CPU%d/R3/InvalidatePage/PDMappings", "The number of times PGMInvalidatePage() was called for a page directory containing mappings (no conflict).");
1830 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePagePDNAs, "/PGM/CPU%d/R3/InvalidatePage/PDNAs", "The number of times PGMInvalidatePage() was called for a not accessed page directory.");
1831 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePagePDNPs, "/PGM/CPU%d/R3/InvalidatePage/PDNPs", "The number of times PGMInvalidatePage() was called for a not present page directory.");
1832 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePagePDOutOfSync, "/PGM/CPU%d/R3/InvalidatePage/PDOutOfSync", "The number of times PGMInvalidatePage() was called for an out of sync page directory.");
1833 PGM_REG_COUNTER(&pPGM->StatR3InvalidatePageSkipped, "/PGM/CPU%d/R3/InvalidatePage/Skipped", "The number of times PGMInvalidatePage() was skipped due to not present shw or pending pending SyncCR3.");
1834 PGM_REG_COUNTER(&pPGM->StatR3PageOutOfSyncSupervisor, "/PGM/CPU%d/R3/OutOfSync/SuperVisor", "Number of traps due to pages out of sync and times VerifyAccessSyncPage calls SyncPage.");
1835 PGM_REG_COUNTER(&pPGM->StatR3PageOutOfSyncUser, "/PGM/CPU%d/R3/OutOfSync/User", "Number of traps due to pages out of sync and times VerifyAccessSyncPage calls SyncPage.");
1836 PGM_REG_PROFILE(&pPGM->StatR3Prefetch, "/PGM/CPU%d/R3/Prefetch", "PGMPrefetchPage profiling.");
1837 PGM_REG_PROFILE(&pPGM->StatR3FlushTLB, "/PGM/CPU%d/R3/FlushTLB", "Profiling of the PGMFlushTLB() body.");
1838 PGM_REG_COUNTER(&pPGM->StatR3FlushTLBNewCR3, "/PGM/CPU%d/R3/FlushTLB/NewCR3", "The number of times PGMFlushTLB was called with a new CR3, non-global. (switch)");
1839 PGM_REG_COUNTER(&pPGM->StatR3FlushTLBNewCR3Global, "/PGM/CPU%d/R3/FlushTLB/NewCR3Global", "The number of times PGMFlushTLB was called with a new CR3, global. (switch)");
1840 PGM_REG_COUNTER(&pPGM->StatR3FlushTLBSameCR3, "/PGM/CPU%d/R3/FlushTLB/SameCR3", "The number of times PGMFlushTLB was called with the same CR3, non-global. (flush)");
1841 PGM_REG_COUNTER(&pPGM->StatR3FlushTLBSameCR3Global, "/PGM/CPU%d/R3/FlushTLB/SameCR3Global", "The number of times PGMFlushTLB was called with the same CR3, global. (flush)");
1842 PGM_REG_PROFILE(&pPGM->StatR3GstModifyPage, "/PGM/CPU%d/R3/GstModifyPage", "Profiling of the PGMGstModifyPage() body.");
1843#endif /* VBOX_WITH_STATISTICS */
1844
1845#undef PGM_REG_PROFILE
1846#undef PGM_REG_COUNTER
1847
1848 }
1849}
1850
1851
1852/**
1853 * Init the PGM bits that rely on VMMR0 and MM to be fully initialized.
1854 *
1855 * The dynamic mapping area will also be allocated and initialized at this
1856 * time. We could allocate it during PGMR3Init of course, but the mapping
1857 * wouldn't be allocated at that time preventing us from setting up the
1858 * page table entries with the dummy page.
1859 *
1860 * @returns VBox status code.
1861 * @param pVM VM handle.
1862 */
1863VMMR3DECL(int) PGMR3InitDynMap(PVM pVM)
1864{
1865 RTGCPTR GCPtr;
1866 int rc;
1867
1868 /*
1869 * Reserve space for the dynamic mappings.
1870 */
1871 rc = MMR3HyperReserve(pVM, MM_HYPER_DYNAMIC_SIZE, "Dynamic mapping", &GCPtr);
1872 if (RT_SUCCESS(rc))
1873 pVM->pgm.s.pbDynPageMapBaseGC = GCPtr;
1874
1875 if ( RT_SUCCESS(rc)
1876 && (pVM->pgm.s.pbDynPageMapBaseGC >> X86_PD_PAE_SHIFT) != ((pVM->pgm.s.pbDynPageMapBaseGC + MM_HYPER_DYNAMIC_SIZE - 1) >> X86_PD_PAE_SHIFT))
1877 {
1878 rc = MMR3HyperReserve(pVM, MM_HYPER_DYNAMIC_SIZE, "Dynamic mapping not crossing", &GCPtr);
1879 if (RT_SUCCESS(rc))
1880 pVM->pgm.s.pbDynPageMapBaseGC = GCPtr;
1881 }
1882 if (RT_SUCCESS(rc))
1883 {
1884 AssertRelease((pVM->pgm.s.pbDynPageMapBaseGC >> X86_PD_PAE_SHIFT) == ((pVM->pgm.s.pbDynPageMapBaseGC + MM_HYPER_DYNAMIC_SIZE - 1) >> X86_PD_PAE_SHIFT));
1885 MMR3HyperReserve(pVM, PAGE_SIZE, "fence", NULL);
1886 }
1887 return rc;
1888}
1889
1890
1891/**
1892 * Ring-3 init finalizing.
1893 *
1894 * @returns VBox status code.
1895 * @param pVM The VM handle.
1896 */
1897VMMR3DECL(int) PGMR3InitFinalize(PVM pVM)
1898{
1899 int rc;
1900
1901 /*
1902 * Reserve space for the dynamic mappings.
1903 * Initialize the dynamic mapping pages with dummy pages to simply the cache.
1904 */
1905 /* get the pointer to the page table entries. */
1906 PPGMMAPPING pMapping = pgmGetMapping(pVM, pVM->pgm.s.pbDynPageMapBaseGC);
1907 AssertRelease(pMapping);
1908 const uintptr_t off = pVM->pgm.s.pbDynPageMapBaseGC - pMapping->GCPtr;
1909 const unsigned iPT = off >> X86_PD_SHIFT;
1910 const unsigned iPG = (off >> X86_PT_SHIFT) & X86_PT_MASK;
1911 pVM->pgm.s.paDynPageMap32BitPTEsGC = pMapping->aPTs[iPT].pPTRC + iPG * sizeof(pMapping->aPTs[0].pPTR3->a[0]);
1912 pVM->pgm.s.paDynPageMapPaePTEsGC = pMapping->aPTs[iPT].paPaePTsRC + iPG * sizeof(pMapping->aPTs[0].paPaePTsR3->a[0]);
1913
1914 /* init cache */
1915 RTHCPHYS HCPhysDummy = MMR3PageDummyHCPhys(pVM);
1916 for (unsigned i = 0; i < RT_ELEMENTS(pVM->pgm.s.aHCPhysDynPageMapCache); i++)
1917 pVM->pgm.s.aHCPhysDynPageMapCache[i] = HCPhysDummy;
1918
1919 for (unsigned i = 0; i < MM_HYPER_DYNAMIC_SIZE; i += PAGE_SIZE)
1920 {
1921 rc = PGMMap(pVM, pVM->pgm.s.pbDynPageMapBaseGC + i, HCPhysDummy, PAGE_SIZE, 0);
1922 AssertRCReturn(rc, rc);
1923 }
1924
1925 /*
1926 * Note that AMD uses all the 8 reserved bits for the address (so 40 bits in total);
1927 * Intel only goes up to 36 bits, so we stick to 36 as well.
1928 */
1929 /** @todo How to test for the 40 bits support? Long mode seems to be the test criterium. */
1930 uint32_t u32Dummy, u32Features;
1931 CPUMGetGuestCpuId(VMMGetCpu(pVM), 1, &u32Dummy, &u32Dummy, &u32Dummy, &u32Features);
1932
1933 if (u32Features & X86_CPUID_FEATURE_EDX_PSE36)
1934 pVM->pgm.s.GCPhys4MBPSEMask = RT_BIT_64(36) - 1;
1935 else
1936 pVM->pgm.s.GCPhys4MBPSEMask = RT_BIT_64(32) - 1;
1937
1938 /*
1939 * Allocate memory if we're supposed to do that.
1940 */
1941 if (pVM->pgm.s.fRamPreAlloc)
1942 rc = pgmR3PhysRamPreAllocate(pVM);
1943
1944 LogRel(("PGMR3InitFinalize: 4 MB PSE mask %RGp\n", pVM->pgm.s.GCPhys4MBPSEMask));
1945 return rc;
1946}
1947
1948
1949/**
1950 * Applies relocations to data and code managed by this component.
1951 *
1952 * This function will be called at init and whenever the VMM need to relocate it
1953 * self inside the GC.
1954 *
1955 * @param pVM The VM.
1956 * @param offDelta Relocation delta relative to old location.
1957 */
1958VMMR3DECL(void) PGMR3Relocate(PVM pVM, RTGCINTPTR offDelta)
1959{
1960 LogFlow(("PGMR3Relocate %RGv to %RGv\n", pVM->pgm.s.GCPtrCR3Mapping, pVM->pgm.s.GCPtrCR3Mapping + offDelta));
1961
1962 /*
1963 * Paging stuff.
1964 */
1965 pVM->pgm.s.GCPtrCR3Mapping += offDelta;
1966
1967 pgmR3ModeDataInit(pVM, true /* resolve GC/R0 symbols */);
1968
1969 /* Shadow, guest and both mode switch & relocation for each VCPU. */
1970 for (unsigned i=0;i<pVM->cCPUs;i++)
1971 {
1972 PVMCPU pVCpu = &pVM->aCpus[i];
1973
1974 pgmR3ModeDataSwitch(pVM, pVCpu, pVCpu->pgm.s.enmShadowMode, pVCpu->pgm.s.enmGuestMode);
1975
1976 PGM_SHW_PFN(Relocate, pVCpu)(pVCpu, offDelta);
1977 PGM_GST_PFN(Relocate, pVCpu)(pVCpu, offDelta);
1978 PGM_BTH_PFN(Relocate, pVCpu)(pVCpu, offDelta);
1979 }
1980
1981 /*
1982 * Trees.
1983 */
1984 pVM->pgm.s.pTreesRC = MMHyperR3ToRC(pVM, pVM->pgm.s.pTreesR3);
1985
1986 /*
1987 * Ram ranges.
1988 */
1989 if (pVM->pgm.s.pRamRangesR3)
1990 {
1991 /* Update the pSelfRC pointers and relink them. */
1992 for (PPGMRAMRANGE pCur = pVM->pgm.s.pRamRangesR3; pCur; pCur = pCur->pNextR3)
1993 if (!(pCur->fFlags & PGM_RAM_RANGE_FLAGS_FLOATING))
1994 pCur->pSelfRC = MMHyperCCToRC(pVM, pCur);
1995 pgmR3PhysRelinkRamRanges(pVM);
1996 }
1997
1998 /*
1999 * Update the two page directories with all page table mappings.
2000 * (One or more of them have changed, that's why we're here.)
2001 */
2002 pVM->pgm.s.pMappingsRC = MMHyperR3ToRC(pVM, pVM->pgm.s.pMappingsR3);
2003 for (PPGMMAPPING pCur = pVM->pgm.s.pMappingsR3; pCur->pNextR3; pCur = pCur->pNextR3)
2004 pCur->pNextRC = MMHyperR3ToRC(pVM, pCur->pNextR3);
2005
2006 /* Relocate GC addresses of Page Tables. */
2007 for (PPGMMAPPING pCur = pVM->pgm.s.pMappingsR3; pCur; pCur = pCur->pNextR3)
2008 {
2009 for (RTHCUINT i = 0; i < pCur->cPTs; i++)
2010 {
2011 pCur->aPTs[i].pPTRC = MMHyperR3ToRC(pVM, pCur->aPTs[i].pPTR3);
2012 pCur->aPTs[i].paPaePTsRC = MMHyperR3ToRC(pVM, pCur->aPTs[i].paPaePTsR3);
2013 }
2014 }
2015
2016 /*
2017 * Dynamic page mapping area.
2018 */
2019 pVM->pgm.s.paDynPageMap32BitPTEsGC += offDelta;
2020 pVM->pgm.s.paDynPageMapPaePTEsGC += offDelta;
2021 pVM->pgm.s.pbDynPageMapBaseGC += offDelta;
2022
2023 /*
2024 * The Zero page.
2025 */
2026 pVM->pgm.s.pvZeroPgR0 = MMHyperR3ToR0(pVM, pVM->pgm.s.pvZeroPgR3);
2027#ifdef VBOX_WITH_2X_4GB_ADDR_SPACE
2028 AssertRelease(pVM->pgm.s.pvZeroPgR0 != NIL_RTR0PTR || !VMMIsHwVirtExtForced(pVM));
2029#else
2030 AssertRelease(pVM->pgm.s.pvZeroPgR0 != NIL_RTR0PTR);
2031#endif
2032
2033 /*
2034 * Physical and virtual handlers.
2035 */
2036 RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesR3->PhysHandlers, true, pgmR3RelocatePhysHandler, &offDelta);
2037 RTAvlroGCPtrDoWithAll(&pVM->pgm.s.pTreesR3->VirtHandlers, true, pgmR3RelocateVirtHandler, &offDelta);
2038 RTAvlroGCPtrDoWithAll(&pVM->pgm.s.pTreesR3->HyperVirtHandlers, true, pgmR3RelocateHyperVirtHandler, &offDelta);
2039
2040 /*
2041 * The page pool.
2042 */
2043 pgmR3PoolRelocate(pVM);
2044}
2045
2046
2047/**
2048 * Callback function for relocating a physical access handler.
2049 *
2050 * @returns 0 (continue enum)
2051 * @param pNode Pointer to a PGMPHYSHANDLER node.
2052 * @param pvUser Pointer to the offDelta. This is a pointer to the delta since we're
2053 * not certain the delta will fit in a void pointer for all possible configs.
2054 */
2055static DECLCALLBACK(int) pgmR3RelocatePhysHandler(PAVLROGCPHYSNODECORE pNode, void *pvUser)
2056{
2057 PPGMPHYSHANDLER pHandler = (PPGMPHYSHANDLER)pNode;
2058 RTGCINTPTR offDelta = *(PRTGCINTPTR)pvUser;
2059 if (pHandler->pfnHandlerRC)
2060 pHandler->pfnHandlerRC += offDelta;
2061 if (pHandler->pvUserRC >= 0x10000)
2062 pHandler->pvUserRC += offDelta;
2063 return 0;
2064}
2065
2066
2067/**
2068 * Callback function for relocating a virtual access handler.
2069 *
2070 * @returns 0 (continue enum)
2071 * @param pNode Pointer to a PGMVIRTHANDLER node.
2072 * @param pvUser Pointer to the offDelta. This is a pointer to the delta since we're
2073 * not certain the delta will fit in a void pointer for all possible configs.
2074 */
2075static DECLCALLBACK(int) pgmR3RelocateVirtHandler(PAVLROGCPTRNODECORE pNode, void *pvUser)
2076{
2077 PPGMVIRTHANDLER pHandler = (PPGMVIRTHANDLER)pNode;
2078 RTGCINTPTR offDelta = *(PRTGCINTPTR)pvUser;
2079 Assert( pHandler->enmType == PGMVIRTHANDLERTYPE_ALL
2080 || pHandler->enmType == PGMVIRTHANDLERTYPE_WRITE);
2081 Assert(pHandler->pfnHandlerRC);
2082 pHandler->pfnHandlerRC += offDelta;
2083 return 0;
2084}
2085
2086
2087/**
2088 * Callback function for relocating a virtual access handler for the hypervisor mapping.
2089 *
2090 * @returns 0 (continue enum)
2091 * @param pNode Pointer to a PGMVIRTHANDLER node.
2092 * @param pvUser Pointer to the offDelta. This is a pointer to the delta since we're
2093 * not certain the delta will fit in a void pointer for all possible configs.
2094 */
2095static DECLCALLBACK(int) pgmR3RelocateHyperVirtHandler(PAVLROGCPTRNODECORE pNode, void *pvUser)
2096{
2097 PPGMVIRTHANDLER pHandler = (PPGMVIRTHANDLER)pNode;
2098 RTGCINTPTR offDelta = *(PRTGCINTPTR)pvUser;
2099 Assert(pHandler->enmType == PGMVIRTHANDLERTYPE_HYPERVISOR);
2100 Assert(pHandler->pfnHandlerRC);
2101 pHandler->pfnHandlerRC += offDelta;
2102 return 0;
2103}
2104
2105
2106/**
2107 * The VM is being reset.
2108 *
2109 * For the PGM component this means that any PD write monitors
2110 * needs to be removed.
2111 *
2112 * @param pVM VM handle.
2113 */
2114VMMR3DECL(void) PGMR3Reset(PVM pVM)
2115{
2116 int rc;
2117
2118 LogFlow(("PGMR3Reset:\n"));
2119 VM_ASSERT_EMT(pVM);
2120
2121 pgmLock(pVM);
2122
2123 /*
2124 * Unfix any fixed mappings and disable CR3 monitoring.
2125 */
2126 pVM->pgm.s.fMappingsFixed = false;
2127 pVM->pgm.s.GCPtrMappingFixed = 0;
2128 pVM->pgm.s.cbMappingFixed = 0;
2129
2130 /* Exit the guest paging mode before the pgm pool gets reset.
2131 * Important to clean up the amd64 case.
2132 */
2133 for (unsigned i=0;i<pVM->cCPUs;i++)
2134 {
2135 PVMCPU pVCpu = &pVM->aCpus[i];
2136
2137 rc = PGM_GST_PFN(Exit, pVCpu)(pVCpu);
2138 AssertRC(rc);
2139 }
2140
2141#ifdef DEBUG
2142 DBGFR3InfoLog(pVM, "mappings", NULL);
2143 DBGFR3InfoLog(pVM, "handlers", "all nostat");
2144#endif
2145
2146 /*
2147 * Reset the shadow page pool.
2148 */
2149 pgmR3PoolReset(pVM);
2150
2151 for (unsigned i=0;i<pVM->cCPUs;i++)
2152 {
2153 PVMCPU pVCpu = &pVM->aCpus[i];
2154
2155 /*
2156 * Re-init other members.
2157 */
2158 pVCpu->pgm.s.fA20Enabled = true;
2159
2160 /*
2161 * Clear the FFs PGM owns.
2162 */
2163 VMCPU_FF_CLEAR(pVCpu, VMCPU_FF_PGM_SYNC_CR3);
2164 VMCPU_FF_CLEAR(pVCpu, VMCPU_FF_PGM_SYNC_CR3_NON_GLOBAL);
2165 }
2166
2167 /*
2168 * Reset (zero) RAM pages.
2169 */
2170 rc = pgmR3PhysRamReset(pVM);
2171 if (RT_SUCCESS(rc))
2172 {
2173 /*
2174 * Reset (zero) shadow ROM pages.
2175 */
2176 rc = pgmR3PhysRomReset(pVM);
2177 if (RT_SUCCESS(rc))
2178 {
2179 /*
2180 * Switch mode back to real mode.
2181 */
2182 for (unsigned i=0;i<pVM->cCPUs;i++)
2183 {
2184 PVMCPU pVCpu = &pVM->aCpus[i];
2185
2186 rc = PGMR3ChangeMode(pVM, pVCpu, PGMMODE_REAL);
2187 AssertRC(rc);
2188
2189 STAM_REL_COUNTER_RESET(&pVCpu->pgm.s.cGuestModeChanges);
2190 }
2191 }
2192 }
2193
2194 pgmUnlock(pVM);
2195 //return rc;
2196 AssertReleaseRC(rc);
2197}
2198
2199
2200#ifdef VBOX_STRICT
2201/**
2202 * VM state change callback for clearing fNoMorePhysWrites after
2203 * a snapshot has been created.
2204 */
2205static DECLCALLBACK(void) pgmR3ResetNoMorePhysWritesFlag(PVM pVM, VMSTATE enmState, VMSTATE enmOldState, void *pvUser)
2206{
2207 if (enmState == VMSTATE_RUNNING)
2208 pVM->pgm.s.fNoMorePhysWrites = false;
2209}
2210#endif
2211
2212
2213/**
2214 * Terminates the PGM.
2215 *
2216 * @returns VBox status code.
2217 * @param pVM Pointer to VM structure.
2218 */
2219VMMR3DECL(int) PGMR3Term(PVM pVM)
2220{
2221 PGMDeregisterStringFormatTypes();
2222 return PDMR3CritSectDelete(&pVM->pgm.s.CritSect);
2223}
2224
2225
2226/**
2227 * Terminates the per-VCPU PGM.
2228 *
2229 * Termination means cleaning up and freeing all resources,
2230 * the VM it self is at this point powered off or suspended.
2231 *
2232 * @returns VBox status code.
2233 * @param pVM The VM to operate on.
2234 */
2235VMMR3DECL(int) PGMR3TermCPU(PVM pVM)
2236{
2237 return 0;
2238}
2239
2240
2241/**
2242 * Find the ROM tracking structure for the given page.
2243 *
2244 * @returns Pointer to the ROM page structure. NULL if the caller didn't check
2245 * that it's a ROM page.
2246 * @param pVM The VM handle.
2247 * @param GCPhys The address of the ROM page.
2248 */
2249static PPGMROMPAGE pgmR3GetRomPage(PVM pVM, RTGCPHYS GCPhys)
2250{
2251 for (PPGMROMRANGE pRomRange = pVM->pgm.s.CTX_SUFF(pRomRanges);
2252 pRomRange;
2253 pRomRange = pRomRange->CTX_SUFF(pNext))
2254 {
2255 RTGCPHYS off = GCPhys - pRomRange->GCPhys;
2256 if (GCPhys - pRomRange->GCPhys < pRomRange->cb)
2257 return &pRomRange->aPages[off >> PAGE_SHIFT];
2258 }
2259 return NULL;
2260}
2261
2262
2263/**
2264 * Save zero indicator + bits for the specified page.
2265 *
2266 * @returns VBox status code, errors are logged/asserted before returning.
2267 * @param pVM The VM handle.
2268 * @param pSSH The saved state handle.
2269 * @param pPage The page to save.
2270 * @param GCPhys The address of the page.
2271 * @param pRam The ram range (for error logging).
2272 */
2273static int pgmR3SavePage(PVM pVM, PSSMHANDLE pSSM, PPGMPAGE pPage, RTGCPHYS GCPhys, PPGMRAMRANGE pRam)
2274{
2275 int rc;
2276 if (PGM_PAGE_IS_ZERO(pPage))
2277 rc = SSMR3PutU8(pSSM, 0);
2278 else
2279 {
2280 void const *pvPage;
2281 rc = pgmPhysGCPhys2CCPtrInternalReadOnly(pVM, pPage, GCPhys, &pvPage);
2282 AssertLogRelMsgRCReturn(rc, ("pPage=%R[pgmpage] GCPhys=%#x %s\n", pPage, GCPhys, pRam->pszDesc), rc);
2283
2284 SSMR3PutU8(pSSM, 1);
2285 rc = SSMR3PutMem(pSSM, pvPage, PAGE_SIZE);
2286 }
2287 return rc;
2288}
2289
2290
2291/**
2292 * Save a shadowed ROM page.
2293 *
2294 * Format: Type, protection, and two pages with zero indicators.
2295 *
2296 * @returns VBox status code, errors are logged/asserted before returning.
2297 * @param pVM The VM handle.
2298 * @param pSSH The saved state handle.
2299 * @param pPage The page to save.
2300 * @param GCPhys The address of the page.
2301 * @param pRam The ram range (for error logging).
2302 */
2303static int pgmR3SaveShadowedRomPage(PVM pVM, PSSMHANDLE pSSM, PPGMPAGE pPage, RTGCPHYS GCPhys, PPGMRAMRANGE pRam)
2304{
2305 /* Need to save both pages and the current state. */
2306 PPGMROMPAGE pRomPage = pgmR3GetRomPage(pVM, GCPhys);
2307 AssertLogRelMsgReturn(pRomPage, ("GCPhys=%RGp %s\n", GCPhys, pRam->pszDesc), VERR_INTERNAL_ERROR);
2308
2309 SSMR3PutU8(pSSM, PGMPAGETYPE_ROM_SHADOW);
2310 SSMR3PutU8(pSSM, pRomPage->enmProt);
2311
2312 int rc = pgmR3SavePage(pVM, pSSM, pPage, GCPhys, pRam);
2313 if (RT_SUCCESS(rc))
2314 {
2315 PPGMPAGE pPagePassive = PGMROMPROT_IS_ROM(pRomPage->enmProt) ? &pRomPage->Shadow : &pRomPage->Virgin;
2316 rc = pgmR3SavePage(pVM, pSSM, pPagePassive, GCPhys, pRam);
2317 }
2318 return rc;
2319}
2320
2321/** PGM fields to save/load. */
2322static const SSMFIELD s_aPGMFields[] =
2323{
2324 SSMFIELD_ENTRY( PGM, fMappingsFixed),
2325 SSMFIELD_ENTRY_GCPTR( PGM, GCPtrMappingFixed),
2326 SSMFIELD_ENTRY( PGM, cbMappingFixed),
2327 SSMFIELD_ENTRY_TERM()
2328};
2329
2330static const SSMFIELD s_aPGMCpuFields[] =
2331{
2332 SSMFIELD_ENTRY( PGMCPU, fA20Enabled),
2333 SSMFIELD_ENTRY_GCPHYS( PGMCPU, GCPhysA20Mask),
2334 SSMFIELD_ENTRY( PGMCPU, enmGuestMode),
2335 SSMFIELD_ENTRY_TERM()
2336};
2337
2338/* For loading old saved states. (pre-smp) */
2339typedef struct
2340{
2341 /** If set no conflict checks are required. (boolean) */
2342 bool fMappingsFixed;
2343 /** Size of fixed mapping */
2344 uint32_t cbMappingFixed;
2345 /** Base address (GC) of fixed mapping */
2346 RTGCPTR GCPtrMappingFixed;
2347 /** A20 gate mask.
2348 * Our current approach to A20 emulation is to let REM do it and don't bother
2349 * anywhere else. The interesting Guests will be operating with it enabled anyway.
2350 * But whould need arrise, we'll subject physical addresses to this mask. */
2351 RTGCPHYS GCPhysA20Mask;
2352 /** A20 gate state - boolean! */
2353 bool fA20Enabled;
2354 /** The guest paging mode. */
2355 PGMMODE enmGuestMode;
2356} PGMOLD;
2357
2358static const SSMFIELD s_aPGMFields_Old[] =
2359{
2360 SSMFIELD_ENTRY( PGMOLD, fMappingsFixed),
2361 SSMFIELD_ENTRY_GCPTR( PGMOLD, GCPtrMappingFixed),
2362 SSMFIELD_ENTRY( PGMOLD, cbMappingFixed),
2363 SSMFIELD_ENTRY( PGMOLD, fA20Enabled),
2364 SSMFIELD_ENTRY_GCPHYS( PGMOLD, GCPhysA20Mask),
2365 SSMFIELD_ENTRY( PGMOLD, enmGuestMode),
2366 SSMFIELD_ENTRY_TERM()
2367};
2368
2369
2370/**
2371 * Execute state save operation.
2372 *
2373 * @returns VBox status code.
2374 * @param pVM VM Handle.
2375 * @param pSSM SSM operation handle.
2376 */
2377static DECLCALLBACK(int) pgmR3Save(PVM pVM, PSSMHANDLE pSSM)
2378{
2379 int rc;
2380 unsigned i;
2381 PPGM pPGM = &pVM->pgm.s;
2382
2383 /*
2384 * Lock PGM and set the no-more-writes indicator.
2385 */
2386 pgmLock(pVM);
2387 pVM->pgm.s.fNoMorePhysWrites = true;
2388
2389 /*
2390 * Save basic data (required / unaffected by relocation).
2391 */
2392 SSMR3PutStruct(pSSM, pPGM, &s_aPGMFields[0]);
2393
2394 for (i=0;i<pVM->cCPUs;i++)
2395 {
2396 PVMCPU pVCpu = &pVM->aCpus[i];
2397
2398 SSMR3PutStruct(pSSM, &pVCpu->pgm.s, &s_aPGMCpuFields[0]);
2399 }
2400
2401 /*
2402 * The guest mappings.
2403 */
2404 i = 0;
2405 for (PPGMMAPPING pMapping = pPGM->pMappingsR3; pMapping; pMapping = pMapping->pNextR3, i++)
2406 {
2407 SSMR3PutU32( pSSM, i);
2408 SSMR3PutStrZ( pSSM, pMapping->pszDesc); /* This is the best unique id we have... */
2409 SSMR3PutGCPtr( pSSM, pMapping->GCPtr);
2410 SSMR3PutGCUIntPtr(pSSM, pMapping->cPTs);
2411 }
2412 rc = SSMR3PutU32(pSSM, ~0); /* terminator. */
2413
2414 /*
2415 * Ram ranges and the memory they describe.
2416 */
2417 i = 0;
2418 for (PPGMRAMRANGE pRam = pPGM->pRamRangesR3; pRam; pRam = pRam->pNextR3, i++)
2419 {
2420 /*
2421 * Save the ram range details.
2422 */
2423 SSMR3PutU32(pSSM, i);
2424 SSMR3PutGCPhys(pSSM, pRam->GCPhys);
2425 SSMR3PutGCPhys(pSSM, pRam->GCPhysLast);
2426 SSMR3PutGCPhys(pSSM, pRam->cb);
2427 SSMR3PutU8(pSSM, !!pRam->pvR3); /* Boolean indicating memory or not. */
2428 SSMR3PutStrZ(pSSM, pRam->pszDesc); /* This is the best unique id we have... */
2429
2430 /*
2431 * Iterate the pages, only two special case.
2432 */
2433 uint32_t const cPages = pRam->cb >> PAGE_SHIFT;
2434 for (uint32_t iPage = 0; iPage < cPages; iPage++)
2435 {
2436 RTGCPHYS GCPhysPage = pRam->GCPhys + ((RTGCPHYS)iPage << PAGE_SHIFT);
2437 PPGMPAGE pPage = &pRam->aPages[iPage];
2438 uint8_t uType = PGM_PAGE_GET_TYPE(pPage);
2439
2440 if (uType == PGMPAGETYPE_ROM_SHADOW)
2441 rc = pgmR3SaveShadowedRomPage(pVM, pSSM, pPage, GCPhysPage, pRam);
2442 else if (uType == PGMPAGETYPE_MMIO2_ALIAS_MMIO)
2443 {
2444 /* MMIO2 alias -> MMIO; the device will just have to deal with this. */
2445 SSMR3PutU8(pSSM, PGMPAGETYPE_MMIO);
2446 rc = SSMR3PutU8(pSSM, 0 /* ZERO */);
2447 }
2448 else
2449 {
2450 SSMR3PutU8(pSSM, uType);
2451 rc = pgmR3SavePage(pVM, pSSM, pPage, GCPhysPage, pRam);
2452 }
2453 if (RT_FAILURE(rc))
2454 break;
2455 }
2456 if (RT_FAILURE(rc))
2457 break;
2458 }
2459
2460 pgmUnlock(pVM);
2461 return SSMR3PutU32(pSSM, ~0); /* terminator. */
2462}
2463
2464
2465/**
2466 * Load an ignored page.
2467 *
2468 * @returns VBox status code.
2469 * @param pSSM The saved state handle.
2470 */
2471static int pgmR3LoadPageToDevNull(PSSMHANDLE pSSM)
2472{
2473 uint8_t abPage[PAGE_SIZE];
2474 return SSMR3GetMem(pSSM, &abPage[0], sizeof(abPage));
2475}
2476
2477
2478/**
2479 * Loads a page without any bits in the saved state, i.e. making sure it's
2480 * really zero.
2481 *
2482 * @returns VBox status code.
2483 * @param pVM The VM handle.
2484 * @param uType The page type or PGMPAGETYPE_INVALID (old saved
2485 * state).
2486 * @param pPage The guest page tracking structure.
2487 * @param GCPhys The page address.
2488 * @param pRam The ram range (logging).
2489 */
2490static int pgmR3LoadPageZero(PVM pVM, uint8_t uType, PPGMPAGE pPage, RTGCPHYS GCPhys, PPGMRAMRANGE pRam)
2491{
2492 if ( PGM_PAGE_GET_TYPE(pPage) != uType
2493 && uType != PGMPAGETYPE_INVALID)
2494 return VERR_SSM_UNEXPECTED_DATA;
2495
2496 /* I think this should be sufficient. */
2497 if (!PGM_PAGE_IS_ZERO(pPage))
2498 return VERR_SSM_UNEXPECTED_DATA;
2499
2500 NOREF(pVM);
2501 NOREF(GCPhys);
2502 NOREF(pRam);
2503 return VINF_SUCCESS;
2504}
2505
2506
2507/**
2508 * Loads a page from the saved state.
2509 *
2510 * @returns VBox status code.
2511 * @param pVM The VM handle.
2512 * @param pSSM The SSM handle.
2513 * @param uType The page type or PGMPAGETYEP_INVALID (old saved
2514 * state).
2515 * @param pPage The guest page tracking structure.
2516 * @param GCPhys The page address.
2517 * @param pRam The ram range (logging).
2518 */
2519static int pgmR3LoadPageBits(PVM pVM, PSSMHANDLE pSSM, uint8_t uType, PPGMPAGE pPage, RTGCPHYS GCPhys, PPGMRAMRANGE pRam)
2520{
2521 int rc;
2522
2523 /*
2524 * Match up the type, dealing with MMIO2 aliases (dropped).
2525 */
2526 AssertLogRelMsgReturn( PGM_PAGE_GET_TYPE(pPage) == uType
2527 || uType == PGMPAGETYPE_INVALID,
2528 ("pPage=%R[pgmpage] GCPhys=%#x %s\n", pPage, GCPhys, pRam->pszDesc),
2529 VERR_SSM_UNEXPECTED_DATA);
2530
2531 /*
2532 * Load the page.
2533 */
2534 void *pvPage;
2535 rc = pgmPhysGCPhys2CCPtrInternal(pVM, pPage, GCPhys, &pvPage);
2536 if (RT_SUCCESS(rc))
2537 rc = SSMR3GetMem(pSSM, pvPage, PAGE_SIZE);
2538
2539 return rc;
2540}
2541
2542
2543/**
2544 * Loads a page (counter part to pgmR3SavePage).
2545 *
2546 * @returns VBox status code, fully bitched errors.
2547 * @param pVM The VM handle.
2548 * @param pSSM The SSM handle.
2549 * @param uType The page type.
2550 * @param pPage The page.
2551 * @param GCPhys The page address.
2552 * @param pRam The RAM range (for error messages).
2553 */
2554static int pgmR3LoadPage(PVM pVM, PSSMHANDLE pSSM, uint8_t uType, PPGMPAGE pPage, RTGCPHYS GCPhys, PPGMRAMRANGE pRam)
2555{
2556 uint8_t uState;
2557 int rc = SSMR3GetU8(pSSM, &uState);
2558 AssertLogRelMsgRCReturn(rc, ("pPage=%R[pgmpage] GCPhys=%#x %s rc=%Rrc\n", pPage, GCPhys, pRam->pszDesc, rc), rc);
2559 if (uState == 0 /* zero */)
2560 rc = pgmR3LoadPageZero(pVM, uType, pPage, GCPhys, pRam);
2561 else if (uState == 1)
2562 rc = pgmR3LoadPageBits(pVM, pSSM, uType, pPage, GCPhys, pRam);
2563 else
2564 rc = VERR_INTERNAL_ERROR;
2565 AssertLogRelMsgRCReturn(rc, ("pPage=%R[pgmpage] uState=%d uType=%d GCPhys=%RGp %s rc=%Rrc\n",
2566 pPage, uState, uType, GCPhys, pRam->pszDesc, rc),
2567 rc);
2568 return VINF_SUCCESS;
2569}
2570
2571
2572/**
2573 * Loads a shadowed ROM page.
2574 *
2575 * @returns VBox status code, errors are fully bitched.
2576 * @param pVM The VM handle.
2577 * @param pSSM The saved state handle.
2578 * @param pPage The page.
2579 * @param GCPhys The page address.
2580 * @param pRam The RAM range (for error messages).
2581 */
2582static int pgmR3LoadShadowedRomPage(PVM pVM, PSSMHANDLE pSSM, PPGMPAGE pPage, RTGCPHYS GCPhys, PPGMRAMRANGE pRam)
2583{
2584 /*
2585 * Load and set the protection first, then load the two pages, the first
2586 * one is the active the other is the passive.
2587 */
2588 PPGMROMPAGE pRomPage = pgmR3GetRomPage(pVM, GCPhys);
2589 AssertLogRelMsgReturn(pRomPage, ("GCPhys=%RGp %s\n", GCPhys, pRam->pszDesc), VERR_INTERNAL_ERROR);
2590
2591 uint8_t uProt;
2592 int rc = SSMR3GetU8(pSSM, &uProt);
2593 AssertLogRelMsgRCReturn(rc, ("pPage=%R[pgmpage] GCPhys=%#x %s\n", pPage, GCPhys, pRam->pszDesc), rc);
2594 PGMROMPROT enmProt = (PGMROMPROT)uProt;
2595 AssertLogRelMsgReturn( enmProt >= PGMROMPROT_INVALID
2596 && enmProt < PGMROMPROT_END,
2597 ("enmProt=%d pPage=%R[pgmpage] GCPhys=%#x %s\n", enmProt, pPage, GCPhys, pRam->pszDesc),
2598 VERR_SSM_UNEXPECTED_DATA);
2599
2600 if (pRomPage->enmProt != enmProt)
2601 {
2602 rc = PGMR3PhysRomProtect(pVM, GCPhys, PAGE_SIZE, enmProt);
2603 AssertLogRelRCReturn(rc, rc);
2604 AssertLogRelReturn(pRomPage->enmProt == enmProt, VERR_INTERNAL_ERROR);
2605 }
2606
2607 PPGMPAGE pPageActive = PGMROMPROT_IS_ROM(enmProt) ? &pRomPage->Virgin : &pRomPage->Shadow;
2608 PPGMPAGE pPagePassive = PGMROMPROT_IS_ROM(enmProt) ? &pRomPage->Shadow : &pRomPage->Virgin;
2609 uint8_t u8ActiveType = PGMROMPROT_IS_ROM(enmProt) ? PGMPAGETYPE_ROM : PGMPAGETYPE_ROM_SHADOW;
2610 uint8_t u8PassiveType= PGMROMPROT_IS_ROM(enmProt) ? PGMPAGETYPE_ROM_SHADOW : PGMPAGETYPE_ROM;
2611
2612 rc = pgmR3LoadPage(pVM, pSSM, u8ActiveType, pPage, GCPhys, pRam);
2613 if (RT_SUCCESS(rc))
2614 {
2615 *pPageActive = *pPage;
2616 rc = pgmR3LoadPage(pVM, pSSM, u8PassiveType, pPagePassive, GCPhys, pRam);
2617 }
2618 return rc;
2619}
2620
2621
2622/**
2623 * Worker for pgmR3Load.
2624 *
2625 * @returns VBox status code.
2626 *
2627 * @param pVM The VM handle.
2628 * @param pSSM The SSM handle.
2629 * @param u32Version The saved state version.
2630 */
2631static int pgmR3LoadLocked(PVM pVM, PSSMHANDLE pSSM, uint32_t u32Version)
2632{
2633 int rc;
2634 PPGM pPGM = &pVM->pgm.s;
2635 uint32_t u32Sep;
2636
2637 /*
2638 * Load basic data (required / unaffected by relocation).
2639 */
2640 if (u32Version >= PGM_SAVED_STATE_VERSION)
2641 {
2642 rc = SSMR3GetStruct(pSSM, pPGM, &s_aPGMFields[0]);
2643 AssertLogRelRCReturn(rc, rc);
2644
2645 for (unsigned i=0;i<pVM->cCPUs;i++)
2646 {
2647 PVMCPU pVCpu = &pVM->aCpus[i];
2648
2649 rc = SSMR3GetStruct(pSSM, &pVCpu->pgm.s, &s_aPGMCpuFields[0]);
2650 AssertLogRelRCReturn(rc, rc);
2651 }
2652 }
2653 else
2654 if (u32Version >= PGM_SAVED_STATE_VERSION_RR_DESC)
2655 {
2656 PGMOLD pgmOld;
2657
2658 AssertRelease(pVM->cCPUs == 1);
2659
2660 rc = SSMR3GetStruct(pSSM, &pgmOld, &s_aPGMFields_Old[0]);
2661 AssertLogRelRCReturn(rc, rc);
2662
2663 pPGM->fMappingsFixed = pgmOld.fMappingsFixed;
2664 pPGM->GCPtrMappingFixed = pgmOld.GCPtrMappingFixed;
2665 pPGM->cbMappingFixed = pgmOld.cbMappingFixed;
2666
2667 pVM->aCpus[0].pgm.s.fA20Enabled = pgmOld.fA20Enabled;
2668 pVM->aCpus[0].pgm.s.GCPhysA20Mask = pgmOld.GCPhysA20Mask;
2669 pVM->aCpus[0].pgm.s.enmGuestMode = pgmOld.enmGuestMode;
2670 }
2671 else
2672 {
2673 AssertRelease(pVM->cCPUs == 1);
2674
2675 SSMR3GetBool(pSSM, &pPGM->fMappingsFixed);
2676 SSMR3GetGCPtr(pSSM, &pPGM->GCPtrMappingFixed);
2677 SSMR3GetU32(pSSM, &pPGM->cbMappingFixed);
2678
2679 uint32_t cbRamSizeIgnored;
2680 rc = SSMR3GetU32(pSSM, &cbRamSizeIgnored);
2681 if (RT_FAILURE(rc))
2682 return rc;
2683 SSMR3GetGCPhys(pSSM, &pVM->aCpus[0].pgm.s.GCPhysA20Mask);
2684
2685 uint32_t u32 = 0;
2686 SSMR3GetUInt(pSSM, &u32);
2687 pVM->aCpus[0].pgm.s.fA20Enabled = !!u32;
2688 SSMR3GetUInt(pSSM, &pVM->aCpus[0].pgm.s.fSyncFlags);
2689 RTUINT uGuestMode;
2690 SSMR3GetUInt(pSSM, &uGuestMode);
2691 pVM->aCpus[0].pgm.s.enmGuestMode = (PGMMODE)uGuestMode;
2692
2693 /* check separator. */
2694 SSMR3GetU32(pSSM, &u32Sep);
2695 if (RT_FAILURE(rc))
2696 return rc;
2697 if (u32Sep != (uint32_t)~0)
2698 {
2699 AssertMsgFailed(("u32Sep=%#x (first)\n", u32Sep));
2700 return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
2701 }
2702 }
2703
2704 /*
2705 * The guest mappings.
2706 */
2707 uint32_t i = 0;
2708 for (;; i++)
2709 {
2710 /* Check the seqence number / separator. */
2711 rc = SSMR3GetU32(pSSM, &u32Sep);
2712 if (RT_FAILURE(rc))
2713 return rc;
2714 if (u32Sep == ~0U)
2715 break;
2716 if (u32Sep != i)
2717 {
2718 AssertMsgFailed(("u32Sep=%#x (last)\n", u32Sep));
2719 return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
2720 }
2721
2722 /* get the mapping details. */
2723 char szDesc[256];
2724 szDesc[0] = '\0';
2725 rc = SSMR3GetStrZ(pSSM, szDesc, sizeof(szDesc));
2726 if (RT_FAILURE(rc))
2727 return rc;
2728 RTGCPTR GCPtr;
2729 SSMR3GetGCPtr(pSSM, &GCPtr);
2730 RTGCPTR cPTs;
2731 rc = SSMR3GetGCUIntPtr(pSSM, &cPTs);
2732 if (RT_FAILURE(rc))
2733 return rc;
2734
2735 /* find matching range. */
2736 PPGMMAPPING pMapping;
2737 for (pMapping = pPGM->pMappingsR3; pMapping; pMapping = pMapping->pNextR3)
2738 if ( pMapping->cPTs == cPTs
2739 && !strcmp(pMapping->pszDesc, szDesc))
2740 break;
2741 AssertLogRelMsgReturn(pMapping, ("Couldn't find mapping: cPTs=%#x szDesc=%s (GCPtr=%RGv)\n",
2742 cPTs, szDesc, GCPtr),
2743 VERR_SSM_LOAD_CONFIG_MISMATCH);
2744
2745 /* relocate it. */
2746 if (pMapping->GCPtr != GCPtr)
2747 {
2748 AssertMsg((GCPtr >> X86_PD_SHIFT << X86_PD_SHIFT) == GCPtr, ("GCPtr=%RGv\n", GCPtr));
2749 pgmR3MapRelocate(pVM, pMapping, pMapping->GCPtr, GCPtr);
2750 }
2751 else
2752 Log(("pgmR3Load: '%s' needed no relocation (%RGv)\n", szDesc, GCPtr));
2753 }
2754
2755 /*
2756 * Ram range flags and bits.
2757 */
2758 i = 0;
2759 for (PPGMRAMRANGE pRam = pPGM->pRamRangesR3; pRam; pRam = pRam->pNextR3, i++)
2760 {
2761 /** @todo MMIO ranges may move (PCI reconfig), we currently assume they don't. */
2762
2763 /* Check the seqence number / separator. */
2764 rc = SSMR3GetU32(pSSM, &u32Sep);
2765 if (RT_FAILURE(rc))
2766 return rc;
2767 if (u32Sep == ~0U)
2768 break;
2769 if (u32Sep != i)
2770 {
2771 AssertMsgFailed(("u32Sep=%#x (last)\n", u32Sep));
2772 return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
2773 }
2774
2775 /* Get the range details. */
2776 RTGCPHYS GCPhys;
2777 SSMR3GetGCPhys(pSSM, &GCPhys);
2778 RTGCPHYS GCPhysLast;
2779 SSMR3GetGCPhys(pSSM, &GCPhysLast);
2780 RTGCPHYS cb;
2781 SSMR3GetGCPhys(pSSM, &cb);
2782 uint8_t fHaveBits;
2783 rc = SSMR3GetU8(pSSM, &fHaveBits);
2784 if (RT_FAILURE(rc))
2785 return rc;
2786 if (fHaveBits & ~1)
2787 {
2788 AssertMsgFailed(("u32Sep=%#x (last)\n", u32Sep));
2789 return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
2790 }
2791 size_t cchDesc = 0;
2792 char szDesc[256];
2793 szDesc[0] = '\0';
2794 if (u32Version >= PGM_SAVED_STATE_VERSION_RR_DESC)
2795 {
2796 rc = SSMR3GetStrZ(pSSM, szDesc, sizeof(szDesc));
2797 if (RT_FAILURE(rc))
2798 return rc;
2799 /* Since we've modified the description strings in r45878, only compare
2800 them if the saved state is more recent. */
2801 if (u32Version != PGM_SAVED_STATE_VERSION_RR_DESC)
2802 cchDesc = strlen(szDesc);
2803 }
2804
2805 /*
2806 * Match it up with the current range.
2807 *
2808 * Note there is a hack for dealing with the high BIOS mapping
2809 * in the old saved state format, this means we might not have
2810 * a 1:1 match on success.
2811 */
2812 if ( ( GCPhys != pRam->GCPhys
2813 || GCPhysLast != pRam->GCPhysLast
2814 || cb != pRam->cb
2815 || ( cchDesc
2816 && strcmp(szDesc, pRam->pszDesc)) )
2817 /* Hack for PDMDevHlpPhysReserve(pDevIns, 0xfff80000, 0x80000, "High ROM Region"); */
2818 && ( u32Version != PGM_SAVED_STATE_VERSION_OLD_PHYS_CODE
2819 || GCPhys != UINT32_C(0xfff80000)
2820 || GCPhysLast != UINT32_C(0xffffffff)
2821 || pRam->GCPhysLast != GCPhysLast
2822 || pRam->GCPhys < GCPhys
2823 || !fHaveBits)
2824 )
2825 {
2826 LogRel(("Ram range: %RGp-%RGp %RGp bytes %s %s\n"
2827 "State : %RGp-%RGp %RGp bytes %s %s\n",
2828 pRam->GCPhys, pRam->GCPhysLast, pRam->cb, pRam->pvR3 ? "bits" : "nobits", pRam->pszDesc,
2829 GCPhys, GCPhysLast, cb, fHaveBits ? "bits" : "nobits", szDesc));
2830 /*
2831 * If we're loading a state for debugging purpose, don't make a fuss if
2832 * the MMIO and ROM stuff isn't 100% right, just skip the mismatches.
2833 */
2834 if ( SSMR3HandleGetAfter(pSSM) != SSMAFTER_DEBUG_IT
2835 || GCPhys < 8 * _1M)
2836 AssertFailedReturn(VERR_SSM_LOAD_CONFIG_MISMATCH);
2837
2838 AssertMsgFailed(("debug skipping not implemented, sorry\n"));
2839 continue;
2840 }
2841
2842 uint32_t cPages = (GCPhysLast - GCPhys + 1) >> PAGE_SHIFT;
2843 if (u32Version >= PGM_SAVED_STATE_VERSION_RR_DESC)
2844 {
2845 /*
2846 * Load the pages one by one.
2847 */
2848 for (uint32_t iPage = 0; iPage < cPages; iPage++)
2849 {
2850 RTGCPHYS const GCPhysPage = ((RTGCPHYS)iPage << PAGE_SHIFT) + pRam->GCPhys;
2851 PPGMPAGE pPage = &pRam->aPages[iPage];
2852 uint8_t uType;
2853 rc = SSMR3GetU8(pSSM, &uType);
2854 AssertLogRelMsgRCReturn(rc, ("pPage=%R[pgmpage] iPage=%#x GCPhysPage=%#x %s\n", pPage, iPage, GCPhysPage, pRam->pszDesc), rc);
2855 if (uType == PGMPAGETYPE_ROM_SHADOW)
2856 rc = pgmR3LoadShadowedRomPage(pVM, pSSM, pPage, GCPhysPage, pRam);
2857 else
2858 rc = pgmR3LoadPage(pVM, pSSM, uType, pPage, GCPhysPage, pRam);
2859 AssertLogRelMsgRCReturn(rc, ("rc=%Rrc iPage=%#x GCPhysPage=%#x %s\n", rc, iPage, GCPhysPage, pRam->pszDesc), rc);
2860 }
2861 }
2862 else
2863 {
2864 /*
2865 * Old format.
2866 */
2867 AssertLogRelReturn(!pVM->pgm.s.fRamPreAlloc, VERR_NOT_SUPPORTED); /* can't be detected. */
2868
2869 /* Of the page flags, pick up MMIO2 and ROM/RESERVED for the !fHaveBits case.
2870 The rest is generally irrelevant and wrong since the stuff have to match registrations. */
2871 uint32_t fFlags = 0;
2872 for (uint32_t iPage = 0; iPage < cPages; iPage++)
2873 {
2874 uint16_t u16Flags;
2875 rc = SSMR3GetU16(pSSM, &u16Flags);
2876 AssertLogRelMsgRCReturn(rc, ("rc=%Rrc iPage=%#x GCPhys=%#x %s\n", rc, iPage, pRam->GCPhys, pRam->pszDesc), rc);
2877 fFlags |= u16Flags;
2878 }
2879
2880 /* Load the bits */
2881 if ( !fHaveBits
2882 && GCPhysLast < UINT32_C(0xe0000000))
2883 {
2884 /*
2885 * Dynamic chunks.
2886 */
2887 const uint32_t cPagesInChunk = (1*1024*1024) >> PAGE_SHIFT;
2888 AssertLogRelMsgReturn(cPages % cPagesInChunk == 0,
2889 ("cPages=%#x cPagesInChunk=%#x\n", cPages, cPagesInChunk, pRam->GCPhys, pRam->pszDesc),
2890 VERR_SSM_DATA_UNIT_FORMAT_CHANGED);
2891
2892 for (uint32_t iPage = 0; iPage < cPages; /* incremented by inner loop */ )
2893 {
2894 uint8_t fPresent;
2895 rc = SSMR3GetU8(pSSM, &fPresent);
2896 AssertLogRelMsgRCReturn(rc, ("rc=%Rrc iPage=%#x GCPhys=%#x %s\n", rc, iPage, pRam->GCPhys, pRam->pszDesc), rc);
2897 AssertLogRelMsgReturn(fPresent == (uint8_t)true || fPresent == (uint8_t)false,
2898 ("fPresent=%#x iPage=%#x GCPhys=%#x %s\n", fPresent, iPage, pRam->GCPhys, pRam->pszDesc),
2899 VERR_SSM_DATA_UNIT_FORMAT_CHANGED);
2900
2901 for (uint32_t iChunkPage = 0; iChunkPage < cPagesInChunk; iChunkPage++, iPage++)
2902 {
2903 RTGCPHYS const GCPhysPage = ((RTGCPHYS)iPage << PAGE_SHIFT) + pRam->GCPhys;
2904 PPGMPAGE pPage = &pRam->aPages[iPage];
2905 if (fPresent)
2906 {
2907 if (PGM_PAGE_GET_TYPE(pPage) == PGMPAGETYPE_MMIO)
2908 rc = pgmR3LoadPageToDevNull(pSSM);
2909 else
2910 rc = pgmR3LoadPageBits(pVM, pSSM, PGMPAGETYPE_INVALID, pPage, GCPhysPage, pRam);
2911 }
2912 else
2913 rc = pgmR3LoadPageZero(pVM, PGMPAGETYPE_INVALID, pPage, GCPhysPage, pRam);
2914 AssertLogRelMsgRCReturn(rc, ("rc=%Rrc iPage=%#x GCPhysPage=%#x %s\n", rc, iPage, GCPhysPage, pRam->pszDesc), rc);
2915 }
2916 }
2917 }
2918 else if (pRam->pvR3)
2919 {
2920 /*
2921 * MMIO2.
2922 */
2923 AssertLogRelMsgReturn((fFlags & 0x0f) == RT_BIT(3) /*MM_RAM_FLAGS_MMIO2*/,
2924 ("fFlags=%#x GCPhys=%#x %s\n", fFlags, pRam->GCPhys, pRam->pszDesc),
2925 VERR_SSM_DATA_UNIT_FORMAT_CHANGED);
2926 AssertLogRelMsgReturn(pRam->pvR3,
2927 ("GCPhys=%#x %s\n", pRam->GCPhys, pRam->pszDesc),
2928 VERR_SSM_DATA_UNIT_FORMAT_CHANGED);
2929
2930 rc = SSMR3GetMem(pSSM, pRam->pvR3, pRam->cb);
2931 AssertLogRelMsgRCReturn(rc, ("GCPhys=%#x %s\n", pRam->GCPhys, pRam->pszDesc), rc);
2932 }
2933 else if (GCPhysLast < UINT32_C(0xfff80000))
2934 {
2935 /*
2936 * PCI MMIO, no pages saved.
2937 */
2938 }
2939 else
2940 {
2941 /*
2942 * Load the 0xfff80000..0xffffffff BIOS range.
2943 * It starts with X reserved pages that we have to skip over since
2944 * the RAMRANGE create by the new code won't include those.
2945 */
2946 AssertLogRelMsgReturn( !(fFlags & RT_BIT(3) /*MM_RAM_FLAGS_MMIO2*/)
2947 && (fFlags & RT_BIT(0) /*MM_RAM_FLAGS_RESERVED*/),
2948 ("fFlags=%#x GCPhys=%#x %s\n", fFlags, pRam->GCPhys, pRam->pszDesc),
2949 VERR_SSM_DATA_UNIT_FORMAT_CHANGED);
2950 AssertLogRelMsgReturn(GCPhys == UINT32_C(0xfff80000),
2951 ("GCPhys=%RGp pRamRange{GCPhys=%#x %s}\n", GCPhys, pRam->GCPhys, pRam->pszDesc),
2952 VERR_SSM_DATA_UNIT_FORMAT_CHANGED);
2953
2954 /* Skip wasted reserved pages before the ROM. */
2955 while (GCPhys < pRam->GCPhys)
2956 {
2957 rc = pgmR3LoadPageToDevNull(pSSM);
2958 GCPhys += PAGE_SIZE;
2959 }
2960
2961 /* Load the bios pages. */
2962 cPages = pRam->cb >> PAGE_SHIFT;
2963 for (uint32_t iPage = 0; iPage < cPages; iPage++)
2964 {
2965 RTGCPHYS const GCPhysPage = ((RTGCPHYS)iPage << PAGE_SHIFT) + pRam->GCPhys;
2966 PPGMPAGE pPage = &pRam->aPages[iPage];
2967
2968 AssertLogRelMsgReturn(PGM_PAGE_GET_TYPE(pPage) == PGMPAGETYPE_ROM,
2969 ("GCPhys=%RGp pPage=%R[pgmpage]\n", GCPhys, GCPhys),
2970 VERR_SSM_DATA_UNIT_FORMAT_CHANGED);
2971 rc = pgmR3LoadPageBits(pVM, pSSM, PGMPAGETYPE_ROM, pPage, GCPhysPage, pRam);
2972 AssertLogRelMsgRCReturn(rc, ("rc=%Rrc iPage=%#x GCPhys=%#x %s\n", rc, iPage, pRam->GCPhys, pRam->pszDesc), rc);
2973 }
2974 }
2975 }
2976 }
2977
2978 return rc;
2979}
2980
2981
2982/**
2983 * Execute state load operation.
2984 *
2985 * @returns VBox status code.
2986 * @param pVM VM Handle.
2987 * @param pSSM SSM operation handle.
2988 * @param u32Version Data layout version.
2989 */
2990static DECLCALLBACK(int) pgmR3Load(PVM pVM, PSSMHANDLE pSSM, uint32_t u32Version)
2991{
2992 int rc;
2993 PPGM pPGM = &pVM->pgm.s;
2994
2995 /*
2996 * Validate version.
2997 */
2998 if ( u32Version != PGM_SAVED_STATE_VERSION
2999 && u32Version != PGM_SAVED_STATE_VERSION_2_2_2
3000 && u32Version != PGM_SAVED_STATE_VERSION_RR_DESC
3001 && u32Version != PGM_SAVED_STATE_VERSION_OLD_PHYS_CODE)
3002 {
3003 AssertMsgFailed(("pgmR3Load: Invalid version u32Version=%d (current %d)!\n", u32Version, PGM_SAVED_STATE_VERSION));
3004 return VERR_SSM_UNSUPPORTED_DATA_UNIT_VERSION;
3005 }
3006
3007 /*
3008 * Call the reset function to make sure all the memory is cleared.
3009 */
3010 PGMR3Reset(pVM);
3011
3012 /*
3013 * Do the loading while owning the lock because a bunch of the functions
3014 * we're using requires this.
3015 */
3016 pgmLock(pVM);
3017 rc = pgmR3LoadLocked(pVM, pSSM, u32Version);
3018 pgmUnlock(pVM);
3019 if (RT_SUCCESS(rc))
3020 {
3021 /*
3022 * We require a full resync now.
3023 */
3024 for (unsigned i=0;i<pVM->cCPUs;i++)
3025 {
3026 PVMCPU pVCpu = &pVM->aCpus[i];
3027 VMCPU_FF_SET(pVCpu, VMCPU_FF_PGM_SYNC_CR3_NON_GLOBAL);
3028 VMCPU_FF_SET(pVCpu, VMCPU_FF_PGM_SYNC_CR3);
3029
3030 pVCpu->pgm.s.fSyncFlags |= PGM_SYNC_UPDATE_PAGE_BIT_VIRTUAL;
3031 }
3032
3033 pPGM->fPhysCacheFlushPending = true;
3034 pgmR3HandlerPhysicalUpdateAll(pVM);
3035
3036 for (unsigned i=0;i<pVM->cCPUs;i++)
3037 {
3038 PVMCPU pVCpu = &pVM->aCpus[i];
3039
3040 /*
3041 * Change the paging mode.
3042 */
3043 rc = PGMR3ChangeMode(pVM, pVCpu, pVCpu->pgm.s.enmGuestMode);
3044
3045 /* Restore pVM->pgm.s.GCPhysCR3. */
3046 Assert(pVCpu->pgm.s.GCPhysCR3 == NIL_RTGCPHYS);
3047 RTGCPHYS GCPhysCR3 = CPUMGetGuestCR3(pVCpu);
3048 if ( pVCpu->pgm.s.enmGuestMode == PGMMODE_PAE
3049 || pVCpu->pgm.s.enmGuestMode == PGMMODE_PAE_NX
3050 || pVCpu->pgm.s.enmGuestMode == PGMMODE_AMD64
3051 || pVCpu->pgm.s.enmGuestMode == PGMMODE_AMD64_NX)
3052 GCPhysCR3 = (GCPhysCR3 & X86_CR3_PAE_PAGE_MASK);
3053 else
3054 GCPhysCR3 = (GCPhysCR3 & X86_CR3_PAGE_MASK);
3055 pVCpu->pgm.s.GCPhysCR3 = GCPhysCR3;
3056 }
3057 }
3058
3059 return rc;
3060}
3061
3062
3063/**
3064 * Show paging mode.
3065 *
3066 * @param pVM VM Handle.
3067 * @param pHlp The info helpers.
3068 * @param pszArgs "all" (default), "guest", "shadow" or "host".
3069 */
3070static DECLCALLBACK(void) pgmR3InfoMode(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs)
3071{
3072 /* digest argument. */
3073 bool fGuest, fShadow, fHost;
3074 if (pszArgs)
3075 pszArgs = RTStrStripL(pszArgs);
3076 if (!pszArgs || !*pszArgs || strstr(pszArgs, "all"))
3077 fShadow = fHost = fGuest = true;
3078 else
3079 {
3080 fShadow = fHost = fGuest = false;
3081 if (strstr(pszArgs, "guest"))
3082 fGuest = true;
3083 if (strstr(pszArgs, "shadow"))
3084 fShadow = true;
3085 if (strstr(pszArgs, "host"))
3086 fHost = true;
3087 }
3088
3089 /** @todo SMP support! */
3090 /* print info. */
3091 if (fGuest)
3092 pHlp->pfnPrintf(pHlp, "Guest paging mode: %s, changed %RU64 times, A20 %s\n",
3093 PGMGetModeName(pVM->aCpus[0].pgm.s.enmGuestMode), pVM->aCpus[0].pgm.s.cGuestModeChanges.c,
3094 pVM->aCpus[0].pgm.s.fA20Enabled ? "enabled" : "disabled");
3095 if (fShadow)
3096 pHlp->pfnPrintf(pHlp, "Shadow paging mode: %s\n", PGMGetModeName(pVM->aCpus[0].pgm.s.enmShadowMode));
3097 if (fHost)
3098 {
3099 const char *psz;
3100 switch (pVM->pgm.s.enmHostMode)
3101 {
3102 case SUPPAGINGMODE_INVALID: psz = "invalid"; break;
3103 case SUPPAGINGMODE_32_BIT: psz = "32-bit"; break;
3104 case SUPPAGINGMODE_32_BIT_GLOBAL: psz = "32-bit+G"; break;
3105 case SUPPAGINGMODE_PAE: psz = "PAE"; break;
3106 case SUPPAGINGMODE_PAE_GLOBAL: psz = "PAE+G"; break;
3107 case SUPPAGINGMODE_PAE_NX: psz = "PAE+NX"; break;
3108 case SUPPAGINGMODE_PAE_GLOBAL_NX: psz = "PAE+G+NX"; break;
3109 case SUPPAGINGMODE_AMD64: psz = "AMD64"; break;
3110 case SUPPAGINGMODE_AMD64_GLOBAL: psz = "AMD64+G"; break;
3111 case SUPPAGINGMODE_AMD64_NX: psz = "AMD64+NX"; break;
3112 case SUPPAGINGMODE_AMD64_GLOBAL_NX: psz = "AMD64+G+NX"; break;
3113 default: psz = "unknown"; break;
3114 }
3115 pHlp->pfnPrintf(pHlp, "Host paging mode: %s\n", psz);
3116 }
3117}
3118
3119
3120/**
3121 * Dump registered MMIO ranges to the log.
3122 *
3123 * @param pVM VM Handle.
3124 * @param pHlp The info helpers.
3125 * @param pszArgs Arguments, ignored.
3126 */
3127static DECLCALLBACK(void) pgmR3PhysInfo(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs)
3128{
3129 NOREF(pszArgs);
3130 pHlp->pfnPrintf(pHlp,
3131 "RAM ranges (pVM=%p)\n"
3132 "%.*s %.*s\n",
3133 pVM,
3134 sizeof(RTGCPHYS) * 4 + 1, "GC Phys Range ",
3135 sizeof(RTHCPTR) * 2, "pvHC ");
3136
3137 for (PPGMRAMRANGE pCur = pVM->pgm.s.pRamRangesR3; pCur; pCur = pCur->pNextR3)
3138 pHlp->pfnPrintf(pHlp,
3139 "%RGp-%RGp %RHv %s\n",
3140 pCur->GCPhys,
3141 pCur->GCPhysLast,
3142 pCur->pvR3,
3143 pCur->pszDesc);
3144}
3145
3146/**
3147 * Dump the page directory to the log.
3148 *
3149 * @param pVM VM Handle.
3150 * @param pHlp The info helpers.
3151 * @param pszArgs Arguments, ignored.
3152 */
3153static DECLCALLBACK(void) pgmR3InfoCr3(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs)
3154{
3155 /** @todo SMP support!! */
3156 PVMCPU pVCpu = &pVM->aCpus[0];
3157
3158/** @todo fix this! Convert the PGMR3DumpHierarchyHC functions to do guest stuff. */
3159 /* Big pages supported? */
3160 const bool fPSE = !!(CPUMGetGuestCR4(pVCpu) & X86_CR4_PSE);
3161
3162 /* Global pages supported? */
3163 const bool fPGE = !!(CPUMGetGuestCR4(pVCpu) & X86_CR4_PGE);
3164
3165 NOREF(pszArgs);
3166
3167 /*
3168 * Get page directory addresses.
3169 */
3170 PX86PD pPDSrc = pgmGstGet32bitPDPtr(&pVCpu->pgm.s);
3171 Assert(pPDSrc);
3172 Assert(PGMPhysGCPhys2R3PtrAssert(pVM, (RTGCPHYS)(CPUMGetGuestCR3(pVCpu) & X86_CR3_PAGE_MASK), sizeof(*pPDSrc)) == pPDSrc);
3173
3174 /*
3175 * Iterate the page directory.
3176 */
3177 for (unsigned iPD = 0; iPD < RT_ELEMENTS(pPDSrc->a); iPD++)
3178 {
3179 X86PDE PdeSrc = pPDSrc->a[iPD];
3180 if (PdeSrc.n.u1Present)
3181 {
3182 if (PdeSrc.b.u1Size && fPSE)
3183 pHlp->pfnPrintf(pHlp,
3184 "%04X - %RGp P=%d U=%d RW=%d G=%d - BIG\n",
3185 iPD,
3186 pgmGstGet4MBPhysPage(&pVM->pgm.s, PdeSrc),
3187 PdeSrc.b.u1Present, PdeSrc.b.u1User, PdeSrc.b.u1Write, PdeSrc.b.u1Global && fPGE);
3188 else
3189 pHlp->pfnPrintf(pHlp,
3190 "%04X - %RGp P=%d U=%d RW=%d [G=%d]\n",
3191 iPD,
3192 (RTGCPHYS)(PdeSrc.u & X86_PDE_PG_MASK),
3193 PdeSrc.n.u1Present, PdeSrc.n.u1User, PdeSrc.n.u1Write, PdeSrc.b.u1Global && fPGE);
3194 }
3195 }
3196}
3197
3198
3199/**
3200 * Service a VMMCALLHOST_PGM_LOCK call.
3201 *
3202 * @returns VBox status code.
3203 * @param pVM The VM handle.
3204 */
3205VMMR3DECL(int) PGMR3LockCall(PVM pVM)
3206{
3207 int rc = PDMR3CritSectEnterEx(&pVM->pgm.s.CritSect, true /* fHostCall */);
3208 AssertRC(rc);
3209 return rc;
3210}
3211
3212
3213/**
3214 * Converts a PGMMODE value to a PGM_TYPE_* \#define.
3215 *
3216 * @returns PGM_TYPE_*.
3217 * @param pgmMode The mode value to convert.
3218 */
3219DECLINLINE(unsigned) pgmModeToType(PGMMODE pgmMode)
3220{
3221 switch (pgmMode)
3222 {
3223 case PGMMODE_REAL: return PGM_TYPE_REAL;
3224 case PGMMODE_PROTECTED: return PGM_TYPE_PROT;
3225 case PGMMODE_32_BIT: return PGM_TYPE_32BIT;
3226 case PGMMODE_PAE:
3227 case PGMMODE_PAE_NX: return PGM_TYPE_PAE;
3228 case PGMMODE_AMD64:
3229 case PGMMODE_AMD64_NX: return PGM_TYPE_AMD64;
3230 case PGMMODE_NESTED: return PGM_TYPE_NESTED;
3231 case PGMMODE_EPT: return PGM_TYPE_EPT;
3232 default:
3233 AssertFatalMsgFailed(("pgmMode=%d\n", pgmMode));
3234 }
3235}
3236
3237
3238/**
3239 * Gets the index into the paging mode data array of a SHW+GST mode.
3240 *
3241 * @returns PGM::paPagingData index.
3242 * @param uShwType The shadow paging mode type.
3243 * @param uGstType The guest paging mode type.
3244 */
3245DECLINLINE(unsigned) pgmModeDataIndex(unsigned uShwType, unsigned uGstType)
3246{
3247 Assert(uShwType >= PGM_TYPE_32BIT && uShwType <= PGM_TYPE_MAX);
3248 Assert(uGstType >= PGM_TYPE_REAL && uGstType <= PGM_TYPE_AMD64);
3249 return (uShwType - PGM_TYPE_32BIT) * (PGM_TYPE_AMD64 - PGM_TYPE_REAL + 1)
3250 + (uGstType - PGM_TYPE_REAL);
3251}
3252
3253
3254/**
3255 * Gets the index into the paging mode data array of a SHW+GST mode.
3256 *
3257 * @returns PGM::paPagingData index.
3258 * @param enmShw The shadow paging mode.
3259 * @param enmGst The guest paging mode.
3260 */
3261DECLINLINE(unsigned) pgmModeDataIndexByMode(PGMMODE enmShw, PGMMODE enmGst)
3262{
3263 Assert(enmShw >= PGMMODE_32_BIT && enmShw <= PGMMODE_MAX);
3264 Assert(enmGst > PGMMODE_INVALID && enmGst < PGMMODE_MAX);
3265 return pgmModeDataIndex(pgmModeToType(enmShw), pgmModeToType(enmGst));
3266}
3267
3268
3269/**
3270 * Calculates the max data index.
3271 * @returns The number of entries in the paging data array.
3272 */
3273DECLINLINE(unsigned) pgmModeDataMaxIndex(void)
3274{
3275 return pgmModeDataIndex(PGM_TYPE_MAX, PGM_TYPE_AMD64) + 1;
3276}
3277
3278
3279/**
3280 * Initializes the paging mode data kept in PGM::paModeData.
3281 *
3282 * @param pVM The VM handle.
3283 * @param fResolveGCAndR0 Indicate whether or not GC and Ring-0 symbols can be resolved now.
3284 * This is used early in the init process to avoid trouble with PDM
3285 * not being initialized yet.
3286 */
3287static int pgmR3ModeDataInit(PVM pVM, bool fResolveGCAndR0)
3288{
3289 PPGMMODEDATA pModeData;
3290 int rc;
3291
3292 /*
3293 * Allocate the array on the first call.
3294 */
3295 if (!pVM->pgm.s.paModeData)
3296 {
3297 pVM->pgm.s.paModeData = (PPGMMODEDATA)MMR3HeapAllocZ(pVM, MM_TAG_PGM, sizeof(PGMMODEDATA) * pgmModeDataMaxIndex());
3298 AssertReturn(pVM->pgm.s.paModeData, VERR_NO_MEMORY);
3299 }
3300
3301 /*
3302 * Initialize the array entries.
3303 */
3304 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_32BIT, PGM_TYPE_REAL)];
3305 pModeData->uShwType = PGM_TYPE_32BIT;
3306 pModeData->uGstType = PGM_TYPE_REAL;
3307 rc = PGM_SHW_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3308 rc = PGM_GST_NAME_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3309 rc = PGM_BTH_NAME_32BIT_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3310
3311 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_32BIT, PGMMODE_PROTECTED)];
3312 pModeData->uShwType = PGM_TYPE_32BIT;
3313 pModeData->uGstType = PGM_TYPE_PROT;
3314 rc = PGM_SHW_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3315 rc = PGM_GST_NAME_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3316 rc = PGM_BTH_NAME_32BIT_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3317
3318 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_32BIT, PGM_TYPE_32BIT)];
3319 pModeData->uShwType = PGM_TYPE_32BIT;
3320 pModeData->uGstType = PGM_TYPE_32BIT;
3321 rc = PGM_SHW_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3322 rc = PGM_GST_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3323 rc = PGM_BTH_NAME_32BIT_32BIT(InitData)(pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3324
3325 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_REAL)];
3326 pModeData->uShwType = PGM_TYPE_PAE;
3327 pModeData->uGstType = PGM_TYPE_REAL;
3328 rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3329 rc = PGM_GST_NAME_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3330 rc = PGM_BTH_NAME_PAE_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3331
3332 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_PROT)];
3333 pModeData->uShwType = PGM_TYPE_PAE;
3334 pModeData->uGstType = PGM_TYPE_PROT;
3335 rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3336 rc = PGM_GST_NAME_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3337 rc = PGM_BTH_NAME_PAE_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3338
3339 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_32BIT)];
3340 pModeData->uShwType = PGM_TYPE_PAE;
3341 pModeData->uGstType = PGM_TYPE_32BIT;
3342 rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3343 rc = PGM_GST_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3344 rc = PGM_BTH_NAME_PAE_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3345
3346 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_PAE)];
3347 pModeData->uShwType = PGM_TYPE_PAE;
3348 pModeData->uGstType = PGM_TYPE_PAE;
3349 rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3350 rc = PGM_GST_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3351 rc = PGM_BTH_NAME_PAE_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3352
3353#ifdef VBOX_WITH_64_BITS_GUESTS
3354 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_AMD64, PGM_TYPE_AMD64)];
3355 pModeData->uShwType = PGM_TYPE_AMD64;
3356 pModeData->uGstType = PGM_TYPE_AMD64;
3357 rc = PGM_SHW_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3358 rc = PGM_GST_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3359 rc = PGM_BTH_NAME_AMD64_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3360#endif
3361
3362 /* The nested paging mode. */
3363 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, PGM_TYPE_REAL)];
3364 pModeData->uShwType = PGM_TYPE_NESTED;
3365 pModeData->uGstType = PGM_TYPE_REAL;
3366 rc = PGM_GST_NAME_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3367 rc = PGM_BTH_NAME_NESTED_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3368
3369 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, PGMMODE_PROTECTED)];
3370 pModeData->uShwType = PGM_TYPE_NESTED;
3371 pModeData->uGstType = PGM_TYPE_PROT;
3372 rc = PGM_GST_NAME_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3373 rc = PGM_BTH_NAME_NESTED_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3374
3375 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, PGM_TYPE_32BIT)];
3376 pModeData->uShwType = PGM_TYPE_NESTED;
3377 pModeData->uGstType = PGM_TYPE_32BIT;
3378 rc = PGM_GST_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3379 rc = PGM_BTH_NAME_NESTED_32BIT(InitData)(pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3380
3381 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, PGM_TYPE_PAE)];
3382 pModeData->uShwType = PGM_TYPE_NESTED;
3383 pModeData->uGstType = PGM_TYPE_PAE;
3384 rc = PGM_GST_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3385 rc = PGM_BTH_NAME_NESTED_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3386
3387#ifdef VBOX_WITH_64_BITS_GUESTS
3388 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, PGM_TYPE_AMD64)];
3389 pModeData->uShwType = PGM_TYPE_NESTED;
3390 pModeData->uGstType = PGM_TYPE_AMD64;
3391 rc = PGM_GST_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3392 rc = PGM_BTH_NAME_NESTED_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3393#endif
3394
3395 /* The shadow part of the nested callback mode depends on the host paging mode (AMD-V only). */
3396 switch (pVM->pgm.s.enmHostMode)
3397 {
3398#if HC_ARCH_BITS == 32
3399 case SUPPAGINGMODE_32_BIT:
3400 case SUPPAGINGMODE_32_BIT_GLOBAL:
3401 for (unsigned i = PGM_TYPE_REAL; i <= PGM_TYPE_PAE; i++)
3402 {
3403 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, i)];
3404 rc = PGM_SHW_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3405 }
3406# ifdef VBOX_WITH_64_BITS_GUESTS
3407 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, PGM_TYPE_AMD64)];
3408 rc = PGM_SHW_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3409# endif
3410 break;
3411
3412 case SUPPAGINGMODE_PAE:
3413 case SUPPAGINGMODE_PAE_NX:
3414 case SUPPAGINGMODE_PAE_GLOBAL:
3415 case SUPPAGINGMODE_PAE_GLOBAL_NX:
3416 for (unsigned i = PGM_TYPE_REAL; i <= PGM_TYPE_PAE; i++)
3417 {
3418 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, i)];
3419 rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3420 }
3421# ifdef VBOX_WITH_64_BITS_GUESTS
3422 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, PGM_TYPE_AMD64)];
3423 rc = PGM_SHW_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3424# endif
3425 break;
3426#endif /* HC_ARCH_BITS == 32 */
3427
3428#if HC_ARCH_BITS == 64 || defined(RT_OS_DARWIN)
3429 case SUPPAGINGMODE_AMD64:
3430 case SUPPAGINGMODE_AMD64_GLOBAL:
3431 case SUPPAGINGMODE_AMD64_NX:
3432 case SUPPAGINGMODE_AMD64_GLOBAL_NX:
3433# ifdef VBOX_WITH_64_BITS_GUESTS
3434 for (unsigned i = PGM_TYPE_REAL; i <= PGM_TYPE_AMD64; i++)
3435# else
3436 for (unsigned i = PGM_TYPE_REAL; i <= PGM_TYPE_PAE; i++)
3437# endif
3438 {
3439 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_NESTED, i)];
3440 rc = PGM_SHW_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3441 }
3442 break;
3443#endif /* HC_ARCH_BITS == 64 || RT_OS_DARWIN */
3444
3445 default:
3446 AssertFailed();
3447 break;
3448 }
3449
3450 /* Extended paging (EPT) / Intel VT-x */
3451 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_EPT, PGM_TYPE_REAL)];
3452 pModeData->uShwType = PGM_TYPE_EPT;
3453 pModeData->uGstType = PGM_TYPE_REAL;
3454 rc = PGM_SHW_NAME_EPT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3455 rc = PGM_GST_NAME_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3456 rc = PGM_BTH_NAME_EPT_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3457
3458 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_EPT, PGM_TYPE_PROT)];
3459 pModeData->uShwType = PGM_TYPE_EPT;
3460 pModeData->uGstType = PGM_TYPE_PROT;
3461 rc = PGM_SHW_NAME_EPT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3462 rc = PGM_GST_NAME_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3463 rc = PGM_BTH_NAME_EPT_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3464
3465 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_EPT, PGM_TYPE_32BIT)];
3466 pModeData->uShwType = PGM_TYPE_EPT;
3467 pModeData->uGstType = PGM_TYPE_32BIT;
3468 rc = PGM_SHW_NAME_EPT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3469 rc = PGM_GST_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3470 rc = PGM_BTH_NAME_EPT_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3471
3472 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_EPT, PGM_TYPE_PAE)];
3473 pModeData->uShwType = PGM_TYPE_EPT;
3474 pModeData->uGstType = PGM_TYPE_PAE;
3475 rc = PGM_SHW_NAME_EPT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3476 rc = PGM_GST_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3477 rc = PGM_BTH_NAME_EPT_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3478
3479#ifdef VBOX_WITH_64_BITS_GUESTS
3480 pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_EPT, PGM_TYPE_AMD64)];
3481 pModeData->uShwType = PGM_TYPE_EPT;
3482 pModeData->uGstType = PGM_TYPE_AMD64;
3483 rc = PGM_SHW_NAME_EPT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3484 rc = PGM_GST_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3485 rc = PGM_BTH_NAME_EPT_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
3486#endif
3487 return VINF_SUCCESS;
3488}
3489
3490
3491/**
3492 * Switch to different (or relocated in the relocate case) mode data.
3493 *
3494 * @param pVM The VM handle.
3495 * @param pVCpu The VMCPU to operate on.
3496 * @param enmShw The the shadow paging mode.
3497 * @param enmGst The the guest paging mode.
3498 */
3499static void pgmR3ModeDataSwitch(PVM pVM, PVMCPU pVCpu, PGMMODE enmShw, PGMMODE enmGst)
3500{
3501 PPGMMODEDATA pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndexByMode(enmShw, enmGst)];
3502
3503 Assert(pModeData->uGstType == pgmModeToType(enmGst));
3504 Assert(pModeData->uShwType == pgmModeToType(enmShw));
3505
3506 /* shadow */
3507 pVCpu->pgm.s.pfnR3ShwRelocate = pModeData->pfnR3ShwRelocate;
3508 pVCpu->pgm.s.pfnR3ShwExit = pModeData->pfnR3ShwExit;
3509 pVCpu->pgm.s.pfnR3ShwGetPage = pModeData->pfnR3ShwGetPage;
3510 Assert(pVCpu->pgm.s.pfnR3ShwGetPage);
3511 pVCpu->pgm.s.pfnR3ShwModifyPage = pModeData->pfnR3ShwModifyPage;
3512
3513 pVCpu->pgm.s.pfnRCShwGetPage = pModeData->pfnRCShwGetPage;
3514 pVCpu->pgm.s.pfnRCShwModifyPage = pModeData->pfnRCShwModifyPage;
3515
3516 pVCpu->pgm.s.pfnR0ShwGetPage = pModeData->pfnR0ShwGetPage;
3517 pVCpu->pgm.s.pfnR0ShwModifyPage = pModeData->pfnR0ShwModifyPage;
3518
3519
3520 /* guest */
3521 pVCpu->pgm.s.pfnR3GstRelocate = pModeData->pfnR3GstRelocate;
3522 pVCpu->pgm.s.pfnR3GstExit = pModeData->pfnR3GstExit;
3523 pVCpu->pgm.s.pfnR3GstGetPage = pModeData->pfnR3GstGetPage;
3524 Assert(pVCpu->pgm.s.pfnR3GstGetPage);
3525 pVCpu->pgm.s.pfnR3GstModifyPage = pModeData->pfnR3GstModifyPage;
3526 pVCpu->pgm.s.pfnR3GstGetPDE = pModeData->pfnR3GstGetPDE;
3527 pVCpu->pgm.s.pfnRCGstGetPage = pModeData->pfnRCGstGetPage;
3528 pVCpu->pgm.s.pfnRCGstModifyPage = pModeData->pfnRCGstModifyPage;
3529 pVCpu->pgm.s.pfnRCGstGetPDE = pModeData->pfnRCGstGetPDE;
3530 pVCpu->pgm.s.pfnR0GstGetPage = pModeData->pfnR0GstGetPage;
3531 pVCpu->pgm.s.pfnR0GstModifyPage = pModeData->pfnR0GstModifyPage;
3532 pVCpu->pgm.s.pfnR0GstGetPDE = pModeData->pfnR0GstGetPDE;
3533
3534 /* both */
3535 pVCpu->pgm.s.pfnR3BthRelocate = pModeData->pfnR3BthRelocate;
3536 pVCpu->pgm.s.pfnR3BthInvalidatePage = pModeData->pfnR3BthInvalidatePage;
3537 pVCpu->pgm.s.pfnR3BthSyncCR3 = pModeData->pfnR3BthSyncCR3;
3538 Assert(pVCpu->pgm.s.pfnR3BthSyncCR3);
3539 pVCpu->pgm.s.pfnR3BthSyncPage = pModeData->pfnR3BthSyncPage;
3540 pVCpu->pgm.s.pfnR3BthPrefetchPage = pModeData->pfnR3BthPrefetchPage;
3541 pVCpu->pgm.s.pfnR3BthVerifyAccessSyncPage = pModeData->pfnR3BthVerifyAccessSyncPage;
3542#ifdef VBOX_STRICT
3543 pVCpu->pgm.s.pfnR3BthAssertCR3 = pModeData->pfnR3BthAssertCR3;
3544#endif
3545 pVCpu->pgm.s.pfnR3BthMapCR3 = pModeData->pfnR3BthMapCR3;
3546 pVCpu->pgm.s.pfnR3BthUnmapCR3 = pModeData->pfnR3BthUnmapCR3;
3547
3548 pVCpu->pgm.s.pfnRCBthTrap0eHandler = pModeData->pfnRCBthTrap0eHandler;
3549 pVCpu->pgm.s.pfnRCBthInvalidatePage = pModeData->pfnRCBthInvalidatePage;
3550 pVCpu->pgm.s.pfnRCBthSyncCR3 = pModeData->pfnRCBthSyncCR3;
3551 pVCpu->pgm.s.pfnRCBthSyncPage = pModeData->pfnRCBthSyncPage;
3552 pVCpu->pgm.s.pfnRCBthPrefetchPage = pModeData->pfnRCBthPrefetchPage;
3553 pVCpu->pgm.s.pfnRCBthVerifyAccessSyncPage = pModeData->pfnRCBthVerifyAccessSyncPage;
3554#ifdef VBOX_STRICT
3555 pVCpu->pgm.s.pfnRCBthAssertCR3 = pModeData->pfnRCBthAssertCR3;
3556#endif
3557 pVCpu->pgm.s.pfnRCBthMapCR3 = pModeData->pfnRCBthMapCR3;
3558 pVCpu->pgm.s.pfnRCBthUnmapCR3 = pModeData->pfnRCBthUnmapCR3;
3559
3560 pVCpu->pgm.s.pfnR0BthTrap0eHandler = pModeData->pfnR0BthTrap0eHandler;
3561 pVCpu->pgm.s.pfnR0BthInvalidatePage = pModeData->pfnR0BthInvalidatePage;
3562 pVCpu->pgm.s.pfnR0BthSyncCR3 = pModeData->pfnR0BthSyncCR3;
3563 pVCpu->pgm.s.pfnR0BthSyncPage = pModeData->pfnR0BthSyncPage;
3564 pVCpu->pgm.s.pfnR0BthPrefetchPage = pModeData->pfnR0BthPrefetchPage;
3565 pVCpu->pgm.s.pfnR0BthVerifyAccessSyncPage = pModeData->pfnR0BthVerifyAccessSyncPage;
3566#ifdef VBOX_STRICT
3567 pVCpu->pgm.s.pfnR0BthAssertCR3 = pModeData->pfnR0BthAssertCR3;
3568#endif
3569 pVCpu->pgm.s.pfnR0BthMapCR3 = pModeData->pfnR0BthMapCR3;
3570 pVCpu->pgm.s.pfnR0BthUnmapCR3 = pModeData->pfnR0BthUnmapCR3;
3571}
3572
3573
3574/**
3575 * Calculates the shadow paging mode.
3576 *
3577 * @returns The shadow paging mode.
3578 * @param pVM VM handle.
3579 * @param enmGuestMode The guest mode.
3580 * @param enmHostMode The host mode.
3581 * @param enmShadowMode The current shadow mode.
3582 * @param penmSwitcher Where to store the switcher to use.
3583 * VMMSWITCHER_INVALID means no change.
3584 */
3585static PGMMODE pgmR3CalcShadowMode(PVM pVM, PGMMODE enmGuestMode, SUPPAGINGMODE enmHostMode, PGMMODE enmShadowMode, VMMSWITCHER *penmSwitcher)
3586{
3587 VMMSWITCHER enmSwitcher = VMMSWITCHER_INVALID;
3588 switch (enmGuestMode)
3589 {
3590 /*
3591 * When switching to real or protected mode we don't change
3592 * anything since it's likely that we'll switch back pretty soon.
3593 *
3594 * During pgmR3InitPaging we'll end up here with PGMMODE_INVALID
3595 * and is supposed to determine which shadow paging and switcher to
3596 * use during init.
3597 */
3598 case PGMMODE_REAL:
3599 case PGMMODE_PROTECTED:
3600 if ( enmShadowMode != PGMMODE_INVALID
3601 && !HWACCMIsEnabled(pVM) /* always switch in hwaccm mode! */)
3602 break; /* (no change) */
3603
3604 switch (enmHostMode)
3605 {
3606 case SUPPAGINGMODE_32_BIT:
3607 case SUPPAGINGMODE_32_BIT_GLOBAL:
3608 enmShadowMode = PGMMODE_32_BIT;
3609 enmSwitcher = VMMSWITCHER_32_TO_32;
3610 break;
3611
3612 case SUPPAGINGMODE_PAE:
3613 case SUPPAGINGMODE_PAE_NX:
3614 case SUPPAGINGMODE_PAE_GLOBAL:
3615 case SUPPAGINGMODE_PAE_GLOBAL_NX:
3616 enmShadowMode = PGMMODE_PAE;
3617 enmSwitcher = VMMSWITCHER_PAE_TO_PAE;
3618#ifdef DEBUG_bird
3619 if (RTEnvExist("VBOX_32BIT"))
3620 {
3621 enmShadowMode = PGMMODE_32_BIT;
3622 enmSwitcher = VMMSWITCHER_PAE_TO_32;
3623 }
3624#endif
3625 break;
3626
3627 case SUPPAGINGMODE_AMD64:
3628 case SUPPAGINGMODE_AMD64_GLOBAL:
3629 case SUPPAGINGMODE_AMD64_NX:
3630 case SUPPAGINGMODE_AMD64_GLOBAL_NX:
3631 enmShadowMode = PGMMODE_PAE;
3632 enmSwitcher = VMMSWITCHER_AMD64_TO_PAE;
3633#ifdef DEBUG_bird
3634 if (RTEnvExist("VBOX_32BIT"))
3635 {
3636 enmShadowMode = PGMMODE_32_BIT;
3637 enmSwitcher = VMMSWITCHER_AMD64_TO_32;
3638 }
3639#endif
3640 break;
3641
3642 default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
3643 }
3644 break;
3645
3646 case PGMMODE_32_BIT:
3647 switch (enmHostMode)
3648 {
3649 case SUPPAGINGMODE_32_BIT:
3650 case SUPPAGINGMODE_32_BIT_GLOBAL:
3651 enmShadowMode = PGMMODE_32_BIT;
3652 enmSwitcher = VMMSWITCHER_32_TO_32;
3653 break;
3654
3655 case SUPPAGINGMODE_PAE:
3656 case SUPPAGINGMODE_PAE_NX:
3657 case SUPPAGINGMODE_PAE_GLOBAL:
3658 case SUPPAGINGMODE_PAE_GLOBAL_NX:
3659 enmShadowMode = PGMMODE_PAE;
3660 enmSwitcher = VMMSWITCHER_PAE_TO_PAE;
3661#ifdef DEBUG_bird
3662 if (RTEnvExist("VBOX_32BIT"))
3663 {
3664 enmShadowMode = PGMMODE_32_BIT;
3665 enmSwitcher = VMMSWITCHER_PAE_TO_32;
3666 }
3667#endif
3668 break;
3669
3670 case SUPPAGINGMODE_AMD64:
3671 case SUPPAGINGMODE_AMD64_GLOBAL:
3672 case SUPPAGINGMODE_AMD64_NX:
3673 case SUPPAGINGMODE_AMD64_GLOBAL_NX:
3674 enmShadowMode = PGMMODE_PAE;
3675 enmSwitcher = VMMSWITCHER_AMD64_TO_PAE;
3676#ifdef DEBUG_bird
3677 if (RTEnvExist("VBOX_32BIT"))
3678 {
3679 enmShadowMode = PGMMODE_32_BIT;
3680 enmSwitcher = VMMSWITCHER_AMD64_TO_32;
3681 }
3682#endif
3683 break;
3684
3685 default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
3686 }
3687 break;
3688
3689 case PGMMODE_PAE:
3690 case PGMMODE_PAE_NX: /** @todo This might require more switchers and guest+both modes. */
3691 switch (enmHostMode)
3692 {
3693 case SUPPAGINGMODE_32_BIT:
3694 case SUPPAGINGMODE_32_BIT_GLOBAL:
3695 enmShadowMode = PGMMODE_PAE;
3696 enmSwitcher = VMMSWITCHER_32_TO_PAE;
3697 break;
3698
3699 case SUPPAGINGMODE_PAE:
3700 case SUPPAGINGMODE_PAE_NX:
3701 case SUPPAGINGMODE_PAE_GLOBAL:
3702 case SUPPAGINGMODE_PAE_GLOBAL_NX:
3703 enmShadowMode = PGMMODE_PAE;
3704 enmSwitcher = VMMSWITCHER_PAE_TO_PAE;
3705 break;
3706
3707 case SUPPAGINGMODE_AMD64:
3708 case SUPPAGINGMODE_AMD64_GLOBAL:
3709 case SUPPAGINGMODE_AMD64_NX:
3710 case SUPPAGINGMODE_AMD64_GLOBAL_NX:
3711 enmShadowMode = PGMMODE_PAE;
3712 enmSwitcher = VMMSWITCHER_AMD64_TO_PAE;
3713 break;
3714
3715 default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
3716 }
3717 break;
3718
3719 case PGMMODE_AMD64:
3720 case PGMMODE_AMD64_NX:
3721 switch (enmHostMode)
3722 {
3723 case SUPPAGINGMODE_32_BIT:
3724 case SUPPAGINGMODE_32_BIT_GLOBAL:
3725 enmShadowMode = PGMMODE_AMD64;
3726 enmSwitcher = VMMSWITCHER_32_TO_AMD64;
3727 break;
3728
3729 case SUPPAGINGMODE_PAE:
3730 case SUPPAGINGMODE_PAE_NX:
3731 case SUPPAGINGMODE_PAE_GLOBAL:
3732 case SUPPAGINGMODE_PAE_GLOBAL_NX:
3733 enmShadowMode = PGMMODE_AMD64;
3734 enmSwitcher = VMMSWITCHER_PAE_TO_AMD64;
3735 break;
3736
3737 case SUPPAGINGMODE_AMD64:
3738 case SUPPAGINGMODE_AMD64_GLOBAL:
3739 case SUPPAGINGMODE_AMD64_NX:
3740 case SUPPAGINGMODE_AMD64_GLOBAL_NX:
3741 enmShadowMode = PGMMODE_AMD64;
3742 enmSwitcher = VMMSWITCHER_AMD64_TO_AMD64;
3743 break;
3744
3745 default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
3746 }
3747 break;
3748
3749
3750 default:
3751 AssertReleaseMsgFailed(("enmGuestMode=%d\n", enmGuestMode));
3752 return PGMMODE_INVALID;
3753 }
3754 /* Override the shadow mode is nested paging is active. */
3755 if (HWACCMIsNestedPagingActive(pVM))
3756 enmShadowMode = HWACCMGetShwPagingMode(pVM);
3757
3758 *penmSwitcher = enmSwitcher;
3759 return enmShadowMode;
3760}
3761
3762
3763/**
3764 * Performs the actual mode change.
3765 * This is called by PGMChangeMode and pgmR3InitPaging().
3766 *
3767 * @returns VBox status code. May suspend or power off the VM on error, but this
3768 * will trigger using FFs and not status codes.
3769 *
3770 * @param pVM VM handle.
3771 * @param pVCpu The VMCPU to operate on.
3772 * @param enmGuestMode The new guest mode. This is assumed to be different from
3773 * the current mode.
3774 */
3775VMMR3DECL(int) PGMR3ChangeMode(PVM pVM, PVMCPU pVCpu, PGMMODE enmGuestMode)
3776{
3777 Log(("PGMR3ChangeMode: Guest mode: %s -> %s\n", PGMGetModeName(pVCpu->pgm.s.enmGuestMode), PGMGetModeName(enmGuestMode)));
3778 STAM_REL_COUNTER_INC(&pVCpu->pgm.s.cGuestModeChanges);
3779
3780 /*
3781 * Calc the shadow mode and switcher.
3782 */
3783 VMMSWITCHER enmSwitcher;
3784 PGMMODE enmShadowMode = pgmR3CalcShadowMode(pVM, enmGuestMode, pVM->pgm.s.enmHostMode, pVCpu->pgm.s.enmShadowMode, &enmSwitcher);
3785 if (enmSwitcher != VMMSWITCHER_INVALID)
3786 {
3787 /*
3788 * Select new switcher.
3789 */
3790 int rc = VMMR3SelectSwitcher(pVM, enmSwitcher);
3791 if (RT_FAILURE(rc))
3792 {
3793 AssertReleaseMsgFailed(("VMMR3SelectSwitcher(%d) -> %Rrc\n", enmSwitcher, rc));
3794 return rc;
3795 }
3796 }
3797
3798 /*
3799 * Exit old mode(s).
3800 */
3801 /* shadow */
3802 if (enmShadowMode != pVCpu->pgm.s.enmShadowMode)
3803 {
3804 LogFlow(("PGMR3ChangeMode: Shadow mode: %s -> %s\n", PGMGetModeName(pVCpu->pgm.s.enmShadowMode), PGMGetModeName(enmShadowMode)));
3805 if (PGM_SHW_PFN(Exit, pVCpu))
3806 {
3807 int rc = PGM_SHW_PFN(Exit, pVCpu)(pVCpu);
3808 if (RT_FAILURE(rc))
3809 {
3810 AssertMsgFailed(("Exit failed for shadow mode %d: %Rrc\n", pVCpu->pgm.s.enmShadowMode, rc));
3811 return rc;
3812 }
3813 }
3814
3815 }
3816 else
3817 LogFlow(("PGMR3ChangeMode: Shadow mode remains: %s\n", PGMGetModeName(pVCpu->pgm.s.enmShadowMode)));
3818
3819 /* guest */
3820 if (PGM_GST_PFN(Exit, pVCpu))
3821 {
3822 int rc = PGM_GST_PFN(Exit, pVCpu)(pVCpu);
3823 if (RT_FAILURE(rc))
3824 {
3825 AssertMsgFailed(("Exit failed for guest mode %d: %Rrc\n", pVCpu->pgm.s.enmGuestMode, rc));
3826 return rc;
3827 }
3828 }
3829
3830 /*
3831 * Load new paging mode data.
3832 */
3833 pgmR3ModeDataSwitch(pVM, pVCpu, enmShadowMode, enmGuestMode);
3834
3835 /*
3836 * Enter new shadow mode (if changed).
3837 */
3838 if (enmShadowMode != pVCpu->pgm.s.enmShadowMode)
3839 {
3840 int rc;
3841 pVCpu->pgm.s.enmShadowMode = enmShadowMode;
3842 switch (enmShadowMode)
3843 {
3844 case PGMMODE_32_BIT:
3845 rc = PGM_SHW_NAME_32BIT(Enter)(pVCpu);
3846 break;
3847 case PGMMODE_PAE:
3848 case PGMMODE_PAE_NX:
3849 rc = PGM_SHW_NAME_PAE(Enter)(pVCpu);
3850 break;
3851 case PGMMODE_AMD64:
3852 case PGMMODE_AMD64_NX:
3853 rc = PGM_SHW_NAME_AMD64(Enter)(pVCpu);
3854 break;
3855 case PGMMODE_NESTED:
3856 rc = PGM_SHW_NAME_NESTED(Enter)(pVCpu);
3857 break;
3858 case PGMMODE_EPT:
3859 rc = PGM_SHW_NAME_EPT(Enter)(pVCpu);
3860 break;
3861 case PGMMODE_REAL:
3862 case PGMMODE_PROTECTED:
3863 default:
3864 AssertReleaseMsgFailed(("enmShadowMode=%d\n", enmShadowMode));
3865 return VERR_INTERNAL_ERROR;
3866 }
3867 if (RT_FAILURE(rc))
3868 {
3869 AssertReleaseMsgFailed(("Entering enmShadowMode=%d failed: %Rrc\n", enmShadowMode, rc));
3870 pVCpu->pgm.s.enmShadowMode = PGMMODE_INVALID;
3871 return rc;
3872 }
3873 }
3874
3875 /*
3876 * Always flag the necessary updates
3877 */
3878 VMCPU_FF_SET(pVCpu, VMCPU_FF_PGM_SYNC_CR3);
3879
3880 /*
3881 * Enter the new guest and shadow+guest modes.
3882 */
3883 int rc = -1;
3884 int rc2 = -1;
3885 RTGCPHYS GCPhysCR3 = NIL_RTGCPHYS;
3886 pVCpu->pgm.s.enmGuestMode = enmGuestMode;
3887 switch (enmGuestMode)
3888 {
3889 case PGMMODE_REAL:
3890 rc = PGM_GST_NAME_REAL(Enter)(pVCpu, NIL_RTGCPHYS);
3891 switch (pVCpu->pgm.s.enmShadowMode)
3892 {
3893 case PGMMODE_32_BIT:
3894 rc2 = PGM_BTH_NAME_32BIT_REAL(Enter)(pVCpu, NIL_RTGCPHYS);
3895 break;
3896 case PGMMODE_PAE:
3897 case PGMMODE_PAE_NX:
3898 rc2 = PGM_BTH_NAME_PAE_REAL(Enter)(pVCpu, NIL_RTGCPHYS);
3899 break;
3900 case PGMMODE_NESTED:
3901 rc2 = PGM_BTH_NAME_NESTED_REAL(Enter)(pVCpu, NIL_RTGCPHYS);
3902 break;
3903 case PGMMODE_EPT:
3904 rc2 = PGM_BTH_NAME_EPT_REAL(Enter)(pVCpu, NIL_RTGCPHYS);
3905 break;
3906 case PGMMODE_AMD64:
3907 case PGMMODE_AMD64_NX:
3908 AssertMsgFailed(("Should use PAE shadow mode!\n"));
3909 default: AssertFailed(); break;
3910 }
3911 break;
3912
3913 case PGMMODE_PROTECTED:
3914 rc = PGM_GST_NAME_PROT(Enter)(pVCpu, NIL_RTGCPHYS);
3915 switch (pVCpu->pgm.s.enmShadowMode)
3916 {
3917 case PGMMODE_32_BIT:
3918 rc2 = PGM_BTH_NAME_32BIT_PROT(Enter)(pVCpu, NIL_RTGCPHYS);
3919 break;
3920 case PGMMODE_PAE:
3921 case PGMMODE_PAE_NX:
3922 rc2 = PGM_BTH_NAME_PAE_PROT(Enter)(pVCpu, NIL_RTGCPHYS);
3923 break;
3924 case PGMMODE_NESTED:
3925 rc2 = PGM_BTH_NAME_NESTED_PROT(Enter)(pVCpu, NIL_RTGCPHYS);
3926 break;
3927 case PGMMODE_EPT:
3928 rc2 = PGM_BTH_NAME_EPT_PROT(Enter)(pVCpu, NIL_RTGCPHYS);
3929 break;
3930 case PGMMODE_AMD64:
3931 case PGMMODE_AMD64_NX:
3932 AssertMsgFailed(("Should use PAE shadow mode!\n"));
3933 default: AssertFailed(); break;
3934 }
3935 break;
3936
3937 case PGMMODE_32_BIT:
3938 GCPhysCR3 = CPUMGetGuestCR3(pVCpu) & X86_CR3_PAGE_MASK;
3939 rc = PGM_GST_NAME_32BIT(Enter)(pVCpu, GCPhysCR3);
3940 switch (pVCpu->pgm.s.enmShadowMode)
3941 {
3942 case PGMMODE_32_BIT:
3943 rc2 = PGM_BTH_NAME_32BIT_32BIT(Enter)(pVCpu, GCPhysCR3);
3944 break;
3945 case PGMMODE_PAE:
3946 case PGMMODE_PAE_NX:
3947 rc2 = PGM_BTH_NAME_PAE_32BIT(Enter)(pVCpu, GCPhysCR3);
3948 break;
3949 case PGMMODE_NESTED:
3950 rc2 = PGM_BTH_NAME_NESTED_32BIT(Enter)(pVCpu, GCPhysCR3);
3951 break;
3952 case PGMMODE_EPT:
3953 rc2 = PGM_BTH_NAME_EPT_32BIT(Enter)(pVCpu, GCPhysCR3);
3954 break;
3955 case PGMMODE_AMD64:
3956 case PGMMODE_AMD64_NX:
3957 AssertMsgFailed(("Should use PAE shadow mode!\n"));
3958 default: AssertFailed(); break;
3959 }
3960 break;
3961
3962 case PGMMODE_PAE_NX:
3963 case PGMMODE_PAE:
3964 {
3965 uint32_t u32Dummy, u32Features;
3966
3967 CPUMGetGuestCpuId(pVCpu, 1, &u32Dummy, &u32Dummy, &u32Dummy, &u32Features);
3968 if (!(u32Features & X86_CPUID_FEATURE_EDX_PAE))
3969 return VMSetRuntimeError(pVM, VMSETRTERR_FLAGS_FATAL, "PAEmode",
3970 N_("The guest is trying to switch to the PAE mode which is currently disabled by default in VirtualBox. PAE support can be enabled using the VM settings (General/Advanced)"));
3971
3972 GCPhysCR3 = CPUMGetGuestCR3(pVCpu) & X86_CR3_PAE_PAGE_MASK;
3973 rc = PGM_GST_NAME_PAE(Enter)(pVCpu, GCPhysCR3);
3974 switch (pVCpu->pgm.s.enmShadowMode)
3975 {
3976 case PGMMODE_PAE:
3977 case PGMMODE_PAE_NX:
3978 rc2 = PGM_BTH_NAME_PAE_PAE(Enter)(pVCpu, GCPhysCR3);
3979 break;
3980 case PGMMODE_NESTED:
3981 rc2 = PGM_BTH_NAME_NESTED_PAE(Enter)(pVCpu, GCPhysCR3);
3982 break;
3983 case PGMMODE_EPT:
3984 rc2 = PGM_BTH_NAME_EPT_PAE(Enter)(pVCpu, GCPhysCR3);
3985 break;
3986 case PGMMODE_32_BIT:
3987 case PGMMODE_AMD64:
3988 case PGMMODE_AMD64_NX:
3989 AssertMsgFailed(("Should use PAE shadow mode!\n"));
3990 default: AssertFailed(); break;
3991 }
3992 break;
3993 }
3994
3995#ifdef VBOX_WITH_64_BITS_GUESTS
3996 case PGMMODE_AMD64_NX:
3997 case PGMMODE_AMD64:
3998 GCPhysCR3 = CPUMGetGuestCR3(pVCpu) & UINT64_C(0xfffffffffffff000); /** @todo define this mask! */
3999 rc = PGM_GST_NAME_AMD64(Enter)(pVCpu, GCPhysCR3);
4000 switch (pVCpu->pgm.s.enmShadowMode)
4001 {
4002 case PGMMODE_AMD64:
4003 case PGMMODE_AMD64_NX:
4004 rc2 = PGM_BTH_NAME_AMD64_AMD64(Enter)(pVCpu, GCPhysCR3);
4005 break;
4006 case PGMMODE_NESTED:
4007 rc2 = PGM_BTH_NAME_NESTED_AMD64(Enter)(pVCpu, GCPhysCR3);
4008 break;
4009 case PGMMODE_EPT:
4010 rc2 = PGM_BTH_NAME_EPT_AMD64(Enter)(pVCpu, GCPhysCR3);
4011 break;
4012 case PGMMODE_32_BIT:
4013 case PGMMODE_PAE:
4014 case PGMMODE_PAE_NX:
4015 AssertMsgFailed(("Should use AMD64 shadow mode!\n"));
4016 default: AssertFailed(); break;
4017 }
4018 break;
4019#endif
4020
4021 default:
4022 AssertReleaseMsgFailed(("enmGuestMode=%d\n", enmGuestMode));
4023 rc = VERR_NOT_IMPLEMENTED;
4024 break;
4025 }
4026
4027 /* status codes. */
4028 AssertRC(rc);
4029 AssertRC(rc2);
4030 if (RT_SUCCESS(rc))
4031 {
4032 rc = rc2;
4033 if (RT_SUCCESS(rc)) /* no informational status codes. */
4034 rc = VINF_SUCCESS;
4035 }
4036
4037 /* Notify HWACCM as well. */
4038 HWACCMR3PagingModeChanged(pVM, pVCpu, pVCpu->pgm.s.enmShadowMode, pVCpu->pgm.s.enmGuestMode);
4039 return rc;
4040}
4041
4042/**
4043 * Release the pgm lock if owned by the current VCPU
4044 *
4045 * @param pVM The VM to operate on.
4046 */
4047VMMR3DECL(void) PGMR3ReleaseOwnedLocks(PVM pVM)
4048{
4049 while (PDMCritSectIsOwner(&pVM->pgm.s.CritSect))
4050 PDMCritSectLeave(&pVM->pgm.s.CritSect);
4051}
4052
4053/**
4054 * Called by pgmPoolFlushAllInt prior to flushing the pool.
4055 *
4056 * @returns VBox status code, fully asserted.
4057 * @param pVM The VM handle.
4058 * @param pVCpu The VMCPU to operate on.
4059 */
4060int pgmR3ExitShadowModeBeforePoolFlush(PVM pVM, PVMCPU pVCpu)
4061{
4062 /** @todo Need to synchronize this across all VCPUs! */
4063
4064 /* Unmap the old CR3 value before flushing everything. */
4065 int rc = PGM_BTH_PFN(UnmapCR3, pVCpu)(pVCpu);
4066 AssertRC(rc);
4067
4068 /* Exit the current shadow paging mode as well; nested paging and EPT use a root CR3 which will get flushed here. */
4069 rc = PGM_SHW_PFN(Exit, pVCpu)(pVCpu);
4070 AssertRC(rc);
4071 Assert(pVCpu->pgm.s.pShwPageCR3R3 == NULL);
4072 return rc;
4073}
4074
4075
4076/**
4077 * Called by pgmPoolFlushAllInt after flushing the pool.
4078 *
4079 * @returns VBox status code, fully asserted.
4080 * @param pVM The VM handle.
4081 * @param pVCpu The VMCPU to operate on.
4082 */
4083int pgmR3ReEnterShadowModeAfterPoolFlush(PVM pVM, PVMCPU pVCpu)
4084{
4085 pVCpu->pgm.s.enmShadowMode = PGMMODE_INVALID;
4086 int rc = PGMR3ChangeMode(pVM, pVCpu, PGMGetGuestMode(pVCpu));
4087 Assert(VMCPU_FF_ISSET(pVCpu, VMCPU_FF_PGM_SYNC_CR3));
4088 AssertRCReturn(rc, rc);
4089 AssertRCSuccessReturn(rc, VERR_IPE_UNEXPECTED_INFO_STATUS);
4090
4091 Assert(pVCpu->pgm.s.pShwPageCR3R3 != NULL);
4092 AssertMsg( pVCpu->pgm.s.enmShadowMode >= PGMMODE_NESTED
4093 || CPUMGetHyperCR3(pVCpu) == PGMGetHyperCR3(pVCpu),
4094 ("%RHp != %RHp %s\n", (RTHCPHYS)CPUMGetHyperCR3(pVCpu), PGMGetHyperCR3(pVCpu), PGMGetModeName(pVCpu->pgm.s.enmShadowMode)));
4095 return rc;
4096}
4097
4098
4099/**
4100 * Dumps a PAE shadow page table.
4101 *
4102 * @returns VBox status code (VINF_SUCCESS).
4103 * @param pVM The VM handle.
4104 * @param pPT Pointer to the page table.
4105 * @param u64Address The virtual address of the page table starts.
4106 * @param fLongMode Set if this a long mode table; clear if it's a legacy mode table.
4107 * @param cMaxDepth The maxium depth.
4108 * @param pHlp Pointer to the output functions.
4109 */
4110static int pgmR3DumpHierarchyHCPaePT(PVM pVM, PX86PTPAE pPT, uint64_t u64Address, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
4111{
4112 for (unsigned i = 0; i < RT_ELEMENTS(pPT->a); i++)
4113 {
4114 X86PTEPAE Pte = pPT->a[i];
4115 if (Pte.n.u1Present)
4116 {
4117 pHlp->pfnPrintf(pHlp,
4118 fLongMode /*P R S A D G WT CD AT NX 4M a p ? */
4119 ? "%016llx 3 | P %c %c %c %c %c %s %s %s %s 4K %c%c%c %016llx\n"
4120 : "%08llx 2 | P %c %c %c %c %c %s %s %s %s 4K %c%c%c %016llx\n",
4121 u64Address + ((uint64_t)i << X86_PT_PAE_SHIFT),
4122 Pte.n.u1Write ? 'W' : 'R',
4123 Pte.n.u1User ? 'U' : 'S',
4124 Pte.n.u1Accessed ? 'A' : '-',
4125 Pte.n.u1Dirty ? 'D' : '-',
4126 Pte.n.u1Global ? 'G' : '-',
4127 Pte.n.u1WriteThru ? "WT" : "--",
4128 Pte.n.u1CacheDisable? "CD" : "--",
4129 Pte.n.u1PAT ? "AT" : "--",
4130 Pte.n.u1NoExecute ? "NX" : "--",
4131 Pte.u & PGM_PTFLAGS_TRACK_DIRTY ? 'd' : '-',
4132 Pte.u & RT_BIT(10) ? '1' : '0',
4133 Pte.u & PGM_PTFLAGS_CSAM_VALIDATED? 'v' : '-',
4134 Pte.u & X86_PTE_PAE_PG_MASK);
4135 }
4136 }
4137 return VINF_SUCCESS;
4138}
4139
4140
4141/**
4142 * Dumps a PAE shadow page directory table.
4143 *
4144 * @returns VBox status code (VINF_SUCCESS).
4145 * @param pVM The VM handle.
4146 * @param HCPhys The physical address of the page directory table.
4147 * @param u64Address The virtual address of the page table starts.
4148 * @param cr4 The CR4, PSE is currently used.
4149 * @param fLongMode Set if this a long mode table; clear if it's a legacy mode table.
4150 * @param cMaxDepth The maxium depth.
4151 * @param pHlp Pointer to the output functions.
4152 */
4153static int pgmR3DumpHierarchyHCPaePD(PVM pVM, RTHCPHYS HCPhys, uint64_t u64Address, uint32_t cr4, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
4154{
4155 PX86PDPAE pPD = (PX86PDPAE)MMPagePhys2Page(pVM, HCPhys);
4156 if (!pPD)
4157 {
4158 pHlp->pfnPrintf(pHlp, "%0*llx error! Page directory at HCPhys=%RHp was not found in the page pool!\n",
4159 fLongMode ? 16 : 8, u64Address, HCPhys);
4160 return VERR_INVALID_PARAMETER;
4161 }
4162 const bool fBigPagesSupported = fLongMode || !!(cr4 & X86_CR4_PSE);
4163
4164 int rc = VINF_SUCCESS;
4165 for (unsigned i = 0; i < RT_ELEMENTS(pPD->a); i++)
4166 {
4167 X86PDEPAE Pde = pPD->a[i];
4168 if (Pde.n.u1Present)
4169 {
4170 if (fBigPagesSupported && Pde.b.u1Size)
4171 pHlp->pfnPrintf(pHlp,
4172 fLongMode /*P R S A D G WT CD AT NX 4M a p ? */
4173 ? "%016llx 2 | P %c %c %c %c %c %s %s %s %s 4M %c%c%c %016llx\n"
4174 : "%08llx 1 | P %c %c %c %c %c %s %s %s %s 4M %c%c%c %016llx\n",
4175 u64Address + ((uint64_t)i << X86_PD_PAE_SHIFT),
4176 Pde.b.u1Write ? 'W' : 'R',
4177 Pde.b.u1User ? 'U' : 'S',
4178 Pde.b.u1Accessed ? 'A' : '-',
4179 Pde.b.u1Dirty ? 'D' : '-',
4180 Pde.b.u1Global ? 'G' : '-',
4181 Pde.b.u1WriteThru ? "WT" : "--",
4182 Pde.b.u1CacheDisable? "CD" : "--",
4183 Pde.b.u1PAT ? "AT" : "--",
4184 Pde.b.u1NoExecute ? "NX" : "--",
4185 Pde.u & RT_BIT_64(9) ? '1' : '0',
4186 Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
4187 Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
4188 Pde.u & X86_PDE_PAE_PG_MASK);
4189 else
4190 {
4191 pHlp->pfnPrintf(pHlp,
4192 fLongMode /*P R S A D G WT CD AT NX 4M a p ? */
4193 ? "%016llx 2 | P %c %c %c %c %c %s %s .. %s 4K %c%c%c %016llx\n"
4194 : "%08llx 1 | P %c %c %c %c %c %s %s .. %s 4K %c%c%c %016llx\n",
4195 u64Address + ((uint64_t)i << X86_PD_PAE_SHIFT),
4196 Pde.n.u1Write ? 'W' : 'R',
4197 Pde.n.u1User ? 'U' : 'S',
4198 Pde.n.u1Accessed ? 'A' : '-',
4199 Pde.n.u1Reserved0 ? '?' : '.', /* ignored */
4200 Pde.n.u1Reserved1 ? '?' : '.', /* ignored */
4201 Pde.n.u1WriteThru ? "WT" : "--",
4202 Pde.n.u1CacheDisable? "CD" : "--",
4203 Pde.n.u1NoExecute ? "NX" : "--",
4204 Pde.u & RT_BIT_64(9) ? '1' : '0',
4205 Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
4206 Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
4207 Pde.u & X86_PDE_PAE_PG_MASK);
4208 if (cMaxDepth >= 1)
4209 {
4210 /** @todo what about using the page pool for mapping PTs? */
4211 uint64_t u64AddressPT = u64Address + ((uint64_t)i << X86_PD_PAE_SHIFT);
4212 RTHCPHYS HCPhysPT = Pde.u & X86_PDE_PAE_PG_MASK;
4213 PX86PTPAE pPT = NULL;
4214 if (!(Pde.u & PGM_PDFLAGS_MAPPING))
4215 pPT = (PX86PTPAE)MMPagePhys2Page(pVM, HCPhysPT);
4216 else
4217 {
4218 for (PPGMMAPPING pMap = pVM->pgm.s.pMappingsR3; pMap; pMap = pMap->pNextR3)
4219 {
4220 uint64_t off = u64AddressPT - pMap->GCPtr;
4221 if (off < pMap->cb)
4222 {
4223 const int iPDE = (uint32_t)(off >> X86_PD_SHIFT);
4224 const int iSub = (int)((off >> X86_PD_PAE_SHIFT) & 1); /* MSC is a pain sometimes */
4225 if ((iSub ? pMap->aPTs[iPDE].HCPhysPaePT1 : pMap->aPTs[iPDE].HCPhysPaePT0) != HCPhysPT)
4226 pHlp->pfnPrintf(pHlp, "%0*llx error! Mapping error! PT %d has HCPhysPT=%RHp not %RHp is in the PD.\n",
4227 fLongMode ? 16 : 8, u64AddressPT, iPDE,
4228 iSub ? pMap->aPTs[iPDE].HCPhysPaePT1 : pMap->aPTs[iPDE].HCPhysPaePT0, HCPhysPT);
4229 pPT = &pMap->aPTs[iPDE].paPaePTsR3[iSub];
4230 }
4231 }
4232 }
4233 int rc2 = VERR_INVALID_PARAMETER;
4234 if (pPT)
4235 rc2 = pgmR3DumpHierarchyHCPaePT(pVM, pPT, u64AddressPT, fLongMode, cMaxDepth - 1, pHlp);
4236 else
4237 pHlp->pfnPrintf(pHlp, "%0*llx error! Page table at HCPhys=%RHp was not found in the page pool!\n",
4238 fLongMode ? 16 : 8, u64AddressPT, HCPhysPT);
4239 if (rc2 < rc && RT_SUCCESS(rc))
4240 rc = rc2;
4241 }
4242 }
4243 }
4244 }
4245 return rc;
4246}
4247
4248
4249/**
4250 * Dumps a PAE shadow page directory pointer table.
4251 *
4252 * @returns VBox status code (VINF_SUCCESS).
4253 * @param pVM The VM handle.
4254 * @param HCPhys The physical address of the page directory pointer table.
4255 * @param u64Address The virtual address of the page table starts.
4256 * @param cr4 The CR4, PSE is currently used.
4257 * @param fLongMode Set if this a long mode table; clear if it's a legacy mode table.
4258 * @param cMaxDepth The maxium depth.
4259 * @param pHlp Pointer to the output functions.
4260 */
4261static int pgmR3DumpHierarchyHCPaePDPT(PVM pVM, RTHCPHYS HCPhys, uint64_t u64Address, uint32_t cr4, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
4262{
4263 PX86PDPT pPDPT = (PX86PDPT)MMPagePhys2Page(pVM, HCPhys);
4264 if (!pPDPT)
4265 {
4266 pHlp->pfnPrintf(pHlp, "%0*llx error! Page directory pointer table at HCPhys=%RHp was not found in the page pool!\n",
4267 fLongMode ? 16 : 8, u64Address, HCPhys);
4268 return VERR_INVALID_PARAMETER;
4269 }
4270
4271 int rc = VINF_SUCCESS;
4272 const unsigned c = fLongMode ? RT_ELEMENTS(pPDPT->a) : X86_PG_PAE_PDPE_ENTRIES;
4273 for (unsigned i = 0; i < c; i++)
4274 {
4275 X86PDPE Pdpe = pPDPT->a[i];
4276 if (Pdpe.n.u1Present)
4277 {
4278 if (fLongMode)
4279 pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a p ? */
4280 "%016llx 1 | P %c %c %c %c %c %s %s %s %s .. %c%c%c %016llx\n",
4281 u64Address + ((uint64_t)i << X86_PDPT_SHIFT),
4282 Pdpe.lm.u1Write ? 'W' : 'R',
4283 Pdpe.lm.u1User ? 'U' : 'S',
4284 Pdpe.lm.u1Accessed ? 'A' : '-',
4285 Pdpe.lm.u3Reserved & 1? '?' : '.', /* ignored */
4286 Pdpe.lm.u3Reserved & 4? '!' : '.', /* mbz */
4287 Pdpe.lm.u1WriteThru ? "WT" : "--",
4288 Pdpe.lm.u1CacheDisable? "CD" : "--",
4289 Pdpe.lm.u3Reserved & 2? "!" : "..",/* mbz */
4290 Pdpe.lm.u1NoExecute ? "NX" : "--",
4291 Pdpe.u & RT_BIT(9) ? '1' : '0',
4292 Pdpe.u & PGM_PLXFLAGS_PERMANENT ? 'p' : '-',
4293 Pdpe.u & RT_BIT(11) ? '1' : '0',
4294 Pdpe.u & X86_PDPE_PG_MASK);
4295 else
4296 pHlp->pfnPrintf(pHlp, /*P G WT CD AT NX 4M a p ? */
4297 "%08x 0 | P %c %s %s %s %s .. %c%c%c %016llx\n",
4298 i << X86_PDPT_SHIFT,
4299 Pdpe.n.u4Reserved & 1? '!' : '.', /* mbz */
4300 Pdpe.n.u4Reserved & 4? '!' : '.', /* mbz */
4301 Pdpe.n.u1WriteThru ? "WT" : "--",
4302 Pdpe.n.u1CacheDisable? "CD" : "--",
4303 Pdpe.n.u4Reserved & 2? "!" : "..",/* mbz */
4304 Pdpe.u & RT_BIT(9) ? '1' : '0',
4305 Pdpe.u & PGM_PLXFLAGS_PERMANENT ? 'p' : '-',
4306 Pdpe.u & RT_BIT(11) ? '1' : '0',
4307 Pdpe.u & X86_PDPE_PG_MASK);
4308 if (cMaxDepth >= 1)
4309 {
4310 int rc2 = pgmR3DumpHierarchyHCPaePD(pVM, Pdpe.u & X86_PDPE_PG_MASK, u64Address + ((uint64_t)i << X86_PDPT_SHIFT),
4311 cr4, fLongMode, cMaxDepth - 1, pHlp);
4312 if (rc2 < rc && RT_SUCCESS(rc))
4313 rc = rc2;
4314 }
4315 }
4316 }
4317 return rc;
4318}
4319
4320
4321/**
4322 * Dumps a 32-bit shadow page table.
4323 *
4324 * @returns VBox status code (VINF_SUCCESS).
4325 * @param pVM The VM handle.
4326 * @param HCPhys The physical address of the table.
4327 * @param cr4 The CR4, PSE is currently used.
4328 * @param cMaxDepth The maxium depth.
4329 * @param pHlp Pointer to the output functions.
4330 */
4331static int pgmR3DumpHierarchyHcPaePML4(PVM pVM, RTHCPHYS HCPhys, uint32_t cr4, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
4332{
4333 PX86PML4 pPML4 = (PX86PML4)MMPagePhys2Page(pVM, HCPhys);
4334 if (!pPML4)
4335 {
4336 pHlp->pfnPrintf(pHlp, "Page map level 4 at HCPhys=%RHp was not found in the page pool!\n", HCPhys);
4337 return VERR_INVALID_PARAMETER;
4338 }
4339
4340 int rc = VINF_SUCCESS;
4341 for (unsigned i = 0; i < RT_ELEMENTS(pPML4->a); i++)
4342 {
4343 X86PML4E Pml4e = pPML4->a[i];
4344 if (Pml4e.n.u1Present)
4345 {
4346 uint64_t u64Address = ((uint64_t)i << X86_PML4_SHIFT) | (((uint64_t)i >> (X86_PML4_SHIFT - X86_PDPT_SHIFT - 1)) * 0xffff000000000000ULL);
4347 pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a p ? */
4348 "%016llx 0 | P %c %c %c %c %c %s %s %s %s .. %c%c%c %016llx\n",
4349 u64Address,
4350 Pml4e.n.u1Write ? 'W' : 'R',
4351 Pml4e.n.u1User ? 'U' : 'S',
4352 Pml4e.n.u1Accessed ? 'A' : '-',
4353 Pml4e.n.u3Reserved & 1? '?' : '.', /* ignored */
4354 Pml4e.n.u3Reserved & 4? '!' : '.', /* mbz */
4355 Pml4e.n.u1WriteThru ? "WT" : "--",
4356 Pml4e.n.u1CacheDisable? "CD" : "--",
4357 Pml4e.n.u3Reserved & 2? "!" : "..",/* mbz */
4358 Pml4e.n.u1NoExecute ? "NX" : "--",
4359 Pml4e.u & RT_BIT(9) ? '1' : '0',
4360 Pml4e.u & PGM_PLXFLAGS_PERMANENT ? 'p' : '-',
4361 Pml4e.u & RT_BIT(11) ? '1' : '0',
4362 Pml4e.u & X86_PML4E_PG_MASK);
4363
4364 if (cMaxDepth >= 1)
4365 {
4366 int rc2 = pgmR3DumpHierarchyHCPaePDPT(pVM, Pml4e.u & X86_PML4E_PG_MASK, u64Address, cr4, true, cMaxDepth - 1, pHlp);
4367 if (rc2 < rc && RT_SUCCESS(rc))
4368 rc = rc2;
4369 }
4370 }
4371 }
4372 return rc;
4373}
4374
4375
4376/**
4377 * Dumps a 32-bit shadow page table.
4378 *
4379 * @returns VBox status code (VINF_SUCCESS).
4380 * @param pVM The VM handle.
4381 * @param pPT Pointer to the page table.
4382 * @param u32Address The virtual address this table starts at.
4383 * @param pHlp Pointer to the output functions.
4384 */
4385int pgmR3DumpHierarchyHC32BitPT(PVM pVM, PX86PT pPT, uint32_t u32Address, PCDBGFINFOHLP pHlp)
4386{
4387 for (unsigned i = 0; i < RT_ELEMENTS(pPT->a); i++)
4388 {
4389 X86PTE Pte = pPT->a[i];
4390 if (Pte.n.u1Present)
4391 {
4392 pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a m d */
4393 "%08x 1 | P %c %c %c %c %c %s %s %s .. 4K %c%c%c %08x\n",
4394 u32Address + (i << X86_PT_SHIFT),
4395 Pte.n.u1Write ? 'W' : 'R',
4396 Pte.n.u1User ? 'U' : 'S',
4397 Pte.n.u1Accessed ? 'A' : '-',
4398 Pte.n.u1Dirty ? 'D' : '-',
4399 Pte.n.u1Global ? 'G' : '-',
4400 Pte.n.u1WriteThru ? "WT" : "--",
4401 Pte.n.u1CacheDisable? "CD" : "--",
4402 Pte.n.u1PAT ? "AT" : "--",
4403 Pte.u & PGM_PTFLAGS_TRACK_DIRTY ? 'd' : '-',
4404 Pte.u & RT_BIT(10) ? '1' : '0',
4405 Pte.u & PGM_PTFLAGS_CSAM_VALIDATED ? 'v' : '-',
4406 Pte.u & X86_PDE_PG_MASK);
4407 }
4408 }
4409 return VINF_SUCCESS;
4410}
4411
4412
4413/**
4414 * Dumps a 32-bit shadow page directory and page tables.
4415 *
4416 * @returns VBox status code (VINF_SUCCESS).
4417 * @param pVM The VM handle.
4418 * @param cr3 The root of the hierarchy.
4419 * @param cr4 The CR4, PSE is currently used.
4420 * @param cMaxDepth How deep into the hierarchy the dumper should go.
4421 * @param pHlp Pointer to the output functions.
4422 */
4423int pgmR3DumpHierarchyHC32BitPD(PVM pVM, uint32_t cr3, uint32_t cr4, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
4424{
4425 PX86PD pPD = (PX86PD)MMPagePhys2Page(pVM, cr3 & X86_CR3_PAGE_MASK);
4426 if (!pPD)
4427 {
4428 pHlp->pfnPrintf(pHlp, "Page directory at %#x was not found in the page pool!\n", cr3 & X86_CR3_PAGE_MASK);
4429 return VERR_INVALID_PARAMETER;
4430 }
4431
4432 int rc = VINF_SUCCESS;
4433 for (unsigned i = 0; i < RT_ELEMENTS(pPD->a); i++)
4434 {
4435 X86PDE Pde = pPD->a[i];
4436 if (Pde.n.u1Present)
4437 {
4438 const uint32_t u32Address = i << X86_PD_SHIFT;
4439 if ((cr4 & X86_CR4_PSE) && Pde.b.u1Size)
4440 pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a m d */
4441 "%08x 0 | P %c %c %c %c %c %s %s %s .. 4M %c%c%c %08x\n",
4442 u32Address,
4443 Pde.b.u1Write ? 'W' : 'R',
4444 Pde.b.u1User ? 'U' : 'S',
4445 Pde.b.u1Accessed ? 'A' : '-',
4446 Pde.b.u1Dirty ? 'D' : '-',
4447 Pde.b.u1Global ? 'G' : '-',
4448 Pde.b.u1WriteThru ? "WT" : "--",
4449 Pde.b.u1CacheDisable? "CD" : "--",
4450 Pde.b.u1PAT ? "AT" : "--",
4451 Pde.u & RT_BIT_64(9) ? '1' : '0',
4452 Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
4453 Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
4454 Pde.u & X86_PDE4M_PG_MASK);
4455 else
4456 {
4457 pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a m d */
4458 "%08x 0 | P %c %c %c %c %c %s %s .. .. 4K %c%c%c %08x\n",
4459 u32Address,
4460 Pde.n.u1Write ? 'W' : 'R',
4461 Pde.n.u1User ? 'U' : 'S',
4462 Pde.n.u1Accessed ? 'A' : '-',
4463 Pde.n.u1Reserved0 ? '?' : '.', /* ignored */
4464 Pde.n.u1Reserved1 ? '?' : '.', /* ignored */
4465 Pde.n.u1WriteThru ? "WT" : "--",
4466 Pde.n.u1CacheDisable? "CD" : "--",
4467 Pde.u & RT_BIT_64(9) ? '1' : '0',
4468 Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
4469 Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
4470 Pde.u & X86_PDE_PG_MASK);
4471 if (cMaxDepth >= 1)
4472 {
4473 /** @todo what about using the page pool for mapping PTs? */
4474 RTHCPHYS HCPhys = Pde.u & X86_PDE_PG_MASK;
4475 PX86PT pPT = NULL;
4476 if (!(Pde.u & PGM_PDFLAGS_MAPPING))
4477 pPT = (PX86PT)MMPagePhys2Page(pVM, HCPhys);
4478 else
4479 {
4480 for (PPGMMAPPING pMap = pVM->pgm.s.pMappingsR3; pMap; pMap = pMap->pNextR3)
4481 if (u32Address - pMap->GCPtr < pMap->cb)
4482 {
4483 int iPDE = (u32Address - pMap->GCPtr) >> X86_PD_SHIFT;
4484 if (pMap->aPTs[iPDE].HCPhysPT != HCPhys)
4485 pHlp->pfnPrintf(pHlp, "%08x error! Mapping error! PT %d has HCPhysPT=%RHp not %RHp is in the PD.\n",
4486 u32Address, iPDE, pMap->aPTs[iPDE].HCPhysPT, HCPhys);
4487 pPT = pMap->aPTs[iPDE].pPTR3;
4488 }
4489 }
4490 int rc2 = VERR_INVALID_PARAMETER;
4491 if (pPT)
4492 rc2 = pgmR3DumpHierarchyHC32BitPT(pVM, pPT, u32Address, pHlp);
4493 else
4494 pHlp->pfnPrintf(pHlp, "%08x error! Page table at %#x was not found in the page pool!\n", u32Address, HCPhys);
4495 if (rc2 < rc && RT_SUCCESS(rc))
4496 rc = rc2;
4497 }
4498 }
4499 }
4500 }
4501
4502 return rc;
4503}
4504
4505
4506/**
4507 * Dumps a 32-bit shadow page table.
4508 *
4509 * @returns VBox status code (VINF_SUCCESS).
4510 * @param pVM The VM handle.
4511 * @param pPT Pointer to the page table.
4512 * @param u32Address The virtual address this table starts at.
4513 * @param PhysSearch Address to search for.
4514 */
4515int pgmR3DumpHierarchyGC32BitPT(PVM pVM, PX86PT pPT, uint32_t u32Address, RTGCPHYS PhysSearch)
4516{
4517 for (unsigned i = 0; i < RT_ELEMENTS(pPT->a); i++)
4518 {
4519 X86PTE Pte = pPT->a[i];
4520 if (Pte.n.u1Present)
4521 {
4522 Log(( /*P R S A D G WT CD AT NX 4M a m d */
4523 "%08x 1 | P %c %c %c %c %c %s %s %s .. 4K %c%c%c %08x\n",
4524 u32Address + (i << X86_PT_SHIFT),
4525 Pte.n.u1Write ? 'W' : 'R',
4526 Pte.n.u1User ? 'U' : 'S',
4527 Pte.n.u1Accessed ? 'A' : '-',
4528 Pte.n.u1Dirty ? 'D' : '-',
4529 Pte.n.u1Global ? 'G' : '-',
4530 Pte.n.u1WriteThru ? "WT" : "--",
4531 Pte.n.u1CacheDisable? "CD" : "--",
4532 Pte.n.u1PAT ? "AT" : "--",
4533 Pte.u & PGM_PTFLAGS_TRACK_DIRTY ? 'd' : '-',
4534 Pte.u & RT_BIT(10) ? '1' : '0',
4535 Pte.u & PGM_PTFLAGS_CSAM_VALIDATED ? 'v' : '-',
4536 Pte.u & X86_PDE_PG_MASK));
4537
4538 if ((Pte.u & X86_PDE_PG_MASK) == PhysSearch)
4539 {
4540 uint64_t fPageShw = 0;
4541 RTHCPHYS pPhysHC = 0;
4542
4543 /** @todo SMP support!! */
4544 PGMShwGetPage(&pVM->aCpus[0], (RTGCPTR)(u32Address + (i << X86_PT_SHIFT)), &fPageShw, &pPhysHC);
4545 Log(("Found %RGp at %RGv -> flags=%llx\n", PhysSearch, (RTGCPTR)(u32Address + (i << X86_PT_SHIFT)), fPageShw));
4546 }
4547 }
4548 }
4549 return VINF_SUCCESS;
4550}
4551
4552
4553/**
4554 * Dumps a 32-bit guest page directory and page tables.
4555 *
4556 * @returns VBox status code (VINF_SUCCESS).
4557 * @param pVM The VM handle.
4558 * @param cr3 The root of the hierarchy.
4559 * @param cr4 The CR4, PSE is currently used.
4560 * @param PhysSearch Address to search for.
4561 */
4562VMMR3DECL(int) PGMR3DumpHierarchyGC(PVM pVM, uint64_t cr3, uint64_t cr4, RTGCPHYS PhysSearch)
4563{
4564 bool fLongMode = false;
4565 const unsigned cch = fLongMode ? 16 : 8; NOREF(cch);
4566 PX86PD pPD = 0;
4567
4568 int rc = PGM_GCPHYS_2_PTR(pVM, cr3 & X86_CR3_PAGE_MASK, &pPD);
4569 if (RT_FAILURE(rc) || !pPD)
4570 {
4571 Log(("Page directory at %#x was not found in the page pool!\n", cr3 & X86_CR3_PAGE_MASK));
4572 return VERR_INVALID_PARAMETER;
4573 }
4574
4575 Log(("cr3=%08x cr4=%08x%s\n"
4576 "%-*s P - Present\n"
4577 "%-*s | R/W - Read (0) / Write (1)\n"
4578 "%-*s | | U/S - User (1) / Supervisor (0)\n"
4579 "%-*s | | | A - Accessed\n"
4580 "%-*s | | | | D - Dirty\n"
4581 "%-*s | | | | | G - Global\n"
4582 "%-*s | | | | | | WT - Write thru\n"
4583 "%-*s | | | | | | | CD - Cache disable\n"
4584 "%-*s | | | | | | | | AT - Attribute table (PAT)\n"
4585 "%-*s | | | | | | | | | NX - No execute (K8)\n"
4586 "%-*s | | | | | | | | | | 4K/4M/2M - Page size.\n"
4587 "%-*s | | | | | | | | | | | AVL - a=allocated; m=mapping; d=track dirty;\n"
4588 "%-*s | | | | | | | | | | | | p=permanent; v=validated;\n"
4589 "%-*s Level | | | | | | | | | | | | Page\n"
4590 /* xxxx n **** P R S A D G WT CD AT NX 4M AVL xxxxxxxxxxxxx
4591 - W U - - - -- -- -- -- -- 010 */
4592 , cr3, cr4, fLongMode ? " Long Mode" : "",
4593 cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "",
4594 cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "Address"));
4595
4596 for (unsigned i = 0; i < RT_ELEMENTS(pPD->a); i++)
4597 {
4598 X86PDE Pde = pPD->a[i];
4599 if (Pde.n.u1Present)
4600 {
4601 const uint32_t u32Address = i << X86_PD_SHIFT;
4602
4603 if ((cr4 & X86_CR4_PSE) && Pde.b.u1Size)
4604 Log(( /*P R S A D G WT CD AT NX 4M a m d */
4605 "%08x 0 | P %c %c %c %c %c %s %s %s .. 4M %c%c%c %08x\n",
4606 u32Address,
4607 Pde.b.u1Write ? 'W' : 'R',
4608 Pde.b.u1User ? 'U' : 'S',
4609 Pde.b.u1Accessed ? 'A' : '-',
4610 Pde.b.u1Dirty ? 'D' : '-',
4611 Pde.b.u1Global ? 'G' : '-',
4612 Pde.b.u1WriteThru ? "WT" : "--",
4613 Pde.b.u1CacheDisable? "CD" : "--",
4614 Pde.b.u1PAT ? "AT" : "--",
4615 Pde.u & RT_BIT(9) ? '1' : '0',
4616 Pde.u & RT_BIT(10) ? '1' : '0',
4617 Pde.u & RT_BIT(11) ? '1' : '0',
4618 pgmGstGet4MBPhysPage(&pVM->pgm.s, Pde)));
4619 /** @todo PhysSearch */
4620 else
4621 {
4622 Log(( /*P R S A D G WT CD AT NX 4M a m d */
4623 "%08x 0 | P %c %c %c %c %c %s %s .. .. 4K %c%c%c %08x\n",
4624 u32Address,
4625 Pde.n.u1Write ? 'W' : 'R',
4626 Pde.n.u1User ? 'U' : 'S',
4627 Pde.n.u1Accessed ? 'A' : '-',
4628 Pde.n.u1Reserved0 ? '?' : '.', /* ignored */
4629 Pde.n.u1Reserved1 ? '?' : '.', /* ignored */
4630 Pde.n.u1WriteThru ? "WT" : "--",
4631 Pde.n.u1CacheDisable? "CD" : "--",
4632 Pde.u & RT_BIT(9) ? '1' : '0',
4633 Pde.u & RT_BIT(10) ? '1' : '0',
4634 Pde.u & RT_BIT(11) ? '1' : '0',
4635 Pde.u & X86_PDE_PG_MASK));
4636 ////if (cMaxDepth >= 1)
4637 {
4638 /** @todo what about using the page pool for mapping PTs? */
4639 RTGCPHYS GCPhys = Pde.u & X86_PDE_PG_MASK;
4640 PX86PT pPT = NULL;
4641
4642 rc = PGM_GCPHYS_2_PTR(pVM, GCPhys, &pPT);
4643
4644 int rc2 = VERR_INVALID_PARAMETER;
4645 if (pPT)
4646 rc2 = pgmR3DumpHierarchyGC32BitPT(pVM, pPT, u32Address, PhysSearch);
4647 else
4648 Log(("%08x error! Page table at %#x was not found in the page pool!\n", u32Address, GCPhys));
4649 if (rc2 < rc && RT_SUCCESS(rc))
4650 rc = rc2;
4651 }
4652 }
4653 }
4654 }
4655
4656 return rc;
4657}
4658
4659
4660/**
4661 * Dumps a page table hierarchy use only physical addresses and cr4/lm flags.
4662 *
4663 * @returns VBox status code (VINF_SUCCESS).
4664 * @param pVM The VM handle.
4665 * @param cr3 The root of the hierarchy.
4666 * @param cr4 The cr4, only PAE and PSE is currently used.
4667 * @param fLongMode Set if long mode, false if not long mode.
4668 * @param cMaxDepth Number of levels to dump.
4669 * @param pHlp Pointer to the output functions.
4670 */
4671VMMR3DECL(int) PGMR3DumpHierarchyHC(PVM pVM, uint64_t cr3, uint64_t cr4, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
4672{
4673 if (!pHlp)
4674 pHlp = DBGFR3InfoLogHlp();
4675 if (!cMaxDepth)
4676 return VINF_SUCCESS;
4677 const unsigned cch = fLongMode ? 16 : 8;
4678 pHlp->pfnPrintf(pHlp,
4679 "cr3=%08x cr4=%08x%s\n"
4680 "%-*s P - Present\n"
4681 "%-*s | R/W - Read (0) / Write (1)\n"
4682 "%-*s | | U/S - User (1) / Supervisor (0)\n"
4683 "%-*s | | | A - Accessed\n"
4684 "%-*s | | | | D - Dirty\n"
4685 "%-*s | | | | | G - Global\n"
4686 "%-*s | | | | | | WT - Write thru\n"
4687 "%-*s | | | | | | | CD - Cache disable\n"
4688 "%-*s | | | | | | | | AT - Attribute table (PAT)\n"
4689 "%-*s | | | | | | | | | NX - No execute (K8)\n"
4690 "%-*s | | | | | | | | | | 4K/4M/2M - Page size.\n"
4691 "%-*s | | | | | | | | | | | AVL - a=allocated; m=mapping; d=track dirty;\n"
4692 "%-*s | | | | | | | | | | | | p=permanent; v=validated;\n"
4693 "%-*s Level | | | | | | | | | | | | Page\n"
4694 /* xxxx n **** P R S A D G WT CD AT NX 4M AVL xxxxxxxxxxxxx
4695 - W U - - - -- -- -- -- -- 010 */
4696 , cr3, cr4, fLongMode ? " Long Mode" : "",
4697 cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "",
4698 cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "Address");
4699 if (cr4 & X86_CR4_PAE)
4700 {
4701 if (fLongMode)
4702 return pgmR3DumpHierarchyHcPaePML4(pVM, cr3 & X86_CR3_PAGE_MASK, cr4, cMaxDepth, pHlp);
4703 return pgmR3DumpHierarchyHCPaePDPT(pVM, cr3 & X86_CR3_PAE_PAGE_MASK, 0, cr4, false, cMaxDepth, pHlp);
4704 }
4705 return pgmR3DumpHierarchyHC32BitPD(pVM, cr3 & X86_CR3_PAGE_MASK, cr4, cMaxDepth, pHlp);
4706}
4707
4708#ifdef VBOX_WITH_DEBUGGER
4709
4710/**
4711 * The '.pgmram' command.
4712 *
4713 * @returns VBox status.
4714 * @param pCmd Pointer to the command descriptor (as registered).
4715 * @param pCmdHlp Pointer to command helper functions.
4716 * @param pVM Pointer to the current VM (if any).
4717 * @param paArgs Pointer to (readonly) array of arguments.
4718 * @param cArgs Number of arguments in the array.
4719 */
4720static DECLCALLBACK(int) pgmR3CmdRam(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
4721{
4722 /*
4723 * Validate input.
4724 */
4725 if (!pVM)
4726 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires a VM to be selected.\n");
4727 if (!pVM->pgm.s.pRamRangesRC)
4728 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Sorry, no Ram is registered.\n");
4729
4730 /*
4731 * Dump the ranges.
4732 */
4733 int rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL, "From - To (incl) pvHC\n");
4734 PPGMRAMRANGE pRam;
4735 for (pRam = pVM->pgm.s.pRamRangesR3; pRam; pRam = pRam->pNextR3)
4736 {
4737 rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL,
4738 "%RGp - %RGp %p\n",
4739 pRam->GCPhys, pRam->GCPhysLast, pRam->pvR3);
4740 if (RT_FAILURE(rc))
4741 return rc;
4742 }
4743
4744 return VINF_SUCCESS;
4745}
4746
4747
4748/**
4749 * The '.pgmmap' command.
4750 *
4751 * @returns VBox status.
4752 * @param pCmd Pointer to the command descriptor (as registered).
4753 * @param pCmdHlp Pointer to command helper functions.
4754 * @param pVM Pointer to the current VM (if any).
4755 * @param paArgs Pointer to (readonly) array of arguments.
4756 * @param cArgs Number of arguments in the array.
4757 */
4758static DECLCALLBACK(int) pgmR3CmdMap(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
4759{
4760 /*
4761 * Validate input.
4762 */
4763 if (!pVM)
4764 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires a VM to be selected.\n");
4765 if (!pVM->pgm.s.pMappingsR3)
4766 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Sorry, no mappings are registered.\n");
4767
4768 /*
4769 * Print message about the fixedness of the mappings.
4770 */
4771 int rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL, pVM->pgm.s.fMappingsFixed ? "The mappings are FIXED.\n" : "The mappings are FLOATING.\n");
4772 if (RT_FAILURE(rc))
4773 return rc;
4774
4775 /*
4776 * Dump the ranges.
4777 */
4778 PPGMMAPPING pCur;
4779 for (pCur = pVM->pgm.s.pMappingsR3; pCur; pCur = pCur->pNextR3)
4780 {
4781 rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL,
4782 "%08x - %08x %s\n",
4783 pCur->GCPtr, pCur->GCPtrLast, pCur->pszDesc);
4784 if (RT_FAILURE(rc))
4785 return rc;
4786 }
4787
4788 return VINF_SUCCESS;
4789}
4790
4791
4792/**
4793 * The '.pgmerror' and '.pgmerroroff' commands.
4794 *
4795 * @returns VBox status.
4796 * @param pCmd Pointer to the command descriptor (as registered).
4797 * @param pCmdHlp Pointer to command helper functions.
4798 * @param pVM Pointer to the current VM (if any).
4799 * @param paArgs Pointer to (readonly) array of arguments.
4800 * @param cArgs Number of arguments in the array.
4801 */
4802static DECLCALLBACK(int) pgmR3CmdError(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
4803{
4804 /*
4805 * Validate input.
4806 */
4807 if (!pVM)
4808 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires a VM to be selected.\n");
4809 AssertReturn(cArgs == 0 || (cArgs == 1 && paArgs[0].enmType == DBGCVAR_TYPE_STRING),
4810 pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: Hit bug in the parser.\n"));
4811
4812 if (!cArgs)
4813 {
4814 /*
4815 * Print the list of error injection locations with status.
4816 */
4817 pCmdHlp->pfnPrintf(pCmdHlp, NULL, "PGM error inject locations:\n");
4818 pCmdHlp->pfnPrintf(pCmdHlp, NULL, " handy - %RTbool\n", pVM->pgm.s.fErrInjHandyPages);
4819 }
4820 else
4821 {
4822
4823 /*
4824 * String switch on where to inject the error.
4825 */
4826 bool const fNewState = !strcmp(pCmd->pszCmd, "pgmerror");
4827 const char *pszWhere = paArgs[0].u.pszString;
4828 if (!strcmp(pszWhere, "handy"))
4829 ASMAtomicWriteBool(&pVM->pgm.s.fErrInjHandyPages, fNewState);
4830 else
4831 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: Invalid 'where' value: %s.\n", pszWhere);
4832 pCmdHlp->pfnPrintf(pCmdHlp, NULL, "done\n");
4833 }
4834 return VINF_SUCCESS;
4835}
4836
4837
4838/**
4839 * The '.pgmsync' command.
4840 *
4841 * @returns VBox status.
4842 * @param pCmd Pointer to the command descriptor (as registered).
4843 * @param pCmdHlp Pointer to command helper functions.
4844 * @param pVM Pointer to the current VM (if any).
4845 * @param paArgs Pointer to (readonly) array of arguments.
4846 * @param cArgs Number of arguments in the array.
4847 */
4848static DECLCALLBACK(int) pgmR3CmdSync(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
4849{
4850 /** @todo SMP support */
4851 PVMCPU pVCpu = &pVM->aCpus[0];
4852
4853 /*
4854 * Validate input.
4855 */
4856 if (!pVM)
4857 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires a VM to be selected.\n");
4858
4859 /*
4860 * Force page directory sync.
4861 */
4862 VMCPU_FF_SET(pVCpu, VMCPU_FF_PGM_SYNC_CR3);
4863
4864 int rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Forcing page directory sync.\n");
4865 if (RT_FAILURE(rc))
4866 return rc;
4867
4868 return VINF_SUCCESS;
4869}
4870
4871
4872#ifdef VBOX_STRICT
4873/**
4874 * The '.pgmassertcr3' command.
4875 *
4876 * @returns VBox status.
4877 * @param pCmd Pointer to the command descriptor (as registered).
4878 * @param pCmdHlp Pointer to command helper functions.
4879 * @param pVM Pointer to the current VM (if any).
4880 * @param paArgs Pointer to (readonly) array of arguments.
4881 * @param cArgs Number of arguments in the array.
4882 */
4883static DECLCALLBACK(int) pgmR3CmdAssertCR3(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
4884{
4885 /** @todo SMP support!! */
4886 PVMCPU pVCpu = &pVM->aCpus[0];
4887
4888 /*
4889 * Validate input.
4890 */
4891 if (!pVM)
4892 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires a VM to be selected.\n");
4893
4894 int rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Checking shadow CR3 page tables for consistency.\n");
4895 if (RT_FAILURE(rc))
4896 return rc;
4897
4898 PGMAssertCR3(pVM, pVCpu, CPUMGetGuestCR3(pVCpu), CPUMGetGuestCR4(pVCpu));
4899
4900 return VINF_SUCCESS;
4901}
4902#endif /* VBOX_STRICT */
4903
4904
4905/**
4906 * The '.pgmsyncalways' command.
4907 *
4908 * @returns VBox status.
4909 * @param pCmd Pointer to the command descriptor (as registered).
4910 * @param pCmdHlp Pointer to command helper functions.
4911 * @param pVM Pointer to the current VM (if any).
4912 * @param paArgs Pointer to (readonly) array of arguments.
4913 * @param cArgs Number of arguments in the array.
4914 */
4915static DECLCALLBACK(int) pgmR3CmdSyncAlways(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
4916{
4917 /** @todo SMP support!! */
4918 PVMCPU pVCpu = &pVM->aCpus[0];
4919
4920 /*
4921 * Validate input.
4922 */
4923 if (!pVM)
4924 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires a VM to be selected.\n");
4925
4926 /*
4927 * Force page directory sync.
4928 */
4929 if (pVCpu->pgm.s.fSyncFlags & PGM_SYNC_ALWAYS)
4930 {
4931 ASMAtomicAndU32(&pVCpu->pgm.s.fSyncFlags, ~PGM_SYNC_ALWAYS);
4932 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Disabled permanent forced page directory syncing.\n");
4933 }
4934 else
4935 {
4936 ASMAtomicOrU32(&pVCpu->pgm.s.fSyncFlags, PGM_SYNC_ALWAYS);
4937 VMCPU_FF_SET(pVCpu, VMCPU_FF_PGM_SYNC_CR3);
4938 return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Enabled permanent forced page directory syncing.\n");
4939 }
4940}
4941
4942#endif /* VBOX_WITH_DEBUGGER */
4943
4944/**
4945 * pvUser argument of the pgmR3CheckIntegrity*Node callbacks.
4946 */
4947typedef struct PGMCHECKINTARGS
4948{
4949 bool fLeftToRight; /**< true: left-to-right; false: right-to-left. */
4950 PPGMPHYSHANDLER pPrevPhys;
4951 PPGMVIRTHANDLER pPrevVirt;
4952 PPGMPHYS2VIRTHANDLER pPrevPhys2Virt;
4953 PVM pVM;
4954} PGMCHECKINTARGS, *PPGMCHECKINTARGS;
4955
4956/**
4957 * Validate a node in the physical handler tree.
4958 *
4959 * @returns 0 on if ok, other wise 1.
4960 * @param pNode The handler node.
4961 * @param pvUser pVM.
4962 */
4963static DECLCALLBACK(int) pgmR3CheckIntegrityPhysHandlerNode(PAVLROGCPHYSNODECORE pNode, void *pvUser)
4964{
4965 PPGMCHECKINTARGS pArgs = (PPGMCHECKINTARGS)pvUser;
4966 PPGMPHYSHANDLER pCur = (PPGMPHYSHANDLER)pNode;
4967 AssertReleaseReturn(!((uintptr_t)pCur & 7), 1);
4968 AssertReleaseMsg(pCur->Core.Key <= pCur->Core.KeyLast,("pCur=%p %RGp-%RGp %s\n", pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
4969 AssertReleaseMsg( !pArgs->pPrevPhys
4970 || (pArgs->fLeftToRight ? pArgs->pPrevPhys->Core.KeyLast < pCur->Core.Key : pArgs->pPrevPhys->Core.KeyLast > pCur->Core.Key),
4971 ("pPrevPhys=%p %RGp-%RGp %s\n"
4972 " pCur=%p %RGp-%RGp %s\n",
4973 pArgs->pPrevPhys, pArgs->pPrevPhys->Core.Key, pArgs->pPrevPhys->Core.KeyLast, pArgs->pPrevPhys->pszDesc,
4974 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
4975 pArgs->pPrevPhys = pCur;
4976 return 0;
4977}
4978
4979
4980/**
4981 * Validate a node in the virtual handler tree.
4982 *
4983 * @returns 0 on if ok, other wise 1.
4984 * @param pNode The handler node.
4985 * @param pvUser pVM.
4986 */
4987static DECLCALLBACK(int) pgmR3CheckIntegrityVirtHandlerNode(PAVLROGCPTRNODECORE pNode, void *pvUser)
4988{
4989 PPGMCHECKINTARGS pArgs = (PPGMCHECKINTARGS)pvUser;
4990 PPGMVIRTHANDLER pCur = (PPGMVIRTHANDLER)pNode;
4991 AssertReleaseReturn(!((uintptr_t)pCur & 7), 1);
4992 AssertReleaseMsg(pCur->Core.Key <= pCur->Core.KeyLast,("pCur=%p %RGv-%RGv %s\n", pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
4993 AssertReleaseMsg( !pArgs->pPrevVirt
4994 || (pArgs->fLeftToRight ? pArgs->pPrevVirt->Core.KeyLast < pCur->Core.Key : pArgs->pPrevVirt->Core.KeyLast > pCur->Core.Key),
4995 ("pPrevVirt=%p %RGv-%RGv %s\n"
4996 " pCur=%p %RGv-%RGv %s\n",
4997 pArgs->pPrevVirt, pArgs->pPrevVirt->Core.Key, pArgs->pPrevVirt->Core.KeyLast, pArgs->pPrevVirt->pszDesc,
4998 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
4999 for (unsigned iPage = 0; iPage < pCur->cPages; iPage++)
5000 {
5001 AssertReleaseMsg(pCur->aPhysToVirt[iPage].offVirtHandler == -RT_OFFSETOF(PGMVIRTHANDLER, aPhysToVirt[iPage]),
5002 ("pCur=%p %RGv-%RGv %s\n"
5003 "iPage=%d offVirtHandle=%#x expected %#x\n",
5004 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc,
5005 iPage, pCur->aPhysToVirt[iPage].offVirtHandler, -RT_OFFSETOF(PGMVIRTHANDLER, aPhysToVirt[iPage])));
5006 }
5007 pArgs->pPrevVirt = pCur;
5008 return 0;
5009}
5010
5011
5012/**
5013 * Validate a node in the virtual handler tree.
5014 *
5015 * @returns 0 on if ok, other wise 1.
5016 * @param pNode The handler node.
5017 * @param pvUser pVM.
5018 */
5019static DECLCALLBACK(int) pgmR3CheckIntegrityPhysToVirtHandlerNode(PAVLROGCPHYSNODECORE pNode, void *pvUser)
5020{
5021 PPGMCHECKINTARGS pArgs = (PPGMCHECKINTARGS)pvUser;
5022 PPGMPHYS2VIRTHANDLER pCur = (PPGMPHYS2VIRTHANDLER)pNode;
5023 AssertReleaseMsgReturn(!((uintptr_t)pCur & 3), ("\n"), 1);
5024 AssertReleaseMsgReturn(!(pCur->offVirtHandler & 3), ("\n"), 1);
5025 AssertReleaseMsg(pCur->Core.Key <= pCur->Core.KeyLast,("pCur=%p %RGp-%RGp\n", pCur, pCur->Core.Key, pCur->Core.KeyLast));
5026 AssertReleaseMsg( !pArgs->pPrevPhys2Virt
5027 || (pArgs->fLeftToRight ? pArgs->pPrevPhys2Virt->Core.KeyLast < pCur->Core.Key : pArgs->pPrevPhys2Virt->Core.KeyLast > pCur->Core.Key),
5028 ("pPrevPhys2Virt=%p %RGp-%RGp\n"
5029 " pCur=%p %RGp-%RGp\n",
5030 pArgs->pPrevPhys2Virt, pArgs->pPrevPhys2Virt->Core.Key, pArgs->pPrevPhys2Virt->Core.KeyLast,
5031 pCur, pCur->Core.Key, pCur->Core.KeyLast));
5032 AssertReleaseMsg( !pArgs->pPrevPhys2Virt
5033 || (pArgs->fLeftToRight ? pArgs->pPrevPhys2Virt->Core.KeyLast < pCur->Core.Key : pArgs->pPrevPhys2Virt->Core.KeyLast > pCur->Core.Key),
5034 ("pPrevPhys2Virt=%p %RGp-%RGp\n"
5035 " pCur=%p %RGp-%RGp\n",
5036 pArgs->pPrevPhys2Virt, pArgs->pPrevPhys2Virt->Core.Key, pArgs->pPrevPhys2Virt->Core.KeyLast,
5037 pCur, pCur->Core.Key, pCur->Core.KeyLast));
5038 AssertReleaseMsg((pCur->offNextAlias & (PGMPHYS2VIRTHANDLER_IN_TREE | PGMPHYS2VIRTHANDLER_IS_HEAD)) == (PGMPHYS2VIRTHANDLER_IN_TREE | PGMPHYS2VIRTHANDLER_IS_HEAD),
5039 ("pCur=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
5040 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias));
5041 if (pCur->offNextAlias & PGMPHYS2VIRTHANDLER_OFF_MASK)
5042 {
5043 PPGMPHYS2VIRTHANDLER pCur2 = pCur;
5044 for (;;)
5045 {
5046 pCur2 = (PPGMPHYS2VIRTHANDLER)((intptr_t)pCur + (pCur->offNextAlias & PGMPHYS2VIRTHANDLER_OFF_MASK));
5047 AssertReleaseMsg(pCur2 != pCur,
5048 (" pCur=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
5049 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias));
5050 AssertReleaseMsg((pCur2->offNextAlias & (PGMPHYS2VIRTHANDLER_IN_TREE | PGMPHYS2VIRTHANDLER_IS_HEAD)) == PGMPHYS2VIRTHANDLER_IN_TREE,
5051 (" pCur=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n"
5052 "pCur2=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
5053 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias,
5054 pCur2, pCur2->Core.Key, pCur2->Core.KeyLast, pCur2->offVirtHandler, pCur2->offNextAlias));
5055 AssertReleaseMsg((pCur2->Core.Key ^ pCur->Core.Key) < PAGE_SIZE,
5056 (" pCur=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n"
5057 "pCur2=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
5058 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias,
5059 pCur2, pCur2->Core.Key, pCur2->Core.KeyLast, pCur2->offVirtHandler, pCur2->offNextAlias));
5060 AssertReleaseMsg((pCur2->Core.KeyLast ^ pCur->Core.KeyLast) < PAGE_SIZE,
5061 (" pCur=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n"
5062 "pCur2=%p:{.Core.Key=%RGp, .Core.KeyLast=%RGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
5063 pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias,
5064 pCur2, pCur2->Core.Key, pCur2->Core.KeyLast, pCur2->offVirtHandler, pCur2->offNextAlias));
5065 if (!(pCur2->offNextAlias & PGMPHYS2VIRTHANDLER_OFF_MASK))
5066 break;
5067 }
5068 }
5069
5070 pArgs->pPrevPhys2Virt = pCur;
5071 return 0;
5072}
5073
5074
5075/**
5076 * Perform an integrity check on the PGM component.
5077 *
5078 * @returns VINF_SUCCESS if everything is fine.
5079 * @returns VBox error status after asserting on integrity breach.
5080 * @param pVM The VM handle.
5081 */
5082VMMR3DECL(int) PGMR3CheckIntegrity(PVM pVM)
5083{
5084 AssertReleaseReturn(pVM->pgm.s.offVM, VERR_INTERNAL_ERROR);
5085
5086 /*
5087 * Check the trees.
5088 */
5089 int cErrors = 0;
5090 const static PGMCHECKINTARGS s_LeftToRight = { true, NULL, NULL, NULL, pVM };
5091 const static PGMCHECKINTARGS s_RightToLeft = { false, NULL, NULL, NULL, pVM };
5092 PGMCHECKINTARGS Args = s_LeftToRight;
5093 cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesR3->PhysHandlers, true, pgmR3CheckIntegrityPhysHandlerNode, &Args);
5094 Args = s_RightToLeft;
5095 cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesR3->PhysHandlers, false, pgmR3CheckIntegrityPhysHandlerNode, &Args);
5096 Args = s_LeftToRight;
5097 cErrors += RTAvlroGCPtrDoWithAll( &pVM->pgm.s.pTreesR3->VirtHandlers, true, pgmR3CheckIntegrityVirtHandlerNode, &Args);
5098 Args = s_RightToLeft;
5099 cErrors += RTAvlroGCPtrDoWithAll( &pVM->pgm.s.pTreesR3->VirtHandlers, false, pgmR3CheckIntegrityVirtHandlerNode, &Args);
5100 Args = s_LeftToRight;
5101 cErrors += RTAvlroGCPtrDoWithAll( &pVM->pgm.s.pTreesR3->HyperVirtHandlers, true, pgmR3CheckIntegrityVirtHandlerNode, &Args);
5102 Args = s_RightToLeft;
5103 cErrors += RTAvlroGCPtrDoWithAll( &pVM->pgm.s.pTreesR3->HyperVirtHandlers, false, pgmR3CheckIntegrityVirtHandlerNode, &Args);
5104 Args = s_LeftToRight;
5105 cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesR3->PhysToVirtHandlers, true, pgmR3CheckIntegrityPhysToVirtHandlerNode, &Args);
5106 Args = s_RightToLeft;
5107 cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesR3->PhysToVirtHandlers, false, pgmR3CheckIntegrityPhysToVirtHandlerNode, &Args);
5108
5109 return !cErrors ? VINF_SUCCESS : VERR_INTERNAL_ERROR;
5110}
5111
5112
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