PGM.cpp revision 347af7572b782ff8715f824f5d70e76a15e3a4d1
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync/* $Id$ */
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync/** @file
5b281ba489ca18f0380d7efc7a5108b606cce449vboxsync * PGM - Page Manager and Monitor. (Mixing stuff here, not good?)
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync */
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync/*
1c94c0a63ba68be1a7b2c640e70d7a06464e4fcavboxsync * Copyright (C) 2006-2007 innotek GmbH
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * This file is part of VirtualBox Open Source Edition (OSE), as
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * available from http://www.virtualbox.org. This file is free software;
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * you can redistribute it and/or modify it under the terms of the GNU
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * General Public License as published by the Free Software Foundation,
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * in version 2 as it comes in the "COPYING" file of the VirtualBox OSE
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * distribution. VirtualBox OSE is distributed in the hope that it will
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * be useful, but WITHOUT ANY WARRANTY of any kind.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync */
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync/** @page pg_pgm PGM - The Page Manager and Monitor
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @section sec_pgm_modes Paging Modes
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
1c94c0a63ba68be1a7b2c640e70d7a06464e4fcavboxsync * There are three memory contexts: Host Context (HC), Guest Context (GC)
1c94c0a63ba68be1a7b2c640e70d7a06464e4fcavboxsync * and intermediate context. When talking about paging HC can also be refered to
1c94c0a63ba68be1a7b2c640e70d7a06464e4fcavboxsync * as "host paging", and GC refered to as "shadow paging".
1c94c0a63ba68be1a7b2c640e70d7a06464e4fcavboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * We define three basic paging modes: 32-bit, PAE and AMD64. The host paging mode
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * is defined by the host operating system. The mode used in the shadow paging mode
3aad980b92149dd95a1ab72ddb8d11d61a28ace6vboxsync * depends on the host paging mode and what the mode the guest is currently in. The
3aad980b92149dd95a1ab72ddb8d11d61a28ace6vboxsync * following relation between the two is defined:
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
3aad980b92149dd95a1ab72ddb8d11d61a28ace6vboxsync * @verbatim
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync Host > 32-bit | PAE | AMD64 |
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync Guest | | | |
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync ==v================================
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync 32-bit 32-bit PAE PAE
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync -------|--------|--------|--------|
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync PAE PAE PAE PAE
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync -------|--------|--------|--------|
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync AMD64 AMD64 AMD64 AMD64
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync -------|--------|--------|--------| @endverbatim
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync *
0a7b20727716a00270f358a1c546473d8c36e8f3vboxsync * All configuration except those in the diagonal (upper left) are expected to
3aad980b92149dd95a1ab72ddb8d11d61a28ace6vboxsync * require special effort from the switcher (i.e. a bit slower).
3aad980b92149dd95a1ab72ddb8d11d61a28ace6vboxsync *
3aad980b92149dd95a1ab72ddb8d11d61a28ace6vboxsync *
3aad980b92149dd95a1ab72ddb8d11d61a28ace6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @section sec_pgm_shw The Shadow Memory Context
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * [..]
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Because of guest context mappings requires PDPTR and PML4 entries to allow
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * writing on AMD64, the two upper levels will have fixed flags whatever the
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * guest is thinking of using there. So, when shadowing the PD level we will
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * calculate the effective flags of PD and all the higher levels. In legacy
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * PAE mode this only applies to the PWT and PCD bits (the rest are
7ccd30dd4bbced565b32c255a11640cd4093abb6vboxsync * ignored/reserved/MBZ). We will ignore those bits for the present.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @section sec_pgm_int The Intermediate Memory Context
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The world switch goes thru an intermediate memory context which purpose it is
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * to provide different mappings of the switcher code. All guest mappings are also
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * present in this context.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The switcher code is mapped at the same location as on the host, at an
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * identity mapped location (physical equals virtual address), and at the
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * hypervisor location.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * PGM maintain page tables for 32-bit, PAE and AMD64 paging modes. This
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * simplifies switching guest CPU mode and consistency at the cost of more
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * code to do the work. All memory use for those page tables is located below
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * 4GB (this includes page tables for guest context mappings).
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgm_int_gc Guest Context Mappings
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * During assignment and relocation of a guest context mapping the intermediate
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * memory context is used to verify the new location.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Guest context mappings are currently restricted to below 4GB, for reasons
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * of simplicity. This may change when we implement AMD64 support.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @section sec_pgm_misc Misc
d605d5391db09e6395a1c091f148f4b86af84bd3vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgm_misc_diff Differences Between Legacy PAE and Long Mode PAE
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * The differences between legacy PAE and long mode PAE are:
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * -# PDPE bits 1, 2, 5 and 6 are defined differently. In leagcy mode they are
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * all marked down as must-be-zero, while in long mode 1, 2 and 5 have the
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * usual meanings while 6 is ignored (AMD). This means that upon switching to
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * legacy PAE mode we'll have to clear these bits and when going to long mode
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * they must be set. This applies to both intermediate and shadow contexts,
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * however we don't need to do it for the intermediate one since we're
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * executing with CR0.WP at that time.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * -# CR3 allows a 32-byte aligned address in legacy mode, while in long mode
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * a page aligned one is required.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync */
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync/** @page pg_pgmPhys PGMPhys - Physical Guest Memory Management.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Objectives:
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - Guest RAM over-commitment using memory ballooning,
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * zero pages and general page sharing.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - Moving or mirroring a VM onto a different physical machine.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgmPhys_Definitions Definitions
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Allocation chunk - A RTR0MemObjAllocPhysNC object and the tracking
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * machinery assoicated with it.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * @subsection subsec_pgmPhys_AllocPage Allocating a page.
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * Initially we map *all* guest memory to the (per VM) zero page, which
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * means that none of the read functions will cause pages to be allocated.
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * Exception, access bit in page tables that have been shared. This must
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * be handled, but we must also make sure PGMGst*Modify doesn't make
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * unnecessary modifications.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync * Allocation points:
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync * - PGMPhysWriteGCPhys and PGMPhysWrite.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - Replacing a zero page mapping at \#PF.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * - Replacing a shared page mapping at \#PF.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * - ROM registration (currently MMR3RomRegister).
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * - VM restore (pgmR3Load).
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync *
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * For the first three it would make sense to keep a few pages handy
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * until we've reached the max memory commitment for the VM.
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync *
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * For the ROM registration, we know exactly how many pages we need
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * and will request these from ring-0. For restore, we will save
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * the number of non-zero pages in the saved state and allocate
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * them up front. This would allow the ring-0 component to refuse
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * the request if the isn't sufficient memory available for VM use.
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync *
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * Btw. for both ROM and restore allocations we won't be requiring
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * zeroed pages as they are going to be filled instantly.
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync *
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync *
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync * @subsection subsec_pgmPhys_FreePage Freeing a page
2e42e0850e182e37277fe28ba5b5d1c37018e783vboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * There are a few points where a page can be freed:
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * - After being replaced by the zero page.
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * - After being replaced by a shared page.
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * - After being ballooned by the guest additions.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - At reset.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - At restore.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * When freeing one or more pages they will be returned to the ring-0
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * component and replaced by the zero page.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * The reasoning for clearing out all the pages on reset is that it will
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * return us to the exact same state as on power on, and may thereby help
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * us reduce the memory load on the system. Further it might have a
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * (temporary) positive influence on memory fragmentation (@see subsec_pgmPhys_Fragmentation).
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * On restore, as mention under the allocation topic, pages should be
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * freed / allocated depending on how many is actually required by the
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * new VM state. The simplest approach is to do like on reset, and free
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * all non-ROM pages and then allocate what we need.
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * A measure to prevent some fragmentation, would be to let each allocation
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * chunk have some affinity towards the VM having allocated the most pages
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * from it. Also, try make sure to allocate from allocation chunks that
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * are almost full. Admittedly, both these measures might work counter to
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * our intentions and its probably not worth putting a lot of effort,
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync * cpu time or memory into this.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgmPhys_SharePage Sharing a page
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The basic idea is that there there will be a idle priority kernel
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * thread walking the non-shared VM pages hashing them and looking for
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * pages with the same checksum. If such pages are found, it will compare
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * them byte-by-byte to see if they actually are identical. If found to be
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * identical it will allocate a shared page, copy the content, check that
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * the page didn't change while doing this, and finally request both the
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * VMs to use the shared page instead. If the page is all zeros (special
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync * checksum and byte-by-byte check) it will request the VM that owns it
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * to replace it with the zero page.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * To make this efficient, we will have to make sure not to try share a page
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * that will change its contents soon. This part requires the most work.
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * A simple idea would be to request the VM to write monitor the page for
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * a while to make sure it isn't modified any time soon. Also, it may
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * make sense to skip pages that are being write monitored since this
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * information is readily available to the thread if it works on the
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * per-VM guest memory structures (presently called PGMRAMRANGE).
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgmPhys_Fragmentation Fragmentation Concerns and Counter Measures
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * The pages are organized in allocation chunks in ring-0, this is a necessity
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * if we wish to have an OS agnostic approach to this whole thing. (On Linux we
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * could easily work on a page-by-page basis if we liked. Whether this is possible
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * or efficient on NT I don't quite know.) Fragmentation within these chunks may
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * become a problem as part of the idea here is that we wish to return memory to
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * the host system.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * For instance, starting two VMs at the same time, they will both allocate the
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * guest memory on-demand and if permitted their page allocations will be
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * intermixed. Shut down one of the two VMs and it will be difficult to return
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * any memory to the host system because the page allocation for the two VMs are
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * mixed up in the same allocation chunks.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync * To further complicate matters, when pages are freed because they have been
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * ballooned or become shared/zero the whole idea is that the page is supposed
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * to be reused by another VM or returned to the host system. This will cause
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * allocation chunks to contain pages belonging to different VMs and prevent
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync * returning memory to the host when one of those VM shuts down.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The only way to really deal with this problem is to move pages. This can
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * either be done at VM shutdown and or by the idle priority worker thread
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * that will be responsible for finding sharable/zero pages. The mechanisms
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * involved for coercing a VM to move a page (or to do it for it) will be
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * the same as when telling it to share/zero a page.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgmPhys_Tracking Tracking Structures And Their Cost
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * There's a difficult balance between keeping the per-page tracking structures
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * (global and guest page) easy to use and keeping them from eating too much
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * memory. We have limited virtual memory resources available when operating in
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * 32-bit kernel space (on 64-bit there'll it's quite a different story). The
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * tracking structures will be attemted designed such that we can deal with up
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * to 32GB of memory on a 32-bit system and essentially unlimited on 64-bit ones.
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsubsection subsubsec_pgmPhys_Tracking_Kernel Kernel Space
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @see pg_GMM
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsubsection subsubsec_pgmPhys_Tracking_PerVM Per-VM
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Fixed info is the physical address of the page (HCPhys) and the page id
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * (described above). Theoretically we'll need 48(-12) bits for the HCPhys part.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Today we've restricting ourselves to 40(-12) bits because this is the current
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * restrictions of all AMD64 implementations (I think Barcelona will up this
b79e4344bf4eb8033fd06d560cd864192728bd0bvboxsync * to 48(-12) bits, not that it really matters) and I needed the bits for
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * tracking mappings of a page. 48-12 = 36. That leaves 28 bits, which means a
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * decent range for the page id: 2^(28+12) = 1024TB.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * In additions to these, we'll have to keep maintaining the page flags as we
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * currently do. Although it wouldn't harm to optimize these quite a bit, like
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * for instance the ROM shouldn't depend on having a write handler installed
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * in order for it to become read-only. A RO/RW bit should be considered so
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * that the page syncing code doesn't have to mess about checking multiple
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * flag combinations (ROM || RW handler || write monitored) in order to
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * figure out how to setup a shadow PTE. But this of course, is second
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * priority at present. Current this requires 12 bits, but could probably
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * be optimized to ~8.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
8fdb63a0d23d1618724f651b8c3d11be48b44d35vboxsync * Then there's the 24 bits used to track which shadow page tables are
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * currently mapping a page for the purpose of speeding up physical
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * access handlers, and thereby the page pool cache. More bit for this
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * purpose wouldn't hurt IIRC.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Then there is a new bit in which we need to record what kind of page
d605d5391db09e6395a1c091f148f4b86af84bd3vboxsync * this is, shared, zero, normal or write-monitored-normal. This'll
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * require 2 bits. One bit might be needed for indicating whether a
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * write monitored page has been written to. And yet another one or
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * two for tracking migration status. 3-4 bits total then.
7ccd30dd4bbced565b32c255a11640cd4093abb6vboxsync *
e190faad3061288ae099cd3ea8a858bd224c00a7vboxsync * Whatever is left will can be used to record the sharabilitiy of a
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * page. The page checksum will not be stored in the per-VM table as
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * the idle thread will not be permitted to do modifications to it.
d605d5391db09e6395a1c091f148f4b86af84bd3vboxsync * It will instead have to keep its own working set of potentially
d605d5391db09e6395a1c091f148f4b86af84bd3vboxsync * shareable pages and their check sums and stuff.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
259e900197d37d0f84ed72e1d4a4eb6c11b4e301vboxsync * For the present we'll keep the current packing of the
259e900197d37d0f84ed72e1d4a4eb6c11b4e301vboxsync * PGMRAMRANGE::aHCPhys to keep the changes simple, only of course,
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * we'll have to change it to a struct with a total of 128-bits at
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * our disposal.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The initial layout will be like this:
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @verbatim
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync RTHCPHYS HCPhys; The current stuff.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync 63:40 Current shadow PT tracking stuff.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync 39:12 The physical page frame number.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync 11:0 The current flags.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync uint32_t u28PageId : 28; The page id.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync uint32_t u2State : 2; The page state { zero, shared, normal, write monitored }.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync uint32_t fWrittenTo : 1; Whether a write monitored page was written to.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync uint32_t u1Reserved : 1; Reserved for later.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync uint32_t u32Reserved; Reserved for later, mostly sharing stats.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync @endverbatim
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The final layout will be something like this:
8fdb63a0d23d1618724f651b8c3d11be48b44d35vboxsync * @verbatim
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync RTHCPHYS HCPhys; The current stuff.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync 63:48 High page id (12+).
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync 47:12 The physical page frame number.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync 11:0 Low page id.
7e10aea6606a51d35041e5a85f9e4f1bd19c4062vboxsync uint32_t fReadOnly : 1; Whether it's readonly page (rom or monitored in some way).
e190faad3061288ae099cd3ea8a858bd224c00a7vboxsync uint32_t u3Type : 3; The page type {RESERVED, MMIO, MMIO2, ROM, shadowed ROM, RAM}.
7e10aea6606a51d35041e5a85f9e4f1bd19c4062vboxsync uint32_t u2PhysMon : 2; Physical access handler type {none, read, write, all}.
0a79c9258d8fae34fa527f125009ab507561b4edvboxsync uint32_t u2VirtMon : 2; Virtual access handler type {none, read, write, all}..
0a79c9258d8fae34fa527f125009ab507561b4edvboxsync uint32_t u2State : 2; The page state { zero, shared, normal, write monitored }.
7e10aea6606a51d35041e5a85f9e4f1bd19c4062vboxsync uint32_t fWrittenTo : 1; Whether a write monitored page was written to.
259e900197d37d0f84ed72e1d4a4eb6c11b4e301vboxsync uint32_t u20Reserved : 20; Reserved for later, mostly sharing stats.
259e900197d37d0f84ed72e1d4a4eb6c11b4e301vboxsync uint32_t u32Tracking; The shadow PT tracking stuff, roughly.
7e10aea6606a51d35041e5a85f9e4f1bd19c4062vboxsync @endverbatim
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * Cost wise, this means we'll double the cost for guest memory. There isn't anyway
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * around that I'm afraid. It means that the cost of dealing out 32GB of memory
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * to one or more VMs is: (32GB >> PAGE_SHIFT) * 16 bytes, or 128MBs. Or another
8c19165bdee6b3a436eba73d9e30fc59f360e313vboxsync * example, the VM heap cost when assigning 1GB to a VM will be: 4MB.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * A couple of cost examples for the total cost per-VM + kernel.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * 32-bit Windows and 32-bit linux:
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * 1GB guest ram, 256K pages: 4MB + 2MB(+) = 6MB
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * 4GB guest ram, 1M pages: 16MB + 8MB(+) = 24MB
89dfdbb56cf9dddad3c7685b41bda1e4e4c1d6f9vboxsync * 32GB guest ram, 8M pages: 128MB + 64MB(+) = 192MB
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * 64-bit Windows and 64-bit linux:
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * 1GB guest ram, 256K pages: 4MB + 3MB(+) = 7MB
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * 4GB guest ram, 1M pages: 16MB + 12MB(+) = 28MB
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * 32GB guest ram, 8M pages: 128MB + 96MB(+) = 224MB
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgmPhys_Serializing Serializing Access
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * Initially, we'll try a simple scheme:
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - The per-VM RAM tracking structures (PGMRAMRANGE) is only modified
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * by the EMT thread of that VM while in the pgm critsect.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * - Other threads in the VM process that needs to make reliable use of
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * the per-VM RAM tracking structures will enter the critsect.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - No process external thread or kernel thread will ever try enter
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * the pgm critical section, as that just won't work.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - The idle thread (and similar threads) doesn't not need 100% reliable
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * data when performing it tasks as the EMT thread will be the one to
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * do the actual changes later anyway. So, as long as it only accesses
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * the main ram range, it can do so by somehow preventing the VM from
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * being destroyed while it works on it...
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - The over-commitment management, including the allocating/freeing
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * chunks, is serialized by a ring-0 mutex lock (a fast one since the
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * more mundane mutex implementation is broken on Linux).
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * - A separeate mutex is protecting the set of allocation chunks so
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * that pages can be shared or/and freed up while some other VM is
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * allocating more chunks. This mutex can be take from under the other
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * one, but not the otherway around.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * @subsection subsec_pgmPhys_Request VM Request interface
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * When in ring-0 it will become necessary to send requests to a VM so it can
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * for instance move a page while defragmenting during VM destroy. The idle
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * thread will make use of this interface to request VMs to setup shared
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * pages and to perform write monitoring of pages.
b79e4344bf4eb8033fd06d560cd864192728bd0bvboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * I would propose an interface similar to the current VMReq interface, similar
ab93606043a9881487aa83be04191d2f4ea24071vboxsync * in that it doesn't require locking and that the one sending the request may
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * wait for completion if it wishes to. This shouldn't be very difficult to
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * realize.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The requests themselves are also pretty simple. They are basically:
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * -# Check that some precondition is still true.
b79e4344bf4eb8033fd06d560cd864192728bd0bvboxsync * -# Do the update.
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * -# Update all shadow page tables involved with the page.
ab93606043a9881487aa83be04191d2f4ea24071vboxsync *
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * The 3rd step is identical to what we're already doing when updating a
9e5c26690d45216629b5f588aced8fcfb68c23b6vboxsync * physical handler, see pgmHandlerPhysicalSetRamFlagsAndFlushShadowPTs.
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync *
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync *
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * @section sec_pgmPhys_MappingCaches Mapping Caches
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * In order to be able to map in and out memory and to be able to support
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * guest with more RAM than we've got virtual address space, we'll employing
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * a mapping cache. There is already a tiny one for GC (see PGMGCDynMapGCPageEx)
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * and we'll create a similar one for ring-0 unless we decide to setup a dedicate
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * memory context for the HWACCM execution.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync *
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * @subsection subsec_pgmPhys_MappingCaches_R3 Ring-3
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync *
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * We've considered implementing the ring-3 mapping cache page based but found
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * that this was bother some when one had to take into account TLBs+SMP and
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * portability (missing the necessary APIs on several platforms). There were
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * also some performance concerns with this approach which hadn't quite been
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * worked out.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * Instead, we'll be mapping allocation chunks into the VM process. This simplifies
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * matters greatly quite a bit since we don't need to invent any new ring-0 stuff,
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * only some minor RTR0MEMOBJ mapping stuff. The main concern here is that mapping
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * compared to the previous idea is that mapping or unmapping a 1MB chunk is more
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * costly than a single page, although how much more costly is uncertain. We'll
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * try address this by using a very big cache, preferably bigger than the actual
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * VM RAM size if possible. The current VM RAM sizes should give some idea for
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * 32-bit boxes, while on 64-bit we can probably get away with employing an
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * unlimited cache.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * The cache have to parts, as already indicated, the ring-3 side and the
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * ring-0 side.
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync *
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * The ring-0 will be tied to the page allocator since it will operate on the
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * memory objects it contains. It will therefore require the first ring-0 mutex
8e342a5c34610667d2b554cb86f1dc2f38a5313cvboxsync * discussed in @ref subsec_pgmPhys_Serializing. We
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * some double house keeping wrt to who has mapped what I think, since both
333e7ab9cd83b32080826d06ac7b1951c684ccb5vboxsync * VMMR0.r0 and RTR0MemObj will keep track of mapping relataions
d605d5391db09e6395a1c091f148f4b86af84bd3vboxsync *
d605d5391db09e6395a1c091f148f4b86af84bd3vboxsync * The ring-3 part will be protected by the pgm critsect. For simplicity, we'll
* require anyone that desires to do changes to the mapping cache to do that
* from within this critsect. Alternatively, we could employ a separate critsect
* for serializing changes to the mapping cache as this would reduce potential
* contention with other threads accessing mappings unrelated to the changes
* that are in process. We can see about this later, contention will show
* up in the statistics anyway, so it'll be simple to tell.
*
* The organization of the ring-3 part will be very much like how the allocation
* chunks are organized in ring-0, that is in an AVL tree by chunk id. To avoid
* having to walk the tree all the time, we'll have a couple of lookaside entries
* like in we do for I/O ports and MMIO in IOM.
*
* The simplified flow of a PGMPhysRead/Write function:
* -# Enter the PGM critsect.
* -# Lookup GCPhys in the ram ranges and get the Page ID.
* -# Calc the Allocation Chunk ID from the Page ID.
* -# Check the lookaside entries and then the AVL tree for the Chunk ID.
* If not found in cache:
* -# Call ring-0 and request it to be mapped and supply
* a chunk to be unmapped if the cache is maxed out already.
* -# Insert the new mapping into the AVL tree (id + R3 address).
* -# Update the relevant lookaside entry and return the mapping address.
* -# Do the read/write according to monitoring flags and everything.
* -# Leave the critsect.
*
*
* @section sec_pgmPhys_Fallback Fallback
*
* Current all the "second tier" hosts will not support the RTR0MemObjAllocPhysNC
* API and thus require a fallback.
*
* So, when RTR0MemObjAllocPhysNC returns VERR_NOT_SUPPORTED the page allocator
* will return to the ring-3 caller (and later ring-0) and asking it to seed
* the page allocator with some fresh pages (VERR_GMM_SEED_ME). Ring-3 will
* then perform an SUPPageAlloc(cbChunk >> PAGE_SHIFT) call and make a
* "SeededAllocPages" call to ring-0.
*
* The first time ring-0 sees the VERR_NOT_SUPPORTED failure it will disable
* all page sharing (zero page detection will continue). It will also force
* all allocations to come from the VM which seeded the page. Both these
* measures are taken to make sure that there will never be any need for
* mapping anything into ring-3 - everything will be mapped already.
*
* Whether we'll continue to use the current MM locked memory management
* for this I don't quite know (I'd prefer not to and just ditch that all
* togther), we'll see what's simplest to do.
*
*
*
* @section sec_pgmPhys_Changes Changes
*
* Breakdown of the changes involved?
*/
/** Saved state data unit version. */
#define PGM_SAVED_STATE_VERSION 5
/*******************************************************************************
* Header Files *
*******************************************************************************/
#define LOG_GROUP LOG_GROUP_PGM
#include <VBox/dbgf.h>
#include <VBox/pgm.h>
#include <VBox/cpum.h>
#include <VBox/iom.h>
#include <VBox/sup.h>
#include <VBox/mm.h>
#include <VBox/em.h>
#include <VBox/stam.h>
#include <VBox/rem.h>
#include <VBox/dbgf.h>
#include <VBox/rem.h>
#include <VBox/selm.h>
#include <VBox/ssm.h>
#include "PGMInternal.h"
#include <VBox/vm.h>
#include <VBox/dbg.h>
#include <VBox/hwaccm.h>
#include <iprt/assert.h>
#include <iprt/alloc.h>
#include <iprt/asm.h>
#include <iprt/thread.h>
#include <iprt/string.h>
#include <VBox/param.h>
#include <VBox/err.h>
/*******************************************************************************
* Internal Functions *
*******************************************************************************/
static int pgmR3InitPaging(PVM pVM);
static DECLCALLBACK(void) pgmR3PhysInfo(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
static DECLCALLBACK(void) pgmR3InfoMode(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
static DECLCALLBACK(void) pgmR3InfoCr3(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs);
static DECLCALLBACK(int) pgmR3RelocatePhysHandler(PAVLROGCPHYSNODECORE pNode, void *pvUser);
static DECLCALLBACK(int) pgmR3RelocateVirtHandler(PAVLROGCPTRNODECORE pNode, void *pvUser);
#ifdef VBOX_STRICT
static DECLCALLBACK(void) pgmR3ResetNoMorePhysWritesFlag(PVM pVM, VMSTATE enmState, VMSTATE enmOldState, void *pvUser);
#endif
static DECLCALLBACK(int) pgmR3Save(PVM pVM, PSSMHANDLE pSSM);
static DECLCALLBACK(int) pgmR3Load(PVM pVM, PSSMHANDLE pSSM, uint32_t u32Version);
static int pgmR3ModeDataInit(PVM pVM, bool fResolveGCAndR0);
static void pgmR3ModeDataSwitch(PVM pVM, PGMMODE enmShw, PGMMODE enmGst);
static PGMMODE pgmR3CalcShadowMode(PGMMODE enmGuestMode, SUPPAGINGMODE enmHostMode, PGMMODE enmShadowMode, VMMSWITCHER *penmSwitcher);
#ifdef VBOX_WITH_STATISTICS
static void pgmR3InitStats(PVM pVM);
#endif
#ifdef VBOX_WITH_DEBUGGER
/** @todo all but the two last commands must be converted to 'info'. */
static DECLCALLBACK(int) pgmR3CmdRam(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
static DECLCALLBACK(int) pgmR3CmdMap(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
static DECLCALLBACK(int) pgmR3CmdSync(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
static DECLCALLBACK(int) pgmR3CmdSyncAlways(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult);
#endif
/*******************************************************************************
* Global Variables *
*******************************************************************************/
#ifdef VBOX_WITH_DEBUGGER
/** Command descriptors. */
static const DBGCCMD g_aCmds[] =
{
/* pszCmd, cArgsMin, cArgsMax, paArgDesc, cArgDescs, pResultDesc, fFlags, pfnHandler pszSyntax, ....pszDescription */
{ "pgmram", 0, 0, NULL, 0, NULL, 0, pgmR3CmdRam, "", "Display the ram ranges." },
{ "pgmmap", 0, 0, NULL, 0, NULL, 0, pgmR3CmdMap, "", "Display the mapping ranges." },
{ "pgmsync", 0, 0, NULL, 0, NULL, 0, pgmR3CmdSync, "", "Sync the CR3 page." },
{ "pgmsyncalways", 0, 0, NULL, 0, NULL, 0, pgmR3CmdSyncAlways, "", "Toggle permanent CR3 syncing." },
};
#endif
#if 1/// @todo ndef RT_ARCH_AMD64
/*
* Shadow - 32-bit mode
*/
#define PGM_SHW_TYPE PGM_TYPE_32BIT
#define PGM_SHW_NAME(name) PGM_SHW_NAME_32BIT(name)
#define PGM_SHW_NAME_GC_STR(name) PGM_SHW_NAME_GC_32BIT_STR(name)
#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_32BIT_STR(name)
#include "PGMShw.h"
/* Guest - real mode */
#define PGM_GST_TYPE PGM_TYPE_REAL
#define PGM_GST_NAME(name) PGM_GST_NAME_REAL(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_REAL_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_REAL_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_32BIT_REAL(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_32BIT_REAL_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_32BIT_REAL_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_32BIT_PT_FOR_PHYS
#include "PGMGst.h"
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
/* Guest - protected mode */
#define PGM_GST_TYPE PGM_TYPE_PROT
#define PGM_GST_NAME(name) PGM_GST_NAME_PROT(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_PROT_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PROT_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_32BIT_PROT(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_32BIT_PROT_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_32BIT_PROT_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_32BIT_PT_FOR_PHYS
#include "PGMGst.h"
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
/* Guest - 32-bit mode */
#define PGM_GST_TYPE PGM_TYPE_32BIT
#define PGM_GST_NAME(name) PGM_GST_NAME_32BIT(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_32BIT_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_32BIT_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_32BIT_32BIT(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_32BIT_32BIT_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_32BIT_32BIT_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_32BIT_PT_FOR_32BIT_PT
#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_32BIT_PT_FOR_32BIT_4MB
#include "PGMGst.h"
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_BIG
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
#undef PGM_SHW_TYPE
#undef PGM_SHW_NAME
#undef PGM_SHW_NAME_GC_STR
#undef PGM_SHW_NAME_R0_STR
#endif /* !RT_ARCH_AMD64 */
/*
* Shadow - PAE mode
*/
#define PGM_SHW_TYPE PGM_TYPE_PAE
#define PGM_SHW_NAME(name) PGM_SHW_NAME_PAE(name)
#define PGM_SHW_NAME_GC_STR(name) PGM_SHW_NAME_GC_PAE_STR(name)
#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_PAE_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_REAL(name)
#include "PGMShw.h"
/* Guest - real mode */
#define PGM_GST_TYPE PGM_TYPE_REAL
#define PGM_GST_NAME(name) PGM_GST_NAME_REAL(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_REAL_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_REAL_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_REAL(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_PAE_REAL_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_REAL_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
/* Guest - protected mode */
#define PGM_GST_TYPE PGM_TYPE_PROT
#define PGM_GST_NAME(name) PGM_GST_NAME_PROT(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_PROT_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PROT_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_PROT(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_PAE_PROT_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_PROT_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
/* Guest - 32-bit mode */
#define PGM_GST_TYPE PGM_TYPE_32BIT
#define PGM_GST_NAME(name) PGM_GST_NAME_32BIT(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_32BIT_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_32BIT_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_32BIT(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_PAE_32BIT_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_32BIT_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_32BIT_PT
#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_32BIT_4MB
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_BIG
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
/* Guest - PAE mode */
#define PGM_GST_TYPE PGM_TYPE_PAE
#define PGM_GST_NAME(name) PGM_GST_NAME_PAE(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_PAE_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PAE_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_PAE_PAE(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_PAE_PAE_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_PAE_PAE_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
#include "PGMGst.h"
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_BIG
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
#undef PGM_SHW_TYPE
#undef PGM_SHW_NAME
#undef PGM_SHW_NAME_GC_STR
#undef PGM_SHW_NAME_R0_STR
/*
* Shadow - AMD64 mode
*/
#define PGM_SHW_TYPE PGM_TYPE_AMD64
#define PGM_SHW_NAME(name) PGM_SHW_NAME_AMD64(name)
#define PGM_SHW_NAME_GC_STR(name) PGM_SHW_NAME_GC_AMD64_STR(name)
#define PGM_SHW_NAME_R0_STR(name) PGM_SHW_NAME_R0_AMD64_STR(name)
#include "PGMShw.h"
/* Guest - real mode */
#define PGM_GST_TYPE PGM_TYPE_REAL
#define PGM_GST_NAME(name) PGM_GST_NAME_REAL(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_REAL_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_REAL_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_AMD64_REAL(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_AMD64_REAL_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_AMD64_REAL_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
/* Guest - protected mode */
#define PGM_GST_TYPE PGM_TYPE_PROT
#define PGM_GST_NAME(name) PGM_GST_NAME_PROT(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_PROT_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_PROT_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_AMD64_PROT(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_AMD64_PROT_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_AMD64_PROT_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PHYS
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
/* Guest - AMD64 mode */
#define PGM_GST_TYPE PGM_TYPE_AMD64
#define PGM_GST_NAME(name) PGM_GST_NAME_AMD64(name)
#define PGM_GST_NAME_GC_STR(name) PGM_GST_NAME_GC_AMD64_STR(name)
#define PGM_GST_NAME_R0_STR(name) PGM_GST_NAME_R0_AMD64_STR(name)
#define PGM_BTH_NAME(name) PGM_BTH_NAME_AMD64_AMD64(name)
#define PGM_BTH_NAME_GC_STR(name) PGM_BTH_NAME_GC_AMD64_AMD64_STR(name)
#define PGM_BTH_NAME_R0_STR(name) PGM_BTH_NAME_R0_AMD64_AMD64_STR(name)
#define BTH_PGMPOOLKIND_PT_FOR_PT PGMPOOLKIND_PAE_PT_FOR_PAE_PT
#define BTH_PGMPOOLKIND_PT_FOR_BIG PGMPOOLKIND_PAE_PT_FOR_PAE_2MB
#include "PGMGst.h"
#include "PGMBth.h"
#undef BTH_PGMPOOLKIND_PT_FOR_BIG
#undef BTH_PGMPOOLKIND_PT_FOR_PT
#undef PGM_BTH_NAME
#undef PGM_BTH_NAME_GC_STR
#undef PGM_BTH_NAME_R0_STR
#undef PGM_GST_TYPE
#undef PGM_GST_NAME
#undef PGM_GST_NAME_GC_STR
#undef PGM_GST_NAME_R0_STR
#undef PGM_SHW_TYPE
#undef PGM_SHW_NAME
#undef PGM_SHW_NAME_GC_STR
#undef PGM_SHW_NAME_R0_STR
/**
* Initiates the paging of VM.
*
* @returns VBox status code.
* @param pVM Pointer to VM structure.
*/
PGMR3DECL(int) PGMR3Init(PVM pVM)
{
LogFlow(("PGMR3Init:\n"));
/*
* Assert alignment and sizes.
*/
AssertRelease(sizeof(pVM->pgm.s) <= sizeof(pVM->pgm.padding));
/*
* Init the structure.
*/
pVM->pgm.s.offVM = RT_OFFSETOF(VM, pgm.s);
pVM->pgm.s.enmShadowMode = PGMMODE_INVALID;
pVM->pgm.s.enmGuestMode = PGMMODE_INVALID;
pVM->pgm.s.enmHostMode = SUPPAGINGMODE_INVALID;
pVM->pgm.s.GCPhysCR3 = NIL_RTGCPHYS;
pVM->pgm.s.GCPhysGstCR3Monitored = NIL_RTGCPHYS;
pVM->pgm.s.fA20Enabled = true;
pVM->pgm.s.pGstPaePDPTRHC = NULL;
pVM->pgm.s.pGstPaePDPTRGC = 0;
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.apGstPaePDsHC); i++)
{
pVM->pgm.s.apGstPaePDsHC[i] = NULL;
pVM->pgm.s.apGstPaePDsGC[i] = 0;
pVM->pgm.s.aGCPhysGstPaePDs[i] = NIL_RTGCPHYS;
}
#ifdef VBOX_STRICT
VMR3AtStateRegister(pVM, pgmR3ResetNoMorePhysWritesFlag, NULL);
#endif
/*
* Get the configured RAM size - to estimate saved state size.
*/
uint64_t cbRam;
int rc = CFGMR3QueryU64(CFGMR3GetRoot(pVM), "RamSize", &cbRam);
if (rc == VERR_CFGM_VALUE_NOT_FOUND)
cbRam = pVM->pgm.s.cbRamSize = 0;
else if (VBOX_SUCCESS(rc))
{
if (cbRam < PAGE_SIZE)
cbRam = 0;
cbRam = RT_ALIGN_64(cbRam, PAGE_SIZE);
pVM->pgm.s.cbRamSize = (RTUINT)cbRam;
}
else
{
AssertMsgFailed(("Configuration error: Failed to query integer \"RamSize\", rc=%Vrc.\n", rc));
return rc;
}
/*
* Register saved state data unit.
*/
rc = SSMR3RegisterInternal(pVM, "pgm", 1, PGM_SAVED_STATE_VERSION, (size_t)cbRam + sizeof(PGM),
NULL, pgmR3Save, NULL,
NULL, pgmR3Load, NULL);
if (VBOX_FAILURE(rc))
return rc;
/*
* Initialize the PGM critical section and flush the phys TLBs
*/
rc = PDMR3CritSectInit(pVM, &pVM->pgm.s.CritSect, "PGM");
AssertRCReturn(rc, rc);
PGMR3PhysChunkInvalidateTLB(pVM);
PGMPhysInvalidatePageR3MapTLB(pVM);
PGMPhysInvalidatePageR0MapTLB(pVM);
PGMPhysInvalidatePageGCMapTLB(pVM);
/*
* Trees
*/
rc = MMHyperAlloc(pVM, sizeof(PGMTREES), 0, MM_TAG_PGM, (void **)&pVM->pgm.s.pTreesHC);
if (VBOX_SUCCESS(rc))
{
pVM->pgm.s.pTreesGC = MMHyperHC2GC(pVM, pVM->pgm.s.pTreesHC);
/*
* Alocate the zero page.
*/
rc = MMHyperAlloc(pVM, PAGE_SIZE, PAGE_SIZE, MM_TAG_PGM, &pVM->pgm.s.pvZeroPgR3);
}
if (VBOX_SUCCESS(rc))
{
pVM->pgm.s.pvZeroPgGC = MMHyperR3ToGC(pVM, pVM->pgm.s.pvZeroPgR3);
pVM->pgm.s.pvZeroPgR0 = MMHyperR3ToR0(pVM, pVM->pgm.s.pvZeroPgR3);
AssertRelease(pVM->pgm.s.pvZeroPgR0 != NIL_RTHCPHYS);
pVM->pgm.s.HCPhysZeroPg = MMR3HyperHCVirt2HCPhys(pVM, pVM->pgm.s.pvZeroPgR3);
AssertRelease(pVM->pgm.s.HCPhysZeroPg != NIL_RTHCPHYS);
/*
* Init the paging.
*/
rc = pgmR3InitPaging(pVM);
}
if (VBOX_SUCCESS(rc))
{
/*
* Init the page pool.
*/
rc = pgmR3PoolInit(pVM);
}
if (VBOX_SUCCESS(rc))
{
/*
* Info & statistics
*/
DBGFR3InfoRegisterInternal(pVM, "mode",
"Shows the current paging mode. "
"Recognizes 'all', 'guest', 'shadow' and 'host' as arguments, defaulting to 'all' if nothing's given.",
pgmR3InfoMode);
DBGFR3InfoRegisterInternal(pVM, "pgmcr3",
"Dumps all the entries in the top level paging table. No arguments.",
pgmR3InfoCr3);
DBGFR3InfoRegisterInternal(pVM, "phys",
"Dumps all the physical address ranges. No arguments.",
pgmR3PhysInfo);
DBGFR3InfoRegisterInternal(pVM, "handlers",
"Dumps physical and virtual handlers. "
"Pass 'phys' or 'virt' as argument if only one kind is wanted.",
pgmR3InfoHandlers);
STAM_REL_REG(pVM, &pVM->pgm.s.cGuestModeChanges, STAMTYPE_COUNTER, "/PGM/cGuestModeChanges", STAMUNIT_OCCURENCES, "Number of guest mode changes.");
#ifdef VBOX_WITH_STATISTICS
pgmR3InitStats(pVM);
#endif
#ifdef VBOX_WITH_DEBUGGER
/*
* Debugger commands.
*/
static bool fRegisteredCmds = false;
if (!fRegisteredCmds)
{
int rc = DBGCRegisterCommands(&g_aCmds[0], ELEMENTS(g_aCmds));
if (VBOX_SUCCESS(rc))
fRegisteredCmds = true;
}
#endif
return VINF_SUCCESS;
}
/* Almost no cleanup necessary, MM frees all memory. */
PDMR3CritSectDelete(&pVM->pgm.s.CritSect);
return rc;
}
/**
* Init paging.
*
* Since we need to check what mode the host is operating in before we can choose
* the right paging functions for the host we have to delay this until R0 has
* been initialized.
*
* @returns VBox status code.
* @param pVM VM handle.
*/
static int pgmR3InitPaging(PVM pVM)
{
/*
* Force a recalculation of modes and switcher so everyone gets notified.
*/
pVM->pgm.s.enmShadowMode = PGMMODE_INVALID;
pVM->pgm.s.enmGuestMode = PGMMODE_INVALID;
pVM->pgm.s.enmHostMode = SUPPAGINGMODE_INVALID;
/*
* Allocate static mapping space for whatever the cr3 register
* points to and in the case of PAE mode to the 4 PDs.
*/
int rc = MMR3HyperReserve(pVM, PAGE_SIZE * 5, "CR3 mapping", &pVM->pgm.s.GCPtrCR3Mapping);
if (VBOX_FAILURE(rc))
{
AssertMsgFailed(("Failed to reserve two pages for cr mapping in HMA, rc=%Vrc\n", rc));
return rc;
}
MMR3HyperReserve(pVM, PAGE_SIZE, "fence", NULL);
/*
* Allocate pages for the three possible intermediate contexts
* (AMD64, PAE and plain 32-Bit). We maintain all three contexts
* for the sake of simplicity. The AMD64 uses the PAE for the
* lower levels, making the total number of pages 11 (3 + 7 + 1).
*
* We assume that two page tables will be enought for the core code
* mappings (HC virtual and identity).
*/
pVM->pgm.s.pInterPD = (PX86PD)MMR3PageAllocLow(pVM);
pVM->pgm.s.apInterPTs[0] = (PX86PT)MMR3PageAllocLow(pVM);
pVM->pgm.s.apInterPTs[1] = (PX86PT)MMR3PageAllocLow(pVM);
pVM->pgm.s.apInterPaePTs[0] = (PX86PTPAE)MMR3PageAlloc(pVM);
pVM->pgm.s.apInterPaePTs[1] = (PX86PTPAE)MMR3PageAlloc(pVM);
pVM->pgm.s.apInterPaePDs[0] = (PX86PDPAE)MMR3PageAlloc(pVM);
pVM->pgm.s.apInterPaePDs[1] = (PX86PDPAE)MMR3PageAlloc(pVM);
pVM->pgm.s.apInterPaePDs[2] = (PX86PDPAE)MMR3PageAlloc(pVM);
pVM->pgm.s.apInterPaePDs[3] = (PX86PDPAE)MMR3PageAlloc(pVM);
pVM->pgm.s.pInterPaePDPTR = (PX86PDPTR)MMR3PageAllocLow(pVM);
pVM->pgm.s.pInterPaePDPTR64 = (PX86PDPTR)MMR3PageAllocLow(pVM);
pVM->pgm.s.pInterPaePML4 = (PX86PML4)MMR3PageAllocLow(pVM);
if ( !pVM->pgm.s.pInterPD
|| !pVM->pgm.s.apInterPTs[0]
|| !pVM->pgm.s.apInterPTs[1]
|| !pVM->pgm.s.apInterPaePTs[0]
|| !pVM->pgm.s.apInterPaePTs[1]
|| !pVM->pgm.s.apInterPaePDs[0]
|| !pVM->pgm.s.apInterPaePDs[1]
|| !pVM->pgm.s.apInterPaePDs[2]
|| !pVM->pgm.s.apInterPaePDs[3]
|| !pVM->pgm.s.pInterPaePDPTR
|| !pVM->pgm.s.pInterPaePDPTR64
|| !pVM->pgm.s.pInterPaePML4)
{
AssertMsgFailed(("Failed to allocate pages for the intermediate context!\n"));
return VERR_NO_PAGE_MEMORY;
}
pVM->pgm.s.HCPhysInterPD = MMPage2Phys(pVM, pVM->pgm.s.pInterPD);
AssertRelease(pVM->pgm.s.HCPhysInterPD != NIL_RTHCPHYS && !(pVM->pgm.s.HCPhysInterPD & PAGE_OFFSET_MASK));
pVM->pgm.s.HCPhysInterPaePDPTR = MMPage2Phys(pVM, pVM->pgm.s.pInterPaePDPTR);
AssertRelease(pVM->pgm.s.HCPhysInterPaePDPTR != NIL_RTHCPHYS && !(pVM->pgm.s.HCPhysInterPaePDPTR & PAGE_OFFSET_MASK));
pVM->pgm.s.HCPhysInterPaePML4 = MMPage2Phys(pVM, pVM->pgm.s.pInterPaePML4);
AssertRelease(pVM->pgm.s.HCPhysInterPaePML4 != NIL_RTHCPHYS && !(pVM->pgm.s.HCPhysInterPaePML4 & PAGE_OFFSET_MASK));
/*
* Initialize the pages, setting up the PML4 and PDPTR for repetitive 4GB action.
*/
ASMMemZeroPage(pVM->pgm.s.pInterPD);
ASMMemZeroPage(pVM->pgm.s.apInterPTs[0]);
ASMMemZeroPage(pVM->pgm.s.apInterPTs[1]);
ASMMemZeroPage(pVM->pgm.s.apInterPaePTs[0]);
ASMMemZeroPage(pVM->pgm.s.apInterPaePTs[1]);
ASMMemZeroPage(pVM->pgm.s.pInterPaePDPTR);
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.apInterPaePDs); i++)
{
ASMMemZeroPage(pVM->pgm.s.apInterPaePDs[i]);
pVM->pgm.s.pInterPaePDPTR->a[i].u = X86_PDPE_P | PGM_PLXFLAGS_PERMANENT
| MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[i]);
}
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.pInterPaePDPTR64->a); i++)
{
const unsigned iPD = i % ELEMENTS(pVM->pgm.s.apInterPaePDs);
pVM->pgm.s.pInterPaePDPTR64->a[i].u = X86_PDPE_P | X86_PDPE_RW | X86_PDPE_US | X86_PDPE_A | PGM_PLXFLAGS_PERMANENT
| MMPage2Phys(pVM, pVM->pgm.s.apInterPaePDs[iPD]);
}
RTHCPHYS HCPhysInterPaePDPTR64 = MMPage2Phys(pVM, pVM->pgm.s.pInterPaePDPTR64);
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.pInterPaePML4->a); i++)
pVM->pgm.s.pInterPaePML4->a[i].u = X86_PML4E_P | X86_PML4E_RW | X86_PML4E_US | X86_PML4E_A | PGM_PLXFLAGS_PERMANENT
| HCPhysInterPaePDPTR64;
/*
* Allocate pages for the three possible guest contexts (AMD64, PAE and plain 32-Bit).
* We allocate pages for all three posibilities to in order to simplify mappings and
* avoid resource failure during mode switches. So, we need to cover all levels of the
* of the first 4GB down to PD level.
* As with the intermediate context, AMD64 uses the PAE PDPTR and PDs.
*/
pVM->pgm.s.pHC32BitPD = (PX86PD)MMR3PageAllocLow(pVM);
pVM->pgm.s.apHCPaePDs[0] = (PX86PDPAE)MMR3PageAlloc(pVM);
pVM->pgm.s.apHCPaePDs[1] = (PX86PDPAE)MMR3PageAlloc(pVM);
AssertRelease((uintptr_t)pVM->pgm.s.apHCPaePDs[0] + PAGE_SIZE == (uintptr_t)pVM->pgm.s.apHCPaePDs[1]);
pVM->pgm.s.apHCPaePDs[2] = (PX86PDPAE)MMR3PageAlloc(pVM);
AssertRelease((uintptr_t)pVM->pgm.s.apHCPaePDs[1] + PAGE_SIZE == (uintptr_t)pVM->pgm.s.apHCPaePDs[2]);
pVM->pgm.s.apHCPaePDs[3] = (PX86PDPAE)MMR3PageAlloc(pVM);
AssertRelease((uintptr_t)pVM->pgm.s.apHCPaePDs[2] + PAGE_SIZE == (uintptr_t)pVM->pgm.s.apHCPaePDs[3]);
pVM->pgm.s.pHCPaePDPTR = (PX86PDPTR)MMR3PageAllocLow(pVM);
pVM->pgm.s.pHCPaePML4 = (PX86PML4)MMR3PageAllocLow(pVM);
if ( !pVM->pgm.s.pHC32BitPD
|| !pVM->pgm.s.apHCPaePDs[0]
|| !pVM->pgm.s.apHCPaePDs[1]
|| !pVM->pgm.s.apHCPaePDs[2]
|| !pVM->pgm.s.apHCPaePDs[3]
|| !pVM->pgm.s.pHCPaePDPTR
|| !pVM->pgm.s.pHCPaePML4)
{
AssertMsgFailed(("Failed to allocate pages for the intermediate context!\n"));
return VERR_NO_PAGE_MEMORY;
}
/* get physical addresses. */
pVM->pgm.s.HCPhys32BitPD = MMPage2Phys(pVM, pVM->pgm.s.pHC32BitPD);
Assert(MMPagePhys2Page(pVM, pVM->pgm.s.HCPhys32BitPD) == pVM->pgm.s.pHC32BitPD);
pVM->pgm.s.aHCPhysPaePDs[0] = MMPage2Phys(pVM, pVM->pgm.s.apHCPaePDs[0]);
pVM->pgm.s.aHCPhysPaePDs[1] = MMPage2Phys(pVM, pVM->pgm.s.apHCPaePDs[1]);
pVM->pgm.s.aHCPhysPaePDs[2] = MMPage2Phys(pVM, pVM->pgm.s.apHCPaePDs[2]);
pVM->pgm.s.aHCPhysPaePDs[3] = MMPage2Phys(pVM, pVM->pgm.s.apHCPaePDs[3]);
pVM->pgm.s.HCPhysPaePDPTR = MMPage2Phys(pVM, pVM->pgm.s.pHCPaePDPTR);
pVM->pgm.s.HCPhysPaePML4 = MMPage2Phys(pVM, pVM->pgm.s.pHCPaePML4);
/*
* Initialize the pages, setting up the PML4 and PDPTR for action below 4GB.
*/
ASMMemZero32(pVM->pgm.s.pHC32BitPD, PAGE_SIZE);
ASMMemZero32(pVM->pgm.s.pHCPaePDPTR, PAGE_SIZE);
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.apHCPaePDs); i++)
{
ASMMemZero32(pVM->pgm.s.apHCPaePDs[i], PAGE_SIZE);
pVM->pgm.s.pHCPaePDPTR->a[i].u = X86_PDPE_P | PGM_PLXFLAGS_PERMANENT | pVM->pgm.s.aHCPhysPaePDs[i];
/* The flags will be corrected when entering and leaving long mode. */
}
ASMMemZero32(pVM->pgm.s.pHCPaePML4, PAGE_SIZE);
pVM->pgm.s.pHCPaePML4->a[0].u = X86_PML4E_P | X86_PML4E_RW | X86_PML4E_A
| PGM_PLXFLAGS_PERMANENT | pVM->pgm.s.HCPhysPaePDPTR;
CPUMSetHyperCR3(pVM, (uint32_t)pVM->pgm.s.HCPhys32BitPD);
/*
* Initialize paging workers and mode from current host mode
* and the guest running in real mode.
*/
pVM->pgm.s.enmHostMode = SUPGetPagingMode();
switch (pVM->pgm.s.enmHostMode)
{
case SUPPAGINGMODE_32_BIT:
case SUPPAGINGMODE_32_BIT_GLOBAL:
case SUPPAGINGMODE_PAE:
case SUPPAGINGMODE_PAE_GLOBAL:
case SUPPAGINGMODE_PAE_NX:
case SUPPAGINGMODE_PAE_GLOBAL_NX:
break;
case SUPPAGINGMODE_AMD64:
case SUPPAGINGMODE_AMD64_GLOBAL:
case SUPPAGINGMODE_AMD64_NX:
case SUPPAGINGMODE_AMD64_GLOBAL_NX:
#ifndef VBOX_WITH_HYBIRD_32BIT_KERNEL
if (ARCH_BITS != 64)
{
AssertMsgFailed(("Host mode %d (64-bit) is not supported by non-64bit builds\n", pVM->pgm.s.enmHostMode));
LogRel(("Host mode %d (64-bit) is not supported by non-64bit builds\n", pVM->pgm.s.enmHostMode));
return VERR_PGM_UNSUPPORTED_HOST_PAGING_MODE;
}
#endif
break;
default:
AssertMsgFailed(("Host mode %d is not supported\n", pVM->pgm.s.enmHostMode));
return VERR_PGM_UNSUPPORTED_HOST_PAGING_MODE;
}
rc = pgmR3ModeDataInit(pVM, false /* don't resolve GC and R0 syms yet */);
if (VBOX_SUCCESS(rc))
rc = pgmR3ChangeMode(pVM, PGMMODE_REAL);
if (VBOX_SUCCESS(rc))
{
LogFlow(("pgmR3InitPaging: returns successfully\n"));
#if HC_ARCH_BITS == 64
LogRel(("Debug: HCPhys32BitPD=%VHp aHCPhysPaePDs={%VHp,%VHp,%VHp,%VHp} HCPhysPaePDPTR=%VHp HCPhysPaePML4=%VHp\n",
pVM->pgm.s.HCPhys32BitPD, pVM->pgm.s.aHCPhysPaePDs[0], pVM->pgm.s.aHCPhysPaePDs[1], pVM->pgm.s.aHCPhysPaePDs[2], pVM->pgm.s.aHCPhysPaePDs[3],
pVM->pgm.s.HCPhysPaePDPTR, pVM->pgm.s.HCPhysPaePML4));
LogRel(("Debug: HCPhysInterPD=%VHp HCPhysInterPaePDPTR=%VHp HCPhysInterPaePML4=%VHp\n",
pVM->pgm.s.HCPhysInterPD, pVM->pgm.s.HCPhysInterPaePDPTR, pVM->pgm.s.HCPhysInterPaePML4));
LogRel(("Debug: apInterPTs={%VHp,%VHp} apInterPaePTs={%VHp,%VHp} apInterPaePDs={%VHp,%VHp,%VHp,%VHp} pInterPaePDPTR64=%VHp\n",
MMPage2Phys(pVM, pVM->pgm.s.apInterPTs[0]), MMPage2Phys(pVM, pVM->pgm.s.apInterPTs[1]),
MMPage2Phys(pVM, pVM->pgm.s.apInterPaePTs[0]), MMPage2Phys(pVM, pVM->pgm.s.apInterPaePTs[1]),
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]),
MMPage2Phys(pVM, pVM->pgm.s.pInterPaePDPTR64)));
#endif
return VINF_SUCCESS;
}
LogFlow(("pgmR3InitPaging: returns %Vrc\n", rc));
return rc;
}
#ifdef VBOX_WITH_STATISTICS
/**
* Init statistics
*/
static void pgmR3InitStats(PVM pVM)
{
PPGM pPGM = &pVM->pgm.s;
STAM_REG(pVM, &pPGM->StatGCInvalidatePage, STAMTYPE_PROFILE, "/PGM/GC/InvalidatePage", STAMUNIT_TICKS_PER_CALL, "PGMGCInvalidatePage() profiling.");
STAM_REG(pVM, &pPGM->StatGCInvalidatePage4KBPages, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/4KBPages", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was called for a 4KB page.");
STAM_REG(pVM, &pPGM->StatGCInvalidatePage4MBPages, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/4MBPages", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was called for a 4MB page.");
STAM_REG(pVM, &pPGM->StatGCInvalidatePage4MBPagesSkip, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/4MBPagesSkip",STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() skipped a 4MB page.");
STAM_REG(pVM, &pPGM->StatGCInvalidatePagePDMappings, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/PDMappings", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was called for a page directory containing mappings (no conflict).");
STAM_REG(pVM, &pPGM->StatGCInvalidatePagePDNAs, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/PDNAs", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was called for a not accessed page directory.");
STAM_REG(pVM, &pPGM->StatGCInvalidatePagePDNPs, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/PDNPs", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was called for a not present page directory.");
STAM_REG(pVM, &pPGM->StatGCInvalidatePagePDOutOfSync, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/PDOutOfSync", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was called for an out of sync page directory.");
STAM_REG(pVM, &pPGM->StatGCInvalidatePageSkipped, STAMTYPE_COUNTER, "/PGM/GC/InvalidatePage/Skipped", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was skipped due to not present shw or pending pending SyncCR3.");
STAM_REG(pVM, &pPGM->StatGCSyncPT, STAMTYPE_PROFILE, "/PGM/GC/SyncPT", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMGCSyncPT() body.");
STAM_REG(pVM, &pPGM->StatGCAccessedPage, STAMTYPE_COUNTER, "/PGM/GC/AccessedPage", STAMUNIT_OCCURENCES, "The number of pages marked not present for accessed bit emulation.");
STAM_REG(pVM, &pPGM->StatGCDirtyPage, STAMTYPE_COUNTER, "/PGM/GC/DirtyPage/Mark", STAMUNIT_OCCURENCES, "The number of pages marked read-only for dirty bit tracking.");
STAM_REG(pVM, &pPGM->StatGCDirtyPageBig, STAMTYPE_COUNTER, "/PGM/GC/DirtyPage/MarkBig", STAMUNIT_OCCURENCES, "The number of 4MB pages marked read-only for dirty bit tracking.");
STAM_REG(pVM, &pPGM->StatGCDirtyPageTrap, STAMTYPE_COUNTER, "/PGM/GC/DirtyPage/Trap", STAMUNIT_OCCURENCES, "The number of traps generated for dirty bit tracking.");
STAM_REG(pVM, &pPGM->StatGCDirtyPageSkipped, STAMTYPE_COUNTER, "/PGM/GC/DirtyPage/Skipped", STAMUNIT_OCCURENCES, "The number of pages already dirty or readonly.");
STAM_REG(pVM, &pPGM->StatGCDirtiedPage, STAMTYPE_COUNTER, "/PGM/GC/DirtyPage/SetDirty", STAMUNIT_OCCURENCES, "The number of pages marked dirty because of write accesses.");
STAM_REG(pVM, &pPGM->StatGCDirtyTrackRealPF, STAMTYPE_COUNTER, "/PGM/GC/DirtyPage/RealPF", STAMUNIT_OCCURENCES, "The number of real pages faults during dirty bit tracking.");
STAM_REG(pVM, &pPGM->StatGCPageAlreadyDirty, STAMTYPE_COUNTER, "/PGM/GC/DirtyPage/AlreadySet", STAMUNIT_OCCURENCES, "The number of pages already marked dirty because of write accesses.");
STAM_REG(pVM, &pPGM->StatGCDirtyBitTracking, STAMTYPE_PROFILE, "/PGM/GC/DirtyPage", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMTrackDirtyBit() body.");
STAM_REG(pVM, &pPGM->StatGCSyncPTAlloc, STAMTYPE_COUNTER, "/PGM/GC/SyncPT/Alloc", STAMUNIT_OCCURENCES, "The number of times PGMGCSyncPT() needed to allocate page tables.");
STAM_REG(pVM, &pPGM->StatGCSyncPTConflict, STAMTYPE_COUNTER, "/PGM/GC/SyncPT/Conflicts", STAMUNIT_OCCURENCES, "The number of times PGMGCSyncPT() detected conflicts.");
STAM_REG(pVM, &pPGM->StatGCSyncPTFailed, STAMTYPE_COUNTER, "/PGM/GC/SyncPT/Failed", STAMUNIT_OCCURENCES, "The number of times PGMGCSyncPT() failed.");
STAM_REG(pVM, &pPGM->StatGCTrap0e, STAMTYPE_PROFILE, "/PGM/GC/Trap0e", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMGCTrap0eHandler() body.");
STAM_REG(pVM, &pPGM->StatCheckPageFault, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time/CheckPageFault", STAMUNIT_TICKS_PER_CALL, "Profiling of checking for dirty/access emulation faults.");
STAM_REG(pVM, &pPGM->StatLazySyncPT, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time/SyncPT", STAMUNIT_TICKS_PER_CALL, "Profiling of lazy page table syncing.");
STAM_REG(pVM, &pPGM->StatMapping, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time/Mapping", STAMUNIT_TICKS_PER_CALL, "Profiling of checking virtual mappings.");
STAM_REG(pVM, &pPGM->StatOutOfSync, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time/OutOfSync", STAMUNIT_TICKS_PER_CALL, "Profiling of out of sync page handling.");
STAM_REG(pVM, &pPGM->StatHandlers, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time/Handlers", STAMUNIT_TICKS_PER_CALL, "Profiling of checking handlers.");
STAM_REG(pVM, &pPGM->StatEIPHandlers, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time/EIPHandlers", STAMUNIT_TICKS_PER_CALL, "Profiling of checking eip handlers.");
STAM_REG(pVM, &pPGM->StatTrap0eCSAM, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/CSAM", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is CSAM.");
STAM_REG(pVM, &pPGM->StatTrap0eDirtyAndAccessedBits, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/DirtyAndAccessedBits", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is dirty and/or accessed bit emulation.");
STAM_REG(pVM, &pPGM->StatTrap0eGuestTrap, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/GuestTrap", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is a guest trap.");
STAM_REG(pVM, &pPGM->StatTrap0eHndPhys, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/HandlerPhysical", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is a physical handler.");
STAM_REG(pVM, &pPGM->StatTrap0eHndVirt, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/HandlerVirtual",STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is a virtual handler.");
STAM_REG(pVM, &pPGM->StatTrap0eHndUnhandled, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/HandlerUnhandled", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is access outside the monitored areas of a monitored page.");
STAM_REG(pVM, &pPGM->StatTrap0eMisc, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/Misc", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is not known.");
STAM_REG(pVM, &pPGM->StatTrap0eOutOfSync, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/OutOfSync", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is an out-of-sync page.");
STAM_REG(pVM, &pPGM->StatTrap0eOutOfSyncHndPhys, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/OutOfSyncHndPhys", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is an out-of-sync physical handler page.");
STAM_REG(pVM, &pPGM->StatTrap0eOutOfSyncHndVirt, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/OutOfSyncHndVirt", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is an out-of-sync virtual handler page.");
STAM_REG(pVM, &pPGM->StatTrap0eOutOfSyncObsHnd, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/OutOfSyncObsHnd", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is an obsolete handler page.");
STAM_REG(pVM, &pPGM->StatTrap0eSyncPT, STAMTYPE_PROFILE, "/PGM/GC/Trap0e/Time2/SyncPT", STAMUNIT_TICKS_PER_CALL, "Profiling of the Trap0eHandler body when the cause is lazy syncing of a PT.");
STAM_REG(pVM, &pPGM->StatTrap0eMapHandler, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Handlers/Mapping", STAMUNIT_OCCURENCES, "Number of traps due to access handlers in mappings.");
STAM_REG(pVM, &pPGM->StatHandlersOutOfSync, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Handlers/OutOfSync", STAMUNIT_OCCURENCES, "Number of traps due to out-of-sync handled pages.");
STAM_REG(pVM, &pPGM->StatHandlersPhysical, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Handlers/Physical", STAMUNIT_OCCURENCES, "Number of traps due to physical access handlers.");
STAM_REG(pVM, &pPGM->StatHandlersVirtual, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Handlers/Virtual", STAMUNIT_OCCURENCES, "Number of traps due to virtual access handlers.");
STAM_REG(pVM, &pPGM->StatHandlersVirtualByPhys, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Handlers/VirtualByPhys", STAMUNIT_OCCURENCES, "Number of traps due to virtual access handlers by physical address.");
STAM_REG(pVM, &pPGM->StatHandlersVirtualUnmarked, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Handlers/VirtualUnmarked", STAMUNIT_OCCURENCES,"Number of traps due to virtual access handlers by virtual address (without proper physical flags).");
STAM_REG(pVM, &pPGM->StatHandlersUnhandled, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Handlers/Unhandled", STAMUNIT_OCCURENCES, "Number of traps due to access outside range of monitored page(s).");
STAM_REG(pVM, &pPGM->StatGCTrap0eConflicts, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Conflicts", STAMUNIT_OCCURENCES, "The number of times #PF was caused by an undetected conflict.");
STAM_REG(pVM, &pPGM->StatGCTrap0eUSNotPresentRead, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/User/NPRead", STAMUNIT_OCCURENCES, "Number of user mode not present read page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eUSNotPresentWrite, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/User/NPWrite", STAMUNIT_OCCURENCES, "Number of user mode not present write page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eUSWrite, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/User/Write", STAMUNIT_OCCURENCES, "Number of user mode write page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eUSReserved, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/User/Reserved", STAMUNIT_OCCURENCES, "Number of user mode reserved bit page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eUSRead, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/User/Read", STAMUNIT_OCCURENCES, "Number of user mode read page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eSVNotPresentRead, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Supervisor/NPRead", STAMUNIT_OCCURENCES, "Number of supervisor mode not present read page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eSVNotPresentWrite, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Supervisor/NPWrite", STAMUNIT_OCCURENCES, "Number of supervisor mode not present write page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eSVWrite, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Supervisor/Write", STAMUNIT_OCCURENCES, "Number of supervisor mode write page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eSVReserved, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/Supervisor/Reserved", STAMUNIT_OCCURENCES, "Number of supervisor mode reserved bit page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eUnhandled, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/GuestPF/Unhandled", STAMUNIT_OCCURENCES, "Number of guest real page faults.");
STAM_REG(pVM, &pPGM->StatGCTrap0eMap, STAMTYPE_COUNTER, "/PGM/GC/Trap0e/GuestPF/Map", STAMUNIT_OCCURENCES, "Number of guest page faults due to map accesses.");
STAM_REG(pVM, &pPGM->StatGCGuestCR3WriteHandled, STAMTYPE_COUNTER, "/PGM/GC/CR3WriteInt", STAMUNIT_OCCURENCES, "The number of times the Guest CR3 change was successfully handled.");
STAM_REG(pVM, &pPGM->StatGCGuestCR3WriteUnhandled, STAMTYPE_COUNTER, "/PGM/GC/CR3WriteEmu", STAMUNIT_OCCURENCES, "The number of times the Guest CR3 change was passed back to the recompiler.");
STAM_REG(pVM, &pPGM->StatGCGuestCR3WriteConflict, STAMTYPE_COUNTER, "/PGM/GC/CR3WriteConflict", STAMUNIT_OCCURENCES, "The number of times the Guest CR3 monitoring detected a conflict.");
STAM_REG(pVM, &pPGM->StatGCPageOutOfSyncSupervisor, STAMTYPE_COUNTER, "/PGM/GC/OutOfSync/SuperVisor", STAMUNIT_OCCURENCES, "Number of traps due to pages out of sync.");
STAM_REG(pVM, &pPGM->StatGCPageOutOfSyncUser, STAMTYPE_COUNTER, "/PGM/GC/OutOfSync/User", STAMUNIT_OCCURENCES, "Number of traps due to pages out of sync.");
STAM_REG(pVM, &pPGM->StatGCGuestROMWriteHandled, STAMTYPE_COUNTER, "/PGM/GC/ROMWriteInt", STAMUNIT_OCCURENCES, "The number of times the Guest ROM change was successfully handled.");
STAM_REG(pVM, &pPGM->StatGCGuestROMWriteUnhandled, STAMTYPE_COUNTER, "/PGM/GC/ROMWriteEmu", STAMUNIT_OCCURENCES, "The number of times the Guest ROM change was passed back to the recompiler.");
STAM_REG(pVM, &pPGM->StatDynMapCacheHits, STAMTYPE_COUNTER, "/PGM/GC/DynMapCache/Hits" , STAMUNIT_OCCURENCES, "Number of dynamic page mapping cache hits.");
STAM_REG(pVM, &pPGM->StatDynMapCacheMisses, STAMTYPE_COUNTER, "/PGM/GC/DynMapCache/Misses" , STAMUNIT_OCCURENCES, "Number of dynamic page mapping cache misses.");
STAM_REG(pVM, &pPGM->StatHCDetectedConflicts, STAMTYPE_COUNTER, "/PGM/HC/DetectedConflicts", STAMUNIT_OCCURENCES, "The number of times PGMR3CheckMappingConflicts() detected a conflict.");
STAM_REG(pVM, &pPGM->StatHCGuestPDWrite, STAMTYPE_COUNTER, "/PGM/HC/PDWrite", STAMUNIT_OCCURENCES, "The total number of times pgmHCGuestPDWriteHandler() was called.");
STAM_REG(pVM, &pPGM->StatHCGuestPDWriteConflict, STAMTYPE_COUNTER, "/PGM/HC/PDWriteConflict", STAMUNIT_OCCURENCES, "The number of times pgmHCGuestPDWriteHandler() detected a conflict.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePage, STAMTYPE_PROFILE, "/PGM/HC/InvalidatePage", STAMUNIT_TICKS_PER_CALL, "PGMHCInvalidatePage() profiling.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePage4KBPages, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/4KBPages", STAMUNIT_OCCURENCES, "The number of times PGMHCInvalidatePage() was called for a 4KB page.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePage4MBPages, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/4MBPages", STAMUNIT_OCCURENCES, "The number of times PGMHCInvalidatePage() was called for a 4MB page.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePage4MBPagesSkip, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/4MBPagesSkip",STAMUNIT_OCCURENCES, "The number of times PGMHCInvalidatePage() skipped a 4MB page.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePagePDMappings, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/PDMappings", STAMUNIT_OCCURENCES, "The number of times PGMHCInvalidatePage() was called for a page directory containing mappings (no conflict).");
STAM_REG(pVM, &pPGM->StatHCInvalidatePagePDNAs, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/PDNAs", STAMUNIT_OCCURENCES, "The number of times PGMHCInvalidatePage() was called for a not accessed page directory.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePagePDNPs, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/PDNPs", STAMUNIT_OCCURENCES, "The number of times PGMHCInvalidatePage() was called for a not present page directory.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePagePDOutOfSync, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/PDOutOfSync", STAMUNIT_OCCURENCES, "The number of times PGMGCInvalidatePage() was called for an out of sync page directory.");
STAM_REG(pVM, &pPGM->StatHCInvalidatePageSkipped, STAMTYPE_COUNTER, "/PGM/HC/InvalidatePage/Skipped", STAMUNIT_OCCURENCES, "The number of times PGMHCInvalidatePage() was skipped due to not present shw or pending pending SyncCR3.");
STAM_REG(pVM, &pPGM->StatHCResolveConflict, STAMTYPE_PROFILE, "/PGM/HC/ResolveConflict", STAMUNIT_TICKS_PER_CALL, "pgmR3SyncPTResolveConflict() profiling (includes the entire relocation).");
STAM_REG(pVM, &pPGM->StatHCPrefetch, STAMTYPE_PROFILE, "/PGM/HC/Prefetch", STAMUNIT_TICKS_PER_CALL, "PGMR3PrefetchPage profiling.");
STAM_REG(pVM, &pPGM->StatHCSyncPT, STAMTYPE_PROFILE, "/PGM/HC/SyncPT", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMR3SyncPT() body.");
STAM_REG(pVM, &pPGM->StatHCAccessedPage, STAMTYPE_COUNTER, "/PGM/HC/AccessedPage", STAMUNIT_OCCURENCES, "The number of pages marked not present for accessed bit emulation.");
STAM_REG(pVM, &pPGM->StatHCDirtyPage, STAMTYPE_COUNTER, "/PGM/HC/DirtyPage/Mark", STAMUNIT_OCCURENCES, "The number of pages marked read-only for dirty bit tracking.");
STAM_REG(pVM, &pPGM->StatHCDirtyPageBig, STAMTYPE_COUNTER, "/PGM/HC/DirtyPage/MarkBig", STAMUNIT_OCCURENCES, "The number of 4MB pages marked read-only for dirty bit tracking.");
STAM_REG(pVM, &pPGM->StatHCDirtyPageTrap, STAMTYPE_COUNTER, "/PGM/HC/DirtyPage/Trap", STAMUNIT_OCCURENCES, "The number of traps generated for dirty bit tracking.");
STAM_REG(pVM, &pPGM->StatHCDirtyPageSkipped, STAMTYPE_COUNTER, "/PGM/HC/DirtyPage/Skipped", STAMUNIT_OCCURENCES, "The number of pages already dirty or readonly.");
STAM_REG(pVM, &pPGM->StatHCDirtyBitTracking, STAMTYPE_PROFILE, "/PGM/HC/DirtyPage", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMTrackDirtyBit() body.");
STAM_REG(pVM, &pPGM->StatGCSyncPagePDNAs, STAMTYPE_COUNTER, "/PGM/GC/SyncPagePDNAs", STAMUNIT_OCCURENCES, "The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit.");
STAM_REG(pVM, &pPGM->StatGCSyncPagePDOutOfSync, STAMTYPE_COUNTER, "/PGM/GC/SyncPagePDOutOfSync", STAMUNIT_OCCURENCES, "The number of time we've encountered an out-of-sync PD in SyncPage.");
STAM_REG(pVM, &pPGM->StatHCSyncPagePDNAs, STAMTYPE_COUNTER, "/PGM/HC/SyncPagePDNAs", STAMUNIT_OCCURENCES, "The number of time we've marked a PD not present from SyncPage to virtualize the accessed bit.");
STAM_REG(pVM, &pPGM->StatHCSyncPagePDOutOfSync, STAMTYPE_COUNTER, "/PGM/HC/SyncPagePDOutOfSync", STAMUNIT_OCCURENCES, "The number of time we've encountered an out-of-sync PD in SyncPage.");
STAM_REG(pVM, &pPGM->StatFlushTLB, STAMTYPE_PROFILE, "/PGM/FlushTLB", STAMUNIT_OCCURENCES, "Profiling of the PGMFlushTLB() body.");
STAM_REG(pVM, &pPGM->StatFlushTLBNewCR3, STAMTYPE_COUNTER, "/PGM/FlushTLB/NewCR3", STAMUNIT_OCCURENCES, "The number of times PGMFlushTLB was called with a new CR3, non-global. (switch)");
STAM_REG(pVM, &pPGM->StatFlushTLBNewCR3Global, STAMTYPE_COUNTER, "/PGM/FlushTLB/NewCR3Global", STAMUNIT_OCCURENCES, "The number of times PGMFlushTLB was called with a new CR3, global. (switch)");
STAM_REG(pVM, &pPGM->StatFlushTLBSameCR3, STAMTYPE_COUNTER, "/PGM/FlushTLB/SameCR3", STAMUNIT_OCCURENCES, "The number of times PGMFlushTLB was called with the same CR3, non-global. (flush)");
STAM_REG(pVM, &pPGM->StatFlushTLBSameCR3Global, STAMTYPE_COUNTER, "/PGM/FlushTLB/SameCR3Global", STAMUNIT_OCCURENCES, "The number of times PGMFlushTLB was called with the same CR3, global. (flush)");
STAM_REG(pVM, &pPGM->StatGCSyncCR3, STAMTYPE_PROFILE, "/PGM/GC/SyncCR3", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMSyncCR3() body.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3Handlers, STAMTYPE_PROFILE, "/PGM/GC/SyncCR3/Handlers", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMSyncCR3() update handler section.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3HandlerVirtualUpdate, STAMTYPE_PROFILE, "/PGM/GC/SyncCR3/Handlers/VirtualUpdate",STAMUNIT_TICKS_PER_CALL, "Profiling of the virtual handler updates.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3HandlerVirtualReset, STAMTYPE_PROFILE, "/PGM/GC/SyncCR3/Handlers/VirtualReset", STAMUNIT_TICKS_PER_CALL, "Profiling of the virtual handler resets.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3Global, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/Global", STAMUNIT_OCCURENCES, "The number of global CR3 syncs.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3NotGlobal, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/NotGlobal", STAMUNIT_OCCURENCES, "The number of non-global CR3 syncs.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3DstCacheHit, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/DstChacheHit", STAMUNIT_OCCURENCES, "The number of times we got some kind of a cache hit.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3DstFreed, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/DstFreed", STAMUNIT_OCCURENCES, "The number of times we've had to free a shadow entry.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3DstFreedSrcNP, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/DstFreedSrcNP", STAMUNIT_OCCURENCES, "The number of times we've had to free a shadow entry for which the source entry was not present.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3DstNotPresent, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/DstNotPresent", STAMUNIT_OCCURENCES, "The number of times we've encountered a not present shadow entry for a present guest entry.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3DstSkippedGlobalPD, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/DstSkippedGlobalPD", STAMUNIT_OCCURENCES, "The number of times a global page directory wasn't flushed.");
STAM_REG(pVM, &pPGM->StatGCSyncCR3DstSkippedGlobalPT, STAMTYPE_COUNTER, "/PGM/GC/SyncCR3/DstSkippedGlobalPT", STAMUNIT_OCCURENCES, "The number of times a page table with only global entries wasn't flushed.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3, STAMTYPE_PROFILE, "/PGM/HC/SyncCR3", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMSyncCR3() body.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3Handlers, STAMTYPE_PROFILE, "/PGM/HC/SyncCR3/Handlers", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMSyncCR3() update handler section.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3HandlerVirtualUpdate, STAMTYPE_PROFILE, "/PGM/HC/SyncCR3/Handlers/VirtualUpdate",STAMUNIT_TICKS_PER_CALL, "Profiling of the virtual handler updates.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3HandlerVirtualReset, STAMTYPE_PROFILE, "/PGM/HC/SyncCR3/Handlers/VirtualReset", STAMUNIT_TICKS_PER_CALL, "Profiling of the virtual handler resets.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3Global, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/Global", STAMUNIT_OCCURENCES, "The number of global CR3 syncs.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3NotGlobal, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/NotGlobal", STAMUNIT_OCCURENCES, "The number of non-global CR3 syncs.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3DstCacheHit, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/DstChacheHit", STAMUNIT_OCCURENCES, "The number of times we got some kind of a cache hit.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3DstFreed, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/DstFreed", STAMUNIT_OCCURENCES, "The number of times we've had to free a shadow entry.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3DstFreedSrcNP, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/DstFreedSrcNP", STAMUNIT_OCCURENCES, "The number of times we've had to free a shadow entry for which the source entry was not present.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3DstNotPresent, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/DstNotPresent", STAMUNIT_OCCURENCES, "The number of times we've encountered a not present shadow entry for a present guest entry.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3DstSkippedGlobalPD, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/DstSkippedGlobalPD", STAMUNIT_OCCURENCES, "The number of times a global page directory wasn't flushed.");
STAM_REG(pVM, &pPGM->StatHCSyncCR3DstSkippedGlobalPT, STAMTYPE_COUNTER, "/PGM/HC/SyncCR3/DstSkippedGlobalPT", STAMUNIT_OCCURENCES, "The number of times a page table with only global entries wasn't flushed.");
STAM_REG(pVM, &pPGM->StatVirtHandleSearchByPhysGC, STAMTYPE_PROFILE, "/PGM/VirtHandler/SearchByPhys/GC", STAMUNIT_TICKS_PER_CALL, "Profiling of pgmHandlerVirtualFindByPhysAddr in GC.");
STAM_REG(pVM, &pPGM->StatVirtHandleSearchByPhysHC, STAMTYPE_PROFILE, "/PGM/VirtHandler/SearchByPhys/HC", STAMUNIT_TICKS_PER_CALL, "Profiling of pgmHandlerVirtualFindByPhysAddr in HC.");
STAM_REG(pVM, &pPGM->StatHandlePhysicalReset, STAMTYPE_COUNTER, "/PGM/HC/HandlerPhysicalReset", STAMUNIT_OCCURENCES, "The number of times PGMR3HandlerPhysicalReset is called.");
STAM_REG(pVM, &pPGM->StatHCGstModifyPage, STAMTYPE_PROFILE, "/PGM/HC/GstModifyPage", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMGstModifyPage() body.");
STAM_REG(pVM, &pPGM->StatGCGstModifyPage, STAMTYPE_PROFILE, "/PGM/GC/GstModifyPage", STAMUNIT_TICKS_PER_CALL, "Profiling of the PGMGstModifyPage() body.");
STAM_REG(pVM, &pPGM->StatSynPT4kGC, STAMTYPE_COUNTER, "/PGM/GC/SyncPT/4k", STAMUNIT_OCCURENCES, "Nr of 4k PT syncs");
STAM_REG(pVM, &pPGM->StatSynPT4kHC, STAMTYPE_COUNTER, "/PGM/HC/SyncPT/4k", STAMUNIT_OCCURENCES, "Nr of 4k PT syncs");
STAM_REG(pVM, &pPGM->StatSynPT4MGC, STAMTYPE_COUNTER, "/PGM/GC/SyncPT/4M", STAMUNIT_OCCURENCES, "Nr of 4M PT syncs");
STAM_REG(pVM, &pPGM->StatSynPT4MHC, STAMTYPE_COUNTER, "/PGM/HC/SyncPT/4M", STAMUNIT_OCCURENCES, "Nr of 4M PT syncs");
STAM_REG(pVM, &pPGM->StatDynRamTotal, STAMTYPE_COUNTER, "/PGM/RAM/TotalAlloc", STAMUNIT_MEGABYTES, "Allocated mbs of guest ram.");
STAM_REG(pVM, &pPGM->StatDynRamGrow, STAMTYPE_COUNTER, "/PGM/RAM/Grow", STAMUNIT_OCCURENCES, "Nr of pgmr3PhysGrowRange calls.");
STAM_REG(pVM, &pPGM->StatPageHCMapTlbHits, STAMTYPE_COUNTER, "/PGM/PageHCMap/TlbHits", STAMUNIT_OCCURENCES, "TLB hits.");
STAM_REG(pVM, &pPGM->StatPageHCMapTlbMisses, STAMTYPE_COUNTER, "/PGM/PageHCMap/TlbMisses", STAMUNIT_OCCURENCES, "TLB misses.");
STAM_REG(pVM, &pPGM->ChunkR3Map.c, STAMTYPE_U32, "/PGM/ChunkR3Map/c", STAMUNIT_OCCURENCES, "Number of mapped chunks.");
STAM_REG(pVM, &pPGM->ChunkR3Map.cMax, STAMTYPE_U32, "/PGM/ChunkR3Map/cMax", STAMUNIT_OCCURENCES, "Maximum number of mapped chunks.");
STAM_REG(pVM, &pPGM->StatChunkR3MapTlbHits, STAMTYPE_COUNTER, "/PGM/ChunkR3Map/TlbHits", STAMUNIT_OCCURENCES, "TLB hits.");
STAM_REG(pVM, &pPGM->StatChunkR3MapTlbMisses, STAMTYPE_COUNTER, "/PGM/ChunkR3Map/TlbMisses", STAMUNIT_OCCURENCES, "TLB misses.");
STAM_REG(pVM, &pPGM->StatPageReplaceShared, STAMTYPE_COUNTER, "/PGM/Page/ReplacedShared", STAMUNIT_OCCURENCES, "Times a shared page was replaced.");
STAM_REG(pVM, &pPGM->StatPageReplaceZero, STAMTYPE_COUNTER, "/PGM/Page/ReplacedZero", STAMUNIT_OCCURENCES, "Times the zero page was replaced.");
STAM_REG(pVM, &pPGM->StatPageHandyAllocs, STAMTYPE_COUNTER, "/PGM/Page/HandyAllocs", STAMUNIT_OCCURENCES, "Number of times we've allocated more handy pages.");
STAM_REG(pVM, &pPGM->cAllPages, STAMTYPE_U32, "/PGM/Page/cAllPages", STAMUNIT_OCCURENCES, "The total number of pages.");
STAM_REG(pVM, &pPGM->cPrivatePages, STAMTYPE_U32, "/PGM/Page/cPrivatePages", STAMUNIT_OCCURENCES, "The number of private pages.");
STAM_REG(pVM, &pPGM->cSharedPages, STAMTYPE_U32, "/PGM/Page/cSharedPages", STAMUNIT_OCCURENCES, "The number of shared pages.");
STAM_REG(pVM, &pPGM->cZeroPages, STAMTYPE_U32, "/PGM/Page/cZeroPages", STAMUNIT_OCCURENCES, "The number of zero backed pages.");
#ifdef PGMPOOL_WITH_GCPHYS_TRACKING
STAM_REG(pVM, &pPGM->StatTrackVirgin, STAMTYPE_COUNTER, "/PGM/Track/Virgin", STAMUNIT_OCCURENCES, "The number of first time shadowings");
STAM_REG(pVM, &pPGM->StatTrackAliased, STAMTYPE_COUNTER, "/PGM/Track/Aliased", STAMUNIT_OCCURENCES, "The number of times switching to cRef2, i.e. the page is being shadowed by two PTs.");
STAM_REG(pVM, &pPGM->StatTrackAliasedMany, STAMTYPE_COUNTER, "/PGM/Track/AliasedMany", STAMUNIT_OCCURENCES, "The number of times we're tracking using cRef2.");
STAM_REG(pVM, &pPGM->StatTrackAliasedLots, STAMTYPE_COUNTER, "/PGM/Track/AliasedLots", STAMUNIT_OCCURENCES, "The number of times we're hitting pages which has overflowed cRef2");
STAM_REG(pVM, &pPGM->StatTrackOverflows, STAMTYPE_COUNTER, "/PGM/Track/Overflows", STAMUNIT_OCCURENCES, "The number of times the extent list grows to long.");
STAM_REG(pVM, &pPGM->StatTrackDeref, STAMTYPE_PROFILE, "/PGM/Track/Deref", STAMUNIT_OCCURENCES, "Profiling of SyncPageWorkerTrackDeref (expensive).");
#endif
for (unsigned i = 0; i < PAGE_ENTRIES; i++)
{
/** @todo r=bird: We need a STAMR3RegisterF()! */
char szName[32];
RTStrPrintf(szName, sizeof(szName), "/PGM/GC/PD/Trap0e/%04X", i);
int rc = STAMR3Register(pVM, &pPGM->StatGCTrap0ePD[i], STAMTYPE_COUNTER, STAMVISIBILITY_USED, szName, STAMUNIT_OCCURENCES, "The number of traps in page directory n.");
AssertRC(rc);
RTStrPrintf(szName, sizeof(szName), "/PGM/GC/PD/SyncPt/%04X", i);
rc = STAMR3Register(pVM, &pPGM->StatGCSyncPtPD[i], STAMTYPE_COUNTER, STAMVISIBILITY_USED, szName, STAMUNIT_OCCURENCES, "The number of syncs per PD n.");
AssertRC(rc);
RTStrPrintf(szName, sizeof(szName), "/PGM/GC/PD/SyncPage/%04X", i);
rc = STAMR3Register(pVM, &pPGM->StatGCSyncPagePD[i], STAMTYPE_COUNTER, STAMVISIBILITY_USED, szName, STAMUNIT_OCCURENCES, "The number of out of sync pages per page directory n.");
AssertRC(rc);
}
}
#endif /* VBOX_WITH_STATISTICS */
/**
* Init the PGM bits that rely on VMMR0 and MM to be fully initialized.
*
* The dynamic mapping area will also be allocated and initialized at this
* time. We could allocate it during PGMR3Init of course, but the mapping
* wouldn't be allocated at that time preventing us from setting up the
* page table entries with the dummy page.
*
* @returns VBox status code.
* @param pVM VM handle.
*/
PGMR3DECL(int) PGMR3InitDynMap(PVM pVM)
{
/*
* Reserve space for mapping the paging pages into guest context.
*/
int rc = MMR3HyperReserve(pVM, PAGE_SIZE * (2 + ELEMENTS(pVM->pgm.s.apHCPaePDs) + 1 + 2 + 2), "Paging", &pVM->pgm.s.pGC32BitPD);
AssertRCReturn(rc, rc);
MMR3HyperReserve(pVM, PAGE_SIZE, "fence", NULL);
/*
* Reserve space for the dynamic mappings.
*/
/** @todo r=bird: Need to verify that the checks for crossing PTs are correct here. They seems to be assuming 4MB PTs.. */
rc = MMR3HyperReserve(pVM, MM_HYPER_DYNAMIC_SIZE, "Dynamic mapping", &pVM->pgm.s.pbDynPageMapBaseGC);
if ( VBOX_SUCCESS(rc)
&& (pVM->pgm.s.pbDynPageMapBaseGC >> PGDIR_SHIFT) != ((pVM->pgm.s.pbDynPageMapBaseGC + MM_HYPER_DYNAMIC_SIZE - 1) >> PGDIR_SHIFT))
rc = MMR3HyperReserve(pVM, MM_HYPER_DYNAMIC_SIZE, "Dynamic mapping not crossing", &pVM->pgm.s.pbDynPageMapBaseGC);
if (VBOX_SUCCESS(rc))
{
AssertRelease((pVM->pgm.s.pbDynPageMapBaseGC >> PGDIR_SHIFT) == ((pVM->pgm.s.pbDynPageMapBaseGC + MM_HYPER_DYNAMIC_SIZE - 1) >> PGDIR_SHIFT));
MMR3HyperReserve(pVM, PAGE_SIZE, "fence", NULL);
}
return rc;
}
/**
* Ring-3 init finalizing.
*
* @returns VBox status code.
* @param pVM The VM handle.
*/
PGMR3DECL(int) PGMR3InitFinalize(PVM pVM)
{
/*
* Map the paging pages into the guest context.
*/
RTGCPTR GCPtr = pVM->pgm.s.pGC32BitPD;
AssertReleaseReturn(GCPtr, VERR_INTERNAL_ERROR);
int rc = PGMMap(pVM, GCPtr, pVM->pgm.s.HCPhys32BitPD, PAGE_SIZE, 0);
AssertRCReturn(rc, rc);
pVM->pgm.s.pGC32BitPD = GCPtr;
GCPtr += PAGE_SIZE;
GCPtr += PAGE_SIZE; /* reserved page */
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.apHCPaePDs); i++)
{
rc = PGMMap(pVM, GCPtr, pVM->pgm.s.aHCPhysPaePDs[i], PAGE_SIZE, 0);
AssertRCReturn(rc, rc);
pVM->pgm.s.apGCPaePDs[i] = GCPtr;
GCPtr += PAGE_SIZE;
}
/* A bit of paranoia is justified. */
AssertRelease((RTGCUINTPTR)pVM->pgm.s.apGCPaePDs[0] + PAGE_SIZE == (RTGCUINTPTR)pVM->pgm.s.apGCPaePDs[1]);
AssertRelease((RTGCUINTPTR)pVM->pgm.s.apGCPaePDs[1] + PAGE_SIZE == (RTGCUINTPTR)pVM->pgm.s.apGCPaePDs[2]);
AssertRelease((RTGCUINTPTR)pVM->pgm.s.apGCPaePDs[2] + PAGE_SIZE == (RTGCUINTPTR)pVM->pgm.s.apGCPaePDs[3]);
GCPtr += PAGE_SIZE; /* reserved page */
rc = PGMMap(pVM, GCPtr, pVM->pgm.s.HCPhysPaePDPTR, PAGE_SIZE, 0);
AssertRCReturn(rc, rc);
pVM->pgm.s.pGCPaePDPTR = GCPtr;
GCPtr += PAGE_SIZE;
GCPtr += PAGE_SIZE; /* reserved page */
rc = PGMMap(pVM, GCPtr, pVM->pgm.s.HCPhysPaePML4, PAGE_SIZE, 0);
AssertRCReturn(rc, rc);
pVM->pgm.s.pGCPaePML4 = GCPtr;
GCPtr += PAGE_SIZE;
GCPtr += PAGE_SIZE; /* reserved page */
/*
* Reserve space for the dynamic mappings.
* Initialize the dynamic mapping pages with dummy pages to simply the cache.
*/
/* get the pointer to the page table entries. */
PPGMMAPPING pMapping = pgmGetMapping(pVM, pVM->pgm.s.pbDynPageMapBaseGC);
AssertRelease(pMapping);
const uintptr_t off = pVM->pgm.s.pbDynPageMapBaseGC - pMapping->GCPtr;
const unsigned iPT = off >> X86_PD_SHIFT;
const unsigned iPG = (off >> X86_PT_SHIFT) & X86_PT_MASK;
pVM->pgm.s.paDynPageMap32BitPTEsGC = pMapping->aPTs[iPT].pPTGC + iPG * sizeof(pMapping->aPTs[0].pPTR3->a[0]);
pVM->pgm.s.paDynPageMapPaePTEsGC = pMapping->aPTs[iPT].paPaePTsGC + iPG * sizeof(pMapping->aPTs[0].paPaePTsR3->a[0]);
/* init cache */
RTHCPHYS HCPhysDummy = MMR3PageDummyHCPhys(pVM);
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.aHCPhysDynPageMapCache); i++)
pVM->pgm.s.aHCPhysDynPageMapCache[i] = HCPhysDummy;
for (unsigned i = 0; i < MM_HYPER_DYNAMIC_SIZE; i += PAGE_SIZE)
{
rc = PGMMap(pVM, pVM->pgm.s.pbDynPageMapBaseGC + i, HCPhysDummy, PAGE_SIZE, 0);
AssertRCReturn(rc, rc);
}
return rc;
}
/**
* Applies relocations to data and code managed by this
* component. This function will be called at init and
* whenever the VMM need to relocate it self inside the GC.
*
* @param pVM The VM.
* @param offDelta Relocation delta relative to old location.
*/
PGMR3DECL(void) PGMR3Relocate(PVM pVM, RTGCINTPTR offDelta)
{
LogFlow(("PGMR3Relocate\n"));
/*
* Paging stuff.
*/
pVM->pgm.s.GCPtrCR3Mapping += offDelta;
/** @todo move this into shadow and guest specific relocation functions. */
AssertMsg(pVM->pgm.s.pGC32BitPD, ("Init order, no relocation before paging is initialized!\n"));
pVM->pgm.s.pGC32BitPD += offDelta;
pVM->pgm.s.pGuestPDGC += offDelta;
for (unsigned i = 0; i < ELEMENTS(pVM->pgm.s.apGCPaePDs); i++)
pVM->pgm.s.apGCPaePDs[i] += offDelta;
pVM->pgm.s.pGCPaePDPTR += offDelta;
pVM->pgm.s.pGCPaePML4 += offDelta;
pgmR3ModeDataInit(pVM, true /* resolve GC/R0 symbols */);
pgmR3ModeDataSwitch(pVM, pVM->pgm.s.enmShadowMode, pVM->pgm.s.enmGuestMode);
PGM_SHW_PFN(Relocate, pVM)(pVM, offDelta);
PGM_GST_PFN(Relocate, pVM)(pVM, offDelta);
PGM_BTH_PFN(Relocate, pVM)(pVM, offDelta);
/*
* Trees.
*/
pVM->pgm.s.pTreesGC = MMHyperHC2GC(pVM, pVM->pgm.s.pTreesHC);
/*
* Ram ranges.
*/
if (pVM->pgm.s.pRamRangesHC)
{
pVM->pgm.s.pRamRangesGC = MMHyperHC2GC(pVM, pVM->pgm.s.pRamRangesHC);
for (PPGMRAMRANGE pCur = pVM->pgm.s.pRamRangesHC; pCur->pNextHC; pCur = pCur->pNextHC)
{
pCur->pNextGC = MMHyperHC2GC(pVM, pCur->pNextHC);
if (pCur->pavHCChunkGC)
pCur->pavHCChunkGC = MMHyperHC2GC(pVM, pCur->pavHCChunkHC);
}
}
/*
* Update the two page directories with all page table mappings.
* (One or more of them have changed, that's why we're here.)
*/
pVM->pgm.s.pMappingsGC = MMHyperHC2GC(pVM, pVM->pgm.s.pMappingsR3);
for (PPGMMAPPING pCur = pVM->pgm.s.pMappingsR3; pCur->pNextR3; pCur = pCur->pNextR3)
pCur->pNextGC = MMHyperHC2GC(pVM, pCur->pNextR3);
/* Relocate GC addresses of Page Tables. */
for (PPGMMAPPING pCur = pVM->pgm.s.pMappingsR3; pCur; pCur = pCur->pNextR3)
{
for (RTHCUINT i = 0; i < pCur->cPTs; i++)
{
pCur->aPTs[i].pPTGC = MMHyperR3ToGC(pVM, pCur->aPTs[i].pPTR3);
pCur->aPTs[i].paPaePTsGC = MMHyperR3ToGC(pVM, pCur->aPTs[i].paPaePTsR3);
}
}
/*
* Dynamic page mapping area.
*/
pVM->pgm.s.paDynPageMap32BitPTEsGC += offDelta;
pVM->pgm.s.paDynPageMapPaePTEsGC += offDelta;
pVM->pgm.s.pbDynPageMapBaseGC += offDelta;
/*
* The Zero page.
*/
pVM->pgm.s.pvZeroPgR0 = MMHyperR3ToR0(pVM, pVM->pgm.s.pvZeroPgR3);
AssertRelease(pVM->pgm.s.pvZeroPgR0);
/*
* Physical and virtual handlers.
*/
RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesHC->PhysHandlers, true, pgmR3RelocatePhysHandler, &offDelta);
RTAvlroGCPtrDoWithAll(&pVM->pgm.s.pTreesHC->VirtHandlers, true, pgmR3RelocateVirtHandler, &offDelta);
/*
* The page pool.
*/
pgmR3PoolRelocate(pVM);
}
/**
* Callback function for relocating a physical access handler.
*
* @returns 0 (continue enum)
* @param pNode Pointer to a PGMPHYSHANDLER node.
* @param pvUser Pointer to the offDelta. This is a pointer to the delta since we're
* not certain the delta will fit in a void pointer for all possible configs.
*/
static DECLCALLBACK(int) pgmR3RelocatePhysHandler(PAVLROGCPHYSNODECORE pNode, void *pvUser)
{
PPGMPHYSHANDLER pHandler = (PPGMPHYSHANDLER)pNode;
RTGCINTPTR offDelta = *(PRTGCINTPTR)pvUser;
if (pHandler->pfnHandlerGC)
pHandler->pfnHandlerGC += offDelta;
if ((RTGCUINTPTR)pHandler->pvUserGC >= 0x10000)
pHandler->pvUserGC += offDelta;
return 0;
}
/**
* Callback function for relocating a virtual access handler.
*
* @returns 0 (continue enum)
* @param pNode Pointer to a PGMVIRTHANDLER node.
* @param pvUser Pointer to the offDelta. This is a pointer to the delta since we're
* not certain the delta will fit in a void pointer for all possible configs.
*/
static DECLCALLBACK(int) pgmR3RelocateVirtHandler(PAVLROGCPTRNODECORE pNode, void *pvUser)
{
PPGMVIRTHANDLER pHandler = (PPGMVIRTHANDLER)pNode;
RTGCINTPTR offDelta = *(PRTGCINTPTR)pvUser;
Assert(pHandler->pfnHandlerGC);
pHandler->pfnHandlerGC += offDelta;
return 0;
}
/**
* The VM is being reset.
*
* For the PGM component this means that any PD write monitors
* needs to be removed.
*
* @param pVM VM handle.
*/
PGMR3DECL(void) PGMR3Reset(PVM pVM)
{
LogFlow(("PGMR3Reset:\n"));
VM_ASSERT_EMT(pVM);
/*
* Unfix any fixed mappings and disable CR3 monitoring.
*/
pVM->pgm.s.fMappingsFixed = false;
pVM->pgm.s.GCPtrMappingFixed = 0;
pVM->pgm.s.cbMappingFixed = 0;
int rc = PGM_GST_PFN(UnmonitorCR3, pVM)(pVM);
AssertRC(rc);
#ifdef DEBUG
PGMR3DumpMappings(pVM);
#endif
/*
* Reset the shadow page pool.
*/
pgmR3PoolReset(pVM);
/*
* Re-init other members.
*/
pVM->pgm.s.fA20Enabled = true;
/*
* Clear the FFs PGM owns.
*/
VM_FF_CLEAR(pVM, VM_FF_PGM_SYNC_CR3);
VM_FF_CLEAR(pVM, VM_FF_PGM_SYNC_CR3_NON_GLOBAL);
/*
* Zero memory.
*/
for (PPGMRAMRANGE pRam = pVM->pgm.s.pRamRangesHC; pRam; pRam = pRam->pNextHC)
{
unsigned iPage = pRam->cb >> PAGE_SHIFT;
while (iPage-- > 0)
{
if (pRam->aPages[iPage].HCPhys & (MM_RAM_FLAGS_RESERVED | MM_RAM_FLAGS_ROM | MM_RAM_FLAGS_MMIO | MM_RAM_FLAGS_MMIO2)) /** @todo PAGE FLAGS */
{
/* shadow ram is reloaded elsewhere. */
Log4(("PGMR3Reset: not clearing phys page %RGp due to flags %RHp\n", pRam->GCPhys + (iPage << PAGE_SHIFT), pRam->aPages[iPage].HCPhys & (MM_RAM_FLAGS_RESERVED | MM_RAM_FLAGS_ROM | MM_RAM_FLAGS_MMIO))); /** @todo PAGE FLAGS */
continue;
}
if (pRam->fFlags & MM_RAM_FLAGS_DYNAMIC_ALLOC)
{
unsigned iChunk = iPage >> (PGM_DYNAMIC_CHUNK_SHIFT - PAGE_SHIFT);
if (pRam->pavHCChunkHC[iChunk])
ASMMemZero32((char *)pRam->pavHCChunkHC[iChunk] + ((iPage << PAGE_SHIFT) & PGM_DYNAMIC_CHUNK_OFFSET_MASK), PAGE_SIZE);
}
else
ASMMemZero32((char *)pRam->pvHC + (iPage << PAGE_SHIFT), PAGE_SIZE);
}
}
/*
* Switch mode back to real mode.
*/
rc = pgmR3ChangeMode(pVM, PGMMODE_REAL);
AssertReleaseRC(rc);
STAM_REL_COUNTER_RESET(&pVM->pgm.s.cGuestModeChanges);
}
/**
* Terminates the PGM.
*
* @returns VBox status code.
* @param pVM Pointer to VM structure.
*/
PGMR3DECL(int) PGMR3Term(PVM pVM)
{
return PDMR3CritSectDelete(&pVM->pgm.s.CritSect);
}
#ifdef VBOX_STRICT
/**
* VM state change callback for clearing fNoMorePhysWrites after
* a snapshot has been created.
*/
static DECLCALLBACK(void) pgmR3ResetNoMorePhysWritesFlag(PVM pVM, VMSTATE enmState, VMSTATE enmOldState, void *pvUser)
{
if (enmState == VMSTATE_RUNNING)
pVM->pgm.s.fNoMorePhysWrites = false;
}
#endif
/**
* Execute state save operation.
*
* @returns VBox status code.
* @param pVM VM Handle.
* @param pSSM SSM operation handle.
*/
static DECLCALLBACK(int) pgmR3Save(PVM pVM, PSSMHANDLE pSSM)
{
PPGM pPGM = &pVM->pgm.s;
/* No more writes to physical memory after this point! */
pVM->pgm.s.fNoMorePhysWrites = true;
/*
* Save basic data (required / unaffected by relocation).
*/
#if 1
SSMR3PutBool(pSSM, pPGM->fMappingsFixed);
#else
SSMR3PutUInt(pSSM, pPGM->fMappingsFixed);
#endif
SSMR3PutGCPtr(pSSM, pPGM->GCPtrMappingFixed);
SSMR3PutU32(pSSM, pPGM->cbMappingFixed);
SSMR3PutUInt(pSSM, pPGM->cbRamSize);
SSMR3PutGCPhys(pSSM, pPGM->GCPhysA20Mask);
SSMR3PutUInt(pSSM, pPGM->fA20Enabled);
SSMR3PutUInt(pSSM, pPGM->fSyncFlags);
SSMR3PutUInt(pSSM, pPGM->enmGuestMode);
SSMR3PutU32(pSSM, ~0); /* Separator. */
/*
* The guest mappings.
*/
uint32_t i = 0;
for (PPGMMAPPING pMapping = pPGM->pMappingsR3; pMapping; pMapping = pMapping->pNextR3, i++)
{
SSMR3PutU32(pSSM, i);
SSMR3PutStrZ(pSSM, pMapping->pszDesc); /* This is the best unique id we have... */
SSMR3PutGCPtr(pSSM, pMapping->GCPtr);
SSMR3PutGCUIntPtr(pSSM, pMapping->cPTs);
/* flags are done by the mapping owners! */
}
SSMR3PutU32(pSSM, ~0); /* terminator. */
/*
* Ram range flags and bits.
*/
i = 0;
for (PPGMRAMRANGE pRam = pPGM->pRamRangesHC; pRam; pRam = pRam->pNextHC, i++)
{
/** @todo MMIO ranges may move (PCI reconfig), we currently assume they don't. */
SSMR3PutU32(pSSM, i);
SSMR3PutGCPhys(pSSM, pRam->GCPhys);
SSMR3PutGCPhys(pSSM, pRam->GCPhysLast);
SSMR3PutGCPhys(pSSM, pRam->cb);
SSMR3PutU8(pSSM, !!pRam->pvHC); /* boolean indicating memory or not. */
/* Flags. */
const unsigned cPages = pRam->cb >> PAGE_SHIFT;
for (unsigned iPage = 0; iPage < cPages; iPage++)
SSMR3PutU16(pSSM, (uint16_t)(pRam->aPages[iPage].HCPhys & ~X86_PTE_PAE_PG_MASK)); /** @todo PAGE FLAGS */
/* any memory associated with the range. */
if (pRam->fFlags & MM_RAM_FLAGS_DYNAMIC_ALLOC)
{
for (unsigned iChunk = 0; iChunk < (pRam->cb >> PGM_DYNAMIC_CHUNK_SHIFT); iChunk++)
{
if (pRam->pavHCChunkHC[iChunk])
{
SSMR3PutU8(pSSM, 1); /* chunk present */
SSMR3PutMem(pSSM, pRam->pavHCChunkHC[iChunk], PGM_DYNAMIC_CHUNK_SIZE);
}
else
SSMR3PutU8(pSSM, 0); /* no chunk present */
}
}
else if (pRam->pvHC)
{
int rc = SSMR3PutMem(pSSM, pRam->pvHC, pRam->cb);
if (VBOX_FAILURE(rc))
{
Log(("pgmR3Save: SSMR3PutMem(, %p, %#x) -> %Vrc\n", pRam->pvHC, pRam->cb, rc));
return rc;
}
}
}
return SSMR3PutU32(pSSM, ~0); /* terminator. */
}
/**
* Execute state load operation.
*
* @returns VBox status code.
* @param pVM VM Handle.
* @param pSSM SSM operation handle.
* @param u32Version Data layout version.
*/
static DECLCALLBACK(int) pgmR3Load(PVM pVM, PSSMHANDLE pSSM, uint32_t u32Version)
{
/*
* Validate version.
*/
if (u32Version != PGM_SAVED_STATE_VERSION)
{
Log(("pgmR3Load: Invalid version u32Version=%d (current %d)!\n", u32Version, PGM_SAVED_STATE_VERSION));
return VERR_SSM_UNSUPPORTED_DATA_UNIT_VERSION;
}
/*
* Call the reset function to make sure all the memory is cleared.
*/
PGMR3Reset(pVM);
/*
* Load basic data (required / unaffected by relocation).
*/
PPGM pPGM = &pVM->pgm.s;
#if 1
SSMR3GetBool(pSSM, &pPGM->fMappingsFixed);
#else
uint32_t u;
SSMR3GetU32(pSSM, &u);
pPGM->fMappingsFixed = u;
#endif
SSMR3GetGCPtr(pSSM, &pPGM->GCPtrMappingFixed);
SSMR3GetU32(pSSM, &pPGM->cbMappingFixed);
RTUINT cbRamSize;
int rc = SSMR3GetU32(pSSM, &cbRamSize);
if (VBOX_FAILURE(rc))
return rc;
if (cbRamSize != pPGM->cbRamSize)
return VERR_SSM_LOAD_MEMORY_SIZE_MISMATCH;
SSMR3GetGCPhys(pSSM, &pPGM->GCPhysA20Mask);
SSMR3GetUInt(pSSM, &pPGM->fA20Enabled);
SSMR3GetUInt(pSSM, &pPGM->fSyncFlags);
RTUINT uGuestMode;
SSMR3GetUInt(pSSM, &uGuestMode);
pPGM->enmGuestMode = (PGMMODE)uGuestMode;
/* check separator. */
uint32_t u32Sep;
SSMR3GetU32(pSSM, &u32Sep);
if (VBOX_FAILURE(rc))
return rc;
if (u32Sep != (uint32_t)~0)
{
AssertMsgFailed(("u32Sep=%#x (first)\n", u32Sep));
return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
}
/*
* The guest mappings.
*/
uint32_t i = 0;
for (;; i++)
{
/* Check the seqence number / separator. */
rc = SSMR3GetU32(pSSM, &u32Sep);
if (VBOX_FAILURE(rc))
return rc;
if (u32Sep == ~0U)
break;
if (u32Sep != i)
{
AssertMsgFailed(("u32Sep=%#x (last)\n", u32Sep));
return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
}
/* get the mapping details. */
char szDesc[256];
szDesc[0] = '\0';
rc = SSMR3GetStrZ(pSSM, szDesc, sizeof(szDesc));
if (VBOX_FAILURE(rc))
return rc;
RTGCPTR GCPtr;
SSMR3GetGCPtr(pSSM, &GCPtr);
RTGCUINTPTR cPTs;
rc = SSMR3GetU32(pSSM, &cPTs);
if (VBOX_FAILURE(rc))
return rc;
/* find matching range. */
PPGMMAPPING pMapping;
for (pMapping = pPGM->pMappingsR3; pMapping; pMapping = pMapping->pNextR3)
if ( pMapping->cPTs == cPTs
&& !strcmp(pMapping->pszDesc, szDesc))
break;
if (!pMapping)
{
LogRel(("Couldn't find mapping: cPTs=%#x szDesc=%s (GCPtr=%VGv)\n",
cPTs, szDesc, GCPtr));
AssertFailed();
return VERR_SSM_LOAD_CONFIG_MISMATCH;
}
/* relocate it. */
if (pMapping->GCPtr != GCPtr)
{
AssertMsg((GCPtr >> PGDIR_SHIFT << PGDIR_SHIFT) == GCPtr, ("GCPtr=%VGv\n", GCPtr));
#if HC_ARCH_BITS == 64
LogRel(("Mapping: %VGv -> %VGv %s\n", pMapping->GCPtr, GCPtr, pMapping->pszDesc));
#endif
pgmR3MapRelocate(pVM, pMapping, pMapping->GCPtr >> PGDIR_SHIFT, GCPtr >> PGDIR_SHIFT);
}
else
Log(("pgmR3Load: '%s' needed no relocation (%VGv)\n", szDesc, GCPtr));
}
/*
* Ram range flags and bits.
*/
i = 0;
for (PPGMRAMRANGE pRam = pPGM->pRamRangesHC; pRam; pRam = pRam->pNextHC, i++)
{
/** @todo MMIO ranges may move (PCI reconfig), we currently assume they don't. */
/* Check the seqence number / separator. */
rc = SSMR3GetU32(pSSM, &u32Sep);
if (VBOX_FAILURE(rc))
return rc;
if (u32Sep == ~0U)
break;
if (u32Sep != i)
{
AssertMsgFailed(("u32Sep=%#x (last)\n", u32Sep));
return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
}
/* Get the range details. */
RTGCPHYS GCPhys;
SSMR3GetGCPhys(pSSM, &GCPhys);
RTGCPHYS GCPhysLast;
SSMR3GetGCPhys(pSSM, &GCPhysLast);
RTGCPHYS cb;
SSMR3GetGCPhys(pSSM, &cb);
uint8_t fHaveBits;
rc = SSMR3GetU8(pSSM, &fHaveBits);
if (VBOX_FAILURE(rc))
return rc;
if (fHaveBits & ~1)
{
AssertMsgFailed(("u32Sep=%#x (last)\n", u32Sep));
return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
}
/* Match it up with the current range. */
if ( GCPhys != pRam->GCPhys
|| GCPhysLast != pRam->GCPhysLast
|| cb != pRam->cb
|| fHaveBits != !!pRam->pvHC)
{
LogRel(("Ram range: %VGp-%VGp %VGp bytes %s\n"
"State : %VGp-%VGp %VGp bytes %s\n",
pRam->GCPhys, pRam->GCPhysLast, pRam->cb, pRam->pvHC ? "bits" : "nobits",
GCPhys, GCPhysLast, cb, fHaveBits ? "bits" : "nobits"));
/*
* If we're loading a state for debugging purpose, don't make a fuss if
* the MMIO[2] and ROM stuff isn't 100% right, just skip the mismatches.
*/
if ( SSMR3HandleGetAfter(pSSM) != SSMAFTER_DEBUG_IT
|| GCPhys < 8 * _1M)
AssertFailedReturn(VERR_SSM_LOAD_CONFIG_MISMATCH);
RTGCPHYS cPages = ((GCPhysLast - GCPhys) + 1) >> PAGE_SHIFT;
while (cPages-- > 0)
{
uint16_t u16Ignore;
SSMR3GetU16(pSSM, &u16Ignore);
}
continue;
}
/* Flags. */
const unsigned cPages = pRam->cb >> PAGE_SHIFT;
for (unsigned iPage = 0; iPage < cPages; iPage++)
{
uint16_t u16 = 0;
SSMR3GetU16(pSSM, &u16);
u16 &= PAGE_OFFSET_MASK & ~( MM_RAM_FLAGS_VIRTUAL_HANDLER | MM_RAM_FLAGS_VIRTUAL_WRITE | MM_RAM_FLAGS_VIRTUAL_ALL
| MM_RAM_FLAGS_PHYSICAL_HANDLER | MM_RAM_FLAGS_PHYSICAL_WRITE | MM_RAM_FLAGS_PHYSICAL_ALL
| MM_RAM_FLAGS_PHYSICAL_TEMP_OFF );
pRam->aPages[iPage].HCPhys = PGM_PAGE_GET_HCPHYS(&pRam->aPages[iPage]) | (RTHCPHYS)u16; /** @todo PAGE FLAGS */
}
/* any memory associated with the range. */
if (pRam->fFlags & MM_RAM_FLAGS_DYNAMIC_ALLOC)
{
for (unsigned iChunk = 0; iChunk < (pRam->cb >> PGM_DYNAMIC_CHUNK_SHIFT); iChunk++)
{
uint8_t fValidChunk;
rc = SSMR3GetU8(pSSM, &fValidChunk);
if (VBOX_FAILURE(rc))
return rc;
if (fValidChunk > 1)
return VERR_SSM_DATA_UNIT_FORMAT_CHANGED;
if (fValidChunk)
{
if (!pRam->pavHCChunkHC[iChunk])
{
rc = pgmr3PhysGrowRange(pVM, pRam->GCPhys + iChunk * PGM_DYNAMIC_CHUNK_SIZE);
if (VBOX_FAILURE(rc))
return rc;
}
Assert(pRam->pavHCChunkHC[iChunk]);
SSMR3GetMem(pSSM, pRam->pavHCChunkHC[iChunk], PGM_DYNAMIC_CHUNK_SIZE);
}
/* else nothing to do */
}
}
else if (pRam->pvHC)
{
int rc = SSMR3GetMem(pSSM, pRam->pvHC, pRam->cb);
if (VBOX_FAILURE(rc))
{
Log(("pgmR3Save: SSMR3GetMem(, %p, %#x) -> %Vrc\n", pRam->pvHC, pRam->cb, rc));
return rc;
}
}
}
/*
* We require a full resync now.
*/
VM_FF_SET(pVM, VM_FF_PGM_SYNC_CR3_NON_GLOBAL);
VM_FF_SET(pVM, VM_FF_PGM_SYNC_CR3);
pPGM->fSyncFlags |= PGM_SYNC_UPDATE_PAGE_BIT_VIRTUAL;
pPGM->fPhysCacheFlushPending = true;
pgmR3HandlerPhysicalUpdateAll(pVM);
/*
* Change the paging mode.
*/
return pgmR3ChangeMode(pVM, pPGM->enmGuestMode);
}
/**
* Show paging mode.
*
* @param pVM VM Handle.
* @param pHlp The info helpers.
* @param pszArgs "all" (default), "guest", "shadow" or "host".
*/
static DECLCALLBACK(void) pgmR3InfoMode(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs)
{
/* digest argument. */
bool fGuest, fShadow, fHost;
if (pszArgs)
pszArgs = RTStrStripL(pszArgs);
if (!pszArgs || !*pszArgs || strstr(pszArgs, "all"))
fShadow = fHost = fGuest = true;
else
{
fShadow = fHost = fGuest = false;
if (strstr(pszArgs, "guest"))
fGuest = true;
if (strstr(pszArgs, "shadow"))
fShadow = true;
if (strstr(pszArgs, "host"))
fHost = true;
}
/* print info. */
if (fGuest)
pHlp->pfnPrintf(pHlp, "Guest paging mode: %s, changed %RU64 times, A20 %s\n",
PGMGetModeName(pVM->pgm.s.enmGuestMode), pVM->pgm.s.cGuestModeChanges.c,
pVM->pgm.s.fA20Enabled ? "enabled" : "disabled");
if (fShadow)
pHlp->pfnPrintf(pHlp, "Shadow paging mode: %s\n", PGMGetModeName(pVM->pgm.s.enmShadowMode));
if (fHost)
{
const char *psz;
switch (pVM->pgm.s.enmHostMode)
{
case SUPPAGINGMODE_INVALID: psz = "invalid"; break;
case SUPPAGINGMODE_32_BIT: psz = "32-bit"; break;
case SUPPAGINGMODE_32_BIT_GLOBAL: psz = "32-bit+G"; break;
case SUPPAGINGMODE_PAE: psz = "PAE"; break;
case SUPPAGINGMODE_PAE_GLOBAL: psz = "PAE+G"; break;
case SUPPAGINGMODE_PAE_NX: psz = "PAE+NX"; break;
case SUPPAGINGMODE_PAE_GLOBAL_NX: psz = "PAE+G+NX"; break;
case SUPPAGINGMODE_AMD64: psz = "AMD64"; break;
case SUPPAGINGMODE_AMD64_GLOBAL: psz = "AMD64+G"; break;
case SUPPAGINGMODE_AMD64_NX: psz = "AMD64+NX"; break;
case SUPPAGINGMODE_AMD64_GLOBAL_NX: psz = "AMD64+G+NX"; break;
default: psz = "unknown"; break;
}
pHlp->pfnPrintf(pHlp, "Host paging mode: %s\n", psz);
}
}
/**
* Dump registered MMIO ranges to the log.
*
* @param pVM VM Handle.
* @param pHlp The info helpers.
* @param pszArgs Arguments, ignored.
*/
static DECLCALLBACK(void) pgmR3PhysInfo(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs)
{
NOREF(pszArgs);
pHlp->pfnPrintf(pHlp,
"RAM ranges (pVM=%p)\n"
"%.*s %.*s\n",
pVM,
sizeof(RTGCPHYS) * 4 + 1, "GC Phys Range ",
sizeof(RTHCPTR) * 2, "pvHC ");
for (PPGMRAMRANGE pCur = pVM->pgm.s.pRamRangesHC; pCur; pCur = pCur->pNextHC)
pHlp->pfnPrintf(pHlp,
"%VGp-%VGp %VHv\n",
pCur->GCPhys,
pCur->GCPhysLast,
pCur->pvHC);
}
/**
* Dump the page directory to the log.
*
* @param pVM VM Handle.
* @param pHlp The info helpers.
* @param pszArgs Arguments, ignored.
*/
static DECLCALLBACK(void) pgmR3InfoCr3(PVM pVM, PCDBGFINFOHLP pHlp, const char *pszArgs)
{
/** @todo fix this! Convert the PGMR3DumpHierarchyHC functions to do guest stuff. */
/* Big pages supported? */
const bool fPSE = !!(CPUMGetGuestCR4(pVM) & X86_CR4_PSE);
/* Global pages supported? */
const bool fPGE = !!(CPUMGetGuestCR4(pVM) & X86_CR4_PGE);
NOREF(pszArgs);
/*
* Get page directory addresses.
*/
PVBOXPD pPDSrc = pVM->pgm.s.pGuestPDHC;
Assert(pPDSrc);
Assert(MMPhysGCPhys2HCVirt(pVM, (RTGCPHYS)(CPUMGetGuestCR3(pVM) & X86_CR3_PAGE_MASK), sizeof(*pPDSrc)) == pPDSrc);
/*
* Iterate the page directory.
*/
for (unsigned iPD = 0; iPD < ELEMENTS(pPDSrc->a); iPD++)
{
VBOXPDE PdeSrc = pPDSrc->a[iPD];
if (PdeSrc.n.u1Present)
{
if (PdeSrc.b.u1Size && fPSE)
{
pHlp->pfnPrintf(pHlp,
"%04X - %VGp P=%d U=%d RW=%d G=%d - BIG\n",
iPD,
PdeSrc.u & X86_PDE_PG_MASK,
PdeSrc.b.u1Present, PdeSrc.b.u1User, PdeSrc.b.u1Write, PdeSrc.b.u1Global && fPGE);
}
else
{
pHlp->pfnPrintf(pHlp,
"%04X - %VGp P=%d U=%d RW=%d [G=%d]\n",
iPD,
PdeSrc.u & X86_PDE4M_PG_MASK,
PdeSrc.n.u1Present, PdeSrc.n.u1User, PdeSrc.n.u1Write, PdeSrc.b.u1Global && fPGE);
}
}
}
}
/**
* Serivce a VMMCALLHOST_PGM_LOCK call.
*
* @returns VBox status code.
* @param pVM The VM handle.
*/
PDMR3DECL(int) PGMR3LockCall(PVM pVM)
{
return pgmLock(pVM);
}
/**
* Converts a PGMMODE value to a PGM_TYPE_* \#define.
*
* @returns PGM_TYPE_*.
* @param pgmMode The mode value to convert.
*/
DECLINLINE(unsigned) pgmModeToType(PGMMODE pgmMode)
{
switch (pgmMode)
{
case PGMMODE_REAL: return PGM_TYPE_REAL;
case PGMMODE_PROTECTED: return PGM_TYPE_PROT;
case PGMMODE_32_BIT: return PGM_TYPE_32BIT;
case PGMMODE_PAE:
case PGMMODE_PAE_NX: return PGM_TYPE_PAE;
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX: return PGM_TYPE_AMD64;
default:
AssertFatalMsgFailed(("pgmMode=%d\n", pgmMode));
}
}
/**
* Gets the index into the paging mode data array of a SHW+GST mode.
*
* @returns PGM::paPagingData index.
* @param uShwType The shadow paging mode type.
* @param uGstType The guest paging mode type.
*/
DECLINLINE(unsigned) pgmModeDataIndex(unsigned uShwType, unsigned uGstType)
{
Assert(uShwType >= PGM_TYPE_32BIT && uShwType <= PGM_TYPE_AMD64);
Assert(uGstType >= PGM_TYPE_REAL && uGstType <= PGM_TYPE_AMD64);
return (uShwType - PGM_TYPE_32BIT) * (PGM_TYPE_AMD64 - PGM_TYPE_32BIT + 1)
+ (uGstType - PGM_TYPE_REAL);
}
/**
* Gets the index into the paging mode data array of a SHW+GST mode.
*
* @returns PGM::paPagingData index.
* @param enmShw The shadow paging mode.
* @param enmGst The guest paging mode.
*/
DECLINLINE(unsigned) pgmModeDataIndexByMode(PGMMODE enmShw, PGMMODE enmGst)
{
Assert(enmShw >= PGMMODE_32_BIT && enmShw <= PGMMODE_MAX);
Assert(enmGst > PGMMODE_INVALID && enmGst < PGMMODE_MAX);
return pgmModeDataIndex(pgmModeToType(enmShw), pgmModeToType(enmGst));
}
/**
* Calculates the max data index.
* @returns The number of entries in the pagaing data array.
*/
DECLINLINE(unsigned) pgmModeDataMaxIndex(void)
{
return pgmModeDataIndex(PGM_TYPE_AMD64, PGM_TYPE_AMD64) + 1;
}
/**
* Initializes the paging mode data kept in PGM::paModeData.
*
* @param pVM The VM handle.
* @param fResolveGCAndR0 Indicate whether or not GC and Ring-0 symbols can be resolved now.
* This is used early in the init process to avoid trouble with PDM
* not being initialized yet.
*/
static int pgmR3ModeDataInit(PVM pVM, bool fResolveGCAndR0)
{
PPGMMODEDATA pModeData;
int rc;
/*
* Allocate the array on the first call.
*/
if (!pVM->pgm.s.paModeData)
{
pVM->pgm.s.paModeData = (PPGMMODEDATA)MMR3HeapAllocZ(pVM, MM_TAG_PGM, sizeof(PGMMODEDATA) * pgmModeDataMaxIndex());
AssertReturn(pVM->pgm.s.paModeData, VERR_NO_MEMORY);
}
/*
* Initialize the array entries.
*/
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_32BIT, PGM_TYPE_REAL)];
pModeData->uShwType = PGM_TYPE_32BIT;
pModeData->uGstType = PGM_TYPE_REAL;
rc = PGM_SHW_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_32BIT_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_32BIT, PGMMODE_PROTECTED)];
pModeData->uShwType = PGM_TYPE_32BIT;
pModeData->uGstType = PGM_TYPE_PROT;
rc = PGM_SHW_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_32BIT_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_32BIT, PGM_TYPE_32BIT)];
pModeData->uShwType = PGM_TYPE_32BIT;
pModeData->uGstType = PGM_TYPE_32BIT;
rc = PGM_SHW_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_32BIT_32BIT(InitData)(pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_REAL)];
pModeData->uShwType = PGM_TYPE_PAE;
pModeData->uGstType = PGM_TYPE_REAL;
rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_PAE_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_PROT)];
pModeData->uShwType = PGM_TYPE_PAE;
pModeData->uGstType = PGM_TYPE_PROT;
rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_PAE_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_32BIT)];
pModeData->uShwType = PGM_TYPE_PAE;
pModeData->uGstType = PGM_TYPE_32BIT;
rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_PAE_32BIT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_PAE, PGM_TYPE_PAE)];
pModeData->uShwType = PGM_TYPE_PAE;
pModeData->uGstType = PGM_TYPE_PAE;
rc = PGM_SHW_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_PAE_PAE(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_AMD64, PGM_TYPE_REAL)];
pModeData->uShwType = PGM_TYPE_AMD64;
pModeData->uGstType = PGM_TYPE_REAL;
rc = PGM_SHW_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_AMD64_REAL(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_AMD64, PGM_TYPE_PROT)];
pModeData->uShwType = PGM_TYPE_AMD64;
pModeData->uGstType = PGM_TYPE_PROT;
rc = PGM_SHW_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_AMD64_PROT(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(PGM_TYPE_AMD64, PGM_TYPE_AMD64)];
pModeData->uShwType = PGM_TYPE_AMD64;
pModeData->uGstType = PGM_TYPE_AMD64;
rc = PGM_SHW_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_GST_NAME_AMD64(InitData)( pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
rc = PGM_BTH_NAME_AMD64_AMD64(InitData)(pVM, pModeData, fResolveGCAndR0); AssertRCReturn(rc, rc);
return VINF_SUCCESS;
}
/**
* Swtich to different (or relocated in the relocate case) mode data.
*
* @param pVM The VM handle.
* @param enmShw The the shadow paging mode.
* @param enmGst The the guest paging mode.
*/
static void pgmR3ModeDataSwitch(PVM pVM, PGMMODE enmShw, PGMMODE enmGst)
{
PPGMMODEDATA pModeData = &pVM->pgm.s.paModeData[pgmModeDataIndex(enmShw, enmGst)];
Assert(pModeData->uGstType == pgmModeToType(enmGst));
Assert(pModeData->uShwType == pgmModeToType(enmShw));
/* shadow */
pVM->pgm.s.pfnR3ShwRelocate = pModeData->pfnR3ShwRelocate;
pVM->pgm.s.pfnR3ShwExit = pModeData->pfnR3ShwExit;
pVM->pgm.s.pfnR3ShwGetPage = pModeData->pfnR3ShwGetPage;
Assert(pVM->pgm.s.pfnR3ShwGetPage);
pVM->pgm.s.pfnR3ShwModifyPage = pModeData->pfnR3ShwModifyPage;
pVM->pgm.s.pfnR3ShwGetPDEByIndex = pModeData->pfnR3ShwGetPDEByIndex;
pVM->pgm.s.pfnR3ShwSetPDEByIndex = pModeData->pfnR3ShwSetPDEByIndex;
pVM->pgm.s.pfnR3ShwModifyPDEByIndex = pModeData->pfnR3ShwModifyPDEByIndex;
pVM->pgm.s.pfnGCShwGetPage = pModeData->pfnGCShwGetPage;
pVM->pgm.s.pfnGCShwModifyPage = pModeData->pfnGCShwModifyPage;
pVM->pgm.s.pfnGCShwGetPDEByIndex = pModeData->pfnGCShwGetPDEByIndex;
pVM->pgm.s.pfnGCShwSetPDEByIndex = pModeData->pfnGCShwSetPDEByIndex;
pVM->pgm.s.pfnGCShwModifyPDEByIndex = pModeData->pfnGCShwModifyPDEByIndex;
pVM->pgm.s.pfnR0ShwGetPage = pModeData->pfnR0ShwGetPage;
pVM->pgm.s.pfnR0ShwModifyPage = pModeData->pfnR0ShwModifyPage;
pVM->pgm.s.pfnR0ShwGetPDEByIndex = pModeData->pfnR0ShwGetPDEByIndex;
pVM->pgm.s.pfnR0ShwSetPDEByIndex = pModeData->pfnR0ShwSetPDEByIndex;
pVM->pgm.s.pfnR0ShwModifyPDEByIndex = pModeData->pfnR0ShwModifyPDEByIndex;
/* guest */
pVM->pgm.s.pfnR3GstRelocate = pModeData->pfnR3GstRelocate;
pVM->pgm.s.pfnR3GstExit = pModeData->pfnR3GstExit;
pVM->pgm.s.pfnR3GstGetPage = pModeData->pfnR3GstGetPage;
Assert(pVM->pgm.s.pfnR3GstGetPage);
pVM->pgm.s.pfnR3GstModifyPage = pModeData->pfnR3GstModifyPage;
pVM->pgm.s.pfnR3GstGetPDE = pModeData->pfnR3GstGetPDE;
pVM->pgm.s.pfnR3GstMonitorCR3 = pModeData->pfnR3GstMonitorCR3;
pVM->pgm.s.pfnR3GstUnmonitorCR3 = pModeData->pfnR3GstUnmonitorCR3;
pVM->pgm.s.pfnR3GstMapCR3 = pModeData->pfnR3GstMapCR3;
pVM->pgm.s.pfnR3GstUnmapCR3 = pModeData->pfnR3GstUnmapCR3;
pVM->pgm.s.pfnHCGstWriteHandlerCR3 = pModeData->pfnHCGstWriteHandlerCR3;
pVM->pgm.s.pszHCGstWriteHandlerCR3 = pModeData->pszHCGstWriteHandlerCR3;
pVM->pgm.s.pfnGCGstGetPage = pModeData->pfnGCGstGetPage;
pVM->pgm.s.pfnGCGstModifyPage = pModeData->pfnGCGstModifyPage;
pVM->pgm.s.pfnGCGstGetPDE = pModeData->pfnGCGstGetPDE;
pVM->pgm.s.pfnGCGstMonitorCR3 = pModeData->pfnGCGstMonitorCR3;
pVM->pgm.s.pfnGCGstUnmonitorCR3 = pModeData->pfnGCGstUnmonitorCR3;
pVM->pgm.s.pfnGCGstMapCR3 = pModeData->pfnGCGstMapCR3;
pVM->pgm.s.pfnGCGstUnmapCR3 = pModeData->pfnGCGstUnmapCR3;
pVM->pgm.s.pfnGCGstWriteHandlerCR3 = pModeData->pfnGCGstWriteHandlerCR3;
pVM->pgm.s.pfnR0GstGetPage = pModeData->pfnR0GstGetPage;
pVM->pgm.s.pfnR0GstModifyPage = pModeData->pfnR0GstModifyPage;
pVM->pgm.s.pfnR0GstGetPDE = pModeData->pfnR0GstGetPDE;
pVM->pgm.s.pfnR0GstMonitorCR3 = pModeData->pfnR0GstMonitorCR3;
pVM->pgm.s.pfnR0GstUnmonitorCR3 = pModeData->pfnR0GstUnmonitorCR3;
pVM->pgm.s.pfnR0GstMapCR3 = pModeData->pfnR0GstMapCR3;
pVM->pgm.s.pfnR0GstUnmapCR3 = pModeData->pfnR0GstUnmapCR3;
pVM->pgm.s.pfnR0GstWriteHandlerCR3 = pModeData->pfnR0GstWriteHandlerCR3;
/* both */
pVM->pgm.s.pfnR3BthRelocate = pModeData->pfnR3BthRelocate;
pVM->pgm.s.pfnR3BthTrap0eHandler = pModeData->pfnR3BthTrap0eHandler;
pVM->pgm.s.pfnR3BthInvalidatePage = pModeData->pfnR3BthInvalidatePage;
pVM->pgm.s.pfnR3BthSyncCR3 = pModeData->pfnR3BthSyncCR3;
Assert(pVM->pgm.s.pfnR3BthSyncCR3);
pVM->pgm.s.pfnR3BthSyncPage = pModeData->pfnR3BthSyncPage;
pVM->pgm.s.pfnR3BthPrefetchPage = pModeData->pfnR3BthPrefetchPage;
pVM->pgm.s.pfnR3BthVerifyAccessSyncPage = pModeData->pfnR3BthVerifyAccessSyncPage;
#ifdef VBOX_STRICT
pVM->pgm.s.pfnR3BthAssertCR3 = pModeData->pfnR3BthAssertCR3;
#endif
pVM->pgm.s.pfnGCBthTrap0eHandler = pModeData->pfnGCBthTrap0eHandler;
pVM->pgm.s.pfnGCBthInvalidatePage = pModeData->pfnGCBthInvalidatePage;
pVM->pgm.s.pfnGCBthSyncCR3 = pModeData->pfnGCBthSyncCR3;
pVM->pgm.s.pfnGCBthSyncPage = pModeData->pfnGCBthSyncPage;
pVM->pgm.s.pfnGCBthPrefetchPage = pModeData->pfnGCBthPrefetchPage;
pVM->pgm.s.pfnGCBthVerifyAccessSyncPage = pModeData->pfnGCBthVerifyAccessSyncPage;
#ifdef VBOX_STRICT
pVM->pgm.s.pfnGCBthAssertCR3 = pModeData->pfnGCBthAssertCR3;
#endif
pVM->pgm.s.pfnR0BthTrap0eHandler = pModeData->pfnR0BthTrap0eHandler;
pVM->pgm.s.pfnR0BthInvalidatePage = pModeData->pfnR0BthInvalidatePage;
pVM->pgm.s.pfnR0BthSyncCR3 = pModeData->pfnR0BthSyncCR3;
pVM->pgm.s.pfnR0BthSyncPage = pModeData->pfnR0BthSyncPage;
pVM->pgm.s.pfnR0BthPrefetchPage = pModeData->pfnR0BthPrefetchPage;
pVM->pgm.s.pfnR0BthVerifyAccessSyncPage = pModeData->pfnR0BthVerifyAccessSyncPage;
#ifdef VBOX_STRICT
pVM->pgm.s.pfnR0BthAssertCR3 = pModeData->pfnR0BthAssertCR3;
#endif
}
#ifdef DEBUG_bird
#include <stdlib.h> /* getenv() remove me! */
#endif
/**
* Calculates the shadow paging mode.
*
* @returns The shadow paging mode.
* @param enmGuestMode The guest mode.
* @param enmHostMode The host mode.
* @param enmShadowMode The current shadow mode.
* @param penmSwitcher Where to store the switcher to use.
* VMMSWITCHER_INVALID means no change.
*/
static PGMMODE pgmR3CalcShadowMode(PGMMODE enmGuestMode, SUPPAGINGMODE enmHostMode, PGMMODE enmShadowMode, VMMSWITCHER *penmSwitcher)
{
VMMSWITCHER enmSwitcher = VMMSWITCHER_INVALID;
switch (enmGuestMode)
{
/*
* When switching to real or protected mode we don't change
* anything since it's likely that we'll switch back pretty soon.
*
* During pgmR3InitPaging we'll end up here with PGMMODE_INVALID
* and is supposed to determin which shadow paging and switcher to
* use during init.
*/
case PGMMODE_REAL:
case PGMMODE_PROTECTED:
if (enmShadowMode != PGMMODE_INVALID)
break; /* (no change) */
switch (enmHostMode)
{
case SUPPAGINGMODE_32_BIT:
case SUPPAGINGMODE_32_BIT_GLOBAL:
enmShadowMode = PGMMODE_32_BIT;
enmSwitcher = VMMSWITCHER_32_TO_32;
break;
case SUPPAGINGMODE_PAE:
case SUPPAGINGMODE_PAE_NX:
case SUPPAGINGMODE_PAE_GLOBAL:
case SUPPAGINGMODE_PAE_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_PAE_TO_PAE;
#ifdef DEBUG_bird
if (getenv("VBOX_32BIT"))
{
enmShadowMode = PGMMODE_32_BIT;
enmSwitcher = VMMSWITCHER_PAE_TO_32;
}
#endif
break;
case SUPPAGINGMODE_AMD64:
case SUPPAGINGMODE_AMD64_GLOBAL:
case SUPPAGINGMODE_AMD64_NX:
case SUPPAGINGMODE_AMD64_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_AMD64_TO_PAE;
break;
default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
}
break;
case PGMMODE_32_BIT:
switch (enmHostMode)
{
case SUPPAGINGMODE_32_BIT:
case SUPPAGINGMODE_32_BIT_GLOBAL:
enmShadowMode = PGMMODE_32_BIT;
enmSwitcher = VMMSWITCHER_32_TO_32;
break;
case SUPPAGINGMODE_PAE:
case SUPPAGINGMODE_PAE_NX:
case SUPPAGINGMODE_PAE_GLOBAL:
case SUPPAGINGMODE_PAE_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_PAE_TO_PAE;
#ifdef DEBUG_bird
if (getenv("VBOX_32BIT"))
{
enmShadowMode = PGMMODE_32_BIT;
enmSwitcher = VMMSWITCHER_PAE_TO_32;
}
#endif
break;
case SUPPAGINGMODE_AMD64:
case SUPPAGINGMODE_AMD64_GLOBAL:
case SUPPAGINGMODE_AMD64_NX:
case SUPPAGINGMODE_AMD64_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_AMD64_TO_PAE;
break;
default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
}
break;
case PGMMODE_PAE:
case PGMMODE_PAE_NX: /** @todo This might require more switchers and guest+both modes. */
switch (enmHostMode)
{
case SUPPAGINGMODE_32_BIT:
case SUPPAGINGMODE_32_BIT_GLOBAL:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_32_TO_PAE;
break;
case SUPPAGINGMODE_PAE:
case SUPPAGINGMODE_PAE_NX:
case SUPPAGINGMODE_PAE_GLOBAL:
case SUPPAGINGMODE_PAE_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_PAE_TO_PAE;
break;
case SUPPAGINGMODE_AMD64:
case SUPPAGINGMODE_AMD64_GLOBAL:
case SUPPAGINGMODE_AMD64_NX:
case SUPPAGINGMODE_AMD64_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_AMD64_TO_PAE;
break;
default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
}
break;
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX:
switch (enmHostMode)
{
case SUPPAGINGMODE_32_BIT:
case SUPPAGINGMODE_32_BIT_GLOBAL:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_32_TO_AMD64;
break;
case SUPPAGINGMODE_PAE:
case SUPPAGINGMODE_PAE_NX:
case SUPPAGINGMODE_PAE_GLOBAL:
case SUPPAGINGMODE_PAE_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_PAE_TO_AMD64;
break;
case SUPPAGINGMODE_AMD64:
case SUPPAGINGMODE_AMD64_GLOBAL:
case SUPPAGINGMODE_AMD64_NX:
case SUPPAGINGMODE_AMD64_GLOBAL_NX:
enmShadowMode = PGMMODE_PAE;
enmSwitcher = VMMSWITCHER_AMD64_TO_AMD64;
break;
default: AssertMsgFailed(("enmHostMode=%d\n", enmHostMode)); break;
}
break;
default:
AssertReleaseMsgFailed(("enmGuestMode=%d\n", enmGuestMode));
return PGMMODE_INVALID;
}
*penmSwitcher = enmSwitcher;
return enmShadowMode;
}
/**
* Performs the actual mode change.
* This is called by PGMChangeMode and pgmR3InitPaging().
*
* @returns VBox status code.
* @param pVM VM handle.
* @param enmGuestMode The new guest mode. This is assumed to be different from
* the current mode.
*/
int pgmR3ChangeMode(PVM pVM, PGMMODE enmGuestMode)
{
LogFlow(("pgmR3ChangeMode: Guest mode: %d -> %d\n", pVM->pgm.s.enmGuestMode, enmGuestMode));
STAM_REL_COUNTER_INC(&pVM->pgm.s.cGuestModeChanges);
/*
* Calc the shadow mode and switcher.
*/
VMMSWITCHER enmSwitcher;
PGMMODE enmShadowMode = pgmR3CalcShadowMode(enmGuestMode, pVM->pgm.s.enmHostMode, pVM->pgm.s.enmShadowMode, &enmSwitcher);
if (enmSwitcher != VMMSWITCHER_INVALID)
{
/*
* Select new switcher.
*/
int rc = VMMR3SelectSwitcher(pVM, enmSwitcher);
if (VBOX_FAILURE(rc))
{
AssertReleaseMsgFailed(("VMMR3SelectSwitcher(%d) -> %Vrc\n", enmSwitcher, rc));
return rc;
}
}
/*
* Exit old mode(s).
*/
/* shadow */
if (enmShadowMode != pVM->pgm.s.enmShadowMode)
{
LogFlow(("pgmR3ChangeMode: Shadow mode: %d -> %d\n", pVM->pgm.s.enmShadowMode, enmShadowMode));
if (PGM_SHW_PFN(Exit, pVM))
{
int rc = PGM_SHW_PFN(Exit, pVM)(pVM);
if (VBOX_FAILURE(rc))
{
AssertMsgFailed(("Exit failed for shadow mode %d: %Vrc\n", pVM->pgm.s.enmShadowMode, rc));
return rc;
}
}
}
/* guest */
if (PGM_GST_PFN(Exit, pVM))
{
int rc = PGM_GST_PFN(Exit, pVM)(pVM);
if (VBOX_FAILURE(rc))
{
AssertMsgFailed(("Exit failed for guest mode %d: %Vrc\n", pVM->pgm.s.enmGuestMode, rc));
return rc;
}
}
/*
* Load new paging mode data.
*/
pgmR3ModeDataSwitch(pVM, enmShadowMode, enmGuestMode);
/*
* Enter new shadow mode (if changed).
*/
if (enmShadowMode != pVM->pgm.s.enmShadowMode)
{
int rc;
pVM->pgm.s.enmShadowMode = enmShadowMode;
switch (enmShadowMode)
{
case PGMMODE_32_BIT:
rc = PGM_SHW_NAME_32BIT(Enter)(pVM);
break;
case PGMMODE_PAE:
case PGMMODE_PAE_NX:
rc = PGM_SHW_NAME_PAE(Enter)(pVM);
break;
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX:
rc = PGM_SHW_NAME_AMD64(Enter)(pVM);
break;
case PGMMODE_REAL:
case PGMMODE_PROTECTED:
default:
AssertReleaseMsgFailed(("enmShadowMode=%d\n", enmShadowMode));
return VERR_INTERNAL_ERROR;
}
if (VBOX_FAILURE(rc))
{
AssertReleaseMsgFailed(("Entering enmShadowMode=%d failed: %Vrc\n", enmShadowMode, rc));
pVM->pgm.s.enmShadowMode = PGMMODE_INVALID;
return rc;
}
}
/*
* Enter the new guest and shadow+guest modes.
*/
int rc = -1;
int rc2 = -1;
RTGCPHYS GCPhysCR3 = NIL_RTGCPHYS;
pVM->pgm.s.enmGuestMode = enmGuestMode;
switch (enmGuestMode)
{
case PGMMODE_REAL:
rc = PGM_GST_NAME_REAL(Enter)(pVM, NIL_RTGCPHYS);
switch (pVM->pgm.s.enmShadowMode)
{
case PGMMODE_32_BIT:
rc2 = PGM_BTH_NAME_32BIT_REAL(Enter)(pVM, NIL_RTGCPHYS);
break;
case PGMMODE_PAE:
case PGMMODE_PAE_NX:
rc2 = PGM_BTH_NAME_PAE_REAL(Enter)(pVM, NIL_RTGCPHYS);
break;
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX:
rc2 = PGM_BTH_NAME_AMD64_REAL(Enter)(pVM, NIL_RTGCPHYS);
break;
default: AssertFailed(); break;
}
break;
case PGMMODE_PROTECTED:
rc = PGM_GST_NAME_PROT(Enter)(pVM, NIL_RTGCPHYS);
switch (pVM->pgm.s.enmShadowMode)
{
case PGMMODE_32_BIT:
rc2 = PGM_BTH_NAME_32BIT_PROT(Enter)(pVM, NIL_RTGCPHYS);
break;
case PGMMODE_PAE:
case PGMMODE_PAE_NX:
rc2 = PGM_BTH_NAME_PAE_PROT(Enter)(pVM, NIL_RTGCPHYS);
break;
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX:
rc2 = PGM_BTH_NAME_AMD64_PROT(Enter)(pVM, NIL_RTGCPHYS);
break;
default: AssertFailed(); break;
}
break;
case PGMMODE_32_BIT:
GCPhysCR3 = CPUMGetGuestCR3(pVM) & X86_CR3_PAGE_MASK;
rc = PGM_GST_NAME_32BIT(Enter)(pVM, GCPhysCR3);
switch (pVM->pgm.s.enmShadowMode)
{
case PGMMODE_32_BIT:
rc2 = PGM_BTH_NAME_32BIT_32BIT(Enter)(pVM, GCPhysCR3);
break;
case PGMMODE_PAE:
case PGMMODE_PAE_NX:
rc2 = PGM_BTH_NAME_PAE_32BIT(Enter)(pVM, GCPhysCR3);
break;
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX:
AssertMsgFailed(("Should use PAE shadow mode!\n"));
default: AssertFailed(); break;
}
break;
//case PGMMODE_PAE_NX:
case PGMMODE_PAE:
GCPhysCR3 = CPUMGetGuestCR3(pVM) & X86_CR3_PAE_PAGE_MASK;
rc = PGM_GST_NAME_PAE(Enter)(pVM, GCPhysCR3);
switch (pVM->pgm.s.enmShadowMode)
{
case PGMMODE_PAE:
case PGMMODE_PAE_NX:
rc2 = PGM_BTH_NAME_PAE_PAE(Enter)(pVM, GCPhysCR3);
break;
case PGMMODE_32_BIT:
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX:
AssertMsgFailed(("Should use PAE shadow mode!\n"));
default: AssertFailed(); break;
}
break;
//case PGMMODE_AMD64_NX:
case PGMMODE_AMD64:
GCPhysCR3 = CPUMGetGuestCR3(pVM) & 0xfffffffffffff000ULL; /** @todo define this mask and make CR3 64-bit in this case! */
rc = PGM_GST_NAME_AMD64(Enter)(pVM, GCPhysCR3);
switch (pVM->pgm.s.enmShadowMode)
{
case PGMMODE_AMD64:
case PGMMODE_AMD64_NX:
rc2 = PGM_BTH_NAME_AMD64_AMD64(Enter)(pVM, GCPhysCR3);
break;
case PGMMODE_32_BIT:
case PGMMODE_PAE:
case PGMMODE_PAE_NX:
AssertMsgFailed(("Should use AMD64 shadow mode!\n"));
default: AssertFailed(); break;
}
break;
default:
AssertReleaseMsgFailed(("enmGuestMode=%d\n", enmGuestMode));
rc = VERR_NOT_IMPLEMENTED;
break;
}
/* status codes. */
AssertRC(rc);
AssertRC(rc2);
if (VBOX_SUCCESS(rc))
{
rc = rc2;
if (VBOX_SUCCESS(rc)) /* no informational status codes. */
rc = VINF_SUCCESS;
}
/*
* Notify SELM so it can update the TSSes with correct CR3s.
*/
SELMR3PagingModeChanged(pVM);
/* Notify HWACCM as well. */
HWACCMR3PagingModeChanged(pVM, pVM->pgm.s.enmShadowMode);
return rc;
}
/**
* Dumps a PAE shadow page table.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param pPT Pointer to the page table.
* @param u64Address The virtual address of the page table starts.
* @param fLongMode Set if this a long mode table; clear if it's a legacy mode table.
* @param cMaxDepth The maxium depth.
* @param pHlp Pointer to the output functions.
*/
static int pgmR3DumpHierarchyHCPaePT(PVM pVM, PX86PTPAE pPT, uint64_t u64Address, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
{
for (unsigned i = 0; i < ELEMENTS(pPT->a); i++)
{
X86PTEPAE Pte = pPT->a[i];
if (Pte.n.u1Present)
{
pHlp->pfnPrintf(pHlp,
fLongMode /*P R S A D G WT CD AT NX 4M a p ? */
? "%016llx 3 | P %c %c %c %c %c %s %s %s %s 4K %c%c%c %016llx\n"
: "%08llx 2 | P %c %c %c %c %c %s %s %s %s 4K %c%c%c %016llx\n",
u64Address + ((uint64_t)i << X86_PT_PAE_SHIFT),
Pte.n.u1Write ? 'W' : 'R',
Pte.n.u1User ? 'U' : 'S',
Pte.n.u1Accessed ? 'A' : '-',
Pte.n.u1Dirty ? 'D' : '-',
Pte.n.u1Global ? 'G' : '-',
Pte.n.u1WriteThru ? "WT" : "--",
Pte.n.u1CacheDisable? "CD" : "--",
Pte.n.u1PAT ? "AT" : "--",
Pte.n.u1NoExecute ? "NX" : "--",
Pte.u & PGM_PTFLAGS_TRACK_DIRTY ? 'd' : '-',
Pte.u & BIT(10) ? '1' : '0',
Pte.u & PGM_PTFLAGS_CSAM_VALIDATED? 'v' : '-',
Pte.u & X86_PTE_PAE_PG_MASK);
}
}
return VINF_SUCCESS;
}
/**
* Dumps a PAE shadow page directory table.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param HCPhys The physical address of the page directory table.
* @param u64Address The virtual address of the page table starts.
* @param cr4 The CR4, PSE is currently used.
* @param fLongMode Set if this a long mode table; clear if it's a legacy mode table.
* @param cMaxDepth The maxium depth.
* @param pHlp Pointer to the output functions.
*/
static int pgmR3DumpHierarchyHCPaePD(PVM pVM, RTHCPHYS HCPhys, uint64_t u64Address, uint32_t cr4, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
{
PX86PDPAE pPD = (PX86PDPAE)MMPagePhys2Page(pVM, HCPhys);
if (!pPD)
{
pHlp->pfnPrintf(pHlp, "%0*llx error! Page directory at HCPhys=%#VHp was not found in the page pool!\n",
fLongMode ? 16 : 8, u64Address, HCPhys);
return VERR_INVALID_PARAMETER;
}
int rc = VINF_SUCCESS;
for (unsigned i = 0; i < ELEMENTS(pPD->a); i++)
{
X86PDEPAE Pde = pPD->a[i];
if (Pde.n.u1Present)
{
if ((cr4 & X86_CR4_PSE) && Pde.b.u1Size)
pHlp->pfnPrintf(pHlp,
fLongMode /*P R S A D G WT CD AT NX 4M a p ? */
? "%016llx 2 | P %c %c %c %c %c %s %s %s %s 4M %c%c%c %016llx\n"
: "%08llx 1 | P %c %c %c %c %c %s %s %s %s 4M %c%c%c %016llx\n",
u64Address + ((uint64_t)i << X86_PD_PAE_SHIFT),
Pde.b.u1Write ? 'W' : 'R',
Pde.b.u1User ? 'U' : 'S',
Pde.b.u1Accessed ? 'A' : '-',
Pde.b.u1Dirty ? 'D' : '-',
Pde.b.u1Global ? 'G' : '-',
Pde.b.u1WriteThru ? "WT" : "--",
Pde.b.u1CacheDisable? "CD" : "--",
Pde.b.u1PAT ? "AT" : "--",
Pde.b.u1NoExecute ? "NX" : "--",
Pde.u & BIT64(9) ? '1' : '0',
Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
Pde.u & X86_PDE_PAE_PG_MASK);
else
{
pHlp->pfnPrintf(pHlp,
fLongMode /*P R S A D G WT CD AT NX 4M a p ? */
? "%016llx 2 | P %c %c %c %c %c %s %s .. %s 4K %c%c%c %016llx\n"
: "%08llx 1 | P %c %c %c %c %c %s %s .. %s 4K %c%c%c %016llx\n",
u64Address + ((uint64_t)i << X86_PD_PAE_SHIFT),
Pde.n.u1Write ? 'W' : 'R',
Pde.n.u1User ? 'U' : 'S',
Pde.n.u1Accessed ? 'A' : '-',
Pde.n.u1Reserved0 ? '?' : '.', /* ignored */
Pde.n.u1Reserved1 ? '?' : '.', /* ignored */
Pde.n.u1WriteThru ? "WT" : "--",
Pde.n.u1CacheDisable? "CD" : "--",
Pde.n.u1NoExecute ? "NX" : "--",
Pde.u & BIT64(9) ? '1' : '0',
Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
Pde.u & X86_PDE_PAE_PG_MASK);
if (cMaxDepth >= 1)
{
/** @todo what about using the page pool for mapping PTs? */
uint64_t u64AddressPT = u64Address + ((uint64_t)i << X86_PD_PAE_SHIFT);
RTHCPHYS HCPhysPT = Pde.u & X86_PDE_PAE_PG_MASK;
PX86PTPAE pPT = NULL;
if (!(Pde.u & PGM_PDFLAGS_MAPPING))
pPT = (PX86PTPAE)MMPagePhys2Page(pVM, HCPhysPT);
else
{
for (PPGMMAPPING pMap = pVM->pgm.s.pMappingsR3; pMap; pMap = pMap->pNextR3)
{
uint64_t off = u64AddressPT - pMap->GCPtr;
if (off < pMap->cb)
{
const int iPDE = (uint32_t)(off >> X86_PD_SHIFT);
const int iSub = (int)((off >> X86_PD_PAE_SHIFT) & 1); /* MSC is a pain sometimes */
if ((iSub ? pMap->aPTs[iPDE].HCPhysPaePT1 : pMap->aPTs[iPDE].HCPhysPaePT0) != HCPhysPT)
pHlp->pfnPrintf(pHlp, "%0*llx error! Mapping error! PT %d has HCPhysPT=%VHp not %VHp is in the PD.\n",
fLongMode ? 16 : 8, u64AddressPT, iPDE,
iSub ? pMap->aPTs[iPDE].HCPhysPaePT1 : pMap->aPTs[iPDE].HCPhysPaePT0, HCPhysPT);
pPT = &pMap->aPTs[iPDE].paPaePTsR3[iSub];
}
}
}
int rc2 = VERR_INVALID_PARAMETER;
if (pPT)
rc2 = pgmR3DumpHierarchyHCPaePT(pVM, pPT, u64AddressPT, fLongMode, cMaxDepth - 1, pHlp);
else
pHlp->pfnPrintf(pHlp, "%0*llx error! Page table at HCPhys=%#VHp was not found in the page pool!\n",
fLongMode ? 16 : 8, u64AddressPT, HCPhysPT);
if (rc2 < rc && VBOX_SUCCESS(rc))
rc = rc2;
}
}
}
}
return rc;
}
/**
* Dumps a PAE shadow page directory pointer table.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param HCPhys The physical address of the page directory pointer table.
* @param u64Address The virtual address of the page table starts.
* @param cr4 The CR4, PSE is currently used.
* @param fLongMode Set if this a long mode table; clear if it's a legacy mode table.
* @param cMaxDepth The maxium depth.
* @param pHlp Pointer to the output functions.
*/
static int pgmR3DumpHierarchyHCPaePDPTR(PVM pVM, RTHCPHYS HCPhys, uint64_t u64Address, uint32_t cr4, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
{
PX86PDPTR pPDPTR = (PX86PDPTR)MMPagePhys2Page(pVM, HCPhys);
if (!pPDPTR)
{
pHlp->pfnPrintf(pHlp, "%0*llx error! Page directory pointer table at HCPhys=%#VHp was not found in the page pool!\n",
fLongMode ? 16 : 8, u64Address, HCPhys);
return VERR_INVALID_PARAMETER;
}
int rc = VINF_SUCCESS;
const unsigned c = fLongMode ? ELEMENTS(pPDPTR->a) : 4;
for (unsigned i = 0; i < c; i++)
{
X86PDPE Pdpe = pPDPTR->a[i];
if (Pdpe.n.u1Present)
{
if (fLongMode)
pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a p ? */
"%016llx 1 | P %c %c %c %c %c %s %s %s %s .. %c%c%c %016llx\n",
u64Address + ((uint64_t)i << X86_PDPTR_SHIFT),
Pdpe.n.u1Write ? 'W' : 'R',
Pdpe.n.u1User ? 'U' : 'S',
Pdpe.n.u1Accessed ? 'A' : '-',
Pdpe.n.u3Reserved & 1? '?' : '.', /* ignored */
Pdpe.n.u3Reserved & 4? '!' : '.', /* mbz */
Pdpe.n.u1WriteThru ? "WT" : "--",
Pdpe.n.u1CacheDisable? "CD" : "--",
Pdpe.n.u3Reserved & 2? "!" : "..",/* mbz */
Pdpe.n.u1NoExecute ? "NX" : "--",
Pdpe.u & BIT(9) ? '1' : '0',
Pdpe.u & PGM_PLXFLAGS_PERMANENT ? 'p' : '-',
Pdpe.u & BIT(11) ? '1' : '0',
Pdpe.u & X86_PDPE_PG_MASK);
else
pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a p ? */
"%08x 0 | P %c %c %c %c %c %s %s %s %s .. %c%c%c %016llx\n",
i << X86_PDPTR_SHIFT,
Pdpe.n.u1Write ? '!' : '.', /* mbz */
Pdpe.n.u1User ? '!' : '.', /* mbz */
Pdpe.n.u1Accessed ? '!' : '.', /* mbz */
Pdpe.n.u3Reserved & 1? '!' : '.', /* mbz */
Pdpe.n.u3Reserved & 4? '!' : '.', /* mbz */
Pdpe.n.u1WriteThru ? "WT" : "--",
Pdpe.n.u1CacheDisable? "CD" : "--",
Pdpe.n.u3Reserved & 2? "!" : "..",/* mbz */
Pdpe.n.u1NoExecute ? "NX" : "--",
Pdpe.u & BIT(9) ? '1' : '0',
Pdpe.u & PGM_PLXFLAGS_PERMANENT ? 'p' : '-',
Pdpe.u & BIT(11) ? '1' : '0',
Pdpe.u & X86_PDPE_PG_MASK);
if (cMaxDepth >= 1)
{
int rc2 = pgmR3DumpHierarchyHCPaePD(pVM, Pdpe.u & X86_PDPE_PG_MASK, u64Address + ((uint64_t)i << X86_PDPTR_SHIFT),
cr4, fLongMode, cMaxDepth - 1, pHlp);
if (rc2 < rc && VBOX_SUCCESS(rc))
rc = rc2;
}
}
}
return rc;
}
/**
* Dumps a 32-bit shadow page table.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param HCPhys The physical address of the table.
* @param cr4 The CR4, PSE is currently used.
* @param cMaxDepth The maxium depth.
* @param pHlp Pointer to the output functions.
*/
static int pgmR3DumpHierarchyHcPaePML4(PVM pVM, RTHCPHYS HCPhys, uint32_t cr4, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
{
PX86PML4 pPML4 = (PX86PML4)MMPagePhys2Page(pVM, HCPhys);
if (!pPML4)
{
pHlp->pfnPrintf(pHlp, "Page map level 4 at HCPhys=%#VHp was not found in the page pool!\n", HCPhys);
return VERR_INVALID_PARAMETER;
}
int rc = VINF_SUCCESS;
for (unsigned i = 0; i < ELEMENTS(pPML4->a); i++)
{
X86PML4E Pml4e = pPML4->a[i];
if (Pml4e.n.u1Present)
{
uint64_t u64Address = ((uint64_t)i << X86_PML4_SHIFT) | (((uint64_t)i >> (X86_PML4_SHIFT - X86_PDPTR_SHIFT - 1)) * 0xffff000000000000ULL);
pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a p ? */
"%016llx 0 | P %c %c %c %c %c %s %s %s %s .. %c%c%c %016llx\n",
u64Address,
Pml4e.n.u1Write ? 'W' : 'R',
Pml4e.n.u1User ? 'U' : 'S',
Pml4e.n.u1Accessed ? 'A' : '-',
Pml4e.n.u3Reserved & 1? '?' : '.', /* ignored */
Pml4e.n.u3Reserved & 4? '!' : '.', /* mbz */
Pml4e.n.u1WriteThru ? "WT" : "--",
Pml4e.n.u1CacheDisable? "CD" : "--",
Pml4e.n.u3Reserved & 2? "!" : "..",/* mbz */
Pml4e.n.u1NoExecute ? "NX" : "--",
Pml4e.u & BIT(9) ? '1' : '0',
Pml4e.u & PGM_PLXFLAGS_PERMANENT ? 'p' : '-',
Pml4e.u & BIT(11) ? '1' : '0',
Pml4e.u & X86_PML4E_PG_MASK);
if (cMaxDepth >= 1)
{
int rc2 = pgmR3DumpHierarchyHCPaePDPTR(pVM, Pml4e.u & X86_PML4E_PG_MASK, u64Address, cr4, true, cMaxDepth - 1, pHlp);
if (rc2 < rc && VBOX_SUCCESS(rc))
rc = rc2;
}
}
}
return rc;
}
/**
* Dumps a 32-bit shadow page table.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param pPT Pointer to the page table.
* @param u32Address The virtual address this table starts at.
* @param pHlp Pointer to the output functions.
*/
int pgmR3DumpHierarchyHC32BitPT(PVM pVM, PX86PT pPT, uint32_t u32Address, PCDBGFINFOHLP pHlp)
{
for (unsigned i = 0; i < ELEMENTS(pPT->a); i++)
{
X86PTE Pte = pPT->a[i];
if (Pte.n.u1Present)
{
pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a m d */
"%08x 1 | P %c %c %c %c %c %s %s %s .. 4K %c%c%c %08x\n",
u32Address + (i << X86_PT_SHIFT),
Pte.n.u1Write ? 'W' : 'R',
Pte.n.u1User ? 'U' : 'S',
Pte.n.u1Accessed ? 'A' : '-',
Pte.n.u1Dirty ? 'D' : '-',
Pte.n.u1Global ? 'G' : '-',
Pte.n.u1WriteThru ? "WT" : "--",
Pte.n.u1CacheDisable? "CD" : "--",
Pte.n.u1PAT ? "AT" : "--",
Pte.u & PGM_PTFLAGS_TRACK_DIRTY ? 'd' : '-',
Pte.u & BIT(10) ? '1' : '0',
Pte.u & PGM_PTFLAGS_CSAM_VALIDATED ? 'v' : '-',
Pte.u & X86_PDE_PG_MASK);
}
}
return VINF_SUCCESS;
}
/**
* Dumps a 32-bit shadow page directory and page tables.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param cr3 The root of the hierarchy.
* @param cr4 The CR4, PSE is currently used.
* @param cMaxDepth How deep into the hierarchy the dumper should go.
* @param pHlp Pointer to the output functions.
*/
int pgmR3DumpHierarchyHC32BitPD(PVM pVM, uint32_t cr3, uint32_t cr4, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
{
PX86PD pPD = (PX86PD)MMPagePhys2Page(pVM, cr3 & X86_CR3_PAGE_MASK);
if (!pPD)
{
pHlp->pfnPrintf(pHlp, "Page directory at %#x was not found in the page pool!\n", cr3 & X86_CR3_PAGE_MASK);
return VERR_INVALID_PARAMETER;
}
int rc = VINF_SUCCESS;
for (unsigned i = 0; i < ELEMENTS(pPD->a); i++)
{
X86PDE Pde = pPD->a[i];
if (Pde.n.u1Present)
{
const uint32_t u32Address = i << X86_PD_SHIFT;
if ((cr4 & X86_CR4_PSE) && Pde.b.u1Size)
pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a m d */
"%08x 0 | P %c %c %c %c %c %s %s %s .. 4M %c%c%c %08x\n",
u32Address,
Pde.b.u1Write ? 'W' : 'R',
Pde.b.u1User ? 'U' : 'S',
Pde.b.u1Accessed ? 'A' : '-',
Pde.b.u1Dirty ? 'D' : '-',
Pde.b.u1Global ? 'G' : '-',
Pde.b.u1WriteThru ? "WT" : "--",
Pde.b.u1CacheDisable? "CD" : "--",
Pde.b.u1PAT ? "AT" : "--",
Pde.u & BIT64(9) ? '1' : '0',
Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
Pde.u & X86_PDE4M_PG_MASK);
else
{
pHlp->pfnPrintf(pHlp, /*P R S A D G WT CD AT NX 4M a m d */
"%08x 0 | P %c %c %c %c %c %s %s .. .. 4K %c%c%c %08x\n",
u32Address,
Pde.n.u1Write ? 'W' : 'R',
Pde.n.u1User ? 'U' : 'S',
Pde.n.u1Accessed ? 'A' : '-',
Pde.n.u1Reserved0 ? '?' : '.', /* ignored */
Pde.n.u1Reserved1 ? '?' : '.', /* ignored */
Pde.n.u1WriteThru ? "WT" : "--",
Pde.n.u1CacheDisable? "CD" : "--",
Pde.u & BIT64(9) ? '1' : '0',
Pde.u & PGM_PDFLAGS_MAPPING ? 'm' : '-',
Pde.u & PGM_PDFLAGS_TRACK_DIRTY ? 'd' : '-',
Pde.u & X86_PDE_PG_MASK);
if (cMaxDepth >= 1)
{
/** @todo what about using the page pool for mapping PTs? */
RTHCPHYS HCPhys = Pde.u & X86_PDE_PG_MASK;
PX86PT pPT = NULL;
if (!(Pde.u & PGM_PDFLAGS_MAPPING))
pPT = (PX86PT)MMPagePhys2Page(pVM, HCPhys);
else
{
for (PPGMMAPPING pMap = pVM->pgm.s.pMappingsR3; pMap; pMap = pMap->pNextR3)
if (u32Address - pMap->GCPtr < pMap->cb)
{
int iPDE = (u32Address - pMap->GCPtr) >> X86_PD_SHIFT;
if (pMap->aPTs[iPDE].HCPhysPT != HCPhys)
pHlp->pfnPrintf(pHlp, "%08x error! Mapping error! PT %d has HCPhysPT=%VHp not %VHp is in the PD.\n",
u32Address, iPDE, pMap->aPTs[iPDE].HCPhysPT, HCPhys);
pPT = pMap->aPTs[iPDE].pPTR3;
}
}
int rc2 = VERR_INVALID_PARAMETER;
if (pPT)
rc2 = pgmR3DumpHierarchyHC32BitPT(pVM, pPT, u32Address, pHlp);
else
pHlp->pfnPrintf(pHlp, "%08x error! Page table at %#x was not found in the page pool!\n", u32Address, HCPhys);
if (rc2 < rc && VBOX_SUCCESS(rc))
rc = rc2;
}
}
}
}
return rc;
}
/**
* Dumps a 32-bit shadow page table.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param pPT Pointer to the page table.
* @param u32Address The virtual address this table starts at.
* @param PhysSearch Address to search for.
*/
int pgmR3DumpHierarchyGC32BitPT(PVM pVM, PX86PT pPT, uint32_t u32Address, RTGCPHYS PhysSearch)
{
for (unsigned i = 0; i < ELEMENTS(pPT->a); i++)
{
X86PTE Pte = pPT->a[i];
if (Pte.n.u1Present)
{
Log(( /*P R S A D G WT CD AT NX 4M a m d */
"%08x 1 | P %c %c %c %c %c %s %s %s .. 4K %c%c%c %08x\n",
u32Address + (i << X86_PT_SHIFT),
Pte.n.u1Write ? 'W' : 'R',
Pte.n.u1User ? 'U' : 'S',
Pte.n.u1Accessed ? 'A' : '-',
Pte.n.u1Dirty ? 'D' : '-',
Pte.n.u1Global ? 'G' : '-',
Pte.n.u1WriteThru ? "WT" : "--",
Pte.n.u1CacheDisable? "CD" : "--",
Pte.n.u1PAT ? "AT" : "--",
Pte.u & PGM_PTFLAGS_TRACK_DIRTY ? 'd' : '-',
Pte.u & BIT(10) ? '1' : '0',
Pte.u & PGM_PTFLAGS_CSAM_VALIDATED ? 'v' : '-',
Pte.u & X86_PDE_PG_MASK));
if ((Pte.u & X86_PDE_PG_MASK) == PhysSearch)
{
uint64_t fPageShw = 0;
RTHCPHYS pPhysHC = 0;
PGMShwGetPage(pVM, (RTGCPTR)(u32Address + (i << X86_PT_SHIFT)), &fPageShw, &pPhysHC);
Log(("Found %VGp at %VGv -> flags=%llx\n", PhysSearch, (RTGCPTR)(u32Address + (i << X86_PT_SHIFT)), fPageShw));
}
}
}
return VINF_SUCCESS;
}
/**
* Dumps a 32-bit guest page directory and page tables.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param cr3 The root of the hierarchy.
* @param cr4 The CR4, PSE is currently used.
* @param PhysSearch Address to search for.
*/
PGMR3DECL(int) PGMR3DumpHierarchyGC(PVM pVM, uint32_t cr3, uint32_t cr4, RTGCPHYS PhysSearch)
{
bool fLongMode = false;
const unsigned cch = fLongMode ? 16 : 8; NOREF(cch);
PX86PD pPD = 0;
int rc = PGM_GCPHYS_2_PTR(pVM, cr3 & X86_CR3_PAGE_MASK, &pPD);
if (VBOX_FAILURE(rc) || !pPD)
{
Log(("Page directory at %#x was not found in the page pool!\n", cr3 & X86_CR3_PAGE_MASK));
return VERR_INVALID_PARAMETER;
}
Log(("cr3=%08x cr4=%08x%s\n"
"%-*s P - Present\n"
"%-*s | R/W - Read (0) / Write (1)\n"
"%-*s | | U/S - User (1) / Supervisor (0)\n"
"%-*s | | | A - Accessed\n"
"%-*s | | | | D - Dirty\n"
"%-*s | | | | | G - Global\n"
"%-*s | | | | | | WT - Write thru\n"
"%-*s | | | | | | | CD - Cache disable\n"
"%-*s | | | | | | | | AT - Attribute table (PAT)\n"
"%-*s | | | | | | | | | NX - No execute (K8)\n"
"%-*s | | | | | | | | | | 4K/4M/2M - Page size.\n"
"%-*s | | | | | | | | | | | AVL - a=allocated; m=mapping; d=track dirty;\n"
"%-*s | | | | | | | | | | | | p=permanent; v=validated;\n"
"%-*s Level | | | | | | | | | | | | Page\n"
/* xxxx n **** P R S A D G WT CD AT NX 4M AVL xxxxxxxxxxxxx
- W U - - - -- -- -- -- -- 010 */
, cr3, cr4, fLongMode ? " Long Mode" : "",
cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "",
cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "Address"));
for (unsigned i = 0; i < ELEMENTS(pPD->a); i++)
{
X86PDE Pde = pPD->a[i];
if (Pde.n.u1Present)
{
const uint32_t u32Address = i << X86_PD_SHIFT;
if ((cr4 & X86_CR4_PSE) && Pde.b.u1Size)
Log(( /*P R S A D G WT CD AT NX 4M a m d */
"%08x 0 | P %c %c %c %c %c %s %s %s .. 4M %c%c%c %08x\n",
u32Address,
Pde.b.u1Write ? 'W' : 'R',
Pde.b.u1User ? 'U' : 'S',
Pde.b.u1Accessed ? 'A' : '-',
Pde.b.u1Dirty ? 'D' : '-',
Pde.b.u1Global ? 'G' : '-',
Pde.b.u1WriteThru ? "WT" : "--",
Pde.b.u1CacheDisable? "CD" : "--",
Pde.b.u1PAT ? "AT" : "--",
Pde.u & BIT(9) ? '1' : '0',
Pde.u & BIT(10) ? '1' : '0',
Pde.u & BIT(11) ? '1' : '0',
Pde.u & X86_PDE4M_PG_MASK));
/** @todo PhysSearch */
else
{
Log(( /*P R S A D G WT CD AT NX 4M a m d */
"%08x 0 | P %c %c %c %c %c %s %s .. .. 4K %c%c%c %08x\n",
u32Address,
Pde.n.u1Write ? 'W' : 'R',
Pde.n.u1User ? 'U' : 'S',
Pde.n.u1Accessed ? 'A' : '-',
Pde.n.u1Reserved0 ? '?' : '.', /* ignored */
Pde.n.u1Reserved1 ? '?' : '.', /* ignored */
Pde.n.u1WriteThru ? "WT" : "--",
Pde.n.u1CacheDisable? "CD" : "--",
Pde.u & BIT(9) ? '1' : '0',
Pde.u & BIT(10) ? '1' : '0',
Pde.u & BIT(11) ? '1' : '0',
Pde.u & X86_PDE_PG_MASK));
////if (cMaxDepth >= 1)
{
/** @todo what about using the page pool for mapping PTs? */
RTGCPHYS GCPhys = Pde.u & X86_PDE_PG_MASK;
PX86PT pPT = NULL;
rc = PGM_GCPHYS_2_PTR(pVM, GCPhys, &pPT);
int rc2 = VERR_INVALID_PARAMETER;
if (pPT)
rc2 = pgmR3DumpHierarchyGC32BitPT(pVM, pPT, u32Address, PhysSearch);
else
Log(("%08x error! Page table at %#x was not found in the page pool!\n", u32Address, GCPhys));
if (rc2 < rc && VBOX_SUCCESS(rc))
rc = rc2;
}
}
}
}
return rc;
}
/**
* Dumps a page table hierarchy use only physical addresses and cr4/lm flags.
*
* @returns VBox status code (VINF_SUCCESS).
* @param pVM The VM handle.
* @param cr3 The root of the hierarchy.
* @param cr4 The cr4, only PAE and PSE is currently used.
* @param fLongMode Set if long mode, false if not long mode.
* @param cMaxDepth Number of levels to dump.
* @param pHlp Pointer to the output functions.
*/
PGMR3DECL(int) PGMR3DumpHierarchyHC(PVM pVM, uint32_t cr3, uint32_t cr4, bool fLongMode, unsigned cMaxDepth, PCDBGFINFOHLP pHlp)
{
if (!pHlp)
pHlp = DBGFR3InfoLogHlp();
if (!cMaxDepth)
return VINF_SUCCESS;
const unsigned cch = fLongMode ? 16 : 8;
pHlp->pfnPrintf(pHlp,
"cr3=%08x cr4=%08x%s\n"
"%-*s P - Present\n"
"%-*s | R/W - Read (0) / Write (1)\n"
"%-*s | | U/S - User (1) / Supervisor (0)\n"
"%-*s | | | A - Accessed\n"
"%-*s | | | | D - Dirty\n"
"%-*s | | | | | G - Global\n"
"%-*s | | | | | | WT - Write thru\n"
"%-*s | | | | | | | CD - Cache disable\n"
"%-*s | | | | | | | | AT - Attribute table (PAT)\n"
"%-*s | | | | | | | | | NX - No execute (K8)\n"
"%-*s | | | | | | | | | | 4K/4M/2M - Page size.\n"
"%-*s | | | | | | | | | | | AVL - a=allocated; m=mapping; d=track dirty;\n"
"%-*s | | | | | | | | | | | | p=permanent; v=validated;\n"
"%-*s Level | | | | | | | | | | | | Page\n"
/* xxxx n **** P R S A D G WT CD AT NX 4M AVL xxxxxxxxxxxxx
- W U - - - -- -- -- -- -- 010 */
, cr3, cr4, fLongMode ? " Long Mode" : "",
cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "",
cch, "", cch, "", cch, "", cch, "", cch, "", cch, "", cch, "Address");
if (cr4 & X86_CR4_PAE)
{
if (fLongMode)
return pgmR3DumpHierarchyHcPaePML4(pVM, cr3 & X86_CR3_PAGE_MASK, cr4, cMaxDepth, pHlp);
return pgmR3DumpHierarchyHCPaePDPTR(pVM, cr3 & X86_CR3_PAE_PAGE_MASK, 0, cr4, false, cMaxDepth, pHlp);
}
return pgmR3DumpHierarchyHC32BitPD(pVM, cr3 & X86_CR3_PAGE_MASK, cr4, cMaxDepth, pHlp);
}
#ifdef VBOX_WITH_DEBUGGER
/**
* The '.pgmram' command.
*
* @returns VBox status.
* @param pCmd Pointer to the command descriptor (as registered).
* @param pCmdHlp Pointer to command helper functions.
* @param pVM Pointer to the current VM (if any).
* @param paArgs Pointer to (readonly) array of arguments.
* @param cArgs Number of arguments in the array.
*/
static DECLCALLBACK(int) pgmR3CmdRam(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
{
/*
* Validate input.
*/
if (!pVM)
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires VM to be selected.\n");
if (!pVM->pgm.s.pRamRangesGC)
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Sorry, no Ram is registered.\n");
/*
* Dump the ranges.
*/
int rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL, "From - To (incl) pvHC\n");
PPGMRAMRANGE pRam;
for (pRam = pVM->pgm.s.pRamRangesHC; pRam; pRam = pRam->pNextHC)
{
rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL,
"%VGp - %VGp %p\n",
pRam->GCPhys, pRam->GCPhysLast, pRam->pvHC);
if (VBOX_FAILURE(rc))
return rc;
}
return VINF_SUCCESS;
}
/**
* The '.pgmmap' command.
*
* @returns VBox status.
* @param pCmd Pointer to the command descriptor (as registered).
* @param pCmdHlp Pointer to command helper functions.
* @param pVM Pointer to the current VM (if any).
* @param paArgs Pointer to (readonly) array of arguments.
* @param cArgs Number of arguments in the array.
*/
static DECLCALLBACK(int) pgmR3CmdMap(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
{
/*
* Validate input.
*/
if (!pVM)
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires VM to be selected.\n");
if (!pVM->pgm.s.pMappingsR3)
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Sorry, no mappings are registered.\n");
/*
* Print message about the fixedness of the mappings.
*/
int rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL, pVM->pgm.s.fMappingsFixed ? "The mappings are FIXED.\n" : "The mappings are FLOATING.\n");
if (VBOX_FAILURE(rc))
return rc;
/*
* Dump the ranges.
*/
PPGMMAPPING pCur;
for (pCur = pVM->pgm.s.pMappingsR3; pCur; pCur = pCur->pNextR3)
{
rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL,
"%08x - %08x %s\n",
pCur->GCPtr, pCur->GCPtrLast, pCur->pszDesc);
if (VBOX_FAILURE(rc))
return rc;
}
return VINF_SUCCESS;
}
/**
* The '.pgmsync' command.
*
* @returns VBox status.
* @param pCmd Pointer to the command descriptor (as registered).
* @param pCmdHlp Pointer to command helper functions.
* @param pVM Pointer to the current VM (if any).
* @param paArgs Pointer to (readonly) array of arguments.
* @param cArgs Number of arguments in the array.
*/
static DECLCALLBACK(int) pgmR3CmdSync(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
{
/*
* Validate input.
*/
if (!pVM)
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires VM to be selected.\n");
/*
* Force page directory sync.
*/
VM_FF_SET(pVM, VM_FF_PGM_SYNC_CR3);
int rc = pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Forcing page directory sync.\n");
if (VBOX_FAILURE(rc))
return rc;
return VINF_SUCCESS;
}
/**
* The '.pgmsyncalways' command.
*
* @returns VBox status.
* @param pCmd Pointer to the command descriptor (as registered).
* @param pCmdHlp Pointer to command helper functions.
* @param pVM Pointer to the current VM (if any).
* @param paArgs Pointer to (readonly) array of arguments.
* @param cArgs Number of arguments in the array.
*/
static DECLCALLBACK(int) pgmR3CmdSyncAlways(PCDBGCCMD pCmd, PDBGCCMDHLP pCmdHlp, PVM pVM, PCDBGCVAR paArgs, unsigned cArgs, PDBGCVAR pResult)
{
/*
* Validate input.
*/
if (!pVM)
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "error: The command requires VM to be selected.\n");
/*
* Force page directory sync.
*/
if (pVM->pgm.s.fSyncFlags & PGM_SYNC_ALWAYS)
{
ASMAtomicAndU32(&pVM->pgm.s.fSyncFlags, ~PGM_SYNC_ALWAYS);
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Disabled permanent forced page directory syncing.\n");
}
else
{
ASMAtomicOrU32(&pVM->pgm.s.fSyncFlags, PGM_SYNC_ALWAYS);
VM_FF_SET(pVM, VM_FF_PGM_SYNC_CR3);
return pCmdHlp->pfnPrintf(pCmdHlp, NULL, "Enabled permanent forced page directory syncing.\n");
}
}
#endif
/**
* pvUser argument of the pgmR3CheckIntegrity*Node callbacks.
*/
typedef struct PGMCHECKINTARGS
{
bool fLeftToRight; /**< true: left-to-right; false: right-to-left. */
PPGMPHYSHANDLER pPrevPhys;
PPGMVIRTHANDLER pPrevVirt;
PPGMPHYS2VIRTHANDLER pPrevPhys2Virt;
PVM pVM;
} PGMCHECKINTARGS, *PPGMCHECKINTARGS;
/**
* Validate a node in the physical handler tree.
*
* @returns 0 on if ok, other wise 1.
* @param pNode The handler node.
* @param pvUser pVM.
*/
static DECLCALLBACK(int) pgmR3CheckIntegrityPhysHandlerNode(PAVLROGCPHYSNODECORE pNode, void *pvUser)
{
PPGMCHECKINTARGS pArgs = (PPGMCHECKINTARGS)pvUser;
PPGMPHYSHANDLER pCur = (PPGMPHYSHANDLER)pNode;
AssertReleaseReturn(!((uintptr_t)pCur & 7), 1);
AssertReleaseMsg(pCur->Core.Key <= pCur->Core.KeyLast,("pCur=%p %VGp-%VGp %s\n", pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
AssertReleaseMsg( !pArgs->pPrevPhys
|| (pArgs->fLeftToRight ? pArgs->pPrevPhys->Core.KeyLast < pCur->Core.Key : pArgs->pPrevPhys->Core.KeyLast > pCur->Core.Key),
("pPrevPhys=%p %VGp-%VGp %s\n"
" pCur=%p %VGp-%VGp %s\n",
pArgs->pPrevPhys, pArgs->pPrevPhys->Core.Key, pArgs->pPrevPhys->Core.KeyLast, pArgs->pPrevPhys->pszDesc,
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
pArgs->pPrevPhys = pCur;
return 0;
}
/**
* Validate a node in the virtual handler tree.
*
* @returns 0 on if ok, other wise 1.
* @param pNode The handler node.
* @param pvUser pVM.
*/
static DECLCALLBACK(int) pgmR3CheckIntegrityVirtHandlerNode(PAVLROGCPTRNODECORE pNode, void *pvUser)
{
PPGMCHECKINTARGS pArgs = (PPGMCHECKINTARGS)pvUser;
PPGMVIRTHANDLER pCur = (PPGMVIRTHANDLER)pNode;
AssertReleaseReturn(!((uintptr_t)pCur & 7), 1);
AssertReleaseMsg(pCur->Core.Key <= pCur->Core.KeyLast,("pCur=%p %VGv-%VGv %s\n", pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
AssertReleaseMsg( !pArgs->pPrevVirt
|| (pArgs->fLeftToRight ? pArgs->pPrevVirt->Core.KeyLast < pCur->Core.Key : pArgs->pPrevVirt->Core.KeyLast > pCur->Core.Key),
("pPrevVirt=%p %VGv-%VGv %s\n"
" pCur=%p %VGv-%VGv %s\n",
pArgs->pPrevVirt, pArgs->pPrevVirt->Core.Key, pArgs->pPrevVirt->Core.KeyLast, pArgs->pPrevVirt->pszDesc,
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc));
for (unsigned iPage = 0; iPage < pCur->cPages; iPage++)
{
AssertReleaseMsg(pCur->aPhysToVirt[iPage].offVirtHandler == -RT_OFFSETOF(PGMVIRTHANDLER, aPhysToVirt[iPage]),
("pCur=%p %VGv-%VGv %s\n"
"iPage=%d offVirtHandle=%#x expected %#x\n",
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->pszDesc,
iPage, pCur->aPhysToVirt[iPage].offVirtHandler, -RT_OFFSETOF(PGMVIRTHANDLER, aPhysToVirt[iPage])));
}
pArgs->pPrevVirt = pCur;
return 0;
}
/**
* Validate a node in the virtual handler tree.
*
* @returns 0 on if ok, other wise 1.
* @param pNode The handler node.
* @param pvUser pVM.
*/
static DECLCALLBACK(int) pgmR3CheckIntegrityPhysToVirtHandlerNode(PAVLROGCPHYSNODECORE pNode, void *pvUser)
{
PPGMCHECKINTARGS pArgs = (PPGMCHECKINTARGS)pvUser;
PPGMPHYS2VIRTHANDLER pCur = (PPGMPHYS2VIRTHANDLER)pNode;
AssertReleaseMsgReturn(!((uintptr_t)pCur & 3), ("\n"), 1);
AssertReleaseMsgReturn(!(pCur->offVirtHandler & 3), ("\n"), 1);
AssertReleaseMsg(pCur->Core.Key <= pCur->Core.KeyLast,("pCur=%p %VGp-%VGp\n", pCur, pCur->Core.Key, pCur->Core.KeyLast));
AssertReleaseMsg( !pArgs->pPrevPhys2Virt
|| (pArgs->fLeftToRight ? pArgs->pPrevPhys2Virt->Core.KeyLast < pCur->Core.Key : pArgs->pPrevPhys2Virt->Core.KeyLast > pCur->Core.Key),
("pPrevPhys2Virt=%p %VGp-%VGp\n"
" pCur=%p %VGp-%VGp\n",
pArgs->pPrevPhys2Virt, pArgs->pPrevPhys2Virt->Core.Key, pArgs->pPrevPhys2Virt->Core.KeyLast,
pCur, pCur->Core.Key, pCur->Core.KeyLast));
AssertReleaseMsg( !pArgs->pPrevPhys2Virt
|| (pArgs->fLeftToRight ? pArgs->pPrevPhys2Virt->Core.KeyLast < pCur->Core.Key : pArgs->pPrevPhys2Virt->Core.KeyLast > pCur->Core.Key),
("pPrevPhys2Virt=%p %VGp-%VGp\n"
" pCur=%p %VGp-%VGp\n",
pArgs->pPrevPhys2Virt, pArgs->pPrevPhys2Virt->Core.Key, pArgs->pPrevPhys2Virt->Core.KeyLast,
pCur, pCur->Core.Key, pCur->Core.KeyLast));
AssertReleaseMsg((pCur->offNextAlias & (PGMPHYS2VIRTHANDLER_IN_TREE | PGMPHYS2VIRTHANDLER_IS_HEAD)) == (PGMPHYS2VIRTHANDLER_IN_TREE | PGMPHYS2VIRTHANDLER_IS_HEAD),
("pCur=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias));
if (pCur->offNextAlias & PGMPHYS2VIRTHANDLER_OFF_MASK)
{
PPGMPHYS2VIRTHANDLER pCur2 = pCur;
for (;;)
{
pCur2 = (PPGMPHYS2VIRTHANDLER)((intptr_t)pCur + (pCur->offNextAlias & PGMPHYS2VIRTHANDLER_OFF_MASK));
AssertReleaseMsg(pCur2 != pCur,
(" pCur=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias));
AssertReleaseMsg((pCur2->offNextAlias & (PGMPHYS2VIRTHANDLER_IN_TREE | PGMPHYS2VIRTHANDLER_IS_HEAD)) == PGMPHYS2VIRTHANDLER_IN_TREE,
(" pCur=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n"
"pCur2=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias,
pCur2, pCur2->Core.Key, pCur2->Core.KeyLast, pCur2->offVirtHandler, pCur2->offNextAlias));
AssertReleaseMsg((pCur2->Core.Key ^ pCur->Core.Key) < PAGE_SIZE,
(" pCur=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n"
"pCur2=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias,
pCur2, pCur2->Core.Key, pCur2->Core.KeyLast, pCur2->offVirtHandler, pCur2->offNextAlias));
AssertReleaseMsg((pCur2->Core.KeyLast ^ pCur->Core.KeyLast) < PAGE_SIZE,
(" pCur=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n"
"pCur2=%p:{.Core.Key=%VGp, .Core.KeyLast=%VGp, .offVirtHandler=%#RX32, .offNextAlias=%#RX32}\n",
pCur, pCur->Core.Key, pCur->Core.KeyLast, pCur->offVirtHandler, pCur->offNextAlias,
pCur2, pCur2->Core.Key, pCur2->Core.KeyLast, pCur2->offVirtHandler, pCur2->offNextAlias));
if (!(pCur2->offNextAlias & PGMPHYS2VIRTHANDLER_OFF_MASK))
break;
}
}
pArgs->pPrevPhys2Virt = pCur;
return 0;
}
/**
* Perform an integrity check on the PGM component.
*
* @returns VINF_SUCCESS if everything is fine.
* @returns VBox error status after asserting on integrity breach.
* @param pVM The VM handle.
*/
PDMR3DECL(int) PGMR3CheckIntegrity(PVM pVM)
{
AssertReleaseReturn(pVM->pgm.s.offVM, VERR_INTERNAL_ERROR);
/*
* Check the trees.
*/
int cErrors = 0;
PGMCHECKINTARGS Args = { true, NULL, NULL, NULL, pVM };
cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesHC->PhysHandlers, true, pgmR3CheckIntegrityPhysHandlerNode, &Args);
Args.fLeftToRight = false;
cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesHC->PhysHandlers, false, pgmR3CheckIntegrityPhysHandlerNode, &Args);
Args.fLeftToRight = true;
cErrors += RTAvlroGCPtrDoWithAll( &pVM->pgm.s.pTreesHC->VirtHandlers, true, pgmR3CheckIntegrityVirtHandlerNode, &Args);
Args.fLeftToRight = false;
cErrors += RTAvlroGCPtrDoWithAll( &pVM->pgm.s.pTreesHC->VirtHandlers, false, pgmR3CheckIntegrityVirtHandlerNode, &Args);
Args.fLeftToRight = true;
cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesHC->PhysToVirtHandlers, true, pgmR3CheckIntegrityPhysToVirtHandlerNode, &Args);
Args.fLeftToRight = false;
cErrors += RTAvlroGCPhysDoWithAll(&pVM->pgm.s.pTreesHC->PhysToVirtHandlers, false, pgmR3CheckIntegrityPhysToVirtHandlerNode, &Args);
return !cErrors ? VINF_SUCCESS : VERR_INTERNAL_ERROR;
}
/**
* Inform PGM if we want all mappings to be put into the shadow page table. (necessary for e.g. VMX)
*
* @returns VBox status code.
* @param pVM VM handle.
* @param fEnable Enable or disable shadow mappings
*/
PGMR3DECL(int) PGMR3ChangeShwPDMappings(PVM pVM, bool fEnable)
{
pVM->pgm.s.fDisableMappings = !fEnable;
size_t cb;
int rc = PGMR3MappingsSize(pVM, &cb);
AssertRCReturn(rc, rc);
/* Pretend the mappings are now fixed; to force a refresh of the reserved PDEs. */
rc = PGMR3MappingsFix(pVM, MM_HYPER_AREA_ADDRESS, cb);
AssertRCReturn(rc, rc);
return VINF_SUCCESS;
}