[linux-block.git] / arch / x86 / kvm / paging_tmpl.h

/*
 * Kernel-based Virtual Machine driver for Linux
 *
 * This module enables machines with Intel VT-x extensions to run virtual
 * machines without emulation or binary translation.
 *
 * MMU support
 *
 * Copyright (C) 2006 Qumranet, Inc.
 *
 * Authors:
 *   Yaniv Kamay  <yaniv@qumranet.com>
 *   Avi Kivity   <avi@qumranet.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 */

/*
 * We need the mmu code to access both 32-bit and 64-bit guest ptes,
 * so the code in this file is compiled twice, once per pte size.
 */

#if PTTYPE == 64
	#define pt_element_t u64
	#define guest_walker guest_walker64
	#define FNAME(name) paging##64_##name
	#define PT_BASE_ADDR_MASK PT64_BASE_ADDR_MASK
	#define PT_DIR_BASE_ADDR_MASK PT64_DIR_BASE_ADDR_MASK
	#define PT_INDEX(addr, level) PT64_INDEX(addr, level)
	#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
	#define PT_LEVEL_MASK(level) PT64_LEVEL_MASK(level)
	#define PT_LEVEL_BITS PT64_LEVEL_BITS
	#ifdef CONFIG_X86_64
	#define PT_MAX_FULL_LEVELS 4
	#define CMPXCHG cmpxchg
	#else
	#define CMPXCHG cmpxchg64
	#define PT_MAX_FULL_LEVELS 2
	#endif
#elif PTTYPE == 32
	#define pt_element_t u32
	#define guest_walker guest_walker32
	#define FNAME(name) paging##32_##name
	#define PT_BASE_ADDR_MASK PT32_BASE_ADDR_MASK
	#define PT_DIR_BASE_ADDR_MASK PT32_DIR_BASE_ADDR_MASK
	#define PT_INDEX(addr, level) PT32_INDEX(addr, level)
	#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
	#define PT_LEVEL_MASK(level) PT32_LEVEL_MASK(level)
	#define PT_LEVEL_BITS PT32_LEVEL_BITS
	#define PT_MAX_FULL_LEVELS 2
	#define CMPXCHG cmpxchg
#else
	#error Invalid PTTYPE value
#endif

#define gpte_to_gfn FNAME(gpte_to_gfn)
#define gpte_to_gfn_pde FNAME(gpte_to_gfn_pde)

/*
 * The guest_walker structure emulates the behavior of the hardware page
 * table walker.
 */
struct guest_walker {
	int level;
	gfn_t table_gfn[PT_MAX_FULL_LEVELS];
	pt_element_t ptes[PT_MAX_FULL_LEVELS];
	gpa_t pte_gpa[PT_MAX_FULL_LEVELS];
	unsigned pt_access;
	unsigned pte_access;
	gfn_t gfn;
	u32 error_code;
};

static gfn_t gpte_to_gfn(pt_element_t gpte)
{
	return (gpte & PT_BASE_ADDR_MASK) >> PAGE_SHIFT;
}

static gfn_t gpte_to_gfn_pde(pt_element_t gpte)
{
	return (gpte & PT_DIR_BASE_ADDR_MASK) >> PAGE_SHIFT;
}

static bool FNAME(cmpxchg_gpte)(struct kvm *kvm,
			 gfn_t table_gfn, unsigned index,
			 pt_element_t orig_pte, pt_element_t new_pte)
{
	pt_element_t ret;
	pt_element_t *table;
	struct page *page;

	down_read(&current->mm->mmap_sem);
	page = gfn_to_page(kvm, table_gfn);
	up_read(&current->mm->mmap_sem);

	table = kmap_atomic(page, KM_USER0);

	ret = CMPXCHG(&table[index], orig_pte, new_pte);

	kunmap_atomic(table, KM_USER0);

	kvm_release_page_dirty(page);

	return (ret != orig_pte);
}

static unsigned FNAME(gpte_access)(struct kvm_vcpu *vcpu, pt_element_t gpte)
{
	unsigned access;

	access = (gpte & (PT_WRITABLE_MASK | PT_USER_MASK)) | ACC_EXEC_MASK;
#if PTTYPE == 64
	if (is_nx(vcpu))
		access &= ~(gpte >> PT64_NX_SHIFT);
#endif
	return access;
}

/*
 * Fetch a guest pte for a guest virtual address
 */
static int FNAME(walk_addr)(struct guest_walker *walker,
			    struct kvm_vcpu *vcpu, gva_t addr,
			    int write_fault, int user_fault, int fetch_fault)
{
	pt_element_t pte;
	gfn_t table_gfn;
	unsigned index, pt_access, pte_access;
	gpa_t pte_gpa;

	pgprintk("%s: addr %lx\n", __func__, addr);
walk:
	walker->level = vcpu->arch.mmu.root_level;
	pte = vcpu->arch.cr3;
#if PTTYPE == 64
	if (!is_long_mode(vcpu)) {
		pte = vcpu->arch.pdptrs[(addr >> 30) & 3];
		if (!is_present_pte(pte))
			goto not_present;
		--walker->level;
	}
#endif
	ASSERT((!is_long_mode(vcpu) && is_pae(vcpu)) ||
	       (vcpu->arch.cr3 & CR3_NONPAE_RESERVED_BITS) == 0);

	pt_access = ACC_ALL;

	for (;;) {
		index = PT_INDEX(addr, walker->level);

		table_gfn = gpte_to_gfn(pte);
		pte_gpa = gfn_to_gpa(table_gfn);
		pte_gpa += index * sizeof(pt_element_t);
		walker->table_gfn[walker->level - 1] = table_gfn;
		walker->pte_gpa[walker->level - 1] = pte_gpa;
		pgprintk("%s: table_gfn[%d] %lx\n", __func__,
			 walker->level - 1, table_gfn);

		kvm_read_guest(vcpu->kvm, pte_gpa, &pte, sizeof(pte));

		if (!is_present_pte(pte))
			goto not_present;

		if (write_fault && !is_writeble_pte(pte))
			if (user_fault || is_write_protection(vcpu))
				goto access_error;

		if (user_fault && !(pte & PT_USER_MASK))
			goto access_error;

#if PTTYPE == 64
		if (fetch_fault && is_nx(vcpu) && (pte & PT64_NX_MASK))
			goto access_error;
#endif

		if (!(pte & PT_ACCESSED_MASK)) {
			mark_page_dirty(vcpu->kvm, table_gfn);
			if (FNAME(cmpxchg_gpte)(vcpu->kvm, table_gfn,
			    index, pte, pte|PT_ACCESSED_MASK))
				goto walk;
			pte |= PT_ACCESSED_MASK;
		}

		pte_access = pt_access & FNAME(gpte_access)(vcpu, pte);

		walker->ptes[walker->level - 1] = pte;

		if (walker->level == PT_PAGE_TABLE_LEVEL) {
			walker->gfn = gpte_to_gfn(pte);
			break;
		}

		if (walker->level == PT_DIRECTORY_LEVEL
		    && (pte & PT_PAGE_SIZE_MASK)
		    && (PTTYPE == 64 || is_pse(vcpu))) {
			walker->gfn = gpte_to_gfn_pde(pte);
			walker->gfn += PT_INDEX(addr, PT_PAGE_TABLE_LEVEL);
			if (PTTYPE == 32 && is_cpuid_PSE36())
				walker->gfn += pse36_gfn_delta(pte);
			break;
		}

		pt_access = pte_access;
		--walker->level;
	}

	if (write_fault && !is_dirty_pte(pte)) {
		bool ret;

		mark_page_dirty(vcpu->kvm, table_gfn);
		ret = FNAME(cmpxchg_gpte)(vcpu->kvm, table_gfn, index, pte,
			    pte|PT_DIRTY_MASK);
		if (ret)
			goto walk;
		pte |= PT_DIRTY_MASK;
		kvm_mmu_pte_write(vcpu, pte_gpa, (u8 *)&pte, sizeof(pte));
		walker->ptes[walker->level - 1] = pte;
	}

	walker->pt_access = pt_access;
	walker->pte_access = pte_access;
	pgprintk("%s: pte %llx pte_access %x pt_access %x\n",
		 __func__, (u64)pte, pt_access, pte_access);
	return 1;

not_present:
	walker->error_code = 0;
	goto err;

access_error:
	walker->error_code = PFERR_PRESENT_MASK;

err:
	if (write_fault)
		walker->error_code |= PFERR_WRITE_MASK;
	if (user_fault)
		walker->error_code |= PFERR_USER_MASK;
	if (fetch_fault)
		walker->error_code |= PFERR_FETCH_MASK;
	return 0;
}

static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *page,
			      u64 *spte, const void *pte)
{
	pt_element_t gpte;
	unsigned pte_access;
	pfn_t pfn;
	int largepage = vcpu->arch.update_pte.largepage;

	gpte = *(const pt_element_t *)pte;
	if (~gpte & (PT_PRESENT_MASK | PT_ACCESSED_MASK)) {
		if (!is_present_pte(gpte))
			set_shadow_pte(spte, shadow_notrap_nonpresent_pte);
		return;
	}
	pgprintk("%s: gpte %llx spte %p\n", __func__, (u64)gpte, spte);
	pte_access = page->role.access & FNAME(gpte_access)(vcpu, gpte);
	if (gpte_to_gfn(gpte) != vcpu->arch.update_pte.gfn)
		return;
	pfn = vcpu->arch.update_pte.pfn;
	if (is_error_pfn(pfn))
		return;
	kvm_get_pfn(pfn);
	mmu_set_spte(vcpu, spte, page->role.access, pte_access, 0, 0,
		     gpte & PT_DIRTY_MASK, NULL, largepage, gpte_to_gfn(gpte),
		     pfn, true);
}

/*
 * Fetch a shadow pte for a specific level in the paging hierarchy.
 */
static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
			 struct guest_walker *walker,
			 int user_fault, int write_fault, int largepage,
			 int *ptwrite, pfn_t pfn)
{
	hpa_t shadow_addr;
	int level;
	u64 *shadow_ent;
	unsigned access = walker->pt_access;

	if (!is_present_pte(walker->ptes[walker->level - 1]))
		return NULL;

	shadow_addr = vcpu->arch.mmu.root_hpa;
	level = vcpu->arch.mmu.shadow_root_level;
	if (level == PT32E_ROOT_LEVEL) {
		shadow_addr = vcpu->arch.mmu.pae_root[(addr >> 30) & 3];
		shadow_addr &= PT64_BASE_ADDR_MASK;
		--level;
	}

	for (; ; level--) {
		u32 index = SHADOW_PT_INDEX(addr, level);
		struct kvm_mmu_page *shadow_page;
		u64 shadow_pte;
		int metaphysical;
		gfn_t table_gfn;

		shadow_ent = ((u64 *)__va(shadow_addr)) + index;
		if (level == PT_PAGE_TABLE_LEVEL)
			break;

		if (largepage && level == PT_DIRECTORY_LEVEL)
			break;

		if (is_shadow_present_pte(*shadow_ent)
		    && !is_large_pte(*shadow_ent)) {
			shadow_addr = *shadow_ent & PT64_BASE_ADDR_MASK;
			continue;
		}

		if (is_large_pte(*shadow_ent))
			rmap_remove(vcpu->kvm, shadow_ent);

		if (level - 1 == PT_PAGE_TABLE_LEVEL
		    && walker->level == PT_DIRECTORY_LEVEL) {
			metaphysical = 1;
			if (!is_dirty_pte(walker->ptes[level - 1]))
				access &= ~ACC_WRITE_MASK;
			table_gfn = gpte_to_gfn(walker->ptes[level - 1]);
		} else {
			metaphysical = 0;
			table_gfn = walker->table_gfn[level - 2];
		}
		shadow_page = kvm_mmu_get_page(vcpu, table_gfn, addr, level-1,
					       metaphysical, access,
					       shadow_ent);
		if (!metaphysical) {
			int r;
			pt_element_t curr_pte;
			r = kvm_read_guest_atomic(vcpu->kvm,
						  walker->pte_gpa[level - 2],
						  &curr_pte, sizeof(curr_pte));
			if (r || curr_pte != walker->ptes[level - 2]) {
				kvm_release_pfn_clean(pfn);
				return NULL;
			}
		}
		shadow_addr = __pa(shadow_page->spt);
		shadow_pte = shadow_addr | PT_PRESENT_MASK | PT_ACCESSED_MASK
			| PT_WRITABLE_MASK | PT_USER_MASK;
		*shadow_ent = shadow_pte;
	}

	mmu_set_spte(vcpu, shadow_ent, access, walker->pte_access & access,
		     user_fault, write_fault,
		     walker->ptes[walker->level-1] & PT_DIRTY_MASK,
		     ptwrite, largepage, walker->gfn, pfn, false);

	return shadow_ent;
}

/*
 * Page fault handler.  There are several causes for a page fault:
 *   - there is no shadow pte for the guest pte
 *   - write access through a shadow pte marked read only so that we can set
 *     the dirty bit
 *   - write access to a shadow pte marked read only so we can update the page
 *     dirty bitmap, when userspace requests it
 *   - mmio access; in this case we will never install a present shadow pte
 *   - normal guest page fault due to the guest pte marked not present, not
 *     writable, or not executable
 *
 *  Returns: 1 if we need to emulate the instruction, 0 otherwise, or
 *           a negative value on error.
 */
static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
			       u32 error_code)
{
	int write_fault = error_code & PFERR_WRITE_MASK;
	int user_fault = error_code & PFERR_USER_MASK;
	int fetch_fault = error_code & PFERR_FETCH_MASK;
	struct guest_walker walker;
	u64 *shadow_pte;
	int write_pt = 0;
	int r;
	pfn_t pfn;
	int largepage = 0;

	pgprintk("%s: addr %lx err %x\n", __func__, addr, error_code);
	kvm_mmu_audit(vcpu, "pre page fault");

	r = mmu_topup_memory_caches(vcpu);
	if (r)
		return r;

	/*
	 * Look up the shadow pte for the faulting address.
	 */
	r = FNAME(walk_addr)(&walker, vcpu, addr, write_fault, user_fault,
			     fetch_fault);

	/*
	 * The page is not mapped by the guest.  Let the guest handle it.
	 */
	if (!r) {
		pgprintk("%s: guest page fault\n", __func__);
		inject_page_fault(vcpu, addr, walker.error_code);
		vcpu->arch.last_pt_write_count = 0; /* reset fork detector */
		return 0;
	}

	down_read(&current->mm->mmap_sem);
	if (walker.level == PT_DIRECTORY_LEVEL) {
		gfn_t large_gfn;
		large_gfn = walker.gfn & ~(KVM_PAGES_PER_HPAGE-1);
		if (is_largepage_backed(vcpu, large_gfn)) {
			walker.gfn = large_gfn;
			largepage = 1;
		}
	}
	pfn = gfn_to_pfn(vcpu->kvm, walker.gfn);
	up_read(&current->mm->mmap_sem);

	/* mmio */
	if (is_error_pfn(pfn)) {
		pgprintk("gfn %x is mmio\n", walker.gfn);
		kvm_release_pfn_clean(pfn);
		return 1;
	}

	spin_lock(&vcpu->kvm->mmu_lock);
	kvm_mmu_free_some_pages(vcpu);
	shadow_pte = FNAME(fetch)(vcpu, addr, &walker, user_fault, write_fault,
				  largepage, &write_pt, pfn);

	pgprintk("%s: shadow pte %p %llx ptwrite %d\n", __func__,
		 shadow_pte, *shadow_pte, write_pt);

	if (!write_pt)
		vcpu->arch.last_pt_write_count = 0; /* reset fork detector */

	++vcpu->stat.pf_fixed;
	kvm_mmu_audit(vcpu, "post page fault (fixed)");
	spin_unlock(&vcpu->kvm->mmu_lock);

	return write_pt;
}

static gpa_t FNAME(gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t vaddr)
{
	struct guest_walker walker;
	gpa_t gpa = UNMAPPED_GVA;
	int r;

	r = FNAME(walk_addr)(&walker, vcpu, vaddr, 0, 0, 0);

	if (r) {
		gpa = gfn_to_gpa(walker.gfn);
		gpa |= vaddr & ~PAGE_MASK;
	}

	return gpa;
}

static void FNAME(prefetch_page)(struct kvm_vcpu *vcpu,
				 struct kvm_mmu_page *sp)
{
	int i, offset = 0, r = 0;
	pt_element_t pt;

	if (sp->role.metaphysical
	    || (PTTYPE == 32 && sp->role.level > PT_PAGE_TABLE_LEVEL)) {
		nonpaging_prefetch_page(vcpu, sp);
		return;
	}

	if (PTTYPE == 32)
		offset = sp->role.quadrant << PT64_LEVEL_BITS;

	for (i = 0; i < PT64_ENT_PER_PAGE; ++i) {
		gpa_t pte_gpa = gfn_to_gpa(sp->gfn);
		pte_gpa += (i+offset) * sizeof(pt_element_t);

		r = kvm_read_guest_atomic(vcpu->kvm, pte_gpa, &pt,
					  sizeof(pt_element_t));
		if (r || is_present_pte(pt))
			sp->spt[i] = shadow_trap_nonpresent_pte;
		else
			sp->spt[i] = shadow_notrap_nonpresent_pte;
	}
}

#undef pt_element_t
#undef guest_walker
#undef FNAME
#undef PT_BASE_ADDR_MASK
#undef PT_INDEX
#undef SHADOW_PT_INDEX
#undef PT_LEVEL_MASK
#undef PT_DIR_BASE_ADDR_MASK
#undef PT_LEVEL_BITS
#undef PT_MAX_FULL_LEVELS
#undef gpte_to_gfn
#undef gpte_to_gfn_pde
#undef CMPXCHG
Commit	Line	Data
6aa8b732 AK	1	/*
	2	* Kernel-based Virtual Machine driver for Linux
	3	*
	4	* This module enables machines with Intel VT-x extensions to run virtual
	5	* machines without emulation or binary translation.
	6	*
	7	* MMU support
	8	*
	9	* Copyright (C) 2006 Qumranet, Inc.
	10	*
	11	* Authors:
	12	* Yaniv Kamay <yaniv@qumranet.com>
	13	* Avi Kivity <avi@qumranet.com>
	14	*
	15	* This work is licensed under the terms of the GNU GPL, version 2. See
	16	* the COPYING file in the top-level directory.
	17	*
	18	*/
	19
	20	/*
	21	* We need the mmu code to access both 32-bit and 64-bit guest ptes,
	22	* so the code in this file is compiled twice, once per pte size.
	23	*/
	24
	25	#if PTTYPE == 64
	26	#define pt_element_t u64
	27	#define guest_walker guest_walker64
	28	#define FNAME(name) paging##64_##name
	29	#define PT_BASE_ADDR_MASK PT64_BASE_ADDR_MASK
	30	#define PT_DIR_BASE_ADDR_MASK PT64_DIR_BASE_ADDR_MASK
	31	#define PT_INDEX(addr, level) PT64_INDEX(addr, level)
	32	#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
	33	#define PT_LEVEL_MASK(level) PT64_LEVEL_MASK(level)
c7addb90	34	#define PT_LEVEL_BITS PT64_LEVEL_BITS
cea0f0e7 AK	35	#ifdef CONFIG_X86_64
cea0f0e7 AK	36	#define PT_MAX_FULL_LEVELS 4
b3e4e63f	37	#define CMPXCHG cmpxchg
cea0f0e7	38	#else
b3e4e63f	39	#define CMPXCHG cmpxchg64
cea0f0e7 AK	40	#define PT_MAX_FULL_LEVELS 2
cea0f0e7 AK	41	#endif
6aa8b732 AK	42	#elif PTTYPE == 32
	43	#define pt_element_t u32
	44	#define guest_walker guest_walker32
	45	#define FNAME(name) paging##32_##name
	46	#define PT_BASE_ADDR_MASK PT32_BASE_ADDR_MASK
	47	#define PT_DIR_BASE_ADDR_MASK PT32_DIR_BASE_ADDR_MASK
	48	#define PT_INDEX(addr, level) PT32_INDEX(addr, level)
	49	#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
	50	#define PT_LEVEL_MASK(level) PT32_LEVEL_MASK(level)
c7addb90	51	#define PT_LEVEL_BITS PT32_LEVEL_BITS
cea0f0e7	52	#define PT_MAX_FULL_LEVELS 2
b3e4e63f	53	#define CMPXCHG cmpxchg
6aa8b732 AK	54	#else
	55	#error Invalid PTTYPE value
	56	#endif
	57
5fb07ddb AK	58	#define gpte_to_gfn FNAME(gpte_to_gfn)
	59	#define gpte_to_gfn_pde FNAME(gpte_to_gfn_pde)
	60
6aa8b732 AK	61	/*
	62	* The guest_walker structure emulates the behavior of the hardware page
	63	* table walker.
	64	*/
	65	struct guest_walker {
	66	int level;
cea0f0e7	67	gfn_t table_gfn[PT_MAX_FULL_LEVELS];
7819026e MT	68	pt_element_t ptes[PT_MAX_FULL_LEVELS];
7819026e MT	69	gpa_t pte_gpa[PT_MAX_FULL_LEVELS];
fe135d2c AK	70	unsigned pt_access;
fe135d2c AK	71	unsigned pte_access;
815af8d4	72	gfn_t gfn;
7993ba43	73	u32 error_code;
6aa8b732 AK	74	};
6aa8b732 AK	75
5fb07ddb AK	76	static gfn_t gpte_to_gfn(pt_element_t gpte)
	77	{
	78	return (gpte & PT_BASE_ADDR_MASK) >> PAGE_SHIFT;
	79	}
	80
	81	static gfn_t gpte_to_gfn_pde(pt_element_t gpte)
	82	{
	83	return (gpte & PT_DIR_BASE_ADDR_MASK) >> PAGE_SHIFT;
	84	}
	85
b3e4e63f MT	86	static bool FNAME(cmpxchg_gpte)(struct kvm *kvm,
	87	gfn_t table_gfn, unsigned index,
	88	pt_element_t orig_pte, pt_element_t new_pte)
	89	{
	90	pt_element_t ret;
	91	pt_element_t *table;
	92	struct page *page;
	93
72dc67a6	94	down_read(&current->mm->mmap_sem);
b3e4e63f	95	page = gfn_to_page(kvm, table_gfn);
72dc67a6 IE	96	up_read(&current->mm->mmap_sem);
72dc67a6 IE	97
b3e4e63f MT	98	table = kmap_atomic(page, KM_USER0);
	99
	100	ret = CMPXCHG(&table[index], orig_pte, new_pte);
	101
	102	kunmap_atomic(table, KM_USER0);
	103
	104	kvm_release_page_dirty(page);
	105
	106	return (ret != orig_pte);
	107	}
	108
bedbe4ee AK	109	static unsigned FNAME(gpte_access)(struct kvm_vcpu *vcpu, pt_element_t gpte)
	110	{
	111	unsigned access;
	112
	113	access = (gpte & (PT_WRITABLE_MASK \| PT_USER_MASK)) \| ACC_EXEC_MASK;
	114	#if PTTYPE == 64
	115	if (is_nx(vcpu))
	116	access &= ~(gpte >> PT64_NX_SHIFT);
	117	#endif
	118	return access;
	119	}
	120
ac79c978 AK	121	/*
	122	* Fetch a guest pte for a guest virtual address
	123	*/
7993ba43 AK	124	static int FNAME(walk_addr)(struct guest_walker *walker,
7993ba43 AK	125	struct kvm_vcpu *vcpu, gva_t addr,
73b1087e	126	int write_fault, int user_fault, int fetch_fault)
6aa8b732	127	{
42bf3f0a	128	pt_element_t pte;
cea0f0e7	129	gfn_t table_gfn;
fe135d2c	130	unsigned index, pt_access, pte_access;
42bf3f0a	131	gpa_t pte_gpa;
6aa8b732	132
b8688d51	133	pgprintk("%s: addr %lx\n", __func__, addr);
b3e4e63f	134	walk:
ad312c7c ZX	135	walker->level = vcpu->arch.mmu.root_level;
ad312c7c ZX	136	pte = vcpu->arch.cr3;
1b0973bd AK	137	#if PTTYPE == 64
1b0973bd AK	138	if (!is_long_mode(vcpu)) {
ad312c7c	139	pte = vcpu->arch.pdptrs[(addr >> 30) & 3];
42bf3f0a	140	if (!is_present_pte(pte))
7993ba43	141	goto not_present;
1b0973bd AK	142	--walker->level;
	143	}
	144	#endif
a9058ecd	145	ASSERT((!is_long_mode(vcpu) && is_pae(vcpu)) \|\|
24993d53	146	(vcpu->arch.cr3 & CR3_NONPAE_RESERVED_BITS) == 0);
6aa8b732	147
fe135d2c	148	pt_access = ACC_ALL;
ac79c978 AK	149
ac79c978 AK	150	for (;;) {
42bf3f0a	151	index = PT_INDEX(addr, walker->level);
ac79c978	152
5fb07ddb	153	table_gfn = gpte_to_gfn(pte);
1755fbcc	154	pte_gpa = gfn_to_gpa(table_gfn);
ec8d4eae	155	pte_gpa += index * sizeof(pt_element_t);
42bf3f0a	156	walker->table_gfn[walker->level - 1] = table_gfn;
7819026e	157	walker->pte_gpa[walker->level - 1] = pte_gpa;
b8688d51	158	pgprintk("%s: table_gfn[%d] %lx\n", __func__,
42bf3f0a AK	159	walker->level - 1, table_gfn);
42bf3f0a AK	160
ec8d4eae	161	kvm_read_guest(vcpu->kvm, pte_gpa, &pte, sizeof(pte));
42bf3f0a AK	162
42bf3f0a AK	163	if (!is_present_pte(pte))
7993ba43 AK	164	goto not_present;
7993ba43 AK	165
42bf3f0a	166	if (write_fault && !is_writeble_pte(pte))
7993ba43 AK	167	if (user_fault \|\| is_write_protection(vcpu))
	168	goto access_error;
	169
42bf3f0a	170	if (user_fault && !(pte & PT_USER_MASK))
7993ba43 AK	171	goto access_error;
7993ba43 AK	172
73b1087e	173	#if PTTYPE == 64
42bf3f0a	174	if (fetch_fault && is_nx(vcpu) && (pte & PT64_NX_MASK))
73b1087e AK	175	goto access_error;
	176	#endif
	177
42bf3f0a	178	if (!(pte & PT_ACCESSED_MASK)) {
bf3f8e86	179	mark_page_dirty(vcpu->kvm, table_gfn);
b3e4e63f MT	180	if (FNAME(cmpxchg_gpte)(vcpu->kvm, table_gfn,
	181	index, pte, pte\|PT_ACCESSED_MASK))
	182	goto walk;
42bf3f0a	183	pte \|= PT_ACCESSED_MASK;
bf3f8e86	184	}
815af8d4	185
bedbe4ee	186	pte_access = pt_access & FNAME(gpte_access)(vcpu, pte);
fe135d2c	187
7819026e MT	188	walker->ptes[walker->level - 1] = pte;
7819026e MT	189
815af8d4	190	if (walker->level == PT_PAGE_TABLE_LEVEL) {
5fb07ddb	191	walker->gfn = gpte_to_gfn(pte);
815af8d4 AK	192	break;
	193	}
	194
	195	if (walker->level == PT_DIRECTORY_LEVEL
42bf3f0a	196	&& (pte & PT_PAGE_SIZE_MASK)
815af8d4	197	&& (PTTYPE == 64 \|\| is_pse(vcpu))) {
5fb07ddb	198	walker->gfn = gpte_to_gfn_pde(pte);
815af8d4	199	walker->gfn += PT_INDEX(addr, PT_PAGE_TABLE_LEVEL);
da928521 AK	200	if (PTTYPE == 32 && is_cpuid_PSE36())
da928521 AK	201	walker->gfn += pse36_gfn_delta(pte);
ac79c978	202	break;
815af8d4	203	}
ac79c978	204
fe135d2c	205	pt_access = pte_access;
ac79c978 AK	206	--walker->level;
ac79c978 AK	207	}
42bf3f0a AK	208
42bf3f0a AK	209	if (write_fault && !is_dirty_pte(pte)) {
b3e4e63f MT	210	bool ret;
b3e4e63f MT	211
42bf3f0a	212	mark_page_dirty(vcpu->kvm, table_gfn);
b3e4e63f MT	213	ret = FNAME(cmpxchg_gpte)(vcpu->kvm, table_gfn, index, pte,
	214	pte\|PT_DIRTY_MASK);
	215	if (ret)
	216	goto walk;
42bf3f0a	217	pte \|= PT_DIRTY_MASK;
42bf3f0a	218	kvm_mmu_pte_write(vcpu, pte_gpa, (u8 *)&pte, sizeof(pte));
7819026e	219	walker->ptes[walker->level - 1] = pte;
42bf3f0a AK	220	}
42bf3f0a AK	221
fe135d2c AK	222	walker->pt_access = pt_access;
	223	walker->pte_access = pte_access;
	224	pgprintk("%s: pte %llx pte_access %x pt_access %x\n",
b8688d51	225	__func__, (u64)pte, pt_access, pte_access);
7993ba43 AK	226	return 1;
	227
	228	not_present:
	229	walker->error_code = 0;
	230	goto err;
	231
	232	access_error:
	233	walker->error_code = PFERR_PRESENT_MASK;
	234
	235	err:
	236	if (write_fault)
	237	walker->error_code \|= PFERR_WRITE_MASK;
	238	if (user_fault)
	239	walker->error_code \|= PFERR_USER_MASK;
73b1087e AK	240	if (fetch_fault)
73b1087e AK	241	walker->error_code \|= PFERR_FETCH_MASK;
fe551881	242	return 0;
6aa8b732 AK	243	}
6aa8b732 AK	244
0028425f	245	static void FNAME(update_pte)(struct kvm_vcpu vcpu, struct kvm_mmu_page page,
489f1d65	246	u64 spte, const void pte)
0028425f AK	247	{
0028425f AK	248	pt_element_t gpte;
41074d07	249	unsigned pte_access;
35149e21	250	pfn_t pfn;
05da4558	251	int largepage = vcpu->arch.update_pte.largepage;
0028425f	252
0028425f	253	gpte = (const pt_element_t )pte;
c7addb90	254	if (~gpte & (PT_PRESENT_MASK \| PT_ACCESSED_MASK)) {
489f1d65	255	if (!is_present_pte(gpte))
c7addb90 AK	256	set_shadow_pte(spte, shadow_notrap_nonpresent_pte);
	257	return;
	258	}
b8688d51	259	pgprintk("%s: gpte %llx spte %p\n", __func__, (u64)gpte, spte);
41074d07	260	pte_access = page->role.access & FNAME(gpte_access)(vcpu, gpte);
d7824fff AK	261	if (gpte_to_gfn(gpte) != vcpu->arch.update_pte.gfn)
d7824fff AK	262	return;
35149e21 AL	263	pfn = vcpu->arch.update_pte.pfn;
35149e21 AL	264	if (is_error_pfn(pfn))
d7824fff	265	return;
35149e21	266	kvm_get_pfn(pfn);
1c4f1fd6	267	mmu_set_spte(vcpu, spte, page->role.access, pte_access, 0, 0,
05da4558	268	gpte & PT_DIRTY_MASK, NULL, largepage, gpte_to_gfn(gpte),
35149e21	269	pfn, true);
0028425f AK	270	}
0028425f AK	271
6aa8b732 AK	272	/*
	273	* Fetch a shadow pte for a specific level in the paging hierarchy.
	274	*/
	275	static u64 FNAME(fetch)(struct kvm_vcpu vcpu, gva_t addr,
97a0a01e	276	struct guest_walker *walker,
05da4558	277	int user_fault, int write_fault, int largepage,
35149e21	278	int *ptwrite, pfn_t pfn)
6aa8b732 AK	279	{
	280	hpa_t shadow_addr;
	281	int level;
ef0197e8	282	u64 *shadow_ent;
fe135d2c	283	unsigned access = walker->pt_access;
ac79c978	284
7819026e	285	if (!is_present_pte(walker->ptes[walker->level - 1]))
ac79c978	286	return NULL;
6aa8b732	287
ad312c7c ZX	288	shadow_addr = vcpu->arch.mmu.root_hpa;
ad312c7c ZX	289	level = vcpu->arch.mmu.shadow_root_level;
aef3d3fe	290	if (level == PT32E_ROOT_LEVEL) {
ad312c7c	291	shadow_addr = vcpu->arch.mmu.pae_root[(addr >> 30) & 3];
aef3d3fe AK	292	shadow_addr &= PT64_BASE_ADDR_MASK;
	293	--level;
	294	}
6aa8b732 AK	295
	296	for (; ; level--) {
	297	u32 index = SHADOW_PT_INDEX(addr, level);
25c0de2c	298	struct kvm_mmu_page *shadow_page;
8c7bb723	299	u64 shadow_pte;
cea0f0e7 AK	300	int metaphysical;
cea0f0e7 AK	301	gfn_t table_gfn;
6aa8b732	302
ef0197e8	303	shadow_ent = ((u64 *)__va(shadow_addr)) + index;
5882842f DE	304	if (level == PT_PAGE_TABLE_LEVEL)
5882842f DE	305	break;
05da4558 MT	306
	307	if (largepage && level == PT_DIRECTORY_LEVEL)
	308	break;
	309
	310	if (is_shadow_present_pte(*shadow_ent)
	311	&& !is_large_pte(*shadow_ent)) {
6aa8b732	312	shadow_addr = *shadow_ent & PT64_BASE_ADDR_MASK;
6aa8b732 AK	313	continue;
	314	}
	315
05da4558 MT	316	if (is_large_pte(*shadow_ent))
	317	rmap_remove(vcpu->kvm, shadow_ent);
	318
cea0f0e7 AK	319	if (level - 1 == PT_PAGE_TABLE_LEVEL
	320	&& walker->level == PT_DIRECTORY_LEVEL) {
	321	metaphysical = 1;
7819026e	322	if (!is_dirty_pte(walker->ptes[level - 1]))
fe135d2c	323	access &= ~ACC_WRITE_MASK;
7819026e	324	table_gfn = gpte_to_gfn(walker->ptes[level - 1]);
cea0f0e7 AK	325	} else {
	326	metaphysical = 0;
	327	table_gfn = walker->table_gfn[level - 2];
	328	}
	329	shadow_page = kvm_mmu_get_page(vcpu, table_gfn, addr, level-1,
fe135d2c	330	metaphysical, access,
f7d9c7b7 AK	331	shadow_ent);
f7d9c7b7 AK	332	if (!metaphysical) {
7ec54588	333	int r;
7819026e	334	pt_element_t curr_pte;
7ec54588 MT	335	r = kvm_read_guest_atomic(vcpu->kvm,
	336	walker->pte_gpa[level - 2],
	337	&curr_pte, sizeof(curr_pte));
d7824fff	338	if (r \|\| curr_pte != walker->ptes[level - 2]) {
35149e21	339	kvm_release_pfn_clean(pfn);
7819026e	340	return NULL;
d7824fff	341	}
7819026e	342	}
47ad8e68	343	shadow_addr = __pa(shadow_page->spt);
aef3d3fe AK	344	shadow_pte = shadow_addr \| PT_PRESENT_MASK \| PT_ACCESSED_MASK
aef3d3fe AK	345	\| PT_WRITABLE_MASK \| PT_USER_MASK;
8c7bb723	346	*shadow_ent = shadow_pte;
6aa8b732	347	}
ef0197e8	348
1c4f1fd6	349	mmu_set_spte(vcpu, shadow_ent, access, walker->pte_access & access,
7819026e MT	350	user_fault, write_fault,
7819026e MT	351	walker->ptes[walker->level-1] & PT_DIRTY_MASK,
35149e21	352	ptwrite, largepage, walker->gfn, pfn, false);
050e6499	353
ef0197e8	354	return shadow_ent;
6aa8b732 AK	355	}
6aa8b732 AK	356
6aa8b732 AK	357	/*
	358	* Page fault handler. There are several causes for a page fault:
	359	* - there is no shadow pte for the guest pte
	360	* - write access through a shadow pte marked read only so that we can set
	361	* the dirty bit
	362	* - write access to a shadow pte marked read only so we can update the page
	363	* dirty bitmap, when userspace requests it
	364	* - mmio access; in this case we will never install a present shadow pte
	365	* - normal guest page fault due to the guest pte marked not present, not
	366	* writable, or not executable
	367	*
e2dec939 AK	368	* Returns: 1 if we need to emulate the instruction, 0 otherwise, or
e2dec939 AK	369	* a negative value on error.
6aa8b732 AK	370	*/
	371	static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
	372	u32 error_code)
	373	{
	374	int write_fault = error_code & PFERR_WRITE_MASK;
6aa8b732	375	int user_fault = error_code & PFERR_USER_MASK;
73b1087e	376	int fetch_fault = error_code & PFERR_FETCH_MASK;
6aa8b732 AK	377	struct guest_walker walker;
6aa8b732 AK	378	u64 *shadow_pte;
cea0f0e7	379	int write_pt = 0;
e2dec939	380	int r;
35149e21	381	pfn_t pfn;
05da4558	382	int largepage = 0;
6aa8b732	383
b8688d51	384	pgprintk("%s: addr %lx err %x\n", __func__, addr, error_code);
37a7d8b0	385	kvm_mmu_audit(vcpu, "pre page fault");
714b93da	386
e2dec939 AK	387	r = mmu_topup_memory_caches(vcpu);
	388	if (r)
	389	return r;
714b93da	390
6aa8b732 AK	391	/*
	392	* Look up the shadow pte for the faulting address.
	393	*/
73b1087e AK	394	r = FNAME(walk_addr)(&walker, vcpu, addr, write_fault, user_fault,
73b1087e AK	395	fetch_fault);
6aa8b732 AK	396
	397	/*
	398	* The page is not mapped by the guest. Let the guest handle it.
	399	*/
7993ba43	400	if (!r) {
b8688d51	401	pgprintk("%s: guest page fault\n", __func__);
7993ba43	402	inject_page_fault(vcpu, addr, walker.error_code);
ad312c7c	403	vcpu->arch.last_pt_write_count = 0; /* reset fork detector */
6aa8b732 AK	404	return 0;
	405	}
	406
72dc67a6	407	down_read(&current->mm->mmap_sem);
05da4558 MT	408	if (walker.level == PT_DIRECTORY_LEVEL) {
	409	gfn_t large_gfn;
	410	large_gfn = walker.gfn & ~(KVM_PAGES_PER_HPAGE-1);
	411	if (is_largepage_backed(vcpu, large_gfn)) {
	412	walker.gfn = large_gfn;
	413	largepage = 1;
	414	}
	415	}
35149e21	416	pfn = gfn_to_pfn(vcpu->kvm, walker.gfn);
72dc67a6	417	up_read(&current->mm->mmap_sem);
d7824fff	418
d196e343	419	/* mmio */
35149e21	420	if (is_error_pfn(pfn)) {
d196e343	421	pgprintk("gfn %x is mmio\n", walker.gfn);
35149e21	422	kvm_release_pfn_clean(pfn);
d196e343 AK	423	return 1;
	424	}
	425
aaee2c94	426	spin_lock(&vcpu->kvm->mmu_lock);
eb787d10	427	kvm_mmu_free_some_pages(vcpu);
97a0a01e	428	shadow_pte = FNAME(fetch)(vcpu, addr, &walker, user_fault, write_fault,
35149e21	429	largepage, &write_pt, pfn);
05da4558	430
b8688d51	431	pgprintk("%s: shadow pte %p %llx ptwrite %d\n", __func__,
97a0a01e	432	shadow_pte, *shadow_pte, write_pt);
cea0f0e7	433
a25f7e1f	434	if (!write_pt)
ad312c7c	435	vcpu->arch.last_pt_write_count = 0; /* reset fork detector */
a25f7e1f	436
1165f5fe	437	++vcpu->stat.pf_fixed;
37a7d8b0	438	kvm_mmu_audit(vcpu, "post page fault (fixed)");
aaee2c94	439	spin_unlock(&vcpu->kvm->mmu_lock);
6aa8b732	440
cea0f0e7	441	return write_pt;
6aa8b732 AK	442	}
	443
	444	static gpa_t FNAME(gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t vaddr)
	445	{
	446	struct guest_walker walker;
e119d117 AK	447	gpa_t gpa = UNMAPPED_GVA;
e119d117 AK	448	int r;
6aa8b732	449
e119d117	450	r = FNAME(walk_addr)(&walker, vcpu, vaddr, 0, 0, 0);
6aa8b732	451
e119d117	452	if (r) {
1755fbcc	453	gpa = gfn_to_gpa(walker.gfn);
e119d117	454	gpa \|= vaddr & ~PAGE_MASK;
6aa8b732 AK	455	}
	456
	457	return gpa;
	458	}
	459
c7addb90 AK	460	static void FNAME(prefetch_page)(struct kvm_vcpu *vcpu,
	461	struct kvm_mmu_page *sp)
	462	{
7ec54588 MT	463	int i, offset = 0, r = 0;
7ec54588 MT	464	pt_element_t pt;
c7addb90	465
e5a4c8ca AK	466	if (sp->role.metaphysical
e5a4c8ca AK	467	\|\| (PTTYPE == 32 && sp->role.level > PT_PAGE_TABLE_LEVEL)) {
c7addb90 AK	468	nonpaging_prefetch_page(vcpu, sp);
	469	return;
	470	}
	471
e5a4c8ca AK	472	if (PTTYPE == 32)
e5a4c8ca AK	473	offset = sp->role.quadrant << PT64_LEVEL_BITS;
7ec54588 MT	474
	475	for (i = 0; i < PT64_ENT_PER_PAGE; ++i) {
	476	gpa_t pte_gpa = gfn_to_gpa(sp->gfn);
	477	pte_gpa += (i+offset) * sizeof(pt_element_t);
	478
	479	r = kvm_read_guest_atomic(vcpu->kvm, pte_gpa, &pt,
	480	sizeof(pt_element_t));
	481	if (r \|\| is_present_pte(pt))
c7addb90 AK	482	sp->spt[i] = shadow_trap_nonpresent_pte;
	483	else
	484	sp->spt[i] = shadow_notrap_nonpresent_pte;
7ec54588	485	}
c7addb90 AK	486	}
c7addb90 AK	487
6aa8b732 AK	488	#undef pt_element_t
	489	#undef guest_walker
	490	#undef FNAME
	491	#undef PT_BASE_ADDR_MASK
	492	#undef PT_INDEX
	493	#undef SHADOW_PT_INDEX
	494	#undef PT_LEVEL_MASK
6aa8b732	495	#undef PT_DIR_BASE_ADDR_MASK
c7addb90	496	#undef PT_LEVEL_BITS
cea0f0e7	497	#undef PT_MAX_FULL_LEVELS
5fb07ddb AK	498	#undef gpte_to_gfn
5fb07ddb AK	499	#undef gpte_to_gfn_pde
b3e4e63f	500	#undef CMPXCHG