vmi_32.c

linux 内核源代码

/*
 * VMI specific paravirt-ops implementation
 *
 * Copyright (C) 2005, VMware, Inc.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
 * NON INFRINGEMENT.  See the GNU General Public License for more
 * details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 *
 * Send feedback to [e-mail]
 *
 */

#include <linux/module.h>
#include <linux/cpu.h>
#include <linux/bootmem.h>
#include <linux/mm.h>
#include <linux/highmem.h>
#include <linux/sched.h>
#include <asm/vmi.h>
#include <asm/io.h>
#include <asm/fixmap.h>
#include <asm/apicdef.h>
#include <asm/apic.h>
#include <asm/processor.h>
#include <asm/timer.h>
#include <asm/vmi_time.h>
#include <asm/kmap_types.h>

/* Convenient for calling VMI functions indirectly in the ROM */
typedef u32 __attribute__((regparm(1))) (VROMFUNC)(void);
typedef u64 __attribute__((regparm(2))) (VROMLONGFUNC)(int);

#define call_vrom_func(rom,func) \
   (((VROMFUNC *)(rom->func))())

#define call_vrom_long_func(rom,func,arg) \
   (((VROMLONGFUNC *)(rom->func)) (arg))

static struct vrom_header *vmi_rom;
static int disable_pge;
static int disable_pse;
static int disable_sep;
static int disable_tsc;
static int disable_mtrr;
static int disable_noidle;
static int disable_vmi_timer;

/* Cached VMI operations */
static struct {
	void (*cpuid)(void /* non-c */);
	void (*_set_ldt)(u32 selector);
	void (*set_tr)(u32 selector);
	void (*set_kernel_stack)(u32 selector, u32 esp0);
	void (*allocate_page)(u32, u32, u32, u32, u32);
	void (*release_page)(u32, u32);
	void (*set_pte)(pte_t, pte_t *, unsigned);
	void (*update_pte)(pte_t *, unsigned);
	void (*set_linear_mapping)(int, void *, u32, u32);
	void (*_flush_tlb)(int);
	void (*set_initial_ap_state)(int, int);
	void (*halt)(void);
  	void (*set_lazy_mode)(int mode);
} vmi_ops;

/* Cached VMI operations */
struct vmi_timer_ops vmi_timer_ops;

/*
 * VMI patching routines.
 */
#define MNEM_CALL 0xe8
#define MNEM_JMP  0xe9
#define MNEM_RET  0xc3

#define IRQ_PATCH_INT_MASK 0
#define IRQ_PATCH_DISABLE  5

static inline void patch_offset(void *insnbuf,
				unsigned long eip, unsigned long dest)
{
        *(unsigned long *)(insnbuf+1) = dest-eip-5;
}

static unsigned patch_internal(int call, unsigned len, void *insnbuf,
			       unsigned long eip)
{
	u64 reloc;
	struct vmi_relocation_info *const rel = (struct vmi_relocation_info *)&reloc;
	reloc = call_vrom_long_func(vmi_rom, get_reloc,	call);
	switch(rel->type) {
		case VMI_RELOCATION_CALL_REL:
			BUG_ON(len < 5);
			*(char *)insnbuf = MNEM_CALL;
			patch_offset(insnbuf, eip, (unsigned long)rel->eip);
			return 5;

		case VMI_RELOCATION_JUMP_REL:
			BUG_ON(len < 5);
			*(char *)insnbuf = MNEM_JMP;
			patch_offset(insnbuf, eip, (unsigned long)rel->eip);
			return 5;

		case VMI_RELOCATION_NOP:
			/* obliterate the whole thing */
			return 0;

		case VMI_RELOCATION_NONE:
			/* leave native code in place */
			break;

		default:
			BUG();
	}
	return len;
}

/*
 * Apply patch if appropriate, return length of new instruction
 * sequence.  The callee does nop padding for us.
 */
static unsigned vmi_patch(u8 type, u16 clobbers, void *insns,
			  unsigned long eip, unsigned len)
{
	switch (type) {
		case PARAVIRT_PATCH(pv_irq_ops.irq_disable):
			return patch_internal(VMI_CALL_DisableInterrupts, len,
					      insns, eip);
		case PARAVIRT_PATCH(pv_irq_ops.irq_enable):
			return patch_internal(VMI_CALL_EnableInterrupts, len,
					      insns, eip);
		case PARAVIRT_PATCH(pv_irq_ops.restore_fl):
			return patch_internal(VMI_CALL_SetInterruptMask, len,
					      insns, eip);
		case PARAVIRT_PATCH(pv_irq_ops.save_fl):
			return patch_internal(VMI_CALL_GetInterruptMask, len,
					      insns, eip);
		case PARAVIRT_PATCH(pv_cpu_ops.iret):
			return patch_internal(VMI_CALL_IRET, len, insns, eip);
		case PARAVIRT_PATCH(pv_cpu_ops.irq_enable_sysexit):
			return patch_internal(VMI_CALL_SYSEXIT, len, insns, eip);
		default:
			break;
	}
	return len;
}

/* CPUID has non-C semantics, and paravirt-ops API doesn't match hardware ISA */
static void vmi_cpuid(unsigned int *eax, unsigned int *ebx,
                               unsigned int *ecx, unsigned int *edx)
{
	int override = 0;
	if (*eax == 1)
		override = 1;
        asm volatile ("call *%6"
                      : "=a" (*eax),
                        "=b" (*ebx),
                        "=c" (*ecx),
                        "=d" (*edx)
                      : "0" (*eax), "2" (*ecx), "r" (vmi_ops.cpuid));
	if (override) {
		if (disable_pse)
			*edx &= ~X86_FEATURE_PSE;
		if (disable_pge)
			*edx &= ~X86_FEATURE_PGE;
		if (disable_sep)
			*edx &= ~X86_FEATURE_SEP;
		if (disable_tsc)
			*edx &= ~X86_FEATURE_TSC;
		if (disable_mtrr)
			*edx &= ~X86_FEATURE_MTRR;
	}
}

static inline void vmi_maybe_load_tls(struct desc_struct *gdt, int nr, struct desc_struct *new)
{
	if (gdt[nr].a != new->a || gdt[nr].b != new->b)
		write_gdt_entry(gdt, nr, new->a, new->b);
}

static void vmi_load_tls(struct thread_struct *t, unsigned int cpu)
{
	struct desc_struct *gdt = get_cpu_gdt_table(cpu);
	vmi_maybe_load_tls(gdt, GDT_ENTRY_TLS_MIN + 0, &t->tls_array[0]);
	vmi_maybe_load_tls(gdt, GDT_ENTRY_TLS_MIN + 1, &t->tls_array[1]);
	vmi_maybe_load_tls(gdt, GDT_ENTRY_TLS_MIN + 2, &t->tls_array[2]);
}

static void vmi_set_ldt(const void *addr, unsigned entries)
{
	unsigned cpu = smp_processor_id();
	u32 low, high;

	pack_descriptor(&low, &high, (unsigned long)addr,
			entries * sizeof(struct desc_struct) - 1,
			DESCTYPE_LDT, 0);
	write_gdt_entry(get_cpu_gdt_table(cpu), GDT_ENTRY_LDT, low, high);
	vmi_ops._set_ldt(entries ? GDT_ENTRY_LDT*sizeof(struct desc_struct) : 0);
}

static void vmi_set_tr(void)
{
	vmi_ops.set_tr(GDT_ENTRY_TSS*sizeof(struct desc_struct));
}

static void vmi_load_esp0(struct tss_struct *tss,
				   struct thread_struct *thread)
{
	tss->x86_tss.esp0 = thread->esp0;

	/* This can only happen when SEP is enabled, no need to test "SEP"arately */
	if (unlikely(tss->x86_tss.ss1 != thread->sysenter_cs)) {
		tss->x86_tss.ss1 = thread->sysenter_cs;
		wrmsr(MSR_IA32_SYSENTER_CS, thread->sysenter_cs, 0);
	}
	vmi_ops.set_kernel_stack(__KERNEL_DS, tss->x86_tss.esp0);
}

static void vmi_flush_tlb_user(void)
{
	vmi_ops._flush_tlb(VMI_FLUSH_TLB);
}

static void vmi_flush_tlb_kernel(void)
{
	vmi_ops._flush_tlb(VMI_FLUSH_TLB | VMI_FLUSH_GLOBAL);
}

/* Stub to do nothing at all; used for delays and unimplemented calls */
static void vmi_nop(void)
{
}

#ifdef CONFIG_DEBUG_PAGE_TYPE

#ifdef CONFIG_X86_PAE
#define MAX_BOOT_PTS (2048+4+1)
#else
#define MAX_BOOT_PTS (1024+1)
#endif

/*
 * During boot, mem_map is not yet available in paging_init, so stash
 * all the boot page allocations here.
 */
static struct {
	u32 pfn;
	int type;
} boot_page_allocations[MAX_BOOT_PTS];
static int num_boot_page_allocations;
static int boot_allocations_applied;

void vmi_apply_boot_page_allocations(void)
{
	int i;
	BUG_ON(!mem_map);
	for (i = 0; i < num_boot_page_allocations; i++) {
		struct page *page = pfn_to_page(boot_page_allocations[i].pfn);
		page->type = boot_page_allocations[i].type;
		page->type = boot_page_allocations[i].type &
				~(VMI_PAGE_ZEROED | VMI_PAGE_CLONE);
	}
	boot_allocations_applied = 1;
}

static void record_page_type(u32 pfn, int type)
{
	BUG_ON(num_boot_page_allocations >= MAX_BOOT_PTS);
	boot_page_allocations[num_boot_page_allocations].pfn = pfn;
	boot_page_allocations[num_boot_page_allocations].type = type;
	num_boot_page_allocations++;
}

static void check_zeroed_page(u32 pfn, int type, struct page *page)
{
	u32 *ptr;
	int i;
	int limit = PAGE_SIZE / sizeof(int);

	if (page_address(page))
		ptr = (u32 *)page_address(page);
	else
		ptr = (u32 *)__va(pfn << PAGE_SHIFT);
	/*
	 * When cloning the root in non-PAE mode, only the userspace
	 * pdes need to be zeroed.
	 */
	if (type & VMI_PAGE_CLONE)
		limit = USER_PTRS_PER_PGD;
	for (i = 0; i < limit; i++)
		BUG_ON(ptr[i]);
}

/*
 * We stash the page type into struct page so we can verify the page
 * types are used properly.
 */
static void vmi_set_page_type(u32 pfn, int type)
{
	/* PAE can have multiple roots per page - don't track */
	if (PTRS_PER_PMD > 1 && (type & VMI_PAGE_PDP))
		return;

	if (boot_allocations_applied) {
		struct page *page = pfn_to_page(pfn);
		if (type != VMI_PAGE_NORMAL)
			BUG_ON(page->type);
		else
			BUG_ON(page->type == VMI_PAGE_NORMAL);
		page->type = type & ~(VMI_PAGE_ZEROED | VMI_PAGE_CLONE);
		if (type & VMI_PAGE_ZEROED)
			check_zeroed_page(pfn, type, page);
	} else {
		record_page_type(pfn, type);
	}
}

static void vmi_check_page_type(u32 pfn, int type)
{
	/* PAE can have multiple roots per page - skip checks */
	if (PTRS_PER_PMD > 1 && (type & VMI_PAGE_PDP))
		return;

	type &= ~(VMI_PAGE_ZEROED | VMI_PAGE_CLONE);
	if (boot_allocations_applied) {
		struct page *page = pfn_to_page(pfn);
		BUG_ON((page->type ^ type) & VMI_PAGE_PAE);
		BUG_ON(type == VMI_PAGE_NORMAL && page->type);
		BUG_ON((type & page->type) == 0);
	}
}
#else
#define vmi_set_page_type(p,t) do { } while (0)
#define vmi_check_page_type(p,t) do { } while (0)
#endif

#ifdef CONFIG_HIGHPTE
static void *vmi_kmap_atomic_pte(struct page *page, enum km_type type)
{
	void *va = kmap_atomic(page, type);

	/*
	 * Internally, the VMI ROM must map virtual addresses to physical
	 * addresses for processing MMU updates.  By the time MMU updates
	 * are issued, this information is typically already lost.
	 * Fortunately, the VMI provides a cache of mapping slots for active
	 * page tables.
	 *
	 * We use slot zero for the linear mapping of physical memory, and
	 * in HIGHPTE kernels, slot 1 and 2 for KM_PTE0 and KM_PTE1.
	 *
	 *  args:                 SLOT                 VA    COUNT PFN
	 */
	BUG_ON(type != KM_PTE0 && type != KM_PTE1);
	vmi_ops.set_linear_mapping((type - KM_PTE0)+1, va, 1, page_to_pfn(page));

	return va;
}
#endif

static void vmi_allocate_pt(struct mm_struct *mm, u32 pfn)
{
	vmi_set_page_type(pfn, VMI_PAGE_L1);
	vmi_ops.allocate_page(pfn, VMI_PAGE_L1, 0, 0, 0);
}

static void vmi_allocate_pd(u32 pfn)
{
 	/*
	 * This call comes in very early, before mem_map is setup.
	 * It is called only for swapper_pg_dir, which already has
	 * data on it.
	 */
 	vmi_set_page_type(pfn, VMI_PAGE_L2);
	vmi_ops.allocate_page(pfn, VMI_PAGE_L2, 0, 0, 0);
}

static void vmi_allocate_pd_clone(u32 pfn, u32 clonepfn, u32 start, u32 count)
{
 	vmi_set_page_type(pfn, VMI_PAGE_L2 | VMI_PAGE_CLONE);
	vmi_check_page_type(clonepfn, VMI_PAGE_L2);
	vmi_ops.allocate_page(pfn, VMI_PAGE_L2 | VMI_PAGE_CLONE, clonepfn, start, count);
}

static void vmi_release_pt(u32 pfn)
{
	vmi_ops.release_page(pfn, VMI_PAGE_L1);
	vmi_set_page_type(pfn, VMI_PAGE_NORMAL);
}

static void vmi_release_pd(u32 pfn)
{
	vmi_ops.release_page(pfn, VMI_PAGE_L2);
	vmi_set_page_type(pfn, VMI_PAGE_NORMAL);
}

/*
 * Helper macros for MMU update flags.  We can defer updates until a flush
 * or page invalidation only if the update is to the current address space
 * (otherwise, there is no flush).  We must check against init_mm, since
 * this could be a kernel update, which usually passes init_mm, although
 * sometimes this check can be skipped if we know the particular function
 * is only called on user mode PTEs.  We could change the kernel to pass
 * current->active_mm here, but in particular, I was unsure if changing
 * mm/highmem.c to do this would still be correct on other architectures.
 */
#define is_current_as(mm, mustbeuser) ((mm) == current->active_mm ||    \
                                       (!mustbeuser && (mm) == &init_mm))
#define vmi_flags_addr(mm, addr, level, user)                           \
        ((level) | (is_current_as(mm, user) ?                           \
                (VMI_PAGE_CURRENT_AS | ((addr) & VMI_PAGE_VA_MASK)) : 0))
#define vmi_flags_addr_defer(mm, addr, level, user)                     \
        ((level) | (is_current_as(mm, user) ?                           \
                (VMI_PAGE_DEFER | VMI_PAGE_CURRENT_AS | ((addr) & VMI_PAGE_VA_MASK)) : 0))

static void vmi_update_pte(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
{
	vmi_check_page_type(__pa(ptep) >> PAGE_SHIFT, VMI_PAGE_PTE);
	vmi_ops.update_pte(ptep, vmi_flags_addr(mm, addr, VMI_PAGE_PT, 0));
}

static void vmi_update_pte_defer(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
{
	vmi_check_page_type(__pa(ptep) >> PAGE_SHIFT, VMI_PAGE_PTE);
	vmi_ops.update_pte(ptep, vmi_flags_addr_defer(mm, addr, VMI_PAGE_PT, 0));
}

static void vmi_set_pte(pte_t *ptep, pte_t pte)
{
	/* XXX because of set_pmd_pte, this can be called on PT or PD layers */
	vmi_check_page_type(__pa(ptep) >> PAGE_SHIFT, VMI_PAGE_PTE | VMI_PAGE_PD);
	vmi_ops.set_pte(pte, ptep, VMI_PAGE_PT);
}

static void vmi_set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
{
	vmi_check_page_type(__pa(ptep) >> PAGE_SHIFT, VMI_PAGE_PTE);
	vmi_ops.set_pte(pte, ptep, vmi_flags_addr(mm, addr, VMI_PAGE_PT, 0));
}

static void vmi_set_pmd(pmd_t *pmdp, pmd_t pmdval)
{
#ifdef CONFIG_X86_PAE
	const pte_t pte = { pmdval.pmd, pmdval.pmd >> 32 };
	vmi_check_page_type(__pa(pmdp) >> PAGE_SHIFT, VMI_PAGE_PMD);
#else
	const pte_t pte = { pmdval.pud.pgd.pgd };
	vmi_check_page_type(__pa(pmdp) >> PAGE_SHIFT, VMI_PAGE_PGD);
#endif
	vmi_ops.set_pte(pte, (pte_t *)pmdp, VMI_PAGE_PD);
}

#ifdef CONFIG_X86_PAE

static void vmi_set_pte_atomic(pte_t *ptep, pte_t pteval)
{
	/*
	 * XXX This is called from set_pmd_pte, but at both PT
	 * and PD layers so the VMI_PAGE_PT flag is wrong.  But
	 * it is only called for large page mapping changes,
	 * the Xen backend, doesn't support large pages, and the
	 * ESX backend doesn't depend on the flag.
	 */
	set_64bit((unsigned long long *)ptep,pte_val(pteval));
	vmi_ops.update_pte(ptep, VMI_PAGE_PT);
}

static void vmi_set_pte_present(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
{
	vmi_check_page_type(__pa(ptep) >> PAGE_SHIFT, VMI_PAGE_PTE);
	vmi_ops.set_pte(pte, ptep, vmi_flags_addr_defer(mm, addr, VMI_PAGE_PT, 1));
}

static void vmi_set_pud(pud_t *pudp, pud_t pudval)
{
	/* Um, eww */
	const pte_t pte = { pudval.pgd.pgd, pudval.pgd.pgd >> 32 };
	vmi_check_page_type(__pa(pudp) >> PAGE_SHIFT, VMI_PAGE_PGD);
	vmi_ops.set_pte(pte, (pte_t *)pudp, VMI_PAGE_PDP);
}

static void vmi_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
{
	const pte_t pte = { 0 };