C++ set_pte函数代码示例

/*H:480
 * (vi) Mapping the Switcher when the Guest is about to run.
 *
 * The Switcher and the two pages for this CPU need to be visible in the Guest
 * (and not the pages for other CPUs).
 *
 * The pages for the pagetables have all been allocated before: we just need
 * to make sure the actual PTEs are up-to-date for the CPU we're about to run
 * on.
 */
void map_switcher_in_guest(struct lg_cpu *cpu, struct lguest_pages *pages)
{
	unsigned long base;
	struct page *percpu_switcher_page, *regs_page;
	pte_t *pte;
	struct pgdir *pgdir = &cpu->lg->pgdirs[cpu->cpu_pgd];

	/* Switcher page should always be mapped by now! */
	BUG_ON(!pgdir->switcher_mapped);

	/* 
	 * Remember that we have two pages for each Host CPU, so we can run a
	 * Guest on each CPU without them interfering.  We need to make sure
	 * those pages are mapped correctly in the Guest, but since we usually
	 * run on the same CPU, we cache that, and only update the mappings
	 * when we move.
	 */
	if (pgdir->last_host_cpu == raw_smp_processor_id())
		return;

	/* -1 means unknown so we remove everything. */
	if (pgdir->last_host_cpu == -1) {
		unsigned int i;
		for_each_possible_cpu(i)
			remove_switcher_percpu_map(cpu, i);
	} else {
		/* We know exactly what CPU mapping to remove. */
		remove_switcher_percpu_map(cpu, pgdir->last_host_cpu);
	}

	/*
	 * When we're running the Guest, we want the Guest's "regs" page to
	 * appear where the first Switcher page for this CPU is.  This is an
	 * optimization: when the Switcher saves the Guest registers, it saves
	 * them into the first page of this CPU's "struct lguest_pages": if we
	 * make sure the Guest's register page is already mapped there, we
	 * don't have to copy them out again.
	 */
	/* Find the shadow PTE for this regs page. */
	base = switcher_addr + PAGE_SIZE
		+ raw_smp_processor_id() * sizeof(struct lguest_pages);
	pte = find_spte(cpu, base, false, 0, 0);
	regs_page = pfn_to_page(__pa(cpu->regs_page) >> PAGE_SHIFT);
	get_page(regs_page);
	set_pte(pte, mk_pte(regs_page, __pgprot(__PAGE_KERNEL & ~_PAGE_GLOBAL)));

	/*
	 * We map the second page of the struct lguest_pages read-only in
	 * the Guest: the IDT, GDT and other things it's not supposed to
	 * change.
	 */
	pte = find_spte(cpu, base + PAGE_SIZE, false, 0, 0);
	percpu_switcher_page
		= lg_switcher_pages[1 + raw_smp_processor_id()*2 + 1];
	get_page(percpu_switcher_page);
	set_pte(pte, mk_pte(percpu_switcher_page,
			    __pgprot(__PAGE_KERNEL_RO & ~_PAGE_GLOBAL)));

	pgdir->last_host_cpu = raw_smp_processor_id();
}

static inline void copy_one_pte(pte_t * old_pte, pte_t * new_pte, int cow)
{
	pte_t pte = *old_pte;
	unsigned long page_nr;

	if (pte_none(pte))
		return;
	if (!pte_present(pte)) {
		swap_duplicate(pte_val(pte));
		set_pte(new_pte, pte);
		return;
	}
	page_nr = MAP_NR(pte_page(pte));
	if (page_nr >= MAP_NR(high_memory) || PageReserved(mem_map+page_nr)) {
		set_pte(new_pte, pte);
		return;
	}
	if (cow)
		pte = pte_wrprotect(pte);
	if (delete_from_swap_cache(page_nr))
		pte = pte_mkdirty(pte);
	set_pte(new_pte, pte_mkold(pte));
	set_pte(old_pte, pte);
	mem_map[page_nr].count++;
}

static void _copy_pte(pte_t *dst_ptep, pte_t *src_ptep, unsigned long addr)
{
	pte_t pte = READ_ONCE(*src_ptep);

	if (pte_valid(pte)) {
		/*
		 * Resume will overwrite areas that may be marked
		 * read only (code, rodata). Clear the RDONLY bit from
		 * the temporary mappings we use during restore.
		 */
		set_pte(dst_ptep, pte_mkwrite(pte));
	} else if (debug_pagealloc_enabled() && !pte_none(pte)) {
		/*
		 * debug_pagealloc will removed the PTE_VALID bit if
		 * the page isn't in use by the resume kernel. It may have
		 * been in use by the original kernel, in which case we need
		 * to put it back in our copy to do the restore.
		 *
		 * Before marking this entry valid, check the pfn should
		 * be mapped.
		 */
		BUG_ON(!pfn_valid(pte_pfn(pte)));

		set_pte(dst_ptep, pte_mkpresent(pte_mkwrite(pte)));
	}
}

/*
 * Trying to stop swapping from a file is fraught with races, so
 * we repeat quite a bit here when we have to pause. swapoff()
 * isn't exactly timing-critical, so who cares (but this is /really/
 * inefficient, ugh).
 *
 * We return 1 after having slept, which makes the process start over
 * from the beginning for this process..
 */
static inline int unuse_pte(struct vm_area_struct * vma, unsigned long address,
	pte_t *dir, unsigned int type, unsigned long page)
{
	pte_t pte = *dir;

	if (pte_none(pte))
		return 0;
	if (pte_present(pte)) {
		unsigned long page_nr = MAP_NR(pte_page(pte));
		if (page_nr >= MAP_NR(high_memory))
			return 0;
		if (!in_swap_cache(page_nr))
			return 0;
		if (SWP_TYPE(in_swap_cache(page_nr)) != type)
			return 0;
		delete_from_swap_cache(page_nr);
		set_pte(dir, pte_mkdirty(pte));
		return 0;
	}
	if (SWP_TYPE(pte_val(pte)) != type)
		return 0;
	read_swap_page(pte_val(pte), (char *) page);
#if 0 /* Is this really needed here, hasn't it been solved elsewhere? */
	flush_page_to_ram(page);
#endif
	if (pte_val(*dir) != pte_val(pte)) {
		free_page(page);
		return 1;
	}
	set_pte(dir, pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot))));
	flush_tlb_page(vma, address);
	++vma->vm_mm->rss;
	swap_free(pte_val(pte));
	return 1;
}

/*H:520
 * Setting up the Switcher PTE page for given CPU is fairly easy, given
 * the CPU number and the "struct page"s for the Switcher code itself.
 *
 * Currently the Switcher is less than a page long, so "pages" is always 1.
 */
static __init void populate_switcher_pte_page(unsigned int cpu,
					      struct page *switcher_page[],
					      unsigned int pages)
{
	unsigned int i;
	pte_t *pte = switcher_pte_page(cpu);

	/* The first entries are easy: they map the Switcher code. */
	for (i = 0; i < pages; i++) {
		set_pte(&pte[i], mk_pte(switcher_page[i],
				__pgprot(_PAGE_PRESENT|_PAGE_ACCESSED)));
	}

	/* The only other thing we map is this CPU's pair of pages. */
	i = pages + cpu*2;

	/* First page (Guest registers) is writable from the Guest */
	set_pte(&pte[i], pfn_pte(page_to_pfn(switcher_page[i]),
			 __pgprot(_PAGE_PRESENT|_PAGE_ACCESSED|_PAGE_RW)));

	/*
	 * The second page contains the "struct lguest_ro_state", and is
	 * read-only.
	 */
	set_pte(&pte[i+1], pfn_pte(page_to_pfn(switcher_page[i+1]),
			   __pgprot(_PAGE_PRESENT|_PAGE_ACCESSED)));
}

/*H:481
 * This clears the Switcher mappings for cpu #i.
 */
static void remove_switcher_percpu_map(struct lg_cpu *cpu, unsigned int i)
{
	unsigned long base = switcher_addr + PAGE_SIZE + i * PAGE_SIZE*2;
	pte_t *pte;

	/* Clear the mappings for both pages. */
	pte = find_spte(cpu, base, false, 0, 0);
	release_pte(*pte);
	set_pte(pte, __pte(0));

	pte = find_spte(cpu, base + PAGE_SIZE, false, 0, 0);
	release_pte(*pte);
	set_pte(pte, __pte(0));
}

/**
 * Replace the PFN of a PTE with the address of the actual page.
 *
 * The caller maps a reserved dummy page at the address with the desired access
 * and flags.
 *
 * This hack is required for older Linux kernels which don't provide
 * remap_pfn_range().
 *
 * @returns 0 on success, -ENOMEM on failure.
 * @param   mm          The memory context.
 * @param   ulAddr      The mapping address.
 * @param   Phys        The physical address of the page to map.
 */
static int rtR0MemObjLinuxFixPte(struct mm_struct *mm, unsigned long ulAddr, RTHCPHYS Phys)
{
    int rc = -ENOMEM;
    pgd_t *pgd;

    spin_lock(&mm->page_table_lock);

    pgd = pgd_offset(mm, ulAddr);
    if (!pgd_none(*pgd) && !pgd_bad(*pgd))
    {
        pmd_t *pmd = pmd_offset(pgd, ulAddr);
        if (!pmd_none(*pmd))
        {
            pte_t *ptep = pte_offset_map(pmd, ulAddr);
            if (ptep)
            {
                pte_t pte = *ptep;
                pte.pte_high &= 0xfff00000;
                pte.pte_high |= ((Phys >> 32) & 0x000fffff);
                pte.pte_low  &= 0x00000fff;
                pte.pte_low  |= (Phys & 0xfffff000);
                set_pte(ptep, pte);
                pte_unmap(ptep);
                rc = 0;
            }
        }

/*
 * Note that this is intended to be called only from the copy_user_page
 * asm code; anything else will require special locking to prevent the
 * mini-cache space being re-used.  (Note: probably preempt unsafe).
 *
 * We rely on the fact that the minicache is 2K, and we'll be pushing
 * 4K of data through it, so we don't actually have to specifically
 * flush the minicache when we change the mapping.
 *
 * Note also: assert(PAGE_OFFSET <= virt < high_memory).
 * Unsafe: preempt, kmap.
 */
unsigned long map_page_minicache(unsigned long virt)
{
	set_pte(minicache_pte, mk_pte_phys(__pa(virt), minicache_pgprot));
	cpu_tlb_invalidate_page(minicache_address, 0);

	return minicache_address;
}

static inline void
remove_mapping_pte_range (pmd_t *pmd, unsigned long address, unsigned long size)
{
	pte_t *pte;
	unsigned long end;

	if (pmd_none (*pmd))
		return;
	if (pmd_bad (*pmd)){
		printk ("remove_graphics_pte_range: bad pmd (%08lx)\n", pmd_val (*pmd));
		pmd_clear (pmd);
		return;
	}
	pte = pte_offset (pmd, address);
	address &= ~PMD_MASK;
	end = address + size;
	if (end > PMD_SIZE)
		end = PMD_SIZE;
	do {
		pte_t entry = *pte;
		if (pte_present (entry))
			set_pte (pte, pte_modify (entry, PAGE_NONE));
		address += PAGE_SIZE;
		pte++;
	} while (address < end);
						  
}

static inline void remap_area_pte(pte_t * pte, unsigned long address, unsigned long size,
	unsigned long phys_addr, unsigned long flags)
{
	unsigned long end;
	unsigned long pfn;

	address &= ~PMD_MASK;
	end = address + size;
	if (end > PMD_SIZE)
		end = PMD_SIZE;
	if (address >= end)
		BUG();
	pfn = phys_addr >> PAGE_SHIFT;
	do {
		if (!pte_none(*pte)) {
			printk("remap_area_pte: page already exists\n");
			BUG();
		}
		set_pte(pte, pfn_pte(pfn, __pgprot(_PAGE_PRESENT | _PAGE_RW | 
					_PAGE_DIRTY | _PAGE_ACCESSED | flags)));
		address += PAGE_SIZE;
		pfn++;
		pte++;
	} while (address && (address < end));
}

static int relocate_restore_code(void)
{
	pgd_t *pgd;
	pud_t *pud;

	relocated_restore_code = get_safe_page(GFP_ATOMIC);
	if (!relocated_restore_code)
		return -ENOMEM;

	memcpy((void *)relocated_restore_code, &core_restore_code, PAGE_SIZE);

	/* Make the page containing the relocated code executable */
	pgd = (pgd_t *)__va(read_cr3()) + pgd_index(relocated_restore_code);
	pud = pud_offset(pgd, relocated_restore_code);
	if (pud_large(*pud)) {
		set_pud(pud, __pud(pud_val(*pud) & ~_PAGE_NX));
	} else {
		pmd_t *pmd = pmd_offset(pud, relocated_restore_code);

		if (pmd_large(*pmd)) {
			set_pmd(pmd, __pmd(pmd_val(*pmd) & ~_PAGE_NX));
		} else {
			pte_t *pte = pte_offset_kernel(pmd, relocated_restore_code);

			set_pte(pte, __pte(pte_val(*pte) & ~_PAGE_NX));
		}
	}
	__flush_tlb_all();

	return 0;
}

/*H:501
 * We do need the Switcher code mapped at all times, so we allocate that
 * part of the Guest page table here.  We map the Switcher code immediately,
 * but defer mapping of the guest register page and IDT/LDT etc page until
 * just before we run the guest in map_switcher_in_guest().
 *
 * We *could* do this setup in map_switcher_in_guest(), but at that point
 * we've interrupts disabled, and allocating pages like that is fraught: we
 * can't sleep if we need to free up some memory.
 */
static bool allocate_switcher_mapping(struct lg_cpu *cpu)
{
	int i;

	for (i = 0; i < TOTAL_SWITCHER_PAGES; i++) {
		pte_t *pte = find_spte(cpu, switcher_addr + i * PAGE_SIZE, true,
				       CHECK_GPGD_MASK, _PAGE_TABLE);
		if (!pte)
			return false;

		/*
		 * Map the switcher page if not already there.  It might
		 * already be there because we call allocate_switcher_mapping()
		 * in guest_set_pgd() just in case it did discard our Switcher
		 * mapping, but it probably didn't.
		 */
		if (i == 0 && !(pte_flags(*pte) & _PAGE_PRESENT)) {
			/* Get a reference to the Switcher page. */
			get_page(lg_switcher_pages[0]);
			/* Create a read-only, exectuable, kernel-style PTE */
			set_pte(pte,
				mk_pte(lg_switcher_pages[0], PAGE_KERNEL_RX));
		}
	}
	cpu->lg->pgdirs[cpu->cpu_pgd].switcher_mapped = true;
	return true;
}

/**
 * 建立临时内核映射
 * type和CPU共同确定用哪个固定映射的线性地址映射请求页。
 */
void *kmap_atomic(struct page *page, enum km_type type)
{
	enum fixed_addresses idx;
	unsigned long vaddr;

	/* even !CONFIG_PREEMPT needs this, for in_atomic in do_page_fault */
	inc_preempt_count();
	/**
	 * 如果被映射的页不属于高端内存，当然用不着映射。直接返回线性地址就行了。
	 */
	if (!PageHighMem(page))
		return page_address(page);

	/**
	 * 通过type和CPU确定线性地址。
	 */
	idx = type + KM_TYPE_NR*smp_processor_id();
	vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
#ifdef CONFIG_DEBUG_HIGHMEM
	if (!pte_none(*(kmap_pte-idx)))
		BUG();
#endif
	/**
	 * 将线性地址与页表项建立映射。
	 */
	set_pte(kmap_pte-idx, mk_pte(page, kmap_prot));
	/**
	 * 当然，最后必须刷新一下TLB。然后才能返回线性地址。
	 */
	__flush_tlb_one(vaddr);

	return (void*) vaddr;
}

/*
 * Create a temporary, one-page mapping for a device.
 * This is used by some device probe routines that
 * need to do peek/write/read tricks.
 */
void *
bus_tmapin(int bustype, int pa)
{
	vaddr_t pgva;
	int off, pte;

	if ((bustype < 0) || (bustype >= BUS__NTYPES))
		panic("bus_tmapin: bustype");

	off = pa & PGOFSET;
	pa -= off;

	pa &= bus_info[bustype].mask;
	pa |= bus_info[bustype].base;

	pte = PA_PGNUM(pa);
	pte |= (bus_info[bustype].type << PG_MOD_SHIFT);
	pte |= (PG_VALID | PG_WRITE | PG_SYSTEM | PG_NC);

	if (tmp_vpages_inuse)
		panic("bus_tmapin: tmp_vpages_inuse");
	tmp_vpages_inuse++;

	pgva = tmp_vpages[1];
	set_pte(pgva, pte);

	return ((void *)(pgva + off));
}

/*
 * copy_user_page
 * @to: P1 address
 * @from: P1 address
 * @address: U0 address to be mapped
 * @page: page (virt_to_page(to))
 */
void copy_user_page(void *to, void *from, unsigned long address, 
		    struct page *page)
{
	__set_bit(PG_mapped, &page->flags);
	if (((address ^ (unsigned long)to) & CACHE_ALIAS) == 0)
		copy_page(to, from);
	else {
		pgprot_t pgprot = __pgprot(_PAGE_PRESENT | 
					   _PAGE_RW | _PAGE_CACHABLE |
					   _PAGE_DIRTY | _PAGE_ACCESSED | 
					   _PAGE_HW_SHARED | _PAGE_FLAGS_HARD);
		unsigned long phys_addr = PHYSADDR(to);
		unsigned long p3_addr = P3SEG + (address & CACHE_ALIAS);
		pgd_t *dir = pgd_offset_k(p3_addr);
		pmd_t *pmd = pmd_offset(dir, p3_addr);
		pte_t *pte = pte_offset_kernel(pmd, p3_addr);
		pte_t entry;
		unsigned long flags;

		entry = pfn_pte(phys_addr >> PAGE_SHIFT, pgprot);
		down(&p3map_sem[(address & CACHE_ALIAS)>>12]);
		set_pte(pte, entry);
		local_irq_save(flags);
		__flush_tlb_page(get_asid(), p3_addr);
		local_irq_restore(flags);
		update_mmu_cache(NULL, p3_addr, entry);
		__copy_user_page((void *)p3_addr, from, to);
		pte_clear(&init_mm, p3_addr, pte);
		up(&p3map_sem[(address & CACHE_ALIAS)>>12]);
	}
}