0%

KVM源代码分析4:内存虚拟化

代码版本:https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git v3.16.37

在虚拟机的创建与运行中pc_init_pci负责(“KVM源代码分析2:虚拟机的创建与运行”),内存初始化也是在这里完成的,还是一步步从qemu说起,在vl.c的main函数中有ram_size参数,由qemu入参标识QEMU_OPTION_m设定,顾名思义就是虚拟机内存的大小,通过machine->init一步步传递给pc_init1函数。在这里分出了above_4g_mem_size和below_4g_mem_size,即高低端内存(也不一定是32bit机器..),然后开始初始化内存,即pc_memory_init,内存通过memory_region_init_ram下面的qemu_ram_alloc分配,使用qemu_ram_alloc_from_ptr。

插播qemu对内存条的模拟管理,是通过RAMBlock和ram_list管理的,RAMBlock就是每次申请的内存池,ram_list则是RAMBlock的链表,他们结构如下:

typedefstruct RAMBlock {
//对应宿主的内存地址
    uint8_t *host;
//block在ramlist中的偏移
    ram_addr_t offset;
//block长度
    ram_addr_t length;
    uint32_t flags;
//block名字
    char idstr[256];
    QLIST_ENTRY(RAMBlock) next;
\#if defined(__linux__) && !defined(TARGET_S390X)
    int fd;
\#endif
} RAMBlock;

typedef struct RAMList {
//看代码理解就是list的head,但是不知道为啥叫dirty...
    uint8_t *phys_dirty;
    QLIST_HEAD(ram, RAMBlock) blocks;
} RAMList;

下面再回到qemu_ram_alloc_from_ptr函数,使用find_ram_offset赋值给new block的offset,find_ram_offset具体工作模型已经在”KVM源代码分析2:虚拟机的创建与运行”,不赘述。然后是一串判断,在kvm_enabled的情况下使用new_block->host = kvm_vmalloc(size),最终内存是qemu_vmalloc分配的,使用qemu_memalign干活。

void \*qemu_memalign(size_t alignment, size_t size){
    void *ptr;
//使用posix进行内存针对页大小对齐
\#if defined(_POSIX_C_SOURCE) && !defined(__sun__)
    int ret;
    ret = posix_memalign(&ptr, alignment, size);
    if (ret != 0) {
        fprintf(stderr, "Failed to allocate %zu B: %sn",
                size, strerror(ret));
        abort();
    }
\#elif defined(CONFIG_BSD)
    ptr = qemu_oom_check(valloc(size));
\#else
//所谓检查oom就是看memalign对应malloc申请内存是否成功
    ptr = qemu_oom_check(memalign(alignment, size));
\#endif
    trace_qemu_memalign(alignment, size, ptr);
    return ptr;
}

以上qemu_vmalloc进行内存申请就结束了。在qemu_ram_alloc_from_ptr函数末尾则是将block添加到链表,realloc整个ramlist,用memset初始化整个ramblock,madvise对内存使用限定。
然后一层层的退回到pc_memory_init函数。

此时pc.ram已经分配完成,ram_addr已经拿到了分配的内存地址,MemoryRegion ram初始化完成。下面则是对已有的ram进行分段,即ram-below-4g和ram-above-4g,也就是高端内存和低端内存。用memory_region_init_alias初始化子MemoryRegion,然后将memory_region_add_subregion添加关联起来,memory_region_add_subregion具体细节“KVM源码分析2”中已经说了,参考对照着看吧,中间很多映射代码过程也只是qemu遗留的软件实现,没看到具体存在的意义,直接看到kvm_set_user_memory_region函数,内核真正需要kvm_vm_ioctl传递过去的参数是什么, struct kvm_userspace_memory_region mem而已,也就是

struct kvm_userspace_memory_region {
__u32 slot;
__u32 flags;
__u64 guest_phys_addr;
__u64 memory_size; /* bytes */
__u64 userspace_addr; /* start of the userspace allocated memory */
};

kvm_vm_ioctl进入到内核是在KVM_SET_USER_MEMORY_REGION参数中,即执行kvm_vm_ioctl_set_memory_region,然后一直向下,到kvm_set_memory_region函数,check_memory_region_flags检查mem->flags是否合法,而当前flag也就使用了两位,KVM_MEM_LOG_DIRTY_PAGES和KVM_MEM_READONLY,从qemu传递过来只能是KVM_MEM_LOG_DIRTY_PAGES,下面是对mem中各参数的合规检查,(mem->memory_size & (PAGE_SIZE - 1))要求以页为单位,(mem->guest_phys_addr & (PAGE_SIZE - 1))要求guest_phys_addr页对齐,而((mem->userspace_addr & (PAGE_SIZE - 1)) || !access_ok(VERIFY_WRITE,(void user *)(unsigned long)mem->userspace_addr,mem->memory_size))则保证host的线性地址页对齐而且该地址域有写权限。
id_to_memslot则是根据qemu的内存槽号得到kvm结构下的内存槽号,转换关系来自id_to_index数组,那映射关系怎么来的,映射关系是一一对应的,在kvm_create_vm “KVM源代码分析2:虚拟机的创建与运行”中,kvm_init_memslots_id初始化对应关系,即slots->id_to_index[i] = slots->memslots[i].id = i,当前映射是没有意义的,估计是为了后续扩展而存在的。
扩充了new的kvm_memory_slot,下面直接在代码中注释更方便:

//映射内存有大小,不是删除内存条if (npages) {
//内存槽号没有虚拟内存条,意味内存新创建if (!old.npages)
        change = KVM_MR_CREATE;
    else { /* Modify an existing slot. */
//修改已存在的内存修改标志或者平移映射地址
//下面是不能处理的状态(内存条大小不能变,物理地址不能变,不能修改只读)
        if ((mem->userspace_addr != old.userspace_addr) ||
            (npages != old.npages) ||
            ((new.flags ^ old.flags) & KVM_MEM_READONLY))
            goto out;
//guest地址不同,内存条平移
        if (base_gfn != old.base_gfn)
            change = KVM_MR_MOVE;
        else if (new.flags != old.flags)
//修改属性
            change = KVM_MR_FLAGS_ONLY;
        else { /* Nothing to change. */
            r = 0;
            goto out;
        }
    }
} else if (old.npages) {
//申请插入的内存为0,而内存槽上有内存,意味删除
    change = KVM_MR_DELETE;
} else /* Modify a non-existent slot: disallowed. */
    goto out;

另外看kvm_mr_change就知道memslot的变动值了:

enum kvm_mr_change {
    KVM_MR_CREATE,
    KVM_MR_DELETE,
    KVM_MR_MOVE,
    KVM_MR_FLAGS_ONLY,
};

在往下是一段检查

if ((change == KVM_MR_CREATE) || (change == KVM_MR_MOVE)) {
    /* Check for overlaps */
    r = -EEXIST;
    kvm_for_each_memslot(slot, kvm->memslots) {
        if ((slot->id >= KVM_USER_MEM_SLOTS) ||
//下面排除掉准备操作的内存条,在KVM_MR_MOVE中是有交集的
            (slot->id == mem->slot))
            continue;
//下面就是当前已有的slot与new在guest线性区间上有交集
        if (!((base_gfn + npages <= slot->base_gfn) ||
              (base_gfn >= slot->base_gfn + slot->npages)))
            goto out;
//out错误码就是EEXIST
    }
}

如果是新插入内存条,代码则走入kvm_arch_create_memslot函数,里面主要是一个循环,KVM_NR_PAGE_SIZES是分页的级数,此处是3,第一次循环,lpages = gfn_to_index(slot->base_gfn + npages - 1,slot->base_gfn, level) + 1,lpages就是一级页表所需要的page数,大致是npages>>09,然后为slot->arch.rmap[i]申请了内存空间,此处可以猜想,rmap就是一级页表了,继续看,lpages约为npages>>19,此处又多为lpage_info申请了同等空间,然后对lpage_info初始化赋值,现在看不到lpage_info的具体作用,看到后再补上。整体上看kvm_arch_create_memslot做了一个3级的软件页表。
如果有脏页,并且脏页位图为空,则分配脏页位图, kvm_create_dirty_bitmap实际就是”页数/8”.

if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
        if (kvm_create_dirty_bitmap(&new) < 0)
            goto out_free;
    }

当内存条的改变是KVM_MR_DELETE或者KVM_MR_MOVE,先申请一个slots,把kvm->memslots暂存到这里,首先通过id_to_memslot获取准备插入的内存条对应到kvm的插槽是slot,无论删除还是移动,将其先标记为KVM_MEMSLOT_INVALID,然后是install_new_memslots,其实就是更新了一下slots->generation的值。

内存的添加说完了,看一下EPT页表的映射,在kvm_arch_vcpu_setup中有kvm_mmu_setup,是mmu的初始化,EPT的初始化是init_kvm_tdp_mmu,所谓的初始化就是填充了vcpu->arch.mmu结构体,里面有很多回调函数都会用到,最终的是tdp_page_fault。

context->page_fault = tdp_page_fault;
context->sync_page = nonpaging_sync_page;
context->invlpg = nonpaging_invlpg;
context->update_pte = nonpaging_update_pte;
context->shadow_root_level = kvm_x86_ops->get_tdp_level();
context->root_hpa = INVALID_PAGE;
context->direct_map = true;
context->set_cr3 = kvm_x86_ops->set_tdp_cr3;
context->get_cr3 = get_cr3;
context->get_pdptr = kvm_pdptr_read;
context->inject_page_fault = kvm_inject_page_fault;

当guest访问物理内存时发生vm-exit,进入vmx_handle_exit函数,根据EXIT_REASON_EPT_VIOLATION走到handle_ept_violation函数,exit_qualification = vmcs_readl(EXIT_QUALIFICATION)获取vm-exit的退出原因,进入kvm_mmu_page_fault函数:vcpu->arch.mmu.page_fault(vcpu, cr2, error_code, false),即是tdp_page_fault,handle_mmio_page_fault的流程不提。

//填充kvm mmu专用的slab
r = mmu_topup_memory_caches(vcpu);
//获取gfn使用的level,即hugepage的问题
force_pt_level = mapping_level_dirty_bitmap(vcpu, gfn);
if (likely(!force_pt_level)) {
    level = mapping_level(vcpu, gfn);
    gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1);
} else
    level = PT_PAGE_TABLE_LEVEL;

//顾名思义,快速处理一个简单的page fault
//即present同时有写权限的非mmio page fault
//参考page_fault_can_be_fast函数
//一部分处理没有写权限的page fault
//一部分处理 TLB lazy
//fast_pf_fix_direct_spte也就是将pte获取的写权限
if (fast_page_fault(vcpu, gpa, level, error_code))
    return 0;
//下面函数主要就一件事情,gfn_to_pfn
if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable))
      return 0;
//direct map就是映射ept页表的过程
r = __direct_map(vcpu, gpa, write, map_writable,
      level, gfn, pfn, prefault);

在try_async_pf中就是gfn转换成hva,然后hva转换成pfn的过程,gfn转换到hva:

static pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic,
             bool *async, bool write_fault, bool *writable)
{
    unsigned long addr = __gfn_to_hva_many(slot, gfn, NULL, write_fault);

    if (addr == KVM_HVA_ERR_RO_BAD)
        return KVM_PFN_ERR_RO_FAULT;

    if (kvm_is_error_hva(addr))
        return KVM_PFN_NOSLOT;

    /* Do not map writable pfn in the readonly memslot. */
    if (writable && memslot_is_readonly(slot)) {
        *writable = false;
        writable = NULL;
    }

    return hva_to_pfn(addr, atomic, async, write_fault,
              writable);
}

gfn2hva本质就是

staticinline unsigned long
__gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
{
    return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
}

而hva_to_pfn则就是host的线性区进行地址转换的问题了,不提。

static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write,
            int map_writable, int level, gfn_t gfn, pfn_t pfn,
            bool prefault)
{
    struct kvm_shadow_walk_iterator iterator;
    struct kvm_mmu_page *sp;
    int emulate = 0;
    gfn_t pseudo_gfn;

    if (!VALID_PAGE(vcpu->arch.mmu.root_hpa))
        return0;
//遍历ept四级页表
    for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
//如果是最后一级,level是hugepage下的level
        if (iterator.level == level) {
//设置pte,页表下一级的page地址就是pfn写入到pte
            mmu_set_spte(vcpu, iterator.sptep, ACC_ALL,
                     write, &emulate, level, gfn, pfn,
                     prefault, map_writable);
            direct_pte_prefetch(vcpu, iterator.sptep);
            ++vcpu->stat.pf_fixed;
            break;
        }

        drop_large_spte(vcpu, iterator.sptep);
//mmu page不在位的情况,也就是缺页
        if (!is_shadow_present_pte(*iterator.sptep)) {
            u64 base_addr = iterator.addr;
//获取指向的具体mmu page entry的index
            base_addr &= PT64_LVL_ADDR_MASK(iterator.level);
            pseudo_gfn = base_addr >> PAGE_SHIFT;
//获取mmu page
            sp = kvm_mmu_get_page(vcpu, pseudo_gfn, iterator.addr,
                          iterator.level - 1,
                          1, ACC_ALL, iterator.sptep);
//将当前的mmu page的地址写入到上一级别mmu page的pte中
            link_shadow_page(iterator.sptep, sp, true);
        }
    }
    return emulate;
}

static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
                         gfn_t gfn,
                         gva_t gaddr,
                         unsigned level,
                         int direct,
                         unsigned access,
                         u64 *parent_pte)
{
    union kvm_mmu_page_role role;
    unsigned quadrant;
    struct kvm_mmu_page *sp;
    bool need_sync = false;

    role = vcpu->arch.mmu.base_role;
    role.level = level;
    role.direct = direct;
    if (role.direct)
        role.cr4_pae = 0;
    role.access = access;
    if (!vcpu->arch.mmu.direct_map
        && vcpu->arch.mmu.root_level <= pt32_root_level) { quadrant="gaddr">> (PAGE_SHIFT + (PT64_PT_BITS * level));
        quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
        role.quadrant = quadrant;
    }
//根据一个hash索引来的
    for_each_gfn_sp(vcpu->kvm, sp, gfn) {
//检查整个mmu ept是否被失效了
        if (is_obsolete_sp(vcpu->kvm, sp))
            continue;

        if (!need_sync && sp->unsync)
            need_sync = true;

        if (sp->role.word != role.word)
            continue;

        if (sp->unsync && kvm_sync_page_transient(vcpu, sp))
            break;

        mmu_page_add_parent_pte(vcpu, sp, parent_pte);
        if (sp->unsync_children) {
            kvm_make_request(KVM_REQ_MMU_SYNC, vcpu);
            kvm_mmu_mark_parents_unsync(sp);
        } else if (sp->unsync)
            kvm_mmu_mark_parents_unsync(sp);

        __clear_sp_write_flooding_count(sp);
        trace_kvm_mmu_get_page(sp, false);
        return sp;
    }
    ++vcpu->kvm->stat.mmu_cache_miss;
    sp = kvm_mmu_alloc_page(vcpu, parent_pte, direct);
    if (!sp)
        return sp;
    sp->gfn = gfn;
    sp->role = role;
//新的mmu page加入hash索引,所以前面的for循环中才能知道gfn对应的mmu有没有
//被分配
    hlist_add_head(&sp->hash_link,
        &vcpu->kvm->arch.mmu_page_hash[kvm_page_table_hashfn(gfn)]);
    if (!direct) {
        if (rmap_write_protect(vcpu->kvm, gfn))
            kvm_flush_remote_tlbs(vcpu->kvm);
        if (level > PT_PAGE_TABLE_LEVEL && need_sync)
            kvm_sync_pages(vcpu, gfn);

        account_shadowed(vcpu->kvm, gfn);
    }
    sp->mmu_valid_gen = vcpu->kvm->arch.mmu_valid_gen;
    init_shadow_page_table(sp);
    trace_kvm_mmu_get_page(sp, true);
    return sp;
}

这样看每次缺页都会分配新的mmu page,虚拟机每次启动是根据guest不停的进行EXIT_REASON_EPT_VIOLATION,整个页表就建立起来了。

坚持原创技术分享,您的支持将鼓励我继续创作!