转自：https://zhuanlan.zhihu.com/p/361173109
当内核需要对申请的page进行回收时，在回收页表前需要解除该page的映射关系，即内核需要知道这个物理页被映射到了哪些进程虚拟地址空间，因此就有了反向映射机制。反向映射一般分为匿名页映射和文件页映射，本文先介绍匿名页反向映射。
基本数据结构
page结构体中涉及反向映射的相关成员
struct page {。。。struct address_space *mapping; /* If low bit clear, points to * inode address_space, or NULL. * If page mapped as anonymous * memory, low bit is set, and * it points to anon_vma object: * see PAGE_MAPPING_ANON below. */。。。/* page_deferred_list().next -- second tail page */ };/* Second double word */union { pgoff_t index; /* Our offset within mapping. */。。。union { atomic_t _mapcount;
mapping因为指针变量是4个字节，因此可以用最后两位来区分不同的映射。对于匿名映射，最低位为PAGE_MAPPING_ANON，指向anon_vma结构体，每个匿名页对应唯一的anon_vma；对于文件映射而言，指向address_space结构体。
index表示页偏移，对于匿名映射，index表示page在vm_areat_struct指定的虚拟内存区域中的页偏移；对于匿名映射，index表示物理页中的数据在文件中的页偏移。
_mapcount记录该page被映射到了多少个vm_struct虚拟内存区域。注意和mm_struct结构体中的map_count做区分，map_count表示mm_strcut中有多少个vm_struct区域。
一般struct anon_vma称为AV，struct anon_vma_chain称为AVC，struct vm_area_struct称为VMA，page找到VMA的路径一般如下：page->AV->AVC->VMA，其中AVC起到桥梁作用，至于为何需要AVC，主要考虑当父进程和多个子进程同时拥有共同的page时的查询效率，具体对比2.6版本时的实现方式。
struct anon_vma
struct anon_vma {struct anon_vma *root; /* Root of this anon_vma tree */struct rw_semaphore rwsem; /* W: modification, R: walking the list *//* * The refcount is taken on an anon_vma when there is no * guarantee that the vma of page tables will exist for * the duration of the operation. A caller that takes * the reference is responsible for clearing up the * anon_vma if they are the last user on release */ atomic_t refcount;/* * Count of child anon_vmas and VMAs which points to this anon_vma. * * This counter is used for making decision about reusing anon_vma * instead of forking new one. See comments in function anon_vma_clone. */unsigned degree;struct anon_vma *parent; /* Parent of this anon_vma *//* * NOTE: the LSB of the rb_root.rb_node is set by * mm_take_all_locks() _after_ taking the above lock. So the * rb_root must only be read/written after taking the above lock * to be sure to see a valid next pointer. The LSB bit itself * is serialized by a system wide lock only visible to * mm_take_all_locks() (mm_all_locks_mutex). */struct rb_root rb_root; /* Interval tree of private "related" vmas */};
struct anon_vma_chain
struct anon_vma_chain {struct vm_area_struct *vma;struct anon_vma *anon_vma;struct list_head same_vma; /* locked by mmap_sem & page_table_lock */struct rb_node rb; /* locked by anon_vma->rwsem */unsignedlong rb_subtree_last;#ifdef CONFIG_DEBUG_VM_RBunsignedlong cached_vma_start, cached_vma_last;#endif};
struct vm_struct中相关成员
struct vm_area_struct {。。。struct list_head anon_vma_chain; /* Serialized by mmap_sem & * page_table_lock */struct anon_vma *anon_vma; /* Serialized by page_table_lock */。。。}
上面几个结构体的关系大致如下：
page通过mapping找到VMA，VMA 遍历自己管理的红黑树rb_root，找到树上的每个节点AVC，AVC通过成员指针anon_vma找到对应的VMA，这个过程就完成了页表映射查找。需要注意的几点：
1.VMA中也有链表anon_vma_chain管理各个AVC，这里主要用在父子进程之间的管理，下文会详细介绍。
2.VMA中有成员指针成员anon_vma，同时AVC中也有成员指针anon_vma，VAC起到桥梁作用所以可以指向VMA和AVC，那VMA中为何又需要指向AV呢？进程创建的流程中一般都是新建AV，然后创建AVC及AMV，然后调用anon_vma_chain_link建立三者之间的关系，但是当一个VMA没有对应页的时候，此时触发pagefault，这里可以快速判断VMA有没有对应的page。
常用接口
staticvoidanon_vma_chain_link(struct vm_area_struct *vma,struct anon_vma_chain *avc,struct anon_vma *anon_vma){ avc->vma = vma; avc->anon_vma = anon_vma; list_add(&avc->same_vma, &vma->anon_vma_chain); anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);}
代码实现
反向映射跟父子进程的写时拷贝有关系，所以先从父子进程创建时对AV，AVC，VMA的创建开始讲。
1.父进程创建匿名页面
当触发pagefault的时候走到handle_pte_fault中，anon_vma_prepare中负责创建AVC和AV并建立彼此的关系；真正将创建的page与av关联在__page_set_anon_map中完成。这样的话父进程新建的page在自己的反向映射中的关系就算完成了。
intanon_vma_prepare(struct vm_area_struct *vma){struct anon_vma *anon_vma = vma->anon_vma;struct anon_vma_chain *avc;。。。if (unlikely(!anon_vma)) {struct mm_struct *mm = vma->vm_mm;struct anon_vma *allocated; avc = anon_vma_chain_alloc(GFP_KERNEL);。。。 anon_vma = find_mergeable_anon_vma(vma); allocated =NULL;if (!anon_vma) { anon_vma = anon_vma_alloc();。。。 allocated = anon_vma; } anon_vma_lock_write(anon_vma);/* page_table_lock to protect against threads */ spin_lock(&mm->page_table_lock);if (likely(!vma->anon_vma)) { vma->anon_vma = anon_vma; anon_vma_chain_link(vma, avc, anon_vma);/* vma reference or self-parent link for new root */ anon_vma->degree++; allocated =NULL; avc =NULL; } spin_unlock(&mm->page_table_lock); anon_vma_unlock_write(anon_vma);。。。 }return0;。。。}staticvoid__page_set_anon_rmap(struct page *page,struct vm_area_struct *vma, unsignedlong address, int exclusive){struct anon_vma *anon_vma = vma->anon_vma;。。。 anon_vma = (void*) anon_vma + PAGE_MAPPING_ANON; page->mapping = (struct address_space *) anon_vma; page->index = linear_page_index(vma, address);}
至于index的含义看linear_page_index的实现应该就明白了。
staticinline pgoff_t linear_page_index(struct vm_area_struct *vma,unsignedlong address){ pgoff_t pgoff;if (unlikely(is_vm_hugetlb_page(vma)))return linear_hugepage_index(vma, address); pgoff = (address - vma->vm_start) >> PAGE_SHIFT; pgoff += vma->vm_pgoff;return pgoff;}
2.父进程创建子进程
当父进程创建子进程的时候，子进程会复制父进程的VMA作为自己的进程地址空间，并且父子进程共享相同的page，知道子进程往自己的地址空间写数据，这就是所谓的COW。这种情况需要完成两件事情：1.子进程需要继承父进程的AVC，AV，VMA及三者之间的关系；2.创建自己的AV，AVC，VMA。
以上实现流程在dup_mm->dup_mmap->anon_vma_fork中完成。
dup_mmap中就是组个创建子进程的vma，并复制父进程对应vma的信息
anon_vma_clone中新建了AVC，将子进程的VMA关联到父进程的AV中，所以父进程AV的rb树上就有了子进程的AVC，通过遍历父进程AV的rb树就能找到子进程的VMA。一个VMA可以包含多个page，但是该区域内的所有page只需要一个AV来反向映射即可。
具体anon_vma_clone代码如下
intanon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src){struct anon_vma_chain *avc, *pavc;struct anon_vma *root =NULL; list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {struct anon_vma *anon_vma; avc = anon_vma_chain_alloc(GFP_NOWAIT | __GFP_NOWARN);if (unlikely(!avc)) { unlock_anon_vma_root(root); root =NULL; avc = anon_vma_chain_alloc(GFP_KERNEL);if (!avc)goto enomem_failure; } anon_vma = pavc->anon_vma; root = lock_anon_vma_root(root, anon_vma); anon_vma_chain_link(dst, avc, anon_vma);/* * Reuse existing anon_vma if its degree lower than two, * that means it has no vma and only one anon_vma child. * * Do not chose parent anon_vma, otherwise first child * will always reuse it. Root anon_vma is never reused: * it has self-parent reference and at least one child. */if (!dst->anon_vma && anon_vma != src->anon_vma && anon_vma->degree <2) dst->anon_vma = anon_vma; }if (dst->anon_vma) dst->anon_vma->degree++; unlock_anon_vma_root(root);return0;
3.子进程发生cow，创建自己的匿名页面
当新创建的子进程写数据时触发pagefault，在wp_page_copy中会创建新的page，此时创建的AV和AVC管理子进程自己的VMA
if (fe->flags & FAULT_FLAG_WRITE) {if (!pte_write(entry))return do_wp_page(fe, entry); entry = pte_mkdirty(entry);}
4.页面回收，解除映射
物理页回收时通过调用try_to_unmap解除一个page的页表映射。对于匿名页面解除映射而言，走
try_to_unmap->rmap_walk->rmap_walk_anon流程。
intrmap_walk(struct page *page, struct rmap_walk_control *rwc){if (unlikely(PageKsm(page)))return rmap_walk_ksm(page, rwc);elseif (PageAnon(page))return rmap_walk_anon(page, rwc, false);elsereturn rmap_walk_file(page, rwc, false);}staticintrmap_walk_anon(struct page *page, struct rmap_walk_control *rwc,bool locked){struct anon_vma *anon_vma; pgoff_t pgoff;struct anon_vma_chain *avc;int ret = SWAP_AGAIN;if (locked) { anon_vma = page_anon_vma(page);/* anon_vma disappear under us? */ VM_BUG_ON_PAGE(!anon_vma, page); } else { anon_vma = rmap_walk_anon_lock(page, rwc); }。。。 pgoff = page_to_pgoff(page);// 遍历AV的红黑树，找到所有的AVC anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {// 通过AVC找到VMAstruct vm_area_struct *vma = avc->vma;// address为该page对应的起始地址unsignedlong address = vma_address(page, vma);。。。 ret = rwc->rmap_one(page, vma, address, rwc->arg);。。。}
rmap_one指向try_to_umap_one，该函数内容比较复杂，这里只截取了页表项解除的操作。
staticinttry_to_unmap_one(struct page *page, struct vm_area_struct *vma,unsignedlong address, void*arg){struct mm_struct *mm = vma->vm_mm; pte_t *pte; pte_t pteval; spinlock_t *ptl;int ret = SWAP_AGAIN;struct rmap_private *rp = arg;enum ttu_flags flags = rp->flags; pte = page_check_address(page, mm, address, &ptl, PageTransCompound(page));。。。/* Nuke the page table entry. */ flush_cache_page(vma, address, page_to_pfn(page));if (should_defer_flush(mm, flags)) {/* * We clear the PTE but do not flush so potentially a remote * CPU could still be writing to the page. If the entry was * previously clean then the architecture must guarantee that * a clear->dirty transition on a cached TLB entry is written * through and traps if the PTE is unmapped. */ pteval = ptep_get_and_clear(mm, address, pte); set_tlb_ubc_flush_pending(mm, page, pte_dirty(pteval)); } else { pteval = ptep_clear_flush(vma, address, pte); }。。。}
