Linux内存管理 (12)反向映射RMAP
专题:Linux内存管理专题
关键词:RMAP、VMA、AV、AVC。
所谓反向映射是相对于从虚拟地址到物理地址的映射,反向映射是从物理页面到虚拟地址空间VMA的反向映射。
RMAP能否实现的基础是通过struct anon_vma、struct anon_vma_chain和sturct vm_area_struct建立了联系,通过物理页面反向查找到VMA。
用户在使用虚拟内存过程中,PTE页表项中保留着虚拟内存页面映射到物理内存页面的记录。
一个物理页面可以同时被多个进程的虚拟地址内存映射,但一个虚拟页面同时只能有一个物理页面与之映射。
不同虚拟页面同时映射到同一物理页面是因为子进程克隆父进程VMA,和KSM机制的存在。
如果页面要被回收,就必须要找出哪些进程在使用这个页面,然后断开这些虚拟地址到物理页面的映射。
匿名页面实际的断开映射操作在rmap_walk_anon中进行的,可以看出从struct page、到struct anon_vma、到struct anon_vma_chain、到struct vm_area_struct的关系。
1. 父进程分配匿名页面
父进程为自己的进程地址空间VMA分配物理内存时,通常会产生匿名页面。
do_anonymous_page()会分配匿名页面;do_wp_page()发生写时复制COW时也会产生一个新的匿名页面。
static int do_anonymous_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, pte_t *page_table, pmd_t *pmd, unsigned int flags) { ... /* Allocate our own private page. */ if (unlikely(anon_vma_prepare(vma)))------------------------------为进程地址空间准备struct anon_vma数据结构和struct anon_vma_chain链表。 goto oom; page = alloc_zeroed_user_highpage_movable(vma, address);----------从HIGHMEM区域分配一个zeroed页面 if (!page) goto oom; ... inc_mm_counter_fast(mm, MM_ANONPAGES); page_add_new_anon_rmap(page, vma, address);----------------------- mem_cgroup_commit_charge(page, memcg, false); lru_cache_add_active_or_unevictable(page, vma); ... }
RMAP反向映射系统中有两个重要的数据结构:一个是struct anon_vma,简称AV;一个是struct anon_vma_chain,简称AVC。
struct anon_vma { struct anon_vma *root; /* Root of this anon_vma tree */----------------指向anon_vma数据机构中的根节点 struct rw_semaphore rwsem; /* W: modification, R: walking the list */------保护anon_vma中链表的读写信号量 /* * 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;------------------------------------------------------------对anon_vma的引用计数 /* * 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 */--------------------指向父anon_vma数据结构 /* * 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 */-----红黑树根节点 }
struct anon_vma_chain数据结构是链接父子进程的枢纽:
struct anon_vma_chain { struct vm_area_struct *vma;-----------------------------------------------指向VMA struct anon_vma *anon_vma;------------------------------------------------指向anon_vma数据结构,可以指向父进程或子进程的anon_vma数据结构。 struct list_head same_vma; /* locked by mmap_sem & page_table_lock */---链表节点,通常把anon_vma_chain添加到vma->anon_vma_chain链表中。 struct rb_node rb; /* locked by anon_vma->rwsem */-------------红黑树节点,通常把anon_vma_chain添加到anon_vma->rb_root的红黑树中。 unsigned long rb_subtree_last; #ifdef CONFIG_DEBUG_VM_RB unsigned long cached_vma_start, cached_vma_last; #endif }
下面分析如何建立AV、AVC、VMA之间的关系:
int anon_vma_prepare(struct vm_area_struct *vma) { struct anon_vma *anon_vma = vma->anon_vma;--------------vma->anon_vma指向struct anon_vma数据结构。 struct anon_vma_chain *avc; might_sleep(); if (unlikely(!anon_vma)) { struct mm_struct *mm = vma->vm_mm; struct anon_vma *allocated; avc = anon_vma_chain_alloc(GFP_KERNEL);------------分配一个struct anon_vma_chain结构。 if (!avc) goto out_enomem; anon_vma = find_mergeable_anon_vma(vma);-----------是否可以和前后vma合并 allocated = NULL; if (!anon_vma) { anon_vma = anon_vma_alloc();-------------------如果无法合并,则重新分配一个结构体 if (unlikely(!anon_vma)) goto out_enomem_free_avc; 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;-------------------------建立struct vm_area_struct和struct anon_vma关联 anon_vma_chain_link(vma, avc, anon_vma);----------建立struct anon_vma_chain和其他结构体的关系。 /* 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); if (unlikely(allocated)) put_anon_vma(allocated); if (unlikely(avc)) anon_vma_chain_free(avc); } return 0; out_enomem_free_avc: anon_vma_chain_free(avc); out_enomem: return -ENOMEM; }
至此已经建立struct vm_area_struct、struct anon_vma、struct anon_vma_chain三者之间的链接,并插入相应链表、红黑树中。
从AVC可以轻松找到VMA和AV;AV可以通过红黑树找到AVC,然后发现所有红黑树中的AV;VMA可以直接找到AV,也可以通过AVC链表找到AVC。
static void anon_vma_chain_link(struct vm_area_struct *vma, struct anon_vma_chain *avc, struct anon_vma *anon_vma) { avc->vma = vma;--------------------------------------------建立struct anon_vma_chain和struct vm_area_struct关联 avc->anon_vma = anon_vma;----------------------------------建立struct anon_vma_chain和struct anon_vma关联 list_add(&avc->same_vma, &vma->anon_vma_chain);------------将AVC添加到struct vm_area_struct->anon_vma_chain链表中。 anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);----将AVC添加到struct anon_vma->rb_root红黑树中。 }
调用alloc_zeroed_user_highpage_movable分配物理内存之后,调用page_add_new_anon_rmap建立PTE映射关系。
void page_add_new_anon_rmap(struct page *page, struct vm_area_struct *vma, unsigned long address) { VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma); SetPageSwapBacked(page);----------------------------------------------------------设置PG_SwapBacked表示这个页面可以swap到磁盘。 atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */-------------设置_mapcount引用计数为0 if (PageTransHuge(page)) __inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES); __mod_zone_page_state(page_zone(page), NR_ANON_PAGES,-----------------------------增加页面所在zone的匿名页面计数 hpage_nr_pages(page)); __page_set_anon_rmap(page, vma, address, 1);--------------------------------------设置这个页面位匿名映射 } static void __page_set_anon_rmap(struct page *page, struct vm_area_struct *vma, unsigned long address, int exclusive) { struct anon_vma *anon_vma = vma->anon_vma; BUG_ON(!anon_vma); if (PageAnon(page))---------------------------------------------------------------判断当前页面是否是匿名页面PAGE_MAPPING_ANON return; /* * If the page isn't exclusively mapped into this vma, * we must use the _oldest_ possible anon_vma for the * page mapping! */ if (!exclusive) anon_vma = anon_vma->root; anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON; page->mapping = (struct address_space *) anon_vma;------------------------------mapping指定页面所在的地址空间,这里指向匿名页面的地址空间数据结构struct anon_vma。 page->index = linear_page_index(vma, address); }
结合上图,总结如下:
- 父进程每个VMA都有一个anon_vma数据结构,vma->anon_vma指向。
- 和VMA相关的物理页面page->mapping都指向anon_vma。
- AVC数据结构anon_vma_chain->vma指向VMA,anon_vma_chain->anon_vma指向AV。
- AVC添加到VMA->anon_vma_chain链表中。
- AVC添加到AV->anon_vma红黑树中。
2. 父进程创建子进程
父进程通过fork系统调用创建子进程时,子进程会复制父进程的进程地址空间VMA数据结构作为自己的进程地址空间,并且会复制父进程的PTE页表项内容到子进程的页表中,实现父子进程共享页表。
多个不同子进程中的虚拟页面会同时映射到同一个物理页面,另外多个不相干进程虚拟页面也可以通过KSM机制映射到同一个物理页面。
fork()系统调用实现在kernel/fork.c中,在dup_mmap()中复制父进程的地址空间和父进程的PTE页表项:
static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) { struct vm_area_struct *mpnt, *tmp, *prev, **pprev; struct rb_node **rb_link, *rb_parent; int retval; unsigned long charge; uprobe_start_dup_mmap(); down_write(&oldmm->mmap_sem); flush_cache_dup_mm(oldmm); uprobe_dup_mmap(oldmm, mm); /* * Not linked in yet - no deadlock potential: */ down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); mm->total_vm = oldmm->total_vm; mm->shared_vm = oldmm->shared_vm; mm->exec_vm = oldmm->exec_vm; mm->stack_vm = oldmm->stack_vm; rb_link = &mm->mm_rb.rb_node; rb_parent = NULL; pprev = &mm->mmap; retval = ksm_fork(mm, oldmm); if (retval) goto out; retval = khugepaged_fork(mm, oldmm); if (retval) goto out; prev = NULL; for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {-------------------for循环遍历父进程的进程地址空间VMA。 struct file *file; if (mpnt->vm_flags & VM_DONTCOPY) { vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, -vma_pages(mpnt)); continue; } charge = 0; if (mpnt->vm_flags & VM_ACCOUNT) { unsigned long len = vma_pages(mpnt); if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ goto fail_nomem; charge = len; } tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); if (!tmp) goto fail_nomem; *tmp = *mpnt;--------------------------------------------------------复制父进程地址空间VMA到刚创建的子进程tmp中。 INIT_LIST_HEAD(&tmp->anon_vma_chain); retval = vma_dup_policy(mpnt, tmp); if (retval) goto fail_nomem_policy; tmp->vm_mm = mm; if (anon_vma_fork(tmp, mpnt))----------------------------------------为子进程创建相应的anon_vma数据结构。 goto fail_nomem_anon_vma_fork; tmp->vm_flags &= ~VM_LOCKED; tmp->vm_next = tmp->vm_prev = NULL; file = tmp->vm_file; ... __vma_link_rb(mm, tmp, rb_link, rb_parent);--------------------------把VMA添加到子进程红黑树中。 rb_link = &tmp->vm_rb.rb_right; rb_parent = &tmp->vm_rb; mm->map_count++; retval = copy_page_range(mm, oldmm, mpnt);---------------------------复制父进程的PTE页表项到子进程页表中。 if (tmp->vm_ops && tmp->vm_ops->open) tmp->vm_ops->open(tmp); if (retval) goto out; } ... }
3. 子进程发生COW
如果子进程的VMA发生COW,那么会使用子进程VMA创建的anon_vma数据结构,即page->mmaping指针指向子进程VMA对应的anon_vma数据结构。
在do_wp_page()函数中处理COW场景情况:子进程和父进程共享的匿名页面,子进程的VMA发生COW。
->发生缺页中断 ->handle_pte_fault ->do_wp_page ->分配一个新的匿名页面 ->__page_set_anon_rmap 使用子进程的anon_vma来设置page->mapping
4. RMAP应用
内核中通过struct page找到所有映射到这个页面的VMA典型场景有:
- kswapd内核线程回收页面需要断开所有映射了该匿名页面的用户PTE页表项。
- 页面迁移时,需要断开所有映射到匿名页面的用户PTE页表项。
try_to_unmap()是反向映射的核心函数,内核中其他模块会调用此函数来断开一个页面的所有映射:
/** * try_to_unmap - try to remove all page table mappings to a page * @page: the page to get unmapped * @flags: action and flags * * Tries to remove all the page table entries which are mapping this * page, used in the pageout path. Caller must hold the page lock. * Return values are: * * SWAP_SUCCESS - we succeeded in removing all mappings------------成功解除了所有映射的PTE。 * SWAP_AGAIN - we missed a mapping, try again later---------------可能错过了一个映射的PTE,需要重来一次。 * SWAP_FAIL - the page is unswappable-----------------------------失败 * SWAP_MLOCK - page is mlocked.-----------------------------------页面被锁住了 */ int try_to_unmap(struct page *page, enum ttu_flags flags) { int ret; struct rmap_walk_control rwc = { .rmap_one = try_to_unmap_one,--------------------------------具体断开某个VMA上映射的pte .arg = (void *)flags, .done = page_not_mapped,-------------------------------------判断一个页面是否断开成功的条件 .anon_lock = page_lock_anon_vma_read,------------------------锁 }; VM_BUG_ON_PAGE(!PageHuge(page) && PageTransHuge(page), page); /* * During exec, a temporary VMA is setup and later moved. * The VMA is moved under the anon_vma lock but not the * page tables leading to a race where migration cannot * find the migration ptes. Rather than increasing the * locking requirements of exec(), migration skips * temporary VMAs until after exec() completes. */ if ((flags & TTU_MIGRATION) && !PageKsm(page) && PageAnon(page)) rwc.invalid_vma = invalid_migration_vma; ret = rmap_walk(page, &rwc); if (ret != SWAP_MLOCK && !page_mapped(page)) ret = SWAP_SUCCESS; return ret; }
内核中有三种页面需要unmap操作,即KSM页面、匿名页面、文件映射页面:
int rmap_walk(struct page *page, struct rmap_walk_control *rwc) { if (unlikely(PageKsm(page))) return rmap_walk_ksm(page, rwc); else if (PageAnon(page)) return rmap_walk_anon(page, rwc); else return rmap_walk_file(page, rwc); }
下面以匿名页面的unmap为例:
static int rmap_walk_anon(struct page *page, struct rmap_walk_control *rwc) { struct anon_vma *anon_vma; pgoff_t pgoff; struct anon_vma_chain *avc; int ret = SWAP_AGAIN; anon_vma = rmap_walk_anon_lock(page, rwc);-----------------------------------获取页面page->mapping指向的anon_vma数据结构,并申请一个读者锁。 if (!anon_vma) return ret; pgoff = page_to_pgoff(page); anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {------遍历anon_vma->rb_root红黑树中的AVC,从AVC得到相应的VMA。 struct vm_area_struct *vma = avc->vma; unsigned long address = vma_address(page, vma); if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg)) continue; ret = rwc->rmap_one(page, vma, address, rwc->arg);-----------------------实际的断开用户PTE页表项操作。 if (ret != SWAP_AGAIN) break; if (rwc->done && rwc->done(page)) break; } anon_vma_unlock_read(anon_vma); return ret; }
struct rmap_walk_control中的rmap_one实现是try_to_unmap_one,最终调用page_remove_rmap()和page_cache_release()来断开PTE映射关系。