Linux Kernel Development有关内存管理

1 Pages

Page的概念来源为处理器Processor的部件MMU(Memory Management Unit)，MMU通过设置好的页表(通过设置CR3寄存器，指向页目录所在的物理内存)对内存进行管理，管理操作包括：

a) 建立线性内存地址与物理内存地址的对应关系，即pa()和va()函数；

b) 管理哪些内存页驻存(Resident)于物理内存中，而哪些内存被交换到Swap文件中；

c) 哪些内存页被映射到哪个进程的虚拟地址空间；

d) 管理哪些内存页存储磁盘上（或者文件系统中）文件的缓存；

数据结构， struct page

   1: /*

10:

11:

12:

13:

14:

15:

16:

17:

18:

19:

20:

21:

22:

23:

24:

25:

26:

27:

28:

29:

31:

32:

33:

34:

35:

36:

37:

38:

39:

40:

41:

42:

43:

44:

45:

46:

47:

48:

49:

50:

51:

52:

54:

56:

57:

58:

59:

60:

61:

62:

63:

64:

65:

66:

67:

68:

69:

70:

71:

72:

73:

74:

75:

77:

78:

81:

85:

86:

87:

88:

89:

每个page结构体对象，代表一个物理内存页。 2: * Each physical page in the system has a struct page associated with 3: * it to keep track of whatever it is we are using the page for at the 4: * moment. Note that we have no way to track which tasks are using 5: * a page, though if it is a pagecache page, rmap structures can tell us 6: * who is mapping it. 7: */ 8: struct page { unsigned long flags; /* Atomic flags, some possibly * updated asynchronously */ atomic_t _count; /* Usage count, see below. */ union { /* * Count of ptes mapped in * mms, to show when page is * mapped & limit reverse map * searches. * * Used also for tail pages * refcounting instead of * _count. Tail pages cannot * be mapped and keeping the * tail page _count zero at * all times guarantees * get_page_unless_zero() will * never succeed on tail * pages. */ atomic_t _mapcount; 30: struct { /* SLUB */ u16 inuse; u16 objects; }; }; union { struct { unsigned long private; /* Mapping-private opaque data: * usually used for buffer_heads * if PagePrivate set; used for * swp_entry_t if PageSwapCache; * indicates order in the buddy * system if PG_buddy is set. */ 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. */ }; 53: #if USE_SPLIT_PTLOCKS spinlock_t ptl; 55: #endif struct kmem_cache *slab; /* SLUB: Pointer to slab */ struct page *first_page; /* Compound tail pages */ }; union { pgoff_t index; /* Our offset within mapping. */ void *freelist; /* SLUB: freelist req. slab lock */ }; struct list_head lru; /* Pageout list, eg. active_list * protected by zone->lru_lock ! */ /* * On machines where all RAM is mapped into kernel address space, * we can simply calculate the virtual address. On machines with * highmem some memory is mapped into kernel virtual memory * dynamically, so we need a place to store that address. * Note that this field could be 16 bits on x86 ... ;) * * Architectures with slow multiplication can define * WANT_PAGE_VIRTUAL in asm/page.h */ 76: #if defined(WANT_PAGE_VIRTUAL) void *virtual; /* Kernel virtual address (NULL if not kmapped, ie. highmem) */ 79: #endif /* WANT_PAGE_VIRTUAL */ 80: #ifdef CONFIG_WANT_PAGE_DEBUG_FLAGS unsigned long debug_flags; /* Use atomic bitops on this */ 82: #endif 83: 84: #ifdef CONFIG_KMEMCHECK /* * kmemcheck wants to track the status of each byte in a page; this * is a pointer to such a status block. NULL if not tracked. */ void *shadow; 90: #endif 91: };

x) flags

   1: enum pageflags {

   2:     PG_locked,        /* Page is locked. Don't touch. */

   3:     PG_error,

   4:     PG_referenced,

   5:     PG_uptodate,

   6:     PG_dirty,

   7:     PG_lru,

   8:     PG_active,

   9:     PG_slab,

  10:     PG_owner_priv_1,    /* Owner use. If pagecache, fs may use*/

  11:     PG_arch_1,

  12:     PG_reserved,

  13:     PG_private,        /* If pagecache, has fs-private data */

  14:     PG_private_2,        /* If pagecache, has fs aux data */

  15:     PG_writeback,        /* Page is under writeback */

  16: #ifdef CONFIG_PAGEFLAGS_EXTENDED

  17:     PG_head,        /* A head page */

  18:     PG_tail,        /* A tail page */

  19: #else

  20:     PG_compound,        /* A compound page */

  21: #endif

  22:     PG_swapcache,        /* Swap page: swp_entry_t in private */

  23:     PG_mappedtodisk,    /* Has blocks allocated on-disk */

  24:     PG_reclaim,        /* To be reclaimed asap */

  25:     PG_swapbacked,        /* Page is backed by RAM/swap */

  26:     PG_unevictable,        /* Page is "unevictable"  */

  27: #ifdef CONFIG_MMU

  28:     PG_mlocked,        /* Page is vma mlocked */

  29: #endif

  30: #ifdef CONFIG_ARCH_USES_PG_UNCACHED

  31:     PG_uncached,        /* Page has been mapped as uncached */

  32: #endif

  33: #ifdef CONFIG_MEMORY_FAILURE

  34:     PG_hwpoison,        /* hardware poisoned page. Don't touch */

  35: #endif

  36: #ifdef CONFIG_TRANSPARENT_HUGEPAGE

  37:     PG_compound_lock,

  38: #endif

  39:     __NR_PAGEFLAGS,

40:

  41:     /* Filesystems */

  42:     PG_checked = PG_owner_priv_1,

43:

  44:     /* Two page bits are conscripted by FS-Cache to maintain local caching

  45:      * state.  These bits are set on pages belonging to the netfs's inodes

  46:      * when those inodes are being locally cached.

  47:      */

  48:     PG_fscache = PG_private_2,    /* page backed by cache */

49:

  50:     /* XEN */

  51:     PG_pinned = PG_owner_priv_1,

  52:     PG_savepinned = PG_dirty,

53:

  54:     /* SLOB */

  55:     PG_slob_free = PG_private,

56:

  57:     /* SLUB */

  58:     PG_slub_frozen = PG_active,

  59: };

x) _count

引用计数，代表有多少处引用到该物理内存页对象。

访问_count时，不要直接访问，调用 page_count()对象进行访问。

x) private, mapping

   1: struct address_space {

   2:     struct inode        *host;        /* owner: inode, block_device */

   3:     struct radix_tree_root    page_tree;    /* radix tree of all pages */

   4:     spinlock_t        tree_lock;    /* and lock protecting it */

   5:     unsigned int        i_mmap_writable;/* count VM_SHARED mappings */

   6:     struct prio_tree_root    i_mmap;        /* tree of private and shared mappings */

   7:     struct list_head    i_mmap_nonlinear;/*list VM_NONLINEAR mappings */

   8:     struct mutex        i_mmap_mutex;    /* protect tree, count, list */

   9:     /* Protected by tree_lock together with the radix tree */

  10:     unsigned long        nrpages;    /* number of total pages */

  11:     pgoff_t            writeback_index;/* writeback starts here */

  12:     const struct address_space_operations *a_ops;    /* methods */

  13:     unsigned long        flags;        /* error bits/gfp mask */

  14:     struct backing_dev_info *backing_dev_info; /* device readahead, etc */

  15:     spinlock_t        private_lock;    /* for use by the address_space */

  16:     struct list_head    private_list;    /* ditto */

  17:     struct address_space    *assoc_mapping;    /* ditto */

  18: } __attribute__((aligned(sizeof(long))));

x) virtual

虚拟地址，如果是高端内存(HighMem)，那么该物理页可能不是长久地映射到内核的内存空间中，所以该字段为NULL。

struct page

这个结构只是用来描述物理内存页，而与该页中存储的内容无关。

物理内存页的可能属主(Owner)有：

user-space processes, 用户态的进程
dynamically allocated kernel data，内核态中动态分配的数据
static kernel code，内核静态代码
the page cache，页缓存

可能程序员对为每个物理内存页都分配一个struct page而感到吃惊，“那得分配多少内存啊，多浪费啊！”。

实际上以4GB内存为例，大概需要40MB的内存来存储所有struct page的对象，相对于它能够管理的4GB物理内存，还是十分微不足道的。

2. Zones

“为什么不利用用户态的3GB的地址空间来映射内核的1GB地址空间映射不下的物理内存？”

因为其实，内核在工作时，是不考虑存在着用户空间的。

内核要处理的任务远比支撑用户空间要复杂。

为什么要使用不同的Zone?

有些硬件设备的DMA能访问的内存空间十分有即，比如x86的ISA总线，只能访问0~16M的物理内存，那么如果这段内存被随便地分配掉了，ISA设备就可能无法工作了。

ISA

ISA插槽是基于ISA总线（Industrial Standard Architecture，工业标准结构总线）的扩展插槽，其颜色一般为黑色，比PCI接口插槽要长些，位于主板的最下端。其工作频率为8MHz左右，为16位插槽，最大传输率16MB/sec，可插接显卡，声卡，网卡以及所谓的多功能接口卡等扩展插卡。其缺点是CPU资源占用太高，数据传输带宽太小，是已经被淘汰的插槽接口。（http://baike.baidu.com/view/13594.htm）

HighMem

对于x86体系结构，以896MB物理内存为界，大于该范围的为高端内存(High Memory)，而小于该范围的为低端内存(Low Memory)。

内核无法直接映射HighMem区域的物理内存，只能以暂时的方式映射其中一小部分使用，当需要使用其他的高端内存时，可以需要打破之前建立的暂时映射，而用于新的映射，就像是内存交换机制一样。