ceph主要数据结构解析2-Rados.h

（1）文件系统id结构：16个字符组成

struct ceph_fsid {unsigned char fsid[16];};

以及对应的比较函数：

static inline int ceph_fsid_compare(const struct ceph_fsid *a,    const struct ceph_fsid *b){return memcmp(a, b, sizeof(*a));}

（2）定义保留的snap的id宏

typedef __le64 ceph_snapid_t;#define CEPH_SNAPDIR ((__u64)(-1))  /* reserved for hidden .snap dir */#define CEPH_NOSNAP  ((__u64)(-2))  /* "head", "live" revision */#define CEPH_MAXSNAP ((__u64)(-3))  /* largest valid snapid */struct ceph_timespec {//自定义的时间描述结构体__le32 tv_sec;__le32 tv_nsec;} __attribute__ ((packed));

（3）对象布局宏：对象怎样放到放置组PG里面去（placement group）

#define CEPH_OBJECT_LAYOUT_HASH     1 //哈希#define CEPH_OBJECT_LAYOUT_LINEAR   2 //线性#define CEPH_OBJECT_LAYOUT_HASHINO  3 //哈希inodestruct ceph_object_layout {//对象布局struct ceph_pg ol_pgid;   /* raw pg, with _full_ ps precision. */__le32 ol_stripe_unit;    /* for per-object parity, if any */} __attribute__ ((packed));

（4）PG布局宏：PG怎样放置到存储设备上去

#define CEPH_PG_LAYOUT_CRUSH  0 //CRUSH算法，伪随机hash#define CEPH_PG_LAYOUT_HASH   1 //hash#define CEPH_PG_LAYOUT_LINEAR 2 //线性#define CEPH_PG_LAYOUT_HYBRID 3 #define CEPH_PG_MAX_SIZE      16  /* max # osds in a single pg */

（5）PG编码到一个64位

struct ceph_pg {__le16 preferred; /* preferred primary osd */参考主osd__le16 ps;        /* placement seed */放置种子__le32 pool;      /* object pool */对象池} __attribute__ ((packed));

下面是两种对象池的宏定义：

#define CEPH_PG_TYPE_REP     1 //冗余#define CEPH_PG_TYPE_RAID4   2 //raid

控制PG数量的函数定义：保证b小于bmask并且bmask是2的n次方的最小值

随着时间b的增加产生一个稳定的映射

static inline int ceph_stable_mod(int x, int b, int bmask){if ((x & bmask) < b)return x & bmask;elsereturn x & (bmask >> 1);}

（6）划时代和版本的组合，被用作存储层的序列化突变

struct ceph_eversion {__le32 epoch;__le64 version;} __attribute__ ((packed));

（7）osd（对象存储设备）映射位的宏定义：

#define CEPH_OSD_EXISTS  (1<<0) //osd存在#define CEPH_OSD_UP      (1<<1) //#define CEPH_OSD_AUTOOUT (1<<2)  /* osd was automatically marked out */自动标出#define CEPH_OSD_NEW     (1<<3)  /* osd is new, never marked in */新的，从未标记extern const char *ceph_osd_state_name(int s);//得到对应的名字，实现如下：const char *ceph_osd_state_name(int s){switch (s) {case CEPH_OSD_EXISTS:return "exists";case CEPH_OSD_UP:return "up";case CEPH_OSD_AUTOOUT:return "autoout";case CEPH_OSD_NEW:return "new";default:return "???";}}/* osd weights.  fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */osd的权重值：最大为0x10000相当于1.0，最小为0#define CEPH_OSD_IN  0x10000#define CEPH_OSD_OUT 0

（8）osd映射标志位定义：

#define CEPH_OSDMAP_NEARFULL (1<<0)  /* sync writes (near ENOSPC) */#define CEPH_OSDMAP_FULL     (1<<1)  /* no data writes (ENOSPC) */#define CEPH_OSDMAP_PAUSERD  (1<<2)  /* pause all reads */暂停所有读#define CEPH_OSDMAP_PAUSEWR  (1<<3)  /* pause all writes */#define CEPH_OSDMAP_PAUSEREC (1<<4)  /* pause recovery *//* * The error code to return when an OSD can't handle a write * because it is too large. */因为数据太大导致OSD不能处理些时返回的错误代码#define OSD_WRITETOOBIG EMSGSIZE

（9）osd的操作

#define CEPH_OSD_OP_MODE       0xf000 //操作模式#define CEPH_OSD_OP_MODE_RD    0x1000 //读#define CEPH_OSD_OP_MODE_WR    0x2000 //写#define CEPH_OSD_OP_MODE_RMW   0x3000 //读修改写#define CEPH_OSD_OP_MODE_SUB   0x4000 //子操作#define CEPH_OSD_OP_TYPE       0x0f00 //操作类型#define CEPH_OSD_OP_TYPE_LOCK  0x0100  //锁#define CEPH_OSD_OP_TYPE_DATA  0x0200 //数据#define CEPH_OSD_OP_TYPE_ATTR  0x0300 //属性#define CEPH_OSD_OP_TYPE_EXEC  0x0400 //执行#define CEPH_OSD_OP_TYPE_PG    0x0500 //PG#define CEPH_OSD_OP_TYPE_MULTI 0x0600 /* multiobject */多对象

下面是定义操作模式与类型组合的枚举：

enum {CEPH_OSD_OP_READ = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 1,中间省略很多...CEPH_OSD_OP_PGLS_FILTER = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_PG | 2,};

以及定义了判断操作是否为每一种类型或模式的操作，如判断是否为锁操作类型的函数如下：

static inline int ceph_osd_op_type_lock(int op){return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_LOCK;}

返回操作名称的函数：

extern const char *ceph_osd_op_name(int op);

（10）osd操作标志定义：有详细的英文注释了

enum {CEPH_OSD_FLAG_ACK =            0x0001,  /* want (or is) "ack" ack */CEPH_OSD_FLAG_ONNVRAM =        0x0002,  /* want (or is) "onnvram" ack */CEPH_OSD_FLAG_ONDISK =         0x0004,  /* want (or is) "ondisk" ack */CEPH_OSD_FLAG_RETRY =          0x0008,  /* resend attempt */CEPH_OSD_FLAG_READ =           0x0010,  /* op may read */CEPH_OSD_FLAG_WRITE =          0x0020,  /* op may write */CEPH_OSD_FLAG_ORDERSNAP =      0x0040,  /* EOLDSNAP if snapc is out of order */CEPH_OSD_FLAG_PEERSTAT_OLD =   0x0080,  /* DEPRECATED msg includes osd_peer_stat */CEPH_OSD_FLAG_BALANCE_READS =  0x0100,CEPH_OSD_FLAG_PARALLELEXEC =   0x0200,  /* execute op in parallel */CEPH_OSD_FLAG_PGOP =           0x0400,  /* pg op, no object */CEPH_OSD_FLAG_EXEC =           0x0800,  /* op may exec */CEPH_OSD_FLAG_EXEC_PUBLIC =    0x1000,  /* op may exec (public) */CEPH_OSD_FLAG_LOCALIZE_READS = 0x2000,  /* read from nearby replica, if any */CEPH_OSD_FLAG_RWORDERED =      0x4000,  /* order wrt concurrent reads */};enum {CEPH_OSD_OP_FLAG_EXCL = 1,      /* EXCL object create */CEPH_OSD_OP_FLAG_FAILOK = 2,    /* continue despite failure */};#define EOLDSNAPC    85  /* ORDERSNAP flag set; writer has old snapc*/#define EBLACKLISTED 108 /* blacklisted */

（11）扩展属性比较操作枚举定义：

enum {CEPH_OSD_CMPXATTR_OP_NOP = 0,CEPH_OSD_CMPXATTR_OP_EQ  = 1,CEPH_OSD_CMPXATTR_OP_NE  = 2,CEPH_OSD_CMPXATTR_OP_GT  = 3,CEPH_OSD_CMPXATTR_OP_GTE = 4,CEPH_OSD_CMPXATTR_OP_LT  = 5,CEPH_OSD_CMPXATTR_OP_LTE = 6};enum {CEPH_OSD_CMPXATTR_MODE_STRING = 1,CEPH_OSD_CMPXATTR_MODE_U64    = 2};

（12）osd操作结构体封装定义，可能是来至数据有效负载：

struct ceph_osd_op {__le16 op;           /* CEPH_OSD_OP_* */__le32 flags;        /* CEPH_OSD_FLAG_* */union {//各种数据包格式定义struct {__le64 offset, length;//偏移量和长度__le64 truncate_size;__le32 truncate_seq;} __attribute__ ((packed)) extent;struct {__le32 name_len;__le32 value_len;__u8 cmp_op;       /* CEPH_OSD_CMPXATTR_OP_* */__u8 cmp_mode;     /* CEPH_OSD_CMPXATTR_MODE_* */} __attribute__ ((packed)) xattr;struct {__u8 class_len;__u8 method_len;__u8 argc;__le32 indata_len;} __attribute__ ((packed)) cls;struct {__le64 count;__le32 start_epoch; /* for the pgls sequence */} __attribute__ ((packed)) pgls;        struct {        __le64 snapid;        } __attribute__ ((packed)) snap;struct {__le64 cookie;__le64 ver;__u8 flag;/* 0 = unwatch, 1 = watch */} __attribute__ ((packed)) watch;struct {__le64 offset, length;__le64 src_offset;} __attribute__ ((packed)) clonerange;};__le32 payload_len;//数据长度} __attribute__ ((packed));struct ceph_osd_reply_head {//回复头部__le32 client_inc;                /* client incarnation */__le32 flags;struct ceph_object_layout layout;__le32 osdmap_epoch;struct ceph_eversion reassert_version; /* for replaying uncommitted */__le32 result;                    /* result code */__le32 object_len;                /* length of object name */__le32 num_ops; //操作数量struct ceph_osd_op ops[0];  /* ops[], object */存放所有操作} __attribute__ ((packed));