4.1.2.1 MASTER_SERV结构体

/*
*Server processes that provide the same service share a common"listen"
*socket to accept connection requests, and share a common pipe to the
*master process to send status reports. Server processes die voluntarily
*when idle for a configurable amount of time, or after servicing a
*configurable number of requests; the master process spawns new processes
*on demand up to a configurable concurrency limit and/or periodically.
*
*The canonical service name is what we use internally, so that we correctly
*handle a request to "reload" after someone changes "smtp"into "25".
*
*We use the external service name from master.cf when reporting problems, so
*that the user can figure out what we are talking about. Of course we also
*include the canonical service name so that the UNIX-domain smtp service
*can be distinguished from the Internet smtp service.
*/
/master/master.h
typedef struct MASTER_SERV {
int     flags;                    /* status, features, etc. */
char   *ext_name;                           /* service endpointname (master.cf) */
char   *name;                          /* service endpointname (canonical) */
int     type;                    /* UNIX-domain, INET, etc.*/
time_t  busy_warn_time;               /* limit "all serversbusy" warning */
int     wakeup_time;            /* wakeup interval */
int    *listen_fd;                      /* incoming requests */
int     listen_fd_count;                  /* nr of descriptors */
union {
struct{
char  *port;          /* inet listen port*/
struct INET_ADDR_LIST *addr;/* inet listenaddress */
}       inet_ep;
#define MASTER_INET_ADDRLIST(s)      ((s)->endpoint.inet_ep.addr)
#define MASTER_INET_PORT(s)      ((s)->endpoint.inet_ep.port)
}       endpoint;
int     max_proc;                   /* upper bound on # processes*/
char   *path;                            /* command pathname*/
struct ARGV *args;                    /*argument vector */
char   *stress_param_val;            /* stress value: "yes" orempty */
time_t  stress_expire_time;          /* stress pulse stretcher */
int     avail_proc;                            /* idle processes */
int     total_proc;                            /* number ofprocesses */
int     throttle_delay;           /* failure recovery parameter */
int     status_fd[2];               /* child status reports */
struct BINHASH *children;                /*linkage */
struct MASTER_SERV *next;            /*linkage */
} MASTER_SERV;

flags 状态标志。

ext_name master.cf中记录的模块名。

name 服务名。

type 模块服务类型，即inet,fifo,或unix。与master.cf中定义的一致。

busy_warn_time:繁忙警告时间。

wakeup_time:唤醒时间，即master.cf中的唤醒时间。

listen_fd:监听的文件描述符数组指针。

listen_fd_count：listen_fd个数，在MASTER_SERV结构体初始化时候确定，对于unix域协议和fifo管道为1，对于inet由master.cf配置文件中的配置决定。

max_proc:最大进程数，即master.cf中的最大进程数。

path:启动服务的命令路径，即master.cf中的command选项。

args:参数数组指针，即master.cf中的args选项。

stress_param_val：表示模块是否满负荷运行，其值为“yes”或空。

业务模块满负荷运转时,master模块重启该模块，并加上“stress=yes”命令行选项，此时stress_param_val为yes。非满负荷运转stress_param_val为空。

avail_proc:可用进程数。

total_proc：当前总进程数。

throttle_delay：当模块当前无法创建子进程时，flags会设置MASTER_FLAG_THROTTLE标志，throttle_delay就是这个标志的保留时间：

serv->flags |= MASTER_FLAG_THROTTLE;

status_fd：汇报子进程状态的管道。master模块要用到3组管道，这里的status_fd用来汇报子进程状态。在/master/master_status.c/master_status_init中构建，另外的两组管道是master_sig_pipe和master_flow_pipe，分别用来报告信号和收信速度。在/master/master_sig.c/master_sigsetup函数

和/master/master_flow.c/master_flow_init函数中被初始化。

children： 用BINHASH哈希链表来记录和管理各个子进程信息。

next：组织MASTER_SERV链表。