Linux C notes (BOOK UNIX)

Chapter 1: Thread

① thread functions

1. pthread_self();
    return current thread id
2. thread existing
    pthread_exit(void * rval_ptr);
    you can get the value of rval_ptr just by  pthread_join(pthread_t tht, void ** rval_ptr)
    pthread_cancel(pthread_t thd);
    cancel a thread as the thread exits with PTHREAD_CANCELED in pthread_exit, this function        only send a signal but not wait until the thread really exits
3. Control a thread
        We can give a function to a thread with which the thread can invoke it when it exits. This
    practice is like atexit in process functions.
         pthread_clean_push
         pthread_clean_pop
        These two functions must be in pairs in your function routines but not in the main                function. And one thing should be mentioned as : the routine should not be returned but
    exit with pthread_exit(arg);
        pthread_detach(pthread_t tid)
        Let a thread be detached
4. Thread synchronization of thread

② thread. synchronics
    1. Mutex

        i.initiation
        pthread_mutex_t variable
        pthread_mutex_init()  this must be followed with destroy function
        pthread_mutex_destroy()
       
        ii. Lock
        pthread_mutex_lock()  if the mutex is locked wait until ...
        pthread_mutex_trylock() return instantly
        pthread_unlock()
    2. WLock

        i. initiation
        pthread_rwlook_t
        pthread_rwlock_init
        pthread_rwlock_destroy
        
        ii. Lock
        pthread_rwlock_rdlock
        pthread_rwlock_wrlook
        pthread_rwlock_unlock
        pthread_rwlock_tryrdlock
        pthread_rwlock_trywrlock        

    3. Cond
    
        i. initiation
        pthread_cond_t pcndt;       
        pthread_cond_init(&pcndt, attr);
        pthread_cond_destroy(&pcndt);       
        
        ii. Wait
        pthread_cond_wait(&pcndt, pthread_mutex_t &mt);       
        pthread_cond_timedwait(....)

        pthread_cond_signal(...)  evoke a thread waiting for the condition
        pthread_cond_broadcast(...) evoke all threads which are waiting for the condition

        example:
          before invoking the cond wait function, it's necessary to lock the mutex...so after the            invoking of wait, the mutex should be released


③  Thread controling
   
    1. Properties of Thread
    
        i. initiation
            pthread_attr_t structure
            pthread_attr_init     initiate the attr as default value
            pthread_attr_destroy
            Every thread has its attributes. like attach states which indicates whether the main              thread  should pay attention to the child thread.
            so.
            pthread_attr_getdetachstate(...)
            pthread_attr_setdetachstate(attr, stat).
            stat is either PTHREAD_CREATE_DETACHED or PTHREAD_CREATE_JOINABLE
    2. Re-Enter
    3. Thread Private Data
        i. initiation
             pthread_key_t thd_ky_t
            pthread_key_create(pthread_key_t,(void(*destructor)))
            

Chapter 2 : signals

①  Register a signal

    1. 
        void * signal(int signo, void (*func)(int));
        To ignore a signal to use
        signal(signo, SIG_IGN) if it returns SIG_IGN , it indicates the signo has been ignored.
    

    2.
        kill(pid, signo);
        raise(signo);
        raise(signo) is equal to kill(getpid(), signo);
        
        pid > 0: send signo to pid process
        pid == 0: send signo to all processes who have the same group ids as sender has
        pid < 0: send signo to processes who have the gpid as abs(pid)]
        pid == -1: send signo to all processes which the sender has privilage to send to

            if the signo is 0, the signal is taken as an empty signal, so signal is often set 0 to test            whether a process still exists.

    3. alarm and pause SIGALRM
        alarm(unsigned int seconds);
            One process only has one alarm. If a alarm is called after another, the first is replace,             the remained will be returned in the return value of the second call.
            If the seconds is 0, the former alarm is canceled.

        pause()
        The process is suspended util a signal received

        setjmp( jmp_buf   jmpbuf );
        longjmp(mark, no);
            The no is the return value of setjmp...
    4. signal sets
        sigset_t *set;
        sigemptyset()
        sigfillset()
        sigadset()
        sigdelset()
        sigismember()
       
        sigprocmask(int how, const sigset_t * set, sigset_t * oset);
         oset : return value showing current signal set
        how  
                SIG_BLOCK, add signal (intersection)
                SIG_UNBLOCK, delete signal (union)
                SIG_SETMASK, replace current signal set (replacement)

        sigpending()
            return signals registered pending and unhandled
       
        sigaction act;
        sigaction(int signo, const sigaction * newact, struct sigaction * oldact);
            replacement of signal()
            mind the sa_flags
            If it is set : AS_SIGINFO the handle should not be set default, instead
            void handler( int signo, siginfo_t * info, void * context)
            Default handle is as following :
            void handler(ing signo);

        sigsuspend()
            The process suspended until a registered signal comes and then revert the current                  signal set to old set


Chapter 3  Advanced I/O

① Blocking I/O

    1. Record Locking
        This practice can lock partly a file.
        fcntl(int fileds, int cmd, struct flock * flkptr);          
        cmd:
            F_GETLK
                testing function, testing the lockptr whether can be set, if undoable write old lock                  into ptr, if doable change the l_type into F_UNLCK
            F_SETLK
                if the current situation is unlockable return instantly. errno is EACCES or EAGAIN
            F_SETLKW
                this is blocking parameter.... of SETLK..if the area is unblockable, the caller process
              pause until a signal pushed
        locktype:
            F_RDLCK
            F_WRLCK
            F_UNCLK
        Inner a process, if a lock exists, a new lock can replace the old lock.

        How to test a lock:
        {
            struct flock lock;
            lock.l_type = type;
            lock.l_start = offset;
            lock.l_whence = whence;
            lock.l_len = len;
            if(fcntl(fd, F_GETLCK,&lock) < 0)
            {
                //error
            }
            if(lock.l_type == F_UNLCK)
            {
                return false;
            }
        }
          Release a lock        
        {
            when the file handle is released, all locks binded to it will be automatically released too.
        }
        lock sharing:
        child process never inherit parents' lock

Chapter 4  Process Msg

①：pipe
    1. pipe
        semiduplex.
        
        pipe(int fileds[2]); 
            Two file descriptors, one for writering , one for reading
        the direction seems fileds[1] --> fileds[0], write(1), read(0)

    2. popen
        
        popen(cmdstring, type);
        pclose()
            cmdstring is a cmd like "ls -a" which runs in shell
            type is direction shows input or output
        note:
            this pipe has also single direction
    3. coprocess       
②：named pipe   
   
③：IPC   
     (msg queue, semaphore , shared memory)
   
    1. queue

        msgget(key_t key, int flag);
        This function creates or retrieve a msg from the msg queue.
        If msgflg specifies both IPC_CREAT and IPC_EXCL and a message queue already exists          for key, then msgget() fails with errno set to EEXIST
        A new message queue is created if key has the value IPC_PRIVATE .
        And if the key is not IPC_PRIVATE  but no queue with the given key exists, new queue           will be created .
        If succeeds, return queue id, vice versa.
       
        msgctl(int msqid, int cmd, struct msgqid_ds * buf);
            IPC_STAT
             retrieve the msgque stuctor into the buf
            IPC_SET
             change the structor related to the id
            IPC_RMID
                remove the msg from queue by given id

        msgsnd(int msqid, const void * ptr, size_t nbytes, int flag);
        msgrecv(int msqid, const void * ptr, size_t nbytes, int flag);
            Mind the ptr, ptr is a structure as following
            struct msgbuf
            {
                long mtype;
                 /* message type, must be > 0 */
                char mtext[n];
                 /* message data */
            };
            This size of mtext is specified by nbytes..
            If flag is set IPC_NOWAIT it functions nonblockingly
            
            as for msgrecv, when received msg is longer than nbytes, if flag is set MSG_NOERROR,          the msg will be truncated, but if no flag set MSG_NOERROR, error




            

Chapter 5 SOCKET Summaries

Chapter 6 File And Directory
    
① Dir Operation

    #include
    #include
    #include
    #include
    int main()
    {
        DIR * pDir = NULL;
        struct dirent * pDirent = NULL;    
        pDir = opendir("./pcapfiles/");
        while((pDirent = readdir(pDir)) != NULL)
        {
            char szPath[1024];
            struct stat buf;
            sprintf(szPath, "./pcapfiles/%s",pDirent->d_name);
             stat (szPath,&buf);
            if( S_ISDIR(buf.st_mode))
            {
                 printf("d_name: %sn",pDirent->d_name);
            }
        }
    }   
    #NOTE this example lists all files and dirs in a given dir, mind the macro S_ISDIR can judge a     #path is a dir or a file