On Fri, Aug 19, 2005 at 20:19:39 +0000, Vincenzo Mallozzi wrote: > In Linux a thread is viewed as a common process, different from the others. > I means that each thread have a different pid. The way you can recognize > threads is by controlling the memory descriptor. > For example, if you want to know if a process is a thread, you have to control > if its memory descriptor is shared with TWO other processes in the system > having TWO different pids. To better explain this concept, I'll use an > example. > > Suppose we are executing a process with pid 100. At any point, this process > create a thread. > The system creates a new process sharing resource with process 100. (for sake > of simplicity, I suppose its pid is 101). The created thread is not the one > created by process 100 but is the so called THREAD MANAGER. The thread > manager creates a new thread (pid 102) with which shares the meomry > descriptor and other resources. > In other word, we have this situation: > > 100 (the process initially executing) > | > | > 101 (the thread manager) > | > | > 102 (the thread created) > > Now, suppose that the process 100 creates three threads. The situation > changes: > > > 100 (the process initially executing) > | > | > 101 (the thread manager) > / | \ > / | \ > 102 103 104 (the threads created) > > As you can see, threads 102 103 and 104 are sons of 101 while 101 is son of > 100. Also, all the processes (from 100 to 104) share the same descriptor. > Furthermore, the threads share also other resources, like terminal, files and > so on (but they can create new file, access new tty, ....) > The thread manager is created to better manage some tasks directed to all of > threads created. An example is the signals directed to the thread. > I ignore the way it manages the signal, perhaps it redirect a signal to the > correct thread or to all of them. But, I'm really sure that a thread can > receive a signal that the other threads don't receive. This is due, perhaps, > to the different pid assigned to them. > > I hope I've explained clearly it. I believe it's actually the initial task, that is turned into the thread manager. So it'd really be: 100 (the thread manager) +-------+---+---+-------+ | | | | 101 102 103 104 Where the first thread_create, invoked from task 100, actually returns to task 101 and 102 is set to execute the function passed to task_create. Because IIRC the "initial task" is required to be equal to all the other threads created. ------------------------------------------------------------------------------- Jan 'Bulb' Hudec <bulb@xxxxxx>
Attachment:
signature.asc
Description: Digital signature
