Topic : Multi-Threaded Programming
Author : LUPG
Page : << Previous 4  Next >>
Go to page :

to alert the programmer, at least during debug phase of the program, of such a case. It might mean nothing, but it might be significant.

A Real Condition For A Condition Variable

A note should be taken about condition variables - they are usually pointless without some real condition checking combined with them. To make this clear, lets consider the server example we introduced earlier. Assume that we use the 'got_request' condition variable to signal that a new request has arrived that needs handling, and is held in some requests queue. If we had threads waiting on the condition variable when this variable is signaled, we are assured that one of these threads will awake and handle this request.

However, what if all threads are busy handling previous requests, when a new one arrives? the signaling of the condition variable will do nothing (since all threads are busy doing other things, NOT waiting on the condition variable now), and after all threads finish handling their current request, they come back to wait on the variable, which won't necessarily be signaled again (for example, if no new requests arrive). Thus, there is at least one request pending, while all handling threads are blocked, waiting for a signal.

In order to overcome this problem, we may set some integer variable to denote the number of pending requests, and have each thread check the value of this variable before waiting on the variable. If this variable's value is positive, some request is pending, and the thread should go and handle it, instead of going to sleep. Further more, a thread that handled a request, should reduce the value of this variable by one, to make the count correct.
Lets see how this affects the waiting code we have seen above.

/* number of pending requests, initially none */
int num_requests = 0;
/* first, lock the mutex */
int rc = pthread_mutex_lock(&a_mutex);
if (rc) { /* an error has occurred */
/* mutex is now locked - wait on the condition variable */
/* if there are no requests to be handled.              */
rc = 0;
if (num_requests == 0)
    rc = pthread_cond_wait(&got_request, &request_mutex);
if (num_requests > 0 && rc == 0) { /* we have a request pending */
        /* do your stuff... */
        /* decrease count of pending requests */
/* finally, unlock the mutex */

Using A Condition Variable - A Complete Example

As an example for the actual usage of condition variables, we will show a program that simulates the server we have described earlier - one thread, the receiver, gets client requests. It inserts the requests to a linked list, and a hoard of threads, the handlers, are handling these requests. For simplicity, in our simulation, the receiver thread creates requests and does not read them from real clients.

The program source is available in the file thread-pool-server.c, and contains many comments. Please read the source file first, and then read the following clarifying notes.

1. The 'main' function first launches the handler threads, and then performs the chord of the receiver thread, via its main loop.
2. A single mutex is used both to protect the condition variable, and to protect the linked list of waiting requests. This simplifies the design. As an exercise, you may think how to divide these roles into two mutexes.
3. The mutex itself MUST be a recursive mutex. In order to see why, look at the code of the 'handle_requests_loop' function. You will notice that it first locks the mutex, and afterwards calls the 'get_request' function, which locks the mutex again. If we used a non-recursive mutex, we'd get locked indefinitely in the mutex locking operation of the 'get_request' function.
You may argue that we could remove the mutex locking in the 'get_request' function, and thus remove the double-locking problem, but this is a flawed design - in a larger program, we might call the 'get_request' function from other places in the code, and we'll need to check for proper locking of the mutex in each of them.
4. As a rule, when using recursive mutexes, we should try to make sure that each lock operation is accompanied by a matching unlock operation in the same function. Otherwise, it will be very hard to make sure that after locking the mutex several times, it is being unlocked the same number of times, and deadlocks would occur.
5. The implicit unlocking and re-locking of the mutex on the call to the pthread_cond_wait() function is confusing at first. It is best to add a comment regarding this behavior in the code, or else someone that reads this code might accidentally add a further mutex lock.
6. When a handler thread handles a reuqest - it should free the mutex, to avoid blocking all the other handler threads. After it finished handling the request, it sohuld lock the mutex again, and check if there are moer requests to handle.

"Private" thread data - Thread-Specific Data

In "normal", single-thread programs, we sometimes find the need to use a global variable. Ok, so good old teach' told us it is bad practice to have global variables, but they sometimes do come handy. Especially if they are static variables - meaning, they are recognized only on the scope of a single file.

In multi-threaded programs, we also might find a need for such variables. We should note, however, that the same variable is accessible from all the threads, so we need to protect access to it using a mutex, which is extra overhead. Further more, we sometimes need to have a variable that is 'global', but only for a specific thread. Or the same 'global' variable should have different values in different threads. For example, consider a program that needs to have one globally accessible linked list in each thread, but note the same list. Further, we want the same code to be executed by all threads. In this case, the global pointer to the start of the list should be point to a different address in each thread.

In order to have such a pointer, we need a mechanism that enables the same global variable to have a different location in memory. This is what the thread-specific data mechanism is used for.

Overview Of Thread-Specific Data Support

In the thread-specific data (TSD) mechanism, we have notions of keys and values. Each key has a name, and pointer to some memory area. Keys with the same name in two separate threads always point to different memory locations - this is handled by the library functions that allocate memory blocks to be accessed via these keys. We have a function to create a key (invoked once per key name for the whole process), a function to allocate memory (invoked separately in each thread), and functions to de-allocate this memory for a specific thread, and a function to destroy the key, again, process-wide. we also have functions to access the data pointed to by a key, either setting its value, or returning the value it points to.

Allocating Thread-Specific Data Block

The pthread_key_create() function is used to allocate a new key. This key now becomes valid for all threads in our process. When a key is created, the value it points to defaults to NULL. Later on each thread may change its copy of the value as it wishes. Here is how to use this function:

/* rc is used to contain return values of pthread functions */
int rc;
/* define a variable to hold the key, once created.         */
pthread_key_t list_key;
/* cleanup_list is a function that can clean up some data   */
/* it is specific to our program, not to TSD                */
extern void* cleanup_list(void*);

/* create the key, supplying a function that'll be invoked when it's deleted. */
rc = pthread_key_create(&list_key, cleanup_list);

Some notes:
1. After pthread_key_create() returns, the variable 'list_key' points to the newly created key.
2. The function pointer passed as second parameter to pthread_key_create(), will be automatically invoked by the pthread library when our thread exits, with a pointer to the key's value as its parameter. We may supply a NULL pointer as the function pointer, and then no function will be invoked for key. Note that the function will be invoked once in each thread, even thought we created this key only once, in one thread.
If we created several keys, their associated destructor functions will be called in an arbitrary order, regardless of the order of keys creation.
3. If the pthread_key_create() function succeeds, it returns

Page : << Previous 4  Next >>