Introduction to Linux Threads – Part I
A thread of execution is often regarded as the smallest unit of processing that a scheduler works on. A process can have multiple threads of execution which are executed asynchronously. This asynchronous execution brings in the capability of each thread handling a particular work or service independently. Hence multiple threads running in a process handle their services which overall constitutes the complete capability of the process.
In this article we will touch base on the fundamentals of threads and build the basic understanding required to learn the practical aspects of Linux threads. Linux Threads Series: part 1 (this article), part 2, part 3.
Why Threads are Required?
Now, one would ask why do we need multiple threads in a process?? Why can’t a process with only one (default) main thread be used in every situation. Well, to answer this lets consider an example : Suppose there is a process, that receiving real time inputs and corresponding to each input it has to produce a certain output. Now, if the process is not multi-threaded ie if the process does not involve multiple threads, then the whole processing in the process becomes synchronous. This means that the process takes an input processes it and produces an output.
The limitation in the above design is that the process cannot accept an input until its done processing the earlier one and in case processing an input takes longer than expected then accepting further inputs goes on hold. To consider the impact of the above limitation, if we map the generic example above with a socket server process that can accept input connection, process them and provide the socket client with output. Now, if in processing any input if the server process takes more than expected time and in the meantime another input (connection request) comes to the socket server then the server process would not be able to accept the new input connection as its already stuck in processing the old input connection. This may lead to a connection time out at the socket client which is not at all desired. This shows that synchronous model of execution cannot be applied everywhere and hence was the requirement of asynchronous model of execution felt which is implemented by using threads.
Difference Between threads and processes
- Processes do not share their address space while threads executing under same process share the address space.
- From the above point its clear that processes execute independent of each other and the synchronization between processes is taken care by kernel only while on the other hand the thread synchronization has to be taken care by the process under which the threads are executing
- Context switching between threads is fast as compared to context switching between processes
- The interaction between two processes is achieved only through the standard inter process communication while threads executing under the same process can communicate easily as they share most of the resources like memory, text segment etc
User threads Vs Kernel Threads
Threads can exist in user space as well as in kernel space.
A user space threads are created, controlled and destroyed using user space thread libraries. These threads are not known to kernel and hence kernel is nowhere involved in their processing. These threads follow co-operative multitasking where-in a thread releases CPU on its own wish ie the scheduler cannot preempt the thread. Th advantages of user space threads is that the switching between two threads does not involve much overhead and is generally very fast while on the negative side since these threads follow co-operative multitasking so if one thread gets block the whole process gets blocked.
A kernel space thread is created, controlled and destroyed by the kernel. For every thread that exists in user space there is a corresponding kernel thread. Since these threads are managed by kernel so they follow preemptive multitasking where-in the scheduler can preempt a thread in execution with a higher priority thread which is ready for execution. The major advantage of kernel threads is that even if one of the thread gets blocked the whole process is not blocked as kernel threads follow preemptive scheduling while on the negative side the context switch is not very fast as compared to user space threads.
If we talk of Linux then kernel threads are optimized to such an extent that they are considered better than user space threads and mostly used in all scenarios except where prime requirement is that of cooperative multitasking.
Problem with Threads
There are some major problems that arise while using threads :
- Many operating system does not implement threads as processes rather they see threads as part of parent process. In this case, what would happen if a thread calls fork() or even worse what if a thread execs a new binary?? These scenarios may have dangerous consequences for example in the later problem the whole parent process could get replaced with the address space of the newly exec’d binary. This is not at all desired. Linux which is POSIX complaint makes sure that calling a fork() duplicates only the thread that has called the fork() function while an exec from any of the thread would stop all the threads in the parent process.
- Another problem that may arise is the concurrency problems. Since threads share all the segments (except the stack segment) and can be preempted at any stage by the scheduler than any global variable or data structure that can be left in inconsistent state by preemption of one thread could cause severe problems when the next high priority thread executes the same function and uses the same variables or data structures.
For the problem 1 mentioned above, all we can say is that its a design issue and design for applications should be done in a way that least problems of this kind arise.
For the problem 2 mentioned above, using locking mechanisms programmer can lock a chunk of code inside a function so that even if a context switch happens (when the function global variable and data structures were in inconsistent state) then also next thread is not able to execute the same code until the locked code block inside the function is unlocked by the previous thread (or the thread that acquired it).