P2L4 Thread Design Considerations - Tasks in Linux

Question 1

Define a ‘task’

Answer 1

• An abstraction that represents an execution context
• The execution context of a kernel level thread

Question 2

A single-threaded process will have _____ task, and a multithreaded process will have _____ tasks.

Answer 2

A single-threaded process will have one task, and a multithreaded process will have many tasks.

Question 3

A task is identified by its ____________

If we have a single threaded process the id of the task and the id of the process will______________

If we have a multithreaded process, each task will have a _________ pid.

If we have a multithreaded process, the process as a whole will be identified by ___________________________

Answer 3

A task is identified by its pid_t pid.

If we have a single threaded process the id of the task and the id of the process will be the same.

If we have a multithreaded process, each task will have a different pid.

If we have a multithreaded process, the process as a whole will be identified by the pid of the first task that was created.

Question 4

True or False?

Linux has one continuous process control block.

Answer 4

False

Linux never had one continuous process control block. Instead, the process state was always maintained through a collection of data structures that pointed to each other. We can see some of the references in the task in struct mm_struct *mm and struct files_struct *files.

Question 5

Describe how Linux creates a new task

Answer 5

clone

To create a new task, Linux supports an operation called clone. It takes a function pointer and an argument (similar to pthread_create) but it also takes an argument sharing_flags which denotes which portion of the state of a task will be shared between the parent and child task.

Question 6

What are the arguments to the function clone that creates a new task?

Answer 6

• function
• stack_ptr
• sharing_flags - which portion of the state of a task will be shared between the parent and child task.

When all set –> like creating a thread - sharing everything

When none set –> like forking a new process - sharing almost nothing

• args

Question 7

True or False

‘fork’ in Linux is basically implemented by clone with all sharing flags cleared.

Answer 7

True

Question 8

The native implementation of threads in Linux is the Native POSIX Threads Library (NPTL). This is a ______ model, meaning that there is a _______ level task for each ____ level thread.

This implementation replaced an earlier implementation LinuxThreads, which was a __________ model.

Answer 8

The native implementation of threads in Linux is the Native POSIX Threads Library (NPTL). This is a ______ model, meaning that there is a kernel level task for each user level thread. This implementation replaced an earlier implementation LinuxThreads, which was a many-to-many model.

Question 9

A multi-threaded process forks a new process. Will this process be single or multi-threaded?

Answer 9

Single threaded process forks a new process -> copy with same address space

Multi-threaded process forks a new process -> single threaded - will only contain a portion of the parent’s address space