Linux Review COMS W4118 Spring 2008 Linux Overview History Distributions Licensing Components Kernel, daemons, libraries, utilities, etc Modules Build Process 2 Core Kernel Applications System Libraries (libc) Modules System Call Interface I/O Related File Systems Networking Process Related Scheduler Memory Management Device Drivers IPC Architecture-Dependent Code Hardware 3 System Calls System calls vs. libraries How to implement (“int 80x”) Syscall interface Trapping into the kernel Dispatch table / jump table Passing parameters Accessing user space Returning values 4 Invoking System Calls user-mode (restricted privileges) … xyz() … app making system call call wrapper routine in std C library kernel-mode (unrestricted privileges) sys_xyz() { … } ret xyz { … int 0x80; … } call int 0x80 iret system call service routine ret system_call: … sys_xyz(); … system call handler 5 Process Address Space 4 GB Privilege-level 0 Kernel space kernel-mode stack 3 GB User-mode stack-area User space Privilege-level 3 Shared runtime-libraries Task’s code and data 0 GB 6 Linux Processes/Tasks Processes/tasks The process descriptor: task_struct Thread context Task States Process relationships Wait queues Kernel threads Context switching Creating processes Destroying processes 7 Linux: Processes or Threads? Linux uses a neutral term: tasks Linux view Tasks represent both processes and threads Threads: processes that share address space Linux "threads" (tasks) are really "kernel threads“ Lighter-weight than traditional processes File descriptors, VM mappings need not be copied Implication: file table and VM table not part of process descriptor 8 The Linux process descriptor pagedir[] task_struct Each process descriptor contains many fields and some are pointers to other kernel structures state *stack mm_struct flags *pgd *mm user_struct exit_code *user pid which may themselves include fields that point to structures files_struct *files *parent *signal signal_struct 9 The Task Structure The task_struct is used to represent a task. The task_struct has several sub-structures that it references: tty_struct – TTY associated with the process fs_struct – current and root directories associated with the process files_struct – file descriptors for the process mm_struct – memory areas for the process signal_struct – signal structures associated with the process user_struct – per-user information (for example, number of current processes) Task States From kernel-header: <linux/sched.h> #define TASK_RUNNING #define TASK_INTERRUPTIBLE #define TASK_UNINTERRUPTIBLE #define TASK_STOPPED #define TASK_TRACED #define EXIT_ZOMBIE #define EXIT_DEAD #define TASK_NONINTERACTIVE64 #define TASK_DEAD 0 1 2 4 8 16 32 128 11 Task List, Run Queue init_task list run_queue Those tasks that are ready-to-run comprise a sub-list of all the tasks, and they are arranged on a queue known as the ‘run-queue’ Those tasks that are blocked while awaiting a specific event to occur are put on alternative sub-lists, called ‘wait queues’, associated with the particular event(s) that will allow a blocked task to be unblocked 12 Kernel Wait Queues waitqueue wait_queue_head_t can have 0 or more wait_queue_t chained onto them waitqueue However, usually just one element wait_queue_t wait_queue_head_t waitqueue Each wait_queue_t contains a list_head of tasks waitqueue All processes waiting for specific "event“ Used for timing, synch, device i/o,13etc. How Do I Block? By calling one of the sleep_on functions: These functions create a wait_queue and place the calling task on it Modify the value of its ‘state’ variable: sleep_on, interruptible_sleep_on, sleep_on_timeout, etc. TASK_UNINTERRUPTIBLE TASK_INTERRUPTIBLE Then call schedule or schedule_timeout The next task to run calls deactivate_task to move us out of the run queue Only tasks with ‘state == TASK_RUNNING’ are granted time on the CPU by the scheduler 14 How Do I Wake Up? By someone calling one of the wake functions: These functions call the curr->func function to wake up the task wake_up, wake_up_all, wake_up_interruptible, etc. Defaults to default_wake_function which is try_to_wake_up try_to_wake_up calls activate_task to move us out of the run queue The ‘state’ variable is set to TASK_RUNNING Sooner or later the scheduler will run us again We then return from schedule or schedule_timeout 15 Wait Queue Options INTERUPTIBLE vs. NON-INTERUPTIBLE: TIMEOUT vs no timeout: Can the task be woken up by a signal? Wake up the task after some timeout interval EXCLUSIVE vs. NON-EXCLUSIVE: Should only one task be woken up? Only one EXCLUSIVE task is woken up Kept at end of the list All NON-EXCLUSIVE tasks are woken up Kept at head of the list Functions with _nr option wake up number of tasks 16 Context Switching Context switching is the process of saving the state of the currently running task and loading the state of the next task to run. This involves saving the task's CPU state (registers), changing the current task value, and loading the CPU state of the new task into the registers. schedule determines the next task to run, calls context_switch, which calls switch_mm to change the process address space, then calls switch_to to context switch to the new task. The Role of the Stack One process must save state where another can find it When the new state is loaded, the CPU is running another process -- the state is the process! The stack pointer determines most of the state Some of the registers are on the stack The stack pointer determines the location of thread_info, which also points to task struct Changing the stack pointer changes the process! 18 Context Switch: FP Registers On context switch: Hardware flag set: TS in cr0 Software flag TS_USEDFPU is cleared in task_struct If task uses floating point instruction and hardware flag is set: Hardware raises “device not available” exception (trap) Kernel restores floating point registers TS is cleared TS_USEDFPU is set in the task_struct for this process Any time it’s set, floating point registers are saved for that process at switch time (but not restored for the next) Bottom line: only done if needed; if only one process uses floating point, no save/restore needed 19 Process Creation and Deletion fork and clone system calls do_fork is kernel implementation: CLONE_VM - share address space CLONE_FS - share root and current working directories CLONE_FILES - share file descriptors CLONE_SIGHAND - share signal handlers CLONE_PARENT – share parent process ID CLONE_THREAD – create thread for process Exit and wait system calls, zombie processes Do_exit, release_task 20 Kernel Threads Linux has a small number of kernel threads that run continuously in the kernel (daemons) No user address space Only execute code and access data in kernel address space How to create: kernel_thread Scheduled in the same way as other threads/tasks Process 0: idle process Process 1: init process Spawns several kernel threads before transitioning to user mode as /sbin/init kflushd (bdflush) – Flush dirty buffers to disk under "memory pressure" kupdate – Periodically flushes old buffers to disk kswapd – Swapping daemon 21 Scheduling Philosophies Priority is the primary scheduling mechanism Priority is dynamically adjusted at run time Processes denied access to CPU get increased Processes running a long time get decreased Try to distinguish interactive processes from noninteractive Bonus or penalty reflecting whether I/O or compute bound Use large quanta for important processes Modify quanta based on CPU use Quantum != clock tick Associate processes to CPUs Do everything in O(1) time 22 Runqueue for O(1) Scheduler priority array priority queue active . . . . . . priority queue expired Higher priority more I/O 800ms quanta lower priority more CPU 10ms quanta priority array priority queue . . . . . . priority queue 23 Basic Scheduling Algorithm Find the highest-priority queue with a runnable process Find the first process on that queue Calculate its quantum size Let it run When its time is up, put it on the expired list Repeat 24 Scheduling Components Static Priority Sleep Average Bonus Interactivity Status Dynamic Priority 25 Time Slice based on Priority Priority: Static Pri Niceness Quantum Highest 100 -20 800 ms High 110 -10 600 ms Normal 120 0 100 ms Low 130 10 50 ms Lowest 139 20 5 ms 26 Priority Array Swapping The system only runs processes from active queues, and puts them on expired queues when they use up their quanta When a priority level of the active queue is empty, the scheduler looks for the next-highest priority queue After running all of the active queues, the active and expired queues are swapped There are pointers to the current arrays; at the end of a cycle, the pointers are switched 27 Real-Time Scheduling Linux has soft real-time scheduling No hard real-time guarantees All real-time processes are higher priority than any conventional processes Processes with priorities [0, 99] are real-time First-in, first-out: SCHED_FIFO Static priority Process is only preempted for a higher-priority process No time quanta; it runs until it blocks or yields voluntarily RR within same priority level Round-robin: SCHED_RR As above but with a time quanta (800 ms) Normal processes have SCHED_OTHER scheduling policy 28 Multiprocessor Scheduling Each processor has a separate run queue Each processor only selects processes from its own queue to run Yes, it’s possible for one processor to be idle while others have jobs waiting in their run queues Periodically, the queues are rebalanced: if one processor’s run queue is too long, some processes are moved from it to another processor’s queue 29 Processor Affinity Each process has a bitmask saying what CPUs it can run on Normally, of course, all CPUs are listed Processes can change the mask The mask is inherited by child processes (and threads), thus tending to keep them on the same CPU Rebalancing does not override affinity 30