minix/kernel/clock.c
Tomas Hruby b4cf88a04f Userspace scheduling
- cotributed by Bjorn Swift

- In this first phase, scheduling is moved from the kernel to the PM
  server. The next steps are to a) moving scheduling to its own server
  and b) include useful information in the "out of quantum" message,
  so that the scheduler can make use of this information.

- The kernel process table now keeps record of who is responsible for
  scheduling each process (p_scheduler). When this pointer is NULL,
  the process will be scheduled by the kernel. If such a process runs
  out of quantum, the kernel will simply renew its quantum an requeue
  it.

- When PM loads, it will take over scheduling of all running
  processes, except system processes, using sys_schedctl().
  Essentially, this only results in taking over init. As children
  inherit a scheduler from their parent, user space programs forked by
  init will inherit PM (for now) as their scheduler.

 - Once a process has been assigned a scheduler, and runs out of
   quantum, its RTS_NO_QUANTUM flag will be set and the process
   dequeued. The kernel will send a message to the scheduler, on the
   process' behalf, informing the scheduler that it has run out of
   quantum. The scheduler can take what ever action it pleases, based
   on its policy, and then reschedule the process using the
   sys_schedule() system call.

- Balance queues does not work as before. While the old in-kernel
  function used to renew the quantum of processes in the highest
  priority run queue, the user-space implementation only acts on
  processes that have been bumped down to a lower priority queue.
  This approach reacts slower to changes than the old one, but saves
  us sending a sys_schedule message for each process every time we
  balance the queues. Currently, when processes are moved up a
  priority queue, their quantum is also renewed, but this can be
  fiddled with.

- do_nice has been removed from kernel. PM answers to get- and
  setpriority calls, updates it's own nice variable as well as the
  max_run_queue. This will be refactored once scheduling is moved to a
  separate server. We will probably have PM update it's local nice
  value and then send a message to whoever is scheduling the process.

- changes to fix an issue in do_fork() where processes could run out
  of quantum but bypassing the code path that handles it correctly.
  The future plan is to remove the policy from do_fork() and implement
  it in userspace too.
2010-03-29 11:07:20 +00:00

255 lines
8.2 KiB
C

/* This file contains the clock task, which handles time related functions.
* Important events that are handled by the CLOCK include setting and
* monitoring alarm timers and deciding when to (re)schedule processes.
* The CLOCK offers a direct interface to kernel processes. System services
* can access its services through system calls, such as sys_setalarm(). The
* CLOCK task thus is hidden from the outside world.
*
* Changes:
* Aug 18, 2006 removed direct hardware access etc, MinixPPC (Ingmar Alting)
* Oct 08, 2005 reordering and comment editing (A. S. Woodhull)
* Mar 18, 2004 clock interface moved to SYSTEM task (Jorrit N. Herder)
* Sep 30, 2004 source code documentation updated (Jorrit N. Herder)
* Sep 24, 2004 redesigned alarm timers (Jorrit N. Herder)
*
* Clock task is notified by the clock's interrupt handler when a timer
* has expired.
*
* In addition to the main clock_task() entry point, which starts the main
* loop, there are several other minor entry points:
* clock_stop: called just before MINIX shutdown
* get_uptime: get realtime since boot in clock ticks
* set_timer: set a watchdog timer (+)
* reset_timer: reset a watchdog timer (+)
* read_clock: read the counter of channel 0 of the 8253A timer
*
* (+) The CLOCK task keeps tracks of watchdog timers for the entire kernel.
* It is crucial that watchdog functions not block, or the CLOCK task may
* be blocked. Do not send() a message when the receiver is not expecting it.
* Instead, notify(), which always returns, should be used.
*/
#include "kernel.h"
#include "proc.h"
#include <minix/endpoint.h>
#include <assert.h>
#include "clock.h"
#ifdef CONFIG_WATCHDOG
#include "watchdog.h"
#endif
/* Function prototype for PRIVATE functions.
*/
FORWARD _PROTOTYPE( void load_update, (void));
/* The CLOCK's timers queue. The functions in <timers.h> operate on this.
* Each system process possesses a single synchronous alarm timer. If other
* kernel parts want to use additional timers, they must declare their own
* persistent (static) timer structure, which can be passed to the clock
* via (re)set_timer().
* When a timer expires its watchdog function is run by the CLOCK task.
*/
PRIVATE timer_t *clock_timers; /* queue of CLOCK timers */
PRIVATE clock_t next_timeout; /* realtime that next timer expires */
/* The time is incremented by the interrupt handler on each clock tick.
*/
PRIVATE clock_t realtime = 0; /* real time clock */
/*
* The boot processor timer interrupt handler. In addition to non-boot cpus it
* keeps real time and notifies the clock task if need be
*/
PUBLIC int bsp_timer_int_handler(void)
{
unsigned ticks;
if(minix_panicing)
return 0;
/* Get number of ticks and update realtime. */
ticks = lost_ticks + 1;
lost_ticks = 0;
realtime += ticks;
ap_timer_int_handler();
assert(!proc_is_runnable(proc_ptr) || proc_ptr->p_ticks_left > 0);
/* if a timer expired, notify the clock task */
if ((next_timeout <= realtime)) {
tmrs_exptimers(&clock_timers, realtime, NULL);
next_timeout = (clock_timers == NULL) ?
TMR_NEVER : clock_timers->tmr_exp_time;
}
if (do_serial_debug)
do_ser_debug();
return(1); /* reenable interrupts */
}
/*===========================================================================*
* get_uptime *
*===========================================================================*/
PUBLIC clock_t get_uptime(void)
{
/* Get and return the current clock uptime in ticks. */
return(realtime);
}
/*===========================================================================*
* set_timer *
*===========================================================================*/
PUBLIC void set_timer(tp, exp_time, watchdog)
struct timer *tp; /* pointer to timer structure */
clock_t exp_time; /* expiration realtime */
tmr_func_t watchdog; /* watchdog to be called */
{
/* Insert the new timer in the active timers list. Always update the
* next timeout time by setting it to the front of the active list.
*/
tmrs_settimer(&clock_timers, tp, exp_time, watchdog, NULL);
next_timeout = clock_timers->tmr_exp_time;
}
/*===========================================================================*
* reset_timer *
*===========================================================================*/
PUBLIC void reset_timer(tp)
struct timer *tp; /* pointer to timer structure */
{
/* The timer pointed to by 'tp' is no longer needed. Remove it from both the
* active and expired lists. Always update the next timeout time by setting
* it to the front of the active list.
*/
tmrs_clrtimer(&clock_timers, tp, NULL);
next_timeout = (clock_timers == NULL) ?
TMR_NEVER : clock_timers->tmr_exp_time;
}
/*===========================================================================*
* load_update *
*===========================================================================*/
PRIVATE void load_update(void)
{
u16_t slot;
int enqueued = 0, q;
struct proc *p;
/* Load average data is stored as a list of numbers in a circular
* buffer. Each slot accumulates _LOAD_UNIT_SECS of samples of
* the number of runnable processes. Computations can then
* be made of the load average over variable periods, in the
* user library (see getloadavg(3)).
*/
slot = (realtime / system_hz / _LOAD_UNIT_SECS) % _LOAD_HISTORY;
if(slot != kloadinfo.proc_last_slot) {
kloadinfo.proc_load_history[slot] = 0;
kloadinfo.proc_last_slot = slot;
}
/* Cumulation. How many processes are ready now? */
for(q = 0; q < NR_SCHED_QUEUES; q++)
for(p = rdy_head[q]; p; p = p->p_nextready)
enqueued++;
kloadinfo.proc_load_history[slot] += enqueued;
/* Up-to-dateness. */
kloadinfo.last_clock = realtime;
}
/*
* Timer interupt handler. This is the only thing executed on non boot
* processors. It is called by bsp_timer_int_handler() on the boot processor
*/
PUBLIC int ap_timer_int_handler(void)
{
/* Update user and system accounting times. Charge the current process
* for user time. If the current process is not billable, that is, if a
* non-user process is running, charge the billable process for system
* time as well. Thus the unbillable process' user time is the billable
* user's system time.
*/
const unsigned ticks = 1;
struct proc * p, * billp;
#ifdef CONFIG_WATCHDOG
/*
* we need to know whether local timer ticks are happening or whether
* the kernel is locked up. We don't care about overflows as we only
* need to know that it's still ticking or not
*/
watchdog_local_timer_ticks++;
#endif
/* Update user and system accounting times. Charge the current process
* for user time. If the current process is not billable, that is, if a
* non-user process is running, charge the billable process for system
* time as well. Thus the unbillable process' user time is the billable
* user's system time.
*/
/* FIXME prepared for get_cpu_local_var() */
p = proc_ptr;
billp = bill_ptr;
p->p_user_time += ticks;
if (priv(p)->s_flags & PREEMPTIBLE) {
p->p_ticks_left -= ticks;
}
if (! (priv(p)->s_flags & BILLABLE)) {
billp->p_sys_time += ticks;
}
/* Decrement virtual timers, if applicable. We decrement both the
* virtual and the profile timer of the current process, and if the
* current process is not billable, the timer of the billed process as
* well. If any of the timers expire, do_clocktick() will send out
* signals.
*/
if ((p->p_misc_flags & MF_VIRT_TIMER)){
p->p_virt_left -= ticks;
}
if ((p->p_misc_flags & MF_PROF_TIMER)){
p->p_prof_left -= ticks;
}
if (! (priv(p)->s_flags & BILLABLE) &&
(billp->p_misc_flags & MF_PROF_TIMER)){
billp->p_prof_left -= ticks;
}
/*
* Check if a process-virtual timer expired. Check current process, but
* also bill_ptr - one process's user time is another's system time, and
* the profile timer decreases for both!
*/
vtimer_check(p);
if (p != billp)
vtimer_check(billp);
/* Update load average. */
load_update();
/* check if the processes still have some ticks left */
check_ticks_left(p);
return 1;
}
PUBLIC int boot_cpu_init_timer(unsigned freq)
{
if (arch_init_local_timer(freq))
return -1;
if (arch_register_local_timer_handler(
(irq_handler_t) bsp_timer_int_handler))
return -1;
return 0;
}