kernel/sched.c

/* */
This source file includes following definitions.
add_to_runqueue
del_from_runqueue
wake_up_process
process_timeout
schedule
sys_pause
wake_up
wake_up_interruptible
__down
__sleep_on
interruptible_sleep_on
sleep_on
add_timer
del_timer
count_active_tasks
calc_load
second_overflow
timer_bh
tqueue_bh
immediate_bh
do_timer
sys_alarm
sys_getpid
sys_getppid
sys_getuid
sys_geteuid
sys_getgid
sys_getegid
sys_nice
show_task
show_state
sched_init
   1 /*
   2  *  linux/kernel/sched.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6 
   7 /*
   8  * 'sched.c' is the main kernel file. It contains scheduling primitives
   9  * (sleep_on, wakeup, schedule etc) as well as a number of simple system
  10  * call functions (type getpid(), which just extracts a field from
  11  * current-task
  12  */
  13 
  14 #include <linux/config.h>
  15 #include <linux/signal.h>
  16 #include <linux/sched.h>
  17 #include <linux/timer.h>
  18 #include <linux/kernel.h>
  19 #include <linux/kernel_stat.h>
  20 #include <linux/fdreg.h>
  21 #include <linux/errno.h>
  22 #include <linux/time.h>
  23 #include <linux/ptrace.h>
  24 #include <linux/delay.h>
  25 #include <linux/interrupt.h>
  26 #include <linux/tqueue.h>
  27 #include <linux/resource.h>
  28 #include <linux/mm.h>
  29 #include <linux/smp.h>
  30 
  31 #include <asm/system.h>
  32 #include <asm/io.h>
  33 #include <asm/segment.h>
  34 #include <asm/pgtable.h>
  35 
  36 #include <linux/timex.h>
  37 
  38 /*
  39  * kernel variables
  40  */
  41 long tick = 1000000 / HZ;               /* timer interrupt period */
  42 volatile struct timeval xtime;          /* The current time */
  43 int tickadj = 500/HZ;                   /* microsecs */
  44 
  45 DECLARE_TASK_QUEUE(tq_timer);
  46 DECLARE_TASK_QUEUE(tq_immediate);
  47 DECLARE_TASK_QUEUE(tq_scheduler);
  48 
  49 /*
  50  * phase-lock loop variables
  51  */
  52 int time_state = TIME_BAD;     /* clock synchronization status */
  53 int time_status = STA_UNSYNC;   /* clock status bits */
  54 long time_offset = 0;           /* time adjustment (us) */
  55 long time_constant = 0;         /* pll time constant */
  56 long time_tolerance = MAXFREQ;  /* frequency tolerance (ppm) */
  57 long time_precision = 1;        /* clock precision (us) */
  58 long time_maxerror = 0x70000000;/* maximum error */
  59 long time_esterror = 0x70000000;/* estimated error */
  60 long time_phase = 0;            /* phase offset (scaled us) */
  61 long time_freq = 0;             /* frequency offset (scaled ppm) */
  62 long time_adj = 0;              /* tick adjust (scaled 1 / HZ) */
  63 long time_reftime = 0;          /* time at last adjustment (s) */
  64 
  65 long time_adjust = 0;
  66 long time_adjust_step = 0;
  67 
  68 int need_resched = 0;
  69 unsigned long event = 0;
  70 
  71 extern int _setitimer(int, struct itimerval *, struct itimerval *);
  72 unsigned long * prof_buffer = NULL;
  73 unsigned long prof_len = 0;
  74 unsigned long prof_shift = 0;
  75 
  76 #define _S(nr) (1<<((nr)-1))
  77 
  78 extern void mem_use(void);
  79 
  80 static unsigned long init_kernel_stack[1024] = { STACK_MAGIC, };
  81 unsigned long init_user_stack[1024] = { STACK_MAGIC, };
  82 static struct vm_area_struct init_mmap = INIT_MMAP;
  83 static struct fs_struct init_fs = INIT_FS;
  84 static struct files_struct init_files = INIT_FILES;
  85 static struct signal_struct init_signals = INIT_SIGNALS;
  86 
  87 struct mm_struct init_mm = INIT_MM;
  88 struct task_struct init_task = INIT_TASK;
  89 
  90 unsigned long volatile jiffies=0;
  91 
  92 struct task_struct *current_set[NR_CPUS];
  93 struct task_struct *last_task_used_math = NULL;
  94 
  95 struct task_struct * task[NR_TASKS] = {&init_task, };
  96 
  97 struct kernel_stat kstat = { 0 };
  98 
  99 static inline void add_to_runqueue(struct task_struct * p)
     /*  */
 100 {
 101 #if 1   /* sanity tests */
 102         if (p->next_run || p->prev_run) {
 103                 printk("task already on run-queue\n");
 104                 return;
 105         }
 106 #endif
 107         if (p->counter > current->counter + 3)
 108                 need_resched = 1;
 109         nr_running++;
 110         (p->next_run = init_task.next_run)->prev_run = p;
 111         p->prev_run = &init_task;
 112         init_task.next_run = p;
 113 }
 114 
 115 static inline void del_from_runqueue(struct task_struct * p)
     /*  */
 116 {
 117         struct task_struct *next = p->next_run;
 118         struct task_struct *prev = p->prev_run;
 119 
 120 #if 1   /* sanity tests */
 121         if (!next || !prev) {
 122                 printk("task not on run-queue\n");
 123                 return;
 124         }
 125 #endif
 126         if (p == &init_task) {
 127                 static int nr = 0;
 128                 if (nr < 5) {
 129                         nr++;
 130                         printk("idle task may not sleep\n");
 131                 }
 132                 return;
 133         }
 134         nr_running--;
 135         next->prev_run = prev;
 136         prev->next_run = next;
 137         p->next_run = NULL;
 138         p->prev_run = NULL;
 139 }
 140 
 141 /*
 142  * Wake up a process. Put it on the run-queue if it's not
 143  * already there.  The "current" process is always on the
 144  * run-queue (except when the actual re-schedule is in
 145  * progress), and as such you're allowed to do the simpler
 146  * "current->state = TASK_RUNNING" to mark yourself runnable
 147  * without the overhead of this.
 148  */
 149 inline void wake_up_process(struct task_struct * p)
     /*  */
 150 {
 151         unsigned long flags;
 152 
 153         save_flags(flags);
 154         cli();
 155         p->state = TASK_RUNNING;
 156         if (!p->next_run)
 157                 add_to_runqueue(p);
 158         restore_flags(flags);
 159 }
 160 
 161 static void process_timeout(unsigned long __data)
     /*  */
 162 {
 163         struct task_struct * p = (struct task_struct *) __data;
 164 
 165         p->timeout = 0;
 166         wake_up_process(p);
 167 }
 168 
 169 /*
 170  *  'schedule()' is the scheduler function. It's a very simple and nice
 171  * scheduler: it's not perfect, but certainly works for most things.
 172  *
 173  * The goto is "interesting".
 174  *
 175  *   NOTE!!  Task 0 is the 'idle' task, which gets called when no other
 176  * tasks can run. It can not be killed, and it cannot sleep. The 'state'
 177  * information in task[0] is never used.
 178  */
 179 asmlinkage void schedule(void)
     /*  */
 180 {
 181         int c;
 182         struct task_struct * p;
 183         struct task_struct * next;
 184         unsigned long timeout = 0;
 185         
 186 #ifdef CONFIG_SMP_DEBUG
 187         int proc=smp_processor_id();
 188         if(active_kernel_processor!=proc)
 189                 panic("active kernel processor set wrongly! %d not %d\n", active_kernel_processor,proc);
 190 #endif
 191 
 192 /* check alarm, wake up any interruptible tasks that have got a signal */
 193 
 194         if (intr_count) {
 195                 printk("Aiee: scheduling in interrupt\n");
 196                 return;
 197         }
 198         run_task_queue(&tq_scheduler);
 199 
 200         need_resched = 0;
 201         cli();
 202         switch (current->state) {
 203                 case TASK_INTERRUPTIBLE:
 204                         if (current->signal & ~current->blocked)
 205                                 goto makerunnable;
 206                         timeout = current->timeout;
 207                         if (timeout && (timeout <= jiffies)) {
 208                                 current->timeout = 0;
 209                                 timeout = 0;
 210                 makerunnable:
 211                                 current->state = TASK_RUNNING;
 212                                 break;
 213                         }
 214                 default:
 215                         del_from_runqueue(current);
 216                 case TASK_RUNNING:
 217         }
 218         p = init_task.next_run;
 219         sti();
 220         
 221 #ifdef CONFIG_SMP
 222         /*
 223          *      This is safe as we do not permit re-entry of schedule()
 224          */
 225         current->processor = NO_PROC_ID;        
 226 #endif  
 227 
 228 /*
 229  * Note! there may appear new tasks on the run-queue during this, as
 230  * interrupts are enabled. However, they will be put on front of the
 231  * list, so our list starting at "p" is essentially fixed.
 232  */
 233 /* this is the scheduler proper: */
 234         c = -1000;
 235         next = &init_task;
 236         while (p != &init_task) {
 237 #ifdef CONFIG_SMP       
 238                 /* We are not permitted to run a task someone else is running */
 239                 if (p->processor != NO_PROC_ID) {
 240                         p = p->next_run;
 241                         continue;
 242                 }
 243 #endif          
 244                 if (p->counter > c)
 245                         c = p->counter, next = p;
 246                 p = p->next_run;
 247         }
 248 
 249         /* if all runnable processes have "counter == 0", re-calculate counters */
 250         if (!c) {
 251                 for_each_task(p)
 252                         p->counter = (p->counter >> 1) + p->priority;
 253         }
 254 #ifdef CONFIG_SMP       
 255         
 256         /*
 257          *      Context switching between two idle threads is pointless.
 258          */
 259         if(!current->pid && !next->pid)
 260                 next=current;
 261         /*
 262          *      Allocate process to CPU
 263          */
 264          
 265          next->processor = smp_processor_id();
 266          
 267 #endif   
 268         if (current != next) {
 269                 struct timer_list timer;
 270 
 271                 kstat.context_swtch++;
 272                 if (timeout) {
 273                         init_timer(&timer);
 274                         timer.expires = timeout;
 275                         timer.data = (unsigned long) current;
 276                         timer.function = process_timeout;
 277                         add_timer(&timer);
 278                 }
 279                 switch_to(next);
 280                 if (timeout)
 281                         del_timer(&timer);
 282         }
 283 }
 284 
 285 asmlinkage int sys_pause(void)
     /*  */
 286 {
 287         current->state = TASK_INTERRUPTIBLE;
 288         schedule();
 289         return -ERESTARTNOHAND;
 290 }
 291 
 292 /*
 293  * wake_up doesn't wake up stopped processes - they have to be awakened
 294  * with signals or similar.
 295  *
 296  * Note that this doesn't need cli-sti pairs: interrupts may not change
 297  * the wait-queue structures directly, but only call wake_up() to wake
 298  * a process. The process itself must remove the queue once it has woken.
 299  */
 300 void wake_up(struct wait_queue **q)
     /*  */
 301 {
 302         struct wait_queue *tmp;
 303         struct task_struct * p;
 304 
 305         if (!q || !(tmp = *q))
 306                 return;
 307         do {
 308                 if ((p = tmp->task) != NULL) {
 309                         if ((p->state == TASK_UNINTERRUPTIBLE) ||
 310                             (p->state == TASK_INTERRUPTIBLE))
 311                                 wake_up_process(p);
 312                 }
 313                 if (!tmp->next) {
 314                         printk("wait_queue is bad (eip = %p)\n",
 315                                 __builtin_return_address(0));
 316                         printk("        q = %p\n",q);
 317                         printk("       *q = %p\n",*q);
 318                         printk("      tmp = %p\n",tmp);
 319                         break;
 320                 }
 321                 tmp = tmp->next;
 322         } while (tmp != *q);
 323 }
 324 
 325 void wake_up_interruptible(struct wait_queue **q)
     /*  */
 326 {
 327         struct wait_queue *tmp;
 328         struct task_struct * p;
 329 
 330         if (!q || !(tmp = *q))
 331                 return;
 332         do {
 333                 if ((p = tmp->task) != NULL) {
 334                         if (p->state == TASK_INTERRUPTIBLE)
 335                                 wake_up_process(p);
 336                 }
 337                 if (!tmp->next) {
 338                         printk("wait_queue is bad (eip = %p)\n",
 339                                 __builtin_return_address(0));
 340                         printk("        q = %p\n",q);
 341                         printk("       *q = %p\n",*q);
 342                         printk("      tmp = %p\n",tmp);
 343                         break;
 344                 }
 345                 tmp = tmp->next;
 346         } while (tmp != *q);
 347 }
 348 
 349 void __down(struct semaphore * sem)
     /*  */
 350 {
 351         struct wait_queue wait = { current, NULL };
 352         add_wait_queue(&sem->wait, &wait);
 353         current->state = TASK_UNINTERRUPTIBLE;
 354         while (sem->count <= 0) {
 355                 schedule();
 356                 current->state = TASK_UNINTERRUPTIBLE;
 357         }
 358         current->state = TASK_RUNNING;
 359         remove_wait_queue(&sem->wait, &wait);
 360 }
 361 
 362 static inline void __sleep_on(struct wait_queue **p, int state)
     /*  */
 363 {
 364         unsigned long flags;
 365         struct wait_queue wait = { current, NULL };
 366 
 367         if (!p)
 368                 return;
 369         if (current == task[0])
 370                 panic("task[0] trying to sleep");
 371         current->state = state;
 372         add_wait_queue(p, &wait);
 373         save_flags(flags);
 374         sti();
 375         schedule();
 376         remove_wait_queue(p, &wait);
 377         restore_flags(flags);
 378 }
 379 
 380 void interruptible_sleep_on(struct wait_queue **p)
     /*  */
 381 {
 382         __sleep_on(p,TASK_INTERRUPTIBLE);
 383 }
 384 
 385 void sleep_on(struct wait_queue **p)
     /*  */
 386 {
 387         __sleep_on(p,TASK_UNINTERRUPTIBLE);
 388 }
 389 
 390 /*
 391  * The head for the timer-list has a "expires" field of MAX_UINT,
 392  * and the sorting routine counts on this..
 393  */
 394 static struct timer_list timer_head = { &timer_head, &timer_head, ~0, 0, NULL };
 395 #define SLOW_BUT_DEBUGGING_TIMERS 1
 396 
 397 void add_timer(struct timer_list * timer)
     /*  */
 398 {
 399         unsigned long flags;
 400         struct timer_list *p;
 401 
 402 #if SLOW_BUT_DEBUGGING_TIMERS
 403         if (timer->next || timer->prev) {
 404                 printk("add_timer() called with non-zero list from %p\n",
 405                         __builtin_return_address(0));
 406                 return;
 407         }
 408 #endif
 409         p = &timer_head;
 410         save_flags(flags);
 411         cli();
 412         do {
 413                 p = p->next;
 414         } while (timer->expires > p->expires);
 415         timer->next = p;
 416         timer->prev = p->prev;
 417         p->prev = timer;
 418         timer->prev->next = timer;
 419         restore_flags(flags);
 420 }
 421 
 422 int del_timer(struct timer_list * timer)
     /*  */
 423 {
 424         unsigned long flags;
 425 #if SLOW_BUT_DEBUGGING_TIMERS
 426         struct timer_list * p;
 427 
 428         p = &timer_head;
 429         save_flags(flags);
 430         cli();
 431         while ((p = p->next) != &timer_head) {
 432                 if (p == timer) {
 433                         timer->next->prev = timer->prev;
 434                         timer->prev->next = timer->next;
 435                         timer->next = timer->prev = NULL;
 436                         restore_flags(flags);
 437                         return 1;
 438                 }
 439         }
 440         if (timer->next || timer->prev)
 441                 printk("del_timer() called from %p with timer not initialized\n",
 442                         __builtin_return_address(0));
 443         restore_flags(flags);
 444         return 0;
 445 #else   
 446         save_flags(flags);
 447         cli();
 448         if (timer->next) {
 449                 timer->next->prev = timer->prev;
 450                 timer->prev->next = timer->next;
 451                 timer->next = timer->prev = NULL;
 452                 restore_flags(flags);
 453                 return 1;
 454         }
 455         restore_flags(flags);
 456         return 0;
 457 #endif
 458 }
 459 
 460 unsigned long timer_active = 0;
 461 struct timer_struct timer_table[32];
 462 
 463 /*
 464  * Hmm.. Changed this, as the GNU make sources (load.c) seems to
 465  * imply that avenrun[] is the standard name for this kind of thing.
 466  * Nothing else seems to be standardized: the fractional size etc
 467  * all seem to differ on different machines.
 468  */
 469 unsigned long avenrun[3] = { 0,0,0 };
 470 
 471 /*
 472  * Nr of active tasks - counted in fixed-point numbers
 473  */
 474 static unsigned long count_active_tasks(void)
     /*  */
 475 {
 476         struct task_struct **p;
 477         unsigned long nr = 0;
 478 
 479         for(p = &LAST_TASK; p > &FIRST_TASK; --p)
 480                 if (*p && ((*p)->state == TASK_RUNNING ||
 481                            (*p)->state == TASK_UNINTERRUPTIBLE ||
 482                            (*p)->state == TASK_SWAPPING))
 483                         nr += FIXED_1;
 484 #ifdef CONFIG_SMP
 485         nr-=(smp_num_cpus-1)*FIXED_1;
 486 #endif                  
 487         return nr;
 488 }
 489 
 490 static inline void calc_load(void)
     /*  */
 491 {
 492         unsigned long active_tasks; /* fixed-point */
 493         static int count = LOAD_FREQ;
 494 
 495         if (count-- > 0)
 496                 return;
 497         count = LOAD_FREQ;
 498         active_tasks = count_active_tasks();
 499         CALC_LOAD(avenrun[0], EXP_1, active_tasks);
 500         CALC_LOAD(avenrun[1], EXP_5, active_tasks);
 501         CALC_LOAD(avenrun[2], EXP_15, active_tasks);
 502 }
 503 
 504 /*
 505  * this routine handles the overflow of the microsecond field
 506  *
 507  * The tricky bits of code to handle the accurate clock support
 508  * were provided by Dave Mills (Mills@UDEL.EDU) of NTP fame.
 509  * They were originally developed for SUN and DEC kernels.
 510  * All the kudos should go to Dave for this stuff.
 511  *
 512  */
 513 static void second_overflow(void)
     /*  */
 514 {
 515     long ltemp;
 516 
 517     /* Bump the maxerror field */
 518     time_maxerror = (0x70000000-time_maxerror <
 519                      time_tolerance >> SHIFT_USEC) ?
 520         0x70000000 : (time_maxerror + (time_tolerance >> SHIFT_USEC));
 521 
 522     /*
 523      * Leap second processing. If in leap-insert state at
 524      * the end of the day, the system clock is set back one
 525      * second; if in leap-delete state, the system clock is
 526      * set ahead one second. The microtime() routine or
 527      * external clock driver will insure that reported time
 528      * is always monotonic. The ugly divides should be
 529      * replaced.
 530      */
 531     switch (time_state) {
 532 
 533     case TIME_OK:
 534         if (time_status & STA_INS)
 535             time_state = TIME_INS;
 536         else if (time_status & STA_DEL)
 537             time_state = TIME_DEL;
 538         break;
 539 
 540     case TIME_INS:
 541         if (xtime.tv_sec % 86400 == 0) {
 542             xtime.tv_sec--;
 543             time_state = TIME_OOP;
 544             printk("Clock: inserting leap second 23:59:60 UTC\n");
 545         }
 546         break;
 547 
 548     case TIME_DEL:
 549         if ((xtime.tv_sec + 1) % 86400 == 0) {
 550             xtime.tv_sec++;
 551             time_state = TIME_WAIT;
 552             printk("Clock: deleting leap second 23:59:59 UTC\n");
 553         }
 554         break;
 555 
 556     case TIME_OOP:
 557         time_state = TIME_WAIT;
 558         break;
 559 
 560     case TIME_WAIT:
 561         if (!(time_status & (STA_INS | STA_DEL)))
 562             time_state = TIME_OK;
 563     }
 564 
 565     /*
 566      * Compute the phase adjustment for the next second. In
 567      * PLL mode, the offset is reduced by a fixed factor
 568      * times the time constant. In FLL mode the offset is
 569      * used directly. In either mode, the maximum phase
 570      * adjustment for each second is clamped so as to spread
 571      * the adjustment over not more than the number of
 572      * seconds between updates.
 573      */
 574     if (time_offset < 0) {
 575         ltemp = -time_offset;
 576         if (!(time_status & STA_FLL))
 577             ltemp >>= SHIFT_KG + time_constant;
 578         if (ltemp > (MAXPHASE / MINSEC) << SHIFT_UPDATE)
 579             ltemp = (MAXPHASE / MINSEC) <<
 580                 SHIFT_UPDATE;
 581         time_offset += ltemp;
 582         time_adj = -ltemp << (SHIFT_SCALE - SHIFT_HZ -
 583                               SHIFT_UPDATE);
 584     } else {
 585         ltemp = time_offset;
 586         if (!(time_status & STA_FLL))
 587             ltemp >>= SHIFT_KG + time_constant;
 588         if (ltemp > (MAXPHASE / MINSEC) << SHIFT_UPDATE)
 589             ltemp = (MAXPHASE / MINSEC) <<
 590                 SHIFT_UPDATE;
 591         time_offset -= ltemp;
 592         time_adj = ltemp << (SHIFT_SCALE - SHIFT_HZ -
 593                              SHIFT_UPDATE);
 594     }
 595 
 596     /*
 597      * Compute the frequency estimate and additional phase
 598      * adjustment due to frequency error for the next
 599      * second. When the PPS signal is engaged, gnaw on the
 600      * watchdog counter and update the frequency computed by
 601      * the pll and the PPS signal.
 602      */
 603     pps_valid++;
 604     if (pps_valid == PPS_VALID) {
 605         pps_jitter = MAXTIME;
 606         pps_stabil = MAXFREQ;
 607         time_status &= ~(STA_PPSSIGNAL | STA_PPSJITTER |
 608                          STA_PPSWANDER | STA_PPSERROR);
 609     }
 610     ltemp = time_freq + pps_freq;
 611     if (ltemp < 0)
 612         time_adj -= -ltemp >>
 613             (SHIFT_USEC + SHIFT_HZ - SHIFT_SCALE);
 614     else
 615         time_adj += ltemp >>
 616             (SHIFT_USEC + SHIFT_HZ - SHIFT_SCALE);
 617 
 618     /* compensate for (HZ==100) != 128. Add 25% to get 125; => only 3% error */
 619     if (time_adj < 0)
 620         time_adj -= -time_adj >> 2;
 621     else
 622         time_adj += time_adj >> 2;
 623 }
 624 
 625 /*
 626  * disregard lost ticks for now.. We don't care enough.
 627  */
 628 static void timer_bh(void * unused)
     /*  */
 629 {
 630         unsigned long mask;
 631         struct timer_struct *tp;
 632         struct timer_list * timer;
 633 
 634         cli();
 635         while ((timer = timer_head.next) != &timer_head && timer->expires <= jiffies) {
 636                 void (*fn)(unsigned long) = timer->function;
 637                 unsigned long data = timer->data;
 638                 timer->next->prev = timer->prev;
 639                 timer->prev->next = timer->next;
 640                 timer->next = timer->prev = NULL;
 641                 sti();
 642                 fn(data);
 643                 cli();
 644         }
 645         sti();
 646         
 647         for (mask = 1, tp = timer_table+0 ; mask ; tp++,mask += mask) {
 648                 if (mask > timer_active)
 649                         break;
 650                 if (!(mask & timer_active))
 651                         continue;
 652                 if (tp->expires > jiffies)
 653                         continue;
 654                 timer_active &= ~mask;
 655                 tp->fn();
 656                 sti();
 657         }
 658 }
 659 
 660 void tqueue_bh(void * unused)
     /*  */
 661 {
 662         run_task_queue(&tq_timer);
 663 }
 664 
 665 void immediate_bh(void * unused)
     /*  */
 666 {
 667         run_task_queue(&tq_immediate);
 668 }
 669 
 670 void do_timer(struct pt_regs * regs)
     /*  */
 671 {
 672         unsigned long mask;
 673         struct timer_struct *tp;
 674         long ltemp, psecs;
 675 
 676         /* Advance the phase, once it gets to one microsecond, then
 677          * advance the tick more.
 678          */
 679         time_phase += time_adj;
 680         if (time_phase <= -FINEUSEC) {
 681                 ltemp = -time_phase >> SHIFT_SCALE;
 682                 time_phase += ltemp << SHIFT_SCALE;
 683                 xtime.tv_usec += tick + time_adjust_step - ltemp;
 684         }
 685         else if (time_phase >= FINEUSEC) {
 686                 ltemp = time_phase >> SHIFT_SCALE;
 687                 time_phase -= ltemp << SHIFT_SCALE;
 688                 xtime.tv_usec += tick + time_adjust_step + ltemp;
 689         } else
 690                 xtime.tv_usec += tick + time_adjust_step;
 691 
 692         if (time_adjust) {
 693             /* We are doing an adjtime thing. 
 694              *
 695              * Modify the value of the tick for next time.
 696              * Note that a positive delta means we want the clock
 697              * to run fast. This means that the tick should be bigger
 698              *
 699              * Limit the amount of the step for *next* tick to be
 700              * in the range -tickadj .. +tickadj
 701              */
 702              if (time_adjust > tickadj)
 703                time_adjust_step = tickadj;
 704              else if (time_adjust < -tickadj)
 705                time_adjust_step = -tickadj;
 706              else
 707                time_adjust_step = time_adjust;
 708              
 709             /* Reduce by this step the amount of time left  */
 710             time_adjust -= time_adjust_step;
 711         }
 712         else
 713             time_adjust_step = 0;
 714 
 715         if (xtime.tv_usec >= 1000000) {
 716             xtime.tv_usec -= 1000000;
 717             xtime.tv_sec++;
 718             second_overflow();
 719         }
 720 
 721         jiffies++;
 722         calc_load();
 723         if (user_mode(regs)) {
 724                 current->utime++;
 725                 if (current->pid) {
 726                         if (current->priority < 15)
 727                                 kstat.cpu_nice++;
 728                         else
 729                                 kstat.cpu_user++;
 730                 }
 731                 /* Update ITIMER_VIRT for current task if not in a system call */
 732                 if (current->it_virt_value && !(--current->it_virt_value)) {
 733                         current->it_virt_value = current->it_virt_incr;
 734                         send_sig(SIGVTALRM,current,1);
 735                 }
 736         } else {
 737                 current->stime++;
 738                 if(current->pid)
 739                         kstat.cpu_system++;
 740                 if (prof_buffer && current->pid) {
 741                         extern int _stext;
 742                         unsigned long ip = instruction_pointer(regs);
 743                         ip -= (unsigned long) &_stext;
 744                         ip >>= prof_shift;
 745                         if (ip < prof_len)
 746                                 prof_buffer[ip]++;
 747                 }
 748         }
 749         /*
 750          * check the cpu time limit on the process.
 751          */
 752         if ((current->rlim[RLIMIT_CPU].rlim_max != RLIM_INFINITY) &&
 753             (((current->stime + current->utime) / HZ) >= current->rlim[RLIMIT_CPU].rlim_max))
 754                 send_sig(SIGKILL, current, 1);
 755         if ((current->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) &&
 756             (((current->stime + current->utime) % HZ) == 0)) {
 757                 psecs = (current->stime + current->utime) / HZ;
 758                 /* send when equal */
 759                 if (psecs == current->rlim[RLIMIT_CPU].rlim_cur)
 760                         send_sig(SIGXCPU, current, 1);
 761                 /* and every five seconds thereafter. */
 762                 else if ((psecs > current->rlim[RLIMIT_CPU].rlim_cur) &&
 763                         ((psecs - current->rlim[RLIMIT_CPU].rlim_cur) % 5) == 0)
 764                         send_sig(SIGXCPU, current, 1);
 765         }
 766 
 767         if (current->pid && 0 > --current->counter) {
 768                 current->counter = 0;
 769                 need_resched = 1;
 770         }
 771         /* Update ITIMER_PROF for the current task */
 772         if (current->it_prof_value && !(--current->it_prof_value)) {
 773                 current->it_prof_value = current->it_prof_incr;
 774                 send_sig(SIGPROF,current,1);
 775         }
 776         for (mask = 1, tp = timer_table+0 ; mask ; tp++,mask += mask) {
 777                 if (mask > timer_active)
 778                         break;
 779                 if (!(mask & timer_active))
 780                         continue;
 781                 if (tp->expires > jiffies)
 782                         continue;
 783                 mark_bh(TIMER_BH);
 784         }
 785         cli();
 786         if (timer_head.next->expires <= jiffies)
 787                 mark_bh(TIMER_BH);
 788         if (tq_timer != &tq_last)
 789                 mark_bh(TQUEUE_BH);
 790         sti();
 791 }
 792 
 793 asmlinkage unsigned int sys_alarm(unsigned int seconds)
     /*  */
 794 {
 795         struct itimerval it_new, it_old;
 796         unsigned int oldalarm;
 797 
 798         it_new.it_interval.tv_sec = it_new.it_interval.tv_usec = 0;
 799         it_new.it_value.tv_sec = seconds;
 800         it_new.it_value.tv_usec = 0;
 801         _setitimer(ITIMER_REAL, &it_new, &it_old);
 802         oldalarm = it_old.it_value.tv_sec;
 803         /* ehhh.. We can't return 0 if we have an alarm pending.. */
 804         /* And we'd better return too much than too little anyway */
 805         if (it_old.it_value.tv_usec)
 806                 oldalarm++;
 807         return oldalarm;
 808 }
 809 
 810 asmlinkage int sys_getpid(void)
     /*  */
 811 {
 812         return current->pid;
 813 }
 814 
 815 asmlinkage int sys_getppid(void)
     /*  */
 816 {
 817         return current->p_opptr->pid;
 818 }
 819 
 820 asmlinkage int sys_getuid(void)
     /*  */
 821 {
 822         return current->uid;
 823 }
 824 
 825 asmlinkage int sys_geteuid(void)
     /*  */
 826 {
 827         return current->euid;
 828 }
 829 
 830 asmlinkage int sys_getgid(void)
     /*  */
 831 {
 832         return current->gid;
 833 }
 834 
 835 asmlinkage int sys_getegid(void)
     /*  */
 836 {
 837         return current->egid;
 838 }
 839 
 840 asmlinkage int sys_nice(long increment)
     /*  */
 841 {
 842         int newprio;
 843 
 844         if (increment < 0 && !suser())
 845                 return -EPERM;
 846         newprio = current->priority - increment;
 847         if (newprio < 1)
 848                 newprio = 1;
 849         if (newprio > 35)
 850                 newprio = 35;
 851         current->priority = newprio;
 852         return 0;
 853 }
 854 
 855 static void show_task(int nr,struct task_struct * p)
     /*  */
 856 {
 857         unsigned long free;
 858         static const char * stat_nam[] = { "R", "S", "D", "Z", "T", "W" };
 859 
 860         printk("%-8s %3d ", p->comm, (p == current) ? -nr : nr);
 861         if (((unsigned) p->state) < sizeof(stat_nam)/sizeof(char *))
 862                 printk(stat_nam[p->state]);
 863         else
 864                 printk(" ");
 865 #if ((~0UL) == 0xffffffff)
 866         if (p == current)
 867                 printk(" current  ");
 868         else
 869                 printk(" %08lX ", thread_saved_pc(&p->tss));
 870 #else
 871         if (p == current)
 872                 printk("   current task   ");
 873         else
 874                 printk(" %016lx ", thread_saved_pc(&p->tss));
 875 #endif
 876         for (free = 1; free < PAGE_SIZE/sizeof(long) ; free++) {
 877                 if (((unsigned long *)p->kernel_stack_page)[free])
 878                         break;
 879         }
 880         printk("%5lu %5d %6d ", free*sizeof(long), p->pid, p->p_pptr->pid);
 881         if (p->p_cptr)
 882                 printk("%5d ", p->p_cptr->pid);
 883         else
 884                 printk("      ");
 885         if (p->p_ysptr)
 886                 printk("%7d", p->p_ysptr->pid);
 887         else
 888                 printk("       ");
 889         if (p->p_osptr)
 890                 printk(" %5d\n", p->p_osptr->pid);
 891         else
 892                 printk("\n");
 893 }
 894 
 895 void show_state(void)
     /*  */
 896 {
 897         int i;
 898 
 899 #if ((~0UL) == 0xffffffff)
 900         printk("\n"
 901                "                         free                        sibling\n");
 902         printk("  task             PC    stack   pid father child younger older\n");
 903 #else
 904         printk("\n"
 905                "                                 free                        sibling\n");
 906         printk("  task                 PC        stack   pid father child younger older\n");
 907 #endif
 908         for (i=0 ; i<NR_TASKS ; i++)
 909                 if (task[i])
 910                         show_task(i,task[i]);
 911 }
 912 
 913 void sched_init(void)
     /*  */
 914 {
 915         /*
 916          *      We have to do a little magic to get the first
 917          *      process right in SMP mode.
 918          */
 919         int cpu=smp_processor_id();
 920         current_set[cpu]=&init_task;
 921 #ifdef CONFIG_SMP       
 922         init_task.processor=cpu;
 923 #endif
 924         bh_base[TIMER_BH].routine = timer_bh;
 925         bh_base[TQUEUE_BH].routine = tqueue_bh;
 926         bh_base[IMMEDIATE_BH].routine = immediate_bh;
 927         enable_bh(TIMER_BH);
 928         enable_bh(TQUEUE_BH);
 929         enable_bh(IMMEDIATE_BH);
 930 }
/* */
root/kernel/sched.c

DEFINITIONS
