kernel/sched.c

/* */
This source file includes following definitions.
math_state_restore
math_emulate
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 
  29 #include <asm/system.h>
  30 #include <asm/io.h>
  31 #include <asm/segment.h>
  32 
  33 #define TIMER_IRQ 0
  34 
  35 #include <linux/timex.h>
  36 
  37 /*
  38  * kernel variables
  39  */
  40 long tick = 1000000 / HZ;               /* timer interrupt period */
  41 volatile struct timeval xtime;          /* The current time */
  42 int tickadj = 500/HZ;                   /* microsecs */
  43 
  44 DECLARE_TASK_QUEUE(tq_timer);
  45 DECLARE_TASK_QUEUE(tq_immediate);
  46 
  47 /*
  48  * phase-lock loop variables
  49  */
  50 int time_status = TIME_BAD;     /* clock synchronization status */
  51 long time_offset = 0;           /* time adjustment (us) */
  52 long time_constant = 0;         /* pll time constant */
  53 long time_tolerance = MAXFREQ;  /* frequency tolerance (ppm) */
  54 long time_precision = 1;        /* clock precision (us) */
  55 long time_maxerror = 0x70000000;/* maximum error */
  56 long time_esterror = 0x70000000;/* estimated error */
  57 long time_phase = 0;            /* phase offset (scaled us) */
  58 long time_freq = 0;             /* frequency offset (scaled ppm) */
  59 long time_adj = 0;              /* tick adjust (scaled 1 / HZ) */
  60 long time_reftime = 0;          /* time at last adjustment (s) */
  61 
  62 long time_adjust = 0;
  63 long time_adjust_step = 0;
  64 
  65 int need_resched = 0;
  66 unsigned long event = 0;
  67 
  68 /*
  69  * Tell us the machine setup..
  70  */
  71 int hard_math = 0;              /* set by boot/head.S */
  72 int x86 = 0;                    /* set by boot/head.S to 3 or 4 */
  73 int fdiv_bug = 0;               /* set if Pentium(TM) with FP bug */
  74 
  75 char x86_vendor_id[13] = "Unknown";
  76 
  77 int ignore_irq13 = 0;           /* set if exception 16 works */
  78 int wp_works_ok = 0;            /* set if paging hardware honours WP */ 
  79 int hlt_works_ok = 1;           /* set if the "hlt" instruction works */
  80 
  81 /*
  82  * Bus types ..
  83  */
  84 int EISA_bus = 0;
  85 
  86 extern int _setitimer(int, struct itimerval *, struct itimerval *);
  87 unsigned long * prof_buffer = NULL;
  88 unsigned long prof_len = 0;
  89 
  90 #define _S(nr) (1<<((nr)-1))
  91 
  92 extern void mem_use(void);
  93 
  94 extern int timer_interrupt(void);
  95 asmlinkage int system_call(void);
  96 
  97 static unsigned long init_kernel_stack[1024] = { STACK_MAGIC, };
  98 static struct vm_area_struct init_mmap = INIT_MMAP;
  99 struct task_struct init_task = INIT_TASK;
 100 
 101 unsigned long volatile jiffies=0;
 102 
 103 struct task_struct *current = &init_task;
 104 struct task_struct *last_task_used_math = NULL;
 105 
 106 struct task_struct * task[NR_TASKS] = {&init_task, };
 107 
 108 long user_stack [ PAGE_SIZE>>2 ] = { STACK_MAGIC, };
 109 
 110 struct {
 111         long * a;
 112         short b;
 113         } stack_start = { & user_stack [PAGE_SIZE>>2] , KERNEL_DS };
 114 
 115 struct kernel_stat kstat = { 0 };
 116 
 117 /*
 118  *  'math_state_restore()' saves the current math information in the
 119  * old math state array, and gets the new ones from the current task
 120  *
 121  * Careful.. There are problems with IBM-designed IRQ13 behaviour.
 122  * Don't touch unless you *really* know how it works.
 123  */
 124 asmlinkage void math_state_restore(void)
     /*  */
 125 {
 126         __asm__ __volatile__("clts");
 127         if (last_task_used_math == current)
 128                 return;
 129         timer_table[COPRO_TIMER].expires = jiffies+50;
 130         timer_active |= 1<<COPRO_TIMER; 
 131         if (last_task_used_math)
 132                 __asm__("fnsave %0":"=m" (last_task_used_math->tss.i387));
 133         else
 134                 __asm__("fnclex");
 135         last_task_used_math = current;
 136         if (current->used_math) {
 137                 __asm__("frstor %0": :"m" (current->tss.i387));
 138         } else {
 139                 __asm__("fninit");
 140                 current->used_math=1;
 141         }
 142         timer_active &= ~(1<<COPRO_TIMER);
 143 }
 144 
 145 #ifndef CONFIG_MATH_EMULATION
 146 
 147 asmlinkage void math_emulate(long arg)
     /*  */
 148 {
 149   printk("math-emulation not enabled and no coprocessor found.\n");
 150   printk("killing %s.\n",current->comm);
 151   send_sig(SIGFPE,current,1);
 152   schedule();
 153 }
 154 
 155 #endif /* CONFIG_MATH_EMULATION */
 156 
 157 unsigned long itimer_ticks = 0;
 158 unsigned long itimer_next = ~0;
 159 
 160 /*
 161  *  'schedule()' is the scheduler function. It's a very simple and nice
 162  * scheduler: it's not perfect, but certainly works for most things.
 163  * The one thing you might take a look at is the signal-handler code here.
 164  *
 165  *   NOTE!!  Task 0 is the 'idle' task, which gets called when no other
 166  * tasks can run. It can not be killed, and it cannot sleep. The 'state'
 167  * information in task[0] is never used.
 168  *
 169  * The "confuse_gcc" goto is used only to get better assembly code..
 170  * Dijkstra probably hates me.
 171  */
 172 asmlinkage void schedule(void)
     /*  */
 173 {
 174         int c;
 175         struct task_struct * p;
 176         struct task_struct * next;
 177         unsigned long ticks;
 178 
 179 /* check alarm, wake up any interruptible tasks that have got a signal */
 180 
 181         if (intr_count) {
 182                 printk("Aiee: scheduling in interrupt\n");
 183                 intr_count = 0;
 184         }
 185         cli();
 186         ticks = itimer_ticks;
 187         itimer_ticks = 0;
 188         itimer_next = ~0;
 189         sti();
 190         need_resched = 0;
 191         p = &init_task;
 192         for (;;) {
 193                 if ((p = p->next_task) == &init_task)
 194                         goto confuse_gcc1;
 195                 if (ticks && p->it_real_value) {
 196                         if (p->it_real_value <= ticks) {
 197                                 send_sig(SIGALRM, p, 1);
 198                                 if (!p->it_real_incr) {
 199                                         p->it_real_value = 0;
 200                                         goto end_itimer;
 201                                 }
 202                                 do {
 203                                         p->it_real_value += p->it_real_incr;
 204                                 } while (p->it_real_value <= ticks);
 205                         }
 206                         p->it_real_value -= ticks;
 207                         if (p->it_real_value < itimer_next)
 208                                 itimer_next = p->it_real_value;
 209                 }
 210 end_itimer:
 211                 if (p->state != TASK_INTERRUPTIBLE)
 212                         continue;
 213                 if (p->signal & ~p->blocked) {
 214                         p->state = TASK_RUNNING;
 215                         continue;
 216                 }
 217                 if (p->timeout && p->timeout <= jiffies) {
 218                         p->timeout = 0;
 219                         p->state = TASK_RUNNING;
 220                 }
 221         }
 222 confuse_gcc1:
 223 
 224 /* this is the scheduler proper: */
 225 #if 0
 226         /* give processes that go to sleep a bit higher priority.. */
 227         /* This depends on the values for TASK_XXX */
 228         /* This gives smoother scheduling for some things, but */
 229         /* can be very unfair under some circumstances, so.. */
 230         if (TASK_UNINTERRUPTIBLE >= (unsigned) current->state &&
 231             current->counter < current->priority*2) {
 232                 ++current->counter;
 233         }
 234 #endif
 235         c = -1000;
 236         next = p = &init_task;
 237         for (;;) {
 238                 if ((p = p->next_task) == &init_task)
 239                         goto confuse_gcc2;
 240                 if (p->state == TASK_RUNNING && p->counter > c)
 241                         c = p->counter, next = p;
 242         }
 243 confuse_gcc2:
 244         if (!c) {
 245                 for_each_task(p)
 246                         p->counter = (p->counter >> 1) + p->priority;
 247         }
 248         if (current == next)
 249                 return;
 250         kstat.context_swtch++;
 251         switch_to(next);
 252         /* Now maybe reload the debug registers */
 253         if(current->debugreg[7]){
 254                 loaddebug(0);
 255                 loaddebug(1);
 256                 loaddebug(2);
 257                 loaddebug(3);
 258                 loaddebug(6);
 259         };
 260 }
 261 
 262 asmlinkage int sys_pause(void)
     /*  */
 263 {
 264         current->state = TASK_INTERRUPTIBLE;
 265         schedule();
 266         return -ERESTARTNOHAND;
 267 }
 268 
 269 /*
 270  * wake_up doesn't wake up stopped processes - they have to be awakened
 271  * with signals or similar.
 272  *
 273  * Note that this doesn't need cli-sti pairs: interrupts may not change
 274  * the wait-queue structures directly, but only call wake_up() to wake
 275  * a process. The process itself must remove the queue once it has woken.
 276  */
 277 void wake_up(struct wait_queue **q)
     /*  */
 278 {
 279         struct wait_queue *tmp;
 280         struct task_struct * p;
 281 
 282         if (!q || !(tmp = *q))
 283                 return;
 284         do {
 285                 if ((p = tmp->task) != NULL) {
 286                         if ((p->state == TASK_UNINTERRUPTIBLE) ||
 287                             (p->state == TASK_INTERRUPTIBLE)) {
 288                                 p->state = TASK_RUNNING;
 289                                 if (p->counter > current->counter + 3)
 290                                         need_resched = 1;
 291                         }
 292                 }
 293                 if (!tmp->next) {
 294                         printk("wait_queue is bad (eip = %p)\n",
 295                                 __builtin_return_address(0));
 296                         printk("        q = %p\n",q);
 297                         printk("       *q = %p\n",*q);
 298                         printk("      tmp = %p\n",tmp);
 299                         break;
 300                 }
 301                 tmp = tmp->next;
 302         } while (tmp != *q);
 303 }
 304 
 305 void wake_up_interruptible(struct wait_queue **q)
     /*  */
 306 {
 307         struct wait_queue *tmp;
 308         struct task_struct * p;
 309 
 310         if (!q || !(tmp = *q))
 311                 return;
 312         do {
 313                 if ((p = tmp->task) != NULL) {
 314                         if (p->state == TASK_INTERRUPTIBLE) {
 315                                 p->state = TASK_RUNNING;
 316                                 if (p->counter > current->counter + 3)
 317                                         need_resched = 1;
 318                         }
 319                 }
 320                 if (!tmp->next) {
 321                         printk("wait_queue is bad (eip = %p)\n",
 322                                 __builtin_return_address(0));
 323                         printk("        q = %p\n",q);
 324                         printk("       *q = %p\n",*q);
 325                         printk("      tmp = %p\n",tmp);
 326                         break;
 327                 }
 328                 tmp = tmp->next;
 329         } while (tmp != *q);
 330 }
 331 
 332 void __down(struct semaphore * sem)
     /*  */
 333 {
 334         struct wait_queue wait = { current, NULL };
 335         add_wait_queue(&sem->wait, &wait);
 336         current->state = TASK_UNINTERRUPTIBLE;
 337         while (sem->count <= 0) {
 338                 schedule();
 339                 current->state = TASK_UNINTERRUPTIBLE;
 340         }
 341         current->state = TASK_RUNNING;
 342         remove_wait_queue(&sem->wait, &wait);
 343 }
 344 
 345 static inline void __sleep_on(struct wait_queue **p, int state)
     /*  */
 346 {
 347         unsigned long flags;
 348         struct wait_queue wait = { current, NULL };
 349 
 350         if (!p)
 351                 return;
 352         if (current == task[0])
 353                 panic("task[0] trying to sleep");
 354         current->state = state;
 355         add_wait_queue(p, &wait);
 356         save_flags(flags);
 357         sti();
 358         schedule();
 359         remove_wait_queue(p, &wait);
 360         restore_flags(flags);
 361 }
 362 
 363 void interruptible_sleep_on(struct wait_queue **p)
     /*  */
 364 {
 365         __sleep_on(p,TASK_INTERRUPTIBLE);
 366 }
 367 
 368 void sleep_on(struct wait_queue **p)
     /*  */
 369 {
 370         __sleep_on(p,TASK_UNINTERRUPTIBLE);
 371 }
 372 
 373 /*
 374  * The head for the timer-list has a "expires" field of MAX_UINT,
 375  * and the sorting routine counts on this..
 376  */
 377 static struct timer_list timer_head = { &timer_head, &timer_head, ~0, 0, NULL };
 378 #define SLOW_BUT_DEBUGGING_TIMERS 1
 379 
 380 void add_timer(struct timer_list * timer)
     /*  */
 381 {
 382         unsigned long flags;
 383         struct timer_list *p;
 384 
 385 #if SLOW_BUT_DEBUGGING_TIMERS
 386         if (timer->next || timer->prev) {
 387                 printk("add_timer() called with non-zero list from %p\n",
 388                         __builtin_return_address(0));
 389                 return;
 390         }
 391 #endif
 392         p = &timer_head;
 393         timer->expires += jiffies;
 394         save_flags(flags);
 395         cli();
 396         do {
 397                 p = p->next;
 398         } while (timer->expires > p->expires);
 399         timer->next = p;
 400         timer->prev = p->prev;
 401         p->prev = timer;
 402         timer->prev->next = timer;
 403         restore_flags(flags);
 404 }
 405 
 406 int del_timer(struct timer_list * timer)
     /*  */
 407 {
 408         unsigned long flags;
 409 #if SLOW_BUT_DEBUGGING_TIMERS
 410         struct timer_list * p;
 411 
 412         p = &timer_head;
 413         save_flags(flags);
 414         cli();
 415         while ((p = p->next) != &timer_head) {
 416                 if (p == timer) {
 417                         timer->next->prev = timer->prev;
 418                         timer->prev->next = timer->next;
 419                         timer->next = timer->prev = NULL;
 420                         restore_flags(flags);
 421                         timer->expires -= jiffies;
 422                         return 1;
 423                 }
 424         }
 425         if (timer->next || timer->prev)
 426                 printk("del_timer() called from %p with timer not initialized\n",
 427                         __builtin_return_address(0));
 428         restore_flags(flags);
 429         return 0;
 430 #else   
 431         save_flags(flags);
 432         cli();
 433         if (timer->next) {
 434                 timer->next->prev = timer->prev;
 435                 timer->prev->next = timer->next;
 436                 timer->next = timer->prev = NULL;
 437                 restore_flags(flags);
 438                 timer->expires -= jiffies;
 439                 return 1;
 440         }
 441         restore_flags(flags);
 442         return 0;
 443 #endif
 444 }
 445 
 446 unsigned long timer_active = 0;
 447 struct timer_struct timer_table[32];
 448 
 449 /*
 450  * Hmm.. Changed this, as the GNU make sources (load.c) seems to
 451  * imply that avenrun[] is the standard name for this kind of thing.
 452  * Nothing else seems to be standardized: the fractional size etc
 453  * all seem to differ on different machines.
 454  */
 455 unsigned long avenrun[3] = { 0,0,0 };
 456 
 457 /*
 458  * Nr of active tasks - counted in fixed-point numbers
 459  */
 460 static unsigned long count_active_tasks(void)
     /*  */
 461 {
 462         struct task_struct **p;
 463         unsigned long nr = 0;
 464 
 465         for(p = &LAST_TASK; p > &FIRST_TASK; --p)
 466                 if (*p && ((*p)->state == TASK_RUNNING ||
 467                            (*p)->state == TASK_UNINTERRUPTIBLE ||
 468                            (*p)->state == TASK_SWAPPING))
 469                         nr += FIXED_1;
 470         return nr;
 471 }
 472 
 473 static inline void calc_load(void)
     /*  */
 474 {
 475         unsigned long active_tasks; /* fixed-point */
 476         static int count = LOAD_FREQ;
 477 
 478         if (count-- > 0)
 479                 return;
 480         count = LOAD_FREQ;
 481         active_tasks = count_active_tasks();
 482         CALC_LOAD(avenrun[0], EXP_1, active_tasks);
 483         CALC_LOAD(avenrun[1], EXP_5, active_tasks);
 484         CALC_LOAD(avenrun[2], EXP_15, active_tasks);
 485 }
 486 
 487 /*
 488  * this routine handles the overflow of the microsecond field
 489  *
 490  * The tricky bits of code to handle the accurate clock support
 491  * were provided by Dave Mills (Mills@UDEL.EDU) of NTP fame.
 492  * They were originally developed for SUN and DEC kernels.
 493  * All the kudos should go to Dave for this stuff.
 494  *
 495  * These were ported to Linux by Philip Gladstone.
 496  */
 497 static void second_overflow(void)
     /*  */
 498 {
 499         long ltemp;
 500         /* last time the cmos clock got updated */
 501         static long last_rtc_update=0;
 502         extern int set_rtc_mmss(unsigned long);
 503 
 504         /* Bump the maxerror field */
 505         time_maxerror = (0x70000000-time_maxerror < time_tolerance) ?
 506           0x70000000 : (time_maxerror + time_tolerance);
 507 
 508         /* Run the PLL */
 509         if (time_offset < 0) {
 510                 ltemp = (-(time_offset+1) >> (SHIFT_KG + time_constant)) + 1;
 511                 time_adj = ltemp << (SHIFT_SCALE - SHIFT_HZ - SHIFT_UPDATE);
 512                 time_offset += (time_adj * HZ) >> (SHIFT_SCALE - SHIFT_UPDATE);
 513                 time_adj = - time_adj;
 514         } else if (time_offset > 0) {
 515                 ltemp = ((time_offset-1) >> (SHIFT_KG + time_constant)) + 1;
 516                 time_adj = ltemp << (SHIFT_SCALE - SHIFT_HZ - SHIFT_UPDATE);
 517                 time_offset -= (time_adj * HZ) >> (SHIFT_SCALE - SHIFT_UPDATE);
 518         } else {
 519                 time_adj = 0;
 520         }
 521 
 522         time_adj += (time_freq >> (SHIFT_KF + SHIFT_HZ - SHIFT_SCALE))
 523             + FINETUNE;
 524 
 525         /* Handle the leap second stuff */
 526         switch (time_status) {
 527                 case TIME_INS:
 528                 /* ugly divide should be replaced */
 529                 if (xtime.tv_sec % 86400 == 0) {
 530                         xtime.tv_sec--; /* !! */
 531                         time_status = TIME_OOP;
 532                         printk("Clock: inserting leap second 23:59:60 GMT\n");
 533                 }
 534                 break;
 535 
 536                 case TIME_DEL:
 537                 /* ugly divide should be replaced */
 538                 if (xtime.tv_sec % 86400 == 86399) {
 539                         xtime.tv_sec++;
 540                         time_status = TIME_OK;
 541                         printk("Clock: deleting leap second 23:59:59 GMT\n");
 542                 }
 543                 break;
 544 
 545                 case TIME_OOP:
 546                 time_status = TIME_OK;
 547                 break;
 548         }
 549         if (xtime.tv_sec > last_rtc_update + 660)
 550           if (set_rtc_mmss(xtime.tv_sec) == 0)
 551             last_rtc_update = xtime.tv_sec;
 552           else
 553             last_rtc_update = xtime.tv_sec - 600; /* do it again in one min */
 554 }
 555 
 556 /*
 557  * disregard lost ticks for now.. We don't care enough.
 558  */
 559 static void timer_bh(void * unused)
     /*  */
 560 {
 561         unsigned long mask;
 562         struct timer_struct *tp;
 563         struct timer_list * timer;
 564 
 565         cli();
 566         while ((timer = timer_head.next) != &timer_head && timer->expires < jiffies) {
 567                 void (*fn)(unsigned long) = timer->function;
 568                 unsigned long data = timer->data;
 569                 timer->next->prev = timer->prev;
 570                 timer->prev->next = timer->next;
 571                 timer->next = timer->prev = NULL;
 572                 sti();
 573                 fn(data);
 574                 cli();
 575         }
 576         sti();
 577         
 578         for (mask = 1, tp = timer_table+0 ; mask ; tp++,mask += mask) {
 579                 if (mask > timer_active)
 580                         break;
 581                 if (!(mask & timer_active))
 582                         continue;
 583                 if (tp->expires > jiffies)
 584                         continue;
 585                 timer_active &= ~mask;
 586                 tp->fn();
 587                 sti();
 588         }
 589 }
 590 
 591 void tqueue_bh(void * unused)
     /*  */
 592 {
 593         run_task_queue(&tq_timer);
 594 }
 595 
 596 void immediate_bh(void * unused)
     /*  */
 597 {
 598         run_task_queue(&tq_immediate);
 599 }
 600 
 601 /*
 602  * The int argument is really a (struct pt_regs *), in case the
 603  * interrupt wants to know from where it was called. The timer
 604  * irq uses this to decide if it should update the user or system
 605  * times.
 606  */
 607 static void do_timer(struct pt_regs * regs)
     /*  */
 608 {
 609         unsigned long mask;
 610         struct timer_struct *tp;
 611 
 612         long ltemp, psecs;
 613 
 614         /* Advance the phase, once it gets to one microsecond, then
 615          * advance the tick more.
 616          */
 617         time_phase += time_adj;
 618         if (time_phase < -FINEUSEC) {
 619                 ltemp = -time_phase >> SHIFT_SCALE;
 620                 time_phase += ltemp << SHIFT_SCALE;
 621                 xtime.tv_usec += tick + time_adjust_step - ltemp;
 622         }
 623         else if (time_phase > FINEUSEC) {
 624                 ltemp = time_phase >> SHIFT_SCALE;
 625                 time_phase -= ltemp << SHIFT_SCALE;
 626                 xtime.tv_usec += tick + time_adjust_step + ltemp;
 627         } else
 628                 xtime.tv_usec += tick + time_adjust_step;
 629 
 630         if (time_adjust)
 631         {
 632             /* We are doing an adjtime thing. 
 633              *
 634              * Modify the value of the tick for next time.
 635              * Note that a positive delta means we want the clock
 636              * to run fast. This means that the tick should be bigger
 637              *
 638              * Limit the amount of the step for *next* tick to be
 639              * in the range -tickadj .. +tickadj
 640              */
 641              if (time_adjust > tickadj)
 642                time_adjust_step = tickadj;
 643              else if (time_adjust < -tickadj)
 644                time_adjust_step = -tickadj;
 645              else
 646                time_adjust_step = time_adjust;
 647              
 648             /* Reduce by this step the amount of time left  */
 649             time_adjust -= time_adjust_step;
 650         }
 651         else
 652             time_adjust_step = 0;
 653 
 654         if (xtime.tv_usec >= 1000000) {
 655             xtime.tv_usec -= 1000000;
 656             xtime.tv_sec++;
 657             second_overflow();
 658         }
 659 
 660         jiffies++;
 661         calc_load();
 662         if ((VM_MASK & regs->eflags) || (3 & regs->cs)) {
 663                 current->utime++;
 664                 if (current != task[0]) {
 665                         if (current->priority < 15)
 666                                 kstat.cpu_nice++;
 667                         else
 668                                 kstat.cpu_user++;
 669                 }
 670                 /* Update ITIMER_VIRT for current task if not in a system call */
 671                 if (current->it_virt_value && !(--current->it_virt_value)) {
 672                         current->it_virt_value = current->it_virt_incr;
 673                         send_sig(SIGVTALRM,current,1);
 674                 }
 675         } else {
 676                 current->stime++;
 677                 if(current != task[0])
 678                         kstat.cpu_system++;
 679 #ifdef CONFIG_PROFILE
 680                 if (prof_buffer && current != task[0]) {
 681                         unsigned long eip = regs->eip;
 682                         eip >>= 2;
 683                         if (eip < prof_len)
 684                                 prof_buffer[eip]++;
 685                 }
 686 #endif
 687         }
 688         /*
 689          * check the cpu time limit on the process.
 690          */
 691         if ((current->rlim[RLIMIT_CPU].rlim_max != RLIM_INFINITY) &&
 692             (((current->stime + current->utime) / HZ) >= current->rlim[RLIMIT_CPU].rlim_max))
 693                 send_sig(SIGKILL, current, 1);
 694         if ((current->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) &&
 695             (((current->stime + current->utime) % HZ) == 0)) {
 696                 psecs = (current->stime + current->utime) / HZ;
 697                 /* send when equal */
 698                 if (psecs == current->rlim[RLIMIT_CPU].rlim_cur)
 699                         send_sig(SIGXCPU, current, 1);
 700                 /* and every five seconds thereafter. */
 701                 else if ((psecs > current->rlim[RLIMIT_CPU].rlim_cur) &&
 702                         ((psecs - current->rlim[RLIMIT_CPU].rlim_cur) % 5) == 0)
 703                         send_sig(SIGXCPU, current, 1);
 704         }
 705 
 706         if (current != task[0] && 0 > --current->counter) {
 707                 current->counter = 0;
 708                 need_resched = 1;
 709         }
 710         /* Update ITIMER_PROF for the current task */
 711         if (current->it_prof_value && !(--current->it_prof_value)) {
 712                 current->it_prof_value = current->it_prof_incr;
 713                 send_sig(SIGPROF,current,1);
 714         }
 715         for (mask = 1, tp = timer_table+0 ; mask ; tp++,mask += mask) {
 716                 if (mask > timer_active)
 717                         break;
 718                 if (!(mask & timer_active))
 719                         continue;
 720                 if (tp->expires > jiffies)
 721                         continue;
 722                 mark_bh(TIMER_BH);
 723         }
 724         cli();
 725         itimer_ticks++;
 726         if (itimer_ticks > itimer_next)
 727                 need_resched = 1;
 728         if (timer_head.next->expires < jiffies)
 729                 mark_bh(TIMER_BH);
 730         if (tq_timer != &tq_last)
 731                 mark_bh(TQUEUE_BH);
 732         sti();
 733 }
 734 
 735 asmlinkage int sys_alarm(long seconds)
     /*  */
 736 {
 737         struct itimerval it_new, it_old;
 738 
 739         it_new.it_interval.tv_sec = it_new.it_interval.tv_usec = 0;
 740         it_new.it_value.tv_sec = seconds;
 741         it_new.it_value.tv_usec = 0;
 742         _setitimer(ITIMER_REAL, &it_new, &it_old);
 743         return(it_old.it_value.tv_sec + (it_old.it_value.tv_usec / 1000000));
 744 }
 745 
 746 asmlinkage int sys_getpid(void)
     /*  */
 747 {
 748         return current->pid;
 749 }
 750 
 751 asmlinkage int sys_getppid(void)
     /*  */
 752 {
 753         return current->p_opptr->pid;
 754 }
 755 
 756 asmlinkage int sys_getuid(void)
     /*  */
 757 {
 758         return current->uid;
 759 }
 760 
 761 asmlinkage int sys_geteuid(void)
     /*  */
 762 {
 763         return current->euid;
 764 }
 765 
 766 asmlinkage int sys_getgid(void)
     /*  */
 767 {
 768         return current->gid;
 769 }
 770 
 771 asmlinkage int sys_getegid(void)
     /*  */
 772 {
 773         return current->egid;
 774 }
 775 
 776 asmlinkage int sys_nice(long increment)
     /*  */
 777 {
 778         int newprio;
 779 
 780         if (increment < 0 && !suser())
 781                 return -EPERM;
 782         newprio = current->priority - increment;
 783         if (newprio < 1)
 784                 newprio = 1;
 785         if (newprio > 35)
 786                 newprio = 35;
 787         current->priority = newprio;
 788         return 0;
 789 }
 790 
 791 static void show_task(int nr,struct task_struct * p)
     /*  */
 792 {
 793         unsigned long free;
 794         static char * stat_nam[] = { "R", "S", "D", "Z", "T", "W" };
 795 
 796         printk("%-8s %3d ", p->comm, (p == current) ? -nr : nr);
 797         if (((unsigned) p->state) < sizeof(stat_nam)/sizeof(char *))
 798                 printk(stat_nam[p->state]);
 799         else
 800                 printk(" ");
 801         if (p == current)
 802                 printk(" current  ");
 803         else
 804                 printk(" %08lX ", ((unsigned long *)p->tss.esp)[3]);
 805         for (free = 1; free < 1024 ; free++) {
 806                 if (((unsigned long *)p->kernel_stack_page)[free])
 807                         break;
 808         }
 809         printk("%5lu %5d %6d ", free << 2, p->pid, p->p_pptr->pid);
 810         if (p->p_cptr)
 811                 printk("%5d ", p->p_cptr->pid);
 812         else
 813                 printk("      ");
 814         if (p->p_ysptr)
 815                 printk("%7d", p->p_ysptr->pid);
 816         else
 817                 printk("       ");
 818         if (p->p_osptr)
 819                 printk(" %5d\n", p->p_osptr->pid);
 820         else
 821                 printk("\n");
 822 }
 823 
 824 void show_state(void)
     /*  */
 825 {
 826         int i;
 827 
 828         printk("                         free                        sibling\n");
 829         printk("  task             PC    stack   pid father child younger older\n");
 830         for (i=0 ; i<NR_TASKS ; i++)
 831                 if (task[i])
 832                         show_task(i,task[i]);
 833 }
 834 
 835 void sched_init(void)
     /*  */
 836 {
 837         int i;
 838         struct desc_struct * p;
 839 
 840         bh_base[TIMER_BH].routine = timer_bh;
 841         bh_base[TQUEUE_BH].routine = tqueue_bh;
 842         bh_base[IMMEDIATE_BH].routine = immediate_bh;
 843         if (sizeof(struct sigaction) != 16)
 844                 panic("Struct sigaction MUST be 16 bytes");
 845         set_tss_desc(gdt+FIRST_TSS_ENTRY,&init_task.tss);
 846         set_ldt_desc(gdt+FIRST_LDT_ENTRY,&default_ldt,1);
 847         set_system_gate(0x80,&system_call);
 848         p = gdt+2+FIRST_TSS_ENTRY;
 849         for(i=1 ; i<NR_TASKS ; i++) {
 850                 task[i] = NULL;
 851                 p->a=p->b=0;
 852                 p++;
 853                 p->a=p->b=0;
 854                 p++;
 855         }
 856 /* Clear NT, so that we won't have troubles with that later on */
 857         __asm__("pushfl ; andl $0xffffbfff,(%esp) ; popfl");
 858         load_TR(0);
 859         load_ldt(0);
 860         outb_p(0x34,0x43);              /* binary, mode 2, LSB/MSB, ch 0 */
 861         outb_p(LATCH & 0xff , 0x40);    /* LSB */
 862         outb(LATCH >> 8 , 0x40);        /* MSB */
 863         if (request_irq(TIMER_IRQ,(void (*)(int)) do_timer, 0, "timer") != 0)
 864                 panic("Could not allocate timer IRQ!");
 865 }
/* */
root/kernel/sched.c

DEFINITIONS
