kernel/sched.c

/* */
This source file includes following definitions.
math_state_restore
schedule
sys_pause
wake_up
wake_up_interruptible
__sleep_on
interruptible_sleep_on
sleep_on
ticks_to_floppy_on
floppy_off
do_floppy_timer
add_timer
count_active_tasks
calc_load
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 #define TIMER_IRQ 0
  15 
  16 #include <linux/config.h>
  17 #include <linux/signal.h>
  18 #include <linux/sched.h>
  19 #include <linux/timer.h>
  20 #include <linux/kernel.h>
  21 #include <linux/sys.h>
  22 #include <linux/fdreg.h>
  23 #include <linux/errno.h>
  24 #include <linux/time.h>
  25 #include <linux/ptrace.h>
  26 
  27 #include <asm/system.h>
  28 #include <asm/io.h>
  29 #include <asm/segment.h>
  30 
  31 int need_resched = 0;
  32 int hard_math = 0;              /* set by boot/head.S */
  33 int ignore_irq13 = 0;           /* set if exception 16 works */
  34 
  35 unsigned long * prof_buffer = NULL;
  36 unsigned long prof_len = 0;
  37 
  38 #define _S(nr) (1<<((nr)-1))
  39 
  40 #define LATCH ((1193180 + HZ/2)/HZ)
  41 
  42 extern void mem_use(void);
  43 
  44 extern int timer_interrupt(void);
  45 extern int system_call(void);
  46 
  47 static unsigned long init_kernel_stack[1024];
  48 static struct task_struct init_task = INIT_TASK;
  49 
  50 unsigned long volatile jiffies=0;
  51 unsigned long startup_time=0;
  52 int jiffies_offset = 0;         /* # clock ticks to add to get "true
  53                                    time".  Should always be less than
  54                                    1 second's worth.  For time fanatics
  55                                    who like to syncronize their machines
  56                                    to WWV :-) */
  57 
  58 struct task_struct *current = &init_task;
  59 struct task_struct *last_task_used_math = NULL;
  60 
  61 struct task_struct * task[NR_TASKS] = {&init_task, };
  62 
  63 long user_stack [ PAGE_SIZE>>2 ] ;
  64 
  65 struct {
  66         long * a;
  67         short b;
  68         } stack_start = { & user_stack [PAGE_SIZE>>2] , 0x10 };
  69 /*
  70  *  'math_state_restore()' saves the current math information in the
  71  * old math state array, and gets the new ones from the current task
  72  */
  73 void math_state_restore(void)
     /*  */
  74 {
  75         if (last_task_used_math == current)
  76                 return;
  77         timer_table[COPRO_TIMER].expires = jiffies+50;
  78         timer_active |= 1<<COPRO_TIMER; 
  79         if (last_task_used_math) {
  80                 __asm__("fnsave %0"::"m" (last_task_used_math->tss.i387));
  81         }
  82         __asm__("fwait");
  83         last_task_used_math = current;
  84         if (current->used_math) {
  85                 __asm__("frstor %0"::"m" (current->tss.i387));
  86         } else {
  87                 __asm__("fninit"::);
  88                 current->used_math=1;
  89         }
  90         timer_active &= ~(1<<COPRO_TIMER);
  91 }
  92 
  93 /*
  94  *  'schedule()' is the scheduler function. It's a very simple and nice
  95  * scheduler: it's not perfect, but certainly works for most things.
  96  * The one thing you might take a look at is the signal-handler code here.
  97  *
  98  *   NOTE!!  Task 0 is the 'idle' task, which gets called when no other
  99  * tasks can run. It can not be killed, and it cannot sleep. The 'state'
 100  * information in task[0] is never used.
 101  */
 102 void schedule(void)
     /*  */
 103 {
 104         int i,next,c;
 105         struct task_struct ** p;
 106 
 107 /* check alarm, wake up any interruptible tasks that have got a signal */
 108 
 109         need_resched = 0;
 110         for(p = &LAST_TASK ; p > &FIRST_TASK ; --p) {
 111                 if (!*p || ((*p)->state != TASK_INTERRUPTIBLE))
 112                         continue;
 113                 if ((*p)->timeout && (*p)->timeout < jiffies) {
 114                         (*p)->timeout = 0;
 115                         (*p)->state = TASK_RUNNING;
 116                 } else if ((*p)->signal & ~(*p)->blocked)
 117                         (*p)->state = TASK_RUNNING;
 118         }
 119 
 120 /* this is the scheduler proper: */
 121 
 122         while (1) {
 123                 c = -1;
 124                 next = 0;
 125                 i = NR_TASKS;
 126                 p = &task[NR_TASKS];
 127                 while (--i) {
 128                         if (!*--p)
 129                                 continue;
 130                         if ((*p)->state == TASK_RUNNING && (*p)->counter > c)
 131                                 c = (*p)->counter, next = i;
 132                 }
 133                 if (c)
 134                         break;
 135                 for(p = &LAST_TASK ; p > &FIRST_TASK ; --p)
 136                         if (*p)
 137                                 (*p)->counter = ((*p)->counter >> 1) +
 138                                                 (*p)->priority;
 139         }
 140         sti();
 141         switch_to(next);
 142 }
 143 
 144 int sys_pause(void)
     /*  */
 145 {
 146         current->state = TASK_INTERRUPTIBLE;
 147         schedule();
 148         return -ERESTARTNOHAND;
 149 }
 150 
 151 /*
 152  * wake_up doesn't wake up stopped processes - they have to be awakened
 153  * with signals or similar.
 154  *
 155  * Note that this doesn't need cli-sti pairs: interrupts may not change
 156  * the wait-queue structures directly, but only call wake_up() to wake
 157  * a process. The process itself must remove the queue once it has woken.
 158  */
 159 void wake_up(struct wait_queue **q)
     /*  */
 160 {
 161         struct wait_queue *tmp;
 162         struct task_struct * p;
 163 
 164         if (!q || !(tmp = *q))
 165                 return;
 166         do {
 167                 if ((p = tmp->task) != NULL) {
 168                         if ((p->state == TASK_UNINTERRUPTIBLE) ||
 169                             (p->state == TASK_INTERRUPTIBLE)) {
 170                                 p->state = TASK_RUNNING;
 171                                 if (p->counter > current->counter)
 172                                         need_resched = 1;
 173                         }
 174                 }
 175                 if (!tmp->next) {
 176                         printk("wait_queue is bad (eip = %08x)\n",((unsigned long *) q)[-1]);
 177                         printk("        q = %08x\n",q);
 178                         printk("       *q = %08x\n",*q);
 179                         printk("      tmp = %08x\n",tmp);
 180                         break;
 181                 }
 182                 tmp = tmp->next;
 183         } while (tmp != *q);
 184 }
 185 
 186 void wake_up_interruptible(struct wait_queue **q)
     /*  */
 187 {
 188         struct wait_queue *tmp;
 189         struct task_struct * p;
 190 
 191         if (!q || !(tmp = *q))
 192                 return;
 193         do {
 194                 if ((p = tmp->task) != NULL) {
 195                         if (p->state == TASK_INTERRUPTIBLE) {
 196                                 p->state = TASK_RUNNING;
 197                                 if (p->counter > current->counter)
 198                                         need_resched = 1;
 199                         }
 200                 }
 201                 if (!tmp->next) {
 202                         printk("wait_queue is bad (eip = %08x)\n",((unsigned long *) q)[-1]);
 203                         printk("        q = %08x\n",q);
 204                         printk("       *q = %08x\n",*q);
 205                         printk("      tmp = %08x\n",tmp);
 206                         break;
 207                 }
 208                 tmp = tmp->next;
 209         } while (tmp != *q);
 210 }
 211 
 212 static inline void __sleep_on(struct wait_queue **p, int state)
     /*  */
 213 {
 214         unsigned long flags;
 215         struct wait_queue wait = { current, NULL };
 216 
 217         if (!p)
 218                 return;
 219         if (current == task[0])
 220                 panic("task[0] trying to sleep");
 221         current->state = state;
 222         add_wait_queue(p, &wait);
 223         save_flags(flags);
 224         sti();
 225         schedule();
 226         remove_wait_queue(p, &wait);
 227         restore_flags(flags);
 228 }
 229 
 230 void interruptible_sleep_on(struct wait_queue **p)
     /*  */
 231 {
 232         __sleep_on(p,TASK_INTERRUPTIBLE);
 233 }
 234 
 235 void sleep_on(struct wait_queue **p)
     /*  */
 236 {
 237         __sleep_on(p,TASK_UNINTERRUPTIBLE);
 238 }
 239 
 240 /*
 241  * OK, here are some floppy things that shouldn't be in the kernel
 242  * proper. They are here because the floppy needs a timer, and this
 243  * was the easiest way of doing it.
 244  */
 245 static struct wait_queue * wait_motor[4] = {NULL,NULL,NULL,NULL};
 246 static int  mon_timer[4]={0,0,0,0};
 247 static int moff_timer[4]={0,0,0,0};
 248 unsigned char current_DOR = 0x0C;
 249 
 250 int ticks_to_floppy_on(unsigned int nr)
     /*  */
 251 {
 252         extern unsigned char selected;
 253         unsigned char mask = 0x10 << nr;
 254 
 255         if (nr>3)
 256                 panic("floppy_on: nr>3");
 257         moff_timer[nr]=10000;           /* 100 s = very big :-) */
 258         cli();                          /* use floppy_off to turn it off */
 259         mask |= current_DOR;
 260         if (!selected) {
 261                 mask &= 0xFC;
 262                 mask |= nr;
 263         }
 264         if (mask != current_DOR) {
 265                 outb(mask,FD_DOR);
 266                 if ((mask ^ current_DOR) & 0xf0)
 267                         mon_timer[nr] = HZ/2;
 268                 else if (mon_timer[nr] < 2)
 269                         mon_timer[nr] = 2;
 270                 current_DOR = mask;
 271         }
 272         sti();
 273         return mon_timer[nr];
 274 }
 275 
 276 void floppy_off(unsigned int nr)
     /*  */
 277 {
 278         moff_timer[nr]=3*HZ;
 279 }
 280 
 281 void do_floppy_timer(void)
     /*  */
 282 {
 283         int i;
 284         unsigned char mask = 0x10;
 285 
 286         for (i=0 ; i<4 ; i++,mask <<= 1) {
 287                 if (!(mask & current_DOR))
 288                         continue;
 289                 if (mon_timer[i]) {
 290                         if (!--mon_timer[i])
 291                                 wake_up(i+wait_motor);
 292                 } else if (!moff_timer[i]) {
 293                         current_DOR &= ~mask;
 294                         outb(current_DOR,FD_DOR);
 295                 } else
 296                         moff_timer[i]--;
 297         }
 298 }
 299 
 300 #define TIME_REQUESTS 64
 301 
 302 static struct timer_list {
 303         long jiffies;
 304         void (*fn)(void);
 305         struct timer_list * next;
 306 } timer_list[TIME_REQUESTS] = { { 0, NULL, NULL }, };
 307 
 308 static struct timer_list * next_timer = NULL;
 309 
 310 void add_timer(long jiffies, void (*fn)(void))
     /*  */
 311 {
 312         struct timer_list * p;
 313         unsigned long flags;
 314 
 315         if (!fn)
 316                 return;
 317         save_flags(flags);
 318         cli();
 319         if (jiffies <= 0)
 320                 (fn)();
 321         else {
 322                 for (p = timer_list ; p < timer_list + TIME_REQUESTS ; p++)
 323                         if (!p->fn)
 324                                 break;
 325                 if (p >= timer_list + TIME_REQUESTS)
 326                         panic("No more time requests free");
 327                 p->fn = fn;
 328                 p->jiffies = jiffies;
 329                 p->next = next_timer;
 330                 next_timer = p;
 331                 while (p->next && p->next->jiffies < p->jiffies) {
 332                         p->jiffies -= p->next->jiffies;
 333                         fn = p->fn;
 334                         p->fn = p->next->fn;
 335                         p->next->fn = fn;
 336                         jiffies = p->jiffies;
 337                         p->jiffies = p->next->jiffies;
 338                         p->next->jiffies = jiffies;
 339                         p = p->next;
 340                 }
 341                 if (p->next)
 342                         p->next->jiffies -= p->jiffies;
 343         }
 344         restore_flags(flags);
 345 }
 346 
 347 unsigned long timer_active = 0;
 348 struct timer_struct timer_table[32];
 349 
 350 /*
 351  * Hmm.. Changed this, as the GNU make sources (load.c) seems to
 352  * imply that avenrun[] is the standard name for this kind of thing.
 353  * Nothing else seems to be standardized: the fractional size etc
 354  * all seem to differ on different machines.
 355  */
 356 unsigned long avenrun[3] = { 0,0,0 };
 357 
 358 /*
 359  * Nr of active tasks - counted in fixed-point numbers
 360  */
 361 static unsigned long count_active_tasks(void)
     /*  */
 362 {
 363         struct task_struct **p;
 364         unsigned long nr = 0;
 365 
 366         for(p = &LAST_TASK; p > &FIRST_TASK; --p)
 367                 if (*p && (*p)->state == TASK_RUNNING)
 368                         nr += FIXED_1;
 369         return nr;
 370 }
 371 
 372 static inline void calc_load(void)
     /*  */
 373 {
 374         unsigned long active_tasks; /* fixed-point */
 375         static int count = LOAD_FREQ;
 376 
 377         if (count-- > 0)
 378                 return;
 379         count = LOAD_FREQ;
 380         active_tasks = count_active_tasks();
 381         CALC_LOAD(avenrun[0], EXP_1, active_tasks);
 382         CALC_LOAD(avenrun[1], EXP_5, active_tasks);
 383         CALC_LOAD(avenrun[2], EXP_15, active_tasks);
 384 }
 385 
 386 /*
 387  * The int argument is really a (struct pt_regs *), in case the
 388  * interrupt wants to know from where it was called. The timer
 389  * irq uses this to decide if it should update the user or system
 390  * times.
 391  */
 392 static void do_timer(struct pt_regs * regs)
     /*  */
 393 {
 394         unsigned long mask;
 395         struct timer_struct *tp = timer_table+0;
 396         struct task_struct ** task_p;
 397 
 398         jiffies++;
 399         calc_load();
 400         if ((VM_MASK & regs->eflags) || (3 & regs->cs)) {
 401                 current->utime++;
 402                 /* Update ITIMER_VIRT for current task if not in a system call */
 403                 if (current->it_virt_value && !(--current->it_virt_value)) {
 404                         current->it_virt_value = current->it_virt_incr;
 405                         send_sig(SIGVTALRM,current,1);
 406                 }
 407         } else {
 408                 current->stime++;
 409 #ifdef CONFIG_PROFILE
 410                 if (prof_buffer && current != task[0]) {
 411                         unsigned long eip = regs->eip;
 412                         eip >>= 2;
 413                         if (eip < prof_len)
 414                                 prof_buffer[eip]++;
 415                 }
 416 #endif
 417         }
 418         if (current == task[0] || (--current->counter)<=0) {
 419                 current->counter=0;
 420                 need_resched = 1;
 421         }
 422         /* Update ITIMER_REAL for every task */
 423         for (task_p = &LAST_TASK; task_p >= &FIRST_TASK; task_p--)
 424                 if (*task_p && (*task_p)->it_real_value
 425                         && !(--(*task_p)->it_real_value)) {
 426                         send_sig(SIGALRM,*task_p,1);
 427                         (*task_p)->it_real_value = (*task_p)->it_real_incr;
 428                         need_resched = 1;
 429                 }
 430         /* Update ITIMER_PROF for the current task */
 431         if (current->it_prof_value && !(--current->it_prof_value)) {
 432                 current->it_prof_value = current->it_prof_incr;
 433                 send_sig(SIGPROF,current,1);
 434         }
 435         for (mask = 1 ; mask ; tp++,mask += mask) {
 436                 if (mask > timer_active)
 437                         break;
 438                 if (!(mask & timer_active))
 439                         continue;
 440                 if (tp->expires > jiffies)
 441                         continue;
 442                 timer_active &= ~mask;
 443                 tp->fn();
 444                 sti();
 445         }
 446         if (next_timer) {
 447                 next_timer->jiffies--;
 448                 while (next_timer && next_timer->jiffies <= 0) {
 449                         void (*fn)(void);
 450                         
 451                         fn = next_timer->fn;
 452                         next_timer->fn = NULL;
 453                         next_timer = next_timer->next;
 454                         (fn)();
 455                 }
 456         }
 457         if (current_DOR & 0xf0)
 458                 do_floppy_timer();
 459 }
 460 
 461 int sys_alarm(long seconds)
     /*  */
 462 {
 463         extern int _setitimer(int, struct itimerval *, struct itimerval *);
 464         struct itimerval new, old;
 465 
 466         new.it_interval.tv_sec = new.it_interval.tv_usec = 0;
 467         new.it_value.tv_sec = seconds;
 468         new.it_value.tv_usec = 0;
 469         _setitimer(ITIMER_REAL, &new, &old);
 470         return(old.it_value.tv_sec + (old.it_value.tv_usec / 1000000));
 471 }
 472 
 473 int sys_getpid(void)
     /*  */
 474 {
 475         return current->pid;
 476 }
 477 
 478 int sys_getppid(void)
     /*  */
 479 {
 480         return current->p_pptr->pid;
 481 }
 482 
 483 int sys_getuid(void)
     /*  */
 484 {
 485         return current->uid;
 486 }
 487 
 488 int sys_geteuid(void)
     /*  */
 489 {
 490         return current->euid;
 491 }
 492 
 493 int sys_getgid(void)
     /*  */
 494 {
 495         return current->gid;
 496 }
 497 
 498 int sys_getegid(void)
     /*  */
 499 {
 500         return current->egid;
 501 }
 502 
 503 int sys_nice(long increment)
     /*  */
 504 {
 505         if (increment < 0 && !suser())
 506                 return -EPERM;
 507         if (increment >= current->priority)
 508                 increment = current->priority-1;
 509         current->priority -= increment;
 510         return 0;
 511 }
 512 
 513 static void show_task(int nr,struct task_struct * p)
     /*  */
 514 {
 515         int i, j;
 516         unsigned char * stack;
 517 
 518         printk("%d: pid=%d, state=%d, father=%d, child=%d, ",(p == current)?-nr:nr,p->pid,
 519                 p->state, p->p_pptr->pid, p->p_cptr ? p->p_cptr->pid : -1);
 520         i = 0;
 521         j = 4096;
 522         if (!(stack = (char *) p->kernel_stack_page)) {
 523                 stack = (char *) init_kernel_stack;
 524                 j = sizeof(init_kernel_stack);
 525         }
 526         while (i<j && !*(stack++))
 527                 i++;
 528         printk("%d/%d chars free in kstack\n",i,j);
 529         printk("   PC=%08X.", *(1019 + (unsigned long *) p));
 530         if (p->p_ysptr || p->p_osptr) 
 531                 printk("   Younger sib=%d, older sib=%d\n", 
 532                         p->p_ysptr ? p->p_ysptr->pid : -1,
 533                         p->p_osptr ? p->p_osptr->pid : -1);
 534         else
 535                 printk("\n");
 536 }
 537 
 538 void show_state(void)
     /*  */
 539 {
 540         int i;
 541 
 542         printk("Task-info:\n");
 543         for (i=0 ; i<NR_TASKS ; i++)
 544                 if (task[i])
 545                         show_task(i,task[i]);
 546 }
 547 
 548 void sched_init(void)
     /*  */
 549 {
 550         int i;
 551         struct desc_struct * p;
 552 
 553         if (sizeof(struct sigaction) != 16)
 554                 panic("Struct sigaction MUST be 16 bytes");
 555         set_tss_desc(gdt+FIRST_TSS_ENTRY,&init_task.tss);
 556         set_ldt_desc(gdt+FIRST_LDT_ENTRY,&init_task.ldt);
 557         set_system_gate(0x80,&system_call);
 558         p = gdt+2+FIRST_TSS_ENTRY;
 559         for(i=1 ; i<NR_TASKS ; i++) {
 560                 task[i] = NULL;
 561                 p->a=p->b=0;
 562                 p++;
 563                 p->a=p->b=0;
 564                 p++;
 565         }
 566 /* Clear NT, so that we won't have troubles with that later on */
 567         __asm__("pushfl ; andl $0xffffbfff,(%esp) ; popfl");
 568         load_TR(0);
 569         load_ldt(0);
 570         outb_p(0x34,0x43);              /* binary, mode 2, LSB/MSB, ch 0 */
 571         outb_p(LATCH & 0xff , 0x40);    /* LSB */
 572         outb(LATCH >> 8 , 0x40);        /* MSB */
 573         request_irq(TIMER_IRQ,(void (*)(int)) do_timer);
 574 }
/* */
root/kernel/sched.c

DEFINITIONS
