kernel/system.c

   1 /* This task handles the interface between the kernel and user-level servers.
   2  * System services can be accessed by doing a system call. System calls are
   3  * transformed into request messages, which are handled by this task. By
   4  * convention, a sys_call() is transformed in a SYS_CALL request message that
   5  * is handled in a function named do_call().
   6  *
   7  * A private call vector is used to map all system calls to the functions that
   8  * handle them. The actual handler functions are contained in separate files
   9  * to keep this file clean. The call vector is used in the system task's main
  10  * loop to handle all incoming requests.
  11  *
  12  * In addition to the main sys_task() entry point, which starts the main loop,
  13  * there are several other minor entry points:
  14  *   get_priv:          assign privilege structure to user or system process
  15  *   set_sendto_bit:    allow a process to send messages to a new target
  16  *   unset_sendto_bit:  disallow a process from sending messages to a target
  17  *   fill_sendto_mask:  fill the target mask of a given process
  18  *   send_sig:          send a signal directly to a system process
  19  *   cause_sig:         take action to cause a signal to occur via a signal mgr
  20  *   sig_delay_done:    tell PM that a process is not sending
  21  *   get_randomness:    accumulate randomness in a buffer
  22  *   clear_endpoint:    remove a process' ability to send and receive messages
  23  *   sched_proc:        schedule a process
  24  *
  25  * Changes:
  26 *    Nov 22, 2009   get_priv supports static priv ids (Cristiano Giuffrida)
  27  *   Aug 04, 2005   check if system call is allowed  (Jorrit N. Herder)
  28  *   Jul 20, 2005   send signal to services with message  (Jorrit N. Herder)
  29  *   Jan 15, 2005   new, generalized virtual copy function  (Jorrit N. Herder)
  30  *   Oct 10, 2004   dispatch system calls from call vector  (Jorrit N. Herder)
  31  *   Sep 30, 2004   source code documentation updated  (Jorrit N. Herder)
  32  */
  33
  34 #include "kernel.h"
  35 #include "system.h"
  36 #include "vm.h"
  37 #include "kernel/clock.h"
  38 #include <stdlib.h>
  39 #include <assert.h>
  40 #include <signal.h>
  41 #include <unistd.h>
  42 #include <minix/endpoint.h>
  43 #include <minix/safecopies.h>
  44
  45 /* Declaration of the call vector that defines the mapping of system calls
  46  * to handler functions. The vector is initialized in sys_init() with map(),
  47  * which makes sure the system call numbers are ok. No space is allocated,
  48  * because the dummy is declared extern. If an illegal call is given, the
  49  * array size will be negative and this won't compile.
  50  */
  51 static int (*call_vec[NR_SYS_CALLS])(struct proc * caller, message *m_ptr);
  52
  53 #define map(call_nr, handler)                                   \
  54     {   int call_index = call_nr-KERNEL_CALL;                           \
  55         assert(call_index >= 0 && call_index < NR_SYS_CALLS);                   \
  56     call_vec[call_index] = (handler)  ; }
  57
  58 static void kernel_call_finish(struct proc * caller, message *msg, int result)
  59 {
  60   if(result == VMSUSPEND) {
  61           /* Special case: message has to be saved for handling
  62            * until VM tells us it's allowed. VM has been notified
  63            * and we must wait for its reply to restart the call.
  64            */
  65           assert(RTS_ISSET(caller, RTS_VMREQUEST));
  66           assert(caller->p_vmrequest.type == VMSTYPE_KERNELCALL);
  67           caller->p_vmrequest.saved.reqmsg = *msg;
  68           caller->p_misc_flags |= MF_KCALL_RESUME;
  69   } else {
  70           /*
  71            * call is finished, we could have been suspended because of VM,
  72            * remove the request message
  73            */
  74           caller->p_vmrequest.saved.reqmsg.m_source = NONE;
  75           if (result != EDONTREPLY) {
  76                   /* copy the result as a message to the original user buffer */
  77                   msg->m_source = SYSTEM;
  78                   msg->m_type = result;         /* report status of call */
  79 #if DEBUG_IPC_HOOK
  80         hook_ipc_msgkresult(msg, caller);
  81 #endif
  82                   if (copy_msg_to_user(msg, (message *)caller->p_delivermsg_vir)) {
  83                           printf("WARNING wrong user pointer 0x%08x from "
  84                                           "process %s / %d\n",
  85                                           caller->p_delivermsg_vir,
  86                                           caller->p_name,
  87                                           caller->p_endpoint);
  88                           cause_sig(proc_nr(caller), SIGSEGV);
  89                   }
  90           }
  91   }
  92 }
  93
  94 static int kernel_call_dispatch(struct proc * caller, message *msg)
  95 {
  96   int result = OK;
  97   int call_nr;
  98
  99 #if DEBUG_IPC_HOOK
 100         hook_ipc_msgkcall(msg, caller);
 101 #endif
 102   call_nr = msg->m_type - KERNEL_CALL;
 103
 104   /* See if the caller made a valid request and try to handle it. */
 105   if (call_nr < 0 || call_nr >= NR_SYS_CALLS) { /* check call number */
 106           printf("SYSTEM: illegal request %d from %d.\n",
 107                           call_nr,msg->m_source);
 108           result = EBADREQUEST;                 /* illegal message type */
 109   }
 110   else if (!GET_BIT(priv(caller)->s_k_call_mask, call_nr)) {
 111           printf("SYSTEM: denied request %d from %d.\n",
 112                           call_nr,msg->m_source);
 113           result = ECALLDENIED;                 /* illegal message type */
 114   } else {
 115           /* handle the system call */
 116           if (call_vec[call_nr])
 117                   result = (*call_vec[call_nr])(caller, msg);
 118           else {
 119                   printf("Unused kernel call %d from %d\n",
 120                                   call_nr, caller->p_endpoint);
 121                   result = EBADREQUEST;
 122           }
 123   }
 124
 125   return result;
 126 }
 127
 128 /*===========================================================================*
 129  *                              kernel_call                                  *
 130  *===========================================================================*/
 131 /*
 132  * this function checks the basic syscall parameters and if accepted it
 133  * dispatches its handling to the right handler
 134  */
 135 void kernel_call(message *m_user, struct proc * caller)
 136 {
 137   int result = OK;
 138   message msg;
 139
 140   caller->p_delivermsg_vir = (vir_bytes) m_user;
 141   /*
 142    * the ldt and cr3 of the caller process is loaded because it just've trapped
 143    * into the kernel or was already set in switch_to_user() before we resume
 144    * execution of an interrupted kernel call
 145    */
 146   if (copy_msg_from_user(m_user, &msg) == 0) {
 147           msg.m_source = caller->p_endpoint;
 148           result = kernel_call_dispatch(caller, &msg);
 149   }
 150   else {
 151           printf("WARNING wrong user pointer 0x%08x from process %s / %d\n",
 152                           m_user, caller->p_name, caller->p_endpoint);
 153           cause_sig(proc_nr(caller), SIGSEGV);
 154           return;
 155   }
 156
 157
 158   /* remember who invoked the kcall so we can bill it its time */
 159   kbill_kcall = caller;
 160
 161   kernel_call_finish(caller, &msg, result);
 162 }
 163
 164 /*===========================================================================*
 165  *                              initialize                                   *
 166  *===========================================================================*/
 167 void system_init(void)
 168 {
 169   register struct priv *sp;
 170   int i;
 171
 172   /* Initialize IRQ handler hooks. Mark all hooks available. */
 173   for (i=0; i<NR_IRQ_HOOKS; i++) {
 174       irq_hooks[i].proc_nr_e = NONE;
 175   }
 176
 177   /* Initialize all alarm timers for all processes. */
 178   for (sp=BEG_PRIV_ADDR; sp < END_PRIV_ADDR; sp++) {
 179     tmr_inittimer(&(sp->s_alarm_timer));
 180   }
 181
 182   /* Initialize the call vector to a safe default handler. Some system calls
 183    * may be disabled or nonexistant. Then explicitely map known calls to their
 184    * handler functions. This is done with a macro that gives a compile error
 185    * if an illegal call number is used. The ordering is not important here.
 186    */
 187   for (i=0; i<NR_SYS_CALLS; i++) {
 188       call_vec[i] = NULL;
 189   }
 190
 191   /* Process management. */
 192   map(SYS_FORK, do_fork);               /* a process forked a new process */
 193   map(SYS_EXEC, do_exec);               /* update process after execute */
 194   map(SYS_CLEAR, do_clear);             /* clean up after process exit */
 195   map(SYS_EXIT, do_exit);               /* a system process wants to exit */
 196   map(SYS_PRIVCTL, do_privctl);         /* system privileges control */
 197   map(SYS_TRACE, do_trace);             /* request a trace operation */
 198   map(SYS_SETGRANT, do_setgrant);       /* get/set own parameters */
 199   map(SYS_RUNCTL, do_runctl);           /* set/clear stop flag of a process */
 200   map(SYS_UPDATE, do_update);           /* update a process into another */
 201   map(SYS_STATECTL, do_statectl);       /* let a process control its state */
 202
 203   /* Signal handling. */
 204   map(SYS_KILL, do_kill);               /* cause a process to be signaled */
 205   map(SYS_GETKSIG, do_getksig);         /* signal manager checks for signals */
 206   map(SYS_ENDKSIG, do_endksig);         /* signal manager finished signal */
 207   map(SYS_SIGSEND, do_sigsend);         /* start POSIX-style signal */
 208   map(SYS_SIGRETURN, do_sigreturn);     /* return from POSIX-style signal */
 209
 210   /* Device I/O. */
 211   map(SYS_IRQCTL, do_irqctl);           /* interrupt control operations */
 212 #if defined(__i386__)
 213   map(SYS_DEVIO, do_devio);             /* inb, inw, inl, outb, outw, outl */
 214   map(SYS_VDEVIO, do_vdevio);           /* vector with devio requests */
 215 #endif
 216
 217   /* Memory management. */
 218   map(SYS_MEMSET, do_memset);           /* write char to memory area */
 219   map(SYS_VMCTL, do_vmctl);             /* various VM process settings */
 220
 221   /* Copying. */
 222   map(SYS_UMAP, do_umap);               /* map virtual to physical address */
 223   map(SYS_UMAP_REMOTE, do_umap_remote); /* do_umap for non-caller process */
 224   map(SYS_VUMAP, do_vumap);             /* vectored virtual to physical map */
 225   map(SYS_VIRCOPY, do_vircopy);         /* use pure virtual addressing */
 226   map(SYS_PHYSCOPY, do_copy);           /* use physical addressing */
 227   map(SYS_SAFECOPYFROM, do_safecopy_from);/* copy with pre-granted permission */
 228   map(SYS_SAFECOPYTO, do_safecopy_to);  /* copy with pre-granted permission */
 229   map(SYS_VSAFECOPY, do_vsafecopy);     /* vectored safecopy */
 230
 231   /* safe memset */
 232   map(SYS_SAFEMEMSET, do_safememset);   /* safememset */
 233
 234   /* Clock functionality. */
 235   map(SYS_TIMES, do_times);             /* get uptime and process times */
 236   map(SYS_SETALARM, do_setalarm);       /* schedule a synchronous alarm */
 237   map(SYS_STIME, do_stime);             /* set the boottime */
 238   map(SYS_VTIMER, do_vtimer);           /* set or retrieve a virtual timer */
 239
 240   /* System control. */
 241   map(SYS_ABORT, do_abort);             /* abort MINIX */
 242   map(SYS_GETINFO, do_getinfo);         /* request system information */
 243   map(SYS_SYSCTL, do_sysctl);           /* misc system manipulation */
 244
 245   /* Profiling. */
 246   map(SYS_SPROF, do_sprofile);         /* start/stop statistical profiling */
 247   map(SYS_CPROF, do_cprofile);         /* get/reset call profiling data */
 248   map(SYS_PROFBUF, do_profbuf);        /* announce locations to kernel */
 249
 250   /* i386-specific. */
 251 #if defined(__i386__)
 252   map(SYS_READBIOS, do_readbios);       /* read from BIOS locations */
 253   map(SYS_IOPENABLE, do_iopenable);     /* Enable I/O */
 254   map(SYS_SDEVIO, do_sdevio);           /* phys_insb, _insw, _outsb, _outsw */
 255 #endif
 256
 257   /* Machine state switching. */
 258   map(SYS_SETMCONTEXT, do_setmcontext); /* set machine context */
 259   map(SYS_GETMCONTEXT, do_getmcontext); /* get machine context */
 260
 261   /* Scheduling */
 262   map(SYS_SCHEDULE, do_schedule);       /* reschedule a process */
 263   map(SYS_SCHEDCTL, do_schedctl);       /* change process scheduler */
 264
 265 }
 266 /*===========================================================================*
 267  *                              get_priv                                     *
 268  *===========================================================================*/
 269 int get_priv(rc, priv_id)
 270 register struct proc *rc;               /* new (child) process pointer */
 271 int priv_id;                            /* privilege id */
 272 {
 273 /* Allocate a new privilege structure for a system process. Privilege ids
 274  * can be assigned either statically or dynamically.
 275  */
 276   register struct priv *sp;                 /* privilege structure */
 277
 278   if(priv_id == NULL_PRIV_ID) {             /* allocate slot dynamically */
 279       for (sp = BEG_DYN_PRIV_ADDR; sp < END_DYN_PRIV_ADDR; ++sp)
 280           if (sp->s_proc_nr == NONE) break;
 281       if (sp >= END_DYN_PRIV_ADDR) return(ENOSPC);
 282   }
 283   else {                                    /* allocate slot from id */
 284       if(!is_static_priv_id(priv_id)) {
 285           return EINVAL;                    /* invalid static priv id */
 286       }
 287       if(priv[priv_id].s_proc_nr != NONE) {
 288           return EBUSY;                     /* slot already in use */
 289       }
 290       sp = &priv[priv_id];
 291   }
 292   rc->p_priv = sp;                          /* assign new slot */
 293   rc->p_priv->s_proc_nr = proc_nr(rc);      /* set association */
 294
 295   return(OK);
 296 }
 297
 298 /*===========================================================================*
 299  *                              set_sendto_bit                               *
 300  *===========================================================================*/
 301 void set_sendto_bit(const struct proc *rp, int id)
 302 {
 303 /* Allow a process to send messages to the process(es) associated with the
 304  * system privilege structure with the given ID.
 305  */
 306
 307   /* Disallow the process from sending to a process privilege structure with no
 308    * associated process, and disallow the process from sending to itself.
 309    */
 310   if (id_to_nr(id) == NONE || priv_id(rp) == id) {
 311         unset_sys_bit(priv(rp)->s_ipc_to, id);
 312         return;
 313   }
 314
 315   set_sys_bit(priv(rp)->s_ipc_to, id);
 316
 317   /* The process that this process can now send to, must be able to reply (or
 318    * vice versa). Therefore, its send mask should be updated as well. Ignore
 319    * receivers that don't support traps other than RECEIVE, they can't reply
 320    * or send messages anyway.
 321    */
 322   if (priv_addr(id)->s_trap_mask & ~((1 << RECEIVE)))
 323       set_sys_bit(priv_addr(id)->s_ipc_to, priv_id(rp));
 324 }
 325
 326 /*===========================================================================*
 327  *                              unset_sendto_bit                             *
 328  *===========================================================================*/
 329 void unset_sendto_bit(const struct proc *rp, int id)
 330 {
 331 /* Prevent a process from sending to another process. Retain the send mask
 332  * symmetry by also unsetting the bit for the other direction.
 333  */
 334
 335   unset_sys_bit(priv(rp)->s_ipc_to, id);
 336
 337   unset_sys_bit(priv_addr(id)->s_ipc_to, priv_id(rp));
 338 }
 339
 340 /*===========================================================================*
 341  *                            fill_sendto_mask                               *
 342  *===========================================================================*/
 343 void fill_sendto_mask(const struct proc *rp, sys_map_t *map)
 344 {
 345   int i;
 346
 347   for (i=0; i < NR_SYS_PROCS; i++) {
 348         if (get_sys_bit(*map, i))
 349                 set_sendto_bit(rp, i);
 350         else
 351                 unset_sendto_bit(rp, i);
 352   }
 353 }
 354
 355 /*===========================================================================*
 356  *                              send_sig                                     *
 357  *===========================================================================*/
 358 int send_sig(endpoint_t ep, int sig_nr)
 359 {
 360 /* Notify a system process about a signal. This is straightforward. Simply
 361  * set the signal that is to be delivered in the pending signals map and
 362  * send a notification with source SYSTEM.
 363  */
 364   register struct proc *rp;
 365   struct priv *priv;
 366   int proc_nr;
 367
 368   if(!isokendpt(ep, &proc_nr) || isemptyn(proc_nr))
 369         return EINVAL;
 370
 371   rp = proc_addr(proc_nr);
 372   priv = priv(rp);
 373   if(!priv) return ENOENT;
 374   sigaddset(&priv->s_sig_pending, sig_nr);
 375   mini_notify(proc_addr(SYSTEM), rp->p_endpoint);
 376
 377   return OK;
 378 }
 379
 380 /*===========================================================================*
 381  *                              cause_sig                                    *
 382  *===========================================================================*/
 383 void cause_sig(proc_nr, sig_nr)
 384 proc_nr_t proc_nr;              /* process to be signalled */
 385 int sig_nr;                     /* signal to be sent */
 386 {
 387 /* A system process wants to send a signal to a process.  Examples are:
 388  *  - HARDWARE wanting to cause a SIGSEGV after a CPU exception
 389  *  - TTY wanting to cause SIGINT upon getting a DEL
 390  *  - FS wanting to cause SIGPIPE for a broken pipe
 391  * Signals are handled by sending a message to the signal manager assigned to
 392  * the process. This function handles the signals and makes sure the signal
 393  * manager gets them by sending a notification. The process being signaled
 394  * is blocked while the signal manager has not finished all signals for it.
 395  * Race conditions between calls to this function and the system calls that
 396  * process pending kernel signals cannot exist. Signal related functions are
 397  * only called when a user process causes a CPU exception and from the kernel
 398  * process level, which runs to completion.
 399  */
 400   register struct proc *rp, *sig_mgr_rp;
 401   endpoint_t sig_mgr;
 402   int sig_mgr_proc_nr;
 403
 404   /* Lookup signal manager. */
 405   rp = proc_addr(proc_nr);
 406   sig_mgr = priv(rp)->s_sig_mgr;
 407   if(sig_mgr == SELF) sig_mgr = rp->p_endpoint;
 408
 409   /* If the target is the signal manager of itself, send the signal directly. */
 410   if(rp->p_endpoint == sig_mgr) {
 411        if(SIGS_IS_LETHAL(sig_nr)) {
 412            /* If the signal is lethal, see if a backup signal manager exists. */
 413            sig_mgr = priv(rp)->s_bak_sig_mgr;
 414            if(sig_mgr != NONE && isokendpt(sig_mgr, &sig_mgr_proc_nr)) {
 415                priv(rp)->s_sig_mgr = sig_mgr;
 416                priv(rp)->s_bak_sig_mgr = NONE;
 417                sig_mgr_rp = proc_addr(sig_mgr_proc_nr);
 418                RTS_UNSET(sig_mgr_rp, RTS_NO_PRIV);
 419                cause_sig(proc_nr, sig_nr); /* try again with the new sig mgr. */
 420                return;
 421            }
 422            /* We are out of luck. Time to panic. */
 423            proc_stacktrace(rp);
 424            panic("cause_sig: sig manager %d gets lethal signal %d for itself",
 425                 rp->p_endpoint, sig_nr);
 426        }
 427        sigaddset(&priv(rp)->s_sig_pending, sig_nr);
 428        if(OK != send_sig(rp->p_endpoint, SIGKSIGSM))
 429         panic("send_sig failed");
 430        return;
 431   }
 432
 433   /* Check if the signal is already pending. Process it otherwise. */
 434   if (! sigismember(&rp->p_pending, sig_nr)) {
 435       sigaddset(&rp->p_pending, sig_nr);
 436       if (! (RTS_ISSET(rp, RTS_SIGNALED))) {            /* other pending */
 437           RTS_SET(rp, RTS_SIGNALED | RTS_SIG_PENDING);
 438           if(OK != send_sig(sig_mgr, SIGKSIG))
 439                 panic("send_sig failed");
 440       }
 441   }
 442 }
 443
 444 /*===========================================================================*
 445  *                              sig_delay_done                               *
 446  *===========================================================================*/
 447 void sig_delay_done(struct proc *rp)
 448 {
 449 /* A process is now known not to send any direct messages.
 450  * Tell PM that the stop delay has ended, by sending a signal to the process.
 451  * Used for actual signal delivery.
 452  */
 453
 454   rp->p_misc_flags &= ~MF_SIG_DELAY;
 455
 456   cause_sig(proc_nr(rp), SIGSNDELAY);
 457 }
 458
 459 /*===========================================================================*
 460  *                               clear_ipc                                   *
 461  *===========================================================================*/
 462 static void clear_ipc(
 463   register struct proc *rc      /* slot of process to clean up */
 464 )
 465 {
 466 /* Clear IPC data for a given process slot. */
 467   struct proc **xpp;                    /* iterate over caller queue */
 468
 469   if (RTS_ISSET(rc, RTS_SENDING)) {
 470       int target_proc;
 471
 472       okendpt(rc->p_sendto_e, &target_proc);
 473       xpp = &proc_addr(target_proc)->p_caller_q; /* destination's queue */
 474       while (*xpp) {            /* check entire queue */
 475           if (*xpp == rc) {                     /* process is on the queue */
 476               *xpp = (*xpp)->p_q_link;          /* replace by next process */
 477 #if DEBUG_ENABLE_IPC_WARNINGS
 478               printf("endpoint %d / %s removed from queue at %d\n",
 479                   rc->p_endpoint, rc->p_name, rc->p_sendto_e);
 480 #endif
 481               break;                            /* can only be queued once */
 482           }
 483           xpp = &(*xpp)->p_q_link;              /* proceed to next queued */
 484       }
 485       RTS_UNSET(rc, RTS_SENDING);
 486   }
 487   RTS_UNSET(rc, RTS_RECEIVING);
 488 }
 489
 490 /*===========================================================================*
 491  *                               clear_endpoint                              *
 492  *===========================================================================*/
 493 void clear_endpoint(rc)
 494 register struct proc *rc;               /* slot of process to clean up */
 495 {
 496   if(isemptyp(rc)) panic("clear_proc: empty process: %d",  rc->p_endpoint);
 497
 498
 499 #if DEBUG_IPC_HOOK
 500   hook_ipc_clear(rc);
 501 #endif
 502
 503   /* Make sure that the exiting process is no longer scheduled. */
 504   RTS_SET(rc, RTS_NO_ENDPOINT);
 505   if (priv(rc)->s_flags & SYS_PROC)
 506   {
 507         priv(rc)->s_asynsize= 0;
 508   }
 509
 510   /* If the process happens to be queued trying to send a
 511    * message, then it must be removed from the message queues.
 512    */
 513   clear_ipc(rc);
 514
 515   /* Likewise, if another process was sending or receive a message to or from
 516    * the exiting process, it must be alerted that process no longer is alive.
 517    * Check all processes.
 518    */
 519   clear_ipc_refs(rc, EDEADSRCDST);
 520
 521 }
 522
 523 /*===========================================================================*
 524  *                             clear_ipc_refs                                *
 525  *===========================================================================*/
 526 void clear_ipc_refs(rc, caller_ret)
 527 register struct proc *rc;               /* slot of process to clean up */
 528 int caller_ret;                         /* code to return on callers */
 529 {
 530 /* Clear IPC references for a given process slot. */
 531   struct proc *rp;                      /* iterate over process table */
 532   int src_id;
 533
 534   /* Tell processes that sent asynchronous messages to 'rc' they are not
 535    * going to be delivered */
 536   while ((src_id = has_pending_asend(rc, ANY)) != NULL_PRIV_ID)
 537       cancel_async(proc_addr(id_to_nr(src_id)), rc);
 538
 539   for (rp = BEG_PROC_ADDR; rp < END_PROC_ADDR; rp++) {
 540       if(isemptyp(rp))
 541         continue;
 542
 543       /* Unset pending notification bits. */
 544       unset_sys_bit(priv(rp)->s_notify_pending, priv(rc)->s_id);
 545
 546       /* Unset pending asynchronous messages */
 547       unset_sys_bit(priv(rp)->s_asyn_pending, priv(rc)->s_id);
 548
 549       /* Check if process depends on given process. */
 550       if (P_BLOCKEDON(rp) == rc->p_endpoint) {
 551           rp->p_reg.retreg = caller_ret;        /* return requested code */
 552           clear_ipc(rp);
 553       }
 554   }
 555 }
 556
 557 /*===========================================================================*
 558  *                              kernel_call_resume                           *
 559  *===========================================================================*/
 560 void kernel_call_resume(struct proc *caller)
 561 {
 562         int result;
 563
 564         assert(!RTS_ISSET(caller, RTS_SLOT_FREE));
 565         assert(!RTS_ISSET(caller, RTS_VMREQUEST));
 566
 567         assert(caller->p_vmrequest.saved.reqmsg.m_source == caller->p_endpoint);
 568
 569         /*
 570         printf("KERNEL_CALL restart from %s / %d rts 0x%08x misc 0x%08x\n",
 571                         caller->p_name, caller->p_endpoint,
 572                         caller->p_rts_flags, caller->p_misc_flags);
 573          */
 574
 575         /* re-execute the kernel call, with MF_KCALL_RESUME still set so
 576          * the call knows this is a retry.
 577          */
 578         result = kernel_call_dispatch(caller, &caller->p_vmrequest.saved.reqmsg);
 579         /*
 580          * we are resuming the kernel call so we have to remove this flag so it
 581          * can be set again
 582          */
 583         caller->p_misc_flags &= ~MF_KCALL_RESUME;
 584         kernel_call_finish(caller, &caller->p_vmrequest.saved.reqmsg, result);
 585 }
 586
 587 /*===========================================================================*
 588  *                               sched_proc                                  *
 589  *===========================================================================*/
 590 int sched_proc(struct proc *p,
 591                         int priority,
 592                         int quantum,
 593                         int cpu)
 594 {
 595         /* Make sure the values given are within the allowed range.*/
 596         if ((priority < TASK_Q && priority != -1) || priority > NR_SCHED_QUEUES)
 597                 return(EINVAL);
 598
 599         if (quantum < 1 && quantum != -1)
 600                 return(EINVAL);
 601
 602 #ifdef CONFIG_SMP
 603         if ((cpu < 0 && cpu != -1) || (cpu > 0 && (unsigned) cpu >= ncpus))
 604                 return(EINVAL);
 605         if (cpu != -1 && !(cpu_is_ready(cpu)))
 606                 return EBADCPU;
 607 #endif
 608
 609         /* In some cases, we might be rescheduling a runnable process. In such
 610          * a case (i.e. if we are updating the priority) we set the NO_QUANTUM
 611          * flag before the generic unset to dequeue/enqueue the process
 612          */
 613
 614         /* FIXME this preempts the process, do we really want to do that ?*/
 615
 616         /* FIXME this is a problem for SMP if the processes currently runs on a
 617          * different CPU */
 618         if (proc_is_runnable(p)) {
 619 #ifdef CONFIG_SMP
 620                 if (p->p_cpu != cpuid && cpu != -1 && cpu != p->p_cpu) {
 621                         smp_schedule_migrate_proc(p, cpu);
 622                 }
 623 #endif
 624
 625                 RTS_SET(p, RTS_NO_QUANTUM);
 626         }
 627
 628         if (proc_is_runnable(p))
 629                 RTS_SET(p, RTS_NO_QUANTUM);
 630
 631         if (priority != -1)
 632                 p->p_priority = priority;
 633         if (quantum != -1) {
 634                 p->p_quantum_size_ms = quantum;
 635                 p->p_cpu_time_left = ms_2_cpu_time(quantum);
 636         }
 637 #ifdef CONFIG_SMP
 638         if (cpu != -1)
 639                 p->p_cpu = cpu;
 640 #endif
 641
 642         /* Clear the scheduling bit and enqueue the process */
 643         RTS_UNSET(p, RTS_NO_QUANTUM);
 644
 645         return OK;
 646 }
 647