vm: include no-caching bits in PTF_ALLFLAGS for flags sanity check.
[minix.git] / kernel / system.c
bloba7f0685048919e247612d91c534b337767034aef
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().
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.
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 * send_sig: send a signal directly to a system process
18 * cause_sig: take action to cause a signal to occur via a signal mgr
19 * sig_delay_done: tell PM that a process is not sending
20 * umap_bios: map virtual address in BIOS_SEG to physical
21 * get_randomness: accumulate randomness in a buffer
22 * clear_endpoint: remove a process' ability to send and receive messages
24 * Changes:
25 * Nov 22, 2009 get_priv supports static priv ids (Cristiano Giuffrida)
26 * Aug 04, 2005 check if system call is allowed (Jorrit N. Herder)
27 * Jul 20, 2005 send signal to services with message (Jorrit N. Herder)
28 * Jan 15, 2005 new, generalized virtual copy function (Jorrit N. Herder)
29 * Oct 10, 2004 dispatch system calls from call vector (Jorrit N. Herder)
30 * Sep 30, 2004 source code documentation updated (Jorrit N. Herder)
33 #include "debug.h"
34 #include "kernel.h"
35 #include "system.h"
36 #include "proc.h"
37 #include "vm.h"
38 #include <stdlib.h>
39 #include <assert.h>
40 #include <signal.h>
41 #include <unistd.h>
42 #include <sys/sigcontext.h>
43 #include <minix/endpoint.h>
44 #include <minix/safecopies.h>
46 /* Declaration of the call vector that defines the mapping of system calls
47 * to handler functions. The vector is initialized in sys_init() with map(),
48 * which makes sure the system call numbers are ok. No space is allocated,
49 * because the dummy is declared extern. If an illegal call is given, the
50 * array size will be negative and this won't compile.
52 PUBLIC int (*call_vec[NR_SYS_CALLS])(struct proc * caller, message *m_ptr);
53 char *callnames[NR_SYS_CALLS];
55 #define map(call_nr, handler) \
56 {extern int dummy[NR_SYS_CALLS>(unsigned)(call_nr-KERNEL_CALL) ? 1:-1];} \
57 callnames[(call_nr-KERNEL_CALL)] = #call_nr; \
58 call_vec[(call_nr-KERNEL_CALL)] = (handler)
60 PRIVATE void kernel_call_finish(struct proc * caller, message *msg, int result)
62 if(result == VMSUSPEND) {
63 /* Special case: message has to be saved for handling
64 * until VM tells us it's allowed. VM has been notified
65 * and we must wait for its reply to restart the call.
67 assert(RTS_ISSET(caller, RTS_VMREQUEST));
68 assert(caller->p_vmrequest.type == VMSTYPE_KERNELCALL);
69 caller->p_vmrequest.saved.reqmsg = *msg;
70 caller->p_misc_flags |= MF_KCALL_RESUME;
71 } else {
73 * call is finished, we could have been suspended because of VM,
74 * remove the request message
76 caller->p_vmrequest.saved.reqmsg.m_source = NONE;
77 if (result != EDONTREPLY) {
78 /* copy the result as a message to the original user buffer */
79 msg->m_source = SYSTEM;
80 msg->m_type = result; /* report status of call */
81 if (copy_msg_to_user(caller, msg,
82 (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);
93 PRIVATE int kernel_call_dispatch(struct proc * caller, message *msg)
95 int result = OK;
96 int call_nr;
98 call_nr = msg->m_type - KERNEL_CALL;
100 /* See if the caller made a valid request and try to handle it. */
101 if (call_nr < 0 || call_nr >= NR_SYS_CALLS) { /* check call number */
102 printf("SYSTEM: illegal request %d from %d.\n",
103 call_nr,msg->m_source);
104 result = EBADREQUEST; /* illegal message type */
106 else if (!GET_BIT(priv(caller)->s_k_call_mask, call_nr)) {
107 printf("SYSTEM: denied request %d from %d.\n",
108 call_nr,msg->m_source);
109 result = ECALLDENIED; /* illegal message type */
110 } else {
111 /* handle the system call */
112 result = (*call_vec[call_nr])(caller, msg);
115 return result;
118 /*===========================================================================*
119 * kernel_call *
120 *===========================================================================*/
122 * this function checks the basic syscall parameters and if accepted it
123 * dispatches its handling to the right handler
125 PUBLIC void kernel_call(message *m_user, struct proc * caller)
127 int result = OK;
128 message msg;
130 caller->p_delivermsg_vir = (vir_bytes) m_user;
132 * the ldt and cr3 of the caller process is loaded because it just've trapped
133 * into the kernel or was already set in schedcheck() before we resume
134 * execution of an interrupted kernel call
136 if (copy_msg_from_user(caller, m_user, &msg) == 0) {
137 msg.m_source = caller->p_endpoint;
138 result = kernel_call_dispatch(caller, &msg);
140 else {
141 printf("WARNING wrong user pointer 0x%08x from process %s / %d\n",
142 m_user, caller->p_name, caller->p_endpoint);
143 result = EBADREQUEST;
146 kernel_call_finish(caller, &msg, result);
149 /*===========================================================================*
150 * initialize *
151 *===========================================================================*/
152 PUBLIC void system_init(void)
154 register struct priv *sp;
155 int i;
157 /* Initialize IRQ handler hooks. Mark all hooks available. */
158 for (i=0; i<NR_IRQ_HOOKS; i++) {
159 irq_hooks[i].proc_nr_e = NONE;
162 /* Initialize all alarm timers for all processes. */
163 for (sp=BEG_PRIV_ADDR; sp < END_PRIV_ADDR; sp++) {
164 tmr_inittimer(&(sp->s_alarm_timer));
167 /* Initialize the call vector to a safe default handler. Some system calls
168 * may be disabled or nonexistant. Then explicitely map known calls to their
169 * handler functions. This is done with a macro that gives a compile error
170 * if an illegal call number is used. The ordering is not important here.
172 for (i=0; i<NR_SYS_CALLS; i++) {
173 call_vec[i] = do_unused;
174 callnames[i] = "unused";
177 /* Process management. */
178 map(SYS_FORK, do_fork); /* a process forked a new process */
179 map(SYS_EXEC, do_exec); /* update process after execute */
180 map(SYS_CLEAR, do_clear); /* clean up after process exit */
181 map(SYS_EXIT, do_exit); /* a system process wants to exit */
182 map(SYS_PRIVCTL, do_privctl); /* system privileges control */
183 map(SYS_TRACE, do_trace); /* request a trace operation */
184 map(SYS_SETGRANT, do_setgrant); /* get/set own parameters */
185 map(SYS_RUNCTL, do_runctl); /* set/clear stop flag of a process */
186 map(SYS_UPDATE, do_update); /* update a process into another */
187 map(SYS_STATECTL, do_statectl); /* let a process control its state */
189 /* Signal handling. */
190 map(SYS_KILL, do_kill); /* cause a process to be signaled */
191 map(SYS_GETKSIG, do_getksig); /* signal manager checks for signals */
192 map(SYS_ENDKSIG, do_endksig); /* signal manager finished signal */
193 map(SYS_SIGSEND, do_sigsend); /* start POSIX-style signal */
194 map(SYS_SIGRETURN, do_sigreturn); /* return from POSIX-style signal */
196 /* Device I/O. */
197 map(SYS_IRQCTL, do_irqctl); /* interrupt control operations */
198 map(SYS_DEVIO, do_devio); /* inb, inw, inl, outb, outw, outl */
199 map(SYS_VDEVIO, do_vdevio); /* vector with devio requests */
201 /* Memory management. */
202 map(SYS_NEWMAP, do_newmap); /* set up a process memory map */
203 map(SYS_SEGCTL, do_segctl); /* add segment and get selector */
204 map(SYS_MEMSET, do_memset); /* write char to memory area */
205 map(SYS_VMCTL, do_vmctl); /* various VM process settings */
207 /* Copying. */
208 map(SYS_UMAP, do_umap); /* map virtual to physical address */
209 map(SYS_VIRCOPY, do_vircopy); /* use pure virtual addressing */
210 map(SYS_PHYSCOPY, do_copy); /* use physical addressing */
211 map(SYS_SAFECOPYFROM, do_safecopy); /* copy with pre-granted permission */
212 map(SYS_SAFECOPYTO, do_safecopy); /* copy with pre-granted permission */
213 map(SYS_VSAFECOPY, do_vsafecopy); /* vectored safecopy */
215 /* Mapping. */
216 map(SYS_SAFEMAP, do_safemap); /* map pages from other process */
217 map(SYS_SAFEREVMAP, do_saferevmap); /* grantor revokes the map grant */
218 map(SYS_SAFEUNMAP, do_safeunmap); /* requestor unmaps the mapped pages */
220 /* Clock functionality. */
221 map(SYS_TIMES, do_times); /* get uptime and process times */
222 map(SYS_SETALARM, do_setalarm); /* schedule a synchronous alarm */
223 map(SYS_STIME, do_stime); /* set the boottime */
224 map(SYS_VTIMER, do_vtimer); /* set or retrieve a virtual timer */
226 /* System control. */
227 map(SYS_ABORT, do_abort); /* abort MINIX */
228 map(SYS_GETINFO, do_getinfo); /* request system information */
229 map(SYS_SYSCTL, do_sysctl); /* misc system manipulation */
231 /* Profiling. */
232 map(SYS_SPROF, do_sprofile); /* start/stop statistical profiling */
233 map(SYS_CPROF, do_cprofile); /* get/reset call profiling data */
234 map(SYS_PROFBUF, do_profbuf); /* announce locations to kernel */
236 /* i386-specific. */
237 #if _MINIX_CHIP == _CHIP_INTEL
238 map(SYS_INT86, do_int86); /* real-mode BIOS calls */
239 map(SYS_READBIOS, do_readbios); /* read from BIOS locations */
240 map(SYS_IOPENABLE, do_iopenable); /* Enable I/O */
241 map(SYS_SDEVIO, do_sdevio); /* phys_insb, _insw, _outsb, _outsw */
243 /* Machine state switching. */
244 map(SYS_SETMCONTEXT, do_setmcontext); /* set machine context */
245 map(SYS_GETMCONTEXT, do_getmcontext); /* get machine context */
246 #endif
248 /* Scheduling */
249 map(SYS_SCHEDULE, do_schedule); /* reschedule a process */
250 map(SYS_SCHEDCTL, do_schedctl); /* change process scheduler */
253 /*===========================================================================*
254 * get_priv *
255 *===========================================================================*/
256 PUBLIC int get_priv(rc, priv_id)
257 register struct proc *rc; /* new (child) process pointer */
258 int priv_id; /* privilege id */
260 /* Allocate a new privilege structure for a system process. Privilege ids
261 * can be assigned either statically or dynamically.
263 register struct priv *sp; /* privilege structure */
265 if(priv_id == NULL_PRIV_ID) { /* allocate slot dynamically */
266 for (sp = BEG_DYN_PRIV_ADDR; sp < END_DYN_PRIV_ADDR; ++sp)
267 if (sp->s_proc_nr == NONE) break;
268 if (sp >= END_DYN_PRIV_ADDR) return(ENOSPC);
270 else { /* allocate slot from id */
271 if(!is_static_priv_id(priv_id)) {
272 return EINVAL; /* invalid static priv id */
274 if(priv[priv_id].s_proc_nr != NONE) {
275 return EBUSY; /* slot already in use */
277 sp = &priv[priv_id];
279 rc->p_priv = sp; /* assign new slot */
280 rc->p_priv->s_proc_nr = proc_nr(rc); /* set association */
282 return(OK);
285 /*===========================================================================*
286 * set_sendto_bit *
287 *===========================================================================*/
288 PUBLIC void set_sendto_bit(const struct proc *rp, int id)
290 /* Allow a process to send messages to the process(es) associated with the
291 * system privilege structure with the given ID.
294 /* Disallow the process from sending to a process privilege structure with no
295 * associated process, and disallow the process from sending to itself.
297 if (id_to_nr(id) == NONE || priv_id(rp) == id) {
298 unset_sys_bit(priv(rp)->s_ipc_to, id);
299 return;
302 set_sys_bit(priv(rp)->s_ipc_to, id);
304 /* The process that this process can now send to, must be able to reply (or
305 * vice versa). Therefore, its send mask should be updated as well. Ignore
306 * receivers that don't support traps other than RECEIVE, they can't reply
307 * or send messages anyway.
309 if (priv_addr(id)->s_trap_mask & ~((1 << RECEIVE)))
310 set_sys_bit(priv_addr(id)->s_ipc_to, priv_id(rp));
313 /*===========================================================================*
314 * unset_sendto_bit *
315 *===========================================================================*/
316 PUBLIC void unset_sendto_bit(const struct proc *rp, int id)
318 /* Prevent a process from sending to another process. Retain the send mask
319 * symmetry by also unsetting the bit for the other direction.
322 unset_sys_bit(priv(rp)->s_ipc_to, id);
324 unset_sys_bit(priv_addr(id)->s_ipc_to, priv_id(rp));
327 /*===========================================================================*
328 * send_sig *
329 *===========================================================================*/
330 PUBLIC void send_sig(endpoint_t proc_nr, int sig_nr)
332 /* Notify a system process about a signal. This is straightforward. Simply
333 * set the signal that is to be delivered in the pending signals map and
334 * send a notification with source SYSTEM.
336 register struct proc *rp;
338 if(!isokprocn(proc_nr) || isemptyn(proc_nr))
339 panic("send_sig to empty process: %d", proc_nr);
341 rp = proc_addr(proc_nr);
342 sigaddset(&priv(rp)->s_sig_pending, sig_nr);
343 mini_notify(proc_addr(SYSTEM), rp->p_endpoint);
346 /*===========================================================================*
347 * cause_sig *
348 *===========================================================================*/
349 PUBLIC void cause_sig(proc_nr, sig_nr)
350 proc_nr_t proc_nr; /* process to be signalled */
351 int sig_nr; /* signal to be sent */
353 /* A system process wants to send a signal to a process. Examples are:
354 * - HARDWARE wanting to cause a SIGSEGV after a CPU exception
355 * - TTY wanting to cause SIGINT upon getting a DEL
356 * - FS wanting to cause SIGPIPE for a broken pipe
357 * Signals are handled by sending a message to the signal manager assigned to
358 * the process. This function handles the signals and makes sure the signal
359 * manager gets them by sending a notification. The process being signaled
360 * is blocked while the signal manager has not finished all signals for it.
361 * Race conditions between calls to this function and the system calls that
362 * process pending kernel signals cannot exist. Signal related functions are
363 * only called when a user process causes a CPU exception and from the kernel
364 * process level, which runs to completion.
366 register struct proc *rp;
367 endpoint_t sig_mgr;
369 /* Lookup signal manager. */
370 rp = proc_addr(proc_nr);
371 sig_mgr = priv(rp)->s_sig_mgr;
373 /* If the target is the signal manager of itself, send the signal directly. */
374 if(rp->p_endpoint == sig_mgr) {
375 if(SIGS_IS_LETHAL(sig_nr)) {
376 panic("cause_sig: signal manager gets lethal signal for itself");
378 sigaddset(&priv(rp)->s_sig_pending, sig_nr);
379 send_sig(rp->p_endpoint, SIGKSIGSM);
380 return;
383 /* Check if the signal is already pending. Process it otherwise. */
384 if (! sigismember(&rp->p_pending, sig_nr)) {
385 sigaddset(&rp->p_pending, sig_nr);
386 if (! (RTS_ISSET(rp, RTS_SIGNALED))) { /* other pending */
387 RTS_SET(rp, RTS_SIGNALED | RTS_SIG_PENDING);
388 send_sig(sig_mgr, SIGKSIG);
393 /*===========================================================================*
394 * sig_delay_done *
395 *===========================================================================*/
396 PUBLIC void sig_delay_done(struct proc *rp)
398 /* A process is now known not to send any direct messages.
399 * Tell PM that the stop delay has ended, by sending a signal to the process.
400 * Used for actual signal delivery.
403 rp->p_misc_flags &= ~MF_SIG_DELAY;
405 cause_sig(proc_nr(rp), SIGSNDELAY);
408 #if _MINIX_CHIP == _CHIP_INTEL
410 /*===========================================================================*
411 * umap_bios *
412 *===========================================================================*/
413 PUBLIC phys_bytes umap_bios(vir_addr, bytes)
414 vir_bytes vir_addr; /* virtual address in BIOS segment */
415 vir_bytes bytes; /* # of bytes to be copied */
417 /* Calculate the physical memory address at the BIOS. Note: currently, BIOS
418 * address zero (the first BIOS interrupt vector) is not considered as an
419 * error here, but since the physical address will be zero as well, the
420 * calling function will think an error occurred. This is not a problem,
421 * since no one uses the first BIOS interrupt vector.
424 /* Check all acceptable ranges. */
425 if (vir_addr >= BIOS_MEM_BEGIN && vir_addr + bytes <= BIOS_MEM_END)
426 return (phys_bytes) vir_addr;
427 else if (vir_addr >= BASE_MEM_TOP && vir_addr + bytes <= UPPER_MEM_END)
428 return (phys_bytes) vir_addr;
430 printf("Warning, error in umap_bios, virtual address 0x%x\n", vir_addr);
431 return 0;
433 #endif
435 /*===========================================================================*
436 * umap_grant *
437 *===========================================================================*/
438 PUBLIC phys_bytes umap_grant(rp, grant, bytes)
439 struct proc *rp; /* pointer to proc table entry for process */
440 cp_grant_id_t grant; /* grant no. */
441 vir_bytes bytes; /* size */
443 int proc_nr;
444 vir_bytes offset, ret;
445 endpoint_t granter;
447 /* See if the grant in that process is sensible, and
448 * find out the virtual address and (optionally) new
449 * process for that address.
451 * Then convert that process to a slot number.
453 if(verify_grant(rp->p_endpoint, ANY, grant, bytes, 0, 0,
454 &offset, &granter) != OK) {
455 printf("SYSTEM: umap_grant: verify_grant failed\n");
456 return 0;
459 if(!isokendpt(granter, &proc_nr)) {
460 printf("SYSTEM: umap_grant: isokendpt failed\n");
461 return 0;
464 /* Do the mapping from virtual to physical. */
465 ret = umap_virtual(proc_addr(proc_nr), D, offset, bytes);
466 if(!ret) {
467 printf("SYSTEM:umap_grant:umap_virtual failed; grant %s:%d -> %s: vir 0x%lx\n",
468 rp->p_name, grant,
469 proc_addr(proc_nr)->p_name, offset);
471 return ret;
474 /*===========================================================================*
475 * clear_endpoint *
476 *===========================================================================*/
477 PUBLIC void clear_endpoint(rc)
478 register struct proc *rc; /* slot of process to clean up */
480 if(isemptyp(rc)) panic("clear_proc: empty process: %d", rc->p_endpoint);
482 /* Make sure that the exiting process is no longer scheduled. */
483 RTS_SET(rc, RTS_NO_ENDPOINT);
484 if (priv(rc)->s_flags & SYS_PROC)
486 priv(rc)->s_asynsize= 0;
489 /* If the process happens to be queued trying to send a
490 * message, then it must be removed from the message queues.
492 clear_ipc(rc);
494 /* Likewise, if another process was sending or receive a message to or from
495 * the exiting process, it must be alerted that process no longer is alive.
496 * Check all processes.
498 clear_ipc_refs(rc, EDEADSRCDST);
502 /*===========================================================================*
503 * clear_ipc *
504 *===========================================================================*/
505 PUBLIC void clear_ipc(rc)
506 register struct proc *rc; /* slot of process to clean up */
508 /* Clear IPC data for a given process slot. */
509 struct proc **xpp; /* iterate over caller queue */
511 if (RTS_ISSET(rc, RTS_SENDING)) {
512 int target_proc;
514 okendpt(rc->p_sendto_e, &target_proc);
515 xpp = &proc_addr(target_proc)->p_caller_q; /* destination's queue */
516 while (*xpp) { /* check entire queue */
517 if (*xpp == rc) { /* process is on the queue */
518 *xpp = (*xpp)->p_q_link; /* replace by next process */
519 #if DEBUG_ENABLE_IPC_WARNINGS
520 printf("endpoint %d / %s removed from queue at %d\n",
521 rc->p_endpoint, rc->p_name, rc->p_sendto_e);
522 #endif
523 break; /* can only be queued once */
525 xpp = &(*xpp)->p_q_link; /* proceed to next queued */
527 rc->p_rts_flags &= ~RTS_SENDING;
529 rc->p_rts_flags &= ~RTS_RECEIVING;
532 /*===========================================================================*
533 * clear_ipc_refs *
534 *===========================================================================*/
535 PUBLIC void clear_ipc_refs(rc, caller_ret)
536 register struct proc *rc; /* slot of process to clean up */
537 int caller_ret; /* code to return on callers */
539 /* Clear IPC references for a given process slot. */
540 struct proc *rp; /* iterate over process table */
542 for (rp = BEG_PROC_ADDR; rp < END_PROC_ADDR; rp++) {
543 if(isemptyp(rp))
544 continue;
546 /* Unset pending notification bits. */
547 unset_sys_bit(priv(rp)->s_notify_pending, priv(rc)->s_id);
549 /* XXX FIXME: Cleanup should be done for senda() as well. For this to be
550 * done in a realistic way, we need a better implementation of senda
551 * with a bitmap similar to s_notify_pending for notify() rather than
552 * a single global MF_ASYNMSG flag. The current arrangement exposes
553 * several performance issues.
556 /* Check if process depends on given process. */
557 if (P_BLOCKEDON(rp) == rc->p_endpoint) {
558 rp->p_reg.retreg = caller_ret; /* return requested code */
559 RTS_UNSET(rp, (RTS_RECEIVING|RTS_SENDING)); /* no longer blocking */
564 /*===========================================================================*
565 * kernel_call_resume *
566 *===========================================================================*/
567 PUBLIC void kernel_call_resume(struct proc *caller)
569 int result;
571 assert(!RTS_ISSET(caller, RTS_SLOT_FREE));
572 assert(!RTS_ISSET(caller, RTS_VMREQUEST));
574 assert(caller->p_vmrequest.saved.reqmsg.m_source == caller->p_endpoint);
577 printf("KERNEL_CALL restart from %s / %d rts 0x%08x misc 0x%08x\n",
578 caller->p_name, caller->p_endpoint,
579 caller->p_rts_flags, caller->p_misc_flags);
583 * we are resuming the kernel call so we have to remove this flag so it
584 * can be set again
586 caller->p_misc_flags &= ~MF_KCALL_RESUME;
587 result = kernel_call_dispatch(caller, &caller->p_vmrequest.saved.reqmsg);
588 kernel_call_finish(caller, &caller->p_vmrequest.saved.reqmsg, result);