xref: /netbsd-src/sys/compat/linux/common/linux_sched.c (revision bfb6cb13d599546df69c7e4d20d70e22e15a549d)
1 /*	$NetBSD: linux_sched.c,v 1.42 2007/02/15 20:32:48 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center; by Matthias Scheler.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Linux compatibility module. Try to deal with scheduler related syscalls.
42  */
43 
44 #include <sys/cdefs.h>
45 __KERNEL_RCSID(0, "$NetBSD: linux_sched.c,v 1.42 2007/02/15 20:32:48 ad Exp $");
46 
47 #include <sys/param.h>
48 #include <sys/mount.h>
49 #include <sys/proc.h>
50 #include <sys/systm.h>
51 #include <sys/sysctl.h>
52 #include <sys/malloc.h>
53 #include <sys/syscallargs.h>
54 #include <sys/wait.h>
55 #include <sys/kauth.h>
56 #include <sys/ptrace.h>
57 
58 #include <machine/cpu.h>
59 
60 #include <compat/linux/common/linux_types.h>
61 #include <compat/linux/common/linux_signal.h>
62 #include <compat/linux/common/linux_machdep.h> /* For LINUX_NPTL */
63 #include <compat/linux/common/linux_emuldata.h>
64 
65 #include <compat/linux/linux_syscallargs.h>
66 
67 #include <compat/linux/common/linux_sched.h>
68 
69 int
70 linux_sys_clone(l, v, retval)
71 	struct lwp *l;
72 	void *v;
73 	register_t *retval;
74 {
75 	struct linux_sys_clone_args /* {
76 		syscallarg(int) flags;
77 		syscallarg(void *) stack;
78 #ifdef LINUX_NPTL
79 		syscallarg(void *) parent_tidptr;
80 		syscallarg(void *) child_tidptr;
81 #endif
82 	} */ *uap = v;
83 	int flags, sig;
84 	int error;
85 #ifdef LINUX_NPTL
86 	struct linux_emuldata *led;
87 #endif
88 
89 	/*
90 	 * We don't support the Linux CLONE_PID or CLONE_PTRACE flags.
91 	 */
92 	if (SCARG(uap, flags) & (LINUX_CLONE_PID|LINUX_CLONE_PTRACE))
93 		return (EINVAL);
94 
95 	/*
96 	 * Thread group implies shared signals. Shared signals
97 	 * imply shared VM. This matches what Linux kernel does.
98 	 */
99 	if (SCARG(uap, flags) & LINUX_CLONE_THREAD
100 	    && (SCARG(uap, flags) & LINUX_CLONE_SIGHAND) == 0)
101 		return (EINVAL);
102 	if (SCARG(uap, flags) & LINUX_CLONE_SIGHAND
103 	    && (SCARG(uap, flags) & LINUX_CLONE_VM) == 0)
104 		return (EINVAL);
105 
106 	flags = 0;
107 
108 	if (SCARG(uap, flags) & LINUX_CLONE_VM)
109 		flags |= FORK_SHAREVM;
110 	if (SCARG(uap, flags) & LINUX_CLONE_FS)
111 		flags |= FORK_SHARECWD;
112 	if (SCARG(uap, flags) & LINUX_CLONE_FILES)
113 		flags |= FORK_SHAREFILES;
114 	if (SCARG(uap, flags) & LINUX_CLONE_SIGHAND)
115 		flags |= FORK_SHARESIGS;
116 	if (SCARG(uap, flags) & LINUX_CLONE_VFORK)
117 		flags |= FORK_PPWAIT;
118 
119 	sig = SCARG(uap, flags) & LINUX_CLONE_CSIGNAL;
120 	if (sig < 0 || sig >= LINUX__NSIG)
121 		return (EINVAL);
122 	sig = linux_to_native_signo[sig];
123 
124 #ifdef LINUX_NPTL
125 	led = (struct linux_emuldata *)l->l_proc->p_emuldata;
126 
127 	led->parent_tidptr = SCARG(uap, parent_tidptr);
128 	led->child_tidptr = SCARG(uap, child_tidptr);
129 	led->clone_flags = SCARG(uap, flags);
130 #endif /* LINUX_NPTL */
131 
132 	/*
133 	 * Note that Linux does not provide a portable way of specifying
134 	 * the stack area; the caller must know if the stack grows up
135 	 * or down.  So, we pass a stack size of 0, so that the code
136 	 * that makes this adjustment is a noop.
137 	 */
138 	if ((error = fork1(l, flags, sig, SCARG(uap, stack), 0,
139 	    NULL, NULL, retval, NULL)) != 0)
140 		return error;
141 
142 	return 0;
143 }
144 
145 int
146 linux_sys_sched_setparam(struct lwp *cl, void *v, register_t *retval)
147 {
148 	struct linux_sys_sched_setparam_args /* {
149 		syscallarg(linux_pid_t) pid;
150 		syscallarg(const struct linux_sched_param *) sp;
151 	} */ *uap = v;
152 	int error;
153 	struct linux_sched_param lp;
154 	struct proc *p;
155 
156 /*
157  * We only check for valid parameters and return afterwards.
158  */
159 
160 	if (SCARG(uap, pid) < 0 || SCARG(uap, sp) == NULL)
161 		return EINVAL;
162 
163 	error = copyin(SCARG(uap, sp), &lp, sizeof(lp));
164 	if (error)
165 		return error;
166 
167 	if (SCARG(uap, pid) != 0) {
168 		kauth_cred_t pc = cl->l_cred;
169 
170 		if ((p = pfind(SCARG(uap, pid))) == NULL)
171 			return ESRCH;
172 		if (!(cl->l_proc == p ||
173 		      kauth_authorize_generic(pc, KAUTH_GENERIC_ISSUSER, NULL) == 0 ||
174 		      kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
175 		      kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
176 		      kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
177 		      kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
178 			return EPERM;
179 	}
180 
181 	return 0;
182 }
183 
184 int
185 linux_sys_sched_getparam(struct lwp *cl, void *v, register_t *retval)
186 {
187 	struct linux_sys_sched_getparam_args /* {
188 		syscallarg(linux_pid_t) pid;
189 		syscallarg(struct linux_sched_param *) sp;
190 	} */ *uap = v;
191 	struct proc *p;
192 	struct linux_sched_param lp;
193 
194 /*
195  * We only check for valid parameters and return a dummy priority afterwards.
196  */
197 	if (SCARG(uap, pid) < 0 || SCARG(uap, sp) == NULL)
198 		return EINVAL;
199 
200 	if (SCARG(uap, pid) != 0) {
201 		kauth_cred_t pc = cl->l_cred;
202 
203 		if ((p = pfind(SCARG(uap, pid))) == NULL)
204 			return ESRCH;
205 		if (!(cl->l_proc == p ||
206 		      kauth_authorize_generic(pc, KAUTH_GENERIC_ISSUSER, NULL) == 0 ||
207 		      kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
208 		      kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
209 		      kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
210 		      kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
211 			return EPERM;
212 	}
213 
214 	lp.sched_priority = 0;
215 	return copyout(&lp, SCARG(uap, sp), sizeof(lp));
216 }
217 
218 int
219 linux_sys_sched_setscheduler(struct lwp *cl, void *v,
220     register_t *retval)
221 {
222 	struct linux_sys_sched_setscheduler_args /* {
223 		syscallarg(linux_pid_t) pid;
224 		syscallarg(int) policy;
225 		syscallarg(cont struct linux_sched_scheduler *) sp;
226 	} */ *uap = v;
227 	int error;
228 	struct linux_sched_param lp;
229 	struct proc *p;
230 
231 /*
232  * We only check for valid parameters and return afterwards.
233  */
234 
235 	if (SCARG(uap, pid) < 0 || SCARG(uap, sp) == NULL)
236 		return EINVAL;
237 
238 	error = copyin(SCARG(uap, sp), &lp, sizeof(lp));
239 	if (error)
240 		return error;
241 
242 	if (SCARG(uap, pid) != 0) {
243 		kauth_cred_t pc = cl->l_cred;
244 
245 		if ((p = pfind(SCARG(uap, pid))) == NULL)
246 			return ESRCH;
247 		if (!(cl->l_proc == p ||
248 		      kauth_authorize_generic(pc, KAUTH_GENERIC_ISSUSER, NULL) == 0 ||
249 		      kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
250 		      kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
251 		      kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
252 		      kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
253 			return EPERM;
254 	}
255 
256 	return 0;
257 /*
258  * We can't emulate anything put the default scheduling policy.
259  */
260 	if (SCARG(uap, policy) != LINUX_SCHED_OTHER || lp.sched_priority != 0)
261 		return EINVAL;
262 
263 	return 0;
264 }
265 
266 int
267 linux_sys_sched_getscheduler(cl, v, retval)
268 	struct lwp *cl;
269 	void *v;
270 	register_t *retval;
271 {
272 	struct linux_sys_sched_getscheduler_args /* {
273 		syscallarg(linux_pid_t) pid;
274 	} */ *uap = v;
275 	struct proc *p;
276 
277 	*retval = -1;
278 /*
279  * We only check for valid parameters and return afterwards.
280  */
281 
282 	if (SCARG(uap, pid) != 0) {
283 		kauth_cred_t pc = cl->l_cred;
284 
285 		if ((p = pfind(SCARG(uap, pid))) == NULL)
286 			return ESRCH;
287 		if (!(cl->l_proc == p ||
288 		      kauth_authorize_generic(pc, KAUTH_GENERIC_ISSUSER, NULL) == 0 ||
289 		      kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
290 		      kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
291 		      kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
292 		      kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
293 			return EPERM;
294 	}
295 
296 /*
297  * We can't emulate anything put the default scheduling policy.
298  */
299 	*retval = LINUX_SCHED_OTHER;
300 	return 0;
301 }
302 
303 int
304 linux_sys_sched_yield(struct lwp *cl, void *v,
305     register_t *retval)
306 {
307 
308 	yield();
309 	return 0;
310 }
311 
312 int
313 linux_sys_sched_get_priority_max(struct lwp *cl, void *v,
314     register_t *retval)
315 {
316 	struct linux_sys_sched_get_priority_max_args /* {
317 		syscallarg(int) policy;
318 	} */ *uap = v;
319 
320 /*
321  * We can't emulate anything put the default scheduling policy.
322  */
323 	if (SCARG(uap, policy) != LINUX_SCHED_OTHER) {
324 		*retval = -1;
325 		return EINVAL;
326 	}
327 
328 	*retval = 0;
329 	return 0;
330 }
331 
332 int
333 linux_sys_sched_get_priority_min(struct lwp *cl, void *v,
334     register_t *retval)
335 {
336 	struct linux_sys_sched_get_priority_min_args /* {
337 		syscallarg(int) policy;
338 	} */ *uap = v;
339 
340 /*
341  * We can't emulate anything put the default scheduling policy.
342  */
343 	if (SCARG(uap, policy) != LINUX_SCHED_OTHER) {
344 		*retval = -1;
345 		return EINVAL;
346 	}
347 
348 	*retval = 0;
349 	return 0;
350 }
351 
352 #ifndef __m68k__
353 /* Present on everything but m68k */
354 int
355 linux_sys_exit_group(l, v, retval)
356 	struct lwp *l;
357 	void *v;
358 	register_t *retval;
359 {
360 #ifdef LINUX_NPTL
361 	struct linux_sys_exit_group_args /* {
362 		syscallarg(int) error_code;
363 	} */ *uap = v;
364 	struct proc *p = l->l_proc;
365 	struct linux_emuldata *led = p->p_emuldata;
366 	struct linux_emuldata *e;
367 
368 	if (led->s->flags & LINUX_LES_USE_NPTL) {
369 
370 #ifdef DEBUG_LINUX
371 		printf("%s:%d, led->s->refs = %d\n", __func__, __LINE__,
372 		    led->s->refs);
373 #endif
374 
375 		/*
376 		 * The calling thread is supposed to kill all threads
377 		 * in the same thread group (i.e. all threads created
378 		 * via clone(2) with CLONE_THREAD flag set).
379 		 *
380 		 * If there is only one thread, things are quite simple
381 		 */
382 		if (led->s->refs == 1)
383 			return sys_exit(l, v, retval);
384 
385 #ifdef DEBUG_LINUX
386 		printf("%s:%d\n", __func__, __LINE__);
387 #endif
388 
389 		led->s->flags |= LINUX_LES_INEXITGROUP;
390 		led->s->xstat = W_EXITCODE(SCARG(uap, error_code), 0);
391 
392 		/*
393 		 * Kill all threads in the group. The emulation exit hook takes
394 		 * care of hiding the zombies and reporting the exit code
395 		 * properly.
396 		 */
397 		mutex_enter(&proclist_mutex);
398       		LIST_FOREACH(e, &led->s->threads, threads) {
399 			if (e->proc == p)
400 				continue;
401 
402 #ifdef DEBUG_LINUX
403 			printf("%s: kill PID %d\n", __func__, e->proc->p_pid);
404 #endif
405 			psignal(e->proc, SIGKILL);
406 		}
407 
408 		/* Now, kill ourselves */
409 		psignal(p, SIGKILL);
410 		mutex_exit(&proclist_mutex);
411 
412 		return 0;
413 
414 	}
415 #endif /* LINUX_NPTL */
416 
417 	return sys_exit(l, v, retval);
418 }
419 #endif /* !__m68k__ */
420 
421 #ifdef LINUX_NPTL
422 int
423 linux_sys_set_tid_address(l, v, retval)
424 	struct lwp *l;
425 	void *v;
426 	register_t *retval;
427 {
428 	struct linux_sys_set_tid_address_args /* {
429 		syscallarg(int *) tidptr;
430 	} */ *uap = v;
431 	struct linux_emuldata *led;
432 
433 	led = (struct linux_emuldata *)l->l_proc->p_emuldata;
434 	led->clear_tid = SCARG(uap, tid);
435 
436 	led->s->flags |= LINUX_LES_USE_NPTL;
437 
438 	*retval = l->l_proc->p_pid;
439 
440 	return 0;
441 }
442 
443 /* ARGUSED1 */
444 int
445 linux_sys_gettid(l, v, retval)
446 	struct lwp *l;
447 	void *v;
448 	register_t *retval;
449 {
450 	/* The Linux kernel does it exactly that way */
451 	*retval = l->l_proc->p_pid;
452 	return 0;
453 }
454 
455 #ifdef LINUX_NPTL
456 /* ARGUSED1 */
457 int
458 linux_sys_getpid(l, v, retval)
459 	struct lwp *l;
460 	void *v;
461 	register_t *retval;
462 {
463 	struct linux_emuldata *led = l->l_proc->p_emuldata;
464 
465 	if (led->s->flags & LINUX_LES_USE_NPTL) {
466 		/* The Linux kernel does it exactly that way */
467 		*retval = led->s->group_pid;
468 	} else {
469 		*retval = l->l_proc->p_pid;
470 	}
471 
472 	return 0;
473 }
474 
475 /* ARGUSED1 */
476 int
477 linux_sys_getppid(l, v, retval)
478 	struct lwp *l;
479 	void *v;
480 	register_t *retval;
481 {
482 	struct proc *p = l->l_proc;
483 	struct linux_emuldata *led = p->p_emuldata;
484 	struct proc *glp;
485 	struct proc *pp;
486 
487 	if (led->s->flags & LINUX_LES_USE_NPTL) {
488 
489 		/* Find the thread group leader's parent */
490 		if ((glp = pfind(led->s->group_pid)) == NULL) {
491 			/* Maybe panic... */
492 			printf("linux_sys_getppid: missing group leader PID"
493 			    " %d\n", led->s->group_pid);
494 			return -1;
495 		}
496 		pp = glp->p_pptr;
497 
498 		/* If this is a Linux process too, return thread group PID */
499 		if (pp->p_emul == p->p_emul) {
500 			struct linux_emuldata *pled;
501 
502 			pled = pp->p_emuldata;
503 			*retval = pled->s->group_pid;
504 		} else {
505 			*retval = pp->p_pid;
506 		}
507 
508 	} else {
509 		*retval = p->p_pptr->p_pid;
510 	}
511 
512 	return 0;
513 }
514 #endif /* LINUX_NPTL */
515 
516 int
517 linux_sys_sched_getaffinity(l, v, retval)
518 	struct lwp *l;
519 	void *v;
520 	register_t *retval;
521 {
522 	struct linux_sys_sched_getaffinity_args /* {
523 		syscallarg(pid_t) pid;
524 		syscallarg(unsigned int) len;
525 		syscallarg(unsigned long *) mask;
526 	} */ *uap = v;
527 	int error;
528 	int ret;
529 	char *data;
530 	int *retp;
531 
532 	if (SCARG(uap, mask) == NULL)
533 		return EINVAL;
534 
535 	if (SCARG(uap, len) < sizeof(int))
536 		return EINVAL;
537 
538 	if (pfind(SCARG(uap, pid)) == NULL)
539 		return ESRCH;
540 
541 	/*
542 	 * return the actual number of CPU, tag all of them as available
543 	 * The result is a mask, the first CPU being in the least significant
544 	 * bit.
545 	 */
546 	ret = (1 << ncpu) - 1;
547 	data = malloc(SCARG(uap, len), M_TEMP, M_WAITOK|M_ZERO);
548 	retp = (int *)&data[SCARG(uap, len) - sizeof(ret)];
549 	*retp = ret;
550 
551 	if ((error = copyout(data, SCARG(uap, mask), SCARG(uap, len))) != 0)
552 		return error;
553 
554 	free(data, M_TEMP);
555 
556 	return 0;
557 
558 }
559 
560 int
561 linux_sys_sched_setaffinity(l, v, retval)
562 	struct lwp *l;
563 	void *v;
564 	register_t *retval;
565 {
566 	struct linux_sys_sched_setaffinity_args /* {
567 		syscallarg(pid_t) pid;
568 		syscallarg(unsigned int) len;
569 		syscallarg(unsigned long *) mask;
570 	} */ *uap = v;
571 
572 	if (pfind(SCARG(uap, pid)) == NULL)
573 		return ESRCH;
574 
575 	/* Let's ignore it */
576 #ifdef DEBUG_LINUX
577 	printf("linux_sys_sched_setaffinity\n");
578 #endif
579 	return 0;
580 };
581 #endif /* LINUX_NPTL */
582