1 /* $NetBSD: linux_sched.c,v 1.37.2.2 2009/06/21 11:22:24 bouyer 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 1.37.2.2 2009/06/21 11:22:24 bouyer 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/sa.h>
54 #include <sys/syscallargs.h>
55 #include <sys/wait.h>
56 #include <sys/kauth.h>
57 #include <sys/ptrace.h>
58
59 #include <machine/cpu.h>
60
61 #include <compat/linux/common/linux_types.h>
62 #include <compat/linux/common/linux_signal.h>
63 #include <compat/linux/common/linux_machdep.h> /* For LINUX_NPTL */
64 #include <compat/linux/common/linux_emuldata.h>
65
66 #include <compat/linux/linux_syscallargs.h>
67
68 #include <compat/linux/common/linux_sched.h>
69
70 int
71 linux_sys_clone(l, v, retval)
72 struct lwp *l;
73 void *v;
74 register_t *retval;
75 {
76 struct linux_sys_clone_args /* {
77 syscallarg(int) flags;
78 syscallarg(void *) stack;
79 #ifdef LINUX_NPTL
80 syscallarg(void *) parent_tidptr;
81 syscallarg(void *) child_tidptr;
82 #endif
83 } */ *uap = v;
84 int flags, sig;
85 int error;
86 #ifdef LINUX_NPTL
87 struct linux_emuldata *led;
88 #endif
89
90 /*
91 * We don't support the Linux CLONE_PID or CLONE_PTRACE flags.
92 */
93 if (SCARG(uap, flags) & (LINUX_CLONE_PID|LINUX_CLONE_PTRACE))
94 return (EINVAL);
95
96 /*
97 * Thread group implies shared signals. Shared signals
98 * imply shared VM. This matches what Linux kernel does.
99 */
100 if (SCARG(uap, flags) & LINUX_CLONE_THREAD
101 && (SCARG(uap, flags) & LINUX_CLONE_SIGHAND) == 0)
102 return (EINVAL);
103 if (SCARG(uap, flags) & LINUX_CLONE_SIGHAND
104 && (SCARG(uap, flags) & LINUX_CLONE_VM) == 0)
105 return (EINVAL);
106
107 flags = 0;
108
109 if (SCARG(uap, flags) & LINUX_CLONE_VM)
110 flags |= FORK_SHAREVM;
111 if (SCARG(uap, flags) & LINUX_CLONE_FS)
112 flags |= FORK_SHARECWD;
113 if (SCARG(uap, flags) & LINUX_CLONE_FILES)
114 flags |= FORK_SHAREFILES;
115 if (SCARG(uap, flags) & LINUX_CLONE_SIGHAND)
116 flags |= FORK_SHARESIGS;
117 if (SCARG(uap, flags) & LINUX_CLONE_VFORK)
118 flags |= FORK_PPWAIT;
119
120 sig = SCARG(uap, flags) & LINUX_CLONE_CSIGNAL;
121 if (sig < 0 || sig >= LINUX__NSIG)
122 return (EINVAL);
123 sig = linux_to_native_signo[sig];
124
125 #ifdef LINUX_NPTL
126 led = (struct linux_emuldata *)l->l_proc->p_emuldata;
127
128 led->parent_tidptr = SCARG(uap, parent_tidptr);
129 led->child_tidptr = SCARG(uap, child_tidptr);
130 led->clone_flags = SCARG(uap, flags);
131 #endif /* LINUX_NPTL */
132
133 /*
134 * Note that Linux does not provide a portable way of specifying
135 * the stack area; the caller must know if the stack grows up
136 * or down. So, we pass a stack size of 0, so that the code
137 * that makes this adjustment is a noop.
138 */
139 if ((error = fork1(l, flags, sig, SCARG(uap, stack), 0,
140 NULL, NULL, retval, NULL)) != 0)
141 return error;
142
143 return 0;
144 }
145
146 int
147 linux_sys_sched_setparam(struct lwp *cl, void *v, register_t *retval)
148 {
149 struct linux_sys_sched_setparam_args /* {
150 syscallarg(linux_pid_t) pid;
151 syscallarg(const struct linux_sched_param *) sp;
152 } */ *uap = v;
153 int error;
154 struct linux_sched_param lp;
155 struct proc *p;
156
157 /*
158 * We only check for valid parameters and return afterwards.
159 */
160
161 if (SCARG(uap, pid) < 0 || SCARG(uap, sp) == NULL)
162 return EINVAL;
163
164 error = copyin(SCARG(uap, sp), &lp, sizeof(lp));
165 if (error)
166 return error;
167
168 if (SCARG(uap, pid) != 0) {
169 kauth_cred_t pc = cl->l_cred;
170
171 if ((p = pfind(SCARG(uap, pid))) == NULL)
172 return ESRCH;
173 if (!(cl->l_proc == p ||
174 kauth_cred_geteuid(pc) == 0 ||
175 kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
176 kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
177 kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
178 kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
179 return EPERM;
180 }
181
182 return 0;
183 }
184
185 int
186 linux_sys_sched_getparam(struct lwp *cl, void *v, register_t *retval)
187 {
188 struct linux_sys_sched_getparam_args /* {
189 syscallarg(linux_pid_t) pid;
190 syscallarg(struct linux_sched_param *) sp;
191 } */ *uap = v;
192 struct proc *p;
193 struct linux_sched_param lp;
194
195 /*
196 * We only check for valid parameters and return a dummy priority afterwards.
197 */
198 if (SCARG(uap, pid) < 0 || SCARG(uap, sp) == NULL)
199 return EINVAL;
200
201 if (SCARG(uap, pid) != 0) {
202 kauth_cred_t pc = cl->l_cred;
203
204 if ((p = pfind(SCARG(uap, pid))) == NULL)
205 return ESRCH;
206 if (!(cl->l_proc == p ||
207 kauth_cred_geteuid(pc) == 0 ||
208 kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
209 kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
210 kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
211 kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
212 return EPERM;
213 }
214
215 lp.sched_priority = 0;
216 return copyout(&lp, SCARG(uap, sp), sizeof(lp));
217 }
218
219 int
220 linux_sys_sched_setscheduler(struct lwp *cl, void *v,
221 register_t *retval)
222 {
223 struct linux_sys_sched_setscheduler_args /* {
224 syscallarg(linux_pid_t) pid;
225 syscallarg(int) policy;
226 syscallarg(cont struct linux_sched_scheduler *) sp;
227 } */ *uap = v;
228 int error;
229 struct linux_sched_param lp;
230 struct proc *p;
231
232 /*
233 * We only check for valid parameters and return afterwards.
234 */
235
236 if (SCARG(uap, pid) < 0 || SCARG(uap, sp) == NULL)
237 return EINVAL;
238
239 error = copyin(SCARG(uap, sp), &lp, sizeof(lp));
240 if (error)
241 return error;
242
243 if (SCARG(uap, pid) != 0) {
244 kauth_cred_t pc = cl->l_cred;
245
246 if ((p = pfind(SCARG(uap, pid))) == NULL)
247 return ESRCH;
248 if (!(cl->l_proc == p ||
249 kauth_cred_geteuid(pc) == 0 ||
250 kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
251 kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
252 kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
253 kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
254 return EPERM;
255 }
256
257 return 0;
258 /*
259 * We can't emulate anything put the default scheduling policy.
260 */
261 if (SCARG(uap, policy) != LINUX_SCHED_OTHER || lp.sched_priority != 0)
262 return EINVAL;
263
264 return 0;
265 }
266
267 int
268 linux_sys_sched_getscheduler(cl, v, retval)
269 struct lwp *cl;
270 void *v;
271 register_t *retval;
272 {
273 struct linux_sys_sched_getscheduler_args /* {
274 syscallarg(linux_pid_t) pid;
275 } */ *uap = v;
276 struct proc *p;
277
278 *retval = -1;
279 /*
280 * We only check for valid parameters and return afterwards.
281 */
282
283 if (SCARG(uap, pid) != 0) {
284 kauth_cred_t pc = cl->l_cred;
285
286 if ((p = pfind(SCARG(uap, pid))) == NULL)
287 return ESRCH;
288 if (!(cl->l_proc == p ||
289 kauth_cred_geteuid(pc) == 0 ||
290 kauth_cred_getuid(pc) == kauth_cred_getuid(p->p_cred) ||
291 kauth_cred_geteuid(pc) == kauth_cred_getuid(p->p_cred) ||
292 kauth_cred_getuid(pc) == kauth_cred_geteuid(p->p_cred) ||
293 kauth_cred_geteuid(pc) == kauth_cred_geteuid(p->p_cred)))
294 return EPERM;
295 }
296
297 /*
298 * We can't emulate anything put the default scheduling policy.
299 */
300 *retval = LINUX_SCHED_OTHER;
301 return 0;
302 }
303
304 int
305 linux_sys_sched_yield(struct lwp *cl, void *v,
306 register_t *retval)
307 {
308
309 yield();
310 return 0;
311 }
312
313 int
314 linux_sys_sched_get_priority_max(struct lwp *cl, void *v,
315 register_t *retval)
316 {
317 struct linux_sys_sched_get_priority_max_args /* {
318 syscallarg(int) policy;
319 } */ *uap = v;
320
321 /*
322 * We can't emulate anything put the default scheduling policy.
323 */
324 if (SCARG(uap, policy) != LINUX_SCHED_OTHER) {
325 *retval = -1;
326 return EINVAL;
327 }
328
329 *retval = 0;
330 return 0;
331 }
332
333 int
334 linux_sys_sched_get_priority_min(struct lwp *cl, void *v,
335 register_t *retval)
336 {
337 struct linux_sys_sched_get_priority_min_args /* {
338 syscallarg(int) policy;
339 } */ *uap = v;
340
341 /*
342 * We can't emulate anything put the default scheduling policy.
343 */
344 if (SCARG(uap, policy) != LINUX_SCHED_OTHER) {
345 *retval = -1;
346 return EINVAL;
347 }
348
349 *retval = 0;
350 return 0;
351 }
352
353 #ifndef __m68k__
354 /* Present on everything but m68k */
355 int
356 linux_sys_exit_group(l, v, retval)
357 struct lwp *l;
358 void *v;
359 register_t *retval;
360 {
361 #ifdef LINUX_NPTL
362 struct linux_sys_exit_group_args /* {
363 syscallarg(int) error_code;
364 } */ *uap = v;
365 struct proc *p = l->l_proc;
366 struct linux_emuldata *led = p->p_emuldata;
367 struct linux_emuldata *e;
368
369 if (led->s->flags & LINUX_LES_USE_NPTL) {
370
371 #ifdef DEBUG_LINUX
372 printf("%s:%d, led->s->refs = %d\n", __func__, __LINE__,
373 led->s->refs);
374 #endif
375
376 /*
377 * The calling thread is supposed to kill all threads
378 * in the same thread group (i.e. all threads created
379 * via clone(2) with CLONE_THREAD flag set).
380 *
381 * If there is only one thread, things are quite simple
382 */
383 if (led->s->refs == 1)
384 return sys_exit(l, v, retval);
385
386 #ifdef DEBUG_LINUX
387 printf("%s:%d\n", __func__, __LINE__);
388 #endif
389
390 led->s->flags |= LINUX_LES_INEXITGROUP;
391 led->s->xstat = W_EXITCODE(SCARG(uap, error_code), 0);
392
393 /*
394 * Kill all threads in the group. The emulation exit hook takes
395 * care of hiding the zombies and reporting the exit code
396 * properly.
397 */
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 return 0;
411
412 }
413 #endif /* LINUX_NPTL */
414
415 return sys_exit(l, v, retval);
416 }
417 #endif /* !__m68k__ */
418
419 #ifdef LINUX_NPTL
420 int
421 linux_sys_set_tid_address(l, v, retval)
422 struct lwp *l;
423 void *v;
424 register_t *retval;
425 {
426 struct linux_sys_set_tid_address_args /* {
427 syscallarg(int *) tidptr;
428 } */ *uap = v;
429 struct linux_emuldata *led;
430
431 led = (struct linux_emuldata *)l->l_proc->p_emuldata;
432 led->clear_tid = SCARG(uap, tid);
433
434 led->s->flags |= LINUX_LES_USE_NPTL;
435
436 *retval = l->l_proc->p_pid;
437
438 return 0;
439 }
440
441 /* ARGUSED1 */
442 int
443 linux_sys_gettid(l, v, retval)
444 struct lwp *l;
445 void *v;
446 register_t *retval;
447 {
448 /* The Linux kernel does it exactly that way */
449 *retval = l->l_proc->p_pid;
450 return 0;
451 }
452
453 #ifdef LINUX_NPTL
454 /* ARGUSED1 */
455 int
456 linux_sys_getpid(l, v, retval)
457 struct lwp *l;
458 void *v;
459 register_t *retval;
460 {
461 struct linux_emuldata *led = l->l_proc->p_emuldata;
462
463 if (led->s->flags & LINUX_LES_USE_NPTL) {
464 /* The Linux kernel does it exactly that way */
465 *retval = led->s->group_pid;
466 } else {
467 *retval = l->l_proc->p_pid;
468 }
469
470 return 0;
471 }
472
473 /* ARGUSED1 */
474 int
475 linux_sys_getppid(l, v, retval)
476 struct lwp *l;
477 void *v;
478 register_t *retval;
479 {
480 struct proc *p = l->l_proc;
481 struct linux_emuldata *led = p->p_emuldata;
482 struct proc *glp;
483 struct proc *pp;
484
485 if (led->s->flags & LINUX_LES_USE_NPTL) {
486
487 /* Find the thread group leader's parent */
488 if ((glp = pfind(led->s->group_pid)) == NULL) {
489 /* Maybe panic... */
490 printf("linux_sys_getppid: missing group leader PID"
491 " %d\n", led->s->group_pid);
492 return -1;
493 }
494 pp = glp->p_pptr;
495
496 /* If this is a Linux process too, return thread group PID */
497 if (pp->p_emul == p->p_emul) {
498 struct linux_emuldata *pled;
499
500 pled = pp->p_emuldata;
501 *retval = pled->s->group_pid;
502 } else {
503 *retval = pp->p_pid;
504 }
505
506 } else {
507 *retval = p->p_pptr->p_pid;
508 }
509
510 return 0;
511 }
512 #endif /* LINUX_NPTL */
513
514 int
515 linux_sys_sched_getaffinity(l, v, retval)
516 struct lwp *l;
517 void *v;
518 register_t *retval;
519 {
520 struct linux_sys_sched_getaffinity_args /* {
521 syscallarg(pid_t) pid;
522 syscallarg(unsigned int) len;
523 syscallarg(unsigned long *) mask;
524 } */ *uap = v;
525 int error;
526 int ret;
527 int ncpu;
528 int name[2];
529 size_t sz;
530 char *data;
531 int *retp;
532
533 if (SCARG(uap, mask) == NULL)
534 return EINVAL;
535
536 if (SCARG(uap, len) < sizeof(int))
537 return EINVAL;
538
539 if (pfind(SCARG(uap, pid)) == NULL)
540 return ESRCH;
541
542 /*
543 * return the actual number of CPU, tag all of them as available
544 * The result is a mask, the first CPU being in the least significant
545 * bit.
546 */
547 name[0] = CTL_HW;
548 name[1] = HW_NCPU;
549 sz = sizeof(ncpu);
550
551 if ((error = old_sysctl(&name[0], 2, &ncpu, &sz, NULL, 0, NULL)) != 0)
552 return error;
553
554 ret = (1 << ncpu) - 1;
555
556 data = malloc(SCARG(uap, len), M_TEMP, M_WAITOK|M_ZERO);
557 retp = (int *)&data[SCARG(uap, len) - sizeof(ret)];
558 *retp = ret;
559
560 error = copyout(data, SCARG(uap, mask), SCARG(uap, len));
561
562 free(data, M_TEMP);
563
564 return error;
565
566 }
567
568 int
569 linux_sys_sched_setaffinity(l, v, retval)
570 struct lwp *l;
571 void *v;
572 register_t *retval;
573 {
574 struct linux_sys_sched_setaffinity_args /* {
575 syscallarg(pid_t) pid;
576 syscallarg(unsigned int) len;
577 syscallarg(unsigned long *) mask;
578 } */ *uap = v;
579
580 if (pfind(SCARG(uap, pid)) == NULL)
581 return ESRCH;
582
583 /* Let's ignore it */
584 #ifdef DEBUG_LINUX
585 printf("linux_sys_sched_setaffinity\n");
586 #endif
587 return 0;
588 };
589 #endif /* LINUX_NPTL */
Cache object: afca633995e6157ed36bc838e30a6540
|