xref: /netbsd-src/sys/rump/librump/rumpkern/emul.c (revision 93bf6008f8b7982c1d1a9486e4a4a0e687fe36eb)
1 /*	$NetBSD: emul.c,v 1.85 2009/03/30 12:47:49 christos Exp $	*/
2 
3 /*
4  * Copyright (c) 2007 Antti Kantee.  All Rights Reserved.
5  *
6  * Development of this software was supported by Google Summer of Code.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS
18  * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
19  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
20  * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
23  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __KERNEL_RCSID(0, "$NetBSD: emul.c,v 1.85 2009/03/30 12:47:49 christos Exp $");
32 
33 #include <sys/param.h>
34 #include <sys/malloc.h>
35 #include <sys/null.h>
36 #include <sys/vnode.h>
37 #include <sys/stat.h>
38 #include <sys/select.h>
39 #include <sys/syslog.h>
40 #include <sys/namei.h>
41 #include <sys/kauth.h>
42 #include <sys/conf.h>
43 #include <sys/device.h>
44 #include <sys/queue.h>
45 #include <sys/file.h>
46 #include <sys/filedesc.h>
47 #include <sys/kthread.h>
48 #include <sys/cpu.h>
49 #include <sys/kmem.h>
50 #include <sys/poll.h>
51 #include <sys/timetc.h>
52 #include <sys/tprintf.h>
53 #include <sys/module.h>
54 #include <sys/tty.h>
55 #include <sys/reboot.h>
56 
57 #include <dev/cons.h>
58 
59 #include <machine/stdarg.h>
60 
61 #include <rump/rumpuser.h>
62 
63 #include <uvm/uvm_map.h>
64 
65 #include "rump_private.h"
66 
67 time_t time_second = 1;
68 
69 kmutex_t *proc_lock;
70 struct lwp lwp0;
71 struct vnode *rootvp;
72 struct device *root_device;
73 dev_t rootdev;
74 int physmem = 256*256; /* 256 * 1024*1024 / 4k, PAGE_SIZE not always set */
75 int doing_shutdown;
76 int ncpu = 1;
77 const int schedppq = 1;
78 int hardclock_ticks;
79 bool mp_online = false;
80 struct vm_map *mb_map;
81 struct timeval boottime;
82 struct emul emul_netbsd;
83 int cold = 1;
84 int boothowto;
85 struct tty *constty;
86 
87 char hostname[MAXHOSTNAMELEN];
88 size_t hostnamelen;
89 
90 u_long	bufmem_valimit;
91 u_long	bufmem_hiwater;
92 u_long	bufmem_lowater;
93 u_long	bufmem;
94 u_int	nbuf;
95 
96 const char *panicstr;
97 const char ostype[] = "NetBSD";
98 const char osrelease[] = "999"; /* paradroid 4evah */
99 const char kernel_ident[] = "RUMP-ROAST";
100 const char *domainname;
101 int domainnamelen;
102 
103 const struct filterops seltrue_filtops;
104 const struct filterops sig_filtops;
105 
106 #define DEVSW_SIZE 255
107 const struct bdevsw *bdevsw0[DEVSW_SIZE]; /* XXX storage size */
108 const struct bdevsw **bdevsw = bdevsw0;
109 const int sys_cdevsws = DEVSW_SIZE;
110 int max_cdevsws = DEVSW_SIZE;
111 
112 const struct cdevsw *cdevsw0[DEVSW_SIZE]; /* XXX storage size */
113 const struct cdevsw **cdevsw = cdevsw0;
114 const int sys_bdevsws = DEVSW_SIZE;
115 int max_bdevsws = DEVSW_SIZE;
116 
117 struct devsw_conv devsw_conv0;
118 struct devsw_conv *devsw_conv = &devsw_conv0;
119 int max_devsw_convs = 0;
120 int mem_no = 2;
121 
122 kmutex_t tty_lock;
123 
124 int
125 copyin(const void *uaddr, void *kaddr, size_t len)
126 {
127 
128 	memcpy(kaddr, uaddr, len);
129 	return 0;
130 }
131 
132 int
133 copyout(const void *kaddr, void *uaddr, size_t len)
134 {
135 
136 	memcpy(uaddr, kaddr, len);
137 	return 0;
138 }
139 
140 int
141 copystr(const void *kfaddr, void *kdaddr, size_t len, size_t *done)
142 {
143 
144 	return copyinstr(kfaddr, kdaddr, len, done);
145 }
146 
147 int
148 copyinstr(const void *uaddr, void *kaddr, size_t len, size_t *done)
149 {
150 
151 	strlcpy(kaddr, uaddr, len);
152 	if (done)
153 		*done = strlen(kaddr)+1; /* includes termination */
154 	return 0;
155 }
156 
157 int
158 copyoutstr(const void *kaddr, void *uaddr, size_t len, size_t *done)
159 {
160 
161 	strlcpy(uaddr, kaddr, len);
162 	if (done)
163 		*done = strlen(uaddr)+1; /* includes termination */
164 	return 0;
165 }
166 
167 int
168 copyin_vmspace(struct vmspace *vm, const void *uaddr, void *kaddr, size_t len)
169 {
170 
171 	return copyin(uaddr, kaddr, len);
172 }
173 
174 int
175 copyout_vmspace(struct vmspace *vm, const void *kaddr, void *uaddr, size_t len)
176 {
177 
178 	return copyout(kaddr, uaddr, len);
179 }
180 
181 int
182 kcopy(const void *src, void *dst, size_t len)
183 {
184 
185 	memcpy(dst, src, len);
186 	return 0;
187 }
188 
189 int
190 uiomove(void *buf, size_t n, struct uio *uio)
191 {
192 	struct iovec *iov;
193 	uint8_t *b = buf;
194 	size_t cnt;
195 
196 	if (uio->uio_vmspace != UIO_VMSPACE_SYS)
197 		panic("%s: vmspace != UIO_VMSPACE_SYS", __func__);
198 
199 	while (n && uio->uio_resid) {
200 		iov = uio->uio_iov;
201 		cnt = iov->iov_len;
202 		if (cnt == 0) {
203 			uio->uio_iov++;
204 			uio->uio_iovcnt--;
205 			continue;
206 		}
207 		if (cnt > n)
208 			cnt = n;
209 
210 		if (uio->uio_rw == UIO_READ)
211 			memcpy(iov->iov_base, b, cnt);
212 		else
213 			memcpy(b, iov->iov_base, cnt);
214 
215 		iov->iov_base = (uint8_t *)iov->iov_base + cnt;
216 		iov->iov_len -= cnt;
217 		b += cnt;
218 		uio->uio_resid -= cnt;
219 		uio->uio_offset += cnt;
220 		n -= cnt;
221 	}
222 
223 	return 0;
224 }
225 
226 void
227 uio_setup_sysspace(struct uio *uio)
228 {
229 
230 	uio->uio_vmspace = UIO_VMSPACE_SYS;
231 }
232 
233 devclass_t
234 device_class(device_t dev)
235 {
236 
237 	if (dev != root_device)
238 		panic("%s: dev != root_device not supported", __func__);
239 
240 	return DV_DISK;
241 }
242 
243 void
244 getnanouptime(struct timespec *ts)
245 {
246 	uint64_t sec, nsec;
247 	int error;
248 
249 	/* XXX: this is wrong, does not report *uptime* */
250 	rumpuser_gettime(&sec, &nsec, &error);
251 	ts->tv_sec = sec;
252 	ts->tv_nsec = nsec;
253 }
254 
255 void
256 getmicrouptime(struct timeval *tv)
257 {
258 	uint64_t sec, nsec;
259 	int error;
260 
261 	/* XXX: this is wrong, does not report *uptime* */
262 	rumpuser_gettime(&sec, &nsec, &error);
263 	tv->tv_sec = sec;
264 	tv->tv_usec = nsec / 1000;
265 }
266 
267 void
268 malloc_type_attach(struct malloc_type *type)
269 {
270 
271 	return;
272 }
273 
274 void
275 malloc_type_detach(struct malloc_type *type)
276 {
277 
278 	return;
279 }
280 
281 void *
282 kern_malloc(unsigned long size, struct malloc_type *type, int flags)
283 {
284 	void *rv;
285 
286 	rv = rumpuser_malloc(size, (flags & (M_CANFAIL | M_NOWAIT)) != 0);
287 	if (rv && flags & M_ZERO)
288 		memset(rv, 0, size);
289 
290 	return rv;
291 }
292 
293 void *
294 kern_realloc(void *ptr, unsigned long size, struct malloc_type *type, int flags)
295 {
296 
297 	return rumpuser_malloc(size, (flags & (M_CANFAIL | M_NOWAIT)) != 0);
298 }
299 
300 void
301 kern_free(void *ptr, struct malloc_type *type)
302 {
303 
304 	rumpuser_free(ptr);
305 }
306 
307 static void
308 gettime(struct timespec *ts)
309 {
310 	uint64_t sec, nsec;
311 	int error;
312 
313 	rumpuser_gettime(&sec, &nsec, &error);
314 	ts->tv_sec = sec;
315 	ts->tv_nsec = nsec;
316 }
317 
318 void
319 nanotime(struct timespec *ts)
320 {
321 
322 	if (rump_threads) {
323 		rump_gettime(ts);
324 	} else {
325 		gettime(ts);
326 	}
327 }
328 
329 /* hooray for mick, so what if I do */
330 void
331 getnanotime(struct timespec *ts)
332 {
333 
334 	nanotime(ts);
335 }
336 
337 void
338 microtime(struct timeval *tv)
339 {
340 	struct timespec ts;
341 
342 	if (rump_threads) {
343 		rump_gettime(&ts);
344 		TIMESPEC_TO_TIMEVAL(tv, &ts);
345 	} else {
346 		gettime(&ts);
347 		TIMESPEC_TO_TIMEVAL(tv, &ts);
348 	}
349 }
350 
351 void
352 getmicrotime(struct timeval *tv)
353 {
354 
355 	microtime(tv);
356 }
357 
358 struct kthdesc {
359 	void (*f)(void *);
360 	void *arg;
361 	struct lwp *mylwp;
362 };
363 
364 static void *
365 threadbouncer(void *arg)
366 {
367 	struct kthdesc *k = arg;
368 	void (*f)(void *);
369 	void *thrarg;
370 
371 	f = k->f;
372 	thrarg = k->arg;
373 	rumpuser_set_curlwp(k->mylwp);
374 	kmem_free(k, sizeof(struct kthdesc));
375 
376 	if ((curlwp->l_pflag & LP_MPSAFE) == 0)
377 		KERNEL_LOCK(1, NULL);
378 	f(thrarg);
379 	panic("unreachable, should kthread_exit()");
380 }
381 
382 int
383 kthread_create(pri_t pri, int flags, struct cpu_info *ci,
384 	void (*func)(void *), void *arg, lwp_t **newlp, const char *fmt, ...)
385 {
386 	char thrstore[MAXCOMLEN];
387 	const char *thrname = NULL;
388 	va_list ap;
389 	struct kthdesc *k;
390 	struct lwp *l;
391 	int rv;
392 
393 	thrstore[0] = '\0';
394 	if (fmt) {
395 		va_start(ap, fmt);
396 		vsnprintf(thrstore, sizeof(thrstore), fmt, ap);
397 		va_end(ap);
398 		thrname = thrstore;
399 	}
400 
401 	/*
402 	 * We don't want a module unload thread.
403 	 * (XXX: yes, this is a kludge too, and the kernel should
404 	 * have a more flexible method for configuring which threads
405 	 * we want).
406 	 */
407 	if (strcmp(thrstore, "modunload") == 0) {
408 		return 0;
409 	}
410 
411 	if (!rump_threads) {
412 		/* fake them */
413 		if (strcmp(thrstore, "vrele") == 0) {
414 			printf("rump warning: threads not enabled, not starting"
415 			   " vrele thread\n");
416 			return 0;
417 		} else if (strcmp(thrstore, "cachegc") == 0) {
418 			printf("rump warning: threads not enabled, not starting"
419 			   " namecache g/c thread\n");
420 			return 0;
421 		} else if (strcmp(thrstore, "nfssilly") == 0) {
422 			printf("rump warning: threads not enabled, not enabling"
423 			   " nfs silly rename\n");
424 			return 0;
425 		} else
426 			panic("threads not available, setenv RUMP_THREADS 1");
427 	}
428 
429 	KASSERT(fmt != NULL);
430 	if (ci != NULL)
431 		panic("%s: bounded threads not supported", __func__);
432 
433 	k = kmem_alloc(sizeof(struct kthdesc), KM_SLEEP);
434 	k->f = func;
435 	k->arg = arg;
436 	k->mylwp = l = rump_setup_curlwp(0, rump_nextlid(), 0);
437 	if (flags & KTHREAD_MPSAFE)
438 		l->l_pflag |= LP_MPSAFE;
439 	rv = rumpuser_thread_create(threadbouncer, k, thrname);
440 	if (rv)
441 		return rv;
442 
443 	if (newlp)
444 		*newlp = l;
445 	return 0;
446 }
447 
448 void
449 kthread_exit(int ecode)
450 {
451 
452 	if ((curlwp->l_pflag & LP_MPSAFE) == 0)
453 		KERNEL_UNLOCK_ONE(NULL);
454 	rump_clear_curlwp();
455 	rumpuser_thread_exit();
456 }
457 
458 struct proc *
459 p_find(pid_t pid, uint flags)
460 {
461 
462 	panic("%s: not implemented", __func__);
463 }
464 
465 struct pgrp *
466 pg_find(pid_t pid, uint flags)
467 {
468 
469 	panic("%s: not implemented", __func__);
470 }
471 
472 void
473 psignal(struct proc *p, int signo)
474 {
475 
476 	switch (signo) {
477 	case SIGSYS:
478 		break;
479 	default:
480 		panic("unhandled signal %d", signo);
481 	}
482 }
483 
484 void
485 kpsignal(struct proc *p, ksiginfo_t *ksi, void *data)
486 {
487 
488 	panic("%s: not implemented", __func__);
489 }
490 
491 void
492 kpgsignal(struct pgrp *pgrp, ksiginfo_t *ksi, void *data, int checkctty)
493 {
494 
495 	panic("%s: not implemented", __func__);
496 }
497 
498 int
499 pgid_in_session(struct proc *p, pid_t pg_id)
500 {
501 
502 	panic("%s: not implemented", __func__);
503 }
504 
505 int
506 sigispending(struct lwp *l, int signo)
507 {
508 
509 	return 0;
510 }
511 
512 void
513 sigpending1(struct lwp *l, sigset_t *ss)
514 {
515 
516 	panic("%s: not implemented", __func__);
517 }
518 
519 int
520 kpause(const char *wmesg, bool intr, int timeo, kmutex_t *mtx)
521 {
522 	extern int hz;
523 	int rv, error;
524 	uint64_t sec, nsec;
525 
526 	if (mtx)
527 		mutex_exit(mtx);
528 
529 	sec = timeo / hz;
530 	nsec = (timeo % hz) * (1000000000 / hz);
531 	rv = rumpuser_nanosleep(&sec, &nsec, &error);
532 
533 	if (mtx)
534 		mutex_enter(mtx);
535 
536 	if (rv)
537 		return error;
538 
539 	return 0;
540 }
541 
542 void
543 suspendsched(void)
544 {
545 
546 	panic("%s: not implemented", __func__);
547 }
548 
549 u_int
550 lwp_unsleep(lwp_t *l, bool cleanup)
551 {
552 
553 	KASSERT(mutex_owned(l->l_mutex));
554 
555 	return (*l->l_syncobj->sobj_unsleep)(l, cleanup);
556 }
557 
558 vaddr_t
559 calc_cache_size(struct vm_map *map, int pct, int va_pct)
560 {
561 	paddr_t t;
562 
563 	t = (paddr_t)physmem * pct / 100 * PAGE_SIZE;
564 	if ((vaddr_t)t != t) {
565 		panic("%s: needs tweak", __func__);
566 	}
567 	return t;
568 }
569 
570 int
571 seltrue(dev_t dev, int events, struct lwp *l)
572 {
573         return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
574 }
575 
576 void
577 selrecord(lwp_t *selector, struct selinfo *sip)
578 {
579 }
580 
581 void
582 selinit(struct selinfo *sip)
583 {
584 }
585 
586 void
587 selnotify(struct selinfo *sip, int events, long knhint)
588 {
589 }
590 
591 void
592 seldestroy(struct selinfo *sip)
593 {
594 }
595 
596 const char *
597 device_xname(device_t dv)
598 {
599 	return "bogus0";
600 }
601 
602 void
603 assert_sleepable(void)
604 {
605 
606 	/* always sleepable, although we should improve this */
607 }
608 
609 void
610 tc_setclock(const struct timespec *ts)
611 {
612 
613 	panic("%s: not implemented", __func__);
614 }
615 
616 void
617 proc_crmod_enter(void)
618 {
619 
620 	panic("%s: not implemented", __func__);
621 }
622 
623 void
624 proc_crmod_leave(kauth_cred_t c1, kauth_cred_t c2, bool sugid)
625 {
626 
627 	panic("%s: not implemented", __func__);
628 }
629 
630 void
631 module_init_md(void)
632 {
633 
634 	/*
635 	 * Nothing for now.  However, we should load the librump
636 	 * symbol table.
637 	 */
638 }
639 
640 /* us and them, after all we're only ordinary seconds */
641 static void
642 rump_delay(unsigned int us)
643 {
644 	uint64_t sec, nsec;
645 	int error;
646 
647 	sec = us / 1000000;
648 	nsec = (us % 1000000) * 1000;
649 
650 	if (__predict_false(sec != 0))
651 		printf("WARNING: over 1s delay\n");
652 
653 	rumpuser_nanosleep(&sec, &nsec, &error);
654 }
655 void (*delay_func)(unsigned int) = rump_delay;
656 
657 void
658 kpreempt_disable(void)
659 {
660 
661 	/* XXX: see below */
662 	KPREEMPT_DISABLE(curlwp);
663 }
664 
665 void
666 kpreempt_enable(void)
667 {
668 
669 	/* try to make sure kpreempt_disable() is only used from panic() */
670 	panic("kpreempt not supported");
671 }
672 
673 void
674 sessdelete(struct session *ss)
675 {
676 
677 	panic("sessdelete() impossible, session %p", ss);
678 }
679 
680 int
681 ttycheckoutq(struct tty *tp, int wait)
682 {
683 
684 	return 1;
685 }
686 
687 void
688 cnputc(int c)
689 {
690 	int error;
691 
692 	rumpuser_putchar(c, &error);
693 }
694 
695 void
696 cnflush(void)
697 {
698 
699 	/* done */
700 }
701 
702 int
703 tputchar(int c, int flags, struct tty *tp)
704 {
705 
706 	cnputc(c);
707 	return 0;
708 }
709 
710 void
711 cpu_reboot(int howto, char *bootstr)
712 {
713 
714 	rumpuser_panic();
715 }
716 
717 /* XXX: static, but not used except to make spcopy.S link */
718 #ifdef __hppa__
719 #undef curlwp
720 struct lwp *curlwp = &lwp0;
721 #endif
722 
723 /*
724  * XXX: from sys_select.c, see that file for license.
725  * (these will go away really soon in favour of the real sys_select.c)
726  * ((really, the select code just needs cleanup))
727  * (((seriously)))
728  */
729 int
730 inittimeleft(struct timespec *ts, struct timespec *sleepts)
731 {
732 	if (itimespecfix(ts))
733 		return -1;
734 	getnanouptime(sleepts);
735 	return 0;
736 }
737 
738 int
739 gettimeleft(struct timespec *ts, struct timespec *sleepts)
740 {
741 	/*
742 	 * We have to recalculate the timeout on every retry.
743 	 */
744 	struct timespec sleptts;
745 	/*
746 	 * reduce ts by elapsed time
747 	 * based on monotonic time scale
748 	 */
749 	getnanouptime(&sleptts);
750 	timespecadd(ts, sleepts, ts);
751 	timespecsub(ts, &sleptts, ts);
752 	*sleepts = sleptts;
753 	return tstohz(ts);
754 }
755