1 /*
2  * Copyright (c) 2006 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Matthew Dillon <dillon@backplane.com>
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
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
15  *    the documentation and/or other materials provided with the
16  *    distribution.
17  * 3. Neither the name of The DragonFly Project nor the names of its
18  *    contributors may be used to endorse or promote products derived
19  *    from this software without specific, prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 #include <sys/types.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
38 #include <sys/stat.h>
39 #include <sys/mman.h>
40 #include <sys/cons.h>
41 #include <sys/random.h>
42 #include <sys/vkernel.h>
43 #include <sys/tls.h>
44 #include <sys/reboot.h>
45 #include <sys/proc.h>
46 #include <sys/msgbuf.h>
47 #include <sys/vmspace.h>
48 #include <sys/socket.h>
49 #include <sys/sockio.h>
50 #include <sys/sysctl.h>
51 #include <sys/un.h>
52 #include <vm/vm_page.h>
53 #include <vm/vm_map.h>
54 #include <sys/mplock2.h>
55 
56 #include <machine/cpu.h>
57 #include <machine/globaldata.h>
58 #include <machine/tls.h>
59 #include <machine/md_var.h>
60 #include <machine/vmparam.h>
61 #include <cpu/specialreg.h>
62 
63 #include <net/if.h>
64 #include <net/if_arp.h>
65 #include <net/ethernet.h>
66 #include <net/bridge/if_bridgevar.h>
67 #include <netinet/in.h>
68 #include <arpa/inet.h>
69 #include <net/if_var.h>
70 
71 #include <stdio.h>
72 #include <stdlib.h>
73 #include <stdarg.h>
74 #include <stdbool.h>
75 #include <unistd.h>
76 #include <fcntl.h>
77 #include <string.h>
78 #include <err.h>
79 #include <errno.h>
80 #include <assert.h>
81 #include <sysexits.h>
82 
83 vm_paddr_t phys_avail[16];
84 vm_paddr_t Maxmem;
85 vm_paddr_t Maxmem_bytes;
86 long physmem;
87 int MemImageFd = -1;
88 struct vkdisk_info DiskInfo[VKDISK_MAX];
89 int DiskNum;
90 struct vknetif_info NetifInfo[VKNETIF_MAX];
91 int NetifNum;
92 char *pid_file;
93 vm_offset_t KvaStart;
94 vm_offset_t KvaEnd;
95 vm_offset_t KvaSize;
96 vm_offset_t virtual_start;
97 vm_offset_t virtual_end;
98 vm_offset_t virtual2_start;
99 vm_offset_t virtual2_end;
100 vm_offset_t kernel_vm_end;
101 vm_offset_t crashdumpmap;
102 vm_offset_t clean_sva;
103 vm_offset_t clean_eva;
104 struct msgbuf *msgbufp;
105 caddr_t ptvmmap;
106 vpte_t	*KernelPTD;
107 vpte_t	*KernelPTA;	/* Warning: Offset for direct VA translation */
108 void *dmap_min_address;
109 u_int cpu_feature;	/* XXX */
110 int tsc_present;
111 int64_t tsc_frequency;
112 int optcpus;		/* number of cpus - see mp_start() */
113 int lwp_cpu_lock;	/* if/how to lock virtual CPUs to real CPUs */
114 int real_ncpus;		/* number of real CPUs */
115 int next_cpu;		/* next real CPU to lock a virtual CPU to */
116 int vkernel_b_arg;	/* -b argument - no of logical CPU bits - only SMP */
117 int vkernel_B_arg;	/* -B argument - no of core bits - only SMP */
118 
119 struct privatespace *CPU_prvspace;
120 
121 static struct trapframe proc0_tf;
122 static void *proc0paddr;
123 
124 static void init_sys_memory(char *imageFile);
125 static void init_kern_memory(void);
126 static void init_globaldata(void);
127 static void init_vkernel(void);
128 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
129 static void init_netif(char *netifExp[], int netifFileNum);
130 static void writepid(void);
131 static void cleanpid(void);
132 static int unix_connect(const char *path);
133 static void usage_err(const char *ctl, ...);
134 static void usage_help(_Bool);
135 static void init_locks(void);
136 
137 static int save_ac;
138 static char **save_av;
139 
140 /*
141  * Kernel startup for virtual kernels - standard main()
142  */
143 int
144 main(int ac, char **av)
145 {
146 	char *memImageFile = NULL;
147 	char *netifFile[VKNETIF_MAX];
148 	char *diskFile[VKDISK_MAX];
149 	char *cdFile[VKDISK_MAX];
150 	char *suffix;
151 	char *endp;
152 	char *tmp;
153 	char *tok;
154 	int netifFileNum = 0;
155 	int diskFileNum = 0;
156 	int cdFileNum = 0;
157 	int bootOnDisk = -1;	/* set below to vcd (0) or vkd (1) */
158 	int c;
159 	int i;
160 	int j;
161 	int n;
162 	int isq;
163 	int pos;
164 	int eflag;
165 	int real_vkernel_enable;
166 	int supports_sse;
167 	size_t vsize;
168 	size_t kenv_size;
169 	size_t kenv_size2;
170 
171 	save_ac = ac;
172 	save_av = av;
173 	eflag = 0;
174 	pos = 0;
175 	kenv_size = 0;
176 
177 	/*
178 	 * Process options
179 	 */
180 	kernel_mem_readonly = 1;
181 	optcpus = 2;
182 	vkernel_b_arg = 0;
183 	vkernel_B_arg = 0;
184 	lwp_cpu_lock = LCL_NONE;
185 
186 	real_vkernel_enable = 0;
187 	vsize = sizeof(real_vkernel_enable);
188 	sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
189 
190 	if (real_vkernel_enable == 0) {
191 		errx(1, "vm.vkernel_enable is 0, must be set "
192 			"to 1 to execute a vkernel!");
193 	}
194 
195 	real_ncpus = 1;
196 	vsize = sizeof(real_ncpus);
197 	sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
198 
199 	if (ac < 2)
200 		usage_help(false);
201 
202 	while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:Ub:B:")) != -1) {
203 		switch(c) {
204 		case 'e':
205 			/*
206 			 * name=value:name=value:name=value...
207 			 * name="value"...
208 			 *
209 			 * Allow values to be quoted but note that shells
210 			 * may remove the quotes, so using this feature
211 			 * to embed colons may require a backslash.
212 			 */
213 			n = strlen(optarg);
214 			isq = 0;
215 
216 			if (eflag == 0) {
217 				kenv_size = n + 2;
218 				kern_envp = malloc(kenv_size);
219 				if (kern_envp == NULL)
220 					errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size);
221 			} else {
222 				kenv_size2 = kenv_size + n + 1;
223 				pos = kenv_size - 1;
224 				if ((tmp = realloc(kern_envp, kenv_size2)) == NULL)
225 					errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2);
226 				kern_envp = tmp;
227 				kenv_size = kenv_size2;
228 			}
229 
230 			for (i = 0, j = pos; i < n; ++i) {
231 				if (optarg[i] == '"')
232 					isq ^= 1;
233 				else if (optarg[i] == '\'')
234 					isq ^= 2;
235 				else if (isq == 0 && optarg[i] == ':')
236 					kern_envp[j++] = 0;
237 				else
238 					kern_envp[j++] = optarg[i];
239 			}
240 			kern_envp[j++] = 0;
241 			kern_envp[j++] = 0;
242 			eflag++;
243 			break;
244 		case 's':
245 			boothowto |= RB_SINGLE;
246 			break;
247 		case 'v':
248 			bootverbose = 1;
249 			break;
250 		case 'i':
251 			memImageFile = optarg;
252 			break;
253 		case 'I':
254 			if (netifFileNum < VKNETIF_MAX)
255 				netifFile[netifFileNum++] = strdup(optarg);
256 			break;
257 		case 'r':
258 			if (bootOnDisk < 0)
259 				bootOnDisk = 1;
260 			if (diskFileNum + cdFileNum < VKDISK_MAX)
261 				diskFile[diskFileNum++] = strdup(optarg);
262 			break;
263 		case 'c':
264 			if (bootOnDisk < 0)
265 				bootOnDisk = 0;
266 			if (diskFileNum + cdFileNum < VKDISK_MAX)
267 				cdFile[cdFileNum++] = strdup(optarg);
268 			break;
269 		case 'm':
270 			Maxmem_bytes = strtoull(optarg, &suffix, 0);
271 			if (suffix) {
272 				switch(*suffix) {
273 				case 'g':
274 				case 'G':
275 					Maxmem_bytes <<= 30;
276 					break;
277 				case 'm':
278 				case 'M':
279 					Maxmem_bytes <<= 20;
280 					break;
281 				case 'k':
282 				case 'K':
283 					Maxmem_bytes <<= 10;
284 					break;
285 				default:
286 					Maxmem_bytes = 0;
287 					usage_err("Bad maxmem option");
288 					/* NOT REACHED */
289 					break;
290 				}
291 			}
292 			break;
293 		case 'l':
294 			next_cpu = -1;
295 			if (strncmp("map", optarg, 3) == 0) {
296 				lwp_cpu_lock = LCL_PER_CPU;
297 				if (optarg[3] == ',') {
298 					next_cpu = strtol(optarg+4, &endp, 0);
299 					if (*endp != '\0')
300 						usage_err("Bad target CPU number at '%s'", endp);
301 				} else {
302 					next_cpu = 0;
303 				}
304 				if (next_cpu < 0 || next_cpu > real_ncpus - 1)
305 					usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
306 			} else if (strncmp("any", optarg, 3) == 0) {
307 				lwp_cpu_lock = LCL_NONE;
308 			} else {
309 				lwp_cpu_lock = LCL_SINGLE_CPU;
310 				next_cpu = strtol(optarg, &endp, 0);
311 				if (*endp != '\0')
312 					usage_err("Bad target CPU number at '%s'", endp);
313 				if (next_cpu < 0 || next_cpu > real_ncpus - 1)
314 					usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
315 			}
316 			break;
317 		case 'n':
318 			/*
319 			 * This value is set up by mp_start(), don't just
320 			 * set ncpus here.
321 			 */
322 			tok = strtok(optarg, ":");
323 			optcpus = strtol(tok, NULL, 0);
324 			if (optcpus < 1 || optcpus > MAXCPU)
325 				usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
326 
327 			/* :lbits argument */
328 			tok = strtok(NULL, ":");
329 			if (tok != NULL) {
330 				vkernel_b_arg = strtol(tok, NULL, 0);
331 
332 				/* :cbits argument */
333 				tok = strtok(NULL, ":");
334 				if (tok != NULL) {
335 					vkernel_B_arg = strtol(tok, NULL, 0);
336 				}
337 
338 			}
339 			break;
340 		case 'p':
341 			pid_file = optarg;
342 			break;
343 		case 'U':
344 			kernel_mem_readonly = 0;
345 			break;
346 		case 'h':
347 			usage_help(true);
348 			break;
349 		default:
350 			usage_help(false);
351 		}
352 	}
353 
354 	writepid();
355 	cpu_disable_intr();
356 	init_sys_memory(memImageFile);
357 	init_kern_memory();
358 	init_globaldata();
359 	init_vkernel();
360 	setrealcpu();
361 	init_kqueue();
362 
363 	vmm_guest = 1;
364 
365 	/*
366 	 * Check TSC
367 	 */
368 	vsize = sizeof(tsc_present);
369 	sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
370 	vsize = sizeof(tsc_frequency);
371 	sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
372 	if (tsc_present)
373 		cpu_feature |= CPUID_TSC;
374 
375 	/*
376 	 * Check SSE
377 	 */
378 	vsize = sizeof(supports_sse);
379 	supports_sse = 0;
380 	sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
381 	init_fpu(supports_sse);
382 	if (supports_sse)
383 		cpu_feature |= CPUID_SSE | CPUID_FXSR;
384 
385 	/*
386 	 * We boot from the first installed disk.
387 	 */
388 	if (bootOnDisk == 1) {
389 		init_disk(diskFile, diskFileNum, VKD_DISK);
390 		init_disk(cdFile, cdFileNum, VKD_CD);
391 	} else {
392 		init_disk(cdFile, cdFileNum, VKD_CD);
393 		init_disk(diskFile, diskFileNum, VKD_DISK);
394 	}
395 	init_netif(netifFile, netifFileNum);
396 	init_exceptions();
397 	mi_startup();
398 	/* NOT REACHED */
399 	exit(EX_SOFTWARE);
400 }
401 
402 /*
403  * Initialize system memory.  This is the virtual kernel's 'RAM'.
404  */
405 static
406 void
407 init_sys_memory(char *imageFile)
408 {
409 	struct stat st;
410 	int i;
411 	int fd;
412 
413 	/*
414 	 * Figure out the system memory image size.  If an image file was
415 	 * specified and -m was not specified, use the image file's size.
416 	 */
417 	if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
418 		Maxmem_bytes = (vm_paddr_t)st.st_size;
419 	if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
420 	    Maxmem_bytes == 0) {
421 		errx(1, "Cannot create new memory file %s unless "
422 		       "system memory size is specified with -m",
423 		       imageFile);
424 		/* NOT REACHED */
425 	}
426 
427 	/*
428 	 * Maxmem must be known at this time
429 	 */
430 	if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
431 		errx(1, "Bad maxmem specification: 64MB minimum, "
432 		       "multiples of %dMB only",
433 		       SEG_SIZE / 1024 / 1024);
434 		/* NOT REACHED */
435 	}
436 
437 	/*
438 	 * Generate an image file name if necessary, then open/create the
439 	 * file exclusively locked.  Do not allow multiple virtual kernels
440 	 * to use the same image file.
441 	 *
442 	 * Don't iterate through a million files if we do not have write
443 	 * access to the directory, stop if our open() failed on a
444 	 * non-existant file.  Otherwise opens can fail for any number
445 	 */
446 	if (imageFile == NULL) {
447 		for (i = 0; i < 1000000; ++i) {
448 			asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
449 			fd = open(imageFile,
450 				  O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
451 			if (fd < 0 && stat(imageFile, &st) == 0) {
452 				free(imageFile);
453 				continue;
454 			}
455 			break;
456 		}
457 	} else {
458 		fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
459 	}
460 	fprintf(stderr, "Using memory file: %s\n", imageFile);
461 	if (fd < 0 || fstat(fd, &st) < 0) {
462 		err(1, "Unable to open/create %s", imageFile);
463 		/* NOT REACHED */
464 	}
465 
466 	/*
467 	 * Truncate or extend the file as necessary.  Clean out the contents
468 	 * of the file, we want it to be full of holes so we don't waste
469 	 * time reading in data from an old file that we no longer care
470 	 * about.
471 	 */
472 	ftruncate(fd, 0);
473 	ftruncate(fd, Maxmem_bytes);
474 
475 	MemImageFd = fd;
476 	Maxmem = Maxmem_bytes >> PAGE_SHIFT;
477 	physmem = Maxmem;
478 }
479 
480 /*
481  * Initialize kernel memory.  This reserves kernel virtual memory by using
482  * MAP_VPAGETABLE
483  */
484 
485 static
486 void
487 init_kern_memory(void)
488 {
489 	void *base;
490 	void *try;
491 	char dummy;
492 	char *topofstack = &dummy;
493 	int i;
494 	void *firstfree;
495 
496 	/*
497 	 * Memory map our kernel virtual memory space.  Note that the
498 	 * kernel image itself is not made part of this memory for the
499 	 * moment.
500 	 *
501 	 * The memory map must be segment-aligned so we can properly
502 	 * offset KernelPTD.
503 	 *
504 	 * If the system kernel has a different MAXDSIZ, it might not
505 	 * be possible to map kernel memory in its prefered location.
506 	 * Try a number of different locations.
507 	 */
508 	try = (void *)(512UL << 30);
509 	base = NULL;
510 	while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
511 		base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
512 			    MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
513 			    MemImageFd, (off_t)try);
514 		if (base == try)
515 			break;
516 		if (base != MAP_FAILED)
517 			munmap(base, KERNEL_KVA_SIZE);
518 		try = (char *)try + (512UL << 30);
519 	}
520 	if (base != try) {
521 		err(1, "Unable to mmap() kernel virtual memory!");
522 		/* NOT REACHED */
523 	}
524 	madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
525 	KvaStart = (vm_offset_t)base;
526 	KvaSize = KERNEL_KVA_SIZE;
527 	KvaEnd = KvaStart + KvaSize;
528 
529 	/* cannot use kprintf yet */
530 	printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
531 
532 	/* MAP_FILE? */
533 	dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE,
534 				MAP_NOCORE|MAP_NOSYNC|MAP_SHARED,
535 				MemImageFd, 0);
536 	if (dmap_min_address == MAP_FAILED) {
537 		err(1, "Unable to mmap() kernel DMAP region!");
538 		/* NOT REACHED */
539 	}
540 
541 	/*
542 	 * Bootstrap the kernel_pmap
543 	 */
544 	firstfree = NULL;
545 	pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base);
546 
547 	mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
548 		 0 | VPTE_R | VPTE_W | VPTE_V);
549 
550 	/*
551 	 * phys_avail[] represents unallocated physical memory.  MI code
552 	 * will use phys_avail[] to create the vm_page array.
553 	 */
554 	phys_avail[0] = (vm_paddr_t)firstfree;
555 	phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
556 	phys_avail[1] = Maxmem_bytes;
557 
558 #if JGV
559 	/*
560 	 * (virtual_start, virtual_end) represent unallocated kernel virtual
561 	 * memory.  MI code will create kernel_map using these parameters.
562 	 */
563 	virtual_start = KvaStart + (long)firstfree;
564 	virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
565 	virtual_end = KvaStart + KERNEL_KVA_SIZE;
566 #endif
567 
568 	/*
569 	 * pmap_growkernel() will set the correct value.
570 	 */
571 	kernel_vm_end = 0;
572 
573 	/*
574 	 * Allocate space for process 0's UAREA.
575 	 */
576 	proc0paddr = (void *)virtual_start;
577 	for (i = 0; i < UPAGES; ++i) {
578 		pmap_kenter_quick(virtual_start, phys_avail[0]);
579 		virtual_start += PAGE_SIZE;
580 		phys_avail[0] += PAGE_SIZE;
581 	}
582 
583 	/*
584 	 * crashdumpmap
585 	 */
586 	crashdumpmap = virtual_start;
587 	virtual_start += MAXDUMPPGS * PAGE_SIZE;
588 
589 	/*
590 	 * msgbufp maps the system message buffer
591 	 */
592 	assert((MSGBUF_SIZE & PAGE_MASK) == 0);
593 	msgbufp = (void *)virtual_start;
594 	for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
595 		pmap_kenter_quick(virtual_start, phys_avail[0]);
596 		virtual_start += PAGE_SIZE;
597 		phys_avail[0] += PAGE_SIZE;
598 	}
599 	msgbufinit(msgbufp, MSGBUF_SIZE);
600 
601 	/*
602 	 * used by kern_memio for /dev/mem access
603 	 */
604 	ptvmmap = (caddr_t)virtual_start;
605 	virtual_start += PAGE_SIZE;
606 }
607 
608 /*
609  * Map the per-cpu globaldata for cpu #0.  Allocate the space using
610  * virtual_start and phys_avail[0]
611  */
612 static
613 void
614 init_globaldata(void)
615 {
616 	int i;
617 	vm_paddr_t pa;
618 	vm_offset_t va;
619 
620 	/*
621 	 * Reserve enough KVA to cover possible cpus.  This is a considerable
622 	 * amount of KVA since the privatespace structure includes two
623 	 * whole page table mappings.
624 	 */
625 	virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
626 	CPU_prvspace = (void *)virtual_start;
627 	virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
628 
629 	/*
630 	 * Allocate enough physical memory to cover the mdglobaldata
631 	 * portion of the space and the idle stack and map the pages
632 	 * into KVA.  For cpu #0 only.
633 	 */
634 	for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
635 		pa = phys_avail[0];
636 		va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
637 		pmap_kenter_quick(va, pa);
638 		phys_avail[0] += PAGE_SIZE;
639 	}
640 	for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
641 		pa = phys_avail[0];
642 		va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
643 		pmap_kenter_quick(va, pa);
644 		phys_avail[0] += PAGE_SIZE;
645 	}
646 
647 	/*
648 	 * Setup the %gs for cpu #0.  The mycpu macro works after this
649 	 * point.  Note that %fs is used by pthreads.
650 	 */
651 	tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace));
652 }
653 
654 
655 /*
656  * Initialize pool tokens and other necessary locks
657  */
658 static void
659 init_locks(void)
660 {
661 
662         /*
663          * Get the initial mplock with a count of 1 for the BSP.
664          * This uses a LOGICAL cpu ID, ie BSP == 0.
665          */
666         cpu_get_initial_mplock();
667 
668         /* our token pool needs to work early */
669         lwkt_token_pool_init();
670 
671 }
672 
673 
674 /*
675  * Initialize very low level systems including thread0, proc0, etc.
676  */
677 static
678 void
679 init_vkernel(void)
680 {
681 	struct mdglobaldata *gd;
682 
683 	gd = &CPU_prvspace[0].mdglobaldata;
684 	bzero(gd, sizeof(*gd));
685 
686 	gd->mi.gd_curthread = &thread0;
687 	thread0.td_gd = &gd->mi;
688 	ncpus = 1;
689 	ncpus2 = 1;	/* rounded down power of 2 */
690 	ncpus_fit = 1;	/* rounded up power of 2 */
691 	/* ncpus2_mask and ncpus_fit_mask are 0 */
692 	init_param1();
693 	gd->mi.gd_prvspace = &CPU_prvspace[0];
694 	mi_gdinit(&gd->mi, 0);
695 	cpu_gdinit(gd, 0);
696 	mi_proc0init(&gd->mi, proc0paddr);
697 	lwp0.lwp_md.md_regs = &proc0_tf;
698 
699 	init_locks();
700 	cninit();
701 	rand_initialize();
702 #if 0	/* #ifdef DDB */
703 	kdb_init();
704 	if (boothowto & RB_KDB)
705 		Debugger("Boot flags requested debugger");
706 #endif
707 	identcpu();
708 #if 0
709 	initializecpu();	/* Initialize CPU registers */
710 #endif
711 	init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
712 
713 #if 0
714 	/*
715 	 * Map the message buffer
716 	 */
717 	for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
718 		pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
719 	msgbufinit(msgbufp, MSGBUF_SIZE);
720 #endif
721 #if 0
722 	thread0.td_pcb_cr3 ... MMU
723 	lwp0.lwp_md.md_regs = &proc0_tf;
724 #endif
725 }
726 
727 /*
728  * Filesystem image paths for the virtual kernel are optional.
729  * If specified they each should point to a disk image,
730  * the first of which will become the root disk.
731  *
732  * The virtual kernel caches data from our 'disk' just like a normal kernel,
733  * so we do not really want the real kernel to cache the data too.  Use
734  * O_DIRECT to remove the duplication.
735  */
736 static
737 void
738 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
739 {
740 	char *serno;
741 	int i;
742 
743         if (diskFileNum == 0)
744                 return;
745 
746 	for(i=0; i < diskFileNum; i++){
747 		char *fname;
748 		fname = diskExp[i];
749 
750 		if (fname == NULL) {
751                         warnx("Invalid argument to '-r'");
752                         continue;
753                 }
754 		/*
755 		 * Check for a serial number for the virtual disk
756 		 * passed from the command line.
757 		 */
758 		serno = fname;
759 		strsep(&serno, ":");
760 
761 		if (DiskNum < VKDISK_MAX) {
762 			struct stat st;
763 			struct vkdisk_info* info = NULL;
764 			int fd;
765 			size_t l = 0;
766 
767 			if (type == VKD_DISK)
768 			    fd = open(fname, O_RDWR|O_DIRECT, 0644);
769 			else
770 			    fd = open(fname, O_RDONLY|O_DIRECT, 0644);
771 			if (fd < 0 || fstat(fd, &st) < 0) {
772 				err(1, "Unable to open/create %s", fname);
773 				/* NOT REACHED */
774 			}
775 			if (S_ISREG(st.st_mode)) {
776 				if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
777 					errx(1, "Disk image %s is already "
778 						"in use\n", fname);
779 					/* NOT REACHED */
780 				}
781 			}
782 
783 			info = &DiskInfo[DiskNum];
784 			l = strlen(fname);
785 
786 			info->unit = i;
787 			info->fd = fd;
788 			info->type = type;
789 			memcpy(info->fname, fname, l);
790 			info->serno = NULL;
791 			if (serno) {
792 				if ((info->serno = malloc(SERNOLEN)) != NULL)
793 					strlcpy(info->serno, serno, SERNOLEN);
794 				else
795 					warnx("Couldn't allocate memory for the operation");
796 			}
797 
798 			if (DiskNum == 0) {
799 				if (type == VKD_CD) {
800 				    rootdevnames[0] = "cd9660:vcd0a";
801 				} else if (type == VKD_DISK) {
802 				    rootdevnames[0] = "ufs:vkd0s0a";
803 				    rootdevnames[1] = "ufs:vkd0s1a";
804 				}
805 			}
806 
807 			DiskNum++;
808 		} else {
809                         warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
810                         continue;
811 		}
812 	}
813 }
814 
815 static
816 int
817 netif_set_tapflags(int tap_unit, int f, int s)
818 {
819 	struct ifreq ifr;
820 	int flags;
821 
822 	bzero(&ifr, sizeof(ifr));
823 
824 	snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
825 	if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
826 		warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
827 		return -1;
828 	}
829 
830 	/*
831 	 * Adjust if_flags
832 	 *
833 	 * If the flags are already set/cleared, then we return
834 	 * immediately to avoid extra syscalls
835 	 */
836 	flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
837 	if (f < 0) {
838 		/* Turn off flags */
839 		f = -f;
840 		if ((flags & f) == 0)
841 			return 0;
842 		flags &= ~f;
843 	} else {
844 		/* Turn on flags */
845 		if (flags & f)
846 			return 0;
847 		flags |= f;
848 	}
849 
850 	/*
851 	 * Fix up ifreq.ifr_name, since it may be trashed
852 	 * in previous ioctl(SIOCGIFFLAGS)
853 	 */
854 	snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
855 
856 	ifr.ifr_flags = flags & 0xffff;
857 	ifr.ifr_flagshigh = flags >> 16;
858 	if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
859 		warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
860 		return -1;
861 	}
862 	return 0;
863 }
864 
865 static
866 int
867 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
868 {
869 	struct ifaliasreq ifra;
870 	struct sockaddr_in *in;
871 
872 	bzero(&ifra, sizeof(ifra));
873 	snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
874 
875 	/* Setup address */
876 	in = (struct sockaddr_in *)&ifra.ifra_addr;
877 	in->sin_family = AF_INET;
878 	in->sin_len = sizeof(*in);
879 	in->sin_addr.s_addr = addr;
880 
881 	if (mask != 0) {
882 		/* Setup netmask */
883 		in = (struct sockaddr_in *)&ifra.ifra_mask;
884 		in->sin_len = sizeof(*in);
885 		in->sin_addr.s_addr = mask;
886 	}
887 
888 	if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
889 		warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
890 		return -1;
891 	}
892 	return 0;
893 }
894 
895 static
896 int
897 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
898 {
899 	struct ifbreq ifbr;
900 	struct ifdrv ifd;
901 
902 	bzero(&ifbr, sizeof(ifbr));
903 	snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
904 		 "tap%d", tap_unit);
905 
906 	bzero(&ifd, sizeof(ifd));
907 	strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
908 	ifd.ifd_cmd = BRDGADD;
909 	ifd.ifd_len = sizeof(ifbr);
910 	ifd.ifd_data = &ifbr;
911 
912 	if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
913 		/*
914 		 * 'errno == EEXIST' means that the tap(4) is already
915 		 * a member of the bridge(4)
916 		 */
917 		if (errno != EEXIST) {
918 			warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
919 			return -1;
920 		}
921 	}
922 	return 0;
923 }
924 
925 #define TAPDEV_OFLAGS	(O_RDWR | O_NONBLOCK)
926 
927 /*
928  * Locate the first unused tap(4) device file if auto mode is requested,
929  * or open the user supplied device file, and bring up the corresponding
930  * tap(4) interface.
931  *
932  * NOTE: Only tap(4) device file is supported currently
933  */
934 static
935 int
936 netif_open_tap(const char *netif, int *tap_unit, int s)
937 {
938 	char tap_dev[MAXPATHLEN];
939 	int tap_fd, failed;
940 	struct stat st;
941 	char *dname;
942 
943 	*tap_unit = -1;
944 
945 	if (strcmp(netif, "auto") == 0) {
946 		/*
947 		 * Find first unused tap(4) device file
948 		 */
949 		tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
950 		if (tap_fd < 0) {
951 			warnc(errno, "Unable to find a free tap(4)");
952 			return -1;
953 		}
954 	} else {
955 		/*
956 		 * User supplied tap(4) device file or unix socket.
957 		 */
958 		if (netif[0] == '/')	/* Absolute path */
959 			strlcpy(tap_dev, netif, sizeof(tap_dev));
960 		else
961 			snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
962 
963 		tap_fd = open(tap_dev, TAPDEV_OFLAGS);
964 
965 		/*
966 		 * If we cannot open normally try to connect to it.
967 		 */
968 		if (tap_fd < 0)
969 			tap_fd = unix_connect(tap_dev);
970 
971 		if (tap_fd < 0) {
972 			warn("Unable to open %s", tap_dev);
973 			return -1;
974 		}
975 	}
976 
977 	/*
978 	 * Check whether the device file is a tap(4)
979 	 */
980 	if (fstat(tap_fd, &st) < 0) {
981 		failed = 1;
982 	} else if (S_ISCHR(st.st_mode)) {
983 		dname = fdevname(tap_fd);
984 		if (dname)
985 			dname = strstr(dname, "tap");
986 		if (dname) {
987 			/*
988 			 * Bring up the corresponding tap(4) interface
989 			 */
990 			*tap_unit = strtol(dname + 3, NULL, 10);
991 			printf("TAP UNIT %d\n", *tap_unit);
992 			if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
993 				failed = 0;
994 			else
995 				failed = 1;
996 		} else {
997 			failed = 1;
998 		}
999 	} else if (S_ISSOCK(st.st_mode)) {
1000 		/*
1001 		 * Special socket connection (typically to vknet).  We
1002 		 * do not have to do anything.
1003 		 */
1004 		failed = 0;
1005 	} else {
1006 		failed = 1;
1007 	}
1008 
1009 	if (failed) {
1010 		warnx("%s is not a tap(4) device or socket", tap_dev);
1011 		close(tap_fd);
1012 		tap_fd = -1;
1013 		*tap_unit = -1;
1014 	}
1015 	return tap_fd;
1016 }
1017 
1018 static int
1019 unix_connect(const char *path)
1020 {
1021 	struct sockaddr_un sunx;
1022 	int len;
1023 	int net_fd;
1024 	int sndbuf = 262144;
1025 	struct stat st;
1026 
1027 	snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1028 	len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1029 	++len;	/* include nul */
1030 	sunx.sun_family = AF_UNIX;
1031 	sunx.sun_len = len;
1032 
1033 	net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1034 	if (net_fd < 0)
1035 		return(-1);
1036 	if (connect(net_fd, (void *)&sunx, len) < 0) {
1037 		close(net_fd);
1038 		return(-1);
1039 	}
1040 	setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1041 	if (fstat(net_fd, &st) == 0)
1042 		printf("Network socket buffer: %d bytes\n", st.st_blksize);
1043 	fcntl(net_fd, F_SETFL, O_NONBLOCK);
1044 	return(net_fd);
1045 }
1046 
1047 #undef TAPDEV_MAJOR
1048 #undef TAPDEV_MINOR
1049 #undef TAPDEV_OFLAGS
1050 
1051 /*
1052  * Following syntax is supported,
1053  * 1) x.x.x.x             tap(4)'s address is x.x.x.x
1054  *
1055  * 2) x.x.x.x/z           tap(4)'s address is x.x.x.x
1056  *                        tap(4)'s netmask len is z
1057  *
1058  * 3) x.x.x.x:y.y.y.y     tap(4)'s address is x.x.x.x
1059  *                        pseudo netif's address is y.y.y.y
1060  *
1061  * 4) x.x.x.x:y.y.y.y/z   tap(4)'s address is x.x.x.x
1062  *                        pseudo netif's address is y.y.y.y
1063  *                        tap(4) and pseudo netif's netmask len are z
1064  *
1065  * 5) bridgeX             tap(4) will be added to bridgeX
1066  *
1067  * 6) bridgeX:y.y.y.y     tap(4) will be added to bridgeX
1068  *                        pseudo netif's address is y.y.y.y
1069  *
1070  * 7) bridgeX:y.y.y.y/z   tap(4) will be added to bridgeX
1071  *                        pseudo netif's address is y.y.y.y
1072  *                        pseudo netif's netmask len is z
1073  */
1074 static
1075 int
1076 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1077 {
1078 	in_addr_t tap_addr, netmask, netif_addr;
1079 	int next_netif_addr;
1080 	char *tok, *masklen_str, *ifbridge;
1081 
1082 	*addr = 0;
1083 	*mask = 0;
1084 
1085 	tok = strtok(NULL, ":/");
1086 	if (tok == NULL) {
1087 		/*
1088 		 * Nothing special, simply use tap(4) as backend
1089 		 */
1090 		return 0;
1091 	}
1092 
1093 	if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1094 		/*
1095 		 * tap(4)'s address is supplied
1096 		 */
1097 		ifbridge = NULL;
1098 
1099 		/*
1100 		 * If there is next token, then it may be pseudo
1101 		 * netif's address or netmask len for tap(4)
1102 		 */
1103 		next_netif_addr = 0;
1104 	} else {
1105 		/*
1106 		 * Not tap(4)'s address, assume it as a bridge(4)
1107 		 * iface name
1108 		 */
1109 		tap_addr = 0;
1110 		ifbridge = tok;
1111 
1112 		/*
1113 		 * If there is next token, then it must be pseudo
1114 		 * netif's address
1115 		 */
1116 		next_netif_addr = 1;
1117 	}
1118 
1119 	netmask = netif_addr = 0;
1120 
1121 	tok = strtok(NULL, ":/");
1122 	if (tok == NULL)
1123 		goto back;
1124 
1125 	if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1126 		if (next_netif_addr) {
1127 			warnx("Invalid pseudo netif address: %s", tok);
1128 			return -1;
1129 		}
1130 		netif_addr = 0;
1131 
1132 		/*
1133 		 * Current token is not address, then it must be netmask len
1134 		 */
1135 		masklen_str = tok;
1136 	} else {
1137 		/*
1138 		 * Current token is pseudo netif address, if there is next token
1139 		 * it must be netmask len
1140 		 */
1141 		masklen_str = strtok(NULL, "/");
1142 	}
1143 
1144 	/* Calculate netmask */
1145 	if (masklen_str != NULL) {
1146 		u_long masklen;
1147 
1148 		masklen = strtoul(masklen_str, NULL, 10);
1149 		if (masklen < 32 && masklen > 0) {
1150 			netmask = htonl(~((1LL << (32 - masklen)) - 1)
1151 					& 0xffffffff);
1152 		} else {
1153 			warnx("Invalid netmask len: %lu", masklen);
1154 			return -1;
1155 		}
1156 	}
1157 
1158 	/* Make sure there is no more token left */
1159 	if (strtok(NULL, ":/") != NULL) {
1160 		warnx("Invalid argument to '-I'");
1161 		return -1;
1162 	}
1163 
1164 back:
1165 	if (tap_unit < 0) {
1166 		/* Do nothing */
1167 	} else if (ifbridge == NULL) {
1168 		/* Set tap(4) address/netmask */
1169 		if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1170 			return -1;
1171 	} else {
1172 		/* Tie tap(4) to bridge(4) */
1173 		if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1174 			return -1;
1175 	}
1176 
1177 	*addr = netif_addr;
1178 	*mask = netmask;
1179 	return 0;
1180 }
1181 
1182 /*
1183  * NetifInfo[] will be filled for pseudo netif initialization.
1184  * NetifNum will be bumped to reflect the number of valid entries
1185  * in NetifInfo[].
1186  */
1187 static
1188 void
1189 init_netif(char *netifExp[], int netifExpNum)
1190 {
1191 	int i, s;
1192 	char *tmp;
1193 
1194 	if (netifExpNum == 0)
1195 		return;
1196 
1197 	s = socket(AF_INET, SOCK_DGRAM, 0);	/* for ioctl(SIOC) */
1198 	if (s < 0)
1199 		return;
1200 
1201 	for (i = 0; i < netifExpNum; ++i) {
1202 		struct vknetif_info *info;
1203 		in_addr_t netif_addr, netif_mask;
1204 		int tap_fd, tap_unit;
1205 		char *netif;
1206 
1207 		/* Extract MAC address if there is one */
1208 		tmp = netifExp[i];
1209 		strsep(&tmp, "=");
1210 
1211 		netif = strtok(netifExp[i], ":");
1212 		if (netif == NULL) {
1213 			warnx("Invalid argument to '-I'");
1214 			continue;
1215 		}
1216 
1217 		/*
1218 		 * Open tap(4) device file and bring up the
1219 		 * corresponding interface
1220 		 */
1221 		tap_fd = netif_open_tap(netif, &tap_unit, s);
1222 		if (tap_fd < 0)
1223 			continue;
1224 
1225 		/*
1226 		 * Initialize tap(4) and get address/netmask
1227 		 * for pseudo netif
1228 		 *
1229 		 * NB: Rest part of netifExp[i] is passed
1230 		 *     to netif_init_tap() implicitly.
1231 		 */
1232 		if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1233 			/*
1234 			 * NB: Closing tap(4) device file will bring
1235 			 *     down the corresponding interface
1236 			 */
1237 			close(tap_fd);
1238 			continue;
1239 		}
1240 
1241 		info = &NetifInfo[NetifNum];
1242 		bzero(info, sizeof(*info));
1243 		info->tap_fd = tap_fd;
1244 		info->tap_unit = tap_unit;
1245 		info->netif_addr = netif_addr;
1246 		info->netif_mask = netif_mask;
1247 		/*
1248 		 * If tmp isn't NULL it means a MAC could have been
1249 		 * specified so attempt to convert it.
1250 		 * Setting enaddr to NULL will tell vke_attach() we
1251 		 * need a pseudo-random MAC address.
1252 		 */
1253 		if (tmp != NULL) {
1254 			if ((info->enaddr = malloc(ETHER_ADDR_LEN)) == NULL)
1255 				warnx("Couldn't allocate memory for the operation");
1256 			else {
1257 				if ((kether_aton(tmp, info->enaddr)) == NULL) {
1258 					free(info->enaddr);
1259 					info->enaddr = NULL;
1260 				}
1261 			}
1262 		}
1263 
1264 		NetifNum++;
1265 		if (NetifNum >= VKNETIF_MAX)	/* XXX will this happen? */
1266 			break;
1267 	}
1268 	close(s);
1269 }
1270 
1271 /*
1272  * Create the pid file and leave it open and locked while the vkernel is
1273  * running.  This allows a script to use /usr/bin/lockf to probe whether
1274  * a vkernel is still running (so as not to accidently kill an unrelated
1275  * process from a stale pid file).
1276  */
1277 static
1278 void
1279 writepid(void)
1280 {
1281 	char buf[32];
1282 	int fd;
1283 
1284 	if (pid_file != NULL) {
1285 		snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1286 		fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1287 		if (fd < 0) {
1288 			if (errno == EWOULDBLOCK) {
1289 				perror("Failed to lock pidfile, "
1290 				       "vkernel already running");
1291 			} else {
1292 				perror("Failed to create pidfile");
1293 			}
1294 			exit(EX_SOFTWARE);
1295 		}
1296 		ftruncate(fd, 0);
1297 		write(fd, buf, strlen(buf));
1298 		/* leave the file open to maintain the lock */
1299 	}
1300 }
1301 
1302 static
1303 void
1304 cleanpid( void )
1305 {
1306 	if (pid_file != NULL) {
1307 		if (unlink(pid_file) < 0)
1308 			perror("Warning: couldn't remove pidfile");
1309 	}
1310 }
1311 
1312 static
1313 void
1314 usage_err(const char *ctl, ...)
1315 {
1316 	va_list va;
1317 
1318 	va_start(va, ctl);
1319 	vfprintf(stderr, ctl, va);
1320 	va_end(va);
1321 	fprintf(stderr, "\n");
1322 	exit(EX_USAGE);
1323 }
1324 
1325 static
1326 void
1327 usage_help(_Bool help)
1328 {
1329 	fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1330 	    "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
1331 	    "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n"
1332 	    "\t[-p file] [-r file]\n", save_av[0]);
1333 
1334 	if (help)
1335 		fprintf(stderr, "\nArguments:\n"
1336 		    "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1337 		    "\t-e\tSpecify an environment to be used by the kernel.\n"
1338 		    "\t-h\tThis list of options.\n"
1339 		    "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1340 		    "\t-I\tCreate a virtual network device.\n"
1341 		    "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1342 		    "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
1343 		    "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n"
1344 		    "\t  \t- numcpus - number of cpus\n"
1345 		    "\t  \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1346 		    "\t  \t  the logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n"
1347 		    "\t  \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1348 		    "\t  \t  the core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n"
1349 		    "\t-p\tSpecify a file in which to store the process ID.\n"
1350 		    "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1351 		    "\t-s\tBoot into single-user mode.\n"
1352 		    "\t-U\tEnable writing to kernel memory and module loading.\n"
1353 		    "\t-v\tTurn on verbose booting.\n");
1354 
1355 	exit(EX_USAGE);
1356 }
1357 
1358 void
1359 cpu_reset(void)
1360 {
1361 	kprintf("cpu reset, rebooting vkernel\n");
1362 	closefrom(3);
1363 	cleanpid();
1364 	execv(save_av[0], save_av);
1365 }
1366 
1367 void
1368 cpu_halt(void)
1369 {
1370 	kprintf("cpu halt, exiting vkernel\n");
1371 	cleanpid();
1372 	exit(EX_OK);
1373 }
1374 
1375 void
1376 setrealcpu(void)
1377 {
1378 	switch(lwp_cpu_lock) {
1379 	case LCL_PER_CPU:
1380 		if (bootverbose)
1381 			kprintf("Locking CPU%d to real cpu %d\n",
1382 				mycpuid, next_cpu);
1383 		usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1384 		next_cpu++;
1385 		if (next_cpu >= real_ncpus)
1386 			next_cpu = 0;
1387 		break;
1388 	case LCL_SINGLE_CPU:
1389 		if (bootverbose)
1390 			kprintf("Locking CPU%d to real cpu %d\n",
1391 				mycpuid, next_cpu);
1392 		usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1393 		break;
1394 	default:
1395 		/* do not map virtual cpus to real cpus */
1396 		break;
1397 	}
1398 }
1399 
1400 /*
1401  * Allocate and free memory for module loading.  The loaded module
1402  * has to be placed somewhere near the current kernel binary load
1403  * point or the relocations will not work.
1404  *
1405  * I'm not sure why this isn't working.
1406  */
1407 int
1408 vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes)
1409 {
1410 	kprintf("module loading for vkernel64's not currently supported\n");
1411 	*basep = 0;
1412 	return ENOMEM;
1413 #if 0
1414 #if 1
1415 	size_t xtra;
1416 	xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK;
1417 	*basep = (vm_offset_t)sbrk(xtra + bytes) + xtra;
1418 	bzero((void *)*basep, bytes);
1419 #else
1420 	*basep = (vm_offset_t)mmap((void *)0x000000000, bytes,
1421 				   PROT_READ|PROT_WRITE|PROT_EXEC,
1422 				   MAP_ANON|MAP_SHARED, -1, 0);
1423 	if ((void *)*basep == MAP_FAILED)
1424 		return ENOMEM;
1425 #endif
1426 	kprintf("basep %p %p %zd\n",
1427 		(void *)vkernel_module_memory_alloc, (void *)*basep, bytes);
1428 	return 0;
1429 #endif
1430 }
1431 
1432 void
1433 vkernel_module_memory_free(vm_offset_t base, size_t bytes)
1434 {
1435 #if 0
1436 #if 0
1437 	munmap((void *)base, bytes);
1438 #endif
1439 #endif
1440 }
1441