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 int tsc_invariant;
112 int tsc_mpsync;
113 int64_t tsc_frequency;
114 int optcpus;		/* number of cpus - see mp_start() */
115 int lwp_cpu_lock;	/* if/how to lock virtual CPUs to real CPUs */
116 int real_ncpus;		/* number of real CPUs */
117 int next_cpu;		/* next real CPU to lock a virtual CPU to */
118 int vkernel_b_arg;	/* -b argument - no of logical CPU bits - only SMP */
119 int vkernel_B_arg;	/* -B argument - no of core bits - only SMP */
120 
121 struct privatespace *CPU_prvspace;
122 
123 static struct trapframe proc0_tf;
124 static void *proc0paddr;
125 
126 static void init_sys_memory(char *imageFile);
127 static void init_kern_memory(void);
128 static void init_globaldata(void);
129 static void init_vkernel(void);
130 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
131 static void init_netif(char *netifExp[], int netifFileNum);
132 static void writepid(void);
133 static void cleanpid(void);
134 static int unix_connect(const char *path);
135 static void usage_err(const char *ctl, ...);
136 static void usage_help(_Bool);
137 static void init_locks(void);
138 
139 static int save_ac;
140 static char **save_av;
141 
142 /*
143  * Kernel startup for virtual kernels - standard main()
144  */
145 int
146 main(int ac, char **av)
147 {
148 	char *memImageFile = NULL;
149 	char *netifFile[VKNETIF_MAX];
150 	char *diskFile[VKDISK_MAX];
151 	char *cdFile[VKDISK_MAX];
152 	char *suffix;
153 	char *endp;
154 	char *tmp;
155 	char *tok;
156 	int netifFileNum = 0;
157 	int diskFileNum = 0;
158 	int cdFileNum = 0;
159 	int bootOnDisk = -1;	/* set below to vcd (0) or vkd (1) */
160 	int c;
161 	int i;
162 	int j;
163 	int n;
164 	int isq;
165 	int pos;
166 	int eflag;
167 	int real_vkernel_enable;
168 	int supports_sse;
169 	size_t vsize;
170 	size_t kenv_size;
171 	size_t kenv_size2;
172 
173 	save_ac = ac;
174 	save_av = av;
175 	eflag = 0;
176 	pos = 0;
177 	kenv_size = 0;
178 
179 	/*
180 	 * Process options
181 	 */
182 	kernel_mem_readonly = 1;
183 	optcpus = 2;
184 	vkernel_b_arg = 0;
185 	vkernel_B_arg = 0;
186 	lwp_cpu_lock = LCL_NONE;
187 
188 	real_vkernel_enable = 0;
189 	vsize = sizeof(real_vkernel_enable);
190 	sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
191 
192 	if (real_vkernel_enable == 0) {
193 		errx(1, "vm.vkernel_enable is 0, must be set "
194 			"to 1 to execute a vkernel!");
195 	}
196 
197 	real_ncpus = 1;
198 	vsize = sizeof(real_ncpus);
199 	sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
200 
201 	if (ac < 2)
202 		usage_help(false);
203 
204 	while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:Ub:B:")) != -1) {
205 		switch(c) {
206 		case 'e':
207 			/*
208 			 * name=value:name=value:name=value...
209 			 * name="value"...
210 			 *
211 			 * Allow values to be quoted but note that shells
212 			 * may remove the quotes, so using this feature
213 			 * to embed colons may require a backslash.
214 			 */
215 			n = strlen(optarg);
216 			isq = 0;
217 
218 			if (eflag == 0) {
219 				kenv_size = n + 2;
220 				kern_envp = malloc(kenv_size);
221 				if (kern_envp == NULL)
222 					errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size);
223 			} else {
224 				kenv_size2 = kenv_size + n + 1;
225 				pos = kenv_size - 1;
226 				if ((tmp = realloc(kern_envp, kenv_size2)) == NULL)
227 					errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2);
228 				kern_envp = tmp;
229 				kenv_size = kenv_size2;
230 			}
231 
232 			for (i = 0, j = pos; i < n; ++i) {
233 				if (optarg[i] == '"')
234 					isq ^= 1;
235 				else if (optarg[i] == '\'')
236 					isq ^= 2;
237 				else if (isq == 0 && optarg[i] == ':')
238 					kern_envp[j++] = 0;
239 				else
240 					kern_envp[j++] = optarg[i];
241 			}
242 			kern_envp[j++] = 0;
243 			kern_envp[j++] = 0;
244 			eflag++;
245 			break;
246 		case 's':
247 			boothowto |= RB_SINGLE;
248 			break;
249 		case 'v':
250 			bootverbose = 1;
251 			break;
252 		case 'i':
253 			memImageFile = optarg;
254 			break;
255 		case 'I':
256 			if (netifFileNum < VKNETIF_MAX)
257 				netifFile[netifFileNum++] = strdup(optarg);
258 			break;
259 		case 'r':
260 			if (bootOnDisk < 0)
261 				bootOnDisk = 1;
262 			if (diskFileNum + cdFileNum < VKDISK_MAX)
263 				diskFile[diskFileNum++] = strdup(optarg);
264 			break;
265 		case 'c':
266 			if (bootOnDisk < 0)
267 				bootOnDisk = 0;
268 			if (diskFileNum + cdFileNum < VKDISK_MAX)
269 				cdFile[cdFileNum++] = strdup(optarg);
270 			break;
271 		case 'm':
272 			Maxmem_bytes = strtoull(optarg, &suffix, 0);
273 			if (suffix) {
274 				switch(*suffix) {
275 				case 'g':
276 				case 'G':
277 					Maxmem_bytes <<= 30;
278 					break;
279 				case 'm':
280 				case 'M':
281 					Maxmem_bytes <<= 20;
282 					break;
283 				case 'k':
284 				case 'K':
285 					Maxmem_bytes <<= 10;
286 					break;
287 				default:
288 					Maxmem_bytes = 0;
289 					usage_err("Bad maxmem option");
290 					/* NOT REACHED */
291 					break;
292 				}
293 			}
294 			break;
295 		case 'l':
296 			next_cpu = -1;
297 			if (strncmp("map", optarg, 3) == 0) {
298 				lwp_cpu_lock = LCL_PER_CPU;
299 				if (optarg[3] == ',') {
300 					next_cpu = strtol(optarg+4, &endp, 0);
301 					if (*endp != '\0')
302 						usage_err("Bad target CPU number at '%s'", endp);
303 				} else {
304 					next_cpu = 0;
305 				}
306 				if (next_cpu < 0 || next_cpu > real_ncpus - 1)
307 					usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
308 			} else if (strncmp("any", optarg, 3) == 0) {
309 				lwp_cpu_lock = LCL_NONE;
310 			} else {
311 				lwp_cpu_lock = LCL_SINGLE_CPU;
312 				next_cpu = strtol(optarg, &endp, 0);
313 				if (*endp != '\0')
314 					usage_err("Bad target CPU number at '%s'", endp);
315 				if (next_cpu < 0 || next_cpu > real_ncpus - 1)
316 					usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
317 			}
318 			break;
319 		case 'n':
320 			/*
321 			 * This value is set up by mp_start(), don't just
322 			 * set ncpus here.
323 			 */
324 			tok = strtok(optarg, ":");
325 			optcpus = strtol(tok, NULL, 0);
326 			if (optcpus < 1 || optcpus > MAXCPU)
327 				usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
328 
329 			/* :lbits argument */
330 			tok = strtok(NULL, ":");
331 			if (tok != NULL) {
332 				vkernel_b_arg = strtol(tok, NULL, 0);
333 
334 				/* :cbits argument */
335 				tok = strtok(NULL, ":");
336 				if (tok != NULL) {
337 					vkernel_B_arg = strtol(tok, NULL, 0);
338 				}
339 
340 			}
341 			break;
342 		case 'p':
343 			pid_file = optarg;
344 			break;
345 		case 'U':
346 			kernel_mem_readonly = 0;
347 			break;
348 		case 'h':
349 			usage_help(true);
350 			break;
351 		default:
352 			usage_help(false);
353 		}
354 	}
355 
356 	writepid();
357 	cpu_disable_intr();
358 	init_sys_memory(memImageFile);
359 	init_kern_memory();
360 	init_globaldata();
361 	init_vkernel();
362 	setrealcpu();
363 	init_kqueue();
364 
365 	vmm_guest = 1;
366 
367 	/*
368 	 * Check TSC
369 	 */
370 	vsize = sizeof(tsc_present);
371 	sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
372 	vsize = sizeof(tsc_invariant);
373 	sysctlbyname("hw.tsc_invariant", &tsc_invariant, &vsize, NULL, 0);
374 	vsize = sizeof(tsc_mpsync);
375 	sysctlbyname("hw.tsc_mpsync", &tsc_mpsync, &vsize, NULL, 0);
376 	vsize = sizeof(tsc_frequency);
377 	sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
378 	if (tsc_present)
379 		cpu_feature |= CPUID_TSC;
380 
381 	/*
382 	 * Check SSE
383 	 */
384 	vsize = sizeof(supports_sse);
385 	supports_sse = 0;
386 	sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
387 	init_fpu(supports_sse);
388 	if (supports_sse)
389 		cpu_feature |= CPUID_SSE | CPUID_FXSR;
390 
391 	/*
392 	 * We boot from the first installed disk.
393 	 */
394 	if (bootOnDisk == 1) {
395 		init_disk(diskFile, diskFileNum, VKD_DISK);
396 		init_disk(cdFile, cdFileNum, VKD_CD);
397 	} else {
398 		init_disk(cdFile, cdFileNum, VKD_CD);
399 		init_disk(diskFile, diskFileNum, VKD_DISK);
400 	}
401 	init_netif(netifFile, netifFileNum);
402 	init_exceptions();
403 	mi_startup();
404 	/* NOT REACHED */
405 	exit(EX_SOFTWARE);
406 }
407 
408 /*
409  * Initialize system memory.  This is the virtual kernel's 'RAM'.
410  */
411 static
412 void
413 init_sys_memory(char *imageFile)
414 {
415 	struct stat st;
416 	int i;
417 	int fd;
418 
419 	/*
420 	 * Figure out the system memory image size.  If an image file was
421 	 * specified and -m was not specified, use the image file's size.
422 	 */
423 	if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
424 		Maxmem_bytes = (vm_paddr_t)st.st_size;
425 	if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
426 	    Maxmem_bytes == 0) {
427 		errx(1, "Cannot create new memory file %s unless "
428 		       "system memory size is specified with -m",
429 		       imageFile);
430 		/* NOT REACHED */
431 	}
432 
433 	/*
434 	 * Maxmem must be known at this time
435 	 */
436 	if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
437 		errx(1, "Bad maxmem specification: 64MB minimum, "
438 		       "multiples of %dMB only",
439 		       SEG_SIZE / 1024 / 1024);
440 		/* NOT REACHED */
441 	}
442 
443 	/*
444 	 * Generate an image file name if necessary, then open/create the
445 	 * file exclusively locked.  Do not allow multiple virtual kernels
446 	 * to use the same image file.
447 	 *
448 	 * Don't iterate through a million files if we do not have write
449 	 * access to the directory, stop if our open() failed on a
450 	 * non-existant file.  Otherwise opens can fail for any number
451 	 */
452 	if (imageFile == NULL) {
453 		for (i = 0; i < 1000000; ++i) {
454 			asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
455 			fd = open(imageFile,
456 				  O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
457 			if (fd < 0 && stat(imageFile, &st) == 0) {
458 				free(imageFile);
459 				continue;
460 			}
461 			break;
462 		}
463 	} else {
464 		fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
465 	}
466 	fprintf(stderr, "Using memory file: %s\n", imageFile);
467 	if (fd < 0 || fstat(fd, &st) < 0) {
468 		err(1, "Unable to open/create %s", imageFile);
469 		/* NOT REACHED */
470 	}
471 
472 	/*
473 	 * Truncate or extend the file as necessary.  Clean out the contents
474 	 * of the file, we want it to be full of holes so we don't waste
475 	 * time reading in data from an old file that we no longer care
476 	 * about.
477 	 */
478 	ftruncate(fd, 0);
479 	ftruncate(fd, Maxmem_bytes);
480 
481 	MemImageFd = fd;
482 	Maxmem = Maxmem_bytes >> PAGE_SHIFT;
483 	physmem = Maxmem;
484 }
485 
486 /*
487  * Initialize kernel memory.  This reserves kernel virtual memory by using
488  * MAP_VPAGETABLE
489  */
490 
491 static
492 void
493 init_kern_memory(void)
494 {
495 	void *base;
496 	void *try;
497 	char dummy;
498 	char *topofstack = &dummy;
499 	int i;
500 	void *firstfree;
501 
502 	/*
503 	 * Memory map our kernel virtual memory space.  Note that the
504 	 * kernel image itself is not made part of this memory for the
505 	 * moment.
506 	 *
507 	 * The memory map must be segment-aligned so we can properly
508 	 * offset KernelPTD.
509 	 *
510 	 * If the system kernel has a different MAXDSIZ, it might not
511 	 * be possible to map kernel memory in its prefered location.
512 	 * Try a number of different locations.
513 	 */
514 	try = (void *)(512UL << 30);
515 	base = NULL;
516 	while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
517 		base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
518 			    MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
519 			    MemImageFd, (off_t)try);
520 		if (base == try)
521 			break;
522 		if (base != MAP_FAILED)
523 			munmap(base, KERNEL_KVA_SIZE);
524 		try = (char *)try + (512UL << 30);
525 	}
526 	if (base != try) {
527 		err(1, "Unable to mmap() kernel virtual memory!");
528 		/* NOT REACHED */
529 	}
530 	madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
531 	KvaStart = (vm_offset_t)base;
532 	KvaSize = KERNEL_KVA_SIZE;
533 	KvaEnd = KvaStart + KvaSize;
534 
535 	/* cannot use kprintf yet */
536 	printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
537 
538 	/* MAP_FILE? */
539 	dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE,
540 				MAP_NOCORE|MAP_NOSYNC|MAP_SHARED,
541 				MemImageFd, 0);
542 	if (dmap_min_address == MAP_FAILED) {
543 		err(1, "Unable to mmap() kernel DMAP region!");
544 		/* NOT REACHED */
545 	}
546 
547 	/*
548 	 * Bootstrap the kernel_pmap
549 	 */
550 	firstfree = NULL;
551 	pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base);
552 
553 	mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
554 		 0 | VPTE_R | VPTE_W | VPTE_V);
555 
556 	/*
557 	 * phys_avail[] represents unallocated physical memory.  MI code
558 	 * will use phys_avail[] to create the vm_page array.
559 	 */
560 	phys_avail[0] = (vm_paddr_t)firstfree;
561 	phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
562 	phys_avail[1] = Maxmem_bytes;
563 
564 #if JGV
565 	/*
566 	 * (virtual_start, virtual_end) represent unallocated kernel virtual
567 	 * memory.  MI code will create kernel_map using these parameters.
568 	 */
569 	virtual_start = KvaStart + (long)firstfree;
570 	virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
571 	virtual_end = KvaStart + KERNEL_KVA_SIZE;
572 #endif
573 
574 	/*
575 	 * pmap_growkernel() will set the correct value.
576 	 */
577 	kernel_vm_end = 0;
578 
579 	/*
580 	 * Allocate space for process 0's UAREA.
581 	 */
582 	proc0paddr = (void *)virtual_start;
583 	for (i = 0; i < UPAGES; ++i) {
584 		pmap_kenter_quick(virtual_start, phys_avail[0]);
585 		virtual_start += PAGE_SIZE;
586 		phys_avail[0] += PAGE_SIZE;
587 	}
588 
589 	/*
590 	 * crashdumpmap
591 	 */
592 	crashdumpmap = virtual_start;
593 	virtual_start += MAXDUMPPGS * PAGE_SIZE;
594 
595 	/*
596 	 * msgbufp maps the system message buffer
597 	 */
598 	assert((MSGBUF_SIZE & PAGE_MASK) == 0);
599 	msgbufp = (void *)virtual_start;
600 	for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
601 		pmap_kenter_quick(virtual_start, phys_avail[0]);
602 		virtual_start += PAGE_SIZE;
603 		phys_avail[0] += PAGE_SIZE;
604 	}
605 	msgbufinit(msgbufp, MSGBUF_SIZE);
606 
607 	/*
608 	 * used by kern_memio for /dev/mem access
609 	 */
610 	ptvmmap = (caddr_t)virtual_start;
611 	virtual_start += PAGE_SIZE;
612 }
613 
614 /*
615  * Map the per-cpu globaldata for cpu #0.  Allocate the space using
616  * virtual_start and phys_avail[0]
617  */
618 static
619 void
620 init_globaldata(void)
621 {
622 	int i;
623 	vm_paddr_t pa;
624 	vm_offset_t va;
625 
626 	/*
627 	 * Reserve enough KVA to cover possible cpus.  This is a considerable
628 	 * amount of KVA since the privatespace structure includes two
629 	 * whole page table mappings.
630 	 */
631 	virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
632 	CPU_prvspace = (void *)virtual_start;
633 	virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
634 
635 	/*
636 	 * Allocate enough physical memory to cover the mdglobaldata
637 	 * portion of the space and the idle stack and map the pages
638 	 * into KVA.  For cpu #0 only.
639 	 */
640 	for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
641 		pa = phys_avail[0];
642 		va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
643 		pmap_kenter_quick(va, pa);
644 		phys_avail[0] += PAGE_SIZE;
645 	}
646 	for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
647 		pa = phys_avail[0];
648 		va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
649 		pmap_kenter_quick(va, pa);
650 		phys_avail[0] += PAGE_SIZE;
651 	}
652 
653 	/*
654 	 * Setup the %gs for cpu #0.  The mycpu macro works after this
655 	 * point.  Note that %fs is used by pthreads.
656 	 */
657 	tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace));
658 }
659 
660 
661 /*
662  * Initialize pool tokens and other necessary locks
663  */
664 static void
665 init_locks(void)
666 {
667 
668         /*
669          * Get the initial mplock with a count of 1 for the BSP.
670          * This uses a LOGICAL cpu ID, ie BSP == 0.
671          */
672         cpu_get_initial_mplock();
673 
674         /* our token pool needs to work early */
675         lwkt_token_pool_init();
676 
677 }
678 
679 
680 /*
681  * Initialize very low level systems including thread0, proc0, etc.
682  */
683 static
684 void
685 init_vkernel(void)
686 {
687 	struct mdglobaldata *gd;
688 
689 	gd = &CPU_prvspace[0].mdglobaldata;
690 	bzero(gd, sizeof(*gd));
691 
692 	gd->mi.gd_curthread = &thread0;
693 	thread0.td_gd = &gd->mi;
694 	ncpus = 1;
695 	ncpus2 = 1;	/* rounded down power of 2 */
696 	ncpus_fit = 1;	/* rounded up power of 2 */
697 	/* ncpus2_mask and ncpus_fit_mask are 0 */
698 	init_param1();
699 	gd->mi.gd_prvspace = &CPU_prvspace[0];
700 	mi_gdinit(&gd->mi, 0);
701 	cpu_gdinit(gd, 0);
702 	mi_proc0init(&gd->mi, proc0paddr);
703 	lwp0.lwp_md.md_regs = &proc0_tf;
704 
705 	init_locks();
706 	cninit();
707 	rand_initialize();
708 #if 0	/* #ifdef DDB */
709 	kdb_init();
710 	if (boothowto & RB_KDB)
711 		Debugger("Boot flags requested debugger");
712 #endif
713 	identcpu();
714 #if 0
715 	initializecpu();	/* Initialize CPU registers */
716 #endif
717 	init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
718 
719 #if 0
720 	/*
721 	 * Map the message buffer
722 	 */
723 	for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
724 		pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
725 	msgbufinit(msgbufp, MSGBUF_SIZE);
726 #endif
727 #if 0
728 	thread0.td_pcb_cr3 ... MMU
729 	lwp0.lwp_md.md_regs = &proc0_tf;
730 #endif
731 }
732 
733 /*
734  * Filesystem image paths for the virtual kernel are optional.
735  * If specified they each should point to a disk image,
736  * the first of which will become the root disk.
737  *
738  * The virtual kernel caches data from our 'disk' just like a normal kernel,
739  * so we do not really want the real kernel to cache the data too.  Use
740  * O_DIRECT to remove the duplication.
741  */
742 static
743 void
744 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
745 {
746 	char *serno;
747 	int i;
748 
749         if (diskFileNum == 0)
750                 return;
751 
752 	for(i=0; i < diskFileNum; i++){
753 		char *fname;
754 		fname = diskExp[i];
755 
756 		if (fname == NULL) {
757                         warnx("Invalid argument to '-r'");
758                         continue;
759                 }
760 		/*
761 		 * Check for a serial number for the virtual disk
762 		 * passed from the command line.
763 		 */
764 		serno = fname;
765 		strsep(&serno, ":");
766 
767 		if (DiskNum < VKDISK_MAX) {
768 			struct stat st;
769 			struct vkdisk_info* info = NULL;
770 			int fd;
771 			size_t l = 0;
772 
773 			if (type == VKD_DISK)
774 			    fd = open(fname, O_RDWR|O_DIRECT, 0644);
775 			else
776 			    fd = open(fname, O_RDONLY|O_DIRECT, 0644);
777 			if (fd < 0 || fstat(fd, &st) < 0) {
778 				err(1, "Unable to open/create %s", fname);
779 				/* NOT REACHED */
780 			}
781 			if (S_ISREG(st.st_mode)) {
782 				if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
783 					errx(1, "Disk image %s is already "
784 						"in use\n", fname);
785 					/* NOT REACHED */
786 				}
787 			}
788 
789 			info = &DiskInfo[DiskNum];
790 			l = strlen(fname);
791 
792 			info->unit = i;
793 			info->fd = fd;
794 			info->type = type;
795 			memcpy(info->fname, fname, l);
796 			info->serno = NULL;
797 			if (serno) {
798 				if ((info->serno = malloc(SERNOLEN)) != NULL)
799 					strlcpy(info->serno, serno, SERNOLEN);
800 				else
801 					warnx("Couldn't allocate memory for the operation");
802 			}
803 
804 			if (DiskNum == 0) {
805 				if (type == VKD_CD) {
806 				    rootdevnames[0] = "cd9660:vcd0a";
807 				} else if (type == VKD_DISK) {
808 				    rootdevnames[0] = "ufs:vkd0s0a";
809 				    rootdevnames[1] = "ufs:vkd0s1a";
810 				}
811 			}
812 
813 			DiskNum++;
814 		} else {
815                         warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
816                         continue;
817 		}
818 	}
819 }
820 
821 static
822 int
823 netif_set_tapflags(int tap_unit, int f, int s)
824 {
825 	struct ifreq ifr;
826 	int flags;
827 
828 	bzero(&ifr, sizeof(ifr));
829 
830 	snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
831 	if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
832 		warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
833 		return -1;
834 	}
835 
836 	/*
837 	 * Adjust if_flags
838 	 *
839 	 * If the flags are already set/cleared, then we return
840 	 * immediately to avoid extra syscalls
841 	 */
842 	flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
843 	if (f < 0) {
844 		/* Turn off flags */
845 		f = -f;
846 		if ((flags & f) == 0)
847 			return 0;
848 		flags &= ~f;
849 	} else {
850 		/* Turn on flags */
851 		if (flags & f)
852 			return 0;
853 		flags |= f;
854 	}
855 
856 	/*
857 	 * Fix up ifreq.ifr_name, since it may be trashed
858 	 * in previous ioctl(SIOCGIFFLAGS)
859 	 */
860 	snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
861 
862 	ifr.ifr_flags = flags & 0xffff;
863 	ifr.ifr_flagshigh = flags >> 16;
864 	if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
865 		warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
866 		return -1;
867 	}
868 	return 0;
869 }
870 
871 static
872 int
873 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
874 {
875 	struct ifaliasreq ifra;
876 	struct sockaddr_in *in;
877 
878 	bzero(&ifra, sizeof(ifra));
879 	snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
880 
881 	/* Setup address */
882 	in = (struct sockaddr_in *)&ifra.ifra_addr;
883 	in->sin_family = AF_INET;
884 	in->sin_len = sizeof(*in);
885 	in->sin_addr.s_addr = addr;
886 
887 	if (mask != 0) {
888 		/* Setup netmask */
889 		in = (struct sockaddr_in *)&ifra.ifra_mask;
890 		in->sin_len = sizeof(*in);
891 		in->sin_addr.s_addr = mask;
892 	}
893 
894 	if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
895 		warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
896 		return -1;
897 	}
898 	return 0;
899 }
900 
901 static
902 int
903 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
904 {
905 	struct ifbreq ifbr;
906 	struct ifdrv ifd;
907 
908 	bzero(&ifbr, sizeof(ifbr));
909 	snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
910 		 "tap%d", tap_unit);
911 
912 	bzero(&ifd, sizeof(ifd));
913 	strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
914 	ifd.ifd_cmd = BRDGADD;
915 	ifd.ifd_len = sizeof(ifbr);
916 	ifd.ifd_data = &ifbr;
917 
918 	if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
919 		/*
920 		 * 'errno == EEXIST' means that the tap(4) is already
921 		 * a member of the bridge(4)
922 		 */
923 		if (errno != EEXIST) {
924 			warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
925 			return -1;
926 		}
927 	}
928 	return 0;
929 }
930 
931 #define TAPDEV_OFLAGS	(O_RDWR | O_NONBLOCK)
932 
933 /*
934  * Locate the first unused tap(4) device file if auto mode is requested,
935  * or open the user supplied device file, and bring up the corresponding
936  * tap(4) interface.
937  *
938  * NOTE: Only tap(4) device file is supported currently
939  */
940 static
941 int
942 netif_open_tap(const char *netif, int *tap_unit, int s)
943 {
944 	char tap_dev[MAXPATHLEN];
945 	int tap_fd, failed;
946 	struct stat st;
947 	char *dname;
948 
949 	*tap_unit = -1;
950 
951 	if (strcmp(netif, "auto") == 0) {
952 		/*
953 		 * Find first unused tap(4) device file
954 		 */
955 		tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
956 		if (tap_fd < 0) {
957 			warnc(errno, "Unable to find a free tap(4)");
958 			return -1;
959 		}
960 	} else {
961 		/*
962 		 * User supplied tap(4) device file or unix socket.
963 		 */
964 		if (netif[0] == '/')	/* Absolute path */
965 			strlcpy(tap_dev, netif, sizeof(tap_dev));
966 		else
967 			snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
968 
969 		tap_fd = open(tap_dev, TAPDEV_OFLAGS);
970 
971 		/*
972 		 * If we cannot open normally try to connect to it.
973 		 */
974 		if (tap_fd < 0)
975 			tap_fd = unix_connect(tap_dev);
976 
977 		if (tap_fd < 0) {
978 			warn("Unable to open %s", tap_dev);
979 			return -1;
980 		}
981 	}
982 
983 	/*
984 	 * Check whether the device file is a tap(4)
985 	 */
986 	if (fstat(tap_fd, &st) < 0) {
987 		failed = 1;
988 	} else if (S_ISCHR(st.st_mode)) {
989 		dname = fdevname(tap_fd);
990 		if (dname)
991 			dname = strstr(dname, "tap");
992 		if (dname) {
993 			/*
994 			 * Bring up the corresponding tap(4) interface
995 			 */
996 			*tap_unit = strtol(dname + 3, NULL, 10);
997 			printf("TAP UNIT %d\n", *tap_unit);
998 			if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
999 				failed = 0;
1000 			else
1001 				failed = 1;
1002 		} else {
1003 			failed = 1;
1004 		}
1005 	} else if (S_ISSOCK(st.st_mode)) {
1006 		/*
1007 		 * Special socket connection (typically to vknet).  We
1008 		 * do not have to do anything.
1009 		 */
1010 		failed = 0;
1011 	} else {
1012 		failed = 1;
1013 	}
1014 
1015 	if (failed) {
1016 		warnx("%s is not a tap(4) device or socket", tap_dev);
1017 		close(tap_fd);
1018 		tap_fd = -1;
1019 		*tap_unit = -1;
1020 	}
1021 	return tap_fd;
1022 }
1023 
1024 static int
1025 unix_connect(const char *path)
1026 {
1027 	struct sockaddr_un sunx;
1028 	int len;
1029 	int net_fd;
1030 	int sndbuf = 262144;
1031 	struct stat st;
1032 
1033 	snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1034 	len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1035 	++len;	/* include nul */
1036 	sunx.sun_family = AF_UNIX;
1037 	sunx.sun_len = len;
1038 
1039 	net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1040 	if (net_fd < 0)
1041 		return(-1);
1042 	if (connect(net_fd, (void *)&sunx, len) < 0) {
1043 		close(net_fd);
1044 		return(-1);
1045 	}
1046 	setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1047 	if (fstat(net_fd, &st) == 0)
1048 		printf("Network socket buffer: %d bytes\n", st.st_blksize);
1049 	fcntl(net_fd, F_SETFL, O_NONBLOCK);
1050 	return(net_fd);
1051 }
1052 
1053 #undef TAPDEV_MAJOR
1054 #undef TAPDEV_MINOR
1055 #undef TAPDEV_OFLAGS
1056 
1057 /*
1058  * Following syntax is supported,
1059  * 1) x.x.x.x             tap(4)'s address is x.x.x.x
1060  *
1061  * 2) x.x.x.x/z           tap(4)'s address is x.x.x.x
1062  *                        tap(4)'s netmask len is z
1063  *
1064  * 3) x.x.x.x:y.y.y.y     tap(4)'s address is x.x.x.x
1065  *                        pseudo netif's address is y.y.y.y
1066  *
1067  * 4) x.x.x.x:y.y.y.y/z   tap(4)'s address is x.x.x.x
1068  *                        pseudo netif's address is y.y.y.y
1069  *                        tap(4) and pseudo netif's netmask len are z
1070  *
1071  * 5) bridgeX             tap(4) will be added to bridgeX
1072  *
1073  * 6) bridgeX:y.y.y.y     tap(4) will be added to bridgeX
1074  *                        pseudo netif's address is y.y.y.y
1075  *
1076  * 7) bridgeX:y.y.y.y/z   tap(4) will be added to bridgeX
1077  *                        pseudo netif's address is y.y.y.y
1078  *                        pseudo netif's netmask len is z
1079  */
1080 static
1081 int
1082 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1083 {
1084 	in_addr_t tap_addr, netmask, netif_addr;
1085 	int next_netif_addr;
1086 	char *tok, *masklen_str, *ifbridge;
1087 
1088 	*addr = 0;
1089 	*mask = 0;
1090 
1091 	tok = strtok(NULL, ":/");
1092 	if (tok == NULL) {
1093 		/*
1094 		 * Nothing special, simply use tap(4) as backend
1095 		 */
1096 		return 0;
1097 	}
1098 
1099 	if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1100 		/*
1101 		 * tap(4)'s address is supplied
1102 		 */
1103 		ifbridge = NULL;
1104 
1105 		/*
1106 		 * If there is next token, then it may be pseudo
1107 		 * netif's address or netmask len for tap(4)
1108 		 */
1109 		next_netif_addr = 0;
1110 	} else {
1111 		/*
1112 		 * Not tap(4)'s address, assume it as a bridge(4)
1113 		 * iface name
1114 		 */
1115 		tap_addr = 0;
1116 		ifbridge = tok;
1117 
1118 		/*
1119 		 * If there is next token, then it must be pseudo
1120 		 * netif's address
1121 		 */
1122 		next_netif_addr = 1;
1123 	}
1124 
1125 	netmask = netif_addr = 0;
1126 
1127 	tok = strtok(NULL, ":/");
1128 	if (tok == NULL)
1129 		goto back;
1130 
1131 	if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1132 		if (next_netif_addr) {
1133 			warnx("Invalid pseudo netif address: %s", tok);
1134 			return -1;
1135 		}
1136 		netif_addr = 0;
1137 
1138 		/*
1139 		 * Current token is not address, then it must be netmask len
1140 		 */
1141 		masklen_str = tok;
1142 	} else {
1143 		/*
1144 		 * Current token is pseudo netif address, if there is next token
1145 		 * it must be netmask len
1146 		 */
1147 		masklen_str = strtok(NULL, "/");
1148 	}
1149 
1150 	/* Calculate netmask */
1151 	if (masklen_str != NULL) {
1152 		u_long masklen;
1153 
1154 		masklen = strtoul(masklen_str, NULL, 10);
1155 		if (masklen < 32 && masklen > 0) {
1156 			netmask = htonl(~((1LL << (32 - masklen)) - 1)
1157 					& 0xffffffff);
1158 		} else {
1159 			warnx("Invalid netmask len: %lu", masklen);
1160 			return -1;
1161 		}
1162 	}
1163 
1164 	/* Make sure there is no more token left */
1165 	if (strtok(NULL, ":/") != NULL) {
1166 		warnx("Invalid argument to '-I'");
1167 		return -1;
1168 	}
1169 
1170 back:
1171 	if (tap_unit < 0) {
1172 		/* Do nothing */
1173 	} else if (ifbridge == NULL) {
1174 		/* Set tap(4) address/netmask */
1175 		if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1176 			return -1;
1177 	} else {
1178 		/* Tie tap(4) to bridge(4) */
1179 		if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1180 			return -1;
1181 	}
1182 
1183 	*addr = netif_addr;
1184 	*mask = netmask;
1185 	return 0;
1186 }
1187 
1188 /*
1189  * NetifInfo[] will be filled for pseudo netif initialization.
1190  * NetifNum will be bumped to reflect the number of valid entries
1191  * in NetifInfo[].
1192  */
1193 static
1194 void
1195 init_netif(char *netifExp[], int netifExpNum)
1196 {
1197 	int i, s;
1198 	char *tmp;
1199 
1200 	if (netifExpNum == 0)
1201 		return;
1202 
1203 	s = socket(AF_INET, SOCK_DGRAM, 0);	/* for ioctl(SIOC) */
1204 	if (s < 0)
1205 		return;
1206 
1207 	for (i = 0; i < netifExpNum; ++i) {
1208 		struct vknetif_info *info;
1209 		in_addr_t netif_addr, netif_mask;
1210 		int tap_fd, tap_unit;
1211 		char *netif;
1212 
1213 		/* Extract MAC address if there is one */
1214 		tmp = netifExp[i];
1215 		strsep(&tmp, "=");
1216 
1217 		netif = strtok(netifExp[i], ":");
1218 		if (netif == NULL) {
1219 			warnx("Invalid argument to '-I'");
1220 			continue;
1221 		}
1222 
1223 		/*
1224 		 * Open tap(4) device file and bring up the
1225 		 * corresponding interface
1226 		 */
1227 		tap_fd = netif_open_tap(netif, &tap_unit, s);
1228 		if (tap_fd < 0)
1229 			continue;
1230 
1231 		/*
1232 		 * Initialize tap(4) and get address/netmask
1233 		 * for pseudo netif
1234 		 *
1235 		 * NB: Rest part of netifExp[i] is passed
1236 		 *     to netif_init_tap() implicitly.
1237 		 */
1238 		if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1239 			/*
1240 			 * NB: Closing tap(4) device file will bring
1241 			 *     down the corresponding interface
1242 			 */
1243 			close(tap_fd);
1244 			continue;
1245 		}
1246 
1247 		info = &NetifInfo[NetifNum];
1248 		bzero(info, sizeof(*info));
1249 		info->tap_fd = tap_fd;
1250 		info->tap_unit = tap_unit;
1251 		info->netif_addr = netif_addr;
1252 		info->netif_mask = netif_mask;
1253 		/*
1254 		 * If tmp isn't NULL it means a MAC could have been
1255 		 * specified so attempt to convert it.
1256 		 * Setting enaddr to NULL will tell vke_attach() we
1257 		 * need a pseudo-random MAC address.
1258 		 */
1259 		if (tmp != NULL) {
1260 			if ((info->enaddr = malloc(ETHER_ADDR_LEN)) == NULL)
1261 				warnx("Couldn't allocate memory for the operation");
1262 			else {
1263 				if ((kether_aton(tmp, info->enaddr)) == NULL) {
1264 					free(info->enaddr);
1265 					info->enaddr = NULL;
1266 				}
1267 			}
1268 		}
1269 
1270 		NetifNum++;
1271 		if (NetifNum >= VKNETIF_MAX)	/* XXX will this happen? */
1272 			break;
1273 	}
1274 	close(s);
1275 }
1276 
1277 /*
1278  * Create the pid file and leave it open and locked while the vkernel is
1279  * running.  This allows a script to use /usr/bin/lockf to probe whether
1280  * a vkernel is still running (so as not to accidently kill an unrelated
1281  * process from a stale pid file).
1282  */
1283 static
1284 void
1285 writepid(void)
1286 {
1287 	char buf[32];
1288 	int fd;
1289 
1290 	if (pid_file != NULL) {
1291 		snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1292 		fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1293 		if (fd < 0) {
1294 			if (errno == EWOULDBLOCK) {
1295 				perror("Failed to lock pidfile, "
1296 				       "vkernel already running");
1297 			} else {
1298 				perror("Failed to create pidfile");
1299 			}
1300 			exit(EX_SOFTWARE);
1301 		}
1302 		ftruncate(fd, 0);
1303 		write(fd, buf, strlen(buf));
1304 		/* leave the file open to maintain the lock */
1305 	}
1306 }
1307 
1308 static
1309 void
1310 cleanpid( void )
1311 {
1312 	if (pid_file != NULL) {
1313 		if (unlink(pid_file) < 0)
1314 			perror("Warning: couldn't remove pidfile");
1315 	}
1316 }
1317 
1318 static
1319 void
1320 usage_err(const char *ctl, ...)
1321 {
1322 	va_list va;
1323 
1324 	va_start(va, ctl);
1325 	vfprintf(stderr, ctl, va);
1326 	va_end(va);
1327 	fprintf(stderr, "\n");
1328 	exit(EX_USAGE);
1329 }
1330 
1331 static
1332 void
1333 usage_help(_Bool help)
1334 {
1335 	fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1336 	    "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
1337 	    "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n"
1338 	    "\t[-p file] [-r file]\n", save_av[0]);
1339 
1340 	if (help)
1341 		fprintf(stderr, "\nArguments:\n"
1342 		    "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1343 		    "\t-e\tSpecify an environment to be used by the kernel.\n"
1344 		    "\t-h\tThis list of options.\n"
1345 		    "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1346 		    "\t-I\tCreate a virtual network device.\n"
1347 		    "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1348 		    "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
1349 		    "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n"
1350 		    "\t  \t- numcpus - number of cpus\n"
1351 		    "\t  \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1352 		    "\t  \t  the logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n"
1353 		    "\t  \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1354 		    "\t  \t  the core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n"
1355 		    "\t-p\tSpecify a file in which to store the process ID.\n"
1356 		    "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1357 		    "\t-s\tBoot into single-user mode.\n"
1358 		    "\t-U\tEnable writing to kernel memory and module loading.\n"
1359 		    "\t-v\tTurn on verbose booting.\n");
1360 
1361 	exit(EX_USAGE);
1362 }
1363 
1364 void
1365 cpu_reset(void)
1366 {
1367 	kprintf("cpu reset, rebooting vkernel\n");
1368 	closefrom(3);
1369 	cleanpid();
1370 	execv(save_av[0], save_av);
1371 }
1372 
1373 void
1374 cpu_halt(void)
1375 {
1376 	kprintf("cpu halt, exiting vkernel\n");
1377 	cleanpid();
1378 	exit(EX_OK);
1379 }
1380 
1381 void
1382 setrealcpu(void)
1383 {
1384 	switch(lwp_cpu_lock) {
1385 	case LCL_PER_CPU:
1386 		if (bootverbose)
1387 			kprintf("Locking CPU%d to real cpu %d\n",
1388 				mycpuid, next_cpu);
1389 		usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1390 		next_cpu++;
1391 		if (next_cpu >= real_ncpus)
1392 			next_cpu = 0;
1393 		break;
1394 	case LCL_SINGLE_CPU:
1395 		if (bootverbose)
1396 			kprintf("Locking CPU%d to real cpu %d\n",
1397 				mycpuid, next_cpu);
1398 		usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1399 		break;
1400 	default:
1401 		/* do not map virtual cpus to real cpus */
1402 		break;
1403 	}
1404 }
1405 
1406 /*
1407  * Allocate and free memory for module loading.  The loaded module
1408  * has to be placed somewhere near the current kernel binary load
1409  * point or the relocations will not work.
1410  *
1411  * I'm not sure why this isn't working.
1412  */
1413 int
1414 vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes)
1415 {
1416 	kprintf("module loading for vkernel64's not currently supported\n");
1417 	*basep = 0;
1418 	return ENOMEM;
1419 #if 0
1420 #if 1
1421 	size_t xtra;
1422 	xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK;
1423 	*basep = (vm_offset_t)sbrk(xtra + bytes) + xtra;
1424 	bzero((void *)*basep, bytes);
1425 #else
1426 	*basep = (vm_offset_t)mmap((void *)0x000000000, bytes,
1427 				   PROT_READ|PROT_WRITE|PROT_EXEC,
1428 				   MAP_ANON|MAP_SHARED, -1, 0);
1429 	if ((void *)*basep == MAP_FAILED)
1430 		return ENOMEM;
1431 #endif
1432 	kprintf("basep %p %p %zd\n",
1433 		(void *)vkernel_module_memory_alloc, (void *)*basep, bytes);
1434 	return 0;
1435 #endif
1436 }
1437 
1438 void
1439 vkernel_module_memory_free(vm_offset_t base, size_t bytes)
1440 {
1441 #if 0
1442 #if 0
1443 	munmap((void *)base, bytes);
1444 #endif
1445 #endif
1446 }
1447