vkernel - enhance the pidfile option and fix memimg file scanning
[dragonfly.git] / sys / platform / vkernel / platform / init.c
CommitLineData
9c059ae3
MD
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 *
2c4a81d7 34 * $DragonFly: src/sys/platform/vkernel/platform/init.c,v 1.56 2008/05/27 07:48:00 dillon Exp $
9c059ae3
MD
35 */
36
44a09dd6
MD
37#include <sys/types.h>
38#include <sys/systm.h>
e4a473f1 39#include <sys/kernel.h>
44a09dd6
MD
40#include <sys/stat.h>
41#include <sys/mman.h>
e4a473f1
MD
42#include <sys/cons.h>
43#include <sys/random.h>
44#include <sys/vkernel.h>
8a9d8995 45#include <sys/tls.h>
884d6eea 46#include <sys/reboot.h>
e4a473f1 47#include <sys/proc.h>
6f7b98e0 48#include <sys/msgbuf.h>
6efe4829 49#include <sys/vmspace.h>
d53b52ae
SZ
50#include <sys/socket.h>
51#include <sys/sockio.h>
702acf06 52#include <sys/sysctl.h>
91be174d 53#include <sys/un.h>
44a09dd6 54#include <vm/vm_page.h>
e130cdcb 55#include <sys/mplock2.h>
44a09dd6 56
702acf06 57#include <machine/cpu.h>
8a9d8995 58#include <machine/globaldata.h>
e4a473f1
MD
59#include <machine/tls.h>
60#include <machine/md_var.h>
61#include <machine/vmparam.h>
870b0161 62#include <cpu/specialreg.h>
8a9d8995 63
d53b52ae
SZ
64#include <net/if.h>
65#include <net/if_arp.h>
66#include <net/ethernet.h>
67#include <net/bridge/if_bridgevar.h>
68#include <netinet/in.h>
69#include <arpa/inet.h>
70
9c059ae3 71#include <stdio.h>
44a09dd6 72#include <stdlib.h>
057f0718 73#include <stdarg.h>
278c4e67 74#include <stdbool.h>
44a09dd6
MD
75#include <unistd.h>
76#include <fcntl.h>
77#include <string.h>
78#include <err.h>
79#include <errno.h>
8a9d8995 80#include <assert.h>
278c4e67 81#include <sysexits.h>
9c059ae3 82
e4a473f1 83vm_paddr_t phys_avail[16];
9c059ae3 84vm_paddr_t Maxmem;
71152ac6 85vm_paddr_t Maxmem_bytes;
39d69dae 86long physmem;
9c059ae3 87int MemImageFd = -1;
ee63ee00
SW
88struct vkdisk_info DiskInfo[VKDISK_MAX];
89int DiskNum;
d53b52ae
SZ
90struct vknetif_info NetifInfo[VKNETIF_MAX];
91int NetifNum;
b02b4170 92char *pid_file;
e4a473f1
MD
93vm_offset_t KvaStart;
94vm_offset_t KvaEnd;
9c059ae3 95vm_offset_t KvaSize;
e4a473f1
MD
96vm_offset_t virtual_start;
97vm_offset_t virtual_end;
791c6551
MD
98vm_offset_t virtual2_start;
99vm_offset_t virtual2_end;
e4a473f1
MD
100vm_offset_t kernel_vm_end;
101vm_offset_t crashdumpmap;
6f7b98e0
MD
102vm_offset_t clean_sva;
103vm_offset_t clean_eva;
104struct msgbuf *msgbufp;
105caddr_t ptvmmap;
e4a473f1 106vpte_t *KernelPTD;
71152ac6 107vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
6f7b98e0 108u_int cpu_feature; /* XXX */
870b0161
MD
109int tsc_present;
110int64_t tsc_frequency;
c5b0b0ba 111int optcpus; /* number of cpus - see mp_start() */
702acf06
JT
112int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
113int real_ncpus; /* number of real CPUs */
114int next_cpu; /* next real CPU to lock a virtual CPU to */
9c059ae3 115
1a8fe9c5
MD
116int via_feature_xcrypt = 0; /* XXX */
117int via_feature_rng = 0; /* XXX */
118
8a9d8995
MD
119struct privatespace *CPU_prvspace;
120
252b2227 121static struct trapframe proc0_tf;
e4a473f1
MD
122static void *proc0paddr;
123
44a09dd6
MD
124static void init_sys_memory(char *imageFile);
125static void init_kern_memory(void);
8a9d8995 126static void init_globaldata(void);
e4a473f1 127static void init_vkernel(void);
a72d8a9f 128static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
ee63ee00 129static void init_netif(char *netifExp[], int netifFileNum);
278c4e67
SG
130static void writepid(void);
131static void cleanpid(void);
91be174d 132static int unix_connect(const char *path);
278c4e67
SG
133static void usage_err(const char *ctl, ...);
134static void usage_help(_Bool);
9c059ae3 135
40d0276c
MD
136static int save_ac;
137static char **save_av;
138
9c059ae3
MD
139/*
140 * Kernel startup for virtual kernels - standard main()
141 */
142int
143main(int ac, char **av)
144{
44a09dd6 145 char *memImageFile = NULL;
d53b52ae 146 char *netifFile[VKNETIF_MAX];
ee63ee00 147 char *diskFile[VKDISK_MAX];
a72d8a9f 148 char *cdFile[VKDISK_MAX];
44a09dd6 149 char *suffix;
702acf06 150 char *endp;
d53b52ae 151 int netifFileNum = 0;
ee63ee00 152 int diskFileNum = 0;
a72d8a9f 153 int cdFileNum = 0;
41556a7a 154 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
44a09dd6 155 int c;
aaf8b91f 156 int i;
a6d77ab5 157 int j;
aaf8b91f 158 int n;
a6d77ab5 159 int isq;
dc7503af 160 int real_vkernel_enable;
aecc8383 161 int supports_sse;
870b0161 162 size_t vsize;
b02b4170 163
40d0276c
MD
164 save_ac = ac;
165 save_av = av;
166
9c059ae3
MD
167 /*
168 * Process options
169 */
7e42c007 170 kernel_mem_readonly = 1;
c5b0b0ba
MD
171#ifdef SMP
172 optcpus = 2;
173#endif
702acf06 174 lwp_cpu_lock = LCL_NONE;
7e42c007 175
27c3aec1 176 real_vkernel_enable = 0;
870b0161
MD
177 vsize = sizeof(real_vkernel_enable);
178 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
dc7503af 179
27c3aec1
MD
180 if (real_vkernel_enable == 0) {
181 errx(1, "vm.vkernel_enable is 0, must be set "
182 "to 1 to execute a vkernel!");
183 }
dc7503af 184
27c3aec1 185 real_ncpus = 1;
870b0161
MD
186 vsize = sizeof(real_ncpus);
187 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
702acf06 188
278c4e67
SG
189 if (ac < 2)
190 usage_help(false);
191
192 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:U")) != -1) {
9c059ae3 193 switch(c) {
aaf8b91f
MD
194 case 'e':
195 /*
196 * name=value:name=value:name=value...
a6d77ab5
MD
197 * name="value"...
198 *
199 * Allow values to be quoted but note that shells
200 * may remove the quotes, so using this feature
201 * to embed colons may require a backslash.
aaf8b91f
MD
202 */
203 n = strlen(optarg);
a6d77ab5 204 isq = 0;
aaf8b91f 205 kern_envp = malloc(n + 2);
a6d77ab5
MD
206 for (i = j = 0; i < n; ++i) {
207 if (optarg[i] == '"')
208 isq ^= 1;
209 else if (optarg[i] == '\'')
210 isq ^= 2;
211 else if (isq == 0 && optarg[i] == ':')
212 kern_envp[j++] = 0;
aaf8b91f 213 else
a6d77ab5 214 kern_envp[j++] = optarg[i];
aaf8b91f 215 }
a6d77ab5
MD
216 kern_envp[j++] = 0;
217 kern_envp[j++] = 0;
aaf8b91f 218 break;
884d6eea
MD
219 case 's':
220 boothowto |= RB_SINGLE;
221 break;
9c059ae3
MD
222 case 'v':
223 bootverbose = 1;
224 break;
225 case 'i':
226 memImageFile = optarg;
227 break;
d869938c 228 case 'I':
d53b52ae 229 if (netifFileNum < VKNETIF_MAX)
40d0276c 230 netifFile[netifFileNum++] = strdup(optarg);
d869938c 231 break;
9c059ae3 232 case 'r':
41556a7a
MD
233 if (bootOnDisk < 0)
234 bootOnDisk = 1;
a72d8a9f 235 if (diskFileNum + cdFileNum < VKDISK_MAX)
40d0276c 236 diskFile[diskFileNum++] = strdup(optarg);
9c059ae3 237 break;
a72d8a9f 238 case 'c':
41556a7a
MD
239 if (bootOnDisk < 0)
240 bootOnDisk = 0;
a72d8a9f 241 if (diskFileNum + cdFileNum < VKDISK_MAX)
40d0276c 242 cdFile[cdFileNum++] = strdup(optarg);
a72d8a9f 243 break;
9c059ae3 244 case 'm':
71152ac6 245 Maxmem_bytes = strtoull(optarg, &suffix, 0);
9c059ae3
MD
246 if (suffix) {
247 switch(*suffix) {
248 case 'g':
249 case 'G':
71152ac6 250 Maxmem_bytes <<= 30;
9c059ae3
MD
251 break;
252 case 'm':
253 case 'M':
71152ac6 254 Maxmem_bytes <<= 20;
9c059ae3
MD
255 break;
256 case 'k':
257 case 'K':
71152ac6 258 Maxmem_bytes <<= 10;
9c059ae3
MD
259 break;
260 default:
71152ac6 261 Maxmem_bytes = 0;
278c4e67 262 usage_err("Bad maxmem option");
9c059ae3
MD
263 /* NOT REACHED */
264 break;
265 }
266 }
267 break;
702acf06
JT
268 case 'l':
269 next_cpu = -1;
270 if (strncmp("map", optarg, 3) == 0) {
271 lwp_cpu_lock = LCL_PER_CPU;
92d2f80a 272 if (optarg[3] == ',') {
702acf06 273 next_cpu = strtol(optarg+4, &endp, 0);
92d2f80a 274 if (*endp != '\0')
278c4e67 275 usage_err("Bad target CPU number at '%s'", endp);
92d2f80a
MD
276 } else {
277 next_cpu = 0;
278 }
702acf06 279 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 280 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
702acf06
JT
281 } else if (strncmp("any", optarg, 3) == 0) {
282 lwp_cpu_lock = LCL_NONE;
283 } else {
284 lwp_cpu_lock = LCL_SINGLE_CPU;
285 next_cpu = strtol(optarg, &endp, 0);
286 if (*endp != '\0')
278c4e67 287 usage_err("Bad target CPU number at '%s'", endp);
702acf06 288 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 289 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
702acf06
JT
290 }
291 break;
c5b0b0ba
MD
292 case 'n':
293 /*
294 * This value is set up by mp_start(), don't just
295 * set ncpus here.
296 */
297#ifdef SMP
298 optcpus = strtol(optarg, NULL, 0);
057f0718 299 if (optcpus < 1 || optcpus > MAXCPU)
278c4e67 300 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
c5b0b0ba
MD
301#else
302 if (strtol(optarg, NULL, 0) != 1) {
278c4e67 303 usage_err("You built a UP vkernel, only 1 cpu!");
c5b0b0ba
MD
304 }
305#endif
306
151f45bf 307 break;
b02b4170
MD
308 case 'p':
309 pid_file = optarg;
310 break;
7e42c007
MD
311 case 'U':
312 kernel_mem_readonly = 0;
313 break;
278c4e67
SG
314 case 'h':
315 usage_help(true);
316 break;
278c4e67
SG
317 default:
318 usage_help(false);
9c059ae3
MD
319 }
320 }
321
b02b4170 322 writepid();
b1b21fa6 323 cpu_disable_intr();
9c059ae3
MD
324 init_sys_memory(memImageFile);
325 init_kern_memory();
8a9d8995 326 init_globaldata();
e4a473f1 327 init_vkernel();
702acf06 328 setrealcpu();
f991237f 329 init_kqueue();
41556a7a 330
870b0161
MD
331 /*
332 * Check TSC
333 */
334 vsize = sizeof(tsc_present);
335 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
336 vsize = sizeof(tsc_frequency);
337 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
338 if (tsc_present)
339 cpu_feature |= CPUID_TSC;
340
341 /*
342 * Check SSE
343 */
344 vsize = sizeof(supports_sse);
27c3aec1 345 supports_sse = 0;
870b0161 346 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
aecc8383 347 init_fpu(supports_sse);
870b0161
MD
348 if (supports_sse)
349 cpu_feature |= CPUID_SSE | CPUID_FXSR;
aecc8383 350
41556a7a
MD
351 /*
352 * We boot from the first installed disk.
353 */
354 if (bootOnDisk == 1) {
355 init_disk(diskFile, diskFileNum, VKD_DISK);
356 init_disk(cdFile, cdFileNum, VKD_CD);
357 } else {
358 init_disk(cdFile, cdFileNum, VKD_CD);
359 init_disk(diskFile, diskFileNum, VKD_DISK);
360 }
d53b52ae 361 init_netif(netifFile, netifFileNum);
6092278a 362 init_exceptions();
9c059ae3
MD
363 mi_startup();
364 /* NOT REACHED */
278c4e67 365 exit(EX_SOFTWARE);
9c059ae3
MD
366}
367
368/*
369 * Initialize system memory. This is the virtual kernel's 'RAM'.
370 */
371static
372void
44a09dd6 373init_sys_memory(char *imageFile)
9c059ae3 374{
44a09dd6 375 struct stat st;
3748859a 376 int i;
9c059ae3
MD
377 int fd;
378
379 /*
380 * Figure out the system memory image size. If an image file was
381 * specified and -m was not specified, use the image file's size.
382 */
71152ac6
MD
383 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
384 Maxmem_bytes = (vm_paddr_t)st.st_size;
385 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
386 Maxmem_bytes == 0) {
376264e3 387 errx(1, "Cannot create new memory file %s unless "
9c059ae3
MD
388 "system memory size is specified with -m",
389 imageFile);
390 /* NOT REACHED */
391 }
392
393 /*
394 * Maxmem must be known at this time
395 */
71152ac6 396 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
376264e3 397 errx(1, "Bad maxmem specification: 32MB minimum, "
9c059ae3
MD
398 "multiples of %dMB only",
399 SEG_SIZE / 1024 / 1024);
400 /* NOT REACHED */
401 }
402
403 /*
404 * Generate an image file name if necessary, then open/create the
405 * file exclusively locked. Do not allow multiple virtual kernels
406 * to use the same image file.
bc3cc25e
MD
407 *
408 * Don't iterate through a million files if we do not have write
409 * access to the directory, stop if our open() failed on a
410 * non-existant file. Otherwise opens can fail for any number
411 * of reasons (lock failed, file not owned or writable by us, etc).
9c059ae3 412 */
3748859a
MD
413 if (imageFile == NULL) {
414 for (i = 0; i < 1000000; ++i) {
0171c06b 415 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
3748859a
MD
416 fd = open(imageFile,
417 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
bc3cc25e 418 if (fd < 0 && stat(imageFile, &st) == 0) {
3748859a
MD
419 free(imageFile);
420 continue;
421 }
422 break;
423 }
424 } else {
425 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
426 }
0171c06b 427 printf("Using memory file: %s\n", imageFile);
9c059ae3 428 if (fd < 0 || fstat(fd, &st) < 0) {
5de37c8b 429 err(1, "Unable to open/create %s", imageFile);
9c059ae3
MD
430 /* NOT REACHED */
431 }
432
433 /*
b72e51d1
MD
434 * Truncate or extend the file as necessary. Clean out the contents
435 * of the file, we want it to be full of holes so we don't waste
436 * time reading in data from an old file that we no longer care
437 * about.
9c059ae3 438 */
b72e51d1
MD
439 ftruncate(fd, 0);
440 ftruncate(fd, Maxmem_bytes);
9c059ae3 441
9c059ae3 442 MemImageFd = fd;
71152ac6 443 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
9f90716f 444 physmem = Maxmem;
9c059ae3
MD
445}
446
447/*
448 * Initialize kernel memory. This reserves kernel virtual memory by using
449 * MAP_VPAGETABLE
450 */
0c4c980f 451
9c059ae3
MD
452static
453void
454init_kern_memory(void)
455{
456 void *base;
0c4c980f 457 void *try;
3748859a 458 char *zero;
0c4c980f
MD
459 char dummy;
460 char *topofstack = &dummy;
e4a473f1
MD
461 vpte_t pte;
462 int i;
9c059ae3
MD
463
464 /*
465 * Memory map our kernel virtual memory space. Note that the
466 * kernel image itself is not made part of this memory for the
467 * moment.
71152ac6
MD
468 *
469 * The memory map must be segment-aligned so we can properly
470 * offset KernelPTD.
0c4c980f
MD
471 *
472 * If the system kernel has a different MAXDSIZ, it might not
473 * be possible to map kernel memory in its prefered location.
474 * Try a number of different locations.
9c059ae3 475 */
0c4c980f
MD
476 try = (void *)0x40000000;
477 base = NULL;
478 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
479 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
480 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
481 MemImageFd, 0);
482 if (base == try)
483 break;
484 if (base != MAP_FAILED)
485 munmap(base, KERNEL_KVA_SIZE);
486 try = (char *)try + 0x10000000;
487 }
488 if (base != try) {
9c059ae3
MD
489 err(1, "Unable to mmap() kernel virtual memory!");
490 /* NOT REACHED */
491 }
a3c35df6 492 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
e4a473f1 493 KvaStart = (vm_offset_t)base;
9c059ae3 494 KvaSize = KERNEL_KVA_SIZE;
e4a473f1 495 KvaEnd = KvaStart + KvaSize;
d4f5f27c
MD
496
497 /* cannot use kprintf yet */
0c4c980f 498 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
e4a473f1
MD
499
500 /*
501 * Create a top-level page table self-mapping itself.
502 *
503 * Initialize the page directory at physical page index 0 to point
504 * to an array of page table pages starting at physical page index 1
505 */
506 lseek(MemImageFd, 0L, 0);
507 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
508 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
509 write(MemImageFd, &pte, sizeof(pte));
510 }
511
512 /*
513 * Initialize the PTEs in the page table pages required to map the
514 * page table itself. This includes mapping the page directory page
515 * at the base so we go one more loop then normal.
516 */
517 lseek(MemImageFd, PAGE_SIZE, 0);
71152ac6 518 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
e4a473f1
MD
519 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
520 write(MemImageFd, &pte, sizeof(pte));
521 }
522
3748859a
MD
523 /*
524 * Initialize remaining PTEs to 0. We may be reusing a memory image
525 * file. This is approximately a megabyte.
526 */
527 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
528 zero = malloc(PAGE_SIZE);
b87d6bd2 529 bzero(zero, PAGE_SIZE);
3748859a
MD
530 while (i) {
531 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
532 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
533 }
534 free(zero);
535
e4a473f1
MD
536 /*
537 * Enable the page table and calculate pointers to our self-map
538 * for easy kernel page table manipulation.
71152ac6
MD
539 *
540 * KernelPTA must be offset so we can do direct VA translations
e4a473f1 541 */
71152ac6
MD
542 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
543 0 | VPTE_R | VPTE_W | VPTE_V);
e4a473f1
MD
544 KernelPTD = (vpte_t *)base; /* pg directory */
545 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
71152ac6 546 KernelPTA -= KvaStart >> PAGE_SHIFT;
e4a473f1
MD
547
548 /*
549 * phys_avail[] represents unallocated physical memory. MI code
550 * will use phys_avail[] to create the vm_page array.
551 */
552 phys_avail[0] = PAGE_SIZE +
553 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
554 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
71152ac6 555 phys_avail[1] = Maxmem_bytes;
e4a473f1
MD
556
557 /*
558 * (virtual_start, virtual_end) represent unallocated kernel virtual
559 * memory. MI code will create kernel_map using these parameters.
560 */
561 virtual_start = KvaStart + PAGE_SIZE +
562 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
563 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
564 virtual_end = KvaStart + KERNEL_KVA_SIZE;
565
566 /*
00835518
MD
567 * kernel_vm_end could be set to virtual_end but we want some
568 * indication of how much of the kernel_map we've used, so
569 * set it low and let pmap_growkernel increase it even though we
570 * don't need to create any new page table pages.
e4a473f1 571 */
00835518 572 kernel_vm_end = virtual_start;
e4a473f1
MD
573
574 /*
575 * Allocate space for process 0's UAREA.
576 */
577 proc0paddr = (void *)virtual_start;
578 for (i = 0; i < UPAGES; ++i) {
579 pmap_kenter_quick(virtual_start, phys_avail[0]);
580 virtual_start += PAGE_SIZE;
581 phys_avail[0] += PAGE_SIZE;
582 }
9c059ae3
MD
583
584 /*
e4a473f1 585 * crashdumpmap
9c059ae3 586 */
e4a473f1
MD
587 crashdumpmap = virtual_start;
588 virtual_start += MAXDUMPPGS * PAGE_SIZE;
589
6f7b98e0
MD
590 /*
591 * msgbufp maps the system message buffer
592 */
593 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
594 msgbufp = (void *)virtual_start;
595 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
596 pmap_kenter_quick(virtual_start, phys_avail[0]);
597 virtual_start += PAGE_SIZE;
598 phys_avail[0] += PAGE_SIZE;
599 }
600 msgbufinit(msgbufp, MSGBUF_SIZE);
601
602 /*
603 * used by kern_memio for /dev/mem access
604 */
605 ptvmmap = (caddr_t)virtual_start;
606 virtual_start += PAGE_SIZE;
607
e4a473f1
MD
608 /*
609 * Bootstrap the kernel_pmap
610 */
611 pmap_bootstrap();
9c059ae3
MD
612}
613
8a9d8995 614/*
e4a473f1
MD
615 * Map the per-cpu globaldata for cpu #0. Allocate the space using
616 * virtual_start and phys_avail[0]
8a9d8995
MD
617 */
618static
619void
620init_globaldata(void)
621{
e4a473f1
MD
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 /*
27c3aec1
MD
654 * Setup the %fs for cpu #0. The mycpu macro works after this
655 * point. Note that %gs is used by pthreads.
e4a473f1 656 */
9ad680a3 657 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
e4a473f1
MD
658}
659
660/*
661 * Initialize very low level systems including thread0, proc0, etc.
662 */
663static
664void
665init_vkernel(void)
666{
667 struct mdglobaldata *gd;
668
669 gd = &CPU_prvspace[0].mdglobaldata;
670 bzero(gd, sizeof(*gd));
671
672 gd->mi.gd_curthread = &thread0;
71152ac6 673 thread0.td_gd = &gd->mi;
e4a473f1 674 ncpus = 1;
b45759e1
MD
675 ncpus2 = 1; /* rounded down power of 2 */
676 ncpus_fit = 1; /* rounded up power of 2 */
677 /* ncpus2_mask and ncpus_fit_mask are 0 */
e4a473f1
MD
678 init_param1();
679 gd->mi.gd_prvspace = &CPU_prvspace[0];
680 mi_gdinit(&gd->mi, 0);
681 cpu_gdinit(gd, 0);
682 mi_proc0init(&gd->mi, proc0paddr);
08f2f1bb 683 lwp0.lwp_md.md_regs = &proc0_tf;
252b2227 684
e4a473f1 685 /*init_locks();*/
e130cdcb
MD
686#ifdef SMP
687 /*
688 * Get the initial mplock with a count of 1 for the BSP.
689 * This uses a LOGICAL cpu ID, ie BSP == 0.
690 */
691 cpu_get_initial_mplock();
692#endif
e4a473f1
MD
693 cninit();
694 rand_initialize();
695#if 0 /* #ifdef DDB */
696 kdb_init();
697 if (boothowto & RB_KDB)
698 Debugger("Boot flags requested debugger");
699#endif
514621ec 700 identcpu();
e4a473f1
MD
701#if 0
702 initializecpu(); /* Initialize CPU registers */
703#endif
704 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
705
706#if 0
707 /*
708 * Map the message buffer
709 */
710 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
711 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
712 msgbufinit(msgbufp, MSGBUF_SIZE);
713#endif
714#if 0
715 thread0.td_pcb_cr3 ... MMU
08f2f1bb 716 lwp0.lwp_md.md_regs = &proc0_tf;
e4a473f1 717#endif
8a9d8995
MD
718}
719
9c059ae3 720/*
ee63ee00
SW
721 * Filesystem image paths for the virtual kernel are optional.
722 * If specified they each should point to a disk image,
723 * the first of which will become the root disk.
5b85cb7d
MD
724 *
725 * The virtual kernel caches data from our 'disk' just like a normal kernel,
726 * so we do not really want the real kernel to cache the data too. Use
727 * O_DIRECT to remove the duplication.
9c059ae3
MD
728 */
729static
730void
a72d8a9f 731init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
44a09dd6 732{
ee63ee00
SW
733 int i;
734
735 if (diskFileNum == 0)
736 return;
737
738 for(i=0; i < diskFileNum; i++){
739 char *fname;
740 fname = diskExp[i];
741
742 if (fname == NULL) {
743 warnx("Invalid argument to '-r'");
744 continue;
745 }
746
747 if (DiskNum < VKDISK_MAX) {
748 struct stat st;
749 struct vkdisk_info* info = NULL;
750 int fd;
751 size_t l = 0;
752
a72d8a9f 753 if (type == VKD_DISK)
6882b7a5 754 fd = open(fname, O_RDWR|O_DIRECT, 0644);
a72d8a9f
MD
755 else
756 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
ee63ee00
SW
757 if (fd < 0 || fstat(fd, &st) < 0) {
758 err(1, "Unable to open/create %s", fname);
759 /* NOT REACHED */
760 }
6882b7a5
MD
761 if (S_ISREG(st.st_mode)) {
762 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
763 errx(1, "Disk image %s is already "
764 "in use\n", fname);
765 /* NOT REACHED */
766 }
767 }
ee63ee00
SW
768
769 info = &DiskInfo[DiskNum];
770 l = strlen(fname);
0171c06b 771
ee63ee00
SW
772 info->unit = i;
773 info->fd = fd;
a72d8a9f 774 info->type = type;
ee63ee00
SW
775 memcpy(info->fname, fname, l);
776
41556a7a 777 if (DiskNum == 0) {
1f1bc20e 778 if (type == VKD_CD) {
a72d8a9f 779 rootdevnames[0] = "cd9660:vcd0a";
1f1bc20e 780 } else if (type == VKD_DISK) {
28b44c8f 781 rootdevnames[0] = "ufs:vkd0s0a";
1f1bc20e
MD
782 rootdevnames[1] = "ufs:vkd0s1a";
783 }
a72d8a9f 784 }
ee63ee00
SW
785
786 DiskNum++;
787 } else {
23e2971a 788 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
ee63ee00 789 continue;
0171c06b
MD
790 }
791 }
44a09dd6
MD
792}
793
d869938c 794static
d53b52ae
SZ
795int
796netif_set_tapflags(int tap_unit, int f, int s)
797{
798 struct ifreq ifr;
799 int flags;
800
801 bzero(&ifr, sizeof(ifr));
802
803 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
804 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
5de37c8b 805 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
d53b52ae
SZ
806 return -1;
807 }
808
809 /*
810 * Adjust if_flags
811 *
812 * If the flags are already set/cleared, then we return
813 * immediately to avoid extra syscalls
814 */
815 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
816 if (f < 0) {
817 /* Turn off flags */
818 f = -f;
819 if ((flags & f) == 0)
820 return 0;
821 flags &= ~f;
822 } else {
823 /* Turn on flags */
824 if (flags & f)
825 return 0;
826 flags |= f;
827 }
828
829 /*
830 * Fix up ifreq.ifr_name, since it may be trashed
831 * in previous ioctl(SIOCGIFFLAGS)
832 */
833 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
834
835 ifr.ifr_flags = flags & 0xffff;
836 ifr.ifr_flagshigh = flags >> 16;
837 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
5de37c8b 838 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
d53b52ae
SZ
839 return -1;
840 }
841 return 0;
842}
843
844static
845int
846netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
847{
848 struct ifaliasreq ifra;
849 struct sockaddr_in *in;
850
851 bzero(&ifra, sizeof(ifra));
852 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
853
854 /* Setup address */
855 in = (struct sockaddr_in *)&ifra.ifra_addr;
856 in->sin_family = AF_INET;
857 in->sin_len = sizeof(*in);
858 in->sin_addr.s_addr = addr;
859
860 if (mask != 0) {
861 /* Setup netmask */
862 in = (struct sockaddr_in *)&ifra.ifra_mask;
863 in->sin_len = sizeof(*in);
864 in->sin_addr.s_addr = mask;
865 }
866
867 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
5de37c8b 868 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
d53b52ae
SZ
869 return -1;
870 }
871 return 0;
872}
873
874static
875int
876netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
877{
878 struct ifbreq ifbr;
879 struct ifdrv ifd;
880
881 bzero(&ifbr, sizeof(ifbr));
882 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
883 "tap%d", tap_unit);
884
885 bzero(&ifd, sizeof(ifd));
886 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
887 ifd.ifd_cmd = BRDGADD;
888 ifd.ifd_len = sizeof(ifbr);
889 ifd.ifd_data = &ifbr;
890
891 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
892 /*
893 * 'errno == EEXIST' means that the tap(4) is already
894 * a member of the bridge(4)
895 */
896 if (errno != EEXIST) {
5de37c8b 897 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
d53b52ae
SZ
898 return -1;
899 }
900 }
901 return 0;
902}
903
904#define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
905
d53b52ae
SZ
906/*
907 * Locate the first unused tap(4) device file if auto mode is requested,
908 * or open the user supplied device file, and bring up the corresponding
909 * tap(4) interface.
910 *
911 * NOTE: Only tap(4) device file is supported currently
912 */
913static
914int
915netif_open_tap(const char *netif, int *tap_unit, int s)
d869938c 916{
d53b52ae
SZ
917 char tap_dev[MAXPATHLEN];
918 int tap_fd, failed;
919 struct stat st;
235a1432 920 char *dname;
d53b52ae
SZ
921
922 *tap_unit = -1;
923
924 if (strcmp(netif, "auto") == 0) {
d53b52ae
SZ
925 /*
926 * Find first unused tap(4) device file
927 */
235a1432 928 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
d53b52ae 929 if (tap_fd < 0) {
235a1432 930 warnc(errno, "Unable to find a free tap(4)");
d53b52ae
SZ
931 return -1;
932 }
933 } else {
934 /*
91be174d 935 * User supplied tap(4) device file or unix socket.
d53b52ae
SZ
936 */
937 if (netif[0] == '/') /* Absolute path */
938 strlcpy(tap_dev, netif, sizeof(tap_dev));
939 else
940 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
941
942 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
91be174d
MD
943
944 /*
945 * If we cannot open normally try to connect to it.
946 */
947 if (tap_fd < 0)
948 tap_fd = unix_connect(tap_dev);
949
d53b52ae 950 if (tap_fd < 0) {
5de37c8b 951 warn("Unable to open %s", tap_dev);
d53b52ae 952 return -1;
d869938c
SW
953 }
954 }
d53b52ae
SZ
955
956 /*
957 * Check whether the device file is a tap(4)
958 */
235a1432
MD
959 if (fstat(tap_fd, &st) < 0) {
960 failed = 1;
961 } else if (S_ISCHR(st.st_mode)) {
962 dname = fdevname(tap_fd);
963 if (dname)
964 dname = strstr(dname, "tap");
965 if (dname) {
966 /*
967 * Bring up the corresponding tap(4) interface
968 */
969 *tap_unit = strtol(dname + 3, NULL, 10);
970 printf("TAP UNIT %d\n", *tap_unit);
971 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
972 failed = 0;
3b6e5cb0
SW
973 else
974 failed = 1;
235a1432
MD
975 } else {
976 failed = 1;
977 }
91be174d
MD
978 } else if (S_ISSOCK(st.st_mode)) {
979 /*
980 * Special socket connection (typically to vknet). We
981 * do not have to do anything.
982 */
983 failed = 0;
d53b52ae 984 } else {
235a1432 985 failed = 1;
d53b52ae
SZ
986 }
987
988 if (failed) {
235a1432 989 warnx("%s is not a tap(4) device or socket", tap_dev);
d53b52ae
SZ
990 close(tap_fd);
991 tap_fd = -1;
992 *tap_unit = -1;
993 }
994 return tap_fd;
995}
996
91be174d
MD
997static int
998unix_connect(const char *path)
999{
1000 struct sockaddr_un sunx;
1001 int len;
1002 int net_fd;
3a6117bb
MD
1003 int sndbuf = 262144;
1004 struct stat st;
91be174d
MD
1005
1006 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1007 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1008 ++len; /* include nul */
1009 sunx.sun_family = AF_UNIX;
1010 sunx.sun_len = len;
1011
1012 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1013 if (net_fd < 0)
1014 return(-1);
1015 if (connect(net_fd, (void *)&sunx, len) < 0) {
1016 close(net_fd);
1017 return(-1);
1018 }
3a6117bb
MD
1019 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1020 if (fstat(net_fd, &st) == 0)
1021 printf("Network socket buffer: %d bytes\n", st.st_blksize);
91be174d
MD
1022 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1023 return(net_fd);
1024}
1025
d53b52ae
SZ
1026#undef TAPDEV_MAJOR
1027#undef TAPDEV_MINOR
1028#undef TAPDEV_OFLAGS
1029
1030/*
1031 * Following syntax is supported,
1032 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1033 *
1034 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1035 * tap(4)'s netmask len is z
1036 *
1037 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1038 * pseudo netif's address is y.y.y.y
1039 *
1040 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1041 * pseudo netif's address is y.y.y.y
1042 * tap(4) and pseudo netif's netmask len are z
1043 *
1044 * 5) bridgeX tap(4) will be added to bridgeX
1045 *
1046 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1047 * pseudo netif's address is y.y.y.y
1048 *
1049 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1050 * pseudo netif's address is y.y.y.y
1051 * pseudo netif's netmask len is z
1052 */
1053static
1054int
1055netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1056{
1057 in_addr_t tap_addr, netmask, netif_addr;
1058 int next_netif_addr;
1059 char *tok, *masklen_str, *ifbridge;
1060
1061 *addr = 0;
1062 *mask = 0;
1063
1064 tok = strtok(NULL, ":/");
1065 if (tok == NULL) {
1066 /*
1067 * Nothing special, simply use tap(4) as backend
1068 */
1069 return 0;
1070 }
1071
1072 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1073 /*
1074 * tap(4)'s address is supplied
1075 */
1076 ifbridge = NULL;
1077
1078 /*
1079 * If there is next token, then it may be pseudo
1080 * netif's address or netmask len for tap(4)
1081 */
1082 next_netif_addr = 0;
1083 } else {
1084 /*
1085 * Not tap(4)'s address, assume it as a bridge(4)
1086 * iface name
1087 */
1088 tap_addr = 0;
1089 ifbridge = tok;
1090
1091 /*
1092 * If there is next token, then it must be pseudo
1093 * netif's address
1094 */
1095 next_netif_addr = 1;
1096 }
1097
1098 netmask = netif_addr = 0;
1099
1100 tok = strtok(NULL, ":/");
1101 if (tok == NULL)
1102 goto back;
1103
1104 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1105 if (next_netif_addr) {
1106 warnx("Invalid pseudo netif address: %s", tok);
1107 return -1;
1108 }
1109 netif_addr = 0;
1110
1111 /*
1112 * Current token is not address, then it must be netmask len
1113 */
1114 masklen_str = tok;
1115 } else {
1116 /*
1117 * Current token is pseudo netif address, if there is next token
1118 * it must be netmask len
1119 */
1120 masklen_str = strtok(NULL, "/");
1121 }
1122
1123 /* Calculate netmask */
1124 if (masklen_str != NULL) {
1125 u_long masklen;
1126
1127 masklen = strtoul(masklen_str, NULL, 10);
1128 if (masklen < 32 && masklen > 0) {
1129 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1130 & 0xffffffff);
1131 } else {
1132 warnx("Invalid netmask len: %lu", masklen);
1133 return -1;
1134 }
1135 }
1136
1137 /* Make sure there is no more token left */
1138 if (strtok(NULL, ":/") != NULL) {
1139 warnx("Invalid argument to '-I'");
1140 return -1;
1141 }
1142
1143back:
2c4a81d7
MD
1144 if (tap_unit < 0) {
1145 /* Do nothing */
1146 } else if (ifbridge == NULL) {
d53b52ae
SZ
1147 /* Set tap(4) address/netmask */
1148 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1149 return -1;
1150 } else {
1151 /* Tie tap(4) to bridge(4) */
1152 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1153 return -1;
1154 }
1155
1156 *addr = netif_addr;
1157 *mask = netmask;
1158 return 0;
1159}
1160
1161/*
1162 * NetifInfo[] will be filled for pseudo netif initialization.
1163 * NetifNum will be bumped to reflect the number of valid entries
1164 * in NetifInfo[].
1165 */
1166static
1167void
1168init_netif(char *netifExp[], int netifExpNum)
1169{
1170 int i, s;
1171
1172 if (netifExpNum == 0)
1173 return;
1174
1175 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1176 if (s < 0)
1177 return;
1178
1179 for (i = 0; i < netifExpNum; ++i) {
1180 struct vknetif_info *info;
1181 in_addr_t netif_addr, netif_mask;
1182 int tap_fd, tap_unit;
1183 char *netif;
1184
1185 netif = strtok(netifExp[i], ":");
1186 if (netif == NULL) {
ee63ee00 1187 warnx("Invalid argument to '-I'");
d53b52ae
SZ
1188 continue;
1189 }
1190
1191 /*
1192 * Open tap(4) device file and bring up the
1193 * corresponding interface
1194 */
1195 tap_fd = netif_open_tap(netif, &tap_unit, s);
1196 if (tap_fd < 0)
1197 continue;
1198
1199 /*
1200 * Initialize tap(4) and get address/netmask
1201 * for pseudo netif
1202 *
1203 * NB: Rest part of netifExp[i] is passed
1204 * to netif_init_tap() implicitly.
1205 */
1206 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1207 /*
1208 * NB: Closing tap(4) device file will bring
1209 * down the corresponding interface
1210 */
1211 close(tap_fd);
1212 continue;
1213 }
1214
1215 info = &NetifInfo[NetifNum];
1216 info->tap_fd = tap_fd;
1217 info->tap_unit = tap_unit;
1218 info->netif_addr = netif_addr;
1219 info->netif_mask = netif_mask;
1220
1221 NetifNum++;
1222 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1223 break;
1224 }
1225 close(s);
d869938c
SW
1226}
1227
bc3cc25e
MD
1228/*
1229 * Create the pid file and leave it open and locked while the vkernel is
1230 * running. This allows a script to use /usr/bin/lockf to probe whether
1231 * a vkernel is still running (so as not to accidently kill an unrelated
1232 * process from a stale pid file).
1233 */
b02b4170
MD
1234static
1235void
bc3cc25e 1236writepid(void)
b02b4170 1237{
bc3cc25e
MD
1238 char buf[32];
1239 int fd;
b02b4170
MD
1240
1241 if (pid_file != NULL) {
bc3cc25e
MD
1242 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1243 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1244 if (fd < 0) {
1245 if (errno == EWOULDBLOCK) {
1246 perror("Failed to lock pidfile, "
1247 "vkernel already running");
1248 } else {
1249 perror("Failed to create pidfile");
1250 }
1251 exit(EX_SOFTWARE);
b02b4170 1252 }
bc3cc25e
MD
1253 ftruncate(fd, 0);
1254 write(fd, buf, strlen(buf));
1255 /* leave the file open to maintain the lock */
b02b4170
MD
1256 }
1257}
1258
1259static
1260void
1261cleanpid( void )
1262{
1263 if (pid_file != NULL) {
bc3cc25e 1264 if (unlink(pid_file) < 0)
b02b4170
MD
1265 perror("Warning: couldn't remove pidfile");
1266 }
1267}
1268
44a09dd6
MD
1269static
1270void
278c4e67 1271usage_err(const char *ctl, ...)
9c059ae3 1272{
057f0718
MD
1273 va_list va;
1274
1275 va_start(va, ctl);
1276 vfprintf(stderr, ctl, va);
1277 va_end(va);
1278 fprintf(stderr, "\n");
278c4e67
SG
1279 exit(EX_USAGE);
1280}
1281
1282static
1283void
1284usage_help(_Bool help)
1285{
1286 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1287 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
1288 "\t[-m size] [-n numcpus] [-p file] [-r file]\n", save_av[0]);
1289
1290 if (help)
1291 fprintf(stderr, "\nArguments:\n"
1292 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1293 "\t-e\tSpecify an environment to be used by the kernel.\n"
1294 "\t-h\tThis list of options.\n"
1295 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1296 "\t-I\tCreate a virtual network device.\n"
1297 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1298 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
1299 "\t-n\tSpecify the number of CPUs you wish to emulate.\n"
1300 "\t-p\tSpecify a file in which to store the process ID.\n"
1301 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1302 "\t-s\tBoot into single-user mode.\n"
1303 "\t-U\tEnable writing to kernel memory and module loading.\n"
1304 "\t-v\tTurn on verbose booting.\n");
1305
1306 exit(EX_USAGE);
9c059ae3
MD
1307}
1308
8a9d8995
MD
1309void
1310cpu_reset(void)
1311{
40d0276c
MD
1312 kprintf("cpu reset, rebooting vkernel\n");
1313 closefrom(3);
b02b4170 1314 cleanpid();
40d0276c 1315 execv(save_av[0], save_av);
8a9d8995 1316}
6f7b98e0
MD
1317
1318void
1319cpu_halt(void)
1320{
40d0276c 1321 kprintf("cpu halt, exiting vkernel\n");
b02b4170 1322 cleanpid();
278c4e67 1323 exit(EX_OK);
6f7b98e0 1324}
702acf06
JT
1325
1326void
1327setrealcpu(void)
1328{
1329 switch(lwp_cpu_lock) {
1330 case LCL_PER_CPU:
1331 if (bootverbose)
1332 kprintf("Locking CPU%d to real cpu %d\n",
1333 mycpuid, next_cpu);
1334 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1335 next_cpu++;
1336 if (next_cpu >= real_ncpus)
1337 next_cpu = 0;
1338 break;
1339 case LCL_SINGLE_CPU:
1340 if (bootverbose)
1341 kprintf("Locking CPU%d to real cpu %d\n",
1342 mycpuid, next_cpu);
1343 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1344 break;
1345 default:
1346 /* do not map virtual cpus to real cpus */
1347 break;
1348 }
1349}
1350