vkernel{,64} - CPU topology support
[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 */
9bea6114
MC
115int vkernel_b_arg; /* -b argument - no of logical CPU bits - only SMP */
116int vkernel_B_arg; /* -B argument - no of core bits - only SMP */
9c059ae3 117
1a8fe9c5
MD
118int via_feature_xcrypt = 0; /* XXX */
119int via_feature_rng = 0; /* XXX */
120
8a9d8995
MD
121struct privatespace *CPU_prvspace;
122
252b2227 123static struct trapframe proc0_tf;
e4a473f1
MD
124static void *proc0paddr;
125
44a09dd6
MD
126static void init_sys_memory(char *imageFile);
127static void init_kern_memory(void);
8a9d8995 128static void init_globaldata(void);
e4a473f1 129static void init_vkernel(void);
a72d8a9f 130static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
ee63ee00 131static void init_netif(char *netifExp[], int netifFileNum);
278c4e67
SG
132static void writepid(void);
133static void cleanpid(void);
91be174d 134static int unix_connect(const char *path);
278c4e67
SG
135static void usage_err(const char *ctl, ...);
136static void usage_help(_Bool);
9c059ae3 137
40d0276c
MD
138static int save_ac;
139static char **save_av;
140
9c059ae3
MD
141/*
142 * Kernel startup for virtual kernels - standard main()
143 */
144int
145main(int ac, char **av)
146{
44a09dd6 147 char *memImageFile = NULL;
d53b52ae 148 char *netifFile[VKNETIF_MAX];
ee63ee00 149 char *diskFile[VKDISK_MAX];
a72d8a9f 150 char *cdFile[VKDISK_MAX];
44a09dd6 151 char *suffix;
702acf06 152 char *endp;
974f8ffe 153 char *tmp;
9bea6114 154 char *tok;
d53b52ae 155 int netifFileNum = 0;
ee63ee00 156 int diskFileNum = 0;
a72d8a9f 157 int cdFileNum = 0;
41556a7a 158 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
44a09dd6 159 int c;
aaf8b91f 160 int i;
a6d77ab5 161 int j;
aaf8b91f 162 int n;
a6d77ab5 163 int isq;
974f8ffe
AHJ
164 int pos;
165 int eflag;
dc7503af 166 int real_vkernel_enable;
aecc8383 167 int supports_sse;
870b0161 168 size_t vsize;
974f8ffe
AHJ
169 size_t kenv_size;
170 size_t kenv_size2;
b02b4170 171
40d0276c
MD
172 save_ac = ac;
173 save_av = av;
974f8ffe
AHJ
174 eflag = 0;
175 pos = 0;
176 kenv_size = 0;
40d0276c 177
9c059ae3
MD
178 /*
179 * Process options
180 */
7e42c007 181 kernel_mem_readonly = 1;
c5b0b0ba
MD
182#ifdef SMP
183 optcpus = 2;
9bea6114
MC
184 vkernel_b_arg = 0;
185 vkernel_B_arg = 0;
c5b0b0ba 186#endif
702acf06 187 lwp_cpu_lock = LCL_NONE;
7e42c007 188
27c3aec1 189 real_vkernel_enable = 0;
870b0161
MD
190 vsize = sizeof(real_vkernel_enable);
191 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
dc7503af 192
27c3aec1
MD
193 if (real_vkernel_enable == 0) {
194 errx(1, "vm.vkernel_enable is 0, must be set "
195 "to 1 to execute a vkernel!");
196 }
dc7503af 197
27c3aec1 198 real_ncpus = 1;
870b0161
MD
199 vsize = sizeof(real_ncpus);
200 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
702acf06 201
278c4e67
SG
202 if (ac < 2)
203 usage_help(false);
204
9bea6114 205 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:Ub:B:")) != -1) {
9c059ae3 206 switch(c) {
aaf8b91f
MD
207 case 'e':
208 /*
209 * name=value:name=value:name=value...
a6d77ab5
MD
210 * name="value"...
211 *
212 * Allow values to be quoted but note that shells
213 * may remove the quotes, so using this feature
214 * to embed colons may require a backslash.
aaf8b91f
MD
215 */
216 n = strlen(optarg);
a6d77ab5 217 isq = 0;
974f8ffe
AHJ
218
219 if (eflag == 0) {
220 kenv_size = n + 2;
221 kern_envp = malloc(kenv_size);
222 if (kern_envp == NULL)
223 errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size);
224 } else {
225 kenv_size2 = kenv_size + n + 1;
226 pos = kenv_size - 1;
227 if ((tmp = realloc(kern_envp, kenv_size2)) == NULL)
228 errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2);
229 kern_envp = tmp;
230 kenv_size = kenv_size2;
231 }
232
233 for (i = 0, j = pos; i < n; ++i) {
a6d77ab5
MD
234 if (optarg[i] == '"')
235 isq ^= 1;
236 else if (optarg[i] == '\'')
237 isq ^= 2;
238 else if (isq == 0 && optarg[i] == ':')
239 kern_envp[j++] = 0;
aaf8b91f 240 else
a6d77ab5 241 kern_envp[j++] = optarg[i];
aaf8b91f 242 }
a6d77ab5
MD
243 kern_envp[j++] = 0;
244 kern_envp[j++] = 0;
974f8ffe 245 eflag++;
aaf8b91f 246 break;
884d6eea
MD
247 case 's':
248 boothowto |= RB_SINGLE;
249 break;
9c059ae3
MD
250 case 'v':
251 bootverbose = 1;
252 break;
253 case 'i':
254 memImageFile = optarg;
255 break;
d869938c 256 case 'I':
d53b52ae 257 if (netifFileNum < VKNETIF_MAX)
40d0276c 258 netifFile[netifFileNum++] = strdup(optarg);
d869938c 259 break;
9c059ae3 260 case 'r':
41556a7a
MD
261 if (bootOnDisk < 0)
262 bootOnDisk = 1;
a72d8a9f 263 if (diskFileNum + cdFileNum < VKDISK_MAX)
40d0276c 264 diskFile[diskFileNum++] = strdup(optarg);
9c059ae3 265 break;
a72d8a9f 266 case 'c':
41556a7a
MD
267 if (bootOnDisk < 0)
268 bootOnDisk = 0;
a72d8a9f 269 if (diskFileNum + cdFileNum < VKDISK_MAX)
40d0276c 270 cdFile[cdFileNum++] = strdup(optarg);
a72d8a9f 271 break;
9c059ae3 272 case 'm':
71152ac6 273 Maxmem_bytes = strtoull(optarg, &suffix, 0);
9c059ae3
MD
274 if (suffix) {
275 switch(*suffix) {
276 case 'g':
277 case 'G':
71152ac6 278 Maxmem_bytes <<= 30;
9c059ae3
MD
279 break;
280 case 'm':
281 case 'M':
71152ac6 282 Maxmem_bytes <<= 20;
9c059ae3
MD
283 break;
284 case 'k':
285 case 'K':
71152ac6 286 Maxmem_bytes <<= 10;
9c059ae3
MD
287 break;
288 default:
71152ac6 289 Maxmem_bytes = 0;
278c4e67 290 usage_err("Bad maxmem option");
9c059ae3
MD
291 /* NOT REACHED */
292 break;
293 }
294 }
295 break;
702acf06
JT
296 case 'l':
297 next_cpu = -1;
298 if (strncmp("map", optarg, 3) == 0) {
299 lwp_cpu_lock = LCL_PER_CPU;
92d2f80a 300 if (optarg[3] == ',') {
702acf06 301 next_cpu = strtol(optarg+4, &endp, 0);
92d2f80a 302 if (*endp != '\0')
278c4e67 303 usage_err("Bad target CPU number at '%s'", endp);
92d2f80a
MD
304 } else {
305 next_cpu = 0;
306 }
702acf06 307 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 308 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
702acf06
JT
309 } else if (strncmp("any", optarg, 3) == 0) {
310 lwp_cpu_lock = LCL_NONE;
311 } else {
312 lwp_cpu_lock = LCL_SINGLE_CPU;
313 next_cpu = strtol(optarg, &endp, 0);
314 if (*endp != '\0')
278c4e67 315 usage_err("Bad target CPU number at '%s'", endp);
702acf06 316 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 317 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
702acf06
JT
318 }
319 break;
c5b0b0ba
MD
320 case 'n':
321 /*
322 * This value is set up by mp_start(), don't just
323 * set ncpus here.
324 */
9bea6114 325 tok = strtok(optarg, ":");
c5b0b0ba 326#ifdef SMP
9bea6114 327 optcpus = strtol(tok, NULL, 0);
057f0718 328 if (optcpus < 1 || optcpus > MAXCPU)
278c4e67 329 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
9bea6114
MC
330
331 /* :core_bits argument */
332 tok = strtok(NULL, ":");
333 if (tok != NULL) {
334 vkernel_b_arg = strtol(tok, NULL, 0);
335
336 /* :logical_CPU_bits argument */
337 tok = strtok(NULL, ":");
338 if (tok != NULL) {
339 vkernel_B_arg = strtol(tok, NULL, 0);
340 }
341
342 }
343
c5b0b0ba 344#else
9bea6114 345 if (strtol(tok, NULL, 0) != 1) {
278c4e67 346 usage_err("You built a UP vkernel, only 1 cpu!");
c5b0b0ba 347 }
9bea6114
MC
348
349 /* :logical_CPU_bits argument */
350 tok = strtok(NULL, ":");
351 if (tok != NULL) {
352 usage_err("You built a UP vkernel. No CPU topology available");
353
354 /* :core_bits argument */
355 tok = strtok(NULL, ":");
356 if (tok != NULL) {
357 usage_err("You built a UP vkernel. No CPU topology available");
358 }
359 }
c5b0b0ba
MD
360#endif
361
151f45bf 362 break;
b02b4170 363 case 'p':
9bea6114 364 pid_file = optarg;
b02b4170 365 break;
7e42c007
MD
366 case 'U':
367 kernel_mem_readonly = 0;
368 break;
278c4e67
SG
369 case 'h':
370 usage_help(true);
371 break;
278c4e67
SG
372 default:
373 usage_help(false);
9c059ae3
MD
374 }
375 }
376
b02b4170 377 writepid();
b1b21fa6 378 cpu_disable_intr();
9c059ae3
MD
379 init_sys_memory(memImageFile);
380 init_kern_memory();
8a9d8995 381 init_globaldata();
e4a473f1 382 init_vkernel();
702acf06 383 setrealcpu();
f991237f 384 init_kqueue();
41556a7a 385
08771751
VS
386 vmm_guest = 1;
387
870b0161
MD
388 /*
389 * Check TSC
390 */
391 vsize = sizeof(tsc_present);
392 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
393 vsize = sizeof(tsc_frequency);
394 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
395 if (tsc_present)
396 cpu_feature |= CPUID_TSC;
397
398 /*
399 * Check SSE
400 */
401 vsize = sizeof(supports_sse);
27c3aec1 402 supports_sse = 0;
870b0161 403 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
aecc8383 404 init_fpu(supports_sse);
870b0161
MD
405 if (supports_sse)
406 cpu_feature |= CPUID_SSE | CPUID_FXSR;
aecc8383 407
41556a7a
MD
408 /*
409 * We boot from the first installed disk.
410 */
411 if (bootOnDisk == 1) {
412 init_disk(diskFile, diskFileNum, VKD_DISK);
413 init_disk(cdFile, cdFileNum, VKD_CD);
414 } else {
415 init_disk(cdFile, cdFileNum, VKD_CD);
416 init_disk(diskFile, diskFileNum, VKD_DISK);
417 }
d53b52ae 418 init_netif(netifFile, netifFileNum);
6092278a 419 init_exceptions();
9c059ae3
MD
420 mi_startup();
421 /* NOT REACHED */
278c4e67 422 exit(EX_SOFTWARE);
9c059ae3
MD
423}
424
425/*
426 * Initialize system memory. This is the virtual kernel's 'RAM'.
427 */
428static
429void
44a09dd6 430init_sys_memory(char *imageFile)
9c059ae3 431{
44a09dd6 432 struct stat st;
3748859a 433 int i;
9c059ae3
MD
434 int fd;
435
436 /*
437 * Figure out the system memory image size. If an image file was
438 * specified and -m was not specified, use the image file's size.
439 */
71152ac6
MD
440 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
441 Maxmem_bytes = (vm_paddr_t)st.st_size;
442 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
443 Maxmem_bytes == 0) {
376264e3 444 errx(1, "Cannot create new memory file %s unless "
9c059ae3
MD
445 "system memory size is specified with -m",
446 imageFile);
447 /* NOT REACHED */
448 }
449
450 /*
451 * Maxmem must be known at this time
452 */
71152ac6 453 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
376264e3 454 errx(1, "Bad maxmem specification: 32MB minimum, "
9c059ae3
MD
455 "multiples of %dMB only",
456 SEG_SIZE / 1024 / 1024);
457 /* NOT REACHED */
458 }
459
460 /*
461 * Generate an image file name if necessary, then open/create the
462 * file exclusively locked. Do not allow multiple virtual kernels
463 * to use the same image file.
bc3cc25e
MD
464 *
465 * Don't iterate through a million files if we do not have write
466 * access to the directory, stop if our open() failed on a
467 * non-existant file. Otherwise opens can fail for any number
468 * of reasons (lock failed, file not owned or writable by us, etc).
9c059ae3 469 */
3748859a
MD
470 if (imageFile == NULL) {
471 for (i = 0; i < 1000000; ++i) {
0171c06b 472 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
3748859a
MD
473 fd = open(imageFile,
474 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
bc3cc25e 475 if (fd < 0 && stat(imageFile, &st) == 0) {
3748859a
MD
476 free(imageFile);
477 continue;
478 }
479 break;
480 }
481 } else {
482 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
483 }
0171c06b 484 printf("Using memory file: %s\n", imageFile);
9c059ae3 485 if (fd < 0 || fstat(fd, &st) < 0) {
5de37c8b 486 err(1, "Unable to open/create %s", imageFile);
9c059ae3
MD
487 /* NOT REACHED */
488 }
489
490 /*
b72e51d1
MD
491 * Truncate or extend the file as necessary. Clean out the contents
492 * of the file, we want it to be full of holes so we don't waste
493 * time reading in data from an old file that we no longer care
494 * about.
9c059ae3 495 */
b72e51d1
MD
496 ftruncate(fd, 0);
497 ftruncate(fd, Maxmem_bytes);
9c059ae3 498
9c059ae3 499 MemImageFd = fd;
71152ac6 500 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
9f90716f 501 physmem = Maxmem;
9c059ae3
MD
502}
503
504/*
505 * Initialize kernel memory. This reserves kernel virtual memory by using
506 * MAP_VPAGETABLE
507 */
0c4c980f 508
9c059ae3
MD
509static
510void
511init_kern_memory(void)
512{
513 void *base;
0c4c980f 514 void *try;
3748859a 515 char *zero;
0c4c980f
MD
516 char dummy;
517 char *topofstack = &dummy;
e4a473f1
MD
518 vpte_t pte;
519 int i;
9c059ae3
MD
520
521 /*
522 * Memory map our kernel virtual memory space. Note that the
523 * kernel image itself is not made part of this memory for the
524 * moment.
71152ac6
MD
525 *
526 * The memory map must be segment-aligned so we can properly
527 * offset KernelPTD.
0c4c980f
MD
528 *
529 * If the system kernel has a different MAXDSIZ, it might not
530 * be possible to map kernel memory in its prefered location.
531 * Try a number of different locations.
9c059ae3 532 */
0c4c980f
MD
533 try = (void *)0x40000000;
534 base = NULL;
535 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
536 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
537 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
538 MemImageFd, 0);
539 if (base == try)
540 break;
541 if (base != MAP_FAILED)
542 munmap(base, KERNEL_KVA_SIZE);
543 try = (char *)try + 0x10000000;
544 }
545 if (base != try) {
9c059ae3
MD
546 err(1, "Unable to mmap() kernel virtual memory!");
547 /* NOT REACHED */
548 }
a3c35df6 549 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
e4a473f1 550 KvaStart = (vm_offset_t)base;
9c059ae3 551 KvaSize = KERNEL_KVA_SIZE;
e4a473f1 552 KvaEnd = KvaStart + KvaSize;
d4f5f27c
MD
553
554 /* cannot use kprintf yet */
0c4c980f 555 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
e4a473f1
MD
556
557 /*
558 * Create a top-level page table self-mapping itself.
559 *
560 * Initialize the page directory at physical page index 0 to point
561 * to an array of page table pages starting at physical page index 1
562 */
563 lseek(MemImageFd, 0L, 0);
564 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
565 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
566 write(MemImageFd, &pte, sizeof(pte));
567 }
568
569 /*
570 * Initialize the PTEs in the page table pages required to map the
571 * page table itself. This includes mapping the page directory page
572 * at the base so we go one more loop then normal.
573 */
574 lseek(MemImageFd, PAGE_SIZE, 0);
71152ac6 575 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
e4a473f1
MD
576 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
577 write(MemImageFd, &pte, sizeof(pte));
578 }
579
580 /*
3748859a
MD
581 * Initialize remaining PTEs to 0. We may be reusing a memory image
582 * file. This is approximately a megabyte.
583 */
584 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
585 zero = malloc(PAGE_SIZE);
b87d6bd2 586 bzero(zero, PAGE_SIZE);
3748859a
MD
587 while (i) {
588 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
589 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
590 }
591 free(zero);
592
593 /*
e4a473f1
MD
594 * Enable the page table and calculate pointers to our self-map
595 * for easy kernel page table manipulation.
71152ac6
MD
596 *
597 * KernelPTA must be offset so we can do direct VA translations
e4a473f1 598 */
71152ac6
MD
599 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
600 0 | VPTE_R | VPTE_W | VPTE_V);
e4a473f1
MD
601 KernelPTD = (vpte_t *)base; /* pg directory */
602 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
71152ac6 603 KernelPTA -= KvaStart >> PAGE_SHIFT;
e4a473f1
MD
604
605 /*
606 * phys_avail[] represents unallocated physical memory. MI code
607 * will use phys_avail[] to create the vm_page array.
608 */
609 phys_avail[0] = PAGE_SIZE +
610 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
611 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
71152ac6 612 phys_avail[1] = Maxmem_bytes;
e4a473f1
MD
613
614 /*
615 * (virtual_start, virtual_end) represent unallocated kernel virtual
616 * memory. MI code will create kernel_map using these parameters.
617 */
618 virtual_start = KvaStart + PAGE_SIZE +
619 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
620 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
621 virtual_end = KvaStart + KERNEL_KVA_SIZE;
622
623 /*
00835518
MD
624 * kernel_vm_end could be set to virtual_end but we want some
625 * indication of how much of the kernel_map we've used, so
626 * set it low and let pmap_growkernel increase it even though we
627 * don't need to create any new page table pages.
e4a473f1 628 */
00835518 629 kernel_vm_end = virtual_start;
e4a473f1
MD
630
631 /*
632 * Allocate space for process 0's UAREA.
633 */
634 proc0paddr = (void *)virtual_start;
635 for (i = 0; i < UPAGES; ++i) {
636 pmap_kenter_quick(virtual_start, phys_avail[0]);
637 virtual_start += PAGE_SIZE;
638 phys_avail[0] += PAGE_SIZE;
639 }
9c059ae3
MD
640
641 /*
e4a473f1 642 * crashdumpmap
9c059ae3 643 */
e4a473f1
MD
644 crashdumpmap = virtual_start;
645 virtual_start += MAXDUMPPGS * PAGE_SIZE;
646
647 /*
6f7b98e0
MD
648 * msgbufp maps the system message buffer
649 */
650 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
651 msgbufp = (void *)virtual_start;
652 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
653 pmap_kenter_quick(virtual_start, phys_avail[0]);
654 virtual_start += PAGE_SIZE;
655 phys_avail[0] += PAGE_SIZE;
656 }
657 msgbufinit(msgbufp, MSGBUF_SIZE);
658
659 /*
660 * used by kern_memio for /dev/mem access
661 */
662 ptvmmap = (caddr_t)virtual_start;
663 virtual_start += PAGE_SIZE;
664
665 /*
e4a473f1
MD
666 * Bootstrap the kernel_pmap
667 */
668 pmap_bootstrap();
9c059ae3
MD
669}
670
671/*
e4a473f1
MD
672 * Map the per-cpu globaldata for cpu #0. Allocate the space using
673 * virtual_start and phys_avail[0]
8a9d8995
MD
674 */
675static
676void
677init_globaldata(void)
678{
e4a473f1
MD
679 int i;
680 vm_paddr_t pa;
681 vm_offset_t va;
682
683 /*
684 * Reserve enough KVA to cover possible cpus. This is a considerable
685 * amount of KVA since the privatespace structure includes two
686 * whole page table mappings.
687 */
688 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
689 CPU_prvspace = (void *)virtual_start;
690 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
691
692 /*
693 * Allocate enough physical memory to cover the mdglobaldata
694 * portion of the space and the idle stack and map the pages
695 * into KVA. For cpu #0 only.
696 */
697 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
698 pa = phys_avail[0];
699 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
700 pmap_kenter_quick(va, pa);
701 phys_avail[0] += PAGE_SIZE;
702 }
703 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
704 pa = phys_avail[0];
705 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
706 pmap_kenter_quick(va, pa);
707 phys_avail[0] += PAGE_SIZE;
708 }
709
710 /*
27c3aec1
MD
711 * Setup the %fs for cpu #0. The mycpu macro works after this
712 * point. Note that %gs is used by pthreads.
e4a473f1 713 */
9ad680a3 714 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
e4a473f1
MD
715}
716
717/*
718 * Initialize very low level systems including thread0, proc0, etc.
719 */
720static
721void
722init_vkernel(void)
723{
724 struct mdglobaldata *gd;
725
726 gd = &CPU_prvspace[0].mdglobaldata;
727 bzero(gd, sizeof(*gd));
728
729 gd->mi.gd_curthread = &thread0;
71152ac6 730 thread0.td_gd = &gd->mi;
e4a473f1 731 ncpus = 1;
b45759e1
MD
732 ncpus2 = 1; /* rounded down power of 2 */
733 ncpus_fit = 1; /* rounded up power of 2 */
734 /* ncpus2_mask and ncpus_fit_mask are 0 */
e4a473f1
MD
735 init_param1();
736 gd->mi.gd_prvspace = &CPU_prvspace[0];
737 mi_gdinit(&gd->mi, 0);
738 cpu_gdinit(gd, 0);
739 mi_proc0init(&gd->mi, proc0paddr);
08f2f1bb 740 lwp0.lwp_md.md_regs = &proc0_tf;
252b2227 741
e4a473f1 742 /*init_locks();*/
e130cdcb
MD
743#ifdef SMP
744 /*
745 * Get the initial mplock with a count of 1 for the BSP.
746 * This uses a LOGICAL cpu ID, ie BSP == 0.
747 */
748 cpu_get_initial_mplock();
749#endif
e4a473f1
MD
750 cninit();
751 rand_initialize();
752#if 0 /* #ifdef DDB */
753 kdb_init();
754 if (boothowto & RB_KDB)
755 Debugger("Boot flags requested debugger");
756#endif
514621ec 757 identcpu();
e4a473f1
MD
758#if 0
759 initializecpu(); /* Initialize CPU registers */
760#endif
761 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
762
763#if 0
764 /*
765 * Map the message buffer
766 */
767 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
768 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
769 msgbufinit(msgbufp, MSGBUF_SIZE);
770#endif
771#if 0
772 thread0.td_pcb_cr3 ... MMU
08f2f1bb 773 lwp0.lwp_md.md_regs = &proc0_tf;
e4a473f1 774#endif
8a9d8995
MD
775}
776
777/*
ee63ee00
SW
778 * Filesystem image paths for the virtual kernel are optional.
779 * If specified they each should point to a disk image,
780 * the first of which will become the root disk.
5b85cb7d
MD
781 *
782 * The virtual kernel caches data from our 'disk' just like a normal kernel,
783 * so we do not really want the real kernel to cache the data too. Use
784 * O_DIRECT to remove the duplication.
9c059ae3
MD
785 */
786static
787void
a72d8a9f 788init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
44a09dd6 789{
ee63ee00
SW
790 int i;
791
792 if (diskFileNum == 0)
793 return;
794
795 for(i=0; i < diskFileNum; i++){
796 char *fname;
797 fname = diskExp[i];
798
799 if (fname == NULL) {
800 warnx("Invalid argument to '-r'");
801 continue;
802 }
803
804 if (DiskNum < VKDISK_MAX) {
805 struct stat st;
806 struct vkdisk_info* info = NULL;
807 int fd;
808 size_t l = 0;
809
a72d8a9f 810 if (type == VKD_DISK)
6882b7a5 811 fd = open(fname, O_RDWR|O_DIRECT, 0644);
a72d8a9f
MD
812 else
813 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
ee63ee00
SW
814 if (fd < 0 || fstat(fd, &st) < 0) {
815 err(1, "Unable to open/create %s", fname);
816 /* NOT REACHED */
817 }
6882b7a5
MD
818 if (S_ISREG(st.st_mode)) {
819 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
820 errx(1, "Disk image %s is already "
821 "in use\n", fname);
822 /* NOT REACHED */
823 }
824 }
ee63ee00
SW
825
826 info = &DiskInfo[DiskNum];
827 l = strlen(fname);
0171c06b 828
ee63ee00
SW
829 info->unit = i;
830 info->fd = fd;
a72d8a9f 831 info->type = type;
ee63ee00
SW
832 memcpy(info->fname, fname, l);
833
41556a7a 834 if (DiskNum == 0) {
1f1bc20e 835 if (type == VKD_CD) {
a72d8a9f 836 rootdevnames[0] = "cd9660:vcd0a";
1f1bc20e 837 } else if (type == VKD_DISK) {
28b44c8f 838 rootdevnames[0] = "ufs:vkd0s0a";
1f1bc20e
MD
839 rootdevnames[1] = "ufs:vkd0s1a";
840 }
a72d8a9f 841 }
ee63ee00
SW
842
843 DiskNum++;
844 } else {
23e2971a 845 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
ee63ee00 846 continue;
0171c06b
MD
847 }
848 }
44a09dd6
MD
849}
850
851static
d53b52ae
SZ
852int
853netif_set_tapflags(int tap_unit, int f, int s)
854{
855 struct ifreq ifr;
856 int flags;
857
858 bzero(&ifr, sizeof(ifr));
859
860 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
861 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
5de37c8b 862 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
d53b52ae
SZ
863 return -1;
864 }
865
866 /*
867 * Adjust if_flags
868 *
869 * If the flags are already set/cleared, then we return
870 * immediately to avoid extra syscalls
871 */
872 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
873 if (f < 0) {
874 /* Turn off flags */
875 f = -f;
876 if ((flags & f) == 0)
877 return 0;
878 flags &= ~f;
879 } else {
880 /* Turn on flags */
881 if (flags & f)
882 return 0;
883 flags |= f;
884 }
885
886 /*
887 * Fix up ifreq.ifr_name, since it may be trashed
888 * in previous ioctl(SIOCGIFFLAGS)
889 */
890 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
891
892 ifr.ifr_flags = flags & 0xffff;
893 ifr.ifr_flagshigh = flags >> 16;
894 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
5de37c8b 895 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
d53b52ae
SZ
896 return -1;
897 }
898 return 0;
899}
900
901static
902int
903netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
904{
905 struct ifaliasreq ifra;
906 struct sockaddr_in *in;
907
908 bzero(&ifra, sizeof(ifra));
909 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
910
911 /* Setup address */
912 in = (struct sockaddr_in *)&ifra.ifra_addr;
913 in->sin_family = AF_INET;
914 in->sin_len = sizeof(*in);
915 in->sin_addr.s_addr = addr;
916
917 if (mask != 0) {
918 /* Setup netmask */
919 in = (struct sockaddr_in *)&ifra.ifra_mask;
920 in->sin_len = sizeof(*in);
921 in->sin_addr.s_addr = mask;
922 }
923
924 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
5de37c8b 925 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
d53b52ae
SZ
926 return -1;
927 }
928 return 0;
929}
930
931static
932int
933netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
934{
935 struct ifbreq ifbr;
936 struct ifdrv ifd;
937
938 bzero(&ifbr, sizeof(ifbr));
939 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
940 "tap%d", tap_unit);
941
942 bzero(&ifd, sizeof(ifd));
943 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
944 ifd.ifd_cmd = BRDGADD;
945 ifd.ifd_len = sizeof(ifbr);
946 ifd.ifd_data = &ifbr;
947
948 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
949 /*
950 * 'errno == EEXIST' means that the tap(4) is already
951 * a member of the bridge(4)
952 */
953 if (errno != EEXIST) {
5de37c8b 954 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
d53b52ae
SZ
955 return -1;
956 }
957 }
958 return 0;
959}
960
961#define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
962
d53b52ae
SZ
963/*
964 * Locate the first unused tap(4) device file if auto mode is requested,
965 * or open the user supplied device file, and bring up the corresponding
966 * tap(4) interface.
967 *
968 * NOTE: Only tap(4) device file is supported currently
969 */
970static
971int
972netif_open_tap(const char *netif, int *tap_unit, int s)
d869938c 973{
d53b52ae
SZ
974 char tap_dev[MAXPATHLEN];
975 int tap_fd, failed;
976 struct stat st;
235a1432 977 char *dname;
d53b52ae
SZ
978
979 *tap_unit = -1;
980
981 if (strcmp(netif, "auto") == 0) {
d53b52ae
SZ
982 /*
983 * Find first unused tap(4) device file
984 */
235a1432 985 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
d53b52ae 986 if (tap_fd < 0) {
235a1432 987 warnc(errno, "Unable to find a free tap(4)");
d53b52ae
SZ
988 return -1;
989 }
990 } else {
991 /*
91be174d 992 * User supplied tap(4) device file or unix socket.
d53b52ae
SZ
993 */
994 if (netif[0] == '/') /* Absolute path */
995 strlcpy(tap_dev, netif, sizeof(tap_dev));
996 else
997 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
998
999 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
91be174d
MD
1000
1001 /*
1002 * If we cannot open normally try to connect to it.
1003 */
1004 if (tap_fd < 0)
1005 tap_fd = unix_connect(tap_dev);
1006
d53b52ae 1007 if (tap_fd < 0) {
5de37c8b 1008 warn("Unable to open %s", tap_dev);
d53b52ae 1009 return -1;
d869938c
SW
1010 }
1011 }
d53b52ae
SZ
1012
1013 /*
1014 * Check whether the device file is a tap(4)
1015 */
235a1432
MD
1016 if (fstat(tap_fd, &st) < 0) {
1017 failed = 1;
1018 } else if (S_ISCHR(st.st_mode)) {
1019 dname = fdevname(tap_fd);
1020 if (dname)
1021 dname = strstr(dname, "tap");
1022 if (dname) {
1023 /*
1024 * Bring up the corresponding tap(4) interface
1025 */
1026 *tap_unit = strtol(dname + 3, NULL, 10);
1027 printf("TAP UNIT %d\n", *tap_unit);
1028 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
1029 failed = 0;
3b6e5cb0
SW
1030 else
1031 failed = 1;
235a1432
MD
1032 } else {
1033 failed = 1;
1034 }
91be174d
MD
1035 } else if (S_ISSOCK(st.st_mode)) {
1036 /*
1037 * Special socket connection (typically to vknet). We
1038 * do not have to do anything.
1039 */
1040 failed = 0;
d53b52ae 1041 } else {
235a1432 1042 failed = 1;
d53b52ae
SZ
1043 }
1044
1045 if (failed) {
235a1432 1046 warnx("%s is not a tap(4) device or socket", tap_dev);
d53b52ae
SZ
1047 close(tap_fd);
1048 tap_fd = -1;
1049 *tap_unit = -1;
1050 }
1051 return tap_fd;
1052}
1053
91be174d
MD
1054static int
1055unix_connect(const char *path)
1056{
1057 struct sockaddr_un sunx;
1058 int len;
1059 int net_fd;
3a6117bb
MD
1060 int sndbuf = 262144;
1061 struct stat st;
91be174d
MD
1062
1063 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1064 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1065 ++len; /* include nul */
1066 sunx.sun_family = AF_UNIX;
1067 sunx.sun_len = len;
1068
1069 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1070 if (net_fd < 0)
1071 return(-1);
1072 if (connect(net_fd, (void *)&sunx, len) < 0) {
1073 close(net_fd);
1074 return(-1);
1075 }
3a6117bb
MD
1076 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1077 if (fstat(net_fd, &st) == 0)
1078 printf("Network socket buffer: %d bytes\n", st.st_blksize);
91be174d
MD
1079 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1080 return(net_fd);
1081}
1082
d53b52ae
SZ
1083#undef TAPDEV_MAJOR
1084#undef TAPDEV_MINOR
1085#undef TAPDEV_OFLAGS
1086
1087/*
1088 * Following syntax is supported,
1089 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1090 *
1091 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1092 * tap(4)'s netmask len is z
1093 *
1094 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1095 * pseudo netif's address is y.y.y.y
1096 *
1097 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1098 * pseudo netif's address is y.y.y.y
1099 * tap(4) and pseudo netif's netmask len are z
1100 *
1101 * 5) bridgeX tap(4) will be added to bridgeX
1102 *
1103 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1104 * pseudo netif's address is y.y.y.y
1105 *
1106 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1107 * pseudo netif's address is y.y.y.y
1108 * pseudo netif's netmask len is z
1109 */
1110static
1111int
1112netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1113{
1114 in_addr_t tap_addr, netmask, netif_addr;
1115 int next_netif_addr;
1116 char *tok, *masklen_str, *ifbridge;
1117
1118 *addr = 0;
1119 *mask = 0;
1120
1121 tok = strtok(NULL, ":/");
1122 if (tok == NULL) {
1123 /*
1124 * Nothing special, simply use tap(4) as backend
1125 */
1126 return 0;
1127 }
1128
1129 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1130 /*
1131 * tap(4)'s address is supplied
1132 */
1133 ifbridge = NULL;
1134
1135 /*
1136 * If there is next token, then it may be pseudo
1137 * netif's address or netmask len for tap(4)
1138 */
1139 next_netif_addr = 0;
1140 } else {
1141 /*
1142 * Not tap(4)'s address, assume it as a bridge(4)
1143 * iface name
1144 */
1145 tap_addr = 0;
1146 ifbridge = tok;
1147
1148 /*
1149 * If there is next token, then it must be pseudo
1150 * netif's address
1151 */
1152 next_netif_addr = 1;
1153 }
1154
1155 netmask = netif_addr = 0;
1156
1157 tok = strtok(NULL, ":/");
1158 if (tok == NULL)
1159 goto back;
1160
1161 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1162 if (next_netif_addr) {
1163 warnx("Invalid pseudo netif address: %s", tok);
1164 return -1;
1165 }
1166 netif_addr = 0;
1167
1168 /*
1169 * Current token is not address, then it must be netmask len
1170 */
1171 masklen_str = tok;
1172 } else {
1173 /*
1174 * Current token is pseudo netif address, if there is next token
1175 * it must be netmask len
1176 */
1177 masklen_str = strtok(NULL, "/");
1178 }
1179
1180 /* Calculate netmask */
1181 if (masklen_str != NULL) {
1182 u_long masklen;
1183
1184 masklen = strtoul(masklen_str, NULL, 10);
1185 if (masklen < 32 && masklen > 0) {
1186 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1187 & 0xffffffff);
1188 } else {
1189 warnx("Invalid netmask len: %lu", masklen);
1190 return -1;
1191 }
1192 }
1193
1194 /* Make sure there is no more token left */
1195 if (strtok(NULL, ":/") != NULL) {
1196 warnx("Invalid argument to '-I'");
1197 return -1;
1198 }
1199
1200back:
2c4a81d7
MD
1201 if (tap_unit < 0) {
1202 /* Do nothing */
1203 } else if (ifbridge == NULL) {
d53b52ae
SZ
1204 /* Set tap(4) address/netmask */
1205 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1206 return -1;
1207 } else {
1208 /* Tie tap(4) to bridge(4) */
1209 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1210 return -1;
1211 }
1212
1213 *addr = netif_addr;
1214 *mask = netmask;
1215 return 0;
1216}
1217
1218/*
1219 * NetifInfo[] will be filled for pseudo netif initialization.
1220 * NetifNum will be bumped to reflect the number of valid entries
1221 * in NetifInfo[].
1222 */
1223static
1224void
1225init_netif(char *netifExp[], int netifExpNum)
1226{
1227 int i, s;
1228
1229 if (netifExpNum == 0)
1230 return;
1231
1232 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1233 if (s < 0)
1234 return;
1235
1236 for (i = 0; i < netifExpNum; ++i) {
1237 struct vknetif_info *info;
1238 in_addr_t netif_addr, netif_mask;
1239 int tap_fd, tap_unit;
1240 char *netif;
1241
1242 netif = strtok(netifExp[i], ":");
1243 if (netif == NULL) {
ee63ee00 1244 warnx("Invalid argument to '-I'");
d53b52ae
SZ
1245 continue;
1246 }
1247
1248 /*
1249 * Open tap(4) device file and bring up the
1250 * corresponding interface
1251 */
1252 tap_fd = netif_open_tap(netif, &tap_unit, s);
1253 if (tap_fd < 0)
1254 continue;
1255
1256 /*
1257 * Initialize tap(4) and get address/netmask
1258 * for pseudo netif
1259 *
1260 * NB: Rest part of netifExp[i] is passed
1261 * to netif_init_tap() implicitly.
1262 */
1263 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1264 /*
1265 * NB: Closing tap(4) device file will bring
1266 * down the corresponding interface
1267 */
1268 close(tap_fd);
1269 continue;
1270 }
1271
1272 info = &NetifInfo[NetifNum];
1273 info->tap_fd = tap_fd;
1274 info->tap_unit = tap_unit;
1275 info->netif_addr = netif_addr;
1276 info->netif_mask = netif_mask;
1277
1278 NetifNum++;
1279 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1280 break;
1281 }
1282 close(s);
d869938c
SW
1283}
1284
bc3cc25e
MD
1285/*
1286 * Create the pid file and leave it open and locked while the vkernel is
1287 * running. This allows a script to use /usr/bin/lockf to probe whether
1288 * a vkernel is still running (so as not to accidently kill an unrelated
1289 * process from a stale pid file).
1290 */
d869938c
SW
1291static
1292void
bc3cc25e 1293writepid(void)
b02b4170 1294{
bc3cc25e
MD
1295 char buf[32];
1296 int fd;
b02b4170
MD
1297
1298 if (pid_file != NULL) {
bc3cc25e
MD
1299 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1300 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1301 if (fd < 0) {
1302 if (errno == EWOULDBLOCK) {
1303 perror("Failed to lock pidfile, "
1304 "vkernel already running");
1305 } else {
1306 perror("Failed to create pidfile");
1307 }
1308 exit(EX_SOFTWARE);
b02b4170 1309 }
bc3cc25e
MD
1310 ftruncate(fd, 0);
1311 write(fd, buf, strlen(buf));
1312 /* leave the file open to maintain the lock */
b02b4170
MD
1313 }
1314}
1315
1316static
1317void
1318cleanpid( void )
1319{
1320 if (pid_file != NULL) {
bc3cc25e 1321 if (unlink(pid_file) < 0)
b02b4170
MD
1322 perror("Warning: couldn't remove pidfile");
1323 }
1324}
1325
1326static
1327void
278c4e67 1328usage_err(const char *ctl, ...)
9c059ae3 1329{
057f0718
MD
1330 va_list va;
1331
1332 va_start(va, ctl);
1333 vfprintf(stderr, ctl, va);
1334 va_end(va);
1335 fprintf(stderr, "\n");
278c4e67
SG
1336 exit(EX_USAGE);
1337}
1338
1339static
1340void
1341usage_help(_Bool help)
1342{
1343 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1344 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
9bea6114
MC
1345 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n"
1346 "\t[-p file] [-r file]\n", save_av[0]);
278c4e67
SG
1347
1348 if (help)
1349 fprintf(stderr, "\nArguments:\n"
1350 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1351 "\t-e\tSpecify an environment to be used by the kernel.\n"
1352 "\t-h\tThis list of options.\n"
1353 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1354 "\t-I\tCreate a virtual network device.\n"
1355 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1356 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
9bea6114
MC
1357 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n"
1358 "\t \t- numcpus - number of cpus\n"
1359 "\t \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1360 "\t \tthe logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n"
1361 "\t \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1362 "\t \tthe core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n"
278c4e67
SG
1363 "\t-p\tSpecify a file in which to store the process ID.\n"
1364 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1365 "\t-s\tBoot into single-user mode.\n"
1366 "\t-U\tEnable writing to kernel memory and module loading.\n"
1367 "\t-v\tTurn on verbose booting.\n");
1368
1369 exit(EX_USAGE);
9c059ae3
MD
1370}
1371
8a9d8995
MD
1372void
1373cpu_reset(void)
1374{
40d0276c
MD
1375 kprintf("cpu reset, rebooting vkernel\n");
1376 closefrom(3);
b02b4170 1377 cleanpid();
40d0276c 1378 execv(save_av[0], save_av);
8a9d8995 1379}
6f7b98e0
MD
1380
1381void
1382cpu_halt(void)
1383{
40d0276c 1384 kprintf("cpu halt, exiting vkernel\n");
b02b4170 1385 cleanpid();
278c4e67 1386 exit(EX_OK);
6f7b98e0 1387}
702acf06
JT
1388
1389void
1390setrealcpu(void)
1391{
1392 switch(lwp_cpu_lock) {
1393 case LCL_PER_CPU:
1394 if (bootverbose)
1395 kprintf("Locking CPU%d to real cpu %d\n",
1396 mycpuid, next_cpu);
1397 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1398 next_cpu++;
1399 if (next_cpu >= real_ncpus)
1400 next_cpu = 0;
1401 break;
1402 case LCL_SINGLE_CPU:
1403 if (bootverbose)
1404 kprintf("Locking CPU%d to real cpu %d\n",
1405 mycpuid, next_cpu);
1406 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1407 break;
1408 default:
1409 /* do not map virtual cpus to real cpus */
1410 break;
1411 }
1412}
1413