kernel: Make SMP support default (and non-optional).
[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.
9c059ae3
MD
33 */
34
44a09dd6
MD
35#include <sys/types.h>
36#include <sys/systm.h>
e4a473f1 37#include <sys/kernel.h>
44a09dd6
MD
38#include <sys/stat.h>
39#include <sys/mman.h>
e4a473f1
MD
40#include <sys/cons.h>
41#include <sys/random.h>
42#include <sys/vkernel.h>
8a9d8995 43#include <sys/tls.h>
884d6eea 44#include <sys/reboot.h>
e4a473f1 45#include <sys/proc.h>
6f7b98e0 46#include <sys/msgbuf.h>
6efe4829 47#include <sys/vmspace.h>
d53b52ae
SZ
48#include <sys/socket.h>
49#include <sys/sockio.h>
702acf06 50#include <sys/sysctl.h>
91be174d 51#include <sys/un.h>
44a09dd6 52#include <vm/vm_page.h>
e130cdcb 53#include <sys/mplock2.h>
44a09dd6 54
702acf06 55#include <machine/cpu.h>
8a9d8995 56#include <machine/globaldata.h>
e4a473f1
MD
57#include <machine/tls.h>
58#include <machine/md_var.h>
59#include <machine/vmparam.h>
870b0161 60#include <cpu/specialreg.h>
8a9d8995 61
d53b52ae
SZ
62#include <net/if.h>
63#include <net/if_arp.h>
64#include <net/ethernet.h>
65#include <net/bridge/if_bridgevar.h>
66#include <netinet/in.h>
67#include <arpa/inet.h>
68
9c059ae3 69#include <stdio.h>
44a09dd6 70#include <stdlib.h>
057f0718 71#include <stdarg.h>
278c4e67 72#include <stdbool.h>
44a09dd6
MD
73#include <unistd.h>
74#include <fcntl.h>
75#include <string.h>
76#include <err.h>
77#include <errno.h>
8a9d8995 78#include <assert.h>
278c4e67 79#include <sysexits.h>
9c059ae3 80
e4a473f1 81vm_paddr_t phys_avail[16];
9c059ae3 82vm_paddr_t Maxmem;
71152ac6 83vm_paddr_t Maxmem_bytes;
39d69dae 84long physmem;
9c059ae3 85int MemImageFd = -1;
ee63ee00
SW
86struct vkdisk_info DiskInfo[VKDISK_MAX];
87int DiskNum;
d53b52ae
SZ
88struct vknetif_info NetifInfo[VKNETIF_MAX];
89int NetifNum;
b02b4170 90char *pid_file;
e4a473f1
MD
91vm_offset_t KvaStart;
92vm_offset_t KvaEnd;
9c059ae3 93vm_offset_t KvaSize;
e4a473f1
MD
94vm_offset_t virtual_start;
95vm_offset_t virtual_end;
791c6551
MD
96vm_offset_t virtual2_start;
97vm_offset_t virtual2_end;
e4a473f1
MD
98vm_offset_t kernel_vm_end;
99vm_offset_t crashdumpmap;
6f7b98e0
MD
100vm_offset_t clean_sva;
101vm_offset_t clean_eva;
102struct msgbuf *msgbufp;
103caddr_t ptvmmap;
e4a473f1 104vpte_t *KernelPTD;
71152ac6 105vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
6f7b98e0 106u_int cpu_feature; /* XXX */
870b0161
MD
107int tsc_present;
108int64_t tsc_frequency;
c5b0b0ba 109int optcpus; /* number of cpus - see mp_start() */
702acf06
JT
110int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
111int real_ncpus; /* number of real CPUs */
112int next_cpu; /* next real CPU to lock a virtual CPU to */
9bea6114
MC
113int vkernel_b_arg; /* -b argument - no of logical CPU bits - only SMP */
114int vkernel_B_arg; /* -B argument - no of core bits - only SMP */
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);
0f85165c 135static void init_locks(void);
9c059ae3 136
40d0276c
MD
137static int save_ac;
138static char **save_av;
139
9c059ae3
MD
140/*
141 * Kernel startup for virtual kernels - standard main()
142 */
143int
144main(int ac, char **av)
145{
44a09dd6 146 char *memImageFile = NULL;
d53b52ae 147 char *netifFile[VKNETIF_MAX];
ee63ee00 148 char *diskFile[VKDISK_MAX];
a72d8a9f 149 char *cdFile[VKDISK_MAX];
44a09dd6 150 char *suffix;
702acf06 151 char *endp;
974f8ffe 152 char *tmp;
9bea6114 153 char *tok;
d53b52ae 154 int netifFileNum = 0;
ee63ee00 155 int diskFileNum = 0;
a72d8a9f 156 int cdFileNum = 0;
41556a7a 157 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
44a09dd6 158 int c;
aaf8b91f 159 int i;
a6d77ab5 160 int j;
aaf8b91f 161 int n;
a6d77ab5 162 int isq;
974f8ffe
AHJ
163 int pos;
164 int eflag;
dc7503af 165 int real_vkernel_enable;
aecc8383 166 int supports_sse;
870b0161 167 size_t vsize;
974f8ffe
AHJ
168 size_t kenv_size;
169 size_t kenv_size2;
b02b4170 170
40d0276c
MD
171 save_ac = ac;
172 save_av = av;
974f8ffe
AHJ
173 eflag = 0;
174 pos = 0;
175 kenv_size = 0;
40d0276c 176
9c059ae3
MD
177 /*
178 * Process options
179 */
7e42c007 180 kernel_mem_readonly = 1;
c5b0b0ba 181 optcpus = 2;
9bea6114
MC
182 vkernel_b_arg = 0;
183 vkernel_B_arg = 0;
702acf06 184 lwp_cpu_lock = LCL_NONE;
7e42c007 185
27c3aec1 186 real_vkernel_enable = 0;
870b0161
MD
187 vsize = sizeof(real_vkernel_enable);
188 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
dc7503af 189
27c3aec1
MD
190 if (real_vkernel_enable == 0) {
191 errx(1, "vm.vkernel_enable is 0, must be set "
192 "to 1 to execute a vkernel!");
193 }
dc7503af 194
27c3aec1 195 real_ncpus = 1;
870b0161
MD
196 vsize = sizeof(real_ncpus);
197 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
702acf06 198
278c4e67
SG
199 if (ac < 2)
200 usage_help(false);
201
9bea6114 202 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:Ub:B:")) != -1) {
9c059ae3 203 switch(c) {
aaf8b91f
MD
204 case 'e':
205 /*
206 * name=value:name=value:name=value...
a6d77ab5
MD
207 * name="value"...
208 *
209 * Allow values to be quoted but note that shells
210 * may remove the quotes, so using this feature
211 * to embed colons may require a backslash.
aaf8b91f
MD
212 */
213 n = strlen(optarg);
a6d77ab5 214 isq = 0;
974f8ffe
AHJ
215
216 if (eflag == 0) {
217 kenv_size = n + 2;
218 kern_envp = malloc(kenv_size);
219 if (kern_envp == NULL)
220 errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size);
221 } else {
222 kenv_size2 = kenv_size + n + 1;
223 pos = kenv_size - 1;
224 if ((tmp = realloc(kern_envp, kenv_size2)) == NULL)
225 errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2);
226 kern_envp = tmp;
227 kenv_size = kenv_size2;
228 }
229
230 for (i = 0, j = pos; i < n; ++i) {
a6d77ab5
MD
231 if (optarg[i] == '"')
232 isq ^= 1;
233 else if (optarg[i] == '\'')
234 isq ^= 2;
235 else if (isq == 0 && optarg[i] == ':')
236 kern_envp[j++] = 0;
aaf8b91f 237 else
a6d77ab5 238 kern_envp[j++] = optarg[i];
aaf8b91f 239 }
a6d77ab5
MD
240 kern_envp[j++] = 0;
241 kern_envp[j++] = 0;
974f8ffe 242 eflag++;
aaf8b91f 243 break;
884d6eea
MD
244 case 's':
245 boothowto |= RB_SINGLE;
246 break;
9c059ae3
MD
247 case 'v':
248 bootverbose = 1;
249 break;
250 case 'i':
251 memImageFile = optarg;
252 break;
d869938c 253 case 'I':
d53b52ae 254 if (netifFileNum < VKNETIF_MAX)
40d0276c 255 netifFile[netifFileNum++] = strdup(optarg);
d869938c 256 break;
9c059ae3 257 case 'r':
41556a7a
MD
258 if (bootOnDisk < 0)
259 bootOnDisk = 1;
a72d8a9f 260 if (diskFileNum + cdFileNum < VKDISK_MAX)
40d0276c 261 diskFile[diskFileNum++] = strdup(optarg);
9c059ae3 262 break;
a72d8a9f 263 case 'c':
41556a7a
MD
264 if (bootOnDisk < 0)
265 bootOnDisk = 0;
a72d8a9f 266 if (diskFileNum + cdFileNum < VKDISK_MAX)
40d0276c 267 cdFile[cdFileNum++] = strdup(optarg);
a72d8a9f 268 break;
9c059ae3 269 case 'm':
71152ac6 270 Maxmem_bytes = strtoull(optarg, &suffix, 0);
9c059ae3
MD
271 if (suffix) {
272 switch(*suffix) {
273 case 'g':
274 case 'G':
71152ac6 275 Maxmem_bytes <<= 30;
9c059ae3
MD
276 break;
277 case 'm':
278 case 'M':
71152ac6 279 Maxmem_bytes <<= 20;
9c059ae3
MD
280 break;
281 case 'k':
282 case 'K':
71152ac6 283 Maxmem_bytes <<= 10;
9c059ae3
MD
284 break;
285 default:
71152ac6 286 Maxmem_bytes = 0;
278c4e67 287 usage_err("Bad maxmem option");
9c059ae3
MD
288 /* NOT REACHED */
289 break;
290 }
291 }
292 break;
702acf06
JT
293 case 'l':
294 next_cpu = -1;
295 if (strncmp("map", optarg, 3) == 0) {
296 lwp_cpu_lock = LCL_PER_CPU;
92d2f80a 297 if (optarg[3] == ',') {
702acf06 298 next_cpu = strtol(optarg+4, &endp, 0);
92d2f80a 299 if (*endp != '\0')
278c4e67 300 usage_err("Bad target CPU number at '%s'", endp);
92d2f80a
MD
301 } else {
302 next_cpu = 0;
303 }
702acf06 304 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 305 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
702acf06
JT
306 } else if (strncmp("any", optarg, 3) == 0) {
307 lwp_cpu_lock = LCL_NONE;
308 } else {
309 lwp_cpu_lock = LCL_SINGLE_CPU;
310 next_cpu = strtol(optarg, &endp, 0);
311 if (*endp != '\0')
278c4e67 312 usage_err("Bad target CPU number at '%s'", endp);
702acf06 313 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 314 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
702acf06
JT
315 }
316 break;
c5b0b0ba
MD
317 case 'n':
318 /*
319 * This value is set up by mp_start(), don't just
320 * set ncpus here.
321 */
9bea6114 322 tok = strtok(optarg, ":");
9bea6114 323 optcpus = strtol(tok, NULL, 0);
057f0718 324 if (optcpus < 1 || optcpus > MAXCPU)
278c4e67 325 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
9bea6114
MC
326
327 /* :core_bits argument */
328 tok = strtok(NULL, ":");
329 if (tok != NULL) {
330 vkernel_b_arg = strtol(tok, NULL, 0);
331
332 /* :logical_CPU_bits argument */
333 tok = strtok(NULL, ":");
334 if (tok != NULL) {
335 vkernel_B_arg = strtol(tok, NULL, 0);
336 }
337
338 }
151f45bf 339 break;
b02b4170 340 case 'p':
9bea6114 341 pid_file = optarg;
b02b4170 342 break;
7e42c007
MD
343 case 'U':
344 kernel_mem_readonly = 0;
345 break;
278c4e67
SG
346 case 'h':
347 usage_help(true);
348 break;
278c4e67
SG
349 default:
350 usage_help(false);
9c059ae3
MD
351 }
352 }
353
b02b4170 354 writepid();
b1b21fa6 355 cpu_disable_intr();
9c059ae3
MD
356 init_sys_memory(memImageFile);
357 init_kern_memory();
8a9d8995 358 init_globaldata();
e4a473f1 359 init_vkernel();
702acf06 360 setrealcpu();
f991237f 361 init_kqueue();
41556a7a 362
08771751
VS
363 vmm_guest = 1;
364
870b0161
MD
365 /*
366 * Check TSC
367 */
368 vsize = sizeof(tsc_present);
369 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
370 vsize = sizeof(tsc_frequency);
371 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
372 if (tsc_present)
373 cpu_feature |= CPUID_TSC;
374
375 /*
376 * Check SSE
377 */
378 vsize = sizeof(supports_sse);
27c3aec1 379 supports_sse = 0;
870b0161 380 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
aecc8383 381 init_fpu(supports_sse);
870b0161
MD
382 if (supports_sse)
383 cpu_feature |= CPUID_SSE | CPUID_FXSR;
aecc8383 384
41556a7a
MD
385 /*
386 * We boot from the first installed disk.
387 */
388 if (bootOnDisk == 1) {
389 init_disk(diskFile, diskFileNum, VKD_DISK);
390 init_disk(cdFile, cdFileNum, VKD_CD);
391 } else {
392 init_disk(cdFile, cdFileNum, VKD_CD);
393 init_disk(diskFile, diskFileNum, VKD_DISK);
394 }
d53b52ae 395 init_netif(netifFile, netifFileNum);
6092278a 396 init_exceptions();
9c059ae3
MD
397 mi_startup();
398 /* NOT REACHED */
278c4e67 399 exit(EX_SOFTWARE);
9c059ae3
MD
400}
401
402/*
403 * Initialize system memory. This is the virtual kernel's 'RAM'.
404 */
405static
406void
44a09dd6 407init_sys_memory(char *imageFile)
9c059ae3 408{
44a09dd6 409 struct stat st;
3748859a 410 int i;
9c059ae3
MD
411 int fd;
412
413 /*
414 * Figure out the system memory image size. If an image file was
415 * specified and -m was not specified, use the image file's size.
416 */
71152ac6
MD
417 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
418 Maxmem_bytes = (vm_paddr_t)st.st_size;
419 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
420 Maxmem_bytes == 0) {
376264e3 421 errx(1, "Cannot create new memory file %s unless "
9c059ae3
MD
422 "system memory size is specified with -m",
423 imageFile);
424 /* NOT REACHED */
425 }
426
427 /*
428 * Maxmem must be known at this time
429 */
71152ac6 430 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
376264e3 431 errx(1, "Bad maxmem specification: 32MB minimum, "
9c059ae3
MD
432 "multiples of %dMB only",
433 SEG_SIZE / 1024 / 1024);
434 /* NOT REACHED */
435 }
436
437 /*
438 * Generate an image file name if necessary, then open/create the
439 * file exclusively locked. Do not allow multiple virtual kernels
440 * to use the same image file.
bc3cc25e
MD
441 *
442 * Don't iterate through a million files if we do not have write
443 * access to the directory, stop if our open() failed on a
444 * non-existant file. Otherwise opens can fail for any number
445 * of reasons (lock failed, file not owned or writable by us, etc).
9c059ae3 446 */
3748859a
MD
447 if (imageFile == NULL) {
448 for (i = 0; i < 1000000; ++i) {
0171c06b 449 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
3748859a
MD
450 fd = open(imageFile,
451 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
bc3cc25e 452 if (fd < 0 && stat(imageFile, &st) == 0) {
3748859a
MD
453 free(imageFile);
454 continue;
455 }
456 break;
457 }
458 } else {
459 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
460 }
0171c06b 461 printf("Using memory file: %s\n", imageFile);
9c059ae3 462 if (fd < 0 || fstat(fd, &st) < 0) {
5de37c8b 463 err(1, "Unable to open/create %s", imageFile);
9c059ae3
MD
464 /* NOT REACHED */
465 }
466
467 /*
b72e51d1
MD
468 * Truncate or extend the file as necessary. Clean out the contents
469 * of the file, we want it to be full of holes so we don't waste
470 * time reading in data from an old file that we no longer care
471 * about.
9c059ae3 472 */
b72e51d1
MD
473 ftruncate(fd, 0);
474 ftruncate(fd, Maxmem_bytes);
9c059ae3 475
9c059ae3 476 MemImageFd = fd;
71152ac6 477 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
9f90716f 478 physmem = Maxmem;
9c059ae3
MD
479}
480
0f85165c
AHJ
481/*
482 * Initialize pool tokens and other necessary locks
483 */
484static void
485init_locks(void)
486{
487
0f85165c
AHJ
488 /*
489 * Get the initial mplock with a count of 1 for the BSP.
490 * This uses a LOGICAL cpu ID, ie BSP == 0.
491 */
492 cpu_get_initial_mplock();
0f85165c
AHJ
493
494 /* our token pool needs to work early */
495 lwkt_token_pool_init();
496
497}
498
9c059ae3
MD
499/*
500 * Initialize kernel memory. This reserves kernel virtual memory by using
501 * MAP_VPAGETABLE
502 */
0c4c980f 503
9c059ae3
MD
504static
505void
506init_kern_memory(void)
507{
508 void *base;
0c4c980f 509 void *try;
3748859a 510 char *zero;
0c4c980f
MD
511 char dummy;
512 char *topofstack = &dummy;
e4a473f1
MD
513 vpte_t pte;
514 int i;
9c059ae3
MD
515
516 /*
517 * Memory map our kernel virtual memory space. Note that the
518 * kernel image itself is not made part of this memory for the
519 * moment.
71152ac6
MD
520 *
521 * The memory map must be segment-aligned so we can properly
522 * offset KernelPTD.
0c4c980f
MD
523 *
524 * If the system kernel has a different MAXDSIZ, it might not
525 * be possible to map kernel memory in its prefered location.
526 * Try a number of different locations.
9c059ae3 527 */
0c4c980f
MD
528 try = (void *)0x40000000;
529 base = NULL;
530 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
531 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
532 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
533 MemImageFd, 0);
534 if (base == try)
535 break;
536 if (base != MAP_FAILED)
537 munmap(base, KERNEL_KVA_SIZE);
538 try = (char *)try + 0x10000000;
539 }
540 if (base != try) {
9c059ae3
MD
541 err(1, "Unable to mmap() kernel virtual memory!");
542 /* NOT REACHED */
543 }
a3c35df6 544 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
e4a473f1 545 KvaStart = (vm_offset_t)base;
9c059ae3 546 KvaSize = KERNEL_KVA_SIZE;
e4a473f1 547 KvaEnd = KvaStart + KvaSize;
d4f5f27c
MD
548
549 /* cannot use kprintf yet */
0c4c980f 550 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
e4a473f1
MD
551
552 /*
553 * Create a top-level page table self-mapping itself.
554 *
555 * Initialize the page directory at physical page index 0 to point
556 * to an array of page table pages starting at physical page index 1
557 */
558 lseek(MemImageFd, 0L, 0);
559 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
560 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
561 write(MemImageFd, &pte, sizeof(pte));
562 }
563
564 /*
565 * Initialize the PTEs in the page table pages required to map the
566 * page table itself. This includes mapping the page directory page
567 * at the base so we go one more loop then normal.
568 */
569 lseek(MemImageFd, PAGE_SIZE, 0);
71152ac6 570 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
e4a473f1
MD
571 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
572 write(MemImageFd, &pte, sizeof(pte));
573 }
574
3748859a
MD
575 /*
576 * Initialize remaining PTEs to 0. We may be reusing a memory image
577 * file. This is approximately a megabyte.
578 */
579 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
580 zero = malloc(PAGE_SIZE);
b87d6bd2 581 bzero(zero, PAGE_SIZE);
3748859a
MD
582 while (i) {
583 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
584 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
585 }
586 free(zero);
587
e4a473f1
MD
588 /*
589 * Enable the page table and calculate pointers to our self-map
590 * for easy kernel page table manipulation.
71152ac6
MD
591 *
592 * KernelPTA must be offset so we can do direct VA translations
e4a473f1 593 */
71152ac6
MD
594 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
595 0 | VPTE_R | VPTE_W | VPTE_V);
e4a473f1
MD
596 KernelPTD = (vpte_t *)base; /* pg directory */
597 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
71152ac6 598 KernelPTA -= KvaStart >> PAGE_SHIFT;
e4a473f1
MD
599
600 /*
601 * phys_avail[] represents unallocated physical memory. MI code
602 * will use phys_avail[] to create the vm_page array.
603 */
604 phys_avail[0] = PAGE_SIZE +
605 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
606 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
71152ac6 607 phys_avail[1] = Maxmem_bytes;
e4a473f1
MD
608
609 /*
610 * (virtual_start, virtual_end) represent unallocated kernel virtual
611 * memory. MI code will create kernel_map using these parameters.
612 */
613 virtual_start = KvaStart + PAGE_SIZE +
614 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
615 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
616 virtual_end = KvaStart + KERNEL_KVA_SIZE;
617
618 /*
00835518
MD
619 * kernel_vm_end could be set to virtual_end but we want some
620 * indication of how much of the kernel_map we've used, so
621 * set it low and let pmap_growkernel increase it even though we
622 * don't need to create any new page table pages.
e4a473f1 623 */
00835518 624 kernel_vm_end = virtual_start;
e4a473f1
MD
625
626 /*
627 * Allocate space for process 0's UAREA.
628 */
629 proc0paddr = (void *)virtual_start;
630 for (i = 0; i < UPAGES; ++i) {
631 pmap_kenter_quick(virtual_start, phys_avail[0]);
632 virtual_start += PAGE_SIZE;
633 phys_avail[0] += PAGE_SIZE;
634 }
9c059ae3
MD
635
636 /*
e4a473f1 637 * crashdumpmap
9c059ae3 638 */
e4a473f1
MD
639 crashdumpmap = virtual_start;
640 virtual_start += MAXDUMPPGS * PAGE_SIZE;
641
6f7b98e0
MD
642 /*
643 * msgbufp maps the system message buffer
644 */
645 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
646 msgbufp = (void *)virtual_start;
647 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
648 pmap_kenter_quick(virtual_start, phys_avail[0]);
649 virtual_start += PAGE_SIZE;
650 phys_avail[0] += PAGE_SIZE;
651 }
652 msgbufinit(msgbufp, MSGBUF_SIZE);
653
654 /*
655 * used by kern_memio for /dev/mem access
656 */
657 ptvmmap = (caddr_t)virtual_start;
658 virtual_start += PAGE_SIZE;
659
e4a473f1
MD
660 /*
661 * Bootstrap the kernel_pmap
662 */
663 pmap_bootstrap();
9c059ae3
MD
664}
665
8a9d8995 666/*
e4a473f1
MD
667 * Map the per-cpu globaldata for cpu #0. Allocate the space using
668 * virtual_start and phys_avail[0]
8a9d8995
MD
669 */
670static
671void
672init_globaldata(void)
673{
e4a473f1
MD
674 int i;
675 vm_paddr_t pa;
676 vm_offset_t va;
677
678 /*
679 * Reserve enough KVA to cover possible cpus. This is a considerable
680 * amount of KVA since the privatespace structure includes two
681 * whole page table mappings.
682 */
683 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
684 CPU_prvspace = (void *)virtual_start;
685 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
686
687 /*
688 * Allocate enough physical memory to cover the mdglobaldata
689 * portion of the space and the idle stack and map the pages
690 * into KVA. For cpu #0 only.
691 */
692 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
693 pa = phys_avail[0];
694 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
695 pmap_kenter_quick(va, pa);
696 phys_avail[0] += PAGE_SIZE;
697 }
698 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
699 pa = phys_avail[0];
700 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
701 pmap_kenter_quick(va, pa);
702 phys_avail[0] += PAGE_SIZE;
703 }
704
705 /*
27c3aec1
MD
706 * Setup the %fs for cpu #0. The mycpu macro works after this
707 * point. Note that %gs is used by pthreads.
e4a473f1 708 */
9ad680a3 709 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
e4a473f1
MD
710}
711
712/*
713 * Initialize very low level systems including thread0, proc0, etc.
714 */
715static
716void
717init_vkernel(void)
718{
719 struct mdglobaldata *gd;
720
721 gd = &CPU_prvspace[0].mdglobaldata;
722 bzero(gd, sizeof(*gd));
723
724 gd->mi.gd_curthread = &thread0;
71152ac6 725 thread0.td_gd = &gd->mi;
e4a473f1 726 ncpus = 1;
b45759e1
MD
727 ncpus2 = 1; /* rounded down power of 2 */
728 ncpus_fit = 1; /* rounded up power of 2 */
729 /* ncpus2_mask and ncpus_fit_mask are 0 */
e4a473f1
MD
730 init_param1();
731 gd->mi.gd_prvspace = &CPU_prvspace[0];
732 mi_gdinit(&gd->mi, 0);
733 cpu_gdinit(gd, 0);
734 mi_proc0init(&gd->mi, proc0paddr);
08f2f1bb 735 lwp0.lwp_md.md_regs = &proc0_tf;
252b2227 736
0f85165c 737 init_locks();
e4a473f1
MD
738 cninit();
739 rand_initialize();
740#if 0 /* #ifdef DDB */
741 kdb_init();
742 if (boothowto & RB_KDB)
743 Debugger("Boot flags requested debugger");
744#endif
514621ec 745 identcpu();
e4a473f1
MD
746#if 0
747 initializecpu(); /* Initialize CPU registers */
748#endif
749 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
750
751#if 0
752 /*
753 * Map the message buffer
754 */
755 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
756 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
757 msgbufinit(msgbufp, MSGBUF_SIZE);
758#endif
759#if 0
760 thread0.td_pcb_cr3 ... MMU
08f2f1bb 761 lwp0.lwp_md.md_regs = &proc0_tf;
e4a473f1 762#endif
8a9d8995
MD
763}
764
9c059ae3 765/*
ee63ee00
SW
766 * Filesystem image paths for the virtual kernel are optional.
767 * If specified they each should point to a disk image,
768 * the first of which will become the root disk.
5b85cb7d
MD
769 *
770 * The virtual kernel caches data from our 'disk' just like a normal kernel,
771 * so we do not really want the real kernel to cache the data too. Use
772 * O_DIRECT to remove the duplication.
9c059ae3
MD
773 */
774static
775void
a72d8a9f 776init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
44a09dd6 777{
ee63ee00
SW
778 int i;
779
780 if (diskFileNum == 0)
781 return;
782
783 for(i=0; i < diskFileNum; i++){
784 char *fname;
785 fname = diskExp[i];
786
787 if (fname == NULL) {
788 warnx("Invalid argument to '-r'");
789 continue;
790 }
791
792 if (DiskNum < VKDISK_MAX) {
793 struct stat st;
794 struct vkdisk_info* info = NULL;
795 int fd;
796 size_t l = 0;
797
a72d8a9f 798 if (type == VKD_DISK)
6882b7a5 799 fd = open(fname, O_RDWR|O_DIRECT, 0644);
a72d8a9f
MD
800 else
801 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
ee63ee00
SW
802 if (fd < 0 || fstat(fd, &st) < 0) {
803 err(1, "Unable to open/create %s", fname);
804 /* NOT REACHED */
805 }
6882b7a5
MD
806 if (S_ISREG(st.st_mode)) {
807 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
808 errx(1, "Disk image %s is already "
809 "in use\n", fname);
810 /* NOT REACHED */
811 }
812 }
ee63ee00
SW
813
814 info = &DiskInfo[DiskNum];
815 l = strlen(fname);
0171c06b 816
ee63ee00
SW
817 info->unit = i;
818 info->fd = fd;
a72d8a9f 819 info->type = type;
ee63ee00
SW
820 memcpy(info->fname, fname, l);
821
41556a7a 822 if (DiskNum == 0) {
1f1bc20e 823 if (type == VKD_CD) {
a72d8a9f 824 rootdevnames[0] = "cd9660:vcd0a";
1f1bc20e 825 } else if (type == VKD_DISK) {
28b44c8f 826 rootdevnames[0] = "ufs:vkd0s0a";
1f1bc20e
MD
827 rootdevnames[1] = "ufs:vkd0s1a";
828 }
a72d8a9f 829 }
ee63ee00
SW
830
831 DiskNum++;
832 } else {
23e2971a 833 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
ee63ee00 834 continue;
0171c06b
MD
835 }
836 }
44a09dd6
MD
837}
838
d869938c 839static
d53b52ae
SZ
840int
841netif_set_tapflags(int tap_unit, int f, int s)
842{
843 struct ifreq ifr;
844 int flags;
845
846 bzero(&ifr, sizeof(ifr));
847
848 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
849 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
5de37c8b 850 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
d53b52ae
SZ
851 return -1;
852 }
853
854 /*
855 * Adjust if_flags
856 *
857 * If the flags are already set/cleared, then we return
858 * immediately to avoid extra syscalls
859 */
860 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
861 if (f < 0) {
862 /* Turn off flags */
863 f = -f;
864 if ((flags & f) == 0)
865 return 0;
866 flags &= ~f;
867 } else {
868 /* Turn on flags */
869 if (flags & f)
870 return 0;
871 flags |= f;
872 }
873
874 /*
875 * Fix up ifreq.ifr_name, since it may be trashed
876 * in previous ioctl(SIOCGIFFLAGS)
877 */
878 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
879
880 ifr.ifr_flags = flags & 0xffff;
881 ifr.ifr_flagshigh = flags >> 16;
882 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
5de37c8b 883 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
d53b52ae
SZ
884 return -1;
885 }
886 return 0;
887}
888
889static
890int
891netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
892{
893 struct ifaliasreq ifra;
894 struct sockaddr_in *in;
895
896 bzero(&ifra, sizeof(ifra));
897 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
898
899 /* Setup address */
900 in = (struct sockaddr_in *)&ifra.ifra_addr;
901 in->sin_family = AF_INET;
902 in->sin_len = sizeof(*in);
903 in->sin_addr.s_addr = addr;
904
905 if (mask != 0) {
906 /* Setup netmask */
907 in = (struct sockaddr_in *)&ifra.ifra_mask;
908 in->sin_len = sizeof(*in);
909 in->sin_addr.s_addr = mask;
910 }
911
912 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
5de37c8b 913 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
d53b52ae
SZ
914 return -1;
915 }
916 return 0;
917}
918
919static
920int
921netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
922{
923 struct ifbreq ifbr;
924 struct ifdrv ifd;
925
926 bzero(&ifbr, sizeof(ifbr));
927 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
928 "tap%d", tap_unit);
929
930 bzero(&ifd, sizeof(ifd));
931 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
932 ifd.ifd_cmd = BRDGADD;
933 ifd.ifd_len = sizeof(ifbr);
934 ifd.ifd_data = &ifbr;
935
936 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
937 /*
938 * 'errno == EEXIST' means that the tap(4) is already
939 * a member of the bridge(4)
940 */
941 if (errno != EEXIST) {
5de37c8b 942 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
d53b52ae
SZ
943 return -1;
944 }
945 }
946 return 0;
947}
948
949#define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
950
d53b52ae
SZ
951/*
952 * Locate the first unused tap(4) device file if auto mode is requested,
953 * or open the user supplied device file, and bring up the corresponding
954 * tap(4) interface.
955 *
956 * NOTE: Only tap(4) device file is supported currently
957 */
958static
959int
960netif_open_tap(const char *netif, int *tap_unit, int s)
d869938c 961{
d53b52ae
SZ
962 char tap_dev[MAXPATHLEN];
963 int tap_fd, failed;
964 struct stat st;
235a1432 965 char *dname;
d53b52ae
SZ
966
967 *tap_unit = -1;
968
969 if (strcmp(netif, "auto") == 0) {
d53b52ae
SZ
970 /*
971 * Find first unused tap(4) device file
972 */
235a1432 973 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
d53b52ae 974 if (tap_fd < 0) {
235a1432 975 warnc(errno, "Unable to find a free tap(4)");
d53b52ae
SZ
976 return -1;
977 }
978 } else {
979 /*
91be174d 980 * User supplied tap(4) device file or unix socket.
d53b52ae
SZ
981 */
982 if (netif[0] == '/') /* Absolute path */
983 strlcpy(tap_dev, netif, sizeof(tap_dev));
984 else
985 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
986
987 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
91be174d
MD
988
989 /*
990 * If we cannot open normally try to connect to it.
991 */
992 if (tap_fd < 0)
993 tap_fd = unix_connect(tap_dev);
994
d53b52ae 995 if (tap_fd < 0) {
5de37c8b 996 warn("Unable to open %s", tap_dev);
d53b52ae 997 return -1;
d869938c
SW
998 }
999 }
d53b52ae
SZ
1000
1001 /*
1002 * Check whether the device file is a tap(4)
1003 */
235a1432
MD
1004 if (fstat(tap_fd, &st) < 0) {
1005 failed = 1;
1006 } else if (S_ISCHR(st.st_mode)) {
1007 dname = fdevname(tap_fd);
1008 if (dname)
1009 dname = strstr(dname, "tap");
1010 if (dname) {
1011 /*
1012 * Bring up the corresponding tap(4) interface
1013 */
1014 *tap_unit = strtol(dname + 3, NULL, 10);
1015 printf("TAP UNIT %d\n", *tap_unit);
1016 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
1017 failed = 0;
3b6e5cb0
SW
1018 else
1019 failed = 1;
235a1432
MD
1020 } else {
1021 failed = 1;
1022 }
91be174d
MD
1023 } else if (S_ISSOCK(st.st_mode)) {
1024 /*
1025 * Special socket connection (typically to vknet). We
1026 * do not have to do anything.
1027 */
1028 failed = 0;
d53b52ae 1029 } else {
235a1432 1030 failed = 1;
d53b52ae
SZ
1031 }
1032
1033 if (failed) {
235a1432 1034 warnx("%s is not a tap(4) device or socket", tap_dev);
d53b52ae
SZ
1035 close(tap_fd);
1036 tap_fd = -1;
1037 *tap_unit = -1;
1038 }
1039 return tap_fd;
1040}
1041
91be174d
MD
1042static int
1043unix_connect(const char *path)
1044{
1045 struct sockaddr_un sunx;
1046 int len;
1047 int net_fd;
3a6117bb
MD
1048 int sndbuf = 262144;
1049 struct stat st;
91be174d
MD
1050
1051 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1052 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1053 ++len; /* include nul */
1054 sunx.sun_family = AF_UNIX;
1055 sunx.sun_len = len;
1056
1057 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1058 if (net_fd < 0)
1059 return(-1);
1060 if (connect(net_fd, (void *)&sunx, len) < 0) {
1061 close(net_fd);
1062 return(-1);
1063 }
3a6117bb
MD
1064 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1065 if (fstat(net_fd, &st) == 0)
1066 printf("Network socket buffer: %d bytes\n", st.st_blksize);
91be174d
MD
1067 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1068 return(net_fd);
1069}
1070
d53b52ae
SZ
1071#undef TAPDEV_MAJOR
1072#undef TAPDEV_MINOR
1073#undef TAPDEV_OFLAGS
1074
1075/*
1076 * Following syntax is supported,
1077 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1078 *
1079 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1080 * tap(4)'s netmask len is z
1081 *
1082 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1083 * pseudo netif's address is y.y.y.y
1084 *
1085 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1086 * pseudo netif's address is y.y.y.y
1087 * tap(4) and pseudo netif's netmask len are z
1088 *
1089 * 5) bridgeX tap(4) will be added to bridgeX
1090 *
1091 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1092 * pseudo netif's address is y.y.y.y
1093 *
1094 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1095 * pseudo netif's address is y.y.y.y
1096 * pseudo netif's netmask len is z
1097 */
1098static
1099int
1100netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1101{
1102 in_addr_t tap_addr, netmask, netif_addr;
1103 int next_netif_addr;
1104 char *tok, *masklen_str, *ifbridge;
1105
1106 *addr = 0;
1107 *mask = 0;
1108
1109 tok = strtok(NULL, ":/");
1110 if (tok == NULL) {
1111 /*
1112 * Nothing special, simply use tap(4) as backend
1113 */
1114 return 0;
1115 }
1116
1117 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1118 /*
1119 * tap(4)'s address is supplied
1120 */
1121 ifbridge = NULL;
1122
1123 /*
1124 * If there is next token, then it may be pseudo
1125 * netif's address or netmask len for tap(4)
1126 */
1127 next_netif_addr = 0;
1128 } else {
1129 /*
1130 * Not tap(4)'s address, assume it as a bridge(4)
1131 * iface name
1132 */
1133 tap_addr = 0;
1134 ifbridge = tok;
1135
1136 /*
1137 * If there is next token, then it must be pseudo
1138 * netif's address
1139 */
1140 next_netif_addr = 1;
1141 }
1142
1143 netmask = netif_addr = 0;
1144
1145 tok = strtok(NULL, ":/");
1146 if (tok == NULL)
1147 goto back;
1148
1149 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1150 if (next_netif_addr) {
1151 warnx("Invalid pseudo netif address: %s", tok);
1152 return -1;
1153 }
1154 netif_addr = 0;
1155
1156 /*
1157 * Current token is not address, then it must be netmask len
1158 */
1159 masklen_str = tok;
1160 } else {
1161 /*
1162 * Current token is pseudo netif address, if there is next token
1163 * it must be netmask len
1164 */
1165 masklen_str = strtok(NULL, "/");
1166 }
1167
1168 /* Calculate netmask */
1169 if (masklen_str != NULL) {
1170 u_long masklen;
1171
1172 masklen = strtoul(masklen_str, NULL, 10);
1173 if (masklen < 32 && masklen > 0) {
1174 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1175 & 0xffffffff);
1176 } else {
1177 warnx("Invalid netmask len: %lu", masklen);
1178 return -1;
1179 }
1180 }
1181
1182 /* Make sure there is no more token left */
1183 if (strtok(NULL, ":/") != NULL) {
1184 warnx("Invalid argument to '-I'");
1185 return -1;
1186 }
1187
1188back:
2c4a81d7
MD
1189 if (tap_unit < 0) {
1190 /* Do nothing */
1191 } else if (ifbridge == NULL) {
d53b52ae
SZ
1192 /* Set tap(4) address/netmask */
1193 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1194 return -1;
1195 } else {
1196 /* Tie tap(4) to bridge(4) */
1197 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1198 return -1;
1199 }
1200
1201 *addr = netif_addr;
1202 *mask = netmask;
1203 return 0;
1204}
1205
1206/*
1207 * NetifInfo[] will be filled for pseudo netif initialization.
1208 * NetifNum will be bumped to reflect the number of valid entries
1209 * in NetifInfo[].
1210 */
1211static
1212void
1213init_netif(char *netifExp[], int netifExpNum)
1214{
1215 int i, s;
1216
1217 if (netifExpNum == 0)
1218 return;
1219
1220 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1221 if (s < 0)
1222 return;
1223
1224 for (i = 0; i < netifExpNum; ++i) {
1225 struct vknetif_info *info;
1226 in_addr_t netif_addr, netif_mask;
1227 int tap_fd, tap_unit;
1228 char *netif;
1229
1230 netif = strtok(netifExp[i], ":");
1231 if (netif == NULL) {
ee63ee00 1232 warnx("Invalid argument to '-I'");
d53b52ae
SZ
1233 continue;
1234 }
1235
1236 /*
1237 * Open tap(4) device file and bring up the
1238 * corresponding interface
1239 */
1240 tap_fd = netif_open_tap(netif, &tap_unit, s);
1241 if (tap_fd < 0)
1242 continue;
1243
1244 /*
1245 * Initialize tap(4) and get address/netmask
1246 * for pseudo netif
1247 *
1248 * NB: Rest part of netifExp[i] is passed
1249 * to netif_init_tap() implicitly.
1250 */
1251 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1252 /*
1253 * NB: Closing tap(4) device file will bring
1254 * down the corresponding interface
1255 */
1256 close(tap_fd);
1257 continue;
1258 }
1259
1260 info = &NetifInfo[NetifNum];
1261 info->tap_fd = tap_fd;
1262 info->tap_unit = tap_unit;
1263 info->netif_addr = netif_addr;
1264 info->netif_mask = netif_mask;
1265
1266 NetifNum++;
1267 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1268 break;
1269 }
1270 close(s);
d869938c
SW
1271}
1272
bc3cc25e
MD
1273/*
1274 * Create the pid file and leave it open and locked while the vkernel is
1275 * running. This allows a script to use /usr/bin/lockf to probe whether
1276 * a vkernel is still running (so as not to accidently kill an unrelated
1277 * process from a stale pid file).
1278 */
b02b4170
MD
1279static
1280void
bc3cc25e 1281writepid(void)
b02b4170 1282{
bc3cc25e
MD
1283 char buf[32];
1284 int fd;
b02b4170
MD
1285
1286 if (pid_file != NULL) {
bc3cc25e
MD
1287 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1288 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1289 if (fd < 0) {
1290 if (errno == EWOULDBLOCK) {
1291 perror("Failed to lock pidfile, "
1292 "vkernel already running");
1293 } else {
1294 perror("Failed to create pidfile");
1295 }
1296 exit(EX_SOFTWARE);
b02b4170 1297 }
bc3cc25e
MD
1298 ftruncate(fd, 0);
1299 write(fd, buf, strlen(buf));
1300 /* leave the file open to maintain the lock */
b02b4170
MD
1301 }
1302}
1303
1304static
1305void
1306cleanpid( void )
1307{
1308 if (pid_file != NULL) {
bc3cc25e 1309 if (unlink(pid_file) < 0)
b02b4170
MD
1310 perror("Warning: couldn't remove pidfile");
1311 }
1312}
1313
44a09dd6
MD
1314static
1315void
278c4e67 1316usage_err(const char *ctl, ...)
9c059ae3 1317{
057f0718
MD
1318 va_list va;
1319
1320 va_start(va, ctl);
1321 vfprintf(stderr, ctl, va);
1322 va_end(va);
1323 fprintf(stderr, "\n");
278c4e67
SG
1324 exit(EX_USAGE);
1325}
1326
1327static
1328void
1329usage_help(_Bool help)
1330{
1331 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1332 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
9bea6114
MC
1333 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n"
1334 "\t[-p file] [-r file]\n", save_av[0]);
278c4e67
SG
1335
1336 if (help)
1337 fprintf(stderr, "\nArguments:\n"
1338 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1339 "\t-e\tSpecify an environment to be used by the kernel.\n"
1340 "\t-h\tThis list of options.\n"
1341 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1342 "\t-I\tCreate a virtual network device.\n"
1343 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1344 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
9bea6114
MC
1345 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n"
1346 "\t \t- numcpus - number of cpus\n"
1347 "\t \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1348 "\t \tthe logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n"
1349 "\t \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1350 "\t \tthe core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n"
278c4e67
SG
1351 "\t-p\tSpecify a file in which to store the process ID.\n"
1352 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1353 "\t-s\tBoot into single-user mode.\n"
1354 "\t-U\tEnable writing to kernel memory and module loading.\n"
1355 "\t-v\tTurn on verbose booting.\n");
1356
1357 exit(EX_USAGE);
9c059ae3
MD
1358}
1359
8a9d8995
MD
1360void
1361cpu_reset(void)
1362{
40d0276c
MD
1363 kprintf("cpu reset, rebooting vkernel\n");
1364 closefrom(3);
b02b4170 1365 cleanpid();
40d0276c 1366 execv(save_av[0], save_av);
8a9d8995 1367}
6f7b98e0
MD
1368
1369void
1370cpu_halt(void)
1371{
40d0276c 1372 kprintf("cpu halt, exiting vkernel\n");
b02b4170 1373 cleanpid();
278c4e67 1374 exit(EX_OK);
6f7b98e0 1375}
702acf06
JT
1376
1377void
1378setrealcpu(void)
1379{
1380 switch(lwp_cpu_lock) {
1381 case LCL_PER_CPU:
1382 if (bootverbose)
1383 kprintf("Locking CPU%d to real cpu %d\n",
1384 mycpuid, next_cpu);
1385 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1386 next_cpu++;
1387 if (next_cpu >= real_ncpus)
1388 next_cpu = 0;
1389 break;
1390 case LCL_SINGLE_CPU:
1391 if (bootverbose)
1392 kprintf("Locking CPU%d to real cpu %d\n",
1393 mycpuid, next_cpu);
1394 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1395 break;
1396 default:
1397 /* do not map virtual cpus to real cpus */
1398 break;
1399 }
1400}
1401