vkernel - Add usage()
[dragonfly.git] / sys / platform / vkernel64 / platform / init.c
CommitLineData
da673940
JG
1/*
2 * Copyright (c) 2006 The DragonFly Project. All rights reserved.
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
16 * distribution.
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * $DragonFly: src/sys/platform/vkernel/platform/init.c,v 1.56 2008/05/27 07:48:00 dillon Exp $
35 */
36
37#include <sys/types.h>
38#include <sys/systm.h>
39#include <sys/kernel.h>
40#include <sys/stat.h>
41#include <sys/mman.h>
42#include <sys/cons.h>
43#include <sys/random.h>
44#include <sys/vkernel.h>
45#include <sys/tls.h>
46#include <sys/reboot.h>
47#include <sys/proc.h>
48#include <sys/msgbuf.h>
49#include <sys/vmspace.h>
50#include <sys/socket.h>
51#include <sys/sockio.h>
52#include <sys/sysctl.h>
53#include <sys/un.h>
54#include <vm/vm_page.h>
55
56#include <machine/cpu.h>
57#include <machine/globaldata.h>
58#include <machine/tls.h>
59#include <machine/md_var.h>
60#include <machine/vmparam.h>
61#include <cpu/specialreg.h>
62
63#include <net/if.h>
64#include <net/if_arp.h>
65#include <net/ethernet.h>
66#include <net/bridge/if_bridgevar.h>
67#include <netinet/in.h>
68#include <arpa/inet.h>
69
70#include <stdio.h>
71#include <stdlib.h>
72#include <stdarg.h>
278c4e67 73#include <stdbool.h>
da673940
JG
74#include <unistd.h>
75#include <fcntl.h>
76#include <string.h>
77#include <err.h>
78#include <errno.h>
79#include <assert.h>
278c4e67 80#include <sysexits.h>
da673940
JG
81
82vm_paddr_t phys_avail[16];
83vm_paddr_t Maxmem;
84vm_paddr_t Maxmem_bytes;
9f90716f 85int physmem;
da673940
JG
86int MemImageFd = -1;
87struct vkdisk_info DiskInfo[VKDISK_MAX];
88int DiskNum;
89struct vknetif_info NetifInfo[VKNETIF_MAX];
90int NetifNum;
91char *pid_file;
92vm_offset_t KvaStart;
93vm_offset_t KvaEnd;
94vm_offset_t KvaSize;
95vm_offset_t virtual_start;
96vm_offset_t virtual_end;
0e6594a8
SW
97vm_offset_t virtual2_start;
98vm_offset_t virtual2_end;
da673940
JG
99vm_offset_t kernel_vm_end;
100vm_offset_t crashdumpmap;
101vm_offset_t clean_sva;
102vm_offset_t clean_eva;
103struct msgbuf *msgbufp;
104caddr_t ptvmmap;
105vpte_t *KernelPTD;
106vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
107void *dmap_min_address;
108u_int cpu_feature; /* XXX */
109int tsc_present;
110int64_t tsc_frequency;
111int optcpus; /* number of cpus - see mp_start() */
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 */
115
116struct privatespace *CPU_prvspace;
117
118static struct trapframe proc0_tf;
119static void *proc0paddr;
120
121static void init_sys_memory(char *imageFile);
122static void init_kern_memory(void);
123static void init_globaldata(void);
124static void init_vkernel(void);
125static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
126static void init_netif(char *netifExp[], int netifFileNum);
278c4e67
SG
127static void writepid(void);
128static void cleanpid(void);
da673940 129static int unix_connect(const char *path);
278c4e67
SG
130static void usage_err(const char *ctl, ...);
131static void usage_help(_Bool);
da673940
JG
132
133static int save_ac;
134static char **save_av;
135
136/*
137 * Kernel startup for virtual kernels - standard main()
138 */
139int
140main(int ac, char **av)
141{
142 char *memImageFile = NULL;
143 char *netifFile[VKNETIF_MAX];
144 char *diskFile[VKDISK_MAX];
145 char *cdFile[VKDISK_MAX];
146 char *suffix;
147 char *endp;
148 int netifFileNum = 0;
149 int diskFileNum = 0;
150 int cdFileNum = 0;
151 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
152 int c;
153 int i;
154 int j;
155 int n;
156 int isq;
157 int real_vkernel_enable;
158 int supports_sse;
159 size_t vsize;
160
161 save_ac = ac;
162 save_av = av;
163
164 /*
165 * Process options
166 */
167 kernel_mem_readonly = 1;
168#ifdef SMP
169 optcpus = 2;
170#endif
171 lwp_cpu_lock = LCL_NONE;
172
173 real_vkernel_enable = 0;
174 vsize = sizeof(real_vkernel_enable);
175 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
176
177 if (real_vkernel_enable == 0) {
178 errx(1, "vm.vkernel_enable is 0, must be set "
179 "to 1 to execute a vkernel!");
180 }
181
182 real_ncpus = 1;
183 vsize = sizeof(real_ncpus);
184 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
185
278c4e67
SG
186 if (ac < 2)
187 usage_help(false);
188
189 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:U")) != -1) {
da673940
JG
190 switch(c) {
191 case 'e':
192 /*
193 * name=value:name=value:name=value...
194 * name="value"...
195 *
196 * Allow values to be quoted but note that shells
197 * may remove the quotes, so using this feature
198 * to embed colons may require a backslash.
199 */
200 n = strlen(optarg);
201 isq = 0;
202 kern_envp = malloc(n + 2);
203 for (i = j = 0; i < n; ++i) {
204 if (optarg[i] == '"')
205 isq ^= 1;
206 else if (optarg[i] == '\'')
207 isq ^= 2;
208 else if (isq == 0 && optarg[i] == ':')
209 kern_envp[j++] = 0;
210 else
211 kern_envp[j++] = optarg[i];
212 }
213 kern_envp[j++] = 0;
214 kern_envp[j++] = 0;
215 break;
216 case 's':
217 boothowto |= RB_SINGLE;
218 break;
219 case 'v':
220 bootverbose = 1;
221 break;
222 case 'i':
223 memImageFile = optarg;
224 break;
225 case 'I':
226 if (netifFileNum < VKNETIF_MAX)
227 netifFile[netifFileNum++] = strdup(optarg);
228 break;
229 case 'r':
230 if (bootOnDisk < 0)
231 bootOnDisk = 1;
232 if (diskFileNum + cdFileNum < VKDISK_MAX)
233 diskFile[diskFileNum++] = strdup(optarg);
234 break;
235 case 'c':
236 if (bootOnDisk < 0)
237 bootOnDisk = 0;
238 if (diskFileNum + cdFileNum < VKDISK_MAX)
239 cdFile[cdFileNum++] = strdup(optarg);
240 break;
241 case 'm':
242 Maxmem_bytes = strtoull(optarg, &suffix, 0);
243 if (suffix) {
244 switch(*suffix) {
245 case 'g':
246 case 'G':
247 Maxmem_bytes <<= 30;
248 break;
249 case 'm':
250 case 'M':
251 Maxmem_bytes <<= 20;
252 break;
253 case 'k':
254 case 'K':
255 Maxmem_bytes <<= 10;
256 break;
257 default:
258 Maxmem_bytes = 0;
278c4e67 259 usage_err("Bad maxmem option");
da673940
JG
260 /* NOT REACHED */
261 break;
262 }
263 }
264 break;
265 case 'l':
266 next_cpu = -1;
267 if (strncmp("map", optarg, 3) == 0) {
268 lwp_cpu_lock = LCL_PER_CPU;
269 if (optarg[3] == ',') {
270 next_cpu = strtol(optarg+4, &endp, 0);
271 if (*endp != '\0')
278c4e67 272 usage_err("Bad target CPU number at '%s'", endp);
da673940
JG
273 } else {
274 next_cpu = 0;
275 }
276 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 277 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
da673940
JG
278 } else if (strncmp("any", optarg, 3) == 0) {
279 lwp_cpu_lock = LCL_NONE;
280 } else {
281 lwp_cpu_lock = LCL_SINGLE_CPU;
282 next_cpu = strtol(optarg, &endp, 0);
283 if (*endp != '\0')
278c4e67 284 usage_err("Bad target CPU number at '%s'", endp);
da673940 285 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
278c4e67 286 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
da673940
JG
287 }
288 break;
289 case 'n':
290 /*
291 * This value is set up by mp_start(), don't just
292 * set ncpus here.
293 */
294#ifdef SMP
295 optcpus = strtol(optarg, NULL, 0);
296 if (optcpus < 1 || optcpus > MAXCPU)
278c4e67 297 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
da673940
JG
298#else
299 if (strtol(optarg, NULL, 0) != 1) {
278c4e67 300 usage_err("You built a UP vkernel, only 1 cpu!");
da673940
JG
301 }
302#endif
303
304 break;
305 case 'p':
306 pid_file = optarg;
307 break;
308 case 'U':
309 kernel_mem_readonly = 0;
310 break;
278c4e67
SG
311 case 'h':
312 usage_help(true);
313 break;
314 case '?':
315 default:
316 usage_help(false);
da673940
JG
317 }
318 }
319
320 writepid();
321 cpu_disable_intr();
322 init_sys_memory(memImageFile);
323 init_kern_memory();
324 init_globaldata();
325 init_vkernel();
326 setrealcpu();
327 init_kqueue();
328
329 /*
330 * Check TSC
331 */
332 vsize = sizeof(tsc_present);
333 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
334 vsize = sizeof(tsc_frequency);
335 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
336 if (tsc_present)
337 cpu_feature |= CPUID_TSC;
338
339 /*
340 * Check SSE
341 */
342 vsize = sizeof(supports_sse);
343 supports_sse = 0;
344 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
345 init_fpu(supports_sse);
346 if (supports_sse)
347 cpu_feature |= CPUID_SSE | CPUID_FXSR;
348
349 /*
350 * We boot from the first installed disk.
351 */
352 if (bootOnDisk == 1) {
353 init_disk(diskFile, diskFileNum, VKD_DISK);
354 init_disk(cdFile, cdFileNum, VKD_CD);
355 } else {
356 init_disk(cdFile, cdFileNum, VKD_CD);
357 init_disk(diskFile, diskFileNum, VKD_DISK);
358 }
359 init_netif(netifFile, netifFileNum);
360 init_exceptions();
361 mi_startup();
362 /* NOT REACHED */
278c4e67 363 exit(EX_SOFTWARE);
da673940
JG
364}
365
366/*
367 * Initialize system memory. This is the virtual kernel's 'RAM'.
368 */
369static
370void
371init_sys_memory(char *imageFile)
372{
373 struct stat st;
374 int i;
375 int fd;
376
377 /*
378 * Figure out the system memory image size. If an image file was
379 * specified and -m was not specified, use the image file's size.
380 */
381
382 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
383 Maxmem_bytes = (vm_paddr_t)st.st_size;
384 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
385 Maxmem_bytes == 0) {
386 err(1, "Cannot create new memory file %s unless "
387 "system memory size is specified with -m",
388 imageFile);
389 /* NOT REACHED */
390 }
391
392 /*
393 * Maxmem must be known at this time
394 */
395 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
396 err(1, "Bad maxmem specification: 32MB minimum, "
397 "multiples of %dMB only",
398 SEG_SIZE / 1024 / 1024);
399 /* NOT REACHED */
400 }
401
402 /*
403 * Generate an image file name if necessary, then open/create the
404 * file exclusively locked. Do not allow multiple virtual kernels
405 * to use the same image file.
406 */
407 if (imageFile == NULL) {
408 for (i = 0; i < 1000000; ++i) {
409 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
410 fd = open(imageFile,
411 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
412 if (fd < 0 && errno == EWOULDBLOCK) {
413 free(imageFile);
414 continue;
415 }
416 break;
417 }
418 } else {
419 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
420 }
421 fprintf(stderr, "Using memory file: %s\n", imageFile);
422 if (fd < 0 || fstat(fd, &st) < 0) {
423 err(1, "Unable to open/create %s", imageFile);
424 /* NOT REACHED */
425 }
426
427 /*
b72e51d1
MD
428 * Truncate or extend the file as necessary. Clean out the contents
429 * of the file, we want it to be full of holes so we don't waste
430 * time reading in data from an old file that we no longer care
431 * about.
da673940 432 */
b72e51d1
MD
433 ftruncate(fd, 0);
434 ftruncate(fd, Maxmem_bytes);
da673940 435
da673940
JG
436 MemImageFd = fd;
437 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
9f90716f 438 physmem = Maxmem;
da673940
JG
439}
440
441/*
442 * Initialize kernel memory. This reserves kernel virtual memory by using
443 * MAP_VPAGETABLE
444 */
445
446static
447void
448init_kern_memory(void)
449{
450 void *base;
451 void *try;
da673940
JG
452 char dummy;
453 char *topofstack = &dummy;
da673940
JG
454 int i;
455 void *firstfree;
456
457 /*
458 * Memory map our kernel virtual memory space. Note that the
459 * kernel image itself is not made part of this memory for the
460 * moment.
461 *
462 * The memory map must be segment-aligned so we can properly
463 * offset KernelPTD.
464 *
465 * If the system kernel has a different MAXDSIZ, it might not
466 * be possible to map kernel memory in its prefered location.
467 * Try a number of different locations.
468 */
469 try = (void *)(512UL << 30);
470 base = NULL;
471 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
472 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
473 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
2c2e847c 474 MemImageFd, (off_t)try);
da673940
JG
475 if (base == try)
476 break;
477 if (base != MAP_FAILED)
478 munmap(base, KERNEL_KVA_SIZE);
479 try = (char *)try + (512UL << 30);
480 }
481 if (base != try) {
482 err(1, "Unable to mmap() kernel virtual memory!");
483 /* NOT REACHED */
484 }
485 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
486 KvaStart = (vm_offset_t)base;
487 KvaSize = KERNEL_KVA_SIZE;
488 KvaEnd = KvaStart + KvaSize;
d4f5f27c
MD
489
490 /* cannot use kprintf yet */
da673940
JG
491 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
492
493 /* MAP_FILE? */
494 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE,
495 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED,
496 MemImageFd, 0);
497 if (dmap_min_address == MAP_FAILED) {
498 err(1, "Unable to mmap() kernel DMAP region!");
499 /* NOT REACHED */
500 }
501
502 firstfree = 0;
2c2e847c 503 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base);
da673940
JG
504
505 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
506 0 | VPTE_R | VPTE_W | VPTE_V);
507
508 /*
509 * phys_avail[] represents unallocated physical memory. MI code
510 * will use phys_avail[] to create the vm_page array.
511 */
2c2e847c 512 phys_avail[0] = (vm_paddr_t)firstfree;
da673940
JG
513 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
514 phys_avail[1] = Maxmem_bytes;
515
516#if JGV
517 /*
518 * (virtual_start, virtual_end) represent unallocated kernel virtual
519 * memory. MI code will create kernel_map using these parameters.
520 */
521 virtual_start = KvaStart + (long)firstfree;
522 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
523 virtual_end = KvaStart + KERNEL_KVA_SIZE;
524#endif
525
526 /*
527 * pmap_growkernel() will set the correct value.
528 */
529 kernel_vm_end = 0;
530
531 /*
532 * Allocate space for process 0's UAREA.
533 */
534 proc0paddr = (void *)virtual_start;
535 for (i = 0; i < UPAGES; ++i) {
536 pmap_kenter_quick(virtual_start, phys_avail[0]);
537 virtual_start += PAGE_SIZE;
538 phys_avail[0] += PAGE_SIZE;
539 }
540
541 /*
542 * crashdumpmap
543 */
544 crashdumpmap = virtual_start;
545 virtual_start += MAXDUMPPGS * PAGE_SIZE;
546
547 /*
548 * msgbufp maps the system message buffer
549 */
550 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
551 msgbufp = (void *)virtual_start;
552 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
553 pmap_kenter_quick(virtual_start, phys_avail[0]);
554 virtual_start += PAGE_SIZE;
555 phys_avail[0] += PAGE_SIZE;
556 }
557 msgbufinit(msgbufp, MSGBUF_SIZE);
558
559 /*
560 * used by kern_memio for /dev/mem access
561 */
562 ptvmmap = (caddr_t)virtual_start;
563 virtual_start += PAGE_SIZE;
564
565 /*
566 * Bootstrap the kernel_pmap
567 */
568#if JGV
569 pmap_bootstrap();
570#endif
571}
572
573/*
574 * Map the per-cpu globaldata for cpu #0. Allocate the space using
575 * virtual_start and phys_avail[0]
576 */
577static
578void
579init_globaldata(void)
580{
581 int i;
582 vm_paddr_t pa;
583 vm_offset_t va;
584
585 /*
586 * Reserve enough KVA to cover possible cpus. This is a considerable
587 * amount of KVA since the privatespace structure includes two
588 * whole page table mappings.
589 */
590 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
591 CPU_prvspace = (void *)virtual_start;
592 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
593
594 /*
595 * Allocate enough physical memory to cover the mdglobaldata
596 * portion of the space and the idle stack and map the pages
597 * into KVA. For cpu #0 only.
598 */
599 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
600 pa = phys_avail[0];
601 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
602 pmap_kenter_quick(va, pa);
603 phys_avail[0] += PAGE_SIZE;
604 }
605 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
606 pa = phys_avail[0];
607 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
608 pmap_kenter_quick(va, pa);
609 phys_avail[0] += PAGE_SIZE;
610 }
611
612 /*
613 * Setup the %gs for cpu #0. The mycpu macro works after this
614 * point. Note that %fs is used by pthreads.
615 */
616 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace));
617}
618
619/*
620 * Initialize very low level systems including thread0, proc0, etc.
621 */
622static
623void
624init_vkernel(void)
625{
626 struct mdglobaldata *gd;
627
628 gd = &CPU_prvspace[0].mdglobaldata;
629 bzero(gd, sizeof(*gd));
630
631 gd->mi.gd_curthread = &thread0;
632 thread0.td_gd = &gd->mi;
633 ncpus = 1;
634 ncpus2 = 1; /* rounded down power of 2 */
635 ncpus_fit = 1; /* rounded up power of 2 */
636 /* ncpus2_mask and ncpus_fit_mask are 0 */
637 init_param1();
638 gd->mi.gd_prvspace = &CPU_prvspace[0];
639 mi_gdinit(&gd->mi, 0);
640 cpu_gdinit(gd, 0);
641 mi_proc0init(&gd->mi, proc0paddr);
642 lwp0.lwp_md.md_regs = &proc0_tf;
643
644 /*init_locks();*/
645 cninit();
646 rand_initialize();
647#if 0 /* #ifdef DDB */
648 kdb_init();
649 if (boothowto & RB_KDB)
650 Debugger("Boot flags requested debugger");
651#endif
652 identcpu();
653#if 0
654 initializecpu(); /* Initialize CPU registers */
655#endif
656 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
657
658#if 0
659 /*
660 * Map the message buffer
661 */
662 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
663 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
664 msgbufinit(msgbufp, MSGBUF_SIZE);
665#endif
666#if 0
667 thread0.td_pcb_cr3 ... MMU
668 lwp0.lwp_md.md_regs = &proc0_tf;
669#endif
670}
671
672/*
673 * Filesystem image paths for the virtual kernel are optional.
674 * If specified they each should point to a disk image,
675 * the first of which will become the root disk.
676 *
677 * The virtual kernel caches data from our 'disk' just like a normal kernel,
678 * so we do not really want the real kernel to cache the data too. Use
679 * O_DIRECT to remove the duplication.
680 */
681static
682void
683init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
684{
685 int i;
686
687 if (diskFileNum == 0)
688 return;
689
690 for(i=0; i < diskFileNum; i++){
691 char *fname;
692 fname = diskExp[i];
693
694 if (fname == NULL) {
695 warnx("Invalid argument to '-r'");
696 continue;
697 }
698
699 if (DiskNum < VKDISK_MAX) {
700 struct stat st;
701 struct vkdisk_info* info = NULL;
702 int fd;
703 size_t l = 0;
704
705 if (type == VKD_DISK)
6882b7a5 706 fd = open(fname, O_RDWR|O_DIRECT, 0644);
da673940
JG
707 else
708 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
709 if (fd < 0 || fstat(fd, &st) < 0) {
da673940
JG
710 err(1, "Unable to open/create %s", fname);
711 /* NOT REACHED */
712 }
6882b7a5
MD
713 if (S_ISREG(st.st_mode)) {
714 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
715 errx(1, "Disk image %s is already "
716 "in use\n", fname);
717 /* NOT REACHED */
718 }
719 }
da673940
JG
720
721 info = &DiskInfo[DiskNum];
722 l = strlen(fname);
723
724 info->unit = i;
725 info->fd = fd;
726 info->type = type;
727 memcpy(info->fname, fname, l);
728
729 if (DiskNum == 0) {
730 if (type == VKD_CD) {
731 rootdevnames[0] = "cd9660:vcd0a";
732 } else if (type == VKD_DISK) {
733 rootdevnames[0] = "ufs:vkd0s0a";
734 rootdevnames[1] = "ufs:vkd0s1a";
735 }
736 }
737
738 DiskNum++;
739 } else {
740 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
741 continue;
742 }
743 }
744}
745
746static
747int
748netif_set_tapflags(int tap_unit, int f, int s)
749{
750 struct ifreq ifr;
751 int flags;
752
753 bzero(&ifr, sizeof(ifr));
754
755 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
756 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
757 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
758 return -1;
759 }
760
761 /*
762 * Adjust if_flags
763 *
764 * If the flags are already set/cleared, then we return
765 * immediately to avoid extra syscalls
766 */
767 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
768 if (f < 0) {
769 /* Turn off flags */
770 f = -f;
771 if ((flags & f) == 0)
772 return 0;
773 flags &= ~f;
774 } else {
775 /* Turn on flags */
776 if (flags & f)
777 return 0;
778 flags |= f;
779 }
780
781 /*
782 * Fix up ifreq.ifr_name, since it may be trashed
783 * in previous ioctl(SIOCGIFFLAGS)
784 */
785 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
786
787 ifr.ifr_flags = flags & 0xffff;
788 ifr.ifr_flagshigh = flags >> 16;
789 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
790 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
791 return -1;
792 }
793 return 0;
794}
795
796static
797int
798netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
799{
800 struct ifaliasreq ifra;
801 struct sockaddr_in *in;
802
803 bzero(&ifra, sizeof(ifra));
804 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
805
806 /* Setup address */
807 in = (struct sockaddr_in *)&ifra.ifra_addr;
808 in->sin_family = AF_INET;
809 in->sin_len = sizeof(*in);
810 in->sin_addr.s_addr = addr;
811
812 if (mask != 0) {
813 /* Setup netmask */
814 in = (struct sockaddr_in *)&ifra.ifra_mask;
815 in->sin_len = sizeof(*in);
816 in->sin_addr.s_addr = mask;
817 }
818
819 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
820 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
821 return -1;
822 }
823 return 0;
824}
825
826static
827int
828netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
829{
830 struct ifbreq ifbr;
831 struct ifdrv ifd;
832
833 bzero(&ifbr, sizeof(ifbr));
834 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
835 "tap%d", tap_unit);
836
837 bzero(&ifd, sizeof(ifd));
838 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
839 ifd.ifd_cmd = BRDGADD;
840 ifd.ifd_len = sizeof(ifbr);
841 ifd.ifd_data = &ifbr;
842
843 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
844 /*
845 * 'errno == EEXIST' means that the tap(4) is already
846 * a member of the bridge(4)
847 */
848 if (errno != EEXIST) {
849 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
850 return -1;
851 }
852 }
853 return 0;
854}
855
856#define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
857
858/*
859 * Locate the first unused tap(4) device file if auto mode is requested,
860 * or open the user supplied device file, and bring up the corresponding
861 * tap(4) interface.
862 *
863 * NOTE: Only tap(4) device file is supported currently
864 */
865static
866int
867netif_open_tap(const char *netif, int *tap_unit, int s)
868{
869 char tap_dev[MAXPATHLEN];
870 int tap_fd, failed;
871 struct stat st;
872 char *dname;
873
874 *tap_unit = -1;
875
876 if (strcmp(netif, "auto") == 0) {
877 /*
878 * Find first unused tap(4) device file
879 */
880 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
881 if (tap_fd < 0) {
882 warnc(errno, "Unable to find a free tap(4)");
883 return -1;
884 }
885 } else {
886 /*
887 * User supplied tap(4) device file or unix socket.
888 */
889 if (netif[0] == '/') /* Absolute path */
890 strlcpy(tap_dev, netif, sizeof(tap_dev));
891 else
892 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
893
894 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
895
896 /*
897 * If we cannot open normally try to connect to it.
898 */
899 if (tap_fd < 0)
900 tap_fd = unix_connect(tap_dev);
901
902 if (tap_fd < 0) {
903 warn("Unable to open %s", tap_dev);
904 return -1;
905 }
906 }
907
908 /*
909 * Check whether the device file is a tap(4)
910 */
911 if (fstat(tap_fd, &st) < 0) {
912 failed = 1;
913 } else if (S_ISCHR(st.st_mode)) {
914 dname = fdevname(tap_fd);
915 if (dname)
916 dname = strstr(dname, "tap");
917 if (dname) {
918 /*
919 * Bring up the corresponding tap(4) interface
920 */
921 *tap_unit = strtol(dname + 3, NULL, 10);
922 printf("TAP UNIT %d\n", *tap_unit);
923 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
924 failed = 0;
3b6e5cb0
SW
925 else
926 failed = 1;
da673940
JG
927 } else {
928 failed = 1;
929 }
930 } else if (S_ISSOCK(st.st_mode)) {
931 /*
932 * Special socket connection (typically to vknet). We
933 * do not have to do anything.
934 */
935 failed = 0;
936 } else {
937 failed = 1;
938 }
939
940 if (failed) {
941 warnx("%s is not a tap(4) device or socket", tap_dev);
942 close(tap_fd);
943 tap_fd = -1;
944 *tap_unit = -1;
945 }
946 return tap_fd;
947}
948
949static int
950unix_connect(const char *path)
951{
952 struct sockaddr_un sunx;
953 int len;
954 int net_fd;
955 int sndbuf = 262144;
956 struct stat st;
957
958 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
959 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
960 ++len; /* include nul */
961 sunx.sun_family = AF_UNIX;
962 sunx.sun_len = len;
963
964 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
965 if (net_fd < 0)
966 return(-1);
967 if (connect(net_fd, (void *)&sunx, len) < 0) {
968 close(net_fd);
969 return(-1);
970 }
971 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
972 if (fstat(net_fd, &st) == 0)
973 printf("Network socket buffer: %d bytes\n", st.st_blksize);
974 fcntl(net_fd, F_SETFL, O_NONBLOCK);
975 return(net_fd);
976}
977
978#undef TAPDEV_MAJOR
979#undef TAPDEV_MINOR
980#undef TAPDEV_OFLAGS
981
982/*
983 * Following syntax is supported,
984 * 1) x.x.x.x tap(4)'s address is x.x.x.x
985 *
986 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
987 * tap(4)'s netmask len is z
988 *
989 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
990 * pseudo netif's address is y.y.y.y
991 *
992 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
993 * pseudo netif's address is y.y.y.y
994 * tap(4) and pseudo netif's netmask len are z
995 *
996 * 5) bridgeX tap(4) will be added to bridgeX
997 *
998 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
999 * pseudo netif's address is y.y.y.y
1000 *
1001 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1002 * pseudo netif's address is y.y.y.y
1003 * pseudo netif's netmask len is z
1004 */
1005static
1006int
1007netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1008{
1009 in_addr_t tap_addr, netmask, netif_addr;
1010 int next_netif_addr;
1011 char *tok, *masklen_str, *ifbridge;
1012
1013 *addr = 0;
1014 *mask = 0;
1015
1016 tok = strtok(NULL, ":/");
1017 if (tok == NULL) {
1018 /*
1019 * Nothing special, simply use tap(4) as backend
1020 */
1021 return 0;
1022 }
1023
1024 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1025 /*
1026 * tap(4)'s address is supplied
1027 */
1028 ifbridge = NULL;
1029
1030 /*
1031 * If there is next token, then it may be pseudo
1032 * netif's address or netmask len for tap(4)
1033 */
1034 next_netif_addr = 0;
1035 } else {
1036 /*
1037 * Not tap(4)'s address, assume it as a bridge(4)
1038 * iface name
1039 */
1040 tap_addr = 0;
1041 ifbridge = tok;
1042
1043 /*
1044 * If there is next token, then it must be pseudo
1045 * netif's address
1046 */
1047 next_netif_addr = 1;
1048 }
1049
1050 netmask = netif_addr = 0;
1051
1052 tok = strtok(NULL, ":/");
1053 if (tok == NULL)
1054 goto back;
1055
1056 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1057 if (next_netif_addr) {
1058 warnx("Invalid pseudo netif address: %s", tok);
1059 return -1;
1060 }
1061 netif_addr = 0;
1062
1063 /*
1064 * Current token is not address, then it must be netmask len
1065 */
1066 masklen_str = tok;
1067 } else {
1068 /*
1069 * Current token is pseudo netif address, if there is next token
1070 * it must be netmask len
1071 */
1072 masklen_str = strtok(NULL, "/");
1073 }
1074
1075 /* Calculate netmask */
1076 if (masklen_str != NULL) {
1077 u_long masklen;
1078
1079 masklen = strtoul(masklen_str, NULL, 10);
1080 if (masklen < 32 && masklen > 0) {
1081 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1082 & 0xffffffff);
1083 } else {
1084 warnx("Invalid netmask len: %lu", masklen);
1085 return -1;
1086 }
1087 }
1088
1089 /* Make sure there is no more token left */
1090 if (strtok(NULL, ":/") != NULL) {
1091 warnx("Invalid argument to '-I'");
1092 return -1;
1093 }
1094
1095back:
1096 if (tap_unit < 0) {
1097 /* Do nothing */
1098 } else if (ifbridge == NULL) {
1099 /* Set tap(4) address/netmask */
1100 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1101 return -1;
1102 } else {
1103 /* Tie tap(4) to bridge(4) */
1104 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1105 return -1;
1106 }
1107
1108 *addr = netif_addr;
1109 *mask = netmask;
1110 return 0;
1111}
1112
1113/*
1114 * NetifInfo[] will be filled for pseudo netif initialization.
1115 * NetifNum will be bumped to reflect the number of valid entries
1116 * in NetifInfo[].
1117 */
1118static
1119void
1120init_netif(char *netifExp[], int netifExpNum)
1121{
1122 int i, s;
1123
1124 if (netifExpNum == 0)
1125 return;
1126
1127 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1128 if (s < 0)
1129 return;
1130
1131 for (i = 0; i < netifExpNum; ++i) {
1132 struct vknetif_info *info;
1133 in_addr_t netif_addr, netif_mask;
1134 int tap_fd, tap_unit;
1135 char *netif;
1136
1137 netif = strtok(netifExp[i], ":");
1138 if (netif == NULL) {
1139 warnx("Invalid argument to '-I'");
1140 continue;
1141 }
1142
1143 /*
1144 * Open tap(4) device file and bring up the
1145 * corresponding interface
1146 */
1147 tap_fd = netif_open_tap(netif, &tap_unit, s);
1148 if (tap_fd < 0)
1149 continue;
1150
1151 /*
1152 * Initialize tap(4) and get address/netmask
1153 * for pseudo netif
1154 *
1155 * NB: Rest part of netifExp[i] is passed
1156 * to netif_init_tap() implicitly.
1157 */
1158 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1159 /*
1160 * NB: Closing tap(4) device file will bring
1161 * down the corresponding interface
1162 */
1163 close(tap_fd);
1164 continue;
1165 }
1166
1167 info = &NetifInfo[NetifNum];
1168 info->tap_fd = tap_fd;
1169 info->tap_unit = tap_unit;
1170 info->netif_addr = netif_addr;
1171 info->netif_mask = netif_mask;
1172
1173 NetifNum++;
1174 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1175 break;
1176 }
1177 close(s);
1178}
1179
1180static
1181void
1182writepid( void )
1183{
1184 pid_t self;
1185 FILE *fp;
1186
1187 if (pid_file != NULL) {
1188 self = getpid();
1189 fp = fopen(pid_file, "w");
1190
1191 if (fp != NULL) {
1192 fprintf(fp, "%ld\n", (long)self);
1193 fclose(fp);
1194 }
1195 else {
1196 perror("Warning: couldn't open pidfile");
1197 }
1198 }
1199}
1200
1201static
1202void
1203cleanpid( void )
1204{
1205 if (pid_file != NULL) {
1206 if ( unlink(pid_file) != 0 )
1207 perror("Warning: couldn't remove pidfile");
1208 }
1209}
1210
1211static
1212void
278c4e67 1213usage_err(const char *ctl, ...)
da673940
JG
1214{
1215 va_list va;
1216
1217 va_start(va, ctl);
1218 vfprintf(stderr, ctl, va);
1219 va_end(va);
1220 fprintf(stderr, "\n");
278c4e67
SG
1221 exit(EX_USAGE);
1222}
1223
1224static
1225void
1226usage_help(_Bool help)
1227{
1228 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1229 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
1230 "\t[-m size] [-n numcpus] [-p file] [-r file]\n", save_av[0]);
1231
1232 if (help)
1233 fprintf(stderr, "\nArguments:\n"
1234 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1235 "\t-e\tSpecify an environment to be used by the kernel.\n"
1236 "\t-h\tThis list of options.\n"
1237 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1238 "\t-I\tCreate a virtual network device.\n"
1239 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1240 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
1241 "\t-n\tSpecify the number of CPUs you wish to emulate.\n"
1242 "\t-p\tSpecify a file in which to store the process ID.\n"
1243 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1244 "\t-s\tBoot into single-user mode.\n"
1245 "\t-U\tEnable writing to kernel memory and module loading.\n"
1246 "\t-v\tTurn on verbose booting.\n");
1247
1248 exit(EX_USAGE);
da673940
JG
1249}
1250
1251void
1252cpu_reset(void)
1253{
1254 kprintf("cpu reset, rebooting vkernel\n");
1255 closefrom(3);
1256 cleanpid();
1257 execv(save_av[0], save_av);
1258}
1259
1260void
1261cpu_halt(void)
1262{
1263 kprintf("cpu halt, exiting vkernel\n");
1264 cleanpid();
278c4e67 1265 exit(EX_OK);
da673940
JG
1266}
1267
1268void
1269setrealcpu(void)
1270{
1271 switch(lwp_cpu_lock) {
1272 case LCL_PER_CPU:
1273 if (bootverbose)
1274 kprintf("Locking CPU%d to real cpu %d\n",
1275 mycpuid, next_cpu);
1276 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1277 next_cpu++;
1278 if (next_cpu >= real_ncpus)
1279 next_cpu = 0;
1280 break;
1281 case LCL_SINGLE_CPU:
1282 if (bootverbose)
1283 kprintf("Locking CPU%d to real cpu %d\n",
1284 mycpuid, next_cpu);
1285 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1286 break;
1287 default:
1288 /* do not map virtual cpus to real cpus */
1289 break;
1290 }
1291}