1 /*
2 * Copyright (c) 2006 The DragonFly Project. All rights reserved.
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
16 * distribution.
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35 #include <sys/types.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
38 #include <sys/stat.h>
39 #include <sys/mman.h>
40 #include <sys/cons.h>
41 #include <sys/random.h>
42 #include <sys/vkernel.h>
43 #include <sys/tls.h>
44 #include <sys/reboot.h>
45 #include <sys/proc.h>
46 #include <sys/msgbuf.h>
47 #include <sys/vmspace.h>
48 #include <sys/socket.h>
49 #include <sys/sockio.h>
50 #include <sys/sysctl.h>
51 #include <sys/un.h>
52 #include <vm/vm_page.h>
53 #include <vm/vm_map.h>
54 #include <sys/mplock2.h>
55 #include <sys/wait.h>
56
57 #include <machine/cpu.h>
58 #include <machine/globaldata.h>
59 #include <machine/tls.h>
60 #include <machine/md_var.h>
61 #include <machine/vmparam.h>
62 #include <cpu/specialreg.h>
63
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 #include <net/if_var.h>
71
72 #include <stdio.h>
73 #include <stdlib.h>
74 #include <stdarg.h>
75 #include <stdbool.h>
76 #include <unistd.h>
77 #include <fcntl.h>
78 #include <string.h>
79 #include <err.h>
80 #include <errno.h>
81 #include <assert.h>
82 #include <sysexits.h>
83 #include <pthread.h>
84
85 #define EX_VKERNEL_REBOOT 32
86
87 vm_phystable_t phys_avail[16];
88 vm_paddr_t Maxmem;
89 vm_paddr_t Maxmem_bytes;
90 long physmem;
91 int MemImageFd = -1;
92 struct vkdisk_info DiskInfo[VKDISK_MAX];
93 int DiskNum;
94 struct vknetif_info NetifInfo[VKNETIF_MAX];
95 int NetifNum;
96 char *pid_file;
97 vm_offset_t KvaStart;
98 vm_offset_t KvaEnd;
99 vm_offset_t KvaSize;
100 vm_offset_t virtual_start;
101 vm_offset_t virtual_end;
102 vm_offset_t virtual2_start;
103 vm_offset_t virtual2_end;
104 vm_offset_t kernel_vm_end;
105 vm_offset_t crashdumpmap;
106 vm_offset_t clean_sva;
107 vm_offset_t clean_eva;
108 struct msgbuf *msgbufp;
109 caddr_t ptvmmap;
110 vpte_t *KernelPTD;
111 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
112 void *dmap_min_address;
113 void *vkernel_stack;
114 u_int cpu_feature; /* XXX */
115 int tsc_present;
116 int tsc_invariant;
117 int tsc_mpsync;
118 int optcpus; /* number of cpus - see mp_start() */
119 int cpu_bits;
120 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
121 int real_ncpus; /* number of real CPUs */
122 int next_cpu; /* next real CPU to lock a virtual CPU to */
123 int vkernel_b_arg; /* no of logical CPU bits - only SMP */
124 int vkernel_B_arg; /* no of core bits - only SMP */
125 int use_precise_timer = 0; /* use a precise timer (more expensive) */
126 struct privatespace *CPU_prvspace;
127
128 tsc_uclock_t tsc_frequency;
129 tsc_uclock_t tsc_oneus_approx;
130
131 extern uint64_t KPML4phys; /* phys addr of kernel level 4 */
132
133 static struct trapframe proc0_tf;
134 static void *proc0paddr;
135
136 static void init_sys_memory(char *imageFile);
137 static void init_kern_memory(void);
138 static void init_globaldata(void);
139 static void init_vkernel(void);
140 static void init_disk(char **diskExp, int *diskFlags, int diskFileNum, enum vkdisk_type type);
141 static void init_netif(char *netifExp[], int netifFileNum);
142 static void writepid(void);
143 static void cleanpid(void);
144 static int unix_connect(const char *path);
145 static void usage_err(const char *ctl, ...) __printflike(1, 2);
146 static void usage_help(_Bool);
147 static void init_locks(void);
148 static void handle_term(int);
149
150 pid_t childpid;
151
152 static int save_ac;
153 static int prezeromem;
154 static char **save_av;
155
156 /*
157 * Kernel startup for virtual kernels - standard main()
158 */
159 int
main(int ac,char ** av)160 main(int ac, char **av)
161 {
162 char *memImageFile = NULL;
163 char *netifFile[VKNETIF_MAX];
164 char *diskFile[VKDISK_MAX];
165 char *cdFile[VKDISK_MAX];
166 char *suffix;
167 char *endp;
168 char *tmp;
169 char *tok;
170 int diskFlags[VKDISK_MAX];
171 int netifFileNum = 0;
172 int diskFileNum = 0;
173 int cdFileNum = 0;
174 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
175 int c;
176 int i;
177 int j;
178 int n;
179 int isq;
180 int pos;
181 int eflag;
182 int real_vkernel_enable;
183 int supports_sse;
184 uint32_t mxcsr_mask;
185 size_t vsize;
186 size_t msize;
187 size_t kenv_size;
188 size_t kenv_size2;
189 int status;
190 struct sigaction sa;
191
192 /*
193 * Currently a bad hack but rtld-elf needs LD_SHAREDLIB_BASE to
194 * be set to force it to mmap() shared libraries into low memory,
195 * so our module loader can link against the related symbols.
196 */
197 if (getenv("LD_SHAREDLIB_BASE") == NULL) {
198 setenv("LD_SHAREDLIB_BASE", "0x10000000", 1);
199 execv(av[0], av);
200 fprintf(stderr, "Must run %s with full path\n", av[0]);
201 exit(1);
202 }
203
204 while ((childpid = fork()) != 0) {
205 /* Ignore signals */
206 bzero(&sa, sizeof(sa));
207 sigemptyset(&sa.sa_mask);
208 sa.sa_handler = SIG_IGN;
209 sigaction(SIGINT, &sa, NULL);
210 sigaction(SIGQUIT, &sa, NULL);
211 sigaction(SIGHUP, &sa, NULL);
212
213 /*
214 * Forward SIGTERM to the child so that
215 * the shutdown process initiates correctly.
216 */
217 sa.sa_handler = handle_term;
218 sigaction(SIGTERM, &sa, NULL);
219
220 /*
221 * Wait for child to terminate, exit if
222 * someone stole our child.
223 */
224 while (waitpid(childpid, &status, 0) != childpid) {
225 if (errno == ECHILD)
226 exit(1);
227 }
228 if (WEXITSTATUS(status) != EX_VKERNEL_REBOOT)
229 return 0;
230 }
231
232 /*
233 * Starting for real
234 */
235 save_ac = ac;
236 save_av = av;
237 eflag = 0;
238 pos = 0;
239 kenv_size = 0;
240
241 /*
242 * Process options
243 */
244 kernel_mem_readonly = 1;
245 optcpus = 2;
246 cpu_bits = 1;
247 vkernel_b_arg = 0;
248 vkernel_B_arg = 0;
249 lwp_cpu_lock = LCL_NONE;
250
251 real_vkernel_enable = 0;
252 vsize = sizeof(real_vkernel_enable);
253 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
254
255 if (real_vkernel_enable == 0) {
256 errx(1, "vm.vkernel_enable is 0, must be set "
257 "to 1 to execute a vkernel!");
258 }
259
260 real_ncpus = 1;
261 vsize = sizeof(real_ncpus);
262 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
263
264 if (ac < 2)
265 usage_help(false);
266
267 while ((c = getopt(ac, av, "c:hsvztTl:m:n:r:R:e:i:p:I:U")) != -1) {
268 switch(c) {
269 case 'e':
270 /*
271 * name=value:name=value:name=value...
272 * name="value"...
273 *
274 * Allow values to be quoted but note that shells
275 * may remove the quotes, so using this feature
276 * to embed colons may require a backslash.
277 */
278 n = strlen(optarg);
279 isq = 0;
280
281 if (eflag == 0) {
282 kenv_size = n + 2;
283 kern_envp = malloc(kenv_size);
284 if (kern_envp == NULL)
285 errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size);
286 } else {
287 kenv_size2 = kenv_size + n + 1;
288 pos = kenv_size - 1;
289 if ((tmp = realloc(kern_envp, kenv_size2)) == NULL)
290 errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2);
291 kern_envp = tmp;
292 kenv_size = kenv_size2;
293 }
294
295 for (i = 0, j = pos; i < n; ++i) {
296 if (optarg[i] == '"')
297 isq ^= 1;
298 else if (optarg[i] == '\'')
299 isq ^= 2;
300 else if (isq == 0 && optarg[i] == ':')
301 kern_envp[j++] = 0;
302 else
303 kern_envp[j++] = optarg[i];
304 }
305 kern_envp[j++] = 0;
306 kern_envp[j++] = 0;
307 eflag++;
308 break;
309 case 's':
310 boothowto |= RB_SINGLE;
311 break;
312 case 't':
313 use_precise_timer = 1;
314 break;
315 case 'v':
316 bootverbose = 1;
317 break;
318 case 'i':
319 memImageFile = optarg;
320 break;
321 case 'I':
322 if (netifFileNum < VKNETIF_MAX)
323 netifFile[netifFileNum++] = strdup(optarg);
324 break;
325 case 'r':
326 case 'R':
327 if (bootOnDisk < 0)
328 bootOnDisk = 1;
329 if (diskFileNum + cdFileNum < VKDISK_MAX) {
330 diskFile[diskFileNum] = strdup(optarg);
331 diskFlags[diskFileNum] = (c == 'R');
332 ++diskFileNum;
333 }
334 break;
335 case 'c':
336 if (bootOnDisk < 0)
337 bootOnDisk = 0;
338 if (diskFileNum + cdFileNum < VKDISK_MAX)
339 cdFile[cdFileNum++] = strdup(optarg);
340 break;
341 case 'm':
342 Maxmem_bytes = strtoull(optarg, &suffix, 0);
343 if (suffix) {
344 switch(*suffix) {
345 case 'g':
346 case 'G':
347 Maxmem_bytes <<= 30;
348 break;
349 case 'm':
350 case 'M':
351 Maxmem_bytes <<= 20;
352 break;
353 case 'k':
354 case 'K':
355 Maxmem_bytes <<= 10;
356 break;
357 default:
358 Maxmem_bytes = 0;
359 usage_err("Bad maxmem option");
360 /* NOT REACHED */
361 break;
362 }
363 }
364 break;
365 case 'l':
366 next_cpu = -1;
367 if (strncmp("map", optarg, 3) == 0) {
368 lwp_cpu_lock = LCL_PER_CPU;
369 if (optarg[3] == ',') {
370 next_cpu = strtol(optarg+4, &endp, 0);
371 if (*endp != '\0')
372 usage_err("Bad target CPU number at '%s'", endp);
373 } else {
374 next_cpu = 0;
375 }
376 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
377 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
378 } else if (strncmp("any", optarg, 3) == 0) {
379 lwp_cpu_lock = LCL_NONE;
380 } else {
381 lwp_cpu_lock = LCL_SINGLE_CPU;
382 next_cpu = strtol(optarg, &endp, 0);
383 if (*endp != '\0')
384 usage_err("Bad target CPU number at '%s'", endp);
385 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
386 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
387 }
388 break;
389 case 'n':
390 /*
391 * This value is set up by mp_start(), don't just
392 * set ncpus here.
393 */
394 tok = strtok(optarg, ":");
395 optcpus = strtol(tok, NULL, 0);
396 if (optcpus < 1 || optcpus > MAXCPU)
397 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
398 cpu_bits = 1;
399 while ((1 << cpu_bits) < optcpus)
400 ++cpu_bits;
401
402 /*
403 * By default assume simple hyper-threading
404 */
405 vkernel_b_arg = 1;
406 vkernel_B_arg = cpu_bits - vkernel_b_arg;
407
408 /*
409 * [:lbits[:cbits]] override # of cpu bits
410 * for logical and core extraction, supplying
411 * defaults for any omission.
412 */
413 tok = strtok(NULL, ":");
414 if (tok != NULL) {
415 vkernel_b_arg = strtol(tok, NULL, 0);
416 vkernel_B_arg = cpu_bits - vkernel_b_arg;
417
418 /* :cbits argument */
419 tok = strtok(NULL, ":");
420 if (tok != NULL) {
421 vkernel_B_arg = strtol(tok, NULL, 0);
422 }
423 }
424 break;
425 case 'p':
426 pid_file = optarg;
427 break;
428 case 'U':
429 kernel_mem_readonly = 0;
430 break;
431 case 'h':
432 usage_help(true);
433 break;
434 case 'z':
435 prezeromem = 1;
436 break;
437 default:
438 usage_help(false);
439 }
440 }
441
442 writepid();
443 cpu_disable_intr();
444 init_sys_memory(memImageFile);
445 init_kern_memory();
446 init_globaldata();
447 init_vkernel();
448 setrealcpu();
449 init_kqueue();
450
451 vmm_guest = VMM_GUEST_VKERNEL;
452
453 /*
454 * Check TSC
455 */
456 vsize = sizeof(tsc_present);
457 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
458 vsize = sizeof(tsc_invariant);
459 sysctlbyname("hw.tsc_invariant", &tsc_invariant, &vsize, NULL, 0);
460 vsize = sizeof(tsc_mpsync);
461 sysctlbyname("hw.tsc_mpsync", &tsc_mpsync, &vsize, NULL, 0);
462 vsize = sizeof(tsc_frequency);
463 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
464 if (tsc_present)
465 cpu_feature |= CPUID_TSC;
466 tsc_oneus_approx = ((tsc_frequency|1) + 999999) / 1000000;
467
468 /*
469 * Check SSE
470 */
471 vsize = sizeof(supports_sse);
472 supports_sse = 0;
473 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
474 sysctlbyname("hw.mxcsr_mask", &mxcsr_mask, &msize, NULL, 0);
475 init_fpu(supports_sse);
476 if (supports_sse)
477 cpu_feature |= CPUID_SSE | CPUID_FXSR;
478
479 /*
480 * We boot from the first installed disk.
481 */
482 if (bootOnDisk == 1) {
483 init_disk(diskFile, diskFlags, diskFileNum, VKD_DISK);
484 init_disk(cdFile, NULL, cdFileNum, VKD_CD);
485 } else {
486 init_disk(cdFile, NULL, cdFileNum, VKD_CD);
487 init_disk(diskFile, diskFlags, diskFileNum, VKD_DISK);
488 }
489
490 init_netif(netifFile, netifFileNum);
491 init_exceptions();
492 mi_startup();
493 /* NOT REACHED */
494 exit(EX_SOFTWARE);
495 }
496
497 /* SIGTERM handler */
498 static
499 void
handle_term(int sig)500 handle_term(int sig)
501 {
502 kill(childpid, sig);
503 }
504
505 /*
506 * Initialize system memory. This is the virtual kernel's 'RAM'.
507 */
508 static
509 void
init_sys_memory(char * imageFile)510 init_sys_memory(char *imageFile)
511 {
512 struct stat st;
513 int i;
514 int fd;
515
516 /*
517 * Figure out the system memory image size. If an image file was
518 * specified and -m was not specified, use the image file's size.
519 */
520 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
521 Maxmem_bytes = (vm_paddr_t)st.st_size;
522 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
523 Maxmem_bytes == 0) {
524 errx(1, "Cannot create new memory file %s unless "
525 "system memory size is specified with -m",
526 imageFile);
527 /* NOT REACHED */
528 }
529
530 /*
531 * Maxmem must be known at this time
532 */
533 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
534 errx(1, "Bad maxmem specification: 64MB minimum, "
535 "multiples of %dMB only",
536 SEG_SIZE / 1024 / 1024);
537 /* NOT REACHED */
538 }
539
540 /*
541 * Generate an image file name if necessary, then open/create the
542 * file exclusively locked. Do not allow multiple virtual kernels
543 * to use the same image file.
544 *
545 * Don't iterate through a million files if we do not have write
546 * access to the directory, stop if our open() failed on a
547 * non-existant file. Otherwise opens can fail for any number
548 */
549 if (imageFile == NULL) {
550 for (i = 0; i < 1000000; ++i) {
551 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
552 fd = open(imageFile,
553 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
554 if (fd < 0 && stat(imageFile, &st) == 0) {
555 free(imageFile);
556 continue;
557 }
558 break;
559 }
560 } else {
561 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
562 }
563 fprintf(stderr, "Using memory file: %s\n", imageFile);
564 if (fd < 0 || fstat(fd, &st) < 0) {
565 err(1, "Unable to open/create %s", imageFile);
566 /* NOT REACHED */
567 }
568
569 /*
570 * Truncate or extend the file as necessary. Clean out the contents
571 * of the file, we want it to be full of holes so we don't waste
572 * time reading in data from an old file that we no longer care
573 * about.
574 */
575 ftruncate(fd, 0);
576 ftruncate(fd, Maxmem_bytes);
577
578 MemImageFd = fd;
579 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
580 physmem = Maxmem;
581 }
582
583 /*
584 * Initialize kernel memory. This reserves kernel virtual memory by using
585 * MAP_VPAGETABLE
586 *
587 * XXX NOTE! MAP_VPAGETABLE is being ripped out and will break VKERNELs
588 * for a while, until we get hardware virtualization working.
589 */
590
591 static
592 void
init_kern_memory(void)593 init_kern_memory(void)
594 {
595 void *base;
596 int i;
597 void *firstfree;
598
599 /*
600 * Memory map our kernel virtual memory space. Note that the
601 * kernel image itself is not made part of this memory for the
602 * moment.
603 *
604 * The memory map must be segment-aligned so we can properly
605 * offset KernelPTD.
606 *
607 * If the system kernel has a different MAXDSIZ, it might not
608 * be possible to map kernel memory in its prefered location.
609 * Try a number of different locations.
610 */
611
612 base = mmap((void*)KERNEL_KVA_START, KERNEL_KVA_SIZE,
613 PROT_READ|PROT_WRITE|PROT_EXEC,
614 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE|MAP_FIXED|MAP_TRYFIXED,
615 MemImageFd, (off_t)KERNEL_KVA_START);
616
617 if (base == MAP_FAILED) {
618 err(1, "Unable to mmap() kernel virtual memory!");
619 /* NOT REACHED */
620 }
621 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
622 KvaStart = (vm_offset_t)base;
623 KvaSize = KERNEL_KVA_SIZE;
624 KvaEnd = KvaStart + KvaSize;
625
626 /* cannot use kprintf yet */
627 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
628
629 /* MAP_FILE? */
630 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE,
631 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED,
632 MemImageFd, 0);
633 if (dmap_min_address == MAP_FAILED) {
634 err(1, "Unable to mmap() kernel DMAP region!");
635 /* NOT REACHED */
636 }
637
638 /*
639 * Prefault the memory. The vkernel is going to fault it all in
640 * anyway, and faults on the backing store itself are very expensive
641 * once we go SMP (contend a lot). So do it now.
642 */
643 if (prezeromem)
644 bzero(dmap_min_address, Maxmem_bytes);
645
646 /*
647 * Bootstrap the kernel_pmap
648 */
649 firstfree = NULL;
650 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base);
651
652 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
653 0 | VPTE_RW | VPTE_V);
654
655 /*
656 * phys_avail[] represents unallocated physical memory. MI code
657 * will use phys_avail[] to create the vm_page array.
658 */
659 phys_avail[0].phys_beg = (vm_paddr_t)firstfree;
660 phys_avail[0].phys_beg = (phys_avail[0].phys_beg + PAGE_MASK) &
661 ~(vm_paddr_t)PAGE_MASK;
662 phys_avail[0].phys_end = Maxmem_bytes;
663
664 #if 0 /* JGV */
665 /*
666 * (virtual_start, virtual_end) represent unallocated kernel virtual
667 * memory. MI code will create kernel_map using these parameters.
668 */
669 virtual_start = KvaStart + (long)firstfree;
670 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
671 virtual_end = KvaStart + KERNEL_KVA_SIZE;
672 #endif
673
674 /*
675 * pmap_growkernel() will set the correct value.
676 */
677 kernel_vm_end = 0;
678
679 /*
680 * Allocate space for process 0's UAREA.
681 */
682 proc0paddr = (void *)virtual_start;
683 for (i = 0; i < UPAGES; ++i) {
684 pmap_kenter_quick(virtual_start, phys_avail[0].phys_beg);
685 virtual_start += PAGE_SIZE;
686 phys_avail[0].phys_beg += PAGE_SIZE;
687 }
688
689 /*
690 * crashdumpmap
691 */
692 crashdumpmap = virtual_start;
693 virtual_start += MAXDUMPPGS * PAGE_SIZE;
694
695 /*
696 * msgbufp maps the system message buffer
697 */
698 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
699 msgbufp = (void *)virtual_start;
700 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
701 pmap_kenter_quick(virtual_start, phys_avail[0].phys_beg);
702 virtual_start += PAGE_SIZE;
703 phys_avail[0].phys_beg += PAGE_SIZE;
704 }
705 msgbufinit(msgbufp, MSGBUF_SIZE);
706
707 /*
708 * used by kern_memio for /dev/mem access
709 */
710 ptvmmap = (caddr_t)virtual_start;
711 virtual_start += PAGE_SIZE;
712 }
713
714 /*
715 * Map the per-cpu globaldata for cpu #0. Allocate the space using
716 * virtual_start and phys_avail[0]
717 */
718 static
719 void
init_globaldata(void)720 init_globaldata(void)
721 {
722 int i;
723 vm_paddr_t pa;
724 vm_offset_t va;
725
726 /*
727 * Reserve enough KVA to cover possible cpus. This is a considerable
728 * amount of KVA since the privatespace structure includes two
729 * whole page table mappings.
730 */
731 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
732 CPU_prvspace = (void *)virtual_start;
733 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
734
735 /*
736 * Allocate enough physical memory to cover the mdglobaldata
737 * portion of the space and the idle stack and map the pages
738 * into KVA. For cpu #0 only.
739 */
740 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
741 pa = phys_avail[0].phys_beg;
742 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
743 pmap_kenter_quick(va, pa);
744 phys_avail[0].phys_beg += PAGE_SIZE;
745 }
746 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
747 pa = phys_avail[0].phys_beg;
748 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
749 pmap_kenter_quick(va, pa);
750 phys_avail[0].phys_beg += PAGE_SIZE;
751 }
752
753 /*
754 * Setup the %gs for cpu #0. The mycpu macro works after this
755 * point. Note that %fs is used by pthreads.
756 */
757 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace));
758 }
759
760
761 /*
762 * Initialize pool tokens and other necessary locks
763 */
764 static void
init_locks(void)765 init_locks(void)
766 {
767
768 /*
769 * Get the initial mplock with a count of 1 for the BSP.
770 * This uses a LOGICAL cpu ID, ie BSP == 0.
771 */
772 cpu_get_initial_mplock();
773
774 /* our token pool needs to work early */
775 lwkt_token_pool_init();
776
777 }
778
779
780 /*
781 * Initialize very low level systems including thread0, proc0, etc.
782 */
783 static
784 void
init_vkernel(void)785 init_vkernel(void)
786 {
787 struct mdglobaldata *gd;
788
789 gd = &CPU_prvspace[0].mdglobaldata;
790 bzero(gd, sizeof(*gd));
791
792 gd->mi.gd_curthread = &thread0;
793 thread0.td_gd = &gd->mi;
794 ncpus = 1;
795 ncpus_fit = 1; /* rounded up power of 2 */
796 /* ncpus_fit_mask are 0 */
797 init_param1();
798 gd->mi.gd_prvspace = &CPU_prvspace[0];
799 mi_gdinit(&gd->mi, 0);
800 cpu_gdinit(gd, 0);
801 mi_proc0init(&gd->mi, proc0paddr);
802 lwp0.lwp_md.md_regs = &proc0_tf;
803
804 init_locks();
805 cninit();
806 rand_initialize();
807 #if 0 /* #ifdef DDB */
808 kdb_init();
809 if (boothowto & RB_KDB)
810 Debugger("Boot flags requested debugger");
811 #endif
812 identcpu();
813 #if 0
814 initializecpu(); /* Initialize CPU registers */
815 #endif
816 init_param2((phys_avail[0].phys_end -
817 phys_avail[0].phys_beg) / PAGE_SIZE);
818
819 #if 0
820 /*
821 * Map the message buffer
822 */
823 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
824 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
825 msgbufinit(msgbufp, MSGBUF_SIZE);
826 #endif
827 #if 0
828 thread0.td_pcb_cr3 ... MMU
829 lwp0.lwp_md.md_regs = &proc0_tf;
830 #endif
831 }
832
833 /*
834 * Filesystem image paths for the virtual kernel are optional.
835 * If specified they each should point to a disk image,
836 * the first of which will become the root disk.
837 *
838 * The virtual kernel caches data from our 'disk' just like a normal kernel,
839 * so we do not really want the real kernel to cache the data too. Use
840 * O_DIRECT to remove the duplication.
841 */
842 static
843 void
init_disk(char ** diskExp,int * diskFlags,int diskFileNum,enum vkdisk_type type)844 init_disk(char **diskExp, int *diskFlags, int diskFileNum, enum vkdisk_type type)
845 {
846 char *serno;
847 int i;
848
849 if (diskFileNum == 0)
850 return;
851
852 for (i=0; i < diskFileNum; i++){
853 char *fname;
854 fname = diskExp[i];
855
856 if (fname == NULL) {
857 warnx("Invalid argument to '-r'");
858 continue;
859 }
860 /*
861 * Check for a serial number for the virtual disk
862 * passed from the command line.
863 */
864 serno = fname;
865 strsep(&serno, ":");
866
867 if (DiskNum < VKDISK_MAX) {
868 struct stat st;
869 struct vkdisk_info *info = NULL;
870 int fd;
871 size_t l = 0;
872
873 if (type == VKD_DISK)
874 fd = open(fname, O_RDWR|O_DIRECT, 0644);
875 else
876 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
877 if (fd < 0 || fstat(fd, &st) < 0) {
878 err(1, "Unable to open/create %s", fname);
879 /* NOT REACHED */
880 }
881 if (S_ISREG(st.st_mode) && (diskFlags[i] & 1) == 0) {
882 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
883 errx(1, "Disk image %s is already "
884 "in use\n", fname);
885 /* NOT REACHED */
886 }
887 }
888
889 info = &DiskInfo[DiskNum];
890 l = strlen(fname);
891
892 info->unit = i;
893 info->fd = fd;
894 info->type = type;
895 info->flags = diskFlags[i];
896 memcpy(info->fname, fname, l);
897 info->serno = NULL;
898 if (serno) {
899 if ((info->serno = malloc(SERNOLEN)) != NULL)
900 strlcpy(info->serno, serno, SERNOLEN);
901 else
902 warnx("Couldn't allocate memory for the operation");
903 }
904
905 if (DiskNum == 0) {
906 if (type == VKD_CD) {
907 rootdevnames[0] = "cd9660:vcd0";
908 } else if (type == VKD_DISK) {
909 rootdevnames[0] = "ufs:vkd0s0a";
910 rootdevnames[1] = "ufs:vkd0s1a";
911 }
912 }
913
914 DiskNum++;
915 } else {
916 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
917 continue;
918 }
919 }
920 }
921
922 static
923 int
netif_set_tapflags(int tap_unit,int f,int s)924 netif_set_tapflags(int tap_unit, int f, int s)
925 {
926 struct ifreq ifr;
927 int flags;
928
929 bzero(&ifr, sizeof(ifr));
930
931 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
932 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
933 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
934 return -1;
935 }
936
937 /*
938 * Adjust if_flags
939 *
940 * If the flags are already set/cleared, then we return
941 * immediately to avoid extra syscalls
942 */
943 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
944 if (f < 0) {
945 /* Turn off flags */
946 f = -f;
947 if ((flags & f) == 0)
948 return 0;
949 flags &= ~f;
950 } else {
951 /* Turn on flags */
952 if (flags & f)
953 return 0;
954 flags |= f;
955 }
956
957 /*
958 * Fix up ifreq.ifr_name, since it may be trashed
959 * in previous ioctl(SIOCGIFFLAGS)
960 */
961 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
962
963 ifr.ifr_flags = flags & 0xffff;
964 ifr.ifr_flagshigh = flags >> 16;
965 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
966 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
967 return -1;
968 }
969 return 0;
970 }
971
972 static
973 int
netif_set_tapaddr(int tap_unit,in_addr_t addr,in_addr_t mask,int s)974 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
975 {
976 struct ifaliasreq ifra;
977 struct sockaddr_in *in;
978
979 bzero(&ifra, sizeof(ifra));
980 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
981
982 /* Setup address */
983 in = (struct sockaddr_in *)&ifra.ifra_addr;
984 in->sin_family = AF_INET;
985 in->sin_len = sizeof(*in);
986 in->sin_addr.s_addr = addr;
987
988 if (mask != 0) {
989 /* Setup netmask */
990 in = (struct sockaddr_in *)&ifra.ifra_mask;
991 in->sin_len = sizeof(*in);
992 in->sin_addr.s_addr = mask;
993 }
994
995 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
996 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
997 return -1;
998 }
999 return 0;
1000 }
1001
1002 static
1003 int
netif_add_tap2brg(int tap_unit,const char * ifbridge,int s)1004 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
1005 {
1006 struct ifbreq ifbr;
1007 struct ifdrv ifd;
1008
1009 bzero(&ifbr, sizeof(ifbr));
1010 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
1011 "tap%d", tap_unit);
1012
1013 bzero(&ifd, sizeof(ifd));
1014 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
1015 ifd.ifd_cmd = BRDGADD;
1016 ifd.ifd_len = sizeof(ifbr);
1017 ifd.ifd_data = &ifbr;
1018
1019 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
1020 /*
1021 * 'errno == EEXIST' means that the tap(4) is already
1022 * a member of the bridge(4)
1023 */
1024 if (errno != EEXIST) {
1025 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
1026 return -1;
1027 }
1028 }
1029 return 0;
1030 }
1031
1032 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
1033
1034 /*
1035 * Locate the first unused tap(4) device file if auto mode is requested,
1036 * or open the user supplied device file, and bring up the corresponding
1037 * tap(4) interface.
1038 *
1039 * NOTE: Only tap(4) device file is supported currently
1040 */
1041 static
1042 int
netif_open_tap(const char * netif,int * tap_unit,int s)1043 netif_open_tap(const char *netif, int *tap_unit, int s)
1044 {
1045 char tap_dev[MAXPATHLEN];
1046 int tap_fd, failed;
1047 struct stat st;
1048 char *dname;
1049
1050 *tap_unit = -1;
1051
1052 if (strcmp(netif, "auto") == 0) {
1053 /*
1054 * Find first unused tap(4) device file
1055 */
1056 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
1057 if (tap_fd < 0) {
1058 warnc(errno, "Unable to find a free tap(4)");
1059 return -1;
1060 }
1061 } else {
1062 /*
1063 * User supplied tap(4) device file or unix socket.
1064 */
1065 if (netif[0] == '/') /* Absolute path */
1066 strlcpy(tap_dev, netif, sizeof(tap_dev));
1067 else
1068 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
1069
1070 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
1071
1072 /*
1073 * If we cannot open normally try to connect to it.
1074 */
1075 if (tap_fd < 0)
1076 tap_fd = unix_connect(tap_dev);
1077
1078 if (tap_fd < 0) {
1079 warn("Unable to open %s", tap_dev);
1080 return -1;
1081 }
1082 }
1083
1084 /*
1085 * Check whether the device file is a tap(4)
1086 */
1087 if (fstat(tap_fd, &st) < 0) {
1088 failed = 1;
1089 } else if (S_ISCHR(st.st_mode)) {
1090 dname = fdevname(tap_fd);
1091 if (dname)
1092 dname = strstr(dname, "tap");
1093 if (dname) {
1094 /*
1095 * Bring up the corresponding tap(4) interface
1096 */
1097 *tap_unit = strtol(dname + 3, NULL, 10);
1098 printf("TAP UNIT %d\n", *tap_unit);
1099 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
1100 failed = 0;
1101 else
1102 failed = 1;
1103 } else {
1104 failed = 1;
1105 }
1106 } else if (S_ISSOCK(st.st_mode)) {
1107 /*
1108 * Special socket connection (typically to vknet). We
1109 * do not have to do anything.
1110 */
1111 failed = 0;
1112 } else {
1113 failed = 1;
1114 }
1115
1116 if (failed) {
1117 warnx("%s is not a tap(4) device or socket", tap_dev);
1118 close(tap_fd);
1119 tap_fd = -1;
1120 *tap_unit = -1;
1121 }
1122 return tap_fd;
1123 }
1124
1125 static int
unix_connect(const char * path)1126 unix_connect(const char *path)
1127 {
1128 struct sockaddr_un sunx;
1129 int len;
1130 int net_fd;
1131 int sndbuf = 262144;
1132 struct stat st;
1133
1134 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1135 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1136 ++len; /* include nul */
1137 sunx.sun_family = AF_UNIX;
1138 sunx.sun_len = len;
1139
1140 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1141 if (net_fd < 0)
1142 return(-1);
1143 if (connect(net_fd, (void *)&sunx, len) < 0) {
1144 close(net_fd);
1145 return(-1);
1146 }
1147 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1148 if (fstat(net_fd, &st) == 0)
1149 printf("Network socket buffer: %ld bytes\n", st.st_blksize);
1150 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1151 return(net_fd);
1152 }
1153
1154 #undef TAPDEV_MAJOR
1155 #undef TAPDEV_MINOR
1156 #undef TAPDEV_OFLAGS
1157
1158 /*
1159 * Following syntax is supported,
1160 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1161 *
1162 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1163 * tap(4)'s netmask len is z
1164 *
1165 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1166 * pseudo netif's address is y.y.y.y
1167 *
1168 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1169 * pseudo netif's address is y.y.y.y
1170 * tap(4) and pseudo netif's netmask len are z
1171 *
1172 * 5) bridgeX tap(4) will be added to bridgeX
1173 *
1174 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1175 * pseudo netif's address is y.y.y.y
1176 *
1177 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1178 * pseudo netif's address is y.y.y.y
1179 * pseudo netif's netmask len is z
1180 */
1181 static
1182 int
netif_init_tap(int tap_unit,in_addr_t * addr,in_addr_t * mask,int s)1183 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1184 {
1185 in_addr_t tap_addr, netmask, netif_addr;
1186 int next_netif_addr;
1187 char *tok, *masklen_str, *ifbridge;
1188
1189 *addr = 0;
1190 *mask = 0;
1191
1192 tok = strtok(NULL, ":/");
1193 if (tok == NULL) {
1194 /*
1195 * Nothing special, simply use tap(4) as backend
1196 */
1197 return 0;
1198 }
1199
1200 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1201 /*
1202 * tap(4)'s address is supplied
1203 */
1204 ifbridge = NULL;
1205
1206 /*
1207 * If there is next token, then it may be pseudo
1208 * netif's address or netmask len for tap(4)
1209 */
1210 next_netif_addr = 0;
1211 } else {
1212 /*
1213 * Not tap(4)'s address, assume it as a bridge(4)
1214 * iface name
1215 */
1216 tap_addr = 0;
1217 ifbridge = tok;
1218
1219 /*
1220 * If there is next token, then it must be pseudo
1221 * netif's address
1222 */
1223 next_netif_addr = 1;
1224 }
1225
1226 netmask = netif_addr = 0;
1227
1228 tok = strtok(NULL, ":/");
1229 if (tok == NULL)
1230 goto back;
1231
1232 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1233 if (next_netif_addr) {
1234 warnx("Invalid pseudo netif address: %s", tok);
1235 return -1;
1236 }
1237 netif_addr = 0;
1238
1239 /*
1240 * Current token is not address, then it must be netmask len
1241 */
1242 masklen_str = tok;
1243 } else {
1244 /*
1245 * Current token is pseudo netif address, if there is next token
1246 * it must be netmask len
1247 */
1248 masklen_str = strtok(NULL, "/");
1249 }
1250
1251 /* Calculate netmask */
1252 if (masklen_str != NULL) {
1253 u_long masklen;
1254
1255 masklen = strtoul(masklen_str, NULL, 10);
1256 if (masklen < 32 && masklen > 0) {
1257 netmask =
1258 htonl(rounddown2(0xffffffff, 1LL << (32 - masklen)));
1259 } else {
1260 warnx("Invalid netmask len: %lu", masklen);
1261 return -1;
1262 }
1263 }
1264
1265 /* Make sure there is no more token left */
1266 if (strtok(NULL, ":/") != NULL) {
1267 warnx("Invalid argument to '-I'");
1268 return -1;
1269 }
1270
1271 back:
1272 if (tap_unit < 0) {
1273 /* Do nothing */
1274 } else if (ifbridge == NULL) {
1275 /* Set tap(4) address/netmask */
1276 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1277 return -1;
1278 } else {
1279 /* Tie tap(4) to bridge(4) */
1280 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1281 return -1;
1282 }
1283
1284 *addr = netif_addr;
1285 *mask = netmask;
1286 return 0;
1287 }
1288
1289 /*
1290 * NetifInfo[] will be filled for pseudo netif initialization.
1291 * NetifNum will be bumped to reflect the number of valid entries
1292 * in NetifInfo[].
1293 */
1294 static
1295 void
init_netif(char * netifExp[],int netifExpNum)1296 init_netif(char *netifExp[], int netifExpNum)
1297 {
1298 int i, s;
1299 char *tmp;
1300
1301 if (netifExpNum == 0)
1302 return;
1303
1304 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1305 if (s < 0)
1306 return;
1307
1308 for (i = 0; i < netifExpNum; ++i) {
1309 struct vknetif_info *info;
1310 in_addr_t netif_addr, netif_mask;
1311 int tap_fd, tap_unit;
1312 char *netif;
1313
1314 /* Extract MAC address if there is one */
1315 tmp = netifExp[i];
1316 strsep(&tmp, "=");
1317
1318 netif = strtok(netifExp[i], ":");
1319 if (netif == NULL) {
1320 warnx("Invalid argument to '-I'");
1321 continue;
1322 }
1323
1324 /*
1325 * Open tap(4) device file and bring up the
1326 * corresponding interface
1327 */
1328 tap_fd = netif_open_tap(netif, &tap_unit, s);
1329 if (tap_fd < 0)
1330 continue;
1331
1332 /*
1333 * Initialize tap(4) and get address/netmask
1334 * for pseudo netif
1335 *
1336 * NB: Rest part of netifExp[i] is passed
1337 * to netif_init_tap() implicitly.
1338 */
1339 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1340 /*
1341 * NB: Closing tap(4) device file will bring
1342 * down the corresponding interface
1343 */
1344 close(tap_fd);
1345 continue;
1346 }
1347
1348 info = &NetifInfo[NetifNum];
1349 bzero(info, sizeof(*info));
1350 info->tap_fd = tap_fd;
1351 info->tap_unit = tap_unit;
1352 info->netif_addr = netif_addr;
1353 info->netif_mask = netif_mask;
1354 /*
1355 * If tmp isn't NULL it means a MAC could have been
1356 * specified so attempt to convert it.
1357 * Setting enaddr to NULL will tell vke_attach() we
1358 * need a pseudo-random MAC address.
1359 */
1360 if (tmp != NULL) {
1361 if ((info->enaddr = malloc(ETHER_ADDR_LEN)) == NULL)
1362 warnx("Couldn't allocate memory for the operation");
1363 else {
1364 if ((kether_aton(tmp, info->enaddr)) == NULL) {
1365 free(info->enaddr);
1366 info->enaddr = NULL;
1367 }
1368 }
1369 }
1370
1371 NetifNum++;
1372 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1373 break;
1374 }
1375 close(s);
1376 }
1377
1378 /*
1379 * Create the pid file and leave it open and locked while the vkernel is
1380 * running. This allows a script to use /usr/bin/lockf to probe whether
1381 * a vkernel is still running (so as not to accidently kill an unrelated
1382 * process from a stale pid file).
1383 */
1384 static
1385 void
writepid(void)1386 writepid(void)
1387 {
1388 char buf[32];
1389 int fd;
1390
1391 if (pid_file != NULL) {
1392 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1393 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1394 if (fd < 0) {
1395 if (errno == EWOULDBLOCK) {
1396 perror("Failed to lock pidfile, "
1397 "vkernel already running");
1398 } else {
1399 perror("Failed to create pidfile");
1400 }
1401 exit(EX_SOFTWARE);
1402 }
1403 ftruncate(fd, 0);
1404 write(fd, buf, strlen(buf));
1405 /* leave the file open to maintain the lock */
1406 }
1407 }
1408
1409 static
1410 void
cleanpid(void)1411 cleanpid( void )
1412 {
1413 if (pid_file != NULL) {
1414 if (unlink(pid_file) < 0)
1415 perror("Warning: couldn't remove pidfile");
1416 }
1417 }
1418
1419 static
1420 void
usage_err(const char * ctl,...)1421 usage_err(const char *ctl, ...)
1422 {
1423 va_list va;
1424
1425 va_start(va, ctl);
1426 vfprintf(stderr, ctl, va);
1427 va_end(va);
1428 fprintf(stderr, "\n");
1429 exit(EX_USAGE);
1430 }
1431
1432 static
1433 void
usage_help(_Bool help)1434 usage_help(_Bool help)
1435 {
1436 fprintf(stderr, "Usage: %s [-hsUvdt] [-c file] [-e name=value:name=value:...]\n"
1437 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
1438 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n"
1439 "\t[-p file] [-r file]\n", save_av[0]);
1440
1441 if (help)
1442 fprintf(stderr, "\nArguments:\n"
1443 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1444 "\t-e\tSpecify an environment to be used by the kernel.\n"
1445 "\t-h\tThis list of options.\n"
1446 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1447 "\t-I\tCreate a virtual network device.\n"
1448 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1449 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
1450 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n"
1451 "\t\t\tnumcpus - number of cpus\n"
1452 "\t\t\tlbits - specify the number of bits within APICID(=CPUID)\n"
1453 "\t\t\t needed for representing the logical ID.\n"
1454 "\t\t\t Controls the number of threads/core:\n"
1455 "\t\t\t (0 bits - 1 thread, 1 bit - 2 threads).\n"
1456 "\t\t\tcbits - specify the number of bits within APICID(=CPUID)\n"
1457 "\t\t\t needed for representing the core ID.\n"
1458 "\t\t\t Controls the number of cores/package:\n"
1459 "\t\t\t (0 bits - 1 core, 1 bit - 2 cores).\n"
1460 "\t-p\tSpecify a file in which to store the process ID.\n"
1461 "\t-r\tSpecify a R/W disk image file, iterates vkd0..n\n"
1462 "\t-R\tSpecify a COW disk image file, iterates vkd0..n\n"
1463 "\t-s\tBoot into single-user mode.\n"
1464 "\t-t\tUse a precise host timer when calculating clock values.\n"
1465 "\t-U\tEnable writing to kernel memory and module loading.\n"
1466 "\t-v\tTurn on verbose booting.\n");
1467
1468 exit(EX_USAGE);
1469 }
1470
1471 void
cpu_smp_stopped(void)1472 cpu_smp_stopped(void)
1473 {
1474 }
1475
1476 void
cpu_reset(void)1477 cpu_reset(void)
1478 {
1479 kprintf("cpu reset, rebooting vkernel\n");
1480 closefrom(3);
1481 cleanpid();
1482 exit(EX_VKERNEL_REBOOT);
1483 }
1484
1485 void
cpu_halt(void)1486 cpu_halt(void)
1487 {
1488 kprintf("cpu halt, exiting vkernel\n");
1489 cleanpid();
1490 exit(EX_OK);
1491 }
1492
1493 void
setrealcpu(void)1494 setrealcpu(void)
1495 {
1496 switch(lwp_cpu_lock) {
1497 case LCL_PER_CPU:
1498 if (bootverbose)
1499 kprintf("Locking CPU%d to real cpu %d\n",
1500 mycpuid, next_cpu);
1501 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1502 next_cpu++;
1503 if (next_cpu >= real_ncpus)
1504 next_cpu = 0;
1505 break;
1506 case LCL_SINGLE_CPU:
1507 if (bootverbose)
1508 kprintf("Locking CPU%d to real cpu %d\n",
1509 mycpuid, next_cpu);
1510 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1511 break;
1512 default:
1513 /* do not map virtual cpus to real cpus */
1514 break;
1515 }
1516 }
1517
1518 /*
1519 * Allocate and free memory for module loading. The loaded module
1520 * has to be placed somewhere near the current kernel binary load
1521 * point or the relocations will not work.
1522 *
1523 * I'm not sure why this isn't working.
1524 */
1525 int
vkernel_module_memory_alloc(vm_offset_t * basep,size_t bytes)1526 vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes)
1527 {
1528 #if 1
1529 size_t xtra;
1530 xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK;
1531 *basep = (vm_offset_t)sbrk(xtra + bytes) + xtra;
1532 bzero((void *)*basep, bytes);
1533 #else
1534 *basep = (vm_offset_t)mmap((void *)0x000000000, bytes,
1535 PROT_READ|PROT_WRITE|PROT_EXEC,
1536 MAP_ANON|MAP_SHARED, -1, 0);
1537 if ((void *)*basep == MAP_FAILED)
1538 return ENOMEM;
1539 #endif
1540 return 0;
1541 }
1542
1543 void
vkernel_module_memory_free(vm_offset_t base,size_t bytes)1544 vkernel_module_memory_free(vm_offset_t base, size_t bytes)
1545 {
1546 #if 0
1547 #if 0
1548 munmap((void *)base, bytes);
1549 #endif
1550 #endif
1551 }
1552
1553 /*
1554 * VKERNEL64 implementation functions using ptrheads.
1555 */
1556 void
vkernel_yield(void)1557 vkernel_yield(void)
1558 {
1559 pthread_yield();
1560 }
1561